16.08.2012 Views

Proceedings of the Junior Scientist Conference 2008

Proceedings of the Junior Scientist Conference 2008

Proceedings of the Junior Scientist Conference 2008

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong><br />

<strong>Junior</strong> <strong>Scientist</strong> <strong>Conference</strong> <strong>2008</strong>


<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong><br />

<strong>Junior</strong> <strong>Scientist</strong> <strong>Conference</strong> <strong>2008</strong><br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

November <strong>2008</strong><br />

Edited by<br />

Hans K. Kaiser and Raimund Kirner


The views and opinions expressed in all <strong>the</strong> papers <strong>of</strong> this book are <strong>the</strong> author’s personal<br />

ones.<br />

The copyright <strong>of</strong> <strong>the</strong> individual papers belongs to <strong>the</strong> authors. Copies cannot be reproduced<br />

for commercial pr<strong>of</strong>it.<br />

Cover page design: Bettina Neunteufl<br />

c○<strong>2008</strong>, Copyright protected.<br />

ISBN 978-3-200-01612-5<br />

Printed in Austria.


Contents<br />

Welcome Message from Peter Skalicky xv<br />

Welcome Message from Hans K. Kaiser xvii<br />

<strong>Conference</strong> Organizers xix<br />

<strong>Conference</strong> Program xxi<br />

1 Information and Communication Technology 1<br />

Implementation <strong>of</strong> Quadrature Modulator Imbalance Compensation Method,<br />

Karel Povalac, Roman Marsálek ...................... 3<br />

Diagnostical analysis <strong>of</strong> voice,<br />

Pavel Sala, Milan Sigmund ......................... 5<br />

Realization <strong>of</strong> <strong>the</strong> OFDM coder and decoder using DSP,<br />

Petr Zelinka, Milan Sigmund ........................ 7<br />

Transmembrane Potential Measurements on Plant Cells using <strong>the</strong> Voltagesensitive<br />

Dye ANNINE-6,<br />

Bianca Flickinger, Thomas Berghöfer, Christian Eing, Wolfgang Frey ... 9<br />

Modeling metamaterial microwave resonators,<br />

Tomas Zvolensky, Zbynek Raida ...................... 11<br />

Bearingless Segment Motor with Buried Magnets,<br />

Thomas Stallinger, Wolfgang Gruber, Wolfgang Amrhein ......... 13<br />

Identification <strong>of</strong> tacit knowledge during realization <strong>of</strong> knowledge creation processes,<br />

Frantisek Babic, Ján Paralic ........................ 15<br />

i


Multifunctional tuned 4th order active filter,<br />

Roman Sotner, Tomas Dostal ........................ 17<br />

Active Semiconductor Devices for Millimetre Waves Applications,<br />

Michal Pokorný, Zbynek Raida ....................... 19<br />

Analysis <strong>of</strong> Power-Bus Structures Using Contour Integral Method,<br />

Martin Stumpf, Zbynek Raida ........................ 21<br />

Creating and Drawing Fonts for Graphic displays with Paged Memory Model,<br />

Filip Adamec, Tomas Fryza ......................... 23<br />

Bandwidth Comparison <strong>of</strong> Microwave Planar Hybrids,<br />

Petr Orság, pr<strong>of</strong>. Ing. Jirí Svacina, CSc. .................. 25<br />

Six prot microwave vector analyzer,<br />

Radek Dvorák, Ing. Tomás Urbanec, Ph.D. ................ 27<br />

Decomposing HEX-Programs: Preliminary Results,<br />

Thomas Krennwallner, Thomas Eiter, Michael Fink ............ 29<br />

Bio-inspired Implementation <strong>of</strong> Linear Vestibulo-Ocular and Opto-Kinetic Reflexes<br />

in a Humanoid Robot,<br />

Igor Labutov, Jizhong Xiao, Theodore Raphan ............... 31<br />

Simulation <strong>of</strong> light environment and its application in practice,<br />

Anna Beneova, Marek Krupa ........................ 33<br />

Towards Synchronization in Multiple Antenna Systems,<br />

Qi Wang, Markus Rupp ........................... 35<br />

Technical criteria for <strong>the</strong> comparison <strong>of</strong> modern ERP system for <strong>the</strong> usage in<br />

orchestra companies at <strong>the</strong> case study <strong>of</strong> Dynamics AX 2009,<br />

Michael Gall, Christian Sterba ....................... 37<br />

Development Of A Quantitative Approach That Uses Fourier Transform Infrared<br />

Spectroscopy (FTIR) To Measure Ambient Aerosol Composition,<br />

Vilma Arriaran, Beth Wittig (Mentor) .................... 39<br />

Optimizing <strong>the</strong> Python interpreter: Identifying performance bottlenecks.,<br />

Stefan Brunthaler, Jens Knoop, n/a ..................... 41<br />

Model for Comparison <strong>of</strong> Types <strong>of</strong> Traffic Intersection Control Systems,<br />

Jan Krcal, Michal Jerabek ......................... 43<br />

An integrated source-to-source approach for WCET analysis,<br />

Adrian Prantl, Jens Knoop ......................... 45<br />

Cooperative Media Access Control in Wireless Networks,<br />

Mathias Offner, Helmut Adam ....................... 47<br />

ii


Jessine: An Approach to Creating S<strong>of</strong>tware Applications with Rule-Based Domain<br />

Knowledge,<br />

Miroslav Sabo ................................ 49<br />

Coordination <strong>of</strong> Agent-based Services,<br />

Lubomir Wassermann ............................ 51<br />

Design <strong>of</strong> Self-organizing Systems Using Evolutionary Methods,<br />

István Fehérvári, Wilfried Elmenreich ................... 53<br />

Towards Automatic Design <strong>of</strong> Competitive Image Filters in FPGAs,<br />

Zdenek Vasícek ............................... 55<br />

Evaluation <strong>of</strong> a State-based Process Management Platform,<br />

Ekaterina Andrianova, Jürgen Dorn .................... 57<br />

Motor task-induced modifications <strong>of</strong> multiple spinal cord reflexes,<br />

Ursula H<strong>of</strong>stoetter, Josef Ladenbauer, Karen Minassian, Milan Dimitrijevic,<br />

Frank Rattay ............................. 59<br />

Approach for a Reliable Cooperative Relaying Process,<br />

Melanie Schranz, Wilfried Elmenreich ................... 61<br />

Retrieval <strong>of</strong> Ocean Water Parameters from Satellite Imagery and <strong>the</strong>ir Dependence<br />

on Atmospheric Correction Models,<br />

Rushane Dyer, Sam Ahmed ......................... 63<br />

Fast Packet Classification Algorithm in Hardware,<br />

Viktor Pus .................................. 65<br />

Radiation <strong>of</strong> Antenna Array in Near-field Area,<br />

Jirí Hermany, Zdenek Novácek ....................... 67<br />

Ontology- and Bayesian-based Threat Probability Determination,<br />

Stefan Fenz, A Min Tjoa ........................... 69<br />

Development <strong>of</strong> Multi-Core Video Decoding Platforms based on High-Level<br />

Architecture Simulations,<br />

Florian Seitner, Michael Bleyer, Margrit Gelautz ............. 71<br />

Comparison between Solute transport parameters estimation from local to field<br />

scale,<br />

Mohamed Kassab, Antonio Coppola, Nicola Lamaddalena, Alberto<br />

Guadagini, Monica Riva .......................... 73<br />

A new method for identification <strong>of</strong> systems <strong>of</strong> arbitrary real order based on solution<br />

<strong>of</strong> fractional differential equations and orthogonal distance fitting,<br />

Tomas Skovranek, Vladimir Despotovic, Igor Podlubny .......... 75<br />

iii


Configurable Meta-search in Human Resource Domain,<br />

Tabbasum Naz, Jürgen Dorn ........................ 77<br />

The Definition <strong>of</strong> Secure Business Processes with Respect to Multiple Objectives,<br />

Johannes Heurix, Thomas Neubauer .................... 79<br />

Generalising Constraint Solving over Finite Domains,<br />

Markus Triska, Nysret Musliu ........................ 81<br />

Default Reasoning on Top <strong>of</strong> Ontologies with dl-Programs,<br />

Minh DAO Tran, Thomas Eiter ....................... 83<br />

Implementation <strong>of</strong> a Design Tool for Automated Generation <strong>of</strong> Four State<br />

Logic Circuits.,<br />

Jakob Lechner, Martin Delvai ........................ 85<br />

Analyzing Moving Objects Behaviour and Predicting <strong>the</strong>ir Future Location,<br />

Ivana Nizetic, Kresimir Fertalj ....................... 87<br />

Efficient Counting with Bounded Treewidth using Datalog,<br />

Stefan Rümmele, Reinhard Pichler ..................... 89<br />

MapFace - A Graphical Editor to Support <strong>the</strong> Semantic Annotation <strong>of</strong> Medical<br />

Text,<br />

Theresia Gschwandtner, Katharina Kaiser, Silvia Miksch ......... 91<br />

Optimization <strong>of</strong> communication by analyzing <strong>the</strong> interlocutors wording,<br />

Gudrun Kellner ............................... 93<br />

A Simulation Environment for Distributed Real-Time Systems in <strong>the</strong> Presence<br />

<strong>of</strong> Malicious Attacks,<br />

Ekarin Suethanuwong, Christian El-Salloum ................ 95<br />

Filtering <strong>of</strong> Phase Shift Terrestrial Laser Scanner Point Clouds,<br />

Clemens No<strong>the</strong>gger, Norbert Pfeifer .................... 97<br />

When and why to use <strong>the</strong> evolutionary algorithms,<br />

Jana Jilková, Zbynek Raida ......................... 99<br />

Hybrid Timing Analysis for ANSI-C Applications with Loops and Function<br />

Calls,<br />

Bernhard Rieder, Peter Puschner ......................101<br />

Deterministic Finite Automaton with Perfect Hashing for Fast Pattern Matching,<br />

Jan Kastil, Jan Korenek ...........................103<br />

Network Coding for Cooperative Communications,<br />

Gordhan Das Menghwar, Christoph F Mecklenbrauker ..........105<br />

iv


Graphics versus spoken language in pedestrian navigation,<br />

Felix Ortag, Georg Gartner .........................107<br />

Hardware Acceleration <strong>of</strong> Protocol Identification,<br />

Petr Kobierský ................................109<br />

Evolutionary Design <strong>of</strong> Wormhole Switched Collective Communications,<br />

Jiri Jaros, Josef Schwarz ..........................111<br />

Component Based Communication Middleware for AUTOSAR,<br />

Dietmar Schreiner, Karl M. Göschka, Jens Knoop .............113<br />

The Acquaintance <strong>of</strong> Hardware Timing Effects: A Sine Qua Non to Validate<br />

Temporal Requirements in Embedded Real Time Systems,<br />

Sven Bünte, Raimund Kirner ........................115<br />

Divide and Measure: CFG Segmentation for <strong>the</strong> Measurement-Based Analysis<br />

<strong>of</strong> Resource Consumption,<br />

Michael Zolda, Raimund Kirner ......................117<br />

Neutralizing Timing Anomalies in Complex Computer Architectures,<br />

Albrecht Kadlec, Raimund Kirner .....................119<br />

Optical Imaging Hidden Objects in Turbid Scattering Media using Principal<br />

Component Analysis,<br />

Binlin Wu, Swapan Gayen, Robert Alfano .................121<br />

Links between Diachronic Thinking and Temporal Cognition Tasks Observed<br />

in Middle Childhood,<br />

Brandy Moore, Patricia Brooks, Laura Rabin, Carmen Carrion ......123<br />

Generalizing Sudoku to Three Dimensions,<br />

Tiffany Lambert, Paula Whitlock ......................125<br />

2 Material Sciences 127<br />

Nonadditivity <strong>of</strong> Quantum Phases for mixed States,<br />

Jürgen Klepp, Stephan Sponar, Stefan Filipp, Matthias Lettner, Gerald<br />

Badurek, Yuji Hasegawa ..........................129<br />

Complexation <strong>of</strong> trivalent Cm and Eu with extraction ligands for <strong>the</strong> Partitioning<br />

& Transmutation strategy,<br />

Sascha Trumm, Petra J. Panak, Andreas Geist, Reinhardt Klenze, Thomas<br />

Fanghänel ..................................131<br />

Complexation <strong>of</strong> Cm(III) with Nitrate and Fluoride at Elevated Temperatures<br />

Studied by Time Resolved Laser Fluorescence Spectroscopy,<br />

Andrej Skerencak, Petra J. Panak, Volker Neck, Reinhardt Klenze, Thomas<br />

Fanghänel ..................................133<br />

v


Development <strong>of</strong> alpha-(Al,Cr)2O3 solid solutions by reactive magnetron sputtering,<br />

Dominic Diechle, Michael Stueber, Harald Leiste, Sven Ulrich ......135<br />

Modification <strong>of</strong> Aluminum Alkoxides with o Esters and Dialkylmalonates as<br />

Precursors for Sol-Gel Derived Hybrid Materials,<br />

Robert Lichtenberger, Stefan O. Baumann, Michael Puchberger, Ulrich<br />

Schubert ...................................137<br />

Structural Aspects <strong>of</strong> Oximate-Modified Titanium Alkoxides,<br />

Stefan O. Baumann, Maria Bendova, Robert Potzmann, Michael Puchberger,<br />

Ulrich Schubert ...........................139<br />

Trace Metal Speciation in Water using a Solid Phase Extraction Flow Injection<br />

Setup online coupled to ICP-AES,<br />

Christoph Puls, Andreas Limbeck .....................141<br />

Analysis <strong>of</strong> Snowpack Properties And Estimation <strong>of</strong> Snow Grain Size Using<br />

Cold Land Processes Field Experiment Data,<br />

Dugwon Seo, Amir Azar, Al Powell, Reza Khanbilvardi ..........143<br />

Supersensitive tin oxide nanosensors for gas detection,<br />

Alexandra Tischner, Anton Köck, Thomas Maier, Christian Edtmaier,<br />

Christian Gspan, Gerhard Kothleitner ...................145<br />

Distortion Investigation during Sintering <strong>of</strong> Copper PIM Parts,<br />

Ijaz Ul Mohsin, Christian Gierl, Herbert Danninger ............147<br />

Electron microscopy <strong>of</strong> LiAlO_2 - a promising substrate for opto-electronic<br />

devices,<br />

Walid Hetaba ................................149<br />

Measuring <strong>of</strong> <strong>the</strong> Vibrations <strong>of</strong> <strong>the</strong> Petrol Engine,<br />

Jan Diblík ..................................151<br />

Silver plated tungsten carbide powders for electrical contact materials with<br />

improved homogeneity,<br />

Robert Christian Hula, Christian Edtmaier ................153<br />

Investigation <strong>of</strong> mechanical properties <strong>of</strong> polymer-derived Si-C-N/Al2O3composites<br />

with low shrinkage,<br />

Thomas Konegger, Antje Liersch ......................155<br />

Preparation <strong>of</strong> Alkyne and Azide modified Spherosilicates,<br />

Angelika Bachinger, Guido Kickelbick ...................157<br />

vi


Learning more about <strong>the</strong> Gene Regulation <strong>of</strong> Hydrolytic enzyme-encoding<br />

Genes in an industrial relevant Fungus,<br />

Matthias G. Steiger, Marion E. Pucher, Astrid R. Mach-Aigner, Robert L.<br />

Mach ....................................159<br />

Investigations <strong>of</strong> Xyr1 (Xylanase regulator 1), <strong>the</strong> main transcriptional regulator<br />

<strong>of</strong> <strong>the</strong> xylanolytic and cellulolytic enzymes in Hypocrea jecorina<br />

(Trichoderma reesei),<br />

Marion E. Pucher, Gudrun E. Bauer, Matthias G. Steiger, Robert L. Mach,<br />

Astrid R. Mach-Aigner ...........................161<br />

Using Confocal Microscopy to Study Mechanotransduction in Renal Epi<strong>the</strong>lial<br />

Cells,<br />

Yi Duan, Sheldon Weinbaum ........................163<br />

Low-Energy School in Indian Himalayas,<br />

Jan Tilinger .................................165<br />

Algorithm <strong>of</strong> Simulation and Optimization <strong>of</strong> Coventional Mechanical Production,<br />

Robert Jurcisin, Dusan Sebo ........................167<br />

A Computer-Controlled Joint Loading System for Quantitative Analysis <strong>of</strong><br />

Mechanoresponsiveness in Articular Cartilage,<br />

Xiang Gu, Daniel Leong, YongHui Li, Francisco Guzman, HuiBin Sun,<br />

Luis Cardoso ................................169<br />

Induced Blood Brain Barrier Electroporation during Deep Brain Stimulation:<br />

In vitro endo<strong>the</strong>lial monolayer model,<br />

Veronica Lopez, Abhisheck Datta, Ronny Amaya, John Tarbell, Marom<br />

Bikson ....................................171<br />

Colloidal gel formation dynamics: light scattering, rheometry and simulation,<br />

Xiujuan Cao, Herman Cummins, Jeffrey Morris (Mentor) .........173<br />

Syn<strong>the</strong>sis Of New Isothiazole Derivatives As Potential Agents For Crop Protection,<br />

Radoslav Flasik, Peter Stanetty .......................175<br />

Site-Selective Time Resolved Laser Fluorescence Spectroscopy on Ca2+-<br />

Bearing Mineral Phases Doped with Europium,<br />

Moritz Schmidt, Thorsten Stumpf, Clemens Wal<strong>the</strong>r, Thomas Fanghänel,<br />

Horst Geckeis ................................177<br />

Nan<strong>of</strong>eatured carbon-copper metal matrix composites,<br />

Michael Kitzmantel, Christoph Eisenmenger-Sittner, Dan Cunningham . . 179<br />

vii


Substitution <strong>of</strong> fossil fuels by using low temperature pyrolysis <strong>of</strong> agricultural<br />

residues in a <strong>the</strong>rmal power plant,<br />

Michael Halwachs .............................181<br />

Long Wave Instabilities in Periodic Structures,<br />

Melanie Todt, Thomas Daxner .......................183<br />

Determination <strong>of</strong> <strong>the</strong> stoichiometry <strong>of</strong> complexes <strong>of</strong> trivalent lanthanides<br />

and extraction-relevant ligands with nanoelectrospray ionization time-<strong>of</strong>flight<br />

mass-spectrometry,<br />

Michael Steppert, Clemens Wal<strong>the</strong>r, Andreas Geist, Horst Geckeis, Thommas<br />

Fanghänel ...............................185<br />

Quantum Optical Coherence Tomography,<br />

Pierre Bouzi, Michael Etienne, Andru Prescod, Roger Dorsinville ....187<br />

Studies on application strategy <strong>of</strong> salts and chitosan for controlling Penicillium<br />

rot <strong>of</strong> Hernandina clementine,<br />

Youssef Khamis, Ligorio Angela, Pentimone Isabella, Sanzani Semona,<br />

DOnghia Anna Maria, Nigro Franco, Ippolito Antonio ..........189<br />

Functionalized Cadmium Sulfide Clusters as Nanosized Building Blocks for<br />

Hybrid Materials - Investigation <strong>of</strong> <strong>the</strong> "[Cd10S4(SPh)12]" Cluster Structure,<br />

Maria Bendova, Michael Puchberger, Ulrich Schubert ..........191<br />

Creating building blocks for Hybrid Materials: Investigation <strong>of</strong> Adhesives for<br />

Zirconia Substrates,<br />

Christoph Lomoschitz, Guido Kickelbick ..................193<br />

Model Structures for <strong>the</strong> Mycotoxin Zearalenone as Mimics in Metabolite Syn<strong>the</strong>sis,<br />

Hannes Mikula, Hans-Peter Karl, Bernhard Klösch, Christian Hametner,<br />

Johannes Fröhlich ..............................195<br />

Realization <strong>of</strong> Inversion-type GaN MOSFETs with Ar Implantation for Device<br />

Isolation,<br />

Clemens Ostermaier, Sang-Il Ahn, Kay Potzger, Manfred Helm, Stefan<br />

Kalchmair, Dionyz Pogany, Jong-Ho Lee, Sung-Ho Hahm, Jung-Hee Lee . 197<br />

Solution Stability <strong>of</strong> Trinuclear Iron Oxo Clusters,<br />

Robert Potzmann, Stefan O. Baumann, Maria Bendova, Michael Puchberger,<br />

Ulrich Schubert ...........................199<br />

Olefinic Bioreduction,<br />

Naseem Iqbal, Marko. Mihovilovic. ....................201<br />

viii


Intraband Auger effect in InAs/InGaAlAs/InP Quantum Dot Structures,<br />

Thomas Gebhard, Deborah Alvarenga, Karl Unterrainer .........203<br />

Effect <strong>of</strong> Various Composition and Dosage <strong>of</strong> Potassium Activator on <strong>the</strong> Properties<br />

<strong>of</strong> Alkali Activated Slag Mortar,<br />

Iva Frybortova, Pavel Rovnanik ......................205<br />

Impact <strong>of</strong> Mineral Dust on PM10 Levels in Austria - Seasonal and local variations<br />

<strong>of</strong> carbonate carbon in ambient PM10 and source samples,<br />

Nicole Jankowski, Hans Puxbaum .....................207<br />

PE-UHMW in hip implants: The influence <strong>of</strong> crosslinking on structural parameters<br />

and micro-mechanical properties,<br />

Ruth Markut-Kohl, Thomas Koch, Vasiliki-Maria Archodoulaki, Sabine<br />

Seidler ....................................209<br />

Syn<strong>the</strong>sis <strong>of</strong> Organo-Electronic Compounds: A Cooperative Sparkling Science<br />

Research Project,<br />

Daniel Lumpi, Caterina Benigni, Aileen Opelt, Ernst Horkel, Johannes<br />

Fröhlich ...................................211<br />

Shadowgrahic Imaging <strong>of</strong> Femto Second Laser Induced Ablation <strong>of</strong> Al,<br />

Shazia Bashir, M:Shahid Rafique, Wolfgang Husinsky ...........213<br />

3D-analysis <strong>of</strong> neuronal and vascular networks by ultramicroscopy,<br />

Nina Jährling, Stefan Kalchmair, Klaus Becker, Andreas Gewies, Edgar<br />

R. Kramer, Hans-Ulrich Dodt ........................215<br />

Regioselective O-Glucuronidation <strong>of</strong> Deoxynivalenol-derived Model Compounds,<br />

Philipp Fruhmann, Ernst Horkel, Christian Hametner, Johannes Fröhlich 217<br />

Optimisation <strong>of</strong> <strong>the</strong> incoupling geometry <strong>of</strong> a pulsed solid-state laser,<br />

Elisabeth Schwarz, Johannes Tauer, Heinrich K<strong>of</strong>ler, Ernst Wintner ....219<br />

Enhancement Mode HEMTs: Evaluation <strong>of</strong> Two Approaches by Numerical<br />

Simulation,<br />

Stanislav Vitanov, Vassil Palankovski ....................221<br />

Simulation <strong>of</strong> Q-switched solid-state laser for ignition,<br />

Franz Trawniczek, Johannes Tauer, Heinrich K<strong>of</strong>ler, Margarita Deneva,<br />

Ernst Wintner ................................223<br />

Priorities in Improving <strong>of</strong> Railway Infrastructure,<br />

Jirí Pospísil .................................225<br />

Raman and Surface Enhanced Raman Spectra <strong>of</strong>, Apigenin and Luteolin,<br />

Charlie Corredor, John Lombardi .....................227<br />

ix


A Sustainable Approach in Hybrid Materials Development: Metal Nanoparticles<br />

Syn<strong>the</strong>sis in Biobased Polymeric Systems,<br />

Colleen Achong, Praveen Kumar Vemula, Kareem Douglas, George John 229<br />

Micro-lenses Fabrication by Solvent-Casting <strong>of</strong> Chalcogenide Glass,<br />

Eric Sanchez, Craig Arnold .........................231<br />

Structure and Dynamics <strong>of</strong> <strong>the</strong> Protein P7 from Bacteriophage φ12,<br />

Ertan Eryilmaz, Ranajeet Ghose ......................233<br />

Dsh is required for <strong>the</strong> phosphorylation <strong>of</strong> Dpr1a by CKIδ,<br />

Evelyn Teran, Joni Seeling .........................235<br />

Momentum <strong>of</strong> Optical Airy Beams,<br />

Henry Sztul, Robert Alfano .........................237<br />

Electrochemistry, Optical Spectroscopy and DFT Calculations <strong>of</strong> Glutathionylcobalamin,<br />

Iya Likhtina, Ronald Birke .........................239<br />

Surface Enhanced Raman Spectroscopy <strong>of</strong> Flavanthrone,<br />

Jingjing Chang, Maria Vega Cañamares, John Lombardi .........241<br />

Nonlinear resistance <strong>of</strong> 2D electrons in crossed electric and magnetic fields,<br />

JingQiao Zhang, Sergey Vitkalov ......................243<br />

Understanding <strong>the</strong> Effects <strong>of</strong> Aerosols in Cloud Microphysics in Coastal Urban<br />

Environments,<br />

Nathan Hosannah, Jorge Gonzalez .....................245<br />

Derivation <strong>of</strong> a Water Potential that Parameterizes Cooperative Effects,<br />

Philipa Njau, Michael Green ........................247<br />

Real-Time Visualization And Quantification Of Glial Progenitor And Glioma<br />

Chemotactic Migration,<br />

Richard Able, Maribel Vazquez .......................249<br />

Surface Enhanced Raman Spectroscopy <strong>of</strong> Pyridine on CdSe/ZnBeSe Quantum<br />

Dots Grown by Molecular Beam Epitaxy,<br />

Richard Livingstone, John Lombardi ....................251<br />

3 Environmental Science and Energy 253<br />

Using RFID Technology in Pedestrian Navigation for Information Transmission<br />

and Data Communication Recording,<br />

Qing Fu, Guen<strong>the</strong>r Retscher ........................255<br />

x


Adaptive Boundary Element Methods Based on Accurate A Posteriori Error<br />

Estimation,<br />

Samuel Ferraz-Leite, Dirk Praetorius ...................257<br />

Risk Habitat Megacity. Recycling Systems in Santiago de Chile,<br />

Tahnee Gonzalez, Klaus-Rainer Bräutigam, Helmut Seifert ........259<br />

Slurry sampling ICP-AES procedure for fast and accurate measurement <strong>of</strong><br />

crustal elements in airborne particulate matter,<br />

Azam Mukhtar, Andreas Limbeck ......................261<br />

Effects <strong>of</strong> Environmental Degradation on <strong>the</strong> Response <strong>of</strong> Reinforced Concrete<br />

Structures,<br />

Paola Simioni, Anna Saetta .........................263<br />

Usage <strong>of</strong> <strong>the</strong> Global Positioning System GPS in <strong>the</strong> fields <strong>of</strong> wea<strong>the</strong>r forecast<br />

and fleet navigation,<br />

Ana Karabatic, Gottfried Thaler, Robert Weber ..............265<br />

Organically Modified Mixed-Oxide Sol-Gel Films with Macro Meso and Micro<br />

Porosity,<br />

Marina Lomoschitz, Ulrich Schubert ....................267<br />

Development <strong>of</strong> a New High Frequency Fatigue Testing Method for Concrete<br />

Bridges,<br />

Johannes Berger, Johann Kollegger, Bernd Koeberl ............269<br />

Antenna far-fields determination from phaseless measurement using <strong>the</strong> realvalued<br />

GA,<br />

Jan Puskely, Zdenek Novácek ........................271<br />

Catalyst loaded porous membranes for environmental applications - Smart<br />

Membranes,<br />

Bin Ren, Ilona Kretzschmar .........................273<br />

Presence <strong>of</strong> <strong>the</strong> Bacteria Legionella in Hot Water Distribution Systems,<br />

Daniela Ocipova, Zuzana Vranayova ....................275<br />

Development and Processing <strong>of</strong> a Multi-Filter Rotating Shadowband Radiometer<br />

Network for Distributed Monitoring <strong>of</strong> Aerosol Optical Depth: Comparisons<br />

Between Conventional Langley Regression and a Novel NASA-<br />

GISS Algorithm,<br />

Miguel Bustamante, Fred Moshary, Barry Gross, Sam Ahmed .......277<br />

Impact Analysis <strong>of</strong> Resort Projects in Alpine Area,<br />

Lu Shen ...................................279<br />

Random Bipartite Graphs and <strong>the</strong>ir Application to Cuckoo Hashing,<br />

Reinhard Kutzelnigg, Michael Drmota (FacultyMentor) ..........281<br />

xi


Adaptive hp-Finite Element Methods for Two-Dimensional Elasticity with<br />

Tresca Friction,<br />

Philipp Dörsek, Jens Markus Melenk ....................283<br />

Thermal Bridge in Detail <strong>of</strong> a Gully in a Warm Flat Ro<strong>of</strong>,<br />

Petr Zahradník ...............................285<br />

Development <strong>of</strong> Novel Flow Sensors using SPICE Simulations,<br />

Samir Cerimovic, Almir Talic, Franz Keplinger ..............287<br />

Pre-heating Vegetable Oil as a Sustainable Energy Development Solution : A<br />

Bi<strong>of</strong>uel-Engine Longevity Test,<br />

Amir Nosrat, Mostafa Morovati, Martin Nolan, Paul A. West (Mentor) . . 289<br />

The "G" Ring Laser in Wettzell, Germany, as an Indicator for Astronomical<br />

and Geophysical Disturbances <strong>of</strong> <strong>the</strong> Planet Earth,<br />

Lucia Plank, Mendes Cerveira P.J. .....................291<br />

Occurrence and Possible Origin <strong>of</strong> Humic-like substances in Vienna Airborne<br />

Particulate Matter,<br />

Barbara Klatzer, Hans Puxbaum, Heidi Bauer, Alex Caseiro .......293<br />

Communication Specialties between Ear and Brain: Analysis <strong>of</strong> <strong>the</strong> Cluster<br />

Functions in <strong>the</strong> Human Cochlea,<br />

Cornelia Wenger, Frank Rattay .......................295<br />

Spatial potentials for renewable energies,<br />

Hartmut Dumke, Stefan Plha, Hannes Schaffer, Andreas Dillinger ....297<br />

Silica nanoparticles connected through ionic linkers,<br />

Marco Litschauer, Marie-Alexandra Neouze ................299<br />

An Adaptive Solver for Parabolic Partial Differential Equations,<br />

Gernot Pulverer ...............................301<br />

ICT - The Big Picture: Intertwined Driving Potentials Innovation, Technology<br />

Penetration and <strong>the</strong> Global Economy,<br />

DI Ranja Reda ...............................303<br />

Activation <strong>of</strong> Neural Networks in <strong>the</strong> Human Spinal Cord with Implanted and<br />

Surface Electrodes: Conclusions from Finite Element Simulations,<br />

Josef Ladenbauer, Ursula S. H<strong>of</strong>stötter, Karen Minassian, Milan R. Dimitrijevic,<br />

Frank Rattay ............................305<br />

Characterisation <strong>of</strong> biomass smoke particle emissions and its application to<br />

source apportionment,<br />

Christoph Schmidl, Hans Puxbaum .....................307<br />

xii


Observations <strong>of</strong> <strong>the</strong> surface evolution <strong>of</strong> Pinka-plain: mass movements and<br />

neotectonics,<br />

Gábor Kovács, Balázs Székely, Sándor Papp ................309<br />

Spray Can Geometry,<br />

Gwen Wilke, Andrew U. Frank .......................311<br />

Anammox Studies Using New York City Centrate to Correlate Performance,<br />

Population Dynamics, and Impact <strong>of</strong> Toxins,<br />

Alex Rosenthal, John Fillos, Krish Ramalingam ..............313<br />

Multi-Perspective Ultramicroscopy with Structured Illumination in Optically<br />

Dense Specimens,<br />

Stefan Kalchmair, Klaus Becker, Nina Jährling, Hans-Ulrich Dodt ....315<br />

Analytical technology for biogas production from straw by multivariate Nearinfrared<br />

spectroscopy,<br />

Chularat Krongtaew, Karin Fackler, Barbara Hinterstoisser, Kurt Messner 317<br />

Playing with maps: The role <strong>of</strong> webcartography in education,<br />

Eszter Simonné-Dombóvári .........................319<br />

Overtopping <strong>of</strong> Dams - A Model Approach,<br />

Michael Pucher, Peter Tschernutter ....................321<br />

Rediscovery <strong>of</strong> a Special Type <strong>of</strong> Separation Pier,<br />

Michael Pucher, Peter Tschernutter ....................323<br />

Possibilities <strong>of</strong> bentonite substitution - development <strong>of</strong> new progressive grouting<br />

materials and durability monitoring.,<br />

Pavla Matulova ...............................325<br />

Combined Active/Passive Microwave Wavelet-Based Approach For Snowmelt<br />

Detection Over Antarctica Ice Shelves,<br />

Nicholas Steiner, Marco Tedesco ......................327<br />

Binary Quadratic Forms over <strong>the</strong> Modular Ring Z/nZ,<br />

Clyde Lewis, Rony Gouraige ........................329<br />

Index <strong>of</strong> Authors 331<br />

xiii


xiv


Welcome Message from Peter Skalicky<br />

By surfing in Wikipedia one finds <strong>the</strong> following definition:<br />

“The scientific community consists <strong>of</strong> <strong>the</strong> total<br />

body <strong>of</strong> scientists, its relationships and interactions.”<br />

And fur<strong>the</strong>r: “Communication between <strong>the</strong> members<br />

is established by disseminating research work and<br />

hypo<strong>the</strong>ses through articles in peer reviewed journals,<br />

or by attending conferences where new research<br />

is presented and ideas are exchanged and discussed.”<br />

So in fact - aside from publications - conferences are<br />

<strong>the</strong> core opportunity for young scientists to become<br />

part <strong>of</strong> <strong>the</strong>ir community, present <strong>the</strong>ir research, make<br />

contacts and make <strong>the</strong>mselves a name.<br />

The <strong>Junior</strong> <strong>Scientist</strong> <strong>Conference</strong> at <strong>the</strong> Vienna University <strong>of</strong> Technology<br />

(TU Wien) is specially designed to make it easier for young researchers to<br />

enter this arena. So it is my pleasure to wish all <strong>of</strong> you good luck and a<br />

good start for <strong>the</strong> presentation <strong>of</strong> your posters and for getting into contact<br />

with each o<strong>the</strong>r.<br />

The <strong>Junior</strong> <strong>Scientist</strong> <strong>Conference</strong> is also a good example for <strong>the</strong> collaboration<br />

within <strong>the</strong> heart <strong>of</strong> Europe. Aside from our friends from New York<br />

where <strong>the</strong> first conference tooks place, <strong>the</strong>re are participants from Austria,<br />

Croatia, <strong>the</strong> Czech Republic, Denmark, Germany, Hungary, Italy and Slovakia.<br />

The Central European Region - in short: Centrope - is a political<br />

commitment to overcome former borders. One promising way to reach<br />

this goal is <strong>the</strong> intensive collaboration between scientists.<br />

Peter Skalicky<br />

Rector at Vienna University <strong>of</strong> Technology<br />

xv


xvi


Welcome Message from Hans K. Kaiser<br />

The idea for such a <strong>Junior</strong> <strong>Scientist</strong> <strong>Conference</strong> originated<br />

in <strong>the</strong> discussions between a delegation from<br />

<strong>the</strong> City College <strong>of</strong> New York and members <strong>of</strong> <strong>the</strong><br />

TU Wien during <strong>the</strong> visit <strong>of</strong> our American colleagues<br />

to Vienna, headed by president Gregory Williams, in<br />

2004. The aim <strong>of</strong> such a conference is to give young<br />

researchers a first possibility to present <strong>the</strong> results <strong>of</strong><br />

<strong>the</strong>ir scientific work to a wider audience. In <strong>the</strong> centre<br />

<strong>of</strong> such a conference is <strong>the</strong> presentation <strong>of</strong> posters<br />

by young scientists. The first conference in this series<br />

took place in New York in March 2005, <strong>the</strong> second<br />

one in Vienna in 2006, and a third conference was organized in New<br />

York in 2007. TU Wien is happy to organize now this fourth event in a -<br />

hopefully - long series <strong>of</strong> similar conferences to come. There was a call<br />

for scientific contributions in all fields <strong>of</strong> Science and Technology both on<br />

Master- and PhD-level. Preference was given to posters from <strong>the</strong> area <strong>of</strong><br />

Environmental Sciences, Information and Communication Technology and<br />

Material Sciences. We are very pleased that <strong>the</strong> announcement <strong>of</strong> our conference<br />

was well received both by universities in our neighbouring countries<br />

and our friends from <strong>the</strong> City College <strong>of</strong> New York. So we are proudly<br />

welcoming more than 160 young researchers from nine different countries.<br />

At <strong>the</strong> heart <strong>of</strong> <strong>the</strong> conference in Vienna will be <strong>the</strong> poster session. Around<br />

this main event <strong>the</strong>re will be three mini-conferences in <strong>the</strong> area <strong>of</strong> Environmental<br />

Sciences, Material Sciences and Information and Communication<br />

Technology. Within <strong>the</strong> conference we are organizing a best poster award:<br />

one for posters on <strong>the</strong> Master-level and one on PhD-level.<br />

We wish all participants an exciting time in Vienna and a successful<br />

conference at <strong>the</strong> Vienna University <strong>of</strong> Technology.<br />

Hans K. Kaiser<br />

Vice Rector for International Relations at Vienna University <strong>of</strong> Technology<br />

xvii


xviii


<strong>Conference</strong> Organizers<br />

<strong>Conference</strong> Chairman<br />

Hans K. Kaiser Vice Rector for International Relations at TU Vienna<br />

Scientific Program Committee<br />

Ewald Benes Institute <strong>of</strong> General Physics, TU Vienna<br />

Jürgen Dorn Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems, TU Vienna<br />

Gerald Futschek Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems, TU Vienna<br />

Georg Gartner Institute <strong>of</strong> Geoinformation and Cartography, TU Vienna<br />

Hermann H<strong>of</strong>bauer Institute <strong>of</strong> Chemical Engineering, TU Vienna<br />

Gün<strong>the</strong>r Karigl Inst.<strong>of</strong> Discrete Ma<strong>the</strong>matics and Geometry, TU Vienna<br />

Raimund Kirner Institute <strong>of</strong> Computer Engineering, TU Vienna<br />

Andreas Kolbitsch Institute for Building Construction and Technology, TU Vienna<br />

Kurt Matyas Institute <strong>of</strong> Management Science, TU Vienna<br />

Helmut Rechberger Institute for Water Quality, Resources and Waste Management, TU<br />

Vienna<br />

Manfred Schrödl Institute <strong>of</strong> Electrical Drives and Machines, TU Vienna<br />

Organizing Committee<br />

Raimund Kirner Assistant Pr<strong>of</strong>essor<br />

Nicole Pacher Assistant to <strong>the</strong> Vice Rector<br />

Edith Scharaditsch Secretary to <strong>the</strong> Vice Rectorate<br />

Simone Souczek Secretary to <strong>the</strong> Vice Rectorate<br />

Melanie Wagner Public Relations and Communication<br />

xix


<strong>Conference</strong> Program<br />

Sunday, November 16, <strong>2008</strong><br />

Time Event Location<br />

05.00 pm -7.00 pm Registration and Poster Setup Prechtl-Saal, Main<br />

Building, Ground<br />

(US: 1st) Floor<br />

07.00 pm (open end) Get toge<strong>the</strong>r at Nelson’s Café University courtyard,<br />

Main Building<br />

Monday, November 17, <strong>2008</strong><br />

Preparation<br />

Time Event Location<br />

08.30 am - 10.00 am Registration and Poster Setup Prechtl-Saal, Main<br />

Building, Ground<br />

(US: 1st) Floor<br />

Morning Session<br />

Time Event Location<br />

10.00 am - 10.30 am Opening Lecture Room 8,<br />

Main Building,<br />

Ground (US:<br />

1st) Floor<br />

10.30 am - 11.15 am Keynote: Pr<strong>of</strong>. Dr. Werner PURGATH-<br />

OFER (Institute <strong>of</strong> Computer Graphics<br />

and Algorithms): "Speedup Concepts<br />

for Realtime Computer Graphics"<br />

11.15 am - 12.00 pm Keynote: Pr<strong>of</strong>. Dr. Karl UNTER-<br />

RAINER (Institute <strong>of</strong> Photonics): "Coherent<br />

Terahertz Generation: New Light<br />

for Nanostructures and Bio-Imaging"<br />

12.00 pm - 02.00 pm Lunch Break Mensa, Freihaus<br />

Bldg., Wiedner<br />

Hauptstr. 8-10<br />

xxi


Afternoon Session (Parallel Sessions)<br />

Session A) Environmental Sciences and Energy<br />

Time Event Location<br />

02.00 pm - 02.45 pm Plenary Speaker: Pr<strong>of</strong>. Dr. Nebojsa<br />

NAKICENOVIC (Institute <strong>of</strong> Power<br />

Systems and Energy Economics): "Energy<br />

Perspectives and Climate Challenge"<br />

02.45 pm - 03.10 pm Student Presentation: Nina<br />

JÄHRLING (Graduate Student, Vienna<br />

University <strong>of</strong> Technology):<br />

"3D-Analysis <strong>of</strong> neuronal and vascular<br />

networks by ultramicroscopy (143)"<br />

03.10 pm - 03.35 pm Student Presentation: Daniela<br />

OCIPOVA (Graduate Student, Technical<br />

University in Kosice): "Presence <strong>of</strong><br />

<strong>the</strong> Bacteria Legionella in Hot Water<br />

Distribution Systems (34)"<br />

03.35 pm - 04.00 pm Student Presentation: Chularat KRO-<br />

NGTAEW (Graduate Student, Vienna<br />

University <strong>of</strong> Technology): "Analytical<br />

Technology for Biogas Production from<br />

Straw by Multivariate Near-Infrared<br />

Spectroscopy"<br />

Lecture Room 8,<br />

Main Building,<br />

Ground Floor<br />

04.00 pm - 04.20 pm Break Prechtl-Saal, Main<br />

Building, Ground<br />

Floor<br />

04.20 pm - 04.45 pm Student Presentation: Christoph<br />

SCHMIDL (Graduate Student, Vienna<br />

University <strong>of</strong> Technology): "Characterisation<br />

<strong>of</strong> Biomass Smoke Particle<br />

Emissions and its Application to Source<br />

Receptor Analysis"<br />

04.45 pm - 05.10 pm Student Presentation: Marina LO-<br />

MOSCHITZ (Graduate Student, Vienna<br />

University <strong>of</strong> Technology): "Sol-Gel<br />

Films with Macro Meso and Micro<br />

Porosity"<br />

xxii<br />

Lecture Room 8,<br />

Main Building,<br />

Ground Floor


Session B) Material Sciences<br />

Time Event Location<br />

02.00 pm - 02.45 pm Plenary Speaker: Asst. Pr<strong>of</strong>. Dr. Ilse-<br />

Christine GEBESHUBER (Institute <strong>of</strong><br />

General Physics): "Scanning Probe Microscopy<br />

across Dimensions"<br />

02.45 pm - 03.10 pm Student Presentation: Clemens OS-<br />

TERMAIER (Graduate Student, Vienna<br />

University <strong>of</strong> Technology): "Realization<br />

<strong>of</strong> Inversion-type Gan MOSFETs<br />

with Ar Implantation for Device Isolation<br />

(114)"<br />

03.10 pm - 03.35 pm Student Presentation: Michael KITZ-<br />

MANTEL (Graduate Student, Vienna<br />

University <strong>of</strong> Technology): "Nan<strong>of</strong>eatured<br />

Carbon-Copper Metal Matrix<br />

Composites (74)"<br />

03.35 pm - 04.00 pm Student Presentation: Melanie TODT<br />

(Graduate Student, Vienna University <strong>of</strong><br />

Technology): "Long Wave Instabilities<br />

in Periodic Structures (91)"<br />

Lecture Room 7,<br />

Main Building,<br />

Ground Floor<br />

04.00 pm - 04.20 pm Break Prechtl-Saal, Main<br />

Building, Ground<br />

Floor<br />

04.20 pm - 04.45 pm Student Presentation: Stanislav VI-<br />

TANOV (Graduate Student, Vienna<br />

University <strong>of</strong> Technology): "Enhancement<br />

Mode HEMTs: Evaluation <strong>of</strong> Two<br />

Approaches by Numerical Simulation<br />

(158)"<br />

04.45 pm - 05.10 pm Student Presentation: Philipp<br />

DÖRSEK (Undergraduate Student,<br />

Vienna Univ. <strong>of</strong> Technology): "Adaptive<br />

hp-Finite Element Methods for<br />

Two-Dimensional Elasticity with Tresca<br />

Friction (49)"<br />

xxiii<br />

Lecture Room 7,<br />

Main Building,<br />

Ground Floor


Session C) Information, Communication and Computer Technology<br />

Time Event Location<br />

02.00 pm - 02.45 pm Plenary Speaker: Assoc. Pr<strong>of</strong>. Dr. Andreas<br />

RAUBER (Institute <strong>of</strong> S<strong>of</strong>tware<br />

Technology and Interactive Systems):<br />

Topic to be announced<br />

02.45 pm - 03.10 pm Student Presentation: Markus<br />

TRISKA (Graduate Student, Vienna<br />

University <strong>of</strong> Technology): "Generalizing<br />

Constraint Solving over Finite<br />

Domains (121)"<br />

03.10 pm - 03.35 pm Student Presentation: Florian SEIT-<br />

NER (Graduate Student, Vienna University<br />

<strong>of</strong> Technology): "Development <strong>of</strong><br />

Multi-Core Video Decoding Platforms<br />

based on High Level Architecture Simulations<br />

(105)"<br />

03.35 pm - 04.00 pm Student Presentation: Qi WANG<br />

(Graduate Student, Vienna University<br />

<strong>of</strong> Technology): "Towards Synchronization<br />

in Multiple Antenna Systems (42)"<br />

Lecture Room 13,<br />

Main Building,<br />

2nd (US: 3rd)<br />

Floor<br />

04.00 pm - 04.20 pm Break Prechtl-Saal, Main<br />

Building, Ground<br />

Floor<br />

04.20 pm - 04.45 pm Student Presentation: Theresia<br />

GSCHWANDTNER (Graduate Student,<br />

Vienna University <strong>of</strong> Technology):<br />

"Map Face - A Graphical Editor to<br />

Support <strong>the</strong> Semantic Annotation <strong>of</strong><br />

Medical Text (139)"<br />

04.45 pm - 05.10 pm Student Presentation: Rushane DYER<br />

(Student, City College New York):<br />

"Retrieval <strong>of</strong> Ocean Water Parameters<br />

from Satellite Imagery and <strong>the</strong>ir Dependance<br />

on Atmospheric Correction Models<br />

(100)"<br />

xxiv<br />

Lecture Room 13,<br />

Main Building,<br />

2nd (US: 3rd)<br />

Floor


Evening Program<br />

Time Event Location<br />

07.00 pm (open end) Enjoy an evening at <strong>the</strong> Wiener Heuriger<br />

Fuhrgassl-Huber<br />

Tuesday, November 18, <strong>2008</strong><br />

1190 Wien,<br />

Neustift/Walde 68<br />

Time Event Location<br />

09.00 am - 10.00 am Presenting and visiting Poster Session 1 Prechtl-Saal, Main<br />

Building, Ground<br />

(US: 1st) Floor<br />

10.00 am - 11.00 am Presenting and visiting Poster Session 2<br />

11.00 am - 11.30 am Break Prechtl-Saal, Main<br />

Building, Ground<br />

Floor<br />

11.30 am - 12.15 pm Keynote: Pr<strong>of</strong>. Dr. Manfred<br />

GRASSERBAUER (Institute <strong>of</strong> Chemical<br />

Technologies and Analytics): " ’Our<br />

Environment - Our Future’. Sustainable<br />

Development in <strong>the</strong> European Union -<br />

Concepts, Approaches and Challenges"<br />

12.15 pm - 12.45 pm Award Ceremony<br />

12.45 pm - 01.00 pm Closing Ceremony/End <strong>of</strong> <strong>Conference</strong><br />

xxv<br />

Lecture Room 8,<br />

Main Building,<br />

Ground Floor


xxvi


Chapter 1<br />

Information and Communication<br />

Technology


Implementation <strong>of</strong> Quadrature Modulator Imbalance<br />

Compensation Method<br />

Karel Povala�, Roman Maršálek (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronics<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech republic<br />

Email: xpoval00@stud.feec.vutbr.cz, marsaler@feec.vutbr.cz<br />

Abstract — Quadrature modulator (demodulator)<br />

is used in transmitting (receiving) part <strong>of</strong> many<br />

devices. Modulator (demodulator) output signal can<br />

be influenced by unwanted amplitude, phase or DC<br />

<strong>of</strong>fset imbalances. The implementation <strong>of</strong> basic<br />

method for imbalances compensation on programmable<br />

logic filed is a main subject <strong>of</strong> <strong>the</strong> paper.<br />

Development kit V2MB1000 with analogue board<br />

Memec P160 was chosen for this purpose. Practical<br />

measurements have been done with digital oscilloscope<br />

and digital signal analyzer.<br />

I. INTRODUCTION – REAL MODULATOR<br />

IQ imbalance problem arises when non-ideal components<br />

injure <strong>the</strong> power balance or phase orthogonality<br />

between inphase (I) and quadrature (Q) branch <strong>of</strong><br />

modulator. Methods for IQ imbalance compensation<br />

are suitable for VHDL implementation in programable<br />

logic field. Figure 1 shows block diagram <strong>of</strong><br />

real modulator with imbalances.<br />

Figure 1: Modulator with imbalances<br />

I and Q signals are usually produced in DSP block<br />

(digital signal processor). Digital signals are converted<br />

into analogue form by digital/analogue converters.<br />

Active Low Pass Filter (ALPF) can add<br />

unwanted amplitude imbalance (KI, KQ) or DC <strong>of</strong>fset<br />

(a1, a2). Phase imbalance (�err) express phase difference<br />

between both branches, which is figured by <strong>the</strong><br />

oscillator block and <strong>the</strong> phase shift block. Adder is<br />

<strong>the</strong> last block <strong>of</strong> quadrature amplitude modulator.<br />

3<br />

Chosen method should adaptively compensate unwanted<br />

above mentioned imbalances.<br />

II. COMPENSATION METHOD [1]<br />

A. MATRIX MODEL<br />

Choosen method was published by Ingolf Held in<br />

paper [1]. IQ imbalance can be characterized by two<br />

parameters: amplitude imbalance K as a power mismatch<br />

between I and Q branch, and phase imbalance<br />

�err as an error <strong>of</strong> orthogonality between I and Q<br />

branch. Situation can be characterized by <strong>the</strong> matrix<br />

equation:<br />

,<br />

�s � �<br />

0<br />

I k � � KI<br />

� � sI �k ��<br />

� ��<br />

. (1)<br />

� � � ,<br />

s � � KQ sin err KQ<br />

cos<br />

Q k � �<br />

� �<br />

� � �� � � err�<br />

��sQ�k��� Where s’I and s’Q denote components <strong>of</strong> unimpaired<br />

signals. KI and KQ represents <strong>the</strong> amplitude<br />

imbalance and �err represents <strong>the</strong> phase imbalance.<br />

B. AMPLITUDE IMBALANCE COMPENSATION CO-<br />

EFICIENT<br />

Estimate <strong>of</strong> amplitude imbalance is based on <strong>the</strong><br />

following equation<br />

K<br />

�<br />

L<br />

�<br />

k �1<br />

est, B L<br />

�<br />

k�1<br />

2<br />

Q<br />

2<br />

I<br />

� �<br />

s k<br />

� �<br />

s k<br />

(2)<br />

The long-time domain preamble has been used as<br />

training input sequence s consisting <strong>of</strong> sI and sQ parts.<br />

Parameter L denotes a number <strong>of</strong> long preamble<br />

samples for computing estimation.<br />

C. PHASE IMBALANCE COMPENSATION COEFI-<br />

CIENT<br />

Estimation <strong>of</strong> phase imbalance coefficient Pest uses<br />

<strong>the</strong> same data symbols s <strong>of</strong> length L. The following<br />

equation is used<br />

L<br />

��<br />

� �� � ��<br />

s k s k<br />

I Q<br />

k �1<br />

est � L<br />

2<br />

� sI�k� k �1<br />

P<br />

(3)


When both <strong>of</strong> compensation coefficients are found<br />

correction <strong>of</strong> imbalances can continue.<br />

III. FPGA IMPLEMENTATION<br />

The Xilinx ISE environment was used for implementation.<br />

The compensation algorithm had to be rebuilt<br />

for VHDL implementation. The basic blocks are<br />

shown on simplified schematic (Figure 2).<br />

Figure 2: Scheme <strong>of</strong> implemented method<br />

Clock speed <strong>of</strong> FPGA was 100 MHz. Digital<br />

Clock Manager (DCM) provided clocking for o<strong>the</strong>r<br />

blocks. Random symbols (s’I and s’Q) were generated<br />

by DATA generator. Data symbols passed through<br />

<strong>the</strong> model <strong>of</strong> imbalances. Next steps were calculating<br />

compensation coefficients, imbalance compensation,<br />

filtering signals by Square Root Raised Cosine filter<br />

(elimination inter symbol interferences) and modulation<br />

on <strong>the</strong> carrier with frequency 500 kHz. All <strong>of</strong><br />

ma<strong>the</strong>matic operations were done in floating point<br />

format. The development kit V2MB1000 has been<br />

used for <strong>the</strong> implementation. Almost 100% logic<br />

block inside FPGA were used. Signal from modulator<br />

has been sent into analogue board Memec P160<br />

with <strong>the</strong> D/A converter. Output analogue signal could<br />

be measured and displayed by digital oscilloscope<br />

and digital signal analyzer.<br />

IV. MEASUREMENTS AND RESULTS<br />

The output analogue signal was measured in time<br />

domain. Measuremt results corresponded with simulations.<br />

Digital signal analyzer (Rohde & Schwarz<br />

4<br />

FSQ3) was used for display constellation diagrams <strong>of</strong><br />

signal before compensation (with amplitude imbalances<br />

KI = 1,1and KQ = 0,9, phase imbalance �err=�<br />

/10rad) and after compensation (without imbalances).<br />

The graphic results are shown on Figure 3.<br />

Figure 3: Constellation diagrams before (left) and<br />

after compensation (right)<br />

It’s evident that both <strong>of</strong> imbalances have been<br />

compensated. The constellation points have been<br />

dispersed as a consequence <strong>of</strong> shorter impulse response<br />

<strong>of</strong> used raised cosine filter.<br />

V. CONCLUSION<br />

Number <strong>of</strong> logic blocks used in FPGA Virtex II<br />

XC2V1000 reached almost 100%. Method was implemented<br />

toge<strong>the</strong>r with data generation, filter for<br />

elimination inter symbol interferences and direct<br />

digital syn<strong>the</strong>sizer generated carrier for modulator.<br />

Measured results proved right function <strong>of</strong> implemented<br />

method.<br />

ACKNOWLEDGMENTS<br />

Implementation described in <strong>the</strong> paper was financially<br />

supported by <strong>the</strong> Czech Grant Agency under<br />

grant No. 102/08/H027 "Advanced Methods, Structures<br />

and Components <strong>of</strong> Electronic Wireless Communication"<br />

and by <strong>the</strong> research program MSM<br />

0021630513 "Advanced Electronic Communication<br />

Systems and Technologies (ELCOM)".<br />

REFERENCES<br />

[1] I. Held, O. Klein, A. Chen and V. Ma. Low<br />

Complexity Digital IQ Imbalance Correction in<br />

OFDM WLAN Receivers. IEEE Integrated System<br />

Solution Corporation. Hsinchu, Taiwan,<br />

2004, p. 1172 – 1176


Diagnostical analysis <strong>of</strong> voice<br />

Pavel Sala, Milan Sigmund (Faculty Mentor)<br />

Faculty <strong>of</strong> Electrical Engineering and Communication<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: xsalap00@stud.feec.vutbr.cz, sigmund@feec.vutbr.cz<br />

Abstract — This paper is focused on methods for<br />

<strong>the</strong> estimation <strong>of</strong> glottal pulses from speech signal<br />

and finding appropriate criterions to describe a<br />

selected diagnosis. Two methods for <strong>the</strong> estimation<br />

<strong>of</strong> <strong>the</strong> glottal flow from <strong>the</strong> speech signal were<br />

programmed. Finally, attention was focused on <strong>the</strong><br />

determination <strong>of</strong> criterions for <strong>the</strong> description <strong>of</strong> a<br />

selected pathological diagnosis and <strong>the</strong> influence <strong>of</strong><br />

<strong>the</strong> stress on <strong>the</strong> glottal flow. The proposal <strong>of</strong> two<br />

criterions for describing <strong>the</strong> influence <strong>of</strong> <strong>the</strong> stress<br />

on <strong>the</strong> glottal flow is an outcome <strong>of</strong> this work.<br />

I. INTRODUCTION<br />

Parameters for <strong>the</strong> estimation <strong>of</strong> some diagnose are<br />

<strong>of</strong>ten extracted directly from a speech signal in time<br />

or frequency domain. A quite new approach consists<br />

in obtaining appropriate parameters from <strong>the</strong> estimate<br />

<strong>of</strong> glottal pulses. The device which can measure<br />

glottal pulses directly is quite expensive, and<br />

<strong>the</strong>refore, <strong>the</strong> development <strong>of</strong> methods for estimating<br />

glottal pulses from speech signal was started.<br />

This paper is organized as follows. After<br />

introductions, two methods for <strong>the</strong> estimation <strong>of</strong><br />

glottal pulses that were programmed are mentioned<br />

in Section II. The Section III is focused on <strong>the</strong><br />

description <strong>of</strong> a selected pathological diagnosis. In<br />

Section IV, two criterions for describing <strong>the</strong><br />

influence <strong>of</strong> <strong>the</strong> stress on <strong>the</strong> glottal flow were<br />

suggested. Finally, concluding remarks are given in<br />

Section V.<br />

II. GLOTTAL PULSE ESTIMATION<br />

Both used methods are derived from <strong>the</strong> speech<br />

production model [1]. Block diagram <strong>of</strong> this model is<br />

showed in Fig. 1.<br />

Figure 1: The speech production model<br />

( ) 2 1 −<br />

( z)<br />

1/<br />

1−<br />

exp(<br />

− cT )<br />

G = z (1)<br />

5<br />

The equation 1 represents a glottis where c is <strong>the</strong><br />

unknown parametr <strong>of</strong> <strong>the</strong> filter and T is <strong>the</strong> sampling<br />

period. The glottal flow goes throught <strong>the</strong> vocal tract<br />

V(z) which is modeled by an all-pole filter defined in<br />

zplaneas<br />

V<br />

( z)<br />

= K −i<br />

1+<br />

∑ c ⋅<br />

i=<br />

i z<br />

1<br />

1 (2)<br />

where ci are coeficients <strong>of</strong> <strong>the</strong> vocal tract model and<br />

K is a number <strong>of</strong> <strong>the</strong>se coefficients. R(z) represents<br />

<strong>the</strong> lips radiation which can be described as<br />

−1<br />

R z = 1 − μ z<br />

(3)<br />

( )<br />

where μ∈〈0.98; 1〉.<br />

The first method was published by Paavo Alku in<br />

1992 and its name is “Iterative Adaptive Inverse<br />

Filtering” (IAIF) [2]. The second method was<br />

published by H. Deng, H. Ward, M. P. Beddoes and<br />

M. Hodgson in 2006 [3].<br />

III. PATHOLOGICAL DIAGNOSIS<br />

Three pathological diagnoses where processed:<br />

− reinek’s edema<br />

− vocal fold edema<br />

− vocal tremor<br />

The signal <strong>of</strong> 10 speakers has been available for each<br />

<strong>of</strong> <strong>the</strong> diagnoses. Each speech signal was splitted to 5<br />

sections and each section was <strong>the</strong>n used for <strong>the</strong><br />

estimate <strong>of</strong> <strong>the</strong> glottal pulse. Normalized forms <strong>of</strong><br />

estimated glottal pulses were displayed in <strong>the</strong> same<br />

graph. It turned out that <strong>the</strong> shape <strong>of</strong> glottal pulses <strong>of</strong><br />

different speakers (for <strong>the</strong> same pathological<br />

diagnose) is different. Therefore a few smaller<br />

groups with similar shapes <strong>of</strong> glottal pulses were<br />

created (for each pathological diagnosis).<br />

IV. STRESS<br />

Two speech signals <strong>of</strong> 9 speakers have been available<br />

[4] (<strong>the</strong> one “stressed” and <strong>the</strong> one “normal” speech<br />

signal <strong>of</strong> <strong>the</strong> same utterance for each speaker). Only


<strong>the</strong> first method [2] for glottal pulses estimation was<br />

used because <strong>of</strong> <strong>the</strong> small length <strong>of</strong> <strong>the</strong> speech signal.<br />

A. MONOTONICITY<br />

The monotonicity <strong>of</strong> each <strong>of</strong> <strong>the</strong> edges (rising and<br />

falling) <strong>of</strong> normalized glottal pulses was tested at<br />

first - concretely <strong>the</strong> number <strong>of</strong> changes <strong>of</strong> edges<br />

directions was observed. The threshold was<br />

determined when <strong>the</strong> error rate <strong>of</strong> <strong>the</strong> classification <strong>of</strong><br />

<strong>the</strong> speech signal was minimal see Fig. 2.<br />

Figure 2: Threshold searching<br />

From <strong>the</strong> monotonicity observation <strong>of</strong> <strong>the</strong> falling<br />

edge <strong>the</strong> minimal error was achieved for <strong>the</strong><br />

threshold value 8% see Fig. 3.<br />

type 2 error (dashed line) [%]<br />

60<br />

40<br />

20<br />

Falling edge<br />

0<br />

0 2 4 6 8 10 12 14 16 18 20<br />

threshold value [%]<br />

40<br />

Figure 3: Error <strong>of</strong> <strong>the</strong> classification for various<br />

thresholds<br />

B. SURFACE<br />

As <strong>the</strong> second classifier <strong>the</strong> surface <strong>of</strong> <strong>the</strong> rising and<br />

<strong>the</strong> falling edge was used. In Fig. 4 glottal pulses <strong>of</strong><br />

<strong>the</strong> normal speech signal (solid line) and <strong>the</strong> stressed<br />

speech signal (dotted line) <strong>of</strong> all 9 speakers are<br />

shown. In almost all <strong>of</strong> <strong>the</strong>se graphs, it is observable<br />

that in <strong>the</strong> right half <strong>of</strong> graphs, <strong>the</strong> surface <strong>of</strong> <strong>the</strong><br />

stressed glottal pulse is smaller than <strong>the</strong> surface <strong>of</strong><br />

<strong>the</strong> normal glottal pulse. If this is not accomplished,<br />

it is necessary to compare <strong>the</strong> left half <strong>of</strong> <strong>the</strong> graphs<br />

too. If <strong>the</strong> left surface <strong>of</strong> <strong>the</strong> examined glottal pulse is<br />

bigger than surface <strong>of</strong> <strong>the</strong> “normal” glottal pulse, it<br />

70<br />

60<br />

50<br />

type1error(solidline)[%]<br />

6<br />

can be determined that <strong>the</strong> glottal pulse is stressed.<br />

This method fails only in one case (speaker 1).<br />

0.9<br />

0.8<br />

0.7<br />

speaker1<br />

-100 -50 0<br />

speaker3<br />

50 100<br />

0.9<br />

0.8<br />

0.7<br />

-100 -50 0<br />

speaker5<br />

50 100<br />

0.9<br />

0.8<br />

0.7<br />

-100 -50 0<br />

speaker7<br />

50 100<br />

0.9<br />

0.8<br />

0.7<br />

-100<br />

1<br />

-50 0<br />

speaker9<br />

50 100<br />

0.8<br />

0.6<br />

-100 -50 0 50 100<br />

0.9<br />

0.8<br />

0.7<br />

speaker2<br />

-100<br />

1<br />

-50 0<br />

speaker4<br />

50 100<br />

0.8<br />

0.6<br />

-100 -50 0<br />

speaker6<br />

50 100<br />

0.9<br />

0.8<br />

0.7<br />

-100<br />

1<br />

-50 0<br />

speaker8<br />

50 100<br />

0.8<br />

0.6<br />

-100 -50 0 50 100<br />

Figure 4: Normal and stressed glottal pulses<br />

CONCLUSION<br />

Two criterions for description <strong>of</strong> <strong>the</strong> stressed speech<br />

signal were proposed. For improvement <strong>of</strong> <strong>the</strong><br />

accuracy <strong>of</strong> <strong>the</strong>se methods or for <strong>the</strong> proposition a<br />

new method for <strong>the</strong> stress description, more samples<br />

<strong>of</strong> <strong>the</strong> speech signal are needed.<br />

REFERENCES<br />

[1] J. Psutka and L. Müller and J. Matoušek and V.<br />

Radová. Mluvíme s počítačem česky, Academia,<br />

Praha, Czech Republic, 2006<br />

[2] H. Pulakka. Analysis <strong>of</strong> Human Voice Production<br />

Using Inverse Filtering. High-Speed Imaging,<br />

Helsinki University <strong>of</strong> Technology, Master’s<br />

Thesis, 2005.<br />

[3] H. Deng and R. K. Ward and M. P. Beddoes and<br />

M. Hodgson. A New Method for Obtaining<br />

Accurate Estimates <strong>of</strong> Vocal-Tract Filters and<br />

Glottal Waves from Vowel Sound. In<br />

<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> IEEE Transaction on audio,<br />

speech, and language processing. Vol. 14, No. 2.<br />

p. 445-455 . March 2006.<br />

[4] M. Sigmund. Introducing <strong>the</strong> database<br />

ExammStress for speech under stress.<br />

<strong>Proceedings</strong> <strong>of</strong> 7 th<br />

IEEE Nordic Signal<br />

Processing Symposium, 290–293, Reykjavik,<br />

2006.


Realization <strong>of</strong> <strong>the</strong> OFDM coder and decoder using DSP<br />

Petr Zelinka, Milan Sigmund (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronics<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: xzelin06@stud.feec.vutbr.cz, sigmund@feec.vutbr.cz<br />

Abstract — The European Standard ETSI EN<br />

300 744 for terrestrial digital video broadcasting<br />

specifies <strong>the</strong> OFDM (orthogonal frequency division<br />

multiplexing) for use in television broadcast stations.<br />

This article describes an OFDM baseband<br />

coder and decoder in 2K mode without errorcorrection<br />

capabilities constructed according to <strong>the</strong><br />

above-mentioned norm. The proper function <strong>of</strong> both<br />

devices is verified using Matlab simulations. The<br />

practical realization comprises two Texas Instruments’<br />

Starter Kits with TMS320C6711 digital<br />

signal processors (DSPs) connected through coaxial<br />

cables.<br />

I. INTRODUCTION<br />

Nowadays set-top-boxes and o<strong>the</strong>r similar devices<br />

use well-established application-specific integrated<br />

circuits (ASIC) for OFDM demodulation. These<br />

parts have predetermined purpose and parameters,<br />

hence, <strong>the</strong>y can’t be used in special applications like<br />

hand-held equipment or computer peripherals.<br />

The coder and decoder described in this article are<br />

configured for 2K mode according to <strong>the</strong> DVB-T<br />

standard and <strong>the</strong>refore minimal number <strong>of</strong> carriers is<br />

used ensuring short symbol period. This fact allows<br />

fast variances <strong>of</strong> <strong>the</strong> transmission channel. The used<br />

algorithms could be adapted for o<strong>the</strong>r modes as well.<br />

II. OFDM BASICS<br />

A typical terrestrial communication channel always<br />

causes certain degree <strong>of</strong> signal distortion – attenuation,<br />

added noise, interferences and fading due to<br />

strong echoes. High-speed digital signals may <strong>the</strong>refore<br />

exhibit severe inter-symbol interferences (ISI)<br />

limiting <strong>the</strong> maximum data rate for one carrier.<br />

OFDM uses many orthogonal carriers providing<br />

good transmission rates within a predetermined HF<br />

bandwidth. The existence <strong>of</strong> a so-called guard interval<br />

greatly improves <strong>the</strong> immunity to <strong>the</strong> echoes.<br />

Individual carriers can be influenced by means <strong>of</strong><br />

amplitude and phase changes <strong>the</strong>refore conveying<br />

desired information.<br />

7<br />

III. PROPOSED CODER<br />

A. SYSTEM PARAMETERS<br />

The system uses a 2K OFDM according to <strong>the</strong> standard<br />

[1] with <strong>the</strong> guard interval <strong>of</strong> ¼ (512 samples).<br />

Carriers are modulated using <strong>the</strong> QPSK or 16-QAM<br />

(uniform) to be able to study <strong>the</strong> impact <strong>of</strong> <strong>the</strong><br />

channel noise on <strong>the</strong> error rate with different<br />

constellation point distances. The system doesn’t<br />

contain blocks for <strong>the</strong> forward error correction.<br />

B. USED PILOT SIGNALS<br />

The modulated carriers are interleaved with continual<br />

and scattered pilots (SP) on <strong>the</strong> proper positions. The<br />

transmission parameters signalling pilots are set to<br />

zero since <strong>the</strong>ir function is not employed in <strong>the</strong> system.<br />

The total number <strong>of</strong> carriers is 1705, hence <strong>the</strong><br />

2048-IFFT is used for <strong>the</strong> transformation into <strong>the</strong><br />

time domain.<br />

C. SPECTRUM ADJUSTMENT<br />

The power spectrum <strong>of</strong> individual carriers follows a<br />

sinc 2 function and <strong>the</strong>ir superposition composes <strong>the</strong><br />

atoll-like shape. The sidelobes, however, show too<br />

slow fading, so an efficient filtration is needed. The<br />

proposed coder uses a windowing method described<br />

in [2] to suppress <strong>the</strong> out-<strong>of</strong>-band radiation. The only<br />

drawback <strong>of</strong> <strong>the</strong> method is a slight increase <strong>of</strong> <strong>the</strong><br />

echo sensitivity since <strong>the</strong> guard interval is shortened<br />

by a small fraction to preserve <strong>the</strong> proper symbol<br />

period.<br />

IV. PROPOSED DECODER<br />

A. MAIN STRUCTURE<br />

The reception <strong>of</strong> <strong>the</strong> OFDM signal comprises <strong>of</strong><br />

two parts: acquisition and tracking. When <strong>the</strong> receiver<br />

is switched-on it has no prior information<br />

about <strong>the</strong> symbol timing and <strong>the</strong>re may also be a<br />

nonzero frequency shift. In order to be able to demodulate<br />

<strong>the</strong> carriers, it is necessary to compensate<br />

<strong>the</strong> amplitude and phase characteristics <strong>of</strong> <strong>the</strong> transmission<br />

channel. The block scheme <strong>of</strong> <strong>the</strong> decoder


was composed (Figure 1) in accordance with [3].<br />

Some receivers use interpolation/decimation blocks<br />

for adjusting fractional timing <strong>of</strong>fsets, never<strong>the</strong>less<br />

this feature was not applied due to its high impact on<br />

required memory capacity and computational power.<br />

baseband<br />

OFDM<br />

signal<br />

A/D<br />

frequency<br />

correction<br />

pre-FFT<br />

analysis:<br />

-symbol<br />

timing<br />

-frequency<br />

guard<br />

interval<br />

removal<br />

Figure 1: Decoder block diagram<br />

B. ACQUISITION<br />

During <strong>the</strong> initial signal acquisition <strong>the</strong> decoder performs<br />

a correlation <strong>of</strong> <strong>the</strong> input signal comparing <strong>the</strong><br />

results against an adaptive threshold in order to find<br />

<strong>the</strong> beginning <strong>of</strong> an OFDM symbol. Subsequently,<br />

<strong>the</strong> first guess <strong>of</strong> <strong>the</strong> frequency <strong>of</strong>fset is computed<br />

using a method described in [2] and a s<strong>of</strong>tware converter<br />

is used for <strong>the</strong> necessary correction. After <strong>the</strong><br />

FFT’s completion, an integer frequency calibration is<br />

done and pilot signals are located. Only two subsequent<br />

symbols are needed to attain scattered pilots’<br />

pattern lock thanks to <strong>the</strong> method [4]. The final stage<br />

<strong>of</strong> <strong>the</strong> acquisition is <strong>the</strong> fine timing correction using<br />

SPs (with resolution limited to one sample period).<br />

C. TRACKING<br />

The receiver is able to perform continuous tracking<br />

and correcting <strong>of</strong> <strong>the</strong> timing and fractional frequency<br />

<strong>of</strong>fset. All received symbols are treated for linear<br />

distortion using a fast equalizer [5] which exploits<br />

SPs only from <strong>the</strong> current symbol and <strong>the</strong> following<br />

correction is applied on this symbol. This allows fast<br />

channel variances.<br />

V. PRACTICAL REALIZATION<br />

post-FFT<br />

alalysis:<br />

-frequency<br />

-timing<br />

-channel<br />

FFT<br />

equalizer<br />

carriers<br />

demodulation<br />

output<br />

data<br />

Two Starter Kits with TMS320C6711 DSPs connected<br />

through coaxial cables were used. Both run<br />

under <strong>the</strong> DSP BIOS operating system in cooperation<br />

with <strong>the</strong> Code Composer Studio v. 3.0. Several libraries<br />

<strong>of</strong> hand-optimized assembler routines were<br />

used on critical parts <strong>of</strong> <strong>the</strong> program.<br />

8<br />

VI. CONCLUSION<br />

The proposed baseband OFDM coder and decoder<br />

according to <strong>the</strong> DVB-T standard were successfully<br />

realized using two evaluation boards with DSPs. The<br />

decoder is adequately resistant to noise and linear<br />

distortion <strong>of</strong> <strong>the</strong> communication channel. Used<br />

algorithms are speed-optimized efficiently employing<br />

hardware resources, still, computational power <strong>of</strong> <strong>the</strong><br />

DSPs limits <strong>the</strong> data rate below <strong>the</strong> standardised<br />

value. A real-time operation would be certainly<br />

possible with faster fixed-point DSPs.<br />

ACKNOWLEDGMENTS<br />

This work was supported by <strong>the</strong> Czech Grant Agency<br />

under project No. 102/08/H027 and by <strong>the</strong> research<br />

program MSM 0021630513 Advanced Electronic<br />

Communication Systems and Technologies<br />

(ELCOM).<br />

References<br />

[1] European Standard ETSI EN 300 744: Digital<br />

Video Broadcasting (DVB); Framing structure,<br />

channel coding and modulation for digital terrestrial<br />

television [online]. Ver. 1.5.1. 11-2004. [cit.<br />

25.4.2007]. Available .<br />

[2] SCHULZE, H.; LUDERS, Ch. Theory and applications<br />

<strong>of</strong> OFDM and CDMA. John Wiley &<br />

Sons Ltd., Chichester, United Kingdom, 2005.<br />

ISBN-13 978-0-470-85069-5.<br />

[3] SPETH, M.; FECHTEL, S.; FOCK, G.; MEYR,<br />

H. Optimum Receiver Design for OFDM-Based<br />

Broadband Transmission-Part II: A Case Study.<br />

In IEEE Transactions on Communications, vol.<br />

49, issue 4, pages 571-578, 2001.<br />

[4] INDUSTRIAL TECHNOLOGY RESEARCH<br />

INSTITUTE. Time-frequency correlation-based<br />

synchronization for coherent OFDM receiver.<br />

Inventor: CHEN Ching-Yung, WANG Yi-Ting,<br />

HUNG Yung-Hua. United States <strong>of</strong> America.<br />

Patent Application Publication No. US<br />

2006/0088133 A1. 27. 4. 2006 [online]. [cit.<br />

25.4.2007]. Available .<br />

[5] FRESCURA, F.; PIELMEIER, S.; REALI, G.;<br />

BARUFFA, G.; CACOPARDI, S. DSP based<br />

OFDM demodulator and equalizer for pr<strong>of</strong>essional<br />

DVB-T receivers. In IEEE Transactions<br />

on Broadcasting, vol. 45, issue 3, pages 323-332,<br />

1999.


Transmembrane Potential Measurements on Plant Cells using <strong>the</strong><br />

Voltage-sensitive Dye ANNINE-6<br />

B. Flickinger, Th. Berghöfer, C. Eing, W. Frey<br />

Institute for Pulsed Power and Microwave Technology (IHM)<br />

Forschungszentrum Karlsruhe GmbH<br />

76344 Eggenstein-Leopoldshafen, Germany<br />

Email: bianca.flickinger@ihm.fzk.de<br />

Abstract — Recently, effects <strong>of</strong> pulsed electric<br />

fields on plant cells have become an interesting<br />

topic <strong>of</strong> research in plant electrophysiology.<br />

Systematic measurements using Nicotiana tabacum<br />

L. cv. Bright Yellow 2 (BY-2) protoplasts<br />

have been performed using a Pulsed Laser Fluorescence<br />

Microscopy (PLFM) setup with nanosecond<br />

time resolution. The field strength dependence<br />

<strong>of</strong> <strong>the</strong> protoplast’s transmembrane<br />

potential shows strong asymmetric saturation<br />

characteristics. To determine <strong>the</strong> azimuthal dependence<br />

<strong>of</strong> <strong>the</strong> transmembrane potential, <strong>the</strong><br />

relative fluorescence change has been measured<br />

in angular intervals <strong>of</strong> 10° along <strong>the</strong> circumference<br />

<strong>of</strong> <strong>the</strong> cell. Pulse-train experiments reveal<br />

that a subsequent re-charging <strong>of</strong> cell membranes<br />

to <strong>the</strong> same voltage is possible within a few seconds<br />

indicating a fast recovery process <strong>of</strong> <strong>the</strong> cell<br />

membrane.<br />

I. INTRODUCTION<br />

The application <strong>of</strong> an external electric field to<br />

biological cells leads to <strong>the</strong> polarisation <strong>of</strong> <strong>the</strong><br />

membrane. This charging process continues until<br />

<strong>the</strong> transmembrane potential reaches a critical<br />

value, where a fast rearrangement <strong>of</strong> <strong>the</strong> membrane’s<br />

molecular structure occurs, leading to <strong>the</strong><br />

formation <strong>of</strong> pores. This field-induced membrane<br />

effect is known as “electroporation”. The created<br />

pores lead to an increase <strong>of</strong> membrane permeability<br />

for ions and even macromolecules like DNA.<br />

Numerous studies with different approaches have<br />

been performed to obtain a better understanding <strong>of</strong><br />

<strong>the</strong> molecular processes involved in electroporation.<br />

Fur<strong>the</strong>rmore, <strong>the</strong>oretical models have been<br />

developed, e.g. <strong>the</strong> single cell electroporation<br />

model [1], to supply <strong>the</strong> experimental knowledge.<br />

But despite <strong>of</strong> all <strong>the</strong>se models little is known<br />

about <strong>the</strong> electroporation process itself. Therefore<br />

a better understanding <strong>of</strong> <strong>the</strong> charging processes <strong>of</strong><br />

biomembranes is needed. Pulsed laser fluorescence<br />

microscopy allows <strong>the</strong> examination <strong>of</strong><br />

9<br />

membrane charging processes with high temporal<br />

resolution <strong>of</strong> a few ns. This knowledge is <strong>of</strong> importance,<br />

because it enables <strong>the</strong> adaptation <strong>of</strong> <strong>the</strong><br />

electric pulse parameters for several applications<br />

ranging from nano- to industrial-scale processes.<br />

II. EXPERIMENTAL SETUP<br />

The experimental setup used for <strong>the</strong> presented<br />

experiments is equivalent to <strong>the</strong> one described by<br />

Frey et al. [2]. Protoplasts prepared from <strong>the</strong> cell<br />

line Nicotiana tabacum L. cv. Bright Yellow 2 (BY-<br />

2) have been stained with ANNINE-6, a fast voltage-sensitive<br />

dye which exhibits fluorescence<br />

intensity changes based on electronic transitions<br />

due to <strong>the</strong> molecular Stark effect within a subnanosecond<br />

time range [3]. For <strong>the</strong> excitation <strong>of</strong><br />

<strong>the</strong> fluorescent dye, a laser pulse with a wavlength<br />

<strong>of</strong> 468 nm delivered by a Nd:YAG-pumped dyelaser<br />

has been used. The voltage pulse driving <strong>the</strong><br />

external electric field is provided by a microscopebased<br />

Blumlein-line generator. Transmembrane<br />

potential dependent fluorescence changes caused<br />

by <strong>the</strong> applied external electric field have been<br />

recorded using a fast intensified CCD camera. To<br />

obtain <strong>the</strong> basic fluorescence F0, <strong>the</strong> fluorescence<br />

response <strong>of</strong> ANNINE-6 has been recorded without<br />

external electric field. For <strong>the</strong> measurement <strong>of</strong> <strong>the</strong><br />

field strength dependent fluorescence intensity F,<br />

an external electric field E has been applied and<br />

<strong>the</strong> fluorescence image has been acquired at a<br />

certain point <strong>of</strong> time t relative to <strong>the</strong> onset <strong>of</strong> <strong>the</strong><br />

pulse. The obtained relative fluorescence change<br />

F/F0 has been converted to transmembrane potential<br />

values �VM using a calibration curve <strong>of</strong> <strong>the</strong><br />

voltage-sensitive dye obtained by spectroscopic<br />

measurements on leech neuron cells [4].<br />

III. RESULTS AND DISCUSSION<br />

The experiments presented in this work have been<br />

performed using a 1 μs rectangular pulse with field<br />

strength amplitudes in a range <strong>of</strong> 0.05 kV/cm to


30 kV/cm. The recordings <strong>of</strong> <strong>the</strong> fluorescence intensity<br />

changes have been acquired at t = 500 ns after<br />

<strong>the</strong> onset <strong>of</strong> <strong>the</strong> electric field pulse.<br />

A. FIELD STRENGTH DEPENDENE OF THE TRANS-<br />

MEMBRANE POTENTIAL<br />

The field strength dependence <strong>of</strong> <strong>the</strong> BY-2 protoplasts<br />

transmembrane potential shows asymmetric<br />

saturation characteristics due to <strong>the</strong> high resting<br />

potential <strong>of</strong> plant cells. At <strong>the</strong> depolarized cell<br />

hemisphere (cathode) <strong>the</strong> fluorescence intensity<br />

decreases until an external electric field <strong>of</strong><br />

~1 kV/cm is reached. With fur<strong>the</strong>r increasing field<br />

strength <strong>the</strong> curve progression shows a clear saturation<br />

effect. At <strong>the</strong> hyperpolarized pole (anode) <strong>of</strong><br />

<strong>the</strong> protoplast, <strong>the</strong> saturation <strong>of</strong> <strong>the</strong> fluorescence<br />

intensity sets in at a lower field strength <strong>of</strong> approximately<br />

0.3 kV/cm.<br />

With <strong>the</strong> assumption <strong>of</strong> a resting potential value <strong>of</strong><br />

-150 mV <strong>the</strong> critical transmembrane potential<br />

where pore formation effects limit a fur<strong>the</strong>r charging<br />

<strong>of</strong> <strong>the</strong> membrane could be calculated to a value<br />

|VM| <strong>of</strong> ~ 350 mV.<br />

B. AZIMUTHAL DEPENDENCE OF THE TRANSMEM-<br />

BRANE POTENTIAL<br />

To determine <strong>the</strong> azimuthal dependence <strong>of</strong> <strong>the</strong><br />

transmembrane potential, <strong>the</strong> relative fluorescence<br />

change has been measured in angular intervals <strong>of</strong><br />

10° along <strong>the</strong> circumference <strong>of</strong> <strong>the</strong> protoplast.<br />

The curve progressions show membrane charging<br />

up to values similar to <strong>the</strong> results obtained by <strong>the</strong><br />

method described in (A.). The flattening and <strong>the</strong><br />

drop <strong>of</strong> <strong>the</strong> curves at <strong>the</strong> depolarized cell hemisphere<br />

at field strengths > 1kV/cm is a clear sign<br />

for <strong>the</strong> limiting <strong>of</strong> <strong>the</strong> membrane charging by <strong>the</strong><br />

formation <strong>of</strong> pores. Due to <strong>the</strong> high resting potential,<br />

<strong>the</strong> pore formation at <strong>the</strong> hyperpolarized cell<br />

pole starts earlier and <strong>the</strong> curves do not show <strong>the</strong><br />

same amplitude compared to <strong>the</strong> cathodic cell<br />

pole. At an external electric field strength <strong>of</strong><br />

~ 1 kV/cm, a statistically verified membrane effect<br />

occuring at <strong>the</strong> hyperpolarized pole could be<br />

observed. Recent <strong>the</strong>ories explain <strong>the</strong> sudden drop<br />

<strong>of</strong> <strong>the</strong> transmembrane potential with an activation<br />

<strong>of</strong> voltage-gated ion channels. The existence <strong>of</strong><br />

hyperpolarisation-activated Ca 2+ -channels plant<br />

cells, which control <strong>the</strong> membrane depolarisation<br />

needed for physiological functions, has been confirmed<br />

by several authors [5].<br />

C. RECHARGING BEHAVIOUR OF THE MEMBRANE<br />

AFTER PULSE EXPOSITION<br />

Pulse train experiments have been performed to<br />

determine <strong>the</strong> charging- and recharging behaviour<br />

10<br />

<strong>of</strong> <strong>the</strong> protoplasts plasmalemma after multiple<br />

pulse exposition at several field strengths. The<br />

results show, that <strong>the</strong> membrane discharges instantaneously<br />

after <strong>the</strong> pulse ends to values which are<br />

supposed to be <strong>the</strong> resting potential <strong>of</strong> <strong>the</strong> cell. The<br />

membrane can be recharged to <strong>the</strong> same value a<br />

few seconds after <strong>the</strong> previous pulse exposition. At<br />

low field strengths both cell poles show <strong>the</strong> same<br />

amplitude <strong>of</strong> <strong>the</strong> induced transmembrane potential.<br />

With increasing field strength, <strong>the</strong> depolarized cell<br />

pole shows higher amplitudes compared to <strong>the</strong><br />

hyperpolarized side. Application <strong>of</strong> electric fields<br />

higher than approximately 1 kV/cm leads to disruption<br />

<strong>of</strong> <strong>the</strong> charging- and discharging-processes<br />

at <strong>the</strong> hyperpolarized cell pole indicated by <strong>the</strong><br />

saturation characteristics <strong>of</strong> <strong>the</strong> curve progression.<br />

The results <strong>of</strong> <strong>the</strong> pulse-train experiments show a<br />

fast recovery process <strong>of</strong> <strong>the</strong> plasmalemma with <strong>the</strong><br />

depolarized pole showing characteristics <strong>of</strong> <strong>the</strong><br />

formation <strong>of</strong> reversible pores whereas <strong>the</strong> curve<br />

progression measured at <strong>the</strong> hyperpolarized pole<br />

indicates <strong>the</strong> formation <strong>of</strong> irreversible pores.<br />

ACKNOWLEDGMENTS<br />

Thanks to Pr<strong>of</strong>. Dr. Peter Nick and Dr. Petra<br />

Hohenberger from <strong>the</strong> Botanical Institute I,<br />

University Karlsruhe for <strong>the</strong>ir assistance.<br />

Faculty mentor: Pr<strong>of</strong>. Dr. Peter Nick<br />

Email:<br />

peter.nick@uni-karlsruhe.de<br />

REFERENCES<br />

[1] W. Krassowska, P.D. Filev, Modeling Electroporation<br />

in a Single Cell, Biophysical Journal,<br />

Vol. 92, 404-417, January 2007.<br />

[2] W. Frey, J.A. White, R.O. Price, P.F. Blackmore,<br />

R.P. Joshi, R. Nuccitelli, S.J. Beebe,<br />

K.H. Schoenbach and J.F. Kolb, Plasma<br />

Membrane Voltage Changes during Nanosecond<br />

Pulsed Electric Field Exposure, Biophysical<br />

Journal, Vol. 90, 3608-3615, May 2006.<br />

[3] B. Kuhn, P. Fromherz, W. Denk, High Sensitivity<br />

<strong>of</strong> Stark-Shift Voltage-Sensing Dyes by<br />

One- or Two-Photon Excitation Near <strong>the</strong> Red<br />

Spectral Edge, Biophysical Journal, Vol. 87,<br />

631-639, July 2004.<br />

[4] B. Kuhn, P. Fromherz, Annelated Hemicyanine<br />

Dyes in a Neuron Membrane: Molecular<br />

Stark Effect and Optical Voltage Recording, J.<br />

Phys. Chem. B, 107, 7903-7913, 2003.<br />

[5] P.J. White, Calcium channels in higher plants,<br />

Biochimica et Biophysica Acta 1465, 171-<br />

189, 2000.


Modeling metamaterial microwave resonators<br />

Tomas Zvolensky and Zbynek Raida (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronics<br />

University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: xzvole00@stud.feec.vutbr.cz, raida@feec.vutbr.cz<br />

Abstract — The work is aimed to verify <strong>the</strong> possibility<br />

<strong>of</strong> modeling, realization and measurement <strong>of</strong><br />

a microwave resonator comprised <strong>of</strong> a metamaterial<br />

equivalent structure. The resonator is designed<br />

in a planar form to be easily implemented in multilayered<br />

planar circuits. For <strong>the</strong> resonator design,<br />

MATLAB scripts for calculating <strong>the</strong> dimensions <strong>of</strong><br />

<strong>the</strong> planar components <strong>of</strong> <strong>the</strong> resonator were developed.<br />

After simulating <strong>the</strong> designed structure in<br />

Zeland IE3D, <strong>the</strong> metamaterial resonator was fabricated<br />

and measured.<br />

I. INTRODUCTION<br />

Throughout last decades, researchers have been<br />

dealing with metamaterials and <strong>the</strong>ir practical implementation<br />

in many fields <strong>of</strong> engineering areas. In<br />

principle, metamaterials are composed media having<br />

various forms (layered media, planar structures,<br />

space neatened structures, etc.). At some frequencies,<br />

metamaterials possess negative values <strong>of</strong> material<br />

constants (i.e., relative permittivity and relative permeability).<br />

Having negative material constants values,<br />

several interesting characteristics can arise (e.g.,<br />

wave propagating through <strong>the</strong> metamaterial structure<br />

can have <strong>the</strong> same phase shift at <strong>the</strong> end <strong>of</strong> <strong>the</strong> structure<br />

as at <strong>the</strong> entrance).<br />

Our interest is focused on using metamaterials in<br />

microwave resonators design. In order to verify <strong>the</strong><br />

characteristics <strong>of</strong> metamaterials, planar resonators<br />

were designed, simulated and measured.<br />

II. SUB-WAVELENGTH CAVITY RESONATOR<br />

Figure 1 shows geometry <strong>of</strong> a two-layer resonator<br />

[1]. The layer <strong>of</strong> <strong>the</strong> thickness d1 is made <strong>of</strong><br />

a conventional material (positive values <strong>of</strong> material<br />

constants), and <strong>the</strong> layer <strong>of</strong> <strong>the</strong> thickness d2 is made<br />

<strong>of</strong> metamaterial. Arrows S1 and S2 show <strong>the</strong> directions<br />

<strong>of</strong> <strong>the</strong> Poyinting vector, arrows k1 and k2 are<br />

directions <strong>of</strong> <strong>the</strong> phase velocity vector.<br />

Metamaterials are left-handed (LH) media, and<br />

conventional materials are right-handed (RH) ones.<br />

Therefore, <strong>the</strong> direction <strong>of</strong> <strong>the</strong> Poyinting vector and<br />

<strong>the</strong> direction <strong>of</strong> <strong>the</strong> phase velocity vector are antiparallel<br />

in <strong>the</strong> metamaterial layer.<br />

11<br />

Resonators designed and investigated in this project<br />

are composed <strong>of</strong> a LH transmission line (TL) and<br />

a RH TL. These resonators are <strong>the</strong>refore called<br />

CRLH TL (Composite Left/Right Handed Transmission<br />

Lines).<br />

When designing such a resonator in <strong>the</strong> planar<br />

form, LH TL is created by series capacitors and shunt<br />

inductors. The resonator exhibits <strong>the</strong>n a nonlinear<br />

phase response. The phase response can be shaped<br />

according to our requirements by changing <strong>the</strong> properties<br />

<strong>of</strong> transmission lines (adjusting RH TL length<br />

and changing inductance and capacity <strong>of</strong> LH TL).<br />

Figure 1: Two-layer resonator [1]<br />

Better characteristics <strong>of</strong> <strong>the</strong> planar resonator (i.e.,<br />

higher modes spacing) can reached by decreasing <strong>the</strong><br />

slope <strong>of</strong> <strong>the</strong> phase response. By increasing <strong>the</strong> slope<br />

<strong>of</strong> <strong>the</strong> phase response, <strong>the</strong> Q factor <strong>of</strong> <strong>the</strong> resonator<br />

can be increased. The advantage <strong>of</strong> this technique is<br />

hidden in <strong>the</strong> fact that changing <strong>the</strong> phase response<br />

influences both <strong>the</strong> essential parameters <strong>of</strong> resonators<br />

– <strong>the</strong> selectivity and <strong>the</strong> higher modes spacing.<br />

III. CIRCUIT MODEL VERIFICATION<br />

Figure 2 shows a planar resonator setup created<br />

by <strong>the</strong> modification <strong>of</strong> <strong>the</strong> structure published in [2].<br />

Gap capacitors used in [2] were replaced interdigital<br />

ones in order to reach higher values <strong>of</strong> <strong>the</strong> series<br />

capacities. Parallel inductivities are implemented in


<strong>the</strong> form <strong>of</strong> short-end shunts. In between shunts, RH<br />

part <strong>of</strong> <strong>the</strong> TL is placed.<br />

Figure 2: Planar resonator setup<br />

In order to obtain proper dimensions <strong>of</strong> <strong>the</strong> shunt<br />

inductor and <strong>the</strong> interdigital capacitor, MATLAB<br />

scripts were implemented. The scripts used a local<br />

optimization algorithm, which changed <strong>the</strong> dimensions<br />

<strong>of</strong> <strong>the</strong> layout to reach <strong>the</strong> required capacity and<br />

inductivity.<br />

The described design procedure was applied considering<br />

<strong>the</strong> dielectric substrate Arlon D600 (<strong>the</strong><br />

height h = 1.54 mm, <strong>the</strong> dielectric constant �r = 6.15,<br />

and <strong>the</strong> loss factor tan � = 0.003).<br />

When <strong>the</strong> dimensions <strong>of</strong> <strong>the</strong> resonator elements are<br />

known, <strong>the</strong> structure can be modeled and optimized<br />

in Zeland IE3D. The optimization is aimed to obtain<br />

<strong>the</strong> dominant resonance at frequency 6 GHz.<br />

0<br />

-5<br />

-10<br />

-15<br />

-20<br />

-25<br />

-30<br />

s11[dB]<br />

0<br />

-10<br />

-20<br />

-30<br />

-40<br />

-50<br />

-60<br />

-70<br />

-80<br />

f[GHz]<br />

4 4,5 5 5,5 6 6,5 7 7,5 8<br />

measured simulated<br />

f[GHz]<br />

4 4,5 5 5,5 6 6,5 7 7,5 8<br />

s21[dB] measured simulated<br />

Figure 3: Measurement and simulation comparison<br />

12<br />

Consequently, <strong>the</strong> optimized resonator was fabricated<br />

and measured. Comparison <strong>of</strong> <strong>the</strong> simulation<br />

and <strong>the</strong> measurement <strong>of</strong> <strong>the</strong> designed structure are<br />

depicted in Figure 3.<br />

Obviously, <strong>the</strong>re is a single lower mode resonating<br />

below frequency 5.4 GHz, and <strong>the</strong>re are a number <strong>of</strong><br />

upper modes resonating above frequency 5.4 GHz.<br />

A detailed comparison shows that:<br />

� The frequency <strong>of</strong> maximum transmission for <strong>the</strong><br />

measured structure is at 6.3 GHz (s21 =–8.05 dB)<br />

and at 6.08 GHz for <strong>the</strong> simulated one (s21 =<br />

–7.94 dB). The drift <strong>of</strong> <strong>the</strong> resonance frequency<br />

is about 3.6 %.<br />

� The frequency <strong>of</strong> <strong>the</strong> minimum reflections is at<br />

5.89 GHz for <strong>the</strong> measured structure (s11 =<br />

–25.12 dB) and at 6.09 GHz for <strong>the</strong> simulated<br />

one (s11 = –27.11 dB). The drift <strong>of</strong> resonance<br />

frequency is about 3.3 %. At <strong>the</strong> frequency<br />

6 GHz (<strong>the</strong> expected resonance), reflections <strong>of</strong><br />

<strong>the</strong> measured structure reach –18.1 dB, <strong>the</strong>refore<br />

impedance matching can be considered as satisfactory.<br />

� The drift <strong>of</strong> <strong>the</strong> resonant frequency can be caused<br />

<strong>of</strong> <strong>the</strong> <strong>the</strong> fabrication <strong>of</strong> <strong>the</strong> structure. An extension<br />

<strong>of</strong> vias through <strong>the</strong> metal joint (moreover<br />

with a different impedance) modified parameters<br />

<strong>of</strong> <strong>the</strong> whole structure and shifted <strong>the</strong> resonance.<br />

IV. CONCLUSIONS<br />

In <strong>the</strong> paper, <strong>the</strong> design, modeling, fabrication<br />

and measurement <strong>of</strong> <strong>the</strong> metamaterial planar resonator<br />

is described. Obtained results are discussed<br />

demonstrating a relatively good agreement.<br />

ACKNOWLEDGEMENTS<br />

The described work was financially supported<br />

by <strong>the</strong> Czech Grant Agency by grants 102/07/0688<br />

and 102/08/H018.<br />

REFERENCES<br />

[1] ENGHETA, N. An idea for thin sub-wavelength<br />

cavity resonators using meta-materials with<br />

negative permittivity and permeability. IEEE Antennas<br />

and Wireless Propagation Letters. 2002,<br />

vol. 1, no. 1, p. 10–13.<br />

[2] ALLEN, C.A., LEONG, K., ITOH, T. Design <strong>of</strong><br />

microstrip resonators using balanced and unbalanced<br />

composite right/left-handed transmission<br />

lines. IEEE Transactions on Microwave Theory<br />

and Techniques. 2006, vol. 54, no. 7, p. 3104 to<br />

3112.


Bearingless Segment Motor with Buried Magnets<br />

Thomas Stallinger, Wolfgang Gruber and Wolfgang Amrhein<br />

Institute <strong>of</strong> Electrical Drives and Power Electronics<br />

Johannes Kepler University<br />

Linz, Austria<br />

Email: {Thomas.Stallinger,Wolfgang.Gruber,Wolfgang.Amrhein}@jku.at<br />

Abstract — Bearingless motors combine contactless<br />

levitation and rotation in a preferably compact<br />

system design, because bearing as well as motor<br />

windings are located on <strong>the</strong> same lamination stack.<br />

The bearingless segment motor features concentrated<br />

coils on separated stator elements, which<br />

reduce <strong>the</strong> stator iron, weight and cost, especially<br />

for constructions with large diameters. In this paper<br />

a bearingless segment motor with buried permanent<br />

magnets is investigated. The ma<strong>the</strong>matical model is<br />

presented and a proper control scheme is introduced.<br />

Finally, <strong>the</strong> performance <strong>of</strong> <strong>the</strong> bearingless segment<br />

motor with buried permanent magnets is proven by<br />

<strong>the</strong> comparison <strong>of</strong> simulation results with measurement<br />

data <strong>of</strong> <strong>the</strong> manufactured prototype.<br />

I. INTRODUCTION<br />

The advantages <strong>of</strong> bearingless drives are evident.<br />

Contactless levitation leads to wearless and lubrication<br />

free operation. Fur<strong>the</strong>rmore, <strong>the</strong> rotating part and<br />

<strong>the</strong> stator can be separated completely by <strong>the</strong> use <strong>of</strong> a<br />

rotor can. This leads to hermetically sealed systems<br />

which can satisfy high demands on cleanliness,<br />

chemical resistance and tightness. Therefore such<br />

motors are <strong>of</strong>ten employed in <strong>the</strong> pharmaceutical,<br />

biomedical, chemical, semiconductor or aerospace<br />

industry.<br />

Bearingless systems have to stabilize all six degrees<br />

<strong>of</strong> freedom by magnetic forces. In general <strong>the</strong>se<br />

systems show a quite complex build-up. The bearingless<br />

slice motor [1] reduces <strong>the</strong> system complexity<br />

because three degrees <strong>of</strong> freedom (two tilting and <strong>the</strong><br />

axial direction) are passively stabilized by means <strong>of</strong><br />

reluctance forces. This becomes feasible due to <strong>the</strong><br />

permanent magnet excited disc shaped rotor. In <strong>the</strong><br />

recent past <strong>the</strong> bearingless segment motor, illustrated<br />

in Figure 1, was introduced [2] as a subtype <strong>of</strong> <strong>the</strong><br />

bearingless slice motor.<br />

II. CONTROL SCHEME<br />

To simplify matters only <strong>the</strong> model for current injection<br />

is given here. The detailed deduction and <strong>the</strong><br />

more complex model for voltage injection can be<br />

found in [4].<br />

13<br />

Figure 1: Drawing <strong>of</strong> a bearingless segment motor<br />

A. FORCE AND TORQUE<br />

A describtion <strong>of</strong> <strong>the</strong> bearing forces (Fx, Fy) and <strong>the</strong><br />

motor torque (Mz) <strong>of</strong> <strong>the</strong> system is given in [3] by<br />

T<br />

� F � � 0 0<br />

x x � �i� r<br />

s<br />

� T<br />

Fy<br />

�x � � � � 0 i 0 �<br />

r<br />

s MQ�xr�is<br />

� � T �<br />

(1)<br />

M z � ��<br />

� x � � 0 0 i<br />

r<br />

s �<br />

�M x i �M<br />

x .<br />

� � � �<br />

L r s N r<br />

Here xr��xr� yr� �r) T represents <strong>the</strong> rotor position by<br />

<strong>the</strong> radial deflection and rotor angle. The vector <strong>of</strong><br />

<strong>the</strong> stator currents is=(is1, is2, is3, … ,ism) T is composed<br />

<strong>of</strong> <strong>the</strong> m phase currents. There is a quadratic (MQ), a<br />

linear (ML) and a term <strong>of</strong> no dependency (MN) concerning<br />

<strong>the</strong> stator currents. Equation (1) can be linearized<br />

via Taylor series approximation and yields to<br />

� Fx<br />

�� � � r<br />

� Fy���<br />

� � T � � � , � � ,<br />

r m �r is �Tc<br />

�r is � Kx �r<br />

iS� r<br />

�<br />

M z ��r�� � �<br />

x . (2)<br />

B. OVERALL MODEL<br />

Adding <strong>the</strong> mechanical equations <strong>of</strong> motion <strong>the</strong> overall<br />

model <strong>of</strong> <strong>the</strong> system in (2) is given by 1<br />

1<br />

�<br />

x��� �M O<br />

K i<br />

E<br />

O<br />

� � O<br />

�x�� � �M<br />

T<br />

� O3×1<br />

�<br />

� � -1 ,<br />

��r, ��<br />

�M Tc iS<br />

�<br />

On× m<br />

E n<br />

�<br />

�<br />

� i<br />

3×3 3 3×m<br />

-1 -1 S<br />

x��r, S� 3×3 m��r�<br />

(3)<br />

represents a zero matrix <strong>of</strong> <strong>the</strong> dimension n × m and<br />

stands for an identity matrix <strong>of</strong> <strong>the</strong> dimension n.


with is as <strong>the</strong> actuating variable. The state vector x is<br />

defined as<br />

x � � x y � x� y�<br />

� �<br />

r r r r r r<br />

T<br />

. (4)<br />

In this equation <strong>of</strong> motion M represents <strong>the</strong> diagonal<br />

mass matrix and Kx(�r) stands for <strong>the</strong> stiffness matrix,<br />

Tm(�r)� is called current-force matrix and<br />

Tc(�r,iS) is <strong>the</strong> reluctance matrix.<br />

C. CONTROL SCHEME<br />

For <strong>the</strong> deduced nonlinear model in (3) a proper<br />

control scheme has to be found. A possible choice is<br />

to use nonlinear feedback techniques [4]. Utilizing<br />

<strong>the</strong> pseudo inverse matrix <strong>of</strong> Tm(�r), which is also<br />

called force-current matrix Km(�r), it is possible to<br />

compensate all nonlinearities with an bias current<br />

vector, depending on <strong>the</strong> rotor angle and <strong>the</strong> current<br />

vector itself, leading to an overall linear and decoupled<br />

model<br />

� O E � � O<br />

x�� x��M T<br />

3×3 3<br />

3×m<br />

� -1 �<br />

-1<br />

S<br />

�M K � � �� x O3×3<br />

m r ��<br />

�<br />

III. OPTIMIZATION<br />

�<br />

i . (5)<br />

Finite element simulations were carried out to optimize<br />

bearing forces and motor torque and to minimize<br />

<strong>the</strong> reluctance forces and <strong>the</strong> cogging torque.<br />

Because <strong>of</strong> calculating capacity and time economy<br />

<strong>the</strong>se optimizations were done in 2D by Femag� and<br />

<strong>the</strong>n verified by Maxwell 3D�. Both <strong>the</strong> rotor and<br />

<strong>the</strong> stator geometry were optimized.<br />

IV. MANUFACTURED MOTOR<br />

To evaluate <strong>the</strong> functionality <strong>of</strong> <strong>the</strong> motor and <strong>the</strong><br />

quality <strong>of</strong> <strong>the</strong> optimization a prototype was built,<br />

visible in Figure 2. The five stator elements (each<br />

holding a concentrated winding coil) are held among<br />

an aluminium framework. Inside <strong>the</strong> rotor <strong>the</strong> buried<br />

permanent magnets are cognizable, forming six magnetic<br />

pole pairs. The used Ni-coated NdFeB magnets<br />

feature a remanence field density BBr <strong>of</strong> 1.2T. The<br />

stator and rotor iron consist <strong>of</strong> laminated V270-35A<br />

steel plates.<br />

Figure 2: Bearingless segment motor prototype<br />

14<br />

The control scheme was implemented on <strong>the</strong> inexpensive<br />

16 bit fixed point digital signal processor<br />

family TMS320 by Texas Instruments. The s<strong>of</strong>tware<br />

is written in ANSI C-code. The design <strong>of</strong> a proper<br />

electronic to evaluate <strong>the</strong> position- and <strong>the</strong> Hallsensors<br />

was also part <strong>of</strong> this work.<br />

V. MEASUREMENTS<br />

An orbit measurement was done while <strong>the</strong> bearingless<br />

segment motor was in operation. The radial<br />

deflection at certain speeds was measured and displayed<br />

in Figure 3. More measurement data is also<br />

given in [5].<br />

Figure 3: Bearingless motor in action - orbit<br />

ACKNOWLEDGMENTS<br />

This research project was funded by <strong>the</strong> Austrian Science<br />

Fund (FWF) under contract P17523-N07. The authors<br />

thank <strong>the</strong> Austrian Government for <strong>the</strong> support. Scientific<br />

advisory support and parts <strong>of</strong> <strong>the</strong> control algorithms were<br />

kindly given by <strong>the</strong> Austrian Center <strong>of</strong> Competence in<br />

Mechatronics (ACCM), a K2 Centre <strong>of</strong> <strong>the</strong> COMET program<br />

<strong>of</strong> <strong>the</strong> Austrian Government.<br />

REFERENCES<br />

[1] Schöb R. and Barletta N., Principle and application<br />

<strong>of</strong> a bearingless slice motor, Proc. <strong>of</strong> <strong>the</strong> 5th<br />

Int. Symp. on Magnetic Bearings (ISMB5), Kanazawa,<br />

Japan, 1996<br />

[2] Gruber W. and Amrhein W., Design <strong>of</strong> a Bearingless<br />

Segment Motor, Proc. <strong>of</strong> <strong>the</strong> 10th Int.<br />

Symp. on Magnetic Bearings (ISMB10), Martigny,<br />

Switzerland, 2006.<br />

[3] Amrhein W., Silber S. and Nenninger K., Levitation<br />

forces in bearingless perm. magnet rotors,<br />

IEEE Trans. on Magnetics, Vol. 35, No. 5, 1996.<br />

[4] Grabner H., Dynamik und Ansteuerkonzepte<br />

lagerloser Drehfeld-Scheibenläufermotoren in<br />

radialer Bauform, Ph. D. Dissertation, Johannes<br />

Kepler University Linz, Trauner Verlag, 2007.<br />

[5] Stallinger T., Lagerloser Segmentmotor mit eingebetteten<br />

Magneten, Diplomarbeit, Johannes<br />

Kepler University Linz, <strong>2008</strong>.


Identification <strong>of</strong> tacit knowledge during realization <strong>of</strong> knowledge<br />

creation processes<br />

František Babi� and Ján Parali� (Faculty Mentor)<br />

Department <strong>of</strong> Cybernetics and Artificial Intelligence<br />

Faculty <strong>of</strong> Electrical Engineering and Informatics, Technical University <strong>of</strong> Košice<br />

Košice, Slovakia<br />

Email: {frantisek.babic, jan.paralic}@tuke.sk<br />

Abstract — The aim <strong>of</strong> this paper is to present<br />

proposed solution for identification <strong>of</strong> tacit knowledge<br />

during knowledge creation processes. The<br />

secondary knowledge arise based on actual situations<br />

and have potential to simplify and solve <strong>the</strong><br />

problematic points in performed processes.<br />

This approach is based on saving performed actions<br />

and activities within processes in form <strong>of</strong> logs into<br />

separate database in predefined format. These logs<br />

<strong>of</strong> events will be fur<strong>the</strong>r analyzed with e.g. Data<br />

Mining methods, SNA methods, statistical methods,<br />

etc.<br />

Events are realized in collaborative system called<br />

KP-Lab System that has been implemented within<br />

integrated IST FP6 project called KP-Lab 1 (2006-<br />

2011).<br />

I. INTRODUCTION<br />

Knowledge creation processes are modelled and<br />

realized in order to acquire new knowledge; this is<br />

<strong>the</strong> main goal <strong>of</strong> <strong>the</strong>se processes. Process models<br />

consist <strong>of</strong> phases and some detail parts, inputs and<br />

resources, participants with assigned roles, etc. During<br />

process execution many interactions between<br />

<strong>the</strong>ses aspects can emerged and it can be reason for<br />

contradictions, deviations and exceptions. In various<br />

cases are <strong>the</strong>se situation evaluated and negative and<br />

several solution are predefined. In our case, we<br />

evaluate <strong>the</strong>se points in process as potentially place<br />

<strong>of</strong> change with possible tacit knowledge.<br />

II. LOG-BASED AWARENESS<br />

Proposed solution is marked as Log-based (History)<br />

awareness (LBA) and provides an asynchronous<br />

functionality that builds on persistent storage <strong>of</strong><br />

events representing activities performed by users <strong>of</strong><br />

various end-user tools in KP-Lab System and based<br />

on <strong>the</strong>m supports analysis and discovery <strong>of</strong> tacit<br />

knowledge from previous activities.<br />

1 http://www.kp-lab.org/<br />

15<br />

KP-Lab System is virtual collaborative system that<br />

provides many important and interesting features to<br />

support collaborative knowledge creation processes.<br />

In this system users/participants can realized <strong>the</strong>ir<br />

activities, e.g. creation <strong>of</strong> process models with fur<strong>the</strong>r<br />

execution, working with shared objects <strong>of</strong> activities,<br />

multimedia annotation (document, video, etc.) based<br />

on free or ontology approach (vocabularies), conferencing<br />

features with creation <strong>of</strong> discussion map and<br />

outcomes export, collaborative semantic modelling,<br />

etc. KP-Lab System use ontologies as common<br />

communication and integration language for all applications<br />

and tools that are implemented and integrated<br />

in whole system. Semantic information (metadata,<br />

ontologies) are stored into knowledge repository<br />

within RDFSuite and content storage is implemented<br />

based on Jackrabbit and JSR-170 technologies.<br />

A. ARCHITECTURE<br />

LBA is integrated on <strong>the</strong> platform level under end<br />

user tool layer and provides several related services,<br />

see Figure 1.<br />

Figure 1: Integration <strong>of</strong> LBA related services into<br />

whole system<br />

Subscription service is responsible for a creation,<br />

modification and a removal <strong>of</strong> <strong>the</strong> subscriptions<br />

which keeps in repository. This module must communicate<br />

both with <strong>the</strong> users (end-user tools) to receive<br />

<strong>the</strong>ir choices or updates <strong>of</strong> <strong>the</strong>ir subscriptions


and with <strong>the</strong> repository in order to allow a persistent<br />

storage <strong>of</strong> <strong>the</strong>se subscriptions.<br />

Role <strong>of</strong> <strong>the</strong> Notification service is to deliver <strong>the</strong> notifications<br />

to <strong>the</strong> corresponding users based on selected<br />

form: email, RSS, online message, etc.<br />

Repository is implemented as transactional database<br />

to store subscription and events for fur<strong>the</strong>r analyses.<br />

LBA services have support analysis and discovery<br />

<strong>of</strong> (possibly tacit) knowledge from previous activities.<br />

B. INTEGRATION WITH KPS<br />

The whole process <strong>of</strong> data collection and analysis can<br />

be shown on an example, using Knowledge Process<br />

Service (KPS), and see Figure 2.<br />

Figure 2: Communication between KPS and envisaged<br />

LBA services [1]<br />

KPS provides a set <strong>of</strong> features and interfaces necessary<br />

for creation, management, and annotation <strong>of</strong><br />

processes composed from various elements. This tool<br />

can generate several types <strong>of</strong> events as Modification<br />

<strong>of</strong> object metadata, Modification <strong>of</strong> object itself<br />

(create, delete) and Modification in process structure.<br />

If e.g. a user wants to be notified about changes in a<br />

particular Task (this means any change <strong>of</strong> Task properties<br />

metadata), <strong>the</strong> process is <strong>the</strong> following (see<br />

Figure 2). The user subscribes via KPS his/her interest<br />

to be notified when a particular Task changes<br />

(Subscription <strong>of</strong> Task changes in Figure 2). This<br />

subscription will be stored in <strong>the</strong> repository (Saving<br />

in database). Whenever any user makes a modification<br />

<strong>of</strong> this particular task, KPS issues an event to be<br />

logged (Modification <strong>of</strong> Task - Event) into repository.<br />

A matching algorithm will match each new event<br />

with all current subscriptions (see Figure 2) and all<br />

matching subscriptions (result <strong>of</strong> Queries in Figure<br />

2) will lead to notifications (Notification <strong>of</strong> realized<br />

event) to all <strong>the</strong>ir subscribers.<br />

Moreover, this event will be taken into account<br />

whenever a relevant analysis <strong>of</strong> logged events will be<br />

16<br />

performed by a user or a tool (Perform History Analysis,<br />

Queries, and Provide results <strong>of</strong> Analysis in<br />

Figure 2).<br />

C. ANALYTICAL APPROACH<br />

The analytical services have been proposed with<br />

several levels <strong>of</strong> complexity and granularity. For a<br />

start we implement basic features as identification <strong>of</strong><br />

all changes within shared objects, list <strong>of</strong> performed<br />

actions by users, performed actions in selected time<br />

interval and some combinations. In <strong>the</strong> next step,<br />

based on fur<strong>the</strong>r acquired logs <strong>of</strong> events, additional<br />

functionalities will be implemented as:<br />

• Visualisation <strong>of</strong> <strong>the</strong> user’s activities (e.g. identification<br />

<strong>of</strong> most active participant, identification<br />

<strong>of</strong> isolated persons in collaborative processes).<br />

• Sequence mining: identification <strong>of</strong> typical sequences<br />

<strong>of</strong> actions that users execute to reach <strong>the</strong><br />

goals or solve <strong>the</strong> problem.<br />

• Identification <strong>of</strong> past patterns – e.g. participants<br />

used several relevant documents for accomplishment<br />

<strong>of</strong> similar type <strong>of</strong> task, so we can recommend<br />

which documents are suitable for actual<br />

task instance.<br />

• Support for building <strong>the</strong> user and task models for<br />

Help services.<br />

• Identification <strong>of</strong> changes in vocabularies, e.g. list<br />

<strong>of</strong> last changes.<br />

• Discovery <strong>of</strong> “best practices” in past knowledge<br />

processes. Based on <strong>the</strong>m new templates <strong>of</strong><br />

various types may be created and stored into Reuse<br />

library.<br />

• Identification <strong>of</strong> objects that were subject <strong>of</strong><br />

(triggered) context-based chat – maybe problematic<br />

part <strong>of</strong> process that needs necessary discussion<br />

about next step, progress or solution.<br />

• Statistics <strong>of</strong> ToDo items creation – evaluation <strong>of</strong><br />

individual work management (user with many<br />

items may be more interested in <strong>the</strong> whole process<br />

as user with only few items).<br />

III. CONCLUSION<br />

LBA services have provided necessary functionalities<br />

for identification <strong>of</strong> tacit knowledge in various<br />

formats to improve knowledge creation processes.<br />

REFERENCES<br />

[1] F. Babi� and J. Wagner. Awareness service based<br />

on events logs. 2nd Workshop on Intelligent and<br />

Knowledge oriented Technologies 2007, Košice,<br />

Slovakia, pages 106-109, <strong>2008</strong>, ISBN 978-80-<br />

89284-10-8.


Multifunctional 4 th order tuned active filter<br />

Roman Šotner, Tomáš Dostál (Faculty Mentor)<br />

Department <strong>of</strong> Radioelectronics<br />

Faculty <strong>of</strong> Electrical Engineering and Communications, University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: xsotne00@stud.feec.vutbr.cz, dostal@feec.vutbr.cz<br />

Abstract — The paper deals about design <strong>of</strong><br />

<strong>the</strong> multifunctional active filter with operational<br />

transconductance amplifiers (OTA-s),<br />

which realize low pass (LP), band pass (BP),<br />

high pass (HP) and band reject (BR) transfer<br />

functions. Mentioned filter is studied with several<br />

circuits models, namely third level based<br />

on voltage controlled current sources and pr<strong>of</strong>essional<br />

macro models <strong>of</strong> commercially available<br />

integrated circuits. The circuit is simulated<br />

by PSpice (OrCad) and adjusting parameters<br />

are discussed. Simulations are compared<br />

with experimental results.<br />

I. INTRODUCTION<br />

Some modern integrated active blocks have<br />

possibility to electronic control <strong>of</strong> <strong>the</strong>ir parameters.<br />

It is possible to used this principle on adjustable<br />

filters and oscillators [1], [3]. Typical<br />

block is operational transconductance amplifier,<br />

for example LT 1228 [5].<br />

II. CIRCUIT DESIGN<br />

The proposed autonomous circuit in Figure 1<br />

was obtained with <strong>the</strong> help <strong>of</strong> symbolic analysis<br />

by Snap [2].<br />

Figure 1: Autonomous circuit<br />

Characteristic equation has <strong>the</strong> following form<br />

Y YYY<br />

�YYY g �YY<br />

g g �Yg<br />

g g �g<br />

g g g �0.<br />

(1)<br />

1 2 3 4<br />

1 2 3<br />

4<br />

1 2<br />

3<br />

4<br />

1 2<br />

3<br />

4<br />

1 2<br />

If Y1 = sC1, Y2 = sC2, Y3 = sC3, Y4 = sC4,<br />

4 3 g4<br />

2 g3g4<br />

2 g2g3g<br />

4 g1g2g<br />

3g4<br />

D( s)<br />

� s � s � s � s � , (2)<br />

C C C C C C C C C C<br />

4<br />

3<br />

4<br />

2<br />

3<br />

4<br />

1<br />

3<br />

2<br />

4<br />

3<br />

4<br />

17<br />

and corresponding circuit diagram is shown in<br />

Figure 2.<br />

Figure 2: Multifunctional 4 th order filter<br />

The transfer functions are<br />

K BR<br />

g1g<br />

2 g 3g<br />

4<br />

K LP ( s)<br />

� , (3)<br />

D(<br />

s)<br />

2<br />

s C 1C<br />

2 g 3g<br />

4<br />

K BP ( s)<br />

� , (4)<br />

D(<br />

s)<br />

4<br />

s C1C<br />

2C<br />

3C<br />

4<br />

K HP ( s)<br />

� , (5)<br />

D(<br />

s)<br />

4<br />

s C1C<br />

2C<br />

2C<br />

4 g1g<br />

2 g 3g<br />

4<br />

( s)<br />

� � . (6)<br />

D(<br />

s)<br />

D(<br />

s)<br />

Inputs, which are not used, must be grounded.<br />

The filter is designed with following specification:<br />

<strong>the</strong> cut-<strong>of</strong>f (center) frequency is<br />

fC = 100 kHz for pass-band with KC = -3 dB, <strong>the</strong><br />

stop-band frequency is fS = 300 kHz, for<br />

KS = -35 dB and Butterworth approximation.<br />

The coefficients <strong>of</strong> <strong>the</strong> denominator <strong>of</strong> <strong>the</strong> transfer<br />

function (2), obtained using <strong>the</strong> computer<br />

tool NAF [4] are b0 = 1,56207.10 26 ,<br />

b1 = 6,49284.10 19 , b2 = 1,34940.10 13 ,<br />

b3 = 1,64280.10 6 , b4 = 1. The transconductances<br />

are: g4 = 7,72 mS, g3 = 3,87 mS, g2 = 2,24 mS,<br />

g1 = 1,13 mS (for C1 = C2 = C3 = C4 = 4,7 nF) .<br />

There is possible apply <strong>the</strong> approximate equations<br />

g3 = g4/2, g2 = g4/3,41, g1 = g4/6,83, <strong>the</strong>n<br />

tuning is very simply. The transconductance


(gm) can be continuously controlled by a DC<br />

control current ISETm = 10.gm [5].<br />

III. EXPERIMENTAL RESULTS<br />

The proposed circuit (Figure 2) was simulated<br />

by PSpice using pr<strong>of</strong>essional macro models<br />

LT 1228 [5]. Resulting magnitude responses<br />

(LP, BP, HP, BR) are given in Figure 3. These<br />

responses are confirming <strong>the</strong>oretical assumptions.<br />

10<br />

K[dB] 0<br />

-10<br />

-20<br />

-30<br />

-40<br />

-50<br />

-60<br />

-70<br />

BP<br />

HP<br />

BR<br />

LP<br />

U CC = ± 15 V<br />

f c = 96 kHz<br />

I SET1 = 111 uA<br />

1,0E+03 1,0E+04 1,0E+05 1,0E+06 1,0E+07<br />

f[Hz]<br />

Figure 3: Magnitude responses<br />

Electronic tuning was tested for LP filter in<br />

Figure 4, changing <strong>the</strong> control current ISET1 in<br />

<strong>the</strong> range given to in Table 1, o<strong>the</strong>r types transfer<br />

functions were tested too.<br />

10<br />

K[dB] U CC = ± 15 V<br />

0<br />

-10<br />

-20<br />

-30<br />

-40<br />

-50<br />

-60<br />

1 2 3 4 5<br />

-70<br />

1,0E+04 1,0E+05 1,0E+06<br />

f[Hz]<br />

1,0E+07<br />

Figure 4: Simulation <strong>of</strong> <strong>the</strong> tuning <strong>of</strong> <strong>the</strong> LP filter<br />

Curve index ISET1 [uA] fC [kHz]<br />

1<br />

2<br />

3<br />

4<br />

5<br />

19<br />

111<br />

323<br />

559<br />

677<br />

17<br />

96<br />

280<br />

485<br />

589<br />

Table 1: Control current versus cut-<strong>of</strong>f frequency<br />

18<br />

The proposed multifunctional 4 th order filter<br />

was practically realized on a solderless connecting<br />

field and <strong>the</strong>n measured. The results<br />

are given to in Figure 5 and compared with<br />

<strong>the</strong> computer simulation above.<br />

simulation<br />

10<br />

f C � 96 kHz<br />

K [dB]<br />

0<br />

U INP (LP, HP) = 50 mV<br />

-10 U INP (BP) = 20 mV<br />

U CC = ± 15 V<br />

-20 R L = 1 M�<br />

I SET1 = 111 uA<br />

-30<br />

HP<br />

-40<br />

BP<br />

-50<br />

measurement<br />

f C � 105 kHz<br />

-60<br />

1,0E+03 1,0E+04 1,0E+05 f [Hz] 1,0E+06<br />

Figure 5: Results <strong>of</strong> simulation and measurement<br />

IV. CONCLUSION<br />

This filter has advantage in more <strong>of</strong> <strong>the</strong> type<br />

transfer function, simplicity in circuit design,<br />

large cut-<strong>of</strong>f frequency range and simply tuning.<br />

Disadvantage is small input voltage level.<br />

Function <strong>of</strong> this filter was verified by computer<br />

simulation and by practical measurement.<br />

ACKNOWLEDGMENTS<br />

Research described in <strong>the</strong> paper was supported<br />

by <strong>the</strong> Grant Agency <strong>of</strong> <strong>the</strong> Czech Republic<br />

project under grant No. 102/08/H027.<br />

REFERENCES<br />

[1] Chen W. K. The circuits and filters handbook.<br />

CRC Press, Florida, 1995<br />

[2] Biolek D., Kolka Z. Snap – symbolic, semisymbolic<br />

and numerical analysis <strong>of</strong> electronic circuits,<br />

on www: http://snap.webpark.cz<br />

[3] Herencsar N., Vrba K. P�eladitelné filtry s OTA<br />

zesilova�i. Elektrorevue 2007/28, ISSN 1213-<br />

1539, 12 s., on www: http://www.elektrorevue.cz<br />

[4] Hajek K., Sedla�ek J., Nafid Program as Powerful<br />

Tool in Filter Education Area. Proc. Of International<br />

<strong>Conference</strong> CBLIS´97, Leicester 1997, s.<br />

PK4 1-10<br />

[5] Linear technology, LT 1228 – Current feedback<br />

amplifier with DC gain control. Linear Technology<br />

Datasheets, 1994. on www:<br />

http://www.linear.com<br />

LP


Active Semiconductor Devices for Millimetre Waves Applications<br />

Michal Pokorný and Zbyn�k Raida (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronics<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czechia<br />

Email: xpokor33@stud.feec.vutbr.cz, raida@feec.vutbr.cz<br />

Abstract — This paper deals with advanced <strong>the</strong>ory<br />

<strong>of</strong> dielectric waveguides and with <strong>the</strong> solid state<br />

physics <strong>of</strong> bulk semiconductors. The research is<br />

aimed to <strong>the</strong> design and numerical modeling <strong>of</strong><br />

active waveguides and antennas using <strong>the</strong> unique<br />

property <strong>of</strong> <strong>the</strong> bulk GaAs – <strong>the</strong> negative differential<br />

resistance. The ma<strong>the</strong>matical model is numerically<br />

implemented using <strong>the</strong> finite element method and<br />

evaluated in COMSOL Multiphysics.<br />

I. INTRODUCTION<br />

The main problem <strong>of</strong> active millimeter-wave (MW)<br />

systems is related to fundamental limitations <strong>of</strong> <strong>the</strong><br />

convectional discrete semiconductor devices which<br />

output power drops drastically at frequencies <strong>of</strong> MW<br />

band. Therefore, alternative principles, like traveling<br />

wave amplification using <strong>the</strong> bulk negative differential<br />

resistance (NDR) that arises in A 3 B 5 semiconductors<br />

(e.g. GaAs), are very perspective. Also losses <strong>of</strong><br />

wave-guiding MW systems are significant. Here, <strong>the</strong><br />

dielectric waveguides are more promising than metallic<br />

ones due to lower cost, easier fabrication and<br />

lower propagation loss [1].<br />

II. SIMPLIFIED MODEL OF WAVE PROPA-<br />

GATION IN EXCITED SEMICONDUCTOR<br />

If <strong>the</strong> uniform doping pr<strong>of</strong>ile and low perturbation <strong>of</strong><br />

GaAs sample is assumed, <strong>the</strong> gradient <strong>of</strong> <strong>the</strong> carrier<br />

concentration is zero, so <strong>the</strong> diffusive processes can<br />

be neglected and <strong>the</strong> free carrier concentration is<br />

always equal to <strong>the</strong> doping concentration. Under<br />

<strong>the</strong>se simplifying considerations, <strong>the</strong> media properties<br />

can be simply described by conductivity σ. If <strong>the</strong><br />

external voltage is applied, <strong>the</strong> electrostatic field E0<br />

arises, and <strong>the</strong> conductivity becomes to have an anisotropic<br />

character due to <strong>the</strong> reduction <strong>of</strong> <strong>the</strong> electron<br />

mobility caused by components <strong>of</strong> <strong>the</strong> E0. Thus <strong>the</strong><br />

conductivity should be expressed as an uniaxial tensorσˆ<br />

. In order <strong>of</strong> fur<strong>the</strong>r simplification, <strong>the</strong> smallsignal<br />

approximation can be used. That introduces<br />

<strong>the</strong> differential form <strong>of</strong> conductivity σD, given by<br />

equation (1) which can be derived from <strong>the</strong> driftdiffusion<br />

scheme [2] <strong>of</strong> <strong>the</strong> semiconductor physics.<br />

19<br />

( )<br />

( ) 2<br />

3<br />

4<br />

4<br />

4vsatE0<br />

− 3�nE0<br />

+ �nEcrit<br />

4 4<br />

E + E<br />

4<br />

qNDE<br />

crit<br />

σ D =<br />

(1)<br />

0 crit<br />

Here q is <strong>the</strong> elementary charge and ND denotes<br />

<strong>the</strong> donor concentration, μn represents <strong>the</strong> original<br />

mobility, vsat is <strong>the</strong> saturation velocity, E is <strong>the</strong> component<br />

<strong>of</strong> <strong>the</strong> electric field in <strong>the</strong> direction <strong>of</strong> <strong>the</strong><br />

current flow, and Ecrit is <strong>the</strong> threshold electric field.<br />

Finally, <strong>the</strong> elements <strong>of</strong> <strong>the</strong> desired tensor σˆ are D<br />

given by <strong>the</strong> formula (1) evaluated for <strong>the</strong> corresponding<br />

component <strong>of</strong> <strong>the</strong> electrostatic field E0<br />

( E )<br />

( )<br />

( )� ��<br />

�σ<br />

D x0<br />

�<br />

ˆ σ =<br />

�<br />

D �<br />

σ D E<br />

(2)<br />

y0<br />

��<br />

σ D Ez<br />

0 �<br />

Considering <strong>the</strong> simplifications above, <strong>the</strong> system<br />

published in [1] can be reduced into a single Helmholz<br />

equation where <strong>the</strong> vector <strong>of</strong> current density is<br />

substituted by <strong>the</strong> product <strong>of</strong> <strong>the</strong> electric field vector<br />

and <strong>the</strong> differential conductivity tensor. Such an<br />

equation is easy to solve, thus <strong>the</strong> analysis <strong>of</strong> a complex<br />

device containing <strong>the</strong> domain <strong>of</strong> <strong>the</strong> active semiconductor<br />

media is less CPU-time and memory consuming.<br />

III. MODELS OF ACTIVE DEVICES<br />

The active waveguide consists <strong>of</strong> <strong>the</strong> GaAs rod,<br />

which is inserted into two open-ended metallic rectangular<br />

waveguides. The GaAs is biased through <strong>the</strong><br />

ohmic contacts placed on <strong>the</strong> top and <strong>the</strong> bottom <strong>of</strong><br />

<strong>the</strong> rod.<br />

Figure 1: FEM model <strong>of</strong> an active waveguide


The active antenna can be created by an openended<br />

dielectric rod, which is excited by an openended<br />

metallic waveguide.<br />

Figure 2: FEM model <strong>of</strong> an active rod antenna<br />

Figure 1 and Figure 2 depict <strong>the</strong> geometry <strong>of</strong> analyzed<br />

structures. The terminating waveguides are <strong>the</strong><br />

standard type WR-10 <strong>of</strong> <strong>the</strong> length 8 mm. The cross<br />

section dimensions <strong>of</strong> <strong>the</strong> GaAs waveguide are<br />

1.10 × 0.55 mm 2 and its length is 24 mm. The donor<br />

concentration in <strong>the</strong> rod volume is ND = 20 ⋅ 10 18 m -3 .<br />

IV. RESULTS<br />

If <strong>the</strong> Poisson equation is solved in <strong>the</strong> domain <strong>of</strong> <strong>the</strong><br />

active semiconductor media, <strong>the</strong>n <strong>the</strong> solution defines<br />

<strong>the</strong> components <strong>of</strong> E0. Consequently, <strong>the</strong> differential<br />

conductivity tensor (2) can be evaluated, and<br />

<strong>the</strong> Helmholz equation can be solved.<br />

S - parameters [dB]<br />

10<br />

5<br />

0<br />

-5<br />

-10<br />

-15<br />

S11<br />

-20<br />

S21<br />

-25<br />

7,5E+10 8,0E+10 8,5E+10 9,0E+10 9,5E+10 1,0E+11 1,1E+11 1,1E+11<br />

f [Hz]<br />

Figure 3: Computed s-parameters <strong>of</strong> <strong>the</strong> active<br />

waveguide<br />

The computed reflection coefficient S11 and <strong>the</strong><br />

transmission coefficient S21 <strong>of</strong> <strong>the</strong> active waveguide<br />

are depicted in Figure 3. The progress <strong>of</strong> <strong>the</strong> S21<br />

coefficient shows that <strong>the</strong> gain <strong>of</strong> <strong>the</strong> rod is about<br />

5 dB, and varies significantly at higher frequencies.<br />

The S11 coefficient has <strong>the</strong> resonating character: <strong>the</strong>re<br />

are well-matched bands on one hand and strong<br />

resonances at higher frequencies on <strong>the</strong> o<strong>the</strong>r hand.<br />

20<br />

Figure 4: Computed gain radiation patterns in dB<br />

for vertical polarization at 88 GHz<br />

The computed gain radiation pattern <strong>of</strong> <strong>the</strong> active<br />

antenna is depicted in Figure 4. The computations<br />

were done for <strong>the</strong> two donor concentrations in order<br />

to compare a passive rod antenna ND = 0 [m -3 ] and an<br />

active one ND = 20⋅10 18 [m -3 ]. The gain increase is<br />

about 4.2 dB. In both cases, <strong>the</strong> magnitude <strong>of</strong> <strong>the</strong><br />

reflection coefficient S11 at <strong>the</strong> antenna input is lower<br />

than –22 dB.<br />

CONCLUSIONS<br />

A simple model <strong>of</strong> electromagnetic wave propagation<br />

in excited semiconductor media was presented.<br />

The model was used in innovative design <strong>of</strong> <strong>the</strong><br />

active dielectric waveguide and <strong>the</strong> active rod antenna.<br />

The future work will be focused on <strong>the</strong> investigation<br />

<strong>of</strong> <strong>the</strong> stability <strong>of</strong> structures in order to avoid<br />

<strong>the</strong> self-oscillation behavior (this effect is ignored in<br />

<strong>the</strong> presented work). Moreover, o<strong>the</strong>r types <strong>of</strong> NDR<br />

media for operating at lower current densities will be<br />

developed.<br />

ACKNOWLEDGMENTS<br />

Research described in this contribution was financially<br />

supported by <strong>the</strong> Czech Grant Agency under<br />

<strong>the</strong> grants no. 102/07/0688 and 102/08/H018, and by<br />

<strong>the</strong> research program MSM 0021630513. The research<br />

is a part <strong>of</strong> <strong>the</strong> COST project IC0603.<br />

REFERENCES<br />

[1] D. Lioubtchenko, S. Tretyakov and S. Dudorov.<br />

Millimeter-Wave Waveguides, Boston: Kluwer<br />

Academic Publishers, 2003<br />

[2] S. Selberherr. Analysis and Simulation <strong>of</strong> Semiconductor<br />

Device, Hiedelberg: Springer-Verlag,<br />

1984.


Analysis <strong>of</strong> Power-Bus Structures Using Contour Integral Method<br />

Martin Štumpf and Zbyn�k Raida (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronics<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: xstump00@stud.feec.vutbr.cz, raida@feec.vutbr.cz<br />

Abstract — The paper is focused on <strong>the</strong> analysis <strong>of</strong><br />

power-bus structures on printed circuit boards<br />

using <strong>the</strong> contour integral method (CIM). The CIM<br />

was implemented in Matlab and <strong>the</strong> program is<br />

capable to analyze <strong>the</strong> parallel-plate structures with<br />

arbitrary material parameters, polygonal shape and<br />

number <strong>of</strong> ports. As an output <strong>the</strong> program provides<br />

<strong>the</strong> frequency dependent voltage distribution between<br />

<strong>the</strong> plates, voltage/current relations between<br />

<strong>the</strong> ports and <strong>the</strong> far-field radiation diagram <strong>of</strong> <strong>the</strong><br />

structure. The evaluation <strong>of</strong> <strong>the</strong> far field radiation<br />

diagrams is based on <strong>the</strong> application <strong>of</strong> <strong>the</strong> field<br />

equivalence principle. All results were validated by<br />

measurement.<br />

I. INTRODUCTION<br />

Power-bus structures serves as <strong>the</strong> charge storage for<br />

<strong>the</strong> integrated circuits (IC) on printed circuit boards<br />

(PCB). The second essential purpose is <strong>the</strong> reduction<br />

<strong>of</strong> <strong>the</strong> switching noise rising from <strong>the</strong> non-ideal<br />

switching within IC. As a consequence <strong>of</strong> <strong>the</strong> switching<br />

noise and non-zero impedance <strong>of</strong> <strong>the</strong> structure,<br />

<strong>the</strong> power-bus structure is a source <strong>of</strong> <strong>the</strong> radiated<br />

electromagnetic emissions.<br />

All today’s analyses <strong>of</strong> power-bus structures are<br />

focused on <strong>the</strong> evaluation <strong>of</strong> voltage/current relations<br />

between <strong>the</strong> ports (e.g. in Z-matrix form) and on <strong>the</strong><br />

far-field radiation diagrams [1]. Recently, several<br />

numerical methods were used. First group <strong>of</strong> <strong>the</strong>se<br />

methods are restricted on simply-shaped structures<br />

only [1], and o<strong>the</strong>rs are based on <strong>the</strong> full-wave numerical<br />

methods as finite-difference-time-domain<br />

method (FDTD) [2] or on <strong>the</strong> circuit extraction based<br />

on <strong>the</strong> mixed-potential integral equation (CEMPIE)<br />

[3]. The numerical approach presented here is based<br />

on <strong>the</strong> 2D contour integral equation [4], usually<br />

called as Weber’s solution for cylindrical waves [5].<br />

In <strong>the</strong> presented paper, <strong>the</strong> resulting Z-matrix is converted<br />

into <strong>the</strong> so-called TouchStone file format,<br />

which allows embedding <strong>the</strong> power-bus structure<br />

into a network analysis. This <strong>the</strong>n enables to simulate<br />

<strong>the</strong> behavior <strong>of</strong> <strong>the</strong> populated power-bus structures<br />

and design <strong>the</strong> proper decoupling. Fur<strong>the</strong>r, <strong>the</strong> radi-<br />

21<br />

ated emissions are evaluated using <strong>the</strong> fieldequivalence<br />

principle [5].<br />

II. FORMULATION OF THE PROBLEM<br />

The problem is firstly formulated as a 2D boundary<br />

value problem. Consequently, Green’s <strong>the</strong>orems and<br />

<strong>the</strong> properties <strong>of</strong> a unit source are applied, getting a<br />

second kind Fredholm integral equation. The voltage<br />

at <strong>the</strong> point Q at <strong>the</strong> periphery C <strong>of</strong> <strong>the</strong> power-bus<br />

structure is given by <strong>the</strong> voltage V(s) and surface<br />

current density iSn(s) along this periphery as [4]:<br />

1<br />

( 2)<br />

( 2)<br />

V ( Q)<br />

� � H ( ) ( ) cos H ( ) ( ) �<br />

1<br />

0 0<br />

2 � k kR V s � � j��<br />

d kR iSn<br />

s ds<br />

j<br />

C<br />

( 1)<br />

where H0 (2) ( . ) and H1 (2) ( . ) are <strong>the</strong> zero-order and firstorder<br />

Hankel functions <strong>of</strong> <strong>the</strong> second kind, respectively.<br />

Next, j =�-1 is <strong>the</strong> imaginary unit, � is <strong>the</strong><br />

angular frequency, �0 is <strong>the</strong> vacuum permeability, k<br />

is <strong>the</strong> complex wavenumber and � denotes <strong>the</strong> angle<br />

between <strong>the</strong> outward normal vector to C and vector<br />

in <strong>the</strong> direction <strong>of</strong> <strong>the</strong> joint line between<br />

source/observe points.<br />

The integral equation is <strong>the</strong>n numerically solved<br />

using CIM. In short only, <strong>the</strong> boundary contour C is<br />

divided into <strong>the</strong> small segments and <strong>the</strong> electric and<br />

magnetic field are assumed to be constant over each<br />

segment. After a few <strong>of</strong> algebraic steps, <strong>the</strong> system <strong>of</strong><br />

linear algebraic equations is obtained. From this<br />

system <strong>of</strong> equations, <strong>the</strong> voltages along <strong>the</strong> boundary<br />

contour C and voltage/current relations between <strong>the</strong><br />

ports can be derived.<br />

A. FAR FIELD RADIATION DIAGRAMS<br />

The far field radiation diagrams <strong>of</strong> <strong>the</strong> parallelplate<br />

structure can be evaluated using <strong>the</strong> fieldequivalence<br />

principle. In this paper, <strong>the</strong> radiated<br />

electric field are numerically computed according to<br />

[1]:<br />

� jk0r<br />

jk0d<br />

e<br />

jkor<br />

'�er<br />

E(<br />

r)<br />

�<br />

( ')<br />

� � ,<br />

4 � M r e e � e ds (2)<br />

S<br />

r s<br />

� r C<br />

where r a r’ denote position vectors <strong>of</strong> <strong>the</strong> observation<br />

and source points, respectively. Next, r is <strong>the</strong><br />

distance <strong>of</strong> <strong>the</strong> observation point from <strong>the</strong> origin, er<br />

and es denote <strong>the</strong> unit vectors having <strong>the</strong> directions <strong>of</strong>


<strong>the</strong> observation point from <strong>the</strong> origin and <strong>of</strong> <strong>the</strong><br />

equivalent magnetic surface current MS(r’) along C,<br />

respectively. The height <strong>of</strong> <strong>the</strong> structure is denoted as<br />

d.<br />

III. EXAMPLES OF RESULTS<br />

The power-bus structure depicted in Fig. 1 includes 2<br />

SMA ports and 13 decoupling capacitors C = 10 nF.<br />

As a substrate, FR-4 with <strong>the</strong> height <strong>of</strong> d = 1.6 mm,<br />

dielectric constant <strong>of</strong> �r = 4.35 and dielectric tangent<br />

loss tan� = 0.035, was used. The parasitic inductance<br />

and resistance <strong>of</strong> <strong>the</strong> decoupling capacitors were<br />

measured for one capacitor as ESL = 640 pH and<br />

ESR = 120 m�.<br />

(0, 100)<br />

(50, 150)<br />

(50, 100)<br />

(25, 80)<br />

(25, 50)<br />

(75, 130) (120, 130)<br />

(75, 80)<br />

(75, 50)<br />

(120, 80)<br />

(0, 0) (100, 0)<br />

(150, 50)<br />

(50, 35) (100, 50)<br />

(25, 20) (75, 20)<br />

Port1<br />

Figure 1: Analyzed and measured power-bus structure<br />

(dimensions in mm)<br />

The far field radiation diagram at f = 1010 MHz,<br />

with respect to an excitation by a unit current<br />

source at <strong>the</strong> position <strong>of</strong> Port2 is depicted in Fig.<br />

2.<br />

Figure 2: Far field radiation diagram at<br />

f = 1010 MHz<br />

(200, 150)<br />

(180, 130)<br />

Port2 (180, 105)<br />

(180, 80)<br />

(180, 50)<br />

(180, 20)<br />

(150, 0)<br />

(200, 0)<br />

22<br />

|Z12| [Ohm]<br />

10 2<br />

10 1<br />

10 0<br />

10 -1<br />

10 -2<br />

10 -3<br />

Measurement<br />

CIM<br />

10<br />

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2<br />

-4<br />

f [GHz]<br />

Figure 3: Frequency response <strong>of</strong> <strong>the</strong> transmission<br />

impedance |z12|<br />

Fig. 3 shows <strong>the</strong> measured and analyzed transmission<br />

impedance |z12| between Port1 and Port2. The<br />

analysis was carried out, as well as measurement,<br />

in <strong>the</strong> frequency range (20�2000) MHz with <strong>the</strong><br />

frequency step 5 MHz.As can be seen, <strong>the</strong> simulation<br />

results were confirmed by measurement very<br />

well.<br />

REFERENCES<br />

[1] M. Leone. The radiation <strong>of</strong> a rectangular powerbus<br />

structure at multiple cavity-mode resonances.<br />

IEEE Transactions on Electromagnetic Compatibility,<br />

vol. 45, no. 3, pp. 486-492, August 2003.<br />

[2] X. Ye, M. Y. Koledintseva, M. Li and J. L.<br />

Drewniak, DC power-bus design using FDTD<br />

modeling with dispersive media and surface<br />

mount technology components. IEEE Transactions<br />

on Electromagnetic Compatibility, vol. 43,<br />

no. 4, pp. 579-587, November 2001.<br />

[3] J. Fan, J. L. Drewniak, H. Shi and J. L. Knighten,<br />

DC power-bus modeling and design with a<br />

mixed-potential integral-equation formulation<br />

and circuit extraction. IEEE Transactions on<br />

Electromagnetic Compatibility, vol. 43, no. 4,<br />

pp. 426-436, November 2001.<br />

[4] T. Okoshi. Planar Circuits for Microwaves and<br />

Lightwaves. Berlin, Germany: Springer-Verlag,<br />

1985.<br />

[5] J. A. Stratton. Electromagnetic Theory. New<br />

York: McGraw-Hill, 1 st edition, 1941.


Creating and Drawing Fonts for Graphic displays with Paged<br />

Memory Model<br />

Filip Adamec (Student), Tomas Fryza (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronic<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: xadame24@stud.feec.vutbr.cz, fryza@feec.vutbr.cz<br />

Abstract — This abstract describes how to draw a<br />

character with variable font and size, on graphical<br />

displays with paged memory model. There are<br />

firstly described some restrictions for minimal draw<br />

algorithm demands, next how to store font data on<br />

desired target and finally how to create very easy<br />

desired font.<br />

I. INTRODUCTION<br />

In many applications (field instruments, information<br />

panels etc.) it is necessary to represent text information<br />

via graphical displays. The displays have two<br />

types <strong>of</strong> memory organizations, paged memory or<br />

linear memory organizations. I can introduce very<br />

easy way to create and draw text on displays with<br />

paged memory model. This model can be seen in<br />

Figure 1.<br />

Figure 1: Paged display RAM<br />

From figure 1 it is obvious, that <strong>the</strong> best way to<br />

write a character bitmap to <strong>the</strong> display RAM memory<br />

is to write it by columns. For additional simplification<br />

we can require <strong>the</strong> high <strong>of</strong> <strong>the</strong> character like a<br />

multiple <strong>of</strong> eight. With that simplification <strong>the</strong> solution<br />

is easy how can be seen in <strong>the</strong> following text.<br />

23<br />

II. CHARACTER FORMAT AND GENRATION<br />

The first problem need to be solved is how to store<br />

<strong>the</strong> font data. Because commonly we can write target<br />

application in C langue, <strong>the</strong> best way how to store<br />

a data is by arrays. We could use two dimensional<br />

arrays, where <strong>the</strong> first dimension is a character number,<br />

that can be <strong>the</strong> ASCII number with some <strong>of</strong>fset,<br />

and <strong>the</strong> second dimension is <strong>the</strong> character bitmap.<br />

We must now consider that we can have nice characters<br />

with variable width. With that condition we must<br />

know width <strong>of</strong> each character and place it as a first<br />

number in character bitmap. The type <strong>of</strong> our array<br />

depends on <strong>the</strong> font size. For <strong>the</strong> eight pixel high font<br />

it will be BYTE variable, for 16 pixel WORD and for<br />

24, 32 DWORD.<br />

Now we can generate our font. Because it is very<br />

hard to create it by own hands, an easy program in<br />

Visual C++ 6.0 was created, which use fonts from<br />

o<strong>the</strong>r programs and from operation system. In this<br />

case many fonts with variable size could be used.<br />

Note that any size could be set, but we need size in<br />

pixel like a multiply <strong>of</strong> eight. The basic <strong>of</strong> this program<br />

is very easy. It prints all printable ASCII character<br />

to <strong>the</strong> memory bitmap on known position and<br />

<strong>the</strong>n read it back pixel by pixel. After read back<br />

whole character, it prints that character into text file<br />

described above. In figure 2 example <strong>of</strong> generated<br />

font for 16 bit font size, is shown.<br />

const unsigned int font[224][17] = {<br />

………<br />

{0xc,0x0,0x1c00,0x1f00,0x3c0,0x2f0,0x23c,0x2e0,0<br />

x23c, 0x2f0,0x3c0,0x1c00,0x1c00}, //A<br />

{0xc,0x0,0x0,0x1ffe,0x1ffe,0x1042,0x1042,0x1042,<br />

0x1042, 0x18e6,4028,1816,0x0}, //B<br />

{0xc,0x0,0x0,0x1e0,0x7f8,0xe1c,0x1806,0x1002,0x1<br />

002, 0x1002,0x1806,0xe1c,0x618}, //C<br />

{0xc,0x0,0x0,0x1ffe,0x1ffe,0x1002,0x1002,0x1002,<br />

0x1806, 0xe1c,0x7f8, 0x1e0,0x0}, //D<br />

………<br />

};<br />

Figure 2: Font example


SIZE<br />

WIDTH<br />

0x0000<br />

0x1c00<br />

0x1f00<br />

0x03c0<br />

0x02f0<br />

0x023c<br />

0x020e<br />

0x023c<br />

0x02f0<br />

0x03c0<br />

0x1f00<br />

0x1c00<br />

Figure 3: Bitmap <strong>of</strong> character “A”<br />

III. DRAW CHARACTER<br />

Now, we have everything prepared for draw our font<br />

on display. We must consider <strong>the</strong> last thing and it is<br />

if we can use a universal algorithm or not. The universal<br />

algorithm can print all font sizes, but is less<br />

exacting than <strong>the</strong> type for fixed font size. I describe<br />

here <strong>the</strong> universal algorithm. The algorithm needs<br />

three global variables, first for current text position<br />

titled cur_line, <strong>the</strong>n for count <strong>of</strong> pixel from beginning<br />

<strong>of</strong> a line cur_pos and <strong>the</strong> variable that represents<br />

actual character position on <strong>the</strong> line cur_column. The<br />

algorithm must change <strong>the</strong> line if <strong>the</strong> actually printed<br />

character could not be printed on actual line and do<br />

nothing if <strong>the</strong> actual number <strong>of</strong> line is higher than<br />

maximum. Also <strong>the</strong> interpret character for jump to<br />

new line is needed. The algorithm in C language is<br />

following.<br />

void TERMIO_PutChar(char character)<br />

{<br />

int i,j;<br />

char pos;<br />

//if jump to <strong>the</strong> new line<br />

if(znak == 0xa)<br />

{<br />

cur_line++;<br />

cur_pos = 0;<br />

cur_column = 0;<br />

return;<br />

}<br />

//subtract <strong>the</strong> font <strong>of</strong>fset<br />

pos = character - FONT_FIRST_CHAR;<br />

//could be <strong>the</strong> character printed on actual line<br />

if((cur_pos + font[pos][0]) > 128)<br />

{<br />

cur_line++;<br />

cur_pos = 0;<br />

cur_column = 0;<br />

}<br />

//if we reach maximum numbers <strong>of</strong> text lines,<br />

//display dependent<br />

if(cur_line > 5)<br />

24<br />

return;<br />

//loop for display pages<br />

for(j = 0; j < FONT_BYTES; j++)<br />

{<br />

//set display page and page position to write<br />

SetPageAdress(cur_line*FONT_BYTES+j);<br />

SetCulomAdress(cur_pos);<br />

//print all data to <strong>the</strong> actual selected<br />

//display page<br />

for( I = 0; I < font[pos][0]; i++)<br />

LCDWrite((font[pos][i+1]>>(8*j))&0xff,true);<br />

}<br />

//change <strong>the</strong> current positions<br />

cur_column++;<br />

cur_pos += font[pos][0];<br />

}<br />

Figure 4: Proposed draw algorithm<br />

The macro constant FONT_FIRST_CHAR is <strong>the</strong><br />

<strong>of</strong>fset for ASCII character, FONT_BYTES is <strong>the</strong><br />

high <strong>of</strong> basic font array type.<br />

IV. RESULTS<br />

The best character size for <strong>the</strong> tested display was 16<br />

pixels high. The 8 pixels high characters are badly<br />

visible and <strong>the</strong> higher pixels high have no meaning,<br />

because we can not represent useless information.<br />

The printed characters for 16 pixels high characters<br />

are shown in figures 5.<br />

Figure 5: Example <strong>of</strong> 16 pixels high characters<br />

ACKNOWLEDGEMENTS<br />

The author would like to thank for <strong>the</strong> financial support<br />

<strong>of</strong> <strong>the</strong> Czech Ministry <strong>of</strong> Education, Youth and<br />

Sports under grants no. MSM 002 163 0513.<br />

REFERENCES<br />

[1] David J. Kruglinski, George Shepherd a Scot<br />

Wingo. Programujeme v Micros<strong>of</strong>t Visual C++.<br />

Computer press 2000, ISBN: 8072263625.<br />

[2] Electronic Assembly, DOGM GRAPHIC SE-<br />

RIS 128x64 DOTS. [online]. 2007 –<br />

[25.9.<strong>2008</strong>] Avaliable: http://www.lcdmodule.de/eng/pdf/grafik/dogm128e.pdf


Bandwidth Comparison <strong>of</strong> Microwave Planar Hybrids<br />

Petr Orság and Ji�í Sva�ina (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronics<br />

Brno University <strong>of</strong> Technology<br />

Czech Republic<br />

E-mail: xorsag04@stud.feec.vutbr.cz, svacina@feec.vutbr.cz<br />

Abstract — The paper deals with some parameters<br />

<strong>of</strong> microwave planar hybrids mainly from <strong>the</strong>ir<br />

attainable frequency bandwidth. The parameters<br />

are especially <strong>the</strong> coupling (C), insertion loss (IL),<br />

isolation (I), <strong>the</strong> VSWR and corresponding bandwidth.<br />

We compare a branch line coupler, rat-race<br />

coupler and De Ronde’s coupler. All three couplers<br />

are designed for centre frequency <strong>of</strong> 3 GHz. Their<br />

main common advantage is a very easy construction,<br />

because <strong>the</strong>y consist from some microstrip<br />

lines eventually a slot line only. Parameters <strong>of</strong> <strong>the</strong>se<br />

planar hybrids were simulated and compared with<br />

measured experimental results.<br />

I. INTRODUCTION<br />

Microwave directional couplers are used frequently<br />

in structures, where we need divide and/or sum microwave<br />

power – in dividers, attenuators, phase<br />

shifters, combiners and o<strong>the</strong>r microwave structures.<br />

Herat we require, that <strong>the</strong>se structures didn’t have big<br />

return loss. Individual types <strong>of</strong> directional couplers<br />

have various shapes and parameters.<br />

Brach line coupler shown in Fig. 1 is a double reflection<br />

symmetrical structure with <strong>the</strong> 1st type <strong>of</strong><br />

directivity. Output signals on port 2 and 3 (coupling<br />

and insertion loss) are in quadrate. Lengths <strong>of</strong> single<br />

microstrips usually opt for �g/4.<br />

Fig. 1. Branch line directional coupler<br />

The structure <strong>of</strong> a rat-race planar coupler is displayed<br />

in Fig. 2. By exciting <strong>the</strong> port 1 (2), output<br />

signals 2 and 3 (1 and 2) are in phase, by exciting <strong>the</strong><br />

port 3 (4) <strong>the</strong> output signals 1 and 4 (2 and 3) are in<br />

opposite phase. Rat-race couplers have usually bigger<br />

frequency bandwidth that branch line couplers.<br />

25<br />

Fig. 2. Rat-race directional coupler<br />

The form <strong>of</strong> so-called composite hybrid (De<br />

Ronde’s coupler) is shown in Fig. 3. This hybrid<br />

consists <strong>of</strong> H shaped microstrip line on <strong>the</strong> top side<br />

<strong>of</strong> <strong>the</strong> dielectric substrate, and <strong>of</strong> an open-ended<br />

slotline on this opposite side. Output signals on ports<br />

2 and 3 are again in quadrate.<br />

Fig. 3. De Ronde’s directional coupler<br />

The main individual electrical parameters <strong>of</strong> all<br />

three hybrids in Fig. 1 to Fig. 3 are defined as follow:<br />

coupling [dB]<br />

P1<br />

1 1<br />

C � 10 log � 20 log � 20 log , (1)<br />

P S k<br />

21<br />

21


insertion loss [dB]<br />

IL<br />

P<br />

1<br />

1 � 10log<br />

� 20log<br />

� 20log<br />

, (2)<br />

P<br />

2<br />

31 S31<br />

and isolation [dB]<br />

1<br />

I �<br />

41<br />

41<br />

1<br />

1�<br />

k<br />

P 1<br />

� 10log<br />

20log<br />

, (3)<br />

P S<br />

where k is <strong>the</strong> voltage coupling coefficient <strong>of</strong> <strong>the</strong><br />

structure (k < 1).<br />

II. RESULTS<br />

All three above mentioned types <strong>of</strong> planar directional<br />

couplers were designed on dielectric substrate<br />

with relative permittivity �r = 3,55 (Arlon AD-350),<br />

centre frequency 3 GHz, characteristic impedance <strong>of</strong><br />

input lines <strong>of</strong> Z0 = 50 �, coupling C = 3 dB for<br />

branch line and rat-race couplers and C = 1,5 dB for<br />

De Ronde’s coupler. The practicable bandwidth <strong>of</strong><br />

<strong>the</strong>se hybrids can be examined in different ways<br />

depending on intended hybrid application. In Fig. 4<br />

<strong>the</strong> measured results <strong>of</strong> coupling frequency dependence<br />

are shown for individual investigated hybrids.<br />

The practicable bandwidth is here taken as <strong>the</strong> difference<br />

<strong>of</strong> frequencies for which <strong>the</strong> actual coupling<br />

differs from <strong>the</strong> nominal value <strong>of</strong> 1 dB.<br />

From <strong>the</strong> Fig. 4 we can see, that <strong>the</strong> so stated<br />

bandwidth is biggest for De Ronde’s hybrid with <strong>the</strong><br />

value <strong>of</strong> about 99 % <strong>of</strong> <strong>the</strong> centre frequency f0.<br />

coupling [dB]<br />

0<br />

-2<br />

-4<br />

-6<br />

-8<br />

-10<br />

0 1 2 3 4 5 6<br />

frequency [GHz]<br />

Fig. 4. Frequency dependence <strong>of</strong> planar hybrid couplings<br />

(— rat-race hybrid, - - - De Ronde’s hybrid,<br />

– – branch line hybrid)<br />

In Tab. I <strong>the</strong> comparison <strong>of</strong> practicable bandwidths<br />

is shown for <strong>the</strong> three investigated planar hybrids.<br />

The criterions used are: <strong>the</strong> coupling value deviation<br />

<strong>of</strong> 1 dB, <strong>the</strong> insertion loss value deviation <strong>of</strong> 1 dB,<br />

<strong>the</strong> return loss > 14 dB (VSWR < 1,5), and <strong>the</strong> isola-<br />

26<br />

tion > 20 dB. From <strong>the</strong>se comparison results we can<br />

see, that <strong>the</strong> largest bandwidth shows <strong>the</strong> De Ronde’s<br />

hybrid in all mentioned criterions. The reason for this<br />

result is, that <strong>the</strong> microstrip – slotline transitions used<br />

in <strong>the</strong> De Ronde’s structure are very broadband [1].<br />

The smallest bandwidth shows <strong>the</strong> branch line coupler<br />

in our research. All couplers shown good return<br />

loss properties with an input VSWR < 1,2.<br />

The directivity (D) <strong>of</strong> <strong>the</strong> hybrids defined as<br />

D = I – C is for all investigated structures > 20dB at<br />

<strong>the</strong> centre frequency.<br />

For easy measurement <strong>of</strong> planar microwave directional<br />

couplers an universal socked was created,<br />

which has until frequency <strong>of</strong> 15 GHz comparable<br />

parameters with soldered connectors.<br />

Bandwidth [% <strong>of</strong> f0]<br />

Parameter Branch line<br />

Raterace<br />

De<br />

Ronde<br />

C ± 1 dB 56 73 99<br />

IL ± 1 dB 26 37 54<br />

RL >14 dB<br />

(PSV 20 dB 10 60 43<br />

Tab. I. Planar hybrid bandwidths comparison<br />

ACKNOWLEDGMENTS<br />

Research described in this paper was supported by<br />

<strong>the</strong> Science Foundation <strong>of</strong> <strong>the</strong> Czech Republic<br />

project under grant No. 102/08/H027 and by <strong>the</strong><br />

National Research Programm No. MS 0021630513.<br />

REFERENCES<br />

[1] L. G. Maloratsky. Passive & Microwave Integrated<br />

Circuits. Burlington, USA, 2004, pages 117-<br />

164<br />

[2] F. C. Ronde. A new class <strong>of</strong> microstrip directional<br />

couplers. In 1970 IEEE MTT-S Int. Microwave<br />

Symposium Digest, p. 184-186<br />

[3] J. Sva�ina. Microwave Integrated Circuits (in<br />

Czech). Textbook <strong>of</strong> Brno University <strong>of</strong> Technology,<br />

2007


����������������������������������<br />

�����������������������������������������������<br />

�������������������������������<br />

�����������������������������<br />

��������������<br />

����������������������������������������������������������<br />

��������������������������������������������������<br />

������������������������������������������������<br />

�����������������������������������������������������<br />

����������������������������������������������������<br />

��������������������<br />

�� ������������<br />

��������������������������������������������������<br />

�������������������������������������������������<br />

�����������������������������������������������������<br />

����������������������������������������������������<br />

���������������������������������������������������<br />

������������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������������������<br />

����������������������������������������������������<br />

�������������������������������������������������<br />

���������������������������������������������������<br />

��������������������������������������������������������<br />

����������������������������������������������������<br />

���������������������������������������������������<br />

���������������<br />

����������������������������������������������������<br />

��������������������������������������������������<br />

�������������������������������������������������<br />

��������������������������������������������������<br />

������������������������������������������������������<br />

�����������������������������������������������������<br />

�����������������������������������������������������<br />

���� � ������� � ������������ � ������� � ��������� � ����<br />

����������������������������������������������������<br />

���������������������������������������������������<br />

�������������������������������������������������<br />

�������������������������������������������������<br />

�������������������������������������������������<br />

��������������������������������������������������<br />

����������������������������������������������������<br />

�������<br />

27<br />

������������������������������������������<br />

����������������������������������������������������<br />

� �<br />

�<br />

� �<br />

�<br />

� �<br />

�<br />

� �<br />

�<br />

�����<br />

�����<br />

�����<br />

�����<br />

��������������������������������������������������<br />

���������������������������������������������<br />

� �<br />

� � � �<br />

| � − � | � � �����<br />

�<br />

⋅<br />

� �<br />

�<br />

�<br />

�<br />

�<br />

� �<br />

| � − � | � ⋅ �<br />

�<br />

� �<br />

| � − � | � ⋅ �<br />

�<br />

����� �<br />

− �<br />

��<br />

� �<br />

�<br />

�<br />

| � ⋅ ��<br />

� � ⋅ ��<br />

| �<br />

�<br />

| � ⋅ ��<br />

� � ⋅ ��<br />

| �<br />

�<br />

| � ⋅ ��<br />

� � ⋅ ��<br />

| �<br />

�<br />

| � ⋅ � � � ⋅ � | �<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

− �<br />

��<br />

� �<br />

�<br />

�<br />

�<br />

�<br />

�<br />

− �<br />

��<br />

� �<br />

�<br />

���<br />

���


������������������������������������������<br />

��������������������������������<br />

������������������������������������������������������<br />

�������������������������������������������������������<br />

���������������������������������������������������<br />

��������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������������������<br />

�������������������� �����������������������������<br />

���������<br />

������������������������������������������������<br />

28<br />

����������������������������������������������������<br />

�����������������������������������������������������<br />

������������������������������� ���������������������<br />

���������������������������������������������������<br />

��������������������������������������������������������<br />

�������������������������������������<br />

����������������������������� ���������������������<br />

�������������������������������������������������������<br />

�����������������������������������������������������<br />

������������������������������������������������<br />

����������������������������������������������������<br />

����� � ��� � �� � ��� � �������� � ������ � �������� � ������<br />

�������������������������������������������������<br />

����������������������������������������������������<br />

��������������������������������������������������<br />

�������������������������������������������������������<br />

����������������������������������������������������<br />

�������������������������������������������������<br />

�����������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������<br />

���������������<br />

����������������������������������������������<br />

����������������������������������������<br />

������������������������������������������<br />

��������������<br />

����������<br />

��� � �������� � �� � ������� � ������� � ��� � ���<br />

����������������������������������������<br />

������������������������������������������<br />

������ � ����������� � ��� � ���������������<br />

�����<br />

��������������������������������������������<br />

�����������������������������������������<br />

���������������������������<br />

��� � �� � ������ � ��������� � ����������� � ���<br />

���������������������������������������<br />

����������������������������������������������<br />

�����������������������������������������<br />

���������������������������������������


Decomposing HEX-Programs: Preliminary Results ∗<br />

Thomas Krennwallner, Thomas Eiter and Michael Fink (Faculty Mentors)<br />

Institute <strong>of</strong> Information Systems, Vienna University <strong>of</strong> Technology, Austria<br />

Email: {tkren,eiter,fink}@kr.tuwien.ac.at<br />

Abstract — We show a program rewriting technique for optimizing<br />

HEX-programs, which are nonmonotonic logic programs<br />

with external atoms tailored to access external knowledge<br />

sources. This approach decomposes a program according<br />

to a partition <strong>of</strong> <strong>the</strong> Herbrand universe in order to be able to<br />

compute <strong>the</strong> subprograms over unrelated parts and thus make<br />

<strong>the</strong> overall program execution faster. First experiments show<br />

promising results.<br />

I. INTRODUCTION<br />

The declarative logic programming paradigm [1] has received<br />

increasing interest in recent years. In this work,<br />

we are interested in logic programs (LPs) called HEXprograms<br />

[2] (Higher-order logic programs with EXternal<br />

atoms). External atoms are special devices for reasoning<br />

with knowledge outside <strong>of</strong> a LP, which may accept input<br />

from <strong>the</strong> program intended for updating this external information.<br />

HEX-programs are thus relevant for reasoning<br />

with hybrid knowledge bases, i.e., knowledge bases integrating<br />

different possibly incompatible semantics, and<br />

have been applied in many fruitful applications like Semantic<br />

Web reasoning. Given <strong>the</strong> inherent nonmonotonic<br />

flavour <strong>of</strong> LPs, in general, we are forced to compute <strong>the</strong><br />

models using a highly complex reasoning process. An<br />

inaugural example for nonmonotonic reasoning is Reiter’s<br />

Nixon diamond.<br />

Example 1. Take <strong>the</strong> logic <strong>the</strong>ory T = {r(n),q(n)} containing<br />

assertions about a former US president: Nixon<br />

(short n) was a republican (r) and was also a quaker (q).<br />

The Nixon diamond expresses that (i) quakers are usually<br />

pacifists, whereas (ii) <strong>the</strong> opposite normally holds for republicans.<br />

Adding (i) and (ii) as additional axioms to a<br />

monotonic <strong>the</strong>ory like T would render it inconsistent. It<br />

is well-known that such exceptional information can only<br />

be dealt with in a nonmonotonic formalism like logic programming.<br />

In spirit <strong>of</strong> dl-programs [3], <strong>the</strong> HEX-program<br />

PN below defines <strong>the</strong> rules (2) and (3) to encode (i) and<br />

(ii), resp. There, a(−p, n) and a(p, n) is used to state that<br />

n is (not) a pacifist, resp. We operate on T using a variant<br />

<strong>of</strong> external atoms shown in [4] <strong>of</strong> form &dl[a, C](X),<br />

whose purpose is to add <strong>the</strong> extension <strong>of</strong> a to T before<br />

querying <strong>the</strong> concept expression C in T .<br />

PN : d(n). (1)<br />

a(p, X) ← d(X), &dl[a, r](X), not &dl[a, −p](X). (2)<br />

a(−p, X) ← d(X), &dl[a, q](X), not &dl[a, p](X). (3)<br />

∗ Supported by <strong>the</strong> Austrian Science Fund (FWF) project P20841.<br />

29<br />

&dl[a,p](X)<br />

&dl[a,–p](X)<br />

¬ e<br />

e e<br />

e ¬ C2 a(–p, X) a(p, X)<br />

e e<br />

e e<br />

&dl[a,r](X) &dl[a,q](X)<br />

d(X)<br />

C 1<br />

Figure 1: Dependency graph <strong>of</strong> PN<br />

PN has two models: {d(n), a(p, n)} and {d(n),<br />

a(−p, n)}. The meaning <strong>of</strong> <strong>the</strong> former is that Nixon is<br />

considered as pacifist, whereas <strong>the</strong> latter expresses <strong>the</strong><br />

opposite, that Nixon is not a pacifist.<br />

A well-known method for incrementally computing <strong>the</strong><br />

models <strong>of</strong> LPs is <strong>the</strong> splitting-set method [5]. In [2], this<br />

technique has been adopted to HEX-programs. The basic<br />

idea is to generate a dependency graph <strong>of</strong> <strong>the</strong> atoms in<br />

a program and split it into subprograms according to <strong>the</strong><br />

connected components <strong>of</strong> <strong>the</strong> graph. The evaluation <strong>of</strong><br />

<strong>the</strong> program can thus be seen as computing <strong>the</strong> models<br />

along a tree <strong>of</strong> smaller subprograms, hence speeding up<br />

<strong>the</strong> overall evaluation due to parallelizing <strong>of</strong> unrelated<br />

program parts. See Figure 1 for <strong>the</strong> dependency graph <strong>of</strong><br />

PN from above, where C1 and C2 are two components <strong>of</strong><br />

PN , and C2 depends on C1.<br />

The class <strong>of</strong> e-unstratified HEX-programs are those<br />

whose dependency graph contains at least one cycle<br />

through a nonmonotonic or an external atom preceded<br />

by not; it is thus inevitable to compute <strong>the</strong> models for<br />

such programs by a guess-and-check methodology. The<br />

program in Example 1 is an instance <strong>of</strong> such programs,<br />

since it contains cycles over negated external atoms.<br />

The efficient computation <strong>of</strong> <strong>the</strong> external atom input<br />

is crucial and contributes tremendously to <strong>the</strong> overall<br />

runtime behaviour <strong>of</strong> unstratified programs. In fact, given<br />

a cyclic external atom &g[p](x) with input predicate p,<br />

<strong>the</strong>re are exponentially many updates for it which include<br />

a particular fact p(c) in <strong>the</strong> worst case. A parsimonious<br />

generation <strong>of</strong> input facts must be preferred.<br />

In general, one cannot decide whe<strong>the</strong>r an input containing<br />

p(c) has an effect on <strong>the</strong> outcome <strong>of</strong> <strong>the</strong> evaluation<br />

for &g, but with knowledge on <strong>the</strong> external computation<br />

source, we can drastically prune <strong>the</strong> search space by decomposing<br />

<strong>the</strong> subprogram containing &g into ℓ copies,<br />

each one referring to exactly one distinct member <strong>of</strong> a<br />

partitioning d1 ˙∪··· ˙∪ dℓ <strong>of</strong> <strong>the</strong> domain <strong>of</strong> <strong>the</strong> program.<br />

This splitting <strong>of</strong> <strong>the</strong> domain is generated by grouping in-


dividuals toge<strong>the</strong>r using independency information for<br />

&g. As a consequence <strong>of</strong> this decomposition scheme,<br />

<strong>the</strong> extensions <strong>of</strong> p in one <strong>of</strong> <strong>the</strong> decomposed program<br />

components can only range over its assigned subdomain.<br />

Since independent individuals cannot occur in different<br />

components, we limit <strong>the</strong> variations <strong>of</strong> extensions <strong>of</strong> p in<br />

<strong>the</strong> guesses to a minimum.<br />

Next, we will see an outline <strong>of</strong> a program rewriting<br />

technique taking <strong>the</strong> above considerations into account.<br />

II. OPTIMIZING HEX-PROGRAMS<br />

Let P be a HEX-program over domain C with rules <strong>of</strong> form<br />

a1 ∨···∨ak ← b1,...,bm, not bm+1,...,not bn. For<br />

a rule r, let H(r) ={a1,...,ak} and B(r) ={b1,. . . ,<br />

bm, not bm+1,...,not bn}. Let D ⊆Cbe a nonempty<br />

finite set <strong>of</strong> constant and predicate symbols.<br />

Let (D, �) be <strong>the</strong> partial ordering <strong>of</strong> <strong>the</strong> lattice <strong>of</strong> partitions<br />

<strong>of</strong> D, where � is <strong>the</strong> usual refinement order, i.e.,<br />

a � b iff ∀x ∈ a ∃y ∈ b: x ⊆ y.<br />

For a partition π ∈D, let e ∈ π, and let p be an mary<br />

predicate symbol. We denote by p[e] <strong>the</strong> set {p(c) |<br />

c ∈ e m }. Fur<strong>the</strong>rmore, let p = p1,...,pk be a list <strong>of</strong><br />

predicate symbols. We define p[e] = �<br />

1≤j≤k pj[e] and<br />

p[π] as �<br />

e ′ ∈π p[e′ ].<br />

Let e ⊆ D, let I be an interpretation, let &g ∈ G<br />

be an external atom over D with associated boolean<br />

function f&g, and let x and c be a ground (i.e., concrete)<br />

input and output list, resp. We say e is a basis<br />

for &g if (1) f&g(I,x, c) = f&g(I ∩ x[e], x, c), and<br />

(2) ∀e ′ ⊇ e: f&g(I ∩ x[e], x, c) =f&g(I ∩ x[e ′ ], x, c).<br />

Given an external atom &g, let ∼&g be an equivalence<br />

relation on D defined as a ∼&g b iff {a, b} ⊆e for a basis<br />

e for &g. We define D/∼&g as <strong>the</strong> quotient set <strong>of</strong> D by<br />

∼&g, which forms a partition <strong>of</strong> D.<br />

Let A be a set <strong>of</strong> atoms. For B = {D/∼&b | &b ∈ A},<br />

we say that π is <strong>the</strong> greatest local split <strong>of</strong> A, denoted<br />

gls(A), ifπ is <strong>the</strong> least upper bound <strong>of</strong> B.<br />

Proposition 1. Let x be a ground input list for &g, A be a<br />

set <strong>of</strong> atoms such that &g ∈ A, and I be an interpretation.<br />

Then, for all π ∈Dsuch that gls(A) � π, f&g(I,x, c) =<br />

f&g(I ∩ x[π], x, c).<br />

In <strong>the</strong> following we assume that predicate symbols ps with superscript s do not occur in C. We denote by ps (e)<br />

<strong>the</strong> facts {ps (c) | c ∈ e} for a set e ⊆ D.<br />

The i-splitting substitution in A is σi A = {p/pi |<br />

p(t) ∈ A}, which maps <strong>the</strong> predicate p <strong>of</strong> every ordinary<br />

atom p(t) in A to pi .Ani-splitting <strong>of</strong> A in P , denoted<br />

P i A , is given by <strong>the</strong> set <strong>of</strong> rules<br />

H(r)σ i A ← B(r)σi A , domi (X1),...,domi (Xr)<br />

for each r ∈ bA(P ), where X1,...,Xr are all output<br />

variables which occur in some external atoms in B(r).<br />

30<br />

k 1 2 3 4 5 6<br />

PN (k) 0.09 0.4 1.7 9.8 — —<br />

P d N (k) 0.09 0.16 0.23 0.31 0.39 0.47<br />

Table 1: Evaluation times in seconds for Example 1<br />

Given <strong>the</strong> dependency graph GP <strong>of</strong> P , let GA P =<br />

(A, E(GP )). We call GA P negatively e-cyclic,ifGA P contains<br />

a cycle through an external atom with a neg. edge.<br />

Let bA(P ) be <strong>the</strong> bottom <strong>of</strong> P wrt. A as in [2].<br />

Let C be <strong>the</strong> set <strong>of</strong> strongly connected components<br />

<strong>of</strong> GP , let C ∈ C, and let gls(C) ={e1,...,eℓ}. The<br />

decomposition <strong>of</strong> P wrt. C, denoted dC(P ), is defined as<br />

�<br />

1≤i≤ℓ<br />

P i C ∪{a ← a i | a ∈ H(r),r ∈ bC(P )}∪dom i (ei)<br />

if G C P is negatively e-cyclic, o<strong>the</strong>rwise it is bC(P ). The<br />

decomposition <strong>of</strong> P , denoted P d ,is �<br />

C∈C dC(P ).<br />

The following result shows <strong>the</strong> equivalence <strong>of</strong> P and<br />

its optimized variant P d .<br />

Theorem 2. There is a 1-to-1 correspondence between<br />

<strong>the</strong> answer sets <strong>of</strong> P and <strong>the</strong> answer sets <strong>of</strong> P d .<br />

III. EXPERIMENTS<br />

For experimental evaluation, we adapted our prototype<br />

implementation dlvhex, 1 a HEX-program reasoner. Syn<strong>the</strong>sizing<br />

Example 1, we created PN (k) =PN ∪{d(ni) |<br />

1


�����������������������������������������������������������<br />

������������������������������������������<br />

�<br />

�������������������������������������������������������������������������������<br />

����������������������������������������������������<br />

��������������<br />

�����������������������������������������<br />

�����������������������������<br />

�<br />

�<br />

��������� �� ���� ����������������� ���� �����<br />

�������� ��������� ����� ���� ����� ��� ���������<br />

�������� ��� �������� ������������� ���� �����<br />

��������� ���� ������������ ��� ������ ��� ��������� ��<br />

������� ������� ������ ��� ���� �������� ��� �����������<br />

���������������������������������������������������<br />

������� ������� ��� �� ������ ������������ ����������<br />

�������������������������������������������������<br />

������� ���� ����� ��� ������������ �� �������<br />

���������� ��� �� ����������������� ������� ������� ���<br />

�������� ������ �������������� ������� �������<br />

��������������������������������������������������<br />

������ ������� ���� ����������� �������� ��� ��������<br />

�����������������������������������������������������<br />

�����������������������������������������������<br />

����� ���� ���������� ��� �� ������ ������ ��� �����������<br />

��������������������������������������������������<br />

����������������������������������������������������<br />

�������������������������������������������������<br />

��������� ���� �������� ��� ��������� ��� ���� ���������<br />

�����������<br />

��� �������������<br />

�� ���������������� ��������� ����� ���� ���� �����<br />

������ ���������� ����� ��������� �� �� ������� ���<br />

��������������������������������������������<br />

���������������������������������������������<br />

������ ����������� ������ ������� ������������<br />

����������������������������������������������������<br />

������������������������������������������������������<br />

���������������������������������������������������<br />

�������� ��� ���������������������� ���� �� ���������<br />

��������� ������������ ���������� ������ �������� ����� ��<br />

������� ������������� ������� ����� ��� �������� ����<br />

������������� ��������� ��������� ���� ���� �����<br />

����� ��������� ����� ������������� ��� ��� ���������<br />

����������������������� ������������������������<br />

�������� ����� ���������� ����������� ����<br />

������������� ���������� ������ �������� �����<br />

������������ ��� ���� ���� ������� ���� ������������<br />

����������������������������������������������������<br />

��� ���� �������������� ������ ��� �����������<br />

�������������<br />

31<br />

��������������������������������<br />

��� ����������<br />

���������������������������������������������������<br />

�<br />

�����������������<br />

�<br />

������ �� �� ������������������������������������<br />

������ ���� ����� ������� ���� ����� �������� ���������� ���<br />

��������� �������������� ��� ���� ��������� �������� �� ����<br />

� � �����<br />

�� ���� ����� ���������� ������������� �������<br />

���� �������� �������������� ��������� ��� ���� ���������<br />

��������� ���������� ������� ��� ����� ����� ���� ����� ��<br />

������������������������ �����������������������������<br />

����������� ������ ���� ����������� ������ ��� ������� ��<br />

����������������������������� ����������<br />

�<br />

��������������<br />

�<br />

������ �������������������������������������������������<br />

����������� ��� ���� �������� ������� ���� ��� ���� ��������<br />

������������������������������������������������������<br />

����������������������������������������������������<br />

������������� ����� ������������������������ ����<br />

���� ������ ���������� ������ ���������� �������� ����<br />

�����������������������������������������������������<br />

����� ������� ���������� ��� ������������ �������������<br />

����������������������� ��������������� �����������<br />

���������������������������� ���������� �<br />

������ ���� ��� �������� � � �����<br />

�� �� �������������<br />


��������������������������������������������������<br />

����� ����� ��� ������� ���� �������� ������������ ������<br />

������ ��� �������� ���� �������� ������ ������� ��� ����<br />

�����������������������������������������������������<br />

��������������������������������<br />

���� ���� ��� �������� ��� ��� �������� ����� ������ ��<br />

�����������������������������������������������������<br />

����������� �������� ������� �������� ����� ��� ���������<br />

������ ����������������������������������������������<br />

�������� �� ��������� ������ �� ������� ���������� ����<br />

����������� ������ ��� �������� ��� ��� ��� ��������� ����<br />

������������ ��� ������� ������������� ��� ���� ������ ��<br />

�������� ������� �� � � ���� ������������������������<br />

�������������������������<br />

�<br />

������������������<br />

�<br />

��������������������������������������������������������<br />

����������� ��� � ������������� �� ����������� ������ ���<br />

������ ��� ����������� ������ �� ��� �� ������� ��� ����<br />

������ ����������� ������ ������������� ��� ���� �������<br />

������������� �����������������������������������<br />

���� � � �����<br />

�� ������ ����������� ���� ����������<br />

�����������������������������������������������<br />

����������������������������������������������������<br />

���������������������������������������<br />

�<br />

�<br />

������������������������������������������<br />

���������<br />

����� ��������<br />

��� ����� ���� ����� �������� ���� ������ ����<br />

������������ ��� �� ������� ����� ������ ��� ������ �������<br />

������������������������������������������������������<br />

���������������������������������������������������<br />

���� ��������� ������������ ������������� ��� ������ �����<br />

���������� ��� ������ ����������������� ��������� ���<br />

������������� ���� ���� ����� ���������� ����<br />

������������ ��� �������� ������������ �������� ��������<br />

���������������������������������������������������<br />

��� ���� ������� ���� ��� ��������� ����� ���� �����<br />

������� ������ ��� ���� ���������� ������� ���� ����<br />

��������� ������ ��� ���������� ��� ������������ ����<br />

� �����������������������������������������������������<br />

����������������������������������������� �����������������������������������������������������<br />

�������������������������������������������������������������������������������<br />

������������������������������������������������������<br />

���� ������ ��� ��������� ����� ���� ��� ������������<br />

������ ��� ������������ ������ ���� ���� ������������<br />

���� ����������<br />

����� ���� ���� ��������� ����������� ����������<br />

���������������������������������������������������<br />

���������� ������ ������������� ������ ���������<br />

������������������������������������������������<br />

����������������������������<br />

32<br />

�����������������������������������������<br />

���� �����������<br />

���� ��� ��������� ��� ��������������� ��� ������ ����<br />

��� �������� ���������������� ��� �������������<br />

����������������� ������� ��� �� ������������<br />

�������� ��� �������������� ����������� ���<br />

��������� ���� �������������������������������<br />

���������������<br />

���� �����������������������������������������������<br />

�������� ����������� ������������ ����������<br />

������<br />

������������������������������������������������<br />

������ ���� ���� ��������� ���������������� �������<br />

���� ������������ ��������� ��� ���������� ���<br />

�������������������������������������������<br />

�����<br />


Simulation <strong>of</strong> light environment and its application in practice<br />

Anna Beneova, Marek Krupa, Ruzena Kralikova<br />

Department <strong>of</strong> environmental studies and information engineering<br />

Technical University in Kosice<br />

Kosice, Slovak republic<br />

Email: anna.beneova@tuke.sk, marek.krupa@tuke.sk, ruzena.kralikova@tuke.sk<br />

Abstract — The condition and development <strong>of</strong><br />

artificial lighting it’s a way <strong>of</strong> expression economical<br />

and energetic situation <strong>of</strong> state, criterion <strong>of</strong><br />

substantial and cultural levels <strong>of</strong> population. Following<br />

present experience come through that <strong>the</strong><br />

enhanced parameters lighting may increase productivity<br />

<strong>of</strong> labour and toge<strong>the</strong>r reach high quality<br />

<strong>of</strong> products in <strong>the</strong> all sectors <strong>of</strong> country economy.<br />

On <strong>the</strong> present, as computer technics bet more<br />

available and its development in <strong>the</strong> last terms has<br />

intensifyed yet, we more apply practices making<br />

use <strong>of</strong> computing graphic if we want to solve problem<br />

<strong>of</strong> luminosity - technical projects. Computer<br />

graphics after alleged methods incoming lighting<br />

get effective centre on calculation and modelling<br />

luminosity- technical (lighting) system, already in<br />

phase its proposal.<br />

I. INTRODUCTION<br />

Nowadays at luminosity- technical modelling<br />

applies access, who be based on computing visualization<br />

three - dimensional scenes <strong>of</strong> engineering<br />

lighting system. In this case realizes luminosity-<br />

engineering calculations with <strong>the</strong> setting precision<br />

without using cargo physical models. Volts computer<br />

visualizations, those aim photo - realistic<br />

reproduction, frequently models in detail and malingers<br />

propagation <strong>of</strong> light in <strong>the</strong> leeway. The<br />

modern visualization programs make it reproduce<br />

brightness, colour and structure surfaces complex<br />

threedimensional leewaies relatively feasibly,<br />

because calculations do include intereflection as<br />

well (inter - bounce) light between various surface<br />

in <strong>the</strong> leeway and various optical effects emergent<br />

by <strong>the</strong> daily, artificial or clustered alight. Simulation<br />

methods are established on classic optic,<br />

<strong>the</strong>rmodynamic, or luminosity- technical models<br />

radiation extensions [2].<br />

II. SIMULATION METHODS<br />

Exist two ground methods being exploited by <strong>the</strong><br />

computer simulation light environment, and o<strong>the</strong>rwise<br />

Monte Carlo method, which in applies<br />

technology following light beam (called raytrac-<br />

33<br />

ing, this title usually utilizes for tracking back<br />

raies, utilizes oneself too concept „ ray casting"<br />

sending light beam, when shaft <strong>of</strong> light go out<br />

from light source), and radiating method (called<br />

radiosity)[1]. From physical aspects are both methods<br />

similar, difference rests in algorithmization.<br />

Volts method following raies work with minimal<br />

plate in stochlastic way (results <strong>of</strong> repeated calculation<br />

can be little different). Volts radiating method<br />

work with bigger surfaces deterministic (results<br />

<strong>of</strong> repeated calculation are always equivalent).<br />

a)<br />

b)<br />

Figure 1: a) Monte Carlo method, b) Radiosity<br />

method (D – walls <strong>of</strong> leeway, S – lighten subject, L<br />

– light source, E – observer)<br />

III. SIMULATION SOFTWARES<br />

When we project lighting system we use as basic<br />

tool for calculation and modeling, for several<br />

years, specialistic computer programs. In last<br />

terms come <strong>the</strong> boom on <strong>the</strong> market with <strong>the</strong> luminosity-<br />

technical s<strong>of</strong>tware following increase in<br />

availability computer technic for everybody projector.<br />

Turn up single programming aids and me-


thod incoming lighting bring chances accurate<br />

calculations <strong>of</strong> lighting system. Following here<strong>of</strong><br />

development come into being dismes luminosity-<br />

technical programs with various levels and purpose<br />

[3]. The last move by <strong>the</strong> development remembered<br />

result is its application not only in proposal<br />

lighting system for living, <strong>of</strong>fice and recreational<br />

spaces, but also its application in engineering<br />

working environment [4].<br />

Among <strong>the</strong> most arranged programs on simulation<br />

light environment belongs besides by o<strong>the</strong>r program<br />

RELUX PROFESSIONAL from <strong>the</strong> Swiss<br />

company Relux informatic AG [5]. Program utilizes<br />

combination <strong>of</strong> methods Monte Carlo and<br />

radiating methods. Function as calculation and<br />

visualization <strong>of</strong> daily and artificial lighting.<br />

IV. APPLICATION SIMULATION IN RELUX<br />

PROFESSIONAL<br />

Last part <strong>of</strong> contribution is above all demonstration<br />

various ascents from specialistic s<strong>of</strong>tware <strong>of</strong><br />

product RELUX PROFESSIONAL 2007 [5]. For<br />

purposes this contribution was carry simulation <strong>of</strong><br />

intensity artificial lighting at chosen workspace.<br />

Options <strong>of</strong> application simulation in Relux Pr<strong>of</strong>essional:<br />

• 3D scene with all <strong>of</strong> element and elements<br />

interior<br />

• attributes illumination intensity<br />

• isot<strong>of</strong>ic graphics illumination intensity<br />

• area map illumination in a given leeway<br />

• 3D map illumination in a given leeway<br />

(Fig. 3)<br />

• 3D rasterization graph illumination intensity<br />

(Fig. 4)<br />

• Wholes 3D sight ratios illumination in <strong>the</strong><br />

under consideration leeway<br />

Figure 3: Three - dimensional maps illumination<br />

34<br />

Figure 4: 3D rasterization graph illumination intensity<br />

V. CONCLUSION<br />

This report was worked - out in <strong>the</strong> flow <strong>of</strong> work<br />

on doctoral <strong>the</strong>sis „ aplication <strong>of</strong> simulation luminosity-<br />

technical quantity by <strong>the</strong> optimalization<br />

lighting system", and within <strong>the</strong> frame project<br />

VEGA 1/3231/06 „ modelling factors <strong>of</strong> working<br />

environment and its optimalization in <strong>the</strong> specified<br />

condition mechanical business.", problem being<br />

solved on Department <strong>of</strong> environmental studies<br />

and information engineering at Technical University<br />

in Kosice.<br />

REFERENCES<br />

[1] SILION, F., PUECH, C.: Radiosity and Global<br />

Illumination, Morgan Kaufmann, ‘94, 149-181<br />

[2] HALÁSZ, Jozef - KRÁLIKOVÁ, Ružena:<br />

Environmentálne informa�né systémy. Košice :<br />

Vienala, 2002. 220 s. ISBN 80-7099-797-4.<br />

[3] BUDAK, V. P. : Vizualizacija raspred�lenija<br />

jarkosti v trjochm�rnych scenach nabljud�nija,<br />

Izdat�lstvo MEI, Moskva 2000<br />

[4] KRUPA, M., KRÁLIKOVÁ, R. : Po�íta�ová<br />

simulácia osvetlenosti pracovných priestorov,<br />

In.: Acta Mechanica Slovaca, 2-B/2005, ISSN<br />

1335-2393<br />

[5] http://www.relux.ch


Towards Synchronization in Multiple Antenna Systems<br />

Qi Wang and Markus Rupp (Faculty Mentor)<br />

Institute <strong>of</strong> Communications and Radio-Frequency Engineering<br />

TU Vienna<br />

Vienna, Austria<br />

Email: {qwang,mrupp}@nt.tuwien.ac.at<br />

Abstract — The presented work deals with synchronization<br />

issues in modern wireless systems employing orthogonal frequency<br />

division multiple access and multiple antenna systems.<br />

We show that current state <strong>of</strong> <strong>the</strong> art algorithms for timing and<br />

frequency <strong>of</strong>fset estimation are capable <strong>of</strong> estimating and compensating<br />

such errors in almost perfect manner even under very<br />

harsh conditions. Throughput simulations <strong>of</strong> a WiMAX compliant<br />

standard support our <strong>the</strong>oretical statement.<br />

I. INTRODUCTION<br />

In <strong>the</strong> past 15 years multiple antenna systems, also called<br />

Multiple-Input Multiple-Output (MIMO) have been a research<br />

focus in wireless communications. Foschini and<br />

Gans [1] and Telatar [2] showed that a much higher<br />

channel capacity is available than originally anticipated.<br />

While many researchers in <strong>the</strong> past years came up with<br />

novel concepts [3]-[6] to make use <strong>of</strong> such capacity, only<br />

few showed in experiments what can be achieved. [7]-<br />

[15]. The problem lies in <strong>the</strong> technical realization <strong>of</strong><br />

many underlaying assumptions.<br />

One <strong>of</strong> such difficulties arises from synchronization in<br />

frequency, phase and sample timing <strong>of</strong> transmitter and<br />

receiver. In simulations typically both ends are perfectly<br />

synchronous while in <strong>the</strong> real world such synchronization<br />

needs to be achieved first. With MIMO systems <strong>the</strong><br />

interference <strong>of</strong> <strong>the</strong> various antenna signals becomes even<br />

larger and thus synchronization algorithms have to be improved<br />

considerably.<br />

Since Orthogonal Frequency Division Multiplex<br />

(OFDM) based transmission plays a major role in future<br />

communications, we concentrated on WiMAX transmissions<br />

to consider principle methods. Also <strong>the</strong> WiMAX<br />

standard supports transmission over more than one antenna.<br />

II. ALGORITHMIC SOLUTION<br />

Unfortunately, timing and frequency <strong>of</strong>fset estimation<br />

are not independent and have to be treated jointly [16]-<br />

[19]. The frequency <strong>of</strong>fset alone in an OFDM system<br />

consists <strong>of</strong> three parts, <strong>the</strong> fractional frequency <strong>of</strong>fset, <strong>the</strong><br />

integer frequency <strong>of</strong>fset as well as <strong>the</strong> residual frequency<br />

<strong>of</strong>fset. The following Figure 1 depicts <strong>the</strong> four steps <strong>of</strong><br />

synchronization required in such system for frequency<br />

and symbol timing. First a coarse symbol timing estima-<br />

35<br />

ton is performed in <strong>the</strong> time domain. With <strong>the</strong> received<br />

signal denoted by r(n), this is achieved by a matched<br />

filter operation:<br />

M(n) =<br />

�Ng+64 k=0<br />

�Ng+64 k=0<br />

r(n − k)r∗ (n − k − 192)<br />

|r(n−k)| 2 + � ,<br />

Ng+64<br />

k=0 |r(n−k−192)| 2<br />

(1)<br />

making use <strong>of</strong> <strong>the</strong> periodic structure in <strong>the</strong> training word.<br />

The correct timing is taken when M(n) reaches its maximum.<br />

In a second step <strong>the</strong> fractional frequency <strong>of</strong>fset<br />

(εFFO) estimation is performed as well in <strong>the</strong> time domain:<br />

εFFO = 1<br />

π arg<br />

⎧<br />

⎨Ng+64<br />

�<br />

� �<br />

j2πn<br />

r(n) exp ·<br />

⎩<br />

N<br />

·<br />

k=0<br />

�<br />

r(n − 128) exp<br />

� j2π(n − 128)<br />

Then <strong>the</strong> integer frequency <strong>of</strong>fset estimation<br />

M(2i) =<br />

N<br />

| �<br />

k X1,k+2iX ∗ 2,k+2i (P1,kP ∗ 2,k )∗ |<br />

�<br />

k |X1,k+2i| 2 + �<br />

k<br />

�� ∗�<br />

. (2)<br />

|X2,k+2i| 2 , (3)<br />

is performed in <strong>the</strong> frequency domain, with <strong>the</strong> received<br />

symbol and pilot symbol in lth OFDM symbol kth subcarrier<br />

in frequency domain represented by Xl,k and<br />

Pl,k. The estimated integer frequency <strong>of</strong>fset 2i appears<br />

when M(2i) gets <strong>the</strong> peak value.<br />

Eventually, <strong>the</strong> residual frequency <strong>of</strong>fset εRFO is<br />

tracked.<br />

Figure 1: The four steps <strong>of</strong> timing and frequency <strong>of</strong>fset<br />

estimation in an OFDM system.


III. RESULTS<br />

In <strong>the</strong> following simulation we measured <strong>the</strong> throughput<br />

<strong>of</strong> a WiMAX compliant OFDM system under frequency<br />

and timing errors when <strong>the</strong> transmission takes place over<br />

a typical pedestrian B channel. The result is depicted in<br />

Figure 2. As <strong>the</strong> figure shows, if we would only correct<br />

for timing or frequency and not taking <strong>the</strong> second part<br />

into consideration, we would loose substantial amounts<br />

in throughput. Only <strong>the</strong> joint approach results in almost<br />

perfect synchronization. The observed throughput loss<br />

due to synchronization estimation is very small even at<br />

low SNR.<br />

perfect synchronized<br />

both timing and frequency<br />

<strong>of</strong>fset corrected<br />

only frequency<br />

<strong>of</strong>fset corrected<br />

only timing<br />

<strong>of</strong>fset corrected<br />

Figure 2: Achieved throughput over SNR under various<br />

frequency and timing estimation errors.<br />

IV. CONCLUSION AND OUTLOOK<br />

In this article we have shown that for OFDM transmission<br />

systems like WiMAX, synchronization algorithms<br />

can be applied to achieve almost optimal synchronization<br />

results and thus maximal throughput. In <strong>the</strong> next<br />

steps such algorithms will be investigated for MIMO systems<br />

employing several transmit and receive antennas.<br />

It is anticipated that such transmission will considerably<br />

hamper <strong>the</strong> quality <strong>of</strong> <strong>the</strong> estimation algorithms and that<br />

<strong>the</strong>y have to be re-designed.<br />

REFERENCES<br />

[1] G.J. Foschini, Layered space-time architecture for wireless<br />

communication in a fading environment when using<br />

multiple antennas, Bell Labs Technical Journal, vol. 1, no.<br />

2, pp. 41-59, 1996.<br />

[2] E. Telatar, Capacity <strong>of</strong> multiple-antenna Gaussian channels,<br />

European Transactions on Telecommunications, vol.<br />

10, pp. 585-595, Nov. 1999.<br />

[3] Erik Larsson, Petre Stoica, und Girish Ganesan, Space-<br />

Time Block Coding for Wireless Communications, Cambridge<br />

University Press, 2003.<br />

[4] Branka Vucetic und Jinhong Yuan, Space-time Coding ,<br />

Wiley & Sons, 2003.<br />

36<br />

[5] H.Bolcskei, D.Gesbert, A.-J. Van Der Veen, Space-Time<br />

Wireless Systems: From Array Processing to MIMO<br />

Communications,, Cambridge University Press, 2006.<br />

[6] G.B. Giannakis, Z.Q. Liu, Space-Time Coding for Broadband<br />

Wireless Communications, Wiley & Sons, 2006.<br />

[7] S. Caban, C. Mehlführer, R. Langwieser, A.L. Scholtz,<br />

M. Rupp, Vienna MIMO Testbed,; EURASIP Journal on<br />

Applied Signal Processing, pp. 1 – 13, 2006.<br />

[8] S. Caban, M. Rupp, Impact <strong>of</strong> transmit antenna spacing<br />

on 2 x 1 Alamouti radio transmission, Electronics Letters,<br />

Vol. 43, 4; pp. 198 – 199, 2007.<br />

[9] M. Rupp, C. Mehlführer, S. Caban, R. Langwieser, L.W.<br />

Mayer, A.L. Scholtz, Testbeds and Rapid Prototyping in<br />

Wireless System Design, EURASIP Newsletter, 17, 3; pp.<br />

32 – 50, 2006.<br />

[10] L.W. Mayer, M. Wrulich, S. Caban, ”Measurements and<br />

Channel Modeling for Short Range Indoor UHF Applications,<br />

Proc. European <strong>Conference</strong> on Antennas and Propagation<br />

(EuCAP 2006), France, Nov. 2006.<br />

[11] C. Mehlführer, S. Caban, M. Rupp, A.L. Scholtz, Effect<br />

<strong>of</strong> Transmit and Receive Antenna Configuration<br />

on <strong>the</strong> Throughput <strong>of</strong> MIMO UMTS Downlink, Proc.<br />

8th International Symposium on DSP and Communication<br />

Systems, Noosa Heads, Sunshine Coast, Australia,<br />

19.12.2005 - 21.12.2005.<br />

[12] C. Mehlführer, S. Geirh<strong>of</strong>er, S. Caban, M. Rupp, A Flexible<br />

MIMO Testbed with Remote Access, Proc. 13th EU-<br />

SIPCO European Signal Processing <strong>Conference</strong>, Antalya,<br />

Turkey, 04.09.2005 - 08.09.2005.<br />

[13] M. Rupp, S. Caban, C. Mehlführer, Challenges in Building<br />

MIMO Testbeds, Proc. 15th EUSIPCO European Signal<br />

Processing <strong>Conference</strong>, Poznan, Poland, pp. 1833 -<br />

1839, 03.09.2007 - 07.09.2007.<br />

[14] C. Mehlführer, S. Caban, M. Rupp, Experimental Evaluation<br />

<strong>of</strong> Adaptive Modulation and Coding in MIMO<br />

WiMAX with Limited Feedback, EURASIP Journal on<br />

Advances in Signal Processing, Article ID 837102, <strong>2008</strong>.<br />

[15] M. Wrulich, S. Caban, M. Rupp, Testbed Measurements<br />

<strong>of</strong> Optimized Linear Dispersion Codes, ITG Workshop on<br />

Smart Antennas, Wien, 26.02.2007 - 27.02.2007.<br />

[16] W.Lei, T.Tao, J.Lu, ”A robust frequency acquisition algorithm<br />

for OFDM systems, The 5th International Symposium<br />

on Wireless Personal Multimedia Communications,<br />

Vol. 1, pp. 145 - 148, vol.1, 27-30 Oct. 2002.<br />

[17] T.Lv; J.Chen, ”ML estimation <strong>of</strong> timing and frequency<br />

<strong>of</strong>fset using multiple OFDM symbols in OFDM systems,<br />

Proc. IEEE Global Telecommunications <strong>Conference</strong>,<br />

GLOBECOM ’03, Vol. 4, pp.2280 - 2284, vol.4,<br />

1-5 Dec. 2003.<br />

[18] L.Wei, Y.Y.Xu, Y.M.Cai, X.Xu, ”Robust frequency <strong>of</strong>fset<br />

estimator for OFDM over fast varying multipath channel,<br />

Electronics Letters, Vol.43, Issue 6, pp.53 - 54, March 15<br />

2007.<br />

[19] Y.Yan; M.Tomisawa, Y.Gong; Y.L.Guan; G.Wu;<br />

C.L.Law, ”Joint timing and frequency synchronization for<br />

IEEE 802.16 OFDM systems, Mobile WiMAX Symposium,<br />

pp.17 21, 25-29 March 2007.


