RADIO SCIENCE, Vol. 2 (New Series), No.

3, March 1967

Propagation of Electromagnetic Waves Near a

Coastline on a Flat Earth 1
R. F. Millar 2

Laboratory of Electromagnetic Theory, Technical University of Denmark, Lyngby, Denmark

(Received August 29, 1966; revised October 28, 1966)

A magnetic line source is placed parallel to and possibly near a land-sea boundary on a flat earth.
The problem is two dimensional, but its solution may be employed to examine the corresponding
three-dimensional configuration which involves a vertical electric dipole. The mathematical model
consists of a perfectly conducting half plane (sea) in the interface between free space and an imper·
fectly conducting dielectric (land). The far field, over both sea and land, is required.
The method employed is to evaluate the rigorous, contour integral solution of Clemmow (1953),
under the assumption that the complex refractive index n of the land is large. This integral is evalu-
ated first for points not too near, but not necessarily far from the coastline. A second field representa-
tion (useful very near the shoreline) as a Bessel-function series is derived.
Analytical comparison is made with recent work of Wait (1963) and Bach Andersen (1963). Good
agreement is found except in the immediate neighborhood of the coastline. The onset of the "recovery
effect" over the sea and the associated rapid phase variations near the coastline, are found for a wave
incident from over the land; the corresponding effects for a wave incident from over the sea are also
noted. These are in qualitative agreement with experiment. Numerical results are exhibited for a
frequency of 10 MHz and n=3.91-i 2.30.

1 . Introduction

Phenomena associated with radio wave propagation across a land-sea boundary have attracted
considerable attention, both theoretical and experimental in nature. However, until recently,
little or no investigation had been made of field behavior in the immediate neighborhood of the
coastline. It is with this aspect of the problem that the present work is chiefly concerned.
With exception of the analysis of Clemmow (1953) (on which the present work is based),
possibly the first relevant theoretical study is that of Wait (1957). He examined the vertical
electric field on both land and sea (including the neighborhood of the coastline) arising from a
distant vertical electric dipole placed on the land. Wait employed a variant of the so-called
"compensation theorem" (Monteath, 1951) in his analysis. The properties of the land and sea
were characterized by an impedance boundary condition, discontinuous at the straight coastline,
which gave rise to a logarithmic singularity in the field at the discontinuity. In three more recent
papers (Wait, 1963; Wait and Jackson, 1963; Wait, 1964), more refined analysis was employed
to study the propagation of a ground wave across various types of coastline. Once again the
transmitter was on the surface of the land, and an impedance boundary condition was assumed
while the coastline was represented by a transition region; explicit consideration was given to
both an abrupt and a linear impedance transition. In a later paper (Wait and Spies, 1964), a
more smoothly varying surface impedance function (which yielded the abrupt discontinuity as a
special case) was employed.
The present analysis was undertaken as a consequence of certain results obtained by Bach
Andersen (1963), who calculated the sky-wave field over the sea from a vertical electric dipole
placed on the land. The land-sea model chosen by Bach Andersen (the same as that selected for
the present analysis) was a perfectly conducting half plane in the interface between two different
dielectrics. Curvature of the earth's surface was neglected. Together with the compensation
theorem, he employed certain approximations, and his results are not valid when the transmitter
is close to the boundary.
1
This research was supported by a grant from the Danish Government Fund for Scientific and Industrial Research (Staten• teknisk-vidensltabelige Fond) to
the Laboratory of Electromagnetic Theory.
'Present address: Division of Radio and Electrical Engineering. National Research Council~ Ottawa. Canada.

261
262 R. F. Millar

The primary purpose of the present work is to attempt to relate the results of Bach Andersen
to those of an exact analysis based on the same model, and to predict the far field over the sea when
a vertical dipole is placed on land close to the coastline. A further object is to examine the extent
of validity of the surface-impedance concept to characterize the surface properties in the neigh-
borhood of the discontinuity.
The following work is based on that of Clemmow (1953), who considered the diffraction of an
H-polarized plane wave by the half plane in the interface as an intermediate step in his study of
diffraction of a line-source field. The former problem involved a pair of dual integral equations,
the solution of which required the factorization (in the Wiener-Hopf sense) of a function closely
related to the Fresnel reflection coefficient of the earth medium. However, for the positions in the
interface of source and field points conside<ed by Clemmow, it was unnecessary to determine the
explicit factors; these involve complete and incomplete elliptic integrals of the first and second
kinds (Stockel, 1963). Extensive use of asymptotic methods was made in the reduction of the
solution to a manageable form and, as a consequence, Clemmow's results are quantitatively signifi-
cant only when both source and field points are sufficiently remote from the diffracting edge.
Although many experimental investigations of radio-wave propagation across actual mixed
land-sea paths have been undertaken (Millington, 1949; Millington and Isted, 1950; Pressey,
Ashwell, and Fowler, 1953; Pressey and Ashwell, 1956), it is only in recent times that model experi-
ments have been performed. Maley and Ottesen (1964) and King and Maley (1965, 1966) have
reported the results of experiments at microwave frequencies in which tap water simulated the
imperfectly conducting land, while aluminum plates on the water surface represented the highly
conducting sea; it is an idealization of this model which was analyzed by Clemmow. Good agree-
ment between experimental data and calculations based on Wait's theory (or modifications thereof)
was found except in the immediate vicinity of the land-sea boundary.
In the present work, the model selected for study consists of a perfectly conducting half plane
(the sea) in the interface between two different dielectric media (land and free space). Although
the actual physical configuration is three dimensional, the corresponding two-dimensional problem
(in which the source is a magnetic line current parallel to the edge of the half plane) will be ex-
amined. It is required to find the field far from the edge due to such a line source which may lie
close to the edge. As a result of the reciprocity principle regarding the interchange of source and
field points, it suffices to examine the field (again, possibly near the edge) when an H-polarized
plane wave is incident on the interface.
The work to be described rests on the hypothesis that the complex refractive index n (4) of
the imperfectly conducting medium is large. Proceeding on this assumption, it is possible to
obtain apparently reasonable (and analytically simple) approximations to the required factors.
To the same order of approximation, it is possible to evaluate Clemmow's contour integral for
points of observation on the surface of the land or sea, and to obtain comparatively simple expres-
sions for the field, provided that the point of observation is sufficiently remote from the disconti-
nuity; the larger the value of n, the closer to the discontinuity the point of observation may come.
To study the field behavior very close to the discontinuity, the contour integral is expressed as a
series of Bessel functions. Here the chief difficulty lies in the determination of the coefficients;
these involve infinite integrals, the integrands of which contain the (only approximately known)
Wiener-Hopf factors. Nevertheless, the two field representations apparently suffice to predict
the field variation everywhere on the surface of land and sea, and enable comparisons to be made
with the results of Wait (1963) and Bach Andersen (1963).
The laboratory report 3 in which the present work is described in more detail, and in which
there was reported disagreement with a result of Wait (1963), has been criticized by Wait (1965).
The reason for the lack of agreement was three-fold. First, an algebraic mistake in the report

3 '"The far field of an H-polarized line source near the edge of a perfectly col!ducting half plane in the interface between two different media. with application to

mixed-path propagation ... Laboratory of Electromagnetic Theory. Technical University of Denmark~ Lyngby. Denmark. Repon R 43~ Jan. 1965. This work was also
described at the Symposium on Electromagnetic Wave Theory. Delft~ The Netherlands. Sept. 6-11. 1965.
Electromagnetic Waves Near a Coastline 263

introduced an erroneous factor of n 112 into the field. Second, a normalized magnetic field was
compared with Wait's normalized electric field. (Such comparison sometimes may be justified,
but not when one refers to the field reflected back towards the transmitter by the coastline.) Third,
the inequality (- a 1) ~ 1 in Wait (1963) was interpreted to be valid as -a~--'; oo. The fact that
Wait assumed the numerical distance from the boundary to the observation point to be small, was
overlooked in the report. When these points are taken into account, it is found that agreement
obtains (see sec. 6.1).

