588 IEEE TRANSACTIONSON MICROWAvE!

THEORY AND TECHNIQUES, JULY 1975

ference in magnitude between the minimum of J,B and the maximum been obtained [7]. In this short paper, a convenient method for an
oftl=z (note thedifferent scale for~,aand.l=m). accurate solution to the disk resonator is developed using dual
The program m written also allows inclusion of loasy media by integral equations [8]. A major advantage lies in the fact that
only a slight modification since c,, K, and x are already written se capacitance, charge densities, and field functions are determined in
complex variables, Limitation in time and money restricted the terms of a quickly convergent series.
number of results that could be obtained.
FORMULATION
REFERENCES
Consider the geometry shown in Fig. 1 for a circular disk resonator
[i] J. C. Minor and D. M. BoI1e, “Modes in the shielded microstrip on of radius “a,” separated from a ground plane by a dielectric material.
a ferrite substrate transversely magnetized in the plane of the sub-
Without loss of generality, the radius is assumed to be unity. The
strate, ” IEEE Trans. Microwave Theorg Tech. (Special Issue on
MJcrowave Integrated Circuits), vol. MTT-19, PP. 570-577, JUIY 1971. disk is charged to potential VO. The potential functions are considered
[2] T. G. Bryant and J. A. Weiss, “Parameters of zuicrostrip transmis- to be ~1 (r,z) and 4Z(r,z) for z > d and O < z < d, respectively.
sion lines and of coupled pairs of. midrostrip lines, ” IEEE Trans. Because of circular symmetry, the Hankel transforms of these
Microwave Theory Tech. (1968 Symposium Issue), vol. MTT-16, PP.
functions may be defined as
1021-1027, Dec. 1968.
[31 G. Barzilai and G. Gerom, “Modes in rectangular waveguidea
partially filled with ferrite, ” IEEE Trans. Antennas Propagat. d,,,(a,z) = m411,z.(r,s)Jo(ti)rdr. (1)
(Special Supplement—Stimposium on Electromagnetic Theorti), vol. / Q
AP-7, pp. S471-S474,Dec. 1959.
[4] D. M. Belle, “Recent advancesin the theory of Planar waveeuidtng Using the boundary conditions & (a,O) = O and & (a, + co) = O,
structures in ferrite substrates,” in Proc. Sem. Int. Dtsposittfs Hyper- the following expressions for potentials are obtained:
freg. a Ferrite, Dig. (Toulouee, l?rance, Mar. 1972), PP. D IV 1-7.
&(cx,s) = A (~) sinh az, O<s.<d (2)

c$I(%z) = ~(a) em [-CY(Z – d)], z > d. (3)

The unknowns A (a) and B (a) are to be determined from the follow-
ing boundary conditions. At the interface z = d,

d, (r,d) = .$a(r,d) . (4)

Capacitance of a Circular Disk for Applications in In particular,

Microwave Integrated Circuits
A (r,d) = A (r,d) = VO, O<r <l. (5)

SURESH R. BORKAR, MZSMBER,

lEEE, AND Aleoatz=d
RICHARD F. H. YANG, mLLOW, IEEE
a$l(r,d) ~, a@(r,d) o
—— r>l. (6)
az 7’ ‘
Abstract—The quasi-static solution for a circular disk separated
from a ground plane by a dielectric substrate is studied using the Clearly from (4)
dual integral equation approach. A simple expression for equivalent
capacitance is determined. A (a) sinh ad = l?(a) (7)

and using (6) and (7), one cm obtain the following dual integral
INTRODUCTION equations