Technical criteria for <strong>the</strong> comparison <strong>of</strong> modern ERP system for<br />

<strong>the</strong> usage in orchestra companies at <strong>the</strong> case study <strong>of</strong> Dynamics<br />

AX 2009<br />

Abstract — Orchestra companies are companies<br />

with 50 (plus or minus 25) employees which do not<br />

create any business value out <strong>of</strong> <strong>the</strong>ir IT system.<br />

The IT systems in <strong>the</strong> company are only operating<br />

resources. In such a company, <strong>the</strong> implementation<br />

<strong>of</strong> a new ERP solution is a critical and cost intensive<br />

step. This research project establishes criteria<br />

and ratios to allow for <strong>the</strong> comparison <strong>of</strong> ERP systems<br />

that are suitable for use in such a company.<br />

The criteria will be split into <strong>the</strong> following areas:<br />

Customizability, integration capacity and, migration<br />

ability. The criteria will be validated against<br />

Dynamics AX 2009 to give a pro<strong>of</strong> <strong>of</strong> concept.<br />

I. INTRODUCTION<br />

In <strong>the</strong> literature some criteria currently exist for <strong>the</strong><br />

comparison <strong>of</strong> architectures <strong>of</strong> ERP systems. [1], [2],<br />

[3] However, <strong>the</strong>se are usually on a very detailed<br />

technical level and are based primarily on <strong>the</strong> requirements<br />

<strong>of</strong> larger companies. This makes it difficult<br />

for small and medium-size enterprises, which<br />

have o<strong>the</strong>r requirements, to use <strong>the</strong>se criteria to decide<br />

upon a particular system. The existing criteria<br />

applied to small and medium size companies do not<br />

always show <strong>the</strong> advantages or disadvantages <strong>of</strong> <strong>the</strong><br />

particular system architectures clearly. An application<br />

<strong>of</strong> comparative criteria which have not been<br />

developed especially for small and medium enterprises<br />

leads to inadequate evaluation results. In this<br />

<strong>the</strong>sis, criteria for <strong>the</strong> comparison <strong>of</strong> ERP systems for<br />

a specific class <strong>of</strong> companies ("orchestra company")<br />

has been developed.<br />

An orchestra company is an enterprise with 25 to<br />

75 employees. Based on <strong>the</strong> definition <strong>of</strong> <strong>the</strong> European<br />

Commission <strong>the</strong>y are ei<strong>the</strong>r small or mediumsized<br />

enterprises. According to <strong>the</strong> definition <strong>of</strong> <strong>the</strong><br />

EU, small businesses are those with less than 50<br />

employees. The annual turnover may not exceed 10<br />

million Euro. A key aspect <strong>of</strong> an orchestra company<br />

is <strong>the</strong> fact that <strong>the</strong> company does not gain direct<br />

value from its IT systems. The entire IT equipment in<br />

Michael Gall und Christian Sterba<br />

Research Group for Industrial S<strong>of</strong>tware (INSO)<br />

Technical University<br />

Vienna, Austria<br />

Email: e9825247@student.tuwien.ac.at<br />

37<br />

<strong>the</strong> company is only an operating resource. These<br />

companies do not have <strong>the</strong>ir own IT department.<br />

This paper is part <strong>of</strong> an ongoing research project.<br />

Based on a literary analysis <strong>of</strong> <strong>the</strong> existing methods<br />

for <strong>the</strong> comparison <strong>of</strong> ERP systems, <strong>the</strong> authoritative<br />

criteria for orchestra companies are identified and<br />

modified. The successful validation <strong>of</strong> this criteria<br />

system has been done in a case study <strong>of</strong> Micros<strong>of</strong>t<br />

Dynamics AX 2009.<br />

II. IDENTIFIED CRITERIA<br />

In an orchestra company decisions are usually made<br />

by <strong>the</strong> decision makers directly without a long and<br />

expensive selection process. [4] In <strong>the</strong> literature<br />

several approaches for <strong>the</strong> evaluation <strong>of</strong> ERP systems<br />

exist, but <strong>the</strong>ir focus is not on its application on <strong>the</strong><br />

chosen enterprise class. [1], [2], [3]. The criteria<br />

identified in this work reflect <strong>the</strong> important decision<br />

criteria which are relevant for orchestra companies .<br />

They can be grouped as followed:<br />

A. CUSTOMIZABILITY<br />

The term customizability includes all activities where<br />

<strong>the</strong> functionality <strong>of</strong> <strong>the</strong> system is changed and it can<br />

be divided fur<strong>the</strong>r into modifiability and extensibility.<br />

[5] The following ratios (including <strong>the</strong>ir domains)<br />

can be defined:<br />

1. What effort is necessary to rework <strong>the</strong> modifications<br />

when switching to a new release <strong>of</strong> <strong>the</strong><br />

system? (0 – no fur<strong>the</strong>r adjustments to 5 –<br />

complete reimplementation)<br />

2. Can modifications been done regardless <strong>of</strong> <strong>the</strong><br />

subcategory (Externally visible, internally<br />

visible or not visible)? (Yes / No)<br />

3. Can modifications be done and stored centrally<br />

? (Yes / No)<br />

4. Do modifications affect <strong>the</strong> process when applying<br />

hot fixes, service packs or new product<br />

versions? (0 – no overhead to 5 – significantly<br />

change)


5. Does an integrated development environment<br />

exist? (Yes / No)<br />

6. Do clear interfaces exist ? (0 – no interfaces to<br />

5 – complete modularization)<br />

7. Can new features be implemented outside <strong>the</strong><br />

existing system ? (‘N’ – Not possible, ‘C’ –<br />

costly or ‘E’ – easy)<br />

8. Does <strong>the</strong> system provide tools to administer<br />

<strong>the</strong> database schema? (Yes / No)<br />

9. Can <strong>the</strong> system access external data structures?<br />

(Yes / No)<br />

B. INTEGRATION CAPACITY<br />

For various reasons <strong>the</strong>re may be a variety <strong>of</strong> o<strong>the</strong>r<br />

systems in <strong>the</strong> enterprise, which need to communicate<br />

with <strong>the</strong> ERP solution. A key aspect here is how<br />

easily is this possible. There are many separate providers<br />

<strong>of</strong> EAI solutions which facilitate communication<br />

between <strong>the</strong> various systems. By taking a few<br />

criteria into consideration, costly integration projects<br />

can be saved.<br />

10. For what processes are services provided?<br />

(‘N’ - None, ‘F’ - Framework, ‘B’ – Basic or<br />

‘A’ - Advanced)<br />

11. Can EAI platforms be integrated? (Yes / No)<br />

12. Are API’s available? (Yes / No)<br />

13. Are data integration capabilities <strong>of</strong> <strong>the</strong> database<br />

system supported? (Yes / No)<br />

14. What file formats are supported? (‘N’ - None,<br />

‘P’ - Proprietary or ‘C’ - Common file formats)<br />

15. Can <strong>the</strong> data and fields be mapped dynamically?<br />

(Yes / No)<br />

16. Do tools for monitoring a data exchange exist?<br />

(Yes / No)<br />

C. MIGRATION ABILITY<br />

The notion <strong>of</strong> migration is defined in this context as<br />

ei<strong>the</strong>r <strong>the</strong> change from a previous system to <strong>the</strong> chosen<br />

ERP solution or as <strong>the</strong> switch between different<br />

product releases <strong>of</strong> <strong>the</strong> same solution.<br />

17. Do tools for data migration exist? (Yes / No)<br />

18. Do application tools to support <strong>the</strong> release<br />

change exist? (Yes / No)<br />

19. Do tools to migrate <strong>the</strong> modifications exist ?<br />

(Yes / No)<br />

III. CASE STUDY DYNAMICS AX 2009<br />

The validation <strong>of</strong> <strong>the</strong> developed criteria has been<br />

carried out in a case study by applying <strong>the</strong> criteria to<br />

<strong>the</strong> product Micros<strong>of</strong>t Dynamics AX 2009. Dynam-<br />

38<br />

ics AX, formerly known as Axapta, came through <strong>the</strong><br />

acquisition <strong>of</strong> Danish s<strong>of</strong>tware company Navision in<br />

<strong>the</strong> possession <strong>of</strong> Micros<strong>of</strong>t. The market placement<br />

<strong>of</strong> Dynamics AX is in <strong>the</strong> segment <strong>of</strong> small and / or<br />

medium-sized enterprises. The current product version<br />

is Dynamics AX 2009 which was used to validate<br />

<strong>the</strong> criteria and ratios. This case study proved<br />

<strong>the</strong> successful application <strong>of</strong> <strong>the</strong> build criteria system.<br />

For <strong>the</strong> given ERP product <strong>the</strong> proposed ratios could<br />

be answered. The results are given in “Table 1”<br />

Ratio AX 2009<br />

1 Revision Effort (0 .. 5)<br />

2 Independent modifications<br />

3 Central modifications<br />

4 Maintenance influence (0 .. 5)<br />

5 Development environment<br />

6 Clear interfaces (0 .. 5)<br />

7 External modifications (N/E/C)<br />

8 Schema administration<br />

9 External data structures<br />

10 Services available (N/F/B/A)<br />

11 EAI integration<br />

12 API’s available<br />

13 Support DB data integration<br />

14 File formats (N/P/C)<br />

15 Dynamic data mapping<br />

16 Monitoring<br />

17 Data migration tools<br />

18 Release change tools<br />

19 Modification migration tools<br />

3<br />

Y<br />

Y<br />

2<br />

Y<br />

3<br />

E<br />

Y<br />

Y<br />

B<br />

Y<br />

Y<br />

N<br />

C<br />

N<br />

Y<br />

N<br />

Y<br />

Y<br />

Table 1: Criteria applied to Dynamics AX 2009<br />

By applying <strong>the</strong> criteria on different ERP solutions<br />

a comparison to support <strong>the</strong> decision making process<br />

is now possible.<br />

REFERENCES<br />

[1] Kennerley, M., Neely A., A framework <strong>of</strong> <strong>the</strong><br />

factors affecting <strong>the</strong> evolution <strong>of</strong> performance<br />

measurement systems, International Journal <strong>of</strong><br />

Operations & Production Management 22(11),<br />

2002, S. 1222-1245<br />

[2] Dumslaff, U.: Ein Vorgehensmodell zur S<strong>of</strong>tware<br />

Evaluation, HMD 31, 1994, S. 89 - 105<br />

[3] Stefanou, C.J.: A framework for <strong>the</strong> ex-ante<br />

evaluation <strong>of</strong> ERP s<strong>of</strong>tware, European Journal <strong>of</strong><br />

Information Systems 10, 2001, S 204 – 215<br />

[4] Barbacci M., SEI Architecture Analysis Techniques<br />

and when to use <strong>the</strong>m, Carnegie Mellon<br />

S<strong>of</strong>tware Engineering Institute, 2002<br />

[5] Gronau, N.. Management von Produktion und<br />

Logistik, 3. überarb. Aufl., Oldenbourg,<br />

München, 1999


DEVELOPMENT OF A QUANTITATIVE APPROACH THAT USES FOURIER<br />

TRANSFORM INFRARED SPECTROSCOPY (FTIR) TO MEASURE AMBIENT<br />

AEROSOL COMPOSITION<br />

Abstract —The goal <strong>of</strong> this research is to develop<br />

a method to collect, identify and quantify organic<br />

and inorganic species present in ambient aerosols.<br />

A sampler to collect both types <strong>of</strong> species on a<br />

single substrate at high efficiency will be developed,<br />

avoiding <strong>the</strong> artifacts that many o<strong>the</strong>r researchers<br />

have experienced. The s<strong>of</strong>tware and<br />

hardware used by <strong>the</strong> FTIR spectrometer will be<br />

optimize to analyze <strong>the</strong> aerosols collected onto a<br />

substrate that is not transparent to IR light and can<br />

<strong>the</strong>refore mask <strong>the</strong> signal <strong>of</strong> <strong>the</strong> aerosols. Because<br />

FTIR is generally not used to quantify chemicals in<br />

solid samples, a chemical method that uses internal<br />

standards added to <strong>the</strong> sample or external analytical<br />

methods to calibrate <strong>the</strong> FTIR will be developed.<br />

Finally, because FTIR has not been used to quantify<br />

chemicals in ambient aerosols before, an algorithm<br />

that relates <strong>the</strong> concentration <strong>of</strong> <strong>the</strong> chemical in <strong>the</strong><br />

ambient air to <strong>the</strong> FTIR absorbance area, chemical<br />

molar absorptivity and sampling flow rate and duration<br />

will be developed. This algorithm will not<br />

depend on <strong>the</strong> path length <strong>of</strong> <strong>the</strong> IR beam through<br />

<strong>the</strong> sample, because this quantity is difficult to<br />

estimate.<br />

I. INTRODUCTION<br />

Particulate matter (PM) is formed by condensation<br />

<strong>of</strong> semi-volatile gases and <strong>the</strong> coagulation <strong>of</strong> even<br />

smaller particles, and <strong>the</strong>y play an important role in<br />

gas phase reactions which require a third body.<br />

These particles are typically formed by <strong>the</strong> condensation<br />

<strong>of</strong> semi-volatile gases onto smaller particles,<br />

by homogeneous nucleation <strong>of</strong> gases to form particles,<br />

and by coagulation <strong>of</strong> smaller particles.<br />

Exposure to ambient fine particulate matter can<br />

cause serious problems to public health. Because<br />

<strong>of</strong> <strong>the</strong>ir small size, fine particles may penetrate in<br />

<strong>the</strong> tracheobronchial and pulmonary region,<br />

causing a range <strong>of</strong> respiratory system responses<br />

from sneezing, coughing, broncho-constriction,<br />

rapid, shallow breathing (Boudel et al., 1994;<br />

Vilma Arriaran and Beth Wittig (Faculty Mentor)<br />

City College <strong>of</strong> New York<br />

City University <strong>of</strong> New York<br />

New York, USA<br />

Email: vilma@ce.ccny.cuny.edu<br />

39<br />

CCOHS, March 22 1999) to lung cancer and<br />

cardiopulmonary disease (Lall et al, 2004) and<br />

even death. For example, in London in 1952, about<br />

4000 people were killed in four days due to<br />

inhalation <strong>of</strong> excessive concentration <strong>of</strong> particles<br />

and SO2 that accumulated in <strong>the</strong> air due to stagnant<br />

wea<strong>the</strong>r conditions (Boudel et al, 1994).<br />

The environment is also adversely affected by exposure<br />

to PM2.5. Deposited PM can corrode stone<br />

buildings and damage <strong>the</strong> cellular structure <strong>of</strong><br />

plants. In suspension <strong>the</strong>y can impair visibility (dust<br />

clouds, haze) .<br />

Researchers have been continuously studying <strong>the</strong><br />

adverse health effects <strong>of</strong> <strong>the</strong> PM and have found<br />

that this effect depends in part on <strong>the</strong> size <strong>of</strong> <strong>the</strong><br />

PM. This prompted more studies on PM2.5 which<br />

creates more adverse health effects due to both<br />

<strong>the</strong>ir size and <strong>the</strong>ir unique chemical composition.<br />

U.S. EPA promulgated a new NAAQS for PM2.5 in<br />

1997. NAAQS set <strong>the</strong> maximum permissible<br />

concentration for fine particulate matter at<br />

15μg/m 3 on an annual basis and 35μg/m 3 on a 24<br />

hour basis (EPA, May 9th <strong>2008</strong>).<br />

The method developed in this work will provide<br />

information that can be used to better identify<br />

sources <strong>of</strong> PM2.5 and develop more effective SIPs<br />

for New York state and <strong>the</strong> Nor<strong>the</strong>astern US (NYS<br />

DEC, <strong>2008</strong>; Solomon, 2004). The method will<br />

speciate and measure <strong>the</strong> concentration <strong>of</strong> <strong>the</strong> organic<br />

and inorganic chemical components present<br />

in fine PM using Fourier Transform Infrared Spectroscopy<br />

(FTIR). The current state <strong>of</strong> <strong>the</strong> science is<br />

unable to identify and quantify all <strong>of</strong> <strong>the</strong> functional<br />

groups in PM2.5 using a single analytical method<br />

and many <strong>of</strong> <strong>the</strong> available methods also require<br />

extraction <strong>of</strong> <strong>the</strong> sample or destroy <strong>the</strong> sample during<br />

analysis. While FTIR is not considered to be<br />

quantitative for liquid and solid samples, through<br />

this research internal and external standards that<br />

make <strong>the</strong> method quantitative will be developed.


II. EXPERIMENTS AND PRELIMINARY RE-<br />

SULTS<br />

PM must be uniformly collected into a small area on<br />

a substrate that can be directly analyzed by FTIR<br />

without introducing artifacts. This will be<br />

accomplished using a sample collection system.<br />

Sample collection systems are designed to draw a<br />

constant flow rate <strong>of</strong> ambient air containing PM2.5<br />

through a substrate for a specified period <strong>of</strong> time,<br />

collecting <strong>the</strong> sample on <strong>the</strong> substrate by impaction,<br />

filtration and sieving.<br />

A. SUBSTRATE SELECTION<br />

Ideal substrates have no signature in <strong>the</strong> MIR region.<br />

Second-best substrates have sharp weak<br />

signatures, especially in MIR regions where targeted<br />

inorganic and organic compounds are expected.<br />

Blank (i.e., clean and prepared for sampling)<br />

substrates will be analyzed using FTIR.<br />

Signatures in <strong>the</strong>ir MIR spectra will be observed to<br />

select <strong>the</strong> substrate that has a sharp and strong<br />

signature that does not obscure <strong>the</strong> sample signature.<br />

Figure 1 shows spectra <strong>of</strong> Teflon and quarts substrates<br />

collected by using transmission and reflectance<br />

spectroscopy.<br />

Absorbance<br />

7<br />

6<br />

5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

3836<br />

3645<br />

Teflon Savillex-ATR<br />

Teflon Savillex-Transmission<br />

Teflon Pall-Transmission<br />

Quartz-ATR<br />

Quartz-Transmission Saturation due to<br />

substrate thickness<br />

3454<br />

3263<br />

3072<br />

2881<br />

2690<br />

2499<br />

2308<br />

2117<br />

1927<br />

Wavenumber (cm -1 )<br />

Sharp and<br />

moderate peaks<br />

Figure 1 Blank substrate IR spectra<br />

B. LIGHT PATH THROUGH SAMPLE<br />

Identical samples will be analyzed by transmission<br />

and reflectance techniques to evaluate <strong>the</strong> diffrence<br />

in peak absorbance areas due to <strong>the</strong> depth <strong>of</strong><br />

penetration <strong>of</strong> <strong>the</strong> light through <strong>the</strong> sample.<br />

C. REFINING ANALYSIS<br />

Due to <strong>the</strong> complex composition <strong>of</strong> PM2.5, its spectra<br />

analysis may show peaks that overlap, due to <strong>the</strong><br />

interactions between functional groups that appear<br />

very close in <strong>the</strong> IR wavenmber range. Different<br />

approaches to split those peaks will be evaluated<br />

1736<br />

1545<br />

1354<br />

1163<br />

972<br />

781<br />

590<br />

399<br />

40<br />

without affecting <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong>ir absorbance<br />

areas.<br />

D. SPECTRAL INTERPRETATION<br />

Functional groups present in <strong>the</strong> sample will be<br />

identified by wavenumber based on spectral library<br />

data and results from literature review. Concentrations<br />

<strong>of</strong> <strong>the</strong> functional groups present in <strong>the</strong> sample<br />

and in <strong>the</strong> actual ambient air will be determined by<br />

developing an algorithm that will use all parameters<br />

calculated after <strong>the</strong> FTIR spectral analysis<br />

A <strong>the</strong>oretical algorithm Equation 1 was developed,<br />

and will need to be validated with internal or external<br />

standards which concentrations are kwnon.<br />

C<br />

xa<br />

ABSxb<br />

1<br />

�<br />

ABS yb �<br />

�<br />

Q � t<br />

x / y<br />

� C<br />

yb<br />

mss<br />

�<br />

�<br />

s<br />

Equation 1<br />

The algorithm calculates <strong>the</strong> concentration <strong>of</strong> <strong>the</strong><br />

functional group in <strong>the</strong> air by relating <strong>the</strong> absorbance<br />

area <strong>of</strong> <strong>the</strong> functional group and <strong>the</strong> standard, <strong>the</strong><br />

relative molar absorptivity, <strong>the</strong> concentration <strong>of</strong> <strong>the</strong><br />

standard in <strong>the</strong> light path, <strong>the</strong> flow and sampling<br />

time, <strong>the</strong> mass <strong>of</strong> <strong>the</strong> sample onto <strong>the</strong> substrate and<br />

<strong>the</strong> density <strong>of</strong> <strong>the</strong> particles. Measurement <strong>of</strong> <strong>the</strong> path<br />

length <strong>of</strong> ei<strong>the</strong>r <strong>the</strong> sample or <strong>the</strong> standard is not<br />

required to validate <strong>the</strong> algorithm. Once algorithm is<br />

validated, it will be applied to calculate concentration<br />

<strong>of</strong> actual ambient PM2.5.<br />

REFERENCES<br />

[1] Boudel, R., Fox, D., Turner, B. and Stern, A.,<br />

1994. Fundamentals <strong>of</strong> air pollution. Academic<br />

Press, 22-34.<br />

[2] Lall, R., Kendall, M., Ito, K. and Thurston, G.,<br />

2004. Estimation <strong>of</strong> historical annual PM2.5 exposures<br />

for health effects assessment. Atmospheric<br />

Environment 38, 5217-5226.<br />

[3] EPA, May 9th <strong>2008</strong>. PM Standards. EPA.<br />

www.epa.gov/air/particlepollution/standards.html<br />

[4] NYS Department <strong>of</strong> Environmental<br />

Conservation, <strong>2008</strong>. Annual monitoring network<br />

plan: New York State ambient air monitoring<br />

program.<br />

[5] Solomon, P and Allen, D., 2004. Special Issue<br />

<strong>of</strong> Aerosol Science and Technology on findings<br />

from <strong>the</strong> fine particulate matter Supersites<br />

Programs. Aerosol Science and Technology 38<br />

(S1), 1-4.<br />

[6] CCOHS, March 22 1999. How do particles<br />

enter <strong>the</strong> respiratory system? Canadian Center for<br />

Occupational Health and Safety.<br />

http://gala.ccohs.ca/oshanswers/chemicals/how_do<br />

.html


Optimizing <strong>the</strong> Python interpreter: Identifying performance<br />

bottlenecks.<br />

Stefan Brunthaler and Jens Knoop (Faculty Mentor)<br />

Institut für Computersprachen<br />

TU Wien<br />

Vienna, Austria<br />

Email: stefan@brunthaler.net, knoop@complang.tuwien.ac.at<br />

Abstract — Efficient optimization <strong>of</strong> <strong>the</strong> Python interpreter<br />

requires detailed information about <strong>the</strong> presence and exact location<br />

<strong>of</strong> performance bottlenecks. This paper presents <strong>the</strong> results<br />

<strong>of</strong> our careful and detailed analysis, complemented by<br />

our selection <strong>of</strong> optimization techniques based on those observations<br />

to overcome <strong>the</strong>se bottlenecks: Speculation and Type<br />

Feedback.<br />

I. MOTIVATION<br />

This abstract paper presents preliminary results <strong>of</strong> ongoing<br />

research. Python is a dynamically, but strongly<br />

typed programming language that enjoys widespread application.<br />

One <strong>of</strong> its main goals is simplicity, which<br />

is also reflected in its run-time execution environment:<br />

an interpreter. Python programs are compiled to byte<br />

code, which models a stack instruction set architecture.<br />

Byte-code execution uses a simple and straight forward<br />

switch-dispatch technique. Similar programming languages,<br />

like Perl share <strong>the</strong>se implementation characteristics.<br />

While this simplicity has merits by itself—positive<br />

effects for maintenance and extensibility—, it is also<br />

detrimental to <strong>the</strong> interpreters performance.<br />

According to [1], <strong>the</strong>re already exist some specific optimizations<br />

for interpreters, such as threaded code [2],[3],<br />

stack caching [4], and superinstructions [5]. All <strong>the</strong>se<br />

optimizations are based on <strong>the</strong> premise that especially instruction<br />

dispatch, i.e. getting from one byte code implementation<br />

to ano<strong>the</strong>r, is <strong>the</strong> most costly operation within<br />

an interpreter. This has also been reported recently in [6].<br />

While this is certainly true for some cases, like Java and<br />

OCaml, we show that it is specifically not true for <strong>the</strong><br />

Python programming language.<br />

In this paper we present our findings <strong>of</strong> a quantitative<br />

analysis <strong>of</strong> Python byte-code frequencies and execution<br />

times. We use <strong>the</strong>se results to argue our choice in optimization<br />

techniques, which will be implemented in subsequent<br />

phases <strong>of</strong> research.<br />

II. QUANTITATIVE ANALYSIS<br />

A. HYPOTHESES<br />

For efficient optimization it is a prerequisite to have a<br />

careful and detailed analysis <strong>of</strong> <strong>the</strong> status quo. Based on<br />

41<br />

its results, estimating <strong>the</strong> potential <strong>of</strong> optimization techniques<br />

is possible. Specifically we are interested in determining<br />

whe<strong>the</strong>r speculation and type feedback are good<br />

candidates for optimizing Python. Our hypo<strong>the</strong>ses for<br />

“measuring” <strong>the</strong>ir optimization potential were:<br />

– For verifying speculation’s potential we measure<br />

raw byte-code frequency in standard libraries.<br />

Since optimization potential is directly proportional<br />

to <strong>the</strong> corresponding byte code frequency, optimization<br />

for infrequent use, i.e. infrequent byte<br />

codes, yields unsatisfactory results.<br />

– For verifying type feedback’s potential we provide<br />

a detailed measurement <strong>of</strong> execution times. In order<br />

to make judgments <strong>of</strong> merits <strong>of</strong> optimization<br />

techniques, it is necessary to locate <strong>the</strong> bottlenecks<br />

within <strong>the</strong> execution pr<strong>of</strong>ile.<br />

B. PROCEDURE<br />

Since we are analyzing for two separate hypo<strong>the</strong>ses, we<br />

also need separate procedures. First we have implemented<br />

a byte code counting system, which recursively<br />

traverses library trees and counts byte code occurrences<br />

per file. These occurrences are grouped by enclosing<br />

modules/directories, in order to check whe<strong>the</strong>r specific<br />

applications, e.g. scientific modules, have different byte<br />

code occurrence patterns. This was done using a standard<br />

Python 2.5 executable as distributed with Ubuntu<br />

8.04. Several libraries were analyzed, such as scipy, a<br />

scientific computing package, PIL, <strong>the</strong> Python Imaging<br />

Library, and twisted, a networking library.<br />

For testing our second hypo<strong>the</strong>sis, we used <strong>the</strong> recent<br />

Python 3.0 beta 3 source code. It already contains<br />

a high resolution timing mechanism based on Intel’s<br />

Time Stamp Counter feature. (Henceforth abbreviated by<br />

TSC.) It counts processor cycles for <strong>the</strong> interpreters main<br />

loop, and its switch-dispatch architecture. We extended<br />

<strong>the</strong> measurement to also include detailed measurement<br />

within <strong>the</strong> separate case/break blocks inside <strong>the</strong> switch<br />

statement. We used this modified interpreter version to<br />

collect samples on <strong>the</strong> pystone benchmark.<br />

Both measurements were done using an Intel Pentium<br />

4 CPU, clocked at 3GHz, running Xubuntu 8.04, with<br />

kernel version 2.6.24-19.


C. RESULTS<br />

For our first part, we counted <strong>the</strong> byte code instructions<br />

used in various libraries, e.g. we found that <strong>of</strong><br />

<strong>the</strong> 664.542 byte codes we counted, 56.832 belonged<br />

to <strong>the</strong> CALL FUNCTION instruction. Likely candidates<br />

for speculation, such as CALL FUNCTION, or<br />

GET ITER/FOR ITER (for generator based speculation)<br />

instructions ranked high, i.e. all <strong>of</strong> <strong>the</strong>m scored above <strong>the</strong><br />

3 rd quartile <strong>of</strong> measured byte code instruction frequencies<br />

1 . Consequently, speculation is a viable option for<br />

parallelization on a chip multiprocessor.<br />

For <strong>the</strong> second part, we ran <strong>the</strong> pystone benchmark in<br />

our modified version <strong>of</strong> <strong>the</strong> Python interpreter, which collected<br />

9 million samples <strong>of</strong> CPU cycle counts for wholeloop,<br />

switch dispatch and operation execution. Figure 1<br />

shows a histogram <strong>of</strong> median cycle counts within those<br />

samples. As we can see, operation execution takes up a<br />

major part <strong>of</strong> <strong>the</strong> CPU cycles, about 50%. (See Table 1<br />

for details.) Python’s execution model relies on run-time<br />

binding functionality to byte codes, e.g. a BINARY ADD<br />

instruction will for every execution instance determine<br />

which actual “add” operation to execute. Several checks<br />

take care <strong>of</strong> properly choosing <strong>the</strong> operation implementation,<br />

in our addition example, <strong>the</strong> Python interpreter will<br />

check argument types, operation availability and implementation,<br />

and whe<strong>the</strong>r a subtype relation holds between<br />

<strong>the</strong> arguments. Nothing is cached and no run-time feedback<br />

enables to cache any <strong>of</strong> those computations. Following<br />

[7], <strong>the</strong> optimization potential for type feedback<br />

in such a case is substantial.<br />

Section Min 1 st Qu Median 3 rd Qu<br />

Op-Execution 75 568 1076 2052<br />

Dispatch 84 812 1296 2084<br />

Whole Loop 84 1576 2568 4156<br />

Table 1: Contains <strong>the</strong> minimum, 1 st quartile, median, and<br />

3 rd quartile values for measured CPU cycles per code<br />

section.<br />

III. CONCLUSIONS<br />

We presented <strong>the</strong> results <strong>of</strong> our quantitative analysis.<br />

In particular, we have shown where Pythons interpreter<br />

spends most <strong>of</strong> its time: in op-code execution. This is a<br />

contradictory result to [6]. We assume that this is due to<br />

<strong>the</strong> JVM instruction set having semantics closer to a real<br />

machine. Never<strong>the</strong>less, without fur<strong>the</strong>r analysis no satisfactory<br />

answer can be found. As a direct consequence,<br />

however, techniques focusing on improving <strong>the</strong> instruction<br />

dispatch <strong>of</strong> <strong>the</strong> interpreter are unlikely to result in<br />

1 Additional material: http://www.complang.tuwien.ac.<br />

at/knoop/jsc08/data.html<br />

42<br />

CPU Cycles (TSC)<br />

0 200 400 600 800 1000 1200<br />

Operation Dispatch<br />

Code Section<br />

Whole Loop<br />

Figure 1: Histogram <strong>of</strong> median CPU cycles per code section<br />

expected speedups. We identified type feedback [7] to be<br />

particularly well suited instead. Our results on measuring<br />

byte code frequency also indicate that speculation is<br />

a viable parallelization technique for Python. Given <strong>the</strong><br />

advent <strong>of</strong> chip-multiprocessors <strong>the</strong> idea <strong>of</strong> adding speculation<br />

to <strong>the</strong> interpreter becomes even more attractive. Interesting<br />

questions remain. For example, taking a second<br />

look at Figure 1, it is striking how many time is spent before<br />

and after <strong>the</strong> switch-dispatch—this requires fur<strong>the</strong>r<br />

investigation. Ano<strong>the</strong>r example would be to drill down<br />

costs within operation execution, e.g. how expensive is<br />

operand fetching or garbage collection.<br />

REFERENCES<br />

[1] M. Anton Ertl and David Gregg. The structure and<br />

performance <strong>of</strong> efficient interpreters. J. Instruction-<br />

Level Parallelism, 5, 2003.<br />

[2] James R. Bell. Threaded code. Commun. ACM,<br />

16(6):370–372, 1973.<br />

[3] M. Anton Ertl. Threaded code variations and optimizations.<br />

In EuroForth, pages 49–55, TU Wien,<br />

Vienna, Austria, 2001.<br />

[4] M. Anton Ertl. Stack caching for interpreters. In<br />

Proc. <strong>of</strong> <strong>the</strong> ACM SIGPLAN 1995 conf. on Programming<br />

language design and implementation, pages<br />

315–327, 1995.<br />

[5] M. Anton Ertl and David Gregg. Combining stack<br />

caching with dynamic superinstructions. In Proc. <strong>of</strong><br />

<strong>the</strong> 2004 workshop on Interpreters, virtual machines<br />

and emulators, pages 7–14, 2004.<br />

[6] Yunhe Shi, Kevin Casey, M. Anton Ertl, and David<br />

Gregg. Virtual machine showdown: Stack versus<br />

registers. ACM Trans. Archit. Code Optim., 4(4):1–<br />

36, <strong>2008</strong>.<br />

[7] Urs Hölzle and David Ungar. Optimizing<br />

dynamically-dispatched calls with run-time type<br />

feedback. In Proc. <strong>of</strong> <strong>the</strong> ACM SIGPLAN 1994 conf.<br />

on Programming language design and implementation,<br />

pages 326–336, 1994.


Model for Comparison <strong>of</strong> Types <strong>of</strong> Traffic<br />

Intersection Control Systems<br />

Jan Krcal, Michal Jerabek, Jana Kalikova (Faculty Mentor)<br />

Faculty <strong>of</strong> Transportation Sciences<br />

Czech Technical University in Prague<br />

Prague, Czech Republic<br />

Email: {krcal,jerabek,kalikova}@fd.cvut.cz<br />

Abstract — This article contains part <strong>of</strong> <strong>the</strong> results<br />

<strong>of</strong> a research which focused on finding <strong>the</strong> best way<br />

to manage a traffic lights controlled crossing. A<br />

simulation model, which includes <strong>the</strong> crossing and<br />

<strong>the</strong> traffic on it, has been created to achieve this.<br />

The simulation model has been created in a<br />

ma<strong>the</strong>matical program "Matlab", which currently<br />

tests static, dynamic and fuzzy control. To assess<br />

<strong>the</strong> expediency <strong>of</strong> <strong>the</strong>se types <strong>of</strong> control, tests under<br />

different levels <strong>of</strong> traffic intensity have been carried<br />

out.<br />

I. INTRODUCTION<br />

The percentage <strong>of</strong> automobile transport in <strong>the</strong><br />

transport as such is nowadays constantly increasing.<br />

As a result <strong>of</strong> this development, <strong>the</strong> requirements <strong>of</strong><br />

quality <strong>of</strong> ei<strong>the</strong>r commercial or private transport are<br />

increasing. The classical crossings are not sufficient<br />

nowadays, new roundabouts and intersections<br />

equipped with Light Signal System (LSS) are<br />

<strong>the</strong>refore more and more common because <strong>the</strong>y<br />

increase <strong>the</strong> effectiveness, traffic-carrying capacity<br />

and safety <strong>of</strong> <strong>the</strong>se points.<br />

The method <strong>of</strong> designing <strong>the</strong> optimal control <strong>of</strong> <strong>the</strong><br />

crossing is very important. This proposal should<br />

ideally enable <strong>the</strong> maximal traffic-carrying capacity<br />

toge<strong>the</strong>r with <strong>the</strong> minimum <strong>of</strong> standing vehicles and<br />

with minimal time loss. As a consequence, it should<br />

also reduce <strong>the</strong> pollution <strong>of</strong> <strong>the</strong> environment caused<br />

by burning <strong>of</strong> fuel.<br />

Apart from <strong>the</strong> fluency <strong>of</strong> <strong>the</strong> transport, <strong>the</strong><br />

optimal type <strong>of</strong> managing <strong>the</strong> crossing should also<br />

contribute to its safety and to <strong>the</strong> decrease <strong>of</strong> ei<strong>the</strong>r<br />

number or seriousness <strong>of</strong> <strong>the</strong> car accidents.<br />

There are four known types <strong>of</strong> <strong>the</strong> controlling<br />

systems nowadays: static (time dependent), dynamic<br />

(traffic dependent), fuzzy and adaptive. The first<br />

three mentioned are used as <strong>the</strong> controlling systems<br />

<strong>of</strong> <strong>the</strong> traffic crossing, whereas <strong>the</strong> adaptive<br />

controlling system is used for <strong>the</strong> whole area. The<br />

aim <strong>of</strong> <strong>the</strong> work presented in <strong>the</strong> article is to compare<br />

<strong>the</strong>se types <strong>of</strong> control for different levels <strong>of</strong> traffic,<br />

and to define and verify which type <strong>of</strong> control is<br />

43<br />

suitable for which (specific) level <strong>of</strong> traffic <strong>of</strong> <strong>the</strong><br />

road.<br />

II. THE SIMULATION MODEL OF A TRAFFIC<br />

CROSSING<br />

In every model, not only in <strong>the</strong> modeling <strong>of</strong> a<br />

traffic crossing, we try to make as precise and real as<br />

possible. It is always necessary to accept some<br />

simplifying (idealizing) assumptions. Some<br />

assumptions were idealized even in this simulation,<br />

but always in a way which does not significantly<br />

influence <strong>the</strong> objectivity <strong>of</strong> <strong>the</strong> simulation.<br />

The ma<strong>the</strong>matical program Matlab 6.5 <strong>of</strong> <strong>the</strong><br />

company MathWorks was used for <strong>the</strong> creation <strong>of</strong><br />

simulation model. The advantage <strong>of</strong> this<br />

ma<strong>the</strong>matical instrument lies primarily in its<br />

additional modules, so-called toolboxes, each <strong>of</strong><br />

which contains a particular part <strong>of</strong> a scientific<br />

interest, i.e. signal processing toolbox, image<br />

processing toolbox etc.<br />

III. INITIAL ASSUMPTION FOR THE<br />

SIMULATION<br />

For illustrative simulation <strong>of</strong> managing <strong>the</strong><br />

crossing in terms <strong>of</strong> our work, we chose three<br />

different options <strong>of</strong> traffic intensity which consider<br />

traffic in various parts <strong>of</strong> <strong>the</strong> day – morning rush<br />

hours, noon traffic and night traffic. As a crossing<br />

type, we chose common four-leg intersection (see<br />

Figure 1), i.e. four entries and exits with following<br />

intensity proportion intensity (V2 = 100%,<br />

V1 = 0,5*V2, V3 = 0,4*V2, V4 = 0,8*V2).<br />

The options <strong>of</strong> traffic intensity consequently are:<br />

1. option I – total intensity at <strong>the</strong> entry 405 uv/h<br />

(Unit Vehicle per hour) (V2 = 150 uv/h)<br />

2. option II – total intensity at <strong>the</strong> entry 810 uv/h<br />

(V2 = 300 uv/h)<br />

3. option III – total intensity at <strong>the</strong> entry 1350 uv/h<br />

(V2 = 500 uv/h)


A signal program for each <strong>of</strong> <strong>the</strong> options was<br />

computed by <strong>the</strong> Webster's method and on <strong>the</strong> basis<br />

<strong>of</strong> <strong>the</strong> program were realized <strong>the</strong> simulations <strong>of</strong> <strong>the</strong><br />

individual types <strong>of</strong> controlling systems.<br />

The length <strong>of</strong> <strong>the</strong> simulation itself was prescribed<br />

to 3600 seconds, i.e. 1 hour. It is highly convenient<br />

for <strong>the</strong> simulation because all values <strong>of</strong> saturated<br />

flow and intensity are given in unit vehicles per hour.<br />

For evaluation <strong>of</strong> <strong>the</strong> efficiency <strong>of</strong> <strong>the</strong> crossing, <strong>the</strong><br />

Performance index value (PI) is used; it considers<br />

also <strong>the</strong> ecological aspects (fuel over-consumption,<br />

higher production <strong>of</strong> CO, CHx, NOx, SO2 etc.).<br />

Following parameters for individual entries are<br />

additionally evaluated in this simulation program:<br />

− The longest queue [uv]<br />

− Period <strong>of</strong> <strong>the</strong> vehicle standing in <strong>the</strong> queue for<br />

<strong>the</strong> longest time [seconds]<br />

− Period <strong>of</strong> <strong>the</strong> vehicle standing in <strong>the</strong> queue for<br />

<strong>the</strong> shortest time [seconds]<br />

− Total delay <strong>of</strong> <strong>the</strong> vehicles [seconds]<br />

− Number <strong>of</strong> vehicles standing in <strong>the</strong> queue [uv]<br />

− Number <strong>of</strong> vehicles passing without stop [uv]<br />

− Average delay in <strong>the</strong> queue [seconds]<br />

− Average delay at <strong>the</strong> entry [seconds]<br />

− Performance Index for individual entry<br />

[seconds]<br />

V2<br />

V3<br />

V1<br />

Figure 1. Model <strong>of</strong> chosen crossing type<br />

V4<br />

44<br />

IV. CONCLUSION<br />

The aim <strong>of</strong> this research was to judge <strong>the</strong><br />

connection between static, dynamic and fuzzy<br />

controlling system <strong>of</strong> <strong>the</strong> crossing, considering <strong>the</strong><br />

simulation model <strong>of</strong> <strong>the</strong> crossing.<br />

It was <strong>the</strong>refore inevitable to create own model<br />

which would enable <strong>the</strong>se simulations. It is necessary<br />

to point out that it is not any commercial product.<br />

The signal program, computed by <strong>the</strong> Webster<br />

method, specified <strong>the</strong> Length <strong>of</strong> <strong>the</strong> Cycle (LC) for<br />

all intensity levels to one value LC = 40s. The<br />

differences could be found only between <strong>the</strong><br />

proportions between signal "go" in corresponding<br />

phases. The values <strong>of</strong> Performance index were<br />

computed for each option <strong>of</strong> intensity and controlling<br />

system.<br />

As a consequence <strong>of</strong> <strong>the</strong> table data (see Table 1),<br />

fuzzy controlling system is <strong>the</strong> most suitable for <strong>the</strong><br />

low intensity as it has 21% better results than <strong>the</strong><br />

static controlling system. The dynamic has 17%<br />

better results. The utility <strong>of</strong> <strong>the</strong> fuzzy control<br />

disappears with <strong>the</strong> increasing intensity, which<br />

means that <strong>the</strong> result <strong>of</strong> <strong>the</strong> middle intensity is worse<br />

than <strong>the</strong> dynamic control and for <strong>the</strong> high intensity<br />

worse even that <strong>the</strong> static control. The dynamic<br />

controlling system has always better results than <strong>the</strong><br />

static control, but it draws near <strong>the</strong> static control with<br />

<strong>the</strong> increasing intensity.<br />

The results related to <strong>the</strong> static controlling system<br />

(100%) in <strong>the</strong> table below.<br />

Intensity [uv/h] 405 810 1350<br />

Static control 100% 100% 100%<br />

Dynamic control 83% 89% 95%<br />

Fuzzy control 79% 96% 101%<br />

Table 1. Table <strong>of</strong> result <strong>of</strong> simulation<br />

REFERENCES<br />

[1] P�ibyl P., Svítek M.: Inteligentní dopravní<br />

systémy, BEN, Praha 2001.<br />

[2] Vysoký P.: Fuzzy �ízení, CVUT, Praha 1996.<br />

[3] Tan K.K., Khalid M., and Yus<strong>of</strong> R.: Intelligent<br />

Traffic Lights Control by Fuzzy Logic,<br />

Malaysian Journal <strong>of</strong> Computer Science, Nov.<br />

1995.<br />

[4] Trabia M. B.: A two-stage fuzzy logic controller<br />

for traffic signals, Transportation Research Part<br />

C 7 (1999), pp. 353 367.<br />

[5] Könönen V, Nevala R, Mäenpää, M.:<br />

Development <strong>of</strong> Fuzzy Signal Controller,<br />

Transportation Engineering, Publication 98,<br />

Espoo 1999, pp. 7.


An integrated source-to-source approach for WCET analysis<br />

Adrian Prantl and Jens Knoop (Faculty Mentor)<br />

Institut für Computersprachen<br />

Technische Universität Wien<br />

Vienna, Austria<br />

Email: {adrian,knoop}@complang.tuwien.ac.at<br />

Abstract — The safety <strong>of</strong> our everyday life depends increasingly<br />

on <strong>the</strong> correct functioning <strong>of</strong> embedded s<strong>of</strong>tware systems<br />

that control more and more appliances. Many <strong>of</strong> <strong>the</strong>se s<strong>of</strong>tware<br />

systems are time-critical. Hence, computations performed<br />

need not only to be correct, but must also return results in a<br />

timely fashion. Worst-case execution time (WCET) analysis<br />

is concerned with computing tight upper bounds for <strong>the</strong> execution<br />

time <strong>of</strong> a system in order to provide formal guarantees for<br />

<strong>the</strong> proper timing behaviour <strong>of</strong> a system. In this paper, we introduce<br />

TuBound – a conceptually new approach, introducing<br />

source-to-source flow analysis, optimization and transformation<br />

<strong>of</strong> analysis information in <strong>the</strong> workflow <strong>of</strong> WCET analysis.<br />

I. MOTIVATION<br />

Embedded s<strong>of</strong>tware systems are virtually ubiquitous today<br />

to control <strong>the</strong> functioning <strong>of</strong> technical devices we<br />

routinely use and rely on in our everyday life. With <strong>the</strong><br />

emergence <strong>of</strong> safety-critical technologies like fly-by-wire<br />

in <strong>the</strong> avionics field or drive-by-wire, which is already<br />

foreseeable in <strong>the</strong> automotive field, <strong>the</strong> correct functioning<br />

<strong>of</strong> <strong>the</strong>se s<strong>of</strong>tware systems is more important than<br />

ever.<br />

In a safety-critical real-time system, <strong>the</strong> timeliness <strong>of</strong><br />

a computation is equally important as its correctness.<br />

Worst-case execution time (WCET) analysis is concerned<br />

with providing formal guarantees for <strong>the</strong> proper timing<br />

behaviour <strong>of</strong> a system by computing tight upper bounds<br />

for <strong>the</strong> execution time <strong>of</strong> a system.<br />

State-<strong>of</strong>-<strong>the</strong>-art WCET analysis tools rely on supporting<br />

analyses and manual annotations to provide <strong>the</strong>m<br />

with information on <strong>the</strong> execution behaviour <strong>of</strong> <strong>the</strong> program,<br />

such as upper bounds for loops. Usually, this is<br />

done on <strong>the</strong> low level binary code, which, toge<strong>the</strong>r with<br />

<strong>the</strong> analysis information, is <strong>the</strong>n combined with a model<br />

<strong>of</strong> <strong>the</strong> target hardware to formulate an optimization problem<br />

describing <strong>the</strong> worst-case behaviour <strong>of</strong> <strong>the</strong> program.<br />

While it is mandatory to perform this last step on <strong>the</strong> final<br />

binary code, a high-level representation, such as <strong>the</strong><br />

source code, generally contains much more information,<br />

such as types, thus rendering possible more powerful<br />

analyses..<br />

The TuBound approach [1], which we pursue in <strong>the</strong><br />

CoSTA 1 project, is to lift annotations and <strong>the</strong> supporting<br />

1 Austrian science fund (FWF) project No P18925-N13: Compiler<br />

support for timing analysis. http://costa.tuwien.ac.at/<br />

45<br />

analyses to <strong>the</strong> source code level <strong>of</strong> a program, while still<br />

performing WCET analysis on <strong>the</strong> machine code level.<br />

In addition to keeping <strong>the</strong> precision <strong>of</strong> low-level WCET<br />

analysis, this has <strong>the</strong> following benefits:<br />

– Convenience and Ease: For <strong>the</strong> user, annotating <strong>the</strong><br />

source code is generally easier and less demanding<br />

as annotating <strong>the</strong> assembler output <strong>of</strong> <strong>the</strong> compiler.<br />

– Reuse and Portability: Source code annotations,<br />

which specify hardware-independent behaviour,<br />

can directly be reused when <strong>the</strong> program is ported<br />

to ano<strong>the</strong>r target hardware.<br />

– Feedback and Tuning: Source code annotations can<br />

be used to present <strong>the</strong> results <strong>of</strong> static analyses to<br />

<strong>the</strong> programmer for inspection and fur<strong>the</strong>r manual<br />

refinement.<br />

The information computed on <strong>the</strong> source level and<br />

annotated in <strong>the</strong> code is <strong>the</strong>n conjointly transformed<br />

throughout <strong>the</strong> compilation and optimization <strong>of</strong> <strong>the</strong> program<br />

to <strong>the</strong> binary code level to make it accessible to <strong>the</strong><br />

WCET analysis component <strong>of</strong> our TuBound tool. Currently,<br />

all optimizations are performed on <strong>the</strong> source code<br />

level, too. The transformed and optimized code is <strong>the</strong>n<br />

fed into a specific WCET-aware variant <strong>of</strong> <strong>the</strong> Gnu-C<br />

compiler [2], which is tailored for preserving <strong>the</strong> validity<br />

<strong>of</strong> code annotations that are provided within <strong>the</strong> compiled<br />

code. The binary code it generates is finally passed to<br />

a retargetable WCET analysis component, which computes<br />

<strong>the</strong> desired upper bound <strong>of</strong> <strong>the</strong> execution time <strong>of</strong><br />

<strong>the</strong> program in <strong>the</strong> worst case. Currently, this is <strong>the</strong><br />

WCET analyzer CALCWCET167.<br />

II. THE SOURCE-TO-SOURCE APPROACH<br />

The heart <strong>of</strong> our high-level approach was implemented<br />

with <strong>the</strong> help <strong>of</strong> <strong>the</strong> source-to-source analysis and transformation<br />

framework SATIrE [3]. This framework integrates<br />

<strong>the</strong> program analyzer generator PAG and SWI-<br />

Prolog with <strong>the</strong> C++ source-to-source compiler framework<br />

LLNL-ROSE.<br />

A. SOURCE-TO-SOURCE ANALYSIS<br />

Utilizing <strong>the</strong> SATIrE framework, we have implemented<br />

<strong>the</strong> following supporting analyses, <strong>the</strong> results <strong>of</strong> which<br />

are attached as annotations to <strong>the</strong> abstract syntax tree and<br />

as pragma-statements in <strong>the</strong> source code:


0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

CalcWCET<br />

GCC<br />

TeBo<br />

PAG-Interval<br />

ICFG<br />

Rose<br />

fibcall<br />

expint<br />

cover<br />

bsort100<br />

qurt<br />

lcdnum<br />

crc<br />

cnt<br />

TuBound: Time spent per phase<br />

st<br />

sqrt<br />

recursion<br />

Seconds/Benchmark<br />

ns<br />

matmult<br />

jfdctint<br />

fdct<br />

whet<br />

– An interprocedural interval analysis that computes<br />

variable-interval pairs for each sequence point in<br />

<strong>the</strong> program.<br />

– A loop-bound analysis that computes upper bounds<br />

for all iteration variable-based loops.<br />

– A generalized loop constraint analysis that uses<br />

constraint logic programming to calculate precise<br />

flow information for nested loops.<br />

B. FLOW CONSTRAINT TRANSFORMATION<br />

While fully automatic analysis is <strong>of</strong>ten sufficient, <strong>the</strong> programmer<br />

can use his knowledge about <strong>the</strong> execution environment<br />

in a semi-automatic process to fur<strong>the</strong>r extend<br />

<strong>the</strong> power <strong>of</strong> <strong>the</strong> analysis. Traditionally, this manual annotation<br />

<strong>of</strong> <strong>the</strong> program is also done on <strong>the</strong> binary code.<br />

The philosophy <strong>of</strong> TuBound is to also lift <strong>the</strong> annotation<br />

process to <strong>the</strong> source level.<br />

Since compiler optimizations can modify <strong>the</strong> control<br />

flow <strong>of</strong> a program, <strong>the</strong>y also have <strong>the</strong> potential to invalidate<br />

source code annotations describing <strong>the</strong> control flow.<br />

In TuBound, <strong>the</strong> FlowTrans component takes care <strong>of</strong> this<br />

by transforming flow constraints according to <strong>the</strong> performed<br />

optimizations. This is realized with <strong>the</strong> help <strong>of</strong> an<br />

optimization trace and a rule system that defines <strong>the</strong> correct<br />

transformation <strong>of</strong> flow constraints for each type <strong>of</strong><br />

optimization. By performing optimizations on <strong>the</strong> source<br />

code, <strong>the</strong> subsequent compiler can be instructed to perform<br />

a direct translation to binary code, thus keeping <strong>the</strong><br />

flow information intact.<br />

III. RESULTS<br />

TuBound successfully took part in <strong>the</strong> WCET Tool Challenge<br />

<strong>2008</strong>, a competition for academic and commercial<br />

WCET analysis tools; detailed results can be found<br />

in [4]. We also evaluated our approach using a selection<br />

<strong>of</strong> standardized WCET-analysis benchmarks compiled<br />

by Mälardalen University. In Figure 1, <strong>the</strong> upper<br />

graph compares <strong>the</strong> run time <strong>of</strong> <strong>the</strong> components <strong>of</strong><br />

1e+09<br />

1e+08<br />

1e+07<br />

1e+06<br />

100000<br />

10000<br />

1000<br />

crc<br />

cnt<br />

Figure 1: Measurements<br />

46<br />

unoptimimized<br />

loop-opt.<br />

lowlevel-opt<br />

loop-opt+lowlevel-opt<br />

fibcall<br />

expint<br />

cover<br />

bsort100<br />

qurt<br />

lcdnum<br />

TuBound: Analyzed WCET<br />

st<br />

sqrt<br />

recursion<br />

Cycles/Benchmark<br />

ns<br />

matmult<br />

jfdctint<br />

fdct<br />

whet<br />

TuBound, whereas <strong>the</strong> lower graph depicts <strong>the</strong> effect <strong>of</strong><br />

optimizations on <strong>the</strong> analyzed WCET <strong>of</strong> <strong>the</strong> benchmarks.<br />

IV. CONCLUSION<br />

With TuBound we have successfully demonstrated <strong>the</strong><br />

feasibility <strong>of</strong> a source-to-source approach to WCET analysis,<br />

by combining <strong>the</strong> advantage <strong>of</strong> low level WCET<br />

analysis with high level source code annotations and optimizing<br />

compilation. The flow constraint transformation<br />

framework FlowTrans ensures that annotations are<br />

transformed according to <strong>the</strong> optimization trace as provided<br />

by <strong>the</strong> high-level optimizer. Recent works such<br />

as [5] show that better integration <strong>of</strong> optimizing compilation<br />

and WCET analysis is indeed <strong>of</strong> high interest for <strong>the</strong><br />

community. With our source-based approach, we now<br />

have a portable solution to close <strong>the</strong> gap between source<br />

code annotations and machine-specific WCET analysis.<br />

REFERENCES<br />

[1] Adrian Prantl, Markus Schordan, and Jens Knoop.<br />

TuBound – A Conceptually New Tool for Worst-Case Execution<br />

Time Analysis. In Proc. 8th International Workshop<br />

on Worst-Case Execution Time Analysis (WCET <strong>2008</strong>),<br />

Prague, <strong>2008</strong>.<br />

[2] Raimund Kirner. Extending Optimising Compilation to<br />

Support Worst-Case Execution Time Analysis. PhD <strong>the</strong>sis,<br />

Technische Universität Wien, 2003.<br />

[3] Markus Schordan. Source-To-Source Analysis with<br />

SATIrE – an Example Revisited. In Proc. <strong>of</strong> Dagstuhl<br />

Seminar 08161: Scalable Program Analysis. Germany,<br />

Dagstuhl, April <strong>2008</strong>.<br />

[4] Niklas Holsti, Jan Gustafsson, and Guillem Bernat (eds).<br />

WCET tool challenge <strong>2008</strong>: Report. In Proc. 8th International<br />

Workshop on Worst-Case Execution Time Analysis<br />

(WCET <strong>2008</strong>), Prague, <strong>2008</strong>.<br />

[5] Heiko Falk, Paul Lokuciejewski, and Henrik Theiling. Design<br />

<strong>of</strong> a WCET-Aware C Compiler. In Soonhoi Ha and<br />

Samarjit Chakraborty, editors, ESTImedia, pages 121–126.<br />

IEEE Computer Society, 2006.


Cooperative Media Access Control in Wireless Networks<br />

Mathias Offner and Helmut Adam (Faculty Mentor)<br />

Networked and Embedded Systems<br />

University <strong>of</strong> Klagenfurt, Austria<br />

Email: m<strong>of</strong>fner@edu.uni-klu.ac.at, helmut.adam@uni-klu.ac.at<br />

Abstract — We present CMAC/DRS, a cooperative relaying enhanced<br />

MAC protocol which solves mandatory tasks in cooperative<br />

networks in a throughput efficient way. Our main focus is <strong>the</strong><br />

relay selection process; we only select relays if requested by <strong>the</strong><br />

destination and try to minimize <strong>the</strong> collision probability <strong>of</strong> messages<br />

in this process. We compare CMAC/DRS with CMAC/ARS<br />

and standard non-cooperative IEEE 802.11. Our simulation results<br />

indicate considerable throughput benefits <strong>of</strong> CMAC/DRS<br />

compared to <strong>the</strong> reference protocols.<br />

I. INTRODUCTION<br />

In cooperative relaying, as depicted in Figure 1, a node denoted<br />

as relay R overhears <strong>the</strong> direct transmission between<br />

S and D and, on demand, forwards this data to D. D combines<br />

both messages to mitigate small scale fading effects.<br />

R is chosen from <strong>the</strong> common neighbors <strong>of</strong> S and D based<br />

on current Channel State Information (CSI) from <strong>the</strong> candidates<br />

to S and D. RTS and CTS messages <strong>of</strong> IEEE 802.11<br />

protocols can be exploited at relaying candidates to estimate<br />

<strong>the</strong>ir CSI to S and D trying to keep <strong>the</strong> overhead introduced<br />

by <strong>the</strong> relay selection small. Fur<strong>the</strong>rmore, a selected relay<br />

needs also to reserve <strong>the</strong> channel in its neighborhood. Thus,<br />

it is appealing to combine cooperative relaying protocols a<br />

similar to IEEE 802.11 MAC protocol [1, 2, 3].<br />

The performance <strong>of</strong> cooperative relaying schemes mainly<br />

depends on <strong>the</strong> relay selection. Distributed relay selection<br />

requires <strong>the</strong> exchange <strong>of</strong> messages which are prone to collide.<br />

Due to collision it can happen that no relay can be<br />

selected and cooperation is not available.<br />

In this paper we introduce Cooperative Medium Access<br />

Control with Dynamic Relay Selection (CMAC/DRS)<br />

which addresses <strong>the</strong> problem <strong>of</strong> relay selection and tries to<br />

avoid collisions between potential relaying nodes. Fur<strong>the</strong>rmore,<br />

it only uses cooperation if required and skips relay<br />

selection o<strong>the</strong>rwise.<br />

S D<br />

R<br />

Figure 1: Basic cooperative relaying scheme<br />

II. PROTOCOL DESCRIPTION: CMAC/DRS<br />

CMAC/DRS which is based on CMAC/ARS [1] works as<br />

depicted in Figure 2: at <strong>the</strong> beginning <strong>the</strong> source broadcasts<br />

an RTS frame to initiate <strong>the</strong> transmission with a certain destination<br />

and reserves <strong>the</strong> channel in its neighborhood for <strong>the</strong><br />

time <strong>of</strong> <strong>the</strong> relay selection. The destination replies with a<br />

Conditional Clear to Send (CCTS) frame if it is ready to<br />

send and occupies <strong>the</strong> channel for <strong>the</strong> expected transmission<br />

time. If <strong>the</strong> direct link between source and destination is bad<br />

a relay is selected distributively. The current relay is confirmed<br />

within <strong>the</strong> data transmission <strong>of</strong> <strong>the</strong> source. The data<br />

packet also extends <strong>the</strong> source’s channel reservation until<br />

47<br />

planned ACK reception time. After data transmission <strong>the</strong> relay<br />

waits for an ACK from <strong>the</strong> destination. If <strong>the</strong> relay does<br />

not receive <strong>the</strong> ACK within a certain period <strong>of</strong> time, it assumes<br />

a failure and forwards <strong>the</strong> overheard message. At <strong>the</strong><br />

end an ACK frame completes <strong>the</strong> cooperative transmission.<br />

A. COOPERATION ON DEMAND<br />

Cooperative Relaying consumes time and energy for relay<br />

selection and overhearing packet transmissions. However,<br />

cooperation is only used when <strong>the</strong> direct transmission<br />

from source to destination does not succeed. Thus, in<br />

CMAC/DRS <strong>the</strong> destination uses <strong>the</strong> RTS message to estimate<br />

<strong>the</strong> current Packet Error Rate (PER) <strong>of</strong> <strong>the</strong> sourcedestination<br />

link. If <strong>the</strong> current PER drops below an application<br />

depended threshold <strong>the</strong> relay selection is skipped which<br />

is reported via <strong>the</strong> CCTS message. Thus cooperation is only<br />

used on demand [4].<br />

B. DISTRIBUTED RELAY SELECTION<br />

Relay selection is a significant topic in cooperative relaying.<br />

Main challenge is to choose a node that can effectively<br />

improve data transmission out <strong>of</strong> a set <strong>of</strong> potential relays.<br />

CMAC/DRS uses three busy tone slots and a static number<br />

<strong>of</strong> RRTS slots to solve this problem. Busy tones are used to<br />

signalize that <strong>the</strong>re exist neighboring nodes that may help to<br />

send packets in a cooperative manner. Moreover those tones<br />

assist to separate relays into sets with different capabilities,<br />

e.g. relays that provide an outage rate <strong>of</strong> 0.01, 0.1 and 0.5.<br />

Busy tones are consulted to estimate <strong>the</strong> number <strong>of</strong> competing<br />

relays [5]. The estimated number <strong>of</strong> relaying candidates<br />

is reported by <strong>the</strong> source using a NRC (Number <strong>of</strong> Relay<br />

Candidates) message. Potential relays use this information<br />

to derive <strong>the</strong>ir slot transmission probability in <strong>the</strong> contention<br />

window. This reduces <strong>the</strong> probability <strong>of</strong> colliding RRTS<br />

(Relay Ready To Send) messages. For instance, if 13 relays<br />

are available each <strong>of</strong> <strong>the</strong>m broadcast an RRTS message<br />

at N available slots with a probability <strong>of</strong> 1/13. This equation<br />

maximizes <strong>the</strong> probability that only one single frame is<br />

send at <strong>the</strong> same time [6].<br />

III. SIMULATION<br />

We implemented CMAC/DRS and CMAC/ARS [1] in a<br />

modified version <strong>of</strong> an 802.11 simulator1 for ad hoc networks.<br />

CMAC/ARS has not been implemented one-to-one<br />

in our simulator because this approach presumes a few assumptions<br />

that can not be applied practically [7]. The main<br />

difference between both implemented cooperative protocols<br />

is that CMAC/ARS uses constant slot transmission probabilities<br />

for RRTS frames. Therefore, <strong>the</strong> number <strong>of</strong> available<br />

slots significantly affects <strong>the</strong> success <strong>of</strong> relay selection.<br />

In order to provide a fair comparison, both CMAC/DRS and<br />

CMAC/ARS use seven RRTS slots.<br />

For our simulations we randomly place nodes on an<br />

100 × 100 m2 area, where each 5 msec a packet with size<br />

1024 bytes is injected. Source and destination <strong>of</strong> this transmission<br />

are random. Relays perform decode and forward.<br />

Fur<strong>the</strong>r assumptions are given in [7]. Figure 3 depicts a<br />

1 http://wiki.uni.lu/secan-lab/802.11+Network+Simulator.html


throughput comparison <strong>of</strong> CMAC/DRS, CMAC/ARS and<br />

802.11. IEEE 802.11 achieves a fairly low throughput<br />

due to many required retransmissions. Throughput <strong>of</strong><br />

CMAC/ARS raises continuously from sparse networks with<br />

only 10 available nodes to networks with 40 radios, afterwards<br />

it drops instantly. The value decreases mainly because<br />

<strong>of</strong> collisions <strong>of</strong> RRTS frames. Consequently, relay<br />

selection fails and data frames cannot be sent in a cooperative<br />

manner. CMAC/DRS reaches at a point <strong>of</strong> 25 network<br />

nodes a throughput value that stays nearly constant<br />

around 620 kB/s. With increasing node density benefits <strong>of</strong><br />

dynamic slot transmission probabilities arise, as <strong>the</strong> probability<br />

<strong>of</strong> RRTS frame collisions is minimized even in dense<br />

networks.<br />

throughput (kB/s)<br />

700<br />

600<br />

500<br />

400<br />

300<br />

200<br />

100<br />

0<br />

0 20 40 60 80 100 120<br />

number <strong>of</strong> network nodes<br />

Figure 2: CMAC/DRS frame exchange sequence<br />

CMAC/ARS<br />

CMAC/DRS<br />

802.11<br />

Figure 3: Throughput comparison<br />

Figure 4 illustrates <strong>the</strong> dropped packet rates <strong>of</strong><br />

CMAC/DRS, CMAC/ARS and IEEE 802.11 in <strong>the</strong> given<br />

scenario. A packet is dropped whenever <strong>the</strong> number <strong>of</strong> retransmissions<br />

exceeds seven. While rates <strong>of</strong> CMAC/ARS<br />

and CMAC/DRS are fairly low and show a dependency on<br />

<strong>the</strong> network density, <strong>the</strong> packet drop rate <strong>of</strong> 802.11 is constantly<br />

around 0.04. Regarding CMAC/ARS we can recognize<br />

that <strong>the</strong> ratio <strong>of</strong> dropped packets raises at networks with<br />

70 or more radios. Again <strong>the</strong> reason are RRTS collisions as<br />

nodes have to transmit directly if relay selection fails. It appears<br />

that more packets are dropped at sparse networks as<br />

<strong>the</strong>re are less potential relays available. Overall simulations<br />

have shown that CMAC/DRS outperforms CMAC/ARS in<br />

many comparison criteria by far and <strong>the</strong>refore it is a promising<br />

state <strong>of</strong> <strong>the</strong> art cooperative MAC layer protocol.<br />

IV. CONCLUSION<br />

This paper shows significant benefits <strong>of</strong> cooperative<br />

communications demonstrated by simulation results.<br />

CMAC/DRS represents an efficient MAC layer protocol<br />

to provide relaying in wireless networks. Fundamentals <strong>of</strong><br />

this proposed scheme are distributed relay selection with<br />

48<br />

ratio <strong>of</strong> dropped packets<br />

0.1<br />

0.01<br />

0.001<br />

1e-04<br />

1e-05<br />

1e-06<br />

CMAC/ARS<br />

1e-07<br />

0 20 40 60 80 100 120<br />

number <strong>of</strong> network nodes<br />

Figure 4: Dropped packets<br />

802.11<br />

CMAC/DRS<br />

dynamic slot transmission probabilities and cooperation on<br />

demand. Evaluations demonstrate gains <strong>of</strong> CMAC/DRS by<br />

terms <strong>of</strong> throughput and outage.<br />

ACKNOWLEDGEMENTS<br />

This work has been sponsored within <strong>the</strong> scope <strong>of</strong> <strong>the</strong> master’s<br />

<strong>the</strong>sis by Orange Labs.<br />

REFERENCES<br />

[1] C.-T. Chou, J. Yang, and D. Wang. Cooperative MAC<br />

Protocol with Automatic Relay Selection in Distributed<br />

Wireless Networks. In Proc. IEEE Intl. Conf. on Pervasive<br />

Computing and Communications Workshops,pages<br />

526–531, Washington, DC, USA, March 2007.<br />

[2] A. Azgin, Y. Altunbasak, and G. Alregib. Cooperative<br />

MAC and routing protocols for wireless ad hoc networks.<br />

In Proc. IEEE GLOBECOM, volume 3, pages<br />

2854–2859, November 2005.<br />

[3] S.Moh,C.Yu,S.M.Park,H.N.Kim,andJ.Park.CD-<br />

MAC: Cooperative Diversity MAC for Robust Communication<br />

in Wireless Ad Hoc Networks. In Proc. IEEE<br />

ICC, pages 3636–3641, Glasgow, Scotland, June 2007.<br />

[4] H. Adam, C. Bettstetter, and S. M. Senouci. Adaptive<br />

relay selection in cooperative wireless networks. In<br />

Proc. IEEE PIMRC, Cannes, France, September <strong>2008</strong>.<br />

[5] Sarkar S. Khanna, S. and I. Shin. An energy measurement<br />

based collision resolution protocol. In Proc. <strong>of</strong><br />

ITC conference, Berlin, Germany, September 2003.<br />

[6] S. S. Lam. Packet switching in a multi-access broadcast<br />

channel with application to satellite communication in<br />

a computer network. PhD <strong>the</strong>sis, University <strong>of</strong> California,<br />

Los Angeles, 1974.<br />

[7] M. Offner. Media access control in wireless cooperative<br />

relaying systems. Master’s <strong>the</strong>sis, University <strong>of</strong><br />

Klagenfurt, <strong>2008</strong>.


Jessine: An Approach to Creating S<strong>of</strong>tware Applications<br />

with Rule-Based Domain Knowledge<br />

Miroslav Sabo and Jaroslav Poruban (Faculty Mentor)<br />

Faculty <strong>of</strong> Electrical Engineering and Informatics<br />

Technical University <strong>of</strong> Kosice<br />

Kosice, Slovak Republic<br />

Email: {miroslav.sabo,jaroslav.poruban}@tuke.sk<br />

Abstract — This paper points out <strong>the</strong> problem <strong>of</strong><br />

tangling <strong>of</strong> implicit knowledge about <strong>the</strong> domain<br />

with <strong>the</strong> core functionality in s<strong>of</strong>tware applications.<br />

The presented framework called Jessine enables<br />

separating <strong>of</strong> rule-based knowledge and objectoriented<br />

functionality in order to facilitate<br />

independent development and evolution. An<br />

application programming interface is provided for<br />

creating, managing and customizing <strong>the</strong> necessary<br />

components <strong>of</strong> rule systems designed with <strong>the</strong><br />

framework. Configuration <strong>of</strong> <strong>the</strong> rule systems<br />

implemented with Jessine is defined via external<br />

files outside an application code. Jessine is a<br />

conceptual framework written in Java and is<br />

utilizing full-fledged rule engine Jess.<br />

I. INTRODUCTION<br />

The complexity <strong>of</strong> s<strong>of</strong>tware applications is steadily<br />

increasing and knowledge management <strong>of</strong><br />

application domain is becoming more important.<br />

Moreover, real-world domains are subject to change<br />

and applications have to cope with <strong>the</strong>se changes in<br />

order to operate properly. Current s<strong>of</strong>tware<br />

engineering practices result in s<strong>of</strong>tware applications<br />

containing implicit knowledge about <strong>the</strong> domain that<br />

is tangled with <strong>the</strong> core functionality.<br />

The development <strong>of</strong> such a s<strong>of</strong>tware is a very<br />

complex task: <strong>the</strong> s<strong>of</strong>tware developer, who is<br />

typically a s<strong>of</strong>tware engineer but not a knowledge<br />

engineer, has to concentrate on two facets <strong>of</strong> s<strong>of</strong>tware<br />

at <strong>the</strong> same time and manually integrate <strong>the</strong>m. This<br />

violates <strong>the</strong> principle <strong>of</strong> separation <strong>of</strong> concerns,<br />

which states that <strong>the</strong> core functionality should be<br />

separated from o<strong>the</strong>r concerns or aspects, such as<br />

rule-based knowledge in this case. However, due to<br />

current s<strong>of</strong>tware engineering methodologies, domain<br />

rules are expressed implicitly, strongly coupled and<br />

even tangled in <strong>the</strong> application code. Thus, changes<br />

in <strong>the</strong> domain imply invasively changing <strong>the</strong> code,<br />

making it hard to understand, maintain, evolve and<br />

reuse.<br />

In my work I am focusing on <strong>the</strong> integration <strong>of</strong><br />

domain knowledge represented by domain rules into<br />

<strong>the</strong> s<strong>of</strong>tware applications during <strong>the</strong> implementation<br />

49<br />

phase <strong>of</strong> <strong>the</strong> s<strong>of</strong>tware development cycle. The<br />

following list mentions <strong>the</strong> main advantages <strong>of</strong><br />

decoupling domain rules from <strong>the</strong> core application:<br />

� Explicit representation <strong>of</strong> domain knowledge.<br />

� Reuse <strong>of</strong> domain rules accross different<br />

processes and applications.<br />

� Reuse <strong>of</strong> s<strong>of</strong>tware applications considering<br />

different sets <strong>of</strong> domain rules.<br />

� Clearer understanding <strong>of</strong> application<br />

behaviour.<br />

� Decreased maintenance and testing costs.<br />

II. FRAMEWORK AS A SOLUTION<br />

To promote <strong>the</strong> aforementioned features, I decided to<br />

use <strong>the</strong> rule engine for storing and processing <strong>of</strong><br />

domain rules, encapsulating and separating that way<br />

<strong>the</strong> domain knowledge from <strong>the</strong> rest <strong>of</strong> <strong>the</strong><br />

application. For seamless non-invasive integration <strong>of</strong><br />

such a separated domain knowledge back to <strong>the</strong><br />

application, I have designed and implemented<br />

specialized application framework [1]. The<br />

framework is utilizing full-fledged rule engine Jess<br />

[2], <strong>the</strong>refore <strong>the</strong> name <strong>of</strong> framework: Jessine.<br />

The main idea which is <strong>the</strong> framework based on is<br />

<strong>the</strong> elimination <strong>of</strong> writing glue code which does not<br />

contribute to any functionality <strong>of</strong> application, but<br />

serves solely to “glue toge<strong>the</strong>r” <strong>the</strong> code <strong>of</strong><br />

application and rule engine, which would not<br />

o<strong>the</strong>rwise be compatible. The good example <strong>of</strong> glue<br />

code would be initialization and configuration <strong>of</strong> rule<br />

engine which are both performed by framework<br />

automatically and <strong>the</strong>refore not needed to be written<br />

by programmer in application code. Instead, various<br />

places in core functionality where some processing<br />

with domain knowledge is required contain simple<br />

method calls to Jessine, which engage <strong>the</strong> rule engine<br />

and appropriate action is executed or value returned.<br />

Moreover, in context <strong>of</strong> Aspect-Oriented S<strong>of</strong>tware<br />

Development technologies, rule systems developed<br />

with Jessine can be considered as implementation <strong>of</strong><br />

<strong>the</strong> cross-cutting functionality concerning domain<br />

knowledge. Since multiple joinpoint models are


available through various AOP implementations,<br />

method calls to Jessine framework can be<br />

encapsulated into <strong>the</strong> aspects and woven into <strong>the</strong><br />

application code by employing some <strong>of</strong> <strong>the</strong>m. In this<br />

manner, by combining Jessine framework with<br />

AOSD technologies, programmers can integrate <strong>the</strong><br />

separated domain knowledge with core functionality<br />

without interfering with application source code.<br />

III. HOW JESSINE WORKS<br />

One <strong>of</strong> <strong>the</strong> main objectives when designing <strong>the</strong><br />

framework was automatization <strong>of</strong> tasks necessary to<br />

perform when using rule systems and leaving <strong>the</strong><br />

programmer only to deal with application related<br />

coding. Ano<strong>the</strong>r one was holding to <strong>the</strong> principle <strong>of</strong><br />

loose coupling [3] between particular components <strong>of</strong><br />

framework.<br />

The Jessine framework is provided with an<br />

Application Programing Interface which exposes<br />

both <strong>of</strong> <strong>the</strong> components that are <strong>the</strong> subjects to be<br />

customized by programmer for specific domain<br />

applications – AbstractEngine and Action.<br />

AbstractEngine is component wrapped around Jess<br />

rule engine and extends it with automatic<br />

initialization and configuration in accordance with<br />

external configuration files defined outside an<br />

application. Ano<strong>the</strong>r component, Action, serves as a<br />

communication bridge between <strong>the</strong> rule system and<br />

<strong>the</strong> application context by enabling domain rules to<br />

access <strong>the</strong> objects from application. All <strong>of</strong> <strong>the</strong> o<strong>the</strong>r<br />

components <strong>of</strong> framework are hidden and may not be<br />

modified by programmer.<br />

There exist no explicit dependencies between<br />

components which make up rule systems in Jessine.<br />

Programmer can develop each component<br />

independently <strong>of</strong> <strong>the</strong> o<strong>the</strong>rs and reuse it among many<br />

rule systems. Bindings between components are<br />

defined declaratively in Jessine configuration file<br />

with Jessine XML-based language.<br />

The framework allows to implement multiple rule<br />

systems in one application, each <strong>of</strong> <strong>the</strong>m specialized<br />

in different area <strong>of</strong> domain knowledge. Every<br />

implementation can be packed into <strong>the</strong> library file<br />

and used toge<strong>the</strong>r with Jessine library in o<strong>the</strong>r<br />

application as a plugin solution.<br />

It is also provided with its own exception and<br />

logging system. When an error occurs, an object <strong>of</strong><br />

Jessine exception is created, handed <strong>of</strong>f to <strong>the</strong><br />

runtime system and <strong>the</strong> event is recorded to <strong>the</strong> log<br />

file. Besides exceptions, logging system records<br />

every important event within context <strong>of</strong> rule systems,<br />

e.g. starting <strong>the</strong> environment <strong>of</strong> framework or setting<br />

<strong>of</strong> configuration directory.<br />

50<br />

IV. RELATED WORKS<br />

When considering literature on developing s<strong>of</strong>tware<br />

applications with rule-based knowledge we find that<br />

all advocate making rules explicit and separating<br />

<strong>the</strong>m from <strong>the</strong> object-oriented core functionality<br />

[4][5][6]. Moreover, many technologies exist that are<br />

targeted towards <strong>the</strong>se goals, although <strong>the</strong>y take<br />

radically different approaches. First <strong>of</strong> all, rule-based<br />

knowledge can be represented separately in <strong>the</strong><br />

object-oriented programming language itself. An<br />

extension to this approach is representing rule-based<br />

knowledge explicitly using object-oriented design<br />

patterns, referred to as <strong>the</strong> Rule Object Pattern [7].<br />

Finally <strong>the</strong>re are dozens <strong>of</strong> both commercial and<br />

academic hybrid systems [8][2], which integrate a<br />

full-fledged rule-based language with a state <strong>of</strong> <strong>the</strong><br />

art object-oriented programming language.<br />

Never<strong>the</strong>less, none <strong>of</strong> <strong>the</strong>m is using rule system for<br />

separating domain knowledge in <strong>the</strong> way I am doing<br />

it in my approach.<br />

ACKNOWLEDGMENTS<br />

This work was supported by VEGA Grant No.<br />

1/4073/07 – Aspect-oriented Evolution <strong>of</strong> Complex<br />

S<strong>of</strong>tware System.<br />

REFERENCES<br />

[1] M. Sabo. Applying Rules in S<strong>of</strong>tware Systems.<br />

Diploma <strong>the</strong>sis, Technical university in Kosice,<br />

Kosice, <strong>2008</strong>.<br />

[2] E. Friedmann-Hill. Jess in Action: Rule-Based<br />

Systems in Java. Manning Publications, 2003.<br />

[3] J. Eder, G. Kappel, M. Schrefl. Coupling and<br />

Cohesion in Object-Oriented Systems. In<br />

<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 1 st <strong>Conference</strong> on Information<br />

and Knowledge Management, 1992.<br />

[4] C.J. Date. What Not How: The Business Rules<br />

Approach to Application Development. Addison-<br />

Wesley, 2000.<br />

[5] R.G. Ross. Principles <strong>of</strong> <strong>the</strong> Business Rule<br />

Approach. Addison-Wesley, 2003.<br />

[6] B. von Halle. Business Rules Applied: Building<br />

Better Systems Using <strong>the</strong> Business Rules<br />

Approach. Wiley, 2002.<br />

[7] A. Arsanjani. Rule Object 2001: A Pattern<br />

Language for Adaptive and Scalable Business<br />

Rule Construction. IBM T.J. Watson Research<br />

Centre, 2001.<br />

[8] M. D’Hondt. Hybrid Aspects for Integrating<br />

Rule-Based Knowledge and Object-Oriented<br />

Functionality. PhD <strong>the</strong>sis, Vrije Universiteit<br />

Brussel, 2004.


�������������������������������������<br />

�<br />

������������������������������������������������������<br />

��������������������������������������������������<br />

�������������������������������<br />

�����������������<br />

�������{Lubomir.Wassermann,Marek.Paralic}@tuke.sk<br />

�<br />

�<br />

��������������������������������������������������<br />

�������������������������������������������������<br />

����������������������������������������������������<br />

������������ ��������� ��������� ��� �������� ���������<br />

���������������������������������������������������<br />

������������������������������������������������<br />

���������������������������������������������������<br />

������������������������������������������������������<br />

�������������� ��� ��������� ���������� ������� ����<br />

������������������������������������������������<br />

�����������������������������������<br />

��� �������������<br />

�������������������������������� ���������� �����<br />

�������� ������� ��� ������ ��� ������� �������� ������<br />

������������������������������������������������������<br />

���� ����� �������� ������ ��������� ����������� ���<br />

������ �����������������������������������������������<br />

����������������������������������������������������<br />

����� ������� ��� ������������������ ���������� ������<br />

����������������������������������������������������<br />

������ �������� ���� ���� ������ ��� ������ ����������� ���<br />

��������� ������ �������� ��� ������� ��� ���������� ���<br />

����������������������������������������������������<br />

���� ����� ����������� ���� ������� ��� ���� ������� ���� ���<br />

����������������������������������������������������������<br />

��� ���� ����� ������� ��������� ������ ���� ��� ����� ���<br />

������� ��� �������� �������� ������� ���� ��� ������ ����<br />

���������� ����� ����� ���� ��� ��������� �������� �����<br />

��������� ��� ���� ������� ��������� ������� ������� ������<br />

������� ��� ����������� ������� ��������� ���� ����� �����<br />

�������������������������������������������������������<br />

��� ��������� ���� ������� ��� ���� �������� ��� ������ ����<br />

�������� ����� ������������� ��� ��������� ���� ��� ������<br />

�������������������������������������������������������<br />

����������������������������������������������������<br />

������ �������� ����� ���� ��� ������� ��������� ������<br />

������� ����� �������� ����� ��������� ���� ������ �����<br />

������������� �������� ���������� ��� ������� ������ ���<br />

�������������������������������������������������<br />

51<br />

���� �������������������������<br />

�����������������������������������������������������<br />

��������������������������������������������������������<br />

������������������<br />

−� ���������������������������������������������<br />

���������������������������������������������������<br />

−� ����������������������������������������������<br />

������������������������������������������������<br />

��������������������������������������������<br />

−� ����������������������������������������������<br />

�������������������������������������<br />

��������������������������������������������������<br />

��������� ��� �������� ���� ������� ��������������<br />

���������� ��������� ������ ���� ���������� ��� �����<br />

���������� �������� ��� ��������� ��� ���������� ��� �����<br />

����������������������������������������������������<br />

����������������������������������������������������<br />

���������������������������������<br />

����������������<br />

������� ����� ��� ����������� ������������ ���������<br />

����������������������������������������������������<br />

��������������������������������<br />

−� ���������������������������������������������<br />

�������������������������������������������������<br />

−� �����������������������������������������������<br />

������������������������������������������<br />

−� ���������������������������������������������<br />

�����������������������������������������<br />

������������������������<br />

�������������������������������������������������<br />

������������������������������� ������� ��� ������ ���<br />

������������� ������� ����� ��� ������� ����� � ����� �����<br />

���������������������������������������������������<br />

����� ��� ������� ��������� ��� ������� ������� ������<br />

�������� ��� ������� ��� ������ ��� �������� ���� ����<br />

������ ��� ������ ����� ������ �������� ��� ������������<br />

������ ������������� ������� �������� ���� ������ ���� ���<br />

������ ��� �������� ������ �������� ���� ����� ����� ����� ���<br />

��������������������������������������������������


������� ��� �������������� ��� ����� ������� ���<br />

�����������������������<br />

�<br />

���������������������������������������<br />

���� �������������������������<br />

������ ������� ��� ������������� ��� ����� ���������<br />

���������������������������<br />

−� ����������������������������������������������<br />

�������������������������������������������<br />

−� �������������������������������������������<br />

�������������������������������������������������<br />

��� �������<br />

����� ���������� ���� ��������� ���������� ������� ���� ���<br />

�����������������������������������������������������<br />

�����������������������������������������������������<br />

����� �������� ������ ���� ������������� ����� ������������<br />

������������������������������������������������������<br />

���� ������� ������������ ������ ��� ����������� ���� ���<br />

��������������������������������������������������������<br />

������������������������������������������������������<br />

��������� ���� ��� ��������� ��� ���������� ������ ����<br />

��������������������������������������������������������<br />

�����������<br />

������� ��� �������������� ����� ����� ���������<br />

����� �������� ����� ������� ��� ���� ������������ ���<br />

�������� ��������� ���� ������ ����� ������� ��� ��������<br />

��������������������������������������������������<br />

�����������������������������������������������������<br />

����������������������������������<br />

−� �����������������������������������������������<br />

−� �����������������������������������������������<br />

������������������������<br />

−� ��������������������������������������������<br />

���������������������<br />

�����������������������������������������������������<br />

����������������������������������������������������<br />

����������������������������������<br />

������ ��������� �������� ��� �������� ���� ����������<br />

����� ���� ��������� �������������� ���� ����� ������<br />

52<br />

���������������������������������������������������<br />

��������������������������������<br />

��� �����������������<br />

���������������� ������ ����������� �������� �����<br />

�����������������������������������������������������<br />

�������������������������������������������������������<br />

��������������������������������������������������������<br />

���� ������ ����� ����� ���� ������ ������� �����������<br />

��������������������������<br />

���������� ��� ��������� �������� ��� ����������� ���<br />

��������� ���������� ��� ������ ������ ��� ���������� �����<br />

�������������� ����� ����� ��� ���������� ��� ������ ���<br />

����������� ��� ������� ������ �������� ������� �����������<br />

������������������������������������������������������<br />

�������������� ������ ���� ��������� ����� ����<br />

�������� �������������� ������� ������������ ���� ������<br />

������ ��� ������� �� ������ ����� ���� ��������� ��� �����<br />

����������������<br />

����������������������������������������������������<br />

������������������������������������������������������<br />

��������������������������������������������������������<br />

�����������������������������������������������������<br />

�������������������������������������������������������<br />

����������������������������������������������������<br />

�����<br />

������ ����� �������� ��������� �������� ��� ���������<br />

����������������������������������������������������<br />

������ ��������� ���� ������� �������� ������� ���������� ���<br />

����������� ��� ������������� ������ ���������� ��� ����<br />

������� ���� ������� ��� ���������� ��� ���������� �����<br />

�����������������������������<br />

����������������<br />

����� ����� ���� ���������� ��� ����� ������ ����<br />

���������� �� ���������������� ���������� ��� ��������<br />

�����������������<br />

�����������<br />

���� ��������� ��� ������� ������� ���� ������� ���<br />

������������� ��� ������ ������������ ���������<br />

���������������������������������������������<br />

���� ���������� �������� �������� ���������<br />

�������������� ��� ������� ��������� ������ ��������<br />

�������� �������� ���������� ����������� ��� ��������<br />

�������� ��� ����������� ������������ ����<br />

� �������������������<br />

������������������������������������������������<br />

������ ���������� �������� �������� ��������<br />

���������� ����������� ��� �������� �������� ���<br />

����������������������������������������������<br />


Design <strong>of</strong> Self-organizing Systems Using Evolutionary Methods<br />

István Fehérvári and Wilfried Elmenreich (Faculty Mentor)<br />

Institute for Networked and Embedded Systems / Mobile Systems Group<br />

University <strong>of</strong> Klagenfurt<br />

Klagenfurt, Austria<br />

Email: {ifeherva,wilfried.elmenreich}@uni-klu.ac.at<br />

Abstract — Having many advantages, self-organizing systems<br />

could mean a solution for intelligent group behavior. Until now<br />

<strong>the</strong>re is no general methodolgy how to design and control <strong>the</strong>se<br />

systems. In this paper we examine a genetic approach for generating<br />

group behavior. Related research proved that it is possible<br />

to create sensor based robot control program using genetically<br />

evolved ANNs. Extending this concept, we propose a method for<br />

designing self-organizing systems by extending this concept to cooperative<br />

behavior among <strong>the</strong> individuals.<br />

I. INTRODUCTION<br />

The power <strong>of</strong> evolutionary methods was shown in many<br />

applications where <strong>the</strong> implementation <strong>of</strong> artificial intelligence<br />

was necessary. As an example genetic algorithms<br />

(and so genetic programming) were used in simple and<br />

more complex games as proposed by Hauptman and Sipper<br />

[1]. Ano<strong>the</strong>r approach is using neural networks to build<br />

programs capable <strong>of</strong> learning. Following this idea adaptive<br />

robot control can be developed at <strong>the</strong> hardware level<br />

for practical motion [2] or taking one step fur<strong>the</strong>r by evolving<br />

neural networks in a genetic manner to get a complete<br />

control program <strong>of</strong> a single robot driven by sensor data. According<br />

to [3] <strong>the</strong> evolved network was significant better<br />

than <strong>the</strong> handcrafted solution.<br />

For self-organizing systems to get <strong>the</strong> highest satisfaction<br />

σsys for <strong>the</strong> system a cooperative behavior should be developed.<br />

This is mostly done by manually adjusting parameters<br />

which in case <strong>of</strong> systems with high complexity could<br />

be very difficult. In <strong>the</strong>se systems, a small change in a parameter<br />

could arise unpredicted behavior due to <strong>the</strong> missing<br />

direct function between element and system. Thus, a lot <strong>of</strong><br />

trials are necessary to find a satisfying solution. Evolutionary<br />

methods provide a means to automatize <strong>the</strong> testing and<br />

optimizing <strong>of</strong> parameters in an intelligent way.<br />

In this paper we present a method where <strong>the</strong> team strategy<br />

is not pre-programmed, but using <strong>the</strong> idea proposed<br />

by Elmenreich and Klinger [3], is generated by genetically<br />

evolving an Artifical Neural Network (ANN). The intention<br />

is to minimize <strong>the</strong> required human interference in creating<br />

<strong>the</strong> model by making it partly or fully generated by itself.<br />

The rest <strong>of</strong> <strong>the</strong> paper is structurized as follows: Section 2<br />

describes self-organizing systems while Section 3 explains<br />

<strong>the</strong> genetic approach and discusses <strong>the</strong> method for <strong>the</strong> given<br />

problem. Current work is discussed in Section 4.<br />

II. SELF-ORGANIZING SYSTEMS<br />

The term “self-organizing” was first introduced in 1947 by<br />

W. Ross Ashby. It refers to pattern-formation processes<br />

within a system caused by <strong>the</strong> behavior <strong>of</strong> its individual entities.<br />

Their basic feature is <strong>the</strong> way how <strong>the</strong>y acquire <strong>the</strong>ir<br />

order and structure. Among <strong>the</strong> several definitions <strong>the</strong> most<br />

suitable could be <strong>the</strong> following:<br />

53<br />

“In self-organizing systems, pattern formation occurs<br />

through interactions internal to <strong>the</strong> system, without intervention<br />

by external directing influences.”[4, p.7]<br />

A good example is a team <strong>of</strong> workers acting on <strong>the</strong>ir own<br />

following a mutual understanding. If <strong>the</strong>re would be common<br />

blueprints or a boss controlling <strong>the</strong>m, it would mean an<br />

external influence resulting in no self-organization. Nature<br />

also has many examples like a school <strong>of</strong> fish where each<br />

individual has knowledge only about its neighbors and so<br />

<strong>the</strong>re is no leader among <strong>the</strong>m. The key part is <strong>the</strong> communication<br />

between <strong>the</strong> individuals satisfying <strong>the</strong>ir own goals<br />

and by doing so a pattern emerges. Through <strong>the</strong>se features<br />

it is expected that <strong>the</strong> overall system shows many advantages<br />

like robustness, adaptability and scalability which<br />

makes self-organizing systems (SOS) an interesting option<br />

for controlling complex systems. However <strong>the</strong> above mentioned<br />

properties and <strong>the</strong> decentralized control makes an<br />

SOS difficult to design and control. Although, Gershenson<br />

[5] provides some ideas for <strong>the</strong> design <strong>of</strong> SOS for technical<br />

applications, <strong>the</strong>re is no general methodology yet to<br />

explain how <strong>the</strong>se should be done.<br />

III. THE GENETIC APPROACH<br />

The main idea behind genetic methods is <strong>the</strong> Darwinian selection<br />

process. The algorithm operates on a pool <strong>of</strong> possible<br />

solutions. It evaluates <strong>the</strong>m based on <strong>the</strong>ir fitness function<br />

which measures <strong>the</strong>ir performance when applied to <strong>the</strong><br />

given problem. While <strong>the</strong> best ones are kept, those with<br />

bad performance will be replaced by <strong>of</strong>fsprings or mutations<br />

<strong>of</strong> <strong>the</strong> pool. For more complex problems where <strong>the</strong><br />

solutions cannot be represented by variables but individual<br />

programs it is hard to define how to mutate or recombine<br />

<strong>the</strong>m while keeping <strong>the</strong> syntactically correct structure. In<br />

<strong>the</strong>se cases normal programming languages like C or Java<br />

do not qualify for this kind <strong>of</strong> programming, however <strong>the</strong><br />

usage <strong>of</strong> ANN could solve <strong>the</strong> problem.<br />

A. NEURAL NETWORK MODEL<br />

We choose a fully connected, time-discrete, recurrent ANN<br />

where each neuron has a connection to every o<strong>the</strong>r neuron<br />

and also to itself via several input connectors. Based on <strong>the</strong><br />

neurons activation value, <strong>the</strong> neurons output is forwarded<br />

via different connections to all o<strong>the</strong>r neurons. Each connection<br />

is assigned a weight and each neuron is assigned<br />

a bias value. There are special input and output neurons<br />

for <strong>the</strong> controlling interface: <strong>the</strong> input neurons produce <strong>the</strong><br />

sensor data on <strong>the</strong>ir output independtly <strong>of</strong> <strong>the</strong>ir input values,<br />

<strong>the</strong> output neurons act like normal neurons in <strong>the</strong> network,<br />

but <strong>the</strong>ir output data is also forwarded to <strong>the</strong> actuators.<br />

O<strong>the</strong>r nodes which are not characterized as input or<br />

output nodes are <strong>the</strong> so-called hidden neurons. Figure 1<br />

shows a schematic structure.<br />

In more complex systems containing many individuals<br />

like self-organizing systems <strong>the</strong>re might be a need for dif-


Figure 1: Neural network model example<br />

ferently arranged sensor subsystems connected to <strong>the</strong> same<br />

network. Robot soccer would be a good example where one<br />

group <strong>of</strong> sensors watch for <strong>the</strong> ball while o<strong>the</strong>r groups for<br />

teammates or opposing partymembers. Their separate inputs<br />

will be connected to <strong>the</strong> common network at <strong>the</strong> input<br />

nodes so <strong>the</strong> output would depend on several external factors<br />

resulting in more compound behavior.<br />

B. EVOLUTION METHOD<br />

Evolution method means a genetic algorithm that looks for<br />

solutions with high fitness values to <strong>the</strong> given problem.<br />

During <strong>the</strong> process many solutions are created and evaluated<br />

in parallel while <strong>the</strong> best ones will be kept for <strong>the</strong> next<br />

generation. First, <strong>the</strong> genomes have to be defined for each<br />

entity. In our case <strong>the</strong>y are represented by <strong>the</strong> weight matrix<br />

and <strong>the</strong> biases <strong>of</strong> each neuron. The next step is <strong>the</strong> definition<br />

<strong>of</strong> <strong>the</strong> fitness function which is a proper formulation <strong>of</strong><br />

what we expect from <strong>the</strong> system. As proved in [3] genetic<br />

selection processes supporting crossover, mutation and elite<br />

selection have higher learning speed than <strong>the</strong> ones without<br />

crossover [6]. To have support <strong>of</strong> all this features we will<br />

use <strong>the</strong> versatile framework given by Pfandler [7].<br />

IV. TESTBED<br />

There are several ways to test <strong>the</strong> proposed method; we<br />

chose <strong>the</strong> RoboCup Soccer Simulator1 as an evaluation tool.<br />

This open-source program is capable <strong>of</strong> simulating a soccer<br />

game between two teams played in real-time. See Figure 2<br />

for a screenshot <strong>of</strong> <strong>the</strong> simulator. As a result <strong>of</strong> each genetic<br />

procedure a population <strong>of</strong> ANNs will be created. These will<br />

be turned into robot soccer team control programs, where<br />

we want to rank <strong>the</strong>m for fur<strong>the</strong>r processes based on <strong>the</strong>ir<br />

in-game performance. In <strong>the</strong> soccer environment <strong>the</strong>re is<br />

no effective way to assign fitness values to <strong>the</strong> entities by<br />

evaluating <strong>the</strong>m one by one, so we will use a Swiss system<br />

tournament [8] organized among <strong>the</strong>m. A standard simulated<br />

robot soccer game takes 12 minutes to finish if played<br />

in real-time. Calculating with a population <strong>of</strong> 100 it would<br />

take 84 minutes to finish with only one iteration. To overcome<br />

this problem and shorten <strong>the</strong> required time for <strong>the</strong><br />

test <strong>the</strong> simulator will be modified to run as fast as possible<br />

using an asynchronic coupling between server and team<br />

clients. Fur<strong>the</strong>r speedup can be achieved using a parallel<br />

evaluation on a multi-computer system.<br />

V. CONCLUSION AND OUTLOOK<br />

The success <strong>of</strong> genetically evolved ANNs used in [3] for<br />

autonomous mobile robot control shows a good potential<br />

1 http://sserver.sourceforge.net<br />

54<br />

Figure 2: Team evaluation using soccer simulator<br />

for applying to self-organizing systems. The goal <strong>of</strong> <strong>the</strong><br />

current project is to create a simulation environment for<br />

self-organizing systems evolved by following <strong>the</strong> proposed<br />

method. Once implemented, experiments will be performed<br />

to optimize and to compare this solution with hand coded<br />

ones in robot soccer environments.<br />

ACKNOWLEDGMENTS<br />

This work was supported by <strong>the</strong> Austrian FWF project<br />

TTCAR under contract No. P18060-N04.<br />

REFERENCES<br />

[1] M. Sipper, Y. Azaria, A. Hauptman, and Y. Shichel. Designing<br />

an evolutionary strategizing machine for game<br />

playing and beyond. IEEE Trans. on Systems, Man and<br />

Cybernetics, Part C, 37(4):583–593, July 2007.<br />

[2] M. W. Han and P. Kopacek. Neural networks for <strong>the</strong><br />

control <strong>of</strong> soccer robots. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 2000<br />

IEEE International Symposium on Industrial Electronics,<br />

pages 571–575, Cholula, Mexico, August 2000.<br />

[3] W. Elmenreich and G. Klingler. Genetic evolution <strong>of</strong><br />

a neural network for <strong>the</strong> autonomous control <strong>of</strong> a fourwheeled<br />

robot. In Sixth Mexican International <strong>Conference</strong><br />

on Artificial Intelligence (MICAI’07), Aguascalientes,<br />

Mexico, November 2007.<br />

[4] S. Camazine, J. Deneubourg, N. R. Franks, J. Sneyd,<br />

G. Theraulaz, and E. Bonabeau. Self-Organization in<br />

Biological Systems, volume 1. Princeton University<br />

Press, Princeton, NJ, USA, 2nd edition, January 2001.<br />

[5] C. Gershenson. Design and control <strong>of</strong> self-organizing<br />

systems. PhD Dissertation, Vrije Universiteit Brussel,<br />

Brussel, Belgium, 2007.<br />

[6] L. A. Meeden. An incremental approach to developing<br />

intelligent neural network controllers for robots. IEEE<br />

Trans. on Systems, Man, and Cybernetics, Part B: Cybernetics,<br />

26(3):474–485, June 1996.<br />

[7] A. Pfandler. Design and implementation <strong>of</strong> a generic<br />

framework for genetic optimization <strong>of</strong> neural networks.<br />

Bachelor’s <strong>the</strong>sis, Vienna University <strong>of</strong> Technology, Vienna,<br />

Austria, 2007.<br />

[8] J. A. Bergstra and L. M. G. Feijs, editors. Algebraic<br />

Methods, volume 2. Springer-Verlag, New York, LLC,<br />

1st edition, May 1991.


Towards Automatic Design <strong>of</strong> Competitive Image Filters in FPGAs<br />

Zdeněk Vaˇsíček, Lukáˇs Sekanina (Faculty Mentor)<br />

Faculty <strong>of</strong> Information Technology<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: {vasicek,sekanina}@fit.vutbr.cz<br />

Abstract — This paper presents a new approach to <strong>the</strong> FPGA<br />

implementation <strong>of</strong> image filters which are utilized to remove <strong>the</strong><br />

salt and pepper noise <strong>of</strong> high intensity. An evolutionary algorithm<br />

(EA) is utilized to find a set <strong>of</strong> image filters which can<br />

be employed in a bank <strong>of</strong> image filters. The main advantage <strong>of</strong><br />

this approach is that <strong>the</strong> bank <strong>of</strong> evolutionary designed filters<br />

requires four times less resources on a chip in comparison with<br />

<strong>the</strong> adaptive median filter while <strong>the</strong> visual quality <strong>of</strong> filtering is<br />

preserved.<br />

In order to design image filters in reasonable time, an FPGAbased<br />

evolutionary platform is utilized. The proposed platform<br />

is based on <strong>the</strong> implementation <strong>of</strong> a search algorithm in <strong>the</strong><br />

PowerPC processor which is available in Xilinx Virtex II Pro<br />

and newer FPGAs. As <strong>the</strong> search algorithm as well as <strong>the</strong> evaluation<br />

<strong>of</strong> candidate solution runs in FPGA, <strong>the</strong> evolutionary design<br />

<strong>of</strong> image filters needs approx. 44 times less time to design a<br />

human-competitive filter when compared to <strong>the</strong> same algorithm<br />

running on <strong>the</strong> common PC.<br />

I. INTRODUCTION<br />

The aim <strong>of</strong> this paper is to introduce a new EA-based<br />

approach to <strong>the</strong> image filter design. We will solely deal<br />

with <strong>the</strong> images corrupted by <strong>the</strong> shot noise, in particular<br />

salt-and-pepper noise. This type <strong>of</strong> noise is caused by<br />

malfunctioning pixels in camera sensors, faulty memory<br />

locations in hardware, or errors in <strong>the</strong> data transmission.<br />

Traditionally, <strong>the</strong> salt-and-pepper noise is removed by<br />

median filter (MF). However, when <strong>the</strong> intensity <strong>of</strong> noise<br />

is increasing (15-90% pixels are corrupted), simple median<br />

filter are not sufficient. Among o<strong>the</strong>rs, adaptive median<br />

filter (AMF) provides good results and simultaneously,<br />

its hardware implementation is straightforward [1].<br />

The problem is that <strong>the</strong>se advanced techniques require<br />

a considerably larger area on a chip in comparison with<br />

simple median filters.<br />

It has been shown that evolutionary design techniques<br />

are able to generate slightly better solutions than standard<br />

MF for this noise intensity [2].Unfortunately, <strong>the</strong> evolutionary<br />

design approach which works up to 10% noise<br />

intensity does not work for higher noise intensities.<br />

In this paper, we will show that by an innovative combination<br />

<strong>of</strong> evolved designs and conventional designs we<br />

are able to significantly reduce <strong>the</strong> overall implementation<br />

cost on a chip in comparison to standard approaches<br />

based on sophisticated filtering schemes, such as AMF.<br />

55<br />

II. IMAGE FILTER GENERATOR<br />

In order to design image filters which can be used in <strong>the</strong><br />

bank <strong>of</strong> filters, we utilize <strong>the</strong> evolutionary platform running<br />

on FPGA introduced in [3]. The system consists<br />

<strong>of</strong> a genetic unit (GU), array <strong>of</strong> reconfigurable elements<br />

(VRC) and fitness calculation unit (FU). The corrupted<br />

image is processed by a pipelined array <strong>of</strong> reconfigurable<br />

elements whose configuration is generated by <strong>the</strong> genetic<br />

unit. While candidate filters are created and evaluated<br />

using a user logic available on <strong>the</strong> FPGA, <strong>the</strong> genetic operations<br />

are carried out in <strong>the</strong> PowerPC processor which<br />

is available on <strong>the</strong> same FPGA. Every image operator is<br />

considered as a digital circuit <strong>of</strong> nine primary 8-bit inputs<br />

and a single 8-bit output, which processes gray-scaled<br />

images. The aim <strong>of</strong> <strong>the</strong> EA is to find a digital circuit<br />

which accomplish <strong>the</strong> requirement task.<br />

The VRC array consists <strong>of</strong> 2-input configurable elements<br />

(CEs), placed in a grid 8 columns and 4 rows. Any<br />

input <strong>of</strong> each CEs may be connected ei<strong>the</strong>r to a primary<br />

input or to <strong>the</strong> output <strong>of</strong> a CEs placed anywhere in <strong>the</strong><br />

preceding column. Each CEs can be programmed to implement<br />

one <strong>of</strong> 16 functions given beforehand. The evaluation<br />

<strong>of</strong> candidate configurations is pipelined in such<br />

manner as <strong>the</strong>re are no idle clock cycles. As <strong>the</strong> EA directly<br />

operates with configurations <strong>of</strong> <strong>the</strong> VRC, a configuration<br />

is considered as a chromosome.<br />

The corrupted image, original image and filtered image<br />

are stored in external SRAM memories. The FU forwards<br />

<strong>the</strong> pixels <strong>of</strong> corrupted image to inputs <strong>of</strong> VRC<br />

and compares <strong>the</strong> pixels <strong>of</strong> filtered image with <strong>the</strong> pixels<br />

<strong>of</strong> original image. The design objective is to minimize<br />

<strong>the</strong> absolute difference between <strong>the</strong> filtered image and <strong>the</strong><br />

original image.<br />

III. PROPOSED APPROACH<br />

In order to create a salt-and-pepper noise filter which<br />

generates filtered images <strong>of</strong> <strong>the</strong> same quality as an AMF<br />

and which is suitable for hardware implementation, we<br />

propose to combine several simple image filters utilizing<br />

<strong>the</strong> 3 × 3 window that are designed by an evolutionary<br />

algorithm. As Figure 2 shows <strong>the</strong> procedure has three<br />

steps: (1) <strong>the</strong> reduction <strong>of</strong> a dynamic range <strong>of</strong> noise, (2)<br />

processing using a bank <strong>of</strong> n filters and (3) deterministic<br />

selection <strong>of</strong> <strong>the</strong> best result.<br />

The first step reduces <strong>the</strong> large dynamic range <strong>of</strong> corrupted<br />

pixels (0/255) using a component which inverts<br />

all pixels with value 255, i.e. all shots are transformed to


PSNR [dB]<br />

45<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

MF (W =3× 3), 268 slices<br />

MF (W =5× 5), 1506 slices<br />

MF (W =7× 7), 4426 slices<br />

AMF (W max =5× 5), 2024 slices<br />

AMF (W =7× 7), 6567 slices<br />

max<br />

AMF (W =9× 9), 16395 slices<br />

max<br />

single filter, 200 slices<br />

3-bank filter, 843 slices<br />

best SW algorithm<br />

5<br />

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80<br />

Noise intensity [%]<br />

Figure 1: Comparison <strong>of</strong> various image filters using a set <strong>of</strong> 25 test images corrupted by salt-and-pepper noise<br />

Filter kernel 3x3<br />

I0<br />

I1<br />

I2<br />

I3<br />

I5<br />

I4<br />

I8<br />

I7<br />

I6<br />

Pre−<br />

processing<br />

filter<br />

Filter<br />

1<br />

Filter<br />

2<br />

Filter<br />

n<br />

Buffer<br />

O0<br />

O1<br />

On<br />

I4<br />

Post−<br />

processing<br />

filter<br />

Filtered image<br />

Figure 2: Proposed architecture for salt-and-pepper noise<br />

removal circuit<br />

have a uniform value. This task is easy to implement in<br />

hardware using comparators.<br />

Preprocessed image <strong>the</strong>n enters a bank <strong>of</strong> n filters that<br />

operate in parallel. The selection <strong>of</strong> suitable filters is <strong>the</strong><br />

crucial step. Since we repeated <strong>the</strong> evolutionary design <strong>of</strong><br />

salt-and-noise filters many times, we have ga<strong>the</strong>red various<br />

implementations <strong>of</strong> this type <strong>of</strong> filter. Thus, we selected<br />

n different evolved filters which exhibit <strong>the</strong> best<br />

filtering quality and utilized <strong>the</strong>m in <strong>the</strong> bank. These filters<br />

were designed by EA using <strong>the</strong> same type <strong>of</strong> noise<br />

and training image and with <strong>the</strong> same aim: to remove<br />

40% salt-and-pepper noise.<br />

Finally, <strong>the</strong> outputs coming from banks 1 ...ntoge<strong>the</strong>r<br />

with <strong>the</strong> original value <strong>of</strong> <strong>the</strong> central pixel are combined<br />

by n +1-input median filter which can easily be implemented<br />

using comparators. As <strong>the</strong> proposed system naturally<br />

forms a pipeline, <strong>the</strong> overall design can operate at<br />

<strong>the</strong> same frequency as a simple median filter when implemented<br />

in hardware.<br />

IV. EXPERIMENTAL RESULTS<br />

Filtering properties <strong>of</strong> <strong>the</strong> proposed bank <strong>of</strong> filters<br />

and MFs as well as AMFs (both with various filtering<br />

windows) were compared on 25 test images (see<br />

http://www.fit.vutbr.cz/ ˜vasicek/imagedb). All images<br />

were corrupted by salt-and-pepper noise <strong>of</strong> intensity 5-<br />

75%. The results were also compared to <strong>the</strong> best known<br />

s<strong>of</strong>tware solution which utilizes filtering windows <strong>of</strong> unlimited<br />

size. Figure 1 contains <strong>the</strong> results obtained as an<br />

average over test set. The visual quality is measured in<br />

terms <strong>of</strong> <strong>the</strong> peek signal-to-noise ratio (PSNR); <strong>the</strong> higher<br />

56<br />

PSNR, <strong>the</strong> better results.<br />

Surprisingly, only three filters utilized in <strong>the</strong> bank are<br />

needed to obtain a filter which produces images <strong>of</strong> practically<br />

undistinguishable visual quality to <strong>the</strong> AMF utilizing<br />

up to 7 × 7-pixel filtering windows.<br />

The implementation cost <strong>of</strong> adaptive median filter is<br />

2024 slices for AMF which utilizes <strong>the</strong> filter window<br />

up to 5 × 5 pixel and 6567 slices for AMF 7 × 7 (Virtex<br />

II Pro considered). The implementation cost <strong>of</strong> proposed<br />

bank consisting <strong>of</strong> three filters (3-bank) requires<br />

500 slices only and 5-bank filter 843 slices only (for details<br />

see [4]).<br />

V. CONCLUSION<br />

The proposed solution which combines image filters designed<br />

by means <strong>of</strong> EA with a simple human designed<br />

preprocessing and post-processing unit can reach <strong>the</strong><br />

quality <strong>of</strong> adaptive median filtering. The bank <strong>of</strong> three<br />

evolved filters which utilizes 3×3-pixel filtering window<br />

is able to accomplish <strong>the</strong> same task as AMF utilizing up<br />

to 7×7-pixel filtering windows; however, using a significantly<br />

lower number <strong>of</strong> FPGA resources. This can lead to<br />

a significant reduction <strong>of</strong> power consumption <strong>of</strong> a target<br />

system.<br />

REFERENCES<br />

[1] H. Hwang and R. A. Haddad. New algorithms for adaptive<br />

median filters. In K.-H. Tzou and T. Koga, editors, Proc.<br />

SPIE Vol. 1606, p. 400-407, Visual Communications and<br />

Image Processing ’91: Image Processing, Kou-Hu Tzou;<br />

Toshio Koga; Eds., pages 400–407, November 1991.<br />

[2] L. Sekanina. Evolvable components: From Theory to<br />

Hardware Implementations. Natural Computing. Springer-<br />

Verlag Berlin, 2004.<br />

[3] Zdenek Vasicek and Lukas Sekanina. An evolvable hardware<br />

system in xilinx virtex ii pro fpga. International Journal<br />

<strong>of</strong> Innovative Computing and Applications, 1(1):63–73,<br />

2007.<br />

[4] Zdenek Vasicek and Lukas Sekanina. An area-efficient alternative<br />

to adaptive median filtering in fpgas. In Proc. <strong>of</strong><br />

<strong>the</strong> 17th Conf. on Field Programmable Logic and Applications,<br />

pages 1–6. IEEE Computer Society, to appear, 2007.


Evaluation <strong>of</strong> a State-based Process Management Platform<br />

Ekaterina Andrianova, Jürgen Dorn<br />

Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Emails: Katerina.Andrianova@gmail.com<br />

juergen.dorn@ec.tuwien.ac.at<br />

Abstract — The first objective <strong>of</strong> <strong>the</strong> analysis presented<br />

in this paper is to define an evaluation procedure<br />

for Business Process Management Systems<br />

(BPMS) and to investigate requirements to BPMS in<br />

different categories according to <strong>the</strong> procedure. The<br />

business process life cycle, different BPM values,<br />

process types, infrastructure requirements and<br />

BPM failure reasons are explored.<br />

The second objective <strong>of</strong> <strong>the</strong> analysis is to investigate<br />

ISIS Papyrus’ BPM approach grounded on a<br />

state-based representation <strong>of</strong> processes and a<br />

learning algorithm sensitive to document content.<br />

We have assessed <strong>the</strong> ISIS Papyrus BPM platform<br />

step by step according to <strong>the</strong> developed procedure.<br />

I. INTRODUCTION<br />

A lot <strong>of</strong> efforts are usually devoted to BPM projects.<br />

Never<strong>the</strong>less, such projects fail quite <strong>of</strong>ten or don’t<br />

live up to <strong>the</strong> expectations due to a variety <strong>of</strong> reasons.<br />

Packaged applications such as ERP also have some<br />

problems for managing business processes. They<br />

don’t deliver <strong>the</strong> process insights and agility needed<br />

by <strong>the</strong> business due to [2].<br />

Potential factors that can lead to a BPM project<br />

failure are important to be investigated. Moreover, a<br />

special procedure should be created in order to help a<br />

BPMS-customer to understand which aspects should<br />

be taken into account when choosing a BPMS and to<br />

help a BPMS-vendor to define a niche for <strong>the</strong> BPMS<br />

and understand its advantages and drawbacks.<br />

At <strong>the</strong> same time many alternative approaches to<br />

BPM have been proposed. One <strong>of</strong> <strong>the</strong>m is ISIS Papyrus’<br />

process management approach grounded on <strong>the</strong><br />

state/event-based representation <strong>of</strong> processes and<br />

learning, sensitive to <strong>the</strong> content <strong>of</strong> documents in a<br />

workflow [4]. The Papyrus platform provides <strong>the</strong><br />

user-trained environment (UTE) that lets execute or<br />

guide a process on <strong>the</strong> basis <strong>of</strong> learned user experience<br />

[4]. Since this approach is quite new for <strong>the</strong><br />

BPMS market, it is essential to assess it and evaluate<br />

strengths and weaknesses and to show which BPM<br />

sector it can potentially occupy.<br />

57<br />

II. RESEARCH METHODOLOGY<br />

The evaluation procedure was created after researching<br />

<strong>the</strong> scientific base <strong>of</strong> BPM, i.e., <strong>the</strong> process life<br />

cycle, classifications <strong>of</strong> processes, and <strong>the</strong> practical<br />

base, i.e., how BPM is understood by companies,<br />

why companies initiate BPM, why BPM projects fail<br />

or don’t live up to <strong>the</strong> expectations. This knowledge<br />

provides <strong>the</strong> foundation for realizing <strong>the</strong> steps that<br />

are important for evaluation <strong>of</strong> BPMS. Then, each<br />

step <strong>of</strong> <strong>the</strong> evaluation procedure was investigated in<br />

order to define requirements to BPMS. Classification<br />

<strong>of</strong> processes was analysed in order to define specific<br />

requirements for each process type. The state-<strong>of</strong>-<strong>the</strong>art<br />

about <strong>the</strong> scientific background was acquainted<br />

by literature study, web-articles and white papers in<br />

<strong>the</strong> BPM field (for instance, [1]). Survey reports such<br />

as [2] and interviews with companies in Austria held<br />

by <strong>the</strong> authors <strong>of</strong> this paper provided <strong>the</strong> up-to-date<br />

view on problems considered in literature.<br />

The Papyrus platform was evaluated according to<br />

<strong>the</strong> developed procedure. Sources such as Papyrus’<br />

documentation [4], workshops and discussions with<br />

<strong>the</strong> Papyrus team as well as partial scenario implementation<br />

were used for evaluation.<br />

III. RESEARCH RESULTS<br />

A. EVALUATION PROCEDURE FOR BPMS<br />

The developed evaluation procedure includes <strong>the</strong><br />

following steps:<br />

1. Process life cycle support<br />

2. BPM values support<br />

3. Infrastructure requirements fulfilment<br />

4. Fulfilment <strong>of</strong> <strong>the</strong> requirements to different<br />

process types<br />

5. Best practices support and failure avoidance<br />

Each step is provided ei<strong>the</strong>r with a chosen structure<br />

for evaluation (<strong>the</strong> first step) or a developed evaluation<br />

schema with requirements (o<strong>the</strong>r steps).<br />

The process classification for <strong>the</strong> step 4 is based<br />

on [2] and [3].


B. ASSESSMENT RESUTLS FOR THE PAPYRUS<br />

PLATFORM<br />

Papyrus’ approach is evaluated as very valuable for<br />

process life cycle support because it brings user experience<br />

into <strong>the</strong> system, is not restricted by modelling<br />

elements and lets optimize a process almost<br />

without human participation. Never<strong>the</strong>less, graphic<br />

representation <strong>of</strong> processes as well as optimization<br />

results are not visible explicitly. This, however, is<br />

essential for a company especially while optimization<br />

is not empowered with a collection <strong>of</strong> process targets.<br />

The Papyrus platform is assessed as oriented on<br />

bringing such values as cross-enterprise process<br />

support, agility <strong>of</strong> process changes and decreasing<br />

process cycle times by automating a part <strong>of</strong> <strong>the</strong> process<br />

and providing operational decision support as<br />

well as by supporting tasks delegation and removing<br />

delays when handed <strong>of</strong>f from person to person. See<br />

“Figure 1”for details.<br />

Figure 1: BPM values support by Papyrus<br />

According to <strong>the</strong> evaluation, <strong>the</strong> platform is flexible<br />

enough, reliable and secure, and fulfils mandatory<br />

infrastructure requirements.<br />

The system suits for managing people- and document-intensive<br />

processes as well as for sequential<br />

workflow and content approval scenarios. It has good<br />

chances to be successful during competitions with<br />

o<strong>the</strong>r BPMS for managing <strong>the</strong>se process types.<br />

Decision-intensive processes can be basically supported<br />

by Papyrus. Never<strong>the</strong>less, <strong>the</strong> platform is not<br />

empowered to support such kind <strong>of</strong> processes in <strong>the</strong><br />

best way as it doesn’t provide decision support on <strong>the</strong><br />

basis <strong>of</strong> business data analysis.<br />

Papyrus’ approach, having a state-based concept,<br />

is potentially very powerful for case management <strong>of</strong><br />

complex processes. This could be Papyrus’ unique<br />

niche as typical workflow modeling BPMS must<br />

58<br />

have failed to support such kind <strong>of</strong> processes. But no<br />

tries have been yet undertaken to explore how system<br />

would operate practically under such conditions.<br />

See support <strong>of</strong> different process types by Papyrus<br />

on “Figure 2”.<br />

Figure 2: Different process types support by Papyrus<br />

The Papyrus platform with <strong>the</strong> UTE concept is especially<br />

efficient for:<br />

� Processes with high reusability <strong>of</strong> actions<br />

� Process flow based on documents content<br />

� Large enterprises with several employers having<br />

<strong>the</strong> same functional role<br />

� Processes where staff changes <strong>of</strong>ten or employers<br />

haven’t enough experience<br />

Since UTE is basically an approach based on inductive<br />

learning, <strong>the</strong> drawback <strong>of</strong> <strong>the</strong> approach is that it<br />

only works if enough examples exist for learning.<br />

Not sufficient process visibility and complexity <strong>of</strong><br />

<strong>the</strong> design and deployment procedure make <strong>the</strong> platform<br />

not a perfect system. Never<strong>the</strong>less, this platform<br />

has a potential to avoid some failures and supports<br />

multiple BPM best practices.<br />

The analysis described in this paper was carried<br />

out in <strong>the</strong> master <strong>the</strong>sis and presented at Papyrus.<br />

REFERENCES<br />

[1] H. Smith, P. Finger, “Business Process Management:<br />

The Third Wave”, 2003.<br />

[2] Forrester company, www.forrester.com, <strong>the</strong> Forrester<br />

Wave reports concerning BPM.<br />

[3] Macehiter Ward-Dutton company, N. Ward-<br />

Dutton, report “Business process management:<br />

beyond technology”, December 2007.<br />

[4] ISIS Papyrus company, www.isis-papyrus.com,<br />

technical and marketing documentation.


Motor task-induced modifications <strong>of</strong> multiple spinal cord reflexes<br />

Ursula S. H<strong>of</strong>stoetter, Josef Ladenbauer, Karen Minassian,<br />

Milan R. Dimitrijevic and Frank Rattay<br />

Institute <strong>of</strong> Analysis and Scientific Computing<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

ursula.h<strong>of</strong>stoetter@gmail.com<br />

Abstract — Dynamic regulation <strong>of</strong> reflexes controlled<br />

by <strong>the</strong> central nervous system plays an integral<br />

part in neurocontrol <strong>of</strong> locomotion. Such modifications<br />

<strong>of</strong> sensory-motor transmission can be<br />

studied by conditioning a test reflex with specific<br />

motor tasks. In <strong>the</strong> present study, we applied a<br />

novel transcutaneous spinal cord stimulation (tSCS)<br />

technique to elicit short-latency test reflexes in<br />

several lower limb muscles simultaneously. During<br />

3 different conditioning-test paradigms in healthy<br />

subjects we could demonstrate that <strong>the</strong> monosynaptic<br />

test reflexes can be characteristically modulated<br />

during <strong>the</strong> execution <strong>of</strong> postural maneuvers.<br />

I. INTRODUCTION<br />

Since successful limb movement depends on effective<br />

interaction <strong>of</strong> sensory flow with motor plans, <strong>the</strong><br />

central nervous system control <strong>of</strong> somatosensory<br />

transmission is an essential element for neural control<br />

<strong>of</strong> locomotion. For its assessment, conditioning-test<br />

paradigms utilizing <strong>the</strong> monosynaptic H reflex are<br />

commonly used [1]. However, elicitation <strong>of</strong> H reflexes<br />

from periphery is methodologically difficult to<br />

be obtained under constant stimulation conditions,<br />

especially during movement at <strong>the</strong> stimulation site.<br />

Recently we described a non-invasive method <strong>of</strong><br />

tSCS effective to simultaneously elicit reflex responses<br />

in multiple lower limb muscles [2]. According<br />

to <strong>the</strong> initiation (posterior roots) and recording<br />

sites (surface <strong>of</strong> a muscle) <strong>the</strong>se responses were<br />

named posterior root-muscle reflexes (PRM reflexes).<br />

PRM reflexes elicited by single stimuli were<br />

shown to be segmental monosynaptic reflexes [3].<br />

We will demonstrate that PRM reflexes can be<br />

elicited in upright standing healthy subjects and used<br />

to assess <strong>the</strong> central gain control <strong>of</strong> short-latency<br />

reflexes at multiple segmental levels simultaneously.<br />

II. MATERIAL AND METHODS<br />

− Subjects:<br />

Three men with intact nervous system were recruited.<br />

The stimulation and recording protocols used<br />

59<br />

in this manuscript were approved by <strong>the</strong> local ethics<br />

committee.<br />

− Stimulation and recording set-up:<br />

tSCS was applied through surface stimulation electrodes<br />

placed over T11-T12 vertebral processes, with<br />

<strong>the</strong> indifferent electrodes over <strong>the</strong> abdomen. A constant-voltage<br />

stimulator was used to generate symmetric,<br />

biphasic rectangular pulses with widths <strong>of</strong> 2<br />

ms. For details see [2].<br />

Electromyographic (EMG) activity was recorded<br />

with surface electrodes placed over bilateral quadriceps<br />

(Q), hamstrings (Ham), tibialis anterior (TA),<br />

and triceps surae (TS).<br />

− Study protocol:<br />

PRM reflexes (> 100 μV) were elicited at 0.2 Hz<br />

in all muscles with subjects in a standing position.<br />

Recordings started with 5 unconditioned PRM reflexes,<br />

followed by 5 responses modified by one <strong>of</strong><br />

three different motor tasks: volitional unilateral dorsiand<br />

plantar flexion; volitional unilateral hip and knee<br />

flexion-extension; leaning backward/forward.<br />

− Data analysis:<br />

The amplitudes <strong>of</strong> <strong>the</strong> 5 unconditioned and <strong>the</strong> 5<br />

conditioned PRM reflexes were averaged. The mean<br />

conditioned responses were <strong>the</strong>n normalized with<br />

respect to <strong>the</strong> unconditioned ones.<br />

III. RESULTS<br />

PRM reflexes could be elicited in all studied muscles<br />

while subjects were standing upright.<br />

As shown in Figure 1, ankle dorsiflexion distinctly<br />

suppressed <strong>the</strong> ipsilateral Ham and TS responses and<br />

facilitated <strong>the</strong> ipsilateral Q response. PRM reflexes<br />

showed modifications during ankle plantar flexion,<br />

without an overall tendency towards ei<strong>the</strong>r facilitation<br />

or suppression.<br />

Figure 2 illustrates that Ham and TS responses <strong>of</strong><br />

<strong>the</strong> moving lower limb were always distinctly suppressed<br />

during both flexion and extension phases <strong>of</strong> a<br />

volitional unilateral multi-joint movement. Attenuation<br />

<strong>of</strong> PRM reflexes in Ham and TS was more pr<strong>of</strong>ound<br />

during flexion than extension phases.


Figure 1: (A) dorsi- and plantar flexion; (B) hip<br />

and knee flexion-extension.<br />

Leaning backward resulted in reduced Ham and<br />

TS response magnitudes and facilitation <strong>of</strong> Q (Figure<br />

3, A). Leaning forward led to <strong>the</strong> facilitation <strong>of</strong> Ham,<br />

TA, and TS (Figure 3, B).<br />

Figure 3: leaning (A) forward; (B) backward.<br />

60<br />

IV. DISCUSSION<br />

PRM reflexes were generally modified in a way to<br />

functionally meet <strong>the</strong> requirements <strong>of</strong> a particular<br />

motor act: For example, dorsiflexion suppressed <strong>the</strong><br />

ankle and hip extensor responses and at <strong>the</strong> same<br />

time facilitated <strong>the</strong> relevant flexors Q and TA.<br />

Volitional multi-joint movements at <strong>the</strong> hip and<br />

knee joints reduced <strong>the</strong> response magnitudes <strong>of</strong> Ham<br />

and TS during flexion and extension with different<br />

degrees <strong>of</strong> attenuation. Hence, mechanisms underlying<br />

<strong>the</strong> reflex modulation do not solely hinge on<br />

muscle stretch and joint position but also on <strong>the</strong><br />

motor task setting <strong>the</strong> role <strong>of</strong> a particular muscle.<br />

Leaning backward requires contraction <strong>of</strong> Q and<br />

TA to maintain equilibrium, as accordingly revealed<br />

by our tests. PRM reflexes <strong>of</strong> Ham and TS, two muscles<br />

that are quiescent during this task, were suppressed,<br />

hinting on ‘active’ inhibition.<br />

Leaning forward is normally accompanied by activation<br />

<strong>of</strong> Ham and TS to avoid falling, as detected in<br />

our recordings. Additionally, we found facilitation <strong>of</strong><br />

TA, probably due to <strong>the</strong> necessity to increase <strong>the</strong><br />

stiffness <strong>of</strong> <strong>the</strong> ankle joint for stabilization.<br />

The presented approach opens a new avenue to<br />

study how motor control is triggered and maintained<br />

in individuals with intact or altered nervous system<br />

functions. Apart from <strong>the</strong> basic sciences value <strong>of</strong><br />

studying human motor control, <strong>the</strong> assessment <strong>of</strong><br />

postural capacities and volitional movements will be<br />

crucial to design rehabilitation strategies in people<br />

with different neurological disorders.<br />

ACKNOWLEDGMENTS<br />

This study was supported by <strong>the</strong> Wings for Life<br />

Spinal Cord Research Foundation, Proj.Nr.<br />

WFL-FR-001/06, and <strong>the</strong> Austrian Science<br />

Fund (FWF), Vienna, Austria, Proj. L512-N13.<br />

REFERENCES<br />

[1] Schieppati M. The H<strong>of</strong>fmann reflex: a means <strong>of</strong><br />

assessing spinal reflex excitability and its descending<br />

control in man. Prog Neurobiol 1987;<br />

28: 345-376.<br />

[2] Minassian K, Persy I, Rattay F, Pinter MM, Kern<br />

H, Dimitrijevic MR. Human lumbar cord circuitries<br />

can be activated by extrinsic tonic input to<br />

generate locomotor-like activity. Hum Mov Sci<br />

2007; 26: 275-295.<br />

[3] Minassian K, Persy I, Rattay F, Dimitrijevic MR,<br />

H<strong>of</strong>er C, Kern H. Posterior root-muscle reflexes<br />

elicited by transcutaneous stimulation <strong>of</strong> <strong>the</strong> human<br />

lumbosacral cord. Muscle Nerve 2007; 35:<br />

327-336.


Approach for a Reliable Cooperative Relaying Process<br />

Melanie Schranz and Wilfried Elmenreich (Faculty Mentor)<br />

Lakeside Labs/Institute for Networked and Embedded Systems<br />

University <strong>of</strong> Klagenfurt<br />

Klagenfurt, Austria<br />

Email: {melanie.schranz,wilfried.elmenreich}@uni-klu.ac.at<br />

Abstract — Cooperative relaying is a method, where nodes<br />

overhearing a communication help in forwarding this information<br />

in case <strong>the</strong> direct transmission between source and destination<br />

fails. In this paper we present a communication protocol<br />

based on cooperative relaying employing a TDMA relay selection<br />

and exchange <strong>of</strong> short control messages. To show its applicability,<br />

we implemented <strong>the</strong> described protocol on <strong>the</strong> wireless<br />

platform WARP as our testbed.<br />

I. INTRODUCTION<br />

Nowadays wireless networks form a powerful area that<br />

cannot be neglected. Primarily <strong>the</strong> task <strong>of</strong> <strong>the</strong>se wireless<br />

networks is to deliver data, whe<strong>the</strong>r it is about music,<br />

videos, sensor data or any o<strong>the</strong>r information. A major<br />

question is, if <strong>the</strong> data packets reliably reach its destination<br />

in a wireless network. With applications such as<br />

voice communication or distributed sensing and control,<br />

reliable communication plays a decisive role.<br />

Wireless links typically suffer from effects such as<br />

shadowing and fading, where <strong>the</strong> latter typically can<br />

cause a wireless link to be unavailable for several 100<br />

milliseconds. Within this timespan, retransmission attempts<br />

issued by <strong>the</strong> source will very likely have no success.<br />

A transmission via a spatially different path, however,<br />

is typically not affected by <strong>the</strong> same fading problems<br />

at <strong>the</strong> same time. Thus <strong>the</strong> concept <strong>of</strong> cooperative<br />

relaying, where a relay node is assisting <strong>the</strong> communication<br />

between a source and a destination node [1], can<br />

greatly improve <strong>the</strong> reliability <strong>of</strong> wireless communication.<br />

Within <strong>the</strong> scope <strong>of</strong> this paper, we consider an approach<br />

where <strong>the</strong> relay node only forwards information<br />

in <strong>the</strong> case <strong>the</strong> direct transmission between source<br />

and destination fails (in contrast to relaying networks in<br />

many telecommunication applications, where <strong>the</strong> relays<br />

are used to extend <strong>the</strong> transmission range via a multi-hop<br />

approach).<br />

In this paper we present a cooperative relaying approach<br />

for improving a wireless multi-hop communication.<br />

Besides elaborating <strong>the</strong> protocol <strong>the</strong> intention is to<br />

evaluate it on a wireless platform called WARP, which<br />

is a universal multiple-input multiple-output wireless test<br />

platform controlled by FPGA for maximum flexibility.<br />

The rest <strong>of</strong> <strong>the</strong> paper is structured as follows: Section 2<br />

introduces to cooperative relaying concepts in more detail.<br />

Section 3 describes a self conceived simple relaying<br />

approach for 802.11-based networks while Section 4 delineates<br />

<strong>the</strong> hardware testbed we are using to implement<br />

61<br />

<strong>the</strong> concepts depicted in Section 3. In <strong>the</strong> end we give a<br />

conclusion and an outlook to future work.<br />

II. COOPERATIVE RELAYING<br />

The term ”cooperative relaying” was already explained in<br />

<strong>the</strong> introduction. The major advantage <strong>of</strong> using a relayed<br />

communication channel is that it raises <strong>the</strong> capacity <strong>of</strong><br />

<strong>the</strong> network by implementing spatial diversity [1]. This<br />

spatial diversity establishes several independent communication<br />

paths. Researches on Multiple Input Multiple<br />

Output (MIMO) systems have depicted that <strong>the</strong> reliability<br />

<strong>of</strong> sending a data packet is enhanced <strong>the</strong> more independent<br />

paths are available [2].<br />

Practically <strong>the</strong> simplest cooperative relaying network<br />

consists <strong>of</strong> three nodes, like shown in Figure 1. The<br />

source (S) has a direct communication (step 1) to destination<br />

(D). If <strong>the</strong> communication fails (steps 2 and 3)<br />

<strong>the</strong> relay (R) has to support D with information.<br />

Figure 1: Cooperative relaying scheme<br />

III. RELAYING APPROACH<br />

We propose <strong>the</strong> following possible solution to achieve an<br />

acceptable reliability in a wireless network: In case a<br />

packet is ready to send, <strong>the</strong> source S ”asks” <strong>the</strong> destination<br />

D via an RTS (ready to send) if a transmission is<br />

possible. When S receives a CTS (clear to send), as a<br />

next step it starts with <strong>the</strong> transmission. S gets an ACK<br />

(acknowledge) if all packets are received by D. In this<br />

(likely) case we have a successful direct communication<br />

like demonstrated in Figure 2.<br />

In <strong>the</strong> o<strong>the</strong>r case, so if a packet did not arrive, we have<br />

<strong>the</strong> situation illustrated in Figure 3. D sends a NACK(not<br />

acknowledge) to S and broadcasts a LOOKforGOODmessage.<br />

This message contains <strong>the</strong> header <strong>of</strong> <strong>the</strong> packet<br />

that could not be received. Attention should be paid to


Figure 2: Packet exchange scheme for successful direct<br />

transmission<br />

Figure 3: Packet exchange scheme for failed direct transmission<br />

<strong>the</strong> source S. In <strong>the</strong> case <strong>of</strong> a none received packet, also S<br />

works as a relay node as assistance for <strong>the</strong> data transmission.<br />

All <strong>the</strong> relay nodes (in our figure <strong>the</strong> situation<br />

is demonstrated with only one relay node R) look<br />

for <strong>the</strong> packet in <strong>the</strong>ir buffers. If <strong>the</strong> packet is in <strong>the</strong><br />

buffer <strong>the</strong> relays send <strong>the</strong>ir ID (IDSENT) in an a priori<br />

defined TDMA schedule, so each node has its own<br />

time slot. In <strong>the</strong> next step D selects one relay node according<br />

to measures <strong>of</strong> <strong>the</strong> channel quality by sending a<br />

SELECT-message that contains <strong>the</strong> ID <strong>of</strong> <strong>the</strong> chosen relay<br />

node. The data will <strong>the</strong>n be transmitted. In <strong>the</strong> case<br />

<strong>of</strong> an failed transmission again, <strong>the</strong> process would start<br />

once with choosing a relay node by sending a LOOKfor-<br />

GOOD message.<br />

The messages CTS, RTS, ACK, NACK, LOOKfor-<br />

GOOD, IDSENT and SELECT are coded with BPSK<br />

(Binary Phase-Shift Keying). These messages are considerably<br />

shorter than <strong>the</strong> data packet. The BPSK modulation<br />

scheme is less vulnerable to noise, so we assume<br />

that <strong>the</strong> control messages can be transmitted even<br />

in case <strong>of</strong> fading problems. The data packet is coded with<br />

16QAM (16 Quadrature Amplitude Modulation), which<br />

is more efficient than BPSK, but requires a better Signalto-Noise<br />

ratio than BPSK.<br />

IV. TESTBED<br />

To evaluate <strong>the</strong> described relay approach, we use a wireless<br />

platform called WARP board1 . The WARP board<br />

consist <strong>of</strong> a Virtex II Pro FPGA <strong>of</strong> Xilinx for <strong>the</strong> processing<br />

<strong>of</strong> <strong>the</strong> developed communication protocol, two<br />

radio boards for <strong>the</strong> transmission over radio frequency <strong>of</strong><br />

1 For fur<strong>the</strong>r information: http://warp.rice.edu<br />

62<br />

2.4 GHz, two antennas and respective circuitry needed<br />

for realizing a communication. Figure 4 depicts a WARP<br />

board, which has been set up to use both transmitters emulating<br />

two different wireless nodes.<br />

Figure 4: WARP-Board: Wireless platform for implementing<br />

<strong>the</strong> relay approach<br />

With <strong>the</strong>se boards we have <strong>the</strong> possibility to modify<br />

physical layer as well as MAC layer. All programming<br />

work is done in C. With <strong>the</strong> comfortable detail <strong>of</strong> a serial<br />

output we have a chance to track <strong>the</strong> communication<br />

channel status with every sent packet for debugging<br />

purposes. Current test configuration consists <strong>of</strong> two<br />

WARP Boards, where each <strong>of</strong> <strong>the</strong> two wireless transmission<br />

units per board is programmed separately, forming a<br />

network <strong>of</strong> four logical nodes.<br />

V. CONCLUSION AND OUTLOOK<br />

The proposed protocol on <strong>the</strong> basis <strong>of</strong> cooperative relaying<br />

promises increased reliability for wireless networks<br />

by mitigating small-scale fading effects. A preliminary<br />

version <strong>of</strong> <strong>the</strong> protocol has been implemented<br />

on <strong>the</strong> WARP platform, but fur<strong>the</strong>r tests have to be done<br />

to achieve expedient results. Though, for an efficient<br />

evaluation we are going to evaluate <strong>the</strong> approach on a<br />

larger wireless network. Fur<strong>the</strong>r future work will also<br />

include researching alternative ways for relay selection,<br />

since <strong>the</strong> proposed TDMA approach comes with considerable<br />

complexity for large networks.<br />

ACKNOWLEDGMENTS<br />

This work was supported by <strong>the</strong> European Regional Development<br />

Fund and <strong>the</strong> Carinthian Economic Promotion<br />

Fund (KWF) under grant 20214/15935/23108 within <strong>the</strong><br />

Lakeside Labs project.<br />

REFERENCES<br />

[1] W. Elmenreich, N. Marchenko, H. Adam, C. H<strong>of</strong>bauer,<br />

G. Brandner, C. Bettstetter, and M. Huemer.<br />

Building blocks <strong>of</strong> cooperative relaying in wireless<br />

systems. Springer e&iJournal, October <strong>2008</strong>.<br />

[2] H. Adam, W. Elmenreich, and S. M. Senouci. A protocol<br />

for cooperative relaying in wireless networks.<br />

submitted to EURASIP Journal on Wireless Communications<br />

and Networking, <strong>2008</strong>.


Retrieval <strong>of</strong> Ocean Water Parameters from Satellite Imagery and<br />

<strong>the</strong>ir Dependence on Atmospheric Correction Models<br />

Rushane Dyer and Sam Ahmed (Faculty Mentor)<br />

Optical Remote Sensing Laboratory<br />

The City College <strong>of</strong> <strong>the</strong> City University <strong>of</strong> New York<br />

New York, USA<br />

Email: rdyer00@ccny.cuny.edu<br />

Abstract — Ocean color algorithms for <strong>the</strong><br />

retrieval <strong>of</strong> chlorophyll [Chl] and mineral<br />

concentrations as well as water inherent optical<br />

properties (IOPs) from satellite reflectance<br />

spectra usually fail in coastal waters because <strong>of</strong><br />

errors in atmospheric correction procedures.<br />

Recent advances in those procedures over coastal<br />

zones opened possibilities for more detailed<br />

studies <strong>of</strong> coastal areas using satellite imagery<br />

which includes analysis <strong>of</strong> spatial and temporal<br />

trends in [Chl]. We report on studies <strong>of</strong> MODIS<br />

data for various parameters for several years in<br />

Western, Central and Eastern areas <strong>of</strong> Long<br />

Island Sound using SeaDAS processing s<strong>of</strong>tware<br />

and <strong>the</strong> SWIR and NIR atmospheric correction<br />

algorithms. Correlations between parameters<br />

retrieved using various procedures were<br />

calculated and analyzed.<br />

I. INTRODUCTION<br />

It is well known that in coastal regions <strong>the</strong> near infrared<br />

(NIR) atmospheric correction algorithm can be<br />

unreliable in <strong>the</strong> retrieval <strong>of</strong> parameters such as chlorophyll<br />

concentrations [Chl], inherent optical properties<br />

(IOPs) and fluorescence line height (FLH). This<br />

is typically due to <strong>the</strong> increase in elastic scattering <strong>of</strong><br />

chlorophyll and minerals along <strong>the</strong> coast. In response<br />

to this problem <strong>the</strong> shortwave infrared<br />

(SWIR) atmospheric correction algorithm was introduced<br />

to be used in retrievals <strong>of</strong> ocean water parameters<br />

within coastal regions [1]. Generally, <strong>the</strong> SWIR<br />

algorithm has shown significant improvement over<br />

<strong>the</strong> NIR algorithm in remote sensing <strong>of</strong> coastal regions.<br />

Ra<strong>the</strong>r than just completely disregarding <strong>the</strong><br />

use <strong>of</strong> <strong>the</strong> NIR algorithm along <strong>the</strong> coast, we have<br />

begun a thorough analysis <strong>of</strong> various parameters<br />

retrieved using both <strong>the</strong> NIR and SWIR algorithms<br />

for <strong>the</strong> region <strong>of</strong> Long Island Sound (LIS). We have<br />

quantified correlations between parameters retrieved<br />

with each algorithm and have determined <strong>the</strong> level <strong>of</strong><br />

sensitivity <strong>the</strong>se parameters have with respect to each<br />

atmospheric model.<br />

63<br />

II. SENSITIVITY OF THE RETRIEVALS<br />

To determine sensitivity <strong>of</strong> [Chl], FLH, and various<br />

o<strong>the</strong>r parameters to <strong>the</strong> atmospheric correction model<br />

we studied MODIS Aqua data <strong>of</strong> [Chl], FLH, absorption<br />

<strong>of</strong> colored dissolved organic matter (CDOM)<br />

and backscattering coefficient for several years for<br />

Western, Central and Eastern areas <strong>of</strong> Long Island<br />

Sound using SeaDAS processing s<strong>of</strong>tware with standard<br />

NIR and recently developed SWIR atmospheric<br />

correction models [2]. The general idea is that high<br />

correlations between <strong>the</strong> retrieved parameters can be<br />

used as indicators <strong>of</strong> more accurate parameter retrieval.<br />

Figure 1 is a typical [Chl] map for <strong>the</strong> regions <strong>of</strong><br />

LIS and NY Bight <strong>the</strong>se were created for both NIR<br />

and SWIR models. Figures 2-3 show correlations<br />

between [Chl] and FLH for <strong>the</strong>se 2 models. Current<br />

MODIS algorithms for [Chl] use wavelengths within<br />

<strong>the</strong> blue and green parts <strong>of</strong> <strong>the</strong> spectra, <strong>the</strong>se areas are<br />

typically <strong>the</strong> most affected by atmospheric correction<br />

models. We can see that <strong>the</strong> correlation <strong>of</strong> retrieved<br />

[Chl] between <strong>the</strong>se 2 models is extremely low,<br />

which is what we would expect.<br />

Figure 1: [Chl] map for LIS


Figure 2: Correlation between [Chl]<br />

Figure 3: Correlation between FLH<br />

Correlation between FLH, on <strong>the</strong> o<strong>the</strong>r hand, is<br />

very good. The current FLH algorithm uses wavelengths<br />

in <strong>the</strong> near infrared region <strong>of</strong> <strong>the</strong> spectra, it<br />

also uses <strong>the</strong> difference between reflectances to calculate<br />

its values, which is less sensitive to atmospheric<br />

models than ratio algorithms such as that <strong>of</strong><br />

[Chl]. This analysis has been extended to parameters<br />

such as backscattering bb(551) and CDOM absorption<br />

at 412 nm ay(412)<br />

ACKNOWLEDGMENTS<br />

This work was supported by grants from NASA,<br />

NOAA and Office <strong>of</strong> Naval Research.<br />

REFERENCES<br />

[1] M. Wang, "Remote sensing <strong>of</strong> <strong>the</strong> ocean<br />

contributions from ultraviolet to nearinfrared<br />

using <strong>the</strong> shortwave infrared bands:<br />

simulations," Appl. Opt. 46, 1535-1547<br />

(2007)<br />

[2] S. Ahmed, A. Gilerson, J. Zhou, R. Dyer, S.<br />

Hlaing, I. Ioannou, B. Gross, F. Moshary,"<br />

Untangling <strong>the</strong> Make-up <strong>of</strong> <strong>the</strong> NIR Reflectance<br />

Peak in Coastal Waters and its Impact<br />

64<br />

on Remote Sensing Retrievals <strong>of</strong> [Chl] and<br />

Fluorescence Height Algorithms,” Proc. <strong>of</strong><br />

SPIE 7105, Remote Sensing <strong>of</strong> <strong>the</strong> Ocean,<br />

Sea Ice, and Large Water Regions <strong>2008</strong>.


Fast Packet Classification Algorithm in Hardware<br />

Abstract — Packet classification is an important operation<br />

for applications such as routers, firewalls or intrusion detection<br />

systems. Many algorithms and hardware architectures for<br />

packet classification have been created, but none <strong>of</strong> <strong>the</strong>m can<br />

compete with <strong>the</strong> speed <strong>of</strong> TCAMs in <strong>the</strong> worst case. I propose<br />

new hardware-based algorithm for packet classification. The<br />

solution is based on problem decomposition and is aimed at<br />

<strong>the</strong> highest network speeds. A unique property <strong>of</strong> <strong>the</strong> algorithm<br />

is <strong>the</strong> constant time complexity in terms <strong>of</strong> external memory accesses.<br />

The algorithm performs exactly two external memory<br />

accesses to classify a packet. Using FPGA and one commodity<br />

SRAM chip, a throughput <strong>of</strong> 150 million packets per second can<br />

be achieved.<br />

I. INTRODUCTION<br />

A classification algorithm contains a set <strong>of</strong> rules ordered<br />

by priority. Each rule defines a condition for all significant<br />

packet header fields. These fields are typically:<br />

Source IP Address, Destination IP Address, Source Port,<br />

Destination Port, Protocol. A condition may be exact<br />

match, prefix match (usually for IP addresses), range<br />

match (for ports), or a wildcard (matching any value).<br />

The goal <strong>of</strong> a packet classification algorithm is to find<br />

<strong>the</strong> matching rule with <strong>the</strong> highest priority. The output <strong>of</strong><br />

<strong>the</strong> algorithm is <strong>the</strong>n <strong>the</strong> number <strong>of</strong> <strong>the</strong> matched rule.<br />

The traditional method <strong>of</strong> classifying packets<br />

makes use <strong>of</strong> Ternary Content Addressable Memories<br />

(TCAMs). However, <strong>the</strong> TCAM is an expensive device<br />

with high power-consumption [1]. Therefore, algorithmic<br />

solutions without <strong>the</strong> use <strong>of</strong> TCAMs has become<br />

a research subject. I propose a new packet classification<br />

method which uses SRAM to store necessary data, and<br />

FPGA to implement <strong>the</strong> algorithm. I will show that my<br />

solution is fully competitive to TCAM.<br />

II. RELATED WORK<br />

As <strong>the</strong> packet classification problem is inherently hard<br />

from a <strong>the</strong>oretical standpoint [2], a large number <strong>of</strong> hardware<br />

and s<strong>of</strong>tware solutions [3, 4] have been proposed.<br />

From <strong>the</strong> wide choice <strong>of</strong> available algorithms, I discuss<br />

only those which are related to my work. All <strong>of</strong> <strong>the</strong>m<br />

belong to <strong>the</strong> family <strong>of</strong> decomposition-based methods. In<br />

decomposition methods, packet classification is divided<br />

into several steps. First step is <strong>the</strong> Longest Prefix Match<br />

(LPM) operation, which is performed independently in<br />

Viktor Puˇs andJanKoˇrenek (Faculty Mentor)<br />

Faculty <strong>of</strong> Information Technology<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: {ipus,korenek}@fit.vutbr.cz<br />

65<br />

each packet header field. Range conditions (such as port<br />

ranges) in <strong>the</strong> ruleset are converted to prefixes, so that<br />

LPM may be performed in all packet header fields. LPM<br />

operation is performed in IP packet routing, so it is well<br />

studied topic. Popular example <strong>of</strong> such algorithm is Tree<br />

Bitmap [5], but <strong>the</strong>re are also many o<strong>the</strong>r solutions [6].<br />

The next step is <strong>the</strong> search process, mapping LPM results<br />

to <strong>the</strong> correct rule number. This is <strong>the</strong> specific part<br />

<strong>of</strong> each decomposition-based method.<br />

III. ALGORITHM<br />

I propose a novel high-speed hardware-suited packet<br />

classification algorithm, which has a modular design and<br />

removes <strong>the</strong> performance drawbacks <strong>of</strong> existing solutions.<br />

The algorithm consists <strong>of</strong> LPM for rule fields followed<br />

by a mechanism to search a rule. LPM and <strong>the</strong><br />

rest <strong>of</strong> <strong>the</strong> classification algorithm have a very simple<br />

unidirectional interface and both parts may be changed<br />

separately, when a better solution is available. Recent<br />

research results for LPM operation have outstanding results<br />

even over 100 Gbps [6], <strong>the</strong>refore I do not propose<br />

any new architecture for LPM and focus on <strong>the</strong> rule<br />

searching mechanism.<br />

The primary goal is to find a solution with <strong>the</strong> constant<br />

packet rate and a good scalability with <strong>the</strong> size <strong>of</strong> <strong>the</strong> rule.<br />

Therefore, I propose to use a perfect hashing mechanism<br />

to provide <strong>the</strong> Rule Table search in a constant time and<br />

utilize <strong>the</strong> <strong>of</strong>f-chip memory to store Perfect Hash Table.<br />

The perfect hash function must find a correct rule number<br />

for every packet. Thanks to LPM used in <strong>the</strong> first step,<br />

<strong>the</strong> packet state space is reduced significantly. The hash<br />

table is stored in <strong>the</strong> <strong>of</strong>f-chip memory and its construction<br />

was described in [7].<br />

The last part <strong>of</strong> <strong>the</strong> algorithm is <strong>the</strong> comparison <strong>of</strong><br />

<strong>the</strong> rule to corresponding packet header fields. As high<br />

throughput memory is needed to read a rule, <strong>the</strong> Rule Tableisstoredinon-chipmemories.<br />

IV. RESULTS<br />

The proposed algorithm was implemented in FPGA and<br />

utilizes one external static memory. I have studied several<br />

rulesets to get information about typical properties<br />

<strong>of</strong> rules, and similarly to [4] have found that <strong>the</strong> number<br />

<strong>of</strong> unique prefixes in each field is usually quite small.<br />

Therefore, <strong>the</strong> data structures for LPMs may be easily


stored in small on-chip memories, ei<strong>the</strong>r BlockRAMs or<br />

distributed memories. Also <strong>the</strong> Rule Table itself is not<br />

greater than a few kilobytes, so <strong>the</strong> external memory is<br />

not needed for its storage. But <strong>the</strong> Perfect Hash Table<br />

may become large, <strong>the</strong>refore, I propose to use an external<br />

memory to store it.<br />

The proposed solution significantly reduces bottleneck<br />

caused by <strong>the</strong> speed <strong>of</strong> external memory. Only two 18-bit<br />

words need to be read for every packet, which is many<br />

times less than one whole rule (or even worse, several<br />

rules, [3]). Moreover, performance <strong>of</strong> <strong>the</strong> algorithm is<br />

not affected by <strong>the</strong> complexity <strong>of</strong> rules. O<strong>the</strong>r fields (i.e.<br />

MAC addresses, TCP flags, etc.) can be added to rules<br />

and <strong>the</strong> throughput remains <strong>the</strong> same, only on-chip Rule<br />

Table size increases linearly. It means that <strong>the</strong> solution<br />

scales well with <strong>the</strong> rule size.<br />

Supposing 300 MHz DDR memory with <strong>the</strong> burst<br />

length <strong>of</strong> two words, throughput <strong>of</strong> <strong>the</strong> solution is always<br />

constant – 150 million packets per second for this particular<br />

memory.<br />

A. MEMORY REQUIREMENTS<br />

I have performed perfect hash function search for several<br />

rulesets from university campus network (fw) toge<strong>the</strong>r<br />

with several syn<strong>the</strong>tic ruleset generated by ClassBench<br />

[8] (synth) to determine <strong>of</strong>f-chip memory requirements.<br />

Memory requirements are shown in Table 1.<br />

Ruleset Rules Memory<br />

fw1 32 740<br />

fw2 58 3 424<br />

fw3 171 116 356<br />

synth1 40 2 740<br />

synth2 49 5 035<br />

synth3 70 2 451<br />

synth4 100 3 827<br />

Table 1: Off-chip memory requirements (in Bytes) for<br />

several rulesets.<br />

The on-chip memory for my solution consists <strong>of</strong> two<br />

basic parts: LPM, which may be <strong>the</strong> same for all<br />

crossproduct-based methods, and Rule Table, which may<br />

be compressed with a simple prefix indexing scheme to<br />

less than a half <strong>of</strong> its original size.<br />

V. CONCLUSION<br />

I have proposed a novel algorithm for fast packet classification<br />

using perfect hash functions. The algorithm introduces<br />

perfect hash functions to packet classification. By<br />

creating custom hash function, I make sure that all packets<br />

are hashed to associated rule. Because only two ex-<br />

66<br />

ternal memory accesses are needed to classify a packet,<br />

150 million packets per second can be processed with<br />

commodity FPGA and SRAM. Therefore, up to 48 Gbps<br />

throughput can be achieved for <strong>the</strong> smallest (40 B) packets.<br />

Moreover, <strong>the</strong> throughput does not depend on ruleset<br />

complexity and is well scalable with number <strong>of</strong> external<br />

memories.<br />

According to my knowledge, <strong>the</strong> proposed algorithm<br />

is <strong>the</strong> first algorithm which requires reasonable amount<br />

<strong>of</strong> memory and has constant processing time even for<br />

complex ruleset. High throughput toge<strong>the</strong>r with constant<br />

processing time makes <strong>the</strong> proposed algorithm fully<br />

competitive to widely used TCAM solutions. As <strong>the</strong> proposed<br />

solution uses commodity SRAM, <strong>the</strong> price and<br />

power consumption is significantly lower than classification<br />

with TCAM memory.<br />

REFERENCES<br />

[1] IDT Generic Part: 75K72100.<br />

http://www.idt.com/?catID=58523&genID=75K72100<br />

, June <strong>2008</strong>.<br />

[2] F. Baboescu, S. Singh, and G. Varghese. Packet classification<br />

for core routers: Is <strong>the</strong>re an alternative to<br />

CAMs? In INFOCOM, 2003.<br />

[3] Sarang Dharmapurikar, Haoyu Song, Jonathan<br />

Turner, and John Lockwood. Fast packet classification<br />

using Bloom filters. In ANCS ’06: <strong>Proceedings</strong><br />

<strong>of</strong> <strong>the</strong> 2006 ACM/IEEE symposium on Architecture<br />

for networking and communications systems, pages<br />

61–70, New York, NY, USA, 2006. ACM.<br />

[4] D. Taylor and J. Turner. Scalable packet classification<br />

using distributed crossproducting <strong>of</strong> field labels.<br />

In IEEE INFOCOM 2005, 24th Annual Joint <strong>Conference</strong><br />

<strong>of</strong> <strong>the</strong> IEEE Computer and Communications<br />

Societies., pages 269–280, July 2005.<br />

[5] Will Ea<strong>the</strong>rton, George Varghese, and Zubin Dittia.<br />

Tree bitmap: hardware/s<strong>of</strong>tware IP lookups with incremental<br />

updates. SIGCOMM Computer Communication<br />

Review, 34(2):97–122, 2004.<br />

[6] H Lee, W Jiang, and V K Prasanna. Scalable<br />

High-Throughput SRAM-Based Architecture for IP<br />

Lookup Using FPGA. In FPL ’08. IEEE, <strong>2008</strong>.<br />

[7] Zbigniew J. Czech, George Havas, and Bohdan S.<br />

Majewski. An optimal algorithm for generating minimal<br />

perfect hash functions. Information Processing<br />

Letters, 43(5):257–264, 1992.<br />

[8] David E. Taylor and Jonathan S. Turner. Classbench:<br />

a packet classification benchmark. IEEE/ACM<br />

Trans. Netw., 15(3):499–511, 2007.


Radiation <strong>of</strong> Antenna Array in Near-field Area<br />

Ji�í Hermany, Zden�k Nová�ek (Faculty Mentor)<br />

Department <strong>of</strong> Radio Electronics<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech republic<br />

Email: xherma09@stud.feec.vutbr.cz, novacek@feec.vutbr.cz<br />

Abstract — The goal <strong>of</strong> this work is to analyze <strong>the</strong><br />

radiation <strong>of</strong> antenna array computation methods<br />

and <strong>the</strong>n to produce <strong>the</strong> radiation <strong>of</strong> antenna array<br />

computation program. The program can display <strong>the</strong><br />

electrical and magnetical intensity and power density<br />

distribution over <strong>the</strong> planar or cylindrical surface<br />

in <strong>the</strong> near-field area. The aim <strong>of</strong> this paper is<br />

to propose some steps which were needed for building<br />

this program up.<br />

I. INTRODUCTION<br />

There is a description <strong>of</strong> a computation method <strong>of</strong><br />

antenna array radiation in this paper. The values <strong>of</strong><br />

an electrical and magnetical intensity or power density<br />

distribution <strong>of</strong> an electric dipole can be evaluated.<br />

II. THEORY<br />

The antenna array analyzed in this paper consists <strong>of</strong> a<br />

number <strong>of</strong> elementary dipoles placed in a row. A part<br />

<strong>of</strong> <strong>the</strong> antenna array is also a reflector which represents<br />

<strong>the</strong> conductive board situated in parallel to <strong>the</strong><br />

row <strong>of</strong> dipoles.<br />

A. ELECTROMAGNETIC FIELD INTENSITY<br />

COMPUTATION<br />

The electromagnetic field radiated in <strong>the</strong> elementary<br />

dipole near-field area can be computed using <strong>the</strong><br />

vector equations (1), (2) and (3) [1]. Its products are<br />

<strong>the</strong> vectors <strong>of</strong> <strong>the</strong> electrical and magnetical intensity<br />

and <strong>the</strong> power density in any chosen point in a space<br />

positioned in <strong>the</strong> near-field area.<br />

When <strong>the</strong> dipole and <strong>the</strong> measuring point positions<br />

are known it is possible to evaluate <strong>the</strong> electrical and<br />

magnetical intensity using equations<br />

� � �<br />

� 3<br />

��<br />

� � � �<br />

k r0<br />

� ( r0<br />

� d s0<br />

) 3(<br />

r0<br />

� d s0<br />

) � r0<br />

� d s0<br />

E � � j � �<br />

2 2<br />

4�� (1)<br />

��<br />

k r<br />

�<br />

k r<br />

� � � �<br />

3(<br />

r0<br />

� d s0<br />

) � r0<br />

� d s0<br />

�<br />

�<br />

� j�t<br />

� j<br />

e ,<br />

3 � 3 �<br />

k r ��<br />

� j�<br />

� 1 jk � �<br />

� j�t<br />

�<br />

H � � e � � �d<br />

0 r0<br />

,<br />

2<br />

s �<br />

(2)<br />

4�<br />

� r r �<br />

� �<br />

67<br />

where k = 2� /� is a wave number, ds is a position <strong>of</strong><br />

a dipole centre, ds0 is a dipole’s orientation and r0 is a<br />

vector from <strong>the</strong> dipole D to <strong>the</strong> point P (Figure 1).<br />

Figure 1: Point P position with regard to <strong>the</strong> position<br />

and orientation <strong>of</strong> dipole D<br />

From <strong>the</strong> values <strong>of</strong> <strong>the</strong> electrical and magnetical<br />

intensity vectors can be computed a power density<br />

vector<br />

*<br />

E H .<br />

(3)<br />

� � �<br />

� � �<br />

When <strong>the</strong> radiation values <strong>of</strong> all dipoles are<br />

known, it is needed to sum up <strong>the</strong> radiation elements<br />

<strong>of</strong> all parts <strong>of</strong> <strong>the</strong> antenna array. The reflector effect<br />

can be substituted by <strong>the</strong> radiation <strong>of</strong> a dipole situated<br />

behind <strong>the</strong> reflector which is driven by <strong>the</strong> same<br />

valued but inverse phased current.<br />

Then <strong>the</strong> radiation elements <strong>of</strong> this so-called mirrored<br />

dipoles are added to radiation elements <strong>of</strong> <strong>the</strong><br />

dipoles in front <strong>of</strong> <strong>the</strong> reflector.<br />

B. SURFACES<br />

The next step is to define a suitable group <strong>of</strong> points<br />

where <strong>the</strong> electromagnetic field values will be computed.<br />

The basic idea is that <strong>the</strong> dipole row with <strong>the</strong><br />

reflector (<strong>the</strong> antenna array) is situated in <strong>the</strong> centre<br />

<strong>of</strong> <strong>the</strong> Cartesian coordinate system when <strong>the</strong> reflector<br />

lies even with <strong>the</strong> z axis <strong>of</strong> <strong>the</strong> coordinate system.<br />

The first group <strong>of</strong> <strong>the</strong> suitable points in space is <strong>the</strong><br />

planar surface, which is defined by its position and<br />

orientation with regard to <strong>the</strong> antenna array position<br />

and orientation. The second one is <strong>the</strong> cylindrical<br />

surface where <strong>the</strong> cylinder axis lies even with <strong>the</strong> z<br />

axis <strong>of</strong> <strong>the</strong> Cartesian coordinate system.<br />

Then <strong>the</strong> values <strong>of</strong> <strong>the</strong> electromagnetic field are<br />

computed for every point <strong>of</strong> <strong>the</strong> planar or <strong>the</strong> cylindrical<br />

surface.


Figure 2: Cylindrical surface with <strong>the</strong> four dipole<br />

antenna array in <strong>the</strong> Cartesian coordinate system<br />

In <strong>the</strong> Figure 2 is shown <strong>the</strong> position <strong>of</strong> <strong>the</strong> cylindrical<br />

surface, where <strong>the</strong> electromagnetic field intensity<br />

distribution is computed, regarding to dipole<br />

antenna array with <strong>the</strong> reflector. Similarly <strong>the</strong>re can<br />

be <strong>the</strong> planar surface.<br />

C. PROGRAM<br />

The program for <strong>the</strong> computation <strong>of</strong> <strong>the</strong> electromagnetic<br />

field distribution was built in MATLAB with<br />

GUI (graphic user interface). It allows not only computing<br />

but also displaying <strong>the</strong> distribution <strong>of</strong> <strong>the</strong><br />

electrical and magnetic intensity and <strong>the</strong> power density<br />

components <strong>of</strong> antenna array radiation on <strong>the</strong><br />

planar or <strong>the</strong> cylindrical surface.<br />

The components can be x, y, z (for Cartesian coordinate<br />

system), r, �, z (for cylindrical coordinate<br />

system) or r, �, � (for spherical coordinate system).<br />

The program also allows saving computed values to<br />

files, saving displayed graphs as pictures, or saving<br />

<strong>the</strong> adjusted parameters <strong>of</strong> antenna array and/or <strong>the</strong><br />

selected surface.<br />

Figure 3: Demonstration program window<br />

68<br />

There is a screenshot <strong>of</strong> <strong>the</strong> program window in <strong>the</strong><br />

Figure 3. It allows setting <strong>the</strong> parameters <strong>of</strong> <strong>the</strong> antenna<br />

array under test (number <strong>of</strong> <strong>the</strong> dipoles and<br />

<strong>the</strong>ir positions and orientations) and <strong>the</strong> dimensions<br />

and resolution <strong>of</strong> <strong>the</strong> selected surface, where <strong>the</strong><br />

electromagnetic intensity distribution will be computed.<br />

III. CONCLUSION<br />

The program allows displaying <strong>the</strong> computed electromagnetic<br />

filed intensity distribution <strong>of</strong> <strong>the</strong> dipole<br />

antenna array on <strong>the</strong> selected planar or cylindrical<br />

surface. It is also possible to find out <strong>the</strong> computed<br />

values in concrete points from <strong>the</strong> displayed intensity<br />

distribution and to export computed data to graphic<br />

or data files.<br />

The created program can be used for <strong>the</strong> presentation<br />

<strong>of</strong> <strong>the</strong> antenna array radiation in a university<br />

laboratory or for <strong>the</strong> graphical representation when a<br />

new antenna is designed.<br />

REFERENCES<br />

[1] STRATTON, J. A.: Teorie elektromagnetického<br />

pole. SNTL, Praha 1961, 592 s.<br />

[2] �ERNOHORSKÝ, D., TICHÝ, J.: Vyza�ování a<br />

ší�ení rádiových vln, II. Díl Antény. Vojenská<br />

akademie Antonína Zápotockého, 1977.<br />

[3] NOVÁ�EK, Z.: Elektromagnetické vlny, antény<br />

a vedení. Elektronické skriptum. Brno: VUT,<br />

FEKT, 2003, 135 s.


Ontology- and Bayesian-based<br />

Threat Probability Determination<br />

Stefan Fenz and A Min Tjoa (Faculty Mentor)<br />

Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {fenz,tjoa}@ifs.tuwien.ac.at<br />

Abstract — Information security risk management is crucial<br />

for ensuring long-term business success and thus numerous<br />

approaches to implementing an adequate information security<br />

risk management strategy have been proposed. The subjective<br />

threat probability determination is one <strong>of</strong> <strong>the</strong> main reasons for<br />

an inadequate information security strategy endangering <strong>the</strong><br />

organization in performing its mission. To address <strong>the</strong> problem<br />

this research project proposes an ontology- and Bayesianbased<br />

approach for determining asset-specific and comprehensible<br />

threat probabilities. The elaborated concepts enable risk<br />

managers to comprehensibly quantify <strong>the</strong> current security status<br />

<strong>of</strong> <strong>the</strong>ir organization.<br />

I. THE SECURITY ONTOLOGY<br />

A conceptual and formal model <strong>of</strong> information security<br />

is required for supporting <strong>the</strong> threat probability determination<br />

in <strong>the</strong> information security risk management process.<br />

Ontologies are one possibility for modeling <strong>the</strong> information<br />

security domain in order to make it accessible<br />

to machines. Therefore, <strong>the</strong> security ontology [1, 2, 3]<br />

was proposed based on <strong>the</strong> security relationship model<br />

described in <strong>the</strong> National Institute <strong>of</strong> Standards and Technology<br />

Special Publication 800-12. Figure 1 shows <strong>the</strong><br />

high-level concepts and corresponding relations <strong>of</strong> our<br />

ontology. A threat gives rise to follow-up threats, represents<br />

a potential danger to <strong>the</strong> organization’s assets and<br />

affects specific security attributes (e.g. confidentiality,<br />

integrity, and/or availability) as soon as it exploits a vulnerability<br />

in <strong>the</strong> form <strong>of</strong> a physical, technical, or administrative<br />

weakness, and it causes damage to certain assets.<br />

Additionally each threat is described by potential<br />

threat origins (human or natural origin) and threat sources<br />

(accidental or deliberate source). For each vulnerability<br />

a severity value and <strong>the</strong> asset on which <strong>the</strong> vulnerability<br />

could be exploited is assigned. Controls have to be<br />

implemented to mitigate an identified vulnerability and<br />

to protect <strong>the</strong> respective assets by preventive, corrective,<br />

deterrent, recovery, or detective measures (control type).<br />

Each control is implemented as asset concept, or as combinations<br />

<strong>the</strong>re<strong>of</strong>. Controls are derived from and correspond<br />

to best-practice and information security standard<br />

controls. To enrich <strong>the</strong> knowledge model with concrete<br />

information security knowledge <strong>the</strong> German IT Grundschutz<br />

Manual is superimposed on <strong>the</strong> security ontology<br />

and more than 500 information security concepts and 600<br />

69<br />

corresponding formal axioms are integrated into <strong>the</strong> ontological<br />

knowledge base. The controls are modeled on<br />

a highly granular level and are thus reusable for different<br />

standards. When implementing <strong>the</strong> controls, a compliance<br />

with various information security standards is implicit.<br />

The coded ontology follows <strong>the</strong> OWL-DL (W3C<br />

Web Ontology Language) standard and ensures that <strong>the</strong><br />

knowledge is represented in a standardized, formal, and<br />

<strong>the</strong>refore machine-interpretable form.<br />

Control Type<br />

<strong>of</strong> Type implemented by<br />

threatens gives rise to affects<br />

Standard Control Control<br />

Severity Scale<br />

corresponds<br />

to<br />

Organization<br />

owned<br />

by<br />

requires<br />

Asset<br />

vulnerability on<br />

requires Level<br />

mitigated by exploited by<br />

Vulnerability<br />

has severity<br />

Threat<br />

Figure 1: Security relationships<br />

has<br />

source<br />

has origin<br />

Security Attribute<br />

Threat Source<br />

Threat Origin<br />

II. BAYESIAN THREAT PROBABILITY DE-<br />

TERMINATION<br />

This section aims at describing <strong>the</strong> connection to <strong>the</strong> security<br />

ontology framework, which provides a foundation<br />

to enrich <strong>the</strong> Bayesian network with concrete knowledge.<br />

Since <strong>the</strong> security ontology provides detailed knowledge<br />

about threat, vulnerability, and control dependencies, this<br />

knowledge could be utilized to build up <strong>the</strong> Bayesian network<br />

for <strong>the</strong> threat probability determination. Figure 2<br />

gives an overview <strong>of</strong> <strong>the</strong> connections between <strong>the</strong> proposed<br />

Bayesian threat probability determination and <strong>the</strong><br />

security ontology. It is assumed that each node has exactly<br />

one <strong>of</strong> a finite set <strong>of</strong> probability states (expressed as<br />

a vector, representing <strong>the</strong> probability distribution among<br />

distinct states, e.g. high, medium, and low). Since <strong>the</strong><br />

threat probability or influencing factors cannot be determined<br />

quantitatively, a qualitative rating is used in this<br />

approach. For each variable a three-point Likert scale<br />

was defined to capture <strong>the</strong> subjective impressions on <strong>the</strong><br />

input variables and to represent <strong>the</strong> results on <strong>the</strong> intermediate<br />

and output variables.<br />

First <strong>of</strong> all, <strong>the</strong> approach has to set up a threat net, in-


Threat Ti<br />

Probability PPTi<br />

sec:canBeConsequenceOf<br />

Threat ST1Ti<br />

Probability PPST1Ti<br />

sec:exploits<br />

….<br />

Threat Ti<br />

Probability PPTi<br />

Vulnerabilities VSTi<br />

Exploitation Probability<br />

PPVSTi<br />

Threat PTnTi<br />

Probability PPPTnTi<br />

...<br />

WPPVSTi<br />

Threat PT1Ti<br />

Probability PPPT1Ti<br />

Threat STnTi<br />

Probability PPSTnTi<br />

WPTnTi<br />

WPT1Ti<br />

Vulnerability V1<br />

Exploitation Probability<br />

PPV1<br />

…..<br />

Vulnerability Vn<br />

Exploitation Probability<br />

PPVn<br />

sec:givesRiseTo<br />

WPPV1<br />

WPPVn<br />

WCE1<br />

Control<br />

Implementation<br />

Effectiveness CE1<br />

sec:implementedBy<br />

sec:mitigatedBy<br />

….<br />

Control Combination<br />

Effectiveness CCEVi<br />

Vulnerability Vi<br />

Exploitation Probability<br />

PPVi<br />

Attacker Motivation<br />

AMVi<br />

WCEn<br />

Control<br />

Implementation<br />

Effectiveness CEn<br />

WCCEVi<br />

WPPVi<br />

WACVi<br />

Attacker Capability<br />

ACVi<br />

WAMVi<br />

Threat Ti<br />

A Priori Probability<br />

APTi<br />

V<br />

WAPTi<br />

WAEVi<br />

Attacker Effectiveness<br />

ΑEVi<br />

Figure 2: Utilizing <strong>the</strong> security ontology for <strong>the</strong> Bayesian<br />

threat probability determination<br />

cluding <strong>the</strong> relations between <strong>the</strong> threats and <strong>the</strong>ir a priori<br />

threat probability. Since each threat modeled in <strong>the</strong> security<br />

ontology is connected by <strong>the</strong> relation sec:givesRiseTo<br />

to follow-up threats (see Figure 2) <strong>the</strong> corresponding<br />

threat net can easily be created. The a priori threat proba-<br />

bility vector �<br />

APTi for each threat Ti is also derived from<br />

<strong>the</strong> security ontology, depending on <strong>the</strong> actual physical<br />

location <strong>of</strong> <strong>the</strong> organization. The sec:Probability concept<br />

and <strong>the</strong> sec:probabilityDistribution property <strong>of</strong> <strong>the</strong><br />

security ontology connect each threat <strong>of</strong> a given physical<br />

location with its a priori probability. Weights for all<br />

threat probability influencing factors (influencing threats<br />

and vulnerabilities) are distributed equally.<br />

For each threat <strong>the</strong> approach has to determine <strong>the</strong> corresponding<br />

vulnerabilities. In <strong>the</strong> security ontology this<br />

relationship is modeled by <strong>the</strong> sec:exploits relation which<br />

allows revealing <strong>the</strong> vulnerabilities <strong>of</strong> a given threat. As<br />

<strong>the</strong> vulnerabilities vector � PPVST is determined by sin-<br />

i<br />

gle vulnerabilities and <strong>the</strong>ir weights, <strong>the</strong> weight <strong>of</strong> each<br />

vulnerability which influences <strong>the</strong> intermediate vulnerabilities<br />

vector � PPVST was determined. Since <strong>the</strong> secu-<br />

i<br />

rity ontology provides a severity rating SVi for each vul-<br />

ontology regarding those control implementations effectiveness<br />

values which are relevant for <strong>the</strong> considered asset/vulnerability<br />

combination.<br />

With <strong>the</strong> security ontology relation sec:mitigatedBy<br />

(see Figure 2) <strong>the</strong> required control implementation combination<br />

which is necessary to mitigate <strong>the</strong> given vulnerability<br />

can be derived. Since each implementation<br />

in <strong>the</strong> recommended control combination has a different<br />

effectiveness � CEi, <strong>the</strong> weight WCEi differs dependently<br />

on <strong>the</strong> implementation’s importance for <strong>the</strong> current<br />

control combination. The security ontology concept<br />

sec:ControlImplementation represents <strong>the</strong> effectiveness<br />

for each control/implementation combination � CEi by a<br />

three-point Likert scale (high, medium, low).<br />

III. RESULTS<br />

The question is if <strong>the</strong> proposed Bayesian threat probability<br />

determination is <strong>the</strong> solution to <strong>the</strong> fundamental information<br />

security risk management problem, namely disposing<br />

<strong>of</strong> realistic probability values. The answer is nei<strong>the</strong>r<br />

yes nor no. The advantage <strong>of</strong> <strong>the</strong> proposed Bayesian<br />

threat probability determination is that it gives <strong>the</strong> risk<br />

manager a methodology to determine <strong>the</strong> threat probability<br />

in a structured and, by incorporating <strong>the</strong> security<br />

ontology, comprehensible way. The calculation schema<br />

is fully documented and each state <strong>of</strong> <strong>the</strong> Bayesian network<br />

can be explained and justified ma<strong>the</strong>matically and<br />

formally taking <strong>the</strong> given input factors into consideration.<br />

However, <strong>the</strong> high dependence on realistic input values<br />

requires fur<strong>the</strong>r research on sound methods to ga<strong>the</strong>r,<br />

store, and provide <strong>the</strong>se crucial threat probability calculation<br />

components.<br />

REFERENCES<br />

nerability (high (3), medium (2), and low(1)), a numerical<br />

weight WPPV for each vulnerability can be deter-<br />

i<br />

mined by dividing <strong>the</strong> severity <strong>of</strong> <strong>the</strong> considered vulnerability<br />

by <strong>the</strong> severity sum <strong>of</strong> all vulnerabilities relevant<br />

SV<br />

to <strong>the</strong> threat: WPPV = � i<br />

n<br />

i<br />

j=1 SVj The exploitation probability <strong>of</strong> each vulnerability variable<br />

is determined by (1) <strong>the</strong> effectiveness <strong>of</strong> <strong>the</strong> implemented<br />

control combination � CCEVi , (2) <strong>the</strong> attacker’s<br />

effectiveness � AEVi in <strong>the</strong> case <strong>of</strong> a deliberate threat<br />

source or by <strong>the</strong> a priori threat probability � APTi in <strong>the</strong><br />

case <strong>of</strong> an accidental threat source. By default all components,<br />

namely � CCEVi , � AEVi , and � APTi , are weighted<br />

equally. While <strong>the</strong> attacker’s effectiveness � AEVi and<br />

[1] Andreas Ekelhart, Stefan Fenz, Markus Klemen, and<br />

Edgar R. Weippl. Security ontology: Simulating<br />

threats to corporate assets. In A. Bagchi and V.<br />

Atluri, editors, Information Systems Security, Second<br />

International <strong>Conference</strong>, ICISS 2006, volume<br />

4332/2006 <strong>of</strong> Lecture Notes in Computer Science,<br />

pages 249–259, Kolkata, India, December 2006.<br />

Springer Berlin / Heidelberg. 978-3-540-68962-1.<br />

[2] Andreas Ekelhart, Stefan Fenz, Gernot Goluch, and<br />

Edgar Weippl. Ontological mapping <strong>of</strong> common<br />

criteria’s security assurance requirements. In H.<br />

Venter, M. El<strong>of</strong>f, L. Labuschagne, J. El<strong>of</strong>f, and R.<br />

von Solms, editors, <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> IFIP TC 11<br />

22nd International Information Security <strong>Conference</strong>,<br />

IFIPSec 2007, pages 85–95, Sandton, South Africa,<br />

May 2007. IFIP. 978-0-387-72366-2.<br />

[3] Thomas Neubauer, Andreas Ekelhart, and Stefan<br />

Fenz. Interactive selection <strong>of</strong> ISO 27001 controls under<br />

multiple objectives. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> IFIP<br />

<strong>the</strong> a priori threat probability � APTi are not rated on an TC 11 23rd International Information Security Con-<br />

asset-specific level, <strong>the</strong> control combination effectiveness ference, IFIPSec <strong>2008</strong>, pages 477–492, Boston, July<br />

�<br />

<strong>2008</strong>. Springer.<br />

CCEVi is determined specifically for <strong>the</strong> considered asset.<br />

Therefore, reasoning algorithms query <strong>the</strong> security<br />

70


Development <strong>of</strong> Multi-Core Video Decoding Platforms based on<br />

High-Level Architecture Simulations<br />

Florian Seitner, Michael Bleyer (Faculty Mentor) and Margrit Gelautz (Faculty Mentor)<br />

Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {seitner, bleyer, gelautz}@ims.tuwien.ac.at<br />

Abstract — The high computational demands <strong>of</strong> state-<strong>of</strong>-<strong>the</strong>art<br />

video coding standards pose serious challenges on strongly<br />

resource-restricted architectures. For reaching <strong>the</strong> performance<br />

specifications, specialized multi-core architectures for<br />

video processing are becoming more and more popular. In this<br />

work, we introduce an high-level simulator for supporting <strong>the</strong><br />

development <strong>of</strong> such decoding platforms. Our system combines<br />

all available information such as hardware measurements, pr<strong>of</strong>ilings<br />

and human expertise. Based on this input, <strong>the</strong> behaviour<br />

<strong>of</strong> <strong>the</strong> final architecture running a parallel video decoder is<br />

estimated. Using this information, adaptations <strong>of</strong> <strong>the</strong> current<br />

hardware or s<strong>of</strong>tware design can be done. The simulator shall<br />

aid in developing efficient and application-optimized decoding<br />

systems.<br />

I. INTRODUCTION<br />

The H.264 video standard [1] is currently used in a wide<br />

range <strong>of</strong> video-related areas such as video content distribution<br />

and television broadcasting. Compared to preceding<br />

standards such as MPEG-2 and MPEG-4 SP/ASP,<br />

improved coding efficiency is reached by introducing<br />

more advanced pixel processing algorithms (e.g. quarterpixel<br />

motion estimation) as well as by <strong>the</strong> use <strong>of</strong> more<br />

sophisticated algorithms for predicting syntax elements<br />

from neighbouring macroblocks (e.g. context-adaptive<br />

VLC). These new coding tools result in significantly increased<br />

CPU and memory loads required for decoding<br />

<strong>the</strong> video stream. In environments <strong>of</strong> limited processing<br />

power such as embedded systems, <strong>the</strong> high computational<br />

demands pose a serious challenge for practical<br />

H.264 implementations. Multi-core systems provide an<br />

elegant and power-efficient solution to overcome <strong>the</strong>se<br />

performance limitations.<br />

The design <strong>of</strong> such a specialized multi-processor decoding<br />

architecture is a non-trivial task. For exploiting<br />

<strong>the</strong> processing power <strong>of</strong> a multi-core system most efficiently,<br />

an equal workload between <strong>the</strong> cores must be<br />

achieved. Apart from <strong>the</strong> system’s usage this also influences<br />

<strong>the</strong> required buffer sizes between <strong>the</strong> cores for<br />

compensating differences in <strong>the</strong> workload. However, <strong>the</strong><br />

significant workload differences in typical video decoding<br />

systems make this balancing a challenging task.<br />

Figure 1 visualizes <strong>the</strong> structure <strong>of</strong> <strong>the</strong> H.264 decod-<br />

71<br />

ing process. The computational complexity <strong>of</strong> a macroblock’s<br />

parsing and deblocking functions is strongly<br />

bitrate dependent. While <strong>the</strong> parsing complexity typically<br />

raises with <strong>the</strong> macroblock’s number <strong>of</strong> bits and<br />

syntax elements, <strong>the</strong> deblocking filter is applied more aggressively<br />

for low bitrates. For <strong>the</strong> pixel-based decoding<br />

tasks (e.g. intra prediction and motion compensation) a<br />

large variety <strong>of</strong> possible macroblock coding modes exists.<br />

The coding options such as <strong>the</strong> prediction type<br />

(i.e. for H.264 skipped, intra and inter prediction) and<br />

<strong>the</strong> macroblock partitioning influence <strong>the</strong> decoding complexity<br />

significantly.<br />

For multi-core video decoding systems, predicting <strong>the</strong><br />

run-time behaviour is not straight forward. Differences<br />

in <strong>the</strong> workload, algorithmic dependencies and stalls due<br />

to resource limitations (e.g. size restrictions <strong>of</strong> communication<br />

buffers) must be considered. Fur<strong>the</strong>rmore, video<br />

decoding platforms are <strong>of</strong>ten based on heterogeneous architectures<br />

for addressing <strong>the</strong> execution behaviour <strong>of</strong> <strong>the</strong><br />

individual decoding tasks more efficiently. For example,<br />

<strong>the</strong> highly conditional parsing and entropy decoding<br />

functions require processors with efficient branch execution.<br />

Depending on which processor a decoding task is<br />

executed, differences in <strong>the</strong> decoder run-time will occur.<br />

The dynamics and heterogeneity <strong>of</strong> video decoding<br />

systems <strong>of</strong>ten result in <strong>the</strong> following questions:<br />

– Can we reach <strong>the</strong> specified decoding requirements<br />

on a specified multi-core architecture?<br />

– Which architecture is required to handle a certain<br />

set <strong>of</strong> videos (i.e. a set <strong>of</strong> streams that represents<br />

<strong>the</strong> common input characteristics <strong>of</strong> an application<br />

such as DVB-T)?<br />

– What is <strong>the</strong> optimal decoder hardware and s<strong>of</strong>tware<br />

partitioning for this architecture?<br />

These questions have a major impact on <strong>the</strong> architecture<br />

decisions and should be addressed before implementing<br />

a video decoding system.<br />

For solving <strong>the</strong>se questions, assumptions about <strong>the</strong> architecture<br />

requirements regarding <strong>the</strong> computational decoding<br />

complexity are typically made. They allow us<br />

to decide on <strong>the</strong> hardware components and <strong>the</strong> s<strong>of</strong>tware


Encoded<br />

Bitstream<br />

Stream<br />

Parsing<br />

Entropy<br />

Decoder<br />

IDCT<br />

Spatial<br />

Prediction<br />

Motion<br />

Compensation<br />

Reference<br />

Frames<br />

+<br />

Deblocking<br />

Decoded<br />

Frame<br />

Figure 1: The H.264 decoding process. After parsing and entropy decoding <strong>the</strong> syntax elements <strong>of</strong> a macroblock, a<br />

spatial or temporal prediction <strong>of</strong> this macroblock is computed. This prediction is added to <strong>the</strong> inverse transformed<br />

(IDCT) residual information and a deblocking filter removes blocking artefacts introduced during <strong>the</strong> quantisation<br />

step.<br />

structure <strong>of</strong> our decoder. Complexity estimation techniques<br />

are <strong>the</strong>refore <strong>of</strong> prime importance. In recent years,<br />

advanced techniques for estimating <strong>the</strong> run-time <strong>of</strong> a program<br />

have evolved.<br />

Static algorithm analysis and path analysis techniques<br />

have been introduced in [2]. These techniques analyse<br />

an algorithm’s definition (e.g. its source code) for estimating<br />

<strong>the</strong> upper and lower run-time bounds <strong>of</strong> a program.<br />

For considering <strong>the</strong> impact <strong>of</strong> <strong>the</strong> input data on<br />

<strong>the</strong> program execution, dynamic pr<strong>of</strong>iling methods [3]<br />

have been developed. These methods observe <strong>the</strong> program’s<br />

execution behaviour during <strong>the</strong> run-time. This<br />

allows us to measure <strong>the</strong> complexity <strong>of</strong> <strong>the</strong> decoding<br />

functions for individual input data. Hardware simulations<br />

and HW/SW-Codesign methods [4] provide accurate<br />

run-time information but require labour intensive<br />

adaptation <strong>of</strong> <strong>the</strong> hardware and decoder s<strong>of</strong>tware before<br />

first run-time estimations are possible.<br />

All <strong>the</strong>se approaches above can provide us with information<br />

about <strong>the</strong> complexity requirements <strong>of</strong> our video<br />

decoding system. However, interpretation <strong>of</strong> <strong>the</strong> complexity<br />

information in <strong>the</strong> context <strong>of</strong> a multi-core system<br />

is not straight-forward. Algorithmic dependencies<br />

and resource limitations result in run-time constraints between<br />

<strong>the</strong> processing units. For highly dynamic multicore<br />

decoder systems, making predictions about <strong>the</strong> parallel<br />

decoding system’s behaviour is hardly possible.<br />

In our work, we introduce a simulator for supporting<br />

<strong>the</strong> development process <strong>of</strong> multi-core video decoding<br />

systems. It estimates <strong>the</strong> basic parameters such as <strong>the</strong><br />

execution time, <strong>the</strong> memory transfers and <strong>the</strong> power consumption<br />

<strong>of</strong> <strong>the</strong> decoder system. Instead <strong>of</strong> running <strong>the</strong><br />

partitioned decoding tasks and using exactly specified interfaces<br />

for connecting <strong>the</strong>m toge<strong>the</strong>r, only an abstract<br />

information about <strong>the</strong> macroblocks’ runtime (e.g. decoding<br />

complexity) is required. This is a major advantage <strong>of</strong><br />

72<br />

our system, since <strong>the</strong> system designer typically gets this<br />

information without building <strong>the</strong> complete hardware architecture<br />

or without partitioning <strong>the</strong> s<strong>of</strong>tware <strong>of</strong> <strong>the</strong> decoder.<br />

Architecture evaluations are possible before <strong>the</strong><br />

decoder architecture is effectively built. Additionally,<br />

<strong>the</strong> Partition Assessment Tool (PAT) allows s<strong>of</strong>tware design<br />

explorations for supporting <strong>the</strong> partitioning <strong>of</strong> <strong>the</strong><br />

decoder s<strong>of</strong>tware.<br />

ACKNOWLEDGMENTS<br />

This work has been supported by ON DEMAND Microelectronics<br />

AG [5] and <strong>the</strong> Austrian Federal Ministry <strong>of</strong><br />

Transport, Innovation, and Technology under <strong>the</strong> FIT-IT<br />

Project 812429.<br />

REFERENCES<br />

[1] ITU-T and ISO/IEC. Advanced video coding for<br />

generic audiovisual services (ITU Rec. H.264 —<br />

ISO/IEC 14496-10). ITU-T and ISO/IEC, March<br />

2005.<br />

[2] Peter P. Puschner and Christian Koza. Calculating<br />

<strong>the</strong> maximum execution time <strong>of</strong> real-time programs.<br />

Journal <strong>of</strong> Real-Time Systems, 1(2):159–176, 1989.<br />

[3] Susan L. Graham, Peter B. Kessler, and Marshall K.<br />

McKusick. gpr<strong>of</strong>: a call graph execution pr<strong>of</strong>iler.<br />

In SIGPLAN Symposium on Compiler Construction,<br />

pages 120–126, 1982.<br />

[4] Peter Voigt Knudsen and Jan Madsen. Pace:<br />

A dynamic programming algorithm for hardware/s<strong>of</strong>tware<br />

partitioning. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> Int.<br />

Workshop on Hardware-S<strong>of</strong>tware Co-Design, pages<br />

85–92, 1996.<br />

[5] ON DEMAND Microelectronics.<br />

http://www.odmsemi.com, <strong>2008</strong>.


Comparison between Solute transport parameters estimation from<br />

local to field scale<br />

Mohamed Kassab 1 , Antonio Coppola 2 , Nicola Lamaddalena 3 , Alberto Guadagini 1 and Monica Riva 1<br />

1<br />

Politecnico di Milano, DIIAR, Piazza L. Da Vinci, Milano, Italy,<br />

2<br />

Dept. Tecnico Economico per la Gestione del Territorio Agr. forestale, University <strong>of</strong> Basilicata, Potenza, Italy<br />

3<br />

Mediterranean Agronomic Institute <strong>of</strong> Bari (IAMB), Italy.<br />

Email: Mohamed.kassab@mail.polimi.it<br />

Abstract — The generalized transfer function<br />

model (GTF) and TDR based time normalized resident<br />

concentrations were combined to characterize<br />

solute transport mechanisms at local and field<br />

scales. A leaching experiment was conducted under<br />

greenhouse, where TDR probes were installed at<br />

three different depths at 37 sites. The field plot was<br />

brought to steady-state water content; a pulse application<br />

<strong>of</strong> 3.87 mm <strong>of</strong> KCl solution was applied.<br />

Measurements <strong>of</strong> water content (�) and impedance<br />

(Z) were simultaneously taken and <strong>the</strong> time series <strong>of</strong><br />

relative resident concentrations for each site where<br />

effectively interpreted in terms <strong>of</strong> GTF model. The<br />

field scale behavior was described by calculating a<br />

local average and an integral average which gave<br />

two significantly different field scale solute transport<br />

behaviors.<br />

I. INTRODUCTION<br />

Sustainable management <strong>of</strong> soil and water resources<br />

depends largely on <strong>the</strong> ability to predict <strong>the</strong> impacts<br />

<strong>of</strong> current and future management practices on soil<br />

and groundwater quality both at local and global<br />

scale.<br />

The complex heterogeneity <strong>of</strong> soil has encouraged<br />

<strong>the</strong> development <strong>of</strong> transport <strong>the</strong>ories based on conceptual<br />

models, such as <strong>the</strong> transfer function approach<br />

[1]. [2] presented a generalized transfer function<br />

models (GTF) which is a four-parameter flexible<br />

transfer function able to describe both <strong>the</strong> convection-dispersive<br />

(CD) and <strong>the</strong> stochastic-convective<br />

(SC) process <strong>of</strong> dispersion in a soil, and solute transport<br />

processes in heterogeneous soils such as those in<br />

which <strong>the</strong> mean travel time increase non-linearly<br />

with depth and when <strong>the</strong> dispersivity is a scale dependent<br />

function. Finally, [3] presented an analytical<br />

solution for time normalized resident concentration<br />

for <strong>the</strong> GTF model.<br />

This work amid at improving <strong>the</strong> link between <strong>the</strong><br />

small scale variability <strong>of</strong> <strong>the</strong> local transport properties<br />

and large scale transport behavior by characterizing<br />

local solute transports mechanisms on several<br />

sites along transect area by using TDR probes at<br />

73<br />

different depths and in terms <strong>of</strong> GTF model, <strong>the</strong>n<br />

integrating local scale measurement and parameters<br />

along a transect area and normalized for <strong>the</strong> different<br />

local water contents, to build a time-integral normalized<br />

resident concentration at each depth to be used<br />

as field-scale curves.<br />

II. MATERIALS AND METHODS<br />

A leaching experiment was carried out in a Terra<br />

Rossa (sand 23.5%, silt 58.7%, clay 17.8%) soil plot<br />

under greenhouse at IAMB, Bari. TDR probes were<br />

installed at three different depths (0-15, 20 and 40<br />

cm) at 37 sites 1m apart along a 40 m transect. The<br />

field plot was brought to steady-state water content<br />

by adapting <strong>the</strong> drip irrigation to apply water at <strong>the</strong><br />

rate <strong>of</strong> 10 mm/day. An automatic irrigation scheduler<br />

was used to maintain <strong>the</strong> desired rate. A pulse application<br />

<strong>of</strong> 3.87 mm <strong>of</strong> KCl solution (23.5 g/l) was<br />

applied. Then, <strong>the</strong> fresh water was newly applied at<br />

10 mm/day to force <strong>the</strong> KCl solution downward into<br />

<strong>the</strong> soil.<br />

III. RESULTS AND DISCUSSIONS<br />

A. SOIL RESIDENT CONCENTRATIONS<br />

Measurements <strong>of</strong> both water content (�) and impedance<br />

(Z) were simultaneously taken to follow <strong>the</strong><br />

KCl solution propagation through <strong>the</strong> soil pr<strong>of</strong>ile<br />

“Figure 1”.<br />

Z(ohm)<br />

80.00<br />

70.00<br />

60.00<br />

50.00<br />

40.00<br />

0-15 cm<br />

20 cm<br />

40 cm<br />

30.00<br />

0.00 100.00 200.00 300.00 400.00<br />

t(hours)<br />

Figure 1: Soil impedance evolution <strong>of</strong> “Site 19”<br />

A closed-form expressions for <strong>the</strong> time normalized<br />

resident concentrations, Crt*, for <strong>the</strong> (GTF model)<br />

based on TDR measurements was applied to estimate<br />

solute transport parameters at both <strong>the</strong> local “Figure<br />

2” and field scale [4], [2] and [3].


