Lecture 5a - Scattering Matrices
Lecture 5a - Scattering Matrices
Instructor
Dr. Raymond Rumpf
(915) 747‐6958
[email protected]
EE 5337
Computational Electromagnetics
Lecture #5a
Scattering Matrices for
Semi‐Analytical Methods
These notes may contain copyrighted material obtained under fair use rules. Distribution of these materials is strictly prohibited
Lecture 5a Slide 1
Outline
• Scattering matrix for a single layer
• Multilayer structures
• Longitudinally periodic structures
• Dispersion analysis
• Alternatives to scattering matrices
Lecture 5a Slide 2
1
12/4/2017
Scattering Matrix
for a Single Layer
R. C. Rumpf, “Improved Formulation of Scattering Matrices for Semi‐Analytical
Methods That is Consistent with Convention,” PIERS B, Vol. 35, pp. 241‐261, 2011.
Lecture 5a Slide 3
Motivation for Scattering Matrices
Scattering matrices offer several important features and
benefits:
• Unconditionally stable method.
• Parameters have physical meaning.
• Parameters correspond to those measured in the lab.
• Can be used to extract dispersion.
• Very memory efficient.
• Can be used to exploit
longitudinal periodicity.
• Mature and proven approach.
• Much greater wealth of
literature available.
However, excellent alternatives to S‐matrices do exist!
Lecture 5a Slide 4
2
12/4/2017
Geometry of a Single Layer
Indicates a point that lies on an interface, but associated with a particular side.
ψ1 ψ i 0 ψ i k0 Li ψ 2
ψ i zi
ψ1 ψ i 0 ψ i k0 Li ψ 2
ψ i zi
Definition of A Scattering Matrix
c1 S11 S12 c1 S11 reflection
c S
2 21 S 22 c 2
S 21 transmission
This is consistent with network theory and experimental convention.
Lecture 5a Slide 6
3
12/4/2017
Field Relations
Field inside the ith layer:
Ex ,i zi
E y ,i zi Wi Wi e λi zi 0 ci
ψ i zi
H x ,i zi Vi Vi 0 e λi zi ci
H y ,i zi
Boundary conditions at the first interface:
ψ1 ψ i 0
W1 W1 c1 Wi Wi ci
V V V
Vi ci
1 1 c1 i
Boundary conditions at the second interface:
ψ i k0 Li ψ 2
Wi Wi e λ i k0 Li 0 ci W2 W2 c2 Note: k0 has been
Vi 0 V2 c2
incorporated to normalize Li.
Vi e λ i k0 Li ci V2
Lecture 5a Slide 7
Derivation of the Scattering Matrix
Solve both boundary condition equations for the intermediate mode
ci ci
coefficients and .
1 1
ci Wi Wi W1 W1 c1 ci e λ i k0 Li 0 Wi Wi W2 W2 c2
ci Vi Vi V1 V1 c1 c i 0 e λ i k0 Li Vi Vi V2 V2 c2
Both of these equations have the term
1
Wi Wi W j W j 1 A ij Bij A ij Wi1W j Vi1V j
V j 2 Bij A ij
See HW2
Vi Vi V j Bij Wi1W j Vi1V j
We set substitute this result into the first two equations and then set
them equal to eliminate the intermediate mode coefficients.
1 Ai1 Bi1 c1 1 e λ i k0 Li 0 Ai 2 Bi 2 c2
2 Bi1 Ai1 c1 2 0 e λ i k0 Li
Bi 2 Ai 2 c2
We write this as two matrix equations and rearrange the terms until
they have the form of a scattering matrix.
c1 ? ? c1
c2 ? ? c2
Lecture 5a Slide 8
4
12/4/2017
The Scattering Matrix
The scattering matrix Si of the ith
layer is defined as:
c1 i c1
S11
i
i
S12
i
S S i
c2 c2
S 21 S 22i
After some algebra, the components of the Si
scattering matrix are computed according to
X B A X A
i 1
S11 A i1 Xi Bi 2 A i21Xi Bi1 i i2
1
i2 i i1 Bi1 A ij Wi1W j Vi1V j
S A X A B A
1
i
12 i1 Xi Bi 2 A i21Xi Bi1 i i2 i2
1
i2 Bi 2 Bij Wi1W j Vi1V j
S A X A B A
1
Xi Bi1A i11Xi Bi 2 1
i
21 i2 i i1 i1 i1 Bi1
S A
i
22 i2 Xi Bi1A i11Xi Bi 2 X B A X A
1
i i1
1
i1 i i2 Bi 2 X i e λ i k0 Li
i is the layer number.
j is either 1 or 2 depending on which external medium is being referenced.
