Periodic Structures

EE625

Periodic Structures and Floquets Theorem

Periodic Structures o Repeated geometry, defined by a unit cell and a uniform periodic spacing Applications o Periodic Array antennas o High-impedance surfaces o Frequency Selective Surfaces (FSS) o Meta-Materials Artificial Materials Left-handed materials Chiral materials Frequency selective material behavior

Prof. S. Gedney

1/37

Periodic Structures

EE625

1D Periodic Surface
Consider a 1D periodic surface (in x) o Structure is infinite in x and periodic o Period d Let u(x) represent a field reacting with the x x+d x+2d periodic surface o Geometric periodicity forces the field to be periodic u ( x + d ) = Cu ( x)

x+3d

x+4d

u ( x + 2d ) = Cu ( x + d ) u ( x + 3d ) = Cu ( x + 2d ) u ( x + nd ) = Cu ( x + ( n 1)d ) C = a complex constant o More generally: u ( x + nd ) = C nu ( x)

o For boundedness, C 1 o In general: C = e + jkd , k = complex constant
Prof. S. Gedney

2/37

Periodic Structures

EE625

Periodic Function
We can define a periodic function P ( x ) , where P ( x ) = e jkx u ( x )

Consequently P ( x + d ) = e jk ( x + d )u ( x + d ) = e jk ( x + d )Cu ( x ) = e jkx e jkd ( e jkd ) u ( x)

= e jkx u ( x ) = P ( x ) Similarly: P ( x + nd ) = P ( x )

P ( x ) is a periodic function in x, with period d

o P ( x ) has the same period as the geometry

o Note that the periodic phase shift and attenuation is normalized out in P ( x ) . o Since P ( x ) is periodic in x, we can express it via a Fourier Series expansion:

P ( x) =

n =

pe
n

2 n x d

Prof. S. Gedney

3/37

Periodic Structures

EE625

P ( x) =

o Substituting: P ( x ) = e jkx u ( x ) , then

n =

pe
n jkx n

2 n x d

u ( x) = u ( x) =
o where

n =

pe

2 n x d

n =

pn e

jk xn x

2 n d o This represents a harmonic expansion of the field u(x). o Each term represents a spatial Floquet Harmonic Infinite series All terms propagate along the periodic axis (forward and backward waves) o Based on Floquets theorem, any planar periodic function can be expanded as an infinite superposition of Floquet harmonics. Here it is presented for a 1D periodic structure. It can be easily generalized to higher dimensions with more complex periodicities.
k xn = k +
Prof. S. Gedney

4/37

Periodic Structures

EE625

1D Periodic Strip Grating

y w x d

Ezinc

Consider a TMz polarized plane wave incident on a co-planar periodic grating of metallic strips. The strips have a width w. The grating has a period d. Incident Electric field:
E
inc z

( x, y ) = e

jk r

=e

jk0 x cos inc + y sin inc

o inc = the angle of the wave off the x-axis. Following Floquets theorem, we can express the incident plane wave as a periodic function in d

( E ( x + d, y) = e
inc z

jk0 ( x + d )cos inc + y sin inc

inc

) = e jk d cos e jk ( x cos
inc o 0

inc

+ y sin inc

= Ezinc ( x, y ) e jko d cos

Prof. S. Gedney

5/37

Periodic Structures

EE625
inc

Therefore, the complex constant C = e jko d cos . o and k = ko cos inc

Our objective is to compute the field scattered by the periodic strip grating. o Pose the scattered field as a function of Floquet Harmonics: E
scat z

( x, y ) =

n =

an ( y ) e

jk xn x

o Each harmonic must satisfy the wave equation o The sum of harmonics is subject to the boundary conditions on the metallic strip surfaces Eztot ( x,0 ) =0
nd x nd + w

Off the PEC surface, Ez and H x are continuous o The coefficient of each harmonic is a function if y due to the boundary condition at y = 0. o The coefficients are independent of z since the source and the geometry are independent of z. The scattered field must satisfy the wave equation: 2 2 2 scat 2 scat Ez + k0 Ez = 0, + 2 + k02 Ezscat = 0 2 y x
Prof. S. Gedney

6/37

Periodic Structures

EE625

Therefore, we expect an ( y ) to be of the form: an ( y ) = cn e

jk yn y

{() y > 0, (+) y < 0}

Plugging this into the wave equation: 2 jk y jk x 2 + 2 + k02 cn e yn e xn = 0 2 x y 2 k02 k x2n k yn = 0

k yn = k02 k x2n
This is referred to as the dispersion relationship. Therefore, E
scat z

( x, y ) = cn e jk
n =

xn x

jk yn y

Thus, the scattered field is expanded via a plane wave expansion of upward and downward traveling waves. The coefficients cn are yet to be determined. o These represent the complex amplitudes of each harmonic, are will be based on the boundary conditions on the strip surface o We will determine these coefficients via a method of moment procedure.
Prof. S. Gedney

7/37

Periodic Structures

EE625

Note that the angular spectrum of the Floquet Harmonic expansion is independent of the coefficients cn . Recall that: 2 n 2 n = ko cos inc + k xn = k + d d Therefore,

2 n k yn = k ko cos inc + d Since n varies from to +, k xn varies dramatically between and + .

