and Ka.

hn Kurss: Xinimum-Scattering Antennas 6‘71

standing
support
mast shielded
the
cables
to
the REFERENCES
re-
ceiver. [I] C. Crampton, et al., “The errors in bearings on a high-frequency
direction-finder caused by reradiation from a nearby vertical
Sample H F calibration curves are reproduced in Fig. mast,” J.Inst. Elect. Engrs. (Lomion),vol. 94, pt. IIIA, pp. 815-
11 wherespacedloopandsimplelooperrorperformance 8% 1947.
[2] -, “The performance of high-frequency direction-finders in
is compared. I t is evidentthattheimprovement in various types of H.M. ships,” ibid., pp. 798-80;, 1947,
bearing accuracy is approximately 2 to 1 and the multi- [3] C. Crampton, “Naval radiodirection-finding, ibid., pp. 132-
153, 1947.
valued regions have been eliminated. The calibration [4] C. Crampton, etal,, ‘lsome principles underlying the design of
plotted xvith automatic equipment at the aerialsystemsfor high-frequency radio direction-finders in H.31.
ships,” ibid. vol. 95, pt. 111, pp. 43745%. Sovember 1948.
time data were taken with about six minutes elapsed [51 R. T. p. iviipple and (Miss) M. Moore, Preliminav theoretical
tirnebetlxTeensimple loop and spaced loop curves on the report on errors caused in bearings by reradiation from a vertical
conductor placed near a spaced loop direction finder,” Admiralty
frequency.
same Signal Establishment, Great Britain, Rept. $1585, December 1943.

Minimum-Scattering Antennas
11’. K. m,SEhTORM3MEER, IEEE, AND H. KURSS, SENIOR hlEhZBER, IEEE

Abstract-Antennas with identical radiation patterns can differ general than Dicke’s treatment. I t includes antennas
in the manner and extent to which they modify an incident wave, with any number of ports and antennas n-ithnonre-
i.e., in the way they scatter. This paper enlarges on the work of
Dicke [l]on this subject, employing, as hedoes, a scattering represen-
ciprocal components.
tation based on spherical modes. However, the approach is more This paper departs, as does Dicke in his work on this
physical and includes antennas with nonreciprocal components and topic, from the conventional description of the electro-
with N (local) ports. magnetic fields about an antenna. These fields will be
A canonical minimum-scattering antenna is de6ned as one which specified as a superposition of spherical modes. As the
becomes “invisible” when the accessible waveguide terminals are
open-circuited. The scattering matrix of such an antenna is shown
calculation of radiated power per unit solid angle, per-
to be unique once N arbitrary orthogonal radiation patterns have haps the central quantity in antenna practice, is gen-
been specified. Neither an impedance nor an admittancematrix erally cumbersome in terms of these spherical modes,
for such an antenna exists. this type of specification has found little application in
It is demonstrated for a large class of antennas terminated by the antenna literature except by way of intermediate
matchedreceivers thatthescattered power is generally greater
thanthe absorbed power, equality being attainedfor rninimum-
mathematical manipulations. Herein,
the
spherical
scattering antennas. modes serve as basis for a network (scattering) repre-
sentation in termsof which general results are obtained
I , IKTRODUCTIOS by linear algebra.
HE DESCRIPTION of the In Section 11, a canonical N-port minimum-scattering

