This action might not be possible to undo. Are you sure you want to continue?
The most basic assumption of the MEI
method is the existence of a local linear relation of the
ﬁelds. Such an assumption is plausible for far ﬁelds, which
is a differential relation of the ﬁelds, known as the So
mmerfeld radiation condition in classical theory. One may
question whether it can still be so assumed on the surface
of the object. However, the success of IEMEI suggests
that it may be true that such local linear equations do ex
ist on the object surface. If this were true, the onsurface
Maxwell differential equations could be found. What an
exciting discovery would that be! Mei has proved the ex
istence of the differential equation of currents on thin wire
structure. The readers may ﬁnd the presentation in the
MAXWELLIAN CIRCUITS article in this encyclopedia. Perhaps,
it is not far away that someone will prove the existence of
differential equations of the surface current density on
conduction surfaces.
BIBLIOGRAPHY
1. K. K. Mei, Unimoment method of solving antenna and scat
tering problems, IEEE Trans. Anten. Propag. AP22(6):760–
766 (Nov. 1974).
2. B. Engquist and A. Majda, Absorbing boundary condition for
the numerical simulation of waves, Math. Comput. 31:629–
651 (1977).
3. J. Fang, Time Domain Finite Difference Computation for
Maxwell’s Equations, Ph.D. dissertation, Univ. California,
Berkeley, CA, 1989.
4. J. P. Berenger, A perfectly matched layer for absorption of
electromagnetic waves, J. Comput. Phys. 114:185–200 (Oct.
1994).
5. K. K. Mei, P. Pous, Z. Chen, Y. Liu, and M. Prouty, Measured
equation of invariance—a new concept in ﬁeld computations,
IEEE Trans. Anten. Propag. 42(3):320–328 (March 1994).
6. J. Jevtic and R. Lee, A theoretic and numerical analysis of the
measured equation of invariance, IEEE Trans. Anten. Propag.
42(8):1097–1105 (Aug. 1994).
7. K. K. Mei and Y. Liu, Comments on ‘‘A theoretic and numer
ical analysis of the measured equation of invariance,’’ IEEE
Trans. Anten. Propag. 43(10):1168–1171 (Oct. 1995).
8. J. Jevtic and R. Lee, How invariant the measured equation of
invariance? IEEE Microwave Guided Wave Lett. 5(2):45–47
(Feb. 1995).
9. K. K. Mei, Comments on ‘‘How invariant the measured equa
tion of invariance?’’ IEEE Microwave Guided Wave Lett.
5(11):417 (Nov. 1995).
10. T. L. Barkdoll and R. Lee, Finite element analysis of bodies of
revolution using the measured equation of invariance, Radio
Sci. 30:803–815 (July–Aug. 1995).
11. Y. L. Luo, K. M. Luk, K. K. Mei, and E. K. N. Yung, Finite
difference analysis of electrically large parabolic reﬂector an
tennas, IEEE Trans. Anten. Propag. 50(3):266–276 (March
2002).
12. R. Pous, Measured Equation of Invariance—a New Concept in
Field Computations, Ph.D. dissertation, Univ. California,
Berkeley, CA, 1992.
13. M. Prouty, Application of the Measured Equation of Invari
ance to Planar Microstrip Structures, Ph.D. dissertation,
Univ. California, Berkeley, CA, 1994.
14. Y. Liu, K. K. Mei, and E. K. N. Yung, Interpolation, extrap
olation and application of the measured equation of invari
ance to scattering by very large cylinders, IEEE Trans. Anten.
Propag. 45(9) (Sept. 1997).
15. O. M. Ramahi, A. Khebir, and R. Mittra, Numerically derived
absorbing boundary condition for solution of open region scat
tering problems, IEEE Trans. Anten. Propag. 39:350–353
(March 1991).
16. B. Stupfel and R. Mittra, A theoretical study of numerical
absorbing boundary conditions, IEEE Trans. Anten. Propag.
43(5):478–486 (May 1995).
17. J. M. Rius, R. Pous, and A. Cardama, Integral formulation of
the measured equation of invariance: A novel sparse matrix
boundary element method, IEEE Trans. Magn. 32(3):962–967
(May 1996).
18. Y. Liu, K. K. Mei, and E. K. N. Yung, Differential formulation
of onsurface measured equation of invariance for 2D con
ducting scatterings, IEEE Microwave Guided Wave Lett.
8(2):99–101 (Feb. 1998).
19. M. Hirose, M. Miyake, J. Takada, and I. Arai, New integral
equation formulation of the measured equation of invariance
and the extension to analyze two dimensional cylinders with
impedance boundary conditions, Radio Sci. 34(1):65–82
(Jan.–Feb. 1999).
20. J. M. Rius, C. P. Carpintero, A. Cardama, and J. R. Mosig,
Integral equation MEI applied to three dimensional arbitrary
surfaces, Electron. Lett. 33(24):2029–2031 (Nov. 1997).
21. N. M. A. Chowdhury, J. Takada, and M. Hirose, Novel formu
lation for scalarﬁeld approach of IEMEI method to solve
threedimensional scattering problem, IEICE Trans. Fund.
Electron. Commun. Comput. Sci. E86A(8):1905–1912 (Aug.
2002).
METHOD OF LINES
ALESSANDRO TOSCANO
LUCIO VEGNI
‘‘Roma Tre’’ University
Rome, Italy
1. INTRODUCTION
Planar passive structures and planar transmission lines
are widely used in microwave and millimeterwave inte
grated circuits. Characterization and modeling of trans
mission lines and their discontinuities are of great
importance in the development of integrated circuits. It
is not economical, or in many cases not even feasible, to
tune the circuits once they are fabricated. Therefore,
accurate methods are needed to model the structures.
Quasistatic methods, equivalent waveguide models,
and equivalentcircuit models have been used in the
modeling of microstrip or slotlines, and discontinuities in
the past [1,2]. A more rigorous approach, which takes into
account all the physical effects including radiation and
surface waves is the spectraldomain Galerkin method
(SDGM) [3]. This method transforms an integral equation
into a linear system of equations, where the solution (e.g.,
electric currents or ﬁelds) has to be computed numerically.
The method is known to be efﬁcient but is restricted, in
2548 METHOD OF LINES
Previous Page
general, to wellshaped structures that involve inﬁnitely
thin conductors as ground planes. The main reason for the
efﬁciency is that the SDGM allows signiﬁcant analytical
preprocessing in comparison to other numerical techniques
such as the ﬁnitedifference method, ﬁniteelement meth
ods, or the method of lines. The method of partial discre
tization or method of lines is a wellestablished technique
for the analysis of different microwave structures.
The method of lines is regarded as a special ﬁnite
difference method but is more effective with respect to
accuracy and computational time than is the regular
ﬁnitedifference method. It basically involves discretizing
a given differential equation in one or two dimensions
while using an analytical solution in the remaining direc
tion. The method of lines (MoL) has the merits of both the
ﬁnite difference method and the analytical method; it does
not yield spurious modes, nor does it have the problem of
‘‘relative convergence.’’ Besides, the method of lines has
the following properties that justify its use:
1. Computational efﬁciency—the semianalytical char
acter of the formulation leads to a simple and
compact algorithm, which yields accurate results
with less computational effort than do other techni
ques.
2. Numerical stability—by separating discretization of
space and time, it is easy to establish stability and
convergence for a wide range of problems.
3. Reduced programming effort—by making use of the
stateoftheart welldocumented and reliable ordin
ary differential equations (ODE) solvers, program
ming effort can be substantially reduced.
4. Reduced computational time—since only a small
amount of discretization lines are necessary in the
computation, there is no need to solve a large system
of equations; hence computing time is small.
2. STEPS IN MoL APPLICATION
Application of MoL usually involves the following four
basic steps:
1. Partitioning the solution region into lines
2. Discretization of the differential equation in one or
two coordinate direction
3. Transformation to obtain decoupled ordinary differ
ential equations
4. Introduction of the boundary conditions and solu
tion of the equations.
2.1. Partitioning the Solution Region into Lines (Step 1)
The method of lines is particularly suitable for modeling a
wide range of transmission lines and integrated struc
tures with multiple layers [4–6]. This involves in the most
general case discretizing the Helmholtz wave equation in
two directions while the other direction is treated analy
tically. We will illustrate the method for planar integrated
structures.
In this section we describe a formulation of the electro
magnetic problem in an arbitrary general isotropic med
ium leading, besides, to the decoupling of the Maxwell’s
equations describing the electromagnetic ﬁeld and to an
equivalentcircuit representation of the layered medium.
The geometry of the problem and coordinate deﬁnitions
of the structure are given in Fig. 1, where elementary
current sources are shown and/or a plane wave is incident
on a multilayer generalized planar structure. First, we
consider the ith layer, which is not inﬁnitely large in the
planar direction (x–z plane) and the homogeneous mate
rial is characterized via the constitutive relation proper of
the dielectric medium considered.
Starting from Maxwell’s equations, the whole ﬁeld may
be obtained from the component in the ^ zzdirection E
z
and
H
z
. For these components the following wave equations
hold
@
2
c
@x
2
÷
@
2
c
@y
2
÷
@
2
c
@z
2
÷k
2
c=0 (1)
where c=E
z
, H
z
, k
2
=o
2
me and m and e are the magnetic
permeability and the electric permittivity, respectively.
The other four components can be expressed in terms of
E
z
and H
z
as follows
@
2
@z
2
÷k
2
_ _
E
x
E
y
_ _
=
@
2
@x @z
÷jkZ
@
@y
@
2
@y @z
jkZ
@
@x
_
¸
¸
¸
_
_
¸
¸
¸
_
E
z
H
z
_ _
(2)
@
2
@z
2
÷k
2
_ _
H
x
H
y
_ _
=
j
k
Z
@
@y
@
2
@x@z
÷j
k
Z
@
@x
@
2
@y@z
_
¸
¸
¸
_
_
¸
¸
¸
_
E
z
H
z
_ _
(3)
where
Z =
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
m
0
=e
_
=Z
0
=
ﬃﬃﬃﬃ
e
r
_
Z
0
=
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
m
0
=e
0
_
_
X
Z
x
y
z y
+
H Lines
E Lines
∧
∧
∧
∧
∧
∧
Figure 1. Geometry of the electromagnetic problem.
METHOD OF LINES 2549
It should be noted that the wave equation (1) and the ﬁeld
relations (2) and (3) are still valid also for inhomogenous
isotropic media characterized by constitutive parameters
m =m
0
and e =e
0
e
r
(y).
Since we are considering the most general case of
application of the method of line, here, for a solution of
the wave equation (1) and for determination of the ﬁeld
components, a discretization along both the ^ xx and the ^ zz
directions is performed. The discretization is performed
using two different sets of lines: the so called e lines and h
lines. The smallest separation between two e lines is s,
while the smallest separation between an e and an h line
is s/2. Different sets of boundary conditions at the two ends
of the generic layer can be considered (e.g., Neumann and/
or Dirichlet, periodic, absorbing boundary). In Fig. 2 we
have depicted the transverse section of the generic dis
cretized along the ^ xx direction.
If N
x
is the number of the e lines, N
x
is also the number
of the h lines and s is the separation between the lines.
Once the transverse section of the generic layer is dis
cretized, for the components of the electric ﬁeld, we have
E
x
(x; y; z) ÷E
x
(y) = E
1
x
(y); E
2
x
(y); E
3
x
(y) · · · E
N
x
(y)
_ ¸
E
y
(x; y; z) ÷E
y
(y) = E
1
y
(y); E
2
y
(y); E
3
y
(y) · · · E
N
y
(y)
_ _
E
z
(x; y; z) ÷ E
z
(y) = E
1
z
(y); E
2
z
(y); E
3
z
(y) · · · E
N
z
(y)
_ ¸
_
¸
¸
¸
_
¸
¸
¸
_
(4)
where, for instance, E
x
(y) is the vector whose elements are
the values of the ^ xx component of the electric ﬁeld sampled
on the e lines. The discretization for the magnetic ﬁeld
is dual.
2.2. Discretization of Differential Equation in One or Two
Coordinate Directions (Step 2)
The ﬁrstorder derivative operators are replaced by differ
ence operators that can be described by the Kro¨neker
products (Q) of the difference operators for the longitudi
nal (d
m
z
) or azimuthal (d
m
y
) line system and the unit
matrices I
x
and I
z
of order N
x
N
x
and N
z
N
z
, respec
tively:
D
m
z
=d
m
z
QI
x
D
m
x
=I
z
Qd
m
x
with m=e; h
_
(5)
The formal expression of the difference operators d
m
z
and
d
m
x
depends on the boundary conditions used. In the case
of Neumann and/or Dirichlet conditions they are given in
Ref. 3, in the case of periodic boundary condition they are
also given in Ref. 3, and for the case of absorbing boundary
conditions they are given in Ref. 7.
For the secondorder derivatives we directly have
DD
m
z
= ÷ d
m
z
QI
x
_ _
.
d
m
z
QI
x
_ _
t
= ÷ d
m
z
QI
x
_ _
.
d
m;t
z
QI
t
x
_ _
= dd
m
z
QI
x
_ _
DD
m
x
= ÷ I
z
Qd
m
x
_ _
.
I
z
Qd
m
x
_ _
t
= ÷ I
z
Qd
m
x
_ _
.
I
t
z
Qd
m;t
x
_ _
= I
z
Qdd
m
x
_ _
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
(6)
It can now be easily shown that the vector c=E
z
; H
z
satisﬁes the following set of coupled differential equations:
@
2
c
@y
2
÷DD
e;h
x
+ c÷DD
e;h
z
+ c÷k
2
c=0 (7)
2.3. Transformation to Obtain Decoupled Ordinary
Differential Equations (Step 3)
The next step is to solve analytically the equations result
ing along the y coordinate. To solve (7) analytically, we
need to obtain a system of uncoupled ordinary differential
equations from the coupled equation (7). To achieve this,
we deﬁne the transformed vector c
+
by letting
c=T
e;h
+ c
+
(8)
and requiring
(T
e;h
)
÷1
+ DD
e;h
x
+ T
e;h
=L
2
x
(T
e;h
)
÷1
+ DD
e;h
z
+ T
e;h
=L
2
z
_
_
_
(9)
where the elements of the diagonal matrices L
2
x
; L
2
z
are the
eigenvalues of DD
e;h
x
and DD
e;h
z
, respectively.
In (8) we advanced the following position
T
e;h
=t
e;h
z
Qt
e;h
x
(10)
equaling t
m
f
and t
m
z
, the eigenvector matrices belonging to
secondorder difference operators dd
m
f
and dd
m
z
, respec
tively.
After some algebraic manipulations, it can be shown
that, for homogeneous layers where m and e do not depend
on y, the general solution for the kth component of the
vector c
+
can be expressed as
c
+
k
(y) =A
k
.
cosh(k
y
k
y) ÷B
k
.
sinh(k
y
k
y) (11)
where A
k
and B
k
are integration constants and depend on
the proper boundary condition to be applied.
x
y
h
h
1
e
1
h
2
e
2
h
3
e
3 h
N
h
N−1
h
N−2
e
N
e
N−1
e
N−2
∧
∧
Figure 2. Transverse section of the generic
layer of height d discretized along the ^ xx direc
tion (—e lines;     h lines).
2550 METHOD OF LINES
From the general solution (11) we can relate to each
other the ﬁeld solution at the interfaces in y
p
and y
p÷1
of
the generic pth layer as follows:
E
+
/
zk
(y
p÷1
)
E
+
/
zk
(y
p
)
_
¸
_
_
¸
_=
fe
(1;1)
k
(y
p
; y
p÷1
)
fe
(2;1)
k
(y
p
; y
p÷1
)
fe
(1;2)
k
(y
p
; y
p÷1
)
fe
(2;2)
k
(y
p
; y
p÷1
)
_
¸
¸
_
_
¸
¸
_
+
E
+
zk
(y
p÷1
)
E
+
zk
(y
p
)
_
_
_
_
(12)
H
+
/
zk
(y
p÷1
)
H
+
/
zk
(y
p
)
_
¸
_
_
¸
_=
fh
(1;1)
k
(y
p
; y
p÷1
)
fh
(2;1)
k
(y
p
; y
p÷1
)
fh
(1;2)
k
(y
p
; y
p÷1
)
fh
(2;2)
k
(y
p
; y
p÷1
)
_
¸
¸
_
_
¸
¸
_
+
H
+
zk
(y
p÷1
)
H
+
zk
(y
p
)
_
_
_
_
(13)
In (12) and (13) the apex prime stands for ﬁrst derivative
with respect the variable y and we have stated the
following positions:
fe
(1;1)
k
(y
p
; y
p÷1
) = ÷k
y
k
.
coth[(y
p
÷y
p÷1
)
.
k
y
k
]
fe
(1;2)
k
(y
p
; y
p÷1
) =k
y
k
.
csch[(y
p
÷y
p÷1
)
.
k
y
k
]
fe
(2;1)
k
(y
p
; y
p÷1
) = ÷k
y
k
.
csch[(y
p
÷y
p÷1
)
.
k
y
k
]
fe
(2;2)
k
(y
p
; y
p÷1
) =k
y
k
.
coth[(y
p
÷y
p÷1
)
.
k
y
k
]
_
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
_
(14)
fh
(1;1)
k
(y
p
; y
p÷1
) =
÷k
y
k
.
coth[(y
p
÷y
p÷1
)
.
k
y
k
]
k
y
k
fh
(1;2)
k
(y
p
; y
p÷1
) =
k
y
k
.
csch[(y
p
÷y
p÷1
)
.
k
y
k
]
k
y
k
fh
(2;1)
k
(y
p
; y
p÷1
) =
÷k
y
k
.
csch[(y
p
÷y
p÷1
)
.
k
y
k
]
k
y
k
fh
(2;2)
k
(y
p
; y
p÷1
) =
k
y
k
.
coth[(y
p
÷y
p÷1
)
.
k
y
k
]
k
y
k
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
(15)
Now we derive from (12) and (13) the discretized trans
verse ^ xx component of the electric and magnetic ﬁelds in
terms of the E
z
and H
z
components as follows:
E
x
(y) = [D
h
z
.
D
e
z
÷o
2
meI]
÷1
.
[D
h
z
.
D
e
x
.
T
e
.
E
+
z
(y) ÷jom
.
T
h
.
H
+
z
/
(y)]
H
x
(y) = [D
e
z
.
D
h
z
÷o
2
meI]
÷1
.
[D
e
z
.
D
h
x
.
T
h
.
H
+
z
(y) ÷joe
.
T
e
.
E
+
/
z
(y)]
_
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
_
(16)
It can now be shown that for the two interfaces A and B in
y
p
and y
p÷1
, respectively, of a generic layer the following
relation in matrix form holds:
H
B
H
A
_ _
=
Y
(a;b)
11
Y
(a;b)
12
Y
(a;b)
21
Y
(a;b)
22
_
_
_
_
+
E
B
E
A
_ _
(17)
where
E
A
=
E
x
(a)
E
z
(a)
_
_
_
_
E
B
=
E
x
(b)
E
z
(b)
_
_
_
_
_
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
_
;
H
A
=
H
x
(a)
H
z
(a)
_
_
_
_
H
B
=
H
x
(b)
H
z
(b)
_
_
_
_
_
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
_
Y
(a;b)
11
Y
(a;b)
12
Y
(a;b)
21
Y
(a;b)
22
_
_
_
_
=B+ Z + A
÷1
A=
a
.
T
e
c
.
T
h
0 0
T
e
0 0 0
0 0 a
.
T
e
c
.
T
h
0 0 T
e
0
_
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
_
B=
b
.
T
h
d
.
T
e
0 0
T
e
0 0 0
0 0 b
.
T
h
d
.
T
e
0 0 T
h
0
_
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
_
Z=
0 /h
(1;1)
(a; b) 0 /h
(1;2)
(a; b)
/e
(1;1)
(a; b) 0 /e
(1;2)
(a; b) 0
0 /h
(2;1)
(a; b) 0 /h
(2;2)
(a; b)
/e
(2;1)
(a; b) 0 /e
(2;2)
(a; b) 0
_
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
_
a = D
h
z
.
D
e
z
÷o
2
meI
_ _
÷1
.
D
h
z
.
D
e
x
c =jom D
h
z
.
D
e
z
÷o
2
meI
_ _
÷1
b = D
e
z
.
D
h
z
÷o
2
meI
_ _
÷1
.
D
e
z
.
D
h
x
c = ÷jom D
e
z
.
D
h
z
÷o
2
meI
_ _
÷1
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
Equation (17) gives the transfer equation for the generic
layer. The transfer equation for a number of layers is
obtained by multiplication of the transmission matrices
proper of the single layers.
Special attention must be paid to the upper isotropic
halfspace. For the radiation condition, we have
c
+
k
(y) =C
k
.
e
÷k
y
k
y
(18)
METHOD OF LINES 2551
In this case we can write
H
F
=Y
(c)
t
+ E
F
(19)
where
Y
(c)
t
=B
(c)
Z
(c)
(A
(c)
)
÷1
A
(c)
a T
e
c T
h
T
e
0
_
¸
_
_
¸
_
B
(c)
b T
h
d T
e
T
h
0
_
¸
¸
_
_
¸
¸
_
Z
(c)
0 fe(y
N
)
fh(y
N
) 0
_
¸
_
_
¸
_
fe(y
N
) and fh(y
N
) are diagonal matrices whose elements
are given by ÷(1=k
y
k
) and ÷k
y
k
, respectively.
The method of lines can be used to analyze homoge
neous and inhomogeneous cylindrical transmission struc
tures [8–10] and circular and elliptic waveguides [11]. The
principal steps involved in applying MoL in cylindrical
coordinates are the same as in Cartesian coordinates.
2.4. Introduction of Boundary Conditions and Solution
of Equations (Step 4)
Here, we illustrate with the use of MoL, an analysis of the
electromagnetic characteristics of planar integrated struc
tures.
In this section we apply the theory described in the
previous sections to two practical geometries. The ﬁrst one
is a microstrip patch antenna fed by a coplanar microstrip
line as shown in Fig. 3. The second one is a layered
structure with three layers in the presence of an aperture
between layers 2 and 3 (Fig. 4). We give two examples in
Sections 2.4.1. and 2.4.2.
2.4.1. Coplanar Microstrip Line. Two different regions
are described by the following transmission matrices:
Region 1 (y
a
ryry
b
):
H
B
H
A
_ _
=
Y
(y
a
;y
b
)
11
Y
(y
a
;y
b
)
12
Y
(ya;y
b
)
21
Y
(ya;y
b
)
22
_
_
_
_
+
E
B
E
A
_ _
(20)
Region 2 (y
b
ryr
N
):
H
F
= Y
(y
b
)
t
_ _
E
F
(21)
On the ground plane we have
E
÷
A
=0 (22)
On the ﬁrst in y =y
b
we must impose the continuity of the
tangential components of the electric ﬁeld. The magnetic
ﬁeld shows a discontinuity due to the presence of electric
impressed sources (J
imp
) on the feedline and of induced or
scattered currents on the metallic path (J
S
):
H
(patch)
F
÷H
(patch)
B
=J
S
H
(feed)
F
÷H
(feed)
B
=J
imp
_
_
_
(23)
Moreover, on the metallic patches the tangential compo
nents of the electric ﬁeld must vanish:
E
÷
F
¸
¸
¸
patch
=0 (24)
From (20) and (22) we get
H
÷
B
=Y
(ya;y
b
)
11
+ E
÷
B
(25)
Region 2
Region 1
B ≡ F
A
Figure 3. Geometry of the layered structure.
Region 2
Region 3
Region 4
Region 1
F ≡ G
D ≡ E
B ≡ C
A
Figure 4. Geometry of the layered structure.
2552 METHOD OF LINES
From the ﬁrst equation of (23) and from (21) and (25), we
can write
H
(patch)
B
H
(feed)
F
_
¸
_
_
¸
_=
H
(patch)
F
H
(feed)
B
_
¸
_
_
¸
_÷
÷J
S
J
imp
_
_
_
_
=
Y
(y
b
)
t
0
0 Y
(ya;y
b
)
11
_
¸
_
_
¸
_
E
(patch)
F
E
(feed)
B
_
¸
_
_
¸
_÷
÷J
S
J
imp
_
_
_
_
(26)
and thus
Y
(y
a
;y
b
)
11
Y
(y
a
;y
b
)
12
Y
(y
a
;y
b
)
21
Y
(y
a
;y
b
)
22
_
¸
_
_
¸
_ +
E
(patch)
F
E
(feed)
B
_
¸
_
_
¸
_
=
Y
(y
c
)
t
0
0 Y
(y
a
;y
b
)
11
_
¸
_
_
¸
_ +
E
(patch)
F
E
(feed)
B
_
¸
_
_
¸
_÷
÷J
S
J
imp
_
_
_
_
or, in a more compact form
(A
11
÷A
12
+ A
÷1
22
+ A
21
) + E
÷
F
÷ J
÷
S
= ÷A
12
+ A
÷1
22
+ J
÷
imp
(27)
with
A
11
=Y
(y
a
;y
b
)
11
÷Y
(y
b
)
t
A
22
=Y
(y
a
;y
b
)
22
÷Y
(y
a
;y
b
)
11
_
_
_
;
A
12
=Y
(y
a
;y
b
)
12
A
21
=Y
(ya;y
b
)
21
_
_
_
It is important to note that in Eq. (27) E
F
has to be set to
zero in the metallic regions, where only J
S
and J
imp
have
components and J
S
and J
imp
must to be set to zero above
the dielectric interface, where, on the contrary, only E
F
has components.
Thanks to this property, all those columns of the matrix
A
11
÷A
12
+ A
÷1
22
+ A
21
that multiply for column void values
of E
F
can be replaced by columns of zeros. Therefore, to
obtain a solution both for E
F
and J
S
, it sufﬁces to operate a
sort of ‘‘fusion’’ of the two vectors in an only one (the vector
X), and to rewrite Eq. (27) as the following linear, matrix
equation:
A + X=B+ J
imp
(28)
It is worth remarking that if we change the location and
the dimensions of the radiating patches and the impressed
source, we have only to compute again the vectors X and
J
imp
(which is not timeconsuming) and numerically solve
the same linear, matrix equation. So, the MoL analysis
technique presented here is found to be sufﬁciently versa
tile to handle important practical conﬁgurations such as
stacked patches, patches with superstrates, without
[unlike e.g., those for the moment method (MoM)] the
expense of very long computer runtimes. We should also
note that the computational efﬁciency is high, and thanks,
to the reduced system (28), a smaller number of discreti
zation lines is necessary.
2.4.2. Layered Structure with Three Layers in Presence of
an Aperture. In this case four different regions are de
scribed by the following transmission matrices:
Region 1:
H
B
H
A
_ _
=
Y
(1;1)
1
Y
(1;2)
1
Y
(2;1)
1
Y
(2;2)
1
_
_
_
_
E
B
E
A
_ _
(29)
Region 2:
H
D
H
C
_ _
=
Y
(1;1)
2
Y
(1;2)
2
Y
(2;1)
2
Y
(2;2)
2
_
_
_
_
E
D
E
C
_ _
(30)
Region 3:
H
F
H
E
_ _
=
Y
(1;1)
3
Y
(1;2)
3
Y
(2;1)
3
Y
(2;2)
3
_
_
_
_
E
F
E
E
_ _
(31)
Region 4:
H
G
=Y
G
.
E
G
(32)
The boundary conditions are
Interface 1:
E
A
=0 (33)
Interface 2:
H
C
÷H
B
=J
imp
H
C
=H
B
_
(34)
Interface 3:
H
E
÷H
D
=J
S1
E
E
¸
¸
slot
=E
D
¸
¸
slot
E
E
¸
¸
ground
=0
_
¸
¸
_
¸
¸
_
(35)
Interface 4:
H
G
÷H
F
=J
S2
E
G
¸
¸
slot
=E
F
¸
¸
slot
E
G
¸
¸
patch
=0
_
¸
¸
_
¸
¸
_
(36)
From (29) and (33) we can write
H
B
=Y
(1;1)
1
+ E
B
(37)
METHOD OF LINES 2553
From (35), (36), and (31) we have
H
F
H
E
_ _
=
H
G
H
D
_ _
÷
÷J
S2
J
S1
_ _
= =
Y
(1;1)
3
Y
(1;2)
3
Y
(2;1)
3
Y
(2;2)
3
_
_
_
_
E
G
E
D
_ _
(38)
and then
Y
G
.
E
F
H
D
_ _
÷
÷J
S2
J
S1
_ _
=
Y
(1;1)
3
Y
(1;2)
3
Y
(2;1)
3
Y
(2;2)
3
_
_
_
_
E
F
E
E
_ _
(39)
From (34), (35), and (30) it follows that
H
D
H
C
_ _
=
H
E
H
B
_ _
÷
÷J
S1
J
imp
_ _
= =
Y
(1;1)
2
Y
(1;2)
2
Y
(2;1)
2
Y
(2;2)
2
_
_
_
_
E
D
E
C
_ _
(40)
and then
H
E
Y
(1;1)
2
.
E
B
_ _
÷
÷J
S1
J
imp
_ _
=
Y
(1;1)
2
Y
(1;2)
2
Y
(2;1)
2
Y
(2;2)
2
_
_
_
_
E
D
E
B
_ _
(41)
Finally, from (39) and (41), after some algebraic manip
ulations, we derive the following linear system, which,
when solved, gives the unknown electric ﬁeld on the slot
and the induced currents J
S1
and J
S2
on the metalliza
tions at interfaces 3 and 4, respectively:
P
1
E
F
÷P
2
E
E
÷J
S1
= ÷P
3
J
imp
Q
1
E
F
÷Q
2
E
E
÷J
S1
=0
_
(42)
Once we solve Eq. (42), the current distributions on the
metallizations and the electric and magnetic ﬁelds can be
computed in every region of the structure.
BIBLIOGRAPHY
1. K. G. Gupta, R. Garg, and I. J. Bahl, Microstrip Lines and
Slotlines, Artech House, Dedham, MA, 1979.
2. R. K. Hoffmann, Integrierte Mikrowellenschaltungen,
SpringerVerlag, Berlin, 1983.
3. T. Itoh, Numerical Techniques for Microwave and Millimeter
Wave Passive Structures, Wiley, New York, 1989.
4. R. Pregla, Analysis of planar microwave structures on mag
netized ferrite substrate, Arch. Elek. Ubertragung. 40:
270–273 (1986).
5. R. Pregla, About the nature of the method of lines, Arch. Elek.
Ubertragung. 41(6):368–370 (1987).
6. R. Pregla, Higher order approximation for the difference
operators in the method of lines, IEEE Microwave Guided
Wave Lett. 5(2):53–55 (1995).
7. B. Engquist and A. Majda, Absorbing boundary conditions for
the numerical simulations of waves, Math. Comput. 31:629–
651 (1977).
8. S. Xiao et al., Analysis of cylindrical transmission lines with
the method of lines, IEEE Trans. Microwave Theory Tech.
44(7):993–999 (1996).
9. R. Pregla, General formulas for method of lines in cylindrical
coordinates, IEEE Trans. Microwave Theory Tech.
43(7):1617–1619 (1995).
10. M. Thorburn et al., Application of the method of lines to
cylindrical inhomogenous propagation structures, Electron.
Lett. 26(3):170–171 (1990).
11. K. Wu and R. Vahldieck, The method of lines applied to planar
transmission lines in circular and elliptical waveguides, IEEE
Trans. Microwave Theory Tech. 37(12):1958–1963 (1989).
METHOD OF MOMENTS
AHMED A. KISHK
University of Mississippi
University, Mississippi
1. INTRODUCTION
A general electromagnetic problem of an object in un
bounded media such as a scatterer or antenna is a complex
mathematical boundary value problem, which so far has
resisted exact analytical treatment, except in such special
cases as circular cylinders [1], spheres [2], and spheroids
[3]. In these cases the surface of the object coincides with
one of the coordinate surfaces of an orthogonal coordinate
system, in which the vector wave equation can be solved
by the method of separation of variables. However, for an
object of arbitrary shape, such an analytic solution is not
feasible and an approximate or numerical solution must
be obtained. The problem can, in general, be formulated
using the integral equation method involving the surface
ﬁeld distributions or the differential equation method
involving the ﬁeld distribution throughout the volume.
The equations describing the problem are exact and can
generally be given as
L(f ) =g (1)
where L is an operator, f is the unknown response of the
system to be determined, and g is the known source. If
the solution to (1) satisﬁes the boundary conditions of the
problem, then f is a unique solution to the problem due to
the source g.
The process of solving (1) might require some approx
imations, and the accuracy of the solution will depend on
the numerical technique and the number of samples used
to describe the structure. Here we are interested in the
method of moments as one of the most widely used
numerical techniques. The method of moments, in gen
eral, reduces the linear operator equation (1) to a system
of ﬁnite linear matrices that can be solved numerically to
obtain the solution. The technique can be used in the
frequency domain or the in the time domain. Here, we will
consider the formulation of the problem using integral
equations in the frequency domain.
The method of moment process is simple and straight
forward. In electromagnetics, the physical problem is
speciﬁed by Maxwell’s equations and the boundary
2554 METHOD OF MOMENTS
conditions are reduced to integral equations. The un
knowns (f) are under the integral, which is the operator
(L). The unknowns are expanded as a series of known
basis functions and unknown constant coefﬁcients.
The method of moment can be understood easily if we
draw an analogy with the numerical integration of a
function. Let
F(x) =
_
b
a
f (x
/
)G(x; x
/
)dx
/
(2)
where F(x) is unknown and the integrand f (x
/
)G(x; x
/
) is
known. If the integration is performed numerically using
the simple rectangular rule, the interval [a,b] can be
divided to N segments of Dx
/
i
. For generality, the segments
are considered unequal as shown in Fig. 1. Then (2) can be
approximated as
F(x) =
N
i =1
f (x
i
)G(x; x
i
)Dx
/
i
(3)
where x
i
is the middle of the segment i and f(x
i
) is the
value of the function at x
i
. Equation (2) can be considered
an integral equation when F(x) is known as the source or
the excitation and f(x) is the unknown function. Then if
the unknown function can be expanded as
f (x
/
) =
N
i =1
a
i
b
i
(x
/
) (4)
where the b
i
values denote known functions and the a
i
values represent unknown constant coefﬁcients. Substi
tute (4) in the integral equation (2) we obtain
_
b
a
N
i =1
a
i
b
i
(x
/
)G(x; x
/
)dx
/
=F(x) (5)
Because the a
i
values are constants, we can write (5) as
N
i =1
a
i
_
b
a
b
i
(x
/
)G(x; x
/
)dx
/
=F(x) (6)
Now we have the completely known integrant, which can
be integrated analytically or numerically. To make the
analogy with the numerical integral using the rectangular
rule, let
b
i
(x
/
) =
1 x
/
=x
i
0 elsewhere
_
(7)
and thus (6) becomes
N
i =1
a
i
G(x; x
/
)Dx
i
=F(x) (8)
which is similar to (3), and one can see that a
i
=f(x
i
).
However, Eq. (8) is one equation in N unknowns. If we
evaluate (8) at each x
i
, we obtain the following system of
equations:
a
1
G(x
1
; x
1
)Dx
1
÷a
2
G(x
1
; x
2
)Dx
2
÷ · · · ÷a
N
G(x
1
; x
N
) =F(x
1
)
a
1
G(x
2
; x
1
)Dx
1
÷a
2
G(x
2
; x
2
)Dx
2
÷ · · · ÷a
N
G(x
2
; x
N
) =F(x
2
)
.
.
.
a
1
G(x
N
; x
1
)Dx
1
÷a
2
G(x
N
; x
2
)Dx
2
÷ · · · ÷a
N
G(x
N
; x
N
) =F(x
N
)
(9)
Equation (9) can be expressed in matrix form as
G(x
1
; x
1
)Dx
1
G(x
1
; x
2
)Dx
2
· · · G(x
1
; x
N
)
G(x
2
; x
1
)Dx
1
G(x
2
; x
2
)Dx
2
· · · G(x
2
; x
N
)
.
.
.
.
.
.
. . .
.
.
.
G(x
N
; x
1
)Dx
1
G(x
N
; x
2
)Dx
2
· · · G(x
N
; x
N
)
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
a
1
a
2
.
.
.
a
N
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
=
F(x
1
)
F(x
2
)
.
.
.
F(x
N
)
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
(10)
or in a compact form as
[Z][I] =[V] (11)
The Z is a square matrix, I is the unknown vector, and Vis
the known excitation vector. The solution of (11) will
provide the unknown vector.
2. SIMPLE ELECTROMAGNETIC SCATTERING PROBLEM
To illustrate, how an electromagnetic problem can be
formulated, consider an arbitrarily shaped conducting
twodimensional scatterer. A twodimensional body is
inﬁnite in the axial direction (here considered to be the z
direction) and has the same physical cross section in any
a
x
1
b
x
2
x
N
x
i
∆x
i
y
f (x)G(x,x ′)
x
Figure 1. Representation of numerical integration using the
rectangular rule.
METHOD OF MOMENTS 2555
plane cut, which is orthogonal to the axis (any plane cut
parallel to the xy plane) as illustrated in Fig. 2. The
scatterer is bounded by the surface S and surrounded by
the free space, which is characterized by the permittivity
and permeability of the free space (e
0
, m
0
). The excitation is
an electromagnetic plane wave of the incident ﬁelds E
inc
and H
inc
. The total electric and magnetic ﬁelds in the
exterior region are denoted by E
0
and H
0
, respectively.
Inside the conducting object the ﬁelds are known and of
zero value. The normal unit vector on the surface S is
n
_
= t
_
z
_
where z
_
and t
_
are the unit vectors in the
direction of z and tangential to the surfaces S parallel
to the xy plane. The time variation e
jot
is implied and
suppressed throughout.
The incoming wave in the 2D case always travels on a
ray, which is orthogonal to the axis of the body. Because of
this, an incoming wave of arbitrary polarization can al
ways be written as the sum of waves, which are transverse
electric (TE) and transverse magnetic (TM) to the axis of
the body. A wave that is transverse magnetic has electric
ﬁeld components only in the z direction and a magnetic
ﬁeld with no zdirected components. The currents, which
result from the arbitrarily polarized waves, can be ob
tained by summing the currents that result from the TE
portion of the wave with those that result from the TM
portion of the wave.
In order to use the freespace Green function in the
solution of this problem, we have to have an unbounded
homogeneous region. The surface equivalence principle is
usually used to create an equivalent problem that con
verts the original problem from inhomogeneous region to a
homogeneous region, but reserves the same boundary
conditions of the original problem on the surface S. In
the equivalent problem, the conducting object is removed
and ﬁlled with the same materials of the exterior region of
the original problem and forces the ﬁelds within the
boundary of S to be zero and the ﬁelds outside S to be
the same ﬁelds and materials of the original problem.
Since the equivalent problem represents an unbounded
homogeneous region, the ﬁelds must be continuous every
where in the space, but by forcing the ﬁelds inside S to be
zero, we have created discontinuity of the ﬁelds that must
be compensated by inserting sources on the surface S.
These sources are electric and magnetic surface currents
that are proportional to the exterior tangential ﬁelds on S.
The surface electric and magnetic currents are
J = n
_
H
0
and M= ÷n
_
E
0
. Since the surface of the
original problem is a perfect conducting surface, the
tangential electric ﬁelds to the surface are zero. Therefore,
the equivalent surface magnetic current M must be zero.
Now, Fig. 3 illustrates the equivalent problem. The total
ﬁelds in the exterior region are due to the equivalent
electric current on the surface and the incident ﬁelds. The
ﬁelds due to the equivalent electric current (E
sc
; H
sc
) are
referred to as the scattered ﬁelds. It should be clear that
the equivalent surface electric current is unknown and
should be determined in order to compute the ﬁelds
(E
sc
; H
sc
).
To formulate the integral equation, the boundary con
dition is applied on the surfaces S. The boundary condition
is expressed as
E
0
[
tan
= ÷n
_
(n
_
E
0
) =0 on S (12)
For simplicity, we consider the TM case. As mentioned
above, for the TM case, only E
z
exists and the magnetic
ﬁeld is transverse to z with no component in the z direc
tion. Therefore, (12) can be written as
E
sc
z
= ÷E
inc
z
on S (13)
In an integral equation form, (13) can be written as
÷jom
_
C
J
z
(r
/
)g(r; r
/
)d‘ = ÷E
inc
z
(r) on S (14)
where C is the contour of the surface S along the tangen
tial direction t and (q; q
/
) are the position vectors in the xy
plane that identify the ﬁeld and source points, respec
tively. The contour integral proceeds in the direction of the
tangential direction t
_
. The function g is the Green func
tion, which is given in terms of the Hankel function of zero
order and second type as
g(r; r
/
) =
1
4j
H
(2)
0
(k
0
[r ÷r
/
[) (15)
where k
0
is the wavenumber. The planewave ﬁelds are
given as
E
inc
z
=A
0
e
÷jk
0
k
_
·q
and H
inc
=(k
_
z
_
)
A
0
Z
0
e
÷jk
0
k
_
·q
(16)
o = ∞
n
)
t
)
s
x
y
j
E
0
= E
inc
+ E
sc
,H
0
=H
inc
+ H
sc
c
0
, j
0
Figure 2. Original scattering problem.
Zero fields
S
J
c
0
, j
0
E
0
= E
inc
+ E
sc
, H
0
=H
inc
+ H
sc
n
)
c
0
, j
0
Figure 3. Surface equivalent problem.
2556 METHOD OF MOMENTS
where A
0
is the amplitude of the incident electric ﬁeld and,
k
_
= ÷ x
_
cos j
inc
÷ y
_
sin j
inc
is the unit vector in the direc
tion of the planewave propagation, which is forming an
angle j
inc
with the x axis. This integral equation is
referred to as an electric ﬁeld integral equation (EFIE).
To solve the surface integral equation (14), the contour
of the scattering body is divided into a number of linear
segments N with width DC. The end points of the seg
ments lie on the actual contours of the body. The center of
the segment i is located at r
i
. The length of the zones is
taken to be less than onetenth of a wavelength. The
currents are expanded in pulse basis functions multiplied
by unknown coefﬁcients as
J
z
=
N
i =1
I
i
P
i
(17)
where I
i
are the unknown constant current coefﬁcients
and P
i
are the pulse basis functions, that deﬁned as
P
i
(r
/
) =
1 over the segment i
0 elsewhere
_
(18)
Substitute (17) in (14). Then the point matching technique
is employed to reduce the integral equations to a system of
linear equations. Equation (14) reduces to a standard
matrix element, which is placed in the proper location in
the matrix. For each point matching j, equation (14) can
be written as
÷
om
0
4
N
i =1
I
i
_
DC
i
H
2
0
(k
0
[r
j
÷r
/
[)d‘
= ÷A
0
e
jk
0
(x
j
cos j
inc
÷y
j
sinj
inc
)
on DC
j
(19)
where (x
j
, y
j
) are the rectangular coordinates of the middle
of the segment j and j
inc
is the angle of the incident plane
wave measured from the x axis. Now, the matrix element
Z
ji
and the excitation vector element V
j
can be written as
Z
ji
= ÷
om
0
4
_
DC
i
H
2
0
(k
0
[r
j
÷r
/
[)d‘ and
V
j
= ÷A
0
e
jk
0
(x
j
cos j
inc
÷y
j
sinj
inc
)
(20)
In the method of moments, the diagonal elements of the
matrix should always have special treatment because the
integrant becomes singular at r
j
=r
/
. The singular term
can be subtracted and integrated analytically, or we may
use the smallargument approximation for the Green
function. When the argument of the Hankel function is
very small, it can be approximated as
H
(2)
0
(k
0
Dr) ’ 1 ÷j
2
p
ln(1:7811k
0
Dr=2) (21)
Now the selfterm (i =j) can be written as
Z
jj
= ÷
om
0
DC
j
4
1 ÷j
2
p
ln
1:781k
0
DC
j
4e
_ _ _ _
(22)
At this point, the MoM matrix is completely known. The
solution of the matrix determines the current coefﬁcients
on the surface of the scattering body. Then these coefﬁ
cients are used to obtain the near or the far scattered
ﬁelds. For the far ﬁelds, farﬁeld approximations can be
considered, and the far scattered ﬁeld after using the
largeargument approximations of the Hankel function
can be expressed as
E
sc
z
(r; j) = ÷om
0
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
1
2prk
0
¸
e
jp=4
e
÷jk
0
r
N
i =1
I
i
DC
i
e
jk
0
r
_
·r
i
(23)
An important measure of the ﬁelds scattered by an
object is the radar cross section (RCS). To determine the
RCS, the scattering object is illuminated by a uniform
plane wave. For the TM case the radar cross section s is
deﬁned by the following equation:
s(f) = lim
r÷o
2pr
[E
sc
z
(r; f)[
2
[E
inc
z
[
2
= lim
r÷o
2pr
[E
sc
z
(r; f)[
2
[A
0
[
2
(24)
In Equation (23), E
z
sc
(r,j) is the z component of the
scattered electric ﬁeld at the point (r,j).
The electric ﬁeld integral equations (EFIE) suffer from
the problem of internal resonance. If we are trying to solve
a waveguide problem with a cross section similar to that of
the scattering problem, the resulting integral equation
will be exactly the same as the integral equation of the
scattering problem. At the cutoff frequencies, the MoM
matrix is singular and an inﬁnite number of solutions
exist for a sourcefree case. At these frequencies, the
scattering problem doesn’t provide a unique solution.
3. POSSIBLE INTEGRAL EQUATIONS
If we consider the equivalent problem in Fig. 3, another
integral equation can be formulated. If we use the mag
netic ﬁeld integral equation in the same way that the
tangential magnetic ﬁeld on the surface S is discontinuous
by the surface equivalent current J
z
, the integral equation
can be written as
n
_
H
0
=J =H
t
=J
z
on S (25)
or in the form
J
z
÷H
sc
t
=H
inc
t
on S (26)
Equation (26) can be written in an operator form as
J
z
÷H
t
(J
z
) =H
inc
t
on S (27)
which is referred to as the magnetic ﬁeld integral equation
(MFIE). The EFIE can be written as
÷
1
Z
0
E
z
(J
z
) =
1
Z
0
E
inc
z
onS (28)
METHOD OF MOMENTS 2557
where H(J) and E(J) are the magnetic ﬁeld and electric
ﬁeld due to the surface electric current J. These operators
will be deﬁned later. The MFIE also suffer from the
internal resonance problem. To avoid the internal reso
nance problem, the EFIE and MFIE are combined
together to form the combined ﬁeld integral equation
(CFIE) as
J ÷H
t
(J
z
) ÷
a
Z
0
E
z
(J)
=H
inc
t
÷
a
Z
0
E
inc
z
on S
(29)
The parameter a is a scalar combination factor less than or
equal to 1. As can be seen, the EFIE is normalized to Z
0
.
This normalization adjusts the value level of the MoM
matrix due to the EFIE part to values comparable to those
of the MoM matrix element of the MFIE in order to obtain
stable numerical inversion of the MM matrix. Note that
all the expressions above are for the TM case of a two
dimensional inﬁnitely extended conducting cylinder and
all the expression are scalar equations. These expressions
can be generalized and presented in vector form. After
that, they should be reduced to scalar form, as will be
described next.
4. GENERAL TWODIMENSIONAL PROBLEM
Consider a composite 2D object parallel to the z axis
present with electric or magnetic harmonic current
sources J and M, respectively. These current sources are
of inﬁnite extent and parallel to the z axis with harmonic
variations in the z direction. Therefore, the total ﬁelds
produced by these harmonic 2D sources must be harmonic
in z as well in order to satisfy the boundary conditions on
the objects [4]. The geometry and notations for such an
object are given in Fig. 4. The whole space is divided into
N÷1 homogeneous regions V
i
, which may be either
dielectric regions with permittivities e
i
and permeabilities
m
i
, or closed conductor regions. These regions are num
bered i =0,1,2,y, N, where i =0 corresponds to the exter
ior region, namely, free space. Lossy materials are
considered by allowing e
i
and m
i
, i =1,2,y, N, to be
complex. Each region V
i
is surrounded by a closed surface
S
i
and associated with an inwarddirected normal unit
vector ^ nn
i
. The surface interface between regions V
i
and V
j
is denoted as S
ij
, iaj. Thus, S
i
constitutes the set of all
interface surfaces S
ij
, where j represents all region num
bers adjacent to region V
i
. Note that S
ij
is the same surface
as S
ji
; however, the normal unit vectors ^ nn
i
and ^ nn
j
are in
opposite directions to each other on the two surfaces. The
concept of the equivalence principle is used to derive a
surface integral equation (SIE) formulation for 2D objects
with N÷1 homogeneous regions. The total ﬁelds in each
homogeneous region are denoted by E
i
and H
i
i =0,1,2,y,
N for the electric and magnetic ﬁelds, respectively. Any
perfect conducting region need not be considered as a
region because the ﬁelds are known to be equal to zero. In
the freespace region V
0
, the total ﬁelds are denoted by
(E
0
, H
0
). From Maxwell’s equations and the equivalence
principle, one can express the ﬁelds in each region in
terms of unknown electric and magnetic equivalent sur
face currents plus the ﬁelds due to the harmonic 2D
sources present in the region.
