Computational Photonics:

Frequency and Time Domain Methods

Steven G. Johnson
MIT Applied Mathematics
 Nano-photonic media (l-scale)
strange waveguides

& microcavities
[B. Norris, UMN] [Assefa & Kolodziejski,
MIT]
3d
structures

[Mangan,
Corning]

synthetic materials
optical phenomena
hollow-core fibers
 Photonic Crystals
periodic electromagnetic media
1887 1987
1-D 2-D 3-D

periodic in periodic in periodic in

one direction two directions three directions

can have a band gap: optical “insulators”

 Electronic and Photonic Crystals
atoms in diamond structure dielectric spheres, diamond lattice
Medium
Periodic

photon frequency
Band Diagram
Bloch waves:

electron energy

wavevector wavevector
 Electronic & Photonic Modeling
Electronic Photonic
• strongly interacting • non-interacting (or weakly),
—entanglement, Coulomb —simple approximations
—tricky approximations (finite resolution)
—any desired accuracy

• lengthscale dependent • scale-invariant

(from Planck’s h) —e.g. size ¥10 fi l ¥10
(except materials may change)
Computational Photonics Problems
• Time-domain simulation
— start with current J(x,t)
— run “numerical experiment” to simulate E(x, t), H(x, t)

• Frequency-domain linear response

— start with harmonic current J(x, t) = e–iwt J(x)
— solve for steady-state harmonic fields E(x), H(x)
— involves solving linear equation Ax=b

• Frequency-domain eigensolver
— solve for source-free harmonic eigenfields
E(x), H(x) ~ e–iwt
— involves solving eigenequation Ax=w2x
 Numerical Methods: Basis Choices
finite difference finite elements
in irregular “elements,”
discretize approximate unknowns
unknowns by low-degree polynomial
on regular grid

df f ( x + Dx) - f ( x - Dx)
ª + O(Dx 2 )
dx Dx
boundary-element methods
† discretize only the
spectral methods
boundaries between
complete basis of homogeneous media
+ smooth functions
(e.g. Fourier series)
…solve
+ integral equation
+ ….. via Green’s functions
 Numerical Methods: Basis Choices
finite difference finite elements
in irregular “elements,”
discretize approximate unknowns
unknowns by low-degree polynomial
on regular grid

boundary-element methods
spectral methods
discretize only the
boundaries between
complete basis of homogeneous media
+ smooth functions
(e.g. Fourier series)
+ …solve
integral equation
+ …..
via Green’s functions

Much easier to analyze, implement, Potentially much more efficient,

generalize, parallelize, optimize, … especially for high resolution
Computational Photonics Problems Numerical Methods: Basis Choices

finite difference
• Time-domain simulation
— start with current J(x,t)
— run “numerical experiment” to spectral methods
simulate E(x, t), H(x, t)
df f ( x + Dx) - f ( x - Dx)
dx
ª
Dx
+ O(Dx 2 ) +
• Frequency-domain linear response
— start with harmonic current J(x, t) = e–iwt J(x) +
†
— solve for steady-state harmonic fields E(x), H(x) + …..
— involves solving linear equation Ax=b
finite elements
• Frequency-domain eigensolver in irregular “elements,”
approximate unknowns
— solve for source-free harmonic eigenfields
by low-degree polynomial
E(x), H(x) ~ e–iwt
— involves solving eigenequation Ax=w2x

boundary-element methods
discretize only the
boundaries between
homogeneous media
 FDTD
Finite-Difference Time-Domain methods

Divide both space and time into discrete grids

— spatial resolution ∆x
— temporal resolution ∆t

Very general: arbitrary geometries, materials,

nonlinearities, dispersion, sources, …
— any photonics calculation, in principle

∂H 1 r ∂E 1 r J
= - —¥ E = —¥ H -
∂t m ∂t e e

dielectric function e(x) = n2(x)

†
 The Yee Discretization (1966)
a cubic “voxel”: ∆x ¥ ∆y ¥ ∆z (i, j+1)
(i+1, j+1, k+1)

Hz
Ey
(i, j, k+1) Hz

Hx
Ez Hy (i, j) (i+1, j)
(i+1, j+1, k)
Ex
Ey
(i, j, k) (i+1, j, k)
Ex

Staggered grid in space:

— every field component is stored on a different grid
 The Yee Discretization (1966)
(i, j+1)
∂H 1
= - — ¥ E fi L
∂t m Ey
Hz
∂H z 1 Ê ∂Ey ∂Ex ˆ
=- Á - ˜
∂t 1
i+ , j +
1 m Ë ∂x ∂y ¯ (i, j) (i+1, j)
2 2 Ex
Ê 1 1 1 1 ˆ
E (i + 1, j + ) - Ey (i, j + ) Ex (i + , j + 1) - Ex (i + , j) ˜
1Á y 2 2 - 2 2
ª- Á ˜
mÁ Dx Dy ˜
Ë ¯
+ O(∆x2) + O(∆y2)
all derivatives become center differences…
 The Yee Discretization (1966)
all derivatives become center differences…
including derivatives in time
1 1
H(n + ) - H(n - )
∂H 1 2 2
= - —¥ E ª
∂t t= nDt m t= nDt
Dt
+ O(∆t2)
Explicit time-stepping:
Dx
stability requires Dt <
# dimensions
(vs. implicit time steps: invert large matrix at each step)
 FDTD Discretization Upshot

• For stability, space and time resolutions are proportional

— doubling resolution in 3d requires
at least 16 = 24 times the work!

• But at least the error goes quadratically with resolution

…right?
…not necessarily!
 Difficulty with a grid:
representing discontinuous materials?

“staircasing”

… how does this affect accuracy?

Field Discontinuity Degrades
Order of Accuracy
E
TE polarization (E in plane: discontinuous)
a

a
Sub-pixel smoothing
Can eliminate
discontinuity
by “grayscaling”
— assign some average e
to each pixel

e?
= discretizing a smoothed structure
— that means we are changing geometry
— can actually add to error
Past sub-pixel smoothing methods
can make error worse!
& convergence is
Three previous smoothing methods
still only linear

[ Dey, 1999 ]
[ Kaneda, 1997 ]
[ Mohammadi, 2005]
 A Criterion for Accurate Smoothing
1st-order errors È 2 1 2˘
from ~ Ú Í De E|| - D( ) D^ ˙
smoothing De Î e ˚

We want the smoothing errors to be zero to 1st order

— minimizes error and 2nd-order convergent!
† Use a tensor e: Ê
e ˆ
Á ˜ E||
(in principal axes:)
Á e ˜
Á -1 ˜
[ Meade et al., 1993 ] Ë e-1 ¯ E^
Consistently the Lowest Error
a

quadratic accuracy! quadratic!

[ Farjadpour et al., Opt. Lett. 2006 ]

… & in 3d too

(notice that
ranking of
other methods
has shuffled!)
A qualitatively different case: corners
still ~lowest error, but not quadratic

zero-perturbation
criterion
not satisfied
due to E divergence
at corner
— analytically,
error ~ ∆x1.404
Yes, but what can you do with FDTD?
Some common tasks:

• Frequency-domain response:
— put in harmonic source and wait for steady-state

• Transmission/reflection spectra:
— get entire spectrum from a single simulation
(Fourier transform of impulse response)

• Eigenmodes and resonant modes:

— get all modes from a single simulation
(some tricky signal processing)
 Transmission Spectra in FDTD
e=1

a e = 12

example: a 90° bend, 2d strip waveguide

transmitted power = energy flux here:

 Transmission Spectra in FDTD
e=1

a e = 12

Gaussian-pulse
current source J
Fourier-transform the fields at each x:
-iwt -iwnDt
Ew = Ú E(t)e dt ª Â E(nDt)e Dt
n
1 *
P(w ) =
2
Ú Re[Ew ¥ H w ]dx
†
 Transmission Spectra in FDTD
must always do two simulations: one for normalization

electric P0(w)
field Ez:

P(w)
transmission = P(w) / P0(w)
 Reflection Spectra in FDTD
e=1

e = 12

1
PR (w ) = Ú Re[(Ew - Ew0 )* ¥ (H w - H w0 )]dx
2

for reflection, subtract incident fields

(from normalization run)