T
electromagnetic
properties of an antenna in terms of radiation antenna is defined essentially by the property that such
(and receiving) patterns is incomplete; antennas an antenna becomes “invisible” when the AT accessible
with identical amplitude and phase patterns can differ waveguide ports are open-circuited. Such an antenna is
in themannerandextentto which theymodifyan shown to be unique (in the sense t h a t all its scattering
incident wave, i.e., in the way they scatter. The possible properties, orequivalently, thecompletescattering
importance of this point in connection with evaluation matrix is uniquely determined) once its 1V orthogonal
of antennas designed tohaveidenticalpatternswas radiation patterns have been specified. Keither the im-
first appreciated by R. H. Dicke [l],who proposed “to pedance matrix nor the admittance matrix of such an
see what can be done in the way of differentiating be- antenna exists. The canonical antenna is generalized by
tween a good and bad antenna on the basis of scatter- cascading it with a “transparent” 2N-port, the result
ing.” The approach taken in this paper is more physical, beingtermedsimply a minimum-scatteringantenna.
mathematicallysimplerand atthesametimemore This process modifies the radiation patterns obtained on
excitation of a particular port of the canonical antenna
Manuscript received May 11, 1964; revised March 11, 1965. This without affecting the manner in which i t scatters.
work was sponsored by the Air Force Office of Scientific Research T h e physical significance of the minimum-scattering
of the Office of Aerospace Research, the department of the Army, antenna concept is examined in Section I11 from several
-4rrrry Research Office, and the department of the Navy, Office of
Naval Research, under Grant .-2F-AFOSR-453-63; and by the Na- points of view. The quantity scattered power is intro-
tional Science Foundation under Grant GP-2761.
\V. K. Kahn is with the Polytechnic Institute of Brooklyn, N. Y . duced as a convenient definite measureof the scattered
H. Kurss is with Adelphi University, Garden City, N. Y . fields. For a large class of antennas, it is demonstrated
that the scattered power is generally greater than the bs = ap. (1)
absorbed power, equality being attained for minimum-
Thegeneral(linear)antennacannowbecharac-
scattering antennas. This is the fundamental fact which
terized in a scattering formalism
justifies the minimum-scattering terminology. The ap-
proach adopted in this paper is compared with that of
Dicke and the theorems demonstrated identified as the
appropriategeneralizationsofDicke'sresultstoan-
tennaswithmorethan a single portandcontaining lossless antenna conservespolver, a+a = b'b, and con-
nonreciprocal components. sequently, the scattering matrixS is unitary,
The implications of the concepts developed for array st;S = I. (3)
theory are discussed briefly.
As the first step towards an ideal antenna, one demands
I I. CANONICAL ~,IIXIMIX-SCATTERING AKTENNAS circuitry such that each accessible port be reflectionless
The network description of a general antenna is in- and decoupled from any of the other accessible ports,
dicated schematically in Fig. 1. The ports at the left, 1.e.,
representingthelocal,accessiblewaveguideports of
sa,= 0. (4
the antenna, are numbered from 1 to N ; the ports on I t follows that the ith columnof Soa corresponds to the
the right, LV+l, . . . , representing the electromagnetic radiation pattern of the antenna when excited a t t h e
fieldsoutsideasphereenclosingtheantenna,arein- ith port alone. Condition (4)will hereafter be assumed.
finite in number. At the N accessible ports normalized, Component equations equivalent to (2) are
rms-voltage,incidentandreflectedwavephasorsare
defined and collected as elementsof the column matrices
b, = +SaBaB, (54
a, and b,, respectively,intheconventionalmanner. bo = Soaa, f Ssaas. (jb)
Outsideasphereenclosingtheantennatheelectro- Correspondinglythe lossless constraint (3) yieldsthe
magnetic field can be represented as a superposition of equations
a complete,denumerablyinfiniteset of orthogonal S3a+S~rr = I,, (64
modefunctionsivithappropriatecomplexcoefficients
[ l ] , [2]. The modes occur in pairs as incoming and out-
sfl
, +spp = 0 (6b)
goingsphericalwaves(analogoustotheincidentand S,o+S,p +
Spp+Spo = I p a , (64
reflectedwaves of uniformwaveguideterminology). where Iaa is an A T dimensional and Ioa a n infinite di-
The corresponding modal coefficients,which completely mensionalunitmatrix.
specify the field, are ordered in an arbitrary manner to E'or an evaluation of the (unnecessary) disturbance
formthe(infinitedimensional)columnmatricesand or scattering of incident fields by an antenna, the ob-
bo, and normalized' so that a8+a3and bB+bB are the in- vious standard is provided by the absence of any an-
cident and reflected powers, respectively. tenna, i.e., free space. In the absence of any antenna
the input scattering amplitudes,1 through N , are mean-
f-" ingless, andcorrespondingentriesinthescattering
+

[( '2I )
(3)
p

I
a

T$;;
$3

(N+31

..
matrix are, therefore, omitted. The remainder
scattering matrix of free space is

