28.06.2013 Views

t - Computational Optics Group

t - Computational Optics Group

t - Computational Optics Group

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Lecture Notes (1 of 4)<br />

“The Finite Element Time Domain Method<br />

(FETD)<br />

for <strong>Computational</strong> Electrodynamics”<br />

Benedikt Oswald<br />

benedikt.oswald@psi.ch<br />

Phone +41 (0)56 310 32 12<br />

WBGB 131<br />

Paul Scherrer Institut (www.psi.ch)<br />

Large Research Facilities Department<br />

Revisions -2010 March 29, ETH


Passion is required for any great work,<br />

and for the revolution passion and<br />

audacity are required in big doses<br />

Ernesto 'Che' Guevara, Letter to his parents.


Acknowledgements<br />

It is my pleasure to acknowledge various contributions to this lecture:<br />

in alphabetic order<br />

Adelmann, Andreas – PSI group leader<br />

Burri, René – world renowned Swiss (Magnum Photos) photographer<br />

Fomins, Aleksejs – summer student contributing to hades3d<br />

Garvey, Terrence – head of ABK<br />

Hafner, Christian – for inviting me<br />

Leidenberger, Patrick – collaborator and esteemed colleague


Literature References<br />

Jin, Jianming; “The Finite Element Method in Electromagnetics”; 2 nd edition, John Wiley & Sons; 2002. ISBN 0-<br />

471-43818-9. Note: reference work<br />

Volakis, John L., and Chatterjee, Arindam, and Kempel, Leo C.; “Finite Element Method for Electromagnetics –<br />

Antennas, Microwave Circuits, and Scattering Applications”; Wiley Interscience & IEEE Press; 1998. ISBN 0-7803-<br />

3425-6. Note: reference work<br />

Novotny, Lukas and Hecht, Bert; “Principles of Nano-<strong>Optics</strong>”; Cambridge University Press; 2006. ISBN 13-978-<br />

0-521-83224-3. Note: motivates applications<br />

Taflove, Allen and Hagness, Susan, C.; “<strong>Computational</strong> Electrodynamics – The Finite Difference Time Domain<br />

Method”, 3 rd edition; Artech House; 2005; ISBN 1-58053-832-0. Note: general inspiration for time domain methods<br />

Zienkiewicz, O. C. and Taylor, R. L.; “The Finite Element Method – Volume 1: The Basics”; 5 th edition;<br />

Butterworth-Heinemann; 2000; ISBN 0 7506 5049 4. Note: fundamentals on finite element methods by its<br />

pioneers in engineering sciences.<br />

And my sincerest apology for all forgotten references, it was not by intention!


Overview<br />

Introduction, Motivation and Governing Equation<br />

Finite Element Method as a Volume Approach<br />

Practical Realization of the Algorithms<br />

Case Studies and Modeling Accuracy


Introduction, Motivation and Governing Equations (1)<br />

Maxwell's equations Constitutive relations<br />

∇× H =j ∂ D<br />

∂ t<br />

∂ B<br />

∇×E=−<br />

∂t<br />

∇⋅D= el<br />

∇⋅B=0<br />

D=⊗E<br />

D=⋅E<br />

j=⋅E<br />

B= 0 ⋅H<br />

Dispersive, scalar<br />

Non-dispersive, scalar<br />

Ohm's law<br />

James Clerk Maxwell<br />

Source wikivisual


Introduction, Motivation and Governing Equations (2)<br />

E lectric field vector wave equation<br />

aka. curl-curl equation:<br />

general dielectric dispersive formulation<br />

∇× 1<br />

<br />

∇×E ∂<br />

∂ t<br />

E<br />

∂2<br />

∂<br />

∗E =− 2<br />

∂t ∂ t J 0<br />

S ubject to boundary conditions and constitutive relations<br />

E=E x , t x ∈ℝ with and 3


Introduction, Motivation and Governing Equations (3)<br />

Why solve Maxwell's equations in the time domain ?<br />

Model transient processes<br />

Model non-linear processes: Raman scattering in near-field optics<br />

Model processes over large bandwidths in one single run<br />

Fields of applications ?<br />

Nano-optics, scanning near-field optical microscope probes, optical antennas<br />