Crt* (-)<br />

time normalized resident concentrations<br />

0.01<br />

0-15 cm measured<br />

20 cm measured<br />

40 cm measured<br />

0-15 cm simulated<br />

20 cm simulated<br />

40 cm simulated<br />

0.00<br />

0 100 200 300 400<br />

Time(h)<br />

Figure 2: The time normalized resident concentrations,<br />

Crt*, for “Site 19”<br />

The GTF scales <strong>the</strong> moments <strong>of</strong> travel times E(t) and<br />

Var(t) according to <strong>the</strong> following relationships:<br />

�1<br />

2�<br />

2 2(�1�<br />

�2<br />

)<br />

E(t,z) z Var(t,z) z CV(t,z) l<br />

(1)<br />

� � � �<br />

� �<br />

��<br />

� , ��<br />

� ; � 2 � �<br />

E(t,l) � l � Var(t,l) � l � CV(t,l) � z �<br />

where, �1; �2 are parameters <strong>of</strong> <strong>the</strong> time moments.<br />

The parameters �� and �� can be calculated from <strong>the</strong><br />

�1 and �2 parameters in <strong>the</strong> previous expressions,<br />

considering that <strong>the</strong> mean, E, and <strong>the</strong> variance, Var,<br />

<strong>of</strong> <strong>the</strong> travel times, t, are related to <strong>the</strong> mean (�) and<br />

variance (�2) <strong>of</strong> <strong>the</strong> ln(t) through [5] as following:<br />

2<br />

E�t, z��<br />

exp���0.<br />

5�<br />

�<br />

(2)<br />

z<br />

z<br />

2<br />

2<br />

Var�t, z��exp�2�<br />

� � ��exp����1� z z<br />

z<br />

(3)<br />

B. PARAMETERS AT LOCAL AND FIELD SCALES<br />

The field scale behavior was described by calculating<br />

a local and integral measurements average in terms<br />

<strong>of</strong> GTF model parameters. A significant different<br />

between <strong>the</strong> integral average (1.2) and local average<br />

(0.587) values was observed “Figure 3”.<br />

�1-�2<br />

1.5<br />

1.2<br />

0.9<br />

0.6<br />

0.3<br />

0<br />

0 10 20 30<br />

local values<br />

Integral average<br />

Local average<br />

Distance (m)<br />

Figure 3: The evolution with <strong>the</strong> distance <strong>of</strong> <strong>the</strong><br />

(������� values obtained along <strong>the</strong> transect.<br />

The time-normalized resident concentrations and <strong>the</strong><br />

dispersivity in terms <strong>of</strong> GTF parameters proposed by<br />

[4] were formulated by [3] as following:<br />

2<br />

2 2<br />

rt*<br />

aln(<br />

t)<br />

�a�<br />

� � � � � �<br />

z a�z<br />

�1<br />

(ln( t)<br />

�z<br />

�z)<br />

(4)<br />

C ( z,<br />

t)<br />

� exp��<br />

2 �<br />

t�1�<br />

2 � 2�<br />

z �<br />

z �<br />

2( �1<br />

��<br />

2 )<br />

z � l �<br />

2<br />

� � � � exp�����1�<br />

(5)<br />

l<br />

2 � z �<br />

Both local and integral average dispersivity values<br />

were significantly different although both <strong>of</strong> <strong>the</strong>m<br />

had <strong>the</strong> same trend with soil depth “Figure 4”. An<br />

integral average curve “Figure 5” was obtained by<br />

weighting Crt*(z, t) by (�� measured at <strong>the</strong> same<br />

depth and time based on [6]:<br />

40<br />

74<br />

Z cm)<br />

� (cm)<br />

0 10 20 30 40<br />

0<br />

10<br />

20<br />

30<br />

40<br />

50<br />

Local average<br />

Integral average<br />

Figure 4: Local and integral average dispersivity<br />

values as a function <strong>of</strong> depth.<br />

37<br />

*<br />

� � �z, x , t � Z fT �z, x , t � (6)<br />

rt *<br />

x �1<br />

C �z, t � � � 37<br />

*<br />

� � � �z, x , t � Z fT �z, x , t �dt<br />

0 x �1<br />

where ZfT*(z,t) is <strong>the</strong> temperature corrected impedance<br />

loads, x is location along <strong>the</strong> transect.<br />

Mean time normal time resident<br />

concentration<br />

<br />

0.03<br />

0<br />

0-15 cm measured<br />

20 cm measured<br />

35 cm measured<br />

0-15 cm simulated<br />

20 cm simulated<br />

35 cm simulated<br />

0 100 200 300 400<br />

Time (h)<br />

Figure 5: The integral average curves for <strong>the</strong> three<br />

depths, along with <strong>the</strong> best fitting obtained in terms<br />

<strong>of</strong> GTF parameters.<br />

REFERENCES<br />

[1] W. A. Jury. Simulation <strong>of</strong> solute transport using a<br />

transfer function model. Water resources research,<br />

18: 363–368, 1982.<br />

[2] R. D. Zhang. Generalized transfer function model<br />

for solute transport in heterogeneous soils. Soil<br />

science society <strong>of</strong> American journal, 64:1595–<br />

1602, 2000.<br />

[3] M. Javaux and M. Vanclooster. Robust estimation<br />

<strong>of</strong> <strong>the</strong> generalized solute transfer function<br />

parameters. Soil science society <strong>of</strong> American<br />

journal, 67:81-91, 2003.<br />

[4] J. Vanderborght, M. Vanclooster, D. Mallants, J.<br />

Diels, and J. Feyen. Determining convective lognormal<br />

solute transport parameters from resident<br />

concentration data. Soil science society <strong>of</strong><br />

American journal ,60:1306–1317, 1996.<br />

[5] T. B. Parkin, J. J. Meisinger, S. T. Chester, J. L<br />

Starr. and J. A. Robinson. Evaluation <strong>of</strong> statistical<br />

estimation methods for lognomally distributed<br />

variables. Soil science society <strong>of</strong> American<br />

journal, 52:323–329, 1988.<br />

[6] M. Javaux. Solute transport in a heterogeneous<br />

unsaturated subsoil: experiments and modeling.<br />

Ph.D. Thesis n45 . Louvain-la-Neuve. Faculte<br />

d'ingenierie biologique, agronomique et environnementale,<br />

Presses universitaires de Louvain.<br />

194 p., 2004.


A new method for identification <strong>of</strong> systems <strong>of</strong> arbitrary real order<br />

based on solution <strong>of</strong> fractional differential equations and orthogonal<br />

distance fitting<br />

Tomas Skovranek 1 , Vladimir Despotovic 2 , Igor Podlubny 1 (Faculty Mentor)<br />

1 Faculty <strong>of</strong> BERG, Institute <strong>of</strong> Control and Informatization <strong>of</strong> Production Processes<br />

Technical University <strong>of</strong> Kosice<br />

Kosice, Slovak Republic<br />

2 Technical Faculty <strong>of</strong> Bor, Department <strong>of</strong> Electromechanical Engineering<br />

University <strong>of</strong> Belgrade<br />

Bor, Serbia<br />

Email: {tomas.skovranek,igor.podlubny}@tuke.sk; vdespotovic@tf.bor.ac.yu<br />

Abstract — In this paper <strong>the</strong> total least squares method<br />

(TLSM), <strong>the</strong> so called orthogonal distance regression (ODR)<br />

will be used for system identification instead <strong>of</strong> <strong>the</strong> classical regression<br />

based on <strong>the</strong> least squares method (LSM). The main<br />

reason is <strong>the</strong> fact that ODR is a suitable tool for fitting lines<br />

and surfaces in multidimensional space, while even <strong>the</strong> use<br />

<strong>of</strong> classical regression in 2D is not so natural. We apply <strong>the</strong><br />

ODR to define <strong>the</strong> optimal data-fitting-line in <strong>the</strong> state space. A<br />

new method for Identification <strong>of</strong> systems <strong>of</strong> arbitrary real order<br />

based on solution <strong>of</strong> fractional differential equations and orthogonal<br />

distance fitting is proposed. In order to define a model<br />

<strong>of</strong> observed data a system <strong>of</strong> fractional-order differential equations<br />

will be considered, which solution represents <strong>the</strong> optimal<br />

fitting <strong>of</strong> given points in space. The sum <strong>of</strong> orthogonal distances<br />

between <strong>the</strong> data points in space and a fitting line is considered<br />

as <strong>the</strong> criterion for optimization. The results are presented in<br />

2-dimensional and 3-dimensional space.<br />

I. INTRODUCTION<br />

System identification is a general term to describe ma<strong>the</strong>matical<br />

tools and algorithms that build dynamical models<br />

from experimental data. Having <strong>the</strong> model <strong>of</strong> a system is<br />

<strong>of</strong>ten very important for analysis, simulation, control system<br />

design etc. In this paper a new method is proposed<br />

for identification <strong>of</strong> systems <strong>of</strong> arbitrary real order based<br />

on solution <strong>of</strong> fractional differential equations. In order<br />

to successfully solve a problem <strong>of</strong> system identification,<br />

parameters <strong>of</strong> <strong>the</strong> differential equations (including orders<br />

<strong>of</strong> differentiation which are considered as arbitrary real or<br />

”fractional”) have to be evaluated, and a criterion must be<br />

determined, by which <strong>the</strong> estimation process will be done<br />

optimally.<br />

LSM was used for long period <strong>of</strong> time among scientists<br />

and researchers as an universal tool for finding bestfitting<br />

curve to a given set <strong>of</strong> points, although this method<br />

has many disadvantages, which lead to development <strong>of</strong><br />

more general methods, for example TLSM or ODR [1, 2].<br />

ODR uses perpendicular (orthogonal) distances between<br />

given points and fitting curve, what is more natural point<br />

<strong>of</strong> view than using vertical <strong>of</strong>fsets, as in LSM.<br />

75<br />

Although <strong>the</strong> fractional calculus and <strong>the</strong> idea <strong>of</strong> fractional<br />

order operators is over three centuries old, <strong>the</strong><br />

interest <strong>of</strong> scientists increased mainly in last decades<br />

[3, 4, 5]. Fractional-order differential equations (FODE)<br />

and <strong>the</strong>ir numerical solution is <strong>the</strong> topic <strong>of</strong> many authors<br />

(e.g. Die<strong>the</strong>lm, He, Podlubny, Lubich, Adomian, Momani,<br />

etc.). It is due to better possibilities <strong>of</strong> description<br />

<strong>of</strong> dynamical systems, as FODEs provide a powerful instrument<br />

for description <strong>of</strong> memory and hereditary properties<br />

<strong>of</strong> systems in comparison to integer-order models,<br />

where such effects are neglected.<br />

II. IDENTIFICATION OF LINEAR FRAC-<br />

TIONAL ORDER SYSTEMS USING OR-<br />

THOGONAL DISTANCE FITTING<br />

The main goal <strong>of</strong> this paper is <strong>the</strong> state space description<br />

<strong>of</strong> a dynamical system with a set <strong>of</strong> FODEs. Indeed, we<br />

want to define a ma<strong>the</strong>matical model <strong>of</strong> this dynamical<br />

system, by finding a solution <strong>of</strong> a set <strong>of</strong> FODEs, which<br />

represents <strong>the</strong> optimal fitting <strong>of</strong> given points in space.<br />

Criterion <strong>of</strong> optimization is <strong>the</strong> sum <strong>of</strong> shortest distances<br />

between given points in space and fitting line. The analysis<br />

is performed in 2D and 3D space.<br />

A. SOLUTION TO A SYSTEM WITH OPTIMIZED COEF-<br />

FICIENTS IN 2D SPACE<br />

Let us first analyze a system <strong>of</strong> two linear FODEs. The<br />

analytical interpretation <strong>of</strong> this system is:<br />

D (α) x = a1 · x + a2 · y + C1<br />

D (β) y = a3 · x + a4 · y + C2,<br />

(1)<br />

where <strong>the</strong> orders <strong>of</strong> differentiation (α, β) are fractional,<br />

in interval 0


where <strong>the</strong> expression (−1) j� � α<br />

j is represented by <strong>the</strong> coefficient<br />

b, and can be evaluated using recurrence rela-<br />

tionship as in [5]:<br />

b (α)<br />

0<br />

=1, b(α)<br />

j<br />

=<br />

�<br />

1 − 1+α<br />

�<br />

b<br />

j<br />

(α)<br />

j−1 , j =1, 2, 3 ....<br />

(3)<br />

Applying definitions (2)-(3) on (1) we obtain a numerical<br />

expression <strong>of</strong> <strong>the</strong> system <strong>of</strong> two linear FODEs:<br />

(a1 − h −α · b (α)<br />

0 ) · xm + a2 · ym<br />

= h −α · � b (α)<br />

j<br />

· xm−j − C1<br />

a3 · xm +(a4 − h −β · b (β)<br />

0<br />

= h −β · � b (β)<br />

j<br />

) · ym<br />

· ym−j − C2<br />

Figure 1: Fitting in 2D space<br />

(4)<br />

B. SOLUTION TO A SYSTEM WITH OPTIMIZED COEF-<br />

FICIENTS IN 3D SPACE<br />

Now we will deal with a system in 3D space, which analytical<br />

interpretation can be written as:<br />

D (α) x = a1 · x + a2 · y + a3 · z + C1<br />

D (β) y = a4 · x + a5 · y + a6 · z + C2<br />

D (γ) z = a7 · x + a8 · y + a9 · z + C3.<br />

(5)<br />

The orders <strong>of</strong> differentiation (α, β, γ) are fractional, in<br />

interval 0


Configurable Meta-search in Human Resource Domain<br />

Abstract — In this paper, we propose a configurable<br />

meta-search engine in <strong>the</strong> human resource<br />

domain (HRD) to provide a platform for job and<br />

applicant tracking processes. The meta-search<br />

engine in HRD has been designed by applying semantic<br />

Web technologies, to solve <strong>the</strong> problems <strong>of</strong><br />

job seekers. We provide <strong>the</strong> solutions for automatic<br />

integration <strong>of</strong> data, structures and processes in<br />

HRD into a meta-search by <strong>the</strong> use <strong>of</strong> domain ontology.<br />

Use <strong>of</strong> domain ontology, HR-XML and multiple<br />

matchers for generation <strong>of</strong> integrated schema<br />

and interface, helps in understanding <strong>the</strong> meaning<br />

<strong>of</strong> terms and improves <strong>the</strong> quality <strong>of</strong> <strong>the</strong> search<br />

interface and search results. Experiments show that<br />

<strong>the</strong>re is improvement in schema/data matching and<br />

integration process, result extraction and identification<br />

process for meta-search engine by use <strong>of</strong> multiple<br />

matchers and domain ontology.<br />

I. INTRODUCTION<br />

A lot <strong>of</strong> information is available online via websites<br />

or databases but this information is heterogeneous<br />

and distributed. It is difficult for a user to visit<br />

every website to access this information. There must<br />

be a way that people can search heterogeneous and<br />

distributed information quickly. Primary tools to<br />

search <strong>the</strong> information on <strong>the</strong> Web are search engines,<br />

subject directories and social network search<br />

engines. Traditional search engines are based on<br />

keyword or phrase search without understanding <strong>the</strong><br />

meaning <strong>of</strong> <strong>the</strong> word and are unable to provide <strong>the</strong><br />

desired results to <strong>the</strong> user. O<strong>the</strong>r traditional search<br />

tools have different index structures, low recall and<br />

precision and do not provide comprehensive coverage<br />

<strong>of</strong> <strong>the</strong> Web. In short, Internet search still needs<br />

improvement. To overcome all <strong>the</strong>se problems, <strong>the</strong>re<br />

exist ano<strong>the</strong>r search tool “meta-search engine” and<br />

considered as an excellent choice for specific topic<br />

search from multiple heterogeneous search engines.<br />

Traditional job search engines lack <strong>the</strong> capability<br />

<strong>of</strong> processing structured queries and search is based<br />

only on statistical methods without any attempt to<br />

understand <strong>the</strong> meaning <strong>of</strong> <strong>the</strong> terms being used. For<br />

job seekers, access to large number <strong>of</strong> job search<br />

engines is difficult and sometimes existing job por-<br />

Tabbasum Naz and Jürgen Dorn (Faculty Mentor)<br />

Institute for Information Systems (184/2)<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {naz,dorn}@dbai.tuwien.ac.at<br />

77<br />

tals produce irrelevant results. As every search portal<br />

has different interface style, it can be confusing for<br />

<strong>the</strong> job seeker in giving job search criteria. Recruiting<br />

organizations also need some improvements in <strong>the</strong><br />

existing job search portals.<br />

Previous research shows that automatic integration<br />

<strong>of</strong> different search engines into a meta-search is<br />

not an easy task. There are many problems in <strong>the</strong><br />

construction <strong>of</strong> meta-search engines. For example<br />

some cannot be automatically constructed and provide<br />

hard wired solutions and o<strong>the</strong>r involve complex<br />

steps for development process. Automatic schema<br />

matching, schema integration, data integration,<br />

global attributes and value selection from multiple<br />

search engines for <strong>the</strong> meta-search interface is a<br />

tricky and time consuming task that needs more<br />

attention. O<strong>the</strong>r meta-search engines also lack <strong>the</strong><br />

capability <strong>of</strong> semantically understanding <strong>the</strong> terms.<br />

To solve <strong>the</strong>se problems in a domain, we have developed<br />

a prototype for a meta-search engine in <strong>the</strong> job<br />

domain that use modern technologies to help <strong>the</strong> job<br />

seekers and recruiting organizations<br />

II. RESEARCH METHODOLOGY<br />

We have developed a prototype for a configurable<br />

job meta-search engine to provide a platform for job<br />

and applicant tracking processes. We identify <strong>the</strong><br />

requirements <strong>of</strong> recruiters and job seekers to improve<br />

<strong>the</strong> efficiency <strong>of</strong> job meta-search. The system supports<br />

automatic integration <strong>of</strong> interfaces, data structures<br />

and processes. There are two processes involved<br />

in <strong>the</strong> construction <strong>of</strong> job meta-search engine<br />

in job domain i) meta-search engine creation process<br />

for recruiting organization and ii) meta-search engine<br />

usage process for job seekers. Our approach can<br />

automatically extract, match and integrate individual<br />

schemas and job results from different job search<br />

engines into one. Job results are converted to a single<br />

format for quick and easy navigation for a job seeker.<br />

Use <strong>of</strong> semantic Web technologies for meta-search<br />

engine in HRD provides convenience to <strong>the</strong> job<br />

seeker and save <strong>the</strong>ir precious time in job search. We<br />

have developed domain ontology in HRD that helps<br />

in <strong>the</strong> translation <strong>of</strong> individual extracted schemas<br />

from individual job search engines to common


schema. We also introduce design patterns for common<br />

components <strong>of</strong> meta-search engines. Presented<br />

design patterns for meta-search engines and <strong>the</strong>ir<br />

components are reusable, extendable and flexible.<br />

Presented design patterns accelerate <strong>the</strong> development<br />

process in meta-search domain and o<strong>the</strong>r related<br />

domains. Moreover, it promises higher quality <strong>of</strong><br />

developed solutions [1].<br />

III. CONTRIBUTIONS<br />

Meta-search engine creation process involves components<br />

i.e. job search engine selector, interface<br />

extractor, XML schema generator, job meta-search<br />

query interface generator. Different search engines<br />

use different representation concepts, data structures<br />

and different granularities <strong>of</strong> knowledge, so domainontology<br />

is used for translation between concepts.<br />

Query interface generator <strong>of</strong> meta-search engine is<br />

responsible for integration <strong>of</strong> meta-data in three steps<br />

i) schema matching and integration ii) data integration<br />

iii) and query interface generation. Schema<br />

matching process is actually a process <strong>of</strong> discovering<br />

correspondences between two schema elements that<br />

correspond to each o<strong>the</strong>r. There can be structural,<br />

syntactic and semantic heterogeneity in schemas.<br />

Manual or semi-automatic schema matching and<br />

mapping is tedious, complex and time consuming<br />

task because <strong>of</strong> open and heterogeneity nature <strong>of</strong><br />

Web data. Automatic schema matching and<br />

integration is always required to reduce manual<br />

efforts but require more researcher’s efforts to<br />

produce high quality techniques and results. Our<br />

main focus is to have an automatic schema/data<br />

matching and integration for meta-search engine and<br />

to resolve <strong>the</strong> semantic conflicts by utilizing <strong>the</strong><br />

commutative techniques from database and ontology<br />

community. We utilize multiple matchers i.e. name<br />

matcher, linguistic matcher, instance level matcher<br />

(top down matcher), number <strong>of</strong> instances matcher,<br />

combinational matcher, synonym matcher, ontology<br />

matcher for schema/data level matching and integration.<br />

Our combinational matcher uses stemming<br />

algorithm and string distance algorithms i.e. Cosine<br />

similarity, Levenshtein distance and Euclidian distance<br />

algorithm to improve <strong>the</strong> matching results. We<br />

use single ontology approach and utilize our domain<br />

ontology as a global ontology for <strong>the</strong> schema/data<br />

matching and integration process. Domain ontology<br />

snippet is used in finding mapping between attributes<br />

<strong>of</strong> two interfaces. A domain-specific ontology also<br />

identifies <strong>the</strong> relationship between <strong>the</strong> values <strong>of</strong> two<br />

interfaces. Identified mappings are stored in XML<br />

format for fur<strong>the</strong>r use by form generator and query<br />

dispatcher component. After schema/data matching<br />

and integration, a form for mete-search engine is<br />

78<br />

generated. The form generation is supported by<br />

XForms which enables <strong>the</strong> generation <strong>of</strong> <strong>the</strong> form<br />

from an XML schema and also <strong>the</strong> easy adaptation to<br />

different user clients.<br />

In HRD, meta-search engine usage process involves<br />

components including query dispatcher, information<br />

extractor (IE), result merger, duplicate<br />

result eliminator and result ranker. Above mentioned<br />

matchers and domain ontology is utilized again by<br />

<strong>the</strong> IE component, to identify <strong>the</strong> different concepts,<br />

data structures <strong>of</strong> a job description and job result<br />

attributes to convert <strong>the</strong>m to a single common format<br />

for <strong>the</strong> job seeker. The system solves problems <strong>of</strong><br />

different representational concepts in different job<br />

search engines [1].<br />

HR domain ontology used in job meta-search engine<br />

construction contains job related attributes from<br />

HR-XML, occupations from Standard Occupation<br />

Classification (SOC) and skill from International Cooperation<br />

Europe Ltd.<br />

Introduced design patterns architect complex<br />

meta-search engines construction processes and<br />

provides us with flexible design philosophy.<br />

IV. CONCLUSION<br />

We have presented an approach for integrating<br />

schema/data from different job portals in a metasearch<br />

in order to support job seeking people to master<br />

<strong>the</strong> large number <strong>of</strong> available job portals. Our<br />

main focus is on <strong>the</strong> automated extraction and integration<br />

<strong>of</strong> <strong>the</strong> structure <strong>of</strong> provided information. This<br />

structure is used in meta-search to integrate <strong>the</strong> different<br />

sources. Application <strong>of</strong> Semantic Web technologies<br />

in meta-search domain proved helpful and<br />

provides automatic and efficient schema/data integration,<br />

query translation, integrated query interface<br />

generation and result identification. Experiments<br />

show that <strong>the</strong>re is improvement in schema/data<br />

matching and integration by use <strong>of</strong> multiple matchers<br />

and domain ontology. It also increases <strong>the</strong> mapping<br />

accuracy. Integrated system provides a unified access<br />

to job seekers to navigate multiple search engines<br />

without visiting <strong>the</strong>m separately and increase <strong>the</strong><br />

Web coverage for <strong>the</strong>m in a short time. Overall metasearch<br />

engine in HRD will be helpful in reducing <strong>the</strong><br />

unemployment rate <strong>of</strong> a country.<br />

REFERENCES<br />

[1] T. Naz. Configurable Meta-search in Human<br />

Resource Domain, PhD Thesis, Institute for Information<br />

Systems, Vienna University <strong>of</strong> Technology,<br />

Austria (To publish in Dec <strong>2008</strong>)


The Definition <strong>of</strong> Secure Business Processes with Respect to<br />

Multiple Objectives<br />

Johannes Heurix, Thomas Neubauer (Faculty Mentor)<br />

Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {jheurix,tneubauer}@securityresearch.at<br />

Abstract — Business processes have gained more and more<br />

importance in today’s business environment, and <strong>the</strong>ir unimpeded<br />

execution is crucial for a company’s success. Since business<br />

processes are permanently exposed to several threats, organizations<br />

are forced to pay attention to security issues. Although<br />

security <strong>of</strong> business activities is widely considered as<br />

important, business processes and security aspects are <strong>of</strong>ten developed<br />

separately. Recent approaches for managing business<br />

process security focus on certain aspects only and neglect o<strong>the</strong>rs,<br />

thus not providing a holistic framework for analyzing process<br />

security and evaluating security safeguards. Often, <strong>the</strong>se<br />

safeguards are evaluated according to technical aspects only;<br />

multiple objectives are not considered.<br />

This work introduces a model-supported, risk-based multiobjective<br />

decision making methodology (MR-MOD) for <strong>the</strong> elicitation<br />

<strong>of</strong> security requirements <strong>of</strong> business processes, for <strong>the</strong><br />

analysis <strong>of</strong> assets, threats, and vulnerabilities, and for <strong>the</strong> selection<br />

<strong>of</strong> appropriate security technologies. Thereby it combines<br />

<strong>the</strong> strengths <strong>of</strong> different methods, including process modeling,<br />

quantitative risk assessment, and multiobjective decision<br />

making techniques, for <strong>the</strong> definition <strong>of</strong> Secure Business Processes.<br />

I. INTRODUCTION<br />

In recent times, business processes have <strong>of</strong>ten been <strong>the</strong><br />

target <strong>of</strong> a number <strong>of</strong> security hazards such as viruses,<br />

hacker attacks, or data <strong>the</strong>ft [1]. Given <strong>the</strong> importance<br />

<strong>of</strong> business processes and <strong>the</strong> fact that <strong>the</strong>y are permanently<br />

exposed to numerous vulnerabilities, it becomes<br />

obvious to include business processes into security considerations.<br />

As business processes generate value and<br />

<strong>the</strong>ir unimpeded execution is vital for <strong>the</strong> success <strong>of</strong> enterprises,<br />

decision makers and security experts are tasked<br />

with revising methods to secure <strong>the</strong>m against external or<br />

internal threats; legal requirements and <strong>the</strong> loss <strong>of</strong> value<br />

sustained after a successful attack result in <strong>the</strong> demand<br />

for appropriate security measures. Loss <strong>of</strong> value can<br />

be ei<strong>the</strong>r <strong>of</strong> monetary nature (e.g. loss <strong>of</strong> pr<strong>of</strong>it due to<br />

<strong>the</strong> interruption <strong>of</strong> business activities) and/or intangible<br />

(e.g. loss <strong>of</strong> reputation). A striking example illustrate<br />

<strong>the</strong> severity <strong>of</strong> security breaches is <strong>the</strong> Denial <strong>of</strong> Service<br />

(DoS) attack <strong>of</strong> February 2000 causing access problems<br />

<strong>of</strong> Yahoo!’s website, costing an estimated half a million<br />

US Dollars in just three hours [2].<br />

79<br />

A major issue with business process security is that<br />

security properties are mainly considered as ’technical’<br />

problems which have to be solved by dedicated security<br />

experts [3], while process managers are tasked with optimizing<br />

<strong>the</strong> core workflows, thus leading to business processes<br />

and security measures developed separately. Ano<strong>the</strong>r<br />

problem is how to measure it. Recent approaches<br />

depend on <strong>the</strong> calculation <strong>of</strong> aggregated values such as<br />

<strong>the</strong> Annualized Loss Expectancy (ALE) [4] or Return<br />

on Investment (ROI) for evaluating security improving<br />

measures. But relying solely on a single value for measuring<br />

security is inappropriate, considering <strong>the</strong> multiple<br />

factors that may play a role in selecting <strong>the</strong> correct security<br />

safeguards including issues such as safeguard costs,<br />

effectiveness, and user acceptance. Part <strong>of</strong> this measurement<br />

problem is <strong>the</strong> issue <strong>of</strong> ’How much (security) is<br />

enough’ [5], as security does not directly generate business<br />

value and does not directly improve <strong>the</strong> net pr<strong>of</strong>it<br />

[1]; investing in security can only prevent negative events<br />

or reduce related adverse effects. Decision makers have<br />

to find <strong>the</strong> optimal balance between <strong>the</strong> expenditure for<br />

security safeguards and <strong>the</strong>ir ability to prevent adverse<br />

events.<br />

II. DESIGN OF A MODEL-SUPPORTED,<br />

RISK-BASED MULTIOBJECTIVE DECI-<br />

SION MAKING PROCESS<br />

The proposed methodology is a risk assessment process<br />

based on process models and security safeguard selection.<br />

Whereas o<strong>the</strong>r existing frameworks <strong>of</strong>ten concentrate<br />

on a certain aspect <strong>of</strong> security and do not address all<br />

requirements for <strong>the</strong> definition <strong>of</strong> secure business processes,<br />

this approach provides a holistic methodology<br />

for eliciting security-related elements based on business<br />

processes, measuring <strong>the</strong> risks, and choosing an optimal<br />

safeguard portfolio to ensure business process security. It<br />

is a model-supported, risk-based multiobjective decision<br />

making process (MR-MOD) and incorporates elements<br />

<strong>of</strong> different disciplines and <strong>the</strong>ir strengths:<br />

Model-based Elicitation Process Process models are<br />

used for deriving security relevant assets and threats<br />

<strong>the</strong>y are exposed to.


Risk Assessment A risk assessment process provides a<br />

structured method to measure information security<br />

risks and to valuate security safeguard effectiveness.<br />

Workshop A workshop environment ensures consideration<br />

<strong>of</strong> <strong>the</strong> different opinions and expertise <strong>of</strong> different<br />

participating domain experts such as security<br />

experts, process owners, and o<strong>the</strong>r stakeholders.<br />

Multiobjective Decision Making A multiobjective decision<br />

making process takes multiple criteria into<br />

consideration and provides solutions that represent<br />

<strong>the</strong> best trade<strong>of</strong>fs <strong>of</strong> opposed factors.<br />

The MR-MOD framework consists <strong>of</strong> three distinctive<br />

phases, each <strong>of</strong> <strong>the</strong>m divided into several sub-steps (figure<br />

1):<br />

Phase 1: Modeling and Identification The first phase<br />

is tasked with <strong>the</strong> modeling and <strong>the</strong> identification<br />

<strong>of</strong> all security relevant entities based on business<br />

process models. This includes <strong>the</strong> creation <strong>of</strong> <strong>the</strong><br />

process model and <strong>the</strong> identification <strong>of</strong> all participating<br />

assets, <strong>the</strong> identification <strong>of</strong> applicable vulnerabilities<br />

and threats <strong>the</strong>y are exposed to, and <strong>the</strong><br />

identification <strong>of</strong> possible safeguards.<br />

Phase 2: Workshop-based Risk Assessment<br />

In <strong>the</strong> next phase, risks are composed as<br />

asset/vulnerability/threat-tuples, cost and benefit<br />

categories <strong>the</strong> safeguards are measured in are<br />

defined, and necessary quantitative data collected<br />

for <strong>the</strong> evaluation process. In addition, constraints<br />

and interdependencies <strong>of</strong> <strong>the</strong> safeguards are<br />

specified.<br />

Phase 3: Multiobjective Decision Making The final<br />

phase deals with <strong>the</strong> generation <strong>of</strong> Pareto-efficient<br />

solutions (safeguard combinations) with respect to<br />

<strong>the</strong> cost/benefit categories using <strong>the</strong> data collected<br />

in <strong>the</strong> previous phase, <strong>the</strong> analysis <strong>of</strong> <strong>the</strong> candidate<br />

solutions by modifying <strong>the</strong> upper and lower<br />

bounds for <strong>the</strong> categories, and <strong>the</strong> final selection<br />

<strong>of</strong> <strong>the</strong> optimal safeguard portfolio according to <strong>the</strong><br />

preferences <strong>of</strong> <strong>the</strong> decision makers.<br />

III. CONCLUSION<br />

In this work, a model aimed to be a holistic framework<br />

for <strong>the</strong> elicitation <strong>of</strong> security holes and <strong>the</strong> selection <strong>of</strong><br />

appropriate methods to satisfy <strong>the</strong> security requirements<br />

<strong>of</strong> business processes has been introduced. Its main advantage<br />

over o<strong>the</strong>r approaches is <strong>the</strong> combination <strong>of</strong> multiple<br />

techniques and <strong>the</strong> inheritance <strong>of</strong> <strong>the</strong>ir strengths to<br />

overcome <strong>the</strong>ir individual weaknesses and shortcomings.<br />

This methodology is supported by a s<strong>of</strong>tware tool.<br />

80<br />

�����������<br />

������������������������������<br />

����������<br />

��������������������������������<br />

�����������<br />

���������������������������������<br />

�������������������������<br />

��������������<br />

����������������������������<br />

��������������������������������<br />

������������������������<br />

��������������������<br />

����������������������������<br />

����������<br />

����������������������������<br />

����������<br />

�����������������������������<br />

���������������������<br />

�������������������������������<br />

���������<br />

�������������������������������<br />

���������<br />

Figure 1: Overview <strong>of</strong> <strong>the</strong> MR-MOD Phases<br />

REFERENCES<br />

[1] T. Neubauer, M. Klemen, and S. Biffl. Secure business<br />

process management: a roadmap. In ARES ’06:<br />

<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> First International <strong>Conference</strong> on<br />

Availability, Reliability and Security, page 8, 20-22<br />

April 2006.<br />

[2] P. Kedrosky. Hackers prey on our insecurities. The<br />

Wall Street Journal, February 2000.<br />

[3] G. Herrmann. Security and integrity requirements <strong>of</strong><br />

business processes - analysis and approach to support<br />

<strong>the</strong>ir realisation. In Consortium on Advanced Information<br />

Systems Engineering, pages 36–47, 1999.<br />

[4] Fips Publication (65), 1979.<br />

[5] K. J. SooHoo. How much is enough? a riskmanagement<br />

approach to computer security. Technical<br />

report, Consortium for Research on Information<br />

Security and Policy (CRISP), June 2000.


Generalising Constraint Solving over Finite Domains<br />

Markus Triska and Nysret Musliu (Faculty Mentor)<br />

Institut für Informationssysteme 184/2<br />

Technische Universität Wien<br />

Vienna, Austria<br />

Email: {triska,musliu}@dbai.tuwien.ac.at<br />

Abstract — Traditionally, finite domain constraint solvers<br />

can typically handle only ra<strong>the</strong>r small values. This limits <strong>the</strong>ir<br />

potential applications. In this project, we devise and study a<br />

generalisation <strong>of</strong> constraint solving over finite domains, where<br />

domains can be unbounded. We present and discuss new application<br />

opportunities for constraint solvers that this feature<br />

allows. We are also interested in proving <strong>the</strong>oretical properties<br />

<strong>of</strong> a new constraint solver that is implemented as part <strong>of</strong> this<br />

project, such as termination and correctness.<br />

I. INTRODUCTION<br />

Constraint programming is a declarative formalism for<br />

describing and solving combinatorial problems from various<br />

areas, such as planning, scheduling and allocation<br />

tasks. It is one <strong>of</strong> <strong>the</strong> most promising developments in<br />

programming languages <strong>of</strong> <strong>the</strong> last decade, and <strong>the</strong> Association<br />

for Computing Machinery (ACM) has identified<br />

it as one <strong>of</strong> <strong>the</strong> strategic directions in computing research<br />

[1].<br />

Formally, a constraint satisfaction problem (CSP) consists<br />

<strong>of</strong>:<br />

– asetX<strong>of</strong> variables, X = {x1,...,xn}<br />

– for each variable xi, asetD(xi) <strong>of</strong> values that xi<br />

can assume, which is called <strong>the</strong> domain <strong>of</strong> xi<br />

– aset<strong>of</strong>constraints, which are simply relations<br />

among variables in X, and which can fur<strong>the</strong>r restrict<br />

<strong>the</strong>ir domains.<br />

From such a description, a constraint solver can ideally<br />

compute solutions without additional instructions.<br />

II. CLP(FD)<br />

In connection with combinatorial optimisation or completion<br />

problems, one <strong>of</strong> <strong>the</strong> most frequently used instances<br />

<strong>of</strong> constraint programming is constraint logic<br />

programming over finite domains, denoted as CLP(FD).<br />

This means that all domains are finite sets <strong>of</strong> integers,<br />

and <strong>the</strong> available constraints include at least <strong>the</strong> common<br />

arithmetic relations between integer expressions.<br />

As an example <strong>of</strong> this approach, consider <strong>the</strong> wellknown<br />

combinatorial problem “Sudoku”, which has recently<br />

gained significant attention. Figure 1 shows an<br />

instance <strong>of</strong> this problem. To give a visual impression <strong>of</strong><br />

<strong>the</strong> values that a constraint solver can safely remove from<br />

81<br />

domains, we proceed as follows: First, we subdivide all<br />

free cells into 9 small regions, numbered in <strong>the</strong>ir natural<br />

order. Each region corresponds to a domain element, i.e.,<br />

an integer between 1 and 9. In Figure 1, a dot is drawn<br />

in those regions that correspond to domain elements that<br />

can be pruned due to <strong>the</strong> given constraints. A subsequent<br />

search can ignore <strong>the</strong>se elements.<br />

1<br />

3<br />

7 5<br />

2 7 4<br />

5 4<br />

9 6<br />

4 6<br />

7 1<br />

1 3<br />

Figure 1: Constraint propagation for a Sudoku puzzle<br />

III. GENERALISING CLP(FD)<br />

Traditionally, finite domain constraint solvers have<br />

mostly been used to tackle problems that involve only<br />

quite small values. This is <strong>the</strong> case in many scheduling,<br />

allocation and combinatorial optimisation tasks<br />

for which constraint-based approaches are very well<br />

suited. A well-known benchmark library for constraints,<br />

CSPLib [2], consists almost exclusively <strong>of</strong> such examples.<br />

On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> need for arbitrary precision integer<br />

arithmetic is widely recognised, and many common<br />

Prolog systems provide transparent built-in support for<br />

arbitrarily large integers.<br />

It thus seems natural to enhance a constraint solver<br />

over finite domains with <strong>the</strong> ability to reason over arbitrarily<br />

large integers. SICStus Prolog already goes in<br />

that direction, using <strong>the</strong> symbolic constants inf and sup<br />

to denote default domain limits, but internally, <strong>the</strong>y still<br />

correspond to quite small integers: The system yields a<br />

representation errors when <strong>the</strong>se limits are exceeded.


In this project, we implement a new constraint solver<br />

over finite domains, in which a s<strong>of</strong>tware implementation<br />

for big integers (“bignums”) is transparently used when<br />

arising values exceed machine-sized integers. We accept<br />

<strong>the</strong> inf/sup notation <strong>of</strong> SICStus Prolog, but <strong>the</strong>se atoms<br />

now denote <strong>the</strong> actual infinities instead <strong>of</strong> abbreviating<br />

underlying finite limits.<br />

As an example where this is useful, consider <strong>the</strong> socalled<br />

“7-11 problem” [3], which already surpasses <strong>the</strong><br />

limits <strong>of</strong> <strong>the</strong> finite domain constraint solver <strong>of</strong> SICStus<br />

Prolog on 32-bit systems. This non-linear problem is also<br />

beyond <strong>the</strong> abilities <strong>of</strong> common CLP(Q) solvers, but can<br />

be solved with our solver.<br />

Fur<strong>the</strong>r, a constraint solver without arbitrary limits can<br />

be freely advocated as a more general and more declarative<br />

alternative for (moded) built-in arithmetic. For example,<br />

consider <strong>the</strong> well known factorial function, implemented<br />

via finite domain constraints:<br />

factorial(0, 1).<br />

factorial(N, F) :-<br />

N #> 0,<br />

F #= F0*N,<br />

N1 #= N - 1,<br />

factorial(N1, F0).<br />

This relation can now be used in all directions. For<br />

example, given n, it can be used to compute f = n!,<br />

and also, given any f �= 0, it can compute all n for<br />

which n! =f:<br />

?- factorial(25, F).<br />

F = 15511210043330985984000000<br />

?- factorial(N, 120).<br />

N = 5<br />

The factorial function quickly exceeds <strong>the</strong> built-in limits<br />

<strong>of</strong> conventional constraint solvers, but not ours.<br />

IV. CURRENT RESULTS AND STATUS<br />

The current version <strong>of</strong> our constraint solver is included<br />

as library(clpfd) in <strong>the</strong> free Prolog system SWI-<br />

Prolog [4]. Partial descriptions <strong>of</strong> <strong>the</strong> system are already<br />

published in two workshop contributions ([5], [6]),<br />

and a fur<strong>the</strong>r paper focusing on termination properties<br />

was recently submitted [7]. The solver is already being<br />

used in a course on logic programming at <strong>the</strong> university<br />

<strong>of</strong> Aachen. O<strong>the</strong>r users include a group working<br />

on worst-case execution time analysis in a program analysis<br />

framework [8]. We work closely toge<strong>the</strong>r with <strong>the</strong>se<br />

groups and sometimes extend <strong>the</strong> solver with features<br />

that are especially useful for <strong>the</strong>ir applications. For example,<br />

we recently implemented a strategy that lets users<br />

quickly count <strong>the</strong> number <strong>of</strong> possible variable instantiations,<br />

without enumerating <strong>the</strong>m explicitly. This feature<br />

82<br />

is needed in execution time analysis, and is also valuable<br />

to quickly skip solutions that are not interesting to users.<br />

Importantly, constraint propagation in our solver always<br />

terminates. While this weakens propagation in<br />

cases with unbounded domains, it yields a simple termination<br />

criterion for constraint logic programs involving a<br />

costly search with <strong>the</strong> built-in predicate labeling/2:<br />

If posting <strong>the</strong> constraints alone terminates, <strong>the</strong>n this posting<br />

followed by labeling also terminates. Terminating<br />

propagation is also very valuable in automatically generated<br />

queries to test <strong>the</strong> solver.<br />

V. WORK IN PROGRESS<br />

Ongoing work on <strong>the</strong> constraint solver includes performance<br />

improvements and correctness considerations. In<br />

addition, we are constantly looking for new real-life tasks<br />

that can now be tackled with our constraint solver.<br />

REFERENCES<br />

[1] Pascal Van Hentenryck and Vijay Saraswat. Strategic<br />

directions in constraint programming. In ACM<br />

Workshop on Strategic Directions in Computing Research,<br />

1996.<br />

[2] Ian P. Gent and Toby Walsh. CSPLib: A benchmark<br />

library for constraints. In CP’99, volume 1713 <strong>of</strong><br />

LNCS, 1999.<br />

[3] Paul Pritchard and David Gries. The seven-eleven<br />

problem. Technical Report TR83-574, Cornell University,<br />

Computer Science Department, September<br />

1983.<br />

[4] Jan Wielemaker. An overview <strong>of</strong> <strong>the</strong> SWI-Prolog<br />

programming environment. In Proc. <strong>of</strong> <strong>the</strong> 13th Int.<br />

Workshop on Logic Prog. Environments, pages 1–16,<br />

December 2003.<br />

[5] Ulrich Neumerkel, Markus Triska, and Jan Wielemaker.<br />

Declarative language extensions for Prolog<br />

courses. In FDPE, <strong>2008</strong>.<br />

[6] Markus Triska, Ulrich Neumerkel, and Jan Wielemaker.<br />

A generalised finite domain constraint solver<br />

for SWI-Prolog. In WLP, <strong>2008</strong>.<br />

[7] Markus Triska, Ulrich Neumerkel, and Jan Wielemaker.<br />

Better termination for Prolog with constraints.<br />

In WLPE (submitted), <strong>2008</strong>.<br />

[8] Adrian Prantl, Jens Knoop, Markus Schordan, and<br />

Markus Triska. Constraint solving for high-level<br />

WCET analysis. In WLPE (submitted), <strong>2008</strong>.


Default Reasoning on Top <strong>of</strong> Ontologies with dl-Programs<br />

DAO Tran Minh and Thomas Eiter (Faculty Mentor)<br />

Institute for Information Systems, Vienna University <strong>of</strong> Technology, Austria<br />

Email: {dao,eiter}@kr.tuwien.ac.at<br />

Abstract — We study <strong>the</strong> usefulness <strong>of</strong> description logic programs<br />

(dl-programs) in implementing Reiter’s default logic on<br />

top <strong>of</strong> a description logic knowledge base (DL-KB). To this<br />

end, we investigate transformations from default <strong>the</strong>ories to dlprograms<br />

based on different established algorithms for computing<br />

default <strong>the</strong>ory extensions, namely <strong>the</strong> select-defaultsand-check<br />

and <strong>the</strong> select-justifications-and-check algorithms.<br />

In each transformation, additional constraints are exploited<br />

to prune <strong>the</strong> search space based on conclusion-conclusion or<br />

conclusion-justification relations. The implementation was deployed<br />

as a new component in <strong>the</strong> dl-plugin for dlvhex, and<br />

evaluated with various experimental test ontologies, which<br />

showed promising results.<br />

I. INTRODUCTION<br />

Default logic [1] has been one <strong>of</strong> <strong>the</strong> most prominent<br />

nonmonotonic reasoning formalisms due to its close relationship<br />

to common-sense reasoning. Ontologies are<br />

a key component <strong>of</strong> <strong>the</strong> future Semantic Web and reasoning<br />

on top also needs to handle default information.<br />

However, nonmonotonic reasoning with an ontology<br />

alone is not possible because description logics<br />

(DLs) which are used as ontology languages are monotonic.<br />

Therefore, incorporating default reasoning into<br />

ontologies is an interesting topic to take into account.<br />

The first attempt in this field was proposed in [2] which<br />

imposes some restrictions regarding decidability. Moreover,<br />

<strong>the</strong>re has not been any implementation or fur<strong>the</strong>r<br />

development on this approach since <strong>the</strong>n. One <strong>of</strong> <strong>the</strong> reasons<br />

was <strong>the</strong> lack <strong>of</strong> support from research on integrating<br />

rules and ontologies at that time [3].<br />

In this work, we investigate <strong>the</strong> problem <strong>of</strong> enabling<br />

default reasoning on top <strong>of</strong> ontologies specified by DL-<br />

KBs, based on an available mechanism that allows for<br />

integrating rules and ontologies, namely dl-programs [4].<br />

In addition to our <strong>the</strong>oretical results, we provide a frontend<br />

as a new component in <strong>the</strong> dl-plugin [5] for <strong>the</strong><br />

HEX-program solver dlvhex1 [6], which implements dlprograms.<br />

II. EMBEDDING DEFAULTS OVER DESCRIP-<br />

TION LOGICS INTO DL-PROGRAMS<br />

In [7], we present three transformations to embed default<br />

reasoning into dl-programs. Due to space constraints, we<br />

present here only one, viz. Υ, which we illustrate in a<br />

simple example ra<strong>the</strong>r than in full details.<br />

1 http://www.kr.tuwien.ac.at/research/systems/dlvhex/<br />

83<br />

Assume that we have a small DL-KB about birds and<br />

penguins: � �<br />

Flier ⊑¬NonFlier, Penguin ⊑ Bird,<br />

L =<br />

Penguin ⊑ NonFlier, Bird(t)<br />

and a set <strong>of</strong> defaults:<br />

� �<br />

Bird(X) :Flier(X)<br />

D =<br />

Flier(X)<br />

Intuitively, we have a DL-KB in which flying and nonflying<br />

objects are distinguished. We know that penguins,<br />

which are birds, do not fly. We also know that birds normally<br />

fly, but this fact cannot be added into L because it<br />

will make L inconsistent. Having a bird named Tweety, t<br />

for short, and no more information, we would like to conclude<br />

that t flies; and later if we learn that t is a penguin,<br />

<strong>the</strong> opposite conclusion should hold. To make this possible,<br />

we add a set <strong>of</strong> defaults D as about on top <strong>of</strong> L. We<br />

call Δ=〈L, D〉 a default <strong>the</strong>ory over a DL-KB. Through<br />

Υ, Δ is transformed to a dl-program KB df<br />

Υ (Δ) = (L, P )<br />

where P consists <strong>of</strong> <strong>the</strong> following dl-rules:<br />

cons_Flier(X) :dom(X),<br />

not out_cons_Flier(X).<br />

out_cons_Flier(X) :dom(X),<br />

not cons_Flier(X).<br />

in_Flier(X) :- cons_Flier(X),<br />

dom(X), DL[Flier+=in_Flier;Bird](X).<br />

:- cons_Flier(X),<br />

DL[Flier+=in_Flier;-Flier](X).<br />

:- not DL[Flier+=in_Flier;-Flier](X),<br />

out_cons_Flier(X).<br />

dom(t).<br />

These rules can be explained as follows: <strong>the</strong> first two<br />

guess if assuming an object can fly is consistent or not<br />

with <strong>the</strong> answer set, i.e., <strong>the</strong> agents’ belief set. The third<br />

rule aims to find which objects can fly under <strong>the</strong> consistency<br />

assumption above, where <strong>the</strong> atom DL[...] incorporates<br />

<strong>the</strong> assumption on Fliers. The last two rules<br />

are two constraints: <strong>the</strong> first prevents cases in which we<br />

guessed that an object can fly is consistent to <strong>the</strong> answer<br />

set but in fact we conclude that it does not fly. Similarly,<br />

<strong>the</strong> second constraint kills all answer sets in which we<br />

guessed that an object cannot fly is consistent with <strong>the</strong><br />

answer set but <strong>the</strong>n we could not conclude that it does<br />

not fly.<br />

With this dl-program, we have <strong>the</strong> desired answer set<br />

I = {dom(t), cons F lier(t), inF lier(t)} which en-


default rules<br />

[A(X);B(X)]<br />

/[C(X)]<br />

dl-rules<br />

HEX-rules<br />

User<br />

dl-rules<br />

C(X):-DL[ ;A](X),<br />

not DL[ ;-B](X).<br />

dfconverter<br />

ontology ontology<br />

dl-rules<br />

HEX-rules<br />

dlconverter<br />

HEXrules<br />

dlvhex<br />

Models/<br />

Extensions<br />

Figure 1: Strategy for implementing <strong>the</strong> df-converter<br />

tails that t flies by default. If we now change to a new<br />

DL-KB L ′ = L ∪{P enguin(t)}, <strong>the</strong>n from <strong>the</strong> dl-<br />

program KB df<br />

Υ (Δ′ )=(L ′ ,P), t can no longer be inferred<br />

to be able to fly, and <strong>the</strong> corresponding answer set<br />

is I ′ = {dom(t), out cons F lier(t)}.<br />

The o<strong>the</strong>r two transformations Π and Ω, <strong>the</strong> pro<strong>of</strong> <strong>of</strong><br />

correctness, and an optimizing technique using pruning<br />

rules are presented in [7].<br />

III. FRONT-END OVERVIEW<br />

The architecture <strong>of</strong> <strong>the</strong> front-end is shown in Figure 1.<br />

Users provide input, including a set <strong>of</strong> default rules along<br />

with a DL-KB in an OWL file, and <strong>the</strong>n get <strong>the</strong> extensions<br />

<strong>of</strong> <strong>the</strong> default <strong>the</strong>ory in terms <strong>of</strong> answer sets.<br />

O<strong>the</strong>r optional inputs can be ei<strong>the</strong>r dl-rules or low-level<br />

HEX-rules. Recall that dlvhex is a solver for HEXprograms.<br />

It receives input in terms <strong>of</strong> HEX-rules and<br />

returns results as answer sets to <strong>the</strong> user. As dlvhex has<br />

a dl-plugin, dl-programs are converted transparently into<br />

HEX-programs, <strong>the</strong> ordinary user is unaware <strong>of</strong> HEXrules.<br />

However, this does not prevent experts from providing<br />

more sophisticated input such as constraints in<br />

terms <strong>of</strong> dl-rules or HEX-rules to reduce <strong>the</strong> search space<br />

and speed up <strong>the</strong> evaluation. In fact, we do provide a<br />

technique in which <strong>the</strong> user can specify more supportive<br />

information, i.e., typing predicates that helps our implementation<br />

gaining significant performance improvement.<br />

Details on this technique can be found in [7].<br />

Utilizing <strong>the</strong> dl-plugin, we implement a df-converter<br />

whose input contains default rules accompanied with a<br />

DL-KB, optional dl-rules, or even HEX-rules. This converter<br />

transforms all <strong>the</strong> default rules into dl-rules based<br />

on different transformations, with <strong>the</strong> help <strong>of</strong> <strong>the</strong> DL-<br />

KB serving as <strong>the</strong> sources <strong>of</strong> individuals for <strong>the</strong> domain<br />

predicates to guarantee <strong>the</strong> safety condition [6].<br />

Then, <strong>the</strong> transformed dl-rules, along with o<strong>the</strong>r input dland<br />

HEX-rules, are transferred to <strong>the</strong> dl-converter. The<br />

rest <strong>of</strong> <strong>the</strong> evaluation will be done by <strong>the</strong> dl-plugin and<br />

dlvhex. For more details on <strong>the</strong> implementation <strong>of</strong> <strong>the</strong> dfconverter,<br />

we refer to <strong>the</strong> Master’s Thesis [7] on which<br />

this abstract is based. The <strong>the</strong>sis also contains experimental<br />

results, and comparisons between different trans-<br />

84<br />

formations, which reveals interesting tasks to improve<br />

<strong>the</strong> overall performance <strong>of</strong> <strong>the</strong> system.<br />

IV. FUTURE WORK<br />

Concerning our implementation <strong>of</strong> <strong>the</strong> front-end, some<br />

issues remain for future work. Firstly, we would like to<br />

investigate more sophisticated pruning rules depending<br />

on <strong>the</strong> structure <strong>of</strong> <strong>the</strong> default <strong>the</strong>ory. Secondly, a closer<br />

look into particular kinds <strong>of</strong> default <strong>the</strong>ories such as normal<br />

or semi-normal default should help to find more effective<br />

transformations.<br />

As our work depends on dlvhex and <strong>the</strong> dl-plugin, <strong>the</strong><br />

experimental results suggest <strong>the</strong> following tasks to increase<br />

<strong>the</strong> system performance.<br />

A caching technique which is only available now for<br />

dl-atoms would give additional benefit if it can be implemented<br />

for cq-programs [8], an extended version <strong>of</strong><br />

dl-programs, and already be deployed in <strong>the</strong> dl-plugin.<br />

dlvhex is now using RacerPro as its only DL-reasoner.<br />

It would also be interesting to look at o<strong>the</strong>r possibilities<br />

interfacing dlvhex with different DL-reasoners such as<br />

KAON2 or Pellet, and <strong>the</strong>n compare <strong>the</strong> results.<br />

Currently, dlvhex takes all grounded dl-programs and<br />

computes <strong>the</strong> answer sets. Ano<strong>the</strong>r challenging task<br />

would be to automatically classify <strong>the</strong> input and do only<br />

necessary rules grounding for a smaller search space.<br />

REFERENCES<br />

[1] R. Reiter. A logic for default reasoning. Artificial<br />

Intelligence, 13(1-2):81–132, 1980.<br />

[2] F. Baader and B. Hollunder. Embedding defaults<br />

into terminological knowledge representation formalisms.<br />

Autom. Reasoning, 14(1):149–180, 1995.<br />

[3] G. Antoniou et al., Combining Rules and<br />

Ontologies: A survey. Technical Report<br />

IST506779/Linköping/I3-D3/D/PU/a1, Linköping<br />

University, Feb 2005.<br />

[4] T. Eiter, G. Ianni, Th. Lukasiewicz, R. Schindlauer,<br />

and H. Tompits. Combining answer set programming<br />

with description logics for <strong>the</strong> semantic web.<br />

Artificial Intelligence, 172(12-13):1495–1539, <strong>2008</strong>.<br />

[5] T. Krennwallner. Integration <strong>of</strong> Conjunctive Queries<br />

over Description Logics into HEX-Programs. Master’s<br />

<strong>the</strong>sis, TU Wien, Oct 2007.<br />

[6] R. Schindlauer. Answer-Set Programming for <strong>the</strong> Semantic<br />

Web. PhD <strong>the</strong>sis, TU Wien, Dec 2006.<br />

[7] M. T. DAO. Default Reasoning on Top <strong>of</strong> Ontologies<br />

with dl-Programs. Master’s <strong>the</strong>sis, TU Wien, Jun<br />

<strong>2008</strong>.<br />

[8] T. Eiter, G. Ianni, T. Krennwallner, R. Schindlauer.<br />

Exploiting Conjunctive Queries in Description Logic<br />

Programs. Annals <strong>of</strong> Ma<strong>the</strong>matics and Artificial Intelligence.<br />

<strong>2008</strong>.


Implementation <strong>of</strong> a Design Tool for<br />

Automated Generation <strong>of</strong> Four State Logic Circuits<br />

Jakob Lechner and Martin Delvai (Faculty Mentor)<br />

Institute <strong>of</strong> Computer Engineering, Embedded Computing Systems Group<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {lechner,delvai}@ecs.tuwien.ac.at<br />

Abstract — Although asynchronous logic design is a long<br />

known research domain it plays a niche role in industry. One<br />

reason is certainly <strong>the</strong> lack <strong>of</strong> design tools that empower hardware<br />

designers to cope with <strong>the</strong> inherent complexity <strong>of</strong> asynchronous<br />

circuits. This short paper describes <strong>the</strong> development<br />

<strong>of</strong> a computer aided conversion process that allows designers<br />

to transform a given synchronous circuit description into an<br />

asynchronous Four State Logic design.<br />

I. INTRODUCTION<br />

In <strong>the</strong> past decades research <strong>of</strong> asynchronous circuitry<br />

has produced various design methodologies and concepts.<br />

One <strong>of</strong> <strong>the</strong>se approaches is called Four State Logic<br />

(FSL). This short paper gives a brief overview <strong>of</strong> a master<br />

<strong>the</strong>sis, which targets <strong>the</strong> development <strong>of</strong> a tool-supported<br />

design flow for FSL circuits. For this purpose <strong>the</strong> following<br />

tasks were performed:<br />

– Provide a sound <strong>the</strong>oretical background about <strong>the</strong><br />

FSL design paradigm focusing on initialisation issues.<br />

In contrast to synchronous designs, where all<br />

registers fire at <strong>the</strong> same point in time, in asynchronous<br />

circuits <strong>the</strong> firing is sequentially, controlled<br />

by (local) handshake signals. Thus, depending<br />

on <strong>the</strong> event sequence, <strong>the</strong> same circuit may<br />

produce different results. Due to <strong>the</strong> fact that <strong>the</strong><br />

initialisation strongly affects this event sequence an<br />

in-depth analysis <strong>of</strong> FSL circuits wrt. initialisation<br />

and <strong>the</strong> resulting behaviour was performed.<br />

– Development <strong>of</strong> a CAD tool which transforms a<br />

conventional synchronous circuit description into<br />

an FSL circuit. Some steps within this transformation<br />

are straightforward (e.g. <strong>the</strong> replacement <strong>of</strong> a<br />

conventional AND gate with an FSL AND gate).<br />

O<strong>the</strong>r tasks however require more attention such as<br />

<strong>the</strong> automated identification <strong>of</strong> source-sink pairs in<br />

complex circuit topologies and <strong>the</strong> introduction <strong>of</strong><br />

appropriate handshake signals.<br />

In this short paper <strong>the</strong> focus is placed on <strong>the</strong> design<br />

tool. In <strong>the</strong> first part <strong>the</strong> basic principles <strong>of</strong> FSL are introduced<br />

whereas <strong>the</strong> second part presents <strong>the</strong> design flow<br />

and <strong>the</strong> developed tool.<br />

85<br />

II. FOUR STATE LOGIC<br />

Four State Logic (FSL) is a representative <strong>of</strong> <strong>the</strong> delayinsensitive<br />

design methodology, where no timing assumptions<br />

have to be made (gate and wire delays can<br />

be unbounded). The key concept <strong>of</strong> FSL is <strong>the</strong> idea to<br />

extend regular boolean data values with temporal information.<br />

With this additional information it’s possible to<br />

determine if a certain data value is valid in <strong>the</strong> current<br />

context. Toge<strong>the</strong>r with acknowledge signals a “local”<br />

handshaking can be established between a data source<br />

and a data sink. Thus a global clock for coordinating<br />

switching actions <strong>of</strong> <strong>the</strong> circuit becomes obsolete.<br />

In FSL consecutive data words are encoded in alternate<br />

phases [1], called <strong>the</strong> even phase (ϕ0) and <strong>the</strong> odd<br />

phase (ϕ1). If all bits <strong>of</strong> an input vector have <strong>the</strong> same<br />

phase <strong>the</strong> input data is said to be consistent and can be<br />

considered valid. Thus, for discovering <strong>the</strong> arrival <strong>of</strong> <strong>the</strong><br />

next data word <strong>the</strong> data sink only needs to check if all<br />

bits <strong>of</strong> <strong>the</strong> input have changed <strong>the</strong>ir phase.<br />

Figure 1 demonstrates this principle. Assume signals<br />

A, B, C and D make up a data word. The figure shows <strong>the</strong><br />

signals’ phases over time as perceived by <strong>the</strong> recipient on<br />

its input ports. As you can see consistent intervals are<br />

interleaved with periods <strong>of</strong> time where <strong>the</strong> phase <strong>of</strong> one<br />

or more signals differs.<br />

Figure 1: Consistent and inconsistent phases.<br />

For representing <strong>the</strong> logical 0 (low) and logical 1<br />

(high) values in two phases a 2-bit encoding is needed:<br />

– ϕ0 : l =00,h=11<br />

– ϕ1 : L =01,H =10


Due to this encoding scheme it’s no longer possible to<br />

transmit one bit <strong>of</strong> information over a single wire. Hence,<br />

FSL needs to employ a “Double-Rail” logic: One single<br />

bit <strong>of</strong> information is sent over two rails.<br />

A. THE HANDSHAKE PRINCIPLE<br />

Registers are important components <strong>of</strong> a circuit because<br />

<strong>the</strong>y are synchronisation points. A register is allowed<br />

to latch new data if two conditions are satisfied: Obviously<br />

new data needs to be available from <strong>the</strong> preceding<br />

registers (upstream stages) and <strong>the</strong> currently stored data<br />

word has to be consumed by <strong>the</strong> successors (downstream<br />

stages). Both conditions can be validated by checking<br />

<strong>the</strong> phases <strong>of</strong> <strong>the</strong> predecessors and <strong>the</strong> successors. The<br />

phases <strong>of</strong> <strong>the</strong> upstream stages can be directly derived<br />

from <strong>the</strong> transmitted (FSL encoded) data, <strong>the</strong> phases <strong>of</strong><br />

<strong>the</strong> downstream stages are provided by additional feedback<br />

(or acknowledge) signals. Now a firing rule for registers<br />

can be formally defined:<br />

Definition 1 A register r is allowed to replace <strong>the</strong> currently<br />

stored value with its current input value if and only<br />

if <strong>the</strong> following two conditions are met:<br />

1. ∀w ∈ N + (r) :Φ(w) =Φ(r)<br />

2. ∀w ∈ N − (r) :Φ(w) �= Φ(r)<br />

N + (r) ... set <strong>of</strong> successor registers,<br />

N − (r) ... set <strong>of</strong> predecessor registers,<br />

Φ(r) ... current phase <strong>of</strong> register r<br />

On top <strong>of</strong> this concise model <strong>of</strong> <strong>the</strong> switching activities<br />

<strong>of</strong> FSL circuits <strong>the</strong> design flow including a tool for<br />

computer-aided conversion <strong>of</strong> synchronous circuits could<br />

be developed.<br />

III. TOOL FOR FSL CIRCUIT GENERATION<br />

The starting point <strong>of</strong> <strong>the</strong> conversion algorithm is a gatelevel<br />

netlist <strong>of</strong> <strong>the</strong> synchronous hardware circuit. This<br />

netlist can be produced with a properly configured syn<strong>the</strong>sis<br />

tool (e.g. Synopsys). Subsequently <strong>the</strong> netlist<br />

needs to be parsed so <strong>the</strong> circuit can <strong>the</strong>n be transformed<br />

into a directed graph, where <strong>the</strong> nodes represent <strong>the</strong> registers<br />

<strong>of</strong> <strong>the</strong> circuit. An edge pointing from one register<br />

v to ano<strong>the</strong>r register w represents a data flow from v to<br />

w. This graph representation <strong>of</strong> <strong>the</strong> circuit is used for all<br />

fur<strong>the</strong>r transformation steps and contains all necessary<br />

information for deciding about <strong>the</strong> structural extensions<br />

needed for building <strong>the</strong> actual FSL design.<br />

Since <strong>the</strong> phase relationship between two registers is<br />

<strong>the</strong> decisive factor ra<strong>the</strong>r than <strong>the</strong> actual phase <strong>of</strong> <strong>the</strong> individual<br />

register, <strong>the</strong> notion <strong>of</strong> tokens was introduced (see<br />

[2]). Thus <strong>the</strong> initialisation can be done be simply placing<br />

tokens at specific edges. A token placed on an edge<br />

between two registers indicates that <strong>the</strong> initial phases <strong>of</strong><br />

86<br />

<strong>the</strong> connected registers are different. Thus <strong>the</strong> token denotes<br />

that new data is available for <strong>the</strong> sink register instantly<br />

after a reset <strong>of</strong> <strong>the</strong> circuit.<br />

As mentioned in <strong>the</strong> introduction <strong>the</strong> initialisation has<br />

a crucial impact on <strong>the</strong> sequence in which registers fire.<br />

In order to investigate <strong>the</strong> circuit’s behaviour in presence<br />

<strong>of</strong> different initialisations a simulation tool was also developed.<br />

Figure 2 shows a screenshot <strong>of</strong> <strong>the</strong> simulator.<br />

A token (black dots) is available on both input edges <strong>of</strong><br />

register data2. Based on <strong>the</strong> firing rule this register is<br />

allowed to latch <strong>the</strong> new input. In <strong>the</strong> simulator <strong>the</strong> firing<br />

action will consume both input tokens and place a new<br />

token on <strong>the</strong> output edge (which <strong>the</strong>n indicates new data<br />

for register data3).<br />

Figure 2: FSL simulator<br />

After <strong>the</strong> designer has decided about a specific token<br />

assignment <strong>the</strong> asynchronous FSL design is completely<br />

defined and can be generated from <strong>the</strong> synchronous<br />

netlist in several steps: First standard logic signals are<br />

replaced by dual-rail encoded FSL signals and combinational<br />

logic gates are substituted for corresponding FSL<br />

gates. Subsequently synchronous registers are replaced<br />

by asynchronous ones and <strong>the</strong> insertion <strong>of</strong> <strong>the</strong> associated<br />

handshaking connections is performed. The chosen initialisation<br />

is applied and an FSL netlist is generated. This<br />

netlist can be used as input for conventional place & route<br />

tools, Xilinx ISE or Altera Quartus e.g., and finally <strong>the</strong><br />

asynchronous design can be downloaded into an FPGA.<br />

Concluding, <strong>the</strong> presented tool converts synchronous<br />

to asynchronous designs with a minimal user interaction<br />

required. The latter is restricted to <strong>the</strong> initialisation,<br />

which allows to tune <strong>the</strong> circuit to specific needs.<br />

REFERENCES<br />

[1] Anthony J. McAuley. Four state asynchronous<br />

architectures. IEEE Transactions on Computers,<br />

41(2):129–142, February 1992.<br />

[2] Daniel H. Linder and James C. Harden. Phased<br />

logic: Supporting <strong>the</strong> synchronous design paradigm<br />

with delay-insensitive circuitry. IEEE Transactions<br />

on Computers, 45(9):1031–1044, September 1996.


���������������������������������������������������������<br />

����������������<br />

�<br />

����������������������������������������������������<br />

��������������������������������<br />

����������������������������������������������������������������������<br />

����������������<br />

�����������������������������������������������<br />

�<br />

�<br />

����������������������������������������������<br />

������������� ������ ��� ��������� ��������������<br />

�������� ���� ������������ ���� ����������� ��� ������<br />

���������� ��������������������� ��� �������� ���������<br />

���� ����� �������� ��� ������� ��������� ��� ����������<br />

��������������������������������������������������<br />

����������������������������������������������������<br />

������������������������������������������������������<br />

����������������������������������������<br />

��� �������������<br />

���� ������� ������� ��� ���� ����� ������ ���� ����<br />

�����������������������������������������������������<br />

�������������������������������������������������������<br />

������������������������������������������������������<br />

�� ������� �������� ��� ����������� ��������� ��� ���<br />

����������������������������������������������������<br />

����������� ���� �������� ��� ����� ��������������� �����<br />

����������������������������������������������������<br />

����������������������������������������������<br />

������������������������������������������������<br />

������� ��� �������� ��� ����� ������������� ���� �������<br />

������� ���� ��� �� ������� ����� �������� ������� ����<br />

�����������������������������������������������������<br />

�������� �� ������� ����������� �������� ������������ ��<br />

������������������������������������������������������<br />

����� ����������� ��� ������ ������������ ���� ������ �����<br />

�����������������������������������������������������<br />

����� ���� ������� ������� ������� �� ������� ����� ���<br />

����������������������������������������������������<br />

��������� ����� ������ ����� ��������� ��� �������� ���<br />

���������������������������<br />

��� ������ ����� ����� �������� ����� ���� ����� ������<br />

��������� ������ �� ����� ������� ��� ��������� ����<br />

��������������������������������������������������������<br />

�����������������������������������������������������������<br />

�����������������<br />

��� ����������������������������<br />

��������� ���� ��������� ��� ������������ ���� ����<br />

�������������� ������������� ������ ��� ��� ������� ���<br />

�����������������������������������������������������������<br />

�������������������������������������������������������<br />

87<br />

���� �������� ��������� ���������� ���� ��������� ���<br />

����������������������������������������������������<br />

���������������������������������������������������������<br />

���������������������������������������������������������<br />

���������� ��� ���� ���������������� ��� ������� ��������<br />

����� ����������� ��� ���������� ��� ���� ������������ ���<br />

���������������������������������������������������<br />

����� ��� ������������� ���� ���������� ���� ������ ��� ����<br />

���������� ������������� ������� ��� ����������� ��������<br />

�����������������������������������������������������<br />

�������������������������������������������������������<br />

�������� �������� ��� ���� ����� ������ ��� ������ �������<br />

���������������������������������������������������������<br />

�������� ������������� ��� ������������� ���� ������ ���<br />

�������������������������������������������������������<br />

�������������������������������������������<br />

��� �������������������<br />

������������������������������������������������������<br />

����� ��������� ��������� ������� ��� ������ ���� ����� ���<br />

����������� ��� ���� ��� �������� ���� ��� ���� ��������� ������<br />

�������� ��� ���� ����� ��� ������� ���� ��� ���������� ����<br />

�������� ������ ��� ���������� ���� ����� �� �������������<br />

��������������������������������������������������<br />

������������������������������������������������������<br />

����� ������������� ��������� ���� �������������������<br />

����������������������������������������������������<br />

������������ ������� ��� ������ ��� ����������� ��� �������<br />

������ ����� ���������� �������� ������ ����� ������� ���<br />

�����������������������������������������������������<br />

������������������������������������������������������<br />

������������������� ���������� ���������� ��� ������ ����<br />

����������� �������� ���� ����� ����������� ��� ������<br />

��������� ������ �������� �������� ���������� ��������<br />

������������� ��� ������ ����������� ����� ��� ������� ����<br />

����������������������������������������������������<br />

���� ���������������������������������<br />

��������������<br />

���������������������������������������������������<br />

��������� ����� ����� ������ ������ ��������� �������������<br />

�������������������������������������������������������<br />

���������� �������� ����������� ����������� ���� ����� ���


���������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������������<br />

��� �����������������������������������<br />

�����������<br />

���������������������������������������������<br />

����������� �������� ����� ��� ����������� �����������<br />

�������������������������������������������������������<br />

����� ��� ��������� �������� �������� ���� ��� �����������<br />

����������������������������������������������������<br />

�����������������������������������������������<br />

����������������������������������������������������������<br />

������������������������������������������������������<br />

��������� ��� ������ ����� ����������� ������� �� ��� ����<br />

������� ��� �������������� ���� ������������ ���� �������<br />

������������������������������� �����������������������<br />

�����������������������������������������������������<br />

��������������������������������������������������<br />

����� ����� ���������� ��� ����� ��������� �������� ���<br />

������� �������� ��� ������ ��� ����� ������������ ������<br />

���������� ������� ������ ���������� ���� ������� ��� �����<br />

��������������������������������������������������<br />

���� ������� ����� ����� ���� ����� ������ ���� ������ �����<br />

�������������������������������������������������<br />

�����������������������������������������������������<br />

����� ������� ������ ��������� ���������� ������� ��������<br />

����� ����������� ���� ���� ��� ������������ �������� ����<br />

�����������������������������������������������������<br />

�������������������������������������������������������<br />

�������������������������������������������������������<br />

��� ��� ����� ��������� ��� ������� ���� ����� ������� �����<br />

���������� ���������� ��������� ��������� ����� ��� ����<br />

�����������������������������������<br />

��� ����������������������������������<br />

���� �������� ��� ��������� ���� ������� ��������� ���� ���<br />

���������������������������������������������������<br />

��������������������������������������������������������<br />

����������������������������������������������������<br />

�������������������������������������������������������<br />

�������������������������������������������������������<br />

��������� ������� ���� �������� ����������� �������������<br />

����� ���� ����������� ����� ��� ������� ������ ����������<br />

����� ��������� ������� ��� �������� ���� ����� ������������<br />

�����������������������������������������������������<br />

������� ������ ���� ����� ������ ��������� ������ ����<br />

������������ ������������ ���� ����������� ��� ������ ����<br />

�����������������������������������������������������<br />

����� ���������<br />

������������� ��� �������� ����� ��� ������������ �����<br />

����������������������������������������������������<br />

������ ��� �������� ����� ��������� ���� ����� ������<br />

����� ���������� ������ ����������� ����������� ������<br />

88<br />

�������������������������������������������������������<br />

����� ����� �������� ��� ���� ��������� ���� ��������� ������<br />

�����������������������������������������������������<br />

�����������������������������������������������<br />

�������������������������������������������������<br />

����������������������������������������������������<br />

��������� ����� ����������� ��������� ������ ����� ����� ���<br />

���������������������������������������������������<br />

�������������������������������������������������������<br />

����������������������������������������������������<br />

��������������������������������������������������������<br />

�����������������������������������<br />

���������������������������������������������<br />

����������� ���� ����� ������� ���������� ������<br />

�����������������������������������������������������<br />

�����������<br />

������� ���� ���� ��� ����� ������������ ��� ������� ���<br />

�����������������������������������������������<br />

�����������������������������������<br />

���������������� ������� ���� ���������� �����������<br />

������� ���������������� �������������������������<br />

�����������������������������������������������<br />

����������������������������������������������<br />

�������������������������<br />

����������������������������������������������������<br />

�������������� ����������� ������ ��������� ����<br />

���������������������������������������������<br />

���������������������������������������������������<br />

���������������������������������������������<br />

������� �������������� ��������� ������������ ����<br />

��������������������������������������������<br />

����������<br />

��������������������������������������������������<br />

����� ��������� ������ ������� ������� ��������<br />

����� ������ ����������������� ������ ���������<br />

�������������������������<br />

����������������������������������������������<br />

������� ���� ��������� ������������� ��� �������<br />

�����������������������������������������������<br />

����������������������<br />

������������������ �������������� ��������� ����������<br />

������������ ���� ��� �������� �������������� �����<br />

���� ���� ��������� ����������� ��� ��������� ����<br />

��������������������������������������������<br />

��������� ��� ���� ���������� ������ ���������<br />

�������������������<br />

������� ������� �� ������� ������ ���� ������� �����<br />

��������������������������������������<br />

�����������������������������������������������������<br />

��������� ��������� ��� ������� ���������� �� ���<br />

������ �����������������������������������������<br />

�������������������������������������<br />


Efficient Counting with Bounded Treewidth using Datalog ∗<br />

Stefan Rümmele and Reinhard Pichler (Faculty Mentor)<br />

Database and Artificial Intelligence Group<br />

Technische Universität Wien<br />

Vienna, Austria<br />

Email: {ruemmele,pichler}@dbai.tuwien.ac.at<br />

Abstract — Bounded treewidth has proven to be a key concept<br />

in identifying tractable fragments <strong>of</strong> inherently intractable<br />

problems. An important result in this context is Courcelle’s<br />

Theorem, stating that any property <strong>of</strong> finite structures definable<br />

in monadic second-order logic (MSO), becomes tractable if <strong>the</strong><br />

treewidth <strong>of</strong> <strong>the</strong> structure is bounded by a constant. An extension<br />

<strong>of</strong> this result to counting problems was done by Arnborg<br />

et al. But both pro<strong>of</strong>s did not yield an implementable algorithm.<br />

Recently Gottlob et al. presented a new approach using<br />

monadic datalog to close this gap for decision problems. In<br />

this paper, we extend this method in order to handle counting<br />

problems as well and apply it to <strong>the</strong> problems #SAT,#CIRCUM-<br />

SCRIPTION and #HORN-ABDUCTION.<br />

I. INTRODUCTION<br />

Many intractable problems can be solved efficiently if<br />

some parameter is bounded by a constant. Downey<br />

and Fellows [1] introduced <strong>the</strong> framework <strong>of</strong> parameterized<br />

complexity to deal with this phenomenon. One<br />

<strong>of</strong> <strong>the</strong> most successful concepts for such a parameter is<br />

bounded treewidth, where <strong>the</strong> treewidth (a measure <strong>of</strong><br />

similarity to a tree) <strong>of</strong> <strong>the</strong> underlying structure is bounded<br />

by a constant. Although <strong>the</strong> systematic study <strong>of</strong> parameterized<br />

complexity basically started with Downey<br />

and Fellows, many ad-hoc results especially related to<br />

fixed-parameter tractability (FPT) have been presented<br />

before. One <strong>of</strong> <strong>the</strong>m is Courcelle’s Theorem [2], stating<br />

that any property <strong>of</strong> finite structures which is expressible<br />

in monadic second-order logic (MSO) can be decided<br />

in linear time, if we restrict ourselves to structures<br />

with bounded treewidth. Previous methods for deriving<br />

a concrete algorithm from <strong>the</strong> MSO description based<br />

on Courcelle’s Theorem [3, 4] translate <strong>the</strong> MSO evaluation<br />

problem over finite structures into an evaluation<br />

problem over colored binary trees, which is <strong>the</strong>n solved<br />

via a finite tree automaton (FTA). Although <strong>the</strong> <strong>the</strong>oretical<br />

value <strong>of</strong> <strong>the</strong>se approaches is indisputable, it turned<br />

out that in practice <strong>the</strong>y lead to a “state explosion” <strong>of</strong> <strong>the</strong><br />

FTA even for very simple MSO formulae (cf. [5]). Therefore,<br />

it was stated in [6] that such algorithms are “useless<br />

for practical applications” and that <strong>the</strong> main benefit <strong>of</strong><br />

Courcelle’s Theorem is in providing “a simple way to<br />

recognize a property as being linear time computable”.<br />

∗This work was supported by <strong>the</strong> Austrian Science Fund (FWF),<br />

project P20704-N18.<br />

89<br />

Recently Gottlob et. al. [7] have proposed an alternative<br />

approach for deriving linear time algorithms from a<br />

MSO property via monadic datalog. In particular <strong>the</strong>y<br />

have shown <strong>the</strong> following <strong>the</strong>orem:<br />

Theorem 1 (Gottlob et al. [7]) Given an arbitrary but<br />

fixed (relational) vocabulary τ and a constant w ≥ 1.<br />

Then every MSO-definable unary query over τ-structures<br />

<strong>of</strong> treewidth w is also definable in <strong>the</strong> quasi-guarded<br />

fragment <strong>of</strong> monadic datalog over τ extended by predicates<br />

for expressing tree decompositions.<br />

Note that <strong>the</strong>ir definition <strong>of</strong> quasi-guarded programs require<br />

that every rule contains an extensional atom called<br />

guard, s.t. every variable in <strong>the</strong> rule is functionally dependent<br />

on <strong>the</strong> guards variables. Additionally <strong>the</strong>y showed<br />

that such a program can be evaluated in linear time.<br />

Our work includes <strong>the</strong> extension <strong>of</strong> Theorem 1 to also<br />

account for counting problems instead <strong>of</strong> handling decision<br />

problems and unary queries only. The formal<br />

framework <strong>of</strong> counting complexity was introduced by<br />

Valiant [8, 9] in order to deal with counting problems,<br />

where one is not only interested in whe<strong>the</strong>r a property<br />

is true or false but instead one wants to count how <strong>of</strong>ten<br />

this property is fulfilled. The first formal integration <strong>of</strong><br />

counting complexity into parameterized complexity <strong>the</strong>ory<br />

was done independently by McCartin [10] and Flum<br />

and Grohe [11]. But again many ad-hoc results related<br />

to fixed-parameter tractability <strong>of</strong> counting problems have<br />

been presented before. One <strong>of</strong> <strong>the</strong>m is a result by Arnborg<br />

et al. [3] extending Courcelle’s Theorem to counting<br />

problems, which are definable via MSO properties.<br />

They showed that each <strong>of</strong> <strong>the</strong>se problems can be solved<br />

in linear time if restricted to problem instances, where<br />

<strong>the</strong> treewidth <strong>of</strong> <strong>the</strong> underlying structure is bounded by<br />

some constant.<br />

II. RESULTS<br />

Of course, <strong>the</strong> statements about implementability <strong>of</strong> <strong>the</strong><br />

algorithms based on Courcelle’s Theorem, which are derived<br />

via colored binary trees and FTA also hold for <strong>the</strong><br />

extension to counting problems. Therefore we adopted<br />

<strong>the</strong> idea <strong>of</strong> [7] and showed:<br />

Theorem 2 Given an arbitrary but fixed (relational) vocabulary<br />

τ and a constant w ≥ 1. Then every counting<br />

problem over τ-structures <strong>of</strong> treewidth w defined via


MSO properties is also definable in <strong>the</strong> quasi-guarded<br />

fragment <strong>of</strong> datalog over τ extended by predicates for<br />

expressing tree decompositions and a summation operator.<br />

The pro<strong>of</strong> was done utilizing a technique from finite<br />

model <strong>the</strong>ory, namely Ehrenfeucht-Fraïssé games for<br />

monadic second-order logic. With <strong>the</strong> help <strong>of</strong> <strong>the</strong>se<br />

games, we could assign a so called type to each node<br />

in <strong>the</strong> tree decomposition, which only depends on <strong>the</strong><br />

types <strong>of</strong> its child nodes. Fur<strong>the</strong>rmore <strong>the</strong> type <strong>of</strong> <strong>the</strong> root<br />

node is all we need to compute <strong>the</strong> result <strong>of</strong> our counting<br />

problem. Therefore we could establish <strong>the</strong> <strong>the</strong>orem<br />

by structural induction over <strong>the</strong> tree decomposition. We<br />

also showed that <strong>the</strong> evaluation <strong>of</strong> a datalog program can<br />

be done in linear time if restricted to <strong>the</strong> quasi-guarded<br />

fragment <strong>of</strong> datalog extended by a summation operator<br />

and assuming unit cost for arithmetic operations. Note<br />

that from this result and from Theorem 2 we can actually<br />

derive Courcelle’s Theorem for counting problems as a<br />

corollary.<br />

To put our <strong>the</strong>orem to work, we applied this datalog<br />

approach to solve <strong>the</strong> three counting problems #SAT,<br />

#CIRCUMSCRIPTION and #HORN-ABDUCTION. The<br />

first one (#SAT), is <strong>the</strong> problem <strong>of</strong> counting all models <strong>of</strong><br />

a propositional formula whereas #CIRCUMSCRIPTION is<br />

<strong>the</strong> problem <strong>of</strong> counting only <strong>the</strong> (subset) minimal models.<br />

#HORN-ABDUCTION is <strong>the</strong> problem <strong>of</strong> counting <strong>the</strong><br />

solutions <strong>of</strong> a propositional abduction problem where <strong>the</strong><br />

underlying <strong>the</strong>ory is restricted to Horn clauses. Note that<br />

<strong>the</strong> #P-completeness <strong>of</strong> #SAT was already mentioned in<br />

[9], whereas <strong>the</strong> #P-completeness <strong>of</strong> #HORN-ABDUC-<br />

TION has been shown recently in [12]. #CIRCUMSCRIP-<br />

TION is #NP-complete and <strong>the</strong>refore lies on <strong>the</strong> second<br />

level <strong>of</strong> <strong>the</strong> counting hierarchy [13].<br />

III. RELATED WORK<br />

Part <strong>of</strong> this work will appear in [14], which contains also<br />

a discussion on a pro<strong>of</strong>-<strong>of</strong>-concept implementation <strong>of</strong> <strong>the</strong><br />

algorithm for #SAT. O<strong>the</strong>r related publications include<br />

[15], where <strong>the</strong> counting version <strong>of</strong> Courcelle’s Theorem<br />

is extended to graphs with bounded clique-width (a measure<br />

<strong>of</strong> similarity to a clique). In [16] an algorithm for<br />

solving #SAT and #GENSAT (a generalization <strong>of</strong> #SAT)<br />

in case <strong>of</strong> bounded treewidth as well as bounded cliquewidth.<br />

Fur<strong>the</strong>rmore <strong>the</strong>y provided a sketch on how to<br />

apply <strong>the</strong>ir approach based on recursive splitting to o<strong>the</strong>r<br />

#P-complete problems. An algorithm for #SAT based<br />

on dynamic programming was presented in [17].<br />

REFERENCES<br />

[1] R.G. Downey and M.R. Fellows. Parameterized<br />

Complexity. Springer, New York, 1999.<br />

90<br />

[2] B. Courcelle. Graph rewriting: An algebraic and<br />

logic approach. In Handbook <strong>of</strong> Theoretical Computer<br />

Science, Volume B, pages 193–242. Elsevier<br />

and MIT Press, 1990.<br />

[3] S. Arnborg, J. Lagergren, and D. Seese. Easy problems<br />

for tree-decomposable graphs. J. Algorithms,<br />

12(2):308–340, 1991.<br />

[4] J. Flum, M. Frick, and M. Grohe. Query evaluation<br />

via tree-decompositions. J. ACM, 49(6):716–752,<br />

2002.<br />

[5] M. Frick and M. Grohe. The complexity <strong>of</strong> firstorder<br />

and monadic second-order logic revisited. In<br />

Proc. LICS’02, pages 215–224. IEEE Computer<br />

Society, 2002.<br />

[6] M. Grohe. Descriptive and parameterized complexity.<br />

In Proc. CSL’99, volume 1683 <strong>of</strong> Lecture Notes<br />

in Computer Science, pages 14–31. Springer, 1999.<br />

[7] G. Gottlob, R. Pichler, and F. Wei. Monadic datalog<br />

over finite structures with bounded treewidth. In<br />

Proc. PODS’07, pages 165–174. ACM, 2007.<br />

[8] L.G. Valiant. The complexity <strong>of</strong> computing <strong>the</strong> permanent.<br />

Theor. Comput. Sci., 8:189–201, 1979.<br />

[9] L.G. Valiant. The complexity <strong>of</strong> enumeration and<br />

reliability problems. SIAM J. Comput., 8(3):410–<br />

421, 1979.<br />

[10] C. McCartin. Parameterized counting problems. In<br />

Proc. MFCS’02, volume 2420 <strong>of</strong> Lecture Notes in<br />

Computer Science, pages 556–567. Springer, 2002.<br />

[11] J. Flum and M. Grohe. The parameterized complexity<br />

<strong>of</strong> counting problems. SIAM J. Comput.,<br />

33(4):892–922, 2004.<br />

[12] M. Hermann and R. Pichler. Counting complexity<br />

<strong>of</strong> propositional abduction. In Proc. IJCAI’07,<br />

pages 417–422, 2007.<br />

[13] Arnaud Durand, Miki Hermann, and Phokion G.<br />

Kolaitis. Subtractive reductions and complete problems<br />

for counting complexity classes. Theor. Comput.<br />

Sci., 340(3):496–513, 2005.<br />

[14] M. Jakl, R. Pichler, S. Rümmele, and S. Woltran.<br />

Fast counting with bounded treewidth. In Proc.<br />

LPAR’08 (to appear).<br />

[15] B. Courcelle, J.A. Makowsky, and U. Rotics. On<br />

<strong>the</strong> fixed parameter complexity <strong>of</strong> graph enumeration<br />

problems definable in monadic second-order<br />

logic. Discrete Applied Ma<strong>the</strong>matics, 108(1–2):23–<br />

52, 2001.<br />

[16] E. Fischer, J.A. Makowsky, and E.V. Ravve. Counting<br />

truth assignments <strong>of</strong> formulas <strong>of</strong> bounded treewidth<br />

or clique-width. Discrete Applied Ma<strong>the</strong>matics,<br />

156(4):511–529, <strong>2008</strong>.<br />

[17] M. Samer and S. Szeider. Algorithms for propositional<br />

model counting. In Proc. LPAR’07, volume<br />

4790 <strong>of</strong> LNCS, pages 484–498. Springer, 2007.


MapFace - A Graphical Editor to Support <strong>the</strong><br />

Semantic Annotation <strong>of</strong> Medical Text<br />

Theresia Gschwandtner, Katharina Kaiser, and Silvia Miksch (Faculty Mentor)<br />

Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {gschwandtner,kaiser,miksch}@ifs.tuwien.ac.at<br />

Abstract — The mapping <strong>of</strong> medical texts to concepts <strong>of</strong> medical<br />

terminology systems is a prerequisite for many tasks <strong>of</strong> automatically<br />

processing <strong>the</strong>se documents. Due to <strong>the</strong> complex<br />

nature <strong>of</strong> this task, <strong>the</strong> results <strong>of</strong> mapping systems still contain<br />

erroneous bits <strong>of</strong> information.<br />

Our editor visualizes <strong>the</strong> annotation <strong>of</strong> <strong>the</strong> text and provides<br />

means to easily navigate and modify it. Thus we are able to<br />

shorten a cumbersome and time-consuming task and subsequently<br />

provide reliable and well-defined information for fur<strong>the</strong>r<br />

processing steps. Even more, <strong>the</strong> visualization features<br />

support a better understanding <strong>of</strong> <strong>the</strong> medical texts.<br />

I. INTRODUCTION<br />

Medical documents contain important information, but<br />

<strong>the</strong>y are <strong>of</strong>ten difficult to understand. Mapping concepts<br />

from medical terminology systems to <strong>the</strong> text supports<br />

<strong>the</strong> intelligibility and disambiguation <strong>of</strong> <strong>the</strong> text by<br />

adding useful metainformation. To this end, programs<br />

such as <strong>the</strong> MetaMap Transfer (MMTx) program [1] create<br />

a rough mapping <strong>of</strong> concepts from <strong>the</strong> Unified Medical<br />

Language System (UMLS) Meta<strong>the</strong>saurus [2] to free<br />

medical text. Due to <strong>the</strong> complexity <strong>of</strong> unstructured text,<br />

it is not always possible to automatically create a correct<br />

mapping. However, <strong>the</strong> complete reliability <strong>of</strong> information<br />

is crucial in medical care - an extremely sensitive<br />

discipline - which makes it necessary for medical experts<br />

to control and correct <strong>the</strong>se results. This led us to meet<br />

<strong>the</strong> following challenges:<br />

– Enabling medical experts without special programming<br />

skills to handle <strong>the</strong> MMTx program.<br />

– Providing means to assure <strong>the</strong> correct affiliation <strong>of</strong><br />

UMLS concepts to text chunks.<br />

– Supporting <strong>the</strong> understanding <strong>of</strong> medical concepts<br />

in <strong>the</strong> text as well as relations between <strong>the</strong>m.<br />

II. THE MAPFACE EDITOR<br />

A variety <strong>of</strong> systems exist annotating text with medical<br />

concepts. In contrast to most o<strong>the</strong>r systems, MMTx not<br />

only supports exact matches between a text token and<br />

a UMLS concept, but also considers term variants as<br />

well as partial matches. Additionally, it computes matchscores<br />

<strong>of</strong> found candidates by combining specific mea-<br />

91<br />

sured values. For a detailed description <strong>of</strong> annotation<br />

systems we refer to [3].<br />

The MMTx program automatically tokenizes <strong>the</strong> text<br />

into phrase chunks and concept chunks; additionally, it<br />

maps UMLS concepts to <strong>the</strong> text (see Figure 1).<br />

Since manually correcting <strong>the</strong>se results would be an<br />

extremely cumbersome and time-consuming task, and as<br />

<strong>the</strong>re has been no satisfactory tool to support it, we have<br />

developed <strong>the</strong> MapFace editor (see Figure 2). By means<br />

<strong>of</strong> this editor we are able to<br />

1. facilitate <strong>the</strong> handling <strong>of</strong> <strong>the</strong> MMTx program, by<br />

– providing a GUI, and<br />

– making <strong>the</strong> MMTx results easily navigable.<br />

2. assure a correct mapping <strong>of</strong> <strong>the</strong> text to UMLS concepts,<br />

by providing means to<br />

– correct <strong>the</strong> tokenization <strong>of</strong> <strong>the</strong> text into concept<br />

chunks and phrase chunks (see Figure 3), and<br />

to<br />

– edit <strong>the</strong> affiliation <strong>of</strong> UMLS concepts to <strong>the</strong>se<br />

concept chunks (see Figure 3).<br />

3. support <strong>the</strong> better understanding <strong>of</strong> <strong>the</strong> text, by<br />

– color-coding <strong>of</strong> semantic types [4],<br />

– highlighting concepts or phrases accordingly,<br />

and<br />

– listing and highlighting relations among medical<br />

concepts in <strong>the</strong> text.<br />

III. EVALUATION<br />

In order to ensure <strong>the</strong> usability <strong>of</strong> <strong>the</strong> MapFace editor we<br />

conducted an evaluation study. In doing so, we went for a<br />

heuristic approach according to Nielsen and Molich [5].<br />

Four evaluators were asked to solve typical tasks with <strong>the</strong><br />

help <strong>of</strong> <strong>the</strong> MapFace editor. They noted and rated each<br />

usability problem <strong>the</strong>y encountered.<br />

The study discovered 32 usability problems, <strong>of</strong> which<br />

78,12% could be fixed immediately. The o<strong>the</strong>r 21.88%<br />

ei<strong>the</strong>r take more effort to fix or <strong>the</strong>y require more general<br />

changes in <strong>the</strong> design. In any case, we will consider all<br />

problems carefully in order to improve <strong>the</strong> quality <strong>of</strong> <strong>the</strong><br />

editor.


Figure 1: The MMTx program tokenizes <strong>the</strong> sentence<br />

into phrase chunks. A set <strong>of</strong> best matching UMLS concepts<br />

toge<strong>the</strong>r with <strong>the</strong>ir associated semantic types (in<br />

italics) are returned for each concept identified within<br />

<strong>the</strong> text. The encircled objects are wrong or ambiguous<br />

results which have to be corrected by means <strong>of</strong> <strong>the</strong> Map-<br />

Face editor.<br />

Figure 2: For a given concept chunk a list <strong>of</strong> best matching<br />

UMLS concepts is displayed in <strong>the</strong> bottom pane. In<br />

<strong>the</strong> right-hand pane <strong>the</strong> user can select color-coded semantic<br />

types to highlight associated text chunks.<br />

IV. CONCLUSION<br />

The MapFace editor provides important means to navigate,<br />

edit, and visualize <strong>the</strong> semantic annotation <strong>of</strong> medical<br />

documents automatically generated by <strong>the</strong> MMTx<br />

program. Enabling medical experts to easily control and<br />

correct <strong>the</strong> semantic annotation assures <strong>the</strong> quality <strong>of</strong> <strong>the</strong><br />

outcome, which in turn improves <strong>the</strong> validity <strong>of</strong> any subsequent<br />

processing step.<br />

With respect to <strong>the</strong> outcome <strong>of</strong> <strong>the</strong> usability evaluation<br />

we will ensure that <strong>the</strong> MapFace editor is not only an<br />

92<br />

Figure 3: Corrections accomplished by means <strong>of</strong> <strong>the</strong><br />

MapFace editor are encircled (compare Figure 1).<br />

important and time-saving means, but also a convenient<br />

tool to work with.<br />

ACKNOWLEDGMENTS<br />

The research leading to <strong>the</strong>se results has received funding<br />

from Fonds zur Förderung der wissenschaftlichen<br />

Forschung FWF (Austrian Science Fund), grant L290-<br />

N04 and from <strong>the</strong> European Community’s Seventh<br />

Framework Programme (FP7/2007-2013) under grant<br />

agreement n ◦ 216134.<br />

REFERENCES<br />

[1] A. R. Aronson. Effective mapping <strong>of</strong> biomedical text<br />

to <strong>the</strong> UMLS Meta<strong>the</strong>saurus: <strong>the</strong> MetaMap program.<br />

In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 25th Annual American Medical<br />

Informatics Association Symposium, pages 17–<br />

21, Washington, D.C., November 2001.<br />

[2] P. L. Schyler, W. T. Hole, M. S. Tuttle, and D. D.<br />

Sherertz. The UMLS Meta<strong>the</strong>saurus: representing<br />

different views <strong>of</strong> biomedical concepts. Bulletin<br />

<strong>of</strong> <strong>the</strong> Medical Library Association, 81(2):217–222,<br />

April 1993.<br />

[3] L. H. Reeve. Semantic Annotation and Summarization<br />

<strong>of</strong> Biomedical Text. PhD Thesis. Drexel University,<br />

College <strong>of</strong> Information Science and Technology,<br />

Philadelphia, Pennsylvania, 2007.<br />

[4] A. T. McCray and S. J. Nelson. The representation<br />

<strong>of</strong> meaning in <strong>the</strong> UMLS. Methods <strong>of</strong> Information in<br />

Medicine, 34(1-2):193–201, March 1995.<br />

[5] J. Nielsen and R. Molich. Heuristic evaluation <strong>of</strong><br />

user interfaces. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> ACM <strong>Conference</strong><br />

on Human Factors in Computing Systems,<br />

pages 249–256, Seattle, Washington, April 1990.


Optimization <strong>of</strong> communication by analyzing <strong>the</strong> interlocutor’s<br />

wording<br />

Gudrun Kellner<br />

E-Commerce Group, Institute <strong>of</strong> S<strong>of</strong>tware Technology and Interactive Systems<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: kellner@ec.tuwien.ac.at<br />

Abstract — Recent research results on communication<br />

science mention a relation between <strong>the</strong> interlocutor’s<br />

perceptional preferences, his or her mental<br />

representations and his or her choice <strong>of</strong> expressions.<br />

However, <strong>the</strong>re is not any tool for analyzing<br />

such.<br />

In my research I combined lexical with computational<br />

linguistic methods to develop a s<strong>of</strong>tware<br />

prototype that is able to analyze text on <strong>the</strong> usage <strong>of</strong><br />

perceptional expressions. This analyzing tool can<br />

help to identify <strong>the</strong> interlocutor’s perceptional preference<br />

for easier meeting his or her way <strong>of</strong> thinking<br />

and <strong>the</strong>reby facilitate <strong>the</strong> understanding.<br />

I. INTRODUCTION<br />

All communication is based on perception. With <strong>the</strong><br />

establishment <strong>of</strong> constructivism <strong>the</strong> knowledge about<br />

individual differences in perception spread. It seems<br />

reasonable to assume that <strong>the</strong> individual form <strong>of</strong><br />

perception also influences <strong>the</strong> language <strong>of</strong> a person.<br />

Interestingly, this <strong>the</strong>ory hasn’t yet shown great affects<br />

in <strong>the</strong> research on individual and personal language<br />

use.<br />

The Neurolinguistic Programming (NLP) <strong>of</strong>fers a<br />

model that seems as a first step in closing this gap:<br />

The idea <strong>of</strong> sensual-specific perception which comes<br />

noticeable in <strong>the</strong> individual language use is described<br />

by <strong>the</strong> keyword “representational systems”. According<br />

to this, every person has one or two sensory systems<br />

preferably used to perceive his or her environment.<br />

From <strong>the</strong> five senses arises <strong>the</strong> classification in<br />

visual, auditory, kines<strong>the</strong>tic, olfactory and gustatory<br />

perception modalities. Each <strong>of</strong> <strong>the</strong>se perception<br />

classes forms a sensomotor complex, in <strong>the</strong> NLP<br />

called representational system. As <strong>the</strong> inner representations<br />

are accessed for expressing thoughts, <strong>the</strong><br />

preferred sensory system <strong>of</strong> a speaker can be told by<br />

his or her diction and wording.<br />

The research about <strong>the</strong> optimization <strong>of</strong> communication<br />

furnished pro<strong>of</strong> that <strong>the</strong> percentage <strong>of</strong> understanding<br />

is higher when a speaker adjusts his diction<br />

to <strong>the</strong> diction <strong>of</strong> his addressee – what applies with a<br />

high probability also to <strong>the</strong> use <strong>of</strong> perceptional<br />

words.<br />

93<br />

Still, for <strong>the</strong> German language, not much research<br />

on <strong>the</strong> relation between an individual’s perceptional<br />

preference and his or her language use has been done<br />

yet. To be able to go deeper into that idea, (1) a lexical<br />

corpus <strong>of</strong> perceptional expressions and (2) a<br />

s<strong>of</strong>tware tool that automatically filters those expressions<br />

from a text are needed. Those two steps shall<br />

be explained in this abstract.<br />

II. A LEXICAL CORPUS OF PERCEPTIONAL<br />

EXPRESSIONS<br />

For a systematic analysis <strong>of</strong> <strong>the</strong> use <strong>of</strong> perceptional<br />

expressions, a corpus is needed, containing a collection<br />

<strong>of</strong> <strong>the</strong> most <strong>of</strong>ten used words and expressions<br />

that can be directly linked to one <strong>of</strong> <strong>the</strong> perceptional<br />

classes mentioned above. That corpus was built with<br />

<strong>the</strong> input <strong>of</strong> 13 different resources and contains more<br />

than 840 different entries with at least 200 words and<br />

expressions per perceptional class. For fur<strong>the</strong>r details<br />

and <strong>the</strong> complete corpus see [1].<br />

During <strong>the</strong> examination <strong>of</strong> <strong>the</strong> corpus it was found<br />

that usually only <strong>the</strong> root word is decisive for <strong>the</strong><br />

assignment to a representational class, but that such a<br />

root word can occur in different combinations. This<br />

follows <strong>the</strong> fact that in <strong>the</strong> German language it is<br />

possible to combine words and <strong>the</strong>reby create new<br />

ones. That possibility causes a problem when searching<br />

for a semantic unit that can also be part <strong>of</strong> a compound<br />

word. For example, when searching for <strong>the</strong><br />

signal word “rot” [red], <strong>the</strong> output should also include<br />

words like “Morgenrot” [dawn] and “rotgrün”<br />

[red and green] containing <strong>the</strong> same semantic root,<br />

but not words like “Brot” [bread] or “Schrot”<br />

[bruised grain] which include <strong>the</strong> same sequence <strong>of</strong><br />

characters, but have no semantic link to <strong>the</strong> searched<br />

word. A search algorithm that also takes compound<br />

words into account would especially be helpful when<br />

scanning a German text for a list <strong>of</strong> signal words.<br />

Therefore, a search algorithm using a new model<br />

describing word composition was developed.


III. THE ANALYZING TOOL<br />

As a consequence <strong>of</strong> <strong>the</strong> multiple combining possibilities<br />

<strong>of</strong> root words, an appropriate analyzing tool<br />

should not only be able to find <strong>the</strong> expressions contained<br />

by <strong>the</strong> corpus, but also all possible combinations<br />

that are based on <strong>the</strong> same semantic root.<br />

To solve this problem, a model <strong>of</strong> possible word<br />

structures was developed. Following that model, it<br />

can be determined which letters or syllables may<br />

occur around a root word without changing its semantic<br />

meaning (and hence <strong>the</strong> assignment to <strong>the</strong><br />

respective perceptional class). This model was introduced<br />

with <strong>the</strong> aim to provide a clear description <strong>of</strong><br />

word composition transferable into an algorithm and,<br />

based on <strong>the</strong> root word, describes <strong>the</strong> structure <strong>of</strong><br />

compound words as follows:<br />

[ [prefix] + root word + [infix] + (word joint) ]<br />

[prefix] + root word + [infix] + (ending)<br />

The following notation was used: Round brackets<br />

mark optional, but maximal single occurrence, square<br />

brackets highlight optional (even multiple) occurrence<br />

<strong>of</strong> elements.<br />

Based on <strong>the</strong> word components mentioned in <strong>the</strong><br />

Duden Grammatik [2], for each element <strong>of</strong> <strong>the</strong> model<br />

a list <strong>of</strong> letter combinations was set up. The analyzing<br />

tool joins <strong>the</strong> corpus <strong>of</strong> perceptional expressions<br />

with <strong>the</strong> model <strong>of</strong> word structure and <strong>the</strong> lists <strong>of</strong><br />

possible word composition elements. After having<br />

identified <strong>the</strong> searched sequence <strong>of</strong> characters in a<br />

word, <strong>the</strong> algorithm analyzes <strong>the</strong> characters around it.<br />

Step by step it checks whe<strong>the</strong>r <strong>the</strong> environment ei<strong>the</strong>r<br />

represents a semantically inseparable part <strong>of</strong> <strong>the</strong><br />

sequence <strong>of</strong> characters (which indicates a different<br />

word meaning) or is <strong>the</strong> rest <strong>of</strong> a compound word<br />

and <strong>the</strong>reby should be part <strong>of</strong> <strong>the</strong> output result.<br />

The tests with <strong>the</strong> tool let presume that <strong>the</strong> occurrence<br />

<strong>of</strong> perceptional expression depends on <strong>the</strong> sort<br />

<strong>of</strong> text and on <strong>the</strong> topic. A user’s test report on some<br />

instant food i.e., had a very high rate <strong>of</strong> gustatory<br />

vocabulary.<br />

IV. FURTHER RESEARCH<br />

At <strong>the</strong> moment I’m working on using this tool in user<br />

modeling. As in <strong>the</strong> internet (and esp. in blogs and<br />

forums) <strong>the</strong>re is quite a lot <strong>of</strong> text which can be directly<br />

linked to a user, this could be a rich source for<br />

modeling <strong>the</strong> user’s mental representations <strong>of</strong> perceptional<br />

information. With such knowledge about <strong>the</strong><br />

user, advertisements with a higher acceptance can be<br />

designed and chosen with respect to <strong>the</strong> user’s individual<br />

perceptional preference.<br />

94<br />

REFERENCES<br />

[1] G. Kellner. Wege der Kommunikationsoptimierung.<br />

Anwendung von NLP im Bereich der<br />

Künstlichen Intelligenz. Master Thesis, Wien,<br />

Austria, 2006.<br />

[2] G. Drosdowski (Ed.). Duden „Grammatik der<br />

deutschen Gegenwartssprache“. Duden, Mannheim,<br />

Germany, 1995.


A Simulation Environment for Distributed Real-Time<br />

Systems in <strong>the</strong> Presence <strong>of</strong> Malicious Attacks<br />

Ekarin Suethanuwong and Christian El-Salloum (Faculty Mentor)<br />

Institute <strong>of</strong> Computer Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {ekarin,salloum}@vmars.tuwien.ac.at<br />

Abstract — The purpose <strong>of</strong> this work is to develop<br />

a simulation environment for distributed real-time<br />

systems for observing system’s behaviour after<br />

malicious attacks. Our simulation framework is<br />

targeted at systems based on <strong>the</strong> Time-triggered<br />

Architecture [1]. The purpose <strong>of</strong> this simulation is<br />

to investigate and increase <strong>the</strong> security <strong>of</strong> existing<br />

dependable distributed real-time systems, especially<br />

when <strong>the</strong>se systems are connected to <strong>the</strong> Internet or<br />

any o<strong>the</strong>r open systems.<br />

I. INTRODUCTION<br />

Nowadays, distributed real-time systems tend to take<br />

more and more responsibilities in many application<br />

domains. For example in <strong>the</strong> automotive or avionics<br />

sector an increasing number <strong>of</strong> functions are<br />

controlled by embedded systems. Current and future<br />

embedded systems tend to be more open than <strong>the</strong>y<br />

were in <strong>the</strong> past. For instance <strong>the</strong>y might be<br />

connected to <strong>the</strong> Internet for remote configuration or<br />

maintenance. Due to <strong>the</strong>se facts, distributed real-time<br />

systems are facing a growing potential for malicious<br />

attacks and security is becoming a major issue in <strong>the</strong><br />

embedded domain.<br />

In order to investigate <strong>the</strong> effect <strong>of</strong> malicious<br />

attacks as well as to find out counter measures in a<br />

distributed real-time system, we are developing a<br />

simulation environment for simulating distributed<br />

real-time systems and malicious attacks.<br />

In this s<strong>of</strong>tware, users can implement different<br />

algorithms, e.g. clock synchronization algorithms<br />

like FTA (Fault-Tolerance Average) [2] or communication<br />

protocols like TTP/C [3] or TT E<strong>the</strong>rnet [4].<br />

The s<strong>of</strong>tware enables users to define <strong>the</strong> structure <strong>of</strong><br />

<strong>the</strong> simulated distributed real-time system and<br />

supports different methods to execute <strong>the</strong> simulated<br />

system and to perform malicious attacks.<br />

This paper is organized as following. In section II,<br />

<strong>the</strong> design <strong>of</strong> simulation framework is presented.<br />

Next, in section III, <strong>the</strong> simulation <strong>of</strong> malicious<br />

attacks is described. The last section will conclude<br />

<strong>the</strong> most important results.<br />

95<br />

II. DESIGN OF SIMULATION FRAMEWORK<br />

Our simulation framework aims to provide a tool<br />

to build and execute a simulation model for<br />

distributed real-time systems and malicious attacks.<br />

Users can build an arbitrary system model <strong>of</strong> distributed<br />

real-time systems via a user interface. The userdefined<br />

distributed real-time systems are executed on<br />

a discrete time scale. The simulation includes updating<br />

<strong>the</strong> internal state <strong>of</strong> nodes as well as <strong>the</strong> transmission<br />

and malicious manipulation <strong>of</strong> messages on <strong>the</strong><br />

network (i.e. it includes <strong>the</strong> simulation <strong>of</strong> nodes, <strong>the</strong><br />

network and <strong>the</strong> attacker). In <strong>the</strong> following we describe<br />

<strong>the</strong> model <strong>of</strong> a simulated system, its execution<br />

and <strong>the</strong> user interface <strong>of</strong> <strong>the</strong> simulation framework.<br />

A. SYSTEM MODEL<br />

In our model a distributed real-time systems is<br />

composed out <strong>of</strong> one or more clusters. Each cluster<br />

consists <strong>of</strong> one network and one or more connected<br />

nodes. The structure and <strong>the</strong> behaviour <strong>of</strong><br />

<strong>the</strong> simulated system are defined by <strong>the</strong> following<br />

parameters:<br />

� Number <strong>of</strong> clusters<br />

� Number <strong>of</strong> nodes in each cluster<br />

� Topology for each cluster<br />

� Transmission delays w.r.t. communication<br />

� Jitter w.r.t. communication<br />

� Drift rate <strong>of</strong> local clocks<br />

� Behavior <strong>of</strong> each node<br />

The behaviour <strong>of</strong> each node is expressed in its<br />

reaction to <strong>the</strong> reception <strong>of</strong> a message and actions<br />

performed at specific instants in time with respect to<br />

its local clock. A reaction is ei<strong>the</strong>r adapting <strong>the</strong><br />

node’s internal state or sending a message with a<br />

specific content derived from its internal state. This<br />

behaviour can be programmed by <strong>the</strong> user <strong>of</strong> <strong>the</strong><br />

simulation environment (e.g., <strong>the</strong> user can program<br />

<strong>the</strong> nodes <strong>of</strong> <strong>the</strong> simulated system to update <strong>the</strong>ir<br />

local clocks according to <strong>the</strong> FTA clock<br />

synchronization protocol).


B. EXECUTION OF THE MODEL<br />

The simulation model is executed on a discrete<br />

time scale called <strong>the</strong> simulation time. A tick in this<br />

time scale represents an instant <strong>of</strong> <strong>the</strong> real time and is<br />

called simulation tick. The length <strong>of</strong> each interval <strong>of</strong><br />

two real-time instants represented by two succeeding<br />

simulation ticks is equal and called <strong>the</strong> granularity <strong>of</strong><br />

<strong>the</strong> simulation time. The granularity <strong>of</strong> <strong>the</strong> simulation<br />

time can be defined by <strong>the</strong> user. A smaller<br />

granularity leads to a better temporal resolution <strong>of</strong><br />

<strong>the</strong> simulation results but induces a higher load for<br />

simulating a given real-time interval.<br />

At each simulation tick <strong>the</strong> simulation framework<br />

will compute <strong>the</strong> new status <strong>of</strong> <strong>the</strong> individual nodes<br />

and <strong>the</strong> network taking into account <strong>the</strong> user defined<br />

topology and behaviour <strong>of</strong> <strong>the</strong> nodes as well as <strong>the</strong><br />

characteristics <strong>of</strong> <strong>the</strong> chosen communication<br />

infrastructure (e.g., latency, bandwidth).<br />

The simulation supports different kinds <strong>of</strong> model<br />

execution:<br />

� Free-running mode: In this mode <strong>the</strong> simulation<br />

runs until a user stops it.<br />

� Break-point mode: Unlike free-running<br />

mode, <strong>the</strong> simulation runs until a specific<br />

condition is met (e.g. a specified point in<br />

time is reached, a collision has occurred, <strong>the</strong><br />

precision exceeds a specified threshold …).<br />

When <strong>the</strong> simulation is stopped or paused, <strong>the</strong> user<br />

can observe or reenter <strong>the</strong> values <strong>of</strong> desired<br />

parameters. Fur<strong>the</strong>rmore after each stop a summary<br />

is created containing defined parameters <strong>of</strong> <strong>the</strong><br />

distributed real-time system, e.g. precision during<br />

run-time, local time value <strong>of</strong> each nodes and failure<br />

time after <strong>the</strong> malicious attack.<br />

C. USER INTERFACE (UI)<br />

The simulation framework will provide an intuitive<br />

graphical user interface. Users can enter parameter<br />

values as well as monitoring for diagnostics and<br />

maintenance. For example, simulation time, local<br />

time <strong>of</strong> each nodes, precision value, correction term<br />

<strong>of</strong> local time, user-defined variable and so forth.<br />

III. SIMULATION OF MALICIOUS ATTACK<br />

In order to investigate and observe both mechanisms<br />

and effects from <strong>the</strong> malicious attacks in <strong>the</strong><br />

simulation model for distributed real-time system, a<br />

priori known malicious attacks are explicitly assumed<br />

and emulated as shown below.<br />

- Injection <strong>of</strong> messages. One or more messages are<br />

added to <strong>the</strong> communication network. A number <strong>of</strong><br />

injected messages might cause a denial <strong>of</strong> service [5].<br />

The injected messages might also cause <strong>the</strong> collision<br />

96<br />

with o<strong>the</strong>r sending messages. So, <strong>the</strong> receiver will<br />

not get messages from <strong>the</strong> sender at <strong>the</strong> expected<br />

message receive instant.<br />

- Changing message content. The content <strong>of</strong> a<br />

message sent from <strong>the</strong> correct sender into correct<br />

receiver might be changed during transmission.<br />

- Changing message latency. The malicious<br />

attacker might attack a temporal accuracy <strong>of</strong><br />

messages [6, pp. 102] by changing <strong>the</strong> message<br />

latency to be delayed or accelerated.<br />

IV. CONCLUSION AND FUTURE WORK<br />

We have purposed a framework for simulating distributed<br />

real-time systems in <strong>the</strong> presence <strong>of</strong> malicious<br />

attacks. This framework will ease <strong>the</strong> development<br />

<strong>of</strong> new security mechanisms and will be employed<br />

at our institute for <strong>the</strong> design <strong>of</strong> future safety<br />

critical architectures and applications.<br />

ACKNOWLEDGMENTS<br />

I am grateful to Christrian El-Salloum to guide me to<br />

work on this challenging research and thankful to<br />

Armin Wasicek and every colleagues for all useful<br />

comments and encouragements.<br />

REFERENCES<br />

[1] H. Kopetz and G. Bauer. The Time-Triggered<br />

Architecture, In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> IEEE Special<br />

Issue on Modeling and Design <strong>of</strong> Embedded<br />

S<strong>of</strong>tware, October 2001.<br />

[2] H. Kopetz and W. Ochsenreiter. Clock Synchronization<br />

in Distributed Real Time Systems, Research<br />

Report Nr. 1/87/II, Institute <strong>of</strong> Computer<br />

Engineering, Vienna University <strong>of</strong> Technology,<br />

Vienna, Austria, March 1987.<br />

[3] H. Kopetz, G. Bauer and W. Steiner. Dependable<br />

Time Triggered Communication, published in<br />

2005 The Industrial Communication Technology<br />

Handbook, pp. 12.1-12.6, February 2005.<br />

[4] H. Kopetz, A.Ademaj, P. Grillinger, K.<br />

Steinhammer and M. Prammer. The Time-<br />

Triggered E<strong>the</strong>rnet (TTE) Design, In Proceeding<br />

s <strong>of</strong> <strong>the</strong> 8 th IEEE International Symposium on<br />

Object-oriented Real-time distributed Computing<br />

(ISORC), Seattle, Washington, USA, 2005.<br />

[5] J. Mirkovic, S. Dietrich, D. Dittrich, P. Reiher.<br />

Internet Denial <strong>of</strong> Service: Attack and Defense<br />

Mechanisms, Prentice Hall, January 2005.<br />

[6] H. Kopetz, Real-Time Systems: Design Principles<br />

for Distributed Embedded Applications,<br />

Kluwer Academic Publishers, 4 th Edition, 1997.


Filtering <strong>of</strong> Phase Shift Terrestrial Laser Scanner Point Clouds<br />

Clemens No<strong>the</strong>gger and Norbert Pfeifer (Faculty Mentor)<br />

Christian Doppler Laboratory for ”Spatial Data from Laser Scanning and Remote Sensing”<br />

Institute <strong>of</strong> Photogrammetry and Remote Sensing<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {cn,np}@ipf.tuwien.ac.at<br />

Abstract — The latest generation <strong>of</strong> phase shift scanners<br />

features an extremely high scanning speed and improved accuracy,<br />

thus making it possible to capture surface detail in <strong>the</strong><br />

millimeter range. However, at <strong>the</strong> highest resolutions <strong>the</strong>se<br />

scanners produce huge amounts <strong>of</strong> data which are quite noisy.<br />

We present a chain <strong>of</strong> preprocessing steps which reduces <strong>the</strong><br />

amount <strong>of</strong> data and <strong>the</strong> noise without loosing detail. Effects influencing<br />

<strong>the</strong> filtering parameters are identified and a method<br />

for <strong>the</strong>ir estimation is described.<br />

I. INTRODUCTION<br />

Phase shift terrestrial laser scanners achieve measurement<br />

rates <strong>of</strong> up to 500 kHz resulting in a huge number<br />

<strong>of</strong> points (up to 600 mio.), which commercial <strong>of</strong> <strong>the</strong> shelf<br />

modeling s<strong>of</strong>tware cannot handle directly. It is <strong>the</strong>refore<br />

necessary to apply thinning to reduce <strong>the</strong> size <strong>of</strong> <strong>the</strong> data.<br />

While being fairly homogeneous locally, <strong>the</strong> point density<br />

at different parts <strong>of</strong> <strong>the</strong> scan can vary significantly,<br />

depending mainly on <strong>the</strong> polar range and angle <strong>of</strong> incidence.<br />

This makes it necessary to set parameters adaptively<br />

if optimal results are to be achieved in <strong>the</strong> processing<br />

<strong>of</strong> <strong>the</strong> point cloud. This, however, is usually not<br />

possible without significant manual work. If this preprocessing<br />

is not done carefully loss <strong>of</strong> detail or even larger<br />

scale model deformation can occur, which can be hard to<br />

detect at later stages <strong>of</strong> <strong>the</strong> processing.<br />

In this paper we propose a chain <strong>of</strong> preprocessing steps<br />

for thinning and smoothing <strong>of</strong> point clouds acquired with<br />

such scanners. The preprocessing steps include surface<br />

normal estimation, surface roughness classification, outlier<br />

detection and removal, thinning, and smoothing. Our<br />

goal is to avoid loss <strong>of</strong> detail and thus achieve millimeter<br />

accuracy <strong>of</strong> <strong>the</strong> model. Since <strong>the</strong> processing is timeconsuming<br />

it can only be applied in a batch process. This<br />

means that all parameters <strong>of</strong> <strong>the</strong> computation must be<br />

known in advance. We show, how <strong>the</strong> parameters needed<br />

in <strong>the</strong> computation can be estimated and set adaptively.<br />

II. ALGORITHM<br />

The filtering strategy consists <strong>of</strong> a sequence <strong>of</strong> four steps:<br />

The first step is <strong>the</strong> analysis <strong>of</strong> <strong>the</strong> surface and a classification<br />

according to <strong>the</strong> local curvature. The second step<br />

consists <strong>of</strong> a computation <strong>of</strong> robust surface normal vec-<br />

97<br />

tors with outlier detection and elimination. The third step<br />

is a thinning step, which is followed by a forth and final<br />

smoothing step. All steps require <strong>the</strong> local neighborhood<br />

<strong>of</strong> a point p. As <strong>the</strong> local neighborhood <strong>of</strong> a point p we<br />

use <strong>the</strong> k points which are closest to p. These k points<br />

can be found efficiently by using <strong>the</strong> kd-tree data structure<br />

[1]. The choice <strong>of</strong> k is crucial, however, if consistent<br />

results are to be achieved. In [2] it is suggested that all<br />

points within a certain radius be used instead. This radius<br />

depends on local point density and curvature. For<br />

locally homogeneous point distribution, however, this is<br />

comparable to choosing an individual k for each point.<br />

The surface analysis we perform is based on <strong>the</strong> covariance<br />

analysis <strong>of</strong> <strong>the</strong> local neighborhood. In [3] it is<br />

shown that <strong>the</strong> surface variation σk - which can be computed<br />

from <strong>the</strong> eigenvalues <strong>of</strong> <strong>the</strong> covariance matrix - can<br />

be used instead <strong>of</strong> <strong>the</strong> local surface curvature for estimating<br />

surface roughness. The surface is <strong>the</strong>n classified into<br />

smooth, rough and intermediate areas by using <strong>the</strong> maximum<br />

surface variation in <strong>the</strong> neighbourhood and applying<br />

thresholds. These thresholds heavily depend on <strong>the</strong><br />

choice <strong>of</strong> k.<br />

Normal vectors can be estimated from <strong>the</strong> covariance<br />

analysis <strong>of</strong> <strong>the</strong> local neighborhood [4]. The problem with<br />

this approach is that it only works for continuous and<br />

smooth surfaces, i.e. surfaces not containing any sharp<br />

features. To address this problem, we use a robust estimator<br />

for <strong>the</strong> covariance. We used <strong>the</strong> highly robust<br />

Fast Minimum Covariance Determinant (FMCD) estimator<br />

[5] for estimating a covariance matrix, from which <strong>the</strong><br />

surface normal vectors can be derived robustly.<br />

Robust estimation can also help with outlier detection<br />

(i.e. data not originating from <strong>the</strong> same underlying process<br />

as <strong>the</strong> majority <strong>of</strong> <strong>the</strong> o<strong>the</strong>r data), since it is not affected<br />

by leverage points [6]. Outliers in TLS may be<br />

caused for example by specular reflections, or if <strong>the</strong> laser<br />

beam hits multiple surfaces. For outlier determination,<br />

we use <strong>the</strong> result from <strong>the</strong> normal vector estimation. A<br />

point is classified as an outlier and deleted, if it is not contained<br />

in <strong>the</strong> final set <strong>of</strong> points which contains all points<br />

that are likely to be part <strong>of</strong> <strong>the</strong> same surface.<br />

In <strong>the</strong> thinning step, <strong>the</strong> estimated normal vectors are<br />

used to determine points having <strong>the</strong> highest probability<br />

<strong>of</strong> being closest to <strong>the</strong> real surface. From a randomly


selected point, neighboring points are projected to <strong>the</strong><br />

point’s normal vector. Along this vector a univariate density<br />

is estimated and <strong>the</strong> point being closest to <strong>the</strong> mode<br />

<strong>of</strong> this distribution is chosen as <strong>the</strong> representative point.<br />

The mode <strong>of</strong> <strong>the</strong> distribution is found by applying <strong>the</strong><br />

mean-shift algorithm [7].<br />

III. NEIGHBORHOOD ESTIMATION<br />

The choice <strong>of</strong> k is crucial for consistent and reliable results<br />

for <strong>the</strong> entire scan. We want to choose k such that<br />

for all flat surfaces <strong>the</strong> surface variation is constant, i.e.<br />

<strong>the</strong> ratio <strong>of</strong> <strong>the</strong> smallest to <strong>the</strong> largest eigenvalue <strong>of</strong> <strong>the</strong><br />

covariance matrix <strong>of</strong> <strong>the</strong> neighborhood must be constant:<br />

λ3/λ1 = const(withλ1 >λ2 >λ3) (1)<br />

For a given start point, we can compute <strong>the</strong> ratio λ3/λ1<br />

for increasing k. At a certain point this ratio will drop<br />

below a threshold. k is <strong>the</strong>n chosen as <strong>the</strong> point where<br />

<strong>the</strong> ratio falls below <strong>the</strong> threshold τ. The value <strong>of</strong> τ can<br />

be chosen empirically. It depends on <strong>the</strong> error bounds for<br />

<strong>the</strong> normal vectors.<br />

For TLS <strong>the</strong> measurement noise is fairly constant, increasing<br />

just slightly with increasing range. The point<br />

density, however, decreases rapidly, proportional to <strong>the</strong><br />

square <strong>of</strong> <strong>the</strong> range. Ano<strong>the</strong>r influence is <strong>the</strong> angle <strong>of</strong> incidence.<br />

As <strong>the</strong> angle <strong>of</strong> incidence increases, <strong>the</strong> point<br />

density decreases. Some scanners do not correct <strong>the</strong> convergence<br />

<strong>of</strong> scan lines at <strong>the</strong> zenith and nadir, thus <strong>the</strong>y<br />

exhibit an increased point density towards <strong>the</strong>se points.<br />

We propose <strong>the</strong> following empirical formula to compute<br />

k:<br />

k = a · cos(α)/(r · sin(θ)) + b (2)<br />

where α is <strong>the</strong> angle <strong>of</strong> incidence, r is <strong>the</strong> range, θ is<br />

<strong>the</strong> zenith angle and a and b are coefficients. The term<br />

sin(θ) must only be used if <strong>the</strong> scanner does not compensate<br />

<strong>the</strong> convergence <strong>of</strong> scan lines.<br />

IV. RESULTS<br />

To evaluate <strong>the</strong> quality <strong>of</strong> <strong>the</strong> point cloud after preprocessing,<br />

we compared a triangulated model <strong>of</strong> <strong>the</strong> preprocessed<br />

points to <strong>the</strong> original point cloud. For a best<br />

fitting model we expect randomly distributed differences<br />

with a magnitude equal to that <strong>of</strong> <strong>the</strong> measurement noise<br />

level, no systematic deformations, but possibly large differences<br />

at erroneously measured points.<br />

We tested <strong>the</strong> method on two datasets, <strong>the</strong> first being<br />

an 80 by 40 cm part <strong>of</strong> a pillar <strong>of</strong> an arcade <strong>of</strong><br />

Schönbrunn Palace. This object mainly consists <strong>of</strong> planar<br />

faces which are bounded by sharp edges. We chose<br />

this object to test our method’s behaviour at <strong>the</strong>se edges.<br />

The second test dataset is a 45 by 35 cm part <strong>of</strong> a historic<br />

stove. This datasets contains surfaces <strong>of</strong> varying<br />

98<br />

curvatures, some sharp edges but no flat surfaces. We<br />

compared <strong>the</strong> model generated in our workflow to models<br />

derived using commercial s<strong>of</strong>tware packages. While<br />

all models were acceptable, we were able to show that<br />

our approach produced <strong>the</strong> smoo<strong>the</strong>st surface with <strong>the</strong><br />

least deformation.<br />

V. CONCLUSION<br />

The processing <strong>of</strong> point clouds from phase shift terrestrial<br />

laser scanners poses a number <strong>of</strong> challenges that<br />

must be overcome if <strong>the</strong> full potential <strong>of</strong> <strong>the</strong> instruments<br />

is to be tapped. Among o<strong>the</strong>rs <strong>the</strong> huge size <strong>of</strong> <strong>the</strong> point<br />

cloud and <strong>the</strong> relatively high noise level have been addressed<br />

in this paper. We demonstrated that our method<br />

achieves <strong>the</strong> goal <strong>of</strong> preserving surface detail while at<br />

<strong>the</strong> same time achieving significant smoothing. We also<br />

showed how it can be applied automatically by providing<br />

a method for estimating <strong>the</strong> parameters and setting <strong>the</strong>m<br />

adaptively <strong>the</strong>m according to known factors.<br />

ACKNOWLEDGMENTS<br />

We would like to thank <strong>the</strong> private management <strong>of</strong><br />

Schönbrunn Palace and Steinmetzbetriebe Bamberger as<br />

members <strong>of</strong> <strong>the</strong> Christian Doppler-Laboratory for supporting<br />

our investigations.<br />

REFERENCES<br />

[1] M. de Berg, M. van Krefeld, M. Overmars, and<br />

O. Schwarzkopf. Computational Geometry Algorithms<br />

and Applications. Springer, 2000.<br />

[2] N.J. Mitra and A. Nguyen. Estimating surface normals<br />

in noisy point cloud data. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong><br />

nineteenth annual symposium on Computational geometry,<br />

pages 322–328. ACM New York, NY, USA,<br />

2003.<br />

[3] M. Pauly, M. Gross, L. P. Kobbelt, E. T. Hochschule,<br />

and S. Zurich. Efficient simplification <strong>of</strong> pointsampled<br />

surfaces. In IEEE Visualization ’02, pages<br />

163–170, 2002.<br />

[4] TK Dey, G. Li, and J. Sun. Normal estimation for<br />

point clouds: a comparison study for a Voronoi based<br />

method. In Point-Based Graphics, 2005. Eurographics<br />

<strong>Proceedings</strong>, pages 39–46, 2005.<br />

[5] P. J. Rousseeuw and K. van Driessen. A fast algorithm<br />

for <strong>the</strong> minimum covariance determinant estimator.<br />

Technometrics, 41(3):212–223, 1999.<br />

[6] P.J. Rousseeuw and A.M. Leroy. Robust Regression<br />

and Outlier Detection. Wiley, 1987.<br />

[7] Y. Cheng. Mean Shift, Mode Seeking, and Clustering.<br />

IEEE Transactions on Pattern Analysis and Machine<br />

Intelligence, pages 790–799, 1995.


�����������������������������������������������<br />

��������������������������<br />

������������������������������<br />

�����������������������������<br />

��������������������<br />

��������������������������������������������������������<br />

������������������������������������������������<br />

�����������������������������������������������������<br />

�� � ��� � �� � ���� � �� � ���� � ��� � ������� � �� � �� � ����� �<br />

�����������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������������������<br />

��������������������������������������������������<br />

����������� ��� � ��� � ����������� � �������� � ��� � ���<br />

������������� � ����������� � ��� � ����������� � ���<br />

����� � �� � ��� � ������������ � ���������� � � � ����<br />

������������ � ��������� � ��� � ��� � ���� � �� � ��������<br />

���������� � ���� � �� � ��������� � ��� � ����������<br />

���������������������������������������������������� �<br />

�� � ��� � ����� � �� � ����� � ������������ � ���������� ����<br />

���������������������������������������������������<br />

����������������������������������������������� �<br />

������������������<br />

�� ������������<br />

��� � ������������ � ���������� � ��� � ����� � ���� � ���<br />

����������������������������������������������������<br />

���������������������������������������������������������<br />

��������������������������������������������������������<br />

��������������������������������������������������<br />

���� ������� � ��� � �� � ��� � �������� � �� � ����� � ��������<br />

����������������������������������������������������<br />

��� � ���� � ����� � ���� � ������� � ���� � �� � ����� � ���<br />

��������������������������������������������������<br />

����������������������������������������������������<br />

������ � ����� � ����� � ������������ � �� � ��� � ���������<br />

�������������������������������������<br />

��������������������������������������������������<br />

�������������������������������������������������<br />

� ����������������������������������������������<br />

���������������<br />

� �����������������������������������<br />

� ��������������������������������<br />

� ��������������������������������������������<br />

������������������������<br />

����������������������������������������������������<br />

�����������������������������������������������������<br />

99<br />

��� � �������� � ���� � ��������� � �������������<br />

���������������������������������������������������<br />

��������������������������������������������������������<br />

������������������������������� � ������ ������������<br />

�����������������������������������������������������<br />

������������������������<br />

��������������������������������������������������<br />

������������������������������������������������������<br />

������������� � �� � ��� � ���� � �� � ��� � �������������<br />

�������������� � �� � ���������� � ���� � �� � ��������<br />

����������������������������������������������<br />

���������������������������������������������������<br />

������������������������������������������������������<br />

����������������������������������������������������<br />

����� � �� � ��������� � ���������� � ���� � ��� � ��������<br />

��������������������������������������<br />

��� ����������������<br />

������������������������������������������������<br />

����������������������������������������������������<br />

��� � �������� � �������� � ����������� � ��� � �� � ����<br />

��������������������������������������������������<br />

������������������������������� � ���� � ����<br />

�������������������������������������������������<br />

���������������������������������������������������<br />

���������������������������������������������������<br />

���������������������������������������������������<br />

�����������������������������������<br />

�����������������������������������������������<br />

��������������������������������


�������������������������������������������<br />

������������<br />

��� ����� ��������������������� ��������� � �������<br />

�������������������������������������������������<br />

��������������������������������������������������<br />

���������������������������������������������������<br />

���� � ���������������������������������������������<br />

�������������������������������������������<br />

������������<br />

��������������������������������������������<br />

��������������������<br />

���������������<br />

������������������ � �� � ���� ������������� �����<br />

����������� � ��������� � �� � ��� � ����� � ������<br />

��������������������������������������������<br />

������������ ������� � ��� � �������� � ��������<br />

��� � ����������� � �������� � ������<br />

�������� � ������� � ��� � ������������� � ����<br />

����������������������������������������������<br />

�������<br />

100<br />

����������<br />

�������������������������������������������������<br />

���������������������������������������������<br />

������� � ������������� � �� ������ � ����� ������<br />

����������������������������������������������<br />

��� � �������� � �������������� � ��������<br />

�������� ��� � ��� � ���������� � ��� � �������<br />

��������� � ������� � �������������������<br />

�������������������������������������������������<br />

���������������<br />

��� ��������� ����������������������������������<br />

������� � ���������� � � � ����������� � ������ � ���<br />

������������������������������������ �������<br />

���������� � �� � ����� � ������������ ������������<br />

���������� � ����������� � ��� � ���������������<br />

�����������������������<br />

�����������������������������������������������<br />

����������������������������������������������<br />

�� ���������� � �������� ����� �������� � ���<br />

��������� ������ � ����������� � ���������<br />

��������������� � ������ � ����������� � ����� � ���<br />

���������������������������������<br />

��� ������������� � ��� � ������������ ����<br />

��������������� ������������� � �� ���������<br />

�����������������������������������������������<br />

������������������<br />

��� ������ ��� �������������������� � � � ����������<br />

������������� � ������ � ����� � ������������������<br />

�����������������������


Hybrid Timing Analysis for ANSI-C Applications<br />

with Loops and Function Calls<br />

Bernhard Rieder, Peter Puschner<br />

Institute <strong>of</strong> Computer Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {bernhard,peter}@vmars.tuwien.ac.at<br />

Abstract — Todays automotive industry uses an increasing<br />

number <strong>of</strong> electronic control units (ECUs) to perform complex<br />

safety-critical tasks.<br />

The systems used to implement <strong>the</strong>se tasks are hard real-time<br />

systems which means that <strong>the</strong> result <strong>of</strong> a calculation is only usefull<br />

when it has <strong>the</strong> correct value and when it arrives on time.<br />

The failure <strong>of</strong> such a system may cause catastrophic results,<br />

such as <strong>the</strong> loss <strong>of</strong> human life.<br />

Worst Case Execution Time (WCET) Analysis is used to determine<br />

<strong>the</strong> maximum execution time <strong>of</strong> a task running on a realtime<br />

system, an information which required to ensure that each<br />

task meets its deadline. The presented work describes a hybrid<br />

WCET analysis method, combining static program analysis and<br />

dynamic execution time measurements.<br />

I. INTRODUCTION<br />

Traditionally, <strong>the</strong> execution time <strong>of</strong> an application was<br />

determined by manually performed end-to-end measurements<br />

using user-generated input data which was supposed<br />

to cover <strong>the</strong> execution path with <strong>the</strong> biggest execution<br />

time. In <strong>the</strong> last years <strong>the</strong> control applications have<br />

become increasingly complex. Now a control application<br />

for a power train has more than 10 40 possible executions<br />

paths ∗ which makes impossible to cover all paths using<br />

end-to-end measurements.<br />

In this article we describe a method to perform segmentation<br />

<strong>of</strong> applications and perform end-to-end measurements<br />

on <strong>the</strong> individual segments. Using <strong>the</strong> WCET<br />

<strong>of</strong> each Program Segment (PS) <strong>the</strong> WCET <strong>of</strong> <strong>the</strong> whole<br />

application is estimated. The presented work extends <strong>the</strong><br />

prototype described in [1] to support loops and function<br />

calls.<br />

This article is structured as follows: Section II. describes<br />

how <strong>the</strong> hybrid WCET calculation method works.<br />

Section III. explains how it is extended to support loops<br />

and function calls, which is <strong>the</strong> main contribution <strong>of</strong> this<br />

work. Section IV. discusses topics that require fur<strong>the</strong>r<br />

research and section V. describes similar work.<br />

II. HYBRID WCET ANALYSIS<br />

The hybrid WCET analysis consists <strong>of</strong> fife steps which<br />

are depicted in Figure 1: ➀ Static Program Analysis is<br />

performed to identify variables and control structures and<br />

∗ Code supplied from an industrial partner from <strong>the</strong> MoDECs project<br />

101<br />

to generate a Control Flow Graph (CFG). ➁ The CFG<br />

is split up into individual Program Segments during <strong>the</strong><br />

CFG Decomposition. ➂ In <strong>the</strong> next phase, <strong>the</strong> Test Data<br />

Generation, model-checking is used to generate test data<br />

to cover all paths <strong>of</strong> each program segment. ➃ These data<br />

are used to perform Execution Time Measurements from<br />

which ➄ <strong>the</strong> WCET is derived during <strong>the</strong> WCET Calculation<br />

step. The individual steps <strong>of</strong> <strong>the</strong> analysis process<br />

are described in more detail in [1] and [2].<br />

Figure 1: Hybrid WCET Analysis Framework<br />

Since <strong>the</strong> CFG partitioning and <strong>the</strong> test data generation<br />

step are key points <strong>of</strong> <strong>the</strong> hybrid WCET analysis approach,<br />

<strong>the</strong>y shall be briefly described here.<br />

The CFG partitioning works by reducing <strong>the</strong> number<br />

<strong>of</strong> decisions and <strong>the</strong>refore paths in <strong>the</strong> analyzed code.<br />

When looking at a series <strong>of</strong> n+m conditional expressions<br />

we can observe 2 n+m execution paths. When <strong>the</strong> code is<br />

partitioned in two program segments PSn and PSm only<br />

2 n +2 m paths exist. Considering a piece <strong>of</strong> code with 16<br />

subsequent if-<strong>the</strong>n-else constructs we can observe a total<br />

<strong>of</strong> 2 16 = 65536 paths while only 2 ∗ 2 8 = 512 paths<br />

exists if we create two program segments (PS). If <strong>the</strong><br />

code is partitioned in four segments, this number drops<br />

to 4 ∗ 2 4 =64. In this simple idealized example <strong>the</strong> number<br />

<strong>of</strong> paths that need to be measured has dropped from<br />

65536 to 64. Figure 2 shows <strong>the</strong> effect <strong>of</strong> segmentation<br />

on an applications from <strong>the</strong> automotive domain [3]. The<br />

first column shows <strong>the</strong> number <strong>of</strong> program segments. In<br />

<strong>the</strong> first line this is 1, which means <strong>the</strong> programs segment<br />

comprises <strong>the</strong> whole application. The second column<br />

shows <strong>the</strong> number <strong>of</strong> paths that need to be measured after<br />

<strong>the</strong> segmentation. The third number is <strong>the</strong> Path bound.


The path bound is <strong>the</strong> maximum number <strong>of</strong> paths within<br />

a single PS and is used for <strong>the</strong> automatic partitioning algorithm.<br />

The last column represents <strong>the</strong> overall analysis<br />

time (including execution time measurements) in seconds.<br />

Instead <strong>of</strong> 1.9 ∗ 10 11 end-to-end measurements we<br />

need only 92 measurements when using 14 program segments<br />

and a fraction <strong>of</strong> <strong>the</strong> time.<br />

PS Paths Path bound Time [s]<br />

1 1.9E+11 unbounded -<br />

5 1455 1000 41402<br />

7 336 100 7794<br />

8 242 50 -<br />

14 92 10 957<br />

Figure 2: Hybrid WCET Analysis Framework<br />

The test data generation in step ➂ uses a heuristic stage<br />

to find <strong>the</strong> majority <strong>of</strong> results very quickly and a modelchecking<br />

stage to find <strong>the</strong> remaining data. Since modelchecking<br />

is complete, which means any existing solution<br />

will be found, we can used it not only for test data generation<br />

but also to identify infeasible paths. When <strong>the</strong> test<br />

data generation finishes we have test data for each path<br />

within a PS or we know that a syntactical possible path<br />

is semantically infeasible. This gives us <strong>the</strong> certainty that<br />

<strong>the</strong> worst-case execution time <strong>of</strong> each PS has been found.<br />

This does not mean that <strong>the</strong> local WCET path is part <strong>of</strong><br />

<strong>the</strong> global WCET path which gives a chance for <strong>the</strong> overestimation<br />

<strong>of</strong> <strong>the</strong> global WCET.<br />

III. LOOPS AND FUNCTION CALLS<br />

While <strong>the</strong> WCET analysis framework created in [3] does<br />

not rely on code modifications o<strong>the</strong>r than inserting Instrumentation<br />

Points (IPs) for measurements extensive alterations<br />

on <strong>the</strong> source code are necessary for <strong>the</strong> measurement<br />

<strong>of</strong> loops and functions calls. The measurement <strong>of</strong><br />

loops uses model checking to find <strong>the</strong> maximum iteration<br />

count <strong>of</strong> <strong>the</strong> loop[4]. After <strong>the</strong> loop bound is known,<br />

<strong>the</strong>re are three possible ways to perform execution time<br />

measurements:<br />

α) The loop has no control flow decisions and can be<br />

covered like a single expression without fur<strong>the</strong>r action.<br />

β) The loop has data dependent control flow and<br />

Paths(loop) Iterations ≤ Path Bound: The loop has to be<br />

unrolled for test data generation but is kept as loop for<br />

execution time measurements without any IPs.<br />

γ) The loop has data dependent control flow and<br />

Paths(loop) Iterations > Path Bound: Handle <strong>the</strong> loop<br />

like β but place an IP inside <strong>the</strong> loop and perform a measurement<br />

for each iteration.<br />

The last approach is <strong>the</strong> most invasive since measurements<br />

can reduce <strong>the</strong> performance considerably[2].<br />

Function calls are handled similar. If it is not necessary<br />

to place IPs into <strong>the</strong> function when Paths(loop) ≤<br />

102<br />

Path Bound <strong>the</strong> function is inlined in <strong>the</strong> model generation<br />

but kept as function for execution time measurements.<br />

If Paths(loop) > Path Bound <strong>the</strong>n <strong>the</strong> function<br />

is inlined which allows <strong>the</strong> placement <strong>of</strong> IPs inside <strong>the</strong><br />

function.<br />

IV. OPEN ISSUES AND OUTLOOK<br />

This work assumes that <strong>the</strong> source code CFG is equivalent<br />

to <strong>the</strong> CFG <strong>of</strong> <strong>the</strong> object code and that instructions<br />

execute with <strong>the</strong> same latency. The first assumption requires<br />

a mapping between object code and source code<br />

to check for equality. The second assumption requires<br />

<strong>the</strong> identification <strong>of</strong> variable latency instructions an <strong>the</strong>ir<br />

consideration in <strong>the</strong> measured execution times.<br />

V. RELATED WORK<br />

Chapter 35 <strong>of</strong> [5] gives a very complete overview <strong>of</strong><br />

WCET analysis and explains different techniques <strong>of</strong> time<br />

acquisition (static, dynamic) as well as different methods<br />

<strong>of</strong> WCET calculation (tree based, path based and IPET)<br />

and much more.<br />

ACKNOWLEDGMENTS<br />

This work has been supported by <strong>the</strong> FIT-IT research<br />

project “ATDGEN - Automatic Test Data Generation<br />

for WCET Measurements” under project number<br />

812653/1729-GLE/BLC.<br />

REFERENCES<br />

[1] I. Wenzel, B. Rieder, R. Kirner, and P. Puschner. Automatic<br />

Timing Model Generation by CFG Partitioning<br />

and Model Checking. In Proc. <strong>of</strong> <strong>the</strong> <strong>Conference</strong><br />

on Design, Automation and Test in Europe, pages<br />

606–611, 2005.<br />

[2] B. Rieder, I. Wenzel, K. Steinhammer, and<br />

P. Puschner. Using a Runtime Measurement Device<br />

with Measurement-Based WCET Analysis. In Proc.<br />

International Embedded Systems Symposium 2007,<br />

pages 15–26, 2007.<br />

[3] I. Wenzel, R. Kirner, B. Rieder, and P. Puschner.<br />

Measurement-Based Timing Analysis. In Proc. 3rd<br />

Int’l Symposium on Leveraging Applications <strong>of</strong> Formal<br />

Methods, Verification and Validation, <strong>2008</strong>.<br />

[4] B. Rieder, P. Puschner, and I. Wenzel. Using Model<br />

Checking to derive Loop bounds <strong>of</strong> general Loops<br />

within ANSI-C applications for measurement based<br />

WCET analysis. In Proc. Sixth Workshop on Intelligent<br />

Solutions in Embedded Systems (WISES’08),<br />

<strong>2008</strong>.<br />

[5] I. Lee, J. Y-T. Leung, and S. H. Son, editors. Handbook<br />

<strong>of</strong> Real-Time and Embedded Systems. Computer<br />

and Information Science Series. Chapman &<br />

Hall/CRC, 1 edition, 7 2007. ISBN 1-58488-678-1.


Deterministic Finite Automaton with Perfect Hashing for Fast<br />

Pattern Matching<br />

Abstract — As todays networks grow very fast, it is necessary<br />

to protect networks by security systems such as firewalls and<br />

Intrusion Detection Systems. Pattern matching is time critical<br />

operation on multigigabit networks for current IDS. Patterns<br />

are <strong>of</strong>ten described by regular expressions. The paper deals<br />

with fast regular expression matching using <strong>the</strong> Deterministic<br />

Finite Automata (DFA) with perfect hash function. We introduce<br />

decomposition <strong>of</strong> <strong>the</strong> statement to <strong>the</strong> transformation <strong>of</strong><br />

<strong>the</strong> alphabet and fast DFA. We introduced perfect hash to reduce<br />

space/speed trade <strong>of</strong>f for DFA transition table.<br />

I. INTRODUCTION<br />

In recent years, Internet becoming very popular to connect<br />

computers all over <strong>the</strong> world. While <strong>the</strong> availability<br />

<strong>of</strong> always-on communication has created many new opportunities<br />

it has also bring new possibilities for <strong>the</strong> malicious<br />

users. Therefore <strong>the</strong> importance <strong>of</strong> <strong>the</strong> network<br />

security is growing. Intrusion Detection Systems (IDS)<br />

are used for <strong>the</strong> detection <strong>of</strong> <strong>the</strong> malicious activity on <strong>the</strong><br />

network.<br />

To define suspicious activities most modern IDS relies<br />

on set <strong>of</strong> rules that are applied to each input packet. The<br />

simplest rules are in <strong>the</strong> form <strong>of</strong> packet type and pattern<br />

<strong>of</strong> content. Looking for such patterns is <strong>the</strong> core operation<br />

for <strong>the</strong> IDS.<br />

Modern IDS cannot process each packet independently,<br />

because an attacker can possibly send part <strong>of</strong> <strong>the</strong><br />

attack at <strong>the</strong> end <strong>of</strong> one packet and <strong>the</strong> o<strong>the</strong>r part at <strong>the</strong> beginning<br />

<strong>of</strong> <strong>the</strong> following packet. To identify such attacks,<br />

IDS must scan each network flow as one stream. For <strong>the</strong><br />

stream reconstruction, we need to store some information<br />

for every flow. If DFA is used for pattern matching, <strong>the</strong><br />

stored information is state. As <strong>the</strong> number <strong>of</strong> flows can<br />

be high and memory size is limited, one state has to be<br />

represented by only small amount <strong>of</strong> memory.<br />

II. PERFECT HASHING<br />

Under <strong>the</strong> concept <strong>of</strong> Perfect Hash a hash function that<br />

has no collisions in <strong>the</strong> given set <strong>of</strong> keys is understood.<br />

Such hash functions are rare and <strong>the</strong>ir identification can<br />

be computationally intensive. However, it allows to determine<br />

<strong>the</strong> position <strong>of</strong> <strong>the</strong> key in hash table in constant<br />

time in <strong>the</strong> worst case. This property can be very helpful<br />

in <strong>the</strong> implementation <strong>of</strong> <strong>the</strong> DFA, because we <strong>of</strong>ten<br />

work with very sparse and large transition table in which<br />

Jan Kastil and Jan Korenek (supervisor)<br />

Faculty <strong>of</strong> Information Technology<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: ikastil@fit.vutbr.cz<br />

103<br />

we need to lookup. As an disadvantage <strong>of</strong> perfect hashing,<br />

<strong>the</strong> fact that we need to store Perfect hash function<br />

in <strong>the</strong> memory can be seen.<br />

Many algorithms exist which aim at finding perfect<br />

hash function. They differ mainly in <strong>the</strong> number <strong>of</strong> bits<br />

which are needed to store Perfect Hash Function (PHF).<br />

Each hash function maps its key into integer interval.<br />

With PHF each key has its own number. It is <strong>of</strong>ten required<br />

<strong>the</strong> output interval to be minimal. Than we spoke<br />

about Minimal Perfect Hash Function (MPHF), which is<br />

PHF which output interval size is equal to <strong>the</strong> number <strong>of</strong><br />

keys. For our research, we chose algorithm presented in<br />

[1] and used it for <strong>the</strong> implementation <strong>of</strong> <strong>the</strong> transition<br />

table. This algorithm requires only about 3 bit per key to<br />

store MPHF.<br />

III. PRINCIPLES OF THE METHOD<br />

In <strong>the</strong> worst case we need to examine each character <strong>of</strong><br />

<strong>the</strong> input stream. Automaton in <strong>the</strong> FPGA can make one<br />

transition per clock cycle and work up to 200 Mhz. To<br />

achieve <strong>the</strong> throughput needed in modern networks we<br />

need to accept more than one character in each transition<br />

ei<strong>the</strong>r by implementing more automata or by means<br />

<strong>of</strong> input stream transformation. For reducing state information<br />

and possible problem with coordination <strong>of</strong> many<br />

automata we use second approach.<br />

Figure 1 shows that our method consist <strong>of</strong> two independent<br />

steps. First, <strong>the</strong> alphabet transformation by a<br />

shared decoder. Second, pattern matching by DFA.<br />

Figure 1: Description <strong>of</strong> our method


A. PERFECT HASH AS TRANSITION FUNCTION<br />

We propose to use DFA as a pattern match algorithm.<br />

Figure 1 shows detailed logical structure <strong>of</strong> DFA. The<br />

computation <strong>of</strong> <strong>the</strong> next state consist <strong>of</strong> three blocks. In<br />

<strong>the</strong> first block, <strong>the</strong> position <strong>of</strong> <strong>the</strong> transition is computed.<br />

The second is transition table itself and it returns <strong>the</strong><br />

value <strong>of</strong> <strong>the</strong> next state and information needed for <strong>the</strong> validation<br />

<strong>of</strong> <strong>the</strong> result. The third block is called next state<br />

logic and it validates results from <strong>the</strong> transition table.<br />

Each transition is represented by starting state, symbol<br />

and ending state. The first two values are used to address<br />

position in <strong>the</strong> transition table and <strong>the</strong> last is stored in addressed<br />

position. Our experiments show that transition<br />

table is extremely sparse. We use hash table to remove<br />

unused spaces from <strong>the</strong> table. In <strong>the</strong> first block, hash from<br />

input state and symbol is computed. Transition table <strong>the</strong>n<br />

returns 3-tuple <strong>of</strong> state, symbol and next state. State and<br />

symbol are compared with actual state and input symbol<br />

in <strong>the</strong> validate block. If classical approach to hash tables<br />

was used, we should have several possible transitions returned<br />

from <strong>the</strong> transition table. In such case, all <strong>of</strong> <strong>the</strong>m<br />

would have to be tested and time complexity <strong>of</strong> <strong>the</strong> third<br />

block would be linear in <strong>the</strong> worst case. We solve this<br />

problem by using MPH in <strong>the</strong> first block.<br />

When we find minimal perfect hash function for <strong>the</strong><br />

transition table <strong>the</strong>n every transition will have one unique<br />

position and no empty space will be in <strong>the</strong> table. We still<br />

have to solve collisions, because nonexistent transitions<br />

are mapped into <strong>the</strong> table and <strong>the</strong>y collide with some existing<br />

transition. As we use MPHF we are sure that if<br />

actual state has outgoing transition with input symbol it<br />

will be situated on returned position. Therefore only one<br />

comparison in validate block is needed.<br />

Perfect hash has to be pre-computed. Perfect hash<br />

function has to be stored in <strong>the</strong> memory and <strong>the</strong>refore<br />

it increase memory requirements. Algorithm described<br />

in [1] is used, because its memory requirements are increasing<br />

linearly.<br />

IV. RESULTS<br />

Memory consumption <strong>of</strong> our approach is O(kn), where<br />

n is number <strong>of</strong> transitions. Due to space limit we used<br />

only two o<strong>the</strong>r methods for <strong>the</strong> comparison. In [2],<br />

next state logic and transition table are implemented as<br />

a treebitmap. The second method is called bit parallel<br />

and was published in [3].<br />

We chose 12 rules from virus rule set from snort and<br />

combine <strong>the</strong>n into one big automaton. We estimated<br />

memory needed to store this automaton by mentioned<br />

methods. Results are shown in table 1.<br />

From Table 1, it can be recognized that our method<br />

requires less memory than o<strong>the</strong>r methods. The difference<br />

grows with <strong>the</strong> number <strong>of</strong> symbols in <strong>the</strong> alphabet <strong>of</strong> <strong>the</strong><br />

automaton.<br />

This result proves that our method is suitable for automata<br />

with big alphabets. Memory required for <strong>the</strong> alphabet<br />

transformation is not taken into account, because<br />

104<br />

n Alphabet<br />

size<br />

Tree<br />

bitmap<br />

[kb]<br />

Perfect<br />

Hash<br />

[kb]<br />

Bit parallel<br />

[kb]<br />

2 777 330 42 1400<br />

3 3213 1400 140 3600<br />

4 13000 6500 700 12000<br />

Table 1: Table shown numbers <strong>of</strong> <strong>the</strong> symbols in <strong>the</strong> alphabet<br />

<strong>of</strong> FA and memory in bits needed to store automaton<br />

it will be equal for all methods.<br />

Our results prove that for accepting more than one<br />

character per clock we have to deal with big alphabets.<br />

V. CONCLUSION AND FUTURE RESEARCH<br />

We propose new regular expression matching algorithm<br />

based on DFA which utilizes Perfect Hash functions for<br />

fast transition table look-up. Finite Automaton with Perfect<br />

Hash can perform transition in constant time even<br />

with extremely big alphabets and allows to store sparse<br />

transition in compact form. Our solution does not implement<br />

any additional compression into <strong>the</strong> transition table<br />

and still has better memory requirements than o<strong>the</strong>r<br />

tested solutions. Moreover, <strong>the</strong> proposed algorithm is<br />

scalable well even for multigigabit networks by increasing<br />

number <strong>of</strong> characters accepted within one clock cycle.<br />

The memory overhead <strong>of</strong> <strong>the</strong> perfect hash function<br />

is extremely small, <strong>the</strong>refore FPGA resources could be<br />

effectively utilized and even large automaton can fit to<br />

on-chip memory.<br />

There are many possible directions for future research.<br />

First, we could concentrate on <strong>the</strong> alphabet transformation.<br />

This would allow us to increase <strong>the</strong> number <strong>of</strong><br />

characters accepted at once and increase <strong>the</strong> throughput.<br />

Second, we could reduce memory requirements <strong>of</strong> our<br />

method by introducing additional compression <strong>of</strong> records<br />

in transition table. It would be also possible to increase<br />

average throughput by skipping characters. The research<br />

in this area will be <strong>the</strong> core <strong>of</strong> my future work.<br />

REFERENCES<br />

[1] Fabiano C. Botelho, Rasmus Pagh, and Nivio Ziviani.<br />

Simple and space-efficient minimal perfect<br />

hash functions. In In Proc. <strong>of</strong> <strong>the</strong> 10th Intl. Workshop<br />

on Data Structures and Algorithms, pages 139–150.<br />

Springer LNCS, 2007.<br />

[2] Nathan Tuck, Timothy Sherwood, Brad Calder, and<br />

George Varghese. Deterministic memory-efficient<br />

string matching algorithms for intrusion detection. In<br />

In IEEE Infocom, Hong Kong, pages 333–340, 2004.<br />

[3] Gonzalo Navarro and Mathieu Raffinot. New techniques<br />

for regular expression searching. Algorithmica,<br />

41(2):89–116, November 2004.


Network Coding for Cooperative Communications<br />

Gordhan Das Menghwar<br />

Christoph Mecklenbräuker (Faculty Mentor)<br />

Institute <strong>of</strong> Communications and Radio Frequency Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: gdas@nt.tuwien.ac.at<br />

Abstract — In this paper, we propose network coding, as an<br />

efficient and low cost option for <strong>the</strong> implementation <strong>of</strong> block-<br />

Markov coding for cooperative communications. Cooperative<br />

communications is a strategy where users, besides transmitting<br />

<strong>the</strong>ir own encoded information, also relay re-encoded versions<br />

<strong>of</strong> o<strong>the</strong>r users’ information to a common destination. Our work<br />

is motivated by <strong>the</strong> recent information-<strong>the</strong>oretic results on user<br />

cooperation. These results show that cooperating users in a<br />

multiple access channel can improve on <strong>the</strong> classical multiple<br />

access channel capacity region under <strong>the</strong> same average transmit<br />

power constraint. Additionally, cooperation provides diversity<br />

gain. The proposed implementation <strong>of</strong> <strong>the</strong> above information<br />

<strong>the</strong>oretic coding with network coding is very simple and<br />

cost effective.<br />

I. INTRODUCTION<br />

Signals transmitted through radio channels are prone to<br />

wave propagation effects like scattering, reflection, refraction,<br />

and diffraction. The result <strong>of</strong> <strong>the</strong>se effects is a<br />

faded signal received at <strong>the</strong> destination. This faded signal,<br />

whose receive power level varies over time, is subject<br />

to noise and interference which result in unreliable<br />

bit decisions. To counter <strong>the</strong>se effects, one strategy is to<br />

increase <strong>the</strong> diversity order <strong>of</strong> <strong>the</strong> communications channel<br />

by providing several independent copies <strong>of</strong> <strong>the</strong> same<br />

transmitted signal.<br />

For this purpose many strategies like space, frequency,<br />

and time diversity have been used. In <strong>the</strong> line <strong>of</strong> same efforts<br />

<strong>of</strong> improving <strong>the</strong> signal quality at receiver, user cooperation<br />

was introduced by Sendonaris et al [1, 2]. Cooperative<br />

diversity is a communications strategy where<br />

users besides transmitting <strong>the</strong>ir encoded information also<br />

relay re-encoded version <strong>of</strong> o<strong>the</strong>r users’ information to a<br />

common destination.<br />

In our work, we use network coding [3] to implement<br />

block-Markov coding used by [4] to achieve diversity and<br />

capacity gain by user cooperation. The structure <strong>of</strong> <strong>the</strong><br />

random codebooks used users u1 and u2 is as follows:<br />

x1 = w1<br />

x2 = w2<br />

� (α1P1)+r � (1 − α1)P1<br />

� (α2P2)+r � (1 − α2)P2<br />

(1)<br />

(2)<br />

for some 0 ≤ αi ≤ 1, where <strong>the</strong> transmission takes place<br />

in blocks. wi represents <strong>the</strong> fresh information in <strong>the</strong> current<br />

block and r is <strong>the</strong> cooperative refinement information<br />

depending on <strong>the</strong> fresh information sent in <strong>the</strong> previ-<br />

105<br />

ous block [4].<br />

Our results show that by using network coding in cooperative<br />

communications we can achieve diversity gain<br />

and <strong>the</strong> said strategy is easy to implement practically.<br />

The paper is structured as follows: In section II we<br />

present our system model. In section III, we describe <strong>the</strong><br />

network coding and receiver structure for that. Section<br />

IV presents our results and finally we conclude in section<br />

V.<br />

II. SYSTEM MODEL<br />

In our cooperative communications setup, two users u1<br />

and u2 wish to send information, denoted by wi where<br />

i = 1, 2, to single destination as shown in Figure 1. We<br />

Figure 1: cooperative communications setup<br />

assume that <strong>the</strong> channels from sources to destination and<br />

<strong>the</strong> inter-user channel are block-fading Rayleigh and receiver<br />

noise is assumed to be AWGN with zero mean and<br />

unit variance. Discrete-time baseband received signal at<br />

each receiver and at each sample time k, is<br />

y0[k] =h1,0x1[k]+h2,0x2[k]+n0[k] (3)<br />

y1[k] =h2,1x2[k]+n1[k] (4)<br />

y2[k] =h1,2x1[k]+n2[k] (5)<br />

y0[k], y1[k] and y2[k] are <strong>the</strong> received signals at destination,<br />

users u1 and u2 respectively. xi is <strong>the</strong> encoded<br />

message transmitted by <strong>the</strong> user ui, wherei =1,2. ni<br />

represent <strong>the</strong> Gaussian noise at <strong>the</strong> receivers with variance<br />

No. Fading coefficients between transmitter i and<br />

receiver j are represented by hi,j. The signal-to-noise<br />

ratio (SNR) is defined as:<br />

si,j = |hi,j| 2 P 2 i<br />

No<br />

(6)


In this model, we assume perfect echo cancellation on<br />

<strong>the</strong> basis <strong>of</strong> <strong>the</strong> fact that each user knows its own signal,<br />

<strong>the</strong>refore it can cancel out its own message’s effect on<br />

<strong>the</strong>ir received signal.<br />

Channel phase information availability at <strong>the</strong> transmitter<br />

side is assumed. With <strong>the</strong> help <strong>of</strong> this phase information,<br />

transmitters can adjust <strong>the</strong> phase <strong>of</strong> <strong>the</strong> signal at <strong>the</strong><br />

transmitting side which results in coherent combination<br />

<strong>of</strong> <strong>the</strong> received signal at <strong>the</strong> destination.<br />

Fur<strong>the</strong>rmore, for practical reasons we impose half duplex<br />

constraint on <strong>the</strong> channel that both <strong>of</strong> <strong>the</strong> users can<br />

not send or receive at <strong>the</strong> same time. Their channel assignment<br />

looks like as in Figure 2. An average trans-<br />

u1 Transmits x1<br />

u2 Listens<br />

Destination Listens<br />

u2 Transmits x2<br />

u1 Listens<br />

Destination Listens<br />

time<br />

u1 Transmits r<br />

u2 Transmits r<br />

Destination Listens<br />

Figure 2: Proposed orthogonal cooperative communications<br />

protocol<br />

mission power constraint <strong>of</strong> Pi is imposed on <strong>the</strong> nodes,<br />

where i = 1, 2. For simplicity, we assume that all <strong>the</strong> terminals,<br />

users u1, u2 and destination are at <strong>the</strong> same distance<br />

d. Fur<strong>the</strong>rmore, we assume perfect channel knowledge<br />

at <strong>the</strong> receivers.<br />

III. NETWORK CODING FOR COOPERATIVE<br />

COMMUNICATIONS<br />

For our two user cooperative communications setup, first<br />

user u1 broadcasts its fresh information in first time slot<br />

which is received both by ano<strong>the</strong>r user and <strong>the</strong> destination.<br />

Same thing is done by user u2 in <strong>the</strong> second time<br />

slot. Finally, in <strong>the</strong> third time slot both <strong>of</strong> <strong>the</strong> users send<br />

refinement information part <strong>of</strong> <strong>the</strong>ir message which is<br />

same for both <strong>of</strong> <strong>the</strong> users. This message is formed by<br />

taking exclusive or (XOR) operation <strong>of</strong> users’ own message<br />

with <strong>the</strong> message received from o<strong>the</strong>r cooperating<br />

user. User power Pi is divided between <strong>the</strong>ir first phase<br />

<strong>of</strong> fresh information and second phase <strong>of</strong> XOR operation,<br />

appropriately.<br />

At <strong>the</strong> receiver side, it detects and saves <strong>the</strong> messages<br />

received by both <strong>of</strong> <strong>the</strong> users during <strong>the</strong>ir fresh information<br />

transmission phase. In <strong>the</strong> refinement phase <strong>of</strong><br />

transmission by users, it gets <strong>the</strong> XOR copy <strong>of</strong> both <strong>of</strong><br />

<strong>the</strong> messages and uses this to get <strong>the</strong> second copy <strong>of</strong> <strong>the</strong><br />

message for each user as follows:<br />

ˆx1 = r ⊕ ˆx2<br />

ˆx2 = r ⊕ ˆx1<br />

IV. OUTAGE PROBABILITY<br />

In block fading case, channels considered follow a<br />

Rayleigh distribution, but remain constant for some specific<br />

time and than <strong>the</strong>y change. As a result <strong>the</strong> achievable<br />

rates change accordingly. Therefore it is difficult to<br />

(7)<br />

(8)<br />

106<br />

maintain constant rate, in this situation outage probability<br />

is better performance measure.<br />

Outage probability is defined as <strong>the</strong> probability that<br />

asystemrateR falls bellow a certain required rate R ′ .<br />

By using cooperative communications concept with network<br />

coding, we not only have capacity gain but, diversity<br />

gain is also in <strong>the</strong> list <strong>of</strong> potential gains provided by<br />

this promising scheme. For <strong>the</strong> reason <strong>of</strong> space constraint<br />

we only show <strong>the</strong> results for diversity gain as shown in<br />

Figure 3. As can be seen from <strong>the</strong> figure that with coop-<br />

Outage probability<br />

10 0<br />

10 −1<br />

10 −2<br />

10<br />

0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8<br />

−3<br />

Cooperative<br />

Non−Cooperative<br />

Required rate R’<br />

Figure 3: Outage probability comparison at 8 dB SNR<br />

erative communications we get 25 percent reduction in<br />

outage as compared to non-cooperative communications.<br />

For <strong>the</strong>se results we have taken <strong>the</strong> inter-user channels<br />

as well as user destination channel operating at <strong>the</strong> same<br />

SNR level <strong>of</strong> 8 dB.<br />

V. CONCLUSION<br />

We proposed a simple XOR-based network code for 3phase<br />

orthogonal cooperative communications protocol.<br />

The simulation results indicate that we benefit from a<br />

22% increase in rate for outage probability <strong>of</strong> 10−2 .<br />

There is 25% <strong>of</strong> reduction in outage at <strong>the</strong> rate <strong>of</strong> 0.55<br />

bits per channel use, where inter-user and user destination<br />

channels are operating at SNR <strong>of</strong> 8 dB.<br />

REFERENCES<br />

[1] A. Sendonaris, E. Erkip, and B. Aazhang. User cooperation<br />

diversity-Part I: System description. IEEE Trans. Commun.,<br />

51(11):1927–1938, Nov 2003.<br />

[2] A. Sendonaris, E. Erkip, and B. Aazhang. User cooperation<br />

diversity-Part II: Implementation aspects and performance analysis.<br />

IEEE Trans. Commun., 51(11):1939–1948, Nov 2003.<br />

[3] R. Ahlswede, N. Cai, S.-Y. R. Li, and R. W. Yeung. Network<br />

information flow. IEEE Transactions on Information Theory,<br />

46(4):1204–1216, July 2000.<br />

[4] J. N. Laneman. Cooperative Diversity in Wireless Networks: Algorithms<br />

and Architectures. PhD <strong>the</strong>sis, Massachusetts Institute <strong>of</strong><br />

Technology, Cambridge, MA, Aug 2002.


Graphics versus spoken language in pedestrian navigation<br />

Felix Ortag and Georg Gartner (Faculty Mentor)<br />

Institute for Geoinformation and Cartography<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {felix.ortag,georg.gartner}@tuwien.ac.at<br />

Abstract — Today’s navigations systems for pedestrians<br />

are mainly based on car navigation. Different<br />

needs <strong>of</strong> car drivers and pedestrians are in part<br />

obvious. Therefore it seems to be useful to do investigations<br />

which might help to improve pedestrian<br />

navigation systems.<br />

The abstracted diploma <strong>the</strong>sis compares two ways<br />

<strong>of</strong> communication between <strong>the</strong> system and its user:<br />

spoken instructions and graphic maps – each as<br />

single method. Which guides <strong>the</strong> user »better« to his<br />

destination? Where does he feel safer? Which<br />

would he prefer? How does <strong>the</strong> mental map differ?<br />

These questions were investigated for indoor and<br />

outdoor use in an empirical test with 31 test persons<br />

who had to walk on test routes using a prototype<br />

navigation system.<br />

I. INTRODUCTION<br />

The main focus <strong>of</strong> cartography lies on graphic representations<br />

<strong>of</strong> <strong>the</strong> real world. O<strong>the</strong>r types <strong>of</strong> media are<br />

mostly considered as an additional improvement, but<br />

not as a replacement. This is because <strong>of</strong> <strong>the</strong> outstanding<br />

qualities <strong>of</strong> <strong>the</strong> visual media in communicating<br />

complex context and large amounts <strong>of</strong> information<br />

in a quick way [1]. [2] mentions interesting<br />

numbers on this: humans need 1.5 seconds to decode<br />

an information which was decoded verbally. In contrast<br />

to this it is possible to process up to 17 information<br />

units in 0.2 seconds when <strong>the</strong>y are transported<br />

on <strong>the</strong> visual channel without verbal coding.<br />

However, this master <strong>the</strong>sis questions that a user in<br />

<strong>the</strong> pedestrian navigation situation needs to know so<br />

much about <strong>the</strong> area surrounding him. From a very<br />

pragmatic point <strong>of</strong> view one could argue that <strong>the</strong><br />

usability <strong>of</strong> a pedestrian navigation system is as important<br />

as <strong>the</strong> special knowledge <strong>the</strong> user derives.<br />

Today <strong>the</strong>re are already navigation systems for pedestrians<br />

sold commercially, but <strong>the</strong>y are <strong>of</strong>ten car<br />

navigation systems with only slight or even no adaptations<br />

to <strong>the</strong> usage situation <strong>of</strong> a person moving by<br />

foot [3, 4]. One <strong>of</strong> <strong>the</strong> many improvements for pedestrians<br />

one could think <strong>of</strong> is a hands free voice only<br />

navigation system. The main idea <strong>of</strong> this master<br />

<strong>the</strong>sis was to prove this idea with empirical means.<br />

107<br />

II. THESES<br />

After literature research it was decided to test <strong>the</strong><br />

following <strong>the</strong>ses for two presentation forms: a map<br />

on a PDA and verbal instructions supplied via an<br />

earphone:<br />

� Subjective navigation quality. The user’s feeling<br />

<strong>of</strong> being on route or <strong>of</strong>f route is independent <strong>of</strong><br />

<strong>the</strong> presentation form.<br />

� Objective navigation quality. There is no difference<br />

in measurable parameters like walking<br />

speed and number <strong>of</strong> errors.<br />

� User acceptance. People prefer <strong>the</strong> system with<br />

auditory instructions.<br />

� Mental map. The mental map – <strong>the</strong> “map in <strong>the</strong><br />

user’s head” – is better concerning topology and<br />

completeness when <strong>the</strong> map is used as presentation<br />

form.<br />

III. EMPIRICAL TEST<br />

In <strong>the</strong> empirical test 31 test persons had to use two<br />

different navigation systems: a PDA displaying a<br />

map and a route to go and a wireless earphone which<br />

gave verbal instructions on how to proceed. The used<br />

devices are shown in Figure 1. Each test person had<br />

to do four test routes, two <strong>of</strong> <strong>the</strong>m outdoor and <strong>the</strong><br />

o<strong>the</strong>r two indoor. In each environment <strong>the</strong> graphic<br />

presentation form was used on one route and <strong>the</strong><br />

verbal one on <strong>the</strong> o<strong>the</strong>r route. The system was not a<br />

real navigation system with working positioning<br />

sensors but a simulation that only worked on <strong>the</strong><br />

selected test routes and with <strong>the</strong> investigator accompanying<br />

<strong>the</strong> test person.<br />

Figure 1: Devices used for <strong>the</strong> test: a tablet PC (for<br />

<strong>the</strong> investigator), a PDA and a Bluetooth earphone.


After doing <strong>the</strong> navigation tasks <strong>the</strong> test persons<br />

were interviewed with a short questionnaire and had<br />

to draw sketch maps <strong>of</strong> <strong>the</strong> outdoor routes.<br />

IV. RESULTS<br />

Among <strong>the</strong> numerous results <strong>the</strong>re were two interesting<br />

significant ones: Women were slower than men<br />

when graphic route presentations (maps) were used<br />

on <strong>the</strong> outdoor routes. In comparison, this difference<br />

did not appear when spoken instructions were used.<br />

The results from <strong>the</strong> interviews support this, as<br />

women overall had a more positive attitude towards<br />

spoken language than men.<br />

Ano<strong>the</strong>r result is that for indoor use <strong>the</strong> test persons<br />

walked significantly faster when <strong>the</strong>y were<br />

guided by voice. Here again <strong>the</strong> results from <strong>the</strong><br />

interview support this finding.<br />

Concerning <strong>the</strong> <strong>the</strong>ses mentioned earlier <strong>the</strong> following<br />

was found out:<br />

� Subjective navigation quality. Basically, this<br />

<strong>the</strong>sis was supported by <strong>the</strong> interview results, but<br />

<strong>the</strong>re seem to be differences between indoor and<br />

outdoor usage and between <strong>the</strong> genders.<br />

� Objective navigation quality. Concerning walking<br />

speed and number <strong>of</strong> errors <strong>the</strong>re were no<br />

significant differences between <strong>the</strong> two presentation<br />

forms.<br />

� User acceptance. The preference for <strong>the</strong> audio<br />

system seems to be true for indoor routes. For<br />

outdoor routes we were not able to show this.<br />

The idea that <strong>the</strong> preference is individually different<br />

seems promising.<br />

� Mental map. Surprisingly <strong>the</strong>re was no difference<br />

in <strong>the</strong> quality <strong>of</strong> sketch maps in conjunction<br />

with <strong>the</strong> presentation form.<br />

V. CONCLUSION<br />

Maps are powerful means to communicate spatial<br />

information and this work – <strong>of</strong> course – did not<br />

change this fact. But it shows that for <strong>the</strong> usage situation<br />

<strong>of</strong> pedestrian navigation a map is not <strong>the</strong> only<br />

possibility to present a route ant that spoken language<br />

can be a real alternative. Some users even prefer it.<br />

The <strong>of</strong>ten discussed and claimed user adaptivity <strong>of</strong><br />

location based services should also include <strong>the</strong> possibility<br />

to choose <strong>the</strong> presentation form.<br />

Ano<strong>the</strong>r open field for fur<strong>the</strong>r research are <strong>the</strong> verbal<br />

instructions itself. As <strong>the</strong>re is a long tradition <strong>of</strong><br />

making maps <strong>the</strong>re are not that many open questions<br />

concerning <strong>the</strong>ir usage in pedestrian navigation systems,<br />

but for spoken instructions <strong>the</strong> opposite is true.<br />

Does <strong>the</strong> user prefer short statements “to <strong>the</strong> point”<br />

108<br />

or longer descriptions with a certain grade <strong>of</strong> redundancy?<br />

Is it important to pronounce street names?<br />

Aside <strong>of</strong> this considerations which mainly focus<br />

on <strong>the</strong> user’s needs, one could also investigate <strong>the</strong><br />

economic and financial aspects <strong>of</strong> <strong>the</strong> topic. Systems<br />

with spoken instructions could be made cheaper and<br />

automated as <strong>the</strong>re is no cartographic work needed.<br />

REFERENCES<br />

[1] R. Däßler. Visuelle Kommunikation mit Karten.<br />

In Visuelle Medienkompetenz. Fachbereich Informationswissenschaften,<br />

FH Potsdam, 2002.<br />

[2] G. Buziek. Dynamic Elements <strong>of</strong> Multimedia<br />

Cartography. In: W. Cartwright, M. P. Peterson<br />

and G. Gartner (Eds.): Multimedia Cartography,<br />

pages 231–245, Springer-Verlag, Berlin, Heidelberg,<br />

1999.<br />

[3] G. Gartner, A. Frank and G. Retscher. Pedestrian<br />

Navigation System in Mixed Indoor/Outdoor<br />

Environment – The Navio Project. In CORP<br />

2004 & Geomultimedia04 <strong>Proceedings</strong>, pages<br />

165–171, 2004.<br />

[4] F. Heidmann and F. Hermann. Benutzerzentrierte<br />

Visualisierung raumbezogener Informationen<br />

für ultraportable mobile Systeme. In Visualisierung<br />

und Erschließung von Geodaten. Beiträge<br />

des Seminars GEOVIS 2003, Kartographische<br />

Schriften, Band 7, pages 121–131, Deutsche Gesellschaft<br />

für Kartographie, Hannover, 2003.


Hardware Acceleration <strong>of</strong> Protocol Identification<br />

Petr Kobiersk´y and Jan Koˇrenek (Faculty Mentor)<br />

Faculty <strong>of</strong> Information Technology<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: {ikobier,korenek}@fit.vutbr.cz<br />

Abstract — A dynamic grow <strong>of</strong> computer networks encourage<br />

rapid development <strong>of</strong> network applications and services.<br />

For providing sufficient network service quality it is important<br />

to shape network flows based on <strong>the</strong>ir application protocol type.<br />

This work analyzes current best protocol identification methods<br />

for use on multi-gigabit networks. Based on analysis, <strong>the</strong> hardware<br />

architecture which accelerates computationally intensive<br />

algorithms parts is proposed. Our solution is able to work on<br />

10 Gbps networks and export application protocol type using<br />

Netflow protocol.<br />

I. INTRODUCTION<br />

Internet is becoming used by more and more applications<br />

and services. Network has to guarantee different levels<br />

<strong>of</strong> quality for each network protocol. For this tasks it<br />

is necessary to identify application protocol in network<br />

flows which is seen, with respect to different protocol<br />

types (ciphered, corporate standards etc.), as a serious<br />

problem.<br />

Standard traffic identification methods associate <strong>the</strong><br />

observed traffic with application according to <strong>the</strong> TCP or<br />

UDP port numbers using IANA list <strong>of</strong> registered ports.<br />

To avoid detection, several applications have begun to<br />

use dynamic port numbers and well-known ports commonly<br />

used by protocols, such as HTTP or SMTP so<br />

new detection methods had to be utilized. Current protocol<br />

identification techniques are based on payload analysis,<br />

but <strong>the</strong>y can be circumvented using variable length<br />

padding and protocol encryption. In recent years, researchers<br />

solved some <strong>of</strong> this issues using statistic based<br />

protocol identification methods [1]. These methods are<br />

able to identify applications with <strong>the</strong> accuracy <strong>of</strong> about<br />

90 – 100 % using statistical information ga<strong>the</strong>red from<br />

network flows.<br />

Current protocol identification methods are not perfect<br />

and each method is suitable for detection <strong>of</strong> different protocol<br />

types. Using a combination <strong>of</strong> <strong>the</strong>se methods is<br />

possible to get better protocol identification accuracy for<br />

all types <strong>of</strong> application protocols. Unfortunately, operations<br />

which have to be performed during protocol identification<br />

are computationally intensive and performance<br />

<strong>of</strong> <strong>the</strong>ir s<strong>of</strong>tware implementation is not sufficient for deployment<br />

on current networks and dedicated accelerated<br />

implementations are required.<br />

This work analyzes computational intensity <strong>of</strong> application<br />

protocol identification operations and allocates<br />

109<br />

<strong>the</strong>m on hardware and s<strong>of</strong>tware resources. Based on <strong>the</strong><br />

allocation, selected operations are mapped on hardware<br />

resources with intention to accelerate current best protocol<br />

identification methods to 10 Gbps networks.<br />

II. PROTOCOL IDENTIFICATION MODEL<br />

Based on <strong>the</strong> knowledge <strong>of</strong> current application protocol<br />

identification methods <strong>the</strong> general model was created.<br />

Proposed model shown in Figure 1 covers all necessary<br />

operations for accurate protocol identification using currently<br />

best identification methods. The model was build<br />

to analyze computational intensity for individual operations<br />

<strong>of</strong> protocol identification. The model consists <strong>of</strong><br />

several cooperating processes which are responsible for<br />

L2-L4 protocol layers processing, holding per-flow information,<br />

L3-L4 protocol reassembling and protocol identification.<br />

Input Interface<br />

L2 – L7<br />

Layers<br />

Identification<br />

Out-order<br />

packet<br />

storage<br />

Pattern Matching<br />

Header<br />

Field<br />

Extraction<br />

L3 and L4 Protocol<br />

Reassembling<br />

Flow Statistic<br />

Computation<br />

Classification<br />

Network<br />

Flow<br />

Identification<br />

Collision<br />

Network Flows<br />

Context<br />

Information<br />

Keeping<br />

Context<br />

Information<br />

Update<br />

Packets<br />

Application Data<br />

Headers<br />

Flow Context<br />

Collision<br />

Flow ID<br />

Position <strong>of</strong> L2-L7<br />

Figure 1: Model <strong>of</strong> application protocol identification<br />

Computationally intensive operations which limit usage<br />

<strong>of</strong> current identification methods to networks up to<br />

1 Gbps were identified based on performed model simulations.<br />

The most critical operation was regular expression<br />

matching with s<strong>of</strong>tware throughput about 18 Mbps<br />

for 200 regular expressions. O<strong>the</strong>r necessary operations<br />

have sufficient throughput for 1 Gbps networks but not<br />

for 10 Gbps networks. The results <strong>of</strong> performed simulations<br />

can be seen in Table 1.<br />

Output Interface


Operation Throughput 10GBase-TX<br />

L2-L4 parsing 21 Mp/s<br />

Reassembling 2.5 Mp/s<br />

Statistics computation 3 Mp/s<br />

Pattern matching 18 Mb/s<br />

Protocol classification 1 Mflows/s<br />

Table 1: S<strong>of</strong>tware throughput <strong>of</strong> selected operations<br />

III. PROPOSED ARCHITECTURE<br />

Hardware and s<strong>of</strong>tware architecture <strong>of</strong> network probe<br />

were proposed based on <strong>the</strong> results <strong>of</strong> model simulations.<br />

This architecture consists <strong>of</strong> a hardware acceleration card<br />

and a host computer. The accelerator performs packet<br />

header analysis, per-flow statistic collection and regular<br />

expression matching. Information about flows (statistics,<br />

matched regular expressions) is exported via PCI bus for<br />

s<strong>of</strong>tware processing. S<strong>of</strong>tware performs classification<br />

task and export information about flows and identified<br />

application protocols using Netflow v9 protocol. This<br />

standardized protocol can be collected by active network<br />

devices which can perform appropriate actions like traffic<br />

shaping and blocking.<br />

The COMBO6 1 card equipped with Xilinx Virtex II<br />

FPGA is used as a hardware accelerator. The firmware<br />

architecture comes out from existing Netflow probe design<br />

[2] with several modifications. For enabling statistical<br />

protocol identification, suitable statistics (with respect<br />

to hardware realization and protocol identification<br />

accuracy) were identified. The firmware architecture was<br />

adapted to be able to collect <strong>the</strong>se statistics on network<br />

flows. The firmware was also extended with powerful<br />

pattern match unit which enables <strong>the</strong> usage <strong>of</strong> signature<br />

based protocol identification methods. The final<br />

firmware architecture is shown in Figure 2.<br />

NetCOPE<br />

Input<br />

Buffers<br />

Sampling<br />

HFE<br />

HFE<br />

HFE<br />

Hash<br />

Generator<br />

External<br />

Memory<br />

Flow state<br />

manager<br />

Pattern<br />

Match<br />

new<br />

External<br />

Memory<br />

Flow<br />

context<br />

block<br />

FPU<br />

FPU<br />

FPU<br />

modified<br />

SW<br />

Buffer<br />

Figure 2: Hardware accelerator firmware architecture<br />

1 http://www.liberouter.org<br />

110<br />

Regular expressions from open-source project L7filter<br />

were used for signature based protocol identification.<br />

Most <strong>of</strong> <strong>the</strong>se signatures were transformed into<br />

NFA representation which can be easily mapped to<br />

FPGA logic and later used to pattern matching [3]. Before<br />

conversion to VHDL NFAs are transformed to lazy<br />

NFA representation which enable processing more characters<br />

in one transition. To save hardware resources <strong>the</strong><br />

lazy NFAs are also determinized and minimized. Thanks<br />

to <strong>the</strong>se optimization a throughput over 10 Gbps was<br />

achieved.<br />

The application protocols are identified based on extracted<br />

attributes using s<strong>of</strong>tware classifier. The decision<br />

tree based classifier was selected after evaluation <strong>of</strong> several<br />

classification algorithms because <strong>of</strong> high throughput<br />

and ability to work with very large training set. Classifier<br />

was evaluated using set <strong>of</strong> pre-classified network<br />

traffic from Cambridge University. For very small set <strong>of</strong><br />

attributes (up to 10 attributes) was able to archive 95 %<br />

accuracy <strong>of</strong> application protocol identification.<br />

IV. CONCLUSION AND FUTURE RESEARCH<br />

The new protocol identification model with respect to<br />

mapping operations between s<strong>of</strong>tware and hardware was<br />

created. A new architecture <strong>of</strong> protocol identification<br />

probe has been proposed based on <strong>the</strong> model. Presented<br />

architecture combines statistic and signature based identification<br />

methods to archive better detection accuracy<br />

and it is also able to archive throughput over 10 Gbps.<br />

The implemented pattern match unit accelerates regular<br />

expression matching more than 400 times compared to<br />

s<strong>of</strong>tware solutions.<br />

The future work will put focus mainly on hardware<br />

probe evaluation with various traffic samples. We also<br />

plan to analyze particular application protocols (mainly<br />

P2P networks) in detail and find suitable classification<br />

attributes and methods for <strong>the</strong>m.<br />

REFERENCES<br />

[1] Andrew W. Moore and Denis Zuev. Internet traffic<br />

classification using bayesian analysis techniques. In<br />

<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 2005 ACM SIGMETRICS international<br />

conference, volume 33, pages 50–60. ACM<br />

Press, June 2005.<br />

[2] Petr Kobiersk´y Martin ˇZádník, Jan Koˇrenek and<br />

Ondˇrej Lengál. Network probe for flexible flow<br />

monitoring. In <strong>Proceedings</strong> <strong>of</strong> IEEE Design and Diagnostics<br />

<strong>of</strong> Electronic Circuits and Systems, pages<br />

213–218, <strong>2008</strong>.<br />

[3] Jan Koˇrenek and Petr Kobiersk´y. Intrusion detection<br />

system intended for multigigabit networks. In <strong>Proceedings</strong><br />

<strong>of</strong> IEEE Design and Diagnostics <strong>of</strong> Electronic<br />

Circuits and Systems, pages 361–364, 2007.


Evolutionary Design <strong>of</strong> Wormhole Switched Collective Communications<br />

Jiri Jaros and Josef Schwarz (Faculty Mentor)<br />

Faculty <strong>of</strong> Information Technology<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: {jarosjir,schwarz}@fit.vutbr.cz<br />

Abstract — In this paper, we describe an evolutionary<br />

technique aimed at scheduling collective<br />

communications on interconnection networks <strong>of</strong><br />

parallel computers. To avoid contention for links<br />

and associated delays, collective communications<br />

proceed in synchronized steps. Minimum number <strong>of</strong><br />

steps is sought for <strong>the</strong> given network topology,<br />

wormhole (pipelined) switching, minimum routing<br />

and given sets <strong>of</strong> sender and/or receiver nodes.<br />

Used technique is able not only re-invent optimum<br />

schedules for known symmetric topologies like<br />

hyper-cubes, but it can find schedules even for any<br />

asymmetric or irregular topologies in case <strong>of</strong> general<br />

many-to-many collective communications.<br />

I. INTRODUCTION<br />

At <strong>the</strong> present time, multiprocessor systems are more<br />

frequently found not only in high-end servers and<br />

workstations, but also in small-scale parallel systems<br />

for high performance control, data acquisition and<br />

analysis, image processing, networking processors,<br />

wireless communication, and game computers. The<br />

importance <strong>of</strong> communication among CPU cores,<br />

processors and computers and <strong>of</strong> related interconnection<br />

networks is recently steadily growing. Whereas<br />

<strong>the</strong> lower bounds on time complexity <strong>of</strong> various<br />

collective communications (CC) can be ma<strong>the</strong>matically<br />

derived for any network topology and <strong>the</strong> given<br />

communication pattern, finding a corresponding<br />

schedule <strong>of</strong> communication is much more difficult,<br />

and in some cases it is not known as yet. The rest <strong>of</strong><br />

<strong>the</strong> paper addresses <strong>the</strong> quest for an optimal communication<br />

schedule based on evolutionary algorithms,<br />

provided that network topology and communication<br />

patterns are given.<br />

II. MODEL OF COMMUNICATION<br />

Any CC can be seen as a set <strong>of</strong> point-to-point communications.<br />

The CC scheduling problem can be<br />

simply described as partitioning this set into as few<br />

subsets as possible that follow one ano<strong>the</strong>r in sequence<br />

<strong>of</strong> synchronized steps, and all communications<br />

in one subset proceed in parallel.<br />

111<br />

The main goal is to avoid any conflicts in shared<br />

resources – links (channels). Several messages between<br />

source-destination pairs, not necessarily <strong>the</strong><br />

neighbours, can proceed concurrently, and can be<br />

combined into a single subset if <strong>the</strong>ir paths are linkdisjoint.<br />

If <strong>the</strong> source and destination nodes are not<br />

adjacent, <strong>the</strong> messages go via some intermediate<br />

nodes, but processors in <strong>the</strong>se nodes are not aware <strong>of</strong><br />

it; <strong>the</strong> messages are routed automatically by <strong>the</strong><br />

routers attached to processors.<br />

To achieve low power consumption and chip area<br />

requirements, <strong>the</strong> wormhole switching and source<br />

based routing have been employed [1].<br />

Four collective communication patterns were examined<br />

for <strong>the</strong>ir high time complexity and impact on<br />

<strong>the</strong> performance degradation. They were one-to-all<br />

broadcast (OAB), one-to-all scatter (OAS), all-to-all<br />

broadcast (AAB) and all-to-all scatter (AAS).<br />

III. OPTIMIZATION ALGORITHM<br />

The proposed technique makes pr<strong>of</strong>its from using<br />

evolutionary algorithms to relax <strong>the</strong> limitation <strong>of</strong><br />

currently known heuristic techniques (adaptive routing,<br />

graph colouring, recursive division or conventional<br />

human design) [3].<br />

A. PRE-PROCESSING PHASE<br />

An input data structure stores a topology description,<br />

a definition <strong>of</strong> CC and subsets <strong>of</strong> communicating<br />

nodes. After an input file is loaded, <strong>the</strong> data have to<br />

be pre-processed. The pre-processor takes a topology<br />

and finds all paths (shortest ones in <strong>the</strong> case <strong>of</strong> minimal<br />

routing) between all source-destination node<br />

pairs and stores <strong>the</strong>m into a special data structure<br />

employing <strong>the</strong> modified Dijkstra’s algorithm.<br />

B. SOLUTION ENCODING<br />

As broadcast and scatter CC are completely different<br />

communication services, candidate solutions are<br />

encoded in separate ways.<br />

The OAS chromosome contains P genes equals to<br />

number <strong>of</strong> communicating nodes; each gene consists<br />

<strong>of</strong> two items: an index <strong>of</strong> one <strong>of</strong> <strong>the</strong> shortest source-


destination paths, and a communication step number.<br />

An index <strong>of</strong> gene represents a destination node for<br />

<strong>the</strong> scattered message.<br />

In <strong>the</strong> case <strong>of</strong> OAB, an indirect encoding has been<br />

used; a chromosome does not include a broadcast<br />

tree, but only instructions how to create it. Each<br />

chromosome consists <strong>of</strong> P genes, one for each destination<br />

node. Individual genes are composed <strong>of</strong> three<br />

items: a source node index, <strong>the</strong> shortest path index,<br />

and a step number.<br />

An AAB/AAS chromosome is created as by extending<br />

<strong>the</strong> vector to a matrix, each row <strong>of</strong> which<br />

corresponds to one <strong>of</strong> OAB/OAS communications.<br />

C. FITNESS FUNCTION DEFINITION<br />

The fitness function is based on counting conflicts<br />

between all point-to-point communications realized<br />

in <strong>the</strong> same steps. The numbers <strong>of</strong> conflicts from<br />

particular steps are summarized. The valid communication<br />

schedule for a given number <strong>of</strong> communication<br />

steps must be conflict-free, i.e. its fitness value<br />

has to reach zero. Valid schedules are ei<strong>the</strong>r optimal<br />

(<strong>the</strong> number <strong>of</strong> steps equals <strong>the</strong> lower bound) or<br />

suboptimal. In practice, we start with higher number<br />

<strong>of</strong> possible steps and if <strong>the</strong> EA is successful <strong>the</strong>n<br />

decrease this value.<br />

D. ACCELERATION HEURISTIC<br />

To improve <strong>the</strong> searching abilities and speed-up <strong>the</strong><br />

convergence, <strong>the</strong> acceleration heuristic has been<br />

utilized. It is based on search space pruning taking<br />

into account:<br />

− a search space restriction due to a limited message<br />

injection capability <strong>of</strong> network nodes<br />

− injection <strong>of</strong> promising building blocks into initial<br />

population<br />

− guided broadcast tree construction<br />

− guided selection <strong>of</strong> message source and utilized<br />

path<br />

IV. EXPERIMENTAL RESULTS<br />

The proposed technique has been verified on well<br />

known hypercube topology, where <strong>the</strong> best known<br />

schedules have been reinvented again (Ho-Kao or<br />

double-tree schedule)[3]. Optimal schedules have<br />

been founded also for o<strong>the</strong>r easily fabricated networks<br />

like 2D mesh, ring, octagon, K-ring, midimew,<br />

ladder, etc. [5]. The special diameter-degree<br />

networks have been examined for <strong>the</strong>ir compactness,<br />

and <strong>the</strong> results have been published in [1].<br />

Moreover, <strong>the</strong> fault-tolerance <strong>of</strong> special networks<br />

has been investigated under a single link or node<br />

fault [2]. The proposed technique has been also ap-<br />

112<br />

plied to fat topologies and irregular topologies like<br />

fat octagon, coated mesh or AMP [3].<br />

Finally, more complex communication services<br />

with overlapping subsets <strong>of</strong> senders and receivers<br />

(many-to-many communications), never solved before,<br />

have been explored and <strong>the</strong>ir results published<br />

in [4].<br />

Last but not least, <strong>the</strong> proposed method was declared<br />

as Human competitive at GECCO 2007 [6].<br />

V. CONCLUSION<br />

Since many <strong>of</strong> optimal CC schedules are still not<br />

known, <strong>the</strong> proposed technique can speed-up many<br />

distributed systems. Moreover, it can be employed to<br />

repair suffered systems by isolating and deactivating<br />

<strong>of</strong> damaged blocks, reloading <strong>of</strong> corrected communication<br />

schedules and restarting <strong>the</strong> system.<br />

ACKNOWLEDGMENTS<br />

This research has been carried out under <strong>the</strong> financial<br />

support <strong>of</strong> <strong>the</strong> research grant “Security-Oriented<br />

Research in Information Technology”, MSM<br />

0021630528 (2007-13).<br />

REFERENCES<br />

[1] J. Jaros and V. Dvorak. An Evolutionary Design<br />

Technique for Collective Communications on<br />

Optimal Diameter-Degree Networks. In <strong>Proceedings</strong><br />

<strong>of</strong> <strong>the</strong> <strong>2008</strong> Genetic and Evolutionary Computational<br />

<strong>Conference</strong> GECCO, pages 1539-<br />

1546, New York, US, ACM, <strong>2008</strong>.<br />

[2] J. Jaros. Evolutionary Design <strong>of</strong> Fault Tolerant<br />

Collective Communications. In Proceeding <strong>of</strong><br />

8th International <strong>Conference</strong>, ICES <strong>2008</strong>, pages<br />

261-272, Berlin, DE, Springer, <strong>2008</strong>.<br />

[3] J. Jaros, M. Ohlidal and V. Dvorak. An Evolutionary<br />

Approach to Collective Communication<br />

Scheduling. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 2007 Genetic<br />

and Evolutionary Computational <strong>Conference</strong>,<br />

pages 2037-2044, New York, US, ACM, 2007.<br />

[4] J. Jaros, M. Ohlidal and V. Dvorak. Complexity<br />

<strong>of</strong> Collective Communications on NoCs. In <strong>Proceedings</strong><br />

<strong>of</strong> <strong>the</strong> 5th International Symposium on<br />

Parallel Computing in Electrical Engineering,<br />

pages 127-132, Los Alamitos, CA 90720-1314,<br />

US, IEEE CS, 2006.<br />

[5] V. Dvorak, J. Jaros and M. Ohlidal. Optimum<br />

Topology-Aware Scheduling <strong>of</strong> Collective Communications.<br />

In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> Sixth International<br />

<strong>Conference</strong> on Networking, pages 6, New<br />

York, US, IEEE CS, 2007.<br />

[6] The “Hummies” URL http://www.geneticprogramming.org/hc2007/cfe2007.html


Component Based Communication Middleware for AUTOSAR<br />

Dietmar Schreiner 1 ,<br />

Karl M. Göschka 2 (Faculty Mentor), and Jens Knoop 1 (Faculty Mentor)<br />

1 Institute <strong>of</strong> Computer Languages, Compilers and Languages Group<br />

2 Institute <strong>of</strong> Information Systems, Distributed Systems Group<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {schreiner,knoop}@complang.tuwien.ac.at<br />

Email: {k.goeschka}@infosys.tuwien.ac.at<br />

Abstract — Due to market demands and technological<br />

progress automotive electronic systems have become highly<br />

complex, distributed, and heterogeneous systems. In consequence,<br />

costs and time-to-market tend to out-grow spendable<br />

budgets, whereas quality seriously suffers. To overcome this situation,<br />

automotive manufacturers agreed upon a new standard<br />

for <strong>the</strong>ir electronic systems—AUTOSAR. In AUTOSAR applications<br />

are built in conformance to <strong>the</strong> component paradigm by<br />

assembling prefabricated application components that utilize<br />

standardized component middleware. Hence, AUTOSAR compliant<br />

systems provide increased reusability, maintainability<br />

and thus reduced time-to-market and over-all costs. However,<br />

<strong>the</strong> middleware itself is still a coarse grained layered s<strong>of</strong>tware<br />

architecture that is hard to adapt and to optimize w.r.t. application<br />

specific needs. Within this paper we provide an overview<br />

on our work on how to apply <strong>the</strong> component paradigm to <strong>the</strong><br />

component middleware itself, and how to capitalize on this redesigned<br />

middleware architecture by automatically syn<strong>the</strong>sizing<br />

custom-tailored, light-weight middleware that still complies<br />

to <strong>the</strong> AUTOSAR standard.<br />

I. MOTIVATION<br />

Driven by steadily increasing requirements <strong>of</strong> innovative<br />

applications, automotive electronic systems have reached<br />

a level <strong>of</strong> complexity that requires a technological breakthrough<br />

in order to manage <strong>the</strong>m cost efficiently and<br />

at high quality. In contrast to o<strong>the</strong>r domains, like e.g.<br />

avionics, automotive electronic systems are produced in<br />

comparatively large quantities (69,1 million vehicles in<br />

2006). Therefore, <strong>the</strong> price per unit has to be as low as<br />

possible, forcing manufacturers to assemble economical,<br />

and thus resource constrained electronic building blocks.<br />

In consequence, automotive s<strong>of</strong>tware has to cope with<br />

harsh restrictions, especially with limitations <strong>of</strong> available<br />

memory and processing power.<br />

The Automotive Open System Architecture<br />

(AUTOSAR) [1], an upcoming industry standard<br />

within <strong>the</strong> automotive domain, reflects <strong>the</strong>se facts by<br />

constituting Component Based S<strong>of</strong>tware Engineering<br />

(CBSE) [2, 3, 4] as development paradigm for automotive<br />

applications. CBSE is well accepted within <strong>the</strong><br />

embedded systems community, as it provides a clear<br />

separation <strong>of</strong> concerns, and hence facilitates extensive<br />

s<strong>of</strong>tware reuse. In AUTOSAR, application concerns are<br />

113<br />

Application<br />

Layer<br />

« component »<br />

AUTOSAR SWC<br />

AUTOSAR compliant BSW Interfaces<br />

Basic S<strong>of</strong>tware<br />

« component »<br />

COMPASS MWC<br />

RTE<br />

« component »<br />

AUTOSAR SWC<br />

« component »<br />

COMPASS MWC<br />

« component »<br />

COMPASS MWC<br />

AUTOSAR<br />

Component Model<br />

COMPASS<br />

Component Model<br />

Figure 1: Component based AUTOSAR middleware<br />

covered by s<strong>of</strong>tware components, while infrastructural<br />

ones are handled within layered component middleware.<br />

This design leads to an increase in application quality,<br />

reusability, and maintainability, and consequently to a<br />

reduction <strong>of</strong> costs and time-to-market. However, <strong>the</strong><br />

AUTOSAR component middleware is specified as layered<br />

s<strong>of</strong>tware that is only customizable on a coarse-grained<br />

level.<br />

In consequence, conventional AUTOSAR middleware<br />

tends to be heavy-weight, which is problematic within<br />

resource constrained automotive embedded systems.<br />

Therefore, our work contributes by applying <strong>the</strong> component<br />

paradigm to AUTOSAR beyond <strong>the</strong> application<br />

layer—to <strong>the</strong> component middleware. AUTOSAR’s conventional,<br />

layered middleware architecture is replaced by<br />

a component based design that completely resembles <strong>the</strong><br />

standard’s functionality, but is more flexible in terms <strong>of</strong><br />

application specific customization.<br />

To capitalize on <strong>the</strong> benefits <strong>of</strong> a component based architecture,<br />

we designed a model driven development process,<br />

that incorporates generation <strong>of</strong> component based<br />

communication middleware. As a result, customtailored,<br />

resource saving middleware can automatically<br />

be syn<strong>the</strong>sized from application models and prefabri-


cated, reusable middleware components.<br />

II. COMPONENT BASED MIDDLEWARE<br />

As adumbrated in Fig. 1, we aim at <strong>the</strong> construction<br />

<strong>of</strong> component based AUTOSAR middleware. This so<br />

called Basic S<strong>of</strong>tware is assembled from standardized<br />

and hence replaceable middleware components, but still<br />

has to provide <strong>the</strong> same interfaces as its conventional<br />

counterpart, to allow seamless integration into existing<br />

environments.<br />

To build component based middleware we developed<br />

a new component model (COMPASS CM) for <strong>the</strong> middleware,<br />

which can not resort to middleware itself. Our<br />

component model specifies a component definition, and<br />

a composition- and interaction-standard. Composition<br />

is based on procedural- and on data-interfaces (function<br />

calls and shared memory); interaction may adhere to<br />

<strong>the</strong> client-server or <strong>the</strong> sender-receiver paradigm. The<br />

type-system <strong>of</strong> COMPASS CM is kept close to that <strong>of</strong><br />

AUTOSAR to ensure full compatibility [5].<br />

III. COMPONENT IDENTIFICATION<br />

After specifying a sound component model for component<br />

based middleware, we identified standardized middleware<br />

building-blocks as much as pared-down versions<br />

for each <strong>of</strong> <strong>the</strong>m. Therefore we investigated a<br />

full-fledged industry implementation <strong>of</strong> a Basic S<strong>of</strong>tware<br />

stack. First we did a manual decomposition based on expert<br />

knowledge. In a second step we developed a contextinsensitive<br />

static analysis that, guided by few manual<br />

source-code annotations, automatically identified a set<br />

<strong>of</strong> valid decompositions containing <strong>the</strong> manually created<br />

one [6, 7].<br />

IV. AUTOMATED MIDDLEWARE SYNTHESIS<br />

One <strong>of</strong> <strong>the</strong> advantages <strong>of</strong> component architectures is <strong>the</strong><br />

ability to easily replace single building-blocks by isomorphic<br />

ones exposing <strong>the</strong> same interfaces. Heavyweight<br />

components may be replaced by pared-down<br />

ones, if <strong>the</strong> application does not require <strong>the</strong> components’<br />

full functionality. Based on this fact, we developed<br />

a model driven process that automatically syn<strong>the</strong>sizes<br />

AUTOSAR middleware from prefabricated middleware<br />

building blocks and application models [8, 9]. The<br />

core <strong>of</strong> <strong>the</strong> defined model driven process is <strong>the</strong> so called<br />

Connector Transformation. It transforms platform independent<br />

models (PIMs), describing <strong>the</strong> application’s<br />

component architecture, into platform specific models<br />

(PSMs), containing application components and component<br />

equivalent middleware structures. The PSMs moreover<br />

render <strong>the</strong> system’s physical structure, so one PSM<br />

is created for each system node.<br />

114<br />

V. RESULTS<br />

To prove <strong>the</strong> proposed methodology, and to assess <strong>the</strong><br />

reduction in size <strong>of</strong> <strong>the</strong> syn<strong>the</strong>sized communication middleware,<br />

we implemented a simple automotive application.<br />

The read-only-memory (ROM) footprint <strong>of</strong> <strong>the</strong> application’s<br />

custom-tailored, syn<strong>the</strong>sized communication<br />

middleware was compared to that <strong>of</strong> <strong>the</strong> conventional<br />

industry implementation to substantiate <strong>the</strong> claimed improvement.<br />

Results showed a reduction <strong>of</strong> <strong>the</strong> middleware’s<br />

ROM footprint by nearly 30% for <strong>the</strong> implemented<br />

application, while a worst-case scenario, where<br />

<strong>the</strong> full AUTOSAR’s functionality was required within<br />

<strong>the</strong> middleware, led to an identical footprint as that <strong>of</strong><br />

<strong>the</strong> conventional one. In addition <strong>the</strong> middleware’s execution<br />

time was reduced by 10% due to omitting forwarding<br />

calls between prescribed layers.<br />

ACKNOWLEDGMENTS<br />

This work has been partially funded by <strong>the</strong> FIT-IT [embedded<br />

systems initiative <strong>of</strong> <strong>the</strong> Austrian Federal Ministry<br />

<strong>of</strong> Transport, Innovation, and Technology] and managed<br />

by Eutema and <strong>the</strong> Austrian Research Agency FFG<br />

within project COMPASS under contract 809444, and by<br />

<strong>the</strong> 7th EU R&D Framework Programme within project<br />

ALL-TIMES under contract No 215068.<br />

REFERENCES<br />

[1] AUTOSAR. Automotive Open System Architecture. http://<br />

www.autosar.org/.<br />

[2] Clemens Szyperski. Component S<strong>of</strong>tware: Beyond Object-<br />

Oriented Programming. Addison-Wesley, January 1998.<br />

[3] Ivica Crnkovic and Magnus Larsson, editors. Building Reliable<br />

Component-Based S<strong>of</strong>tware Systems. Artech House, 2002.<br />

[4] Rob C. van Ommering, Frank van der Linden, Jeff Kramer, and<br />

Jeff Magee. The koala component model for consumer electronics<br />

s<strong>of</strong>tware. IEEE Computer, 33(3):78–85, 2000.<br />

[5] Dietmar Schreiner and Karl M. Göschka. A component model<br />

for <strong>the</strong> AUTOSAR Virtual Function Bus. In COMPSAC ’07: <strong>Proceedings</strong><br />

<strong>of</strong> <strong>the</strong> 31st Annual International Computer S<strong>of</strong>tware and<br />

Applications <strong>Conference</strong>, volume 2, pages 635–641. IEEE, 2007.<br />

[6] Dietmar Schreiner and Karl M. Göschka. Building component<br />

based s<strong>of</strong>tware connectors for communication middleware<br />

in distributed embedded systems. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 2007<br />

ASME/IEEE International <strong>Conference</strong> on Mechatronic and Embedded<br />

Systems and Applications (MESA07), ASME/IEEE, 2007.<br />

CD-ROM.<br />

[7] Dietmar Schreiner, Markus Schordan, Gergö Barany, and Karl M.<br />

Göschka. Source code based component recognition in s<strong>of</strong>tware<br />

stacks for embedded systems. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> <strong>2008</strong><br />

IEEE/ASME International <strong>Conference</strong> on Mechatronic and Embedded<br />

Systems and Applications (MESA08), IEEE. IEEE, <strong>2008</strong>.<br />

to appear.<br />

[8] Dietmar Schreiner and Karl M. Göschka. Syn<strong>the</strong>sizing communication<br />

middleware from explicit connectors in component based<br />

distributed architectures. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 6th International<br />

Symposium on S<strong>of</strong>tware Composition (SC 2007), LNCS. Springer,<br />

2007. to appear.<br />

[9] Dietmar Schreiner and Karl M. Göschka. Explicit connectors in<br />

component based s<strong>of</strong>tware engineering for distributed embedded<br />

systems. In SOFSEM 2007: Theory and Practice <strong>of</strong> Computer Science,<br />

<strong>Proceedings</strong>, volume 4362 <strong>of</strong> LNCS, pages 923–934. LNCS,<br />

Springer, Jan 2007.