2. Formulation of the Problem

Let, the plane y= 0 in a rectangular Cartesian coordinate system Oxyz coincide with the inter-
face between free space (y > 0) and an imperfectly conducting dielectric (y :o:;;; 0), which will repre·
sent the earth. The half plane x ~ 0, y= 0, - oo < z < oo is assumed to be perfectly conducting,
and corresponds to the surface of the sea. A time-harmonic, H-polarized plane wave is incident
on the interface from y > 0. The propagation vector is normal to the z axis, and makes an angle
a with the positive x axis (see fig. 1). Thus the problem is two dimensional and scalar, the only
unknown being the z component of magnetic field, Hz. A time-dependence factor eiwt will be sup·
pressed. If r, 0 are the polar coordinates of a point in y > 0 (0= 0 coinciding with the positive x
axis) then, in accordance with (83) of Clemmow (1953), the total magnetic field is

(l)

i cos j a cos ! f3 e -ikr cos(IJ-Pl

in which (2)
Hs TrL 1 (cos a) fc L1 (cos f3)(cos /3 +cos a) tf3,

. n 2 sin a- (n2 - cos 2 a) 1f2

and p(sm a)= 2
n sin a+ (n 2 - cos 2 a)112
· (3)

Here k is the wave number in y > 0, p denotes the Fresnel reflection coefficient of the earth, and n
is its complex refractive index:

n={E'/E icr'/(we)}ltz. (4)

The wave number in the imperfectly conducting earth is k'(= nk). The square roots in (3) are real
and positive when their arguments are real and positive. The quantities € ', cr' are, respectively,
the permittivity and conductivity of the earth medium, while E denotes the permittivity of free space.
(Rationalized M.K.S. units are employed.) Then arg n lies between 0 and -i?T. It is assumed

FIGURE I. Model for examination of mixed-path problem.

Perfectly conducting half plane y= 0, x ;;. 0, representing sea, lies in interface
between free space {:r > 0) and imperfectly conducting dielectric eanh (r < 0).
H·polarized plane wave is incident at angle "· (r,l/) are polar coordiMtes of point
of observation. E'. €1' denote permittivity and conductivity of earth.
264 R. F. Millar

that the magnetic permeability of the earth is the same as that of free space; C is a contour in the
complex ~ plane, illustrated in figure 2.
The function Lt is defined implicitly by (80) in Clemmow's paper:

(5)

The complex variables A, ~ are related by A= cos ~ which maps the strip 0 ~ Re ~ ~ 1T into the
A plane cut on the real axis from 1 to oo and from 1 to oo. The function Ut(A)Lt(A) is analytic and
single-valued in the A plane cut from - 1 and - n to - oo in the upper half plane and from 1 and n
to oo in the lower half plane. The contour C transforms into Ct, as shown in figure 3.
The function L 1 is an L-function and U1 a U-function in the sense of Clemmow; that is, Lt is free
of singularities and zeros throughout the region below C 1 and of algebraic growth at infinity therein,
while U1 has similar properties above Ct. Moreover, because U1 (A)L 1(A) is an even function of A,
by proper normalization one can write

{6)

a relation of subsequent use. Because the pole P of the reflection coefficient p(sin does noy lie m
in 0 ~ Re f3 ~ 1T (see fig. 2), it follows that (1- A2) 112 + n - 1 (1- A2/n 2)112 # 0 on the upper sheet in
the cut A plane. (Here, and henceforth, the upper sheet of the Riemann surface for (1- A2)1' 2
+ n - 1 (1- A2 /n 2)112 is that which correspond§ to 0 ~ Re f3 ~ 1r.)
The next task is to determine an approximation to Lt which is sufficiently simple to permit
evaluation of the integral in (2).

c •
p
0 -A. •.
0 rr n~
). plane

X
f3 plane FIGURE 3. Path of integration C1o singularities and branch
cuts, in A (=cos {I) plane.
Pole P lies on lower sheet of Riemann surface, near i. - - L

FIGURE 2. Path of integration C in complex~ plane, with

indentation below pole at ~ = Tr- a.
P denotes pole of reftection coefficient of earth. Branch points at cos {J=-n
are denoted by ctllllses (X).
Electromagnetic Waves Near a Coastline 265

3. An Approximation to L1(A.)

In many applications, the conductivity (<T') of the earth is sufficiently high, or the angular fre-
quency (w) sufficiently low, that lnl Pl. Here it will be assumed that this is the case; however,
as often occurs when asymptotic methods are employed, the results apparently are useful even for
lnl as small as 4 or 5.
In a footnote, Clemmow (1953) has pointed out that (1- A.2)1/2 + 1/n is a valid approximation to
(1- A. 2) 1i 2 + n - 1(1- A. 2/n 2) 1i 2 when In I P 1, and that this function has been factored by Senior (1952).
However, the factors obtained by Senior are too complicated to be of use in the integral (2). Essen-
tially the same function was considered by Bazer and Karp (1962), who also quoted a compact
formula for one of the factors, given earlier by Fock; this, too, is more complicated than is desirable
for present purposes. Of course, the exact factors obtained by Stockel (1%3) lead to an intract-
able integral when inserted into (2). Therefore, it seems necessary to proceed as follows.
Let h(A.) be defined by

(7)

Then, in accordance with (5), it is required to find Ut. L 1 such that

h(A.) = 2U1(A.)L1(A.)/(l + n 2 ). (8)

The function h(A.) is defined, continuous and non zero on the contour Ct. and the determination
of U1 and L1 requires the solution of a homogeneous Hilbert problem (Noble, 1958, ch. 4). 'fhen

L1 (A.)= 2- 112 (1 + n 2 )1i 2 L(A.), (9)

with 1.
log L(A.)=--
27Tl
J"'
-oo
log h(z)
z-
A. dz, lm A.< 0. (10)

Here it may be assumed that the logarithm denotes that branch of the function which vanishes
when its argument is unity; the integral converges because h(z)=1+0(z- 2) as z~±oo and the
total variation of arg h(z) along C1 is zero. The path of integration coincides with C1 (see fig. 3)
but is described in the opposite sense. The integral in (10) defines an analytic function for all A.
below Ci which tends to zero as lXI ~ oo so L1(A.) is an L-function in the sense of Clemmow 1953).

Because of the presence of the logarithm, the integration in (10) is difficult to perform. How-
ever, differentiation of (10) with respect to A., followed by an integration by parts, gives

L'(A.) =--1-f"" h'(z)/h(z) d I ,

L(A.) 27Ti -"' z- A. z, m 1\. < 0 ' (11)

in which a prime implies differentiation; h'(z)/h(z) may be calculated from (7) and is O(z- 3) as
I z I ~oo.
A suitable approximation to L '(A.)/ L(A.) is found in the following way. First, the integration
contour shown in figure 3 is deformed in the upper half plane into a loop enclosing the branch cuts
which run from z=- nand -1 to -oo. The denominator of h'(z)/h(z) contains the factor (n 2 - z2 ) 1i 2
+ n2 (l- z2)112 and the new integration contour may be deformed away from the point z =- 1 so
that the term n 2(l- z2) 1i 2 predominates to any desired degree when n is sufficiently large. Further-
more, in these same circumstances, the loop can be taken to lie in Re 'A~ 0. Denote the resulting
contour by r and the domain outside f by Dr. Then
266 R. F. Millar

2
L'(A.) n -1J z
- -=-- dz (12)
L(A.) 27ri r (n 2 - z 2)(1 - z 2)(z- A.)

if A. lies within Dr. (A closer examination of the approximation suggests that the error thereby
introduced is at most of order 100/l n I percent.) The contour r encloses simple poles at z=- n
and z=-1, and the integral is readily evaluated by the method of residues. Moreover, because
of the behavior of h(z) at infinity, it is easy to see that (10) implies log L(A.) ~ 0 as I A. I~ 00 in Dr.
Thus one finds

L(A.) = (1 + A.)112(n + A.)- 112, A. in Dr (13)

and, according to (9),

(14)

Here, only the dominant term in n has been retained.

The approximation (14) is not valid for points inside the loop f; in particular, it does not
describe the behavior of L 1 (A.) near the branch points at A.=-1 and A.=-n. To obtain an approxi-
mation for L,(A.) which is valid when A. lies within r, one may use (5) and (6). Thus

L 1 (A.)= [(1- p)L 1( - A.)] -1 (15)

for all A..

