2/18/2020

Computational Science:
Computational Methods in Engineering

Slab Waveguide Analysis

Outline

• Slab Waveguides
• Formulation
• Solution
• Implementation in MATLAB
• More About Resolution and Spacer Regions

1
 2/18/2020

Slab Waveguides

Slide 3

Refractive Index n
Light travels at different speeds when it is inside different materials.

Frequency is constant.
Speed changes.
Wavelength changes.

The factor by which light slows c

down is called the refractive index. n
v
Slide 4

2
 2/18/2020

Snell’s Law
z
Snell’s law quantifies the angles
of light rays at an interface.
1 n1

x
n1 sin 1  n2 sin  2
n2
2

Slide 5

Critical Angle c
z There exists a special angle, the
critical angle, where the ray in
the low-index medium is at 90°.
1  90 n1 n1 sin 1  n2 sin  2
n1 sin 90  n2 sin  c
x n1  n2 sin  c
sin  c  n1 n2
 c  sin 1  n1 n2 
n2 2  c
 c  sin 1  n1 n2 
where n2  n1
Slide 6

3
 2/18/2020

Total Internal Reflection (TIR)

z When a light ray is incident onto
an interface at an angle greater
than the critical angle, the light
n1 completely reflects and no light
is transmitted.
No transmission!
x This is called total internal
reflection (TIR).
c
n2
2 2  c

Slide 7

The Slab Waveguide

If we “sandwich” a slab of high-index material between two materials
with lower refractive index, we form a slab waveguide.

n1 Conditions
n2  n1
TIR
and
n2 n2  n3
TIR

n3

Slide 8

4
 2/18/2020

Ray Tracing Picture

The round trip phase of a ray
must be an integer multiple of
2. Otherwise the wave will
m  2  interfere with itself and escape
from the slab.

Because of this, only certain

  angles are allowed to
propagate in the waveguide.

This is the origin of discrete

modes in a waveguide.

  k0 neff  k0 n sin 

Slide 9

Rigorous Analysis

  k0 neff  k0 n sin 
Slide 10

5
 2/18/2020

Slab Vs. Channel Waveguides

Slab waveguides confine Channel waveguides confine
energy in only one transverse energy in both transverse
direction. directions.

Confinement
Confinement

Slide 11

Mathematical Form of Solution

 
x E  x, y , z   A  x  e  j  z
Amplitude Profile

Wave oscillations

y z   phase constant

Slide 12

6
 2/18/2020

Formulation

Slide 13

What is Formulation?
Formulation is the initial analytical work we do before implementing a computer code.

Usually we start with the governing equation(s) and end with the matrix equation to be
solved.

Governing Equations Matrix Equation

 
  E   j H
  Av   v
  H  j E

Slide 14

7
 2/18/2020

Governing Equations
Since this is an electrodynamics problem, we start with Maxwell’s curl equations.

 
  E   j H
 
  H  j E

Vector Curl
The curl of a vector is a measure of the vector field’s tendency to circulate about an axis. The curl quantity is
directly along this axis and the magnitude measures the strength of the circulation.

  A Ay   Ax Az   Ay Ax 

 A   z   aˆ x     aˆ y     aˆ z
 y z   z x   x y 

Slide 15

Expand Governing Equations

(1 of 2)
If we expand the first equation into its vector components, we get
 
  E   j H
 Ez E y   Ex Ez   E y Ex 
   aˆ x     aˆ y     aˆ z   j  H x aˆ x  H y aˆ y  H z aˆ z 
 y z   z x   x y 

The vector components on each side must be equal.

Ez E y
x -component:    j H x
y z
Ex Ez
y -component:    j H y
z x
E y E x
z -component:    j H z
x y
Slide 16

8
 2/18/2020

Expand Governing Equations

(2 of 2)
There are now six equations. Ez E y
   j H x
y z
Ex Ez
   j H y
z x
  E y Ex
  E   j H x

y
  j H z

  H z H y
  H  j E y

z
 j Ex

H x H z
  j E y
z x
H y H x
  j Ez
x y
Slide 17

How to Reduce Dimensions

It is always good practice to minimize the number of dimensions utilized in a numerical
analysis.
x
x Material changes as a function of x. The mode
profile will change as a function of x. We must
retain this dimension.
y
Device is uniform. Wave does not propagate
in this direction. Mode profile is uniform.