2 2 0

Define:

k2 k2 , k02 k x2n xn 0 k yn = j k x2n k02 , k02 k x2n Assuming k0 is real, then k yn is defined by a finite set of purely propagating modes, and an infinite set of evanescent modes.

Prof. S. Gedney

8/37

Periodic Structures

EE625

Note that multiple propagating modes can be scattered off the periodic surface. The angle of the propagating modes are defined as:
n inc k xn = ko cos + , d / 0
s

k yn = k 0
1

n inc 1 cos + d / 0

Examples: o inc = 30 d
0.50
0.53590

k yn = tan kx n

n
-1 0 1 -1 0 1

k xn / k0
-1.134 0.866 2.866 -0.192 0.866 2.732

k yn / k 0
0.535j 0.5 2.686j 0.003354 0.5 2.542j

nscat = tan 1 ( k y / k x
n

Evanescent 30o Evanescent 179.8o 30 o Evanescent

The scattered angle is off the negative x-axis

Prof. S. Gedney

9/37

Periodic Structures

EE625

d
1.00

n
-2 -1 0 1 -4 -3 -2 -1 0 1

k xn / k0
-1.134 -0.134 0.866 1.866 -1.134 -0.634 -0.134 0.366 0.866 1.366

k yn / k 0
0.535j 0.991 0.5 1.575 j 0.535 j 0.773 0.991 0.931 0.5 -0.931 j

nscat = tan 1 ( k y / k x
n

2.00

Evanescent 97.7o 30 o Evanescent Evanescent -129.3o -97.7 o 68.6o 30o Evanescent

Prof. S. Gedney

10/37

Periodic Structures

EE625

d = 0.50

y incident
m=0

d = 0.53590 y

m=0

incident
m = 1

x
m=0
m=0

m = 1

d = 1.00

y
m = 1

d = 2.00

incident
m = 1

y
m = 2 incident m = 3

x
m=0
m=0

x
m = 3 m = 2

m = 1

m = 1

Prof. S. Gedney

11/37

Periodic Structures

EE625

Observations o For very small cells ( < 0 / 2 ), only the specular reflection (n = 0) will be non-evanescent. o As the unit cell size is increased, new propagating modes will appear. These modes will begin propagating along the +ve x-axis, and as frequency increase, will migrate towards the specular reflection. Note that inc also impacts the frequency at which higher-order harmonics will begin to propagate, and the reflection angle. o The harmonics propagate off at discrete angles Practical application: Diffraction Grating Frequency selective surface Dichroic reflector antenna o Evanescent modes are surface wave modes Propagate along the surface of the structure Purely attenuate away from the surface

Prof. S. Gedney

12/37

Periodic Structures

EE625

1D Periodic Green Function

The total electric field must satisfy the boundary condition that Eztot on the strip surface = 0. o Can also be expressed that Ezscat = Ezinc on the strip surface Ezscat is represented by an infinite summation of Floquet harmonics. o We need to solve for the amplitudes of each harmonic to satisfy the boundary condition. o This can be done in the same way as performing a Fourier series expansion of a function. o The difficulty is the function is a rectangular pulse, which needs an infinite number of harmonics to converge. An alternate approach is to derive a Green function for a 1D periodic line source, and then pursue a method of moment solution using the EFIE. o This is referred to as the periodic Green function (PGF).

