You are on page 1of 12

IEEE

TRANSACTIONS

63 1

ON ANTENNAS AND PROPAGATION, VOL. A'-28, NO. 5 , SEPTEMBER 1980

A Uniform GTD Analysis of the Diffraction of Electromagnetic Waves by a


Smooth Convex Surface

Absmcr-The problem of the diffraction of an arbitrary ray optical


electromagnetic field by a smooth perfectly conducting
convex surface is
investigated. A pure ray opticalsolution to this problem has been
developed by Keller within the framework of his geometrical theory of
fails in the
diffraction (GTD). However,theoriginalGTDsolution
transition region adjacent to the shadow boundary where the diffracted
A uniformGTDsolutionis
developed
fieldplaysasignificantrole.
which remains valid within the shadow boundary transition region,and
which reduces to the GTD solution outside this transition regionwhere
the latter solution is valid. The construction of this uniform solution is
based on an asymptoticsolutionobtained
previously for a simpler
canonicalproblem. The present uniform GTDsolution can be conveniently and efficiently applied tomany practical problems.Numerical
results based on this uniform GTD solution are
shown to agreevery well
with experiments.

I. INTRODUCTION

HE PROBLEM of the diffraction of an arbitrary ray optical electromagnetic field by a smooth perfectly conducting
convex surface of slowly varying curvature
is investigated in
this paper. The convex surface is taken to be electrically large.
A pure ray optical solutiont o this problem has been developed
by Keller [ 1 ] - [ 3 ] within the framework of his geometrical
theory of diffraction (GTD). However, the GTD solution fails
in the transition region adjacent to the shadow boundary(SB)
wherethediffracted
field playsarathersignificantrole.
In
thepresentinvestigation,auniform
GTD solutionwhich is
convenient and accurate for engineering applications is developed for this problem such that it overcomes the limitations
of the GTD within theSB transition region. Besides being valid
within the SB transitionregion,thisuniform
GTD solution
also automatically reduces to theGTD solution exterior to the
SB transition region where the latter solution is indeed
valid.
The geometrical configuration of the diffraction problem
being investigated is illustrated in Fig. 1. Referring t o Fig. 1,
one notes that an extension of the incident ray beyond the
point of grazing at Q l on the convex surface defines the SB
which divides the space exterior to the surface into the lit and
shadow regions. According to the GTD, the total exterior
field
is associated with the usual geometrical optics (GO) incident
and reflected rays in the deep lit region, i.e., region I of Fig.
1, whereas in the deep shadow region, i.e., region 111 of Fig. 1,
it is associated entirely with the surface diffracted rays. Region
I1 on either side of the SB is the SB transition region in which
the field descriptionchangesrapidlybutcontinuouslyfrom
the GO field in I to the surface diffracted field in 111. The anManuscriptreceived June 13, 1979; revised April 10, 1980.This
work was supported in part by the AssociateJoint Services Electronics
Program under Contract N00014-78-C-0049, and in part by the Naval
Air Development Center and the Ohio State University Research Foundation under Contract N62269-76C-0554.
The authors are with the ElectroScience Laboratory, Department of
Electrical Engineering,Ohio State University, Columbus,OH 43212.

,
\

DIFFRACTED 1
RAY

L I T ZONE

1,
PL = FIELDPOINT

IN L I T ZONE

Ps = F I E L D POINT IN SHADOW ZONE

Fig. 1

Rays and regions associated with scattering by smooth convex


surface.

gular extent of I1 is of the order ( 2 / / ~ p ~ ( Q l ) )rad,


' ' ~ where k
is the wavenumber of the isotropic homogeneous medium exterior to the convex surface, and p g ( Q l ) is the surface radius
of curvature along the incident ray direction at Q l . Regions
IV, V, and VI lie in the close vicinity of the shadowed part of
the surface which is alsoacausticof
the surface diffracted
rays; these regions constitute the caustic (or surface) boundary
layer. It is noted that region IV is in the lit zone; whereas V
and VI are in the shadow zone. Specifically, theGTD solution
fails in region I1 and it also fails in the surface boundary layer,
whereas the present uniform GTD solution
fails only within
the surfaceboundarylayer.Aseparateapproximateasymptotic solution will be reported in the future to describe the
fieldswithin the surface boundary layer in a form which is
suitable for applications.
The ansatz orformulation of the presentuniformGTD
solution which employs GTD rays is based on an asymptotic
solution given recently by Pathak [4] for the canonical problem of plane wave diffraction by a perfectly conducting circular cylinder. The work in [4] extends andimproves our earlier
work of Fock [ 61 .
work [ 51 ,and it draws upon the pioneering
Inthepresentworkit
is assumed thattheincident
highfrequency electromagnetic field can be described according to
ray optics [7].Thus the wavefront of the incident ray optical
field is describedlocallyby
an arbitraryquadraticsurface,
and the field is polarized transverse to the ray direction. It is
noted that plane, cylindrical, conical, and spherical type
incident wavefronts are simplyspecial cases of the arbitrary quadratic wavefront. It is also assumed that the field point and the
caustics of the incident ray system are not in theclose vicinity
of the surface, and that the amplitude of the incident ray field
does not exhibit a rapid spatial variation invicinity
the
of Q1.
First, the GTD solution to the problem under investigation
is reviewed in Section I1 for the sakeof completeness. The uniform GTD solution for this problem
is developed in Section
111. Several examples illustrating the usefulness and accuracy
of this uniform GTD solution are presented in Section IV. An
d w rtime dependence is assumed and suppressed in the present analysis; only here doesf refer t o time.