0
y
y z
z
Device is uniform. Wave 
propagates in this direction so   j
wave phase is increasing. z
Slide 18

9
 2/18/2020

 y  0 Apply
Since nothing is changing in the y direction, any derivative with respect to y must be zero.

Ez E y E y
   j H x    j H x
y z z
Ex Ez Ex Ez
   j H y    j H y
z x z x
E y Ex E y
   j H z   j H z
x y x

H z H y H y
  j E x   j E x
y z z
H x H z H x H z
  j E y   j E y
z x z x
H y H x H y
  j Ez  j E z
x y x

Slide 19

Two Distinct Mode Types

Our revised governing equations have We will analyze the Ey mode
separated into two distinct mode types. Mode Type 1 – Ey Mode
H x H z
E y   j E y
   j H x z x
z E y
Ex Ez    j H x
   j H y z
z x E y
E y   j H z
  j H z x
x
Mode Type 2 – Hy Mode
H y
  j E x Ex Ez
z    j H y
z x
H x H z
  j E y H y
z x   j Ex
H y z
 j E z H y
x  j Ez
x
Slide 20

10
 2/18/2020

What About ?
The guided mode has the following mathematical form
 
E  x, y , z   A  x  e  j  z
Now calculate the partial derivative with respect to z and see what happens.
       
E  x, y, z    A  x  e  j  z   A  x  e  j z  e  j  z A x
z z z z
 
  j  A  x  e  j  z   j  E  x, y , z 
 

E x, y , z 

It can be concluded that for this slab waveguide analysis,


  j
z
Slide 21

1D Governing Equations
The equations for the Ey mode were
 H z
Hx   j E y
z x

 E y   j H x
z
E y
  j H z
x

These equations were derived simply replacing/z with –j.

dH z
 j H x   j E y
dx The partial derivative has become an
j  E y   j H x ordinary derivative because there is only one
dE y independent variable remaining.
  j H z
dx

Slide 22

11
 2/18/2020

Normalize the Parameters

Before converting our equations to matrix form, we should normalize the spatial coordinate
x to put it in terms of wavelength in some manner.
??? x
x 
0

Additionally, it will be mathematically convenient to normalize by multiplying x by the free

space wave number k0.

x  k0 x 2  
k0  
0 n

Slide 23

Normalizing Maxwell’s Equations

Start with the following equation,
dH z
 j H x   j E y
dx
and replace x with x k0 .
dH z
 j  H x  k0  j E y
dx
Next, divide both sides of the equation by k0.
 dH z j
j Hx   Ey
k0 dx k0
Recognizing that  = k0neff, this equation becomes
dH z j
 jneff H x   Ey
dx k0
j 0 r 
 Ey  j 0  r Ey
  0 0 0
Slide 24

12
 2/18/2020

Normalized Equations
Applying the normalizations to all equations to get
dH z 
 jneff H x   j 0  r Ey

dx 0
0
jneff E y   j H
0 r x
dE y 
  j 0 r H z
dx 0

Last, at optical frequencies, the magnetic response is negligible so r = 1.

dH z 
 jneff H x   j 0  r Ey
dx 0
0
jneff E y   j H
0 x
dE y 
  j 0 Hz
dx 0
Slide 25

Final Governing Equation

Solve the last two equations for Hx and Hz.
dH z 
 jneff H x   j 0  r Ey
dx 0
0 0
jneff E y   j H  H x  neff E
0 x 0 y
dE y 0  0 dE y
j H  Hz  j
dx 0 z 0 dx