s
of the

.. O
-
- (
___I-__
I ' IBB ) (7)
(N1 1
(4
in virtue of (1).
A canonicalminimum-scatteringantennahaving
Fig. 1. A general antenna. specified radiation patterns Sg, could, therefore, tenta-
tively be defined as a lossless device approaching (l),
In the absenceof any antenna, the origin of spherical i.e. (i), a s closely as possible, suitable adjustments at
coordinates is j u s t an ordinary pointof space. At sucha the accessible ports of the antenna beingpermitted.
point incident power cannot be absorbed, and conse- The power u-hich is received by the antenna and reaches
quently must be reflected. Somewhat surprisingly then, the local ports is t o be reflected back into the antenna
this ordinar). point of space corresponds to some sort andreradiatedin a manner 11-hich approximatesthe
ofperfectreflectorforsphericalmodes. The require- reflection that u-ould have taken place in the absenceof
any antenna. The sense in lvhich (1) is t o be approached
ment that the fields themselvesremainfinite a t this
need not be discussed a t this point, since, for any given
point yields in particular that [ l ]
S3, consistent with (6a) there exists a unique lossless
antenna whichonopencircuit a t ports 1 through N
1 Sote: a- denotes the conjugate of the transpose of a. scatters exactly as required. The last statement will now
1965 Kahn and Kurss: Minimum-Scadterin.gAntennas . 673

be proved, and the scattering matrix of the canonical is that neither an impedance nor an admittance matrix
antenna found. corresponding to S, exists.
On open circuit a t ports 1 through N Consider the effect of inserting a lossless, transparent
[3] 2 ~ - p o r t ,
b, = aa. (8)
Employing (5) and (8) to eliminate a , and b , in favor
of ag and ba, one obtains
+
bp = (S,TS~S Spdaa. (9) in series with a canonical antenna S,, as illustrated in
Fig. 2. The resultant antenna has the scattering matrix
The scattering relation (9) coincides with the standard,
free space (I) or, equivalently,(7) only if
+ s,, = r3,,
sBasaS (10)
which may be solved for S B ~Equations
. (6a)-(6c) then Antennascharacterized by- scattering matrices of the
serveto fix S uniquely,sinceonpremultiplying (IO) form(18) will betermedsimplyminimum-scattering
by .Sa,+ one obtains, in viewof (6a) and (Sb), antennas. This terminology is justified by the fact that
such an antenna is rendered invisible by a particular
Safi = sa,+. (11)
lossless termination of the accessible waveguide ports.
Ithastherefore'beendemonstratedthattheunique Since the /3/3 component of the scattering matrix de-
scattering matrix of the canonical minimum-scattering termines the n-ay an antenna scatters, as discussed in
antenna with an arbitrarilyprescribed set of orthogonal the next section, S, and 3, scatter identically. In the
radiation patterns which form the columnsof Sg, is special case

Clearly such an antenna is reciprocal?

S,T = s, (13)
if and only if each entry in Sgais purel). real.
Certain formal properties of S, are pointed o u t next, (31
a fulldiscussion of its physicalsignificancebeingre- .. .. ..
served until Section 111.
T h e eigenvalues of S, are defined by
(S, - u,I)s, = 0, (14) Fig. 2. Transparent 2X-port in series with a
canonical minimum-scattering antenna.
where s n is the eigenvector corresponding to the eigen-
value u,,.Since S, is unitary, its eigenvalues lie on the
111. CHARACTERISTICS O F ~ ~ I N I b I U M -
I
unit circle, uni = 1. Since S, is Hermitean, its eigen-
SCATTERIKG -~TENNAS
values are real. Consequently,
un = * 1, 12 = 1, 2, 3, * . * . (15)
Thescatteringmatrix
scattering
antenna having
of canonical
A
a minimum-
' specified (orthogonal)
radiationpatterns wasconstructedinthepreceding
The multiplicity of the eigenvalue - 1 can be shown
section. The construction was found to be unique. In
to be precisely N. For, expanding(14) with u,,= - 1, one
this section, the physicalsignificance of this constructis
obtains with some manipulation
examined from several points of view.
S,a+Sd = - .S?l,; (164 137hen electromagneticwaves a areincident on an
antenna with scattering matrix S as indicated in (5),
s , 3 S $ a + ~ n=
~ S~B. (16b) a, at the local ports and aB from the surrounding space,
Equation (16b) implies t h a t sng is in the X-dimensional the reflected waves b can be interpreted as follows. T h e
sub-space spanned by the N columns of SB=while (16a) term ba= Suoag is the wave absorbed,S;7aaais the wave
associates a unique snn with each s,~. In view of (6a) radiated, and S ~ is ~the wave
a ~ reflected back into the
eachcolumn of S8- is a solution of (16b). There are surrounding space. The scatteredfield f p due to the an-
therefore exactly A T linearly independent eigenvectors tenna is defined as the difference between the reflected
of S, with eigenvalues - 1. fields when an antenna is present and when no antenna
An incidental result which follows directly from (15) is present. The former field is Ssaas and the latter field
is 1,~Jaa.As a definite measure of this scattered field
* ST denotes the transpose of S.
674 IEEE TRANSACTIONS O N ALVTENNAS AND PROPAGATION September