According to the surface equivalence principle, we can
break the original problem into a number of auxiliary
problems that are equal to the number of the nonperfect
conducting regions. To obtain the ith auxiliary problem,
the boundaries of region V
i
are replaced by equivalent
surface currents radiating in a homogeneous medium
with the constitutive parameters of region V
i
using elec
tric currents for the conductor boundaries and equivalent
electric and magnetic currents for the dielectric bound
aries. The electric and magnetic currents appearing on
opposite sides of a dielectric interface in different auxiliary
problems are taken equal in magnitude and opposite in
direction to ensure the continuity of the tangential com
ponents of the ﬁelds on these boundaries, as in the original
problem. In this procedure the ﬁelds produced by the
equivalent currents within the region boundaries are the
same as those in the original problem, while the zero ﬁeld
is produced outside these boundaries. The electric and
magnetic surface currents along the boundaries are
J
i
= ^ nn
i
H
i
and M
i
= ÷ ^ nn
i
E
i
on S
i
(30)
Both J
i
and M
i
have components in both the longitudinal
^ zz and transverse t
_
directions
J
i
M
i
_ _
=
J
i
z
M
i
z
_ _
^ zz ÷
J
i
t
M
i
t
_ _
^
tt on S
i
(31)
where ^ nn
i
is the unit normal to S
i
. The currents on the
surface S
i
are the sum of the currents on all the bound
aries S
ij
, where jai and j represents the entire region
numbers adjacent to the region V
i
J
i
=
\j
J
ij
onthe boundaries of regioni (32)
and similarly for M
i
. On the conductor boundaries the
magnetic current is zero. We can now obtain the electric
n
1
n
N
n
2
n
0
S
01
V
0
n
i
S
0i
V
1
V
2
V
i
V
N
S
12
S
2i
S
02
S
02
S
N2
S
Ni
c
1
,j
1
c
1
,j
1
c
0
,j
0
c
2
,j
2
c
N
,j
N
S
ON
Figure 4. Cross section of multiregion cylindrical object.
2558 METHOD OF MOMENTS
and magnetic ﬁelds E
i
(r; k
z
) and H
i
(r; k
z
) due to the
electric current J
i
(r
/
; k
z
) by using
E
i
(r; k
z
) = ÷joA
i
(r; k
z
) ÷j
1
oe
i
m
i
VV + A
i
(r; k
z
) (33)
H
i
(r; k
z
) =
1
m
i
VA
i
(r; k
z
) (34)
where the vector potential A is
A
i
(r; k
z
) =
m
i
j4
_
l
/
J
i
(r
/
; k
z
)H
(2)
0
(k
r
[r ÷r
/
[)dl
/
(35)
The V operator is V=V
t
÷jk
z
^ zz; V
t
=(@=@x) ^ xx ÷(@=@y) ^ yy and
k
r
=
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
k
2
÷k
2
z
_
. For each value of k
z
A[ ÷k,k], the far ﬁeld is
radiating on a cone around the z axis with cone angle y =
cos
÷1
(k
z
/k) [4]. Equations (33) and (34) can be given in an
operator form E
i
(J
i
) and H
i
(J
i
). The vector tangential
operator can be presented in terms of their operator
components as
E
i
tan
(J
i
) =[E
i
t
(J
i
t
) ÷E
i
t
(J
i
z
)] t
_
÷[E
i
z
(J
i
t
) ÷E
i
z
(J
i
z
)] ^ zz (36)
H
i
tan
(J
i
) =[H
i
t
(J
i
t
) ÷H
i
t
(J
i
z
)] t
_
÷H
z
(J
i
t
) ^ zz (37)
After some mathematical manipulations the electric and
magnetic ﬁelds due to the electric currents can be ex
pressed in operator form. Expressions are given for E
i
t
(J
i
t
)
E
i
z
(J
i
t
), E
i
t
(J
i
z
), and E
i
z
(J
i
z
), which are the components of the
vector operator E
i
tan
(J
i
) for the tangential E ﬁeld at a point
r on S
i
, and in the same fashion we get H
i
t
(J
i
t
) H
i
z
(J
i
t
), and
H
i
t
(J
i
z
) for the tangential Hﬁeld vector operator H
i
tan
(J
i
).
The electric and magnetic ﬁeld operators due to the
electric current components are expressed as
E
z
(J
t
) =
÷jk
z
k
r
4oe
0
e
jk
z
z
_
C
J
t
H
(2)
1
(k
r
Dr)( t
_
/
+D^ rr)dl
/
(38)
E
t
(J
z
) =
÷jk
r
k
z
4oe
0
e
jk
z
z
_
C
J
z
H
(2)
1
(k
r
Dr)( t
_
+D^ rr)dl
/
(39)
E
z
(J
z
) = ÷
k
2
r
4oe
0
e
jkzz
_
C
J
z
H
(2)
0
(k
r
Dr)dl
/
(40)
E
t
(J
t
) = ÷
e
jkzz
4oe
0
_
C
k
2
J
t
{(t
/
_
+ t
_
)H
(2)
0
(k
r
Dr)
÷
k
r
Dr
H
(2)
1
(k
r
Dr)(t
/
_
+ t
_
)
÷k
r
[k
r
H
(2)
0
(k
r
Dr)
÷
2
Dr
H
(2)
1
(k
r
Dr)]
(t
/
_
+ D^ rr)( t
_
+D^ rr)]dl
/
(41)
H
t
(J
z
) = ÷
jk
r
4
e
jk
z
z
_
C
J
z
H
(2)
1
(k
r
Dr)(n
_
+D^ rr)dl
/
(42)
H
z
(J
z
) =0 (43)
H
z
(J
t
) = ÷
k
r
4j
e
jk
z
z
_
C
J
t
H
(2)
1
(k
r
Dr)(n
/
_
+D^ rr)dl
/
(44)
H
t
(J
t
) = ÷
k
z
4
e
jk
z
z
_
C
J
t
H
(2)
0
(k
r
Dr)( t
_
+ n
/
_
)dl
/
(45)
where H
(2)
1
( ) is the Hankel function of the ﬁrst order and
second type and Dr=[r ÷r
/
[. The hat is used to indicate
the unit vectors. In Eqs. (38)–(45) the prime is used to
indicate the source coordinates. The ﬁelds due to the
magnetic current M
i
(q
/
; k
z
) can be obtained using duality
[5, Chap. 3, Sect. 32]. Now the boundary conditions must
be enforced on each boundary on the object, which can be
used as surface integral equations after substituting the
proper operators as follows
÷[E
i
tan
(J
i
) ÷E
j
tan
(J
j
) ÷E
i
tan
(M
i
) ÷E
j
tan
(M
j
)]
=[E
i
tan
(J
i;inc
) ÷E
i
tan
(M
i;inc
) ÷E
j
tan
(J
j;inc
) ÷E
j
tan
(M
j;inc
)]
on S
ij
(46)
÷ ^ nn
i
[H
i
(J
i
) ÷H
j
(J
j
) ÷H
i
(M
i
) ÷H
j
(M
j
)]
= ^ nn
i
[H
i
(J
i;inc
) ÷H
i
(M
i;inc
) ÷H
j
(J
j;inc
) ÷H
j
(M
j;inc
)]
on S
ij
(47)
if the boundary S
ij
is a perfect conducting boundary and
the region V
j
is a perfect conducting region, Eqs. (46) and
(47) become
÷E
i
tan
(J
i
)
÷E
i
tan
(M
i
) =E
i
tan
(J
i;inc
) ÷E
i
tan
(M
i;inc
) on S
ij
(48)
÷ ^ nn
i
[H
i
(J
i
) ÷H
i
(M
i
)] = ^ nn
i
[H
i
(J
i;inc
) ÷H
i
(M
i;inc
)]
on S
ij
(just inside V
i
)
(49)
Only one of the equations (48) and (49) can be used. If
Eq. (48) is used on the perfect electric conductor (PEC)
with (46) and (47) on the dielectric boundary, the formula
tion is called EPMCHW formulation, and if Eq. (49) is
used just inside the PEC boundary with (46) and (47) on
the dielectric boundary, the formulation is referred to as
HPMCHW [6]. The combined ﬁeld formulation on the
conductor can be obtained if Eqs. (48) and (49) are
combined together and used as one equation with (46)
and (47) [6]. The combined ﬁeld formulation is referred to
as CPMCHW.
Following the method of moments, the object contour
C is divided into N linear segments with length DC
‘
,
‘ =1; 2; . . . ; N as in Ref. 9, and each current component is
expanded into N pulse basis function p
‘
. In equation form,
the unknown currents can be expressed as
J
i
=
N
‘
(I
‘;i
z
t
_
÷I
‘;i
z
^ zz)P
‘
on all S
ij
(50)
METHOD OF MOMENTS 2559
and
M
i
=
N÷N
d
‘ =N÷1
(M
‘;i
t
t
_
÷M
‘;i
z
^ zz)P
‘
only onthe dielectric boundaries
(51)
where I
{‘;i]
{
t
z
]
and M
{‘;i]
{
t
z
]
are the unknown electric and magne
tic current coefﬁcients, respectively. The pulse function is
p
‘
=1 on the subdomain DC
‘
and zero elsewhere. N
d
is the
number of segments on the dielectric boundaries. Substi
tuting (50) and (51) into the operators deﬁned in (38)–(45)
and then satisfying (46) to (47) at the match point (middle
of the segments), the integral equations reduce to a matrix
of order 2(N÷N
d
), which can be written in the form
Z
ij; i
zz
÷Z
ij;j
zz
Z
ij; i
zt
÷Z
ij; j
zt
Y
ij; i
zz
÷Y
ij; j
zz
Y
ij; i
zt
÷Y
ij; j
zt
Z
ij; i
tz
÷Z
ij; j
tz
Z
ij; i
tt
÷Z
ij; j
tt
Y
ij; i
tz
÷Y
ij; j
tz
Y
ij; i
tt
÷Y
ij; j
tt
Y
ij; i
tz
÷Y
ij; j
tz
Y
ij; i
tt
÷Y
ij; j
tt
÷
1
Z
i
Z
ij; i
tz
÷
1
Z
j
Z
ij; j
tz
÷
1
Z
i
Z
ij; i
tt
÷
1
Z
j
Z
ij; j
tt
Y
ij; i
zz
÷Y
ij; j
zz
Y
ij; i
zt
÷Y
ij; j
zt
÷
1
Z
i
Z
ij; i
zz
÷
1
Z
j
Z
ij; j
zz
÷
1
Z
i
Z
ij; i
zt
÷
1
Z
j
Z
ij; j
zt
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
I
ij
z
I
ij
t
M
ij
z
M
ij
t
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
=
V
ij; i
inc
z
÷V
ij; j
inc
z
V
ij; i
inc
t
÷V
ij; j
inc
t
I
ij; i
inc
t
÷I
ij; j
inc
t
I
ij; i
inc
z
÷I
ij; j
inc
z
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
(52)
where Z
ij;i
ab
and Y
ij;i
ab
denote matrix elements obtained from
the operators E
i
a
(J
b
) and H
i
a
(J
b
), respectively, on the sur
face S
ij
from the region i. Therefore, the ﬁrst sufﬁx of the
subscript refers to the ﬁeld component, and the second
sufﬁx of the subscript refers to the current component. The
ﬁrst pair of the superscripts refers to the surface boundary
ij and the second sufﬁx, to the region number i or j. This
matrix is built assuming that all the boundaries are
dielectric boundaries, but on the perfect conducting parts
the columns and the rows that correspond to the magnetic
currents and the magnetic ﬁeld, respectively, which are in
the third and fourth columns and rows, respectively, must
be removed, and in the ﬁrst and second columns of (52) the
parts that correspond to perfect conducting regions i or j
must be forced to zero.
The quantities I
ij
t
, I
ij
z
, M
ij
t
, and M
ij
z
are the unknown
expansion coefﬁcients of the electric and magnetic cur
rents, respectively. The righthandside (RHS) columns
are the excitation vectors, where V
ij;i
t
, V
ij;i
z
, I
ij;i
t
, and I
ij;i
z
denote the electric and magnetic ﬁelds, respectively, on
the surface S
ij
due to all electric and magnetic sources in
the region i. The matrix elements can be written as
Z
ij
zz
=
÷
k
2
r
4k
0
_
DC
j
H
(2)
0
(k
r
Dr
ij
)dt
/
iOj
÷
k
2
r
4k
0
DC
j
1 ÷
2j
p
ln
DC
j
gk
r
4
_ _
÷1
_ _ _ _
i =j
_
¸
¸
¸
¸
_
¸
¸
¸
¸
_
(53)
Z
ij
tz
=
÷
jk
r
k
z
4oe
0
_
DC
j
H
(2)
1
(k
r
Dr
ij
)( t
_
i
+ D^ rr
ij
)dt
/
iOj
0 i =j
_
¸
_
¸
_
(54)
Z
ij
zt
=
÷
jk
z
k
r
4oe
0
_
DC
j
H
(2)
1
(k
r
Dr
ij
)( t
_
j
+ D^ rr
ij
)dt
/
iOj
0 i =j
_
¸
_
¸
_
(55)
Y
ij
tt
=
÷
k
z
4
_
DC
j
H
(2)
0
(k
r
Dr
ij
)( t
_
i
+ n
_
j
)dt
/
iOj
0 i =j
_
¸
_
¸
_
(56)
Z
ij
tt
=
÷
1
4k
0
_
DC
j
{k
2
( t
_
j
+ t
_
i
)H
(2)
0
(k
r
Dr
ij
)
÷
k
r
Dr
ij
H
(2)
1
(k
r
Dr
ij
)( t
_
j
+ t
_
i
)
÷k
r
k
r
H
(2)
0
(k
r
Dr
ij
) ÷
2
Dr
ij
H
(2)
1
(k
r
Dr
ij
)
_ _
( t
_
j
+ D^ rr
ij
)( t
_
i
+ D^ rr
ij
)]dt
/
iOj
2k
r
4k
0
H
(2)
1
k
r
DC
i
2
_ _
÷
k
2
0
DC
i
4k
0
1 ÷
2 j
p
ln
DC
i
gk
r
4
_ _
÷1
_ _
i =j
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
(57)
Y
ij
zz
=0 (58)
Y
ij
zt
=
÷
k
r
4j
_
DC
j
H
(2)
1
(k
r
Dr
ij
)(n
_
j
+ D^ rr
ij
)dt
/
iOj
÷
1
2
i =j
_
¸
¸
_
¸
¸
_
(59)
Y
ij
tz
=
÷
jk
r
4
_
DC
j
H
(2)
1
(k
r
Dr
ij
)(n
_
i
+ D^ rr
ij
)dt
/
iOj
1
2
i =j
_
¸
¸
¸
_
¸
¸
¸
_
(60)
These integrals are performed numerically where
Dr
ij
=[r
i
÷r
/
[ and r
/
for the segment j. The excitation
vector can be obtained in a manner similar to that for the
matrix elements due to known electric and magnetic
current sources. In the case of planewave incidence the
excitation matrix elements are given as
H
i;inc
t
=
E
m
Z
0
cos y
inc
(E
p
sin a
inc
÷E
n
cos a
inc
) + t
_
i
e
jkr(r
_
+r
i
)
(61)
H
i;inc
z
=
1
Z
0
sin y
inc
sin a
inc
e
jk
r
(r
_
+r
i
)
(62)
2560 METHOD OF MOMENTS
E
i;inc
z
=
E
m
Z
0
sin y
inc
cos a
inc
e
jkr(r
_
+r
i
)
(63)
E
i;inc
t
=
E
m
Z
0
(E
p
cos a
inc
÷E
n
sin a
inc
) + t
_
i
e
jk
r
(r
_
+r
i
)
(64)
where the superscript i denotes the middle of segment i
and a
inc
is the angle between E
inc
and the plane of inci
dence (k–z plane).
Once the MoM matrix equation is solved the far ﬁelds
due to the currents on the outer surface boundary, and the
current sources in the exterior region can be computed.
The ﬁeld will be along a cone of halfangle y =arcos(k
z
/k
0
)
around the structure [4].
5. SAMPLE RESULTS
Sample results are obtained using the formulations pre
sented above to show the accuracy of the numerical method
versus the exact solution for circular cylinder objects.
Figure 5 shows the scattered ﬁeld of a coated conducting
cylinder with radii ka=2 and kb=3 and e
r
=4 for TM and
TE polarization due to a plane wave incident by an angle of
451 from the z axis. The numerical solution is compared
with the exact solution. Excellent agreement is observed.
In the preceding sections we have demonstrated the
formulations of the integral equation for simple twodi
mensional objects using the simplest possible basis func
tion. This basis function is a slowly conversion function.
Other basis functions can be used with better conversion
rates. In the following section we will present literature
review for the developments in the method of moments.
6. ADVANCED METHOD OF MOMENTS
Antennas, which usually consist of threedimensional (3D)
radiating elements (sources), can be present with two
dimensional (2D) structures (cylinders of inﬁnite extent).
Such problems can be analyzed by using a spectrum of
twodimensional solutions (S2DS), as described in Refs. 7
and 8. Generally, any antenna consisting of slots or dipoles
in or in the vicinity of 2D structures of arbitrary cross
sectional shape and material combination can be analyzed
by the S2DS technique.
The most signiﬁcant part of an S2DS analysis is to
solve repeatedly the special spectraldomain problem ob
tained by Fouriertransforming the sources in the z direc
tion of the structure. This spectraldomain problem can be
interpreted as a harmonic 2D spatial problem where the
sources (and the resulting ﬁelds) have harmonic z varia
tion of the form exp( ÷jk
z
z), where k
z
=0 corresponds to
the standard 2D problem. This general harmonic 2D
problem must be solved for a large number of values of
the spectral variable k
z
in order to be able to inverse
transform to 3D spatial domain. The radiation pattern,
however, can normally be found directly without inverse
transformation from the 2D harmonic solution [7].
The harmonic 2D problem is conveniently solved for
each k
z
by the method of moments (MoM). Planewave
scattering from 2D composite objects [9,10] and scattering
from an impedance cylinder with arbitrary cross section
under oblique planewave incidence were analyzed, where
the equivalent electric and magnetic currents were solved
for by pulse expansion and point matching.
7. USE OF IMPEDANCE BOUNDARY CONDITIONS
The concept of the impedance boundary condition (IBC),
which was ﬁrst described by Leontovich in 1948 [11], has
long been used in a variety of electromagnetic wave
scattering problems. For closed bodies, it has been shown
that the IBC is a valid approximation to the exact condi
tion whenever (1) the refractive index is large compared
with unity, (2) the total radii of curvature are much
greater than the skin depth, (3) the body is much thicker
than the skin depth, and (4) the spatial variation of the
index of refraction is slow in comparison to the local
wavelength [12–15]. Consequently, the IBC concept has
long been utilized to solve many electromagnetic pro
blems. Garbacz [16] investigated a special case of bistatic
10
5
0
−5
−10
−15
−20
−25
−30
0 60 120 180 240 300 360
A
m
p
l
i
t
u
d
e
(
d
B
)
10
5
0
−5
−10
−15
−20
−25
−30
0 60 120 180 240 300 360
A
m
p
l
i
t
u
d
e
(
d
B
)
ç
0
ç
0
E
z
(num.)
p H
z
(num.)
E
z
(exact)
p H
z
(exact)
Ez (num.)
p H
z
(num.)
E
z
(exact)
p H
z
(exact)
Figure 5. Far scattered ﬁeld from a coated conducting cylinder with radii ka=2, kb=3, e
r
=4: (a)
TM; (b) TE.
METHOD OF MOMENTS 2561
scattering from a class of lossy dielectric spheres subject to
the impedance boundary condition, using the classical
modal solution derived from Mie’s series. Using the same
approach, Wait and Jackson [17] presented many com
puted results for the scattering by an impedance sphere
assuming various surface impedances. The usefulness of
the IBC in determining the scattered ﬁelds has also been
demonstrated by Cassedy and Fainberg [18] for ﬁnite
cylinders.
The electromagnetic scattering from bodies made of
electrically dissimilar materials has also been investigated
using the IBC in conjunction with the method of moments
(MoM). Different geometries have been considered, such
as circular and elliptical cylinders [19,20] and spheroidal
bodies [21]. The introduction of dissimilar materials has
opened the door for more practical uses of the IBC in
antenna problems. Iskander et al. [22] used the IBC to
simulate transverse corrugations in a small conical horn.
A modiﬁcation of the IBC concept to treat the problem of
homogeneous imperfectly conducting objects, with small
radii of curvature, has been described by Mitzner [23].
Different integral equation formulations have been
developed to treat the problem of scattering from impe
dance bodies of revolution. Iskander et al. [22] employed
the electric ﬁeld integral equation formulation (EFIE).
Later, Graglia and Uslenghi [24] examined impedance–
coated bodies of revolution using the generalized MFIE
formulation. In these earlier studies the problem of inter
nal resonances had not been addressed. However, the
EFIE and MFIE formulations were shown to be illposed
and lead to illconditioned matrices when solved by the
MoM [25–28]. The internal resonance problem has been
found to be similar to the one encountered with perfectly
conducting bodies, which was reported in 1978 [29] and
treated by the combined ﬁeld integral equation (CFIE)
formulation. The problem of the internal resonances with
impedance bodies of revolution was reported and treated
by many investigators using CFIE. Apparently, the inter
nal resonance problem is not the only problem associated
with impedance bodies. Rusch and Pogorzelski [30] re
ported another problem even when applying CFIE formu
lation with bodies having both thin and thick parts.
Accordingly, a mixedﬁeld solution that combines different
integral equations on different parts of the scatterer must
be employed. They [30] recommended that the EFIE be
applied on the thin parts of the scatterer in order to
provide a correct solution.
The calculation of scattering from bodies composed of
lossless or low loss dielectric materials and surface im
pedance objects using a mix of exact and impedance
boundary conditions is treated with different types of
integral equations by Kishk [31,32].
Some complicated structures can be modeled approxi
mately using the concept of surface impedance, such as
corrugated objects or objects coated with lossy material or
thin dielectric layers, which can even be loaded with metal
strips. The surface impedance model deals with the outer
boundary of the structure in terms of equivalent surface
impedance, which can be obtained from the expected local
relation between the tangential components of the electric
and magnetic ﬁelds on the outer boundary. This relation
can be found approximately at any surface point from the
solution of a canonical problem, which is similar to the
local geometry around this point. The equivalent surface
impedance is generally anisotropic, even if the coating is
isotropic, in particular at an outer surface that has two
different principal curvatures. Also, structures with peri
odic surface discontinuities such as corrugations or strip
loaded coatings can be modeled using the anisotropic
surface impedance concept, if the periods of the corruga
tions or strips are smaller than half the wavelength. The
advantage with the surface impedance concept is that the
numerical analysis of the object becomes simpler and
takes less time. This is because the exact geometry of
the loads needs not to be modeled to facilitate description
of the problem and so that the number of unknowns can be
reduced dramatically.
The impedance boundary conditions (IBCs) are a valid
approximation under certain conditions [15]; more refer
ences on IBC can be found in Kishk [32]. The use of IBC
can simplify the analysis of many complex electromag
netic problems. However, in other problems the use of IBC
is questionable [33]. In order to widen the applicability of
the IBC, generalized impedance boundary conditions
(GIBCs) was proposed [34] and later improved for coated
2D structures [35] at the expense, however, of consider
able analytical complications, which requires specialized
expertise to work with such problems. So far, GIBC has
been used only with coated metallic surfaces without
corrugations. On the other hand, IBC has been used
successfully to analyze corrugated horns and waveguides
[22]. Anisotropic surface impedances have also been used
to deﬁne soft and hard surfaces [36].
Many papers have formulated the problem of electro
magnetic scattering from 2D impedance structures due to
a normally incident plane wave [37–39]. However, few
have considered oblique incidence [40,41]. In [40], the
ﬁnite element method was used to model arbitrarily
shaped objects with isotropic impedance surfaces. Gordon
and Kishk [41] used the ﬁnitedifference method in the
frequency domain for elliptic isotropic impedance cylin
ders. The problem of scattering from a twodimensional
object of arbitrary cross section and anisotropic surface
impedance due to an oblique plane wave incidence has
been formulated [10]. The formulation is based on the
surface integral equation and solved using the method of
moment with pulse basis functions and point matching.
With the proper implementation of this simple expansion
and testing, accurate numerical solutions are obtained.
The numerical solution is veriﬁed with the exact solution
of a circular cylinder [42]. Different surface integral for
mulations are generated and found useful in the veriﬁca
tion of the numerical solutions for arbitrary objects.
8. FULLWAVE ANALYSIS OF MICROWAVE CIRCUITS
In conjunction with the widespread development of wire
less information networks such as cellular, personal
communication services (PCS), satellite communication
systems, mobile computing, and other new systems and
services, monolithic microwave integrated circuits
2562 METHOD OF MOMENTS
(MMICs) have advanced signiﬁcantly. The primary moti
vation for developing MMIC in wireless communications
such as PCS is to achieve higher quality, longer battery
lifetime, lower cost, and lighter terminals [43]. Dielectric
resonators (DRs) with high permittivity and low tempera
ture coefﬁcients are smaller than waveguide or coaxial
resonators, are easily fabricated, and are compatible with
MMIC implementation [44,45]. Using dielectric resona
tors, instead of metal cavities, in a multilayered medium
coupled to a microstrip line or a slot aperture eliminates
the need for microstrip to cavity adapters and provides
greater ﬂexibility to realize complicated bilateral and
multilayered printed circuits, thus allowing very compact,
highdensity circuit integration.
With the advent of the highdielectricconstant and
lowtemperaturecoefﬁcient ceramic materials, DR appli
cations of the dielectric resonator in the design of passive
and active microwave circuits [46,47] have spread toward
low frequencies for mobile communications such as PCS at
1.8–2 GHz. An experimental study of a dielectric resona
tor, of relative dielectric constant 80 and of resonant
frequency B1.7 GHz, coupled to a microstrip line can be
found in Ref. 48. The method of moments (MoM) has been
used to calculate the characteristics of an isolated di
electric resonator that is of the revolutiontype body
[49,50]. This model includes both dielectric loss and
radiation loss. Numerical analysis of incircuit parameters
of a dielectric resonator is more challenging. A study of the
DR antennas excited by a coaxial probe or slot aperture
has been conducted in both theory and measurement at
the University of Mississippi [51–53], where theoretical
investigations have considered mainly structures consist
ing of bodyofrevolution dielectric resonators combined
with wires or slots.
The MoM fullwave analysis procedure is developed to
analyze the dielectric resonator of arbitrary shape in a
multilayered medium coupled to a microstrip circuit. A
new mixed potential integral equation (MPIE) formulation
has been developed for the analysis of electromagnetic
problems involving conducting or dielectric objects of
arbitrary shape embedded in a planar stratiﬁed medium
[54]. In the new MPIE formulation, the dyadic kernel of
the vector potential is kept in the simplest form originally
developed by Sommerfeld, yet the scalar potential, which
is represented by a double dot product of a dyadic kernel
with a dyadic charge density, remains compatible with the
original implementation. The triangular patch model,
originally developed for arbitrarily shaped objects in free
space [55–59], is employed in conjunction with the new
MPIE formulation to model the problem of a dielectric
resonator excited by microstrip circuit. The numerical
procedure has been modiﬁed to handle the potential dyadic
kernels and the dyadic charge density. A matchedload
simulation procedure has been used to extract the network
S parameters of a DR microstrip circuit. The diameters
(scattering) of the Q circles have been measured to deter
mine the coupling coefﬁcients and the Q factors of the
dielectric resonator excited by a microstrip circuit. The
validity of the new MPIE formulation and the numerical
procedure has been veriﬁed by comparing the S para
meters obtained and the available measurement data [54].
Ceramic materials of high dielectric constant and low
loss can exhibit a very high Q factor. Therefore, they have
been widely used as resonators in microwave ﬁlters and
oscillators. To understand the behavior of these elements
and to determine the best coupling mechanism, it is
necessary to determine the electromagnetic ﬁeld distribu
tion inside the cavity. The resonant frequency and the Q
factor have to be computed accurately before the electro
magnetic ﬁelds in the vicinity of the resonators can be
obtained. It is therefore important to be able to determine
the resonant frequencies and Q factors of the desired
modes.
Many different approaches to the analysis of dielectric
resonators have been described in the literature [60–76].
Some of these methods are based on simpliﬁcations of the
geometry, such as the perfect magnetic conducting (PMC)
walls method [60–62]. Dielectric waveguide methods [63],
as well as their perturbation corrections and the variation
improvements [64] for cylindrical resonators, have also
been developed. In 1975, Van Bladel reported a rigorous
asymptotic method for evaluating the modes of dielectric
resonators of arbitrary shape and high permittivity
[65,66]. In addition, radial and axial modematching
methods [67] for shielded resonators as well as asymptotic
expansion methods [68] have been reported. Also, general
modematching approaches using Green’s dyadic func
tions or transverse modes in expanding the interior and
exterior ﬁelds [69] have been successful approaches for
this kind of problem. Glisson, Kajfez, and James intro
duced the use of the method of moments for the analysis of
an open dielectric resonator [70–72]. The method was
applied to dielectric bodies of revolution of arbitrary cross
section and is useful for any azimuthal variation. Their
results indicate that their method has yielded highly
accurate values of resonant frequencies and Q factors for
homogeneous dielectric resonators. The ﬁnite integration
method was also used for axisymmetric shielded resona
tors [73]. Another approach to the electromagnetic reso
nance of open dielectric resonators is the nullﬁeld method
[74,75] developed by Wenxin Zheng and Staffan Strom. In
that approach, the resonance problem is solved by search
ing for zeros of the determinant of the socalled Q matrix
for the dielectric resonators.
A previously developed numerical method [76] based on
the method of moments is used to search for the complex
resonant frequency, from which the resonant frequency
and radiation Q factor can be computed, of both homo
geneous and inhomogeneous dielectric resonators in free
space. The geometries considered here are rotationally
symmetric, so the bodyofrevolution (BoR) approach is
employed. The equivalence principle is used, and the sur
face integral equations are formulated for the problem.
Then the method of moments is used to reduce the integral
equations to a matrix equation. The ‘‘natural’’ resonant
frequencies are deﬁned as the frequencies at which the
determinant of the moment matrix vanishes. Also, since
the rotationally symmetric structure supports independent
azimuthal modes, the azimuthal variation of the unknown
equivalent currents is expanded in Fourier series. This
allows one to search for the zeros of the moment matrix for
modes having a particular azimuthal variation.
METHOD OF MOMENTS 2563
To compute the Q factor of nonradiating cavities, loss
less structure and search for the resonant frequencies are
considered. Once the resonant frequency is known, the
surface currents are computed, which are then used to
compute ﬁeld distributions inside the cavity. Then, the
perturbation method is employed to compute the stored
energy and dissipated power from which we compute the
Q factor [77].
For the solution of the EM modeling problems involving
arbitrary geometries, the mixed potential surface integral
equation (MPSIE) formulation and the method of mo
ments (MoM) solution procedure are employed [55–58].
The EPMCHW form [78] of the MPSIE formulation is
used. EMscattering and radiation have been the subject of
numerous studies covering the whole spectrum ranging
from LF (lowfrequency) to the optical frequencies.
Although analytical solution methods are available for
some canonical cases, they are limited to simple geome
tries, for example, an object with boundaries of constant
coordinates. Such solutions can be a useful tool in EM
modeling for validation and comparison of more general
solutions, and for the development of approximate asymp
totic techniques. Very often an analytically solvable object
is either completely a conductor or purely a dielectric, but
not a composite. For an arbitrarily shaped conducting or
dielectric object, however, there is no analytic solution,
and one must resort to numerical methods.
For EM problems there numerous numerical methods
have been developed and can be classiﬁed in many ways
[79,80]. The variety of the electromagnetic numerical
methods is due to the extremely wide frequency range
and diverse classes of EM problems that are of interest,
such as multilayer, periodic, and frequencyselective
structures, to name just a few. Needless to say, the variety
of the methods testiﬁes to the fact that no single method
works for all problems, and that each numerical method is
well suited for the analysis of particular types of problems.
The MoM (method of moments) [81], FEM (ﬁniteelement
method) [82], and FDTD (ﬁnitedifference time domain)
[83–85] techniques are the most powerful fullwave meth
ods used in this frequency range. The surface integral
equations are derived from the equivalence principle [86]
and appropriate boundary conditions.
Although the ﬁrst reported dielectric resonator an
tenna was used in a millimeterwave array [87], most of
the research has focused on individual DRA elements of
various common shapes [88–90]. More extensive refer
ences on DRAs are found in Ref. 91. Unlike other classes
of antennas, a DRA has a minimum of metal parts and
consists mainly of dielectric materials mounted over a
metal ground plane. When compared with traditional
metallic antennas, DRA’s are small and low cost, and
they have relatively small losses, especially in the milli
meterwave region, where the conductor losses become
severe. Aperturecoupled DRAs provide compatibility with
MMIC fabrication as well as isolation between the feeding
and radiating structures. Possible advantages of DRAs
over microstrip antennas of similar size are wider
bandwidth and higher powerhandling capabilities. Di
electric radiators also offer the freedom to choose a mode
for broadside radiation or another for endﬁre radiation.
DRAs can be designed for wide bandwidth, desired ﬁeld
patterns, and a linear or circular polarization by choosing
appropriate dielectric constants of the DR elements, ele
ment shapes, and feeding structures. One of the main
disadvantages of DRAs, however, is the difﬁculty of their
analysis. Although the early work on DRAs was based on
experiments and simple models of a DR element [92–94],
more rigorous and accurate numerical methods have been
indispensable for better design of DRAs. One example of
such a demand is a more compact and wideband handset
antenna used in the wireless communication, for which
the DRA is a good candidate.
At the University of Mississippi, the MPSIE/MoM
procedure [56–58] and BoR (body of revolution)/MoM [95]
developed for the analysis of general EM modeling has
been applied to the study dielectric resonator antennas
[96–102]. The MPSIE/MoM formulation has been further
developed by D. R. Wilton et al. to become a standard
procedure for SIE/MoM and VIE (volume integral equa
tions)/MoM, and a package of standard subroutines has
been available [59]. For problems with a ﬁnite number of
homogeneous dielectric regions in combination with con
ducting bodies, SIE/MoM may be advantageous over VIE/
MoM, FEM, or FDTD because it discretizes only the sur
face of the problem space rather than the whole volume.
The wavelength within the dielectric may be much shorter
than that of the free space, and thus girding the whole
computational space uniformly with the resolution set by
the critical geometry within the dielectric region results in
a very inefﬁcient situation. Therefore most numerical
methods reported for DRAs have used SIE/MoM methods
and taken further computational advantage of BoR models
[96–102], by which their applications are limited to the
problems of axially symmetric shapes.
However, the SIE/MoMcan be quite tedious and difﬁcult
to implement when applied to arbitrary conﬁgurations
with complex geometries or inhomogeneous dielectrics.
By arbitrary conﬁgurations, we refer to an arbitrary num
ber of dielectric regions, arbitrary compositions of conduc
tors and dielectrics, general excitations, and other factors
as well as arbitrary shapes. Although numerous MoM
codes have been developed, when the arbitrariness of the
geometries modeled is considered, the number dwindles
down quickly, and the codes capable of handling the more
general geometries often involve years of work and are
expensive to obtain. A 3Dsurface of arbitrary shape is most
suitably modeled by using the triangular patches or RWG
(Rao–Wilton–Glisson) basis functions compared with many
other approaches [102–107]. EM modeling of arbitrarily
shaped conductors [58] or homogeneous dielectric [108] has
been accomplished using the RWG basis functions.
Arbitrarily shaped composite bodies of conductors and
dielectrics have often been treated with the constraint of a
multilayer environment [109–113], where the dielectric
layers are assumed to be of inﬁnite extent. EM modeling of
arbitrarily shaped 3D composite objects using RWG basis
functions has been also reported in the literature
[113–119]. However, the arbitrariness of the conﬁgura
tions treated by them is more or less conﬁned and no
systematic procedure for extension of the formulation has
been provided. For example, for some cases the number of
2564 METHOD OF MOMENTS
dielectric regions is limited to two, including the free
space, and arbitrary shape is allowed only for conductors
in presence of axially symmetric dielectric bodies [120], or
only junctions consisting of up to three intersecting sur
faces are allowed. Moreover the procedures employed to
deal with the composite objects of conducting and di
electric objects are more involved than the standard
MPIE/MoM formulation.
9. TIMEDOMAIN INTEGRAL EQUATIONS
The method of moments is also applied in the time domain
for conducting and dielectric bodies. Several formulations
have been used for the solution of the timedomain inte
gral equation to calculate the electromagnetic scattering
from arbitrarily shaped 3D structures. For the 3D struc
tures the triangular basis function is used [121]. The time
derivative in EFIE formulation of the magnetic vector
potential is implemented by differentiating all the terms
in the EFIE with respect to time; this magnetic vector
potential term is approximated by secondorder ﬁnite
differences [122]. However, the results become oscillatory
for late times, which could be eliminated by approximating
the average value of the current [123]. The incident ﬁeld
should be differentiated, but an impulse or step function
for the incident ﬁeld cannot be used as an excitation. In
addition, to overcome this problem, a backward ﬁnite
difference approximation for the magnetic vector potential
term has been used for the explicit technique [124]. Many
numerical results using the explicit method with forward/
backwarddifference schemes have been shown in
[121,124,125]. An implicit scheme has been proposed to
solve scattering problems [126–130]. The explicit method
requires a very small timestep, and the computed time
domain response becomes unstable. This is due to the
accumulation of numerical error, and it consumes much
computation time. This can be overcome using an implicit
method. In the implicit method the timestep is larger than
that for the explicit case. Therefore numerical error due to
the approximation of a time derivative with the use of
ﬁnite difference is increased. Jung and Sarkar used a
central ﬁnitedifference methodology, which is more accu
rate and provides stable solutions [131]. In this work, the
time domain CFIE formulation is used. The EFIE and
MFIE are converted into matrix equations and are solved
for CFIE solution. The goal is to ﬁnd that the CFIE gives a
unique solution of transient scattering problems when the
incident wave includes resonant frequencies of the scat
terer. The solution technique developed in this work is
capable of handling either an explicit or implicit scheme of
the EFIE, MFIE, and CFIE.
BIBLIOGRAPHY
1. J. J. Faran, Scattering of cylindrical waves by a cylinder, J.
Acoust. Soc. Am. 25:155–156 (1953).
2. S. B. Adler and R. S. Johnson, New backscattering computa
tion and tables for dielectric and metal spheres, Appl. Opt.
1:655–660 (1962).
3. S. Asano and G. Yamamota, Light scattering by spheroidal
particle, Appl. Opt. 14:29–49 (1975).
4. A. A. Kishk, P. Slattman, and P.S. Kildal, Radiation from 3D
sources in the presence of 2D objects of arbitrary cross
sectional shape, Appl. Comput. Electromagn. Soc. (ACES) J.
14(1):17–24 (March 1999).
5. R. F. Harrington, Field Computation by Moment Methods,
Macmillan, New York, 1968.
6. A. A. Kishk and L. Shafai, Different formulations for numer
ical solution of single or multibodies of revolution with mixed
boundary conditions, IEEE Trans. Anten. Propag. AP34:666–
673 (May 1986).
7. P.S. Kildal, S. Rengarajan, and A. Moldsvor, Analysis of
nearly cylindrical antennas and scattering problems using a
spectrum of twodimensional solutions, IEEE Trans. Anten.
Propag. 44:1183–1192 (Aug.1996).
8. K. Forooraghi and P.S. Kildal, Transverse radiation pattern
of a slotted waveguide array radiating between ﬁnite height
baafﬂes in terms of a spectrum of twodimensional solutions,
IEE Proc. H 140:52–58 (1993).
9. A. A. Kishk and P. M. Goggans, Electromagnetic scattering
from twodimensional composite objects, ACES J. 9:32–39
(1994).
10. A. A. Kishk and P.S. Kildal, Electromagnetic scattering
from two dimensional anisotropic impedance objects under
oblique plane wave incidence, ACES J. 10(3):81–92 (1995).
11. M. A. Leontovich, Investigation of Propagation of Radio
Waves, Part II, Moscow, 1948.
12. T. B. A. Senior, Impedance boundary conditions for imper
fectly conducting surfaces, Appl. Sci. Res. B 8:418–436
(1961).
13. T. B. A. Senior, Impedance boundary conditions for statisti
cally rough surfaces, Appl. Sci. Res. B 8:437–462 (1961).
14. T. B. A. Senior, A note on impedance boundary conditions,
Can. J. Phys. 40:663–665 (1962).
15. T. B. A. Senior, Approximate boundary conditions, IEEE
Trans. Anten. Propag. AP29:826–829 (1981).
16. R. J. Garbacz, Bistatic scattering from a class of lossy
dielectric spheres with surface impedance boundary condi
tions, Phys. Rev. 33:A14–A16 (1964).
17. J. R. Wait and C. M. Jackson, Calculation of the bistatic
scattering cross section of a sphere with an impedance
boundary condition, Radio Sci. J. Res. 69D:299–315 (1965).
18. E. S. Cassedy and J. Fainberg, Back scattering cross sections
of cylindrical wires of ﬁnite conductivity, IRE Trans. Anten.
Propag. AP8:1–7 (1960).
19. N. G. Alexopoulos and G. A. Tadler, Accuracy of the Leonto
vich boundary condition for continuous and discontinuous
surface impedance, J. Appl. Phys. 46:3326–3332 (1975).
20. N. G. Alexopoulos and G. A. Tadler, Electromagnetic scatter
ing from an elliptic cylinder loaded by continuous and
discontinuous surface impedance, J. Appl. Phys. 46:1128–
1134 (1975).
21. N. G. Alexopoulos, G. A. Tadler, and P. L. E. Uslenghi,
Scattering from spheroidal composite objects, J. Franklin
Inst. 309:147–162 (1980).
22. K. A. Iskander, L. Shafai, A. Fransen, and J. E. Hansen,
Application of impedance boundary conditions to numerical
solution of corrugated circular horns, IEEE Trans. Anten.
Propag. AP30:366–372 (1982).
23. K. M. Mitzner, An Integral equation approach to scattering
from a body of ﬁnite conductivity, Radio Sci. 2:1459–1470
(1967).
METHOD OF MOMENTS 2565
24. R. D. Graglia and P. L. E. Uslenghi, Electromagnetic scat
tering by impedance bodies of revolution, Proc. Natl. Radio
Science Meeting, URSI/B63, Houston, 1983, p. 115.
25. A. A. Sebak and L. Shafai, Eﬁeld, Hﬁeld and combined ﬁeld
solutions for bodies of revolution with impedance boundary
conditions, Proc. Natl. Radio Science Meeting, URSI/B165,
Boston, 1984, p. 211.
26. A. A. Sebak and L. Shafai, Performance of various integral
equation formulations for numerical solution of scattering
by impedance objects, Can. J. Phys. 62:605–615 (1984).
27. J. R. Rogers, Numerical solutions of illposed impedance
boundary condition integral equations, 1984 Int. Symp.
Digest, APS111, 1984, pp. 347–350.
28. L. N. MedgyesiMitschang and J. M. Putnam, Integral
equation formulation for imperfectly conducting scatterers,
IEEE Trans. Anten. Propag. AP33:206–214 (1985).
29. J. R. Mautz and R. F. Harrington, Hﬁeld, Eﬁeld, and
combinedﬁeld solutions for conducting bodies of revolution,
Arch. Elektron. Ubertragungstech. 32:159–164 (1978).
30. W. V. T. Rusch and R. J. Pogorzelski, A mixedﬁeld solution
for scattering from composite bodies, IEEE Trans. Anten.
Propag. AP34:955–958 (1986).
31. A. A. Kishk, Formulation of impedance surfaces coated with
dielectric materials, Comput. Phys. Commun. 145–156
(1991).
32. A. Kishk, Electromagnetic scattering from composite objects
using a mixture of exact and impedance boundary condi
tions, IEEE Trans. Anten. Propag. AP39(6):826–833 (June
1991).
33. J. R. Wait, Use and misuse of impedance boundary condi
tions in electromagnetics, Proc. PIERS Symp. Boston, July
1989, p. 358.
34. T. B. A. Senior and J. L. Volakis, Derivation and application
of a class of generalized boundary conditions, IEEE Trans.
Anten. Propag. AP37:1566–1572 (1989).
35. D. J. Hoppe and Y. RahmatSami, Scattering by superquad
ric dielectriccoated cylinders using higher order impedance
boundary conditions, IEEE Trans. Anten. Propag. 40:1513–
1523 (1992).
36. P.S. Kildal, Artiﬁcially soft and hard surfaces in electro
magnetics, IEEETrans. Anten. Propag. 38:1537–1544 (1990).
37. P.S. Kildal, A. A. Kishk, and A. Tengs, Reduction of forward
scattering from cylindrical objects using hard surfaces,
IEEE Trans. Anten. Propag. 44:1509–1520 (Nov. 1996).
38. P. M. Goggans, A Combined MethodofMoments and Ap
proximate Boundary Condition Solution for Scattering from
a Conducting Body with a DielectricFilled Cavity, Ph.D.
dissertation, Auburn Univ., 1990.
39. A. A. Kishk and P. M. Goggans, Electromagnetic scattering
from twodimensional composite objects, ACES J. 9(1):32–39
(1994).
40. R. Gordon and A. A. Kishk, An efﬁcient ﬁnite element
method for determining the scattering from lossy cylinders
using the impedance boundary condition and an absorbing
boundary condition due to oblique incident, Proc. 9th An
nual Review of Progress in Applied Computational Electro
magnetics Conf. 1993, pp. 871–878.
41. J. Yan, R. K. Richard and A. A. Kishk, Electromagnetic
scattering from impedance elliptic cylinders using ﬁnite
difference method (oblique incidence), J. Electromagn.
15:157–173 (1995).
42. A. W. Glisson, M. Orman, and D. Koppel, Electromagnetic
scattering by a body of revolution with a general anisotropic
impedance boundary condition, IEEE Anten. Propagation
Soc. Int. Symp. Digest, 4:1997–2000 (June 1992).
43. P. Kaveh and A. H. Levesque, Wireless Information Net
works, Wiley, New York, 1995.
44. D. Kajfez and P. Guillon, Dielectric Resonators, Vector
Fields, Oxford, MS, 1990.
45. L. E. Larson, RF and Microwave Circuit Design for Wireless
Communications, Artech House, Norwood, MA, 1996.
46. A. Podcameni and L. F. M. Conrado, Design of microwave
oscillators and ﬁlters using transmissionmode dielectric
resonators coupled to microstrip lines, IEEE Trans. Micro
wave Theory Tech. MTT33:1329–1332 (Dec. 1985).
47. A. M. Pavio and M. A. Smith, A 2040GHz pushpush
dielectric resonator oscillator, IEEE Trans. Microwave The
ory Tech. MTT33:1346–1349 (Dec. 1985).
48. D. Kajfez and J. Guo, Precision measurement of coupling
between microstrip and TE
01
mode dielectric resonator,
Electron. Lett. 30:1772–1773 (Oct. 1994) (also, the data
that were not published).
49. A. W. Glisson, D. Kajfez, and J. James, Evaluation of modes
in dielectric resonators using a surface integral equation
formulation, IEEE Trans. Microwave Theory Tech. MTT
31:1023–1029 (Dec. 1983).
50. D. Kajfez, A. W. Glisson, and J. James, Computed model ﬁeld
distributions for isolated dielectric resonators, IEEE Trans.
Microwave Theory Tech. MTT32:1609–1616 (Dec. 1984).
51. G. P. Junker, Analysis of Dielectric Resonator Antennas
Excited by a Coaxial Probe or Narrow Slot Aperture, Ph.D.
dissertation, Univ. Mississippi, Aug. 1994.
52. G. P. Junker, A. A. Kishk, and A. W. Glisson, Input impedance
of dielectric resonator antennas excited by a coaxial probe,
IEEE Trans. Anten. Propag. AP42:960–966 (July 1994).
53. G. P. Junker, A. A. Kishk, D. Kajfez, A. W. Glisson, and J.
Guo, Input impedance of microstripslotcoupled dielectric
resonator antennas mounted on thin dielectric layers, Int. J.
Microwave Millimeterwave Comput. Aided Eng. 6:174–182
(May 1996).
54. J. Y. Chen, A. A. Kishk, and A. W. Glisson, Application of
new MPIE formulation to the analysis of a dielectric reso
nator embedded in a multilayered medium coupled to a
microstrip circuit, IEEE Trans. Microwave Theory Tech.
49:263–279 (Feb. 2001).
55. A. W. Glisson, On the Development of Numerical Techniques
for Treating ArbitrarilyShaped Surfaces, Ph.D. thesis,
Univ. Mississippi, 1978.
56. S. M. Rao, Electromagnetic Scattering and Radiation of
ArbitrarilyShaped Surfaces by Triangular Patch Modeling,
Ph.D. thesis, Univ. Mississippi, 1980.
57. S. M. Rao, D. R. Wilton, and A. Glisson, Electromagnetic
scattering by surfaces of arbitrary shape, IEEE Trans.
Anten. Propag. AP30:409–418 (May 1982).
58. S. V. Yesantharao, EMPACK: A Software Toolbox of Potential
Integrals for Computational Electromagnetics, master’s the
sis, Univ. Houston, 1989.
59. S. M. Rao, C. C. Cha, R. L. Cravey, and D. L. Wilkes,
Electromagnetic scattering from arbitrary shaped conduct
ing bodies coated with lossy materials of arbitrary thickness,
IEEE Trans. Anten. Propag. AP39:627–631 (May 1991).
60. H. Y. Yee, Natural resonant frequencies of microwave di
electric resonators, IEEE Trans. Microwave Theory Tech.
MTT13:256 (March 1965).
61. Y. Garault and P. Guillon, Best approximation for
design of natural resonance frequencies of microwave
2566 METHOD OF MOMENTS
dielectric disc resonants, Electron. Lett. 10:505–507 (Nov.
