IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-29, NO.

1, JANUARY 1981 63

A Full-Wave Analysis Method for Open Microstrip Structures

TATSUO ITOH, SENIOR MEMBER, E E E , AND WOLFGANG MENZEL

Abstract-A method for analyzing characterisitcs of openmicro- since we solve the problem fortheFouriertransforms of
strip disk structures is presented. The method is based on the spectral unknown current distributions on the strip or aperture elec-
domain immittance matrix approach, and all thewavephenomena tric fields on the substrate, thefar field radiation patterns may
associated with the structures areincorporated. The method provides be simply extracted.
a number of unique and convenient features both in analytical and
numerical phases. A numerical example illustrating the usefulness of 11. FORMULATION OF THE PROBLEM
the method is included. Some numerical results are compared with
experimental data. Althoughthemethod may be applied tootherprinted
circuit structures, we will use the simple rectangular open
microstrip disk in Fig. 1 for the formulation. Themicrostrip is
I. INTRODUCTION located on the substrate of relative dielectric constant er and
I BER of theoretical analyses of printed circuit thickness d. Thestandard spherical coordinate (P, 8, 4) is
A antenna structures appeared recently (for example; [ 11). superimposed on the (x, y , z ) system so that the radiation
NUM
Many of these analyses are based on the quasi-static model of pattern is expressed in terms of 8 and 4. 8 is measured from
printed circuit structures. One typically computes the resonant the z axis, and 4 is the angle measured in the xy plane from
frequency of the hypothetical closed resonator derived from the x axis. The broadside of the antenna 0, axis) is, therefore,
the actual structure by placing a magnetic side wall extending given by 8 = 4 = ~ / 2 The . top surface of the substrate is taken
fromthe microstrip to theground plane. Magnetic current to be y = 0. In conventional analysis we would derive coupled
components computed from this resonator model are used to homogeneous integral equations for
current
distributions
calculate the radiation patterns. The results so obtained indi- J x ( x , zand) J,(x, 2):
cate goodagreement with experimental data, and hence the
theory seems quite useful. j\[ZXX(X -x’, z - zf)Jx(x’, 2 ’ )
Independent of the efforts on antennas, microstrip, and
other printed circuit structures have been used in microwave
and millimeterwave integratedcircuits. As thefrequency of
+
Z,,(X - x’, z - z’)J,(x’, z’)] dx’ dz‘
operation is increased, it has been realized, and now it is
=O
widely known,thatthe quasi-staticanalyses of microstrip
circuit elements are not accurate enough, and a more rigorous
full-wave analysis is required. For instance, as for the analysis p z x ( x - x f ,z - Z‘>J&’, z’>
of microstrip disk resonators, a number of improved theoreti-
cal analyses appeared in thepast [ 21 -[ 41 . It has been reported 1- Z,,(X - X I , z - z‘)J,(x‘, z’)] dx’ dz‘
that the results by the full-wave analysis agree extremely well
with results measured at high frequencies [ 41 . =O (2)
This paper presents a full-wave analysis of the open printed
circuit structuressuch as those encountered inmicrostrip where the integrations are over the strip and the equations are
antennas as an eigenvalue problem with complex eigenvalue valid for (x, z ) on the strip. As is well-known, (1) and (2) state
(resonant frequency). Since all the wave phenomena are incor- that E, and E, are zero on the strip. These equations can be
porated in the analysis, it is believed useful for microstrip type solved forthe complexresonant frequency and thecurrent
antenna applications at higher frequencies. Themethod is components J, and J,, provided all the dyadic Green’s func-
based onthe spectraldomain immittancematrixapproach tion components Z,,, Z,, = Z,,, and Z,, are available which
are functions of the frequency. In the radiating structure, this
developed recently [ 5 ] . In the formulation process the direc-
tions parallel t o the substrate surface are completely separted resonant frequency becomes complex, and its imaginary part
from the normal direction by the use of the equivalent net- accounts for theenergy loss due to radiation.
work for spectral waves. As we will see shortly, the formula- Instead of (1)and(2), we will introducethe spectral
tion process is so simple that it may be accomplished almost domain method. The domain of (1) and (2) is first extended to
by inspection. Themethod is quite versatile andmay be the infinite space and then Fourier transforms are taken. We
obtain
applied to almost anytype of printedstructures including
coplanar, slot,and microstrip-slot resonators.Additionally,
Z,x(,, P>J;(., P) + -%&, P)Jz(a, 8) = m a , P , O ) (3)
Manuscript received February 12,1980; revised June 23,1980. This ZZX(,, P P X @ , 8) + L ( a , P>J,(Q, P) = i&, P,O) (4)
work was supported in part by the U.S. A r m y Research Office under
Grant DAAG29-786-0145. where the Fourier transform is defined via
T. Itoh is with the Department of Electrical Engineering, University
of Texas, Austin, TX 78712.
W. Menzel is with theHochfrequenztechnik, AEG-Telefunken,
D-7900 Uh,Germany.
$(a,8) = /I I[ $(X, ~ ) d ( ~ ~ +dx”dz.
)

