Analysis of Composite Metallic and Dielectric Structures - WIPL-D code

Branko Kolundzija
1
, Jovan Ognjanovic
2
, Tapan Sarkar
3
1
Dept of EE, University of Belgrade, P.O.Box 35-54, 11120 Belgrade, Yugoslavia, [email protected]
2
Institut Vinca, P.O.Box 522, 11001 Belgrade, Yugoslavia, [email protected]
3
Dept of EECS, Syracuse University, 121 Link Hall, Syracuse, NY 13244, [email protected]
Abstract: WIPL-D code uses the Galerkin method to solve surface integral equations for electric and magnetic currents
over dielectric boundary surfaces and perfect electric conductors. Flexible geometrical modeling is performed by truncated
cones for wires and bilinear surfaces for other surfaces. Efficient approximation of currents is achieved by using
polynomial entire-domain expansions that automatically satisfy the continuity equation. Special care is devoted to:
a) modeling of wire-to-plate junctions and protrusions of wires through dielectric surfaces, b) automatic segmentation of
large surfaces and c) automatic parameterization of 3Dgeometries. Thus very efficient and user friendly code is obtained,
enabling solution of real life problems at personal computers.
INTRODUCTION
Very often problems in radar scattering, antenna and microwave fields involve frequency analysis of material objects made
of linear and isotropic conductors, dielectrics and their combinations. In some cases, but rarely, the problem includes linear
and isotropic magnetic materials. Hence, such material objects are termed composite metallic and dielectric structures.
Generally speaking, the main objective of the frequency-domain analysis is to evaluate the distribution of the
electric and magnetic fields or their sources inside and/or outside the material object at one frequency. This objective can
be achieved in different ways. In what follows a particular way is given, which represents a basis for WIPL-D code [1-5].
WIPL-D code is based on the solution of surface integral equations (SIEs) for equivalent currents over dielectric
boundary surfaces and electric currents over perfect electric conductors (PECs). Flexible geometrical modeling is
performed by truncated cones for wires and bilinear surfaces for other surfaces. Special care is devoted to: a) modeling of
wire-to-plate junctions, b) modeling of protrusions of wires through dielectric surfaces, c) automatic segmentation of large
surfaces, d) and automatic parameterization of 3Dgeometries. Efficient approximation of currents is achieved by using
polynomial entire-domain expansions (i.e. rooftop sub-domain expansions in the particular case) that automatically satisfy
the continuity equation, particularly at arbitrary multiple metallic and/or dielectric junctions. The integral equations are
solved by using the Galerkin method, which is a special case of the method of moments (MoM). Special care is devoted to
efficient evaluation of impedance integrals covering: a) arbitrarily thin wires, b) extremely narrow patches (the ratio of two
sides can go up to 1000), and c) materials of high dielectric and magnetic constants (magnitude of these constants can go
up to 1000), etc. Thus very efficient code is obtained, requiring 3-4 unknowns per wavelength along electrically long wires,
and 10-20 unknowns per wavelength squared along large metallic surfaces, enabling solution of real life problems at
personal computers. Two of these problems are shown at the end: a) scattering from Mirage at 1 GHz, in which case its
electrical length is more than 40 wavelengths, and b) input impedance of stacked microstrip antenna in the cavity, made in
multi-layer technology.
FORMULATION OF THE PROBLEM
Inhomogeneous dielectrics can be always represented as part by part homogeneous dielectrics. Hence, any composite
metallic and dielectric structure can be represented as the electromagnetic system consisting of a finite number of finite-size
linear, homogeneous, and isotropic regions, situated in an unbounded linear, homogeneous, and isotropic environment, as
shown in Figure 1a. Most often, this environment is a vacuum, but it also can be another medium (e.g., water).
Figure 1. Decomposition of a multiple-region problem into single-region problems:
a) original problem, b) equivalent problem for region i.
Some of the regions can be PEC. In the interior of any such region, the electromagnetic field is always zero.
Hence, all of them will collectively be denoted as region 0 and referred to as zero-field regions. As special cases of zero-
field regions, we shall consider metallic wires and plates. Plates forming an open surface can be considered as a degenerate
case of a zero-field region, where one dimension (thickness) of the region is zero.
In all the remaining regions generally there exist electromagnetic fields. These regions will collectively be referred
to as non-zero-field regions. We denote the total number of such regions by n. The medium filling out region i is assumed
to be described by the complex permittivity,
( ) i
, and permeability,
( ) i
, i = 1,,n, which include all losses. In any of
these regions, there may exist impressed electric and magnetic fields,
) (
inc
i
E and
) (
inc
i
H , i = 1,,n, whose angular frequency
is .
Let us consider an arbitrary region with a non-zero electromagnetic field, e.g., region i shown in Figure 1a.
According to the surface equivalence theorem, the influence of all sources outside region i can be replaced by equivalent
currents placed at the boundary surface of region i, in which case the field outside region i becomes zero, as shown in
Figure 1b. Hence, the region outside region i is denoted as region 0-i (i.e., region 0 with respect to region i, in the
equivalent problem for region i). Since the field outside region i is zero it can be homogenized with respect the region i, i.e.
it can be filled by the same material as region i. Thus, a multiple-region problem (consisting of n-regions) is decomposed
into n single-region problems.
The densities of equivalent currents at the boundary surface between regions i and j are
) (i
ij sij
H n J ,
) (i
ij sij
E n M , (1a,b)
where
ij
n is unit normal directed from region j to region i, and
) (i
E and
) (i
H are the electric and magnetic fields at the
boundary surface, just inside region i. If we consider equivalent currents for region j,
sji
J and
sji
M , they are evaluated
according to equations (1), but with interchanged indices i and j, i.e.,
) ( j
ji sji
H n J ,
) ( j
ji sji
E n M . (2a,b)
Since there are no currents between regions i and j, the fields
) (i
H and
) ( j
H , and the fields
) (i
E and
) ( j
E satisfy
boundary conditions
( ) 0
) ( ) (