1974).
62. Y. Konishi, N. Hosino, and Y. Utsumi, Resonant frequency of
a TE
01
dielectric resonator, IEEE Trans. Microwave Theory
Tech. MTT24:112–114 (Feb. 1976).
63. T. Itoh and R. Rudokas, New method for computing the
resonant frequencies of dielectric resonator, IEEE Trans.
Microwave Theory Tech. MTT25:52–54 (Jan. 1977).
64. D. Kajfez and P. Guillon, eds., Dielectric Resonators Noble
Publishing, Atlanta, 1998.
65. J. Van Bladel, On the resonances of a dielectric resonator of
very high permittivity, IEEE Trans. Microwave Theory Tech.
MTT23:199–208 (Feb. 1975).
66. J. Van Bladel, The excitation of dielectric resonators of very
high permittivity, IEEE Trans. Microwave Theory Tech.
MTT23:208–217 (Feb. 1975).
67. Y. Kobayashi, N. Fukuoka, and S. I. Yoshida, Resonant
models for a shielded dielectric resonator, Electron. Com
mun. Jpn. 64B(11):44–51 (1981).
68. K. A. Zaki and C. Chen, Coupling of nonaxially symmetric
hybrid modes in dielectric resonators, IEEE Trans. Micro
wave Theory Tech. MTT35:1136–1142 (Dec. 1987).
69. R. E. Collin and D. A. Ksienski, Boundary element method
for dielectric resonators and waveguides, Radio Sci.
22:1155–1167 (Dec. 1987).
70. A. W. Glisson, D. Kajfez, and J. James, Evaluation of modes
in dielectric resonators using a surface integral equation
formulation, IEEE Trans. Microwave Theory Tech. MTT
31:1023–1029 (Dec. 1983).
71. D. Kajfez, A. W.Glisson, and J. James, Computed model ﬁeld
distributions for isolated dielectric resonators, IEEE Trans.
Microwave Theory Tech. MTT32:1609–1616 (Dec. 1984).
72. A. A. Kishk, A. W. Glisson, and D. Kajfez, Computed
resonant frequency and far ﬁelds of isolated dielectric discs,
Proc. 1993 IEEE APS Int. Symp. Ann Arbor, MI, June 1993,
pp. 408–411.
73. J. E. Lebaric and D. Kajfez, Analysis of dielectric resonator
cavities using the ﬁnite integration technique, IEEE Trans.
Microwave Theory Tech. 37:1740–1748 (Nov. 1989).
74. W. Zheng and S. Strom, The nullﬁeld approach to electro
magnetic resonance of composite objects, Comput. Phys.
Commun. 157–174 (March 1991).
75. W. Zheng, Computation of complex resonance frequencies of
isolated composite objects, IEEE Trans. Microwave Theory
Tech. MTT37:953–961 (June 1989).
76. A. A. Kishk, Electromagnetic scattering from composite
objects using a mixture of exact and impedance boundary
conditions, IEEE Trans. Anten. Propag. AP39(6):826–833
(June 1991).
77. A. A. Kishk, D. Kajfez, and S. Chebolu, Resonant frequency
and Qfactor of axisymmetric composite microwave cavities,
IEEE Trans. Microwave Theory Tech. 50(10):2287–2293
(Oct. 2002).
78. A. A. Kishk and L. Shafai, Different formulations for numer
ical solution of single or multibodies of revolution with
mixed boundary conditions, IEEE Trans. Anten. Propag.
AP34:666–673 (May 1986).
79. E. K. Miller and G. J. Burke, Lowfrequency computational
electromagnetics for antenna analysis, IEEE Proc. 80:24–43
(Jan. 1992).
80. T. H. Hubing, Survey of Numerical Electromagnetic Model
ing Techniques, Tech. Report TR91001.3, Electromagnetic
Compatibility Laboratory, Univ. Missouri—Rolla, 1991.
81. R. F. Harrington, Field Computation by Method of Moments,
Krieger, Malabar, FL, 1968.
82. R. Coccioli, T. Itoh, G. Pelosi, and P. P. Silvester, Finite
element methods in microwaves: A selective bibliography,
IEEE Anten. Propag. Mag. 38:34–48 (Dec. 1996).
83. K. S. Kunz and R. J. Luebbers, The Finite Difference Time
Domain Method for Electromagnetics, CRC Press, 1993.
84. A. Taove, Computational Electromagnetics: The FiniteDif
ference TimeDomain Method, Artech House, 1995.
85. K. L. Shlager and J. B. Schneider, A selective survey of the
ﬁnitedifference timedomain literature, IEEE Anten. Pro
pag. Mag. 37:39–57 New York, (Aug. 1995).
86. R. F. Harrington, Time Harmonic Electromagnetic Fields,
McGrawHill, 1963.
87. M. T. Birand and R. V. Gelsthrope, Experimental millimetric
array using dielectric radiators fed by means of dielectric
waveguide, Electron. Lett. 17:633–635 (Sept. 1981).
88. M. W. McAllister and S. A. Long, Rectangular dielectric
resonator antenna, Electron. Lett. 19:218–219 (March 1983).
89. S. A. Long, M. W. McAllister, and L. C. Shen, The resonant
cylindrical dielectric cavity antenna, IEEE Trans. Anten.
Propag. AP31:406–412 (May 1983).
90. M. W. McAllister and S. A. Long, Resonant hemispheri
cal dielectric antenna, Electron. Lett. 20:657–658 (March
1984).
91. A. Petosa, A. Ittipiboon, Y. M. M. Antar, and D. Roscoe,
Recent advances in dielectricresonator antenna technology,
IEEE Anten. Propag. Mag. 40:35–48 (June 1998).
92. S. B. Cohn, Microwave bandpass ﬁlters containing highQ
dielectric resonators, IEEE Trans. Microwave Theory Tech.
MTT16:218–217 (April 1968).
93. T. Itoh and R. S. Rudokas, New method for computing the
resonant frequencies of dielectric resonators, IEEE Trans.
Microwave Theory Tech. MTT25:52–54 (Jan. 1977).
94. D. Kajfez and P. Guillon, eds., Dielectric Resonators, Artech
House, 1986.
95. A. W. Glisson and C. M. Butler, Analysis of a wire antenna in
the presence of a body of revolution, IEEE Trans. Anten.
Propag. AP28:604–609 (Sept. 1980).
96. (a) A. A. Kishk, H. A. Auda, and B. C. Ahn, Radiation
characteristics of cylindrical dielectric resonator antennas,
IEEE Anten. Propag. Soc. Newsl. 31:7–16 (1989); (b) A. A.
Kishk, M. R. Zunoubi, and D. Kajfez, A numerical study of a
dielectric disk antenna above a grounded dielectric sub
strate, IEEE Trans. Anten. Propag. AP41:813–821 (June
1993).
97. G. P. Junker, A. A. Kishk, A. W. Glisson, and D. Kajfez, Effect
of an air gap around the coaxial probe exciting a cylindrical
dielectric resonator antenna, Electronics Letters, vol. 30, pp.
177–178 (Feb. 1994).
98. A. A. Kishk, G. Zhou, and A. W. Glisson, Analysis of di
electricresonator antennas with emphasis on hemispherical
structures, IEEE APS Newsl. 36:20–31 (April 1994).
99. G. Junker, A. Kishk, and A. Glisson, Input impedance of
dielectric resonator antennas excited by a coaxial probe,
IEEE Trans. Anten. Propag. AP42:960–966 (July 1994).
100. G. P. Junker, A. A. Kishk, and A. W. Glisson, Input impe
dance of an aperture coupled dielectric resonator antenna,
IEEE Trans. Anten. Propag. AP44:600–607 (May 1996).
101. G. P. Junker, A. A. Kishk, and A. W. Glisson, Multiport
network description and radiation characteristics of coupled
dielectric resonator antennas, IEEE Trans. Anten. Propag.
AP46:425–433 (March 1998).
METHOD OF MOMENTS 2567
102. L. N. MedgyesiMitschang and J. M. Putnam, Electromag
netic scattering from axially inhomogeneous bodies of revo
lution, IEEE Trans. Anten. Propag. AP32:797–806 (Aug.
1984).
103. N. N. Wang, J. H. Richmond, and M. C. Gilreath, Sinusoidal
reaction formulation for radiation and scattering from con
ducting surfaces, IEEE Trans. Anten. Propag. 23:376–382
(May 1975).
104. G. J. Burke and A. J. Poggio, Numerical Electromagnetic
Code (NEC): Method of Moments, Naval Ocean Systems
Center, San Diego, CA, Tech. Document 116, July 1977.
105. B. J. Rubin, General solution for propagation, radiation, and
scattering in arbitrary 3D inhomogeneous structures, IEEE
Anten. Propag. Mag. 34:17–25 (Feb. 1992).
106. D. I. Kaklamani and N. K. Uzunoglu, Analysis of dielectri
cally loaded radiators using entiredomain Galerkin techni
que, IEEE Anten. Propag. Mag. 39:30–54 (Oct. 1997).
107. B. M. Notaros and B. D. Popovic, General entiredomain
method for analysis of dielectric scatters, IEE Proc. Micro
wave Anten. Propag. 143:498–504 (Dec. 1996).
108. K. Umashanka, A. Taove, and S. M. Rao, Electromagnetic
scattering by arbitrary shaped threedimensional homoge
neous lossy dielectric objects, IEEE Trans. Anten. Propag.
AP34:758–766 (June 1986).
109. K. A. Michalski and D. Zheng, Electromagnetic scattering
and radiation by surfaces of arbitrary shape in layered
media, Part I: Theory, IEEE Trans. Anten. Propag. AP
38:335–344 (March 1990).
110. K. A. Michalski and D. Zheng, Electromagnetic scattering
and radiation by surfaces of arbitrary shape in layered
media, Part II: Implementation and results for contiguous
halfspaces, IEEE Trans. Anten. Propag. AP38:345–352
(March 1990).
111. J. Y. Chen, Full Wave Analysis of a Dielectric Resonator
Embedded in a Uniaxial Multilayered Medium Coupled to a
Microstrip Circuit, Ph.D. thesis, Univ. Mississippi, 1996.
112. J. Chen, A. A. Kishk, and A. W. Glisson, Application of a new
MPIE formulation to the analysis of a dielectric resonator
embedded in a multilayered medium coupled to a microstrip
circuit, IEEE Trans. Microwave Theory Tech. MTT49:263–
279 (Feb. 2001).
113. S. M. Rao, C. C. Cha, R. L. Cravey, and D. L. Wilkes,
Electromagnetic scattering from arbitrary shaped conduct
ing bodies coated with lossy materials of arbitrary thickness,
IEEE Trans. Anten. Propag. AP39:627–631 (May 1991).
114. T. K. Sarkar, S. M. Rao, and A. R. Djordevic, Electromagnetic
scattering and radiation from ﬁnite microstrip structures,
IEEE Trans. Microwave Theory Tech. MTT38:1568–1575
(Nov. 1990).
115. S. M. Rao, T. K. Sarkar, P. Midya, and A. R. Djordevic,
Electromagnetic radiation and scattering from ﬁnite
conducting and dielectric structures: Surface/surface formu
lation, IEEE Trans. Anten. Propag. AP39:1034–1037 (July
1991).
116. J. M. Putnam and L. N. MedgyesiMitschang, Generalized
method of moments for threedimensional penetrable scat
terers, J. Opt. Soc. Am. A 11:1384–1398 (April 1994).
117. J. M. Putnam and M. B. Gedera, CARLOS3DTM: A general
purpose threedimensional methodofmoments scattering
code, IEEE Anten. Propag. Mag. 35:69–71 (April 1993).
118. T. E. Durham and C. G. Christodoulou, Integral equation
analysis of dielectric and conducting bodies of revolution in
the presence of arbitrary surfaces, IEEE Trans. Anten.
Propag. AP43:674–680 (July 1995).
119. A. J. Movahedi, Electromagnetic Radiation and Scattering
by a Dielectric Loaded Conducting Body Coupled to an
Arbitrary Wire, Ph.D. thesis, Univ. Mississippi, 1995.
120. S. U. Hwu and D. R. Wilton, Electromagnetic Scattering and
Radiation by Arbitrary Conﬁgurations of Conducting Bodies
and Wires, Tech. Report TR8717, Applied Electromagnetics
Laboratory, Univ. Houston, 1989.
121. S. M. Rao, Time Domain Electromagnetics, Academic Press,
Orlando, FL, 1999.
122. S. M. Rao and D. R. Wilton, Transient scattering by con
ducting surfaces of arbitrary shape, IEEE Trans. Anten.
Propag. AP39:56–61 (1991).
123. D. A. Vechinski and S. M. Rao, A stable procedure to
calculate the transient scattering by conducting surfaces of
arbitrary shape, IEEE Trans. Anten. Propag. AP40:661–
665 (1992).
124. S. M. Rao and T. K. Sarkar, An alternative version of the
timedomain electric ﬁeld integral equation for arbitrarily
shaped conductors, IEEE Trans. Anten. Propag. AP41:831–
834 (1993).
125. P. J. Davies, On the stability of timemarching schemes for
the general surface electricﬁeld integral equation, IEEE
Trans. Anten. Propag. AP44:1467–1473 (1996).
126. S. M. Rao and T. K. Sarkar, Timedomain modeling of two
dimensional conducting cylinders utilizing an implicit
schemeTM incidence, Microwave Opt. Technol. Lett.
15:342–347 (1997).
127. S. M. Rao and T. K. Sarkar, An efﬁcient method to evaluate
the time domain scattering from arbitrarily shaped conduct
ing bodies, Microwave Opt. Technol. Lett. 17:321–325 (1998).
128. S. M. Rao and T. K. Sarkar, Transient analysis of electro
magnetic scattering from wire structures utilizing an im
plicit timedomain integralequation technique, Microwave
Opt. Technol. Lett. 17:66–69 (1998).
129. S. M. Rao, D. A. Vechinski, and T. K. Sarkar, Transient
scattering by conducting cylinders implicit solution for the
transverse electric case, Microwave Opt. Technol. Lett.
21:129–134 (1999).
130. T. K. Sarkar, W. Lee, and S. M. Rao, Analysis of transient
scattering from composite arbitrarily shaped complex struc
tures, IEEE Trans. Anten. Propag. AP48:1625–1634 (2000).
131. B. H. Jung and T. K. Sarkar, Timedomain CFIE for the
analysis of transient scattering from arbitrarily shaped 3D
conducting objects, Microwave Opt. Technol. Lett. 34(2):289–
296 (2002).
MICROSTRIP ANTENNA ARRAYS
PETER S. HALL
The University of Birmingham
Edgbaston, Birmingham, UK
1. INTRODUCTION
Microstrip antenna arrays are formed from collections of
microstrip antenna elements, which may then be inter
connected and fed using microstrip transmission lines. For
information on microstrip antenna elements, the reader is
referred to other articles. In this article it is assumed that
2568 MICROSTRIP ANTENNA ARRAYS
the elements may be a microstrip patch or a section of
microstrip transmission line that radiates. This article is
concerned with the architectures of microstrip arrays and
some specific issues that arise in their design, such as
surface waves and mutual coupling.
In using microstrip for both the radiating elements and
the feed network, a lowproﬁle planar structure results.
The minimum thickness will be obtained when the feed is
coplanar with the radiators (Fig. 1), which is to say that
the array and the feed are manufactured by the process of
photolithography on one side of a copperclad microwave
substrate. Improved performance and more functionality
will be obtained if multilayer structures are used. For ex
ample, dual polarization, with separate output connectors
for each, can be achieved using two feed networks in a
multilayer conﬁguration. It is the combination of a low
proﬁle and simple manufacturing process that is the main
attraction of microstrip antenna arrays. This results in
low cost and ease of installation, which means that they
have found application in many areas, including satellites,
aircraft and groundbased systems for communications,
navigation, and radar.
There is a large body of literature on microstrip arrays.
Much of this is gathered together in several books that
have been published [1–5].
2. DESCRIPTION OF ARRAY ARCHITECTURES
Good overviews of array architectures can be found in two
other texts [2, Chaps. 1, 13, 14; 3, Chap. 6]. The primary
classiﬁcations of microstrip arrays are parallel or series
feeding. These are now discussed and other variants are
described.
2.1. Parallel or CorporateFed Arrays
A parallel, or corporate, feed network consists of a micro
strip transmission line that splits successively from the
input to the radiating elements so that all elements are
connected [6,7]. Figure 1 illustrates a microstrip patch ar
ray with four elements and a parallel feed. The feed input
is on the righthand side of the substrate and splits into
two near the center. These two lines then split into two,
and the four lines are then attached to the patches. An
impedance match is maintained at each splitter by the use
of quarterwavelength matching sections, although short
tapered sections have also been used. In this case, the
splitters produce equal amplitudes at each of their out
puts. This then results in an equalamplitude excitation of
each patch. If unequal splitters are used, then some con
trol of the amplitude distribution can be obtained to allow
the sidelobes to be controlled. Similarly, in the array of
Fig. 1, the lengths from the input to each of the patches
are equal, resulting in an equiphase excitation and a beam
that points broadside, or normal, to the array plane. If the
lengths of line between each splitter and the next or the
patches are unequal, then control of the array excitation
phase will be obtained, either allowing the beam to be
shaped or causing it to point away from broadside.
The parallelfed array produces a single pencilshaped
beam, with beamwidths inversely proportional to the
array aperture dimensions. The radiation pattern is rela
tively insensitive to changes in frequency, in that, provid
ed the feed is matched throughout, the array excitation
does not change. In addition, for a broadside beam, the
angle at which the beam points does not vary with fre
quency. The input impedance and gain bandwidth are
controlled primarily by the bandwidth of the radiating
elements.
2.2. SeriesFed Arrays
In a seriesfed array, the radiating elements are arranged
in a line and connected to a transmission line that runs
close by them. Figure 2 shows an example of a fourpatch
seriesfed array. The input connector is attached to the
righthand end of the line. The amplitude distribution
along the array is controlled by the splitter ratios. In the
case shown in which the splitters all have the same ge
ometry, the amplitudes will have an exponentially reduc
ing distribution. While each patch is fed with the same
proportion of the feedline power, that power reduces at
each splitter.
If a matched load is placed on the lefthand end of the
feedline, a travelingwave array is formed. In this case, as
the patches are equispaced, the phase change between
Microstrip patches
Feed network
Microwave
substrate
Ground
plane
Figure 1. Microstrip patch array (typical dimensions for opera
tion at 12GHz, patch size =9.7 7.2 mm, patch spacing=17mm,
substrate height =1.59mm, e
r
=2.32, feedline width=0.5 mm at
patches, 2.5mm at input) [57].
Figure 2. Seriesfed patch array (typical dimensions same as
those for Fig. 1).
MICROSTRIP ANTENNA ARRAYS 2569
elements down the array is constant. The relative element
phase will then determine the beam pointing angle in the
plane of the array. If the frequency is changed, then the
relative phase will change and the beam angle will
change. This type of array is thus called a frequency
scan array.
If the feedline is not terminated in a matched load, then
a resonant array is formed. The radiating elements should
be placed a feedline wavelength apart. The array has a
good input match albeit across a narrow beamwidth, and
the beam points to broadside.
The beamshape of a single seriesfed microstrip array
is broad in the dimension across the array and narrow
along the plane of the array. A pencilshaped beam can be
formed by placing many series arrays side by side. In this
case, a onedimensional corporate feed can be made co
planar with the array; or alternately, a series feed can be
used to feed a set of series feeds. The crossfed array [8] is
a twodimensional series array using equalwidth feed
lines that is center fed and automatically produces a ta
pered distribution with low sidelobes. Dual polarization
can be obtained from a twodimensional patch array [9] in
which square patches are seriesconnected on all four
sides, so that the horizontal connections produce horizon
tal polarization and the vertical connections, vertical po
larization. The chain array structures give linear
polarization and have rectangular [10] and triangular or
honeycomb [11] shapes. The patches may be connected di
rectly to the feedline to form a comblike array [12], cou
pled by a small airgap in the parasitically coupled patch
array [13], or seriesconnected by a feedline connecting the
centers of their sides [14].
As the feedline itself will radiate, the patches shown in
Fig. 2 can be dispensed with and a meander array formed.
There are many forms. Figure 3a shows a silhouette of the
rampart line array [15] as an example. The function of
radiation can be understood from the following descrip
tion, which is based on an equivalent source model of mi
crostrip radiation. Each of the four corners in each period
radiates. The radiation has a polarization that is oriented
451 to the incoming line, as shown in Fig. 3b. If these four
vectors, whose relative phase is determined by the inter
vening linelengths, are summed, then the total amount of
radiation and its polarization can be determined. An array
with uniform linewidth will result in an exponentially re
ducing amplitude distribution and a beam that is either
broadside, if there is no matched load present on the line
end; or a frequencyscanned beam, if there is a load. By
varying the distance between the bends, longitudinal,
transverse, or circular polarization can be obtained.
Many other forms of series array have been designed,
including the serpent line [16,17], the rectangular loop
line, [10], the Franklin line [18], and the circularly polar
ized chain antenna [19] (see Fig. 4). The dimensions can be
determined from the Figure caption or the following. For
the serpent line
sin y =
l
m
s
l
0
l
m
÷
l
0
2s
(1)
where y is the beam angle, l
m
and l
0
are the line and
freespace wavelengths, respectively; and l
m
and s are as
indicated in Fig. 4a. For the chain antenna
sin y =
2s ÷l
p
÷l
w
÷l
0
d
(2)
where s, l
p
, l
w
and d are as indicated in Fig. 4d.
(a)
(b)
(c)
(d)
A B
s
d
A
s
B
r r
w
p
d
Figure 4. Microstrip series array: (a) serpent line [16,17], (l =
l
m
); (b) rectangular loop line [10] (d=l
m
, r =0.4l
m
); (c) Franklin
line [18] (length along line A–B=2l
m
); (d) circularly polarized
chain antenna [19] (2l
p
÷l
w
=0.25l
m
).
(a)
(b)
d
d
Figure 3. Rampart line seriesfed microstrip array: (a) rampart
line array (typical dimensions for operation at 15GHz, linewidth
=1 mm, section length=0.5l
m
, substrate height =0.793mm, e
r
=2.32 [15]); (b) rightangle bend showing polarization of radia
tion (miter depth=0.5 line width); l
m
=microstrip line wave
length.
2570 MICROSTRIP ANTENNA ARRAYS
2.3. LeakyWave Arrays
A halfwavelengthwide microstrip line can also be made
to radiate by feeding with an asymmetric step [20] to
excite the ﬁrstorder asymmetric mode, which then leaks
to radiation down the array length. Synthesis of the
radiation pattern is performed by modulating the line
parameters, such as width or height, or by inserting
radiating slots periodically down the length. A double
layer leakywave microstrip antenna has also been
described [21].
2.4. Ultrawideband Arrays
Linear microstrip patch arrays can be made to operate
over multioctave bandwidths using the log periodic
principle [22] (Fig. 5 shows the architecture). A micro
strip line is electromagnetically coupled to an array
of patches whose dimensions and hence resonant frequen
cy are scaled patch to patch by a constant factor t,
where
t =
l
m÷1
l
m
=
w
m÷1
w
m
=
d
m÷1
d
m
and l, w and d are the patch length, width, and spacing,
respectively. The overall bandwidth is determined by the
number of patches and the scale factor. For example, an
array designed with t =1.05 and 36 elements, achieved a
5 : 1 bandwidth and had a gain of around 10 dB and an
input return loss of –10 dB over a 4–20GHz range. Similar
to the design of the log periodic dipole array, the element
feeding arrangement is crucial. In the microstrip array, it
is necessary to have electromagnetic coupling to prevent
the formation of stopbands in the wanted frequency range
due to the higher patch resonances. Reference 23 shows a
log periodic array that consists of a feedline with directly
connected patches. A useful, but limited, bandwidth of
about 2 : 1 is achieved.
The array of Fig. 5 has a broadside beam. An endﬁre
beam has also been demonstrated [24].
2.5. Multilayer Arrays
Multilayer arrays are designed to
*
Reduce the perturbation to the pattern of radiation
from the feed network in a coplanar form
*
Allow multiple functions, such as dual polarization
*
Prevent radiation from, and to protect active circuits
in, active integrated antennas
Figure 6 shows an exploded diagram of a multilayer
microstrip array. It can be seen that the front face contains
only microstrip patch elements. The feed network is con
structed in triplate stripline and located below the array
substrate. The three layers are then glued together to
form the integral array and feed. Zurcher [25] gives design
techniques for highefﬁciency multilayer arrays, by using
foambased lowpermittivity dielectrics. An example in
which multiple patch layers has been used to increase
bandwidth is given by Legay and Shafai [26].
The connections from the ends of the feedline to the
elements can be made either using vias or by aperture
coupling. The via is a metallic rod or pin that is attached to
the end of the feedline, at its lower point, and to a point on
the patch that has the appropriate impedance, at its upper
end. A clearance hole must be etched in the patch ground
plane. The via can also be made using a plated through
hole. The aperture coupling is formed by having a (usual
ly) rectangular aperture in the ground plane beneath the
patch and above the feedline. Aperture coupling is now
widely used as it is more reliable than a via connection.
Examples of aperturecoupled arrays are given in Refs. 27
and 28.
2.6. Sequentially Fed Arrays
Multiple feeding of microstrip patch antennas has been
shown to improve both the symmetry and the polarization
Figure 5. Log periodic microstrip patch array (typical dimen
sions for a 36element array for operation at 4–18GHz, smallest
patch size 3.672.92mm, smallest spacing=3.67, t =1.05, sub
strate height, feed=0.793mm, patch=1.586mm, e
r
=2.32) [22].
Figure 6. Multilayer microstrip array (patch and feed details
same as those in Fig. 1).
MICROSTRIP ANTENNA ARRAYS 2571
purity of the radiation pattern and the input impedance
bandwidth [29,30], in addition to allowing the generation
of circular and multiple polarizations. Figure 7 illustrates
this principle. The circular patch shown on the left is fed
at two points. This patch can, in principle, be fed at a sin
gle point. However, some asymmetry of the pattern re
sults, due to the presence of higherorder modes in the
patch. Provided that the two points shown in Fig. 7 are fed
with signals that are of equal amplitude and in antiphase,
that is, at 01 and 1801, respectively, then a symmetric ra
diation pattern is formed that is linearly polarized in the
horizontal direction. The symmetry results from cancella
tion of the higherorder modes in the patch. If two further
feeds are added as shown in the central patch example,
and if these are fed with independent signals having equal
amplitude and in antiphase, then a dual linearly polarized
patch is formed. If these four points are fed with equal
amplitude and in a phase sequence of 01, 901, 1801, and
2701, then circular polarization with highly symmetric
patterns and low crosspolarization will result. It has also
been shown that the threepointfed element shown on the
right of Fig. 7, when fed with equal amplitude and a phase
sequence of 01, 1201, and 2401, will also result in symmet
ric patterns but with a considerably wider input return
loss bandwidth.
These multiplefed patches can be used in arrays using
the feed structures described earlier. However, a very sim
pliﬁed feed results if the multiple feedpoints are distrib
uted among the elements in an array. Figure 8a shows
four circular patches fed by a corporate feed. As the patch
es are fed diagonally, each will radiate with a diagonal
polarization. The top left patch will have a polarization
inclined to ÷451, while the top right one will have ÷451,
and so on. However, the feed is arranged to provide the top
left patch with a phase of 01 and the top left with a phase
of 901, shifting the top power splitter to the left by a line
length equivalent to 451. These two patches together thus
generate circular polarization. The bottom pair is likewise
arranged and generates circular polarization in the same
hand as the top pair. However, it can be seen that these
patches are fed from the top rather than the bottom, thus
providing an inversion. By offsetting the central power
splitter, the two pairs are fed to reinforce. The result is a
feed that produces a phase sequence of 01, 901, 1801, and
2701, to feed patches that have a similar rotation se
quence.
Although this illustrates the general polarization prin
ciple, it is usual when generating circular polarization to
use circularly polarized elements as shown in Figs. 8b and
8c. There are slight pattern asymmetries induced by the
displacements of the linearly polarized patches in Fig. 8a,
which can be removed in this way. Figure 8b shows an
arrangement in which pairs of patches are rotated and
Fig. 8c, one in which all four elements are rotated.
Figure 9 shows the performance improvement of
sequential rotation applied to an eightelement array.
Figure 9a shows that the boresight 3dB axial ratio band
width has been increased by several factors and Fig. 9b,
that the input return loss bandwidth has also significantly
increased. Further information on the effects of sequential
rotation of the array radiation pattern is provided in Refs.
31 and 32.
Dual polarization arrays can also beneﬁt from sequen
tial rotation [33]. Figure 10 shows two examples. Figure
10a shows a 2 2element array of square patches where
dual linear polarization is achieved by feeding orthogonal
edges of the patch. As noted above, this leads to poor cross
polarization in the radiation patterns and to isolation be
tween the input ports. By arranging each successive patch
to ensure rotated feeding, the isolation is improved, and
the pattern problems are balanced out. For the array
shown in Fig. 10a, improvements of nearly 10 dB in the
isolation are noted. Figure 10b shows how a linear array
can be similarly fed; in this case rotation is applied to the
two groups of four patch elements.
Figure 11 shows how sequential rotation can be applied
to dual circularly polarized arrays. Feed 1 has positive
feed progression, while feed 2 is negative, with appropri
ate feedpoint rotation to offset the phasing. In this case,
crosspolarization is reduced and input match is improved;
however, isolation is unchanged from a conventionally fed
array.
2.7. Phased Arrays
The microstrip array can be used to create phased arrays
by the addition of phase shifters [34]. In general, due to
the planar nature of the microstrip, this medium is ideal
for integration of active devices, and there are examples
where multilayer structures incorporate the phase shift
ing elements on layers below the radiating elements.
Many arrays have been built on lowdielectricconstant
substrates, and this can lead to low costs, although the
array cost will then be dependent primarily on the active
device cost and hence on array size. Alternatively, more
thought has been given to the integration of both the ra
diating patch and the phase shifter onto a semiconducting
substrate. This has the potential to allow cost reduction
Figure 7. Multiple feeding of microstrip patches.
(a) (b) (c)
Figure 8. Sequentially rotated patch arrays (typical dimensions
for operation at 8 GHz, patch diameter =12.25mm, patch spacing
=21.4mm, patch notch size =21 mm, substrate height, feed=
1.586mm, e
r
=2.32) [32,33].
2572 MICROSTRIP ANTENNA ARRAYS
through chip mass production or production of the com
plete array on a single wafer, which is possible at milli
metric wavelengths. However to date no highvolume
chips have been manufactured that adopt this approach.
One of the reasons for this is the high dielectric constant of
semiconducting substrates, which cause substantial wave
trapping in the form of surface waves, which lead to the
scan blindness phenomenon. This aspect of phased array
design is discussed later in Section 3.3.
2.8. Conformal Arrays
Microstrip is an ideal medium in which to make conformal
arrays as the radiating elements are relatively thin and
the substrates can be formed in a variety of shapes [35–
39]. Cylindrical arrays are relatively straightforward as
the substrate can be easily bent around the cylinder. Con
ical arrays have also been made in this way.
Synthesis of the radiation pattern of conformal arrays
is more difﬁcult, and procedures vary with the geometry.
More details are given in the Section 3.1.
2.9. MultipleBeam Arrays
Many forms of beamformer lend themselves to printed
circuit production methods and can thus be easily
integrated with microstrip arrays [40]. For example,
Fig. 12 shows a Butler matrix designed to give four beams
from the fourelement patch array to which it is connected
[41]. The Butler matrix is formed from a number of inter
connected hybrid couplers, which present a set of excita
tions to the linear array that give rise to multiple beams;
that is, the amplitudes are equal and the phase difference
between elements is such to produce beams spaced in the
plane of the array. The array in Fig. 11 consists of four
microstrip patches. In principle, each element can assume
any form, including linear arrays of patches of series
fed discontinuity arrays described earlier. This allows a
twodimensional array to be formed with a fan of pencil
beams. Such a set of beams is ideal when coverage of a
horizontal area in front of the array is needed; a radar
antenna for collision avoidance is an example where the
multiple beams deployed across the highway enable target
discrimination to be achieved.
The Butler matrix forms, in general, a set of orthogonal
beams, in which the beam spacing and sidelobe level are
predetermined to minimize the coupling between the
beams. The Blass matrix gives more ﬂexibility and per
mits the shaping of beams and arbitrary sidelobe level and
overlap to be obtained. However, power is lost in resistive
terminations on the beamformer circuit.
Both of these beamformers are circuitbased and permit
integration with a microstrip array in either coplanar
(Fig. 12) or multilayer (Fig. 6) format. One conﬁguration
in which a Blasslike beamformer is integrated with a
patch array is shown in Fig. 13 [42]. The radiating ele
ments are seriesconnected patch arrays. Beneath each
patch of each series array lies a feedline, and coupling be
tween the feedlines is through the fringing ﬁelds. Between
(a)
(b)
Figure 10. Examples of dualpolarized sequentially rotated
arrays.
2
1
2
1
2
1 2
1
1 = 0°
2 = 0° 2 = –90°
1 = 90°
2 = –180°
1 = 180°
2 = –270°
1 = 270°
Figure 11. Sequential rotation for dual circularly polarized
arrays.
Frequency (GHz)
2.1 2.2 2.3 2.4 2.5 2.6
A
x
i
a
l
r
a
t
i
o
(
d
B
)
0
2
4
6
8
10
12
14
Frequency (GHz)
2.15 2.20 2.25 2.30 2.35 2.40 2.45 2.50 2.55
V
S
W
R
1.0
1.2
1.4
1.6
1.8
2.0
Figure 9. Performance of sequentially rotated arrays (     conventional;         sequential);
(eight elements, substrate height =4 mm, e
r
=2.26).
MICROSTRIP ANTENNA ARRAYS 2573
the coupling areas in the feedline are meandered line
lengths that set the phase shifts between elements. Each
feed has a different meander length, and this excites a
phase difference between the seriesconnected patch ar
rays proportional to the meander lengths of the other
feeds. Each feed excites the whole array. A set of beams
in the plane of the feedlines is created, with as many
beams as there are feedlines. Coupling between the feed
lines is minimized by the cancellation of the power coupled
out of each seriesconnected array into the other feedlines,
provided, of course, that an orthogonal beam set is excited.
If the set is not orthogonal, then power will be absorbed in
the matched terminations on the feedlines and the reﬂec
tion coefﬁcient seen at the input to the feedlines will be
affected.
A circular conformal version of this multibeam array
has been developed that is capable of creating a set of
beams ﬁlling the azimuth plane of the array [43].
2.10. Reﬂectarrays
The microstrip reﬂectarray concept is illustrated in
Fig. 14. It results in a reﬂector antenna whose reﬂector
surface is ﬂat and hence conformal, but whose bandwidth
is much narrower than that of a conventional parabolic
metal reﬂector [44,45].
The reﬂectarray works on the principle that the reﬂec
tion phase of each element is appropriate for the produc
tion of a uniform phase front in the mainbeam direction.
As the path length from the feed to the central element is
different from the path length to the outer elements, the
reﬂection phase of each element must vary from the center
to the edge of the array. This reﬂection phase can be con
trolled in a number of ways [46,47]. Figure 15a shows how
the patch size can be used. Figure 15b shows patches
whose phase is controlled by a section of transmission line
attached to the patch.
2.11. Active and Grid Arrays
The insertion of active devices into the radiating elements
of microstrip arrays gives rise to a distinct class of array
antennas known as active integrated antennas, with per
formance very different from that of conventional micro
strip arrays. Several books and review papers have
described design techniques and performance capabilities
[5,48,49].
Microstrip antennas can be integrated either with an
oscillating transistor or a power ampliﬁer to form trans
mitting elements or with detecting diodes, downconver
sion mixers, or lownoise ampliﬁers to form receive
Figure 13. Microstrip multiplebeam array (typical dimensions
for operation at 8.3 GHz with ﬁve beams, patch size =11.8
5.9 mm, spacing along resonant arrays =24.1mm, spacing be
tween resonant arrays =18.75mm, linewidths =0.5 mm, sub
strate height, feed=0.793mm, patch=1.586mm, e
r
=2.32) [42].
Feed
Array
Figure 14. Microstrip reﬂectarray.
Figure 12. Printed circuit Butler matrix and patch array.
(a) (b)
Figure 15. Microstrip reﬂectarray techniques: (a) size modula
tion; (b) tab length modulation.
2574 MICROSTRIP ANTENNA ARRAYS
elements. Figure 15 shows examples of both classes. All of
these can then be used to make active arrays.
Active arrays can be conﬁgured with conventional ar
ray spacing or very close spacing. Those with conventional
spacing have been used to demonstrate the principles of
such arrays, which include the integration of frontend
functions with conventional beamforming. Close spacing
has been used primarily in power combining arrays,
where the motivation is to group together as many active
elements as possible to create high powers.
Figure 16 shows an example of a conventionally spaced
array that attempts to combine transmit and receive
action with dual polarization [50]. The inclusion of active
devices into antennas in general renders the resulting
component nonreciprocal; that is, either transmit or re
ceive. Most array examples are of one type or the other. In
Fig. 16 a transmit power combing array is formed in the
horizontal polarization by the use of two patch oscillators.
The transistor and its source and gate loading lines can be
seen attached to the outer edges of the two patches. Re
ception is in the vertical polarization where a lownoise
ampliﬁer and front end would normally be connected to
the output port. The critical issue in duplex communica
tion systems is the separation, or isolation, of the transmit
and receive signals. This isolation determines the maxi
mum transmit power and hence the range of the commu
nication link. Isolation of around 25dB is achieved by the
use of orthogonal polarizations, and an additional 25 dB
comes with sequential rotation in the receive–feed net
work. An 8 8 array of such elements could achieve a kil
ometer range. Active integrated transmit arrays also
allow simple forms of beam scanning. The array of Fig.
16 can be phasescanned if the freerunning frequency of
each patch oscillator is varied to form an appropriate
phase front across the array. As the array is frequency
locked, each element continues to oscillate at a common
frequency, while the radiated phase is determined by the
difference between the (national) freerunning frequency,
controlled, for example, by the oscillator bias voltage, and
the radiated frequency. Figure 17a shows this arrange
ment. Wideangle scanning of a fourelement array has
been demonstrated using this technique [51].
Figure 17b shows an arrangement of patch oscillators
in which variation of the phase of the end elements gives
rise to beam scanning [52]. The freerunning frequencies
of these elements can be controlled through their bias
voltage, or by a varactor diode, through what is called the
tuning port. All elements are locked to their neighbours
through the mutual fringing ﬁelds, called soft locking, or
through coupling circuits, called hard locking. A hard
locking arrangement is shown. The end elements, A and
B, are tuned in the opposite direction. This makes the
phase of one increase while the other decreases. This
phase difference is transmitted down the array and re
sults in a linear phase front giving a wellcontrolled
scanned beam. Scanning of twodimensional arrays using
this technique has proved harder to achieve.
Such scanning methods give rise to limitations in the
maximum data rate that can be transmitted [53]. When
data are applied by modulating the central element of a
linear form of Fig. 17, the delays introduced by propaga
tion of the locking signal along the array causes signal
distortions. Frequency shift keying at rates of around
40 Mbps (million bits per second) have been shown to in
troduce significant amounts of amplitude modulation into
the otherwise frequency modulation for a fourelement
array. Modulation of all array elements overcomes this
problem, but increases the complexity of the array.
3. DESIGN ISSUES
3.1. Pattern Synthesis
Synthesis of the radiation pattern shape of microstrip lin
ear arrays follows the fundamental methods deﬁned for
general arrays [54]. In principle, a uniform phase front is
initially speciﬁed to produce a beam in the desired direc
tion, and then beamwidth and sidelobe level are controll
˙

ed by the array length and amplitude distribution.
Figure 16. Microstrip duplex active integrated antenna array
(typical dimensions for operation at 4.1 GHz, patch size =24
24mm, interconnecting linewidth=1.55mm, substrate height =
0.793mm, e
r
=2.32, FET=atf26,884, FET stub width=0.5 mm,
lengths =5, 19mm) [50].
(a)
(b)
Voltage
controlled
oscillators
Voltage
controlled
oscillators
Antennas
Antennas
Tuning ports
Tuning ports
Coupling
circuits
A B
Injected signal
Figure 17. Active integrated antenna array conﬁgurations for
beam scanning.
MICROSTRIP ANTENNA ARRAYS 2575
For planar twodimensional arrays, this technique is ap
plied in each principal plane. For twodimension confor
mal arrays, such as the cylinder or cone, such separation
of the distributions in the principal planes is not possible,
and more complicated synthesis procedure are necessary
[35].
If the linear array is seriesfed, then the synthesis pro
cedure is modiﬁed to take account of the reducing power
down the array as a proportion of it is coupled to radiation
at each element [55]. While the techniques described
above form the basis of synthesis, there are limits to the
performance achievable. In particular, the sidelobe level is
limited by manufacturing tolerances. While these have
been quantiﬁed for general arrays [56], a specific study
has shown that in patch arrays it is the precision with
which the patch resonant frequency can be deﬁned that
sets the lower limit to sidelobe performance. Variations in
the size of the patch, which affect the resonant frequency,
are caused by variations in the photoetching process. The
variations are the order of the thickness of the copper
cladding on the substrate. These variations give rise to
errors in the phase of the radiated signal, which degrade
sidelobe level. This effect is more pronounced as the fre
quency increases but can be made more manageable by
using thinner copper cladding. The study showed that a
lower sidelobe level of around 30 dB could be achieved in
practice without special measures.
3.2. Feed Loss and Radiation
In large microstrip arrays, losses in the feed network limit
the maximum achievable gain [57,58]. There are two
sources of such loss: (1) the resistive losses in the dielec
tric substrate and metallic conductors give rise to loss; and
(2) if a coplanar feed is used, then radiation loss can be
come significant, which can, in addition to limiting gain,
significantly compromise the radiation pattern control.
Figure 18a shows calculated microstrip line loss for
various substrates typically used in arrays and a variety
of line impedances. It can be seen that line loss increases
with increasing dielectric constant and impedance and re
ducing substrate thickness. Although typically less than
0.2 dB/line wavelength, this can be significant for arrays
many wavelengths in size. Figure 18b illustrates typical
radiation levels. Radiation loss from a rightangle bend is
given and is seen to reduce with increasing dielectric con
stant and impedance and reducing substrate thickness.
Again, while less than 0.5dB per bend, a large corporate
feed will have many bends, and splitters and radiation loss
can be substantial.
For large microstrip arrays, these losses can become
substantial. For example, for a substrate with thickness
h/l =0.06 and e
r
=2.32, total losses of the order of 4dB
may occur for an array size of 16 elements and up to 10dB
for 256 elements. A similar situation occurs for e
r
=1.06
representing foam or airspaced substrates. These losses
limit the array gain as shown in Fig. 19. As the array size
is increased, two things happen: (1) the array gain in
creases because of the increasing aperture size; (2) how
ever, for arrays of 10 wavelengths square and greater, this
gain increase is offset by significant loss. Eventually this
Line impedance (Ω)
20 40 60 80 100 120 140 160
20 40 60 80 100 120 140 160
L
i
n
e
l
o
s
s
(
d
b
/
λ
m
)
0.00
0.04
0.08
0.12
0.16
0.20
0.24
iii
iii
ii
ii
i
i
Line impedance (Ω)
R
i
g
h
t
a
n
g
l
e
b
e
n
d
r
a
d
i
a
t
i
o
n
l
o
s
s
(
d
B
)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
iii
iii
ii
+
+
ii
i
i
(a)
(b)
Figure 18. Losses in microstrip arrays: (a) calculated line loss;
(b) calculated radiation loss from rightangle bend (     e
r
=1.0;
    e
r
=2.32); curves (i) h/l =0.06, (ii) h/l =0.03, (iii) h/l =0.01;
smooth 901 bend, e
r
=2.32, h/l =0.06,; bend radius, Kr =0.2l
0
;
÷r =0.4l
0
[57].
Array size (D/λ)
1 10 100
G
a
i
n
(
d
B
)
0
10
20
30
40
Figure 19. Microstrip array gain (efﬁciency—solid line =100%,
dotted line =50%, dotted–dashed line =10%; K calculated, ’
measured, feed impedance 120O, D measured feed impedance
200O; h/l =0.06, e
r
=2.1) [57].
2576 MICROSTRIP ANTENNA ARRAYS
increase in loss becomes greater than the increase in gain
and a maximum of gain of around 35dBi occurs.
Feed radiation also affects the radiation pattern. Figure
20 shows calculated radiation patterns for 16 16 patch
arrays. The solid line is the envelope of sidelobe peaks
from the patch radiators for a uniform amplitude and
phase aperture distribution. The dotted lines show side
lobes, due to the feed radiation, that are significantly
higher than those due to the patches.
3.3. Mutual Coupling
Mutual coupling between the radiating elements in arrays
is a significant factor that both perturbs the aperture dis
tribution and affects the input impedance of the elements.
Figure 21 shows the mutual coupling between microstrip
patch elements [59]. It is seen that a coupling in the E
plane is stronger than in the H plane, and for e
r
=2.55 and
spacing of 40.25l
0
, as typically used in ﬁxedbeam arrays,
it is less than ÷20 dB.
In scanning arrays, the position is more complicated
and in some cases, mutual coupling leads to undesirable
blindness problems. Such effects are difﬁcult to both mea
sure and compute. The performance of patch elements in
inﬁnite arrays has been computed by the method of
moments and Floquet mode theory and latterly by the ﬁ
niteelement method. Figures 22 and 23 show computed
results for rectangular patches in ﬁnite arrays [60,61].
Figure 22a shows the input reﬂection coefﬁcient of the
center element of various size arrays or e
r
=2.55 substrate
against scan angle. It can be seen that mutual coupling
Theta (deg)
Theta (deg)
–30
–10
–20
(a)
(b)
–20
–10
–30
(
d
B
)
(
d
B
)
–90 –60 –30 0 30 60 90
–90 –60 –30 0 30 60 90
Figure 20. Calculated radiation pattern of 16 16element
patch array (     sidelobe envelope of patches): E plane—copo
larized feed pattern; H plane—crosspolarized feed pattern [57].
Patch spacing s/z
0.00 0.25 0.50 0.75 1.00 1.25

S
1
2

2
(
d
B
)
–45
–40
–35
–30
–25
–20
–15
–10
–5
0
e plane measured
h plane measured
e plane calculated
h plane measured
W S
L
E plane
W
S
L
H plane
Figure 21. Mutual coupling between coaxially fed microstrip
patches (     calculated, K measured; W=10.57cm, L=
6.55cm, d=0.1588cm, e
r
=2.55, f =1.41GHz) [59].
Theta (deg)
Theta (deg)

R

E plane 7 x 7
H plane 7 x 7
E plane infinite array
H plane infinite array
E
f
f
i
c
i
e
n
c
y
1 x 1
3 x 3
7 x 7
11 x 11
(a)
(b)
90 60 30 0
90 60 30 0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
Figure 22. Calculated performance of microstrip patches in
arrays: (a) reﬂection coefﬁcient magnitude versus scan angle;
(b) efﬁciency of ﬁnite patch array versus Eplane scan angle (e
r
=
2.55, substrate thickness d=0.04l
0
, patch size L=W=0.3l
0
,
patch spacing a=b=0.5l
0
, feed position X= ÷L/2, Y=0) [61].
MICROSTRIP ANTENNA ARRAYS 2577
causes the reﬂection coefﬁcient to increase with scan angle
for both E and Hplane scans. Results are similar to those
for the inﬁnite array case for scan angles ranging from 01
to 601. The effect of this change in reﬂection coefﬁcient is to
reduce array efﬁciency at large scan angles, and this effect
is worse for large arrays (Fig. 22b). In practice, some
phased arrays incorporate circulators between the trans
mitter and the radiating element to absorb this reﬂected
power. Others incorporate active matching, but this is a
complicated and expensive solution.
Mutual coupling increases as the dielectric constant
increases (Fig. 23). The attempts to integrate microstrip
patch arrays onto semiconducting substrates led to simi
lar computations for patch arrays on substrates with e
r
=
12.8. While Hplane scanning performance (Fig. 23a) is
similar to that on lowdielectricconstant substrates,
Eplane scanning shows that the reﬂection coefﬁcient
becomes greater than 1.0 at scan angles near 401. This
is a classical scan blindness condition in which it is
impossible to radiate power as the element is highly
mismatched. The fact that the reﬂection coefﬁcient is
greater than 1.0 does not violate the principle of energy
conservation. The result is for one element of the array
only. Other elements will have reﬂection coefﬁcient less
than 1.0 to achieve a power balance. At this blind angle,
array efﬁciency is, of course, very low, as shown in Fig. 23b.
Similar computations have also been done for other
types of patches, including circular patches [62], one and
twoprobefed circular patches [63], and proximity coupled
patches [64].
BIBLIOGRAPHY
1. J. R. James, P. S. Hall, and C. Wood, Microstrip Antenna The
ory and Design, Peter Perigrinus, IEE, London, 1981.
2. J. R. James and P. S. Hall, eds., Handbook of Microstrip An
tennas, Peter Perigrinus, IEE, London, 1989.
3. D. M. Pozar and D. H. Schaubert, eds., Microstrip Antennas,
IEEE Press, selected reprint volume, New York, 1995.
4. K. F. Lee and W. Chen, eds., Advances in Microstrip and
Printed Antennas, Wiley, New York, 1997.