Let A denote the reflection of r in the origin; then A is a loop containing the branch cuts from
z= n and z= 1 to oo. Denote by DA, the domain outside A; for sufficiently large n, this domain
includes the interior of f. Then if A. is in DA,- A. is in Dr, so (3), (14), and (15) yield

(16)

which, in particular, describes the behavior of L 1(A.) near A.=-n and A.=-l.
The approximation to L1 (.\) given by (14) and (16) has branch points in the upper half plane at
.\ = - 1,- n. Provided that n is sufficiently large, .\ = 0 will lie in both Dr and DA. Thus (14) gives

(17)

while (16) implies L, (0) = 2- 1/2n 1/2(1 + n- 1). (18)

These two values of L1(0) are equal, to order 1/n. Furthermore, L 1(0) may be computed
exactly from (5) and (6):

(19)

Thus, the approximate values agree with the exact value, to order 1/n, at A.= 0. This is a
point where the approximation might be expected to be relatively poor.
Finally, it is observed that the approximation to L 1(A.) vanishes only with n2(1- A. 2)1/2+(n2- A. 2)1/2.
The zero (A.=-n(1+n 2)- 1/ 2) of this function coincides with the pole of the reflection coefficient
p; it does not lie on the upper sheet of the Riemann surface, but is situated close to A.=-1 on the
lower sheet (fig. 3).
Thus it is seen that the approximation embodied in (14) and (16) possesses all the essential
properties if n is large.
Electromagnetic Waves Near a Coastline 267

4. The Field Away From the Coastline

In the present section, the integral in (2) will be evaluated for y= 0 with the assistance of the
approximation to Lt derived in the previous section. (The restriction to a point of observation on
the earth's surface seems to be necessary in order to make further progress.) It will be seen that
useful, relatively simple expressions may be obtained for the field on the surface of the land or
sea, provided that the point of observation is sufficiently remote (the minimum distance being of
the order of a fraction of a free-space wavelength) from the coastline.
The function H • is defined by (2). If attention is confined to points on the surface of the earth,
then y= 0 and, in terms of the integration variable>..,

(20)

Here A.o =cos a, A(a) =(cos i a)/Lt (cos a), and the integration contour is that of figure 3, described
in the opposite sense.
It is necessary to give separate consideration to the cases x < 0 and x > 0; this will be done in
the following subsections.

4.1. The Field on the Land (x < 0)

It is convenient to rotate the branch cuts in the upper half plane through 1/2 1r in the clockwise
r2
sense, and to deform the integration contour in (20) into loops ft and about the cuts, as illustrated
in figure 4. Then the domain within which (14) is not a valid approximation is rotated likewise.
No singu1arities of the integrand are captured in this deformation of the contour of integration and,
because (16) provides a valid approximation to Lt(A.) at all points on the integration contours, one
finds that the integral in (20) may be approximated by 2t12 n(It + I 2),
eiK>.. (n + >..)11 2 d>..
where
J
Iv = r n2(l-A_2)t/2+(n2-A_2)t/2. >..+A.o' v=1, 2,
v
(21)

and K =- kx > 0.
The integral I2 will contain a factor e-inK. Thus, if-Klmn is sufficiently large, I 2 may be
neglected in comparison with It. For example, je-inKj :::o.l if-Kimn;::2.3; this value will be
considered sufficient to justify the neglect of I2. However, for sufficiently small K, I2 will be
comparable in magnitude to It. Nevertheless, when the land conductivity is adequately high, or
the frequency sufficiently low, the neglect of I2 is justified except for points within a small fraction
of a wavelength of the discontinuity. Therefore, it will be assumed that

0 FIGURE 4. Deformed branch cuts in upper and lower half

-.:\. planes.
The approximation (14) is not valid inside the shaded area .

.:1. plane
268 R. F. Millar

Ln (22)
H 8 = - A(a)/r.
1T'

The denominator of the integrand in (21) may be rationalized, and the resulting expression
expanded in partial fractions to yield

. (23)

in which lv = lv(K) is determined by

(24)

(a term in the integrand, single valued in a neighborhood of f 1 and so contributing nothing to the
integral, having been omitted). The coefficients Bv are determined exactly for all n but, because
n _is large, th~ following approximations suffice:

Br = i n/(b- Ao), B2 = i n/(b + A.o), B3 = n/(A.~- b 2), (25)

wherein (26)

and (27)

Because In! is assumed to be large, the factor (n +A.) - 112 in the integrand of lv may he replaced
by n - 112 if K # 0. (This step may he established with more rigor by representing (n +A.) - 112, A. '
on f1, by

inverting the orders of integration, and expanding lv by the method of steepest descent for In! ~ l.)
Then
e iK:>..(l- 1\\ 2)1/2dA.
= -1/2
}v n
i r, A_-Y
#0
o,v
,K • (28)

This expression is valid for complex K if Re K ;;;;.: 0.

Now,

(Jv(K)e -iK'v ]'=in -1/2e -iK'v J (l- A_2)1/2eiK>-dA., (29)

rt
wherein a prime denotes differentiation with respect to K. It is easy to verify that the integral in
(29) is equal to 7r-· 1 [H~2)"(K)+ Hlr>(K)], if K # 0, ReK ;;l!; 0 (see(4l}); here Hb2>is the Hankel function
of the second kind of order zero. Therefore, because }1(-ioo)=O and lv(oo)=O if v=2, 3, one
obtains
(30)

in which * denotes- ioo if v= l, oo if v=2 or 3. Integration of the first term by parts and use of
the identities
Electromagnetic Waves Neor a Coastline 269

(31)

leads to

(32)

provided that - Kimn? 2.3. The summation index v runs between 1 and 3.
It will be noted that this approximation for H s has a logarithmic singularity at K = 0, although
the integral / 1 remains finite. However, there is no disagreement because (32) is not valid in a
neighborhood of K= 0. (Moreover, the term H~l(K) will be shown to be extraneous and of no
significance.)

Provided that K > 1, each Hankel function in the integrands of (32) may be represented with
sufficient accuracy bythe leading term in its asymptotic expansion; the approximation, of course,
improves with increasing K. The integrals which result may be expressed in terms of the complex
Fresnel integrai F, defined by

F(z} = i"' e-i>..•d>... (33)

It is found that

(34)

in which

(35)

Then for K ;:: 1, (32) becomes

H•""" :,2 A{a) {H3>{K)- 23/27T- 112e-t/wi ~ B.,(1- "-) (1 + ,,..)112eiKC., F(Z.,)}. (36)
n

It may be observed that Z 1 = (2K)ll2 sin ! a 8 where, as defined by (16) of Clemmow (1953),
sin aa (1 + n2) -t/2 and aa is the Brewster angle for the imperfect conductor. Because In I is
assumed to be large, Zt = (K/2)112 sin aB (= Yot in the notation of Clemmow).__ Thus,- iZ'i approxi-
mates to the numerical distance of the point of observation from the boundary. It will be seen later
(91} that the terms in (36) which involve Zt and Z2 account for the ground-wave attenuation experi-
enced by the field in passing over the imperfect conductor. The major dependence on_ the angle of
incidence a is provided by the term which involves Z 3 ; this term vanishes both for a=O and for
a=TT.

One might be tempted to interpret the Hankel function in (36) in terms of a (fictitjeus) line
source along the boundary. However, it is easily seen that this function only arises because of
the approximation employed for Lt. If (2) is evaluated asymptotically for 0=1r and kr~ 1 by
deforming C into the path of steepest descent through {3 = 1T, it may be shown that H 3 = O(K- 312) as
K - oo. On the other hand, the Hankel function in (36) yields a term of order K- 112 while the sum
of the Fresnel integrals is O(K- 312 ); moreover, for fixed K, the terms which involve Fresnel integrals
dominate when lnl is sufficiently large. Therefore this "line-source" term will be neglected, and
it will be assumed that
270 R. F. Millar

(37)

This result will be discussed later in some detail.

4.2. The Field on the Sea (x > 0)

Here it is desirable to rotate the branch cuts from A= I and A= n through hr in the clockwise
direction, and to deform the integration contour in (20) into loops At. A 2 about the corresponding
cuts (fig. 4). In so doing, the pole at A=- Ao is captured in the negative sense; its contribution
may be evaluated exactly with the aid of (2). Furthermore, the appropriate approximation to L,
now is provided by (I4). There results

(38)

wherein K = kx > 0,

v=I, 2, (39)

and the relation (I5) has been employed.

Since K2 contains a factor e-inx, it is negligible if- K Im n > 2.3. Also, the factor (n + A) 112
in the integrand of K 1 may be replaced by nitz, a step which may b~ justified in a manner analogous
to that discussed previously in connection with lv· One finds

(40)

and, by employing the relation

(4I)

and assuming that K;:: I, it is seen that

(42)

in which (43)

Thus, forK;::: 1, In! sufficiently large,- KIm n:::: 2.3, (3B) and (42) imply that
2el/4rrieiK cos a
H3 = [1- p (sin a)]eiK cos a (44)
7Tl/2nl/2L, (cos a) F(Zo)

and, in accordance with (1), the total magnetic field will be.

