Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
6/10/2015
1
Pioneering 21st Century Electromagnetics and Photonics
Module 3 –Spatially-Variant Lattices
Course Website:http://emlab.utep.edu/scSVL.htm
EM Lab Website:http://emlab.utep.edu/
Short CourseLecture Outline
• Fourier decomposition of lattices
• Rotation matrices
• Algorithm for synthesis of spatially-variant lattices
• Improving efficiency of the algorithm
• Implementation tips and tricks
Slide 2SVL Short Course -- Module 3
6/10/2015
2
Fourier Decomposition
of Lattices
Short Course
Recall Complex Fourier Series
SVL Short Course -- Module 3 Slide 4
Periodic functions can be expanded into a Fourier series.
For 1D periodic functions, this is
22 2
2
1
px pxj j
p pp
f x a e a f x e dx
For 2D periodic functions, this is
2 2 2 2
, ,
1, ,x y x y
px qy px qyj j
p q p qp q A
f x y a e a f x y e dAA
For 3D periodic functions, this is
2 2 2 2 2 2
, , , ,
1, , , ,x y z x y z
px qy rz px qy rzj j
p q r p q rp q r V
f x y z a e a f x y z e dVV
6/10/2015
3
Short Course
Fourier Expansion in Terms of Grating Vectors
SVL Short Course -- Module 3 Slide 5
2 2
,, x y
px qyj
p qp q
f x y a e
, , ,,, ,x yj K p q x K p q y jK p q r
p qp q p q
f x y a e a p q e
2,
2,
xx
yy
pK p q
qK p q
,K p q
p
q
Short Course
Visualize the Terms
SVL Short Course -- Module 3 Slide 6
,, , jK p q r
p q
f x y a p q e
p
q
p
q
6/10/2015
4
Short Course
Lattices Can be Decomposed into a Set of Planar Gratings
7SVL Short Course -- Module 3
p
q
Using the concept of the complex Fourier series, we decompose the unit cell into a set of planar gratings. Each planar grating is described by a grating vector and a complex amplitude apq.pqK
Short Course
Two Parts to the Decomposition
8SVL Short Course -- Module 3
,, , jK p q r
p q
f x y a p q e
Input
2 2ˆ ˆ,
x y
p qK p q x y
2D, FFT ,a p q f x y
Calculated using FFT. Calculated analytically.
p
q
Fourier Coefficients
Grating Vectors
p
q
x
y
6/10/2015
5
Short Course
The Grating Amplitudes
9SVL Short Course -- Module 3
p
q
2D-FFT ,p qa
Short Course
Impact of Number of Planar Gratings
10
1×1
3×3
5×5
7×7
9×9
21×21
31×31
11×11
SVL Short Course -- Module 3
6/10/2015
6
Short Course
3D Decomposition
11
3D Unit Cell
3D Spatial Harmonics
ˆ ˆ ˆ
2 integer
2 integer
2 integer
x y z
xx
yy
zz
K K x K y K z
pK p
qK q
rK r
SVL Short Course -- Module 3
RotationMatrices
6/10/2015
7
Short Course
Definition of a Rotation Matrix
a
b ab R
Rotation matrix [R] is defined as
The rotation matrix should not change the amplitude. This implies that [R] is unitary.
1H
H H
R R
R R R R
b a
I
SVL Short Course -- Module 3 13
Short Course
2D Rotation Matrix
ab
cos sin
sin cosx
R
yy
xb
b
a
a
SVL Short Course -- Module 3 14
ab R
6/10/2015
8
Short Course
3D Rotation Matrices
1 0 0
0 cos sin
0 sin cosxR
Rotation matrices for 3D coordinates can be written directly from the previous result.
cos 0 sin
0 1 0
sin 0 cosyR
cos sin 0
sin cos 0
0 0 1zR
a
b
b b
a
a
SVL Short Course -- Module 3 15
Short Course
Composite Rotation Matrix
We can combine multiple rotation matrices into a single composite rotation matrix [R].
ab R
SVL Short Course -- Module 3 16
z y xR R R R Vector rotation using the composite rotation matrix is
6/10/2015
9
Algorithm for Synthesis of
Spatially-Variant Lattices
Short Course
Step 1 – Generate Input to Algorithm
18
Unit Cell Lattice SizeUnit Cell
OrientationLattice Spacing Fill Fraction
A grayscale unit cell allows easier adjustment of fill fraction.
SVL Short Course -- Module 3
6/10/2015
10
Short Course
Step 2 – Build a Grayscale Unit Cell
SVL Short Course -- Module 3 19
A grayscale unit cell allows for better control of fill fraction in the end.
Short Course
Step 3 – Decompose Unit Cell Into a Set of Planar Gratings
20SVL Short Course -- Module 3
6/10/2015
11
Short Course
Step 4 – Truncate the Set of Planar Gratings
21
We must retain only a minimum number of spatial harmonics.