5. K. C. Gupta and P. S. Hall, eds., Analysis and Design of In
tegrated Circuit Antenna Modules, Wiley Series in Microwave
and Optical Engineering, Wiley, New York, 2000.
6. Ref. 1, pp. 116, 161.
7. P. S. Hall and J. R. James, Design of microstrip antenna
feeds—Part 2: Design and performance limitations of triplate
corporate feeds, Proc. IEE 128(Part H):26–34 (1981).
8. J. C. Williams, Cross fed printed aerials, Proc. 7th European
Microwave Conf., Copenhagen, 1977, pp. 292–296.
9. A. G. Derneryd, Linearly polarised microstrip antennas,
IEEE Trans. Anten. Propag. AP24:846–851 (1976).
10. M. Tiuri, J. Henriksson, and S. Tallquist, Printed circuit radio
link antenna, Proc. 6th European Microwave Conf., Rome,
1976, pp. 280–282.
11. R. Hill, Printed planar resonant arrays, Proc. IEE Int. Conf.
Antennas and Propagation, York, 1987, pp. 473–476.
12. J. R. James and P. S. Hall, Microstrip antennas and arrays;
Part 2—new array design technique, IEE J. Microwaves
Optics Acousties, 1(5):175–180 (Sept. 1977).
13. E. R. Cashen, R. Frost, and D. E. Young, Improvement Relat
ing to Aerial Arrangements, British Provisional Patent (EMI
Ltd.) Speciﬁcation, 1294024.
14. T. Metzler, Microstrip series arrays, IEEE Trans. Anten.
Propag. AP29:174–178 (1981).
15. P. S. Hall, Microstrip linear array with polarisation control,
Proc. IEE 103(Part H):215–224 (1983).
16. G. von Trentini, Flachantenna mit periodisch gebogenem lei
ter, Frequenz 14:230–243 (1960).
17. D. J. Skidmore and G. Morris, Design and performance of
covered microstrip serpent antennas, Proc. IEE Int. Conf. An
tennas and Propagation, Norwich, 1983, pp. 295–300.
18. S. Nishimura, K. Nakano, and T. Makimoto, Franklin type
microstrip line antenna, Proc. IEEE APS Symp., Seattle,
1979, pp. 134–137.
19. J. Henriksson, K. Markus, and M. Tiuri, Circularly polarised
travelling wave chain antenna, Proc. 9th European Micro
wave Conf., Brighton, 1979.
20. W. Menzel, New travelling wave antenna in microstrip, Proc
8th European Microwave Conf., Paris, 1978, pp. 302–306.

R

0.0
0.2
0.4
0.6
0.8
1.0
1.2
E plane 7 x 7
H plane 7 x 7
E plane infinite array
H plane infinite array
Theta (deg)
Theta (deg)
E
f
f
i
c
i
e
n
c
y
0.0
0.2
0.4
0.6
0.8
1.0
1 × 1
3 × 3
7 × 7
11 × 11
(a)
(b)
90 60 30 0
90 60 30 0
Figure 23. Calculated performance of microstrip patches in ar
rays: (a) reﬂection coefﬁcient magnitude versus scan angle; (b)
efﬁciency of ﬁnite patch array versus Eplane scan angle
(e
r
=12.8, substrate thickness d=0.06l
0
, patch size L=0.107l
0
,
W=0.15l
0
, patch spacing a=b =0.5l
0
, feed position X= ÷L/2,
Y=0) [61].
2578 MICROSTRIP ANTENNA ARRAYS
21. P. W. Chen, C. S. Lee, and V. Nalbandian, Planar double layer
leaky wave microstrip antenna, IEEE Trans. AP50(6):832–
835 (June 2002).
22. P. S. Hall, Multioctave bandwidth log periodic microstrip
array antenna, Proc. IEE 133(Part H) (2):127–137 (April
1986).
23. H. Pues, J. Bogaers, R. Pieck, and A. Van de Capelle, Wide
band quasilogperiodic microstrip antenna, Proc. IEE
128(Part H) (3):159–163 (1981).
24. P. S. Hall and A. Sparrow, Microstrip log periodic antenna
array with endire beam, Electron. Lett. 23(17):912–913 (Aug.
13, 1987).
25. J.F. Zurcher, The SSFIP—a global concept for high perfor
mance broadband planar antennas, Electron. Lett.
24(23):1433–1435 (Nov. 1988).
26. H. Legay and L. Shafai, Newstacked microstrip antenna with
large bandwidth and high gain, IEE Proc. Microwave Anten.
Propag. 141(3):199–204 (1994).
27. F. Croq and D. M. Pozar, Millimeter wave design of wide band
aperture coupled stacked microstrip antennas, IEEE Trans.
Anten. Propag. AP39(12):1770–1776 (Dec. 1991).
28. J. C. MacKinchan, P. A. Miller, M. R. Straker, and J. S.
Dahele, A wide bandwidth microstrip subarray for array ap
plications using aperture coupling, IEEE Antennas and Prop
agation Society Int. Symp. APS Digest, 1989, pp. 878–881.
29. P. S. Hall, J. S. Dahele, and J. R. James, Design principles of
sequentially fed wide bandwidth circularly polarised micro
strip antennas, Proc. IEE 136(Part H) (5):381–389 (Oct. 1989).
30. P. S. Hall, Application of sequential feeding to wide band
width circularly polarised microstrip patch arrays, Proc. IEE
136(Part H)(5):390–398 (Oct. 1989).
31. P. S. Hall, Dual polarisation antenna arrays with sequentially
rotated feeding, Proc. IEE 139(Part H) (5):465–471 (Oct.
1992).
32. M. S. Smith and P. S. Hall, Analysis of radiation pattern ef
fects in sequentially rotated arrays, IEE Proc. Microwave An
ten. Propag. 141(4):313–320 (Aug. 1994).
33. P. S. Hall and M. S. Smith, Sequentially rotated arrays with
reduced sidelobe levels, IEE Proc. Microwave Anten. Propag.
141(4):321–325 (Aug. 1994).
34. J. J. Schuss, et al., Design of the iridium phased array an
tennas, Proc. IEEE APS Int. Symp., Ann Arbor, 1993, pp.
218–221.
35. R. C. Hansen, Conformal Antenna Array Design Handbook,
AD A110091, Jan, 1982.
36. R. E. Munson, Conformal microstrip antennas and micro
strip phased arrays, IEEE Trans. Anten. Propag. 74–79 (Jan.
1974).
37. K. R. Jacobsen, Radiation from microstrip antennas mounted
on two dimensional objects, IEEE Trans. Anten. Propag. AP
32:1255–1259 (1984).
38. G. G. Sanford, Conformal microstrip phased array for aircraft
tests with ATS6, IEEE Trans. Anten. Propag. AP26:642–646
(1978).
39. Ref. 2, pp. 1153–1191, 1227–1256.
40. A. R. Mishra, K. K. Sood, and A. Kumar, Cavity backed mi
crostrip patch array feed for multiple beam applications, Elec
tron. Lett. 34(1):4–6 (Jan. 8, 1998).
41. B. P. Pirollo, E. Okon, R. A. Lewis, P. Gardner, G. Ma, L. Piel,
L. Flamini, P. S. Hall, and K. Davies, Millimetric technologies
for future vehicle communications, Proc. Int. Conf. Micro
waves and MillimeterWave Technology, Beijing, Aug 2002,
pp. 1–6.
42. P. S. Hall and S. J. Vetterlein, Integrated multiple beam mi
crostrip array, Microwave J. 35(1) (Jan. 1992).
43. P. S. Hall and S. J. Vetterlein, Advances in microstrip multiple
beam arrays, Proc. 7th IEE Conf. Antennas and Propagation,
York, April 1991, pp. 129–132.
44. J. Huang, Miccrostrip reﬂectarray, Proc. IEEE APS Int.
Symp., London, Ontario, June 1991, pp. 612–615.
45. Y. Zhuang, J. Litva, C. Wu, and K.L. Wu, Modelling studies of
microstrip reﬂectarrays, IEE Proc. Microwaves Anten. Prop
ag. 142(1) (Feb. 1995).
46. T. A. Metzler, Stub loaded microstrip reﬂectarrays, IEEE An
tennas and Propagation Society Int. Symp., APS Digest,
1995, Vol. 1, pp. 18–23.
47. J. Huang and R. J. Pogorzelski, A Ka band reﬂectarray with
elements having variable rotation angles, IEEE Trans. Anten.
Propag. AP46:650–656 (May 1998).
48. A. Mortazawi, T. Itoh, and J. Harvey, eds., Active Antennas
and Quasioptical Arrays, IEEE Press, 1999.
49. K. Chang, R. A. York, P. S. Hall, and T. Itoh, Active integrated
antennas, IEEE Trans. Microwave Theory Tech. (50th anni
versary issue) 50(3):937–944 (March 2002) (invited paper).
50. M. J. Cryan, P. S. Hall, K. S. H. Tsang, and J. Sha, Integrated
active antenna with full duplex operation, IEEE Trans. Mi
crowave Theory Tech. MTT45(10):1742–1748 (Oct. 1997).
51. I. L. Morrow, P. S. Hall, and J. R. James, Measurement and
modelling of a microwave active patch phased array for wide
angle scanning, IEEE Trans. Anten. Propag. AP45(2):297–
304 (Feb. 1997).
52. P. Liao and R. A. York, A new phaseshifterless beam scan
ning technique using arrays of coupled oscillators, IEEE
Trans. Microwave Theory Tech. MTT41(10):1810–1815 (Oct.
1993).
53. C. Kykkotis, P. S. Hall, and H. GhafouriShiraz, Performance
of active antenna oscillator arrays under modulation for com
munication systems, IEE Proc. Microwaves Anten. Propag.
145(4):313–320 (Aug. 1998).
54. J. D. Kraus and R. J. Marhefka, Antennas For All Applica
tions, 3rd ed., McGrawHill, 2002, pp. 90–164.
55. B. B. Jones, F. Y. M. Chow, and A. W. Seeto, The synthesis
of shaped patterns with seriesfed microstrip patch
arrays, IEEE Trans. Anten. Propag. AP30(6):1206–1212
(Nov. 1982).
56. D. M. Pozar and B. Kaufman, Design considerations for low
sidelobe microstrip arrays, IEEE Trans. Anten. Propag. AP
38(8):1176–1185 (Aug. 1990).
57. P. S. Hall and C. M. Hall, Coplanar corporate feed effects in
microstrip patch array design, Proc. IEE 135(Part H) (3):180–
186 (June 1988).
58. E. Levine, G. Malamud, S. Shtrikman, and D. Treves, A study
of microstrip array antennas with the feed network, IEEE
Trans. Anten. Propag. AP37(4):426–434 (April 1989).
59. D. M. Pozar, Input impedance and mutual coupling of rect
angular microstrip antennas, IEEE Trans. Anten. Propag.
AP30(6):1191–1196 (Nov. 1982).
60. D. M. Pozar, Finite phased arrays of rectangular microstrip
patches, IEEE Trans. Anten. Propag. AP34(5):658–665 (May
1986).
61. D. M. Pozar, Scanning characteristics of inﬁnite arrays of
printed antenna subarrays, IEEE Trans. Anten. Propag.
AP40(6):666–674 (June 1992).
62. K. Solbach, Phased array simulation using circular patch ra
diators, IEEE Trans. Anten. Propag. AP34(8):1053–1058
(Aug. 1986).
MICROSTRIP ANTENNA ARRAYS 2579
63. J. T. Aberle and D. M. Pozar, Analysis of inﬁnite arrays of one
and twoprobefed circular patches, IEEE Trans. Anten. Prop
ag. AP38(4):421–432 (April 1990).
64. J. S. Herd, Modelling of wideband proximity coupled micro
strip array elements, Electron. Lett. 26:1282–1284 (Aug.
1990).
MICROSTRIP ANTENNAS
HOTON HOW
Hotech, Inc.
Belmont, Massachusetts
Microstrip circuitry consists of a metal strip or patch on a
dielectric substrate backed by a metal ground plane. Mi
crostrip antennas are ﬁnding increasing popularity owing
to their advantages in size, cost, conformity to the sup
porting structure, low proﬁle, and ease of fabrication. By
using simple etching techniques it is possible to fabricate a
wide variety of microstrip circuits including antenna ar
rays, feeding networks, and active devices such that pre
amps (preampliﬁers) or distributed transmitters can be
conveniently placed next to the antenna elements. In
addition, diode phase shifter circuits can also be etched
on the substrate to form singleboard phased arrays.
This article describes the microwave properties of patch
antennas.
The lateral dimensions of microstrip patch antennas
are large in comparison with the width of conventional
microstrip lines. The main purpose of a patch antenna is
to conﬁne microwave energy within a compact region with
a dimension compatible to the wavelength. Such conﬁne
ment can be achieved via the resonant behavior of a ﬁnite
guiding structure supporting propagation waves. As a
consequence, electromagnetic energy radiates away from
that part of the patch resonator that is open to free space,
or the periphery of the patch antenna, in a manner similar
to a slot opening cut along the side of a waveguide. In or
der to continuously radiate microwave energy, the patch
needs to be electrically connected to a ‘‘feeder line.’’ When
the characteristic impedance of the feeder line matches
the impedance of radiation waves, power is thus dumped
into the conﬁned space of the patch resonator without
causing much reﬂection. This means that all the input
microwave power delivered to the patch antenna is radi
ated away into free space.
At resonance electromagnetic excitations reinforce
within the conﬁned space of the patch antenna, attaining
high intensity far exceeding than at the feederline region.
Thus, radiation from the feeder line, or from a regular
microstrip transmission line of the same length, is insig
nificant in comparison to a microstrip antenna driven at
resonance. In microstrip circuitry power also radiates
from open/short circuits and from discontinuities, such
as corners. However, the radiated power is relatively
small, since the radiation impedances there are normally
very different from the characteristic impedance of the
microstrip transmission line encompassing these discon
tinuities.
This article introduces two calculational methods ca
pable of quantitatively describing antenna performance.
The ﬁrst method, the socalled resonant cavity model, is
less rigorous, but nevertheless has the advantage of being
more analytical and, hence, can be applied with ease. For
the cavity model perfect metal boundaries are assumed at
the metal patch and at the groundplane positions, and
magnetic wall boundary conditions are assumed at the
periphery, or at the sidewall position of the patch antenna.
Due to fringeﬁeld penetration, magnetic walls are located
slightly beyond the boundary contour of the patch geom
etry. This fringeﬁeld effect can thus be accounted for by
considering the patch to ﬁctitiously possess a boundary
whose dimension extends outward by a distance consis
tent with the spatial extent of the fringe ﬁelds. Radiation
and material imperfections are then considered as pertur
bations to a lossless cavity. In this manner, engineering
parameters of the microstrip patch antenna, such as the
resonant frequency, farﬁeld radiation pattern, input im
pedance, radiation linewidth, directivity, and efﬁciency,
are all calculable.
Although simple and straightforward, the drawback of
the aforementioned analysis is that surface waves can
hardly be accommodated by the cavity model, even if one
is attempted in a perturbative manner. The reason for this
limitation is that, unlike spatial waves, surface waves lack
in closedform expressions, and hence their existence is
inconsistent with the nature of the cavity model, which
requires calculations to be carried out analytically. Sur
face waves are identiﬁed as simple poles appearing in the
integrand of a spectrumdomain analysis characteristic of
the guiding structure of a microwave circuit supporting
electromagnetic wave propagation along the lateral direc
tions. Analogous to spatial radiation waves, while the as
ymptotic behaviors of surface waves can be readily
deduced in analytic forms in the farﬁeld zone, surface
waves can hardly be accessed analytically in the near
ﬁeld, since the decomposition of currents into multipole
expansions adopted by a conventional spatialwave anal
ysis is not available, rendering the failure for cavitymode
analysis. For this reason, the cavity model is considered
inappropriate for a microstrip structure involving a thick
substrate, or if the dielectric constant of the substrate is
large, because it is known that under these conditions ex
citation of surface waves is significant.
The second calculational method concerns the Green
function analysis. In general a Green function is deﬁned
as the solution of a (linear) differential/integral equation
resulting from a (deltafunctionlike) point source satisfy
ing the imposed (homogeneous) boundary conditions.
Thus, under an arbitrary source excitation, the solution
of the equation can be composed of a superposition of the
Green functions, still satisfying the boundary conditions of
the structure. In the presence of a point (dipole) current
source in association with the geometry of a stratiﬁed
structure consisting of dielectric/magnetic layers, the ex
cited electromagnetic ﬁeld is termed the Green function
dyad, due to the vectorial nature of the source ﬁeld
and the observer ﬁeld. The construction of dyadic Green
2580 MICROSTRIP ANTENNAS
functions of a microstrip circuit is straightforward, and
this article emphasizes its physical meanings and inter
pretations, as appearing in the original work by So
mmerfeld [1].
Sommerfeld solved for the ﬁrst time the Green func
tions associated with a horizontal point dipole and a ver
tical point dipole in the presence of a semiinﬁnite
conductor halfspace, our planet Earth [1]. Radiowaves
are therefore radiated into free space either directly from
the point dipole source or indirectly from the image dipole
source induced by the (imperfect) ground plane, the Earth
surface. They are called spatial (radiation) waves, because
they exhibit a 1/r spatial dependence, where r denotes the
distance from the original/image source point to the ob
server point. In addition, Sommerfeld also showed a sec
ond kind of radiation ﬁeld that is tied to the air–Earth
interface exhibiting a 1=
ﬃﬃﬃ
r
_
dependence, where r denotes
the 2D distance between the source point projected on the
Earth surface to the observer point located also on the
Earth surface [1]. These are called surface waves because
they decay exponentially when departing away from the
air–Earth interface. This article explains how the spatial
wave and the surfacewave solutions arrive in association
with a (horizontal) point dipole source in the presence of a
layered dielectric/magnetic structure.
By using the dyadic Green functions, one can then for
mulate the general ﬁeld solution in a stratiﬁed structure
excited by (external) current sources in terms of an inte
gral equation, which is subsequently solved by applying
the numerical (Galerkin) method. The external current
sources are expressed in the form of surface conductor in
homogeneities appearing at the microstrip patch location.
This article outlines procedures to calculate the engineer
ing parameters of a microstrip antenna of an arbitrary
geometry, and it illustrates calculational results on circu
lar microstrip antennas, showing plots in radiation
frequencies, bandwidths, farﬁeld patterns, input imped
ances, effects in feederline positions, and interference be
tween antennas. The dyadic Green functions involving
anisotropic/gyromagnetic materials are introduced in Sec
tion 3, where the transfer matrix technique is explained. A
transfer matrix translates the tangential components of
the electromagnetic ﬁeld over one layer of a stratiﬁed
structure satisfying the boundary conditions therein.
More recent developments on printed circuit antennas
are reviewed in Section 4, including the important broad
band techniques employing stacked parasitic elements to
achieve high gain and low crosspolarization level. Fractal
antennas involving a patch geometry exhibiting a frac
tional dimension are introduced in Section 5. More so
phisticated treatments on microstrip antennas may be
found in Ref. 2.
1. CAVITY MODEL
1.1. Introduction
A microstrip patch antenna is a narrowband device, and
the bandwidth covers typically about 5% of the center ra
diation frequency. When the bandwidth Df can be related
to the quality factor Q of a resonator, then
Q=2pf
maximumstoredenergy
averagedpower dissipation
_ _
=
f
Df
(1)
which implies that power dissipation is not significant and
the circuit of a microstrip patch antenna can be approx
imated as a lowloss cavity resonator. This suggests that
the performance of a microstrip patch antenna can be an
alyzed using a perturbation method; that is, the zeroth
order solution of the antenna is described in terms of the
eigenmodes, or normal modes, of a lossless cavity. Losses
are then added to the analysis as ﬁrstorder perturbations,
including conductor loss, dielectric loss, and radiation loss.
Magnetic loss can also be included if ferrites are used as
the substrate material.
Dissimilar to radiation loss, surfacewave loss can
hardly be treated as perturbations in the cavity model.
Radiation and surface waves represent leaky waves from
the antenna; the former leaks directly from air and the
latter, from the guided structure of the air–substrate in
terface. Surface waves lack analytic expressions and are
not compatible with the closedform representation of cav
ity modes. Surfacewave modes are identiﬁed as being as
sociated with simple poles appearing in the integrand of a
spectrumdomain analysis characteristic of the guiding
structure of surface discontinuities in the circuit geome
try. Surfacewave modes are readily included with a full
wave analysis, as discussed in the following section,
introducing Green’s function formulation.
In contrast, the cavity model is easy to apply, allowing
engineering parameters of a microstrip patch antenna to
be calculated analytically, including bandwidth, input im
pedance, radiation efﬁciency, and nearﬁeld/farﬁeld radi
ation patterns. Furthermore, the physical meaning of
normal modes is evident in the cavity model. For exam
ple, the left/righthand polarized radiations from a micro
strip ferrite patch antenna can be readily calculated by
using the cavity model. However, the cavity model will
provide satisfactory calculations only if surfacewave loss
is insignificant. Surfacewave loss is not important for a
thin substrate exhibiting low dielectric constant, but not
for a thick substrate with high dielectric constant. Also,
the cavity model fails to describe the coupling effects be
tween microstrip circuits placed close to each other. For
these complex situations a fullwave analysis is needed,
and the cavity model is too simple to be applicable. The
fullwave analysis involving the use of dyadic Green’s
functions is discussed in Section 2.
1.2. Resonant Frequency
Let a metal patch be deposited on top of a dielectric sub
strate backed by a ground plane. The patch geometry con
sidered in this section is either rectangular or circular.
Only dielectric substrates are considered in this section.
Formulation of cavity normal modes for patch antennas
fabricated on ferrite substrates can be found in Ref. 3. In
the cavity model one assumes a lossless substrate in which
the cavity is formed that is bounded by either electric or
magnetic walls. The metal surfaces of the microstrip patch
MICROSTRIP ANTENNAS 2581
and the ground plane are approximated as perfect con
ductors, or electric walls, and the peripheral surface sur
rounding the antenna cavity directly under the patch
boundary is a magnetic wall. This is illustrated in Fig. 1,
in which a rectangular microstrip patch antenna is mod
eled as a resonator cavity with its periphery shown shaded
as a magnetic wall. An electric wall is deﬁned so that the
tangential component of the electric ﬁeld vanishes at the
wall surface, and a magnetic wall is deﬁned so that
the tangential component of the magnetic ﬁeld vanishes
at the wall surface. Thus, a normal metal surface ap
proaches an electrical wall if the value of conductivity goes
to inﬁnity. A magnetic wall imposes ﬁctitious boundary
conditions that insulate the inside of the cavity from the
outside, allowing no electromagnetic energy to exchange
across it.
The active volume of the cavity is slightly exceeds that
implied by the physical dimension of the patch geometry,
since fringe ﬁelds are excited at the periphery of the an
tenna. Therefore, the effective volume of the cavity con
sists of both regions directly under the patch pattern and
its surrounding containing the fringe ﬁeld. It is shown in
Fig. 2 that the effective radius of the cavity is larger than
that of the metal patch of the microstriop antenna. Thus,
the resonant length along the side of a rectangular patch
antenna, denoted as L
/
=L÷2D, is larger than the phys
ical length L and the effective increment in length due to
the fringeﬁeld effect is 2D with [4]
D
d
=0:412
e
eff
÷0:300
e
eff
÷0:258
_ _
L=d÷0:262
L=d÷0:813
_ _
(2)
where d is the thickness of the substrate, e
eff
denotes the
effective dielectric constant of the patch cavity given by [5]
e
eff
=
e
r
÷1
2
÷
e
r
÷1
2
1÷
10d
L
_ _
÷1=2
(3)
and e
r
is the dielectric constant of the substrate material.
The rationale behind Eq. (3) is that, depending on the ra
tio of d/L, electromagnetic ﬁelds can extend into air di
rectly above the metal patch, and so e
eff
oe
r
, and e
eff
Ee
r
if
d/L51. The resonant frequency of the cavity resonator is
therefore
f
n
=
cn
2L
/
ﬃﬃﬃﬃﬃﬃﬃ
e
eff
_ (4)
where c denotes the speed of light in vacuum and n the
order of the resonant mode (n=1 for the fundamental
mode).
On the basis of the same fringeﬁeld consideration, we
expect that for a circular patch antenna the effective res
onant radius of the cavity, denoted as R
/
=R÷D, is larger
than the physical radius of the metal patch, denoted as R,
and [6]
D=
d
p
ln
pR
2d
_ _
÷1:7726
_ _
(5)
as shown in Fig. 2. A similar expression for e
eff
, or Eq. (3),
is expected for a circular microstrip cavity. However, such
an expression is lacking in the present literature. We
therefore use e
r
as e
eff
, bearing in mind that this approx
imation is true only if Rbd. The resonant frequency of the
cavity is therefore
f
mn
=
cX
/
mn
2pR
/
ﬃﬃﬃﬃ
e
r
_ (6)
where X
/
mn
denotes the nth zero of the derivative of the
Bessel function J
m
(x) of order m.
We note that the expressions given above for resonant
frequencies for rectangular and circular microstrip patch
antennas [Eqs. (4) and (6)] are derived from the cavity
model assuming that standing modes are excited within
the cavity. At high frequencies this quasistatic picture is
no longer valid and surface waves of high orders are ex
cited propagating away from the cavity region guided by
the microstrip structure. As such, magneticwall boundary
conditions are no longer adequate at the periphery of the
antenna cavity. For these situations we must resort to a
rigorous fullwave analysis to suitably address the leaky
feature of the cavity generating surface waves, for exam
ple, by using the dyadic Green functions to be described in
Section 2. To simplify notations in the following, we will
d
W
L
(a 0) (b 0)
(0 0)
Figure 1. Feeding the rectangular patch antenna by a microstrip
line. The resonant cavity is located directly under the patch
bound by a magnetic wall shown shaded in the ﬁgure.
2(R + ∆)
2R
Figure 2. Feeding the circular patch antenna by a coaxial line.
Because of the fringeﬁeld effect, the effective radius of the reso
nator is larger than the resonator’s physical value.
2582 MICROSTRIP ANTENNAS
use e
r
for e
eff
and L and R for L
/
and R
/
, respectively, with
the difference understood.
1.3. Normal Modes and FeederLine Excitations
The normal modes excited in a microstrip resonator cavity
are normally TE waves exhibiting no nodal points along
the z axis, the direction perpendicular to the substrate
surface. Thus, the normalmode solutions show no z de
pendence, and the electromagnetic components can be
uniquely derived from E
z
satisfying the following Helm
holtz equation
(V
2
t
÷k
2
mn
)E
z
=0 (7)
where V
t
denotes the transverse part of the del operator
(with respect to the z axis), k
mn
is given by
k
mn
=o
mn
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
m
0
e
0
e
r
_
(8)
o
mn
is the angular frequency, and {m, n} refers to the in
dex of the normal mode. The boundary condition imposed
on E
z
is that the derivative of E
z
along the normal direc
tion of the boundary vanishes at the magnetic wall. Once
E
z
is solved from Eq. (7), the corresponding H ﬁeld is
H
t
=(jom
0
)
÷1
e
z
V
t
E
z
(9)
where e
z
denotes the unit vector along the z axis.
Let the rectangular patch be located at 0rxrL, and
0ryrW (see Fig. 1). The normal modes are therefore
E
mn
z
(x; y) = cos
mpx
L
cos
npy
W
(10)
with the corresponding modal wavenumber
k
mn
=
mp
L
_ _
2
÷
np
W
_ _
2
_ _
1=2
(11)
For a circular cavity of radius R the normal modes are (see
Fig. 2)
E
mn
z
(r; f) =J
m
(k
mn
r) exp(jmf) (12)
and the modal wavenumber is determined by the zeroing
condition of the derivative of Bessel functions of order m
J
/
m
(k
mn
R) =0 (13)
and (r, f, z) denotes the cylindrical coordinate.
Let us consider the patch antenna be excited by either a
microstrip line at the edge of the patch (Fig. 1) or by a
coaxial feeder directly under the patch (Fig. 2). The angu
lar frequency of the external driving ﬁeld is o=2pf. We
expect that the cavity will be driven to attain the highest
intensity when the excitation frequency approaches the
normalmode frequencies [Eq. (8)]. This phenomenon is
generally known as the forced oscillation of a resonator,
and it occurs commonly in many branches of physics.
Thus, at resonance, the cavity is driven by the external
ﬁeld undergoing the forced oscillation giving rise to a
maximum efﬁciency in radiation. We ﬁrst consider the ex
citation of the cavity by a microstrip feeder line. The
boundary conditions for the cavity are that electric walls
are located at metal boundaries and magnetic walls at the
periphery not adjacent to the microstrip feeder line. At the
input port of the feeder line adjacent to the cavity periph
ery we deﬁne a window where we assume that the exci
tation currents are uniformly distributed, inducing a
uniform magnetic ﬁeld there:
H=ne
z
h
0
(14)
here n denotes the unit vector pointing outward along the
normal direction of the window and h
0
is a constant spec
ifying the amplitude of the excitation current. The excita
tion ﬁeld following Eqs. (7) and (9) can be uniquely solved
subject to the boundary condition of Eq. (14).
For a rectangular patch depicted in Fig. 1, let the mi
crostrip feeder be connected to the patch at the location
deﬁned by arxrb, and y =0. The excitation ﬁeld is there
fore
E
z
(x; y) =jkzh
0
o
m=0
A
m
cos
mpx
L
cos b
m
(W÷y) (15)
H
x
(x; y) =h
0
o
m=0
A
m
b
m
cos
mpx
L
sin b
m
(W÷y) (16)
H
y
(x; y) = ÷h
0
o
m=0
A
m
mp
L
sin
mpx
L
cos b
m
(W÷y) (17)
where
b
m
=
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
k
2
÷
mp
L
_ _
2
_
(18)
k =o
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
m
0
e
0
e
r
_
z =
ﬃﬃﬃﬃﬃﬃﬃﬃ
m
0
e
0
e
r
_
(19)
A
m
=
4
mp
sin
mp(b ÷a)
2L
cos
mp(b÷a)
2L
b
m
sin b
m
W
; (mO0) (20)
A
0
=
b ÷a
Lk sin kW
(21)
Note that when o=o
mn
, then A
m
=N, indicating that
when the frequency of the external driving ﬁeld equals
one of the normalmode frequencies, only that normal
mode is excited in the cavity responsible for radiation.
This statement is valid only if the cavity is lossless. How
ever, when losses are included with the patch resonator,
other normal modes can also be excited at resonance, but
with much smaller amplitudes.
For a circular patch we consider the microstrip feeder
line to be located at ÷arfra, and r=R. The excitation
MICROSTRIP ANTENNAS 2583
ﬁeld can then be written as
E
z
(r; f) =jzh
0
o
n=0
C
n
J
n
(kr) cos(nf) (22)
H
r
(r; f) =
h
0
kr
o
n=1
C
n
nJ
n
(kr) sin(nf) (23)
H
f
(r; f) =h
0
o
n=0
C
n
J
/
n
(kr) cos(nf) (24)
C
n
=
2 sin na
npJ
/
n
(kR)
; (nO0) (25)
where
C
0
=
a
pJ
/
0
(kR)
(26)
and z and k are as given in Eq. (19). Again, from Eqs. (25)
and (26) it is seen that the coefﬁcient of C
n
goes to inﬁnity
if the driving ﬁeld frequency approaches the normalmode
frequencies, which are speciﬁed by Eq. (13), resulting in
forced oscillation of a lossless cavity resonator.
Let us consider the excitation of the patch antenna by a
coaxial line inserted directly under the patch. We assume
that this line shows an inner ﬁlament of zero diameter
ending in a point charge at the surface of the patch (see
Fig. 2). The excitation current density is therefore
J
e
(x; y) =I
0
d(x ÷x
0
)d(y ÷y
0
)e
z
for a rectangular patch
(27)
J
e
(r; f) =
I
0
d(r ÷r
0
)d(f ÷f
0
)
re
z
for acircular patch (28)
In the presence of a driving current, Eq. (7) is modiﬁed to
include an inhomogeneous term, the source, on the right
hand side:
(V
2
t
÷k
2
mn
)E
z
=jom
0
J
e
.
e
z
(29)
The boundary conditions are the same as before, leading
to the normalmode solutions, namely, electric walls at
metal boundaries and magnetic walls at the periphery of
the resonator cavity. Thus, Eq. (29) can be solved in terms
of the normalmode solutions:
E
z
(x; y) =jom
0
m;n
E
mn
z
(x; y)
k
2
÷k
2
mn
_
L
0
dx
/
_
W
0
dy
/
J
e
(x
/
; y
/
)E
mn
z
(x
/
; y
/
)
+
_
L
0
dx
/
_
W
0
dy
/
[E
mn
z
(x
/
; y
/
)[
2
(30)
E
z
(r; f) =jom
0
m;n
E
mn
z
(r; f)
k
2
÷k
2
mn
_
R
0
dr
/
_
2p
0
r
/
df
/
J
e
(r
/
; f
/
)E
mn
z
(r
/
; f
/
)
+
_
R
0
dr
/
_
2p
0
r
/
df
/
[E
mn
z
(r
/
; f
/
)[
2
(31)
These two equations apply for a rectangular patch and a
circular patch, respectively, and the normalmode solu
tions, E
mn
z
(x; y) and E
mn
z
(r; f), are as given in Eqs. (10) and
(12). We note that Eqs. (30) and (31) imply that when k
approaches k
mn
, the normalmode {m, n} acquires an inﬁ
nite amplitude, as implied by a lossless cavity. The corre
sponding magnetic ﬁeld can be derived using Eq. (9).
Thus, for the rectangular patch, we have
E
z
(x; y) =
4jom
0
I
0
LW
o
m;n=1
cos(mpx
0
=L) cos(mpx=L) cos(npy
0
=W) cos(npy=W)
k
2
÷k
2
mn
(32)
H
x
(x; y) =
4I
0
LW
o
m;n=1
np
W
cos(mpx
0
=L) cos(mpx=L) cos(npy
0
=W) sin(npy=W)
k
2
÷k
2
mn
(33)
H
y
(x; y) =
÷4I
0
LW
o
m;n=1
mp
L
cos(mpx
0
=L) sin(mpx=L) cos(npy
0
=W) cos(npy=W)
k
2
÷k
2
mn
(34)
For the circular patch, we have
E
z
(r; f) =
jom
0
I
0
p
o
m=0
o
n=1
(2 ÷d
m0
)k
2
mn
k
2
÷k
2
mn
J
m
(k
mn
r)J
m
(k
mn
r
0
) cos[m(f ÷f
0
)]
(k
2
mn
R
2
÷m
2
)[J
m
(k
2
mn
R
2
)]
2
(35)
H
r
(r; f) =
÷I
0
pr
o
m=1
o
n=1
m(2 ÷d
m0
)k
2
mn
k
2
÷k
2
mn
J
m
(k
mn
r)J
m
(k
mn
r
0
) sin[m(f ÷f
0
)]
(k
2
mn
R
2
÷m
2
)[J
m
(k
2
mn
R
2
)]
2
(36)
H
f
(r; f) =
I
0
p
o
m=0
o
n=1
(2 ÷d
m0
)k
3
mn
k
2
÷k
2
mn
J
/
m
(k
mn
r)J
m
(k
mn
r
0
) cos[m(f ÷f
0
)]
(k
2
mn
R
2
÷m
2
)[J
m
(k
2
mn
R
2
)]
2
(37)
where d
ij
denotes the Kro¨necker delta function and d
m0
=1
if m=0, and d
m0
=0, otherwise.
In Eqs. (32)–(34) and (35)–(37) the k
mn
values are as
given by Eqs. (11) and (13), respectively, and z and k are as
deﬁned in Eq. (19). Once the excitation ﬁelds are known
[Eqs. (16,17,18,22,23,24,32,33,34,35,36,37)], losses of var
ious kinds, and hence the quality factor Q of the patch
cavity can be calculated as discussed below.
1.4. Input Impedance
Having solved the electromagnetic ﬁelds inside a lossless
cavity, we can relax the assumption of perfect electric and
magnetic walls and allow electromagnetic waves to extend
2584 MICROSTRIP ANTENNAS
beyond the wall boundaries. This results in ohmic loss and
radiation (surfacewave) loss. Dielectric loss and magnetic
loss occur in the interior of the patch cavity, which can be
accounted for by assuming a lossy medium possessing a
complex permittivity and a complex permeability, respec
tively. We consider ﬁrst the radiation loss.
The Kirchhoff–Huygens principle, which is a vector an
alog of Green’s theorem, states that the electromagnetic
ﬁeld inside a closed volume V can be derived by the volume
charge and current distributions inside V and the surface
charges and currents distributed on the enclosing surface
of V, denoted as S. The effective electric surface current
density K
e
magnetic surface current density K
m
and elec
tric surface charge density S are respectively [7]
K
e
= ÷nH; K
m
=nE; S= ÷en
.
E (38)
where n denotes the unit outward vector normal to S and e
is the permittivity. Thus, if we consider the outside volume
of the patch antenna cavity as the volume V, we conclude
that the effective magnetic current density appearing on
the magnetic wall is
K
m
= ÷2ne
z
E
z
(39)
where n denotes the unit outward vector normal to
the magnetic wall whose sign changes as comparing to
Eq. (38), and the factor of 2 in Eq. (39) accounts for the
presence of the ground plane, K
e
and S do not show up on
a magnetic wall.
The radiation ﬁeld arising from the magnetic current
density K
m
can be derived by using the same formula de
scribing the electric current density K
e
but converted from
the duality rule [7]. The duality rule states that the elec
tromagnetic theory remains valid if one changes all the
electric quantities into the corresponding magnetic quan
tities and the magnetic quantities into the negative of the
corresponding electric quantities. Thus, the vector poten
tial associated with K
m
can be expressed as [8]
A
m
(r) =
e
0
d
4p
_
Cm
d‘
/
K
m
(r
/
)
[r ÷r
/
[
e
÷jk
0
[r÷r
/
[

e
0
d
4p
exp(÷jk
0
r)
r
_
C
m
d‘
/
t
/
E
z
(r
/
) exp(jk
0
e
^ rr
.
r
/
)
(40)
where d denotes the thickness of the substrate, C
m
is the
contour of the magnetic wall, t is the unit vector along the
contour C
m
in the counterclockwise sense, e
r
is the unit
vector pointing toward the observation point r, and
k
0
=o(m
0
e
0
)
1=2
(41)
is the wavenumber in air. In Eq. (40) we denote ‘‘primed’’
symbols as referring to the source point, and have as
sumed [r ÷r
/
[@d.
In the farﬁeld zone we have
H(r) = ÷joA
m
(r) (42)
E(r) = ÷z
0
e
r
H(r) (43)
where
z
0
=
m
0
e
0
_ _
1=2
(44)
is the wave impedance in air ( =377O). We derive, there
fore
E(r) 
je
0
k
0
d
4p
exp(÷jk
0
r)
r
_
Cm
d‘
/
(÷t
/
.
e
f
e
y
÷t
/
.
e
y
e
f
)E
z
(r
/
) exp(jk
0
e
r
.
r
/
)
(45)
where e
y
and e
f
denote the unit vectors along the y and f
directions at the observation point r. By using Eq. (45) the
farﬁeld differentialpower radiation from a circular patch
antenna fed by a microstrip line [Eq. (22)] is, for example
E(r) 
azh
0
k
0
d
p
R exp(÷jk
0
r)
r
o
m=÷o
j
÷m
J
m
(kR)
J
/
m
(kR)
sin ma
ma
e
÷jmf
.
e
y
J
/
m
(k
0
R sin y) ÷je
f
cos y
mJ
m
(k
0
R sin y)
k
0
R sin y
_ _
(46)
where it is understood that the ratio of sinma to ma is 1
when m=0. Expressions of E(r) for the other rectangular
patch geometry under both the microstrip and the coaxial
line feeder excitation conﬁgurations, or for the same
circular patch geometry under the coaxialline feeder ex
citation conﬁguration, can be derived in a similar manner,
but their explicit expressions are not given here.
The farﬁeld differentialpower radiation pattern can
be calculated by using the following equation:
dP
r
dO
= lim
r÷o
r
2
[E[
2
2z
0
(47)
The total radiation power is then
P
r
=
_
p
0
sin y dy
_
2p
0
df
dP
r
dO
(48)
Conductor loss and dielectric loss can also be derived from
the zerothorder normalmode solutions of the cavity [7].
The conductor loss is given by
P
c
=R
s
__
Sc
[H[
2
ds (49)
where S
c
denotes the patch surface and R
s
is the surface
resistance
R
s
=
ﬃﬃﬃﬃﬃﬃﬃﬃﬃ
om
0
2s
_
(50)
MICROSTRIP ANTENNAS 2585
and s is the conductivity of metal. The dielectric loss is
given by
P
d
=
oe
r
e
0
tan d
2
__
V
c
_
[E
z
[
2
dv (51)
where V
c
denotes the volume of the cavity and tan d is the
dielectric loss tangent.
The antenna efﬁciency or radiation efﬁciency is deﬁned
as the ratio of the radiated power to the input power
e =
P
r
P
r
÷P
c
÷P
d
100% (52)
The total stored electric energy is
W
e
=
e
r
e
0
4
__
V
c
_
[E
z
[
2
dv =
P
d
2otand
(53)
At resonance, because the stored magnetic energy W
m
equals the stored electric energy W
e
, we have thus the
total stored electromagnetic energy
W
T
=W
n
÷W
m
=
P
d
otand
(54)
The Q factor of the cavity is, from the definition of Eq. (1)
Q=
W
T
P
T
÷P
c
÷P
d
(55)
From which the VSWR bandwidth can be calculated. We
deﬁne S
max
to be the maximum value of VSWR that can be
tolerated in the radiation band. We then have [9]
VSWRbandwidth(%) =
100(S
max
÷1)
Q
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
S
max
_ (56)
Typically, S
max
=2.
The input susceptance of the antenna B can be calcu
lated for a lossless patch cavity, which is thus a zeroth
order quantity. In contrast, the input conductance G must
be calculated from the total loss of the cavity, which rep
resents a ﬁrstorder quantity. Thus
jB=
÷I
V
(57)
G=
P
r
÷P
c
÷P
d
[V[
2
(58)
where I and Vare the averaged input current and voltage
at the input feeder position. For coaxial feeder with exci
tation current given by Eqs. (27) and (28), assuming a thin
inner ﬁlament shown in Fig. 2, the current I is known and
the voltage V can be calculated from E
z
evaluated at the
input position multiplied by the thickness of the substrate
d. We have therefore
B
÷1
=
4dom
0
LW
o
m;n=1
cos
2
(mpx
0
=L) cos
2
(npy
0
=W)
k
2
÷k
2
mn
(59)
B
÷1
=
dom
0
p
o
m=0
o
n=1
(2 ÷d
m0
)k
2
mn
k
2
÷k
2
mn
[J
m
(k
mn
r
0
)]
2
(k
2
mn
R
2
÷m
2
)[J
m
(k
2
mn
R
2
)]
2
(60)
for rectangular and circular patch antennas, respectively.
When excited by a microstrip feeder the input voltage is
obtained by averaging E
z
over the width of the feederline
window joining the cavity multiplied by the thickness of
the substrate d as dictated by Faraday’s law. The input
current can be derived by applying Ampe`re’s law at the
feederline window, and the result is h
0
times the trans
verse length of the window, specifically, h
0
(b ÷a) for t
he rectangular antenna, and h
0
(2aR) for the circular
antenna. We have therefore
B
÷1
=
zd
L
ctnkW÷
o
m=1
8
(mp)
2
L
2
(b ÷a)
2
k
b
m
_
ctnb
m
W sin
mp(b ÷a)
2L
cos
mp(b÷a)
2L
_ _
2
_
(61)
B
÷1
=
zd
2pR
J
0
(kR)
J
/
0
(kR)
÷
o
n=1
2 sin
2
na
n
2
a
2
J
n
(kR)
J
/
n
(kR)
_ _
(62)
for rectangular and circular patch antennas, respectively.
Here k
mn
in Eq. (60) is as given by Eq. (13) and b
m
in
Eq. (61) is as given by Eq. (18).
From Eqs. (59)–(62) we note that when excited at the
normalmode frequencies, the input susceptance B=0,
and hence B
÷1
=N. These points are called antiresonance
points when referred to in a Smith chart plot (e.g., see
Fig. 4 in Section 2.3). In contrast, the resonant points are
deﬁned to be purely resistive so that at resonance the in
put resistance of the antenna is intended to match the
Im(k
p
)
Re(k
p
)
–k
1
–k
0
–p
k
0
k
1
Q
0
Q
1
p
P
W
0
Figure 3. Integration contours for Sommerfeld integrals illus
trating contribution from spatialwave and surfacewave excita
tions.
2586 MICROSTRIP ANTENNAS
feederline impedance, thereby resulting in zero reﬂection.
Therefore, B
÷1
=0 at resonance. The resonance frequency
occurs slightly above the antiresonance frequency, which
requires participation of normal modes of all orders, al
though one normal mode responsible for antiresonance is
excited with the largest amplitude. Thus, by definition, at
resonance the capacitive part of the stored energy equals
the inductive part, rendering the overall input reactance
to be zero. In calculating the susceptance, or the stored
electromagnetic energy of the cavity, we should also in
clude the capacitive contribution from the nearﬁeld exci
tation. However, this contribution has already been
accounted for as the increment in the effective resonance
length of the antenna cavity [Eqs. (2) and (5)]. The input
impedance is
Z=(G÷jB)
÷1
(63)
We note that, in reality, at antiresonance B is ﬁnite be
cause of the presence of losses occurring at the antenna
patch cavity.
One severe drawback of the cavity model is that it is
not able to address the surfacewave loss. James and
Henderson [10] estimated that surfacewave excitation is
not important if d/l
0
o0.09 for e
r
=2.3 and d/l
0
o0.03 for
e
r
=10, where l
0
is the freespace wavelength. The crite
rion given by Wood [11] is more quantitative: d/l
0
o0.07
for e
r
=2.3, and d/l
0
o0.0023 for e
r
=10, if the antenna is
to launch no more than 25% of the total radiated power as
surface waves. Work by Fonseca and Giarola showed that
the size of the patch is also a parameter [12], as will be
discussed in more detail in the next section. When ferrite
material is used as the substrate, magnetic loss can be
estimated using a formula similar to Eq. (51):
P
m
=
om
//
2
__
V
c
_
[H[
2
dv (64)
where m
//
denotes the imaginary part of the permeability.
In concluding this section we note that the cavity model
depicts a semiempirical picture where the parameters
have been adjusted to ﬁt experiments, for example, the
effective dielectric constant e
eff
[Eq. (3)], and the increment
in the resonant length of the patch D [Eqs. (2) and (5)].
Nevertheless, the calculated radiation pattern and input
impedance compared very well with measurements [13].
2. DYADIC GREEN FUNCTION
2.1. Introduction
As we have mentioned in Section 1, the open structure of a
microstrip patch antenna can be rigorously accounted for
only in a fullwave analysis resorting to numerical solu
tions. Maxwell equations can be explicitly solved numer
ically in either the frequency or time domain using the
generic 3D ﬁniteelement and ﬁnitedifference methods
[14]. However, it is more informative to use the dyadic
Green functions, since the electromagnetic ﬁelds generat
ed by a point dipole current source has already been solved
‘‘analytically’’ in the same microstrip geometry, which is
termed the dyadic Green function. The electromagnetic
ﬁelds excited by a patch antenna can then be composed as
superposition of the point dipole solutions in the context of
a conventional Green’s function method. The numerical
technique comes in only when Galerkin method is called
for to solve the resultant integral equations relating the
unknown current variables to the local electric ﬁelds dis
tributed across the metal patch boundaries.
The physical meaning of Green’s function is clear, and
surface waves are treated with equal importance as spa
tial waves in the Green’s function formulation. Further
more, the Green function solution usually requires 2D
calculations, as in contrast to the generic 3D computa
tional methods. This is true when the metal thickness
compares much smaller than the thickness of the sub
strate, as is usually the case. Material losses can be readi
ly included with the Green function if complex
permittivity and permeability values are used. It turns
out that efﬁcient CAD tools can be constructed using the
dyadic Green’s function solutions, which calculate engi
neering parameters of a microstrip antenna, including ra
diation frequency, farﬁeld pattern, efﬁciency, and input
impedance, and analyze the crosstalk problem inherent in
common microstrip circuitries.
To illustrate the physics in the application of a Green
function, we have decided not to regenerate many math
ematical formulas in this section. Instead, we concentrate
on Sommerfeld’s approach to Green function analysis [1],
since it lends itself to more physical understanding of the
problem. We wish to introduce the methodology leading to
the formulation of the dyadic Green function for a general
stratiﬁed structure, consisting of ﬁnite number of dielec
tric and magnetic layers as constituents. We assume this
layered structure is inﬁnite in both horizontal and vertical
directions, although it is possible to include ﬁnite sub
strate and radiation space by employing periodic bound
ary conditions, for example. Also, we assume the current
distribution is twodimensional, resulting in 2D analysis
of the Galerkin elements. The ﬁnite conductivity of the
ground plane can be accounted for by invoking complex
permittivity of the conductor layer [1]. Calculational re
sults are cited mainly from Ref. 15. Background material
can be found in Ref. 16.
2.2. Point Dipole Solutions
We start by introducing the vector potential A, and scalar
potential V, in electrodynamics subject to a Lorentz gauge
[17], which has been implicitly used by Sommerfeld [1]:
H=
1
m
0
VA (65)
E= ÷joA ÷VV (66)
V
.
