RADC-TR-66-351, Volume I

Final Report
MATRIX METHODS FOR SOLVING FIELD PROBLEMS

Volume I - Matrix Techniques and Applications

"' Roger F. Hairrin-gton

Long-Fei Chang
Arlon T. Adams, et al

Syracuse University

TECHNICAL REPORT NO. RADC-TR- 66-351

August 1966

Distribution of his document is unli mited

DDC

'OCT 10 1966
I"A W -1W-9it0 U S E
,oR R,)ERAL 3CIENnFIC AND
TECHNICAL IN''ORMATION C
7ri0ar0oPLabe ome Air Development Center
| .oO $ jY'- '
IO plM Re earch and Technology Division
• . Air Force Systems Command
ADDJ
j V Gri fiss Air Force Base, New York
I

Then US Government dtawi.gs. speci.kit-ots. or other data are used for any purpose other
than a definitely related gov-.-mment procurement operation, the gosernment thereby incurs
no responsibility nor any obligation whatsoever; and the fact that the government may have
formulated, furnished, or in any way supplied the said drawings, specifications, or other
data is not to be .egarded, by implication or otherwise, as in any manner licensing the
holder or any otner person or corporation, or conveying any rights or perm:ssion to manu-
facture, use, or sell any patented invention that may in any way be related thereto.

oi se o s

'YFW ~ Ti
. A .........

I 'Wb~oe/AvAuCi,. 1

Ii far

Dontrtunti op.RtJ rdsry

 MATRIX METHODS FOR SOLVING FIELD PROBLEMS
- Volume I - Matrix Techniques and Applications

Roger F. Harrington
Long-Fei Chang
Arlon T. Adams, et al

Syracuse University

Distribution o this docr.nent is unlimited

m'~ "

j
°
M~in-'i=t =, ' .~nR ~q~h .'-~d-' o -
 F01MORD

This report was prar by Syracts. fiversity, Electrical Engineer-

ing Departzant, Syracuse, Nev York undor Coutract No. AM3(602)-3724,
Project No. 4519; Task No, 51901, covering the period March 1965 throvqh
W March 1966.
RADC Project Engineer was John J. Patti-(CIKRa);
ChaPpters I, II and III, on Basic Concepts, Formulation of Electro-
sagnetic Problems, and Wire An tennas and Scatterers of Arbitrary Shape,
w'ze vritten by Roger F. Harrington. Chapter IV, on Calculations for
Linear Wires, 'as writtza by Joseph Kautz and R. F. Harrin ton. Chapter
V, on Scattering by Conducting Cylinders, was written by Robert Wallenberg,
and Chapter VI, on Scattering by Dielectrie Cylixders, as ritten by
Arlon Adens. The next chbpt r on Scatterijy by Bodies of Revolution is
by Thomas 3ristol, and the final chapter on Aastri Inversion is by Long-
Fei Chang. The material in each chapter is nt racessarily the sole work
3 of the author, since the project was a coordinited effort of all persons
working on it.

This technical relort has been revieved and is approved.

Approwd:
.I~HN JA TI
Project Engineer

Approved:
CA M. COSEL
Colonel, USAF
Chief, Communications Division

FOR THE COMMANDER:

ING
$ ASELMAN
Chief, A n cd Studies Group

A
ii

- _ - _
 ABSTRACT

This reporu poresents general pro~zedures for solving field problems

of engineering interes-t using digitaLl computer techniues. The basic

concept is to represent a boundary-value problem by a superposition

integral, approxina-e the integrai ecuation by a matrix equation, and

invert the mtrix for a solution. The theory is d xtcribed in terms of

the method of moments , _aich is eauivralent, to the variational method.

For electromaoetic antenna and scattering- vroblems., th._e rnethod

gives a matrix whose elements can be interpret.ed. as generali zed impedances.

T'hese impedances are closely related to those used in the theory of

loaded antennas and~ scatterers, and hence such lo-aded structres canz also

be treated. A solhution for wire antennas and scetterers of arbitrary

shape is formulated in detail, and calculations for linear wire antennas

and scattertrs, both loaded and unloaded, have been made.

Additiconal poroblems t1,reated by these procedures are two-dimensional

scattering by conducting cylinders and bydieleatric cylinders, - d

three-dimensional scattering by bodies of :ievoluation. These problems

are used to show the effect of various approdmati xis in the sclution,

in an attempt11 to draw some general conclusions as to the 1-esIt approxi-

mations. Special procedures for inverting matrices have also been considered,

to take into account any symmetry properties present in the matriCes.

A considerable saving in computation time can often be made by properly

utilizing these syrmetries.

 are

I.BASIC CONCEPTS- .. .................... 1

A- Introduction1
B. Formlation. of Pro3blems 1
C. Example: Electrostatics I
D. M-4ethod of Moments 5
E. Rxapple: Charged Conducting Plvte 8
F.Varliational Intei'pretation 'I
G. References 1

179

B. Conducting Bodi~es 20
C, Dielectric Bodies 295
D. -Magn-eti--'c Bodies 28
E. Bodies 'thDielectric adMagnetic 50
-;- 1easuaremeant
G. Discussion 38
H.References -4
Tfl1. WIR A'~jj N)~~~~ FAW9AYSAE-. -,h

'It.±ntrowlckiOn 4
Eoz
Fb.- tion of"tht-e Prble= 144
C.- Wir-e Ant9nn
D. Wire Scattrrs 5

Ealo zio%, -1 58
G. Ref erences &2e

i. cpscr±rion-mS FOR Lu1710 Si'E .~

--------------. . 96
A- PS ecevise-linear Current,. Calarkin' Method 6
BA.Point Matching and Finie Diff-erences
C. linear Antennas with Aritrary Feed 68
D. Linear Wire Scatterers P
S. loaded Wire Scatt,'erers
F.References 9)0
V. SC A iR!BY CIIIXCIN CWhIIPq OF ARIEflRARY SHAPE . . . 9
AIntroduction
fPulse Paturon Annroxhation uw
-arren-
C.Higher Approinations to Current Density
B.Solution U3sing GalerkIn' a Methd 11)9
Ba. Ev.alu-,-ation ofL integrals f13
F. eferences up,
VI. SCAThERINSI BY DILCTPJC CYIDR.
im . . . . . . - . . .t~
A. Introduction 1
B. Formulation of the Problem ( fl4 to z) 109
C. Evalluation of Partial Matrix Elements f'rom
Magp-etic Currents12
D. Evtaluation of Partial Matrix filnrnxts from
1
Electri-c Curxrents -1 Q
ED.Thalitv. 132

I,. CPA IRi PF~ BODIIES (IF REMfl-rxI l.. ....... .. .

yfl ~
r-----f U

C. s~rar ta' nr':I~- ta_

pl-c-Met-hod of !4--ments
r~~~e~~a
Sluio -- actr%
3&
£JL-i;t mI.. ir in 5- ..

5.Fear -Antenn

fl na'a' ~ TrAvi
 T1he objective oI this effort;, was to deveLob pracea tichnioue-s
-for solving a vide class of electroagnti sca-dterins q!fS, ro

A niediadgeneral &Jt-rU&aCh is ta- to eiectronaunetzr iel

probemsin terms of 11U=e2!20!1 analy5s. porac -tSLLS
aR- ; f
m:atrix euuti ons renresentirc resoo the iitrnt
j~4.er-di..Afrrr'iiaj

gouatins of the n &lar bein g arm The- accurac-y of t'-

solution diatj-ijrd JIs very-j goad tar most p~priflg-

SOM cr.e- exit.-__

anoears that s-m s mre hnam rvo -z

in o -the..atJ t .=Sv
ti

scatter11-ng wnwajpmns am-esaliha a gv~ri fram2W3'' Wtf

mek o amac are descibe in Vole I t-r f

Th e r ess 01 zivanS eIfhnt are drectly s-caeto the

urn aef -tZ i- aten In tharcrfap . a

w~ee ~zrcLresut 3ire- nresete ii 'IVo=-

1 the flanaI

ec s.
 MATRIX MET5.ODS FOR SOLVING FIELD PROBLEMS

A. introduction. The object of this project is to develop practical

techniques for solving electromagnetic scattering problems using digital

computers. The general approach is to approximate the integro-differential

equations describing the boundary value problem by matrix equations and

to invert the matrix equations.

The theory is conveniently discussed using the concepts of linear

spaces. The equations of interest are of the general type

L(f) = g (1-1)

where L is a known linear operator, g is a known function, and f is the

unknown to be determined. The general procedure will be to take an

equation involving f'unctions, such as (1-1), and to approximate it by

a matrix equation. The solution to an inLomogeneous matrix equation is

found by matrix inversion. Most computer libraries have subroutines for

matrix inversion, and hence a problem will be considered solved once the

elements of' a well conditioned matrix are evaluated.

B. Formulation of Problems. General procedures for solutions will

be discussed in the language of linear spaces and operators. Hence, the

problems to be considered should be formulated in this notation. Given

a deterministic problem, we wish to put it into the form L(f g,

identifying the opcrator L, its domain (the functions f on which it

operates), and its range (the functions g resulting from the operation).
 satisfy1
Furthermore, we usually need an inner pyoduct <fg>, definel to

<fig> = <9,f> (1-2)

<af + Pg,h> = qfh> + V-gh> (1-3)

<f ,f> > 0 if f 0

o
0 if f 0 (1-4)

where a and 1 are scalars, and * denotes complex conjugate. We some-

times need the adjoint operator La and its domain, defined by

<Lf,g> = <f,Lag> (1-5)

for all f in the domain of L. An operator is self adjoint if La = L and

the domain of La is that of L.

Properties of the solution depend upon properties of the operator.

An operator is real if Lf is real whenever f is real. An operator is

positive definite if
<f*,Lf> > 0 (1-6)

for all f 1 0 in its domain. It is positive semidefinite if > is replaced

by > in (1-6), negative definite if > is replaced by < in (1-6), etc.

We shall identify other properties of operators as we need them.

If the solution to L(f) = g exists and is unique for all g, then

the inverse operator L-1 exists such that

f (g) (A-7)

If g is known, then (1-7) represents the solution to the original problem.

However, (1-7) is itself an inhomogeneous equation for g if f is known,

and its solution is L(f) = g.

The usual definition of inner product in Hilbert space corresponds to
<f*,g> in our notation. For this report it is more convenient to show
the conjugate operation explicitly wherever it occurs, and to define
the adjoint operator without conjugation.

-2-
In general, it is desirable to think of L and L-1 as a pair of operators,

each one of which is the inverse of the other. The choice in any parti-

cular problem is one of convenience.

C. Example: Electrostatics. Consider a volume density of charge

p(x,y,z) in unbounded Euclidean space of constant permittivity E, and

its associated electrostatic potential $(x,y,z). The differential

eauation for the problem is Poisson's eauation

-= (i8)
subject to the boundary condition that r$-*! as r--o for every p of

finite extent and where C1 is a constant. In operator otation, (1-8) is

LO P (1-9)
where -2
L = v

subject to the above boundary condition. The well-known solution to

(!-8) is

O(~Y7
,(x~~~~y~~z)f ERf
, XY.Z
', a.,=
3xt dy' dz
z=
(1-211

I 2 22
where R = (x-x') + (y-y') + (z-z*) is the distance from a source

point (x',y',zt) to a field point (x,y,z). Hence, the inverse operator

is
L-j dx' dy' dz' 1

It is important to realize that (1-12) is inverse to (1-10) only fo' the

stated boundary conditions. The boundary conditions of a differential

operator should be considered an integral part of the operator, and "f

-3-
they are changed the operator is changed. Hence, L1 -will also be

different if the hnnnhif-'ry r-tnA 4t~nz e%, T. &-Ar "ttr'a

A suitable inner product for electrostatic problems is

i, 2> jff0(x~yz, 2 (x,y,z) dx dy dz (1-13)

That (1-13) satisfies (1-2), (1-3), and (1-4) is easily verified. The
choice of inner product is not unique. For example, the integrand of

(1-13) could be multiplied by an arbitrary positive function (weighting

function) w(x, y, z) > 0 and it would still be an acceptable scalar

product. However, the particular choice (1-13) makes the operators L
-1 hie(-3, ae h tr
and I:' self adjoint, as we shall now show.

Let and 2 represent arbitrary functions in the domain of L,

and form the left-hand side of (1-5)

<V I = Jj (-E V20 1 ) 02 d (1-1+)

where dT = dx dy dz. Green's identity is

-T fcj*)ds

where S is the surface bounding the volume V. We consider S to be a

sphere of radius r, such that in the limit r--o the volume V includes

all space. Let = i and -- = 02. The boundar- condition is O-iC

as r- , which reauires r 2 g/n--.CI as r--c. Since ds varies only as

r2 the surface integral in (1-15) vanishes, and we have

-ffr - , Nrr-

-4-
Since e is constanlt, (1-11) is symmetrical in a
and and
d02 (1-!7)

We reauired L2 satisfy the same boundary conditions as

to hence L
02l
is self adjoint. If L is self adjoint, so is L " , since

<41';2 > = <l, L- > (--18)

2

The mathematical concept of self adjointneas of an operator corresponds

to the nhysical principle of reciprocity.

It is evidcat from (1-10) and (1-12) that L and I are real

operators. We shall show that they are also positive definite, that is,
-1
they satisfy (1-6). Again we need tc show iZ only for L or L " , since

(1-18) is valid for = 21.2"(The ccnjugate operation is not needed

since L is real.) Form

and use t-e vector identity

9"Q90 V
VO(1-20)

and the divergence theorem. The result is

fff dr -ff F ds (1
V S

where S bounds V. Again take S a sphere of infinite radius, and the last

termn of (1-21) vanishes because of the boundary condition r4-C1 as r--

Now
1,V> (Vlj (1-.22)

and, since E > 0, L is positive definite. The mathematical concept of

 positive defiuiteness :8 often related to the physical concepts of work

or energy. In (1-19), the right-hwd side is proportional to the electro-

static energy.

D. Method of Moments. We now discuss a general procedure for solv-

ing linear equations, called the method of moments. Consider the inhomo-

geneaus equation
L(f) - g (1-23)

where L is a linear operator, g is known and f is to be determined.

Let f be expanded In a series of functions f, f2 Y f5 ' ... in the domain

of L, as
f a.f.(-
i

where the a. are constants. We shall call the f. expansion functions

or basis functions. For exact solutions, (1-24) is normally ean infinite

sunmation and the f.1 form a complete set of basis fAnetions. For
approximate solutions, (1-24) is a fnite siumation. Substi;.uing

(1-24) into (1-23), and using the linearity of L, we have

a. ((f)
2g

It is assumed that a suitable inner product <fg> has been determined

for the Problem. Now define a set of weighting fanctions, or testing

9olction, wI w2' w ... in the range of L, and take the inner product
of (i-2) witIh each w,. The result is

I a ..Lf.> = <w.,g>
w., (1-26)
V. Inmti-fs

j 1, 2, 3, ... This set of eauations can be written in matrix form

as
 ([SJ 1 ] La1 =- [s] ('-27)
where

[L ']w , < 1, 2> . .

. .. .. . . . . . . . . . . (1-28)

0I
(L "2 II <WO j9
:
B (1-29)

If the matrix [8] is nonsingu2.ar its inverse (1-1 exists. The a. are
1
then given by
ta] it[] (1-30)

and the solution for f is given by (l-24).For concise expression of this

result, define the matrix of functions

Ri' =( -3 ... I (1-31)

and write

if [0 [ [
R. -1']I4[g.J (1-32)

This solution may be exact or approximate, depending upon the choice of

the f.
1 and -..
1

If the matrix [1] is of infinite order, it can be inverted only in

special cases, for example, if it is diagonal. The classical eigen-

function method leads to a diagonal matrix, and can be thought of as a

special case of the method of moments. If the sets f. and wiI are finite,

the matrix is of finite order and can be inverted by known methads.

-7-
 A principal task in the solution of any particular problem is the

choice of .
1- and w..
! The f. should be chosen so that they are relative-

ly independent functions and so that some superposition (1-24) can approx-

imate f reasonably well. The w. should be chosen so that they are also

relatively independent aid so that the products <w. ,g> test relatively
1

independent properties of g. We shall say more about this in Section

i-F when we discuss the stationaryI nature of the solution.

Some additional factors which affect the choice of f.1 and w.1 are

W the a-curacy of solution des-red, (2) the ease of evaluation of the

matrix elements, (3) the size of the matrix that can be inverted, and

() the realization of a well conditioned matrix [2].

-I
E. E-aple: Cha-rged Conducting Plate.
Figure !-I represents a
U-e condtctin- plate, 2a meters on a side, in the plane 0

with center at the ori.gin. Let o(x,y) represent the surface charge

density- on t tte. assumed infinite-l'y thin. The electrostatic potential

at uny Iit: in space is

a a
4 ER

where R (x-x') + (-e, ) C . The boundary condition is $= V

(-onsLnt) on -, Ia---. Th intc- ral equation for the problem is there-

V - dx' dy' + (1-34)

'I'I (-Yl
x' <a, : a. The charge density o(x y) is the unknown to be determined.

A pX;LzractV, a"* i"nterest is the capaci tance

-U-
 z

'conducting plate

Fig. 1-1 A square conducting plate-

 c # 4 dx] dyokxy) (-5
-a -a

which can be calcmlated once cx,y) is found.

For an approximate soluti in, consider the plate to be divided into

N saare subareas, as shoew in Fig. i-1. Let the su-bareas be de-oted

AS, As . Define fhnctions

AS
1 As 2 , a
on &s.
II-.

H~ e all other As.

and let a
txy) a. r" (i-37:)

where -a. are constants to be eval uated. In other wirds,w-e are apvroxi-
mating a by a constant over each subarea. if (1-37) is st in

the -integral eqation (1 k and the resultant equation satisfied a4 the

center (x ,y.)) of each As., we obtain the matrLx of' eati-ns

N
vV ." ...J--a A - ,2 .. (,
i-i

where

X..= I dx I
I- !#' 2

Note tht I. is the tentia atth cente of As.. e 0 a --- t car e

-density over "'Dr.)=

-"+ A solution
-s. ,'az of th. s-.uitaneous
hgivesa e-uations
r (d-n&]
t gives

an a-oroxation -to the char- density on the niate a0corA-.ng ta (! 7).

Tne cMrre aD-roxiation to -- -

gives the capmacit%an-eor-:
tte Plate.

To Pat the above solution into the notation of the netind or

Omeflts lIet
f(xy) d(X : (-lu

L(r)=f dx f dyt
, ~' -
IL -a Ar - ~2 o

Then. (1-25) is ecrivaient to (1-3lu). Define thne inner product as

a a
g>fdx f4Y f%'x~y) g-(xIy) (-

The unkzn f'=ai arprorcinted by (1-37) tmee the f.are defined by

(1-51. he eighti fiactions are defined by

a a
<-.:> f dxf dyw(x)fxy)=f.,)(1 )

-a -a

Oiese vjdo not exi:st6 as ordinai-y 1 5 trcti-ons, bu are s cU frac-tions.

Tn particulnr,
rv (x,y)l = 8(x-x.. Ny-yj)(4)

vbhere 8(x) is the Dinec delta t'!%m io taw the elements of the matr-ix

(1-28) are gi-ven by (-),and those of (1-29) are

[gz
I'j-
fV
-11
 Hence. (1-27) is equivalent to (1-38), and (1-30) represents the solution.