Lecture 5a Slide 9
Scattering Matrices in the Literature
For some reason, the computational electromagnetics community has:
(1) deviated from convention, and (2) formulated inefficient scattering
matrices.
ci1 S11 S12 ci S 22 transmission
ci S 21 S 22 ci 1
S12 reflection S 21
5
12/4/2017
Limitation of Conventional S‐Matrix Formulation
Note that the elements of a scattering matrix are a function of
materials outside of the layer.
This makes it difficult to interchange scattering matrices arbitrarily.
For example, there are only three unique layers in the multilayer
structure below, yet 20 separate computations of scattering
matrices are needed.
Three unique layers
20 layer stack
Lecture 5a Slide 11
Solution
To get around this, we will surround each layer with external regions of
zero thickness. This lets us connect the scattering matrices in any
order because they all calculate fields that exist outside of the layers in
the same medium. This will have no effect electromagnetically as long
as we make the external regions have zero thickness between layers.
Li
Lecture 5a Slide 12
6
12/4/2017
Visualization of the Technique
We calculate the scattering matrices for just the unique layers.
Three unique layers
Then we just manipulate these same three scattering matrices to
“build” the global scattering matrix.
Gaps between the layers are made to have zero thickness so
they have no effect electromagnetically.
Revised Geometry of a Single Layer
same medium
ψ1 ψ i 0 ψ i k0 Li ψ 2
ψ i zi
r,g r,g
ψ1 ψ i 0 ψ i k0 Li ψ 2
r,g ψ i zi r,g
Lecture 5a Slide 14
7
12/4/2017
Calculating Revised Scattering Matrices
The scattering matrix Si of the ith
layer is still defined as:
c1 i c1
S i i
S12
S S i 11i
c2 c2 S 21 S 22i
But the equations to calculate the
elements reduce to Si
X B A X A B
i 1
S11 A i Xi Bi A i1Xi Bi i i
1
i i i i A i Wi1Wg Vi1Vg
S A X B X A B A B
1
i
12 i Xi Bi A i1 i i i i i
1
i i
Bi Wi1Wg Vi1Vg
S21i S12
i • Layers are symmetric so the scattering matrix
elements have redundancy.
i
S 22 S11 i
• Scattering matrix equations are simplified. X i e λ i k0 Li
• Fewer calculations.
• Less memory storage.
Lecture 5a Slide 15
Layers in TMM are Actually Four‐Port
Networks
We have written the scattering matrices as 22 block matrices.
S11 S12
S
21 S 22
For TMM, this actually expands to a 44 element scattering matrix.
s11 s12 s13 s14 Each mode provides an I/O mechanism
s24
S11 S12 s21 s22 s23 and there are two modes on each side
S in each direction.
21 S 22 s31 s32 s33 s34
s41 s42 s43 s44 1 3
s s s s14
S11 11 12 S12 13
s21 s22 s23 s24
s s s s34 2
S 21 31 32 S 22 33
s41 s42 s43 s44 4
Lecture 5a Slide 16
8
12/4/2017
Scattering Matrices of Lossless Media
If a scattering matrix is composed of materials that have no loss and
no gain, the scattering matrix must conserve power. That is, all
incident power must either reflect or transmit.
This implies that the scattering matrix is unitary.
If the scattering matrix is unitary, it must obey the following rules:
S H S 1
S H S SS H S 1S SS 1 I
Note: If the regions external to the layer are different from each other, the scattering
matrices will not be unitary. This is because the field amplitudes will be different even
though the field carries the same amount of power.
Lecture 5a Slide 17
Hints About Stability in These Formulations
S11 S12
S
S 21 S 22
• X describes propagation through an entire layer.
Don’t divide by X or your code can become
unstable.
Lecture 5a Slide 18
9
12/4/2017
Multilayer Structures
Lecture 5a Slide 19
Solution Using Scattering Matrices
The scattering matrix method consists of working through the device
one layer at a time and calculating an overall scattering matrix.
S 1 S 2 S 3 S 4 S 5
S device S1 S 2 S 3 S 4 S 5
Redheffer star product.
NOT matrix multiplication!
Lecture 5a Slide 20
10
12/4/2017
Derivation of the Redheffer Star Product
We start with the equations for the two adjacent scattering matrices.
A A B B
c1 S11 S12 c1 c2 S11 S12 c2
A B
c2 S 21 S22 c2 c3 S 21 S 22 c3
A B
We expand these into four matrix equations.
