Beruflich Dokumente
Kultur Dokumente
To cite this article: S. Sohail H. Naqvi & N.C. Gallagher (1990): A General Solution to the
Scattering of Electromagnetic Waves from a Strip Grating, Journal of Modern Optics, 37:10,
1629-1643
To link to this article: http://dx.doi.org/10.1080/09500349014551791
1990,
VOL .
37,
NO .
10, 1 62 9-1643
S. SOHAIL H . NAQVI
Department of Electrical and Computer Engineering,
University of New Mexico, Albuquerque, NM 87131, U .S .A .
and N . C . GALLAGHER
School of Electrical Engineering, Purdue University,
West Lafayette, IN 47907, U .S .A.
(Received 10 November 1989 ; revision received and accepted
13 February 1990)
Abstract. We describe a new robust approach for the analysis of strip gratings,
both of finite and infinite conductivity, for the TE and TM cases . The field
distributions in the plane of the grating are expanded in a Fourier series, whose
coefficients are derived as the solution to an infinite-dimensional system of linear
equations . Various configurations of the scatterer are considered and it is shown
that even in cases where the Tsao-Mittra SIT procedures fails to converge and
the moment method requires a large matrix to arrive at a solution, our method
yields reasonable results even for small matrix sizes . The accuracy of the solution
procedure is analysed by considering the mean-square error in the field
magnitudes as a function of the truncation size of the infinite system of linear
equations .
1.
Introduction
The scattering of electromagnetic waves from periodic metallic structures is a
classic problem that has been looked at by numerous researchers over the past few
decades [1-8] . The geometry of the problem under consideration in this paper is
given in figure 1 . The basic assumptions are that we have infintely thin periodic strips
of metal of finite or infinite conductivity . These strips are of infinite length and are
located in the xy plane . A plane wave is obliquely incident at an angle 0 from the
normal and we wish to determine the electric field distribution everywhere . The
standard approach to the solution of this problem is to formulate the E-field or
H-field integral equations . A matrix equation is then obtained by using the method
of moments [9]. The infinite system of linear equations obtained can then be solved
using any of the techniques described in [10] . The C-G method [11] was proposed
by Hestenes and Stiefel nearly 30 years ago for the solution of a system of linear
equations . However, it is only in the last decade that this method has been applied to
the electromagnetic scattering problem [12,13] . A different approach to the solution
of this problem was proposed by Tsao and Mittra [5] . Formulating the problem in
the spectral domain they obtained a set of algebraic equations for the coefficients of
the Electric field and current distributions . These equations are solved using the
spectral iteration technique developed by the authors .
We find that both the moment method and the spectral iteration method have
`regions of operation' associated with them when applied to the perfectly conducting
0950-0340/90 $3-00 1990 Taylor & Francis Ltd .
1 63 0
Region 1
Y
r. .
Region 3
1631
2.
Problem formulation
Moharram and Gaylord in [20] discuss the formulation of the general problem of
the scattering of a plane wave from a periodic surface . The total field in region 1 can
be written in the most general form as the sum of an incident plane wave and multiple
backward diffracted orders .
00
F1 (x,y,z)=exp(-ik 1 r )+
Rexp(-ik li r ),
(1)
with harmonic exp (iwt) assumed and suppressed . Similarly for region 3 we have
only the forward diffracted orders and
F3 (x,Y, z)= Y-
(2)
where r=xil+yk'+z4 and k, 9 and 4 are unit vectors in the x, y and z directions
respectively . Thus the scattered field can be treated as a sum of plane waves with
wave vectors k l ,, . Where the incident electric field lies parallel to the metallic strips
(TE case), the field F(x, y, z) represents the electric field expansion . For the dual case
of the incident magnetic field lying parallel to the strips (TM case), the field F(x, y, z)
represents the magnetic field . In either case the only component of the field present is
in the y direction . F 1 and F 3 represent these y components .
In the limiting case of zero grating modulation, we have an infinite metallic sheet
and only the n=0 diffracted mode is present and has to be phase matched to the
incident field at the z=0 boundary . Thus
k 1 k= kio k .
Since the scattering surface is periodic, each diffracted mode must satisfy the
'Floquet condition' whereby the scattered field is also periodic in the x direction with
period d. We then have
k l k =k 1 . 4- n
(3)
Simplifying we obtain
sing=sing-
,
d
(4)
Since the waves in regions 1 and 3 are travelling in opposite directions, for the z
components of the wave vector we have
k 1 .4=-k10 . =k 30 .~~
and
k1
=-k 3i '4 .