For numerical results, the AJi of (1-39) must be evaluated. Let

2b = 2a/denote the side length of each Asi. The potential at the

center of As. due to unit charge density over its own surface is
b b
11
X., I xl dy _ _

-b E +y

2 bi n (1 + ) 2b (o.88137) (1-48)

This derivation used Dwight2 200.01 and 731.2. The potential at the

center of As due to unit ch irge density over As i can be similarly

evaluated, but for most purposes need not be. Usually it is sufficiently

accurate to consider the charge on As i as a point charg-. and use

q b(1-49)

ire (Xj - xi ) +(yj - yi)

This approximation is 3.8 percent in error for adjacent subareas, and

has even less error for nonadjacent ones. Table 1-1 shows calculations

solving (1-38) and (1-39) for various numbers of subareas. The second

column uses approximation (1-49), the third column evaluates 2..

31 more

precisely. A good estimate of the exact capacitance is C/2a = 4000

micromicrofarads. Figure 1-2 shows a plot of the charge density along

the subareas nearest the center line of the square plate. Note that

c exhibits the well-known square root singularity at the edges of the

S plate.

This example illustrates two simple but useful approximations, (1)

the use of expansion functions f.i each of which exists over only a

-12-
 %F0 6

0
C
C0
0

0,

EiEC
lopold/Apusp
6.w0
 Tlab'le 1-3. Capaeitance of~ a unit square plate
(micromicro farad/meter)-

No. ofL C/2a C/2a

subareas (approx. .. ) (eatj.

1 35150 51-50
o 5730 368o
1.0 3820 57(70
36 5920 3870
100 3950

restricted region of space (method of" subsections), and (2) the satis-

faction of the approximatle equation at specific points in the space

(point matching method). These approximations, tirough crude, give

acceptable results for many purposes.

F. Variational Interpretation. The special case of the method of

moments for which the approximating functions f.are equaal to the weight-

ingly functLions w.i is known as Galerkin's miethod. That Galerkin's method

is equivalent to the Rayleigh-Ritz variational. method is well known3,4

That the treeral method o' moments is -also a variational lethod -is usually

not noted, but- the proof is essentially the same.

We first interDret the nmcEthod 01' ino-ents according to the concepts

of Lnca- spaces. Let (Uf) dernote the rang-e olf L, (Lf.) denote the

unw.,
sparce spanned by Lim Lf.,

on
Te 111thod o" t-cients, Eqs.
(.-othe
and (w. denote the space .spanned by the

(1U_-26-), e-quates the projection of Lf

projection of the approximate Lf on /(w,). Figure

I- -eusns tils i'-'tov'al ( denotes approximation to f). In the

o Gierin~ mtho,
speialcae (w) = jf). Because -the process
 /(LF)
L LF
a oxLF , error AL~
Pr~oec--11-0-F0 4l(wn~)

Fig- 1-3 Pictoral representation

of the method
Of moments in function
space.
of obtaining projections minimizes the error, the method of moments is

an error minimizIra Drocedure. Because the error termi is orthogonal to

the projections, it is of second order. This same conclusion is obtain-

ed by meqzis of the calculus of variations. 4The derivation will not

be given here, but we shall sutarize the results.

Given an operator equation Lf = g, it is desired to determine a

fA-unctional of IL
p(f) = <fy,> (1-50)
This f'unctional may be f itself' if g' is an impulse '~ncin LeL

be the adjoint operator to L., and define aun associated fiunc'tion Vt by

Laf 9'(-1

It can then be shown'ta

P f,g'> <f (1-52N

is a variational fonnula with st'ationary val.ue (1-50) when f is a

Solution to Lf = g and f1 a solution to (1-51'. For an approximate

evaluation ofI (1-450), let

andsubtitute ito (1-52). It can t-hen be shown4 that the necessary

and sufficient, ecaations for p tCo De stationary are those of (1-261,

thtis, the method of moments. Minhis variational procedture is knovn

as the iRayleighi-Mitz method, and hence the method of moments I's ident-

The. second eauation of (153) gives us some additional isight into

how to choose the we~h~i~ fnctions. For good results they she-l ad

be chosen so that some lincar combination ct..n closely represent the

 assciaed ield f that is, the field of whatever g' appears in (-5,D).

For example, if we want -. itself, then g' is an, impualse functio4n, and

V is teGreen's function. This is a Doorly behaved functi-on, difficult

to approximatLe byasimni-e setk of w.,. Hence, we should expect calculations

of t.he fMi'ed it'self- to converge less slowly than calculioEJn.3 of othe

PrpmeCer which are as so ciateAd witL.h a we3-behavedf'

i. D. K. Reitanr and T. J. ?T1iag..ins, "Accte Deerntino h

Caacttan-cc of- a T-hin Rectariz;nular Plate,!! ATTrans.? vo.l. 715, pt.4,

DwB.icnht, " les tfntegrals and 0'her Matb-,rna+!cal Data,"

11 an't pany , Ne~w York 1-I?

aviilypo
K-m--l-r
xi l Mehos o I alrAnalysis,"I
trnsate by C. Benster, Jo-hn Wil ey and Sons, inc. , Nlew York,164
u-napter IV.

iS. Jons "A Cr5 tioucof -;-m ariational Meth-od in Sca-titer-

ni:Problems," Trans. ±Rr, vol. A-! , 1556, pp. 295, 01.-
3,no
 II. OPIOJTAlJON -'3? ETZCTROMADJ=-.IC PROBL4

A. Introduction. Mo0 e1ngineering problems in. electromagnetic tner

Consist of' a source, some rn+ar, and a measnr-_.. ent to he rerfor-c-1

Antnnasysem robemsandraar scattering prviees are -v4ca examnples.

Bar'pi3Uly rth qaj.tio-n 2 -- rol'-ra of -h imdat'o '

eah elem.ent Olt Matt-er wi th every oe--r -lernent-, with- n source, and
wih thle meastir~na QCvrice. Th-%e empio
-r. describing ti-he irse an

II oogencjsotperator eouaio
_1.o qainf'.'-A hw

I.n hat-r I, he opera'.Or Mfllation cai be trans f 0 a into a materizx

taton. -ti' -era cti o .],Wrnter- .m;nz-Nf in ar iS inrcer

-
~ of~n-- matt. caateri -s

i dract-erizea by an excitatiofn manl- euln f' theh !~te

-. z z.: - --- a----

~~S~hb ~ ar-Ira it ionexanrdursntrc

z o
o Tr2 W or -ea
-urg divi
--..- -c Jisp-' chrc - zedn an e ai
?' -- trUfl -' l.

1- -o it -su- -:il - re * fl-

rixns lr . Jin-Iic.atc withr armn

ss Ofrrx

nt~- f4ttA hl4- L T1! tJrer --

p? IS ~D'S
** '- '-chs onttsan elrs - -ansf! inteest

W._ FW -- ------

n r.;c--- ie------- P+--

v'e~ _ -'r~tr~
=.rrpn n .04t
* - +a~~n nr~~ -'cAt-- -r - - r

n-- -a- MAond~G w

Z.~=~ -
 feld quantity corresponds to a weighted combination of the network currents.

flielectrac oodies are cha-racter-ized by two gener-alized network matrices,

one independent of the dielectrick vermittivity of the body, and the othe
dependent on. the: permttivity. Electromaitcexiainoftebd

corresponds to voltage excitation of the two net;,works connect',ed in se-ries,

and a N 1 dmeasurement again corresponds to a weighted -comabination of

t'.-he terminal currents. Magnetic bodies are the dual3 problem to dielectric

bodies, and have a dual representation inr teArms of gener&' I zed networks

Hence, they also are characrterized by two matrices,, one inderendent of'

the permeabili +y amA _ e 1 on the permeability.

aerer Electro-
ma~iticexctaton f te bdy corresuonds tocurrent excitation of the

two netwvorks connected in shunt, and a field meas-urement corresvonels to

a weighted comination of the terminal voltages. Bodies havi--ng both di-

electLric and ragnetic nrovet]'s are characterized by three network mtrices,

one of w4hich. is independentl' of the permttivit-y and permeability, one

dependent on the pritiyand one dep-endent on the vexmeability.

Field excitation of the body corresponds to voltage excitIation off some

porits connected inseries, and current excitation of other wort;.s connectea

in shuY~nt4. Field measurement corepnst a !IC.gntect conb ina-tion of th

currents at those term-inal-s excit_ ed by voltage sources, and of! the vroltages

at those termials excited bV current sources.-

The classical cigenfunction method of soltion corresuonds to obtain-

ing a network matrixJ- having only diagonal elements. The in verse matrix

is also diagonal, with olemn wmaal to the recirnrocal of' the elements

of t-a original Matrix inversion is then trval n the infnie

order of the matrix uni-uortant. However, the eigenf[UnctiLons are known

-i-
 fOrnay ;-; felv 0-FCiCal body Shapes, and are oftekldfin~ a-l

If the network matrix can be approximated by a well-conditioned matrix of

finite order, the inversion is readily accomplished by known algorithms.

For example, a high speed computer can invert a matrix of the order one

hundred by one hundred in a few mninutes. The realization of a well-behaved

network matrix of reasonable order therefore constitutes an engineering

soluation to -'*-he problem. This Chapter deals primarily with the formulation

othe desired matrix eciuations and their network interpret'ation.

B. Conducting Bodies. Figur-e 2-1 represents the general problem

of a material body in the presence of an elect-romaagnetic source. For

t'his section, it is assumed that the body is a perfect electric conductor.

Let Z' -be the impressed electric field, that is, the field produced by the
S
source when the body is absent. Let k- be the scatterea field, that is,

the field produced by currenits on the body. The scattered field is

related to the conduction c-urrent j on the body S according to

=L~(~ 'r')(r(Z,Lr') ds'(-)

'where I- %,,')
( is the t'ensor Green' s function relat$-.ng a current element

at r' to its electric field at r. The totL&al field at any point in space

is i+~ and the boundaryj condit".ion is thatt the tangential. components

x (~ ~) =0 on S, where u is the unit no ltoS.ene h

app-ropDriat-1e eauution for determining j is

10 ~x L(1-- xi on S (2-2)

-Where L, is the operaftor defined by (2-1). A suitable inner product

satisfying (1-2), (1-3), and (-)is

-20-
 E E3 u

tsource v
S

Fig- 2-1 A material body excited by an external source.

 Rd (2-3)1
tJ
which is the quantity defined as reaction. 2 ' 5 Note that (2-3) involves

only the tangential components of , since Z must be tangential to S.

The method of moments is now applied to (2-2). Let the carrent &T

be expanded In s series of functions J1 u, . " ." defined on S, as

where the I are complex constants. A substitutim of (2-1) into (2-2),

n
and an application of the linearity of 41, yi -lds

uxZ L")-x W-4 -

25
n e%
n

Now define a set of testing f-unctions X, , , ... , which are vectors

tangential to S. i.e., they are current-typ-pe vectors. The method of

moments requires that (2-5) be valid for the inner product i.th each - x

that is

for all m. The unit normal u nas been dropped from (2-6) since ) and

4are a!:,. +aqentii to S.

1:07s define the fo..wing tr

.. . - .
( 1.

[z.3= E'-42
-V I - "V.,L?

-22-
and rewrite 114-6) in natri4x fain- as

Due Solution frthe evtmunsion coeff i cents i

r -

t~-mti nes of-7- . o~ezj- t- mat:, of C.li.. n.

Einfl

Ani I--+hi; Te-

M-4e4ie]

-- Ek reP-u 10 hr~~a
e c- c. ~ erttfo_fo

vtjge trratin aa aam ei-'.s

N-port newok ~ - Thi ~s-!Yri s-a-zThF- 2 M

aOud~fla anove ji
the corrinese.ncageoitagde, curet aprc matritinx
c atri

basia.-1y Ld-; evnT eon'~ r ed by-r v o eea

L eg -% n rf -o

am - t -jn n
esse ffr~.~~ncac hi tbei ssz e ~-i-:l 76

a-1ato- -- U -,g h

a.-nrnnsadth ep'ori- fixnid-;W-oilee-an S -ho

y~lt% citaio of tthe net:xu-7,

accoaingto(2i-

--- '
 +

+ 1'-mnJ
V4) depends on geometry
+ 3 and wavelength
V3 O

Fig. 2-2 Network analogue for a conducting body in

an arbitrary impressed field.
 C. Dielectric Bodies. Now let the bony of Fig. 2-1 be a dielectric

of permitti which ma
c.ty .A M' p.i:t!ir' or even a tensor.
1or

Again the impressed electric field is denoted by and the scattered

field by Bs. The scattered field is produced by the polarization currents

J over the body according to

= L( £ ) = fff (r') re(,r') dr' (2-14)

V

which differs from (2-1) only in that the integral is over a volume V

instead of the surface S. The total field at any point is given by .i + ,

and within V the polarization current is given by

= - e)( + ZS) (2-15)

where E is the Dermittivity of free space. Using (2-14) for s, one

can rearrange (2-15) into the form

L() _L in V (2-16)

where Ac = E - eo This is the appropriate equation for determining ,.

The left-hand side of (2-16) could be redefined as a single operation on

j, but it is more convenient to consider it as the sum of two operations.

A suitable inner product for thi. problem is

jj r E dt (2-17)
V

which is again a reaction.

For the method of moments, let the polarization current j be expanded

in a series of functions Ji' j' "3' defined over V, as

, ZnV, (2-18)

-25-
where the In are complex constants. Substituting (2-18) into (2-16),

and using the linearity of L. one obtans

Z InILe ( ) - ~=~(2-19)
n

Define a set of testing functions Jl ?-' 3' ... , over V, and require

the equality (2-19) to hold for the inner products vith each M" Hence

n <W ,LJ> - -.- -<Wm

n

for all m. Now define matrices [In] and [Vm] as in (2-7), and two

impedance matrices, one [Zmn] as in (2-8), and the other

.,m] W,Jl/"> <WVJ21t > ...]

2 'J /
'-W - 1 L- <W 2 'j 2 / - (2-21)

. ..
S I . . . . ....
.

Equation (2-20) can then be written in matrix form as

[Zmn + Zn [In] = [Vm (2-22)

The solution for the expansion coefficients In is again given by (2-10)

where
[Yk = r(ZR +Z -1] (2-2)

.nterms of the matrix (2-12) of expansion fuDctions, the polarization

current is given by (2-13). Again the solution is approximate or exact,

depending upon the n and -W.

In terms of generalized network parameters, one can interpret this

solution as two impedance matrices connected in series as shown in Fig. 2-3.

One matrix, [ZI, depends only on the geometry of the body and the wavelength,

-26-
 07 _ tZmn]
dependent 1113 independent
on c of f

Fig. 2-3 Network analogue for 0 dielectric body in an

arbitrary impressed field.

-27-
=d the other =atrix, rWi et p~tti Tj y
L'EJ)P depad alaoon
-

The impressed electric field determineb the voltage excitation of the

netmorks connected in series. The resultant terminal currents then

determine the coefficients I n of the polarization current. Note that as

e--ac, the network analogy of Fig. 2-3 reduces to that for the conduct-

ing body, Fig. 2-2.

D. Magnetic Bodies. If the body of Fig. 2-1 is magnetic but not

dielectric, the problem is dual to that for the dielectric body, discussed

above. To be specific, let i denote the impressed magnetic field, and

s the scattered field, produced by the magnetization current X over the

body. The relationship of to M is the dual to (2-14) for the dielectric

case, that is,

= LX1)= r(Z') d ' (2-24)

V

Let g denote the permeability of the body, and -xco that of free space.

The magnetic current of magnetization is then dual to (2-15), or

= jW(V - g.o)(Hi + s) (2-25)

Finally, combining (2-24) and (2-25), one has the equation for determining

M
L -4 ir. V (2-26)
jW4L

which is dual to (2-16). Define an inner product

To <'H> - 9d-r(2-27)
V

The minus sign is certainly not necessary in (2-27)- but is used so that

the inner product corresponds to that defined in the reaction concept.

-28-
 Tv apply the method of moments, let the maietization current & be

expanded in a series of functions , , ... , defined over V, as

9vj VZ41 (2-28)

where the V n are complex constants. Fr=n (2-28) and (2-26), it follows

that

Again define a set of testing functions F F' -3' ... , over V. and take

inner products of (2-29) with each Wm . Then, dual to (2-20), one has

n,!
Z n
m 1 <mH> (2-0

for al-I m. The following generalized network matrices are now defined

V1 1 [In] = - <WiHi>i

- >(2-31)

inn 111I111 W1IV? .

< 2i 1> <WT'2 .fV (2-32)

. . . . . . . . . . . . . . .

. . .>/ ,- >
. 1 21 . . .

I 2'L> 2'-2'P (2-33)

The matrix eauation for (2-30) is then

[Ym + r ] Vn] = [r 2
fy LM (2-[4)

-29-
which is dual to (2-22). The solution for the [ ]is the inverse of

(2-34). The magnetic current of the maetization is given by (2-28) once

the In are found.

The network representation of this solution shown in Fig. 2-4

is the dual case of Fig. 2-3. The matrix [Y I depends only on the geometry
mfl
of the body and the wavelength, and the matrix [Y I depends on the
mn
permeability. The impressed magnetic field determines the current excitation

of the networks connected in shunt. The resultant terminal voltages

then correspond to the expansion coefficients V for the magnetic current.

n

E. Bodies both Dielectric and Magnetic. If the body has both e and

different from their free-space values, a combination of the preceding

two analyses must be used. For this purpose, in addii.ion to the relation-

ship between electric current and electric field, (2-.A), and magnetic

current and magnetic field, (2-24,), one needs the relautionship between

electric current and magnetic field

. = () - if/ ,{) Pr(r,r') dT' (2-35)

V

where "(rr') is the tensor Green's function which relates an element

of J at r' to its magnetic field at r, and the correslonding relationship

between magnetic current and electric field

z= -Nz)(2-36)

where N is the same ouerator defined in (2-55). The minus sign difference

between (2-35) and (2-36) reflects the minus sign difference between the

two curl equations of Maxwell. The integral equation fbr the problem

is now a matrix of (2-16) and (2-26), with the appropriate interaction

termis added. To be explic.it

-30-
Im

I, Vl I

^mn]
dependent
I *1,
~

11I v
£ymnJ
iindependent

Fig. 2-4 Network analogue for a magnetic body in an

arbitrary impressed field.
 e N1 ?A] P
P&l r

is the equation to be solved for . end M When the body is both dielectric

and magnetic.

Equation (2-37) is just a more complicated case of an operator equation.

This is evident if the following matrices are defined

g (2-58)

C (2-59)
L .

71 00

and (2-37) rewritten as

(f) + ;f) =-g (2-L)

Of course, Ir and 7C could be combined into a single operator, and (2-40)

written as (1-1). The appropriate inner product for this problem. is

<).',g>~ gd =
=f f fi - E - M- F) d-r 2-i
P:j JI4H -B i(-1

V v

which is precisely- the general definition of . 2,3

A solution by the method of moments is obtained as follows. Define

%-__.ectric" expansion and wei±hting as
.ctions

-= IJ'" , = We .
=i n n .

10 f o0 (2-42)

and "magnetic" expansion and wei-hting-- fn tions as

g-i-nfu-a

-Jm-
 lo w =1 0
n n 2-43)

The expansion for f is then of the form

f U
n
I
(fe + Vn) (2-4h)

where the I and V correspond to the a of (i-24). Follow-ing the method

n n n
of mments, one obtains the matrix equation

F+:
L^:I: +1 (2-4)
Here the various matrices have the same definitions as in sections II-C and

flI-D. The additional matrices [Bmn] and [Cmn] describe the interaction

between electric and magnetic currents, and the superscript i has been

added to the source terms (right hand side' to distinguish them fromt

the response terms.

The general-zed network representation of Eq. (2-45) is shown in

Fig. 2-5. The network denoted [Lmn] again depends only on geometry

and wavelength, not on F or The network ,Z I in series with the

V.
mn-
voltage sources depends on e, and the network rn in shunt with the

current sources depends on u. The impressed electric field determines

the voltage sources according to (2-7), and the impressed magnetic field

determines the current scurces according to (2-31).