j i
ij
H H n , ( ) 0
) ( ) (

j i
ij
E E n . (3a,b)
After expressing the field vectors in terms of the equivalent currents according to (1) and (2), and using
ij ij
n n , the
equivalent currents are related as
sji sij
J J ,
sji sij
M M . (4a,b)
Thus, n-single single region problems are mutually coupled through conditions (3) and (4). It is shown that
satisfaction of these conditions guarantee uniqueness of the solution for both, sources and fields. In that case, the main
objective of the frequency-domain analysis is to evaluate the distribution of equivalent electric and magnetic currents at
boundary surfaces of single-region problems, which satisfy conditions (3) and (4).
FIELD INTEGRAL EQUATION
Field integral equations are derived from boundary conditions (3). Let us start with condition (3b). The total electric field in
region i can be expressed as
) (
0
) ( ) (
) , (
i
inc
n
k
sik sik
i i
E M J E E +

, (5)
where ) , (
) (
sik sik
i
M J E represents the scattered field inside region i, which is produced by currents placed on the boundary
surface between regions i and k, and
) (i
inc
E is the corresponding incident field. The scattered field, just inside region i, due
to the currents placed on the boundary surface of regions i and k, is
) ( ) ( ) , (
) ( ) ( ) ( ) (
sik
i
sik
i i
sik sik
i
K L Z M J M J E + , (6)
where
) (i
Z is the intrinsic impedance of the medium filling region i, and
) (i
L and
) (i
K are operators defined by
' d ) ( ) (
] [
1
) ( ) ( ) (
) (
2 ) (
) ( ) (
ik
S
i
ik sik sik
i
i
ik sik sik
i
S R g R g L
ik

1
1
]
1

r J r J J , (7)
' d ) ( ) ( ) (
) ( ) (
ik
S
i
ik sik sik
i
S R g K
ik

r J M , (8)
g R
e
R
i
R
i
( )
( )
( )

4
,
ik
R r r ,
( ) ( ) ( ) i i i
j (9)
r
ik
is the vector position of the source point and r is the vector position of the field point,
sik
acts on r
ik
(inside surface S
ik
)
and acts on r. Finally, after replacing (5) and (6) written for regions i and j into (3b), the integral equation is obtained in
the form
[ ] [ ] ( )
) ( ) (
0
) ( ) ( ) (
0
) ( ) ( ) (
) ( ) ( ) ( ) (
j
inc
i
inc ij
n
j k
k
sjk
i
sjk
i j
n
i k
k
sik
i
sik
i i
ij
K L Z K L Z E E n M J M J n

'

