348 IEEE TRANSACTIONS ON AKTENNAS AND PROPAGATION,’VOL. AP-19, NO.

3, XAY 1971

A Generalized Expansion for Radiated and

Scattered Fields

. Abstract-At a given frequencyevery perfectly conducting ob-

induced on the scatt.erer surface or t.he far-scattered field
stacle has associatedwith it aparticular set of surface currents
pattern, are represented as weighted sums of eigenfunc-
and corresponding radiated fields which are characteristic of the
tions (here called characteristic waves or modes) appro-
obstacle shape and independent of any specijic excitation. These
priatetothescatterershape.
characteristic modes form a useful basis set in which to expand I t is commonknowledge
that. the eigenfunction method ca.n be used for obstacles
fields radiated or scattered at a great distance from the obstacle.
Once these modes are known for a given obstacle, the scattering
whose surfaces coincide with certain of the separable co-
of plane wavesincidentfrom arbitrarysource directionsinto
ordinate surfaces, e.g., the sphere; less common, however,
arbitrary receiver directions may be evaluated concisely. To sup-
is the realization that characteristicmodes exist for a
port the theory, a method is described for determining character-
much larger class of obstacle shapes and that these can
istic mode currents on thin wires of general shape and is applied
toseveralshapestogenerate certainbackscattering be used to describe their scattering properties in an ele-
andinput
admittance data. Wherever possible comparison is made with gantly compact fashion. Section I1 briefly reviews the
existing data. theoretical basis of characteristic modes and states some
of their properties. From t,hese defining propert,ies a
I. INTRODUCTION method is described in Section I11 for det.ermining the
characterist,ic modes associated v d h t.hin finite wire ob-
HE prediction of the fields scattered by an arbit.rary stacles, and in Section 117, several exa,mples are present,ed
T obstacle in its resonant, region remains one of the vex- for comparison with exist.ing data. SectionV concludes the
ing problemsof electromagnetic and acoustic theory. True, paper 1T-it.h some remarks and observations.
t.he applicat.ion of digital-computer methods to this prob- An timeconvention is assumed and is suppressed
lemhas provided ans%-ers heretofore unobtainable,but for convenience. Column vectors are denoted by (.) and
too often these answers go no deeper than a set of num- matrices by [a].

bers, providing little insight tothenatural phenomena

11. THEORETIC.4L RESTJLTS
underlying them. In the course of studies pursued at the
Ohio Stat,e University, involving low-frequency shipboard It. is useful to view a scattering obstacle as a device
antennas and radar chaff elements, we ran headlong into which transforms an incident wave converging in upon
the need for a scattering theory pa.rticularly adapted to an origin within it into a scattered wave diverging away
arbitrarily shaped, perfectly conducting obstacles in their from it. [l], [a]. Such a scattering obsta.cle is represented
resonant regions which would providemore than a. nu- mathemat.ically by its scat.tering operator S or by its re-
merical solution. This paper presents an approach to the lated perturbation operat.or P , discussed [ 3 ] briefly later.
subject and offers some results t.o support. them. I n t.he a.bsence of t.he obstacle, any continuous wave
From our viewpoint, a resonant smttering obstacle is fi incoming upon the origin is transformed without. dis-
atuned device, not, unlikea t,uned resonantcircuit, in turbance into an outgoing wave f”. That, is,
which the entire shape and composition of the obstacle f” = I f i (1)
a.re thetuning mechanisms, therebyimplying that a
mathematical model which purports to describe the reso- where I is the identity operator and may be viewed as
nance phenomenon must somehow take into account the the scattering operatorof unbounded free spa.ce. By plac-
obstacle shape as an integrated whole. Thus the model ing theobstacle at the origin, thisidentityoperatoris
which we chose is based upon the eigenfunction expansion perturbed to form t.he scat.tering operator X, indicating
method in which pertinent functions, such as the current that the wavefiis now transformed into anout.going wave
which is perturbed from f”.That. is,
Manuscript received October 21, 1969; revised September 23,
1970. This work was supported in part by the Air Force Avionics
f” f Af” = sf”. (2)
Laboratory, Air Force Systems Command, Wright-Patterson AFB, The perturbation is represent,ed by the operator (S- I),
Ohio, and the Ohio St.ate University Research Foundation under
Contract F 33616-68C-1252. so t,hat
R. J. Garbacz is with t.he Depart,ment of Electrical Engineering,
Ohio State University, Columbus, Ohio 43210.
R. H. Turpin was with the Departmentof Electrical Engineering,
Af” = ( s- 1)fi. (3)
Ohio St,ate University, Columbus, Ohio 43210. He is now wit.h the The opera.tor + ( S - I ) is defined t.0 be the pert.urbation
Depart,ment of Electrical Engineering, Indiana University-Purdue
University, Indianapolis, Ind. 46205. operator,denotedby P , and the perturbation Af”/Z of
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
GARBACZ AND TURPIN : EXPANSION FOR RAJJIATED AND SCA-ED FIELDS 349

