Electromagnetic Waves PDF

Electromagnetic
An Introductory Course
Waves
M.D. Verweij, P.M. van den Berg, H. Blok

Electromagnetic Waves
— An Introductory Course
About the authors
Martin Verweij (1961) graduated in 1992 from Delft University of
Technology, Delft, Netherlands. From 1993-1997, he was a research
fellow of the Royal Netherlands Academy of Arts and Sciences at the
Laboratory of Electromagnetic Research of the Delft University of
Technology, and since 1998 he is associate professor in the same
laboratory. He has written a range of papers on integral transformation
methods for electromagnetic, acoustic and elastic waves. He has been
teaching various classes on basic and advanced electromagnetic wave
theory since 1994 and was chosen Best Teacher of Electrical
Engineering in 2003.
Peter van den Berg (1943) was a member of the scientific staff of the
Laboratory of Electromagnetic Research, Delft University of
Technology, Delft, Netherlands. He graduated in 1971 from the Delft
University of Technology, and during the academic year 1973-1974 he
was Visiting Lecturer in the Department of Mathematics, University of
Dundee, Scotland. He was appointed as a full professor at the Laboratory
of Electromagnetic Research in 1981. Since 2003 he is a research
professor in the Faculty of Applied Sciences of the Delft University of
Technology. During his career, he has written an impressive amount of
papers on the numerical analysis of forward and inverse electromagnetic
wavefield problems, and he has been teaching various classes on
electromagnetic, acoustic and elastic waves.
Hans Blok (1935) was a member of the scientific staff of the Laboratory
of Electromagnetic Research, Delft University of Technology, Delft,
Netherlands. He graduated in 1970 from the Delft University of
Technology, and he was appointed full professor at the Laboratory of
Electromagnetic Research in 1980. He was dean of the Faculty of
Electrical Engineering, Delft University of Technology in the period
1980-1982. During the academic year 1983-1984 he was a visiting
scientist at Schlumberger-Doll Research, Ridgefield, Connecticut,
U.S.A. During his career, he has written a number of papers on
resonators and optical waveguides, and he has been teaching various
classes on electromagnetic waves and signal theory. He is
emeritus professor since 2000.
— An Introductory Course
M.D. Verweij
P.M. van den Berg
H. Blok
VSSD
© VSSD
First edition 1999
Second edition 2001-2006
Published by:
VSSD
Leeghwaterstraat 42, 2628 CA Delft, The Netherlands
tel. +31 15 278 2124, telefax +31 15 278 7585, e-mail: hlf@vssd.nl
internet: http://www.vssd.nl/hlf
URL about this book: http://www.vssd.nl/hlf/e016.htm (a.o. to download the auxiliary
Repetitive Guide)
A collection of digital pictures and an elctronic version can be made available for
lecturers who adopt this book. Please send a request by e-mail to hlf@vssd.nl
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical, photo-
copying, recording, or otherwise, without the prior written permission of the publisher.
ISBN 90-407-1836-9 EAN 9789040718366
NUR 924
Keywords: electromagnetic waves
Preface
The course ”Electromagnetic Waves” offers an introduction in the the-

oretical concepts of electromagnetic waves. This course book contains the
basic material on time-varying wavefields and their applications in electri-
cal engineering, e.g., electromagnetic compatibility, communication and re-
mote sensing. A prerequisite to this course is a standard course ”Electricity
and Magnetism” where, from experimental laws, the Maxwell equations for
time-varying electromagnetic fields are formulated as a system of partial
differential equations.
Chapter 1 reviews the necessary mathematical background, while Chap-
ter 2 introduces the fundamental mathematical equations: the Maxwell equa-
tions, the constitutive relations and boundary conditions. The main line of
the course is the construction of solutions to these equations in some sim-
ple configurations. The concept of an electromagnetic wave is introduced in
Chapter 3, where one-dimensional waves are discussed. A wave phenomenon
can only be understood in connection with an electromagnetic source that
generates a wave. For the excitation of one-dimensional waves, the planar-
electric-current sheet is chosen. As a simple example of one-dimensional wave
propagation, the parallel-plate waveguide is discussed shortly. In Chapter 4,
the two-dimensional waves are studied, in particular specific properties as
interference, Fresnel reflection/transmission factors, Brewster’s angle and
total reflection are treated. In Chapter 5, the consequences of a weakly in-
homogeneous medium are discussed and the theory of electromagnetic rays
is introduced. Further, in Chapters 6 and 7, the theory of transmission lines
and electromagnetic waveguides is treated. In view of communication ap-
plications, the closed parallel-plate waveguide and the open dielectric-slab
waveguide are described in full detail. Finally, Chapter 8 deals with the
excitation of two-dimensional waves and the concept of the far-field approx-
imation is introduced.
The student who has successfully completed the present introductory
course on electromagnetic waves, has learned the basic concepts of electro-
magnetic wave propagation. By simplifying the problems in such a way
vi preface
that a description in terms of one-dimensional and two-dimensional waves

suffices, more attention can be given to the physical understanding of the
propagation phenomena. However, it is stressed that in more realistic con-
figurations of present-day technology, a full three-dimensional description
of electromagnetic wavefields is needed. In this context, it is noted that
the methodology of handling the radiation and scattering of electromagnetic
waves in three-dimensional configurations will be treated in more advanced
courses of the electrical engineering curriculum.
The authors acknowledge Dr. E. C. Slob for compiling the original set of
exercises, problems and answers; Dr. M. D. Verweij for contributing to the
material of Chapters 4 and 5, revising the exercises, problems and answers,
and preparing the final print version; and Mr. K. F. I. Haak for re-checking
the answers.
Delft, January 1999 H. Blok

P.M. van den Berg
Preface to the second edition

This edition is identical to the first edition, except that a number of
errors have been corrected. The authors acknowledge their collegues of the
Laboratory of Electromagnetic Research and in particular Dr. D. Quak and
Mr. P. Jorna for reporting most of these errors.
Delft, September 2001 M.D. Verweij

P.M. van den Berg
H. Blok
Contents
Preface v
1 Introduction 1
1.1. Cartesian vectors and their properties . . . . . . . . . . . . . 4
1.1.1. Addition, subtraction and multiplication of vectors . . 4
1.1.2. Differentiation with respect to a parameter . . . . . . 6
1.1.3. Differentiation with respect to the spatial coordinates 6
1.2. Exercises and problems . . . . . . . . . . . . . . . . . . . . . 17
2 The Electromagnetic Field Equations 21

2.1. Force exerted on an electric point charge . . . . . . . . . . . . 21
2.2. Maxwell’s equations in vacuum . . . . . . . . . . . . . . . . . 24
2.3. Maxwell’s equations in matter . . . . . . . . . . . . . . . . . . 25
2.4. The constitutive relations . . . . . . . . . . . . . . . . . . . . 28
2.5. The system of field equations . . . . . . . . . . . . . . . . . . 31
2.6. The boundary conditions . . . . . . . . . . . . . . . . . . . . 32
2.7. Frequency-domain representations . . . . . . . . . . . . . . . 36
2.7.1. The frequency-domain field equations . . . . . . . . . 37
2.8. Polarization state . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.9. Poynting’s theorem . . . . . . . . . . . . . . . . . . . . . . . . 41
3 One-dimensional Electromagnetic Waves 49

3.1. The planar electric-current sheet as emitter . . . . . . . . . . 50
viii contents
3.2. Steady-state analysis . . . . . . . . . . . . . . . . . . . . . . . 54

3.2.1. Lossless medium . . . . . . . . . . . . . . . . . . . . . 55
3.2.2. Lossy medium . . . . . . . . . . . . . . . . . . . . . . 57
3.3. Transient emission into a lossless medium . . . . . . . . . . . 59
3.4. Reflection and transmission problem . . . . . . . . . . . . . . 62
3.4.1. Electric field analysis . . . . . . . . . . . . . . . . . . . 64
3.4.2. Magnetic field analysis . . . . . . . . . . . . . . . . . . 65
3.5. Shielding problem . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.1. Electric field analysis . . . . . . . . . . . . . . . . . . . 69
3.6. Parallel-plate waveguide . . . . . . . . . . . . . . . . . . . . . 72
4 Two-dimensional Electromagnetic Waves 81

4.1. Plane waves in a homogeneous medium . . . . . . . . . . . . 83
4.1.1. Uniform plane waves . . . . . . . . . . . . . . . . . . . 87
4.2. Interference of two plane waves . . . . . . . . . . . . . . . . . 90
4.2.1. Steady-state analysis: lossless case . . . . . . . . . . . 92
4.3. Reflection of a plane wave by an electrically impenetrable half-
space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.4. Reflection and transmission of a plane wave incident upon a
plane interface . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.4.1. Uniform plane waves in the frequency domain . . . . . 108
5 Electromagnetic Rays in a Two-dimensional Medium 121

5.1. Homogeneous, lossless medium . . . . . . . . . . . . . . . . . 122
5.2. Parallel polarization . . . . . . . . . . . . . . . . . . . . . . . 124
5.3. Perpendicular polarization . . . . . . . . . . . . . . . . . . . . 128
5.4. Ray trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.4.1. Ray trajectories in a horizontally layered medium . . . 133
5.4.2. Ray trajectories in a radially layered medium . . . . . 138
contents ix
6 Transmission Lines 145

6.1. TEM-waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
6.3. Coaxial line . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6.4. Propagation properties . . . . . . . . . . . . . . . . . . . . . . 155
6.4.1. Two-conductor transmission line . . . . . . . . . . . . 156
6.4.2. Lossless transmission line: steady-state analysis . . . . 158
6.4.3. Transients on lossless transmission lines . . . . . . . . 160
7 Electromagnetic Waveguides 167

7.2. Propagation properties of modes in a parallel-plate waveguide 178
7.3. Dielectric slab waveguides . . . . . . . . . . . . . . . . . . . . 182
7.4. Propagation properties of guided modes in a dielectric slab
waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
8 Excitation of Two-dimensional Electromagnetic Waves 203

8.1. The sheet emitter with a parallel electric current . . . . . . . 204
8.1.1. The far-field approximation . . . . . . . . . . . . . . . 207
8.2. The sheet emitter with a perpendicular electric current . . . . 213
8.2.1. The far-field approximation . . . . . . . . . . . . . . . 216
Answers to Exercises 223
Bibliography 241
Index 243
Chapter 1
Introduction
Electromagnetic wavefields have a wide range of applications: from com-

munication to medical treatment, from environmental sensing to energy ra-
diation.
When following an electromagnetic wave on its course, we start with its
excitation by an electromagnetic source. Some of them are natural sources
such as the sun and stars, others are artificial ones (a transmitting antenna,
a laser). Once it has been generated, the wave propagates along a certain
path from the source to the receiver. Depending on the properties of the
medium through which the wave passes, this propagation can lead to contin-
uous refraction by spatial changes in the medium parameters (for example,
the atmosphere), or to discontinuous refraction by an abrupt change in the
medium (for example, an interface between two different media). Finally,
the wave motion is picked up by an electromagnetic receiver (a receiving
antenna, an optical detector).
Each of these aspects is the subject of theoretical and experimental in-
vestigation. Usually, when the attention is focussed on a particular detail,
the remaining circumstances are chosen as simple as possible. For example,
when one wants to investigate the directional characteristics of a transmit-
ting antenna, the surrounding medium will be taken of the utmost simplicity,
as far as its electromagnetic properties are concerned, and of infinite extent.
When studying refraction phenomena during the propagation of an electro-
magnetic wave, the source will be taken a simple one, while the influence
2 introduction
of the receiver will be neglected at all. These simplifications are dictated

by the impossibility to take into account the influence of all parameters
simultaneously.
The basic laws of macroscopic electromagnetic theory were formulated
by James Clerk Maxwell and can be found in his famous book (Maxwell
1873). For a survey of the history of the subject the reader is referred to
Whittaker (Whittaker 1953). From the theory it follows that there exist
electromagnetic waves that travel with a finite speed which in vacuo seems
to be a universal constant, independent of the state of motion in which the
observer carries out his or her experiments. (The latter is not the case for
waves in matter.) Since through a wave motion with constant speed the
changes in position in space and the changes in time are interrelated in a
rigid manner, electromagnetic waves in vacuum can serve to interconnect the
space-time observations for two observers in relative motion. This concept
has led Einstein (Einstein 1956) to the theory of relativity. We shall
confine our analysis of electromagnetic waves to the case where the sources
that generate the wavefield, and the observer are at rest with respect to the
material media in the configuration.
As in any type of wave motion, the physical quantities that describe
the electromagnetic waves, depend on position and time. Their time depen-
dence in the domain where the source is acting is impressed by the excitation
mechanism of the source. The subsequent dependence on position and time
elsewhere is governed by propagation laws. The physical laws that underly
the properties of waves are induced from a series of basic standard experi-
ments. To carry out these experiments, an observer must be able to register
the position and the instant at which an observation is made. To register
the position the existence of an isotropic background space is preassumed.
In this space, distance can be measured along three mutually perpendicu-
lar directions with one and the same position- and orientation-independent
standard measuring rod. To register instants, the existence of a position-
and orientation-independent standard clock is preassumed. The standard
measuring rod is used to define, at a certain position which is denoted as
the origin O, an orthogonal Cartesian reference frame consisting of three
base vectors {i1 , i2 , i3 } that are of unit length each. The orientations of
these three base vectors form a mutually perpendicular, right-handed triad
(Fig. 1.1). (The property that each base vector specifies geometrically a
introduction 3
length and an orientation, makes it a vectorial quantity, or a vector; nota-

tionally, vectors will be represented by bold face symbols.) Let {x1 , x2 , x3 }
denote the three numbers that are needed to specify the position of an ob-
server, then the vectorial position of the observer x is the linear combination
(Fig. 1.2)
x = x1 i1 + x2 i2 + x3 i3 . (1.1)
i1
......
.......
....
...
...
...
...
...
...
....
.
.
...... ............
.
.
.
O
.
....
. . .......
...
.. .......
..
...
..... .......
. .......
....... .........
..
..............
.
. ...........
.
i2 i3
Figure 1.1. Standard measuring rod and Cartesian reference frame {O, i1 , i2 , i3 }
in three-dimensional space.
....
...
..
...
...
...
..
................
....
... .. .......
.. .. .......
.... ...
....... ...
.......
....... ... .......
...... ... .......
.......... ... .......
. ....... ... .......
. ....... .
.
.... ..
. .......
....... ..
. ..........
... ....... ... ....... .
. .......... ...... ....
. .... ....... ....... .
.... i1 .....
.
........
..
x
.......
....... .......
...... .
....
... .
. ... .......................... ...
.
. .... ...... .
. .
. ...
.... ............... .... ..
.
....
.... ..........
.... . ..
.
O
... ..
.
...
i2 ......
.
..........
...........
..
. .........
............. ....
. . ...
...
.
... ........
.. ....... i3 ... .........
....... ....... .
....
...... . .......
................
.
.....
.. ..
..
.... ... .
. ..
......
x1 ....
.......
.......
....
.........
...
.
....... ....... .... ....... .
..........
.
....... . .................
....... .... .......
.......
.......
....... ...... .......
x3 ....... ....
. x2 ......
....
....... .......
.......... ......
.
Figure 1.2. Cartesian coordinates {x1 , x2 , x3 }.