. 2el/4rrieiK cos a
Hz= 2e~~< cos a- F(Zo). (45)
'TT 112 n 112 L 1 (cos a)
Electromagnetic Waves Neor a Coastline 271

It is now worth while to examine the cases in which a lies close to 1T· or 0; these correspond
to incidence from over the land and sea, respectively.

4.2.1. a Near 1r

If a is close to 1T, Zo can be small even for relatively large K.

Then

F(Z o) = t 1r 112 e- 1, 4 .,.;- Z o + O(Z~) (46)

and, because cos a is near -I, (I6) provides the necessary approximation to Lt(cos a). If only
the dominant terms in n are retained, (45) becomes
2neix cos a sin a 25/2el/4'1ri eix cos a cos 1/2 a
Hz= +---:-- ------Kl/2 (47)
I + n sin a 1rl/2 I+ n sin a ·

This approximation vanishes when a= 1T, in agreement with the exact solution, because in-
cident and reflected waves annihilate one another. The result (47) is valid, subject to the present
assumptions and restrictions (one of which is that K >I), provided that higher order terms in the
series for F(Zo) are negligible with respect to the fir;t two (this being so if K cos 2 < O.I, say). ta
The second term in (47) indicates the well-known "recovery effect." If a# 1T and [n[~ oo, then
the approximation becomes exact and gives Hz= 2eix cos a, which is the solution for the case of an
infinite, perfectly conducting plane.
It must be stressed that (47) has no validity for very small K, although it remains finite as
K ~ 0. However, it should conform to the behavior of Hz when a is close to 1T for points of obser·
vation no closer than about 1/a wavelength or so from the discontinuity except, possibly, at very
high frequencies or for very low soil conductivities.

4.2.2. a Not Near 1r

If a is sufficiently remote from 1T, the approximation (I4) may be employed for L1 (cos a) in
(45). For large [n[,- K lm n;::: 2.3, K;::: l, the dominant behavior is

2el/47ri ix cos a
Hz= 2eiK cos a- e F(Zo). (48)
1T112 n cos ta
Thus the effect of the imperfectly conducting earth is a term of order 1/n in the field on the sea.
This may be interpreted in terms of reflection at the coastline. Furthermore, if Zo ~ l, so that

F(Zo)- (2iZo)- 1 exp (- iZ5), (49)

(50)

and the perturbation decreases with distance from the coastline as K- 112 • This is in contrast with
the field on the surface of the land which, at sufficiently large distances from the coast and for
adequately large values of the numerical distance (- iZr), differs from that appropriate to an
imperfectly conducting dielectric by a term of order K- 312 , a result which may be deduced from (37).
On the other hand, if the numerical distance is small, the form of this difference is just- I times
the second term in (50). These results will be employed in section 6.1.
272 R. F. Millar

5. The Field Near the Coastline

To obtain a complete description of the field behavior, it is necessary to study the solution for
points very close to the coastline. Although the dominant terms derived in section 4 remain finite
at K = 0, it has been shown that these results are not quantitatively significant for small values of
K. Other than that lnl be large, the most important assumption in section 4 is that terms involving
the factor e-i71K be negligible; this exponential is of magnitude less than 0.1 if - Kimn;;;::: 2.3. In
addition, it was assumed that K was s~fficiently large to justify replacement of a Hankel function
by its asymptotic form. Even for K as small as unity, this should give reasonable accuracy, and
the resulting form of the field is relatively simple; but this assumption on the magnitude of K is
not an inherent limitation of the method.
It remains to investigate the regiori within a small fraction of a (free-space) wavelength of the
coastline. This will be done by deriving the first few terms in the Bessel-function expansion of
the solution. First it is shown that the solution of Clemmow (1953) may be expressed as a series
of increasing powers of (kr) 112, the coefficients of which depend on 8. Following this, it is demon-
strated that these terms may be rearrangtd into a series of elementary cylindrical wave functions
of period 47T in 8. It is found that the coefficients of the wave functions of integral order may be
expressed in terms of infinite integrals involving L 1 • The wave functions of order half an odd
integer involve only asymptotic properties of Lt; in this respect, the present findings correspond
with those of Stockel (1963).
Again, the starting point is (2) and, stated briefly, the series of increasing powers of (kr) 1i 2 is
obtained by deforming the contour C into the line Re f3 = 8 so that the resulting integral may be
interpreted as a Fourier integral in the variable K(= kr). The expansion of such an integral for
small K has been considered elsewhere (Millar, 1966), so a detailed derivation is unnecessary.
Evidently the coefficients of the various powers of K1i 2 depend on (} and involve the function Lt.
Therefore, although in theory (} can lie anywhere in the interval (0, 7T), in practice it is convenient
to choose a value such that the simpler approximation (14) for L 1 is valid on the deformed contour.
The coefficients ·thus determined should correspond to the exact values to the same extent that
(14) approximates to Lt. Once they have been determined, 8 may be permitted to range over the
interval (0, 7T). (It can be argued that, if a value of (} were chosen which required use of (16) for
L~, then, for sufficiently large values of In I, essentially the same values would be obtained for the
coefficients; of course, were Lt known precisely, then the values found would be exactly the same
for any choice of e.)
The total field will be expressed in terms of known functions (such as plane waves and Fresnel
integrals) plus an infinite series of elementary cylindrical wave functions. Because plane waves
and Fresnel integrals may be expanded in series of this type, it is apparent that different represen-
tations will arise depending on the manner in which the total field is decomposed. The optimum
splitting of the field depends on o:. When o: is small, for purposes of calculation it is desirable to
assume that 0 < 7T- o:, and when o: is near 7T it is convenient to take (} > 7T- o:. It will be seen that
the coefficients of the series then are as small as possible. Moreover, these coefficients are inde-
pendent of 0 and, although they will have been calculated subject to the assumption that(}< 7r- o:
or 0 > 7T- o: as the case may be, the series will be valid throughout 0 ~ 0 ~ 7T.
In the following section will be derived the series appropriate to the case of small o:; the cor-
responding results for values of o: near 7T will be given in section 5.2.

5.1. The Case of Small o:; 0 < 7T- o:

Deform C into a contour C' on which Re f3 = 0; because 8 < 7T- o:, the pole at f3 = 7T- o: is
captured in the deformation. C' forms one border of the two semi-infinite strips within which
the integral in (2) converges (fig. Sa). If () were greater than the real part of the branch point then
it would be necessary to add a branch-cut integral to the integral on C'; in essence, this is the
Electromagnetic Waves Near a Coastline
273

-2 r. FIGURE 5. Shaded domain in f3 plane (a) maps into cut '

r_ plane (b) under transformation (51).
Branch points at cos IF-" are denoted by crosaes (X).
"$plane

(a) (b)

branch-cut integral which has been ignored earlier, but its contribution probably would not be
negligible for very small K. However, in the approximate evaluation of (2) it will be assumed that
6 is small, so the branch-cut integral will not be required.
Introduce a new variable ' by the transformation

' = cos (/3 - 6) - I , Re sin .(/3- 6) < 0, (51)

~hich maps the strip 6- 7T < Re j3 < 6 into the ' plane cut along the real axis from- 2 to- oo and
from 0 to+ oo (see fig. 5b). Then C' transforms into the contour f(= f _ + fo+ f +). Because the
integrand in (2) is analytic near j3 = 6, the integral on f 0 tends to zero with the radius and hence-
forth will be ignored. Then, according to (2),

Hs= [1-p (sin a)]eiKcos<t~+a>+H, (52)

in which the first term arises from the pole at j3 = 7T- a,

~
H =-:- M(a)(H+ + H_) cos! a, (53)
7T

M(a) = l/L1 (cos a); (54)

(55)

and N('· a 6) = _ M(/3) cos !/3 (56)

' ' (cos j3+cos a) sin (j3-0)

Equations (l) and (52) show that

Hz= eiK cos (8-a) + eiK cos (8+a) + H, (57)

so it is apparent that H is the perturbation of the field from the value taken when the land is per-
fectly conducting.
Integrals of the form (55) have been examined elsewhere (Millar, 1966). It was there shown
that, for a suitable class of functions N, the integral can be expanded in a series of increasing
powers of K; coefficients of nonintegral powers of K are determined entirely by asymptotic prop-
erties of N as '~ oo on f + and f _, while coefficients of integral powers involve integrals on f"f-
and r_.
In order to apply these results in the present circumstances, it is necessary to determine the
behavior of N('; a, 6) for large values of ' on f + and f -· Consider first that ' is on f +· The
 R. F. Millar
274

expansion of 1/sin (fJ- 0) in inverse powers of' is readily found with the aid of (51). The remaining
factors of N are first expanded in inverse powers of cos {3. Also 1/cos f3 may be expanded in
inverse powers of,, the coefficients of which depend on 8:

(58)

for 1,1~ 1 on r+.