The Acquaintance <strong>of</strong> Hardware Timing Effects: A Sine Qua Non to<br />

Validate Temporal Requirements in Embedded Real Time Systems ∗<br />

Sven Bünte and Raimund Kirner (Faculty Mentor)<br />

Real Time Systems Group<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {sven,raimund}@vmars.tuwien.ac.at<br />

Abstract — The validation <strong>of</strong> embedded real time systems not<br />

only has to include <strong>the</strong> investigation <strong>of</strong> temporal behavior <strong>of</strong> <strong>the</strong><br />

s<strong>of</strong>tware but also <strong>of</strong> <strong>the</strong> hardware to a large extent. We present<br />

a measurement-based analysis approach in which delays due to<br />

hardware effects <strong>of</strong> <strong>the</strong> system under test can be captured with<br />

high resolution on <strong>the</strong> one hand and how to relate this information<br />

to a model describing <strong>the</strong> control flow <strong>of</strong> <strong>the</strong> s<strong>of</strong>tware on<br />

<strong>the</strong> o<strong>the</strong>r hand.<br />

I. INTRODUCTION<br />

Embedded Systems are computer systems that are included<br />

in devices that serve a small number <strong>of</strong> dedicated<br />

functions. Today’s life would be quite different without<br />

<strong>the</strong>m: The clock radio in <strong>the</strong> morning, <strong>the</strong> c<strong>of</strong>fee machine<br />

during breakfast, MP3-Players, mobile phones, cars, airplanes<br />

- <strong>the</strong>y all include embedded systems. Modern cars<br />

comprise up to 70 microcontrollers, which is just a fraction<br />

compared to <strong>the</strong> amount <strong>of</strong> what can be found in airplanes<br />

or spacecrafts. There are estimations, that 98% 1<br />

<strong>of</strong> all computer systems are embedded.<br />

For <strong>the</strong> design <strong>of</strong> embedded control s<strong>of</strong>tware, not only<br />

functional requirements have to be adhered to but also<br />

temporal demands. An airbag for instance must be released<br />

in time to be useful. Systems <strong>of</strong> this kind are referred<br />

to as Real Time Systems. The validation <strong>of</strong> real<br />

time systems has to include techniques to assure that temporal<br />

requirements are met. There are basically two major<br />

methodologies: formal verification on <strong>the</strong> one hand<br />

and testing (i.e. performing time measurements) on <strong>the</strong><br />

o<strong>the</strong>r hand. The former can guarantee safe and precise<br />

upper bounds on execution times but is not feasible for<br />

complex hardware or s<strong>of</strong>tware. In contrast, <strong>the</strong> process <strong>of</strong><br />

performing measurements with random test data is easy<br />

to conduct but will most likely miss <strong>the</strong> Worst Case Execution<br />

Time (WCET) for <strong>the</strong>re are too many possible runs<br />

such that each <strong>of</strong> <strong>the</strong>m will be tested eventually.<br />

Our research focuses a hybrid approach in which test<br />

cases are not generated randomly but in a systematic way<br />

∗ The research leading to <strong>the</strong>se results has received funding from <strong>the</strong><br />

Austrian Science Fund (Fonds zur Förderung der wissenschaftlichen<br />

Forschung) within <strong>the</strong> research project Formal Timing Analysis Suite<br />

<strong>of</strong> Real-Time Systems (FORTASRT) under contract P19230-N13.<br />

1 http://www.jimturley.com<br />

115<br />

using formal methods. The ultimate goal is to give <strong>the</strong><br />

s<strong>of</strong>tware developer a testing tool that can provide adequately<br />

precise timing estimations <strong>of</strong> <strong>the</strong> system under<br />

test in an efficient way. The programmer should be able<br />

to press <strong>the</strong> check button before he leaves <strong>the</strong> <strong>of</strong>fice in<br />

order to get <strong>the</strong> results by <strong>the</strong> time he gets back to work<br />

<strong>the</strong> next day.<br />

The following section describes <strong>the</strong> hybrid approach<br />

in more detail and illustrates what our research project is<br />

concerned about. The last section shows one important<br />

part in <strong>the</strong> big picture, namely measuring, analyzing, and<br />

predicting <strong>the</strong> temporal behavior <strong>of</strong> hardware effects.<br />

II. MEASUREMENT BASED TIMING ANALY-<br />

SIS<br />

Figure 1: Data flow within FORTAS<br />

The Formal Timing Analysis Suite (FORTAS) project 2<br />

follows a measurement-based timing analysis approach<br />

based on [1]. We use formal methods like model checking<br />

to examine <strong>the</strong> S<strong>of</strong>tware Under Test (SUT) and to<br />

generate proper test cases which form <strong>the</strong> input for measurements<br />

on <strong>the</strong> target platform.<br />

Figure 1 depicts a top-level view on our framework: an<br />

embedded control program written in C is translated into<br />

an abstract graph representation <strong>of</strong> <strong>the</strong> program’s control<br />

flow structure. In a next step, we use FShell [2] for generating<br />

a set <strong>of</strong> test cases in oder to cover feasible paths<br />

in <strong>the</strong> graph. The model is <strong>the</strong>n iteratively refined by<br />

adding timing information from <strong>the</strong> measurements that<br />

are performed using <strong>the</strong> test cases as input. This enables<br />

2 http://www.fortastic.net


techniques to take timing effects <strong>of</strong> <strong>the</strong> target platform<br />

selectively into account. The refinement loop stops at <strong>the</strong><br />

time <strong>the</strong> s<strong>of</strong>tware developer considers <strong>the</strong> result (<strong>the</strong> Expected<br />

Time System) to be adequately precise. A positive<br />

effect <strong>of</strong> <strong>the</strong> introduced procedure is that if <strong>the</strong> s<strong>of</strong>tware<br />

is locally modified, many test cases can most likely be<br />

reused.<br />

Our approach inevitably yields to large sets <strong>of</strong> test<br />

data. We <strong>the</strong>refore designed an architecture [3] that (a)<br />

stores <strong>the</strong>m and (b) manages meta information in order<br />

to use <strong>the</strong>m conveniently.<br />

III. THE TEMPORAL BEHAVIOR OF HARD-<br />

WARE<br />

t1<br />

1<br />

n1 CFG<br />

n2 n3<br />

n4<br />

n5<br />

Trace element 1:<br />

instruction info<br />

t<br />

t start<br />

finish<br />

t2<br />

2<br />

t3<br />

3<br />

t4<br />

4<br />

t5<br />

5<br />

Trace element 2:<br />

instruction info<br />

t<br />

t start<br />

finish<br />

t6<br />

6<br />

t7<br />

7<br />

t8<br />

8<br />

...<br />

...<br />

...<br />

Trace element k:<br />

instruction info<br />

t<br />

t start<br />

finish<br />

t<br />

m-2<br />

t m-1<br />

t m<br />

m-2 m-1 m<br />

Measured path<br />

n1 n3 n4 n5<br />

tbegin<br />

tend<br />

tbegin<br />

tend<br />

tbegin<br />

tend<br />

tbegin<br />

tend<br />

Trace<br />

CPU cycles<br />

Figure 2: The relation between a trace, measured path,<br />

and CFG<br />

Recall that one <strong>of</strong> our goals is to estimate <strong>the</strong> WCET<br />

<strong>of</strong> <strong>the</strong> system under test. The hardware has to be part <strong>of</strong><br />

<strong>the</strong> analysis to get a reasonable degree <strong>of</strong> precision. As<br />

an example, <strong>the</strong> latency when reading data from memory<br />

can differ by <strong>the</strong> factor <strong>of</strong> 100, depending on whe<strong>the</strong>r <strong>the</strong><br />

data is located in <strong>the</strong> cache or not. A similar effect can<br />

be observed in <strong>the</strong> processor pipeline. A precondition<br />

to include hardware effects <strong>of</strong> this kind into <strong>the</strong> analysis<br />

using <strong>the</strong> measurement-based approach is to sample an<br />

execution run at a high rate in order not to miss anything.<br />

In our experiments we perform measurements on a Infineon<br />

Tricore TC1796 microcontroller. It includes an instruction<br />

cache, a processor pipeline and <strong>the</strong> OCDS level<br />

2 interface which conforms to <strong>the</strong> NEXUS 3 standard.<br />

Using both, <strong>the</strong> on-chip debug support and <strong>the</strong> Lauterbach<br />

LA-7690 Powertrace device, we reproduce a trace<br />

<strong>of</strong> an execution run that includes cycle-accurate timing<br />

and control-flow information. As depicted in Figure 2, a<br />

trace is a sequence <strong>of</strong> references to machine instructions.<br />

Fur<strong>the</strong>rmore, trace elements are mapped to time-stamped<br />

CPU cycles in which <strong>the</strong> respective instructions are processed<br />

in <strong>the</strong> execution unit <strong>of</strong> <strong>the</strong> pipeline. As men-<br />

3 http://www.nexus5001.org/<br />

116<br />

tioned, <strong>the</strong> FORTAS model relates to <strong>the</strong> Control Flow<br />

Graph (CFG) <strong>of</strong> <strong>the</strong> s<strong>of</strong>tware under investigation. In order<br />

to get a measured path, <strong>the</strong> trace is associated with<br />

CFG nodes in a fur<strong>the</strong>r step, since each node might represent<br />

several machine instructions. Thus, detailed timing<br />

information can be related to <strong>the</strong> FORTAS model in<br />

a flexible way.<br />

The demonstrated measurement and post-processing<br />

method has been implemented and shows positive results<br />

with respect to its applicability.<br />

A. FUTURE WORK<br />

As shown, we are able to capture delays due to hardware<br />

effects. The next step is to use <strong>the</strong> collected data for <strong>the</strong><br />

generation <strong>of</strong> new test cases. This implies that we have<br />

to speculate on what favorable test cases could be. In<br />

<strong>the</strong> context <strong>of</strong> WCET analysis <strong>the</strong> goal is quite clear: we<br />

want to get as many cache misses and pipeline stalls as<br />

possible in order to discover expensive paths within our<br />

model. One problem is to perceive which paths that have<br />

not yet been measured might lead to those time consuming<br />

effects. Ano<strong>the</strong>r challenge is to differentiate whe<strong>the</strong>r<br />

an observed delay results from a cache miss, a pipeline<br />

stall or both.<br />

Beside WCET analysis, we also want to provide <strong>the</strong><br />

s<strong>of</strong>tware developer with timing pr<strong>of</strong>iles that can be used<br />

for performance optimization. Here, <strong>the</strong> Average Case<br />

Execution Time (ACET) might be <strong>of</strong> interest as well. We<br />

still do not know precisely what timing information <strong>the</strong><br />

programmer considers to be useful. This is also one <strong>of</strong><br />

our current research topics. Once this is known, our objective<br />

will be to provide this information by designing<br />

an appropriate analysis.<br />

REFERENCES<br />

[1] Ingomar Wenzel, Bernhard Rieder, Raimund Kirner,<br />

and Peter Puschner. Automatic Timing Model Generation<br />

by CFG Partitioning and Model Checking. In<br />

Proc. <strong>Conference</strong> on Design, Automation, and Test in<br />

Europe, Mar. 2005.<br />

[2] Andreas Holzer, Christian Schallhart, Michael<br />

Tautschnig, and Helmut Veith. FShell: Systematic<br />

Test Case Generation for Dynamic Analysis and<br />

Measurement. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 20th International<br />

<strong>Conference</strong> on Computer Aided Verification<br />

(CAV <strong>2008</strong>), Lecture Notes in Computer Science,<br />

Princeton, NJ, USA, July <strong>2008</strong>. Springer.<br />

[3] Sven Bünte and Michael Tautschnig. A Benchmarking<br />

Suite for Measurement-Based WCET Analysis<br />

Tools. In First International <strong>Conference</strong> on S<strong>of</strong>tware<br />

Testing, Verification and Validation (ICST), Lillehammer,<br />

Norway, April <strong>2008</strong>. IEEE Computer Society<br />

Press.


Divide and Measure: CFG Segmentation for <strong>the</strong><br />

Measurement-Based Analysis <strong>of</strong> Resource Consumption<br />

Michael Zolda and Raimund Kirner (Faculty Mentor)<br />

Institut für Technische Informatik<br />

Technische Universität Wien<br />

Wien, Austria<br />

Email: {michaelz, raimund}@vmars.tuwien.ac.at<br />

Abstract — A computer system is a good computer system if<br />

it correctly performs <strong>the</strong> task it was intended to perform. This<br />

is not even half <strong>of</strong> <strong>the</strong> truth: Non-functional requirements are<br />

abundant in <strong>the</strong> world <strong>of</strong> s<strong>of</strong>tware and system engineering, even<br />

if <strong>the</strong>y are not always stated explicitly. In our work we are concerned<br />

with <strong>the</strong> measurement-based analysis <strong>of</strong> resource consumption.<br />

Examples <strong>of</strong> resources are time, energy, or memory<br />

space. In <strong>the</strong> context <strong>of</strong> our measurement-based approach for<br />

s<strong>of</strong>tware analysis, we face <strong>the</strong> problem <strong>of</strong> breaking <strong>the</strong> s<strong>of</strong>tware<br />

under examination into smaller parts <strong>of</strong> manageable size,<br />

a process dubbed CFG Segmentation.<br />

I. INTRODUCTION<br />

In our measurement-based approach for resourceconsumption<br />

<strong>of</strong> embedded s<strong>of</strong>tware, we perform measurements<br />

on <strong>the</strong> real physical computer system, and subsequently<br />

integrate <strong>the</strong> measurement results into a resource<br />

consumption model that is ready for expert inspection<br />

or use in a higher level analysis <strong>of</strong> <strong>the</strong> system.<br />

In order to obtain an accurate picture <strong>of</strong> <strong>the</strong> system behavior,<br />

we would like to achieve full measurement coverage<br />

<strong>of</strong> all feasible operation sequences <strong>of</strong> <strong>the</strong> s<strong>of</strong>tware<br />

under test. We <strong>the</strong>refore start from <strong>the</strong> Control Flow<br />

Graph (CFG) <strong>of</strong> <strong>the</strong> s<strong>of</strong>tware under examination, a fundamental<br />

program representation where nodes represent<br />

<strong>the</strong> operations <strong>of</strong> <strong>the</strong> s<strong>of</strong>tware, and where directed edges<br />

represent possible successive execution (see Figure 1 for<br />

an example <strong>of</strong> a CFG). Thus, each feasible sequence <strong>of</strong><br />

operations <strong>of</strong> <strong>the</strong> s<strong>of</strong>tware corresponds to a path in <strong>the</strong><br />

CFG.<br />

In <strong>the</strong>ory, it would <strong>the</strong>refore suffice to examine all<br />

CFG paths, but due to <strong>the</strong> huge number <strong>of</strong> paths in <strong>the</strong><br />

CFGs <strong>of</strong> real s<strong>of</strong>tware, this approach is practically infeasible.<br />

For example, <strong>the</strong> CFG <strong>of</strong> a simple actuator controller<br />

that is part <strong>of</strong> our benchmarks contains about 10 44<br />

paths, which is clearly beyond measurement feasibility.<br />

To cope with such huge numbers <strong>of</strong> paths, we split <strong>the</strong><br />

CFG into connected subgraphs <strong>of</strong> manageable size (“segments”),<br />

deal with <strong>the</strong>se subgraphs individually, and subsequently<br />

merge <strong>the</strong> data that was obtained for each segment<br />

into a global resource consumption model.<br />

A (simple) segment is a connected subgraph <strong>of</strong> a CFG,<br />

and is characterized by its sets <strong>of</strong> entry edges (edges lead-<br />

117<br />

20783664<br />

20850528 20795424<br />

27719392<br />

30565936<br />

27719552 14453136<br />

20795568 20850368<br />

13861344 13802240<br />

13858944<br />

14793904<br />

24922544<br />

17667232<br />

17677824 17699296<br />

17699248<br />

14955392 14955456<br />

17710032<br />

17731072 17752272<br />

Figure 1: A CFG with four segments (indicated by thick,<br />

dashed borders). Thin borders, on <strong>the</strong> o<strong>the</strong>r hand, mark<br />

intermediate segments generated by our proposed greedy<br />

segmentation algorithm.<br />

ing into <strong>the</strong> subgraph), its set <strong>of</strong> exit edges (edges leading<br />

out <strong>of</strong> <strong>the</strong> subgraph), and <strong>the</strong> number <strong>of</strong> segment paths<br />

(paths leading through <strong>the</strong> subgraph).<br />

Our goal is to partition <strong>the</strong> CFG into as few segments<br />

as possible, such that <strong>the</strong> segment path total does not exceed<br />

a given feasibility limit.<br />

To illustrate our idea, consider Figure 1, which shows<br />

a tiny CFG that has been partitioned into four segments.<br />

The total number <strong>of</strong> segment paths to be examined is<br />

3+6+7+8 = 24. Without segmentation we would<br />

have to examine all 99 CFG paths. The number <strong>of</strong> paths<br />

to be examined is thus lowered by a factor <strong>of</strong> 4 in this<br />

tiny example, but <strong>the</strong> effect is much more pronounced<br />

for realistically-sized CFGs, where we use segments in<br />

<strong>the</strong> magnitude <strong>of</strong> hundreds or thousands <strong>of</strong> paths.<br />

II. PROPOSED APPROACH<br />

In [1, 2], Wenzel et al. present an approach for CFG<br />

segmentation that produces segments with a single entry<br />

edge. Such segments are advantageous from a composi-<br />

17762832


14144128<br />

13974624<br />

14172032<br />

13911376<br />

14631232<br />

14147008<br />

14728752<br />

14742400<br />

14020320<br />

14212496<br />

14518304<br />

13847536<br />

14309808<br />

14384736<br />

13879344 14061328 14518496<br />

14122272<br />

14890208<br />

14880336<br />

14048752<br />

14894128<br />

14781568 14215328<br />

13945200<br />

Figure 2: A CFG with a lattice-shaped structure.<br />

bility point <strong>of</strong> view, but yield problems with large proper<br />

interval structures [3]. For example, <strong>the</strong> lattice-shaped<br />

CFG in Figure 2 allows only <strong>the</strong> two trivial segmentations<br />

<strong>of</strong> ei<strong>the</strong>r putting everything into one single segment,<br />

or putting each node an individual segment.<br />

We <strong>the</strong>refore abandon <strong>the</strong> single entry restriction and<br />

allow any number <strong>of</strong> entry edges. However, for composability<br />

reasons, we do not want <strong>the</strong> entry/exit interface<br />

to become unjustifiably large. Because we later want to<br />

associate measurement results with all entry/exit combinations<br />

<strong>of</strong> a segment, a reasonable size metric for <strong>the</strong> interface<br />

<strong>of</strong> a segment is<br />

h(S) =entries(S) · exits(S). (1)<br />

To perform a segmentation <strong>of</strong> this more general form,<br />

we propose an iterative greedy algorithm that builds segments<br />

bottom-up from smaller segments, starting with<br />

single-node segments. For each iteration, <strong>the</strong> algorithm<br />

considers all segments that could be created by merging<br />

pairs <strong>of</strong> adjacent segments and picks <strong>the</strong> least-cost one.<br />

As costs functions, we currently consider <strong>the</strong> following<br />

class <strong>of</strong> functions:<br />

costs(S) =h(S) a · paths(S) b , (2)<br />

We have already mentioned that we would like to keep<br />

<strong>the</strong> entry/exit interfaces <strong>of</strong> segments as small as reasonably<br />

possible. This explains <strong>the</strong> h(S) part <strong>of</strong> <strong>the</strong><br />

costs function, which is weighted by <strong>the</strong> tunable exponent<br />

a. On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> paths(S) part <strong>of</strong> <strong>the</strong> costs<br />

function assures that <strong>the</strong> algorithm will produce roughly<br />

equally-sized segments. Again, this part <strong>of</strong> <strong>the</strong> costs<br />

function can be tuned by adjusting an exponent.<br />

The example in Figure 1 indicates both, <strong>the</strong> final segments<br />

produced by our algorithm (thick, dashed borders),<br />

and <strong>the</strong> intermediate segments generated during <strong>the</strong> individual<br />

iterations (thin borders).<br />

118<br />

III. PRELIMINARY RESULTS AND FUTURE<br />

WORK<br />

We have implemented <strong>the</strong> segmentation algorithm presented<br />

above as a component <strong>of</strong> <strong>the</strong> timing analysis<br />

framework that is currently being developed within <strong>the</strong><br />

FORTAS project. Our implementation is already capable<br />

<strong>of</strong> handling real industrial application code. A logical<br />

next step would be a detailed evaluation <strong>of</strong> <strong>the</strong> produced<br />

segments for a larger number <strong>of</strong> benchmarks and <strong>the</strong> usage<br />

<strong>of</strong> <strong>the</strong> obtained insight to tune <strong>the</strong> costs function presented<br />

above. It is even conceivable that careful examination<br />

<strong>of</strong> such results point at completely new forms <strong>of</strong><br />

costs functions.<br />

Performing segmentation is just one tasks in<br />

measurement-based analysis <strong>of</strong> resource consumption.<br />

O<strong>the</strong>r tasks include <strong>the</strong> generation <strong>of</strong> suitable test<br />

data that can trigger <strong>the</strong> execution <strong>of</strong> individual paths [1],<br />

performing <strong>the</strong> actual measurements, construction <strong>of</strong> <strong>the</strong><br />

complete timing model, and o<strong>the</strong>rs. All <strong>the</strong>se tasks, and<br />

especially <strong>the</strong>ir integration, are important future research<br />

issues.<br />

IV. SUMMARY<br />

In this work, we have explained <strong>the</strong> problem <strong>of</strong> CFG segmentation,<br />

as it arises in <strong>the</strong> context <strong>of</strong> measurementbased<br />

resource consumption analysis. We have introduced<br />

segmentation as a possible solution to handle <strong>the</strong><br />

explosion <strong>of</strong> CFG paths and have pointed at <strong>the</strong> shortcomings<br />

<strong>of</strong> previous segmentation approaches. Lastly,<br />

we have sketched a novel segmentation algorithm that<br />

overcomes <strong>the</strong>se shortcoming.<br />

ACKNOWLEDGMENTS<br />

The research leading to <strong>the</strong>se results has received funding<br />

from <strong>the</strong> Austrian Science Fund (Fonds zur Förderung<br />

der wissenschaftlichen Forschung) within <strong>the</strong> research<br />

project “Formal Timing Analysis Suite <strong>of</strong> Real-Time<br />

Systems” (FORTAS-RT) under contract P19230-N13.<br />

REFERENCES<br />

[1] Ingomar Wenzel, Bernhard Rieder, Raimund Kirner,<br />

and Peter Puschner. Automatic Timing Model Generation<br />

by CFG Partitioning and Model Checking. In<br />

Proc. <strong>Conference</strong> on Design, Automation, and Test in<br />

Europe, Mar. 2005.<br />

[2] Ingomar Wenzel, Raimund Kirner, Bernhard Rieder,<br />

and Peter Puschner. Measurement-Based Timing<br />

Analysis. To be presented at <strong>the</strong> 3rd Int. Symp. on<br />

Leveraging Applications <strong>of</strong> Formal Methods, Verification<br />

and Validation, Oct. <strong>2008</strong>.<br />

[3] Steven S. Muchnick. Advanced Compiler Design<br />

and Implementation. Morgan Kaufmann, August<br />

1997.


Neutralizing Timing Anomalies in<br />

Complex Computer Architectures ∗<br />

Albrecht Kadlec and Raimund Kirner (Faculty Mentor)<br />

Institut für Technische Informatik<br />

University <strong>of</strong> Technology, Vienna<br />

Vienna, Austria<br />

Email: albrecht@vmars.tuwien.ac.at<br />

Abstract — The complex designs <strong>of</strong> current computer architectures<br />

mostly stem from complex optimizations performed at<br />

run-time. The interactions <strong>of</strong> <strong>the</strong>se optimizations lead to timing<br />

anomalies. This paper presents a first approach to actually<br />

eliminate timing anomalies by transforming <strong>the</strong> code in a way<br />

that avoids <strong>the</strong> adverse effects <strong>of</strong> timing anomalies on <strong>the</strong> processor<br />

state. This enables static timing analysis to be extended<br />

to more complex types <strong>of</strong> computer architectures.<br />

I. INTRODUCTION<br />

Safety critical real time systems have to guarantee timeliness<br />

in addition to correctness. To ensure this, schedulability<br />

analysis was established for multi-tasking systems.<br />

To provide schedulability analysis with hard facts on <strong>the</strong><br />

Worst Case Execution Time (WCET) <strong>of</strong> a single task,<br />

static timing analysis was established (more precisely:<br />

static WCET analysis). But static WCET analysis in turn<br />

relies on timing composability for complexity reduction:<br />

it requires that single basic blocks can be analyzed separately<br />

and <strong>the</strong> results be added to form an overall WCET<br />

for a certain execution path through <strong>the</strong> program.<br />

Timing anomalies compromise timing composability,<br />

as <strong>the</strong>y produce timing artefacts, that are only analyzable<br />

if <strong>the</strong> whole path is considered at once. This in turn drives<br />

up overall analysis complexity, as <strong>the</strong> number <strong>of</strong> paths is<br />

much larger than <strong>the</strong> number <strong>of</strong> basic blocks, with more<br />

analysis time needed for a path than for a single basic<br />

block. This easily renders timing analysis intractable, so<br />

timing composability must be restored.<br />

Timing anomalies were first described by Lundqvist<br />

and Stenström for out-<strong>of</strong>-order architectures [1]. They<br />

defined timing anomalies as unintuitive timing behaviour,<br />

where <strong>the</strong> local worst case does not correspond<br />

to <strong>the</strong> global worst case. Their work was extended among<br />

o<strong>the</strong>rs by Wenzel [2], who showed <strong>the</strong> existence <strong>of</strong> timing<br />

anomalies for less complex architectures without out<strong>of</strong>-order<br />

execution.<br />

In [3] we identified <strong>the</strong> processor state as <strong>the</strong> carrier<br />

and <strong>the</strong>refore a necessary condition for late effects, and<br />

∗ This work has been partially supported by <strong>the</strong> Austrian Science<br />

Fund (Fonds zur Förderung der wissenschaftlichen Forschung) within<br />

<strong>the</strong> research project “Compiler-Support for Timing Analysis” (CoSTA)<br />

under contract P18925-N13.<br />

119<br />

listed several properties to categorize <strong>the</strong> influence <strong>of</strong><br />

<strong>the</strong> various CPU state. The second necessary condition<br />

for timing anomalies is choice in <strong>the</strong> CPU-internal algorithms,<br />

i.e., <strong>the</strong> possibility <strong>of</strong> functionally identical,<br />

but not timing-identical decisions based on this processor<br />

state (e.g. all sorts <strong>of</strong> speculation).<br />

As a result, <strong>the</strong> processor pipeline was identified as <strong>the</strong><br />

biggest obstacle for static timing analysis, as it connects<br />

subsystems such as cache or jump prediction, has a high<br />

decision density and <strong>the</strong> decisions immediately influence<br />

several o<strong>the</strong>r subsystems.<br />

II. THEORETICAL FOUNDATION<br />

Our key observation is <strong>the</strong> fact that we do not need to<br />

avoid timing variations in general, but that we only need<br />

to make sure, that <strong>the</strong> state is identical after all possible<br />

executions - i.e. after all possible different decisions during<br />

<strong>the</strong> execution <strong>of</strong> <strong>the</strong> instructions so far.<br />

Timing variations can be accounted for (overestimated)<br />

by WCET analysis. The different possible<br />

state <strong>of</strong> <strong>the</strong> CPU after execution <strong>of</strong> an instruction is responsible<br />

for <strong>the</strong> combinatorial explosion <strong>of</strong> interactions<br />

with later instructions, which makes WCET analysis intractable.<br />

However, since CPUs must be deterministic in <strong>the</strong> execution<br />

<strong>of</strong> <strong>the</strong> instructions, <strong>the</strong> difference in state can only<br />

be a difference in time. Given enough time, <strong>the</strong> CPU<br />

state must converge to a single output state after each<br />

instruction is executed. So we must make sure that <strong>the</strong><br />

following instructions, which might be influenced by <strong>the</strong><br />

respective state, can only observe that state after it has<br />

converged - when it is stable.<br />

This can be done by inserting NOP instructions, until<br />

<strong>the</strong> state has stabilized, i.e., until all effects <strong>of</strong> <strong>the</strong> instruction<br />

onto <strong>the</strong> CPU state have occurred. By induction this<br />

avoids <strong>the</strong> state space explosion in WCET analysis, no<br />

matter which decisions were taken in <strong>the</strong> execution <strong>of</strong><br />

any instruction. The remaining small timing variations<br />

<strong>the</strong>mselves are easily accounted-for by <strong>the</strong> WCET analysis.


III. PRACTICAL SOLUTIONS<br />

A. RATE NOP INSERTION<br />

The easiest way to achieve our goal is to exploit <strong>the</strong> fact<br />

that <strong>the</strong> fetch width closely resembles <strong>the</strong> average issue<br />

width for a given processor: making <strong>the</strong> fetch width much<br />

larger overly increases <strong>the</strong> costs <strong>of</strong> <strong>the</strong> memory subsystem,<br />

making it smaller voids <strong>the</strong> expenses for execution<br />

units. However, <strong>the</strong> fetch width must be somewhat larger<br />

to recover from pipeline break instructions, cache misses<br />

and mis-speculations.<br />

Now if we fill this overcapacity with NOPs, we can<br />

easily control <strong>the</strong> issuing stage <strong>of</strong> <strong>the</strong> CPU: starting with<br />

an empty prefetch window, each instruction that enters<br />

<strong>the</strong> prefetch window is immediately ready and can be executed<br />

without reordering. If we do not have an instruction<br />

to issue in this cycle, we still have to insert a NOP<br />

to keep <strong>the</strong> prefetch window from filling up with <strong>the</strong> following<br />

instructions, that may not be executed yet.<br />

B. SPARSE NOP INSERTION<br />

We implemented a sparse NOP insertion algorithm<br />

within a compile-time list scheduler for fictious ARMbased<br />

superscalar out-<strong>of</strong>-order architectures. For detailed<br />

evaluations, <strong>the</strong> issue-width <strong>of</strong> <strong>the</strong> fictious CPU can be<br />

configured as well as <strong>the</strong> depth <strong>of</strong> <strong>the</strong> prefetch window<br />

and <strong>the</strong> number <strong>of</strong> respective functional units.<br />

Now whenever a variable-time instruction is scheduled,<br />

<strong>the</strong> succeeding instructions, that may be scheduled<br />

to <strong>the</strong> same unit are marked with a minimum program<br />

counter before which <strong>the</strong>y may not be scheduled. The<br />

distance to that program counter is filled with o<strong>the</strong>r ready<br />

instructions for independent functional units or NOPs.<br />

This mechanism makes sure that <strong>the</strong> instruction does not<br />

enter <strong>the</strong> prefetch window before <strong>the</strong> effects <strong>of</strong> <strong>the</strong> previous<br />

instructions on <strong>the</strong> state are completed.<br />

C. PRELIMINARY MEASUREMENTS<br />

First preliminary measurement indicate that <strong>the</strong>re is a<br />

significant but acceptable penalty in code size and dynamically<br />

executed instructions associated with our algorithm.<br />

However, we cannot supply definite numbers<br />

yet, as <strong>the</strong> algorithm still seems to expose bugs, so <strong>the</strong><br />

measurements are not yet reliable.<br />

IV. CONCLUSION<br />

We illustrated <strong>the</strong> problem <strong>of</strong> timing anomalies in <strong>the</strong><br />

wider context <strong>of</strong> WCET and schedulability analysis for<br />

hard real time systems. In prior work we already refined<br />

<strong>the</strong> description <strong>of</strong> timing anomalies from <strong>the</strong> timing domain<br />

to <strong>the</strong> state domain.<br />

Now we reasoned about <strong>the</strong> necessary preconditions<br />

for timing anomalies to occur in <strong>the</strong> state domain, and<br />

separated <strong>the</strong> tolerable effects that can be handled by<br />

120<br />

static WCET analysis from <strong>the</strong> adversary effects, that<br />

lead to state space explosion within <strong>the</strong> WCET analysis.<br />

Finally we designed two algorithms that avoid <strong>the</strong>se<br />

adversary effects and implemented one <strong>of</strong> <strong>the</strong>m. First<br />

measurements for a fictious four way issue superscalar<br />

out-<strong>of</strong>-order ARM implementation seem very positive,<br />

but unfortunately <strong>the</strong> numbers are not reliable enough to<br />

be published, yet.<br />

We will continue on this track and evaluate <strong>the</strong> algorithm<br />

for different configurations <strong>of</strong> our fictious architecture<br />

family. We will <strong>the</strong>n look for fur<strong>the</strong>r improvements<br />

<strong>of</strong> <strong>the</strong> algorithm, or more efficient alternative solutions.<br />

REFERENCES<br />

[1] Thomas Lundqvist and Per Stenström. Integrating<br />

path and timing analysis using instruction-level simulation<br />

techniques. In Proc. ACM SIGPLAN Workshop<br />

on Languages, Compilers, and Tools for Embedded<br />

Systems (LCTES), pages 1–15, June 1998.<br />

[2] Ingomar Wenzel. Principles <strong>of</strong> timing anomalies in<br />

superscalar processors. Master’s <strong>the</strong>sis, Technische<br />

Universität Wien, Vienna, Austria, 2003.<br />

[3] Albrecht Kadlec and Raimund Kirner. On <strong>the</strong><br />

difficulty <strong>of</strong> building a precise timing model for<br />

real-time programming. In 14. Kolloquium Programmiersprachen<br />

und Grundlagen der Programmierung,<br />

pages 99–105, Timmendorfer Strand, Germany,<br />

Oct. 2007.


Optical Imaging Hidden Objects in Turbid Scattering Media using<br />

Principal Component Analysis<br />

Binlin Wu, S. K. Gayen (Faculty Mentor) and R. R. Alfano (Faculty Mentor)<br />

The Institute for Ultrafast Spectroscopy and Lasers (IUSL)<br />

The City College <strong>of</strong> The City University <strong>of</strong> New York<br />

New York, USA<br />

Email: bwu@sci.ccny.cuny.edu<br />

Abstract — We have developed a new imaging<br />

approach for 3D localization and characterization<br />

<strong>of</strong> objects in turbid media using <strong>the</strong> principal component<br />

analysis (PCA). The aim is to develop an<br />

approach for imaging inside a breast using light. A<br />

novel algorithm is presented based on PCA for<br />

optical imaging (OIPCA) <strong>of</strong> objects located in scattering<br />

media. OIPCA can be used with any geometry<br />

and any forward model in steady-state or timeresolved<br />

imaging. We use a model medium to demonstrate<br />

this method using simulated data.<br />

I. INTRODUCTION<br />

Reliable and fast approaches for inverse image reconstruction<br />

are important. In this paper, we present<br />

a novel and new algorithm based on PCA for imaging.<br />

We refer to this approach as Optical Imaging<br />

using Principal Component Analysis (OIPCA).<br />

OIPCA determines <strong>the</strong> number and strengths <strong>of</strong> targets<br />

with <strong>the</strong> leading eigenvalues. The light propagations<br />

from different inhomogeneities to <strong>the</strong> source or<br />

detector planes are uncorrelated. The locations <strong>of</strong> <strong>the</strong><br />

objects are retrieved by fitting principal components<br />

intensity distributions (PCID) to Green’s functions.<br />

A multi-source and multi-detector scheme is used.<br />

This paper focuses on using PCA as a new<br />

method for optical imaging <strong>of</strong> objects in scattering<br />

media for breast applications.<br />

II. ALGORITHMS AND SIMULATIONS<br />

A. DIFFUSION APPROXIMATION<br />

Using <strong>the</strong> diffusion approximation, <strong>the</strong> light intensity<br />

perturbation on <strong>the</strong> detector plane to <strong>the</strong> first order <strong>of</strong><br />

Born Approximation can be written as [1][2]<br />

n<br />

�<br />

j�1<br />

��<br />

� G(<br />

r , r ) � G(<br />

r , r ) , (1)<br />

sca<br />

d<br />

j<br />

j<br />

where a j cV �� � � and � � �DcV<br />

j for absorptive<br />

and scattering inhomogeneities respectively.<br />

G ( r<br />

d<br />

, r<br />

j<br />

) and G(<br />

r<br />

j<br />

, r<br />

s<br />

) are Green’s functions for<br />

j<br />

s<br />

121<br />

light propagation from objects to detectors and<br />

sources to objects respectively. The light perturbation<br />

can be rewritten as<br />

�<br />

x ( r , r ) � a ( r ) s ( r ) , (2)<br />

d<br />

s<br />

j<br />

j<br />

d<br />

where s j(<br />

rs<br />

) � � jG(<br />

rj,<br />

rs<br />

) , a j(<br />

rd<br />

) � � jG(<br />

rd<br />

, rj<br />

) ,<br />

j<br />

� and � are scaling factors and � j � j � � j . Using<br />

a multi-source and multi-detector scheme, one has<br />

X � AS , where rows <strong>of</strong> S are virtual sources which<br />

are <strong>the</strong> inhomogeneities illuminated by <strong>the</strong> incident<br />

light and A is a mixing matrix.<br />

B. OPTICAL IMAGING USING PCA<br />

PCA [3] is a data analysis tool widely used to extract<br />

hidden uncorrelated signals by finding <strong>the</strong> largest<br />

variances (eigenvalues). We assume <strong>the</strong> virtual<br />

sources are not correlated. So we mean-center <strong>the</strong><br />

matrix X , solve <strong>the</strong> eigenvalue equation for <strong>the</strong><br />

covariance matrix. Then we sort and select principal<br />

components (PC) with <strong>the</strong> leading eigenvalues and<br />

project <strong>the</strong> matrix X onto <strong>the</strong> selected PCs with<br />

noise reduced. The uncorrelated matrix is <strong>the</strong> virtual<br />

T T �1<br />

source matrix S . A � XS ( SS ) . Locations <strong>of</strong><br />

<strong>the</strong> targets can be found by fitting ei<strong>the</strong>r <strong>of</strong> A and S<br />

or both to <strong>the</strong> respective Green’s functions using<br />

least squares fitting:<br />

{ [<br />

1<br />

( ) ( , )]<br />

2<br />

, ,<br />

[<br />

1<br />

( ) ( , )]<br />

2<br />

}<br />

min � �<br />

�<br />

j<br />

s<br />

j<br />

r<br />

s<br />

� G r<br />

j<br />

r<br />

s<br />

r<br />

j<br />

�<br />

j<br />

�<br />

j r<br />

s<br />

. (3)<br />

�<br />

�<br />

� �<br />

j<br />

a<br />

j<br />

r<br />

d<br />

� G r<br />

d<br />

r<br />

j<br />

r<br />

d<br />

C. RESULTS OF SIMULATED DATA<br />

The sample is a 40mm thick uniform slab <strong>of</strong> scattering<br />

medium, as shown in Figure 1. Its absorption and<br />

diffusion coefficients are � a<br />

�1<br />

� 1/<br />

300mm<br />

and<br />

D � 1/<br />

3mm<br />

, respectively. An absorptive and a<br />

scattering objects were placed at (50, 60, 15) mm and<br />

(30, 30, 25) mm, respectively. The incident CW<br />

beam scanned at 21� 21 grid points covering<br />

s


2<br />

80 � 80mm<br />

area. Light on <strong>the</strong> o<strong>the</strong>r side was recorded<br />

by CCD camera at 41� 41 grid points covering<br />

<strong>the</strong> same area. 5% Gaussian noise was added to<br />

<strong>the</strong> simulated intensity change.<br />

Figure 1: Light intensity on one side is measured<br />

when a point source scans on <strong>the</strong> o<strong>the</strong>r side.<br />

The eigenvalues found by PCA were sorted and<br />

<strong>the</strong> leading four PCIDs on <strong>the</strong> source and detector<br />

planes were calculated. The absorptive object has<br />

only one component while <strong>the</strong> scattering object has<br />

one centrosymmetric and two dumbbell-shaped<br />

components on <strong>the</strong> source and detector planes. The<br />

centrosymmetric PCIDs for <strong>the</strong> scattering and absorptive<br />

objects are shown in Figure 2.<br />

By least squares fitting <strong>the</strong> intensity pr<strong>of</strong>iles with<br />

respective Green’s functions, <strong>the</strong> absorptive object is<br />

found to be at (50, 60, 15.3) mm and <strong>the</strong> scattering at<br />

(30, 30, 25.1) mm.<br />

(a)<br />

(b)<br />

(c)<br />

122<br />

(d)<br />

Figure 2: Centrosymmetric PCIDs <strong>of</strong> <strong>the</strong> scattering<br />

inhomogeneity are shown in (a), <strong>of</strong> <strong>the</strong> absorptive<br />

inhomogeneity in (c). Normalized PCIDs on <strong>the</strong><br />

source plane are shown in <strong>the</strong> second columns <strong>of</strong> (a)<br />

and (c), and on detector plane in <strong>the</strong> first columns.<br />

(b) and (d) are showing <strong>the</strong> horizontal pr<strong>of</strong>iles <strong>of</strong><br />

intensity distributions on <strong>the</strong> source and detector<br />

planes (circles) and <strong>the</strong>ir least squares fittings (solid).<br />

III. SUMMARY<br />

OIPCA can determine <strong>the</strong> number, strength and type<br />

<strong>of</strong> targets without any forward model, which is helpful<br />

to achieve a “bind test” <strong>of</strong> targets embedded in<br />

turbid media such as a tumor in a breast. Due to <strong>the</strong><br />

high accuracy, OIPCA is a potential novel tool for<br />

breast cancer detection in early stages. OIPCA runs<br />

very fast which make it possible for in-vivo real-time<br />

detection. OIPCA can also be used for pre-evaluation<br />

or noise removal for o<strong>the</strong>r imaging and reconstruction<br />

modalities.<br />

ACKNOWLEDGMENTS<br />

The research is supported by <strong>the</strong> U. S. Army Medical<br />

Research and Materials Command, IUSL and The<br />

City University <strong>of</strong> New York (CUNY) organized<br />

research programs. This research by Binlin Wu is<br />

made possible by <strong>the</strong> support and mentoring from<br />

Pr<strong>of</strong>essor S. K. Gayen and Pr<strong>of</strong>essor R. R. Alfano.<br />

The author Binlin Wu thanks Dr. Cai, Dr M. Xu and<br />

Dr. Alrubaiee for helpful discussions.<br />

REFERENCES<br />

[1] M. A. O’Leary, D. A. Boas, B. Chance and A. G.<br />

Yodh, “Experimental images <strong>of</strong> heterogeneous<br />

turbid media by frequency-domain diffusingphoton<br />

tomography,” Opt. Lett. 20, 426–428<br />

(1995).<br />

[2] S. R. Arridge, “Photon-measurement density<br />

functions. Part I: Analytical forms”, Applied Optics,<br />

Vol. 34, Issue 31, pp. 7395-7409 (1995)<br />

[3] I. T. Jolliffe, Principal Component Analysis.<br />

New York: Springer, 1986.


Links between Diachronic Thinking and Temporal Cognition<br />

Tasks Observed in Middle Childhood<br />

Brandy Moore, Patricia Brooks, Laura Rabin (Faculty Mentor) and Carmen Carrion<br />

City University <strong>of</strong> New York<br />

Graduate Center, and Brooklyn College<br />

New York, United States<br />

Email: lrabin@brooklyn.cuny.edu<br />

Abstract — The present study will assess<br />

developments in temporal language, source<br />

memory, and diachronic concepts <strong>of</strong><br />

transformation, syn<strong>the</strong>sis, and tendency in children<br />

ages 5 to 10. These constructs broadly refer to <strong>the</strong><br />

interpretation <strong>of</strong> events and understanding <strong>of</strong> time<br />

in relation to <strong>the</strong> past, present, and future. Research<br />

has not investigated all <strong>of</strong> <strong>the</strong>se constructs<br />

simultaneously within this age group. One-hundred<br />

children participated in this study. Participants<br />

performed a series <strong>of</strong> tasks thought to reflect an<br />

understanding for temporal cognition, syn<strong>the</strong>sis,<br />

source memory, and verbal and nonverbal ability.<br />

Preliminary data reveals a correlation between<br />

most tasks <strong>of</strong> temporal cognition. The data indicates<br />

that temporal cognition is significantly related to<br />

life experience and not to children’s nonverbal<br />

intelligence or age.<br />

I. INTRODUCTION<br />

The construct <strong>of</strong> diachronic thinking can be thought<br />

<strong>of</strong> as <strong>the</strong> ability <strong>of</strong> an individual to naturally visualize<br />

<strong>the</strong> past, present and future. Prior research suggests<br />

that this developmental change occurs early on in<br />

childhood during <strong>the</strong> pre-school years (Atance &<br />

O’Neill, 2005). This learned response occurs through<br />

continual exposure to daily life in which children<br />

develop a sense <strong>of</strong> time (Nelson, 2007). At<br />

approximately three years <strong>of</strong> age, children begin to<br />

think about <strong>the</strong> not-now (<strong>the</strong> past) as a point <strong>of</strong><br />

reference to something that already occurred;<br />

children at this stage do not possess <strong>the</strong> capacity to<br />

talk about specific experiences so <strong>the</strong>y envision <strong>the</strong><br />

future as non-existent or as unimagined (Nelson,<br />

1989a, 1997, 2000). This challenge elicits <strong>the</strong><br />

understanding in children that one remains constant<br />

while still possessing <strong>the</strong> ability to project oneself to<br />

<strong>the</strong> past and <strong>the</strong> future (Nelson, 2000).<br />

According to McCormack and Hoerl (1999) <strong>the</strong>se<br />

experiences give rise to <strong>the</strong> concept <strong>of</strong> temporal<br />

order. It has been predicted, however, that temporal<br />

order is independent <strong>of</strong> children’s nonverbal<br />

intelligence and that it may be fur<strong>the</strong>r related to <strong>the</strong>ir<br />

123<br />

verbal capacity and experiences in daily life (Nelson,<br />

2007). Source memory on <strong>the</strong> o<strong>the</strong>r hand refers to <strong>the</strong><br />

ability <strong>of</strong> an individual to recall <strong>the</strong> context by which<br />

information was obtained, such as when and/or<br />

where an object was encountered (Cycowicz,<br />

Friedman, Snodgrass, & Duff, 2000). The results <strong>of</strong><br />

<strong>the</strong>ir study suggest that source memory in children<br />

tends to improve with age.<br />

The present study seeks to demonstrate that<br />

developments in temporal cognition, diachronic<br />

transformation, syn<strong>the</strong>sis, are interrelated and<br />

independent <strong>of</strong> age and nonverbal intelligence.<br />

II. METHODS<br />

Participants in <strong>the</strong> current study were 100 children<br />

between <strong>the</strong> ages <strong>of</strong> 5 and 10. The children were<br />

recruited ei<strong>the</strong>r from a local school or were brought<br />

in by Brooklyn College undergraduate students for<br />

course extra credit. The children were given $10 gift<br />

cards for <strong>the</strong>ir participation. Parents and guardians<br />

signed an informed consent sheet and children gave<br />

verbal assent. The children completed Boucher,<br />

Pons, Lind, and Williams’ (2007) diachronic thinking<br />

tasks:<br />

A. TENDENCY<br />

Participants were asked to draw inferences regarding<br />

prior and subsequent moments in time upon<br />

examining a picture <strong>of</strong> a man at <strong>the</strong> beach. In <strong>the</strong> first<br />

task participants deduced from looking at <strong>the</strong> image<br />

what is happening and what is going to happen in <strong>the</strong><br />

picture. In task 2 <strong>the</strong> children had to determine<br />

whe<strong>the</strong>r it was summer or winter and provide reasons<br />

for <strong>the</strong>ir response. Subsequently, participants had to<br />

determine if a surfer at <strong>the</strong> beach is a child or an<br />

adult. Lastly, <strong>the</strong> children were asked to explain why<br />

<strong>the</strong> man on <strong>the</strong> beach was lying on a towel and also<br />

why he brought an umbrella with him. For <strong>the</strong> third<br />

task <strong>the</strong> children were shown a blank piece <strong>of</strong> paper<br />

and asked to think about what might be observed<br />

fur<strong>the</strong>r along <strong>the</strong> beach. In <strong>the</strong> final task <strong>of</strong> this<br />

section, participants were asked to place six picture


cards in order depicting <strong>the</strong> man’s past and future<br />

activities.<br />

B. TRANSFORMATION<br />

In <strong>the</strong> first task participants were asked to draw a<br />

tree, and <strong>the</strong>n asked to show <strong>the</strong> tree’s life cycle by<br />

also drawing a before and after image. In tasks 2 and<br />

3 participants drew qualitatively different trees and<br />

placed picture cards <strong>of</strong> a tree in <strong>the</strong> appropriate<br />

developmental stages.<br />

C. SYNTHESIS<br />

Participants were asked to syn<strong>the</strong>size a group <strong>of</strong><br />

pictures. In <strong>the</strong> first task <strong>the</strong> children were shown 4<br />

different sets <strong>of</strong> cards (Boucher, et al., 2007). The<br />

pictures depicted parts <strong>of</strong> a face, house, pot plant, or<br />

farm. For <strong>the</strong> second task <strong>the</strong> children viewed six<br />

pictures depicting <strong>the</strong> day <strong>of</strong> a man at <strong>the</strong> beach. The<br />

first image was <strong>of</strong> <strong>the</strong> man arriving at <strong>the</strong> beach and<br />

<strong>the</strong> final image was <strong>of</strong> <strong>the</strong> man in his bed sleeping.<br />

Children were asked what do all <strong>of</strong> <strong>the</strong> pictures<br />

toge<strong>the</strong>r show. The test was terminated when <strong>the</strong><br />

participants showed a capacity for syn<strong>the</strong>sis.<br />

D. OTHER TASKS AND NEUROPSYCHOLOGICAL<br />

TESTS<br />

Upon completion <strong>of</strong> <strong>the</strong>se tasks, participants<br />

undertook Friedman’s temporal cognition tasks<br />

(1990; 2000) <strong>the</strong>reby judging pictured events and<br />

different times <strong>of</strong> <strong>the</strong> year. A similar task was also<br />

administered to judge <strong>the</strong> temporal order <strong>of</strong> daily<br />

events. The Peabody Picture Vocabulary Test-Fourth<br />

Edition (PPVT-4) was utilized to estimate<br />

participants’ verbal intelligence and <strong>the</strong> Test <strong>of</strong><br />

Nonverbal Intelligence (TONI-3) was used to<br />

measure nonverbal reasoning skills. Participants were<br />

also given a recognition test to assess whe<strong>the</strong>r <strong>the</strong>y<br />

recalled seeing animate versus inanimate images<br />

from a previous test trial. As a final point,<br />

participants were asked to determine where <strong>the</strong>y had<br />

observed images that were previously presented. This<br />

test was given to assess participants’ source memory<br />

capacity.<br />

III. RESULTS AND DISCUSSION<br />

Pearson partial coefficients were computed using age<br />

and TONI-3 scores as covariates. Preliminary data<br />

indicates that most tasks <strong>of</strong> temporal cognition are<br />

inter-correlated. There were only two intercorrelations<br />

<strong>of</strong> diachronic thinking but no significant<br />

correlations between diachronic thinking and<br />

temporal cognition tasks. The Peabody tests <strong>of</strong> verbal<br />

intelligence were significantly correlated with most<br />

tasks <strong>of</strong> temporal order. Conversely, <strong>the</strong> TONI-3<br />

124<br />

tasks <strong>of</strong> nonverbal reasoning skills were not intercorrelated<br />

with temporal cognition.<br />

The current data support <strong>the</strong> hypo<strong>the</strong>sis that temporal<br />

order developments are dependent upon experience<br />

and not upon <strong>the</strong> developmental stage. Results reveal<br />

that children’s temporal cognition is independent <strong>of</strong><br />

<strong>the</strong>ir level <strong>of</strong> nonverbal intelligence and age. Data<br />

collection is ongoing and we expect that findings will<br />

enhance understanding <strong>of</strong> <strong>the</strong> typical development <strong>of</strong><br />

various cognitive skills and inform treatment <strong>of</strong><br />

autism spectrum and o<strong>the</strong>r childhood disorders.<br />

ACKNOWLEDGMENTS<br />

This work was supported in part by <strong>the</strong> New York<br />

City Louis Stokes Alliance.<br />

REFERENCES<br />

[1] C. M. Atance, & D. K. O'Neill, Preschoolers’<br />

talk about future situations. First Language, 1,<br />

pages 5-18, 2005.<br />

[2] K. Nelson. Young Minds in Social Worlds:<br />

Experience, Meaning, and Memory. February<br />

2007.<br />

[3] J. Boucher, F. Pons, S. Lind, & D. Williams.<br />

Temporal cognition in children with autistic<br />

spectrum disorders: Tests <strong>of</strong> diachronic<br />

perspective taking. Journal <strong>of</strong> Autism and<br />

Developmental Disorders, 37, pages 1413-1429,<br />

2007.<br />

[4] T. McCormack, & C. Hoerl. Memory and<br />

Temporal Perspective: The Role <strong>of</strong> Temporal<br />

Frameworks in Memory Development.<br />

Developmental Review, 19, pages 154-182,<br />

March 1999.<br />

[5] W. J. Friedman. The development <strong>of</strong> children's<br />

knowledge <strong>of</strong> <strong>the</strong> times <strong>of</strong> future events. Child<br />

Development, 7, 1, pages 913-932, 2000.<br />

[6] L. Brown, R. Sherbenou, & S. Johnsen. Test <strong>of</strong><br />

nonverbal intelligence, a language free measure<br />

<strong>of</strong> cognitive ability. (3rd ed.). Austin: Pro-ed,<br />

1997.<br />

[7] K. Nelson. Narrative, time and <strong>the</strong> emergence <strong>of</strong><br />

<strong>the</strong> encultured self. Culture & Psychology, 6,<br />

pages 183-196, 2000.<br />

[8] Y. M. Cycowicz, D. Friedman, J. G. Snodgrass,<br />

and M. Duff. Recognition and Source Memory<br />

for pictures in children and adults.<br />

Neuropsychologia, pages 255-267, New York,<br />

NY, August 2000.


Generalizing Sudoku to Three Dimensions<br />

Abstract —The aim <strong>of</strong> this paper is to<br />

generalize <strong>the</strong> logic puzzle Sudoku from twodimensions<br />

into three-dimensions. Our<br />

objective is to find <strong>the</strong> most efficient way to<br />

generate and solve <strong>the</strong>se puzzles. Some <strong>of</strong><br />

our o<strong>the</strong>r goals are to find how many<br />

possible solutions exist, and if and how ca-n<br />

we generate logic solvable puzzles? We have<br />

designed a set <strong>of</strong> rules and constraints for a<br />

three-dimensional Sudoku puzzle and have<br />

found m-any unique solutions using <strong>the</strong><br />

method <strong>of</strong> simulated annealing.<br />

I. INTRODUCTION<br />

Tiffany A Lambert (Dr. Paula Whitlock)<br />

Brooklyn College<br />

The City University <strong>of</strong> New York<br />

Brooklyn, NY, USA<br />

Email: LT0180@bcmail.brooklyn.cuny.edu<br />

Sudoku is a logic puzzle where every row,<br />

column and block must contain <strong>the</strong> numbers 1-9,<br />

which means no number can be repeated more<br />

than once in a row, column or block. Sudoku can<br />

actually be changed to consist <strong>of</strong> smaller or larger<br />

blocks and can easily be defined.<br />

The general problem <strong>of</strong> solving Sudoku is<br />

known to be NP-Complete [6], and <strong>the</strong>re have<br />

been many papers and methods published that try<br />

to solve Sudoku optimally. The Sudoku examples<br />

most people are familiar with are called logicsolvable<br />

because a logical reasoning approach<br />

will lead to a solution. O<strong>the</strong>rs, however, can be<br />

solved using many o<strong>the</strong>r algorithms. There has<br />

also been plenty <strong>of</strong> computational work done to<br />

find <strong>the</strong> number <strong>of</strong> unique Sudoku puzzle<br />

possibilities [1] [5].<br />

II. METHOD<br />

Generalizing a three-dimensional Sudoku can be<br />

more challenging because it has a much larger<br />

state-space than a regular two-dimensional<br />

Sudoku. Our design has 6 faces, each where n =<br />

4. Every face <strong>of</strong> <strong>the</strong> puzzle is an individual puzzle<br />

within itself and follows <strong>the</strong> traditional rules <strong>of</strong><br />

Sudoku. Our added constraint was on <strong>the</strong> cube’s<br />

edges, which relate <strong>the</strong> faces to one ano<strong>the</strong>r. The<br />

edges on each adjacent cell must be identical.<br />

125<br />

A. SIMULATED ANNEALING<br />

Rhyd Lewis uses a technique called Simulated<br />

Annealing [3] to solve Sudoku. Simulated<br />

Annealing is an algorithm that employs both<br />

iterative improvement, as well as Monte Carlo<br />

methods [2]. The algorithm is based on making<br />

random swaps toward an optimal solution. It<br />

calculates an initial score and accepts a new<br />

move if <strong>the</strong> score has decreased. The algorithm<br />

accepts a move based on <strong>the</strong> following probabilistic<br />

function:<br />

−ΔE<br />

T<br />

P = e (1)<br />

where T is <strong>the</strong> temperature and �E is <strong>the</strong> change<br />

in energy or cost. The cost approaches 0 as <strong>the</strong><br />

algorithm approaches a solution.<br />

B. NEW CONSTRAINTS<br />

Lewis’ Sudoku solver is capable <strong>of</strong> producing a<br />

solution for both partially filled, as well as<br />

completely empty grids. We altered <strong>the</strong><br />

parameters and constraints in order to<br />

accommodate our three-dimensional model. We<br />

added additional constraints that accommodated<br />

our new rules for <strong>the</strong> edges. When coding this,<br />

we used <strong>the</strong> original two-dimensional array, but<br />

2 2<br />

increased <strong>the</strong> size to 2n × 2n<br />

. The array was<br />

divided into four quadrants, one for each face,<br />

with <strong>the</strong> fourth quadrant unused. For practical<br />

purposes, we have worked with only three faces<br />

<strong>of</strong> <strong>the</strong> cube.<br />

In order to configure <strong>the</strong> edge constraint into <strong>the</strong><br />

total cost, we added an edge score to <strong>the</strong> row and<br />

column score. For every edge cell that does not<br />

meet <strong>the</strong> constraint, we added one point to <strong>the</strong><br />

total edge score. This edge score told us how<br />

close we were to a solution.<br />

C. COOLING SCHEDULES<br />

In <strong>the</strong> original paper by Lewis <strong>the</strong> temperature<br />

was cooled by 1% after every Markov chain,<br />

which was efficient for two-dimensional 9 x 9<br />

Sudoku grids, but after we added our new


constraints and increased <strong>the</strong> size <strong>of</strong> <strong>the</strong> statespace,<br />

<strong>the</strong> algorithm was no longer efficient. We<br />

tested several constant cooling rates less than<br />

99%. These faster cooling rates did improve <strong>the</strong><br />

time, but if <strong>the</strong> substance was cooled to quickly,<br />

it did not produce <strong>the</strong> desired results. This led us<br />

to experiment with dynamic cooling schedules<br />

such that, <strong>the</strong> temperature decreases by larger<br />

amounts initially, and as <strong>the</strong> cooling rate<br />

gradually reduces, <strong>the</strong> temperature decreases at a<br />

much slower rate toward <strong>the</strong> end. To achieve this,<br />

we made <strong>the</strong> cooling rate a function <strong>of</strong> cost, so<br />

that as <strong>the</strong> change in cost increases from its initial<br />

state, so does <strong>the</strong> cooling rate.<br />

III. RESULTS<br />

The runtimes for both rule constraints produce<br />

similar outcomes. The average times using<br />

constant cooling rates are shown below. Reheats<br />

were performed if <strong>the</strong> algorithm could not find a<br />

solution on <strong>the</strong> first run within <strong>the</strong> allocated time.<br />

The results are near linear with <strong>the</strong> exception <strong>of</strong><br />

when it reaches its cooling limit. From <strong>the</strong> figures<br />

above, it is clear that <strong>the</strong> anything less than 80%<br />

causes <strong>the</strong> system to cool too quickly.<br />

Trial and error led us to new cooling schedules<br />

that have been <strong>the</strong> most efficient thus far. This<br />

schedule improved <strong>the</strong> runtime by more than<br />

90% from <strong>the</strong> initial cooling schedule used in<br />

Metaheuristics [3].<br />

Time (sec)<br />

900<br />

800<br />

700<br />

600<br />

500<br />

400<br />

300<br />

200<br />

100<br />

0<br />

Average Runtimes<br />

variable 80% 85% 90% 95% 99%<br />

Cooling Rate<br />

Figure 3: Average runtimes <strong>of</strong> cooling schedules<br />

IV. DISCUSSION<br />

According to Taking Sudoku seriously [1], and<br />

The ma<strong>the</strong>matics <strong>of</strong> Sudoku [7], we know that as<br />

n increases, so does <strong>the</strong> number <strong>of</strong> puzzle<br />

combinations. Exactly how many unique puzzles<br />

exist is a problem that still needs to be explored.<br />

As we increase to three dimensions, <strong>the</strong> number<br />

<strong>of</strong> possibilities grows factorially, but several<br />

solutions can be eliminated due to repetition.<br />

126<br />

The issue <strong>of</strong> logic solvability is still a question<br />

that needs to be fur<strong>the</strong>r explored. It has not been<br />

proven, but <strong>the</strong> smallest known amount <strong>of</strong><br />

numbers given to find a unique Sudoku solution<br />

is 17 [7]. Computing <strong>the</strong> minimum number <strong>of</strong><br />

givens for a unique solution is an optimization<br />

problem that is NP-complete.<br />

A. CONCLUSION<br />

From our experiments with <strong>the</strong> cooling schedule,<br />

we have determined that both, change in score,<br />

and cooling rates are key factors in finding an<br />

optimal solution. As <strong>the</strong> difference between <strong>the</strong><br />

initial score and <strong>the</strong> current score becomes larger,<br />

<strong>the</strong> cooling rate should become stricter in order to<br />

restrict jumps to a less optimal solution. By<br />

changing <strong>the</strong> cooling rate dynamically, we reduce<br />

<strong>the</strong> amount <strong>of</strong> Marcov chains that are started,<br />

thus decreasing <strong>the</strong> run time.<br />

ACKNOWLEDGEMENTS<br />

We acknowledge <strong>the</strong> Louis Stoke Alliance for<br />

Minority Participation and <strong>the</strong> National Science<br />

Foundation for providing us with this grant. We<br />

would also like to thank Dr. Murray Gross for his<br />

insight and recommendations.<br />

REFERENCES<br />

[1] Felgenhauer, Bertram, and Frazer Jarvis.<br />

"Ma<strong>the</strong>matics <strong>of</strong> Sudoku I." Ma<strong>the</strong>matical<br />

Spectrum 39(2006): 15-22.<br />

[2] Kirkpatrick, S., C. D. Gelatt Jr., and M. P.<br />

Vecchi. "Optimization by Simulated Annealing."<br />

Science 220(1983): 671 - 680.<br />

[3] Lewis, Rhyd. "Metaheuristics can solve<br />

Sudoku puzzles." Journal <strong>of</strong> Heuristics 13(2006):<br />

387-401.<br />

[4] Mantere, T and J. Koljonen. “Solving and<br />

rating Sudoku puzzles with genetic algorithms.”<br />

<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 12 th Finnish Artificial<br />

Intelligence <strong>Conference</strong> (2006): 86-92.<br />

[5] Russell, Ed, and Frazer Jarvis. "Ma<strong>the</strong>matics<br />

<strong>of</strong> Sudoku II." Ma<strong>the</strong>matical Spectrum 39(2006):<br />

54-58.<br />

[6] Santos-Garcia, G. and M. Palomino.<br />

“Solving Sudoku Puzzles with Rewriting Rules.”<br />

Electronic Notes in Theoretical Computer<br />

Science 176(2007): 79-93.<br />

[7] Taalman, L. “Taking Sudoku Seriously.”<br />

Math Horizons 15(2007): 5-9.<br />

[8] “Sudoku Solver by Logic.” (2006)<br />

http://www.sudokusolver.co.uk


Chapter 2<br />

Material Sciences


Nonadditivity <strong>of</strong> Quantum Phases for mixed States<br />

J. Klepp, S. Sponar, S. Filipp, M. Lettner, G. Badurek and Yuji Hasegawa<br />

Atominstitut<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {jklepp,hasegawa}@ati.ac.at<br />

Abstract — In a neutron polarimetry experiment mixed state<br />

relative phases between spin eigenstates are determined. We<br />

consider evolutions leading to purely geometric, purely dynamical<br />

and combined phases. It is experimentally demonstrated<br />

that <strong>the</strong> sum <strong>of</strong> <strong>the</strong> geometric and dynamical phases - both obtained<br />

in separate measurements - is not equal to <strong>the</strong> associated<br />

total phase which is obtained from a single measurement, unless<br />

<strong>the</strong> system is in a pure state. Therefore, surprisingly, mixed<br />

state phases are not additive.<br />

I. INTRODUCTION<br />

Evolving quantum systems acquire two kinds <strong>of</strong> phase<br />

factors: (i) <strong>the</strong> dynamical phase which depends on <strong>the</strong><br />

dynamical properties <strong>of</strong> <strong>the</strong> system - like energy or time -<br />

during a particular evolution, and (ii) <strong>the</strong> geometric phase<br />

which only depends on <strong>the</strong> evolution path <strong>the</strong> system<br />

takes in state space on its way from <strong>the</strong> initial to <strong>the</strong> final<br />

state [1, 2]. Due to its robustness against noise [3] <strong>the</strong><br />

geometric phase is an excellent candidate to be utilized<br />

for logic gate operations in quantum information science<br />

[4]. Thus, a rigorous investigation <strong>of</strong> all its properties is<br />

<strong>of</strong> great importance.<br />

In addition to an approach by Uhlmann [5] a new concept<br />

<strong>of</strong> phase for mixed input states based on interferometry<br />

was developed by Sjöqvist et al. [6]. Here,<br />

each eigenvector <strong>of</strong> <strong>the</strong> initial density matrix independently<br />

acquires a geometric phase. The total mixed state<br />

phase is a weighted average <strong>of</strong> <strong>the</strong> individual phase factors.<br />

This concept is <strong>of</strong> great significance for such experimental<br />

situations or technical applications where pure<br />

state <strong>the</strong>ories may imply strong idealizations. Theoretical<br />

predictions have been tested using NMR and singlephoton<br />

interferometry [7, 8]. Here, we report on measurements<br />

<strong>of</strong> nonadiabatic and noncyclic geometric, dynamical<br />

and combined phases. These depend on noise<br />

strength in state preparation, defining <strong>the</strong> degree <strong>of</strong> polarization,<br />

<strong>the</strong> purity, <strong>of</strong> <strong>the</strong> neutron input state. In particular,<br />

our experiment demonstrates for <strong>the</strong> first time that<br />

<strong>the</strong> geometric and dynamical mixed state phases Φg and<br />

Φd, resulting from separate measurements, are not additive<br />

[9], because <strong>the</strong> phase resulting from a single, cumulative,<br />

measurement differs from Φg +Φd. These striking<br />

results are published in [10].<br />

II. THEORETICAL CONSIDERATIONS<br />

A polarized neutron beam propagating in y-direction,<br />

interacting with static magnetic fields �B(y), undergoes<br />

129<br />

Figure 1: Schematic view <strong>of</strong> <strong>the</strong> measurement setup with<br />

overall guide field Bz, polarizer P, three DC-coils to im-<br />

plement unitary operations U1, U †<br />

1 ,Uφ, analyzer A and<br />

detector D. Greek letters denote spin rotation angles.<br />

Shifting <strong>the</strong> second coil induces an additional dynamical<br />

phase η resulting in intensity oscillations. The desired<br />

phase φ is determined from <strong>the</strong>ir minima and maxima.<br />

Zeeman splitting. This results in a momentum shift<br />

k± � k0∓Δk, where k0 is <strong>the</strong> momentum <strong>of</strong> <strong>the</strong> free particle<br />

and Δk = mμ|�B(y)|/¯h 2 k. Δk can be detected from<br />

spinor precession. We focus on <strong>the</strong> resulting evolution <strong>of</strong><br />

superposed spin eigenstates sometimes described as Larmor<br />

precession <strong>of</strong> <strong>the</strong> polarization vector �r = 〈ϕ|�σ|ϕ〉,<br />

where �σ =(σx,σy,σz) is <strong>the</strong> Pauli vector operator.<br />

Consider <strong>the</strong> experimental setup shown in Figure 1.<br />

In [11] it is stated that with such an apparatus one can<br />

obtain <strong>the</strong> pure state Pancharatnam (total) phase [1] between<br />

spin eigenstates <strong>of</strong> neutrons, induced by a SU(2)<br />

transformation Uφ(ξ,δ,ζ). The phase can be written as<br />

a function <strong>of</strong> <strong>the</strong> maximum Imax and minimum Imin <strong>of</strong> <strong>the</strong><br />

intensity oscillations, exhibited by applying an auxiliary<br />

phase shift η. The intensity only depends on <strong>the</strong> SU(2)<br />

parameters ξ, δ and ζ - set by choosing <strong>the</strong> spin rotation<br />

angles in <strong>the</strong> second coil and <strong>the</strong> additional propagation<br />

distance within <strong>the</strong> guide field Bz, respectively. η is varied<br />

by stepwise translation <strong>of</strong> <strong>the</strong> second coil.<br />

A neutron beam with incident purity r = |�r| along <strong>the</strong><br />

+z-axis (�r =(0, 0,r)) is described by <strong>the</strong> density operator<br />

ρin(r) =1/2(1l + rσz). For mixed input states,<br />

0 ≤ r


sity, one obtains <strong>the</strong> mixed state phase [12]. For r =1,<br />

Figure 2: Evolutions Uφ associated to: a) Purely (noncyclic)<br />

geometric phase (2ξ = π/2). b) Combinations <strong>of</strong><br />

dynamical and geometric phase (0 < 2ξ


Complexation <strong>of</strong> trivalent Cm and Eu with extraction ligands for <strong>the</strong><br />

Partitioning & Transmutation strategy<br />

S. Trumm a,b , P. J. Panak a,b , A. Geist a , R. Klenze a , Th. Fanghänel b,c<br />

a<br />

Institut für Nukleare Entsorgung, Forschungszentrum Karlsruhe, D-76021 Karlsruhe,<br />

Germany<br />

b<br />

Ruprecht-Karls Universität, 69120 Heidelberg, Germany<br />

c<br />

Institute for Transuranium Elements, European Commission, 76125 Karlsruhe, Germany<br />

The partitioning and transmutation (P&T) strategy [] i is a prosperous strategy for <strong>the</strong> treatment<br />

<strong>of</strong> spent nuclear fuels. The aim is to reduce <strong>the</strong> long-term radiotoxicity <strong>of</strong> <strong>the</strong> nuclear material<br />

by separating long-lived actinides (Np, Pu, Am and Cm isotopes) and transmuting <strong>the</strong>m into<br />

shorter-lived isotopes by nuclear fission. In this way <strong>the</strong> long-term radiotoxicity <strong>of</strong> high-level<br />

nuclear wastes can be significantly reduced, which has a positive impact on <strong>the</strong> storage in<br />

final geological repositories.<br />

The separation <strong>of</strong> trivalent americium and curium from <strong>the</strong> lanthanides is a key step within<br />

this process. Due to <strong>the</strong>ir similar chemical behavior, highly selective ligands are necessary to<br />

separate trivalent actinides from lanthanides. This separation step, is not feasible using<br />

common oxygen-donor ligands extractants, only s<strong>of</strong>t donor ligands bearing nitrogen or sulfur<br />

atoms induce <strong>the</strong> desired selectivity.<br />

[ ] ii<br />

Alkylated 2,6-Bis(1,2,4-triazine-3-yl)pyridines (BTPs) and 6,6’-Bis(1,2,4-triazin-3yl)bipyridines<br />

(BTBPs, figure 1) have shown high performance in <strong>the</strong> separation <strong>of</strong> trivalent<br />

minor actinides (americium, curium) from lanthanides and were <strong>the</strong> first N-donor extracting<br />

agents to extract and separate trivalent actinides and lanthanides from nitric acideous<br />

[ ],[ ]<br />

solutions. iii iv However, <strong>the</strong> origin <strong>of</strong> <strong>the</strong> observed selectivity is not yet completely<br />

understood. To set <strong>the</strong> foundation for <strong>the</strong> basic understanding <strong>of</strong> <strong>the</strong> selectivity, fundamental<br />

research towards <strong>the</strong> complexation <strong>of</strong> <strong>the</strong>se ligands with trivalent actinides and lanthanides is<br />

necessary.<br />

Fig. 1: Structure <strong>of</strong> BTP and BTBP ligands<br />

BTP BTBP<br />

In this work we investigated <strong>the</strong> complexation <strong>of</strong> Cm(III) and Eu(III) using <strong>the</strong> time-resolved<br />

laser fluorescence spectroscopy (TRLFS). This method allows monitoring complex species in<br />

organic solution at very low concentrations. To quantify <strong>the</strong> differences in complexation<br />

between BTPs and BTBPs, complexation constants have been derived from titration<br />

experiments. Two new BTP ligands have been syn<strong>the</strong>sized carrying substituents on <strong>the</strong><br />

pyridine ring altering <strong>the</strong> electronic structure <strong>of</strong> <strong>the</strong> aromatic system. This allowed us for <strong>the</strong><br />

first time to investigate <strong>the</strong> influence <strong>of</strong> electronic parameters on <strong>the</strong> complexation and<br />

extraction chemistry <strong>of</strong> Cm(III) and Eu(III). Fur<strong>the</strong>r we were able to gain <strong>the</strong>rmodynamic<br />

parameters <strong>of</strong> complexation <strong>of</strong> BTBP ligands with trivalent Cm and Eu by measuring <strong>the</strong><br />

equilibrium between <strong>the</strong> complex species in solution at different temperatures.<br />

131


The results show a difference between <strong>the</strong> complexation constants <strong>of</strong> BTP ligands with <strong>the</strong><br />

two metal ions in <strong>the</strong> range <strong>of</strong> three to five orders <strong>of</strong> magnitude, which is in excellent<br />

agreement with <strong>the</strong> selectivity found in extraction experiments. The complexation <strong>of</strong> Eu(III)<br />

shows a significant dependence on <strong>the</strong> electron density <strong>of</strong> <strong>the</strong> coordinating nitrogen atoms.<br />

Electron withdrawing substituents cause a decrease <strong>of</strong> complexation constants whereas<br />

electron donating substituents lead to higher ones. Also in extraction experiments <strong>the</strong><br />

influence <strong>of</strong> <strong>the</strong> substituents is observed. This information is important for <strong>the</strong> understanding<br />

<strong>of</strong> selectivity and also for <strong>the</strong> optimization <strong>of</strong> extraction ligands with respect to future<br />

industrial processing. The results <strong>of</strong> our studies on <strong>the</strong> complexation <strong>of</strong> BTBP ligands with<br />

Cm(III) and Eu(III) show higher changes in <strong>the</strong> free enthalpy <strong>of</strong> complexation for Cm(III) than<br />

for Eu(III). The <strong>the</strong>rmodynamically favored complexation <strong>of</strong> trivalent Cm is again in good<br />

agreement with extraction experiments.<br />

i<br />

Actinide and fission product partitioning and transmutation, status and assessment report; OECD<br />

Nuclear Energy Agency; Paris, 1999.<br />

ii<br />

Nash, K. L. Solvent Extr. Ion Exch. 1993, 11 (4), 729�768.<br />

iii<br />

Kolarik Z.; Müllich U.; Gassner F. Solvent Extr. Ion Exch. 1999, 17 (1), 23�32.<br />

iv<br />

Kolarik Z.; Müllich U.; Gassner F. Solvent Extr. Ion Exch. 1999, 17 (5), 1155�1170.<br />

132


COMPLEXATION OF Cm(III) WITH NITRATE AND FLUORIDE AT ELEVATED<br />

TEMPERATURES STUDIED BY<br />

TIME-RESOLVED LASER FLUORESCENCE SPECTROSCOPY<br />

A. Skerencak a,c , P. J. Panak a,c , V. Neck a , R. Klenze a , Th. Fanghänel b,c<br />

a Institut für Nukleare Entsorgung, Forschungszentrum Karlsruhe, D-76021 Karlsruhe, Germany<br />

b Institute for Transuranium Elements, European Commission, 76125 Karlsruhe, Germany<br />

c Ruprecht-Karls Universität, 69120 Heidelberg, Germany<br />

Fundamental understanding <strong>of</strong> <strong>the</strong> geochemical processes in <strong>the</strong> near and far field <strong>of</strong> a nuclear waste repository<br />

and <strong>the</strong>ir <strong>the</strong>rmodynamic quantification is necessary for <strong>the</strong> assessment <strong>of</strong> <strong>the</strong> long-term migration behavior <strong>of</strong><br />

actinides in <strong>the</strong> geosphere. The migration <strong>of</strong> <strong>the</strong> nuclides is limited by <strong>the</strong>ir low solubility and interaction with<br />

inorganic material (i.e. interaction with mineral surfaces (sorption, incorporation)). Conversely, aqueous<br />

complexation with proper ligands can increase <strong>the</strong>ir mobility. As natural waters contain a variety <strong>of</strong> inorganic<br />

ligands such as OH - , CO3 2- , SO4 2- , Cl - , F - , H2PO4 - , H3SiO4 - , HCO3 - , and NO3 - , <strong>the</strong> complexation <strong>of</strong> actinides with<br />

inorganic ligands is <strong>of</strong> particular interest for a better understanding <strong>of</strong> <strong>the</strong> migration behaviour <strong>of</strong> actinides under<br />

natural conditions [1] . Extensive knowledge on <strong>the</strong> participating chemical reactions and <strong>the</strong> corresponding<br />

<strong>the</strong>rmodynamic data is needed. In <strong>the</strong> past few years considerable progress in establishing a <strong>the</strong>rmodynamic<br />

database for lanthanides and actinides has been achieved [2] . Never<strong>the</strong>less, <strong>the</strong>se data is mostly restricted to<br />

ambient temperatures. Due to <strong>the</strong> heat production in <strong>the</strong> close vicinity <strong>of</strong> radioactive waste, complexation<br />

reactions <strong>of</strong> actinides at elevated temperatures are <strong>of</strong> major importance for modelling <strong>the</strong> migration behaviour <strong>of</strong><br />

actinides under near field conditions over a long time period.<br />

We utilized a newly designed high temperature and high pressure cell coupled with time resolved laser<br />

fluorescence spectroscopy (TRLFS). TRLFS is a versatile tool for determining <strong>the</strong> speciation <strong>of</strong> actinides in <strong>the</strong><br />

submicromolar concentration range and provides information on reaction mechanisms without disturbing <strong>the</strong><br />

chemical equilibrium [3] . Because <strong>of</strong> its outstanding spectroscopic properties, Cm(III) is chosen as a<br />

representative for trivalent actinides. In this work we have investigated <strong>the</strong> complexation <strong>of</strong> Cm(III) with<br />

fluoride in <strong>the</strong> temperature range <strong>of</strong> 20 to 90°C and with nitrate in <strong>the</strong> temperature region <strong>of</strong> 20 to 200°C. The<br />

speciation is done via peak deconvolution. The equilibrium constants are determined and extrapolated to zero<br />

ionic strength using linear SIT Regression. Modeling <strong>of</strong> <strong>the</strong> data is done according to <strong>the</strong> Van't H<strong>of</strong>f equation at<br />

temperatures up to 90°C. Higher temperatures required an extended Van't H<strong>of</strong>f equation for adequate modeling.<br />

From this approaches valuable <strong>the</strong>rmodynamic data (�rH 0 m, �rS 0 m, �rG 0 m, �rC 0 p,m) is derived at elevated<br />

temperatures.<br />

The results show a strong increase in complexation <strong>of</strong> Cm(III) with increasing temperatures resulting in a<br />

significant increase <strong>of</strong> <strong>the</strong> stability constants at elevated temperatures. Due to high temperatures in <strong>the</strong> close<br />

vicinity <strong>of</strong> a nuclear waste repository, <strong>the</strong> increased complexation leads to an increased mobility <strong>of</strong> <strong>the</strong> actinides<br />

under near field conditions. Therefore, <strong>the</strong> determined <strong>the</strong>rmodynamic data is crucial for a reliable description <strong>of</strong><br />

<strong>the</strong> migration behaviour <strong>of</strong> radionuclides under near field conditions and has to be taken into account for <strong>the</strong> long<br />

term safety assessment <strong>of</strong> a nuclear waste repository.<br />

[1] Fanghänel, Th., Kim, J.-I., Journal <strong>of</strong> Alloys and Compounds, Vol. 271-273, 728-737, (1998).<br />

[2] Guillaumont, R., Fanghänel, Th., Neck, V., Fuger, J., Palmer, D.A., Gren<strong>the</strong>, I., Rand, M.H., Chemical<br />

Thermodynamics 5, Nuclear Energy Agency, Elsevier Science Publisher, (2003).<br />

[3] Edelstein, N. M., Klenze, R., Fanghänel, Th., Hubert, S., Coord. Chem. Rev. 250, 948, (2006).<br />

133


134


Development <strong>of</strong> �-(Al,Cr)2O3 solid solutions<br />

by reactive magnetron sputtering<br />

Dominic Diechle, Michael Stüber, Harald Leiste and Sven Ulrich (Faculty Mentor)<br />

Institute for Materials Research I<br />

Forschungszentrum Karlsruhe<br />

Karlsruhe, Germany<br />

Email: dominic.diechle@imf.fzk.de<br />

Abstract — Hard, tough, wear and corrosion<br />

resistant thin film materials are very important for<br />

many applications such as cutting tools. We present<br />

a new combinatorial approach for <strong>the</strong> deposition <strong>of</strong><br />

solid solution streng<strong>the</strong>ned �-(Al,Cr)2O3 thin films<br />

grown by reactive r.f. magnetron sputtering in an<br />

argon-oxygen atmosphere. The deposition<br />

experiments are carried out with a Leybold Z 550<br />

PVD machine for <strong>the</strong> sputtering from a segmented<br />

target. The growth, phase formation and<br />

constitution <strong>of</strong> <strong>the</strong> thin films has been characterized<br />

by various methods.<br />

I. INTRODUCTION<br />

The development <strong>of</strong> superior coatings for high<br />

performance cutting tools is a key element for<br />

significant advances in metal working. Aluminum<br />

oxide thin films, mostly deposited by chemical<br />

vapour deposition (CVD) methods, are an established<br />

technology in this application field. Recently, <strong>the</strong><br />

physical vapour deposition (PVD) syn<strong>the</strong>sis <strong>of</strong><br />

aluminum oxide and derivate coatings is attracting<br />

large scientific and technical interest. Especially<br />

aluminum-chromium oxide thin films are promising<br />

candidates for fur<strong>the</strong>r significant improvement.<br />

Physical vapour deposition allows a nanoscale design<br />

<strong>of</strong> thin films made <strong>of</strong> ceramics, metals and <strong>the</strong>ir<br />

alloys, far away from <strong>the</strong>rmodynamic equilibrium<br />

conditions. The “Holleck triangle” classifies hard<br />

materials with regard to <strong>the</strong>ir bonding [1]. For <strong>the</strong><br />

work presented here, we intend to combine two<br />

predominant ionic hard materials in new coating<br />

structures. The bonding in aluminum oxide and<br />

chromium oxide is a mixture <strong>of</strong> ionic and covalent<br />

bonding. Thus, <strong>the</strong> properties <strong>of</strong> <strong>the</strong>se materials are a<br />

very good compromise between <strong>the</strong>rmal stability and<br />

chemical inertness due to <strong>the</strong> ionic bonding share and<br />

an excellent high temperature toughness and<br />

hardness due to <strong>the</strong> covalent bonding share. The<br />

section �-Al2O3/�-Cr2O3 in <strong>the</strong> Al-Cr-O phase<br />

diagram shows two important regions, refeering to a<br />

miscibility gap, and, to a solid solution, respectively<br />

[2]. In <strong>the</strong> field <strong>of</strong> <strong>the</strong> miscibility gap below 1300 °C<br />

135<br />

<strong>the</strong> phase consists <strong>of</strong> two coexisting solid states �-<br />

Al2O3 (corundum) and �-Cr2O3 (escolaite) [3].<br />

Above <strong>the</strong> miscibility gap exists <strong>the</strong> solid solution<br />

(Al1-x,Crx)2O3 <strong>of</strong> �-Al2O3 and �-Cr2O3. The materials<br />

science approach behind <strong>the</strong> physical vapour<br />

deposition process is to quench <strong>the</strong> vapour phase in<br />

<strong>the</strong> solid solution at deposition temperatures<br />

significantly below 1300 °C. A metastable corundum<br />

type solid solution <strong>of</strong> composition (Al1-x,Crx)2O3 is<br />

obtained at non-equilibrium conditions. Witthaut et<br />

al. [4] deposited (Al1-x,Crx)2O3 coatings in corundum<br />

structure by reactive magnetron sputtering (where<br />

aluminum was sputtered in r.f. mode and chromium<br />

in d.c. mode simultaneously). We introduce a new<br />

combinatorial approach for <strong>the</strong> deposition <strong>of</strong> Al-Cr-O<br />

thin films in order to elaborate <strong>the</strong> conditions <strong>of</strong> (Al1x,Crx)2O3<br />

thin film growth in corundum structure in<br />

more detail.<br />

II. EXPERIMENTAL DETAILS<br />

The coatings were deposited in a Leybold Z 550<br />

PVD machine. The experimental setup for <strong>the</strong><br />

deposition experiments was shown in Figure 1. We<br />

used a combinatorial approach with a segmented<br />

target consisting <strong>of</strong> aluminum and chromium plates.<br />

The diameter <strong>of</strong> <strong>the</strong> target was 75 mm, <strong>the</strong> target to<br />

substrate distance was 50 mm, and <strong>the</strong> cathode power<br />

was 500 W in r.f. mode. The total gas pressure was<br />

kept constant at 0.65 Pa in all experiments. The<br />

atmosphere consisted <strong>of</strong> 80 at.-% argon and 20 at.-%<br />

oxygen. During <strong>the</strong> reactive deposition process we<br />

controlled <strong>the</strong> substrate temperature in <strong>the</strong> range from<br />

180 °C to 600 °C with a heating plate (resistance<br />

wire) and a connected d.c. power supply. In addition<br />

we were able to induce a substrate bias up to -400 V<br />

with a second r.f. power supply. The coatings<br />

presented here were deposited at -100 V substrate<br />

bias at a substrate temperature <strong>of</strong> 500 °C.<br />

Commercial cemented carbide substrates and silicon<br />

wafers were coated. Before deposition <strong>the</strong> substrates<br />

were cleaned in an ultrasonic bath in acetone for 15<br />

minutes and plasma-etched in pure argon r.f.<br />

powered plasma for ten minutes. Five cemented


carbide substrates (12.7 mm x 12.7 mm in lateral<br />

dimension) were coated in each deposition process.<br />

These samples had a different chemical composition<br />

and thus a different microstructure, constitution and<br />

properties.<br />

N S<br />

N<br />

Al Cr<br />

Ar/O2 plasma<br />

1 2 3 4 5<br />

substrate holder, heatable<br />

Figure 1: Experimental setup<br />

III. CHARACTERISATION<br />

The coatings were characterized by determining <strong>the</strong>ir<br />

thickness by using a standard calo test, <strong>the</strong>ir Vickers<br />

micro hardness, <strong>the</strong>ir chemical composition by<br />

Electron Probe Microanalysis (EPMA) with a<br />

Camebax Microbeam equipment and <strong>the</strong>ir<br />

microstructure by X-Ray Diffraction (XRD) with a<br />

Seifert C-3000 diffractometer running in <strong>the</strong> Bragg-<br />

Bretano geometry using CuK� radiation.<br />

IV. GROWTH AND CONSTITUTION OF THE<br />

ALXCRYOZ THIN FILMS<br />

Between <strong>the</strong> AlxCryOz layer and <strong>the</strong> substrate a<br />

metallic adhesion interface layer <strong>of</strong> 300 nm Cr and 1<br />

μm Al/Cr, with a mixing ration Al/Cr according to<br />

<strong>the</strong> sample position under <strong>the</strong> target, was applied.<br />

The AlxCryOz layer had a minimum thickness <strong>of</strong><br />

about 3 μm.<br />

The highest growth rate was 1.13 μm/h at sample<br />

position three and <strong>the</strong> lowest growth rate was 0.81<br />

μm/h at sample position one.<br />

The lowest Vickers micro hardness was 2150<br />

HV0.05 at sample position five and <strong>the</strong> highest<br />

Vickers micro hardness was 2620 HV0.05 at sample<br />

position two. Ramm et al. [5] reported values<br />

between 1800 and 2500 HVpl for arc evaporated Al-<br />

Cr-O layers.<br />

The chemical analysis by EPMA revealed an<br />

argon content lower than 0.01 at.-%. The oxygen<br />

content <strong>of</strong> <strong>the</strong> coatings varied between 61.14 at.-% at<br />

sample position one and 59.61 at.-% at sample<br />

position five. The aluminum + chromium-content<br />

varied slightly over <strong>the</strong> sample positions, 38.60 at.-%<br />

at sample position one and 39.92 at.-% at sample<br />

position five. Thus we had thin films with nearly<br />

136<br />

stoichiometric (Al1-x,Crx)2O3 composition. The<br />

maximum Al/(Al+Cr) ratio was 0.71 at sample<br />

position one and <strong>the</strong> maximum Cr/(Al+Cr) ratio was<br />

0.75 at sample position five.<br />

We used <strong>the</strong> X-Ray Diffraction in Bragg-Bretano<br />

geometry for <strong>the</strong> phase analysis. The cemented<br />

carbide substrates consisted <strong>of</strong> tungsten carbide and<br />

cobalt. In <strong>the</strong> XRD spectrum we clearly identified<br />

tungsten carbide peaks (PDF No. 25-1047 and PDF<br />

No. 5-728 from <strong>the</strong> ICDD-data base). A comparison<br />

<strong>of</strong> <strong>the</strong> measured XRD spectrum with <strong>the</strong> peaks <strong>of</strong><br />

corundum (PDF No. 42-1468 from <strong>the</strong> ICDD-data<br />

base) and escolaite (PDF No. 38-1479 from <strong>the</strong><br />

ICDD-data base) showed six Bragg peaks positioned<br />

between <strong>the</strong> �-Al2O3 and �-Cr2O3 peaks <strong>of</strong> <strong>the</strong> same<br />

crystal lattice plane. Thus we assumed corundumtype<br />

growth <strong>of</strong> <strong>the</strong> solid solution.<br />

V. CONCLUSIONS<br />

Thin films <strong>of</strong> corundum-type solid solutions (Al1x,Crx)2O3<br />

were deposited at non-equilibrium<br />

conditions at a deposition temperature <strong>of</strong> 500 °C with<br />

a new combinatorial approach by reactive magnetron<br />

sputtering in an atmosphere containing 20 at.-%<br />

oxygen. The metastable solid solutions were<br />

nanocrystalline with a high degree <strong>of</strong> crystallinity.<br />

We investigated <strong>the</strong> growth rate, chemical<br />

composition and <strong>the</strong> phases <strong>of</strong> <strong>the</strong> coatings.<br />

REFERENCES<br />

[1] H. Holleck. Basic principles <strong>of</strong> specific<br />

applications <strong>of</strong> ceramic materials as protective<br />

layers. Surface and Coatings Technology, 43/44:<br />

245–258, December 1990.<br />

[2] T. M. Besmann and N. S. Kulkarni.<br />

Thermochemical analysis and modeling <strong>of</strong> <strong>the</strong><br />

Al2O3-Cr2O3, Cr2O3-SiO2, and Al2O3-Cr2O3-SiO2<br />

systems relevant to refractories . Journal <strong>of</strong> <strong>the</strong><br />

American Ceramic Society, 89(2):638–644,<br />

February 2006.<br />

[3] W. Sitte. Investigation <strong>of</strong> <strong>the</strong> miscibility gap <strong>of</strong><br />

<strong>the</strong> system chromia-alumina below 1300 °C.<br />

Materials Science Monographs, 28A:451-456,<br />

1985.<br />

[4] M. Witthaut, R. Cremer, K. Reichert and D.<br />

Neuschütz. Preparation <strong>of</strong> Cr2O3-Al2O3 solid<br />

solutions by reactive magnetron sputtering.<br />

Mikrochimica Acta, 133:191-196, 2000.<br />

[5] J. Ramm, M. Ante, T. Bachmann, B. Widrig, H.<br />

Brändle and M. Döbeli. Pulse enhanced electron<br />

emission (P3e) arc evaporation and <strong>the</strong><br />

syn<strong>the</strong>sis <strong>of</strong> wear resistant Al-Cr-O coatings in<br />

corundum structure. Surface and Coatings<br />

Technology, 202:876-883, 2007.


Modification <strong>of</strong> Aluminum Alkoxides with �-Keto Esters and<br />

Dialkylmalonates as Precursors for Sol-Gel Derived Hybrid<br />

Materials<br />

Robert Lichtenberger, Stefan O. Baumann, Maria Bendova, Michael Puchberger,<br />

and Ulrich Schubert<br />

Institute <strong>of</strong> Materials Chemistry<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: rlichten@mail.zserv.tuwien.ac.at<br />

Abstract — Aluminum alkoxides were modified<br />

with a series <strong>of</strong> �-ketoesters and dialkylmalonates<br />

to give complexes <strong>of</strong> <strong>the</strong> general formulas<br />

[Al(OR)2(�-diketonate)]2 and Al(�-diketonate)3. The<br />

use <strong>of</strong> ligands bearing additional functional groups<br />

allows syn<strong>the</strong>sis <strong>of</strong> precursors for inorganicorganic<br />

hybrid materials were syn<strong>the</strong>sized. The<br />

influence <strong>of</strong> varying <strong>the</strong> ester-alkoxo groups on<br />

structure and reactivity was studied using NMR, IR<br />

and XRD techniques. Transesterification as a possible<br />

side reaction affecting <strong>the</strong> structure <strong>of</strong> <strong>the</strong><br />

products was studied and found to be crucial for <strong>the</strong><br />

reaction with sterically less demanding malonates.<br />

I. INTRODUCTION<br />

Substitution <strong>of</strong> OR groups in metal alkoxides by<br />

organic ligands is a powerful tool for tuning <strong>the</strong><br />

structural and chemical behavior <strong>of</strong> <strong>the</strong> alkoxides,<br />

e.g. <strong>the</strong> reactivity during hydrolysis in sol-gel processing<br />

[1]. The use <strong>of</strong> ligands bearing additional<br />

functional groups allows introducing organic functionalities<br />

into precursors for metal oxide based solgel<br />

materials. This opens new possibilities for <strong>the</strong><br />

syn<strong>the</strong>sis <strong>of</strong> sol-gel derived inorganic-organic hybrid<br />

materials, where <strong>the</strong> inorganic and organic parts <strong>of</strong><br />

<strong>the</strong> material are directly linked. Understanding <strong>the</strong><br />

chemistry and structures <strong>of</strong> those precursors is a<br />

prerequisite for <strong>the</strong> deliberate design and use in <strong>the</strong><br />

preparation <strong>of</strong> novel metal oxide based materials.<br />

II. RESULTS<br />

The modification <strong>of</strong> aluminum alkoxides with a<br />

series <strong>of</strong> �-ketoesters and dialkylmalonates and <strong>the</strong><br />

influence <strong>of</strong> variation <strong>of</strong> <strong>the</strong> ester-alkoxo groups was<br />

studied. Fur<strong>the</strong>rmore <strong>the</strong> structure, stability, and<br />

reactivity <strong>of</strong> <strong>the</strong> complexes was investigated, in particular<br />

<strong>the</strong>ir tendency to undergo transesterification.<br />

137<br />

For <strong>the</strong> application <strong>of</strong> �-ketoesters and dialkylmalonates<br />

slightly different reaction conditions had<br />

to be applied.<br />

A. MODIFICATION WITH �-KETO ESTERS<br />

In contrast to procedures from <strong>the</strong> literature [2] for<br />

<strong>the</strong> modification <strong>of</strong> aluminum alkoxides with �diketones,<br />

<strong>the</strong> reaction <strong>of</strong> [Al(O i Pr)3]4 with one<br />

equivalent <strong>of</strong> �-ketoester at room temperature yields<br />

exclusively Al(�-ketoesterate)3 and unreacted alkoxide<br />

for all ligands used (Figure 1).<br />

O<br />

meac-H<br />

O O<br />

R =<br />

O O O O<br />

O<br />

etac-H<br />

i prac-H<br />

R<br />

t buac-H aaa-H<br />

Figure 1: �-Keto esters used<br />

O<br />

O<br />

meaa-H<br />

To obtain <strong>the</strong> desired [Al(O i Pr)2(�-ketoesterate)]2<br />

complexes, three syn<strong>the</strong>tic pathways were developed.<br />

1. In situ heating <strong>of</strong> <strong>the</strong> reaction solution containing<br />

Al(�-ketoesterate)3 and [Al(O i Pr)3]4<br />

2. Reaction <strong>of</strong> isolated Al(�-ketoesterate)3 with<br />

[Al(O i Pr)3]4 at elevated temperatures<br />

3. Pre-deoligomerization <strong>of</strong> [Al(O i Pr)3]4 at elevated<br />

temperatures and subsequent addition <strong>of</strong><br />

�-ketoester at room temperature<br />

The third reaction procedure opens <strong>the</strong> possibility<br />

to use <strong>the</strong>rmally sensitive ligands, e.g. 2-<br />

(methacryloyloxy)ethyl acetoacetate (meaa-H), for<br />

<strong>the</strong> syn<strong>the</strong>sis <strong>of</strong> [Al(O i Pr)2(meaa)]2 as a potential<br />

precursor for inorganic-organic hybrid materials.


The crystal structure <strong>of</strong> [Al(O i Pr)2( t buac)]2 (Figure<br />

2) shows <strong>the</strong> dimeric nature <strong>of</strong> <strong>the</strong> compound with<br />

one octahedrally and one tetrahedrally coordinated<br />

Al center, bridged by two isopropoxo groups.<br />

C18<br />

C19<br />

C17<br />

C21<br />

O7<br />

C15<br />

C20<br />

C23<br />

O8<br />

C16<br />

C22<br />

C24<br />

O6<br />

O10<br />

C13<br />

C11<br />

C14 C1<br />

Al2<br />

O9<br />

O5<br />

C26<br />

C10<br />

C25<br />

C28<br />

O3<br />

Figure 2: Thermal ellipsoid (50% probability) plot<br />

<strong>of</strong> [Al(O i Pr)2( t buac)]2<br />

Transesterification as side reacting was only found<br />

to be relevant in <strong>the</strong> reaction <strong>of</strong> Al(d t bum)3 with<br />

liberated isopropanol at elevated temperatures.<br />

B. MODIFICATION WIH DIALKYLMALONATES<br />

Al(O i Pr)3 Al(O t Bu)3 and Al(OEt)3 were analogously<br />

modified with a series <strong>of</strong> dialkylmalonates (Figure<br />

3).<br />

CH 3<br />

dmem-H<br />

O4<br />

O O<br />

C7<br />

C8<br />

C27<br />

C2<br />

C12<br />

C9<br />

R R<br />

O O<br />

R =<br />

Al1<br />

O1<br />

O2<br />

detm-H d i prm-H d t bum-H<br />

Figure 3: Dialkylmalonates used<br />

The modification <strong>of</strong> [Al(O i Pr)3]4 with dialkylmalonates<br />

was possible without additional reagents<br />

for <strong>the</strong> deprotonation <strong>of</strong> malonates, but in contrast to<br />

<strong>the</strong> modification with �-ketoesters, elevated temperatures<br />

and extended reaction times were needed.<br />

NMR studies revealed that <strong>the</strong> complexes<br />

[Al(O i Pr)2(d i prm)]2 and [Al(O i Pr)2(d t bum)]2 are<br />

isostructural to [Al(O i Pr)2( t buac)]2 (Figure 2).<br />

For <strong>the</strong> reaction <strong>of</strong> Al(O i Pr)3 with dialkylmalonates<br />

bearing smaller alkoxo groups, viz. dmem-H<br />

and detm-H, transesterification was observed to a<br />

considerable extent. Therefore, for <strong>the</strong> preparation <strong>of</strong><br />

Al(dmem)3 (Figure 4) Al(O t Bu)3 was used as a starting<br />

compound, giving no transesterification.<br />

C3<br />

C4<br />

C5<br />

C6<br />

138<br />

O3<br />

C2<br />

C4<br />

C1<br />

O4<br />

C3<br />

O1<br />

C8<br />

C5<br />

O2<br />

O5<br />

Al1<br />

Figure 4: Thermal ellipsoid (50% probability) plot<br />

<strong>of</strong> Al(dmem)3<br />

III. CONCLUSION<br />

Aluminum alkoxides were successfully modified<br />

with a series <strong>of</strong> �-keto esters and dialkylmalonates.<br />

For both types <strong>of</strong> ligands, dimeric complexes with<br />

two ligands and trisubstituted monomeric compounds<br />

were prepared and characterized. The influence <strong>of</strong><br />

<strong>the</strong> ester-alkoxo groups was studied and found to<br />

have no influence on <strong>the</strong> structure but to be crucial<br />

for <strong>the</strong> tendency to undergo transesterification.<br />

For <strong>the</strong> modification with �-keto esters additionally<br />

ligands bearing polymerizable groups were successfully<br />

introduced to give products applicable as<br />

precursors for sol-gel derived inorganic-organic<br />

hybrid materials.<br />

Transesterification was found to be a relevant side<br />

reaction in particular for <strong>the</strong> modification <strong>of</strong><br />

Al(O i Pr)3 with sterically less demanding dialkylmalonates.<br />

ACKNOWLEDGMENTS<br />

The author thanks R. Hellein and M. Schwabl for<br />

<strong>the</strong>ir practical work in <strong>the</strong> field <strong>of</strong> <strong>the</strong> dialkylmalonate<br />

modification.<br />

REFERENCES<br />

[1] U. Schubert. Sol-Gel Processing <strong>of</strong> Metal Compounds.<br />

In Comprehensive Coordination Chemistry<br />

II, pages 629-656. Elsevier, Amsterdam,<br />

Ne<strong>the</strong>rlands, 2004<br />

[2] J. H. Wengrovius, M. F. Garbauskas, E. A. Williams,<br />

R. C. Going, P. E. Donahue, and J. F.<br />

Smith. Aluminum Alkoxide Chemistry Revisited:<br />

Syn<strong>the</strong>sis, Structure, and Modification <strong>of</strong><br />

Several Aluminum Alkoxide and Siloxide Complexes.<br />

J. Am. Chem. Soc., 108(5):982–989,<br />

1986.<br />

O6<br />

C6<br />

C7


Structural Aspects <strong>of</strong> Oximate-Modified Titanium Alkoxides<br />

Stefan O. Baumann, Maria Bendova, Robert Potzmann,<br />

Michael Puchberger, and Ulrich Schubert<br />

Institute <strong>of</strong> Materials Chemistry, Vienna University <strong>of</strong> Technology,<br />

Getreidemarkt 9, 1060 Vienna, Austria<br />

Email: sbaumann@mail.zserv.tuwien.ac.at<br />

Abstract — Titanium alkoxides were reacted with<br />

oximes in order to give modified precursors for solgel<br />

processing. The substituent’s influence on <strong>the</strong><br />

obtained derivates was monitored in case <strong>of</strong> anisaldehyde<br />

oxime as crystal structures and calculated<br />

models were compared. The reaction <strong>of</strong> acetone<br />

oxime with titanium isopropoxide resulted in a<br />

dimeric complex and - as partial hydrolysis occurred<br />

– in a hexanuclear oxo cluster.<br />

I. INTRODUCTION<br />

Organically modified titanium alkoxides serve as<br />

precursors for sol-gel derived inorganic-organic<br />

hybrid materials. Substitution <strong>of</strong> alkoxo groups<br />

with organic ligands on <strong>the</strong> one hand lowers reactivity<br />

and thus facilitates control during sol-gel<br />

processing, and affects <strong>the</strong> material’s structure. On<br />

<strong>the</strong> o<strong>the</strong>r hand, modification <strong>of</strong> those molecular<br />

precursors also permits <strong>the</strong> introduction <strong>of</strong> new<br />

functional groups that influence <strong>the</strong> chemical<br />

properties and provide reactive sites, <strong>of</strong>fering <strong>the</strong><br />

possibility <strong>of</strong> post-syn<strong>the</strong>sis modification or allowing<br />

later crosslinking. Understanding <strong>the</strong> influence<br />

<strong>of</strong> <strong>the</strong> introduced organic groups on both chemistry<br />

and structure <strong>of</strong> <strong>the</strong> modified alkoxide precursors<br />

establishes is a key requirement for <strong>the</strong> development<br />

<strong>of</strong> new titania based hybrid materials.<br />

II. RESULTS<br />

A lot <strong>of</strong> work has been done on <strong>the</strong> modification<br />

<strong>of</strong> titanium alkoxides with oxygen containing coordinating<br />

groups, viz. carboxylates and �-diketonates. [1]<br />

Nitrogen bearing ligands, e.g. oximes, have been less<br />

investigated, [2-4] although oximate ligands show<br />

interesting coordination behaviour and allow new<br />

ways in transition metal alkoxide modification.<br />

A. BINDING MODES<br />

The effect <strong>of</strong> altered substitution modes in o-, m-<br />

and p-anisaldehyde oxime on <strong>the</strong> reaction products<br />

with titanium isopropoxide was investigated.<br />

Binding <strong>of</strong> different isomers to <strong>the</strong> metal atom in<br />

<strong>the</strong> crystal structures (Figure 1) was compared<br />

139<br />

with <strong>the</strong> results <strong>of</strong> quantum mechanical calculations.<br />

Figure 1: Crystal structures <strong>of</strong> <strong>the</strong> reaction products<br />

<strong>of</strong> titanium isopropoxide with o-, m- and panisaldehyde<br />

oxime<br />

B. PARTIAL HYDROLYSIS PRODUCTS<br />

The reaction <strong>of</strong> titanium isopropoxide with acetone<br />

oxime yielded crystals <strong>of</strong> <strong>the</strong> dimeric complex with a<br />

ligand to metal ratio <strong>of</strong> 2/1. When small amounts <strong>of</strong><br />

water were introduced to <strong>the</strong> system, a hexanuclear<br />

oxo cluster crystallized from <strong>the</strong> reaction mixture.<br />

The structures derived by single crystal X-ray diffraction<br />

(Figure 2) were correlated to NMR studies in<br />

solution.<br />

Figure 2: Crystal structures <strong>of</strong> <strong>the</strong> reaction products<br />

<strong>of</strong> titanium isopropoxide with acetone oxime


III. CONCLUSION<br />

With regard to <strong>the</strong> syn<strong>the</strong>sis <strong>of</strong> new hybrid materials<br />

via <strong>the</strong> sol-gel pathway <strong>the</strong> investigation <strong>of</strong> precursors<br />

derived by modification <strong>of</strong> metal alkoxides with<br />

organic ligands is a crucial point. The structural investigation<br />

<strong>of</strong> differently substituted anisaldehyde<br />

oximate derivates is as well as hydrolysis effects <strong>of</strong><br />

titanium alkoxides modified with acetone oximate<br />

<strong>of</strong>feris new snapshots <strong>of</strong> <strong>the</strong> route from <strong>the</strong> molecular<br />

precursor to <strong>the</strong> final material.<br />

ACKNOWLEDGMENTS<br />

Financial support by <strong>the</strong> Austrian Fonds zur Förderung<br />

der wissenschaftlichen Forschung (FWF) is<br />

gratefully acknowledged.<br />

REFERENCES<br />

[1] U. Schubert. Chemical modification <strong>of</strong> titanium<br />

alkoxides for sol-gel processing. Journal <strong>of</strong> Materials<br />

Chemistry, 15: 3701-3715, 2005.<br />

[2] R. C. Mehrotra, A. K. Rai, A. K. Singh and R.<br />

Bohra. Oxime and hydroxylamine derivatives <strong>of</strong><br />

metals and metaloids. Inorganica Chimica Acta<br />

13(1): 91-103, 1975.<br />

[3] U. Thewalt and R. Friedrich. Cp2Ti IV Complexes<br />

with oximato ligands. Z. Naturforsch., 46b: 475-<br />

482, 1991.<br />

[4] M. G. Davidson, A. L. Johnson, M. D. Jones, M.<br />

D. Lunn, M. F. Mahon. Titanium(IV) complexes<br />

<strong>of</strong> oximes – Novel binding modes. Polyhedron<br />

26(5): 975-980, 2007.<br />

140


Trace Metal Speciation in Water using a Solid Phase Extraction<br />

Flow Injection Setup online coupled to ICP-AES<br />

Christoph Puls and Andreas Limbeck (Faculty Mentor)<br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

Vienna University <strong>of</strong> Technologies<br />

Vienna, Austria<br />

Email: e9625356@stud3.tuwien.ac.at<br />

Abstract — The presented speciation method employs<br />

a Flow-Injection (FI) system consisting <strong>of</strong><br />

three different Solid-Phase-Extraction (SPE) cartridges<br />

that retain passing metal species, separating<br />

<strong>the</strong>m into <strong>the</strong> fractions neutral, anionic and cationic.<br />

During direct elution into <strong>the</strong> detecting ICP-<br />

AES system, transient signals are recorded simultaneously<br />

for Mg, Al, Ca, Cr, Mn, Fe, Co, Cu, Zn, Cd<br />

and Pb. Thereby, both <strong>the</strong> total metal concentrations<br />

and <strong>the</strong> distribution into <strong>the</strong> three classes are<br />

acquired in one run. By automatization <strong>the</strong> presented<br />

method minimizes potential error sources.<br />

Due to <strong>the</strong> occurring preconcentration, enhanced<br />

detection limits could be achieved.<br />

The accuracy <strong>of</strong> <strong>the</strong> method was verified using<br />

standard addition series. Finally <strong>the</strong> applicability <strong>of</strong><br />

<strong>the</strong> technique to molten snow samples was demonstrated.<br />

I. INTRODUCTION<br />

In environmental and especially aquatic systems, <strong>the</strong><br />

determination <strong>of</strong> trace metals and assessment <strong>of</strong> <strong>the</strong>ir<br />

toxic potential cannot be reduced to <strong>the</strong> measurement<br />

<strong>of</strong> <strong>the</strong>ir total concentration, since trace metals tend to<br />

be complexed by varying ligands, changing <strong>the</strong>ir<br />

chemical form and thus affecting <strong>the</strong>ir chemical<br />

reactivity and biological availability [1],[2] . Due to <strong>the</strong><br />

correlation <strong>of</strong> trace metal toxicity and biological<br />

availability to fractions retained on different sorbent<br />

materials [3],[4] , solid phase extraction (SPE) techniques<br />

have been widely employed.<br />

The use <strong>of</strong> flow injection (FI) techniques for all<br />

kinds <strong>of</strong> automatization and <strong>the</strong>ir adaption to enhance<br />

sensitivity and precision <strong>of</strong> ultra trace elemental<br />

analysis has been widely demonstrated [5]-[7] . Both<br />

chemical speciation analysis by FI-systems coupled<br />

on-line to varying detector systems has been comprehensively<br />

studied [8] and a few systems able to<br />

discriminate between organic and inorganic complexes<br />

as well as <strong>the</strong> free ionic form have been described<br />

[9] , but <strong>the</strong> incorporation <strong>of</strong> an SPE-system<br />

able to achieve this fractionation into an FI-system<br />

coupled on-line to a multi-elemental detector remains<br />

to be achieved.<br />

141<br />

II. EXPERIMENTAL<br />

A. REAGENTS AND MATERIALS<br />

High purity water was obtained by double distillation<br />

<strong>of</strong> deionized water and used throughout. All used<br />

chemicals were <strong>of</strong> p.a. grade purity or higher. Due to<br />

<strong>the</strong> need to reuse <strong>the</strong> SPE-cartridges and -sorbents<br />

several times styrene-divinylbenezene (S-DVB)<br />

based materials were preferred. The employed sorbent<br />

materials were Strata-X, Strata-X-AW and<br />

Strata-X-C (all from Phenomenex). All sorbents were<br />

delivered in cartridges containing 30 mg and space<br />

for 1 ml <strong>of</strong> sample solution. They were shortened and<br />

fitted with a male luer tip for incorporation into <strong>the</strong><br />

FI-system.<br />

The FI-Setup included three six-port two position<br />

valves. All parts were connected with PTFE and<br />

TYGON pump tubings. The eluent and <strong>the</strong> sample<br />

flow were kept up through two channels <strong>of</strong> a<br />

SPETEC PERIMAX 12 peristaltic pump. A Spectr<strong>of</strong>lame<br />

P ICP-AES from SPECTRO A.I. was used<br />

for detection.<br />

B. FI-PROCEDURE<br />

The FI-system for fractionation <strong>of</strong> metal species into<br />

hydrophobic organic complexes, anionic complexes<br />

and cations consisted <strong>of</strong> three six-port valves and<br />

three SPE cartridges.<br />

After conditioning all cartridges to neutralize <strong>the</strong><br />

pH <strong>of</strong> <strong>the</strong> SPE sorbents with (NH4)2CO3 buffer solution,<br />

<strong>the</strong> sample was pumped through <strong>the</strong> cartridges,<br />

thus concentrating <strong>the</strong> analytes on <strong>the</strong> respective SPE<br />

sorbents. After washing all sorbent cartridges with<br />

buffer solution, elution <strong>of</strong> <strong>the</strong> cartridges was carried<br />

out consecutively in reversed order with 2 M HNO3.<br />

The elution sequence was automatized and controlled<br />

by a serial communication program (VCom).<br />

In course <strong>of</strong> <strong>the</strong> measuring sequence 110 data points<br />

were collected for each element. This transient signals<br />

each comprised three peaks, with peak areas<br />

corresponding to <strong>the</strong> respective trace metal fraction.


III. RESULTS<br />

A. DETECTION LIMITS AND REPRODUCIBILITY<br />

Statistical characterization <strong>of</strong> <strong>the</strong> total metal quantification<br />

was performed employing standard solutions<br />

spiked with different ligands to account for differing<br />

trace metal species.<br />

For determination <strong>of</strong> <strong>the</strong> reproducibility standard<br />

solutions were analyzed in <strong>the</strong> course <strong>of</strong> two measurement<br />

series occurring on different days. Within<br />

each series three standards were analyzed at intervals<br />

<strong>of</strong> 2 hours resulting in a total <strong>of</strong> six measurements.<br />

Detection limits calculated via <strong>the</strong> 3σ-criterion and<br />

relative standard deviations from reproducibility<br />

measurements are given in Table 1.<br />

Element LOD (3σ) RSD (n = 6)<br />

Mg 1.2 μg/l 4.8 %<br />

Al 3.3 μg/l 3.1 %<br />

Ca 3.5 μg/l 3.7 %<br />

Cr 0.7 μg/l 3.5 %<br />

Mn 0.2 μg/l 4.6 %<br />

Fe 1.2 μg/l 4.5 %<br />

Co 1.3 μg/l 2.9 %<br />

Cu 1.3 μg/l 4.7 %<br />

Zn 1.4 μg/l 5.9 %<br />

Cd 1.1 μg/l 3.8 %<br />

Pb 12 μg/l 6.4 %<br />

Table 1: Detection limits and relative<br />

standard deviations.<br />

B. SPECIES SEPARATION<br />

To investigate <strong>the</strong> separation <strong>of</strong> trace metals into <strong>the</strong><br />

desired fractions, standard solutions were spiked with<br />

0.1 mM <strong>of</strong> different organic ligands at varying pH<br />

values. The investigated ligands were<br />

8-hydroxyquinoline, tryptophan, ethylene-diaminetetraacetate<br />

and nitrilotriacetic acid.<br />

Measurements <strong>of</strong> <strong>the</strong>se syn<strong>the</strong>tic samples confirmed<br />

<strong>the</strong> ability <strong>of</strong> <strong>the</strong> system to fractionate trace<br />

metal complexes according to <strong>the</strong>ir affinity to <strong>the</strong><br />

employed SPE sorbents.<br />

C. SNOW SAMPLES<br />

Snow Samples were collected at different urban<br />

locations, stored frozen and molten prior to analysis.<br />

Analysis was carried out using <strong>the</strong> previously described<br />

setup, to ensure <strong>the</strong> accuracy <strong>of</strong> trace metal<br />

quantification, standard addition measurements were<br />

performed.<br />

Total metal concentrations were within <strong>the</strong> range<br />

reported previously for urban snow samples. Additionally,<br />

trace metals could be assigned to fractions<br />

according to <strong>the</strong>ir natural speciation.<br />

142<br />

IV. CONCLUSION<br />

The presented FI-system is able to separate neutral<br />

hydrophobic and anionic complexes from <strong>the</strong> cationic<br />

constituents by retaining and simultaneously<br />

preconcentrating each <strong>of</strong> <strong>the</strong> three fractions separately<br />

on <strong>the</strong> corresponding SPE-sorbent.<br />

Experiments involving syn<strong>the</strong>tic ligand-spiked<br />

standard solutions confirmed <strong>the</strong> achieved separation<br />

efficiency. Due to <strong>the</strong> inherent preconcentration<br />

abilities <strong>of</strong> <strong>the</strong> presented FI-system <strong>the</strong> achievable<br />

limits <strong>of</strong> detection for quantification by ICP-AES<br />

were improved. On-line coupling <strong>of</strong> <strong>the</strong> FI-system to<br />

<strong>the</strong> nebulizer <strong>of</strong> <strong>the</strong> ICP-AES and automatized control<br />

<strong>of</strong> <strong>the</strong> employed valves minimized contamination<br />

risks and led to good reproducibility <strong>of</strong> determined<br />

results.<br />

To demonstrate <strong>the</strong> applicability <strong>of</strong> <strong>the</strong> developed<br />

method to snow samples, samples collected at different<br />

sites were analyzed. Observed total metal concentrations<br />

were within <strong>the</strong> range reported previously for<br />

snow samples collected at urban sites.<br />

REFERENCES<br />

[1] Hart, B. T.; Hines T., In Trace Elements In Natural<br />

Waters; Salbu, B.; Steinnes, E., Eds.; CRC<br />

Press: Boca Raton, FL, 1995; pp 203-221.<br />

[2] Campbell, P. G. C., In Metal Speciation and<br />

Bioavailability in Aquatic Systems; Tessier, A.;<br />

Turner, D. R., Eds.; John Wiley & Sons: New<br />

York, 1995 pp 45-102.<br />

[3] Chakrabarti, C. L. et al., Anal. Chim. Acta 267<br />

(1993), 47-64.<br />

[4] Burba, P.; Rocha, J.; Klockow, D., Fresenius J.<br />

Anal. Chem. 349 (1994), 800-807.<br />

[5] Sperling, M.; Yin, X.; Welz, B., J. Anal. At.<br />

Spectrom. 6 (1991), 295 – 300.<br />

[6] Liu, Z.-S.; Huang, S.-D., Anal. Chim. Acta 281<br />

(1993), 185-90.<br />

[7] Benkhedda, K. et al, J. Anal. At. Spectrom. 15<br />

(2000), 429-434.<br />

[8] Das, A. K.; de la Guardia, M.; Cervera, M. L.,<br />

Talanta 55 (2001), 1-28.<br />

[9] Groschner, M.; Appriou, P., Anal. Chim. Acta<br />

297 (1994), 369-376.


Analysis <strong>of</strong> Snowpack Properties<br />

And Estimation <strong>of</strong> Snow Grain Size<br />

Using Cold Land Processes Field Experiment Data<br />

Dugwon Seo 1 , Amir E Azar 1 , Al Powell 1 and Reza Khanbilvardi 2<br />

NOAA-CREST<br />

The City College <strong>of</strong> New York<br />

New York, U.S.A.<br />

Email: seo59ny@gmail.com<br />

Abstract — We focused on analysis <strong>of</strong> snow grain<br />

size behavior with respect to o<strong>the</strong>r snow parameters<br />

such as snow depth, density, and temperature which<br />

are collected by NASA Cold-Land Process Experiment<br />

data. We derive a pattern which can be used<br />

to approximate <strong>the</strong> range <strong>of</strong> grain size variations.<br />

The analysis showed that <strong>the</strong> snow density, grain<br />

size and temperature usually increase towards <strong>the</strong><br />

snow-ground interface. Overall, <strong>the</strong> grain size<br />

variation is highly correlated with both density and<br />

temperature. The correlation is generally higher<br />

between grain size and temperature as compared<br />

with grain size and density. Thus, snowpack temperature<br />

pr<strong>of</strong>ile might be estimated by a linear<br />

function having <strong>the</strong> top and bottom temperature.<br />

Using snowpack temperature, <strong>the</strong> grain size evolution<br />

can be approximated.<br />

I. INTRODUCTION<br />

Snow albedo plays a major role in earth's energy<br />

balance. Seasonal snow covers more than thirty percent<br />

<strong>of</strong> earth’s land surface [1]. Also, estimation <strong>of</strong><br />

snowpack properties is significantly important in<br />

regional scale for various problems such as flood<br />

predictions and water resource management. Melting<br />

snow is responsible for majority <strong>of</strong> <strong>the</strong> spring floods.<br />

Microwaves are sensitive to snowpack properties<br />

(depth, water equivalent, grain size, and wetness).<br />

Many studies have been conducted to establish a<br />

relationship between snowpack properties and microwave<br />

electromagnetic scattering signatures [2].<br />

One <strong>of</strong> <strong>the</strong> approaches was simulation <strong>of</strong> <strong>the</strong> microwave<br />

emission <strong>of</strong> snowpack [3]. It proposed a single<br />

layer snow emission model (Helsinki University <strong>of</strong><br />

Technology, HUT) for simulating <strong>the</strong> snowpack<br />

brightness temperature. Similarly, Wiseman and<br />

Matzler developed a multi-layer microwave emission<br />

model (MEMLS) [4].<br />

In <strong>the</strong> mentioned algorithms and models, <strong>the</strong>re are<br />

key parameters <strong>of</strong> snow such as snow depth, density,<br />

water content, and grain size. These parameters play<br />

a significant role in microwave emission and con-<br />

143<br />

tribute to <strong>the</strong> measured emission by sensors. The goal<br />

<strong>of</strong> this study is to analyze <strong>the</strong> behavior <strong>of</strong> snow grain<br />

size variations with respect to o<strong>the</strong>r snowpack parameters,<br />

which influence <strong>the</strong> microwave scattering<br />

<strong>of</strong> snowpack. We analyzed <strong>the</strong> data from one <strong>of</strong> <strong>the</strong><br />

most comprehensive snow survey; NASA CLPX, in<br />

order to establish <strong>the</strong> potential relationship <strong>of</strong> snow<br />

pack properties with respect to each o<strong>the</strong>r.<br />

II. ANALYSIS<br />

A. STUDY AREAS<br />

The Cold Land Process Experiment (CLPX) consisted<br />

<strong>of</strong> multiple sites in Colorado. This study area<br />

(Colorado) is selected because <strong>the</strong> central Rocky<br />

Mountains have large physiographic gradients and<br />

<strong>the</strong>y provide different terrain, snow, soil, and ecological<br />

characteristics. Study area <strong>of</strong> CLPX data<br />

collection is designed in one area within ano<strong>the</strong>r,<br />

based on 5-level set. Different scale <strong>of</strong> level is designed<br />

to provide a “scale bridge” between <strong>the</strong> levels<br />

for better understanding at one scale to o<strong>the</strong>r scales.<br />

B. PROFILE OF SNOW PROPERTIES<br />

In order to have a better understanding <strong>of</strong> snowpack<br />

properties and <strong>the</strong>ir behavior with respect to each<br />

o<strong>the</strong>r, <strong>the</strong> pr<strong>of</strong>iles <strong>of</strong> snow depth, density, temperature,<br />

and grain size were analyzed.<br />

The pr<strong>of</strong>ile <strong>of</strong> snow temperature, density, and<br />

snow grain size for various snow pits in all ISAs<br />

located in Fraser and Rabbit Ear were drawn. Figure<br />

1 illustrates changes in snowpack characteristics<br />

(density, temperature and grain size). It is observed<br />

that, with except to very top layer <strong>of</strong> snow, <strong>the</strong><br />

snow temperature, density, and averaged grain size<br />

increase towards <strong>the</strong> snow-ground interface. The<br />

older snow has larger grain size and <strong>the</strong> fresh snow<br />

on top has smaller grain size. But, <strong>the</strong> very top<br />

layer <strong>of</strong> <strong>the</strong> snow (top 10cm) shows different<br />

characteristics. In cases, <strong>the</strong> snow grain size is<br />

larger in very top layer as compared with <strong>the</strong>


underlying layer. This can be explained by daily<br />

fluctuations air temperature.<br />

height(cm)<br />

ffsp12 snow temp<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

-8 -6 -4 -2<br />

snow temp(c)<br />

height(cm)<br />

ffsp12 grain sz<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

0 1 2 3<br />

snow grain size(mm)<br />

height(cm)<br />

ffsp12 density<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

0 100 200 300<br />

snow density(kg/m 3 )<br />

Figure 1:Example <strong>of</strong> variation <strong>of</strong> snow temperature,<br />

grain size, and density in <strong>the</strong> snow pr<strong>of</strong>ile<br />

On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> bottom layer has<br />

different characteristics. The average grain size,<br />

snow temperature, and density <strong>of</strong> <strong>the</strong> bottom layer<br />

are larger than o<strong>the</strong>r layers. In addition, <strong>the</strong> snow<br />

temperature in <strong>the</strong> bottom layer falls in a certain<br />

range with a relatively small deviation. The<br />

majority temperature <strong>of</strong> bottom layer is between<br />

0 o C and -2 o C in Rabbit Ear and Fraser.<br />

To quantify <strong>the</strong> variation <strong>of</strong> all snow<br />

parameters with respect to each o<strong>the</strong>r, <strong>the</strong><br />

correlation coefficients between those parameters<br />

over <strong>the</strong> pr<strong>of</strong>ile was calculated for all <strong>the</strong> snow<br />

pits within ISAs <strong>of</strong> Fraser and Rabbit Ear.<br />

C. METHODOLOGY<br />

As mentioned in <strong>the</strong> previous section, <strong>the</strong> snow<br />

temperature and snow grain size are highly correlated<br />

in <strong>the</strong> pr<strong>of</strong>ile <strong>of</strong> snowpack. Thus, using <strong>the</strong><br />

temperature, <strong>the</strong> grain size pr<strong>of</strong>ile can be estimated.<br />

Also, it was mentioned that <strong>the</strong> temperature for <strong>the</strong><br />

bottom layer <strong>of</strong> snow tends to be in a certain range<br />

(between 0 o C and -2 o C). Thus, <strong>the</strong> intercept <strong>of</strong> <strong>the</strong><br />

regression line can be a fixed value which is<br />

determined based on <strong>the</strong> characteristics <strong>of</strong> <strong>the</strong> snow<br />

covered area.<br />

Fraser<br />

ƒ(snow temperature) Averaged Linear Equations<br />

IOP1 (Feb, 2002) Z=0.25*T + 2.4<br />

IOP2 (Mar, 2002) Z = 0.22*T + 2<br />

IOP3 (Feb, 2003) Z = 0.14*T + 1.8<br />

IOP4 (Mar, 2003) Z = 0.3*T + 1.4<br />

Rabbit Ear<br />

ƒ(snow temperature) Averaged Linear Equations<br />

IOP1 (Feb, 2002) Z = 0.12T + 1.4<br />

IOP2 (Mar, 2002) Z = 0.13*T + 1.2<br />

IOP3 (Feb, 2003) Z = 0.07*T + 1<br />

IOP4 (Mar, 2003) Z = 0.07*T + 0.8<br />

Table 1: Equations to estimate snow grain size (Z)<br />

using snow temperature (T).<br />

144<br />

In this study, we derived <strong>the</strong> regression line<br />

between snow grain size and snow temperature for<br />

each <strong>of</strong> <strong>the</strong> snow pits located in all ISAs for<br />

selected periods. Then, we defined a new<br />

regression line by averaging <strong>the</strong> slope <strong>of</strong> <strong>the</strong><br />

regression lines for all <strong>the</strong> snow pits. Finally, we<br />

examined <strong>the</strong> derived regression line for<br />

estimating <strong>the</strong> snow grain size using snow<br />

temperature pr<strong>of</strong>ile for different periods in various<br />

locations (Table 1).<br />

III. CONCLUSION<br />

It was observed that, with except very top layer <strong>of</strong><br />

snow, <strong>the</strong> snow temperature, density, and averaged<br />

grain size increase towards <strong>the</strong> snow-ground<br />

interface. Using <strong>the</strong> results form <strong>of</strong> <strong>the</strong> analysis,<br />

we equation to estimate snow grain size using<br />

snow temperature. The spatial changes in<br />

snowpack properties in this area add to <strong>the</strong><br />

complexity <strong>of</strong> <strong>the</strong> problem. In addition, <strong>the</strong><br />

changes in <strong>the</strong> snow pr<strong>of</strong>ile could be due to<br />

existence <strong>of</strong> a very stratified layer in <strong>the</strong> bottom<br />

and a layer <strong>of</strong> recently fallen fresh snow on <strong>the</strong><br />

top. In short, <strong>the</strong> approach for estimation <strong>of</strong> snow<br />

grain size using <strong>the</strong> snow temperature was<br />

relatively successful and <strong>the</strong> source <strong>of</strong> <strong>the</strong> errors<br />

was identified.<br />

ACKNOWLEDGMENTS<br />

The authors would like to thank NOAA and <strong>the</strong> City<br />

University <strong>of</strong> New York for funding this research<br />

project.<br />

REFERENCES<br />

[1] Robinson, D., Dewey, K., and Heim, R., “Global<br />

Snow Cover Monitoring: An Update,” Bulltin <strong>of</strong><br />

Americamn Meteorological Society, vol.74, no.9, pp.<br />

1689-1696, 1993.<br />

[2] B.E. Goodison, and A.E. Walker, “Canadian<br />

development and use <strong>of</strong> snow cover information from<br />

passive microwave satellite data,” Passive microwave<br />

remote sensing <strong>of</strong> land- atmosphere interactions, B. J.<br />

Choudhury, Y. H. Kerr, E. G. Njoku and P. Pampaloni,<br />

VSP, pp. 245-262, 1995.<br />

[3] Pullianen, J. T., Grandell, J., and Hallikainen M,<br />

“HUT snow emission model and its applicability to<br />

snow water equivalent retrieval,” IEEE Transaction on<br />

Geosciences and Remote sensing, vol.37, no.3, pp.<br />

1378-1390, 1999.<br />

[4] Matzler, C., and Wisemann, A., “Microwave<br />

Emission Model <strong>of</strong> Layered Snowpack,” Remote<br />

Sensing <strong>of</strong> Environment, 70, pp. 307- 316, 1999.


Supersensitive tin oxide nanosensors for gas detection<br />

Alexandra Tischner a , Anton Köck (Faculty Mentor) a , Thomas Maier a , C. Edtmaier b , C.<br />

Gspan c , G. Kothleitner c<br />

a Nano-System-Technologies, Austrian Research Centers GbmH – ARC, 1220 Vienna, Austria<br />

b Institute <strong>of</strong> Chemical Technologies and Analytic, Vienna University <strong>of</strong> Technology, Austria<br />

c Graz Centre for Electron Microscopy, Austria<br />

Abstract — Supersensitive gas sensors based on<br />

nanocrystalline SnO2-films and single-crystalline<br />

SnO2-nanowires, both integrated on Si, are presented.<br />

The sensing mechanism is based on a<br />

change in <strong>the</strong> electrical conductance due to adsorption<br />

<strong>of</strong> gas molecules. The sensors are able to detect<br />

CO and CH4 gases down to concentrations <strong>of</strong> less<br />

than 5 ppm, which are both gases relevant for<br />

safety and environmental issues.<br />

I. INTRODUCTION<br />

Triggered by various applications like industrial<br />

process control, safety systems, environmental monitoring<br />

or disease diagnostics in medicine, <strong>the</strong>re is an<br />

increasing demand for high sensitive gas detecting<br />

devices. Over <strong>the</strong> last decades, especially metal oxides<br />

have become established gas sensing materials<br />

for a number <strong>of</strong> applications [1-4]. The most prominent<br />

candidate among all sensing materials is SnO2,<br />

and many SnO2 based sensing devices have been<br />

realized so far [5-7]. A most powerful strategy for <strong>the</strong><br />

improvement <strong>of</strong> sensor performance is <strong>the</strong> implementation<br />

<strong>of</strong> nanostructures as gas sensing components<br />

due to <strong>the</strong>ir high surface to volume ratio which<br />

allows a strong interaction between <strong>the</strong> ambient<br />

gases and <strong>the</strong> material. Several sensors based on<br />

nanocrystalline metal oxide films [8-10] or single<br />

crystalline nanowires [11, 12] have been realized so<br />

far.<br />

II. SENSORS BASED ON TIN OXIDE<br />

NANOCRYSTALLINE FILMS<br />

The thin film gas sensors are fabricated by a spray<br />

pyrolysis process on silicon substrates. A solution <strong>of</strong><br />

tin chloride pentahydrate and ethyl acetate is sprayed<br />

on Si-substrates with a 750 nm thick SiO2 layer on<br />

top, which are heated to a temperature <strong>of</strong> 500°C in<br />

ambient atmosphere. Produced film thickness is 20-<br />

100 nm. Subsequently <strong>the</strong> SnO2-films are mesastructured<br />

by means <strong>of</strong> photolithography and reactive<br />

Email: alexandra.tischner@arcs.ac.at<br />

anton.koeck@arcs.ac.at<br />

145<br />

CO flow [sccm]<br />

Fig. 1: Nanocrystalline sensor element with<br />

SnO2-bars on SiO2/Si-substrate.<br />

10<br />

9<br />

5,80E+03<br />

8<br />

7<br />

5,64E+03<br />

6<br />

5<br />

5,48E+03<br />

4<br />

3<br />

5,32E+03<br />

2<br />

1<br />

CO flow [sccm]<br />

Sensor R [Ω]<br />

5,16E+03<br />

0<br />

5,00E+03<br />

0 500 1000 1500 2000<br />

Time [s]<br />

Fig. 2: Nanocrystalline sensor response to 4ppm<br />

CO<br />

ion etching; sensors are accomplished by evaporation<br />

<strong>of</strong> Ti/Au-contacts (Fig. 1).<br />

Gas sensing performance is investigated in an automated<br />

measurement setup, which allows precise<br />

adjustment <strong>of</strong> <strong>the</strong> gaseous environment with syn<strong>the</strong>tic<br />

air as background gas. The following results<br />

have been found: The sensors, which are operated at<br />

temperatures <strong>of</strong> 250–400°C, are able to detect CO<br />

and CH4 down to a concentration <strong>of</strong> less than 5 ppm<br />

[13]. Fig. 2 shows <strong>the</strong> response <strong>of</strong> <strong>the</strong> nanocrystalline<br />

sensor to CO gas pulses with a concentration <strong>of</strong> only<br />

4 ppm at an operating temperature <strong>of</strong> 350°C.<br />

Sensor Resistance [Ω]


Despite <strong>the</strong> excellent sensitivity performance, stability<br />

and selectivity to specific gas species are a problem<br />

for <strong>the</strong> presented ultrathin SnO2-gassensor: as<br />

can be seen from Fig. 2 <strong>the</strong> sensor signal drifts In<br />

addition <strong>the</strong> sensor shows cross sensitivity with water.<br />

III. STABILITY AND SELECTIVITY<br />

A. NANOWIRES<br />

With respect to sensor stability single crystalline<br />

nanowires are basically superior to nanocrystalline<br />

sensing layers. Thus SnO2 nanowires have been<br />

fabricated and implemented as sensing elements.<br />

The SnO2-nanowires are realized in a two-stage<br />

procedure: First <strong>of</strong> all nanocrystalline SnO2-films<br />

are fabricated by <strong>the</strong> spray pyrolysis process described<br />

above. A tempering process in Aratmosphere<br />

(800-1100°C) leads to growth <strong>of</strong> single<br />

crystalline SnO2-nanowires. Fig. 3 shows <strong>the</strong><br />

SEM-photograph <strong>of</strong> a single SnO2-nanowire sensor<br />

with diameter <strong>of</strong> ~ 85 nm and a length <strong>of</strong><br />

~ 55 μm. Fig. 4 represents <strong>the</strong> response <strong>of</strong> <strong>the</strong><br />

SnO2-nanowire sensor at an operating temperature<br />

<strong>of</strong> 200°C when exposed to CH4-gas pulses with a<br />

concentration <strong>of</strong> only 3 ppm. The nanowire sensors<br />

show extremely high sensitivity towards CO<br />

and CH4, have very short response and recovery<br />

times and can be operated at lower temperatures<br />

than <strong>the</strong> nanocrystalline sensors.<br />

CH4 Flow [sccm]<br />

Fig. 3: SEM-photograph <strong>of</strong> a single SnO2nanowire<br />

sensor<br />

1,0<br />

3,00E+06<br />

0,9<br />

2,98E+06<br />

0,8<br />

2,96E+06<br />

0,7<br />

2,94E+06<br />

0,6<br />

2,92E+06<br />

0,5<br />

2,90E+06<br />

0,4<br />

2,88E+06<br />

0,3<br />

2,86E+06<br />

0,2<br />

2,84E+06<br />

0,1<br />

CH4 flow<br />

Sensor R [Ω]<br />

2,82E+06<br />

0,0<br />

2,80E+06<br />

300 600 900 1200 1500 1800 2100 2400<br />

Time [s]<br />

Fig. 4: Nanowire sensor response to 3ppm CH 4.<br />

Sensor Resistance [Ω]<br />

146<br />

B. SURFACE DOPING<br />

With respect to selectivity issues, additional surface<br />

doping by sputter deposition <strong>of</strong> a few monolayers <strong>of</strong><br />

functional layers is a powerful strategy both for ultrathin<br />

sensing layers as well as for nanowires. First Ptdoping<br />

results demonstrate improved drift behaviour<br />

and very high sensitivity as well as lower cross sensitivity<br />

towards humidity in comparison with undoped<br />

sensor devices.<br />

IV. CONCLUSION<br />

The high sensitivity as well as <strong>the</strong> low cost production<br />

process, <strong>the</strong> small size and excellent response<br />

and recovery times show <strong>the</strong> high potential <strong>of</strong> <strong>the</strong><br />

presented nanosensor devices for gas detection applications.<br />

REFERENCES<br />

[1] H. Meixner, J.Gerblinger, U.Lampe, M.Fleischer,<br />

Sens. Act. B 23 (1995) 119-125.<br />

[2] M. Fleischer, H. Meixner, Sens. Act. B 43 (1997)<br />

1-10.<br />

[3] G. Korotcenkov, Sens. Act. B 107 (2005) 209-<br />

232.<br />

[4] G. Korotcenkov, Mater. Sci. Eng. B 139 (2007)<br />

1-23.<br />

[5] W. Göpel, K.D. Schierbaum, Sens. Act. B 26-27<br />

(1995) 1-12.<br />

[6] J. Kappler, A. Tomescu, N. Barsan, U. Weimar,<br />

Thin Solid Films 391 (2001) 186-191.<br />

[7] G. Korotcenkov, V. Brynzari, S. Dmitriev, Mater.<br />

Sci. Eng. B 56 (1999) 195-204.<br />

[8] B. Schumacher, D. Szaboó, S. Schlabach, R.<br />

Ochs, H. Müller, M. Bruns, Mater. Res. Soc.<br />

Symp. Proc. 900E (2006) 6.1-6.6.<br />

[9] V. Malyshev, A. Pislyakov, Sens. Act. B 123<br />

(2007) 71-81.<br />

[10] E. Comini, Analytica Chimica Acta 568 (1-2)<br />

(2006) 28-40.<br />

[11] F. Patolsky, C.M. Lieber, Mat. Today 8 (2005)<br />

20-28.<br />

[12] C. Li, D. Zhang, X. Liu, S. Han, T. Tang, J.<br />

Han, C. Zhou, Appl. Phys Lett. 81 (10) (2002)<br />

1869-1871.<br />

[13] A. Tischner, T. Maier, C. Stepper, A. Köck,<br />

Sens. Act. B 134 (<strong>2008</strong>) 796-802.


��������������������������������������������������������������<br />

�<br />

��������������� � ����������������� � ���������������� � �<br />

� �������������������������������������������������<br />

�����������������<br />

������ imohsin@mail.zserv.tuwien.ac.at<br />

�<br />

�<br />

��������� ������ ����������� �������� �������<br />

���������� ��� ������� ���� ������ ���� ��������������<br />

���������� ������������� ��� ������� ������� ������ ���<br />

����������������������������������������������<br />

������ ����� ����� ��� ������� ���� ������ ��������<br />

����������� ���� ������� ���������� ����� ���������<br />

������������������������������������������������<br />

���� �������� ���� ����� ��������� ������ ���������<br />

���������������������������������������������������<br />

�������� ������� ���� ������������ �������� �������<br />

����� �������� ��� ��������������� � �� ���� ��� ��������<br />

��������������������������������������������������<br />

�������� ��� �������������� ��������� ����� ������������<br />

�����������������������������������������������<br />

���������������������������������������������������<br />

���������������������������������������������������<br />

���� ������� ����� �������� ���� �������� ���� ��� �����<br />

�����������������������<br />

��� ��������������<br />

���������������������������������������������������<br />

�������� ��� ������� ����������� ���� �������� �������<br />

���� ��������� ����� ���� ������� ����� �����������<br />

����������������������������������������������������<br />

���������� ������� ���� ���� �������� ������ ���� ���<br />

������� ��� ������ ��� �������� ���������� ��������� ���<br />

�������������������������������������������������<br />

�������������������������������������������������<br />

����� ��� ����������� �������� ���� ���������� ������<br />

�������������������������������<br />

��� �������� ������ ���� ����������� ������� ��������<br />

���������������������������������������������������<br />

������ ���� �������������� ���� ���� ����������� ���<br />

���������� ������������ ������� ������� ������ ���<br />

�����������������������������������������������<br />

����������������������������������������������<br />

������� ������������ �� ���� ��������������� ���������<br />

���������������������������������������������������<br />

������� ����������� ������ ��������� ��� �������������<br />

��������������������<br />

���� ��������������<br />

���� ���� ����������� ��� ���������� ������������<br />

������� ������� ������ �������������� ��������<br />

147<br />

�<br />

����������������������������������������������<br />

�������� ����������� ���� ���� �������� ����� ��������<br />

������������������������������������������������<br />

����� ���� ���������� ���������� ��������� ����� �����<br />

������������ � �������������� � �������������������� � ��<br />

����������� �� ������������������ � ������������� �� ��<br />

���� ��� �������� ��� ��������� ������������ ��������<br />

���� ������������ ����� ���� ����� ���� �������� ����<br />

��������� ���� �� �� ��������������������������<br />

��������� � �������� ����� ����� ��������� ��� �����<br />

�������� �������� ����� ��������� ����� �������� �����<br />

� � ������������ �� ������ �� ������ �� ��������� � ������<br />

������������������������������������������������<br />

�������� �� ��� ��������� ���� ����� ������ ���������<br />

������������������������������ � ������������������<br />

����� ��������� ����� ����� ���������� ���� ���������<br />

�����������������������������������������������<br />

����� ���������������������<br />

������� ����������� ���� ����������� ����� ������� ������<br />

����������������������������������������������������� � ��<br />

������ ���� ��� ����������� ��� ��������� ������������ ���<br />

�����������������������������������������������������<br />

���������������������������������������������������������<br />

��������� ���� ���������� ����� �������� ���� ��������� ��� ����<br />

���������������������������������������������������������<br />

����� ������� ����� ���� ��������� ���� ���������� �������<br />

������������������������������������������������������������<br />

������������������������������������������������������<br />

������� ����� ���� ��������� ���� ���������� ��� �������<br />

����������������������������������������������������<br />

������ ���� ����� ���� ���� ��������� �������� ���� ����� ���<br />

������������������������������������������������������<br />

������������������������������������������������������<br />

��������������������������������������������������������<br />

����� ��� �� ����������� ������� ����� �������� ��� ������ ����<br />

���������� ����� � � ����� ������ ��� ������������ ����� ����<br />

������ ����� ��������� ���� ����������� ����������� �����<br />

��������� ��� ������� ��� ���� ��������� ��������� ������<br />

����������������������������������������������������<br />

������� ������ ������������� ���� ����������� ��� ������ ���<br />

������ ��� ����� ��� ���� ������� �� ������� ��������� ���<br />

����������������������������������������������������<br />

����������������������������������������������


��������<br />

���������������������������������������������������� � ��<br />

�<br />

��� ����� ������ ����� ��� �������������������������<br />

����������������������������������������������������<br />

���������������� � ����������������������<br />

�<br />

�����������<br />

������� ������ ��������� ���������<br />

���������� ����� ����� ����� �����<br />

�������� ������� ������� ������� �������<br />

��������� ��������� ��������� ���������<br />

���� ���� ���� ����<br />

���� ����� �������� �������� ��������<br />

���� ����� ������� �������� ��������<br />

����<br />

�<br />

����� �������� �������� ��������<br />

��������������������������������������������<br />

�<br />

�����������������������������������������������<br />

��������������������������������� � ������������������<br />

�<br />

�<br />

����������������������� ������������������������<br />

����������������������<br />

������<br />

����� ������<br />

������� ������<br />

���� ������<br />

�<br />

����������������������������������� �� ����������������<br />

���������������������������������<br />

�<br />

��<br />

�<br />

148<br />

��������������������������������������������������� � ��<br />

�<br />

��������<br />

�<br />

��<br />

��<br />

��<br />

��<br />

���<br />

���<br />

��������������<br />

����������������<br />

��������������<br />

��������������<br />

��������������������<br />

��������������<br />

��������������������<br />

��� ��� ��� ��� ��� ��� ��� ��� ���<br />

��������������<br />

�<br />

��������������<br />

������������������� ��������������<br />

������������������������������������������������<br />

������� ����������� ����� ��� ��� ������<br />

������<br />

�����������<br />

�����������������������������������������������������<br />

������������������������������������������������������<br />

���� ��������� ������� ����� ����� �� ������� ��� ����� ���<br />

��������������������������������������������������<br />

����������������������������������������������������<br />

������������������������������������������������������<br />

������������������������������������������������<br />

�����������<br />

�������������������������������������������������<br />

������������������������������������������������<br />

���������������������������������������������<br />

��������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������������<br />

�������������������������<br />

�<br />

�����<br />

����� ���<br />

��� ��� ��� ���<br />

���<br />

���<br />

���<br />

���<br />

���<br />

�<br />

���<br />

���<br />

����<br />

����<br />

����<br />

����<br />


Electron microscopy <strong>of</strong> LiAlO2 -<br />

a promising substrate for opto-electronic devices<br />

Walid Hetaba and Peter Schattschneider (Faculty Mentor)<br />

Institute for Solid State Physics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: walid.hetaba@student.tuwien.ac.at<br />

Abstract — γ-LiAlO2 is a promising substrate for GaN epitaxy.<br />

Inclusions in γ-LiAlO2-specimen were analysed using<br />

energy loss near edge structure (ELNES) <strong>of</strong> O K-edge. The<br />

ELNES spectra were simulated for different materials using<br />

WIEN97 program. It is shown that <strong>the</strong> inclusions can be assigned<br />

to LiAl5O8. As an effect <strong>of</strong> electron beam radiation damage<br />

it can be seen that γ-LiAlO2 undergoes a transformation to<br />

Al2O3. Experimental O K-edge spectra contain an additional<br />

peak at 531 eV, which can be attributed to molecular oxygen<br />

O2. This result suggests that <strong>the</strong> electron beam induced LiAlO2<br />

decomposition occurs through <strong>the</strong> loss <strong>of</strong> O2.<br />

I. INTRODUCTION<br />

(100)γ-LiAlO2 is considered to be a promising substrate<br />

which allows for growth <strong>of</strong> both c- and m-plane oriented<br />

GaN layers [1]. In contrast to conventional substrates<br />

like SiC <strong>the</strong> advantage <strong>of</strong> (100)γ-LiAlO2 is <strong>the</strong> very<br />

small lattice mismatch <strong>of</strong> about 1.4%. Thus, <strong>the</strong> GaN<br />

layers grown on this substrate show a low defect density.<br />

Also, it allows <strong>the</strong> growing <strong>of</strong> non-polar m-plane GaN<br />

which shows absence <strong>of</strong> internal electrostatic fields along<br />

<strong>the</strong> growing direction. This makes it possible to generate<br />

new efficient white light-emitting diodes (LEDs) [2].<br />

At certain growth conditions <strong>the</strong> LiAlO2-single crystalls<br />

contain a core <strong>of</strong> inclusions. Also at <strong>the</strong> GaN/LiAlO2 interface<br />

<strong>the</strong>re are inclusions which are partly surrounded<br />

by cavities in <strong>the</strong> LiAlO2-matrix. These inclusions at<br />

<strong>the</strong> interface are suspected to be <strong>the</strong> cause <strong>of</strong> a spontaneous<br />

separation <strong>of</strong> thick GaN layers. This effect may be<br />

used in <strong>the</strong> future to generate GaN layers for homoepitaxy<br />

which would allow to produce almost defect free<br />

GaN wafers.<br />

Preliminary transmission electron microscopy (TEM)<br />

analysis using electron diffraction showed that <strong>the</strong> inclusions<br />

mentioned above consist <strong>of</strong> LiAl5O8 [3]. But also<br />

<strong>the</strong> formation <strong>of</strong> Al2O3 can occur. The aim <strong>of</strong> this research<br />

project was to see if <strong>the</strong>se materials can be clearly<br />

distinguished using Oxygen K-edge ELNES spectra.<br />

II. SIMULATIONS AND EXPERIMENTS<br />

Simulations were carried out using <strong>the</strong> full potential linear<br />

augmented plane wave code WIEN97. The unit cells<br />

<strong>of</strong> <strong>the</strong> three materials are <strong>of</strong> different size resulting in dif-<br />

149<br />

ferent simulation parameters and time consumption for<br />

simulation. The used parameters are <strong>the</strong>n verified by<br />

analysing monople-, dipole- and quadrupole-transition<br />

terms.<br />

TEM analysis comprising scanning transmission electron<br />

microscopy (STEM) and EELS was carried out on<br />

a 200 kV JEOL JEM2200FS microscope at Humboldt-<br />

University Berlin. In Figure 1 <strong>the</strong> simulated and experimental<br />

spectra are compared. Considering <strong>the</strong> known<br />

limitations <strong>of</strong> <strong>the</strong> simulation program, <strong>the</strong> very good<br />

agreement <strong>of</strong> simulation and experimental spectra is visible.<br />

Intensity [a.u.]<br />

14000<br />

12000<br />

10000<br />

8000<br />

6000<br />

4000<br />

2000<br />

O-K ELNES Simulation vs. Experiment<br />

LiAlO2 Experiment<br />

LiAlO2 Simulation<br />

LiAl5O8 Experiment<br />

LiAl5O8 Simulation<br />

Al2O3 Experiment<br />

Al2O3 Simulation<br />

0<br />

520 530 540 550 560<br />

Energy Loss [eV]<br />

570 580 590 600<br />

Figure 1: Comparison <strong>of</strong> simulated and experimental<br />

ELNES spectra <strong>of</strong> oxygen K-edge. From top to bottom<br />

LiAlO2, LiAl5O8 and Al2O3.<br />

III. RESULTS<br />

The simulations show that <strong>the</strong> materials γ-LiAlO2,<br />

LiAl5O8 and α-Al2O3 can be clearly distinguished using<br />

O K-edge ELNES. The investigations showed that <strong>the</strong><br />

inclusions in γ-LiAlO2 consist <strong>of</strong> LiAl5O8. It was also<br />

found that <strong>the</strong> specimen is damaged due to electron beam<br />

irradiation. Comparison <strong>of</strong> ELNES simulations and experimental<br />

spectra did not only show <strong>the</strong> already known<br />

transformation <strong>of</strong> LiAlO2 into LiAl5O8 [4] but also an<br />

unexpected transition to Al2O3 which is shown in Figure<br />

2. At <strong>the</strong> beginning <strong>the</strong> two peaks <strong>of</strong> LiAlO2 can


Intensity [a.u.]<br />

10000<br />

9000<br />

8000<br />

7000<br />

6000<br />

5000<br />

4000<br />

O-K ELNES<br />

3000<br />

500 520 540 560 580 600<br />

Energy Loss [eV]<br />

5s<br />

10s<br />

15s<br />

20s<br />

25s<br />

Figure 2: Spectra acquired at intervals <strong>of</strong> five seconds.<br />

The transformation <strong>of</strong> LiAlO2 to Al2O3 is clearly visible.<br />

The π ∗ -resonance peak <strong>of</strong> molecular oxygen at 531 eV is<br />

marked with an arrow.<br />

be clearly identified, while <strong>the</strong> intensity <strong>of</strong> <strong>the</strong> first peak<br />

is decreasing until <strong>the</strong> peak is transformed to <strong>the</strong> shoulder<br />

in <strong>the</strong> rising edge <strong>of</strong> Al2O3. These changes can be<br />

attributed to <strong>the</strong> following reactions:<br />

5 LiAlO2 −→ 2 Li2O ↑ + LiAl5O8 (1)<br />

2 LiAl5O8 −→ Li2O ↑ +5Al2O3. (2)<br />

The spectra contain an additional peak at 531 eV which is<br />

marked with an arrow in Figure 2. This peak is observed<br />

as long as <strong>the</strong> transformation from LiAlO2 to Al2O3<br />

takes place. This suggests that it can be associated with<br />

<strong>the</strong> loss <strong>of</strong> Li2O. The peak can be attributed to a so-called<br />

π ∗ -resonance which occurs when an electron is excited<br />

from a σ-molecular orbital to a π ∗ -orbital [5]. Li2O has<br />

a similar structure as H2O which means that <strong>the</strong>re are<br />

only σ-bondings and no π ∗ -resonance can occur. Thus,<br />

<strong>the</strong> peak at 531 eV has to be attributed to molecular oxygen.<br />

The molecular orbital structure <strong>of</strong> O2 is shown in<br />

Figure 3. All this means that in <strong>the</strong> local environment <strong>of</strong><br />

<strong>the</strong> beam cross-section O2 is set free due to heating <strong>of</strong><br />

<strong>the</strong> sample. This causes a change in <strong>the</strong> chemical composition<br />

<strong>of</strong> <strong>the</strong> sample so that it becomes more Al-rich.<br />

IV. CONCLUSION<br />

For <strong>the</strong> first time <strong>the</strong> inclusions in γ-LiAlO2 are investigated<br />

using oxygen K-edge ELNES spectra. It is<br />

suggested that <strong>the</strong> inclusions at <strong>the</strong> GaN/LiAlO2 interface<br />

cause <strong>the</strong> spontaneous separation <strong>of</strong> thick GaN layers<br />

which in <strong>the</strong> future may be used to generate GaN<br />

wafers for homoepitaxy. An unexpected transfomation<br />

<strong>of</strong> LiAlO2 into Al2O3 was observed. The ocurrence <strong>of</strong><br />

an additional peak at 531 eV which can be attributed to<br />

a π ∗ -resonance allows a better understandig <strong>of</strong> this transition.<br />

Fur<strong>the</strong>r investigation <strong>of</strong> <strong>the</strong> LiAlO2 transforma-<br />

150<br />

Figure 3: Molecular orbitals <strong>of</strong> O2. An electron can be<br />

excited from a σ-orbital into a partly occupied π ∗ -orbital.<br />

tion would allow an adaption <strong>of</strong> <strong>the</strong> production paramters<br />

for fabrication <strong>of</strong> defect-free LiAlO2 for subsequent GaN<br />

epitaxy. With this substrate a new generation <strong>of</strong> high<br />

quality LEDs could be produced.<br />

ACKNOWLEDGMENTS<br />

The Author wants to thank W. Neumann and <strong>the</strong><br />

members <strong>of</strong> his workgroup for crystallography at <strong>the</strong><br />

Humboldt-University Berlin for <strong>the</strong>ir support with <strong>the</strong><br />

experimental part <strong>of</strong> this work.<br />

REFERENCES<br />

[1] H.P. Maruska, D.W. Hill, M.C. Chou, J.J. Gallagher,<br />

and B.H. Chai. Free-standing non-polar gallium nitride<br />

substrates. Opto-Electronics Review, 11(1):7–<br />

17, 2003.<br />

[2] P. Waltereit, O. Brandt, A. Trampert, H.T. Grahn,<br />

J. Menniger, M. Ramsteiner, M. Reiche, and K.H.<br />

Ploog. Nitride semiconductors free <strong>of</strong> electrostatic<br />

fields for efficient white light-emitting diodes. Nature,<br />

406:865–868, 2000.<br />

[3] A. Mogilatenko, W. Neumann, E. Richter, M. Weyers,<br />

B. Veličkov, and R. Uecker. Mechanism<br />

<strong>of</strong> LiAlO2 decomposition during <strong>the</strong> GaN growth<br />

on (100)γ-LiAlO2. Journal <strong>of</strong> Applied Physics,<br />

102:023519, 2007.<br />

[4] M.H. Auvrey-Gély, A. Dunlop, and L.W. Hobbs. Irradiation<br />

damage in lithium ceramics. Journal <strong>of</strong> Nuclear<br />

Materials, 133&134:230–233, 1985.<br />

[5] J. Stöhr. NEXAFS Spectroscopy. Springer Verlag,<br />

Berlin, 1996.


�������������������������������������������������<br />

�<br />

�������������������������������������������������<br />

�������������������������������<br />

�������������������������������������������������������������������������������������<br />

���������������������<br />

�������������������������������������������������������������<br />

�<br />

�<br />

�������������������������������������������������<br />

���������� ����������� ��� ������ �������� ������<br />

������� �������� ��������������� ����� ����� ������<br />

�������� ������� ����������� ���� ���������� ���<br />

������������� ������� ����� ��������� ������ ����� ��<br />

������� ���� ���������� ����������� ��� ���� ������� ����<br />

����������������������������������������������������<br />

��������������������������������������������������<br />

������ ��� �������� ������� ��� ���� ������ ������ �����<br />

�������� ����������� ��������� ��� �� ���� �������� �����<br />

���������� ���� ��������� ���� ��� ����� ��������� �����<br />

�����������<br />

��� �������������<br />

���� ����� ���������� ��� ����� ����� ���� ��� ������� ��<br />

���������� ����������� �������� ��� ����������<br />

����������� ��� ������ �������� ��������� ���� ������������<br />

����������� ������� ��� ����� ���������� ���� ������������<br />

���� ���������� ���������� ���� ���� ��� ��������������<br />

��������������� ������� ��������������� ���� ����� ���<br />

�����������������<br />

���� ������������������������<br />

���� ������� ����������� ������������ �������������<br />

�������������� �������� ��� ������� �������� ����<br />

������ ���� ������ ��� ����� ������� ��� ������� ��� ��� �������<br />

���������������� ����� �������� ����� ����� ����<br />

������������������������<br />

��������������������������������������������������������<br />

������������ ��������� ��� ���� �������������� �����<br />

������������������������������������������������������<br />

��� ����� ����� ������������ ��� ���� ���������� ���<br />

��������� �� ������ ������� ������������� ����� �� ������<br />

�������������������������������������������������������<br />

�������������������������������������������������������<br />

�����������������������������������������������������<br />

������� ��� ������� ��� ���� ������ ������� ���� ���������� ���<br />

���������� �������� ���� ���� ��� ������� ����������<br />

�����������������������������������������������������<br />

���������������<br />

151<br />

����� ������������������<br />

���� ���������� ����������� ���� ��������� ������ ��<br />

������������ ����������������� ����� ����� ������� ����<br />

�������������<br />

�<br />

�������������������������<br />

��<br />

��<br />

�<br />

�<br />

��<br />

���<br />

���<br />

����������������<br />

��� ��� �� � � �� ��<br />

����������������<br />

����������<br />

�<br />

�<br />

��������������������������������������������������<br />

����������<br />

�<br />

��������������������������������������������������<br />

������ ��������� �������� ��� �������� ����� �������� ����<br />

������������� ��� ����� ������������ ������������ �����<br />

���������� ��������� ��� ���������� ����� �� ���������<br />

���������������������������������������������������<br />

�� �������� ����� ������� ��� ������������� ��������� ����<br />

���������� ��� ����������� �������� ���� ����� ������� ���<br />

������������������������������������������������������<br />

������ ��� ���� ������ ���� ����� ������� ���� �������� ���<br />

���� ����� ��� ����� ����� �������� ������ ����������<br />

������������<br />

���� ������������������������������<br />

���������� �������� ������ ������ ����� ���� �������<br />

�������������������������������������������������������<br />

�������� ���� ����������� ���� �������� ����� ��������� �����<br />

������ ��������� ��� �������� �������� ��������


���������������������������������������������������<br />

�������������������������������������������������������<br />

��������������������������������������<br />

�<br />

�<br />

�<br />

������������������������������������<br />

�<br />

���� �������� ���� ����������� ������ ����<br />

���������������������������������������������������<br />

������� ��� ���� ������ ����� ��� ���� �������� �� ����� ���� ���<br />

������������������������������������������������������<br />

�����������������������������������������������������<br />

������������������������������������������������������<br />

���� ���������� ��� ����� ���� ������ ���� ������� ���<br />

��������������������������������������������������������<br />

���������������<br />

�<br />

����������������<br />

�<br />

�<br />

�<br />

�<br />

��<br />

��<br />

��������<br />

������<br />

��������<br />

���������������<br />

�������<br />

��<br />

� ��� � ��� � ��� � ���<br />

��������<br />

�<br />

����������������������������������������������������<br />

���<br />

�<br />

��� ��������������������<br />

�������<br />

���������<br />

���������������<br />

��������������<br />

��������� ����� ������� ��� ���������� ��� ����� ������<br />

�������������������������������������������������������<br />

�� ���������� ���������� ����������� ������� ���������<br />

������������������������������������������������<br />

��������� ��� ���� ������� �������� ��� ������� ������ ���<br />

����������� ��� ��� ��������� ����� ���� ����������� ������<br />

������������������������������������������������������<br />

�������������������������������������������������<br />

������� ���������� ��� �� ������� ��� ����� ��������� ������<br />

������������������������������������<br />

�<br />

152<br />

�<br />

���������������������������������������������������<br />

��� ���� ����������� �������� ��� ����� ��������<br />

�<br />

��������������������������������������������������<br />

�<br />

���� ������������<br />

�� ������� ���� ���������� ���������� ���� ��������� ����<br />

�������������������������������������������������������<br />

�����������������������������������������������������<br />

����� ���������� ������ ������� ����������� �����������<br />

����� ��������� ��������� ������ ����� ��� ������ ��� ����<br />

��������� ������������� �������������������������������<br />

�������������������������������<br />

����������������<br />

����������������������������������������������<br />

��� ���� ������ ������� ��� ���� ������ ���������<br />

�������� ������ ������ ���� ������������ ����<br />

����������������������������������������������<br />

�����������<br />

��������<br />

������<br />

����������<br />

���� ��� ��������� �������� ��������������� ���������<br />

�����������������������������������������<br />

�������� ���������� �� ���� ����������������� �����<br />

������������������������� �� �����������������<br />

���� ��� ������� ��� ������� ����������� ��� ��������<br />

����������������������������������������������<br />

����������������������������������������������<br />

�����������������������������������������������<br />

��������������������������������������������<br />

�������������������������������������������<br />

�����������������������������������������������<br />

�����������������������������������������<br />


Silver plated tungsten carbide powders for<br />

electrical contact materials with improved homogeneity<br />

Robert Christian Hula and Christian Edtmaier (Faculty Mentor)<br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: robert.hula@tuwien.ac.at<br />

Abstract — Electrical contact materials based on<br />

tungsten carbide and silver (WCAg) are most commonly<br />

produced by infiltration techniques as <strong>the</strong>y<br />

are superior to mixing-pressing-sintering. Never<strong>the</strong>less,<br />

infiltrated WCAg shows inhomogeneity and<br />

more or less porosity. Moreover, <strong>the</strong> process is not<br />

suitable for high volume fractions <strong>of</strong> silver or submicron<br />

tungsten carbide powders. A new technique<br />

using chemical silver plating <strong>of</strong> tungsten carbide<br />

was investigated. The coated composite powders<br />

were consolidated using uniaxial cold pressing<br />

followed by liquid phase sintering. Products obtained,<br />

showed distinct higher homogeneity than<br />

industrial infiltrated products.<br />

I. INTRODUCTION<br />

Tungsten carbide–silver composites (WCAg) are<br />

common electrical contact materials e.g. for heavyduty<br />

circuit breakers and vacuum contactors [1]. The<br />

combination <strong>of</strong> high <strong>the</strong>rmal and electrical conductivity<br />

(Ag) with hardness and refractory properties<br />

(WC) leads to high arc resistance and low chopping<br />

currents [2]. Typically, <strong>the</strong> silver fraction is in <strong>the</strong><br />

range <strong>of</strong> 38 to 67 w%, most likely around 40 w%.<br />

Properties <strong>of</strong> WCAg contact materials are standardized<br />

in ASTM B 663-89.<br />

Industrial production mostly uses infiltration techniques<br />

<strong>of</strong> tungsten carbide preforms with liquid silver.<br />

As <strong>the</strong> green body has to have distinct<br />

cohesiveness, complete infiltration is not possible.<br />

Therefore <strong>the</strong> products show residual pores and also<br />

poor homogeneity on a microscopic level. Beside<br />

that, it is not possible to use submicron tungsten<br />

carbide and also high volume fractions <strong>of</strong> silver are<br />

not achievable using infiltration techniques.<br />

Chemical plating methods <strong>of</strong>fer a new approach<br />

for improved homogeneity <strong>of</strong> WCAg [3]. By covering<br />

single tungsten carbide particles with silver by<br />

wet chemical deposition, idealized dispersion <strong>of</strong> <strong>the</strong><br />

components can be expected.<br />

153<br />

II. EXPERIMENTAL<br />

The substrate was tungsten carbide powder <strong>of</strong> 3 μm<br />

average particle diameter and an oxygen concentration<br />

<strong>of</strong> 550 μg/g (delivered by H.C.Starck). The<br />

powder was chemically plated, pressed and sintered<br />

as described beneath.<br />

A. CHEMICAL PLATING<br />

In order to minimize impurities, a method using<br />

hydrazine hydrate as <strong>the</strong> reducing agent was used.<br />

Following equation 1, hydrazine completely dissipates<br />

during reaction with silver.<br />

Ag + + N2H4 + 4OH - --> Ag + N2 + 4H2O (1)<br />

To maximize yield and reduction strength, a high pH<br />

value is preferable. Therefore alkaline solutions <strong>of</strong><br />

ammonia and ammonium bicarbonate have been<br />

applied [4]. Ammonia also has <strong>the</strong> advantage to<br />

stabilize silver containing solutions due to formation<br />

<strong>of</strong> silver diamine complex.<br />

The reaction was performed in a flat flange glass<br />

reaction vessel with PTFE stirrer at room temperature.<br />

While stirring, <strong>the</strong> silver nitrate (99,97 %, Umicore)<br />

containing ammonia solution was added dropwise<br />

to <strong>the</strong> suspension <strong>of</strong> tungsten carbide in alkaline<br />

hydrazine hydrate solution. Table 1 shows <strong>the</strong> composition<br />

<strong>of</strong> <strong>the</strong> used solutions. The plated powders<br />

were separated by filtration, washed with water and<br />

ethanol and were dried in a desiccator under vacuum<br />

at 50 °C.<br />

Component Solution A<br />

mol/L<br />

AgNO3<br />

N2H4.H2O<br />

NH4HCO3<br />

NH4OH<br />

WC<br />

0.127<br />

-<br />

1.897<br />

3.340<br />

-<br />

Solution B<br />

mol/L<br />

-<br />

0.082<br />

-<br />

-<br />

0.105<br />

Table 1: Composition <strong>of</strong> <strong>the</strong> chemical plating<br />

solutions


Figure 1: Silver plated tungsten carbide<br />

(SEM, 200x (top) and 5000x (bottom))<br />

Characterisation was done by SEM (Philips<br />

XL30, 20kV), <strong>the</strong> silver concentration was<br />

checked by volumetric method (Volhard) and <strong>the</strong><br />

oxygen content was measured by N/O/H-analyzer<br />

(LECO TCH 600).<br />

B. CONSOLIDATION<br />

The plated powders were cold pressed (300 MPa)<br />

and liquid phase sintered under flowing hydrogen (1-<br />

2 L/min) at 1100 °C (dwell time: 30 minutes). The<br />

samples were characterized by light optical microscopy<br />

(LOM) and density measurements.<br />

III. RESULTS<br />

The plated WCAg powder is shown in Fig. 1. Silver<br />

was deposited homogeneous forming smooth surfaces<br />

and showing no detachment. The silver concentration<br />

<strong>of</strong> 40,2 w% was within <strong>the</strong> desired range and<br />

<strong>the</strong> oxygen concentration was 1808 μg/g.<br />

After consolidation, <strong>the</strong> composite had some porosity<br />

on <strong>the</strong> surface and a density <strong>of</strong> 10,5 g/cm 3<br />

(80 % <strong>of</strong> <strong>the</strong>oretical value). Never<strong>the</strong>less, <strong>the</strong> samples<br />

had much higher density inside. The microstruc<br />

ture was highly homogeneous without any substantial<br />

silver or tungsten carbide culmination.<br />

154<br />

Figure 2: Comparison <strong>of</strong> microstructure: Sintered<br />

product from plated WCAg (left) and industrial<br />

infiltrated WCAg (right) with same composition.<br />

(LOM, 200x (top), 500x (middle) 1000x (bottom))<br />

In Fig. 2, consolidated WCAg, made from silver<br />

plated tungsten carbide is shown next to industrial<br />

infiltrated WCAg, made out <strong>of</strong> <strong>the</strong> same tungsten<br />

carbide powder and with <strong>the</strong> same silver concentration<br />

like <strong>the</strong> plated one. A significant improvement<br />

<strong>of</strong> homogeneity due to <strong>the</strong> use <strong>of</strong> silver plated tungsten<br />

carbide is distinguishable. Therefore a positive<br />

effect on <strong>the</strong> arc erosion resistance may be expected<br />

but have not been measured yet.<br />

REFERENCES<br />

[1] J. R. Davis (Ed.). Metals Handbook, volume 3.<br />

ASM International, Metals Park, Ohio, USA,<br />

10th edition, 1990<br />

[2] S. Temborius and M. Lindmayer. Stromnullverhalten<br />

unter Vakuum-Lastschalter-Bedingungen.<br />

In 15. Fachtagung Kontaktverhalten und Schalten,<br />

pages 1–10, Karlsruhe, Germany, 1999.<br />

[3] R. C. Hula. Silberbeschichtung von Wolframcarbidpulvern<br />

zur Anwendung in elektrischen Kontaktwerkst<strong>of</strong>fen.<br />

Diploma Thesis, TU Wien, Vienna,<br />

Austria, February 2007.<br />

[4] A. S. Kozlov and T. Palanisamy and D. Narasimhan.<br />

Electroless silver plating, Patent<br />

WO02/04711A2, Honeywell, 2002.


Investigation <strong>of</strong> mechanical properties <strong>of</strong> polymer-derived<br />

Si-C-N/Al2O3-composites with low shrinkage<br />

Thomas Konegger and Antje Liersch (Faculty Mentor)<br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: konegger@mail.tuwien.ac.at<br />

Abstract — The mechanical characteristics <strong>of</strong> a<br />

ceramic composite material using a polysilazanederived<br />

binder phase and alumina filler were investigated.<br />

Green bodies with high stability and good<br />

machinability were manufactured by warm-pressing<br />

<strong>of</strong> <strong>the</strong> polysilazane-coated filler powder, <strong>the</strong>reby<br />

<strong>the</strong>rmally cross-linking <strong>the</strong> polymer which was used<br />

in a volume fraction <strong>of</strong> 20 to 40 Vol.%. Shrinkage<br />

during pyrolysis was very low, reaching a maximum<br />

<strong>of</strong> 2.3 % at 1200 °C.<br />

I. INTRODUCTION<br />

An inherent problem <strong>of</strong> ceramic materials is <strong>the</strong> need<br />

for high process temperatures during sintering, for<br />

Si3N4 usually exceeding 1700 °C. For Si-based ceramic<br />

compounds, <strong>the</strong> <strong>the</strong>rmal conversion <strong>of</strong> Sicontaining<br />

polymeric precursors, for example polysilanes,<br />

polycarbosilanes <strong>of</strong> polysilazanes, is an alternative,<br />

resulting in covalent ceramics in amorphous<br />

state [1]. A disadvantage is <strong>the</strong> high shrinkage rate<br />

resulting in cracks, which can be reduced by an application<br />

<strong>of</strong> inert ceramic fillers like B4C or Si3N4 [2].<br />

Due to <strong>the</strong> polymeric nature <strong>of</strong> <strong>the</strong> material after<br />

cross-linking, a good machinability in green state is<br />

given. A possible method <strong>of</strong> forming <strong>of</strong> said materials<br />

is <strong>the</strong> precoating <strong>of</strong> <strong>the</strong> filler powder with <strong>the</strong><br />

polymer and subsequent cross-linking <strong>of</strong> <strong>the</strong> polymeric<br />

contents by warm-pressing, followed by a<br />

pyrolytic conversion <strong>of</strong> <strong>the</strong> polymeric precursor [3].<br />

The combination <strong>of</strong> a polysilazane precursor with<br />

zirconia or alumina/zirconia powders led to specimens<br />

with high green strength and a microhardness<br />

<strong>of</strong> up to 1080 HV0.5 after pyrolysis at 1400 °C. The<br />

application <strong>of</strong> zirconia/alumina resulted in a decrease<br />

in shrinkage but also in lower material strength [4].<br />

The aim <strong>of</strong> this study is <strong>the</strong> examination <strong>of</strong> ceramic<br />

composites with low shrinkage and good machinability<br />

in green state, enabling a near-net-shape<br />

production <strong>of</strong> ceramic parts. This particular investigation<br />

focuses on <strong>the</strong> determination <strong>of</strong> mechanical<br />

properties <strong>of</strong> polysilazane-alumina-composites that<br />

can be achieved by <strong>the</strong> variation <strong>of</strong> process parameters<br />

including polymer volume fraction, curing<br />

conditions and pyrolysis temperature.<br />

155<br />

II. EXPERIMENTAL<br />

A. MATERIALS<br />

A commercially available liquid poly(vinyl)silazane<br />

(KiON HTT 1800, Clariant) was used as preceramic<br />

polymer binder. As ceramic filler, Al2O3 powder<br />

(Almatis CT 3000 SG) with a median particle size<br />

(D50) <strong>of</strong> 0.8 μm was used.<br />

B. FABRICATION OF SPECIMENS<br />

The alumina powder was attrition milled for 2 h,<br />

dried and sieved. The prepared powder was dispersed<br />

in dried acetone, mixed with <strong>the</strong> polysilazane and<br />

stirred for one hour. The amount <strong>of</strong> polysilazane was<br />

calculated to yield between 20 and 40 Vol.% after <strong>the</strong><br />

cross-linking step. After removal <strong>of</strong> <strong>the</strong> solvent under<br />

reduced pressure, <strong>the</strong> particulate mixture was ground<br />

and sieved through a 200 μm sieve.<br />

The cross-linking <strong>of</strong> <strong>the</strong> specimens was accomplished<br />

by warm-pressing <strong>of</strong> <strong>the</strong> polymer-coated<br />

powders at pressures between 25 and 55 MPa at<br />

200 °C for 4 h, resulting in cylindrical specimens<br />

with a diameter <strong>of</strong> 40 mm.<br />

Pyrolytic conversion was carried out in a tube furnace<br />

under nitrogen at temperatures <strong>of</strong> 1000 and<br />

1200 °C for 4 hours, with a heating rate <strong>of</strong> 36 K/h.<br />

C. CHARACTERIZATION<br />

The density <strong>of</strong> <strong>the</strong> specimens was determined by <strong>the</strong><br />

Archimedes method. A helium pycnometer (Quantachrome<br />

Ultrapycnometer 1000) was used to determine<br />

<strong>the</strong> open porosity <strong>of</strong> selected samples.<br />

The samples were cut and embedded in epoxy<br />

resin before grinding and polishing to 1 μm. Examination<br />

<strong>of</strong> <strong>the</strong> microstructure was carried out by light<br />

optical microscopy (Olympus GX51).<br />

Four point bending strength was determined with<br />

polished samples on <strong>the</strong> basis <strong>of</strong> DIN EN 843-1 on a<br />

universal testing machine (Zwick 1474).<br />

The hardness <strong>of</strong> <strong>the</strong> material was determined by<br />

<strong>the</strong> Vickers method with a load <strong>of</strong> 0.5 kg (Leco LM-<br />

100).


III. RESULTS AND DISCUSSION<br />

By <strong>the</strong> described method <strong>of</strong> production, specimens<br />

with a high stability in green state could be manufactured.<br />

Ceramography showed a homogeneous distribution<br />

<strong>of</strong> components in <strong>the</strong> ceramic/ceramiccomposite.<br />

The observed shrinkage <strong>of</strong> <strong>the</strong> specimens during<br />

<strong>the</strong>rmal transformation was very low, rising with<br />

volume fraction and process temperature to a maximum<br />

<strong>of</strong> 2.3 % at a polymer content <strong>of</strong> 40 Vol.% and<br />

a pyrolysis temperature <strong>of</strong> 1200 °C.<br />

By helium pycnometry and calculation <strong>of</strong> <strong>the</strong> <strong>the</strong>oretical<br />

density <strong>of</strong> <strong>the</strong> material, an open porosity between<br />

21 and 43 % could be determined for polysilazane<br />

contents <strong>of</strong> 40 Vol.% and 20 Vol.%, respectively.<br />

These values are in accordance to <strong>the</strong> relative<br />

density <strong>of</strong> <strong>the</strong> material determined by <strong>the</strong> Archimedes<br />

method.<br />

It could be shown that a good compaction <strong>of</strong> <strong>the</strong><br />

specimens during cross-linking is crucial in obtaining<br />

high flexural strength values. Figure 1 shows <strong>the</strong><br />

determined flexural strength <strong>of</strong> <strong>the</strong> material as a<br />

function <strong>of</strong> <strong>the</strong> polymer binder volume fraction. As<br />

can be seen, a high increase in material strength can<br />

be accomplished by an increase in binder content.<br />

Due to <strong>the</strong> polymeric nature <strong>of</strong> <strong>the</strong> polysilazane, <strong>the</strong><br />

mechanism <strong>of</strong> compaction includes viscous flow <strong>of</strong><br />

<strong>the</strong> material during warm-pressing. An increase in<br />

polymer content leads to an improved compaction,<br />

thus leading to a reduction <strong>of</strong> porosity <strong>of</strong> <strong>the</strong> pyrolyzed<br />

product and a higher structural integrity <strong>of</strong> <strong>the</strong><br />

material.<br />

Figure 1: Flexural strength <strong>of</strong> specimens as function<br />

<strong>of</strong> polysilazane volume fraction<br />

Figure 2 shows <strong>the</strong> influence <strong>of</strong> <strong>the</strong> pressure applied<br />

during <strong>the</strong>rmal cross-linking by warm-pressing<br />

on <strong>the</strong> material’s flexural strength. An increase in<br />

pressure leads to an increase in flexural strength,<br />

caused by <strong>the</strong> aforementioned improvement <strong>of</strong> densification<br />

which remains present through <strong>the</strong> course <strong>of</strong><br />

pyrolytic transformation and sintering.<br />

156<br />

Figure 2: Flexural strength <strong>of</strong> specimens as function<br />

<strong>of</strong> pressure during <strong>the</strong>rmal cross-linking<br />

The hardness <strong>of</strong> <strong>the</strong> material increased with process<br />

temperature and polymer content, averaging at<br />

385 HV0.5 for specimens with 40 Vol.% <strong>of</strong> polysilazane<br />

pyrolyzed at 1200 °C. Significant differences<br />

by a variation <strong>of</strong> cross-linking pressure could not be<br />

observed in <strong>the</strong> investigated range.<br />

It could be shown that a good compaction <strong>of</strong><br />

specimens during <strong>the</strong>rmal cross-linking is crucial to a<br />

high mechanical stability <strong>of</strong> <strong>the</strong> investigated composite<br />

material. By careful variation <strong>of</strong> process parameters,<br />

a material could be produced that combines a<br />

high green stability with a very low shrinkage during<br />

<strong>the</strong>rmal consolidation and mechanical characteristics<br />

that can be tailored to specific applications in <strong>the</strong><br />

field <strong>of</strong> structural ceramics.<br />

REFERENCES<br />

[1] P. Greil. Active-filler controlled pyrolysis <strong>of</strong><br />

preceramic polymers. Journal <strong>of</strong> <strong>the</strong> American<br />

Ceramic Society, 78(4):835-848, 1995.<br />

[2] H.J. Kleebe. Microstructure and stability <strong>of</strong><br />

polymer-derived ceramics; <strong>the</strong> Si-C-N system.<br />

physica status solidi (a), 166(1):297-<br />

313, 1998.<br />

[3] D. Galusek, J. Sedlacek and R. Riedel.<br />

Al2O3-SiC composites prepared by warm<br />

pressing and sintering <strong>of</strong> an organosilicon<br />

polymer-coated alumina powder. Journal <strong>of</strong><br />

<strong>the</strong> European Ceramic Society, 27(6):2385-<br />

2392, 2007.<br />

[4] T. Konegger, A. Liersch and R. Felzmann.<br />

Influence <strong>of</strong> oxide ceramic fillers on mechanical<br />

characteristics <strong>of</strong> polysilazanederived<br />

ceramic materials. In 2 nd International<br />

Congress on Ceramics – Global<br />

Roadmap for Ceramics & ICC2 <strong>Proceedings</strong>,<br />

Paper 3-P-10, Verona, Italy, July <strong>2008</strong>.


Preparation <strong>of</strong> Alkyne and Azide modified Spherosilicates<br />

Angelika Bachinger and Guido Kickelbick<br />

Institute <strong>of</strong> Materials Chemistry<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: angelika.bachinger+e165@tuwien.ac.at<br />

Abstract — The aim <strong>of</strong> this work is <strong>the</strong> use <strong>of</strong><br />

Huisgen 1,3-dipolar cycloaddition reaction for <strong>the</strong><br />

eight corner functionalization <strong>of</strong> spherosilicates in<br />

order to prepare new nanocomposites or literature<br />

known nanocomposites by a simple and efficient<br />

reaction, that is based on <strong>the</strong> building block system.<br />

I. INTRODUCTION<br />

Their small and defined sizes as well as <strong>the</strong> high<br />

density <strong>of</strong> functional groups and <strong>the</strong>ir rigid framework<br />

make silica cages very popular in materials<br />

chemistry. Many different syn<strong>the</strong>tic techniques<br />

have been used for <strong>the</strong>ir functionalization, such as<br />

silylation 1-6 , hydrosilation 6-11 , substitution reactions<br />

12,13 and a lot more. However, all <strong>of</strong> <strong>the</strong>se<br />

reactions require <strong>the</strong> absence <strong>of</strong> specific functional<br />

groups, a fact that constricts <strong>the</strong>ir applicability.<br />

Thus, <strong>the</strong> well-established and insensitive click<br />

reactions were applied on such systems by Binder<br />

et al 14 . These reactions are characterized by <strong>the</strong>ir<br />

high efficiency, simplicity, modularity and especially<br />

<strong>the</strong>ir broad applicability – due to <strong>the</strong>ir insensitivity<br />

towards most organic functionalities and<br />

selectivity 15 . Binder and his co-workers applied<br />

one <strong>of</strong> <strong>the</strong> most versatile click reactions – <strong>the</strong> Huisgen<br />

1,3-dipolar cycloaddition reaction – for <strong>the</strong><br />

one-corner functionalization <strong>of</strong> POSS cages 14 .<br />

Considering that this reaction has been shown to<br />

allow reactions also in sterically hindered environments<br />

16 , we assumed that <strong>the</strong> eight corner<br />

modification <strong>of</strong> spherosilicates and POSS cages<br />

might be possible by this approach.<br />

II. RESULTS AND DISCUSSION<br />

For <strong>the</strong> preparation <strong>of</strong> eight corner modified<br />

spherosilicates it was assumed that highly efficient<br />

functionalization reactions are required. Thus, a<br />

literature known spherosilicate having <strong>the</strong> instable<br />

silicon-chlorine bond on all eight corners was prepared<br />

and a substitution reaction with ethinyl magne-<br />

157<br />

siumbromide yielded <strong>the</strong> corresponding alkyne modified<br />

spherosilicate.<br />

The product was characterized by spectroscopic<br />

techniques such as 1 H NMR (Figure 1), 13 C NMR,<br />

29 Si NMR and FT-IR.<br />

Figure 1: 1 H NMR <strong>of</strong> alkyne modified spherosilicate<br />

Syn<strong>the</strong>sis <strong>of</strong> a spherosilicate with azide groups<br />

was conducted via tosylation <strong>of</strong> a literature known<br />

OH modified spherosilicate and subsequent substitution<br />

reaction with sodium azide.<br />

The product was characterized by 1 H NMR<br />

(Figure 2), 13 C NMR (Figure 2), 29 Si NMR and<br />

FT-IR.<br />

Figure 2: 1 H NMR (top) and 13 C NMR (bottom) <strong>of</strong><br />

azide modified spherosilicate


Click reaction <strong>of</strong> <strong>the</strong> alkyne modified spherosilicate<br />

with a model azide compound revealed partial<br />

success.<br />

ACKNOWLEDGMENTS<br />

We thank Dr. Sorin Ivanovici for providing good<br />

ideas and Dr. Michael Puchberger for NMR measurements.<br />

REFERENCES<br />

(1) Hoebbel, D.; Wieker, W.<br />

Zeitschrift fuer Anorganische und Allgemeine<br />

Chemie 1971, 384, 43-52.<br />

(2) Hasegawa, I.; Kuroda, K.; Kato, C.<br />

Bulletin <strong>of</strong> <strong>the</strong> Chemical Society <strong>of</strong> Japan 1986, 59,<br />

2279-83.<br />

(3) Hoebbel, D.; Weber, C.; Schmidt,<br />

H.; Krueger, R.-P. Journal <strong>of</strong> Sol-Gel Science and<br />

Technology 2002, 24, 121-129.<br />

(4) Hasegawa, I.; Ishida, M.; Motojima,<br />

S. Syn<strong>the</strong>sis and Reactivity in Inorganic and<br />

Metal-Organic Chemistry 1994, 24, 1099-110.<br />

(5) Harrison, P. G.; Hall, C. Main<br />

Group Metal Chemistry 1997, 20, 515-529.<br />

(6) Laine, R. M. Journal <strong>of</strong> Materials<br />

Chemistry 2005, 15, 3725-3744.<br />

(7) Feher, F. J.; Wyndham, K. D.;<br />

Soulivong, D.; Nguyen, F. Journal <strong>of</strong> <strong>the</strong> Chemical<br />

158<br />

Society, Dalton Transactions: Inorganic Chemistry<br />

1999, 1491-1498.<br />

(8) Laine, R. M.; Zhang, C.; Sellinger,<br />

A.; Viculis, L. Applied Organometallic Chemistry<br />

1998, 12, 715-723.<br />

(9) Zhang, C.; Laine, R. M. Journal <strong>of</strong><br />

<strong>the</strong> American Chemical Society 2000, 122, 6979-<br />

6988.<br />

(10) Zhang, L.; Abbenhuis, H. C. L.;<br />

Yang, Q.; Wang, Y.-M.; Magusin, P. C. M. M.;<br />

Mezari, B.; van Santen, R. A.; Li, C. Angewandte<br />

Chemie, International Edition 2007, 46, 5003-5006.<br />

(11) Holzinger, D.; Kickelbick, G.<br />

Journal <strong>of</strong> Polymer Science, Part A: Polymer Chemistry<br />

2002, 40, 3858-3872.<br />

(12) Hasegawa, I.; Niwa, T.; Takayama,<br />

T. Inorganic Chemistry Communications 2005, 8,<br />

159-161.<br />

(13) Kim, K.-M.; Ouchi, Y.; Chujo, Y.<br />

Polymer Bulletin (Berlin, Germany) 2003, 49, 341-<br />

348.<br />

(14) Binder, W. H.; Petraru, L.; Sachensh<strong>of</strong>er,<br />

R.; Zirbs, R. Monatshefte fuer Chemie 2006,<br />

137, 835-841.<br />

(15) Kolb, H. C.; Finn, M. G.;<br />

Sharpless, K. B. Angewandte Chemie, International<br />

Edition 2001, 40, 2004-2021.<br />

(16) Binder, W. H.; Sachsenh<strong>of</strong>er, R.<br />

Macromolecular Rapid Communications 2007, 28,<br />

15-54.