% COMPUTE 2D-FFTERF = fftshift(fft2(ER))/(Nx*Ny);
% TRUNCATE SPATIAL HARMONICSp0 = 1 + floor(Nx/2); %p position of zero-order harmonicq0 = 1 + floor(Ny/2); %q position of zero-order harmonicp1 = p0 - floor(P/2); %start of p rangep2 = p0 + floor(P/2); %end of p rangeq1 = q0 - floor(Q/2); %start of q rangeq2 = q0 + floor(Q/2); %end of q rangeERF = ERF(p1:p2,q1:q2); %truncate harmonics
ERF128×128before truncation
ER128×128 ERF7×7after truncation (P=Q=7)
ERF128×128before truncation
ER128×128 ERF7×7after truncation (P=Q=7)
SVL Short Course -- Module 3
Short Course
Step 5 – Spatially Vary Each Planar Grating Individually and Sum the Results
22
Construct spatially variant K-function
Compute grating phase on low resolution grid
Interpolate grating phase into high resolution grid
Compute spatially variant planar grating
Add planar grating to overall lattice
For each spatial harmonic…
SVL Short Course -- Module 3
6/10/2015
12
Short Course
Step 5 – For Each Gratinga) Construct K-Function
23
Unit Cell Orientation (x,y)
Lattice Spacing a(x,y)
Uniform Kpq Function
+ =
Intermediate Kpq
+ =
Intermediate Kpq K-Function
pqK
SVL Short Course -- Module 3
Short Course
Step 5 – For Each Gratingb) Solve for the Grating Phase
24
Construct Matrix Equation
,
,
x x pq
y y pq
D kA b
D k
Solve Using Least Squares
1
pq
Φ A b
Reshape Back to a 2D Grid
,pq pq x yΦ
Interpolate to a Higher Resolution Grid Using interp2()
T T A A A b A b
2, 2 2, ,pq pqx y x y
SVL Short Course -- Module 3
All of this is performed by
svlsolve(Kx,Ky,dx,dy)
or
svlsolve(Kx,Ky,Kz,dx,dy,dz)
6/10/2015
13
Short Course
Step 5 – For Each Gratingc) Sum All of the Spatially-Variant Gratings
25
Generate the pqth spatially variant grating
Add this analog planar grating to the overall analog lattice
2,exppq pq pqr a j r
analog analog pqr r r
SVL Short Course -- Module 3
Short Course
Step 6 – Incorporate Spatially-Variant Fill Fraction
26
At this point, we have the analog lattice. Set any imaginary components to zero.
cosr f r
SVL Short Course -- Module 3
1 analogbinary
2 analog
r rr
r r
analog analogRer r
From the analog lattice, we calculate the binary lattice with a spatially-variant fill fraction.
From analog lattices that have a smooth cosine looking profile, the threshold can be estimated as
6/10/2015
14
Short Course
Summary of Spatially Variant Algorithm
27
1. Define the spatial variance: orientation, lattice spacing, fill fraction,...
2. Build the baseline unit cell.3. Decompose unit cell into a set of planar gratings.4. Truncate the set of planar gratings to a minimal set.5. Loop over each spatial harmonic
a. Construct K-function that is uniform across the grid according to the grating vector of the spatial harmonic.
b. Solve for grating phase from the K-function.c. Add the planar grating to the overall lattice.
6. Incorporate spatially-variant fill fraction.
SVL Short Course -- Module 3
Improving Efficiency of the
Algorithm
6/10/2015
15
Short Course
Collinear Spatial Harmonics
SVL Short Course -- Module 3 29
If a spatial harmonic is collinear to another (i.e. their grating vectors are parallel), we can calculate its grating phase by scaling the grating phase of the other. Therefore, we only have to solve for one of them.
1K
2 1K aK
3 1K bK
1 1K
2 2 1
2 21 1
2 1
K aK
Ka a
a
3 1b
We solve this numerically.
We just scale the solution from .1K
Again, we simply scale the solution from .1K
Short Course
Identifying Collinear Planar Gratings
SVL Short Course -- Module 3 30
121 spatial harmonics 40 “unique” spatial harmonics.
The rest are collinear.
6/10/2015
16
Short Course
Performance Gain by Eliminating Collinear Gratings
SVL Short Course -- Module 3 31
69% for 2D
59% for 3D
Short Course
Eliminating Gratings According to Their Amplitude
SVL Short Course -- Module 3 32
thresholdpqa a
We neglect all planar gratings with amplitude apq less than some threshold.
threshold 0.02max pqa a A threshold that works in many cases is one that is around 2% of the maximum apq in the expansion.
6/10/2015
17
Short Course
Overall Efficiency Improvement
SVL Short Course -- Module 3 33
Truncation by Grating Amplitude
Truncation by Coplanar K
ImplementationTips and Tricks
6/10/2015
18
Short Course
Grid Strategy
35SVL Short Course -- Module 3
Short Course
Calculating the Grid Parameters
36
• High Resolution Unit Cell Grid– Need enough points for the FFT to converge.
• Low Resolution Lattice Grid– Need enough resolution to resolve the spatial variance. Usually this is N>10
grid cells per unit cell.
• High Resolution Lattice Grid– Need enough resolution to resolve the shortest period planar grating.
coarse a N
FFT a N P
minfine min
2
max pq
aa
N K
SVL Short Course -- Module 3
6/10/2015
19
Short Course
Arrays of Discontinuous Metallic Elements (1 of 2)
SVL Short Course -- Module 3
Place metallic elements at the intersections.
37
Short Course
Arrays of Discontinuous Metallic Elements (2 of 2)
SVL Short Course -- Module 3 38
When the lattice is composed of an array of discontinuous metallic elements, it is not efficient to decompose the unit cells into planar gratings.
6/10/2015
20
Short Course
Arrays of Metallic Elements Over Curved Surfaces
SVL Short Course -- Module 3 39