The expansion for large values of !A.! of an integral such as that which appears in (10) also
was considered elsewhere (Millar, 1966). However, in the present case, the approximation (14)
is employed for L 1(A.), and the appropriate expansion for M(/3) is obtained from (14) by straight·
forward means without recourse to the general method. (It can be verified that the first few co·
efficients in the expansion of the approximation to M(fJ) agree, to order 1/n, with those obtained
by means of the method described in the cited paper.)
If M(/3) is expanded as

M(/3) =
.
2: en/cos 10 f3 (59)
n=O

then, when 'is on r +• one finds

N('; a, 8)=2-l/2iel/2fJi{-3/2 f
n=O
g 10 ,-n. (60)

Here {-312 is real and positive on r +>

go=fo,

(61)

and fo=eo,

h = 2l eo+ e1- eo cos a,

12=-gl eo+ 2l (e!-eo cos a) +e2 e1 cos a+ e0 cos2 a, (62)

If ' is on r -· it is found that

N('; a, 8) 2-1/2i,-3/2e-1/2fJi f
n=O
g~,-n. (63)

Here ,- 3/ 2 is real and negative, and g~ may be obtained from gn by substituting 8 for 8 therein.
Electromagnetic Waves Near a Coastline 275

The series in increasing powers of K may be found by using the following result (Millar, l%6):
If fo(x) is integrable and

fo(x) - x-y f
v=O
avx-v, x~oo, (64)

wherein 0 < y < l, then if K > 0,

L"' fo(x)e-iK:rdx- ~ (tm + Um), K~O. (65)

Here (66)

(67)

8=1-y, (68)

(69)

and rdenotes the gamma function.

In the present application, y =i and ao = 0; moreover, it is permissible to replace the signs
of asymptotic equality by those of equality. It is to be noted that the coefficients of nonintegral
powers of K are determined wholly by the expansion coefficients av, while the coefficients of in-
tegral powers involve the integration of fo. According to (53), (55), (60), and (63), it is seen that

i ~
H=-M(a)e-iK cos j a ,L bmf2Kml2 , (70)
1T m=O

in which

bm = (- i)m(Nm+l + N,;.+l ), l
bm+t/2 = (- i)m+12-l/2el/41Ti1T(gm + g;,.)Jr ( m + ~) 'J (71)

for m=O, 1, 2, . ., with

Nm+! =-,l f. {
~m N(~; a, L
8)-2-l/2iel/28i~-3/2 m-1 gv~-v } d~, (72)
m. ~ -

(73)

wherein the sums contain no terms if m = 0.

By means of a straightforward calculation, it is possible to rewrite the right-hand side of (70)
as· a series of elementary cylindrical wave functions. To this end, assume that

(74)
m=O m=O

where } denotes the Bessel function of the first kind. By employing the power series expansion
276 R. F. Millar

of this function and of e-iK, and identifying the coefficients of like powers of K on each side of (74),
one easily finds that

(75)

The coefficients of the Bessel functions of integral order involve Nm + N;,. for various values of
m and they will not be written explicitly. An outline of a method for simplifying these coefficients
is as follows. It is convenient to deform +, r r-
into contours which run in the ' plane from 0 to
ooeiD and from ooe-i8 to 0, respectively. (ooe:!:iil correspond to {3 =± ioo; figure Sa indicates that no
branch point singularities are captured in the deformation, and the pole at {3 = 7T- a is not captured
because () < 7r- a.) Then each coefficient Am may be expressed as the sum of two integrals in
the {3 plane. The integrands differ because of the different () dependence of the expansions for
N('; a, 8) on f + and f -· However, the term which involves N explicitly is common to both inte·
grands (72) and (73); thus the corresponding integrals may be replaced by one along the imaginary
axis in the {3 plane. The terms which differ in the two integrands may be integrated explicitly.
If this program is carried out, it is found that, for m=0,1,2,3,and 4,

(76)

in which Eo= 1 and Em=2, m= 1,2,3, . . . , Tm(z) denotes the Chebyshev polynomial of 9rder m
and argument z[Tm (cos ())=cos m8], while powers of I are to be interpreted as follows: Im = I m,
m=0,1,2, . . . , with

M({3) h }
Im = f-•.,. cosm {3 cos ! {3
i"' {
cos
{3 + ·
m
L
__!!_{3 d{3.
cos a n=l cosn
(77)

The series (which is absent when m=O) represents the first m terms in the expansion of M({3)
-+(cos {3 +cos a) in inverse powers of cos {3; the coefficients hn can be found easily in terms of the en
with the aid of (59).
Although the form (76) of Am has been verified only to m = 4, it is conjectured to be generally
valid.
To compute the coefficient Am, it is thus necessary to evaluate the integrals Iv for v= 0,1,2, ... ,
m. This will be described in section 5.3.

5.2. The Case of a Near 7T; () > 1r-a

It has been mentioned that when a is near 7T, the optimum splitting of the field is obtained by
assuming that () > 7T- a in the calculation. Although repetition of the analysis of section 5.1 under
the assumption that () > 7T- a would lead to the desired result, it may be obtained more readily in
the following manner.
The function H 6 is determined by (52), wherein the Bessel-function expansion of H may be
found through (70) and (74). By employing the expansion

eiK cos(8+al= ~ EmimJm(K) COS m(O+a), (78)

m=O
Electromagnetic Waves Near a Coastline 277

and the relation (15), it is seen that H8 may be written in the form

2[1- p(sin a)] "" imJm(K) sin m(J sin ma,

L (79)
m=l

wherein Bm12=Am/2, m= 1,3,5, . . . , ) (80)

Bm=Am+Cm, m 0,1,2,. .,

and Cm=Emmm-I cos ma/{Lt(- cos a) cos fa}. (81)

The function Hz is given by (1) and (79). Although the result is more complicated in ap·
pearance than that given by (57), (70), and (74), it is, in fact, more suitable for purposes of numerical
computation when a is near?T. This is because the coefficients Am are unbounded at a=?T. The
Cm also are unbounded at a=?T, hut the Bm are bounded.

5.3. Evaluation of Coefficients Am

The computation of the coefficients Am necessitates the evaluation of the integrals Im. The
path of integration is the imaginary f3 axis which, in the A.(= cos /3) plane, corresponds to a loop
about the branch cut from A 1 to A=+ oo. Thus the contour lies in the domain within which the
approximation (14) for Lt is valid and will be employed. Then, in order that the integrals converge,
it is necessary that the coefficients hn in (77) be given the values derived from the approximation
to Lt.
Exact evaluation of lm does not seem to be possible, even with L 1 approximated by (14).
Therefore, asymptotic methods will be employed which should yield a good approximation to the
integrals provided that In! is sufficiently large. Unfortunately, any approximation (either in L 1
or in the evaluation of the integrals) restricts the range of Kover which the series solution is useful.
It is found that the coefficients of the Bessel-function series increase in magnitude with the sum-
mation index m, although the series itself is ultimately convergent. Thus, unless K is close to
zero (so only the first few terms are of importance), small relative errors in the coefficients of higher
order turn out to be of magnitude comparable to that expected of the solution and the result be-
comes quantitatively worthless. .
The computation of the Am is an even more tedious task than the previous calculations; details
may be found in the cited report. 3 It is found that

Ao ""'2in- 1 [n 112 a/sin a+ !n- 112 {log n+ log 2-C+ 1-.p(- !)} J, (82)

A 1 ""'2iA 0 cos a+ 4n- 112{log n +log 2- C- 2- !/1(-! )} , (83)

and A2 = 2Ao+ 2iAt cos a+ 16in112/3. (84)

Here C(= 0.5772157 .) is Euler's constant and

'"The farlield of an H-polarized line source near the edge of a perfectly conducting half plane in the interface between two different media, with application to mixed
path propagation." Laboratory of Electromagnetic Theory. Technical University of Denmark, Lynghy, Denmark. Report R 43, Jan. 1965. This work ,... also de·
scribed at the Sympo&ium on Electromagnetic Wave Theory, Delft, The Netherlands, Sept. 6-11, 1965.