the outgoing function from its value in the absence of the

obstacle is ca.lled the scattered wave f”. Thus
f = Pfi. (4)
On the sphere Z enclosing the obstacle and infinit.ely
far from it,, the incoming exciting f u n d o n f”’ has a far-
field pattern function Fm;(O,+), and the resulting out.going
smttered field f” has a far-field patt.ern function Fmo(8,+).
What we seek is that set. of pattern functions { F,(O,Q))
associa.ted with incoming waves which are tra.nsformed
by the obstacle into replicas of themselves, with at. most.
a magnit.ude and phase change. That is, we seek to satisfy
the characteristic equation
anFn (e,+) = PFn (e,+) (5)
where a, are complex characteristic values of magnit.ude
less than or equal to unity. The normal, complet.ely con-
tinuous property of the operator P leads to a. spectral
theorem [3], [4] which assures us that { F,(O,+) } forms
a denumera.blyinfinite complet,e ort.hogona1 set on the
sphere 2 with respect to radiated pom7er. It, is convenient Fig. 1. Geometrical relat,ionships between characteristicvalues s,,
to order the functions F,(O,+), which are called charac- and a, of S and P = a(&’ - I),respectively.
teristic-pa.t.tern funct.ions, such that) they correspond to
characteristicvalues a, which satisfy I a 1 1 2 1 a3 I 2 n being scat.tered best. and those near n/2 or 3n/2 being
--
I a3 1 2 -.I t is also convenient to normalize the F, (e,+) scattered least. If CY, = n, t.he nt.h characterist>icwa.ve is
so that they each represent. unit radia.ted power. Together in scat.t,ering resonance; if an = s / 2 or 3n/2, thenth
with their orthogonality property, this means that characteristic wave is in scattering null, i.e., behaves as if
# z F,(e,+) X Fm*(e,+) sin e de = ~~s~~ CI+ t.he obstacle mere not present,. IY0t.e that in accorda.nce
(6) with our mode ordering scheme, t.hese 1att.er modes do
where, ,a is t.he Kronecker de1t.a and 2 0 = 12Or. not occur for finite n.
Although every finite obstacle possesses an infinite set Since the characterist.ic-patkern functions F, (e,+) form
of a, and F,, t.he scattering operation is such that acomplete ort,honormalset
’
(with respect. toradiated
W
power) onthe sphere 2 , any reasonably well behaved
(PT)= C I a n 1’ (7) scat,t.ering patt.ernF(O,+) on 2 may be expanded in t,erms
n=l of the F, (e,+). That is,
remains finite, implying that only the first N waves are m

significant enough to include in pract.ice, and we may t,reat

n=l
the scattering operation as effectively a finit,e dimensional
one to a good approximation. It. turns out that (PT)is where t.he coefficients A , are found in t.erms of a reaction
proportional to the total power sca.t.tered by the obstacle [5] bet-ween the primary source current. and the nth out.-
(under plane wave illumination)averagedover all pos- going modal electric field (specified by Fn(O,+) analytically
sible tumble aspects. continued in to the primarysource region). If the primary
The complex characterist.icvalues a, associated with source is a unit planewaveincident,upon the obstacle
any loss-free scatterer all must lie on t.he half unit circle from an a,rbitrary direction (ei,+;), t.he sca.t.tered electric-
centered in the complex plane at (-$,io), i.e., field intensity observed in adirection (Os,@) on Z is
given by
a, = I cos CY,
I eian, n/2 5 CY,
i 3n/2. (8a)
An alternative expression, suited t.0 Smith chart plot.ting
(Fig. I), is
where hi is a unit vector specifying t,he polarizat,ion st.ate
-1
of the incident plane wave and 5(8i,+i;Bs,@) is the dyadic
an = 1$-zx, (Sb) scattering patt.ern,
where, incidentally, t.he real constants h, may be related m

to extrema1 properties of certain real quadratic forms [SI.

n==l
The angle CY, is a measure of the effect,iveness of the ob-
sta.cle in scattering the nt,h-order converging charact,er- Thus, knowledge of the eharacterist.ic-pa.ttern functions
istic wa.ve, those waves associated with angles a, nearest Fn(O,+) andthe associated characteristic va,lues a, (or
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
350 IEEE TRANSACITONS ON ANTENNAS AND PROPAGATION, MAY 1971