001 8-926X/80/0900-063 lS00.750 1980 IEEE

632

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-28, NO. 5;SEPTEMBER 1980

11. THE GTD SOLUTION-A REVIEW


This review serves to introduce the GTD format and notation that will be employed in the developmentof the uniform
GTD solution. Consider a ray system which
is incident on a
convex surface; it produces a system
of rays reflected from
that surface, and the field of these incident and reflected rays
constitutes the usual geometrical optics(GO) field. At grazing,
the reflectedraymergeswith
theincidentraytoforma
shadow boundary (SB) which divides the space exterior to the
convex surface into the lit and shadow regions,
as shown in
Fig. 1. Also, the incident ray whichgrazes the surface launches
Keller's surface rays that propagate along
geodesic paths on
the convex surface while continually shedding energy, via diffraction, along the forward tangents to the surface ray paths.
The field of these rays which are shed (or diffracted) from the
surface is known as the surface diffracted field. The GO field
is zero within the shadow region behind the obstacle; thus the
field within the shadow zone is produced entirely by the surface diffracted rays. These surface diffracted rays may also be
present in the lit regionif the surface of the obstacle is closed;
however, their field is generally negligible compared to the GO
field if the closed surface is sufficiently large in terms of the
wavelength.
The GO solution for the lit region is brieflyreviewed in
subsection A of this section, and likewise a brief review of the
surfacediffractedraysolutionfor
the shadowzone is presented in subsectionB.

The superscripts i and r refer to the incident and reflectedGO


fields, respectively; thus E' is the incident field and Er is the
reflected field. Note that pl'.' and p z K p rare the principal radii
of curvatures' of. the incident or reflected wavefront surface
dA0 at ro, and sZpris the distancealong the incident or reflected ray from O; .to ;as shown in Fig. 2. Next, requiring
V * E = 0 leads to ( ? z * r * E ~ z=* 0,
r ) which implies that the field
in (2) is polarizedtransverse to the ray direction (?'), as
shown in Fig. 3. The square root quantity in (2) is the ray divergence factorwhichindicatesthemannerinwhichthe
energy spreads along the ray path; it is a consequence of the
conservation of energy in a ray tube (or pencil). The leading
term in the Luneberg-Kline expansion for the corresponding
X YgF'9r in which
magnetic field intensity is Zztr iZtr
is given by (2) and YO is the characteristic admittance of the
medium. From the boundary condition i2 X [E' E'] = 0 on
the perfectly conducting surface, where is the surface normal
at the point of reflection QR as shown in Fig. 3, one obtains
the following if the reference point at O; is moved t o QR :

i?lr

Thus thereflected field becomes

A . The Geometrical Optics FieldSolution for the Lit Region


Luneberg [ 71 and Kline [ 81 developed an asymptotic highfrequency solution of Maxwell's equations in which the fields
are expanded in inverse powers of the angular frequency w .
The leading term in this expansion is regarded as the GO field.
The details of such an expansion are discussed elsewhere [ 71 [IO] and are only summarized here. According
t o Luneberg
and Kline, the electric field intensity 3 in a source-free homogeneous isotropic medium can beexpressed for large w by

in which r is the position vector of the field point, k = w / c ,


and c is the speed of light in the given medium. Substituting
(1) into the vector Helmholtz equation satisfied by E, namely
( V 2 k 2 ) E = 0, leads to the usual eikonal and transport equations, I V I ) l 2 = land
[ V 2 $ + 2VJ/*V]En = -V2En-1
(with E-' = O), respectively. The surfaces of constant $ are
referred t o aswavefronts,and
thefamily of all wavefronts
describe a system of associated rays which are straight lines in
a homogeneous medium. The rays are everywhere orthogonal
to the wavefrontsinanisotropicmedium.Integratingthe
transport equation for the
n = 0 case fromsomereference
point 70 to i: yields the GO or ray optical field:'

1 The positive branch of each of the square roots in (2) is chosen


with siJ > 0 in the direction of wave propagation.

with

Rs,h = T 1,

(5b)

where P I ' . p2', and sr are measured with respect to the reference point 70 which is now moved t o Q R , and R , and R h are
the acousticsoftandhardreflectioncoefficients.
Theunit
vector d l in (Sa) is perpendicular to the plane of incidence,
whereas the unit vectors $11' and 611' are in the plane of incidenceasshowninFig.
3. The principalradii of curvatures
( p l r ,p 2 ' ) oftherepectep
wavefrontandtheir
associated
principal directions ( X l r , X 2 r ) are given in [ 1 1I and are expressed more compactly here as
1
P 1 ,2r

Pm'

+
Pg(&

(QR

(li[

COS

6'

pg2

ei

4f

2 p m j J = f I p m i J I for a{dwerging/converging} pair of adjacent


rays. Here, m = 1, 2.

633

PATHAK e t al.: DIFFRACTION O F ELECTROMAGNETIC WAVES

= ULilT VECTOR
AXIAL

ALONG T H EC E U T R A LO R

il-r

d A o I S THE
IhCREMENTAL
WAVEFRONT
AREA

7,

OF T H E
RAY
P E N C IA
LTHE
REFERENCE
POIVT

d l

I S T H E I ~ C R E M E N T AW
L A V E F R O NATR E A
O F T H E R A Y P E N C IALTTHFEI E LP
DO I N T

Fig. 4. Geometry for description of wavefront reflected from a curved

Fig. 2. Astimagtic tube of rays (i.e., ray pencil).

surface.

where
l/pb' = (sin2 ao/P1')

+ (cos2 (Y,/p2i)

(8)

and pbiAis the radiusof curvature of the incident wavefront in


the (f, b ) plane (i.e., in the plane tangent
to the surface!at
QR fvr 8'
n/2. Furthermore, the principal directions
XIy
and X2' of the reflectedwavefrontapproach
the following
values for grazing incidence:
-+

n* =

U N I TO U T W A R DN O R M A LV E C T O R
C O N V ES
XU R F A C E
AT O R

Fig. 3.

where R 1 ,

TO T H E

-?lr-+;

(at Q R )

22r= (-ir x z l r )

Reflection by a curved surface.