The analytical study of disk resonators is of considerable impor- ~+ Sinh ad

tance for applications in integrated circuits. In order to determine the
resonant frequency of such structures, it becomes necessary to
!Om [,ginh ad + c, cosh @l
j(a) ●
Jo(cw)da = Vo, O < r <1 (8)

obtain the value of capacitance [1 ~, [2]. Recently, the determina- and

tion of capacitance for a circular disk resonator was accomplished
using computer calculations based on a numerical approach in ‘j(a) .JO(ar)& = O, r>l (9)
spectral domain [1], [3]. Although capacitance was determined /o
readily, it appears that the determination of actual surface charge where
densities and potential functions may warrant inversion of matrices
f(a) = a’A (a) [sinh ad + e, cosh ad]. (lo)
of large orders.
The main complication in such a class of problems arises because of
the mixed boundary conditions involved. Various approaches have
been put forth in the past to circumvent this complexity. Rikitake
[4] used the relaxation method for studying electromagnetic induc- %,+(r-, z)
tion in a plane sheet with a circular aperture. For a two-dmensional
problem in Cartesian coordinates, use was made of conformal
mapping [5]. A method using multiple partial images has been
reported [6]. The capacitance of disk resonator in free space has also

Manuscript received March 13, 1974; revised March 20, 1975.

S. R. Borkar was with the Department of Electrical Engineering,
Illinois Institute of Technology, Chicago, 111. He is now with the Zenith
Radio Corp., Chicago, 111.60639.
R. F. H. Yang was with the Department of Electrical Engineering, Ground plate~
Illinois Institute of Technology, Chicago, Ill. He is now a Consultant
in electromagnetic at 10021 West 146th St., Orland Park, Ill. 60462. Fig. 1. Geometry of the problem.
SHORT PAPERS 589

Equations (8) and (9) are special casea (V = O) of the general dual integrations and using an integral formula [12, p. 692], the total
integral equations charge on the disk is obtained as

r(l). r(n+l)
‘G(p)j(p)&(rp)dP = g(r), 0<? ’<1 (11) Q = 2.eo~ am. (20)
/ o m-O 2r(-n+l)r(nz +2)r(7n+ l)”
. Since r(–n) + = for n = 0,1,2,..0, one gets
f(P)J”(?T)@ = 0, ?>1. (12)
/o
(21)
Such equations have been treated previously in the literature. Under
certain conditions, these equations can be s@ved by employing
Mellin transforms [9] or by converting them to Fredhohn integral It maybe interesting to note that the value of the total charge on the
equations [10]. Here, use is made of a certain dwccntinuous property disk and hence the capacitance is uniquely determined by ao only
of integrals involving Bessel functions. The unknown function ~(a) is while the charge density and potential functions are dependent on all
assumed to be of the form the ~. This implies that the transform of charge density should be of
the form (.ll (a) /a) for obtaining stationary value of capacitance
l-n -1
LOJ.
f(a) = d-kg awlzm+k(a) (13)
m-o The method followed for obtaining the capacitance for the case
d > a is stiller to the aforementioned approach. In this case, (8) is
where k is reel, >0.
rewritten as
This automatically satisfies (9) and the general solution can be
co
determined from (8). For the case a(r) = A w’ atmlicable here, the
G(a)f(ci)Jo(cw) da = (e, + 1) VO (22)
solution is obtein~”as the following series [11, p: i15]: /o