F. Measurement. Figure 2-6 represents a general problem of electro-

magnetic engineering, one consisting of a source, a receiver, and material

bodies. Explicitly, Fig. 2-6 shows a source connected to a transmitting

-33-
 M[Lm]

dependent_____inepndn

on cadt
I~mn]

and mantcprpris
dielctri

ayn
dependent
# v
receiver

source

Fig. 2-6 A typical electromagnetic eng-naering problem_

 I

antenna, a measurement device connected to a receiving antenna, and a

scattering object. However, any of the material bodies mav be absentj

her ce a general problem includes the transmitting antenna, the receiving

antenna, and free-space scattering as special cases. The source and

receiver may, of course, be at the same point in space, as, for example,

in an input, impedance proble-n.

The voltage at the receiver terminals due to a current source at

the transmitter terminals can be expressed in terms of the transfer

impedance Zrt between the two sets of terminals. 1 Y3 In terms of reaction,

this transfer impedance is given by

Z = ft Et J - Ht ) d. (2-46)
rt

where Et,H t is the field produced when I t is applied to the transmitter

terminals, and j, Mr is the current resulting when Ir is . - lied to the

receive?" terminals. If e and 4 are scalars, or even symmetric tensors,

the usual reciprocity theorem applies, and (2-46) can be rewritten as

zt =trJj (E - • Mt) d (2-47)

Rr , Hr is the field produced by Ir appli-ed to the receiver terminals,

and Jt, Mt is the cur.-ent produced by I t applied to the transmitter

terminls. The important fact to note is that the measurement is of the

form of an inner product of the current caused by an excitation of the

transmitter with a field caused by an excitation of the receiver. Hence,

in the general notation of the method of moments,

measurement = <f, gr> (2-.8)

-36-
where gr is function depending on the measurement to be performed. For
rr
example, g is a Er, Hr in (2-47). Equatior -2-48) is the general expression

for a functional of f. In terms of the solution obtained by the method

of moments, equation (1-32),

measurement = <f,gr> =[ (2-49)

where [r I, '
Lg = <!,>t
n
Lg- <ff-1 ,r>!
>I

19i

This measurement matrix is of the same form as the excitation matrix [g]

of (1-29), except that the inner products are with the fn instead of the
w
n
As an example of a measurement in a field problem, consider a conduct-

ing body in a plane-wave incident field, and let the measurement be the

bistatiz radar echo at some distant receiver. By definition, let Echo

be the field quantity whose malpitude squaxed is the conventional echo

area, that is

a = jEchoj 2 (2-=1)

A formula for Echo is obtained by letting both the transmitter and receiver

of Fig. 2-6 recede to infinity according to an appropriate limi'ing

procedure. The result is

Echo - pA " "t ds (2-52)

where - ,k = 21T/k, k is a normalized (unit amplitude) plane wave

from the receiver, and j is the current oe the c;n t scetterer when

excited by a normalized plane wave from the transmitter. The cuirent on

-37-
the scatterer has already been determined by the method of' moments in

section TI-B, the result being ±j ±LV JLA* i

(2-52) gives

Echo [Y (Vi (2-53)

2 lr lf i

where lVr] is given by

n 0'
- 412 ,Er> (2-54)

This Is the same forin as -the voltage excitation matrix (2-7), excepDt

that repDlace the V-In the Galerfkin method ( = ) the two voltage
m..atrices ['] and [V ]Are just different excitations of the same generalized
n n
network. In the method of' moments, [Vr] may be interpreted as the excitation

o--, an 'ad-oiiit f-eneralized network, defined as the transpose of [z ].

G. Discussion. The method of momento is so general that almost any

-oluti-on can be inter-preted as an application of it. The use of eigen-

fktions for t1he expansion functions and adjoint eigenfunctions for the

weighting functlions is the class ical eiigenfunction method. Point-matching

procedui-csc. that is, satisfaction of the operator equiation at specific

poi~nts on u body, i-rc- i-u i valent to using iinpulse functions as weighting

-uneti-ol-s in the method of -ilomer -s- Methods which approximate the operator,

for tevznple. the miethod of nsar!e interpretable as special cases of

mehod
tA1~,~cera Heceone should view the method of moments as a

uniltyin,' convet ra'ther than as a particular techinique of solution. In

apprlying t~h method to engineecring problems, one is confronted with many

possible choices and approximations. The particular technique of approxi-

mation called the method of subsections, introduced in Section I-E, has

been found quite useful for digital computation.

Basically, the method of subsections can be thought of as dividing

the object into a number of pieces, and calculating the interactions

according to the method of moments. One of the first applications to time-

harmonic electromagnetic fields was the calculation of scattering by

square conducting cylinders. General computer programs for basically

the same method, but with somewhat better approximations; have been used

for cylinders of arbitrary shape. 5 Solutions of this type have also been

obtained for dielectric cylinders,6 for conducting bodies of revolution,7

and for wire anternas of arbitrary shape. 8 An analysis of wire antennas

and scatterers of arbitrary shape, both loaded and unloaded, are formulated

using the methods of this section in the next chapter.

In terms of the method of moments, the method of subsections involves

using expansion functicns which each exist over separate sections of the
body, and using weihtig functions which test the fielid in each section

(often by Doint matching). In the case of conducting bcdies5 this is

ecuivalent to obtaining the ne1 -f-impedance of each subsection of current,

and the mutual impedance of each element with every other element of current.

In the case of dielectric (or magnetic) bodies, the interpretation is

similar, except that the self imlnedance (or admittance in the magmetic

case) can be divided into two comonents, one dependent on £ (or ti) and

the other independent of e (or g). The mutual terms are always independent

of e (or j) Hence, the careted networks (left-hand ones) of Figures 2-5S,

t;n
-_- -s "'- is xr ss1-
r-4 -a
2-4. and -become nonin eract.ng elements, t is, expressible as diagonal
matrices. For exalmple the network representation of Fig. 2-5 becomes

that of Fie. 2-7 when the method of subsections is used. The effect of e

- -
 V.

depend22
OfP-7

* [LmnJ
independent
A/; of e andp

depend__ _ _

Fig. 2-7 Network analogue for an arbitrary body when

the method of subsections is used.

-40-
and p is then just a loading of the appropriate terminals of the general-

ized network for the body.

The use of these techniques for computation gives rise to a number

of practical difficulties, which can be overcome by careful analysis.

For example, the field of a discontinaous distribution of current has

singularities, and the order of differentiation and integration cannot

always be interchanged. The usual tensor Green-s function often cannot

be used, but must be interpreted as symbolic of the more fundamental

vector and scalar potential formulas. 9 The accuracy of a solution and

the ease of comatation depend to a large extent on the ingenuity and

care used in the formulation of the problem. The method of subsections

is basically a small body technique, since the matrices involved become

too big for large bodies. The treatment of large bodies requires great

care in the choice of expansion functions and testing functions, and

even then the interaction matrix usually is valid for only a wubclass

of excitations. Perhars the best approach to the large bodies is a per-

turbation one, that is, the use of an approximate solution as one of the

expansion functions. The method of moments then gives a correction to

this approximate solution. A disadvantage to this approach is that the

generality of the solution is lost, being valid for only one excitation

of the body. M.uch more work remains to be done on the large body problM.

H. References.

1. R. F. Harrington, "Theory of Loaded Scatterers." Proc. EE (Iondon),

vol. III, no. 4, April 194, pp. 617-623.

2. V. H. Rumsey, "Ibe Reaction Concept in Electromagnetic Theory,"

Phys. Rev., ser. 2, vol. 94, June 15, 1954, pp. 1483-1491.

3. R. F. Harring+ton, Time-Harmonic Electromagnetic Fields, McGraw-

Hill Book Co., New york, 19!, pp. 116-120 and. ;-W-371.

-41-
 K. K. Mei and J. Van Blladel, "Scatt-ering froma Perfectly Conduct-
ing Rec tangualar Cylinde-rs , h Trans. , VTol PP. -- ,Y,

5. M. G. Andreasen, "Scattering from Paxa!Ill Met '1n

with Arbitrary Cross Sections, WrEE- Trans. , vol1. PIP- 12. joverbe t -6
PP. 7k6(-754.
6. J. H. Richmr-nd, "Scoater.ingr by a Cylnder
('!r1--i of" Arbitrary--
Cross Section," Tk&E Tra-ns., vrol. A-5 a 95,p~5'

Ixctn .G £ nrae, Sctern from Bodie%$ f R -,- mn byariLy

Exa--t,eet-o-l, 1 RP Tas.,-e- I1 AP-D A -. March r~

d.K. K. Mei, "On the Integral Ecupationis of Thin J eAtna

1rr
71 a5 rr1,* ; r --, ;A Z71

J.G.Van Bidl "Sm Remrk on GrP-! Thr- fr Infiit

pae, T1RZ Trans., VOl. *
IL Zitt lQ±,
itveb -prr -6

U1Pw
 ±1±-r. WJSLAN" ttantp WLAThILIX? i ' ARBITRP3RY SHAPEL

A. Int-roduction * general a nays

S', S o0 tf*lmt

conaductina vwires wnd siutdi an arb+ rxr

S -- r. r -F +h( fTYr hp vara, _ '-a4 d - p

Ainr ±4zr b-etween -

antenasan4 i--
satt~~rlntrljasa -- -.
0!per-r -'ZSlf

-e iiIlf th IS'inf
- . ul-

r-Usr.Tri
zn~~irpap -Cm~e r~* ed.a ft 5______t

.a m-5- -U -

7, c-.i-U- vrn are ar al -

intearl eoratyon iL'nhiar tchnioucsT- ta ral -0 t-airi - fz

-eals - s. - -. ~in el -s - -eu to

-s -as ct.u zt avs rnp

z ca in sn ~Y:- -~'
con.V- I - --- n Inc-"Z nr

cccauseits _0y
cnnens.a rIr±ct - n-~ Vclt~l eUa ^eoner

arDS --1 l0- e13rI jTT eaic a~ -~

, t Wfli (a r a. a - - 'ra- -

-- tipf -rx ;-------Z

-,- nflnm-ri- - w~n~

n-n t- e Is can e
uthudt ofa--b -n

itn 0c-l Uul- -v -- .o 7-lTa +hp. -tw

.- a 4- %lt~r? -inC___
souice
tl .. .~ nla.a szcr r-3
 acurrent soureo- eAn-h port in-in n alciiati th ove crcit

voizages at al31 ncort-s. Ths procedure 4 vvs02Vcurrent elmet

r i

einnty pace. Vine A.~+ _ _ . n inverse -y1 tn--e imnn-dn.cearx

ar-t- nceurMa rx is M the po~rt c'r- et s ncurnman -Mc ibtr1

onthe wire) r %-A "rA +

-n n

np- con c n oavrr Sim.a

Zi -Za- as ~ . r.- ~ ~ JA~A

Zs- T z - - - C-- -

- -

- - -- --
* - -fl ;.o

omvintneinre-tin ~ ~ -s era rdrz

w rc an- -lir rafl

SaCsa:-: ann
 N

wire axis

T)
(a) (b)

Fig. 3-I (a) A wire scatterer- (b) The wire axis

divided into N segmnens.
the wire axis. 5)The boundary condition (-)i ple otea~a

component of P at the wi re surf:ace.- othi s apprximtoE . (-)to

axis

:MV s

A -T?

-~-~~ ---- arnR te

ace wre anis an 11'Cfr r

a jzf.%AZT on e-

-n.l -s,- nvc.

as txavataa Sleme-1over

~
- .,-...n_~_ ~ ~ 4 _ - -5A

- - - - S - e ( E-4 1 0 2 tI Y - m..Y -

-- m V%~r -- new rfl-rn n- - m -

- - - - - Tin -2 en

are 1 -CM

an i- Z.

--
n-i C~fS1~a
n~ ~C - . ..--
-
r~t~~-1 -
zmnnn1-r.-S -
 J-II ~ -k-
d At
'FO-di

u- ite - n-t m th -," cm On en -.e~ t1 b.aN

e * &s abe 10)v as ther euos ther (3-5) id orre

an Ny wi ymi
tairrmsv0; e =1e oec tnaLZ WJXW!k

en yd-
e. i- - -zle

- - -

one
can e -1 ainrces a

V.-

oea- --- a -' , h

'm-Izi =lm- All!; cly [idIr lai r4IIr w -d bc~

issoeb asir~ ~ jrPo
a)w
 t
Consider two represefttative e..eizt.
r 'L 1

in Fig. 3-2. The integrals in (3-11) and (3-12) are of the same form, and

are denoted by
1 :r--jkRm
*(n,m) CI
1- J R (-16)
n mn
A9n

Symbols + and - are used over m and n when appropri.ate. Evaluation of the
*, in general is considered in Section III-F. Let element n of Fig. 3-2

consist of a current filament I(n), and tuo charge filaments of net

charge

q(+n) £I(n) q(ii) I (n) (3-17)

where q = caLn.The vector potential at m due to I(n) is, by (3-11),

A = (n) in 4(n,gn) (3-18)

The scalar potentials at m and R due to the charges (3-17) are, by (3-12),
+ +.+ n+m
-
"--+ '
(
4we n'('J
([In)
t(npa) - I(n)
(3-19)
= e [I(n) - I()
i(n) $(,)

Substituting from (3-18) and (3-19) into (3-10), and forming 2mn=
(m
im/I(n), one obtains

Zn = jwgbr -m 4(nm) + .- (n,+) - *.-(F.,+

t.+ (320

p-kn,) + *(Em)] (3-a0)

This result applies for self impedances (m = n) as well as for mutual

impedances.
The wire object is completely characterized by its impedance matrix,

subject, of course to the approximations involved. The object is defined bj

-h8-
 r
+
AAm m

Fig. 32 Two segments of a wire

Scatterer
21+] points on the wire axis, plus the wire radius. 1he impedance

elements are calculated by (3-20), and the voltage matrix is determined

by the impressed field, rxcording to (3-14). The current at N points on

che scatterer is then given by the current matrix, obtained from the

inversion of (3-15) as

[W] [Y] IV] Yj" = [Z] 1 (3-21)

Once the current distribution is known, pardameters of interest such as

field patterns, input impedances, echo areas, etc., can be calculated

by n-merically eval-uating the conventional formulas.

C. Wire Antennas. A wire antenna is obtained when the wire is excited

by a voltage source at one or more points along its length. Hence, for-
an antenna excited in the n-th interval, the applied voita m

is [v s ] -0

Vn;
V (0-22!

0 I

i.e.. all elements zero except the n-th. which is equal to the source

voltage. The current distribution is given by (0-21), which for the [V]

of (3-22) becomes
[I] =Vn Yi

4n
(3-23)

Hence, the n-th column of the admittance matrix is the current distribution

for a unit voltage source applied in the n-th interval. Inversion of

I
I
the impedance ;atrix tnerezore gives simultaneously the current distribu-
tions when the antenna i-s excited in a arbitrarj interval along its
I z os- Lm

length. The diagonal elements Y of the adittance matrix are the input
admittances of the wire object fed in thz n-th i-terval, and the Y are
mn
f ,nces
,l.et between a rort in the n-th interal and one in

the n-th interval.

The radiation pattern of a wire antenna is obtahr-d by treating the

antenna as an array of N current elements I(n)A.n By standard forncaIdas,

the far-zone vector pote1 '3 given by

gi

rr0o jA os
=leti Z I(n) Ati e n (5-2k)
n

-where and are the ra'l 4 us vectors to the distant field point and

to the source points, -re ectiveiy,

4 and En is the angle between

zo and: ?. far-zCne field components are

Th

=9 9 =jA it (5-25)

where 9 and $ are the conventional spherical coordinate angles.

An alternative formla for the radiation pattern, more convenient

for comptation, can be obtained by reciprocity. Figare 3- represents

a distant ctn.rent element I (subscripts r denote "receiver"), ad.justed

to product the unit plane wave

-Jk rn
=u e 3-2C1

in the vicinity of Une antenna. Here u is a unit vector sDecifying the

pcl. Urization of the wave, k r is a wa-e number vector pointing in the

direction of travel of the wave, ana r is the radius vector to a point

n on the antenna4 By reciprocity,

 nrr antenna

x V

Fig. 3--3 A wire antenna and distant dipole.

 r T (3-27)
antenna

where E is thore ucomponent of P ir-on the ant-enna, alnd I -is the clurrent

on the antenna. The -constant 1/12: is that. needed to. nroduce a n1,lne

wave of unit ainnlitu.de el the cn--4giwihi

te
LL 0 (z -,pi

nu
mmerical appo iton to -7) is obt"Lained by defining a -voltage

matrix
V- -

weeZ J im- byaniC

r I-
* I-

0 t.0

where [LVJ aenotes z%,as t.ranroDose of 1V..

Note that Iisthe Sa' t = matri.;x. as ) thatS, A. is the 'vo-tarze

matrix for D'&ne-ave exy:K t.ation of- the wire. S'atiation C(3-30' r'l

v1 dfor an-oiblzrary excarivon RISJ1 i rr-- v- s-ric~- to ~

the

sin-gle source excit.ation 3-2)

The power gtdn pattern for- th- u component of teradiation frield

igien b ~r E(

Jr-in

where ij=I~ is the initrirsic impecance of spacep and J the poer

is
-in
anvt to th e ante na (dentt- conjugate;
 In
B.=
e I;'I [i*j Re hi 5 ]I9*j)
Cr] (3-32)

For the special c;-se (3-22) of a sin-gle; :ource, P. becomes simply

in

Re~vnI Yn). -Using (3-30) and (3-32) in (531, ri has

g(G)= 4e 1 (i(9]
j

where [r ))is given by (5x-29);r for various angles of incidence @,$

If tetotal power gain natt-ern i;s desirtt, the g' s for two orthnogonal

pej'--ri-Zatiorn may b-e added tzovj-th.

Di. Wirwe Scatterers. Oonrs~aer now thne fIeld Sca+-nread yawr

objectL
a in ild.-aeicdnFigure 3- rtmaresents a sca-!tt-erer

and two -distnt+ czarrent eements, I -47 transmitting Point _and

Ti at the receiving point -.The IL- is adjuasted to pwd:-,,uce a uni-t

n±an"e wave at the scatterer

wehere the cotation.- !*s anaiaogous to that of- (5-) he vol-tage excit*Jation

(-14
matrx isthen

acidL curn Uhe -is giv-en by (3-21) with [VI,=tti The f1i-old uLroaea
by[icnthen be found. by conventional +-;dhi crs.

The distant scattered fiold canfltz~o be evaluated by- rcj-pr.C~vtty h

same as in the anencase. A dipole -1- at the rec-eI-i"g point is

 z
rr

wire scatterer
x

Fig. 3-4 Definitions for plane-wovs. scattering.

adjustCed +npreduce the wuit plane wave (326)" atthe scattere-r. T-i4e6

scatArndft ield is then. given by (55)with r repla-cca by [ ,ta

Apretrof" inte-rest is the bisticcross sec-J.- U-

dfned as that area for -Which Ifte incident aecnan sfiin oe

thiS is

c=4~rr

For the -. c-nosrtic- cross secr"-on. set =(V]i (-f.T0 '-v

section d-enns on t-he polarization of the i nctaen waeAa n

receiver. A bett-er descrintion ofT the scatterer Can beMade= in tIerms

ofMf
scr a mari.