angles an),permits a compact bilinear expression for the d l make rather novel use of the least. mean-square cri-
field scattered in any direct,ion due to a plane wave inci- terion by minimizing t.he phase variation along the wire
dent from a.ny diredon. rather than the field differences. This is accomplished by
The properties just described are-common to all loss- requiring tha.t 6he nth characteristiccurrent. I,, repre-
free obst.acles, but to simplify the determination of t,he sent.ed by the vector (I,) in the chosen ba.sis system, be
characterist,ic waveswe restxict ourselves to perfectly con- real and that the associated characteristic field E,, repre-
ducting obstacles. I n these cases it is always possible, a t sent.ed by the vect.or (E,) in the chosen basis system, be
angular frequency W , to define real (i.e., equiphase) char- equiphase over the wire surface and lag Z, in phase by an
acteristic-current. dist,ributions J,, induced on the closed angle cy,. The problem is t,o determine t.he phase angles an
surface S of the obst.acle by the nth cha.racteristic wave and the associated cha.racteristic current dist.ributions I,
such t.hat, int.he absence of the obstacle, they each radiate for the significant, modes. The characterist,ic-pattern func-
.a field pattern F, (0,4) on Z and an electric-field intensit,y tions F,(O,4) derived from the Z,, together with the char-
which is equiphase inside and on S and lags J, by pre- acteristic values a, derived from t.he a,, then provide a
cisely 6he angle a,.l For a part.icular surface S, determin- concise eva.lua.t.ionof t,he scattered elect.ric field according
ing that, unique set. of characteristic-current dist.ributions to (10) and (11).
{ Jn) which possesses t,his property, makes it possible to To t,his end, (12) is rewritt.en as
construct the bilinear expansionformula (10) for pla.ne
[ Z ] ( I ) = (E’)e-ia (13)
wave scattering from the corresponding perfectly conduct-
ing obstacle. Based on t.his property, a met.hod is devel- where a constant phase factor e-ia is removed from the
oped in the following section for the case of thin wire field vector and is shom-n explicitly. This phase fa,ctor is
obstacles. absorbed into t.he impeda.nce matrix, i.e.,
111. TECHNIQUE FOR DETERMINIXGCHARACTERISTIC rz’(a)l(n= (E’) (14)
MODES OF THINWIRES where [Z’(a)] = e-[Z].
A modal impedance ma.trix [Z] can be derived for a If a and ( I ) of (13) and (14) were t.he phase and cor-
wire such that responding real current vectorassociated with a character-
istic mode, i.e., if a = an and ( I ) = (In),(E’) would be
a real vector. At Ohis point, however, both a, and (I,) are
‘where ( E ) is a vector represenhg the tangential compo- unknown. Thusit is expect.ed that.,initially, (E’) has
nent of t.heelect,ric-field int,ensit,y a t $1 pointsonthe some residual phase factor which is a function of the dis-
wire due to a current I existing on the wire surface. ( I ) tance I along t.he mire. To reflect this fact, (14)is rewritten
isavector whose components are the coefficients of a as
Fourier expansion of t,he current I in terms of N ort.ho-
normal basis functions (AT 5 171). The ijth element of [ Z ‘ ( a ) ] ( I )= (Re ( E ’ ) ) i(Im ( E ’ ) ) . (15)+
the impedance matrix t.hen gives the tangential electric Letting ( 6 ) = (Im ( E ’ ]) , and assuming t,hat ( I ) is a
field at the ith point on t.he wire due to the jth basis real vector,
function of current. distributed over thewire lengt,h. Thus
[Z] is an M X N matrix and ( I ) and ( E ) are column C1I. {Z’(QI)11(0 = (6) (16)
vectors of dimensions N and 171,respectively. where [In1 {Z‘(cy) ] ] is a real mat.rix. The real vector (6)
I n a common applicat.ion of (12) to scatteringproblems is thereby related to the deviation of the phase of (E’)
[6],[7], the N-dimensional current, vector ( I ) is sought from an equiphase condit.ion. Ideally, ( 6 ) equals the null
which generat,es a scatt.ered field equal t.0 the negative of vector when ( I ) = ( I , ) . We shall approximate this null
the incident. field (i.e., ( E ) = - (Ei) ) at N p0int.s along vector with one which is minimum in a least mean-square
the wire, t.hereby approximating t,he exact boundary con- sense, i.e. ,
dition of zero totaltangent,ial elect.ric field everywhere
along the wire in a point-matching sense. A less common = (Qt(6) (17)
application [SI of (12) is the approximation of the exact. is to be minimized. (The superscript t denotes the t.rans-
boundary condition in a least mean-square sense. In this pose operation.)
met.hod, an N-dimensional currentvector ( I ) is sought, Following the suggestion of (17), (16) is mult.iplied by
whichgeneratesa scattered field that. differs from the its transpose, with the result
negative of the incident field sucht.hat t.he norm
+
11 ( E ) (Ei)I I of the difference vector, when a.veraged
over M > N points on the wire, is a minimum. Applica-
tion of (12) to the determination of characterist.ic modes