4 introduction
The numbers {x1 , x2 , x3 } are denoted as the orthogonal Cartesian coordi-

nates of the point of observation. The time coordinate is denoted by t.
One of the purposes of the basic standard experiments is to define the
units in terms of which the measured physical quantities are expressed. In ac-
cordance with international convention, we employ the International System
of Units (Système International d’Unités), abbreviated to SI, for expressing
the measured physical quantities. The mathematical framework by which
the results from the standard basic experiments are cast into the macroscopic
physical laws that govern the wave motion is furnished by vector calculus.
For this reason, the next section summarizes those properties of Cartesian
vectors that are needed in our further analysis.
1.1. Cartesian vectors and their properties
The mathematical framework of the theory of electromagnetic waves is

furnished by vector calculus. For this reason we summarize those properties
of Cartesian vectors that are needed in our further analysis.
1.1.1. Addition, subtraction and multiplication of vectors
Vectors can be subjected to the algebraic operations of addition, sub-

traction and multiplication. Let the components of v be given by v1 , v2
and v3 , and those of w by w1 , w2 and w3 , then the components of the sum
(difference) of v and w is given by
v ± w = (v1 ± w1 )i1 + (v2 ± w2 )i2 + (v3 ± w3 )i3 . (1.2)
The product of the scalar ϕ and the vector v is given by
ϕ v = ϕ v1 i1 + ϕ v2 i2 + ϕ v3 i3 . (1.3)
The scalar (dot) product of the vectors v and w is given by
v · w = v1 · w1 + v2 · w2 + v3 · w3 = w · v , (1.4)
cartesian vectors and their properties 5
The length of a vector v is denoted as

1 1
|v| = (v · v) 2 = (v12 + v22 + v32 ) 2 . (1.5)
The vector (cross) product of the vectors v and w is given by
v × w = (v2 w3−v3 w2 )i1 + (v3 w1−v1 w3 )i2 + (v1 w2−v2 w1 )i3 = −w × v , (1.6)
or in matrix notation

i
1 i2 i3

v × w = v1 v2 v3 . (1.7)

w1 w2 w3
The scalar triple product of three vectors u, v and w is given by
u · (v × w) = u1 (v2 w3 −v3 w2 ) + u2 (v3 w1 −v1 w3 ) + u3 (v1 w2 −v2 w1 ) , (1.8)
or in matrix notation

u
1 u2 u3

u · (v × w) = v1 v2 v3 . (1.9)

w1 w2 w3
The scalar triple product has the property
u · (v × w) = v · (w × u) = w · (u × v) (1.10)
= −w · (v × u) = −v · (u × w) = −u · (w × v) .
The vectorial triple product can be written as
u × (v × w) = (u · w)v − (u · v)w . (1.11)
As regards the differentiation of a vector, two cases have to be distin-

guished: differentiation with respect to a parameter, and differentiation with
respect to the spatial (Cartesian) coordinates of the space in which the vector
function is defined.
6 introduction
1.1.2. Differentiation with respect to a parameter
Let ϕ = ϕ(t) a scalar function and assume that ϕ is a differentiable

function of the parameter t (in electromagnetics often the time coordinate).
Then, the derivative ∂t ϕ = ∂ϕ/∂t is also a scalar function. Let v = v(t) be
a vector function and assume that v is a differentiable function of the para-
meter t. Let v1 = v1 (t), v2 = v2 (t), and v3 = v3 (t) denote the components
of v, then the derivative ∂t v of v is the vector
∂t v = (∂t v1 )i1 + (∂t v2 )i2 + (∂t v3 )i3 . (1.12)
Let ϕ = ϕ(t) be a differentiable scalar function of the parameter t and let

v = v(t) and w = w(t) be differentiable vector functions of the parameter
t, then we have the following differentiation rules:
∂t (ϕ v) = (∂t ϕ)v + ϕ∂t v , (1.13)
∂t (v × w) = (∂t v) × w + v × ∂t w . (1.14)
1.1.3. Differentiation with respect to the spatial coordinates
Let ϕ be a scalar function and assume that ϕ = ϕ(x) = ϕ(x1 , x2 , x3 ) is a

differentiable function of the spatial (Cartesian) coordinates x1 , x2 and x3 .
Then, the derivatives ∂1 ϕ = ∂ϕ/∂x1 , ∂2 ϕ = ∂ϕ/∂x2 and ∂3 ϕ = ∂ϕ/∂x3 are
also scalar functions. In this context, the gradient of ϕ = ϕ(x) is introduced
as
grad ϕ = ∇ϕ = (∂1 ϕ)i1 + (∂2 ϕ)i2 + (∂3 ϕ)i3 , (1.15)
where
∇ = i1 ∂1 + i2 ∂2 + i3 ∂3 (1.16)
is the operator of Hamilton, the so-called nabla operator or del operator.
This operator is a vector and acts as a spatial differentiation with respect to
the three spatial coordinates.
Let v be a vector function and assume that v = v(x) = v(x1 , x2 , x3 ) is a
differentiable function of the spatial (Cartesian) coordinates x1 , x2 and x3 .
The derivative ∂1 v is the vector
∂1 v = (∂1 v1 )i1 + (∂1 v2 )i2 + (∂1 v3 )i3 . (1.17)

Similarly, we have
∂2 v = (∂2 v1 )i1 + (∂2 v2 )i2 + (∂2 v3 )i3 , (1.18)
∂3 v = (∂3 v1 )i1 + (∂3 v2 )i2 + (∂3 v3 )i3 . (1.19)

These three derivatives operating on the vector function v can be combined
in the divergence operator, defined as
div v = ∇ · v = ∂1 v1 + ∂2 v2 + ∂3 v3 , (1.20)
and in the curl operator, defined as
curl v = ∇ × v = (∂2 v3 −∂3 v2 )i1 + (∂3 v1 −∂1 v3 )i2 + (∂1 v2 −∂2 v1 )i3 . (1.21)
We note that ∇ is a vector operator satisfying two sets of rules:
• vector rules;
• partial differentiation rules, including differentiation of a product.
We now summarize the rules for the differentiation with respect to the
spatial coordinates of the scalar functions ϕ = ϕ(x) and ψ = ψ(x), and of
the vector functions v = v(x) and w = w(x).
∇(ϕ + ψ) = ∇ϕ + ∇ψ , (1.22)
∇ · (v + w) = ∇ · v + ∇ · w , (1.23)
∇ × (v + w) = ∇ × v + ∇ × w , (1.24)
∇(ϕ ψ) = (∇ϕ) ψ + ϕ ∇ψ , (1.25)
∇ · (ϕ v) = (∇ϕ) · v + ϕ ∇ · v , (1.26)
∇ × (ϕ v) = (∇ϕ) × v + ϕ ∇ × v , (1.27)
∇ · (v × w) = (∇ × v) · w − v · (∇ × w) , (1.28)
∇ × (v × w) = (w · ∇)v − w ∇ · v − (v · ∇)w + v ∇ · w , (1.29)
∇(v · w) = w × (∇ × v) + (w · ∇)v + v × (∇ × w) + (v · ∇)w .
(1.30)
We note that we have assumed that the functions ϕ, ψ, v and w are differen-
tiable functions of the spatial coordinates. When we assume that ϕ = ϕ(x)
8 introduction
is also a twice differentiable function of x, we have the rules:
∇ · (∇ϕ) = (∇ · ∇)ϕ = (∂12 + ∂22 + ∂32 )ϕ , (1.31)

∇ × (∇ϕ) = 0 , (1.32)
∇ · (∇ × v) = 0 , (1.33)
∇ × (∇ × v) = ∇(∇ · v) − (∇ · ∇)v . (1.34)
Subsequently, we present the rules for the spatial differentiation of a spatially

dependent function f = f (|x|):
x
∇|x| = , (1.35)
|x|
∇|x|n = n|x|n−2 x , (1.36)
x
∇f (|x|) = ∂f (|x|) , (1.37)
|x|
where ∂f is the derivative of f with respect to its argument. Further, we
have:
∇·x = 3, (1.38)
∇×x = 0, (1.39)
(∇ · ∇)|x| n
= n(n + 1)|x| n−2
, (1.40)
and when a is a constant vector:
∇(a · x) = a , (1.41)
(a · ∇)x = a , (1.42)
(a × ∇) × x = −2a . (1.43)
Interpretation of grad ϕ
We consider a continuously differentiable scalar function ϕ = ϕ(x) and
we take the dot product of its gradient ∇ϕ (del ϕ) and an infinitesimal
increment of length
dx = dx1 i1 + dx2 i2 + dx3 i3 . (1.44)
Thus we obtain
(∇ϕ) · dx = (∂1 ϕ)dx1 + (∂2 ϕ)dx2 + (∂3 ϕ)dx3 = dϕ , (1.45)

∇ϕ .
....
.........
..............
.
... .
..
...
...
..
... .... ... .................. ....
.
.... ....... .
.......... ...... ...
........ ...... ..
....... ...
.
..........
.......
......
......
.
. ..
....... ..
.... ϕ(x) = C + ∆C
....... ...... ....
....... ...... ... ....
...................
....... ....... .............
........ . .. ............
.... ...
......... .... .............. ......
...
............
........ ......... .
............................................. ...
.. ........... ...........
... ..............................................................
..
.....
.
...
................................... ..
........... ...
..
. ......... ......
.. .....
.. ....... .
....... ...... ...
.
ϕ(x) = C
...... ......
....... ...... ...
..
....... ......
...... ..... ..............................
.......
..........
.
.......
......
...
s .
.. .
.
.........
dx
.................... ............................... ..............
............
........................
...............
...... . .................. ........
....... ..
..... .. ...........
............. P s
................... ...
..... . ............................ .............
..... .........................................
.....
.
......
.....
.
.....
..
Q
xP .......
.....
.....
.
...
.....
.....
.....
i1 ..
......... .
...
.....
.....
... ....
.....
... ..........
... .......
......
.
.... ....
....... ............ ..
i2 ..........
........... O ...........
...
i3
Figure 1.3. The gradient vector.
the change in the scalar function ϕ corresponding to a change in position

dx. Now consider P and Q to be two points on a surface ϕ(x1 , x2 , x3 ) =
constant. These points are chosen so that Q is a distance dx from P. Then,
moving from P to Q, the change in ϕ(x1 , x2 , x3 ) = constant is given by
dϕ = (∇ϕ) · dx = 0 , (1.46)
since we stay on the surface ϕ(x1 , x2 , x3 ) = constant. This shows that ∇ϕ is

perpendicular to dx. The vectorial distance dx may have any direction from
P as long as it stays in the surface ϕ = constant, point Q being restricted to
this surface. For vanishing dx, we observe that ∇ϕ is oriented in a direction
of the normal to the surface ϕ = constant (see Fig. 1.3).
If we now permit dx to take us from one surface ϕ = C, C being a
constant, to an adjacent surface ϕ = C + dC, then,
dϕ = dC = (∇ϕ) · dx . (1.47)
10 introduction
For a given dϕ, |dx| is a minimum when it is chosen parallel to ∇ϕ, or, for
a given |dx|, the change in the scalar function ϕ is maximized by choosing
dx parallel to ∇ϕ. This identifies ∇ϕ as a vector having the direction of
the maximum space rate of change of ϕ.
Very often the notion of directional derivative occurs. When τ is a unit
vector, the quantity τ · ∇ϕ is called the directional derivative of ϕ in the
direction of τ , and equals the rate of change of ϕ in the direction of τ , viz.,
τ · ∇ϕ = ∂τ ϕ = τ1 ∂1 ϕ + τ2 ∂2 ϕ + τ3 ∂3 ϕ . (1.48)
When τ is the tangent along a surface ϕ = constant, we obtain
τ · ∇ϕ = ∂τ ϕ = 0 , (1.49)
which is consistent with Eq. (1.46).
Interpretation of div v
We consider a continuously differentiable vector function v = v(x). The
divergence operator ∇ · v (del dot v) results in a scalar quantity indicating
the outflow of a vector field. It can be obtained from the limiting behavior
of the net outflow integral for a vanishing small enclosed volume. To show
this we first compute the net outflow of a vector field v over the elementary
domain with volume dV = dx1 dx2 dx3 at the center of the elementary domain
(see Fig. 1.4). This latter point is given by xP = { 12 dx1 , 12 dx2 , 12 dx3 }. By
Taylor’s theorem, the field component v1 is
v1 (x) = v1 (xP ) + (∂1 v1 )(x1 − 12 dx1 ) + (∂2 v1 )(x2 − 12 dx2 ) + (∂3 v1 )(x3 − 12 dx3 )
+ higher order terms . (1.50)
The surface integral of the normal component of v (in the direction of the
outward normal) over the top surface {x1 = dx1 , 0 < x2 < dx2 , 0 < x3 <
dx3 } of the volume element, shown in Fig. 1.4, is
dx2 dx3
v1 (dx1 , x2 , x3 ) dA =[v1 (xP ) + 12 (∂1 v1 ) dx1 ]dx2 dx3
x2 =0 x3 =0
+ higher order terms . (1.51)

i1
......
.........
... .... ..
...
....
..
...
...
...
...
...
dx1
. .
...
.........
.
.... .... ...................................................
..
.
.
.
.....
.
..
.... ...
.
.
..........................
..........................
............ ∂V .
.
.....
.
. . .... ......
..... .... ..... ... ......
..
. . ..... .......
..
....
. . .
.
.
. ... ...... ..............
..... .... .... ....
..... .....
............... .. ..... ...
.... ........................................................ ..... ...
... ... ........................... .
..
...... ...
........................... .....
... ... ......... ...
... ... .
. ...
... ... .... ...
... ... .
. ...
... ... ..
. ...
... .. .. ...
s
. .
... ... .... ....
... .
. .
. ..
...
...
...
xP ..
.
....
.
.
..
.
....
.
.
r ........................................
ν
..
... ....
... .... .... ...
... ... ... ...
... ............................
.
. .
.
. ...
... .
. . . ...
O ..... ......................... .
.
... .. ...... . ...
...
... ..
....
.
.
..
...... ................................ .
.
.
.
.
.
. .
. .. .
..... ...
..
.
dx3
............................
.........................
... .
.... .
. ..... ...................................
...
... .........
....
.....
.
.
.
.
....
. .
........
.....
.
. ........ .... i3
.... ....... ..
. ..
.....
.. .... ....
................................................... .....
..... ........................... ... .........
..
.
.
.....
... .....
................. dx2 ...........................
..................................
i2
Figure 1.4. Elementary domain in three-dimensional space.
The surface integral of the normal component of v (in the direction of the
outward normal) over the bottom surface {x1 = 0, 0 < x2 < dx2 , 0 < x3 <
dx3 } of the volume element, shown in Fig. 1.4, is
dx2 dx3
− v1 (0, x2 , x3 ) dA = − [v1 (xP ) − 12 (∂1 v1 ) dx1 ]dx2 dx3
x2 =0 x3 =0
+ higher order terms , (1.52)
the negative sign in front of the integral coming in because the outward
pointing component of v is integrated, and the outward pointing component
of v for the bottom surface is −v1 . The sum of the surface integrals over these
two faces is therefore simply ∂1 v1 dx1 dx2 dx3 , to the order of approximation
considered here. The contributions to the other faces depend on v2 and v3
and can be computed in a similar way. The net outflow integral from the
volume

element is therefore
⊂⊃ ν · v dA = (∂1 v1 + ∂2 v2 + ∂3 v3 )dx1 dx2 dx3 = ∇ · v dV , (1.53)
x∈∂V
12 introduction
in which ∂V denotes the boundary surface of the elementary domain, and ν

denotes the normal to ∂V and is oriented away from the elementary domain
V. The analogue of the divergence is the net outflow integral per unit volume
at the point x and is written as

⊂⊃ ν · v dA
div v = lim x∈∂V = ∂1 v1 + ∂2 v2 + ∂3 v3 = ∇ · v , (1.54)
V →0 V
in which
V = dV (1.55)
x∈V

is the volume of the domain V. The integral x∈∂V ν · v dA is called the
flux of the vector field v through the closed surface ∂V.
Gauss’ integral theorem

Let v be a continuously differentiable vector function of position de-
fined in a bounded domain D. Let, further, ∂D denote the boundary of
D (Fig. 1.5). Because of the additive property, the net outflow integral for
the domain D must equal the sum of the outflow integrals for all elemen-
tary domains included in the domain D. Then, we arrive at Gauss’ integral
theorem, viz.,
∂D
ν
Figure 1.5. Configuration for the application of Gauss’ integral theorem.