Learning more about <strong>the</strong> Gene Regulation <strong>of</strong> Hydrolytic enzymeencoding<br />

Genes in an industrial relevant Fungus<br />

Matthias G. Steiger, Marion E. Pucher, Astrid R. Mach-Aigner and Robert L. Mach<br />

Institute <strong>of</strong> Chemical Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: msteiger@mail.tuwien.ac.at<br />

Abstract — Hydrolytic enzymes are needed for<br />

plant biomass degradation in order to make <strong>the</strong>se<br />

renewable resources available for industrial purposes.<br />

The filamentous fungus Trichoderma reesei<br />

has a high secretory capacity <strong>of</strong> those enzymes.<br />

With Xyr1 (Xylanase regulator 1) a very potent<br />

genetic regulator <strong>of</strong> <strong>the</strong> xylanolytic and cellulolytic<br />

enzyme system was found. This study reports <strong>the</strong><br />

action <strong>of</strong> transcription factors, namely Ace1 and<br />

Cre1, on <strong>the</strong> transcription <strong>of</strong> xyr1. Independent <strong>of</strong><br />

<strong>the</strong> carbon source xyr1 transcription is repressed<br />

by Ace1 and subject to carbon catabolite repression<br />

mediated by Cre1.<br />

I. INTRODUCTION<br />

Trichoderma reesei is a filamentous ascomycete<br />

which degrades plant biomass and frequently occurs<br />

in soil. With high effectivity polyssacharides are<br />

metabolized by <strong>the</strong> hydrolytic enzyme system <strong>of</strong><br />

Trichoderma [1]. Due to <strong>the</strong> high secretion capacity<br />

<strong>of</strong> hydrolytic enzymes (up to 100 g/L [2]) this fungus<br />

is <strong>of</strong>ten used in industrial applications in a broad<br />

range <strong>of</strong> industries (e.g. textile, food and feed or<br />

paper industries). Today this fungus is also considered<br />

as a powerful organism for <strong>the</strong> production <strong>of</strong><br />

bi<strong>of</strong>uels (like ethanol) [3] and biochemical building<br />

blocks (e.g. succinate, aspartic acid, itaconic acid,<br />

glycerol, xylitol). Understanding <strong>the</strong> genetic regulation<br />

<strong>of</strong> his fungus is crucial for any fur<strong>the</strong>r targeted<br />

strain improvement. With Xyr1 (Xylanase regulator<br />

1) an important genetic regulator was found [3]. This<br />

regulator is responsible for <strong>the</strong> activation <strong>of</strong> more<br />

than 10 genes involved in xylan and cellulose degradation<br />

and has also a regulatory impact on D-xylose<br />

metabolism. Ace1 ano<strong>the</strong>r transcription factor (Activator<br />

<strong>of</strong> Cellulase expression) has a misleading name<br />

because its impact on cellulase and xylanase expression<br />

was shown to be repressing instead <strong>of</strong> activating<br />

[4]. The consensus binding sequence <strong>of</strong> Ace1 was<br />

found to be AGGCA. Ano<strong>the</strong>r important regulator is<br />

<strong>the</strong> carbon catabolite repressor Cre1 which can be<br />

described as a wide domain repressor <strong>of</strong> particular<br />

hydrolase-encoding genes [5]. Sugars easy to metabolize,<br />

like glucose, repress <strong>the</strong> transcription <strong>of</strong><br />

159<br />

genes which encode enzymes needed for <strong>the</strong> utilization<br />

<strong>of</strong> alternative carbon sources, this process is<br />

known as carbon catabolite repression. Cre1 consensus<br />

binding sequence to DNA is SYGGRG. In this<br />

report we describe <strong>the</strong> influence <strong>of</strong> Ace1 and Cre1 on<br />

<strong>the</strong> transcription <strong>of</strong> Trichoderma reesei xyr1 gene.<br />

II. RESULTS AND DISCUSSION [6]<br />

A. PROMOTER ANALYSIS OF XYR1 GENE<br />

In order to get an idea which transcriptional regulators<br />

could have an influence on <strong>the</strong> transcription <strong>of</strong><br />

xyr1 gene a promoter analysis was carried out. The<br />

genome <strong>of</strong> Trichoderma reesei is sequenced and <strong>the</strong><br />

structure gene <strong>of</strong> xyr1 can be found on scaffold 11<br />

(sc11) in reverse orientation between positions<br />

204723 and 201774. The region 1400 bp (sc11:<br />

201774 - 200374) upstream <strong>of</strong> <strong>the</strong> structural gene<br />

was investigated for binding sites <strong>of</strong> transcription<br />

regulators.<br />

Figure 1: In silico analysis <strong>of</strong> <strong>the</strong> xyr1 promoter<br />

(-1400 to -1 bp upstream <strong>the</strong> structural gene) for<br />

binding sites <strong>of</strong> transcriptional regulators; Cre1<br />

(black vertical bars), Ace1 (white vertical bars)<br />

As can be seen in Figure 1 two potential binding sites<br />

<strong>of</strong> Ace1 and ten <strong>of</strong> Cre1 were found in <strong>the</strong> promoter<br />

region.<br />

B. ACE1 A REPRESSOR OF XYR1 TRANSCRIPTION<br />

To prove <strong>the</strong> influence <strong>of</strong> Ace1 on <strong>the</strong> xyr1 mRNA<br />

formation <strong>the</strong> transcription level <strong>of</strong> xyr1 was studied<br />

in an ace1 �������������������������ace1) compared<br />

to <strong>the</strong> parental strain (Table 1). Results reveal that a<br />

�ace1 strain has always 4 to 6 times higher transcription<br />

levels compared to <strong>the</strong> parental strain. This effect<br />

was found on all carbon sources tested. This fact<br />

provides evidence that Ace1 has a negative influence<br />

on <strong>the</strong> transcriptional regulation <strong>of</strong> xyr1 independent


<strong>of</strong> <strong>the</strong> carbon source. This finding is consistent with<br />

<strong>the</strong> presence <strong>of</strong> two Ace1 binding sites in <strong>the</strong> promoter<br />

sequence <strong>of</strong> xyr1 gene and <strong>the</strong> so far known repressing<br />

action <strong>of</strong> Ace1 on hydrolase expression in<br />

Trichoderma reesei.<br />

carbon source<br />

multiple amounts <strong>of</strong><br />

xyr1 mRNA<br />

Xylose 4,1 ± 0,4<br />

Glucose 6,2 ± 0,4<br />

Sophorose 5,1 ± 0,4<br />

Xylobiose 6,0 ± 0,5<br />

no carbon source 4,0 ± 0,3<br />

Table 1: Increase <strong>of</strong> xyr1 transcription level in a<br />

�ace1 mutant strain compared to parental strain on<br />

various carbon sources<br />

C. TRANSCRIPTION OF XYR1 IS REGULATED VIA<br />

CRE1-MEDIATED CARBON CATABOLITE<br />

REPRESSION<br />

Investigating whe<strong>the</strong>r Cre1 has also an impact on<br />

xyr1 transcription, transcription levels obtained on<br />

glucose, a sugar mediating carbon catabolite repression<br />

(CCR), were compared to <strong>the</strong> levels received on<br />

no carbon source, xylobiose and xylose, two sugars<br />

inducing xylanase formation.<br />

Figure 2: xyr1 transcript formation on no carbon<br />

source (NC), glucose (G), xylose (XO) and xylobiose<br />

(XB)<br />

Results show decreased xyr1 transcription on glucose<br />

compared to inducing substances (Figure 2). To<br />

investigate whe<strong>the</strong>r <strong>the</strong> reduced transcript level <strong>of</strong><br />

xyr1 on glucose is due to Cre1-mediated CCR, we<br />

cultured a Cre1 negative strain (Rut-C30) on <strong>the</strong><br />

same carbon sources as <strong>the</strong> parental strain. A transcript<br />

analysis <strong>of</strong> xyr1 in RutC-30 revealed similar<br />

levels for all conditions tested (data not shown),<br />

indicating a general release from Cre1- dependent<br />

CCR (i.e. de-repression).<br />

160<br />

Summing up, Ace1 acts as a repressor <strong>of</strong> <strong>the</strong> xyr1<br />

transcription. xyr1 is subject to CCR mediated by<br />

Cre1. A Cre1-dependent regulation <strong>of</strong> <strong>the</strong> Xyr1 expression<br />

can now also explain <strong>the</strong> Cre1-related expression<br />

<strong>of</strong> hydrolytic enzyme-encoding genes in<br />

Trichoderma, which do not bear functional Cre1<br />

binding sites in <strong>the</strong>ir own promoters. These aspects<br />

help to gain fur<strong>the</strong>r insight into <strong>the</strong> regulatory enigma<br />

<strong>of</strong> hydrolase production in Trichoderma.<br />

ACKNOWLEDGMENTS<br />

This study was supported by a grant from <strong>the</strong> Austrian<br />

Fonds zur Förderung Wissenschaftlicher Forschung<br />

(P20192-B03) given to R.L. Mach which is<br />

gratefully acknowledged.<br />

REFERENCES<br />

[1] B. Nidetzky, W. Steiner and M. Claeyssens. Cellulose<br />

hydrolysis by <strong>the</strong> cellulases from Trichoderma<br />

reesei: Adsorptions <strong>of</strong> two cellobiohydrolases,<br />

two endocellulases and <strong>the</strong>ir core proteins<br />

on filter paper and <strong>the</strong>ir relation to hydrolysis.<br />

(303): 817-823, November 1994.<br />

[2] J. R. Cherry and A. L. Fidantsef. Directed evolution<br />

<strong>of</strong> industrial enzymes: an update. (14): 438-<br />

443, July 2003.<br />

[3] B. Hahn-Hagerdal, M. Galbe, M. F. Gorwa-<br />

Grauslund, G. Liden and G. Zacchi. Bio-ethanol-<br />

-<strong>the</strong> fuel <strong>of</strong> tomorrow from <strong>the</strong> residues <strong>of</strong> today.<br />

(24): 549-56, December 2006.<br />

[4] A. R. Stricker, K. Grosstessner-Hain, E. Würleitner,<br />

and R. L.Mach. Xyr1 (Xylanase regulator 1)<br />

regulates both <strong>the</strong> hydrolytic enzyme system and<br />

D-xylose metabolism in Hypocrea jecorina. Eukaryot<br />

Cell, 5:2128-2137, December 2006.<br />

[5] N. Aro, M. Ilmen, A. Saloheimo and M. Penttila.<br />

ACEI <strong>of</strong> Trichoderma reesei Is a Repressor <strong>of</strong><br />

Cellulase and Xylanase Expression. (69): 56-65,<br />

January 2003.<br />

[6] R. L. Mach, J. Strauss, S. Zeilinger, M. Schindler<br />

and C. P. Kubicek. Carbon catabolite repression<br />

<strong>of</strong> xylanase I (xyn1) gene expression in Trichoderma<br />

reesei. (21): 1273-1281, July 1996.<br />

[7] A. R. Mach-Aigner, M. E. Pucher, M. G. Steiger,<br />

G. E. Bauer, S. J. Preis and R. L. Mach. Transcriptional<br />

regulation <strong>of</strong> xyr1, encoding <strong>the</strong> main<br />

regulator <strong>of</strong> <strong>the</strong> xylanolytic and cellulolytic enzyme<br />

system in Hypocrea jecorina. AEM.01143-<br />

08, September <strong>2008</strong>.


Investigations <strong>of</strong> Xyr1 (Xylanase regulator 1), <strong>the</strong> main transcriptional<br />

regulator <strong>of</strong> <strong>the</strong> xylanolytic and cellulolytic enzymes in<br />

Hypocrea jecorina (Trichoderma reesei)<br />

Marion E. Pucher, Gudrun E. Bauer, Matthias G. Steiger, Robert L. Mach and Astrid R. Mach-<br />

Aigner<br />

Department <strong>of</strong> Gene Technology and Applied Biochemistry, Institute <strong>of</strong> Chemical Engineering<br />

Vienna University <strong>of</strong> Technology<br />

A-1060 Vienna, Austria<br />

Email: mpucher@mail.zserv.tuwien.ac.at<br />

Abstract — In Trichoderma reesei (teleomorph<br />

Hypocrea jecorina), Xyr1 (Xylanase regulator 1) is<br />

<strong>the</strong> main transcription activator <strong>of</strong> hydrolaseencoding<br />

genes, such as xyn1, xyn2, bxl1, cbh1,<br />

cbh2, egl1, and bgl1. We showed that Xyr1 is permanently<br />

available in <strong>the</strong> cell, and no de novo syn<strong>the</strong>sis<br />

<strong>of</strong> this factor is needed for a first induction <strong>of</strong><br />

xyn1 transcription. The constitutive expression <strong>of</strong><br />

<strong>the</strong> transcription factor leads to clearly elevated<br />

xylanolytic enzyme activities compared to <strong>the</strong> parental<br />

strain QM9414. These hydrolytic enzymes<br />

may be used for <strong>the</strong> efficient hydrolysis <strong>of</strong> plant<br />

biomass waste into monomeric compounds which<br />

have an attractive potential as renewable source <strong>of</strong><br />

energy and basic chemicals.<br />

I. INTRODUCTION<br />

Xylans are heteropolysaccarides with a backbone<br />

��� �-1,4-linked xylopyranosyl units, which compose<br />

20 – 35 % <strong>of</strong> <strong>the</strong> approximate 830 Gt <strong>of</strong> annually<br />

formed renewable plant biomass. Filamentous<br />

ascomycetes <strong>of</strong> <strong>the</strong> genus Trichoderma<br />

mainly act as saprophytes <strong>the</strong>reby degrading a<br />

wide range <strong>of</strong> bio-polymeric substrates. A large<br />

number <strong>of</strong> hydrolytic enzymes secreted by Trichoderma<br />

reesei (teleomorph Hypocrea jecorina)<br />

have received broad industrial interest. Within <strong>the</strong><br />

last 20 years a large number <strong>of</strong> hydrolytic enzymes<br />

have been cloned from T. reesei and o<strong>the</strong>r Trichoderma<br />

spp., but only few <strong>of</strong> <strong>the</strong>m, namely xyn1<br />

and xyn2 (encoding <strong>the</strong> two main Endo-�-1,4xylanases<br />

I and II), have received detailed attention<br />

concerning <strong>the</strong>ir regulation <strong>of</strong> expression (e.g.<br />

[1], [2]). Whereas in Aspergillus <strong>the</strong> xylanolytic<br />

and cellulolytic system is strictly co-regulated via<br />

<strong>the</strong> inducer D-xylose (e.g. [3]), <strong>the</strong> hydrolytic<br />

enzymes <strong>of</strong> T. reesei are not. Their differential<br />

expression has been reported in several studies<br />

(e.g. [4]). In contrast to <strong>the</strong> diversity <strong>of</strong> inducers<br />

and induction mechanisms, it was recently shown<br />

that transcriptional regulation <strong>of</strong> <strong>the</strong> major hydro-<br />

161<br />

lytic enzyme-encoding genes XYNI and XYNII,<br />

as well as cellobiohydrolases CBHI, CBHII, and<br />

EGLI are strictly dependent on Xyr1 (Xylanase<br />

regulator 1). Fur<strong>the</strong>rmore, it was demonstrated that<br />

Xyr1 plays an essential role in <strong>the</strong> regulation <strong>of</strong><br />

<strong>the</strong> D-xylose reductase expression. Altoge<strong>the</strong>r, it<br />

can be summarized that Xyr1 is <strong>the</strong> main transcription<br />

activator <strong>of</strong> <strong>the</strong> hydrolase system in T.<br />

reesei [5].<br />

II. RESULTS [7]<br />

A. NO DE NOVO SYNTHESIS OF XYR1 IS NEEDED<br />

FOR A FIRST INDUCTION OF XYN1<br />

We questioned if de novo syn<strong>the</strong>sis <strong>of</strong> Xyr1 is<br />

necessary for an initial induction <strong>of</strong> xyn1 transcription.<br />

To answer this, <strong>the</strong> parental strain was replaced<br />

to D-xylose or glucose, with or without <strong>the</strong><br />

translation inhibitor cycloheximide. Transcriptional<br />

analysis showed that xyn1 transcription<br />

levels are not significantly altered, independent <strong>of</strong><br />

cycloheximide supplementation. This is remarkable<br />

because xyn1 transcription, which is strictly<br />

dependent on Xyr1, is induced although <strong>the</strong> addition<br />

<strong>of</strong> a translation inhibitor prevents de novo<br />

syn<strong>the</strong>sis <strong>of</strong> Xyr1. We assume that a low constitutive<br />

level <strong>of</strong> Xyr1 is present in <strong>the</strong> cell under all<br />

growth conditions.<br />

B. CONSTITUTIVE ACTIVATION OF XYR1<br />

To investigate <strong>the</strong> effect <strong>of</strong> a constitutive activated<br />

Xylanase regulator 1, <strong>the</strong> structural gene xyr1 (Gen-<br />

Bank accession no. AF479644) was put under <strong>the</strong><br />

control <strong>of</strong> <strong>the</strong> nag1 promotor <strong>of</strong> H. atroviridis. This<br />

nag1::xyr1 fusion and <strong>the</strong> vector pAN7 [6], conferring<br />

hygromycin B resistance, were transformed into<br />

<strong>the</strong> genome <strong>of</strong> H. jecorina �xyr1 strain [5]. The<br />

resulting nx transformant strains showed similar<br />

growth rates on malt extract medium compared to <strong>the</strong><br />

QM9414 pa������� ������� ��� ���� �xyr1 strain. The


conidiospores are white to brownish instead <strong>of</strong> dark<br />

green and conidiation is slightly delayed in <strong>the</strong> nx<br />

strains compared with <strong>the</strong> parental strain.<br />

C. MODIFIED XYLANASE EXPRESSION DUE TO<br />

THE CONSTITUTIVE ACTIVATED XYR1<br />

The parental and <strong>the</strong> nx strain were cultivated in<br />

medium containing xylan. We observed endo-�-1,4xylanase<br />

activity at an earlier culture stage (Fig. 1A)<br />

���� �-xylosidase activity was clearly elevated (Fig.<br />

1B) in <strong>the</strong> nx transformant strain. NaOH-soluble<br />

protein measurements showed that growth characteristics<br />

slightly differed in <strong>the</strong> strains tested (Fig. 1C),<br />

most likely due to <strong>the</strong> delayed xylanase production in<br />

<strong>the</strong> parental strain. Complete clearing (i.e., enzymatic<br />

degradation <strong>of</strong> <strong>the</strong> insoluble xylan compounds) <strong>of</strong> <strong>the</strong><br />

xylan-containing medium by <strong>the</strong> nx transformant<br />

strain was visible after approximately 50 h <strong>of</strong> cultivation,<br />

whereas <strong>the</strong> parental strain could not clear <strong>the</strong><br />

medium during <strong>the</strong> duration <strong>of</strong> <strong>the</strong> experiment.<br />

Figure 1: Analysis <strong>of</strong> xylanase enzyme activity<br />

formation. Endo-�-���� ���������� ���� ���� �xylosidase<br />

(B) activities in culture supernatants <strong>of</strong><br />

<strong>the</strong> QM9414 parental strain (grey diamonds) and<br />

<strong>the</strong> nx mutant strain (black squares) (A to C). Detection<br />

<strong>of</strong> enzyme activity formation in both<br />

strains was accomplished after 0, 12, 24, 36, 48,<br />

60, and 72 h <strong>of</strong> cultivation in shake flasks on 1%<br />

(wt/vol) xylan after inoculation with spores. One<br />

unit <strong>of</strong> activity is defined as <strong>the</strong> amount <strong>of</strong> enzyme<br />

required to release 1 micromole <strong>of</strong> xylose reducing<br />

sugar equivalents or 4-nitrophenyl residues per<br />

minute under <strong>the</strong> defined assay conditions. (C)<br />

162<br />

Determination <strong>of</strong> biomass formation in medium<br />

containing insoluble xylan by measurement <strong>of</strong><br />

NaOH-soluble protein. All data are means <strong>of</strong> results<br />

from three independent biological replicates.<br />

Error bars indicated standard deviations.<br />

ACKNOWLEDGMENTS<br />

This study was supported by a grant from <strong>the</strong> Austrian<br />

Fond zur Förderung Wissenschaftlicher Forschung<br />

(P 20192-B03) given to R.L.M., which is<br />

gratefully acknowledged. I want to thank Astrid R.<br />

Mach-Aigner and Robert L. Mach for supporting my<br />

work.<br />

REFERENCES<br />

[1] R. L. Mach, J. Strauss, S. Zeilinger, M. Schindler,<br />

C. P. Kubicek. Carbon catabolite repression <strong>of</strong><br />

xylanase I (xyn1) gene expression in Trichoderma<br />

reesei. Mol Microbiol, 21(6):1273-1281,<br />

September 1996.<br />

[2] E. Würleitner, L. Pera, C. Wacenovsky, A. Czifersky,<br />

S. Zeilinger, C. P. Kubicek, and R. L.<br />

Mach. Transcriptional regulation <strong>of</strong> xyn2 in Hypocrea<br />

jecorina. Eukaryot Cell, 2(1):150-158,<br />

February 2003.<br />

[3] M. M. Gielkens, E. Dekkers, J. Visser, and L. H.<br />

de Graaff. Two cellobiohydrolase-encoding<br />

genes from Aspergillus niger require D-xylose<br />

and <strong>the</strong> xylanolytic transcriptional activator XlnR<br />

for <strong>the</strong>ir expression. Appl Environ Microbiol,<br />

65(10):4340-4345, October 1999.<br />

[4] S. Zeilinger, R. L. Mach, M. Schindler, P. Herzog<br />

and C. P. Kubicek. Different inducibility <strong>of</strong> expression<br />

<strong>of</strong> <strong>the</strong> two xylanase genes xyn1 and<br />

xyn2 in Trichoderma reesei. J Biol Chem,<br />

271(41):25624-25629, October 1996.<br />

[5] A. R. Stricker, K. Grosstessner-Hain, E. Würleitner,<br />

and R. L.Mach. Xyr1 (Xylanase regulator 1<br />

regulates both <strong>the</strong> hydrolytic enzyme system and<br />

D-xylose metabolism in Hypocrea jecorina. Eukaryot<br />

Cell, 5(12):2128-2137, December 2006.<br />

[6] P. J. Punt, R. P. Oliver, M.A. Dingemanse, P. H.<br />

Pouwels, and C. A. van den Hondel. Transformation<br />

<strong>of</strong> Aspergillus based on <strong>the</strong> hygromycin B<br />

resistance marker from Escherichia coli. Gene,<br />

56(1):117-124, May 1987.<br />

[7] A. R. Mach-Aigner, M. E. Pucher, M. G. Steiger,<br />

G. E. Bauer, S. J. Preis, and R. L. Mach. Transcriptional<br />

regulation <strong>of</strong> xyr1, encoding <strong>the</strong> main<br />

regulator <strong>of</strong> <strong>the</strong> xylanolytic and cellulolytic enzyme<br />

system in Hypocrea jecorina. Appl Environ<br />

Microbiol, 74(21):AEM.01143-08v1, November<br />

<strong>2008</strong>.


Using Confocal Microscopy to Study Mechanotransduction<br />

in Renal Epi<strong>the</strong>lial Cells<br />

Yi Duan and Sheldon Weinbaum<br />

Department <strong>of</strong> Biomedical Engineering<br />

The City College and Graduate Center <strong>of</strong> <strong>the</strong> City University <strong>of</strong> New York<br />

New York, United States<br />

Email: yduan@ccny.cuny.edu<br />

Abstract — The study <strong>of</strong> mechanotransduction in<br />

renal epi<strong>the</strong>lial cells (REC) is critical in understanding<br />

<strong>the</strong> coupling between mechanical forces<br />

and biochemical reactions. In <strong>the</strong> current study, we<br />

used confocal microscopy to study <strong>the</strong> dynamic<br />

cytoskeletal responses <strong>of</strong> RECs to a laminar fluid<br />

shear stress (FSS). Confluent RECs were exposed to<br />

a FSS <strong>of</strong> 1 dyne/cm 2 for 5h. Immun<strong>of</strong>luorescence<br />

images demonstrated a reinforcement <strong>of</strong> peripheral<br />

actin networks, a disappearance <strong>of</strong> stress fibers,<br />

and an enhanced linearly distributed ZO-1. In addition,<br />

FSS caused an accumulation <strong>of</strong> paxillin at <strong>the</strong><br />

cell base. Our results showed that FSS-induced<br />

cytoskeletal reorganization <strong>of</strong> RECs is, surprisingly,<br />

exactly opposite that observed in endo<strong>the</strong>lial cells.<br />

I. INTRODUCTION<br />

The most important unresolved mystery in kidney<br />

proximal tubule is that <strong>the</strong> fractional reabsorption <strong>of</strong><br />

water is reliably at 65% despite <strong>the</strong> large variation<br />

<strong>of</strong> <strong>the</strong> glomerular filtration rate. Guo et al. [1] proposed<br />

that <strong>the</strong> brush border microvilli, which are<br />

composed <strong>of</strong> a bundle <strong>of</strong> actin filaments, serve a<br />

mechanosensory function and proposed that <strong>the</strong><br />

torque (bending moment) on <strong>the</strong> microvilli modulated<br />

transport activities. Clearly this hypo<strong>the</strong>sis<br />

demonstrated that an intact actin cytoskeleton was<br />

essential to this transport activity. However, in<br />

marked contrast to endo<strong>the</strong>lial cells (EC), <strong>the</strong>re are<br />

very few studies on cultured epi<strong>the</strong>lial cells in<br />

which cytoskeletal reorganization has been examined<br />

in response to laminar FSS under carefully<br />

monitored flow conditions. In <strong>the</strong> present study, we<br />

exposed REC to defined laminar FSS for 5hs and<br />

investigated <strong>the</strong> structural responses <strong>of</strong> RECs to<br />

FSS by examining <strong>the</strong> localization <strong>of</strong> F-actin (cytoskeleton),<br />

tight junction (TJ) protein ZO-1 and<br />

focal adhesion (FA) protein paxillin.<br />

II. EXPERIMENTAL METHODS<br />

Cell cultures and flow experiments: Mouse RECs<br />

were grown to confluence on glass coverslips for 5<br />

163<br />

days and were exposed to a FSS <strong>of</strong> 1 dyne/cm 2 for<br />

5h. The system was kept in 37 o C.<br />

Immun<strong>of</strong>luorescence and confocal microscopy:<br />

After exposed to FSS, cells were stained with <strong>the</strong><br />

following monoclonal antibodies: 1) ZO-1, 2) Ecadherin,<br />

and 3) paxillin. F-actin was labelled by<br />

phalloidin. Fluorescence images were captured using<br />

a confocal microscope (Zeiss, LSM510).<br />

III. KEY RESULTS<br />

Cytoskeleton Reorganization: To study actin cytoskeleton<br />

reorganization after application <strong>of</strong> FSS,<br />

distribution <strong>of</strong> F-actin stress fibers was examined<br />

in REC and compared with no-flow controls. We<br />

witnessed a marked change <strong>of</strong> F-actin organization.<br />

In control, numerous long, thick cytosolic<br />

stress fibers were found regularly at <strong>the</strong> basal aspect<br />

<strong>of</strong> REC (Figure 1C) and a relatively thin<br />

circumferential actin network at cell-cell contacts<br />

at <strong>the</strong> basal surface (Figure 1A). Gaps were <strong>of</strong>ten<br />

found in between <strong>the</strong> cells. Exposure <strong>of</strong> RECs to<br />

5h FSS causes dramatically diminished stress fibers<br />

at <strong>the</strong> basal surface (Figure 1D) and a reinforcement<br />

<strong>of</strong> actin filaments at peripheral cell borders<br />

near <strong>the</strong> apical surface (Figure 1B).<br />

TJ: To investigate FSS-stimulated TJ distribution in<br />

epi<strong>the</strong>lial cells, TJ-associated protein ZO-1 was<br />

analyzed. In static control, ZO-1 does not even<br />

form a cell-cell contact structure but appears as<br />

isolated rings surrounding individual cells (Figure<br />

2A), though <strong>the</strong>se cells are cultured until fully confluent.<br />

A continuous distribution with dramatically<br />

reinforced junction staining pattern <strong>of</strong> ZO-1 was<br />

observed after exposure to 5h FSS (Figure 2B).<br />

FAs: Paxillin, an adaptor protein at FAs was examined<br />

to study <strong>the</strong> effect <strong>of</strong> FSS in <strong>the</strong> modulation <strong>of</strong><br />

FAs. Under static conditions, paxillin appears ei<strong>the</strong>r<br />

weak or absent in <strong>the</strong> basement membrane<br />

(Figure 3A). After 5h FSS, a predominant upregulation<br />

<strong>of</strong> paxillin was found at both cell periphery<br />

and inside <strong>of</strong> <strong>the</strong> cells (Figure 3B).


Apical<br />

Basal<br />

CTL<br />

A B<br />

C<br />

Figure 1: Confocal images <strong>of</strong> F-actin showing actin<br />

reorganization <strong>of</strong> REC exposed to 5h FSS. CTL,<br />

control; FSS, fluid shear stress. Bar, 10μ.<br />

Figure 2: Reorganization <strong>of</strong> TJs in response to FSS.<br />

CTL, control; FSS, fluid shear stress. Bar, 10μ.<br />

Figure 3. Confocal fluorescence images for antipaxillin<br />

binding before (A) and after (B) 5h FSS.<br />

CTL, control; FSS, fluid shear stress. Bar, 10μ.<br />

IV. DISCUSSION<br />

FSS<br />

The experiments in this paper have led to two important<br />

and unexpected observations. First, <strong>the</strong><br />

cytoskeletal reorganization in response to FSS in<br />

cultured confluent epi<strong>the</strong>lial cells is nearly diametrically<br />

opposite to that observed in [2] for ECs<br />

which were also subjected to FSS for 5h. In <strong>the</strong><br />

D<br />

CTL FSS<br />

A B<br />

CTL FSS<br />

A B<br />

164<br />

latter paper, provided an intact glycocalyx is present,<br />

<strong>the</strong>re was a dense peripheral actin band, few<br />

basal stress fibers, a dispersed distribution <strong>of</strong> vinculin<br />

and clearly defined continuous TJs in <strong>the</strong> control<br />

state as indicated by <strong>the</strong> localization <strong>of</strong> ZO-1. This<br />

organization was greatly altered after FSS with a<br />

breakdown <strong>of</strong> peripheral actin bands, a formation <strong>of</strong><br />

stress fibers, a movement <strong>of</strong> vinculin to cell borders<br />

and a disruption <strong>of</strong> TJs. These observations are just<br />

opposite that observed in <strong>the</strong> present study. Second,<br />

culture RECs, though confluent, do not form TJ in<br />

<strong>the</strong>ir control state. Quite remarkably, a FSS <strong>of</strong> only<br />

1 dyne/cm 2 appears to be a critical precondition for<br />

<strong>the</strong> formation <strong>of</strong> TJs. This FSS is only one tenth that<br />

applied in <strong>the</strong> EC studies <strong>of</strong> [2].<br />

Why do confluent EC and REC respond in such a<br />

different manner to FSS? The most important clue<br />

is <strong>the</strong> fact that, <strong>the</strong> cells, though, confluent and<br />

touching at <strong>the</strong>ir base, do not have TJ as indicated<br />

by <strong>the</strong> distribution <strong>of</strong> ZO-1. There is a strong expression<br />

<strong>of</strong> stress fibers at <strong>the</strong> base <strong>of</strong> <strong>the</strong> cell, as<br />

indicated by Figure 1B, which causes a firm adhesion<br />

<strong>of</strong> <strong>the</strong> cell to its substrate. This creates a tension<br />

in <strong>the</strong> cell membrane, due to <strong>the</strong> compressive<br />

resistance <strong>of</strong> <strong>the</strong> internal cytoskeleton, which in turn<br />

produces a rounding <strong>of</strong> <strong>the</strong> apical surface and a<br />

pulling away <strong>of</strong> a cell from its neighbor at its basolateral<br />

surface. The cells have <strong>the</strong> appearance <strong>of</strong> tall<br />

columnar domes. Cell junctions can not form until<br />

<strong>the</strong>re is a disruption <strong>of</strong> <strong>the</strong> stress fibers at <strong>the</strong> basal<br />

surface and a release <strong>of</strong> this membrane tension.<br />

Even a small FSS will cause <strong>the</strong>se tall columnar<br />

cells to tilt and <strong>the</strong>ir basolateral surfaces to come in<br />

contact. This also causes a release <strong>of</strong> actin stress<br />

fibers at <strong>the</strong> basal surface and <strong>the</strong> formation <strong>of</strong> a<br />

DAPB as part <strong>of</strong> AJ assembly as shown in Figure 1.<br />

The present study demonstrated for <strong>the</strong> first time<br />

<strong>the</strong> importance <strong>of</strong> FSS in <strong>the</strong> formation <strong>of</strong> TJ in<br />

RECs in comparison with <strong>the</strong> cytoskeleton reorganization<br />

<strong>of</strong> ECs under <strong>the</strong> same duration <strong>of</strong> flow.<br />

Whe<strong>the</strong>r this dramatic structural change participates<br />

in <strong>the</strong> process <strong>of</strong> FSS-induced transporter activation<br />

or translocation will be examined in <strong>the</strong> future.<br />

REFERENCES<br />

[1] Guo P, Weinstein AM, Weinbaum S. A hydrodynamic<br />

mechanosensory hypo<strong>the</strong>sis for brush border<br />

microvilli. Am J. Physiol Renal Physiol.<br />

279(4): F698-712, 2000.<br />

[2] Thi MM, Tarbell JM, Weinbaum S, Spray DC.<br />

The role <strong>of</strong> glycocalyx in reorganization <strong>of</strong> <strong>the</strong><br />

actin cytoskeleton under fluid shear stress: a<br />

“bumper-car” model. Proc. Natl Acad Sci USA.<br />

101 (47): 16483-8, 2004.


Low-energy School in Indian Himalayas<br />

Jan Tilinger (mentor: Ing. Frantisek Kulhanek, CSc. - kulhanek@fsv.cvut.cz)<br />

Faculty <strong>of</strong> Civil Engineering<br />

Czech Technical University in Prague, www.cvut.cz<br />

Czech Republic<br />

Email: jan.tilinger@suryaschool.org<br />

www.suryaschool.org<br />

Abstract — We have build low-energy solar school<br />

in village Kargyak in nor<strong>the</strong>rn part <strong>of</strong> Indian Himalayas<br />

as a main part <strong>of</strong> Project Surya. Construction<br />

was made by local materials and with use <strong>of</strong> traditional<br />

building techniques with respect to harmonious<br />

proportions and adaptation to <strong>the</strong> landscape<br />

typical for local structures. The project responds to<br />

<strong>the</strong> existing socio-demographic situation in this<br />

remote region. Literacy is far below average in<br />

India; only few local people out <strong>of</strong> hundreds can<br />

read and write. The current situation leads to migration<br />

out <strong>of</strong> <strong>the</strong> region which dramatically disrupts<br />

not only <strong>the</strong> local culture but also traditional<br />

family relations.<br />

Village Kargyak is located in Zanskar mountain<br />

range at altitude <strong>of</strong> 4200 metres; it is a village in <strong>the</strong><br />

highest location within this very isolated region.<br />

Kargyak is located 80 kilometres (3 days walk) from<br />

Padum, administrative centre <strong>of</strong> <strong>the</strong> region, where <strong>the</strong><br />

regular road ends.<br />

Climatic conditions are quite uncommon in<br />

Kargyak. Village is located in rain shadow <strong>of</strong> nearby<br />

mountains in extremely dry climate with substantial<br />

proportion <strong>of</strong> sunshine – more than 300 days a year<br />

are classified as sunny. The landscape has character<br />

<strong>of</strong> mountain desert: ridges, rubble and sand mountainsides;<br />

sporadic green vegetation is only alongside<br />

water courses or on artificially irrigated fields. There<br />

was not any systematic monitoring <strong>of</strong> <strong>the</strong> local climate<br />

carried out yet (project carries out limited<br />

measurements).<br />

The building is designed in compliance<br />

with <strong>the</strong> local architectonic principles, with emphasis<br />

on usage <strong>of</strong> local natural materials and traditional<br />

building technologies as part <strong>of</strong> doctoral research on<br />

CTU. Detail research <strong>of</strong> local procedures, material<br />

properties and traditional architecture was done. In<br />

<strong>the</strong> region is still tradition <strong>of</strong> making and using clay<br />

bricks and o<strong>the</strong>r natural materials like stone, poplar<br />

and willow wood, low bushes, straw and yak, sheep<br />

and goat dunk. O<strong>the</strong>r imported materials like glass<br />

and timber species for assembly construction are also<br />

used for <strong>the</strong> construction.<br />

165<br />

Due to <strong>the</strong> use <strong>of</strong> solar energy will be possible<br />

to teach in <strong>the</strong> school almost all year round which<br />

would not be possible o<strong>the</strong>rwise. In summer 2006<br />

realization team made an agreement with Kargyak<br />

village, chosen in previous years. Cooperation was<br />

based on <strong>the</strong> fact that <strong>the</strong> village itself as well as<br />

neighbouring villages lacked school education and<br />

only few children had <strong>the</strong> opportunity to study in<br />

boarding schools in lowlands. In autumn 2006 was<br />

built a greenhouse in order to test cooperation with<br />

<strong>the</strong> villagers and also in order to verify <strong>the</strong> possibility<br />

<strong>of</strong> <strong>the</strong> use <strong>of</strong> solar energy, research <strong>of</strong> quality <strong>of</strong> local<br />

building materials and building technologies and last<br />

but not least in order to obtain fresh vegetable during<br />

Zanskar winter. Data logger was placed in <strong>the</strong> greenhouse<br />

to record inside and outside temperature as<br />

well as air humidity. The temperature in <strong>the</strong> greenhouse<br />

reached 20��� ������� ����� ��� ���� ������������<br />

on average and vegetable was grown in <strong>the</strong> greenhouse<br />

during all winter season, even though <strong>the</strong> temperature<br />

in surroundings was up to 27���������������<br />

The lowest air temperature measured in <strong>the</strong> greenhouse<br />

was only 3.6��������������<br />

In February 2007 realisation team travelled<br />

to <strong>the</strong> construction site to ensure delivery <strong>of</strong> timber<br />

on frozen river from village <strong>of</strong> Ichar to Kargyak as<br />

well as to document functionality <strong>of</strong> <strong>the</strong> greenhouse,<br />

assemble temperature. In spring 2007 continued<br />

transport <strong>of</strong> building tools, timber and provisions for<br />

workers living in <strong>the</strong> village. All <strong>the</strong> work was due to<br />

inaccessibility <strong>of</strong> any machinery done manually or by<br />

simple hand tools. In spring and during summer 2007<br />

were dug trenches in <strong>the</strong> volume <strong>of</strong> 1600 cubic metres<br />

<strong>of</strong> stony soil, built three abutment walls <strong>of</strong> total<br />

length 120 metres and done o<strong>the</strong>r groundwork.<br />

Foundations were surveyed, dug and lined and on <strong>the</strong><br />

edge <strong>of</strong> <strong>the</strong> site was built a toilet for workers.<br />

During <strong>the</strong> summer 2007 production <strong>of</strong> clay<br />

bricks sized 20x20x40 centimetres was optimized.<br />

Intensity <strong>of</strong> labour on clay bricks production: four<br />

workers produce about 50 clay bricks a day/ 28 kilos<br />

each brick. Bricks are dried at <strong>the</strong> place <strong>of</strong> production,<br />

three weeks later piled into pyramids and a<br />

month after <strong>the</strong> production transported on a horse or


yak back approximately 1500 metres to <strong>the</strong> building<br />

site. Samples <strong>of</strong> clay bricks were at <strong>the</strong> end <strong>of</strong> year<br />

2007 tested in <strong>the</strong> laboratories <strong>of</strong> Faculty <strong>of</strong> Civil<br />

Engineering in Prague; obtained basic engineering<br />

characteristics <strong>of</strong> <strong>the</strong> bricks are: heat conductivity<br />

coefficient - � = 0,317 W/mK, specific <strong>the</strong>rmal<br />

capacitance - cp = 1,39 106 J/m3K, specific gravity -<br />

� = 2054.32 kg/m3, dynamic modulus <strong>of</strong> elasticity -<br />

E = 1.37 GPa.<br />

Volume solution <strong>of</strong> <strong>the</strong> school building is<br />

based on <strong>the</strong> shape <strong>of</strong> bevelled ashlar, partially embedded<br />

in <strong>the</strong> ground. Faced wall which is substantially<br />

glassed-in south-facing (azimuth 195�� ����� ����<br />

west from south) allows direct use <strong>of</strong> solar energy.<br />

Ground floor <strong>of</strong> <strong>the</strong> house is centrally symmetric<br />

with <strong>the</strong> entrance area in west-facing wall.<br />

The school building consists <strong>of</strong> three classrooms,<br />

corridor, lobby and two toilets. Strip foundations <strong>of</strong><br />

<strong>the</strong> building are made <strong>of</strong> shaped stone. Vertical<br />

framework consists <strong>of</strong> timber poles and stone wall.<br />

Cladding is filled with adobe bricks. Thermal insulation<br />

is ensured by a layer <strong>of</strong> straw between <strong>the</strong> stone<br />

and adobe walls, which runs around <strong>the</strong> building<br />

periphery.<br />

Figure 1: Section<br />

The ro<strong>of</strong> structure was made entirely according<br />

to local traditions and its construction is<br />

based on low incidence <strong>of</strong> rainfall in <strong>the</strong> area. Horizontal<br />

supporting structure is made <strong>of</strong> poplar wood,<br />

heat insulation layer is made <strong>of</strong> straw mixed with yak<br />

dirt and branches tabulated with soil and covered<br />

with clay (hydro insulation layer) with mound <strong>of</strong><br />

sandy soil.<br />

Because <strong>the</strong> characteristics <strong>of</strong> indoor climate<br />

are determined by <strong>the</strong> local customs <strong>the</strong> anticipated<br />

use <strong>of</strong> <strong>the</strong> building is non-standard (by European<br />

norms); as a result it is difficult to carry out<br />

classification <strong>of</strong> energy intensity <strong>of</strong> <strong>the</strong> building<br />

according to established European conventions. It is<br />

assumed that major part <strong>of</strong> <strong>the</strong> heat loss will be cov-<br />

166<br />

ered by solar acquisition and heat acquisition from<br />

indoor sources. The main effort is to use <strong>the</strong> solar<br />

energy at <strong>the</strong> most: air heated by <strong>the</strong> sun will be<br />

distributed around <strong>the</strong> building by especially designed<br />

system <strong>of</strong> air circulation.<br />

Project Surya will follow in long term observations<br />

<strong>of</strong> indoor environment characteristics in<br />

<strong>the</strong> school building and observation <strong>of</strong> efficiency <strong>of</strong><br />

solar system.<br />

ACKNOWLEDGMENTS<br />

The school construction was supported by donation<br />

<strong>of</strong> private donors. The project was run by Czech<br />

Civic Association Surya and <strong>the</strong> construction part <strong>of</strong><br />

<strong>the</strong> project was finished in September <strong>2008</strong> with<br />

great help <strong>of</strong> volunteers from more <strong>the</strong>n 20 different<br />

countries.<br />

REFERENCES<br />

[1] ŠUBRT, R., TILINGER, J.: Pasivní stavba<br />

školy v Himalájích, Energie kolem nás,<br />

6/2006<br />

[2] TILINGER, J.: Solární škola v Kargyaku,<br />

������������������������������������<br />

[3] TILINGER, J., KULHÁNEK, F.:<br />

Nízkoenergetická solární škola v indickém<br />

Himaláji, In: Tepelná ochrana budov 2006,<br />

sborník mezinárodní konference, Praha 2006<br />

[4] Stavebníctvo a bývanie, pr��������������������<br />

��������������������������������������<br />

2279/2000, vydává MEDIA/ST, s.r.o., Žilina<br />

�����������������������������������������<br />

Pasívna stavba školy v ������������������������<br />

80, text+foto: Ing. Roman Šubrt, Ing. Jan<br />

Tilinger<br />

[5] Stavebníctvo a b�����������������������������<br />

146 stran, ��������������������������������<br />

2279/2000, vydává MEDIA/ST, s.r.o., Žilina<br />

�����������������������������������������<br />

Pasívna stavba školy v ������������������������<br />

82, 83, text: Ing. Roman Šubrt, foto Daniel<br />

Šubrt<br />

[6] TILINGER, J.: Realizace nízkoenergetické<br />

solární školy v indickém Himaláji, Tepelná<br />

ochrana budov 2007, sborník mezinárodní<br />

konference, Podbánské 2007<br />

[7] �����������������-����������������– ERA -<br />

Brno 2002 - ISBN:80-86517-21-7<br />

[8] Minke, Gernot - Earth Construction<br />

Handbook: The Building Material Earth in<br />

Modern Architecture - 2000 -<br />

80.237.211.43


ALGORITHM OF SIMULATION AND OPTIMALIZATION OF<br />

CONVENTIONAL MECHANICAL PRODUCTION<br />

Robert Jurcisin, and Dusan Sebo (Faculty Mentor)<br />

Faculty <strong>of</strong> Mechanical Engineering<br />

Technical University in Kosice<br />

Kosice, Slovakia<br />

Email: {robert.jurcisin,dusan.sebo}@tuke.sk<br />

Abstract — The main idea <strong>of</strong> <strong>the</strong> article is <strong>the</strong><br />

assignment <strong>of</strong> components to machines by<br />

ma<strong>the</strong>matical algorithm. Algorithm is able to<br />

calculate each queue which is arranged to<br />

production and calculate continuous period <strong>of</strong><br />

production <strong>of</strong> each component. Mentioned<br />

algorithm is able to determine structure <strong>of</strong><br />

machining time needs for fulfilled order according<br />

to individual components and machines or separately<br />

for each component. There are also o<strong>the</strong>r<br />

given information for far<strong>the</strong>r technical - economic<br />

analyses in global aspect.<br />

I. INTRODUCTION<br />

Following article deals with problems <strong>of</strong> engineering<br />

and production scheduling in conventional machineindustry<br />

production. For filling <strong>of</strong> production was<br />

used method FIFO (First in First Out) which is suitable<br />

for conventional machine-industry production.<br />

In this algorithm, we were looking for an optimal<br />

solution with <strong>the</strong> shortest technological times <strong>of</strong><br />

production. This <strong>the</strong>sis has also wide possibilities to<br />

use in technical-economic analysis and also pricing<br />

<strong>of</strong> machinery products, technical preparation <strong>of</strong><br />

production using expert methods and so on.<br />

II. SIMULATION AND OPTIMALIZATION<br />

OF CHARGING OF CONVENTIONAL<br />

MECHANICAL PRODUCTION<br />

The entrance for simulation algorithm is a matrix <strong>of</strong><br />

assignment tij, which sets how much time is needed<br />

for proceeding <strong>the</strong> operations <strong>of</strong> i –part <strong>of</strong> component<br />

on j – part <strong>of</strong> machine.<br />

The second entrance for initialization formula <strong>of</strong><br />

algorithm is <strong>the</strong> matrix iBiv, which determines how<br />

many components <strong>of</strong> <strong>the</strong> kind <strong>of</strong> i – part is in v –<br />

part <strong>of</strong> product (number <strong>of</strong> components is taken over<br />

from bill <strong>of</strong> material <strong>of</strong> composition drawing <strong>of</strong><br />

product).<br />

The third entrance is matrix uBuv (number <strong>of</strong> components<br />

<strong>of</strong> product A1....v, which is necessary to<br />

produce for <strong>the</strong> certain order DvQ).<br />

167<br />

We will still follow interval <strong>of</strong> time jump <strong>of</strong> finishing<br />

work Tk find from initialization formula,<br />

where for minus item hold <strong>the</strong> rule <strong>of</strong> counting <strong>of</strong>f<br />

for ƒ k ksi VZj, if in <strong>the</strong> last step <strong>the</strong> component<br />

was on <strong>the</strong> same machine.<br />

Production system in which 5 production equipments<br />

are situated has a right to manage and charge production<br />

by computer even if its processes are simply<br />

controlled and dispatching control does not exceed<br />

human abilities.<br />

It would be more complicated if <strong>the</strong>re were tenths <strong>of</strong><br />

machines and equipments and thousands <strong>of</strong> components<br />

in production system and conventional small<br />

batch production for which mentioned algorithm is<br />

suitable for.<br />

III. CASE STUDY<br />

We will illustrate <strong>the</strong> case study <strong>of</strong> production<br />

scheduling in conventional mechanical production on<br />

virtual company in which <strong>the</strong>re are situated four<br />

following machines:<br />

VZ1: Centre la<strong>the</strong> SV 18RA<br />

VZ2: Table milling FC 50V with system NS 350<br />

VZ3: Centre la<strong>the</strong> SU 80A<br />

VZ4: Grinding machine BP1/600<br />

Figure 2: Disposition <strong>of</strong> production system


We are concerned with conventional mechanical<br />

production in which machines do not have NC or<br />

CNC control but are controlled by human factor.<br />

Ma<strong>the</strong>matically we try to find optimal solution with<br />

<strong>the</strong> shortest time <strong>of</strong> production. According to<br />

combinatory analysis we have n! possibilities for<br />

choosing <strong>the</strong> order <strong>of</strong> components. In our case we<br />

have exactly 120 possibilities. Therefore we showed<br />

<strong>the</strong> principle <strong>of</strong> algorithm only for illustration on<br />

three examples.<br />

Si/VZj Vz1 Vz2 Vz3 Vz4 A1 A2<br />

S1 38,6 6,9 3,6 3,8 16 16<br />

S2 77,6 16,3 8,3 14,3 8 8<br />

S3 27,5 5,3 - - 20 20<br />

S4 10,3 - 3,4 5,3 70 70<br />

S5 6,7 - 5,7 - 10 10<br />

Q1=5 Q2=5<br />

Table 1: Table with input data<br />

IV. EXAMPLE OF CALCULATIONS<br />

T0 = 0<br />

F1 = S2, S3 ,S4,S5<br />

Vz1 = S1 T1S1 = 38,6. (16.5+16.5) = 6 176<br />

F2 = 0 Vz2 = 0<br />

F3 = 0 Vz3 = 0<br />

F4 = 0 Vz4 = 0<br />

T1 = 6 176<br />

F1 = S3, S4, S5 Vz1 = S2 T2S2 = 77,6. (8.5+8.5) = 6 208<br />

F2 = 0 Vz2 = S1 T2S1 = 6, 9. (16.5+16.5) = 1 104<br />

F3 = 0 Vz3 = 0<br />

F4 = 0 Vz4 = 0<br />

T2 = 1 104<br />

F1 = S3, S4, S5 Vz1 = S2 T3S2 = 6 208 – 1 104 = 5 104<br />

F2 = 0 Vz2 = 0<br />

F3 = 0 Vz3 = S1 T3S1 = 3, 6. (16.5+16.5) = 347, 4<br />

F4 = 0 Vz4 = 0<br />

…..,<br />

∑T = T1+T2+T3+T4+....+T15 = 6 176 + 1 104 + 347, 4 +<br />

608 + 4 148, 6 + 1 304 + 664 + 1 144 + 2 388 + 1 060 +<br />

6 150 + 670 + 1 710 + 570 + 3140 = 31 184 norm<br />

minutes<br />

Total machine time by setting <strong>of</strong> components from<br />

S1 – S5 is 31 184 norm minutes, what equally<br />

rounded is 520 norm hours.<br />

1.Example 31 184 norm min 520 norm hour<br />

2. Example 26 624 norm min 444 norm hour<br />

3. Example 28 052 norm min 467 norm hour<br />

Table 2: Table with results <strong>of</strong> calculations<br />

168<br />

V. CONCLUSION<br />

The best combination <strong>of</strong> charging <strong>the</strong> system by<br />

FIFO method requires 444 norm hours for<br />

machining both orders. As we sad it is three<br />

examples <strong>of</strong> 120 possibilities. Selected algorithms<br />

show how important it is for mechanical companies<br />

with conventional machines and also it can be a<br />

feedback for own technological preparation <strong>of</strong><br />

production.<br />

This project will be programmed into integrated<br />

s<strong>of</strong>tware package, where <strong>the</strong> whole program will be<br />

synergically built on three autonomous programs.<br />

The first program will ensure expert calculations <strong>of</strong><br />

technological times for machining based on method<br />

<strong>of</strong> primitives. The second program will ensure<br />

managing <strong>of</strong> stocks from production point <strong>of</strong> view.<br />

The third program will ensure algorithm which on<br />

<strong>the</strong> basis <strong>of</strong> initialization formula calculates <strong>the</strong><br />

shortest time for machining production batch or<br />

whole order. This s<strong>of</strong>tware package will work with<br />

<strong>the</strong> help <strong>of</strong> interface such as autonomous universal<br />

system on existing systems and will be tested in real<br />

mechanical production.<br />

ACKNOWLEDGMENTS<br />

I would like to acknowledge my pr<strong>of</strong>essor<br />

Dusan Sebo for his support and his advice.<br />

REFERENCES<br />

[1] Šebo, D.: LOGISTIKA AKO NÁSTROJ<br />

MANAŽERSKEJ ČINNOSTI, Edícia vedeckej<br />

a odbornej literatúry – Strojnícka fakulta, ISBN<br />

978-80-8073-776-4, Košice, 2007.<br />

[2] Malindžák, D.: VÝROBNÁ LOGISTIKA I.,<br />

Vydavateľstvo ŠTROFFEK KOŠICE, ISBN 80-<br />

967636-6-0, Košice, 1997.<br />

[3] Leščišin, M, Stern, J, Dupaľ, A.:<br />

MANAŽMENT VÝROBY, 1. Vydanie EU v<br />

Bratislave, ISBN 80-968881-0-2, Bratislava,<br />

2002.<br />

[4] TER-Manuelliac, A.: Matematické modely teórie<br />

zásob, In., Matematicko – ekonomický obzor č.<br />

2, Praha, 1972.<br />

[5] Leščišín, M, Stern, J, Dupaľ, A.:<br />

MANAŽMENT VÝROBY, SPRINT –vfra Bratislava,<br />

ISBN 80-89085-00-6, Bratislava, <strong>2008</strong>.


A Computer-Controlled Joint Loading System<br />

for Quantitative Analysis <strong>of</strong> Mechanoresponsiveness in Articular Cartilage<br />

Xiang Gu1,2,3, Daniel Leong1,2,3, Yong Hui Li2, Francisco Guzman1, Hui Bin Sun2,3 , and<br />

Luis Cardoso1,2<br />

1 Dept. <strong>of</strong> Biomedical Engineering, City College <strong>of</strong> New York, New York, U.S.A<br />

2 The Graduate Center, City University <strong>of</strong> New York, New York, U.S.A<br />

3 Mount Sinai School <strong>of</strong> Medicine, New York, U.S.A<br />

Email: cardoso@engr.ccny.cuny.edu<br />

Abstract —Motion and mechanical load <strong>of</strong> <strong>the</strong> joint<br />

have been associated with complex beneficial and<br />

detrimental effects on cartilage during <strong>the</strong> progression<br />

<strong>of</strong> degenerative joint diseases. The development <strong>of</strong> in<br />

vivo computer-controlled joint loading system is<br />

critical to studying how motion and load regulate<br />

cartilage homeostasis. The goal <strong>of</strong> this study was to<br />

develop a joint loading system to assess <strong>the</strong> biological<br />

response <strong>of</strong> cartilage due to well quantified<br />

underloading, physiological, and overloading<br />

conditions. A computer-controlled joint loading<br />

system was designed to apply precisely controlled<br />

cyclic motion and compressive loading protocols to<br />

<strong>the</strong> rat’s right knee, and under-, physiological, and<br />

over-loading conditions were created to test <strong>the</strong><br />

mechanoresponsiveness <strong>of</strong> <strong>the</strong> cartilage. In contrast,<br />

continuous passive motion suppressed catabolic<br />

effects induced by immobilization, while <strong>the</strong> overload<br />

led to <strong>the</strong> catabolic responses similar to those caused<br />

by immobilization. The current data was consistent<br />

with our previous in vitro studies where moderate load<br />

had anticatabolic effects, while underloading or<br />

overloading conditions may lead to cartilage<br />

degeneration.<br />

I. INTRODUCTION: Joint diseases such as<br />

Osteoarthritis (OA) are common cause <strong>of</strong> disability<br />

worldwide. Physical activity and weight bearing<br />

condition play an important role in <strong>the</strong> regulation <strong>of</strong><br />

joint homeostasis throughout life. Quantitative<br />

assessment <strong>of</strong> physiological dosage <strong>of</strong> mechanical load<br />

necessary to preserve <strong>the</strong> homeostasis <strong>of</strong> articular<br />

cartilage, and parametric characterization <strong>of</strong><br />

deleterious and beneficial joint loading regimens are<br />

<strong>of</strong> great interest from both a basic research and clinical<br />

practice point <strong>of</strong> view. Prevention and treatment <strong>of</strong><br />

OA depend on understanding <strong>the</strong> onset, progression <strong>of</strong><br />

cartilage degeneration associated with joint loading<br />

abnormalities, and <strong>the</strong> wide use <strong>of</strong> motion-based<br />

physical <strong>the</strong>rapies such as continuous passive motion<br />

(CPM) requires more scientific guidelines for optimal<br />

protocols (frequency, intensity, and duration, etc).<br />

The development <strong>of</strong> in vivo animal models is critical<br />

to studying how motion and load regulate cartilage<br />

homeostasis, by revealing underlying mechanisms<br />

mediating <strong>the</strong> biomorphological effects in articular<br />

joints in response to external mechanical stimuli. To<br />

date, few experimental devices <strong>of</strong>fer a non-invasive<br />

approach to providing precise control <strong>of</strong> joint loading<br />

169<br />

parameters, or wide range <strong>of</strong> well-quantified loading<br />

environments on small animal models such as rats. In<br />

<strong>the</strong> present work, we introduced a novel computercontrolled<br />

loading system capable <strong>of</strong> applying cyclic<br />

motions on a single anes<strong>the</strong>tized rat’s knee joint under<br />

axial compressive load. The system fulfills <strong>the</strong> control<br />

requirements <strong>of</strong> 1) Knee joint position and angular<br />

displacement (Initial/Final angle, Flexion); 2) Cyclic<br />

motion frequency; 3) Magnitude <strong>of</strong> <strong>the</strong> axial<br />

compressive loading force along <strong>the</strong> tibia; 4) Realtime<br />

monitoring <strong>of</strong> <strong>the</strong> system operation. We loaded<br />

<strong>the</strong> rats with under-, physiological, and over- loading<br />

conditions to determine <strong>the</strong> ability <strong>of</strong> <strong>the</strong> system in<br />

implementing defined joint loading protocols.<br />

II. MATERIALS AND METHODS: Motion<br />

Design. The animal is anes<strong>the</strong>tized using a mask on a<br />

custom built flat bed plane, <strong>the</strong> hip is maintained at a<br />

fixed height and <strong>the</strong> right hind femur is stabilized by a<br />

clamp coupled on <strong>the</strong> bed plane. The cyclic motion <strong>of</strong><br />

<strong>the</strong> right knee is implemented by a computer-controlled<br />

linear actuator. An adaptor attached to <strong>the</strong> shaft <strong>of</strong> <strong>the</strong><br />

linear actuator translates <strong>the</strong> linear motion <strong>of</strong> <strong>the</strong><br />

actuator into <strong>the</strong> rotational motion experienced by <strong>the</strong><br />

animal’s knee. The user defines <strong>the</strong> loading parameters<br />

by inputting <strong>the</strong> initial and final angles (in deg) and <strong>the</strong><br />

cyclic frequency (in cycle/min) on <strong>the</strong> LabView control<br />

panel. By adjusting <strong>the</strong> extension <strong>of</strong> two springs on<br />

each side <strong>of</strong> <strong>the</strong> tibia, <strong>the</strong> axial compressive force is<br />

applied on <strong>the</strong> knee. In vivo animal study. Five<br />

months old male Sprage Dawley rats (n=5/group) were<br />

randomly assigned to five testing groups <strong>of</strong> animals by<br />

analyzing <strong>the</strong> gene expression <strong>of</strong> selected proinflamatory<br />

and anti-inflamatory proteins due to<br />

different motion conditions. For <strong>the</strong> immobilization<br />

group (IM), animals were immobilized for six hours<br />

using a cast made <strong>of</strong> cotton and steel mesh as described<br />

in Coutinho et al. (2002) [1]. A second group <strong>of</strong><br />

animals (CPM) was immobilized for 2.5 hours,<br />

remobilized using CPM (70° flexion, 30 sec/cycle) for<br />

one hour, and immobilized again for 2.5 more hours.<br />

The third group was treated similar to <strong>the</strong> CPM group,<br />

while <strong>the</strong> joints were overloaded (OVL) during <strong>the</strong> one<br />

hour motion treatment by applying an axial load<br />

equivalent to ~ 2X body weight. A fourth group <strong>of</strong><br />

animals (SHAM) was treated exactly <strong>the</strong> same way as<br />

those in <strong>the</strong> CPM group except that <strong>the</strong> animals on <strong>the</strong><br />

device did not received any motion or loading. A fifth


group not subjected to any treatment was used as naïve<br />

controls (NAI). Real-time PCR. After each rat’s<br />

respective experiment endpoint, <strong>the</strong> lateral and femoral<br />

condyles were dissected, rinsed in 1X PBS/DEPCtreated<br />

water, and frozen in liquid nitrogen. The<br />

isolation <strong>of</strong> total RNA was conducted using <strong>the</strong> Rneasy<br />

mini protocol. The RNA was <strong>the</strong>n reverse transcribed<br />

(RT) with reverse transcriptase using Oligo(dT), and<br />

<strong>the</strong> RT products were amplified with real-time PCR,<br />

using GAPDH for normalization. The primers used<br />

were: MMP-3, MMP-13, Collagen II.<br />

III. RESULTS: Figure 1 shows <strong>the</strong> schematic<br />

system design. It consists <strong>of</strong> a linear actuator with<br />

micro-step drive and encoder, a knee joint loading<br />

apparatus for controlled movement and adjustable<br />

compressive force on <strong>the</strong> rat’s knee joint, a magnetic<br />

encoder for real-time joint angle readout, a load cell for<br />

real-time compressive force monitoring, a USB port<br />

analog/digital data acquisition module, and an<br />

anes<strong>the</strong>sia machine (not shown). Figure 2 shows <strong>the</strong><br />

experimental settings, <strong>the</strong> system covers <strong>the</strong><br />

physiological range <strong>of</strong> motion (ROM) from 55° to<br />

125°, <strong>of</strong>fers up to 4 cycles/min speed for 70° and up to<br />

5X body weight axial compressive load. Figure 3<br />

shows <strong>the</strong> LabView s<strong>of</strong>tware interface with user inputs<br />

and three channels <strong>of</strong> real-time measurement. Figure 4<br />

shows gene expressions <strong>of</strong> Pro-inflammatory mediators<br />

(MMP-3 and MMP-13) and anti-inflammatory<br />

mediator Collagen II in response to immobilization,<br />

physiological, and overloading conditions. In contrast,<br />

overload resulted in catabolic responses (increased<br />

MMP, decreased collagen II) similar to those caused by<br />

immobilization, while CPM reversed <strong>the</strong> catabolic<br />

responses caused by immobilization.<br />

IV. DISCUSSION: The development <strong>of</strong> <strong>the</strong><br />

animal model incorporated with <strong>the</strong> novel joint loading<br />

system addresses <strong>the</strong> quantification <strong>of</strong> <strong>the</strong> motion and<br />

load in <strong>the</strong> characterization <strong>of</strong> <strong>the</strong> consequential tissue<br />

responses. The flexibility <strong>of</strong> <strong>the</strong> system allows for test<br />

or evaluation <strong>of</strong> a broad range <strong>of</strong> clinically relevant<br />

protocols and motion-based hypo<strong>the</strong>ses. The outcome<br />

from <strong>the</strong> animal study provided supporting evidence<br />

for validation <strong>of</strong> <strong>the</strong> system designed for extensive<br />

experiments in biomechanics research. The expression<br />

<strong>of</strong> MMP-3, MMP-13, and Collagen II were found<br />

similar during immobilization and overloading, and<br />

<strong>the</strong>y had <strong>the</strong> opposite behavior compared to <strong>the</strong> results<br />

<strong>of</strong> passive motion, suggesting continuous excessive<br />

stress on <strong>the</strong> joint in motion can result in degeneration<br />

<strong>of</strong> articular cartilage. The current data was consistent<br />

with our previous in vitro studies [2] where moderate<br />

load had anticatabolic effects, while underloading [3]<br />

or overloading conditions may lead to cartilage<br />

degeneration. This novel loading system can contribute<br />

to quantitative investigations <strong>of</strong> mechanoresponsiveness<br />

in joint tissues, and can help better<br />

understand <strong>the</strong> effects <strong>of</strong> physical activity and weight<br />

bearing condition on joint homeostasis, in particular,<br />

170<br />

<strong>the</strong> role <strong>of</strong> important pro- and anti- inflammatory<br />

effectors in cartilage in regard <strong>of</strong> <strong>the</strong> mechanical<br />

loading environment.<br />

1.<br />

2.<br />

3.<br />

4.<br />

REFERENCES<br />

[1] Coutinho E.L., Gomes A.R., Franca C.N., and<br />

Salvini T.F.. A new model for <strong>the</strong> immobilization<br />

<strong>of</strong> <strong>the</strong> rat hind limb. Braz J Med Biol Res; 35:<br />

1329-1332, 2002<br />

[2] Sun HB, and Yokota H. CITED2-mediated<br />

Regulation <strong>of</strong> MMP-1 and MMP-13 in Human<br />

Chondrocytes under Flow Shear.J. Biol. Chem.,<br />

278(47): 47275-47280, November 21 2003<br />

[3] Gu, X; Leong, D; Li, Y.H.; Guzman, F; Sun<br />

H. B; and Cardoso, L. Computer-Controlled Joint<br />

Loading System For Biomechanics Research In A<br />

Rat Model. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 2007 Annual<br />

Biomedical Engineering Society Meeting, Los<br />

Angeles, California, September 2007


��������������������������������������������������������������<br />

��������������������������������������������������<br />

�<br />

������������������������������������������������������������������������������������<br />

���������<br />

�������������������������������������<br />

������������������������������<br />

������������������<br />

���������������������������������<br />

�<br />

�<br />

��������� �� ����� ������ ������������ ������� ����<br />

����� ��������� ����� ���������� �������� ��� ����<br />

���������� ��� ������������ ��������� ���������<br />

������������ ���� ���������� ����������� ���������<br />

������ ������ ���������� ����������� ��� �������<br />

������� ��������� ��� ����� ��������� �� ������������<br />

������� ��� ������ ���� �������� ��� ������������ ���������<br />

������� ��� ������� ������� ������������ ����� �������<br />

������������ ������ ����� ����� ��� �� ������ ���� ����<br />

������ ������ �������� ������ ������������� �������<br />

������������ ��� ��������� ��� ��������� ��� ����<br />

���������������������������������������������������<br />

��� ���� ������ ����������� ���� ����� ������� ��� ���������<br />

������ ����� ������� ����� ��������� ���� ���� ������ �����<br />

������ ��� ����������������� �������� �������� ����<br />

���������� ���������� ��� ���������� ������ �������� ���<br />

��������� ������� ��� ���� ������������������� ���� ����<br />

�����������������������������������������<br />

��� �������������<br />

�������������������������������������������������<br />

������������������������������������������������������<br />

����� ���������� ���� �������� ���������� ������ �����<br />

���������� ����������� ���������������� ���� �����<br />

���������������������������������������������<br />

��������� ��� �� ��������� ���������� ���� ����� ������ ���<br />

���������������<br />

��������������������������������������������������<br />

���� ������� ��� ������� ��� ���� ���������� ���<br />

������������� �������� ���� ���������� ����<br />

�������������������������������������������������<br />

����������� ����� ��� ������� ��� ����������������<br />

����������������������������������������������������<br />

��������������������������������������������������<br />

��� ����� ������� ��� ������� ��� ���������� ��������<br />

���� �������� ��������� ������� ����� ������� ���<br />

���������������� ��� ������������ ����������� ����<br />

�������� ����� ��������� ����� ����� ��� ��������� �����<br />

�������������������������������������� ����������<br />

��� ���������������� ����� ���� � ����� �����������<br />

������������ ��� ���������� ��� ������������� ��� ����<br />

��������� ��������� ������� ����� ��������� ����������<br />

171<br />

������� ������������ ������������ �������������������<br />

����������������������������������������������������<br />

���� ���������� ���������������� ��� ������������<br />

�������������� ��� ���� ������ ����� ������������<br />

��������������������������������������������������<br />

�������������������������������������������������<br />

���� ���� ������ ��� ���� �������� ��� ���������� ���������<br />

������� ��� ����� ������������ ����������� ��������<br />

��������� ������� ����� ���������� ������� �� ����������<br />

����������������������������������������������������<br />

����� �������� ��� �� �������� ������� �����<br />

���������������� ���� ������ ����� ����������� ������<br />

�������������������������������������������<br />

���� ����������������������<br />

����� ����� ��������� ������ ��������� ���������<br />

������ ����������� ��� �������� ���������� �������<br />

�������������������������������������������������<br />

������������������������������������������������������<br />

������������� ���������� ������ ����� ������� �����<br />

����������������������������������������������������<br />

��������������������������������������������������<br />

������������ ��� �������� ��������� ��� ������� ��������<br />

�������������������������������������������������<br />

������ ��� ������������� ����������� ����� ������� �����<br />

������� ��������� ����� ��������� ������� ����� ��� ���<br />

���� ��� �����������������������������������������<br />

�������������������������������������������������������<br />

����������������������������������������������������<br />

�����������<br />

�������� ����������� ����� �������� ��� ���� �����<br />

�������� ���� ����� ���� ������������ ������������<br />

������ ������������� ���� ���������� ���� ����������<br />

����� ����� ���� ��� ���� ������ ������ ����<br />

����������� ����� ����� ������� ��� ����� ����� ���<br />

������� ������ ���� ��������� �� ��� ��������� ��� ����<br />

���� ��� �� ���� ���� ����� ������ ���� �� ����� ����<br />

��������� ��� ��������� ������������ ���� �����������<br />

������������������������������������������������<br />

�������� ���� �������� ����� ����������� �������� ���<br />

����������������������������������������������������


�<br />

�<br />

����������������������������������������������������<br />

�������������������<br />

���� ���������� ������������� �������������� ���<br />

�������� �� ������ ���� ������������ ���������<br />

������������� ������� ���� ����������� � ���� ������ �����<br />

����������������������������������������������������<br />

��� �� ������� ������� �� ������������� ������ ������ ����<br />

��������������������������������������������������<br />

��������������������������������������<br />

�����������������������������������������������<br />

�������������������������������������������������<br />

��� ���������� ���� ����������� ���� ������ ����������<br />

�����������������������������������������������������<br />

���� ��� ��� ���������� �������������<br />

���������������������������������������������������<br />

�������������������������������������������<br />

���� �������� ��� ������ ���� ����������� ��� ����<br />

��������� ������ ������������ ��� �� ���� �����������<br />

�������� ����� ���� ������������� �������� �������<br />

������������ � ���� ��������� ��� ���������� �����<br />

������������������������������������<br />

�<br />

���������������������������������������<br />

����� ��������<br />

��� ������������ ���������������� ���� ���������<br />

�����������������������������������������������������<br />

������������������������������������������������<br />

�����������������������������������������������������<br />

�����������������������������������������������������<br />

����� ��� ���� ���� ����������� �������������<br />

����������� ��������� ��� ����������� ����������������<br />

���� ����� ������� ��� �������� ��� ������ ����� �������<br />

��������������������������������������������������<br />

�����������������������������������������������������<br />

���������������������������������������������������<br />

������������������������������������������������������<br />

��������������<br />

�<br />

�����������<br />

����<br />

����<br />

����<br />

����<br />

����<br />

����<br />

������� ��������<br />

��������� �������<br />

� �� �� �� ��� ��� ���<br />

���������<br />

�������������������������������<br />

172<br />

�<br />

�<br />

��������������������������������������������������<br />

��������� ��� �������� ���� ���� ���������� ���� ���<br />

�������������������������������������������������������<br />

��������������������������������������������������<br />

��������������������������������������������������<br />

��� ����� ������ ��� ������� ����� ����������������� ���<br />

������ ��� ���� ���������� ������� ���� ��������� ����<br />

��������������������������������������������<br />

�<br />

�<br />

����������������������������������������<br />

��������������������������������������������������<br />

��� �������� ��� ���� ����������� ��� ���� ������ ���������<br />

�������� ������� ��� ���������� ��� ���� ����� ��������<br />

�������������������������������������������������������<br />

��������������������������������������������������<br />

������������������������������������������������<br />

����������������<br />

����� ����� ���� ���������� ��� ���� ���� ����<br />

�����<br />

�����������<br />

����� ���������� ���� ��������� ����� �������� �����<br />

��������� ����� ������ ����� ������ ���� ������<br />

���������������� ��� ������ ��������������<br />

������������ ������� ��������� ���� ��������������<br />

���������� ��� ��������������������� ����<br />

���������������������������<br />

���� ������� ������ �������� ������ ������� ���� �������<br />

������ ������������ ������ ������ ����������������<br />

��������� ��������� ��� ���������� ��� ������<br />

����������� ����������� �������������� �������� ���<br />

����������������������������<br />

���� ������� ������������ ���������� ��������<br />

������������ ������������ �������� ����������<br />

��� ������ �������� ��� ������ ��������� ��� �� �����<br />

���� ���������������� ������ ��� ����<br />

���������������������������������������������������<br />

����������������������<br />

���� ��������� ����� ������ ����� ���� ������� ������ ����<br />

������������������������������������������������<br />

�������������������<br />

��<br />


Colloidal gel formation dynamics:<br />

light scattering, rheometry and simulation<br />

Xiujuan Cao, Herman Z. Cummins and Jeffrey Morris<br />

Levich Institute<br />

The City College <strong>of</strong> New York, Graduate Center, CUNY<br />

New York, United States<br />

Email: xcao@che.ccny.cuny.edu<br />

Abstract — Effects <strong>of</strong> particle concentration and ionic<br />

strength on <strong>the</strong> gelation <strong>of</strong> colloidal silica particles are studied<br />

by dynamic and static light scattering (DLS, SLS) and rheological<br />

measurements. The combination <strong>of</strong> <strong>the</strong>se experimental<br />

approaches gives us a simultaneous observation <strong>of</strong> microstructure<br />

and macroscopic rheology. The hydrodynamic radius Rh,<br />

measured through DLS, was found to grow exponentially. Approaching<br />

<strong>the</strong> gel point, <strong>the</strong> fractal dimension Df was determined<br />

to be 1.8 by SLS. The storage modulus G ′ and loss modulus<br />

G ′′ become measurable and increase as <strong>the</strong> gel forms.<br />

Stokesian and Brownian Dynamics simulation techniques are<br />

applied to study <strong>the</strong> evolution <strong>of</strong> particle aggregation and gelation,<br />

with a focus on understanding <strong>the</strong> role <strong>of</strong> hydrodynamics.<br />

I. INTRODUCTION<br />

In an aqueous colloidal silica suspension, <strong>the</strong>re is negative<br />

charge on <strong>the</strong> surface <strong>of</strong> <strong>the</strong> particles, which stabilizes<br />

<strong>the</strong> dispersion against flocculation by inducing electrostatic<br />

repulsion. The addition <strong>of</strong> salt solution reduces<br />

<strong>the</strong> repulsive barrier <strong>of</strong> potential energy, destabilizing <strong>the</strong><br />

kinetically stable system. The resulting aggregation <strong>of</strong><br />

<strong>the</strong> particles leads to gelation: <strong>the</strong> dynamics <strong>of</strong> this process<br />

in general and <strong>the</strong> specific role <strong>of</strong> hydrodynamic interactions<br />

are <strong>of</strong> primary interest in this study.<br />

II. RESULTS AND DISCUSSION<br />

We used <strong>the</strong> aqueous colloidal silica suspension Ludox−<br />

SM (pH = 10.47) with radius 7.0 nm and 32 mass% SiO2<br />

(corresponding to a volume fraction ϕ =18.4%) from<br />

Grave Davison. The aggregation is initiated by <strong>the</strong> addition<br />

<strong>of</strong> aqueous NaCl under stirring.<br />

A. DYNAMIC LIGHT SCATTERING<br />

Dynamic light scattering, which probes <strong>the</strong> aggregation<br />

kinetics, measures <strong>the</strong> time-averaged time correlation<br />

function <strong>of</strong> <strong>the</strong> intensity I(q,t) <strong>of</strong> light scattered by <strong>the</strong><br />

sample in <strong>the</strong> direction described by <strong>the</strong> scattering vector<br />

q, |q| =(4πn/λ)sin(θ/2) (n is <strong>the</strong> refractive index). In<br />

our experiment, <strong>the</strong> samples are tested with λ = 488 nm,<br />

n = 1.33, and with scattering angle θ =90 o . Normalization<br />

<strong>of</strong> <strong>the</strong> intensity correlation function g2(t) is<br />

g2(t) =/ 2 . (1)<br />

173<br />

For a polydisperse system, g2(t) can be expressed as<br />

|g2(t)| =1+aexp[−(2t/τ) β ], (2)<br />

where 1/τ = Dq2 , and D = kT/6πηRh. Equation (2)<br />

can be used to fit <strong>the</strong> intensity correlation function data<br />

in order to obtain a first estimate <strong>of</strong> Rh.<br />

After destabilization <strong>of</strong> <strong>the</strong> system, hydrodynamic radius<br />

Rh was measured through DLS, and found to grow<br />

exponentially (Figure 1). Approaching <strong>the</strong> gel point, Rh<br />

diverges, apparently because <strong>the</strong> particles form a spacefilling<br />

network. The initial aggregation rate is very slow<br />

and <strong>the</strong> rate increases as <strong>the</strong> size <strong>of</strong> <strong>the</strong> cluster increases;<br />

large clusters have more possibilities to connect to particles<br />

and clusters [1].<br />

Figure 1: Evolution <strong>of</strong> hydrodynamic radius Rh<br />

B. STATIC LIGHT SCATTERING<br />

Static light scattering averages <strong>the</strong> scattering intensities<br />

detected and observed at one specific scattering angle. It<br />

is applied to determine <strong>the</strong> fractal dimension Df (Figure<br />

2). At <strong>the</strong> beginning <strong>of</strong> aggregation, <strong>the</strong> I(q) curves<br />

do not demonstrate power-law behavior. As <strong>the</strong> clusters<br />

grow, <strong>the</strong> power-law dependence <strong>of</strong> I(q) is seen over<br />

most <strong>of</strong> <strong>the</strong> observation regime and <strong>the</strong> slope is <strong>the</strong>n determined<br />

to be <strong>the</strong> fractal dimension Df =1.8 [2].<br />

C. RHEOLOGICAL APPROACH<br />

Rheological measurements are a measure <strong>of</strong> <strong>the</strong> mechanical<br />

properties <strong>of</strong> <strong>the</strong> colloidal network. As gel forms,


Figure 2: Structure evolution by SLS<br />

<strong>the</strong> storage (G ′ ) and loss (G ′′ ) modulus were measured<br />

as function <strong>of</strong> <strong>the</strong> time at fixed frequency 1.0 rad/s, as<br />

shown in (Figure 3). After gel formation, dynamics are<br />

largely due to <strong>the</strong> unconnected particles, particle chains<br />

and particle clusters among <strong>the</strong> network, which continue<br />

to make contributions to <strong>the</strong> stiffness <strong>of</strong> system as <strong>the</strong>y<br />

attach to <strong>the</strong> space-filling network (<strong>the</strong> increase in storage<br />

modulus G ′ ). Finally, each particles is trapped in a<br />

small region and cannot move any fur<strong>the</strong>r. The increase<br />

in <strong>the</strong> storage modulus becomes very slow and ultimately<br />

reaches to a constant value [3].<br />

Figure 3: Storage and loss modulus evolution<br />

D. SIMULATION RESULTS<br />

Stokesian and Brownian Dynamics simulation techniques<br />

[4] are applied to study <strong>the</strong> relationship between<br />

microstructure and mechanical properties <strong>of</strong> <strong>the</strong> colloidal<br />

gel. At a solid volume fraction <strong>of</strong> φ = 10%, <strong>the</strong> system<br />

forms compact structures; for lower φ =1%, <strong>the</strong><br />

particles evolve into small clusters (Figure 4 and Figure<br />

5). The fractal structure <strong>of</strong> colloidal gel is predicted<br />

depending on <strong>the</strong> balance <strong>of</strong> long-range repulsive forces<br />

and short-range attractive forces with <strong>the</strong> force form<br />

F = −∇V, (3)<br />

174<br />

where<br />

V = 40[( 2<br />

r )12 − ( 2<br />

r )6 ]+40 e−r/2<br />

r<br />

. (4)<br />

Figure 4: Cluster aggregation for φ = 10%,t= 100.<br />

Figure 5: Cluster aggregation for φ =1%,t= 100.<br />

ACKNOWLEDGMENTS<br />

I would like to thank my advisor Pr<strong>of</strong>essor Jeffrey Morris,<br />

who is very supportive and knowledgable. I would<br />

like to thank Dr. Xiangnan (Allen) Ye for his patient<br />

guidance on <strong>the</strong> experimental work.<br />

REFERENCES<br />

[1] M. Y. Lin, H. M. Lindsay, D. A. Weitz, R. C. Ball,<br />

R. Klein, and P. Meakin. Universal reaction-limited<br />

colloid aggregation. Phys. Rev. A., 41(4):2005–2020,<br />

1990.<br />

[2] D. W. Schaefer and J. E. Martin. Fractal geometry <strong>of</strong><br />

collodal aggregates. Phys. Rev. Lett., 52(26):2371–<br />

2374, 1984.<br />

[3] V. Prasad, V. Trappe, A. D. Dinsmore, P. N. Segre,<br />

L. Cipelletti, and D. A. Weitz. Universal features<br />

<strong>of</strong> <strong>the</strong> fluid to solid transition for attractive colloidal<br />

particles. Faraday Discuss., 123:1–12, 2002.<br />

[4] J. F. Brady and G. Bossis. Stokesian dynamics. Ann.<br />

Rev. Fluid Mech., 20:111–157, 1988.


�����������������������������<br />

������������������������������������<br />

����������������<br />

�<br />

�����������������������������������<br />

��������������������������������������������������������������������������<br />

������������������������������������<br />

�����������������������������<br />

�<br />

��������������������������������������������������������������������������������������������������������<br />

���������������������������������������������������������������������������������������������������������<br />

���� ������ ��� ����� �������� ��� ���� ������������ ��� ���� ���������� ������������ ��� ����������� ������ ��� ����<br />

�����������������������������������������������������������������������������������������������������������<br />

������������������������������������������ � �����������������������������������������������������������������<br />

��������������������������������������������������������������������������������������������������������������<br />

��������������������� �<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

���������� ��������� ��� ������������� �������������� ��� ���� ������������ ���� ��� ���� ���������� ��� �����������<br />

����������������������������������������������������������������������������������������������������������<br />

�������������������<br />

�<br />

�<br />

�<br />

�<br />

�������������������������������������������������<br />

��������������������������������������������������������������������������������������<br />

���������������������������������������������������������������������������������������������������<br />

�����������������������<br />

175<br />

�<br />


��������<br />

�<br />

��� ������������� ������������� ���������� ��� �������� �� ���� �� ��������� ����� ���� ����������� ������<br />

���������������������������������������������������������������������<br />

�<br />

������������������������������������������������������������������������������������������������������������<br />

�������������������������������������������������������������������������������������������������������<br />

����������������������������<br />

�<br />

�<br />

�<br />

���� ���������� ������ ��������� ����� �������������������������������������� �� ������� ���� ��� ��� ��������������<br />

������������� ������ ��������� ������ ������ ����������� ���� ���� ����� �������� ���� ������������ ����� ��� ���� ��<br />

����������������������������������������������������������������������������������������������������������<br />

��������� ������ ����� ������ �������� ���� ������ ������������ ������� ������������ ��������� ���� ����������<br />

����������������������������������������������������������������������<br />

�<br />

�������������������������������������������������������������������������������������������������������������<br />

��������������������������������������������������������������<br />

���������������������������������������������������������������������������������������������������������������<br />

����������������������������������������������������<br />

�<br />

���������������������������� ���������������������������������������������������������������������������<br />

��������������������������������������������������������<br />

176


Site-Selective Time Resolved Laser Fluorescence Spectroscopy on<br />

Ca 2+ -Bearing Mineral Phases Doped with Europium<br />

Moritz Schmidt a , Thorsten Stumpf a,b , Clemens Wal<strong>the</strong>r a , Horst Geckeis a , Thomas Fanghänel b,c<br />

a)<br />

Institut für Nukleare Entsorgung, Forschungszentrum Karlsruhe, P.O. Box 3640, 76021<br />

Karlsruhe (Germany)<br />

b)<br />

Physikalisch-Chemisches Institut, Ruprecht-Karls-Universität Heidelberg, Im Neuenheimer<br />

Feld 253, 69120 Heidelberg (Germany)<br />

c)<br />

European Commission, Joint Research Centre, Institute for Transuranium Elements, P.O. Box<br />

2340, 76125 Karlsruhe (Germany)<br />

Email: moritz.schmidt@ine.fzk.de<br />

Abstract — The aim <strong>of</strong> <strong>the</strong> presented work is to<br />

understand <strong>the</strong> interaction behaviour <strong>of</strong> trivalent<br />

actinides with Ca 2+ -bearing mineral phases on a<br />

molecular level. A comparison <strong>of</strong> <strong>the</strong> sorption behaviour<br />

<strong>of</strong> Eu 3+ (as an inactive homologue <strong>of</strong> <strong>the</strong><br />

trivalent actinides with convenient fluorescence<br />

properties) with aragonite and gypsum is presented.<br />

Results from time resolved laser fluorescence spectroscopy<br />

show a significantly different mode <strong>of</strong><br />

interaction in <strong>the</strong> two cases. While in aragonite only<br />

structural incorporation is observed, on gypsum<br />

nothing but inner-sphere sorption complexes can be<br />

found.<br />

I. INTRODUCTION<br />

Assessing <strong>the</strong> geochemical long-term safety <strong>of</strong> a<br />

nuclear waste repository requires a molecular level<br />

understanding <strong>of</strong> radionuclide behavior in <strong>the</strong> geosphere.<br />

In particular, <strong>the</strong> interaction <strong>of</strong> radionuclides<br />

with mineral phases (adsorption, solid solution formation)<br />

strongly affects <strong>the</strong>ir mobility and sequestration.<br />

The retardation capabilities <strong>of</strong> <strong>the</strong> mineral<br />

phases in <strong>the</strong> geotechnical and geological barrier are<br />

a key component <strong>of</strong> <strong>the</strong> safety assessment <strong>of</strong> every<br />

nuclear waste repository in deep geological formations[1].<br />

The actual mode <strong>of</strong> interaction (e.g. formation<br />

<strong>of</strong> inner and outer sphere complexes and structural<br />

incorporation) is a major controlling factor <strong>of</strong><br />

this retardation capability. Recent results on <strong>the</strong> interaction<br />

<strong>of</strong> Eu 3+ and Cm 3+ with calcite show <strong>the</strong><br />

formation <strong>of</strong> solid solutions by exchange <strong>of</strong><br />

Eu 3+ /Cm 3+ for Ca 2+ [2, 3]. Thus <strong>the</strong> question arose if<br />

this mechanism is generally applicable for Ca 2+ -<br />

bearing mineral phases.<br />

177<br />

II. RESULTS<br />

In order to shed light on this question, Ca 2+ -<br />

bearing mineral phases homogeneously doped with<br />

Eu 3+ were syn<strong>the</strong>sized in a mixed-flow-reactor. The<br />

investigated mineral phases are aragonite, a metastable<br />

CaCO3 modification, and gypsum CaSO4·2H2O.<br />

Europium fluorescence spectroscopy is a versatile<br />

tool for <strong>the</strong> elucidation <strong>of</strong> <strong>the</strong> local structure, hydration<br />

state and substitution mechanism in crystalline<br />

and non-crystalline systems (e.g. [2, 4-6]). The applied<br />

site-selective excitation <strong>of</strong> <strong>the</strong> 7 F0 � 5 D0 transition<br />

at low temperatures (


Figure 1: Excitation spectra <strong>of</strong> <strong>the</strong> 7 F0 --> 5 D0<br />

transition in Eu 3+ doped aragonite (circles) and gypsum<br />

(crosses).<br />

Time resolved laser fluorescence spectroscopy results<br />

clearly show that Eu 3+ exhibits completely opposed<br />

sorption behavior on both investigated Ca 2+<br />

mineral phases. In <strong>the</strong> aragonite system formation <strong>of</strong><br />

solid solutions by substitution for Ca 2+ is <strong>the</strong> only<br />

observed form <strong>of</strong> interaction, which is evident by <strong>the</strong><br />

corresponding emission spectrum and fluorescence<br />

lifetime. However, in <strong>the</strong> gypsum system no incorporation<br />

occurs, but only one inner-sphere sorption<br />

species can be found. On macroscopic level ICP-MS<br />

results show a significantly weaker Eu 3+ uptake by<br />

gypsum compared to aragonite.<br />

These results clearly show that <strong>the</strong> exchange<br />

mechanism observed for calcite and aragonite is not<br />

generally applicable to o<strong>the</strong>r Ca 2+ -bearing mineral<br />

phases. Apparently o<strong>the</strong>r parameters such as <strong>the</strong><br />

binding strength <strong>of</strong> <strong>the</strong> anion have to be taken into<br />

consideration. It is, however, evident that radically<br />

different modes <strong>of</strong> interaction as observed for <strong>the</strong>se<br />

two mineral phases will have a considerable impact<br />

on <strong>the</strong>ir retardation capabilities, and thus on <strong>the</strong>ir<br />

desirability in <strong>the</strong> near and far field <strong>of</strong> a potential<br />

waste repository.<br />

ACKNOWLEDGMENTS<br />

This work was co-financed by <strong>the</strong> Helmholtz Gemeinschaft<br />

Deutscher Forschungszentren (HGF) by<br />

supporting <strong>the</strong> Helmholtz-Hochschul-Nachwuchsgruppe<br />

"Aufklärung geochemischer Reaktionsmechanismen<br />

an der Wasser/Mineralphasen Grenzfläche".<br />

REFERENCES<br />

[1] N. M. Edelstein, R. Klenze, T. Fanghänel,<br />

and S. Hubert, "Optical properties <strong>of</strong><br />

Cm(III) in crystals and solutions and <strong>the</strong>ir<br />

178<br />

application to Cm(III) speciation," Coordination<br />

Chemistry Reviews, vol. 250, pp.<br />

948-973, 2006.<br />

[2] M. Schmidt, T. Stumpf, M. M. Fernandes,<br />

C. Wal<strong>the</strong>r, and T. Fanghaenel, "Charge<br />

compensation in solid solutions," Angewandte<br />

Chemie Int. Ed., vol. 47, pp. 5846-<br />

5850, <strong>2008</strong>.<br />

[3] M. M. Fernandes, M. Schmidt, T. Stumpf,<br />

C. Wal<strong>the</strong>r, D. Bosbach, R. Klenze, and T.<br />

Fanghaenel, "Site-selective time resolved<br />

laser fluorescence spectroscopy <strong>of</strong> Eu 3+<br />

doped calcite," Journal <strong>of</strong> Colloid and Interface<br />

Science, vol. 321, pp. 323-332,<br />

<strong>2008</strong>.<br />

[4] K. Binnemans and C. Görller-Walrand,<br />

"Crystal field analysis <strong>of</strong> EuCl3·6H2O,"<br />

Journal <strong>of</strong> Alloys and Compounds, vol. 250,<br />

pp. 326-331, 1997.<br />

[5] B. Piriou, M. Fedor<strong>of</strong>f, J. Jeanjean, and L.<br />

Bercis, "Characterization <strong>of</strong> <strong>the</strong> sorption <strong>of</strong><br />

europium(lll) on calcite by site-selective<br />

and time-resolved luminescence spectroscopy,"<br />

Journal <strong>of</strong> Colloid and Interface Science,<br />

vol. 194, pp. 440-447, 1997.<br />

[6] T. Stumpf and T. Fanghänel, "A timeresolved<br />

laser fluorescence spectroscopy<br />

(TRLFS) study <strong>of</strong> <strong>the</strong> interaction <strong>of</strong> trivalent<br />

actinides (Cm(III)) with calcite," Journal <strong>of</strong><br />

Colloid and Interface Sciences, vol. 249, pp.<br />

119-122, 2002.


Nan<strong>of</strong>eatured carbon-copper metal matrix composites<br />

Michael Kitzmantel, Christoph Eisenmenger-Sittner (Faculty Mentor) and Dan Cunningham *<br />

Institute for Solid State Physics, Vienna University <strong>of</strong> Technology, Austria<br />

* Department <strong>of</strong> Engineering Mechanics, The Pennsylvania State University, USA<br />

Email: michael.kitzmantel@student.tuwien.ac.at<br />

Abstract — In electronic devices materials with a<br />

high <strong>the</strong>rmal conductivity and a coefficient <strong>of</strong> <strong>the</strong>rmal<br />

expansion matching to <strong>the</strong> ones <strong>of</strong> Si, GaAs or<br />

alumina are used as heat sinks, heat spreaders and<br />

substrates. Besides <strong>the</strong> high <strong>the</strong>rmal conductivity a<br />

low coefficient <strong>of</strong> <strong>the</strong>rmal expansion (CTE) is a<br />

crucial property for <strong>the</strong> long-term stability and<br />

reliability <strong>of</strong> <strong>the</strong> products. Today commonly used<br />

materials such as Mo-Cu, W-Cu, Kovar, Al/SiC,<br />

AlN or diamond exhibit several limitations in respect<br />

to limited <strong>the</strong>rmal conductivity (Kovar), high<br />

density (Mo-Cu, W-Cu), bad machinability (Al/SiC,<br />

AlN) or high price (diamond).<br />

The wetting and infiltration behavior <strong>of</strong> copper<br />

and some copper alloys containing carbide forming<br />

elements such as Ti, Mo, Cr or B on carbon nan<strong>of</strong>ibers<br />

was investigated. These additives promote<br />

wetting on <strong>the</strong> carbon surface. It is shown how<br />

carbon nan<strong>of</strong>ibers (CNF) and nanotubes (CNT)<br />

can be used to reinforce a copper matrix. Several<br />

ways <strong>of</strong> producing metal matrix composites (MMC)<br />

with carbon nanoparticles as building blocks were<br />

investigated, including liquid metal infiltration,<br />

powder metallurgical approaches and chemical<br />

coating <strong>of</strong> <strong>the</strong> nan<strong>of</strong>ibers.<br />

I. INTRODUCTION<br />

As <strong>the</strong> building elements <strong>of</strong> micorelectronics get<br />

smaller and have to cope with growing power density,<br />

more efficient cooling are a hot topic. Suitability<br />

for cycling temperatures as well as long-term<br />

stability is also crucial for future materials.<br />

However, it is difficult to fabricate carbon/metal<br />

composites because <strong>of</strong> <strong>the</strong> poor wettability between<br />

carbon and molten metal alloys [1].<br />

Miscellaneous investigations with graphite flakes<br />

and regular carbon fibers have been reported to show<br />

promising results [2]. Due to <strong>the</strong> lack <strong>of</strong> chemical<br />

interactions between carbon and <strong>the</strong> matrix metal, <strong>the</strong><br />

manipulation and investigation <strong>of</strong> <strong>the</strong> interface plays<br />

an essential role.<br />

Several potentially interesting matrices degrade<br />

<strong>the</strong> fibers [3] at high temperatures and <strong>the</strong>refore<br />

179<br />

development has concentrated on matrices chemically<br />

inert towards carbon, such as copper.<br />

The non-wetting behavior <strong>of</strong> <strong>the</strong> C/Cu system [4],<br />

leads to several problems to be addressed in <strong>the</strong> production<br />

<strong>of</strong> MMCs. How can <strong>the</strong> wetting behavior be<br />

promoted, and how to infiltrate carbon fiber preforms<br />

with <strong>the</strong>se metals? One approach is to alloy <strong>the</strong> liquid<br />

metal with an element that will react with <strong>the</strong> carbon<br />

fibers to change <strong>the</strong>ir wettability but that will not<br />

produce serious degradation. To investigate <strong>the</strong> wetting<br />

behavior, <strong>the</strong> sessile drop technique was successfully<br />

used in various studies [5]. Ano<strong>the</strong>r way is to<br />

functionalize <strong>the</strong>� carbon fibers <strong>the</strong>mself to build a<br />

catalytically-active surface and <strong>the</strong>n deposit <strong>the</strong> metal<br />

chemically on <strong>the</strong> carbon [6].<br />

II. EXPERIMENTAL<br />

A. INFILTRATING CARBON SKELETONS<br />

Clustering <strong>of</strong> <strong>the</strong> nanoparticles is a mayor problem<br />

when using a high volume percent (>20%) <strong>of</strong><br />

CNT/CNF as a filler material in <strong>the</strong> composite. This<br />

can be overcome by using carbon preforms and infiltrate<br />

<strong>the</strong>se skeletons by <strong>the</strong> copper matrix. Wetting<br />

and infiltration experiments were performed in a high<br />

temperature sessile drop device (HTSDD) [7] and a<br />

high vacuum sintering furnace (ASTRO APF-0716-<br />

MM) at 10 -3 Pa.<br />

Figure 1: Setup HTSDD (left) and ASTRO (right)<br />

Figure 1 shows <strong>the</strong> setups were <strong>the</strong> preform is topped<br />

with <strong>the</strong> metal and heated up to <strong>the</strong> metal melting<br />

temperature. In <strong>the</strong> sintering furnace external<br />

mechanical pressure (up to 80kPa) was applied to<br />

enhance <strong>the</strong> infiltration.


B. COPPER PLATING OF CNF<br />

Coating <strong>the</strong> CNF with copper with subsequent hot<br />

pressing is ano<strong>the</strong>r successful way to receive a<br />

composite with homogeneous dispersed filler<br />

material. The technique chosen is electroless<br />

chemical copper plating depositing copper from a<br />

CuSO4 solution [6]. Pretreatment including activation<br />

and sensitization <strong>of</strong> <strong>the</strong> CNF was studied. The plating<br />

process was enhanced by ultrasonic agitation in a<br />

conventional cleaning device (Branson 5510).<br />

III. RESULTS AND DISCUSSION<br />

The preforms investigated were produced by Future<br />

Carbon GmbH [9] in form <strong>of</strong> CNT foams, CNF felts<br />

and CNT papers (BuckyPaper).<br />

Infiltration depths up to 200 m were observed for<br />

alloys containing carbide forming elements like Ti,<br />

Cr, B and Mo. No conclusive differences between<br />

various pre-treatments (<strong>the</strong>rmal annealing, surface<br />

modification in Boric acid) <strong>of</strong> <strong>the</strong> skeletons were<br />

found. In <strong>the</strong> infiltration zone <strong>the</strong> carbon particles are<br />

gradual consumed by carbide formation during <strong>the</strong><br />

process. Figure 2 shows <strong>the</strong> infiltration zone <strong>of</strong> an<br />

infiltrated CNF felt with clearly visible carbon nan<strong>of</strong>iber<br />

structures.<br />

Figure 2: CNF infiltrated by a Cu-Ti alloy. The arrow<br />

indicates <strong>the</strong> gradient <strong>of</strong> carbide formation.<br />

Good infiltration results were also obtained for CNT<br />

skeletons, both foams and BuckyPaper, when<br />

infiltrating with Cu-Ti and Cu-B alloys. The<br />

significantly tinier structures <strong>of</strong> nanotubes seem not<br />

to hold very stable positions within <strong>the</strong> skeleton<br />

during <strong>the</strong> infiltration process.<br />

Goal<br />

Figure 3: Cu-B infiltrated CNT foam, microstructure.<br />

180<br />

The metal meanders through <strong>the</strong> preform, but tiny<br />

CNT clews remain within <strong>the</strong> infiltration zone <strong>of</strong><br />

several hundred m.<br />

The future goal is a complete infiltration also <strong>of</strong> <strong>the</strong>se<br />

agglomerates. Figure 3 shows <strong>the</strong> successful CNT<br />

foam infiltration by a Cu-B alloy and a sketch <strong>of</strong> <strong>the</strong><br />

CNT microstructure still to be infiltrated.<br />

Figure 4: Copper coated CNF, SEM image.<br />

The o<strong>the</strong>r approach followed is copper coating <strong>of</strong> <strong>the</strong><br />

CNFs and subsequent hot pressing. The deposited Cu<br />

keeps <strong>the</strong> CNF in distance which improves <strong>the</strong> particle<br />

dispersion in <strong>the</strong> composite. Homogeneous coating<br />

for a composite with 40 vol.% CNF succeeded.<br />

Figure 4 shows a SEM image <strong>of</strong> Cu coated CNF.<br />

IV. OUTLOOK<br />

Future investigations will include optimization <strong>of</strong> <strong>the</strong><br />

carbide interlayer to exploit <strong>the</strong> <strong>the</strong>oretically<br />

predicted composite limitations.<br />

ACKNOWLEDGMENTS<br />

This work was supported by <strong>the</strong> Austrian Research<br />

Centers Seibersdorf and <strong>the</strong> Center for Innovative<br />

Products at <strong>the</strong> Pennsylvania State University.<br />

Dr. Neubauer and Pr<strong>of</strong>. Smid are kindly acknowledged<br />

REFERENCES<br />

[1] N. Eustathopoulos: Acta Metallurgica Vol.46,<br />

No.7 (1998) 2319-2327<br />

[2] H. Weidmueller, T. Weissgarber et al.: Materials<br />

Science Forum Vols. 534–536 (2007) 853–856<br />

[3] R.B. Barclay: Journal <strong>of</strong> Mat. Sci. 6 (1971) 1076<br />

[4] Naidich, Yu and Kolesnichenko G.A.: Russian<br />

Metallurgy (Metally) No.4 (1968) 141<br />

[5] R. Voitovitch, A. Mortensen and F. Hodaj: Acta<br />

Metallurgica Vol.47, No.4 (1999) 1117-1128<br />

[6] L.M. Ang et al.: Carbon 38 (2000) 363-372<br />

[7] B. Schwarz, C. Eisenmenger-Sittner, H. Steiner:<br />

Vacuum, vol 82 issue 2 (<strong>2008</strong>) 186-188<br />

[8] Future Carbon GmbH, Bayreuth, Germany


Substitution <strong>of</strong> fossil fuels by using low temperature pyrolysis <strong>of</strong><br />

agricultural residues in a <strong>the</strong>rmal power plant<br />

Michael Halwachs and Hermann H<strong>of</strong>bauer (Faculty Mentor)<br />

Institute <strong>of</strong> Chemical Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: halwachs@mail.zserv.tuwien.ac.at<br />

Abstract — Externally heated rotary kiln pyrolysis<br />

reactor is used as a new process technology for <strong>the</strong><br />

conversion <strong>of</strong> biomass into useful primary energy<br />

products. A 3 MW pyrolysis pilot plant is presently<br />

being investigated using agricultural residues<br />

(straw as <strong>the</strong> primary feedstock). Several analytical<br />

methods are applied to provide an insight into <strong>the</strong><br />

complex process <strong>of</strong> pyrolysis. Fundamentals for an<br />

advanced pyrolysis model approach will be obtained<br />

by <strong>the</strong> results <strong>of</strong> <strong>the</strong> pilot plant.<br />

I. INTRODUCTION<br />

The global warming due to <strong>the</strong> increasing CO2 emission<br />

and dependency on fossil fuels, as well as <strong>the</strong><br />

high-energy price level have resulted in an increasing<br />

demand in renewable energy source. Biomass, as a<br />

renewable energy sources, has <strong>the</strong> potential to contribute<br />

to <strong>the</strong> future energy mix in many countries.<br />

The desire <strong>of</strong> <strong>the</strong> conversion <strong>of</strong> biomass is to generate<br />

useful primary products. The choice <strong>of</strong> <strong>the</strong> conversion<br />

process depends upon <strong>the</strong> type and quantity<br />

<strong>of</strong> biomass feedstock, <strong>the</strong> desired form <strong>of</strong> energy, end<br />

use requirements, environmental standards and economic<br />

conditions. The two main processes to convert<br />

biomass into useful products are <strong>the</strong>rmo-chemical<br />

processes and bio-chemical processes. Thermochemical<br />

processes include combustion, gasification,<br />

liquefaction, hydrogenation and pyrolysis. Biochemical<br />

processes are aerobe and anaerobe decomposition,<br />

which include <strong>the</strong> alcoholic and methane<br />

fermentation.<br />

II. PRINCIPLE OF PYROLYSIS<br />

A. GENERAL ASPECTS<br />

Pyrolysis is a process to convert biomass directly into<br />

solid, liquid and gaseous products by <strong>the</strong>rmal decomposition<br />

in absence <strong>of</strong> oxygen. Pyrolysis is a<br />

very complex process. Many different reactions take<br />

place and can be influenced by numerous factors.<br />

The influencing parameters are chemical or physical<br />

pretreatment <strong>of</strong> raw material, heating rate, reactor<br />

type, pyrolysis temperature, residence time, pyrolysis<br />

181<br />

atmosphere and particle size [1]. Biomass pyrolysis<br />

can be divided into four different stages: moisture<br />

evolution, hemicellulose decomposition, cellulose<br />

decomposition and lignin decomposition. [2] Pyrolysis<br />

is always a step in combustion and gasification<br />

processes where it is followed by total or partial<br />

oxidation <strong>of</strong> <strong>the</strong> primary products. An advantage <strong>of</strong><br />

<strong>the</strong> pyrolysis process is that pyrolysis products are<br />

easily storable and transportable.<br />

Figure 1: Typical yields <strong>of</strong> organic liquid, reaction<br />

water, gas and char from fast pyrolysis <strong>of</strong> wood,<br />

wt% on dry feed basis [3]<br />

B. REACTION KINETICS<br />

Pyrolysis <strong>of</strong> ligno-cellulosic biomass is a very complex<br />

process <strong>of</strong> interdependent reactions. Never<strong>the</strong>less,<br />

it can be reduced to a simple kinetic model,<br />

which is shown in Figure 2.<br />

Figure 2: Simple kinetic model for biomass pyrolysis<br />

[4]


The yield <strong>of</strong> main products: liquid, char and gas<br />

depends especially on <strong>the</strong> pyrolysis temperature. At<br />

high temperatures liquid products are obtained, lower<br />

temperatures favour high char yields (see Figure 1).<br />

III. PILOT PLANT AND EXPERIMENTAL<br />

WORK<br />

The externally heated rotary kiln pyrolysis reactor in<br />

Dürnrohr is an innovative process technology which<br />

can also be used for high capacities. The design fuel<br />

power is about 3 MW, <strong>the</strong> pyrolyis gas capacity is<br />

about 1.5 MW. Approximately 0.6 t/h straw can be<br />

processed in <strong>the</strong> rotary kiln. The combustible straw is<br />

characterized by a high ash fraction and his culm<br />

shape. The process is operated at low temperatures<br />

(350 to 630 °C) to prevent an entry <strong>of</strong> corrosive ash<br />

elements (K,Cl,S, etc.) and additional emissons in <strong>the</strong><br />

steam boiler <strong>of</strong> <strong>the</strong> coal fired power plant. An energetic<br />

use <strong>of</strong> <strong>the</strong> pyrolysis-charcoal (approximately<br />

40-50% <strong>of</strong> <strong>the</strong> original heating value) occurs separately<br />

in a fluidised bed reactor. Several analytical<br />

methods are used to get more insight about <strong>the</strong> behaviour<br />

during pyrolysis. The main compounds <strong>of</strong><br />

<strong>the</strong> pyrolysis gas are measured with a permanent gas<br />

measurement system. For <strong>the</strong> analytical determination<br />

<strong>of</strong> water and organic liquids in <strong>the</strong> pyrolysis gas<br />

phase two complementary methods are used: Gravimetric<br />

determination and GC-MS determination.<br />

First test runs were carried out at pyrolysis gas temperatures<br />

ranging from 600 to 630 °C.<br />

IV. RESULTS AND DISCUSSION<br />

Organic liquids can be classified due to <strong>the</strong> viscosity.<br />

Thereby liquids differ in low viscous pyrolysis oils<br />

and high viscous tar.<br />

Figure 3: Typical chromatogram (GC-MS) from<br />

<strong>the</strong> pyrolysis gas<br />

182<br />

The components are a mixture <strong>of</strong> alcohols, furans,<br />

aldehydes, esters, phenols, organic acids and oligomer<br />

carbohydrate and lignin products. The composition<br />

depends on <strong>the</strong> raw material and <strong>the</strong> pyrolysis<br />

process. The pyrolysis gas mainly contains H2, CO2,<br />

CO, CH4, trace amounts <strong>of</strong> larger gaseous organics<br />

compounds and water vapor. The water content results<br />

from <strong>the</strong> biomass humidity, about 10 %, and<br />

from <strong>the</strong> reaction water. To control <strong>the</strong> water content,<br />

<strong>the</strong> water content <strong>of</strong> <strong>the</strong> used biomass should be<br />

lower than 10 wt.%. The charcoal consists <strong>of</strong> carbon,<br />

volatile components, ash and partly tar. The content<br />

<strong>of</strong> carbon is one <strong>of</strong> <strong>the</strong> most important parameter for<br />

<strong>the</strong> quality <strong>of</strong> <strong>the</strong> charcoal, as well as <strong>the</strong> water content,<br />

ash content, elemental analysis, particle size,<br />

and energy density. All <strong>of</strong> <strong>the</strong> quality characteristics<br />

depend on <strong>the</strong> used biomass and <strong>the</strong> pyrolysis process<br />

as well as <strong>the</strong> purpose <strong>of</strong> use.<br />

Straw Charcoal<br />

C 42.85 68.80<br />

H 5.30 1.93<br />

N 0.44 0.73<br />

Natrium (water soluble) 0.005 0.090<br />

Kalium (water soluble) 1.07 3.75<br />

Volatile Compounds 70.97 15.25<br />

Gross Heating Value (Ho) 17111 25332<br />

Table 1: Typical straw and charcoal composition<br />

V. CONCLUSION<br />

The findings <strong>of</strong> <strong>the</strong> pilot plant will deliver fundamentals<br />

for <strong>the</strong> development <strong>of</strong> an advanced pyrolysis<br />

model. Fur<strong>the</strong>rmore, <strong>the</strong> results will be <strong>the</strong> basis for a<br />

scale up to a 30 MW capacity. For an improved description<br />

<strong>of</strong> this highly complex process it is necessary<br />

to carry out more runs in <strong>the</strong> pilot plant in <strong>the</strong><br />

future.<br />

REFERENCES<br />

[1] Pütün, A. E. (2007). Comparision between <strong>the</strong><br />

"slow" and <strong>the</strong> "fast" pyrolysis <strong>of</strong> tabacco<br />

residue. Industrial Crops and Products (26),<br />

307-314.<br />

[2] Yang, H., et al. (2007). Charateristics <strong>of</strong><br />

hemicellulose, cellulose and lignin pyrolysis.<br />

Fuel (86), 1781-1788.<br />

[3] T<strong>of</strong>t A.J. (1996) A Comparison <strong>of</strong> Integrated<br />

Biomass to Electricity Systems, PhD Thesis,<br />

Aston University, Birmingham, UK<br />

[4] Kaltschmitt, M., & Hartmann, H. (2001). Energie<br />

aus Biomasse. Springer Verlag


Long Wave Instabilities in Periodic Structures<br />

Melanie Todt and Thomas Daxner (Faculty Mentor)<br />

Institute <strong>of</strong> Lightweight Design and Structural Biomechanics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: mt@ilsb.tuwien.ac.at<br />

Abstract — The aim <strong>of</strong> <strong>the</strong> present work is <strong>the</strong> implementation<br />

<strong>of</strong> an efficient method for detecting <strong>the</strong> onset <strong>of</strong> buckling in spatially<br />

periodic structures loaded under compression. All calculations<br />

made are based on <strong>the</strong> Bloch Wave method which allows<br />

for detecting <strong>the</strong> critical state <strong>of</strong> <strong>the</strong> structure by using a single,<br />

geometrically representative cell. The Bloch Wave method is<br />

applied to different periodic structures, viz., an open cell foam,<br />

a closed cell foam, and a two-dimensional periodic lattice.<br />

I. INTRODUCTION<br />

Materials with a periodic microstructure, like composite<br />

materials or cellular solids, are widely used in engineering<br />

applications. Loaded under compression <strong>the</strong> onset<br />

<strong>of</strong> instability <strong>of</strong> <strong>the</strong>se solids may be related to a buckling<br />

type instability. For large structures <strong>the</strong> evaluation<br />

<strong>of</strong> <strong>the</strong> critical load and <strong>the</strong> prediction <strong>of</strong> <strong>the</strong> corresponding<br />

buckling mode can be a numerically expensive task.<br />

Exploiting <strong>the</strong> periodicity <strong>of</strong> <strong>the</strong> structure can reduce <strong>the</strong><br />

numerical requirements significantly.<br />

II. STABILITY CRITERIA<br />

Two different criteria are used for evaluating <strong>the</strong> critical<br />

state <strong>of</strong> periodic structures, namely a microscopic stability<br />

criterion and a macroscopic stability criterion. In both<br />

cases only a single unit cell, which can be as small as <strong>the</strong><br />

smallest periodic unit <strong>of</strong> <strong>the</strong> structure, is used for predicting<br />

<strong>the</strong> critical state. The general <strong>the</strong>ory for both concepts<br />

can be found in [1].<br />

A. MICROSCOPIC STABILITY CRITERION<br />

The microscopic stability criterion is based on <strong>the</strong> fact<br />

that <strong>the</strong> buckling eigenmode <strong>of</strong> a periodic structure can<br />

be expressed in terms <strong>of</strong> <strong>the</strong> displacement field:<br />

δu ∼ (X ∼ )=U ∼ (X ∼ )exp[i( m1<br />

h1<br />

X + m2<br />

h2<br />

Y + m3<br />

Z)], (1)<br />

h3<br />

where hi are <strong>the</strong> unit cell dimensions, and U is a peri-<br />

∼<br />

odic function with <strong>the</strong> same periodicity as <strong>the</strong> unit cell.<br />

Equation (1) is <strong>the</strong> so called Bloch Wave representation<br />

<strong>of</strong> <strong>the</strong> buckling eigenmode. The incremental equilibrium<br />

condition can be written as:<br />

⎡<br />

⎣<br />

K ≈ 11 K ≈ 12 K ≈ 1I<br />

K ≈ 21 K ≈ 22 K ≈ 2I<br />

K ≈ I1 K ≈ I2 K ≈ II<br />

⎤ ⎛<br />

⎦ ⎝<br />

Δu ∼ 1<br />

Δu ∼ 2<br />

Δu ∼ I<br />

⎞ ⎛<br />

⎠ = ⎝<br />

ΔF ∼ 1<br />

ΔF ∼ 2<br />

ΔF ∼ I<br />

⎞<br />

⎠<br />

(2)<br />

183<br />

with Δu ∼ 1 and Δu ∼ 2 being <strong>the</strong> increments <strong>of</strong> <strong>the</strong> nodal displacement<br />

vectors <strong>of</strong> <strong>the</strong> Bloch Wave (BW) master nodes<br />

and <strong>of</strong> <strong>the</strong> BW slave nodes, respectively. The BW master<br />

and slave nodes are located on <strong>the</strong> outer boundaries <strong>of</strong> <strong>the</strong><br />

cell and sit on opposite sides <strong>of</strong> <strong>the</strong> cell. Δu ∼ I is <strong>the</strong> increment<br />

<strong>of</strong> <strong>the</strong> nodal displacement vector <strong>of</strong> <strong>the</strong> internal<br />

nodes <strong>of</strong> <strong>the</strong> cell. The partition <strong>of</strong> <strong>the</strong> tangent stiffness<br />

matrix K ≈ is implied by <strong>the</strong> partition <strong>of</strong> <strong>the</strong> increment Δu ∼<br />

<strong>of</strong> <strong>the</strong> nodal displacement vector. The internal degrees <strong>of</strong><br />

freedom can be condensed out and Δu ∼ 1 and Δu ∼ 2 can be<br />

coupled using Equation (2). The static stability criterion<br />

can now be expressed as:<br />

˚K ≈ δu ∼ 1 = 0 ∼<br />

(3)<br />

involving only <strong>the</strong> part <strong>of</strong> <strong>the</strong> buckling eigenmode corresponding<br />

to <strong>the</strong> BW master nodes δu ∼ 1. ˚ K ≈ is <strong>the</strong> reduced<br />

stiffness matrix. For details see, e.g., [1], [2], or [3].<br />

B. MACROSCOPIC STABILITY CRITERION<br />

The macroscopic loss <strong>of</strong> stability is indicated by <strong>the</strong> loss<br />

<strong>of</strong> rank one convexity <strong>of</strong> <strong>the</strong> homogenized tangent moduli<br />

tensor LH ijkl <strong>of</strong> <strong>the</strong> single unit cell:<br />

with<br />

det|L H ijkl(Λ)njnl| =0 (4)<br />

n ∼ = {cos(ϕ), sin(ϕ)cos(ψ), sin(ϕ)sin(ψ)} T<br />

0 ≤ ϕ ≤ π 0 ≤ ψ ≤ 2π (5)<br />

for <strong>the</strong> 3D case. The buckling eigenmode has infinite<br />

wavelength and its band direction is given by n ∼ see, e.g.,<br />

[4].<br />

III. OPEN CELL AND CLOSED CELL FOAM<br />

The open cell foam is modeled on <strong>the</strong> basis <strong>of</strong> <strong>the</strong> spacefilling<br />

Kelvin cell depicted in Figure 1 (left) which is<br />

assigned geometric properties reported in <strong>the</strong> literature,<br />

e.g., [2], [5], and [6]. The cell is elongated in x-direction<br />

(rise direction), see Figure 1, and subjected to uniaxial<br />

loading and a set <strong>of</strong> triaxial loading cases.<br />

The critical stresses and <strong>the</strong> corresponding buckling<br />

modes depend on <strong>the</strong> number <strong>of</strong> base cells that form<br />

<strong>the</strong> periodic structure and <strong>the</strong> multiaxiality <strong>of</strong> <strong>the</strong> loading<br />

case. For triaxial loading cases close to pure hydrostatic<br />

pressure <strong>the</strong> corresponding buckling mode is independent<br />

<strong>of</strong> <strong>the</strong> number <strong>of</strong> base cells and corresponds to a buckling


Figure 1: Unit cells representing a open cell foam (left) and a<br />

closed cell foam (right).<br />

mode local to a single unit cell. The results obtained for<br />

<strong>the</strong> Kelvin cell are in good agreement with <strong>the</strong> results reported<br />

in [2].<br />

Specifically, <strong>the</strong> results received for uniaxial loading<br />

in transverse direction (y-direction) are depicted in Figure<br />

2. Both stability criteria are used to predict <strong>the</strong> critical<br />

stress and <strong>the</strong> corresponding buckling mode. For<br />

columns consisting <strong>of</strong> up to 60 base cells <strong>the</strong> critical<br />

stress and <strong>the</strong> corresponding buckling mode, which is<br />

local to a single cell, are independent <strong>of</strong> <strong>the</strong> number <strong>of</strong><br />

base cells. For configurations consisting <strong>of</strong> more than 60<br />

cells <strong>the</strong> critical stress drops asymptotically towards <strong>the</strong><br />

critical stress σ H 1c predicted by <strong>the</strong> macroscopic stability<br />

criterion as <strong>the</strong> mode wavelength tends towards infinity.<br />

Figure 2: Results <strong>of</strong> <strong>the</strong> micro- and macroscopic stability criteria<br />

in transverse direction (y- orz-direction).<br />

The closed cell model is based on <strong>the</strong> space-filling<br />

Weaire-Phelan cell shown in Figure 1 (right), see, e.g.,<br />

[7]. The cell was subjected to uniaxial loading and pure<br />

hydrostatic pressure. For both cases <strong>the</strong> critical buckling<br />

mode is local to a single unit cell and, like <strong>the</strong> critical<br />

stress, independent <strong>of</strong> <strong>the</strong> number <strong>of</strong> base cells that form<br />

<strong>the</strong> periodic structure. For multi-cell models consisting<br />

<strong>of</strong> up to 4 × 4 × 4 base cells it is possible to show that<br />

local cell wall buckling is <strong>the</strong> dominant buckling mode,<br />

as predicted by <strong>the</strong> Bloch Wave method.<br />

IV. TWO-DIMENSIONAL PERIODIC LATTICE<br />

A square two-dimensional periodic lattice is modeled using<br />

four different unit cells. These four unit cells are sub-<br />

184<br />

jected to uniaxial and bi-axial compressive loads. The<br />

critical stresses and <strong>the</strong> corresponding buckling modes<br />

depend on <strong>the</strong> size <strong>of</strong> <strong>the</strong> structure (number <strong>of</strong> base cells<br />

in each direction) and on <strong>the</strong> multiaxiality <strong>of</strong> <strong>the</strong> loading.<br />

The results received for all four unit cells are in good<br />

agreement with and confirmed by accompanying eigenvalue<br />

analyses <strong>of</strong> Finite Element models containing multiple<br />

base cells.<br />

V. SUMMARY<br />

By analyzing various regular periodic structures, it was<br />

possible to demonstrate how <strong>the</strong> Bloch Wave method<br />

helps to reduce significantly <strong>the</strong> numerical effort for predicting<br />

buckling in periodic structures. While opencell<br />

topologies showed buckling modes <strong>of</strong> infinite wavelength,<br />

<strong>the</strong> examined closed-cell Weaire-Phelan structure<br />

showed no signs <strong>of</strong> long-wavelength buckling.<br />

REFERENCES<br />

[1] N. Triantafyllidis, M. D. Nestorović, and M. W.<br />

Schraad. Failure surfaces for finitely strained twophase<br />

periodic solids under general in-plane loading.<br />

Journal <strong>of</strong> Applied Mechanics, 73(3):505–515, May<br />

2006.<br />

[2] L. Gong, S. Kyriakides, and N. Triantafyllidis. On<br />

<strong>the</strong> stability <strong>of</strong> Kelvin cell foams under compressive<br />

loads. Journal <strong>of</strong> <strong>the</strong> Mechanics and Physics<br />

<strong>of</strong> Solids, 53(4):771–794, April 2005.<br />

[3] M. Todt. Long wave instabilities in periodic structures.<br />

Diploma <strong>the</strong>sis, Vienna University <strong>of</strong> Technology,<br />

Vienna, <strong>2008</strong>.<br />

[4] G. Geymonat, S. Müller, and N. Triantafyllidis. Homogenization<br />

<strong>of</strong> nonlinearly elastic materials, microscopic<br />

bifurcation and macroscopic loss <strong>of</strong> rank-one<br />

convexity. Archive for Rational Mechanics and Analysis,<br />

122:231–290, 1993.<br />

[5] L. Gong, S. Kyriakides, and W.-Y. Jang. Compressive<br />

response <strong>of</strong> open cell foams. Part I: Morphology<br />

and elastic properties. International Journal <strong>of</strong> Solids<br />

and Structures, 42(5-6):1355–1379, March 2005.<br />

[6] L. Gong and S. Kyriakides. Compressive response <strong>of</strong><br />

open cell foams. Part II: Initiation and evolution <strong>of</strong><br />

crushing. International Journal <strong>of</strong> Solids and Structures,<br />

42(5-6):1381–1399, March 2005.<br />

[7] T. Daxner, R.D. Bitsche, and H.J. Böhm. Spacefilling<br />

polyhedra as mechanical models for solidified<br />

dry foams. Materials Transactions, 47(9):2213–<br />

2218, 2006.


Determination <strong>of</strong> <strong>the</strong> stoichiometry <strong>of</strong> complexes <strong>of</strong> trivalent<br />

lanthanides and extraction-relevant ligands with nanoelectrospray<br />

ionization time-<strong>of</strong>-flight mass-spectrometry<br />

Michael Steppert, Clemens Wal<strong>the</strong>r, Andreas Geist, Horst Geckeis, Thomas Fanghänel<br />

Institute for Nuclear Waste Disposal<br />

Forschungszentrum Karlsruhe<br />

76021 Karlsruhe, Germany<br />

Email: michael.steppert@ine.fzk.de<br />

Abstract — The minor actinides (neptunium, americium<br />

and curium) and plutonium control <strong>the</strong><br />

long-term radiotoxicity <strong>of</strong> spent nuclear fuel. While<br />

direct disposal <strong>of</strong> this waste requires repositories<br />

which are save for >100.000 years, partitioning <strong>of</strong><br />

<strong>the</strong> long lived products and <strong>the</strong>ir subsequent transmutation<br />

to short lived or stable nuclides reduces<br />

storage time scales to only some hundreds <strong>of</strong> year:<br />

plutonium separation is already performed on <strong>the</strong><br />

industrial scale by <strong>the</strong> PUREX process. However,<br />

efficient transmutation requires also partitioning <strong>of</strong><br />

<strong>the</strong> trivalent actinides Am and Cm from lanthanide<br />

fission products, which is not yet feasible in large<br />

scale. The present work focuses investigations on a<br />

possible candidate for An(III)/Ln(III) separation:<br />

bis-triazinyl-pyridine (BTP). By means <strong>of</strong> nanoelectrospray<br />

mass-spectrometry complex formation <strong>of</strong><br />

BTP with selected elements <strong>of</strong> <strong>the</strong> lanthanide series<br />

is investigated<br />

I. INTRODUCTION<br />

Besides uranium, spent nuclear fuel contains plutonium,<br />

minor actinides (MA, neptunium, americium,<br />

curium), and fission products (FP). While<br />

most <strong>of</strong> <strong>the</strong> fission products decay within <strong>the</strong> first<br />

100 years, <strong>the</strong> radiotoxicity <strong>of</strong> spent nuclear fuel is<br />

dominated by <strong>the</strong> contributions <strong>of</strong> Pu and <strong>the</strong> MA<br />

at time scales exceeding 300y (see Fig. 1). Uranium<br />

and plutonium are separated from <strong>the</strong> spent<br />

fuel by liquid-liquid extraction in <strong>the</strong> PUREX<br />

process. In <strong>the</strong> subsequent DIAMEX process trivalent<br />

lanthanides and actinides [1] are coextracted.<br />

However, for an efficient transmutation <strong>the</strong> lanthanides<br />

must be separated from <strong>the</strong> chemically<br />

very similar trivalent actinides Am and Cm. To<br />

<strong>the</strong>se ends, <strong>the</strong> organic complexing agent bistriazinyl-pyridine<br />

(BTP) was developed [2]. Distribution<br />

ratios between organic and aqueous<br />

phase using this highly selective ligand are depicted<br />

in Figure 2. Though BTP works well in <strong>the</strong><br />

laboratory and has passed already first upscaling<br />

tests [3] <strong>the</strong> origin <strong>of</strong> its exceptionally high<br />

185<br />

Figure 1: Radiotoxicity, i.e. <strong>the</strong> measure <strong>of</strong> <strong>the</strong><br />

cumulative negative health effects <strong>of</strong> radionuclides<br />

due to ionizing radiation after ingestion, <strong>of</strong> 1t <strong>of</strong><br />

spent nuclear fuel after discharge from a pressurized<br />

light water reactor (40GW/t)<br />

selectivity towards trivalent actinides is not yet<br />

understood.<br />

For a better understanding, <strong>the</strong> interaction <strong>of</strong> BTP<br />

with trivalent lanthanides was studied with nanoelectrospray-ionization<br />

mass spectrometry.<br />

Figure 2: Distribution coefficients for extraction to<br />

<strong>the</strong> organic phase with BTP-ligand


With <strong>the</strong> nano-electrospray method, ions that are<br />

present in a solution can be transferred to <strong>the</strong> gas<br />

phase almost non invasively (so called “s<strong>of</strong>t ionisation”)<br />

[4][5] avoiding fragmentation reactions <strong>of</strong> <strong>the</strong><br />

ions. These s<strong>of</strong>t ionization conditions make <strong>the</strong> nanoelectrospray<br />

ionization mass spectrometry <strong>the</strong> ideal<br />

method to determine <strong>the</strong> stoichiometry <strong>of</strong> complexes.<br />

The low invasiveness <strong>of</strong> measurements with <strong>the</strong> used<br />

ALBATROS ESI-TOF spectrometer could be shown<br />

in <strong>the</strong> investigation <strong>of</strong> polynuclear Zr(IV) hydroxide<br />

complexes in [6].<br />

II. EXPERIMENTAL<br />

Solutions <strong>of</strong> lanthanide nitrates (5 x 10 -5 M) and<br />

BTP with ligand-to-metal ratios <strong>of</strong> 2 in octanol<br />

were prepared for each ESI-measurement with <strong>the</strong><br />

ALBATROS ESI-TOF mass spectrometer [7]-[9].<br />

III. RESULTS<br />

The ESI measurements for <strong>the</strong> interaction <strong>of</strong> <strong>the</strong><br />

lanthanide nitrates with BTP in octanol show that <strong>the</strong><br />

relative abundances <strong>of</strong> <strong>the</strong> complexes [Ln(BTP)-<br />

(NO3)2 1+ ], [Ln(BTP)2(NO3)2 + ] and [Ln(BTP)3-<br />

(NO3) 2+ ] are very different for Ln= La, Pr, Eu, Tb<br />

and Lu, respectively. Figure 3 shows <strong>the</strong> ratios <strong>of</strong> <strong>the</strong><br />

different detected species <strong>of</strong> <strong>the</strong> observed lanthanides<br />

normalized to 100% for [Ln(BTP)(NO3)2 1+ ],<br />

[Ln(BTP)2(NO3)2 + ] and [Ln(BTP)3(NO3) 2+ ].<br />

IV. DISCUSSION<br />

The complexation <strong>of</strong> lanthanide ions by BTP was<br />

studied before by ESI-TOF [10]. However, while<br />

<strong>the</strong>se investigations were perfomed in water/methanol,<br />

mixtures <strong>the</strong> present work focuses on <strong>the</strong> diluent<br />

octanol. Since <strong>the</strong> diluent has a drastic influence on<br />

<strong>the</strong> speciation <strong>of</strong> <strong>the</strong> metal ligand complexes investigations<br />

must be performed in a diluent relevant to <strong>the</strong><br />

actual extraction process, in this case octanol. The<br />

relative abundances <strong>of</strong> <strong>the</strong> Ln(BTP)3-complexes <strong>of</strong><br />

<strong>the</strong> different lanthanides (Figure 3) correlate with <strong>the</strong><br />

distribution ratios (Figure 2): We found that a high<br />

distribution ratios, i.e. an efficient extraction from <strong>the</strong><br />

aqueous into <strong>the</strong> organic phase correlates with a high<br />

abundance <strong>of</strong> <strong>the</strong> respective Ln(BTP)3-complex (e.g.<br />

for <strong>the</strong> case <strong>of</strong> Tb). In contrast, formation <strong>of</strong><br />

Ln(BTP)(1-2)-complexes result in lower distribution<br />

ratios and high separation factors from <strong>the</strong> actinides<br />

Cm and Am. Both, electrostatic and steric effects<br />

seem to play a role in <strong>the</strong> formation <strong>of</strong> <strong>the</strong> different<br />

complexes: For <strong>the</strong> lanthanides La through Tb <strong>the</strong><br />

decreasing ionic radius causes an increasing electrostatic<br />

attraction and a stabilization <strong>of</strong> <strong>the</strong> 1:3 complex.<br />

Fur<strong>the</strong>r contraction <strong>of</strong> <strong>the</strong> ions from Tb through<br />

186<br />

Lu favours formation <strong>of</strong> 1:2 and 1;1 complexes, most<br />

likely due to steric hindrance.<br />

Similar results as for <strong>the</strong> 1:2 metal to ligand ratio<br />

were obtained for <strong>the</strong> ligand-to-metal ratio 10. The<br />

better <strong>the</strong> lanthanide is extracted to <strong>the</strong> organic phase<br />

by BTP, <strong>the</strong> more Ln(BTP)3-complexes are detected.<br />

Figure 3: Ratios <strong>of</strong> <strong>the</strong> different detected species <strong>of</strong><br />

<strong>the</strong> observed lanthanides normalized to 100%. Top<br />

left: Structure <strong>of</strong> Ln(BTP)3-complexes.<br />

ACKNOWLEDGMENTS<br />

I would like to thank Pr<strong>of</strong>. Thomas Fanhänel for <strong>the</strong><br />

opportunity to work on this exciting topic. I thank<br />

Clemens Wal<strong>the</strong>r and Andreas Geist for supervising<br />

my work and for <strong>the</strong> helpful discussions. I am grateful<br />

for <strong>the</strong> technical support <strong>of</strong> Sebastian Büchner<br />

and <strong>the</strong> help <strong>of</strong> Markus Fuß through <strong>the</strong> first steps<br />

with <strong>the</strong> ESI device.<br />

REFERENCES<br />

[1] C. Madic et al. J. All. Com., 444-445, 23-27,<br />

2007.<br />

[2] Z. Kolarik et. al. Solvent. Extr. Ion. Exch., 17,<br />

5, 1155-1170, 1999.<br />

[4] J. Fenn. J. Biomol. Tech., 13,101–118, 2002.<br />

[3] C. Hill et al., J. Nucl. Sci. Technol., 3, 309-312,<br />

2002.<br />

[5] M.Wilm, M. Mann. Anal. Chem., 68, 1–8,<br />

1996.<br />

[6] C. Wal<strong>the</strong>r et al. Anal. Bioanal. Chem., 388, 409-<br />

431, April 2007.<br />

[7] T. Bergmann, T.P. Martin, H. Schaber. Rev. Sci.<br />

Instrum., 60, 347–349, 1989.<br />

[8] T. Bergmann, T.P. Martin, H. Schaber. Rev. Sci.<br />

Instrum., 61, 2592–2600, 1990.<br />

[9] T. Bergmann, et. al. Rev. Sci. Instrum., 61,<br />

2585–2591, 1990<br />

[10] S. Colette, et al. Inorg. Chem., 42, 2215-2226,<br />

2003.


Quantum Optical Coherence Tomography<br />

Abstract- Quantum entanglement, which is a<br />

unique phenomenon <strong>of</strong> Quantum Mechanics, is<br />

being used in many areas like Quantum cryptography,<br />

imaging and computing. In this project,<br />

which is <strong>the</strong> preliminary work for Quantum Optical<br />

Coherence Tomography (QOCT), we have used <strong>the</strong><br />

process <strong>of</strong> spontaneous parametric down conversion<br />

(SPDC) to generate entangled photon<br />

pairs. These newly created photons, separated by a<br />

polarizing beam splitter (PBS), are <strong>the</strong>n sent<br />

separately to a single photon detector via different<br />

paths. Using a LabView user interface s<strong>of</strong>tware, we<br />

recorded <strong>the</strong> amount <strong>of</strong> photons entering each<br />

detector over a period <strong>of</strong> time, along with a<br />

coincidence count between <strong>the</strong> two detectors. Using<br />

those data, we obtained a visibility curve, which<br />

indicates <strong>the</strong> degree <strong>of</strong> correlation between <strong>the</strong><br />

entangled photons. We have achieved a visibility <strong>of</strong><br />

94%, which allowed us to move on to <strong>the</strong> next step:<br />

QOCT.<br />

I. INTRODUCTION<br />

The field <strong>of</strong> Quantum Cryptography has<br />

received renewed interest during <strong>the</strong> past<br />

decade due to <strong>the</strong> advent <strong>of</strong> a new source<br />

<strong>of</strong> light. This source, made <strong>of</strong> entangled<br />

photon pairs allows for enhanced security<br />

and robustness <strong>of</strong> <strong>the</strong> system.<br />

This project is divided into two parts. First,<br />

we generate and characterize entangled<br />

photon pairs through parametric down<br />

conversion. Then we move on to QOCT.<br />

II. EXPERIMENTAL SETUP<br />

The diagram below is a simplified version<br />

<strong>of</strong> <strong>the</strong> actual experimental setup. The idea<br />

is to generate two beams and let <strong>the</strong>m<br />

propagate on different paths. A sample is<br />

PIERRE BOUZI<br />

Dr. ROGER DORSINVILLE<br />

Department <strong>of</strong> Electrical Engineering<br />

City College <strong>of</strong> New York<br />

New York, USA<br />

Email: pbouzi00@ccny.cuny.edu<br />

187<br />

placed in a fixed arm while <strong>the</strong> o<strong>the</strong>r path<br />

is varied over a small distance. When <strong>the</strong><br />

two path lengths are equal, we obtain an<br />

interference pattern which reveals <strong>the</strong><br />

image inside <strong>the</strong> sample.<br />

Figure1. Simplified QOCT setup<br />

The results we have obtained so far are<br />

show next.<br />

III. RESULTS<br />

We have obtained a visibility <strong>of</strong> 94%, as<br />

shown in <strong>the</strong> figure below.<br />

Coincidence counts in 30 sec<br />

laser<br />

source<br />

Prism<br />

Mirror<br />

PBS<br />

KTP<br />

V isib ility (~94 % )<br />

160<br />

140<br />

120<br />

100<br />

80<br />

60<br />

40<br />

20<br />

0<br />

0 50 100 150 200 250 300 350 400<br />

QWP angles in degree<br />

Detector<br />

Figure2. Degree <strong>of</strong> correlation between <strong>the</strong><br />

entangled photons<br />

Mirror


Such a curve, which indicates <strong>the</strong> degree<br />

<strong>of</strong> correlation between <strong>the</strong> photons, was<br />

obtained using <strong>the</strong> following formula:<br />

V= max c min c<br />

max c min c<br />

100 ,where <strong>the</strong> subscript<br />

c stands for coincidence count.<br />

With such a high visibility, we are ready to<br />

move on to <strong>the</strong> o<strong>the</strong>r part <strong>of</strong> <strong>the</strong><br />

experiment: QOCT.<br />

IV. CONCLUSION<br />

We have completed <strong>the</strong> first part <strong>of</strong> <strong>the</strong><br />

project successfully, as indicated by <strong>the</strong><br />

visibility plot. Now we are working on <strong>the</strong><br />

QOCT part, where we hope to match a<br />

simulation obtained at Boston University<br />

[1]. A similar technique already exists in<br />

188<br />

imaging, but by using entangled photon<br />

pairs it is possible to achieve a two fold<br />

increase in resolution.<br />

V. ACKNOWLEDGMENTS<br />

I would like to thank our sponsors at Corning<br />

Incorporated for <strong>the</strong>ir support, as well as my<br />

colleagues and my supervisor.<br />

VI. REFERENCES<br />

[1] Mark C. Booth, Giovanni Di Giuseppe,<br />

Bahaa E. A. Saleh, Alexander V.<br />

Sergienko, and Mavin C. Teich,<br />

“Polarization-sensitive quantum-optical<br />

coherence tomography”


����������������������������������������������������������<br />

����������������������������������������������������<br />

������������������������������������������������������������������������������������������������ ������������������������������������������������������������������������������� ����������������������������<br />

������������������������������������������������������������������������������������� ���������������������������������������������������������������������������������� �� ������������������������<br />

���������������������������������������������������� �<br />

��� � �������� ��������������������������� � ��� � ���<br />

���������� � ���������� � ��� � ��������� � ��������<br />

������������������������������������������������<br />

���� � ������� � �� � ����� � ������ � �������� � ��������<br />

��������������������������������������������������<br />

��� � ����������� ����������� � ������ � ��� �����<br />

�������������������������������������������������<br />

��������������������������������������������������<br />

��������������������������������������������������<br />

���������������������������������������������<br />

����������������������������������������������������<br />

���� � �� � ���� � �� � �������� � ������� � ����������� � ���<br />

�������� � ������������� � ��� � ���������� � �����<br />

���������������������������������������������������<br />

��� � ����������� � ����������� � ���� � �� � ����<br />

���������������������������������������������<br />

�� ������������<br />

��������������������������������� �������������<br />

���������� ��� ��� � ��������� � ����������������������<br />

�����������������������������������������������������<br />

��� � ��� � ���� � � � ����������� � �������� � ��������<br />

����������� � ����� � �� � ��� � ���� � ������ � ���������<br />

�������������������������������������������������������<br />

������ � ��������� � ������� � ����� � ������� � ��� � ���<br />

�������������������������<br />

���������������������������������������������������<br />

����������������������������������������������������<br />

������� � ������� � �� � ��� � ����������� � ������������<br />

���������� � ����� � ����� � �� � ����� � ������ � ����<br />

����������������������������������������������������<br />

�����������������������������������������������������<br />

�������������������������������������������������<br />

�������������������������������������������������<br />

�������� � �� � �� � ��������� � �� � �������� � �����<br />

��������������������������������������������������<br />

������������������������������<br />

������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������<br />

��� ��������������������<br />

��� ������������ ����� ���������� ��� ������� � ����<br />

������������������������������������������������<br />

�����������������������������������������������<br />

189<br />

���������<br />

���������<br />

�����<br />

������<br />

�������<br />

������<br />

�������<br />

������ �� � � �<br />

������ �� � � �<br />

����� �� � � �<br />

����� �� � � �<br />

������ �� � � �<br />

���������� �� � � �<br />

�������� �� � � �<br />

��������� �� � � �<br />

����������� �� � � �<br />

����� � � � �<br />

�������� ���� � � �<br />

�������<br />

����<br />

������<br />

�������<br />

������������� ����������������������������<br />

���������������������������������<br />

�<br />

�� �����������������������<br />

��� � ����� � ��� � ���� � ��������� � ���� � ���������<br />

�������� � ��� � ������� � ���� � ��������� � ���<br />

�������������� � � � ���������� � �� � ����� � ��������������<br />

����������������������������<br />

���������������������������������������������������<br />

���������������������������������������������������<br />

��������������������������������������<br />

�� ������������������������������������<br />

��������������������������������������������������<br />

��������� � ��� � ������� � ����� � ���� � ��� � �����<br />

���������� � ���� � �� � ���������� � ���������� � ������<br />

����������������������������������������������<br />

��������� � ����� � ������� � ����� � ���� � ������ � ���<br />

������������������������������������������������<br />

���������<br />

�� ��������������������������<br />

��������������������������������������������������<br />

��������������������������������������������������<br />

��������� � ��� � ��� � ���� � �� � ��������� � ��� � ����<br />

��������������������<br />

�������������������������������������������������<br />

������������������������������������������


��������������������<br />

�� � ������� � �������������� � ��� � ������� � �� � �����<br />

����� � �� � �������� � �� � �������� � �������� � �����<br />

������ � �� � ���������� � ���� � �������� � ������<br />

��������������������������������������<br />

���� �������<br />

�� ������������������������<br />

��������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������������������������<br />

������������������������������������������������<br />

������������������������������������������������<br />

���������������������������������������������������<br />

�� �����������������������������������<br />

��� � ���������� � ���� � ��������� � �� � �������� � ����<br />

������������������������������������������������<br />

��������������������������������������������������<br />

���������������������������������������������<br />

��������������������������<br />

�� ��������������������������<br />

������������ � �������������� � ��� � �������������<br />

�����������������������������������������������������<br />

������� � ���� ��� ����� � ��������������������� ���<br />

�������������������������<br />

������ � �� � ���������� � �� � ������ � �����������<br />

�������������������������������������������������<br />

����� � �� � ����� � ����� � ������ � ���� � ������� � �����<br />

����������������������������������������������������<br />

���������������<br />

��� �����������<br />

��������������������������������������������������<br />

��������������������������������������������������<br />

190<br />

�� � �������� � �� � ����� � ��������� � ���������� � ���<br />

�����������������������������������������������<br />

����������� � ��� � ������ � ������ � �� � ��� � ���������<br />

����� � �� � ���� � � � ����� � ���� � ��� � ��������� � ����<br />

�����������������������������������������������������<br />

��������������������������������������������������<br />

����� � ������������ � ���� � ���� � ����������� � ����<br />

���������� � �� � ������������ � ������� � �� � ��������<br />

����������������������������������������������������<br />

����������������������������������������������������<br />

����������������������������������������������<br />

������������������������������������������������<br />

�������������������������������������������������<br />

����������������������������<br />

��������������� ��������������������������������<br />

������������ � �� � ���� � ����� � ��� � �������� � ���<br />

���������� � ���� � ��� � ����������� � �� � ������ � ����� � ����<br />

����� � ��������� � ����������� � ��� � ���������� � ����<br />

��������� � ����������� � ��������� � � � ���� � �� � ������<br />

���������������������������������������������������<br />

�������������<br />

�� ������������ ���� � �������������������� � ���� �����<br />

��������� � �� � ����������� � ���� � ������ � ���<br />

�������������������������������������������������<br />

������������� � ������� � �� � ����� � � � ������ � ���<br />

����������� � ������� � ������ � ����� � ��� � ���� � ����<br />

������������������������������������������������<br />

��� � ���� � ��������� � ����������� � ����� � ���� � ����<br />

���������������������������������������������������<br />

����� � ��� � ���� � �� � �� � ��������� � �� � ��� � ����� � ���<br />

�����������������������������������������������������<br />

����������������������������������������������<br />

����������<br />

�������������� � �� � ������� � �� � �������� � ���<br />

������������������������������������������<br />

�������������������������������������������<br />

������������������������������ �������������<br />

������������������������������������<br />

��� ����������������� � �� � �� � �������� � �� � ���<br />

������������������������������������<br />

��������������������������������������<br />

�� � ������������� � �� � �������� � �� � ��������<br />

�����������������������������������������<br />

�� � �������� ����������� � ����� � ���� � ���<br />

�����������������������������������������<br />

��� ��� � ���������� � �� � �� � ���������� � �� � ���<br />

����������������������������������������<br />

��� � ������� � �� � ������� � ����������� � �����<br />

������������������������������������������<br />

����������������������������<br />

��� ��� � �� � ������ � �� � ����� � �� � �� � ��� � ������<br />

�������������������������������������������<br />

����������������������������������������<br />

����������������������������������� ������<br />

��������������������������


Functionalized Cadmium Sulfide Clusters as Nanosized Building<br />

Blocks for Hybrid Materials – Investigation <strong>of</strong> <strong>the</strong><br />

“[Cd10S4(SPh)12]” Cluster Structure<br />

Maria Bendova, Michael Puchberger and Ulrich Schubert (Faculty Mentor)<br />

Institute <strong>of</strong> Materials Chemistry<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: mbendova@mail.zserv.tuwien.ac.at<br />

Abstract — The goal <strong>of</strong> our work is to syn<strong>the</strong>size<br />

molecular CdS clusters with polymerizable ligands<br />

(“functional” clusters) in order to use <strong>the</strong>m as comonomers<br />

in polymerization reactions. In this way,<br />

cluster-crosslinked polymers might be formed which<br />

possess special physical, e.g. optoelectronic, properties.<br />

In this contribution, we present our results on <strong>the</strong><br />

investigation <strong>of</strong> “[Cd10S4(SPh)12]” (1) which is<br />

formed by <strong>the</strong>rmal treatment <strong>of</strong> <strong>the</strong> precursor<br />

(Me4N)4[Cd10S4(SPh)16] cluster. According to previous<br />

literature studies, it should have four free Cd<br />

coordination sites available for fur<strong>the</strong>r functionalization.<br />

However, we proved now that <strong>the</strong> core<br />

structure <strong>of</strong> <strong>the</strong> cluster is not maintained during <strong>the</strong><br />

pyrolysis reaction and that it grows unspecifically.<br />

Thus, it is not suitable for fur<strong>the</strong>r functionalization.<br />

I. INTRODUCTION<br />

Molecular semiconductor clusters with a defined<br />

number <strong>of</strong> atoms and a defined molecular structure<br />

are <strong>of</strong> particular interest for investigations on <strong>the</strong><br />

variation <strong>of</strong> properties related to <strong>the</strong> change in size.<br />

There is a series <strong>of</strong> defined molecular CdS clusters<br />

with increasing size. The two smallest clusters<br />

([Cd10S4(SPh)16] 4- [1] and [Cd17S4(SPh)28] 2- [2]) are<br />

syn<strong>the</strong>sized via kinetic trapping method, using powdered<br />

S or a S 2- salt as <strong>the</strong> sulfur source. The cluster<br />

with 32 Cd atoms ([Cd32S14(SPh)40] 4- ) can be prepared<br />

in different ways: (i) by solvo<strong>the</strong>rmal syn<strong>the</strong>sis<br />

with thiourea as <strong>the</strong> sulfur source [3], or (ii) by solidstate<br />

formation <strong>of</strong> “[Cd10S4(SPh)12]” (1) from<br />

[Cd10S4(SPh)16] 4- and subsequent crystallization from<br />

pyridine/DMF solution [4]. The currently biggest<br />

cluster, which could be characterized by X-ray diffraction,<br />

has 54 Cd atoms ([Cd54S32(SPh)48(H2O)4] 4- )<br />

and was syn<strong>the</strong>sized by <strong>the</strong> solvo<strong>the</strong>rmal method<br />

using thiourea as sulfur source [5].<br />

Few contributions were published concerning <strong>the</strong><br />

structure <strong>of</strong> 1. The early one [6] concludes from<br />

different measurements (<strong>the</strong>rmogravitmetric, X-ray<br />

powder diffraction, luminescence and absorption<br />

191<br />

spectra, 113 Cd and 1 H NMR) that 1 consists <strong>of</strong> charge<br />

neutral cluster [Cd10S4(SPh)12] that tends to aggregate<br />

to different degrees in <strong>the</strong> solid and solution. This<br />

observation was used by Hiratani et al. [7], who<br />

reacted 1 with ionic organic bromides with long alkyl<br />

chains and studied intercalative cation accumulation.<br />

He states to have a cluster with Cd10 core. Herron et<br />

al. [4] syn<strong>the</strong>sized <strong>the</strong> [Cd32S14(SPh)36.DMF4] cluster<br />

from <strong>the</strong> solution <strong>of</strong> 1 in pyridine by addition <strong>of</strong><br />

DMF. This claims that both 1 and<br />

[Cd32S14(SPh)36.DMF4] undergo a rearrangement in<br />

pyridine solution caused by rapid fracture <strong>of</strong> <strong>the</strong><br />

cluster cores which leads to a dynamic mixture <strong>of</strong><br />

species with lower nuclearities.<br />

We used mostly different NMR measurements and<br />

crystallization experiments to characterize <strong>the</strong> intermediate<br />

product 1.<br />

II. RESULTS AND DISCUSSION<br />

A. NMR MEASUREMENTS<br />

1. 1 in d5-pyridine, which showed one set <strong>of</strong> PhS<br />

and one set <strong>of</strong> pyridine signals, and no H correlation<br />

to Cd, probably due to fast ligand exchange.<br />

2. The pyridine solution <strong>of</strong> 1 in d-pyridine was<br />

evaporated and dissolved again in CD2Cl2.<br />

Two sets <strong>of</strong> PhS signals were observed, one <strong>of</strong><br />

<strong>the</strong>m showed correlation to Cd. NMR spectra<br />

changed its pattern after 2 weeks and crystals<br />

were formed (Cd17 cluster connected via PhS<br />

bridges, building a 1D polymer).<br />

3. 1 in d-DMSO after heating at 80 °C shows 3<br />

different PhS ligands and 3 Cd signals. Maybe<br />

[Cd32S14(SPh)36.DMSO4] was formed.<br />

4. Backward reactions with 4 equivalents <strong>of</strong><br />

PhSH/Et3N or NaSPh/Me4NCl or Me4NSPh in<br />

d-acetonitrile or d-DMSO showed no explicit<br />

results, but <strong>the</strong>re was a certain similarity with<br />

expected spectra (Figure 1).


5. Solid-state 113 Cd NMR showed only one intense<br />

broad and several small signals, although<br />

<strong>the</strong> spectra <strong>of</strong> [Cd10S4(SPh)16] 4- revealed<br />

sharp peaks.<br />

Figure 1: 1 H NMR spectra <strong>of</strong> a backward reaction <strong>of</strong><br />

1 with NaSPh (up) and <strong>of</strong> <strong>the</strong> starting cluster<br />

[Cd10S4(SPh)16] 4- (down).<br />

These measurements allowed <strong>the</strong> conclusion that<br />

<strong>the</strong> substance 1 has not a defined core structure. In<br />

coordinating solvents (like pyridine, DMF, DMSO),<br />

some rearrangement <strong>of</strong> <strong>the</strong> clusters takes place which<br />

was also observed earlier [4]. Reactions in noncoordinating<br />

solvents (CH2Cl2) did not lead to <strong>the</strong><br />

formation <strong>of</strong> one defined cluster species which supports<br />

our conclusion.<br />

B. CRYSTALLIZATION EXPERIMENTS<br />

We could successfully crystallize different clusters<br />

from different solutions <strong>of</strong> 1. The results from singlecrystal<br />

XRD are summarized in Table 1.<br />

Preparation Cluster core<br />

1 + 1:4 PhSH/Et3N in acetonitrile,<br />

filtration, cooling to -20 °C<br />

Cd10<br />

1 + excess pyridine, evaporation,<br />

dissolution in CH2Cl2<br />

Cd17<br />

1 in pyridine, addition <strong>of</strong> DMF Cd32 [4]<br />

1 in pyridine, addition <strong>of</strong> toluene Cd54<br />

1 in DMF, heating to 70 °C, filtration,<br />

addition <strong>of</strong> acetone<br />

Cd54<br />

Table 1: Summary <strong>of</strong> crystallization experiments.<br />

The first experiment (addition <strong>of</strong> 1:4 PhSH/Et3N in<br />

acetonitrile) clearly shows that part <strong>of</strong> <strong>the</strong> starting<br />

cluster [Cd10S4(SPh)16] 4- retains its core structure<br />

during <strong>the</strong> <strong>the</strong>rmal decomposition, while <strong>the</strong> rest<br />

forms aggregates or clusters with o<strong>the</strong>r cores, which<br />

are not soluble in acetonitrile after addition <strong>of</strong><br />

PhSH/Et3N and thus could not be obtained in this<br />

way.<br />

The latter experiments can be explained by two<br />

different effects: (i) different species are present in 1<br />

immediately after <strong>the</strong> <strong>the</strong>rmal decomposition, and (ii)<br />

<strong>the</strong>y rearrange to o<strong>the</strong>r stable species in solutions <strong>of</strong><br />

coordinating solvents [4].<br />

192<br />

In conclusion, <strong>the</strong> core structure <strong>of</strong> 1 is not uniform.<br />

Different clusters can be obtained from its<br />

solutions. So, <strong>the</strong> assumed four free coordination<br />

sites cannot be used for fur<strong>the</strong>r functionalization.<br />

ACKNOWLEDGMENTS<br />

The authors thank <strong>the</strong> Austrian Science Fund (FWF)<br />

for financial support and Pr<strong>of</strong>. Kurt Mereiter from<br />

Vienna University <strong>of</strong> Technology for help with single-crystal<br />

XRD.<br />

REFERENCES<br />

[1] I. G. Dance, A. Choy and M. L. Scudder. J. Am.<br />

Chem. Soc., 106: 6285-95, 1984.<br />

[2] G. S. H. Lee, D. C. Craig, I. Ma, M. L. Scudder,<br />

T. D. Bailey and I. G. Dance. J. Am. Chem. Soc.,<br />

110: 4863-4, 1988.<br />

[3] N. Zheng, X. Bu, J. Lauda and P. Feng. Chem.<br />

Mater., 18: 4307-11, 2006.<br />

[4] N. Herron, J. C. Calabrese, W. E. Farneth and Y.<br />

Wang. Science, 259(5100): 1426-8, 1993.<br />

[5] N. Zheng, X. Bu, H. Lu, Q. Zhang and P. Feng.<br />

J. Am. Chem. Soc., 127: 11963-5, 2005.<br />

[6] W. E. Farneth, N. Herron and Y. Wang. Chem.<br />

Mater., 4: 916-22, 1992.<br />

[7] T. Hiratani and T. Konishi. Angew. Chem. Int.<br />

Ed., 43: 5943-6, 2004.


Creating building blocks for Hybrid Materials:<br />

Investigation <strong>of</strong> Adhesives for Zirconia Substrates<br />

Christoph Lomoschitz and Guido Kickelbick<br />

Institute <strong>of</strong> Materials Chemistry<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: chlomo@mail.zserv.tuwien.ac.at<br />

Abstract — Phosphorus based coupling agents<br />

(P-CAs) comprising different chain lengths were<br />

prepared via a multistep syn<strong>the</strong>sis approach. P-CAs<br />

were used for modification <strong>of</strong> zirconia nanometer<br />

sized particles and investigated towards bonding<br />

(IR and 31 P MAS NMR), surface coverage (TGA)<br />

and surface properties (nitrogen sorption at 77 K).<br />

Results suggest that bonding <strong>of</strong> P-CAs is <strong>of</strong> covalent<br />

character. Fur<strong>the</strong>more it was found that different<br />

chain lengths influence <strong>the</strong> surface properties<br />

significantly.<br />

I. INTRODUCTION<br />

Transition metal oxides are a widespread class <strong>of</strong><br />

substances <strong>of</strong>fering a variety <strong>of</strong> properties such as<br />

reasonable hardness, magnetism or photo-catalytical<br />

activity. However homogeneous incorporation <strong>of</strong><br />

such advanced ceramics into a polymeric matrix,<br />

thus creating hybrid materials, is <strong>of</strong>ten ra<strong>the</strong>r<br />

delicate and governed by phase separation. This<br />

problem can be overcome by introducing molecular<br />

layers <strong>of</strong> adhesives, so called coupling agents. Very<br />

prominent systems comprise alkyl thiol films on<br />

gold or noble metals as well as alkylsilicon based<br />

films on metal oxides. 1 The affinity <strong>of</strong> phosphates<br />

and/or phosphonates to transition metal oxides was<br />

also subject for several studies in literature. 2-4<br />

However, fundamental studies concerning P-CAs<br />

are only limited.<br />

In this paper we present a convenient multistep<br />

syn<strong>the</strong>sis approach to obtain methacryl based P-<br />

CAs <strong>of</strong> convenient purity with reasonable yields.<br />

The prepared linkers were grafted onto zirconia<br />

surfaces, observing ra<strong>the</strong>r high reactivity <strong>of</strong> P-CAs<br />

towards zirconia. The surface properties <strong>of</strong><br />

modified zirconia were significantly changed.<br />

Evaluation <strong>of</strong> analytical data suggested reasonably<br />

high grafting densities <strong>of</strong> P-CAs thus resulting<br />

partially in self-assembled monolayer like systems.<br />

193<br />

II. RESULTS AND DISCUSSION<br />

Two different classes <strong>of</strong> P-CAs were syn<strong>the</strong>sized<br />

being methacryloyloxy alkyl phosphates and<br />

methacryloyloxy alkyl phosphonates (Figure 1).<br />

Phosphates were syn<strong>the</strong>sized via a simple<br />

condensation reaction <strong>of</strong> <strong>the</strong> corresponding �methacryloyloxy-�-alkanol<br />

with POCl3 followed by<br />

hydrolysis. Phosphonate homologues were<br />

syn<strong>the</strong>sized using a multistep approach comprising<br />

<strong>the</strong> Arbuzov reaction followed by deprotection,<br />

functionalization and mild hydrolysis via<br />

trimethylsilyl bromide.<br />

O<br />

O<br />

Br<br />

n<br />

O<br />

O<br />

P(OEt) 3 O P KOH/MeOH HO P<br />

n O<br />

n O<br />

150°C<br />

O<br />

O<br />

O<br />

O<br />

HO n O<br />

O<br />

O<br />

O<br />

P<br />

n OH<br />

OH<br />

1. TMSBr<br />

2. MeOH<br />

+<br />

CH 2 Cl 2<br />

1. POCl3/NEt3<br />

O<br />

O<br />

2. H2O<br />

HO<br />

P<br />

O<br />

OH<br />

n O<br />

O<br />

O<br />

O<br />

NEt CH 3 2Cl Cl<br />

2<br />

O<br />

P<br />

n O<br />

O<br />

Figure 1: Syn<strong>the</strong>sis <strong>of</strong> P-CAs; up: phosphonic acid<br />

bottom: phosphoric acid mono ester<br />

Mono alkyl phosphoric acids as well as<br />

phosphonic acids <strong>of</strong>fer <strong>the</strong> possibility <strong>of</strong> tridentate<br />

bonding to a substrate caused by two possible<br />

condensation sites (P-OH) and an electron donating<br />

site (P=O). 5 In literature bonding states <strong>of</strong> P-CAs on<br />

metal oxides are intensely debated. Typical results <strong>of</strong><br />

modified metal oxide surfaces suggest a variety <strong>of</strong><br />

different bonding modes present. 6<br />

Zirconia particles (Aldrich,


CH 2<br />

3000<br />

CH 2<br />

3000<br />

MUPnA<br />

MHPnA<br />

MPPnA<br />

MEPnA<br />

2500<br />

2500<br />

MDPA<br />

MEPA<br />

C=O<br />

C=C<br />

2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 C=O<br />

C=C<br />

2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 P=O / C-O-C<br />

P-O<br />

P-OH<br />

CH 2<br />

1000 3000<br />

P=O / C-O-C<br />

P-O<br />

P-OH<br />

1000<br />

CH 2<br />

3000<br />

MUPnA@ZrO 2<br />

MHPnA@ZrO 2<br />

MPPnA@ZrO 2<br />

MEPnA@ZrO 2<br />

C=O<br />

C=C<br />

2500 2000 1500<br />

Wavenumber cm-1 2500 2000 1500<br />

Wavenumber cm-1 2500 2000 1500<br />

Wavenumber cm-1 MDPA@ZrO 2<br />

MEPA@ZrO 2<br />

2500<br />

C=O<br />

C=C<br />

2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 2000 1500<br />

Wavenumber cm-1 P=O / C-O-C<br />

P-O<br />

P-OH<br />

1000<br />

P=O / C-O-C<br />

P-O<br />

P-OH<br />

1000<br />

Figure 2: IR Spectra <strong>of</strong> syn<strong>the</strong>sized phosphonic acids<br />

(top) and mono alkyl phosphates (bottom); right:<br />

pure left: grafted on zirconia<br />

From <strong>the</strong> IR spectra <strong>of</strong> P-CAs, pure and grafted on<br />

zirconia, it is clearly visible that <strong>the</strong> methacrylic<br />

moiety, crucial for attachment to a polymer matrix,<br />

did not change after <strong>the</strong> grafting process, as<br />

evidenced by <strong>the</strong> C=O (1720 cm -1 ) and C=C (1630<br />

cm -1 ) stretching vibrations. Changes in <strong>the</strong> P-O<br />

region 950-1200 cm -1 (decrease <strong>of</strong> P-OH and P=O<br />

band) indicate covalent multidentate bonding <strong>of</strong> P-<br />

CA on <strong>the</strong> zirconia surface featuring a variety <strong>of</strong><br />

different bonding modes according to published<br />

results. 2, 3 The results obtained could be confirmed<br />

using 31 P solid state NMR (data not shown) where an<br />

upfield shift <strong>of</strong> <strong>the</strong> main resonance could be detected.<br />

TG /%<br />

100.0<br />

99.0<br />

98.0<br />

97.0<br />

96.0<br />

95.0<br />

94.0<br />

93.0<br />

Phosphonate<br />

C2<br />

C3<br />

C6<br />

C11<br />

Phosphate<br />

C2<br />

C10<br />

100 200 300 400 500 600 700<br />

Temperature /°C<br />

Figure 4: TGA graphs <strong>of</strong> modified zirconia powder<br />

Thermogravimetric measurements were carried out<br />

in order to investigate grafting densities <strong>of</strong> modified<br />

zirconia particles (Figure 4). It could be observed that<br />

coverages range from 55 to 85 % for phosphonates<br />

and from 35 to 45 % <strong>of</strong> a <strong>the</strong>oretically dense<br />

194<br />

monolayer for phosphates assuming a <strong>the</strong>oretical<br />

bonding area <strong>of</strong> 24 Å 2 per phosphonate or phosphate<br />

headgroup respectively. 3 These results suggest that<br />

phosphonates possess a higher affinity to zirconia<br />

than phosphates under <strong>the</strong> conditions used for<br />

grafting process. In addition it was observed that<br />

longer alkyl spacers favour higher grafting densities.<br />

This effect can be ascribed to van-der-Waals<br />

interactions between <strong>the</strong> alkyl chains similar as<br />

known from self assembled monolayers. 1<br />

Nitrogen sorption measurements at 77 K show a<br />

significant change in surface properties. It can be<br />

clearly observed that CBET constants decrease with<br />

increasing chain length <strong>of</strong> P-CA which is equivalent<br />

to an increasing hydrophobicity <strong>of</strong> <strong>the</strong> surface, as<br />

CBET is proportional to <strong>the</strong> adsorption enthalpy <strong>of</strong><br />

nitrogen (Figure 5).<br />

CBET<br />

70<br />

65<br />

60<br />

55<br />

50<br />

45<br />

40<br />

35<br />

30<br />

0 2 4 6 8 1012 No. Methylene<br />

Phosphonate<br />

Phosphate<br />

Unmodified<br />

Figure 5: CBET constants <strong>of</strong> modified zirconia<br />

powders<br />

ACKNOWLEDGMENTS<br />

This work was financed by IvoclarVivadent AG. We<br />

thank Dr. Burtscher, Dr. Moszner and Dr. Salz for<br />

encouragement and scientific support.<br />

REFERENCES<br />

[1] A. Ulman, Chem. Rev. 1996, 96, 1533-1554.<br />

[2] W. Gao, L. Dickinson, C. Grozinger, F. G. Morin<br />

and L. Reven, Langmuir 1996, 12, 6429-<br />

6435.<br />

[3] G. Guerrero, P. H. Mutin and A. Vioux, Chem.<br />

Mater. 2001, 13, 4367-4373.<br />

[4] J. T. Woodward, A. Ulman and D. K. Schwartz,<br />

Langmuir 1996, 12, 3626-3629.<br />

[5] P. H. Mutin, G. Guerrero and A. Vioux, J. Mater.<br />

Chem. 2005, 15, 3761-3768.<br />

[6] F. Brodard-Severac, G. Guerrero, J. Maquet, P.<br />

Florian, C. Gervais and P. H. Mutin, Chem.<br />

Mater. <strong>2008</strong>, 20, 5191-5196.


�������������������������������������������������������������<br />

���������������������<br />

�<br />

������������� � ����������������� � ����������������� � �������������������� � ������������������� � �<br />

�<br />

� �������������������������������������������������������������������������������������������<br />

� �������������������������������������������������������������������������������������������������<br />

���������������������������������������<br />

�<br />

�<br />

��������� ��������������� ����������������������<br />

������ ��� ���������� ����� ��� ������������������� ���<br />

�������� ���� �������� ���� ���� ���������� ���<br />

������������ ������������� ������ ����������� �����<br />

������������ ��� ������� ���� ����� ������� ������ ����<br />

�������� ���� ���������� ����������� ����� �������� ����<br />

�������� ���������� ������������� ��� ��������� ��� ���<br />

�������� ����� �������� ������������� ��������� ���� ����<br />

������� ��������� ����� ���������� ���� ������ ������<br />

����������� �� �������� ��������� ��������� ���������<br />

�������������������������������������������������<br />

���������������������������������������������������<br />

����� ���� ��� ���������� ������ ��������� �����������<br />

��������� ����� ��������� ������ ��� ����� ����� �����<br />

��������� ������� �� ���������� �������� �������� ���<br />

�������������������������������������<br />

��� �������������<br />

������ ����������������������������� ��������� ������<br />

����� ��� ��������������������� ����� ������������ ����<br />

�������������������������������������������������������<br />

���� ��� ������ ���������� ��� ������ ��������� ����� ���<br />

���������� ���� ������� ����� ����������� �����������<br />

������������������������������������������������������<br />

����� ����� �������� ��� ���������� ��� ������ �����������<br />

��������� ���� ���� ���������� ������������ �������� ����<br />

������ ����� �������� ����� ���� ���� ������������ ����<br />

���������������������������������������<br />

�<br />

�<br />

�<br />

�����������������������������������<br />

�<br />

������������ ��� ������� ���������� ����� ���������<br />

������������� ��� ��������� ������ ������� ���� �����<br />

������� ��������� ����� ������� �������� ��� ������ ����<br />

���������������������������������������������������<br />

���� ��� ���� ����� ������������ �������� ������ ����������<br />

�����<br />

195<br />

���� ��������������<br />

������ ������ ����������� ����� ������������ ����� ������<br />

������������������������������������� ���������������<br />

����� ����� �����������������������������������������<br />

������ ������ �������������� ��� ������������ ���� ������<br />

�����������������������������������������������������<br />

����� ������� ����� ���������� ��� ������� ���� �������<br />

���������������������������������������������������<br />

���� ��������� �������� ���������� ������� �� ������ ����<br />

����������������������������<br />

�<br />

�<br />

�<br />

�������������������������������������<br />

�<br />

����������������������������������������������������<br />

���� ���� ������� ��������� �� ������ ���� ������������<br />

��������������������������������������������<br />

����� ��������������<br />

���������������������������������������������������<br />

���� ��������� �� ������������ ����� ������������ ���� ���<br />

���������� ������ ��������� ����������� ��� ������������<br />

����������������������� ������ ������������������� ��<br />

����������� �������� ���� ����������� ��� �������������<br />

�����������������������������������������������������<br />

�������������� ��� ���� ��������� ������� ����� ������<br />

���������������������������������������� ������ ��������<br />

������������������������������������������������������<br />

������ ����������� ���� ������ ���� ���������� ��� ����<br />

�������������� ������������ ������ �����������������<br />

������������������� �������� ��� ������ ������������ ���<br />

������� ���� ������ ����� �� ������ ���� �������������<br />

������������


�<br />

����������������������������������������<br />

�����������������������������������������<br />

�<br />

�����������������������������������������������������<br />

������������������ ��� ������������� ��� ���� �����������<br />

������������������������������������������������������<br />

���������� �� ���������� ����������� ��� ��������� �� ����<br />

���������� ����������� ������� ����� �������� ���� ����<br />

������ ���������� ��� ���� ������������ ��� ����������<br />

������� ���� �������� ��� ������ ������������ ����� ��<br />

�������� �������� ���� ������� ����� �� ������ ����������<br />

������������<br />

�<br />

�<br />

������������������������������������������<br />

������������������������������������<br />

���� ������������<br />

����������������������������������������������������<br />

���� ��������� ���� ���������������� ��������� ����� ����<br />

�������������������������������������������������������<br />

����������������������������������������������������<br />

������ ������������ ������������������������������<br />

����������������������<br />

�<br />

�<br />

������������������������������������<br />

�<br />

�<br />

196<br />

��� �����������<br />

������ ������ �������� ���������� ��� ����� ��� �������<br />

������������������������������������������������������<br />

��������� ���������� ������������ ����� ������������<br />

������������������������������<br />

���� ������������������<br />

�������������������������������������������������������<br />

���� ��������� �������� ��������� ��� �������� ���� ���<br />

������� ������������������� ���� ������������� ����<br />

������������ ������ ����������� ������������ �����������<br />

����� �������� ������������� ���� ��������� �����������<br />

����� ���� ��������� ������� �������� ���� ���� �������<br />

����������� ���������� ��� ������ ��� �������� ������ ���<br />

���������������<br />

����� �����������<br />

�<br />

���� ���������������������������������������������<br />

������������<br />

���� ������� �������� ������� �������������� ����������<br />

������������������������<br />

���� ��������������������������������<br />

�����������������������<br />

� �������������������������������������������������<br />


Realization <strong>of</strong> Inversion-type GaN MOSFETs with Ar Implantation<br />

for Device Isolation<br />

Clemens Ostermaier a,c , Sang-Il Ahn a , Kay Potzger b , Manfred Helm b , Stefan Kalchmair c ,<br />

Dionyz Pogany c (Faculty Mentor), Jong-Ho Lee a , Sung-Ho Hahm a , and Jung-Hee Lee a<br />

a<br />

School <strong>of</strong> Electrical Engineering & Computer Science, Kyungpook National University, Daegu,<br />

Republic <strong>of</strong> Korea<br />

b<br />

Institute <strong>of</strong> Ion Beam Physics and Materials Research, Forschungszentrum Dresden-Rossendorf,<br />

Dresden, Germany<br />

c<br />

Institute for Solid State Electronics, Vienna University <strong>of</strong> Technology, Vienna, Austria<br />

Email: Clemens.Ostermaier@TUWien.ac.at<br />

Abstract — We report an inversion-type metaloxide-semiconductor<br />

field-effective transistor<br />

(MOSFET) on MOCVD-grown p-type GaN with<br />

silicon - implanted source/drain regions and device<br />

isolation by argon implantation. This isolation<br />

technique allows a planar process without<br />

surface leakage current caused by loss <strong>of</strong> nitrogen<br />

at <strong>the</strong> surface during <strong>the</strong> high-temperature annealing<br />

required for <strong>the</strong> activation <strong>of</strong> silicon dopant.<br />

The maximum measured drain output current was<br />

100 �A/mm while <strong>the</strong> device isolation current was<br />

below 10 nA/mm over <strong>the</strong> whole range <strong>of</strong> device<br />

operation.<br />

I. INTRODUCTION<br />

GaN is known for its high power device capability<br />

due to <strong>the</strong> high saturation velocity, high mobility<br />

in <strong>the</strong> two-dimensional electron gas layer, large<br />

bandgap and high <strong>the</strong>rmal conductivity. Therefore<br />

heterostructure devices with nitride-compounds<br />

showed incredible performances during <strong>the</strong> last<br />

years. Those devices are usually depletion mode<br />

devices and need <strong>the</strong>refore a higher implementation<br />

to be used in electronic circuits. On <strong>the</strong> o<strong>the</strong>r<br />

hand, inversion-type GaN metal-oxidesemiconductor<br />

field-effect transistors (MOSFET)<br />

are usually normally <strong>of</strong>f and fur<strong>the</strong>r have <strong>the</strong> advantage<br />

that <strong>the</strong> <strong>of</strong>f-state drain current is reduced<br />

by <strong>the</strong> large band gap [1,2]. This could lead to a<br />

competitive on/<strong>of</strong>f current ratio, which is <strong>the</strong> determining<br />

factor <strong>of</strong> a power switch. On <strong>the</strong> o<strong>the</strong>r<br />

side, source/drain implantation is very difficult in<br />

GaN with its strong atomic binding energy and its<br />

high melting point. Hence, <strong>the</strong> annealing temperature<br />

for a full crystal recovery should be around<br />

1800°C and requires high nitrogen partial pressure.<br />

Usually used annealing processes with temperatures<br />

around 1000°C at normal ambient pressure<br />

lead to outdiffusion <strong>of</strong> nitrogen, which is<br />

known to create donor like vacancies causing n+<br />

197<br />

doping in <strong>the</strong> surface region [3,4]. In this paper we<br />

present <strong>the</strong> fabrication <strong>of</strong> GaN MOSFET devices<br />

with Ar+ device isolation as shown schematically<br />

in Fig. 1. The high fluence <strong>of</strong> argon atoms prevents<br />

<strong>the</strong> current flow in <strong>the</strong> implanted surface<br />

layer [5].<br />

Figure 1: Schematic <strong>of</strong> <strong>the</strong> structure<br />

II. DEVICE FABRICATION<br />

For device fabrication we grew by metal-organic<br />

chemical vapour deposition (MOCVD) a 1 �m<br />

Mg-doped p-type GaN layer on a 2 �m GaN buffer<br />

grown on sapphire substrate. Silicon implantation<br />

[6,7] was performed at 45 keV and 100 keV with<br />

an overall fluence <strong>of</strong> 1x10 16 cm -2 into a 300 nm<br />

thick SiO2 channel mask making a 2 �m channel<br />

length. The silicon atoms were activated during a 5<br />

minutes annealing process at 1100°C under N2 and<br />

Ar ambient and different capping modes. The<br />

summary <strong>of</strong> <strong>the</strong> secondary mass ion spectroscopy<br />

(SIMS) pr<strong>of</strong>iles for different annealing techniques<br />

are shown in Fig. 2. After wet removal <strong>of</strong> <strong>the</strong> mask<br />

and cleaning <strong>the</strong> surface in H2SO4:H2O2, a 60 nmthick<br />

PECVD-SiO2 film was deposited as gate<br />

oxide. The interface characteristic <strong>of</strong> such an interface<br />

was known to have an interface and fixed<br />

charge density below 3x10 12 cm -2 and an interface<br />

trap density <strong>of</strong> around 10 11 cm -2 eV -1 close to band<br />

edge. The characterization <strong>of</strong> <strong>the</strong> interface was<br />

performed by UV-induced capacitance-voltage


(CV) measurements and <strong>the</strong> CV conduction<br />

method [8,9]. Source and drain contacts were<br />

provided by dry-etching, Ti/Al/Ni/Au deposition<br />

by electron-beam evaporator and annealing in an<br />

rapid <strong>the</strong>rmal annealing process at 700°C. In order<br />

to isolate <strong>the</strong> MOSFET devices, Ar + with a fluence<br />

<strong>of</strong> 5x10 14 cm -2 was implanted at 40 keV and 140<br />

keV. The average implantation depth <strong>of</strong> <strong>the</strong> argon<br />

atoms exceeded that one <strong>of</strong> silicon to guarantee a<br />

sufficient isolation for all tested annealing methods,<br />

as shown in Fig. 2.<br />

Source-Drain implanted Si atoms (cm -3 )<br />

1E21<br />

1E20<br />

0 100 200 300 400<br />

Depth (nm)<br />

not annealed<br />

SiO capping in N 2 2<br />

no capping in Ar<br />

GaN wafer capping in N2 simulated Ar pr<strong>of</strong>il<br />

1E19<br />

1E18<br />

Figure 2: Measured Si and simulated Ar SIMS<br />

pr<strong>of</strong>iles for different ambients and capping<br />

Finally, a Ni/Au gate film was deposited with a 3<br />

�m gate overlapping on each side over <strong>the</strong> nonimplanted<br />

channel area to ensure sufficient channel<br />

modulation.<br />

III. RESULTS AND DISCUSSION<br />

The output characteristics <strong>of</strong> <strong>the</strong> GaN MOSFET<br />

(Fig. 3) show a transistor behavior with a positive<br />

ID (VD) slope until 10 V. The low current value is<br />

determined by <strong>the</strong> large amount <strong>of</strong> interface states<br />

at <strong>the</strong> dielectric/channel interface. The maximum<br />

applied gate voltage is around 13 V, since <strong>the</strong> gate<br />

leakage current <strong>of</strong> <strong>the</strong> device becomes dominant<br />

for higher voltages. However, <strong>the</strong> drain current<br />

increases to 104 mA/mm for higher drain voltage<br />

up to 40 V, due to drain induced barrier lowering<br />

(DIBL). Below <strong>the</strong> threshold voltage, <strong>the</strong> device<br />

showed a wide dynamic range over four orders <strong>of</strong><br />

magnitude. The <strong>of</strong>f-state breakdown voltage <strong>of</strong> <strong>the</strong><br />

MOSFET is 38 V.<br />

The argon implantation was tested between active<br />

areas with a pitch <strong>of</strong> 100 �m. Since <strong>the</strong> mask for<br />

<strong>the</strong> silicon implantation included only <strong>the</strong> gate<br />

finger regions, <strong>the</strong> whole o<strong>the</strong>r area was implanted<br />

by silicon and <strong>the</strong>n isolated by argon atoms. The<br />

Ar + isolation showed extremely low leakage current<br />

values around 10 nA/mm at 40 V as shown in<br />

Isolation implanted Ar atoms (cm -3 )<br />

198<br />

<strong>the</strong> inset <strong>of</strong> Fig. 3. It is <strong>the</strong>refore an easy-to-apply<br />

and efficient isolation method for GaN MOSFET<br />

devices in comparison to non-planar MESA etching<br />

isolation. In comparison to source/drain island<br />

isolation, <strong>the</strong> leakage is lowered and surface leakage<br />

current is prevented by a shallow implantation.<br />

However, no additional annealing steps can be<br />

applied after <strong>the</strong> argon isolation, since this would<br />

cause most argon atoms to escape from GaN resulting<br />

in an ineffective isolation.<br />

Drain Current I D (�A/ mm)<br />

125<br />

100<br />

75<br />

50<br />

25<br />

Buffer current (A/mm)<br />

1E-9<br />

1E-10<br />

-40 -20 0 20 40<br />

Voltage (V)<br />

�V G = -1 V<br />

0<br />

0 2 4 6 8<br />

Drain Voltage V D (V)<br />

V G = 13 V<br />

Figure 3: Output IV characteristic and buffer leakage<br />

current.<br />

ACKNOWLEDGMENTS<br />

This research was supported by <strong>2008</strong> Brain Korea<br />

21 (BK21) and <strong>the</strong> Korea Science and Engineering<br />

Foundation (KOSEF) grant funded by <strong>the</strong> Korea<br />

government (MEST) (No.M10600000273-06J0000-<br />

27310).<br />

REFERENCES<br />

[1] H.-B. Lee et al., IEEE Electronic Device Letters<br />

27(2): 81-83, February 2006<br />

[2] W. Huang, et al., Journal <strong>of</strong> Electronic Materials<br />

35(4):726-732, 2006<br />

[3] J.C. Zolper et al., Journal <strong>of</strong> Crystal Growth<br />

178:157–167, 1997<br />

[4] B. Boudart, Journal <strong>of</strong> Raman Spectroscopy,<br />

33:283–286, 2002<br />

[5] H. Boudinov et al., Applied Physic Letters<br />

78(7):943-945, February 2001<br />

[6] D.W. Jenkins et al., Physical Review B<br />

39(5):3317-3329, February 1989<br />

[7] C.J. Sun et al., Joural <strong>of</strong> Applied Physics 76(1):<br />

236-241, July 1994<br />

[8] C. Ostermaier et al., Physica Status Solidi (c) 5<br />

(6):1992-1994, May <strong>2008</strong><br />

[9] E.H. Nicollian et al., MOS Physics and Technology,<br />

Wiley, New York, 1982


Solution Stability <strong>of</strong> Trinuclear Iron Oxo Clusters<br />

Robert Potzmann, Stefan O. Baumann, Maria Bendova, Michael Puchberger,<br />

Ulrich Schubert<br />

Institute <strong>of</strong> Materials Chemistry, Vienna University <strong>of</strong> Technology,<br />

1060 Vienna, Austria<br />

e-mail: rpotzman@mail.zserv.tuwien.ac.at<br />

Abstract — A large number <strong>of</strong> clusters with a Fe3O<br />

core was syn<strong>the</strong>sised varying <strong>the</strong> ligands and anions.<br />

Detailed investigation in solid state and solution<br />

elucidate <strong>the</strong> close relation between <strong>the</strong> different<br />

coordination compounds and show a complex<br />

equilibrium between <strong>the</strong>m.<br />

I. INTRODUCTION<br />

Iron oxide clusters with bridging carboxylates <strong>of</strong><br />

<strong>the</strong> general formula [Fe3O(OOCR)6(L)3]X (figure<br />

1) also named “basic iron acetate” are well known<br />

for more than three centuries. A huge number <strong>of</strong><br />

publications has been published in <strong>the</strong> field <strong>of</strong><br />

electron transfer effects, intramolecular spin coupling,<br />

magnetic behavior, model compounds for<br />

iron storage proteins and <strong>the</strong> use as oxidation catalyst<br />

for aliphates. [1] Due to <strong>the</strong>se wide spread applications<br />

and <strong>the</strong> simple syn<strong>the</strong>sis numerous clusters<br />

have been prepared. Clusters with a wide<br />

variety <strong>of</strong> different coordinating ligands and diverse<br />

acids to a serious <strong>of</strong> anions. [2][3][4] This work<br />

focuses on <strong>the</strong> stability <strong>of</strong> methacrylate-modified<br />

basic iron acetate in solution and opens new insights<br />

in <strong>the</strong> behavior <strong>of</strong> three membered iron oxo<br />

clusters when <strong>the</strong> anion or <strong>the</strong> coordinating ligands<br />

are modified.<br />

Figure 1: [Fe3O(OOCCH3CCH2)6(H2O)3]NO3<br />

199<br />

II. RESULTS<br />

Although basic iron acetate is well known for over<br />

hundred years little is known about <strong>the</strong> stability in<br />

solution and <strong>the</strong> influence <strong>of</strong> <strong>the</strong> coordinating<br />

ligands on <strong>the</strong> cluster behavior. We systematically<br />

syn<strong>the</strong>sized a large number <strong>of</strong> methacrylatesubstituted<br />

clusters with pyridine, methanol, ethanol<br />

and water as co-ligends and nitrate, chloride<br />

and methacrylate as counter-anions. Fur<strong>the</strong>rmore,<br />

detailed analysies were carried out to investigate<br />

interactions <strong>of</strong> <strong>the</strong> cluster with <strong>the</strong> solvent.<br />

A. SYNTHESIS AND CRYSTAL STRUCTURE<br />

Water-substituted derivatives were easily prepared<br />

by combining <strong>the</strong> iron salt with sodium methacrylate<br />

in water. O<strong>the</strong>r derivatives were syn<strong>the</strong>sised<br />

by exchanging <strong>the</strong> ligand in chlor<strong>of</strong>orm by common<br />

syn<strong>the</strong>tic techniques.<br />

B. SOLUTION BEHAVIOUR<br />

The stability and behaviour <strong>of</strong> <strong>the</strong> clusters in solution<br />

was investigated by solution NMR and IR in different<br />

solvents. Coordinating solvents led to spontaneous<br />

exchange <strong>of</strong> <strong>the</strong> coordinated ligand against solvent<br />

molecules while in non-coordinating solvents<br />

<strong>the</strong> counter-anion partly coordinates to <strong>the</strong> cluster<br />

core (figure 2). By varying <strong>the</strong> ligands, <strong>the</strong> exchange<br />

against <strong>the</strong> anion could be suppressed accompanied<br />

by a tremendous change <strong>of</strong> <strong>the</strong> solubility and reactivity.


Figure 2: Solvent effect on trinuclear iron oxo<br />

clusters a) using a coordinating solvent b) using a<br />

coordinating anion<br />

C. SELF ORGANISATION<br />

Dynamic light scattering (DLS) proved <strong>the</strong> existance<br />

<strong>of</strong> agglomerates in <strong>the</strong> size between 100 and 500 nm.<br />

These agglomerates were investigated by solution<br />

small angle X-ray diffraction (SAXS) and it turned<br />

out that <strong>the</strong>y show <strong>the</strong> ordering in <strong>the</strong> nanometer<br />

regime.<br />

III. CONCLUSION<br />

In this abstract we have described <strong>the</strong> syn<strong>the</strong>sis <strong>of</strong><br />

methacrylate, modified trinuclear iron oxo clusters, a<br />

new class <strong>of</strong> basic iron acetates. The stability <strong>of</strong> such<br />

compounds in different solvents was investigated<br />

resulting in a detailed knowledge about <strong>the</strong> influence<br />

<strong>of</strong> <strong>the</strong> coordinating ligand and <strong>the</strong> anion. Finally, <strong>the</strong><br />

self organisation tendencies <strong>of</strong> <strong>the</strong>se molecules were<br />

described.<br />

ACKNOWLEDGMENTS<br />

Thank to <strong>the</strong> Fonds zur Förderung der wissenschaftlichen<br />

Forschung (FWF) for funding this project.<br />

REFERENCES<br />

[1]F. E. Sowrey, C. J. McDonald and R. D. Cannon;<br />

J. Chem. Soc. Faraday Trans. 94 (1998) 1571-74<br />

[2] M.Losalda, M. A. Mendiola and M. T. Sevilla;<br />

Inorganic Chimica Acta 255 (1997) 125-31<br />

[3] M. A. Kiskin, F. A. Fomina and I. I. Moiseev;<br />

Russian Chemical Bulletin Int. Ed. 53 (2004)<br />

2508-18<br />

200<br />

[4] A. M. Bond, R. J. H. Clark and A. H. White; J.<br />

Chem. Soc. Dalton Trans. (1998) 1845-52


Olefinic Bioreduction<br />

Naseem Iqbal, Marko D. Mihovilovic.<br />

Vienna University <strong>of</strong> Technology<br />

Institute <strong>of</strong> Applied Syn<strong>the</strong>tic Chemistry, Getreidemarkt 9/163-OC, 1060 Wien (Austria)<br />

Phone (+43) 1 58801-15451<br />

E-mail: naseem.iqbal@ias.tuwien.ac.at<br />

Enzyme mediated reactions have become a powerful methodology in asymmetric syn<strong>the</strong>sis and this area<br />

has received considerable attention within <strong>the</strong> past decade. 1 Currently, biocatalysis <strong>of</strong>fers complementary<br />

solutions to metal-assisted transformations in chiral catalysis and is considered one <strong>of</strong> <strong>the</strong> important<br />

strategies <strong>of</strong> “green chemistry”. In particular <strong>the</strong> high sustainability <strong>of</strong> protein based catalytic entities<br />

represents a major advantage. Recent trend studies indicate a significant potential for growth <strong>of</strong> <strong>the</strong><br />

contribution by “white biotechnology” to <strong>the</strong> industrial scale syn<strong>the</strong>sis <strong>of</strong> chiral compounds via<br />

biotransformations at <strong>the</strong> expense <strong>of</strong> classical chiral auxiliary strategies (with recently re-discovered<br />

“organocatalysis” as yet to be adopted methodology for industrial applications). 2 A particularly appealing<br />

target for application <strong>of</strong> <strong>the</strong> outlined strategy in <strong>the</strong> area <strong>of</strong> bioreductions is <strong>the</strong> emerging field <strong>of</strong> formal<br />

biocatalytic olefin hydrogenation. This biotransformation is merely a nucleophilic attack by hydride<br />

originating from NAD(P)H across an activated C=C double bond via a Michael-type reaction. Enzymes<br />

involved are referred to as olefin reductases (OREDs). Recently, old yellow enzyme (OYE) from<br />

Saccharomyces carlsbergensis was re-discovered for <strong>the</strong> chemo- and stereoselective conversion <strong>of</strong> enones<br />

and nitroolefins into saturated compounds. 3 Homologs <strong>of</strong> this enzyme from Shewanella oneidensis<br />

(SYEs) were shown to be catalytically active when recombinantly expressed as GST fusion proteins. 4<br />

Whole cell and crude protein extract biotransformation with different substrates (unsaturated cyclic<br />

ketones, unsaturated diesters, substituted imides) for SYE-3(BL21 DE3) and SYE-4(BL21 DE3)<br />

(Shewanella Yellow Enzyme) were performed. The results (conversion and ee) were rationalized by<br />

chiral GC.<br />

������������������������������������������������������������<br />

1<br />

Biocatalysis in <strong>the</strong> Pharmaceutical and Biotechnology Industries, Patle, R.N. (ed.) CRC Press, Boca<br />

Raton, 2007. Enzyme Catalysis in Organic Syn<strong>the</strong>sis, Drauz, K.; Waldmann, H. (eds.) Wiley-VCH,<br />

Weinheim 2002.<br />

2<br />

Business assessment by Frost&Sullivan, Chem. Eng. News 2003, 81.<br />

3<br />

Swiderska, M.A.; Stewart, J.D. J. Mol. Catal. B: Enzym. 2006, 42, 52. Swiderska, M.A.; Stewart, J.D.<br />

Org. Lett. 2006, 8, 6131.<br />

4<br />

Brige, A.; van den Hemel, D.; Carpentier, W.; de Smet, L.; van Beeumen, J.J. Biochem. J. 2006, 394,<br />

335.<br />

201<br />


202


��������������������������������������������<br />

�����������������������<br />

�<br />

�������������������������������������������������������<br />

��������������������������������<br />

��������������������������������<br />

����������������<br />

�����������������������������������<br />

�<br />

�<br />

��������� �� ���������� ������������� ���� �������<br />

����� ������������� ����� ���������� ���<br />

������������������ �������� ���� ������������ ����<br />

������������������������������������������������������<br />

���������� ��� �� ������ ��� ������ ������������ ������<br />

���������������������������������������������������<br />

����� ��������� ����������� ��� ������ ��� �������� ����<br />

���������������������������������������������������<br />

����������� ���������� ���� ����� �������� ���� �������<br />

����� ��� ���� ���������������� ��������� ����� �����<br />

��������������<br />

��� �������������<br />

������������ �������� ��� ���� ����� ������� ����� �������<br />

���� ���������� ������ ����������� ��� �������� �����<br />

����������������������������������������������������<br />

������������� ���������� ��� ������ ����������� ���<br />

���������� ������������ ���� ����� ������ ����������<br />

������ ���� ��������� ��� ����������� ����� ���������<br />

������������� ��������� ��� ���� ������ ������� ��� ���<br />

�������� ������������ ���� ���� ����� ��������� ����� ���<br />

��������������������������������������������������<br />

�������������������������������������������������<br />

����������������������<br />

���� ���������������������������<br />

�������������������������������������������������<br />

����� ��� ������������� ������ ������ �������� ��� ����<br />

��� ��������� ������� ���������������� ��� ���� �����<br />

������������������������������������������������<br />

��������� ���� �������� ��� ���� ���� ���������� �����<br />

������ �������� �� ����������������������������<br />

�������� ���� ������� ��� ���� �� �� �� ��� � �����������<br />

�������������� ����� ����������� ��� ������� ������<br />

����������� �������������� ������������� ���������<br />

������������������������������������������<br />

���������� ��� ������������ ������������� ������ �� ���<br />

��������������������������������������������������<br />

������� ��������� ����� ���� ���������� ����������� ��� ���<br />

������� ��� ������ �������� ���� ������� ������ ���� ����<br />

����������������������������������������������������<br />

����� ��� ������������� ����� �� ����� ��� ���� �����<br />

����������������������������������������������������<br />

203<br />

����������������������������������������������������<br />

��� ������� ����� ������ �� ���������� ����������<br />

��������� ������ ���������� ������������ ����������� ���<br />

����������������������������������������������������<br />

��������������������������������������������������<br />

���������������������������������������������������<br />

�������������������������������<br />

������������������������������������������������<br />

�����������������������������������������������<br />

�����������������<br />

�����������������������������������������������<br />

���������� ��� ��� ���������� ������ ��� ���� �����������<br />

�����������������������������������������������������<br />

���������������������������������������������������<br />

������������<br />

��<br />

��<br />

�<br />

�<br />

�����������������<br />

����<br />

����<br />

����<br />

��� ��� ��� ��� ��� ���<br />

������������<br />

�<br />

����<br />

����<br />

����<br />

�����������������������������������������������<br />

���������<br />

�<br />

�<br />


����������������������������������������������������<br />

��������������������������������������������������<br />

�����������������������������������������������������<br />

�����������������������������������������������������<br />

������ ��� ��������� ������ ���� ����������� ���������<br />

������������<br />

���� �������� ��������� ��������� ����� �������� ����<br />

�������������� ����������� ����� ��� ����������� ��� ���<br />

������� ��� �������� ������ ��� ���� ����� ����������<br />

��������������� ������������ ��� �������� ����� ����<br />

�����������������������������������<br />

�����������<br />

����������������������������������������������<br />

���� ��� ������� � ��� ���� ��� ������ ������ ���� �����<br />

��������<br />

�����������������������������������������������������<br />

��������<br />

�<br />

204


Effect <strong>of</strong> Various Composition and Dosage <strong>of</strong> Potassium<br />

Activator on <strong>the</strong> Properties <strong>of</strong> Alkali Activated Slag Mortar<br />

Iva Frýbortová and Pavel Rovnaník (Faculty Mentor)<br />

Institute <strong>of</strong> Chemistry<br />

Brno University <strong>of</strong> Technology, Faculty <strong>of</strong> Civil Engineering<br />

Brno, Czech Republic<br />

frybortova.i@fce.vutbr.cz<br />

Abstract — Alkali activated slag is a clinker-free<br />

binding material capable <strong>of</strong> replacing Portland<br />

cement in concrete production. In this paper properties<br />

<strong>of</strong> alkali activated blast furnace slag mortars<br />

were studied. Potassium silicate and hydroxide<br />

solution <strong>of</strong> various silicate moduli (Sio2/K2O) was<br />

used as an activator. Activator dosage was 4, 8, 10<br />

and 20 % with respect to <strong>the</strong> mass <strong>of</strong> slag. Compressive<br />

strength <strong>of</strong> <strong>the</strong> mortar samples was measured<br />

28 days after casting. It was found that increasing<br />

amount <strong>of</strong> activator added to <strong>the</strong> mixture<br />

leads to an increase in compressive strength. No or<br />

very little efflorescence was observed on <strong>the</strong> samples.<br />

I. INTRODUCTION<br />

Utilization <strong>of</strong> waste materials is an issue that has<br />

been paid increasing attention to not only for economical<br />

but also environmental reasons. Portland<br />

cement plants are responsible for 5 % <strong>of</strong> worldwide<br />

CO2 emissions; with every 1 ton <strong>of</strong> cement<br />

produced 0.66 ton <strong>of</strong> CO2 is released to <strong>the</strong> atmosphere<br />

[1]. Alkali activated aluminosilicates, such<br />

as metallurgical slags, fly ash and o<strong>the</strong>r industrial<br />

by-products can be used as binding materials substituting<br />

Portland cement in concrete. Latent hydraulic<br />

properties <strong>of</strong> <strong>the</strong>se materials are activated<br />

by addition <strong>of</strong> alkaline substances, such as sodium<br />

or potassium hydroxides, silicates, carbonates etc.<br />

Alkali activated blast furnace slag shows very<br />

good mechanical properties, high resistance to<br />

chemical attack, freezing and thawing, high temperatures<br />

and alkali-aggregate reaction [2–5].<br />

Properties <strong>of</strong> <strong>the</strong> final product are determined by<br />

many factors, such as chemical and phase composition<br />

<strong>of</strong> <strong>the</strong> slag, slag fineness, type and concentration<br />

<strong>of</strong> activator, water to slag ratio, curing conditions<br />

etc [6].<br />

Properties <strong>of</strong> blast furnace slag mortars activated<br />

by potassium activator <strong>of</strong> various composition and<br />

dosage were studied in this paper.<br />

205<br />

II. EXPERIMENTAL<br />

A ground blast furnace slag (Kotouc, Czech Republic)<br />

with specific surface 380 m 2 /kg was used.<br />

Chemical composition <strong>of</strong> <strong>the</strong> slag is in Table 1.<br />

Component Content [%]<br />

SiO2 34.74<br />

Al2O3 5.91<br />

Fe2O3 0.39<br />

CaO 40.27<br />

MgO 9.60<br />

MnO 0.36<br />

Stotal<br />

0.39<br />

Na2O 0.29<br />

K2O 0.41<br />

Table 1: Chemical composition <strong>of</strong> blast furnace slag.<br />

Water solution <strong>of</strong> anhydrous potassium silicate<br />

Portil K (Henkel, Germany) and potassium hydroxide<br />

was used as an alkaline activator. Silicate<br />

modulus MS <strong>of</strong> <strong>the</strong> activator, i.e. molar ratio <strong>of</strong><br />

SiO2 to K2O, was 0.8; 1.0 and 1.2. The amount <strong>of</strong><br />

activator in solid state added to <strong>the</strong> mixture was 4,<br />

8, 10 and 20 % with respect to <strong>the</strong> mass <strong>of</strong> slag.<br />

Standard silica sand was used as an aggregate.<br />

Mortar samples <strong>of</strong> 40×40×160 mm dimensions<br />

were cast and 24 hours after demoulding immersed in<br />

water. 28 days after casting <strong>the</strong> samples were subjected<br />

to compressive strength test. Pieces <strong>of</strong> samples<br />

left after strength testing were partly immersed in<br />

water due to efflorescence observation.<br />

III. RESULTS<br />

Table 2 and Figure 1 show 28 day compressive<br />

strengths <strong>of</strong> <strong>the</strong> specimens. It can be seen that for all<br />

silicate moduli increasing amount <strong>of</strong> alkaline activator<br />

results in <strong>the</strong> strength growth. Therefore, it can be<br />

stated that <strong>the</strong> required ultimate strength <strong>of</strong> <strong>the</strong> mixture<br />

can be adjusted by <strong>the</strong> amount <strong>of</strong> activator,<br />

which also works partly as a plasticizer. On <strong>the</strong> o<strong>the</strong>r


hand, <strong>the</strong> activator concentration should be kept as<br />

low as possible due to several reasons:<br />

− alkaline activator solution is very corrosive to<br />

<strong>the</strong> material <strong>of</strong> storing and mixing equipment,<br />

− alkaline activator is <strong>the</strong> most expensive constituent<br />

<strong>of</strong> <strong>the</strong> mixture,<br />

− alkalis represent ecological load when occur in<br />

waste water,<br />

− excessive amount <strong>of</strong> alkalis in <strong>the</strong> mixture can<br />

lead to efflorescence.<br />

% <strong>of</strong><br />

MS = SiO2/K2O<br />

activator 0.8 1.0 1.2<br />

4 41.0 48.9 58.2<br />

8 77.2 66.8 60.7<br />

10 84.1 84.3 77.7<br />

20 90.1 90.4 86.3<br />

Table 2: Compressive strength (28 day) <strong>of</strong> alkali<br />

activated slag mortar samples in MPa.<br />

Table 2 shows that even with only 4 % <strong>of</strong> alkaline<br />

activator, based on <strong>the</strong> mass <strong>of</strong> slag, <strong>the</strong> compressive<br />

strength can reach 58.2 MPa (for MS = 1.2). The<br />

mixtures with 20 % <strong>of</strong> activator had <strong>the</strong> highest compressive<br />

strength, i.e. 90.1; 90.4 and 86.3 MPa, for<br />

silica modulus equal to 0.8; 1.0 and 1.2, respectively.<br />

However, <strong>the</strong> compressive strength difference between<br />

<strong>the</strong> mixtures with 10 % and 20 % <strong>of</strong> <strong>the</strong> alkaline<br />

activator is not substantial.<br />

For mixtures with low amount <strong>of</strong> activator <strong>the</strong><br />

strength is ra<strong>the</strong>r variable for different silica moduli<br />

contrary to <strong>the</strong> mixtures with 10 or 20 % <strong>of</strong> activator,<br />

see Figure 1.<br />

Efflorescence occurred, to a very small extent, only<br />

on samples with 20 % <strong>of</strong> activator.<br />

Compressive strength [MPa]<br />

100<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

0,8 1,0 1,2<br />

Ms = SiO2/K2O<br />

4 %<br />

8 %<br />

10 %<br />

20 %<br />

Figure 1 Compressive strength (28 day) <strong>of</strong> blast furnace<br />

slag activated by 4, 8, 10 and 20 % <strong>of</strong> activator<br />

(with respect to <strong>the</strong> mass <strong>of</strong> slag) with silicate<br />

modulus 0.8; 1.0 and 1.2.<br />

206<br />

IV. CONCLUSION<br />

Properties <strong>of</strong> alkali activated blast furnace slag depend<br />

on several factors; one <strong>of</strong> <strong>the</strong>m is type and dosage<br />

<strong>of</strong> alkaline activator. In this study anhydrous<br />

potassium silicate and potassium hydroxide in<br />

amount 4, 8, 10 and 20 %, based on <strong>the</strong> mass <strong>of</strong> slag,<br />

was used as an activator. Silicate moduli <strong>of</strong> <strong>the</strong> prepared<br />

solutions were 0.8; 1.0 and 1.2.<br />

It was found that it is possible to reach ra<strong>the</strong>r high<br />

compressive strength after 28 days with only 4 % <strong>of</strong><br />

alkaline activator, i.e. 58.2 MPa for silicate modulus<br />

1.2. The higher amount <strong>of</strong> activator was added to <strong>the</strong><br />

mixture <strong>the</strong> higher compressive strength was obtained<br />

for all silicate moduli. Maximum strength measured<br />

was 90.4 MPa for samples with 20 % <strong>of</strong> activator and<br />

silicate modulus 1.0. Very small difference in<br />

strengths between <strong>the</strong> mixtures with 10 and 20 % <strong>of</strong><br />

activator has been observed.<br />

Excessive amount <strong>of</strong> alkalis causes efflorescence,<br />

especially when sodium activators are used. Larger<br />

ionic radius <strong>of</strong> potassium does not allow it to crystallize<br />

on <strong>the</strong> surface to such an extent. Therefore, very<br />

little marks <strong>of</strong> efflorescence occurred only on <strong>the</strong><br />

surface <strong>of</strong> specimens with 20 % <strong>of</strong> activator.<br />

ACKNOWLEDGMENTS<br />

This paper was elaborated with <strong>the</strong> financial support<br />

<strong>of</strong> <strong>the</strong> project granted by Ministry <strong>of</strong> Education,<br />

Youth and Sport (MSM 0021630519).<br />

REFERENCES<br />

[1] J. Davidovits, High-Alkali Cements for 21st Century<br />

Concretes, American Concrete Institute, SP-<br />

144 (Concrete Technology): 383-97, 1994.<br />

[2] P. Rovnanikova, P. Bayer, Corrosion Resistance<br />

<strong>of</strong> Alkali-activated Aluminosilicate Materials,<br />

Innovations and Developments in Concrete Materials<br />

and Construction, 373-381, 2002.<br />

[3] T. Bakharev, J. G. Sanjayan, Y. B. Cheng, Sulfate<br />

attack on alkali-activated slag concrete, Cement<br />

and Concrete Research, 32: 211-216, 2003.<br />

[4] T. W. Cheng, J. P. Chiu, Fire-resistant geopolymer<br />

produced by granulated blast furnace slag,<br />

Minerals Engineering, 16:, 205-210, 2003.<br />

[5] A. Fenandez-Jimenez, F. Puertas, The alkali-silica<br />

reaction in alkali activated granulated slag mortars<br />

with reactive aggregate, Cement and Concrete<br />

Research, 32(7): 1019-1024, 2002.<br />

[6] Talling B., Krivenko P., Blast Furnace Slag – The<br />

Ultimate Binder, Waste Materials Used in Concrete<br />

Manufacturing, 235-289 1997.