1 Note that those modes which are in scattering null, i.e., t.hose
for which a, = ~ / or 2 3 ~ / 2 ,correspond to J,, on S, which are the
mall currents of the ideal cavity formed by the interior of S when
in resonance at. angular frequency o.
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
GARBACZ AND TURPIN: EXPAISION FOR RADIATED AND SCATTERED FIELDS 351

Fig. 2. Characteristic values an for circular loop of radius 0.25X with wire radius of 0.0025h.

where [B (a)] is a real symmet.ric matrix of dimensions Equa.tion (21) determines the funct.iona1 dependence
T X N and assume for convenience that, the nommaliza-
h of the characteristic values en on a. At this point it. may
t.ion ( I )t ( I ) = 1 applies. Then (18) becomes be well to remember that en ( a ) is a measure of the devia-
tion of the phase of ( E ) from an equiphasecondition.
[ B ( a ) - 4a)Il(I)= ( 0 ) ( 20) The smaller t,he €,(a) the closer t.his idealcondition is
approached. In this light it is clear that we are interested
where [I] is the identity matrix and(0) is the null vector.
in those valuesof a which cause relat>iveminima in e m i n ( a ) ,
The equation
where e m i n ( a ) denotes the smallest of the N characteristic
det [ B ( a ) - e ( a ) I ] = 0 (21) values en(.') for a specified a. These relative minima., de-
noted by e,,,in(an), will occur for a = an,n = 1,2, - ..,AT,
is recbgnized as the characteristic equation of t.he matrix corresponding to each of the charact.erist,ic vect.ors
[ B ( a ) ] . The N roots € ( a )= e f l ( a ) , n = 1,2,...,N, of (Imin(a,))of (20). The angle an is the phase lag of the
this equat.ion are the chara.cteristic values of [ B ( a ) ] ,and nth equiphase surface field ( E f l )and the associated char- '

t.he associated N vectors ( I ) = ( I n ( a ) )are its character- acteristic vect.or (In,in ( a , ) ) , hereafter denoted simply by
istic vectors. Because [ B ( a ) ] is real and symmehric the ( I , ) , describes the nth characteristic current distribution.
e n ( ( . ) must be real, and because of the mean-square con- For each pair a, and ( I n ) ,the equiphase constraint on
dition enforced by (18) theymust be nonnegat*ive. To the tangent.ia1 electric field is satisfied in a least mean-
assist in the following discussion, the characterist.ic values square sense at t.he N selected points along the wire. The
€,,(a) have been plottedinFig. 2 as a. funct.ion of the nth characteristic-mode current. distribut.ion ma.y now be
phase angle a for a specific example to be developed lat.er. expressed explicitly as a Fourier expansion in terms of the
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
 _. x- ~