Anohe
naamterofinte-rest i s -thj- total scattUerinc, across szenc-3

qdefined as rae rat-io of1 th-e total- scarrereic. power to- the MOWe-raesv
ofthe in cidn wave 77n., rr-
IT-0 -- -,rn 3y0--

for any eitin;hencie the scattered Power is iien y(-2-wt

Br]rep-laced by [V.t]L. The inci-aent plower density is irso

C-=r, Re - v ri['i:(-5

Noe oa
a 1 dpedeton the polarizatio:n ofr the incident wavle.

E. Discussion. ToIne solu1at*ivon of, this chanmter is a first order

solution 4- ihe anproimate integ-rUI eauaition, (5) Inth

solution 1±) the veto otential was evraluatedU using a stenu funct-ion
__ __ cti -t 'W- t-e ma - _n__ rh: teta-wa

using a sterD fctilon anr#.ar n the. 5 s~'

-e b -.I-I% ~ tn

C~--~_- & - nr -im 4- fr ~ nt... tn

usc.
tt.~ trse~ a jf~jsrDonf 1-a~ nh 7 -wsn =d--. m -mg

..--. t-------t-a- -u- n -

-M- -OM hi -1 - - -

ea- - _ -

--- s~~~--' m- aIt--r-em

a

*d -. ---aS T z-- Z

- -a - _ e, at rl -- m of aw~ wi-e trean c -

asa- -rvl w . w z j.

-zcj w-Z I g- S -hd

Z., - = t-&S t- % nr-

-i-ha~ -iz~ -ore - I - -Ar Is 2-0

t'ra v-al suciamsubroutie Whil e thei rvdfiae- 4uc__ the

-a-i- u-j

thewlre-it-jct aV'u %on the basis cAexy nenrr- 11'pers-a

fl -rg:j

urnmer Soluti1Z.onl W-tfl no0 SttC..o. t'CL..a.S h.rn

?e-t- nee nnr -rti.%-

atfttlIj:-:f~nz-p rri !re
'L- zonle

oua-t-ities. such -as randia-tion- nattern-s ana epgnaz ar-ea-s, Amhich ar-e re-Latn-ei-r

enl
1S iV to =--all maiisons rte~cretcsrbt-n
 The solution is in the form of an impedance or admittance representation

of the scatterer. Hence, the general procedures for determining scatter-

ing from loaded bodies, available in the literature, 2 can be used directly.

This means that the behavior of wire antennas and scatterers loaded by

lumped impedance elements in any of the N intervals can be calculated

from the present solution by matrix manipulations. Furthermore, the effect

of finite conductivity of the wire, or, more generally, arbitrary boundary

conditions at tne wire surface, can be analyzed by considering the wire

to be appropriately loaded in all intervals along its length.

F, Evaluation of_. An accurate evaluation of the scalar * function

of (3-16) is desired. Let the coordinate origin be located at the point

m, and the path of integration lie along the z axis. Then

a -jkR mn
*(mn) =
(m~-n) f e Rmn dz' (3-39f
~a R
where
2a =a n (3-40)

)P2 + (z - zI)2 m n
%nn
R, (5-41)
a 2 + (z') M

and a = wire radius. The geometry foi. these formulas is given in Fig. 3-5.

One approximation to the *'s can be obtained by expanding the

exponential in a Maclaurin series, giving

k2
1 (i _jk R + ...)dz' (5-42)

The first term is identical with the static potential of a filament of

charge. The second term is independent of Rm. Hence, a two term

-58-
xz

3-
emtyfrevlaigqrm)
Fig.~Rm

+-r9z
 approximation to (3-39) is

1 log z r+ + (z+a) 2 jk

I1z -a+ jp/+ z-)2J

If r = +z 2 is large, then
e-jkrA-4
4(mn) = e (-44

For a first order solution, one can take (51'-43) as applying for small r,

say r < 2a/3, and (3-44) for large r, say r > 291o3.

For higher order approximations, more rapid convergence can be obtained

by taking a phase term e - j k r out of the integrand. Then

e -j a -jk(mn - r)

-a
)2
- jkr --
C (i_ jk(R-r) k2 (R, -r
R+ ... )dz'
-az
(3-145)
Term by term integration gives

e-jkrk - r
-e 'T-
a7
,. mBn;,--- I1 - 2 -r1)

2
~-. -2r1,2 + I

+ kj 3r 12) r1)
4 A ++Il)
r12 + .. (3-46)
Uwhere
+ a + + (z+ a)(-47)
'1=
log L + p +~z 2 (z347)
 -2

r 2
a+ /12 2 au-z /2 ~2
52a ~ ~ ~2 - p('z
.jp Zj a -

IL= 222

An expanso othtye (-6)itheoret i-cally valid for all r, but it

rails numericafljy for in,-ge r because it involves subtract-ions of

num-be-rs almost equal in nagnitude. For p < a, on;,- should set p =a in

the expansion.

.An expression su-itable for large r is otained by expanding (3-39)

in a M1-aclaurin series in z'I as

=IO + r'(ozt + 1flQ) + ...'J dz'

-a (3-5')

where
P-3K -P + (Z -z')2

*4 (z - t

When. a five termn ap-ansion of (55)is int--egrated term by term.., there

results

= ir A jkaA + ka) 2 +u jka 'A5 + (ka)+A,) (5-2

wnere
=1 1 (2 + 1 1a 7;01z +
5(I - -- ) [5 -( ri I-;
55(7
0 o40rk~

[-1 + - 30(G)+
132]+lcz5[
6r r rr
 lz
1 2 1c2 2 z 4

A3 (Cx) [+z( Z)2 4

454
A,, - Y
.. 120r

For accuracy of better than one verCent, one can use 3 i611
0 for r < 1:00.

and (3-52,1 for r >~ 10a.

An alternative derivation of t.he +ype of 3-2 can --e obtiained as

follows. For r > z', one has the exoansion

-jkRI L
Mn
_____ -'~(2n+) fkz' h (kr! P(.
nn nr'

whllere jare thle siuheri cal- Besse'! 4 ~cin ofP the first*kin,~ (:'
are the stpherical Hakel functions of the second kidrn P. ~) are

the Eezgendre polyomiJ.al. If (3-5k Js substituted into (3-3) no

integrated temby tlerm, t;,here;- results

-Where
kcc
b
n 2a
/I~ 'on
; W dx f(3-56)

Equatc~n (-55)caii -be rearranged into tmhe form olf (-2.atog h

recurrence forrnulas for h(2 n P. make com Utation directly from(-5)

nn
aiost as easj,.

~'. Referyencez.

.P. T'Harring~ UW Time-larmonic ElectromaziOtic Fields," McrGraW-

Hil Boolk (.,New Yfork, 16,Clmaaps. A.and 7.
QU......g~n "Theory c+' Loaded Scatterers," Pron. TZE_-,b(London).
V~l. a1, no. l96i4, pp. 61 -623.
-,April1
 TV. CALUI!ATIONS FOR INAR WIRES

A. Piecewise-linear C(irrent, Galerkin's Method. A linear wiri antenna

or scatterer is one constructed of a straight piece of wire, or of several

colinear pieces of wire. Again the wire cross section is assumed circular,

and its diameter small compared to wavelength. This is a special case

of the poroblem consiIered in Chapter III, and some calculations wire maxe

usingq that theory. However, to obtain faster convergence, the f2al calcu-

lations wrre made using a piecewise-linear approximation to the current.

This is eu-ivalent to using "trirgle" functions as expansion fur.ctionys

in the method of moments. For this section, the weighting ftnct..-ons were

taken to be the same triangle functions, and hence the solution 2orresponds

to Galerkin's method.

Figure 1 shows a straight section of wire, and defines the coordinate

system. The wire extendb from z = 0 to z = L along the z axis, and is

of radius a. it is assumed that only the axial component of the current

on the wire is significant, and it is expressed in terms of the net current

I(z) at any point z. As discussed in Chapter III, the problem is represented

by the o-erator euation

where E is the z component of the impressed electric field at the wire

surface. The operator L is determined by the usual vector poItential method,

and is given by
1 ) dz'
LI+ (d k- L r-'
_z (4-21

where

-65-
ANGLE OF F2A-Z~ L
INCIDENCE

Z: 0

Fig. 4-1 LINEAR ANTENNA

 zr2'2ak LA+ si2
G(x)=- dig(-
+sn 9

T'he boundary conditions for -the current are

5IlCrC flO c at0 endsOf re

F-rconivenience, cosdr tewre to be(--)mtr ongE.Le

1(z) be arnrc-xincated bya seriHes nl 'triange ThaCtions

T T TV'-7(-

where

Nte satisfiess Ihe lboa- j crJ -- -L-,s.s Thea-t).xma

vnin4 -a T-
Tf arej jac naar, e

define by-, successive integers. Foiowin~ alerkn 2s metln (hy ~tlo

of rnzts withi w. = ,one btains teuu arxeuto

IL

-Where+

7 M(rib-'!3-.,- ;

V. = £ .Lv.3)L ir ) dz

i-i5
 "e e 1

jTa Isue kt-d were ev at edo IC-f)ijn±~ieJ r a

souionr r~creto t~ ieaeoj to (4).As AiV'Ea

22SecItion_- IP this so'l S 1n -~~n~fn n~rnl _

T.I was oe-IedaS- - cn' Vsn

curnara~tn t~n-i~te caksn as r%-A e et_

shmr i-1ie bct-o-s of cmhantnr- g1~nnr~n Was s~t~

- t - it' -m -an-p

rn..S-e .a:mtas -,a te t

the mmA n- -ntv ann c tnus .- ar ~-~ n the i - n- - r-

~
~~ta fl i' -- __ I---

~
was~~~~ ~~ ~ inth;nslonf~mmt
-LutlO-f Unnifg!TtflC_ er~eie
At irect attennt at rc-'-=----m-c n
-, -r- REU

- n- s! iub - -I - -mw =. -- -_tP6-- -

i-n--:z s-.r a
4-.t --

Wns
oit retei tElrs I =g e wir.
i..m.-~ s

icA4 s atcemdwy-r-he~jj

o,
nit i -an-yfly 'varyng ield n nf an r.-r% tinin

t-es-t it-

V%!- - L 1z - *mll
I- A * *-,tx* j-

Em- -Tas - e VA s are cure maanu that. o Set on ~-

N :- - - , --- it-e-d,

and sca'Lar 4- z aa. !'---ap r~r -: .- ' -n t -T% - aA 3-n

 L(z) - ( z
2a

inc-rme-ts alunE the wire have been seated to unity hence the finite

all!- eeneP ac-p-z~-oxananion the secndeiatve±

-:2 -'hd -- S f :I'eV-r

T, b. A-

i.
on---t -- meho *-n-,t.
xrq--lr- -Ive satsid artrer

Z logth ire- fce heo ............. I I'

-
9. -K

A.t.- giveoi-- t- esrpaed bvVe thati

ishea en

to *cod- -(4-91.
 It was found th~at. whlen the wire Segments were less th1n3./3 jn lng th

there was, no siiii .nAifference

In the solutilon using (411.f0M
znt using ( ! ). S i nc +tine -r r i--s so - w at S l .

(4-8)most actual conttoswre made usin(-1) oerte ch-omue

-la - i~---- .- i 1 1- Ir-t . l ~ d a

have btinn fl I-d l . !lll--

nf' ± nw-
* -- TicO izr--itn

- -l r-, - -

- im-- -hm SIM V- -f:

snows atez N - 'r

'aa Sea nt = -

-- - -- LAZ -
- A - -- - - Ins n e ZZ i z z ~Z Z

- itr-
- - - - n m ----

is ;ra gm-~-n
a rM_ m c ic - ~
 I
I

flit-,i

Tee
ieMRl16,afte SAtfa Pcil
,8r-
. Soo'rcs at centet -

Source . from cenler...

Source .". from center-...

/!'" '" ,
.2~ I t

Fig. 4-3a Input resistance for o linear antenna, L/2a zX42, as the feed point varies.

.3 r / / .'

Al /
I ii
'5"\
Fig. 4-3b Input reacstance for o linear antenina, L/2o 74.2, as the feed point varies.

Fig .51O
I LI)" 1.5 2.0
L/X
hit I 'Iil

4-bIpu ectne foralnaratt, 'L/ 742 sI edpin ais

=, , ;, /,70
70
 0I

It

+0 0
It

0 0

+00

81 8(y

710
 1,gures 4- o47 illustrate how the current distribut-ion on the

antenna. changes es th-Ie -feed point changes. Only the magnitude is plotted,

but the real and imaginary Parts were also calculated. In each case the

source strength was one volt. Fig-are I- is for L XFgure

F It-6 is

for L =.5,and Figure 1:-7 is for L 2X The case L x/- shwe

little change in -the shape of the current distribution as the source

was moved,, and hence- was not pbo0tted.

Fil gures -8tc --3- show,, how the gain pattern for the linear antenna
changes as the feed point changes. These curves were norimalized so that

they represent power gain over an isotropic radiator. Fi~Urae '-is for

L ,Figure h9is for L = l.5X, and Figure -OifoL=2X Te

case L = X,/2 was not plotted since -;-ne gain pattern is relively-r-,

insensitive to feed -Position. Note 'now t',he an te n na tends to behave a-:s

a traveling wave anzennaas thie feed is =1c-ed to one end, nhis is

particularly evident in the L =2%case.

D. Linear Wi-re Ecatterers. Conuutatiolls Jf _ lincar w-ire scatzerers

excited by plane waves at, various atngles of incidence have also llee mde

for L/2a ratios varying from 10 t-o 2000, a3nd L/X values uD to 2.1. The
following results for the case L/2a =71 -.2 G = 10) are -

illustrate tL-he- ge-neral behavior of -ire scatterers

Figure L-1LI shows e: ho area/1". vs. L/i and comrpares -the resul with

the solition of Y. Y. 2c Profeso -U~ -.

Lutoi- c~"a

typ~e solution using- Galerkin's metnc. with t he -wire divided in'to t-wo

W sections.
fctUions.
On each section a constant
As can be seen from: Fig.
nlus a sinusoid was used for expansion
-lte accura-Jy ofr her solution

-isgood over the range of L/,X consida'ered. This indicat.es thatg long wle

pro ab~ c-an De ac. fatelvreae ifdvie it s .ents of l1ength up

-72-
 44

-I

7-,73
 i -

- ~CCC

~JCT

10

744
 ( I
-J

111
I I:
1~ I(~J
-1~--
S

C I U) K VJi
( I; -
V
V

I a

( I I- I
-3
L.Jili~z~~i N
w

-J

V
£
0

0
V
£

I
4
(I
/4
I 0

K 'I *0V

It,
E

C
V

75
 'L IL

/ CL

r s z
fn0

T~4,

N. -~74
 IIwo

ii In

~NLIL
IF/

if-
4F0

II

jar
_4Iw

CL
 04
1.0-

0.4-

b -"

0.06-

O4

O.Oi .OOI °

.006,

.004-

Matrix solution
.002- y Y Hu'S solution ao -0

0 11 Ik
0 ooD 0-5 LS
LI)A
Fig 4-11 EcIto area of a wire scatterer, length to diamter roio 74,2.
to a wavelength, if constants plus sinusoids are used as~ expnansion fuanctions.

However, _uch a procedure gives rather~ difficult integrals to evaluate.

and the soluition of thiS Daper is easier t3 compute if' the Wire is not too

long.

Figures 4-12 to L-14 -- nw ho~w the current d.Istribution on the wire

changes as the angle of incidance of t.he Di ane wave excitation chaznges. For

these clurves the mPtgr-ztud e of the component- of' E parallel to the wre was

one vt't oer wavel.s,,tth Only 1:1-- ;.agnitude ol[ the current is shown, but!_

terea'l and imaginary components were also calculated. Figure L-22 is

for t'.e cse5 L X wit+ angles of' -'ncidencr varying otr

In st~ers of 15' 7igu-re L-13 is th or~~c.gset forL~.X and

Figure L-14~ is for L = 2%. The current on a wire L = X/2 was relatively-

insensii- fo &nzleof incidence, and is not show..

Fi;rures 3-15 to 4-7show the bistatic radar- cross sectio n vatterns

for the same cases as the current was shown ir Figures 4-1-2 to te-14 * The

angl e of incidence is shown in each case by an arro-'w. I oteta there

Is a 1.,arge lobe at an angle of scattLAer equal to the ang' of incidence,

oDut on the olher side of the direction normaa to the w-ire. Thscorrc-s-

vonds to snecular diffracticn from the~ wi re, anzd is mnmi'e pronounced as

-~e wi -e becomes longer. Figure !-15 -is for wires of l1engthn E.= mFgr

L11 o .X and Figure 4-17 for L = 2X. Again the case L

is not shownr "-ecause the shape :)f the cross section palL-erns is relativl~v

insensiti-ve to the angle of incidence.

E. Loaded Wire Scatterer-s. T-he impedance elements character'zing

a"-;near wire are basically the samme paramtr as th-meac lmns

used in tetheoryj o--; loaded scatterers.3 Hence, it is relatiVely easy

to cal-culate the behavior of- a wire with lumped i-vedances at 'It~s alonz
 Utr

alA48i of 4ncierice =90* b! AN*i of~E 7

Z-
f
orpo ieefIm(~! mvl e aamqh
 4 4/
2.

L.44 1.a 3L/14 LL1L

a) Angle of incidence : 90 b)Angle of incidence :75*

4~4-
LAL?311 L. L/2 IL/

c) Angle of incidence =600 d) Angle of incidence z45*

4 V
U-4 L .1
LA L/2 3L^

e) Angle of incidence =300 f) Angie of incidence= l5*

U Fig. 4- 13 Current (magnitude) on a linear antenna ~L =74.2 ItL =1L5 X

fr,r plane wave excitation~ (IEZi 1 one volt per wavelength)

82
 a 4-
L/4 L/2 L/ L L/4 L!2 3L/4

a) Angle of incidence z 900 b)Arigle of incidence =750

4- 4,

4.

2I

U/4 L/ /4
L14 Liz &A2

e)Angle of incidence z300 f) Angle of incidence z150

Fig. 4 -14 Current (magnitude) on a linear antenna -z74.2 1L z2 X

for plane wove excitation (I Ea = one volt per wavelength)
 ,if
/<

a) Angie of kcdnce gagl

9(- of kcwen750

<-1- /r /

0 Angie of kcdwc.fSoo d)Anale of ix~dnce*5'

*)r*of kxwg*s,3 f)A"~ of kh6c is-

Fig.4 15 Sistotic roo avs section pcrnsrm ;.-q a linea ant -742,L-X,cs the anl of rcidwu vaim
0* a) Angi df h~iwics 906W bAre* of -7mne'7

C)Aargh of Vsc,'6O'i d)Argl of k~~a45'

*) of
di~m~~3 fiAIIIS dd gsi5

Fo;4.1 0 1iti tor M~U "Chm Pcttm fOr 0 ft Ogm 7.u4.2,.L - Mk, O ft On*k f k a im
 0) Args of hkcbdwi '9(e 5 An"l of kwofc 75*

c'An* of icdsflco 6 d)AngW of e45

4an

*)AngI of uocdmnc fArg*

Mp od masic"5*

Fi-4i17 B*istc ~ross

emnSW;n pattmrv orM
[. Lh 7,2L20
fMr, Q- of ksCa v~rim
 ts length. Azain calculatic--s for /arto ffo 0t 00 .z

for L/X ratios vp to 2.1I, were made. The following results for tKhe case

= 7
112a 2 = 10) are representative of' the -eneral behavior of loaded

scattcrers.

Figurei 4-18 shows the echo area/ - for a center loaded dipole with

various loads Zj.chosen to correspond to those used byProfessor Hu

2
for her calculations. Again Professor Hu's results are goodl, but

apparently become worse as ZL-W. However, this discrepancy is no-t as

bad as it;. seems, since the echo area of an onen-circui-ied dipole is very

sensitive to the capacitance across the gap. For an infinitesimal gap,

the canacitance is infinite, just as it is -in- inear antenna theory.