(5)
1 63 2
k
z=0
Y
r. .
Figure 2 .
Decomposition of wave vectors, of the incident wave and the zero order reflected
and transmitted waves, into rectangular components .
Figure 2 displays graphically the result of applying the formalism of [20] to the strip
graing problem . In regions 1 and 3 the wave vectors have magnitudes
Ik11=lkt .l,
Ik31=Ik3n1
Iki1=1k31 .
(6)
We now wish to find the coefficients R and T such that the total wave satisfies the
boundary conditions .
In the TE polarized case only the following field components are present
Ey , Hx, HZ ,
J, .
Hy , Ex,
E2, Jx,
where H is the magnetic field and J is the induced surface current on the metallic
strips.
2 .1 . Assumptions
(a) We have an infinitely thin metallic grating .
(b) The grating is periodic and of infinite dimension .
(c) The incident wave is TM or TE polarized with wave vector in the xz plane .
The procedure developed in this paper can be conveniently extended to
consider the case of an arbitrarily polarized incident field [21] .
(d) Regions 1 and 3 separated by the metallic strips are filled with the same
homogeneous material . We assume the material is air in this problem . An
identical procedure can be developed for the case where regions 1 and 3
contain different materials [22] . The procedure has also been extended to
consider a strip grating placed on a dielectric slab of some thickness h
[22,231 .
1633
In the TE case we wish to obtain an expression for the surface current density on
the metallic strips . For the dual TM case we require an expression for the electric
field distribution in the z = 0 plane . We apply the following boundary conditions :
(i) The tangential electric field is continuous across the z = 0 boundary . In the
TE case we get then
Rn +
b o = T n.
(7 a)
-R n +Sn0 .
(7 b)
Tn =
(8)
Y Cn exp(-ik tn xk),
(9)
n=-ao
where
(10)
Cn = BnRn,
Bn=
).
(k1
(11)
n=-co
where
R 1,, +S no =R n .
E1x(x)= E
Cn exp(-ik ln xk),
(9 a)
n=-co
where
Cn = 1 kln 4Rln,
CUE
(10a)
= BnR1n,
and
I
Bn=
Coe
k1n 4 .
(11 a)
1634
k2 > ( kl .1)2 ,
k 2 < (k 1 8)2 .
(12)
E(x) =R,J(x),
on metallic surface .
3.
Solution
In order to solve for the reflected and diffracted waves we use the following two
equations at the z = 0 plane .
For the TE case
cc
El y l .=O =
Jylz=o -
` Cn exp(-lk ln xk) .
n=-oo
.=0=
.1
E1
Y Cn exp(-ik ln x *),
n=-m
J1xIZ=o =
n= - oo
(Rln+5no)exp( - ikln x k) .
As a first step, let us describe the iterative procedure based on the Tsao-Mittra SIT
approach . The procedure is described in detail for the TE case . The solution in the
TM case can be obtained in a similar manner .
We utilize the periodicity of the grating and use equation (1) to write the above
equations for the TE case as
El ylz=o =
n=-ao
2n
in -Xx /,
00
Jy lz=0 =
Cn exp(-ik l x k)exp/ in
n=-ao
d x' .
These equations can be recognized as being in the form of a Fourier Series . The
iteration procedure is as follows :
Jy)(x)=
n= - 00
1635
2it
C ;,) exp (-ik l xk) exp i d x
Now make a new estimate J Y(' ) (x) of J (Y' )(x) by applying the constraint that the
surface current density is zero where there is no metal . So
;,)(x)S(x),
Jyl)(x) =J
where
S(x) = (0, in gap,
j 1, else .
The Fourier coefficients for the truncated J field can thus be obtained by
convolving the Fourier coefficients of the square wave S(x) with the Fourier
coefficients of the field Jy)(x) . Thus if we write
00
Jyl) (x) =
,
d
M= - .0
then
00
Cm)-
1
i21t
C"
exp [i27t(n-m)]
n = _ W (n-m)
n#m
c
c(n-m)
-exp Ci
C ( ) .
+d d
This electric field, however, does not satisfy the boundary condition for the
electric field on the metallic surface .