⊂⊃ ν · v dA = ∇ · v dV = div v dV , (1.56)
x∈∂D x∈D x∈D
in which ν is the unit vector normal to ∂D and oriented away from D.

Two modifications of Gauss’ integral theorem are obtained as follows.
With v = a ϕ, in which a is an arbitrary constant vector and ϕ is a contin-
uous differentiable scalar function, we obtain

⊂⊃ ν ϕ dA = ∇ϕ dV = grad ϕ dV , (1.57)
With v = a × w, in which a is an arbitrary constant vector and w is a

continuous differentiable vector function, we obtain

⊂⊃ ν × w dA = ∇ × w dV = curl w dV . (1.58)
Interpretation of curl v
We consider a continuously differentiable vector function v = v(x). The
curl operator ∇ × v (del cross v) results in a vectorial quantity indicating
the rate of circulation of a vector field. It can be obtained from the limiting
behavior of the net circulation integral around a vanishing small surface
area. To show this we first consider the case that the elementary area is
perpendicular to the i1 -axis. The center point of the area is given by xP =
{ 12 dx1 , 12 dx2 , 12 dx3 }. The circulation integral for the path shown in Fig. 1.6
is
dx2 dx3
τ · vdl = v2 ( 12 dx1 , l, 0) dl + v3 ( 12 dx1 , dx2 , l) dl
x∈∂A x2 =0 x3 =0
dx2 dx3
− v2 ( 12 dx1 , l, dx3 ) dl − v3 ( 12 dx1 , 0, l) dl ,
x2 =0 x3 =0
(1.59)
14 introduction
i3
.....
........
..........
...
....
..
...
...
...
...
...
... τ ..
∂A
.....
............................
dx3 ....
.. .. .....
.....
.....
... ....
.
...
...
...
...
...
...
...
... .... ..
r xP
........
....... ..
.. ......
... .......
..... .. ... ..
..
...
...
...
...
...
...
...
... .. ..
.............................................................................................................................................................................................................................
i2
O
.. ..
dx2
Figure 1.6. Net circulation integral.
in which τ is the unit vector along the path and l is the arclength along the
path. Substituting the Taylor’s series expansion (see Eq. (1.50)) for v2 and
v3 we find

τ · vdl = v2 (xP ) dx2 − 12 (∂3 v2 )dx2 dx3 + v3 (xP ) dx3 + 12 (∂2 v3 )dx2 dx3
x∈∂A
−v2 (xP ) dx2 − 12 (∂3 v2 )dx2 dx3 − v3 (xP ) dx3 + 12 (∂2 v3 )dx2 dx3
= (∂2 v3 − ∂3 v2 )dx2 dx3 (1.60)
plus higher order terms. Subsequently, the net circulation integral for the
elementary surface area perpendicular to the i2 -axis is (∂3 v1 − ∂1 v3 )dx3 dx1
and the net circulation integral for the elementary surface area perpendicular
to the i3 -axis is (∂1 v2 − ∂2 v1 )dx1 dx2 . Hence, the analogue of the curl is the
net circulation integral per unit of surface area at the point x and is written
as
τ · v dl
ν · curl v = lim x∈∂A = ν · (∇ × v) , (1.61)
A→0 A
in which
A= dA (1.62)
x∈A
is the area of the surface A and ∂A is the closed boundary of the surface
area A. In Eq. (1.61), ν is the unit vector normal to the surface area A and
oriented toward the side of advance of a right-hand screw as it is turned in
the direction of τ around ∂A.
Stokes’ integral theorem

Let v be a continuously differentiable vector function of position defined
on a bounded surface S. Let, further, ∂S denote the closed boundary of S
(Fig. 1.7). We divide the surface S into its elements of area dA and add
all the net circulation integrals for the elementary areas. Then, we arrive at
Stokes’ integral theorem, viz.,

τ · v dl = (ν × ∇) · v dA = ν · curl v dA , (1.63)
x∈∂S x∈S x∈S
∂S
Figure 1.7. Configuration for the application of Stokes’ integral theorem.

16 introduction
in which ν is the unit vector normal to the surface area S and is oriented to-
ward the side of advance of a right-hand screw as it is turned in the direction
of τ around ∂S.
Two modifications of Stokes’ integral theorem are obtained as follows.
With v = a ϕ, in which a is an arbitrary constant vector and ϕ is a contin-
uous differentiable scalar function, we obtain

τ ϕ dl = (ν × ∇)ϕ dA = ν × grad ϕ dA , (1.64)
x∈∂S x∈S x∈S
With v = a × w, in which a is an arbitrary constant vector and w is a
continuous differentiable vector function, we obtain

τ × w dl = (ν × ∇) × w dA . (1.65)
x∈∂S x∈S
introduction 17
1.2. Exercises and problems
Exercise 1.1
Calculate the following expressions and eliminate as many vector products
as possible
(a) a × a
(b) (a × b) · c − a · (b × c)
( c ) (a × b) × c
(d) (a × b) · (c × d)
( e ) (a × b) × (c × d).
Exercise 1.2
What is the direction of the vector c = a × b with respect to the vectors a
and b? Calculate (a × b) · a and (a × b) · b.
Exercise 1.3
A two-dimensional rectangle in the plane x3 = 0, with dimensions d1 , d2 ,
can said to be spanned by two vectors along the two base vectors of the
Cartesian reference frame, viz., d1 = d1 i1 , d2 = d2 i2 . Let the unit normal
ν point in the positive x3 -direction. Give a geometrical interpretation of
A = ν · (d1 × d2 ).
Exercise 1.4
A three-dimensional brick in space, with dimensions d1 , d2 , d3 , can said to be
spanned by three vectors along the Cartesian reference frame base vectors,
viz., d1 = d1 i1 , d2 = d2 i2 , d3 = d3 i3 . What is the geometrical meaning of
the scalar triple product V = d1 · (d2 × d3 )?
Exercise 1.5
Sketch the following vector fields by selecting several points and by drawing
at each of these points an arrow with a direction that corresponds to the
direction of the vector field and with a length that is proportional to the
magnitude of the vector field
(a) v(x) = −x1 i1
(b) v(x) = x2 i1
18 introduction
( c ) v(x) = −x1 i1 − x3 i3
(d) v(x) = x1 i3 − x3 i1 .
Exercise 1.6
Determine the divergence and the rotation of the vector fields in Exercise
1.5.
Exercise 1.7
The operator ∇ · ∇ in Eq. (1.31) is a scalar operator that occurs often
in physics. It has therefore been given a name; it is called the Laplacian
(operator) and is often denoted as ∇2 . Given the at least twice differentiable
vector function A(x), write ∇2 A out in components.
Exercise 1.8
Show that if an arbitrary vector function E(x) can be written as E = −∇V ,
where V (x) is a differentiable scalar function, the vector field E is curl free,
i.e. irrotational, and hence ∇ × E = 0.
Exercise 1.9
Show that if an arbitrary vector function D(x) can be written as D =
∇ × C, where C(x) is a differentiable vector function, the vector field D(x)
is divergence free, i.e. solenoidal, and hence ∇ · D = 0.
Exercise 1.10
The distance between two points with position vectors x and x is given
by the length of the vector from x to x , d(x, x ) = |x − x|. When ∇
and ∇ denote partial differentiation with respect to x and x , respectively,
determine
(a) ∇d , ∇ d
(b) ∇d−1 , ∇ d−1
( c ) ∇ · ∇d , ∇ · ∇ d
(d) ∇ · ∇d−1 , ∇ · ∇ d−1 .
Exercise 1.11
Given the smooth surface S with unit normal ν, and a vector S = E × H.
Show that the normal component ν · S of S can also be written as ν · S =
ν · [(ν × E) × (ν × H)].
exercises and problems 19
Exercise 1.12
Given the smooth surface S with unit normal ν. Show that an arbitrary
vector H can be decomposed into H = (ν · H)ν + (ν × H) × ν. Give a
geometrical interpretation of the two terms.
Problem 1.1
Calculate the volume of a three-dimensional parallepiped spanned by the
arbitrarily oriented vectors a, b and c.
Problem 1.2
Given the smooth surface S with unit normal ν. Decompose the operator
∇ in two parts, one part tangential and the other normal to the surface S.
Problem 1.3

Use Gauss’ theorem to calculate x∈∂D ν dA, where ν is the outward unit
normal to the closed boundary ∂D of the domain D (see Fig. 1.5).
Problem 1.4

Use Stokes’ theorem to calculate x∈∂S τ dl, where τ denotes the unit tan-
gent along the closed boundary contour ∂S of the surface S. The integration
runs in the direction of circulation that forms a right-handed system with
the unit normal ν to S (see Fig. 1.7).
Chapter 2
The Electromagnetic Field

Equations
The electromagnetic field equations take on their simplest form in a vac-

uum domain. In such a domain, they follow, in principle, from a series of
basic experiments in which an electric point charge is employed as measuring
device, and the force exerted on it is measured. Through this force, the elec-
tric and the magnetic field strengths, as experienced by an observer located
in a vacuum domain, are introduced. The electromagnetic field equations in
matter are introduced through an axiomatic procedure that leaves in tact
their structure in a vacuum domain in case the point of observation is located
in a vacuum domain. Since external sources are essentially composed of mat-
ter, they are encompassed in the volume densities of electric and magnetic
currents that describe the electromagnetic action of matter. The induced
parts of the latter quantities then describe the passive reaction of a piece of
matter to an electromagnetic field.
2.1. Force exerted on an electric point charge
From experience it is known that electrically charged particles exert forces

on each other. These forces depend on the relative position and the relative
22 the electromagnetic field equations
state of motion of the charged particles. We take a single electrically charged

particle and use this as a measuring device for defining, in a vacuum domain,
the electric and the magnetic field strengths of an electromagnetic field. It is
assumed that an observer who is handling this measuring device can measure
the force exerted on it (for example, by counterbalancing the force by an
adjustable mechanical one). The experiments show the following:
• the force F is proportional to the strength q of the point charge;

• the force contains a term that is independent of the velocity v of the
point charge with respect to the observer;
• the force contains a term that is proportional to the velocity of the
point charge with respect to the observer and has a direction perpen-
dicular to it.
In accordance with this, and with the conventions of the International Sys-
tem of Units, we postulate the expression (see Fig. 2.1)
F = qE + qµ0 v × H . (2.1)
In this expression
F = force (N),
q = electric charge (C),
v = velocity of point charge with respect to the observer (m/s),
µ0 = permeability in vacuum (H/m),
E = electric field strength (V/m),
H = magnetic field strength (A/m).
The value of µ0 is fixed by SI as µ0 = 4π × 10−7 H/m. Assuming that the

value of q is so small that the reaction of the test point charge on the sources
that generate the electromagnetic field can be neglected, Eq. (2.1) defines
the value of the electric field strength E = E(x, t) and the value of the
magnetic field strength H = H(x, t) at the position of observation x with
respect to a given, fixed, orthogonal Cartesian reference frame and the time
of observation t. Since the observer must be free to position and to move his
or her measuring device, the operational definition of Eq. (2.1) can only be
applied in a vacuum domain and cannot be used in matter.
force exerted on an electric point charge 23
........
........
......
F
......
.
..
........
.
......
......
......
......
.. .........
.
......
......
......
......
......
i1 x u ......
......
.
......
.
......
....
......... ..
...
.
.....
.......... ........................................................... v
.......
...
..
... .......
....... q
......
... ......
... ......
... . .
...... .
... ......
.......
... .......
... .......
... .. ......
.
... .....
......
.. .......
.........
.
...
........ .............
...... .......
.......
.......
.......
O .......
.......
.......
...
.........
. .......
.......
. . .
...
...... .........
.
........... ............
i2 i3
Figure 2.1. Force F on an electric point charge of strength q and moving with
velocity v with respect to an observer at rest in the reference frame {0, i1 , i2 , i3 }.
By first choosing, in Eq. (2.1), v = 0, one determines by measuring F the

value of E = E(x, t). By giving, subsequently, v three linearly independent
directions, one can determine H = H(x, t) by measuring F and using the
already known values of E. The quantity
F = qE (2.2)
is sometimes denoted as the electric force; the quantity
F = qµ0 v × H (2.3)
is sometimes denoted as the magnetic force or Lorentz force (after H.A.

