LMB Cambridge - Helsinki Institute of Physics

Impact of silicon pixel detectors on 

structural biology 

Wasi Faruqi 

MRC Laboratory of Molecular Biology, 

Hills Road, 

Cambridge CB2 0QH, 

UK 

IWORID 10 Helsinki, 1 st July, 2008 

6/26/2008 

LMB, Cambridge

Kings College Chapel, Cambridge 

6/26/2008 


MRC Laboratory of Molecular Biology, 

Cambridge 

6/26/2008 


Direct Detection in Silicon Pixel 

• Hybrid Pixel Detectors 

Medipix2 

Detectors 

http://medipix.web.cern.ch/MEDIPIX/ 

• CMOS Detectors 

Monolithic Active Pixel Sensors (MAPS) designed at the 

STFC Rutherford Lab. 

Pixellated silicon, readout built into each pixel. 

http://mi3.shef.ac.uk/ 

6/26/2008 


Acknowledgements 

LMB Cambridge 

•Richard Henderson 

•Greg McMullan 

•Shaoxia Chen 

CERN (Medipix2) 

•Lukas Tlustos, 

•Xavi Llopart, 

•M.Campbell 

•RAL-STFC (MAPS) 

•R.Turchetta 

•M. Prydderch, 

•MI3 Collaboration (RCUK) 

6/26/2008 


Main techniques used in structure 

determination 

X-ray (Protein) Crystallography 

Electron Cryo-Microscopy 

Nuclear Magnetic Resonance 

(along with other supporting technologies) 

6/26/2008 


Main scientific aims 

•Structural analysis of proteins and 

macromolecular complexes to atomic or near 

atomic resolution 

•Electron Cryo-Microscopy frequently used 

as a complementary technique to X-ray 

Crystallography 

•Example: Hepatitis B virus solved by both 

techniques – Example in later slide 

6/26/2008 


Electron Detectors & X-ray Detectors 

Electron Detection: What are the main differences 

between X-ray Detectors and Electron Detectors? 

Electrons 100 – 300 keV 

X-rays 10 – 20 keV 

Electrons very easily scattered and stopped by matter, 

So, need to: 

(a) install detector in a vacuum chamber, and 

(b) entrance window not possible on the detector 

Cannot use gas detectors 

6/26/2008 


Schematic of Microscope 

LMB-Cambridge 

6/26/2008 


Scientific Background to Electron Cryo-Microscopy 

Three Main types of Analysis (and the resolution attained 

in the analysed structures): 

1. Single Particle (molecule) Analysis 3.8 -10 Å 

2. Electron Crystallography of ordered specimen, i.e. 2-D 

crystals ~3Å … near-atomic resolution 

3. Electron Tomography 20-100 Å … cell biology 

Electron energy preferred: 300 keV (less lens 

aberrations, less multiple scattering in sample, less 

absorption) 

6/26/2008 


Electron Cryo-Microscopy 

•Image molecules in native aqueous 

environment in vitreous ice (prevent de-hydration) 

•Trap important conformations in intermediate states of a 

kinetic cycle by rapid freezing : equivalent to time-resolved 

measurements 

•Images have low contrast : need sophisticated software and 

lots of averaging 

•Radiation damage to specimen a severe limitation – hence 

need very good detectors! 

6/26/2008 


Single Particle Analysis 

No crystals required .. Makes it possible, e.g. to analyse 

structure of membrane proteins, which are very difficult to 

crystalise, 

Can be applied to large macromolecular complexes 

• Powerful technique when used in conjunction with atomic 

structures obtained with x-ray crystallography 

•Applications: Virus particles, ribosomes, etc 

•‘Best’ Resolution with this technique :Rotavirus capsid 3.8 Ǻ 

•6.5 Million copies of the capsid protein used for the map 

•From 8400 virus particles (excellent symmetry) 

6/26/2008 


Rotavirus Epidemiology 

Deaths due 

to rotavirus 

(diarrhea) 

6/26/2008 

Causes ~600,000 deaths annually 

Infection occurs from feces. Only 10 to 100 infectious particles suffice. 