238-469 0-67-2
278 R. F. Millar

(= f' (t ), where f' denotes the derivative of the r function).

In (82), (83), and (84), certain inverse powers of n have been neglected. The approximations
for the coefficients Am (m;;;. 3) will contiun errors involving positive powers of n which, therefore,
become large if n is large. A discussion of the effect of such errors may be found in the cited
report. 3 If these errors are to have a negligible effect, it turns out that the value of K must be
sufficiently small. In the numerical calculations to be described later, K will be so limited that
terms involving Am for m ;;;. 3 would have little effect on the sum.
It is of interest to compute the field for K = 0. With the aid of (57), (70), and (74), it is seen that

(85)

if K=O. It may be noted that the analyses of Clemmow (1953) and Senior (1957) suggest that
Hz= 2 when K= 0 and a= 0, although neither analysis is quantitatively significant for small values
of K. However, (85) indicates that the imperfectly conducting medium does exert some influence
on the field at K=O when a=O. In this case, L 1 may be approximated by (14) and, if a= K=O,

Hz = 2- 2/(7rn) (86)

plus higher-order terms in 1/n; this result should be exact to terms of order 1/n inclusive.
It is seen that when a= K = 0, the effect of the imperfectly conducting earth is a term of order
1/n. On the other hand, if a= 1r and K = 0, (85) gives Hz = 0. This result is to be expected be-
cause p(O) =-I if In I is finite. Thus the total field must vanish everywhere for grazing incidence
over the imperfect conductor. Evidently the field at (and near) K = 0 is strongly dependent on the
angle of incidence and quite different effects are to be expected near K = 0 depending upon whether
a is in the neighborhood of 0 or 1r.

6. Analytical Comparison With Other Results

In this section, comparisons will be made with results of Wait and Bach Andersen, which have
been derived by different means.

6. 1. Comparison With Results of Wait

Wait (1963) expresses the effect of the mixed path in terms of the change (Azm) in the mutual
impedance between unelevated transmitting and receiving dipoles, referred to the case of a homo-
geneous earth (zm). The transmitting dipole is placed on the imperfectly conducting earth, far
from the coastline. For propagation normal to the boundary, Wait finds that

fum/Zm = ~g(kx), (87)

wherein x denotes distance from the receiver to the boundary and is positive or negative according
to whether the receiver is on sea or land. The function g is a combination of cylinder functions of
orders 0 and 1. If the parameters in Wait's solution are specialized to the case considered here,
it is found that flo= e- 1/ 4 7ri /n.
Wait's result (87) actually is a measure of the perturbation of the vertical component of elec-
tric field at the receiver caused by the sea. Thus, for purposes of comparison, the vertical electric
Electromagnetic Waves Near a Coastline 279

field of the present solution must be computed and suitably normalized. (In the report 3 on which
this work is based, a normalized magnetic field was compared with the results of Wait. This
procedure, which is justifiable in some circumstances, but not in cases where the field contains a
reflected component, led to the reported disagreement with Wait's result.)

Let (88)

in which Ey denotes the y component of the total electric field, £i corresponds to the first two
terms in (1) and £s may be obtained from H 8 by Maxwell's equations. Define f(8; K) by

f(8; K) =lim £sf£i. (89)

Then f(8; K) is the quantity which should correspond to lizm/Zm; the interested reader is referred
to section 9.1 of the cited report 3 where this point is discussed. For points on the sea 8= 0, while
8= 7T describes points on the land; f(7T; K) is a measure of the wave reflected back toward the
transmitter by the coastline.
Expressions for f(O; K). and f(7T; K) will be obtained from the formulas of section 4, which are
not valid in the immediate neighborhood of the coastline. When 8 = 0, H 8 is given approximately
by the second term of (47), in which a may be set equal to 7T except in the factor cos ! a. From
this, E 8 is easily derived. It is found that

(90)

for sufficiently large K. This is identical with the asymptotic form of Wait's result (Wait, 1963,
eq. 26).
When 8 = 71", it is necessary to use the expression (37) for H•, from which E 8 may be calculated.
If it is assumed that I Z , I ~ I (that is, the numerical distance is small), but K itself is sufficiently
large that asymptotic expansions for F (Z v)(v= 2, 3) may be employed, one finds

(91)

which is identical with (24) of Wait (1963) for the case of normal incidence.
Thus it is seen that if K is sufficiently large (of the order of 2 or 3, say, at least) so that certain
asymptotic expansions provide valid approximations, then the present results agree with those of
Wait for the field on the sea. Moreover, if it is further assumed that the numerical distance to
an observation point on the land is small, there is agreement with Wait's result for the field on
the land.
If no restriction is placed on the magnitude of the numerical distance, but K is still sufficiently
large, it may be shown that

Ao e-i2K
. ;;;;---;. -mG(p), (92)
2 v(:.:7r) K

wherein (93)

(94)
280 R. F. Millar

is the numerical distance, and erfc is the complement of the error function. This expression for
f(1r; K) is in complete agreement with the generalized result of Wait (1965, eq. 2a).
The agreement between the results of the present analysis and those of Wait gives added
support to the use of an impedance boundary condition to describe the land properties when n is
large. Of course, in the very immediate neighborhood of the coastline, the results do not cor·
respond. Both are singular at the discontinuity, but the singularities differ; the result of Wait
becomes logarithmically infinite, while in the present solution Ey tends to infinity as K - 112 •
The magnetic field Hz remains bounded in the neighborhood of the coastline. Therefore,
when discussing field behavior near the coast (in particular, for purposes of numerical calculations)
it is more convenient to discuss the variation of Hz than that of Ey.
In analogy with (89), define A( 8; K) by

(95)

wherein H 1 denotes the first two terms of (1). In the neighborhood of the coastline, A may be
determined with the aid of (57), (70), (74), (75), and (82). If () 0, and only the dominant terms in
n are retained,

(96)

while, to the same order in n,

e-iK
A(1r; K)=-Jo(K)+. (97)
n1T

In (96), Bessel functions and powers of K of order greater than! have been ignored, while those
of order 1 and greater are neglected irr (97).
Although valid only for small K, (96) and (97) are indicative of a recovery of phase by the field
as it crosses the coastline from land to sea, for 17"/4 < arg {e 114 -rrin- 312} < 51T/8 because -rr/4 < arg n
< 0. For large K, A(O; K) also is given by the right.hand side of (90). Thus, the net phase recovery
over the perfectly conducting sea is about 17"/4; this appears to be in good agreement (as it must
be) with figure ll of Clemmow (1953).
One other point is worth mention. It has just been remarked that

for large K and large n; but it is possible to determine the behavior of A(O; K) for large K and all n
from (2), merely by evaluating the integral by the method of steepest descent and employing (15).
One finds

(98)

6.2. Comparison With Results of Bach Andersen

Bach Andersen (1963) calculated the vertical electric component of the sky-wave field over
the sea from a vertical electric dipole placed on the land. The land-sea model chosen by Bach
Andersen was the same as that selected for the present analysis, but his time·dependence factor
was e-iwt. Bach Andersen employed certain approximations together with the compensation
Electromagnetic Waves Near a Coastline 281

theorem (Monteath, 1951), and his results are not valid when the transmitter is close to the boundary.
The initial purpose of the present analysis was to determine rigorously the far field over the sea
when a vertical dipole was placed on land close to the coastline, and so deduce the domain of
validity of Bach Andersen's formulas.
By invoking a reciprocity principle, the results of the present analysis involving the field near
the coastline due to an incident plane wave can be rephrased in terms of Bach Andersen's con-
figuration. His results are expressed in terms of a normalized far-field pattern A' which, when
expressed in the present notation, is given by

(99)

Now a denotes the direction of the receiving dipole over the sea, while the+ or sign is selected
accordingly as the transmitter is on land or sea.
If the transmitter is on land, and a~ 0 (so the far field is observed on the sea surface), then,
provided that certain exponential terms are negligible and K ;::: 1, H8 is determined by (37). If
only the dominant terms in n are retained in this equation, it is found that

A' (100)