0 1981 IEEE
0018-926X/81/0100-0063$00.75
64 IEEE TRANSACTIONSON ANTENNAS AND PROPAGATION, VOL. XP-29, NO. 1, JANUAR? 1981

Y X

Fig. 1. Openmicrostripresonant structure. Fig. 2. Coordinatetransformation.

Notice also tiat the right-hand sides of (3) and (4) are no y = w

longer zero because they are the Fourier transformsof E , and

E, on the substrate surface which are obviously nonzero ex-
y2
Cept-on the strip. _Although(3) and ):( contag four unknowns,
J,, J,, E,, and E,, two of them, E, and E, may eliminated
yTM2
later in the solution process, and we can derive a homogeneous
set of equations which can be solved for the complex resonant
frequency. - - -
The impedance coefficients Z,,, Z,,, and Z,, may be Y
1
y = 0

obtained by writing field distributions in the substrate and air

region and applying the interface condition at theair-substrate
surface. This process is straightforward. However, the applica-
tion of the spectral domain immittance approach [ 5 ] is quite
illustrative forsuch a process as described below. These
c_oefficierts are actually impedance matrix elements that relate
J , and J, with E, and E,. We will make use of equivalent
transmission lines forthe derivation. To this end,we first
recognize that,fromthe definition of the inverse Fourier
transform
all the field components are superpositions over CY and of
inhomogeneous(in y ) waves propagating in the direction 6
from the z axis where 6 = cos-' (P/E), $ = d m : . F_or
each 6, the waves _may-be cJecomposed into TM to y ( E y , E,,
g,) and TE to y(H,,, E,, H , ) where the coordinates u and u
are as shown in Fig. 2 and related with (x, z) via
y~~ t
y = -d
Fig. 3. Spectraldomainequivalentcircuits.
where y t = a* + p2 - eik2 is the propagation constant in the
y direction in the ith region. Also el = E, and e2 = 1. y i is real
for decaying wave and is imaginary for wave propagaing in the
y direction. All the boundary conditions which these TE and
TM waves are required to satisfy are incorpated in the equiva-
lent circuits. Specifically, the electric conductor boundary at
y = -d is represented by the short circuit conditions in the
equivalent circuits, whereas the radiation condition at y + 00
corresponds to infinitely long transmission lines f o r y > 0. The
continuity condition of tangential electric fields at y = 0 can
be seen from the fact that the voltages on both sides of the
junction y = 0 are equal, while thediscontinuity of the
magnetic field on both sides of the strip is represented by the
equivalent current sources J , and Ju for thespectral wave.
In these equivalent circuits, the voltages, E , and E, at y = 0
are related to the currentsources via

u=zsin6--xcos6 (7) Z"(a:,P, 0) = Z O ' ( a : , P)J,@, P) (11)

zJ=zcos6+xsin6. (8) P, 0) = Zoh(a, nJ>J,(a:,
0). (1 2)
Next, we recognize that, if there were a current component It is easy to see that Z o e and Z o h are input impedances look-
J,, it generates only the TM fields, and only the TE fields are ing into theequivalent circuits at y ='O and hence are expressed
generated by the ru. Hence it is possible to draw equivalent as
circuits for the TM and TE fields as shown in Fig. 3. The wave
admittances ineach region are