1 *
P(w ) =
2
Ú Re[Ew ¥ H w ]dx
 Transmission/Reflection Spectra
e=1

a e = 12

T 1–T–R

wa / 2πc = a / l
 Dimensionless Units

Maxwell’s equations are scale invariant

— most useful quantities are dimensionless ratios
like a / l, for a characteristic lengthscale a
— same ratio, same e, m = same solution
regardless of whether a = 1µm or 1km

Our (typical) approach:

pick characteristic lengthscale a
– measure distance in units of a
– measure time in units of a/c
– measure w in units of 2πc/a = a / l
– ....
 Absorbing Boundaries:
Perfectly Matched Layers
“perfect” absorber: PML Artificial absorbing material
overlapping the computation

Theoretically reflectionless

… but PML is no longer perfect

with finite resolution,
so “gradually turn on” absorption
over finite-thickness PML
Computational Photonics Problems Numerical Methods: Basis Choices

finite difference
• Time-domain simulation
— start with current J(x,t)
— run “numerical experiment” to spectral methods
simulate E(x, t), H(x, t)
df f ( x + Dx) - f ( x - Dx)
dx
ª
Dx
+ O(Dx 2 ) +
• Frequency-domain linear response
— start with harmonic current J(x, t) = e–iwt J(x) +
†
— solve for steady-state harmonic fields E(x), H(x) + …..
— involves solving linear equation Ax=b
finite elements
• Frequency-domain eigensolver in irregular “elements,”
approximate unknowns
— solve for source-free harmonic eigenfields
by low-degree polynomial
E(x), H(x) ~ e–iwt
— involves solving eigenequation Ax=w2x

boundary-element methods
discretize only the
boundaries between
homogeneous media
 A Maxwell Eigenproblem
r r 1∂ r w r
—¥E =- H =i H First task:
c ∂t c get rid of this mess
r r 1∂ r r0 w r
— ¥ H =e E + J = -i eE
c ∂t c

dielectric function e(x) = n2(x)

1 r Êw ˆ 2 r
+ constraint
—¥ —¥H =Á ˜ H r
Ë c¯ —⋅ H = 0
e
eigen-operator eigen-value eigen-state
Electronic & Photonic Eigenproblems
Electronic Photonic
Ê h2 2 ˆ 1 r Ê w ˆ 2 r

Á- — + V ˜y = Ey —¥ —¥H =Á ˜ H
Ë 2m ¯ e Ë c¯
nonlinear eigenproblem simple linear eigenproblem
(V depends on e density |y|2) (for linear materials)

—many well-known
computational techniques

Hermitian = real E & w, … Periodicity = Bloch’s theorem…

A 2d Model System

dielectric “atom”
e=12 (e.g. Si)

square lattice,
period a
a
a
E
TM
H
 Periodic Eigenproblems
if eigen-operator is periodic, then Bloch-Floquet theorem applies:

r r i ( k ⋅ x -wt ) r
r r
r
can choose: H( x ,t) = e Hkr ( x )

planewave
periodic “envelope”

Corollary 1: k is conserved, i.e. no scattering of Bloch wave

r
Corollary 2: Hrk given by finite unit cell,
so w are discrete wn(k)
 A More Familiar Eigenproblem
e=1

a e = 12 y
x
band diagram / dispersion relation
w
find the normal modes light cone
(all non-guided modes)
of the waveguide:
i(kx-wt )
H(y,t) = H k (y)e
(propagation constant k
a.k.a. b)
k
 Solving the Maxwell Eigenproblem
Finite cell Ë discrete eigenvalues wn 1 wn 2
(— + ik) ¥ (— + ik) ¥ H n = 2 H n
e c
Want to solve for wn(k),
& plot vs. “all” k for “all” n,
constraint: (— + ik) ⋅ H = 0
1

0.9
† where:
0.8

0.7
H(x,y)#ei(k⋅x – wt)
0.6

0.5

0.4
Photonic Band Gap
0.3

0.2

0.1
TM bands
†
0

1 Limit range of k: irreducible Brillouin zone

2 Limit degrees of freedom: expand H in finite basis

3 Efficiently solve eigenproblem: iterative methods

Solving the Maxwell Eigenproblem: 1
1 Limit range of k: irreducible Brillouin zone
—Bloch’s theorem: solutions are periodic in k
M

2p ky
first Brillouin zone G X a
kx
= minimum |k| “primitive cell”