, (14)
(e, + I)a-l sinh ad
G(a) = (23)
[sinh ad + e cosh ad]”
8.=0, forn>O; 60=1
Equation (23) is substituted in (16) and k is chosen to be ~. In this
case, the expression for capacitance is
I% = Lo,n; cn’ = S -k,.%
m-o z8/2#2eo
c=— . m. (24)
Vo
(15)
As before, capacitance is determined by only one coefficient. The
and transform of charge density for numerical calculations is given by
cr-llz ●
JIIZ(a), which is the same as (sin a/a) obtained elsewhere [3].
(2u + 4n + 2k)-’L~,n The capacit&nce values can be calculated by completing the
integrations in (16) and by substituting in (14) and (21) or (24).
.
/
,m
{P’+W(P) -1 )p-’~,+~+~(p) “~,+w+~(p) dp. (16) For d ~ a, the integrations were performed on the digital computer
and for one set of parameters d and e,, the computational time re-
quired was approximately 3 s on UNIVAC 1108. The integrand
From practical considerations, k is chosen such that {p2_W2 (p) -1)
decreases-asymptotically as the square of the argument and the inte-
is es small as possible.
gration was termirmted at the upper limit of 30. Hence the capaci-
tance values obtained for d < a are slightly lower than exact values.
SOLUTION AND DISCUSSION
A more accurate determination of capacitance may require extending
the upper limit of integration.,
Initially, an attempt is made to obtain the solution for the condi-
When d ~ a, the integration for L~,n can be obtained as a quickly
tion d <1. To make pz4W (p) approach 1 for small values of d, (8)
convergent series in powers of (1/d). Thk is indicated in the Appen-
is rewritten as
dix. It can be noted that as d + O, (14) and (21) show that the
. capacitance approaches (~,~0ra2/d) which is the value obtained when
G(cY)~(a)~o(ar) da = ~ VO (17)
/ o fringing fields are neglected. Similarly from (14), (24), and (A-3),
for d ~ co, the zeroth-order term or, in other words, the asymptotic
where value of (Cd/wrcP) is obtained as (4 (c, + 1) d/.s,~a) which is (8d/~a)
for G. = 1.
e, (ad)’1 tanh ad
G(a) = (18) In Fig, 2, the normalized capacitance values are plotted as func-
tanh ad -1- ●, “
tions of dwtance d. The stationary values obtained elsewhere [3] are
also indicated for comparison. -Good agreement is evident. In Fig. 3,
Equation (18) is substituted in (16) and k is chosen to be “1.” The
the resonant frequency using the equivalent value of static capacitance
coefficients are then obtained by performing the integrations in (16).
obtained here is plotted for e, = 2.65. The curves by Mao et al. [13]
Insertion in (14) yields the ~. By retracing through the equations,
and Itoh and Mittra [1] are also drawn. The experimental values
~:(adz) and hence the potential functions Ol,,(r,z) can be deter-
reported in [1] are also shown. It can be noted that excellent agree-
ment is obtained with the experimental results.
In particular, if one is primarily interested in calculating the
capacitance of the disk resonator, the total charge density on the
disk is giVSYl by t~ONGLUDING REMARKS

. The dual integral equation approach seems to be highly suitable

wr(r,d) = @E am ‘Jw+l (a) %lo(ar) &l (19)
/o for determining the potential function and capacitance of a circular
m-a
disk resonator. No inversion of matrices is needed. The computx
which is zero for r >1 as expected. After interchanging the order of tional effort requiiwd is minimal. The complex situation of an arbl-
590 IEEE TRANSACTIONS ONMICROWAVE TEEORYAND TJ3CHNIQUES, JULY 1975

40 1 1 I I I
30 – — This theory
----- Itoh and M#tira [j,[3]

20 -

10 -

“m
~5 -

s
u

2–

I
01 0.5 I 2 5 10

Fig. 2. Capacitance of a circular disk with dielectric substrate.

— This theory
\

8 - \

\
d z 1.5875 mm
~“ 7 – \
o \
.
$ \
~ \

~6
\
\
: \
2 \

ii
5 -

.,~
12
a,mm

Fig. 3. Resonant frequency of disk rescmator.

trary potentiaf function on the disk can also be studied using this If needed, using the relation [17, p. 21]
approach.
(–l)~-kw 1 ‘nu_el z.exp (–u) du
It
involving
may also be possible
mixed boundary
to extend
conditions,
the method
e.g.,
to other
striplines or
problems
micro-
2
n-1 n’
.—
r (s) / ~ 1 +sexp (–w)

strips [14] and circular strip resonators. For the latter case, the
formulation would involve the discussion of triple integral equations. and [12, p. 325]
The solutions for these have been attempted in the literatme [15].
m U’–l exp ( —M-J)
du = r (,) @(p;P;#)
APPENDIX
/ ~ 1 –pexp(–u)