A÷jom
0
eV =0 (67)
This results in uncoupled equations for A and Vas follows
(V
2
÷k
2
)A= ÷m
0
J (68)
(V
2
÷k
2
)V =
÷1
e
0
r (69)
MICROSTRIP ANTENNAS 2587
where J and r are the current and charge densities, re
spectively, and normally r=0 (oscillating charges can
hardly be realized physically). We note that Eq. (68) does
not specify A uniquely for a ﬁnite volume under consider
ation [17]. However, this gauge freedom is almost ﬁxed for
a system of inﬁnite volume, since the only sourcefree ra
diation for the entire space is the incoming waves from
inﬁnity, which can be readily checked out and excluded
from the solution of A by performing proper gauge trans
formation. From Eq. (67) the scalar potential Vis obtained
from the divergence of A, and hence only the vector po
tential A needs to be solved.
We are now solving the vector potential, A(x; y; z)
[Eq. (68)], induced by a point dipole current source,
J
0
(x; y; z), in the background of a dielectric/magneticlay
ered structure. A point dipole is also called a Hertzian di
pole, which can be approximated by a dipole antenna
whose arms are much smaller than the wavelength of ra
diation [1]. Assume the point dipole to be located at (0, 0,
z
0
). The current density associated with the point dipole is
therefore
J
0
(x; y; z) =I
0
d‘
.
d(x)d(y)d(z ÷z
0
) (70)
where I
0
d‘ ( =ﬁnite) denotes the strength of the dipole.
Without loss of generality, we assume that I
0
is along the x
direction. From Eq. (68) we know that A
x
is nonzero, cor
responding to the radiation ﬁeld directly from the hori
zontal dipole. This is the ﬁeld arising from the point dipole
of Eq. (70) in empty space. However, in the presence of the
layered structure, A
z
is also nonzero, due to oblique re
ﬂection of A
x
from the layer interfaces, corresponding to
the radiation ﬁeld from a vertical dipole in the absence of
the stratiﬁed structure. Finally, A
y
is identically zero as
implied by the symmetry of the problem.
We require the tangential components of E and H to be
continuous across the layer interfaces, and these bound
ary conditions need to be expressed in terms A satisfying
Eqs. (65) and (66). According to Sommerfeld [1], these
boundary conditions can be deduced by integrating Eqs.
(65) and (66) with respect to the transverse coordinates x
and y, and the constants of integration are justiﬁed as ze
ros by letting x and y go to inﬁnity. For example, suppose
that one boundary condition requires the derivative of the
function f(x, y, z) with respect to x to be continuous across
an interface. If f(x, y, z) vanishes as x goes to inﬁnity, either
decreasing exponentially to zero as for a decaying wave or
averaging to zero as for an oscillating wave, we can then
integrate this boundary condition with respect to x to con
clude that the function f(x, y, z) itself must be continuous
across the interface. As such, the boundary conditions im
posed on the vector potential A=A
x
e
x
÷A
z
e
z
as derived by
Sommerfeld are the continuity conditions on the following
four quantities [1],
emA
x
; em
@A
x
@z
_ _
; emA
z
;
@A
x
@x
÷
@A
z
@z
(71)
There are therefore four boundary conditions at each layer
interface.
Recognizing the fact that the layered structure is ho
mogeneous in the transverse directions, say, x and y, it
implies that the Helmholtz equation, Eq. (68), can be con
veniently solved in the transverse Fourier spectral do
main. For a given transverse spectral vector, k
t
=[k
x
; k
y
],
we denote the corresponding spectraldomain vector po
tential components as
~
AA
x
(k
x
; k
y
; z) and
~
AA
z
(k
x
; k
y
; z), which
relate to A
x
(x,y,z) and A
x
(x,y,z) via double Fourier inte
grals, respectively. For each layer, which does not contain
the dipole source, we solve the Helmholtz equation,
Eq. (68), to obtain the following solutions
~
AA
x
(k
x
; k
y
;
z) =a exp( ÷gz) ÷b exp(÷gz) (72)
~
AA
j
(k
x
; k
y;
z) =c exp( ÷gz) ÷d exp(÷gz) (73)
where a,b,c,d are four unknowns to be determined by the
boundary conditions and g is given by
g
2
=k
2
x
÷k
2
y
÷emo
2
(74)
where e and m are respectively the permittivity and per
meability of the layer under consideration. For the upper
most layer a=0=c, and for the lowermost layer b =0=d,
as required by the boundary conditions at z =7N. For the
layer that contains the dipole source, we integrate Eq. (68)
from z =z
÷
0
to z =z
÷
0
and note that the double Fourier
transform of the function d(x)d(y) is (2p)
÷1
; we then derive
the following discontinuity requirement on
@
~
AA
x
(k
x
; k
y
; z)=@z on both sides of the plane z =z
0
:
@
~
AA
x
(k
x
; k
y
; z
÷
)
@z
÷
@
~
AA
x
(k
x
; k
y
; z
÷
)
@z
=
÷I
0
d‘
2p
(75)
We therefore insert a ﬁctitious interface at z =z
0
assuming
different a,b,c,d values for the two subregions above and
below this interface [Eqs. (72) and (73)]; g is the same
across the ﬁctitious interface [Eq. (74)]. Thus, the bound
ary conditions imposed on this ﬁctitious interface are the
same as before except that the requirement on the conti
nuity of the quantity em(@A
x
/@z) is now being replaced by
Eq. (75). If the z
0
plane occurs at one layer interface of the
stratiﬁed structure, no ﬁctitious interface is needed;
we need only to replace the continuity requirement on
em(@A
x
/@z) by Eq. (75) where g differs across the interface.
Thus, we arrive at 4N unknowns with 4N boundary
conditions for N interfaces, including the ﬁctitious one,
if there is one. We replace in the boundary conditions
[Eq. (71)] the operator @/@x by ÷jk
x
, and @/@z by 7g,
whichever is applicable according to Eqs. (72) and (73).
The 4N boundary conditions converts into algebraic equa
tions, and hence the 4N unknowns can be solved. The am
plitudes of these unknowns are all proportional to the
dipole strength I
0
d‘, which can be conveniently chosen to
be 1, as deﬁned by the Green function. The vector poten
tials A
x
(x, y, z) and A
z
(x, y, z) can then be solved from
~
AA
x
(k
x
; k
y
; z) and
~
AA
z
(k
x
; k
y
; z) by applying the inverse two
dimensional Fourier transforms. The electric and magnet
ic ﬁelds can ﬁnally be obtained by using Eqs. (65)–(67).
2588 MICROSTRIP ANTENNAS
Material imperfection results in losses of various kinds
whose effects can be analytically modeled by letting the
permittivity and permeability be complex numbers.
The dielectric loss is described by a loss tangent, tand,
and the permittivity takes the form of e
0
e
r
(1 ÷j tand),
where e
r
denotes the dielectric constant. For a partially
magnetized magnetic substrate the permeability is m
0
(m
/
÷
jm
//
), and m
/
and m
//
are the real and imaginary parts of the
relative permeability [3]. For a metal conductor the per
mittivity contains both the displacement current and the
conduction current, and hence the permittivity is modiﬁed
as e
0
÷js/o where s denotes the conductivity. This is the
permittivity that was explicitly considered by Sommerfeld
[1] solving the electromagnetic ﬁelds generated by a
Hertzian dipole in response to Earth’s surface.
If perfect metal is used as the ground plane, the elec
tromagnetic ﬁeld will not penetrate into it. Therefore, the
boundary condition on the metal surface is that the tan
gential components of the electric ﬁeld vanish, and it im
plies
A
x
=0;
@A
x
@x
÷
@A
z
@z
=0 (76)
Thus, we have 4N÷2 unknowns and 4N÷2 boundary
conditions for a layered structure possessing a perfect
metal ground plane.
Actually, the interface boundary conditions [Eq. (71)]
specify the reﬂection and transmission of electromagnetic
waves from one layer to another, and
~
AA
x
is proportional to
the magnetic ﬁeld component and
~
AA
z
is proportional to the
electric ﬁeld component. Obliqueangle reﬂection and
transmission can be readily expressed by using Snell’s
law [18] and hence the coefﬁcients of
~
AA
x
and
~
AA
z
[Eqs. (72)
and (73)] are determined for each layer with relation to its
preceding and succeeding layers. Thus, circumventing the
need to solve the boundary conditions explicitly [Eq. (71)],
all the a,b,c,d coefﬁcients for the layers are correlated to
each other and only four unknowns remain, corresponding
to those at the outermost layers (top and bottom), which
can then be solved by using the boundary conditions at the
ﬁctitious interface imposed by the point dipole [Eq. (75)].
Suppose the last layer is a perfect metal. For this case the
a,b,c,d coefﬁcients of the layer adjacent to the metal sur
face satisfy the following relationships
b= ÷a; d=c (77)
which are recognized as the total reﬂection condition.
Equation (77) can also be derived from Eq. (76).
For completeness we list below the transverse spectral
domain vector potential induced by a horizontal point di
pole located on top of a dielectric substrate backed by a
perfect metal ground plane:
~
AA
x
(k
x
; k
y
; z) =
m
0
2p
exp
(÷g
0
z)
D
TE
~
AA
z
(k
x
; k
y
; z) =
m
0
2p
(e
r
÷1)jk
x
exp(÷g
0
z)=(D
TE
D
TM
)
for z > 0
(78)
~
AA
x
(k
x
; k
y
; z) =
m
0
2p
sinh
g(z ÷d)
D
TE
sinh gd
~
AA
z
(k
x
; k
y
; z) =
m
0
2p
(e
r
÷1)jk
x
cosh
g(z ÷d)
D
TE
D
TM
cosh gd
for 0z > ÷h
(79)
and D
TE
and D
TM
are given by
D
TE
=g
0
÷g coth gd (80)
D
TM
=e
r
g
0
÷g tanh gd (81)
Here the dipole is located at the plane z =0, the dielectric
substrate is of a thickness d and permittivity e, and g
0
and
g are given by Eq. (74) with the zero subscript (0) referring
to air. The zeros of Eqs. (80) and (81) correspond to sur
facewave TE and TM modes, respectively. While there
exists at least one TM surface mode, it is not always the
case that TE surface modes will be excited. The threshold
for TE mode excitation is
k
0
d(e
r
÷1)
1=2
o
p
2
or
f o
75
d
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
e
r
÷1
_ (82)
where f is expressed in unit of GHz, and d in unit of mm.
Surfacewave modes will be further discussed in the
following section.
2.3. Sommerfeld Integrals
Within each layer the dipole ﬁeld solutions are obtained by
performing inverse Fourier transform of
~
AA
x
(k
x
; k
y
; z) and
~
AA
z
(k
x
; k
y
; z) over the twodimensional k
x
–k
y
plane, or,
equivalently, k
r
–k
f
plane. Here (k
r
,k
f
) denotes the polar
coordinate and (k
x
,k
y
) the Cartesian coordinate. For a lay
ered structure consisting of isotropic materials,
~
AA
x
and
~
AA
z
can be expressed as the following functionals
~
AA
x
(k
x
; k
y
; z) =f (k
r
; z) (83)
and
~
AA
z
(k
x
; k
y
; z) =k
x
g(k
r
; z) = cos k
f
k
r
g(k
r
; z) (84)
as can be checked from the boundary condition of Eq. (71).
In other words,
~
AA
x
is an even function of k
x
and
~
AA
z
is an
odd function of k
x
, as implied by the symmetry of the
problem. Therefore, after integration over k
f
, we obtain
A
x
(r; f; z) =
_
o
0
J
0
(k
r
r)k
r
f (k
r
; z) dk
r
(85)
A
z
(r; f; z) =j cos f
_
o
0
J
1
(k
r
r)k
2
r
g(k
r
; z) dk
r
(86)
Although Eqs. (85) and (86) can be numerically integrated,
as will be discussed below, it is informative to discuss the
physical meanings implied by these integrals. Equations
MICROSTRIP ANTENNAS 2589
(85) and (86) are called Sommerfeldtype integrals whose
significance was ﬁrst addressed by Sommerfeld [1]. So
mmerfeld noticed that the seemly real integration starting
at the ﬁxed point k
r
=0 can be converted into a complex k
r
integration over a path W, which closes at inﬁnity (see
Fig. 3)
A
x
(r; f; z) =
1
2
_
W
H
(2)
0
(k
r
r)k
r
f (k
r
; z) dk
r
(87)
A
z
(r; f; z) =
j cos f
2
_
W
H
(2)
1
(k
r
r)k
2
r
g(k
r
; z) dk
r
(88)
where H
(2)
0
and H
(2)
1
are Hankel functions of the second
kind of orders 0 and 1, respectively. In Fig. 3, the contour
W is detoured slightly in the complex k
r
plane not to run
into branchcuts and poles such that the integrands re
main ﬁnite and singlevalued. In Fig. 3 we illustrate the
case that the layered structure involves only two layer
substances, the air, denoted by subscript 0 (zero), and a
dielectric layer, denoted by subscript 1 (one). For example,
the respective permittivities for air and the dielectric layer
are denoted as e
0
and e
1
(the permeability of the dielectric
layer is assumed to be the same as that of air, m
0
). In Fig. 3,
k
0
and k
1
are the branch points given by
k
0
=(e
0
m
0
)
1=2
; k
1
=(e
1
m
0
)
1=2
(89)
and p is the smallest zero assumed by the function D
TM
[Eq. (81)]. It is understood that more branchpoints, and
hence more branchcuts, will appear in Fig. 3 if more layers
are presented in the layered structure. Also, more simple
poles will appear if D
TE
and D
TM
of Eqs. (80) and (81) ad
mit more zeros. In Fig. 3 the branchpoint at k
r
=0 results
from the Hankel functions H
(2)
0
[Eq. (87)]. However, for Eq.
(88) this branchpoint, and hence its associated branchcut,
is replaced by a simple pole due to the different singular
behavior of H
(2)
1
at the origin.
According to Sommerfeld, the contour W shown in
Fig. 3 can be analytically deformed into three components
or more surrounding the respective branchcuts and simple
pole, denoted as Q
0
, Q
1
, and P. Sommerfeld showed that
[1] contour integrals of Q
0
and Q
1
give rise to spatialwave
radiation and for a large (spherical) distance r, they ex
hibit the following asymptotic dependence:
Q
0
(Q
1
) o
exp(÷jk
0
r)
r
(90)
Actually, Q
0
gives rise to the freespace dipole radiation,
and Q
1
is the radiation wave diffracted by the dielectric
layer (or from the image dipole induced by that layer) [1].
Assuming that the air–dielectric interface is located at
z =0, then, for z40, contribution form contour P has the
following asymptotic expressions:
P oH
(2)
0
(pr) exp[÷
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
p
2
÷k
2
0
_
z]; for A
x
; [Eq: (86)]
(91)
P ocos fH
(2)
1
(pr) exp[÷
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
p
2
÷k
2
0
_
z]; for A
z
; [Eq: (87)]
(92)
These are surface waves tied to the interface and decrease
in a rate proportional to 1=
ﬃﬃﬃ
r
_
in the lateral directions, in
contrast to the spatialwave radiation, which exhibits a 1/r
dependence [Eq. (90)].
The significance of Fig. 3 is that through Sommerfeld’s
interpretations [1], spatial and surface waves have been
treated on an equal basis that for a given point dipole
source in a layered structure the farﬁeld solutions can be
expressed explicitly by using Eq. (90) and Eqs. (91), (92),
respectively, to show radiations either propagating away
semispherically in free space, or guiding away concen
trically on the surface of the layered structure. By using
Sommerfeld’s formulation, it is now possible to include a
surfacewave analysis with the resonator cavity model
discussed in Section 1, and the farﬁeld surface wave
radiation pattern and the total surface wave loss from
the cavity can be calculated by using the perturbation
theory in a manner similar to that for the spatial radiation
waves. Specifically, when the zerothorder solution is
known for an isolated cavity, the farﬁeld solution for sur
face waves can be expressed as the ﬁrstorder perturba
tion by using Eq. (40) with the spatial dependence
exp( ÷jk
0
r)/r for A
z
by H
(2)
0
(pr) exp[÷
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
p
2
÷k
2
0
_
z], and for
A
x
by cos fH
(2)
1
(pr) exp[÷
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
p
2
÷k
2
0
_
z] [see Eqs. (90)–(92)].
However, research in this direction is lacking in the cur
rent literature. Having discussed the analytic features of
Sommerfeldtype integrals, in the following we discuss
how Sommerfeldtype integrals are evaluated numerically.
In the presence of material losses, the branchpoints
and simple pole(s) associated with spatial and surface
waves, respectively, acquire imaginary components, which
are then pushed off from the real axis toward the lower
half of the complex k
r
plane. As such, numerical integra
tion of Eqs. (65) and (66) can be properly carried out.
However, care needs to be taken in order to avoid large
truncation errors. Integration of k
r
along the real positive
axis can be distinguished in three regions
Region1 : 0ok
r
ok
0
Region2 : k
0
ok
r
ok
c
Region3 : k
c
ok
r
oo
where k
c
denotes a cutoff wavenumber to be discussed be
low. In region 1 the integrands are well behaved. However,
at the resonant frequencies of a metal patch dictated by
the cavity model both the numerator and denominator of
the integrand vanish, although their ratio remains ﬁnite.
We call these geometric resonant points quasisingularities
[15]. Near these quasisingular points the numerator and
denominator need to be expanded in Taylor series on
which their common zeros cancel out.
All the surface poles are contained in region 2, and con
ventionally, the upper bound of region 2, k
c
, is deﬁned to be
10 times the real part of the largest surface pole occurring
at the integrand. In region 2 the integrand behaves wildly
in the vicinity of a surface pole. When coming across a
surface pole, the integrand transits from positive inﬁnity
to negative inﬁnity, resulting in sharp cancellation dur
ing numerical integration. To circumvent this difﬁculty,
2590 MICROSTRIP ANTENNAS
we expand the integrand in Laurent series in the
vicinity of a surface pole k
r
=p, and the quasisingular
terms (with negative exponents) are then evaluated ana
lytically (remember p is now a complex number). Equiva
lently, the singular part of integration is obtained via
residual calculations. After subtracting the singular part
from the integrand, the integrand becomes regular, which
can then be numerically integrated in region 2.
In region 3 we are involved with integration of Bessel
functions at inﬁnity, which oscillate indefinitely as inﬁnity
is approached without exhibiting a strict period. This ren
ders the conventional extrapolation scheme inaccurate. To
perform integration in this region we consider asymptotic
expansion of Bessel functions:
J
m
(x) =
ﬃﬃﬃﬃﬃﬃ
2
px
_
cos x ÷
mp
2
÷
p
4
_ _ _
o
k =0
(÷1)
k
2
2k
G m÷2k÷
1
2
_ _
(2k)!G m÷2k÷
1
2
_ _ x
2k
÷sin x ÷
mp
2
÷
p
4
_ _
o
k=0
(÷1)
k
2
2k ÷1
G m÷2k÷
3
2
_ _
(2k÷1)!G m÷2k ÷
1
2
_ _ x
2k ÷1
_
¸
¸
_
(93)
where the G terms denote gamma functions. As such, inte
grands are written in series containing terms of the form
_
o
kc
A
p
sin(ax ÷b)
x
p
dx (94)
which can be readily evaluated by exploiting sine and
cosine integrals and their derivatives if p is a positive
integer, or error functions and their derivatives if p is a
positive halfinteger larger than 1 [15].
Having discussed the Green functions due to a point
dipole source in a layered structure, we mention how these
Green functions are modiﬁed in the presence of material
losses. For a lossy material one may use complex permit
tivity e
0
e
r
(1÷j tand) and complex permeability m
0
(m
/
÷jm
//
)
to describe its lossy characteristics, where e
r
denotes the
dielectric constant, tan d represents the loss tangent, and
m
/
and m
//
are the real and imaginary parts of the relative
permeability [3]. So far we have considered perfect metal
boundary conditions on bounding a layered structure at
the outermost surface [Eq. (77)]. For imperfect metal
boundaries, for example, our Earth’s surface, one may in
clude the layer explicitly in the formulation, but using a
complex permittivity e
0
÷js/o, where s denotes the con
ductivity [1]. Alternatively, one can apply Snell law to
specify the relationships between the a,b,c,d, coefﬁcients
resulting from reﬂections of incoming waves from the sur
face of an imperfect metal, in a manner similar to that de
riving Eq. (77). Or, if the conductivity is large, say, a good
metal such as copper, and the skin depth in penetration
compares much smaller than the wavelength, one can even
apply the perturbation theory. Thus, one ﬁrst derives the
Green’s functions of the layered structure assuming
perfect metal boundaries followed by ﬁrstorder modiﬁca
tion that the metal boundary is withdrawn a distance d
into the interior of the imperfect metal bulk with the re
cessed volume to be ﬁlled by air [19]. Here d
c
denotes the
(complex) skin depth
d
c
=(1 ÷j)
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
2=om
0
s
_
(95)
This procedure is parallel to Wheeler’s incremental imped
ance [20,21], since it is recognized that the definition of a
Green’s function is a transimpedance. To be explicit, in the
presence of a good conductor ground plane Eq. (80) is mod
iﬁed to include its ﬁrstorder Tayler expansion in d, or
Wheeler’s incremental impedance, and the result is [19]
D
TE
=g
0
÷g coth gd÷d
c
(1 ÷tanh gd)=(1 ÷ tanh gd) (96)
and D
TM
is invariant under the ﬁrstorder consideration.
The construction of Green functions for a layered structure
containing gyromagnetic materials will be discussed
in Section 4.
2.4. Numerical Solutions
The dyadic Green function, G(r; r
/
), is deﬁned as the elec
tric ﬁeld at location r produced by a unit point dipole lo
cated at r
/
. We are solving the current distribution over a
microstrip metal patch of negligible thickness deposited
on a layered structure backed by a ground plane. There
fore, r and r
/
are both located at the same metal patch (or
located at different but mutually coupled metal patches),
and essentially we are solving a twodimensional problem
with the third dimension, the z direction, being absorbed
into the Green function. This results in the following
integral equation
t
e
_
S
G(q; q
/
)
.
J(q
/
) d
2
q
/
÷R
s
J(r)
_ _
=E
e
(q) (97)
where S denotes the metal patch with surface impedance
R
s
=(1÷j)
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
om
0
=2s
_
. The ﬁrst term in Eq. (97) is the elec
tric ﬁeld induced by the background layered structure,
and the second term relates to ohm’s current (induction
and conduction), and E
e
is the electric ﬁeld generated by
external currents. Of special concern, in Eq. (97) t
e
results
from integration along the z axis, and hence it deﬁnes the
effective thickness of the metal patch, to be either the skin
depth thickness, d =(2/mo
0
s)
1/2
, or the physical thickness
of the metal patch t, whichever is smaller. For a perfect
metal patch t
e
=0. Thus, t
e
represents the singular be
havior of the current distribution exhibiting a deltafunc
tionlike proﬁle in the thickness direction. However, we
expect that t
e
will not enter the ﬁnal expressions when
evaluating engineering parameters of the antenna, as is
the case shown in Eqs. (98), (100), (101), and (105) below.
In Eq. (97) we have used q and q
/
as the twodimensional
position vectors on the metal patch for which the z depen
dence is understood.
We denote {J
mn
(q)] as a set of a complete orthonormal
vector basis for currents on the metal patch. We deﬁne
here {J
mn
(q)] to be the regular part of the surface current
density and the singular part has been factored out in
MICROSTRIP ANTENNAS 2591
Eq. (99) as t
e
÷1
, characterizing, again, the deltafunction
like distribution of surface current along the z direction.
Thus, the dimension of J
mn
is amperes/meter. We apply
the Galerkin’s method to convert the integral equation
[Eq. (97)], into the following matrix form
m
/
;n
/
[B
mnm
/
n
/ ][a
m
/
n
/ ] =[b
mn
]; (98)
where [a
mn
] are the unknown coefﬁcients to be solved
expressed in terms of the current basis {J
mn
(q)]:
J(q) =t
÷1
e
mn
a
mn
J
mn
(q) (99)
[B
mnm
/
n
/ ] are the matrix element derived from Eq. (97)
B
mnm
/
n
/ =
_
S
d
2
q
_
S
d
2
q
/
J
+
mn
(q)
.
[G(q; q
/
) ÷R
s
I]
.
J
m
/
n
/ (q
/
)
(100)
where I denotes the identity dyad. The inhomogeneous
term of Eq. (98), [b
mn
], is associated with current driving
given by
b
mn
=
_
S
d
2
qJ
+
mn
(q)
.
E
e
(r) =t
÷1
e
_
V
d
3
rE
mn
(r)
.
J
e
(r)
+
=
_
V
d
3
r
_
S
d
2
q
/
J
mn
(q
/
)
.
G(r; q
/
)
.
J
e
(r)
+
(101)
where E
mn
is the electric ﬁeld generated by J
mn
, J
e
is the
driving current density, and integration is over the whole
volume, V. In Eq. (101) the reciprocity theorem has been
used, which states that the response of a system to a
source is unchanged if source and observer are inter
changed [22]. The unit drive current density may be spe
ciﬁed as
J
e
(q; z) =e
z
d(x ÷x
0
)d(y ÷y
0
)[S(z) ÷S(z ÷d)]=2
for rectangular patch
=e
z
d(r ÷r
0
)d(f ÷f
0
)[S(z) ÷S(z ÷d)]=(2r)
for circular patch
(102)
J
e
(q; z) =e
z
[S(x ÷a) ÷S(x ÷b)]d(y)[S(z)
÷S(z ÷d)]=[4(a÷b)] for rectangular patch
=e
z
d(r ÷R)[S(f÷a) ÷S(f ÷a)][S(z)
÷S(z ÷d)]=(8ar) for circular patch (103)
These last two equations denote coaxial and microstrip
feeders, respectively. We note that the current densities
given in Eq. (103) are Huygens currents resulting fromthe
external current fed at the microstrip input port [Eq. (38)].
By using Huygens principle the two feeder conﬁgurations,
coax line and microstrip line, are cast in similar expres
sions [Eqs. (102) and (103)] [23]. In Eqs. (102) and (103)
S(x) denotes the step function and
S(x) =0 if xo0
=0:5 if x =0
=1 if x > 0
(104)
The input impedance is therefore, assuming unit excita
tion current
Z
in
= ÷t
e
_
S
d
2
qE
+
e
(q)
.
J(q)
= ÷
m;n;m
/
;n
/
[b
mn
]
+
[B
mnm
/
n
/ ]
÷1
[b
m
/
n
/ ]
(105)
The radiation ﬁeld associated with a point (Hertzian)
dipole located at the air–dielectric interface of a microstrip
structure is [16]
E
y
= ÷j
Z
0
l
0
cos fg
y
(y)
e
ik
0
r
r
;
E
f
=j
Z
0
l
0
sin fg
f
(y)
e
ik0r
r
(106)
where
g
y
(y) =
T cos y
T ÷j c
r
cos y cot(Tk
0
d)
(107)
g
f
(y) =
cos y
cos y ÷jT cot(Tk
0
d)
(108)
and
T=
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
e
r
÷sin
2
y
_
(109)
The radiation pattern for a given patch current distribu
tion J(q) [Eq. (106)] is therefore
E
rad
(r)
= ÷j
z
0
l
0
e
÷jk
0
r
r
_
d
2
qe
jk
.
r
J(q)
_ _
.
[g
y
(y)e
f
e
z
e
y
÷g
f
(y)e
f
e
f
]
(110)
where z
0
=(m
0
/e
0
)
1/2
and k
0
=2p/l
0
=o(e
0
m
0
)
1/2
are, resp
ectively, impedance and wavenumber in air.
Before performing numerical calculations, we must
ﬁrst determine which current basis is best suited for a
given microstrip patch geometry implementing numerical
calculations. For example, for a given microstrip of width
expanded by interval [ ÷1,1] one may use the triangular
functions, Chebyshev polynomials, Legendre polynomials,
and so on, as the functional basis to expand currents, but
which one would provide the most advantages? The an
swer relies on the physical restriction imposed on current
ﬂow on a microstrip patch; the current basis needs to
show zero (longitudinal) components when crossing the
boundary of the patch; that is, the currents are not allow
2592 MICROSTRIP ANTENNAS
to ﬂow over metal boundaries. Under this restriction the
freedom in choosing a current basis is largely ﬁxed, which
leads to the functional set of current potentials [24].
The simplest approach is to deﬁne a set of current po
tential {c
mn
(q)] satisfying the 2D Helmholtz equation
(V
2
t
÷k
2
mn
)c
mn
(q) =0 (111)
from which the current basis {J
mn
(q)] is derived as
J
mn
(q) =V
t
c
mn
(q) (112)
The boundary condition imposed on {J
mn
(q)] is that the
current is not allowed to ﬂow across the metal patch
boundary, or, at the patch boundary
n
.
J
mn
=0 (113)
where n denotes the unit outward normal at the metal
patch boundary:
n
.
V
t
c
mn
=0 at patchboundary (114)
This equation determines the eigenvalues k
mn
denoted in
Eq. (111), and the current basis speciﬁed by Eq. (112) form
a complete basis for vector ﬁelds on the metal patch.
Other advantages follow by deriving the current basis
from current potentials:
1. The current potentials are intrinsic functions of the
patch, so that the modal numbers {n.m} reﬂect its
symmetry. For example, for a rectangular microstrip
antenna loaded with isotropic material the symme
try of the patch dictates currents with opposite par
ities do not couple. For a circular microstrip antenna
loaded with isotropic materials currents with differ
ent angular modal numbers (monopole, dipole,
quadrapole, etc.) do not couple. This results in large
ly reduced matrix size in Galerkin elements de
manding numerical evaluation, thereby saving
calculational efforts considerably.
2. Due to Green’s theorem, in calculating a Galerkin
element, most integrations required by (2D) Fourier
and (2D) inverse Fourier transforms, for example
[Eqs. (85) and (86)], can be performed analytically,
leaving behind only onefold, or at most twofold, in
tegrals that ﬁnally resort to numerical methods.
This also saves calculational efforts.
3. By using current potentials Galerkin elements ex
pressed by Eq. (100) reduce to scalar function mul
tiplications rather than the originally proposed
inner products between tensor and vector ﬁelds,
thereby simplifying calculations.
On the basis of current potentials, we have solved
the coupling problems between circular microstrip
patch antennas, as introduced as follows.
For a rectangular patch located at 0rxrL, 0ryrW,
the current potentials are
c
mn
(x; y) =
4
LW
cos
mpx
L
cos
npy
W
; m; n=1; 2; 3; . . . (115)
and for a circular patch located at 0 _ q _ R, they are
c
mn
(r; f) =
ﬃﬃﬃ
p
_
b
mn
R 1 ÷
m
2
b
2
mn
R
2
_ _
1=2
J
m
(b
mn
R)
_
_
_
_
÷1
J
m
(b
mn
r) exp(jmf)
(116)
and b
mn
denotes the root of the derivative of Bessel func
tion, J
/
m
(x):
J
/
m
(b
mn
R) =0 (117)
In the following we present some calculational results
for circular microstrip patch antennas [15]. For rectangu
lar patch antennas, see Ref. 25. The ﬁrst calculation ap
plies to the published data of a circular antenna
characterized by the following parameters [13]: R=
6.75 cm, d=0.1588cm, e
r
=2.62, and tand =0.001. The
1.0
2.0
–2.0
–1.0
–0.5
–0.25
0.0
0.25
0.5
0.76 GHz
0.794 GHz
0.84 GHz
0
.
2
5
0
.
5
1
.
0
2
.
0
Figure 4. Input impedance loci of the antenna observed by Lo
et al. The solid line is fromcalculation, and small circles represent
measurements from Lo et al. [13].
–90°
0 –20 –40 –20 0 (dB)
–60°
–30°
90°
60°
30°
0°
[ = 0°
Figure 5. Radiation proﬁle of the antenna observed by Lo et al.
[13] in the f=01 plane.
MICROSTRIP ANTENNAS 2593
calculated resonance frequency of the fundamental mode
is 0.7936GHz [15], which compares exactly to the mea
sured value of 0.794 GHz [13]. This is contrasted to the
cavity model calculation presented in Section 1, which
predicts a resonant frequency of 0.805 GHz [13]. The cal
culated input impedance of the antenna is shown in
Ref. 15, which compares nicely with measurements shown
as small circles in Fig. 4 [13]. Figures 5 and 6 show the
calculated radiation pattern of the antenna in the f=01
and f=901 planes, respectively. We note that only the co
polarized radiations are generated from the fundamental
mode excitation; the crosspolarized ﬁeld cancels out for
the two m=1 and m= ÷1 modes at the fundamental res
onant frequency.
The second patch antenna geometry considered is a
microstrip disk of radius R=1 cm, which is fed by either a
coaxial or a microstrip line [15]. The substrate has dielec
tric constant, e
r
=2.2, loss tangent tand =0.001, whose
thickness d is subject to vary. Figures 7–9, assuming
a =0.2 rad, show the calculated resonant frequency, input
impedance, and radiation linewidth of the fundamental
mode as a function of the substrate thickness, respectively.
In Fig. 7 the calculated resonant frequency of the antenna
decreases monotonically with the substrate thickness d,
indicating the effective dimension of the patch resonator
increases with d, as expected for a leaky cavity. Figure 8
shows that the input impedance of the antenna is rela
tively a constant, unless d becomes very small, say, small
er than 0.05 cm. In Fig. 9 we see that the radiation
bandwidth increases with d, and hence the Q factor of
the antenna resonator decreases with d. This ﬁnding is
consistent with that revealed from Fig. 7 exhibiting the
leaky feature of the antenna patch cavity.
Figure 10 shows the calculated input impedance loci of
the above circular antenna of substrate thickness d=
0.1 cm fed by a coaxial line located at (r
0
, 0). In Fig. 10
the parameter r
e
is deﬁned as r
e
=r
0
/R, and r
e
=1.0, 0.8,
0.6, and 0.416. For each feeder location the two resonant
0 0 (dB) –20
–90°
–60°
–30°
–40 –20
90°
60°
30°
0°
[ = 90°
Figure 6. Radiation proﬁle of the antenna observed by Lo et al.
[13] in the f =901 plane.
0 0.1 0.2 0.3 0.4
Substrate Thickness d (cm)
5
5.2
5.4
5.6
5.8
6
C
e
n
t
r
a
l
R
a
d
i
a
t
i
o
n
F
r
e
q
u
e
n
c
y
(
G
H
z
)
Figure 7. Calculated radiation frequency as a function of sub
strate thickness (fed by microstrip line).
0 0.1 0.2 0.3
200
300
400
500
Substrate Thickness d (cm)
R
a
d
i
a
t
i
o
n
R
e
s
i
s
t
a
n
c
e
(
Ω
)
Figure 8. Calculated radiation resistance as a function of sub
strate thickness (fed by microstrip line).
0
0
0.1
0.2
0.3
0.4
0.1 0.2 0.3
Substrate Thickness d (cm)
R
a
d
i
a
t
i
o
n
L
i
n
e
w
i
d
t
h
(
G
H
z
)
Figure 9. Calculated radiation linewidth as a function of sub
strate thickness (fed by microstrip line).
2594 MICROSTRIP ANTENNAS
frequencies shown in Fig. 10 are the resonance frequen
cies of the probe inductance in parallel with the detuned
patch resonator forming a parallelresonant circuit, and
the resonance frequency of the probe inductance in series
with the detuned patch resonator forming a seriesreso
nant circuit. The ﬁeld distribution of the patch at both
frequencies shows very little difference, and the patch is
operated at the same resonance mode, yet more or less
detuned from its resonance frequency, which occurs be
tween these two frequencies. It is seen in Fig. 10 that
these resonant frequencies appear only if 0.416rr
e
r1.
For r
e
r0.416 no patch resonance can be possibly excited.
Figure 11 shows the calculated radiation frequency as a
function of r
e
, or r
0
. It is seen in Fig. 11 that radiation
frequency remains roughly constant for 0.6rr
e
r1, which
increases rapidly when r
e
is further reduced. Figure 12
shows the calculated input resistance as a function of r
e
, or
r
0
. From Fig. 12 we see that the input impedance decreas
es with r
e
, or r
0
, and hence it is possible to design 50 O
input resistance of the antenna by feeding the antenna at
r
e
=0.442 (or r
e
=0.488 for 75 O input impedance).
Finally, we consider the interaction between two iden
tical circular microstrip patch antennas excited by micro
strip feeders at equal amplitude and phase. Let the
antennas be deployed in parallel exhibiting the same pa
rameters considered before: R=1cm, d=0.1 cm, e
r
=2.2,
and the feeder lines possess 50 O resistance (width
0.312 cm). The separation between the antennas is desig
nated as R
12
measured between their respective centers.
Figure 13 shows the calculated and measured input im
pedance for the coupled case, R
12
=2 cm (the patches are
touching each other), and the uncoupled case, R
12
=5 cm
(the patches are separated by 3cm at their edges). Mea
surements were performed with respect to patch antennas
fabricated using RT/duroid 5880 material (Rogers Co.,
Chandler, AZ). The measured resonant frequencies were
5.514 and 5.561 GHz for the coupled and uncoupled cases,
respectively, which compare almost exactly to their calcu
lated values of 5.5137 and 5.5642GHz. The resonant fre
quencies of the patches and their input impedances have
also been calculated as a function of the patch separation,
R
12
. These are shown in Figs. 14 and 15, respectively, with
measurements shown as solid squares. From Fig. 14 it is
seen that the resonant frequency changes most rapidly
when R
12
is small, say, when 2rR
12
r3cm. Further in
crease in R
12
does not significantly change the resonance
frequency. However, the input impedance does show long
range interference between two patch antennas. As shown
–1.0
–0.5
–0.25
0.0
0.25
0.5
1.0
2.0
–2.0
0
.
2
5
0
.
5
1
.
0
2
.
0
r
e
= 1.0 r
e
= 0.8 r
e
= 0.6 r
e
= 0.416
Start Frequency: 5.2 GHz Stop Frequency: 6.0 GHz
Figure 10. Input impedance loci of the antenna for several co
axial feeder positions.
0.4 0.6 0.8 1
5.56
5.58
5.60
5.62
Coax Feed Position (cm)
C
e
n
t
r
a
l
R
a
d
i
a
t
i
o
n
F
r
e
q
u
e
n
c
y
(
G
H
z
)
Figure 11. Calculated radiation frequency as a function of coax
ial feeder position.
0.4 0.6 0.8 1
Coax Feed Position (cm)
0
50
100
150
200
250
R
a
d
i
a
t
i
o
n
R
e
s
i
s
t
a
n
c
e
(
Ω
)
Figure 12. Calculated radiation resistance as a function of
coaxial feeder position.
MICROSTRIP ANTENNAS 2595
in Fig. 15, the input impedance is still increasing when the
two antennas are separated by 5 cm, although the rate of
increase has slowed when compared with its initial rate at
R
12
=2 cm.
We conclude that the leaky feature of an antenna cavity
can be well characterized by using a fullwave analysis
outlined in Section 2. Among many numerical methods,
the dyadic Green function approach might prove to be the
simplest one to apply, not only because the analysis is two
dimensional but also because the evaluation of the scalar
Galerkin elements involves only onefold Sommerfeld
type integrals. Surfacewave generation is significant for
a thick dielectric substrate with large dielectric constant.
The coupling between microstrip elements is of longrange
nature, although the radiation frequency of a patch is less
inﬂuenced by its neighboring antenna elements.
3. ANISOTROPIC/GYROMAGNETIC LAYER
In Sections 1 and 2 we considered microstrip patch an
tennas fabricated on isotropic materials. Anisotropic sub
strates, due to either the process in material preparation
or crystalline asymmetry, have been included in the fab
rication of microwave integrated circuits (MICs). Also,
electromagnetic wave propagation in a fully saturated fer
rite substrate is anisotropic, called the gyromagnetic effect.
For the former case the dielectric properties of the sub
strate can be described in terms of a permittivity tensor of
rank 2, and for the latter case Polder permeability tensor
results under the smallsignal approximations [26]. While
many authors have applied numerical calculations to mi
crowave circuitries containing anisotropic substrates
[27–31], in this section we consider wave propagation in
a gyromagnetic substrate [32]. Specifically, we consider
the electromagnetic wave propagation involving the mi
crostrip geometry where the metal circuit is fabricated on
top of a ferrite substrate for which the bias magnetic ﬁeld
is applied along an arbitrary direction that is not neces
sarily the symmetry axis. In order to minimize magnetic
loss, the ferrite substrate needs to be fully saturated so
that an external DC magnetic ﬁeld is applied aligning
magnetic domains achieving the singledomain conﬁgura
tion, thereby eliminating domain wall motion to be other
wise induced accompanying electromagnetic wave
propagation [33]. This external magnetic bias ﬁeld can
be supplied by material anisotropy as occurring to a hexa
ferrite substrate considered in Ref. 32.
In contrast to the permittivity tensor, Polder tensor
elements are highly frequencydependent and exhibit the
resonance response in frequency and in the bias magnetic
2 2.5 3 3.5 4 5 4.5
5.50
5.52
5.54
5.56
5.58
5.60
Patch Separation R
12
(cm)
R
e
s
o
n
a
n
t
F
r
e
q
u
e
n
c
y
(
G
H
z
)
Figure 14. Calculated and measured resonant frequencies of two
coupled patch antennas as a function of their separation distance.
2.5 3 4 5 4.5 3.5
Patch Separation R
12
(cm)
50
100
150
250
I
m
p
e
d
a
n
c
e
(
Ω
)
2
200
Figure 15. Calculated and measured input impedances at reso
nant of two coupled patch antennas as a function of their sepa
ration distance.
Uncoupled case (R
12
= 5 cm)
Coupled case (R
12
= 2 cm)
–1.0
–2.0
–0.5
–0.25
0.0
0.25
0.5
1.0
2.0
0
.
2
5
0
.
5
1
.
0
2
.
0
5.514 GHz
5.561 GHz
Start Frequency: 5.2 GHz Stop Frequency: 6.0 GHz
Figure 13. Calculated and measured input impedance loci of two
coupled patch antennas.
2596 MICROSTRIP ANTENNAS
ﬁeld [33]. For example, when ferrimagnetic resonance
(FMR) is encountered, the effective permeability changes
from a large positive number to a large negative number
together with a nonzero imaginary part accounting for
magnetic loss [26]. Magnetic loss at resonance is para
metrized as FMR linewidth whose occurrence arises from
magnetic relaxation processes accompanying the preces
sional motion of spins at resonance [33]. Most useful mag
netic microwave devices are designed to operate near
FMR so that the rapid change in magnetic permeability
is exploited, either to obtain frequency tuning capability
or to remove the degeneracy in electromagnetic modes,
thereby inducing nonreciprocity in wave propagation [33].
For example, ferrite phase shifters [34], resonators [35],
and ﬁlters [26] are operational according to the ﬁrst prin
ciple [33] and circulators and isolators, according to the
second [36].
Before solving the AC problem on wave propagation
involving a ferrite substrate, one is required to ﬁrst solve
the DC problem to calculate the demagnetizing ﬁeld aris
ing from the ﬁnite dimensions of the ferrite geometry [33].
In a cubic ferrite sample material anisotropy is seldom
important, since it is small in comparison to the external
bias ﬁeld. In a hexaferrite substrate the internal anisot
ropy ﬁeld can be as large as 50 kOe, and hence it can no
longer be neglected [37]. Actually, hexaferrites are pur
posely introduced with a microwave circuit to alleviate, or
even eliminate [38], the external biasﬁeld requirement at
high frequencies. In a hexaferrite material magnetic an
isotropy appears in the form of an easy axis or an easy
plane. For an Mtype hexaferrite the c axis is an easy axis,
and the magnetization vector prefers to be aligned along
the c axis so as to lower its free energy [33]. For a Ytype
hexaferrite the ab plane is an easy plane, and the mag
netization vector is energetically favorable to stay on the
ab plane [38]. The total effective ﬁeld inside a ferrite sub
strate can be calculated from the system Lagrangian, in
cluding magnetic anisotropy, FMR linewidth, exchange
coupling, demagnetization, and magnetostriction [33].
In Section 2 we have considered the Green functions of
a layered structure containing isotropic materials that
were constructed using the vector potentials as introduced
by Sommerfeld [1]. In this section we consider the Green
function construction by using the planewave solutions
for which the substrate/superstrate may contain anisotro
pic layers. While planewave solutions are trivial within
an isotropic dielectric layer, planewave solutions for an
unbounded ferrite bulk magnetized along an arbitrary di
rection are also available [26], whose properties can be
found in Ref. 32. For a given frequency there are four in
dependent electromagnetic modes in a substrate material
that are all degenerate if the substrate shows isotropy for
wave propagation. However, for an anisotropy substrate,
say, a ferrite fully biased by a magnetic ﬁeld (or, a gyro
magnetic material), wave propagation for these four
modes becomes nondegenerate, assuming different effec
tive permeability values for different modes, resulting in
different propagation speeds and polarizations. Similar to
the isotropic case, the directions of electric ﬁeld, magnetic
induction, and wave propagation for each mode excited in
an anisotropic substrate are mutually perpendicular to
each other, as dictated by Maxwell equations. For a gyro
magnetic substrate, the magnetic ﬁeld is no longer aligned
with magnetic induction, although it is still required to be
perpendicular to the direction of the electric ﬁeld [32].
One may apply the transfermatrix technique to con
struct the Green functions of a layered structure. A trans
fer matrix is formulated in the transverse spectrum
domain, which translates the surface impedance, which
is itself a 22 matrix, from one layer interface to another,
assuming the tangential components of the electromag
netic ﬁelds to be continuous across the interfaces in the
absence of circuit (conductor) inhomogeneities. The trans
fer matrices for a given layer can be constructed in terms
of the four electromagnetic modes, or the planewave so
lutions in the spectral domain, in a manner similar to the
a,b,c,d coefﬁcients introduced by Eqs. (72) and (73) (except
that the g values differ for the four modes in an anisotropic
layer). The surface impedance matrix is deﬁned as the ra
tio of the tangential electric ﬁeld to the tangential mag
netic ﬁeld at a given layer interface, and hence it is a 22
matrix. The outermost layers are either air or a metal
surface of ﬁnite conductivity, giving rise to the (imperfect)
opencircuited or shortcircuited boundaries for the lay
ered structure, respectively. The surface impedance ma
trices for the outmost two layers can thus be obtained, in a
manner similar to that used to obtain Eq. (77). Via the
transfer matrices deﬁned for the layers, these two outer
most surface impedance matrices are translated sequen
tially onto an interface containing a point source
responsible for the Green function definition (a Green
function is deﬁned to be the response excited by the point
source). After applying Ampe`re law the Green functions
evaluated at this interface position can thus be obtained
[32]. To be explicit, assuming the point source is located at
the origin, the Green functions, or the Green dyads, are
deﬁned by
e
t
(x; y)
=
1
4p
2
_
o
÷o
dk
x
_
o
÷o
dk
y
_
o
÷o
dx
/
_
o
÷o
dy
/
exp[ik
x
(x ÷x
/
)]
exp[ik
y
(y ÷y
/
)]G(k
x
; k
y
)j(x
/
; y
/
)
(118)
and after applying Ampe`re law at the interface z =0 the
Green functions are obtained as
G(k
x
; k
y
) ={[Z(0
÷
; k
x
; k
y
)]
÷1
÷[Z(0
÷
; k
x
; k
y
)]
÷1
]
÷1
0 1
÷1 0
_ _
(119)
Here j(x
/
; y
/
) denotes current density, e
t
(x; y) tangential
electric ﬁeld, and (k
x
,k
y
) the transverse spectral vector. In
Eq. (119) Z(0
÷
) denotes the impedance matrix that is
transferred from the topmost layer, and Z(0
÷
) from the
bottommost layer, and Z(0
÷
), Z(0
÷
), and G are all 22
matrices. Thus, once the Green function is known [Eq.
(119)], the microstrip circuit involving a layered structure
MICROSTRIP ANTENNAS 2597
containing anisotropic materials can be solved in a man
ner identical to that shown in Section 2.
As we have just mentioned, wave propagation is aniso
tropic in an anisotropic medium, and in performing nu
merical calculations Galerkin elements need to be all
evaluated in general, since symmetry of the system ceas
es to apply. This is illustrated in Fig. 16, where the in
duced current proﬁles along the width direction of a
microstrip line fabricated on a hexaferrite substrate no
longer show left–right symmetry, due to the gyromagnetic
nature of a hexaferrite material [32]. In Fig. 16 the sub
strate was a singlecrystal Ytype hexaferrite of composi
tion Ba
2
MgZnFe
12
O
22
, and the easy plane coincides with
the substrate surfaces, the xy plane. The hexaferrite sub
strate was characterized by the following parameters: sat
uration magnetization 4pM
S
=2.86kG, anisotropy ﬁeld
H
A
=7.94kOe, dielectric constant e
f
=18, dielectric loss
tangent, tan d
f
=0.01, and FMR linewidth DH=100Oe.
The width of the fabricated microstrip line was w=0.0051
in. and length ‘ =0:0157 in. and the thickness of the hexa
ferrite substrate was d=0.010 in. More calculations and
measurements on this fabricated microstrip line can be
found in Ref. 32.
4. FRACTAL ANTENNAS
Traditional approaches to the analysis and design of an
tenna systems are based on Euclidean geometry. More re
cently, however, there has been a considerable amount of
interest in developing new types of antennas that employ
fractal rather than Euclidean geometries. This new and
rapidly growing ﬁeld is referred to as fractal antenna en
gineering, and this topic is briefly introduced in this
section.