Devices based on plasmonic resonances<br />

Laser assisted field emission for advanced photo-cathodes in accelerators (SwissFEL)<br />

Also, traditional microwave electromagnetics


Introduction, Motivation and Governing Equations (4)<br />

Dispersive Dielectric Material – Debye Model<br />

used for:<br />

alcohols<br />

Water<br />

water in soils<br />

∞<br />

= s − ∞<br />

<br />

t= ∞ t <br />

<br />

<br />

= ∞<br />

1 j <br />

relative permittivity at infnite frequency (often 1.0)<br />

change in relative permittivity due to pole<br />

pole relaxation time<br />

e− t<br />

⋅U t<br />

Source: wikipedia<br />

Peter Joseph<br />

William<br />

Debye;Dutch<br />

physicist and<br />

physical chemist;<br />

1884-1966.<br />

Debye model in time domain<br />

U t = unit step function


Introduction, Motivation and Governing Equations (5)<br />

Dispersive Dielectric Material – Drude Model Source: wikipedia<br />

used for:<br />

Plasmas<br />

Metals in the visible region<br />

p<br />

<br />

2<br />

p = − ∞<br />

2 − j <br />

∞ relative permittivity at infnite frequency (often 1.0)<br />

t= ∞ t 2<br />

p<br />

p<br />

pole angular frequency<br />

angular relaxation frequency<br />

Paul Karl Ludwig<br />

Drude; German<br />

Physicist; 1863-<br />

1906.<br />

1−e − p t ⋅U t Drude model in time domain


Introduction, Motivation and Governing Equations (6)<br />

Dispersive Dielectric Material – Lorentz Model<br />

used for:<br />

Solid in the visible<br />

Metals in the visible<br />

∞<br />

= s − ∞<br />

p<br />

p<br />

= ∞ <br />

pole angular frequency<br />

t= ∞ t 2<br />

p<br />

2<br />

p 2<br />

2 j p p − 2<br />

relative permittivity at infnite frequency (often 1.0)<br />

change in relative permittivity due to pole<br />

damping coeffcient associated with pole<br />

p<br />

2 − p<br />

Source: wikipedia<br />

Hendrik Antoon<br />

Lorentz; Dutch<br />

Physicist; 1853-<br />

1928.<br />

2 ⋅e− p t 2 2<br />

⋅sin p−<br />

pt<br />

⋅U t <br />

Lorentz model in time domain


Introduction, Motivation and Governing Equations (7)<br />

Dispersive Dielectric Material – Drude-Lorentz Model – especially for metals in the visible<br />

= ∞ −<br />

2<br />

p 2 − j <br />

Vial et al., Physical Review B, Vol 71, pp. 085416, 2006:<br />

Gold dielectric data modeled by Drude & Drude-Lorentz<br />

2<br />

p<br />

2<br />

2 j − p<br />

p 2<br />

source ditto


Introduction, Motivation and Governing Equations (8)<br />

Why finite elements ? Rationale ? Concept ?<br />

In order to numerically solve<br />

ordinary differential equations (ODE)<br />

partial differential questions (PDE)<br />

Integral equations (IE)<br />

defined on multi-dimensional, geometrically complicated domains, we<br />

subdivide the complicated domain into elements of simple shape<br />

approximate the solution via functions of know shape but unknown coefficients, aka. Degrees of freedom<br />

formulate systems of equations to calculate the DoF


Introduction, Motivation and Governing Equation (9)<br />

Appeal of the finite element method – level of detail (LoD)<br />

The figure represents an S EM comparison of<br />

the emitter shape from a double-oxidation<br />

mould with the emitter shape from a single<br />

oxidation mould. Fig a) is a 45 degree image of<br />

the emitter from the double oxidation mould,<br />

showing the rounded joins between pyramid<br />

faces. Fig. b) shows the blunted tip of this<br />

emitter from above. Fig c) and d) are the<br />

equivalent views for the emitter from the single<br />

oxidation mould, which has a very sharp tip<br />

(not more than 5 nm - this is about the limit of<br />

the S E M resolution for looking at tips) and it<br />

also has undesirable, sharp ridges where the<br />

pyramid faces meet.<br />

From: Fabrication of all-metal field<br />

emitter arrays with controlled apex sizes<br />

by molding, E. Kirk, S. Tsujino, T. Vogel,<br />

K. Jefimovs, J. Gobrecht and A. Wrulich.<br />

Journal of Vacuum Science and<br />

Technology B, 27(4), 2009.<br />

c)<br />

a)<br />

200 nm<br />

20 nm<br />

1μm 20 nm<br />

b)