-+

(gal
(at Q R ) .

of (YO.
The limiting valuesin (9a) and (9b) are independent
The total GO field E at PL (see Fig. 1) in the lit region is the
sum of the incident and reflected ray optical fields; hence

> 0 for a convex surface and

C6c)
Thequantities R and R2 constitute>hepri?cipalradii
of
curvatures of the surface at Q R , and U1 and U2 denote the
corresponding principal surface directions at
Q R . The radius
of curvature of the surface at QR is p g ; it is measured in the
plane of incidence which contains ?, fi, a n i i, where i is tangent to the surface. Also, p t is the radius of curvature of the
plane con!ai;ing
a?d the binormal
surfaceAat QR inthe
vector b. The unit vectors i, i , b , U 1 , and U2 are shpwn in
Fig. 4(a) together wit) th: angle wo between i and U 2 ;the
principal directions ( X l i , X 2 ' ) of the incident wavefront and
(YO are shown in Fig. 4(b). The angle of incidence 8' is defined
air.
by si' = -cos 8' = -;
It is noted that theGO representation of (2) fails at caustics
which are the intersection of the paraxial rays (comprising the
1-2 and 3-4 as shown in Fig.
ray tube or pencil) at the lines
2.. Uponcrossing acausticinthedirectionofpropagation,
( p z r r s i S r )changes sign under the radical in ( 2 ) , and a phase
jump of +n/2 results. Furthermore, the reflected field ?i of
(4) fails in the SB transition region. However, it is important
to note that near the SB (i.e., as 0' n/2),p l r and p I r of (6)
approach the following limiting values:

-+

(9b)

(10)

B. The Surface Diffracted RayField Solution for the


Shadow Region
The incident ray at grazing launches a set of surface rays
which propagate along a geodesic path on the convex surface,
thereby carrying energy into the shadow region. The field associated with these surface rays attenuates (Le., decays exponenof rays
tially)due t o acontinuoussheddingordiffraction
from the surface rays along the fonvard tangents
to the geodesic surface ray paths, as shown in Fig. 5. An analysis of this
surfacediffracted field is discussed indetail elsewhere [ 11,
[ 2 ] , [ 101, hence only the essential features are summarized
here. Let i 1 , 2 denote the unit vectors tangent to the surface
denotetheunitoutwardnormal
ray at Q 1 , 2 , andlet i 1 , 2
vectors to the sprface at Q l , 2 as in Fig. 5. Then the unit binormalvectors b1.2 = i 1 , 2 X A1,2 and the field canbeexpressedinvariantlyinterms
of ( t 1 , 2 ; j t l , 2 ; h 1 , 2 ) which are
fixed in the surface ray. The diffracted ray pencil
possesses
caustics at Q2 and PC as seen in Fig. 5 ; these caustics are the
same as those at 3-4 and 1-2 of the typical ray pencil in Fig.
2. The diffracted electric field Ed arriving at Ps may be repreis given by (2) with the
sented as a ray optical field, hence
superscripts i and r replaced by d t o denote the "diffracted"
ray. Then p l d and p 2 d become the caustic distances associated
(TO)
withthediffractedwavefrontatsomereferencepoint
which lies between Q2 and P,, as in Fig. 2. However, if the

zd

63 4

IEEE
TRANSACTIONS
ON

ANTENNAS AND PROPAGATION, VOL. AP-28, NO. 5 , SEPTEMBER 1980

hard type field contributions, respectively. D p S t h ( 4 1 , 2 ) are


referred t o as the surfacediffractioncoefficientswhich
deof DpSp(Ql)and
scribe the
diffraction
at
The
forms
D p S p h ( Q 2must
)
be identical via reciprocity. The factor
-jkr-J~:apSSh(r)dr

is the ratio of the surface ray field incident at Q 2 (prior to


diffraction from Q2) to the surface ray field launched by the
incident field at Q1 ; it is obtained by integrating

Fig. 5.

(b)
Diffraction by a smooth curved surface. (a) Top view. (b) Side
view.

referenceposition ro is movedback tothe causticat Q 2 ,


then p l d -+ 0 , and p z d now becomes the distance from Qz
t o PC. Since Z d ( P s ) is independent of 70,one may define the
following limit:

Therefore,
the
electric
field
region (see Fig. 1) becomes

Ed point
at
a

P, in the shadow

Q2.

Theform of thesolutionin(12a),(12b),and(12c)has
been verified via asymptotic solutions t o appropriate canonical
problems [3], [ 121 ; also, these canonical solutions lead to the
specific
expressions
for
DpSphand
which
to
the
first
order are given by [ 3 ] , [ 121

(?)

[Dps] =

113

2rrk

(1 2 4
? ( Q l , Q2 ) is dyadic
a
transfer
function
which
relates
the
fielddiffractedfrom
Q2 to the field incidentat
dyadic
quantity
is expressed
by
[ 31 , [ IO]

from Ql t o Q 2 , with a,,h2 being theintensity of the pth


surface ray field3 anddt is the incremental arc length between
Q 1 and Q 2 . This factor is based on the assumption that the
rate of decrease of energy flux at any point along the surface
ray due t o diffraction is proportional t o its value at that point;
theconstant of proportionality is chosen as 2anS3h,where
apsps.l is the attenuation coefficient. The factor e-kr denotes
the dominant phase delay of the surface ray field from Q l t o

aps

3
pg

This

41.

aP

It

e-i(7V15)

[Ai(-qp)l

(!!)
(?)
I3

(1W

$(?r/6),

113 e1n/6.
.

=
Pg

The Ai denotes
the
Keller type Airy function,
and
Ai is
its derivative. The values q p and T p arethoseforwhich
Ai(-qp) = 0 and Ai(-qp) = 0. The values of Ai(-qp) and
Ai(-qp) are given in [ 131 - [ 151. It is noted that p g refers t o
the surface radius of curvature in the ray direction (i.e., in the,
inclusion
the(12c)
ri, t* In
plane).
couple
firstof
theonly
of
modes (i.e., p = 1, 2) is sufficientforobtainingaccurate
results in the deep shadow region. However, this solution fails
( 2c) in the SB transition region; it also fails near the caustic at Q 2 .