Lorentz).
It is emphasized that the thus determined values of E and H have the
meaning of electric and magnetic field strengths, respectively, only for the
observer who carries out the measurements and interprets them according
to Eq. (2.1). A second observer who is in relative motion with respect to
the first one, measures different values of the electric and the magnetic field
strengths. In particular, if the second observer moves with the point charge
of the first, we have, denoting the latter’s quantities by primed symbols,

v = 0, and hence, F = qE , i.e., the second observer interprets the present
electromagnetic field as a pure electric one. (In accordance with the theory
of relativity, we have put q = q.) The interrelations that exist between the
values of the electric and the magnetic field strengths of two observers who
are in relative motion are investigated in the theory of relativity.
2.2. Maxwell’s equations in vacuum
Using the electric point charge as a measuring device one can investigate
the properties of an electromagnetic field in a vacuum domain. In particular,
it proves to be fruitful to investigate the relationships between the changes of
the field in space and the changes of the field in time. Quantitatively, this is
most simply done by mutually comparing the first-order partial derivatives
with respect to the spatial coordinates with the first-order partial derivatives
with respect to time. Carrying out the analysis, it is found that
−∇ × H + ε0 ∂t E = 0 , (2.4)
∇ × E + µ0 ∂t H = 0 . (2.5)
In these equations
E = electric field strength (V/m),

H = magnetic field strength (A/m),
ε0 = permittivity in vacuum (F/m),
µ0 = permeability in vacuum (H/m).
The value of ε0 follows from

1
ε0 = , (2.6)
µ0 c20
where c0 = 299792458 m/s is the electromagnetic wave speed in vacuum.
Substituting the values of µ0 and c0 , one obtains ε0 = 8.8541878 × 10−12 .
Maxwell’s equations in matter 25
Equations (2.4) and (2.5) are known as Maxwell’s equations (in vacuum).
Further, it is found that
∇·E = 0, (2.7)
∇·H = 0. (2.8)
The latter equations are not independent of Maxwell’s equations and are a
kind of compatibility relations. By noting that ∇ · (∇ × H) = 0, it follows
from Eq. (2.4) that ∂t ∇ · E = 0. Hence, if at some instant ∇ · E = 0, this
would be so at any instant. In particular, this applies when we study the
causal fields that are generated by sources that are switched on at a certain
instant, starting from an initially vanishing field. The same reasoning applies
to Eq. (2.5), from which, since ∇ · (∇ × E) = 0, it follows that ∂t ∇ · H = 0.
2.3. Maxwell’s equations in matter
Methodologically the simplest way to account for the presence of matter

is to retain the vectorial nature of the electromagnetic field equations and
introduce in the right-hand sides of Eqs. (2.4) and (2.5) vectorial quantities
that differ from zero only in a domain occupied by matter (and necessarily
reduce to zero in a vacuum domain). Let us write (Fig. 2.2)
−∇ × H + ε0 ∂t E = −J mat , (2.9)
∇ × E + µ0 ∂t H = −K mat
. (2.10)
Here,
J mat = volume density of material electric current (A/m2 ),

K mat = volume density of material magnetic current (V/m2 ).
As we have said,
J mat = K mat = 0 in a vacuum domain . (2.11)

.................................
.............. .......
.........
........ .....
...
......... .....
...
...
...... ...
........ ...
...
..... ...
...
....
J mat
..
.. ..
...
.... ...
.....
. .

3 ..
.
...
..
. .
.
...

.. .
..
... ..
s
. .
... ...
... ...
...
...
.
...
... @ ...
...
...
vacuum
...
... @ ...
...
...
...
...
R K mat
@ ...
....
... ...
...
... ....
..... ..
..
....
matter ...
.....
.....
.
... .....
... ...
......
... ......
... .....
... ......
......
...
..... ....
........
...... .....
........ ........
..................................................
Figure 2.2. Piece of matter embedded in vacuum, with volume density of electric
current J mat and volume density of magnetic current K mat .
First of all, we shall distinguish in J mat and K mat between an active

part and a passive part. The active parts (also denoted as the source parts,
or external parts) describe the action of sources that generate the field and
whose physical behavior is either irrelevant to or outside the scope of our
analysis; they will be denoted as J ext and K ext , respectively. The active
parts are taken to be field independent. The passive parts (also denoted
as the induced parts) describe the reaction of matter to the presence of an
electromagnetic field and are typically field dependent; they will be denoted
by J ind and K ind , respectively. Hence, we have
J mat = J ind + J ext , (2.12)
mat ind ext
K = K +K . (2.13)
The term with K ext has been introduced for symmetry and convenience
reasons. It is noted that K ext has no physical meaning.
Traditionally, the induced parts are written in a somewhat different man-
ner, viz.
J ind = J + ∂t P , (2.14)
ind
K = µ0 ∂t M , (2.15)
Maxwell’s equations in matter 27
where
J = volume density of electric current (A/m2 ),

P = electric polarization (C/m2 ),
M = magnetization (A/m).
The latter quantities are particularly useful for describing the electromag-
netic properties of matter in case only fields varying slowly in time are
present. Substitution of Eqs. (2.12) - (2.15) in Eqs. (2.9) and (2.10) leads to
−∇ × H + ε0 ∂t E + J + ∂t P = −J ext , (2.16)
∇ × E + µ0 ∂t H + µ0 ∂t M = −K ext
. (2.17)
Further, it is customary to introduce the quantities
D = ε0 E + P , (2.18)
B = µ0 (H + M ) . (2.19)
where
D = electric flux density (C/m2 ),

B = magnetic flux density (T).
With the aid of Eqs. (2.18) and (2.19), Eqs. (2.16) and (2.17) can be rewrit-
ten as
−∇ × H + J + ∂t D = −J ext , (2.20)
∇ × E + ∂t B = −K ext
. (2.21)
Equations (2.20) and (2.21) are known as Maxwell’s equations (in matter).
These equations constitute an incomplete system of equations since the num-
ber of equations is less than the number of unknown quantities, assuming
that the right-hand sides, which are representative of the action of exter-
nal sources, are known. The supplementing equations are furnished by the
constitutive relations that describe the reaction of passive matter to the pen-
etration of an electromagnetic field.
As far as the compatibility relations are concerned, we observe that ap-
plication of the divergence operator to Eqs. (2.20) and (2.21) yields
∇ · J + ∂t ∇ · D = −∇ · J ext , (2.22)
∂t ∇ · B = −∇ · K ext
. (2.23)
Historically, the volume density of electric charge (in C/m3 ) is introduced

as
ρ=∇·D. (2.24)
2.4. The constitutive relations
The electromagnetic constitutive relations are representative for the

macroscopic electromagnetic properties of matter. In their general form,
they constitute an equivalent of three vectorial relations between the five
vectorial quantities {J , D, B, E, H}. For reasons that are connected with
the transfer of energy by electromagnetic wavefields, the standard form ex-
presses {J , D, B} in terms of {E, H}. Several terminological aspects of this
relationship are enumerated below. In view of the assumed passivity of the
medium, we require that for any type of matter we have {J , D, B} → 0 as
{E, H} → 0
When the values of {J , D, B} are linearly related to the values of {E, H},
we denote the medium as linear. If this is not the case, the medium is denoted
as non-linear.
When the operators that express the values of {J , D, B} in terms of
the values of {E, H} are time invariant, the medium is denoted as time
invariant. Otherwise, the medium is time variant or parametrically affected.
When the constitutive relations express the values of {J , D, B} at some
instant in terms of the values of {E, H} at the same instant only, the medium
is denoted as instantaneously reacting. When, on the other hand, the values
of {J , D, B} are expressed in terms of the values of {E, H} at all previous
the constitutive relations 29
instants, the medium is said to show relaxation; the property that only
the past is involved in relaxation phenomena, is known as the principle of
causality.
When the values of {J , D, B} at some position are related to the values of
{E, H} at the same position only, the medium is denoted as locally reacting.
Almost all media are locally reacting in their electromagnetic behavior.
If at a point in space the constitutive operators are orientation invariant,
the medium is denoted as isotropic at that point. If this property does not
apply, the medium is denoted as anisotropic.
In a domain in space where the constitutive operators are shift invariant,
the medium is denoted as homogeneous; in a domain in space where the
shift invariance does not apply, the medium is denoted as inhomogeneous or
heterogeneous.
For a wide class of materials, the quantities J and D only depend on E
(and not on H), while the quantity B only depends on H (and not on E).
For a medium that is, in addition, linear, time invariant, instantaneously
reacting, locally reacting and isotropic in its electromagnetic behavior, we
then have
J (x, t) = σ(x)E(x, t) , (2.25)

D(x, t) = ε(x)E(x, t) , (2.26)
B(x, t) = µ(x)H(x, t) , (2.27)
where
σ = conductivity (S/m),
ε = permittivity (F/m),
µ = permeability (H/m).
For electromagnetic fields that vary relatively slowly in time (frequency of

operation less than a few MHz), the assumption of an instantaneously re-
acting material is realistic. In a (sub)domain where these constitutive co-
efficients σ = σ(x), ε = ε(x) and µ = µ(x) change indeed with position,
the medium is inhomogeneous; in a domain where they are constant, the

medium is homogeneous.
In vacuum we have
σ = 0,
1
ε = ε0 = ,
µ0 c20
µ = µ0 .
In the tables of physical constants (see Table 2.1) one customarily spec-
ifies the dielectric properties of a material through
εr = ε/ε0 = relative permittivity
and its magnetic properties through
µr = µ/µ0 = relative permeability;
εr and µr are dimensionless.
Table 2.1. Relative permittivities and conductivities for

electromagnetic fields that vary slowly in time (at 20 ◦ C)
εr = ε/ε0 σ (S/m)
air 1.0006
glass 6 1.0 × 10−12
quartz 4.6 1.0 × 10−12
bakelite 4.8 1.0 × 10−9
dry earth 5 10−4
wet earth 10 10−2
fresh water 81 10−3
seawater 81 4
salt water (20%) 81 20
aluminium 1 3.5 × 107
copper 1 5.8 × 107
silver 1 6.17 × 107
the system of field equations 31
2.5. The system of field equations
Substitution of the constitutive relations in Maxwell’s equations (2.20)

and (2.21) yields a system of two vectorial equations in two vectorial un-
knowns, viz., the electric field strength and the magnetic field strength.
These electromagnetic field equations are
−∇ × H + σE + ε∂t E = −J ext , (2.28)

∇ × E + µ∂t H = −K ext
. (2.29)
Let the subscripts {1, 2, 3} denote the components of a vector along the
{i1 , i2 , i3 }-axes of the chosen reference frame, respectively. Then the elec-
tromagnetic field is governed by the six scalar equations for the three compo-
nents of the electric field strength and the three components of the magnetic
field strength. This system of electromagnetic field equations is
−(∂2 H3 − ∂3 H2 ) + σE1 + ε∂t E1 = −J1ext , (2.30)
−(∂3 H1 − ∂1 H3 ) + σE2 + ε∂t E2 = −J2ext , (2.31)
−(∂1 H2 − ∂2 H1 ) + σE3 + ε∂t E3 = −J3ext , (2.32)
∂2 E3 − ∂3 E2 + µ∂t H1 = −K1ext , (2.33)

∂3 E1 − ∂1 E3 + µ∂t H2 = −K2ext , (2.34)
∂1 E2 − ∂2 E1 + µ∂t H3 = −K3ext , (2.35)
and is amenable to a mathematical solution procedure, either analytically
or numerically.
The transfer of electromagnetic power is governed by the Poynting vector
S = E × H (in W/m2 ) , (2.36)
where the three components are defined as

S1 = E2 H3 − E3 H2 , (2.37)
S2 = E3 H1 − E1 H3 , (2.38)
S3 = E1 H2 − E2 H1 . (2.39)
The Poynting vector quantifies the amount of electromagnetic power flow

per unit area.
2.6. The boundary conditions
In those domains in a medium where the constitutive parameters change

continuously with position, the electromagnetic field vectors are continu-
ously differentiable functions of position and satisfy the differential equations
(2.28) and (2.29). Across certain boundary surfaces in a configuration, the
electromagnetic field quantities may exhibit a discontinuous behavior. Since
at those positions they are no longer continuously differentiable, the electro-
magnetic field equations cease to hold. To interrelate the electromagnetic
wavefield quantities at either side of the interface, a certain set of boundary
conditions is needed.
Upon crossing the interface of two adjacent media that differ in their
electromagnetic properties, the electric and the magnetic field strengths are,
in general, no longer continuously differentiable in a domain that contains
(part of) an interface, and Eqs. (2.28) and (2.29) cease to hold. Assum-
ing that the properties of the media under consideration and the position
of the interface are time invariant, the non-differentiability is restricted to
the dependence on the spatial variables. To solve electromagnetic wavefield
problems in domains that contain abrupt boundaries, the electromagnetic
field equations must be supplemented by conditions that interrelate the field
values at either side of the interface, the so-called boundary conditions.
The interface of two media

Let ∂D denote the interface and assume that ∂D has everywhere a unique
tangent plane. Let, further, ν denote the unit vector along the normal to ∂D
such that upon traversing ∂D in the direction of ν, we pass from the domain
D(2) to the domain D(1) , D(1) and D(2) being located at either side of ∂D
(Fig. 2.3). Suppose, now, that some (or all) electromagnetic wave quantities
jump across ∂D. In the direction parallel to ∂D, all electromagnetic wave
quantities still vary in a continuously differentiable manner, and hence the
the boundary conditions 33
partial derivatives parallel to ∂D give no problem in Eqs. (2.28) and (2.29).