Impact <strong>of</strong> Mineral Dust on PM10 Levels in Austria –<br />

Seasonal and local variations <strong>of</strong> carbonate carbon in ambient<br />

PM10 and source samples<br />

Nicole Jankowski and Hans Puxbaum<br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: Nicole.Jankowski@tuwien.ac.at<br />

Abstract — Mineral dust contributing to PM10<br />

levels in Austria arises from different sources. The<br />

aim <strong>of</strong> this study was to identify <strong>the</strong>se sources and<br />

characterise <strong>the</strong>m chemically. Paved and unpaved<br />

street dust, bare soil samples, dust from construction<br />

sites and gritting material were considered as<br />

main contributors <strong>of</strong> mineral dust to ambient air<br />

concentrations. The grab samples were resuspended<br />

and <strong>the</strong> PM10 fractions were collected on<br />

two different filter materials. Metals, soluble ions<br />

and various carbonaceous constituents were determined.<br />

The chemical mass balance (CMB) model<br />

and a macrotracer concept were used to estimate<br />

<strong>the</strong> contribution to ambient PM10 levels.<br />

I. INTRODUCTION<br />

Exceedings <strong>of</strong> <strong>the</strong> short term air quality standard<br />

<strong>of</strong> particulate matter smaller than an aerodynamic<br />

diameter <strong>of</strong> 10μm (PM10), set by <strong>the</strong> Commission<br />

<strong>of</strong> <strong>the</strong> European Communities [1], has been reported<br />

for many urban settlements in Europe [2].<br />

To develop effective reduction measures, more<br />

information about <strong>the</strong> sources and <strong>the</strong>ir contribution<br />

to ambient particle concentration is required.<br />

Various programs have been launched (e.g. CAFE<br />

program – Clean Air for Europe [3]), to find <strong>the</strong><br />

sources responsible for <strong>the</strong> exceedings. Generally,<br />

<strong>the</strong> sources <strong>of</strong> fine particulate matter can ei<strong>the</strong>r be<br />

natural or man-made; for mineral dust both origins<br />

are possible. Important sources for mineral dust<br />

are: natural soil dust, paved and unpaved street<br />

dust, fugitative dust from construction activities<br />

and gritting materials used in winter.<br />

II. EXPERIMENTAL<br />

Sufficient sample material was collected around <strong>the</strong><br />

stations for ambient air monitoring. The dried material<br />

was <strong>the</strong>n resuspended in a <strong>the</strong>refore constructed<br />

apparatus, and <strong>the</strong> PM10 fraction was collected with<br />

PM10 sampling heads on two filters in parallel. For<br />

<strong>the</strong> analysis <strong>of</strong> metals a cellulose membrane filter<br />

207<br />

was used, for <strong>the</strong> determination <strong>of</strong> ions and carbonaceous<br />

constituents quartz fibre filters were applied.<br />

Gravimetric analysis was performed prior and after<br />

<strong>the</strong> sampling after 48h <strong>of</strong> equilibration in a room<br />

with 20±1°C and 50±5% RH, according to EN 12341<br />

[4].<br />

Samples were analysed by X-Ray fluorescence<br />

for Al and Si, after digestion <strong>of</strong> 1 up to 3 filters<br />

with aqua regia ICP-HRMS and GF-AAS were<br />

applied to determine Mg, Ca, Ti, V, Cr, Mn, Fe,<br />

Co, Ni, Cu, Zn, Sr, Cd, Sn, Ba, Pb and As. The<br />

soluble ions Na + , K + , NH4 + , Cl - , NO3 - , SO4 2- were<br />

determined by anion and cation chromatography.<br />

The composite parameters total carbon and its<br />

fractions organic carbon, elemental carbon and<br />

carbonate carbon were determined with different<br />

combustion methods and <strong>the</strong> subsequent determination<br />

<strong>of</strong> <strong>the</strong> evolved CO2 [5, 6]. The anhydrosugar<br />

levoglucosan (a tracer for wood smoke) and<br />

<strong>the</strong> sugar alcohol arabitol (a tracer for fungal<br />

spores) were analysed with HPLC [7]. Cellulose<br />

was determined with a photometric method [8],<br />

HULIS (humic like substances) were separated by<br />

solid phase extraction (SPE) and analysed as dissolved<br />

organic carbon [9].<br />

III. RESULTS<br />

The analysis <strong>of</strong> street dust samples taken in March<br />

at <strong>the</strong> sampling stations revealed high levels <strong>of</strong> carbonate<br />

carbon (CC). Figure 1 shows a TLT-plot<br />

(<strong>the</strong>rmo-optical method with laser correction) <strong>of</strong> a<br />

street dust sample highly burdened with CC in comparison<br />

to <strong>the</strong> ambient air sample taken at <strong>the</strong> same<br />

day and station “Kendlerstraße” in Vienna, Austria.<br />

The laser signal indicates increasing transmission and<br />

<strong>the</strong> burn <strong>of</strong>f <strong>of</strong> elemental carbon, mainly to be found<br />

in <strong>the</strong> ambient PM10 sample. Organic carbon burns<br />

<strong>of</strong>f at lower temperatures (below 350°C), a charring<br />

artefact could also be corrected with <strong>the</strong> laser signal.<br />

The quantification <strong>of</strong> CC results in 8% <strong>of</strong> total mass<br />

in street dust samples; calculated as CO3 2- <strong>the</strong> mass<br />

fraction in street dust 41%.


dCO 2/dT Street Dust<br />

dCO2/dT PM10<br />

1.1<br />

1.0<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

Vienna, Kendlerstraße March<br />

Organic Carbon<br />

Elemental Carbon<br />

Carbonate Carbon<br />

50 100 150 200 250 300 350 400 450 500 550 600 650<br />

0<br />

700<br />

Temperature [°C]<br />

Figure 1: TLT-plot <strong>of</strong> street dust sample and ambient<br />

air sample taken at <strong>the</strong> sampling site “Kendlerstraße”<br />

An investigation <strong>of</strong> <strong>the</strong> samples revealed, that <strong>the</strong><br />

sampled street dust material taken at “Kendlerstraße”<br />

mostly consists <strong>of</strong> dolomite (Ca- and MgCO3), which<br />

was used as gritting material during winter in <strong>the</strong><br />

sampling period.<br />

The source contribution <strong>of</strong> fine particulate matter<br />

arising from street dust to <strong>the</strong> ambient air concentration<br />

at that day results in about 70% (calculated from<br />

<strong>the</strong> slope between <strong>the</strong> correlation <strong>of</strong> <strong>the</strong> CO3 2- , Ca 2+<br />

and Mg 2+ content <strong>of</strong> <strong>the</strong> street dust and ambient air<br />

sample). These results were confirmed with <strong>the</strong> CMB<br />

model as well.<br />

Figure 2 shows <strong>the</strong> local and seasonal variation <strong>of</strong><br />

carbonates at two sampling sites in Vienna. The<br />

traffic related site was obviously more affected by<br />

<strong>the</strong> use <strong>of</strong> gritting materials than <strong>the</strong> urban background<br />

site.<br />

2- in PM10<br />

% CO 3<br />

40<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0<br />

1<br />

2<br />

3<br />

4<br />

5<br />

6<br />

7<br />

8<br />

9<br />

10<br />

11<br />

12<br />

13<br />

14<br />

12123412121212121212345678912345123456<br />

Comparison <strong>of</strong> Carbonate Content at 2 sampling sites<br />

Jan Feb Mar AprMayJun JulAugSep Oct Nov Dec<br />

Urban traffic related Vienna Urban background Vienna<br />

Figure 2: Comparison <strong>of</strong> CO3 2- content at a traffic<br />

related and an urban background site in Vienna<br />

Figure 2 demonstrates that some single events<br />

(point 8 in January and point 2 in March) can amount<br />

in about 35% CO3 2- <strong>of</strong> total PM10 mass (about 55%<br />

calculated as Ca- and MgCO3).<br />

The mean contribution <strong>of</strong> all mineral dust sources<br />

to ambient PM10 samples estimated by <strong>the</strong> macrotracer<br />

concept and <strong>the</strong> CMB model was in <strong>the</strong> range<br />

<strong>of</strong> 10 up to 70%, showing <strong>the</strong> high variation <strong>of</strong> this<br />

source due to special occurrences like gritting and<br />

construction sites.<br />

4<br />

3<br />

2<br />

1<br />

Laser [V]<br />

208<br />

ACKNOWLEDGMENTS<br />

This work was part <strong>of</strong> <strong>the</strong> AQUELLIS IGL – project<br />

(Emissions <strong>of</strong> industry, craft and agriculture)<br />

funded by <strong>the</strong> Austrian Federal Ministry <strong>of</strong> Agriculture,<br />

Forestry, Environment and Water Management<br />

and <strong>the</strong> AQUELLA Vienna, AQUELLA<br />

Graz, AQUELLA Salzburg and AQUELLA Linz<br />

projects. Special thanks to all members <strong>of</strong> <strong>the</strong><br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

at <strong>the</strong> Vienna University <strong>of</strong> Technology for doing<br />

parts <strong>of</strong> <strong>the</strong> analysis work.<br />

REFERENCES<br />

[1] European Air Qualitative Directive 199/30/EC<br />

[2] X. Querol, A. Alastuey, C.R. Ruiz, B. Artinano,<br />

H.C. Hansson, R.M. Harrison, E. Buringh, H.M.<br />

ten Brink, M. Lutz, P. Bruckmann, P. Straehl, J.<br />

Schneider. Speciation and origin <strong>of</strong> PM10 and<br />

PM2.5 in selected European cities. Atmospheric<br />

Environment 38, pages 6547-6555, 2004.<br />

[3] J.-G. Bartaire. Clean Air for Europe. Pollution<br />

Atmospherique 172, pages 493-497, 2001.<br />

[4] EN12341, Air Quality – Determination <strong>of</strong> <strong>the</strong><br />

PM10 fraction <strong>of</strong> suspended particulate matter –<br />

Reference method and field testing procedure to<br />

demonstrate reference equivalence <strong>of</strong> measurement<br />

methods, 1998.<br />

[5] H. Puxbaum, Thermo-Gasanalysator zur Charakterisierung<br />

von Kohlenst<strong>of</strong>f- und Schwefelverbindungen<br />

in Luftgetragenen Stäuben.<br />

Fresenius Zeitschrift für Analytische Chemie<br />

298, 250-259, 1979.<br />

[6] H. Puxbaum, J. Rendl., Ein automatisches<br />

Analysatorensystem zur Bestimmung von<br />

Kohlenst<strong>of</strong>f und Schwefel in Luftgetragenen<br />

Stäuben. Microchemia Acta, 263-272, 1983.<br />

[7] A. Caseiro, I.L. Marr, M. Claeys, A. Kasper-<br />

Giebl, H. Puxbaum, C.A. Pio, Determination<br />

<strong>of</strong> saccharides in atmospheric aerosol using<br />

anion-exchange high-performance liquid<br />

chromatography and pulsed-amperometric detection.<br />

Journal <strong>of</strong> Chromatography A. 1171<br />

(1-2), 37-45, 2007.<br />

[8] H. Puxbaum, M. Tenze-Kunit, Size distribution<br />

and seasonal variation <strong>of</strong> atmospheric cellulose.<br />

Atmospheric Environment 37, 3693-<br />

3699, 2003.<br />

[9] A. Limbeck, M. Handler, B. Neuberger, B.<br />

Klatzer, H. Puxbaum, Carbon-Specific-<br />

Analysis <strong>of</strong> Humic like Substances in Atmospheric<br />

Aerosol and Precipitation Samples.<br />

Analytical Chemistry 77 (22), 7288-7293,<br />

2005.


PE-UHMW in hip implants: The influence <strong>of</strong> crosslinking on<br />

structural parameters and micro-mechanical properties<br />

R. Markut-Kohl, T. Koch, V.-M. Archodoulaki, S. Seidler<br />

Institute <strong>of</strong> Materials Science and Technology<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: rmarkut@mail.tuwien.ac.at<br />

Abstract - Ultra-high molecular weight polyethylene<br />

(PE-UHMW), commonly used in hip implants,<br />

is <strong>the</strong> limiting factor within <strong>the</strong> pros<strong>the</strong>sis<br />

due body reactions on wear debris [1]. With <strong>the</strong><br />

objective <strong>of</strong> prolonging <strong>the</strong> durability <strong>of</strong> <strong>the</strong> hip<br />

replacement, radiation-crosslinking <strong>of</strong> <strong>the</strong> PE-<br />

UHMW hip component is more and more applied.<br />

In this way <strong>the</strong> amount <strong>of</strong> wear in <strong>the</strong> pros<strong>the</strong>sis is<br />

reduced [2]. Influences <strong>of</strong> chemical constitution<br />

and morphology on <strong>the</strong> micro-mechanical properties<br />

<strong>of</strong> PE-UHMW hip components due to in vivo<br />

use and in vitro testing are discussed. Fur<strong>the</strong>rmore<br />

influences <strong>of</strong> manufacturing and sterilization<br />

processes on <strong>the</strong> material properties are investigated.<br />

I. INTRODUCTION<br />

The most important medical application <strong>of</strong> PE-<br />

UHMW is <strong>the</strong> replacement <strong>of</strong> damaged cartilage in<br />

total joint replacement surgery. Since 1998 radiation<br />

crosslinked PE-UHMW, developed to optimize<br />

<strong>the</strong> wear behavior in <strong>the</strong> pros<strong>the</strong>sis, replaces<br />

conventional non-crosslinked PE-UHMW in hip<br />

joint replacement [1]. Manufacturing, irradiation<br />

(sterilization as well as crosslinking procedures)<br />

and subsequent <strong>the</strong>rmal treatment (in order to<br />

reduce <strong>the</strong> oxidation tendency <strong>of</strong> <strong>the</strong> material [3])<br />

significantly alter <strong>the</strong> microstructure [4]. This<br />

results in changes <strong>of</strong> <strong>the</strong> mechanical properties <strong>of</strong><br />

<strong>the</strong> material [5]. During <strong>the</strong> life cycle <strong>of</strong> <strong>the</strong> implant<br />

shelf aging and in vivo use influence <strong>the</strong><br />

properties <strong>of</strong> <strong>the</strong> material additionally.<br />

The investigations are part <strong>of</strong> a PhD-project which<br />

aims to correlate structural parameters at molecular<br />

level and at nanometer scale to <strong>the</strong> micro- and<br />

macro-mechanical behavior <strong>of</strong> orthopedic grade<br />

PE-UHMW. Assuming that surrounding medium<br />

and elevated temperature in <strong>the</strong> artificial joint are<br />

involved in changes <strong>of</strong> <strong>the</strong> material, <strong>the</strong> influences<br />

<strong>of</strong> time, temperature, surrounding medium and<br />

mechanical load on <strong>the</strong> properties <strong>of</strong> PE-UHMW<br />

will be figured out. In this contribution, influences<br />

<strong>of</strong> chemical constitution and morphology on <strong>the</strong><br />

209<br />

microhardness will be shown. A depth pr<strong>of</strong>ile <strong>of</strong><br />

<strong>the</strong> micro-hardness was taken over <strong>the</strong> crosssection<br />

<strong>of</strong> <strong>the</strong> PE-UHMW hip implants revealing<br />

local variations within <strong>the</strong> components which can<br />

be attributed to changes in <strong>the</strong> degree <strong>of</strong> crystallinity<br />

and <strong>the</strong> oxidative degradation state.<br />

II. MATERIALS AND METHODS<br />

The physical and mechanical properties <strong>of</strong> three<br />

highly crosslinked PE-UHMW hip components at<br />

different states (shelf aged, after in vitro testing<br />

and after two years in vivo) are compared. As a<br />

reference, medical grade PE-UHMW and a noncrosslinked<br />

retrieved hip implant were investigated.<br />

Fourier Transformed Infrared Spectroscopy<br />

(FTIR) was used to analyze oxidized species,<br />

molecules diffused into <strong>the</strong> material and to evaluate<br />

irradiation conditions. The characterization <strong>of</strong><br />

<strong>the</strong> morphology was carried out via Differential<br />

Scanning Calorimetry (DSC). Instrumented<br />

Nanoindentation technique was used for micro<br />

mechanical characterization.<br />

III. RESULTS AND CONCLUSIONS<br />

Crosslinking and oxidative degradation influence<br />

<strong>the</strong> degree <strong>of</strong> crystallinity and <strong>the</strong> melting behavior<br />

<strong>of</strong> <strong>the</strong> polymer, thus reflecting manufacturing<br />

processes and in vivo use. Oxidative degradation<br />

leads to chain scission and recrystallization. The<br />

resulting higher degree <strong>of</strong> crystallinity can be<br />

correlated to higher hardness values. These dependencies<br />

are shown in non-crosslinked and<br />

highly crosslinked materials.<br />

The depth pr<strong>of</strong>ile <strong>of</strong> <strong>the</strong> non-crosslinked explant<br />

reveals a maximum in microhardness, which can<br />

be correlated to a greater extend <strong>of</strong> oxidation and a<br />

higher degree <strong>of</strong> crystallinity.<br />

The shelf aged highly crosslinked PE-UHMW<br />

implant and <strong>the</strong> in vivo used one show no changes<br />

in hardness, crystallinity and indication <strong>of</strong> oxidative<br />

degradation in <strong>the</strong>ir depth pr<strong>of</strong>iles. The in


vitro tested implant shows oxidative degradation<br />

and a higher degree <strong>of</strong> crystallinity in a region <strong>of</strong><br />

elevated micro-hardness. Such characteristics have<br />

not been reported in <strong>the</strong> field <strong>of</strong> highly crosslinked<br />

PE-UHMW used for implants jet.<br />

The formation <strong>of</strong> crosslinks hinders <strong>the</strong> crystallization<br />

and leads to a reduced degree <strong>of</strong> crystallinity.<br />

Beside this, <strong>the</strong> in vitro tested implant shows a<br />

different melting behavior as shown in Figure 1.<br />

Heat Flow [W/g] (endo up)<br />

Heat Flow [W/g] (endo up)<br />

3,0<br />

2,5<br />

2,0<br />

1,5<br />

1,0<br />

0,5<br />

0,0<br />

3,0<br />

2,5<br />

2,0<br />

1,5<br />

1,0<br />

0,5<br />

0,0<br />

1st Heating Run (10 K/min)<br />

100 120 140<br />

Temperature [°C]<br />

Shelf aged (surface) Shelf aged (back)<br />

In vivo (surface) In vivo (back)<br />

In vitro (2 mm sub-surface) In vitro (back)<br />

non-sterilized PE-UHMW<br />

2nd Heating Run (10 K/min)<br />

100 120 140<br />

Temperature [°C]<br />

Figure 1: DCS results <strong>of</strong> all material conditions<br />

The different properties <strong>of</strong> <strong>the</strong> in vitro tested<br />

highly crosslinked implant can be attributed to<br />

severe chemical degradation, a different crosslinking<br />

state, different manufacturing processes and/or<br />

lot to lot variations.<br />

Although adsorbed and diffused chemical compounds<br />

from <strong>the</strong> surrounding media are observed<br />

in <strong>the</strong> in vivo used samples and <strong>the</strong> in vitro tested<br />

one, no plasticizer effect as hypo<strong>the</strong>size in [6], was<br />

found. Obviously 2a in vivo and <strong>the</strong> duration <strong>of</strong> in<br />

vitro testing are too short that a sufficient amount<br />

<strong>of</strong> molecules can diffuse in <strong>the</strong>se crosslinked hip<br />

components. One o<strong>the</strong>r explanation could be, that<br />

since <strong>the</strong> oxidation level in <strong>the</strong> non-crosslinked<br />

210<br />

explant after 10a in vivo is high, plasticizing effects<br />

interfere with oxidation peaks. For <strong>the</strong> time<br />

being no fur<strong>the</strong>r verifications are possible because<br />

only one material in this state is available.<br />

REFERENCES<br />

[1] S. M. Kurtz; The UHMWPE Handbook: Ultra<br />

High Molecular Weight Polyethylene in Total<br />

Joint Replacement; Elsevier Academic Press<br />

2004<br />

[2] D.W. Manning, P. P. Chiang, J. M. Martell,<br />

J. O. Galante, W. H. Harris; In Vivo Comparative<br />

Wear Study <strong>of</strong> Traditional and<br />

Highly Cross-linked Polyethylene in Total<br />

Hip Arthoplasty; The Journal <strong>of</strong> Arthoplasty<br />

Vol. 20 No. 7 (2005) 880-886<br />

[3] G. Lewis; Properties <strong>of</strong> crosslinked ultrahigh-molecular-weight<br />

polyethylene; Biomaterials<br />

22 (2001) 371-401<br />

[4] S. M. Lee, S. W. Choi, Y. C. Nho, H. H.<br />

Song; Modification <strong>of</strong> Microstructures and<br />

physical Properties <strong>of</strong> Ultra High Molecular<br />

Weight Polyethylene by Electron Beam Irradiation;<br />

Journal <strong>of</strong> Polymer Science: Part<br />

B: Polymer Physics 43 (2005) 3019-3029<br />

[5] L. A. Pruitt; Deformation, yielding, fracture<br />

and fatigue behaviour <strong>of</strong> conventional and<br />

highly cross-linked ultra high molecular<br />

weight polyethylene; Biomaterials 26 (2005)<br />

905-915<br />

[6] L. Costa, P. Bracco, E. Brach del Prever, M.<br />

P. Luda, L. Trossarelli; Analysis <strong>of</strong> products<br />

diffused into UHMWPE pros<strong>the</strong>tic components<br />

in vivo; Biomaterials 22 (2001) 307-<br />

315


������������������������������������������<br />

�������������������������������������������������<br />

�<br />

������������ � ������������������ � �������������� � �������������� � ���������������������� � �<br />

� �������������������������������������������������������������������������������������������<br />

� �������������������������������������������������������������������������������������������������<br />

�������������������������������������<br />

�<br />

�<br />

�������������������������������������������������<br />

������������������������������������������������������<br />

�����������������������������������������������������<br />

����������������������������<br />

��� ����� ����� �� ����������� ���������� ��������� ��� ��<br />

���� ��� ������ ��������� �������� ���������� ��� ����<br />

������������������������������������������������<br />

����������������������������������������������������<br />

�������������� ������������������������ ��� ���������<br />

������� ����������������� ���������� ���� ������� ���<br />

���������� ������������ �������������� �������� ���<br />

�������������������������������������������� ��������<br />

��������� ���� ���������� ����� ���� ��������� ��� ����<br />

������������������������������������������������<br />

��� �������������<br />

�������� ������������ ���� ���������� ��� ������� ��� ����<br />

�������������������������������������������������������<br />

��������������������������������������������������������<br />

������������������ ��������������������� ���������������<br />

��������������������������������������������������<br />

�������������������������������������������������������<br />

���������������������������������<br />

���� �����������������<br />

��� ������ ��� ������ ������������ ������������� �����<br />

���� ���������� ���� ���� ������������ ���� ����������� ���<br />

������������������������������������������������������<br />

�������������������������������������������������������<br />

��� ��������� ������� ��� ������������ ���������������<br />

�������������������<br />

�<br />

� �<br />

�<br />

���<br />

��<br />

�<br />

�<br />

�<br />

��<br />

��� �������� ����� ����� �� ���� ��� �������� ������ ���������<br />

������� ������� ���������� ����� ����� ����� ������<br />

���������������� ���� ������������� ����������� ���<br />

������� ��� ����������� ���� ����� ������������� ���� ����<br />

����� ��� ���� ��������� ��������� ��� ������ ��� ������� ����<br />

�������������������������������������������������<br />

����������������������������������������������������<br />

���������������������������������������������������<br />

���������������������������������������������������<br />

�� ����������������������� ��� ���������� ���������<br />

��������������������������� ��� ������ ��� ������� ���<br />

�������������������������������������������������<br />

����� ������������������<br />

���������������������������������������������������<br />

��� ���� ��������� �������� ��������� ��� �������� ����<br />

���������� ������ ����������� ������������ �����������<br />

����� �������� ������������� ���� ��������� ����� �����<br />

�������������� ��� ���������� ����� ���� ����� ��������<br />

������������������������������������������������������<br />

���������������������� �������������������������������<br />

����������������������������������������������������<br />

����������������������������������������<br />

�����������<br />

���� ������������������������������������������<br />

���� �������������������������������������������<br />

��������������������������������������������<br />

���� �������������������������������<br />

� ������������������������������������������������<br />

��� ������ ������<br />

�<br />

������<br />

�����<br />

��<br />

��<br />

�<br />

�������� � �����������<br />

�<br />

�������������������������������������������������<br />

211<br />

�<br />

��� ������<br />

�<br />

�<br />

���<br />

�������������<br />

�����


212


Shadowgrahic Imaging <strong>of</strong> Femto Second Laser<br />

Induced Ablation <strong>of</strong> Al<br />

Shazia Bashir, M.Shahid Rafique and Wolfgang Husinsky<br />

Institut für Allgemeine Physik, Vienna University <strong>of</strong> Technology,<br />

Abstract- Shadowgraphy technique is used to<br />

investigate <strong>the</strong> ablation <strong>of</strong> Al as a function <strong>of</strong> <strong>the</strong><br />

laser fluence. Ablation is done with an 800 nm, 25fs<br />

Ti: Sapphire laser. The shadowgraphs <strong>of</strong> ablation<br />

are captured by CCD and imaging is done by <strong>the</strong><br />

probe beam having wavelength 400 nm (produced<br />

by frequency doubling <strong>of</strong> Ti: Sapphire laser).The<br />

experimental data shows that <strong>the</strong> phenomenological<br />

development <strong>of</strong> <strong>the</strong> ablation dynamics is dependent<br />

on <strong>the</strong> number <strong>of</strong> pulses and laser fluence. The<br />

phenomenon <strong>of</strong> particles emission, nucleation<br />

growth, bursting <strong>of</strong> bubbles and droplets at <strong>the</strong><br />

target surface can be attributed to hydrodynamic<br />

instability.<br />

I. Introduction<br />

Ultra fast time-resolved optical shadowgraphy with<br />

pump-probe technique allowing time- and spaceresolved<br />

measurements <strong>of</strong> <strong>the</strong> changes in <strong>the</strong><br />

expansion dynamics <strong>of</strong> <strong>the</strong> femtosecond laser<br />

ablation is very useful in picoseconds to<br />

nanosecond time scale. This imaging analysis can<br />

provide information about <strong>the</strong> ablation efficiency as<br />

a function <strong>of</strong> <strong>the</strong> laser fluence .Then this<br />

information about <strong>the</strong> ejected species can be<br />

correlated with Raman spectroscopy, time <strong>of</strong> flight<br />

mass spectrometry, SEM and AFM analysis. This<br />

ultrafast shadow graphic imaging can be used to<br />

investigate femtosecond laser ablation <strong>of</strong> variety <strong>of</strong><br />

materials like CaF2,<br />

Si, Al, Au, CR-39 etc<br />

A-1040 Wien, Austria<br />

Email: bashir@iap.iap.tuwien.ac.at<br />

213<br />

II. Layout<br />

Schematic <strong>of</strong> shadowgraphy setup is shown in<br />

“Figure 1”. A Ti: saphire 800 nm 25 fs laser<br />

operated at 1 KHz repetition rate has been used in<br />

<strong>the</strong> experiment. Beam was split into two parts with<br />

<strong>the</strong> help <strong>of</strong> a beam splitter. First part (pump beam)<br />

was focused to a spot size <strong>of</strong> 50 �m in diameter<br />

onto a target using 36 cm focal-length lens to<br />

produce ablation. Second part is converted into 400<br />

nm with a BBO crystal and is used as a probe beam.<br />

A fast response CCD along with fly capture<br />

s<strong>of</strong>tware has been used to grab <strong>the</strong> image made by<br />

<strong>the</strong> probe beam.<br />

III. Results<br />

Figure 2 is a sequence <strong>of</strong> frames <strong>of</strong> ablated Al with<br />

2<br />

laser fluence 3.6 J/cm for different number <strong>of</strong><br />

pulses. These frames are taken from an 80 second<br />

movie which recorded <strong>the</strong> real time ablation <strong>of</strong> Al<br />

when irradiated with femtosecond pump beam<br />

operated at a repetition rate <strong>of</strong> 1 kHz. The sequence<br />

shows a gradual appearance <strong>of</strong> <strong>the</strong> ablated material<br />

with <strong>the</strong> increased number <strong>of</strong> pulses or exposure<br />

time. The sequences clearly indicate <strong>the</strong> liquid–<br />

phase expulsion, and vaporization .increases with<br />

increasing number <strong>of</strong> pulses There seems to be a<br />

nucleation growth and bursting <strong>of</strong> bubbles and<br />

droplets which can be attributed to hydrodynamic<br />

instability. One <strong>of</strong> <strong>the</strong> basic quantities in laser<br />

ablation is temperature rise induced by<br />

<strong>the</strong> absorbed laser radiation on material surface.


This localized temperature rise can be calculated by<br />

solving three dimensional heat equations. In <strong>the</strong><br />

liquid-phase expulsion technique, high pressure<br />

gas-jet is formed which expels <strong>the</strong> liquefied<br />

material as clear from “Figure.2.”Molten material<br />

accumulates near <strong>the</strong> target surface due to surface<br />

tension. Melting and evaporation are widely<br />

exploited in <strong>the</strong> treatment <strong>of</strong> materials by laser<br />

pulses. It has been recognized that shorter laser<br />

pulses may strongly superheat <strong>the</strong> melt, causing a<br />

homogeneous nucleation <strong>of</strong> gas bubbles. Their rate<br />

rises near <strong>the</strong> critical temperature. In <strong>the</strong> end, <strong>the</strong><br />

liquid is believed to disrupt into a mixture <strong>of</strong> gas<br />

and droplets by this phase explosion. The<br />

development <strong>of</strong> <strong>the</strong>se melt ejections are<br />

complicated depending upon <strong>the</strong> energy deposition<br />

and material ejection mechanism.<br />

REFERENCES<br />

Femto second<br />

laser<br />

CCD<br />

Interference<br />

filter<br />

HR mirror<br />

Imaging lens<br />

[1] X. Zeng, X.L. Mao, R. Greif, R.E. Russo<br />

Appl. Phys. A 80, 237–241 (2005)<br />

[2] H-.Dachraoui W. Husinsky Applied Phys<br />

Lett. 89, 104102(2006)<br />

[3] N. Zhang, X. Zhu, J. Yang, X. Wang,<br />

Phys. Rev. Lett. 99, 167602 (2007)<br />

800nm<br />

Beam<br />

splitter<br />

Ablated<br />

material<br />

Target<br />

Unexpose<br />

BBO<br />

crystal<br />

Focussing<br />

lens<br />

Probe beam<br />

400nm<br />

Figure 1 : A Schematic <strong>of</strong> shadowgraphy experimental<br />

set up<br />

214<br />

5000<br />

5500 5600<br />

5000<br />

7500 9500 9800 11000<br />

7500 12000 15000 9500 30000 300<br />

66000 80000<br />

30000 57000<br />

100 μm<br />

Figure 2: An ablation sequence at fluence <strong>of</strong> 3.6 J/cm 2 .


3D-analysis <strong>of</strong> neuronal and vascular networks by ultramicroscopy<br />

Nina Jährling 1,2,3 , Stefan Kalchmair 1,2 , Klaus Becker 1,2 , Andreas Gewies 4 ,<br />

Edgar R. Kramer 5 and Hans-Ulrich Dodt 1,2<br />

1 FKE, Dept. <strong>of</strong> Bioelectronics, TU Vienna, Vienna, Austria<br />

2 Center <strong>of</strong> Brain Research, Medical Univ. <strong>of</strong> Vienna, Vienna, Austria<br />

3 University <strong>of</strong> Oldenburg, Dept. <strong>of</strong> Neurobiology, Oldenburg, Germany<br />

4 TU Munich, Klinikum rechts der Isar, 3. Med. Dept., Munich, Germany<br />

5 ZMNH, Dept. <strong>of</strong> Molecular Neurobiology, Hamburg, Germany<br />

Email: nina.jaehrling@meduniwien.ac.at<br />

Abstract — Ultramicroscopy has recently been shown as a<br />

useful imaging technique for three-dimensional visualization<br />

<strong>of</strong> biological specimens. It allows 3D-reconstructions <strong>of</strong> up<br />

to cm-sized samples with μm resolution. We present ultramicroscopic<br />

3D reconstructions <strong>of</strong> whole immunhistochemically<br />

labelled mouse embryos, giving detailed insight into <strong>the</strong>ir neuronal<br />

and vascular networks. We fur<strong>the</strong>r obtained reconstructions<br />

<strong>of</strong> parts <strong>of</strong> <strong>the</strong> blood vessel network. For analysis images<br />

were vectorized parts. Generally, ultramicroscopy is especially<br />

suitable for investigations <strong>of</strong> biological networks, because mechanical<br />

slicing aretfacts are avoided.<br />

I. INTRODUCTION<br />

In <strong>the</strong> last decade a lot <strong>of</strong> innovative imaging techniques<br />

were developed, giving new impulses in biological sciences.<br />

One <strong>of</strong> <strong>the</strong>se techniques is ultramicroscopy, which<br />

allows three-dimensional reconstructions <strong>of</strong> cm-sized<br />

objects with micrometer resolution [1, 2, 3]. Thus, ultramicroscopy<br />

bridges a gap between confocalmicroscopy<br />

and macroscopic imaging techniques, like computer tomography<br />

and optical resonance tomography [4].<br />

II. ULTRAMICROSCOPY<br />

Ultramicroscopy is based on <strong>the</strong> principle <strong>of</strong> light sheet<br />

illumination, i.e. <strong>the</strong> specimen is illuminated perpendicular<br />

to <strong>the</strong> imaging pathway by a fine sheet <strong>of</strong> laserlight,<br />

formed by clinderical lenses (Fig. 1). Thus, out <strong>of</strong> focus<br />

parts <strong>of</strong> <strong>the</strong> specimen are not illuminated, so that <strong>the</strong>y<br />

cannot contribute to image blurring or stray light generation.<br />

Therefore, many drawbacks <strong>of</strong> histology, like<br />

mechanical distortions and misalignments <strong>of</strong> slices are<br />

avoided.<br />

Since ultramicroscopy requires specimens to be transparent,<br />

most specimens have to be chemically cleared before.<br />

To render <strong>the</strong> specimens transparent, <strong>the</strong>y are incubated<br />

in a medium, having approximatley <strong>the</strong> same<br />

refractive index as protein. Thus, light scattering is<br />

strongly reduced, and <strong>the</strong> specimens appear translucent.<br />

This clearing procedure was invented 1914 by <strong>the</strong> German<br />

anatomist W. Spalteholz [5].<br />

215<br />

Figure 1: The sample is illuminated from side by a laser<br />

forming a thin sheet <strong>of</strong> light. A 488 nm diode laser with<br />

500 mW power (Coherent, Germany) was used for excitation.<br />

Fluorescent light is <strong>the</strong>refore only emitted from<br />

a thin plane. The light is <strong>the</strong>n collected by an objective<br />

lens and projected onto a camera target. The excitation<br />

light was blocked by an optical filter (525/50nm). Fig.<br />

from [3].<br />

We immunhistochemically labelled mouse embryos with<br />

a conjugated secondary fluorescent antibody before<br />

clearing. For imaging, an Olympus XL 4x objective with<br />

a N.A. <strong>of</strong> 0.28 was used. Images were recorded with<br />

CoolSnap K4 CCD-camera. 3D image reconstruction<br />

was done with 3D reconstruction s<strong>of</strong>tware (Amira 4.1,<br />

Mercury Computer Systems, Germany).<br />

III. RESULTS AND DISCUSSION<br />

A. PROJECT: NERVE GROWTH<br />

In an ongoing study we applied ultramicroscopy to investgations<br />

<strong>of</strong> nerve growth in mouse embryos, labelled


with neur<strong>of</strong>ilament-160 antibody, conjugated with <strong>the</strong><br />

fluorescent dye Alexa488 (Fig. 2). By detecting <strong>the</strong> fluorescence<br />

signal, neuronal structures can be precisley localized<br />

in three dimensions.<br />

Figure 2: 3D-reconstruction <strong>of</strong> limb-innervating nerves<br />

<strong>of</strong> a mouse embryo E12.5 (PN:peroneal nerve, TN: tibial<br />

nerve). Size <strong>of</strong> scale bar 100 μm.<br />

B. PROJECT: VASCULAR NETWORKS<br />

Fur<strong>the</strong>rmore, we obtained reconstructions <strong>of</strong> blood vessel<br />

networks (Fig. 3). In this ongoing project, we plan to<br />

investigate <strong>the</strong> complexity <strong>of</strong> vascular networks in mouse<br />

embryos. Therefore <strong>the</strong> data are vectorized (Fig. 3).<br />

Figure 3: Blood vessel network in a mouse embyro<br />

E12.5. Parts <strong>of</strong> <strong>the</strong> network were vectorized for fur<strong>the</strong>r<br />

analysis. Size <strong>of</strong> scale bar 100 μm.<br />

IV. CONCLUSION<br />

Ultramicroscopy has been validated as a valuable tool<br />

for <strong>the</strong> three-dimensional analysis <strong>of</strong> complex anatomical<br />

structures. As in ultramicroscopy mechanical artefacts<br />

are avoided, even large specimen can be threedimensonally<br />

reconstructed with outstanding resolution.<br />

216<br />

ACKNOWLEDGMENTS<br />

The study was supported by SFB 391 and <strong>the</strong> Hertie<br />

Foundation.<br />

REFERENCES<br />

[1] H.U. Dodt, U. Leischner, A.Schierloh,<br />

N. Jährling, C.P. Mauch, J.M. Deussing, M. Eder,<br />

W. Zieglgänsberger, and K. Becker. Ultramicroscopy:<br />

three-dimensional visualization <strong>of</strong><br />

neuronal networks in <strong>the</strong> whole mouse brain. Nature<br />

Methods, 4(4):331–336, April 2007.<br />

[2] K. Becker, N. Jährling, E.R. Kramer, F. Schnorrer,<br />

J.M. Deussing, and H.U. Dodt. Ultramicroscopy:<br />

3D reconstruction <strong>of</strong> large microscopic specimens.<br />

Journal <strong>of</strong> Biophotonics, 1(1):36–42, January <strong>2008</strong>.<br />

[3] N. Jährling, K. Becker, E.R. Kramer, and H.U. Dodt.<br />

3D-Visualization <strong>of</strong> nerve fiber bundles by ultramicroscopy.<br />

Medical Laser Application, <strong>2008</strong> in press.<br />

[4] J.R. Walls, L. Coultas, J. Rossant, and R.M. Henkelmann.<br />

Three-dimensional analysis <strong>of</strong> vascular development<br />

in <strong>the</strong> mouse embryo. PLoS one, 1(1):1,<br />

August <strong>2008</strong>.<br />

[5] W. Spalteholz. Über das Durchsichtigmachen von<br />

menschlichen und tierischen Präparaten. S.Hierzel,<br />

Leipzig, Germany, 1914.


�����������������������������������������������������������<br />

����������������<br />

�<br />

�������������������������������������������������������������������������<br />

�����������������������������������������<br />

��������������������������������<br />

�����������������<br />

�����������������������������������������<br />

�<br />

�<br />

�������������������������������������������������<br />

�� ������� ������ ���� ����������� ������������ ���<br />

������� ����� ��� ������� ������� ���� ������� ���������<br />

���� ��������� ��� ��������������� ��� ����� ��������� ����<br />

�������� ��� ���� ������������ ������������ ���� ������<br />

��������� ���� ����� ������� ���� ������������<br />

����������� ������� ��� ������������ ��� ������ ���<br />

����������������������������������������������������<br />

���� ����� ������ ��� ���� ���������� ���������������<br />

���������������������� ��������������������������<br />

���� ���������� ��������� ��� ���� ��� ���� �������������<br />

��� ���� ���������������� ���������� �������� ���� �����<br />

��������������������������������������������������<br />

����������������������������������������������������<br />

�����������������<br />

��� ����������������������������<br />

�������������������������� �������������������������<br />

�����������������������������������������������������<br />

��������������������������������������������������<br />

�������� ����� �������� �������� ����������� ���<br />

����������������������������������������������������<br />

����������������������������<br />

���� ����� ������� ��������� ��� ������ ���������<br />

����������������������������������������������������<br />

�����������������������������������������������������<br />

��������������������������������������������������<br />

������� ���� ������ ���� ������� �� ����� ������ ���<br />

��������������������������������������������������<br />

���������� ��� ������������ ��� ���� ���� ������ ����<br />

������������ �������������� ������ ��� �� ��� ���� ������<br />

����������������������������������������������������<br />

����������<br />

�<br />

�<br />

����������<br />

���������������<br />

217<br />

���� ����������������<br />

������������������������������������������������������<br />

���� ������������ ��� ���� ���������������� ��������� ����<br />

����������������������������������<br />

�<br />

�<br />

�<br />

� �� � ��<br />

�<br />

��<br />

�<br />

�<br />

��<br />

����� �������� ������<br />

��������������������������������������<br />

�����������������������������������������������������<br />

��������������������������������������������������������<br />

��� ����� ����������� ���������� ��� ���� ����������� �����<br />

����� ����� ����������� ������������ ������ �������<br />

������ ���������� �������� ��� ����� ������������ ���<br />

������������������������������������������������������<br />

�����������������������������������������������������<br />

����������������������������������������������������<br />

�������������������������������������������<br />

����� ����������������<br />

��� �������������<br />

��� ������ ������������� ���� �������� ��� �����������<br />

��������������������������������������������������������<br />

��������� ��� ������ ��� �� ������� ���������� ����<br />

��������������� ���������������� ������� ���� �����<br />

������ ��� ���������� ���������� ���� ���� ����������������<br />

���������� ���� ���������� ��� ������������ ���� �������� ����<br />

�������������� ���������� ������ ������������<br />

�����������������������������������������������������<br />

�������������������������������������������������������<br />

���������������������������������������������������<br />

����������������������������������


��� ���������<br />

���� ��������� ������� �������� ��� ��� �������� �������� ����<br />

������ �� ����������� ��������� ����� ������ ���������� ���<br />

�����������������������������������������������������<br />

����������� ������� ��� ������ ����� ���������� ��� ����<br />

����������� ����� ��� ����� ��� ���� ��������� �������� ����<br />

������� ����� ��������� ��� ���� ���� ��� ����������<br />

�����������������������������������<br />

�<br />

��<br />

��<br />

�<br />

�<br />

�<br />

�<br />

�<br />

�<br />

� �� �<br />

�<br />

��<br />

�<br />

�<br />

��<br />

�<br />

��<br />

�<br />

�<br />

���������<br />

��<br />

�<br />

�<br />

��<br />

�<br />

������������������������������������<br />

��������������������������������<br />

�����������������������������������������<br />

���� ������������� ��� ���� ���������������� ���������<br />

������� ������� ��� �������� ���� ������� ��� �����������<br />

���� �������� �������� ��� ���� �������� ���������� ����<br />

����� ��������� ��� ���� ������ ������������ ���������<br />

�������������������������������������������������<br />

��� �����������������<br />

������������������������������������������������������<br />

����������������������������������������������������<br />

����������������������������������������������������<br />

���� �������������� ���� ��������� ������� ��� �������� ����<br />

������ ���� ����������� ���������� ��� ���� ������<br />

�����������<br />

�<br />

����������������������������<br />

�����������������������������<br />

�<br />

�<br />

��<br />

218<br />

�<br />

��<br />

����������������������<br />

�����������������<br />

���� ������� ������ ����� ��� ������ ������� ����� ����<br />

�������� ������� ���� ����������� ����� ��� ����<br />

������������� ����� ������� �� ���������� ��������<br />

��������� ��������� ����� �������� ���� ������� ��� ����<br />

��������� �������� ��������� ��� �������� ���� ���������<br />

���� ����������� ������������ ����������� ����� ��������<br />

��������������������������<br />

�����������<br />

����������������������������������������������������<br />

�����������������������������������<br />

����������������������������������������������������<br />

��������������������������������������<br />

������������������������������������<br />

� ������������������������������������������������<br />


Optimisation <strong>of</strong> <strong>the</strong> incoupling geometry <strong>of</strong> a pulsed<br />

solid-state laser<br />

E. Schwarz, J. Tauer, H. K<strong>of</strong>ler and E. Wintner<br />

Institut für Photonik<br />

Technische Universität Wien<br />

Vienna, Austria<br />

Email: e9925266@student.tuwien.ac.at,<br />

ernst.wintner+e387@tuwien.ac.at<br />

Abstract — The topic <strong>of</strong> this work is <strong>the</strong> analysis <strong>of</strong><br />

<strong>the</strong> incoupling geometry for an ignition laser<br />

concept. The ignition laser is a compact high peak<br />

power, passively Q-switched, longitudinally diodepumped<br />

solid-state laser. The incoupling optic<br />

consists <strong>of</strong> <strong>the</strong> pump fiber and an aspheric<br />

collimating lens. For optimal operation <strong>of</strong> a pulsed<br />

solid-state laser not only <strong>the</strong> usual parameters<br />

(pump power, OC mirror and absorber) should be<br />

taken into account, also <strong>the</strong> incoupling geometry<br />

has to be adapted to <strong>the</strong> structural conditions.<br />

I. INTRODUCTION<br />

A fully functional laser ignition concept for gas<br />

engines was developed in seven years by <strong>the</strong> “Laser<br />

Ignition Group” [1-4] at <strong>the</strong> “Institut für Photonik”.<br />

The outcome was a compact high peak power,<br />

passively Q-switched, longitudinally diode-pumped<br />

solid-state laser. One day this “laser spark plug”<br />

should replace conventional spark plugs because <strong>of</strong><br />

<strong>the</strong> fact that conventional ignition has reached it limit<br />

in terms <strong>of</strong> efficiency and reduction <strong>of</strong> pollutant<br />

emissions<br />

II. EXPERIMENTAL SETUP<br />

The components <strong>of</strong> <strong>the</strong> laser system used in this work<br />

are <strong>the</strong> pump fiber, an aspheric incoupling lens, <strong>the</strong><br />

active medium, a passive saturable absorber and <strong>the</strong><br />

output mirror (see Figure 1).<br />

Figure 1: Experimental setup<br />

219<br />

The reflectivity <strong>of</strong> <strong>the</strong> output mirror should be about<br />

50% whereas <strong>the</strong> initial transmission <strong>of</strong> <strong>the</strong> absorber<br />

should approximately be 40%. As active medium a<br />

Nd:YAG crystal with a diameter <strong>of</strong> 4 mm and a<br />

length <strong>of</strong> 15 mm was used and Cr 4+ :YAG as<br />

saturable absorber medium.<br />

For pumping, a laser diode emitting at 808 nm with<br />

an output power <strong>of</strong> up to 400 W and a pulse duration<br />

<strong>of</strong> up to 300 μs was employed. The length <strong>of</strong> a laser<br />

pulse is about 1 ns.<br />

The incoupling geometry consists <strong>of</strong> <strong>the</strong> pump fiber<br />

and an aspheric collimating lens with a focal length<br />

<strong>of</strong> 8 mm. Fur<strong>the</strong>rmore, <strong>the</strong> distance between pump<br />

fiber and collimating lens was continuously variable.<br />

III. EXPERIMENTAL RESULTS<br />

By changing <strong>the</strong> distance between fiber and lens <strong>the</strong><br />

dependency <strong>of</strong> <strong>the</strong> pulse energy and <strong>the</strong> temporal<br />

pulse position <strong>of</strong> <strong>the</strong> laser pulse on <strong>the</strong> configuration<br />

<strong>of</strong> <strong>the</strong> incoupling optic was analysed. The distance<br />

between fiber and lens primarily is related to <strong>the</strong><br />

diameter <strong>of</strong> <strong>the</strong> pump beam.<br />

Figure 2: Pulse energy as a function <strong>of</strong> distance V.


That means, for increasing distance between fiber<br />

and lens <strong>the</strong> pump beam diameter is decreasing,<br />

consequently <strong>the</strong> pump volume becomes smaller and<br />

less energy is deposited in <strong>the</strong> crystal.<br />

In <strong>the</strong> case <strong>of</strong> decreasing distance between fiber and<br />

lens, <strong>the</strong> pump beam diameter becomes smaller.<br />

Hence <strong>the</strong> pump volume enlarges, more energy can<br />

be deposited in <strong>the</strong> crystal and consequently <strong>the</strong> pulse<br />

energy is higher (see Figure 2).<br />

For optimal operation <strong>of</strong> a pulsed solid-state laser not<br />

only <strong>the</strong> usual parameters (pump power, OC mirror<br />

and absorber) should be taken into account, also <strong>the</strong><br />

incoupling geometry has to be adapted to <strong>the</strong><br />

structural conditions.<br />

REFERENCES<br />

[1] H. K<strong>of</strong>ler, J. Tauer, G. Tartar, K. Iskra, J.<br />

Klausner, G. Herding and E. Wintner, An<br />

innovative solid-state laser for engine ignition.<br />

Laser Phys. Lett. 4, 2007. No. 4: p. 322-327.<br />

[2] Tauer, Johannes, Development <strong>of</strong> an Ignition<br />

Laser. 2006, TU-Wien.<br />

[3] Weinrotter, M., Laser Ignition <strong>of</strong> Internal<br />

Combustion Engines- Basic Laser and Ignition<br />

Optics Developments, Engine Application and<br />

Optical Diagnostics. 2006, TU Wien.<br />

[4] Kopecek, H., "Laser Ignition <strong>of</strong> Gas Engines".<br />

2004, Vienna University <strong>of</strong> Technology.<br />

220


Enhancement Mode HEMTs: Evaluation <strong>of</strong> Two Approaches by<br />

Numerical Simulation<br />

Stanislav Vitanov and Vassil Palankovski (Faculty Mentor)<br />

Advanced Material and Device Analysis Group<br />

Institute for Microelectronics, TU Wien<br />

Vienna, Austria<br />

Email: {vitanov,palankovski}@iue.tuwien.ac.at<br />

Abstract — Normally-<strong>of</strong>f operation <strong>of</strong> high electron mobility<br />

transistors is desired for many reasons, however proved to<br />

be difficult to achieve, despite <strong>the</strong> rapid development <strong>of</strong> <strong>the</strong> depletion<br />

mode devices. Amongst <strong>the</strong> few approaches proposed so<br />

far, we focus on two, which promise high performance. The first<br />

device features an InGaN cap layer, while <strong>the</strong> later relies on<br />

gate recess technology. We perform DC and AC analyses, using<br />

our device simulator calibrated against experimental data,<br />

and compare <strong>the</strong> performance <strong>of</strong> <strong>the</strong> devices based on simulation<br />

results.<br />

I. INTRODUCTION<br />

The transport and material properties <strong>of</strong> GaN and AlN<br />

and <strong>the</strong>ir heterostructures have encouraged <strong>the</strong> research<br />

<strong>of</strong> AlGaN/GaN based transistors for various applications<br />

in <strong>the</strong> last decade. Consequently, outstanding results<br />

have been reported for <strong>the</strong> depletion mode (D-mode)<br />

high electron mobility transistors (HEMTs). However,<br />

for several applications enhancement mode (E-mode) devices<br />

are essential. In analog electronics <strong>the</strong>y supersede<br />

<strong>the</strong> negative voltage supply and also assure a safe state<br />

in case <strong>of</strong> power loss. In digital electronics, <strong>the</strong>y allow<br />

complementary logic.<br />

Despite <strong>the</strong> interest in E-mode operation, <strong>the</strong> excellent<br />

results as in D-mode devices remain to be seen. The first<br />

E-mode transistor was reported back in 1996 by Khan et<br />

al. [1], and several o<strong>the</strong>r refinements followed. In this<br />

paper we focus on two recently proposed approaches.<br />

The first relies on gate recess combined with fluorine<br />

based surface treatment, while <strong>the</strong> second incorporates<br />

a thin InGaN layer, which raises <strong>the</strong> conduction band.<br />

For <strong>the</strong> study <strong>of</strong> <strong>the</strong> devices we use our two-dimensional<br />

device simulator MINIMOS-NT.<br />

II. DEVICE STRUCTURE WITH INGAN-CAP<br />

The InGaN/AlGaN/GaN device structure was proposed<br />

by Mizutani et al. [2]. A 3 μm thick GaN layer is grown<br />

on sapphire substrate. A 20 nm thick Al0.25Ga0.75N supply<br />

layer follows (5 nm undoped, 10 nm highly-doped,<br />

and 5 nm undoped). A 5 nm non-intentionally doped<br />

In0.2Ga0.8N layer is deposited next. The gate length<br />

lg=1.9 μm, source-gate distance is 1.5 μm, and gatedrain<br />

distance is 2.4 μm.<br />

221<br />

III. DEVICE STRUCTURE WITH RECESSED<br />

GATE<br />

The gate recess structure and its fabrication is reported<br />

by Palacios et al. [3]. The 11 nm thick GaN channel is<br />

grown on-top <strong>of</strong> a 1 nm thick In0.1Ga0.9N back-barrier.<br />

A 1 nm thick AlN layer between <strong>the</strong> channel and <strong>the</strong><br />

25 nm Al0.33Ga0.67N is grown in order to improve <strong>the</strong><br />

electron mobility. After <strong>the</strong> AlGaN surface treatment a<br />

12 nm gate recess is performed, resulting in a gate-tochannel<br />

distance <strong>of</strong> 13 nm. The gate length lg is 160 nm,<br />

source-gate distance is 0.6 μm, and gate-drain distance is<br />

0.9 μm.<br />

IV. SIMULATION SETUP<br />

ThedevicesimulatorMINIMOS-NT has proven to be a<br />

suitable tool for <strong>the</strong> analysis <strong>of</strong> heterostructure devices<br />

[4]. Recently, it has been used for <strong>the</strong> study <strong>of</strong> a whole<br />

generation <strong>of</strong> AlGaN/GaN HEMTs [5]. Since <strong>the</strong> longitudinal<br />

electric field in <strong>the</strong> channel reaches peak values<br />

<strong>of</strong> above 500 kV/cm, <strong>the</strong> hydrodynamic transport model<br />

is used to properly model electron transport and energy<br />

relaxation. Self-heating effects are accounted for by using<br />

a global self-heating model, which calculates a spatially<br />

constant lattice temperature. The value <strong>of</strong> <strong>the</strong> sheet<br />

charge at <strong>the</strong> AlGaN/GaN interface induced by <strong>the</strong> polarization<br />

effects is derived from <strong>the</strong> DC characteristics.<br />

Additional charges at <strong>the</strong> InGaN and AlGaN interfaces<br />

are accounted for [6].<br />

V. SIMULATION RESULTS<br />

Figure 1 shows <strong>the</strong> results for <strong>the</strong> transfer characteristics<br />

<strong>of</strong> both devices. After a calibration <strong>of</strong> <strong>the</strong> sheet charges<br />

a good agreement is achieved. The InGaN/GaN device<br />

exhibits lower current, however a higher threshold voltage<br />

is achievable, without recessing <strong>the</strong> InGaN cap layer<br />

[2]. The threshold voltage <strong>of</strong> <strong>the</strong> recess device can be<br />

increased too (Figure 2) by increasing <strong>the</strong> recess depth.<br />

Figure 3 compares <strong>the</strong> DC transconductance gm for<br />

both devices. The decrease in <strong>the</strong> measured gm <strong>of</strong> <strong>the</strong><br />

InGaN/AlGaN/GaN transistor at higher gate bias, might<br />

be due to non-idealities in <strong>the</strong> source and drain ohmic<br />

contacts, which are not considered in <strong>the</strong> simulation.


I D [A/mm]<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

exp. data<br />

simulation<br />

0.0<br />

−1.0 −0.5 0.0 0.5 1.0 1.5 2.0<br />

V GS [V]<br />

gate recess<br />

InGaN cap<br />

Figure 1: Comparison <strong>of</strong> <strong>the</strong> measured (symbols) and<br />

simulated (lines) transfer characteristics at Vds=5 V.<br />

As expected, <strong>the</strong> recessed gate device exhibits a higher<br />

gm due to <strong>the</strong> much shorter gate length lg and <strong>the</strong> reduced<br />

gate-to-channel separation.<br />

We also perform an AC analysis <strong>of</strong> <strong>the</strong> transistors.<br />

The recessed gate structure (lg=0.16 μm) exhibits transit<br />

cut-<strong>of</strong>f frequency fT=85 GHz, while our simulation<br />

gives fT=10 GHz for <strong>the</strong> InGaN/AlGaN/GaN device<br />

(lg=1.9 μm). Note, that <strong>the</strong> product fT×lg=19 GHz·μm<br />

is higher than 14.4 GHz·μm <strong>of</strong> <strong>the</strong> recessed-gate device.<br />

Our simulation <strong>of</strong> an InGaN cap structure with lg=0.8 μm<br />

shows that fT=30 GHz can be achieved.<br />

VI. CONCLUSION<br />

After a calibration against experimental data our device<br />

simulator provides reliable results for <strong>the</strong> DC and AC<br />

performance <strong>of</strong> <strong>the</strong> investigated novel HEMTs. This allows<br />

us to perform fur<strong>the</strong>r device optimization, e.g. for<br />

higher current capability, higher threshold voltage, and<br />

better AC performance.<br />

ACKNOWLEDGMENTS<br />

The authors acknowledge support from <strong>the</strong> Austrian Science<br />

Fund (FWF), Project START Y247-N13.<br />

REFERENCES<br />

[1] M. Khan, Q. Chen, C. Sun, J. Yang, M. Blasingame,<br />

M. Shur, and H. Park. Enhancement and depletion<br />

mode GaN/AlGaN heterostructure field effect transistors.<br />

Appl.Phys.Lett. 68(4):514–516, Jan. 1996.<br />

[2] T. Mizutani, M. Ito, S. Kishimoto, and F. Nakamura.<br />

AlGaN/GaN HEMTs with thin InGaN cap layer for<br />

normally <strong>of</strong>f operation. IEEE Electron Device Lett.<br />

28(7):549–551, July 2007.<br />

222<br />

I D [A/mm]<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

12 nm gate recess depth<br />

14 nm gate recess depth<br />

16 nm gate recess depth<br />

0.0<br />

−0.5 0.0 0.5<br />

V [V] GS<br />

1.0 1.5 2.0<br />

Figure 2: Simulated transfer characteristics at Vds=5 V<br />

for HEMTs with different gate recess depths.<br />

DC g m [S/mm]<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

exp. data<br />

simulation<br />

gate recess<br />

InGaN cap<br />

0.0<br />

−1.0 −0.5 0.0 0.5<br />

V [V] GS<br />

1.0 1.5 2.0<br />

Figure 3: Comparison <strong>of</strong> <strong>the</strong> measured (symbols) and<br />

simulated (lines) DC transconductance gm at Vds=5 V.<br />

[3] T. Palacios, C. Suh, A. Chakraborty, S. Keller,<br />

S. DenBaars, and U. Mishra. High-performance Emode<br />

AlGaN/GaN HEMTs. IEEE Electron Device<br />

Lett. 27(6):428–430, June 2006.<br />

[4] V. Palankovski and R. Quay. Analysis and Simulation<br />

<strong>of</strong> Heterostructure Devices. Wien – New York,<br />

Springer, 2004.<br />

[5] S. Vitanov, V. Palankovski, S. Murad, T. Rödle,<br />

R. Quay, and S. Selberherr. Predictive Simulation<br />

<strong>of</strong> AlGaN/GaN HEMTs. Tech.Dig. IEEE Compound<br />

Semiconductor IC Symp. pp. 131–134, Nov. 2007.<br />

[6] S. Vitanov and V. Palankovski. Normally-<strong>of</strong>f Al-<br />

GaN/GaN HEMTs with InGaN cap layer: A simulation<br />

study. Solid-State Electron. <strong>2008</strong>, (in press).


Simulation <strong>of</strong> Q-switched solid-state laser ignition<br />

F. Trawniczek, J. Tauer, H. K<strong>of</strong>ler, M. Deneva and E. Wintner<br />

Institut für Photonik<br />

Technische Universität Wien<br />

Vienna, Austria<br />

Email: franz.trawniczek@gmx.at<br />

ernst.wintner+e387@tuwien.ac.at<br />

Abstract — The main topic <strong>of</strong> this work covers <strong>the</strong><br />

analysis and <strong>the</strong> simulation <strong>of</strong> a Q-switched laser<br />

system. The calculations including all relevant<br />

parameters are based on <strong>the</strong> rate equations for a<br />

4-level laser system.<br />

The approximation <strong>of</strong> <strong>the</strong> pump function, <strong>the</strong><br />

analysis <strong>of</strong> <strong>the</strong> Q-modulator and <strong>the</strong> detailed<br />

overview <strong>of</strong> one single roundtrip in <strong>the</strong> laser system<br />

are indispensable to obtain all important values <strong>of</strong><br />

<strong>the</strong> laser system.<br />

Using an established code, <strong>the</strong> optimal performance<br />

<strong>of</strong> <strong>the</strong> high peak power Q-switched ignition laser is<br />

obtained by <strong>the</strong> application <strong>of</strong> numerical<br />

procedures (Runge-Kutta) and <strong>the</strong> adaption <strong>of</strong><br />

several important parameters (reflectivities <strong>of</strong> <strong>the</strong><br />

two used mirrors, best time for opening <strong>of</strong> <strong>the</strong> Qmodulator,<br />

...).<br />

I. INTRODUCTION<br />

The exact analysis <strong>of</strong> <strong>the</strong> system <strong>of</strong> rate equations<br />

and <strong>the</strong> four level laser system in combination with<br />

<strong>the</strong> attributes <strong>of</strong> <strong>the</strong> used active medium Nd:YAG<br />

(neodymium-doped yttrium aluminium garnet) are<br />

<strong>the</strong> background for understanding <strong>the</strong> laser system.<br />

The approximation <strong>of</strong> <strong>the</strong> actual pump function as a<br />

trapezoid function is essential for obtaining significant<br />

results. It is important to analyze all <strong>the</strong> parameters<br />

<strong>of</strong> a Q-switched laser system for <strong>the</strong> simulation.<br />

II. CALCULATION<br />

In this diploma <strong>the</strong>sis a certain system <strong>of</strong> rate equations<br />

is used to describe to processes in <strong>the</strong> ignition<br />

laser system is:<br />

dN<br />

dt<br />

2 � Rp(<br />

t)<br />

� BqN<br />

2<br />

N2<br />

�<br />

�<br />

(1)<br />

223<br />

dq<br />

dt<br />

q<br />

BqN Va �<br />

�<br />

� 2<br />

c<br />

(2)<br />

Equations (1) and (2) describe <strong>the</strong> effects in <strong>the</strong> laser<br />

system due to stimulated and spontaneous emission,<br />

<strong>the</strong> influence <strong>of</strong> <strong>the</strong> pump function and <strong>the</strong> active<br />

volume. By <strong>the</strong> help <strong>of</strong> <strong>the</strong>se equations it is possible<br />

to calculate one single roundtrip in <strong>the</strong> laser system<br />

to obtain values for <strong>the</strong> gain and <strong>the</strong> losses in <strong>the</strong><br />

<strong>the</strong>rein.<br />

Fur<strong>the</strong>rmore, <strong>the</strong>y are necessary to define <strong>the</strong> threshold<br />

conditions. There are some more calculations<br />

needed like <strong>the</strong> function that describes <strong>the</strong> opening <strong>of</strong><br />

<strong>the</strong> Q-modulator being responsible for <strong>the</strong> differences<br />

between active and passive Q-switched systems.<br />

The additional losses in <strong>the</strong> system are considered by<br />

<strong>the</strong> introduction <strong>of</strong> logarithmic losses in <strong>the</strong> calculation<br />

<strong>of</strong> one single roundtrip.<br />

III. SIMULATION<br />

Applying a code, written in Borland PASCAL, <strong>the</strong><br />

main behavior <strong>of</strong> <strong>the</strong> laser ignition system is simulated<br />

to find <strong>the</strong> conditions for optimal performance <strong>of</strong><br />

<strong>the</strong> high peak power laser.<br />

Numerical procedures (Runge- Kutta) are needed to<br />

solve <strong>the</strong> coupled nonlinear differential equations. It<br />

is necessary to diversify several important parameters<br />

in <strong>the</strong> program (reflectivities <strong>of</strong> <strong>the</strong> two mirrors, best<br />

time for opening <strong>of</strong> <strong>the</strong> Q-modulator, pumping energy,<br />

opening time <strong>of</strong> <strong>the</strong> Q-modulator, ...) to obtain<br />

only one single pulse with high peak power (MW)<br />

and a very short duration (~ 1ns).<br />

A Q-modulator is introduced to meet demands <strong>of</strong> <strong>the</strong><br />

ignition laser. It is an active Q-switched system,<br />

where it is possible to choose <strong>the</strong> time for opening <strong>of</strong><br />

<strong>the</strong> Q-modulator to obtain only one single high peak<br />

pulse.


starting time <strong>of</strong> pulses,<br />

�s<br />

90<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

0<br />

60%<br />

70<br />

80%<br />

90%<br />

0 10 20 30 40 50 60 70<br />

60%<br />

70%<br />

80%<br />

90%<br />

E pump, mJ<br />

Figure 1: Starting times <strong>of</strong> pulses without Q-switch<br />

as a function <strong>of</strong> pump energy<br />

Figure 1 shows <strong>the</strong> starting times <strong>of</strong> <strong>the</strong> pulses in a<br />

system without Q-modulator (free lasing) as a function<br />

<strong>of</strong> <strong>the</strong> pump energy that is applied. In this<br />

graphic <strong>the</strong> dependence is shown for four different<br />

reflectivities <strong>of</strong> <strong>the</strong> outcoupling mirror.<br />

In <strong>the</strong> case <strong>of</strong> active Q-switching, one important<br />

question is to find <strong>the</strong> optimal time for switching to<br />

obtain maximum output power.<br />

P out, W<br />

800000<br />

600000<br />

400000<br />

200000<br />

P out, W<br />

0<br />

289 291 293 295 297 299 301 303<br />

t_Q, �s<br />

Figure 2: Peak power as a function <strong>of</strong> time for<br />

switching <strong>the</strong> Q-Modulator<br />

Figure 2 shows <strong>the</strong> best time for switching <strong>the</strong> Q-modulator<br />

to obtain maximum power. It is necessary to<br />

lock one parameter in position and slightly change<br />

some o<strong>the</strong>rs to reach finally <strong>the</strong> best conditions <strong>of</strong> <strong>the</strong><br />

whole system.<br />

REFERENCES<br />

[1] Orazio Svelto. Principles <strong>of</strong> Lasers. Plenum<br />

Press New York, USA, Fourth Edition, 1998<br />

[2] Wulfhard Lange. Einführung in die Laserphysik.<br />

Wissenschaftliche Buchgesellschaft,<br />

Deutschland, zweite Auflage, April 2002<br />

224


Priorities in Improving <strong>of</strong> Railway Infrastructure<br />

Jiří Pospíšil(Karel Baudyš)<br />

Faculty <strong>of</strong> Transportation Sciences<br />

Czech Technical University in Prague<br />

Prague, Czech Republic<br />

Email: {xjpospisil,baudys}@fd.cvut.cz<br />

Abstract — Article brings new point <strong>of</strong> view on<br />

problems <strong>of</strong> public transport systems improving.<br />

New solution is based on some ideal supply in ideal<br />

infrastructure net. Introduced method assesses<br />

priorities to arcs, junctions or rolling stock to<br />

enabled efficient improving <strong>of</strong> whole system <strong>of</strong><br />

public transport stepwise. One time made<br />

improvements are useful later, <strong>the</strong>y are according<br />

to final ideal system. It enables efficient investment,<br />

fastest supply improvements on beginning <strong>of</strong><br />

process and goal-directed approach to ideal system.<br />

I. INTRODUCTION<br />

Railway has potential to support a device to main<br />

traffic flows. But can be ecological and efficient only<br />

if will be competitive. Public transport must <strong>of</strong>fer<br />

frequent, fast connection between all spots in<br />

network. Such supply can be created thanks to<br />

periodic timetables. Temporal availability is<br />

guaranteed by appropriate period between trains.<br />

Spatial availability is guaranteed by suitable<br />

coordination <strong>of</strong> train time positions (time slots) on<br />

tracks and in junctions. Based on Integrated Periodic<br />

Timetable (IPT), lines in particular area are<br />

interconnected so that <strong>the</strong>re are realised optimal<br />

changing connections, i.e. without waiting times.<br />

II. INTEGRATED PERIODIC TIMETABLE<br />

IPT is based on principle „in each hour – <strong>the</strong> same<br />

minute – in each junction – in each direction“.<br />

Figure 1 – Principle <strong>of</strong> Integrated Periodic Timetable<br />

Optimal connections between particular lines in IPT<br />

junctions need system travelling times between<br />

225<br />

junctions. System travelling time includes regular<br />

travelling time + stop-times in stations between<br />

junctions + proportionate part <strong>of</strong> changing times in<br />

junctions + time allowance. To build a net-supply<br />

with synergic effect basic condition must be fulfilled:<br />

- System travelling time = half <strong>of</strong> multiple period<br />

- In addition, IPT edge condition must be fulfilled:<br />

Arc equation tH = n . ½ tT forn=1,2,3,…(1)<br />

Circle equation ∑ tH = n .tT forn=1,2,3,…(2)<br />

tH = Arc length (system travelling time) and<br />

tT =Period<br />

To enable equal connection in both directions all<br />

lines must have unified symmetry axis, i.e. time, in<br />

which trains <strong>of</strong> one line are crossing each o<strong>the</strong>r. In<br />

long-distance transport in Europe this axis is usually<br />

set to minute 00, eventually according to period 30.<br />

For suburban trains is used symmetry axis 00 and 30<br />

eventually also 15 and 45.<br />

Concept <strong>of</strong> periodic transport consists in relevant<br />

location <strong>of</strong> IPT-junctions in net, in achieving system<br />

travelling times on arcs and unified symmetry axis on<br />

all lines. Transport chains are attractive supply in net.<br />

Supply <strong>of</strong> long-distance trains in Czech Republic<br />

bases on IPT-principle, passenger count increases. In<br />

infrastructure area we first build and than we make<br />

some proposal for transport conceptions. It’s not<br />

optimal. Due to life-cycle <strong>of</strong> infrastructure and its<br />

high costs could be better first plan “ideal” supply <strong>of</strong><br />

public transport system (problem <strong>of</strong> transport<br />

planning) and than determine requirements for<br />

infrastructure and rolling stock.<br />

In praxis (with actual infrastructure and rolling<br />

stock) is not easy to realise ideal supply, very <strong>of</strong>ten is<br />

it impossible. High investment is necessary. But<br />

<strong>of</strong>ten is invested uncoordinatedly without net synergy<br />

effects. There are some political wishes witch evoke<br />

direct technical requirement for process <strong>of</strong> improving<br />

<strong>of</strong> public transport system:<br />

- Minimize total passenger time in system<br />

- Have a circles corresponding to circle equation as<br />

soon as possible and ideal arcs IPT (net bindings)<br />

- During improving maximize relationship between<br />

contributions and cost (first contributions for lower<br />

costs, high costs and <strong>the</strong>ir contributions later)<br />

- High stability <strong>of</strong> timetable (supply) during process<br />

– It’s impossible to change timetable all <strong>the</strong> time


III. ASSESSMENT OF PRIORITIES IN RAILWAY<br />

INFRASTRUCTURE IMPROVING<br />

Method assesses priorities in railway infrastructure<br />

improving has two inputs: contemporary and ideal<br />

net. By simple table adjustment with all circles and<br />

arcs <strong>of</strong> net(s) is possible to go from contemporary<br />

(actual) to ideal net. Each step represents next<br />

priority. It’s necessary to choose right arc from right<br />

circle and this one improve – shorten for needed<br />

length to make possible to travel in circles without<br />

waiting times. It’s not necessary to shorten arc to<br />

ideal (shortest) length. So is possible leave <strong>the</strong> most<br />

expensive parts <strong>of</strong> arcs (e.g. new tunnels) and make<br />

<strong>the</strong>m later.<br />

Method takes into account system travelling times<br />

(arc length), investment needs (infrastructure, rolling<br />

stock) and a passenger flows. Most important value<br />

is time. Passengers and costs are only significant.<br />

Method give priority to low-cost improvements,<br />

witch enable faster transition to ideal net with ideal<br />

change possibilities in railway junction, but not with<br />

ideal length <strong>of</strong> arcs. It isn’t problem. Improvements<br />

shorten arc length anyway and also passengers<br />

travelling on arcs go shorter time. Final lifetime and<br />

costs <strong>of</strong> improvements are by any priorities <strong>the</strong> same.<br />

IV. FIRST RESULTS<br />

I have tested simplified method (unitary valuation <strong>of</strong><br />

costs and traffic flows) on chosen net <strong>of</strong> longdistance<br />

railway lines in west Bohemia and in<br />

Moravia. In both cases give method successful<br />

outputs how show tables. For confrontation I used<br />

chooses with biggest timesavings on arcs,<br />

lowest/highest investment and randomised chooses.<br />

“K1-K6” represents all available circles in net. “x”<br />

indicates circle with null waiting times in all junction<br />

on this circle.<br />

Steps Travel. time Waiting t. Timesavings Price k1 k2 k3 k4 k5 k6<br />

1 2680 80 120 20 x x x x<br />

2 2640 120 120 30 x x<br />

3 2610 30 240 40 x x x<br />

4 2520 30 330 70 x x x<br />

5 2460 0 420 90 x x x x x x<br />

6 2370 0 510 120 x x x x x x<br />

7 2280 30 570 150 x x x x x<br />

8 2190 0 690 180 x x x x x x<br />

Table 1: Method for priorities assessment<br />

Steps Travel. time Waiting t. Timesavings Price k1 k2 k3 k4 k5 k6<br />

1 2680 80 120 20 x x x x<br />

2 2650 110 120 30 x x<br />

3 2530 50 300 70 x x<br />

4 2440 50 390 100 x x<br />

5 2380 80 420 120 x x x<br />

6 2320 110 450 140 x<br />

7 2280 30 570 150 x x x x x<br />

8 2190 0 690 180 x x x x x x<br />

Table 2: Political priority for TEN-corridors<br />

226<br />

In table 1 have good connections on circles longest<br />

continuity. After 5th step we have net with system<br />

travelling times, without waiting times in junctions.<br />

Some arcs are still longer <strong>the</strong>n ideal length. These<br />

arcs need higher investment and bigger infrastructure<br />

adjustments. There can’t be compared appropriate<br />

steps, but timesaving for accordant costs. Travelling<br />

time + waiting time correspond to total time <strong>of</strong><br />

passengers in public transport system.<br />

There are some opened questions in method still:<br />

- Efficient choose <strong>of</strong> junctions and arcs IPT<br />

- Determination <strong>of</strong> all circles in net<br />

- Traffic flows (net lay-out)<br />

- Efficient valuation <strong>of</strong> improvements<br />

- Problem <strong>of</strong> arc elongation, withdrawal or insertion<br />

Introduced method enables manual intervention<br />

(political choose) providing misalignment from ideal<br />

path <strong>of</strong> improvements. But <strong>the</strong>re is enabled also<br />

limitation on total investment level or investment<br />

level for saved time value.<br />

V. CONCLUSION<br />

In method are included not only infrastructure but<br />

rolling stock improvements too. It’s about<br />

improvement <strong>of</strong> whole public transport system.<br />

Regional transport systems (including Buses) could<br />

be included or solved separately.<br />

Impartial assessment <strong>of</strong> priorities (in <strong>the</strong> sense first<br />

contributions for lower costs, high costs and <strong>the</strong>ir<br />

contributions later) could bring more efficiency to<br />

whole public transport system, better competitive<br />

advantage <strong>of</strong> service (net connections) and enable its<br />

sustainable development. Supply <strong>of</strong> service will be<br />

better thanks investment to infrastructure and<br />

according to this to rolling stock too. Method<br />

assesses time horizons <strong>of</strong> realization respective <strong>of</strong><br />

introduction <strong>of</strong> new supply according to investment<br />

budget. Introduced method could be suitable for<br />

employers <strong>of</strong> public transport service or for managers<br />

<strong>of</strong> transport infrastructure.<br />

REFERENCES<br />

[1] M. Bär. Merkblatt zum integralen Taktfahrplan.<br />

TU Dresden, 2001.<br />

[2] K. Baudyš, V. Janoš. Optimalizační metody<br />

používané při sestavě integrálního taktového<br />

grafikonu. Konference “Věda o dopravě”, FD<br />

ČVUT Praha, 2001.<br />

[3] O. Pastor, A. Tuzar. Teorie dopravních<br />

systémů. Praha ASPI, 2007.<br />

[4] Z. Michl, J. Pospíšil. Is railway ecological?<br />

Summer school „Transportation and Climate<br />

Change“, Ostritz, Germany, 2007.


Raman and Surface Enhanced Raman Spectra <strong>of</strong> Apigenin and Luteolin<br />

Charlie Corredor and John R. Lombardi (Faculty Mentor)<br />

Department <strong>of</strong> Chemistry and Center for Study <strong>of</strong> Structures and Interfaces (CASI),<br />

The City College <strong>of</strong> New York<br />

New York, N.Y. 10031<br />

Email: Ccorred00@ccny.cuny.edu<br />

Abstract — The FT-Raman and surfaceenhanced<br />

Raman (SER) spectra <strong>of</strong> three flavonoids,<br />

namely chrysin, apigenin and luteolin, have been<br />

obtained. The SERS spectra were obtained on citrate<br />

reduced Ag colloids. Assignments <strong>of</strong> <strong>the</strong> experimentally<br />

obtained normal vibrational modes were aided<br />

by density fuctional <strong>the</strong>ory (DFT) calculations using<br />

<strong>the</strong> B3LYP functional and <strong>the</strong> 6-31+G* basis set.<br />

Excellent fits were obtained for <strong>the</strong> observed spectra<br />

with little or no scaling. The most intense lines in<br />

<strong>the</strong> three flavonoids are those in <strong>the</strong> C=O stretching<br />

region and around 1250cm -1 . The first ones are <strong>of</strong>ten<br />

weakened by proximity <strong>of</strong> <strong>the</strong> metal surface, while<br />

<strong>the</strong> latter are not affected by <strong>the</strong> Ag. On <strong>the</strong> o<strong>the</strong>r<br />

hand, <strong>the</strong> lines at lower wavenumbers, assigned to inplane<br />

ring deformation, are strongly enhanced by <strong>the</strong><br />

surface, indicating a perpendicular orientation <strong>of</strong> <strong>the</strong><br />

flavonoids on <strong>the</strong> Ag surface. The spectra <strong>of</strong> <strong>the</strong><br />

flavonoids are compared, and a case study <strong>of</strong> application<br />

to detect weld, a mixture <strong>of</strong> apigenin and luteolin,<br />

in a textile is presented.<br />

I. INTRODUCTION<br />

Phenolic derivatives <strong>of</strong> flavone are responsible<br />

for <strong>the</strong> brightly colored pigments <strong>of</strong> many fruits<br />

and vegetables. Polyphenols are found in high concentrations<br />

in wine, tea, grapes and in a wide variety<br />

<strong>of</strong> o<strong>the</strong>r plants and have been associated with prevention<br />

<strong>of</strong> heart disease and cancer i . Due to <strong>the</strong>ir polyphenolic<br />

nature, flavonoids exhibit strong antioxidant<br />

properties and have been widely used as ingredients<br />

in pharmaceutical products. They can be subdivided<br />

into: flavonols (e.g. quercetin), flavones (e.g.<br />

apigenin, luteotin), flavanols (e.g. catechin) and<br />

is<strong>of</strong>lavones (e.g. genistein). Flavones are also <strong>the</strong><br />

main components <strong>of</strong> various natural dyes (e.g. weld,<br />

fustic, quercitron), mainly yellow but ranging from<br />

brown to green and olive-green, used in textile industries<br />

since ancient times ii,iii . Present in plants as<br />

glycosides,<br />

<strong>the</strong>y can be used as mordant dyes by hydrolysis <strong>of</strong><br />

glycoside bonds and chelating with metals such as<br />

aluminum or copper iv .<br />

227<br />

Figure 1 shows <strong>the</strong> structure <strong>of</strong> chrysin (5,<br />

7-dihydroxyflavone), apigenin (5, 7, 4’trihydroxyflavone)<br />

and luteolin (5, 7, 3’, 4’tetrahydroxyflavone).<br />

They differ only in that apegenin<br />

and luteolin have a 4’ or 3’ and 4’ OH group on<br />

<strong>the</strong> bezene ring. Among those, chrysin has been<br />

studied least by spectroscopic techniques. Even<br />

though it contains two hydroxyl groups, this molecule<br />

is relatively insoluble in water, making <strong>the</strong> preparation<br />

<strong>of</strong> <strong>the</strong> samples difficult. Chrysin and its complexes<br />

with Al 3+ , Ga 3+ and In 3+ have been studied by<br />

IR, UV-vis and NMR v . The IR spectra shows (OH)<br />

bands at 3600-2400 cm -1 , (C=O) at 1655 cm -1 , aromatic<br />

(C=C) at 1505 cm -1 , o<strong>the</strong>r bands at 1460 cm -1 ,<br />

1168 cm -1 , (CH) at 850 cm -1 and 815 cm -1 . Only <strong>the</strong><br />

UV resonance Raman (UVRR) spectrum <strong>of</strong> chrysin<br />

has been reported in <strong>the</strong> study on extractable compounds<br />

in Scots pine wood, where <strong>the</strong> flavonoid was<br />

used as a reference vi . The reported Raman bands are<br />

<strong>the</strong> symmetric aromatic ring stretching at 1605 cm -1 ,<br />

<strong>the</strong> C=C ring stretching at 1649 cm -1 , and <strong>the</strong> asymmetric<br />

ring vibration at ~1500 cm -1 .<br />

On <strong>the</strong> o<strong>the</strong>r hand, apigenin and luteolin are<br />

<strong>the</strong> main coloring matters <strong>of</strong> <strong>the</strong> natural dyestuff<br />

weld. It originates from <strong>the</strong> leaves and stem <strong>of</strong> Reseda<br />

luteola L. found in Central Europe, India and<br />

China. Before <strong>the</strong> discovery <strong>of</strong> America, weld was<br />

<strong>the</strong> dyestuff most used in Western Europe. In 1775,<br />

quercitron became available in large quantities and<br />

superseded weld and fustic as <strong>the</strong> principal yellow<br />

dye vii . Apigenin and luteolin are very abundant in<br />

textile dyes from various samples and HPLC, GCMS,<br />

IR, Raman and SERS on silver colloid data have<br />

been reported viii,ix,x,xi . Their structure differs by only<br />

one hydroxyl group, so <strong>the</strong>y exhibit similar structural<br />

properties. Luteolin shows (O-H) vibrations around<br />

3220 cm -1 and (C=O) at 1655 cm -1 in IR. FT-Raman<br />

spectra <strong>of</strong> luteolin and apigenin in solid state show<br />

similar trend <strong>of</strong> <strong>the</strong> region corresponding to (C=O)<br />

and (C=C) vibrations at 1660-1550 cm -1 , but differences<br />

in <strong>the</strong> 1400-1000 cm -1 region. This is explained<br />

by appearance <strong>of</strong> different (O-H) and (C-O)<br />

motions due to <strong>the</strong> different number <strong>of</strong> hydroxyl<br />

groups. SERS <strong>of</strong> luteolin on silver colloid showed a<br />

strong downward shift <strong>of</strong> <strong>the</strong> aromatic (C=C) vibra-


tion at 1608 cm -1 in solid to 1577 cm -1 in colloid.<br />

This was explained by <strong>the</strong> effect <strong>of</strong> polymerization <strong>of</strong><br />

<strong>the</strong> catechol-like moiety on <strong>the</strong> Ag nanoparticles.<br />

Apigenin was reported to have fewer differences<br />

between <strong>the</strong> Raman spectrum <strong>of</strong> <strong>the</strong> solid and <strong>the</strong><br />

SERS on <strong>the</strong> colloid 11 .<br />

The limited number <strong>of</strong> studies on chrysin by<br />

Raman and SERS spectroscopy may be due to its<br />

poor solubility in water as well as <strong>the</strong> strong fluorescence<br />

upon excitation, which obscure regularly weak<br />

Raman spectra. Our recent study on o<strong>the</strong>r flavonoid<br />

molecules including pentahydroxyflavone quercitin<br />

has shown effectiveness <strong>of</strong> SERS in quenching <strong>the</strong><br />

fluorescence and inducing spectra intensity by <strong>the</strong><br />

few orders <strong>of</strong> magnitude xii,xiii . This method also allows<br />

using very little sampling which is crucially<br />

important in art painting and textile conservation xiv as<br />

well as in bioresearch and forensic science xv .<br />

In this article we report FTR and SERS on Ag<br />

colloid <strong>of</strong> all three molecules as well as Ag electrode<br />

SERS spectra <strong>of</strong> luteolin. These studies are aided in<br />

<strong>the</strong>ir interpretation by DFT calculations. We provide<br />

comparative analysis <strong>of</strong> <strong>the</strong> spectra <strong>of</strong> molecules with<br />

relation to each o<strong>the</strong>r. Our objective is to apply techniques<br />

similar to our past studies to obtain spectra <strong>of</strong><br />

flavonoids and to fur<strong>the</strong>r study <strong>the</strong>ir structures and<br />

possibly seek better methods <strong>of</strong> analysis suitable for art<br />

preservation and o<strong>the</strong>r fields. In order <strong>the</strong> better<br />

interpret <strong>the</strong> results, we have also conducted studies <strong>of</strong><br />

<strong>the</strong> related compounds 7-hydroxyflavone, and 3’,4’-<br />

Dihydroxyflavone. These are presented in <strong>the</strong> supplementary<br />

materials, and referred to when needed.<br />

II. FIGURES<br />

Figure 1.<br />

(a) Chrysin<br />

(b) Apigenin<br />

(c) Luteotin<br />

REFERENCES<br />

i<br />

Oliveira Brett, A.M., Ghica, M-E., Electroanalysis 15,<br />

22 (2003).<br />

228<br />

ii Joosten, I., Boomel, M., R., Keijzer, R., H., Reschreiter,<br />

H.; Micro Analysis on Hallstatt Textules: Colour and<br />

Condition; J. Microchim. Acta. 155, 169-174 (2006)<br />

iii Zhang, X., Boytner, R., Cabrera, J., L., Laurson, R.;<br />

Identification <strong>of</strong> Yellow Dye Types in Pre-Columbian<br />

Andean Textiles; J. Anal. Chem. 2007, 79, 1575-1582<br />

iv Ferreira ESB, Hulme AN, McNab H, Quye A; The<br />

Natural Constituents <strong>of</strong> Historical Textile Dyes; Advance<br />

Article, www.rsc.org, 2004<br />

v Pusz, J., Nitka, B., Zielinska, A., Wawer, I.; Syn<strong>the</strong>sis<br />

and physicochemical properties <strong>of</strong> <strong>the</strong> Al(III), Ga(III) and<br />

In(III) complexes with chrysin; Microchem. J. 65 (2000)<br />

245-253<br />

vi Nuopponen, M., Willfor, S., Jaaskelainen, A. S.,<br />

Vuorinen, T.; A UV resonance Raman (UVRR) spectroscopic<br />

study on <strong>the</strong> extractable compounds in Scots pine<br />

(Pinus sylvestris) wood Part II. Hydrophilic compounds;<br />

Spectrochimica Acta Part A 60 (2004) 2963-2968<br />

vii Judith H. H<strong>of</strong>enk de Graaff , The Colorful Past, 165-<br />

231, Abegg-Stiftung and Archetype Publications, Riggisberg<br />

and London, 2004.<br />

viii Chen K, Leona M, Vo-Dinh KC, Yan F, Wabuyele<br />

MB, Vo-Dinh T. Application <strong>of</strong> surface-enhanced Raman<br />

scattering (SERS) for <strong>the</strong> identification <strong>of</strong> athranquinone<br />

dyes used in works <strong>of</strong> art. Journal <strong>of</strong> Raman Spectroscopy,<br />

2006; 37: 520-527<br />

ix Colombini, M. P., Andreotti, A., Baraldi, C., Degano,<br />

I., Lucejko, J. J; Colour fading in textiles: A model study<br />

on <strong>the</strong> decomposition <strong>of</strong> natural dyes; Microchem. J. 85<br />

(2007) 174-182<br />

x Inglett, G. E.; Infrared spectra <strong>of</strong> naturally occurring<br />

flavonoids; J. Org. Chem (1958), 23, 93-94<br />

xi Jurasekova, Z., Garcia-Ramos, J. V., Dominigo, C.,<br />

Sanchez-Cortes, S.; Surface-Enhanced Raman scattering<br />

<strong>of</strong> flavonoids; J. Raman. Spectrosc. 2006; 37: 1239-1241<br />

xii Wang, M., Teslova, T., Xu, F., Spataru, T., Lombardi,<br />

J. R., Birke, R. L; Raman and Surface Enhanced Raman<br />

Scattering <strong>of</strong> 3-Hydroxyflavone; . Phys. Chem. C, 111<br />

(7), 3038 -3043, 2007<br />

xiii Teslova , T.; Corredor , C.; Livingstone , R.; Spataru<br />

T.; Birke R.L.; Lombardi ,J.R.; M. V. Cañamares,<br />

M.V.; Leona, M.; Raman and surface-enhanced Raman<br />

spectra <strong>of</strong> flavone and several hydroxy derivatives; J.<br />

Ram. Spectrosc. 38, 802 (2007).<br />

xiv Z, Wiggins K, Russell C, Chabli S, Rosengarten A.<br />

Evaluation <strong>of</strong> Raman Spectroscopy for <strong>the</strong> analysis <strong>of</strong><br />

colored fibers: a collaborative study. J. Forensic Sci.<br />

2005, 50(5) 1028-1038<br />

xv Review Papers, 14 th International Forensic Science<br />

Symposium, Interpol- Lyon, 19-22 October 2004, (N. N.<br />

Daeid Editor)


A Sustainable Approach in Hybrid Materials Development:<br />

Metal Nanoparticles Syn<strong>the</strong>sis in Biobased Polymeric Systems<br />

Colleen Achong, Praveen Kumar Vemula, Kareem Douglas, and George John *<br />

Department <strong>of</strong> Chemistry, City College <strong>of</strong> <strong>the</strong> City University <strong>of</strong> New York, 160 Convent Avenue,<br />

New York, NY 10031 USA. Fax: +1-212-650-6107; Tel:+1-212-650-8353<br />

* Corresponding author Email: john@sci.ccny.cuny.edu<br />

Abstract —In situ syn<strong>the</strong>sis <strong>of</strong> metal nanoparticles<br />

using various forms <strong>of</strong> self-assembled organic s<strong>of</strong>t<br />

materials such as molecular gels and liquid crystals<br />

proved to be an efficient approach in preparing<br />

hybrid materials. The present study shows <strong>the</strong> successful<br />

extension <strong>of</strong> our previous strategies to generate<br />

metal nanoparticles (MNPs) in polymeric<br />

systems, ano<strong>the</strong>r form <strong>of</strong> s<strong>of</strong>t matter. We have syn<strong>the</strong>sized<br />

a biobased polymer, poly(cardanyl acrylate)<br />

(PCA) from cardanol (byproduct <strong>of</strong> cashew<br />

industry) equipped with usaturated side chains<br />

suitable for oxidative drying. Autoxidation (crosslinking)<br />

<strong>of</strong> PCA produced various free-radicals<br />

which were used as a tool to reduce metal salts for<br />

preparing and stabilizing <strong>the</strong> gold and silver<br />

nanoparticles in situ. This sustainable approach<br />

enabled us to avoid external hazardous reducing<br />

and stabilizing agents.<br />

I. INTRODUCTION<br />

Design and syn<strong>the</strong>sis <strong>of</strong> organic-inorganic hybrid<br />

materials from biobased monomers/polymers have a<br />

tremendous impact on industry and <strong>the</strong> environment.<br />

Strategic utilization <strong>of</strong> renewable resources as starting<br />

material for generating value-added products and<br />

building blocks in chemistry will have broad impact<br />

in industrial economy as well as in sustainable development.<br />

We 1-4 and o<strong>the</strong>rs 5-6 have shown <strong>the</strong> efficient<br />

use <strong>of</strong> renewable resources for developing novel<br />

monomers, polymers and hierarchically assembled<br />

s<strong>of</strong>t nanomaterials. Polymers are one among <strong>the</strong> most<br />

important products <strong>of</strong> <strong>the</strong> chemical industry used for<br />

versatile applications in everyday life. Herein we<br />

report <strong>the</strong> syn<strong>the</strong>sis and stabilization <strong>of</strong> AuNPs and<br />

AgNPs in biobased poly(cardanyl acrylate), PCA.<br />

Focal point <strong>of</strong> present study is, naturally occurring<br />

cross-linking <strong>of</strong> unsaturated alkyl chains which is<br />

known as autoxidation or drying process produces<br />

free-radicals in situ; such free-radicals were used as a<br />

tool to reduce metal salts to generate MNPs in situ,<br />

which allowed us to avoid using external hazardous<br />

reducing and stabilizing agents.<br />

229<br />

II. RESULTS AND DISCUSSION<br />

Cardanol [obtained from <strong>the</strong>rmal treatment <strong>of</strong><br />

cashew nut shell liquid] is well known to exist as a<br />

mixture <strong>of</strong> four components differing in <strong>the</strong> degree <strong>of</strong><br />

unsaturation in <strong>the</strong> side chain: 5% <strong>of</strong> 3-(pentadecyl)phenol,<br />

49% <strong>of</strong> 3-(8Z-pentadecenyl)-phenol, 17% <strong>of</strong><br />

3-(8Z,11Z-pentadecadienyl)-phenol, and 29% <strong>of</strong> 3-<br />

(8Z,11Z,14-pentadecatrienyl)-phenol (Figure 1).<br />

Figure 1. Chemical structures and syn<strong>the</strong>tic<br />

scheme <strong>of</strong> a) PCA and b) PPDA. i) acryloyl chloride,<br />

NaOH, toluene, 8 hr; ii) azobisisobutyronitrile,<br />

toluene, 60ºC. c) General mechanism for<br />

free-radical mediated cross-linking <strong>of</strong> cardanyl<br />

polymer side chains.<br />

Cardanyl acrylate (CA) was syn<strong>the</strong>sized using a<br />

reported procedure. Subsequently, solution polymerization<br />

was achieved using azobisisobutyronitrile<br />

(AIBN) in toluene to obtain PCA (Figure 1a). The<br />

PCA was dissolved in chlor<strong>of</strong>orm and could easily<br />

be cast into a thin transparent and sticky film by <strong>the</strong><br />

solution-casting technique. On heating to 65ºC for 30<br />

min or upon exposure to <strong>the</strong> air (ambient conditions)<br />

for 10 hours <strong>the</strong> sticky film converted into non-sticky<br />

(scratch-free) transparent insoluble film. The crosslinking<br />

<strong>of</strong> PCA unsaturated hydrocarbon chains<br />

possibly occurred through <strong>the</strong> hydrogen abstraction<br />

(at allylic carbon center) proceeded by a cascade <strong>of</strong>


eactions with bond breaking and forming to generate<br />

an insoluble polymer network (Figure 1c). This was<br />

thoroughly characterized using various techniques<br />

such as infra-red spectroscopy, 1 H-NMR and differential<br />

scanning calorimeter as reported earlier. It is<br />

proposed that cross-linking was occurred through<br />

hydroperoxidation <strong>of</strong> allylic radical centers <strong>of</strong> cardanol<br />

alkyl chain, during this process various freeradicals<br />

were generated in situ (Figure 1c). In case <strong>of</strong><br />

PCA, <strong>the</strong> side chains are brought close to each o<strong>the</strong>r<br />

by <strong>the</strong> acrylate polymer backbone and precisely align<br />

<strong>the</strong> allylic side chains for fur<strong>the</strong>r cross-linking process<br />

as shown schematically in Figure 1a.<br />

Figure 2. a) Images <strong>of</strong> glass slides coated with<br />

PCA containing AuNPs (left), AgNPs (middle)<br />

and un-reduced metal salts (right). b) Absorption<br />

spectra <strong>of</strong> PCA (i) after addition <strong>of</strong> chloroauric<br />

acid, (ii) after AuNPs formed. Similarly, (iii) after<br />

addition <strong>of</strong> silver benzoate and (iv) after AgNPs<br />

formed.<br />

Polymer films which contained in situ prepared<br />

MNPs were characterized using different techniques<br />

such as ultraviolet-visible (UV-vis) spectrophotometer,<br />

transmission (TEM) and scanning electron microscopy<br />

(SEM). The absorption spectra were recorded<br />

for metal salts contained polymer films before<br />

and after drying process (Figure 2b). Absorption<br />

spectrum <strong>of</strong> AuNPs contained polymer was showed<br />

characteristic surface plasmon resonance band at 555<br />

nm, whereas such peak was absent when we immediately<br />

measured after <strong>the</strong> addition <strong>of</strong> metal salts<br />

(before drying process occurs, Fig. 2b: i & ii). This is<br />

clearly suggesting <strong>the</strong> formation <strong>of</strong> AuNPs during <strong>the</strong><br />

cross-linking process. Similarly, absorption spectra<br />

<strong>of</strong> in situ syn<strong>the</strong>sized AgNPs in polymer have shown<br />

peak at 460 nm, and this peak is noticed for <strong>the</strong><br />

AgNPs due to <strong>the</strong> characteristic surface plasmon<br />

230<br />

resonance effect originating from <strong>the</strong> quantum size<br />

effect <strong>of</strong> AgNPs. In this instance also prior to autoxidation<br />

(cross-linking) process such absorption band<br />

in UV was absent (Figure 2b: iii &iv) which is<br />

clearly suggesting that indeed MNPs are generated<br />

during <strong>the</strong> oxidative drying <strong>of</strong> <strong>the</strong> polymer film. The<br />

absorbance maximum does not change over a long<br />

period which indicates that MNPs are prevented from<br />

coagulation due to <strong>the</strong> stabilization <strong>of</strong> NPs in <strong>the</strong><br />

polymer matrix. The SEM and TEM analysis <strong>of</strong> <strong>the</strong><br />

in situ syn<strong>the</strong>sized AuNPs and AgNPs in PCA indicated<br />

that <strong>the</strong> syn<strong>the</strong>sized MNPs were nearly monodispersed<br />

with <strong>the</strong> average size <strong>of</strong> 18 nm for AuNPs,<br />

and 13 nm for AgNPs. Polydispersity <strong>of</strong> <strong>the</strong> particles<br />

were calculated from TEM data.<br />

III. CONCLUSIONS<br />

In conclusion, renewable resources based monomer<br />

(cardanol from cashew nut shell liquid) was used<br />

to prepare a biobased polymer. In situ syn<strong>the</strong>sis <strong>of</strong><br />

MNPs has been achieved with this polymer without<br />

using any external reducing and stabilizing agents.<br />

Free-radicals which were generated during oxidative<br />

drying <strong>of</strong> cardanol unsaturated chains have been used<br />

as a tool for MNPs syn<strong>the</strong>sis, such hypo<strong>the</strong>sis was<br />

confirmed by syn<strong>the</strong>sizing saturated analogue polymer<br />

which failed to generate MNPs. MNPs embedded<br />

polymers films were stable at ambient conditions<br />

for a longer period. The preparation <strong>of</strong> organicinorganic<br />

hybrid materials would have applications<br />

in developing materials with tunable optical, electrical,<br />

and catalytic properties. Potential antibacterial<br />

activity <strong>of</strong> AgNPs embedded biobased polymers are<br />

currently under investigation in our laboratory.<br />

ACKNOWLEDGMENTS<br />

We thank Science Interdepartmental Electron Microscope<br />

and Imaging Center at CCNY.<br />

REFERENCES<br />

[1] P. K. Vemula, and G. John, Chem. Commun.<br />

2218 (2006).<br />

[2] P. K. Vemula, U. Aslam, V. A. Mallia, and G.<br />

John, Chem. Mater. 19, 138 (2007).<br />

[3] V. A. Mallia, P. K. Vemula, G. John, A. Kumar,<br />

and P. M. Ajayan, Angew. Chem. Int. Ed. 46,<br />

3269 (2007).<br />

[4] P. K. Vemula, V. A. Mallia, K. Bizati, and G.<br />

John, Chem. Mater. 19, 5203 (2007).<br />

[5] F. W. Lichtenthaler, Acc. Chem. Res. 35, 728<br />

(2002).<br />

[6] A. Corma, S. Iborra, and A. Velty, Chem. Rev.<br />

107, 2411 (2007).


Micro-lenses Fabrication by Solvent-Casting <strong>of</strong> Chalcogenide<br />

Glass<br />

Eric Sanchez 1 and Craig Arnold 2<br />

1 The City College <strong>of</strong> New York, New York, NY 10031, USA<br />

2 Department <strong>of</strong> Mechanical and Aerospace Engineering, Princeton University, Princeton, NJ<br />

08540, USA<br />

Email: esanche05@ccny.cuny.edu<br />

Abstract — A solution-based method for <strong>the</strong> fabrication<br />

<strong>of</strong> chalcogenide glass optical elements is<br />

described. A micro-syringe system using manual<br />

disposition has been applied to <strong>the</strong> fabrication <strong>of</strong><br />

micro-lenses that range from 7μm to 350 μm. The<br />

micro-lenses are made As2S3 which has demonstrated<br />

important optical properties for mid infrared<br />

wavelengths <strong>of</strong> light. The method itself can be<br />

divided in two steps: disposition <strong>of</strong> chalcogenide<br />

solution, and heat treatment. The produced microlenses<br />

are measured by pr<strong>of</strong>ilometry and photographed<br />

using a light microscope.<br />

I. INTRODUCTION<br />

Quantum Cascade Lasers (QCLs) have emerged as a<br />

leading method to produce mid-infrared (Mid-IR)<br />

laser light for sensing applications for health and <strong>the</strong><br />

environment [1]. Due to <strong>the</strong> small size <strong>of</strong> QCLs, <strong>the</strong>y<br />

have <strong>the</strong> potential to significantly reduce overall<br />

device size and even to be directly integrated for onchip<br />

sensing. However, in order to successfully implement<br />

this technology, corresponding optics must<br />

be scaled down creating a need for micro-lenses and<br />

waveguides at <strong>the</strong>se frequencies. Chalcogenide<br />

glasses have become an important material for Mid-<br />

IR technology due to <strong>the</strong>ir favorable physical and<br />

optical properties [2]. Through novel processing<br />

approaches, <strong>the</strong>se materials promise a path for developing<br />

micro-lenses for Mid-IR technology.<br />

In developing optical elements, <strong>the</strong> method used<br />

plays a critical role in reaching <strong>the</strong> small feature<br />

sizes. In this presentation, we describe a solutionbased<br />

method for <strong>the</strong> fabrication <strong>of</strong> chalcogenide<br />

glass micro-lenses. This particular method consists <strong>of</strong><br />

two components: micro-syringe dispensing and heat<br />

treatment. The micro-syringe is an attractive technique<br />

due to its low cost and relative ease <strong>of</strong> implementation.<br />

II. METHOD<br />

Chalcogenide glasses are amorphous compounds<br />

containing sulfur, selenium, or tellurium, and are<br />

231<br />

known to be low-loss mid-infrared materials [3].<br />

Chalcogenide glasses micro-lenses are fabricated by<br />

disposing As2S3 on a hydrophobic surface in controlled<br />

conditions. In this case, we use InP, which is<br />

one <strong>of</strong> <strong>the</strong> mid-IR technology semi-conductor materials<br />

used as substrate for QCL’s and o<strong>the</strong>r devices.<br />

This solution-based method for fabrication <strong>of</strong><br />

chalcogenide glass micro-lenses is divided into four<br />

parts:<br />

- Solution preparation<br />

- Surface conditioning<br />

- Dispensing<br />

- Heat treatment<br />

The chalcogenide solution is produced by grinding<br />

bulk As2S3 (typically 99.9% pure) into a fine powder<br />

using a ceramic mortar and pestle; 2.50±0.03 g <strong>of</strong> <strong>the</strong><br />

powder is combined with 10±0.1 ml <strong>of</strong> butylamine<br />

and allowed to dissolve for four days. An ultrasonic<br />

homogenizer is used to expedite this dissolving<br />

process. The solution is <strong>the</strong>n centrifuged at 3000 rpm<br />

for 3 minute to remove any suspended particulates or<br />

impurities. Exposure <strong>of</strong> solution or samples to ambient<br />

light and atmospheric moisture is kept to a<br />

minimum throughout <strong>the</strong> preparation procedure [3].<br />

After this, <strong>the</strong> As2S3 solution is ready for use in <strong>the</strong><br />

micro-lenses fabrication.<br />

The method described is only effective in a hydrophobic<br />

surface. Because <strong>of</strong> <strong>the</strong> InP hydrophilic nature,<br />

this must be treated first with 10% hydr<strong>of</strong>luoric<br />

acid (HF) for 10 seconds to remove surface native<br />

oxide thus becoming hydrophobic [4].<br />

In <strong>the</strong> dispensation process, 30μl <strong>of</strong> As2S3 solution<br />

are placed in a pneumatic micro-syringe dispensator<br />

(EFD 1500XL) with a tip <strong>of</strong> aperture <strong>of</strong> 25.4μm. The<br />

tip is placed at 0.25mm above <strong>the</strong> InP surface in a<br />

semi dust-free atmosphere. Next, As2S3 solution is<br />

dispensed on <strong>the</strong> InP surface with a positive pressure<br />

<strong>of</strong> 0.5 psi for 1μs. this is continually performed manually<br />

displacing <strong>the</strong> InP surface about a 1mm in <strong>the</strong><br />

x-axis and recording <strong>the</strong> micro-drops’ position.<br />

In <strong>the</strong> heat treatment phase, <strong>the</strong> micro-droplets are<br />

introduced in a vacuum oven and baked under vacuum<br />

at 50°C for 45 minutes in dark due to <strong>the</strong> pho-


tosensitivity <strong>of</strong> chalcogenide glasses. After this, <strong>the</strong><br />

micro-droplets become plano-convex micro-lenses.<br />

Following deposition and treatment, <strong>the</strong> micro-lenses<br />

are analyzed with a light microscope to estimate size,<br />

shape and quality <strong>of</strong> <strong>the</strong> droplets. After this is complete,<br />

<strong>the</strong> best samples undergo pr<strong>of</strong>ilometry to determine<br />

<strong>the</strong>ir specific dimensions and focal length.<br />

III. RESULTS<br />

The micro-lenses fabricated can be divided in two<br />

categories. The first one range in from 120μm to<br />

350μm in diameter. The yield <strong>of</strong> successful lenses<br />

from this group is approximately 85%. The second<br />

category is micro-lenses that range from 7μm to<br />

30μm in diameter. However, <strong>the</strong> yield <strong>of</strong> successful<br />

lenses from this latter group is only about 10% due to<br />

<strong>the</strong> presence <strong>of</strong> dust and o<strong>the</strong>r contamination. Thus<br />

fur<strong>the</strong>r study must be performed to increase <strong>the</strong> yield<br />

and improve <strong>the</strong> technique.<br />

Table 1 shows measurements obtained through<br />

pr<strong>of</strong>ilometry <strong>of</strong> four lenses in two different InP substrates.<br />

Figure 1 shows plot <strong>of</strong> data obtained through<br />

superficial pr<strong>of</strong>iling <strong>of</strong> <strong>the</strong> C2 micro-lens.<br />

Figure 2 shows a microscopic picture at 100x<br />

magnification <strong>of</strong> a lens in this range. Although pr<strong>of</strong>ilometry<br />

on this have not been done yet, though calibrated<br />

images, we estimate its diameter is about 8<br />

μm, with a height <strong>of</strong> about 0.6 μm corresponding to a<br />

focal length <strong>of</strong> about 20μm. As it mentioned before,<br />

fur<strong>the</strong>r improvement on <strong>the</strong> technique is necessary to<br />

fabricate lens in <strong>the</strong> range <strong>of</strong> 7-30 μm.<br />

Lens Diameter<br />

(μm)<br />

Height<br />

(μm)<br />

Focal<br />

Length<br />

(μm)<br />

C2 147 8.1 337<br />

C3 137 7.6 312<br />

A4 203 8.6 617<br />

A5 331 22 644<br />

Table 1: Micro-lenses’ dimensions<br />

232<br />

Pr<strong>of</strong>ile (μm)<br />

10<br />

8<br />

6<br />

4<br />

2<br />

0<br />

-2<br />

100 150 200 250 300<br />

Distance (μm)<br />

Figure 1: Pr<strong>of</strong>ilometry <strong>of</strong> C2 Micro-lens<br />

Figure 2: Micro-lens at 100x magnification<br />

ACKNOWLEDGMENTS<br />

The authors acknowledge partial financial support <strong>of</strong><br />

Mid-Infrared Technologies for Health and <strong>the</strong> Environment<br />

(MIRTHE). The authors also acknowledge<br />

assistance form Shanshan Song and Candice Tsay.<br />

REFERENCES<br />

[1] C. Gmachl, F. Capasso, D. L. Sivco, and A. Cho.<br />

Recent progress in quantum cascade lasers and<br />

Applications. Reports on Progress in Physics. 64,<br />

1533–1601 October 2001.<br />

[2] H. Hisakuni and K. Tanaka Optical fabrication <strong>of</strong><br />

microlenses in chalcogenide glasses. Optics letters,<br />

Vol. 20, No. 9, May 1995.<br />

[3] S. Song, S. Howard, Z. Liu, A. Dirisu, C.<br />

Gmachl, C. Arnold. Mode tuning <strong>of</strong> quantum<br />

cascade lasers through optical processing <strong>of</strong><br />

chalcogenide glass claddings. Optics letters. Applied<br />

Physics Letters, 89, 041115, July 2006.<br />

[4] H. Zhao, L. Yu, Y. Huang. Investigation <strong>of</strong> a<br />

chemically treated InP(1 0 0) surface during hydrophilic<br />

wafer bonding process. Materials<br />

Science and Engineering B, 128 (2006) 93–97,<br />

January 2006.


Abstract<br />

Structure and Dynamics <strong>of</strong> <strong>the</strong> Protein P7 from<br />

Bacteriophage 12<br />

Towards attaining <strong>the</strong> complete structural<br />

and functional characterization <strong>of</strong> all proteins<br />

<strong>of</strong> <strong>the</strong> cystoviral PX we have obtained <strong>the</strong><br />

crystal structure <strong>of</strong> <strong>the</strong> P7 protein (from <strong>the</strong><br />

12 cystovirus) at 1.8 Å resolution. We find<br />

that both <strong>the</strong> N-terminal core and <strong>the</strong><br />

dynamic C-terminus tail <strong>of</strong> P7 play a role in<br />

RNA recognition leading to a significant<br />

reduction <strong>of</strong> <strong>the</strong> degree <strong>of</strong> disorder in <strong>the</strong> Cterminal<br />

tail [1].<br />

I. INTRODUCTION<br />

The cystoviruses ( 6-14)[2] constitute an<br />

unique group <strong>of</strong> enveloped bacteriophages<br />

that utilize a three-segmented doublestranded<br />

RNA genome, infect strains <strong>of</strong> <strong>the</strong><br />

plant pathogen, Pseudomonas phaseolicola<br />

and share an overall organization and<br />

replicative mechanism analogous to <strong>the</strong><br />

reoviruses[3]. The innermost layer <strong>of</strong> <strong>the</strong><br />

cystoviral virion consists <strong>of</strong> an arrangement<br />

<strong>of</strong> four proteins: P1 (major capsid protein),<br />

P2 (RNA-directed RNA polymerase), P4<br />

(packaging ATPase) and P7 (putative<br />

assembly co-factor) (Figure 1).<br />

Figure 1: Organization <strong>of</strong> cystoviruses.<br />

These proteins that are encoded by <strong>the</strong> Lsegment<br />

<strong>of</strong> viral genome constitute <strong>the</strong><br />

polymerase complex (PX) that is responsible<br />

for genome replication and transcription. Of<br />

<strong>the</strong> four PX proteins (Figure 1), highresolution<br />

structures are available for <strong>the</strong><br />

RNA-directed RNA polymerase (RdRp) P2<br />

Ertan Eryilmaz, Ranajeet Ghose<br />

The City College <strong>of</strong> New York<br />

New York, USA<br />

{eertan,rghose}@sci.ccny.cuny.edu<br />

233<br />

from 6 and <strong>the</strong> hexameric ATPase motor<br />

P4 from 12. The least characterized <strong>of</strong> <strong>the</strong><br />

PX proteins is <strong>the</strong> protein P7. Little is known<br />

about its structure except that it exists in<br />

solution as an elongated dimer[4]. We solved<br />

<strong>the</strong> structure <strong>of</strong> N-terminal core <strong>of</strong> P7 from<br />

12, characterized <strong>the</strong> dynamics <strong>of</strong> <strong>the</strong> Cterminal<br />

tail and <strong>the</strong> core, and provided<br />

insight into its possible role in RNA<br />

recognition [1].<br />

II. RESULTS<br />

The Highly Flexible C-terminal Tail <strong>of</strong> P7<br />

Interacts Minimally with <strong>the</strong> Protein<br />

Core[1]<br />

The 15 N, 1 H HSQC spectrum <strong>of</strong> fullyprotonated<br />

full-length P7 (1-169, P7fl) from<br />

12 revealed a set <strong>of</strong> highly intense peaks in<br />

a background <strong>of</strong> broad resonances indicative<br />

<strong>of</strong> a highly unstructured region in <strong>the</strong> protein.<br />

These intense resonances were not seen in<br />

a construct <strong>of</strong> P7 (P7 C, 1-129) lacking <strong>the</strong><br />

last 40 residues from <strong>the</strong> C-terminal tail<br />

(Figure 2).<br />

Figure 2. The flexible C-terminal tail <strong>of</strong> P7 interacts<br />

minimally with <strong>the</strong> protein core. (a) Overlay <strong>of</strong> 15 N, 1 H<br />

TROSY spectra (800 MHz) <strong>of</strong> 15 N,REDPRO- 2 H-labeled<br />

P7fl (black) and P7 C (red). Also shown are <strong>the</strong> intense<br />

resonances corresponding to <strong>the</strong> C-terminal tail<br />

residues in a 15N,1H HSQC spectrum (600 MHz, cyan)<br />

<strong>of</strong> fully-protonated P7fl. (b) Resonance assignments <strong>of</strong><br />

<strong>the</strong> flexible C-terminal tail resonances in fully-protonated<br />

full-length P7 (p7fl).


Structure <strong>of</strong> P7 Core<br />

P7 forms a novel / -fold[1]. The<br />

dimerization surface consists <strong>of</strong> interactions<br />

involving helices 2 and 3 on each<br />

monomer (Figure 3) and is composed <strong>of</strong> <strong>the</strong><br />

largely hydrophobic residues.<br />

Figure 3. Structure <strong>of</strong> <strong>the</strong> P7 C biological unit. (a)<br />

P7 C forms a symmetric homodimer in solution. The<br />

monomers are colored blue and red.<br />

Residual Structure <strong>of</strong> P7 Tail and RNA<br />

Interactions<br />

Although <strong>the</strong> 15 N-{ 1 H} NOE values for <strong>the</strong> Cterminal<br />

tail resonances were very low<br />

indicative <strong>of</strong> extensive disorder on <strong>the</strong> ps-ns<br />

timescale, <strong>the</strong> values for <strong>the</strong> residues<br />

comprising <strong>the</strong> stretch between Leu136 and<br />

Gln145 on <strong>the</strong> N-terminal end <strong>of</strong> <strong>the</strong> tail were<br />

significantly higher indicating a higher<br />

degree <strong>of</strong> local order. (see Figure 4(a)).<br />

Figure 4. (a) Steady-state 15 N- { 1 H} NOE data at<br />

600MHz for <strong>the</strong> 41 C-terminal tail residues in fully<br />

protonated full-length P7. (b) The difference in <strong>the</strong><br />

deviation <strong>of</strong> <strong>the</strong> C and C chemical shifts from <strong>the</strong><br />

corresponding random coil values for <strong>the</strong> 41 C-terminal<br />

residues in fully-protonated full-length P7 (P7fl).<br />

234<br />

Given <strong>the</strong> requirement <strong>of</strong> P7 in maintaining<br />

genome packaging efficiency and<br />

transcriptional fidelity we wanted to test <strong>the</strong><br />

possible role <strong>of</strong> P7 in RNA recognition; we<br />

have recorded a series <strong>of</strong> 15 N, 1 H HSQC<br />

experiments with varying amounts <strong>of</strong> a 5-nt<br />

RNA constructs corresponding to <strong>the</strong> 5’-ends<br />

<strong>of</strong> <strong>the</strong> plus-strands <strong>of</strong> <strong>the</strong> 12 genome. Both<br />

<strong>the</strong> core and <strong>the</strong> tail resonances showed<br />

significant chemical shift changes. Majority<br />

<strong>of</strong> <strong>the</strong> residues <strong>of</strong> <strong>the</strong> tail were located near<br />

<strong>the</strong> C-terminal end <strong>of</strong> <strong>the</strong> protein (159-163)<br />

(Figure 5(a)), and <strong>the</strong> moving residues <strong>of</strong> <strong>the</strong><br />

core domain make continuous surfaces<br />

(Figure 5(b)).<br />

Figure 5. Interactions <strong>of</strong> <strong>the</strong> C-terminal tail and Nterminal<br />

core with RNA. (a) 15 N, 1 H HSQC spectrum <strong>of</strong><br />

fully protonated 15 N-labeled P7fl with varying RNA<br />

concentrations. Resonances showing significant<br />

chemical shiftchanges are shown. (b) The surface maps<br />

<strong>of</strong> N-terminal core; two specific surfaces display large<br />

chemical shift changes in <strong>the</strong> presence <strong>of</strong> RNA – key<br />

residues are labeled and <strong>the</strong> color<br />

REFERENCES<br />

[1] Eryilmaz, E., Benach, J., Su, M., Seetharaman, J.,<br />

Dutta, K., Wei, H., Gottlieb, P., Hunt, J.F., and<br />

Ghose, R. (<strong>2008</strong>). Structure and dynamics <strong>of</strong> <strong>the</strong><br />

P7 protein from <strong>the</strong> bacteriophage f12. J. Mol. Biol.<br />

382, 402-422.<br />

[2] Mindich, L., Qiao, X., Qiao, J., Onodera, S.,<br />

Romantschuk, M. & Hoogstraten, D. (1999).<br />

Isolation <strong>of</strong> additional bacteriophages with<br />

genomes <strong>of</strong> segmented double-stranded RNA. J.<br />

Bacteriol. 181, 4505-8.<br />

[3] Mertens, P. (2004). The dsRNA viruses. Virus Res.<br />

101, 3-13.<br />

[4] Huiskonen, J. T., de Haas, F., Bubeck, D.,<br />

Bamford, D. H., Fuller, S. D. & Butcher, S.<br />

J.(2006). Structure <strong>of</strong> <strong>the</strong> bacteriophage f6<br />

nucleocapsid suggests a mechanism for sequential<br />

RNA packaging. Structure 14, 1039-48.


Dsh is required for <strong>the</strong> phosphorylation <strong>of</strong> Dpr1a by<br />

CKIδ<br />

Evelyn Teran, and Joni M. Seeling<br />

Queens College<br />

City University <strong>of</strong> New York<br />

New York,United States <strong>of</strong> America<br />

Abstract<br />

Colon cancer is <strong>the</strong> third most common form <strong>of</strong><br />

cancer and <strong>the</strong> second leading cause <strong>of</strong> cancerrelated<br />

death in <strong>the</strong> Western world causing<br />

655,000 deaths worldwide per year [1]. The Wnt<br />

signaling pathway plays an important role during<br />

embryonic development and colon cancer. The<br />

PDZ-Binding domain <strong>of</strong> protein XDpr1a interacts<br />

with <strong>the</strong> PDZ domain <strong>of</strong> protein Dsh to inhibit <strong>the</strong><br />

Wnt signaling pathway. β-catenin, a protein found<br />

in abundance in colorectal cancerous cells is reduced<br />

in abundance by XDpr1a which also decreases<br />

activation <strong>of</strong> Wnt responsive genes. We<br />

have found that XDpr1a is phosphorylated by<br />

CKIδ, an activator <strong>of</strong> Wnt signaling, in <strong>the</strong> presence<br />

<strong>of</strong> Dsh. PDZ-binding mutants <strong>of</strong> XDpr1a<br />

which do not bind Dsh are not phosphorylated by<br />

CK1δ in <strong>the</strong> presence <strong>of</strong> Dsh. Mutation <strong>of</strong> <strong>the</strong> PDZ-<br />

B binding loop <strong>of</strong> <strong>the</strong> Dsh PDZ domain prevents<br />

Dsh from promoting XDpr1a phosphorylation,<br />

whereas mutation within <strong>the</strong> PDZ domain but outside<br />

<strong>of</strong> <strong>the</strong> PDZ-B binding loop does not.<br />

Introduction<br />

The Wnt signaling pathway is involved in<br />

a large range <strong>of</strong> embryonic patterning events and<br />

cancer. One <strong>of</strong> <strong>the</strong> most distinctive effects <strong>of</strong> <strong>the</strong><br />

Wnt proteins is <strong>the</strong>ir ability to induce formation <strong>of</strong><br />

<strong>the</strong> dorsoventral body axis in <strong>the</strong> model organism<br />

Xenopus laevis. It has been shown that activation <strong>of</strong><br />

<strong>the</strong> Wnt/�-catenin signaling pathway is commonly<br />

associated with <strong>the</strong> development <strong>of</strong> colorectal cancer<br />

and various types <strong>of</strong> cancers [2]. In <strong>the</strong> absence<br />

<strong>of</strong> <strong>the</strong> Wnt ligand, glycogen synthase kinase 3�<br />

(GSK-3�) phosphorylates �-catenin in <strong>the</strong> �catenin<br />

degradation complex. This leads to <strong>the</strong><br />

ubiquitination and degradation <strong>of</strong> �-catenin. Hence,<br />

�-catenin levels are maintained low. However,<br />

when a Wnt ligand is present, <strong>the</strong> ligand binds to<br />

<strong>the</strong> seven pass transmembrane receptor Frizzled<br />

(Fz), and <strong>the</strong> signal activates <strong>the</strong> protein Dishevelled<br />

(XDsh) by phosphorylation. Activated XDsh<br />

destabilizes <strong>the</strong> �-catenin degradation complex, and<br />

�-catenin <strong>the</strong>refore accumulates in <strong>the</strong> cytoplasm.<br />

As a result, �-catenin enters <strong>the</strong> nucleus, where it<br />

forms a complex with members <strong>of</strong> T-cell factor/lymphoid-enhancing<br />

factor (TCF/LEF) and<br />

upregulates TCF/LEF-dependent transcription <strong>of</strong><br />

target genes such as c-Myc and siamois.<br />

Email: Evelyn0021@aol.com<br />

235<br />

Regulatory subunit B56 <strong>of</strong> protein phosphatase<br />

2A(PP2A), interacts with APC in <strong>the</strong> Wnt<br />

pathway [3], PP2A is a heterotrimeric intracellular<br />

serine-threonine phosphatase. PP2A contains a<br />

conserved catalytic (C), structural (A), and a variable<br />

regulatory (B) subunit [3]. There are three<br />

families <strong>of</strong> B subunits; B55, B56 and B72. Each<br />

family subunit has multiple is<strong>of</strong>orms, leading to<br />

many distinct holoenzymes, which explains <strong>the</strong><br />

diversity and multiple functions <strong>of</strong> PP2A. The B<br />

subunit provides substrate specificity to PP2A and<br />

subcellular localization. Fur<strong>the</strong>rmore, it found that<br />

casein kinase I (CKI), more specifically CKI� and<br />

CKI�, phosphorylate and destabilize <strong>the</strong> �-catenin<br />

degradation complex through <strong>the</strong> dissociation <strong>of</strong><br />

PP2A [4].<br />

Dishevelled (XDsh) is a key activator <strong>of</strong><br />

<strong>the</strong> Wnt pathway that destabilizes <strong>the</strong> �-catenin<br />

degradation complex. XDsh contains a central Post<br />

synaptic density-95, Discs-large, Zonula occludens-1<br />

(PDZ) domain. The PDZ domain is involved<br />

in protein-protein interaction which usually<br />

binds to <strong>the</strong> extreme C-terminus <strong>of</strong> transmembrane<br />

proteins containing PDZ-binding motifs (PDZ-B)<br />

[4]. Each domain <strong>of</strong> XDsh binds to specific proteins<br />

that transduce Wnt signaling to different<br />

downstream pathways. Dapper (XDpr1a), functions<br />

as an inhibitory protein in <strong>the</strong> pathway by<br />

inhibiting XDsh functions during Xenopus embryogenesis.<br />

XDpr1a also downregulates levels <strong>of</strong><br />

�-catenin accumulation in <strong>the</strong> cytoplasm and <strong>the</strong>refore<br />

also downregulates Wnt ligand mediated transcription.<br />

XDpr1a contains a PDZ-binding domain<br />

at its C-terminal end which binds to <strong>the</strong> PDZ domain<br />

<strong>of</strong> XDsh.<br />

Materials and Methods<br />

Transcription /Translation<br />

In vitro transcription and translation was<br />

performed with [ 35 S]methionine (Amersham Biosciences<br />

Corp., Piscataway, NJ) using TNT SP6<br />

Quick Coupled Transcription/Translation System<br />

(Promega, Madison, WI) according to <strong>the</strong> manufacturer’s<br />

instructions. Unlabeled XDsh was prepared<br />

with <strong>the</strong> TNT SP6 Quick Coupled kit using cold<br />

methionine.<br />

In Vitro phosphorylation assay<br />

[ 35 S]methionine labeled XDpr1a and ei<strong>the</strong>r<br />

[ 35 S]methionine labeled or unlabeled XDsh<br />

were mixed also containing MgCl2 and ATP. Rat


CKIδ lacking its autoinhibitory carboxy-terminal<br />

domain (New England Biolabs, Beverly, MA) was<br />

added to a final concentration. Samples were resolved<br />

using SDS-PAGE and visualized by autoradiography<br />

Results We have found that Dapper is phosphorylated<br />

by CK1 in <strong>the</strong> presence <strong>of</strong> Disheveled. PDZbinding<br />

mutants <strong>of</strong> Dapper are not phosphorylated<br />

by CK1 in <strong>the</strong> presence <strong>of</strong> Disheveled and <strong>the</strong> Disheveled<br />

mutation <strong>of</strong> <strong>the</strong> PDZ domain is still able<br />

to promote dapper phosphorylation, but a mutation<br />

N-terminal to <strong>the</strong> PDZ domain is unable.<br />

Mutant XDpr1a and XDsh constructs were obtained<br />

from Randall Moon’s lab. We analyzed<br />

three XDpr1a mutants. The first mutant, DprΔLZ,<br />

lacks <strong>the</strong> 36 amino acid leucine zipper region, <strong>the</strong><br />

second mutant XDpr1aMNTV contains a<br />

Thr822Asn point mutation in <strong>the</strong> PDZ-B domain,<br />

and <strong>the</strong> third mutant XDpr1aΔMTTV lacks <strong>the</strong><br />

PDZ-B domain.<br />

The results in Figure 1 show that DprΔLZ protein<br />

behaved similar to wild-type XDpr1a and underwent<br />

a gel-shift <strong>of</strong> hyperphosphorylation in <strong>the</strong><br />

presence <strong>of</strong> CKIδ. XDpr1aΔMTTV did not undergo<br />

a mobility shift in <strong>the</strong> presence <strong>of</strong> CKIδ, suggesting<br />

that it is not phosphorylated by CKIδ.<br />

XDprMNTV behaved similar to XDpr1aΔMTTV,<br />

and did not exhibit a mobility shift as well. Therefore,<br />

<strong>the</strong>se data suggest that unmodified PDZ-B<br />

domain in XDpr1a is required for <strong>the</strong> XDshdependent<br />

phosphorylation <strong>of</strong> XDpr1a by CK1δ.<br />

In addition, we examined mutations in<br />

XDsh that prevent its interaction with XDpr1a, to<br />

determine if <strong>the</strong>y had any effects on <strong>the</strong> phosphorylation<br />

<strong>of</strong> XDpr1a by CKI. In a paper by Cheyette<br />

(et al., 2002), it was found that an N317T point<br />

mutation in Dsh’s PDZ-B binding domain<br />

(XDsh*α) (Figure.2), disrupts its interaction with<br />

XDpr1a. On <strong>the</strong> o<strong>the</strong>r hand, a triple mutation in a<br />

PDZ domain loop upstream <strong>of</strong> XDsh’s PDZ-B<br />

binding domain (XDsh***β-β) does not. We<br />

wanted to find out whe<strong>the</strong>r a mutation <strong>of</strong> <strong>the</strong> PDZ<br />

domain <strong>of</strong> XDsh could prevent XDsh from promoting<br />

a mobility shift and hyperphosphorylation <strong>of</strong><br />

XDpr1a in <strong>the</strong> presence <strong>of</strong> CK1δ. Our results show<br />

that XDpr1a did not exhibit a mobility shift in <strong>the</strong><br />

presence <strong>of</strong> XDsh*α but in fact XDpr1a exhibited a<br />

mobility shift in <strong>the</strong> presence <strong>of</strong> wild type XDsh<br />

and XDsh***β-β. The inability <strong>of</strong> a XDsh*� to<br />

promote <strong>the</strong> phosphorylation <strong>of</strong> XDpr1a by CKI<br />

due to its point mutation in <strong>the</strong> PDZ-B binding<br />

domain <strong>of</strong> its PDZ domain, suggests that Dsh must<br />

bind to XDpr1a to promote its phosphorylation.<br />

236<br />

Figure 1: XDpr1a lacking PDZ-B is not phosphor<br />

ylated by CKI�<br />

Figure 2: Mutation <strong>of</strong> <strong>the</strong> Dsh PDZ-B binding loop<br />

abrogates XDpr1a phosphorylation by CKI�.<br />

Conclusion<br />

We have found that an interaction <strong>of</strong> PDZ<br />

domain <strong>of</strong> Dsh and <strong>the</strong> PDZ-binding domain <strong>of</strong><br />

XDpr1a are required for gel-shift, indicating phosphorylation.<br />

When <strong>the</strong> domains were mutated or<br />

deleted, XDpr1a’s phosphorylation by CKI� was<br />

not functional. By doing TNT assays, we have<br />

shown that a deletion (� MTTV) and point mutation<br />

(MNTV) <strong>of</strong> <strong>the</strong> PDZ-binding domain in dapper<br />

prevents it from binding to Dsh. Conversely, we<br />

have shown that a triple mutation (***β-β) outside<br />

<strong>the</strong> binding pocket <strong>of</strong> Dsh appears to affect gel shift<br />

to XDpr1a, whereas a point mutation (α) within <strong>the</strong><br />

PDZ domain <strong>of</strong> Dsh does not inhibit its gel shift.<br />

Acknowledgements<br />

I’d like to thank my mentor Dr.Joni<br />

Seeling as well as my funding programs<br />

LSAmp and Bridges to Doctorate.<br />

References<br />

[1] World Health Organization, February<br />

2006.<br />

[2] X.Li, H.Yost, D.Virshup and J.Seeling<br />

(2001) The EMBO Journal, 4122-4131<br />

[3] Z.Gao, J.Seeling, V.Hill, A.Yochum and<br />

D.Virshup (2002) PNAS, 1182-1187<br />

[4]Cheyette, B.N., J. S. Waxman, et al. (2002).<br />

Dev Cell 2(4): 449-61


Momentum <strong>of</strong> Optical Airy Beams<br />

H. I. Sztul and R. R. Alfano (Faculty Mentor)<br />

Institute <strong>of</strong> Ultrafast Spectroscopy and Lasers<br />

The Graduate Center and City College <strong>of</strong> <strong>the</strong> City University <strong>of</strong> New York<br />

New York, USA<br />

Email: hsztul@gc.cuny.edu<br />

Abstract — The optical Airy beam is a unique solution to<br />

<strong>the</strong> wave equation that curves as it travels through space. We<br />

analyze and describe <strong>the</strong> evolution <strong>of</strong> <strong>the</strong> Poynting vector and<br />

angular momentum <strong>of</strong> <strong>the</strong> optical Airy beam as it propagates<br />

through space. We show that while <strong>the</strong> total angular momentum<br />

<strong>of</strong> <strong>the</strong> Airy beam is zero, <strong>the</strong> angular momentum <strong>of</strong> <strong>the</strong> main<br />

intensity peak and <strong>the</strong> Airy “tail” are non-zero.<br />

I. INTRODUCTION<br />

Until recent laser beams have been known to travel along<br />

straight paths. A new class <strong>of</strong> laser beam known as<br />

<strong>the</strong> Airy beam, that is a solution to <strong>the</strong> paraxial wave<br />

equation, follows a parabolic trajectory, curving through<br />

space [1, 2, 3]. These beams <strong>of</strong> light are diffraction-free<br />

accelerating beams <strong>of</strong> light that have <strong>the</strong> spatial pr<strong>of</strong>ile<br />

<strong>of</strong> <strong>the</strong> Airy function. The 1-D and 2-D Airy solution was<br />

most recently shown experimentally in optical beams <strong>of</strong><br />

light [2]. The Airy beam is said to be free <strong>of</strong> diffractive<br />

spreading and freely accelerating in <strong>the</strong> absence <strong>of</strong> any<br />

external potential [1, 4].<br />

In this paper <strong>the</strong> Poynting vector and angular momentum<br />

<strong>of</strong> <strong>the</strong> Airy beam as it propagates through space is<br />

explored.<br />

Figure 1: The numerically calculated Poynting vector in<br />

<strong>the</strong> sx − sy plane, <strong>of</strong> <strong>the</strong> finite-energy Airy wave with<br />

a = 0.15 at ξ =0(a), ξ =0.025 (b), ξ =0.050 (c),<br />

ξ =0.075 (d). The intensity <strong>of</strong> <strong>the</strong> Airy field is shown in<br />

<strong>the</strong> background <strong>of</strong> each frame.<br />

II. POYNTING VECTOR OF AIRY BEAMS<br />

The (2 + 1)D paraxial wave equation has <strong>the</strong> nondispersive<br />

solution that limits infinite energy flow <strong>of</strong> <strong>the</strong><br />

form [1, 3, 4],<br />

φ(sx,sy,ξ) =<br />

�<br />

Ai sx − (ξ/2) 2 �<br />

+ iaξ<br />

�<br />

Ai sy − (ξ/2) 2 �<br />

+ iaξ<br />

�<br />

× exp asx + asy − (aξ 2 /2) − i(ξ 3 /12)<br />

−i(a 2 electrodynamics [5, 6]. This vector is routinely examined<br />

for plane waves but has received considerable attention in<br />

<strong>the</strong> literature with regard to Laguerre-Gaussian beams <strong>of</strong><br />

light that have helical wavefronts [7, 9].<br />

The Poynting vector is defined as [6]:<br />

�S =(c/4π)<br />

�<br />

ξ/2) + iξ(sx + sy)/2 . (1)<br />

where sx = x/x0 and sy = y/y0 are normalized transverse<br />

coordinates, ξ = z/k(x0 + y0) is a normalized<br />

propagation distance, x0 and y0 are normalization constants,<br />

and k =2πn/λ0. The background <strong>of</strong> Figure 1(ad)<br />

shows <strong>the</strong> intensity in <strong>the</strong> transverse plane at various<br />

ξ-positions when a =0.15 and xo = yo =0.1.<br />

The rate <strong>of</strong> electromagnetic energy flow per unit area,<br />

or <strong>the</strong> Poynting vector, is a commonly known quantity in<br />

� E × � B, (2)<br />

where c is <strong>the</strong> speed <strong>of</strong> light. Given a vector potential<br />

�A =ˆɛu(sx,sy,ξ)exp[ik2ξ(x0 + y0)], where ˆɛ is an arbitrary<br />

polarization and u(sx,sy,ξ) is <strong>the</strong> Airy field amplitude<br />

given by Eq. (1), we can use <strong>the</strong> � E and � B-fields<br />

in <strong>the</strong> Lorenz gauge, as given by Ref. [8], to calculate <strong>the</strong><br />

time-averaged Poynting vector, < � S>, numerically.<br />

Figure 1 (a-d) shows <strong>the</strong> numerically computed ˆsxand<br />

ˆsy-components <strong>of</strong> <strong>the</strong> Poynting vector for a =0.15<br />

at ξ =0.025, 0.050, 0.075, and 0.1, respectively. The direction<br />

and magnitude <strong>of</strong> <strong>the</strong> arrows (shown in red) correspond<br />

to <strong>the</strong> direction and magnitude <strong>of</strong> <strong>the</strong> energy flow<br />

in <strong>the</strong> transverse plane. The intensity <strong>of</strong> <strong>the</strong> Airy field is<br />

shown in <strong>the</strong> background <strong>of</strong> each frame to show <strong>the</strong> direction<br />

<strong>of</strong> <strong>the</strong> energy flow in relation to <strong>the</strong> peaks <strong>of</strong> <strong>the</strong><br />

237


Figure 2: The numerically calculated angular momentum<br />

in <strong>the</strong> ˆ ξ-direction <strong>of</strong> <strong>the</strong> finite-energy Airy wave given by<br />

eq. (3) with a = 0.15 at ξ =0.025 (a), ξ =0.050 (b),<br />

ξ =0.075 (c), ξ =0.100 (d). Reds are positive values,<br />

blues are negative values, and green is zero.<br />

Airy beam. The flow <strong>of</strong> energy <strong>of</strong> <strong>the</strong> main peak at ξ>0<br />

is consistently pointed at 45 o relative to <strong>the</strong> sx −sy plane<br />

at all ξ-locations. In contrast, <strong>the</strong> direction <strong>of</strong> <strong>the</strong> energy<br />

flow for <strong>the</strong> Airy tails, or <strong>the</strong> peaks oriented along <strong>the</strong><br />

horizontal or vertical axis approaches a direction perpendicular<br />

to that axis. The net energy flow is measured,<br />

however, to be constant and pointed in <strong>the</strong> direction that<br />

<strong>the</strong> main peak moves, i.e 45 o or along <strong>the</strong> line sx = sy.<br />

III. ANGULAR MOMENTUM OF AIRY<br />

BEAMS<br />

It is well known that as �p ∝ � E × � B from which follows<br />

that angular momentum density about <strong>the</strong> ˆ ξ-direction is<br />

jξ(sx,sy,ξ) =<br />

� �<br />

�r × �E × B�<br />

ξ<br />

= sx · Ssy + sy · Ssx. (3)<br />

The numerically computed < � S > shown in Figure<br />

1 is used to calculate <strong>the</strong> angular momentum in <strong>the</strong> ˆ ξdirection.<br />

Figure 2 (a-d) shows <strong>the</strong> ˆ ξ-component <strong>of</strong> <strong>the</strong><br />

angular momentum density with a =0.15 at ξ =0.025,<br />

0.050, 0.075, and 0.100, respectively. At ξ =0<strong>the</strong> computed<br />

angular momentum is zero so it is not shown. In<br />

Figure 2 reds are positive values (clockwise), blues are<br />

negative values (counter-clockwise), and green is zero.<br />

As <strong>the</strong> beam propagates, <strong>the</strong> net angular momentum in<br />

<strong>the</strong> ξ-direction is always zero. The spatial distribution <strong>of</strong><br />

<strong>the</strong> angular momentum is changing however, and locally<br />

has non-zero values <strong>of</strong> angular momentum. Not only is<br />

<strong>the</strong> angular momentum changing in <strong>the</strong> Airy tails, but<br />

238<br />

<strong>the</strong>re are also changes to <strong>the</strong> angular momentum in <strong>the</strong><br />

main Airy peak. This change <strong>of</strong> angular momentum is<br />

a torque that corresponds to <strong>the</strong> force present due to <strong>the</strong><br />

changing linear momentum.<br />

IV. CONCLUSIONS<br />

In this paper we analyze <strong>the</strong> spatial evolution <strong>of</strong> <strong>the</strong> Airy<br />

solution to <strong>the</strong> paraxial wave equation and show that<br />

while momentum is changing, energy and momentum<br />

are conserved. We point out that <strong>the</strong> linear and angular<br />

momenta are changing as <strong>the</strong> Airy beam propagate in<br />

<strong>the</strong> ξ-direction which should implications when analyzing<br />

<strong>the</strong> velocity <strong>of</strong> this field. Some form <strong>of</strong> <strong>the</strong> velocity,<br />

be it phase, energy, or signal velocity, should be changing<br />

as <strong>the</strong> beam propagates and should be investigated in future<br />

studies. These beams have promise for applications<br />

in optical trapping, imaging, and spectroscopy where a<br />

sample might interact with a changing momentum and<br />

spatially varying angular momentum.<br />

ACKNOWLEDGMENTS<br />

This work was supported in part by organized research<br />

at CCNY, NASA URC - Center for Optical Sensing and<br />

Imaging (COSI) at CCNY (NASA Grant No.: NCC-<br />

1-03009), and DOD Center <strong>of</strong> Nanoscale Photonics at<br />

CCNY.<br />

REFERENCES<br />

[1] M.V. Berry and N.L. Balazs, “Nonspreading wave<br />

packets,” Am. J. Phys. 47, 264 (1979).<br />

[2] G.A. Siviloglou, J. Broky, A. Dogariu, and D. N.<br />

Christodoulides, “Observation <strong>of</strong> Accelerating Airy<br />

Beams,” Phys. Rev. Lett. 99, 213901 (2007).<br />

[3] G. A. Siviloglou and D. N. Chistodoulides, “Accelerating<br />

finite energy Airy beams,” Opt. Lett. 32, 979<br />

(2007).<br />

[4] D. M. Greenberger, “Comment on ’Non-Spreading<br />

Wave Packets’, ” Am. J. Phys. 48, 256 (1980).<br />

[5] J.D. Jackson, Classical Electrodynamics, Wiley,<br />

New York, 1962.<br />

[6] Max Born and Emil Wolf, Principles <strong>of</strong> Optics,<br />

7th Ed., Cambridge University Press, Cambridge,<br />

1999.<br />

[7] L. Allen and M.J. Padgett, “The Poynting vector<br />

in Laguerre-Gaussian beams and <strong>the</strong> interpretation<br />

<strong>of</strong> <strong>the</strong>ir angular momentum desnity,” Opt. Comm.<br />

184, 67-71 (2000).<br />

[8] L. Allen, M. J. Padgett, and M. Babiker, Progress<br />

in Optics XXXIX, pp. 291-372 (1999).<br />

[9] H.I. Sztul and R. R. Alfano, “Double-slit interference<br />

with Laguerre-Gaussian beams,” Opt. Lett. 31,<br />

999-1001 (2006).


Electrochemistry, Optical Spectroscopy and DFT Calculations <strong>of</strong><br />

Glutathionylcobalamin<br />

Iya Likhtina and Ronald L. Birke (Faculty Mentor)<br />

Department <strong>of</strong> Chemistry and Center for Analysis <strong>of</strong> Structures and Interfaces (CASI), The City<br />

College and The Graduate School and University Center <strong>of</strong> City University <strong>of</strong> New York<br />

Email: ilikhtina@gc.cuny.edu<br />

Abstract - The bioinorganic chemistry <strong>of</strong> vitamin<br />

B12 is an active area <strong>of</strong> research [1]. Forms <strong>of</strong><br />

vitamin B12 in <strong>the</strong> reduced state are powerful reducing<br />

agents which have been used for dehalogenation<br />

<strong>of</strong> organic halide compounds in aqueous<br />

solution such as <strong>the</strong> reductive dehalogenation <strong>of</strong><br />

brom<strong>of</strong>orm for environmental remediation [2]. In<br />

this work we studied <strong>the</strong> cyclic voltammetry (CV),<br />

UV-Visible spectroscopy, and circular dichroism<br />

(CD) <strong>of</strong> GSCbl in solution. Also, DFT calculations<br />

(B3LYP/6-311G*) have been used to calculate <strong>the</strong><br />

properties <strong>of</strong> glutathionylcobalamin, GS-Cbl. UV-<br />

Vis spectroscopy was employed to monitor <strong>the</strong> formation<br />

<strong>of</strong> GSCbl which showed well-defined<br />

isosbestic points. Time dependent (TD) DFT calculations<br />

were again used to calculated <strong>the</strong> electronic<br />

excitation energies found in <strong>the</strong> electronic spectra.<br />

I. INTRODUCTION<br />

A corrin macrocyclic ring is <strong>the</strong> base ligand<br />

for <strong>the</strong> structure <strong>of</strong> glutathionylcob(III)alamin<br />

which contains a Co3+ ion. The ring resembles<br />

<strong>the</strong> porphyrin ring but has one less carbon. With<br />

side chains and Co ion in <strong>the</strong> center <strong>of</strong> <strong>the</strong> corrin<br />

ring , <strong>the</strong> molecules are called cobalamins. Four <strong>of</strong><br />

<strong>the</strong> six coordination sites are occupied by<br />

equatorial nitrogens <strong>of</strong> <strong>the</strong> corrin ring, and <strong>the</strong>re<br />

are two axial ligands above and below <strong>the</strong> ring.<br />

The lower ligand is a dimethylbenzimidazole<br />

group in cobalamins. The sixth coordination site<br />

which is above <strong>the</strong> corrin can be substituted by<br />

many groups such as a cyano group, a hydroxyl<br />

group, a methyl group, a 5’-deoxyadenozyl group<br />

and in our case a glutathionyl group giving<br />

glutathionylcobalamin, GS-Cbl.<br />

In biological tissues glutathione, GSH, is a<br />

major intracellular reducing agent and GS-Cbl(III)<br />

is one <strong>of</strong> <strong>the</strong> major forms <strong>of</strong> B12 in mammalian<br />

cells. This species appears to be <strong>the</strong> natural<br />

precursor <strong>of</strong> <strong>the</strong> cobalamin coenzymes. All <strong>of</strong><br />

above evidence suggests that GS-Cbl(III) is an<br />

important intermediate for biological processes<br />

involving vitamin B12 species.<br />

239<br />

Electrochemistry has proven to be a valuable<br />

tool for investigating <strong>the</strong> electron-transfer steps <strong>of</strong><br />

<strong>the</strong> three oxidation states <strong>of</strong> cobalt and <strong>the</strong>ir<br />

coupling with ligand exchange reactions in <strong>the</strong><br />

oxidoreduction chemistry <strong>of</strong> various vitamin B12<br />

derivatives. There is only one paper on <strong>the</strong><br />

electrochemistry <strong>of</strong> GSCbl and this is from <strong>the</strong><br />

absorbed state [3]. We studied <strong>the</strong> cyclic<br />

voltammetry (CV ) on glassy carbon and mercury<br />

electrodes as a function <strong>of</strong> potential scan rate, pH<br />

and concentration <strong>of</strong> GSCbl. Digital simulation <strong>of</strong><br />

<strong>the</strong> electrode process was investigated for possible<br />

mechanisms. The CVSIM program was utilized to<br />

perform <strong>the</strong> simulation <strong>of</strong> <strong>the</strong> CV curves. The<br />

shape <strong>of</strong> a cyclic voltammetric curve reflects both<br />

electron transfer at <strong>the</strong> electrode and solution<br />

chemical reaction that are coupled to <strong>the</strong> electron<br />

transfer. Both UV-Vis and circular dichroism (CD)<br />

electronic spectroscopic techniques were used to<br />

elucidate <strong>the</strong> electronic transitions in <strong>the</strong> visible<br />

and ultraviolet regions <strong>of</strong> <strong>the</strong> spectrum.<br />

II. EXPERIMENTAL<br />

A. RESULTS<br />

The CV experiments were performed with <strong>the</strong><br />

CH 660A workstation. A three electrodes<br />

configuraton was used in all experiments. The<br />

working electrodes were a mercury drop from a<br />

BAS controlled growth Hg electrode (CGME) or a<br />

glassy carbon electrode. A Ag/AgCl/Cl (3M)<br />

electrode was used as <strong>the</strong> reference electrode and a<br />

platinum electrode was used as <strong>the</strong> counter<br />

electrode. The electrode process found on glassy<br />

carbon and mercury electrodes at pH 4 indicates a<br />

one-electron reduction after which <strong>the</strong> Co(II)-S<br />

bond breaks. The one-electron reduction<br />

mechanism has been confirmed by digital<br />

simulation at pH 4. Figure 1 shows a good fit<br />

between <strong>the</strong> simulation <strong>of</strong> <strong>the</strong> one-electron<br />

reduction, dotted line, and <strong>the</strong> experimental<br />

results. The reduction potential <strong>of</strong> GS-Cbl(III) at<br />

pH 4 is at ca. -0.9V vs. <strong>the</strong> reference electrode.


This shows that <strong>the</strong> glutathionyl ligand is a better σ<br />

-donating ligand than water but poorer than<br />

cyanide or alkyl carbanions. We also studied <strong>the</strong><br />

equilbrium reaction between GSH and Cbl(III)<br />

between pH 3 and 5 by Uv-Vis spectrocopy.<br />

Finally <strong>the</strong> structure and excited states <strong>of</strong> GS-<br />

Cbl(III) was investiagted by DFT calculations.<br />

B. COMPUTATIONAL DETAILS<br />

We have made DFT electronic structure calculations<br />

on <strong>the</strong> glutathionylcobalamin. The model <strong>of</strong> GS-<br />

Cbl(III) that we used includes <strong>the</strong> full corrin ring<br />

with all side-chain groups replaced by hydrogen<br />

atoms and with <strong>the</strong> dimethylbenzimidazole base<br />

ligand replaced by imidazole. Since <strong>the</strong> model does<br />

not include <strong>the</strong> negative phosphate-containing side<br />

chain, <strong>the</strong> GS-Cbl(III) model is a singly plus charged<br />

species. The calculations were made with <strong>the</strong><br />

Gaussian 03 package <strong>of</strong> programs using <strong>the</strong> B3LYP<br />

hybrid functional level <strong>of</strong> <strong>the</strong>ory and <strong>the</strong> 6-311G**<br />

basis set.<br />

The geometric optimization <strong>of</strong> <strong>the</strong> model shows a<br />

very good fit with <strong>the</strong> XAFS [4] and X-ray<br />

diffraction [5] results for Co-S, Co-N(equatorial) and<br />

Co-N(axial) bond distances as indicated in Table 1.<br />

TD-DFT TD-DFT GS-Cbl (III) RS-Cbl (III)<br />

R GS-Cbl GS-Cbl EXAFS, Å X-ray Diff,<br />

(III) z = (II) z = 0,<br />

Å<br />

+1, Å Å<br />

Co - S 2.2958 2.31460 2.28±0.05 2.267<br />

Co - Nax 2.0685 2.05710 2.15±0.03 2.049<br />

Co - Neq<br />

(5-mem)<br />

Co - Neq<br />

(5-mem)<br />

Co - Neq<br />

(6-mem)<br />

Co - Neq<br />

(6-mem)<br />

1.9020 1.90490 1.89±0.01 1.885<br />

1.8880 1.90582 1.891<br />

1.93645 1.95293 1.902<br />

1.93645 1.95160 1.914<br />

Table1. Comparison <strong>of</strong> Co-X bond distances in<br />

GS-Cobalamin. There are two sets (5-mem and 6mem)<br />

<strong>of</strong> equatorial Co-N bonds in <strong>the</strong> molecule.<br />

The excited state transitions calculated by TD-<br />

DFT show a good fit with <strong>the</strong> experimental results<br />

from <strong>the</strong> UV-Vis and CD optical experiments. The<br />

vertical excitations are depicted by <strong>the</strong> transitions<br />

between molecular orbitals.<br />

240<br />

CONCLUSION<br />

The GC-Cbl(III) molecule has<br />

electrochemistry, optical spectroscopy, and electron<br />

structure which is very similar to <strong>the</strong> biologically<br />

active vitamin B12 coenzymes, methylcobalamin and<br />

5’-deoxyadenozylcobalamin. This is consistent with<br />

<strong>the</strong> supposition that it is <strong>the</strong> storage form <strong>of</strong> <strong>the</strong><br />

vitamin in <strong>the</strong> human body .<br />

Current, mkAmp<br />

1 0<br />

8<br />

6<br />

4<br />

2<br />

0<br />

-2<br />

- 4<br />

-0.6 -0.7 -0.8 -0.9 -1 .0 -1 .1 - 1 .2<br />

P otent ia l, V<br />

Figure1. Simulated and experimental CV <strong>of</strong> GS-<br />

Cbl on mercury electrode. Solid line: experimental<br />

CV, dotted line: simulated CV.<br />

REFERENCES<br />

[1] J. M. Pratt, In: R. Banerjee (Ed) , Chemistry and<br />

Biochemistry <strong>of</strong> B12., John Wiley & Sons,<br />

New York , 1999, p. 104<br />

[2] Eric A. Betterton, Robert G. Arnold, Ronald J.<br />

Kuhler and Gregory A. Santo Environmental<br />

Health Perspectives, Vol. 103, Supplement 5:<br />

Biodegradation (Jun., 1995), pp. 89-91.<br />

[3] L. Xia, A. G. Cregan, L. A. Berben, N. Brash,<br />

Inorg. Chem. 43 (2004) 6848-6857.<br />

[4] Eva M. Scheuring, Irit Sagi and Mark R.<br />

Chance, Biochem (1994) 33, 6310-6315.<br />

[5] Robert K. Suto et al., Inorg. Chem., 2001, 40,<br />

2686-2692.


Surface Enhanced Raman Spectroscopy <strong>of</strong><br />

Flavanthrone<br />

Jingjing Chang 1,2 , Maria Vega Cañamares 1 , John R. Lombardi 1<br />

1 Department <strong>of</strong> Chemistry, City College <strong>of</strong> New York<br />

N.Y., N.Y. 10031<br />

2 State Key Laboratory <strong>of</strong> Supramolecular Structure and Materials, Jilin University<br />

Changchun, P. R. China, 130012<br />

Email: changjj@sci.ccny.cuny.edu<br />

Abstract —Flavanthrone is among <strong>the</strong> oldest<br />

syn<strong>the</strong>tic vat dyes. It is <strong>of</strong> considerable interest<br />

due to its high symmetry. As it has a point <strong>of</strong> inversion,<br />

no spectral lines are common to <strong>the</strong> infrared<br />

and Raman spectrum. In this work, flavanthrone<br />

has been studied by FT-Raman and Surface-enhanced<br />

Raman scattering (SERS) spectroscopy.<br />

The SERS spectra were recorded at 633 and<br />

785 nm on Ag nanoparticles. The SERS spectra<br />

show lines which are forbidden in normal Raman<br />

spectra. This is due to <strong>the</strong> reduction in symmetry<br />

originated by <strong>the</strong> interaction <strong>of</strong> <strong>the</strong> molecules with<br />

<strong>the</strong> metallic surface. The B3LYP method <strong>of</strong> <strong>the</strong><br />

Density Functional Theory (DFT) was used to<br />

identify those new lines and determine <strong>the</strong>ir origin.<br />

I. INTRODUCTION<br />

Flavanthrone is one <strong>of</strong> <strong>the</strong> oldest syn<strong>the</strong>tic vat dyes.<br />

It was syn<strong>the</strong>sized by Rene Bohn in 1901 [1]. Due to<br />

its high light fastness, flavanthrone is especially<br />

favoured as an artists’ color [2]. As shown in Figure<br />

1, we illustrate a recent installation in which flavanthrone<br />

has been infused into a wax sculpture to produce<br />

a luminescent orange glow.<br />

This yellow dye is <strong>of</strong> considerable interest in spectroscopy<br />

due to its high symmetry. As it has a point<br />

<strong>of</strong> inversion, no spectral lines are common to both<br />

<strong>the</strong> infrared and Raman spectrum.<br />

Most organic dyes exhibit an intense fluorescence<br />

emission. This makes <strong>the</strong>ir analysis by normal<br />

Raman spectroscopy difficult, as <strong>the</strong> fluorescence<br />

background can overwhelm <strong>the</strong> Raman bands. The<br />

use <strong>of</strong> surface-enhanced Raman spectroscopy<br />

(SERS), however, allows <strong>the</strong> analysis <strong>of</strong> organic<br />

dyes. The great advantage <strong>of</strong> this technique is <strong>the</strong><br />

combination <strong>of</strong> a highly increased Raman sensitivity<br />

with <strong>the</strong> fluorescence quenching that occurs on rough<br />

metallic surfaces [3, 4].<br />

In this work, we present <strong>the</strong> first SERS analysis <strong>of</strong><br />

one <strong>of</strong> <strong>the</strong> oldest syn<strong>the</strong>tic vat dyes, flavanthrone,<br />

241<br />

toge<strong>the</strong>r with <strong>the</strong> DFT analysis <strong>of</strong> its vibrational<br />

normal modes.<br />

Figure 1: Contemporary Art with Flavanthrone.<br />

Title: "In Front <strong>of</strong> <strong>the</strong> Plane, No. 8". Artist: Johannes<br />

Girardoni, 2007/<strong>2008</strong>, Installed at: Lukas Feichtner<br />

Gallery, Vienna<br />

II. EXPERIMENT<br />

Ag colloid was prepared following <strong>the</strong> method <strong>of</strong><br />

Lee and Meisel [5] by reduction <strong>of</strong> silver nitrate with<br />

sodium citrate. To fur<strong>the</strong>r concentrate <strong>the</strong> colloid for<br />

use, a volume <strong>of</strong> 1 ml <strong>of</strong> <strong>the</strong> original colloid was<br />

centrifuged at 8000 rpm for 5 minutes. Then, 900 μl<br />

<strong>of</strong> <strong>the</strong> supernatant was discarded and <strong>the</strong> settled<br />

portion was resuspended in 900 μl <strong>of</strong> original colloid.<br />

This process was repeated 5 times. SERS<br />

measurements were made simply by adding a 1 �l<br />

drop <strong>of</strong> <strong>the</strong> colloid and 2 �l drop <strong>of</strong> a 10 -4 M<br />

flavanthrone solution, followed by <strong>the</strong> addition <strong>of</strong> 1<br />

�l <strong>of</strong> a 0.5 M KNO3 solution. The potassium nitrate<br />

was used for aggregating <strong>the</strong> colloid so it would<br />

become active in SERS [6]. The SERS<br />

measurements were taken directly from <strong>the</strong> drop<br />

focusing by a microscope .<br />

III. RESULTS AND DISCUSSION<br />

The SERS spectra <strong>of</strong> flavanthrone obtained by using<br />

<strong>the</strong> excitation <strong>of</strong> 633 and 785 nm are displayed in<br />

Figure 2 in comparison with <strong>the</strong> FT-Raman spectrum.


Raman Intensity<br />

477<br />

509<br />

N<br />

O<br />

637<br />

O<br />

N<br />

1296<br />

1322<br />

1425<br />

* *<br />

*<br />

*<br />

* *<br />

817<br />

926<br />

1018<br />

1042<br />

1160<br />

400 600 800 1000 1200 1400 1600<br />

a<br />

1800<br />

Wavenumber (cm -1 )<br />

1560<br />

1601<br />

Figure 2: Comparison <strong>of</strong> <strong>the</strong> FT-Raman spectrum<br />

<strong>of</strong> flavanthrone (a) and SERS spectra <strong>of</strong> Flavanthrone<br />

(10 -4 M) with excitation at 785nm (b) and<br />

633nm (c) in <strong>the</strong> region 400-1800 cm -1 . Some forbidden<br />

lines are marked with *. Inset: <strong>the</strong> structure <strong>of</strong><br />

<strong>the</strong> flavanthrone molecule<br />

The DFT wavenumbers have been scaled by a factor<br />

<strong>of</strong> 0.98 (table 1). As can be seen, <strong>the</strong> fit <strong>of</strong> <strong>the</strong><br />

DFT wavenumbers and <strong>the</strong> FT-Raman wavenumbers<br />

is excellent, and we may <strong>the</strong>refore have great confidence<br />

in <strong>the</strong> assignments.<br />

The position <strong>of</strong> <strong>the</strong> bands and <strong>the</strong> normal modes<br />

assigned to <strong>the</strong> corresponding vibrations are compared<br />

in table 1. The most intense bands in <strong>the</strong> FT-<br />

Raman spectrum (Fig. 2a) are at 1183, 1384 and<br />

1596 cm -1 , which involve <strong>the</strong> �(C=C), �(C=N) and<br />

�(CH) vibrations. O<strong>the</strong>r important bands appear at<br />

1160, 1260, 1419, 1596 and 1619 cm -1 . The first two<br />

bands are assigned to �(CH) motions. The third one<br />

is due to a �(C=N) motion, and <strong>the</strong> last two bands<br />

correspond to �(C=C) vibrations. It is also shown<br />

that <strong>the</strong> majority <strong>of</strong> <strong>the</strong> most intense bands in <strong>the</strong> FT-<br />

Raman spectrum are totally symmetric modes (ag).<br />

Both <strong>of</strong> <strong>the</strong> SERS spectra <strong>of</strong> 633 (Fig. 2c) and<br />

785nm (Fig. 2b) excitation are shown in Fig. 2 in<br />

comparison with <strong>the</strong> FT-Raman spectrum (Fig. 2c).<br />

Several <strong>of</strong> <strong>the</strong> prominent bands, such as 1619, 1596,<br />

1419, 1384, 1260, 1183 and 1160 cm -1 are present in<br />

both (giving <strong>the</strong> FT-Raman measurements). All <strong>of</strong><br />

<strong>the</strong>m are assigned to ag modes, with <strong>the</strong> exception <strong>of</strong><br />

<strong>the</strong> last one. The relative intensities <strong>of</strong> many bands<br />

have been drastically changed. The intensities <strong>of</strong> <strong>the</strong><br />

bands at 1619, 1596, 1384 and 1183 cm -1 are<br />

distinctly decreased. On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> bands<br />

located at 1419, 1355, 1312, 1287, 1260 and 1160<br />

cm -1 are strongly enhanced on <strong>the</strong> Ag nanoparticles<br />

surface. Moreover, <strong>the</strong> bands at 1502, 1456, 1208,<br />

951, 926 and 902 cm -1 appear in <strong>the</strong> SERS spectra <strong>of</strong><br />

flavanthrone but are not shown in FT-Raman spectrum.<br />

Only <strong>the</strong> first one involves <strong>the</strong> totally symmetric<br />

mode (ag) all o<strong>the</strong>rs are due to <strong>the</strong> non-totally<br />

symmetric modes bu or bg.<br />

c<br />

b<br />

242<br />

Descrip<br />

tion<br />

�(CC)<br />

�(CC)<br />

�(CC)<br />

�(CN)<br />

�(CN)<br />

�(CN)<br />

�(CC)<br />

�(CC)<br />

�(CN)<br />

�(CH)<br />

�(CH)<br />

�(CH)<br />

�(CH)<br />

�(C=O)<br />

�(CH)<br />

�(CNC)<br />

Sym DFT<br />

ag<br />

ag<br />

ag<br />

bu<br />

ag<br />

ag<br />

ag<br />

ag<br />

bu<br />

ag<br />

bu<br />

ag<br />

bu<br />

bu<br />

bg<br />

bu<br />

1609<br />

1595<br />

1497<br />

1447<br />

1417<br />

1383<br />

1358<br />

1325<br />

1285<br />

1278<br />

1193<br />

1177<br />

1166<br />

940<br />

923<br />

902<br />

FTR SERS<br />

633<br />

1619<br />

1596<br />

1419<br />

1384<br />

1355<br />

1312<br />

1287<br />

1260<br />

1183<br />

1160<br />

1628<br />

1601<br />

1502<br />

1456<br />

1425<br />

1405<br />

1371<br />

1322<br />

1296<br />

1273<br />

1208<br />

1188<br />

1160<br />

951<br />

926<br />

902<br />

SERS<br />

785<br />

1621<br />

1598<br />

1498<br />

1455<br />

1418<br />

1401<br />

1368<br />

1318<br />

1285<br />

1270<br />

1207<br />

1186<br />

1159<br />

949<br />

927<br />

901<br />

Table 1: Wavenumbers (in cm -1 ) and assignments<br />

<strong>of</strong> some forbidden lines and <strong>the</strong> most intense lines<br />

in <strong>the</strong> FT-Raman and SERS spectra <strong>of</strong> flavanthrone.<br />

The DFT wavenumbers are scaled by a<br />

factor <strong>of</strong> 0.98<br />

ACKNOWLEDGMENTS<br />

We gratefully acknowledge Marco Leona for<br />

possibility to use <strong>the</strong> SERS instrument, and Manfred<br />

Schreiner, Wilfreid Vetter, Johannes Giradoni, for<br />

providing samples and photographs <strong>of</strong> <strong>the</strong> art work.<br />

We are indebted to <strong>the</strong> National Institute <strong>of</strong> Justice<br />

(Department <strong>of</strong> Justice Award #2006-DN-BX-K034)<br />

REFERENCES<br />

[1] E.R. Riegel, J.A. Kent, Riegel’s Handbook <strong>of</strong><br />

Industrial Chemistry, J.A. Kent, Ed, 10th Edition,<br />

Springer, 2003.<br />

[2] W. Herbst, K. Hunger, Industrial Organic Pigments.<br />

Production, Properties, Applications.<br />

Third, Completely Revised Edition. Wiley-VCH<br />

Verlag GmbH & Co. KGaA , Weinheim, Germany,<br />

2004.<br />

[3] Birke, R.L., Lombardi, J.R. Surface Enhanced<br />

Raman Scattering. In Spectroelectrochemistry:<br />

Theory and Practice, Gale RJ (ed). Plenum:New<br />

York, 1988; 263–348.<br />

[4] Moskovits M. Rev. Mod. Phys 1985; 57: 783.<br />

[5] Lee PC, Meisel D. J.Phys.Chem. 1982; 86: 3391.<br />

[6] M.V. Cañamares, J.V. Garcia-Ramos, J.D. Gomez-Varga,<br />

C. Domingo, S. Sanchez-Cortes,<br />

Langmuir 2005; 21: 8546.


Nonlinear resistance <strong>of</strong> 2D electrons in crossed electric and<br />

magnetic fields<br />

JingQiao Zhang and Sergey Vitkalov (Faculty Mentor)<br />

Physics Department<br />

City College <strong>of</strong> New York<br />

New York, United States<br />

Email: {phyzjqk,vitkalov}@sci.ccny.cuny.edu<br />

Abstract — Electric field induced, dramatic reduction <strong>of</strong><br />

longitudinal resistivity <strong>of</strong> two dimensional electrons placed in<br />

strong magnetic field is studied in broad range <strong>of</strong> magnetic<br />

fields B and temperatures T . The data are found to be in good<br />

agreement with <strong>the</strong>ory, considering <strong>the</strong> strong nonlinearity <strong>of</strong><br />

<strong>the</strong> resistivity as result <strong>of</strong> non-uniform spectral electron diffusion<br />

induced by <strong>the</strong> electric field. Comparison with <strong>the</strong> <strong>the</strong>ory<br />

gives inelastic scattering time τin <strong>of</strong> <strong>the</strong> 2D electrons. In temperature<br />

range T =2− 20K for overlapping Landau levels,<br />

<strong>the</strong> inelastic scattering rate 1/τin is found to be proportional<br />

to T 2 , indicating dominant contribution <strong>of</strong> <strong>the</strong> electron-electron<br />

interaction to <strong>the</strong> inelastic electron relaxation.<br />

I. INTRODUCTION<br />

In this report we study and discuss in details effect <strong>of</strong><br />

a small electric field Edc on longitudinal resistance <strong>of</strong><br />

two-dimensional electrons placed in a strong magnetic<br />

field. At <strong>the</strong> magnetic field <strong>the</strong> density <strong>of</strong> states <strong>of</strong> <strong>the</strong><br />

2D electrons is modulated due to Landau quantization <strong>of</strong><br />

<strong>the</strong> electron motion. Strong reduction <strong>of</strong> <strong>the</strong> longitudinal<br />

resistance <strong>of</strong> <strong>the</strong> 2D electrons in response to <strong>the</strong> dc electric<br />

field has been observed recently [1, 2]. The effect is<br />

attributed [2] to a spectral diffusion <strong>of</strong> <strong>the</strong> electrons induced<br />

by <strong>the</strong> electric field [3]. The diffusion is limited<br />

by an electron inelastic relaxation, which moves <strong>the</strong> system<br />

back to <strong>the</strong> <strong>the</strong>rmal equilibrium. A reasonable agreement<br />

between <strong>the</strong> <strong>the</strong>ory [3] and <strong>the</strong> experiment has been<br />

found[2]. The finding opens new possibilities to study inelastic<br />

electron processes and nonlinear electron kinetics<br />

<strong>of</strong> <strong>the</strong> low dimensional systems.<br />

II. EXPERIMENTAL SETUP<br />

Our samples are cleaved from a wafer <strong>of</strong> a high-mobility<br />

GaAs quantum well grown by molecular beam epitaxy<br />

on semi-insulating (001) GaAs substrates. The width <strong>of</strong><br />

<strong>the</strong> GaAs quantum well is 13 nm. Two AlAs/GaAs type-<br />

II superlattices grown on both sides <strong>of</strong> <strong>the</strong> well served<br />

as barriers, providing a high mobility <strong>of</strong> <strong>the</strong> 2D electrons<br />

inside <strong>the</strong> well at a high electron density. Two samples<br />

(N1 and N2) were studied with electron density n1 = 12.2<br />

×10 15 m −2 , n2=8.2 ×10 15 (m −2 ) and mobility μ1= 93<br />

m 2 /Vs, μ2=85 (m 2 /Vs) at T=2.7K.<br />

Measurements were carried out between T=0.3K and<br />

243<br />

Figure 1: Schematic view <strong>of</strong> experimental setup.<br />

T=20 K in a He-3 insert in a superconducting solenoid.<br />

Magnetic field up to 1 T was applied perpendicular to <strong>the</strong><br />

2D electron layers patterned in a form <strong>of</strong> d=50 μm wide<br />

Hall bars with a distance <strong>of</strong> 250 μm between potential<br />

contacts. The shape <strong>of</strong> <strong>the</strong> samples and <strong>the</strong> experimental<br />

setup are shown in Fig.1. The differential longitudinal<br />

resistance rxx is measured at a frequency <strong>of</strong> 77 Hz in <strong>the</strong><br />

linear regime.<br />

III. THEORY AND NUMERICAL SIMULA-<br />

TIONS<br />

The net longitudinal conductivity <strong>of</strong> <strong>the</strong> 2D electrons σnl<br />

is a sum <strong>of</strong> conductivities σ(ɛ) <strong>of</strong> <strong>the</strong> levels with energy ɛ<br />

over all possible energies, weighted with <strong>the</strong> first derivative<br />

<strong>of</strong> <strong>the</strong> distribution function ∂f/∂ɛ [?]:<br />

�<br />

σnl = σ(ɛ)(−∂f/∂ɛ)dɛ, (1)<br />

In <strong>the</strong> leading approximation for a classically strong<br />

magnetic field (ωcτtr ≫ 1) <strong>the</strong> longitudinal conductivity<br />

σ(ɛ) at an energy ɛ reads [3]:


σ(ɛ) =σD ˜ν 2 (ɛ), (2)<br />

where σD = e 2 ν0v 2 F /2ω2 cτtr is <strong>the</strong> dc Drude conductivity<br />

in strong magnetic field B, ˜ν(ɛ) =ν(ɛ)/ν0 is dimensionless<br />

density <strong>of</strong> states (DOS), τtr and ν0 = m/π¯h 2<br />

are transport scattering time and <strong>the</strong> density <strong>of</strong> states at<br />

zero magnetic field.<br />

As a result <strong>of</strong> <strong>the</strong> energy conservation, <strong>the</strong> diffusion<br />

motion <strong>of</strong> <strong>the</strong> electron in real space originates a spectral<br />

diffusion <strong>of</strong> <strong>the</strong> electron kinetic energy in <strong>the</strong> energy<br />

space. The spectral diffusion is described by <strong>the</strong> Fokker-<br />

Plank type equation [3]:<br />

− ∂f<br />

∂t + E2 σD dc<br />

ν0˜ν(ɛ) ∂ɛ<br />

� 2<br />

˜ν (ɛ)∂ɛf(ɛ) � = f(ɛ) − fT (ɛ)<br />

τin<br />

(3)<br />

The left side <strong>of</strong> <strong>the</strong> equation describes <strong>the</strong> spectral diffusion<br />

<strong>of</strong> a spherical part <strong>of</strong> <strong>the</strong> electron distribution function<br />

f induced by <strong>the</strong> electric field E in <strong>the</strong> presence <strong>of</strong><br />

<strong>the</strong> elastic impurity scattering.<br />

The numerical calculations are done in several steps.<br />

The first step is to find <strong>the</strong> density <strong>of</strong> electron states ν(ɛ),<br />

which are obtained using a Gaussian form <strong>of</strong> <strong>the</strong> DOS.<br />

In <strong>the</strong> second step we use <strong>the</strong> DOS to calculate numerically<br />

<strong>the</strong> distribution function f(ɛ) using <strong>the</strong> eq.3 in <strong>the</strong><br />

limit t ≫ τin. In <strong>the</strong> third step <strong>the</strong> normalized nonlinear<br />

conductivity σnl/σD is calculated using <strong>the</strong> eq.1 for<br />

different electric field. The results is compared with <strong>the</strong><br />

normalized resistance Rxx/R0. The inelastic scattering<br />

time τin is found from <strong>the</strong> best fit between <strong>the</strong> normalized<br />

resistance Rxx/R0 and <strong>the</strong> calculated normalized<br />

conductivity σnl/σD.<br />

IV. RESULTS AND DISCUSSION<br />

At small magnetic fields <strong>the</strong> separation between Landau<br />

levels ¯hωc is less than <strong>the</strong> effective width <strong>of</strong> <strong>the</strong> levels<br />

Γ = ¯h/τq. At low temperatures <strong>the</strong> width Γ is predominantly<br />

determined by <strong>the</strong> elastic impurity scattering<br />

<strong>of</strong> <strong>the</strong> 2D electrons. At <strong>the</strong> small magnetic field <strong>the</strong><br />

density <strong>of</strong> states ν(ɛ) is weakly oscillating with <strong>the</strong> energy<br />

ɛ, making <strong>the</strong> spectral diffusion to be also a weakly<br />

modulated function <strong>of</strong> <strong>the</strong> energy. We consider a regime<br />

<strong>of</strong> high temperatures: kT ≫ ¯hωc. In this regime <strong>the</strong><br />

electron-electron interaction is proposed to be <strong>the</strong> dominant<br />

mechanism <strong>of</strong> <strong>the</strong> inelastic electron relaxation at<br />

relatively low T[3]. The <strong>the</strong>ory predicts <strong>the</strong> temperature<br />

dependence <strong>of</strong> <strong>the</strong> inelastic scattering rate 1/τin to be<br />

proportional to square <strong>of</strong> <strong>the</strong> temperature T 2 .<br />

Figure 2 shows <strong>the</strong> comparision between <strong>the</strong> experimental<br />

results and numerical simulation <strong>of</strong> <strong>the</strong>ory. The<br />

results are in good agreement with <strong>the</strong> <strong>the</strong>ory.<br />

244<br />

Figure 2: (Color online), (a) Dependence <strong>of</strong> normalized<br />

longitudinal resistance Rxx/R0 on electric current at different<br />

temperatures as labeled. Solid lines are experimental<br />

curves. Symbols present result <strong>of</strong> numerical calculations<br />

<strong>of</strong> <strong>the</strong> resistance. (b) Dependencies <strong>of</strong> normalized<br />

density <strong>of</strong> states ˜ν(ɛ) =ν(ɛ)/ν0 and electron distribution<br />

function f on electron energy ɛ counted with<br />

respect to Fermi energy μ. The quantum scattering time<br />

τq=3.42 (ps) is obtained from comparison <strong>of</strong> <strong>the</strong> resistance<br />

at magnetic field B=0.2 (T) with eq.(1). The distribution<br />

function is solution <strong>of</strong> eq.(3) using <strong>the</strong> DOS,<br />

<strong>the</strong> temperature and <strong>the</strong> electric current shown in <strong>the</strong> figure.<br />

(c) Dependence <strong>of</strong> inelastic scattering time τin and<br />

quantum scattering time τq on temperature. B=0.2 (T);<br />

sample N2.<br />

ACKNOWLEDGMENTS<br />

This work was supported by National Science Foundation:<br />

DMR 0349049 and by Russian Fund for Basic Research,<br />

project No.08-02-01051<br />

REFERENCES<br />

[1] A. A. Bykov, Jing qiao Zhang, Sergey Vitkalov,<br />

A. K. Kalagin and A. K. Bakarov. Phys.Rev.B,<br />

72:245307, 2005.<br />

[2] Jing qiao Zhang, Sergey Vitkalov, A. A. Bykov,<br />

A. K. Kalagin and A. K. Bakarov. Phys.Rev.B,<br />

75:081305(R), 2007.<br />

[3] I. A. Dmitriev, M.G. Vavilov, I. L. Aleiner, A. D.<br />

MirlinandD.G.Polyakov.Phys.Rev.B, 71:115316,<br />

2005.


Understanding <strong>the</strong> Effects <strong>of</strong> Aerosols in Cloud Microphysics in<br />

Coastal Urban Environments<br />

Nathan Hosannah and Jorge E. Gonzalez (Faculty Mentor)<br />

Mechanical Engineering Department<br />

City College / The City University <strong>of</strong> New York Graduate Center<br />

New York, United States <strong>of</strong> America<br />

Email: nhosannah@gmail.com , gonzalez@me.ccny.cuny.edu<br />

Abstract — Several studies have found evidence <strong>of</strong><br />

warm-season rainfall increases over and downwind<br />

<strong>of</strong> cities. This induced precipitation (PCP) has been<br />

attributed mostly to induced updraft <strong>of</strong> warm air<br />

masses. Aerosols are abundant in urban environments<br />

and it has been hypo<strong>the</strong>sized that <strong>the</strong>y play a<br />

role in <strong>the</strong> water balance <strong>of</strong> cities. High concentrations<br />

<strong>of</strong> cloud condensation nuclei (CCN) may induce<br />

precipitation in humid urban environments.<br />

However precipitation may be reduced due to excess<br />

CCN’s or by large aerosols. The objective <strong>of</strong><br />

<strong>the</strong> present research is to improve our understanding<br />

<strong>of</strong> <strong>the</strong> role <strong>of</strong> aerosols in cloud processes <strong>of</strong><br />

complex coastal urban environments.<br />

I. INTRODUCTION<br />

There is increasing evidence that anthropogenic<br />

activities can significantly alter precipitation processes.<br />

Urbanization is an example <strong>of</strong> anthropogenic<br />

forcing. Recent studies [1] provide evidence that<br />

urban environments can modify or induce precipitation<br />

under a specific set <strong>of</strong> conditions. In <strong>the</strong> past 30<br />

years, several observational and climatological studies<br />

have found evidence <strong>of</strong> warm-season rainfall<br />

increases <strong>of</strong> 9%–17% over and downwind <strong>of</strong> major<br />

urban cities [2]. This work shows that historical<br />

records indicate an increase in <strong>the</strong> frequency <strong>of</strong> intense<br />

rain showers in recent decades in correlation<br />

with <strong>the</strong> population growth.<br />

A. AEROSOLS<br />

Aerosols may play an important role in precipitation<br />

increase. Our research is aimed to understand how<br />

aerosols may impact precipitation in NYC. Aerosols<br />

in <strong>the</strong> atmosphere have direct and indirect effects on<br />

<strong>the</strong> Earth's climate (Figure 1). The direct effect is<br />

related to <strong>the</strong>ir optical properties. Aerosols scatter<br />

and/or absorb solar and terrestrial radiation. The level<br />

<strong>of</strong> scattering and absorption depends on <strong>the</strong>ir physical<br />

and chemical characteristics. Consequently, aerosols<br />

act to modify <strong>the</strong> Earth's radiation budget and<br />

thus influence <strong>the</strong> warming/cooling <strong>of</strong> <strong>the</strong> planet.<br />

245<br />

B. MICROPHYSICS AND IMPACTS ON PCP<br />

Microphysical processes are cloud processes<br />

(growth, evaporation, etc.) which take place on <strong>the</strong><br />

scale <strong>of</strong> <strong>the</strong> individual aerosol (polluted particle) or<br />

precipitation particle as opposed to <strong>the</strong> scale <strong>of</strong> <strong>the</strong><br />

visual cloud. These processes include collision, coalescence,<br />

and droplet growth. Aerosols may ei<strong>the</strong>r<br />

enhance or reduce urban precipitation, depending on<br />

<strong>the</strong>ir composition and particle size. Cloud microphysics<br />

is affected by changes in <strong>the</strong> aerosol concentration.<br />

High concentrations yield more cloud water,<br />

but less rainwater in <strong>the</strong> atmosphere.<br />

Figure 1: Effects <strong>of</strong> aerosols on clouds, and<br />

microphysical processes (Image by Dr. S. N. Tripathi:<br />

Department <strong>of</strong> Civil Engineering Indian Institute<br />

<strong>of</strong> Technology)


II. HYPOTHESIS<br />

Based on analysis <strong>of</strong> background information, it can<br />

be inferred that aerosols may influence (enhance or<br />

decrease) precipitation in urban environments without<br />

added convection. To test this hypo<strong>the</strong>sis, data <strong>of</strong><br />

particle size distributions and precipitation levels<br />

from ground and satellite observations can be analyzed,<br />

and be ingested into refined mesoscale models<br />

with improved microphysics model (i.e. RAMS or<br />

WRF).<br />

III. MODELLING<br />

Mesoscale models can be used to interpret data<br />

from observations and to improve our understanding<br />

<strong>of</strong> microphysical processes. Through modeling,<br />

optimum conditions for precipitation to occur<br />

can be inferred with simulated or observed in-situ<br />

data particle size distribution and composition<br />

obtained from satellite and ground observations.<br />

The Regional Atmospheric Modeling System<br />

(RAMS), for example, contains a cloud microphysics<br />

component that will be used to simulate<br />

precipitation in a coastal urban environment [3].<br />

Idealized simulations utilize <strong>the</strong> extreme effects <strong>of</strong><br />

<strong>the</strong> microphysical parameterizations to expose <strong>the</strong><br />

impacts. The results suggest that cloud microphysics<br />

impact quasi-equilibrium temperature and moisture<br />

pr<strong>of</strong>iles substantially, but <strong>the</strong> relative humidity<br />

is only weakly affected [4]. Figure 2 shows how<br />

CCN size distribution may vary during clear skies<br />

(straight line), and cloudy skies (single dash curve,<br />

and bold solid curve).<br />

Figure 2: Variation <strong>of</strong> particle size distribution [5].<br />

Figure 3 shows results from a RAMS PCP simulation<br />

using <strong>the</strong> data from Figure 2. This work is expanding<br />

upon <strong>the</strong> work by Comarazamy et al. for <strong>the</strong> case <strong>of</strong><br />

New York City to investigate <strong>the</strong> role <strong>of</strong> aerosols in<br />

PCP in complex urban coastal environments.<br />

246<br />

Figure 3: Simulation <strong>of</strong> PCP in polluted (right) and<br />

non-polluted air (left). Rainwater mixing ratio in<br />

polluted air is less than a third <strong>of</strong> that in clear air [5].<br />

ACKNOWLEDGMENTS<br />

This work was financially supported by <strong>the</strong><br />

NASA COSI and NOAA CREST programs at<br />

<strong>the</strong> City College <strong>of</strong> New York.<br />

REFERENCES<br />

[1] J. Sheperd, H. Peirce, & A. Negri. Rainfall<br />

Modification by Major Urban Areas: Observations<br />

from Space Bourne Rain Radar on<br />

<strong>the</strong> TRMM Satellite. Journal <strong>of</strong> Applied Meteorology,<br />

Volume 41 pages 689–701, 2002.<br />

[2] F. Huff and S. Changnon. Precipitation modification<br />

by major urban areas. Bulletin<br />

American Meteoroligcal Society, Volume<br />

54, 1220 – 1232, 1973.<br />

[3] S. Saleeby and W. Cotton. A Large-Droplet<br />

Mode and Prognostic Number Concentration<br />

<strong>of</strong> Cloud Droplets in <strong>the</strong> Colorado State<br />

University Regional Atmospheric Modeling<br />

System (RAMS). Part I: Module Descriptions<br />

and Supercell Test Simulations. Journal<br />

<strong>of</strong> Applied Meteorology, Volume 43<br />

pages 182–195, 2004.<br />

[4] W. Grabowski. Cloud Microphysics and <strong>the</strong><br />

Tropical Climate: Cloud-Resolving Model<br />

Perspective. Journal <strong>of</strong> Climate, Volume 13<br />

pages 2306 – 2322, 1999.<br />

[5] D. Comarazamy, J. Gonzalez, C. Tepley, S.<br />

Raizada, and V. Pandya. The effects <strong>of</strong> Atmospheric<br />

particle Concentration on Cloud<br />

Microphysics over Arecibo. 2006.


Derivation <strong>of</strong> a Water Potential that Parameterizes Cooperative Effects<br />

Abstract - We propose a method for keeping track <strong>of</strong><br />

positions, distances and angles <strong>of</strong> water molecules<br />

during a simulation. Using FORTRAN 90, we have<br />

produced a program that keeps track <strong>of</strong> atoms by<br />

listing <strong>the</strong>m in terms <strong>of</strong> adjacency matrices. We read<br />

in coordinates from an xyz data file. Molecules are<br />

rotated using Euler angles and translated using a<br />

displacement vector. At <strong>the</strong> current stage <strong>the</strong><br />

program calculates distances <strong>of</strong> relative atomic<br />

positions, creates adjacency relationships <strong>of</strong><br />

molecules, outputs a graph <strong>of</strong> radial and angular<br />

distributions <strong>of</strong> molecules and an adjacency matrix<br />

for each atom. We also have a Monte Carlo<br />

algorithm. We find that it is necessary to devise new<br />

potentials for water that include parameterization <strong>of</strong><br />

cooperative effects. It has been observed that <strong>the</strong><br />

potential for hydrogen bonding varies with number <strong>of</strong><br />

donor/acceptor neighbors to a given central pair <strong>of</strong><br />

water molecules. The final program will allow us to<br />

simulate extended systems at nearly quantum<br />

mechanical accuracy.<br />

INTRODUCTION<br />

Philipa A. Njau and Michael E. Green<br />

City College <strong>of</strong> New York <strong>of</strong> <strong>the</strong><br />

City University <strong>of</strong> New York<br />

New York, USA<br />

phnjau@gmail.com<br />

Water models are necessary as water surrounds o<strong>the</strong>r<br />

molecules in biological chemistry. In heterogeneous<br />

solutions one cannot use <strong>the</strong> dielectric <strong>of</strong> water for<br />

simulations. There are about fifty potentials in <strong>the</strong><br />

literature for modeling water. Which is used depends<br />

on <strong>the</strong> kind <strong>of</strong> accuracy you are looking for and <strong>the</strong><br />

available computer resources. For studying<br />

molecular clefts and surfaces and <strong>the</strong>refore for <strong>the</strong><br />

most important applications in biophysics, <strong>the</strong>re is a<br />

need for a potential that parameterizes cooperative<br />

effects. Previous studies by Green and Znamenskiy<br />

in our group showed that hydrogen bond strength<br />

varies with <strong>the</strong> number <strong>of</strong> neighbors a given central<br />

pair <strong>of</strong> water molecules has [1,2,3,4] . Green and<br />

Znamenskiy have already started <strong>the</strong> work <strong>of</strong><br />

producing such a potential. The idea is that “<strong>the</strong><br />

potential <strong>of</strong> interaction <strong>of</strong> two water molecules is a<br />

function <strong>of</strong> <strong>the</strong> number <strong>of</strong> donor and acceptor<br />

neighbors” [1,2] . In this work, lengths and angles for<br />

36 different clusters were ‘scanned’ using density<br />

functional <strong>the</strong>ory for each possible arrangement <strong>of</strong><br />

donor acceptor hydrogen bonds to a given central<br />

hydrogen bonding pair. Four numbers (defined<br />

247<br />

below), K1K2K3K4, where K1 and K4 can be 1 or 2<br />

and K2 and K3 can be 0, 1 or 2 defined a cluster. As a<br />

result <strong>the</strong>re is a possibility <strong>of</strong> 2x3x3x2=36 distinct<br />

clusters (not sixty-four), due to symmetry (see figure<br />

1). K1= <strong>the</strong> number <strong>of</strong> times <strong>the</strong> central<br />

Figure 1: A 2222 cluster. Reprinted from<br />

Znamenskiy and Green 1<br />

donor acts as a donor; K1 can donate one more time.<br />

Hence it<br />

has values <strong>of</strong> 1 and 2, not 0, since it is already <strong>the</strong><br />

central donor, K2=number <strong>of</strong> times <strong>the</strong> central donor<br />

water molecule is an acceptor, for 0, 1, or 2 bonds.<br />

The central acceptor K3 can be a donor for 0, 1, or 2<br />

bonds and <strong>the</strong> acceptor K4 can accept 1 or 2 bonds but<br />

not 0. For example, an isolated pair has 1001 where<br />

<strong>the</strong> central donor donates once and <strong>the</strong> acceptor<br />

accepts once; <strong>the</strong> cluster with <strong>the</strong> maximum number<br />

<strong>of</strong> neighbors is 2222 (figure 1). A graph (figure 2) <strong>of</strong><br />

energy versus O-O distance was generated; let K1 - K2<br />

– K3 + K4 = KN; KN shows <strong>the</strong> different distanceenergy<br />

graphs that are found 1 . This turns out to be a<br />

key index, giving a first order correction (up to ±3kT<br />

from <strong>the</strong> TIP3P curve) to <strong>the</strong> potential. The task is to<br />

somehow include this parameter in a derivation <strong>of</strong> a<br />

potential; this way cooperative effects will be<br />

included. A second order effect would allow <strong>the</strong><br />

potential to vary with an angle: this has yet to be<br />

included. However, so far <strong>the</strong>re is enough data to<br />

know that we can ignore <strong>the</strong> angles, as <strong>the</strong><br />

computational time will be greatly extended, with<br />

little effect on <strong>the</strong> results, as <strong>the</strong> energy involved is<br />

less than <strong>the</strong>rmal energy. The graph shows that


increase in KN correlates with decrease in electron<br />

density in <strong>the</strong> bond. Moreover, <strong>the</strong> energy ratio <strong>of</strong><br />

variation with KN to <strong>the</strong>rmal energy, a total range <strong>of</strong><br />

6kT, suggests that it is worth pursuing this potential.<br />

We want a set <strong>of</strong> parameters that describes <strong>the</strong><br />

number <strong>of</strong> types <strong>of</strong> neighbors. From <strong>the</strong> energydistance<br />

curve we can describe a potential for a given<br />

KN .<br />

Figure 2: Graph <strong>of</strong> Energy <strong>of</strong> H-bond versus O-O<br />

distant for all 36 clusters. Reprinted from<br />

Znamenskiy and Green 1<br />

METHODS<br />

FORTRAN 90 has been used for <strong>the</strong> program. The<br />

Cartesian coordinates <strong>of</strong> atoms are read from a data<br />

file. Adjacency relationships are inferred by<br />

functions that perform standard calculations <strong>of</strong> inner<br />

product vector operations. Adjacency relationships<br />

are established and a table <strong>of</strong> adjacency is output. We<br />

also have a Monte Carlo algorithm set up for a<br />

simulation employing a dummy potential for <strong>the</strong> time<br />

being.<br />

RESULTS<br />

We have set up a framework for a simulation.<br />

The program now runs, and we are proceeding to<br />

insert <strong>the</strong> new potential for tests on standard<br />

systems, so far water. We will soon add o<strong>the</strong>r<br />

potentials that include proton transfer so that we<br />

can track, for example, pKa values.<br />

CONCLUSIONS<br />

We have created a program that keeps track <strong>of</strong><br />

adjacent atoms in a given simulation. Once we insert<br />

<strong>the</strong> potential we will incorporate it into <strong>the</strong> program.<br />

In <strong>the</strong> meantime we use a dummy potential to test our<br />

program. If <strong>the</strong>se routines work well in a standard<br />

248<br />

molecular dynamics or Monte Carlo simulation <strong>the</strong>n<br />

we will add proton transfer and salt bridge potentials,<br />

now under development, prior to applying it to some<br />

interesting system such as an ion channel or o<strong>the</strong>r<br />

large scale simulation. We are in <strong>the</strong> process <strong>of</strong><br />

getting this program to read data generated from<br />

studies done by Znamenskiy and Green. The data are<br />

<strong>the</strong> points on <strong>the</strong> graph (figure 2). This program<br />

should be able to read <strong>the</strong> data and, as its output,<br />

produce a simulation. If we have a good potential, our<br />

simulations should be able to reproduce<br />

experimentally observed <strong>the</strong>rmodynamic properties<br />

<strong>of</strong> water. If we calculate accurately <strong>the</strong> heat capacity<br />

<strong>of</strong> water, that will be a severe test, as it is a second<br />

derivative <strong>of</strong> <strong>the</strong> partition function.<br />

ACKNOWLEDGEMENTS<br />

We thank Vasiliy Znamenskiy for useful<br />

scientific discussions. This work has been<br />

supported in part by PSC-CUNY grants, and by<br />

a grant <strong>of</strong> computer time from <strong>the</strong><br />

Environmental Molecular Science Laboratory <strong>of</strong><br />

Pacific Northwest National Laboratory, a DOE<br />

facility.<br />

REFERENCES<br />

[1] Green M.E., Znamenskiy V. Quantum<br />

Calculations on Hydrogen Bonds in Certain<br />

Water Clusters Show Cooperative Effects. J<br />

Chem Theory Comp. (2006).3, 103-114<br />

[2] V.S. Znamenskiy and M.E. Green,<br />

Topological changes <strong>of</strong> hydrogen bonding <strong>of</strong><br />

water with acetic acid: AIM and NBO studies. J.<br />

Phys. Chem. A 108, 6543-6553 (2004)<br />

[3] J. Lu, J. Yin, and M.E. Green, A model for<br />

ion channel voltage gating with static S4<br />

segments, Ferroelectrics, 220, 249-271 (1999)<br />

[4] J. Lu and M.E. Green, Simulation <strong>of</strong> water in<br />

a small pore: Effect <strong>of</strong> Electric field and density<br />

II: Immobilized Molecules, J. Phys. Chem. B<br />

103, 2776-2780 (1999)


REAL-TIME VISUALIZATION AND<br />

QUANTIFICATION OF GLIAL PROGENITOR<br />

AND GLIOMA CHEMOTACTIC MIGRATION<br />

Richard A. Able Jr. 1 and Dr. Maribel Vazquez 2<br />

1 Department <strong>of</strong> Biochemistry, <strong>the</strong> Graduate Center CUNY 2<br />

2 Department <strong>of</strong> Biomedical Engineering, <strong>the</strong> City College <strong>of</strong> New York<br />

New York, NY, USA<br />

Rable00@ccny.cuny.edu<br />

Abstract -<br />

Previous research has indicated that progenitor<br />

precursor cells have distinct migration patterns<br />

in brain tissue, which may be induced mainly by<br />

specific cytokines. Our experiments analyzed and<br />

quantified <strong>the</strong> preferential migration <strong>of</strong> tumor<br />

cells towards four specific growth factors<br />

concentrations. Using transwell assays, <strong>the</strong><br />

chemotactic migration <strong>of</strong> <strong>the</strong> Ntva-mouse<br />

derived glial progenitor cells was compared to<br />

<strong>the</strong> migration <strong>of</strong> cells derived from a Human<br />

Glioblastoma multiforme cell line (U-87MG).<br />

This investigation illustrated that glial<br />

progenitor motility approached zero at higher<br />

ligand concentrations, while gliomas remained<br />

motile. In order to fur<strong>the</strong>r investigate <strong>the</strong><br />

directionality <strong>of</strong> migration, a new set <strong>of</strong><br />

experiments made use <strong>of</strong> a micr<strong>of</strong>luidic system in<br />

which to monitor chemotaxis <strong>of</strong> individual cells.<br />

Introduction<br />

Brain tumors can be highly aggressive and<br />

debilitating for many patients and lead to an<br />

untimely death in just a few months.<br />

Unfortunately, due to <strong>the</strong> location <strong>of</strong> many brain<br />

tumors, <strong>the</strong>rapy with ionizing radiation,<br />

chemo<strong>the</strong>rapeutic agents and surgery has limited<br />

rewards (1). Fur<strong>the</strong>rmore, <strong>the</strong> ability <strong>of</strong> cancer<br />

cells to undergo metastasis- invasion and<br />

migration <strong>of</strong> new tissue- makes <strong>the</strong>m even more<br />

illusive and difficult to treat.<br />

Here, glial progenitor cells (PDGF, Ras,<br />

LacZ), have been engineered to perform special<br />

functions. RCAS-PDGF infected cells are driven<br />

by a PDGF autocrine loop and can not detect any<br />

exogenous PDGF gradients. The RCAS-Ras<br />

infected cells’ downstream Ras pathways are<br />

activated independently <strong>of</strong> its receptors. And, <strong>the</strong><br />

RCAS-LacZ cells, which express markers for <strong>the</strong><br />

transgenic �-galactosidase gene, do not actually<br />

249<br />

have any experimentally-induced mutations to its<br />

growth factor receptors, and is used as <strong>the</strong><br />

control cell line. The U87 Malignant Glioma cell<br />

line is a human glioblastoma; astrocytoma type<br />

<strong>of</strong> epi<strong>the</strong>lial morphology. This grade III tumor<br />

was purchased from ATCC and was used for<br />

comparative purposes.<br />

Methods<br />

Boyden Migration Assay<br />

The migration <strong>of</strong> cells in response to signaling<br />

from different concentrations <strong>of</strong> growth factors<br />

was analyzed via transwell assay as described<br />

previously (2). Growth factors were added to <strong>the</strong><br />

lower well <strong>of</strong> <strong>the</strong> 24-well Boyden chamber<br />

Falcon Companion TC Plate (BD Bioscience,<br />

MA). An incubation time <strong>of</strong> 12 hours at 37 0 C<br />

was used for every assay. The nuclei <strong>of</strong> cells that<br />

migrated to <strong>the</strong> underside <strong>of</strong> <strong>the</strong> filter were<br />

counted using an inverted light microscope under<br />

20x magnification (Nikon TE300, Morrell<br />

Instruments, NY).<br />

Channels used for this study were 1.5cm in<br />

length, and approximately 100�m by 100�m in<br />

cross-section. Channels were filled with a<br />

solution <strong>of</strong> 2�g/mL <strong>of</strong> Matrigel® (BD<br />

Bioscience, MA) containing a desired cytokine<br />

concentration. A 10�L volume <strong>of</strong> Matrigel was<br />

pipetted into <strong>the</strong> loading reservoir, termed <strong>the</strong><br />

Northpole, until <strong>the</strong> entire volume was seen<br />

exiting <strong>the</strong> opposite reservoir, designated <strong>the</strong><br />

Southpole. Systems were <strong>the</strong>n incubated at 37 o C<br />

for one hour to allow polymerization. Afterwards,<br />

a rectangular section <strong>of</strong> PDMS was cut out<br />

around <strong>the</strong> Southpole to generate a larger<br />

reservoir. A cell solution having a density <strong>of</strong><br />

10 6 /mL was pipetted into this reservoir.