352 IEEE TRANS.4CTIONS ON ANTENN-kS AND PROPAGATION, MAY 1971

TABLE I relat.ive minima of this e m i n ( a ) which are sought. To em-

PEASE ANGLES an FOR CIRCVLAR LOOP O F R a D I W s
0.25X WITH WIRE R ~ I W0.0025X
S phasize this point e m i n ( a ) is shorn for the preceding ex-
ample as a heavy line in Fig. 2.
h d y t . i c a 1 Method Numerical Method
Mode n (degrees) (degrees) IT. EXAMPLES
OF CHARACTERISTIC
M O D E S AND
USE m-SCATTERIKG
THEIR SOLUTIOKS
103.33 103.4
107.5 108.13 In this section the technique described in Section I11
263.1 263 .83 is used to determinecha.racteristiccurrentdistributions
269.75 269.7
-270 .0 269 .99 and characterist.ic values of circular and elliptical loops,
st.raight wires, circular arcs, and one-turn helices. I n each
case the wire obstacle is considered perfectly conducting
aforementioned set. of orthonormal basis functions (e.g., and its radius a is assumed to be much smaller t.han its
cosine and sine functions), the coefficients of which are length L and the wavelength X. The assumption is made
the components of the real vector (1%). that t.he surface current denshy on the thin wire ha.s only
In applying t.his procedure it is convenient to break up an axia.1 component and is distributed uniformly a.round
the matrix [B (a)] of (19), separating out those parts the circumference of the vrire, i.e., I, = 2raJn.
which are independent. of a. I n addition, when both even More than one characteristic mode generally is needed
a.nd odd basis functions are used for the currentexpansion to describe the far field of a scatterer. Often, however,
and these functions are ordered so that, the first X, com- one mode is suEicient, as is the case, for example, uit.11
ponents of the current vector ( I ) correspond to the even short. wires, i.e., wires of first. resonant length or less, and
. and the last N o components correspond to the odd basis this single-mode character holds true forquite general
+
functions(where N = hTe N o is the total number of wire configurations. For short. wires which are effective
basis functions), the matrix [B(cr)J has a block diagonal scatterers, the dominant mode has associated with it a.
form, a,nd the even and odd characteristic-mode current phase angle a, within about f a / 4 of the a = T resonant
dist,ributions can be derived separately from the respect.ive angle; higher modes cha.racteristic of such a scatterer are
even and odd characteristic equations. These details are found to have phase angles very near r / 2 or 3 ~ / 2imply- ,
given in [SI. ing that their contributions to scattering are relatively
An example of the behavior of the characteristic values small and usually negligible. As wires increase in length,
€,(a) as functions of a has been given in Fig. 2. The scat- more modes are required to describe their scattering prop-
terer in t.his example is a circula,r loop of radius O.%X, ertiesaccurately. The number of modes needed is also
with a wire radius of 0.0025X. The loop is approximated dependent, though to a lesser degree, on t.he nire geometry.
by 120 straight-wire segments, and five cosine functions Using the characteristic modes derived for several of
formthe basis of the Fourier expansion of thecurrent the examples, we calculated the backscattering cross sec-
distributions. Due to the rotational symmetry of the loop tions averaged over all possible tumbleaspects of the
a.nd the resulting second-order degeneracy of the modes, obstacle, assuming asphericalprobability densit.y func-
the characteristic valuesassociated w4t.h the sine functions tionfor aspect, angle. This is referred to below as t.he
are the sa.me as t.hose associated with the cosine functions. tumble average backscattering cross section [SI. Also in
The characteristicphase angles* a,, n = 0,1,2,3,4, cor- certain cases we caIcuIat.ed the input. admittance of a gap
responding to the relative minima. e m i n ( a n ) , are given in located at a point. Z a.long the wire obstacle according t.0
Table I where they are compared with those angles ca.lcu- the formula [SI
lated [SI analytically by independent. met.hods. m
The circular loop is a very special example in that. the Y(Z) = - a, I In(Z)I*. (22)
n=l
charact,eristic-mode current.distributionsandthe basis
functions used for their expansion are one and the sa.me, Wherever possible, t.he results are compa.red with solutions
i.e., cosine and sine functions. For this reason each char- derived by independent.methods. Many more examples
acteristicvalue en(a) of [B ( a ) ] assumes aminimum may be found in [ 9 ]
emin(an)for its respective angle an a.s seen in Fig. 2. For
more general scat.terers the cha,racteristic value behavior A . Cireuhr Loop
is not, this simple but no extra complications arise. In any Characteristic modes were determined for acircular
case, the smallest characteristic value e m i n ( a ) is the only loop -4th a loop radius to wire radius ratio R / a = 100.
one of interest, and regardless of which E, (a)happens to The first four phase angles a, are shown in Fig. 3 for a
be the smallest for the choice of a under study, it is the 4: 1 range of frequencies. The curves compare fa.vorably
with independently derived results presented in [3]. The
associated cha.racterist.iccurrents are t,heCOS nd,and sinn+
2
Because the characteristic currents turn out to be the familiar
cos n# and sin n@ functions in this example, we have chosen to functions and are not plott,ed here. Backscatteringand
order the modes on n in the conventional manner, beginning wit,h t.umble average cross sect.ion data derived using these
n = 0. In more general problems one would order the mod= so as
to make I cos an 1 a nonincreasing sequence wit.h increasing n. modes are shown in Fig. 4. The edge aspect results are

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
GSILBACZ AND TURPIN: EXP.4KSION FOR RADI.4TED AND SCATTERED FIELDS 353

0.1 0.2 0.3 0.4

LOOP RADIUS,

----- AVERAGE,SAME-SENSE LINEAR TRANSMITANDRECEIVE

0 0 0 0 0 HARRINGTON'S SOLUTION, EDGE ASPECT

I O I

1 ol
\
a a Y

\ I
I
10-4 I
0.1 0.2 0.3 (
LOOP RADIUS, "X/

Fig. 4. Backscattering and humble average cross sect.ion versusfrequency of circular loop m4t.h loop radius to
wire radius ratio R / a = 100,

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
 -.
354 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, UY 1971

I .- HARRINGTON'S
SOLUTION I t
a , . SINE MODE

I//Qo, COSINE MODE

P-
=" J
-4!.10 0.15 0.lO 0.!5 O.?, 0,;s 0.i 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
LOOP RADIUS "/x AXIAL RATIO, %
Fig. 5. Input admittance versus frequency of gap in circular loop Fig. 6 . Phase angles a,, n = 0,1,2, for elliptical loops with circum-
with loop radius t o wire radius ratio R / a = 100. ference C = 1.1X and wwe radius a = 0.00175X.