one may introduce the concept

of scattered pon-er [ l ] The powers absorbed in the two cases are equal
Ps = I f6 12. (21) I ball2= I I? = I SCgag1*1* = I b,2I2. (30)
An important connection between the scattered pon-er I t follows that the power radiation patterns of such an
and the pon-er absorbed antenna are symmetrical in this way with respect to
the origin. An elementary dipole illustrates these fea-
PA = I 1'ba (22)
tures.
by a large class of antennas will now be demonstrated. At this point, one may contrast the approach of the
For a canonical minimum-scattering antenna (12) one present paper with thatof Dicke [l].Dicke showed t h a t
immediatelqr obtains (27) is validfor any reciprocalantennawithonly a
single local port ( N = 1) when a =smr an eigenvector of
fg = - sBO$fl,+ag. (23)
S. Employing the Gram identity cited in the Appendix
This states that f g is a linear combination of the col- (38), his proof can be generalized to cases N > 1. The
umns of SBa. In other words, the pattern of the scat- fields scattered by an antenna due to an incident eigen-
tered radiation is a linear combination of the orthog- wave s m may be designated
onal patterns in which the antenna normally radiates.
fmg = (Spg - IO~)S,$, ??z = 1, 2, * * (31)
Furthermore,oneverifies,foranarbitraryincident
wave a, t h a t Intheabsence of orthogonalityrelationsamongthe
Ps = PA (24) f,s, a demonstration of (27) for eigenwaves carries no
since by definitions and(11) implication for cases of more general incidence, includ-
ingthecase of greatestpracticalinterestinwhich
1 I
P A = b,12 = S4gagI2= Sga+agI2 I
(25) matched receivers are attached at ports 1 through N .
while from (6a) and (23) Dicke then investigates antennas for which the scat-
tered pori-er is a minimum (in fact, P s = P A ) simul-
I Sg,+aa I2 I
= SpO$o,+as l2
= fg 1' = Ps. I(26) taneously for each eigenvector sm of S. From this con-
Both (23) and (24) remain valid for minimum-scatter- sideration he is also led to the representation (12) for
ing antennas,in general. his special case.
The physical mechanism through which an antenna Finally, a brief remarkmaybeenteredregarding
absorbs power from a wave a6 incident from the sur- arrays of canonical minimum-scattering antennas. As-
rounding space is destructive interference, i.e., the an- sume the elements to be so spaced that the spherescir-
tenna scattersso as to cancel someof the incident fields. cumscribed about each element do not overlap. Each
I t m a y be conjectured, therefore, that quite generally element, on open circuit, produces no disturbancein
any incident field, and therefore the array as awhole is
PS 2 PA. ( 2 7 ) invisible. Nevertheless, the array is not necessarilya
Thatsuch is thecaseforantennasequippedwith canonical minimum-scattering antenna, since (4) and
matched receivers (a,= 0) provided consequently the orthogonality relations (6a) are not
necessarily satisfied [4].
I S4aoI = SBu+asI I (28) Theproperties of anarray of suchantennasare
is demonstrated in the Appendix. Equation (28) holds peculiarly susceptibletoexactcomputation: 1) the
for all real excitations of arbitrary reciprocal antennas fields radiated by a given element (with all others open-
or arbitrary excitation of reciprocal antennas with real circuited) are rigorously the same as those which would
patterns3 For minimum-scattering antennas equality is be radiated by the element isolated ; and 2) the open
achieved in ( 2 7 ) . circuit voltage computed from this field a t each one of
When the canonical minimum-scatteringantennais re- the other elements is rigorously that which would be
ciprocal, S, is pure real(11),and several most significant computed foran isolated antenna.These properties
new features appear. The first has to do with the sym- arepresentlybeingexploitedin work dealingwith
metry of the antenna patterns. Given an incident plane mutual coupling.
wave, the function representinga second wave incident Inconclusion,theanalysis of minimum-scattering
from the opposite direction is the complex conjugate of antennas has been greatly simplified and the range of
the first. From the formof the spherical mode functions this concept considerably extended beyond results pre-
and (l),i t follows that the modal coefficients represent- viouslypublished.
ing the second wave a g z are the complex conjugates of -APPENDIX
the coefficients representing thefirst aB1, i.e.,
BETWEEKPA
A RELATION AND Ps
ag? = agl*. (29)
Let
8 The physical significance of real patterns S d is made evident
in the discussion of reciprocal minimum-scattering antennas con-
nected with (29) and (30).
1965 Rao: YAar gmiy- T y p e in a Waveguide 675