1
ioh =
Y+h + Y_*

where Y+= and Y - e are input admittances looking down and

ITOH AND MENZEL: METHOD FOR OPEN MICROSTRIF' STRUCTURES
up at y = 0 in the TM equivalent circuit, and Y+h and Y - h are
corresponding quantities in the TE circuit:
The final step of the formulationconsists of the mapping from
the u , u cordiante for spectral waves corresponding to each?.
Becatse of the coordinate transform rgations (7) an$ (8), E.
and E, are linearly related t c E u and E,. Similarly, J , and J,
are superposition of Ju and J v . When these relations are used,
the impedance matrix elements in (3) and(4) are given by
where N , and N , are transforming ratios given by
(Y P
=sin6,Nz= = cos 6.
dGTjF
Notice that f o e and f o h are functions of CY? + -P2, and the
ratio of (Y to P enters only throughN , and N,.
111. DERIVATION OF THE EIGENVALUE EQUATION
We now have two ways to solve the original open microstrip
problem. The conventional one is to use the integral equations
(1) and (2) in the spectral impedance matrix elements given in
(17)-(19). Anotherapproach is the direct useof algebraic
equations (3) and (4) as all the necessary quantities have been
derived in the previous section. As discussed earlier, (3) and
(4) contain four unknowns. Two of them, Ex and E,, how-
ever, are eliminated in thesolution process based onthe
Galerkin's mehod. To this end, the unknownspectral current
components J , andare expanded in terms of linearcombi-
nations of known basis functions
M
m=1
N
In selecting basis functions, we ensure that they are-Fourier
t_ransforms of functions with finite support. That is, Jxm and
Jzn are Fourier transforms of J,,(x, z) and J z n ( x , z) which
are nonzero only on the strip. Let us now substitue (21) and
( 2 2 ) into(3) and (4) andtaketheinnerproducts of the
resulting equations with each of basis function as the standard
Galerkin's procedure calls for.The result is the following
homogeneous matrix equation
M _.
N
2
m= 1
KpmXXcm +
n=l
KpnXZdn= 0, p = 1,2, - 0 , M
M N
m=l n= 1
65
where the typical matrix element is given by
(24)
Since K p n X z , etc., are functions of a frequency, a nontrivial
solution of (23) is derived by seekingacomplex frequency
that makes the determinant of the coefficient matrix of (23)
zero. The corresponding eigenvector (e,, d,) specifies the cur-
rent distributions on thestrip.
Notice thatthe right-handside of (23) is zero. This is
explained by writing a more specific process in the Galerkin's
procedure. When a basis function is multiplied with (3) or (4),
we typically have
(20)
by virtue of the Parseval's relation. The right-hand side van-
ishes because J , and E, are nonzero only over regions of ( x , z)
complementary to each other, that is, J, is zero outside the
strip and E, is zero on the strip.
Equation (23) is exact if M = N-+ 03. However, in practice,
M and N must be finite, and such tmncatio_n introdyes an
approximation. If individual basis functions J x m and J,, are
chosensuch that their inverse Fourier transformsinclude
qualitative natures of the true unknown current distributions,
it is possible to use only a few basis functions to obtain good
results, andthecomputationtime can be reduced.Another
important feature for time saving is to choose the basis func-
tions which are expressed in closed forms. Although this is not
always possible with stripsthat have generalshapes,in the
present rectangular strip one may use J,,(x, z) and J z n ( x , z)
which reasonablyrepresentqualitative natures of thetrue
components and still whose Fourier transforms are analytically
obtainable. Choice of the basis functions have been studied in
a number of recent publications [6] , [ 71 . One possible choice
for J x m and Jzm is
where r = 1, 2 , -..and s = 1, 2 , -.. Any combination of r and s
provides a specific basis function. Therefore, the index rn for
the basis function is given by a combination of r and s. For
instance, we can choose rn = 1 for r = s = 1. The Fourier
transforms of ( 2 6 ) and (27) are readily available as combina-
.66 IEEE TRANSACTIONS ON ANTENNAS
PROPAGATION,
AND VOL. AP-29, NO. 1, JANUARY 1981