A fractal is a recursive object that has a fractional di
mension showing shape similarity in scale; that is, a
fractal preserves its shape when viewed in different scales.
Many objects, including antennas, can be designed using
the recursive nature of a fractal. The term fractal, which
means broken or irregular fragments, was originally
coined by Mandelbrot [39] to describe a family of complex
shapes that possess an inherent selfsimilarity in their
geometric structure. Since the pioneering work of Man
delbrot and others, a wide variety of applications for
fractals have been found in many branches of science
and engineering. One such area is fractal electrodynam
ics, in which fractal geometry is combined with electro
magnetic theory for the purpose of investigating a new
class of radiation, propagation, and scattering problems.
One of the most promising areas of fractal electrodynamics
is in its application to antenna theory and design.
In an Euclidean geometry, broadband and frequency
independent antennas were developed and analyzed in
the early 1960s, and some convoluted shapes were devel
oped whose performance indicated the dependence of the
antenna radiation properties on its physical size relative
to wavelength. Spirals and logperiodic structures are
some examples of successful developments giving rise to
frequencyindependent radiations. Fractals might join
some of those designs because of their selfscaling prop
erties. Actually, fractal antennas have been shown to pro
vide two advantages: to generate multiband radiations
and to perform beamforming capabilities. As such, the
current fractal antenna engineering is applied in two ar
eas, and the fractal pattern is either used by an isolated
antenna radiator or included as elements in an antenna
array (a multipole conﬁguration). As described in Section
1, since the radiation frequency is inversely proportional
to the linear dimension of a radiator, a fractal antenna can
thus emit multiband radiation [40]. Furthermore, a fractal
antenna can also be viewed as arrays of arrays, and hence
it possesses beamforming capability, if designed and oper
ated properly [41]. It is these novel properties that make
fractal antennas very attractive, allowing phased arrays
to be fabricated with conformity and reduced sizes, result
ing in convenience and ﬂexibility.
Concerning these particular issues, PuenteBaliarda et
al. described [40] the behavior of the ﬁrst fractal multi
band antenna—the Sierpinski monopole. Such a monopole
displayed similar behavior at several bands in terms of
both input return loss and radiation patterns. Some steps
further in the ﬁeld of multiband fractal antennas were
reported by Cohen [43], who also discussed other inter
esting features regarding small and frequencyindepen
dent fractal antennas. In [40] the behavior of both the
Sierpinski monopole and dipole have been investigated,
and their performance has been compared with a conven
tional triangular (bowtie) antenna.
Figure 17 shows the original Sierpinski triangular pat
tern. The Sierpinski pattern is named after the Polish
mathematician Sierpinski, who described some of the
main properties of this fractal shape in 1916 [42]. The
original pattern is constructed by subtracting a central
inverted triangle from a main triangle shape (see Fig. 17).
After the subtraction, three equal triangles remain on the
structure, each one half the size of the original one. One
can iterate this same subtraction procedure on the re
maining triangles, and if the iteration is carried out an
0 0.2 0.4 0.6 0.8 1
Position (Relative)
–1
–0.5
0.5
0
1
Transverse
Longitudinal
C
u
r
r
e
n
t
(
R
e
l
a
t
i
v
e
)
H
0
= 5 KOe
f = 20 GHz
Figure 16. Calculated current proﬁle for the longitudinal and
the transverse components across the strip width of a microstrip
line fabricated on top of a hexaferrite.
2598 MICROSTRIP ANTENNAS
inﬁnite number of times, the ideal fraction Sierpinski pat
tern is obtained. In such an ideal structure, each of its
three main parts is exactly equal to the whole object, but
scaled by a factor of 2, and so are each of the three trian
gles that compose any of those parts. Because of these
particular similarity properties, shared with many other
fractal shapes, it is said that the Sierpinski pattern is a
selfsimilar structure.
In Fig. 17 the Sierpinski pattern is shown with six lev
els in iterations, and the smallest scale is a single triangle
shown shaded in Fig. 17. Sierpinski triangles constitutes
the simplest geometry for a 2D fractal. In the monopole
conﬁguration, the feeder is located at the top vertex of the
big (main) triangle. In this antenna geometry, if one ne
glects the contribution of the central holes to the antenna
performance and admits that the current ﬂowing from the
feeder should concentrate over a region that is comparable
in size to the wavelength, a behavior similar to sixscaled
bowtie antennas (each one operating at its resonant fre
quency) could be expected. The scale factor among the six
iterations is D=2; therefore, one should look for similar
ities at frequencies also spaced by a factor of 2. A Sierpin
ski dipole or quadrupole can be formed by joining two or
four Sierpinski monopoles at the vertex points, respec
tively, showing improved frequency independency with
reduced beamwidth in radiation. A Sierpinski quadrupole
is shown in Fig. 18, which has been used as the company
logo by Hotech, Inc. In conclusion, the most important
features of a fractal array include the frequencyindepen
dent multiband characteristics, schemes for realizing low
sidelobe designs, systematic approach to thinning, and the
ability to develop rapid beamforming algorithm, and
fractal antennas are still active under current research
and development.
5. RECENT DEVELOPMENTS
5.1. Wideband Techniques
A microstrip patch antenna is inherently narrowband,
stemming from the cavity model, where an electromag
netic resonator conﬁning energy locally in space with little
dissipation is necessarily a highQ, and hence narrow
band, device. For a highQ resonator the generation of
surface waves is insignificant. For a thin substrate this
condition amounts to d/l
0
o0.07 for e
r
=2.3 and d/l
0
o
0.023 for e
r
=10, as stated at the end of Section 1 [10,11].
For a microstrip antenna fabricated on a thin substrate,
the impedance bandwidth is typically 1–3%. This is con
trasted with the bandwidth of 16% of a halfwave dipole
with a radius:length ratio equal to 0.01, and 70% of a me
diumlength helix operating in the axial mode.
A number of techniques have been proposed to increase
the bandwidth of a microstrip patch antenna. These tech
niques are generally classiﬁed into three categories: (1) a
straightforward approach based on the use of a thick sub
strate whose dielectric constant is small, (2) an approach
where a matching network is designed to enhance the
bandwidth [44], and (3) a method using parasitically cou
pled elements in a variety of ways to produce closely
spaced multiple resonances of the antenna [45,46].
By using a thick substrate whose dielectric constant is
considerably different from that of air, surface waves will
be generated and inevitably reduce the radiation efﬁciency
and introduce interference between array elements. To
resolve this problem, air cavities or holes may be intro
duced with the substrate to effectively reduce the dielec
tric constant of the patch. For example, Gauthier et al. [47]
machined closely spaced holes in a Duroid substrate un
derneath a microstrip patch to lower the effective dielec
tric constant of the antenna. Using the micromachining
technique, Zheng et al. measured a 12.8% impedance
bandwidth on cavitybacked microstrip patch antennas
fabricated on silicon wafers [48].
The bandwidth of the antenna is determined primarily
by the rate of transition that the imaginary part of the
impedance changes sign at resonance, as discussed at the
end of Section 1. Thus, it is possible to introduce a can
cellation mechanism on the input inductance so as to
Figure 17. Fractal Sierpinski pattern.
Figure 18. A quadrupole fractal.
MICROSTRIP ANTENNAS 2599
smooth the impedance variation. For example, the induc
tance associated with the long wire lead of the coaxial
probe will limit the bandwidth to o10% for a thick sub
strate. By etching a small circular slot around the probe
on the patch, capacitance is thereby introduced, canceling
the probe inductance to produce a bandwidth of 16% [49].
More recently, with the use of a Ushaped slot, a substan
tial increase in bandwidth (32%) has been demonstrated
[50]. Alternatively, an Lshaped probewire has been
shown to result a bandwidth of 28% [51].
The third method to achieving broadband operation is
to couple the microstrip patch antenna parasitically with
other dielectric resonators characterized by approximate
ly the same resonant frequency of the patch. For an iso
lated patch the resonant frequency is determined by its
lateral dimension. However, in the presence of fringe ﬁelds
at the patch periphery the boundary of the antenna is
neither sharply nor rigidly deﬁned, leading to a slightly
larger effective dimension of the patch, as described by
Eqs. (2) and (5) for rectangular and circular microstrip
patches, respectively. Thus, in contrast to a metal cavity,
patch antennas have soft boundaries, which in turn give
rise to ﬁnite bandwidth in radiation; the softer the bound
ary, the wider the bandwidth. For this reason a patch an
tenna fabricated on a thick substrate will show a wide
bandwidth. Similarly, when coupled together many soft
boundary microstrip resonators/patches, the overall radi
ation bandwidth is consequently enlarged.
The disadvantage of using a thick substrate is that,
besides its cost, surface waves may be generated in the
substrate so as to reduce the antenna feeding efﬁciency.
This disadvantage can be overcome by using multiple
electromagnetically (EM) coupled patches. EMcoupled
patches can be deployed either sidebyside (laterally cou
pled geometry) or layerbylayer (vertically coupled geom
etry). For laterally coupled patches the antenna size will
increase considerably, ultimately restricting it usage from
largearray applications [52].
The twolayer EM coupled patch antenna consisting of
a driven (feeder) patch in the bottom and a parasitic (ra
diating) patch on the top has been investigated by several
authors. For circular [53], equitriangular [54], and rect
angular [55] patches, experimental results have shown an
enhanced gain and impedance bandwidth with low cross
polarization levels as compared to the conventional single
layer microstrip antenna. Actually, as analogous to Yagi
antenna, the gain of a stacked antenna can be increased to
above 20 dBi at any scan angle if the thickness of the sub
strate and multiple superstrate layers is chosen properly
[56]. Therefore, using stacked parasitic elements in mi
crostrip arrays could improve the overall array perfor
mance, offering a higher gain and broader bandwidth,
meeting with the same array design criteria of a conven
tional singlelayer microstrip array but with fewer array
elements.
5.2. Size Reduction
Microstrip antennas have a number of advantages over
conventional antennas, namely, small size, light weight,
low production cost, and natural conformal. However, for
many applications, for example, handheld mobile commu
nication systems, halfwave microstrip antennas etched
on a lowcost dielectric substrate are still too large to be
accommodated on the portable phone. A wellknown ap
proach to reduce the size of a halfwave patch to a quarter
wave is to introduce an electrically shortcircuiting wall at
one of the radiating edges. Hiraswa and Haneishi [57] has
shown that the length of the patch can be made sufﬁcient
ly shorter than a quarterwave by replacing the shortcir
cuiting wall with a shortcircuiting pin at the corner of the
patch. Wang and Lin have shown that by replacing the
shortcircuiting pin with a chip resistor of low resistance,
the antenna size can be further reduced in addition to an
increase in bandwidth [58].
The other approach to reduce antenna size is to use a
meandered geometry of the patch antenna [57]. By mean
dering the patch, the effective electrical length is greater
than the physical length. Consequently, the resonant fre
quency of the meandered antenna can be much lower than
that of a conventional design using the same physical
length [59].
Finally, we note that it is not only possible to achieve
high gain for printed circuit antennas, it is also feasible to
shape the radiation pattern in some prescribed manner.
Some very interesting phenomena, such as radiation into
the horizon, radiation pattern monodirectionality, and az
imuthdependent radiation, have been found possible [60].
Furthermore, laterally coupled patches can serve as adap
tive array antennas, because the phase of radiation from a
parasitic element can be adjusted via a varactor diode in
serted at some feeding position [61]. The important fea
ture of an adaptive array antenna is that it can provide
beamsteering as required by many communication and
trafﬁc control systems [62].
BIBLIOGRAPHY
1. A. Sommerfeld, Partial Differential Equations, Academic
Press, New York, 1962.
2. J. R. James and P. S. Hall, eds., Handbook of Microstrip
Antennas, Vols. 1, 2. Peter Peregrinus, London, 1989.
3. H. How, T.M. Fang, and C. Vittoria, Intrinsic modes of
radiation in ferrite patch antennas, IEEE Trans. Microwave
Theory Tech. MTT42(6):988 (1994).
4. E. O. Hammerstad, Equations for microstrip circuit design,
Proc. 5th European Microstrip Conf., Hamburg, Sept. 1975,
pp. 268–272.
5. M. V. Schneider, Microstrip lines for microwave integrated
circuits, Bell Syst. Tech. J. 48:1421–1444 (1969).
6. L. C. Shen, Resonant frequency of circular disc, printed
circuit antenna, IEEE Trans. Anten. Propag. AP25:596
(1977).
7. J. A. Stratton, Electromagnetic Theory, McGrawHill, New
York, 1941.
8. R. S. Elliott, Antenna Theory and Design, PrenticeHall,
Englewood Cliffs, NJ, 1981.
9. A. G. Derneryd, Microstrip array antenna, Proc. 6th Europe
an Microwave Conf., 1976, pp. 339–343.
10. J. R. James and A. Henderson, Highfrequency behavior of
microstrip opencircuit terminations, IEE J. Microwaves Opt.
Acoust. 3:205–218 (1979).
2600 MICROSTRIP ANTENNAS
11. C. Wood, Analysis of microstrip circular patch antennas, IEE
Proc. 128(H):69–76 (1981).
12. S. B. A. Fonseca and A. J. Giarola, Microstrip disk antennas,
Part I: efﬁciency of space wave launching, IEEE Trans. Anten.
Propag. AP32:561–567 (1984).
13. Y. T. Lo, D. Solomon, and W. F. Richards, Theory and exper
iment on microstrip antennas, IEEE Trans. Anten. Propag.
AP27:137–145 (1979).
14. R. C. Booton, Jr., Computational Method for Electromagnetics
and Microwaves, Wiley, New York, 1992.
15. H. How and C. Vittoria, Computer aided design tools for
microstrip circuitries, in N. K. Das and H. L. Bertoni, eds.,
Directions for the Next Generation of MMIC Devices and Sys
tems, Plenum Press, New York, 1997, pp. 399–406.
16. J. R. Mosig, R. C. Hall, and F. E. Gardiol, Numerical analysis
of microstrip patch antennas, in J. R. James and P. S. Hall,
eds., Handbook of Microstrip Antennas, Peter Peregrinus,
London, 1989, pp. 393–453.
17. J. D. Jackson, Classical Electrodynamics, Wiley, New York,
1975.
18. S. Y. Liao, Microwave Devices and Circuits, PrenticeHall,
Englewood, Cliffs, NJ, 1985.
19. H. How, C. Vittoria, and T. Fang, New formulation of dyadic
Green’s function: applied to a microstrip, IEEE Trans. Micro
wave Theory Tech. MTT41(8):1580 (1994).
20. K. C. Kupta, R. Garg, and I. J. Bahl, Microstrip Lines and
Slotline, Artech House, Norwood, MA, 1979.
21. H. How, R. Seed, C. Vittoria, D. B. Chrisey, and J. S. Horwitz,
Microwave characteristics of superconducting coplanar wave
guide resonator, IEEE Trans. Microwave Theory Tech. MTT
41(3):255 (1993).
22. R. F. Harrington, Time Harmonic Electromagnetic Fields,
McGrawHill, New York, 1961.
23. H. How and C. Vittoria, Green’s function calculations on cir
cular microstrip patch antennas, IEEE Trans. Anten. Propag.
AP49:393 (1991).
24. H. How and C. Vittoria, Radiation modes in dielectric circular
patch antennas, IEEE Trans. Microwave Theory Tech. MTT
42(10):1939 (1994).
25. D. M. Pozar, Input impedance and mutual coupling of rect
angular microstrip antennas, IEEE Trans. Anten. Propag.
AP30:1191 (1982).
26. B. Lax and K. J. Button, Microwave Ferrites and Ferrimag
netics, McGrawHill, New York, 1962.
27. N. Alexo´poulos, Integratedcircuit structures on anisotropic
substrates, IEEE Trans. Microwave Theory Tech. MTT33:847
(1985).
28. T. Q. Ho and B. Becker, Frequencydependent characteristics
of shielded broadside coupled microstrip lines on anisotropic
substrates, IEEE Trans. Microwave Theory Tech. MTT
39:1021 (1991).
29. G. W. Hanson, A numerical formulation of dyadic Green’s
functions for planar bianisotropic media with application to
printed transmission lines, IEEE Trans. Microwave Theory
Tech. MTT44:144 (1996).
30. M. R. Rosa, M. L. Albuquerque, A. G. D’Assumcio, R. G. Maia,
and A. J. Giarola, Full wave analysis of microstrip lines on
ferrite and anisotropic dielectric substrate, IEEE Trans.
Magn. Mag25:2944 (1989).
31. ELB. ElSharawy and R. W. Jackson, Coplanar waveguide
and slot line on magnetic substrates: Analysis and experi
ment, IEEE Trans. Microwave Theory Tech. MTT36:1071
(1988).
32. H. How, X. Zuo, E. Hokanson, and C. Vittoria, Calculated and
measured characteristics of a microstrip line fabricated on a
Ytype hexaferrite substrate, IEEE Trans. Microwave Theory
Tech. MTT50(5):1280–1288 (2002).
33. H. How, Magnetic microwave devices, in J. G. Webster, ed.,
Encyclopedia of Electrical and Electronics Engineering,
Wiley, New York, 1999, Vol. 12, pp. 31–45.
34. H. How, P. Shi, C. Vittoria, E. Hokenson, M. H. Champion, L.
C. Kempel, and K. D. Trott, Steerable phased array antennas
using singlecrystal YIG phase shifters—theory and experi
ments, IEEE Trans. Microwave Theory Tech. MTT48:1544
(2000).
35. D. M. Pozar, Radiation and scattering characteristics of mi
crostrip antennas on normally biased ferrite substrates, IEEE
Trans. Anten. Propag. AP30:1084 (1992).
36. H. How, S. A. Oliver, S. W. McKnight, P. M. Zavracky, N. E.
McGruer, and C. Vittoria, Theory and experiment of thinﬁlm
junction circulator, IEEE Trans. Microwave Theory Tech.
MTT46:1645 (1998).
37. W. H. Von Aulock, ed., Handbook of Microwave Ferrite Mate
rials, Academic Press, New York, 1965.
38. N. Zeina, H. How, and C. Vittoria, Selfbiasing circulators op
erating at Ka band utilizing Mtype hexagonal ferrites, IEEE
Trans. Magn. Mag28:3219 (1992).
39. B. B. Mandelbrot, The Fractal Geometry of Nature, Freeman,
New York, 1983.
40. C. PuenteBaliarda, J. Romey, and R. Pous, On the behavior of
the Sierpinski multiband fractal antenna, IEEETrans. Anten.
Propag. AP46:517–524 (1998).
41. D. H. Werner, R. L. Haupt, and P. L. Werner, Fractal antenna
engineering: The theory and design of fractal antenna arrays,
IEEE Trans. Anten. Propag. AP41:37–59 (1999).
42. H. O. Peitgen, H. Juergens, and D. Saupe, Chaos and
Fractals, SpringerVerlag, New York, 1983.
43. N. Cohen, Fractal antennas, Commun. Quart. 53–66 (1996).
44. H. An, K. J. C. Nauwelaers, and A. R. Van de Capelle, Broad
band microstrip antenna design with the simpliﬁed real fre
quency technique, IEEE Trans. Anten. Propag. AP42(2):129–
136 (Feb. 1994).
45. R. Q. Lee, K. F. Lee, and J. Bobinchak, Characteristics of a
two layer electromagnetically coupled rectangular patch an
tenna, Electron. Lett. 23:1070–1072 (Sept. 1987).
46. C. K. Aanandan, P. Mohanan, and K. G. Nair, Broadband gap
coupled microstrip antenna, IEEE Trans. Anten. Propag. AP
38(10):1581–1586 (Oct. 1990).
47. G. P. Gauthier, A. Courtay, and G. M. Rebeiz, Microstrip
antenna on synthesized lowdielectric constant substrates,
IEEE Trans. Anten. Propag. AP45(8):1310–1314 (Aug.
1997).
48. M. Zheng, Q. Chen, P. S. Hall, and V. F. Fusco, Broadband
microstrip patch antennas on micromachined silicon sub
strates, Electron. Lett. 34(1):3–4 (1998).
49. P. S. Hall, Probe compensation in thick microstrip patches,
Electron. Lett. 23:606–607 (1987).
50. K. F. Lee, K. M. Luk, K. F. Tong, S. M. Shum, T. Huynh, and
R. Q. Lee, Experimental and simulation studies of coaxially
fed Ulot rectangular patch antenna, IEE Proc. Microwave
Anten. Propag. 144:354–358 (1997).
51. K. M. Luk, C. L. Mak, Y. L. Chow, and K. F. Lee, Broadband
microstrip patch antenna, Electron. Lett. 34(15):1442–1443
(1998).
52. S. Dey and R. Mittra, A compact broadband microstrip an
tenna, Microwave Opt. Technol. Lett. 11(6):295–297 (1996).
MICROSTRIP ANTENNAS 2601
53. A. Sabban, A new broadband stacked twolayer microstrip
antenna, IEEE APS Int. Symp. Digest, 1983, pp. 63–66.
54. P. S. Bhatnagar, J. P. Daniel, K. Mahdjoube, and C. Terret,
Experimental study on stacked triangular microstrip anten
nas, Electron. Lett. 22:864–865 (1986).
55. R. Q. Lee and K. F. Lee, Experimental study of the twolayer
electromagnetically coupled rectangular patch antenna,
IEEE Trans. Anten. Propag. AP38:1298–1302 (1990).
56. H. Y. Yang and G. N. Alexopoulos, Gain enhancement meth
ods for printed circuit antennas through multiple superst
rates, IEEE Trans. Anten. Propag. AP35:860–863 (1987).
57. K. Hiraswa and M. Haneishi, Analysis, Design, and Measure
ment of Small and LowProﬁle Antennas, Peter Peregrinus,
London, 1992, Chap. 5.
58. K. Wang and Y. Lin, Small broadband rectangular microstrip
antenna with chipresistor loading, Electron. Lett.
33(19):1593–1594 (1997).
59. S. Dey and R. Mittra, Compact microstrip patch antenna,
Microwave Opt. Technol. Lett. 13(1):12–14 (1996).
60. H. Y. Yang and N. G. Alexopoulos, Generation of nearly hemi
spherical and high gain azimuthal symmetric patterns with
printed circuit antennas, IEEE Trans. Anten. Propag. 35:972–
977 (1987).
61. D. Cailleu, N. Haese, and P. A. Rolland, Microstrip adaptive
array antenna, Electron. Lett. 32(14):1246–1247 (1996).
62. R. J. Dinger, A planar version of a 4.0GHz reactively steered
adaptive array, IEEE Trans. Anten. Propag. 34(3):427–431
(1986).
MICROSTRIP ANTENNAS, BROADBAND
K. M. LUK
C. L. MAK
City University of Hong Kong
Kowloon, Hong Kong
R. CHAIR
The University of Mississippi
Oxford, Mississippi
H. WONG
City University of Hong Kong
Kowloon, Hong Kong
K. F. LEE
The University of Mississippi
Oxford, Mississippi
1. INTRODUCTION
The microstrip patch antenna offers many attractive fea
tures, including low proﬁle, light weight, conformal struc
ture, and inexpensive fabrication costs. The major
drawback of this antenna is narrow bandwidth in its ba
sic form. This article collects different approaches for
bandwidth enhancement of linearly polarized microstrip
antennas from conventional means to presentday meth
ods, with emphasis on advanced techniques proposed by
the authors (Section 2), particularly on the Lprobe cou
pling technique. However, due to limited spacing, the
development of wideband elliptically polarized microstrip
antennas will not be covered here. Interested readers can
refer to some of the more recent designs reported in the
literature [1–4].
The article also highlights wideband techniques for de
signing dualpolarized patch antennas (Section 3) and
electrically small patch antennas (Section 4). In Section 3,
the popular aperture coupling method for a dualpolarized
patch antenna is described ﬁrst. Emphasis is then placed
on the wideband Lprobecoupled dualpolarized patch an
tenna. A technique for improving the isolation between
the input ports is also presented in that section. Section 4
focuses on miniature patch antennas, including patches
with shorting (shortcircuiting) pins or shorting walls.
Several wideband methods such as the use of stacked
geometry, and the Lprobe feed, are examined.
The wideband antennas discussed in Sections 3 and 4
are essential in today’s wireless communication systems.
For instance, the polarization diversity technique is found
useful in cellular systems, and therefore most of the base
station antennas require a 7451 dualpolarized feature as
well as broadband characteristic. Modern mobile devices
are getting smaller and smaller in size. Thus the dimen
sions of the antennas inside these devices have to be re
duced accordingly. With the novel wideband techniques
discussed, both dualpolarized patch antennas and minia
ture patch antennas can be operated with widebandwidth
performance, which will be particularly useful in future
highspeed wireless communications. The advanced wide
band techniques proposed by the authors can also be ap
plied to the design of dualband and circularpolarized
patch antennas.
2. LINEARLY POLARIZED WIDEBAND MICROSTRIP
PATCH ANTENNAS
According to The New IEEE Standard Dictionary of Elec
trical and Electronics Terms [5], the bandwidth of an an
tenna is deﬁned as ‘‘the range of frequencies within which
its performance, in respect to some characteristics, con
forms to a speciﬁed standard.’’ This characteristic can be
selected as the input impedance or radiation pattern. For
the former, which is more sensitive to frequency, a com
mon standard is that the voltage standingwave ratio
VSWR, or simply SWR, should be less than a certain val
ue, usually 1.5 (G(f)=0.2) or 2 ([G(f )[ =
1
3
), where G is the
reﬂection coefﬁcient. Equivalently, the return loss S
11
(f)
should be less than –14 or –10 dB, respectively. Moreover,
we usually express the impedance bandwidth in terms of
percentage (BW), which is calculated by
BW=
2(f
U
÷f
L
)
f
U
÷f
L
100%
where f
U
and f
L
are respectively the upper and lower cutoff
frequencies of the operating bandwidth. Key methods de
veloped to solve the inherently narrow bandwidth issue of
linearly polarized microstrip patch antennas are dis
cussed below.
2602 MICROSTRIP ANTENNAS, BROADBAND
2.1. Utilization of Electrically Thick Substrate with
Low Permittivity
The patch antenna works like a lossy dielectricloaded cav
ity resonator bounded by two electric and four magnetic
walls. Assuming that the only loss is due to radiation, we
have the antenna quality factor Q=2pf
r
U/P
rad
, where U is
the total stored energy trapped inside the cavity between
the patch and the ground plane at the resonant frequency
f
r
. P
rad
is the total radiated power at f
r
. This quality factor
is similar to that of a parallelresonant circuit having
lumped circuit elements and can also be determined from
the input resistance R(f) and input reactance X(f) at a fre
quency close to f
r
. It is understood that a lower value of Q
will result in a wider bandwidth. For a patch antenna em
ploying a thin substrate, a higher value of Q [6] and a
narrower bandwidth will be obtained. The quality factor
can be lowered by simply increasing the substrate
thickness; however, the surface wave introduced will
deteriorate the overall performance, including the band
width. Fortunately, a substrate with lower value of per
mittivity e =e
r
e
0
can minimize the surfacewave power. It is
known that the foam or air substrate has the lowest value
of relative permittivity e
r
close to unity. Thus a patch an
tenna with electrically thick substrate could exhibit a wide
bandwidth [7]. Examples are described in Refs. 8 and 9. In
Ref. 8, the bandwidth of a patch antenna over a large
ground plane with an air substrate thickness 0.14–0.27 l
0
was shown to be about 40% (VSWRr2). In Ref. 9, a similar
concept was implemented. The effective substrate thick
ness was increased by the use of a Wshaped ground plane.
A 12% (VSWRr1.5) bandwidth was reported.
2.2. Bandwidth Improvement by Related Impedance
Matching in Different Feeding Mechanisms
2.2.1. Proximity or Electromagnetic Coupling. The prox
imity coupling technique for patch antennas was proposed
by Pozar and Kaufman in 1987 [10]. Energy is electro
magnetically coupled from the feedline through a sub
strate to the radiating element such as a patch [10] or a
printed dipole [11]. It has the merit of being less sensitive
to fabrication errors. For the proximitycoupled patch as
described in Ref. 10 and shown in Fig. 1, there is a rela
tively small opencircuited tuning stub connected in par
allel with the feedline, which provides related matching to
the feedpoint; hence the bandwidth can be increased to
13% (VSWRr2).
Another proximitycoupled patch with wider band
width was proposed in 1998 [12]. The geometry is shown
in Fig. 2. Instead of adding a tuning stub, the feedline is
Radiating patch
Supporting substrate
Ground plane
Dielectric substrate
Microstrip feedline
Opencircuited stub
Figure 1. Proximitycoupled patch antenna. (From Pozar and
Kaufman [10], reprinted with permission from IEE.)
40
25
2
2 2
15
15
5
Ground plane
Feedline
Uslotted patch
13
100 Ω line
1.36
19
14
4.75
15
=2.33
1.6 1.6
3
F
o
a
m
s
u
b
s
t
r
a
t
e
Unit: mm
50 Ω line
c
r
Figure 2. Geometry of proximitycoupled U
slot patch antenna. (From Mak et al. [12],
reprinted with permission from IEE.)
MICROSTRIP ANTENNAS, BROADBAND 2603
terminated by a Pshaped stub, which is underneath the
radiating patch. The patch is etched with a Ushaped slot,
the function of which will be elaborated later. A bandwidth
of 20% (VSWRr2) and a gain of 7.5 dBi were achieved for
this singlelayer patch antenna. Figures 3 and 4 show the
measured VSWR and gain of this antenna. Figure 5 shows
the radiation patterns at 4.3GHz. The antenna has a
broadside radiation pattern with beamwidth of about 601
in both E plane and H plane. The crosspolar level is below
–15 dB. This singleelement antenna has been modiﬁed
and employed for the design of wideband antenna array
with excellent performance [13,14].
2.2.2. Aperture Coupling Method. The aperture cou
pling method was proposed 1985 by Pozar again [15].
The basic geometry is shown in Fig. 6. The microstrip
feedline and the patch are located on different sides of the
ground plane. Energy from the microstrip feedline is elec
tromagnetically coupled to the patch through an aperture
on the ground plane. The coupling aperture can be of dif
ferent shapes, sizes, and locations with respect to the
patch for different coupling characteristics [16,17]. An ad
vantage of this feeding method is that the feedline is iso
lated from the radiating element by the ground plane.
Thus radiation from the feedline in the broadside direction
is insignificant. Also this method do not have the probe
inductance problem arising in the coaxial probe feed case.
Disadvantages are comparative complexity in fabrication
and relatively high level of backlobe radiation.
From Ref. 15, the bandwidth obtained was only 14%.
There are two simple ways for tuning the antenna: the
aperture shape and size and the stub length. So it is pos
sible to obtain a better bandwidth if appropriate matching
is achieved. One example was presented in 2002 [18], with
the patch excited by an Hshaped aperture. An impedance
bandwidth of 56.2% (VSWRr2) and a 3 dB gain band
width of 24% were achieved.
2.2.3. MicrostripLine Edge Feed and Coaxial Probe
Feed. The microstripline edge feed and the coaxial probe
feed are fundamental feeding methods for microstrip
patch antennas, since they are the most direct way to
transfer energy from the feedline to the patch and vice
versa. Some methods have been proposed in the literature
to improve the matching condition and the bandwidth as
well. In Ref. 19, a patch is excited by a coupled linefeed. It
was reported that the bandwidth can be increased by a
8
7
6
5
4
3
2
1
2 3 4 5 6
Without Πstub
and Uslot
With Πstub
and Uslot
(S = 15 mm)
Frequency (GHz)
Without Πstub
but with Uslot
With Πstub but
without Uslot
S
W
R
Figure 3. Variation of VSWR of the proximitycoupled patch an
tenna. (From Mak et al. [12], reprinted with permission from
IEE.)
3.5 4.0 4.5 5.0
0
2
4
6
8
Frequency (GHz)
G
a
i
n
(
d
B
i
)
Figure 4. Gain of the proximitycoupled Uslot patch antenna.
(From Mak et al. [12], reprinted with permission from IEE.)
10 dB
0
−10
−20
−30
−40
Figure 5. Radiation pattern at 4.3GHz. (From Mak et al. [12],
reprinted with permission from IEE.)
Radiating patch
Coupling aperture
Microstrip feed line
Dielectric substrate
Ground plane
Supporting substrate
Figure 6. Basic geometry of aperturecoupled patch antenna.
2604 MICROSTRIP ANTENNAS, BROADBAND
factor of 42.5 as compared to the conventional edge feed
method.
In using the coaxial probe feed, there are even more
methods available to improve the matching condition of
the singlepatch conﬁguration such as introducing a slot
on the patch, adding a capacitor patch, or using the
Lshaped probe. These techniques will be discussed in
the following sections.
2.3. Bandwidth Widening Technique Using Parasitic
Elements
The original geometry of a microstrip patch antenna
consists of only one patch radiator, which is driven by
the appropriate feeding techniques. The bandwidth of
such a conﬁguration can be very narrow. A significant
improvement in bandwidth can be achieved by adding one
or more parasitic patches next to or stacked above the
Driven patch
Substrate
Ground plane
Parasitic patches
Feed
Figure 8. Examples of parasitic patches in stacked conﬁgura
tion.
Ground plane
Radiating patch
Capacitor patch
Feeding probe
Figure 9. Capacitively fed patch antenna (side view).
17 mm 10.5 mm 8.5 mm 155° 80° 14% 12.5 mm
R
1
R
2
Bandwidth (VSWR ≤ 2) R r 0 0
1
R
1
R
2
R
r
Feedpoint
0
0
1
Figure 10. Circular patch with an arc
shaped slot. (From Luk et al. [41], reprint
ed with permission from IEE.)
2 3 4 5 6
10
8
6
4
2
0
Frequency (GHz)
V
S
W
R
Figure 11. Measured VSWR of the offset feed circular patch with
arcshaped slot. (From Luk et al. [41], reprinted with permission
from IEE.)
Nonradiating edges
Driven patch
Substrate
Couplinggap
Ground plane
Parasitic patches
Feed
Figure 7. Examples of parasitic patches in coplanar conﬁgura
tion.
MICROSTRIP ANTENNAS, BROADBAND 2605
driven patch. The parasitic elements can be of slightly
different dimensions and are excited by electromagnetic
coupling from the driven patch. By adding parasitic patch
es, the bandwidth can be improved because there are one
or more resonant frequencies introduced by the parasitic
patches. If these resonant frequencies are close to the
original resonant frequency of the driven patch, they could
combine together to yield a wider bandwidth. Obviously,
the major drawback of using parasitic elements is the en
largement in volume of the antenna.
2.3.1. Coplanar Structure. In the coplanar geometry,
the parasitic patches are placed on the same layer and
adjacent to the driven patch as shown in Fig. 7. The co
axial probe feed [20–22] or the aperturecoupled feed [23]
can be chosen to excite the driven patch. When using the
coaxial probe feed method, 15% bandwidth (VSWRr2)
can be yielded when two parasitic patches are placed next
to the nonradiating edge of the driven patch [21]. In ad
dition, if we add four parasitic patches gapcoupled to the
four edges of the driven patch, around 25% bandwidth
(VSWRr2) can be obtained [21]. As for using the aper
turecoupled feed, 23% bandwidth (VSWRr2) was
achieved [23] when two parasitic elements were gap
coupled to the nonradiating edges of the fed patch.
2.3.2. Stacked Structure. In the stacked geometry, the
parasitic patches are suspended above the driven patch
and thus are placed in another upper layer with respect to
the driven patch as shown in Fig. 8. Again, both coaxial
probe feed [24,25] and aperturecoupled feed [26–31]
methods for the driven patch are commonly used. Newer
feed methods for the driven patch have been proposed: the
coplanar waveguide feed (CPW) [32,33] and the Lshaped
strip feed [34]. In using the coaxial probe feed method,
more than 25% bandwidth (VSWRr2) can be achieved
[25]. If the aperturecoupled feed is employed, around 40%
bandwidth (VSWRr2) was obtained [28,29].
2.4. Wideband Patch Antenna with Probe
Inductance Cancellation
As mentioned in Section 2.1, employing electrically thick
substrate could increase the impedance bandwidth by
about 20% (VSWRr2) [7] when the substrate thickness
is around 0.15l
0
. However, in the probe feed case, the
bandwidth is limited to about 10% (VSWRr2) because of
Frequency (GHz)
3.5 3.9 4.3 4.7 5.1 5.5
10
9
8
7
6
5
4
3
2
1
V
S
W
R
Measurement
FDTD simulation XXXXXXXX
Figure 13. Measured and FDTDsimulated VSWR of rectangu
lar Uslot patch antenna. (From Lee et al. [47], reprinted with
permission from IEE.)
Slots
Slots
Feed
Feed
Figure 14. Examples of Uslot equivalent patches [52–54].
h air
Ground plane
L F
15 2.1 5
W Parameters W
S
L
S
a b c h Bandwidth (VSWR ≤ 2)
32.4% 4.3 2.2 19.5 12 26 35.5 Unit/mm
a
L
s
b
W
L
c
W
S
c
c
F
Figure 12. Wideband rectangular Uslot
patch antenna. (From Lee et al. [47], reprint
ed with permission from IEE.)
2606 MICROSTRIP ANTENNAS, BROADBAND
the larger probe inductance [35,36] introduced by the
longer probe required. This undesired inductance results
in antenna mismatch. Moreover, the unwanted radiation
from the longer probe is a major source of high crosspo
larization in the H plane and also causes an unsymmetric
copolarization in the E plane. Thus a shorter probe and a
thicker substrate are desired in designing wideband patch
antennas. In one method, employing a Wshaped ground
plane [9] to increase the (effective) substrate thickness
without lengthening the feeding probe, a bandwidth of
15% (VSWRr1.5) can be achieved. Apart from retaining a
short feeding probe, there are other effective methods to
compensate the unwanted probe inductance in wideband
patch antenna design.
2.4.1. Capacitive Coupling Patch Antenna. In the design
of capacitively coupled patch antenna, a small patch,
which is connected to the feeding probe, is introduced be
tween the ground plane and the radiating patch as shown
in Fig. 9 [37–39]. The structure is somewhat similar to the
probefed stackedpatch antenna, but the small patch to
gether with the radiating patch works as a capacitor
that is used to cancel out the unwanted probe inductance
within the antenna structure itself. Thus a simpler
matching network can be used at the back of the ground
plane. About 5% impedance bandwidth (VSWRr2) was
reported [37].
2.4.2. Coaxial ProbeFed Slotted Patch. Without using
an additional capacitor patch between the ground and the
radiating patch, the unwanted probe inductance can be
canceled out by introducing appropriate slots, surround
ing the feedpoint, on the patch [40]. The method can retain
the advantage of a singlepatch conﬁguration. Not long
ago, the authors introduced differentshaped slots on the
patch, including an arcshaped slot and a Ushaped
slot, for wideband operation [41,42]. The latter case with
Ushaped slot has received much attention in the litera
ture [12,13,43–51]. Many similar designs have also been
proposed such as the Eshaped patch [52,53], and the
backtoback Lshaped slotted patch [54].
2.4.2.1. Circular Patch with an ArcShaped Slot. In a
procedure reported in 1997 [41], an arcshaped slot was
cut inside the circular patch as shown in Fig. 10. It is
found that the offset feed (r40) will give a better band
width (VSWRr2) of 14% than the 11% of the center feed
case (r =0). Figure 11 shows the measured VSWR
response.
y
y
x
W
x
W
y
S
H
L
v
2R
L
h
W
x
L
h
L
v
25 0.5 36%
−Z
Lshaped probe
Ground
plane
Bandwidth (VSWR ≤ 2) S R H W
y
Parameters
Unit/mm 30 6.6 5.5 10.5 2
Figure 15. Basic geometry of Lshaped
probefed patch antenna. (From Luk et al.
[55], reprinted with permission from IEE.)
Frequency (GHz)
3
1
2
3
4
5
6
7 10
5
0
−5
G
a
i
n
(
d
B
i
)
−10
−15
−20
Copol.
Xpol.
VSWR
V
S
W
R
3.5 4 4.5 5 5.5 6
Figure 16. VSWR and gain of Lshaped probefed rectangular
patch antenna. (From Luk et al. [55], reprinted with permission
from IEE.)
MICROSTRIP ANTENNAS, BROADBAND 2607
2.4.2.2. Rectangular Patch with a UShaped Slot. The de
sign was ﬁrst proposed in 1995 by Huynh and Lee [42]. An
impedance bandwidth of around 30% (VSWRr2) can be
easily obtained [47]. A typical example is described below
with the basic geometry shown in Fig. 12. Note that the
thickness is around 0.08 l
0
. Figure 13 shows the measured
and simulated [ﬁnitedifference timedomain (FDTD)]
results of VSWR of this wideband antenna.
The arcshaped or Ushaped slot, surrounding the feed
point, on the patch provides a capacitive component in the
input impedance of the antenna, compensating the strong
inductive component from the long feeding probe. There
fore, appropriate matching can be obtained for wideband
performance. Apart from the broadband characteristics,
the Uslotted patch preserves the singlepatch lowproﬁle
feature in comparison to the geometry with parasitic
patches. The thickness of such a wideband singlepatch
antenna is only r0.08 l
0
. Due to such an attractive
feature, many Uslot equivalent patch antennas have
been designed for bandwidth enhancement. For examples,
an Eshaped patch [52,53] and the backtoback L
shaped slotted patch [54] were proposed, as shown in
Fig. 14. Similar wide bandwidths of around 30% were re
ported.
2.5. Wideband Patch Antenna with LShaped or
TShaped Probe
It is believed that using slightly thicker substrate with low
permittivity for patch antenna design is the most effective
way to achieve wideband operation, at the same time re
taining the attractive original lowproﬁle characteristic.
Of course, the probefed Uslotted patch can easily achieve
such merit of wideband and lowproﬁle structure. Howev
er, the fabrication process is not so convenient, and the
ﬁnetuning process of the Ushaped slot is not an easy
task. Such problems may increase the fabrication cost and
time.
2.5.1. LProbeCoupled Patch Antenna. More recently, a
novel Lprobe coupling technique was invented, and
awarded patents from the People’s Republic of China
and the United States. This design, ﬁrst proposed in
1998 [55], preserves the original advantages of patch an
tennas with an improvement in impedance bandwidth
without increasing the fabrication difﬁculties since no
slot is needed on the patch and no soldering process is
required. Many investigations have been undertaken on
Lshaped probe coupling [49,50,56–66], including the cir
cular polarization design and the dualband design.
The basic geometry of the Lprobecoupled patch an
tenna is shown in Fig. 15. This is a singlelayer patch an
tenna with a rectangular patch of width W
x
=30 mm
(0.45l
0
) and length W
y
=25 mm (0.375l
0
). Note that the
operating center frequency is f
0
E4.5 GHz. The copper
made Lshaped probe is constructed by simply bending a
straight copper wire of radius R=0.5 mm (0.0075l
0
) into
an L shape. When this Lshaped probe is then connected
to the inner conductor of a 50O launcher without touching
the patch, as shown in Fig. 15, it becomes an effective
coupling probe for the radiating patch. The Lshaped
probe has a vertical arm length L
v
=10.5 mm (0.1575l
0
)
and a horizontal arm length L
h
=5.5mm (0.0825l
0
), which
are exciting the fundamental TM
01
mode of the rectangu
lar patch. A supporting foam layer (e
r
E1) is employed
with thickness H=6.6mm (0.099l
0
), which is about one
tenth of the operating wavelength. Moreover, from the top
view, the distance between the lower edge of the patch and
the horizontal arm of the Lshaped probe is SE2 mm. The
values of these parameters are chosen after a series of ex
tensive appraisals, from HE3.3 mm (0.05l
0
) to HE16 mm
(0.25l
0
) and are believed to provide an optimum perfor
mance of bandwidth and gain.
Both the measured VSWR and gain (copolarization
gain at y =01 and crosspolarization gain at yE451 in
(a)
0dB
0dB
0dB
0dB
0dB
−30
−20
−10
−30
−20
−10
−30
−20
−10
−30
−20
−10
−30
−20
−10
−30
−20
−10
0dB
HPlane EPlane
EPlane
(b)
HPlane
(c)
HPlane
Copol. Xpol.
EPlane
Figure 17. Farﬁeld radiation patterns of Lshaped probefed
patch antenna: (a) 4 GHz; (b) 4.53GHz; (c) 5.34GHz. (From Mak
et al. [56], r 2000 IEEE.)
2608 MICROSTRIP ANTENNAS, BROADBAND
the H plane) are shown in Fig. 16, where y is the azimuth
angle. A wide bandwidth of up to 36% (VSWRr2) with
7.5 dBi gain is obtained. The crosspolarization level in the
H plane is slightly higher at the two ends of the band of
operation, and about 15 dB down from the copolarization
level is attained at the center region of the band of oper
ation. The farﬁeld radiation patterns measured at 4, 4.53,
and 5.34 GHz are shown in Fig. 17. It can be observed that
b h
1.6
Parameters W L H S a
Unit/mm 30 25 8.3 5 10 1 49%
Bandwidth (VSWR ≤ 2)
(b)
z
y
H
a
S b
h
Microstrip
feedline
Dielectric
substrate
(a)
x
W
L
z
Patch
Ground
plane
Ground plane
Foam
Foam
Foam
Patch(P1)
Lstrip
Dielectric substrate
Microstrip
feedline
Figure 18. Geometry of Lshaped stripfed
rectangular patch antenna. (From Mak et al.
[34], reprinted with permission from IEE.)
Parameters H W
y
W
x
T
h
T
v
h
30 9
y
x
W
y
Wx
S
Tprobes
Rectangular patch
S
Microstrip feedline
Ground plane
h
y
z
Microstrip feedline
Tshaped probes
Dielectric substrate
(c
r
= 2.33)
H
T
h
T
v
z
g
/2
Quarterwave
transformer
5.6 1.6 6.6 25 Unit /mm 41.6%
Bandwidth (VSWR ≤ 2)
Figure 19. Geometry of the twoTprobecou
pled patch antenna. (From Mak et al. [67], re
printed with permission from IEE.)
MICROSTRIP ANTENNAS, BROADBAND 2609
the copolarization patterns are stable across the band
of operation, with the mainbeam toward the broadside
direction.
Apart from the rectangular patch coupled by the L
shaped probe as described, circular and triangular patch
es are also investigated [57,58], which are also able to
showthe wideband characteristic. It is believed that the L
shaped probe coupling technique is also suitable for other
common patch shapes such as elliptical patch, semidisk,
and circular ring, as well as the bowtie patch and the
notched patch for different requirements. The wideband
performance of the Lprobecoupled patch antenna is
achieved by two basic insights: (1) the use of an electri
cally thick substrate (E0.1l) [7], which provides a lower Q
factor, thus potentially giving a wider bandwidth; and (2)
probe inductance compensation [37–42], where, from the
transmissionline perspective, the vertical arm between
the patch and the ground forms an open circuit stub of
length less than a quarterwavelength; thus a capacitive
reactance is provided. This is similar in concept to the ca
pacitive coupled patch antenna [37–39]. Moreover, it is
necessary to note that the total length of the Lshaped
probe is less than a quarterwavelength (i.e., L
v
÷
L
h
o0.25l), thus maintaining the patch as the only radi
ating element. Experiments also show that the L probe
itself without the patch does not act as an efﬁcient radiator
across the operation band.
In the process of ﬁnding an optimum conﬁguration, it is
found that the bandwidth generally increases with the
value of H, the separation between the patch and ground.
However, an optimum gain is obtained when H is around
0.1l. Furthermore, the gain is highly sensitive to the value
of H. Measurements also show that the gain is usually
higher when part of the L probe is not under the patch, as
shown in Fig. 15, than when the whole L probe is com
pletely covered by the patch [56]. It is essential to mention
that when varying the value of H, it is necessary to
slightly alter the other parameters such as L
v
, L
h
, and
S simultaneously in order to obtain a wideband perfor
mance again.
It has been shown that the Lshaped probe is effective
for exciting linearly polarized patch antennas. For ellipti
cal polarized patch antenna designs, some studies have
been performed on the use of the Lprobe coupling method
[63–65]. For dualband operation, an Lprobecoupled
patch antenna operating at GSM/PCS band has also
been investigated [59]. As for combining the Lprobe
coupling method with other bandwidth enhancement
techniques, there are Lprobecoupled Uslotted patch an
tennas [49,50], together with a parasitic patch [60], or
with a shortcircuiting post [61].
2.5.2. Extensions of the LProbe Coupling Technique.
Further development based on the original Lshaped probe
design was made in order to provide more alternative
coupling methods that maintain the fundamental features
and enhance the ﬂexibility of the design. One of these is an
Lshaped strip coupling method [34]; the concept of the L
shaped probe is extended to accommodate a microstrip
feedline without the process of drilling the substrate and
soldering the Lshaped probe to the feedline. The use of a
microstrip feedline can allow the array design using low
loss microwave substrates and inexpensive PCB sub
strates. The second design is based on the Tshaped probe
coupling [67], which allows the vertical arm to point in a
direction orthogonal to both arms of the original Lshaped
probe. The feeding probe appears as a letter ‘‘T’’ when ex
amined from the side view. Details of the two feeding tech
niques are discussed below.
2.5.2.1. LStripCoupled Patch Antenna. A singlelayer
patch antenna coupled by an Lshaped strip is shown in
Fig. 18. A rectangular patch, with WL=3025mm, is
designed to operate at f
0
=4.5GHz. The air substrate has a
thickness of H=8.3mm. A 50O microstrip feedline is
etched on a microwave substrate (e
r
=2.33) of thickness
h=1.6mm, which is located symmetrically with respect to
the patch. A stepped feedline, referred to as an ‘‘Lshaped
strip,’’ is introduced at the end of the 50O microstrip
feedline. The width of the Lshaped strip is chosen to be
the same as that of the feedline. Moreover, referring
to Fig. 18b, the vertical and horizontal portions of this
Lshaped strip, with dimensions a=5mm and b=10mm,
0
10
copol.
xpol.