T(Ql,Qz)=~l~zTS+;21;12Th,

in which
N

Ts,~l=

D,(Qd

[e -jkr-&aps(r)dr

P=l

,/~]Dpsoh(Q2),

with dQ being the width of the surface ray tube (or strip) as
in Fig. 5 and t being the geodesic arc length from Q l t o Qz
on the surface. It is clear from (1 2b) that the
ray field in 12a)
is polarized transverse to the ray path. It is assumed in (12a)
and (12b) that the & and 6 2 components of Ed propagate independently of each other. The sum in (12c) indicates that the
surface ray field is actually composed of a set of surface ray
modes as indicated earlier, and p refers to the modal index.
The superscripts s and k in (12c) denote the acoustic soft and

<

111. THE CONSTRUCTION OF A UNIFORM GTD

SOLUTION
Auniform GTD solution which remains valid within the
shadow boundary transition region is developed in this section.
It basically departs from the pure ray optical field approximation inherent in the GTD in order t o correct for the failure of
3 uS,h f ApSh(t)e-ikr in which Aps,h(t) isslowly varying in comparison to e-lkt. Here t is the surface ray dlstance measured from 0 1 .

635

PATHAK et a[.: DIFFRACTION O F ELECTROMAGNETIC WAVES

the GTD within the shadow boundary (SB) transition region;


outside this SB transition region it reduces t o t h e usual GTD
solution where the latteris indeed valid. The precise manner in
which a uniform theory accomplishes such a task differs with
each ansatz. While different formulations of uniform solutions
might even lead to the same answers exactly at the SB, and
also exterior to the SB transition region where they all must
reduce to the GTD solution, their behavior within the transition region may notnecessarily be the same[ 191.
As mentionedinSection
1: the ansatzemployed in this
paper is based on an asymptotic solution given by Pathak [ 4 ]
forthe canonicalproblemofplane
wave scatteringby a
smoothperfectlyconductingcircularcylinder.Anappendix
is included here t o correct a few minor typographical errors
presentin [ 4 ] . T h e presentansatzleads to a uniformGTD
solution for the general problem of the scattering ofa ray optical electromagnetic field by a smooth perfectly conducting
convex surface of any shape, such that the solution thus obtained is convenient and accurate for engineering applications.
The starting point in the development of the solution
t o this
general problem is the uniform GTD solution for the far-zone
field u of a scalar point source radiating in the presence of an
acoustic soft o r hard smooth convex cylinder. The latter solution is developed in detail in [ 161, hence only the important
aspectsofthatsolutionanditsdevelopmentarebriefly
reviewed here. The field u is obtained in [ 161 as follows. First,
thesolutionforthetwo-dimensional
case in [4] isdirectly
generalized t o obtain a solution for the three-dimensional case
of the cylinderilluminated byanobliquelyincident
scalar
plane wave via the usual separation of variables technique. The
to express the three-dimensional
latter procedure allows one
field solution as a product of the transverse or effective twodimensional field solution and an axial (2) traveling wave field
variation. Next, this solution for the problem of scattering
of
an obliquely incident plane wave by a convex cylinder is seen
via reciprocity t o also directly furnish the far-zone field u of
a scalar point source (or spherical wave) radiating in the presence of the same convex cylinder as in Fig. 6. This far-zone
field u is explicitly given by [ 161

;.

the square roots in (15a) and (15b) is chosen. The quantity


u z denotes the incident ray field of the point source. Referring

to Fig. 6, itis seen that ur may be expressed as follows:

- co

e-jkso

(1 6c)

->

SO

in which CO is a known constant (complex) amplitude of the


spherical wave generated by the point source atP.The various
parameters occurring in (15a) and (15b) are defined as

tL = - 2 m ( Q ~ ) [ f ( Q ~ ) ] - l /COS
~ 8';
~ ( Q R=
) 1

(17 4

1%

p g 2 ( ~ Rcos2
)
e' - P , ~ ( Q ~cos2
) ei
&(QR M Q R )
R 1 (QR )R 2 (QR)

1.

(1 7b)
In (17b), the f(&) is approximated by unity for all practical
purposes because f- 1 near the SB where 8' + n/2, whereas
0, EL < 0, irrespective of whether
far from the SB where 8'
f is approximated by unity or not, so that (15a) always properly reduces to the geometricalopticsresult (via (25a) and
--f

for P = PS in the shadow zone. Again, the positive branch of

.ccs8,

Fig. 6. Ray paths associated with problemof scattering of an obliquely


incident plane waveby a smooth convex cylinder. (a) Incident and
P in lit zone. (b) Surface difreflected ray system for field point
fracted ray path for field pointP in shadow zone.

u'(Q1)

for P = PL in the lit zone and

5' = f.:d

(b)

636

IEEE
TRANSACTIONS
ON

ANTENNAS AND PROPAGATION, VOL. AP-28, NO. 5 , SEPTEMBER 1980

(25 b)). Also,

t =

$'

GO reflected ray path in the lit region (which also includesthe


lit portion of the SB transition region) as in Fig. 6(a), and it
propagates along the surface diffracted ray path in the shadow
region (including the shadowed portion
of the SB transition
region) as in Fig. 6(b). The automatic reduction of the uniform GTD result in ( 15a) and ( 15b) to thescalar soft and hard
GTD solutionspresentin
(IO) and(12)asthe
field point
moves exterior to the SB transition region is easilyverified
by noting that sr % p1,zr and S"
pz" in the far zone, and
that the F and 8 S . h functions take the following limitingvalues
[131, [41:

dl'

X L = 2kLL cos2 8'


[ l - F ( o ) ] "0,
kLd(Ed)'

P=

2m(Ql)W22>

for u B 0

(which is true in the lit


and shadow
zones
exterior to the SB transition region)
(254

Thequantities, L L , L", d d q ( Q l ) / d q ( Q 2 ) ,
occurring in the above equations are given by
LL = I,

Ld =so;

zone exterior t o t h e SB
transition region)

[;+ p,oss']
2sin2

Plr=

eo

-l

P z r = 1,
p2" = so

+ t.

The p7 in (21) represents the principal surface radiusof curvathe circumferential direction at QR.The functions F
ture
and Ps,h are defined as

.
i
n

for 6 % 0

F(6) = 2jfiei6

drePiT2,

6 > 0 (for 6

<0

see end of this section)

1,

(which is true in the shadow zone


exterior to
the
SB transition
region).
(25~)

The N in the summation of (25c) is identical to theN in (12c).


Furthermore, the limit of the result in ( 1 5a) as the field point
approaches SB from the lit side is identical to the limit of the
result in (15b) as the field pointapproaches SB fromthe
shadow side; thus the total field is continuous across SB. I t is
noted that, as the field point approaches SB,

soft case
3

hard case.