In any approximate solu-tion, the g-p capacitance depends on the avnroxi-

mations made. In Professor Huts solution, no fuanction capa'rhle o

representing a singularity in charge distribution is used, hence erresults

give a low gap capacitance. In the method of subsections used for the

calculations, the gap capacit;ance increases with the niambe.- of- subsect'ions

chosen. This is because the function. exvansion (4-5) can come closer

to representing a singular-ity as -the niumber of subsec'tions increases.

-Actdually, a very small adjus;tMent of the gkp capacitance in Professor

Hut s results would bring 'then i-nto close agreement with the results of

this report, evuen for the caseZL

Fig-are 1--19 shows the ech-o area/Xe for a center lowaded divole

resonated by reactances at various values of La/k. 'The general theory

of-resonant scasleerers is availa e in

i the literaure. By defniio,
.

a resonant loaded scatterer is one for -which the load impuedance is a

reacta,-nce eaual Le neLgtJ o h in~u rectne oh

.. ,en fed as an anten-na. This definition gives a maximum echo are a whien

the open-circuit echo from the scatterer 'is much smaller than the short-
3.4
3.0-

((C)

2.0-

4-4

ILII

.02

0"2A. .8 dipole.
- 142
ir
fra center WodO
Fig. 4-19 Echo
 circuit echo. Such a condition is usually met for small scatterers.

but ot for ones of dimensions k!2 or greater. Figure 4-19 illustrates

this in the following way. Curve (a) is for a center-loaded dipole

cont'nually tuned to resonance. Curves (b) and (c) are for the divoiE!

tuned to resonance by an inductor at L/X - 0.25 and 0.55, r.espectil.

,

C-rves (e) and (f) are for the dipole tuned to resonance by a capacitcr

at L/k= 0.55 and 0.65. The fact that. curves (e) and (-) overshoot

';na (.a) ind-icates that resonance defined in tems of input imnedarnce

does not necessarily -Ive maxmu-. echo ar-ea. Curve (d) is for an oven-

Clicluitadiole. Tne c'-"n-re for a short-circlited dipole would le iin

between cur1-ves (c) and (e)l.

. W. P. IKLzg
Press, . Carie, "fhe -eor
Massachusetts, l-6.of Linea- ntenas
eaard Univers t y
f

Y
1% u, "Back-scattering Cross Sections o - a Cen
( J--in-A_-e_m-enna,- IFE Tras.
o , v¥ol. P-6, 'i-n--,a.-v a-
flti
t.- Lo d - e - -

R.K. Iia rringtSon, iTneory of0 ae-cterr, rc t

Zr. :a zrngton,
_. tS.3ra Resonant Scatterers and Their Use for
-L ?azuren-ia 12 T Trans. vo =-7 3 j

UF
 V. SCATIERNG BY CONNUCING CYLfl1DERO
OP ARBITRAURY SHAPE~

A. introduction. Thdis chav-ter iLs int.-ended to illust.rate a variel-y

of amthematical techniques wlhi ch nay be used for calclata.ing scattering

from conducting flyI nders off arbitrary cross section. Variolus appro'x -

nations have been t61ried and the results w-4 U. be illustl-rated.

-An int"egrO eqva t ion fona-l

1 ation of thne soluftion has been appflied to

a nu-mber of- examtples by K. K_ Mei and M. G. Andreasen7 Andreasen-'t s

examoles 4ivev bod-e s with curiving surfaces, corners, and even t-wo

symmretricall-y plac-ed bodies. Th0fin detail given for the cnr rent density

Indicates a ...are -zxlb of Oiaswere consiclerei -on tacnsatrr

An estimate of-: iblished
pu curves will be given for comprSon.

A diffnrnini- e~aton anroc has benfrm "le by Rflr

and hse been found to give fairly good results. However, it -was felt"

thatl the. integra' equation for-iulatian -was mo--st prom-ising SO emipflasIs

was Placed on i4t.

By the usual vector rotenial method the integral equation for W.