These are substituted into the first equation to get a single equation containing only Ey.
This is why it was called the Ey mode.
dH z 
 jneff H x   j 0  r Ey
dx 0
  0  d   0 dE y  0
 jneff  neff E y    j   j  r Ey
  0  
dx   0 dx   0

d 2Ey d 2 Ey
2
neff Ey    r Ey   r E y  neff
2
Ey
dx 2 dx 2
Slide 26

13
 2/18/2020

Eigen-Value Problem
For optical problems, people like to put everything in terms of refractive index. This is
related to the relative permittivity through r = n2.

d 2 Ey
 n 2 E y  neff
2
Ey
dx 2

The governing equation can be rearranged to the form of a standard eigen-value

problem 𝐀𝐱 = 𝜆𝐱.
d2
A  n2  x 
d 2
 dx 2
 dx 2  n  x   E y  x   neff E y  x 
2 2
x  Ey  x 
 
  neff
2

Slide 27

Matrix Form
We go term-by-term to write the equation in matrix form.

 d2 
 dx 2  n  x   E y  x   neff E y  x 
2 2

 

D 2
x  n 2  e y  neff
2
ey
or

D 2
x  ε  e y  neff
2
ey

Eigen Matrix Eigen Value

Slide 28

14
 2/18/2020

Solution

Slide 29

Solving the Eigen-Value Problem

V  Eigen-vector matrix
D 2
x  ε  e y  neff
2
ey 
λ  Eigen-value matrix
 
  e y 1   ey2 1   eyM  1  
1

  e1 2 
   2   M  
 y     ey  2  
  2 
 ey  2   
 M  
 e1  3
 ey  3  e  3   M = # modes
V   y    y  Usually M = Nx
        

  1     
 ey  N x  1 ey  N x  1 eyM   N x  1 
 2

 e1  N    2   M  
 y   ey  N x    ey  N x   

x


 n   
2
1
 eff 
 
 n   Eigen-vectors and eigen-values come in pairs.
2
2
 
D x   eff

   Do not mix up their pairing!
 
n 
2
M 
 eff 

Slide 30

15
 2/18/2020

Visualizing the Solution

The columns of the eigen-vector matrix are pictures of the modes.
 
  x
 
 
 
  z
 
V 
 
 
 
 
 
 
 

The eigen-values are the effective

refractive indices of the modes squared.
  neff
2

Slide 31

Implementation
in MATLAB

Slide 32

16
 2/18/2020

Implementation Outline

1. Initialize MATLAB
2. Dashboard (materials, dimensions, etc.)
3. Calculate Grid
4. Build Device on Grid
5. Perform Finite-Difference Analysis
6. Visualize the Results

Slide 33

Dashboard
% slabdemo.m

% INITIALIZE MATLAB
close all;
clc;

b n2 How big should we

clear all;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
make b? %% DASHBOARD
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 Enough to allow

a n1 the mode to decay to

zero before reaching
% FREE SPACE WAVELENGTH
lam0 = 1.0;
the boundary.
% SLAB PARAMETERS
n1 = 2.0;

b n2 n2 = 1.0;
a = 3*lam0;

% GRID
b = 5*lam0;
What grid resolution NRES = 10;
should we use? dx = lam0/NRES;

 Convergence % NUMBER OF MODES TO CALCULATE

M = 5; Slide 34

17
 2/18/2020

Build Device on Grid

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% BUILD WAVEGUIDE
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% COMPUTE GRID

b n2 Sx = a + 2*b;
Nx = ceil(Sx/dx);
Sx = Nx*dx;

xa = [0.5:Nx-0.5]*dx;
nx1
a n1 Sx nx
xa = xa - mean(xa);

% COMPUTE START AND STOP INDICES

nx2 nx = round(a/dx);
nx1 = round((Nx - nx)/2);
nx2 = nx1 + nx - 1;

b n2 % BUILD N
N = zeros(Nx,1);
N(1:nx1-1) = n2;
N(nx1:nx2) = n1;
N(nx2+1:Nx) = n2;

Slide 35

Perform Finite-Difference Analysis

 2 1 
 1 2 1  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
  %% PERFORM FD ANALYSIS
 1 2 1  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1  
D2x    
 k0 x 2  1 2 1  % CALCULATE k0
  k0 = 2*pi/lam0;
 1 2 1 
 1 2 