Introduction, Motivation and Governing Equation (10)<br />

mesh – element types


Introduction, Motivation and Governing Equations (11)<br />

Ultra-short History of the Finite Element Method in Electromagnetics<br />

...admittedly somewhat arbitrary...<br />

Zoltan Cendes Boris Galerkin<br />

1870 - Rayleigh variational methods<br />

1909 - Ritz ditto<br />

1915 - Galerkin weighted residuals<br />

1943 - Courant piecewise continuous trial functions<br />

1947 - Prager-Synge ditto<br />

1964 - Zienkiewicz ditto<br />

Source CNN<br />

1967 - Zienkiewicz et al. solution of three-dimensional field problems<br />

1970 - Silvester et al. non-linear magnetic field analysis<br />

1980ies - many authors electrodynamics problems, problems with spurious modes<br />

1990ies - Cendes, Webb, edge element to avoid spurious modes<br />

Lee, Nédelec et al. finite element time domain methods<br />

Source wikipedia<br />

Source times<br />

Olek Zienkiewicz<br />

2000nds many authors PML, ABC, boundary integrals, dispersive media in time domain


el anejo dieléctrico (1)<br />

0<br />

0<br />

derivation of the Drude operator L<br />

we need to evaluate: 0<br />

2<br />

p p<br />

2<br />

p p<br />

∂ 2<br />

∂ t 2 u t⊗E = 0<br />

2<br />

p p<br />

2<br />

p p<br />

∂ 2<br />

∂ t 2 −e− p t u t⊗E =− 0<br />

∂ 2<br />

∂ t 2 1−e− p t u t ⊗E<br />

∂<br />

∂ t t ⊗E = 0<br />

− 0<br />

we use: e − p t t⊗E=∫ −∞<br />

thus<br />

0<br />

2<br />

p p<br />

∞<br />

2<br />

p<br />

p<br />

2<br />

p<br />

p<br />

∂<br />

∂ t<br />

2<br />

p<br />

p<br />

∂ E<br />

∂ t<br />

[ ∂<br />

∂ t e− p t ut ⊗E ]<br />

1 st part<br />

2 nd part<br />

∂<br />

∂t [− p e− p t u te − p t t ⊗E]<br />

e − pt ' t ' E t−t ' dt '=E t<br />

∂ 2<br />

∂ t 2 −e− p t 2 ∂<br />

u t⊗E =0 p<br />

∂ t e− p t ut ⊗E −0 2<br />

p<br />

p<br />

∂ E<br />

∂ t


el anejo dieléctrico (2)<br />

we combine the 1 st and 2 nd part<br />

0<br />

0<br />

2<br />

p p<br />

2<br />

p p<br />

∂ 2<br />

∂ t 2 1−e− p t 2 ∂<br />

u t ⊗E= 0 p<br />

∂t e− p t u t⊗E<br />

∂ 2<br />

∂ t 2 1−e− p t 2<br />

u t ⊗E= 0 p[−<br />

p e − p t u te −p t t]⊗E<br />

using the identity from the preceeding slide again, eventually we obtain<br />

0<br />

2<br />

p p<br />

thus<br />

∂ 2<br />

∂ t 2 1−e− p t 2<br />

u t ⊗E= 0 p[<br />

E− p e − p t u t⊗E ]<br />

L Drude = 0<br />

2<br />

p<br />

p<br />

∂ 2<br />

∂t 2 1−e− p t 2<br />

u t⊗E =0 p[<br />

E− p e − p t u t ⊗ E ]


End of 1 st Part<br />

The revolution is not an apple that falls when it is ripe. You have to make it fall (Che)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!