Ford > a, the expression (16) for L~,~can berewrittenm where @(6;P;~) is Lerch’s exponent [18], the second series in (A-3)
m
may be rewritten as

!
sinhpd — coshpd
(47L + l)-%?,. = “ P-l
o “” sinh pd + e, cosh pd

J2m+1/2(P) J2n+1/2(P) dp. (A-1)

To obtain the result ss a series in (l/d), it is desirable to use the

following expansion
sinh pd — cosh pd
= ~~ (-l)’ A ‘exp (-2pds).
‘(=)’[(-’fi)’(2m+2n+
2z+1)’11
sinh pd + E, cosh pd ()er+l
ACKNOWLEDGMENT
(A-2)

Using the formulas for Weber–Schafheitlin integral [16, p. 402], one The authors would like to thank R. Venzon for his advice in com-
gets putational techniques.

(–1)’r(27n+27t +21 +1). r(2m+2n +21 +2) .5 ( _~)8 e ‘ @@n+.+2+1/2).

(4n + l)-’Lm,. = Z2
()C. — 1 ~42d(4d)2@fi+W! r(2m + Z + :). I?(2n + 1 + ~). r(2m -t-2n + 1 +2) .-l ()e,+l
(A-3)
SHORT PAPERS 591

REFERENCES the Green’s function for an open or covered microstrip without using
extended image theoryor a computer algorithm, valid for any b/h
[1} T. Itoh and R. Mittr~, “Analysis of a microstrip disk resonator,”
Arch. Elek. ~bertragung., vol. 27, no. 11, pp. 456-45S, Nov. 1973. and retaining the logarit~lc singularity, is now given. The approach
[2] I. Wolff and N, Knoppik, “Rectangulara ndcircularm icrostripdisk’ used is conceptually similar to that of Kaden [5], in which the
capacitors and resonators,” IEEE Trans. Microroa#e Theory Tech., capacitance of a pair of circular cylindrical wires above a dielectric
VOI. M’I!’J!-22, Pp. 857–S64, Oct. 1974. coated ground plane wes determined.
18] T. Iioh and R, Mittra, ,,A new ~eth~d for ccdeulating the O%P%@~-
tance of a circular disk for microwave integrated circuits, ” IEEE
Trans. Microwave Theory Tech. (Short Papers), vol. MTT-21, pp.
II. FREE-SPACE GREEN’S FUNCTION
431-432, June 1973.
[41 T. Rikitake, “Electromagnetic induction in a perfectly conducting
The free-space Green’s function satisfying
plate with a circular hole, ” J. Geornag. Geoe2ect., vol. 16, pt. 1, pp.
31-36, 1964. vz+o = -fi(z,y) (1)
[5] H. A. Wheeler, “Transmission-line properties of parallel wide strips
by a conformal-mapping approximation, ” IEEE Trans. Microwave is [6]
Theory Tech., vol. MTT-12, pp. 286-2S9, May 1964,
[6] P. 8ilvester, “TEM wave properties of microstrip transmission
lines, ” Pr’oc. Inst. Elec. En@., vol. 115, pp. 43-4s, Jan. 196S. 40 = ~. log (r) (2)
[7] F. Kottler, Handbuch der F’h~sik, pt. XII. Berlin: Springer, 1927,
p. 485.
where
[S] A. Erdelyi and I. N. Sneddon, “Fractional integration and dual
integral equations, ” Can. J. Math., vol. 14, no. 4, pp. 6S5–693, 1962.
[9] E. C. Titchmarsh, Introduction to the Theory of Fourier Integrals.
New York: Oxford, 1937, p. 337.
[10] C. J. Tranter, Bessel Functions with Some Phgsical Applications.
New York: Hart, 1968, p. 91. is the two-dimensional Dirac delta function, cois the permittivit y of
[11] —, Integral Transforms in Mathematical Phgsics. London: free space, and r = (Z2 -I- @)1/2. An integral representation of
Methuen, 1956.
log (z) with z = y +jz and jz = –1 is [7] ‘
[12] I. S. Gradshteyn and I. M. Ryzhik, Tables of Integrals, Series and
Products. New York: Academic, 1965.
[13] S. Mag, S. Jones, and G. D. Vendelin, “Millimeter-wave integrated
‘exp (—A) — exp (—Lz) ~X
log (z) = (. (3)
circuits, ‘‘ IEEE Trans. Microwaoe Theorg Tech. (Special Issue on Jo A
Microwave Integrated Circuits), vol. MT T-16, pp. 455-461, July
196S. where the integral converges provided Re (z) >0. On taking the
[14] M. Sbimasaki and T, Kiyono, “An analysis of microstrip transmis- real part of (3) and substituting for log (r) in (2), the free-space
sion lines by integral equation approach, ” Etectran. Commun.
Japan, vol. 54-B, Feb. 1971.
Green’s function takes the form
[15] J. C. Cooke, “The solution of triple integral equations in opera-
‘@exp (–A I y 1) cos (kc) – exp (–k) ~X
tional form, ” Quart. J. Mech. APP1. Math., vol. 18, pt. 1, PP. 57-72. ~o=L (4)
Feb. 1965. 2rr.o / ~ A
[16] G. N. Watson, A Treatise on the Theor~ of Beesel Functions, Lon-
don: Cambridge, 1945. Note that in (4) the x and y variables are separated so that deriva-
[17] E. C. Titchmarsh, The Theorv of the Riemann Zeta-Function. tives may be easily obtained. Now the integrsJ representation (4)
London: Oxford, 1951.
[18] W. Magnus, I?. Oberhettinger, and R. P. Soni, FormuJas and is used to derive the electrostatic Green’s function, the results of
Theorems for the Special Functions of Mathematical Ph~sics. Berlin: which may be applied to obtain approximately the characteristics
Springer, 1956. of the lowest order ‘(quasi-TEM” mode of a microstrip.