A
c1 S11 A
c1 S12 c2 Eq. 1 B
c2 S11 B
c2 S12 c3 Eq. 3
A A
c2 S 21 c1 S 22 c2
Eq. 2 B B
c3 S 21 c2 S 22 c3
Eq. 4
We substitute Eq. (2) into Eq. (3) to get an equation with only . c2
We substitute Eq. (3) into Eq. (2) to get an equation with only . c2
I S S c
B
11
A
22 2
B A
S11 B
S 21 c1 S12 c3 Eq. 5
I S S c
A
22
B
11 2
A
1
A B
S 21 c S 22 S12 c 3
Eq. 6
c c 2
2
We eliminate and by substituting these equations into Eq. (1)
and (4). We then rearrange terms into the form of a scattering matrix.
c1 ? ? c1 Overall, this is just algebra. We start with 4 equations and 6
c3 ? ? c3 unknowns and reduce it to 2 equations with 4 unknowns.
Lecture 5a Slide 21
Redheffer Star Product
Two scattering matrices may be combined into a single scattering
matrix using Redheffer’s star product.
S A A
S12 S B B
S12
S AB
S A
S B S A 11A S B 11B
S 21 S22A
S 21 S22B
The combined scattering matrix is then
1
AB
A A
B A B A
S11 S11 S12 I S11 S 22 S11 S 21
S11 AB AB 1
S12 AB A
S12 B A B
S12 I S11 S 22 S12
AB
S AB
S 21 S22AB 1
S21 S21 I S22 S11
AB B A B A
S 21
1
S22AB S22B S21B I S 22A S11
B A B
S 22 S12
R. Redheffer, “Difference equations and functional equations in transmission-line theory,”
Modern Mathematics for the Engineer, Vol. 12, pp. 282-337, McGraw-Hill, New York, 1961.
Lecture 5a Slide 22
11
12/4/2017
Putting it All Together (1 of 2)
First, we calculate the device scattering matrix by iterating through
each layer of the device and combining the scattering matrices using
the Redheffer star product.
S
3
S S
2
S
N
1
L1 L2 L3
LN
0 0
S device S 1 S 2 S 3 S N
Lecture 5a Slide 23
Putting it All Together (2 of 2)
Second, we must connect the device scattering matrix to the external
regions to get the global scattering matrix. We use connection
scattering matrices to do this.
S
trn
S ref
S
3
S S
2
S
N
1
0 L1 L2 L3
LN 0
0 0 0 0
S global
S
ref
S S S S S
1 2 3 N trn
Lecture 5a Slide 24
12
12/4/2017
Reflection/Transmission Side Scattering Matrices
The reflection‐side scattering matrix is
s ref
ref 1
S11 A B ref ref r ,I r,g
ref 1
A ref Wg1Wref Vg1Vref r ,I r,g
S12 2 A ref 1 1
B ref W Wref V Vref
S 21ref 0.5 A ref B ref A ref1 B ref g g
The transmission‐side scattering matrix is
s trn
B trn A trn1
trn
S11 r,g r ,II
1 1
S12 0.5 A trn B trn A trn1 B trn
trn A trn W Wtrn V Vtrn
g g r,g r ,II
1 1
B trn W Wtrn V Vtrn
S21 2A trn1
trn g g
Lecture 5a Slide 25
Summary of Using Scattering Matrices
S ref S device S
trn
S
3
S S 2 S N
1
0 L1 L2 L3
LN 0
0 0 0 0
Device in gap medium
13
12/4/2017
Longitudinally Periodic
Devices
Lecture 5a Slide 27
Longitudinally Periodic Devices
Suppose we just calculated the scattering matrix for the unit cell of a
longitudinally periodic device.
Unit Cell
S
1
S S S
A B C
ABC
There exists a very efficient way of calculating the global scattering
matrix of a longitudinally periodic device without calculating and
combining all the individual scattering matrices.
S
8
S S S S S S S S S S S S S S S S S S S S S S S S
A B C A B C A B C A B C A B C A B C A B C A B C
S
8
S
1
S
1
S
1
S
1
S
1
S
1
S
1
S
1
Both are inefficient!!!
Lecture 5a Slide 28
14
12/4/2017
Cascading and Doubling
We can quickly build an overall scattering matrix that describes
hundreds and thousands of unit cells.
We start by calculating the scattering matrix for a single unit cell.
S 1 S A S B S C A BC
Next, we keep connecting the scattering matrix to itself to keep
doubling the number of unit cells it describes.
S 4 S 2 S 2
S 8 S 4 S 4
and so on… What if you have a non‐integer
power of 2 number of layers?
Lecture 5a Slide 29
Block Diagram for Modified Cascading and
Doubling Algorithm
Convert N to binary
Inputs
1
10110
S S-matrix of one unit cell
N Number of times to repeat unit cell
Initialize Algorithm
0 I
S global S bin S1
I 0
Output
starting with the least significant digit.
Done?