The Fock type Airy functions V ( T )and W 2 ( 7 ) are defined in


[4]and
[ 131,andthe
Miller type Airy fynction Ai(7) =
V ( T ) /is~as in [ 131. Tabulated values of Ps,h are available
in [ 131 ; also, the Fresnel integr? present in F is well tab-dated
[ 131, [ 151. Plots of F ( 6 ) and Ps,h((l) are illustrated in [ 4 ] . It
is noted that the result in (15a) and (15b) is valid outside the
paraxial regions of the cylinder, i.e., for 00 f 0, n (see Fig. 6);
the field within the paraxial regions must be obtained via separate considerations which are notdiscussed in this paper.
The scattered field in (15a) and (15b) propagates along the

(25d)

foru-+O(u=OonSB).

Also, the limiting value 0: the field at the SB is more easily


evaluated if one defines PS,12(6)in terms of the related functionsp*(6) and q * ( 6 ) as in [4] :

(notethat

= 0 at SB). Fromtheabovelimitingforms

of

F(u) and Ps,h(6), one notes that the F ( u ) term in (15a) and
( 15b) plays a dominant role in the immediate neighborhoodof

SB, and it is entirely responsible for ensuring the continuity of

637

PATHAK e t al.: DIFFRACTION O F ELECTROMAGNETIC


WAVES

the total field at SB. On the other hand, the

i,,h@)

term in

( 15a) and (15b) plays a dominant role as the field point moves
far from SB (where [ 1 - F ( a ) ] + 0), and it is therefore en-

tirely responsible for reducing (15a) and (15b) uniformly to


the GTD solution for field points exterior to the SB transition
region.
Next, it is observed that the GTD solution of Section 11,
which is a first-order asymptotic solution to terms in inverse
powers of kpg, is valid for cylindrical, spherical, or any other
smooth convex shape. Thus it is valid for torsional surface rays
in that effects of torsion do not occur explicitly to the f i i t
orderforthescattering
problemconsideredhere?
Inaddition, it is observed in the GTD solution that a simple relationship exists between the vector electr_omagnetic and the scalar
acoustic problems; namely, R and T of the GTD solution .in
(5a) and (5b) and (12a) and (12b) area simple combination of
the corresponding scalar or acoustic (soft and hard) functions
Rs,h, and Ts,h for the lit and shadow regions, respectively. As
mentioned previously, these aspects of the GTD solution including its validity for torsionalsurface rays have been verified
via the rigorous asymptotic solutionsto several canonical problems [ 3 ] , [ 121. Based on these observations and the important
observation that the uniform GTD result in (15a) and (15b)
for the far-zone field of a point source radiating in the presence of an acoustic soft or hard convex cylinder is already in
a ray format, it is reasonable to conjecture the following in
regard to the uniformGTD solution for the moregeneral case.
a) The uniform GTD solution for the general electromagnetic problem of the scattering of aray optical electromagnetic field incident on a smooth perfectly conducting arbitrary
convex surface can also be expressed in exactly the same ray
type format as the ordinary GTD solution of Section I1 (see
(10) and (12a)).
b) The dyadics E and 7 in (5a) and (5b) and (12a) and
(12b) of the oldinary-GTD solution must be replaced by the
more general R and dyadics-in the uniform GTD solution.
Of course, one requires that -g +-R in the lit region outside
the SB transition region and.T + T in the shadow region outside the SB transition region. Furthermore,thefunctional
forms of R and f are assumed to be the same for cylindrical,
spherical, or any other convex shape, as is true of thei? and T
dyads in theGTD solution for theelectromagnetic case.
c) The uniform GTD solution for the electromagnetic case
may also be simply expressed in terms of the corresponding
scalar (or acoustic) soft and hard cases, respectively, exactly as
in the GTD solution [see (5a) and (12b)l.
Thus the form of the uniform GTD solution for the total
electric field E may be expressed via the conjectures in a) and
b) above as

"

for PL in the lit region

"

The incident ray field E' is the same a s j n (2) pr (10). From
conjecture c) above, one mayexpress R and T in terms of
their corresponding acoustic soft and hard functions as5

the

for

lit region

(29)

and

+ is,(9)

',shadow
the
for
WQ2)

region.

(30)

The various parameters occurring in (29) and (30)are already


defined in (17a)-( 17g).
If one specializes (26a) and (26b) for the electromagnetic
case to the far zone(i.e., sr 9 p 1 , 2 ~and sd 3 p z d ) and also to
an electromagnetic spherical wave (or point source) type illumination, then the distance parameters f L and Ld required in
(17f) and (1 7g) are given for this special case by L L = 1 and
Ld = SO as in (18) and (19) for the scalar point source illumination case. For thegeneral problem of finding the field in the
near zone of anarbitrary convexsurfaceilluminated
by an
arbitrary ray optical electromagnetic field, the LL and L d are
different from those in (18) and (19). In fact, one only needs
to findthese distance parametersinorder to completethe
solction to this generalproblemasexpressedin
(26a)and
(26b) (together with (27)-(30)). Recalling that the role of the
functions F ( X L ) and F ( x d ) is to ensure the continuity of the
total field at SB, one may then evaluate L L and Ld appearing
in X L and X d , respectively, by actually enforcing the continuity of the total field at SB. This procedure is exactly analogous
tothat employed previously by KouyoumjianandPathak
[ 1 1 ] in their development of a uniform GTD solution for edge
diffraction. Let P S B denote a field point on SB. The continuity of the totalfield at SB requires that

(26a)
in which E ( P L ) and E(P,) are given by (26a) and (26b). Employing the limiting form of (25d) for theF function, and the
definition given in (25e) for the i , , h function, into (29) and

the
for P, in shadow

region.

(26b)

4 In this scattering problem, the source and field points are not too
close to the surface. If the source and/or the field points lie in the immediate vicinity of the surface, then surface ray torsionexplicitly
modifies the surface ray fields [20], [ 211.

5 The canonical cylinder solution for the electromagnetic case which


can be constructed from the acoustic soft and hard solutions (pertaining to the axial electric and magnetic fields) mentioned earlier show the
presence of cross termsinthe dyadic
of (27). However, these cross
terms vanish on the SB and in the deep lit region; they are generally
negligible elsewhere and are thus neglected.