Symmetry: 

60 x 13 

=780 VP6 molecules 

6/26/2008 

LMB, Cambridge 

Zhang et al. (2008)

Side Chains 

3.8 Å 

X-ray crystallography (2F o 

–F c 

) 

Cryo-EM 

6/26/2008 

Settembre et al. 


Xing Zhang et al. (2008)

Electron Crystallography 

• Averaging done in crystal… many identical scattering 

particles 

• Main application: membrane proteins (but not 

exclusively) 

• Near-atomic resolution (~2.5 Å) achieved. 

6/26/2008 


Bacteriorhodopsin 

•Membrane protein family 

•Light driven proton pump; photocycle 

•Structural changes at atomic resolution studied 

by trapping intermediates in reaction cycle 

•Diffraction data collected with LMB CCD 

camera 

6/26/2008 


Electron Diffraction Studies on Bacteriorhodopsin 

(with R.Henderson and S. Subramaniam) 

6/26/2008 


A 7-second diffraction pattern from bacteriorhodopsin with spots visible to 2Å -1

BR Model 

6/26/2008 

LMB, Cambridge Henderson, et al JMB, 213, 899-929 (1990)

BR 

6/26/2008 


Courtesy R.Henderson

Electron Tomography 

1. View specimen from large number of angles, 

but limits due to radiation damage 

2. Combine views into 3-D reconstruction 

3. Image needs re-focusing and re-aligning after 

each step due to imperfections in stage construction 

4. Automation and electronic detectors essential! 

6/26/2008 


Reconstruction 

Tomography - Reconstruction of a 3D Model 

Based on a Series of Projections 

Projections 

Reconstruction 

6/26/2008 


A. Koster et al. 1997

Investigation into Mitotic Proteins using 

Electron Tomography 

John Kilmartin and Sam Li (LMB) 

Identify proteins common to yeast and 

mammalian cells using mass spectrometry. 

Explore function with biochemical techniques and 

study structure with electron tomography. 

6/26/2008 


6/26/2008 


Recent structures obtained using 

Cryo-EM on next slide 

a. 70S E.coli ribosome complexed with mRNA and fMettRNA 

(11.5 Ǻ) 

b. Hepatitis B virus (7.4 Ǻ) 

c. Actin filaments decorated with myosin heads (30-35 

Ǻ) 

d. 2D crystal Light Harvesting Complex II (3.4 Ǻ) 

Baker & Henderson‘Electron cryomicroscopy’ 

International Tables for Crystallography, 451-479,(2002) 

6/26/2008 


Molecular structures obtained by Electron Cryo-Microscopy Magnification 

Micrographs:170K, Models: 1.2 Million 

Montage 

70S Ribosome 

6/26/2008 

Hepatitis B Virus Actin-Myosin (Muscle) Light Harvesting Comple 


Hepatitis B Structure 

Cryo-EM X-ray Crystallography 

Andrew Leslie LMB 

6/26/2008 


Scientific Background to Electron Cryo-Microscopy 

Three Main types of Analysis (and the resolution attained 

in the analysed structures): 

1. Single Particle (molecule) Analysis 4 -10 Å 

2. Electron Crystallography of ordered specimen, i.e. 2-D 

crystals ~3Å … near-atomic resolution 

3. Electron Tomography 20-100 Å … cell biology 

Electron energy preferred: 300 keV (less lens aberrations, 

less multiple scattering in sample, less absorption) 

6/26/2008 


High Resolution Imaging Detector 

Requirements for Cryo-EM 

1. Electronic detector with computer control.. eliminate film! 

2. Number of independent pixels : 4000 by 4000 

3. Pixel Size 10 – 50 µm (has to fit in commercial microscopes) 

4. High sensitivity with no noise – ability to add multiple frames 

5. Radiation damage; should be able to withstand 

at least 1 MRad this would be ~ a one year dose for cryo-EM 

6. Readout time preferably short 

6/26/2008 


Medipx2(Quad) in 300 kV Microscope 

Mounting 

6/26/2008 


300 kV EM with detector installed 

6/26/2008 


Medipix Quad

Detectors: Quality Factors 

Sensitivity: Detective Quantum Efficiency (DQE), 

(S/N) 2 output /(S/N)2 input (=1 for perfect detector) 