comparedwith solut,ions obt,ained byHarringtonand ordering of the a, is used here as was used for the circu-
Maut.z [lo]. lar loop. For an axial rat.io of unit,y the angles a, are, of
The real and in1agina.y parts of the input admit.tance course, those shorn in Fig. 3 for the corresponding circu-
of an ideal gap in the loop are plot,ted in Fig. 5 and are lar loop. As the axial ra,tio is variedfromunit,y the
compared with the Harrington and Maut.z solution [lo]. rotational symmetry of t.he loop is lost, and t.he modes
There is excellent. agreement in the conductive pa.rt of the are no longer doubly degenerate. Backscattering and tum-
admittance but, the susceptive pa.rt. is consistentlytoo ble average cross section for these examples are shown in
induct.ive. At present ixe a.re not certain of t.he cause of Fig. 7.
this error, but most likely it, reflects t.he neglect of higher Input admittance data mere calcuhted [9] for a gap
order modes, all of which are capacitive (i.e., those angles located both on the major and minor axes of the elliptical
a, which hoverabout 270"). However, the da.t,a verify loop but are not presented here since no compara.tivedat.a
the soundness of (22) a.nd the principles behind it. could be found in the literature.

B. Elliptical Loops C . Straight F i r e

It. should be noted that the characteristic modes ofa. The characteristic phase angles for a st,raight wire wit.h
circula.r loop are doubly degenerate, t,hat is, tn-o charac- a length to dia.meter ratio L/2a = 100 are shown in Fig. 8
teristic currents of the forms cos n4 and sin nd are a.ssoci- as functions of frequency. The characteristic current dis-
ated with each a, of Fig. 3 (excepting n = 0). It is of tributions I,, for a wire of lengt,h L = 1 . 3 a.re shown in
interest, therefore, to observe t.he effect. which distortions Fig. 9 where each current has been normalized to radiate
from the circula,r geometry have on the modes. For ex- unit pon-er. Fig. 10 gives backscattering and t.umble a.ver-
ample, one may study themodes of a planar ellipticalloop age cross sectionda.taobtained using t.he characteristic
whose axial ratios diverge from unity. Fig. 6 shows the modes associated withthe phase angles of Fig. 8. The
phase angles associated with the lowest order modes of broadsidedata. are comparedwithresultsobtained by
elliptical loops as functions of their axia.1 ratios. The cir- Harrington and Mautz [ll].
cumference is held constant at 1.1A. The wire radius is h comparison of the phase angle curves with the sca.tr
a = 0.00173 and the circumference to wire radius ratio tering behavior as a function of length of the straightwire
is NOT, i.e., the wire is of the same dimensions as tha.t clea.rly denlonstra.tes the dependence of the scat.t,ering or
used in t.he preceding circular loop example for a loop r a d i a h g efficacy of a mode upon its cha.racteristic value
radius R = 0.175X. For convenience the samesubscript) or phase. I n each case considered (where one mode is t.he
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
GARB.4CZ .LXD TURPIN: EXP-4NSION FOR RADIATED AXD SCAWERED FIELDS
355

/-
/

0. I

/ /'
- S=+=Ool.
(
BROADSIDE
ASPECT
8-POLARIZATION

/
-- BROADSIDE
ASPECT.
.#. - POLARIZATION
--- + =O') , 4 -
i
EDGE ASPECT e- 900.
POLARIZATION
-.- EDGE ASPECT
(e=+=9Oo),+-POLARIZATlON

-----
/ 0
AVERAGE,SAME-SENSE
LINEAR
TRANSMIT
RECEIVE
0
AN0

MEASURED VALUES

A X I AR
L ATIO, A
/'

Fig. i . Backscattering andtumble average cross sections of elliptical loops with circumference C = 1.1X and
mire radius a = 0.00175X.

WIRE LENGTH, L/X

Fig. 8. Phase angles a,, n = 1,2,3,4, versus frequency for straight wire witif.hlength to diameter ratio L/2a = 100.

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
. . ~. . ..
356 IEEE TRANSACTIONS OW ANTENNAS -4ND PROPAGATION, MAY 1971

0-
-a
I-
VI

W
I-
o
24
o

-0.20’ I I
0
% L/2
OISTANCE 1 ALONG W I R E

Fig. 9. Normalizedcharacterisi,ic currentdistributions I,,, n =

1,2,3,4, for st.raightwire length L = 1.5X with L/2a = 100.