† reduced by symmetry
irreducible Brillouin zone:

2 Limit degrees of freedom: expand H in finite basis

3 Efficiently solve eigenproblem: iterative methods

Solving the Maxwell Eigenproblem: 2a
1 Limit range of k: irreducible Brillouin zone
2 Limit degrees of freedom: expand H in finite basis (N)

N
H = H(xt ) = Â hm bm (x t ) ˆ 2
solve: A H = w H
m=1

2
finite matrix problem: Ah = w Bh

f g = Ú f* ⋅g Aml = bm Aˆ bl Bml = bm bl

3 Efficiently solve eigenproblem: iterative methods

Solving the Maxwell Eigenproblem: 2b
1 Limit range of k: irreducible Brillouin zone
2 Limit degrees of freedom: expand H in finite basis
— must satisfy constraint: (— + ik) ⋅ H = 0

Planewave (FFT) basis Finite-element basis

constraint, boundary conditions:
iG⋅xt
H(x t ) = Â HG e † Nédélec elements
G [ Nédélec, Numerische Math.
35, 315 (1980) ]
constraint: H G ⋅ (G + k) = 0
nonuniform mesh,
uniform “grid,” periodic boundaries,
more arbitrary boundaries,
simple code, O(N log N) [ figure: Peyrilloux et al.,
J. Lightwave Tech.
21, 536 (2003) ]
complex code & mesh, O(N)

† 3 Efficiently solve eigenproblem: iterative methods

Solving the Maxwell Eigenproblem: 3a
1 Limit range of k: irreducible Brillouin zone
2 Limit degrees of freedom: expand H in finite basis
3 Efficiently solve eigenproblem: iterative methods
2
Ah = w Bh
Slow way: compute A & B, ask LAPACK for eigenvalues
— requires O(N2) storage, O(N3) time
Faster way:
— start with initial guess eigenvector h0
— iteratively improve
— O(Np) storage, ~ O(Np2) time for p eigenvectors
(p smallest eigenvalues)
Solving the Maxwell Eigenproblem: 3b
1 Limit range of k: irreducible Brillouin zone
2 Limit degrees of freedom: expand H in finite basis
3 Efficiently solve eigenproblem: iterative methods
2
Ah = w Bh
Many iterative methods:
— Arnoldi, Lanczos, Davidson, Jacobi-Davidson, …,
Rayleigh-quotient minimization
Solving the Maxwell Eigenproblem: 3c
1 Limit range of k: irreducible Brillouin zone
2 Limit degrees of freedom: expand H in finite basis
3 Efficiently solve eigenproblem: iterative methods
2
Ah = w Bh
Many iterative methods:
— Arnoldi, Lanczos, Davidson, Jacobi-Davidson, …,
Rayleigh-quotient minimization
for Hermitian matrices, smallest eigenvalue w0 minimizes:
“variational 2 h' Ah minimize by preconditioned
theorem” w = min
0 conjugate-gradient (or…)
h h' Bh
 Band Diagram of 2d Model System
a (radius 0.2a rods, e=12)
1

frequency w (2πc/a) = a / l
0.9

0.8

0.7

0.6

0.5

0.4
Photonic Band Gap
0.3

0.2
TM bands
0.1

irreducible Brillouin zone G X M G

M
r E gap for
k X TM
G H n > ~1.75:1
 Origin of Gap in 2d Model System
orthogonal: node in high e 1

0.9

0.8

Ez 0.7

0.6

0.5

0.4
Photonic Band Gap
0.3

0.2
TM bands
lives in high e 0.1

0
G X M G
Ez
E gap for
TM
H n > ~1.75:1
– +
 The Iteration Scheme is Important
(minimizing function of 104–108+ variables!)
2 h' Ah
w = min
0 = f (h)
h h' Bh

Steepest-descent: minimize (h + a —f) over a … repeat

Conjugate-gradient: minimize (h + a d)
† — d is —f + (stuff): conjugate to previous search dirs
Preconditioned steepest descent: minimize (h + a d)
— d = (approximate A-1) —f ~ Newton’s method
Preconditioned conjugate-gradient: minimize (h + a d)
— d is (approximate A-1) [—f + (stuff)]
 The Iteration Scheme is Important
(minimizing function of ~40,000 variables)
1000000