The result (100) may be identified completely with that of Bach Andersen (1963, eq. 7), when
the different notation and time dependence employed there is taken into account. Thus the result
of Bach Andersen (obtained by approximating the magnetic field on the surface of the sea by the
value it would have if the earth were perfectly conducting) is a valid approximation to the exact
result provided that n is sufficiently large, K > l, and terms of order e-inK are negligible.
For a"''
0, a comparison of the present ;esults with those of Bach Andersen is also possible.
Once again, if K ;::: I and certain exponential terms negligible, then (37) is appropriate. Further-
more, the term which involves F(Z 2 ) is negligible in comparison with the others if n is sufficiently
large. If c* (essentially the complex conjugate of c as defined by Bach Andersen} is defined by

c* 2i (l -cos a) n 2 ,

and if a is sufficiently remote from 1r that L1 may be represented by (14), then it is found that

l 2 ·• I * 2 -1/4ni
A ' =-(l + p) cos a + t Vc. • ~F(Z 3
)+ e .~* . e-IK(J-cosa)..iz•l:?(Z)
~ u· 1 • (101)
2 -t 1r 112 c -t

Provided that the different notations and time dependences are taken into account, and the
factors cos a in the latter two terms are replaced by unity, then (101) may be identified with Bach
Andersen's equation (4). Therefore, his result is valid if n is sufficiently large, K;;;::: l, terms of
order e-inK are negligible, and cos a = 1.
Bach Andersen determined also another solution by employing the solution of Clemmow (1953)
in his surface integrals; Clemmow's solution is not valid up to the boundary. The result of this
calculation (Bach Andersen, 1963, eq. 11) is, in the present notation,

1 1. e-tf+rri-iVc* {2iVc* F(Z3) e- 114ni e-IK(l-cosa) . }

A'=2(l+p)cosa+2t(1+p) Vc* c*-i · 1r112 +2 1r112 • c*-i eizfF(Zt).
(102)

Bach Andersen's work implies the equivalence of (101) and (102) for small a. For larger a,
and large n, one sees that (e- 1' 4"'LiVc*)/Vc* =-i. Thus the essential difference between (101)
282 R. F. Millar

and (102) is that the factors cos a in the latter two terms of (101) are replaced by (1 + p)/2 in Bach
Andersen's equation. But for a sufficiency remote from 0 and -rr, and n large, (1 + p)/2 = 1 ¥-cos a.
Thus, Bach Andersen's equation (11) apparently is valid only for small a, and not, as surmised, for
larger values.
The expression (101) remains finite asK~ 0, although it does not provide a valid approximation
in the immediate neighborhood of K=O where neglected terms become important. Nevertheless,
if n is sufficiently large, the error remains small. For if K =a= 0 in (101), then A'= l. On the
other hand, in the same circumstances (85) yields
1
A'= 1- --rrn3/2
- [n 112 +l.2 n- 112 {log n+ log 2-C+ 1-lji ( - !)}

which differs by little from unity if n is sufficiently large.

7. Numerical Results

In order to illustrate the results of the present analysis, some numerical calculations have
been performed. The values selected for the parameters E', a-' and w are those employed by
Bach Andersen (1963): €=10- 8 /(36-rr) f/m, E'/E=10, a-'=10- 2 mhos/m, w=2-rrf,f=10 MHz, so
that n=3.91-i2.30=4.53 exp (-i30.45°).
Most attention will be directed towards the configuration considered by Bach Andersen; in
particular, the function A', defined by (99), is computed for a= 0°, 4°, 100, and 20° for positions of
the transmitter on both sides of the coastline (Bach Andersen considered only a land-based trans-
mitter). Furthermore, A(O; K) (95) is computed for e= 0 and e= 1r in the vicinity of K = 0.
Figures 6a, b, illustrate the variation in amplitude and phase of A' with K(=Jkxj) on both sides
of the boundary close to the coastline. In the immediate neighborhood of the coastline, the Bessel-
function representation has been employed. For larger values of K the results of section 4 have
been used; when the transmitter is on the sea, the field is determined by (45) and for a dipole on
the surface of the land the relevant equation is (37). However, because the value selected for n is
not very large, it was found to be desirable to include second terms in the asymptotic expansions
of the 1, (see (24)). The effect of so doing is to multiply the right-hand side of (37) by the differential

{ 1. d}
operator 1 +-!:.- ·
2 n dK
Because of the different forms of the Bessel-function expansion on each side of the boundary,
both lA' I and arg A' have infinite derivatives as K ~ 0 on the sea; the effect is too small to be evi-
dent in figure 6a.
The smooth transition from the representation for A' in the neighborhood ·of K = 0 to those
appropriate for larger K is gratifying, for it indicates the internal consistency of the present work.
However, at best, these results will correspond to the exact values only to the extent that the
approximation to L 1 conforms to the true values. Nonetheless, it is felt that the curves in figure 6
are probably in good agreement with the exact values.
The curves in figure 6a for kx < 0 join smoothly with those illustrated in figure 3 of Bach
Andersen (1963), while those of figure 6b for kx < 0 merge with values of arg A' calculated from
Bach Andersen's formulas. Moreover, if Bach Andersen's numerical calculations are extended
to kx=O, it is found that even at the coastline the two values of JA'J differ by less than 10 percent,
and the difference between the values of arg A' at kx= 0 is only 4°. The continuation of the
curves in figure 6 for larger values of kx over the sea takes the form of a standing wave arising
from interference between the incident field and the small wave reflected by the coastline.
It is evident that there is little variation in JA'J near K=O, especially over the land. Even
for the relatively small value of n employed here, JA'J is close to cos a, which is the value it would
Electromagnetic Waves Near a Coastline 283

take if the land were a perfect conductor. On the other hand, the variation of arg A' is of more
interest. (Arg A' would be zero if the land were perfectly conducting.) From figure 6b, it is
apparent that rapid changes in arg A' take place near K = 0, although the total change in phase
is only a few degrees. When the transmitting dipole is on the sea, there is some interference in
the far field between the direct wave and a reflection from the coastline. This oscillation in
phase increases in amplitude as the coastline is approached. When K = 0. 7, the maximum phase
deviation of about + 3.5° is attained. Then, as K ~ 0 on the sea, arg A' decreases rapidly, passes
through zero, and is about -4° at K = 0. As K increases from zero over the land, the decrease
in arg A' continues, but at a gradually decreasing rate, and eventually approaches the value ap-
propriate to the imperfectly conducting land.
It is easy to see that the variations in arg A' illustrated in figure 6b (now interpreted in terms of
a distant transmitter over the sea and a receiver near the coastline) are in agreement with some of
the observations reported by Pressy, Ashwell and Fowler (1953); these refer to the phase velocity
of the field near a discontinuity in the earth's conductivity when the transmitter is far distant over
a region of high conductivity. For, with the help of (99), it is not difficult to show that on y= 0 the
phase (1/1) of Hz is given by

1/J=kx cos a+wt+c/>,

where x > 0 for points on the sea, x < 0 for points on the land, and cf>= arg A'. The component of
phase velocity of Hz (or of Ey) parallel to the earth's surface (vp) may be obtained by equating dl/1 to
zero. Because· the field is incident from x > 0, vp=- dx/dt, so

vp/c=[cos a+dcf>/dx]- 1 ,

argA'

5'

!P:I

1.1
X

0.8
-2.0 -1.0 0 1.0 20 kx
land sea

-15'

FIGURE 6a. Variation of lA' I with kx near kx = 0, for several

values of a. FIGURE 6b. Variation of arg A' with kx near kx = 0, for
Crosses (X) are points calculated from Bessel-function series; circles (Q) denote several values of a.
values computed according to formulas of section 4. Curves give .. best fit." Crosses. circles. and curves have same significance as in figure 6a.
284 R. F. Miller

where c is the speed of light in free space. As the point of observation moves away from the
transmitter and approaches the coastline, vp oscillates about the value c/cos a which is appropriate
to the perfect conductor or free space. When kx = 0.7, vp=c/cos a. Thereafter, Vp decreases to
0 at x = 0. The phase velocity is discontinuous at x = 0 because of the idealized nature of the
present theoretical treatment of the problem.
The sudden decrease in phase velocity, noted by Pressey, Ashwell and Fowler (1953), certainly
is indicated by the present calculations. The most rapid changes appear to take place before· the
coastline is reached.
The case in which a wave is incident from over the imperfectly conducting land (or when the
transmitter is near the coastline and the receiver far away on the land) is illustrated in figures 7a,b,
which show the variation with kx of A. For small values of kx, the Bessel-function series (79)
has been employed. Here a -7T, while 8=0 or 7T according to whether x > 0 or x < 0. For larger
values of kx, H8 is represented by (37) or (44) as the case may be. Here, too, it is seen that a smooth
transition takes place between the two types of field representation for values of jkxj somewhat
less than unity.
It is apparent that !AI is almost constant over the land near the sea, although it increases
slightly as the coastline is approached (fig. 7a). Over the land, arg A varies essentially like ei2kx(x
.CO), indicating the presence of a wave reflected by the edge. However, !AI is quite small, tending
to zero as K- 312 as K - oo on the land, so the coastline and sea have little effect on the field on the
land.
When the point of observation crosses from land to sea, a recovery effect is immediately notice-
able. Then A increases abruptly in both amplitude and phase; the initial slopes of !AI and arg A
are infinite for the reason mentioned previously in connection with A'. The infinite slope of !AI
at kx = 0 + is not shown in figure 7a; A soon assumes its asymptotic behavior which here is

A(O; kx) - 0.1737 ei74.2o(kx)l/2.