, (10)
In a similar way we obtain the integral equation from boundary condition (3a), which is dual to integral equation (10) (not
written here due to lack of the space).
The obtained set of equations actually represents a general form of the so-called PMCHW formulation. In the case
when one of two regions having a common boundary surface is a PEC, the magnetic currents are equal to zero at the
boundary surface and the first of above equations degenerates into the electric field integral equation (EFIE). Note that the
EFIE provides solution not only for closed metallic bodies, but also for open metallic surfaces and metallic wires. (It is
well-known that the usage of the EFIE at internal resonant frequencies of the PEC can result in spurious solutions. This
problem is practically eliminated using higher order basis functions.) Particularly, in the case of wires the EFIE is based on
the extended boundary conditions and thin wire approximation. Thus a set of PMCHW and EFIE equations for the
unknown electric and magnetic currents is obtained. This set is solved by the method of moments (MoM).
In order to obtain an efficient method for analysis of general structures, special care is devoted to the choice of
basis functions, which is performed into two steps: geometrical modeling and approximation of currents.
GEOMETRICAL MODELING
Geometrical modeling of wires is performed by right truncated cones. A right truncated cone is determined by position
vectors and radii of its beginning and its end,
1
r and
1
a , and
2
r and
2
a , respectively, as shown in Figure 2a. The
parametric equations of the cone axis and its local radius can be written in the form
2 1
2
1
2
1
) ( r r r
s s
s
a
+
+

,
2 1
2
1
2
1
) ( a
s
a
s
s a
+
+

, 1 1 s , (11a,b)
where s is local coordinate along a cone generatrix. In order to define parametric equation of the cone surface, let us adopt
a local cylindrical coordinate system in such a manner that the z-coordinate axis coincides with the cone axis. In that case
the parametric equation of the cone surface can be written as
) ( ) ( ) ( ) , ( +

i r r s a s s
a
, 1 1 s , , (12a,b,c)
where is the circumferential angle, measured from the x-axis, and ) (

i is the radial unit vector, perpendicular to the

cone axis.
As special cases, the truncated cone degenerates into a right cylinder (
2 1
a a ), an ordinary cone ( 0
2
a ), a flat
disc ( 0
2
a ,
2 1
r r ), and a frill (
2 1
r r ). The right truncated cone and its degenerate forms can be used for modeling of
cylindrical wires with flat (frill-like) or conical changes of the wire radius, as well as of flat and conical wire ends and
feeds.
Figure 2. Basic building elements: (a) right truncated cone, and (b) bilinear surface.
Metallic and dielectric surfaces are modeled by bilinear surfaces. A bilinear surface is, in general, a nonplanar
quadrilateral, which is defined uniquely by its four arbitrarily spaced vertices, as shown in Fig. 2b. Hence, it can be used for
efficient modeling of both flat and curved surfaces. The parametric equation of such an isoparametric element can be
written in the form

4
) 1 )( 1 (
+
4
) 1 )( 1 (
+
4
) 1 )( 1 (
+
4
) 1 )( 1 (
) , (
22 21 12 11
s p s p s p s p
s p
+ + + +
r r r r r
1 1 p , 1 1 s (13a,b,c)
where r
11
, r
12
, r
21
, and r
22
are the position vectors of its vertices, and p and s are local coordinates. After elementary
transformations this equation can be written as
ps s p s p
ps s p c
r r r r r + + + ) , ( , 1 1 p , 1 1 s (14.a,b,c)
where
c
r =(r
11
+r
12
+r
21
+r
22
)/4,
p
r =(-r
11
-r
12
+r
21
+r
22
)/4,
s
r =(-r
11
+r
12
-r
21
+r
22
)/4 and
ps
r =(-r
11
+r
12
-r
21
+r
22
)/4. Depending on
the values of the vectors
p
r ,
s
r , and
ps
r , a bilinear surface takes different degenerate forms: flat quadrilateral (
p
r ,
s
r ,
and
ps
r are coplanar), rhomboid (
ps
r = 0), rectangle (
ps
r = 0 and
s p
r r = 0) , etc.
Geometrical modeling of wire-to-plate junctions is performed using specific segmentation technique, as shown in
Figure 3. Each plate at the junction is subdivided into two, three or four plates in such a manner that short edges of
subplates surround wire end. In addition, it is forced that total current flowing out from the junction domain (consisting of
the short edges and the wire end) be zero. The same technique is applied in the case when wire protrude through dielectric
surface.
Figure 3. Wire-to-plate junctions.
Figure 4. Segmentation of large plates.
The maximum allowed length of edges of bilinear surfaces is two wavelengths. (Namely, the current over patches
of this size still can be successfully approximated by single entire-domain expansion.) If any of patch edges is longer than
two wavelengths the patch is subdivided into a set of sub-patches in such a manner that only edges longer than two
wavelengths are subdivided into minimum number of edges shorter than two wavelengths.
The geometrical modeling of composite structures by truncated cones and bilinear surfaces is a tedious job, even
in the case of simple shapes. For example, a simple sphere is completely specified by only one parameter, its radius.
However, a good geometrical model of the sphere is made of at least 24 or 54 bilinear surfaces (the latest is shown in
Figure 5.
Figure 5. Automatic parameterization of 3D geometries.
On the other side, there are some complicated structures that can also be completely specified with a few
parameters, e.g., quadrifilar (wire) helicoidal antenna mounted on a dielectric rod. This difficulty can be overcome by
introducing an automated segmentation (parameterization) of various 3D geometries. Many of them belong to five general
classes, which we termed: the generalized sphere, the generalized circle, the generalized reflector, the two-generatrix body
(2G-body), and the generalized helix.
APPROXIMATION OF CURRENTS
Currents along wires are approximated by polynomials that automatically satisfy continuity equation at wire ends and
junctions. The starting approximation is simple polynomial in the form