-9
0.2
I
0.4
I
0.6
I
0.8
I
1.0 1.2
I 1
1.4
I
I6
WIRELENGTH I N WAVELENGTHS - p/x
o.o#
-BROADSIDE ASPECT, MODE SOLUTION
(b)

F! I
AVERAGE,SAME-SENSE LINEAR TRANSMITTER
AND RECEIVERPOLARIZATION
o o o o o BROADSIDE ASPECT,HARRINGTON’S
-----
SOLUTION
I
Fig. 11. Input admitt.anceversusfrequency of centrallylocated
gap in st,raight wire withL/Za = 100.

dominant contributor) a peak in t,he average cross section

0.001
0.3 0.5 0.7 0.9
I
1.1
I
1.3 1.5 I.7 corresponds to a wire length for which t,he phase associ-
WIRE LENGTH, A
/‘ ated wiith the dominant mode is equalto or approximately
Fig. 10. Broadside and tumble average
backscattering cross equal to r.
sect.ions versus frequency of straight wire with L/2a = 100. The real and imaginary parts of t,he input admitkance
of a. center gap in the dipole are plotted in Fig. 11 and
may be compared with Harrington and Mautz’s solution
[E] for L/2a. = 74.2. As in the case of the circular loop,
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
GARBICZ AKD TURPIN: EXPANSION FOR RADIATED AND SCATTERED FIELDS 357

WIRE
LENGTH,

Fig. 12. Phase an,n = 1,2,3,4, versus frequency for one-turn helix with wire lengt,h to diameter rat.io L/2a =
100 and pitch p = loa.

Fig. 13 gives backscatteringand t,umble average cross

sect.ions for linear polarizations. I t is interesting to com-
pa.re Figs. 8 and 12 and observe the relat.ionship between
bandwidt.hs of the first resonant. peaks and the slopes of
t.he phase angle curves at 180”.

V. CONCLUSIONS
0.1
We have presented a case for the use of cha.racterist.ic
“/x* modes as a basis set in which t.0 expand the fields radia.ted
OR or scattered at a far distance from an arbit.rary obstacle.
7 2 To supportour theoret,ical conclusions, cha,ract,eristic
modes were calculatedfort,hin wire scat.terers and t,he
0.01
resultsderived using them were found to bein good
- 8 + 0’;
( =
BROADSIDE ASPECT
= +POLARIZATION )
agreement wit,h established data.
We. hope that the chracteristic-mode viewpoint will
--
18; + ‘0’;8-
BROADSIDE ASPECT
POLARIZATION
lead to a better understanding of scatt.ering by obsta,cles
----- AVERAGE, SAME-SENSE in their resonant region. I n particular, analysis by char-
0.001 TRANSMIT LINEAR AN[
acteristic modes may give some insight to the change in
RECEIVE
scat,t.ering due to cert,ain controlled perturbations in ob-
st.acle shape. Or, t,hey may aid the design of singly or
0.4 0.6 0.8 1.0 1.2 1.4 1.6 I multiplyloaded scat.t.erers byproviding clues as to the
WIRELENGTH, L/~
most. effective locationsforsurface slot.s, etJc., to affect
Fig. 13. Backscattering andtumble average cross sectionsvelsus cerhin modes overothers. Theymay be useful in the
frequency of oneturn helix with L/2a = 100 and p = 10a for
linear polarizat.ions. synthesis of far-field patternsradiated fromnonplana,r
apertures.
Further work on characteristic modes is warranted by
t,he conduct.ance exhibit.s excellent. agreement butthe the encouraging resu1t.s described in this paper. I n partic-
susceptance appears consistently too inductive. ular, the calculation of characteristic modes of rotationally
synlmet,ric, perfectly conducting obstacles3 is a logical ex-
D. Helices
A one-turn helix with a wire length to diameter ratio A recent conversat.ion nit.h Prof. R. F. Harrington indicat.es
L/2a. = 100 and a. pitch p = 10a was investigat.ed as a that he has successfully at,tacked this problem and together with
function of frequency. The phase angles corresponding to J. R. Maut,z is preparing a Technical Report, “Theory and com-
putat.ion of charact.eristic-modes forconducting bodies,” for
the first, four charact,erist.ic modes are shorn in Fig. 12. Air Force Cambridge Research Laboratories.
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.
358 IEEE TRANSACTIONS ON ANTENNAS AXD PROPAGATION, VOL. AP-19, NO. 3, MAY 1971