The rapid increase in the phase of A over the sea corresponds to an increase in the phase ve-
locity .of the field as the wave crosses the coastline. This, too, agrees with the observations of
Pressey, Ashwell, and Fowler (1953); see also figure ll of Clemmow (1953).

II\I
so· ar 1\ (0, kx)
0.3
.."
"
sea
1.0 2.0 kx

"
"
-2.0 -1.0 0 1.0 2.0 kx
land sea

FIGURE 7b. Variation of phase of A near kx = 0.

FIGURE 7a. Behavior of IAI near kx = 0. Points (X) and (+) are calculated from appropriate Bessel-function series:
Crosses (X) are calculated from Bessel-function series; circles (0) denote values circles (0) denote values computed from formulas of section 4. Curves give
computed according to formulas of section 4. t.;urves give ""best fit. •• "best fit."
Electromagnetic Waves Near a Coastline 285

It may be noted that here the most rapid changes take place after the coastline is crossed.
This result, together with that corresponding to a = 0, suggests that when a is near 0 or 1r the most
rapid changes take place over the better conductor.
The observations of Pressey, Ashwell, and Fowler (1953), noted above, were made over a
naturally occurring mixed path. However King and Maley (1965) recently have reported the
results of model experiments over mixed paths, the model being essentially that selected for the
present analysis. Favorable agreement was obtained between the experimental measurements
and computed values based on the theory of Wait (1963), except in the immediate neighborhood
of the discontinuity where the impedance concept breaks down. Because of the agreement
of the present results with those of Wait, noted in section 6.1, it may be concluded that the ex·
perimental values of King and Maley and the present results are in agreement. No attempt has
been made to compare theory with experiment in the immediate neighborhood of the discon-
tinuity, where accurate experimental measurements are difficult to obtain.

8. Concluding Remarks

Propagation of electromagnetic waves across a land-sea boundary has been examined in the
present paper. The sea was represented by a perfectly conducting half plane in the interface
between free space and the imperfectly conducting earth.
Diffraction of an H-polarized plane wave by this configuration was studied by Clemmow (1953),
and an approximate reduction of his solution for large values of n, has been provided here. Near
the edge of the half plane, the solution is represented by its Bessel-function expansion; at more
distant points, a representation in terms of Fresnel integrals is derived.
The present results have been compared analytically with those of Wait (1963) and Bach
Andersen (1963); in both cases, agreement obtains except in the immediate neighborhood of the
coastline. This indicates that the use of impedance boundary conditions and certain approxi-
mations in connection with the compensation theorem lead to valid results even in regions where
one might not expect to obtain them. Moreover, the agreement between (91) and Wait's result
suggests that the form of the wave reflected back toward a land-based transmitter by the coast-
line, expect in the immediate neighborhood of the discontinuity, is insensitive to the particular
analytic characterization of the discontinuity, provided that n is large.
The value of the present work may lie in the fact that it provides a description, albeit ap-
proximate, of field behavior in the immediate neighborhood of the coastline when a realistic model
of the discontinuity is employed.
Analytical and numerical calculations have indicated that very rapid variations are to be
expected in the phase of the field near the coastline. Correspondingly large variations in the
amplitude do not seem to occur (at least for the value of n here employed) when the field is incident
over the sea, although there is certainly a noticeable effect. It is suggested that when a is near 0
or 1r, the most rapid changes take place over the better conductor.

The author is grateful to Professor H. Lottrup Knudsen for the invitation to work in his labora-
tory and for his encouragement during the period of this work. It is also a pleasure to acknowledge
the many helpful and informative discussions on the problem which were held with a number of
colleagues in the laboratory.

9. References

Bach Andersen, J. (1963), The radiation field from a vertical dipole on an inhomogeneous ground, Electromagnetic Theory
and Antennas, ed. E. C. Jordan, 1099-1112 (Pergamon Press, London).
286 R. F. Millar

Bazer, ]., and S. N. Karp (1962), Propagation of plane electromagnetic waves past a shoreline, J. Res. NBS 66D (Radio
Prop.), No. 3, 319-334.
Clemmow, P ..C. (1953), Radio propagation over a flat earth across a boundary separating two different media, Phil. Trans.
Roy. Soc. (London) A246, No. 905, 1-55.
King, R. ]., and S. W. Maley (1965), Model experiments on propagation of groundwaves across an abrupt boundary at per·
pendicularincidence, Radio Sci. J. Res. NBS 69D, No. 10, 1375-1381.
King, R. J., and S. W. Maley (1966), Model experiments on propagation of groundwaves across an abrupt boundary at oblique
incidence, Radio Sci. I (New series), No.1, 111-115.
Maley, S. W., and H. Ottesen (1964), An experimental study of mixed-path groundwave propagation, Radio Sci. J. Res.
NBS 68D, No. 8, 915-918.
Millar, R. F. (1966), On the asymptotic behavior of two classes of integrals, Soc. Ind. Appl. Math. Rev. 8, No. 2, 188-195.
Millington, G. (1949), Ground-wave propagation over an inhomogeneous smooth earth, Proc. Inst. Elec. Engrs. 96, III,
53-64.
Millington, G., and G. A. Isted (1950), Ground-wave propagation over an inhomogeneous smooth earth. Part 2. Experimen·
tal evidence and practical implications, Proc. Inst. Elec. Engrs. 97, III, No. 48, 209-217.
Monteath, G. D. (1951), Application of the compensation theorem to certain radiation and propagation problems, Proc.
Inst. Elec. Engrs. 98, IV, 23-30.
Noble, B. (1958), Methods Based on the Wiener-Hopf Technique for the Solution of Partial Differential Equations (Perga-
mon Press, New York).
Pressey, B. G., G. E. Ashwell, and C. S. Fowler (1953), The measurement of the phase velocity of ground-wave propagation
· at low frequencies over a land path, Proc. Inst. Elec. Engrs. I 00, III, No. 64, 73-84.
Pressey, B. G., and G. E. Ashwell (1956), The deviation of low-frequency ground waves at a coast-line, Proc. Inst. Elec.
Engrs. I 03B, No. 10, 535-541.
Senior, T. B. A. (1952), Diffraction by a semi-infinite metallic sheet, Proc. Roy. Soc. (London) A2I3, 436-458.
Senior, T. B. A. (1957), Radio propagation over a discontinuity in the earth's electrical properties, I, Proc. Inst. Elec. Engrs.
1 04C, 43-53.
Stockel, H. (1963), Die Beugung einer ebenen elektromagnetischen Welle an einer bzw. zwei idealleitenden Halbebenen in
der Trennebene zwischen zwei homogenen, isotropen Medien, Ann. Physik, Series 7, 12, No. 3-4, 156-185.
Wait, J. R. (1957), Amplitude and phase of the low-frequency ground wave near a coastline, J. Res. NBS 58 (Radio Prop.),
No. 5, 237-242.
Wait, J. R. (1963), Oblique propagation of groundwaves across a coastline, Part I, J. Res. NBS 67D (Radio Prop.), NQ. 6,
617-624.
Wait, J. R. (1964), Oblique propagation of groundwaves across a coastline, Part III, Radio Sci. J. Res. NBS 68D, No.3,
291-296.
Wait, J. R. (1965), Nature of the electromagnetic field reflected from a coastline, Electronics Letters 1, No.3, 65-66.
Wait, J. R., and C. M. Jackson (1963), Oblique propagation of groundwaves across a coastline, Part II; J. Res. NBS 67D
(Radio Prop.), No. 6, 625-630.
Wait, J. R., and K. P. Spies (1964), Propagation of radio waves past a coast line with a gradual change of surface impedance,
IEEE Trans. Ant. Prop. AP-I2, No.5, 57Q-575.
Paper (2-3-196)