100000
E E E E E
10000 E EEE
EEEE
EEEEEE
EEE
EE
EE
E
1000 J E
EE
E
EE
EE
E
EE
EE
EE
E
Ñ E
EE
EE
E
EE
EE
E
EE
EE
E
EE
EE
E
E
100
Ñ
J
E
EE
E
EE
EE
E
EE
EE
E
E
no preconditioning
% error

Ñ Ñ
J E
EE
E
EE
EE
E
EE
E
10 J Ñ E
E
EE
EE
E
E
J ÑÑ E
EE
E
EE
E
J ÑÑ E
E
EE
E
J ÑÑ E
E
EE
E
1 JJ ÑÑÑÑ E
E
EE
EE
E
E
Ñ
JJJJJ ÑÑÑÑÑÑ E
E
EE
JJ ÑÑ
ÑÑ
ÑÑ
ÑÑ
Ñ
ÑÑ
ÑÑ
Ñ
ÑÑ
ÑÑ
Ñ EE
E
0.1 JJ Ñ
ÑÑ
ÑÑ
ÑÑ
Ñ
ÑÑ
ÑÑ EE
E
ÑÑ
Ñ
ÑÑ
ÑÑ
Ñ
ÑÑ E
E
J ÑÑ
Ñ
ÑÑ
ÑÑ
Ñ EE
E
J ÑÑ
ÑÑ
Ñ EE
E
E
0.01 ÑÑ
ÑÑ
Ñ EE
E
J ÑÑ
ÑÑ EE
E
Ñ
ÑÑ
Ñ EE
E
E
Ñ
Ñ
ÑÑ E
E
J ÑÑ
Ñ
0.001 ÑÑ
preconditioned J
J
ÑÑ
Ñ
ÑÑ
0.0001
conjugate-gradient J
J
no conjugate-gradient
0.00001 J
J
0.000001
1 10 100 1000
# iterations
The Boundary Conditions are Tricky
E|| is continuous

E^ is discontinuous
(D^ = eE^ is continuous)

Any single scalar e fails:

(mean D) ≠ (any e) (mean E)
e?
Use a tensor e:
Êe ˆ
Á ˜ E||
Á e ˜
Á -1 ˜
Ë e-1 ¯ E^
 The e-averaging is Important
100

H
H
backwards averaging
10 H H H H H
B
J B
B H H
H H
correct averaging
B H H
% error

J
J
B B B changes order
B no Baveraging
J
1 J J
B B of convergence
J
J J
B B
from ∆x to ∆x2
tensor averaging J J
0.1
J J

0.01
10 100 (similar effects
resolution (pixels/period) in other E&M
numerics & analyses)
 Gap, Schmap?
1

a 0.9

0.8

0.7

frequency w
0.6

0.5

0.4
Photonic Band Gap
0.3

0.2
TM bands
0.1

0
G X M G

But, what can we do with the gap?

 Intentional “defects” are good

microcavities waveguides (“wires”)

3D Photonic C rysta l with Defects

 Intentional “defects” in 2d
(Same computation, with supercell = many primitive cells)

a
Microcavity Blues
For cavities (point defects)
frequency-domain has its drawbacks:

• Best methods compute lowest-w bands,

but Nd supercells have Nd modes
below the cavity mode — expensive

• Best methods are for Hermitian operators,

but losses requires non-Hermitian
 Time-Domain Eigensolvers
(finite-difference time-domain = FDTD)

Simulate Maxwell’s equations on a discrete grid,

+ absorbing boundaries (leakage loss)

• Excite with broad-spectrum dipole ( ) source

Dw

signal processing Response is many

complex wn sharp peaks,
[ Mandelshtam,
J. Chem. Phys. 107, 6756 (1997) ] one peak per mode

decay rate in time gives loss

 Signal Processing is Tricky
signal processing
complex wn
?
a common approach: least-squares fit of spectrum
0.8 450
E
0.6 EE 400
EE