DQE(0) zero spatial frequency 

DQE(spatial frequency) 

Resolution: Modulation Transfer Function (MTF) 

Framing Speeds… inverse of readout time 

Radiation Hardness … useful lifetime 

Dynamic Range … ability to record very weak and very 

strong parts of an image simultaneously (diffraction only) 

Defects …… Faults in fabrication, etc 

6/26/2008 


Monte Carlo simulation of electron trajectories in silicon. 

Detector thickness = 300 microns, pixel=55 microns 

Extension of simulations to include energy deposition 

6/26/2008 


McMullan, et al Ultramicroscopy, 

107,(2007), 401-413

MTF at Nyquist Frequency 

6/26/2008 


McMullan, et al 

Ultramicroscopy, 107, (2007), 

401-413

Single Electron Clusters; 240 incident electrons at 120 keV 

250 

Counts/electron vs Ext Vthl 

High Threshold 

Low Threshold 

200 

No counts 

1 count 

Counts/electron 

150 

100 

2 counts 

50 

0 

1190 1240 1290 1340 1390 1440 1490 

Ext Vthl (1190 __ 120 keV) 

3 

4 

counts 

6/26/2008 


Increased probability of several pixels counting an electron at 

lower thresholds. Seed pixel in centre of array. E=120 keV 

(a) (b) (c) (d) (e) 

(f) (g) (h) (i) (j) 

6/26/2008 

(k) (l) (m) (n) (o) 


DQE(0) and DQE(Nyquist) 

DQE 

6/26/2008 


Threshold(keV)

Some examples of Single Particle 

Imaging using the Medipix2_Quad 

Detector 

All images recorded at 120 keV 

McMullan & Faruqi 

Nucl. Instr. and Meth. A 591 (2008) 129–133 

6/26/2008 


Rotavirus with 1.6(left) and 160(right) electrons/pixel 

0.04 electron/Å 2 (at specimen) 4 electrons/Å 2 

Rotavirus imaged with Medipix2 Quad 

6/26/2008 


Single Lambda Phage on MPX2 Quad 

6/26/2008 


TMV: Results of image alignment 

Negatively stained TMV 

protein stacked disks: 

(a) first image in a series of 

65, 

(b) simple sum of the 65 

images (blurred), 

(c) aligned sum of the 

images (sharp), 

(d) sample movement in Å. 

The scale bar in (b) 

indicates 500 Å 

6/26/2008 


Tobacco Mosaic Virus (from 65 images) 

Magn. =180K 

6/26/2008 


MAPS CMOS Detector 

- charged particles 

• no bias voltages 

• charge diffusion 

• 100% fill factor 

Epilayer 

Substrate 

6/26/2008 

Turchetta et al 

NIM A458 (2001) 677-689 


Monolithic Active Pixel Sensor (MAPS) 

General Background 

Monolithic Active Pixel Sensor (MAPS ) –designed at 

RAL* 

Size: 525 by 525, 25 μm square pixels 

Non-Radhard, standard 0.5 μm CMOS technology 

Each pixel contains four diodes 

Electrons drift to one of the four diodes in pixel 

Charge summed from all diodes and converted to a voltage 

One ADC per column; all pixels in a row read out in 

parallel 

*Prydderch, et al Nucl.Instr. & Meth. A512,(2003),358 

6/26/2008 


Monte Carlo simulation of electron trajectories in silicon. 