S global
Loop through all binary digits
no
yes Update S(N)
digit = 1? global
S
global
S
bin
S
no
Perform Doubling
S bin S bin S bin
Lecture 5a Slide 30
15
12/4/2017
Example of Cascading and Doubling Algorithm
Step 0 – Calculate scattering matrix for one unit cell
S1 S A S B SC ABC
Inputs to algorithm:
S 1 scattering matrix for a single unit cell
N 22 number of unit cells to combine
Example of Cascading and Doubling Algorithm
2’s digit = 1 Update S(global) S global S global S bin S(global) now encompasses 2 unit cells
4’s digit = 1 Update S(global) S global S global S bin S(global) now encompasses 6 unit cells
16’s digit = 1 Update S(global) S global S global S bin S(global) now encompasses 22 unit cells
Oops! This algorithm performs one unnecessary doubling operation.
Lecture 5a
How can we fix this? Slide 32
16
12/4/2017
Dispersion Analysis
Lecture 5a Slide 33
Dispersion Analysis (1 of 2)
An overall scattering matrix is calculated that describes the unit cell.
uc uc
c0 S11 S12 c0
uc S uc same as S1 on previous slide
c N 1 S 21 S22uc c N 1
The terms are rearranged in “almost” the form of a transfer matrix.
0 S12 uc uc
c N 1 S11 I c0
uc
uc
I S 22 c N 1 S 21 0 c0
If the device is infinitely periodic in the z direction, then the
following periodic boundary condition must hold.
Lecture 5a Slide 34
17
12/4/2017
Dispersion Analysis (2 of 2)
We substitute the periodic boundary condition into our rearranged
equation to get
uc
S11 uc
I c0 jk z z
0 S12 c 0
uc e uc
S 21 0 c0 I S 22 c0
This is a generalized eigen‐value problem.
S uc I c
A 11uc x 0
S 21 0 c 0
Ax Bx
0 S12 uc
B
uc
e jk z z
I S 22
[V,D] = eig(A,B);
Eigen vectors Eigen values
Bloch modes kz’s
Lecture 5a Slide 35
Who Cares?
Given kz, we can
1. Calculate the effective properties of the unit cell.
k z r,eff r,eff
This is an over
simplification and beyond
the scope of this course.
2. Construct band diagrams.
Lecture 5a Slide 36
18
12/4/2017
Alternatives to
Scattering Matrices
Lecture 5a Slide 37
Transmittance Matrices (T‐Matrices)
The T‐matrix method is the transfer matrix method where forward
and backward waves are distinguished.
c trn T11 T12 cinc
left
c right T T
inc 21 22 c ref
Benefits
• Much faster (5 to 10 times)
• Unconditionally stable
Drawbacks
• Less memory efficient
• Cannot exploit longitudinal
periodicity
• Less popular in the literature
M. G. Moharam, Drew A. Pommet, Eric B. Grann, “Stable implementation of the rigorous coupled-wave analysis for surface-
relief gratings: enhanced transmittance matrix approach,” J. Opt. Soc. Am. A, Vol. 12, No. 5, pp. 1077-1086, 1995.
Lecture 5a Slide 38
19
12/4/2017
Hybrid Matrices (H‐Matrices)
The h‐matrix method is borrowed from electrical two‐port networks.
V1 V1
h11 h12
V1 h11 h12 I1 I2 V2
I h
V2 0 I1 0
2 21 h22 V2 I
h21 2
I
h22 2
I1 V2 0
V2 I1 0
In the framework of fields, the h‐matrix is defined as
Ex ,i 1 H x ,i 1 Claimed Benefits
E i i • Improved numerical stability
y ,i 1 H11 H12
H y ,i 1 • More concise formulation
H x ,i H i H 22i Ex ,i • Simpler to implement
21 • Improved numerical efficiency
H y ,i E y ,i (30% better than ETM)
• Unconditionally stable
Eng L. Tan, “Hybrid-matrix algorithm for rigorous coupled-wave analysis of
multilayered diffraction gratings,” J. Mod. Opt., Vol. 53, No. 4, pp. 417-428, 2006.
Lecture 5a Slide 39
R‐Matrices
The R‐matrix method is essentially the impedance matrix framework
borrowed from electrical two‐port networks.
V1 V1
z11 z12
V1 z11 z12 I1 I1 I2
V z
I2 0 I1 0
2 21 z22 I 2 z21
V2
z22
V2
I1 I2 0
I2 I1 0
In the framework of fields, the h‐matrix is defined as
Ex ,i 1 H x ,i 1 Claimed Benefits
E i i
y ,i 1 R11 R12 H y ,i 1 • Unconditionally stable
• Improved numerical efficiency
E x ,i R i R 22i H x ,i
21
E y ,i H y ,i
Lifeng Li, “Bremmer series, R-matrix propagation algorithm, and numerical modeling of
diffraction gratings,” J. Opt. Soc. Am. A, Vol. 11, No. 11, pp. 2829-2836, 1994.
Lecture 5a Slide 40
20