s
N
i
i
i
s a s I
0
) ( , 1 1 s , (14a,b)
where a
i
are unknown coefficients and N
s
is order of approximation. Coefficients a
0
and a
1
can be expressed in terms of
other unknown coefficients a
i
, i = 2,.., N
s
, and values of current at the wire ends, I
1
= I(-s) and I
2
= I(s). In that case
expansion (14a) can be written as

+ +
s
N
i
i i
s S a s N I s N I s I
2
2 1
) ( ) ( ) ( ) ( , (15)
where node basis function, N(s), and segment basis functions, S
i
(s), i = 2,.., N
s
, are expressed as
2
1
) (
s
s N

,

'

odd
even
1
) (
i
i
s s
s
s S
i
i
i
, (16a,b)
Note that node basis functions are equal to unity at one node (wire end) and to zero at another node, while segment basis
functions are equal to zero at both nodes. Obviously, the continuity equation at free wire end is satisfied by omitting the
corresponding node basis function. The continuity equation at junction of two or more wires is satisfied by grouping the
corresponding node basis functions into doublets. These doublet actually represent triangle (rooftop) basis functions. One
doublet automatically satisfy continuity equation at a junction of two wires, two doublets automatically satisfy continuity
equation at a junction of three wires, and so on.
The surface current over a bilinear surface is decomposed into its p and s-components. However, the p-current
component can be treated as the s-current component defined over the same bilinear surface with interchanged p and s-
coordinates. Thus, in the general case, a distribution of surface currents can be represented as a sum of s-components
defined over bilinear surfaces which overlap or are interconnected.
The initial approximations for s-components of electric and magnetic currents over a bilinear surface are

p
s
N
i
N
j
ij ij s
s p a s p
0 0
) , ( ) , ( F J , ) , ( ) , (
0 0
s p b s p
p s
N
i
N
j
ij ij s

F M , (17a,b)
where N
p
and N
s
are the degrees of approximations along the coordinates, a
ij
and b
ij
are unknown coefficients, and F
ij
are
initial basis functions, defined as
j i
s p
s
ij
s p s p
a a
a
F

) , ( ,
p
s p
p

) , ( r
a ,
s
s p
s

) , ( r
a . (18a,b,c)
(a
p
and a
s
are the so-called unitary vectors.) Let us introduce new unknowns c
i1
and c
i2
, i=0,...,N
p
, defined as
c a
i ij
j
j
N
s
1
0
1

( ) ,

s
N
j
ij i
a c
0
2
, (19a,b)
Coefficients for any i, a
i0
and a
i1
can be expressed in terms of other unknown coefficients, a
ij
, j = 2,3,,N
s
, c
i1
and c
i2
. In
that case expansion for electric currents (17a) can be written as

'

+ +
p
s
N
i
N
j
ij ij i i i i s
s p a s p c s p c s p
0 2
2 1
) , ( ) , ( ) , ( ) , ( P E E J , (20)
where edge basis functions, E
i
(p,s), i = 0,,N
p
, and patch basis functions, P
ij
(p,s), i = 0,,N
p
, j = 2,..,N
s
, are expressed as
) ( ) , ( s N p s p
i
s p
s
i
a a
a
E