−10
−20
−30
−40
3 3.5
with singleT with doubleT
Frequency (GHz)
4 4.5 5 5.5
VSWR
V
S
W
R
G
a
i
n
(
d
B
i
)
6
6
4
5
3
2
1
Figure 20. VSWR and gain of Tprobe coupled patch antenna.
(From Mak et al. [67], reprinted with permission from IEE.)
HPlane 4.5GHz
Copol. Xpol.
0 dB
−20
−10
EPlane
0 dB
−10
−20
Figure 21. Farﬁeld radiation patterns of the twoTprobecou
pled patch antenna. (From Mak et al. [67], reprinted with per
mission from IEE.)
2610 MICROSTRIP ANTENNAS, BROADBAND
respectively, excite the TM
01
mode of the patch by electro
magnetic coupling. Note that the horizontal portion of the
Lshaped strip is supported by a foam material (e
r
E1)
while the vertical portion is located under the patch at an
inset distance S=1mm from the edge of the patch. The
VSWR and gain (copolarization gain and crosspolariza
tion gain) are measured. A wide bandwidth of up to 49%
(VSWRr2) and 7dBi gain are achieved, details can be
found in Ref. 34.
2.5.2.2. TShaped ProbeCoupled Patch Antenna. The
second feeding method derived from the idea of the
Lshaped probe is the Tshaped probe [67]. A single patch
coupled by two identical Tshaped probes is discussed in
this section. The geometry is shown in Fig. 19. The objec
tive here is to improve the radiation performance by can
cellation of unwanted probe radiation [68], caused by the
vertical arm when using only one probe, without reducing
the achievable bandwidth.
S
Plastic post
t
0
Patch
A
D
C
B
0
S
1
W
1
t
2
L
h
Top view
Vertical
Horizontal
Side view
t
3
L
v
h
0
t
1
air
Plastic post
W
0
Lshaped probe
Ground plane
Feed network
(on dielectric substrate, c
r
=2.65)
W
0
Ground plane
Lshaped probe
Patch
(Unit: mm, t
0
=2, t
1
=2, t
2
=2, t
3
=1.5, h
0
=17, L
v
=9, W
0
=180, W
1
=62, L
h
=38, S=15, S
1
=92, 0=45°)
Figure 23. Geometry of a dualpolarized
antenna with a dual feed. (From Wong
et al. [71], r 2004 IEEE.)
Top view
Feed lines
Port 1 Port 2
Patch
Crossedslot
c
c
Ground plane
Port 1
Port 2
Feed lines
Side view
r2
r1
Crossedslot
Figure 22. Conﬁguration of a dualpolar
ized stackedpatch antenna with aperture
coupled feed. (From Edimo et al. [70],
reprinted with permission from IEE.)
MICROSTRIP ANTENNAS, BROADBAND 2611
In Fig. 19, a microstrip feedline, with a quarterwave
length transformer designed to operate at 4.5 GHz, is
etched onto a dielectric substrate of thickness h=
1.6 mm and e
r
=2.33. Two identical Tshaped probes,
T
v
=5.6 mm and T
h
=9mm, are connected to the feedline
below the ground plane and separated by l
g
/2, where l
g
is
the guide wavelength at 4.5 GHz. A rectangular patch of
dimensions 30 25 mm is located at a height of H=
6.6 mm. Also, the position of the patch is such that the
two Tshaped probes are inset the same distance S from
the patch edges and are symmetric with respect to the y
axis, as shown in the top view. In addition, because of the
l
g
/2 separation, the two Tshaped probes simultaneously
excite the same TM
01
mode of the patch.
Figure 20 shows the variations of VSWR with frequen
cy, the broadside copolarization gain and the crosspolar
ization gain in the H plane with yE451, together with that
using one single T probe [67]. It is seen from the ﬁgure
that both antennas have impedance bandwidth
(VSWRr2) of around 40%. Also both antennas have
copolarization gain greater than 6.5dBi across most of
their bands of operation, and the patch with a single T
probe has an average gain of 7dBi. Most importantly, by
using two T probes to excite a single patch, the crosspo
larization in the H plane is significantly suppressed and is
20 dB or more below the copolarization gain between 4.2
and 5.15 GHz.
The farﬁeld radiation patterns at 4.5GHz are shown
in Fig. 21. It can be observed that the copolarization pat
terns in the H plane are symmetric about the broadside
direction and have low backlobe radiation. Also, the cross
polarization patterns in the E plane are well below –25 dB.
Moreover, with the use of the additional outofphase T
shaped probe, the dent in the E plane (when using a single
T) disappears and the Eplane copolarized pattern be
comes more symmetric. The Hplane crosspolarized pat
tern has also been effectively suppressed to less than –
20 dB in all directions.
3. DUALPOLARIZED WIDEBAND PATCH ANTENNA
3.1. DualPolarized ApertureCoupled Stacked
Patch Antenna
A dualpolarized aperturecoupled stackedpatch anten
na is shown in Fig. 22. Two square stacked patches are
fed by two crossed and stacked microstrip lines through
two crossed and nonresonate slots in the ground plane. A
thin substrate is placed between the microstrip lines to
enhance the port decoupling. The use of the stacked
patch design is to obtain a wideband operation [69]. By
applying two orthogonal crossed slots located beneath
the center of the patch, symmetric farﬁeld copolarization
patterns are achieved and high isolation can be obtained
[70]. This structure of dualpolarized patch antenna
yields about 30% impedance bandwidth (for VSWRr2)
and around 28 dB isolation across the operating band
width.
3.2. A DualPolarized LProbeCoupled Patch Antenna
A square patch antenna can be excited with 7451 polar
izations by using only two orthogonal Lshaped probes.
However, this antenna has poor input port isolation due to
strong coupling between the two vertical arms of the L
probes. To tackle this problem, a dualfeed concept is in
troduced here to improve the isolation.
A modiﬁed dualpolarized patch antenna coupled by
two pair of Lshaped probes is shown in Fig. 23. The
A
D C
B
Port 1
Port 2
Figure 24. Outlook of feed network. (From Wong et al. [71], r
2004 IEEE.)
−60
−50
−40
−30
−20
−10
0
1.36 1.45 1.54 1.62 1.71 1.80 1.89 1.98 2.07 2.16 2.25
Frequency/GHz
S
2
1
/
d
B
Antenna without dualfeed

Antenna with dualfeed
Figure 25. Isolation of dualpolarized Lprobefed patch antenna
with and without dual feed. (From Wong et al. [71], r 2004
IEEE.)
1
2
3
4
5
6
7
8
9
10
1.36 1.45 1.54 1.62 1.71 1.80 1.89 1.98 2.07 2.16 2.25
Frequency/GHz
S
W
R
S11
S22
Figure 26. Measured SWR against frequency for dualpolarized
Lprobefed patch antenna with dual feed. (From Wong et al. [71],
r 2004 IEEE.)
2612 MICROSTRIP ANTENNAS, BROADBAND
antenna employs a pair of L probes to excite one polariza
tion, say, ÷451 polarization. Then, another pair excites
the ÷451 polarization. The operating center frequency is
f
0
E1.8GHz. The patch, which is made of aluminum with
dimensions of W
1
=62 mm (0.37l
0
) W
1
=62 mm (0.37l
0
)
t
1
=2 mm (0.012l
0
), is supported by two plastic posts
and is placed h
0
=17 mm (0.1l
0
) above the ground plane.
Four copper Lshaped probe feeds with radius of 1 mm are
located below the patch and are connected to a feed net
work. Each probe is designed to have an input impedance
of 50 O. The horizontal and vertical lengths of each probe
are L
h
=38 mm (0.22l
0
) and L
v
=9 mm (0.05l
0
), respec
tively. As shown in the ﬁgure, the four Lshaped probes
are located at points A, B, C, and D with AB=BC=CD=
Horizontalplane copol
Horizontalplane xpol
Verticalplane copol
Verticalplane xpol
0
30
60
90
120
150
180
(a)
210
240
270
300
330
0 −10 −20 −30 −40
Horizontalplane copol
Horizontalplane xpol
Verticalplane copol
Verticalplane xpol
0
30
60
90
120
150
180
(b)
210
240
270
300
330
0 −10 −20 −30 −40
Figure 27. Measured farﬁeld patterns at
1.8 GHz: (a) port 1; (b) port 2. (From Wong
et al. [71], r 2004 IEEE.)
MICROSTRIP ANTENNAS, BROADBAND 2613
DA. Probes A and C are excited with 1801 phase difference,
while probes B and D are also excited with antiphase.
Probes A and C are used to excite the ÷451 polarization,
while probes B and D are for the ÷451 polarization. Fig
ure 24 shows the outlook of feed network that has a di
electric substrate of e
r
=2.65.
3.2.1. Experimental Results. Figure 25 shows measured
results of S21 for both antennas with or without the dual
feed. It is observed that across the GSM1800 operating
bandwidth from 1.71 to 1.88GHz, the S
21
of the antenna
without the dual feed is about ÷19 dB and the corre
sponding value of the antenna with the dual feed is less
than ÷30 dB. The 10 dB improvement is contributed by
the dual feed suppression technique. The VSWR of the
antenna with the dual feed (Fig. 26) is less than 2 in the
frequency range from 1.56 to 1.99 GHz, and an impedance
bandwidth of 23.8% is obtained. The VSWR is less than
1.5 from 1.66 to 1.94GHz, and a 15% impedance band
width is obtained that is wide enough to cover the
GSM1800 bandwidth.
The measured farﬁeld patterns at 1.8GHz for the an
tenna with dual feed are shown in Fig. 27. The 3dB
beamwidths are 611 and 621 in the vertical and horizon
tal planes, respectively, for the ÷451 polarization. The
corresponding values for the ÷451 polarization are 621
and 601. The backlobe levels for both ÷451 and ÷451 po
larizations are about ÷15 dB, which may be quite high.
This is due to the use of a relatively smaller ground plane
(180 180mm) in the measurement. The antenna has the
average gain of 8.5 dBi.
3.3. DualPolarized LProbe Antenna Array
A basic element of a dualpolarized Lprobe patch antenna
has been presented in Section 3.2. Although the single el
ement can have good selfisolation, in the array environ
ment, the existence of coupling between array elements
results in poor input port isolation. To solve this problem,
we [71] also introduced techniques to enhance input ports
isolation by reducing coupling between elements. To en
hance the isolation of a dualpolarized patch antenna ar
ray, one solution is to use metallic walls surrounding each
element [71]. The function of the walls is to act as a barrier,
which avoids coupling between different elements. The
conﬁguration of a dualpolarized patch antenna array
with surrounding metallic walls is shown in Fig. 28. The
separation between two elements is 133mm (0.8l
0
). For
each element, a square aluminum patch with thickness of
1mm and lengths of 58mm is coupled by four identical
Lshaped probes located at each corner of the patch. The
patch is supported by plastic posts and placed 22mmabove
the ground plane. Each probe with radius of 1mm, vertical
h
2
Patch
t
0
t
1
h
0
h
1
t
0 t
3
Feed network
(on dielectric substrate, c
r
=2.65)
Plastic post
Ground plane
Side view
Lshaped probe
Metallic walls
Port1
W
0
W
2
S
3
Top view
Vertical
Horizontal
Feed network
Port2
(Unit: mm, t
0
=2, t
1
=2, t
3
=1.5, h
0
=22, h
1
=9, h
2
=30, W
0
=180, W
2
=300, S
3
=133)
Figure 28. Conﬁguration of a dualpolar
ized patch antenna array with surrounding
metallic walls. (From Wong et al. [71], r
2004 IEEE.)
2614 MICROSTRIP ANTENNAS, BROADBAND
length of 9mm, and horizontal length of 38mm is placed
13mm below the patch. Each probe is designed to have a
characteristic input impedance of 50O. Metallic walls with
thickness of 2 mm surround the patch and act as an open
ended cavity with dimensions of 12012030mm. The
size of ground plane is 180300mm. All Lshaped probes
are connected to the feed network, and placed underneath
the ground plane. Ports 1 and 2 are the input ports of the
÷451 and ÷451 polarizations, respectively.
3.3.1. Results. Measured results of S
11
, S
12
, and S
22
for
the antenna array are shown in Figs. 29 and 30. As seen in
Fig. 29, the impedance bandwidth is 23.8% from 1.50 to
1.93 GHz, for VSWRr2, and is 12% from 1.66 to 1.89 GHz
for VSWRr1.5. Across the frequency range from 1.71 to
1.88 GHz, the value of S
12
is less than ÷30 dB as shown in
Fig. 30. This indicates that the vertical metallic walls are
able to reduce the coupling between different elements.
The measured farﬁeld patterns for the center frequency
of 1.8GHz are shown in Fig. 31. The 3dB beamwidths are
311 and 641 in the vertical and horizontal planes, respec
tively, for the ÷451 polarization. The corresponding val
ues for the ÷451 polarization are 301 and 651. The
backlobe levels for both ÷451 and ÷451 polarizations
are about ÷25 dB. The performance of this antenna array
is suitable for using as base station antennas in mobile
communication systems.
3.4. DualPolarized Patch Antenna with Hybrid
Feeding Mechanism
A hybrid feeding mechanism for dualpolarized patch an
tennas have been proposed [72,73]. They combine an ap
erture coupling feed [15–17] and a capacitive coupling feed
[37–39] to enhance the isolation between input ports. One
of these hybrid dualpolarized patch antennas [72] is dis
cussed here. The geometry of the antenna is shown in Fig.
32. The square patch, separated from the ground plane by
a thick foam substrate, is fed by two perpendicular strip
lines etched on the feed layer, one through a small circular
hole and the other through a nearresonance aperture
formed in the upper ground plane. These two lines have
−70
−60
−50
−40
−30
−20
−10
0
1.36 1.45 1.54 1.62 1.71 1.80 1.89 1.98 2.07 2.16 2.25
Frequency/GHz
I
s
o
l
a
t
i
o
n
S
2
1
/
d
B
Measured S21
Simulated S21
Figure 29. VSWR of antenna array with me
tallic walls surrounding elements. (From
Wong et al. [71], r 2004 IEEE.)
1
2
3
4
5
6
7
8
9
10
1.36 1.45 1.54 1.62 1.71 1.80 1.89 1.98 2.07 2.16 2.25
Frequency/GHz
S
W
R
Simulated S11
Measured S11
Simulated S22
Measured S22
Figure 30. Isolation of antenna array with
metallic walls surrounding elements. (From
Wong et al. [71], r 2004 IEEE.)
MICROSTRIP ANTENNAS, BROADBAND 2615
characteristic impedance Z
c
=50 O. At port 1, an Lshaped
probe is connected at the end of the line via the small cir
cular hole to cancel the inductance introduced by the thick
substrate. At port 2, an openended stub is needed to tune
out the large inductive reactance delivered to the feedline
by the aperture.
3.4.1. Measured Results. Figure 33 shows the measured
VSWR and input port isolation of the antenna. It is ob
served that, for VSWRr1.5, the frequency range is 1.68–
2.06 GHz, or BW=20.5% for port 1 and 1.59–2.03 GHz,
BW=24.9% for port 2, respectively. Over the bandwidth,
input port isolation is below ÷25 dB. The radiation pat
terns at 1.8GHz for the two ports are shown in Fig. 34.
The 3 dB beamwidths for the Lprobe coupling case (port
1) are 701 and 711 in the E and H planes, respectively. The
corresponding values for the aperturefed case (port 2) are
601 and 721, respectively. The measured gain at each of the
0
330
30
300
270
240
210
180
150
120
90
60
−40 −30 −20 −10 0
0
330
30
300
270
240
210
180
150
120
90
60
40 30 20 10 0
(a)
(b)
Horizontalplane copol
Horizontalplane xpol
Verticalplane copol
Verticalplane xpol
Horizontalplane copol
Horizontalplane xpol
Verticalplane copol
Verticalplane xpol
Figure 31. Measured farﬁeld patterns at
1.8 GHz: (a) port 1; (b) port 2. (From Wong
et al. [71], r 2004 IEEE.)
2616 MICROSTRIP ANTENNAS, BROADBAND
two ports is greater than 6.0 dBi over the bandwidth. It
can be observed that the level of backradiation is high,
which is due to a relatively small ground plane used in the
measurement and can be reduced through the use of a
larger ground plane. The techniques described in Refs. 74
and 75 are also the efﬁcient ways to reduce backradiation.
The stripline feed structure can reduce backradiation con
siderably. On the other hand, backradiation from the ap
erture is trapped between the upper and lower metallic
planes, resulting in the appearance of higher modes,
which enhances the crosspolarization components. It is
seen in Fig. 34 that the crosspolarization is quite high.
The crosspolarization and efﬁciency can be improved with
the use of shorting pins between the two ground planes
[76]. The high crosspolarization can also be overcome in
the array design using a balanced feed mechanism [77].
4. MINIATURE WIDEBAND PATCH ANTENNAS
Extensive studies have been dedicated to size reduction
techniques of microstrip patch antennas [78–81]. Several
techniques have been proposed to reduce the size of the
conventional halfwave patch. One approach involves the
use of expensive highdielectricconstant materials [82].
Other approaches are to use either a shorting (shortcir
cuiting) wall [83,84] or a shorting pin [78–81,85–87] or
various methods to increase the current pathlength ﬂow
ing on the patch [88]. However, a narrow bandwidth of less
than 6% (VSWRr2) is achieved on these designs. In this
section, we will summarize several designs of the minia
ture patch antenna with bandwidth enhancement. Band
width enhancements of small patch antennas with more
than 10% (VSWRr2) impedance bandwidth have been
reported in a number of papers. The designs can be divid
ed into two major categories, one based on the quarter
wave patch (shorting wall) structure and the other one
based on the shorting pin structure.
4.1. Wideband Patch Antennas Design Based on Quarter
Wave Patch Structure
We know that placing a short circuit at the zero electric
ﬁeld line of a halfwave rectangular patch can reduce the
size of the patch since half of the patch can be removed.
The narrow bandwidth of only about 6% (VSWRr2) is
achieved on this quarterwave patch antenna.
Patch
Foam
Ground plane
H1
Ls
Port 1
Lv
Lh
Lprobe Aperture
H3
Wa
Stub
Port 2
d1
Feed substrate c
r
Ground plane
Hole
d2 PW
PL
H2 La
d1 = 30, H2 =1.6, H3 = 18, PL = PW = 59.5, d2 = 8)
(Unit: mm, H1 = 10, Lh = 34, Lv = 12, La = 48, Wa = 1, Ls = 10,
Figure 32. Antenna geometry. (From Guo
et al. [72], r 2002 IEEE.)
−10
−20
−30
−40
I
s
o
l
a
t
i
o
n
(
d
B
)
S
W
R
−50
−60
−70
2.2 2.1 2 1.9
Frequency (GHz)
1.8 1.7 1.6 1.5
1
1.5
2
2.5
3
S11 S22 S21
Figure 33. Measured reﬂection and coupling coefﬁcients. (From
Guo et al. [72], r 2002 IEEE.)
MICROSTRIP ANTENNAS, BROADBAND 2617
4.1.1. QuarterWave Patch with ShortCircuiting Pin.
Figure 35 shows the conventional quarterwave patch an
tenna and the quarterwave patch antenna with shorting
pin. By applying a shorting pin, as much as 20.7% band
width is achieved on the quarterwave patch antenna [89].
The patch size is 30 mm (L) 20 mm (W), with one of the
longer edge shortcircuited to the ground via a shortcir
cuiting wall. The feed position is d
f
=5.5 mm away from
the shorting wall, and the feed radius is r
f
=0.5 mm. The
shorting pin of radius r
s
=0.5mm is placed d
s
=11 mm
away from the open edge.
Figure 36 shows the VSWR of the quarterwave patch
antenna with shorting pin for different substrate thick
ness. The substrate thickness (h) varies from 3mm
(0.375l
0
) to 9 mm (0.837l
0
). The resonant frequency de
creases when the substrate thickness increases. The band
width varies from 9% when h=3 mm (0.375l
0
) at
3.76 GHz to 20.7% when h=9 mm (0.837l
0
) at 2.79 GHz,
where the feed location has varied in order to obtain the
widest bandwidth. Table 1 summarizes the feedprobe lo
cation d
f
, shorting pin location d
s
, feedprobe radius r
f
, and
shorting pin radius r
s
.
Figure 37 shows the measured radiation pattern when
h=9mm. It is observed that the shorting wall leads to
asymmetric radiation pattern and causes the f=01 plane
copolarized radiation pattern (E
y
) tilting away from the
broadside direction, which is similar to the quarterwave
patch antenna. Also, the radiation of the shorting wall and
the shorting pin cause the crosspolarized ﬁeld (E
y
) in the
f=901 plane to increase, and this becomes greater when
the substrate thickness h is increased. This antenna
achieves a gain of more than 3 dBi when the substrate
thickness varies from h=3 to 9 mm. Table 1 summarizes
the performance of the antenna with different substrate
thicknesses.
4.1.2. Shorted TwoLayer Patch Antennas. The design of
this type of antennas is based on the twolayer patch an
tenna. A shorting wall is placed at the line of symmetry
of this twolayer patch antenna as shown in Fig. 38a.
Figures 38b–38d show different conﬁgurations of the
shortcircuited twolayer patch antennas. The differences
among the three antennas are the upper patch shape and
the width of the shorting wall. The impedance bandwidth
achieves more than 8% (VSWRr2) in all designs [90–92].
The projection dimensions of the antenna are, width W
=17.5mm and length L=17.5 mm. The substrate be
tween the upper patch and the lower patch has a thick
ness of h
1
=2 mm. The antennas are excited in the TM
10
mode by a coaxial feed, where the feed position is 13.5 mm
away from the radiating edge. The lower patch is held by a
foam substrate of thickness h
2
=6 mm. The area of the
shorted twolayer patch antenna has been reduced by 92%
when compared with the conventional rectangular patch.
The bandwidth has been enhanced to 8.3%. In cases with
higher resonant frequency, the bandwidth can attain 11%.
(a) (b)
h
L
W
L
W
d
f
d
s
d
f
Shorting pin
Shorting wall
Foam Foam
x
y
Top view
Side view
Shorting wall
[
Figure 35. Geometry of quarterwave patch
antenna with shortcircuiting pin. (From Mok
et al. [89], reprinted with permission from
IEE.)
Port 1 Port 2
Hxpol Expol
Hcopol Ecopol
Hxpol Expol
Hcopol Ecopol
0
30
60
90
120
150
180
210
240
270
300
330
0
30
60
90
120
150
180
210
240
270
300
330
Figure 34. Measured patterns for ports 1
and 2 at 1.8GHz, 10dB/div. (From Guo et al.
[72], r 2002 IEEE.)
2618 MICROSTRIP ANTENNAS, BROADBAND
Figure 39 shows the radiation pattern of the shorted two
layer patch antenna. Again, a relatively high crosspolar
ization level is measured as a result of the radiation of the
coaxial feed. It is also observed that the shorting wall
leads to the asymmetric radiation pattern. As the size has
been greatly reduced, the average gain obtained is only
1.5 dBi for the conﬁguration shown in Fig. 38d.
4.1.3. Other Structures with Shorting Walls
4.1.3.1. LProbe Feed. The Lshaped probe coupling
technique [55] has been demonstrated to enhance the
bandwidth of the traditional patch antennas in Section
2.5. Here, the bandwidth is enhanced by applying the L
probe coupling techniques incorporated with the quarter
wave patch antenna [93], as shown in Fig. 40.
The patch is excited in the TM
01
mode by the Lprobe
fed. The patch size has length L=0.17l
0
, width W=
0.46l
0
, and height H=0.1l
0
. Figure 41 shows the imped
ance bandwidth and gain, where the impedance band
width achieves as much as 39% (VSWRr2) with an
average gain of 7 dBi. The radiation pattern is stable
across the passband and is radiated in the broadside di
rection with a beam squint of B151 due to the asymmetric
structure of the antenna.
4.1.3.2. Capacitive Coupling. This is another feeding
method introduced to enhance the bandwidth of regular
patch antenna, as discussed in Section 2.4.1, and also for
the miniature patch antenna. A circular disk is connected
to the coaxial probe, which acts as a capacitor patch to
compensate the inductance introduced from the feeding
probe [94]. The structure is shown in Fig. 42.
The capacitance from the capacitor patch and the in
ductance from the coaxial probe are in resonance near the
patch resonance. As shown in Fig. 43, there are two res
onances: one from the series resonance by the feed mech
anism and the other one from the patch. Once the
resonances are close to each other, a bandwidth of 35%
(VSWRr2) is achieved. The average gain obtained is 4dBi
across the passband. The cross polarization level is as high
as the copolar value.
4.1.3.3. Stacked Patch. In the stackedpatch techniques
(Section 2.3.2), where the parasitic patch has a close res
onance to the driven patch, the bandwidth is enhanced
because the two resonators have two closely packed reso
nant frequencies. Several designs have been reported on
the bandwidth enhancement of the quarterwave patch
antenna using stacked structure [95–97]. Figures 44a and
44b show two different conﬁgurations. The structure
shown in Fig. 44a consists of a driven lower patch with a
parasitic element placed horizontally on top of it, and a
common shorting wall is connected to both patches
[95,96]. The driven patch is larger than the parasitic
patch. Figure 44b shows another conﬁguration with a tilt
ed parasitic patch where the shorted edge is common to
the shorted edge of the driven patch [97].
Table 1. Parameter Study of the QuarterWave Patch
Antenna with ShortCircuiting Pin
h
(mm)
d
f
(mm)
d
s
(mm)
f
0
(GHz)
Bandwidth
(%)
Maximum
Gain (dBi)
3 5 8 3.75 9.1 3.4
4 3 10 3.88 12.4 3.5
5 2 14 3.6 17.2 3.3
6 3 16 3.49 19.2 3.7
7 3 18 3.17 20.2 4
8 5 20 2.95 20.7 3
9 3 22 2.79 20 3.6
Source: From Mok et al. [89], reprinted with permission from IEE.
0
45
90
135
180
225
270
315
0
45
90
135
180
225
270
315
10dB/Div
[=0° Plane [=90° Plane
E
0
E
[
Figure 37. Radiation pattern of quarterwave patch antenna
with shortcircuiting pin. (From Mok et al. [89], reprinted with
permission from IEE.)
1
2
4
5
3.5
S
W
R
4.5 4.3 4.1 3.9 3.7 3.3 3.1 2.9 2.7 2.5
1.5
2.5
3
3.5
4.5
Frequency (GHz)
h=3 mm
h=4 mm
h=5 mm
h=6 mm
h=7 mm
h=8 mm
h=9 mm
Figure 36. VSWR of quarterwave patch an
tenna with shortcircuiting pin. (From Mok
et al. [89], reprinted with permission from
IEE.) (This ﬁgure is available in full color at
http://www.mrw.interscience.wiley.com/erfme.)
MICROSTRIP ANTENNAS, BROADBAND 2619
Figure 45 shows the VSWR of the quarterwave patch
antenna shown in Fig. 44a. Two resonances come from two
stacked resonators. A maximum of 30% of impedance
bandwidth is obtained when the antenna thickness is
0.082l
0
. Table 2 summarizes the relationship between
the antenna thickness and the impedance bandwidth.
In the second design, based on the stackedpatch struc
ture, the upper patch is tilted by 13.61 and the impedance
bandwidth enhances to 44% (VSWRr2). Both radiation
patterns are stable across the passband with a shift of 451
in the Eplane copolarized ﬁeld. This is due mainly to the
asymmetric structure of the quarterwave patch antenna.
The crosspolarization levels in the Hplane are quite
high, which is typical of the behavior of the quarter
wave patch antenna. Figure 46 shows the radiation pat
tern of the titled upper patch antenna.
315
270
225
180
EPlane
135
90
45
0
315
270
225
180
135
90
45
0
E
0
E
[
10dB/Div
HPlane
Figure 39. Radiation pattern of shortcir
cuited twolayer patch antennas. (From
Chair et al. [91], reprinted with permis
sion from IEE.)
L=17.5
W=17.5
13.5
13.5
L=17.5
W=17.5
13.5
L=17.5
W=17.5
Line of Symmetry
(c) (d)
Top view
Side view
Top view
Side view
Unit: mm
Unit: mm
(b)
(a)
h
2
=6
h
1
=2
h
2
=6
h
1
=2
h
2
=6
h
1
=2
Figure 38. Geometry of shortcircuited two
layer patch antennas. (From Chair et al.
[90,91], reprinted with permission from IEE.)
2620 MICROSTRIP ANTENNAS, BROADBAND
4.2. Wideband Patch Antenna Design Based on the
ShortCircuiting Pin Structure
The use of a shorting pin to adjust the resonant frequency
of a patch antenna is well known [98]. As the design is
similar, but not identical, to the PIFA (planar invertedF
antenna) design, it was shown that the size of the tradi
tional patch antenna can be reduced by adding a shorting
pin near the feed location [81]. The size reduces by more
than 80%; however, narrow impedance bandwidth is its
drawback, and only 1.2% is achieved. With the use of a
thicker substrate, around 0.06l, the bandwidth can be
improved to 6% [99].
4.2.1. Multiple Shorting Pins
4.2.1.1. SingleLayer Patch. In Ref. 99, by using more
shorting pins, the bandwidth of the antenna can be in
creased. Figures 47a and 47b show the antennas structure
with two and three shorting pins, respectively. The shape
of the patch can be either circular or rectangular.
The bandwidth (VSWRr2) with 7.9% and 9.4% is
achieved when two and three shorting pins are used as
shown in Figs. 47a and 47b, respectively. Unlike the single
Lprobefeed
Shorting wall
Side view
Top view
Figure 40. Lprobefed patch antenna with shortcircuiting wall.
(From Guo et al. [93], reprinted with permission from IEE.)
6
5
4
3
2
1
3.0 3.5 4.0 4.5 5.0 5.5
0
2
4
6
8
10
S
W
R
Frequency (GHz)
G
a
i
n
(
d
B
i
)
(i)
(ii)
Figure 41. VSWR and gain of Lprobefed patch antenna with
shortcircuited wall. (From Guo et al. [93], reprinted with per
mission from IEE.)
Top view
Shorting wall
Side view
Circular capacitor patch
Circular capacitor patch
Figure 42. Capacitively fed patch antenna with shortcircuiting
wall. (From Voipio et al. [94], r 1998 IEEE.)
0
−5
−10
−15
−20
−25
−30
1600 1800 2000 2200 2400 2600 2800 3000
S
1
1
(
d
B
)
Frequency (MHz)
Figure 43. Measured return loss of capacitively fed patch an
tenna with shortcircuiting wall. (From Voipio et al. [94], r1998
IEEE.)
(a)
Side view
Top view
Shorting wall
Shorting wall
(b)
Figure 44. Stacked patch structures with shortcircuiting wall.
(From Zaid et al. [96], r 1999 IEEE and Bonefacic et al. [97],
reprinted with permission from IEE.)
Table 2. Parameter Study on Patch Thickness and
Impedance Bandwidth
Patch Thickness Impedance Bandwidth (VSWRr2) (%)
0.039l
0
15.4
0.043l
0
18.6
0.055l
0
23
0.082l
0
29.5
Source: From Zaid et al. [96], r 1999 IEEE.
MICROSTRIP ANTENNAS, BROADBAND 2621
shorting pin, the multiple shorting pins should be posi
tioned further apart. However, the tradeoff is the increase
in patch size. Since the number of the shorting pins in
creases, the performance of the antenna tends toward a
quarterwave antenna behavior. The Hplane crosspolar
ization level is still relative high because the presence of
the shorting pins.
4.2.2. SlotLoaded Patch
4.2.2.1. UShaped Slot. Lee [42] proposed a Ushaped
slot on the patch to enhance the bandwidth of the tradi
tional probe feed patch, as discussed in Section 2.4.2.2. In
Ref. 100, a design combining the U shaped slot and the
shorting pin that can reduce the size and enhance the
bandwidth of the antenna is presented. The structure is
shown in Fig. 48.
This patch antenna can achieve an impedance band
width to 30% (VSWRr2) with substrate thickness of
0.084l
0
. The size of the proposed antenna is only 21% of
the original halfwave patch antenna. As the size of this
antenna has been greatly reduced, the average gain is
9
8
7
6
5
4
3
2
1
1.26 1.404 1.548 1.692 1.836 1.98 2.124 2.268 2.412 2.556 2.7
Frequency (GHz)
V
S
W
R
Simulation
Experimental
Figure 45. VSWR of stacked quarterwave patch antenna. (From
Zaid et al. [96], r 1999 IEEE.)
0
−10
−20
−30
−40
−180 −120 −60 0 60 120 180
R
e
l
a
t
i
v
e
p
o
w
e
r
(
d
B
)
Angel (deg)
Eplane, copolarisation
Eplane, crosspolarisation
Hplane, copolarisation
Hplane, crosspolarisation
Figure 46. Radiation pattern of stacked
quarterwave patch antenna. (From Bone
facic et al. [97], reprinted with permission
from IEE.)
(a) (b)
h
c
r
Feed Feed
Shorting pins
Shorting pins
Top view
Side view
Figure 47. Structures with multiple shortcircuiting pins. (From
Waterhouse et al. [99], r 1998 IEEE.)
Side view
Top view
Shorting Pin
UShaped slot
Feed
Feed
Figure 48. Structure of shortcircuited Uslot patch antenna.
(From Shackelford et al. [100], reprinted with permission from
IEE.)
2622 MICROSTRIP ANTENNAS, BROADBAND
around 2 dBi. Figure 49 shows the radiation pattern of the
antenna. The radiation pattern is stable across the band,
whereas the strongest radiation has tilted 521 away from
the broadside direction. The high crosspolarization level
in the Hplane is obtained due to the radiation of the
shorting pin and coaxial feed.
4.2.2.2. Rectangular Notch. Figure 50 shows the short
ed rectangular shaped slot semidisk microstrip antenna
[101]. The antenna is designed be used in the IMT2000
(1885–2200 MHz), DECT (1880–1900 MHz), and Bluetooth
wireless communication (2400–2483.5 MHz) systems.
With an appropriate rectangular slot size and locations
of the shorting pins, the impedance bandwidth achieves
32.4% (VSWRr2), which covers the frequency range from
1.86 to 2.58GHz. The size reduction is due mainly to the
shorting pin, which acts as an inductive loading. Enhance
ment of the electric ﬁelds by the thin rectangular slot can
further reduce the size of the antenna. With the thin rect
angular slot, the shorted semidisk patch antenna excites
dualfrequency resonance. Therefore, a broad bandwidth
is obtained by optimizing the size of the rectangular slot in
order to merge the two resonances closely packed. Figure
51 shows the return loss and input impedance of the an
tenna. The thickness of the antenna is about 0.07l
1
, and
the size is less than 0.35l
1
, where l
1
is the wavelength at
the lowest frequency (1.86GHz).
5. CONCLUDING REMARKS
This article begins with the introduction of several
common bandwidthwidening methods for microstrip
antennas. Features of each method are discussed with
(a) (b)
0
315
0
315
270 270
225
225
180 180
135
135
90 90
45
45
Figure 49. Radiation pattern of shortcir
cuited Uslot patch antenna. (From Shackel
ford et al. [100], reprinted with permission
from IEE.)
Rectangular notch
Shorting pin
Top view
Side view
Figure 50. Structure of shortcircuited notch patch antenna.
(From Wang et al. [101], r 2002 John Wiley & Sons, Inc.)
2.58GHz
R
e
t
u
r
n
l
o
s
s
(
d
B
)
Input impedance
Return loss
0
−5
−10
−15
−20
−25
1.5 2.0 2.5 3.0
Simulated
Measured
Frequency (GHz)
Simulated
Measured
1.86GHz
2.58GHz
0.1 1
1
2
3
4
25 25
50 50
75
−75
−50 −50
−25 −25
5
0.1
0.1
0 0
0.2
0.2
0.3
0.3
0.4
0.4
0.5
0.5
0.2 2 3 4 5 0.30.4 0.5
Figure 51. Return loss and input im
pedance of shortcircuited notch patch.
(From Wang et al. [101], r 2002 John
Wiley & Sons, Inc.)
MICROSTRIP ANTENNAS, BROADBAND 2623
related illustrations and references. One of the basic
means of increasing bandwidth is to employ electrically
thick substrate with low dielectric constant, but the prob
lem associated with increasing the substrate thickness is
the increased reactance associated with the longer probe
feed required. This limits the achievable bandwidth of mi
crostrip antennas to only a few percent. Several wideband
techniques designed by the authors are then discussed,
including the advanced technique of employing the L
probe coupling technique. The article proceeds to discuss
the wideband performance on two popular antenna de
signs: the wideband dualpolarized patch antennas de
signs and the wideband miniature patch antennas
designs. Throughout the article, experimental results are
presented and referenced.
BIBLIOGRAPHY
1. S. D. Targonski and D. M. Pozar, Design of wideband circu
larly polarized aperturecoupled microstrip antennas, IEEE
Trans. Anten. Propag. AP41(2):214–220 (Feb. 1993).
2. K. L. Wong and T. W. Chiou, Broadband singlepatch circu
larly polarized microstrip antenna with dual capacitively
coupled feeds, IEEE Trans. Anten. Propag. AP49(1):41–44
(Jan. 2001).
3. N. Herscovici, Z. Sipus, and D. Bonefacic, Circularly polar
ized singlefed wideband microstrip patch, IEEE Trans. An
ten. Propag. AP51(6):1277–1280 (June 2003).
4. M. Sumi, K. Hirasawa, and S. Shi, Two rectangular loops fed
in series for broadband circular polarization and impedance
matching, IEEE Trans. Anten. Propag. AP52(2):551–554
(Feb. 2004).
5. G. P. Kurpis and C. J. Booth, The New IEEE Standard Dic
tionary of Electrical and Electronics Terms, 5th ed., IEEE,
Jan. 1993.
6. J. R. James and P. S. Hall, Handbook of Microstrip Antennas,
Peter Peregrinus, 1989.
7. E. Chang, S. A. Long, and W. F. Richards, An experimental
investigation of electrically thick rectangular microstrip an
tennas, IEEE Trans. Anten. Propag. AP34(6):767–772
(June 1986).
8. N. Herscovici, A wideband singlelayer patch antenna,
IEEE Trans. Anten. Propag. AP46(4):471–474 (April 1998).
9. K. L. Wong, C. L. Tang, and J. Y. Chiou, Broadband probe
fed patch antenna with a Wshaped ground plane, IEEE
Trans. Anten. Propag. AP50(6):827–831 (June 2002).
10. D. M. Pozar and B. Kaufman, Increasing the bandwidth of a
microstrip antenna by proximity coupling, Electron. Lett.
23(8):368–369 (April 1987).
11. H. G. Oltman and D. A. Huebner, Electromagnetically cou
pled microstrip dipoles, IEEE Trans. Anten. Propag. AP
29(1):151–157 (Jan. 1981).
12. C. L. Mak, K. M. Luk, and K. F. Lee, Proximitycoupled U
slot patch antenna, Electron. Lett. 34(8):715–716 (April
1998).
13. K. L. Lau, K. M. Luk, and K. F. Lee, Wideband Uslot mi
crostrip patch antenna array, IEE Proc. Microwaves Anten.
Propag. 148(1):41–44 (Feb. 2001).
14. C. Kidder, M. Li, and K. Chang, Broadband Uslot patch
antenna with a proximitycoupled double Pshaped feed line
for arrays, IEEE Anten. Wireless Propag. Lett. 1:2–4 (2002).
15. D. M. Pozar, Microstrip antenna aperturecoupled to a mi
crostripline, Electron. Lett. 21(2):49–50 (Jan. 1985).
16. X. H. Yang and L. Shafai, Characteristics of aperture cou
pled microstrip antennas with various radiating patches and
coupling apertures, IEEE Trans. Anten. Propag. AP
43(1):72–78 (Jan. 1995).
17. V. Rathi, G. Kumar, and K. P. Ray, Improved coupling for
aperture coupled microstrip antennas, IEEE Trans. Anten.
Propag. AP44(8):1196–1198 (Aug. 1996).
18. H. S. Shin and N. Kim, Wideband and highgain onepatch
microstrip antenna coupled with Hshaped aperture, Elec
tron. Lett. 38(19):1072–1073 (Sept. 2002).
19. M. D. van Wyk and K. D. Palmer, Bandwidth enhancement
of microstrip patch antennas using coupled lines, Electron.
Lett. 37(13):806–807 (June 2001).
20. C. Wood, Improved bandwidth of microstrip antennas using
parasitic elements, IEE Proc.H: Microwaves Anten. Propag.
127(4):231–234 (Aug. 1980).
21. G. Kumar and K. C. Gupta, Nonradiating edges and four
edges gapcoupled multiple resonator broadband microstrip
antennas, IEEE Trans. Anten. Propag. AP33(2):173–178
(Feb. 1985).
22. C. K. Aanandan and K. G. Nair, Compact broadband micro
strip antenna, Electron. Lett. 22(20):1064–1065 (Sept. 1986).
23. T. M. Au, K. F. Tong, and K. M. Luk, Characteristics of ap
erturecoupled coplanar microstrip subarrays, IEE Proc. Mi
crowaves Anten. Propag. 144(2):137–140 (April 1997).
24. R. Q. Lee and K. F. Lee, Experimental study of the twolayer
electromagnetically coupled rectangular patch antenna,
IEEE Trans. Anten. Propag. AP38(8):1298–1302 (Aug.
1990).
25. R. B. Waterhouse, Design of probefed stacked patches,
IEEE Trans. Anten. Propag. AP47(12):1780–1784 (Dec.
1999).
26. F. Croq and A. Papiernik, Stacked slotcoupled printed an
tenna, IEEE Microwave Guided Wave Lett. 1(10):288–290
(Oct. 1991).
27. F. Croq and D. M. Pozar, Millimeterwave design of wide
band aperturecoupled stacked microstrip antennas, IEEE
Trans. Anten. Propag. AP39(12):1770–1776 (Dec. 1991).
28. T. M. Au, K. F. Tong, and K. M. Luk, Analysis of offset dual
patch microstrip antenna, IEE Proc. Microwaves Anten.
Propag. 141(6):523–526 (Dec. 1994).
29. S. D. Targonski, R. B. Waterhouse, and D. M. Pozar, Design
of wideband aperturestacked patch microstrip antennas,
IEEE Trans. Anten. Propag. AP46(9):1245–1251 (Sept.
1998).
30. H. Legay and L. Shafai, New stacked microstrip antenna
with large bandwidth and high gain, IEE Proc. Microwaves
Anten. Propag. 141(3):199–204 (June 1994).
31. K. F. Tong, T. M. Au, K. M. Luk, and K. F. Lee, Twolayer
ﬁvepatch broadband microstrip antennas, Electron. Lett.
31(19):1621–1622 (Sept. 1995).
32. W. S. T. Rowe and R. B. Waterhouse, Broadband CPWfed
stacked patch antenna, Electron. Lett. 35(9):681–682 (April
1999).
33. C. H. Cheng, K. Li, and T. Matsui, Stacked patch antenna
fed by a coplanar waveguide, Electron. Lett. 38(25):1630–
1631 (Dec. 2002).
34. C. L. Mak, K. M. Luk, and K. F. Lee, Microstrip linefed
Lstrip patch antenna, IEE Proc. Microwaves Anten. Propag.
146(4):282–284 (Aug. 1999).
35. F. Abboud, J. P. Damiano, and A. Papiernik, Simple model
for the input impedance of coaxfed rectangular microstrip
2624 MICROSTRIP ANTENNAS, BROADBAND
patch antenna for CAD, IEE Proc.H: Microwaves Anten.
Propag. 135(5):323–326 (Oct. 1988).
36. B. M. Alarjani and J. S. Dahele, Feed reactance of rectan
gular microstrip patch antenna with probe feed, Electron.
Lett. 36(5):388–390 (March 2000).
37. G. A. E. Vandenbosch and A. R. Van De Capelle, Study of the
capacitively fed microstrip antenna element, IEEE Trans.
Anten. Propag. AP42(12):1648–1652 (Dec. 1994).
38. G. A. E. Vandenbosch, Capacitive matching of microstrip
antennas, Electron. Lett. 31(18):1535–1536 (Aug. 1995).
39. G. A. E. Vandenbosch, Network model for capacitively fed
microstrip element, Electron. Lett. 35(19):1597–1599 (Sept.
1999).
40. P. S. Hall, Probe compensation in thick microstrip patches,
Electron. Lett. 23(11):606–607 (May 1987).
41. K. M. Luk, Y. W. Lee, K. F. Tong, and K. F. Lee, Experimental
studies of circular patch with slots, IEE Proc. Microwaves
Anten. Propag. 144(6):421–424 (Dec. 1997).
42. T. Huynh and K. F. Lee, Singlelayer singlepatch wideband
microstrip antenna, Electron. Lett. 31(16):1310–1312 (Aug.
1995).
43. K. L. Wong and W. H. Hsu, Broadband triangular microstrip
antenna with Ushaped slot, Electron. Lett. 33(25):2085–
2086 (Dec. 1997).
44. K. F. Tong, K. M. Luk, K. F. Lee, and R. Q. Lee, A broad
band Uslot rectangular patch antenna on a microwave sub
strate, IEEE Trans. Anten. Propag. AP48(6):954–960 (June
2000).
45. K. M. Luk, K. F. Lee, and W. L. Tam, Circular Uslot patch
with dielectric superstrate, Electron. Lett. 33(12):1001–1002
(June 1997).
46. Y. X. Guo, K. M. Luk, K. F. Lee, and Y. L. Chow, Double U
slot rectangular patch antenna, Electron. Lett. 34(19):1805–
1806 (Sept. 1998).
47. K. F. Lee, K. M. Luk, K. F. Tong, S. M. Shum, T. Huynh, and
R. Q. Lee, Experimental and simulation studies of the coax
ially fed Uslot rectangular patch antenna, IEE Proc. Mi
crowaves Anten. Propag. 144(5):354–358 (Oct. 1997).
48. M. Clenet and L. Shafai, Multiple resonances and polarisat
ion of Uslot patch antenna, Electron. Lett. 35(2):101–103
(Jan. 1999).
49. K. M. Luk, Y. X. Guo, K. F. Lee, and Y. L. Chow, Lprobe
proximity fed Uslot patch antenna, Electron. Lett.
34(19):1806–1807 (Sept. 1998).
50. Y. X. Guo, K. M. Luk, and K. F. Lee, Uslot circular patch
antennas with Lprobe feeding, Electron. Lett. 35(20):1694–
1695 (Sept. 1999).
51. W. X. Zhang, C. S. Pyo, S. I. Jeon, S. P. Lee, and N. H. Myung,
A new type of wideband slotfed Uslotted patch antenna,
Microwave Opt. Technol. Lett. 22(6):378–381 (Sept. 1999).
52. F. Yang, X. X. Zhang, X. Ye, and Y. RahmatSamii, Wide
band Eshaped patch antennas for wireless communica
tions, IEEE Trans. Anten. Propag. AP49(7):1094–1100
(July 2001).
53. K. L. Wong and W. H. Hsu, A broadband rectangular patch
antenna with a pair of wide slits, IEEE Trans. Anten. Prop
ag. AP49(9):1345–1347 (Sept. 2001).
54. Z. N. Chen, Experimental investigation of impedance char
acteristics of patch antennas with ﬁnitesize substrates, Mi
crowave Opt. Technol. Lett. 25(2):107–110 (April 2000).
55. K. M. Luk, C. L. Mak, Y. L. Chow, and K. F. Lee, Broadband
microstrip patch antenna, Electron. Lett. 34(15):1442–1443
(July 1998).
56. C. L. Mak, K. M. Luk, K. F. Lee, and Y. L. Chow, Experi
mental study of a microstrip patch antenna with an L
shaped probe, IEEE Trans. Anten. Propag. AP48(5):777–
783 (May 2000).
57. K. M. Luk, L. K. Au Yeung, C. L. Mak, and K. F. Lee, Cir
cular patch antenna with an Lshaped probe, Microwave
Opt. Technol. Lett. 20(4):256–257 (Feb. 1999).
58. C. L. Mak, K. M. Luk, and K. F. Lee, Wideband triangular
patch antenna, IEE Proc. Microwaves Anten. Propag.
146(2):167–168 (April 1999).
59. K. M. Luk, C. H. Lai, and K. F. Lee, Wideband Lprobefeed
patch antenna with dualband operation for GSM/PCSbase
stations, Electron. Lett. 35(14):1123–1124 (July 1999).
60. B. L. Ooi and C. L. Lee, Broadband airﬁlled stacked Uslot
patch antenna, Electron. Lett. 35(7):515–517 (April 1999).
61. C. L. Mak, K. F. Lee, and K. M. Luk, Smallsize wideband
microstrip antenna: the shorted Uslot Lprobe (SUL) patch,
Proc. Millennium Conf. Antennas and Propagation
(AP2000), Davos, Switzerland, April 2000, Vol. 1, p. 361.
62. Y. X. Guo, C. L. Mak, K. M. Luk, and K. F. Lee, Analysis and
design of Lprobe proximity fed patch antennas, IEEE
Trans. Anten. Propag. AP49(2):145–149 (Feb. 2001).