tension of the work presented here. A method for finding tering problems,” IRE Trans. Antennas Propagat., vol. AP-3,
Oct. 1955, pp. 193-199.
the modes associated with loss-free dielectric obstacles [6] J. H. Richmond,“Digit.alcomputersolution of the rigorous
presents an interesting challenge [13], [14], as does the equations for scattering problems,” Proe. IEEE, vol. 53,
Aug. 1965, pp. 796-804.
development. of a parallel theory for obstacles in elect.ro- [ T I R. F. Harrington, Field ComputationbyMomentMethods.
static and magnet.ostatic fields. I n network theory, char- New York: Macmillan, 1968.
[8] R. H. Turpin, “Basist.ransformation, least square and char-
acteristic modes may find use in the analysis of N-ports, acteristic-mode techniques for t.hin wire scattering analysis,”
pa,rticularly in large power dist,ribution systems. We hope Ph.D. dissertation, Ohio State Univ., Columbus, 1969.
[9] R. J. Garbacz and R. Wickliff, “Introduction to characteristic
that this paperd l stimulate research in these diredons. modes for chaff applications,” Ohio State Univ., Electro-
Science Lab., Air Force Avionics Lab., Wright-Patterson
AFB, Ohio, Tech. Rep. 2584-6, Apr. 30, 1970.
REFEREKCES [lo] R. F. Harringtonand J. Mautz,“Computations for loaded
wire loops,” Rome Air Dev. Ctr., Griffiss AFB, N. Y . , Tech.
[I] C. G.Montgomery, R. H. Dicke, and E. M . Purcell, E&. ReD. RADGTR-67-8. Feb. 1967.
P.rincipZes of Micrmmve Circuits (M.I.T. Radiat.ion Lab. Ser.). [ll] 2, “Matrix met,hod for solving field problems,” Rome
New York: McGram-Hill, 1948, pp. 317-333, 405411. Air Dev.Ctr., Griffiss AFB, N. Y., Tech.Rep.RADC-TR-
121 R. G. Newton, Scaltering Theory of Waves and Pa.rticles. New 66-351, vol. 11, Aug. 1966.
York: McGraw-Hill, 1966, pp. 177-213,300-304, 452-457. [12] -, “Straight wires with arbitrary excit.ation and loading,”
[3] R. J. Garbacz, “A generalized expanslon for radlatedand IEEE Tram.Antennas Propaad.. vol. AP-15. Julv 1967,
_ I I I ”

smttered fields,” Ph.D. dissertation, Ohio State Univ., Colum- pp. 502-515.
bus, 1968. [13] P. C. Waterman, “New formulation of acousticscat.tering,”
[4] Mr. Schmeidler, Linear Opera-tors in Hilbert Space. New J. Awust. Soc.,Amer., vol. 45, June 1969, pp. 1417-1429.
York: -4cademic Press, 1965, pp. 58-60. [14] -, “Scat,tenng by dielectric obst.acles,” AZta Freq. (Special
[5]M. H. Cohen, “Application of the reactionconcept to scat- Issue), vol. 38, 1969, pp. 348-352.

Diffraction by Arbitrary Cross-Sectional

Semi-Infinite Conductor

Abstract-This paper investigatesthe problem of diffraction the sametechnique. I n almost all cases, an assumpt.ion
of an electromagnetic wave by a semi-iniinite conductor by means is made t.hat the plane is infinitely thin. There are, how-
of the computer method based on an integral equation formulation.
The crosssection of the conductorcan have an arbitrary shape, ever, a fern works available in which finite thickness is
provided the face shined by the incident wave is a uniform plane considered. Jones [a], for example,consideredsucha
thick problem which, however,is useful only for the thick-
in the region far from the edge. In that plane region, the problem
is simply that of reflection from a conductor wall for which a solu-
ness smaller t,han O.lX, where X is the wa,velength.
tion is known.This known part of the solution is analytically excluded
More recent.ly, t y o papers [3], [4] have appeared which
to derive an integralequationdescribing the edge effect over a
treat the thick half-plane problem without t,he restrict.ion
limited region. The limited region is u n k n o w n a t kst and must be
determined from the results of some numerical computations. As a kl << 1, where k is the wavenumber and 1 is the thickness.
numerical example, a thick half-plane conductor is treated in which The work of Lee and Mittra [4], in particular, is intcr-
the electric field polarization is parallel to the conductor, i.e., the
esting; it is basedon the generalized scatteringmatrix
case of an incident E wave. procedure, which is useful for a.nalyzing waveguide dis-
continuity problems, and t.he Wiener-Hopf technique.
I. INTRODUCTIOK The present paper investigates t,he same kind of prob-
l e m by means of a computer method based on an inte-
T HE DIFFRACTION of a planewavebyasemi- gral equat.ion formulationdevelopedbyMeiandVan
infinite conductor is oneof t.he classica.1problems. Cop- Blade1 [SI andothers.Here,the
son [1] analyzed this problem by mea.ns of t.he Wiener- semi-inhite conductor may be
cross section of the
of arbitrary shape, pro-
Hopf t,echnique. Since that time, several modifications of vided the face shined by an incident m v e is. a uniform
the original half-plane problem have been invest,igated by plane in the region far from t.he edge. The author previ-
ously considered this problem in the case where the inci-
Manuscript received June 22, 1970; revised August. 7, 1970. dent wave hits upon t.he semi-idnit>e conductor at right
Theauthor is v4t.h theDepartment, of Communicat,ion Engi- angles [SI. I n this case, however, a.n arbitrary incident
neering, Faculty of Engineering, Osaka University, Suits, Osaka,
Japan 565. angle was notpermittedbemuse somekinds of semi-
Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on July 06,2023 at 04:26:07 UTC from IEEE Xplore. Restrictions apply.