tions of the Bessel function of order zero. Notice also that the circuits do not, however, tell any directional information in
correct singularities of the current distributionat the edges are t_he xz plane.Such is provided by the weighting functions
incorporated in (26) and ( 2 7 ) . The current becomes singular of J , and J,.
order l/& at the edges parallel to the current and zero of What is presented in the previous paragraph is rather well-
fi at the edges normal to it,where R is the distance from the known. In the present eigenvalue problem, it is necessary to
edge. considercomplex k’s because only the complex frequency
w, = w, -I-j oi can satisfy the system which loses energy by
IV. FEATURES OF THE METHOD radiation. In the computation, we keep 01 and p t o be real.
The present methodincorporates a number of unique Therefore, surface wave poles arenot crossed. However,
features. Let us describe numerical aspects first and then go to since they are located near theintegration surface, their
moreimportant analyticalaspects. It has been proven in a effects are contained in the formulation.
number of papers on microstrip and other printed line struc- Once the problem-is solved, far field radiation patterns are
turesthatthe spectraldomain method is numerically quite given from Ex and E , as they are Fourier transforms of the
efficient [4] , [ 5 ] . For instance, accurate solutions are ob- electric field:
tained by using a relatively small size matrix (23) such as
M = N = 1 or 2. In contrast in many space domain analyses Eo($, 0 ) a sin ~ i z ( a0),
which typically deal with the coupled integral equations (1)
and ( 2 ) , the size of the matrix to be inverted is quite large if, E@($, e ) a COS Q COS e.E,(O, p) + sin Q ~ , ( Q I0),
for instance, the point matching is used. On the other hand,if
Galerkin type procedures are used in the space domain, it is with
necessary first to perform the inverse Fouriertransforms,
CY = k sin 0 cos Q
whichare extremely timeconsuming, to get the Green’s
functions and then to carry out the convolution integrals. 0 = k cos 0 (29)
The time saving feature of the spectral domain method is
caused essentially by two elements. First, it is possible to use and ,?, 2,given by (3), (4),and (21), ( 2 2 ) . In the E-plane
basis functions which incorporate certain qualitative natures of (e
= n/2)
the true current distributions such as the edge condition. This
eliminates use of a large numberof basis functions to accurately
represent unknown current distributions. Second, in the
spectral domain approach, we deal with algebraic rather than
Eo 0: x
n
d,Z,(o, k COS e)Jan(O,k cos e ) (30)