, ) ( ) , ( s S p s p
j
i
s p
s
ij
a a
a
P

, (21a,b)
Note that component of patch basis functions normal to the patch edges is equal to zero at these edges. Component
of E
i
normal to patch edges is not equal to zero only at edge s=-1, at which it is equal to
1
p
i
p a , where
p
a is constant
function along the edge. Obviously the continuity equation along this edge must be satisfied independently for each order i.
At free patch edge it is satisfied by omitting the corresponding edge basis functions (of all orders). The continuity equation
for i
th
order at junction of two or more patches is satisfied by grouping the corresponding edge basis functions of i
th
order
into doublets of i
th
order. These doublets represent generalized rooftop basis functions. One doublet of i
th
order
automatically satisfy the continuity equation for i
th
order at a junction of two patches, two doublets of i
th
order
automatically satisfy the continuity equation for i
th
order at a junction of three patches, and so on.
In the case of multiple dielectric junction the continuity equation must be satisfied in each region. In that case one
doublet of i
th
order automatically satisfy the continuity equation for i
th
order at a junction of two patches in each region.
However, according to (4a) all these doublets must enter to the final solution with the same weighting coefficients. Hence,
all these doublets are grouped into one multiplet basis function of i
th
order. In the case of composite metallic and dielectric
junction the continuity equation for i
th
order is satisfied by proper combination of doublets and multiplets.
Test Procedure
In order to determine the unknown current coefficients, the first of coupled integral equations (10a) is tested by electric
current basis functions, and the second of coupled integral equations (10b) is tested by magnetic current basis functions.
Thus obtained matrix elements represent linear combinations of two types of impedance integral, Z
kl
L
and Z
kl
K
, defined as
( )
l
S
l k l k
L
kl
dS L L Z
l

, F F F F , ( )
l
S
l k l k
K
kl
dS K K Z
l

, F F F F , (22)
where L and K are linear operators given by (7) and (8), F
k
is the k-th basis function defined over the surface S
k
, and F
l
is
the l-th test function defined over the surface S
l
. (Before applying the operators in the above expression the notation of (7)
and (8) is changed: order number of region is omitted, index of boundary surface ik is replaced by index of basis function
k, and position vector of field point, r, is replaced by position vector of test surface, r
l
.) After applying the operators L
and K and after simple transformation the above impedance integrals are obtained in relatively simple symmetrical form

1
1
]
1

+
l k
S
l k
S
l l l k k k l l k k
L
kl
dS dS R g Z ) ( ) ( ) (
1
) ( ) (
2
r F r F r F r F , (23)
( ) [ ]


l k
S
l k
S
l l k k l k
K
kl
dS dS
dR
R dg
R
Z
) ( 1
) ( ) ( r F r F r r , (24)
Having in mind that
1
1

j i
s p
ij
s jp
a a
F , dpds dS
s p
a a (25)
impedance integrals (23) and (24) can be represented as linear combination of only two classes of quadruple integrals
I p s p s g R dp ds dp ds
i j i j
L
k
i
k
j
l
i
l
j
k k l l
k k l l
k k l l

j ( )
1
1
1
1
1
1
1
1
, (26)
I p s p s
R
dg R
dR
dp ds dp ds
i j i j
K
k
i
k
j
l
i
l
j
k k l l
k k l l
k k l l

1
1
1
1
1
1
1
1
1
( )
. (27)
Special care is devoted to the evaluation of these integrals.
Examples
Let us consider two real life problems. Figure 6 shows monostatic RCS at f=600MHz for Mirage, the model of which is
shown in inset. At this frequency electrical length of the airplane is about 25 wavelengths. Number of unknowns used in the
analysis is 7760. The analysis is performed at Pentium 2 with 512 Mb of RAM.) Figure 7 shows input impedance of
rectangular stacked patch antenna enclosed in cavity. Good agreement between WIPL-D results and measured from Ref. 11
can be observed. (Number of unknowns in this case is 421 with symmetry option applied.)

Figure 6. RCS of Mirage at f = 600 MHz.

Figure 7. Input impedance of rectangular stacked patch antenna enclosed in cavity.
REFERENCES
1. B.M. Kolundzija, J.S. Ognjanovic, and T.K. Sarkar, WIPL-D: Electromagnetic modeling of composite metallic and
dielectric structures, Software and User's Manual, Boston: Artech House, 2000.
2. B. M. Kolundzija and A. R. Djordjevic, Spiral super-quadric generatrix and bodies of two generatrices in automated
parameterization of 3-D geometries, IEEE Transaction on Microwave Theory and Techniques, vol. 45, no. 5, May
1997, pp. 864-866.
3. B. M. Kolundzija: Automatic mesh generation using single- and double-node segmentation techniques, (feature
article), IEEE Antennas and Propagat. Magazine, vol. 40, no. 4, pp. 30-38, Aug. 1998.
4. B. M. Kolundzija: "Electromagnetic modeling of composite metallic and dielectric structures", IEEE Trans.
Microwave Theory Tech., vol. MTT-47, pp. 1021-1032, July 1999.
5. G. G. Gentili, L.E. Garcia-Castillo, F. Perez-Martinez, M. Salazar-Palma: Improved Greens function formulation for
the analysis of rectangular stacked patch antennas enclosed in a cavity, Digest of IEEE AP-S Int. Symposium, vol. 2,
pp. 1070-1073, Baltimore, 1996.