III. OPEN MICROSTRIP GREEN’S FUNCTION

To ~o are added functions satisf ying the two-dimensional Laplace’s

equation, exhibiting the same z behavior as (4) and together with
+0 satisfying the appropriate boundary conditions. Thus with
A Note on Green’s Function for Microstrip reference to Fig. 1, the Green’s function for the open microstrip
may be chosen as
..
SHIMON COEN, STUDENT MEMBER, IEEE
+1 (w) = -J-
2rreo
Abstract—The electrostatic Green% function for the open or cov-
ered microstrip line is obtained by an integral representation of the
free space Green’s function, the results of which may be applied to
obtain approximately the characteristics of the lowest order ‘{quasi-
TEMt’ mode of microstrip. (5)

for y >0, and

I. INTRODUCTION
$2 (X,y) = J--
2rT60er
The electrostatic Green’s function for the open microstrip line
may be obtained from extended image theory as illustrated by
. ‘.f2J) ew [–X(V - A)] +f, (h) exp [A(u – A)]
Silvester [1] and later by Weeks [2]. Weiss and Bryant [3] derived
the covered microstrip Green’s function by using a computer al- /o A
gorithm and this was refined by Farrar and Adams [4]. However,
-00s (b) dx (6)
in [4] the Green’s function is not explicitly given for all values of
b/hi and the logarithmic singularity inherent in their series representat-
ion is not immediately apparent. A direct method of obtaining

—x
Manuscript received September 19, 1974; revised February 1S, 1975.
The author is with the Department of Electrical Engineering, Uni-
versit y of Waterloo, Waterloo, Ont., Canada. ‘.-h
1 b is the separation of the ground planes and h is the height of the
line source from the bottom ground plane. Fig. 1, Green’s function geometry for the open microstrip.