0.4 EE
E
350 fit to:
E E
E
0.2 E
EE
EE
E
E
E E E
E
EE
EE
E E
300 E E
A
0 E E E E E E E E E
EEE
E
EE
E
EE E
EE
E
EE
EE
E
EE
E
EEEE
E
EE
EE
E
EE
E
EE
E
EE
EE
E
EE
E
EE
E
EE
E
EE
E
EE
EE
E
EE
E
EE
E
EE
EE
E
EE
E
EE
E
EE
EE
E
EE
E
EE
EE
E
EE
E
EE
E
EE
EE
E
EE
E
EE
E
EE
EE
E
EE
E
EE
E
EE
E
E
E
E E E
EE
E
(w - w 0 ) 2 + G 2
E
EEE
EE
E E
EE
E
E 250
EE E
E
E E EE
E
E E
E
-0.2 E 200

-0.4
E EE
EE
EE FFT 150
E E

E
-0.6 100 E E
E
-0.8 E E
E
50
EE
†
E E
EE
-1 E EE EEEE
EEEEEEEEEEEEEEEEEEE
0
0 1 2 3 4 5 6 7 8 9 10 0 0.5 1 1.5 2 2.5 3 3.5 4

Decaying signal (t) Lorentzian peak (w)

 Fits and Uncertainty
problem: have to run long enough to completely decay
1 40000
E
EE
E
0.8 EE
EE E 35000
EE

0.6 EE
EE
EE EE
E
EE
EE
E
EE
E
actual
EE EE EE
E 30000
EE EE EE
E
0.4 EE EE E
EE
EE EE E
E
E
EE EE EE EE
E E E E EE
E E EE EE 25000
0.2 E E E E E E E E E E E E E E
0 E E E E E E E E E E E E E E E E E E E E 20000

E E E E E E E
-0.2 E E E E E E E
E E EE E
E E E E EE 15000
E EE EE EE E
EE EE E
-0.4 EE
EE EE EE
E
E
E
EE E E
EE EE E
-0.6 E
EE
EE
EE
EE
EE
E
E
EE
EE
E
E 10000
signal
EE
EE
-0.8 E
EE
E
EE
E
E 5000
portion
E E E
-1 E 0 E E E E E E E E
0 1 2 3 4 5 6 7 8 9 10 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5

Portion of decaying signal (t) Unresolved Lorentzian peak (w)

There is a better way, which gets complex w to > 10 digits

 Unreliability of Fitting Process
Resolving two overlapping peaks is
near-impossible 6-parameter nonlinear fit
(too many local minima to converge reliably)
1200

1000
E
sum of two peaks
There is a better
800
E
E
way, which gets
600
complex w
for both peaks
400
w = 1+0.033i E E to > 10 digits
E E
w = 1.03+0.025i
200
E E
E E
E EE
E
E E EE E EE
0 E E E EE E E E E EE E E E E E E E EE E E E E EE E E

0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
Sum of two Lorentzian peaks (w)
 Quantum-inspired signal processing (NMR spectroscopy):
Filter-Diagonalization Method (FDM)
[ Mandelshtam, J. Chem. Phys. 107, 6756 (1997) ]

Given time series yn, write: y n = y(nDt) = Â ak e -iw k nDt

k
…find complex amplitudes ak & frequencies wk
by a simple linear-algebra problem!

Idea: pretend y(t) is autocorrelation of a quantum system:

†
ˆ ∂ ˆ -iHˆ Dt / h
H y = ih y time-∆t evolution-operator: U =e
∂t
say: y n = y (0) y (nDt) = y (0) Uˆ n y (0)

†
Filter-Diagonalization Method
[ Mandelshtam, J. Chem. Phys. 107, 6756 (1997) ]
(FDM)

y n = y (0) y (nDt) = y (0) Uˆ n y (0) ˆ

U =e -iHˆ Dt / h

We want to diagonalize U: eigenvalues of U are eiw∆t

…expand U in basis of†|y(n∆t)>:

U m,n = y (mDt) Uˆ y (nDt) = y (0) Uˆ mUˆ Uˆ n y (0) = y m +n +1

Umn given by yn’s — just diagonalize known matrix!

 Filter-Diagonalization Summary
[ Mandelshtam, J. Chem. Phys. 107, 6756 (1997) ]

Umn given by yn’s — just diagonalize known matrix!