integral equations. Itis not necessary t o carry out the convolu- and in the H-plane (19 = ~ / 2 )
tion integrals, and the spectral domain impedance functions
are given in closed forms. Of course, the price we have to pay Eo 0: sin Q d,.Zz,(k COS Q, O)J,n(k COS @,O). (31)
for such features is that the inner products to compute matrix n
elements K p m X X ,etc., are infinite integrals. However, the
V. NUMERICAL RESULTS
integrands of theseintegralsdecay as fastas Or2 and pd2
when the basis functions like those obtained from (26) and A Fortran program has been made to perform the calcula-
(27) are used, and hence the inner products can be computed tions described above. The integrationoveLthe CY - 0 plane has to
withoutmuchdifficulty. Nevertheless, thenumber of itera- be done numerically.As& 0: l/@andZ,, o:p2/d-for
tions for seeking a root should be minimized. Often, this can krge agreements QI and 0 (and similar relations are valid for J,,
be as small as five to ten when a good subroutine is used. Z,,, and?,,), the integrationconverges with l / ( c $ ? d m .
Besides, since the basis functions are notfunctions of the Numerical andexperimental evaluations have been done
frequency, they need to be evaluated only once, contributing with RT-Duriod substrates with E, = 2.35 and a thickness of
to the reductionof computation time. d = 1.58 mm. Table Isummarizes computedresonant fre-
Let us now turn our attention to more.analytica1 features quencies fortwo differentmicrostrip sizes. For anarrow
of the method. We go back to-( 17)-( 1%) and study their con- element ( I = 1 cm, w = 0.2 cm), the real part of the computed
struction. We recognize that Z o e and Zo* are obtained from resonant frequency does not change much by using different
the equivalent circuits which extend in the y direction. The number of basis functions. The imaginary part indicating the
$formation in the x and z direction come in Z x x , Z x zand , energy lost by radiation is extremely small.
Z,, only throughthe transforming ratios N , and N,. Also Experiments have been performed for the patch antenna
theinformation on the microstrip is contained only in the made of a microstrip of I = 1 cm and w = 1.5 cm. First, the
basis functions. This is not special at all because Z,,, etc., are antenna element was fed by a lOO-.Q microstrip line connected
theFourier transform of the Green’s functions which are directly (or inductively) to the element. The measured reso-
independent of the sourceshapes and relatedonly tothe nantfrequency c f , ) was 8.75 GHz. Next, we createda gap
location and the direction of the source. between the antenna element and the feed line to get a weak
When t_he reson_ances in the y-direction occur, the denomi- capacitive coupling. The measured resonantfrequency then
nator of Zoe or ZO* becomes zero as a surface wave pole is was 8.29 GHz which results in adifferenc of 1.4 percent,
encountered. The radiation phenomena are associated with the compared with_ the theoretical value obtained by using two
+
imaginary part of 72. Therefore, the visible region (a2 0’ < T,,,, and one J x m , that is, 8.41 GHz. This can be due to the
k2) of the ab plane is responsible for radiation, and the sur- uncertainty of E,. (k2 percent) or the substrate thickness d.
face wave poles occur on circles with radius d w - b e t w e e n In any case, it is felt that more basis functions have to be
&k and &k. The invisible region ?C+ O2 > k2 is respon- taken into account forthis very wide element.
sible for the stored energy in the near region. The equivalent The bandwidth of the capacitively coupled element was 0.6
N ITOH
FOR AND
METHOD
MENZEL: 67

GHz in the experiment compared with the theoretical one of

0.54 GHz. The latter was derived from the definitionof Q:

TABLE I
CALCULATED RESONANT FREQUENCY e=-. U P
(32)
20i
Structure No. of basis functions Resonant frequency C G E d
a t 4 44 Jz jx
The differencebetween the resonantfrequencies in the
1 0.2 1 a 9.616 +j 0.171 x inductive and capacitive coupling cases seems to be due mainly
1 1 9.608 +j (1.128 x l o 4 to the relatively high inductance of the impedance step when
the element is coupled directly to the feed line. This shows the
necessity of getting moreinformationonthediscontinuity
1 0 8.586 + j 0.28 effects involved in microstrip antenna circuits.
The radiation patterns in the E- and H-planes were meas-
1 1.5 8.05 1 1 ’ + j 0.21 sured with the capacitively coupled element at 8.29 GHz. Fig.
2* 1 8.41 +j 0.272 4,shows the measured and computed (with two Jzm and one
J z m ) patterns.Theexperimental E-plane patterncontains
E= = 2.35 , d = 1.58 mn large ripples caused by the small substrate size and radiation
from the feed.
w i t h different x-dependences