63. W. K. Lo, J. L. Hu, C. H. Chan, and K. M. Luk, Circularly
polarized patch antenna with an Lshaped probe fed by a
microstrip line, Microwave Opt. Technol. Lett. 24(6):412–414
(March 2000).
64. W. K. Lo, J. L. Hu, C. H. Chan, and K. M. Luk, Lshaped
probefeed circularly polarized microstrip patch antenna
with a cross slot, Microwave Opt. Technol. Lett. 25(4):251–
253 (May 2000).
65. W. K. Lo, C. H. Chan, and K. M. Luk, Circularly polarized
patch antenna array using proximitycoupled Lstrip line
feed, Electron. Lett. 36(14):1174–1175 (July 2000).
66. J. S. Jeon, Design of wideband patch antennas for PCSand
IMT2000 service (Technical Feature), Microwave J. (July
2002).
67. C. L. Mak, K. F. Lee, and K. M. Luk, Broadband patch an
tenna with a Tshaped probe, IEE Proc. Microwaves Anten.
Propag. 147(2):73–76 (April 2000).
68. A. Petosa, A. Ittipiboon, and N. Gagnon, Suppression of un
wanted probe radiation in wideband probefed microstrip
patches, Electron. Lett. 35(5):355–357 (March 1999).
69. F. Croq and A. Papiernik, Large bandwidth aperturecou
pled microstrip antenna, Electron. Lett. 26:1293–1294 (Aug.
1990).
70. M. Edimo, A. Sharaiha, and C. Terret, Optimized feeding of
dual polarized broadband aperturecoupled printed anten
na, Electron. Lett. 28:1785–1787 (Sept. 1992).
71. H. Wong, K. L. Lau, and K. M. Luk, Design of dualpolarized
Lprobe patch antenna arrays with high isolation, IEEE
Trans. Anten. Propag. AP52:45–52 (Jan. 2004).
72. Y. X. Guo, K. M. Luk, and K. F. Lee, Broadband dual polar
ization patch element for cellularphone base station, IEEE
Trans. Anten. Propag. AP50:251–253 (Feb. 2002).
73. T. W. Chiou and K. L. Wong, Broadband dualpolarized sin
gle microstrip patch antenna with high isolation and low
cross polarization, IEEE Trans. Anten. Propag. AP50:399–
401 (March 2002).
74. S. D. Targonski, R. B. Waterhouse, D. M. Pozar, Wideband
aperture coupled microstrip patch array with backlobe re
duction, Electron. Lett. 33(24):2005–2006 (Nov. 1997).
75. D. Sievenpiper, L. Zhang, R. F. J. Broas, N. G. Alexopolous,
and E. Yablonovitch, Highimpedance electromagnetic sur
MICROSTRIP ANTENNAS, BROADBAND 2625
faces with a forbidden frequency band, IEEE Trans. Micro
wave Theory Tech. MTT47:2059–2074 (Nov. 1999).
76. P. Brachat and J. M. Baracco, Dualpolarization slotcoupled
printed antennas fed by stripline, IEEE Trans. Anten. Prop
ag. AP43:738–742 (July 1995).
77. K. Levis, A. Ittipiboon, and A. Petosa, Probe radiation can
cellation in wideband probefed microstrip arrays, Electron.
Lett. 36(7):606–607 (March 2000).
78. T. K. Lo, C. O. Ho, Y. Hwany, K. K. W. Lam, and B. Lee,
Miniature aperturecoupled microstrip antenna of very high
permittivity, Electron. Lett. 33:9–10 (Jan. 1997).
79. S. Pinhas and S. Shtrikman, Comparison between computed
and measured bandwidth of quarterwave microstrip radia
tors, IEEE Trans. Anten. Propag. AP36:1615–1616 (Nov.
1988).
80. J. S. Kuo and K. L. Wong, A lowcost microstriplinefed
shortedpatch antenna for a PCS base station, Microwave
Opt. Technol. Lett. 29(3):146–148 (May 2001).
81. R. B. Waterhouse, Small microstrip patch antenna, Electron.
Lett. 31:604–605 (1995).
82. C. L. Tang, H. T. Chen, and K. L. Wong, Small circular mi
crostrip antenna with dualfrequency operation, Electron.
Lett. 33(13):1112–1113 (June 1997).
83. T. Y. Wu and K. L. Wong, On the impedance bandwidth of a
planar invertedF antenna for mobile handsets, Microwave
Opt. Technol. Lett. 32(4):249–251 (Feb. 2002).
84. F. R. Hslao, H. T. Chen, T. W. Chlou, G. Y. Lee, and K. L.
Wong, A dualband planar invertedF patch antenna with a
branchline slit, Microwave Opt. Technol. Lett. 32(4):310–
312 (Feb. 2002).
85. H. K. Kan and R. B. Waterhouse, Small printedwing anten
na suitable for wireless handset terminals, Microwave Opt.
Technol. Lett. 30(4):226–229 (June 1997).
86. S. H. Yeh, S. T. Fang, and K. L. Wong, Dualband shorted
patch antenna for dual ISMband application, Microwave
Opt. Technol. Lett. 32(1):79–80 (Jan. 2002).
87. K. L. Wong, C. L. Tang, and H. T. Chen, Compact meandered
circular microstrip antenna with a shorting pin, Microwave
Opt. Technol. Lett. 15(3):147–149 (June 1997).
88. S. Dey and R. Mittra, Compact microstrip patch antenna,
Microwave Opt. Technol. Lett. 13(1):12–14 (Sept. 1996).
89. W. C. Mok, R. Chair, K. M. Luk, and K. F. Lee, Wideband
quarterwave patch antenna with shorting pin, IEE Proc.
Microwaves Anten. Propag. 150(1):56–60 (Feb. 2003).
90. R. Chair, K. M. Luk, and K. F. Lee, Miniature multilayer
shorted patch antenna, Electron. Lett. 36:3–4 (Jan. 2000).
91. R. Chair, K. M. Luk, and K. F. Lee, Miniature shorted dual
patch antenna, IEE Proc. Microwaves Anten. Propag.
147(4):273–276 (Aug. 2000).
92. R. Chair, K. M. Luk, and K. F. Lee, Radiation efﬁciency
analysis on small antenna by Wheeler cap method, Micro
wave Opt. Technol. Lett. 33(2):112–113 (April 2002).
93. Y. X. Guo, K. M. Luk, and K. F. Lee, Lprobe proximityfed
shortcircuited patch antennas, Electron. Lett. 35(24):2069–
2070 (Nov. 1999).
94. V. Voipio, J. Ollikainen, and P. Vainikainen, Quarterwave
patch antenna with 35% bandwidth, Proc. 1998 IEEE APS
Int. Symp., June 21–26, 1998, Vol. 2, pp. 790–793.
95. J. Ollikainen, M. Fischer, and P. Vainikainen, Thin dualres
onant stacked shorted patch antenna for mobile communi
cations, Electron. Lett. 35(16):437–438 (March 1999).
96. L. Zaid, G. Kossiavas, J.Y. Dauvignac, J. Cazajous, and
A. Papiemik, Dualfrequency and broadband antennas
with stacked quarter wavelength elements, IEEE Trans.
Anten. Propag. AP47(4):654–660 (April 1999).
97. D. Bonefacic, J. Bartolic, and D. Kocen, Stacked shorted
patch antenna with tilted parasitic radiator, Electron. Lett.
37(18):1109–1110 (Aug. 2001).
98. S. S. Zhong and Y. T. Lo, Single element rectangular micro
strip antenna for dual frequency operation, Electron. Lett.
19(8):298–300 (1983).
99. R. B. Waterhouse, S. D. Targonski, and D. M. Kokotoff, De
sign and performance of small printed antennas, IEEE
Trans. Anten. Propag. AP46(11):1629–1633 (Nov. 1998).
100. A. K. Shackelford, K. F. Lee, K. M. Luk, and R. Chair, Uslot
patch antenna with shorting pin, Electron. Lett. 37(12):729–
730 (June 2001).
101. Y. J. Wang, Y. B. Gan, and C. K. Lee, A broadband and com
pact microstrip antenna for IMT2000, DECT, and Bluetooth
integrated handsets, Microwave Opt. Technol. Lett.
32(3):204–207 (Feb. 2002).
MICROSTRIP ANTENNAS, COMPACT
KINLU WONG
National Sun YatSen
University
Kaohsiung, Taiwan
1. INTRODUCTION
Microstrip antennas are usually formed by printing a con
ducting patch on a grounded dielectric substrate, and
show the attractive features of lightweight, low proﬁle,
easy fabrication, and conformability to mounting hosts [1].
In order to meet the miniaturization requirement of mo
bile communication equipment [2], researchers have de
voted much attention to compact microstrip antennas.
Many related compact designs with broadband, dualfre
quency, dualpolarized, circularly polarized, and gainen
hanced operations have been reported since the late 1990s
[3]. This article addresses these innovative designs for
compact microstrip antennas.
In Section 2, promising design techniques for reducing
the size of microstrip antennas at a ﬁxed operating fre
quency are introduced. These techniques include the uses
of a highpermittivity substrate, a shortcircuited patch, a
slotted or meandered patch, a slotted or meandered slot
ted ground plane, and chip capacitor loading. Some re
ported compact microstrip antenna designs are described
and discussed in this section. Section 3 discusses compact
broadband microstrip antenna designs. The design tech
niques for achieving broadband operation with reduced
antenna size are described. The related design techniques
include the use of a shortcircuited patch with a thick air
layer substrate, stacked shorted patches, chip resistor
loading, and slot loading in the radiating patch or ground
plane. Some interesting designs are presented.
Sections 4 and 5 present the designs of compact dual
frequency and dualpolarized microstrip antennas, respec
tively. Compact microstrip antennas with dualfrequency
operation have attracted much attention recently (as of
2004), and many related designs have been available in
2626 MICROSTRIP ANTENNAS, COMPACT
the open literature. The advances in compact circularly
polarized (CP) microstrip antennas are considered in
Section 6. A variety of design examples of the compact
CP microstrip antennas are presented. The designs for
achieving gainenhanced compact microstrip antennas are
included in Section 7. Some design examples are demon
strated. Finally, in Section 8, concluding remarks are
made, and some future studies of compact microstrip
antennas are addressed.
2. DESIGN TECHNIQUES FOR COMPACT
MICROSTRIP ANTENNAS
2.1. Use of a HighPermittivity Substrate
In general, microstrip antennas are a halfwavelength
structure and are operated at the fundamental resonant
mode of TM
01
or TM
10
, with a resonant frequency given by
(valid for a rectangular microstrip antenna with a thin
microwave substrate)
f
0
ﬃ
c
2L
ﬃﬃﬃﬃ
e
r
_ (1)
where c is the speed of light, L is the patch length of a
rectangular microstrip antenna, and e
r
is the relative per
mittivity of the grounded microwave substrate. From (1),
it is known that the radiating patch of the microstrip an
tenna has a resonant length approximately proportional
to 1=
ﬃﬃ
e
_
r
, and the use of a microwave substrate with a
larger permittivity thus can result in a smaller physical
antenna length required at a ﬁxed operating frequency.
For comparison, taking two microstrip antenna designs
with substrates of different relative permittivities of 3 and
27 as an example, the required length of the microstrip
patch of the latter design will be only about onethird of
that of the former one, as predicted from (1); that is, the
required patch size of the latter design is only about 10%
that of the former one. This result suggests that an an
tenna size reduction as large as about 90% can be ob
tained, if the design with e
r
=27 is used instead of the case
with e
r
=3 for a ﬁxed operating frequency.
2.2. Use of a ShortCircuited Patch
Figure 1 shows the geometry of a shortcircuited rectan
gular microstrip antenna with a shortcircuiting (short
ing) pin. For the case that the shortcircuiting pin is
absent, the rectangular microstrip antenna is usually op
erated as a halfwavelength antenna, and the fundamen
tal resonant frequency is given in (1). When there is a
shortcircuiting (SC) pin placed at x =L/2, y =0 (center of
the patch edge) and the feed position is chosen from the
centerline (x axis), the ﬁrst resonant frequency occurs at
about 0.38f
0
[4] (When there is more than one SC pin at
the edge or a SC wall is used, the ﬁrst resonant frequency
occurs close to or at about 0.5f
0
. In this case, the short
circuited microstrip antenna is operated as a quarter
wavelength antenna.). This behavior suggests that the
SCpinloaded rectangular microstrip antenna is operated
with a resonant length less than one quarterwavelength,
and a greater reduced antenna size than the case with a
SC wall can be obtained.
With the SCpinloading technique, the antenna size
reduction is due mainly to the shifting of the null voltage
point at the center of the rectangular patch (excited at
TM
01
mode) and the circular patch (operated at TM
11
mode) to their respective patch edges, which causes the
shortcircuited patches to resonate at a much lower fre
quency. Thus, at a given operating frequency, the required
patch dimensions can be significantly reduced, and the
reduction in the patch size is limited by the distance be
tween the null voltage point in the patch and the patch
edge. For this reason, compared to the case of SCpinload
ed rectangular and circular patches, it is expected that an
equilateral triangular microstrip patch excited at its fun
damental mode (TM
10
mode), where the null voltage point
is at twothirds of the distance from the triangle tip to the
bottom edge of the triangle, will have a much larger
reduction in the resonant frequency, when applying the
SCpinloading technique. For comparison with the pre
diction, a compact triangular microstrip antenna with a
SC pin loaded at the triangle tip was constructed and test
ed [5]. In this study, it is found that, for operating at about
the same frequency, the side length of the compact (short
circuited) triangular microstrip antenna is only about 25%
that of the conventional (regularsize) triangular micro
strip antenna. This means a reduction of about 75% of the
linear dimension of the antenna, or the size of the compact
antenna is only about 7% of that of the conventional mi
crostrip antenna. This reduction in antenna size is greater
than those reported for the compact rectangular and cir
cular microstrip antennas using the same technique.
2.3. Use of a Slotted or Meandered Patch
Compact operation of microstrip antennas can be obtained
by embedding slots in the radiating patch or inserting slits
at the boundary of the radiating patch. In this case, the
effective surface current path in the radiating patch can be
greatly lengthened or meandered, thereby leading to a
large decrease in the antenna’s fundamental resonant fre
quency. Figure 2 shows some slotted or meandered patch
es suitable for the design of compact microstrip antennas.
In Fig. 2a, the embedded slot is a crossslot, whose two
Figure 1. Geometry of a compact microstrip antenna with short
circuitingpin loading.
MICROSTRIP ANTENNAS, COMPACT 2627
orthogonal arms can be of unequal or equal lengths. This
kind of slotted patch causes meandering of the patch sur
face current path in two orthogonal directions and is suit
able for achieving compact circularly polarized radiation
[6,7] or compact dualfrequency operation with orthogonal
polarizations [8]. Similarly, the designs with a pair of bent
slots [9] (Fig. 2b), a group of four bent slots [10] (Fig. 2c),
four 901spaced inserted slits [11] (Fig. 2d), a perforated
square patch or a squarering patch with a crossstrip [12]
(Fig. 2e), a circular slot [13] (Fig. 2f), a square slot [14]
(Fig. 2g), and a group of meandering slits at the boundary
of a circular patch [15] (Fig. 2h) or at the nonradiating
edges of a rectangular patch [16] (Fig. 2i) have been suc
cessfully applied to achieve compact circularly polarized
or compact dualfrequency microstrip antennas.
2.4. Use of a Slotted or Meandered Ground Plane
The technique shown in Section 2.3 for lengthening the
excited patch surface current path to lower the antenna’s
fundamental resonant frequency can also be applied to the
antenna’s ground plane [17]. Figures 3a and 3b show, re
spectively, the geometries of compact microstrip antennas
with a meandered ground plane and a slotted ground
plane. The design in Fig. 3a is intended for achieving lin
early polarized radiation for a compact microstrip anten
na, while that in Fig. 3b can be used to generate dual
linearly polarized waves for compact dualpolarized oper
ation. In addition to a large decrease in the antenna’s fun
damental resonant frequency obtained, similar to the
behavior described in Section 2.3, the impedance band
widths of the compact microstrip antennas shown in Fig. 3
are all greater than that of the regularsize microstrip an
tenna. This behavior is due largely to the slots embedded
in the antenna’s ground plane, which effectively lowers the
quality factor of the antenna. Moreover, an increase in the
radiation efﬁciency of the antenna has also been obtained,
and an enhanced antenna gain has been measured. From
the results obtained in Ref. 17, for the case of using an FR4
substrate, the measured impedance bandwidth of the
design shown in Fig. 3a is about 1.85 times that (3.7 vs.
2.0%) of a corresponding conventional (regularsize) mi
crostrip antenna without embedded slots in the antenna’s
ground plane, and a gain enhancement of 1.6dBi (4.6 vs.
3.0dBi) is also obtained. However, a larger fronttoback
ratio (F/B) was also measured. From the measured results,
the backward radiation is increased by about 7dBi, com
pared to the conventional microstrip antenna. This in
crease in the backward radiation is contributed by the
embedded slots in the ground plane and the decreased
groundplane size in wavelength.
2.5. Use of a Chip Capacitor Loading Technique
With the use of one or more chip capacitors loaded into the
microstrip antenna (see Fig. 4), compact microstrip an
tennas can be obtained [18,19]. For the construction of this
kind of chip capacitorloaded microstrip antenna, a via
hole at a suitable position is ﬁrst drilled through the sub
strate. Then, a chip capacitor of suitable capacitance is
loaded into the via hole, with its two ends connected to the
radiating patch and the ground plane. It has been report
ed that a large frequency reduction of about 65% can be
obtained using the chip capacitor loading technique [19].
Moreover, with the large frequency reduction obtained,
the radiated power from the antenna is almost the same.
This suggests that no significant change in the antenna
Figure 2. Some reported slotted or meandered patches suitable
for the design of compact microstrip antennas.
Figure 3. Geometries of compact microstrip antennas with (a) a
meandered ground plane and (b) a slotted ground plane.
2628 MICROSTRIP ANTENNAS, COMPACT
gain of such a compact microstrip antenna compared to a
conventional microstrip antenna at a ﬁxed operating fre
quency is expected.
3. COMPACT BROADBAND MICROSTRIP ANTENNAS
With the size reduction at a ﬁxed operating frequency, the
impedance bandwidth of a microstrip antenna is usually
decreased. To obtain an enhanced impedance bandwidth,
one can simply increase the antenna’s dielectric substrate
thickness to compensate the decreased electrical thickness
of the substrate due to the lowered operating frequency or
apply the compact broadband design techniques of using
chip resistor loading, a shortcircuited patch with a thick
airlayer substrate, stacked shortcircuited patches, and
slot loading. These compact broadband design techniques
are described in this section.
3.1. Use of a Chip Resistor Loading Technique
Figures 5a and 5b show two promising designs of chipre
sistorloaded microstrip antennas. Two different feed
mechanisms using a probe feed and an inset microstrip
line feed have been implemented and analyzed [20]. This
kind of microstrip antenna design is achieved by replacing
the shortcurcuiting pin in a shortcircuited microstrip
antenna, as seen in Fig. 1, with a chip resistor of low re
sistance (generally 1 O). In this case, with the same an
tenna parameters, the antenna size reduction obtained
can be even greater for a design using SCpin loading de
scribed in Section 2.2. Moreover, the obtained impedance
bandwidth can be increased by a factor of six compared to
a design using shortingpin loading. For the case of using
an FR4 substrate of thickness 1.6 mm and relative per
mittivity 4.4, the impedance bandwidth can reach 10% in
the Lband operation [20]. However, as a result of the in
troduced ohmic loss of the chip resistor loading, the an
tenna gain is decreased, and is estimated to be about
2 dBi, compared to a shorted patch antenna with a SC pin.
3.2. Use of a ShortCircuited Patch with a Thick
AirLayer Substrate
Broadband shortcircuited microstrip antennas fed by
using a probe feed, a capacitively coupled feed [21] or an
Lprobe feed [22] have been reported. Figure 6 shows typ
ical geometries of this kind of shortcircuited microstrip
antenna. The design shown in Fig. 6a uses a simple probe
feed with a long probe pin, which may introduce a large
inductance to the input impedance of the antenna, thus
resulting in poor impedance matching for frequencies
Figure 4. Geometry of a compact microstrip antenna with chip
capacitor loading.
Figure 5. Geometries of chipresistorloaded microstrip anten
nas with (a) a probe feed and (b) an inset microstripline feed.
Ground plane
Shorted patch
viahole Shorting wall
Capacitive
feed
Lprobe or
Lstrip
(a)
(b)
(c)
Figure 6. Geometries of shortcircuited microstrip antennas
with a thick airlayer substrate fed using (a) a probe feed, (b) a
capacitively coupled feed, and (c) an Lprobe or Lstrip feed.
MICROSTRIP ANTENNAS, COMPACT 2629
across the desired operating band. To solve the problem,
the designs shown in Figs. 6b and 6c can be applied.
The design shown in Fig. 6b uses a capacitively coupled
feed, which usually consists of either a circular or a rect
angular metal plate to capacitively couple the electromag
netic energy from the source to the shortcircuited
radiating patch. By further incorporating a capacitive
load to this shortcircuited (SC) patch, it has been dem
onstrated that the overall length of the SC microstrip an
tenna with an airlayer substrate can be reduced from a
quarterwavelength to less than oneeighth wavelength
[21]. Such a design with a volume of 20 8 4mm
3
has
been constructed, and an impedance bandwidth of
178MHz centered at 1.8GHz has been obtained, which
meets the bandwidth requirement of the DCS (Digital
Communication System) cellular communication system.
For the geometry shown in Fig. 6c, the capacitive cou
pling of the electromagnetic energy from the source to the
SC patch is achieved by using an L probe or an L strip. It
has been reported that, with the use of a foam substrate of
thickness about 0.1l
0
(where l
0
is the freespace wave
length of the center operating frequency), an impedance
bandwidth of 39% can be obtained for an Lprobefed SC
microstrip antenna [22]. In this design, the L probe incor
porated with the SC patch introduces a capacitance com
pensating some of the large inductance introduced by the
long probe pin in the thick foam substrate, which makes it
possible for achieving good impedance matching over a
wide frequency range. However, a beam squint of about
15–601 in the Eplane radiation pattern obtained has also
been observed, which is probably attributed to the asym
metric current distribution of the shorted patch, due to the
presence of the SC wall and the L probe.
3.3. Use of Stacked ShortCircuited Patches
The impedance bandwidth of a microstrip antenna is in
general proportional to the antenna volume measured in
wavelengths. However, by using two stacked shortcircuit
ed (SC) patches and making both patches radiate as equal
ly as possible and have a radiation quality factor as low as
possible, one can obtain enhanced impedance bandwidth
for a ﬁxed antenna volume [23]. Figure 7 shows two typ
ical geometries of stacked SC microstrip antennas for
broadband operation. In Fig. 7a, the two stacked SC
patches have different shortcircuiting (SC) walls. By se
lecting a proper distance between the two offset SC walls,
one can achieve a wide impedance bandwidth for the an
tenna. For the geometry shown in Fig. 7b, a common SC
wall is used for the two stacked SC patches. In this case,
impedance matching is achieved mainly by selecting a
proper feed position and a proper distance between the
two SC patches. It should also be noted that, for the two
geometries shown in Fig. 7, the upper SC patch can be
considered to be a parasitic element coupled to the lower
shorted patch, the driven element. In the design, the two
SC patches are usually selected to have approximately the
same, but unequal, dimensions. The substrates between
the two SC patches and between the lower SC patch and
the ground plane can be with air, foam or dielectric ma
terials. Also, a partial SC wall or a SC pin can be used in
place of the offset SC walls or the common SC wall shown
in Fig. 7.
3.4. Use of a SlotLoading Technique
By embedding suitable slots in the radiating patch of a
microstrip antenna, enhanced bandwidth with a reduced
antenna size can be obtained [24]. A typical design with a
slotted equilateraltriangular patch is shown in Fig. 8. It is
found that, by embedding a pair of branchlike slots of
proper dimensions, the ﬁrst two broadside radiation
modes of TM
10
and TM
20
of the equilateraltriangular mi
crostrip antenna can be perturbed such that their reso
nant frequencies are both lowered and close to each other
to form a wideimpedance bandwidth. In this design, the
obtained impedance bandwidth is usually about equal to
or less than 2.0 times that of a corresponding conventional
microstrip antenna.
Ground plane
Shorted patch
viahole
Feed position
Shorting wall
(a)
Ground plane
Shorted patch
viahole
Feed position
Shorting wall
(b)
Figure 7. Geometries of stacked shortcircuited microstrip an
tennas with (a) offset shortcircuiting walls and (b) a common
shortcircuiting wall.
Figure 8. Geometry of a slotted equilateral triangular patch for
the design of compact broadband microstrip antenna.
2630 MICROSTRIP ANTENNAS, COMPACT
4. COMPACT DUALFREQUENCY
MICROSTRIP ANTENNAS
Dualfrequency operation has been an important subject
in microstrip antenna designs [25], and many such de
signs are known. To achieve dualfrequency operation for
reducedsize or compact microstrip antennas, many prom
ising designs have been reported [3]. Some typical com
pact dualfrequency designs are presented in this section.
The two operating frequencies can have the same polar
ization planes [4,26–30] or orthogonal polarization planes
[8,9]. In Section 4.1, two kinds of compact dualfrequency
design techniques of using slot loading and SCpin loading
for the case with same polarization planes are introduced.
In Section 4.2, the designs of compact dualfrequency mi
crostrip antennas with orthogonal polarization planes are
described.
4.1. With Same Polarization Planes
The design using a slotloading technique is ﬁrst de
scribed. Figures 9a and 9b show the promising geometries
of a slotloaded, meandered rectangular microstrip anten
na [26] and a slotloaded, bowtie microstrip antenna [27]
for compact dualfrequency operation with same polariza
tion planes. In the design shown in Fig. 9a, the radiation
characteristics of the antenna operated in the TM
10
and
TM
30
modes are similar and have the same polarization
planes. These two modes can be excited with good imped
ance matching using a single probe feed, and owing to the
meandering of the rectangular patch with slits inserted at
the patch’s nonradiating edges, the resonant frequencies
f
10
and f
30
of the two operating modes can be significantly
lowered, with the radiation characteristics slightly affect
ed. This indicates that a large antenna size reduction can
be obtained by using the present design compared to the
slotloaded patch without slits for ﬁxed dualfrequency
operation.
Based on its compactness in antenna size for a ﬁxed
operating frequency and the fact that its radiation char
acteristics are similar to those of a regularsize rectan
gular microstrip antenna, the bowtie microstrip antenna
is also a good candidate for achieving compact dualfre
quency operation. A typical design obtained by using a
slotloading technique is shown in Fig. 9b. In this design,
a pair of narrow slots is embedded close to the radiating
edges of the bowtie patch. The bowtie patch has a ﬂare
angle of a and a patch width of W. The linear dimension
of the bowtie patch in the resonant direction is ﬁxed to be
L. A pair of narrow slots having dimensions of 1 mm‘
are embedded in the bowtie patch and placed close to the
radiating edges at a distance of 1 mm. A single probe feed
is located along the centerline of the bowtie patch. It is
found that both the TM
10
and TM
30
modes are strongly
perturbed, and that their respective resonant frequen
cies decrease with increasing ﬂare angle of the bowtie
patch. In addition, there exists a feed position for
good impedance matching of the two operating frequen
cies. In addition to the compact dualfrequency operation
obtained, the two operating frequencies have the
same polarization planes, and good crosspolarization
radiation is observed, especially for the Eplane radia
tion [27].
In Fig. 10, the geometries of shortcircuited microstrip
antennas for compact dualfrequency operation with same
polarization planes are presented. These designs are
achieved by applying the SCpin loading technique
[4,28–30]. By incorporating a SC pin in the centerline of
a rectangular microstrip patch and exciting the patch
through a suitable feed position also chosen from the cen
terline (see Fig. 10a), a good matching condition for both
the ﬁrst two resonant frequencies of the microstrip anten
na can be obtained, which makes possible the dualfre
quency operation of such a compact microstrip antenna
through a single coax feed. The obtained frequency ratios
between the two operating frequencies have been reported
to be about 2.0–3.2 [4]. For the case of using a circular
microstrip antenna (Fig. 10b), the frequency ratio ob
tained for the compact dualfrequency operation is about
2.55–3.83 [28], while that for a triangular microstrip an
tenna is about 2.5–4.9 [29]. As for the case of using a bow
tie microstrip antenna (see Fig. 10d), it is possible to
obtain a frequency ratio of about 5.0 [30].
Figure 9. Geometries of (a) a slotloaded, meandered rectangu
lar microstrip antenna and (b) a slotloaded, bowtie microstrip
antenna for compact dualfrequency operation with same polar
ization planes.
MICROSTRIP ANTENNAS, COMPACT 2631
4.2. With Orthogonal Polarization Planes
In this section, rectangular microstrip antennas with
promising embedded slots or inserted slits for achieving
compact dualfrequency operation with orthogonal polar
ization planes [8,9] are described. Figure 11 shows four
promising designs of this kind of compact dualfrequency
microstrip antenna. In the design shown in Fig. 11a, the
embedded slot is a crossslot of equal arm lengths. It is
found that, by increasing the crossslot length, both the
antenna’s ﬁrst two excited resonant frequencies (f
10
and
f
01
) can be lowered with the frequency ratio almost un
changed. Note that the frequency ratio obtained is close to
the aspect ratio of the rectangular patch. It has also been
found that the reduction in the two excited resonant fre
quencies results in a patch size reduction of about 23% for
a given dualfrequency design [8].
With the loading of a pair of bent slots, compact dual
frequency operation for a rectangular microstrip antenna
has been reported [9]. Figure 10b shows the antenna ge
ometry studied. It has been determined that, for ﬁxed
dualfrequency operation, the required patch size is only
about 68% that of the design using a simple patch without
bent slots. This corresponds to a 32% patch size reduction.
The measured radiation patterns of the two operating fre
quencies were also measured. Although the excited patch
surface current paths are significantly altered to lower the
desired resonant frequencies, no special distortion of the
radiation patterns is observed and the crosspolarization
radiation is at an acceptable level.
Figure 10c shows the design of using four inserted slits
at the patch edges of a rectangular patch [3]. The four in
serted slits are of the same length. When the slit length
increases, the resonant frequencies of the ﬁrst two
resonant modes are lowered, similar to the cases with a
crossslot (Fig. 11a) or a pair of bent slots (Fig. 11b).
Experimental results have shown that, for ﬁxed dualfre
quency operation, the patch size of the present design can
be reduced to 56% of that using a simple patch without
inserted slits [3]. With the use of four Tshaped slots (Fig.
11d), a patch size reduction of about 38% has been ob
tained [3].
5. COMPACT DUALPOLARIZED MICROSTRIP ANTENNAS
This section describes the dual linearly polarized opera
tion of a compact square microstrip antenna with a slotted
radiating patch. It is demonstrated that such a slotted
microstrip antenna with a group of four symmetric bent
slots can perform excellent dualpolarized radiation [31],
while the antenna size is significantly reduced for opera
tion at a ﬁxed frequency. Figure 12 shows the geometries
of two possible designs. In Fig. 12a, the design is with bent
slots parallel to the patch’s central lines (denoted design
A), while the design in Fig. 12b is with bent slots parallel
to the patch’s diagonals (design B). The square patch has a
side length of L, and the four bent slots are of the same
dimensions and have a narrow width of 1 mm. The two
arms of each bent slot have the same length ‘, and are
perpendicular to each other. The feed arrangement in de
sign A excites 01 ( ^ xxdirected) and 901 ( ^ yydirected) linearly
polarized waves, while the feed arrangement in design B
radiates 7451 slanted linearly polarized waves.
For design A, experimental results [31] indicate that
the resonant frequency is lowered by about 25% compared
to that of a simple square patch without bent slots. This
lowering in the resonant frequency corresponds to an an
tenna size reduction of about 44%. Furthermore, the an
tenna shows an input isolation (S
21
) of less than ÷39.2 dB
across the obtained impedance bandwidth, which is better
than that ( ÷31.9 dB) of a simple square patch. This
Figure 10. Geometries of shortcircuited micro
strip antennas for compact dualfrequency op
eration with same polarization planes.
2632 MICROSTRIP ANTENNAS, COMPACT
behavior suggests that improved input isolation can be
obtained while resonant frequencies are significantly low
ered for achieving compact operation.
For design B, an experimental study was also conduct
ed. The side lengths of the square patch and microwave
substrate used were the same as those in the study of de
sign A. A reduction of about 16% compared to that of a
simple square patch has been obtained [31]. This lowering
in the resonant frequency corresponds to about 30% re
duction in patch size for design B compared to the design
with a simple square patch for ﬁxed dualpolarized oper
ation.
6. COMPACT CIRCULARLY POLARIZED
MICROSTRIP ANTENNAS
Various CP designs with a compact patch size at a ﬁxed
operating frequency have been reported [3]. Figure 13
shows some reported patches suitable for the design of
compact circularly polarized microstrip antennas. In Fig.
13a, the design is with a crossslot of unequal arm lengths
[32]. By incorporating a probe feed at 451 to the two arms
of the crossslot as shown in the ﬁgure, it has been found
that CP radiation can be obtained at frequencies much
lower than the fundamental resonant frequency of the
antenna without a crossslot. That is, for a ﬁxed operating
frequency, the antenna can perform CP radiation with a
smaller antenna size than can a regularsize microstrip
antenna without a crossslot.
Figure 13b shows a square patch with four inserted
slits for compact CP operation. Two pairs of the slits are
cut in the centerlines of the square patch. Each pair of slits
has equal lengths, but the total lengths of the two pairs of
slits are unequal, which can split the resonant mode of
interest into two orthogonal neardegenerate modes for CP
radiation [11]. By embedding a group of four bent slots in a
cornertruncated square microstrip patch, a singlefeed,
compact circularly polarized microstrip antenna can easily
be obtained. Figure 13c shows the geometry of the micro
strip patch studied. The two arms of each bent slot are
aligned parallel to the centerlines of the square patch. By
increasing the arm length of the bent slots, the fundamen
tal resonant frequency of the slotted square patch is sig
nificantly decreased. By further truncating a suitable size
of the patch corners, the antenna can perform CP opera
tion. Results show that, for a given CP operating frequen
cy, the antenna can have a patch size reduction of more
than 50% compared to the conventional CP design using a
cornertruncated square patch without slots [10].
The method of producing singlefeed CP operation of a
square microstrip antenna by truncating a pair of patch
corners used in Fig. 13c can also be applied to a modiﬁed
square microstrip patch with four inserted slits of equal
lengths to achieve compact CP operation with relaxed
manufacturing tolerances. The geometry of such a micro
strip patch is shown in Fig. 13d. The compactness of the
proposed CP design is achieved because of the inserted
slits at the patch corners of the square patch. These in
serted slits result in meandering of the excited fundamen
talmode patch surface current path, which effectively
lowers the resonant frequency of the modiﬁed square
patch, similar to the design using four inserted slits of dif
ferent lengths at the boundary of a square patch (see Fig.
13b). Instead of using different slit lengths for CP excita
tion, which usually requires a very small slit length differ
ence for the case of a large patch size reduction, the present
design uses the perturbation of truncating a pair of patch
corners, with the inserted slits to be of equal lengths. Ex
perimental results show that the required size of the trun
cated corners for CP operation increases with increasing
reduction in patch size [33]. This behavior gives the
present design a relaxed manufacturing tolerance for
Figure 11. Geometries of slotted rectangular
microstrip antennas for compact dualfre
quency microstrip antennas with orthogonal
polarization planes.
MICROSTRIP ANTENNAS, COMPACT 2633
achieving a compact circularly polarized microstrip
antenna.
Compact CP operation of the circular microstrip anten
na with a crossslot of equal arm lengths in the patch
center and a pair of peripheral cuts at opposite sides of the
patch boundary has been proposed and experimentally
studied [6]. Figure 13e shows the geometry of the
microstrip patch studied. Experimental results show
that the present compact CP antenna has a center fre
quency about 10.4% lower than that of a corresponding
regularsize antenna. This lowering of the center frequen
cy corresponds to a patch size reduction of about 20% by
using the present compact design in place of the regular
size design at a ﬁxed operating frequency.
The CP design for exciting a circularly polarized mi
crostrip antenna using a 50O inset microstripline feed
has been reported [34]. The proposed design is shown in
Fig. 13f. In this design, a narrow slit is inserted at the
patch edge. Because of the combined effects of the inset
microstrip line and the inserted slit, two orthogonal near
degenerate modes for CP radiation can easily be excited.
Also, the excited patch surface currents are meandered in
the present design, and the CP operating frequency ob
tained is greatly lowered; thus, compact CP radiation can
be achieved for the present design.
7. COMPACT MICROSTRIP ANTENNAS WITH
ENHANCED GAIN
Most compact microstrip antenna designs show decreased
antenna gain owing to the antenna size reduction. To
overcome this disadvantage and obtain an enhanced an
tenna gain, several designs for gainenhanced compact
microstrip antennas with the loading of a highpermit
tivity dielectric superstrate have been demonstrated [35–
37]. Figure 14 shows two promising designs of gain
enhanced compact broadband microstrip antenna. In
Fig. 14a, a gainenhanced compact broadband rectangu
lar microstrip antenna is shown [35]. It has been demon
strated that, by adding a highpermittivity superstrate
layer, the antenna gain of a chipresistorloaded micro
strip antenna can be increased to about the same level of a
conventional microstrip antenna. The radiating patch has
Figure 12. Geometries of compact dualpolarized square micro
strip antennas: (a) design with bent slots in parallel with the
patch’s central lines (design A); (b) design with bent slots in par
allel with the patch’s diagonals (design B).
Figure 13. Some reported patches suitable for the design of
compact circularly polarized microstrip antennas.
2634 MICROSTRIP ANTENNAS, COMPACT
a 1O chip resistor loaded at one of the patch edges. For
operating at 1.84GHz, a microwave substrate of e
r
=3.0
and h
1
=1.524mm is covered with a highpermittivity ce
ramic superstrate of e
r
=79 and h
2
=3.05mm. The superst
rate thickness is determined by trying various thicknesses
to optimize the enhanced gain with acceptable radiation
patterns. Conventional and chipresistorloaded rectangu
lar microstrip antennas were fabricated and measured for
comparison. It is found that, at the ﬁxed frequency f =
1.84GHz, the patch size of the resistor and superstrate
loaded microstrip antenna can be reduced to 6.05% times
that of a conventional microstrip antenna. The measured
input impedance bandwidths (10dB returnloss band
width) are 19.6MHz (1.07%) for the conventional anten
na, 145.6MHz (7.91%) for the resistorloaded antenna, and
120MHz (6.52%) for the proposed resistor and superstrate
loaded antenna. The maximum relative received power for
the proposed antenna is ÷46.9dBm; that is, the proposed
antenna has a net power increment of 10.4dB compared
with the resistorloaded antenna ( ÷57.3dBm), and about
1.3dB lower than that ( ÷45.6dBm) of the conventional
antenna (Detailed results of the gain and radiation pat
terns can be found in Ref. 35.). A gainenhanced compact
microstrip antenna is thus achieved.
The design of a highgain, compact microstrip antenna
with CP radiation has been reported [36]. The small size of
the microstrip antenna results from both the highpermit
tivity superstrate loading and slits cut in the patch. In
addition, the antenna gain is enhanced by choosing the
superstrate thickness to be about onequarter wavelength
in the superstrate layer. A typical design of 30% lower an
tenna size (projection area) and 5.2dB higher antenna
gain compared to the conventional CP design has been
obtained. Figure 14b shows the proposed compact circu
larly polarized microstrip antenna with an enhanced gain.
The two pairs of unequal slits ‘
x
and ‘
y
(‘
x
> ‘
y
) are cut in
the patch to split the fundamental resonant mode into two
orthogonal neardegenerate modes for CP operation. The
case with ‘
x
> ‘
y
and feed position shown in the ﬁgure is
for righthand CP operation. When the superstrates of
various thicknesses are loaded onto the microstrip anten
na, its resonant frequency is decreased, and the required
slit lengths and 50O feed position need to be readjusted to
achieve good CP operation. Highpermittivity (e
r2
=79) ce
ramic superstrates of various thicknesses were loaded
onto a compact circularly polarized microstrip antenna.
When no slits and superstrate are present, a square patch
with dimensions 26.2 26.2 mm
2
has a fundamental
resonant frequency of 2697 MHz. For antennas with var
ious superstrate loadings, the slit length ‘
y
is ﬁxed and ‘
x
is adjusted to achieve CP operation. For good impedance
matching, the feed position is slightly varied along the
patch diagonal. The results show that the microstrip an
tenna has a maximum relative received power when the
superstrate thickess is about onequarter of the wave
length of the wave propagating in the superstrate layer.
Compared to a conventional CP design using a nearly
square patch without slits and superstrate loading at the
same operating frequency (2272MHz), the proposed an
tenna has a 30% smaller patch size and a 5.21 dB greater
antenna gain [36]. In addition, the results show that the
CP bandwidth determined from 3dB axial ratio is about
32 MHz or 1.4%, which is larger than that obtained for a
conventional CP design.
8. CONCLUDING REMARKS
Avariety of compact microstrip antennas with broadband,
dualfrequency, dualpolarized, circularly polarized, and
gainenhanced operations have been presented. These
compact microstrip antennas are attractive for applica
tions in presentday wireless communication systems, in
which portable communication products are becoming
smaller and smaller. However, in order to be more com
petitive and attractive in practical applications, the re
duction in the construction cost of compact microstrip
antennas is yet an important topic for intensive studies.
To reduce the construction cost, it is most effective to
use an inexpensive substrate such as the FR4 substrate
for compact microstrip antennas. However, the FR4 sub
strate has a high loss, especially for higher operating fre
quencies, which usually leads to some degradation in
antenna performance. Thus, some new dielectric materi
als with low cost and low loss for the substrate of compact
microstrip antennas should be developed. Such substrate
materials can make possible highperformance, lowcost
compact microstrip antennas, which are very attractive
for practical applications.
Figure 14. Geometries of gainenhanced (a) compact broadband
microstrip antenna and (b) compact circularly polarized micro
strip antenna.
MICROSTRIP ANTENNAS, COMPACT 2635
Another important topic for further studies is the
groundplane effects on compact microstrip antennas. Tra
ditionally, compact or even regularsize microstrip anten
nas that have been studied are usually with the condition
that the ground plane is symmetrically centered below the
microstrip patch. However, for practical applications, the
antenna may be required to be mounted on arbitrary po
sitions on the ground plane of a wireless communication
device, for example, along the edge of the ground plane of
the system circuit board. In this condition, the effects of
the ground plane on the performance of the compact mi
crostrip antenna cannot be ignored. This is because the
ground plane is an integral part of the compact microstrip
antenna, and thus will greatly affect the impedance and
radiation characteristics of the antenna. However, related
studies on this issue are relatively scanty, and more efforts
should be taken in this ﬁeld of compact microstrip anten
na designs.
BIBLIOGRAPHY
1. K. L. Wong, Design of Nonplanar Microstrip Antennas and
Transmission Lines, Wiley, New York, 1999.
2. K. L. Wong, Planar Antennas for Wireless Communications,
Wiley, New York, 2003.
3. K. L. Wong, Compact and Broadband Microstrip Antennas,
Wiley, New York, 2002.
4. K. L. Wong and W. S. Chen, Compact microstrip antenna with
dualfrequency operation, Electron. Lett. 33:646–647 (April
10, 1997).
5. K. L. Wong and S. C. Pan, Compact triangular microstrip an
tenna, Electron. Lett. 33: 433–434 (March 13, 1997).
6. W. S. Chen, C. K. Wu, and K. L. Wong, Compact circularly
polarized circular microstrip antenna with cross slot and pe
ripheral cuts, Electron. Lett. 34:1040–1041 (May 28, 1998).
7. N. Ishitobi and N. Misawa, Microstrip Antenna, U.S. Patent
6,452,552 B1, (Sept. 17, 2002).
8. K. L. Wong and K. P. Yang, Small dualfrequency microstrip
antenna with cross slot, Electron. Lett. 33:1916–1917 (Nov. 6,
1997).
9. K. L. Wong and K. P. Yang, Compact dualfrequency micro
strip antenna with a pair of bent slots, Electron. Lett. 34:225–
226 (Feb. 5, 1998).
10. W. S. Chen, C. K. Wu, and K. L. Wong, Compact circularly
polarized microstrip antenna with bent slots, Electron. Lett.
34:1278–1279 (June 25, 1998).
11. K. L. Wong and J. Y. Wu, Singlefeed small circularly polar
ized square microstrip antenna, Electron. Lett. 33:1833–1834
(Oct. 23, 1997).
12. W. S. Chen, C. K. Wu, and K. L. Wong, Squarering microstrip
antenna with a cross strip for compact circular polarization
operation, IEEE Trans. Anten. Propag. 47:1566–1568 (Oct.
1999).
13. H. D. Chen, A dualfrequency rectangular microstrip antenna
with a circular slot, Microwave Opt. Technol. Lett. 18:130–132
(June 5, 1998).
14. W. S. Chen, Singlefeed dualfrequency rectangular micro
strip antenna with square slot, Electron. Lett. 34:231–232
(Feb. 5, 1998).
15. K. L. Wong, C. L. Tang, and H. T. Chen, A compact meandered
circular microstrip antenna with a shorting pin, Microwave
Opt. Technol. Lett. 15:147–149 (Jun. 20, 1997).
16. J. H. Lu and K. L. Wong, Slotloaded, meandered rectangular
microstrip antenna with compact dualfrequency operation,
Electron. Lett. 34:1048–1050 (May 28, 1998).
17. J. S. Kuo and K. L. Wong, A compact microstrip antenna with
meandering slots in the ground plane, Microwave Opt. Tech
nol. Lett. 29:95–97 (April 20, 2001).
18. M. C. Liang, Frequency Tunable Patch Antenna Device, U.S.
Patent 6,462,712 B1 (Oct. 8, 2002).
19. M. C. Liang, Y. L. Kuo, Y. M. Yen, and W. C. Lai, Capacitor
loaded frequency control scheme for circular patch antenna,
Electron. Lett. 36:1757–1758 (Oct. 12, 2000).
20. K. L. Wong and Y. F. Lin, Small broadband rectangular mi
crostrip antenna with chipresistor loading, Electron. Lett.
33:1593–1594 (Sept. 11, 1997).
21. C. R. Rowell and R. D. Murch, A capacitively loaded PIFA for
compact mobile telephone handset, IEEE Trans. Anten. Prop
ag. 45:837–842 (May 1997).
22. Y. X. Guo, K. M. Luk, and K. F. Lee, Lprobe proximityfed
shortcircuited patch antenna, Electron. Lett. 35:2069–2070
(Nov. 25, 1999).
23. J. Ollikainen, M. Fischer, and P. Vainikainen, Thin dualres
onant stacked shorted patch antenna for mobile communica
tions, Electron. Lett. 35:437–438 (March 18, 1999).
24. S. T. Fang and K. L. Wong, Microstrip Antenna, U.S. Patent
6,400,322 B2, (June 4, 2002).
25. S. Maci and G. Bifﬁ Gentili, Dualfrequency patch antennas,
IEEE Anten. Propag. Mag. 39:13–20 (Dec. 1997).
26. J. H. Lu and K. L. Wong, Slotloaded, meandered rectangular
microstrip antenna with compact dualfrequency operation,
Electron. Lett. 34:1048–1050 (May 28, 1998).
27. K. L. Wong and W. S. Chen, Slotloaded bowtie microstrip
antenna for dualfrequency operation, Electron. Lett.
34(18):1713–1714 (Sep. 3, 1998).
28. C. L. Tang, H. T. Chen, and K. L. Wong, Small circular mi
crostrip antenna with dualfrequency operation, Electron.
Lett. 33:1112–1113 (June 19, 1997).
29. S. C. Pan and K. L. Wong, Dualfrequency triangular micro
strip antenna with a shorting pin, IEEE Trans. Anten. Prop
ag. 45:1889–1891 (Dec. 1997).
30. J. George, K. Vasudevan, P. Mohanan, and K. G. Nair, Dual
frequency miniature microstrip antenna, Electron. Lett.
34:1168–1170 (June 11, 1998).
31. G. S. Row, S. H. Yeh, and K. L. Wong, Compact dualpolarized
microstrip antennas, Microwave Opt. Technol. Lett. 27:
284–287 (Nov. 20, 2000).
32. H. Iwasaki, A circularly polarized smallsize microstrip an
tenna with a cross slot, IEEE Trans. Anten. Propag. 44:
1399–1401 (Oct. 1996).
33. W. S. Chen, C. K. Wu, and K. L. Wong, Novel compact circu
larly polarized square microstrip antenna, IEEE Trans.
Anten. Propag. 49:340–342 (March 2001).
34. W. S. Chen, K. L. Wong, and C. K. Wu, Inset microstripline
fed circularly polarized microstrip antennas, IEEE Trans.
Anten. Propag. 48:1253–1254 (Aug. 2000).
35. C. Y. Huang, J. Y. Wu, C. F. Yang, and K. L. Wong, Gainen
hanced compact broadband microstrip antenna, Electron.
Lett. 34:138–139 (Jan. 22, 1998).
36. C. Y. Huang, J. Y. Wu, and K. L. Wong, Highgain compact
circularly polarized microstrip antenna, Electron. Lett.
34:712–713 (April 16, 1998).
37. H. Nakano and Y. Hirachi, Patch Antenna with Dielectric
Separated from Patch Plane to Increase Gain, U.S. Patent
6,492,950 B2 (Dec. 10, 2002).
2636 MICROSTRIP ANTENNAS, COMPACT
Next Page