638

IEEE TRANSACTIONS ON ANTENNAS ANDPROPAGATION, VOL. AP-28, NO. 5 , SEPTEMBER 1980

(30), allows one to write (31) as follows:

Alternatively, away from SB, only the value of s and p2' in


of (36) is replaced by sd and pzd for field points in the
shadow zone, with everything else in that equation still being
evaluated at SB; likewise in the lit zone, the distances so and s
in LL of (36) are replaced by the distances 1 from the reference point at FO to QR and sr, respectively. It is noted that
besides the large parameter rn, the kLLld must also be large6
in the asymptotic development of the present uniform GTD
solution. This completes the constructionof the uniform GTD
solution in (26a)and (26b) for a diverging wavefront.
If the incident wavefront is of the conver&g, or converging-diverging type,thenthe
parameters LLpd in (36) can
become negative. It has not been investigated in detail how the
general solution can be completed when LLBd becomes negative. However, if one of the principal directions of the incident
wavefront coincides with one of the principal planes of the
With the distance from Q1 t o PSB defied ass, oneobtains
surface at grazing, then one can treat a converging or converging-diver
ing type wavefront, for which LLrd < 0, by res' ISB = s
(334
placing F ( X f,d ) with F*( I XLtdI). Note that the asterisk
d
s ISB =s.
(33b) on F*( I X L s d I ) denotes the complex conjugate operator. The
use of F*( I XLld I ) when L L J d < 0 leads to a continuous total
It is noted that if PL + P S B , then 0' + ~ / 2EL, + ?, QR + field at SB in this case, and its use may be justified as in the
Q l , p l r + O ( s e e ( 7 a ) ) , & ~ ' + A l , & ~ l ' + A l , a n d ~ l ~ b l . L i kedge
e - diffraction problem [ 171 via ananalyticcontinuation
yise, i.f Ps + PSB, then EA + 0, Q2 + Q , t + 0 , ; ~ + h1, and procedure to include negative values of XLJd(or L L p d ) while
b2 + b l . Furthermore, at SB, p 2 r = p2d: The functionsp * ( S ) simultaneously satisfying the radiation condition for the scatand q * ( 6 ) are continuous everywhere including 6 = 0. Let the tered field.
incident field be described by a diverging wavefrontin the
Finally, exterior t o the SB transition region, the uniform
direction of propagation at Ql (i.e., at grazing); thus, one may GTD result of (26a) and (26b) does indeed recover the ordiwrite
nary GTD results of (10) and (12a) as may be verified by employing the limiting forms of (25a)-(25c) into (29)and (30).
It is noted that a solution for the SB transition region pertaining to the same general problem as the oneconsidered here
has also been given in [ 181 ; however, that solution is not uniform in that it does not properly reduce to the GTD solution
(34)
exterior to the SB transition region. Furthermore, that solution employs a "pseudo" ray path for thereflected field in the
where SO denotes the distance si from some reference point at lit region whichdoesnot
satisfy the generalized Fermat's
?o to thAepoint Ql. The quantities al(70) and ~ l l ( 7 0 )which principle, and secondly it does notpredict the correctpolarizaare the b 1. and 1 directed amplitudes of E' at the reference tionforthe
reflectedfield, especially in the off principal
location ( 7 0 ) are assumed known. It follows from (34) that
planes and in the deep lit
region.

Ld

IV. DISCUSSION AND NUMERICAL RESULTS

The incident wave caustic distances p l , 2 ' are measured from


the reference point at ' 0 to the respective causticlocations.
Incorporating (34) and(35) into (32)yields

The value o f p z r at SB equals p b i as pointed out in (7b).


The distance parameter is a slowly varying quantity near
SB, and ( 1 - F ) in (29) and (30) vanishes sufficiently rapidly
as the field point moves far from SB; it is therefore convenient
to use the form in (36) for
and L d even away from SB.

t!.'

A uniform GTD solution hasbeen obtained for the problem


of the scattering of a ray optical electromagnetic field by a
smooth perfectly conducting convex surface as shown in Fig.
1. This result is explicitly given in (26a) for thelit region and
in (26b) for the shadow region, together with the parameters
defined in (17a)-( 17g), (27)-(301, and also in (36). While the
behavior across only a single shadow boundary (SB) is discussed in this paper in connection with the open convex surface of Fig. 1, the present theory can just as easily treat the
scattering by a closed convex surface. Basically, one treats a
closed convex surface via theuniform GTD inthe same
manner as one would via the ordinary GTD._The ogly differi? and T dyads in
ence between the two approaches is that the
the GTD solution_are replaced by the more general and more
accurate R and f dyads in the uniform GTD solution. It is
noted of course that the usual GTD solution fails within the
In general, the present solution appears to be accurateeven for
kLLld as small as three; in some special cases k L L d can be made as

small as unity.

639

PATHAK et al.: DIFFRACTION OF ELECTROMAGNETIC WAVES

(a)

(b)

Fig. 7. Comparison between patterns calculated via uniform GTD (UTD) and ordinary GTD solutions for radiation by
electric (a) and magnetic (b) dipoles parallel to axis of a perfectly conducting circular cylinderwith Q = l h and p' = 2X.
shadow boundary transition regions whereas the uniform GTD
solution does not. When a surface of revolution is illuminated
by a plane wave which is incident along the axis of revolution
of the surface, a caustic of the surface diffracted rays lies on
this axis. The uniform GTD solution fails (as does the usual
GTD solution) in the neighborhood of such a caustic. However, if the field point is in the near zone such that the
caustic
and the shadow boundarydirectionsare
widely separated,
then one can employ the methodof equivalent ring currents to
evaluate the field in the neighborhood of such a caustic. The
equivalent currents in this case are found indirectly from the
uniform GTD solution. Such a procedure will be reported in a
separate paper together with another approximate technique
which would yield the field forthe special case when the
caustic lies on the shadow boundary. The other minor restrictions on the uniform GTD solution are mentioned at the end
of Sections I and 111.
It is interestingto note that theE;" of (1 7a) may be approximated for convenience by -2m(&) cos
0' (upon arbitrarily
setting f-'I3(&)
= 1) without affecting the accuracy of the
solution. Furthermore, it is easily verified that the f . L , d parameter of (36) simplifies to the following for cylindrical or
spherical type wave illumination:

In the case of plane wave illumination, I

+ 00 and so +
so
that LL = s' and L d = sd. Using LL'd as in (37) for plane,
cylindrical, o r spherical wave illumination, it is easily verified
that the uniform GTD result in (26a) and (26b) satisfies reciprocity. Finally,thisuniformGTD
solution is simple y d
accurate to use since it is given in terms of the F and Ps,h
type functions which are tabulated. These important aspects

of the present uniform GTD solution are illustrated below by


applying it t o several interesting problems.
Consider an infinitely long circular cylinder of radius a as
shown in Fig. 7. The ordinary GTD solution for this geometry
is shown in Fig. 7(a) foran electric dipole(acousticsoft
boundary condition) and in Fig. 7(b) for amagnetic dipole
(acoustichard boundary condition) mounted
parallel to the
cylinder axis. As is well known, theGTD solution breaks down
near the shadow boundaries. This is readily apparent in these
two figures. Nevertheless, this solution has often been applied
to obtain a result for the scattering from a cylinder. The uniform GTD solution presented in this paper, however, can be
quickly and accurately computed. The uniform GTD or UTD
result iscomparedwiththeordinary
GTD resultin Fig. 7.
The validity of this result is shown by comparing it with the
exact eigenfunction solution as shown in Fig. 8. The agreement is excellent and hence confirms the validity of the uniform GTD (UTD) solution.
An example of a magnetic dipole in the presence of an elliptic cylinder is considered next, asillustrated in Fig. 9(a).
The pattern is a conic cut about the cylinder axis. The dipole
is parallel to the cylinderaxis. This could represent a slot
mounted near an aircraft fuselage, engine, or store. The calculatedresult iscompared against aresult obtainedfrom a
moment method solution7 in Fig. 9(b). The agreement is very
good, again verifying the validity of the present solution.
In order to validate this solution in terms of a more complex surface,consider the circular cone configuration illustrated in Fig. 10.A half-wave dipole is located in the near zone
of the cone, and thenear-zone field is measured as the receiver
moves azimuthly around the axis of the cone. Both a vertical
and horizontal dipole are treated with the
resulting patterns
shown in Fig. 1 I(a) and (b), respectively. In each case the
calculated and measured results are in good agreement. Note
thatthe receiver polarization was aligned withthat of the
transmitter in bothcases.
7 This solution has been kindly

furnished by Dr. Nan Wang.

640

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-28, NO. 5 , SEPTEMBER 1980

(a)
(b)
Fig. 8. Comparison of UTD solution of Fig. 7 with exact modal series solution.

Fig. 9.

Radiation pattern of a magnetic dipole located parallel to axis


of an elliptic cylinder.

The last example is used to show a cylinder being illuminatedby a complex wavefront that can berepresentedin
terms of an astigmatic tube of rays. The source of the astigpencil) which impinges
matic tube of rays (or the quadratic ray
on the cylinder is an edge diffracted field of the plate mounted
on the cylinder such that the
cylinder is not in the shadow
boundary transition region of the edge diffracted fields. The
geometry used is illustrated in Fig. 12. The antenna is a slot
mounted parallel to the cylinder axis in the centerof the plate.
The calculated and measured results for this configuration are

shown in Fig. 12. There is good agreement between measured


and calculated results, thus confirming the validity of the uniform GTD solution for a cylinder illuminated by a more general wavefront.
Although only numericalresults for developablesurfaces
have been illustrated here, the present UTD solution is valid
for an arbitrary smooth convex surface. Examples illustrating
the usefulness of this solution for anondevelopableconvex
surfacesuchasaprolatespheroid
will bereportedinthe
future.

641

PATHAK ef al.: DIFFRACTION O F ELECTROMAGNETIC WAVES


Im B

FIELD P O I Y T

18.15X

ItiiSIDE 0 < +<

Fig. 13. Location of

ps, pp, and SDP for shadowregion analysis [ 4 ] .

Fig. 10.Geometryforcircularcone.

36

72

+-ANGLE IDEGREESI
108 144
180 216
252

288

324

b) Fig. 5(b) and (c)in [4] are plots of thefunctions


e+i(n/4)p(() and e f ( n / 4 ) q ( t ) , respectively,instead of their
complex conjugates as erroneously indicated.
c) Fig. 3 in [4] should be replaced by Fig. 13 in this paper.

360

0-

0-

REFERENCES

-20

(a)
d - A N G L E IDEGREES)
44

252

216

180

288

324

360

151

161

(b)
Fig. 11. Comparison of measured (dashed curve) and calculated (solid
curve) radiation patterns of an electric dipole mounted near a circu10). (a)Verticalpolarization.
(b) Horizontal
larcone(seeFig.
polarization.

11 11

I
\

MEASURED
-_--CALCULATED

1121

1131

ANGLE ( DEGREES )

Fig. 12.Comparisonofmeasuredandcalculatedradiationpatterns
for a slot mounted in a plate-cylinder configurationwith slot parallel
to cylinder axis.

APPENDIX

This appendix serves t o correct a few minor typographical


errorsappearing in [ 4 J on which the ansatz of the present
UTD solution is based. It is noted that these errors do not in
any way change the analytical development or the final results
given in [ 41 . The corrections tothese errors are as follows.
a) The factor 2 inside the square root sign present in [4,
eq. (20) and (55b)l shouldactually beplace!
outsidethe
square root; thus the corrected definition for
Ps,h in [4,eq.
(20) and (55b)l should be as given in (25e) of this paper.