(e) We require that
Eili,) = R.Jy1)(x),
on strips .
expCic
m#n
+aC (S n o+R"la))+RSBRknla) .
(f) Repeat until desired accuracy is obtained .
(m-n)
d
-1
(m-n)
d
1636
(a) Let
R[m] =Rm,
- 00<m<00,
C[m] = Cm ,
- o < m <
B(m, n) _
B m,
0,
m = n,
min,
00,
- co < m, n <
0C) .
(b) Obtain R ( using the Fourier series coefficient integration formula on the
initial guess Eiy) (x).
(c) Using equation (10) we obtain the Fourier series coefficients for the current
density
0 0) = BR (()) ,
and
C(1)=A1C(0),
where
i2n(n-m)
(1-expIic
d (n-m)1),
d-c
min,
m = n.
d '
(d) We now obtain a better approximation of the electric field from the current
density
R( 1 a) = B -1 C (1) =B -1 A 1 BR(0) .
(e) Applying the boundary conditions, the Fourier series coefficients of the
electric field after one iteration are given by
RM =A2R(1a ) + p ,
where
A2[m, n] =
-A 1 [m,n],
m#n,
c
d +R.Bn ,
m=n .
and
P[m]=
A2 [m,0],
m :A 0,
(c-d)
d ,
M=O .
Consequently the electric-field Fourier coefficients, after one iterative step, are
given by
)
R" =QR( + P,
1637
where
Q=A 2 B- 'A1B.
(15a)
In the TM case, we begin with a guess R( for the current density and obtain the
current density R" l) after one iteration as
Rc1>=Q'Rc + P',
where
Q'=AZB - 'A'lB.
(15b)
and
Downloaded by [INASP - Pakistan ] at 06:14 08 December 2011
CexpIic
d (n-m)I-1),
2R8
A + B.
AZ[m,n]=
min,
m = n,
-A 1 [m,n],
min,
d-c
d '
= n'
and
M :A
0,
P[M] =
d B0
d'
M=0 .
The matrices Q and Q, are fixed once the geometry of the scatterer and the angle of
incidence of the incoming wave is defined .
In the TE case, if the initial guess is the solution to the problem, then
R") = R( O ) = QR(0) +P.
Or we have the system of linear equations
[I - Q]R = P.
(16)
where I is the identity matrix . A similar system of equations can be derived for the
TM case .
Thus any solution to the problem must satisfy equation (16) . Since we know the
solution to be unique [24, 27], we can solve equation (16) for the Fourier series
coefficients of the required field components .
4. Results
To implement this procedure on computer, we need to truncate the infinite
dimensional matrices . The electric field is now represented by 2N+ 1 Fourier series
coefficients signifying the i = - N, . . . , N diffracted modes in equation (13) . Thus,
each of the matrices A 1, A2, B is truncated to a 2N+ I by 2N+ 1 matrix and the Q
1638
x
x
Figure 3 . Magnitude of electric field in the z=0 plane . The incident plane wave is TE
polarized with 0=60, d=1 . 12 and c=0-9d. R,=O . (a) N=5, (b) N=20 .
Figure 4 . Magnitude of surface current density distribution in the z=0 plane . The incident
plane wave is TM polarized with 0=45, d=5 . 022 and c=0-75d. R,=O . (a) N=5,
(b) N=20 .
matrix, given by equation (15 a), is also of dimension 2N+1 by 2N+1 . A solution to
the problem is obtained by solving the system of linear equations
[I - Q]R = P.
using the Gauss elimination procedure [9] .
In general, for the perfectly conducting grating case, the electric and current field
distribution for the TM case behaved in a similar manner to the respective dual
current and electric field distributions calculated in the case of TE incidence . For
cases where the period of the grating was near the wavelength of incident wave
(figure 3), only a few terms in the expansion were needed to obtain a good
2 .4-
1639
2 .4-
1 .6-
_sa-
0 .8-
0.8-
0
1
I
-0.2d
0
0.2d 0 .4d
c
x
1
0
0 .8d
B = 800
2 .4-
2 .4-
1
1
I
I I
0 .2d 0 .4d Old 0.8d c
x
0 =450
d = 1 .5X
d = 5 .02X
c = O .6d
c = 0 .75d
1 .6-
0 .8-
0
-0 .2d
---
I
I
I
1
0
0 .2d 0.4d
c
0.8d
x
1JI
-0 .1dO
I
I
I
0.2d 0 .4d 0 .6d
x
(' I
c 0.9d
Figure 5 . Magnitude of electric field in the z = 0 plane calculated for the different cases using
matrix size N=20 . The incident wave is TM polarized . R,=0 . (a) 0=0, d=1 . 5 .1,
c=0. 6d . (b) 0=0, d=1 . 5 .1, c=0.9d. (c) 0=60, d=1 . 5,, c=0 . 6d . (d) 0=45, d=5 . 02.1,
c=0-75d.