The partial derivatives perpendicular to ∂D, on the contrary, meet functions
that show a jump discontinuity across ∂D; these give rise to surface Dirac
delta distributions (surface impulses) located on ∂D. Distributions of this
kind would, however, physically be representative of surface sources located
on ∂D. In the absence of such surface sources, the absence of surface impulses
in the partial derivatives perpendicular to ∂D should be enforced. The latter
is done by requiring that these normal derivatives only meet functions that
are continuous across ∂D. To investigate the consequences of this reasoning,
we write
∇ = (ν · ∇)ν + [∇ − (ν · ∇)ν] = (ν · ∇)ν + (ν × ∇) × ν , (2.40)
in which (ν · ∇)ν is the normal part of the derivatives in the ∇ operator,

while (ν × ∇) × ν represents the tangential part. Considering Eqs. (2.28)
and (2.29), and assuming that there are no external sources present at the
interface ∂D, the quantities ∇ × H and ∇ × E may not contain any Dirac
distribution at ∂D, when crossing the interface in the normal direction. This
means that in the expressions
(ν · ∇)ν × H = (ν · ∇)(ν × H) , (2.41)

(ν · ∇)ν × E = (ν · ∇)(ν × E) , (2.42)
...
.........
∂D
........
.......
.......
D(1) .
......
......
..
...
.
......
......
ν .
.....
.
......
.
D(2)
.....
.....
@
I
@ ..
..
.
.....
.....
.....
.....
.....
.....
.
.....
....
....
...
.....
...
...
...
...
....
.
...
...
...
...
...
...
...
....
.
..
...
...
..
Figure 2.3. Interface between two media with different electromagnetic properties.
the quantities ν × H and ν × E may not jump, when crossing ∂D in the

normal direction. Hence, this reasoning leads to the requirements
ν × H is continuous across ∂D , (2.43)
and
ν × E is continuous across ∂D . (2.44)
Equations (2.43) and (2.44) are the boundary conditions at a sourcefree in-
terface between two different media. In these equations only the components
of the electric and magnetic field strengths tangential to ∂D occur.
In view of the continuity of the tangential components of E and H across
the interface ∂D, the normal component of the Poynting vector S, ν · S =
ν · (E × H) = ν · [(ν × E) × (ν × H)], is continuous across this interface,
as it should be on physical grounds.
Electrically impenetrable object

A material body, occupying the domain D in space, is denoted as elec-
trically impenetrable if in it the electric field strength is negligibly small,
while the boundary condition of the continuity of the electric field strength
across the boundary surface ∂D of the object is maintained. Consequently,
the boundary condition upon approaching the boundary surface ∂D of such
a body via its exterior is given by (Fig. 2.4)
lim ν × E(x + δν, t) = 0 for any x ∈ ∂D , (2.45)
δ↓0
where ν is the unit vector along the normal to ∂D pointing away from D.
It is not allowed to prescribe boundary conditions for the tangential part of
the magnetic field strength in this case. In fact, the tangential part of the
magnetic field strength will, in general, exhibit a discontinuity upon crossing
∂D. Electrically impenetrable materials arise as limiting cases of materials
whose conductivity and/or permittivity go to infinity.
In view of the vanishing tangential components of E at the boundary
∂D, the normal component of the Poynting vector S, ν · S = H · (ν × E),
vanishes at this boundary as well.
the boundary conditions 35
ν ∂D..................................
.............................................
......
.....
@
I .. .
......
...
...
...
@ ...
..... ...
*
..
.
.... ...
x + δν .... ..
....
@ .
...... ..
1 ...
. ..
.....
. ..
D
. ..

x ..
.
....
.
.....
....
.
.
. .

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

...

... ..
..
...
....
...
.
..
... ...
...
O ...
..
.
...
impenetrable object ...
......
.
... ...
... ...
..... ..
.....
....
.... ....
... .....
.....
...
... ........
.
... ......
......
... ......
... ......
....
..... .
...
........
....... ......
.......... .........
.......................................
Figure 2.4. Limiting procedure approaching the boundary

of an impenetrable object.
Magnetically impenetrable object

A material body, occupying the domain D in space, is denoted as mag-
netically impenetrable if in it the magnetic field strength is negligibly small,
while the boundary condition of the continuity of the magnetic field strength
across the boundary surface ∂D of the object is maintained. Consequently,
the boundary condition upon approaching the boundary surface ∂D of such
a body via its exterior is given by (Fig. 2.4)
lim ν × H(x + δν, t) = 0 for any x ∈ ∂D , (2.46)

δ↓0
where ν is the unit vector along the normal to ∂D pointing away from D.
It is not allowed to prescribe boundary conditions for the tangential part
of the electric field strength in this case. In fact, the tangential part of the
electric field strength will, in general, exhibit a discontinuity upon crossing
∂D. Magnetic impenetrable materials arise as limiting cases of materials
whose permeability goes to infinity.
In view of the vanishing tangential components of H at the boundary
∂D, the normal component of the Poynting vector S, ν · S = −E · (ν × H),
vanishes at this boundary as well.
2.7. Frequency-domain representations
In reality, any electromagnetic field is transient in nature: it is generated

by a source that has been switched on at some instant t = t0 in the finite
past and decays in magnitude as t → ∞. An essential feature is further the
physical condition of causality that requires that an effect (the presence of
an electromagnetic field) is causally related to its cause (the action of the
source). This implies that the field that is causally related to the action
of a source that is switched on at the instant t = t0 , necessarily vanishes
prior to this instant, i.e., for t < t0 , at any point in space. In time-invariant
configurations the behavior of such transient fields can advantageously be
described via their (complex) frequency-domain representations. Further, t0
may be chosen arbitrarily, hence, we simply take t0 = 0. Through these
representations the principle of causality can automatically be taken into
account, while the dependence on t (the fourth coordinate in space-time) is
replaced by the dependence on a (transform) parameter; the resulting reduc-
tion in dimensionality of the problem is often advantageous. The Laplace
transformation combines the two features.
Let f = f (x, t) denote an electromagnetic field quantity that is causally
related to the action of a source that has been switched on at the instant
t = 0, then its Laplace transform (with respect to time) is defined as
∞
fˆ(x, s) = exp(−st)f (x, t)dt for Re(s) > s0 , (2.47)
t=0
where s is the complex Laplace transform parameter and fˆ is analytic (i.e.,

differentiable with respect to s) in the right half of the complex s-plane
Re(s) > s0 . An important property of the Laplace transformation is that
the Laplace transform of ∂t f (x, t) is sfˆ(x, s) if f (x, t) is causal.
Once fˆ = fˆ(x, s) has been evaluated, f = f (x, t) can be recovered by
the application of the Bromwich inversion integral
s0 +j∞
1 ˆ s)ds for all t ,
f (x, t) = exp(st)f(x, (2.48)
2πj s=s0 −j∞
that automatically yields the value zero for t < 0 and for which numeri-
cal (Fast Fourier Transform) routines are available. In Eq. (2.48), j is the
imaginary unit (j 2 = −1).
frequency-domain representations 37
In a number of cases one can take s0 = 0 and consider fˆ = fˆ(x, s) for

imaginary values s = jω, where ω is the (real) angular frequency. Then,
Eq. (2.47) can be rewritten as
∞
fˆ(x, jω) = exp(−jωt)f (x,t)dt for all real ω , (2.49)
t=0
and Eq. (2.48) as

∞
1
f (x, t) = exp(jωt)fˆ(x, jω)dω for all t. (2.50)
2π ω=−∞
Equations (2.49) and (2.50) express the Fourier transformation for causal
fields.
Note that the length of a complex vector v̂ with components v̂1 , v̂2 and
v̂3 is defined as
|v̂| = (v̂ · v̂ ∗ ) 2 = (v̂1 v̂1∗ + v̂2 v̂2∗ + v̂3 v̂3∗ ) 2 ,

1 1
(2.51)
which is in accordance with the definition of Eq. (1.5) for real vectors. The
asterisk denotes complex conjugate.
2.7.1. The frequency-domain field equations
We subject the electromagnetic field equations (2.28) and (2.29) to a

Laplace transformation over the interval {t ∈ IR; t > 0}. We are only
interested in the causal field generated by sources that are switched on at
the instant t = 0, we arrive at
ext
−∇ × Ĥ + (σ+sε)Ê = −Ĵ , (2.52)
ext
∇ × Ê + sµĤ = −K̂ , (2.53)
or written out in their components as
−(∂2 Ĥ3 − ∂3 Ĥ2 ) + (σ+sε)Ê1 = −Jˆ1ext , (2.54)

−(∂3 Ĥ1 − ∂1 Ĥ3 ) + (σ+sε)Ê2 = −Jêxt , 2 (2.55)
−(∂1 Ĥ2 − ∂2 Ĥ1 ) + (σ+sε)Ê3 = −Jˆ3ext , (2.56)
∂2 Ê3 − ∂3 Ê2 + sµĤ1 = −K̂1ext , (2.57)

∂3 Ê1 − ∂1 Ê3 + sµĤ2 = −K̂2ext , (2.58)
∂1 Ê2 − ∂2 Ê1 + sµĤ3 = −K̂3ext . (2.59)
The steady-state analysis

The behavior of electromagnetic waves is often characterized by the re-
sults of a steady-state analysis. In such an analysis, all electromagnetic field
quantities are taken to depend sinusoidally on time with a common angular
frequency ω. Each purely real quantity f (x, t) can then be associated with a
complex counterpart fˆ(x, jω) and a common time factor exp(jωt). In doing
so, the original quantity is found from its complex counterpart as

f (x, t) = Re fˆ(x, jω)exp(jωt) . (2.60)
Substitution of these representations for the field quantities occurring

in the time-domain field equations (2.28) and (2.29) yields, except for the
common time factor exp(jωt), the set of field equations
ext
−∇ × Ĥ + (σ+jωε)Ê = −Ĵ , (2.61)
ext
∇ × Ê + jωµĤ = −K̂ . (2.62)
or written out in their components as
−(∂2 Ĥ3 − ∂3 Ĥ2 ) + (σ+jωε)Ê1 = −Jˆ1ext , (2.63)

−(∂3 Ĥ1 − ∂1 Ĥ3 ) + (σ+jωε)Ê2 = −Jêxt ,
2 (2.64)
−(∂1 Ĥ2 − ∂2 Ĥ1 ) + (σ+jωε)Ê3 = −Jˆ3ext , (2.65)
∂2 Ê3 − ∂3 Ê2 + jωµĤ1 = −K̂1ext , (2.66)

∂3 Ê1 − ∂1 Ê3 + jωµĤ2 = −K̂2ext , (2.67)
∂1 Ê2 − ∂2 Ê1 + jωµĤ3 = −K̂3ext , (2.68)
which are similar to Eqs. (2.52) - (2.59), if we take s = jω. Hence, we

interpret the steady-state analysis as a limiting case of the Laplace transform
analysis, in which s → jω .
polarization state 39
2.8. Polarization state
An important aspect of a time-harmonic electromagnetic field defined in

the previous section is its polarization state (not to confuse with the electric
polarization P ). The polarization state of a field vector (here we will choose
the electric field strength of the electromagnetic field) is described by the
locus of the endpoint of the field vector E(x, t) as time progresses. From
(see Eq. (2.60))

E(x, t) = Ê (x, jω) cos(ωt) + Ê (x, jω) sin(ωt) , (2.69)
where

Ê (x, jω) = Re Ê(x, jω) , (2.70)

Ê (x, jω) = −Im Ê(x, jω) , (2.71)
we observe that the field vector E(x, t) for any instant is located in the plane

through Ê (x, jω) and Ê (x, jω) while its endpoint describes an ellipse in
that plane. The electric field strength (and also) the electromagnetic field is
then said to be elliptically polarized at that point (Fig. 2.5).

When Ê (x, jω) and Ê (x, jω) have the same direction, i.e.,
∗
Ê(x, jω) × Ê (x, jω) = 0 , (2.72)
the ellipse degenerates into a straight line. The electric field strength is then

said to be linearly polarized at that point (Fig. 2.5). Finally, when Ê (x, jω)

and Ê (x, jω) have equal magnitudes and are perpendicular to each other,
i.e.,
Ê(x, jω) · Ê(x, jω) = 0 , (2.73)
the ellipse changes into a circle. The electric field strength is then said to be
circularly polarized at that point (Fig. 2.6).
ωt = 12 π ..
ωt = 12 π........ ...
.
.....................................................
.......... ....... ......
........ ............ .....
....... ........
t ..........
..........
. ..
..
.....
....
... ...................
..
.. .
...
.
..
..
......
.
.
.
.
.
..

...
...
...
Ê ........... ..
... .
. ... .........
.....
.
.
..
.
.
.
.
..
Ê ...
...
.
............... ωt = 0
........
.......
.
..
.
.. ... .. .
....
...
.
.
.
.
..
.
.
. ...
.
..
Ê...........
.. .
. ..
. .
...
. .
.. ..
. ...
. . . ...
... ... ... ...
... ... ... .....
.. .. .. ...
...
s..
. Ê .......... ....
s...
...
ωt = π ...
.
..
.
..... ....... ....... ....... ....... ....... ...............................................................................................
... .. ..
..
ωt = 0 .
...
... .. ... ...
.... .. ...
.... . .... .
...
.
.. ... ... .
.. ...
.... .. ...
... ... ...
. .
... .... ...
... .. ... .
...
...
... .
.
..
..
...
.
....
.....
.....
ωt = π...............
.
... .....
... .. ..... .
... ... ...... ...
..... ......
.....
...... .
. .
............ .........
....... .. ....... .
.........
...........
.....................................
.
.
... ωt = 32 π
3 ...
ωt = 2π
Figure 2.5. Elliptical polarization (left) and linear polarization (right)

of the electric field strength.
ωt = 12 π
...........................................................
......... .......
....... ......... ......
t . .........
. .. ... .. ......
.......... ....
.....
...
.
....... .
.
.....
....
......
..
Ê .
.
.
..
.
.
.
...
...
...
...
. .
.... ..
. ...
...
.... .... ...
... ...
... ... ...
...
...
s ...
.......... ....... ....... ....... ....... .......................................................................................... Ê .
ωt = π ...
...
... .......... ....
.
.
ωt = 0
...
... .... ...
... . ...
... .... ...
... ...
... .
... ...
... .... ...
.... . ...
..... ..
.....
. ..
.....
...... ... .....
.....
......
....... ......
......... .. .......
....................................................
...
ωt = 32 π
Figure 2.6. Circular polarization of the electric field strength.

poynting’s theorem 41
At a fixed position x, the polarization is said to be right- or left-handed

according to whether Ê , Ê , and the direction of propagation, form a right-
or lefthanded triad, respectively.
2.9. Poynting’s theorem
To calculate instantaneous power, we shall derive the instantaneous

Poynting theorem from the field equations (2.28) and (2.29). We take the
dot product of Eq. (2.28) with E and the dot product of Eq. (2.29) with H,
and add the results. Using the vector identity −E · (∇ × H) + H · (∇ × E)
= ∇ · (E × H), we find
∇ · (E × H) + σE · E + εE · ∂t E + µH · ∂t H = −E · J ext − H · K ext . (2.74)
Noticing that E · ∂t E = ∂t ( 12 E · E) and H · ∂t H = ∂t ( 12 H · H), we can write
∇·(E×H)+σE·E+∂t ( 12 εE·E)+∂t ( 12 µH·H) = −E·J ext −H·K ext . (2.75)
In order to arrive at the physical interpretation of the different terms in

this relation, we integrate Eq. (2.75) over an elementary domain D that is
bounded by the surface ∂D with outward normal ν (Fig. 2.7). We obtain
the power relation
P out + Ẇ h + ∂t (W e + W m ) = P ext . (2.76)
The term

P out
= ∇ · (E × H)dV
x∈D
= ν · (E × H)dA = ν · SdA , (2.77)
x∈∂D x∈∂D
where Gauss’ integral theorem has been used, is written as an integral over
the surface ∂D through which D is in contact with its surroundings. There-
fore, it seems natural to interpret this term as the instantaneous power that
is, across ∂D, transferred from D toward its surroundings.
.............................................
∂D................................... ......
.....
...
ν ....
..
.
.
.....
... ...
...
...
..... ...
@
I
@ .. .
........
. ...
..... ..
..
...
.
....
..
. ..
..
.
.
... ...
.
...... ...
...
. ..
.
D
.. ...
...
... ...
... ...
...
. ..
.
... ...
... ...
... ...
..
. .....
... ..
.. ...
.. ...
... ...
... σ, ε, µ .
.....
.
.
.... .....
....
... .....
... .....
... .........
.
... ......
... .....
... ......
......
....
..... .
...
.........
...... ...
......... .........
...........................................
Figure 2.7. Representative elementary domain D.
The term