Immediately, <strong>the</strong> system was imaged using an<br />

inverted microscope to verify <strong>the</strong> locations <strong>of</strong><br />

viable cells at <strong>the</strong> reservoir-�Lane interface and<br />

<strong>the</strong>n incubated at 37 o C.<br />

Results<br />

Boyden experiments show all three progenitor<br />

cell lines to have single digit migrating cells at<br />

<strong>the</strong> 12hr time point accompanied by a<br />

progressive increase in <strong>the</strong> number <strong>of</strong> migrating<br />

cells and an increase in distance traveled, as time<br />

increases. GPC-Lacz, after a 36hrs incubation<br />

time had ~59 cells invade <strong>the</strong> matrix. Of <strong>the</strong> 59<br />

cells to penetrate <strong>the</strong> matrix 55/59 migrated a<br />

distance found between <strong>the</strong> interface and 200�m.<br />

The o<strong>the</strong>r 4/59 was found between 200�m and<br />

400�m having a maximum extension at ~256�m.<br />

GPC-PDGF, after a 36hrs incubation time had<br />

~53 cells invade <strong>the</strong> matrix. Of <strong>the</strong> 53 cells to<br />

penetrate <strong>the</strong> matrix 47/53 migrated a distance<br />

found between <strong>the</strong> interface and 200�m. The<br />

o<strong>the</strong>r 5/53 was found between 200�m and<br />

400�m having a maximum extension at ~348�m.<br />

GPC-Ras, after a 36hrs incubation time had ~50<br />

cells invade <strong>the</strong> matrix. Of <strong>the</strong> 50 cells to<br />

penetrate <strong>the</strong> matrix 42/50 migrated a distance<br />

found between <strong>the</strong> interface and 200�m. The<br />

o<strong>the</strong>r 8/50 was found between 200�m and<br />

400�m having a maximum extension at ~230�m.<br />

Fur<strong>the</strong>r, cells were exposed to 10pM<br />

TGF-�. GPC-PDGF had ~125 cells invade <strong>the</strong><br />

matrix in response to 10pM TGF-� after 45hrs.<br />

100/125 was in <strong>the</strong>


Surface Enhanced Raman Spectroscopy <strong>of</strong> Pyridine on<br />

CdSe/ZnBeSe Quantum Dots Grown by Molecular Beam Epitaxy<br />

Abstract — Using Surface Enhanced Raman<br />

Spectroscopy (SERS), we observed large Raman<br />

5 6<br />

enhancements (10 -10 ) for pyridine molecules<br />

adsorbed on a II-VI semiconductor CdSe/ZnBeSe<br />

sample <strong>of</strong> uncapped self-assembled quantum dots<br />

produced by molecular beam epitaxy. When a<br />

monolayer <strong>of</strong> pyridine is adsorbed on <strong>the</strong>se structures,<br />

excitation at 488 nm produces intense Raman<br />

spectra a very large enhancement <strong>of</strong> <strong>the</strong> A 1 and B2<br />

modes. Examination <strong>of</strong> individual Raman frequencies<br />

provides information as to <strong>the</strong> mode <strong>of</strong> attachment<br />

<strong>of</strong> <strong>the</strong> molecules to <strong>the</strong> CdSe surface. These<br />

results demonstrate that <strong>the</strong> CdSe quantum dot<br />

system provides a high degree <strong>of</strong> molecular specificity<br />

and large Raman enhancement factor, enabling<br />

sensitive detection and identification <strong>of</strong> SERS<br />

active molecules.<br />

I. INTRODUCTION<br />

Surface enhanced Raman scattering (SERS) has been<br />

found to be <strong>of</strong> value in surface and interface Raman<br />

studies in metals. The SERS effect is characterized<br />

by an enormous increase in <strong>the</strong> Raman intensity by<br />

many orders <strong>of</strong> magnitude for species adsorbed on<br />

rough metal surfaces, and <strong>the</strong> technique displays a<br />

number <strong>of</strong> important advantages: sensitivity, selectivity,<br />

non-destructive detection, and feasibility for insitu<br />

studies and on metal substrates. The SERS effect<br />

has been shown to arise from a product <strong>of</strong> surface<br />

plasmon and charge transfer resonances 1 .<br />

In recent work we have observed SERS from III-V<br />

(InAs/GaAs) semiconductor quantum dots 2 . This<br />

suggests a technique for sensitive, high-resolution<br />

study <strong>of</strong> <strong>the</strong> interaction <strong>of</strong> adsorbate molecules with<br />

quantum dot surfaces. Fur<strong>the</strong>rmore, <strong>the</strong> technique <strong>of</strong><br />

molecular beam epitaxy (MBE) is useful in constructing<br />

and testing <strong>of</strong> solid state surface, made<br />

under controlled conditions. In this work, we report<br />

on <strong>the</strong> observation <strong>of</strong> surface enhancement <strong>of</strong> a Raman<br />

signal from pyridine molecules adsorbed on a<br />

self-assembled CdSe quantum dot array produced by<br />

molecular beam epitaxy (MBE). The high degree <strong>of</strong><br />

spectral selectivity, coupled with <strong>the</strong> high sensitivity<br />

Richard Livingstone and John Lombardi<br />

City College <strong>of</strong> New York<br />

New York, New York, 10031 USA<br />

Email: livingstonerm@aol.com<br />

251<br />

afforded by <strong>the</strong> ra<strong>the</strong>r large enhancement factors,<br />

makes such systems good candidates for possible use<br />

in detecting and identifying SERS active molecules.<br />

II. EXPERIMENTAL<br />

The quantum dot (QDs) assemblies were grown by<br />

MBE on GaAs (001) substrates in an ultrahigh<br />

vacuum (UHV) Riber system. This was followed<br />

by <strong>the</strong> growth <strong>of</strong> a ZnBeSe epilayer on which <strong>the</strong><br />

CdSe quantum dots were deposited. The entire<br />

growth process was monitored in situ by reflection<br />

high-energy electron diffraction (RHEED).<br />

Samples were removed from <strong>the</strong> MBE growth<br />

chamber, cut into smaller rectangles<br />

(approximately 3mm) and stored in liquid nitrogen<br />

to prevent possible ripening until used. One <strong>of</strong> <strong>the</strong><br />

cut pieces <strong>of</strong> <strong>the</strong> quantum dot samples would <strong>the</strong>n<br />

be removed allowed to attain room temperature<br />

conditions, and <strong>the</strong>n immersed in a small portion<br />

(10ml) <strong>of</strong> <strong>the</strong> liquid organic substance for 20<br />

minutes. The sample was sonicated in commercial<br />

acetonitrile solution for 5 minutes to remove<br />

excess liquid and mounted near <strong>the</strong> entrance slit <strong>of</strong><br />

a SPEX (0.75 meter) model 1401 double<br />

monochromator and slit width 200 nm. Light from<br />

a Spectra Physics Model 2560 argon ion laser was<br />

focused on to <strong>the</strong> sample. All spectra reported<br />

below were taken at 488 nm. Typical power levels<br />

at <strong>the</strong> sample were measured to be approximately<br />

30 mW.<br />

III. RESULTS<br />

Figure 1 shows <strong>the</strong> SERS spectrum <strong>of</strong> <strong>the</strong> pyridine<br />

molecules adsorbed on <strong>the</strong> CdSe quantum dots, toge<strong>the</strong>r<br />

with <strong>the</strong> Raman spectrum <strong>of</strong> <strong>the</strong> quantum dots<br />

sample before <strong>the</strong> deposition <strong>of</strong> <strong>the</strong> molecule and <strong>the</strong><br />

Raman spectrum <strong>of</strong> <strong>the</strong> pyridine solution. Numerous<br />

lines characteristic <strong>of</strong> pyridine appear, as can be seen<br />

by comparison with <strong>the</strong> normal Raman spectrum <strong>of</strong><br />

pyridine liquid. These are listed in <strong>the</strong> table, along<br />

with <strong>the</strong> lines observed on Ag. We also present a<br />

brief summary <strong>of</strong> <strong>the</strong> assignment <strong>of</strong> <strong>the</strong> normal modes<br />

3 <strong>of</strong> pyridine, including <strong>the</strong>ir Wilson number and<br />

symmetry in C2v. On <strong>the</strong> quantum dot surface most <strong>of</strong>


<strong>the</strong> lines correspond to lines previously observed<br />

ei<strong>the</strong>r in <strong>the</strong> liquid or on a Ag surface. Note that in<br />

addition to <strong>the</strong> expected enhancement <strong>of</strong> several a1<br />

modes, <strong>the</strong> bands at 665 (b2), 728 (b1) and 946 (b1)<br />

cm -1 are also considerably enhanced relative to <strong>the</strong><br />

liquid spectrum. The line at 728 cm -1 suffers from<br />

interference from <strong>the</strong> bare quantum dot line at 742<br />

cm -1 , and should <strong>the</strong>refore be considered questionable.<br />

The same is true for <strong>the</strong> a1 line at 897 cm -1 ,<br />

while <strong>the</strong> line at 1221 cm -1 is sufficiently distinct<br />

from <strong>the</strong> QD line at 1203 cm -1 to be acceptable. The<br />

observation <strong>of</strong> relatively intense non-totally symmetric<br />

modes is usually diagnostic <strong>of</strong> a charge-transfer<br />

contribution to <strong>the</strong> observed enhancement, most<br />

probably from <strong>the</strong> quantum dots to <strong>the</strong> molecule.<br />

Intensity (Arbitrary Units)<br />

605<br />

Raman <strong>of</strong> CdSe Quantum Dots<br />

SERS <strong>of</strong> Pyridine adsorbed on CdSe Quantum Dots<br />

Raman <strong>of</strong> Pyridine<br />

653<br />

665<br />

728<br />

708 753 812<br />

742<br />

809 897<br />

993<br />

923<br />

946<br />

887<br />

945<br />

887<br />

1004<br />

1236<br />

1069<br />

1150<br />

1192<br />

600 700 800 900 1000 1100 1200<br />

Wavenumbers (cm-1)<br />

1202<br />

1203<br />

1219<br />

1221<br />

Fig 1: Normal Raman and SERS spectrum <strong>of</strong><br />

pyridine on Ag and CdSe quantum dots<br />

excited at 488 nm.<br />

Wilson #<br />

C2v<br />

Symm<br />

SERS<br />

Pyridine<br />

on Ag<br />

SERS<br />

Pyridine on<br />

CdSe QDs<br />

6a a1 623(mw) 616(w)<br />

6b b2 649 (w) 665(m)<br />

698 715(w)<br />

4, QD b1 753(w) 728(m)<br />

809(m)<br />

10a, QD a1 877(vw) 897 (s)<br />

923(w)<br />

17b b1 940 (w) 946 (w)<br />

1 a1 1005(vs) 1004(vs)<br />

12 a1 1034(s)<br />

18a,18b a1,b2 1063(w)<br />

15 b2 1148(vw)<br />

QD 1202<br />

9a a1 1214(m) 1221(s)<br />

252<br />

Table showing Pyridine lines as observed in Normal<br />

Raman, SERS on a Ag surface and on CdSe/ZnBeSe<br />

self-assembled quantum dots.<br />

IV. CONCLUSION<br />

The observed Raman spectrum shows several<br />

differences from <strong>the</strong> normal Raman spectrum, and<br />

is closer in some ways to <strong>the</strong> SERS observed on<br />

Ag nanoparticles. Among o<strong>the</strong>r things, <strong>the</strong><br />

prominence <strong>of</strong> non-totally symmetric lines<br />

indicates <strong>the</strong> importance <strong>of</strong> charge-transfer<br />

contributions to <strong>the</strong> overall enhancement. These<br />

enhancements are far from <strong>the</strong> region <strong>of</strong> surface<br />

plasmon resonance for CdSe and so <strong>the</strong> magnitude<br />

<strong>of</strong> enhancement must be attributed to some o<strong>the</strong>r<br />

resonance. We suggest that this is likely charge<br />

transfer resonance between <strong>the</strong> molecule and <strong>the</strong><br />

quantum dot.<br />

ACKNOWLEDGEMENTS<br />

This work was supported by <strong>the</strong> MARC/RISE<br />

Fellowship, National Science Foundation under<br />

Cooperative Agreement No. RII-9353488, grant<br />

No. CHE-0091362, CHE-0345987. We are also<br />

indebted to <strong>the</strong> National Institute <strong>of</strong> Justice<br />

(Department <strong>of</strong> Justice award No. 2006-DN-BX-<br />

K034)<br />

REFERENCES<br />

[1] Lombardi, J. R.; Birke, R.L.; Lu, T.; Xu, J., J.<br />

Chem. Phys.. 1986, 84, 4357.<br />

[2] Quagliano, L., J.Am.Chem.Soc., 2004, 126, 7393-<br />

7398..<br />

[3] Lombardi, J.R.; Birke, R.L Foucault R; Vivoni A,<br />

J.Phys.Chem. B, 2003, 107, 5547-5557.


Chapter 3<br />

Environmental Science and Energy


Using RFID Technology in Pedestrian Navigation<br />

for Information Transmission and<br />

Data Communication Recording<br />

Qing Fu and Guen<strong>the</strong>r Retscher (Faculty Mentor)<br />

Institute <strong>of</strong> Geodesy and Geophysics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: fu@mail.tuwien.ac.at<br />

or gretsch@pop.tuwien.ac.at<br />

Abstract — The development <strong>of</strong> pedestrian<br />

navigation and guidance services are<br />

challenging research topics and have been<br />

investigated by several researchers worldwide.<br />

Most <strong>of</strong> <strong>the</strong> developed systems <strong>the</strong>reby rely on<br />

<strong>the</strong> use <strong>of</strong> satellite positioning (GNSS),<br />

sometimes also in combination with o<strong>the</strong>r<br />

sensors and positioning methods. In our<br />

approach in <strong>the</strong> project UCPNAVI (Ubiquitous<br />

Carthograpy for Pedestrian Navigation) active<br />

RFID (Radio Frequency Identification) in<br />

combination with GNSS and INS (Inertial<br />

Navigation Systems) are employed for<br />

positioning. RFID can be employed in areas<br />

where no satellite positioning is possible due to<br />

obstructions, e.g. in urban canyons and indoor<br />

environments. This paper presents our approach<br />

for positioning <strong>of</strong> a pedestrian by using RFID.<br />

I. INTRODUCTION<br />

Radio-frequency identification, or RFID for<br />

short, is a technology for data transmission via<br />

radio waves without line <strong>of</strong> sight contact.<br />

Nowadays, <strong>the</strong> technology <strong>of</strong> RFID is also used<br />

for positioning, because <strong>the</strong> location estimation<br />

can be based on RSSI (short for received signal<br />

strength indication) which is a measurement <strong>of</strong><br />

<strong>the</strong> power present in a received radio signal. The<br />

receiver can compute its position using various<br />

methods based on RSSI (e.g. a range-based<br />

positioning system based on trilateration,<br />

modeling <strong>of</strong> indoor positioning systems based on<br />

location fingerprinting). Totally, three different<br />

methods have been employed, i.e., cell-based<br />

positioning, trilateration using ranges to <strong>the</strong><br />

surrounding RFID tags deduced from received<br />

signal strength measurements and RFID location<br />

fingerprinting. In our experiment for <strong>the</strong><br />

positioning <strong>of</strong> a pedestrian from <strong>the</strong> underground<br />

255<br />

station “Karlsplatz” to one <strong>of</strong> <strong>the</strong> <strong>of</strong>fices in <strong>the</strong><br />

building <strong>of</strong> <strong>the</strong> Vienna University <strong>of</strong> Technology<br />

on Gusshausstrasse 27-29 <strong>the</strong>se three methods<br />

have been combined. The results showed that<br />

<strong>the</strong>se approaches using RFID are suitable to<br />

navigate <strong>the</strong> user (see [1]). In <strong>the</strong> following <strong>the</strong><br />

principles and methods <strong>of</strong> <strong>the</strong> positioning using<br />

active RFID would be described briefly.<br />

II. POSITIONING USING ACTIVE RFID<br />

Thanks to <strong>the</strong> RFID technology, data can be<br />

transmitted from RFID tags to a reader via<br />

radio waves without line <strong>of</strong> sight contact. The<br />

transmitted data might include <strong>the</strong> ID and <strong>the</strong><br />

information <strong>of</strong> <strong>the</strong> position <strong>of</strong> <strong>the</strong> RFID tags. In<br />

<strong>the</strong> presented work only one reader and a large<br />

number <strong>of</strong> active tags are used. The moving<br />

object is equipped with a RFID reader and <strong>the</strong><br />

unmoving objects in <strong>the</strong> surrounding<br />

environment are equipped with RFID tags. The<br />

user is equipped with a portable RFID reader<br />

(e.g. a reader in form <strong>of</strong> a PCMCIA card that<br />

can be plugged into a laptop). If <strong>the</strong> user moves<br />

inside <strong>the</strong> surrounding <strong>of</strong> <strong>the</strong> RFID tags, <strong>the</strong><br />

tags’ ID, signal strength RSSI, known location<br />

and time can be obtained. The position is a<br />

function f as shown in (1):<br />

position = f (ID, RSSI, known location) (1)<br />

at a certain time t. Then <strong>the</strong> position <strong>of</strong> <strong>the</strong> user<br />

can be calculated by using various methods<br />

based on RSSI and known location <strong>of</strong> <strong>the</strong> tag<br />

with a certain ID. The measured data is<br />

recorded in a database. Three methods, i.e.,<br />

cell-based positioning, trilateration and RFID<br />

location fingerprinting, are employed<br />

depending on <strong>the</strong> current enviroment <strong>of</strong> <strong>the</strong> user<br />

(see Figure 1).


Figure 1: Information transmission and data<br />

communication recording for <strong>the</strong> active RFID<br />

positioning system<br />

1. CELL-BASED POSITIONING<br />

Cell-based positioning is an algorithm to<br />

determine <strong>the</strong> location <strong>of</strong> <strong>the</strong> user in a cell<br />

around <strong>the</strong> RFID tag with a size defined by <strong>the</strong><br />

maximum range <strong>of</strong> <strong>the</strong> RFID signals. The<br />

achievable positioning accuracies depend on<br />

<strong>the</strong> size <strong>of</strong> <strong>the</strong> cell, i.e., up to 20 m using our<br />

long range RFID equipment. Therefore this<br />

method is only well suited for areas where<br />

accuracy is not that import, such as urban<br />

outdoor environment.<br />

2. TRILATERATION<br />

Trilateration is a method <strong>of</strong> determining <strong>the</strong><br />

absolute positions <strong>of</strong> moving objects (e.g. a<br />

user with a portable RFID reader) using <strong>the</strong><br />

know location <strong>of</strong> two or more unmoving<br />

objects (i.e., tags), and <strong>the</strong> range between <strong>the</strong><br />

moving object and <strong>the</strong> unmoving objects. To<br />

accurately and uniquely determine <strong>the</strong><br />

absolute position <strong>of</strong> a RFID reader on a 2D<br />

plane two ranges are necessary. Trilateration<br />

achieves a positioning accuracy <strong>of</strong> � 2-3 m<br />

and can be employed in <strong>the</strong> transition zone<br />

between outdoor to indoor (e.g. at <strong>the</strong> entrance<br />

<strong>of</strong> a building).<br />

3. RFID LOCATION FINGERPRINTING<br />

The principle <strong>of</strong> operation <strong>of</strong> RFID location<br />

fingerprinting is similar to that used in WiFi<br />

networks. Here <strong>the</strong> signal strength is used<br />

directly to obtain <strong>the</strong> location <strong>of</strong> <strong>the</strong> user.<br />

Positioning accuracies on <strong>the</strong> 1 to 2 m level<br />

can be achieved. Fingerprinting can <strong>the</strong>refore<br />

be used in indoor environments as it provides<br />

<strong>the</strong> highest level <strong>of</strong> positioning accuracy.<br />

256<br />

4. INTELLIGENT INTEGRATION<br />

The most challenging task is <strong>the</strong> selection <strong>of</strong> one<br />

<strong>of</strong> <strong>the</strong> positioning methods in <strong>the</strong> current<br />

environment <strong>of</strong> <strong>the</strong> user by using an intelligent<br />

s<strong>of</strong>tware package. For this purpose, <strong>the</strong> use <strong>of</strong><br />

knowledge-based systems will be investigated.<br />

III. COMPARISON OF THE METHODS<br />

As mentioned in <strong>the</strong> introduction all <strong>the</strong> three<br />

methods <strong>of</strong> positioning were used for <strong>the</strong><br />

positioning in a complex urban and indoor<br />

environment. Table 1 shows a comparison <strong>of</strong><br />

<strong>the</strong> methods with <strong>the</strong>ir suitable enviroment<br />

and achievable positioning accuracy.<br />

Method Environment Accuracy<br />

Cell-based<br />

positioning<br />

Trilateration<br />

Location<br />

fingerprinting<br />

Outdoor � 20 m<br />

Transition<br />

zone<br />

� 2-3 m<br />

Indoor � 1-2 m<br />

Table 1: Comparison <strong>of</strong> <strong>the</strong> three methods.<br />

IV. CONCLUSION<br />

RFID tags transmit useful information such as<br />

ID, position and signal strength to <strong>the</strong> RFID<br />

reader. This information is recorded in a<br />

database and can be used for determination <strong>of</strong><br />

<strong>the</strong> position <strong>of</strong> <strong>the</strong> RFID reader. Based on <strong>the</strong><br />

principle <strong>of</strong> RSSI measurement three approaches<br />

could be employed for <strong>the</strong> positioning, i.e., cellbased<br />

positioning, trilateration and location<br />

fingerprinting. The test results showed that with<br />

a combination <strong>of</strong> <strong>the</strong>se three methods and a<br />

knowledge-based selection it is possible to<br />

navigate a pedestrian in a complex urban and<br />

indoor environment.<br />

ACKNOWLEDGMENTS<br />

This research is supported by <strong>the</strong> research<br />

project P19210-N15 “UCPNAVI” founded by<br />

<strong>the</strong> Austrian Science Fund (FWF).<br />

REFERENCES<br />

[1] Fu Q., <strong>2008</strong>. Active RFID for Positioning<br />

Using Trilateration and Location<br />

Fingerprinting Based on RSSI, in: Papers<br />

presented at <strong>the</strong> ION GNSS <strong>Conference</strong>,<br />

September 16-19, <strong>2008</strong>, Savannah, Georgia,<br />

USA, CD-Rom <strong>Proceedings</strong>, 14 pgs.


Adaptive Boundary Element Methods Based on<br />

Accurate A Posteriori Error Estimation<br />

Samuel Ferraz-Leite and Dirk Praetorius (Faculty Mentor)<br />

Institute for Analysis and Scientific Computing<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {Samuel.Ferraz-Leite,Dirk.Praetorius}@tuwien.ac.at<br />

Abstract — The boundary element method is one strategy to<br />

solve partial differential equations <strong>of</strong> elliptic type. As model<br />

problem, we consider <strong>the</strong> computation <strong>of</strong> <strong>the</strong> charge density φ<br />

and <strong>the</strong> capacitance C <strong>of</strong> a thin electrified plate. We introduce<br />

an intelligent algorithm based on certain error estimators. The<br />

mesh-refinement is steered automatically in <strong>the</strong> sense that <strong>the</strong><br />

mesh is locally refined, where <strong>the</strong> error appears to be large. Numerical<br />

experiments show that <strong>the</strong> new method reveals <strong>the</strong> optimal<br />

order <strong>of</strong> convergence and is <strong>the</strong>refore significantly faster<br />

than a standard uniform approach. All ma<strong>the</strong>matical results are<br />

valid in a quite general framework and thus apply to a large<br />

problem class, including, e.g., <strong>the</strong> Laplace problem, <strong>the</strong> Stokes<br />

system, and <strong>the</strong> Lamé equation.<br />

I. INTRODUCTION<br />

Computer aided simulations – and <strong>the</strong>refore usually <strong>the</strong><br />

solution <strong>of</strong> partial differential equations (PDEs) – have<br />

established <strong>the</strong>mselves as cost and time efficient methods<br />

in research and development.<br />

Certain PDEs <strong>of</strong> elliptic type can be solved by <strong>the</strong> socalled<br />

boundary element method (BEM). Compared to<br />

<strong>the</strong> more popular finite element method (FEM), <strong>the</strong> essential<br />

disadvantage <strong>of</strong> BEM is that it leads to large dense<br />

matrices. Ma<strong>the</strong>matical strategies to overcome this disadvantage,<br />

e.g. fast multipole methods, have been developed<br />

and are usually employed by engineers nowadays.<br />

Besides <strong>the</strong> fact that one only has to discretize <strong>the</strong> boundary<br />

<strong>of</strong> <strong>the</strong> simulation domain, one key advantage <strong>of</strong> BEM<br />

is its generically higher order <strong>of</strong> convergence when compared<br />

to FEM. With respect to <strong>the</strong> mesh-size h <strong>of</strong> <strong>the</strong><br />

discretization Th <strong>of</strong> <strong>the</strong> simulation domain optimal convergence<br />

for lowest-order approximations reads:<br />

– FEM: O(h) as h → 0,<br />

– BEM: O(h3/2 ) as h → 0.<br />

However, <strong>the</strong>se convergence rates are only observed<br />

if <strong>the</strong> unknown solution is sufficiently smooth. This is<br />

generically not <strong>the</strong> case in practice. One possibility to recover<br />

<strong>the</strong> optimal order <strong>of</strong> convergence is to estimate <strong>the</strong><br />

simulation error, and to refine <strong>the</strong> spatial discretization<br />

only locally, where <strong>the</strong> error appears to be large.<br />

In <strong>the</strong> context <strong>of</strong> BEM, only few a posteriori error estimators<br />

and adaptive strategies have been proposed in<br />

<strong>the</strong> literature. All <strong>of</strong> <strong>the</strong>m imply a significant implementational<br />

and computational overhead. In <strong>the</strong> following,<br />

257<br />

we introduce a simple and efficient approach <strong>of</strong> practical<br />

relevance.<br />

II. MODEL PROBLEM<br />

We consider <strong>the</strong> Dirichlet screen problem<br />

Δu =0 in R 3 \ Γ,<br />

u = f on Γ ⊆ R 2 ×{0},<br />

u = O(�x� −1 (1)<br />

) as x →∞,<br />

which describes <strong>the</strong> potential u away from an electrified<br />

thin plate Γ loaded with potential f, see [1]. This problem<br />

is equivalent to Symm’s integral equation<br />

Vφ(x) :=− 1<br />

�<br />

1<br />

4π Γ �x − y� φ(y) dsy = f(x). (2)<br />

Then, φ is <strong>the</strong> charge density on <strong>the</strong> plate Γ which<br />

is known to show singularities along <strong>the</strong> edges. Of<br />

special interest in physics is <strong>the</strong> so-called capacitance<br />

C = − 1<br />

�<br />

4π Γ φds, where <strong>the</strong> charge density φ now<br />

solves Symm’s integral equation (2) with f =1.<br />

III. MAIN RESULTS<br />

A. A POSTERIORI ERROR ESTIMATION<br />

The h-h/2-strategy is a basic technique for <strong>the</strong> a posteriori<br />

error estimation, well-known from <strong>the</strong> context <strong>of</strong><br />

ordinary differential equations. In [2], this approach is<br />

proposed in <strong>the</strong> context <strong>of</strong> BEM. Let φ denote <strong>the</strong> exact<br />

(but in general unknown) solution <strong>of</strong> (2). One <strong>the</strong>n considers<br />

ηH := �φh − φh/2� (3)<br />

to estimate <strong>the</strong> error �φ − φh�,whereφh is <strong>the</strong> Galerkin<br />

solution with respect to a mesh Th and φ h/2 is <strong>the</strong><br />

Galerkin solution for a mesh T h/2 obtained from a uniform<br />

refinement <strong>of</strong> Th. We stress that ηH is always a<br />

lower bound<br />

ηH ≤�φ− φh�, (4)<br />

even with known constant 1. Under <strong>the</strong> saturation assumption<br />

�φ − φh/2� ≤σ �φ − φh� with some constant<br />

σ ∈ (0, 1), <strong>the</strong>re holds<br />

�φ − φh� ≤<br />

1<br />

√ 1 − σ 2 ηH. (5)


This means, ηH also gives an upper bound for <strong>the</strong> error.<br />

However, for boundary element methods, <strong>the</strong> energy<br />

norm �·�is non-local and thus <strong>the</strong> error estimator ηH<br />

does not provide information about <strong>the</strong> local error. Recent<br />

localization techniques from [3] allow to replace <strong>the</strong><br />

energy norm by mesh-size weighted L 2 -norms. For instance,<br />

<strong>the</strong> estimator μH defined by<br />

μ 2 H<br />

= �<br />

T ∈Th<br />

μ 2 H,T<br />

:= �<br />

T ∈Th<br />

hT �φh − φh/2� 2 L2 (T ) , (6)<br />

where hT denotes <strong>the</strong> diameter <strong>of</strong> an element T ∈Th, is<br />

equivalent to ηH, see [4].<br />

B. ADAPTIVE ALGORITHM<br />

Based on <strong>the</strong> error estimators ηH and μH from <strong>the</strong> previous<br />

section, we now introduce an adaptive algorithm.<br />

With a fixed parameter θ ∈ (0, 1) as well as an initial<br />

mesh Th, our strategy reads as follows: Until ηH is sufficiently<br />

small, do:<br />

1. Refine Th uniformly to obtain Th/2. 2. Compute discrete solutions φh and φh/2 as well as<br />

corresponding error estimators ηH and μH.<br />

3. Find minimal set M⊆Th such that<br />

θμ 2 H = θ �<br />

μ<br />

T ∈Th<br />

2 H,T ≤ �<br />

μ<br />

T ∈M<br />

2 H,T (7)<br />

4. Refine T ∈Mto obtain new mesh Th.<br />

IV. NUMERICAL EXPERIMENTS<br />

Figure 1 shows <strong>the</strong> discrete charge density φh computed<br />

over an adaptively generated mesh Th which discretizes<br />

an L-shaped plate. The solution shows strong singularities,<br />

i.e. peaks, at all edges and <strong>the</strong> convex corners <strong>of</strong> <strong>the</strong><br />

simulation domain. Our algorithm leads to a well aligned<br />

mesh showing refinements towards <strong>the</strong> edges, thus resolving<br />

<strong>the</strong> singularities efficiently.<br />

Figure 2 shows <strong>the</strong> true error �φ − φh� as well as <strong>the</strong><br />

error estimators ηH and μH in <strong>the</strong> uniform and adaptive<br />

case. All quantities are plotted over <strong>the</strong> number <strong>of</strong> boundary<br />

elements N. In <strong>the</strong> double logarithmic plot <strong>the</strong> convergence<br />

rate is <strong>the</strong> slope <strong>of</strong> a straight line. We stress<br />

that for uniform meshes <strong>the</strong> equivalence h ∼ N −1/2 implies<br />

that <strong>the</strong> optimal order <strong>of</strong> convergence is O(N −3/4 ).<br />

A uniform approach only reveals a convergence order <strong>of</strong><br />

O(N −1/4 ) due to <strong>the</strong> singularities <strong>of</strong> <strong>the</strong> unknown solution<br />

φ. Our proposed adaptive algorithm, however, recovers<br />

<strong>the</strong> optimal order <strong>of</strong> convergence O(N −3/4 ).<br />

REFERENCES<br />

[1] V. J. Ervin, E. P. Stephan, and S. Abou El-Seoud. An<br />

improved boundary element method for <strong>the</strong> charge<br />

density <strong>of</strong> a thin electrified plate in R 3 . Ma<strong>the</strong>matical<br />

Methods in <strong>the</strong> Applied Sciences, 13(4):291–303,<br />

1990.<br />

258<br />

Figure 1: Discrete charge density φh on L-shaped plate<br />

and enlargement <strong>of</strong> <strong>the</strong> solution at <strong>the</strong> lower left corner.<br />

error and estimators<br />

10 0<br />

10 −1<br />

10 −2<br />

10 1<br />

error (unif.)<br />

η H (unif.)<br />

μ H (unif.)<br />

error (adap.)<br />

η H (adap.)<br />

μ H (adap.)<br />

3/4<br />

10 2<br />

1<br />

number <strong>of</strong> elements<br />

Figure 2: Error and error estimators plotted over <strong>the</strong> number<br />

<strong>of</strong> boundary elements used.<br />

[2] S. Ferraz-Leite. A Posteriori Fehlerschätzer für die<br />

Symmsche Integralgleichung in 3D. Diploma Thesis<br />

(in German), Institute for Analysis and Scientific<br />

Computing, Vienna University <strong>of</strong> Technology,<br />

Vienna, Austria, October 2007.<br />

[3] C. Carstensen and D. Praetorius. Averaging<br />

techniques for <strong>the</strong> effective numerical solution <strong>of</strong><br />

Symm’s integral equation <strong>of</strong> <strong>the</strong> first kind. SIAM<br />

Journal on Scientific Computing, 27(4):1226–1260<br />

(electronic), 2006.<br />

[4] S. Ferraz-Leite and D. Praetorius. Simple a posteriori<br />

error estimators for <strong>the</strong> h-version <strong>of</strong> <strong>the</strong> boundary<br />

element method. Technical report, ASC Report<br />

01/2007, July 2007.<br />

1<br />

10 3<br />

1/4<br />

10 4


Risk Habitat Megacity - Recycling Systems in Santiago<br />

de Chile<br />

Tahnee González, Klaus-Rainer Bräutigam, Helmut Seifert<br />

Forschungszentrum Karlsruhe, POB 3640, D-76021<br />

Karlsruhe, Germany<br />

Email: Tahnee.gonzalez@itc-tab.fzk.de<br />

Helmut.seifert@itc-tab.fzk.de<br />

Klaus-Rainer.Braeutigam@itas.fzk.de<br />

Abstract — Within <strong>the</strong> research initiative “Risk<br />

Habitat Megacity” a study about sustainability<br />

within <strong>the</strong> field <strong>of</strong> waste management for <strong>the</strong> city <strong>of</strong><br />

Santiago de Chile will be work out. A flow chart <strong>of</strong><br />

<strong>the</strong> current waste flows in Santiago is presented and<br />

recycling systems are briefly described, including<br />

<strong>the</strong> contribution <strong>of</strong> <strong>the</strong> informal primary collectors,<br />

who supply more than 90% <strong>of</strong> <strong>the</strong> total materials<br />

recycled in <strong>the</strong> city.<br />

I. INTRODUCTION<br />

Megacities are centres <strong>of</strong> complex and highly dynamic<br />

processes, such as population growth and<br />

density, change <strong>of</strong> land use, informal and formal<br />

economies, etc. However, <strong>the</strong>ir real impact on global<br />

environmental change is currently not well understood.<br />

How are interactions between <strong>the</strong>se processes?<br />

How could <strong>the</strong> risks associated with megacities be<br />

turned into opportunities for a sustainable development?<br />

These are some <strong>of</strong> <strong>the</strong> questions that <strong>the</strong> research<br />

initiative “Risk Habitat Megacity” <strong>of</strong> <strong>the</strong><br />

German Helmholtz-Association intended to answer.<br />

The initiative is interdisciplinary and integrative<br />

and it means to establish a conceptual framework<br />

based on three <strong>the</strong>oretical concepts:<br />

1. The concept <strong>of</strong> sustainable development: defines<br />

criteria towards basic sustainability<br />

goals.<br />

2. The risk concept: focus on extend and severity<br />

<strong>of</strong> <strong>the</strong> problems and <strong>the</strong>ir threat for future sustainability.<br />

3. The governance concept: focus on stakeholders<br />

and policy makers that are important<br />

for implementation <strong>of</strong> strategies.<br />

The initiative applies <strong>the</strong>se three concepts, to several<br />

typical megacity issues, such as land use management,<br />

socio-spatial differentiation, energy system,<br />

transportation, air quality and health, water resources<br />

and services, and waste management.<br />

This research focuses on <strong>the</strong> Latin America Region,<br />

which is characterized by being <strong>the</strong> developing<br />

region more urbanized and with more wealth distri-<br />

259<br />

bution inequalities [1]. The pilot city chosen for <strong>the</strong><br />

study is Santiago de Chile because it shows <strong>the</strong> main<br />

problems <strong>of</strong> large cities in <strong>the</strong> region, additionally<br />

international institutions like <strong>the</strong> Economic Commission<br />

for Latin America and <strong>the</strong> Caribbean<br />

(ECLAC/CEPAL) are located in Santiago and with<br />

cooperation <strong>of</strong> <strong>the</strong> diverse Chilean partners it would<br />

be feasible to transfer <strong>the</strong> developed conceptual<br />

framework to o<strong>the</strong>r large cities in this geographical<br />

area.<br />

II. OBJECTIVES<br />

As a part <strong>of</strong> <strong>the</strong> “Risk Habitat Megacity” study, a<br />

doctoral work has being proposed in <strong>the</strong> waste management<br />

application field.<br />

The general objective <strong>of</strong> this research is to develop<br />

a waste management concept in <strong>the</strong> Metropolitan<br />

Region <strong>of</strong> Santiago (MRS), based on <strong>the</strong> Integrative<br />

Helmholtz Sustainability Concept [2] considering <strong>the</strong><br />

integration <strong>of</strong> <strong>the</strong> informal sector.<br />

This study is searching for answers to <strong>the</strong> following<br />

questions:<br />

1. Which is <strong>the</strong> role <strong>of</strong> <strong>the</strong> informal sector in <strong>the</strong><br />

current waste management system?<br />

2. How sustainable is <strong>the</strong> current waste management<br />

situation in <strong>the</strong> MRS?<br />

3. Which technical systems will guarantee a<br />

more sustainable management <strong>of</strong> municipal<br />

solid waste?<br />

4. How can <strong>the</strong> informal sector be integrated into<br />

<strong>the</strong> formal waste labour market and technical<br />

systems?<br />

This paper focuses on <strong>the</strong> investigation <strong>of</strong> question<br />

number 1.<br />

III. METHODOLOGY<br />

In order to obtain information related with recycling<br />

systems and to understand <strong>the</strong> current situation <strong>of</strong> <strong>the</strong><br />

informal sector, field research was carried out in<br />

Santiago de Chile. This research included technical<br />

interviews and visits to <strong>the</strong> most important stakeholders<br />

in waste management, such as municipalities


with recycling programs, industries, non governmental<br />

organizations and leaders <strong>of</strong> scavengers associations.<br />

IV. WASTE RECYCLING IN MRS<br />

A. BACKGROUND<br />

Recycling has taken place in Chile since 1970 [3],<br />

but more as an informal activity carried out by primary<br />

collectors (known in Chile as cartoneros) who<br />

collect valuable materials on <strong>the</strong> streets, than a formal<br />

organized action.<br />

B. RECYCLING SYSTEMS<br />

Figure 1 shows <strong>the</strong> diagram for <strong>the</strong> mass flow from<br />

<strong>the</strong> municipal solid waste in Santiago. About 90 % <strong>of</strong><br />

this waste is correctly disposed <strong>of</strong> in one <strong>of</strong> <strong>the</strong> three<br />

landfills in <strong>the</strong> area (2.5 Million tonnes per year) (see<br />

flow 1).<br />

According to <strong>of</strong>ficial data <strong>the</strong> recycling rate in<br />

2004 in <strong>the</strong> Metropolitan Region was 9% [4]. Currently<br />

formal recycling actions are developed only in<br />

five municipalities (MRS in total consists <strong>of</strong> 52<br />

municipalities), consisting basically on segregated<br />

collection <strong>of</strong> valuable materials which are separated,<br />

classified and given to production companies <strong>of</strong><br />

paper and cardboard, plastics, glass and scrap, as<br />

most important ones. The contribution <strong>of</strong> <strong>the</strong>se initiatives<br />

corresponds only to 1% <strong>of</strong> <strong>the</strong> total recycled<br />

(3,100 tons per year – flow 9 in Figure 4).<br />

Ano<strong>the</strong>r formal initiative consists on several containers<br />

located in public places such as supermarkets,<br />

churches, parks, etc. where people can bring <strong>the</strong>ir<br />

materials and drop <strong>the</strong>m <strong>of</strong>f directly into <strong>the</strong> containers.<br />

It has been estimated that <strong>the</strong> contribution <strong>of</strong><br />

<strong>the</strong>se recycling scheme is 6.7% (19,000 tons per year<br />

– flow 3 in Figure 1) being glass <strong>the</strong> material mostly<br />

collected.<br />

Municipal<br />

Solid<br />

Waste<br />

2,8<br />

Mio. Ton<br />

pro year<br />

Bins<br />

Bags<br />

Drop <strong>of</strong>f<br />

Systems<br />

Formal<br />

Collection<br />

1<br />

Landfills<br />

Informal<br />

Collection<br />

Paper<br />

Figure 1: Diagram Flow <strong>of</strong> Municipal Solid Waste<br />

in <strong>the</strong> Metropolitan Region <strong>of</strong> Santiago<br />

On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> largest contribution to recycling<br />

(92.3%) is carried out by <strong>the</strong> informal sector,<br />

with a total <strong>of</strong> 260,000 tons per year (flow 2 in Figure<br />

1). Figure 2 compares <strong>the</strong> contributions <strong>of</strong> <strong>the</strong><br />

3<br />

4<br />

2<br />

260<br />

three recycling schemes in MRS. The importance <strong>of</strong><br />

informal collectors is evident.<br />

100%<br />

90%<br />

80%<br />

70%<br />

60%<br />

50%<br />

40%<br />

30%<br />

20%<br />

10%<br />

0%<br />

Paper /<br />

cardboard<br />

Figure 2: Materials recycled in RMS with three<br />

different recycling systems<br />

V. CONCLUSIONS<br />

� Informal sector contributes with more than<br />

90% <strong>of</strong> <strong>the</strong> total recycling in MRS.<br />

� Integration <strong>of</strong> informal sector into formal<br />

waste management schemes can contribute to<br />

achieve higher recycling rates.<br />

ACKNOWLEDGMENTS<br />

I would like to acknowledge funding <strong>of</strong> <strong>the</strong><br />

present PhD <strong>the</strong>sis given by <strong>the</strong> Initiative and<br />

Networking Fund <strong>of</strong> <strong>the</strong> Helmholtz Association.<br />

Additionally funding for <strong>the</strong> foreign stay in<br />

Santiago de Chile was given by <strong>the</strong> Karlsruhe<br />

House <strong>of</strong> Young <strong>Scientist</strong>s.<br />

REFERENCES<br />

Municipal Program<br />

Charity<br />

Informal Sector<br />

Glass Metals Plastics<br />

[1] United Nation. The millennium development<br />

goals: A Latin America and Caribbean perspective.<br />

United Nation Publications, pages 41, 61,<br />

Santiago de Chile, Chile, August 2005.<br />

[2] Kopfmüller, J. et al. 2001 Nachhaltige Entwicklung<br />

integrativ betrachtet. Konstitutive Elemente,<br />

Regeln, Indikatoren. Berlin, Germany, 2001.<br />

[3] Szantó Marcel. Gestión de Residuos Sólidos<br />

Domiciliarios en Santiago de Chile. Pontifica<br />

Universidad Católica de Valparaíso, Grupo de<br />

Residuos Sólidos, p. 21, Valparaíso, Chile, July<br />

2006.<br />

[4] Szantó Marcel. Caracterización de residuos sólidos<br />

domiciliarios en la Región Metropolitana.<br />

Pontifica Universidad Católica de Valparaíso,<br />

Grupo de Residuos Sólidos, p. 93, Valparaíso,<br />

March 2006.


Slurry sampling ICP-AES procedure for fast and accurate<br />

measurement <strong>of</strong> crustal elements in airborne particulate matter<br />

Azam Mukhtar and Andreas Limbeck<br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: amukhtar@mail.zserv.tuwien.ac.at<br />

Abstract — Slurry sampling ICP-AES method for<br />

fast and accurate determination <strong>of</strong> crustal elements<br />

(Ca, Si, Fe, Mg, Ti, Al) in airborne particulate<br />

matter is presented. Decomposition and mineralization<br />

<strong>of</strong> <strong>the</strong> filter material including deposited<br />

particulate matter was obtained by treatment <strong>of</strong> <strong>the</strong><br />

samples with a mixture <strong>of</strong> nitric acid and perchloric<br />

acid. The prepared slurry solutions were comprised<br />

<strong>of</strong> <strong>the</strong> dissolved part <strong>of</strong> <strong>the</strong> sample as well as <strong>the</strong><br />

suspended insoluble material (such as Si oxides),<br />

and were measured by ICP-AES using La as internal<br />

standard. Quantification was performed using<br />

aqueous standards. Analysis <strong>of</strong> <strong>the</strong> certified reference<br />

material 2709 ® (San Joaquin soil) revealed a<br />

good agreement for <strong>the</strong> elements Ca, Mg, Fe and T.<br />

The contents <strong>of</strong> Al and Si were underestimated<br />

which might be due to loss <strong>of</strong> volatile species during<br />

sample preparation and/or influences <strong>of</strong> <strong>the</strong> matrix.<br />

I. INTRODUCTION<br />

The short term standard for particulate matter PM10 is<br />

violated in many European cities [1] , <strong>the</strong>refore <strong>the</strong><br />

current air quality status <strong>of</strong> PM10 is under investigation<br />

in many countries in particular to understand <strong>the</strong><br />

source terms <strong>of</strong> <strong>the</strong> particulate matter and to propose<br />

a catalogue <strong>of</strong> measures to avoid fur<strong>the</strong>r exceeding <strong>of</strong><br />

<strong>the</strong> limiting value.<br />

The crustal elements Si, Al, Fe, Ca, Mg and Ti have<br />

been shown to be important constituents <strong>of</strong> airborne<br />

particulate matter. For measurement <strong>of</strong> <strong>the</strong>se elements<br />

in atmospheric samples, preferably nondestructive<br />

multi-elemental techniques, such as X-ray<br />

fluorescence (XRF), particle induced x-ray emission<br />

(PIXE) or instrumental neutron activation analysis<br />

(INAA) were used [2,3] . However although frequently<br />

applied, all <strong>of</strong> <strong>the</strong>se methods suffer from distinct<br />

limitations. X-ray fluorescence does not allow <strong>the</strong><br />

quantitative determination at <strong>the</strong> lowest level <strong>of</strong><br />

μg g -1 and requires a sophisticated calibration.<br />

Similarly neutron activation analysis encounters with<br />

matrix problems. So <strong>the</strong> application <strong>of</strong> <strong>the</strong>se techniques<br />

for routine analysis <strong>of</strong> aerosol samples is<br />

difficult. However, <strong>the</strong> use <strong>of</strong> inductively coupled<br />

261<br />

plasma atomic emission spectroscopy (ICP-AES) in<br />

comparison to above mentioned techniques, yield<br />

enhanced selectivity, reproducibility and in some<br />

cases also sensitivity, but <strong>the</strong>se improvements are<br />

always associated with distinctly increased efforts in<br />

sample pretreatment and measurement<br />

[4,5] . The<br />

present study describes <strong>the</strong> development <strong>of</strong> a slurry<br />

ICP-AES procedure for fast and accurate multielement<br />

measurement <strong>of</strong> Si, Al, Fe, Ca, Mg and Ti in<br />

solid environmental samples.<br />

II. EXPERIMENTAL<br />

A. INSTRUMENTATION<br />

ICP-AES measurement <strong>of</strong> slurry samples was performed<br />

using a Spectr<strong>of</strong>lame P optical emission<br />

spectrometer (Spectro, Germany). The instrument<br />

was equipped with a pneumatic nebulizer and a<br />

cyclonic spray chamber and was controlled by<br />

Spectro Smart Analyzer S<strong>of</strong>tware.<br />

B. REAGENTS ANS SAMPLES<br />

Nitric acid, perchloric acid and all o<strong>the</strong>r reagents<br />

used were <strong>of</strong> analytical grade. For method development<br />

a certified multi-element solution (Ca, Fe, Mg,<br />

Ti, 100 mg/L) and stock solutions <strong>of</strong> Al and Si (1000<br />

mg/L) procured from Merck (Germany) were used.<br />

Working solutions were prepared by dilution with<br />

metal free water. Certified reference material 2709 ®<br />

(CRM, San Joaquin Soil) from NIST (National Institute<br />

<strong>of</strong> Standards and Technology, Gai<strong>the</strong>rsburg,<br />

MD, USA) was used to verify <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong><br />

method for multi-element determination. PM10 samples<br />

were collected in Vienna using standard procedures<br />

and GN4-Matricel filter as sampling substrate.<br />

C. PRE-TREATMENT PROCEDURE<br />

Slurry solutions <strong>of</strong> PM10 test samples were prepared<br />

by treatment <strong>of</strong> PM10 samples with a mixture <strong>of</strong> conc.<br />

nitric acid and perchloric acid in bernas type Teflonlined<br />

bomb heated in refractory oven at 110 o C for 1<br />

hr and subsequently temperature was increased to<br />

150 o C for 1 hr ensuring an enhanced decomposition


<strong>of</strong> carbonaceous material, and simultaneously <strong>the</strong><br />

excess <strong>of</strong> nitric acid was evaporated. After cooling<br />

<strong>the</strong> samples to room temperature, derived solution <strong>of</strong><br />

perchloric acid including undissolved particles was<br />

completely transferred to 20 mL polypropylene tubes<br />

and diluted to a final volume <strong>of</strong> 20 mL with 2.5%<br />

(v/v) HNO3 in bi-distilled water.<br />

III. RESULTS<br />

A- METHOD OPTIMIZATION<br />

The slurry method demands <strong>the</strong> complete mineralization<br />

<strong>of</strong> organic material, dissolution <strong>of</strong> soluble sample<br />

constituents and <strong>the</strong> homogeneous suspension <strong>of</strong><br />

insoluble material (e.g. Si and Al oxides) [6] . To<br />

achieve this condition, various concentrations <strong>of</strong><br />

nitric acid and perchloric acid at various temperatures<br />

were used. Finally, treatment <strong>of</strong> PM10 samples with 2<br />

mL <strong>of</strong> nitric acid and 500 μL <strong>of</strong> perchloric acid at<br />

110 o C for 1 hr and subsequently increasing temperature<br />

to 150 o C for 1 hr were found to give slurry solutions<br />

with comparable characteristics to aqueous<br />

solutions, enabling <strong>the</strong> use <strong>of</strong> aqueous standards for<br />

quantification.<br />

Instrumental parameters with respect to flow rate,<br />

gas pressure and RF power were also optimized to<br />

get maximum signal intensity, minimum background<br />

signal, high reproducibility and to overcome<br />

matrix and spectral interferences. Lanthanum<br />

was used as internal standard for correction<br />

<strong>of</strong> changing conditions during measurement.<br />

B- ANALYSIS OF REAL SAMPLES<br />

Prior to <strong>the</strong> application <strong>of</strong> <strong>the</strong> method to real sample<br />

<strong>the</strong> accuracy <strong>of</strong> <strong>the</strong> procedure was tested by <strong>the</strong><br />

analysis <strong>of</strong> <strong>the</strong> CRM. For this purpose 20 mg <strong>of</strong><br />

<strong>the</strong> CRM was analyzed using proposed method<br />

and <strong>the</strong> quantification was performed using aqueous<br />

standards. The results derived for <strong>the</strong> investigated<br />

elements were in good agreement with certified<br />

contents except for Al and Si (Table 1). The<br />

insufficient results for Al and Si were probably<br />

due to loss <strong>of</strong> volatile species during sample pretreatment<br />

and/or influences <strong>of</strong> remaining matrix<br />

during measurement.<br />

Finally <strong>the</strong> developed procedure was applied to <strong>the</strong><br />

analysis <strong>of</strong> PM10 samples from Vienna, confirming<br />

that <strong>the</strong> investigated crustal elements were major<br />

constituents <strong>of</strong> urban aerosol samples.<br />

262<br />

Mg<br />

Al<br />

Si<br />

Ca<br />

Fe<br />

Ti<br />

Elements Content (%)<br />

Measured<br />

1.52±0.032<br />

5.90±0.6<br />

24.8±7<br />

1.85±0.04<br />

3.46±0.12<br />

0.328±0.05<br />

Content (%)<br />

Certified<br />

1.51±0.05<br />

7.50±0.06<br />

29.66±0.23<br />

1.89±0.05<br />

3.50±0.11<br />

0.342±0.024<br />

Table 1: Comparison <strong>of</strong> measured and certified<br />

values <strong>of</strong> standard reference material<br />

IV- CONCLUSION<br />

The presented slurry sampling method enables fast<br />

and accurate analysis <strong>of</strong> Ca, Mg, Fe and Ti in airborne<br />

particulate matter and can widely be employed<br />

to o<strong>the</strong>r solid and environmental samples. Future<br />

work will be dedicated to <strong>the</strong> problems associated<br />

with measurement <strong>of</strong> Al and Si, so that this method<br />

can also be applied to for accurate determination <strong>of</strong><br />

<strong>the</strong>se elements.<br />

REFERENCES<br />

[1] Querol et al, Speciation and origin <strong>of</strong> PM10 and<br />

PM2.5 in selected European cities, Atmos. Envor. 38:<br />

6547-6555, 2004.<br />

[2] V. Ariola , et al, Elemental characterization <strong>of</strong><br />

PM10, PM2.5 and PM1 in <strong>the</strong> town <strong>of</strong> Genoa<br />

(Italy), chemosphere, 62: 226-232, 2006.<br />

[3] A. Najat, Saliba, H. Kouyoumdjian, M. Roumie,<br />

Effect <strong>of</strong> local and long-range transport emissions<br />

on <strong>the</strong> elemental composition <strong>of</strong> PM10–2.5<br />

and PM2.5 in Beirut, Atmospheric Environment,<br />

41: 6497–6509, 2007.<br />

[4] A.P. Krushevska, R.M. Barnes, Determination<br />

<strong>of</strong> low silicon concentration in food and coral soil<br />

by inductively coupled plasma atomic emission<br />

spectrometry, J. Anal. At. Spectrum. 9: 981-984,<br />

1994.<br />

[5] P. Klemens, K.G. Heumann, Development <strong>of</strong><br />

an ICP-HRIDMS method for accurate determination<br />

<strong>of</strong> traces <strong>of</strong> silicon in biological and clinical<br />

samples, Fresenius J. anal. Chem. 371: 758-763,<br />

2001.<br />

[6] M.J. Cal-Prieto et al, Slurry sampling for direct<br />

analysis <strong>of</strong> solid materials by electro<strong>the</strong>rmal atomic<br />

absorption spectrometry (ETAAS), a literature<br />

review from 1990 to 2000, Talanta, 56: 1-51,2002.


Effects <strong>of</strong> Environmental Degradation on <strong>the</strong> Response <strong>of</strong> Reinforced<br />

Concrete Structures<br />

Paola Simioni and Anna Saetta (Faculty Mentor)<br />

Department <strong>of</strong> Architectural Construction<br />

University IUAV <strong>of</strong> Venice<br />

Venice, Italy<br />

Email: paolasim@dipolo.dicea.unifi.it; saetta@iuav.it<br />

Abstract — Durable design and maintenance <strong>of</strong><br />

r.c. structures have become a great societal concern,<br />

especially in <strong>the</strong> last decades, considering <strong>the</strong><br />

extensive use <strong>of</strong> concrete for infrastructure facilities.<br />

Significant efforts have been made by <strong>the</strong> scientific<br />

community in order to enhance an efficient<br />

distribution <strong>of</strong> resources for <strong>the</strong> assessment and<br />

prevention <strong>of</strong> <strong>the</strong> environmental degradation which<br />

a structure may suffer during its service life.<br />

In this paper, <strong>the</strong> major effects <strong>of</strong> aggressive environments<br />

on <strong>the</strong> structural response <strong>of</strong> r.c. members<br />

are analysed. Particular attention is paid to reinforcement<br />

corrosion that may increase <strong>the</strong> vulnerability<br />

<strong>of</strong> <strong>the</strong> structural elements to natural hazards.<br />

I. INTRODUCTION<br />

The safety and serviceability assessment <strong>of</strong> r.c. structures<br />

is strongly related to <strong>the</strong>ir time-variant degradation,<br />

which may produce significant changes in <strong>the</strong><br />

safety coefficients, with respect to both <strong>the</strong> ultimate<br />

and service limit states. These degradation processes<br />

may be ascribed to <strong>the</strong> diffusive attack <strong>of</strong> environmental<br />

aggressive agents, such as chloride and sulphate<br />

ions, or to carbonation, alkali-aggregate reaction,<br />

freeze-thaw cycles, etc. In reinforced concrete,<br />

<strong>the</strong> most detrimental deterioration mechanisms are<br />

those leading to reinforcement corrosion.<br />

Experimental tests performed on corroded r.c.<br />

members have shown <strong>the</strong> reduction <strong>of</strong> <strong>the</strong>ir load<br />

carrying capacity and ductility as <strong>the</strong> level <strong>of</strong> corrosion<br />

increases, e.g. [1]. Such effects are herein considered<br />

in <strong>the</strong> framework <strong>of</strong> <strong>the</strong> Risk Management,<br />

and both <strong>the</strong> static (at a “micro-level”) and dynamic<br />

(at a “macro-level”) behaviours <strong>of</strong> r.c. elements suffering<br />

from corrosion attacks have been modelled.<br />

II. CORROSION RISK MANAGEMENT<br />

Combining <strong>the</strong> general Risk Management procedure<br />

developed within <strong>the</strong> International Graduate College<br />

802, e.g. [2], with <strong>the</strong> methodology proposed in [1]<br />

for assessing <strong>the</strong> residual service life <strong>of</strong> corrosionaffected<br />

r.c. structures, a proposal for a Corrosion<br />

Risk Management Process is herein presented.<br />

263<br />

The procedure consists <strong>of</strong> three main parts, Figure<br />

1. First, <strong>the</strong> Corrosion Detection phase is performed<br />

through inspections, if necessary repeated at certain<br />

time intervals. Preliminary visual inspection allows<br />

<strong>the</strong> identification <strong>of</strong> <strong>the</strong> deterioration mechanisms,<br />

mapping damage and selecting <strong>the</strong> sites for testing.<br />

Afterwards <strong>the</strong> exposure class is recognized and insitu<br />

testing is carried out. When corrosion is definitely<br />

detected, <strong>the</strong> Corrosion Assessment phase<br />

starts. After <strong>the</strong> estimate <strong>of</strong> <strong>the</strong> time for corrosion<br />

initiation, <strong>the</strong> main effects <strong>of</strong> corrosion are evaluated,<br />

i.e. loss <strong>of</strong> steel section, reduction <strong>of</strong> steel-concrete<br />

bond strength and degradation <strong>of</strong> concrete cover.<br />

Then <strong>the</strong> analyses on <strong>the</strong> corroded structure are carried<br />

on, resulting in <strong>the</strong> diagnosis and prognosis on<br />

its safety condition, which are <strong>the</strong> input data for <strong>the</strong><br />

final phase <strong>of</strong> Corrosion Treatment.<br />

CORROSION DETECTION<br />

yes no<br />

CORROSION ASSESSMENT<br />

Initiation time<br />

ANALYSIS<br />

DIAGNOSIS<br />

PROGNOSIS<br />

CORROSION TREATMENT<br />

Inspection<br />

time<br />

Residual steel section<br />

Residual bond strength<br />

Residual concrete section<br />

Management<br />

strategy<br />

Figure 1: Corrosion Risk Management Process<br />

In <strong>the</strong> present work, <strong>the</strong> Corrosion Risk Assessment<br />

phase is considered, at two different levels: a<br />

micro-level, consisting on <strong>the</strong> improvement <strong>of</strong> a<br />

durability model which accounts for <strong>the</strong> main effects<br />

<strong>of</strong> corrosion and is able to predict <strong>the</strong> structural response<br />

<strong>of</strong> corroded r.c. members; a macro-level,<br />

aiming at <strong>the</strong> safety assessment <strong>of</strong> existing structures<br />

under seismic excitations. The major results are<br />

briefly discussed in <strong>the</strong> following sub-sections.


A. MICRO-LEVEL ANALYSIS<br />

A model representing <strong>the</strong> loss <strong>of</strong> steel-concrete bond<br />

as a function <strong>of</strong> <strong>the</strong> corrosion degree has been formulated<br />

and implemented in a research code enclosing<br />

<strong>the</strong> coupled mechanical-environmental damage<br />

model developed in [3]. In particular, a bond damage<br />

parameter has been introduced, [4], being a simple<br />

tool for describing bond degradation phenomena.<br />

The model reliability has been assessed by simulating<br />

experimental pull-out tests and beam tests,<br />

performed at different levels <strong>of</strong> corrosion, obtaining a<br />

good agreement with <strong>the</strong> numerical results, e.g. [4].<br />

B. MACRO-LEVEL ANALYSIS<br />

Corrosion may also induce <strong>the</strong> shift <strong>of</strong> <strong>the</strong> failure<br />

mechanism from ductile to a more brittle type. This<br />

effect becomes <strong>of</strong> great concern especially in seismic<br />

areas, where ductility and <strong>the</strong> actual collapse mechanism<br />

are key issues in safety assessment.<br />

Pushover analyses have been performed in <strong>the</strong><br />

framework <strong>of</strong> lumped plasticity. By means <strong>of</strong> <strong>the</strong>oretical<br />

as well as experimental relationships, <strong>the</strong><br />

effects <strong>of</strong> corrosion have been considered by modifying<br />

<strong>the</strong> moment-rotation relationships <strong>of</strong> <strong>the</strong> plastic<br />

hinges including steel section loss, reduction <strong>of</strong> steel<br />

ultimate deformation and concrete cover degradation.<br />

An applicative example <strong>of</strong> <strong>the</strong> variation, respect to<br />

<strong>the</strong> pristine state, <strong>of</strong> <strong>the</strong> seismic behaviour <strong>of</strong> r.c.<br />

structures affected by a moderate level <strong>of</strong> corrosion<br />

due to a carbonation attack, is proposed in [5]. Moreover,<br />

<strong>the</strong> case <strong>of</strong> a non-symmetric corrosion attack<br />

may result in detrimental torsional effects, [6].<br />

III. PROBABILISTIC MODELLING<br />

It is well-known that, due to <strong>the</strong> high level <strong>of</strong> uncertainty<br />

affecting <strong>the</strong> performance <strong>of</strong> deteriorating<br />

structures, <strong>the</strong> time-variant structural safety can be<br />

realistically assured only in probabilistic terms.<br />

In <strong>the</strong> present research, <strong>the</strong> durability model, [3],<br />

has been adopted to evaluate <strong>the</strong> probability <strong>of</strong> corrosion<br />

initiation including statistical uncertainties in <strong>the</strong><br />

most influencing material and environmental parameters<br />

(identified by a preliminary sensitivity<br />

analysis). In particular, it is assumed that corrosion<br />

starts when <strong>the</strong> carbonation coefficient at <strong>the</strong> depth <strong>of</strong><br />

<strong>the</strong> concrete cover exceeds <strong>the</strong> threshold value 0.1.<br />

Figure 2 shows <strong>the</strong> results for different w/c ratios and<br />

a mean value <strong>of</strong> relative humidity <strong>of</strong> 70%.<br />

IV. CONCLUDING REMARKS<br />

The assessment <strong>of</strong> existing r.c. structures affected by<br />

environmental degradation is a complex issue due to<br />

<strong>the</strong> large numbers <strong>of</strong> factors coming into play. Considerable<br />

progress has been made in <strong>the</strong> attempt <strong>of</strong> a<br />

durable design and efficient maintenance planning<br />

264<br />

and repair, e.g. [7]. However fur<strong>the</strong>r research is necessary,<br />

in particular concerning a realistic model <strong>of</strong><br />

loss <strong>of</strong> bond, [8]. Future work will be devoted to <strong>the</strong><br />

improvement <strong>of</strong> <strong>the</strong> suggested bond model and to <strong>the</strong><br />

merger <strong>of</strong> <strong>the</strong> micro and macro-level analyses for a<br />

detailed analysis <strong>of</strong> plastic hinge zones.<br />

probability <strong>of</strong> corrosion initiation<br />

1.0<br />

0.8<br />

0.6<br />

0.4<br />

w/c = 0.6<br />

0.2<br />

0.0<br />

w/c = 0.5<br />

w/c = 0.4<br />

0 10 20 30 40 50 60 70 80<br />

initiation time [yr]<br />

Figure 2: Probability <strong>of</strong> corrosion initiation<br />

ACKNOWLEDGMENTS<br />

The present research is carried on in <strong>the</strong> framework<br />

<strong>of</strong> <strong>the</strong> IGC 802 on “Risk Management <strong>of</strong> Natural and<br />

Civilization Hazards on Buildings and Infrastructure”,<br />

with a partnership between <strong>the</strong> University <strong>of</strong><br />

Florence and <strong>the</strong> TU <strong>of</strong> Braunschweig.<br />

REFERENCES<br />

[1] J. Rodriguez and C. Andrade. A validated users<br />

manual for assessing <strong>the</strong> residual service life <strong>of</strong><br />

concrete structures. GEOCISA, Madrid, 2001.<br />

[2] T. Pliefke, S. Sperbeck, M. Urban, U. Peil, H.<br />

Budelmann. A standardized methodology for<br />

managing disaster risk - An attempt to remove<br />

ambiguity. 5 th Int. Prob. Workshop, Ghent, 2007.<br />

[3] A. Saetta, R. Scotta, R. Vitaliani. Coupled Environmental-Mechanical<br />

Damage Model <strong>of</strong> RC<br />

Structures. J. Eng. Mech., 125(8):930-940, 1999.<br />

[4] L. Berto, P. Simioni, A. Saetta. Numerical modelling<br />

<strong>of</strong> bond behaviour in RC structures affected<br />

by reinforcement corrosion. Eng. Struct,<br />

30:1375–1385, 2007.<br />

[5] A. Saetta, P. Simioni, L. Berto, R. Vitaliani. Seismic<br />

response <strong>of</strong> corroded r.c. structures. Int. fib<br />

Symp. <strong>2008</strong>, Amsterdam, <strong>the</strong> Ne<strong>the</strong>rlands.<br />

[6] L. Berto, R. Vitaliani, A. Saetta, P. Simioni. Seismic<br />

assessment <strong>of</strong> existing RC structures affected<br />

by degradation phenomena. Submitted, <strong>2008</strong>.<br />

[7] Model Code for Service Life Design. Fib Bulletin<br />

34, CEB-FIP; 2006.<br />

[8] M. Faber, D. Straub, M. Maes. A computational<br />

framework for risk assessment <strong>of</strong> RC structures<br />

using indicators. Computer-Aided Civil and Infr.<br />

Eng., 21:216–230, 2006.


Usage <strong>of</strong> <strong>the</strong> Global Positioning System GPS in <strong>the</strong> fields <strong>of</strong><br />

wea<strong>the</strong>r forecast and fleet navigation<br />

Ana Karabatic, Gottfried Thaler and Robert Weber<br />

Institute <strong>of</strong> Geodesy and Geophysics<br />

Technical University <strong>of</strong> Vienna<br />

Vienna, Austria<br />

Email: {anna,gthaler,rweber}@mars.hg.tuwien.ac.at<br />

Abstract — The Global Navigation Satellite System<br />

GPS, which was established by <strong>the</strong> US military<br />

in <strong>the</strong> early 1970´s, has become very important for<br />

civil navigation and also for geosciences. Nowadays<br />

a huge number <strong>of</strong> applications make use <strong>of</strong> <strong>the</strong><br />

GPS.<br />

In this paper we will present results <strong>of</strong> two basically<br />

different applications, carried out at our institute.<br />

The first one deals with <strong>the</strong> topic wea<strong>the</strong>r forecast<br />

based on <strong>the</strong> atmospheric influences on microwave<br />

signals. The second application is about using GPS<br />

in <strong>the</strong> field <strong>of</strong> fleet navigation in urban areas.<br />

I. INTRODUCTION<br />

Since <strong>the</strong> invention <strong>of</strong> <strong>the</strong> Global Positioning System<br />

GPS by <strong>the</strong> US military in <strong>the</strong> early 1970´s, <strong>the</strong> importance<br />

<strong>of</strong> this system has also grown for civil usage.<br />

The Russians established <strong>the</strong>ir own system<br />

GLONASS and fur<strong>the</strong>r systems like Galileo (Europe)<br />

and COMPASS (China) will come. All toge<strong>the</strong>r<br />

<strong>the</strong>se systems are called Global Navigation Satellite<br />

Systems or GNSS. The fundamental observation<br />

equations for code and phase measurements are equal<br />

to all systems and read as follows<br />

and<br />

(1)<br />

(2)<br />

is <strong>the</strong> Pseudorange, <strong>the</strong> real<br />

distance, <strong>the</strong> clock error <strong>of</strong> <strong>the</strong> receiver, <strong>the</strong><br />

clock error <strong>of</strong> <strong>the</strong> satellite, c <strong>the</strong> speed <strong>of</strong> light, <strong>the</strong><br />

wavelength <strong>of</strong> <strong>the</strong> signal and <strong>the</strong> unknown ambiguity<br />

factor. All GNSS applications are based on<br />

<strong>the</strong>se two observation equations for Code and Phase.<br />

Below we present two different applications for<br />

GNSS carried out at our institute.<br />

II. GNSS FOR WEATHER FORECAST<br />

One <strong>of</strong> many applications <strong>of</strong> <strong>the</strong> Global Positioning<br />

System is <strong>the</strong> investigation <strong>of</strong> atmospheric influences<br />

on microwave signals. The availability <strong>of</strong> real-time<br />

265<br />

or near real-time tropospheric data derived from GPS<br />

measurements to be assimilated in existing meteorological<br />

models represents an important issue. While<br />

<strong>the</strong> zenith hydrostatic delay <strong>of</strong> GPS microwave signals<br />

is usually well sizeable, <strong>the</strong> wet component,<br />

describing <strong>the</strong> rapid variable water vapour content <strong>of</strong><br />

<strong>the</strong> troposphere has to be estimated from <strong>the</strong> observation<br />

data. High resolution meteorological analysis <strong>of</strong><br />

<strong>the</strong> humidity field is an important precondition for a<br />

better monitoring <strong>of</strong> local and regional extreme precipitation<br />

events and for forecasts with improved<br />

spatial resolution.<br />

Figure 1: Hourly ZWD estimates<br />

In this presentation we present <strong>the</strong> results <strong>of</strong> continuous<br />

measurements <strong>of</strong> a regional network consisting<br />

<strong>of</strong> 8 GPS reference stations, located in Carinthia,<br />

Austria. The network has been extended with surrounding<br />

stations <strong>of</strong> <strong>the</strong> IGS and EUREF-network.<br />

The aim is to provide GNSS based measurements <strong>of</strong><br />

<strong>the</strong> tropospheric water vapour content with a temporal<br />

resolution <strong>of</strong> one hour and a temporal delay <strong>of</strong><br />

less than one hour (Fig. 1) to use <strong>the</strong>m within <strong>the</strong><br />

INCA (Integrated Nowcasting through Comprehensive<br />

Analysis) system, operated by <strong>the</strong> Austrian Meteorological<br />

Service (ZAMG).<br />

III. GNSS FOR FLEET-NAVIGATION<br />

Ano<strong>the</strong>r application for GNSS, which was carried<br />

out within <strong>the</strong> project called PPos – Taxi [1] at our<br />

institute, can be assigned to <strong>the</strong> field <strong>of</strong> fleet navigation.<br />

This time GNSS-Signals are used to compute<br />

<strong>the</strong> positions <strong>of</strong> hundreds <strong>of</strong> GPS receivers, in this


special case <strong>the</strong> positions <strong>of</strong> taxis from a taxi fleet in<br />

Vienna.<br />

To perform this task we need <strong>the</strong> positions and<br />

clock errors <strong>of</strong> <strong>the</strong> GPS satellites, as well as at least<br />

four Pseudorange measurements to four GPS satellites.<br />

If <strong>the</strong> Code observation equation in (1) is used position<br />

accuracies <strong>of</strong> about ± 5-8 m can be achieved.<br />

In <strong>the</strong> case <strong>of</strong> using <strong>the</strong> Phase observation equation in<br />

(2) positioning on <strong>the</strong> cm-level is possible. However,<br />

solving <strong>the</strong> integer ambiguities in (2) is not possible<br />

in real-time approaches like fleet navigation. It is<br />

possible to improve <strong>the</strong> positioning accuracy with<br />

Code measurements by using Differential GNSS<br />

approaches. In this case correction terms, so called<br />

Pseudorange-corrections PRCs, are transmitted to <strong>the</strong><br />

user by reference station networks. This approach<br />

allows for accuracies <strong>of</strong> 1 – 2 meters. Figure 2 shows<br />

<strong>the</strong> results <strong>of</strong> a test drive as differences to a reference<br />

trajectory.<br />

Figure 2: Position differences <strong>of</strong> test drive<br />

Ano<strong>the</strong>r problem concerning positioning accuracy<br />

is <strong>the</strong> so called Multipath-effect, especially in urban<br />

areas like Vienna. Multipath occurs, when <strong>the</strong> satellite<br />

signal is reflected at glass, facades, trees, water<br />

surfaces etc. The travel time <strong>of</strong> <strong>the</strong> reflected signal is<br />

longer than <strong>the</strong> travel time <strong>of</strong> <strong>the</strong> direct signal. This<br />

results in an increased Pseudorange which again<br />

results in a position error. Errors due to Multipath<br />

can reach up to 50 – 100 meters [2].<br />

Nowadays <strong>the</strong>re are many different ways to mitigate<br />

Multipath-effects [3]. One way, which was<br />

examined in detail in this application, is <strong>the</strong> analysis<br />

<strong>of</strong> <strong>the</strong> Signal-to-Noise-Ratio SNR <strong>of</strong> <strong>the</strong> received<br />

Signal. Reflection or diffraction <strong>of</strong> <strong>the</strong> signal usually<br />

leads to a decrease in <strong>the</strong> SNR. The SNR <strong>of</strong> <strong>the</strong> signal<br />

is also related to <strong>the</strong> elevation (angle <strong>of</strong> <strong>the</strong> satellite<br />

above <strong>the</strong> horizon) <strong>of</strong> <strong>the</strong> satellite. Due to <strong>the</strong><br />

longer travel path through <strong>the</strong> atmosphere signals<br />

from lower elevations have lower SNR-values too.<br />

These two facts can be used to detect Multipath and<br />

fur<strong>the</strong>r on to mitigate it [2]. Figure 3 shows an exam-<br />

266<br />

ple <strong>of</strong> <strong>the</strong> relationship between SNR and elevation<br />

and candidates for Multipath-signals.<br />

Figure 3: Relationship SNR and Multipath<br />

Finally Figure 4 shows <strong>the</strong> results <strong>of</strong> a test drive in<br />

Vienna using DGNSS corrections and Multipath<br />

mitigation.<br />

Figure 4: Corrected car path <strong>of</strong> test ride<br />

REFERENCES<br />

[1] M. Haberler-Weber, G. Thaler, R. Weber: Efficient<br />

Taxi Allocation based on SISNet and<br />

DGPS-corrected Vehicle Positioning, Toulouse<br />

Space Show 08, Toulouse, 22.-25.04.<strong>2008</strong><br />

[2] G. Thaler, Mltipathreduktion bei der<br />

Fahrzeugpositionierung im innerstädtischen<br />

Bereich, Diplomarbeit, TU Wien, Institut für<br />

Geodäsie und Geophysik, 2007<br />

[3] B. H<strong>of</strong>mann-Wellenh<strong>of</strong>, H. Lichtenegger, J.<br />

Collins, GPS Theory and Practice, Springer<br />

Wien New York, Fünfte überarbeitete Auflage,<br />

2001


Organically Modified Mixed-Oxide Sol-Gel Films with Macro<br />

Meso and Micro Porosity<br />

Marina Lomoschitz, Ulrich Schubert<br />

Institute <strong>of</strong> Materials Chemistry<br />

Vienna University <strong>of</strong> Technology,<br />

1060 Vienna, Austria<br />

Email: mjosip@mail.tuwien.ac.at<br />

Abstract — Highly porous mixed oxide sol-gel films<br />

with multiple organic functionality and different<br />

pore-size regimes were manufactured by a simple<br />

dip coating method using different metal precur-<br />

sors, different organic groups and combinations <strong>of</strong><br />

several porogens. Post-syn<strong>the</strong>sis modification with<br />

phosphonic acids allows introduction <strong>of</strong> a fur<strong>the</strong>r<br />

organic functionality. Spectroscopic methods like<br />

NMR and IR were used for qualitative analysis <strong>of</strong><br />

<strong>the</strong> organic functionalities. Macro- and mesoporos-<br />

ity were verified using transmission electron mi-<br />

croscopy (TEM) and atomic force microscopy<br />

(AFM). Nitrogen sorption measurements at 77 K<br />

were used for quantification <strong>of</strong> <strong>the</strong> micro- meso<br />

pore surface area. For investigation and better<br />

understanding <strong>of</strong> material formation, small angle X-<br />

ray scattering (SAXS) measurements were per-<br />

formed. The degree <strong>of</strong> post-functionalization was<br />

determined by <strong>the</strong>rmogravimetric analysis (TGA).<br />

I. INTRODUCTION<br />

Surfactant templated mesoporous sol-gel materials<br />

have potential applications in catalysis, separation<br />

and sensing because <strong>of</strong> <strong>the</strong>ir high surface area,<br />

controllable pore size and narrow pore size distri-<br />

bution. Adsorption applications may require ex-<br />

traordinary materials properties such as combina-<br />

tion <strong>of</strong> different adsorption sites. Different metal<br />

sites and different organic functionalities may<br />

increase <strong>the</strong> interactions <strong>of</strong> adsorbate mixtures in<br />

one material. Ano<strong>the</strong>r very important criterion for<br />

adsorption is high mass transport which can be<br />

achieved by macropores.<br />

In <strong>the</strong> current work we present <strong>the</strong> combina-<br />

tion <strong>of</strong> approaches for creating films with polymo-<br />

dal (macro-, meso- and micropores) porosity as<br />

well as complex materials compositions (different<br />

metals, different organic groups) 1 in one material.<br />

Thus, a combination <strong>of</strong> (i) several porogens, (ii)<br />

co-condensation <strong>of</strong> different metal alkoxide pre-<br />

cursors (Fig.1) and (iii) post-syn<strong>the</strong>sis modifica-<br />

tion was used to obtain highly porous mixed oxide<br />

films with multiple organic functionality and dif-<br />

ferent pore-size regimes.<br />

Fig. 1 Sol-gel precursors and templates<br />

II. RESULTS AND DISCUSSION<br />

For creating macro porosity polystyrene (PS)<br />

spheres <strong>of</strong> 100 nm diameter were deposited onto<br />

cleaned glass slides from a dispersion in water by<br />

dip coating. This is a simple and convenient<br />

method which allows formation <strong>of</strong> densely packed<br />

layers <strong>of</strong> polymer spheres by self-assembly. 2 In a<br />

subsequent step <strong>the</strong> slides were immersed in a sol<br />

PS + H 2O<br />

sol<br />

toluene<br />

ΔT post-functionallized<br />

micro-meso-macro<br />

porous material<br />

P OH<br />

OH<br />

O<br />

R<br />

micro porous<br />

material<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

meso pore<br />

inorganic precursor<br />

+ template<br />

polystyrene<br />

sphere dispersion<br />

removal <strong>of</strong><br />

polystyrene<br />

micro-meso-macro<br />

porous hybrid material<br />

macro pore<br />

PS + H 2O<br />

sol<br />

toluene<br />

ΔT post-functionallized<br />

micro-meso-macro<br />

porous material<br />

P OH<br />

OH<br />

O<br />

R<br />

micro porous<br />

material<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr Si<br />

Z r<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

S i<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Z r<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr S i<br />

Zr<br />

Si<br />

Zr<br />

S i<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

Zr<br />

Si<br />

Zr<br />

Si<br />

Zr<br />

meso pore<br />

inorganic precursor<br />

+ template<br />

polystyrene<br />

sphere dispersion<br />

removal <strong>of</strong><br />

polystyrene<br />

micro-meso-macro<br />

porous hybrid material<br />

macro pore<br />

Fig. 2 Schematic illustration <strong>of</strong> <strong>the</strong> film preparation procedure<br />

Si<br />

O<br />

O O<br />

O<br />

Si<br />

O<br />

O CH 3<br />

O<br />

Si<br />

O<br />

O<br />

O<br />

Zr<br />

O<br />

O O<br />

O<br />

Ti<br />

O<br />

O O<br />

O<br />

O<br />

O<br />

O<br />

O<br />

OH<br />

n<br />

TEOS MTES PTES Zr(OPr) 4 Ti(OiPr) 4<br />

Brij 56<br />

267


<strong>of</strong> hydrolyzed alkoxide precursors. During this<br />

step <strong>the</strong> PS spheres are coated with an organically<br />

modified mixed oxide gel network.<br />

After sol-gel network stabilization by drying,<br />

PS porogen was removed by extraction. The<br />

calcination at adequate temperature allowed removal<br />

<strong>of</strong> residual PS and surfactant without removing <strong>the</strong><br />

organic groups covalently bonded to <strong>the</strong> silicon atoms<br />

<strong>of</strong> <strong>the</strong> gel network. All steps <strong>of</strong> film manufacturing<br />

are schematically shown in Fig. 2.<br />

The formation <strong>of</strong> uniform macropores was<br />

demonstrated using TEM. Figure 3 shows that <strong>the</strong><br />

films had uniform ellipsoidal macropores after calcination.<br />

The anisotropic shrinkage during calcination,<br />

from originally 100 nm to 70 nm in <strong>the</strong> present case,<br />

is a common effect during heat treatment <strong>of</strong> porous<br />

films, where a higher tendency <strong>of</strong> film pore shrinking<br />

perpendicular to surface was noticed. 3<br />

Fig. 3 TEM image <strong>of</strong> macro pores<br />

The stability <strong>of</strong> organic groups covalently bonded<br />

to silica matrix during calcination was monitored by<br />

IR spectroscopy (Fig. 4).<br />

ν OH<br />

ν CH template<br />

ν C-O-C<br />

ν Si-CH 3<br />

4000 3500 3000 2500 2000 1500 1000 500<br />

Wavenumber / cm -1<br />

ν Zr-O-Si<br />

ν Si-O-Si<br />

Si-CH 3<br />

130°C<br />

200°C<br />

300°C<br />

400°C<br />

Fig. 4 IR spectra <strong>of</strong> a film deposited onto a KBr<br />

disk, during temperature treatment<br />

To quantify <strong>the</strong> surface area <strong>of</strong> <strong>the</strong> porous films,<br />

meso and microporosity was investigated by N2<br />

sorption measurements at 77K (Fig. 5). The Brun-<br />

268<br />

auer-Emmett-Teller (BET) method was applied to<br />

calculate <strong>the</strong> total surface area, while <strong>the</strong> t-plot<br />

method was used to determine microporosity. The<br />

average mesopore sizes were calculated according to<br />

<strong>the</strong> BJH model 5 and NLDFT was applied on adsorption<br />

iso<strong>the</strong>rms to evaluate <strong>the</strong> size <strong>of</strong> <strong>the</strong> micropores.<br />

Volume Adsorbed / g/cm³<br />

120<br />

100<br />

80<br />

60<br />

40<br />

0,0 0,2 0,4 0,6 0,8 1,0<br />

Fig. 5 N2 sorption iso<strong>the</strong>rm and pore size distribution<br />

Phosphonic acids are known to coordinate tightly<br />

to titanium and zirconium and selective modification<br />

<strong>of</strong> <strong>the</strong> metal sites in MO2 (M = Ti, Zr)/SiO2 composites<br />

could be accomplished in aqueous solution. 6 The<br />

post-functionalization <strong>of</strong> <strong>the</strong> calcined samples with<br />

phosphonic acids, was verified by IR spectroscopy<br />

and <strong>the</strong>rmogravimetric analysis (TGA).<br />

ACKNOWLEDGMENTS<br />

This work was supported by <strong>the</strong> European Space<br />

Agency (ESA) and <strong>the</strong> Austrian Forschungsförderungsgesellschaft<br />

(FFG) in <strong>the</strong> framework <strong>of</strong> <strong>the</strong><br />

Austrian Space Application Programme (ASAP).<br />

REFERENCES<br />

dV / d log (d)<br />

0,06<br />

0,05<br />

0,04<br />

0,03<br />

0,02<br />

0,01<br />

0,00<br />

0 5 10 15 20 25 30 35 40<br />

Pore diameter / nm<br />

p/p°<br />

[1] P. C. Angelome and G. J. d. A. A. Soler-Illia, J.<br />

Mater. Chem. 2005, 15, 3903-3912.<br />

[2] M. Sakurai, A. Shimojima, M. Heishi and K.<br />

Kuroda, Langmuir 2007, 23, 10788-10792.<br />

[3] R. A. Farrell, N. Petkov, H. Amenitsch, J. D.<br />

Holmes and M. A. Morris, J. Mater. Chem. <strong>2008</strong>, 18,<br />

2213-2220.<br />

[4] K. S. W. Sing, D. H. Everett, R. A. W. Haul, L.<br />

Moscou, R. A. Pierotti, J. Rouquerol and T. Siemieniewska,<br />

Pure Appl. Chem. 1985, 57, 603-619.<br />

[5] J. C. Groen, L. A. A. Peffer and J. Perez-Ramirez,<br />

Microporous and Mesoporous Materials 2003, 60, 1-<br />

17.<br />

[6] P. H. Mutin, V. Lafond, A. F. Popa, M. Granier,<br />

L. Markey and A. Dereux, Chem. Mater. 2004, 16,<br />

5670-5675.


Development <strong>of</strong> a New High Frequency Fatigue Testing Method<br />

for Concrete Bridges<br />

V<br />

Dyn<br />

Johannes Berger and Johann Kollegger (Faculty Mentor)<br />

Institute for Structural Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: johannes.berger+e212@tuwien.ac.at<br />

Abstract — This paper describes a new method<br />

for high frequency fatigue testing, based on <strong>the</strong><br />

resonance effect. Because <strong>of</strong> <strong>the</strong> stiffness, mass<br />

and damping features, <strong>the</strong> test specimen itself<br />

becomes <strong>the</strong> major element <strong>of</strong> <strong>the</strong> testing facility.<br />

The aim was to increase <strong>the</strong> test frequency from<br />

1 Hz to 10 Hz with an energy consumption which<br />

is a fractional amount compared to conventional<br />

methods (servo-hydraulic facilities).<br />

Fur<strong>the</strong>rmore, an analytical and numerical<br />

analysis for a fatigue testing facility was carried<br />

out, to analyse <strong>the</strong> material behaviour <strong>of</strong> a new<br />

bridge restoration method.<br />

I. INTRODUCTION<br />

Fatigue is one <strong>of</strong> <strong>the</strong> main reasons <strong>of</strong> failure in<br />

mechanical components and structures subjected to<br />

cycling loading. The testing <strong>of</strong> structural components<br />

under cyclic loading constitutes one <strong>of</strong> <strong>the</strong> most<br />

important fields <strong>of</strong> experimental mechanics [1]. At<br />

<strong>the</strong> Institute for Structural Engineering at Vienna<br />

University <strong>of</strong> Technology, a new method for<br />

dynamic fatigue tests for tendons has been developed<br />

and patented [2]. The new method is based on <strong>the</strong><br />

physical principle <strong>of</strong> <strong>the</strong> resonance effect, and is now<br />

also being used for reinforced concrete. For <strong>the</strong><br />

development <strong>of</strong> a new bridge restoration method, by<br />

concrete overlay in compound with <strong>the</strong> structure, it is<br />

necessary to carry out dynamic fatigue tests for <strong>the</strong><br />

fatigue verification, in order to gain information<br />

about <strong>the</strong> long-term behaviour <strong>of</strong> <strong>the</strong> compound.<br />

II. PRINCIPLE OF OPERATION<br />

The basic concept <strong>of</strong> <strong>the</strong> test arrangement is to<br />

choose an excitation frequency (e.g. vibration<br />

generator) which correlates to <strong>the</strong> resonance<br />

frequency <strong>of</strong> <strong>the</strong> testing facility. Then <strong>the</strong> applied<br />

force is multiplied due to resonance, whereas <strong>the</strong><br />

dynamic magnification factor (VDyn) depends on<br />

<strong>the</strong> damping ratio (x), as in Eq. (1).<br />

1<br />

�<br />

��� ��<br />

2<br />

2 (1 )<br />

(1)<br />

269<br />

K<br />

M<br />

x<br />

Figure 1: (a) SDOF-System, (b) Model <strong>of</strong> <strong>the</strong><br />

testing facility<br />

The testing facility is designed as a single degree<br />

<strong>of</strong> freedom system (SDOF), shown in Figure 1(a).<br />

The differential equation <strong>of</strong> this system is<br />

described in Equation (2), where M is <strong>the</strong> mass<br />

and K is <strong>the</strong> stiffness. The eigenvalue (w) and <strong>the</strong><br />

natural frequency (f) can be determined easily, see<br />

Eq. (3) and Eq. (4).<br />

2<br />

� x(t)<br />

M �K�x 2<br />

(t) �0<br />

�t<br />

��<br />

f<br />

K2<br />

K<br />

M<br />

�<br />

�<br />

2 ��<br />

(2)<br />

(3)<br />

(4)<br />

III. DESIGN OF THE TESTING FACILITY<br />

To obtain practical values <strong>of</strong> dynamic fatigue tests <strong>of</strong><br />

bridges, experiments which have already been<br />

performed at <strong>the</strong> Technical University <strong>of</strong> Graz and<br />

<strong>the</strong> University <strong>of</strong> Innsbruck, were analysed. In<br />

addition, a bridge restoration with concrete overlay<br />

has been analysed for <strong>the</strong> fatigue verification. Based<br />

on this experience and on <strong>the</strong> basis <strong>of</strong> <strong>the</strong> fatigue<br />

verification according to EN 1992-1-1 [3], <strong>the</strong><br />

specimen has been designed.<br />

In order to design <strong>the</strong> load-bearing <strong>of</strong> a bridge as<br />

good as possible, <strong>the</strong> cantilever and field area are<br />

modelled for <strong>the</strong> specimen as a tension and a<br />

compression area supplement.<br />

M<br />

K1<br />

K2<br />

K3 K3<br />

K4 K4<br />

(a) (b)


IV. ANALYTICAL AND NON-LINEAR<br />

NUMERICAL SIMULATION<br />

To obtain high efficiency <strong>the</strong> testing facility, shown<br />

in Figure (2), was designed in such a way that <strong>the</strong><br />

resonance effect will set in at a frequency <strong>of</strong> about 10<br />

Hz. The first investigations revealed, that regarding<br />

<strong>the</strong> resonance frequency state I (concrete without<br />

cracks) and state II (concrete with cracks) show<br />

considerable differences. The frequency ranges<br />

between 26,21 Hz in state I and 13,52 Hz at <strong>the</strong><br />

compression area supplement and 8,65 Hz at <strong>the</strong><br />

tension area supplement in state II. These frequencies<br />

were determined by analytical methods.<br />

Figure 2: Experimental set-up<br />

However, since <strong>the</strong>re will be no pure state II<br />

condition, <strong>the</strong> calculations need to be done with a<br />

realistic model. This is necessary to avoid resonance<br />

overlapping <strong>of</strong> <strong>the</strong> specimen with o<strong>the</strong>r parts <strong>of</strong> <strong>the</strong><br />

testing facility. These calculations are made by using<br />

a finite element program, taking into account a<br />

nonlinear material behaviour and a concrete cracking<br />

model, see Figure (3). With <strong>the</strong> previously<br />

determined stiffness, <strong>the</strong> resonance frequency was<br />

determined analytically and numerically by using <strong>the</strong><br />

finite element method in a 3D- model.<br />

Fur<strong>the</strong>rmore, it was essential to determine <strong>the</strong><br />

magnification factor <strong>of</strong> <strong>the</strong> system, whe<strong>the</strong>r <strong>the</strong><br />

capacity <strong>of</strong> <strong>the</strong> vibration generator is sufficient or<br />

not.<br />

270<br />

Figure 3: Material behaviour, nonlinear 2D FE-<br />

Analysis<br />

V. ACHIEVEMENT<br />

Finally seven specimens with <strong>the</strong> following<br />

dimensions were built:<br />

� Length: 5,0m<br />

� Width: 1,0m<br />

� Thickness: 0,4m (including 16cm <strong>of</strong> top<br />

concrete as decking)<br />

Four million load cycles for each specimen were<br />

carried out for <strong>the</strong> fatigue test. This means that with a<br />

frequency <strong>of</strong> 10 Hz, it takes 111 hours to finish one<br />

test.<br />

VI. CONCLUSION<br />

High frequency fatigue tests, based on <strong>the</strong> resonance<br />

effect, can be carried out for reinforced concrete<br />

(RC) structures. As a result <strong>of</strong> <strong>the</strong> variable stiffness<br />

<strong>of</strong> RC in state II, <strong>the</strong>re is a big range <strong>of</strong><br />

eigenfrequencies <strong>of</strong> <strong>the</strong> testing facility. The limits are<br />

given by state I and state II conditions. The tests can<br />

be performed 8 to 26 times faster by using <strong>the</strong> new<br />

developed test set-up, compared to <strong>the</strong> servo<br />

hydraulic testing procedure.<br />

REFERENCES<br />

[1] B. Koeberl and J. Kollegger. High frequency<br />

testing facility for stay cables and tendons.<br />

Structural Engineering International, August<br />

<strong>2008</strong>.<br />

[2] Kollegger J, Koeberl B, Pardatscher H, Vill M.<br />

Verfahren zur Durchführung von Dauerschwingversuchen<br />

an einem Prüfkörper sowie eine Vorrichtung<br />

zur Durchführung des Verfahrens.<br />

Austrian Patent AT 501 168 B1, 2006.<br />

[3] EN 1992-1-1. Eurocode 2: Bemessung und Konstruktion<br />

von Stahlbeton und Spannbetontragwerken,<br />

Teil 1-1: Allgemeine Bemessungsregeln<br />

und Regeln für den Hochbau, Dezember 2004.


Antenna far-fields determination from phaseless<br />

measurement using <strong>the</strong> real-valued GA<br />

Jan Puskely and Zden�k Nová�ek (Faculty Mentor)<br />

Departure <strong>of</strong> Radio Electronics<br />

Brno University <strong>of</strong> Technology<br />

Czech Republic<br />

Email: xpuske01@stud.feec.vutbr.cz, novacek@feec.vutbr.cz<br />

Abstract — In this paper, we present an<br />

optimization procedure exploiting in a functional<br />

minimization method that utilizes <strong>the</strong> phaseless<br />

electric field data over two plane surfaces for <strong>the</strong><br />

reconstruction <strong>of</strong> <strong>the</strong> phase distribution on <strong>the</strong><br />

aperture <strong>of</strong> <strong>the</strong> measured antenna. The<br />

performance <strong>of</strong> a real-valued genetic algorithm<br />

applied to <strong>the</strong> functional minimization is presented.<br />

Analyses <strong>of</strong> <strong>the</strong> different pairing and mating<br />

methods which influence <strong>the</strong> GA behavior are<br />

included.<br />

I. INTRODUCTION<br />

When dealing with a near-field to far-field (NF-<br />

FF) transformation for an antenna measurement,<br />

traditional techniques require not only <strong>the</strong> amplitude<br />

but also <strong>the</strong> phase information <strong>of</strong> <strong>the</strong> NF data in order<br />

to calculate accurately <strong>the</strong> far-field pattern <strong>of</strong> <strong>the</strong><br />

antenna under test (AUT). However, it is difficult to<br />

obtain <strong>the</strong> complex (amplitude and phase)<br />

measurement at millimeter-wave frequencies. Also,<br />

at low frequencies, <strong>the</strong> amplitude-only measurement<br />

techniques are economical to employ.<br />

Thus, this work is focused on calculating <strong>the</strong> fields<br />

on <strong>the</strong> antenna aperture that characterize <strong>the</strong> AUT or<br />

generate <strong>the</strong> FF information using amplitude-only<br />

information <strong>of</strong> <strong>the</strong> electric field collected over two<br />

plane surfaces in <strong>the</strong> near field. In this work, a realvalued<br />

genetic algorithm (R-VGA) [3] has been<br />

developed and tuned to predict <strong>the</strong> radiation pattern<br />

<strong>of</strong> <strong>the</strong> AUT.<br />

II. PRINCIPLE OF THE METHOD<br />

The reconstruction method using <strong>the</strong> amplitudeonly<br />

data on two planes is based on <strong>the</strong> phaseretrieval<br />

techniques. Reconstructing <strong>the</strong> phase <strong>of</strong> <strong>the</strong><br />

NF data is <strong>the</strong> main goal in those techniques. This<br />

usually leads to an iterative scheme in which <strong>the</strong><br />

phase <strong>of</strong> <strong>the</strong> known amplitude near-field data on two<br />

different planes is reconstructed in a forward–<br />

backward fashion. The presented method exploits<br />

two known principles, <strong>the</strong> functional minimization<br />

[2] and <strong>the</strong> Fourier iterative algorithm (FIA) [1].<br />

271<br />

In <strong>the</strong> functional minimization scheme using GA<br />

for finding <strong>the</strong> global minimum, <strong>the</strong> random guess is<br />

refined in each step by <strong>the</strong> minimization difference<br />

between <strong>the</strong> calculated amplitudes and measured<br />

ones. Since <strong>the</strong> GA can slow down <strong>the</strong> convergence,<br />

we use it to find an initial estimation only. In order to<br />

improve <strong>the</strong> initial estimation, <strong>the</strong> FIA is applied,<br />

which minimizes <strong>the</strong> functional also. The process<br />

continues until some stopping criteria are satisfied.<br />

By <strong>the</strong> Fast Fourier Transform (FFT), coefficients<br />

<strong>of</strong> <strong>the</strong> plane wave spectrum (PWS) are calculated and<br />

<strong>the</strong> far-field pattern can be obtained consequently.<br />

A. REAL-VALUED GENETIC ALGORITHM<br />

Genetic Algorithm (GA) is a stochastic search<br />

method widely used in electromagnetism to find <strong>the</strong><br />

global minimum in a multidimensional domain. The<br />

set <strong>of</strong> <strong>the</strong> parameters forms <strong>the</strong> population, which is<br />

evolved by means <strong>of</strong> <strong>the</strong> selection, <strong>the</strong> crossover, and<br />

<strong>the</strong> mutation genetic operators [3]. The fitness<br />

function F to be minimized is <strong>of</strong> <strong>the</strong> following form:<br />

N<br />

N<br />

2 ~ 2<br />

�E1�i, j�<br />

� M 1�i,<br />

j�<br />

�<br />

~ 2<br />

M �i, j�<br />

F � ��<br />

� ... (1)<br />

i�1<br />

j�1<br />

1<br />

In (1), E1(i, j) is <strong>the</strong> computed complex intensity and<br />

M ~ 1(i, j) is <strong>the</strong> measured amplitude in <strong>the</strong> point i, j on<br />

<strong>the</strong> first scanning plane. For <strong>the</strong> second scanning<br />

plane, a similar scheme is applied.<br />

Case<br />

#1<br />

#2<br />

#3<br />

#4<br />

#5<br />

#6<br />

#7<br />

Selection Crossover Mutation<br />

strategy strategy Pc (%) Pm (%)<br />

Random<br />

Rank weight<br />

Cost weight<br />

Tournament<br />

Tournament<br />

Tournament<br />

Tournament<br />

Blending<br />

Blending<br />

Blending<br />

Blending<br />

Linear<br />

Heuristic<br />

Extrapolation<br />

2<br />

70 10-30<br />

Table 1: The strategies and parameters considered<br />

to tune up <strong>the</strong> real-valued GA [3]<br />

For <strong>the</strong> proposed problem, <strong>the</strong> influence <strong>of</strong> <strong>the</strong><br />

selection and crossover strategies and <strong>the</strong> influence<br />

<strong>of</strong> <strong>the</strong> population size on <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong> results


are investigated. The chosen strategies are<br />

summarized in Table 1. The optimizations were<br />

repeated 10 times and were stopped in <strong>the</strong> 3000 th<br />

iteration.<br />

III.RESULTS<br />

The radiation pattern reconstruction was<br />

performed for <strong>the</strong> horn antenna T1-R100. The<br />

antenna was analyzed at <strong>the</strong> working frequency <strong>of</strong><br />

12.4 GHz.<br />

With regard to <strong>the</strong> fact that <strong>the</strong> tournament<br />

selection and <strong>the</strong> linear crossover <strong>of</strong>fer <strong>the</strong> most<br />

accurate results, <strong>the</strong>y were chosen (shown in Fig. 1).<br />

Since <strong>the</strong> computational cost depends on <strong>the</strong><br />

population size, we try to find out <strong>the</strong> optimal<br />

population size to achieve <strong>the</strong> accurate solutions<br />

saving CPU time. The population sizes were chosen<br />

12, 24, 48 and 96 individuals.<br />

Figure 1: The analysis <strong>of</strong> <strong>the</strong> influence <strong>of</strong> <strong>the</strong><br />

different crossover operators: #4, #5, #6, #7.<br />

Figure 2: The performance <strong>of</strong> <strong>the</strong> R-VGA as a<br />

function <strong>of</strong> <strong>the</strong> population size; results for case #5.<br />

As shown in Fig. 2, larger population <strong>of</strong>fers <strong>the</strong><br />

fast convergence but only for a cost <strong>of</strong> <strong>the</strong> higher<br />

expense <strong>of</strong> <strong>the</strong> CPU time. The population size <strong>of</strong> 24<br />

individuals was chosen as an optimum since<br />

providing <strong>the</strong> best performance ratio.<br />

The far-field results <strong>of</strong> <strong>the</strong> E plane obtained by <strong>the</strong><br />

R-VGA and <strong>the</strong> FIA are shown in Fig. 3. The<br />

272<br />

radiation pattern achieved by <strong>the</strong> direct<br />

transformation <strong>of</strong> <strong>the</strong> complex electric field intensity<br />

is represented by <strong>the</strong> dotted line. The radiation<br />

pattern calculated from <strong>the</strong> initial estimation<br />

(achieved by <strong>the</strong> R-VGA) is represented by <strong>the</strong><br />

dashed line, and <strong>the</strong> radiation pattern achieved by<br />

applying <strong>the</strong> FIA is represented by <strong>the</strong> solid line.<br />

Figure 3: Reconstructed far-field pattern; E-plane<br />

IV. CONCLUSION<br />

Method for <strong>the</strong> reconstruction <strong>of</strong> <strong>the</strong> radiation<br />

pattern based on <strong>the</strong> functional minimization using<br />

GA for finding <strong>the</strong> initial estimation has been<br />

presented. The behaviour <strong>of</strong> <strong>the</strong> R-VGA has been<br />

analyzed. The confrontation <strong>of</strong> <strong>the</strong> R-VGA and<br />

particle swarm optimization (PSO), which showed<br />

<strong>the</strong> best performance till now, shows that <strong>the</strong> R-<br />

VGA exhibits comparable convergence properties.<br />

ACKNOWLEDGMENTS<br />

Research described in this contribution was<br />

financially supported by <strong>the</strong> Czech Grant Agency<br />

under <strong>the</strong> grants no. 102/07/1084 and<br />

102/08/H018, and by <strong>the</strong> research program MSM<br />

0021630513.<br />

REFERENCES<br />

[1] R. G. Yaccarino and Y. R. Samii, “Phase-less bipolar<br />

planar near-field measurements and<br />

diagnostics <strong>of</strong> array antennas,” IEEE Trans.<br />

Antennas Propagat., vol. 47, pp. 574–583, Mar.<br />

1999.<br />

[2] Bucci, O. M., D’elia, G., Leone, G., Pierri, R.<br />

Far-field pattern determination from <strong>the</strong> nearfield<br />

amplitude on two surfaces. IEEE<br />

Transaction on Antennas and Propagation.<br />

1990, vol. 38, no. 11, p. 1772–1779.<br />

[3] Haupt, R. L., Haupt, S. E. Practical genetic<br />

algorithm. A Wiley-Interscience Publication,<br />

1998, ISBN 0-471-18873-5


Catalyst loaded porous membranes for<br />

environmental applications - Smart Membranes<br />

Abstract —This project involves <strong>the</strong> fabrication<br />

and testing <strong>of</strong> microporous, polymer membranes<br />

designed to remove minute amounts <strong>of</strong> toxic air<br />

pollutants such as formaldehyde from air streams.<br />

The hypo<strong>the</strong>sis to be tested is that active silver<br />

contained within <strong>the</strong> porous polymer membranes<br />

results in high formaldehyde retention. The silver<br />

caps in <strong>the</strong> porous membrane absorb formaldehyde<br />

from <strong>the</strong> air, while in fact at <strong>the</strong> same time <strong>the</strong> silver<br />

caps are oxidized and become Ag2O, which will<br />

initiate a gas-phase/solid reaction with<br />

formaldehyde.<br />

I. INTRODUCTION<br />

Porous materials are useful in a wide range <strong>of</strong><br />

applications,, for example, as membranes for<br />

separation and purification, as high surface area<br />

adsorbents, as solid supports for sensors and catalysts,<br />

as scaffolds for tissue engineering, and as lowdielectric<br />

constant materials for microelectronic<br />

devices. These materials have also been actively<br />

explored as photonic band gap (PBG) materials for<br />

use in optoelectronics and as new types <strong>of</strong><br />

mechanical materials that may exhibit negative<br />

Poisson ratios.<br />

Generally, colloidal crystals have been used as<br />

templates to form organic, inorganic, or metallic<br />

porous structures depending on <strong>the</strong> nature <strong>of</strong> <strong>the</strong><br />

applications. Unlike lithography based techniques<br />

that have shown mostly two-dimensional features, a<br />

colloidal self-assembly approach can produce<br />

complex and regular three-dimensional structures<br />

including channel-like, spherical, ellipsoidal, and<br />

rectangular shapes as well as more complex forms<br />

such as star shaped assemblies.<br />

This project involves <strong>the</strong> fabrication <strong>of</strong> ordered<br />

microporous cylinders by colloidal self-assembly[1]<br />

in microcapillaries. Microporous structures can<br />

enhance mass transport to and efficiency <strong>of</strong> catalysts.<br />

Such shape-controlled colloidal crystals are <strong>of</strong><br />

critical importance for practical use <strong>of</strong> bulk colloidal<br />

crystals.<br />

Bin Ren and Pr<strong>of</strong>. Ilona Kretzschmar<br />

Department <strong>of</strong> Chemical Enigneering<br />

City College <strong>of</strong> New York<br />

New York, USA<br />

Email: renbin83@gmail.com<br />

273<br />

II. EXPERIMENTAL<br />

The fabrication <strong>of</strong> <strong>the</strong> porous membranes in our work<br />

consists <strong>of</strong> four parts: monolayer assembly, silver<br />

coating deposition, cylindrical colloidal crystal<br />

assembly, and porous membrane fabrication. The<br />

first two steps are <strong>the</strong> preparation work for <strong>the</strong> last<br />

one, which provide half capped silver coatings on<br />

each <strong>of</strong> <strong>the</strong> spherical particle templates.<br />

A. MONOLAYER ASSEMBLY<br />

The monolayer <strong>of</strong> closed packed polystyrene(PS)<br />

particles is obtained on a glass slide by dragging a<br />

glass slide with a syringe pump from New Era<br />

Systems. Inc. The proper dragging speed is set<br />

according to <strong>the</strong> humidity and <strong>the</strong> size <strong>of</strong> PS<br />

particles. The deposiiton process only takes a few<br />

minutes.<br />

B. SILVER COATING DEPOSITION<br />

A bench top metal evaporator 308R system from Ted<br />

Pella Inc. is used in which <strong>the</strong> source materials<br />

(silver and titanium) are evaporated in a vacuum, and<br />

<strong>the</strong> vacuum allows vapor particles to travel directly<br />

to <strong>the</strong> target object (monolayers), where <strong>the</strong>y<br />

condense back to a solid state. The thickness <strong>of</strong><br />

coating is controlled by <strong>the</strong> deposition time.<br />

C. COLLOIDAL CRYSTAL ASSEMBLY<br />

This process is carried out in a 100 �m OD/50 �m ID<br />

microcapillary template. The PS particles are washed<br />

<strong>of</strong>f from <strong>the</strong> glass slide by sonication in deionized<br />

water. Uniform cylindrical colloidal crystals <strong>of</strong><br />

modified PS spheres are self-assembled inside <strong>the</strong><br />

microcapillary by capillary force. The water is<br />

evaporated from <strong>the</strong> open end <strong>of</strong> <strong>the</strong> capillary.<br />

D. FABRICATION OF POROUS MEMBRANE<br />

A polyurethane prepolymer is infiltrated into <strong>the</strong> asprepared<br />

PS colloidal crystals. The prepolymer is<br />

cured by UV exposure. The PS particles are removed<br />

from <strong>the</strong> particle polymer membrane by treatment


with organic solvents resulting in <strong>the</strong> formation <strong>of</strong><br />

inverse opal structures. Silver does not dissolve in <strong>the</strong><br />

organic solvents and cannot leave <strong>the</strong> pores due to<br />

<strong>the</strong> small size <strong>of</strong> connecting holes in an inverse opal.<br />

III. RESULTS<br />

The monolayer is easy to obtain, which is shown in<br />

Figure 1(left). Figure 1 (right) shows <strong>the</strong> surface <strong>of</strong><br />

silver capped particles as smooth right after silver<br />

deposition. However, <strong>the</strong> surface becomes rough<br />

shortly after because <strong>of</strong> silver cap oxidation.<br />

Figure 1: Optical microscope image <strong>of</strong> 2.5 �m PS<br />

monolayer (left), SEM image <strong>of</strong> 2.5 �m silver<br />

capped PS particle(right).<br />

Fierce vibration is employed when <strong>the</strong> silver<br />

capped particles are assembled. Because <strong>the</strong> silver<br />

capped particles are much heavier than <strong>the</strong><br />

uncapped sPS particles. The shinning parts in<br />

Figure 2 are <strong>the</strong> silver capped particles.<br />

Figure 2: Optical microscope image <strong>of</strong> sectioned<br />

cylindrical colloidal crystals with silver capped<br />

particles (a). Interface (b).<br />

Theoretically, in an evaporation-induced colloidal<br />

self-assembly, <strong>the</strong> thickness (or <strong>the</strong> number <strong>of</strong> layers)<br />

<strong>of</strong> deposited colloidal crystals on a substrate is<br />

proportional to <strong>the</strong> meniscus height, which is mainly<br />

related to <strong>the</strong> lateral capillary force. The lateral<br />

capillary force is given by <strong>the</strong> derivative:<br />

F ��d�W / dL<br />

(1)<br />

which yields:<br />

F �� 2 ��Q1Q2qK1(<br />

qL)<br />

( rk�L) (2)<br />

K1 is <strong>the</strong> modified Bessel function.<br />

In Figure 3 <strong>the</strong> interface between different sizes <strong>of</strong><br />

pores and <strong>the</strong> uniform close packing <strong>of</strong> pores are<br />

clearly visible in <strong>the</strong> SEM images. The dark lines<br />

274<br />

next to <strong>the</strong> porous membranes are PMMA<br />

microcapillary and PS template residues.<br />

Figure 3: SEM images <strong>of</strong> <strong>the</strong> polymeric inverted<br />

replica after <strong>the</strong> PS particles and microcapillaries<br />

have been removed by THF. All <strong>the</strong> images show <strong>the</strong><br />

interface between 10 �m and 2.5 �m sPS particles.<br />

In Figure 4, <strong>the</strong> inner part <strong>of</strong> <strong>the</strong> cylindrical colloidal<br />

crystal is visible and shows that <strong>the</strong> pores are<br />

arranged uniformly. This fur<strong>the</strong>r confirms that <strong>the</strong> PS<br />

particles with silver caps arranged into a closed<br />

packed fcc lattice throughout <strong>the</strong> capillary without<br />

forming a hollow core.<br />

Figure 4: SEM image <strong>of</strong> porous membrane<br />

containing silver caps.<br />

Future studies will explore <strong>the</strong> use <strong>of</strong> mixed<br />

TiO2/silver caps as catalyst material and study <strong>the</strong><br />

retention <strong>of</strong> formaldehyde in <strong>the</strong> membranes.<br />

ACKNOWLEDGMENTS<br />

This work was supported by <strong>the</strong> National Science<br />

Foundation under Award Number #0625753.<br />

REFERENCES<br />

[1] Moon, J. H.; Kim, S.; Yi, G-R.; Lee, Y-H.; Yang,<br />

S-M.; Langmuir 2004, 20, 2033-2035.


Presence <strong>of</strong> <strong>the</strong> Bacteria Legionella in Hot Water Distribution<br />

Systems<br />

Daniela Ocipova and Zuzana Vranayova<br />

Institute <strong>of</strong> Building and Environmental Engineering<br />

Technical University <strong>of</strong> Kosice<br />

Kosice, Slovak Republic<br />

Email: daniela.ocipova@tuke.sk, zuzana.vranayova@tuke.sk<br />

Abstract — The presence <strong>of</strong> <strong>the</strong> bacteria Legionella<br />

in water systems, especially in <strong>the</strong> hot water<br />

distribution system, represents in terms <strong>of</strong> health<br />

protection <strong>of</strong> inhabitants <strong>the</strong> crucial problem which<br />

can not be overlooked. The aim <strong>of</strong> my <strong>the</strong>sis is <strong>the</strong><br />

research focused on <strong>the</strong>oretical and experimental<br />

analysis <strong>of</strong> hot water distribution systems from <strong>the</strong><br />

point <strong>of</strong> view <strong>of</strong> microbiological risk on <strong>the</strong> basis <strong>of</strong><br />

confrontation between <strong>the</strong> <strong>the</strong>oretical analysis and<br />

syn<strong>the</strong>sis <strong>of</strong> ga<strong>the</strong>red information in conditions <strong>of</strong><br />

<strong>the</strong> Slovak Republic. There are a lot <strong>of</strong> guidelines<br />

and regulations developed in many individual<br />

countries for <strong>the</strong> design, operation, and<br />

maintenance <strong>of</strong> tap water systems to avoid <strong>the</strong><br />

growth <strong>of</strong> bacteria Legionella. The core <strong>of</strong> <strong>the</strong><br />

article consists in my investigation <strong>of</strong> Legionella<br />

contamination <strong>of</strong> hot water in a cross-sectional<br />

survey in Kosice, <strong>the</strong> Slovak Republic.<br />

I. INTRODUCTION<br />

Legionellosis is a collection <strong>of</strong> infections that<br />

emerged in <strong>the</strong> second half <strong>of</strong> <strong>the</strong> 20th century, and<br />

that are caused by Legionella pneumophila and<br />

related Legionella bacteria. Water is <strong>the</strong> major<br />

natural reservoir for legionella, and <strong>the</strong> bacteria are<br />

found worldwide in many different natural and<br />

artificial aquatic environments, such as cooling<br />

towers; water systems in hotels, homes, ships and<br />

factories; respiratory <strong>the</strong>rapy equipment; fountains;<br />

misting devices; and spa pools. About 20% <strong>of</strong> <strong>the</strong><br />

cases <strong>of</strong> legionellosis detected in Europe are<br />

considered to be travel-related; <strong>the</strong>se cases present a<br />

particular set <strong>of</strong> problems because <strong>of</strong> difficulties in<br />

identifying <strong>the</strong> source <strong>of</strong> infection. The severity <strong>of</strong><br />

legionellosis varies from mild febrile illness (Pontiac<br />

fever) to a potentially fatal form <strong>of</strong> pneumonia<br />

(Legionnaires’ disease) that can affect anyone, but<br />

principally affects those who are susceptible due to<br />

age, illness, immunosuppression or o<strong>the</strong>r risk factors,<br />

such as smoking [1].Outbreaks <strong>of</strong> disease have<br />

generally occurred when <strong>the</strong> concentrations in water<br />

systems have been high and aerosol has been<br />

produced. The aim <strong>the</strong>refore is to minimise <strong>the</strong><br />

possibility for infectious doses to be produced as a<br />

275<br />

result <strong>of</strong> operation <strong>of</strong> water systems. It is important<br />

that appropriate measures are taken to guard against<br />

conditions which may encourage Legionella<br />

multiplication.<br />

II. LEGISLATION<br />

In <strong>the</strong> last years particular parts <strong>of</strong> <strong>the</strong> ‘STN EN<br />

806 Specifications for installations inside<br />

buildings conveying water for human<br />

consumption’ have been accepted by <strong>the</strong> Slovak<br />

Republic. Installations have to be operated and<br />

maintained without having an adverse effect on<br />

conveying water for human consumption and<br />

water quality. Safety quality and suitability <strong>of</strong><br />

procedures accepted for safeguarding <strong>of</strong> system<br />

performance according to EN 806 and EN 1717<br />

should be regularly controlled. Conditions <strong>of</strong><br />

installations operation have to be compared with<br />

conditions <strong>of</strong> design and assembly in order to<br />

insure <strong>the</strong>ir functionality.<br />

Potable water is safeguarded by basic formula:<br />

� Cold water must remain cold<br />

� Hot water must remain hot<br />

� Water must not stagnate in water pipes<br />

longer than it is necessary<br />

III. AIMS AND METHODS<br />

To assess <strong>the</strong> potential public health impact <strong>of</strong><br />

Legionella colonization at domestic level, as well as<br />

public level, a descriptive multicentric study was<br />

undertaken to identify and qualify <strong>the</strong> levels <strong>of</strong> <strong>the</strong><br />

microorganism in a substantial number <strong>of</strong> Slovak<br />

domestic and public hot water samples.<br />

We addressed three specific aims:<br />

1. to estimate <strong>the</strong> frequency <strong>of</strong> Legionella<br />

colonization and severity <strong>of</strong> contamination at<br />

different levels<br />

2. to identify <strong>the</strong> potential risk factors for<br />

contamination relative to distribution systems<br />

and water characteristics<br />

3. to define relative role <strong>of</strong> each risk factor and<br />

suggest possible remediation.


Lastly, risk for legionellosis will be retrospectively<br />

evaluated by collecting information about pneumonia<br />

symptoms recorded by residents at buildings [2].<br />

A. SAMPLE COLLECTION<br />

From February to October 2006, a total <strong>of</strong> 46 water<br />

samples were collected from private homes, hospitals<br />

and boiler houses <strong>of</strong> Kosice, representative samples<br />

<strong>of</strong> Eastern Slovakia.The selection was made on <strong>the</strong><br />

basis <strong>of</strong> <strong>the</strong> water distribution systems inside <strong>the</strong><br />

town and buildings and heater types in each area.<br />

After we identified each building, we asked a random<br />

family, or a work collective to participate in <strong>the</strong><br />

study, i.e. to complete our questionnaire and give<br />

informed consensus for water collection. Laboratory<br />

examinations and Legionella analysis were made by<br />

<strong>the</strong> Regional Health Office – referential centre for<br />

potable water in Kosice. The hot water samples were<br />

drawn from <strong>the</strong> bathroom outlets in <strong>the</strong> case <strong>of</strong><br />

residential houses (shower heads or bathroom taps)<br />

in <strong>the</strong> sterile 1-L glass bottles after a short flow time<br />

(to eliminate cold water inside <strong>the</strong> tap or flexible<br />

shower pipe). To neutralize residual free chlorine,<br />

sodium thiuosulphate was added into sterile bottles<br />

for bacteriological analysis, whereas acid-preserved<br />

glass bottles were used for chemical determinations.<br />

The collection bottles were returned to <strong>the</strong> laboratory<br />

immediately after sampling for bacteriological<br />

examination by a Membrane filtration. Filters<br />

Millipore were used for 10 ml sample volumes. Adjusted<br />

samples were inoculated on <strong>the</strong> medium<br />

GVPC surface [3].<br />

B. POSITIVE SAMPLES<br />

Legionellas presence was detected in 8 samples out<br />

<strong>of</strong> analysed drinking water samples. Positive findings<br />

were recorded in 8 samples <strong>of</strong> PWH (potable water<br />

hot)In waters for human consumption (potable water<br />

cold - PWC) volume <strong>of</strong> legionellas were detected,<br />

from sporadic colonies <strong>of</strong> 20 CCU/100ml up to<br />

massive colonizations in <strong>the</strong> quantity 6700<br />

CCU/100ml per a sample. In water for human<br />

consumption (PWH) volume <strong>of</strong> legionellas were<br />

detected, from sporadic colonies <strong>of</strong> 200CCU/100ml<br />

up to massive colonizations in <strong>the</strong> quantity 14600<br />

CCU/100ml per a sample. Legionellas presence was<br />

detected in 8 samples <strong>of</strong> analysed PWH samples, i.e.<br />

in 17,4 %.<br />

C. RESULTS<br />

We repeated sampling after <strong>the</strong>rmal disinfection in<br />

contaminated places. After 12 days <strong>the</strong> level <strong>of</strong><br />

legionella colonies was almost <strong>the</strong> same as before<br />

this measure (Figure 1).<br />

276<br />

Figure 1: Legionella results – positive samples<br />

Much worse results were obtained in similar survey<br />

in Italy or Germany �2]. In this case 36 - 68 % <strong>of</strong><br />

samples were positive. In case <strong>the</strong> <strong>the</strong>rmal<br />

disinfection in contaminated places was not done <strong>the</strong><br />

concentration <strong>of</strong> bacterias will have an exponential<br />

character. The measures have proved that <strong>the</strong> <strong>the</strong>rmal<br />

disinfection is not a suitable system treatment. New<br />

strategies are tend to permanent disinfection due to<br />

<strong>the</strong> fact that spasmodic disinfection is not enough<br />

reliable to ensure <strong>the</strong> required standard.<br />

IV. CONCLUSION<br />

By collecting <strong>the</strong> samples we verify that <strong>the</strong>rmal<br />

disinfection is not <strong>the</strong> systematic solution and it is<br />

inevitable to find a new complex solution. Expenses<br />

on <strong>the</strong> elimination <strong>of</strong> Legionella from <strong>the</strong> water<br />

distribution systems are very high and <strong>the</strong> results are<br />

not <strong>of</strong>ten sufficient. Our control preventive measures<br />

have been in progress in order to find an effective<br />

way to suppress <strong>the</strong> spread <strong>of</strong> <strong>the</strong> Legionella bacteria<br />

in water systems. Never<strong>the</strong>less, it should be <strong>the</strong><br />

common goal <strong>of</strong> designers and operators to reduce<br />

<strong>the</strong> risk <strong>of</strong> Legionella bacteria in <strong>the</strong> installation<br />

inside buildings. To prevent tragic events it is<br />

obviously necessary to monitor <strong>the</strong> issue in <strong>the</strong> world<br />

and pay attention to precautionary regulations.<br />

ACKNOWLEDGMENTS<br />

This work was funded by NATO ESP.NUKR.CLG<br />

982978.<br />

REFERENCES<br />

[1] P. Borella et al. Environmental diffusion <strong>of</strong><br />

Legionella spp. and legionellosis frequency<br />

among patients with pneumonia: results <strong>of</strong> a<br />

multicentric Italian survey, 15;.493–503, 2003<br />

[2] F. J. Bartram et al.: LEGIONELLA and <strong>the</strong><br />

prevention <strong>of</strong> legionellosis, EWGLI,2007.<br />

[3] Z. Vranayová: Stav kontaminácie legionellou<br />

v rozvodoch teplej vody. Sanhyga, Bratislava<br />

SSTP st: 43 -52, October 2007


Development and Processing <strong>of</strong> a Multi-Filter Rotating Shadowband Radiometer<br />

Network for Distributed Monitoring <strong>of</strong> Aerosol Optical Depth: Comparisons Between<br />

Conventional Langley Regression and a Novel NASA-GISS Algorithm<br />

Miguel Bustamante, Fred Moshary, Barry Gross and Sam Ahmed<br />

NOAA-CREST Optical Remote Sensing Laboratory, The City College <strong>of</strong> New York, NY<br />

10031 and CUNY Graduate Center 365 Fifth Ave. NY, NY 10016<br />

Email: mbustamante@cuny.ccny.edu<br />

ABSTRACT<br />

Information about global distributions <strong>of</strong> aerosol<br />

optical thickness (AOT) is necessary to quantify <strong>the</strong><br />

aerosol radiative forcing. The development and<br />

processing <strong>of</strong> suitable radiometer networks is central<br />

to this effort. Modern robotic solar instruments for<br />

retrievals <strong>of</strong> aerosols optical depth and o<strong>the</strong>r<br />

microphysical parameter retrievals, such as <strong>the</strong><br />

CIMEL Sky Scanning Radiometers part <strong>of</strong> <strong>the</strong><br />

Aeronet Network are crucial to this effort but are<br />

unfortunately quite sparse and expensive. It has been<br />

proposed that <strong>the</strong> use <strong>of</strong> more cost effective devices<br />

likes <strong>the</strong> Multi-Filter Rotating Shadow-band<br />

Radiometer in a network might prove sufficient.<br />

Therefore it is crucial to develop appropriate<br />

algorithms that can be used in unattended processing<br />

which allows us to assess satellite retrieval<br />

performance.<br />

1 INTRODUCTION<br />

Atmospheric aerosols play an important role in <strong>the</strong><br />

Earth’s radiative budget [1] and affect <strong>the</strong> global<br />

climate through radiative forcing and chemical<br />

perturbations. Information about global distributions<br />

<strong>of</strong> aerosol optical thickness (AOT) is necessary to<br />

quantify <strong>the</strong> aerosol radiative forcing and can play an<br />

important role in improving atmosphere correction for<br />

land and ocean products. The used <strong>of</strong> modern robotic<br />

solar instruments for retrievals <strong>of</strong> aerosols optical<br />

depth, such as <strong>the</strong> Multi-Filter Rotating Shadow-band<br />

Radiometer [2] is crucial in <strong>the</strong> assessment <strong>of</strong> aerosol<br />

retrievals from satellites. As an example <strong>of</strong> <strong>the</strong><br />

usefulness <strong>of</strong> <strong>the</strong> MFRSR network we are developing,<br />

we plot in figure 1 a comparison <strong>of</strong> <strong>the</strong> MFRSR<br />

radiometer retrievals against <strong>the</strong> operational retrieval.<br />

We observe MODIS under-estimates AOD for<br />

Princeton (rural) and over-estimates for urban sites<br />

like NYC and MECC Brooklyn. We find that over<br />

urban areas <strong>the</strong>re is a strong bias in <strong>the</strong> retrieval <strong>of</strong><br />

277<br />

AOD from satellite instruments due to poor<br />

estimation due to surface reflection. This observation<br />

was crucial in trying to develop a better surface<br />

model <strong>of</strong> urban areas for better aerosol retrieval. For<br />

example, an overestimate <strong>of</strong> <strong>the</strong> aerosol retrieval<br />

points to an underestimate in <strong>the</strong> ground reflection If<br />

we only had available <strong>the</strong> CIMEL site at CCNY, we<br />

could not confirm our hypo<strong>the</strong>sis. However, our<br />

ground measurements at Princeton and MECC were<br />

performed with relatively inexpensive and portable<br />

MFRSR and <strong>the</strong> development <strong>of</strong> a suitable MFRSR<br />

network would be very advantageous. However,<br />

processing MFRSR data using calibrations method<br />

which are not robust affects <strong>the</strong> accuracy <strong>of</strong> <strong>the</strong><br />

measurements which are not at <strong>the</strong> level <strong>of</strong> Aeronet.<br />

Some calibration approach assumes stability in <strong>the</strong><br />

aerosol loading over <strong>the</strong> entire day. Evidently, <strong>the</strong><br />

operational method <strong>of</strong> calibration <strong>of</strong> <strong>the</strong>se instruments<br />

by Langley Regression is extremely difficult in<br />

regions where <strong>the</strong> aerosol loading varies significantly.<br />

However, a new processing technique which<br />

combines <strong>the</strong> use <strong>of</strong> both <strong>the</strong> direct and <strong>the</strong> diffuse<br />

components should lead to improved results. This<br />

method only assume that <strong>the</strong> aerosol spectral ratios<br />

(i.e. aerosol type) be uniform. To this end, we have<br />

� MODIS<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

0<br />

ME<br />

PR<br />

NYC<br />

0 0.2 0.4<br />

�<br />

aer<br />

0.6 0.8 1<br />

Figure 1 Comparison <strong>of</strong> MFRSR AOD vs. MODIS AOD<br />

between Sites: Princeton, Medgar Evers and CCNY for Sep-<br />

Oct 2007 period. Princeton is under estimated compared to<br />

CCNY and MECC.


Error<br />

begun a massive effort to outfit <strong>the</strong> metropolitan area<br />

with <strong>the</strong>se instruments and explore <strong>the</strong> processing<br />

performance <strong>of</strong> <strong>the</strong> GISS algorithm against <strong>the</strong> more<br />

traditional Langley Regression approach. We are<br />

interested in assessing <strong>the</strong> performance <strong>of</strong> <strong>the</strong><br />

MFRSR retrievals against <strong>the</strong> Aeronet CIMEL<br />

instrument and embark on using <strong>the</strong> network data to<br />

assess <strong>the</strong> diversity <strong>of</strong> aerosols over <strong>the</strong> New York<br />

Metropolitan area. At present, <strong>the</strong> network includes<br />

5 sites which will ultimately be extended to form<br />

both an urban and suburban ring which will allow us<br />

to look at transport between <strong>the</strong>se two rings for a<br />

better determination <strong>of</strong> aerosol sources.<br />

2. DATA ANALYSIS<br />

A new regression technique similar to <strong>the</strong> Langley<br />

approach but relying on <strong>the</strong> spectral stability <strong>of</strong><br />

aerosol extinction instead <strong>of</strong> optical depth stability<br />

was developed at NASA GISS [1,2]. The main<br />

point in <strong>the</strong> new scheme is <strong>the</strong> fact that <strong>the</strong><br />

calibration is based on <strong>the</strong> stability <strong>of</strong> aerosol type<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

Fractional Error<br />

0<br />

0 50 100 150 200 250 300 350<br />

Julian Date<br />

and not aerosol loading. That this is in general an<br />

improvement is seen in figure 2 which shows that <strong>the</strong><br />

normalized optical depth variability approximately<br />

three times larger than <strong>the</strong> optical depth ratio<br />

variability.<br />

3. INTERCOMPARISONS<br />

tau 8 70<br />

Ratio 8 70<br />

Figure 2 calculating <strong>the</strong> fractional error between AOD at 870 nm<br />

channel vs. <strong>the</strong> AOD ratios <strong>of</strong> Diffuse/Direct clearly shows an<br />

improvement <strong>of</strong> AOD.<br />

The most important parameter to assess is <strong>the</strong> aerosol<br />

optical depth retrieval in <strong>the</strong> visible channel. In figure<br />

3, we plot <strong>the</strong> retrieved AOD at 500 nm for both <strong>the</strong><br />

GISS processing (left panel) and <strong>the</strong> normal Langley<br />

regression approach. We note in that <strong>the</strong> GISS<br />

approach leads to less variance from <strong>the</strong> Aeronet<br />

baseline but <strong>the</strong>re is persistence over bias. This is<br />

symptomatic <strong>of</strong> <strong>the</strong> need to compensate for <strong>the</strong> optical<br />

278<br />

opacity [2] factor which seems to difficult over urban<br />

scenes. On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> Langley approach does<br />

not show bias but clearly produces larger errors.<br />

AOT500<br />

0.35<br />

0.3<br />

0.25<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

0<br />

-0.05<br />

-0.1<br />

Giss - CIMEL tau<br />

100 200<br />

Julian Day<br />

300<br />

4. CONCLUSIONS<br />

The processing <strong>of</strong> MFRSR data with <strong>the</strong> improved<br />

algorithm processing method based on calibration <strong>of</strong><br />

<strong>the</strong> long channel (870) using a combination <strong>of</strong> direct<br />

and diffuse radiation does not require good stability<br />

in <strong>the</strong> optical depth during <strong>the</strong> calibration stage.<br />

Instead, <strong>the</strong> calibration assumes only that <strong>the</strong> optical<br />

depth ratios are stable which we showed is a more<br />

robust assumption.<br />

5. AKNOWLEDGMENTS<br />

This work is supported by grants from NOAA<br />

#NA17AE1625 and NASA #NCC-1-03009. The views,<br />

opinions, and findings contained in this report are those <strong>of</strong><br />

<strong>the</strong> author(s) and should not be construed as an <strong>of</strong>ficial<br />

NOAA, NASA or U.S. Government position, policy, or<br />

decision.<br />

6. REFERENCES<br />

AOT500<br />

0.35<br />

0.25<br />

0.15<br />

0.05<br />

-0.05<br />

-0.15<br />

0 200<br />

Julian Day<br />

400<br />

[1] Alexandrov M., et al. “Remote Sensing <strong>of</strong><br />

Atmospheric Aerosols and Trace Gases by Means <strong>of</strong><br />

Multifilter Rotating Shadowband Radiometer”. Part<br />

I,II: J. Atmos. Sci., 59, pp524-542, pp544-566, 2002.<br />

[2] Alexandrov et. al.: ”Separation <strong>of</strong> fine and coarse<br />

aerosol modes in MFRSR data sets”. J.Geophys. Res.<br />

110, D13204, doi:10.1029/2004JD005226, 2005.<br />

0.3<br />

0.2<br />

0.1<br />

0<br />

-0.1<br />

Lang - CIMEL ave<br />

Figure 3 Inter-comparison <strong>of</strong> AOD retrieval at 500nm using a) novel GISS<br />

algorithm b) current Langley algorithm


��������������������������������������������������<br />

�<br />

���������������������������������������������������������<br />

������������������������������������������������<br />

��������������������������������<br />

����������������<br />

����������������������������������������<br />

�<br />

�<br />

��������� �� ���� ������ �������� ��� ���� �������<br />

��������� ��� ���� �������� ������� ������� ��� �������<br />

�������� �������� ��������� ���������� ��� ���� ��������<br />

����� ������������ ����� ���� ��� ���� ����� ����� ���<br />

������������������������������������������������<br />

�<br />

����������������������������������������������������<br />

����������������������������������������������������<br />

���� ����� ������� ������ ������� ���� ����� ������� ��� ����<br />

������������������� ��������� ����� �����������������<br />

�����������������������������������������������������<br />

������������������������������������������<br />

�<br />

���� ������������ �������� ���������� ��� ���� ������� ������<br />

����������������������������������������������������<br />

����� ������ ������� ������������ ������ ����������� ��� ����<br />

�����������������������������������������������������<br />

����������������������������������������������������<br />

�������� ����������� ����� ������ ������ ��������� ����<br />

������ �������� ���� ���� ������� ���������� ������ ���<br />

����� ������������ ����� ������������ ���� �������� ������<br />

����� ��������� ������ ���� ��������� ���� ����������<br />

����������������������������������������������������<br />

�������������������������������������������������<br />

������ ������������� ���� ��������� ���� ����� ����<br />

������������� ������������ �������� �������������������<br />

��������������������������������������������������������<br />

�����������������������������������������������������<br />

�����������������������������<br />

�<br />

���� �������� ���������� ������ ��� ������ ��� ������� �����<br />

���� ��������� ���������� ������ �������� ��� ����������<br />

������������� ��������� ���� ����� �������������� ���� �����<br />

���������������� �������� ���� ������� ��������� ����<br />

������������ ����� ��� ��������� ��� ���������� ���������<br />

��������������������������������������������������<br />

��� ����� ������� ����� ������ ����� ���� �������� ��������� ����<br />

�������������������������������������������������������<br />

���� ��� ����������������� ������� ��� ��� ��� ��������� ����<br />

����������� ���� ����������� �������� ������ ��� ��������<br />

����������������������������������������������������<br />

������������������������������������������������������<br />

���������������������������������������������������������<br />

���������� ���������������� ��� �������� ���� ������������<br />

����������������������������������������������������<br />

�� ������� ����� ������������� ���������� ���� ��������<br />

279<br />

������ ��� ������� ��� ����� ������ ������ ���� ��� ���� ����������<br />

����� ����������� �������� ��� ���� ��������� ��� ������� �����<br />

���� ���������� ������� ����������� ���� ��������� ���<br />

������������������������������������������������������<br />

���������������������������������������������������<br />

������������ ���������� ������������ ��������� ������<br />

����������� ��������� ������ ���� ������� ���������� ��<br />

������������������������������������������������������<br />

���� ���������� ��������� ���������� ��� ��������� ����<br />

������������������������������������������������������<br />

����������������������������������������������������<br />

�����������������������������������������������������<br />

��������������������������������������������������<br />

�<br />

�������������������������������������������������������<br />

������������ ������� ���� ��������� �������� ����� �����<br />

���������������������������������������������������������<br />

���������������������������������������������������<br />

��������������������������������������������������<br />

������������������������������������������������������<br />

������������������������������������������������������<br />

��������������������������������������������������������<br />

��� ���� �������� ������ ��� ��������� ��� ���� ��������<br />

�������������


�<br />

�<br />

�<br />

������������������������������������<br />

�����������<br />

���� ��� �������� ����������� ����������� ��� ���� �����<br />

��������������������� ��� ��������� �������� ���<br />

�����������������������<br />

�����������������������������������������������������<br />

����������������������������������������������<br />

�������������������������������������������<br />

�������������������������������������������������<br />

����� ���� ������ ���� ������� ��� ����������� ����<br />

���������������� �������������� ���� ����������<br />

�������������������������������������������<br />

����������<br />

�<br />

280


Random Bipartite Graphs and <strong>the</strong>ir Application to Cuckoo Hashing<br />

Reinhard Kutzelnigg and Michael Drmota (FacultyMentor)<br />

Institute <strong>of</strong> Discrete Ma<strong>the</strong>matics and Geometry<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {reinhard.kutzelnigg,michael.drmota}@tuwien.ac.at<br />

Abstract — Cuckoo hashing is a hash table data structure<br />

introduced in [1], that <strong>of</strong>fers constant worst case search time.<br />

This paper presents a precise average case analysis <strong>of</strong> Cuckoo<br />

hashing, and extends <strong>the</strong> analysis <strong>of</strong> [2]. In particular we determine<br />

<strong>the</strong> probability that <strong>the</strong> construction <strong>of</strong> a cuckoo hash table<br />

succeeds and give an upper bound for <strong>the</strong> construction time<br />

that is linear in <strong>the</strong> size <strong>of</strong> <strong>the</strong> table. As a fur<strong>the</strong>r contribution,<br />

we analyse modified versions <strong>of</strong> this algorithm with improved<br />

performance. Finally, we provide an asymptotic analysis <strong>of</strong> <strong>the</strong><br />

search costs <strong>of</strong> all this variants <strong>of</strong> cuckoo hashing and compare<br />

this results with <strong>the</strong> well known properties <strong>of</strong> double hashing<br />

and linear probing.<br />

I. INTRODUCTION<br />

Cuckoo hashing introduced in [1] uses two tables <strong>of</strong> size<br />

m to store n = m(1 − ε) keys, where ε ∈ (0, 1) holds.<br />

The algorithm is based on two hash functions h1 and h2,<br />

both map a key to a unique position in <strong>the</strong> first resp. second<br />

table. These are <strong>the</strong> only allowed storage locations<br />

<strong>of</strong> this key and, hence search operations need at most<br />

two look-ups. To insert a key x, we put it into its primary<br />

storage cell h1(x) <strong>of</strong> <strong>the</strong> first table. If this cell was<br />

empty, <strong>the</strong> insertion is complete. O<strong>the</strong>rwise, <strong>the</strong>re exists<br />

akeyy such that h1(x) =h1(y). We move this key to<br />

its secondary position h2(y) in <strong>the</strong> second table. If this<br />

cell was previously occupied too, we proceed with rearranging<br />

keys in <strong>the</strong> same way until we hit an empty cell.<br />

Obviously, <strong>the</strong>re are <strong>of</strong> course situations where we might<br />

enter an endless loop because <strong>the</strong> same keys are moved<br />

again and again. In such a case, <strong>the</strong> whole data structure<br />

is rebuild by using two new hash functions. As a strong<br />

point, this is a rare event [2, 3].<br />

Our analysis is based on <strong>the</strong> assumption, that <strong>the</strong> storage<br />

locations <strong>of</strong> <strong>the</strong> keys form a sequence <strong>of</strong> pairs <strong>of</strong> independent<br />

uniform random integers. If a rehash is necessary,<br />

we assume that all new hash values are independent<br />

from all previous attempts.<br />

Consider <strong>the</strong> bipartite graph possessing m labelled<br />

nodes <strong>of</strong> each type and n labelled edges, where each edge<br />

connects <strong>the</strong> two possible storage locations <strong>of</strong> a fixed key.<br />

This graph is called <strong>the</strong> cuckoo graph. The shape <strong>of</strong> it determines,<br />

whe<strong>the</strong>r <strong>the</strong> construction <strong>of</strong> <strong>the</strong> data structure<br />

is successful or not, since we enter an endless loop iff<br />

<strong>the</strong>re is a connected component with more than one cycle<br />

[2]. The fur<strong>the</strong>r analysis rests on a generating function<br />

approach to <strong>the</strong> cuckoo graph. We than apply a double<br />

saddle point method to obtain asymptotic results.<br />

281<br />

II. SUCCESS<br />

Theorem 1 ([3, 4]) Suppose that ε ∈ (0, 1) is fixed.<br />

Then <strong>the</strong> probability that a cuckoo hash <strong>of</strong> n = ⌊(1 −<br />

ε)m⌋ data points into two tables <strong>of</strong> size m succeeds, is<br />

equal to<br />

1 − (2ε2 − 5ε + 5)(1 − ε) 3<br />

12(2 − ε) 2ε3 �<br />

1 1<br />

+ O<br />

m m2 �<br />

. (1)<br />

�<br />

2<br />

This probability decreases to 3 + o(1), ifn equals m.<br />

III. ASYMMETRIC CUCKOO HASHING<br />

Assume that we choose <strong>the</strong> tables in such a way, that <strong>the</strong><br />

first table holds more memory cells than <strong>the</strong> second one<br />

(cp. [1]). Thus, we expect that <strong>the</strong> number <strong>of</strong> keys actually<br />

stored in <strong>the</strong> first table increases, what leads to improved<br />

search and insertion performance.<br />

Theorem 2 ([5]) Suppose that c ∈ [0, 1) and ε ∈ (1 −<br />

√ 1 − c 2 , 1) are fixed. Then, <strong>the</strong> probability that an asymmetric<br />

cuckoo hash <strong>of</strong> n = ⌊(1 − ε)m⌋ data points<br />

into two tables <strong>of</strong> size m1 = ⌊m(1 + c)⌋ respectively<br />

m2 =2m − m1 succeeds, is equal to<br />

1 −<br />

h(c, ε)<br />

m<br />

+ O<br />

� 1<br />

m 2<br />

�<br />

. (2)<br />

Note that it is possible to calculate <strong>the</strong> function h(c, ε),<br />

see [5]. The analysis reveals a major drawback <strong>of</strong> this<br />

modification. The algorithm requires a load factor less<br />

than √ 1 − c 2 /2 and this bound decreases if <strong>the</strong> asymmetry<br />

increases.<br />

IV. SIMPLIFIED CUCKOO HASHING<br />

Instead <strong>of</strong> using two separate tables, it is straightforward<br />

to “glue” <strong>the</strong>m toge<strong>the</strong>r and use one table <strong>of</strong> double size<br />

only (cp. [1]). Fur<strong>the</strong>r, both hash functions address <strong>the</strong><br />

whole table. As a result <strong>of</strong> this change, <strong>the</strong> probability<br />

that <strong>the</strong> first hash function hits an empty cell increases,<br />

hence we expect a better performance for search and insertion<br />

operations.<br />

Theorem 3 ([5]) Suppose that ε ∈ (0, 1) is fixed. Then,<br />

<strong>the</strong> probability that a simplified cuckoo hash <strong>of</strong> n =<br />

⌊(1 − ε)m⌋ data points into a table <strong>of</strong> size 2m succeeds,<br />

is equal to<br />

(5 − 2ε)(1 − ε)2<br />

1 −<br />

48ε3 �<br />

1 1<br />

+ O<br />

m m2 �<br />

. (3)


We now conclude that <strong>the</strong> success probability <strong>of</strong> simplified<br />

cuckoo hashing is slightly decreased compared to<br />

<strong>the</strong> standard algorithm, but <strong>the</strong> practical behaviour is almost<br />

identical in this aspect.<br />

V. SEARCH<br />

Similar to <strong>the</strong> analysis <strong>of</strong> linear probing and uniform<br />

probing in [6], we consider hashing without deletions.<br />

Theorem 4 ([5]) Under <strong>the</strong> assumptions <strong>of</strong> Theorem 2,<br />

assume that an asymmetric (resp. standard) cuckoo hash<br />

table has been constructed successfully. Then, <strong>the</strong> expected<br />

number <strong>of</strong> inspected cells <strong>of</strong> an successful search<br />

is asymptotically given by<br />

2 − 1+c<br />

2α<br />

�<br />

1 − e −2α/(1+c)�<br />

+ O � m −1� , (4)<br />

where α = n/(2m) denotes <strong>the</strong> load factor <strong>of</strong> <strong>the</strong> table.<br />

Fur<strong>the</strong>rmore, <strong>the</strong> expected number <strong>of</strong> steps <strong>of</strong> an unsuccessful<br />

search is asymptotically given by<br />

2 − e −2α/(1+c) + O � m −1� . (5)<br />

Theorem 5 ([5]) Under <strong>the</strong> assumptions <strong>of</strong> Theorem 3,<br />

consider a simplified cuckoo hash data structure. Then<br />

<strong>the</strong> expected number <strong>of</strong> inspected cells <strong>of</strong> an successful<br />

search is bounded between<br />

2−<br />

1 − e−α<br />

α<br />

+O � m −1� 1 − e−2α<br />

and 2−<br />

2α +O � m −1� ,<br />

(6)<br />

where α = n/(2m) denotes <strong>the</strong> load factor <strong>of</strong> <strong>the</strong> table.<br />

Fur<strong>the</strong>rmore, <strong>the</strong> expected number <strong>of</strong> steps <strong>of</strong> an unsuccessful<br />

search is given by 1+α.<br />

Figure 1 displays <strong>the</strong> asymptotic behaviour <strong>of</strong> a successful<br />

search, depending on <strong>the</strong> load factor α. For simplified<br />

cuckoo hashing, <strong>the</strong> grey area shows <strong>the</strong> span between<br />

<strong>the</strong> upper and lower bound. The corresponding<br />

curve is obtained experimentally. Fur<strong>the</strong>r, Figure 2 shows<br />

a plot according to an unsuccessful search. We conclude<br />

that simplified cuckoo hashing <strong>of</strong>fers <strong>the</strong> best average<br />

performance over all algorithms considered here, for all<br />

feasible load factors.<br />

VI. INSERTION<br />

Until now, no exact analysis <strong>of</strong> <strong>the</strong> insertion cost <strong>of</strong><br />

cuckoo hashing is known. However, it is possible to establish<br />

an upper bound.<br />

Theorem 6 ([4]) Under <strong>the</strong> assumptions <strong>of</strong> Theorem 2<br />

resp. 3, an upper bound <strong>of</strong> <strong>the</strong> number <strong>of</strong> expected memory<br />

accesses during <strong>the</strong> construction <strong>of</strong> a standard resp.<br />

simplified cuckoo hash table is given by<br />

min<br />

�<br />

4,<br />

− log(1 − 2α)<br />

2α<br />

�<br />

n + O(1), (7)<br />

where α =(1− ε)/2 denotes <strong>the</strong> load factor and <strong>the</strong><br />

constant implied by O(1) depends on α.<br />

282<br />

steps<br />

1.4<br />

1.3<br />

1.2<br />

1.1<br />

1<br />

�<br />

stand. cuckoo<br />

�<br />

�<br />

double hashing<br />

�<br />

simpl. cuckoo<br />

0 0.1 0.2 0.3 0.4<br />

�<br />

�<br />

�<br />

lin. probing<br />

asymm. cuckoo, c =0.3<br />

Figure 1: Comparison <strong>of</strong> successful search.<br />

steps lin. probing<br />

1.8<br />

1.6<br />

1.4<br />

1.2<br />

1<br />

stand. cuckoo<br />

double hashing<br />

asymm. cuckoo, c =0.3<br />

simpl. cuckoo<br />

0 0.1 0.2 0.3 0.4<br />

Figure 2: Comparison <strong>of</strong> unsuccessful search.<br />

Numerical results show that <strong>the</strong> expected performance is<br />

by far below this upper bound. Fur<strong>the</strong>r, we notice that<br />

<strong>the</strong> simplified version <strong>of</strong>fers better average performance<br />

for all investigated settings compared to o<strong>the</strong>r variants <strong>of</strong><br />

cuckoo hashing.<br />

ACKNOWLEDGMENTS<br />

Research supported by <strong>the</strong> Austrian Science Foundation<br />

FWF, project S9604 and by <strong>the</strong> EU FP6-NEST-<br />

Adventure Programme, Contract number 028875.<br />

REFERENCES<br />

[1] R. Pagh and F. F. Rodler. Cuckoo hashing. Journal<br />

<strong>of</strong> Algorithms, 51(2):122–144, 2004.<br />

[2] L. Devroye and P. Morin. Cuckoo hashing: Fur<strong>the</strong>r<br />

analysis. Information Processing Letters, 86(4):215–<br />

219, 2003.<br />

[3] R. Kutzelnigg. Bipartite random graphs and cuckoo<br />

hashing. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 4th Colloquium on<br />

Ma<strong>the</strong>matics and Computer Science, DMTCS, pages<br />

403–406, 2006.<br />

[4] Michael Drmota and Reinhard Kutzelnigg. A precise<br />

analysis <strong>of</strong> cuckoo hashing. preprint, <strong>2008</strong>.<br />

[5] R. Kutzelnigg. An improved version <strong>of</strong> cuckoo hashing:<br />

Average case analysis <strong>of</strong> construction cost and<br />

search operations. In <strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> 19th internatinal<br />

workshop on combinatorial algoritms, pages<br />

253–266, <strong>2008</strong>.<br />

[6] Donald E. Knuth. The Art <strong>of</strong> Computer Programming,<br />

Volume III: Sorting and Searching. Addison-<br />

Wesley, Boston, second edition, 1998.<br />

�<br />

�<br />

α<br />

α<br />


Adaptive hp-Finite Element Methods for Two-Dimensional<br />

Elasticity with Tresca Friction<br />

Philipp Dörsek and Jens Markus Melenk (Faculty Mentor)<br />

Institute for Analysis and Scientific Computing<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: philipp.doersek@tuwien.ac.at<br />

Abstract — We consider a frictional contact problem, given<br />

as a variational inequality <strong>of</strong> <strong>the</strong> second kind. Using a primaldual<br />

variational formulation, we analyse its discretisation by<br />

<strong>the</strong> hp-version <strong>of</strong> <strong>the</strong> finite element method (hp-FEM). With<br />

<strong>the</strong> aid <strong>of</strong> a novel hp-mortar projection operator, we obtain<br />

a priori error estimates for this discretisation that are explicit<br />

in <strong>the</strong> mesh width h and <strong>the</strong> approximation order p. We extend<br />

residual error indicators, which are known from <strong>the</strong> classical<br />

h-version FEM, to <strong>the</strong> hp-FEM and prove <strong>the</strong>ir reliability.<br />

These indicators are employed to steer an hp-adaptive algorithm.<br />

Numerical experiments for a 2D elasticity problem<br />

with Tresca friction demonstrate exponential convergence <strong>of</strong><br />

<strong>the</strong> hp-FEM, i. e., highly accurate solutions are obtained with<br />

very small computational effort.<br />

I. INTRODUCTION<br />

Ma<strong>the</strong>matical models and numerical simulations are an<br />

indispensable tool in science and engineering. In engineering,<br />

for example, <strong>the</strong>y can contribute tremendously<br />

to cost reduction in product design since <strong>the</strong> manufacture<br />

and <strong>the</strong> testing <strong>of</strong> costly prototypes can be limited to<br />

those that proved successful in <strong>the</strong> simulation.<br />

Typically, <strong>the</strong> ma<strong>the</strong>matical models employed are very<br />

complex and can <strong>the</strong>refore be tackled only numerically.<br />

The key to obtaining highly accurate and reliable numerical<br />

results are adaptive algorithms. The purpose <strong>of</strong> <strong>the</strong><br />

<strong>the</strong>sis [1] is <strong>the</strong> development <strong>of</strong> such an algorithm for<br />

a class <strong>of</strong> problems that arise in technical applications,<br />

namely, a friction problem.<br />

II. THE FRICTION PROBLEM<br />

We want to simulate an elastic body that has frictional<br />

contact with a fixed object. While <strong>the</strong> behaviour <strong>of</strong> elastic<br />

bodies can be described by partial difference equations,<br />

<strong>the</strong> presence <strong>of</strong> frictional contact leads to a partial<br />

differential inequality, more precisely, a variational inequality<br />

<strong>of</strong> <strong>the</strong> second kind. The transition from an equation<br />

to an inequality leads to several difficulties in <strong>the</strong> numerical<br />

simulation. For example, while <strong>the</strong> Newton algorithm<br />

is <strong>the</strong> standard and well-established tool for solving<br />

nonlinear system <strong>of</strong> equations, <strong>the</strong> presence <strong>of</strong> a nondifferentiable<br />

functional in <strong>the</strong> variational inequalities considered<br />

here necessitates very different algorithms. One<br />

283<br />

approach is <strong>the</strong> primal “energy minimising” method described<br />

in detail in [2]. We focus, however, on a different<br />

idea: The special structure <strong>of</strong> <strong>the</strong> nondifferentiable functional<br />

allows one to derive a so-called primal-dual formulation<br />

leading to a saddle point problem as described<br />

in [3]. This is numerically attractive, as, by first solving<br />

a linear problem in <strong>the</strong> interior <strong>of</strong> <strong>the</strong> elastic body to construct<br />

<strong>the</strong> Schur complement <strong>of</strong> <strong>the</strong> system matrix, we<br />

only have to solve a single variational inequality <strong>of</strong> <strong>the</strong><br />

first kind on <strong>the</strong> contact boundary, which is a relatively<br />

small problem.<br />

III. CONVERGENCE<br />

In linear FEM, high order methods, that is, polynomials<br />

<strong>of</strong> high degree, have been successfully used to obtain<br />

accurate solutions with small computational effort.<br />

The standard approach in finite element methods is to<br />

use conforming methods, that is, <strong>the</strong> discrete solutions<br />

are required to be admissible for <strong>the</strong> continuous problem.<br />

In contrast to h-version methods, which have been<br />

in use since <strong>the</strong> 1980s (see [4]), it is not clear how to<br />

achieve this for polynomials <strong>of</strong> higher degree, which are<br />

used in hp-finite element methods, as we have to discretise<br />

a pointwise constraint. This can be explained as follows:<br />

Let K be <strong>the</strong> set <strong>of</strong> functions defined on <strong>the</strong> interval<br />

(0, 1) that are bounded pointwise by 1. It is easy to check<br />

whe<strong>the</strong>r a piecewise affine function is an element <strong>of</strong> K –<br />

it suffices to check <strong>the</strong> constraint in <strong>the</strong> nodes. But already<br />

for piecewise quadratic functions, it is not possible<br />

to ascertain membership in K by checking <strong>the</strong> pointwise<br />

constraint in a few a priori known points. Thus, if one<br />

wishes to use higher order methods, one has to consider<br />

non-conforming approximations, that is, <strong>the</strong> discrete set<br />

ˆK is no longer a subset <strong>of</strong> <strong>the</strong> continuous set K.<br />

This raises <strong>the</strong> question <strong>of</strong> how to define ˆ K. In [5],<br />

it was shown that an effective approach is to relax <strong>the</strong><br />

pointwise constraint by requiring <strong>the</strong> boundedness only<br />

at certain quadrature points. Using a <strong>the</strong>orem due to<br />

Glowinski (see [4]), it is <strong>the</strong>n possible to show convergence.<br />

In addition to <strong>the</strong> abstract convergence obtained by <strong>the</strong><br />

<strong>the</strong>orem cited above, it is essential to give estimates <strong>of</strong><br />

<strong>the</strong> rate <strong>of</strong> convergence, assuming that <strong>the</strong> solution <strong>of</strong>


<strong>the</strong> continuous problem is sufficiently regular. A result<br />

due to Haslinger (see [3]) shows that, as is typical for<br />

mixed finite element methods, such an estimate has to<br />

be based on an inf-sup condition. Using methods from<br />

functional analysis and Sobolev spaces, we were able to<br />

derive a non-uniform inf-sup condition in [1, Theorem<br />

3.33], which gives a better bound than a corresponding<br />

result in [6]. Finally, we obtain an error estimate which<br />

shows that, assuming that <strong>the</strong> meshes resolve all singularities,<br />

we can expect exponential convergence rates.<br />

IV. ADAPTIVITY<br />

For <strong>the</strong> successful application <strong>of</strong> hp-finite element methods,<br />

it is essential to have a mesh resolving all features<br />

<strong>of</strong> <strong>the</strong> solution such as singularities arising, for example,<br />

at <strong>the</strong> transition from sliding to sticking on <strong>the</strong> contact<br />

boundary. Since <strong>the</strong>se transition points are unknown,<br />

adaptivity is essential. Adaptive algorithms are steered<br />

by local error indicators. The error indicator employed<br />

error estimate<br />

1000<br />

100<br />

10<br />

1<br />

100 1000<br />

displacement degrees <strong>of</strong> freedom<br />

10000<br />

uniform h-FEM<br />

adaptive h-FEM<br />

uniform p-FEM<br />

adaptive hp-FEM<br />

Figure 1: Error plot for a model friction problem<br />

in this work is <strong>of</strong> residual type, generalised appropriately<br />

from <strong>the</strong> version for <strong>the</strong> h-FEM as given in [7] to <strong>the</strong><br />

present hp-context.<br />

The key issue in hp-adaptivity, to decide whe<strong>the</strong>r to<br />

do an h-refinement, more appropriate for a singular, or<br />

a p-refinement, more appropriate for a regular solution.<br />

This decision is realised as follows: We estimate <strong>the</strong><br />

local smoothness <strong>of</strong> <strong>the</strong> solution by expanding <strong>the</strong> approximate<br />

solution into a Legendre series. Depending<br />

on <strong>the</strong> decay rate <strong>of</strong> <strong>the</strong> coefficients in this expansion,<br />

calculated by a least squares method, we do an h- or<br />

a p-refinement. This method, analysed in [8], yields<br />

very good empiric convergence rates in our numerical<br />

experiments, and can thus be recommended for practical<br />

computations: In particular, as can be seen in Figure 1,<br />

<strong>the</strong> hp-adaptive method delivers much more accurate results<br />

for comparable problem sizes than uniform and hadaptive<br />

methods, and yields exponential convergence.<br />

284<br />

Figure 2: Deformation for a model friction problem<br />

REFERENCES<br />

[1] Philipp Dörsek. hp-Finite Element Methods for Variational<br />

Inequalities. Master’s <strong>the</strong>sis, Vienna University<br />

<strong>of</strong> Technology, Sep <strong>2008</strong>.<br />

[2] Ralf Kornhuber. Adaptive monotone multigrid methods<br />

for nonlinear variational problems. Advances<br />

in Numerical Ma<strong>the</strong>matics. B. G. Teubner, Stuttgart,<br />

1997.<br />

[3] I. Hlaváček, J. Haslinger, J. Nečas, and J. Lovíˇsek.<br />

Solution <strong>of</strong> variational inequalities in mechanics,<br />

volume 66 <strong>of</strong> Applied Ma<strong>the</strong>matical Sciences.<br />

Springer-Verlag, New York, 1988. Translated from<br />

<strong>the</strong> Slovak by J. Jarník.<br />

[4] Roland Glowinski. Numerical methods for nonlinear<br />

variational problems. Springer Series in Computational<br />

Physics. Springer-Verlag, New York, 1984.<br />

[5] Matthias Maischak and Ernst P. Stephan. Adaptive<br />

hp-versions <strong>of</strong> BEM for Signorini problems. Appl.<br />

Numer. Math., 54(3-4):425–449, 2005.<br />

[6] Padmanabhan Seshaiyer and Manil Suri. Uniform<br />

hp convergence results for <strong>the</strong> mortar finite element<br />

method. Math. Comp., 69(230):521–546, 2000.<br />

[7] Weimin Han. A posteriori error analysis via duality<br />

<strong>the</strong>ory, volume 8 <strong>of</strong> Advances in Mechanics<br />

and Ma<strong>the</strong>matics. Springer-Verlag, New York, 2005.<br />

With applications in modeling and numerical approximations.<br />

[8] T. Eibner and J. M. Melenk. An adaptive strategy for<br />

hp-FEM based on testing for analyticity. Comput.<br />

Mech., 39(5):575–595, 2007.


Thermal Bridge in Detail <strong>of</strong> a Gully in a Warm Flat Ro<strong>of</strong><br />

Petr Zahradník and František Kulhánek (Faculty Mentor)<br />

Department <strong>of</strong> Building Structures<br />

Czech Technical University in Prague<br />

Prague, The Czech Republic<br />

Email:petr.zahradnik@fsv.cvut.cz and Frantisek.kulhanek@fsv.cvut.cz<br />

Abstract — Details are <strong>the</strong> most problematic parts<br />

<strong>of</strong> car park ro<strong>of</strong>s. These details cause <strong>the</strong>rmal<br />

bridges – anchoring, perforating elements, dilatation<br />

etc. In this document, <strong>the</strong>re are two different<br />

solutions <strong>of</strong> a gully described. Both badly and well<br />

designed details from a <strong>the</strong>rmal-bridge-solutionviewpoint.<br />

Linear iso<strong>the</strong>rms show <strong>the</strong> risk <strong>of</strong> condensation<br />

on surfaces <strong>of</strong> <strong>the</strong> details. Final calculation<br />

is compared to <strong>the</strong> Czech Standard �SN 73<br />

0540 Thermal protection <strong>of</strong> buildings and is evaluated<br />

in relation to durability and reliability.<br />

I. INTRODUCTION<br />

Almost all problems with durability in ro<strong>of</strong>s are<br />

located in details <strong>of</strong> it. One <strong>of</strong> <strong>the</strong> biggest problem<br />

besides leakage is condensation and, as a result, fungi<br />

growth. Mostly we can see it around details. In following<br />

article, differences between two solutions are<br />

described (both, however, quite usual).<br />

II. OBJECTIVENESS OF AN ANALYSIS<br />

Due to objectivness reasons, both variants <strong>of</strong> a solution<br />

<strong>of</strong> a gully are calculated with <strong>the</strong> same ro<strong>of</strong><br />

layer-composition. Ro<strong>of</strong> with foamglass <strong>the</strong>rmal<br />

insulation has been used as one <strong>of</strong> <strong>the</strong> most safe types<br />

<strong>of</strong> ro<strong>of</strong> construction [1], suitable for example for<br />

traffic public use. 220 mm <strong>of</strong> foamglass has been<br />

used (corresponds to recommended U-value in <strong>the</strong><br />

Czech Republic [2], values in Table 1).<br />

Basic requirements <strong>of</strong> �SN 73 0540 [2] value<br />

U-value required 0,24 W/(m 2 .K)<br />

U-value recommended 0,16 W/(m 2 minimal internal surface temperature (at<br />

.K)<br />

this temperature, max.80% relative<br />

humidity is guaranteed)<br />

13,57 o C<br />

Table 1: Basic Standard [2] requirements<br />

O<strong>the</strong>r conditions used in <strong>the</strong> calculation are described<br />

in Table 2.<br />

285<br />

Conditions used in a calculation value<br />

external design temperature �e o<br />

-15 C<br />

internal design air-temperature �ai o<br />

21 C<br />

external design relative air-humidity �e 84%<br />

internal design air-temperature �i 50%<br />

Table 2: Conditions used in calculation based on [2]<br />

III. NUMERIC ANALYSIS OF TWO VARI-<br />

ANTS OF A GULLY DETAIL<br />

Numeric analysis <strong>of</strong> this <strong>the</strong>rmal bridge has been<br />

done with use <strong>of</strong> Czech calculation tool Cube 3D [3].<br />

Three-dimensional analysis is necessary in this case<br />

due to non-linearity <strong>of</strong> such a bridge.<br />

Analysis shows linear iso<strong>the</strong>rm as a graphic result.<br />

Important value in this case is temperature <strong>of</strong><br />

13,57 o C, which is border-value for risk <strong>of</strong> vapour<br />

condensation and fungi growth.<br />

A. UNSUITABLE, BUT USUAL SOLUTION OF A<br />

GULLY DETAIL<br />

Results <strong>of</strong> this detail solution will be compared to<br />

ano<strong>the</strong>r, more sophisticated, solution.<br />

There are two main mistakes in this solution:<br />

1. <strong>the</strong>re is no additional <strong>the</strong>rmal insulation layer<br />

around <strong>the</strong> pipe;<br />

2. <strong>the</strong> gully and a part <strong>of</strong> <strong>the</strong> pipe have been directly<br />

buried in concrete.<br />

Figure 1: Unsuitable, but usual detail <strong>of</strong> a gully


Figure 2: Linear Iso<strong>the</strong>rm <strong>of</strong> a gully<br />

Main problem <strong>of</strong> this solution is very low internal<br />

surface temperature, which is 9,96 ºC. It causes water<br />

vapour condensation and supports expectable fungi<br />

growth. This solution does not dramatically fulfil<br />

Standard [1] requirement. Due to missing dilatation<br />

between pipe and <strong>the</strong> concrete ceiling, we can expect<br />

cracking and crumbling <strong>of</strong> <strong>the</strong> concrete and deformation<br />

<strong>of</strong> <strong>the</strong> pipe.<br />

This solution is unacceptable from both energetic<br />

and hygienic viewpoint.<br />

B. SUITABLE SOLUTION OF A GULLY DETAIL<br />

Thanks to well done, continuous <strong>the</strong>rmal insulation<br />

layer we can expect much higher surface temperature.<br />

1. we calculate 60mm thick insulation around<br />

<strong>the</strong> pipe. The insulation makes (1) dilatation<br />

space for any volume changes in construction<br />

and (2) acoustic barrier. There should be at<br />

least 0,5m <strong>of</strong> an insulation on <strong>the</strong> pipe.<br />

2. <strong>the</strong> gully is located to an insulation “ring”,<br />

which is made <strong>of</strong> foamglass;<br />

3. load-bearing part <strong>of</strong> <strong>the</strong> detail – covering grid<br />

– is embedded to <strong>the</strong> concrete, which eliminates<br />

any external forces by <strong>the</strong> traffic on <strong>the</strong><br />

gully.<br />

Figure 3: Suitable detail <strong>of</strong> a gully<br />

286<br />

Figure 4: Linear Iso<strong>the</strong>rm <strong>of</strong> a gully<br />

This variant <strong>of</strong> <strong>the</strong> detail already fulfils Standard [1]<br />

requirements according to internal surface temperature<br />

and risk <strong>of</strong> fungi growth. The lowest surface<br />

temperature (in area <strong>of</strong> connection <strong>of</strong> a pipe to a<br />

ceiling) is calculated at 18,36 ºC which is higher than<br />

required (13,57 ºC).<br />

IV. CONCLUSION<br />

Due to badly designed or made detail, we can expect<br />

not only unreasonable energy losses, but risk <strong>of</strong> illnesses<br />

in case <strong>of</strong> low internal surface temperature<br />

that can cause fungi growth, as well. Degradation <strong>of</strong><br />

construction and decrease <strong>of</strong> its durability is additional<br />

negative effect.<br />

At commonly made details like gully, described in<br />

previous text, we can find serious <strong>the</strong>rmal bridges.<br />

Some <strong>of</strong> those solutions are unacceptable from both<br />

energetic and hygienic viewpoint. Obviously, it has<br />

big influence on durability and reliability <strong>of</strong> a construction.<br />

Well designed detail and its correct application, influence<br />

energy performance more, <strong>the</strong> better insulation<br />

<strong>the</strong> building has.<br />

ACKNOWLEDGMENTS<br />

This article has been made with support <strong>of</strong> Research<br />

Project MSM 6840770001 Reliability, Optimalisation<br />

and Durability <strong>of</strong> Building Materials and Construction.<br />

REFERENCES<br />

[1] Erfahrungen mit Foamglas an wärmegedämmten<br />

Parkdächern und Verkehrsflächen in Neubau und<br />

Sanierung, Pittsburgh Corning Europe, 2004;<br />

[2] Standard �SN 73 0540 Thermal Protection <strong>of</strong><br />

Buildings (April 2007);<br />

[3] S<strong>of</strong>tware Cube 3D, Svoboda S<strong>of</strong>tware, Kladno,<br />

The Czech Republic.


Development <strong>of</strong> Novel Flow Sensors using SPICE Simulations<br />

Samir �erimovi� 1,2 , Almir Tali� 1,2 , Franz Keplinger 2<br />

1 FISS, Austrian Academy <strong>of</strong> Sciences, Wiener Neustadt, Austria<br />

2 ISAS, Vienna University <strong>of</strong> Technology, Vienna, Austria<br />

Email: {samir.cerimovic,almir.talic,franz.keplinger}@tuwien.ac.at<br />

Abstract — A novel control loop scheme for calorimetric<br />

flow sensors combining an electronic twostate<br />

controller with <strong>the</strong> electro-<strong>the</strong>rmal transduction<br />

system is presented. For a convenient controller<br />

design, we developed a comprehensive SPICE<br />

model <strong>of</strong> <strong>the</strong> <strong>the</strong>rmal system able to fully cover its<br />

static and dynamic behavior. Such on-<strong>of</strong>f control<br />

schemes can be optimized for a specified flow range<br />

and by that outperform classical analogue controllers.<br />

Experiments confirm that this approach is a<br />

convenient method for <strong>the</strong> development <strong>of</strong> optimized<br />

controllers for any calorimetric flow transducer<br />

and specific measurement tasks.<br />

I. INTRODUCTION<br />

A typical calorimetric flow sensor consists <strong>of</strong> a<br />

miniaturized heat source and spatially separated<br />

temperature sensors (<strong>the</strong>rmistors), all embedded in a<br />

thin membrane (Figure 1). The sensor relies on flow<br />

dependent heat transfer altering <strong>the</strong> temperature<br />

distribution near <strong>the</strong> heater [1]. Operating it with<br />

constant heating power, a high sensitivity is feasible<br />

only within a limited flow range. Due to efficient<br />

convective cooling at higher flow rates <strong>the</strong> output<br />

characteristic becomes ambiguous. In order to avoid<br />

this ambiguity a constant temperature difference<br />

between <strong>the</strong> membrane and <strong>the</strong> fluid is necessary.<br />

This can be achieved by controlling <strong>the</strong> heater voltage<br />

with an electronic controller [2]. Figure 2 shows<br />

<strong>the</strong> measured sensor output characteristic for both<br />

operational modes. The output signal is proportional<br />

to <strong>the</strong> temperature difference <strong>of</strong> <strong>the</strong> <strong>the</strong>rmistors. In<br />

<strong>the</strong> constant over-temperature mode a standard PI<br />

controller was applied, resulting in an extended<br />

measurement range but poorer sensitivity at higher<br />

flow rates. In order to achieve better sensitivity, we<br />

Figure 1: Sensor layout with two temperature<br />

sensors (MT1, MT2) and a heater (H).<br />

287<br />

investigated <strong>the</strong> application <strong>of</strong> a two-state controller.<br />

For this purpose a SPICE model <strong>of</strong> <strong>the</strong> sensor was<br />

developed using PSpice as simulation tool [3].<br />

II. SPICE MODELING<br />

In order to obtain <strong>the</strong> model parameters <strong>the</strong> dependence<br />

<strong>of</strong> stationary and dynamic sensor properties on<br />

flow velocity was experimentally analyzed. Figure 3<br />

depicts <strong>the</strong> SPICE model diagram <strong>of</strong> <strong>the</strong> sensor electro-<strong>the</strong>rmal<br />

transduction system. The input variables<br />

are <strong>the</strong> heater voltage UH, <strong>the</strong> ambient temperature<br />

�LAB B and <strong>the</strong> flow velocity v. After <strong>the</strong> steady state<br />

value is calculated <strong>the</strong> transient behavior is approximated<br />

using <strong>the</strong> sequence <strong>of</strong> low pass filters. The<br />

time constants <strong>of</strong> <strong>the</strong> last two <strong>of</strong> <strong>the</strong>m depend on <strong>the</strong><br />

flow velocity. Two additional <strong>the</strong>rmistors (ST1, ST2)<br />

measure <strong>the</strong> ambient temperature. The output variables<br />

are <strong>the</strong> <strong>the</strong>rmistor resistances. Using this SPICE<br />

model, it is possible to investigate <strong>the</strong> interaction <strong>of</strong><br />

sensor and its evaluation and controlling circuit in<br />

order to find <strong>the</strong> best controller parameters. Figure 4<br />

shows <strong>the</strong> scheme <strong>of</strong> a novel operational mode using<br />

two-state closed loop controller which keeps <strong>the</strong><br />

average excess temperature (��) <strong>of</strong> <strong>the</strong> membrane<br />

constant. The on-<strong>of</strong>f-controller establishes a train <strong>of</strong><br />

voltage pulses at <strong>the</strong> heater. This approach <strong>of</strong>fers <strong>the</strong><br />

quotient <strong>of</strong> high to low pulse duration (TH/TL)<br />

as<br />

output quantity, in addition to <strong>the</strong> temperature difference<br />

signal (Uout).<br />

III. RESULTS<br />

The simulation results are depicted in Figure 5. Due<br />

to convective cooling, <strong>the</strong> pulse duration increases<br />

Figure 2: Relative output signal <strong>of</strong> <strong>the</strong> flow sensor<br />

with and without analogue controller.


U H<br />

v<br />

f ( U , � , v)<br />

H LAB<br />

steady state<br />

� LAB<br />

f ( U , � , v)<br />

H LAB<br />

steady state<br />

v<br />

and <strong>the</strong> pulse gap becomes smaller with increasing<br />

flow velocity. On <strong>the</strong> o<strong>the</strong>r hand, reducing <strong>the</strong> amplitude<br />

<strong>of</strong> <strong>the</strong> heating pulse (Uh) at constant velocity<br />

evokes <strong>the</strong> same effect. In order to keep <strong>the</strong> average<br />

excess temperature (��) <strong>of</strong> <strong>the</strong> membrane constant,<br />

<strong>the</strong> controller must increase <strong>the</strong> average heating<br />

power and hence <strong>the</strong> quotient <strong>of</strong> high to low pulse<br />

duration (TH/TL) increases too. Thus, altering <strong>the</strong><br />

amplitude <strong>of</strong> <strong>the</strong> heating pulse one can influence <strong>the</strong><br />

slope <strong>of</strong> <strong>the</strong> output characteristic in order to achieve<br />

better sensitivity.<br />

The measurement results for different amplitudes <strong>of</strong><br />

<strong>the</strong> heating pulses (Uh) are presented in Figure 6. The<br />

course <strong>of</strong> <strong>the</strong> output characteristic is in good agreement<br />

with simulated results. The slight difference in<br />

absolute values <strong>of</strong> <strong>the</strong> heating pulse amplitude results<br />

from non-ideal circuit components.<br />

In addition to <strong>the</strong> ratio <strong>of</strong> high to low pulse duration<br />

(TH/TL), <strong>the</strong> temperature difference signal can be<br />

used too, yielding <strong>the</strong> similar output characteristic as<br />

obtained with PI-Controller (Figure 2). For low flow<br />

velocities, this signal provides <strong>the</strong> highest sensitivity<br />

whereas <strong>the</strong> ratio TH/TL becomes <strong>the</strong> preferable output<br />

quantity at higher flow rates.<br />

IV. CONCLUSION<br />

TP TP TP<br />

v v<br />

TP TP TP TP<br />

v v<br />

We presented a novel control loop scheme combining<br />

an electronic two-state controller with <strong>the</strong> electro<strong>the</strong>rmal<br />

transduction system <strong>of</strong> <strong>the</strong> micromachined<br />

calorimetric flow sensors. Closed loop control <strong>of</strong> <strong>the</strong><br />

sensor membrane excess temperature with <strong>the</strong> twostate<br />

controller results in a pulse form <strong>of</strong> <strong>the</strong> heater<br />

� th2<br />

� LAB<br />

flow direction<br />

� LAB<br />

Figure 3: Block diagram <strong>of</strong> <strong>the</strong> SPICE model <strong>of</strong><br />

<strong>the</strong> sensor electro-<strong>the</strong>rmal transduction system.<br />

~�� ref<br />

~��<br />

-<br />

Comparator Amplifier<br />

G<br />

Transducer<br />

membrane<br />

Heater<br />

Flow velocity<br />

Figure 4: Scheme <strong>of</strong> a novel operational mode<br />

using two-state closed loop controller.<br />

� th1<br />

Thermistors<br />

ST2<br />

MT2<br />

MT1<br />

ST1<br />

Output quantities:<br />

Uout ~ �th2 - �th1<br />

288<br />

Figure 5: Simulated output characteristic.<br />

Figure 6: Measured output characteristic.<br />

voltage. This approach <strong>of</strong>fers <strong>the</strong> quotient <strong>of</strong> high to<br />

low pulse duration (TH/TL) as output quantity in<br />

addition to <strong>the</strong> temperature difference signal. The<br />

slope <strong>of</strong> output characteristic and hence <strong>the</strong> flow<br />

sensitivity can be optimized for a certain flow range<br />

through suitable selection <strong>of</strong> <strong>the</strong> heating voltage<br />

amplitude.<br />

ACKNOWLEDGMENTS<br />

We gratefully acknowledge partial financial<br />

support by <strong>the</strong> Austrian Science Fund FWF<br />

(research grant L234-N07).<br />

REFERENCES<br />

[1] N. Sabaté, J. Santander, L. Fonseca, I. Gràcia, C.<br />

Cané, Multi-range silicon micromachined flow<br />

sensor, Sensors and Actuators A 110 (2004)<br />

282–288.<br />

[2] A. Glaninger, A. Jachimowicz, F. Kohl, R.<br />

Chabicovsky, G. Urban, Wide range semiconductor<br />

flow sensors, Sensors and Actuators A 85<br />

(2000) 139–146.<br />

[3] PSpice (Cadence) is based on SPICE circuit<br />

simulation s<strong>of</strong>tware developed by UC Berkeley.


Pre-heating Vegetable Oil as a Sustainable Energy<br />

Development Solution : A Bi<strong>of</strong>uel-Engine<br />

Longevity Test<br />

Amir Hossein Nosrat, Mostafa Morovati, Martin Nolan, and Paul A.West (Faculty Mentor)<br />

Grove School <strong>of</strong> Engineering<br />

City College <strong>of</strong> New York<br />

New York, USA<br />

Email: sase@ccny.cuny.edu<br />

Abstract — In many parts <strong>of</strong> <strong>the</strong> world,<br />

including Africa and India, diesel engines are<br />

relied on heavily for mechanical power, which<br />

is used to grind grain, pump water, and<br />

generate electricity. In <strong>the</strong>se parts <strong>of</strong> <strong>the</strong><br />

world, single cylinder engines like <strong>the</strong> Listertype<br />

engine are popular for <strong>the</strong>ir reliability<br />

and longevity. These communities, however,<br />

<strong>of</strong>ten have limited access to diesel, driving up<br />

<strong>the</strong> costs and burden <strong>of</strong> running <strong>the</strong>se engines.<br />

The Bi<strong>of</strong>uel-Engine Longevity Test (BELT)<br />

investigates Jatropha oil as a sustainable<br />

energy solution for <strong>the</strong>se rural communities..<br />

I. INTRODUCTION<br />

The most common stationary diesel engines in<br />

rural Africa and India are single-cylinder, low-<br />

RPM designs similar to <strong>the</strong> British Lister<br />

engine. They are older, purely mechanical<br />

designs popular for <strong>the</strong>ir reliability, longevity<br />

and inexpensive price. Common engine sizes<br />

are between 5 and 16 horsepower, which can<br />

provide a community <strong>of</strong> 300 to 1500 people<br />

all <strong>of</strong> its agricultural processing and potable<br />

water pumping. [1]<br />

The most important setback related to <strong>the</strong><br />

use <strong>of</strong> Jatropha Oil is its viscosity. If <strong>the</strong> fuel is<br />

fed to <strong>the</strong> engine at room temperatures, <strong>the</strong><br />

engine can bear critical amounts <strong>of</strong> wear in<br />

vital components such as its fuel line and<br />

cylinder. This problem can be potentially fixed<br />

by pre-heating <strong>the</strong> Jatropha Oil to an ideal<br />

temperature <strong>of</strong> 110 o C. Jatropha Oil is known<br />

to have viscosity levels matching that <strong>of</strong><br />

regular diesel at this designated range.<br />

II. OBJECTIVE<br />

Accordingly, <strong>the</strong> objective <strong>of</strong> <strong>the</strong> BELT<br />

program is to devise an economical and<br />

practical pre-heating kit capable <strong>of</strong> fueling <strong>the</strong><br />

engine on Jatropha Oil at <strong>the</strong> desired<br />

temperature. The project includes three<br />

different teams located at CCNY, Manhattan<br />

College and Columbia Universities. Each team<br />

is charged with <strong>the</strong> task <strong>of</strong> devising <strong>the</strong>ir own<br />

289<br />

Viscosity (cS)<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

Viscosity vs. Temperature<br />

Jatropha<br />

Canola<br />

Diesel<br />

0<br />

20.0 60.0 100.0<br />

Temperature (C)<br />

140.0<br />

Figure 1: Viscosity as a function <strong>of</strong><br />

temperature<br />

pre-heating kit and providing experimental<br />

data on <strong>the</strong> practicality <strong>of</strong> <strong>the</strong> design during a<br />

500-hour test using waste vegetable oil<br />

(WVO), a fluid with similar characteristics as<br />

Jatropha Oil. The apparatus is <strong>the</strong>n to be<br />

modified into a 'field kit' that is intended for<br />

use at a pilot site in Uganda.<br />

III. METHOD<br />

The experiment consists <strong>of</strong> mechanical,<br />

electrical and data acquisition components.<br />

The Listeroid engine itself is attached to an ST<br />

generator by a groove belt. The generator is<br />

<strong>the</strong>n wired to an electric board consisting <strong>of</strong><br />

ten 300-Watt bulbs which act as a variable<br />

engine load. This composition mimics a field<br />

setup <strong>of</strong> <strong>the</strong> Listeroid engine where it is<br />

usually loaded by a pump, mill, or generator.<br />

The innovative portion <strong>of</strong> <strong>the</strong> experiment<br />

lies within <strong>the</strong> pre-heating kit. The CCNY<br />

BELT team is designing a dual stage preheating<br />

kit where <strong>the</strong> vegetable fuel is preheated<br />

to 110 C before <strong>the</strong> fuel pump as well<br />

as before <strong>the</strong> fuel injector, which is<br />

approximately 2 feet <strong>of</strong> fuel line away from<br />

<strong>the</strong> fuel pump.<br />

This approach willdecrease <strong>the</strong> load on <strong>the</strong><br />

pump cycle and improve <strong>the</strong> efficiency <strong>of</strong> <strong>the</strong><br />

engine's <strong>the</strong>rmodynamic cycle.


A). CONTROLS AND PRE-HEATING<br />

SYSTEM<br />

The process <strong>of</strong> pre-heating <strong>the</strong> bi<strong>of</strong>uel before<br />

it is injected into <strong>the</strong> engine is a vital part <strong>of</strong><br />

<strong>the</strong> BELT project. The first heating element,<br />

<strong>the</strong> VOW2, which is installed before <strong>the</strong><br />

pump, serves as a “ball-park” pre-heater,<br />

effectively increasing <strong>the</strong> temperature <strong>of</strong> <strong>the</strong><br />

bi<strong>of</strong>uel from ambient conditions to about<br />

90�C�95� C , so that <strong>the</strong> fuel pump does<br />

less work transporting <strong>the</strong> warmer bi<strong>of</strong>uel<br />

(with lower viscosity and <strong>the</strong>refore, less<br />

energy loss).<br />

The bi<strong>of</strong>uel <strong>the</strong>n travels through <strong>the</strong> pump<br />

until it reaches second heating element, <strong>the</strong><br />

cartridge heater, which is installed right before<br />

<strong>the</strong> fuel is injected into <strong>the</strong> engine. The<br />

cartridge heater will <strong>the</strong>n serve as a precision<br />

heating device, raising <strong>the</strong> temperature <strong>of</strong> <strong>the</strong><br />

pre-heated fuel to <strong>the</strong> desired range <strong>of</strong><br />

110�C�115� C . The purpose <strong>of</strong> <strong>the</strong> control<br />

system, as <strong>the</strong> name suggests, is to control <strong>the</strong><br />

fuel heating process. If left unattended, <strong>the</strong><br />

two heating elements may raise <strong>the</strong> bi<strong>of</strong>uel<br />

temperature to undesirable values which<br />

undermines <strong>the</strong> energy efficiency purpose <strong>of</strong><br />

this project. This mechanism will result in a<br />

double-loop. A <strong>the</strong>oretical PID analysis <strong>of</strong> an<br />

impulse response reveals that <strong>the</strong> system is<br />

stable and can reach <strong>the</strong> target temperature<br />

within approximately 5 seconds. [2]<br />

The temperature output <strong>of</strong> an individual<br />

heating element was calculated to be<br />

T mi<br />

0<br />

x<br />

q s ''<br />

(1)<br />

T m (x)<br />

Figure 2: Theoretical illustration <strong>of</strong> preheating<br />

element<br />

where Tme is <strong>the</strong> exit temperature, Tmi is <strong>the</strong><br />

inlet temperature, qs” is <strong>the</strong> heat flux, P is <strong>the</strong><br />

perimeter <strong>of</strong> <strong>the</strong> fuel line, L is <strong>the</strong> pre-heater<br />

length, m� is <strong>the</strong> mass flow rate and cp is <strong>the</strong><br />

specific heat <strong>of</strong> <strong>the</strong> Jatropha oil. [3]<br />

B). TESTING PARAMETERS<br />

In order to ensure compatible data sets<br />

between <strong>the</strong> three BELT teams, <strong>the</strong> following<br />

290<br />

general categories will be used as various<br />

testing parameters:<br />

1. Emissions<br />

2. Lubricating oil<br />

3. Fuel<br />

4. Engine Loading<br />

5. Thermal/Pressure<br />

IV. DISCUSSION AND FUTURE WORK<br />

The CCNY team has to collect sufficient data<br />

that will provide accurate information on <strong>the</strong><br />

viability <strong>of</strong> using pre-heated Jatropha oil for<br />

third world communities. Even collecting data<br />

from <strong>the</strong> current pre-heating mechanism has to<br />

be used to refine <strong>the</strong> design into a much more<br />

robust device that can be implemented outside<br />

<strong>of</strong> <strong>the</strong> laboratory and in remote environments.<br />

The team is also expected to participate in <strong>the</strong><br />

expansion <strong>of</strong> a BELT pilot site in Uganda that<br />

has been constructed at a secondary school<br />

through collaboration between Columbia<br />

University, and Pilgrim (a Ugandan NGO).<br />

Additionally, <strong>the</strong> team looks to expand such<br />

that ano<strong>the</strong>r Listeroid engine may be installed<br />

by <strong>the</strong> end <strong>of</strong> <strong>the</strong> <strong>2008</strong>-2009 academic year.<br />

Future pre-heating designs with additional<br />

data will fur<strong>the</strong>r serve <strong>the</strong> BELT project's<br />

objectives.<br />

ACKNOWLEDGEMENTS<br />

The authors would like to thank Matt<br />

Basinger, PhD student at Columbia University,<br />

for project mentoring and data findings.<br />

REFERENCES<br />

[1]Nate Beckett, Manuel Donnay, Seth<br />

Karpkinski, S. Carl Nolen, and Rena<br />

Rudavsky. Straight Vegetable Oil<br />

(SVO) Engine Modification Kit.<br />

Columbia University Department <strong>of</strong><br />

Mechanical Engineering, pp. 7-12,<br />

June 2006<br />

[2] Jiji, Latif M. Heat Convection.<br />

New<br />

York: Springer, 2006.<br />

[3] Ogata, Katsuhiko. System Dynamics.<br />

4 th edition. New Jersey: Pearson<br />

Prentice Hall, 2004.


The “G” Ring Laser in Wettzell, Germany, as an Indicator for Astronomical<br />

and Geophysical Disturbances <strong>of</strong> <strong>the</strong> Planet Earth<br />

Lucia Plank and Paulo Jorge Mendes Cerveira (Faculty Mentor)<br />

Institute <strong>of</strong> Geodesy and Geophysics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: luciplank@hotmail.com<br />

Abstract — In <strong>the</strong> last years, <strong>the</strong> absolute measurement<br />

<strong>of</strong> Earth’s rotation using inertial sensors<br />

has become reality. Laser gyroscopes, e.g. <strong>the</strong> “G”<br />

Wettzell ring laser, use <strong>the</strong> Sagnac effect to sense<br />

<strong>the</strong> instantaneous spin and orientation <strong>of</strong> <strong>the</strong> Earth.<br />

Variations in <strong>the</strong> latter are <strong>of</strong> astronomical and<br />

geophysical origin. The astronomical part, e.g. <strong>the</strong><br />

Oppolzer terms, is detected from a dynamical approach<br />

and supports current <strong>the</strong>ories up to a very<br />

high accuracy (


III. RESULTS<br />

A. DIURNAL POLAR MOTION<br />

The lunisolar and planetary torques on an heterogeneous,<br />

quasi-ellipsoidal and tilted Earth cause a<br />

nearly circular motion <strong>of</strong> <strong>the</strong> instantaneous rotation<br />

pole (IRP) in an Earth-fixed frame (see figure 2), also<br />

known as Oppolzer terms. Space geodetic techniques<br />

are basically not able to distinguish between spacerelated<br />

(nutation terms) and Earth-related (Oppolzer<br />

terms) components. Ring lasers have direct access to<br />

<strong>the</strong> Earth-related motions and are <strong>the</strong>refore sensitive<br />

to diurnal polar motion [3]. Detailed <strong>the</strong>oretical investigations<br />

<strong>of</strong> diurnal polar motion for an elastically<br />

and anelastically deformable Earth with a liquid outer<br />

core are available. Differences between <strong>the</strong>ory and<br />

measurement are small.<br />

Figure 2: Oppolzer terms for <strong>the</strong> period 22.9.2006-<br />

13.2.2007, derived from space geodetic techniques.<br />

CIP stands for Celestial Intermediate Pole.<br />

B. LOCAL ORIENTATION CHANGES<br />

The “G” ring laser works also as a north-south tiltmeter<br />

[2], i.e. it divulges small angular variations<br />

arising from local tilts due to ground deformation.<br />

As is shown in figures 3 and 4, <strong>the</strong> solid Earth tides<br />

in north-south direction can be nicely monitored. A<br />

big difference <strong>of</strong> <strong>the</strong> ring laser versus some tiltmeter<br />

is that <strong>the</strong> former is only sensitive to <strong>the</strong> geometrical<br />

deformation and does not respond to attractional<br />

variations.<br />

292<br />

Figure 3: <strong>the</strong>oretical scalograms <strong>of</strong> <strong>the</strong> relative<br />

Sagnac frequency (units <strong>of</strong> mas) for <strong>the</strong> Oppolzer<br />

terms (left) and solid Earth tides (right).<br />

Figure 4: relative Sagnac frequency (units <strong>of</strong> mas)<br />

from <strong>the</strong> original ring laser data (left), and residual<br />

<strong>of</strong> relative Sagnac frequency, after removal <strong>of</strong><br />

Oppozer terms and solid Earth tides (right).<br />

ACKNOWLEDGMENTS<br />

The author is indebted to <strong>the</strong> Bundesamt für Kartographie<br />

und Geodäsie (BKG), and in particular to<br />

Pr<strong>of</strong>. U. Schreiber and Dr. T. Klügel for giving access<br />

to <strong>the</strong> ring laser data from <strong>the</strong> Geodätisches<br />

Observatorium Wettzell, Germany.<br />

REFERENCES<br />

[1] E.J. Post, Sagnac Effect, Reviews <strong>of</strong> Modern<br />

Physics, Vol. 39 (2), pp. 475-493 (1967)<br />

[2] V. Rautenberg, H.P. Plag, M. Burns, G.E. Stedman<br />

and H.U. Jüttner, Tidally induced Sagnac<br />

signal in a ring laser, Geophys. Res. Lett., Vol.<br />

24 (8), pp. 893-896 (1997)<br />

[3] K.U. Schreiber, T. Klügel, A. Velikoseltsev, W.<br />

Schlüter and G.E. Stedman, The Large ring laser<br />

G for Continuous Earth Rotation Monitoring,<br />

submitted to PAGEOPH (<strong>2008</strong>)


Occurrence and Possible Origin <strong>of</strong><br />

Humic-like substances in<br />

Vienna Airborne Particulate Matter<br />

BARBARA KLATZER, Alex Caseiro, Heidi Bauer, Hans Puxbaum.<br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: bklatzer@mail.tuwien.ac.at<br />

The presence <strong>of</strong> Humic-like substances (HULIS),<br />

also referred to as organic macromolecules and<br />

polyacids, in airborne particulate matter was reported<br />

as early as 1980 [1]. Since <strong>the</strong>n, numerous<br />

studies have been conducted to shed light on<br />

chemical structure, physical properties and formation<br />

pathways <strong>of</strong> <strong>the</strong>se macromolecular compounds<br />

that contribute to atmospheric organic<br />

carbon with up to 50 %. Figure 1 visualises <strong>the</strong><br />

increasing interest since <strong>the</strong> term HULIS was<br />

coined [2] in 1998.<br />

20<br />

10<br />

0<br />

2<br />

0<br />

1998<br />

1999<br />

number <strong>of</strong><br />

publications<br />

containing <strong>the</strong> term<br />

"HULIS" in title or<br />

keywords<br />

1<br />

0<br />

2<br />

2000<br />

2001<br />

2002<br />

4<br />

2<br />

9<br />

11<br />

2003<br />

2004<br />

2005<br />

2006<br />

Figure 1: Histogram <strong>of</strong> publications<br />

18<br />

11<br />

2007<br />

<strong>2008</strong><br />

Atmospheric HULIS levels have been quantified<br />

by using a two step separation technique with<br />

subsequent carbon-specific detection [3] and contributions<br />

<strong>of</strong> HULIS to organic carbon between 6<br />

and 40 % have been revealed for background sites<br />

in a European west east transect and for intra urban<br />

and remote sites, respectively [4, 5]. For urban<br />

and suburban sampling sites in Eastern Austria<br />

2004 seasonal behaviour <strong>of</strong> HULIS levels were<br />

investigated in airborne particulate matter (PM10)<br />

by quantification with <strong>the</strong> TU Vienna HULIS<br />

Analyser [3]. The first separation step assures <strong>the</strong><br />

removal <strong>of</strong> inorganic ions and organic hydrophilic<br />

compounds, while an ion-exchange step removes<br />

mono- and dicarboxylic acids. Aerosol samples on<br />

quartz fibre filters are extracted first with ultrapure<br />

293<br />

water and dilute alkali. Monthly averaged HULIS<br />

levels are depicted in figure 2.<br />

µg HULIS C/m³<br />

2,5<br />

2,0<br />

1,5<br />

1,0<br />

0,5<br />

0,0<br />

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov<br />

Urban fringe Intra urban<br />

Figure 2: Monthly averaged HULIS C concentration<br />

(n=2-14)<br />

Lowest concentration levels were found during<br />

summer season with 0.2 μg HULIS C/m³. The<br />

highest monthly mean occurred in December at <strong>the</strong><br />

urban fringe site with 2.2 μg C/m³. Seasonal averaged<br />

winter enrichment factors <strong>of</strong> 2 and 4, have<br />

been found for <strong>the</strong> intra urban and <strong>the</strong> urban fringe<br />

site, respectively.<br />

HULIS C<br />

2,5<br />

2,0<br />

1,5<br />

1,0<br />

0,5<br />

R 2 = 0,82<br />

R 2 = 0,78<br />

0,0<br />

0,00 0,05 0,10 0,15 0,20 0,25 0,30 0,35 0,40 0,45 0,50<br />

levoglucosan<br />

Schafbergbad (urban fringe) Rinnböckstrasse<br />

Linear (Schafbergbad (urban fringe)) Linear (Rinnböckstrasse)<br />

Figure 3: Correlation between HULIS C and<br />

levoglucosan for intra urban and urban fringe site<br />

Monthly HULIS averages strongly correlated with<br />

levels <strong>of</strong> <strong>the</strong> biomass combustion tracer levoglucosan<br />

for both sites (R²~0.8), shown in figure 3, what<br />

indicates biomass combustion products to be a<br />

source for HULIS formation.<br />

Dec


REFERENCES<br />

[1] Simoneit, B.R.T., 1980.Eolian particulates<br />

from oceanic and rural areas - <strong>the</strong>ir<br />

lipids, fulvic and humic acids and residual<br />

carbon. Physics and Chemistry <strong>of</strong> <strong>the</strong> Earth<br />

(1956-1998) 12 (Adv. Org. Geochem. 1979),<br />

343-52.<br />

[2] Havers, N., Burba, P., Lambert, J.,<br />

Klockow, D., 1998. Spectroscopic characterization<br />

<strong>of</strong> humic-like substances in airborne<br />

particulate matter. Journal <strong>of</strong> Atmospheric<br />

Chemistry, 29(1), 45-54.<br />

[3] Limbeck, A., Handler, M., Neuberger, B.,<br />

Klatzer, B., Puxbaum, H., 2005.Carbon-<br />

Specific Analysis <strong>of</strong> Humic-like Substances in<br />

Atmospheric Aerosol and Precipitation Samples.<br />

Analytical Chemistry 77(22), 7288-<br />

7293.<br />

[4] Feczko, T., Puxbaum, H., Kasper-Giebl,<br />

A., Handler, M., Limbeck, A., Gelencser, A.,<br />

Pio, C., Preunkert, S., Legrand, M. 2007.<br />

Determination <strong>of</strong> water and alkaline extractable<br />

atmospheric humic-like substances with<br />

<strong>the</strong> TU Vienna HULIS analyzer in samples<br />

from six background sites in Europe. Journal<br />

<strong>of</strong> Geophysical Research, [Atmospheres]<br />

112(D23), D23S10/1-D23S10/9.<br />

[5] Wonaschütz, A., Hitzenberger, R.,Bauer<br />

H., Klatzer, B., Puxbaum, H., <strong>2008</strong>. Separation<br />

<strong>of</strong> „brown“and black carbon in wood<br />

smoke impacted atmospheric aerosol with <strong>the</strong><br />

integrated sphere technique, Environmental<br />

Science & Technology, submitted.<br />

294<br />

ACKNOWLEDGEMENTS<br />

Financial funding <strong>of</strong> <strong>the</strong> AQUELLA Wien project,<br />

which <strong>the</strong> work presented here was part <strong>of</strong>, by <strong>the</strong><br />

municipal government <strong>of</strong> Vienna is greatly appreciated.