Prof. S. Gedney

13/37

Periodic Structures

EE625

Periodic Green Function

Consider an infinite 1D array of electric line sources that are illuminated by a plane wave o The line sources are effectively radiating a scattered field due the plane wave illuminating the periodic structure. o Floquets theorem states that all observables will have the same periodicity as the structure, and have a cell-to-cell phase shift equal to that of the source. o The phase shift will be k0 d cos inc This leads to an array of infinite line sources, periodically placed with spacing inc d, and with phase shift: e jn = e jnk0d cos The periodic Green function (PGF) satisfies the wave equation:

G p + k G p = ( x nd ) e jn
2 2 0
n =

We can apply superposition, and solve for one mode at a time: 1 2 (2 Gp ( r ) = H 0 ) k0 ( x nd ) + y 2 e jn 4 j n=

Prof. S. Gedney

14/37

Periodic Structures

EE625

EFIE Solution for the Scattering by a 1D Periodic Strip Array

Now, consider the scattering by a 1D periodic array of PEC strips co-planar in the y = 0 plane. y Ezinc The EFIE is formulated as: Ezscat ( x,0 ) = Ezinc ( x,0 )
nd x nd + w

w x
nd x nd + w

where,

E
w

scat z

= jk00 Az

Az = J z ( x ) G p ( x, y | x ) dx 1 2 (2 G p ( x, y | x ) = H 0 ) k0 ( x x nd ) + y 2 e jn 4 j n= Note that the integration is only over 1 cell. This cell is referred to as the unit cell. The current from the unit cell is repeated by the PGF. Namely, the PGF effectively radiates an infinite number of currents from all the periodic cells.
0

Prof. S. Gedney

15/37

Periodic Structures

EE625

Method of Moment Discretization

The unit cell strip alone is discretized into N linear segments. The current density can be expanded using pulse basis functions over each linear segment: J z i Pi ( x; xi , xi +1 )
i =1 N

o Only the current over the unit cell must be expanded o The PGF effectively replicates each pulse basis function as an infinite number that are periodically displaced o The PGF also applies the correct phase shift to every periodically displaced basis function. It is sufficient to use delta test functions o The test functions are located at the cell centers T j = x x cj

Prof. S. Gedney

16/37

Periodic Structures

EE625
2 nd ) + 02 e jn dx (x x

The EFIE becomes: xi +1 k00 N ( inc c i H 0 2 ) k0 Ez ( x j ) = 4 i =1 xi n= or,

k00 N inc c Ez ( x j ) = i n 4 i =1 =

c j

xi +1

xi

( H0

2)

(k

x c x nd e jn dx j

This leads to a linear system of equations: e inc = Z where, einc = Ezinc ( x c ) j j Z j ,i k = 0 0 4

n =

e jn

xi +1

xi

( H0

2)

(k

x c x nd dx j

= k0 d cos inc

Prof. S. Gedney

17/37

Periodic Structures

EE625

Z j ,i

k = 0 0 4

n =

e jn

xi +1

xi

( H0

2)

(k

x c x nd dx j

Observation: o Every impedance element requires numerical integration and an infinite summation o The convergence of this series is extremely slow: 1 j2 jk0 n d 2 lim H 0 ( k0 x x nd ) e = O nd n k0 n d o We need to accelerate the rate of convergence. o There are a number of methods to do this. We will look at one method based on using the Spectral Periodic Green function

Prof. S. Gedney

18/37

Periodic Structures

EE625

Spectral Periodic Green Function

Introduce the Fourier transform pair:

F ( f ( x )) = f ( kx ) =
1

f ( x ) e jk x x dx

1 F f ( kx ) = f ( x ) = f ( k x ) e jk x x dk x 2 Next, consider again the infinite line source array:

J z ( x) =

We can perform the Fourier transform of J z ( x ) :

n =

( x nd ) e

jn

J z ( kx ) =
=

n =

( x nd ) e jn e jk x dx
x

n =

jn

e jk x nd

We can prove that this is is equivalent to the Fourier series expansion: 2 2 n k x d d = n e jn e jkxnd d n= =
Prof. S. Gedney

19/37

Periodic Structures

EE625

Proof

2 d
2 d

2 n kx d d n =

jk x nd = an e n=

2 d

2 n kx d d n =
2 0 + d d md

jk x md = e

2 d

ae
0 n =

jk x nd

e jk x md dx

2 j e d

2 am d
= e jn e jk x nd Q.E.D. n=

am = e jm 2 2 n kx d n= d d Therefore, 2 J z ( kx ) = d

n =

kx

2 n d d

Prof. S. Gedney

20/37

Periodic Structures

EE625

Now, we can return to the spatial domain via the inverse Fourier transform: 2 1 2 n jkx x 1 e J z ( x ) = F J z ( kx ) = kx d n 2 kx = d d =

1 j = e d n=

2 n + d d

1 jk xn x = e d n= Finally: Jz ( x) =
n =

( x nd ) e

jn

1 jk xn x = e d n=

Observations: o The infinite summation of periodically displaced line sources is equal to an infinite summation of current sheets. o Note that: Each current sheet physically radiates from the y = 0 plane The effective wave number of the current sheets are the Floquet wave numbers Each current sheet is linearly independent The field radiated by each current sheet is a plane wave!
Prof. S. Gedney