Ẇ h = σE · EdV (2.78)
x∈D
is a loss term and represents the electromagnetic power that is irreversibly
dissipated into heat.
The two terms
We = 1
2 εE · EdV (2.79)
x∈D
and
Wm = 1
2 µH · HdV (2.80)
x∈D
occur only as ∂t (W e + W m ) and are interpreted as the amount of energy
that is reversibly stored in the electric field and the magnetic field in D,
respectively.
Finally, the term

P ext
=− (E · J ext + H · K ext )dV (2.81)
x∈D
poynting’s theorem 43
represents the electromagnetic power that is generated by the electromag-

netic sources in D whose action has been accounted for by the volume densi-
ties of exterior electric current and magnetic current. The minus sign in this
term is typical for the fact that for the sources to deliver a positive power
to the system, the currents must be oriented opposite to the fields. Notice,
that in this physical interpretation the law of conservation of energy holds.
Now that Eq. (2.76) has been physically interpreted, we return to the
local equation (2.75). In view of Eqs. (2.78) - (2.80) we introduce the volume
density of electromagnetic power that is irreversibly dissipated into heat
ẇh = σE · E , (2.82)
the volume density of reversibly stored electric field energy
we = 12 εE · E , (2.83)
the volume density of reversibly stored magnetic field energy
wm = 12 µH · H . (2.84)
Then, Poynting’s theorem is written as
∇ · S + ẇh + ∂t (we + wm ) = −E · J ext − H · K ext ; (2.85)
it expresses the local conservation of energy.
Steady-state analysis
In the steady-state analysis we define the time average Poynting’s vector
ST as the average of the time domain quantity S(x, t) = E(x, t)×H(x, t).
Let T = 2π/ω be the period in time of the fields, then we define the time
average

1 t +T
ST = E(x, t) × H(x, t)dt . (2.86)
T t=t
Using the complex representations of the type of Eq. (2.60), it follows that
∗

ST = 1
2 Re Ê × Ĥ , (2.87)
in which Ê = Ê(x, jω) and Ĥ = Ĥ(x, jω) are the complex electric and
∗
magnetic field vectors. The vector Ê × Ĥ is known as the complex Poynting
vector.
To calculate the time-average power, we start with Eqs. (2.61) and (2.62),
and subsequently we take the dot product of the complex conjugate of
∗
Eq. (2.61) with Ê and the dot product of Eq. (2.62) with Ĥ , and add
the results. We then find
∗ ∗ ∗ ext∗ ∗ ext
∇·(Ê × Ĥ )+(σ −jωε)Ê · Ê +jωµĤ · Ĥ = −Ê · Ĵ − Ĥ · K̂ . (2.88)
This equation is known as the complex Poynting’s theorem. The conserva-

tion of energy follows from this equation by taking half of its real part in
accordance to Eq. (2.87), viz.

∗ ∗ ext∗ ∗ ext
1
2 Re ∇ · (Ê × Ĥ ) + 1
2 σ Ê · Ê = − 12 Re Ê · Ĵ + Ĥ · K̂ .
(2.89)
The term
∗
ẇh T = 12 σ Ê · Ê (2.90)
represents the time average of the volume density of heat dissipated by elec-
tromagnetic field, while the right-hand side of Eq. (2.89) can be interpreted
as the time average of the volume density of electromagnetic power that is
generated by the specified source in D whose action has been accounted by
the volume densities of exterior electric current and magnetic current.
the electromagnetic field equations 45
Exercise 2.1
How many unknown quantities occur in Maxwell’s equations in matter?
Exercise 2.2
Substitution of Eq. (2.24) in Eq. (2.22), with J = 0, yields the local form
of the conservation law of electric charge, ∇ · J ext = −∂t ρ. If we integrate
both sides of this equation over the source domain Dext we obtain the global

x∈Dext ∇ · J dV = −∂t
ext
form, x∈Dext ρ dV . Use Gauss’ law to give
a physical interpretation of the global form of the law of conservation of
charge.
Exercise 2.3
Let the plane x3 = 0 with unit normal ν = i3 be the source-free interface
between two media with different electromagnetic properties. Apply the
reasoning, as it is used to derive the boundary conditions given in Section
2.6, to Eqs. (2.30) - (2.35) and derive the boundary conditions for E1 , E2 ,
H1 and H2 .
Exercise 2.4
Apply the reasoning, as it is used to derive the boundary conditions given in
Section 2.6, to Eqs. (2.22) and (2.23), and derive the boundary conditions
for J , D and B at a source-free interface.
Exercise 2.5
Find the polarization state (linear, circular or elliptical) of the following
fields given at a fixed position in space
(a) Ê = ji1 + i2
(b) Ê = (1 + j)i1 + (1 − j)i2
( c ) Ê = (1 + j)i2 − ji3
(d) Ê = (2 + j)i1 + (3 − j)i3
( e ) Ê = ji1 + j2i2 .
Exercise 2.6
Give the dimensions of the following quantities in SI-units
(a) D · E , B·H
(b) J · E , E×H
(c) S , ST .
Exercise 2.7
1
Given the two fields Ê = (i1 + ji2 ) and Ĥ = ( µε00 ) 2 (−ji1 + i2 ) at a fixed
point in space, calculate ST . Then calculate E(t), H(t) and S(t).
Problem 2.1
Substitution of Eq. (2.24) in Eq. (2.22) for a conducting medium, with J ext =
0, yields ∇·J +∂t ρ = 0. In an isotropic homogeneous domain we can express
∇ · J in terms of ρ as ∇ · J = σε ρ. Substitution of this relation leads to a
first order ordinary differential equation for the electric charge density. The
factor τ = σε is called the charge relaxation time of the medium and is a
measure for the time it takes the medium to restore its equilibrium state
once it has been disturbed by an electromagnetic wave.
(a) Given a charge distribution ρ0 at t = 0, solve the ordinary differential
equation for the electric charge density in the time domain.
(b) Use Table 2.1 to find the relaxation times for copper, sea water, fresh
water and glass.
In the frequency domain we can write the equation for the charge density
as (jω + σε )ρ̂ = ρ0 , where the right-hand side comes from the initial charge
distribution considered as a source input. If σ ωε the medium can be
considered a conductor for that frequency; this is equivalent to ωε σ
1,
σ
where the factor ωε is known as the loss tangent also denoted as tan(δ).
( c ) Find the frequency range for which ‘wet earth’ can be regarded a con-
ductor. Use Table 2.1 and the condition tan(δ) > 10.
Problem 2.2
Let the plane x3 = 0 be the interface S between two media with different
electromagnetic properties, D(1) with x3 < 0 and D(2) with x3 > 0, re-
spectively. In D(1) we have the constitutive constants ε(1) and µ(1) (while
σ (1) = 0) and in D(2) we have the constitutive constants ε(2) , µ(2) and σ (2) .
Let further in D(1) , in the limit of x3 ↑ 0, the electric flux density and the
(1) (1)
electric current density be given by D̂1 = 1 and Jˆ1 = 0. Give the values
(2) (2)
of D̂1 and Jˆ1 in the limit of x3 ↓ 0.
Problem 2.3
Show that an elliptically polarized field can be decomposed in two circularly
polarized fields, one left-handed and the other right-handed.
Hint: let Ê = (ai1 + bi2 ) with two arbitrary complex numbers a and
b, then let the superposition of two circularly polarized waves be Ê =
(a i1 + ja i2 ) + (b i1 − jb i2 ) and solve for a , b in terms of a, b.
Problem 2.4
Let the time-harmonic varying electric and magnetic field strengths at a
1
fixed point in space be given by Ê = i1 and Ĥ = ( µε00 ) 2 i2 . Determine
(a) E(t)
(b) H(t)
( c ) S(t).
Use the time-domain expressions for the electric and magnetic fields and the
Poynting vector to calculate
(d) ST
( e ) ∂t (we + wm )T .
Problem 2.5
Assume that the electromagnetic sources are switched on during a finite
interval (0 < t < T ). Integrate both sides of Eq. (2.76) over all positive
times and give a physical interpretation of the resulting terms in analogy of
Eqs. (2.77) - (2.81).
Chapter 3
One-dimensional
In Chapter 2 we observed that the space variations of the electric and

magnetic field components are related to the time-variations of the magnetic
and electric field components through a system of field equations. This inter-
dependence gives rise to the phenomenon of electromagnetic wave propaga-
tion. In the general case, electromagnetic wave propagation involves electric
and magnetic fields having more than one component, each dependent on
all three coordinates, in addition to time. However, a simple and very useful
type of wave that serves as a building block in the study of the electromag-
netic field consists of electric and magnetic fields that are perpendicular to
each other and to the direction of propagation. These waves are known as
plane waves. By orienting the coordinate axes such that the electric field
strength is in the i1 -direction and the magnetic field strength is in the i2 -
direction, the propagation direction is in the i3 -direction. Plane waves do
not exist in practice because they cannot be produced by finite-sized an-
tennas. At large distances from the sources, however, the waves can be
approximated as plane waves. Furthermore, the first principles of guiding of
electromagnetic waves along transmission lines and the ones of many other
wave phenomena can be studied in terms of plane waves. In fact, any elec-
tromagnetic field can be composed through the superposition of an infinite
number of different plane waves.
To illustrate the phenomenon of interaction of electric and magnetic fields
50 one-dimensional electromagnetic waves
giving rise to plane electromagnetic wave propagation, and the principle of

radiation of electromagnetic waves from an antenna, we shall consider a
simple, idealized, electromagnetic source. This hypothetical source is the
planar electric-current sheet with a uniform (but time-varying) impressed
electric current distribution. Such a current sheet acts as an active device
and emits in a homogeneous medium one-dimensional electromagnetic waves.
3.1. The planar electric-current sheet as emitter
A reference frame is introduced such that the sheet coincides with the
plane x3 = 0 (Fig. 3.1). Let the impressed electric current be uniformly in
the x2 -direction and flowing in the direction of decreasing x1 , then
Jˆ1ext = −Iˆ∆ (s)δ(x3 ) , Jˆ2ext = 0 , Jˆ3ext = 0 , (3.1)
where δ(x3 ) denotes the one-dimensional unit impulse (Dirac distribution),

and Iˆ∆ (in A/m) is the electric current per unit length along the x2 -direction.
Since the electric-current sheet carries no magnetic current, we have
K̂1ext = 0 , K̂2ext = 0 , K̂3ext = 0 . (3.2)
Because the exciting electric current is independent of x1 and x2 , also the

generated electromagnetic field will be independent of x1 and x2 and, hence,
∂1 = 0 and ∂2 = 0. With this, Eqs. (2.54) - (2.56) and (2.57) - (2.59) are
satisfied when
Ê3 = 0 , Ĥ3 = 0 . (3.3)
However, since Ê2 and Ĥ1 are not related to the source, we can take
Ê2 = 0 , Ĥ1 = 0 . (3.4)
The remaining equations for Ê1 and Ĥ2 are

the planar electric-current sheet as emitter 51
...
...
σ, ε, µ ...
...
σ, ε, µ
...
E. ...
... E
.
... ... ...
.......... ...
... ..........
... ...
..
... ......
...
i1 ..
...
... ....... ...
.... ... ....
S ..........................................×f O r r............................................ S
f
............................
...
...
.... i3
...
H ...
...
...
H
...
...
...
...
...
...
...
...
...
...
....... ..
x3 < 0 ........
......
x3 > 0
J ext .
Figure 3.1. Electric-current sheet with impressed current as an emitter of

one-dimensional electromagnetic waves.
∂3 Ĥ2 + (σ+sε)Ê1 = Iˆ∆ (s)δ(x3 ) , (3.5)

∂3 Ê1 + sµĤ2 = 0 , (3.6)
where σ, ε and µ are constants, since a homogeneous medium is assumed.

Eliminating Ĥ2 from these two differential equations of the first order we
end up with the second order differential equation,
∂3 ∂3 Ê1 − γ 2 Ê1 = −sµIˆ∆ (s)δ(x3 ) , (3.7)
for the electric field strength Ê1 , where
1
γ = [(σ + sε)sµ] 2 with Re(γ) ≥ 0. (3.8)
In the domain outside the source distribution, i.e., for x3 = 0, the modified
Helmholtz equation (3.7) has a zero right-hand side and admits exponential
functions of the type exp(±γx3 ) as their solutions. From Eq. (3.6) it follows
that the magnetic field also admits these exponential functions as their so-
lutions. The quantity of γ is the (s-domain) propagation coefficient. With
52 one-dimensional electromagnetic waves
the chosen value of the square root in the right-hand side, the propagation
factor exp(−γx3 ) is bounded as x3 → ∞ and unbounded as x3 → −∞, while
exp(γx3 ) is bounded as x3 → −∞ and unbounded as x3 → ∞. Since, due
to causality, the electromagnetic field in the half-space x3 < 0 must remain
bounded as x3 → −∞ and the electromagnetic field in the half-space x3 > 0
must remain bounded as x3 → ∞, we write
Ê1 = ê− (s) exp(γx3 ) for x3 < 0 , (3.9)

−
Ĥ2 = ĥ (s) exp(γx3 ) for x3 < 0 , (3.10)
and
Ê1 = ê+ (s) exp(−γx3 ) for x3 > 0 , (3.11)

+
Ĥ2 = ĥ (s) exp(−γx3 ) for x3 > 0 . (3.12)
As we will discuss later, Eqs. (3.9) - (3.10) represent an electromagnetic

wave propagating in the negative x3 -direction, while Eqs. (3.11) - (3.12)
represent an electromagnetic wave propagating in the positive x3 -direction.
Substitution of Eqs. (3.9) - (3.12) in Eq. (3.6) and use of Eq. (3.8) lead to
ĥ− = −Y ê− , (3.13)

+ +
ĥ = Y ê , (3.14)
where Y = Y (s) is given by

1
σ + sε 2
Y = . (3.15)
sµ
Y has the dimension of an admittance and it is therefore denoted as the

wave admittance. We further observe that through Eqs. (3.13) and (3.14)
the magnetic field is expressed in terms of the electric field. The analysis
where the electric field is considered as the fundamental unknown is called
the planar electric-current sheet as emitter 53
the electric field analysis. One can also write
ê− = −Z ĥ− , (3.16)