% BUILD DX2
DX2 = sparse(Nx,Nx);
DX2 = spdiags(+1*ones(Nx,1),-1,DX2);
 n 1 
  DX2 = spdiags(-2*ones(Nx,1), 0,DX2);
 n  2  DX2 = spdiags(+1*ones(Nx,1),+1,DX2);
N n  3  DX2 = DX2 / (k0*dx)^2;
 
  
 n  N x   % MAKE N DIAGONAL

N = diag(sparse(N(:)));

D
 N e
% SOLVE EIGEN-VALUE PROBLEM
2
x
2
 neff
2
ey A = DX2 + N^2;

y [V,D] = eig(full(A));
NEFF = sqrt(diag(D));
A

Slide 36

18
 2/18/2020

Visualize the Results

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% VISUALIZE
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% SORT MODES
[~,ind] = sort(real(NEFF),'descend');
V = V(:,ind);
NEFF = NEFF(ind);

% OPEN FIGURE WINDOW

figure('Color','w');
hold on;

% DRAW SLAB WAVEGUIDE

x = [0 2*(M+1) 2*(M+1) 0 0];
y = [ -b-a/2 -b-a/2 b+a/2 b+a/2 -b-a/2 ];
fill(x,y,0.9*[1 1 1]);
y = [ -a/2 -a/2 a/2 a/2 -a/2 ];
fill(x,y,0.5*[1 1 1]);

% DRAW AND LABEL MODES

for m = 1 : M
x0 = 2*m;
y0 = (a + b)/2;
x = x0 + 3*V(:,m);
y = linspace(-b-a/2,b+a/2,Nx);
line(x,y,'Color','w','LineWidth',4);
h = line(x,y,'Color','b','LineWidth',2);
text(x0,y0,['Mode ' num2str(m)],'Rotation',-90,...
'HorizontalAlignment','center','VerticalAlignment','bottom');
text(x0,-y0,['n_{eff} = ' num2str(NEFF(m),'%4.2f')],'Rotation',-90,...
'HorizontalAlignment','center','VerticalAlignment','bottom');
end

% SET GRAPHICS VIEW

hold off;
h2 = get(h,'Parent');
set(h2,'XTick',[]);
axis equal tight;
ylabel('x axis');
Slide 37

More About Resolution

and Spacer Regions

Slide 38

19
 2/18/2020

Convergence Study for NRES

Notes
• Higher-order modes converge slower.
• Higher-order modes have a smaller neff.

0
x 
NRES

Slide 39

Spacer Region b
Metal
Remember the Dirichlet boundary conditions?
Values outside of the grid are forced to zero.
b n2
This means we really are simulating a slab
waveguide inside of a large metal waveguide.
a n1
It is only possible to get an accurate simulation
of the slab waveguide when the metal
b n2 waveguide is large enough.

We must choose b to be large enough to

Metal ensure the metal waveguide is insignificant.

Slide 40

20
 2/18/2020

Effect of Spacer Region Size

If the spacer region b is too small, the outer

metal waveguide becomes significant and the
results for the slab are not accurate.

Slide 41

Conditions for Large Evanescent Fields: Thin

Waveguides

Thin dielectric waveguides have large

evanescent fields.

The spacer region b must be big enough to

sufficiently encompass the evanescent field
in order to give an accurate simulation.

Slide 42

21
 2/18/2020

Conditions for Large Evanescent Fields: Modes Near

Cutoff

Guided modes operating near cutoff have

very large evanescent fields.

The spacer region b must be big enough to

sufficiently encompass the evanescent field
in order to give an accurate simulation.

Slide 43

Convergence Study for Spacer Region b

Notes
• Under normal circumstances, the
spacer region size can be ~0.250.
• Modes near cutoff require larger
spacer regions to resolve.
• Thin waveguides may require larger
spacer regions.
• Always check for convergence of
spacer region size.

Slide 44

22