Dolari zation (i )i

r0 j

contouar

~~~1 ~ he
)a nrurac c, r- An dtr
f the conduc.tor ands(

-is the i-c-iaent field.

Wen h the f4 0 1AL dueta T -heAn tIh- bistati_ width 4

Le(0)=On2pH 52
 .d is comouted using the large arnnnment formula for the Ha1nke2. P-Nnct--ion'-.

Since p > > p' we approximate

ta -al - p1 os($0$4

where and $'are defined in Fi?-'g. r;-! T hen

for an- incident electic fedostrength .I. The scattering width may

aso be calculated wsing recaprcit AR J-n-nspred in Secion. ni-l--F for

as

wfiret of arbit4rary sn-are.

B. Pu Se Function Awmrxination to Current. Ttsini-est~l santion

to ( -i ISofl in which the inerlover th isrface is ~poae

as a sum o: Enejzrals ovr N small 4-ntervals. in each interval the

current is ass--med uniorm and the ~i~is matchA"ed at one noint wi-tin

eachiitr~
Figur5-1 llustrates -the divisionofasnetilboyno

intrIvals and defines the notation. The i -th segment is gvnb h

coo rdi nates o f a; wh i ch noin-t he integral ecuation is satisfi;-d

at

'Ph~
t ng aden oints of th+1, era are 'en-.eA w- -ente

of ofthe interval is defined as

egh i.ad..-ih

as- U.4.
he I 'Ieng-= t
+=

-9.-an-
 p

44-

Fig. 5-! Symmetrical Scatterer.

 L jJ')I (P)

N
PI
= A i (5-6)
r=!

within ,g
i
P=

0 outside /-.

J i0
=J(i) P.H(_j'(kQ- -Ci
a'- dks (7)

The resulting set of equations, assuming an incident plane wave of

strergth r, from x is

N
N ' i =g j= , 2, ... ,N (5-8)
i=l

where
ji,9,. H(2 "
o0)(k.6ds
-!j- ! s 59
..
1

and
-jkx
gj = 4e (5-1c)

The diagonal matrix elements &.. must be integrated analytically

ince the .ntegrand is singular. This integratin is illustrated in

ESection V-E.l. 'When the field and source points are different (j # i)

no :ingulerities are involved and to a crude approximation

-a4 -
 2.. H2)(k,, - i) & .-i 1

These are the approximations it is believed K. K. Mei made in his solution.

For symmetric bodies with normal incidence

i +- JQ"Q+_i Z = !AeiN+L
3 (51I

so that the matrix equation (5-8) may be reduced to

N/2 -kx
11. a. = he j, (.5-15)

where

ii11 0 1iN+I-i
)
=ii+ Ho( (kli-Nli k,6(114

and

2'. =2. +2
j1 ji j,N+l-i

2) - I) + (2) -
(k+3 (k Ia. -QN ~ 1 i~j
(5-15)

Note there are the same number of points on both halves of the scatterer

so that N is even.

Inversion of the matrix equations (5-8) or (5-13) gives the ai which

represent the current density within the i-th interval. This current

may also be found by computing the magnetic field outside the conductor.
Hence enc- - x X 4j() + x (5-16)

where n. is the normal to the conductor at the j-th point. This method

requires calculating two matrices, the column matrix [H] whose elements -are

-95-
 -. ikx.
Hjx x z e (j-7)

and the square matrix 11. whose elements are

-where z is the unit vector in t he z direction. The matrix eoa:ation for

isis then

.e nt4*s HL are evaluated in Section V-E. 2, using

Tile singular matrix el1em

can analytic 4itegration. Equation (51)is i:n itself a very crude

approximlatlOn. However, it is easifly evaluated using analytic difff erenti-a-

tion of the argument since no singularities are involved. The recta:u.a

component.s of the normral are easily approximated. as

n = -(Y.
-1.

The solution was first progranumed using (5-3) and (-l)for thc-

matrix elements. The results for the currentE. density and echo area are

shown in Figures 5-2a and 5-k.1ote that even wit-h such crude anrrox.-

mainthe echo widt.h as calculated using the cc 's is itl good agreement41

w~t~Axdreasen's. The currentl. Q.sitYr as ca]'c~lated by the a. 's is

cons ideraolyj in error. However, as calculated by Qj x ;qj Jis in good

agreement, being in error only in th-e region of unequal in-tervals and

rapid curvature of the ellipse. No more current calculations were naee

using this method. The reason for the imore accurate current evaluation

is probably related to the f'act 4+-hat this Drocedure is somewhat of an

____ ____
____ ____
_- - --------
 0
zw

0 x

o x~
0 x

0K x

48-

IN
00

00

Inn

ft ft p
0
10
0

CP

C.

C~C,
_ite tLioa vroce!ss wher-e only ai approximation to L has been made.

For a better evalutation of the i 4(i/) of (=Z_011, the addition

theorem for Hankel functions-

EJ -kt H%)( cos m .5-21)

n=O

2 M 0O

i s used to expand the kernel of- tChe integr'a1 equation around each p. in

terms of the variaible of integ3.-tiob, t' along tChe scatterer contour.

Figure -5deffines the notation whe-re ~'is now the angle between t1 and

= - Each interva2'% is t.aken. to be flat so tha~t

Cos and -(5-22"

The result off the integration for t,-he £.. is given in Section VT-E. 5. An

upper limit of four seemied.sfi i for convergence of the integrated

Series (.5-49).

dZ:
It should be not-ed that the elements X1.. an~ are most, efficiently

thne same time since they involve the same values of

comvuted atdC

'(lj- p. I) The sy-itvof the scatterer may also be taken into

acunt using (5-12) and (11---).

When the scatterer sur-face curves rapidly-f, the matrix element (5-49)

may be altered by defi'iing anle -nd 10 as shown. n Fi&

Fz. The

diagonal eleinentr remain as i~i (5-L5) but for J 4 i, the matrix eleinenet s

b~ecome
 Pxe Ij
0x

3 Fig. 5-3 Definition of variables I' ad ifon scatterer.

 yiy

Xi y~
X. N N

xix

Fig. 5-4 I!!fistrotiofl defining angles ~'and ~'on

scatterer.

-101-
 Z() [a+ ,3T -- 21- " 2)-(klkl) Cos r'

4
M=O kit+ l L k&L- 1
m -1A L(clr r# +o
2 I'

M=O L

(5-2)

where
COS~~~T- ' -

and a similar eouaation for cos

Figure 5-5 shows thz solution for the a. 's for three different cases.

in all three, (5-k3) is used as the singular point element. The solution

given oy dots used (5- 9) for the off diagonal e-_7ment- as did the solution

given by circles. In zhe latter,imre points were added about s .- where

the ellipse curves rapidly. The solution given by crosses used (5-25)

for off diagonal elements. in all three solutions, a "J " occurs

where the spacing between points changes drastically. This problem remained

throu.hout the work and is proba2bly due to an inadequate integration

rrozed-re. Fur-rtner work will be done to e!-:.-minate this problem. Echo

area has not been shown since the results plotted. as in Fig. :-2b.

A further calculation was done where (5-.51 was used foi to

sinvAlar matrix elements and (5-Il) for the I.. when kitI > 1.5. For

the L.. where k I -1.5=, the better approximation (5-25) was used. The

results plot identically with Fig. f-4, illustrating that the current

a
ansitv oinz 's highly dependent on the field near that voint and

not veri sensitive to the fiel4 far awayg.

-102-
 2Q
za

two

SA

10
00

0C i

cit S
C.)

![ -0

]f0

-- - ,- (O-
o" -103-
 C. Higher Approximations to Curre__t Dentsity A first possible iprove-

ment in the form of the assumed current is to constrain the a- 's in every.

other interval to be linearly related to the a 's in the two adjoining

intervals. If one beins with (5-13) for a syetrical body, the nmber

of u nkow n a i 's is reduced from N/2 to approximately half as many and

the integra equation is then satisfied at only every other point. From

Fig. 9-6 we see

A+ t - + -Or + A
a= a. , (P, A
I(+s z +A.A + ' i+!

. - + &A
-L. 1(5-25)

Where the defiriions of the ai's are obvious. The colu matr x of

the C. Ts becomes

=1 - k +. 2

Ak
az3

L A - a -

The resultin atrix equations t-o be solved are

re ccnjand eaatnar.-x1co-m sove atrixe [I is Iare

H j

sauare matrix w;here M =( )/, whicn is the case cons!dered belo"-

soulae -Were'
=-atr.x
aik
The elJ nts of the reduced matrix may be calculated as
!.

~
m+ 2 x 2m-l,2n-1 + I&C
2n-2 en

ImiI 2m-l,l L

UN4 2m-l,(N,'2)+

n M m~l . ,M

z are
and those of rE]

-~kxml
E . ne M~l .,M (-

-,f a plane wave from x=

The result of using (5-143) and (52)for the i.. matrix elements

and (5-28) and (5-29) f'or the reduced matrices is shown in Fig. 5-7.

The field has now only been matched at 19 points but the resuli are

quite good
It is also possible to expand the current density in a Taylor series

aon't each in terms of the body surface coordinate t. Keeping only

the linear t erm an t

3 t= G.+
t 1+~
1 - _1_t (5-30)

where the derivatt.v e term is approximated as shown. Integra",ion over

o,.e interval then gives a cont11ribution to the ~.-1, i, and i + ± elements

........-
 z
C,)

40
0

I CL
4-

04

~co

-- j 0

00

-107-
of a row of the matrix. A matrix element is not computed directly

but is obta.ied as a result of integrating over three intervals on the

scatterer surface and summing the proper contributions. The contributions

to the matrix elements of the j-th row es a result of integrating over

the i-th interval (col-un) is

. (2 ).1 a - aJ-i r 2)
if 4 I) dkt
H()(kIQQ_ + H ( 5-1)

where 2. has been defined in (5-25). The first integral has been computed

previously for both j = i and j i and the results given by (5-43) and

(5-23). The seccnd integral is evaluated in Section V-E.4.

A piecewise linear appr3ximation to the current density may also

be made. In this case, trianrgles are chosen as expansion functions in

the method of moments and the weighting functions are again delta functions

Thus N
= Z
i=l
.T_. (5-32)

wnhere
.her i!+ -Ak. < t < 0

1 +
k.
i- +---+

AR.

4 val then con+ributes to the matrix elements

int ezor a t ioan over -Lhe -- t,h intI~er

of the j-th row as

0
( o2- ) k - '(2) (ki dt:
- ,')°xkt i +" -
cvf.) - A-k -J o ".j_" ,
Cdk f

+ k--rl-
J 4- t H - ,II dt
-' 0
The integrals here have all been computed when doing a Taylor series

tproximatiov to uie current uensiLy. in Uns case, howv he :atrix

nt,

elements have been constructed from somewhat different contributions.

For symmetric scatterers use is again made of (5-15).

The dots in Fig. f-8 give the result for the Taylor series apDroxi-

m.at-Loi und the circles are the result for the piecewise linear approxi4mation.

D. Solution Using Galerkins ethod. A sclution for the 24 polariza-

tiun was also attempted using the expansion and weighting functions to De

the samie in the method of moments.

Cnoosing * pulse approxLmation to the current

f1 within Zt.
I L

w_ = (5)
0 outside

and

YI Z... = j (t") Ei() dkt" IN

where

z.. = f (±"'
ti.
rP(4. - t(-7

iJ

with 1-i(t') given by (5-7). "Here primed variales d source"

"et

coordinates and double orimed-- variables denote field coordinates. For

integration purposes the HanKel fnction with j i is expanded in both

l:source" and "fii_.
d" variables in Section V-E.5. Tne interval is

considered "fiat" as in,Fig. I-.n th..reuliting exp a.sion, integration

of the small argument Bessel function yields terms identical to those

-10g-
 z 2r

In
c
1 0

00
I
00

3C

~ CL

-0

is _

Cu CD

i 1

M -
wi-thin -tne br--ackez-, of t5i.) Thle "'-m)ete result i s o'-," b~ecaus-

of its length. The sinalilar Doinint ~enration- is sta .h-frard

and

is done In Section V-B. 6.

Thne (-erms of the column matrix on, the righ-t hand side of' C; .;C

mu-:t he areftully' evWaluated. Tus the in-terval is divided i.nto two

s -raight li1ne sepmment s as in i~ 5-;-. The resul of: the iteg7ration 'is

x.
,1y

= SGi ~ '~&S (.

X.1

d~/d37
nd d~r - eadda osan vrm 14

2 RX

Thee N3/-lcationsy of--are rgin'~smetods--n! oayerl 6b-erecdtoJ

adatie whenira~e as~ yne~ia.Z~no'~- sA*--

It- e is asoeti n s to er s the Ha7e

- auco
s be e-s
dfmv the

ecuatnd y~ecorites Siate

oI' body
i T etrJ- 5-7- 4mri2- imi
 44

I1 1 0

C~j d
 +2 2 V, g((221

V)n V 1()+ (g.H(' 7),2

o0

where
(P. x - x.,zr-
%7)=
OX y-y
-dy

and the nal-± unXns a-re evtua te d ai

et, x yx yV V

y= . hs is eatya Taylor- ser-ies i;,n four varia--ble.-s includ-i"-

second Aeriva.Ptive t-. The ineg volve-powers of x~-.

and (y -r v a)nd are performed in a fashion. sitLar to that. used- to

evaluate(53)

Mne result of programming this method pDlots very close to tne solutio

9_CIIin Fig9-9 'Thera are sev,,eral short'comins t o tbis methodA. The

exv.ancra ecuattons contain rmcany tCerms, althoaugh mxsit are easi ly progranren.

However, a piecewise Linear appr-oximation. to the -current density requires

t.erms m,oth variabcLs x and y. Insvect1ion of- Galerkin's method indictes

thnat i-ereasultin nu-ber term in the it-w*aab±L'n *oldb rx

f onr-aa1oe -

. Evaluation o nte--als.

EvahtiaWtI2
of'b [I J'w ppp3YimatJ oon n ent. -i

I vrtav7-" a smeLii iflter inV' s ~a the - -al

approximal J011
!21- ~ -( .kt %,~

may be used No isth

+n variable oineainalong the i-th! seg9ment

01f -ttlC sccatterer with or-igin at thme roin dfined by C ;ece(59

becomes

I.. H~~-kt')
k d'kt' + r H'2 (kt') dkt (5E2
0

IntCegrating

kZ=a( kAZ. 2 -kNi'. LCA

1Li k Z. 2k
kA t

-( ~Liog~yr)Ij.~Z
7 '

kAI 2-kA

where

2. Elvaluation of Va 'efr Pulse Approimation to Cu-rr-ent'-. Tb evaluate

a_ s'13 argin-enti a-pprxiion to th Hankel fucinmay agair. be

used. Hoewever since the singula-rity Js xnonintegpable, t-be inerlis

frtexuressed wit'h the fI'eld -jint ofZ!th surface ofT the conductor

a small di;stanr n Thus

ci'.f

-k-
-n1.

!.sing

and
 U(22
tAl+1

The intearal then becomes

-' 2j 1 -z -. z
n± T. X~ [t an - -n

Lett.ingv n-0Q

cu -as -~ulli (5-1 6)

E a-. -, a' s-,: Ap-oimaai n u r n . S m

fl* tht el fLrmn on-- =a be ma -heb t4F&

-k~ hl3tAo~ c 't

71~ ln jL 1, l r ____ -t

Ct.,t~itcJ(~~I 3
m-
On integration

kn 2.l[ + - 1m 1) H( ka 1 o2
ktZ6
2) [r-C (nil)11 (E73Y

4. Evaluation of Linear- Tern- ±n Expansion for Current. The integration

-is done ex-actly as in. -Sections V-E. I and VT-E. 5. For -the singuar- tenus

R f +

2 kAL2 ~k~.2 ~kL. .2 1kt2

1,21 2 7a i- k . 2 2 k-'z2
zlogi- l~ __
kur 2-

For, integration over zrnsingu.iar intervals

L. m--2 r .1 2

kZ(:Jn2' [5c3f- i 2 '

rn' -- k - J1 csnt
 o--f aas~unkeal
-. 1n.--i~aS-4-1- t -- r- - ad I i -. l a r 2 a

in m-s its o-

along witn tacM-

spe-i-I ense uaZec-rv..USN

In' - - r n rd--

* t 5 z re ~ r-I an S;r l

r o- =izac -a -n bewee Ao flld

ds~ ~~ 1 0 j4.F=; imze t= n an :p nSo- a-.. im c

fi
I&!Y~' zo-fre,-

s, r.V-b u-i ai flom ~ s~~if-. ay fteI- cn-r CD.,&~tn !r sr -

V- v - -Z I t , - -- - - -

mt,I to D neaainc~rIbt -arn oread p

c ate---iS
C0 r m; r-, fl.ti vr s- a----ezHne
function is now written as

(2 -'2 2 2 (t-t 2
H((1- lVg 2 ( 2

for t > t' and with t and t' interchanged for t < t'. The integration

to be carried out is then

+t I

Iii 1o H (t-t dkt' H( )[k(t-t) dkt dkt

kt
-k. .1. 0.
11
(5-56)

After the first integration'is carrii-d out the second may be reduced

considerably. The final result is

i (kX)2 (1 - 1( i

+j 1 (kA)2 (3-2 o (7 - )(I - 1 ( )2) 19 (ki)2)

(5-57)

F. References

1. K. K. Mei and J. Van Bladel, "Scattering from Perfectly Conduct-

ing Rectangular Cylinders," IEEE Trans., vol. AP-2.1, March, 1963,
pp. 185-192.

2. M. G. Andreasen, "Scattering from Parallel Metallic Cylinders

with Arbitrary Cross Sections," IEEE Trans., vol. AP-12, Novem'er, 1964,
pp. 746-754.
3. R. F. Harrington, "On the Calculation of Scattering by Conducting
Cylinders," IE Trans., vol. AP-13, September, 1965, pp. 812-813.

4. R. F. Harrington, Time-Harmonic Electromagnetic Fields, McGraw-

Hill Book Co., New York, 1961, p. 232.

5. J. A. Stratton, Electromagnetic Theory, McGraw-Hill Book Co.,

New York, 1941, 49- 374.

-L18-
 rI. SCATTERING BY DIET ECTRIC CYLII ERS

A. Introduction. In this section, the problem of scattering by di-

electric cylinders is discussed and a particular solution is formulated

in terms of electric and magnetic surface currents -n the obstacle. The

fields inside and outside are repre sented in terms of the field radiated

by electric and magnetic surface currents distributed over the surface

of the obstacle.

The two dimensional curve representing the surface of the obstacle

is approximated with straight line segments and currents are assmed to

be constant over each straight line segment or to be delta functions

located at the center of the strip. The fields due te infinite strips

of electric or vna-netic current, with Z or K directed either in the

z direction or along the straight line segment are calculated. A dual

formulation is used to extend the results to the TE polarization.

B. Formulation of the Problem (24 to z). Consider the problem

indicated in Fig. 6-1. A plane wave with electric field in the S direction

(OMn to z) is incident upon an infinite dielectric cylinder of arbitrary

cross section, pert ittivity e, and permeability [L.

We cwn represent the fields in terms of incident and scattered fields

E inc
Ez =ES +
z z

H-in (6-1)
H = S + Hinc
P P P

For purposes of calculating the fields external to the cylinder we

can use the eauivalence principle, replacing the induced polarization

-119-
 y

Fig. 6-1 Cykid Scatterer and Coordiiote System

-120-
 and meinetiztio'~soures. /:Za and
urcsV x M wi th surfae
fac curents
u---T = x

and x j.. d istribated over the surface of the obst-acle. These sources

E-re such as to -aroduc-, zero fields iIpde (if '%,he currents are exact).

Since the.s e equivalent sources- nroduce zero fie.ds inside. we can replace

t~ ~r~
clint '~t'ia an th exernal field will notl be affected.

If we choose the material1 as free space, thean Lile Problem becomes a nrl-

Reneous one of calcutlating the fielids due to --~~iite

t-%ni hommogeneous

fre Space.

,Sinn
1,, fr prposes of-_-' cuat.I ng the intev-rial f-!els we can
replace the effacits of- te inci;dent 4 --a~ wit surface cur-ents
,

+Inx H and - =~+Z x distriout-ed over tChe surface of the cy-linder.

These sou-rcer-s are such as toproduce zero fields outside the cylinder

-ifr- cirrentws are exact), so that we can rep2.lc th

Ue free space ouitside

th y~inder with any material aand the internal fields will1 notw be affected.

Tfwe choose tUhe material to be of -pe-r-Iitvify e -and permeabil-ity t, then

the probl em becomes the 1hom-geneous one of c-alculating the fields due

to -~ ~in infinite homogeneo:_ svace of characteristics jir:

Vee-tor-'- -tential functi ois can be cal1culated for tbese two homogeneous

pr-Yblems. Denoting potentials outside by subscripts 1.t-hen

contour
r o
y~~j~t) H-'(kop) 4 *(65

contour

d- differential element of, length arvound contour ofI

cylinder

k J'E i
0
The fields outside are given by

60
- (6-4a)
~= + V xA,L j dV XV X

The 'vectorpotentials for the fields inside are similar except the material

wave i.mber and material constants appear. This

=V H(2 .)(kdi
P)

S._.- -/ () (kp) d"

(6-5)
V .Z,+ -,

where k = b.v

Since there are no surface currents in the actual problem, the

tangential B and p_" fields must 1.e continuous at the surface of the

cylinder. On the surface of the _ylinder

Equations (6-6) may be expressed using (6-4) and (6-5) in terms of twa

integro-differential operator equations-

r IIinc
V Ai + -,VX X

(6-7)

-122-
Vhich are of the fo=m!

4 .Mf) +22'-2' =2 g 6-

where L l-IP . P areP integro-differentiJal operators, P ff are

-the unk-.mwns 4,~and g, a:) are the knowrn f tncti-,o's ii x _ana

Ecations (6-? may be rewrit'ten in the forta

U J5 4 V inc
L -; -2 Faj

Where V' -4- the fieldi inside du-. to the- source in infinite hom-geneous

media wih constanats jxand e anc jli is the f ield outside due to the 1

source in f-ree svace. Like defintions hold for the fields.

f

In order to convert the operator eauations to a set of matrix -

ecquat.ions the contour of the sceaerer is divided into N straight_ line

segment-s as in Fig. 6-2 The current. is then ass'imed u~ni form within

each, interval so

a.-P
C-1

-Where =i f ihn i-tb segment

I0 outside i-th segaent'.

Usiag (6-10) and satlisfying (6-) t-the center off each segment ;dives
two sets o-4 matrix equations. o Mplrzto

-123-
 y

XH YIAI

Dielectric
Scatterer

Magnetic currents are assumed Counterclockwise.

Electric currents are assumed 0% ~ecticm.
&

Xi 4(x! + x!.)

Cos~j xi

Fig.6-2 Dielectric scatterer and notation.

 EtT.E

ZIP

- r

Vi2l J- I - - .

other ieh bma-Zc matrix

n is elements ulIst fetain

,:r.i Dartie
the n to fidnte-atrix iWteeti
elements ieda I rh e

_~ ~ ~ uet
~~ -
urn X -Aa~n fmgiuel - Cure
nthe i-ie

.. ntrvsh tohje
e hat
edds~ ter abi Dt''L ofn tben-,I.ht e- - to the

unt' mthree submat an the ir satrixnelement. Abua eto2.utin

no~heTEixlriatonas~
r csed n I-E. Tee-reoba5ed-
 It is assumed that the magnetic currents are all directed counter

clockwise. Then, locating ourselves at the midpoint of a segment and

facing in the direction of the magnetic irrent arrow, the fields for

the subscript i case (free space) wili be just to the right and the

fieldt for the subscript 2 case wi be just to the left.

Consider the case i=j, that is the fields are to be evaluated at the

midvoint of the j-th segnenlt due to magnetic currents in the j-tb segment.

The orientation of the coordn ate system is ninnportant for this vrbii

and the configuration may be si-mplif- ed to that of Fig. 6-3-

The magnetic field is obtained from the vector Dotential % where

(Y2H(2)p) jj'
j 0
-d/2
and
x2 +y

Now in dielectric media with constants p and c

SVx VxZF(6-16)

Since in Fig. 6-3 the tangential magnetic field H., and the partial

matrix element are

a> identical, then

-i e2 42a)
2)
HI H(kp) tc' (6-17)
'2jj j,.7, j o
-d/2

This field may be evaluated in the following py. let the field point

lie on the y as, at a point (O,y), evaluate f H(2)kP" dx', differentiate

twi.ce -with respect to y, and then let y .vO. Hence

.. ........ 'J %a.

 y p

Fig. 6-3 Magnetic%. Current Source

 = un
~(,0) ~ ~fH2)(k)dt (6-18)

Taking the first two terms of each series in the series expansion of
2
H(2)(kp), using integration formulas Dwight 623 and Dwight 623.2,
0

differentiating and taking the limit as y--C, the partial matrix element

becomes

1(1I
[Er
4 4r
[3 -log 2j (6-19)

To calculate the tangential magnetic field H X(OfO) in the subscript

1 case, consider the same limiting process with the field point approaching

(0,0) along the negative y axis; by symmetry the tangential 11 field

will be identical to that obtained in the subscript 2 case, since the

magnetic current source is tne negative of the source in the subscript

2 case. Thus we obtain the field Hxl by replacing the unit magnetic

current with its negative and g, e, k with gOP Eo, k Hence

0 =-j (k d) 8~ -1 1. [3 lg kod (6-20)

ljj 2n 0 -(iR 1 % ~F

The tangential electric field on both sides of a strip of magnetic

current of Strength unity must also be obtained. From Fig. 6-3, if we

assume that the electric field to the left of the midpoint is in the

direction, then from symmetry considerations, the electric field just