+ +
ê = Z ĥ , (3.17)
where Z = Z(s) is given by

1
sµ 2
(3.18)
Z= .
σ + sε
Z has the dimension of an impedance and it is therefore denoted as the

wave impedance. We further observe that through Eqs. (3.16) and (3.17) the
electric field is expressed in terms of the magnetic field. The analysis where
the magnetic field is considered as the fundamental unknown is called the
magnetic field analysis.
Obviously, we have
YZ =1. (3.19)
The values of the as yet undetermined coefficients in Eqs. (3.9) - (3.12)

follow from the application of the excitation conditions
lim Ĥ2 − lim Ĥ2 = Iˆ∆ , (3.20)

x3 ↓0 x3 ↑0
which is a consequence of Eq. (3.5), and
lim Ê1 − lim Ê1 = 0 , (3.21)

x3 ↓0 x3 ↑0
which is a consequence of Eq. (3.6). Substitution of Eqs. (3.9) - (3.12) in

Eqs. (3.20) and (3.21) leads to
ĥ+ − ĥ− = Iˆ∆ (3.22)
and
ê+ − ê− = 0 . (3.23)
Combining Eqs. (3.22) and (3.23), with Eqs. (3.16) and (3.17), we arrive at
54 THEORIES OF LIGHT: PARTICLES VERSUS WAVES
cation" theories by a series of experiments reported in Book I. Among New-

ton's other objections to the wave theory, perhaps the most famous is that if
light "consisted in Pression or Motion either in an instant or in time, it would
bend into the Shadow,"51 as with water waves and sound waves. But no such
bending (no diffraction) had been observed.
The wave theory, various forms of which were supported by Newton's
critics Hooke and Huygens, was the rival to Newton's corpuscular theory.
That there are objections to such theories seems to be at least part of New-
ton's reason for favoring a particle theory. Putting Queries 28 and 29 together,
then, we might say that Newton is proposing a weak inference to a particle
theory from two sets of considerations: the explanatory success of the par-
ticle theory (Query 29) and the objections to wave theories (Query 28). But
how exactly is this inference supposed to proceed? What form does it take?
If Newton is making such an inference—and I think it is plausible to
suppose that he is — he does not spell it out. On the basis of the discussions in
Queries 28 and 29 it is, I think, reasonable to say that Newton is arguing in
some such way as this:
The hypothesis that light is corpuscular explains a range of observed opti-

cal phenomena.
The rival wave hypothesis has such and such difficulties.
Therefore (probably)
Light is corpuscular.
The inference is intended to be weak, not strong. But even so, exactly how it is
supposed to go and whether it is reasonable even as a weak inference is not at
all clear.
In what follows I will construct an idealized version of this argument,
which, although I cannot claim it to be Newton's, may nevertheless reflect
some features of his thought. (See Essay 3.) The argument will have two
essential components: objections to the wave theory, and an appeal to the
explanatory power of the particle theory. Moreover, it will be an argument
whose conclusion is drawn with probability. This probability will be reasona-
bly high but not high enough to achieve the certainty, or virtual certainty, of a
"deduction from phenomena." In constructing the argument I will assume
that the usual axioms of the probability calculus are satisfied. Probability can
be construed here as representing rational degrees of belief.
Let us assume to begin with that light is either a particle phenomenon or a
wave phenomenon. Newton himself offers no explicit argument for such an
assumption, although here is one that he was in a position to offer (and that is
very similar to one in fact offered by his wave-theoretic opponents in the nine-
teenth century; see Essay 3):
51. Ibid.
NEWTON'S CORPUSCULAR QUERY AND EXPERIMENTAL PHILOSOPHY 55
Light is observed to travel in straight lines with uniform speed.

In other cases, when something travels with uniform speed in a straight line
this motion is always observed to be caused by a series of bodies or by a
series of wave pulses produced in a medium (e.g., sound waves, water
waves).
Therefore
Light is either a particle or a wave phenomenon.
Let us assume that the premises are true, that they report "phenomena," and
that this is a "deduction from phenomena," so that the conclusion has the
highest certainty possible in experimental philosophy. Such an inference, a
type of causal simplification, would be permitted by Newton's Rule 2 requir-
ing that like causes be assigned to like effects, as far as possible. If in other
cases motion is caused by particles or waves, then, unless there is evidence to
the contrary, we should infer like causes in the case of the motion of light.
Let T1 be the hypothesis that light consists of particles, T2 the hypothesis
that light consists of waves, O the observed fact that light travels in straight
lines with uniform speed, and b the accepted background information, which
includes the information in the second premise above. We might express the
results of this argument probabilistically, as follows:
That is, the probability that either T1 or T2 is true, given O and b, is equal to
1. (With minor alterations an argument similar to that which follows can be
made if equation (1) is changed to say that the probability in question is close
to 1; see Essay 3.)
Let us suppose that by appeal to certain other observed facts about light —
call them O'—we can show that the probability of the wave theory is low, say
less than 1/2. Which facts these are, and how this is to be shown, will be taken
up in a moment. For the present let us simply write
Now if p(T1 or T2/O&b) = 1, then p(T1 or T2/O&O'&b) = 1. Therefore from

(1) and (2), since T1 and T2 are incompatible, we can infer
Let O1, . . . ,On be various observed facts about light (e.g., reflection, refrac-
tion, variety of colors, fits of easy reflection and transmission, etc.) other
than those in O and O'. We would like to know how probable T1 is when these
facts are considered in addition to O&O'&b. Is p(T 1 /O 1 , . . . ,On&O'&b) also
high? Suppose that explanations of O1, . . . ,On by theory Tl can be con-
structed in such a way that O1, On follow deductively (in the ordi-
nary sense) from T1 together possibly with the background information b.
But if the particle theory Tl (together possibly with b) deductively entails
O1, On, then it follows from the probability calculus that p(T1:/
O1, . . . ,On&O&O'&b) p(T1/O&O'&b). So from (3), given the existence
of such explanations, we derive
which can be construed as the conclusion of the argument. It says that the
particle theory is probable (more probable than not) given not just a few
chosen optical phenomena but a range of them, including ones explained by
that theory.
The explanations of O1, . . . ,On provided by the particle theory do not
create the high probability for that theory, but they do sustain it. (In Essay 4 I
argue that, in general, derivational explanations, even those subject to certain
further conditions, will not by themselves suffice to guarantee high probabili-
ty, although they can increase it.) These explanations permit an inference
from (3) to (4). In an attempt to establish high probability for the particle
theory on the basis of a range of optical phenomena, this is an essential role
played by such explanations. To create high probability in the first place, a
type of eliminative argument is used in which the wave and particle theories
exhaust the probability but the probability of the wave theory is low. How is
the latter to be established?
Newton offers arguments of two types, one direct, the other indirect. An
argument appealing to diffraction is an example of the former. In the case of
other wave motions such as sound waves and water waves, diffraction into the
shadow of an obstacle is observed, but no such diffraction into the shadow
had been observed by Newton or others in the case of light (though Newton
had observed diffraction away from the shadow).52 So if we include in the
background information the fact of observed diffraction with other wave
phenomena and the absence of such observations in the case of light, then the
probability of the wave theory, given b, is low.
Second, Newton offers a more indirect type of argument. He points out
that to explain certain observed optical phenomena, the wave theory intro-
duces auxiliary assumptions that are either refuted by, or made very improba-
ble by, observations. For example, to explain differences in refrangibility of
rays emerging from a prism, wave theories introduce the auxiliary hypothesis
that the prism modifies rather than separates the rays. Newton argues that
this modification assumption is refuted or at least made extremely unlikely by
further refraction experiments. (See Experiment 5, Book I, pp. 34ff.) Can we
52. See Roger H. Stuewer, "A Critical Analysis of Newton's Work on Diffraction," Isis 61
(1970), pp. 188-205.
infer from this that the wave theory is improbable? We can if we suppose that
the probability of the modification assumption, given the wave theory and
the observations, is close to 1. Letting M be the modification assumption, T2
the wave theory, and O' the results of various of Newton's refraction experi-
ments, if p(M/O&O'&b) is close to zero and p(M/T2&O&O'&b) is close to 1,
then p(T2/O&O'&b) is close to zero. Even more generally, we get the same
result if we suppose simply that p(M/T2&O&O'&b) is much, much larger
than p(M/O&O'&b), without needing to suppose that the former probability
is close to 1. (See Essay 3.)
Accordingly, there are two sorts of arguments to show that the probability
of the wave theory is low, that is, (2). Once this is shown, we can infer the high
probability of the particle theory (3), and its continued high probability in
light of the various observed facts that it explains, that is, (4).
How "Newtonian" is the previous argument? In one respect, quite un-
Newtonian, since it explicitly invokes numerical probabilities and the proba-
bility calculus, neither of which, of course, Newton does. However, in certain
other respects it reflects what Newton seems to be doing in Queries 28 and 29.
Assuming, as I have been, that Newton intends to provide some reasons for
believing the particle theory, albeit "weak" ones, it gives a basis for inferring
that theory with probability rather than certainty. Moreover, in doing so it
takes into account Newton's criticisms of the wave theory and the explanatory
virtues of the particle theory (each of which Newton himself emphasizes),
showing how both contribute to the probability of the particle theory.
One objection that might be offered is that this argument is a type of
eliminative one, whereas Newton at one point rejects (a certain form of)
eliminative reasoning. In a letter to Oldenburg of July 1672 he writes:
I cannot think it effectual for determining truth to examine the several ways by
which phenomena may be explained, unless there can be a perfect enumeration of
those ways. You know, the proper method for inquiring after the properties of
things is to deduce them from experiments. And I told you that the theory, which I
propounded [the theory of the heterogeneity of light rays] was evinced to me, not by
inferring 'tis thus because not otherwise, that is, not by deducing it only from a
confutation of contrary suppositions, but by deriving it from experiments conclud-
ing positively and directly.53
Newton here seems to be rejecting eliminative arguments of this form:
E: Each of the hypotheses h1, . . . ,hn, if true, will correctly explain phe-
nomenon p.
But hypotheses h2,. . . ,hn are false.
Therefore, hypothesis h1 is true.
53. I. B. Cohen, ed., Newton's Papers and Letters (Cambridge, Mass., 1978), p. 93.
Such arguments, which infer the truth of an hypothesis from the falsity of
competitors, are fallacious, unless a complete enumeration can be made of all
the competitors. Newton appears to be thinking of deductive interpretations
of E in which if the premises are true the conclusion must also be true. And he
is correct in saying that arguments of form E, thus construed, are fallacious
unless a complete enumeration of hypotheses is given. They are also falla-
cious if construed nondeductively, that is, as being such that the premises
make the conclusion probable without entailing it. Unless some suitable as-
sumption is made about the probability of the disjunction of hypotheses that
are mentioned in the first premise, the conclusion that the probability of hl is
high cannot be drawn.
However, the particular eliminative argument I have constructed is not of
form E. The first step in the argument, which leads to (1) above, is not an
explanatory step, but one involving causal simplification. The claim is not
that the particle and wave theories will both explain the finite rectilinear
motion of light, but that in other observed cases when something travels with
uniform speed in a straight line this motion is caused by a series of bodies or a
series of wave pulses in a medium. Also, the claim in the first step is indeed
exhaustive, since it assigns a probability of 1 to the disjunction of hypotheses.
But even if it were not exhaustive in this sense, even if (1) were changed to read
"p(T 1 or T2/O&b) is close to but not equal to 1," a fallacy would not emerge
(although other changes would need to be made in the argument).
I conclude that reconstructing what Newton does in Queries 28 and 29 in
the form of a probabilistic argument that takes us from (1) to (4) above is in
conformity with certain important aspects of Newton's methodology. It com-
bines his explanatory reasoning in Query 29 with his criticisms of the wave
theory in Query 28 to provide some reason, though not the highest possible in
experimental philosophy, to believe the corpuscular hypothesis. Although it is
an eliminative argument it is not one of the type Newton rejects. Because the
hypothesis in question is inferred with a probability not sufficiently high to be
a virtual certainty, Newton could not construe the argument as a "deduction
from phenomena." While we should search for phenomena that will sanction
such a "deduction," we should acknowledge that, assuming its premises are
true, the present argument does provide a legitimate "weak" reason for believ-
ing the corpuscular hypothesis.
6. STRONG VERSUS WEAK INFERENCES: AN ASSESSMENT

OF ONE TENET OF NEWTONIAN METHODOLOGY
A "strong" inference furnishes the proposition inferred with the highest evi-
dence possible in experimental philosophy; a "weak" inference furnishes some
evidence but not the highest possible. I shall suppose that this difference can
be interpreted as a difference over probabilities (construed as representing

degrees of rational belief). In a strong inference from A to B, the probability
of B given A is close to or equal to 1. In a weak inference the probability is
high (say greater than 1/2), but is not close to or equal to 1. For Newton, both
strong and weak inferences are based on "phenomena."
Now I take it to be a tenet of Newtonian methodology that in experimental
philosophy "deductions from phenomena," and only these, are strong infer-
ences. Accordingly, I want to raise two questions of assessment: (a) Are
Newton's "deductions from phenomena" guaranteed to be strong inferences?
(b) Must other kinds of inferences fail to be strong?
To answer the first question we must return to the definition of "deduction
from phenomena" offered in section 2. Deductions from phenomena, we
recall, include ordinary deductions, inductions, and causal simplification.
Inductions are inferences from all observed members of a class to some
members of the class that have not been observed, or to all members of the
class. Let me simplify the discussion by considering deductive and inductive
inferences but omitting causal simplification, which does not lend itself so
readily to a general probabilistic treatment. Also, I shall discuss cases involv-
ing only deductions (in the ordinary sense) and those involving only induc-
tions.
Deductive cases: Let O1, . . . ,On be descriptions of phenomena the con-
junction of which, together with background information b, deductively im-
plies h. Then p(h/O 1 , . . . ,On&b) = 1. So here we have a "strong" inference
from the O's and b to h.
Inductive cases: To discuss these I shall first introduce some probability
considerations and afterward apply them to the sorts of cases particularly
relevant to Newtonian induction. Let h be a proposition that together with
background information b deductively entails some observational statements
O1,O2, . . . . The following claims are provable.54
(a) tells us that if the prior probability of h is not zero, then as the number n
of observed consequences of h and b gets larger and larger, the probability
that the n + 1 observational consequence of h and b will be true gets higher
and higher, approaching 1 as a limit. (b) tells us that if the prior probability of
h is not zero, then as the numbers m and n get larger, the probability that the
54. See John Earman, "Concepts of Projectibility and Problems of Induction," Nous 19
(1985), pp. 521-535.
next m observational consequences are true, given that n observational conse-

quences obtain, gets higher and higher and approaches 1 as a limit.
To introduce the third probability result some restrictions will need to be
made on h and Oi. Let h be some universal generalization of the form (x)(Fx
Gx). Let the O's be "instances" of h of the form Fai Gai. The following is
provable:
(c) gives a set of sufficient conditions for the probability of (x)(Fx Gx),
given observed instances of the form Fa, Gai, to approach 1 as a limit.
Now let us apply these three probability results to Newtonian inductions.
In all three cases let us consider h's of the form (x)(Fx Gx), and O's instances
of the form Fai Gai. (a) tells us that if the prior probability of (x)(Fx Gx) is
not zero, then as the number of observed instances of (x)(Fx Gx) increases,
the probability that the next instance will obtain gets higher and higher,
approaching 1 as a limit. A similar claim can be made for (b). (a) and (b) —so
construed — correspond to Newton's inductions from some observed members
of a class to some other member(s) of that class. (c) corresponds to Newton's
inductions from some observed members of a class to all members. In all
three cases the probability in question approaches 1 as a limit, under certain
very weak assumptions. Intuitively, the probability that the next instance will
satisfy (x)(Fx Gx), that the next m instances will, and that all instances will,
gets higher and higher as more and more instances are observed. We get more
and more certainty in these cases with more and more observed instances of
(x)(Fx Gx).
However, it is not the case that for every number n of observed instances,
the probability that the next instance will satisfy (x)(Fx Gx), that the next m
instances will (for any m), and that all instances will, is close to 1. Consider
just the latter, and suppose that the prior probability of (x)(Fx Gx) is low,
and the O's are such that, with sufficiently small n, the prior probability of
the conjunction of O's is high. If (x)(Fx Gx) and b entails the O's, then by
Bayes' theorem,
Now if p((x)(Fx Gx)/b) is low and the O's are such that, with sufficiently
55. For proof see Earman, op. cit., p. 529.