VI. CONCLUSION

0 We have presented a full-wave method for analyzing open

printedcircuitstructures. Theformulation is based onthe
W
F spectral domain immittance matrix derived from the spectral
2
m
domain equivalent circuits. In the solution process, Galerkin’s
D
method is used. The method contains several attractive fea-
= 10 turesfromboth analyticalandnumerical points of view.
P
I
C Although amicrostrip structure is treated in the paper, the
0
m method itself is quite general and is applicable to other types
I

D
m
of printed structures such as the coplanar and slot lines. Also,
L

it is quite straightforward to extend the method to structures

20 involving stratified substrates, several radiating elements, and
even thosecontaining conductor elementsat different inter-
faces of stratified substrate [5] . It will be constructive to test
the present method with simpler basis functions. Such studies
may be useful for more complicated strip structures for which
30
less sophisticated basis functions need to be used.
DEGREE
(a) ACKNOWLEDGMENT
0
The authors thank Dr. B. Rembold of AEG-Telefunken, for
W
rn
r- his giving them the opportunity to complete the
present work.
3
-
<
m REFERENCES
g 10 [I] J. R. James and G . J. Wilson, “Microstrip antennas and arrays. Pt.
2
-i I-Fundamental action and limitations,” IEEJ. Microwaves, Opt.,
c
0
m Acoust., vol. 1, pp.157-164,Sept.1977.
I [2] T. Itoh and R. Mittra, “Analysis of a microstrip disk resonator,”
0
m
I
Arch. Elek. Ubertragung., vol. 27, pp.45-58. NOV.1973.
PI I . Wolff and N. Knoppik,“Rectangular and circularmicrostrip
20
disk capacitors and resonators,” IEEE Trans. Microwave Theory
Tech., vol. MTT-22, pp. 857-864,Oct.1974.
[41 T. Itoh,“Analysis of microstripresonators,” IEEE Trans.Mi-
crowave Theory Tech., vol. MTT-22, pp. 946-952, NOV.1974.
[SI -, “Spectral
domain immittance approach for dispersion
characteristics of shieldedmicrostrips with tuning septums,”
30
presented at the 9th
European
Microwave Conf., Brighton,
DEGREE England,Sept.17-21,1979.
(b) [61 R . H. Jansen, “High-speed computation of single and coupled
microstrip parameters including dispersion, high-order modes, loss
Fig. 4. Radiation patterns. - theory, --- measured. (a) EAane. (b) and finite strip thickness,” IEEE Truns. Microwave Theory Tech.,
H-plane. VOl. MTT-26, pp, 75-82, Feb. 1978.
(71 E. F. Kuesterand D. C . Chang, “An appraisal of methods for
computation of the dispersion characteristics of open-microstrip,”
lEEE Trans. Microwave Theory Tech.,vol. MTT-27, pp. 691-694,
July 1979.
68 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-29, NO. 1, JANUARY 1981

TatsuoItoh (S’69-M’69SMS74) received the Engineers of Japan, Sigma Xi, and Commissions B and C of USNC/URSI.
Ph.D. degree in electrical engineering from the He is a Registered Professional Engineer in the State of Texas.
University of Illinois, Urbana, in 1969.
From September 1966 to April 1976 he was
with the Electrical Engineering Department, WolfgangMenzel wasbornin Brilon-Wald,
University of Illinois. From April 1976 to August West Germany, on December 10, 1948. He
1977 he was a Senior Research Engineer in the received the Dipl. Ing. degree from the Technical
Radio Physics Laboratory, SRI International. University of Aachen, Germany, in 1974, and the
Menlo Park, CA. From August 1977 to June 1978 Dr. Ing. degree from the University of Duisburg,
he was an Associate Professor at the University Germany, in 1977.
of Kentucky, Lexington. In July 1978 he joined From 1974-1979, he
was working atthe
the faculty at The University of Texas at Austin, where he is nowan University of Duisburg on microstrip circuit
Associate Professor of Electrical Engineering and Director of the problems. Since 1979, he
has joined AEG-
Microwave Laboratory. During the summer 1979, he was a Guest Telefunken in Ulm. Germany, working on in-
Researcher at AEG-Telefunken, Ulm, West Germany. tegrated millimeter-wave components and planar
Dr. ltoh is a member of the Institute of Electronics and Communication antennas