21/37

Periodic Structures

EE625

Finding the radiated waves:

H x+ = H + e

jk xn x jk y y

J zn =

1 jk xn x e d
H x = H e
jk xn x

jk y y

Prof. S. Gedney

22/37

Periodic Structures

EE625

Finding the radiated waves:

y H x+ = H + e
jk xn x jk y y

J zn =

1 jk xn x e d
H x = H e
jk xn x

jk y y

Boundary Condition:

y H + H = zJ zn

1 jk xn x jk yn y , y>0 2d e e H xn = 1 e jkxn x e jk yn y , y<0 2d From the dispersion relationship: k yn = k02 k x2n

Prof. S. Gedney

23/37

Periodic Structures

EE625

Normal H can be found via Gauss law: Hn = 0

kn H n = 0 k xn H xn k yn H
yn

=0

H =

yn

k xn k yn

H xn

We can also derive the Magnetic Vector Potential due to the plane wave: Azn Azn H n = An = x y = xH xn + yH yn y x Given, H xn , H yn from above:
1 jk x jk y e xn e yn j 2dk yn Finally, the periodic Green function is derived from the summation of all harmonics of the vector potential: jk x jk y e xn e yn G p ( x, y ) = j 2dk yn n = Azn =
Prof. S. Gedney

24/37

Periodic Structures

EE625

Observations o The spectral domain PGF is a series of Floquet harmonics! o Each harmonic has an amplitude proportional to 1/ k yn 2 n o Asymptotically, lim k yn j n d
o lim G pn
n

2 n 2 n x y d d

4 n o Therefore, the series is converging as 1/n, rather than the spatial PGF, which converges as 1/ n . o Note, this is a dramatic improvement, however, it is still quite slow! o Fortunately, the convergence can be accelerated This will happen naturally when convolving the PGF with the currents A number of techniques have been introduced to accelerate the convergence even more.

Prof. S. Gedney

25/37

Periodic Structures

EE625

EFIE Solution for the Scattering by a 1D Periodic Strip Array Solution with the Spectral Green function
We can return to the MoM solution of the EFIE for the 1D strip array.
y

Ezinc
w x

The EFIE is formulated as: Ezscat ( x,0 ) = Ezinc ( x,0 )

nd x nd + w

nd x nd + w

where,
w

Ezscat = jk00 Az
Az = J z ( x ) G p ( x, y | x ) dx
0

G p ( x, y | x ) =

n =

jk xn ( x x ) jk yn y

e j 2dk yn

Prof. S. Gedney

26/37

Periodic Structures

EE625

Again, we assume point basis functions and point test functions The EFIE becomes:

E or,

inc z

( x ) = jk
N c j 0 0 i =1 i N xi +1 c j

xi +1

jk xn x c x j

xi n =

j 2dk yn e
jk xn x c x j

dx

k E ( x ) = 0 0 i 2d i =1 n =
inc z

xi

k yn

dx

This leads to a linear system of equations: e inc = Z where, einc = Ezinc ( x cj ) j

Z j ,i

k = 0 0 2d

n = xi

xi +1

jk xn x c x j

k yn

dx

This integration can be computed analytically! k xn i i sinc 2 jk xn ( xcj xic ) k Z j ,i = 0 0 e k yn 2d n=

Prof. S. Gedney

27/37

Periodic Structures

EE625

Z j ,i

k = 0 0 2d

n =

k xn i i sinc 2 jk xn ( xcj xic ) e k yn

Observation: o The sinc function is effectively the Fourier-transform of the pulse basis function. o The result is effectively the product of the spectral PGF harmonic and the Fourier harmonic of the pulse basis function o Due to the properties of the sinc function, the series now has an asymptotic convergence of O (1/ n 2 ) , which is much improved.

What if we used a smoother basis how would this impact the convergence? o Due to the product in the spectral domain, the spectral response of the current basis effectively filters the spectral Green function.

Prof. S. Gedney

28/37

Periodic Structures

EE625

Computing the Scattered Fields

The method of moment solution is used to compute the unknown coefficients of the current basis functions. Given the solution, we can approximate the currents, and consequently the scattered field: xi +1 jk ( x x ) jk yn y N e xn e Ezscat = jk00 Az = jk00 i dx j 2dk yn i =1 xi n =
k = 0 o 2d where,
xi +1