Detector thickness = 300 microns, pixel=55 microns 

Extension of simulations to include energy deposition 

6/26/2008 


McMullan, et al Ultramicroscopy, 

107,(2007), 401-413

MAPS in 300 keV Mounting 

6/26/2008 


MI-3 Consortium Members 

Professor NM Allinson, University of Sheffield 

Professor PP Allport, University of Liverpool 

Dr RL Bates, University of Glasgow 

Professor AR Cossins, University of Liverpool 

Professor MM El Gomati, University of York 

Dr AR Faruqi, MRC Laboratory of Molecular Biology 

Dr MJ French, CCLRC 

Professor A Holland, Open University 

Professor RJ Ott, Institute of Cancer Research 

Dr V O'Shea, University of Glasgow 

Professor RD Speller, University College London 

Dr R Turchetta, CCLRC 

Dr K Wells, University of Surrey 

6/26/2008 


Imaging of 100 mesh grid in MAPS 

6/26/2008 


ADC Response for Single Electrons 

at 40 keV and 120 keV 

6/26/2008 


Sensitivity: 

Noise: 

MAPS Summary at 120 keV 

~50 ADC Units/electron 

~2 ADC Units 

Signal/Noise: 20-25 

Radiation Hardness: 

Active area 

10-15 kRad . Needs improvement! 

525 x 525 pixels need larger areas 

Faruqi, Henderson, Turchetta et al Nucl. Instr.& Meth 546, 

170-175, (2005) 

6/26/2008 


Radhard 

Technology 

512 x 512, 25 µm 

Radiation Damage to STAR250 

(FillFactory/Cypress Corp.) at 300 keV 

C 

Radiation Dose: 

A: 200kRad 

(annealed for 4 

weeks) 

B: 200 kRad 

C: 1000 kRad 

Contrast values 

labelled in bottom 

left image 

B 

B 

C 

A 

A 

6/26/2008 


Startracker & Film response to electrons 10 - 300 keV 

50 

Response - Film values normalised 

40 

30 

20 

10 

0 

0 50 100 150 200 250 300 

Electron energy (keV) 

Startracker 

Film 

6/26/2008 


Large Area Sensor : Specifications 

Pixel Size 40µm 

Array Size 1.4K by 1.4K 

Array Dimensions 5.6cm x 5.6cm 

Epi Thickness 15µm 

Frame Rate 

10 f/s max 

Noise floor 

28e 

Radiation tolerance ~

Large Area Sensor Wafer 

Courtesy of 

Andy Clark & 

Renato Turchetta 

MI3 

Collaboration 

6/26/2008 


Large Area Sensor - Mounted 

Courtesy of 

Andy Clark & 

Renato Turchetta 

MI3 

Collaboration 

6/26/2008 


6/26/2008 

Detectors for Protein Crystallography : Main 

Requirements 

High Efficiency, low noise (high DQE) 

Large active area; can get better S/N by increasing distance 

Excellent spatial resolution; need to resolve ~500 orders (3K x3K) 

Very large dynamic range (strong & weak spots) 

High rate capability (no dead time, shutterless operation) 

No spatial distortions or non-uniformity of response; any corrections should 

be stable over long periods 

Ability to operate at a wide range of wavelengths for MAD, 0.6Ǻ –2.5Ǻ 

Low cost, reliable (low maintenance) 

Continuous readout to eliminate beam shutter (closed during readout) 

PILATUS 1M and 6M (PSI, Villigen) - talks during conference 

Medipix3: Does it offer special advantages for Crystallography? 


Medipix3 

Silicon sensor, 300 or 500 µm for very high efficiency 

1 or 2 counters per pixel – continuous R/W with no dead time 

Pixel Size : 55 µm square, very well suited for microcrystallography. 

Can be re-arranged to be 110 µm for medical work 

with higher energy x-rays 

Pixel level intelligence – expect improved resolution (by reducing 

effects of charge sharing between pixels) 

Counting rates ~ 10 6 counts/second/pixel 

Chips to be 4-side buttable (in future) for extended tiling 

For more details, see Michael Campbell during IWORID10 or visit: 

http://medipix.web.cern.ch/MEDIPIX/

Medipix3 – charge summing concept 

The winner takes 

all 

• The incoming 

quantum is assigned as 

a single hit 

• Charge processed is 

summed in every 4 

pixel cluster on an 

event-by-event basis 

55μ 

Medipix 

Collaboration 

6/26/2008 


55μ 

4 

5 

6 7 

DIGITAL CIRCUITRY 

4. Control logic 

(124) 