A few omitted steps:
—Generalized eigenvalue problem (basis not orthogonal)
—Filter yn’s (Fourier transform):
small bandwidth = smaller matrix (less singular)

• resolves many peaks at once

• # peaks not known a priori
• resolve overlapping peaks
• resolution >> Fourier uncertainty
 Do try this at home
FDTD simulation:
http://ab-initio.mit.edu/meep/

Bloch-mode eigensolver:
http://ab-initio.mit.edu/mpb/

Filter-diagonalization:
http://ab-initio.mit.edu/harminv/

Photonic-crystal tutorials (+ THIS TALK):

http://ab-initio.mit.edu/
/photons/tutorial/
 Meep (FDTD) MPB (Eigensolver)
• Arbitrary e(x) — including
dispersive, loss/gain, • Arbitrary periodic e(x) —
and nonlinear [c(2) and c(3)] anisotropic, magneto-optic, …
(lossless, linear materials)
• Arbitrary J(x,t)
• PML/periodic/metal bound. • 1d/2d/3d

• 1d/2d/3d/cylindrical • band diagrams, group velocities

perturbation theory, …
• power spectra • eigenmodes

Free/open-source • fully scriptable interface

software (GNU)
• built-in multivariate optimization,
• MPI parallelism integration, root-finding, …
• exploit mirror symmetries • field output (standard HDF5 format)
 Unix Philosophy
combine small, well-designed tools, via files

Input text file MPB/Meep standard formats

(text + HDF5)
Disadvantage:
— have to learn several programs
Visualization / Analysis
Advantages: software
— flexibility (Matlab, Mayavi [vtk],
— batch processing, shell scripting command-line tools, …)
— ease of development
 Unix Philosophy
combine small, well-designed tools, via files

Input text file MPB/Meep standard formats

(text + HDF5)

GNU Guile scripting interpreter

(Scheme language)
Visualization / Analysis
software
Embed a full scripting language: (Matlab, Mayavi [vtk],
— parameter sweeps command-line tools, …)
— complex parameterized geometries
— optimization, integration, etc.
— programmable J(x, t), etc.
— … Turing complete
 A Simple Example (MPB)
e=1

a e = 12 y
find the normal modes wn(k) x
of the waveguide:
i(kx-wt )
H(y,t) = H k (y)e

Need to specify:
• computational cell size/resolution
† • geometry, i.e. e(y)
• what k values
• how many modes (n = 1, 2, … ?)
 A File Format Made of Parentheses
Need to specify:
• computational cell size/resolution
(set! geometry-lattice (make lattice (size no-size 10 no-size)
(set! resolution 32)

• geometry, i.e. e(y)

• what k values
• how many modes (n = 1, 2, … ?) 1 pixel

10 (320 pixels)
e=1

a e = 12 y
x
 A File Format Made of Parentheses
Need to specify:
• computational cell size/resolution
• geometry, i.e. e(y)
(set! geometry (choose units of a)
(list
(make block (size infinity 1 infinity)
(center 0 0 0)
(material (make dielectric (epsilon 12))))))
• what k values
• how many modes (n = 1, 2, … ?)

e=1

a e = 12 y
x
 A File Format Made of Parentheses
Need to specify:
• computational cell size/resolution
• geometry, i.e. e(y)
(units of 2π/a)
• what k values
(set! k-points
(interpolate 10 (list (vector3 0 0 0) (vector3 2 0 0))))

(built-in function)
• how many modes (n = 1, 2, … ?)

e=1

a e = 12 y
x
 A File Format Made of Parentheses
Need to specify:
• computational cell size/resolution
• geometry, i.e. e(y) …Then run:
• what k values (run)
• how many modes (n = 1, 2, … ?)
(set! num-bands 5) or only TM polarization:
(run-tm)

or only TM, even modes:

(run-tm-yeven)

e=1

a e = 12 y
x
 Simple Example (MPB) Results
e=1

a e = 12 y
find the normal modes wn(k) x
of the waveguide:

red = even
blue = odd
 Do try this at home
FDTD simulation:
http://ab-initio.mit.edu/meep/

Bloch-mode eigensolver:
http://ab-initio.mit.edu/mpb/

Filter-diagonalization:
http://ab-initio.mit.edu/harminv/

Photonic-crystal tutorials (+ THIS TALK):

http://ab-initio.mit.edu/
/photons/tutorial/