N xi+1 jkx x e jk xn x e jk yn y i e n dx k n = i =1 yn xi

xi

jk xn x

dx = i e

jk xn xic

sinc k xn i 2

Therefore,

scat z

k0o = n = 2dk yn

i =1

i i

jk xn xic

jkx x jk y y i sinc k xn e n e n 2

=
Prof. S. Gedney

n =

cne

jk xn x jk yn y

29/37

Periodic Structures

EE625

The scattered electric field is thus expanded via a series of Floquet harmonics: E With amplitudes:
scat z

n =

cne
i

jk xn x jk yn y

cn = And with:

k0o 2d

i
i =1

e k yn

jk xn xic

sinc k xn i 2
2

n n inc inc k xn = ko cos + , k yn = k0 1 cos + d / 0 d / 0 Therefore, the scattered field are the Floquet harmonics expected, and cn determine the amplitude of each mode. Note that each mode carries real power away from the grating at the Floquet angles determined by inc and d / 0 Evanescent modes have purely imaginary power along the y-direction

Prof. S. Gedney

30/37

Periodic Structures

EE625

Power Spectrum and Conservation of Power

Only propagating modes will carry power away from the plane of the strip - that is along the normal projection. Compute the power density relative to the normal direction in the back scatter region (y > 0) for propagating modes: 1 1 1 Pybn = Re ( Ezn H xn ) = Re Ezn Ez 2 2 jo y n 2 1 jk xn x jk yn y k yn jkxn x + jk yn y 1 cn cn e e = Re cn e e = 2 sin (n ) 2 koo o where k yn = ko sin (n ) , and n is the angle off the positive x-axis. In the forward scattering region for n 0 propagating modes: 1 1 1 f Pyn = Re ( Ezn H xn ) = Re Ezn E zn 2 2 jo y 2 1 jk xn x jk yn y k yn * jk xn x jk yn y 1 cn cn e e = Re cn e e = 2 sin (n ) 2 koo o
Prof. S. Gedney

31/37

Periodic Structures

EE625

Note that the forward scattered power is negative, reflecting the power flow along the negative y-direction. When n = 0 the diffracted field (or transmitted field) is actually a superposition of the incident wave and the Floquet mode: tot 1 tot 1 1 Pyf0 = Re ( Eztot H xp ) = Re Ez 0 Ez 0 0 2 2 jo y 2 i i i 1 1 cn jk x x jk i y + jk i y k y sin ( i ) = Re ( c0 + Eo ) e jk x x e y ( c0 + Eo ) e e y = 2 2 o koo To understand this, consider the case where the periodic structure reduces to a plane. Both the reflected and the transmitted Floquet harmonics have an amplitude of -1. However, the transmitted field is a superposition of the Floquet mode and the incident field and consequently has an amplitude of 0. The incident power is simply 2 1 1 Eo inc Pyinc = Re Ezinc H x = sin ( i ) 2 2 o

Prof. S. Gedney

32/37

Periodic Structures

EE625

A power check simply requires that the sum of the powers carried in the Floquet modes should sum to the incident power. Or, more appropriately:
n = P 1

P2

Pyn
f

P where P and P2 reflect the bounds of the propagating modes, and the minus sign 1 accounts for the direction of power flow. Note that since Pyinc is negative, we the quotient above will be positive Note that a power check is a necessary, but not sufficient condition, that the solution is correct. We can also define the normalized powers of each mode: f Pyn Pybn f Pyn = i , Pybn = i Py Py Then, for power conservation:
n = P 1

n = P 1 inc y

P2

Pybn

=1

P2

yn

n = P 1

P2

b yn

=1

Prof. S. Gedney

33/37

Periodic Structures

EE625

Examples
y w x d

Ezinc

d = 0.8 0 ,

w = 0.4 0 , inc = 30

50 segments (way over sampled!)

Prof. S. Gedney

34/37

Periodic Structures

EE625

Refraction angles (off ve x-axis): 0 = 30

1 = 67.42
Floquet Amplitudes: e1 = 0.14511 0.2285 j
e0 = 0.79076 + 0.18673 j

Powers: Pyinc = 6.63605 104 W/m 2

Pyb1 = 8.66667 105 W/m 2
f Py1 = 8.66667 105 W/m 2

Pyb0 = 4.3809110 4 W/m 2

f Py0 = 5.21917 105 W/m 2
P2 P2 f b Power check: Pyn Pyn / Pyinc = 1.00002 n = P n = P1 1

Prof. S. Gedney

35/37

Periodic Structures

EE625

Magnitude of the Current Distribution

Prof. S. Gedney

36/37

Periodic Structures

y w

Ezinc
x d

EE625

d = 0.5 1.2 0 ,

w = d / 2 , inc = 30 , 25 segments

f f Pyb0 Py0 Py1/ b , i, i i Py Py Py

f Py2/ b

Pyi

Prof. S. Gedney

37/37