Analysis of a Circular Microstrip Disk Antenna with a Thick Dielectric Substrate

WENG CHO CHEW AND JIN AU KONG, SENIOR MEMBER, IEEE

Absrruct-The problem of a circular microstrip disk excited by a of view it is important to analyze both thin- and thick-sub-
probe is solved using rigorous analysis. The disk is assumed to have strate microstrip antennas and to investigate their respective
zero thickness, and the current on the probe is taken to be uniform. regions of validity in theory and applicability. Existing tech-
Using vector Hankel transforms the problem is formulated in terms niques for calculating theinputimpedance of microstrip
of vector dual-integral equations, from which the unknown current disk antennasareonly valid forthin dielectric substrates
can be solved for. Dueto thesingularnature ofthecurrent dis-
[2], [4],[ 8 ] . This is partly due to the use of the free-space
tributionarising fromprobe excitation,the direct application of
Galerkin’s basis function expansion method gives aslowly convergent Green’s function in calculating theradiation loss andthus
result. Therefore the singularpart of the current is removed since the neglecting the dielectric effects [2], [4],andpartly to the
singularity is known u priori. The unknown current to he solved for is perturbative arguments employed [ 81. Intheperturbative
then regular and tenable to Galerkin’s methodof analysis.It is shown approach the microstrip disk antenna is thought of as a per-
that this analysis agreeswith the single-modeapproximation when the turbation of a magnetic wall cavity, implying its high-Qfac-
dielectric substrate layer is thin, and that it deviates from the single- tor. The high-Q factor allows the approximation of the field
mode approximation whenthe substrate layer is thick. Excellent inside the microstrip antenna witha single mode.
agreement of both the computed real and imaginary parts of the input In this paper we shall present a method to calculate the in-
impedance with experimental data is noted. The radiation patterns put i m p e d a n c e of a circular microstrip disk under a probe ex-
and the current distributions on the disk are also.presented.
citation which is viable beyond the range of validity of the
single-mode approximation in [ 81. We shall make use of vec-
I. INTRODUCTION tor Hankel transforms (VHT’s) in our formulation. The VHT
is devised by the authorsin [ 71 where its associated properties
T HE RECENT advent of the microstrip antenna has aroused
interest in both the experimental and theoretical analysis
of the antenna [ 11-[ 81. Due to the inherentnarrow bandwidth
are also derived. It is found that the use of VHT’s simplifies
the otherwise complicated analysis.
of the microstrip antenna, it is desirable to find methods to
increase the bandwidth. It is well-known that, by increasing 11. FORMULATION
the thickness of the dielectric substrate, the bandwidth of the We wish to obtain the field solution of a circular printed-
antenna can be increased [ 71, [ 81. From a theoretical point circuit antenna under a probe excitation. To do this we first
have to obtain the field in the upper half-space due to a finite-
Manuscript received May 1, 1980; revised August10,1980.This radius vertical probe embedded in the first layer of a stratified
work was supported by Schlumberger-Doll ResearchCenterand the half-space as shown in Fig. 1.
Joint Services Electronics Program under Contract DUG-29-78C-0020.
The authors are with the Research Laboratory of Electronics, D e Using the dyadic Green’s function formalism we can show
partment of Electrical Engineering and Computer Science, Massachu- that the 2 component of the electric field, due to a vertical
setts Instituteof Technology, Cambridge, MA 02139. probe of length L and radius R with uniform current I in an

0018-926X/81/0100-0068$00.75 0 1981 IEEE