to the right is in the negative z direction. The strength of the field

is most easily obtained by asing the dual of Ampere's law

-128-
 dZ = K (6-21)

where K is the magnetic current enclosed. The result for a unit magnetic

current is easily obtained as

= -=jZ= - 1g (6-22)
li 2jj 2

Now consider the evaluation of the fields when j r i. Replace the

magnetic surface current distribution with a magnetic dipole elemen' of

1 located at the midpoint of the segment.

strength d. Here d. is the

length of the i-th segment. For purposes of calculation, assume a magietic

current element located at the origin and oriented in the x direction as

in Fig. 6-3. Tben to a crude approximation

d" H12(kP2)

1
E-VxF = VxVxZ (6-23)

Thus
kd. (2)'

kd. '2)' (kp)

k1 cos H' (6-24)
p iT' kP o

kd. (2)"
IV = - -sin 0 H kP)

"low if we locate the magnetic current element at the center of the

i-th strip with an orientacion corresponding to that of the i-th strip

we can derive fields at the center of the j-th strip by replacing P

with IP-1 and 0 with .. - i" Then

-129-
 :=- - sin (0.. - 0.) i4 cp..i ) (6-25)

= sin - p..) + H, cos (0. - ij)

"k. Is - i-Oj
=kd [ in (O.- 0) sin (0i. "0) H(2)"(klpji)

cos (0 - Oj) cos (Oij -i ) (2)(

+ k 13j Hi (~p..)1
(ki (6-26)

The elements and j can be obtained by replacing k, e, g

in (6-26) with ko,

0 , L0 "

D. Evaluation of Partial Matrix Elements from Electric Currents. We

wish to calculate the fields at the j-th point due to a unit electric

current at the i-th segment.

First consider the case when j = i as in Fig. 6-3. Then

A= 1 fH(2) (kP) dx' (6-27)

so

Ez = -2 (kx ') (6-28)

0
Since Ez and E. are identical, we have on integrating dire;tly

2 =-kk2 ~d H(2)(kd) + ifd [H(2) (k) H0(

0 (d o
-(2)()H 2 l
(6-29)

here H1 and are Struve functions. Alternately, if H(2)(kx) is

approximated by its series expansion (retaining the first two terms of

each series), and integrated directly, the following result is obtained

-130-
 E
E2j j
kd
= - (-)(1 - )2
ikd)

,2(-d )j 2()2 + 4 (6-30)

The element Ej j is obta-ned by replacing K by K in (6.30).

The tangential magnetic field may be obtained by applying Ampere's law

in a manner similar to that used in Section VT-C to calculate the

tangential electric field. The result is

jj= j 1= (6-31)

Now consider the fields when j A i. Replace the electric surface

current T with an electric filament directed in the z direction with

current I = d, and located at the midpoint of segment i. Consider a

filament located at the origin. Then

di (2)
A. = H
H- (kp) (6-32)

Hence

kd. (2)
kdT Hi)(kP) (6-55)

Hp=

The partial matrix elements for the subscript 2 case are then

d H (2) k j i)
2ji = 0

(6-311)

= 11(2) (k I P J)31
di

-131-
 7.he elements i and Hj"i are obtained by replacing k and q by k°

and % in (6-34). The matrix elements in (6-i1) have thus essentially

been established.

E. Duality. For incident fields TE to z, the problem may be solved

by duality. In this case, all magnetic fields will be in the z direction

and the electric fields will lie in the xy plane. Thus the scattered

fields can be calculated by assuming magnetic zurrent sheets in the z

direction and electric currents sheets directed along the contour of

the scatterer in the xj plane. Again the problem is resolved into two

homogeneous problems, one involving free space and the other involving

an infinite homogeneous space of characteristics V,e. Replacing g with

E and j with M according to a duality principle4 yields problems

similar to those discussed in VI-A to VI-D, which can be solved using

the functions developed in VI-C and VI-D.

F. References.

1. R. F. Harrington, "Time Harmonic Electromagnetic Fields," McGraw-

Hill Book Co., New York, 1961, p. 106.

2. H. B. Dwight, "Tables of Integrals," Macmillan Company, New York.

3. N. W. McLachlan, "Bessel Functions for Engineers," Clarendon Press.

4. R. F. Harrington, op. cit., p. 99.

-152-
 VII. SCATTERING FROM BODIES OF REVOWUTION

A. IItroduction. The objective of this work is to determine matrix

methods of solving for the surface currents induced on conducting bodies

of revolution by incident electromagnetic waves. These currents can then

be usod to calculate the scattered fields.

The general problem has been solved numerically by Andreasen for

bodies with maximum cross-section perimeters ar twenty wavelengths. His

met> )d consists nf obtaining approximate solutions of the integral

equations for the induced currents.

In this section, the general solution for the body currents will

be fonmalated by the method of momeats. From this, two specific solutions,

namely, a field matching solution and a solution by Galerkin's method,

will be described.

B. Description of Problem. Consider an electrotagnetic wave incident

on the surface of an arbitrary conducting body of revlIution. The

ccordiraes for the body and the inciden wave are illustrated in Fig.

7-1 and 7-2 respectively. The incident field (') induces warface
field (Zs
currents on the body whichare the sources of the scattered

The boundary conditions require that the total tangential Z be zero on

the body mrface or

= R
R( - (7-1)
Itang "ttang

For converience, (tl' '2 ) are defined as unit vectors of the local

orthogonal coordinate system defined on the body surface. t corresp-,nds

to the t vector and t2 is 8 in Fig. 7-1. in terms of these coordnates,

(7-1) becomes

-133-
 Field

LOW

3 Fig. 7-1 Coordinlates for Body of Rev"oln.

 I

x
Fig. 7-2 Coordinaes of hkcident Pia Wove.

-135-
 The scattered fields are expressed in terms of the surface currents

as

Ahere the operator L is defined below. Equation (7-2), becomes

t,n~ SI)r = (7-44)

The pxroblem is now reduced to solving for

fi~

Tdefine the operator L, (-)is expressed in te-x-as of theagei

vectir a-ad scalar DotCentials. The vector potential1 for the induced surface

cu re n t ( , is

R opt

body

where in terms of the spherical coordiin-tes of Fig. -1

R = z zI=[r2 + ri 2 2r(Cos 0 Cos r.2 +

sin 0 sin 9' co31

Mhe scalar potential is given by

s - '. . Z da ( 7-6 )
body

whiere Vr.4 is the differential operacor tangent to the body surface.= In

Uerrns of the local body coordinates, t14his operation on ar arbiturary vecto r

.is
- ~ t sin v
Vt++
where p is the cylindrical coordinate radius and v is the counte -clockwise

angle between t and the z axis.

Now the scattered field is expressed in terms of the induced current

as

.s,.

where V is the gri ent operator on the body surface

he definition of Ls i li-ed by (7-7)

C. Rxpansion of the lcident Field and S-urface urrt. An arbitr&l

plane wave can be exoanded in a discrete set of cylindrical modes. Eah

mode has a different azimuthal wave -rnr'er and each propagates with the

sa e velocity along the 2 -axis of Fig- 7-2. The advantage of this

expansion is that the current induced by each mode can be comit ed

separately and the total current cnp 4

then be found by su-e ri:'osing a1

of the significant rode current-s. Uis proce is necessary for

an -in l rge bodies on available com-puters.

The details of the expansion are given by Am.drees en. 2 The prpagatin
vector E of the incident wave is cbose to be in the x-z plane of Fig-. 7-2.

Ten the incident wave is separated into a wave volir-zed nrallel -to

the x-Z olane an. a wave polarized ternendicular

to thjs pDane.

The exnansion for the tangential electric field comnonent of the

parallel polarized wave the warfa;e of te boo- is Ula
t= > Et!t) 00s (4 ) + E,4t) sin (=--) (7)
M=0----
 -a -- - --- _

where

jkz cos P.~

- co's 09- sin v j'Qr'. sin

-
ojj
d]

jkz co's 9 ..mtl si -0"

~t) Cos0 CosG z- Jek

X SinG9.e
se

qj Js the im-uedaz ce of free s=ace. e is Reu- nn-s nunmber. Ja tshCe

Bessel PAncitioa off" Qrdnr m 9A thea mu

incares the differen:tiaraon

with' respect to the- argament.

For- the perpendicular vola-rized -the t-angential field i

2

a ~ S - 0XE
2 (t Cos 4O

Eu~)= - i sin ve eJM kin&

jkz -cos 9..

CDMbnin (7-) d (9 Taird- fieaid is

- z(- a
Cosco

Im

sin 4d - E 0 ')cc.4-7-0
-T%,e total current inue tu the
-, ld sn -- tne IO'n

Its - t JItJ- s r i

L I ,c a=

are,

Cu
tena ofe th e:;-. e cae n

jie to- ser-a y -o-e t- za

=~U rttsn -1 J Z,(t) (71 1s

ecr the ofen-ea. culrolariation h t-so o~.o. lscse

aidosltionisb
reSxe3aw el ofoIMmeter Mae~C oenr -±. i-mOt~isr then
A Z_. Or-3 7 l2 <A .. I fZ

ajtre o-,1 of h oeb

0 -- __ied b- S nowaein sice 5 ir of

eac node of eofe -a -1r~

aStnc .. -f-
s. -.
t'zej - %.AI aur& o.-i

IZo
tecurftnt a-n dfl'm-%.onsgve thep rill -m - ol - -o

and~~l -th souto - . Ga~l -L 's nntfm- are.a'.a =---s

 It-e coordinates of the body currents are cal.led the "source coordinates"

and are the primed coordinates in Fig. 7-1. The coordinates of the incident

and scattered fields are denoted by the unprine "field coordinates."

From (7-14), for a given mode, say the m-th mode, the curre..1, induced by

either incident field nolarization is of the form

s (r_')= 'Jt(t') h (mo') + 0, Z,(tt) h,(mo') (-

A procedure analogous to the development of (7-14) is used to express

the incident field corresponding to (7-15) as

ZI__ - t E'L(t) h (mo) + 0 E,(t) h (mo) (7-16)

where h,(mp) and h0 (4 are the respective trigonometric functions in

(7-12) or (7-13) for whichever mode and polarization is bein- considered.

The scattered fields due to

W
S.,
L sj (X')
=~t =-- b pda' [k
L e ~T ejkR
Jt J
bo~dyI
(7-17)

The t and 0 scattered field components are found by taking the inner product

of (7-1-7) with the t and 0 unit vectors at the field points. The result

{
for either polarization is

a S 1 + Gn+l)
in v sin v' (G_
ta

2cos v cos v, cj ,(t,) - sin v (Gm%-l - Gm+,)

f,I,(t' ' dr' + 2 VW +i VIJ

fJt)P 2ft Rj,t) Pt
a
a

+ (t, ' dt'

PI- I (7-18)
ii

-p wJ
 L.j ,(f ho -sin v, (Gm1 - Gm) 4(t')

-+ ' ~ (t,) + t) m dt'f

ta
a
(7-19)

where
Gm j r" e R-0 cos mo, do' (7-20)
0
0

% =R(0 = 0)

To apply the method of moments to either polarization of the m-th

mode, the mode current terms are approximated as

N
t(t) a. Ti(t')
i-1

N
X i t t (7-21)
('

# i=l

where Ti(t' ) and i(D(t')are the approximating functions in the method of

moments. A suitable set of weighting functions (l' w2: ...ywN ) and

an inner product are defined. The approximate currents (7-21) are

substituted in (7-18) and (7-19) and the inner products of (7-4) with

the weighting functions are formed as

-141-

U
 =: -~(7-22)

where the current and field terms are d-efined in (7-15) and (7-16)o Both

of the matrices of (7-22) are complex. The left sides of the respective

real and imaginary matrices of (7-22) can be rearranged as

ij ] "
I W3,1> i 1,2,...,N

eN 1: 1 ,

] k (7-23)

The solutions for the complex currents, a i and Pi are found by inverting

the [1] matrices. Thus

L
[] lII (7-24)
L] Le0 ]
where the definitions of the matrices in (7-24) are implied from (7-23).

E. Field Matching Solution. The field matching solution amounts

to satisfying (7-1) at N points along the body perimeter from t to

(see Fig. 7-1) for each polarization of each mode. The perimeter length

from t to t is divided into N segments. The length of the k-th segment

a b
T isAtk and the center point is t k as illustrated in Fig. 7-1. Since in

(7-22) the 0 variations in the scattered and incident fields are the same,

the h(mo) terms can be cancelled by choosing the weighting functions to be

-142-
 Wk= t-h k = 1, 2, ...

Then the inner product is defined as

tb

<A ABdt 7-25)

t
a

where A and B are matrices and A is the transpose of A. Now (7-22) is

seen to be equivalent to (7-1) at the N points along the body perimeter.

The current approximating functions in (7-21) can be chosen to be

Tk ( t k(t l)
I ; t, 1 on 6tk

0 elsewhe -e

This amounts to assuming that the currents are of constant amplitude on

each interval. Other current expansions, such as a piecewise linear

approximation, are possible but the computations are more cumbersome.

Substitution of the approximate currents into (7-18) and (7-19)

give approximations to the scattered fields at the sampling points.

These are substituted in (7-22) and the approximate current amplitudes

are computed from (7-24).

The accuracy of the results improves as the number of sampling

points is increased. The scattered field at each field point on the body

is calculated as the sum of the contributions to the currents on each

o1 the source intervals. For each contribution, (7-18) and (7-19) must

be evaluated. These calculations will usually be approximate. When

the field point (t i) is several segments away from the source sampling

point (ti), sufficient accuracy may be obtained if the integrands of

(718) and (7-19) are assumed constant over the sampled source intervals

-143-
 Pnra
u t- -their resaective values at the center of -each source interval.
nUa

In general form, the integrals are then approximated as

When the field and source segments are within a few segments of each

other, it may be necessary to subdivide each source interval into sub-

interva.s. The contributions to the scattered field from the currents

on each subinterval can be determined by evaluating the integrands at

the center of each subinterval and evaluating the respective integrals

as M
r 1)
ktj kj
tj=l

where M is the number of subintervals of the k-tb source interval. When

t. = t' the G terms (7-20) have singularities since R 0. The

contribution from_ the currents outside of the small region

S6 <~ 9P +6

can be determined as in the above case. The contributions from the

currents within the remaining small surface aroiu~d t1he field point can

be determined by an approximate analytic procedure. This is discussed

in section VII-G.

The current derivative tems in (7-18) and (7-19) can be approximated

by finite difference approximations. For example, the derivative of the

t-component of current at the k-th sampling point is

-144-
 j(t') l- %-
attk+ 1 + t t;- !

2 + At --

Higher order approximations could be used if necessary.

The evaluations of the G. terms defined in (7-20) and 6G /t are

kk
discussed in Section VII-G.

F. Solution by Galerkin's Method. The proposed solution by Galerkin's

method is equivalent to requiring

w h.ere ),- (Z)> - i(r), i s(&) (7-26)

w () is the approximation to the current on the k-th body segment.

To obtain this formulation from the method of moments, the weighting

functions wk are equal to the current approximating f _ons. For example

let

= T(t') = (k(t') = 1 on

0 elsewhere.

The inner product is defined by (7-25). Equations (7-22) are now the

equations of Galerkin' s metnod.

The constraint imposed on the solution by (7-26) is more stringent

than that of the field matching solution. Tnerefore it may be ossib' e

to obtain a higher degree of accuracy in the variational solution with

the same number of body intervals as in the field matching method. However,

the calculations of the matrix terms in (7-23) are more cumbersome.

Many numerical techniques which are analogous to those described in the

previous section can be used to simplify the computations.

-14.5-
Polution, one of the major computations is the evaluation of G and 6)Gjt

in (7-18) and (7-19). G is defined in (7-20) as

ir-jkR
G - eW- Cos O, (7-20)
0

where_____ 2 02 __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

R = Ir + r, - 2rr' (cos @ cos @' + sin 0 sin Q' cos

(7-27)

Using the derivative chain rule

c)G- 6GN
-- _ m r + M--3 (7-28)

Applying (7-20), (7-28) is expressed as

resu frm+ f2 Cos-o).

0 0
0

wher th e th aprorit f cin f r ' , ' hc

+th ,r e nteg i + t cos

fr dcos in e'(7-2 )
0 0

where the f.'s are the appropriate functions of (r, r', 9, 0) which

result from (7-28).-

'For the integration techniques discussed in Section V[II-E, (7-20)

and (7-29) are integrals over the source loop at the middle of the source

inter-LU (see Fig. 7-1).

.e i2tegrals can now be expressed in a general form as

7 -.k m = 0,1, 2,...

e 0
-o=
n 10' Cos ,' do' p = 0, 1
n0 0O
n = 1,2,53
(7-30)

when the integrands of (7-30) have si.alarities at R = 0, approximate

0

-146-
analytic integration over the singlarities is Iiecessary. The details

of this method are given in the references.P

For the integrals which do not include sinf-alarities, the complexity

of the integrand prohibits exact analytic aaluation. Mmerical integration

is possible but the computation is cumbersome. Another approach is to

ex-pand e-JR'n in functions of the integration variable '. Substitution

of these expansions then allows approximate evaluation of (7-30) by

simple analytic integrations. This is the method which is proposed for

the body of revolution problem.

A
To illustrate the expansion. .onsider

e- jkR
cos kR
- = Rn J. sin nkR (-

Denoting either cos kR or sin kR as T(kR), each term in (7-31) can be

ex-panded as
N
T(kR) _ [A+ Be-O' + Ce " - '
.,n L- a. i(O') -
0 <' -
<-
(7-32)

for n = 1, 2, 3 and R > e which is a sma!! positive constant to be

determined below. The bracketed terms of (7-32) are chosen so that

[A- Be4 + Ce 0O < < (7-3)

Fiure 7-3 is a typical curve of 1/Rn versus $' for closely spaced loops.

A, B and a are chosen so that

A +Be-at0
1n -

(7-34)
>i at.0 =*f,0;
U
 A+Be-:¢

-(A+B)

Rn

R-(A+Be-C

0° I 9',0 ° I

I II I I

Fig. 7-3 Graphical IIkistr6iton of the Approximation of -I

-14RN
then

! -

a I 5)n
i R( R(OY' n
3j

3. 1

B=-C
_ 1!

A Be (7-35)

The constants C and 9 are chosen so that

1 (A+ Be-4' ce- W

-.'

and

Ce-'---- for0' >

If ""is -Droeriy chosen, (7-33) is satisfied when

k')
- -. n(A+ Be

LR(;)n

L n

To compl ete the ex-panzion, (7-32) is rearranged as

-1 9
 N
T(kR) '736
i=O 11 =A +,eW + CeI -56

-T(kR)

Zince the T(kR) are sinusoidal functions which are even about ' =,

it is convenient to choose the -. (' ) as trigonometric functions With

the same symmetry. For example

CL.nstant, cos i ', sin (i -)

Equation (-36) is a minimum square error a.proximation and the coefficients

a. are detsrmined by satisfying (7-36) for N val!ues of $'.

Figure 7-L gives the results of representative expansions of cos kR/R

and ces kR/l 5 as

' Ce '] [a+c +~o2 !
fA($)=[A+Be
+e4t+ +efa- cos 14' + - cos '+

a3 sin 2 a14 si

(7-37)

Denoting the expanded anction as f(T1 '), the error at any 0! is define d

as
A' T

TP ),ax

From the error curves in Fig. 7-4, the average mag:itude of e for
fror

either extansion is less than 0.3 percent. For these cases, Rn -

0.061k. it was found that for R > 0.067, the average error decreased,

so the cases presented give-anL estimate of the maxim errors that can

be expected for expansions of bodies of re;outtion with diameters less

-150-
 R -0 co67xs ks -
r = 0.235K r' = 0.206K
-
-_.o A 2. ',-

B 1k r 11-22_
Eq. (7-32) satisfied at -_ -z -_

4,,,O . -0 0 0 a

cos kR cos kR
Iiz

$4= 90
-0-
°°0 _
0!=2-- S I i.-'6

oi;= 60P0
oo t_ --J-5
. r

-I .0.

Error a,
(Percent) R

Fig. 7-4 Rwuft ot Couted Ew

than 0.5 and for Rm > g = O.067X. For larger diameters, it may be

necessary, to increase the number of t-rms in (7-37).

H. References.

1. M. G. Andreasen, "Scattering from Bodies of Revolution," IEi

Trans. on Antennas and. -Propagation, vol. AP-13, no. 2, March, 19W,-
pp. 305-310.

2. Mo G. Andreasen, "Scattering from Rotationally Symmetric Metallic

Bodi 1os," Final Report, Subcontract 294, AF i9(628)-OO, TRG West, Menlo
Park, California, April, 1964.

-152-
 VIII. NVM-.RSION OF MATRICES

A. Introduction. The inversion of a matrix is closely related to

the solution of a system of linear algebraic equations. The n lineai

equations with n unknowns

all,, +ax 2 +.. . +abx yl

a2 1 x1 + a 2 2 x 2 + "+a 2 nxn =y (8-1)

ax + a2x + . + ax =
nll1 n2x2 nn n

may be written as a matrix equation

Ax = y (8-2)

where A is an n x n matrix, x = (xl, x 2 , ... x.) an unknown vector

and y = (y1 , Y2 "', y) a known vector. This matrix equation can

be conveniently written as a partitioned matrix (AIl) with a column

vector [2i]
-y
(Ali) C]L = 0 (8-5)

where I denotes an identity matrix.

The elr entary row operations are defined as

a) multiply any row by a constant;

b) interchange any two rows;

c) add one row multiplied by a constant c to any other row,

or their combination.

When the elementary row operations are performed on (8-1), or equivalently,

on the pextitioned matrix (All) of (8-3) they do not change the solution

of the system of Eqs. (8-1) or the matrix equation (8-3).

-155-
 Let L be a matrix representing some combination or product of the

elementary operations described above. After these operations on (8-3),

we have
(LAI L) []=0 (8-4)
-y

If we carry out these elementary rowoperations so that LA becomes an

identity matrix, i.e., LA = I, then the inverted matrix A is found to

be A7l= L
(8-5)

B. Method of Gauss-Jordan Reduction. The method of Gauss-Jordan

reduction is 9.systematic procedure designed to execute the method outlined

in Section A. Let A be an n x n matrix to be inverted. The partitioned

mtrix of (8-3) can be written as an n x 2n matrix

all a 12 ... a,,-- a,, 1 1 0 ... 0 O0

a ... a 0 1 ... 0 0

aa2 a
"
-- II"2n

n-l,l n-1,2 an-l,n-1 an-.,nI 0 0 1 0

ai
ni an2 ... n.r -_L an
n.,n 0 0 ... 0 i
(8-6)

Search the first column a., (i = 1, 2, ... , n) for the lirgest

element in :agnitude, say a

jaj > Ja.1 1 (i 1, 2, ... , n) (8-7)

IIf akl = 0,then the first column of A is zero, A is a singular matxix

and hence A has no inverse in the usual sense. Therefore we may assn.e

akl j 0. Interchange the first and k-th rows, and then divide the firs-

row through by akl. If the interchange of the first and k-th rows is

followed by an interchange of the first and k-th columns of the second

-154-
submatrix. then the resulting mat.rix becomes more 'dly.
oake

sur. that the interchange of cuwuIs of the second submatrix does not

disturb the matrix equation (8-3) or (8-L), it is required that Yl and

yk be interchanged also. Finally we use the first row to eliminate all but

the first term of the first columm by the elementary row operations. Thus

((1)AI()
L x__
(L= 0 (8-8)

where L , = M()y [M(1) is a permittation matrix

which transforms y = Y, "(1) = (YI Y2 ,

Y3' '

Yk-l' Yl' Yk+l-' "'"Yn)11

(L(l)A'L(1))--1 bl 1 ... bl lrn 1 bln bl1 0 ... 0 01

0 b ... b, bn b 1 ... 0
22-l~ 21

0~~ b n-12n-!
""-.,
. bn-!,n1 n-!-,l0..1 0
bn2 ... bnn_! b 0 0 1

(8-9)

The elements of the matrix (8-9) can be written in terms of the elements

of the matzix (8-6) by inspection.