Consequently SgBag= bo, andthe definitions of the From the lossless constraint
scattered power PS and the absorbed power P-4 then
S,qa+bb= 0 = bgfS,aa, (39)
read and
PS = Ibg - a@\'; p-4 = 1 balz. (33) Sga+s~a= Iaa. (40)
Hence the above determinantal inequality (38) leads to
For alossless antenna, conservationof energy requires
I(ao+bs)1' 5 I bg lz( I 1' - I sga+aa 1')- (41)
I agI2 = I b,12 + 1 bg12. ( 34)
If, now, (28) of the text
Hence,
1 SaDas I = I SBora' S I (42)
is satisfied,
I (++bo>l2 5 I b g lz(I ab l2 - I ba 1') (43)
or, in viewof conservation of energ)r (34), one concludes
Thus Ps > _ P Aif and only if (37). Thus (28) or (42) is a sufficient condition for the
Re (ao+bd Ibo 1 12. (36) validity of PS2 P.4.

Clearly it is sufficient to show t h a t REFERENCES

[ l ] C. G. Montgomery, R. H. Dicke, and E. M. Purcell, Principles
I (ao+ba)I 5 I bo 12. (35) of microm-ave circuits, in Radiutiotz LaboratorySeries,
New York: McGraw-Hill, pp 317-333, 1948.
vol 8.
[2] N. Marcuvitz, Waveguide handbook, in RadiationLabmatory
Consider thedeterminant of theGram [SI matrix Series, vol 10. New York: McGraw-Hill, sec 2.8, 1951.
formed from ag, b 3 , and the M columns of S3a, i.e., [3] 11'. K. Kahn, Scattering equivalentcircuits for common sym-
metrical junctions, IRE Trans. on Circuit Theory, vol CT-3, pp
121-127, Jun 1956.
[4] S. Stein, On cross-coupling in multiple-beam antennas, IRE
Tra.ns. on Antentus and Propagation, vol AP-IO, pp 548-557,
Sen 1962.
[SI F.'R.Gantmacher, The Theory of Matrices, vol 1. Yew York:
Chelsea, p 251, 1959.

A Two-Element Yagi-Type Array in a Parallel-Plate Waveguide-

Theoretical and Experimental Studies

Abstract-A theory is developed forthecurrent distribution, INTRODUCTION

impedance, and radiation characteristics of a two-element Yagi-type
array, consisting of a driven antenna anda tuned parasitein a parallel-
plate waveguide. By optimizing the spacing between elements and
the tuning reactance of the parasite, front-to-back ratios as high a s
+31 dB (parasite as reflector) or as low as - 19 dB (parasite as direc-
I N A R E C E S T P A P E R , Lewin [ l ] has considered
the radiation from a linear antenna in a parallel-
plate waveguide using a modal expansion technique.
The inherent inadequacy of this method, as compared
tor) can be obtained for antennas with Q =O.OlX and h =0.25X, where
a and h are the radius andheight of the antenna, respectively.
to integral equation methods commonly used for solving
Extensive experimental measurements have been made on the free-space antenna problems, is that it cannot beex-
tended for analyzing the current distribution and im-
antenna array, and a brief description of the apparatus is included.
The experimental results agree well with theory. Possible applica-pedance characteristics of coupled antennas, or of an
tions of this type of antenna array a s variable power-dividers and
array of antenna elements.
attenuators in strip lines and other circuits are explained.
In this paper, an integral equation technique been has
employed to study the behaviorof a two-element Yagi-
Manuscript received July 24, 1964; revised March 2, 1965. This type array inside a parallel-platewaveguide.Bythe
remrch was supported by the National Science Foundation Grant method of phase-sequence components, the two simul-
GP-851, and by the Air Force Cambridge Research Labs. Contract
AF19(604)-4118. Sections of this paper were presented at the 1964 taneous integral equations governing the current distri-
Spring URSI Meeting. Washington, D. C. bution on the coupled antennas are resolved into two
The author is with the Gordon McKay Laboratory,Harvard
University, Cambridge, Mass. independent integral equations, each of which has, in