approximation to the field . Note that in all cases the incident field was a unit electricfield amplitude TM- or TE-polarized plane wave . Since the conservation of energy
criterion does not guarantee the solution to be correct [28], our criterion for
convergence here was the satisfication of the boundary conditions by the electric and
current fields . If only the reflection coefficient is desired, a good approximation can
be obtained using N= 5 . As the period of the grating increased with respect to the
wavelength (figure 4), more terms were required in the expansion to obtain an
accurate description of the field .
In figure 5 we display the electric fields calculated for different scatterer
configurations in the case of a TM polarized incident plane wave . To eliminate the
ripples due to Gibb's phenomenon, the electric field Fourier coefficients are first
multiplied with a hamming window of length 2N+ 1 .
1640
M=0
and
E100 (NP)
EN(NP)I
}2 ,
where
EN(x)
in
27c
x .
NP represents the total number of sampling points for the region 0 < x < d. For a
particular matrix size, the error increases upon decreasing the strip width, increasing
Figure 6 . Mean squared error in the electric field magnitude for different strip widths and
angles of oblique incidence of TE polarized plane wave . RS =O .
: 0=0, d= 1-5A,
1641
1 .5
1 .25-
1 .0-
0.75k
- .t-
0 .5-
0 .25-
A
0
I
0.2d
(I i
1
0.4d c 0 .6d 0.8d
x
Figure 7 . Magnitudes of electric and RR x current fields in the z=0 plane calculated using
matrix size N=20 . We have normal incidence, d=0.97 .1, c=0-5d, R,=5000 and the
incident wave is TE polarized .
2 .5
2 .0-
1 .0
0 .5-
Figure 8 . Magnitudes of electric and Ra x current fields in the z=0 plane calculated using
matrix size N=20 . We have normal incidence, (.i.-d)=3 x 10 - ', c=0-5d, Rg =500I
and the incident wave is TE polarized .
the period of the grating relative to the wavelength, or increasing the angle of oblique
incidence . Irrespective of the scatterer or the angle of incidence of the plane wave, it
can be observed that this error decreases monotonically with increasing matrix size .
The trade-off involved however is that the computer time increases correspondingly
with increasing matrix size .
Hall and Mittra in [8] consider imperfectly conducting strips . We ran our
program for the cases considered and obtained similar results for the reflection
coefficient in all cases except for the cases where we had normal incidence and the
period of the grating was nearly equal to the wavelength of the incident wave . When
1642
the wavelength of incident wave is equal to the period of the grating, one of the terms
in the diagonal matrix B is equal to 0 . Thus B -1 can not be computed . To avoid this
problem we considered 2-d= + 3 x 10 - ' . As expected, almost identical results were
obtained for the two cases . In figures 7 and 8 we consider the electric and current field
distributions in the case of a TE-polarized plane wave normally incident on a strip
grating with resistivity R S = 500 . Again to reduce ripples due to Gibb's phenomenon,
the current-field Fourier coefficients are first multiplied by a hamming window of
length 2N+ 1 . Although the period of the grating used in figure 8 was only 3% larger
than the one used in figure 7, the fields changed significantly and the magnitude of
the reflection coefficient also decreased sharply . It can be observed that in both cases
the boundary condition for the tangential electric field on the conducting strips is
well satisfied by the field R,J 5,(x) .
5.
Conclusions
We have introduced a new robust method for the solution of the scattered field
distributions in the case of a plane wave obliquely incident on a metallic strip grating .
An infinite-dimensional system of linear equations is derived which is equivalent to
that arrived at using the moment method solution . A solution is obtained by
truncating the infinite-dimensional matrix and using Gauss elimination to solve this
system of linear equations . In this paper we consider both TM and TE polarized
incident plane waves . The solution procedure is general enough to treat gratings
both of finite and infinite conductivity .