We proceed to obtain a new mat-- ix equation

(L(2)AIL ()) [I ] 0 (8-10)

from (8-8) by the elementary row operations such that the submatrix L(2)A

has zero elements for the first and second column except the first and

second diagvnal terms which are normalized to unity. This can be achieved

by the folLowing steps::

1) Find an integer k such that

Ibk 21 > lb 21 (i = 2, 3, ... , n-l, n)

-155-
 The Ibk,21 may be assumed to be non-zero otherwise the original matrix

is singular.

2) interchange the second and k-th rows and also interchange the

second and k-th columns of the second submatriy,

3) divide the (new) second row by bk,

4) use the second row to eliminate the second column by the

elementary operations except, of course, the second element of the column,

5) interchange y2 and y1C.

It is readily seen that these steps lead to the matrix equation (8-10).

Assuming that the matrix A is nonsingular, then after repeating the

process n times, we obtain

(Tn)AI (n) x =0 (8-1!)

-Y

where L(n)A is an identity matrix, and (n) is a vector resulting from

the rearrangement of the original y. Let M be a permutation matrix

transforming y into (n)

y (8-12)
Since M is a permutation matrix so that J., L(n) d L(n)
= i adL are related

by L(n) = (n)M (8-13)

Equations (8-12) and (8-13) imply that if y, and yj are interchanged in

then b kiand b of' Ltn, (k = 1, 2, ... , n) must be interchanged.

IJ C. Commatative Matrices. When a boundary value problem has some

degree of' symmetry such as rotational symmetry or reflectional symmetry,

the matrix obtained by the method of approximation described in previous

chapters often exhibits a special structure. This special structure of

-156-
 - a c ie by the ±= Lawt ti matGrix

commutes with another matrix which represents the nature of the symmetry.

The product of two matrices A and B depends, in general, on the order

of the factors, i.e., AB A BA. In the very special situation when the

product is independent of the order of the factor i.e., AB = BA, -L en

the matrices A and B are said to commute. A case of special interest

to us is when a given matrix com__irtes with a diagonal matrix.

Any matrix that commutes with a diagonal matrix having distinct

diagonal terms is necessarily a diagonal matrix.

To prove this statement let A = [' ,] be an arbitrary n x n matrix

and D be an n x n diagonal matrix with distinct diagonal terms:

D = [d. 5. . j d.
d] if i j
2. 23

Since AD = DA, we have

[ . F) .( d .F 3 [ 1 j

which is equivalent to

n n

q-~ 8 k3
dkl L d- 6ik dkj~
k=l k=l

This equation leads to

[a(d. - d.)]=i

where [0] denotes a zero matrix. The above equality holds only if

a 0 if d d for i i j

which shows that the matrix A = [a-. ] must be diagonal.

It follows from the proof given above that if a matrix A commutes,

with a diagonal matrix D whose diagonal terms of equal value appear

....e -3y, then +h tr A =" b a. e!j-+ M atv with zero off-

diagonal submatrices- -,or example, ir A is a b x b matrix Uij] and D

is a diagonal matrix:

D= d1 0 0 0 0

0 d2 0 0 0 i]
I0 0 d 0 0 0

0 0 0 d3 0 0

0 0 0 0 d- 0

0 0 0 0 0 d3

then AD = DA imlies that

A = Fail 0 0 0 0 01

10 a2 2 a2 3 0 0
%2 %3 0 0
0 0 0 a4 a,, (8-a,)
1a 0a-- 0xn. 6

L[0 0 0 a a 65 a6

It is well known that a matrix equation remains invariant under a

similarity transformation which includes unitaay and orthogonal trans-

formations as its special cases. For example the matrix equation

AIC = CA (8-15)

which denotes the conmratability of the matrices A and C. after a

similarity transformation (by a non-singular matrix P) takes the same

form
BD - DB (8-16)

where

-358-
 B P 'A? and D= P CP (-7

-Now suppose that C of (8-15) can be diagonalized by' a similarity

transformation and that the diagonaLized mat-rix D has distinct diagonal

terms. Then the matrix B of (8-16) obtained by the same similarity

transformation from matrix A must be diagonal also. On the other hand,

if D)is a -diagonal matrix with some repeated diagonal terms as shown in

(8-13), then B must be a supermatrix with zero off-diagonal submatrices

as shown -in (8-14).

The invariant property of a matrix equation under the similari-ty

transformation sometimes can be ussad to simplify the inversion of' a matriX.

Let A be a matrix to be inverted. Suppose we can find a mat rix C which

commutes wIth A and-whose eigenvalues and the corresponding eigenelements

can oe easily found so that a nonsingular matrix P., which willI diagonalizze

C by & 'similaritly transformation, can be obtained. In addition, if the

diagonalized matrix has distinct diagonal terms, then the matrix A is

reduced by the same similarity transformation to a diagonal matrix which

can be inverted easily. The desired inverted matrix A7 can be obotained

by reversing the original- simila-rity transformation. In case the diagonal

terms of the diagonalized matrix obtained by the similarity transformation

are not distinct, the matri>y A is reduced by the same similarity -trans-

formation to L. supermatrJx with zero off-diagonal submattrices. The inverted

matrix A is then obtained by first inverting 'rnese diagonal submatrices

larity transformation.
and then reversing the original- sim-iA

The pr A..acticab4ility of the methzod described above depends on whether -

we can find a second matrix which co 1=,dtes with the Matrix to be inverted

and vhose eigenvalues and eigenelements are 'rno-mn %j easily obtLai nable

- 150-
so that a nonsingualar matrix caa be construct-ed to diagonalize the matrix.

TO ftind such mat-rix, we consider the a pacLiLic configuration of a four-

port network -hown in Fing. 8-1. It is easily, seenx that. when the network~
undergoes a rotation of nr/2, n, r112 thelctiasrctere in

unchanged. Let e = (e, e, e, ej denote the set of excitations, r

the set of corres~onding responses in the net-won't due to the excitation

and S the matrix representation of t.he network parameters det-erinhed by

the electrical- structure of the network. T7hen the eqraation relating

the excitations and res-oonses of' the network is given by

- Se 818-
"

Ljet R be a rotating opoerator signifying a rtatio fr2i the

counterclockwise direction. A rot-ation of the s!Lt of IeXItions i n the

counterclockwise direction by an an;-Ire of 4z2, keeping the ne-twiork

stationar, gives a set of- new responses

r'- SRe kr-±~

On thle -other hand, the set of -new responses r t is ecrual to the set of

responses ob-tained by rot'ating thne whole system-, includina both the

exciltatio-J: and netwA.ork, -c0=ZAreClockwi--se 'by an angl-e ol- R-/2:

r rL= M~e (-

Thus b y (819 2n-0 we find, for every excittion e

P= S

the lly t- S iratrix -Which rve'resezfls the electri~lcstructure of

 lea I

R2

47t
-±0±0

Fig. 8-I A four port network with rototionvil symmetry. 3

rather co-liatt while the R matrix which characterizes the s 'etry

of the network or the problem is relatively si.mple. This feature is

estiecially evident when the problem has a high degree of s;r-ietlry.

D. Example. Circular ioop Antenna. The boundary value proble of

a circular loop antenna possesses a high degree of symmetry, namely

the rotational syzmetry about the center of the loop. One of the m-t

effective aproaches for obtaining an approxim ate solution is to divide

the circul oop in N eaual sections and rega.d the antenn-a as an n-oort

network. 'The matri x equation obtained by aoproxn-~zr. intgro-

differential exuation describin.g the bond-ary value proul is a n x n

M-trix (ze.dance matrix)

The problem is to invert the matrix A to obtin an admittance matrix.

it is seen from Fig. 8-2 that the boundarj value proble is

invarant with respect to a rotation of the system by an 2'e
2l.iN

(k =, 2 ., in coun-tercockwize sen-. Since this rotation can

bereresentmed the C mti

LS...
10 0 1 0
. .- .. .J . . .
.. o. I

10 0 0 0 . . . 0 1
L. . .. . Fi

he ma-tri-ces A and C ith each othe:

Next we consider the dialonazation of the C For tishtrix.

purpose it is necessary to obtain the eigeLvalues and eigenellemernts of

 2

N-tN

Fig. 8-2 Circular loop antenna regorded as an N port

network.
 C matrix. The characteristic equation

(C - XiVx = 0 (8-26)

gives the N eigenvalues and the corresponding eigenelements

e N, k =0, 1, 2, ... ,111-1 (8-27)

(N -1J 2&1k
1 e k = 0, 1, 2, . N- (8-28)
L 9=0 FN

wh L'e I denotes a unit element in the N-dimensional Euclidean space.

The diagonalizing matrix P can be obtained from the set of normalized

eigenelements by regarding them as the column elements of P:

P [ 1 [e - ] (k,1 =0,1 , 2, ... , N-I)

(8-29)

The inverse of P is seen to be

. 21rki
p-1 = pe N (kA = 0, 1, 2, ... , N-1)

(6-30)
TL 1 t P is a unitEry matrix can be seen from the following relation:

N-I 2vZk 2 .Im

(FPkm
(p*p)-J =j _ -N J -N IN
1=o
lN-i J 20(m
N - X)
[g
=
'
e N

m The mat yx C
can be diagonalized by the following similarity transformation:

2nk
= [e N ki]NxN (k,i = 0, 1, 2, ... , N-i) (8-71)

-164-
w.ich has N distincs diagonal terms equal to the eigenvalues of C. It

follows that B = P1AP must be a diagonal matrix also.

We now evaluate the diagonal terms of B. The elements of B, in general,

can be calculated by the formula

N-l . 2ir(km - 1i)

Rg [i ]*;*[ak ] [yk,]- 1]
IT , - e N (8-32)
(-
i ,k=O

According to the results in Section C, we see that the off diagonal

terns vanish, i.e.,

P m= 0 if2 8 m

For the diagonal terms, we have

N-I j 2_(k-i)

P22 ZE
i k=O
.e N(8-33)

This expression can be simplified by making ase of the commutability of

matrices A and C.

It follows from the relation AC = CA that the elements a.,k of A

depend only on the relative positions of i and k. Let p denote the

relative position of i and k and write

a -=
a. (8-34)

A moment of careful examination of (8-33) and Fig. 8-2 reveals that (8-33)

may be regrouped as follows:

-165-
Integral number assumed The partial sum of (8..33) corresponding
by p to p
S2:ip
N e N if N is even

Na e Nif N is odd

[N] +1 N ae N~
e p

0 N e N
p

2ggp
N-e N

SifN is even

N-c p e if N is odd
P

Wnere the square bracket [Q] denotes the maximum integer of Q, e.g..

L2.1341= 2.
m t.. of these terms over p yields

=N2 , 1N i
5(2 S M e j - N(--

p=-[N12j (8-35)

If, in addition, the reciprocity property holds as in the problem

of circular loop antenna, then

aA %i (8-36)
or
a_ = a (8-37)
-p p

-166-
 voneipntl, k-'.,5 can bt! written as

[N12]
2 - b(p) - 8(p - a) cos (8-38)

which represent the diagonal terms of matrix B. Thus after the similarity

transformatio' of matrix A, we obtain a diagonal matrix

B = P- AP = [gj bgm] (8-39)

The inverse of this matrix is easily seen to be

B A P= 5=Pm (8-40)

-1.
Therefore, A1 is given by

A7 1 PB-'P'l
2:mik 2nim
= [e N I l [e I
kk

N-1 2,rk(2-m)
N
E[ e
: T.e N ] = [a,] (8-hi)
k=O

Thus the elements aim of A- can be identified as

N-1 . 2i (2-m)
a , et N (8-42)
k=0

It suffices to evaluate th3_ elements of the first row of A- 1 since the

elements of the other rows can be constructed from these elements. For

the first row, we have

N-1 2,km
a e (8-43)
-17
AigaLi for thlose r-robleins which satsfyte
Q.L recproit ;I3trty4w1 h

N- I

where kkare gi-.ven, by (8-38).

As an example, we consider the inversion of a .4 x 4 symetric

Ara a - -a
0 L
a a1
C a,1
cx
>2 acx,(8-4:)

ja, a2 a1, 0

whicn commutes wi th11 rotation maetr ix- ooterator

10 0 0
10 0 0 1
LI 0 v0 ol
By (8-38), mat-rix A can be diagonaliLzed by a simila-rity trans-

formation to give a diagonal miatri;3x B w-ith,';A the following diagonal -t-erm-;s:.

A00 a+a~,.
= ad c-+a

p11=f~=a
- (o-
a -2a,+ a

By t;-he elements off t1he -first row ae foumnd t..be

d
.= - t r-- -
+-
00 k

-o+ B
00 4eL

2 -0 -
The inverted matrix A - 1 is found according to (8-h_ with the eAirnt_

of the first row given by (8-48):

=a000 aa 11 0
a2
02
a.
a 3 0

a0!
0
00
a1 ! a 01a 2 (8-49)
(-9
02 01 00 %l
Laoa2
L% aa 2 aa1 a0
a0

E. Example. Linear Antenna. As another application, we consider

the problem of a linear antenna. The electrical property of the antenna

remains unchanged by rotating the antenna with respect to the mid-point

by an angle 1 . e matrix equation intended to approximate the inzegro-

differential equation describing the boundary value problem can be arranged

by a proper choice of the approximating fDanctions (elements) to have a

matrix A = qj] which conautes with the matrix representing the rotation:

R 0 0
I,
... 01 i
;o o ... .
I......... 1 (8-50)
to 1 - - - 0 0
Fo 0 . .. o,01

that is

AR= IA (8-51)

We sball show that the inversion of matrix A can be reduced to

the inversion of its diagonil r-Wmatrices. To be specific Let An , R2

and In denote matrices of order 2n. The characteristic equation for

B2n
! 2n - I =~ o(8-52)

112n - =-
'L-ni
gives the eigenvalues +1 and -1, each of which repeats n-times.

For the eigenvalue X = 1 and x = (yi" '' ""

.n P "" ) the

equation

n
(.R -%12 )x =0 (8-53)

Tields the following 2celations to be satisfied by the eigenelements:

l- 2n =o

- E2n-l (8-54)

En- 'n+l =

The corresponding eigenelements, after normalization, can be choEen as

U 1 (1, O, 0, ..., O, 0, 1)

% (0., 11 0, 0 O y0)
,
• ,2 (8-55)
_1

'12
u - (0, .., , i, O, O, . 85)

Similarly for the eigen-alue X = -1, we find the n normalized eigenelements

as follows:

, O)
(C, .- ., O, 1, -it O, ...1

n--.
_ i (0., ...1 1, OP 0, -1, ***1 0)
2 T1 (8-56)

U2n - (i, O , 0, 0, 0, -1)

-170-
 The matrix Pn which diagonalizes R by a sinilarity transformation

may be col structed from these eigenelemeats by regaiding them as colmn

elements of P n
2n*

F2
1 0 0 1.0i

I. ... . . .I (8-5.7)
t 7)

0 0 ... 0 -i
00 ... 0 0
L

which can be put into a more compact form

= [ (8-58)
.1- _ j

The inverse is seen to be

- [1 R~] (8-59)

L~ .59Ifl

;-ot. t --.
letht 2n - 2n' R2 2n and Rn Rn
Putting F

An J- (8-6i)

and using the fact that An commutes with R2n , i.e.,

2n 2n -2n2n
We find that A2n can be rewritten as

n= [A RnA'iRn] (8-63)

LA2- ~ U~

The similarity transformetion of by Pn yields

B2n 7-P2 1 A2nP_ JI 8-4

it(A R 2 , )R 1
L o Rn(A - j

The inverse is
.(A + R.A 2 1) o

-1 -i-1

8-6@)

It follows that the inverse matrix A- is fund to be

'2n

2n 2nB2n n
[ (A,+RnA ) + (Anl2n )-- [(Al2-- -(Al- R )] R

R (,R(A, A-%A [. (A,R]l ln l'

= R--A 2 .- 2 1) R1 L(+(;'-RA 2 ,I

(8&66)
Thus, we have reduced the inversion of An given by (8-6) '--1 the inversion

of the subma.trices (All + RnA2 ,) and (All- R-A21 )

F. References

1. E. P. Wigner, "Group Theory and Its Application to the quantum

Mechanics of Atomic Spectra," Academic Press, New York, 1959.

2. F. R. Gantmacher, "Thr Theory of Matrices," Vol. I, Chelsea

Publishing Company, New York, 1959.

-172-
 3. J. S. Frame, "Matrix Operations and Generalized inversems-Part I,"
___Sectrum,. vol. 1,r~. !,Kar. IC-A015j1.
4. B. A. mid, "Appnications of Group iTheory in the Study of
Symmetrical Wavegaide J1rnctions," Ph.D. Dissertation, Electrical Engineering
Deartment, Stanf-ord University, March, -92.

im

Lj2

-
 UNCIASSIFIED
S-cuitv COwi.ifiction
DOCUME4T CONTROL DATA - R&D
(Security c.niftcaflat of na. bodyj* M
otwfw nc Md~z4 rwot*Umum b e antsad s ~mno jmfB
l 1*0di
c.aif!;1
RIGINiATIN C ACTIVITY (C~roromt.*, Ig EhATSSUA C LAW?:ICATION
Syracuse University Research Institute Unclassified
Syracuse, Neu York 1-210 i
- N/A
3 REPORT TITLE

Matrix Methods for Solving Field Probils

Volume I, Matrix Techniques and Applications

. DESCRIPTIVE NOTES a of repr 1 cl dna.)

Final Report, Mrch 1-65 to March 1966
Mst nrne. _rst nn-e. fmi"e1
S AUTHOR(S) .(
Harrington, Roger F. Wallenberg, Robert
Chang, Long-Fei Bristol, Thcmas
Adams, AxIon T. M~utz, Joseph
-
6- REPORT DATE j 70 Or floES
TOTAL NO- -. 10 oor
, grs
August 16 5i_
BeCCONTRACT OROt A ! PI0)(602 )1-3Teu92OnKa4s WE l~s~S

b PROJECT r.0- 's19

-- Task 110. 451901 6S&5. O-11Itf REpOlt! %O(C) (Afy cans aaew W lyZetnpjl j

RADC-T-66-35-1
10 A V A IL ASILITYLINITAT!C4 NOTICES j
Distribution of this doc=--ent is un!ii-ited.

I- SUPPLEMENTARY N0 TZ 12- SPCONVtRN U.TARY ACTIfVFY

John J. Patti z Rom A..- Deyela1ent Center (17WE-U-)
Project Engineer (F2CRR) Griffzts Air Force Base, Sew York l3bt0

13 ABSTRACT

This report presents general procedures for solving field problems of

engineering interest using dgitI comuter techniques. The basic coneis to
represent a boundary-value problem by a sunern-ostion integral, a--poxite the
integral equation by a matrix equation, and invert the =atrix for a soluaion.
The tbeory is described in tern of the ethod of ets, which is ecuivalet
to the variational method.
For electromagnetic antenna and scattering problems, the method gives a
matrix whose elements can be interpreted as generalized ie ces. These
impedances are closely related to those used in the theory of loded antennas a
scatterers, end hence such loaded structures can also be trtd. A S t
vi;-e antennas and scatterers of a--bi-r-y shape is forwlated in detai, and
calculations for linear inre antennas and scatterers, both loaded and -itdM,
have been made.
Additional problems treated by these procedures are J-=81 scater-
i by conducting cylinders and by dielectric cinders, and th-- ional
scattering by b dies of revolution. These problcrs are used to show the effect
of various approximations in the aoiu- n, -inan attest to draw some ge-eral
conclusions &s to the best approximatics. Special -rocedures for inve-r-i--g
matrices have also been considered, to take into accou-nt any sy~netry properitiesj
present in tbe matrices. A considerable sawing in coztation time can often be
made by pr-operly utilizing these symetries.

D D 'A 4113
Jf W147
&curix
rAWIFME
Qansicision
 ________ ~'CIasificatoj11
K!:Y WORDS ~~ A LINK C
ROL RDLE ROL

f
________________ W WTe WT

Nuamerical Analysis I
j Ccmiputer Techniqueu i
Anatennes Electromagnetic Waves
Matrix Algebra

INSTRUCTIONS
1. ORIGINATING ACTIvITY:. Enter the name and address imposed by security classification, using standard Statements
of the contractor, subcontractor, grantee, Department of De- such as:
thne activity or other organization (corporate autnor) issuing ()"alidrqetr a bancpe fti
report.report from DDC."
2a.REPRT LASI~CATON:Entr
ECUTY te o'~ (2) "Foreign announcement and dissemination of this
all security classificrition of the report. Indicate whether rpr yDCi o uhrza
"Restricted Data" is includei. Marking is to be in accord-reotbDDisntahrze.
ance wit appropriate security regulations. (3) 11U. S. Government agencies may obtain copies of
this report directly forn DDC. Other qualified DDC
2b. GROUP. Automatic downgrading is specifted in DoD Di-ussshlreettrog
rective 5200. 10 and Afr'ed F'orces Industrial Manual. Enter
the group niimber. Ak -. when applicable, show that optional
markings have been usied fvr Group 3 and Group 4 as author- (4) "U. S. milit-ry agencies my obtain copies of this
ized.
raotdiret. - from DDC Other qualified users
3. REPORT TITLE: Enter the complete report title in all shall requimt through
Capital 'Lcters. Titles in all cases shouid be unclossified.
If a mesningful title cannot be selected without classifica-
tion, stow title classification in all capitals in parenthesis (5) "All distribution of this report is co-trolled. Q'L:9-
xrnmedi-tely following the title. ified DDC users shall request through
4. DEI.CRIPTI1172 NOTES.~ If appropriate, enter the type of
report. e.g., inter.m, progress, summary, annual, 0r fial If the report has been furnished to the Office of Technical
Give the inclusive dates when a specific reporting period is Services, Department of Commerce, for riale to the public. indi-
covered. cate this fact and enter the price, if known,
5. AUTHOR(S): Enter the name(s) of author(s) as shrwn on IL SUPPLEMENTARY NOTES: Us.- for additional explana-
or in the report. Entet lost name, f-irst name, middle - titial. tory notes.
If military, show rank and branch of arvice. The name of
the principal 0 z'lthor is an absolute minimum requirement. 12. SPONS07 iuG MILITARY -rJVfl1Y: Enter the name of
6. he dte f th reort s ly,
EnerDT---
F-1PRT e departmental project office or laiboratory sponsoring (pay-
6. PPOREterDAT.
he ateof te rportas aying for) the research ae development. Jr'clude address.
morth, year; or month, year. If more than one date appears
on the fepoft, use dete of publication. 13. ABSTRACT: Enter an abstract giving a brief and factual
7a. OTAF PGES:Thetotl
NUMER pae cunt summary of the document indicative of the report, even though
shoul TOTLo normERO pagiwonpceusie
.SThtoal cntth it may also appear elsewhere in the body of the technical re-
shou~l agiatin
oilo nomal
prcedires i~., eterthe port. If adciitional space is required, a continuation sheet shall
number of pake'. containing information.b atchd
7b. NTJMVER OF REFERENCES; Enter the totpl nur-ber of It is highly desirable that the abstract of classified reports
references cited in- the report. be unclassified. Each paragraph of the abstract sbe end with
8a. CONTRACT OR GRANT NUMBER; If qppropripte, enter an indication of the military security classification of the in-
the appl'cable number of the contract or grant tinder which formation in the paragraph, represented Gs (Ts). Cs). (c). or(U).
the report was written. There is no limitotion -%-.the le'.gth of !he abstract. OHow-
8b, 8c, & 8d. PROJECT NUME.ER: Enter the appropriate ever, the suggested length is from 150 to 229 words.
military department ideatification, such as project inumber, 14KEWOD:eywrsaetcncl mnigutes
subproject number, system numbers, task number, etc.14KEWOD:eyw.sar eincl mnngutrs
or srort phrases- that characterize a report grid may be used as

rproject
~' REOR"
9a. RIGNAT NUBERS):Entr te ofi- index entriet%for cantaloging the report.. Key words must be
cial report nufrbt by which the document will be identified Selectee so that no security classification is required. Identi-
and controlled bj, the originati-g activity. This numlmrein ust fiers, such as equipment mtxlel designation, trade ame, military
be unique to this report. code niste, geographic location, may be used as key
96. OTHER REPORT NUMBEF.,S): If the report has been words but will be followed by an indication of technical con-
asr.'gnf-d any other report numbers (either by the originator text. The a-wsignwent of links, rules, and weights is optxonal. 4
or &y' Me sponsor). also enter this number(s).
10. AVAILARILITY/LIMITATION NOTICESk Enter any urn-
122 tions on furthei aisse-mination of the rep-irt, ither bhan these

UNCLASSIFIED
Security Classificatiem