Communication Specialties between Ear and Brain:<br />

Analysis <strong>of</strong> <strong>the</strong> Cluster Functions in <strong>the</strong> Human Cochlea<br />

Abstract — Although <strong>the</strong> somatic region <strong>of</strong> afferent human<br />

cochlear neurons is quite unique, our contemporary knowledge<br />

on <strong>the</strong> neural coding principles in mammalian cochlear neurons<br />

is primarily based on animal experiments. However two<br />

outstanding human particularities are expected to essentially<br />

affect <strong>the</strong> neural code resulting in a specific human physiologic<br />

hearing performance.<br />

Two ma<strong>the</strong>matical models are presented to account for <strong>the</strong><br />

anatomical differences which are <strong>of</strong> major relevance for <strong>the</strong><br />

propagation <strong>of</strong> an action potential.<br />

I. INTRODUCTION<br />

The somatic region <strong>of</strong> afferent human cochlear neurons<br />

is quite unique. Firstly many <strong>of</strong> <strong>the</strong> cochlear neurons are<br />

ga<strong>the</strong>red to clusters with 2-4 [1] neurons having a common<br />

insulation by myelin (B in Figure 1). Secondly 94%<br />

<strong>of</strong> human spiral ganglion cells are mostly surrounded by<br />

only one to several layers <strong>of</strong> satellite cells [2], whereas<br />

in cats 95% <strong>of</strong> spiral ganglion cells are firmly myelinated<br />

[3], which represents almost a reverse proportion<br />

between man and cat (A in Figure 1).<br />

Figure 1: (A) Schema <strong>of</strong> <strong>the</strong> cochlear neurons in cat<br />

and man; myelinated (grey) and nonmyelinated (white).<br />

(B) Scanning electron microscopy <strong>of</strong> <strong>the</strong> 2kHz (left) and<br />

1kHz (right) region. [4]<br />

Nerve fiber model analysis and computer simulations<br />

are valuable tools to enlighten <strong>the</strong> functional relevance<br />

<strong>of</strong> <strong>the</strong> particular human case which is not accesible to<br />

neurophysiologic experiments as used in animals. Based<br />

Cornelia Wenger and Frank Rattay<br />

Institute <strong>of</strong> Analysis and Scientific Computing<br />

Technical University <strong>of</strong> Vienna<br />

Vienna, Austria<br />

corneliawenger@hotmail.com<br />

295<br />

on [5] a ma<strong>the</strong>matical model <strong>of</strong> <strong>the</strong> neuron clusters is developed<br />

to study <strong>the</strong>ir influence on <strong>the</strong> spiking behavior<br />

<strong>of</strong> <strong>the</strong> involved neurons.<br />

II. METHODS<br />

In our compartment model <strong>the</strong> processes are split up in<br />

nodes and internodes. Fur<strong>the</strong>rmore <strong>the</strong> presomatic region<br />

represents three compartments whereas <strong>the</strong> soma<br />

and its postsomatic region are single units. All compartments<br />

are cylindrically approximated beside <strong>the</strong> spherical<br />

soma. Neurons which share a cluster are similarly<br />

spatially arranged and are assumed to have <strong>the</strong> same geometrical<br />

and electrical properties. The surrounding passive<br />

cluster membrane enwraps <strong>the</strong> whole nonmyelinated<br />

soma region from <strong>the</strong> pre- to <strong>the</strong> postsomatic compartment.<br />

To simulate <strong>the</strong> reduced transmembrane voltage <strong>of</strong> <strong>the</strong><br />

n-th compartment <strong>of</strong> a neuron Vn = Vi,n −Ve,n −Vrest,n<br />

<strong>the</strong> Hodgkin-Huxley equations were adapted [5]. For<br />

analyzing <strong>the</strong> neuron cluster <strong>the</strong> electrical circuit is en-<br />

hanced to obtain <strong>the</strong> differential equations for <strong>the</strong> trans-<br />

membrane potential <strong>of</strong> <strong>the</strong> cluster V Cl<br />

n , presented in (1),<br />

and <strong>the</strong> enclosed neurons V k n ,k=1..4 in (2). For more<br />

details see [6].<br />

dV k n<br />

dt =<br />

�<br />

− I k ion,n +<br />

V k n+1 − V k n<br />

Rk n+1 /2+Rk n/2 +<br />

V k n−1 − V k n<br />

Rk n−1 /2+Rk n/2 +<br />

V Cl<br />

n+1 − V Cl<br />

n<br />

Rk n+1 /2+Rk V Cl<br />

n−1 − V Cl<br />

n<br />

+<br />

n/2 Rk n−1 /2+Rk �<br />

/C<br />

n/2<br />

k n<br />

(1)<br />

dV Cl<br />

n<br />

dt =<br />

�<br />

− I Cl<br />

ion −<br />

j�<br />

[I<br />

k=1<br />

k ion + Ik c ]+<br />

V Cl<br />

n−1 − V Cl<br />

n<br />

RCl n−1 /2+RCl<br />

+<br />

n /2<br />

V Cl<br />

n+1 − V Cl<br />

n<br />

RCl n+1 /2+RCl n /2 ]<br />

�<br />

/ cCl n<br />

nmCl n<br />

(2)<br />

III. RESULTS<br />

A first model <strong>of</strong> a single human cochlear neuron showed<br />

strong effects on <strong>the</strong> excitation patterns <strong>of</strong> <strong>the</strong> neuron by<br />

small changes in certain sensitive geometrical and electrical<br />

parameters. The second extended model demonstrated<br />

<strong>the</strong> influence <strong>of</strong> <strong>the</strong> cluster on <strong>the</strong> spiking behavior<br />

<strong>of</strong> <strong>the</strong> enclosed cochlear neurons.


A. NONMYELINATED SOMA REGION<br />

Five parameters have been tested: <strong>the</strong> intracellular resistivity<br />

ρi, <strong>the</strong> number <strong>of</strong> myelin sheets at <strong>the</strong> soma<br />

nmsoma, <strong>the</strong> diameter <strong>of</strong> <strong>the</strong> peripheral process d(1), <strong>the</strong><br />

length <strong>of</strong> <strong>the</strong> last peripheral internode li(6) and that <strong>of</strong><br />

<strong>the</strong> presomatic region lsomap. To analyze <strong>the</strong> influence<br />

<strong>of</strong> changes in <strong>the</strong>se parameters standard values have been<br />

defined. As an example Figure 2 presents <strong>the</strong> transmembrane<br />

voltage <strong>of</strong> <strong>the</strong> last peripheral internode at <strong>the</strong> left,<br />

<strong>the</strong> first presomatic region in <strong>the</strong> middle and <strong>the</strong> soma at<br />

<strong>the</strong> right.<br />

Figure 2: Variation <strong>of</strong> d(1). The thinnest diameter fails<br />

to excite <strong>the</strong> soma.<br />

The simulation starts with a thinner peripheral process<br />

with a diameter <strong>of</strong> 0.2 μm which is also <strong>the</strong> stepsize leading<br />

to five different cases up to <strong>the</strong> standard value for<br />

d(1) <strong>of</strong> 1 μm.<br />

The excitation with <strong>the</strong> thinnest diameter is <strong>the</strong> fastest<br />

and strongest but leaks <strong>the</strong> second rising phase after a<br />

short fall <strong>of</strong> Vn (Figure 2 left). Although it seems that<br />

at <strong>the</strong> first presomatic soma compartment all cases lead<br />

to an ongoing propagation, <strong>the</strong> thinnest diameter fails<br />

to excite <strong>the</strong> soma. Even if it looks like <strong>the</strong> velocity<br />

and strength increases from thinnest to thickest diameter<br />

as detected in <strong>the</strong> middle subframe it turns out that<br />

<strong>the</strong> strongest excitation is provided by <strong>the</strong> thickest diameter.<br />

Never<strong>the</strong>less two cases have increased velocity<br />

compared to <strong>the</strong> standard case value.<br />

B. NEURON CLUSTER<br />

At first <strong>the</strong> <strong>the</strong>oretical case <strong>of</strong> a single neuron with a<br />

soma ensheathing cluster has been tested, in order to see<br />

whe<strong>the</strong>r <strong>the</strong> cluster sheets support or hinder <strong>the</strong> spiking<br />

behavior. Altered cluster diameters revealed a decreased<br />

influence <strong>of</strong> <strong>the</strong> cluster on <strong>the</strong> neuron with increasing<br />

size. Increasing <strong>the</strong> myelin layers <strong>of</strong> <strong>the</strong> cluster and <strong>the</strong><br />

neuron show that in most cases <strong>the</strong> neuron is positively<br />

effected, resulting in an increased and accelerated excitation.<br />

Fur<strong>the</strong>rmore multiple neuron clusters have been analyzed<br />

with <strong>the</strong> goal to examine <strong>the</strong> influence <strong>of</strong> <strong>the</strong> nerve<br />

cells on each o<strong>the</strong>r and to elucidate if one neuron could<br />

be triggered to produce a nerve impulse under subthreshold<br />

conditions when it is supported by properly stimu-<br />

296<br />

lated o<strong>the</strong>r neurons within <strong>the</strong> same cluster (Figure 3).<br />

Figure 3: Voltages in a four neuron cluster at <strong>the</strong> soma for<br />

properly excited neurons at <strong>the</strong> left. One graph represents<br />

a neuron under subthreshold condition in <strong>the</strong> middle and<br />

right.<br />

On <strong>the</strong> left all four neurons are properly stimulated<br />

with identical strengths, i1a-i1d, but different onsets, t1at1d.<br />

Observing <strong>the</strong> maximum values V k max <strong>the</strong> effect <strong>of</strong><br />

<strong>the</strong> neurons on each o<strong>the</strong>r can be verified, especially <strong>the</strong><br />

fourth is pr<strong>of</strong>iting from <strong>the</strong> excitation <strong>of</strong> <strong>the</strong> o<strong>the</strong>rs.<br />

The o<strong>the</strong>r subframes show three properly stimulated<br />

neurons and one with a subthreshold stimulus with different<br />

stimulus onsets. Although <strong>the</strong> influence on <strong>the</strong> excitation<br />

<strong>of</strong> <strong>the</strong> fourth neuron can be clearly observed <strong>the</strong><br />

o<strong>the</strong>r neurons are still not able to produce a nerve impulse<br />

at an adjacent one under subthreshold conditions.<br />

REFERENCES<br />

[1] S. Tylstedt, A. Kinnefors, and H. Rask-Anderson.<br />

Neural interaction in <strong>the</strong> human spiral ganglion: A<br />

tem study. Acta oto-laryngologica, 117:505–512,<br />

1997.<br />

[2] C.Y. Ota and R.S. Kimura. Ultrastructural study <strong>of</strong><br />

<strong>the</strong> human spiral ganglion. Acta oto-laryngologica,<br />

89:53–62, 1980.<br />

[3] H. Spoendlin. Degeneration behaviour <strong>of</strong> <strong>the</strong><br />

cochlear nerve. Arch. Klin. Exp. Ohren Nasen<br />

Kehlkopfheilkd., 200:275–291, 1971.<br />

[4] R. Glueckert, K. Pfaller, A. Kinnefors, H. Rask-<br />

Anderson, and A. Schrott-Fischer. The human spiral<br />

ganglion: New insights into ultrastructure, survival<br />

rate and implications for cochlear implants. Audio<br />

Neurothology, 10:258–273, 2005.<br />

[5] F. Rattay, P. Lutter, and H. Felix. A model <strong>of</strong><br />

<strong>the</strong> electrically excited human cochlear neuron: I.<br />

contribution <strong>of</strong> neural substructures to <strong>the</strong> generation<br />

and propagation <strong>of</strong> spikes. Hearing Research,<br />

153:43–63, 2001.<br />

[6] C. Wenger. Analysis <strong>of</strong> <strong>the</strong> cluster functions in <strong>the</strong><br />

human cochlea. Master’s <strong>the</strong>sis, Technical University<br />

<strong>of</strong> Vienna, <strong>2008</strong>.


Spatial potentials for renewable energies<br />

Study results concerning 3 federal states in eastern Austria<br />

A. Dillinger, H. Dumke, S. Plha, H. Schaffer and<br />

Vicerektor Hon. Pr<strong>of</strong>. Dipl.-Ing. Dr.techn. Gerhard Schimak<br />

Department <strong>of</strong> Spatial Development, Infrastructure & Environmental Planning<br />

Center Regional Planning and Regional Development<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

e0025325@mail.student.tuwien.ac.at<br />

Abstract — The study “Spatial potentials for renewable<br />

energies” analysed <strong>the</strong> distribution and<br />

amounts <strong>of</strong> renewable energy production in 3 federal<br />

states in eastern Austria with main focus on<br />

biomass, geo<strong>the</strong>rmal energy and wind power potentials.<br />

As many EU guidelines, policy papers, national<br />

laws, local mission statements demand an<br />

enormous increase in <strong>the</strong> production <strong>of</strong> renewable<br />

energy and bi<strong>of</strong>uel, accurate data on <strong>the</strong>ir spatial<br />

distribution and dimension is urgently needed.<br />

Therefore all results were visualised in maps down<br />

to <strong>the</strong> level <strong>of</strong> <strong>the</strong> communities, which represents an<br />

innovative approach since most previous studies<br />

included only figure calculations. The main aims<br />

were to find “potential focus regions” and to assess<br />

<strong>the</strong> possible additional energy output especially out<br />

<strong>of</strong> local resources in <strong>the</strong>se areas.<br />

I. MAIN IMPORTANT RESULTS -BIOMASS<br />

The future <strong>of</strong> biomass-related energy production lies<br />

within a vast pool <strong>of</strong> ra<strong>the</strong>r small-scaled decentralised<br />

facilities, using local resources from <strong>the</strong><br />

nearby environment. The aggregation <strong>of</strong> forest, agriculture<br />

(including derelict areas), grassland and reed<br />

indicated clearly <strong>the</strong> respective focus areas, representing<br />

typical landscape units on <strong>the</strong> one hand and<br />

<strong>the</strong> per-capita (hectares per person) distribution on<br />

<strong>the</strong> o<strong>the</strong>r. Using a simple model showing <strong>the</strong> probable<br />

share <strong>of</strong> potential energetic use in total (in competition<br />

to industrial use or food production) <strong>the</strong><br />

energy output <strong>of</strong> 38 Petajoule seems possible in a<br />

short-time period. This amount represents a heat<br />

capacity for about 1.8 million people. In order to<br />

increase efficient biomass energy production without<br />

huge transport costs, rural communities with large<br />

areas, but low population densities have <strong>the</strong> best<br />

endogenous preconditions, whereas urban regions<br />

will always be bound to immense imports <strong>of</strong> resources.<br />

297<br />

The aggregation <strong>of</strong> <strong>the</strong> different potential biomass<br />

areas showed that <strong>the</strong> most promising regions are <strong>the</strong><br />

nor<strong>the</strong>rn and sou<strong>the</strong>rn (Figure 1) parts <strong>of</strong> Lower<br />

Austria. The eastern part <strong>of</strong> Lower Austria and <strong>the</strong><br />

nor<strong>the</strong>rn part <strong>of</strong> <strong>the</strong> Burgenland (close to <strong>the</strong> Neusiedler<br />

Lake) can ra<strong>the</strong>r benefit from agricultural<br />

Biomass. The east shore <strong>of</strong> <strong>the</strong> lake also provides a<br />

very specific resource - large reed fields - which are<br />

ecologically very valuable, but can also provide heat<br />

production.<br />

Figure 1: Extract <strong>of</strong> potential energy<br />

output per person<br />

II. MAIN IMPORTANT RESULTS -WIND<br />

POWER<br />

Due to <strong>the</strong> massive enlargements <strong>of</strong> wind parks in <strong>the</strong><br />

recent years <strong>the</strong> three federal states already produce<br />

electricity for about 560.000 households. This equals<br />

90% <strong>of</strong> <strong>the</strong> total Austrian production. The study<br />

mapped all existing single wind engines and supporting<br />

and hindering factors were added. A consistent<br />

and complete GIS model was created to maintain<br />

ongoing surveys. Buffers have been introduced to<br />

show hindering or prohibiting factors for new wind


mills like existing (or zoned) building land, nature<br />

reserves <strong>of</strong> all kinds or aircraft fields. Supporting<br />

factors are zoned wind enlargement areas (only in <strong>the</strong><br />

Burgenland zoned by law) and a strong average wind<br />

power. Suitable areas for new wind parks are visualised<br />

in dark blue colour on <strong>the</strong> map (Figure 2).<br />

Figure 2: Extract <strong>of</strong> suitable areas for wind power<br />

This remaining space for additional engines is ra<strong>the</strong>r<br />

limited, but still about 440 new wind mills – which<br />

can conduct electricity for ano<strong>the</strong>r 500.000 households<br />

- are possible here. This amount <strong>of</strong> energy is<br />

comparable with <strong>the</strong> output <strong>of</strong> three large water<br />

power plants along <strong>the</strong> Danube. The most appropriate<br />

areas for new wind plants are <strong>the</strong> nor<strong>the</strong>rn and<br />

eastern parts <strong>of</strong> Lower Austria and <strong>the</strong> nor<strong>the</strong>rn part<br />

<strong>of</strong> <strong>the</strong> Burgenland.<br />

III. MAIN IMPORTANT RESULTS -GEO-<br />

THERMAL HEAT<br />

Geo<strong>the</strong>rmal energy is <strong>of</strong>ten underestimated or even<br />

not existent in <strong>the</strong> current energy debate. In Austria<br />

<strong>the</strong>re are only 12 geo<strong>the</strong>rmal plants, two <strong>of</strong> <strong>the</strong>m<br />

producing electricity, <strong>the</strong> o<strong>the</strong>rs heat. But <strong>the</strong> future<br />

potentials are very promising. A single geo<strong>the</strong>rmal<br />

plant could cover <strong>the</strong> demand <strong>of</strong> over 1.000 households<br />

producing a constant heat flow without any<br />

emissions. But <strong>the</strong> land use patterns have to fit accurately<br />

to use geo<strong>the</strong>rmal energy efficiently: a specific<br />

structure <strong>of</strong> consumption, a needed density <strong>of</strong> buildings<br />

in combination with existing or newly constructed<br />

heating systems (pipes, single connections<br />

and steering logistics) are crucial. In <strong>the</strong> study area<br />

<strong>the</strong>re exist four geological regions with good geological<br />

preconditions: <strong>the</strong> both in <strong>the</strong> North also <strong>of</strong>fer<br />

<strong>the</strong> above mentioned specifics concerning consumption.<br />

The raster analysis (Figure 3) layered <strong>the</strong> existing<br />

average local heat connections (orange to red<br />

298<br />

squares) and <strong>the</strong> geologically promising areas (blue<br />

grid). This model allows a realistic estimation on <strong>the</strong><br />

number <strong>of</strong> household that could be served by geo<strong>the</strong>rmal<br />

heat soon. These approximately 60.000<br />

households are mostly located in <strong>the</strong> nor<strong>the</strong>rn and<br />

eastern parts <strong>of</strong> <strong>the</strong> city <strong>of</strong> Vienna and <strong>the</strong> adjacent<br />

areas. At <strong>the</strong> moment <strong>the</strong> largest petrol company in<br />

Austria (OMV) is checking if its closed oil wells<br />

could be used for geo<strong>the</strong>rmal matters. As geo<strong>the</strong>rmal<br />

energy suffers from enormous investment (drilling)<br />

costs and misses economic efficiency especially in<br />

<strong>the</strong> OCR electricity production, this re-use <strong>of</strong> closed<br />

oil wells would be a sustainable and reasonable solution.<br />

Still geo<strong>the</strong>rmal energy seems to be ra<strong>the</strong>r suitable<br />

for urban areas. Additionally new technologies<br />

are very <strong>of</strong>ten underrated – as <strong>the</strong> very first experiences<br />

with solar energy showed.<br />

Figure 3: Extrakt <strong>of</strong> geo<strong>the</strong>rmal potentials meet<br />

urban structures<br />

Fur<strong>the</strong>r information: www.pgo.wien.at/pgo_d.html<br />

ACKNOWLEDGMENTS<br />

This study was done by order <strong>of</strong><br />

Planungsgemeinschaft Ost | DI Walter Pozarek,<br />

DI Hans Schulz, Eva Maria Danzer-Horvath<br />

www.pgo.wien.at<br />

was conducted by<br />

mecca - Technisches Büro für Raum- und Landschaftsplanung<br />

| DI Hartmut Dumke, Mag.<br />

Stefan Plha, Dr. Hannes Schaffer, DI Andreas<br />

Dillinger, DI Beate Schaffer<br />

www.mecca-consulting.at<br />

and kindly supported by<br />

more than 30 different institutions and organisations


Silica nanoparticles connected<br />

through ionic linkers<br />

Marco Litschauer, Marie-Alexandra Neouze<br />

Institute <strong>of</strong> Materials Chemistry<br />

Vienna University <strong>of</strong> Technology<br />

1060 Vienna, Austria<br />

Email: malitsch@mail.zserv.tuwien.ac.at<br />

Abstract — To obtain networks build up by silica<br />

nanoparticles, connected through ionic linkers, two<br />

different syn<strong>the</strong>tic approaches were investigated.<br />

First, a bissilylated functional ionic compound was<br />

syn<strong>the</strong>sized and afterwards linked to silica<br />

nanoparticles. In <strong>the</strong> second approach, silica<br />

nanoparticles were modified with ei<strong>the</strong>r 3chloropropyltrimethoxysilane<br />

or N-(3trimethoxysilylpropyl)imidazol<br />

before being connected<br />

via nucleophilic substitution. Both strategies<br />

are compared in order to determine <strong>the</strong> most efficient<br />

method to get large networks.<br />

I. INTRODUCTION<br />

The development and introduction <strong>of</strong> nanoparticles<br />

on <strong>the</strong> one hand and functionalized ionic linkers on<br />

<strong>the</strong> o<strong>the</strong>r hand provided materials chemistry two new<br />

versatile compounds.<br />

Functionalized ionic linkers, were intensively studied<br />

when connected to an inorganic matrix, e.g. silica<br />

[1]. Many different applications were pointed out for<br />

such composites or hybrid materials, among o<strong>the</strong>rs<br />

catalytic properties, e.g. for Mizoroki-Heck reactions<br />

[2]. For metal oxide nanoparticles, <strong>the</strong> syn<strong>the</strong>sis<br />

routes, mainly sol-gel reactions [3], are well established.<br />

These particles in <strong>the</strong> nanometer range <strong>of</strong>fer,<br />

in contrast to <strong>the</strong> bulk material, interesting electronic<br />

or magnetic properties [4]. Combination <strong>of</strong> ionic<br />

species with nanoparticles drove to powerful synergies<br />

[5].<br />

The aim <strong>of</strong> this work is to syn<strong>the</strong>sise 3-dimensional<br />

silica nanoparticles networks with ionic linkers by<br />

combining <strong>the</strong>se two interesting materials. The originality<br />

<strong>of</strong> this new compound lies in <strong>the</strong> covalent<br />

bonding. The two different syn<strong>the</strong>tic pathways to<br />

obtain <strong>the</strong>se networks are fur<strong>the</strong>r named strategy A<br />

and B.<br />

In strategy A (Scheme 1) <strong>the</strong> fist step was to syn<strong>the</strong>size<br />

<strong>the</strong> functionalized ionic bridging agent,<br />

1,3-di(propyltrimethoxysilyl)imidazolium iodid, via<br />

a nucleophilic substitution reaction. This reaction<br />

occurs without solvent at room temperature quanti-<br />

299<br />

tatevely [6]. Afterwards this bissilylated compound<br />

was condensed on silica nanoparticles. The nanoparticles<br />

were prepared through a Stöber process, where<br />

<strong>the</strong> size <strong>of</strong> <strong>the</strong> particles can easily be controlled by<br />

<strong>the</strong> syn<strong>the</strong>sis parameters, like pH or concentration <strong>of</strong><br />

<strong>the</strong> precursor.<br />

O<br />

O<br />

Si<br />

O<br />

SiO 2<br />

O<br />

N<br />

+<br />

N<br />

Si<br />

O<br />

I O<br />

-<br />

N<br />

+<br />

I -<br />

N<br />

+<br />

SiO 2<br />

Scheme 1: Strategy A.<br />

Strategy B (Scheme 2) is <strong>the</strong> inversive equivalent<br />

<strong>of</strong> strategy A. Here, previously syn<strong>the</strong>sized silica<br />

nanoparticles are separately modified, through a<br />

condensation reaction, with ei<strong>the</strong>r<br />

3-chloropropyltrimethoxysilane or N-(3trimethoxysilylpropyl)imidazol<br />

[7]. Afterwards<br />

<strong>the</strong>se two different nanoparticles suspensions are<br />

mixed toge<strong>the</strong>r for <strong>the</strong> nucleophilic substitution to<br />

take place.<br />

O<br />

O<br />

Si<br />

O<br />

SiO2<br />

SiO 2<br />

+ +<br />

N<br />

Cl N<br />

Cl<br />

+<br />

N<br />

Scheme 2: Strategy B.<br />

N<br />

SiO2<br />

SiO 2<br />

O<br />

O<br />

Si<br />

O<br />

SiO 2


II. RESULT AND DISCUSSION<br />

To prove <strong>the</strong> formation <strong>of</strong> <strong>the</strong> imidazolium linker<br />

TGA measurements were carried out. Indeed <strong>the</strong><br />

<strong>the</strong>rmal stability <strong>of</strong> imidazolium species are remarkably<br />

high, e.g. around 300 °C [8, 9]. For strategy<br />

B, it proves <strong>the</strong> occurrence <strong>of</strong> <strong>the</strong> nucleophilic<br />

substitution, as <strong>the</strong> <strong>the</strong>rmal stability <strong>of</strong> <strong>the</strong> product is<br />

higher as <strong>the</strong> one <strong>of</strong> <strong>the</strong> reactants (Figure 1). The<br />

mass loss appearing around 100 °C are due to water<br />

desorption and SiO2-dehydration phenomena.<br />

Remaining mass / weight%<br />

100<br />

80<br />

60<br />

40<br />

20<br />

0<br />

100 200 300 400 500 600 700 800<br />

Temperature / °C<br />

Figure 1: TGA analysis; dotted lines: silica<br />

nanoparticles modified with ei<strong>the</strong>r chloropropyl or<br />

propyimidazole; solid line: product <strong>of</strong> strategy A.<br />

Solid state 29 Si CP MAS NMR confirmed that a<br />

condensation reaction between <strong>the</strong> linker and <strong>the</strong><br />

silica nanoparticles took place, formation <strong>of</strong> Si-O-Si<br />

bonds. The characteristic signals for <strong>the</strong> material are<br />

around -115 ppm, quaternary units issued from <strong>the</strong><br />

condensation <strong>of</strong> <strong>the</strong> nanoparticles precursor, and -70<br />

ppm, tertiary units originating from <strong>the</strong> condensation<br />

<strong>of</strong> <strong>the</strong> Si(OR) units <strong>of</strong> <strong>the</strong> bissilylated compound.<br />

As <strong>the</strong> condensation on <strong>the</strong> nanoparticles was already<br />

performed, for strategy B, <strong>the</strong> appropriate experiment<br />

is 13 C CP MAS NMR that confirmed <strong>the</strong> formation<br />

<strong>of</strong> <strong>the</strong> imidazolium bridge, via creation <strong>of</strong> an N + -C<br />

bond.<br />

Figure 2: TEM <strong>of</strong> silica nanoparticles linked via<br />

strategy A (left) and strategy B (right) [10].<br />

By looking at <strong>the</strong> TEM pictures, which were recorded<br />

for both strategies (Figure 2), one can see,<br />

300<br />

that for strategy A small “island-like” systems (50 –<br />

100 nm) were observed. The micrograph <strong>of</strong> strategy<br />

B indicated that much more nanoparticles are linked,<br />

leading to bigger domains. Thus strategy B seems to<br />

be more efficient.<br />

This is confirmed by SAXS experiments, comparing<br />

both methods. Fitting <strong>of</strong> <strong>the</strong> diffraction curves demonstrated<br />

a short-range ordering <strong>of</strong> <strong>the</strong> nanoparticles<br />

in <strong>the</strong> case <strong>of</strong> method B.<br />

III. CONCLUSION<br />

The formation <strong>of</strong> a 3D network consisting <strong>of</strong> silica<br />

nanoparticles and imidazolium based ionic linkers<br />

was described. For <strong>the</strong> creation <strong>of</strong> <strong>the</strong>se hybrids two<br />

strategies were developed. Strategy B was examined<br />

to be <strong>the</strong> more efficient leading to a more extended,<br />

short-range ordered, network.<br />

ACKNOWLEDGMENTS<br />

The authors are thanking Pr<strong>of</strong>. Schubert for his<br />

support and <strong>the</strong> Hochschuljubiläumstiftung der<br />

Stadt Wien for financial support.<br />

REFERENCES<br />

[1] M.-A. Neouze, J. Le Bideau, P. Gaveau, et<br />

al., Chemistry <strong>of</strong> Materials, 18, 3931-3936,<br />

2006.<br />

[2] V. Polshettiwar, P. Hesemann, and J.J.E.<br />

Moreau, Tetrahedron Letters, 48, 5363-<br />

5366, 2007.<br />

[3] W. Stoeber, A. Fink, and E. Bohn, Journal<br />

<strong>of</strong> Colloid and Interface Science, 26, 62-9,<br />

1968.<br />

[4] C.P. Collier, R.J. Saykally, J.J. Shiang, et<br />

al., Science (Washington, D. C.), 277, 1978-<br />

1981, 1997.<br />

[5] A. Taubert and Z. Li, Dalton Transactions,<br />

723-727, 2007.<br />

[6] G. Cerveau, R.J.P. Corriu, E. Framery, et<br />

al., Journal <strong>of</strong> Materials Chemistry, 12,<br />

3021-3026, 2002.<br />

[7] C.S.J. Cazin, M. Veith, P. Braunstein, et al.,<br />

Syn<strong>the</strong>sis, 622-626, 2005.<br />

[8] P. Migowski and J. Dupont, Chemistry--A<br />

European Journal, 13, 32-39, 2006.<br />

[9] P. Wasserscheid and T. Welton, Ionic Liquids<br />

in Syn<strong>the</strong>sis. 2003. 364 pp.<br />

[10] M. Litschauer and M.-A. Neouze, Journal<br />

<strong>of</strong> Materials Chemistry, 18, 640-646, <strong>2008</strong>.


An Adaptive Solver for Parabolic Partial Differential Equations<br />

Abstract — TIMETRAVELLER is a numerical solver for <strong>the</strong><br />

solution <strong>of</strong> initial/boundary value problems (IBVPs) <strong>of</strong> partial<br />

differential equations (PDEs). The transverse method <strong>of</strong><br />

lines generates a sequence <strong>of</strong> boundary value problems in<br />

ordinary differential equations (ODEs) resulting from semidiscretization<br />

in time. To <strong>the</strong>se it is possible to apply a code<br />

for boundary value problems (BVPs) to find <strong>the</strong> numerical solution<br />

in both, space and time adaptively. The BVPs are solved<br />

using BVPSUITE which has already been successfully applied<br />

to a variety <strong>of</strong> problems, see for example [1], [2], and [3].<br />

In <strong>the</strong> following an adaptive technique for mesh generation in<br />

two-point BVPs (2pBVPs) is presented, which tries to find a<br />

suitable mesh density function on a coarse mesh before increasing<br />

<strong>the</strong> number <strong>of</strong> mesh points to satisfy <strong>the</strong> tolerance requirements.<br />

I. INTRODUCTION<br />

Based on <strong>the</strong> encouraging results for <strong>the</strong> collocation code<br />

SBVP 2.0 [4], <strong>the</strong> first part <strong>of</strong> this study is concerned<br />

with <strong>the</strong> solution <strong>of</strong> IBVPs <strong>of</strong> parabolic partial differential<br />

equations. The idea is to design a solver which is<br />

adaptive in both, time and space. In <strong>the</strong> second part, an<br />

algorithm for mesh adaptation for 2pBVPs is presented<br />

which follows <strong>the</strong> idea <strong>of</strong> first refining <strong>the</strong> mesh density<br />

function φ and estimating <strong>the</strong> number <strong>of</strong> points once <strong>the</strong><br />

density has stabilized.<br />

II. THE TRANSVERSE METHOD OF LINES<br />

Consider partial differential equations <strong>of</strong> <strong>the</strong> following<br />

type:<br />

� �<br />

2 ∂u<br />

∂ u ∂u<br />

(x, t) =F (x, t), (x, t),u(x, t) (1)<br />

∂t ∂x2 ∂x<br />

The main interest lies in <strong>the</strong> solution <strong>of</strong> initial/boundary<br />

value problems <strong>of</strong> partial differential equations (1) with<br />

initial and boundary conditions<br />

Dipl.-Ing. Gernot Pulverer<br />

Advisor: Ao. Univ. Pr<strong>of</strong>. Dr. Ewa B. Weinmüller<br />

Institute for Analysis and Scientific Computing<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: gernotpulverer@gmail.com<br />

u(x, 0) = u0(x), (2)<br />

B(u(a, t),u(b, t)) = 0, (3)<br />

with a ≤ x ≤ b. In order to solve Problem (1) – (3),<br />

<strong>the</strong> transverse method <strong>of</strong> lines approach is used. The<br />

idea behind <strong>the</strong> transverse method <strong>of</strong> lines is to use a<br />

discretization in time such that each timestep <strong>of</strong> <strong>the</strong><br />

301<br />

resulting procedure implies a boundary value problem<br />

for an ordinary differential equation which can be solved<br />

using a solver designed for this class. Therefore, <strong>the</strong><br />

well established BVP-solver BVPSUITE is very suitable.<br />

BVPSUITE is capable <strong>of</strong> solving fully implicit systems<br />

<strong>of</strong> singular and regular boundary value problems for<br />

ordinary differential equations <strong>of</strong> arbitrary, mixed order.<br />

For <strong>the</strong> discretization in time <strong>the</strong> implicit Euler<br />

method is used, yielding<br />

u(x, tn+1) − u(x, tn)<br />

=<br />

Δtn<br />

� 2 ∂ u<br />

F<br />

∂u<br />

(x, tn+1),<br />

∂x2 ∂x (x, tn+1), u(x, tn+1)<br />

with initial/boundary conditions<br />

u(x, 0) = u0(x),<br />

B(u(a, tn+1),u(b, tn+1)) = 0.<br />

�<br />

,<br />

Setting vn(x) :=u(x, tn), it is easy to find that each<br />

time step corresponds to an ordinary differential equation<br />

<strong>of</strong> second order in vn+1:<br />

0=G(v ′′<br />

n+1(x), v ′ n+1(x), vn+1(x), vn(x)) (4)<br />

with boundary conditions<br />

B(vn+1(a), vn+1(b)) = 0. (5)<br />

As an example consider <strong>the</strong> imaginary time<br />

Schrödinger equation:<br />

∂u<br />

∂t − ∂2u + Vu=0,<br />

∂x2 x ∈ [−π, π], t>0. (6)<br />

Here V : Rm → R is a smooth potential that is 2πperiodic.<br />

The following initial/boundary conditions are<br />

used for <strong>the</strong> calculations:<br />

u(x, 0) = u0(x), −π ≤ x ≤ π (7)<br />

u(−π, t) = 0, 0 ≤ t ≤ b (8)<br />

u(π, t) = 0, 0 ≤ t ≤ b (9)


For numerical calculations Vu = (1 − cos x)u and<br />

u0(x) = sin(4x) +cos(x) +1. Results are shown in<br />

Figures 1 and 2.<br />

u(x,t)<br />

2<br />

1<br />

0<br />

−1<br />

−2<br />

0<br />

x<br />

2<br />

Figure 1: The evolution in time for <strong>the</strong> imaginary<br />

time Schrödinger equation with initial pr<strong>of</strong>ile sin(4x) +<br />

cos(x) +1. The numerical solution was calculated using<br />

time tolerance Tt = 10 −3 and spatial tolerance<br />

Ts =10 −5 .<br />

0<br />

0.1<br />

t<br />

0.2<br />

0 0.05 0.1 0.15 0.2 0.25<br />

t<br />

Figure 2: The values for t where <strong>the</strong> corresponding<br />

boundary value problems had to be solved to satisfy <strong>the</strong><br />

tolerance requirement Tt =10 −3 .<br />

For difficult problems <strong>the</strong> number <strong>of</strong> time steps is very<br />

large. As every time step requires <strong>the</strong> solution <strong>of</strong> a full<br />

two point ODE boundary value problem, an efficient<br />

mesh adaptation algorithm is necessary.<br />

III. AN ADAPTIVE MESH SELECTION AL-<br />

GORITHM<br />

The efficient solution <strong>of</strong> <strong>the</strong> sub-problem (4), (5) implies<br />

<strong>the</strong> implementation <strong>of</strong> an adaptive algorithm for <strong>the</strong> numerical<br />

solution <strong>of</strong> 2pBVPs for ODEs, a well explored<br />

area in which some well established standard codes exist,<br />

see for instance [5] and references <strong>the</strong>rein. Never<strong>the</strong>less,<br />

most codes are using a technique for <strong>the</strong> equidistribution<br />

<strong>of</strong> <strong>the</strong> error function that changes <strong>the</strong> grid density function<br />

and <strong>the</strong> number <strong>of</strong> mesh points in <strong>the</strong> same step. But<br />

<strong>the</strong> estimation <strong>of</strong> <strong>the</strong> number <strong>of</strong> mesh points which is necessary<br />

to solve <strong>the</strong> BVP to <strong>the</strong> desired tolerance requirement<br />

is based on <strong>the</strong> density function <strong>of</strong> <strong>the</strong> last step and<br />

not on <strong>the</strong> updated one. So one is at great risk to solve to<br />

a much higher accuracy than requested, yielding longer<br />

302<br />

runtimes and much denser meshes than necessary.<br />

As a remedy for this difficulty an algorithm is designed<br />

to first adapt <strong>the</strong> mesh density φ on a coarse mesh and<br />

<strong>the</strong>n determine <strong>the</strong> number <strong>of</strong> points to satisfy <strong>the</strong> tolerance<br />

requirements.<br />

To ensure that this method works, it is necessary that <strong>the</strong><br />

mesh density φ stabilizes ra<strong>the</strong>r quickly. In Figure 3 <strong>the</strong><br />

evolutions <strong>of</strong> φ and N are shown for 12 iterations for a<br />

typical BVP. Note that after <strong>the</strong> sixth iteration <strong>the</strong> algorithm<br />

could have been stopped since <strong>the</strong> tolerance was<br />

satisfied.<br />

φ<br />

3<br />

2<br />

1<br />

12<br />

10<br />

8<br />

6<br />

Iteration<br />

4<br />

2<br />

Evolution <strong>of</strong> φ<br />

0<br />

x<br />

0.5<br />

1<br />

N<br />

200<br />

150<br />

100<br />

50<br />

Evolution <strong>of</strong> N<br />

2 4 6 8 10 12<br />

Iteration<br />

Figure 3: Evolution <strong>of</strong> <strong>the</strong> density function φ (left) and<br />

<strong>the</strong> number <strong>of</strong> mesh points N (right). Both are stable<br />

after eight iterations.<br />

ACKNOWLEDGMENTS<br />

This work was supported by <strong>the</strong> project ”Development<br />

<strong>of</strong> adaptive s<strong>of</strong>tware for singular boundary value problems”<br />

<strong>of</strong> <strong>the</strong> Austrian Science Fund (FWF) under grant<br />

no. P17253-N12, project leader: A.o. Univ. Pr<strong>of</strong>. Dr.<br />

Ewa B. Weinmüller.<br />

REFERENCES<br />

[1] I. Rach˚unková, O. Koch, G. Pulverer, and<br />

E. Weinmüller. On a singular boundary value problem<br />

arising in <strong>the</strong> <strong>the</strong>ory <strong>of</strong> shallow membrane caps.<br />

Math. Anal. and Appl., 332:523–541, 2007.<br />

[2] S. Stanek, G. Pulverer, and E. Weinmüller. Analysis<br />

and numerical solution <strong>of</strong> positive and dead core<br />

solutions <strong>of</strong> singular two-point boundary value problems.<br />

Comp. Math. Appl., 56:1820–1837, <strong>2008</strong>.<br />

[3] G. Kitzh<strong>of</strong>er, O. Koch, P. Lima, and E. Weinmüller.<br />

Efficient numerical solution <strong>of</strong> <strong>the</strong> density pr<strong>of</strong>ile<br />

equation in hydrodynamics. J. Sci. Comput., 32:411–<br />

424, 2006.<br />

[4] W. Auzinger, O. Koch, D. Praetorius, G. Pulverer,<br />

and E. Weinmüller. Performance <strong>of</strong> collocation<br />

s<strong>of</strong>tware for singular boundary value problems.<br />

Techn. Rep. Nr. 4/04, Inst. for Analysis and Scientific<br />

Computing, Vienna Univ. <strong>of</strong> Technology, Austria,<br />

2004.<br />

[5] W. Auzinger, G. Kneisl, O. Koch, and E. Weinmüller.<br />

A collocation code for boundary value problems in<br />

ordinary differential equations. Numer. Algorithms,<br />

33:27, 2003.


ICT – The Big Picture:<br />

Intertwined Driving Potentials:<br />

Innovation, Technology Penetration and <strong>the</strong> Global Economy<br />

DI Ranja Reda<br />

Vienna University <strong>of</strong> Technology<br />

Institute for Ma<strong>the</strong>matical Methods in Economics<br />

Vienna, Austria<br />

Email: ranja.reda@fam.tuwien.ac.at<br />

Abstract — The goal <strong>of</strong> this paper is to provide a<br />

novel analytical study <strong>of</strong> <strong>the</strong> Information Communication<br />

Technologies (ICT) sector. Future trends and<br />

<strong>the</strong> impact <strong>of</strong> this sector on <strong>the</strong> Global Economy are<br />

presented. Subsequently, <strong>the</strong> intertwined ICT driving<br />

potentials Innovation, Technology Penetration<br />

and <strong>the</strong> Global Economy are analyzed under <strong>the</strong><br />

spotlight <strong>of</strong> <strong>the</strong> recent subprime crisis.<br />

Recent studies showed that <strong>the</strong> global economy in<br />

2007 was not in a recession phase. Fur<strong>the</strong>rmore,<br />

for <strong>2008</strong> <strong>the</strong>re is no recession phase to be expected<br />

ei<strong>the</strong>r. Some sectors, e.g. Information Communication<br />

Technology, even exhibit a promising growth<br />

rate. However, in some sectors, especially industrial<br />

and infrastructure sector, <strong>the</strong> subprime crisis<br />

started to turn <strong>the</strong> global economy <strong>the</strong> o<strong>the</strong>r way<br />

around. This work presents an analytical study <strong>of</strong><br />

<strong>the</strong> Information Communication sector, which is<br />

considered to be <strong>the</strong> main driving potential <strong>of</strong> <strong>the</strong><br />

global economy. On <strong>the</strong> o<strong>the</strong>r hand, a healthy<br />

economy will boost <strong>the</strong> technology penetration<br />

which in turn will enhance economic growth. The<br />

intertwining <strong>of</strong> <strong>the</strong> global economy and technological<br />

progress in <strong>the</strong> communication sector is analyzed.<br />

Finally, <strong>the</strong> impact <strong>of</strong> <strong>the</strong> subprime crisis on <strong>the</strong><br />

global economy, and subsequently its connection to<br />

<strong>the</strong> communication technology sector is discussed.<br />

I. INTRODUCTION<br />

With <strong>the</strong> beginning <strong>of</strong> <strong>the</strong> third millennium, we experienced<br />

a unique technological progress rate, accompanied<br />

by a big transformation from <strong>the</strong> old<br />

economy to <strong>the</strong> new economy to <strong>the</strong> current agile<br />

economy, where technology and economy are more<br />

and more intertwined.<br />

From <strong>the</strong> service point <strong>of</strong> view, <strong>the</strong> advance from<br />

peer to peer VoIP, to triple play and finally to quadruple<br />

play enabled everyone using <strong>the</strong> appropriate<br />

access technology to be online, any time, everywhere.<br />

This was supported by <strong>the</strong> progress <strong>of</strong> wire-<br />

303<br />

less broadband technologies. Accordingly, more than<br />

70% <strong>of</strong> <strong>the</strong> global business turned to be online business,<br />

making use <strong>of</strong> <strong>the</strong> available advanced communication<br />

technologies.<br />

II. ICT, THE BIG PICTURE<br />

Figure 1: ICT, Technological Penetration and <strong>the</strong><br />

Global Economy [ 1 ]<br />

(Analysis to be provided in <strong>the</strong> full paper)<br />

III. INFORMATION TECHNOLOGY AND<br />

COMMUNICATION: THE MARKET<br />

POINT OF VIEW<br />

Information Communication Technology is considered<br />

as <strong>the</strong> motivator and facilitator for <strong>the</strong> global<br />

economy.<br />

The world entered <strong>the</strong> 21st century on a wave <strong>of</strong><br />

technological optimism, far from bringing <strong>the</strong> world<br />

to a halt during <strong>the</strong> economic difficulties. Information<br />

and Communication Technologies (ICT) and<br />

especially <strong>the</strong> Communication part, seem capable <strong>of</strong><br />

generating a new level <strong>of</strong> global prosperity.<br />

The progress <strong>of</strong> <strong>the</strong> world ICT Market is illustrated<br />

in Fig. 3.


Fig. 3: Worldwide ICT Market<br />

Convergence between <strong>the</strong> IT and telecom markets<br />

will be analyzed in <strong>the</strong> full paper.<br />

IV. COMMUNICATION MARKET<br />

The global communication market reached a record<br />

<strong>of</strong> over 1100 Billion € (see Fig.4). It showed <strong>the</strong><br />

following break down according to <strong>the</strong> main global<br />

economical domains [ 2 ], [ 3 ]:<br />

Fig. 4: Global Communication Market<br />

by Region<br />

From <strong>the</strong> growth rate point <strong>of</strong> view, BRIC are sustaining<br />

an amazing growth rate <strong>of</strong> 7,8 % p.a.,<br />

whereas <strong>the</strong> 4 Tigers are moving into <strong>the</strong> future with<br />

speed and skill, showing a remarkable growth rate <strong>of</strong><br />

more than 6%.<br />

Economical Domain Billion € %<br />

Europe 346 30,7<br />

US 243 21,6<br />

304<br />

Japan 164 14,68<br />

BRIC 97 9<br />

4 Tigers 11 1<br />

RoW 264 23<br />

BRIC: Brazil, Russia, India and China<br />

4 Tigers: Hong Kong, Singapore, S. Korea and Taiwan<br />

RoW: Rest <strong>of</strong> <strong>the</strong> World<br />

V. SUBPRIME CRISIS<br />

With <strong>the</strong> US as its epicentre, in late 2006 <strong>the</strong> subprime<br />

mortgage crisis triggered a global liquidity<br />

crunch putting downward pressure on economic<br />

growth. As <strong>the</strong> International Monetary Fund stated,<br />

<strong>the</strong> worldwide losses stemming from <strong>the</strong> US subprime<br />

mortgage crisis could hit 945 billion dollars as<br />

<strong>the</strong> impact spreads in <strong>the</strong> global economy.<br />

Today, trading strategies involve complex structures,<br />

high volumes <strong>of</strong> securities, enormous trading frequencies<br />

and high-level quantitative methods. Therefore,<br />

sophisticated technological tools like pricing<br />

s<strong>of</strong>tware and trading platforms are absolutely necessary<br />

to make faster, more accurate trading decisions.<br />

These technological tools played a fundamental role<br />

in <strong>the</strong> early stages <strong>of</strong> <strong>the</strong> US subprime mortgage<br />

crisis which now quickly spreads all over <strong>the</strong> world<br />

having a great impact on global economy. As early as<br />

2003, US investment guru Warren Buffet referred to<br />

quantitative instruments as “financial weapons <strong>of</strong><br />

mass destruction”.<br />

(Fur<strong>the</strong>r analysis, conclusion and references: to be<br />

provided in <strong>the</strong> full paper)<br />

VI. REFERENCES<br />

[1] R. Reda, ICT, e-Business, e-Commerce and<br />

Online Services. Technological Evolution, Future<br />

Trends and Impact to <strong>the</strong> Global Economy,<br />

<strong>Proceedings</strong> <strong>of</strong> <strong>the</strong> Networking and Electronic<br />

Commerce Research <strong>Conference</strong> <strong>2008</strong> (NAEC<br />

<strong>2008</strong>), Lake Garda, Italy, Sept. <strong>2008</strong><br />

[2] European Information Technology Observatory<br />

2007, EITO, Published by European Economic<br />

Interest Grouping (EEIG) , ISSN 0947-4862<br />

[3] R. Reda & N. Jordan, “Signpots for <strong>the</strong> Future<br />

<strong>of</strong> Mobile Communication” , e&I, elektronik<br />

und informationstechnik, heft 9.2006,published<br />

by Springer Wien New York, ISSN: 0932-383X<br />

EIEIEE 123(9) 361-408, a1-a44(2006)


Activation <strong>of</strong> Neural Networks in <strong>the</strong> Human Spinal Cord with<br />

Implanted and Surface Electrodes:<br />

Conclusions from Finite Element Simulations<br />

Josef Ladenbauer, Ursula S. H<strong>of</strong>stoetter, Karen Minassian,<br />

Milan R. Dimitrijevic and Frank Rattay<br />

Institute <strong>of</strong> Analysis and Scientific Computing<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

josef.ladenbauer@gmail.com<br />

Abstract — Using surface electrodes it is possible<br />

to communicate with neural networks in <strong>the</strong> human<br />

spinal cord in a similar way as with a small single<br />

implanted electrode, which is essentially closer to<br />

<strong>the</strong> neural elements. This surprising result is promising<br />

to become <strong>of</strong> great importance for paralyzed<br />

people because with <strong>the</strong> new non-invasive method<br />

some inherent functions <strong>of</strong> <strong>the</strong> spinal cord (neural<br />

circuits called pattern generators) controlling stepping<br />

and standing can be activated. With <strong>the</strong> help <strong>of</strong><br />

<strong>the</strong> activating function, a ma<strong>the</strong>matical concept<br />

derived from nerve fiber models, <strong>the</strong> presented<br />

computer simulations show a quite similar excitation<br />

situation for both stimulation methods.<br />

I. INTRODUCTION<br />

Since 1998 <strong>the</strong>re is an increasing hope to find a new<br />

clinical method for people which are paralyzed as a<br />

consequence <strong>of</strong> spinal cord injury. This hope is based<br />

on <strong>the</strong> observation that a train <strong>of</strong> constant pulses (e.g.<br />

7 Volt, 30 Hz) from a single active electrode, implanted<br />

close to <strong>the</strong> spinal cord, can result in stepping<br />

like movements in paralyzed patients, even when<br />

<strong>the</strong>ir lower spinal cord is completely separated from<br />

<strong>the</strong> brain [1].<br />

Based on previous work it can be assumed that this<br />

surprising phenomenon is caused by neural networks<br />

in <strong>the</strong> lumbar spinal cord, which are indirectly activated<br />

by artificially generated neural impulses from<br />

<strong>the</strong> largest diameter nerve fibers originating in <strong>the</strong><br />

lower limbs [2,3]. These fibers enter <strong>the</strong> spinal cord<br />

in <strong>the</strong> vicinity <strong>of</strong> <strong>the</strong> active electrode within <strong>the</strong> socalled<br />

posterior roots. Every action potential (AP) in<br />

<strong>the</strong> population <strong>of</strong> <strong>the</strong> stimulated nerve fibers operates<br />

as input to <strong>the</strong> neural network. In case <strong>of</strong> proper input<br />

sequences for walking, <strong>the</strong> network sends patterned<br />

signals to <strong>the</strong> muscles <strong>of</strong> <strong>the</strong> legs, leading to rhythmic<br />

contraction and relaxation phases.<br />

Nobody could believe that this principle <strong>of</strong> artificial<br />

pattern generator activation would be possible with<br />

305<br />

surface electrodes, because such electrodes are much<br />

fur<strong>the</strong>r away from <strong>the</strong> target neurons and additionally<br />

<strong>the</strong>y are essentially bigger in size. These two properties<br />

are assumed not to allow <strong>the</strong> focused stimulating<br />

effects needed to activate <strong>the</strong> pattern generator.<br />

Motivated by some experimental successes (Fig. 1),<br />

[4], we demonstrate with <strong>the</strong> presented simulation<br />

study that <strong>the</strong> general concerns are wrong and focusing<br />

is strong enough to activate <strong>the</strong> neural pattern<br />

generators by <strong>the</strong> same principle using surface electrodes<br />

instead <strong>of</strong> implanted ones.<br />

Figure 1: Changes <strong>of</strong> <strong>the</strong> H<strong>of</strong>fmann reflex during<br />

stimulation <strong>of</strong> <strong>the</strong> spinal cord with surface electrodes.<br />

II. METHOD AND RESULTS<br />

The artificial activation <strong>of</strong> <strong>the</strong> pattern generators in<br />

<strong>the</strong> human lumbar spinal cord is caused by fiber<br />

excitation in <strong>the</strong> vicinity <strong>of</strong> <strong>the</strong> stimulating electrode.<br />

A single target fiber can be simulated as a compartment<br />

model, where <strong>the</strong> transmembrane voltage Vn <strong>of</strong><br />

<strong>the</strong> n-th compartment in response to an external potential<br />

Ve can be calculated as<br />

dVn<br />

Vn<br />

� 1 �Vn<br />

Vn<br />

� 1 �Vn<br />

�[<br />

�Iion,<br />

n �<br />

�<br />

dt<br />

Rn<br />

� 1/<br />

2 � Rn<br />

/ 2 Rn<br />

� 1/<br />

2 � Rn<br />

/ 2 (1)<br />

Ve,<br />

n � 1 �Ve,<br />

n Ve,<br />

n � 1�<br />

Ve,<br />

n<br />

+<br />

� ] / Cn<br />

Rn<br />

� 1/<br />

2 � Rn<br />

/ 2 Rn<br />

� 1/<br />

2 � Rn<br />

/ 2<br />

where C is <strong>the</strong> membrane capacitance and R/2 <strong>the</strong><br />

internal resistance between <strong>the</strong> center and <strong>the</strong> border


Figure 2: A-D describe <strong>the</strong> situation <strong>of</strong> an implanted dipole, where <strong>the</strong> lower electrode contact in (A)<br />

stimulates a sensory nerve fiber which enters <strong>the</strong> spinal cord at a position marked by arrow. At this position,<br />

which is marked also by a dotted line in (B) and (C) <strong>the</strong> activating function is quite large and initiates<br />

an AP that propagates in both directions (D). The 2 diagrams at <strong>the</strong> left side describe <strong>the</strong> same target<br />

fiber in <strong>the</strong> same region when stimulated with surface electrodes. The electrical potential pr<strong>of</strong>ile is similar<br />

to case B, and <strong>the</strong> activating function is comparable with (C) in <strong>the</strong> region <strong>of</strong> interest.<br />

<strong>of</strong> <strong>the</strong> compartment. Ion membrane current Iion,n is<br />

assumed to be proportional to Vn (linear analysis) or<br />

can be calculated with a model <strong>of</strong> <strong>the</strong> Hodgkin-<br />

Huxley type. The term<br />

Ve,<br />

n � 1 �Ve<br />

, n Ve,<br />

n � 1 �Ve<br />

, n<br />

f n �[<br />

� ] / Cn<br />

(2)<br />

Rn � 1 / 2 � Rn / 2 Rn � 1 / 2 � Rn<br />

/ 2<br />

is called <strong>the</strong> generalized activating function [5]. f is<br />

<strong>the</strong> driving force for every compartment and an AP is<br />

expected to be generated in a region where f>0.<br />

The excitation <strong>of</strong> selected target fibers is simulated in<br />

a two step procedure. First <strong>the</strong> potential distribution<br />

is calculated with <strong>the</strong> finite element s<strong>of</strong>tware COM-<br />

SOL Multiphysics for implanted or surface electrodes.<br />

In <strong>the</strong> second step, <strong>the</strong> potential distribution<br />

along <strong>the</strong> pathway <strong>of</strong> a specific fiber is used as input<br />

data for <strong>the</strong> compartment model, where <strong>the</strong> ion currents<br />

in <strong>the</strong> active membrane are calculated with <strong>the</strong><br />

CRRSS model [6].<br />

The potential distribution along a typical nerve fiber<br />

pathway <strong>of</strong> a sensory neuron is shown in <strong>the</strong> leftmost<br />

part <strong>of</strong> Fig. 2 for surface stimulation with a pair <strong>of</strong><br />

small active electrodes, see [4] for stimulation details.<br />

For <strong>the</strong> same neuron geometry (Fig. 2A) <strong>the</strong><br />

potential distribution was calculated again with finite<br />

elements and found a ra<strong>the</strong>r similar relation. The<br />

stimulation is focused within a region <strong>of</strong> strong curvature<br />

and spike initiation is also supported by <strong>the</strong><br />

change <strong>of</strong> extracellular conductance at <strong>the</strong> entry point<br />

<strong>of</strong> <strong>the</strong> fiber into <strong>the</strong> spinal cord (arrow in Fig. 2A).<br />

III. CONCLUSION<br />

With appropriate positions for surface and implanted<br />

electrodes, similar characteristic shapes <strong>of</strong> <strong>the</strong> acti-<br />

306<br />

vating functions for <strong>the</strong> same sensory target neurons<br />

are obtained. The shapes <strong>of</strong> <strong>the</strong> activating functions<br />

predict corresponding spike initiation regions and<br />

<strong>the</strong>reby a comparable neural input for <strong>the</strong> pattern<br />

generator networks within <strong>the</strong> spinal cord. Thus,<br />

similar effects can be expected in both stimulation<br />

strategies, ei<strong>the</strong>r using surface electrodes or implants.<br />

REFERENCES<br />

[1] M. R. Dimitrijevic, Y. Gerasimenko, and M. M.<br />

Pinter. Evidence for a spinal central pattern generator<br />

in humans. Annals <strong>of</strong> <strong>the</strong> New York Academy<br />

<strong>of</strong> Sciences, 860: 360–376, 1998.<br />

[2] F. Rattay, K. Minassian and M. R. Dimitrijevic.<br />

Epidural electrical stimulation <strong>of</strong> posterior structures<br />

<strong>of</strong> <strong>the</strong> human lumbosacral cord: 2. Quantitative<br />

analysis by computer modeling. Spinal<br />

Cord, 38: 473–489, 2000.<br />

[3] K. Minassian. Modeling <strong>of</strong> a human spinal pattern<br />

generator for locomotion and its activation<br />

by electrical epidural stimulation. PhD Thesis.<br />

TU Vienna, 2004<br />

[4] U.S. H<strong>of</strong>stoetter, K. Minassian, C. H<strong>of</strong>er, W.<br />

Mayr, F. Rattay and M.R. Dimitrijevic. Modification<br />

<strong>of</strong> reflex responses to lumbar posterior<br />

root stimulation by motor tasks in healthy subjects.<br />

Artificial Organs 32(8):644–648, <strong>2008</strong>.<br />

[5] F. Rattay. The basic mechanism for <strong>the</strong> electrical<br />

stimulation <strong>of</strong> <strong>the</strong> nervous system. Neuroscience<br />

89, 335-346, 1999.<br />

[6] F. Rattay, R. J. Greenberg and S. Resatz. Neuron<br />

Modeling. In Handbook <strong>of</strong> Neuropros<strong>the</strong>tic Methods.<br />

Eds. W.E. Finn and P.G. LoPresti. CRC<br />

Press, 39-71, 2003.


Characterisation <strong>of</strong> biomass smoke particle emissions and its<br />

application to source receptor analysis<br />

Christoph Schmidl and Hans Puxbaum<br />

Institute <strong>of</strong> Chemical Technologies and Analytics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: Christoph.schmidl@tuwien.ac.at<br />

Abstract — An extensive study testing various<br />

biomass combustion systems and biomass fuel types<br />

was performed in order to investigate particulate<br />

matter emissions. Detailed chemical pr<strong>of</strong>iles were<br />

developed and applied in PM10 source apportionment<br />

studies using a chemical mass balance approach.<br />

Chemical species were identified that enable<br />

<strong>the</strong> differentiation between smoke <strong>of</strong> hard- and<br />

s<strong>of</strong>twood types. Fur<strong>the</strong>rmore it could be shown that<br />

<strong>the</strong> use <strong>of</strong> two different biomass combustion pr<strong>of</strong>iles,<br />

one for manually and one for automatically<br />

fired systems, brings a significant improvement to<br />

source apportionment results.<br />

I. INTRODUCTION<br />

Wood smoke is increasingly being seen as an important<br />

component <strong>of</strong> airborne particulate matter<br />

(PM). Recent source apportionment studies in<br />

Austria (AQUELLA) and Europe (CARBOSOL)<br />

showed that wood smoke is <strong>the</strong> main primary<br />

particle source during <strong>the</strong> winter half year [1].<br />

Emissions <strong>of</strong> this source are dominated by organic<br />

and elemental carbon, toge<strong>the</strong>r constituting around<br />

70% <strong>of</strong> total particle emissions. Therefore it is<br />

obvious that wood combustion is one <strong>of</strong> <strong>the</strong> major<br />

primary sources <strong>of</strong> carbonaceous particles in <strong>the</strong><br />

atmosphere as well. The source apportionment<br />

studies mentioned above applied a macrotracer<br />

concept using levoglucosan as single tracer for<br />

wood combustion [2]. Additionally <strong>the</strong> Austrian<br />

ambient dataset was analysed with a CMB model<br />

[3,4] approach to evaluate <strong>the</strong> macrotracer results.<br />

However experience with CMB showed problems<br />

with micro components, i.e. organic micro tracers,<br />

probably due to low concentrations and partly<br />

limited stability in <strong>the</strong> atmosphere and resulting<br />

difficulties in analysis.<br />

II. EXPERIMENTAL<br />

To make use <strong>of</strong> <strong>the</strong> advantages <strong>of</strong> macrotracer and<br />

CMB model we developed a combination <strong>of</strong> <strong>the</strong><br />

two approaches. The number <strong>of</strong> species was reduced<br />

so that we are running <strong>the</strong> CMB model on a<br />

307<br />

number <strong>of</strong> 14 macro-compounds omitting organic<br />

micro tracers. This gave good correlations for<br />

most compounds except potassium where <strong>the</strong><br />

measured values were consistently around one<br />

order <strong>of</strong> magnitude higher than <strong>the</strong> calculated (see<br />

marking in figure 1a).<br />

Figure 1: Measured/Calculated comparison (CMB);<br />

Original model result(a); with wood boiler pr<strong>of</strong>ile(b).<br />

Hence, fur<strong>the</strong>r improvement <strong>of</strong> source pr<strong>of</strong>iles was<br />

necessary. We <strong>the</strong>refore investigated a number <strong>of</strong><br />

wood combustions facilities – reaching from smallscale<br />

stoves in <strong>the</strong> power range <strong>of</strong> some kilowatts<br />

(kW) to mid-scale boilers (50kW) – regarding to<br />

<strong>the</strong>ir particle emissions. A sampling system was<br />

designed, that allows PM samples to be collected<br />

directly out <strong>of</strong> <strong>the</strong> chimney <strong>of</strong> wood combustion<br />

appliances. The exhaust gas was diluted with precleaned<br />

air to simulate ambient conditions. For size<br />

segregation an impactor – type low volume sampling<br />

head (Digitel) was used. PM10 samples were collected<br />

afterwards on 8 filters, 6 quartz fibre and 2<br />

cellulose filters, in parallel at room temperature. All<br />

samples were analysed for a number <strong>of</strong> organic and<br />

inorganic compounds. From <strong>the</strong> obtained dataset<br />

source pr<strong>of</strong>iles for different wood and application<br />

types were developed and analysed with respect to


differences and similarities. Figure 2 gives an overview<br />

<strong>of</strong> <strong>the</strong> sampling equipment.<br />

Exhaust<br />

gas<br />

Dilution<br />

Dilution air<br />

(filtered, dried)<br />

Figure 2: Scheme <strong>of</strong> PM10 sampling system.<br />

III. RESULTS<br />

Mixing<br />

Cooling<br />

Condensation<br />

Separation<br />

PM10/PM2.5<br />

Collection<br />

8 Filters parallel<br />

Results showed that almost all differences in emission<br />

pr<strong>of</strong>iles within <strong>the</strong> group <strong>of</strong> manually fired small<br />

scale combustion units are rendered insignificant,<br />

mainly due to <strong>the</strong> high variability <strong>of</strong> <strong>the</strong> combustion<br />

process, which is in good agreement with previous<br />

studies [5]. Therefore it was decided to use one<br />

source pr<strong>of</strong>ile for this group. Never<strong>the</strong>less specific<br />

compounds were identified that allow a distinction <strong>of</strong><br />

smoke from hardwood- and s<strong>of</strong>twoods. Levoglucosan<br />

and mannosan, both anhydrosugars originating<br />

from pyrolysis <strong>of</strong> cellulose and hemicelluloses, were<br />

found to differ significantly between wood types.<br />

Expressed as ratio <strong>the</strong> difference becomes even more<br />

marked, with higher ratios <strong>of</strong> around 15 found in<br />

hardwoods and lower ratios <strong>of</strong> around 3 in s<strong>of</strong>twoods.<br />

Figure 3 gives a scatter plot <strong>of</strong> levoglucosan<br />

and mannosan results from tested appliance and<br />

wood types displaying this difference.<br />

Figure 3: Scatter plot <strong>of</strong> levoglucosan/mannosan<br />

ratios for different wood and appliance types<br />

Combustion conditions in automatically fired systems,<br />

however, are pretty stable after a short start-up<br />

phase and <strong>the</strong>refore also emissions are more constant.<br />

As <strong>the</strong> efficiency <strong>of</strong> combustion is much higher, <strong>the</strong><br />

chemical composition <strong>of</strong> emissions differs significantly<br />

to manually fired appliances. Emissions from<br />

wood stoves are dominated by carbonaceous fractions<br />

(total carbon ~ 70%) while wood boilers mainly<br />

emit inorganic material (sum <strong>of</strong> ions ~ 60%).<br />

We <strong>the</strong>refore decided to use two different wood<br />

combustion pr<strong>of</strong>iles in CMB modelling, one for<br />

308<br />

small scale manually fired wood stoves, and <strong>the</strong> o<strong>the</strong>r<br />

for automatically fired systems (wood boilers). The<br />

enhanced model gave similar results for most compounds<br />

and a much better fit for potassium which can<br />

be explained by <strong>the</strong> high emission concentration <strong>of</strong><br />

potassium in <strong>the</strong> new source pr<strong>of</strong>ile (see figure 1b).<br />

Wood smoke source pr<strong>of</strong>iles and results from CMBbased<br />

source apportionment for Austrian PM10 data<br />

are presented.<br />

ACKNOWLEDGMENTS<br />

This study was funded by <strong>the</strong> European Union Interreg<br />

III Programme, <strong>the</strong> Austrian Federal Ministry <strong>of</strong><br />

Agriculture, Forestry, Environment and Water Management,<br />

<strong>the</strong> government <strong>of</strong> Lower Austria and klimaaktiv<br />

– Holzwärme.<br />

The authors want to thank all members <strong>of</strong> <strong>the</strong><br />

AQUELLA working group at <strong>the</strong> Institute <strong>of</strong> Chemical<br />

Technologies and Analytics at <strong>the</strong> Vienna University<br />

<strong>of</strong> Technology for doing analysis work. A<br />

special thanks goes to <strong>the</strong> institutions providing <strong>the</strong><br />

test stands equipment: <strong>the</strong> Institute <strong>of</strong> Chemical Engineering,<br />

Vienna University <strong>of</strong> Technology, <strong>the</strong><br />

Austrian Bioenergy Centre and <strong>the</strong> BLT Wieselburg.<br />

REFERENCES<br />

[1] H. Puxbaum, A. Caseiro, A. Sánchez-Ochoa, A.<br />

Kasper-Giebl, M. Claeys, A. Gelencsér, M. Legrand,<br />

S. Preunkert. C. Pio. Levoglucosan levels<br />

at background sites in Europe for assessing <strong>the</strong><br />

impact <strong>of</strong> biomass combustion on <strong>the</strong> European<br />

aerosol background. Journal <strong>of</strong> Geophysical Research<br />

112(D23), D23S05/1-D23S05/11, 2007.<br />

[2] B.R.T. Simoneit, J.J. Schauer, C.G. Nolte, D.R.<br />

Oros, V.O. Elias, M.P. Fraser, W.F. Rogge, G.R.<br />

Cass. Levoglucosan, a tracer for cellulose in<br />

biomass burning and atmospheric particles. Atmospheric<br />

Environment 33, 173 – 182, 1999.<br />

[3] J.G. Watson. Overview <strong>of</strong> receptor model principles.<br />

Journal <strong>of</strong> Air Pollution Control Association<br />

34, 619-623, 1984.<br />

[4] J.J Schauer, W.F. Rogge, L.M. Hildemann, M.A.<br />

Mazurek, G.R. Cass, B.R.T. Simoneit. Source<br />

apportionment <strong>of</strong> airborne particulate matter using<br />

organic compounds as tracers. Atmospheric<br />

Environment, 30, 3837-3855, 1996.<br />

[5] C. Schmidl, I.L. Marr, A. Caseiro, P. Kotianova,<br />

A. Berner, H. Bauer, A. Kasper-Giebl, H. Puxbaum.<br />

Chemical characterization <strong>of</strong> fine particle<br />

emissions from wood stove combustion <strong>of</strong> common<br />

woods growing in mid-European Alpine regions.<br />

Atmospheric Environment 42, 126-141,<br />

<strong>2008</strong>.


Observations <strong>of</strong> <strong>the</strong> surface evolution <strong>of</strong> Pinka-plain:<br />

mass movements and neotectonics<br />

Gábor Kovács, Dr. Balázs Székely (Faculty Mentor), Dr. Sándor Papp (Faculty Mentor)<br />

Institute <strong>of</strong> Geography and Earth Sciences<br />

Eötvös Loránd University<br />

Budapest, Hungary<br />

Email: s_kovacs_gabor@hotmail.com<br />

Abstract — Pinka Plain is located at <strong>the</strong> western<br />

part <strong>of</strong> <strong>the</strong> Pannonian Basin. The margin <strong>of</strong> <strong>the</strong><br />

plateau on north is a scarp surface <strong>of</strong> <strong>the</strong> valley <strong>of</strong><br />

Arany Creek. In <strong>the</strong> first half <strong>of</strong> <strong>the</strong> 1990’s a<br />

landslide arose on a ski-track situated 1 km west<br />

from <strong>the</strong> city <strong>of</strong> Szomba<strong>the</strong>ly. Our goal was to<br />

investigate this phenomenon, especially <strong>the</strong> inner<br />

structure <strong>of</strong> <strong>the</strong> moved mass by shallow boreholes,<br />

and to measure recent velocity <strong>of</strong> movement with<br />

fixed points. In <strong>the</strong> period <strong>of</strong> investigation this <strong>the</strong><br />

toe <strong>of</strong> <strong>the</strong> landslide didn’t move, but o<strong>the</strong>r<br />

landforms in <strong>the</strong> vicinity reveal that <strong>the</strong> scarp<br />

surface is still active. Comparison <strong>of</strong> our results<br />

with tectonic data and <strong>the</strong> crystalline basement<br />

structure reveal <strong>the</strong> neotectonic origin <strong>of</strong><br />

movements in <strong>the</strong> study area. Therefore <strong>the</strong> scarp is<br />

expected to be active for long term.<br />

I. INTRODUCTION<br />

The surface evolution <strong>of</strong> western part <strong>of</strong> Pannonian<br />

Basin is largely determined by <strong>the</strong> uplift <strong>of</strong> <strong>the</strong><br />

Eastern Alps and <strong>the</strong> tectono-geomorphic processes<br />

<strong>of</strong> <strong>the</strong> basin. The reason to examine this area was<br />

<strong>the</strong> abundant occurrence <strong>of</strong> mass movements and<br />

gullies on <strong>the</strong> nor<strong>the</strong>rn scarp. Our assumption was<br />

that <strong>the</strong> above processes are, at least partly, <strong>of</strong><br />

neotectonic origin.<br />

The general morphology south from <strong>the</strong> K�szegi<br />

Mountains shows typical SSE slopes with low relief<br />

cut <strong>of</strong>f by 30–60 meter high scarps followed by<br />

streams. We examined <strong>the</strong> surface development<br />

processes on <strong>the</strong> steep scarp <strong>of</strong> <strong>the</strong> nor<strong>the</strong>rn border<br />

<strong>of</strong> Pinka Plain. Previous studies examined<br />

individual mass movements [1], general<br />

morphology, and general <strong>the</strong> tectonic setting <strong>of</strong> <strong>the</strong><br />

area [2] but <strong>the</strong> relation <strong>of</strong> <strong>the</strong>se to <strong>the</strong> surface<br />

evolution has not been examined yet. Our goal was<br />

to investigate this relationship.<br />

An old ski-track is situated between<br />

Szomba<strong>the</strong>ly, and village <strong>of</strong> Sé. In <strong>the</strong> 1990’s, a<br />

landslide happened that hampered its fur<strong>the</strong>r use.<br />

The understanding <strong>of</strong> this mass movement was <strong>the</strong><br />

main reason for our investigation.<br />

309<br />

II. THE INVESTIGATION OF THE LANDSLIDE<br />

The slope became sensitive to <strong>the</strong> erosion due to <strong>the</strong><br />

wood cutting on <strong>the</strong> ski-track. In <strong>the</strong> literature [3]<br />

authors consider <strong>the</strong> steep, surface parallel layering<br />

<strong>of</strong> <strong>the</strong> sediments as <strong>the</strong> most important reason for<br />

<strong>the</strong> previously occurred mass movements in our<br />

study area.<br />

A. SHALLOW BOREHOLES<br />

We drilled shallow boreholes along <strong>the</strong> landslide to<br />

check this assumed parallel structure. The clay and<br />

sand layers alternating almost parallel to <strong>the</strong><br />

surface, and <strong>the</strong> straight oxidated clay layer as <strong>the</strong><br />

margin between <strong>the</strong> wet and dry sediments verified<br />

this assumption.<br />

We couldn’t analyze <strong>the</strong> role <strong>of</strong> variation in<br />

precipitation, because <strong>the</strong> exact time <strong>of</strong> <strong>the</strong><br />

landslide, being far away from urban area, is not<br />

known. According to different maps about this area,<br />

<strong>the</strong> landslide occurred in <strong>the</strong> first half <strong>of</strong> <strong>the</strong> 1990’s.<br />

B. OTHER PHENOMENA<br />

Breakdowns and ruptures also indicate <strong>the</strong> slope<br />

instability, but haven’t been investigated due to <strong>the</strong><br />

possible artificial influence. The evidences <strong>of</strong> creep<br />

are deformed artificial structures, for example tilted<br />

fences or ruptures on walls. We tried to measure <strong>the</strong><br />

velocity <strong>of</strong> movement with fixed checkpoints, but<br />

<strong>the</strong> moderate wea<strong>the</strong>r in winter 2006-2007 and<br />

2007-<strong>2008</strong> retarded this.<br />

On <strong>the</strong> steep slope a lot <strong>of</strong> similar gullies with<br />

different state <strong>of</strong> development can be find, too. The<br />

investigated gully is conspicuously characterized by<br />

high erosive power: <strong>the</strong>re are no sediment or even a<br />

layer <strong>of</strong> death fallen leaves on <strong>the</strong> valley floor after<br />

low amount <strong>of</strong> precipitation. The gully represents<br />

cascade long pr<strong>of</strong>ile with numerous knickpoints due<br />

to <strong>the</strong> alternating erodibility <strong>of</strong> sediments in base<br />

rock. Visible layers in <strong>the</strong> valley side also show <strong>the</strong><br />

near parallelism to <strong>the</strong> surface.


III. THE DISTRIBUTION OF PHENOMENA<br />

The geomorphologic features occur on <strong>the</strong> entire 6<br />

km long nor<strong>the</strong>rn steep margin <strong>of</strong> <strong>the</strong> Pinka-plain.<br />

The apparent distribution <strong>of</strong> <strong>the</strong> mass movements is<br />

strongly correlates with <strong>the</strong> area <strong>of</strong> clear-cuts,<br />

especially where <strong>the</strong> base rock is <strong>the</strong> upper<br />

Pannonian sediment with <strong>the</strong> aforementioned<br />

alternating stratification. We investigated also <strong>the</strong><br />

spatial distribution <strong>of</strong> <strong>the</strong> gullies.<br />

Where <strong>the</strong> base rock is <strong>the</strong> Pleistocene gravel,<br />

<strong>the</strong>re <strong>the</strong> relief is low, and <strong>the</strong> surface is less<br />

dissected. However, if <strong>the</strong> base rock is <strong>of</strong> upper<br />

Pannonian strata <strong>of</strong> alternating sediments, <strong>the</strong><br />

Tihany Formation, <strong>the</strong> relief is higher, and <strong>the</strong><br />

appearance <strong>of</strong> gullies is more frequent. The reason<br />

for that is, that <strong>the</strong> precipitation immediately<br />

infiltrate into <strong>the</strong> gravel sediment without relevant<br />

erosion, contrary to <strong>the</strong> Pannonian base rock that<br />

include compact impermeable clay and sand layers<br />

with higher erodibility.<br />

IV. RELATION TO THE STRUCTURAL<br />

LINEAMENTS<br />

A conclusion <strong>of</strong> our investigation <strong>of</strong> <strong>the</strong> surface<br />

development is that <strong>the</strong> steep slope is a result <strong>of</strong><br />

neotectonic uplifting or an asymmetric incision.<br />

The topography pretends like tilted platforms with<br />

stream pattern with two main directions. This form<br />

repeats between <strong>the</strong> different gravel platforms and<br />

also inside <strong>the</strong>m. The neotectonic structure was<br />

already considered in previous investigations [4]<br />

just like <strong>the</strong> incision <strong>the</strong>ory [5]. These contrasting<br />

<strong>the</strong>ories and <strong>the</strong> active nor<strong>the</strong>rn slope <strong>of</strong> Pinkaplain<br />

motivated us to investigate <strong>the</strong> area by GIS<br />

technology.<br />

The asymmetrical valley was analyzed using data<br />

<strong>of</strong> <strong>the</strong> densely spaced exploration wells by<br />

geostatistical evaluation <strong>of</strong> methods [6]. Data for<br />

three bedding plane and <strong>the</strong> surface was used for<br />

generating 3D DTMs. All models shows <strong>the</strong> scarp,<br />

less and less steep but situated more and more<br />

towards N.<br />

Based on <strong>the</strong> results outlined above, we assume<br />

that <strong>the</strong> investigated scarp is related to a listric fault.<br />

V. SUMMARY AND OUTLOOK<br />

The steep scarp <strong>of</strong> <strong>the</strong> Pinka plain is found to be an<br />

actively forming slope. Our observations endorse<br />

<strong>the</strong> concept <strong>of</strong> <strong>the</strong> neotectonic origin <strong>of</strong> <strong>the</strong><br />

movements. Accordingly, <strong>the</strong> instability <strong>of</strong> <strong>the</strong><br />

slope is expected to be present in <strong>the</strong> future,<br />

<strong>the</strong>refore surface development phenomena are<br />

310<br />

going to happen. Geophysical measurements would<br />

be useful to get more detailed insights <strong>of</strong> <strong>the</strong><br />

faulting.<br />

It would be interesting to study <strong>the</strong> structural<br />

movements, <strong>the</strong> impact <strong>the</strong> recent surface<br />

morphology, and also <strong>the</strong> effects <strong>of</strong> <strong>the</strong> differential<br />

uplift: 0 mm/yr vertical movement marker isoline is<br />

assumed to cross our study area [7]. A fur<strong>the</strong>r<br />

interesting question: Are <strong>the</strong>re any relation between<br />

<strong>the</strong> general morphology and <strong>the</strong> similar structure <strong>of</strong><br />

<strong>the</strong> pre-Tertiary basement [8]? We also plan to<br />

investigate <strong>the</strong> active periods <strong>of</strong> <strong>the</strong> surface<br />

development phenomenon.<br />

ACKNOWLEDGEMENTS<br />

The study has been carried out in <strong>the</strong> framework <strong>of</strong><br />

No. T47104 project <strong>of</strong> <strong>the</strong> Hungarian National<br />

Research Fund OTKA. We thank <strong>the</strong> assistance in<br />

fieldwork <strong>of</strong> ZS. CSÁKÁNY, P. GADÁNYI, A.<br />

HORVÁTH, GY. ISZTIN, P. KOVÁCS, Z. MAGYAR, A.<br />

NAGY, G. ORBÁN, É. SPANNER, B. TÓTH. In <strong>the</strong><br />

evaluation <strong>of</strong> <strong>the</strong> data B. KOHÁN, DR. J. KOVÁCS<br />

and A. ZÁMOLYI gave useful advices. All help is<br />

gratefully acknowledged.<br />

REFERENCES<br />

[1] SZILÁGYI E. 1989. Vas megye<br />

felszínmozgásainak katasztere. Földr. Ért. 38.<br />

1-2. pp. 33–54.<br />

[2] HORVÁTH, F. et al. 2006. A Pannon-medence<br />

jelenkori geodinamikájának atlasza: Eurokonform<br />

térképsorozat és magyarázó. Magyar<br />

Ge<strong>of</strong>izika, 47, 133-137.<br />

[3] KECSKÉS T. 1968. A szomba<strong>the</strong>lyi<br />

dombcsúszás. Vasi Szemle 22. 4. pp. 557–566<br />

[4] ÁDÁM L. 1962. A Rábántúli kavicstakaró. In:<br />

ÁDÁM L. et al.: Néhány dunántúli<br />

geomorfológiai körzet jellemzése. Földr. Ért.<br />

11. 1. pp. 41–52.<br />

[5] JASKÓ S. 1964. A nyugat-vas megyei<br />

barnak�szénterület. Földt. Kut. 7. 2–3 pp. 24–<br />

48.<br />

[6] FÜST A. 1997. Geostatisztika. Eötvös Kiadó.<br />

Budapest. pp. 18-69.<br />

[7] JOÓ, I. 1992. Recent vertical surface<br />

movements in <strong>the</strong> Carpathian Basin. –<br />

Tectonophysics. 202. pp. 129–134.<br />

[8] KILÉNYI, É. and ŠEFARA, J. (ed.) 1989. Pre-<br />

Tertiary basement contour map <strong>of</strong> <strong>the</strong><br />

Carpathian basin beneath Austria,<br />

Czechoslovakia and Hungary. 1: 500 000.<br />

ELGI Budapest, Ge<strong>of</strong>yzika Brno, Bratislava<br />

Branch


Spray Can Geometry<br />

Gwen Wilke and Andrew U. Frank (Faculty Mentor)<br />

Institute for Geoinformation and Cartography<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: {wilke,frank}@geoinfo.tuwien.ac.at<br />

Abstract — A model <strong>of</strong> axiomatic projective geometry for geographic<br />

information systems (GIS) is proposed that incorporates<br />

positional uncertainty. We define ma<strong>the</strong>matical models <strong>of</strong><br />

extended points and lines that reflect <strong>the</strong> extended and uncertain<br />

character <strong>of</strong> geographic features. Geometric operations<br />

in this model exhibit tolerance for positional uncertainty. The<br />

axiomatic approach ensures <strong>the</strong> consistency <strong>of</strong> geometric reasoning<br />

despite <strong>the</strong> uncertainties in <strong>the</strong> locations.<br />

I. INTRODUCTION<br />

Geometric functionality in current vector based geographic<br />

information systems (GIS) is based on infinitely<br />

small points and infinitely thin lines. This is in contrast<br />

to <strong>the</strong> fact that geographic features and <strong>the</strong>ir representation<br />

are extended and uncertain in location. Object representation<br />

in GIS is discrete and <strong>of</strong> limited computational<br />

accuracy and <strong>of</strong>ten destroys <strong>the</strong> consistency <strong>of</strong> geometric<br />

reasoning. A more realistic model <strong>of</strong> geometry is needed.<br />

The present approach is based on Menger’s axiomatic<br />

system for projective geometry [1]. We refer to a geometry<br />

as an algebra, consisting <strong>of</strong> a set <strong>of</strong> primitive objects<br />

and operations and a set <strong>of</strong> axioms. Every interpretation<br />

<strong>of</strong> primitives that complies with <strong>the</strong> axioms is considered<br />

to be a model for <strong>the</strong> geometry. In <strong>the</strong> proposed work<br />

interpretations <strong>of</strong> <strong>the</strong> primitive objects point and line are<br />

extended in all dimensions. Interpretations <strong>of</strong> <strong>the</strong> primitive<br />

operations connection and intersection exhibit tolerance<br />

for positional uncertainty in <strong>the</strong>ir input.<br />

The advantage <strong>of</strong> <strong>the</strong> axiomatic approach over o<strong>the</strong>r<br />

more pragmatic efforts in <strong>the</strong> past is that <strong>the</strong> implementation<br />

<strong>of</strong> basic operations is sufficient to derive compound<br />

objects and operations, analytical functions, and tests for<br />

relations. It guarantees concistency in geometric reasoning.<br />

II. MOTIVATION<br />

The idea for <strong>the</strong> proposed research emerged from a<br />

keynote speech given by Lotfi Zadeh: The name Spray<br />

Can Geometry refers to “a primitive world in which figures<br />

are drawn with a spray can, with no ruler or compass<br />

available” [2]. Many conventional approaches to<br />

geometric uncertainty modeling for GIS are built upon<br />

underlying exact operations and idealized object representation.<br />

By contrast, <strong>the</strong> present work aims at a flexible<br />

311<br />

representation that tolerates uncertainty in location while<br />

preserving <strong>the</strong> consistency <strong>of</strong> <strong>the</strong> axiomatic system.<br />

The Spray Can Geometry for GIS assumes interpretations<br />

<strong>of</strong> geometric primitives that reflect <strong>the</strong> extended<br />

character <strong>of</strong> geographic features and <strong>the</strong> positional uncertainty<br />

in <strong>the</strong>ir representation. These Spray Can primitives<br />

must meet a number <strong>of</strong> requirements in order to be viable<br />

for practical application in GIS. The most important requirement<br />

for <strong>the</strong> processing in a computer is closedness<br />

<strong>of</strong> operations. This is ensured by <strong>the</strong> axiomatic approach:<br />

Every set <strong>of</strong> interpretations that complies with geometric<br />

axioms automatically is operationally closed. Apart from<br />

this, two main criteria have been identified:<br />

1. Spray Can Points and Spray Can Lines are uncertain<br />

in all dimensions.<br />

2. Spray Can Operations exhibit tolerance for positional<br />

uncertainty in <strong>the</strong>ir input.<br />

A model that violates criterion 1 does not account for<br />

<strong>the</strong> real character <strong>of</strong> geographic features. A model that<br />

violates criterion 2 is isomorphic to <strong>the</strong> exact model; <strong>the</strong><br />

resulting geometric reasoning equals exact reasoning and<br />

produces similar problems in a descete representation.<br />

III. HYPOTHESIS<br />

Our hypo<strong>the</strong>sis is that an axiomatic model <strong>of</strong> 2D projective<br />

geometry can be defined that incorporates geometric<br />

operations with tolerance for positional uncertainty.<br />

These operations apply to objects that reflect <strong>the</strong><br />

extended character <strong>of</strong> geographic features and <strong>the</strong> positional<br />

uncertainty in <strong>the</strong>ir representation in GIS.<br />

IV. BUILDING A MODEL<br />

The proposed approach is based on Mengers axiomatic<br />

system for projective geometry [1]. Projective geometry<br />

provides a workable research domain for geometric uncertainty<br />

modeling and is preferable for GIS: Euclidean<br />

geometry can be naturally embedded in projective geometry;<br />

Menger’s axiomatization is smaller and formulated<br />

in simple algebraic terms; It does not depend on dimension<br />

and can easily be extended to 3D.<br />

In a first step <strong>of</strong> <strong>the</strong> modeling process, existing ma<strong>the</strong>matical<br />

tools for geometric uncertainty modeling are inspected<br />

and tested for compliance with <strong>the</strong> criteria listed<br />

in chapter 2. Among <strong>the</strong>se models are Rosenfeld’s fuzzy


geometry [3], fuzzy plane geometry introduced by Buckley<br />

and Eslami [3], fuzzy rough sets [4], probability metrics,<br />

and Poston’s fuzzy geometry [5].<br />

In a second step interpretations <strong>of</strong> primitive geometric<br />

objects and operations are defined based on <strong>the</strong> findings<br />

<strong>of</strong> step one. The primitive geometric objects <strong>of</strong> projective<br />

plane geometry are points, lines, universe (“everything“),<br />

and vacuum (”nothing”). Primitive operations comprise<br />

intersection (∧) and connection (∨). In Menger’s system<br />

<strong>the</strong> behavior <strong>of</strong> projective primitives is axiomized by two<br />

projective laws that are dual to each o<strong>the</strong>r:<br />

X ∨ ((X ∨ Y ) ∧ Z) = X ∨ ((X ∨ Z) ∧ Y ) (1)<br />

X ∧ ((X ∧ Y ) ∨ Z) = X ∧ ((X ∧ Z) ∨ Y ) (2)<br />

where X, Y, Z are primitive objects. The special role <strong>of</strong><br />

universe U and vacuum V is determined by four axioms:<br />

U ∨ X = U, U ∧ X = X (3)<br />

V ∧ X = V, V ∨ X = X (4)<br />

for all primitive objects X. Initially three <strong>of</strong> <strong>the</strong> six primitives<br />

are defined. The remaining primitives must be chosen<br />

appropriately to ensure consistency <strong>of</strong> <strong>the</strong> axioms.<br />

Additionally, an approximate version <strong>of</strong> equality is defined.<br />

The set <strong>of</strong> interpretations is acceptable as a projective<br />

model, if Menger’s axioms can be verified. As a<br />

result, consistency and operational closedness is ensured.<br />

We applied <strong>the</strong> proposed procedure to a statistical interpretation<br />

<strong>of</strong> Spray Can Objects [6]. In <strong>the</strong> following<br />

section this interpretation is discussed to show <strong>the</strong><br />

methodology <strong>of</strong> <strong>the</strong> research. The presented example<br />

complies with Menger’s axioms. Primitive objects are<br />

uncertain in all dimensions, but primitive operations do<br />

not exhibit tolerance for positional uncertainty. Consequently,<br />

<strong>the</strong> model is not satisfactory for GIS applications.<br />

It is subject <strong>of</strong> ongoing research to prove or disprove<br />

<strong>the</strong> existence <strong>of</strong> a model that satisfies all criteria. It<br />

is expected that additionally a degree <strong>of</strong> uncertainty has<br />

to be assigned to <strong>the</strong> axioms in order to define a working<br />

model.<br />

V. EXAMPLE<br />

For a given exact projective point p <strong>the</strong> Spray Can Point<br />

˜p is defined by <strong>the</strong> probability density function (pdf)<br />

<strong>of</strong> a Gaussian normal distribution on <strong>the</strong> homogeneous<br />

plane Hp in p (cf. Figure 1). The Spray Can Dual<br />

scd <strong>of</strong> ˜p is defined by scd.˜p = ˜p.d, where d denotes<br />

<strong>the</strong> duality operation in <strong>the</strong> exact model and . denotes<br />

<strong>the</strong> composition <strong>of</strong> functions. The Spray Can Connection<br />

˜∨ <strong>of</strong> two Spray Can Points ˜p and ˜q is specified by<br />

˜p˜∨˜q = scd. � d(p ∨ q) � ∼ , where ∨ denotes <strong>the</strong> connection<br />

operation in <strong>the</strong> exact model and p and q are <strong>the</strong> corresponding<br />

exact points. The definition <strong>of</strong> <strong>the</strong>se primitives<br />

is sufficient to derive all remaining primitives. As we<br />

312<br />

have proven in [6] <strong>the</strong> proposed set <strong>of</strong> interpretations <strong>of</strong><br />

projective primitives complies with Mengers axioms.<br />

Figure 1: The projective plane is isomorphic to <strong>the</strong> unit<br />

sphere S 2 modulo opposite points. (a) The homogeneous<br />

plane Hp in p ∈ S 2 . (b) A Spray Can Point ˜p in p.<br />

VI. FUTURE RESEARCH<br />

The main types <strong>of</strong> positional uncertainty in GIS are measurement<br />

errors and vaguely defined locations. Uncertain<br />

geometric primitives showing <strong>the</strong>se types <strong>of</strong> uncertainty<br />

can be expressed in terms <strong>of</strong> probabilistic and possibilistic<br />

constraints [7]. The proposed work contributes to <strong>the</strong><br />

long term goal <strong>of</strong> describing geometric constructions <strong>of</strong><br />

objects under different types <strong>of</strong> uncertainty by generalized<br />

constraint propagation rules.<br />

REFERENCES<br />

[1] Leonard Blumenthal and Karl Menger. Studies in<br />

Geometry. W. h. Freeman and Company, 1970.<br />

[2] Lotfi A. Zadeh. From fuzzy logic to extended<br />

fuzzy logic - <strong>the</strong> concept <strong>of</strong> f-validity<br />

and <strong>the</strong> impossibility principle. In FUZZ-<br />

IEEE 07, www.fuzzieee2007.org/ZadehFUZZ-<br />

IEEE2007London.pdf, 2007.<br />

[3] John N. Mordeson and Premchand S. Nair. Fuzzy<br />

Ma<strong>the</strong>matics - An Introduction for Engineers and<br />

<strong>Scientist</strong>s. Physica-Verlag. A Springer-Verlag Conpany.,<br />

1998.<br />

[4] Didier Dubois and Henri Prade. Rough fuzzy sets<br />

and fuzzy rough sets. International Journal <strong>of</strong> General<br />

Systems, 17(2-3):191–209, 1990.<br />

[5] Tim Poston. Fuzzy Geometry. PhD <strong>the</strong>sis, University<br />

<strong>of</strong> Warwick, 1971.<br />

[6] Gwen Wilke and Andrew U. Frank. Projective spray<br />

can geometry - towards an axiomatic approach to error<br />

modelling for vector based geographic information<br />

systems. GEOWS 2009, accepted for publication.<br />

[7] Lotfi A. Zadeh. Generalized <strong>the</strong>ory <strong>of</strong> uncertainty<br />

(gtu) - principal concepts and ideas. Computational<br />

Statistics & Data Analysis, 51:15–46, 2006.


Anammox Studies Using New York City Centrate to Correlate<br />

Performance, Population Dynamics and Impact <strong>of</strong> Toxins<br />

Alex Rosenthal, Krish Ramalingam, and John Fillos (Faculty Mentor)<br />

Department <strong>of</strong> Civil Engineering<br />

The City College <strong>of</strong> New York<br />

New York, The United States <strong>of</strong> America<br />

Email: arosent00@ccny.cuny.edu<br />

Abstract — Anaerobic ammonium oxidation<br />

(Anammox) is a promising new technology for <strong>the</strong><br />

removal <strong>of</strong> nitrogen from wastewater. The City<br />

College <strong>of</strong> New York has been operating a bench<br />

scale Anammox sequencing batch reactor since<br />

January <strong>2008</strong> with <strong>the</strong> objective to enrich Anammox<br />

biomass in anaerobic digester centrate and to assess<br />

<strong>the</strong> impact <strong>of</strong> various environmental factors on<br />

Anammox performance. Also, with <strong>the</strong> data from<br />

batch activity tests it may be possible to correlate<br />

specific Anammox activity with <strong>the</strong> microbial dynamics<br />

<strong>of</strong> <strong>the</strong> reactor. These data will help <strong>the</strong> City<br />

<strong>of</strong> New York design a process that can be used to<br />

alleviate <strong>the</strong>ir nitrogen loading problem.<br />

I. INTRODUCTION<br />

Wastewater flows are a major source <strong>of</strong> nitrogen<br />

pollution to coastal ecosystems. In order to comply<br />

with current environmental regulations, The New<br />

York City Department <strong>of</strong> Environmental Protection<br />

(NYCDEP) is in <strong>the</strong> process <strong>of</strong> upgrading its water<br />

pollution control plants (WPCPs) from <strong>the</strong> conventional<br />

step-feed activated sludge process to biological<br />

nitrogen removal (BNR) processes. One promising<br />

technology under consideration is <strong>the</strong> use <strong>of</strong> anaerobic<br />

ammonia oxidation (Anammox) to remove nitrogen<br />

from <strong>the</strong> high strength centrate produced from<br />

<strong>the</strong> dewatering <strong>of</strong> anaerobic digested sludge. The<br />

Anammox process, i.e. <strong>the</strong> microbiological conversion<br />

<strong>of</strong> ammonium and nitrite to dinitrogen gas, is a<br />

very recent addition to our understanding <strong>of</strong> <strong>the</strong> biological<br />

nitrogen cycle. Compared to conventional<br />

BNR processes, <strong>the</strong> Anammox process requires no<br />

organic carbon addition and has decreased oxygen<br />

and alkalinity requirements. By some estimates,<br />

Anammox results in a 90% operating cost reduction<br />

compared to existing removal technologies [1].<br />

II. AIMS<br />

The evaluation <strong>of</strong> <strong>the</strong> Anammox process was first<br />

initiated in New York City on December 2005<br />

through January <strong>2008</strong> at a pilot facility at <strong>the</strong> 26th<br />

313<br />

Ward Nutrient Research Facility using inoculum<br />

from <strong>the</strong> DEMON reactor in Strass, Austria.<br />

Though <strong>the</strong> applicability <strong>of</strong> <strong>the</strong> process was demonstrated,<br />

<strong>the</strong> present study is designed to address<br />

and elucidate inhibitory concerns that became<br />

evident during <strong>the</strong> pilot study and thus fur<strong>the</strong>r<br />

demonstrate its viability to treat centrate at New<br />

York City wastewater treatment plants. This study<br />

began on January <strong>2008</strong> and is being conducted at<br />

<strong>the</strong> City College <strong>of</strong> New York (CCNY) using a<br />

bench scale continuous flow sequencing batch<br />

reactor (SBR) as an enrichment reactor for kinetic<br />

tests on Anammox biomass. The seed for <strong>the</strong> SBR<br />

was obtained from <strong>the</strong> previously operated pilot<br />

facility.<br />

An additional aspect <strong>of</strong> this research effort is to<br />

correlate <strong>the</strong> observed gross activities within <strong>the</strong><br />

SBR to <strong>the</strong> active biology present. Using molecular<br />

analyses, changes in activity will be related to<br />

changes in microbial composition and may serve<br />

as a predictor for major process upsets.<br />

III. FINDINGS<br />

The CCNY SBR operates anoxically from a decanted<br />

volume <strong>of</strong> 10 liters up to a fill volume <strong>of</strong> 20<br />

liters, a 20 day solids retention time (SRT), 35˚C,<br />

and a pH controlled between 7.5-7.8. The feed to<br />

<strong>the</strong> SBR is anaerobic digester centrate from <strong>the</strong><br />

Wards Island Wastewater Treatment Plant that is<br />

previously treated partially in a separate bench<br />

scale nitritation reactor. The nitritation reactor<br />

operates at a 42 hour SRT, 21˚C, and dissolved<br />

oxygen (DO) concentration <strong>of</strong> 2 mg/L, producing<br />

an effluent with a NO2-N concentration <strong>of</strong> 150-200<br />

mg/L and effluent NH3-N depending on <strong>the</strong> influent<br />

NH3-N concentration. To maintain <strong>the</strong><br />

stoichiometry required to support <strong>the</strong> Anammox<br />

process, nitrite is added to <strong>the</strong> effluent as needed.<br />

The Anammox reactor currently operates at a<br />

nitrogen loading rate <strong>of</strong> 0.75 g/L*day and a nitrogen<br />

removal efficiency <strong>of</strong> 88% which is comparable<br />

to previously studied bench scale SBRs reported<br />

in <strong>the</strong> literature[2]. Average stoichiometry


<strong>of</strong> <strong>the</strong> main reactor has been 1.00 NH3 : 1.21 NO2 -<br />

: -0.13 NO3 - , a slight deviation from <strong>the</strong> published<br />

stoichiometry <strong>of</strong> 1.00 NH3 : 1.32 NO2 - : -0.26 NO3 -<br />

which is most likely due to <strong>the</strong> complex matrix <strong>of</strong><br />

<strong>the</strong> feed that supports a complex ecology [3].<br />

Parallel to <strong>the</strong> SBR study, short duration batch<br />

tests are being conducted to assess <strong>the</strong> impact <strong>of</strong><br />

different environmental factors such as pH, DO,<br />

NO2 - , and Cl - toxicity. The seed for <strong>the</strong>se batch<br />

tests is <strong>the</strong> waste solids required to sustain <strong>the</strong><br />

SRT selected for <strong>the</strong> SBR. Figure 1 shows typical<br />

results from a batch study confirming <strong>the</strong><br />

stoichiometry expected from <strong>the</strong> Anammox process.<br />

Figures 2 and 3 display preliminary results <strong>of</strong><br />

<strong>the</strong> impact <strong>of</strong> pH and Cl - on Anammox activity.<br />

Phylogenetic analysis using <strong>the</strong> planctomycetales<br />

primer set (Pla46F, 1492R) and <strong>the</strong><br />

Anammox target primer set (Pla46F, 820R) both<br />

detected <strong>the</strong> presence <strong>of</strong> <strong>the</strong> Anammox bacteria,<br />

Candidatus Brocadia sp. 40. This was followed by<br />

community analysis using Denaturing Gradient<br />

Gel Electrophoresis (DGGE) which confirmed that<br />

Candidatus Brocadia sp. 40 was <strong>the</strong> dominant<br />

Anammox performing species in <strong>the</strong> reactor.<br />

However, <strong>the</strong> dominant clone found in <strong>the</strong> inoculating<br />

culture from Strass was Kuenenia Stuttgartiensis<br />

and <strong>the</strong> dominant species at <strong>the</strong> New York<br />

pilot plant was Brocadia Fulgida. This observation<br />

has been found in o<strong>the</strong>r inoculation/reactor pairs<br />

and suggests varied selective pressures between<br />

one wastewater environment and ano<strong>the</strong>r [4].<br />

Weekly samples from <strong>the</strong> Anammox reactor<br />

have been extracted since May <strong>2008</strong> and are being<br />

analyzed using qPCR. This time interval includes<br />

two major process upsets and extended periods <strong>of</strong><br />

progressively increased loading. It is hypo<strong>the</strong>sized<br />

that variations in reactor performance may<br />

be related to temporal variations in microbial<br />

composition and activity. It may be possible to<br />

detect those changes using qPCR <strong>of</strong> mRNA.<br />

[N] (mg/L)<br />

50<br />

40<br />

30<br />

20<br />

10<br />

Anammox Activity Test<br />

0<br />

0 15 30 45 60<br />

Time (min)<br />

NH3-N<br />

NO2-N<br />

NO3-N<br />

Figure 1. Typical results from Anammox batch<br />

activity tests<br />

314<br />

SUR (ug/mgVSS*hr)<br />

Figure 2. pH effect on Anammox activity. pH was<br />

adjusted from original values <strong>of</strong> 7.5-7.8 using CO2<br />

and NaCO3.<br />

SUR (ug/mgVSS*hour<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

Figure 3. Cl - effect on Anammox activity. Cl - was<br />

added as NaCl.<br />

ACKNOWLEDGMENTS<br />

This research is funded by <strong>the</strong> New York City<br />

Department <strong>of</strong> Environmental Protection. Special<br />

thanks go to Alan Deur, Keith Beckmann,<br />

and Rolando Villacres for <strong>the</strong>ir continued<br />

guidance and support.<br />

REFERENCES<br />

pH effect on Anammox Activity<br />

5<br />

0<br />

7.00 8.00<br />

pH<br />

NH3-N<br />

NO2-N<br />

Cl - Effect on Anammox Activity<br />

15<br />

10<br />

5<br />

0<br />

0 mg/L 500 mg/L 1000 mg/L 2000 mg/L<br />

Additional Chloride<br />

NH3-N<br />

NO2-N<br />

[1] M. Jetten et al. Microbiology and application <strong>of</strong><br />

<strong>the</strong> anaerobic ammonium oxidation (‘anammox‘)<br />

process. Current Opinion in Biotechnology,<br />

12:283-288, 2001.<br />

[2] A. Dapena-Mora et al. Stability <strong>of</strong> <strong>the</strong> ANAM-<br />

MOX process in a gas-lift reactor and a SBR.<br />

Journal <strong>of</strong> Biotechnology, 110:159-170, 2004.<br />

[3] M. Strous et al. The sequencing batch reactor as<br />

a powerful tool for <strong>the</strong> study <strong>of</strong> slowly growing<br />

anaerobic ammonium-oxidizing microorganisms.<br />

Appl Microbiol Biotechnol, 50:589-596, 1998.<br />

[4] W. van der Star et al. Startup <strong>of</strong> reactors for<br />

anoxic ammonium oxidation: Experiences from<br />

<strong>the</strong> first full-scale anammox reactor in Rotterdam.<br />

Water Research, 41:4149-4163, 2007.


Multi-Perspective Ultramicroscopy with Structured Illumination in<br />

Optically Dense Specimens<br />

Stefan Kalchmair, Klaus Becker, Nina Jährling and Hans-Ulrich Dodt (Faculty Mentor)<br />

Institute <strong>of</strong> Solid State Electronics, Dept. <strong>of</strong> Bioelectronics<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: stefan.kalchmair@tuwien.ac.at<br />

Abstract — Three-dimensional imaging <strong>of</strong> biological<br />

specimens allows visualization <strong>of</strong> <strong>the</strong> spatial<br />

structures in a relatively short amount <strong>of</strong> time.<br />

Structured illumination ultramicroscopy is a technique<br />

to obtain good 3D reconstruction even in<br />

optically dense specimens. However, in large samples<br />

scattering <strong>of</strong> <strong>the</strong> excitation beam and especially<br />

<strong>the</strong> illumination pattern ultimately limits <strong>the</strong> functionality<br />

<strong>of</strong> this method. We present a method for<br />

increasing <strong>the</strong> image quality by illuminating from<br />

multiple perspectives with a specially designed<br />

specimen rotator. For comparison <strong>of</strong> conventional<br />

and multi-perspective ultramicroscopy <strong>the</strong> central<br />

part <strong>of</strong> an entire mouse brain is imaged. Our results<br />

show significantly enhanced contrast and reduced<br />

artefacts.<br />

I. INTRODUCTION<br />

Imaging <strong>of</strong> fluorescence labelled specimens has<br />

become a standard tool in life sciences. A fluorescent<br />

dye is injected into <strong>the</strong> specimen or a fluorescent<br />

protein genetically expressed, excited with a laser<br />

and examined under a microscope. Ultramicroscopy<br />

(UM) is a technique which allows three-dimensional<br />

imaging <strong>of</strong> macroscopic samples with micrometer<br />

resolution [1, 2]. In ultramicroscopy, a cleared<br />

specimen is illuminated from <strong>the</strong> side with a thin<br />

sheet <strong>of</strong> laser light. Only <strong>the</strong> focal plane <strong>of</strong> <strong>the</strong> detection<br />

optics is illuminated, so that no out-<strong>of</strong>-focus<br />

light is generated and only structures from a thin slice<br />

<strong>of</strong> <strong>the</strong> specimen are recorded.<br />

We improved this technique by <strong>the</strong> introduction <strong>of</strong><br />

structured illumination ultramicroscopy (SI-UM) [3].<br />

The light sheet is spatially filtered to form single<br />

beams <strong>of</strong> light with large spaces in between. After<br />

recording each image, <strong>the</strong> illumination mask is<br />

shifted by a fraction <strong>of</strong> one single beam width and<br />

<strong>the</strong> next image is recorded (Figure 1). By processing<br />

<strong>the</strong> entire set <strong>of</strong> images, an improved image is calculated<br />

with reduced background fluorescence and<br />

enhanced contrast.<br />

However, in large samples or optically dense specimens,<br />

scattering <strong>of</strong> <strong>the</strong> structured light sheet limits<br />

315<br />

Figure 1: Optical setup <strong>of</strong> <strong>the</strong> multi-perspective<br />

SI-UM. The fluorescent sample is excited from <strong>the</strong><br />

left side with a structured light beam. The sample<br />

is mounted on <strong>the</strong> magnetically coupled platform<br />

and rotated after each recorded SI-UM image.<br />

<strong>the</strong> functionality <strong>of</strong> <strong>the</strong> method. SI-UM depends on<br />

<strong>the</strong> clear visibility <strong>of</strong> <strong>the</strong> pattern in <strong>the</strong> focal plane <strong>of</strong><br />

<strong>the</strong> detection optics. In a scattering sample, <strong>the</strong> structured<br />

light sheet is clearly visible in <strong>the</strong> outer area <strong>of</strong><br />

<strong>the</strong> specimen, but appears strongly degraded deep<br />

inside <strong>the</strong> sample. Also shadowing effects become<br />

stronger as <strong>the</strong> beam propagates through <strong>the</strong> tissue.<br />

This paper describes a method to reduce <strong>the</strong> influence<br />

<strong>of</strong> light sheet degradation. By merging images<br />

from multiple perspectives, we can increase <strong>the</strong> final<br />

image quality. For this purpose we developed a computer<br />

controlled specimen rotator that allows precise<br />

sample movement inside <strong>the</strong> liquid filled specimen<br />

chamber.<br />

II. MATERIALS AND METHODS<br />

The fluorescent samples were excited by a solid state<br />

laser (Cyan Excelsior, Spectra Physics, � = 488 nm,<br />

50mW). The expanded beam was focused vertically<br />

by a cylindrical lens to form a thin light sheet. A<br />

mask was placed in <strong>the</strong> focus <strong>of</strong> <strong>the</strong> cylindrical lens<br />

to form <strong>the</strong> illumination pattern, which <strong>the</strong>n was


projected into <strong>the</strong> specimen. The microscope (Zeiss)<br />

was equipped with a bandpass filter (pass band: 505-<br />

530nm, Zeiss) positioned above <strong>the</strong> objective. Images<br />

were recorded with a CCD Camera (CoolSnap<br />

K4, 2048x2048 pixels, 12bit, Roper Scientific). To<br />

achieve sufficient transparency for most biological<br />

specimens, chemical clearing is necessary. Clearing<br />

was done by incubation in benzylbenzoat/ benzylalcohol<br />

(BABB), which has <strong>the</strong> same refractive index<br />

as protein.<br />

The specimens were mounted on a custom made<br />

specimen rotator. To avoid image distortions after<br />

reconstruction, displacements along <strong>the</strong> optical axis<br />

had to be below <strong>the</strong> vertical resolution <strong>of</strong> <strong>the</strong> microscope,<br />

which was about 30 �m. To protect <strong>the</strong> motor<br />

chamber against fluid in <strong>the</strong> BABB filled specimen<br />

chamber, <strong>the</strong> rotating specimen platform was magnetically<br />

coupled to <strong>the</strong> motor (Figure 2). To ensure<br />

smooth movements a chemically resistant bearing<br />

was used.<br />

III. RESULTS<br />

We recorded images from <strong>the</strong> central region <strong>of</strong> a<br />

mouse brain by using tissue aut<strong>of</strong>luorescence. If <strong>the</strong><br />

sample is illuminated from only one direction, <strong>the</strong><br />

illumination pattern appears dim and gradually degraded<br />

from top to bottom (Figure 3 a). This effect is<br />

mainly due to scattering caused by tissue outside <strong>the</strong><br />

field <strong>of</strong> view. The processed image shows sharp<br />

edges at <strong>the</strong> surface <strong>of</strong> <strong>the</strong> specimen but blurred<br />

features deep inside <strong>the</strong> sample (Figure 3 b). After<br />

multi-perspective imaging, <strong>the</strong> single SI images are<br />

superimposed computationally (Figure 3 c). The<br />

processed image now appears sharp in most areas <strong>of</strong><br />

<strong>the</strong> image (Figure 3 d). The white arrow in Figure 3 d<br />

marks a feature deep inside <strong>the</strong> sample, which was<br />

hardly noticeable in Figure 3 b and is now clearly<br />

visible. Fur<strong>the</strong>r, artefacts from shadowing effects<br />

were significantly reduced.<br />

Figure 2: Schematic <strong>of</strong> <strong>the</strong> specimen rotator.<br />

316<br />

Figure 3: Comparison <strong>of</strong> conventional and multiperspective<br />

SI-UM. (a) Single-perspective image<br />

with structured illumination. (b) Result <strong>of</strong> post<br />

processing <strong>of</strong> 40 shifted SI images. (c) Multiperspective<br />

image. (d) Combination <strong>of</strong> two post<br />

processed multi-perspective SI-UM images.<br />

Future improvements are possible by computational<br />

analysis <strong>of</strong> <strong>the</strong> single SI images for automatic alignment<br />

<strong>of</strong> <strong>the</strong> superposition boarders. Optimization <strong>of</strong><br />

<strong>the</strong> specimen rotator should also result in increased<br />

accuracy and recording speed.<br />

IV. CONCLUSIONS<br />

We have demonstrated that multi-perspective illumination<br />

improves image quality <strong>of</strong> SI-UM<br />

images in optically dense specimens. Artefacts<br />

from shadowing effects are also significantly reduced.<br />

Especially for large specimens this approach<br />

is advantageous.<br />

REFERENCES<br />

[1] H.-U. Dodt, U. Leischner, A. Schierloh, N. Jährling,<br />

C. P. Mauch, K. Deininger, J. M. Deussing,<br />

M. Eder, W. Zieglgänsberger, and K. Becker. Ultramicroscopy:<br />

three-dimensional visualization<br />

<strong>of</strong> neuronal networks in <strong>the</strong> whole mouse brain.<br />

Nat Meth, 4:331-336, April 2007.<br />

[2] K. Becker, N. Jährling, E. R. Kramer, F. Schnorrer<br />

and H.-U. Dodt. Ultramicroscopy: 3D reconstruction<br />

<strong>of</strong> large microscopical specimens. J.<br />

Biophoton. 1:36-42, January <strong>2008</strong>.<br />

[3] S. Kalchmair. Image Improvement by Structured<br />

Illumination in Ultramicroscopy. MA <strong>the</strong>sis,<br />

Vienna University <strong>of</strong> Technology, June <strong>2008</strong>.


Analytical technology for biogas production from<br />

straw by multivariate Near-infrared spectroscopy<br />

Chularat Krongtaew 1 , Karin Fackler 1 , Barbara Hinterstoisser 2<br />

and Kurt Messner 1 (Faculty Mentor)<br />

1 Institute <strong>of</strong> Chemical Engineering, Vienna University <strong>of</strong> Technology, Vienna, Austria<br />

2 Department <strong>of</strong> Material Sciences and Process Engineering, University <strong>of</strong> Natural Resources<br />

and Applied Life Sciences, Vienna, Austria<br />

Email: ckrongta@mail.zserv.tuwien.ac.at<br />

Abstract —An analytical procedure to evaluate<br />

<strong>the</strong> key process parameters <strong>of</strong> wheat (Triticum<br />

aestivum L.) and oat (Avena sativa L.) straw pretreatment<br />

as well as <strong>the</strong> biogas fermentation <strong>of</strong><br />

treated straw using Fourier transform nearinfrared<br />

(FT-NIR) spectra in combination with<br />

chemometrics was developed. Not only weight<br />

loss, residual lignin content and hydrolysable<br />

polysaccharides in terms <strong>of</strong> reducing sugars <strong>of</strong><br />

delignified straw, but also biogas productivity <strong>of</strong><br />

pre-treated straw and <strong>the</strong> parameters, total solids<br />

and volatile solids, were accurately calculated<br />

based on substrate quality after pretreatment. The<br />

multivariate ma<strong>the</strong>matical analysis was used as an<br />

efficient facility for spectral data processing. The<br />

findings in this work suggest FT-NIR spectroscopy<br />

as an alternative and low-budget analytical tool<br />

providing robust and validated prediction models<br />

for lignocellulose-to-energy conversion in biogas<br />

production.<br />

I. INTRODUCTION<br />

Owing to <strong>the</strong> world’s energy shortage, a bulk volume<br />

industry <strong>of</strong> new rapidly growing agricultural biomass-to-energy<br />

conversion is urgently required.<br />

These processes need a robust and low-budget process<br />

analytical chemistry (PAC) approach to come<br />

across all hindrances for efficiently industrial anaerobic<br />

digestion process analyses. Great advantages<br />

<strong>of</strong> process analytical chemistry will be accomplished<br />

if analyses can be done prior to bulk production in<br />

terms <strong>of</strong> both growing plants and building up <strong>the</strong><br />

digester. As pretreatment step <strong>of</strong> lignocellulose was<br />

required to diminish <strong>the</strong> rate-limiting step <strong>of</strong> overall<br />

anaerobic digestion processes, in <strong>the</strong> present study<br />

<strong>the</strong> crucial parameters <strong>of</strong> both pretreatment and anaerobic<br />

digestion steps <strong>of</strong> wheat and oat straw were<br />

evaluated by means <strong>of</strong> Fourier transform nearinfrared<br />

(FT-NIR) analysis with ma<strong>the</strong>matical treatment<br />

<strong>of</strong> spectral data. This led to achievement on<br />

evaluating biogas productivity based on substrate<br />

quality by practical and inexpensive means.<br />

317<br />

II. MATERIALS AND METHODS<br />

Delignifying pretreatment <strong>of</strong> wheat and oat straw<br />

performed in aqueous suspension (2 g/ 20ml) with<br />

different acidic and alkaline pre-treatment systems.<br />

Afterwards, straw particles were washed with deionized<br />

water and dried at 50°C before <strong>the</strong>y were<br />

ei<strong>the</strong>r milled with a Retsch ZM 1000 ultracentrifugal<br />

mill (80 �m) or enzymatically hydrolysed (Viscozyme®L,<br />

www.novozymes.com). After hydrolysis,<br />

<strong>the</strong> amount <strong>of</strong> reducing sugars in <strong>the</strong> supernatant<br />

was determined by means <strong>of</strong> a modified<br />

dinitrosalicylic acid (DNS) assay. Extraction and<br />

lignin analyses <strong>of</strong> milled sample were carried out<br />

according to TAPPI methods.<br />

For biogas potential test both solid and liquid<br />

parts after straw pretreatment was used as substrate<br />

and 150 ml digested sludge was added as inoculum.<br />

The anaerobic digestion performed at 37°C and<br />

accumulative biogas generation was recorded daily.<br />

Total solids and volatile solids <strong>of</strong> digested sludge<br />

were determined after <strong>the</strong> gas generation ceased.<br />

NIR spectroscopy <strong>of</strong> delignified straw (80 μm)<br />

was recorded by EQUINOX 55 (Bruker Optics<br />

Inc., Germany) using a fiber optic probe [1]. The<br />

appearance absorbance, log [1/ Reflectance], <strong>of</strong><br />

treated straw samples was measured with 8 cm -1<br />

resolution, 100 scans, <strong>the</strong> wave number ranging<br />

from 10000 to 4000 cm -1 with four replication.<br />

The 17 smoothing points and 2 nd order polynomial<br />

fit <strong>of</strong> spectra were applied [2]. Statistical analyses<br />

<strong>of</strong> NIR spectra were performed using OPUS 6.0<br />

(www.brukeroptics.de) and Unscrambler 9.7<br />

(www.camo.com), <strong>the</strong> chemometric s<strong>of</strong>tware.<br />

III. RESULTS AND DISCUSSION<br />

For <strong>the</strong> highly lignified and acetylated wheat straw<br />

samples (i.e. non-treated sample with 20.68 % total<br />

lignin content), <strong>the</strong> minima at 5980 cm -1 attributed<br />

to lignin, and at 5800 cm -1 attributed to lignin and<br />

xylan moieties [3] were intense as shown in Figure<br />

1A. The similar results and spectral features were<br />

obtained in oat straw samples (Figure 1B). Princi-


pally, <strong>the</strong>re were two major characteristics <strong>of</strong> delignified<br />

straw samples found from <strong>the</strong> NIR 2 nd derivative:<br />

1) <strong>the</strong> lignin minimum near 5980 cm -1 and <strong>the</strong><br />

overlapping <strong>of</strong> lignin and xylan minimum at 5800<br />

cm -1 were less intense due to <strong>the</strong> delignifying steps<br />

in which lignin and hemicellulose xylan were degraded,<br />

2) <strong>the</strong> minima at 5960 and 5990 cm -1 assigned<br />

to xylan disappeared due to <strong>the</strong> degradation<br />

or deacetylation. Therefore, <strong>the</strong> lignin and carbohydrate<br />

content within delignified straw after pretreatment<br />

were certainly analysed from <strong>the</strong> FT-NIR<br />

spectra by means <strong>of</strong> quantitative analysis through<br />

partial least-squares (PLS) regression models as<br />

shown in Table 1.<br />

As some <strong>of</strong> hemicelluloses, extractives, mineral<br />

salts in straw were solubilised after pretreatment,<br />

volatile solid content in treated straw increased and<br />

led to higher biogas yield for anaerobic digestion.<br />

Total solids and volatile solids <strong>of</strong> digested sludge<br />

that significantly influenced biogas yield were powerfully<br />

analysed by FT-NIR spectroscopy (Table 1).<br />

Component range<br />

Spectral processing<br />

No. <strong>of</strong> Sample<br />

Loading factors<br />

Calibration<br />

R 2 cal<br />

Figure 1: NIR 2 nd derivative <strong>of</strong> delignified straw:<br />

(A) wheat, and (B) oat straw samples.<br />

NIR Reducing sugars Total residual lignin Total solids Volatile solids Biogas<br />

Wave number [cm -1 ] 10001.1-6398.7<br />

5800.9-5199.2<br />

4601.3-3999.7<br />

115.80-704.20 mg/g<br />

First derivative + MSC 1<br />

10001.1-7598.2 10001.1-9399.4 10001.1-8199.9<br />

7000.4-4597.5 8801.6-8199.9 7602.1-7000.4<br />

7602.1-5797.0 5800.9-5199.2<br />

5.23-21.45 % 93.60-99.73 % 94.50-99.20 %<br />

First derivative Second derivative First derivative + MSC<br />

38<br />

38<br />

32<br />

5<br />

4<br />

5<br />

1<br />

10001.1-8199.9<br />

7000.4-6398.7<br />

5800.9-4597.5<br />

219.0-386.5 ml/g<br />

MSC<br />

38<br />

4<br />

1<br />

30<br />

2<br />

RMSEE 2<br />

Validation<br />

R 2 val<br />

0.96<br />

38.50 mg/g<br />

0.92<br />

0.98<br />

0.64 %<br />

0.97<br />

0.96<br />

0.47 %<br />

0.93<br />

0.27 %<br />

0.75<br />

20.9 ml/g<br />

RMSECV/RMSEP 3<br />

53.70 mg/g<br />

0.73 %<br />

0.80 %<br />

0.33 %<br />

1 2 3<br />

Multiplicative scattering correction, Root mean square error <strong>of</strong> estimation, Root mean square error <strong>of</strong> cross validation/Root mean square error <strong>of</strong> prediction<br />

0.85<br />

0.88<br />

0.65<br />

23.4 ml/g<br />

Table 1: Partial least-squares (PLS) regression models calculated from FT-NIR spectra<br />

It was obvious that <strong>the</strong> analytical procedure <strong>of</strong> <strong>the</strong><br />

pre-treated substrate by means <strong>of</strong> NIR spectra combined<br />

with multivariate analysis could be applied to<br />

<strong>the</strong> entire process – from <strong>the</strong> pre-treatment <strong>of</strong> lignocellulose<br />

to anaerobic digestion. Ma<strong>the</strong>matical<br />

spectral data processing (partial least squares regression<br />

– PLSR) toge<strong>the</strong>r with derivatization,<br />

multiplicative scattering correction (MSC), minmax<br />

normalization, constant <strong>of</strong>fset elimination as<br />

data pre-treatment was applied in order to achieve<br />

optimized, robust, and validated models. This study<br />

provided a potential analytical approach based on<br />

substrate quality which can be feasibly broadened to<br />

<strong>the</strong> large scale industrial application.<br />

ACKNOWLEDGMENTS<br />

The Technology Grants from <strong>the</strong> Austrian Federal<br />

Ministry for Science and Research is honourably<br />

acknowledged.<br />

318<br />

Lignin decrease<br />

5990<br />

5980<br />

5960<br />

5990<br />

5980<br />

5960<br />

REFERENCES<br />

5885<br />

5885<br />

5840<br />

5840<br />

Lignin decrease 5800<br />

5800<br />

5750<br />

Lignin 8.32%<br />

Lignin 11.29%<br />

Lignin 16.97%<br />

Lignin 20.68%<br />

4.E-06<br />

2.E-06<br />

0.E+00<br />

-2.E-06<br />

-4.E-06<br />

-6.E-06<br />

-8.E-06<br />

[1] K. Fackler, M. Schwanninger, C. Gradinger,<br />

B. Hinterstoisser and K. Messner. Fungal decay<br />

<strong>of</strong> spruce and beech wood assessed by<br />

near infrared spectroscopy in combination<br />

uni- and multivariate data analysis. Holzforschung<br />

61: 680-687, 2007.<br />

[2] A. Savitzky and M.J.E. Golay. Smoothing<br />

and differentiation <strong>of</strong> data by simplified<br />

least squares procedures. Analytical chemistry<br />

36(8):1627-1639, 1964.<br />

[3] J.S. Shenk, J.J. Workman and M.O. Westerhaus.<br />

Application <strong>of</strong> NIR Spectroscopy to<br />

Agricultural Products. Handbook <strong>of</strong> Near-<br />

Infrared Analysis. (D.A. Burns, E.W. Ciurczak,<br />

eds.). Dekker, New York pp. 419-<br />

474, 2001.<br />

5750<br />

5700<br />

5700<br />

Wave number [cm -1 ]<br />

(A)<br />

Lignin 8.83%<br />

Lignin 9.90%<br />

Lignin 15.64%<br />

Lignin 18.51%<br />

(B)<br />

4.E-06<br />

2.E-06<br />

0.E+00<br />

-2.E-06<br />

-4.E-06<br />

-6.E-06<br />

-8.E-06<br />

derivative<br />

NIR 2 nd<br />

NIR 2 nd derivative


PLAYING WITH MAPS:<br />

THE ROLE OF WEBCARTOGRAPHY IN EDUCATION<br />

Univ.Ass. MSc. Eszter Simonné-Dombóvári and<br />

Univ.Pr<strong>of</strong>. Mag. Dr. Georg Gartner (Faculty Mentor)<br />

Institute for Geoinformation and Cartography<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: eszter@cartography.tuwien.ac.at,<br />

georg.gartner@tuwien.ac.at<br />

Ass.Pr<strong>of</strong>. MSc. Dr. José Jesús Reyes Nuñez (Faculty Mentor)<br />

Department <strong>of</strong> Cartography and Geoinformatics<br />

Eötvös Loránd University<br />

Budapest, Hungary<br />

Email: jesus@lundens.elte.hu<br />

Abstract — My PhD <strong>the</strong>sis is about developing<br />

map games in <strong>the</strong> geographic education for 4th, 6th<br />

and 8th grade <strong>of</strong> High Schools. It analyses<br />

geography and cartography education and <strong>the</strong>ir<br />

teaching and learning methods. It also researches<br />

possibilities <strong>of</strong> not traditional teaching tools in<br />

Austria, Hungary and in o<strong>the</strong>r countries <strong>of</strong><br />

European Union. My aim is to make an interactive<br />

website with applications to learn cartographical<br />

skills. They can be used by pupils to practice map<br />

reading and geographical concepts.<br />

I. INTRODUCTION<br />

As far back as <strong>the</strong> history <strong>of</strong> geography teaching<br />

goes, conscientious teachers have sought new and<br />

better ways to facilitate and accelerate geography<br />

learning. We all know that people’s best learning<br />

experience comes when <strong>the</strong>y are engaged in activities<br />

that <strong>the</strong>y enjoy and care about [1].<br />

Edutainment, <strong>the</strong> method <strong>of</strong> amusing teaching, can<br />

help practicing and enlarging <strong>the</strong> store <strong>of</strong> learning in<br />

particular by map reading and geographical concepts.<br />

The edutainment games have some basic elements,<br />

such as educational content, interactive components<br />

and attractive interface that can catch <strong>the</strong> attention <strong>of</strong><br />

<strong>the</strong> users. But one <strong>of</strong> <strong>the</strong> most important rules <strong>of</strong> <strong>the</strong><br />

map games is <strong>the</strong> interactivity which is possible with<br />

<strong>the</strong> special opportunities <strong>of</strong> <strong>the</strong> web environment like<br />

Flash, Java, JavaScript, SVG. This new techniques<br />

let <strong>the</strong> developers using interactive and multimedia<br />

elements to make dynamic websites [2].<br />

These days we witness an unexpected<br />

development <strong>of</strong> Information and Communication<br />

Technology (ICT). Its rule and importance is<br />

determinative even in schoolwork [3]. Using<br />

319<br />

computer and online-medium in geography education<br />

was summarized by A. Koller. He expressed that<br />

more and newer means have to be used in education<br />

beside traditional tools. Certainly this brings a new<br />

situation for teachers and students as well [4].<br />

Nowadays <strong>the</strong> number <strong>of</strong> geographical and<br />

cartographical websites – e.g. homepages <strong>of</strong><br />

institutes, geography portals, lexicons, guides,<br />

methodological assistance, databases, map- and<br />

outline map catalogs – is continuously increasing.<br />

The main aim <strong>of</strong> <strong>the</strong>se is to help <strong>the</strong> work <strong>of</strong> students<br />

and teachers in preparation for <strong>the</strong> lessons or to raise<br />

<strong>the</strong>ir attention. A website containing map games can<br />

be included to this category, because it is planned to<br />

entertain, but at <strong>the</strong> same time it is also able to<br />

develop <strong>the</strong> user’s cartographical skills. It is a great<br />

tool to practice what <strong>the</strong>y have learnt previously, that<br />

is <strong>the</strong> curriculum itself. It is recommended to people<br />

<strong>of</strong> different ages, who prefer games and putting <strong>the</strong>ir<br />

memory and general knowledge into test.<br />

II. EXPERIENCES<br />

Nowadays not only <strong>the</strong> 3D, complicated, virtual<br />

reality games but also <strong>the</strong> amusing, simple, easy-touse,<br />

non-violent, short, intellectual type <strong>of</strong><br />

entertainment games is popular. The main types <strong>of</strong><br />

traditional cartographic edutainment are <strong>the</strong><br />

following [5]:<br />

One: Outline map game. The users have to identify<br />

map features or to position <strong>the</strong>m on a map.<br />

Two: Puzzle. The elements can be equal in size or<br />

administrative areas like countries or counties.<br />

Three: Cross-words. It is a combination <strong>of</strong> <strong>the</strong><br />

normal cross-word and <strong>the</strong> outline map.


In <strong>the</strong> poster different educational s<strong>of</strong>tware and<br />

websites are presented, including a website<br />

developed for schools. This application is <strong>the</strong> first<br />

result <strong>of</strong> my PhD research (Figure 1). It was<br />

originally prepared for Hungarian education, but my<br />

aim is to implement this program into German and<br />

o<strong>the</strong>r languages too.<br />

A. AIMS<br />

The aim was to make a program which can help<br />

learning, practising and exercising cartographical<br />

knowledge, an educational tool which is easily<br />

accessible on internet. If it is based on <strong>the</strong><br />

requirement <strong>of</strong> <strong>the</strong> secondary school, it can help <strong>the</strong><br />

teachers’ work in preparation for <strong>the</strong> lessons as well<br />

as to raise students’ attention. This develops spatial<br />

orientation which is a geographical competence in<br />

education. This is <strong>the</strong> ability to orientate oneself:<br />

topographical orientation skills, map-reading<br />

competence, orientation in real space and reflection<br />

upon spatial perception [6].<br />

Figure 1: GeoLearn website,<br />

http://cartography.tuwien.ac.at/geolearn<br />

B. TECHNIQUES<br />

The interactive web applications will be developed<br />

with scripting languages (PHP and JavaScript) on <strong>the</strong><br />

base <strong>of</strong> MySQL-database to be portable. The lists <strong>of</strong><br />

topographic requirements are saved as tables in a<br />

MySQL-database with <strong>the</strong> same structure and PHP<br />

was used to implement <strong>the</strong> user management, to<br />

work with HTML-forms and to make database<br />

connection. The client-side JavaScript realizes <strong>the</strong><br />

interactivity <strong>of</strong> <strong>the</strong> website. This technical<br />

environment can help making a dynamic website<br />

with cartographical contents and a cartographical<br />

information system [7].<br />

C. PLANS<br />

According to <strong>the</strong> feedbacks, users would like to see<br />

more functions on <strong>the</strong> website, e.g. separate screen<br />

for teachers with <strong>the</strong> possibility to adjust, modify <strong>the</strong><br />

pre-defined database. There will be a complete web<br />

application for using and learning geographical<br />

320<br />

names including <strong>the</strong>se new functions. The fur<strong>the</strong>r<br />

plans are to make o<strong>the</strong>r applications about<br />

cartographical skills to this website, e.g. tests, outline<br />

maps, puzzles.<br />

III. FEEDBACK<br />

Some teachers built <strong>the</strong> program into <strong>the</strong>ir geography<br />

class <strong>the</strong>matic. They made tests with this application.<br />

They found it helpful to memorize topographic<br />

information. Pupils and students used it for preparing<br />

for tests and final exams. Students outside school<br />

were also interested in it. The number <strong>of</strong> <strong>the</strong> visitors<br />

on <strong>the</strong> website shows it (more than 1400 visitors in 6<br />

months on Hungarian website). They found this<br />

website very useful and entertaining. Last but not<br />

least, we could compare <strong>the</strong> exam results <strong>of</strong> students<br />

who practised with our program or not. The results <strong>of</strong><br />

<strong>the</strong> program users were 15% better on average.<br />

REFERENCES<br />

[1] M. Resnick. Edutainment? No Thanks. I Prefer<br />

Playful Learning. Retrieved March 08, <strong>2008</strong>,<br />

from http://www.roboludens.net/Edut_Articoli/<br />

Playful_Learning.pdf<br />

[2] L. Zentai, E. Dombóvári. Edutainment in<br />

cartography. In: <strong>Proceedings</strong> <strong>of</strong> International<br />

Cartographic Association (Eds.), Joint ICA<br />

Commissions Seminar, pages 60-65, Madrid-<br />

Budapest, July 2005.<br />

[3] Eurydice. Key Data on Information and<br />

Communication Technology in Schools in<br />

Europe. Retrieved February 03, <strong>2008</strong>, from<br />

http://www.eurydice.org/ressources/Eurydice/pdf<br />

/0_integral/048EN.pdf<br />

[4] A. Koller. Computereinsatz und Online-Medien<br />

im GW-Unterricht. In: W. Sitte und H.<br />

Wohlschlägl (Eds.): Beiträge zur Didaktik des<br />

„Geographie und Wirtschaftskunde“-Unterrichts<br />

(= Materialien zur Didaktik der Geographie und<br />

Wirtschaftskunde, 16). Universität Wien, pp. 59-<br />

75, Wien, 2001.<br />

[5] E. Dombóvári. Master <strong>the</strong>sis. Interaktív<br />

térképrejtvények. Eötvös Loránd University,<br />

Budapest, Juni 2005.<br />

[6] Deutsche Gesellschaft für Kartographie (DGfK)<br />

(Eds.) Educational Standards in Geography for<br />

<strong>the</strong> Intermediate School Certificate. Retrieved<br />

February 10, <strong>2008</strong>, from<br />

http://www.geographie.de/<br />

[7] M. Gede, E. Simonné-Dombóvári. The blind<br />

mouse – mute map game. In: <strong>Proceedings</strong> <strong>of</strong> 5th<br />

International <strong>Conference</strong> on Geographic<br />

Information Systems (ICGIS<strong>2008</strong>), pages 461-<br />

466, Istanbul, July <strong>2008</strong>.


Overtopping <strong>of</strong> Dams – A Model Approach<br />

Michael Pucher and Peter Tschernutter (Faculty Mentor)<br />

Institute <strong>of</strong> Hydraulic And Water Resources Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: michael.pucher@kw.tuwien.ac.at<br />

Abstract — The design <strong>of</strong> dams with full or partial<br />

overtoppable crests already has a long history.<br />

Since recently <strong>the</strong>ir number is increasing constantly,<br />

but <strong>the</strong>re are still open questions concerning<br />

<strong>the</strong>ir design. Design formulas in <strong>the</strong> literature<br />

give different results. One <strong>of</strong> <strong>the</strong> reasons could be<br />

<strong>the</strong> underlying model test assumptions. Model test<br />

on two different sets were performed and <strong>the</strong>ir results<br />

compared to recommendations from literature.<br />

I. INTRODUCTION<br />

The problem <strong>of</strong> flood discharges in high dams is<br />

usually solved by means <strong>of</strong> a concrete spillway separated<br />

from <strong>the</strong> dam. But overflowable dams can be<br />

� less expensive – less “over-design” - especially<br />

when used during construction <strong>of</strong> diversion<br />

structures<br />

� aes<strong>the</strong>tically pleasing<br />

Figure 1: Types <strong>of</strong> failures <strong>of</strong> bank protections.<br />

Erosion <strong>of</strong> single stone (left) or <strong>the</strong> bank as a<br />

whole (right) [1]<br />

Of late, it has repeatedly been attempted to reduce<br />

<strong>the</strong> cost <strong>of</strong> higher dams <strong>of</strong> this type by admitting <strong>the</strong>ir<br />

overtopping. The design <strong>of</strong> dams with full or partial<br />

overtoppable crests has already a long history. This<br />

type <strong>of</strong> dam can be found on river dykes, with a<br />

sideways overflow section, or in <strong>the</strong> case <strong>of</strong> retention<br />

dams, with a vertical flow direction towards <strong>the</strong> crest.<br />

Because <strong>of</strong> <strong>the</strong> recent floods overtoppable dykes or<br />

retentions dams have undergone a revival. The main<br />

scientific questions are <strong>the</strong>:<br />

� flow characteristic on <strong>the</strong> overflow section<br />

� forces on <strong>the</strong> lining or single stone<br />

� location <strong>of</strong> overtopping section<br />

321<br />

� best and economic type <strong>of</strong> slope protection<br />

� design <strong>of</strong> special features (e.g. crest, dam base)<br />

To find answers for our questions a literature review<br />

and physical model tests have been undertaken.<br />

II. STATE OF THE ART<br />

In [1] <strong>the</strong> basic hydraulic mechanisms at and over a<br />

overflow section are discussed. A lot <strong>of</strong> formulas for<br />

overflow sections in terms <strong>of</strong> hydraulics and stability<br />

are based on research on ramps [2]. The mechanisms<br />

<strong>of</strong> failure and forces on <strong>the</strong> lining or single stone are<br />

well described in [1, 3, 4]. Recommendations on <strong>the</strong><br />

location <strong>of</strong> <strong>the</strong> overflow section are given in (5).<br />

Usually <strong>the</strong> design <strong>of</strong> a well defined section is recommended.<br />

Due to differential settlements <strong>the</strong> discharge<br />

at preferred flow paths can exceed <strong>the</strong> design<br />

discharge. Thus <strong>of</strong>ten causing failures.<br />

Figure 2: Comparison <strong>of</strong> rip-rap formulas [4]<br />

For <strong>the</strong> design <strong>of</strong> over-flow dam sections a multitude<br />

<strong>of</strong> slope-protection methods can be found in <strong>the</strong><br />

literature [1, 5].<br />

� Placing <strong>of</strong> a lining (bank protection) like riprap,<br />

placed riprap, paver stones, open-stone asphalt<br />

and more<br />

� Consolidation with soil cement<br />

� Composite methods with geotextiles or gabions<br />

For most <strong>of</strong> <strong>the</strong> construction methods design formulas<br />

can be found in <strong>the</strong> literature [4, 5, 6, 7]. However,<br />

<strong>the</strong>se design formulas (see for example “Figure<br />

2”give results in a wide range [4, 5]. This is not surprising<br />

since most <strong>of</strong> <strong>the</strong> formulas are based on<br />

model tests, where <strong>the</strong> settings and boundary condi-


tions are difficult to translate to any o<strong>the</strong>r test. For<br />

special design features individual project design and<br />

<strong>the</strong> literature <strong>of</strong>ten propose quite divergent solutions<br />

[4].<br />

III. PHYSICAL MODEL TESTS<br />

Due to <strong>the</strong> differences in formulas and <strong>the</strong>ir underlying<br />

model tests it was decided that basic research in<br />

terms <strong>of</strong> an extended “model family” is needed to<br />

clarify <strong>the</strong> influence <strong>of</strong> <strong>the</strong> model type on results. The<br />

first two sets <strong>of</strong> physical model tests were performed<br />

by means <strong>of</strong> a sectional model with a slope <strong>of</strong> 2H:1V<br />

in a 50 cm wide flume <strong>of</strong> <strong>the</strong> Hydraulic Laboratory.<br />

For <strong>the</strong> first set (1) a box model substituted <strong>the</strong><br />

dam body, while in <strong>the</strong> second (2) set a “half-dam”<br />

mode was used (see “Figue 3”) [8].<br />

1 2<br />

Figure 3: Model sets<br />

Testing <strong>the</strong> first set we found that <strong>the</strong> failure mode<br />

was <strong>of</strong>ten a slip surface along <strong>the</strong> border <strong>of</strong> <strong>the</strong> lining<br />

and <strong>the</strong> filter. That means that <strong>the</strong> whole rip-rap<br />

slipped one to two centimeters. Thus opening a joint<br />

at <strong>the</strong> dam crest and destabilizing <strong>the</strong> single stones.<br />

By increasing <strong>the</strong> discharge <strong>the</strong> riprap was destroyed<br />

beginning from top down to bottom. Sometimes <strong>the</strong><br />

slip would occur earlier but <strong>the</strong> rip rap would stay<br />

stable until a higher discharge (see “Figure 4”). By<br />

protecting <strong>the</strong> upper fourth <strong>of</strong> <strong>the</strong> riprap we could<br />

improve <strong>the</strong> performance reasonably.<br />

Figure 4: Results on a slope H2:V1 (tan�=0,5)<br />

In <strong>the</strong> series with 1V:2H slopes on <strong>the</strong> “half-dam”<br />

model, with a slightly different time-step <strong>of</strong> discharge<br />

increase, no slip surface occurred. Higher<br />

discharge levels than in <strong>the</strong> first set could be<br />

achieved. As <strong>the</strong> failure mode, <strong>the</strong> erosion <strong>of</strong> single<br />

stones could be determined. An interesting effect is<br />

322<br />

that during <strong>the</strong> erosion process temporarily a quasi<br />

stable system <strong>of</strong> “cascades”, before <strong>the</strong> final erosion,<br />

is formed.<br />

IV. CONCLUSION AND OUTLOOK<br />

The research indicates that <strong>the</strong> model test setup might<br />

have significant influence on <strong>the</strong> results. The “Halfdam”<br />

model setup, with a small discharge increase,<br />

results in higher critical discharge levels than in <strong>the</strong><br />

box setup, while no slip surface occurred. Generally,<br />

<strong>the</strong> results are in range with to <strong>the</strong> ones found in <strong>the</strong><br />

literature. It is not yet possible to recommend or<br />

discard any design formula. Since we found such<br />

variation in our results we will continue our efforts.<br />

An extended series will be run in a big flume to<br />

check on 3D effects (prefered flow patchs etc.).<br />

REFERENCES<br />

[1] M. Bosshard, Überflutbarkeit kleiner Dämme,<br />

Versuchsanstalt für Wasserbau, Hydrologie und<br />

Glaziologie der ETH Zürich, 1991<br />

[2] G. Platzer, Hydraulische Grundlagen für die<br />

Dimensionierung ebener und muldenförmiger<br />

Blocksteinrampen mit einer Neigung von 1 : 10,<br />

Dissertation an der Technischen Universität<br />

Wien, 1994<br />

[3] A.Biberstein, H. Wörschnig, J.Queißer, H.H.<br />

Bernahrt, Überströmbare Dämme – landschaftsverträgliche<br />

Ausführungsvarianten für den dezentralen<br />

Hochwasserschutz in Baden-<br />

Württemberg, Institut für Bodenmechanik und<br />

Felsmechanik, Universität Karlsruhe, Förderkennzeichen<br />

BWC 20005 und BWT 22004 Land<br />

Baden-Württemberg, 2004<br />

[4] S. Dornack, Überströmbare Dämme – Beitrag zur<br />

Bemessung von Deckwerken aus Bruchsteinen,<br />

Wasserbauliche Mitteilungen Heft 20, Inst. f.<br />

Wasserbau und Techn. Hydromechanik, 2001<br />

[5] Landesanstalt für Umweltschutz (LfU) Baden-<br />

Württemberg-Oberirdische Gewässer, Gewässerökologie<br />

90, Überströmbare Dämme, 2006<br />

[6] Rathgeb A., Hydrodynamische Bemessungsgrundlagen<br />

für Lockerdeckwerke an überströmbaren<br />

Erdämmen, Mitteilungen, Institut für Wasserbau,<br />

Universität Stuttgart, Heft 109, 2001<br />

[7] Queißer J., Entwicklung landschaftsverträglicher<br />

Bauweisen für überströmbare Dämme, Mitteilungen<br />

des Instituts für Wasser und Gewässerentwicklung<br />

– Bereich Wasserwirtschaft und<br />

Kulturtechnik – der Universität Karlsruhe (TH),<br />

Heft 233, 2006<br />

[8] A. Kainrath, C. Henzinger, Interdiszilpinäre Seminararbeit<br />

– Überströmbare Dämme, Institut für<br />

Wasserbau und Ingenieurhydrologie, <strong>2008</strong>


Rediscovery <strong>of</strong> a Special Type <strong>of</strong> Separation Pier<br />

Michael Pucher and Peter Tschernutter (Faculty Mentor)<br />

Institute <strong>of</strong> Hydraulic and Water Resources Engineering<br />

Vienna University <strong>of</strong> Technology<br />

Vienna, Austria<br />

Email: michael.pucher@kw.tuwien.ac.at<br />

Abstract — At low head river hydropower plants<br />

<strong>the</strong> optimal flow to <strong>the</strong> turbines is <strong>of</strong> great importance<br />

for maximum power production. In <strong>the</strong> case <strong>of</strong><br />

<strong>the</strong> lateral arrangement <strong>of</strong> <strong>the</strong> power house, <strong>the</strong><br />

flow must be redirected toward <strong>the</strong> turbine inlet. In<br />

order to achieve <strong>the</strong> flow around <strong>the</strong> separation<br />

pier as loss free as possible, great attention is put<br />

on its design. In <strong>the</strong> literature only a few design<br />

rules can be found. Optimal geometries are mainly<br />

found by model tests. In <strong>the</strong> course <strong>of</strong> full model<br />

tests <strong>of</strong> a projected bay power plant, a special pier<br />

type was rediscovered as <strong>the</strong> best design solution<br />

for inflow optimization.<br />

I. INTRODUCTION<br />

For thousands <strong>of</strong> years hydroelectric power plants<br />

have already been used for environmental friendly<br />

energy production. In order to reach an optimal energy<br />

output – i.e. minimize energy loss in “Figure 1”<br />

- a careful design and, usually, model tests are necessary.<br />

This work is concerned with <strong>the</strong> structural design<br />

<strong>of</strong> bay power plant intakes. Bay power plants<br />

“Figure 2” belong to <strong>the</strong> low head power plants and<br />

form a special type <strong>of</strong> run-<strong>of</strong>-river power plants<br />

where <strong>the</strong> power house is arranged besides <strong>the</strong> weir<br />

in a lateral bay. Because <strong>of</strong> <strong>the</strong> lateral arrangement <strong>of</strong><br />

<strong>the</strong> power house, <strong>the</strong> current flow must be redirected<br />

toward <strong>the</strong> turbine inlet. In order to achieve <strong>the</strong> flow<br />

around <strong>the</strong> separation pier as loss-free as possible,<br />

great attention is put on its design.<br />

Figure 1: Magnitude <strong>of</strong> energy losses [1]<br />

First a literature review was undertaken to assess<br />

<strong>the</strong> state <strong>of</strong> <strong>the</strong> art in separation pier design.<br />

323<br />

In a fur<strong>the</strong>r step existing structures were evaluated.<br />

The main part <strong>of</strong> this paper focuses on <strong>the</strong> full model<br />

tests <strong>of</strong> a projected bay power plant.<br />

Figure 2: Bay plant. Typical Setting and Flow [2]<br />

II. STATE OF THE ART<br />

In [2, 3] an overview <strong>of</strong> low head hydropower plant<br />

design can be found, while [4, 5] try to give more<br />

detailed insight into bay plant design, flow criteria<br />

and energy losses. Views from <strong>the</strong> point <strong>of</strong> a turbine<br />

maker concerning criteria for determining favourable<br />

flow conditions are given in [6], while <strong>the</strong> newest<br />

development in <strong>the</strong> trash rack design is shown in [7].<br />

For one <strong>of</strong> <strong>the</strong> only detailed separation pier design<br />

rules, consult <strong>the</strong> still up to date [8], while in [9]<br />

computational fluid dynamics design with optimization<br />

algorithms are used to determine <strong>the</strong> optimal<br />

shape <strong>of</strong> <strong>the</strong> pier. In [2,5,8,9] different shapes <strong>of</strong><br />

separation piers are discussed. After reviewing proposed<br />

and existing separation pier geometries it becomes<br />

apparent that since each hydropower plant is a<br />

prototype, favourable geometries are, in spite <strong>of</strong> basic<br />

rules and recommendations, mainly found by model<br />

tests. In [10, 11] a special type <strong>of</strong> “split”-separation<br />

pier (pier with a guidance wall) has been used successfully<br />

since <strong>the</strong> 1960s, although no design rules<br />

and support in <strong>the</strong> literature can be found.<br />

III. PHYSICAL MODEL TESTS<br />

The physical model tests were performed in a temporary<br />

flume <strong>of</strong> approximately 25 m length and 3,75 m<br />

width in <strong>the</strong> institute’s own laboratory. The chosen


model scale was 1 : 40 in accordance with Froud´s<br />

Law,whichresultedinamaximumdesignflow<strong>of</strong><br />

20 l/s for two turbines. The projected hydro power<br />

plants performance in various fields like maximum<br />

flood discharge, energy dissipation, bed load transport<br />

and especially turbine inflow was tested.<br />

Figure 3: Fisher & Franke evaluation [2]<br />

The resulting inflow conditions were evaluated<br />

qualitatively using strings visualizing flow conditions.<br />

Qualitative performance evaluation could be<br />

achieved by examining velocity (v) distributions<br />

from a 10 x 10 raster at each turbine intake according<br />

to <strong>the</strong> Fisher & Franke (F&F) criteria [in 5]. By using<br />

different criteria it was possible to order <strong>the</strong> alternatives<br />

according to <strong>the</strong> quality <strong>of</strong> <strong>the</strong> inflow conditions<br />

(“Figure 3”,”Table 1”).<br />

Rank Geometry �-value F&F<br />

1 1,021 6,37<br />

7 1,046 1,51<br />

9<br />

Original<br />

1,077 -3,39<br />

Table 1: Ranking <strong>of</strong> Pier Geometries [4]<br />

The original design showed unfavourable inflow<br />

conditions into and at <strong>the</strong> turbine next to <strong>the</strong> weir. In<br />

a set <strong>of</strong> model tests first <strong>the</strong> intake shape was<br />

changed and <strong>the</strong>n <strong>the</strong> separation pier geometry was<br />

varied. The width <strong>of</strong> <strong>the</strong> separation pier could not be<br />

increased, as calculated according to [8]; <strong>the</strong>refore<br />

only slight improvements could be obtained with<br />

standard pier geometries (see “Table 1” rank 7 and<br />

9). Finally, only special solutions with additional<br />

guidance walls [8, 10, 11] in front <strong>of</strong> separation piers<br />

resulted in acceptable inflow conditions to <strong>the</strong> turbines.<br />

Up to a point all guidance walls/split pier<br />

solutions resulted in similar performance. In fur<strong>the</strong>r<br />

324<br />

tests it became apparent that <strong>the</strong> position <strong>of</strong> <strong>the</strong> guidance<br />

wall is <strong>the</strong> dominating factor.<br />

IV. OUTLOOK<br />

Fur<strong>the</strong>r research based on 3D simulation and fur<strong>the</strong>r<br />

physical model tests will be undertaken to clarify <strong>the</strong><br />

underlying hydraulic processes. Through understanding<br />

<strong>the</strong> hydraulics, it will be possible to define design<br />

rules that can be used in small scale hydro projects<br />

were 3D and model tests are too expensive.<br />

REFERENCES<br />

[1] J. Erlach: Betrachtung der Gesamtenergiebilanz<br />

einer Wasserkraftanlage mit Aufschlüsselung der<br />

Einzelverluste, in: Beiträge zum 2. Seminar<br />

Kleinwasserkraft; Praxis und aktuelle Entwicklung,<br />

Stuttgart, 1.10.1999, (Mitteilungen des Inst.<br />

F. Strömungsm. und Hydr. Strömungsm.<br />

[2] H. Blind: Wasserbauten aus Beton, Berlin: Ernst,<br />

Verl. für Architektur und techn. Wiss., 1987<br />

[3] J. Giesecke & E. Monsony: Wasserkraftanlagen:<br />

Planung, Bau und Betrieb, Berlin: Springer, 4.,<br />

aktual. u. erw. edition, 2005<br />

[4] T. Mayr: Bauliche und konstruktive Gestaltung<br />

von Buchtenkraftwerken, Diplomarbeit <strong>2008</strong>, Institut<br />

für Wasserbau und Ingenieurhydrologie,<br />

Technische Universität Wien, <strong>2008</strong><br />

[5] D. Godde: Experimentelle Untersuchungen zur<br />

Anströmung von Rohrturbinen. Wasserbau und<br />

Wasserwirtschaft: Berichte der Versuchsanstalt<br />

Obernach und des Lehrstuhls für Wasserbau und<br />

Wassermengenwirtschaft der Technischen Universität<br />

München;75, 1994), ISSN 0947-7187.<br />

[6] A. Nichtawitz: Kriterien und Methoden zur Optimierung<br />

der Einlaufströmung von Wasserturbinen,<br />

in: Konstruktiver Wasserbau - Landschaftswasserbau,<br />

Band 18, Institut für konstruktiven<br />

Wasserbau, TU Wien, 2004<br />

[7] H. Meusburger, Energieverluste an Einlaufrechen<br />

von Flusskraftwerken, Zürich: Eidgen.<br />

Techn. Hochschule, 2002 (Diss. ETH;14891).<br />

[8] G. Rouvé: Der Krafthaustrennpfeiler; Strömungsverhältnisse<br />

an gekrümmten Wänden,<br />

Karlsruhe: Eigenverlag, 1958 (Arbeit aus dem<br />

Theodor- Rehbock- Flussbaulaboratorium; 145).<br />

[9] G. Demny: Erschließung der automatischen<br />

Strömungsoptimierung zur Lösung von Gestaltungsaufgaben<br />

im Wasserbau. Aachen: Rheinisch-Westfälische<br />

Techn. Hochsch., Diss. 2004<br />

[10] RWE, Hydro power design maps, Personal<br />

communication<br />

[11] e.on, Hydro power design maps, Personal communication


Possibilities <strong>of</strong> bentonite substitution – development <strong>of</strong> new<br />

progressive grouting materials and durability monitoring.<br />

Pavla Matulova, (Rostislav Drochytka)<br />

Institute <strong>of</strong> Technology <strong>of</strong> Building Materials and Components<br />

Brno University <strong>of</strong> Technology<br />

Brno, Czech Republic<br />

Email: matulova.p@fce.vutbr.cz<br />

Abstract — The cracks and caverns <strong>of</strong> different<br />

types belong to <strong>the</strong> most frequent failures within<br />

concrete, reinforced concrete and masonry<br />

structures. These can appear also in <strong>the</strong> bottom<br />

surface, in rocks and grounds under <strong>the</strong> structure.<br />

Very efficient method <strong>of</strong> <strong>the</strong>se failures rehabilitation<br />

and repair is <strong>the</strong> grouting with a medium which has<br />

better physical properties than <strong>the</strong> original<br />

structure. Paper is focused on <strong>the</strong> monitoring<br />

durability <strong>of</strong> new progressive grouting mixtures.<br />

The paper describes <strong>the</strong> possibilities <strong>of</strong> grouting<br />

strategy and especially <strong>the</strong> development <strong>of</strong> new<br />

progressive materials.<br />

I. INTRODUCTION<br />

The grouting is <strong>the</strong> process <strong>of</strong> pumping <strong>the</strong> liquid<br />

with variable viscosity into <strong>the</strong> ground, into <strong>the</strong><br />

fissured or loose ground, concrete or masonry in<br />

order to increase <strong>the</strong> density <strong>of</strong> <strong>the</strong>se materials. The<br />

classical technology <strong>of</strong> grouting is well known for<br />

some centuries. The principle <strong>of</strong> this technology is<br />

<strong>the</strong> filling <strong>of</strong> pores and cavities by <strong>the</strong> grouting<br />

mixture. The mixing <strong>of</strong> <strong>the</strong> grouting mixture with <strong>the</strong><br />

original material forms a composite which has new<br />

physical properties. This is most frequently <strong>the</strong> way<br />

to improve <strong>the</strong> strength and <strong>the</strong> imperviousness <strong>of</strong> <strong>the</strong><br />

material. Ano<strong>the</strong>r typical application is <strong>the</strong><br />

waterpro<strong>of</strong>ing <strong>of</strong> cracks and gaps (to secure <strong>the</strong> water<br />

tightness <strong>of</strong> cracks and gaps). Basic characteristics <strong>of</strong><br />

grouting are: good workability, volume stability,<br />

good penetration properties, good pump ability, great<br />

resistance against erosion, sufficient compression<br />

strength. And technological and technical variables<br />

are grouting material curing time <strong>of</strong> <strong>the</strong> material, way<br />

<strong>of</strong> fastening, fastening <strong>of</strong> spacing distance, grouting<br />

pressure, grouting time for one connecting opening,<br />

grouted volume for one connecting opening,<br />

sequence <strong>of</strong> works. The problem <strong>of</strong> building<br />

structures subsoil hardening, <strong>of</strong> filling <strong>the</strong> cavities<br />

and <strong>the</strong> formed caverns became to be a very topical<br />

subject in <strong>the</strong> Czech Republic, especially after <strong>the</strong><br />

year 2002, when <strong>the</strong> Czech Republic suffered under<br />

extensive floods, which caused <strong>the</strong> significant<br />

disturbance <strong>of</strong> foundation subsoil. These problems<br />

325<br />

can be successfully solved by <strong>the</strong> utilization <strong>of</strong><br />

grouting technologies. This concerns <strong>the</strong> application<br />

<strong>of</strong> large scale grouting and possibilities are looked<br />

for to decrease <strong>the</strong> final price under keeping <strong>the</strong><br />

demanded parameters for <strong>the</strong> grout mixtures.<br />

Considering <strong>the</strong> fact, that <strong>the</strong> floods came also in<br />

year 2006, though <strong>the</strong>y were not so great, <strong>the</strong> subject<br />

<strong>of</strong> large scale grouting is still very topical. The<br />

number <strong>of</strong> structures rehabilitated by grouting<br />

materials increases <strong>the</strong> demands on <strong>the</strong> research and<br />

development <strong>of</strong> new building materials. Large-scale<br />

grouting in <strong>the</strong> ground work is connected with<br />

increasing demand to solve <strong>the</strong> question <strong>of</strong> large<br />

scale utilization <strong>of</strong> industrial wastes in <strong>the</strong> largest<br />

possible extent. The utilization <strong>of</strong> wastes would not<br />

only partially solve <strong>the</strong> problem with wastes disposal<br />

but <strong>the</strong> wastes application would have positive effect<br />

on <strong>the</strong> price <strong>of</strong> <strong>the</strong> work.<br />

II. THE METHODICS OF REALIZED<br />

WORK<br />

The Institute <strong>of</strong> Technology <strong>of</strong> Building Materials<br />

and Components at <strong>the</strong> Faculty <strong>of</strong> Civil Engineering,<br />

Brno University <strong>of</strong> Technology in <strong>the</strong> Czech<br />

Republic is already years concerned with <strong>the</strong> solution<br />

<strong>of</strong> optimization problems connected with <strong>the</strong><br />

selection <strong>of</strong> <strong>the</strong> material type and <strong>the</strong> dose <strong>of</strong> waste<br />

materials. The methodics <strong>of</strong> realized work were<br />

devided into three main parts. First <strong>of</strong> all <strong>the</strong> atention<br />

was paid on substitution <strong>of</strong> filler by 10% till 50% <strong>of</strong><br />

waste material. We have performed <strong>the</strong> experimental<br />

erification <strong>of</strong> <strong>the</strong> partial substitution effect <strong>of</strong> filler<br />

by <strong>the</strong> grouting mixture. The second stage was<br />

focused on substitution <strong>of</strong> bending materials. And in<br />

<strong>the</strong> third stage <strong>of</strong> realized work was concerning to<br />

bentonite stage. The fresh mixtures <strong>of</strong> all selected<br />

formulae were tested by following basic tests:<br />

Consistency <strong>of</strong> <strong>the</strong> mixture, initial and final setting<br />

time, bending strength on test samples (40/40/160<br />

mm), compression strength on test samples<br />

(40/40/160 mm), volume mass, shrinkage during<br />

hardening on test samples (40/40/160 mm). And <strong>the</strong><br />

most successful mixture were monitoring also by <strong>the</strong><br />

durability view.


III. EXPERIMENTAL PART<br />

Aim <strong>of</strong> this paper is to describe <strong>the</strong> durability<br />

monitoring. The Attention was paid first <strong>of</strong> all<br />

physico-mechanical parameters <strong>of</strong> <strong>the</strong> hardened<br />

mixture after 180 days expoatation in corosive<br />

solution.<br />

corosive solution. concetration<br />

sulphate SO3 2- 10 000 mg/l<br />

chloride NaCl 1 000 mg/l<br />

cyclic action Chloride NaCl 1 000 mg/l<br />

CO2<br />

80 mg/l<br />

Table 1: expoation in corosive solution.<br />

RESULTS IN GRAPHS<br />

25<br />

20<br />

15<br />

10<br />

5<br />

0<br />

-5<br />

-10<br />

10<br />

5<br />

0<br />

-5<br />

-10<br />

-15<br />

-20<br />

22,2<br />

17,6<br />

10,5<br />

17,5<br />

12,8<br />

2,2<br />

5,9<br />

9,9<br />

9,4<br />

-0,7-1,2<br />

-3,2 -2,9 -3,1<br />

-5,1 -4,6<br />

-5,7<br />

-7,4<br />

sírany 2-<br />

sulphate NaCl NaCl cyklicky Agresivní CO2<br />

SO3<br />

NaCl cyclic action Chloride NaCl CO2<br />

S ref SP SS SP zeolit 75% SP sádra SS zeolit 75%<br />

Figure 1: Tensile strength change under <strong>the</strong><br />

exploration <strong>of</strong> corrosive solution<br />

-7,5<br />

-9,6<br />

-5,4-5,3-5,7<br />

-9,8<br />

-7,9<br />

-9,4<br />

1,7 1,6<br />

-0,31 -0,3 -0,3<br />

-5,9<br />

-17 -17,3<br />

-4,3<br />

-5,3<br />

2,9<br />

-5,8<br />

5,4<br />

1,2<br />

8,8<br />

-10,7 -10,3<br />

2-<br />

sulphate SO3<br />

NaCl cyclic action Chloride NaCl CO2<br />

sírany NaCl NaCl cyklicky Agresivní CO2<br />

S ref SP SS SP zeolit 75% SP sádra SS zeolit 75%<br />

Figure 1: Compressive tensile strength change<br />

under <strong>the</strong> exploration <strong>of</strong> corrosive solution<br />

1,5<br />

-4<br />

-3,9<br />

4<br />

0,9<br />

326<br />

MICRO-STRUCTURAL ANALYSES IS ELECTRON MICROSCOPY<br />

Figure 1: Structure <strong>of</strong> grouting matrix – exploation<br />

<strong>of</strong> chloride NaCl 1 000 mg/l (enlarged -I.4000x<br />

II. 5000x), 200 days exploatation<br />

IV. CONCLUSION<br />

It was fur<strong>the</strong>r proved that it is possible with <strong>the</strong><br />

suitably selected substitution <strong>of</strong> <strong>the</strong> original filler<br />

bonding and bentonite part by waste materials to<br />

manufacture grouting materials for utilization in<br />

practice which fulfill <strong>the</strong> given conditions. Corrosive<br />

resistance <strong>of</strong> modifacated mixtures by by-product is<br />

comparable with <strong>the</strong> resistance <strong>of</strong> reference basic<br />

mixture, and in some case (see graphs were corrosive<br />

resistance by modificated mixture exceeded as<br />

well).The first stage shown very good results and in<br />

following research was provided that it is possible to<br />

substitute also bending material There were 40% <strong>of</strong><br />

bonding material (cement) decreasing in <strong>the</strong><br />

modificated mixture. The research describes <strong>the</strong><br />

possibilities <strong>of</strong> development <strong>of</strong> new rehabilitation<br />

grouting materials containing different wastes (fly<br />

ash, washing wastes, etc.) under respecting <strong>the</strong><br />

increase <strong>of</strong> quality. In praxis this research could case<br />

lower prices <strong>of</strong> large scale grouting works.<br />

ACKNOWLEDGMENTS<br />

The given problems are solved in <strong>the</strong> framework <strong>of</strong><br />

<strong>the</strong> Grant Project no.103/05/H044 called:<br />

“Stimulation <strong>of</strong> Doctorands Scientific Development<br />

in <strong>the</strong> Branch <strong>of</strong> Building-Materials Engineering”<br />

and in <strong>the</strong> framework <strong>of</strong> <strong>the</strong> research project MSM<br />

0021630511 called: “Progressive Building Materials<br />

with Utilization <strong>of</strong> Secondary Raw Materials and<br />

<strong>the</strong>ir Effect on Service Life <strong>of</strong> Structures”.<br />

REFERENCES<br />

[1] EMMONS, P. H., DROCHYTKA, R., JE�ÁBEK<br />

Rehabilitation and Maintenance <strong>of</strong> Concrete in Pictures. Brno<br />

CERM 1999, ISBN 0-87629-286-4 (in Czech)<br />

[2] P ŠÍMA, J. Research and development <strong>of</strong> new progressive<br />

grouting system. Dissertation, 2007 (in Czech) Brno 2007


Combined Active/Passive Microwave Wavelet-Based Approach<br />

For Snowmelt Detection Over Antarctica Ice Shelves<br />

Nicholas Steiner and Marco Tedesco (Faculty Mentor)<br />

The Graduate Center <strong>of</strong> The City University <strong>of</strong> New York<br />

New York, USA<br />

Email: NSteiner@gc.cuny.edu<br />

Abstract — The presence <strong>of</strong> wet snow on Antarctica<br />

can be monitored using Earth orbiting active<br />

microwave sensors, such as QuickScat, using several<br />

threshold-based techniques. The goal <strong>of</strong> this<br />

study is to apply several signal thresholding techniques<br />

to microwave backscattering data and<br />

evaluates <strong>the</strong>m based on comparison to passive<br />

sensors and ground measurements Since effects <strong>of</strong><br />

climate change will be most evident in Polar Regions,<br />

monitoring changes in <strong>the</strong> cryosphere is<br />

imperative to understanding global climate change.<br />

The seasonal geographic extent <strong>of</strong> melted snow and<br />

<strong>the</strong> temporal persistence <strong>the</strong>se melt areas are a<br />

direct indicator <strong>of</strong> <strong>the</strong> climactic conditions at <strong>the</strong><br />

poles. Active microwave sensors give us <strong>the</strong> ability<br />

to measure <strong>the</strong>se quantities with a high temporal<br />

and spatial resolution, compared to passive sensors.<br />

The SeaWinds scatterometer mounted on <strong>the</strong><br />

QuickScat satellite provides daily, near global,<br />

coverage <strong>of</strong> Antarctica from 1999 to <strong>the</strong> present.<br />

This data will be compared with Special Sensor<br />

Microwave/Imager (SSM/I) passive microwave<br />

derived melting products and measurements taken<br />

at ground stations on Antarctica.<br />

I. INTRODUCTION<br />

Since effects <strong>of</strong> climate change will be most evident<br />

in Polar Regions [1], monitoring changes in <strong>the</strong><br />

cryosphere is imperative to understanding global<br />

climate change. The seasonal geographic extent <strong>of</strong><br />

melted snow and <strong>the</strong> temporal persistence <strong>the</strong>se melt<br />

areas are a direct indicator <strong>of</strong> <strong>the</strong> climactic conditions<br />

at <strong>the</strong> poles. Active microwave sensors give us <strong>the</strong><br />

ability to measure <strong>the</strong>se quantities with a high temporal<br />

and spatial resolution, compared to passive sensors.<br />

The introduction, through a change <strong>of</strong> phase, <strong>of</strong><br />

liquid water into a volume <strong>of</strong> snow will dramatically<br />

alter its microwave properties. The absorption <strong>of</strong><br />

incoming microwave energy will dominate scattering,<br />

leading to a marked decrease in <strong>the</strong> observed<br />

backscatter signal. The SeaWinds scatterometer<br />

mounted on <strong>the</strong> QuickScat satellite provides daily,<br />

near global, coverage <strong>of</strong> Antarctica from 1999 to <strong>the</strong><br />

present. This data is readily available from <strong>the</strong><br />

NASA Scatterometer Climate Record Pathfinder<br />

(SCP) [2].<br />

327<br />

II. METHODS<br />

Identifying <strong>the</strong> presence <strong>of</strong> wet snow is most <strong>of</strong>ten<br />

achieved by defining a threshold value for change in<br />

measured signal and assuming that all signals above<br />

that value represent wet snow. Methods for deriving<br />

a valid threshold have been put forth by various<br />

authors, and include a constant threshold, one based<br />

on wintertime signal statistics [3] as well as a novel<br />

technique based on <strong>the</strong> thresholding <strong>of</strong> wavelettransform<br />

melt onset detection [4]. To our knowledge,<br />

a wavelet-transform based edge detection approach<br />

has never been applied to active microwave<br />

observations <strong>of</strong> Antarctica. The application <strong>of</strong> all<br />

previously mentioned techniques will be performed<br />

on QuickScat data from 1999 to <strong>the</strong> present to obtain<br />

<strong>the</strong> spatial extent <strong>of</strong> Antarctic and <strong>the</strong> persistence <strong>of</strong><br />

this melt, <strong>the</strong> melting index (MI). This dataset will<br />

be used to compare with ground based measurements<br />

as well as passive measurements over <strong>the</strong> same time<br />

period, to evaluate <strong>the</strong> effectiveness <strong>of</strong> each thresholding<br />

technique.<br />

III. RESULTS<br />

A constant threshold <strong>of</strong> 3 dB, based on an approximate<br />

<strong>of</strong> 8 standard deviations from <strong>the</strong> average winter<br />

signal is suggested in many approaches. This<br />

corresponds to layer <strong>of</strong> wet snow with a 3.8 cm<br />

depth. Results obtained with this threshold value<br />

leads to <strong>the</strong> detection <strong>of</strong> melt in areas where it is not.<br />

Patterns in <strong>the</strong>se ‘false’ melting areas appear to be<br />

consistent to <strong>the</strong> local topography, so that <strong>the</strong>se may<br />

be due to local elevation variations. The effects <strong>of</strong><br />

<strong>the</strong>se are most evident on <strong>the</strong> Wilkes ice shelf and<br />

may account for a 20% overestimation. These effects<br />

are removed by increasing <strong>the</strong> threshold value<br />

to 7 Db “Figure 1”.


Backscattering Coefficient (dB)<br />

0<br />

-10<br />

-20<br />

-30<br />

Winter<br />

150 200 250 300 350 400 450 500<br />

Day<br />

Average<br />

Thresh = 3 dB<br />

Thresh = 7 dB<br />

Figure 1: A single pixel record <strong>of</strong> backscattering<br />

measurements for 2000-2001 illustrating several<br />

melting thresholds<br />

Results for <strong>the</strong> melting index and extent calculated<br />

with a 7 dB threshold agree, to a large extent, to<br />

recently published SSMI results [5], although <strong>the</strong>re<br />

does appear to be a discrepancy in defining <strong>the</strong> melt<br />

year ”Figure 2”. The melting index does not appear<br />

to be as well matched, and <strong>the</strong> absolute values do not<br />

have overlapping ranges.<br />

Figure 2: Melting Index and Extent calculated<br />

using a variable threshold based on wintertime<br />

values.<br />

REFERENCES<br />

[1] Stainforth, D.A., T. Aina, C. Christensen, M.<br />

Collins, N. Faull, D.J. Frame, J.A. Kettleborough,<br />

S. Knight, A. Martin, J.M. Murphy, C. Piani,<br />

D. Sexton, L.A. Smith, R.A. Spicer, A.J.<br />

Thorpe, and M.R. Allen, Uncertainty in predictions<br />

<strong>of</strong> <strong>the</strong> climate response to rising levels <strong>of</strong><br />

greenhouse gases, Nature, 433, 403-406, 2005.<br />

[2] Brigham Young University Center for Remote<br />

Sensing; NASA Scatterometer Climate Record<br />

Pathfinder;http://www.scp.byu.edu/<br />

328<br />

[3] Ashcraft, I. & Long, D. Comparison <strong>of</strong> methods<br />

for melt detection over Greenland using active<br />

and passive microwave measurements. Int. J.<br />

Remote Sens. 27, 2469-2488 (2006).<br />

[4] Liu, H., Wang, L. & Jezek, K. C. Wavelettransform<br />

based edge detection approach to derivation<br />

<strong>of</strong> snowmelt onset, end and duration from<br />

satellite passive microwave measurements. Int. J.<br />

Remote Sens. 26, 4639-4660 (2005).<br />

[5] Tedesco, M. Updated <strong>2008</strong> Surface Snowmelt<br />

Trends in Antarctica. Eos Trans.AGU 89, 126<br />

(<strong>2008</strong>).


Binary Quadratic Forms<br />

over <strong>the</strong> Modular Ring �/n� Clyde W. Lewis, Dr. Rony Gouraige<br />

Lehman College &<br />

Bronx Community College<br />

Dept. <strong>of</strong> Ma<strong>the</strong>matics<br />

New York, United States<br />

cwlewis777@aol.com rgouraige@yahoo.com<br />

Abstract � Given a binary quadratic form<br />

2 2<br />

ax +bxy+cy<br />

where a, b, c � �.<br />

<strong>the</strong>n it is a known result <strong>of</strong> number<br />

<strong>the</strong>ory that <strong>the</strong> reduction <strong>of</strong> Binary quadratic forms<br />

depends on <strong>the</strong> coefficients a, b, and c, while <strong>the</strong><br />

equivalence <strong>of</strong> two Binary quadratic forms outwardly<br />

depends on <strong>the</strong> variables x and y. We now relate <strong>the</strong>se<br />

two observations to <strong>the</strong> notion <strong>of</strong> binary quadratic<br />

forms over <strong>the</strong> modular ring �/ n�,<br />

and ask, how<br />

much <strong>of</strong> what is known about <strong>the</strong> reduction <strong>of</strong> <strong>the</strong>se<br />

forms depend on a, b, and c being integers? At what<br />

point in establishing equivalence do we use <strong>the</strong> fact<br />

that x and y are integers? These are questions that if<br />

answered may explain <strong>the</strong> behavior <strong>of</strong> <strong>the</strong>se forms not<br />

apparent to us over an integral domain.<br />

Elements in �� /n�<br />

<strong>of</strong> <strong>the</strong> form<br />

2 2<br />

x �� y<br />

Before we investigate <strong>the</strong> general equivalence and<br />

reduction results <strong>of</strong> binary quadratic forms over <strong>the</strong><br />

modular ring �/ n�,<br />

we consider a classic result in<br />

number <strong>the</strong>ory by Fermat, which states that an odd<br />

prime p may be written as a sum <strong>of</strong> two squares <strong>of</strong><br />

integers,<br />

2 2<br />

p=x +y � p�1mod4.<br />

� �<br />

We may extend this fact and state that every prime<br />

in <strong>the</strong> factorization <strong>of</strong> a natural number a which is<br />

congruent to 3 modulo 4, i.e.<br />

p � 3 �mod4�, has to be occurring to an even power in order for a to<br />

be written as a sum <strong>of</strong> two squares. By generalizing<br />

this <strong>the</strong>orem over <strong>the</strong> modular ring �/ n�,<br />

we are<br />

asking, which elements in �/ n�<br />

can be written as a<br />

sum <strong>of</strong> two squares. And conveniently we begin with<br />

<strong>the</strong> basic binary quadratic form<br />

2 2<br />

Q(x, y) = x + y<br />

which is appropriately a sum <strong>of</strong> two squares.<br />

The first step in characterizing those n for which<br />

every element in �/ n�<br />

is a sum <strong>of</strong> two squares is to<br />

329<br />

compute n up to some considerable value and observe<br />

a pattern from which we may develop some function<br />

<strong>of</strong> n.<br />

“Table 1” lists <strong>the</strong> elements <strong>of</strong> �/ n� which are<br />

sums <strong>of</strong> two squares for<br />

n �� 2, 3, 4, 8,10,12, 25<br />

Modular ring<br />

� / n�<br />

Sums <strong>of</strong> squares in<br />

� / n�<br />

�/2� All elements are sums <strong>of</strong> squares<br />

�/3� All elements are sums <strong>of</strong> squares<br />

�/4� �0� �1� �2� � � � � � �<br />

�/8� �0� �1� �2� �4� �5� � � � � � � � � � �<br />

�/10� All elements are sums <strong>of</strong> squares<br />

�/12� �0� �1� �2� �4� �5� �6� �8� �9� �10� � � � � � � � � � � � � � � � � � �<br />

�/25� All elements are sums <strong>of</strong> squares<br />

Table 1.<br />

Clearlywehavenotcomputednupto a value that<br />

would indicate a concrete pattern, but we may be able<br />

to include or exclude certain propositions that may or<br />

may not warrant investigation. For example one could<br />

ask, “for which n is it true that every element in<br />

�/ n�<br />

is a sum <strong>of</strong> two squares? “ And after an initial<br />

observation <strong>of</strong> “Table 1” one could conjecture, “If n<br />

is a product <strong>of</strong> distinct primes <strong>the</strong>n every element in<br />

�/ n�<br />

can be written as a sum <strong>of</strong> two squares.” Now<br />

we have confirmed that this is true for all values <strong>of</strong> n,<br />

where 2 � n � 24, and we have also confirmed that<br />

for every n in �/ n�for<br />

n between 2 and 24 where<br />

<strong>the</strong> elements can be written as a sum <strong>of</strong> two squares, is<br />

a product <strong>of</strong> distinct primes. However this conjecture<br />

collapses at n= 25,<br />

since in �/25� all elements<br />

may be written as a sum <strong>of</strong> two squares but 25 is a<br />

2<br />

product <strong>of</strong> repeated primes i.e., 25 �� 5 .


Now a reasonable conjecture is, “If n is a prime,<br />

<strong>the</strong>n every element in �/ n�<br />

is a sum <strong>of</strong> two<br />

squares”. Now we again have confirmed that every<br />

element in �/ n�<br />

in which n is a prime is indeed a<br />

sum <strong>of</strong> two squares for 2 � n � 25. This observation<br />

forces one to consider <strong>the</strong> “prime directive” in number<br />

<strong>the</strong>oretic investigations, and that is, we always begin<br />

with an exploration <strong>of</strong> <strong>the</strong> prime numbers. We do this<br />

for two reasons. First, patterns are generally easier to<br />

notice for primes numbers. Second, patterns for prime<br />

numbers are regularly used to figure out patterns for<br />

all numbers, since <strong>the</strong> Fundamental Theorem <strong>of</strong><br />

Arithmetic says that <strong>the</strong> primes are <strong>the</strong> central<br />

building blocks for all numbers. So if we only look at<br />

n, when n is a prime p, <strong>the</strong>n <strong>the</strong> conjecture holds for<br />

2 � p � 25, but until we are able to prove or disprove<br />

this conjecture for all primes p, in �/ p�<br />

we are only<br />

certain it is true for 2 � p � 25. A pro<strong>of</strong> <strong>of</strong> this<br />

conjecture is pending.<br />

Binary Quadratic Forms over �� /n�.<br />

Let<br />

2 2<br />

f �x,y � = ax +bxy+cy<br />

<strong>the</strong>n we denote <strong>the</strong> associated matrix <strong>of</strong> f<br />

�a �b � 2<br />

b<br />

2�<br />

=F<br />

c �<br />

�<br />

where <strong>the</strong> entries on <strong>the</strong> diagonal are <strong>the</strong> coefficients<br />

<strong>of</strong> <strong>the</strong> squared terms and whose entries <strong>of</strong>f <strong>the</strong><br />

diagonal are half <strong>the</strong> coefficients <strong>of</strong> <strong>the</strong> cross product<br />

term. In addition, since an associated matrix <strong>of</strong> G is<br />

uniquely determined by <strong>the</strong> coefficients <strong>of</strong> <strong>the</strong> binary<br />

quadratic form g, we may conclude that f is<br />

equivalent to g if <strong>the</strong>re exist an element<br />

M SL � such that,<br />

� 2<br />

� �<br />

t<br />

M FM = G<br />

We now attempt to determine how much <strong>of</strong> <strong>the</strong><br />

notion <strong>of</strong> equivalence <strong>of</strong> binary quadratic forms carry<br />

over in <strong>the</strong> modular ring �/ p�.<br />

keep in mind, if p, is<br />

prime <strong>the</strong>n our conjecture, “If p is a prime, <strong>the</strong>n every<br />

element in �/ p�<br />

is a sum <strong>of</strong> two squares”, will<br />

satisfy <strong>the</strong> squared terms <strong>of</strong> a binary quadratic form.<br />

We begin with an example <strong>of</strong> two forms we are<br />

familiar with.<br />

2 2 2 2<br />

f �x,y �= x + y and g�x,y �=<br />

x +2xy+2y<br />

we know <strong>the</strong> unimodular matrix M that takes f to g is<br />

<strong>of</strong> <strong>the</strong> form<br />

�1 1�<br />

M= � �<br />

�0 1�<br />

�1� 330<br />

and <strong>the</strong> matrix equation in (1) illustrates <strong>the</strong><br />

equivalence between f and g. If we demonstrate this<br />

example over <strong>the</strong> modular ring �/2� we see that<br />

But<br />

t<br />

M F M G<br />

�1 0��1 0��1 1� �1 1�<br />

�<br />

�<br />

�1 1�� ��0 1�� ��0 1� �<br />

�<br />

�1 2�<br />

�<br />

�� 2 � �� 0<br />

in �/2� so <strong>the</strong> form associated G is<br />

but again<br />

<strong>the</strong>refore<br />

2<br />

� � � 2<br />

g x, y x + 2xy<br />

�� 2 � �� 0<br />

2<br />

� � � 2<br />

g x, y x<br />

and <strong>the</strong> equivalence between f and g is unfamiliar to<br />

us in �/2� since <strong>the</strong> desired form was not obtained.<br />

Preliminary results regarding <strong>the</strong> equivalence <strong>of</strong><br />

binary quadratic forms over �/ p�<br />

looks promising<br />

but one simple case may not show <strong>the</strong> behavior <strong>of</strong><br />

<strong>the</strong>se forms over �/ p�<br />

for all p. However, based on<br />

our first few attempts it seems that if <strong>the</strong> greatest<br />

coefficient <strong>of</strong> <strong>the</strong> entire matrix equation in (1) is less<br />

than p in �/ p�<br />

<strong>the</strong>n <strong>the</strong> equivalence amid two<br />

binary quadratic forms over <strong>the</strong> modular ring �/ p�<br />

is analogous to <strong>the</strong> equivalence <strong>of</strong> <strong>the</strong>se forms over<br />

<strong>the</strong> integers �. And if <strong>the</strong> greatest coefficient <strong>of</strong> <strong>the</strong><br />

entire matrix equation in (1) is equal to or greater than<br />

p in �/ p�<br />

<strong>the</strong>n <strong>the</strong> matrix equation changes.<br />

REFERENCES<br />

[1] William J. Leveque, Topics in Number Theory.<br />

Dover Publications, Inc. Mineola, New York<br />

2002.<br />

[2] Stillwell, John. Elements <strong>of</strong> Number Theory.<br />

Springer. New York 2003.<br />

[3] Armenda’riz, E.P., & McAdam, S.J., Elementary<br />

Number Theory. Macmillan. New York 1980.<br />

[4] Buell, D.A., Binary Quadratic Forms: Classical<br />

Theory and Modern Computations. Springer-<br />

Verlag. New York. 1989.


Index <strong>of</strong> Authors<br />

Able, Richard, 249<br />

Achong, Colleen, 229<br />

Adam, Helmut, 47<br />

Adamec, Filip, 23<br />

Ahmed, Sam, 63, 277<br />

Ahn, Sang-Il, 197<br />

Alfano, Robert, 121, 237<br />

Amaya, Ronny, 171<br />

Amrhein, Wolfgang, 13<br />

Andrianova, Ekaterina, 57<br />

Angela, Ligorio, 189<br />

Anna Maria, DOnghia, 189<br />

Antonio, Ippolito, 189<br />

Archodoulaki, Vasiliki-Maria, 209<br />

Arnold, Craig, 231<br />

Arriaran, Vilma, 39<br />

Azar, Amir, 143<br />

Bünte, Sven, 115<br />

Babic, Frantisek, 15<br />

Bachinger, Angelika, 157<br />

Badurek, Gerald, 129<br />

Bashir, Shazia, 213<br />

Bauer, Gudrun E., 161<br />

Bauer, Heidi, 293<br />

Baumann, Stefan O., 137, 139, 199<br />

Becker, Klaus, 215, 315<br />

Bendova, Maria, 139, 191, 199<br />

Beneova, Anna, 33<br />

Benigni, Caterina, 211<br />

Berger, Johannes, 269<br />

Berghöfer, Thomas, 9<br />

Bikson, Marom, 171<br />

Birke, Ronald, 239<br />

Bleyer, Michael, 71<br />

331<br />

Bouzi, Pierre, 187<br />

Bräutigam, Klaus-Rainer, 259<br />

Brooks, Patricia, 123<br />

Brunthaler, Stefan, 41<br />

Bustamante, Miguel, 277<br />

Cao, Xiujuan, 173<br />

Cardoso, Luis, 169<br />

Carrion, Carmen, 123<br />

Caseiro, Alex, 293<br />

Cerimovic, Samir, 287<br />

Chang, Jingjing, 241<br />

Coppola, Antonio, 73<br />

Corredor, Charlie, 227<br />

Cummins, Herman, 173<br />

Cunningham, Dan, 179<br />

Dörsek, Philipp, 283<br />

Danninger, Herbert, 147<br />

DAO Tran, Minh, 83<br />

Datta, Abhisheck, 171<br />

Daxner, Thomas, 183<br />

Delvai, Martin, 85<br />

Deneva, Margarita, 223<br />

Despotovic, Vladimir, 75<br />

Diblík, Jan, 151<br />

Diechle, Dominic, 135<br />

Dillinger, Andreas, 297<br />

Dimitrijevic, Milan, 59<br />

Dimitrijevic, Milan R., 305<br />

Dodt, Hans-Ulrich, 215, 315<br />

Dorn, Jürgen, 57, 77<br />

Dorsinville, Roger, 187<br />

Dostal, Tomas, 17<br />

Douglas, Kareem, 229


Duan, Yi, 163<br />

Dumke, Hartmut, 297<br />

Dvorák, Radek, 27<br />

Dyer, Rushane, 63<br />

Edtmaier, Christian, 145, 153<br />

Eing, Christian, 9<br />

Eisenmenger-Sittner, Christoph, 179<br />

Eiter, Thomas, 29, 83<br />

El-Salloum, Christian, 95<br />

Elmenreich, Wilfried, 53, 61<br />

Eryilmaz, Ertan, 233<br />

Etienne, Michael, 187<br />

Fackler, Karin, 317<br />

Fanghänel, Thomas, 131, 133, 177<br />

Fanghänel, Thommas, 185<br />

Fehérvári, István, 53<br />

Fenz, Stefan, 69<br />

Ferraz-Leite, Samuel, 257<br />

Fertalj, Kresimir, 87<br />

Filipp, Stefan, 129<br />

Fillos, John, 313<br />

Fink, Michael, 29<br />

Flasik, Radoslav, 175<br />

Flickinger, Bianca, 9<br />

Fröhlich, Johannes, 195, 211, 217<br />

Franco, Nigro, 189<br />

Frank, Andrew U., 311<br />

Frey, Wolfgang, 9<br />

Fruhmann, Philipp, 217<br />

Frybortova, Iva, 205<br />

Fryza, Tomas, 23<br />

Fu, Qing, 255<br />

Göschka, Karl M., 113<br />

Gall, Michael, 37<br />

Gartner, Georg, 107<br />

Gayen, Swapan, 121<br />

Gebhard, Thomas, 203<br />

Geckeis, Horst, 177, 185<br />

Geist, Andreas, 131, 185<br />

Gelautz, Margrit, 71<br />

Gewies, Andreas, 215<br />

332<br />

Ghose, Ranajeet, 233<br />

Gierl, Christian, 147<br />

Gonzalez, Jorge, 245<br />

Gonzalez, Tahnee, 259<br />

Gouraige, Rony, 329<br />

Green, Michael, 247<br />

Gross, Barry, 277<br />

Gruber, Wolfgang, 13<br />

Gschwandtner, Theresia, 91<br />

Gspan, Christian, 145<br />

Gu, Xiang, 169<br />

Guadagini, Alberto, 73<br />

Guzman, Francisco, 169<br />

Hahm, Sung-Ho, 197<br />

Halwachs, Michael, 181<br />

Hametner, Christian, 195, 217<br />

Hasegawa, Yuji, 129<br />

Helm, Manfred, 197<br />

Hermany, Jirí, 67<br />

Hetaba, Walid, 149<br />

Heurix, Johannes, 79<br />

Hinterstoisser, Barbara, 317<br />

H<strong>of</strong>stötter, Ursula S., 305<br />

H<strong>of</strong>stoetter, Ursula, 59<br />

Horkel, Ernst, 211, 217<br />

Hosannah, Nathan, 245<br />

Hula, Robert Christian, 153<br />

Husinsky, Wolfgang, 213<br />

Iqbal, Naseem, 201<br />

Isabella, Pentimone, 189<br />

Jährling, Nina, 215, 315<br />

Jankowski, Nicole, 207<br />

Jaros, Jiri, 111<br />

Jerabek, Michal, 43<br />

Jilková, Jana, 99<br />

John, George, 229<br />

Jurcisin, Robert, 167<br />

Köck, Anton, 145<br />

Kadlec, Albrecht, 119<br />

Kaiser, Katharina, 91


Kalchmair, Stefan, 197, 215, 315<br />

Karabatic, Ana, 265<br />

Karl, Hans-Peter, 195<br />

Kassab, Mohamed, 73<br />

Kastil, Jan, 103<br />

Kellner, Gudrun, 93<br />

Keplinger, Franz, 287<br />

Khamis, Youssef, 189<br />

Khanbilvardi, Reza, 143<br />

Kickelbick, Guido, 157, 193<br />

Kirner, Raimund, 115, 117, 119<br />

Kitzmantel, Michael, 179<br />

Klösch, Bernhard, 195<br />

Klatzer, Barbara, 293<br />

Klenze, Reinhardt, 131, 133<br />

Klepp, Jürgen, 129<br />

Knoop, Jens, 41, 45, 113<br />

Kobierský, Petr, 109<br />

Koch, Thomas, 209<br />

Koeberl, Bernd, 269<br />

K<strong>of</strong>ler, Heinrich, 219, 223<br />

Kollegger, Johann, 269<br />

Konegger, Thomas, 155<br />

Korenek, Jan, 103<br />

Kothleitner, Gerhard, 145<br />

Kovács, Gábor, 309<br />

Kramer, Edgar R., 215<br />

Krcal, Jan, 43<br />

Krennwallner, Thomas, 29<br />

Krongtaew, Chularat, 317<br />

Krupa, Marek, 33<br />

Kutzelnigg, Reinhard, 281<br />

Labutov, Igor, 31<br />

Ladenbauer, Josef, 59, 305<br />

Lamaddalena, Nicola, 73<br />

Lambert, Tiffany, 125<br />

Lechner, Jakob, 85<br />

Lee, Jong-Ho, 197<br />

Leiste, Harald, 135<br />

Leong, Daniel, 169<br />

Lettner, Matthias, 129<br />

Lewis, Clyde, 329<br />

333<br />

Li, YongHui, 169<br />

Lichtenberger, Robert, 137<br />

Liersch, Antje, 155<br />

Likhtina, Iya, 239<br />

Limbeck, Andreas, 141, 261<br />

Litschauer, Marco, 299<br />

Livingstone, Richard, 251<br />

Lombardi, John, 227, 241, 251<br />

Lomoschitz, Christoph, 193<br />

Lomoschitz, Marina, 267<br />

Lopez, Veronica, 171<br />

Lumpi, Daniel, 211<br />

Mach, Robert L., 159, 161<br />

Mach-Aigner, Astrid R., 159, 161<br />

Maier, Thomas, 145<br />

Markut-Kohl, Ruth, 209<br />

Marsálek, Roman, 3<br />

Matulova, Pavla, 325<br />

Mecklenbrauker, Christoph F, 105<br />

Melenk, Jens Markus, 283<br />

Menghwar, Gordhan Das, 105<br />

Messner, Kurt, 317<br />

Mihovilovic., Marko., 201<br />

Miksch, Silvia, 91<br />

Mikula, Hannes, 195<br />

Minassian, Karen, 59, 305<br />

Mohsin, Ijaz Ul, 147<br />

Moore, Brandy, 123<br />

Morovati, Mostafa, 289<br />

Moshary, Fred, 277<br />

Mukhtar, Azam, 261<br />

Musliu, Nysret, 81<br />

Naz, Tabbasum, 77<br />

Neck, Volker, 133<br />

Neouze, Marie-Alexandra, 299<br />

Neubauer, Thomas, 79<br />

Nizetic, Ivana, 87<br />

Njau, Philipa, 247<br />

Nolan, Martin, 289<br />

Nosrat, Amir, 289<br />

No<strong>the</strong>gger, Clemens, 97<br />

Novácek, Zdenek, 67, 271


Ocipova, Daniela, 275<br />

Offner, Mathias, 47<br />

Opelt, Aileen, 211<br />

Orság, Petr, 25<br />

Ortag, Felix, 107<br />

Ostermaier, Clemens, 197<br />

Palankovski, Vassil, 221<br />

Panak, Petra J., 131, 133<br />

Papp, Sándor, 309<br />

Paralic, Ján, 15<br />

Pfeifer, Norbert, 97<br />

Pichler, Reinhard, 89<br />

Plank, Lucia, 291<br />

Plha, Stefan, 297<br />

Podlubny, Igor, 75<br />

Pogany, Dionyz, 197<br />

Pokorný, Michal, 19<br />

Pospísil, Jirí, 225<br />

Potzger, Kay, 197<br />

Potzmann, Robert, 139, 199<br />

Povalac, Karel, 3<br />

Powell, Al, 143<br />

Praetorius, Dirk, 257<br />

Prantl, Adrian, 45<br />

Prescod, Andru, 187<br />

Puchberger, Michael, 137, 139, 191, 199<br />

Pucher, Marion E., 159, 161<br />

Pucher, Michael, 321, 323<br />

Puls, Christoph, 141<br />

Pulverer, Gernot, 301<br />

Pus, Viktor, 65<br />

Puschner, Peter, 101<br />

Puskely, Jan, 271<br />

Puxbaum, Hans, 207, 293, 307<br />

Rümmele, Stefan, 89<br />

Rabin, Laura, 123<br />

Rafique, M:Shahid, 213<br />

Raida, Zbynek, 11, 19, 21, 99<br />

Ramalingam, Krish, 313<br />

Raphan, Theodore, 31<br />

Rattay, Frank, 59, 295, 305<br />

Reda, DI Ranja, 303<br />

334<br />

Ren, Bin, 273<br />

Retscher, Guen<strong>the</strong>r, 255<br />

Rieder, Bernhard, 101<br />

Riva, Monica, 73<br />

Rosenthal, Alex, 313<br />

Rovnanik, Pavel, 205<br />

Rupp, Markus, 35<br />

Sabo, Miroslav, 49<br />

Saetta, Anna, 263<br />

Sala, Pavel, 5<br />

Sanchez, Eric, 231<br />

Schaffer, Hannes, 297<br />

Schmidl, Christoph, 307<br />

Schmidt, Moritz, 177<br />

Schranz, Melanie, 61<br />

Schreiner, Dietmar, 113<br />

Schubert, Ulrich, 137, 139, 191, 199, 267<br />

Schwarz, Elisabeth, 219<br />

Schwarz, Josef, 111<br />

Sebo, Dusan, 167<br />

Seeling, Joni, 235<br />

Seidler, Sabine, 209<br />

Seifert, Helmut, 259<br />

Seitner, Florian, 71<br />

Semona, Sanzani, 189<br />

Seo, Dugwon, 143<br />

Shen, Lu, 279<br />

Sigmund, Milan, 5, 7<br />

Simioni, Paola, 263<br />

Simonné-Dombóvári, Eszter, 319<br />

Skerencak, Andrej, 133<br />

Skovranek, Tomas, 75<br />

Sotner, Roman, 17<br />

Sponar, Stephan, 129<br />

Stallinger, Thomas, 13<br />

Stanetty, Peter, 175<br />

Steiger, Matthias G., 159, 161<br />

Steiner, Nicholas, 327<br />

Steppert, Michael, 185<br />

Sterba, Christian, 37<br />

Stueber, Michael, 135<br />

Stumpf, Martin, 21


Stumpf, Thorsten, 177<br />

Suethanuwong, Ekarin, 95<br />

Sun, HuiBin, 169<br />

Székely, Balázs, 309<br />

Sztul, Henry, 237<br />

Talic, Almir, 287<br />

Tarbell, John, 171<br />

Tauer, Johannes, 219, 223<br />

Tedesco, Marco, 327<br />

Teran, Evelyn, 235<br />

Thaler, Gottfried, 265<br />

Tilinger, Jan, 165<br />

Tischner, Alexandra, 145<br />

Tjoa, A Min, 69<br />

Todt, Melanie, 183<br />

Trawniczek, Franz, 223<br />

Triska, Markus, 81<br />

Trumm, Sascha, 131<br />

Tschernutter, Peter, 321, 323<br />

Ulrich, Sven, 135<br />

Vasícek, Zdenek, 55<br />

Vazquez, Maribel, 249<br />

Vega Cañamares, Maria, 241<br />

Vemula, Praveen Kumar, 229<br />

Vitanov, Stanislav, 221<br />

Vitkalov, Sergey, 243<br />

Vranayova, Zuzana, 275<br />

Wal<strong>the</strong>r, Clemens, 177, 185<br />

Wang, Qi, 35<br />

Wassermann, Lubomir, 51<br />

Weber, Robert, 265<br />

Weinbaum, Sheldon, 163<br />

Wenger, Cornelia, 295<br />

Whitlock, Paula, 125<br />

Wilke, Gwen, 311<br />

Wintner, Ernst, 219, 223<br />

Wu, Binlin, 121<br />

Xiao, Jizhong, 31<br />

Zahradník, Petr, 285<br />

335<br />

Zelinka, Petr, 7<br />

Zhang, JingQiao, 243<br />

Zolda, Michael, 117<br />

Zvolensky, Tomas, 11

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!