We have shown how, using our method, an arbitrarily accurate description of the
electric and current fields can be obtained for any configuration of the strip grating
and for any angle of oblique incidence of the plane wave (figure 1) . Different scatterer
configurations are considered . In each case it is shown that the error in the magnitude
of the fields in the plane of the strip grating depends on the size of the truncated
matrix . It is shown that this error decreases monotonically as the matrix size
increases .
References
[1] PRIMICH, R . I ., 1957, IRE Trans . Antennas Propagat ., 25, 176-182 .
[2] NEUREUTHER, A . R., and ZAKI, K ., 1968, Radio Science, 3, 1158-1167 .
[3] HILL, D . A ., and WAIT, J . R ., 1976, Can . J . Phys ., 54, 353-361 .
[4] CHEN, C . C ., 1970, IEEE Trans . microw . Theory Tech ., 18, 627-632 .
[5] TSAO, C .-H ., and MITTRA, R., 1982, IEEE Trans . Antennas Propagat ., 30, 303-308 .
[6] CHRISTODOULOU, C . G ., and KAUFFMAN, J . F ., 1986, IEEE Trans . Antennas Propagat .,
34, 144-154 .
[7] UCHIDA, K ., NoDA, T ., and MATSUNAGA, T ., 1987, IEEE Trans . Antennas Propagat ., 35,
46-52 .
[8] HALL, R . C ., and MITTRA, R ., 1985, IEEE Trans . Antennas Propagat ., 33, 1009-1011 .
[9] HARRINGTON, R . F ., 1967, Proc . IEEE, 55, 136-149 .
[10] SARKAR, T . K ., SIARKIEWICZ, K . R ., and STRATTON, R . F ., 1981, IEEE Trans . Antennas
Propagat., 29, 847-856 .
[11] HESTENES, M . R ., and STIEFEL, E ., 1952, J . Res . natn . Bur . Stand ., 49, 409-435 .
[12] SARKAR, T . K ., and RAO, S . M ., 1982, IEEE Trans . Antennas Propagat ., 32, 398-403 .
[13] PETERSON, A . F ., and MITTRA, R ., 1985, J. opt . Soc . Am ., 2, 971-977 .
[14] RICHMOND, J . H ., 1980, IEEE Trans. Antennas Propagat., 28, 883-887 .
[15] WILTON, D . R ., and GovIND, S ., 1977, IEEE Trans . Antennas Propagat ., 25, 845-850 .
[16] Wu, T. K ., 1987, IEEE Trans . Antennas Propagat ., 35, 205-207 .
[17] VAN DEN BERG, P . M ., 1984, IEEE Trans. Antennas Propagat ., 32, 1063-1070 .
[18] MACKAY, A . J ., and MCCOWEN, A ., 1987, IEEE Trans . Antennas Propagat ., 35.218-220 .
1643
[19] KAS, A ., and Yip, E . L ., 1987, IEEE Trans . Antennas Propagat ., 35, 147-152 .
[20] GAYLORD, T. K ., and MOHARRAM, M . G ., 1985, Proc . IEEE ., 73, 894-937 .
[21] NAQVI, S . S . H ., and GALLAGHER, N . C ., 1990, IEEE Trans . Antennas Propagat ., 38,
414-415 .
[22] NAQvI, S . S . H ., 1988, Electromagnetic Wave Scattering from Planar Periodic Metallic
Surfaces, Ph .D . Dissertation, Purdue University .
[23] NAQVI, S . S . H ., and GALLAGHER, N . C ., 1989, J . opt . Soc . Am . (to be published) .
[24] BORN, M ., and WOLF, E ., 1985, Principles of Optics, sixth edition (New York :
Pergammon), p . 6 .
[25] SENIOR, T . B . A ., 1979, IEEE Trans . Antennas Propagat ., 27, 808-813 .
[26] ORTA, R., TASCONE, T ., and ZICH, R ., 1985, Electromagnetics, 5, 307-329 .
[27] PETIT, R ., 1980, Electromagnetic Theory of Gratings, Topics in Current Physics (Berlin :
Springer), pp . 56-57 .
[28] PETIT, R ., and TAYEB, G ., 1987, Application and Theory of Periodic Structures,
Diffraction Gratings, and Moire Phenomena III, SPIE proceedings Vol. 815 .