5. 2x15bit counters 

/ shift registers 

(480) 

6. Configuration 

latches (152) 

7. Arbitration 

circuits (100) 

Total digital 856 

2 

1 3 

17 October 2006 Michael Campbell 

55μ 

ANALOG CIRCUITRY 

1. Preamplifier (24) 

2. Shaper (134) 

3. Discriminators 

and Threshold 

Adjustment 

Circuits (72) 

Total analog 230

The Medipix3 Consortium 

• University of Canterbury, Christchurch, New Zealand 

• CEA, Paris, France 

• CERN, Geneva, Switzerland, 

• DESY-Hamburg, Germany 

• Albert-Ludwigs 

Ludwigs-Universität Freiburg, , Germany, 

• University of Glasgow, Scotland, UK 

• Leiden Univ., The Netherlands 

• NIKHEF, Amsterdam, The Netherlands 

• Laboratory of Molecular Biology, Cambridge, England, UK 

• Mid Sweden University, Sundsvall, , Sweden 

• Czech Technical University, Prague, Czech Republic 

• ESRF, Grenoble, , France 

• Universität Erlangen-Nurnberg 

Nurnberg, Erlangen, , Germany 

• University of California, Berkeley, USA 

• VTT, Information Technology, Espoo, , Finland 

• ISS, Forschungszentrum Karlsruhe, Germany 

• Diamond Light Source, Oxfordshire, UK 

6/26/2008 


Microcrystallography – A new application? 

Useful technique for very small crystals, ~20 µm 3 

(large crystals more difficult to grow in many cases) 

ESRF Microfocus Beam ID13, Focused beam size ~1 µm 

Energy=13 keV, 

Flux = 3x10 10 photons/sec/µm 2 

XylanaseII structure determined to 1.5 Ǻ, diffraction pattern 

next slide. 

MW=21kDa, 40Ǻ x 39Ǻ X 57Ǻ 

Detector needs to have small pixels, high DQE 

Riekel, Schertler… Current Opinion Strc Biol (2005), 15, 556-562 

6/26/2008 


Micro-diffraction from Xylanase II 

Data collected on 

MarCCD165 

Pixel size ~79 µm 

square 

Spot size limited by 

detector resolution and 

not by beam or crystal 

6/26/2008 


Summary: X-ray Detectors 

Potential Advantages 

Direct detection in Silicon – better resolution as no light scattering, 

even better resolution with Medipix3 

Fast readout – fast framing (few milliseconds) possible 

Excellent S/N – noiseless readout, high dynamic range 

Detector and electronics separate – choose detector material (Si, 

GaAs, CdTe, …) for optimum efficiency 

See also: PILATUS contribution to this conference 

Downside 

Large area detectors difficult and expensive to build 

Individual detectors, ~2 cm 2 , tiled to obtain larger areas but gaps in 

between chips/modules leads to some dead space 

Technology not yet mature (?) – problems of ‘yield’ 

6/26/2008 


Summary: Detectors for Electron 

Microscopy 

• Medipix2 is a superb detector up to 120 keV 

• But, it may prove expensive to design 4K square arrays 

without dead spaces 

• Higher energies (300 keV) may be feasible but with higher 

density compounds, e.g. Cd(Zn)Te 

• CMOS detectors offer a good chance of a radiation hard, 

4Kx4K square detector – but needs a lot more R&D effort 

6/26/2008 


Acknowledgements 

LMB Cambridge 

•Richard Henderson 

•Greg McMullan 

•David Cattermole 

•Shaoxia Chen 

CERN (Medipix2) 

•Lukas Tlustos, 

•Xavi Llopart, 

•M.Campbell 

http://medipix.web.cern.ch/MEDIPIX/ 

•RAL-STFC (MAPS) 

•R.Turchetta, et al 

•M. Prydderch, et al 

•MI3 Collaboration (RCUK) 

•http://mi3.shef.ac.uk/ 

6/26/2008

LMB Cambridge - Helsinki Institute of Physics

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

Delete template?

Save as template?