1191

J . B.Keller,Geometricaltheoryofdiffraction,
J . Opt.SOC.
Am., vol. 52,pp. 116-130, 1962.
-,
Diffraction by a convex cylinder, IRE Trans. Antennas
Propagat., vol. AP-24, pp. 312-321, 1956.
B. R. Levyand J. B. Keller, Diffraction by a smooth object,
Comm. Pure and Appl. Math., vol. 12, pp. 159-209;1959.
P. H. Pathak, An asymptotic analysis of the scattering of plane
waves by asmoothconvexcylinder,
RadioSci., vol.14,pp.
4 1 9 4 3 5 , May-June 1979.
P.H.Pathak,R.J.Marhefka,andW.D.Burnside,High
frequency scattering by curved surfaces, ElectroSci. Lab., Ohio
State Univ., Columbus, Tech. Rep. 3390-5, June 1974.
V. A . Fock,Diffraction,refractionandreflection
of waves:
Thirteen papers, Air Force Cambridge Res. Center Rep. AFCRCTN-57-102, (AD 1 17276), 1957, also Electromagnetic Diffracrion
andPropagationProblems.
New York: Pergamon,1965.
R. K. Luneberg, MarhematicalTheory OfOptics. Providence, R1:
Brown Univ., 1964.
IM. Kline,
Anasymptoticsolution
of Maxwellsequation,
Comm. Pure and Appl. Math., vol. 4, pp. 225-263, 1951.
S. W. Lee,GeometricalOptics,Electromagnet.Lab.,Univ.
Illinois, Tech. Rep. 78-2, 1978.
R . G. Kouyoumjian, The geometrical theory of diffraction and its
application, in Numerical and Asymproric Techniques in ElectroMagnetics, R. Mittra,Ed.Berlin:Springer-Verlag,1975,ch.6.
R . G. KouyoumjianandP.H.Pathak:
A uniformgeometrical
theory of diffraction for an edge
in a perfectly-conducting surface,
Proc. IEEE. vol.62,pp.1448-1461,1974.
D. R. Voltmer, Diffraction
by doubly curved convex surfaces,
Ph.D. dissertation, Ohio State Univ., Columbus, 1970.
I,
N.A.Logan,Generalresearch
in diffractiontheory,vol.
LMSD-288087;andvol.
11. LMSD-288088,MissilesandSpace
Div., Lockheed Aircraft Corp., 1959.
N. A. Logan and K. S . Yee, Electromagneric Waves, R. E. Langer,
Ed.Madison,WI:Univ.Wisconsin,1962.
I. A.Stegun, Handbook of Mathematical
M.Abramowitzand
Fzmctions. Washington, DC: Nat. Bur. Stand., 1970.
A uniform
P. H . Pathak, W . D. Burnside,andR.J.Marhefka,
GTD analysisof the scattering of electromagneticwaves by a
smoothconvexsurface,ElectroSci.Lab.,OhioStateUniv.,
Columbus, Tech. Rep. 784583-4, 1978.
J . D.
Cashman,
R.
G . Kouyoumjian,
and
P.
H. Pathak,
for an Edge in a
Comments on A Uniform Theory of Diffraction
perfectly Conducting Surface, IEEE Trans. Antennas Propagat.,
vol. AP-25, pp. 447-451, May 1977.
G . L. James, GeometricalTheory
of DIffactionforElectromagneric Maves. Henfordshire,England:PeterPeregrinusLtd.,
1979.
Y.Rahmat-SamiiandR.Mittra,Spectralanalysis
of highfrequency diffraction of an arbitrary incident fieldby a half-planeComparison with four asymptotic techniques, Radio Sei., vol. 13,
pp. 31-48, Jan.-Feb.1978.
P. H . Pathak and N. N. Wang, An analysis of the mutual coupling
between antennas on a smooth convex surface, ElectroSci. Lab.,

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-28, NO. 5, SEPTEMBER 1980

642

Dept. Elect.Eng.,
Columbus, Ohio State Univ., Final Rep.
784583-7, Oct. 1978; also, IEEE T r a m . Anrennas Propagar.,
submitted for publication.
[21] P. H. Pathak, N. N. Wang, W. D. Burnside, and R. G.
Kouyoumjian, A uniform GTD solution for the radiation from
sources on a convex surface, IEEE Trans. Antennas Propagar..
submitted for publication.

problems of diffraction by discontinuities in the geometrical as well as in


the electrical properties of a surface; the latter category involved surface
wave structures. He has also worked on the analysis of reflector and
corrugated horn antennas. Recently, he has co-authored a chapter on the
subject of high-frequency diffraction for two books.
Dr. Pathak is a member of Sigma Xi and an associate member of the
U.S. Commission B of URSI.

Prabhakar H. Pathak (76)


was
born
in
Poona, India, on March 21, 1942. He received
the B.Sc. degree in physics from the University
of Bombay, India, in 1962, the B.S. degree in
electrical engineering from Louisiana State
University, Baton Rouge, in 1965, and the M.S.
and Ph.D. degrees in electrical engineering from
the Ohio State University, Columbus, in 1970
and 1973, respectively.
From 1965 to 1966 he was an instructor in the
Department of Electrical Engineering at the
University of Mississippi, Oxford. He spent the summer of 1966 as an
Electronics Engineer at the Boeing Co., in Renton, WA. Since 1968 he
has been with the Ohio State University ElectroScience Laboratory,
initially as a student employee and later as a member of their full time
research staff after graduating in 1973. Hiswork at this laboratory has
primarily dealt with the development of uniform asymptotic solutions
which improve and extend the geometrical theory of diffraction solutions
for solving antenna and scattering problems associated with complex
structures, such as aircraft, spacecraft, and ships. He has analyzed

Walter D. Burnside (S6%M72), for a biography and photograph please


see page 327 of the May 1980 issue of this TRANSACTIONS.

Ronald J. Marhefka (S67-76) was born


in
Cleveland, OH, on June 2, 1947. He received the
B.S.E.E. degree from Ohio University, Athens,
in 1969, andthe M.S. and Ph.D. degrees in
electrical engineering from the Ohio State
University, Columbus, in 1971and 1976, respectively.
Since 1969 he has been with the Ohio State
University ElectroScience Laboratory. His research interest is in the area of applications of
diffraction techniques to antenna and scattering
problems, including airborne and shipboard antenna analyses.
Dr. Marhefka is a member of Tau Beta Pi, Eta Kappa Nu, Phi Kappa
Phi, and Sigma Xi.

You might also like