small n, p(O 1 , . . . ,On/b) is high, then the probability on the left will be
small, despite the fact that all the observed O's satisfy (x)(Fx Gx). One case
of this sort involves Goodmanesque properties such as "grue," where the prior
probability of the proposition "All emeralds are grue" is very low but where,
given appropriate background information, the probability that observed
items are grue if they are emeralds is very high. Strange Goodmanesque
properties or classes can prevent the probability on the left from being high
for a given n. But as n increases without bound, the probability on the left will
approach 1 as a limit, strange properties notwithstanding.
However, Goodmanesque properties are not the only things that can pre-
vent the probability on the left from being high for a given n. Recall the proof
of Proposition 1 of the Opticks. Here an induction is made from observations
of differences in refrangibility of blue and red rays in an experiment with the
sorts of prisms used by Newton to differences in refrangibility of any differ-
ently colored rays in any sort of refraction, whether or not the latter is
produced by a prism. A critic of Newton might argue as follows: (i) The
number n of observed instances of Proposition 1 (that lights that differ in
color differ in degrees of refrangibility) is quite low. (If we count as instances
here the results of types of experiments, rather than specific trials, then the
critic has some justification, since Newton cites only two experiments.) (ii)
The critic might agree that the probability of getting the results Newton
obtains with these types of experiments with prisms is high, while supposing
that obtaining analogous refraction results with other sorts of prisms or
without prisms is improbable.56 (iii) The critic might argue, on the basis of
background information b, that the prior probability of Newton's Proposi-
tion 1 is very low. At least, the critic might complain, Newton does nothing to
dispel doubts expressed in points (i) through (iii). But unless such doubts are
removed, the probability of Newton's Proposition 1, given the results of the
experiments Newton mentions, cannot be assumed to be high. The most we
can say is that this probability will increase toward 1 as the number of ob-
served instances of the proposition increases.
Confining our attention to ordinary deductions and inductions, we can
now answer the question "Are Newton's 'deductions from phenomena' guar-
anteed to be strong inferences?" in the following way. If they are deductions
(in the ordinary sense), they are so guaranteed. Any deductive inference from
O1, ,On and b to h—no matter what number n is — guarantees that the
probability of h given O1, . . . ,On and b is maximal. By contrast, it is not the
case that every particular inductive inference is guaranteed to be strong, no
matter how many instances are involved and no matter what the character of
56. See Simon Schaffer, "Glass Works: Newton's Prisms and the Uses of Experiment," in
David Gooding, Trevor Pinch, and Simon Schaffer, eds., The Uses of Experiment (Cambridge,
England, 1989), pp. 67-104.
the properties or classes in question. If Newton's methodology requires a

claim to the contrary, then it is mistaken. However, the previous probability
results show that (under certain weak assumptions), as more and more in-
stances are observed, then no matter what the character of properties or
classes in question, the strength of the inference is guaranteed to increase and
to approach the highest strength in the limit.
Accordingly, there are several ways to interpret Newton's methodology (or
to modify that methodology) so as to avoid the problems above. First, instead
of saying that every inductive inference from phenomena is a strong one,
Newton could say that some are, namely, those based on sufficiently many
instances (provided that the prior probability of (x)(Fx Gx) is not zero).
Second, Newton could restrict those inductions he will allow in the category
of "deductions from phenomena" to ones based on sufficiently many in-
stances. On both of these proposals, however, no particular number can be
chosen that will count as "sufficiently many." In each case this will depend on
the prior probability of (x)(Fx Gx) and on the prior probability of the
conjunction of instances. Third, Newton could attempt to impose conditions
on the character of the properties or classes that are subject to induction so
that inductions involving such properties or classes will guarantee maximal
probability no matter how many instances have been observed. Newton does
not formulate any such conditions. Whether it would be possible to do so
seems very dubious to me, though I shall not pursue this here. Finally, New-
ton could abandon entirely an "absolute" claim about the strength of induc-
tions in favor of a "comparative" one. He could say simply that, under
minimal assumptions, the strength of an induction increases as more and
more instances are observed.
Now turning to the other side of the coin we need to ask whether in
experimental philosophy there are strong inferences that are not "deductions
from phenomena." Is Newton correct in implying that only "deductions from
phenomena" can have this feature?
Let us return to result (a) above. (What I say here will be applicable to (b)
as well, mutatis mutandis.) Although (a) allows h and Oi to be of forms
(x)(Fx Gx) and Fa Ga, respectively, it does not require this. All that is
necessary is that h and b deductively imply Oi. Accordingly, h might be some
proposition that Newton would classify as an hypothesis, for example, that
light consists of particles. This hypothesis is not "deduced from phenomena."
Let the background information b include Newton's first law of motion that
in the absence of forces particles travel with uniform speeds in straight lines.
Hypothesis h + b deductively implies (O1) that in the absence of forces light
travels in straight lines, and (O2) that in the absence of forces light travels with
uniform speed. Now result (a) allows us to conclude that the probability that
some consequence of an "hypothesis" (in the Newtonian sense) obtains gets
higher and higher, approaching 1 as a limit, as more and more consequences
of that hypothesis are observed. The only assumption needed is that the prior
probability of this hypothesis is not zero. This, of course, does not imply that
the probability of the "hypothesis" itself approaches certainty, but only that
the probability of its deductive consequences does.
Newton does not appear to be thinking of cases in which we make infer-
ences to deductive consequences of "hypotheses." But such inferences can be
strong ones, or at least they can get stronger and stronger as more and more
consequences are observed to hold. To be sure, Newton could claim that he is
classifying as "inductive" an inference from some observed consequences of h
to other not yet observed consequences. But his inductions appear to be
simply inductive generalizations from observed F's that are G's to other or all
F's being G's.
Let us turn, then, to result (c) involving the probability of h itself. And let
us consider the more general case in which h is any proposition that, together
with b, deductively implies O1, O2, Here we cannot obtain the result
that lim p(h/O 1 , . . . ,On&b) = 1 because we cannot in general assume that
n
lim
n—
p(O 1 , . . . ,On/b) = p(h/b). Indeed, the following are provable:
(d) Let h (together with b) entail O1, O2, . . . . If h has at least one
incompatible competitor h' that together with b also entails O1, O2,
. . . , and whose probability on b is greater than zero, then lim
p(h/O 1 , . . . ,On&b) 1.
(e) Let h together with b entail O1, O2, . . . . If h has at least one incom-
patible competitor h' that together with b also entails O1, O2, . . . ,
and is such that p(h'/b) p(h/b), then for any n, no matter how
large, p(hlO 1 , . . . ,On&b) .5.57
So if h has competitors that, like h, deductively imply all the observable

phenomena, then h's probability will not approach 1 as a limit; and if the
prior probability of one of the competitors is at least as great as h's prior
probability, then h's probability will not increase beyond .5, no matter how
many deductive consequences of h are observed to be true.
However, the quest for strong inferences to Newtonian "hypotheses" is not
necessarily doomed. We need not insist that the limit of the probability of h
be 1, but only that the probability of h given the observations be "very high"
and remain so with more and more observations. To this end, I shall employ
the concept of a partition of propositions on b, which is a set of mutually
exclusive propositions, the probability of whose disjunction on b is 1, and the
probability of each of which on b is not zero. The following is provable (see
Essay 4):
57. See Earman, op. cit., pp. 528-529.

(f) If h, h1, . . . , hk form a partition on b, then for any O, and for each
hi ( h) in the partition, and for any number r greater than or equal to
0 and less than 1, p(h/O&b) > r if and only if p(h i /O&b) <
1 - r. i=1
Now suppose that we have some observed phenomena O, O1, ,On and
background information b, and we want to make a strong inference to h by
showing that the probability of h given the observed phenomena and back-
ground information is greater than some threshold value r for "very high"
probability. Using theorem (f), the following strategy is possible:
Strategy for showing that h has a very high probability (greater than some threshold
value r for very high probability), given observed phenomena O, O1, . . . , On and
background information b:
1. Find some partition on b—h, h1, . . . ,hk—that includes h.

2. Show that phenomenon O is such that for each proposition hi, h in the
partition, p(ht /O&b) < 1 - r.
i 1
3. Show that O1, . . . ,On are derivable from h (together with b).
If we complete steps 1 and 2 in this strategy, then, in accordance with theorem

(f), we will have shown that p(h/O&b) > r. By completing step 3, we show
that p(h/O& O1 , . . . ,On&b) > r, since O1, . . . ,On are derivable from h to-
gether with b.
The question of interest is whether this strategy is applicable to proposi-
tions Newton would regard as hypotheses. In fact, it seems applicable to the
hypothesis Newton considers in Query 29 of the Opticks—that light consists
of particles. Indeed, the probabilistic argument I constructed in the last sec-
tion can be suitably modified and shown to be a legitimate variant of this
form. Let me recall the basic steps.
We began with the observation O that light travels in straight lines with
uniform speed, which, together with the background information b, yields a
probability of 1 that light is corpuscular or undulatory, that is,
Accordingly, T1 and T2 form a partition on O&b, since T1 and T2 are incom-

patible.
Second, we found some other observed facts O' that cast doubt on the
wave theory T2. We noted this by writing p(T2 /O&O'&b) < 1/2. But this is
too modest, even for Newton, since the actual facts cited, Newton thought,
cast much more doubt on T2 than this. The two mentioned were diffraction
and refraction. If light is a wave phenomenon, then, like water waves and
sound waves, it ought to be diffracted into the shadow; but Newton observed
no such diffraction. Second, Newton (as well as defenders of the wave theory)
believed that, given the wave theory and the observations of differences in
degrees of refraction, the probability that light is modified by the refracting
prism is very high, say close to 1. But on the basis of his own refraction
experiments, Newton pretty clearly thought he had refuted the modification
assumption, that is, the probability of this assumption, given his experimen-
tal results, is close to zero. So, where M = the modification assumption, and
O' includes the results of Newton's refraction experiments, we have the result
ihat p(M/T2&O'&O&b) is close to 1, whereas p(M/O'&O&b) is close to zero.
It follows that p(T2 /O'&O&b) is close to zero. Letting O' also contain the
observed absence of diffraction into the shadow and b also contain observed
diffraction in the case of sound and water, we write
This completes the second step in the strategy.

From (1), since the probability of T1 or T2 is 1, it remains 1 if we add O'. So
we have
Since T1 and T2 are incompatible, from (2) and (3) we infer
Since T1 + b deductively implies other optical phenomena O1, . . . ,On, from

(4) we derive
which completes the third and final step of the strategy.

Again, I must stress that it is not my claim that this is Newton's actual
argument in Queries 28 and 29. Besides the attribution of the probability
calculus, the main stumbling block lies in the use of the first step in the
strategy, leading to (1) above (even if (1) were weakened by replacing "equals"
with "is close to"). Although Newton considers only the wave and particle
theories, he does not explicitly claim that the probability of their disjunction
on the evidence is maximal (or even very high). Still in the previous section I
indicated what type of argument for this claim Newton could have given that
would be compatible with his general methodology. If the strategy is launched
by this assumption in step (1), then Newton's own arguments against the wave
theory can be used to justify the steps leading to the final (5).
Let us suppose that (1) is justified by inference from observed phenomena.
And let us assume that the remaining steps are also valid, so that the argu-
ment does establish the very high probability of a proposition given certain
observations and background information. If so it provides the basis for a
"strong" inference to that proposition from those observations and back-
ground information. Is the argument a "deduction from phenomena"?
In certain ways it seems quite different from the sorts of arguments New-
ton has in mind when he uses this expression. First, unlike the "deductions"
that Newton gives, it contains an inference to a disjunction of propositions in
the first step. Second, the argument is eliminative, whereas the "deductions"
Newton offers are not. Indeed, he rejects (certain types of) eliminative argu-
ments. Third, and most important, it makes use of the probability calculus,
which Newton never does. The inferences to (3), (4), and (5) are justified by
principles of probability. Whether Newton would have been willing to classify
such inferences as "deductive" is unclear.
Yet reasons might be offered for classifying the argument as a "deduction
from phenomena." First, the previous characterization of causal simplifica-
tion (as well as that of induction) does not preclude an inference to a disjunc-
tion of propositions. Second, although it is eliminative, it is not an elimina-
tive argument of the type that Newton rejects. Indeed, if the previous point is
accepted, it is an eliminative argument that uses causal simplification to
establish a disjunction of propositions and then to argue against one of the
disjuncts. Third, the probability principles generating steps (3), (4), and (5)
might be thought of as, or as akin to, mathematical principles, which for
Newton can serve as a basis for "deductions."
Accordingly, assuming the argument in question is valid, the following
possibilities emerge:
1. In a broad sense the argument is a "deduction from phenomena." If so
it does not refute the Newtonian claim that only "deductions from phenome-
na" guarantee strong inferences. However, if we construe it as a "deduction
from phenomena," then with this argument we must deny that the Newtonian
corpuscular hypothesis is an hypothesis. With this argument we will have
"deduced" the corpuscular hypothesis from the phenomena and thus rendered
it no longer hypothetical.
2. In a narrower sense (one that excludes probability arguments) the argu-
ment is not a "deduction from phenomena." Yet it provides the basis for a
"strong" inference to the corpuscular hypothesis. So if this narrower sense is
Newton's, then we need to reject his idea that only "deductions from phenom-
ena" can provide the highest certainty in experimental philosophy.
7. CONCLUSIONS
1. Although neither Newton's professed methodology, nor his actual prac-

tice, form consistent sets, my suggestion is that interpretation (3) in section 3
reflects a good deal of both. On that interpretation, the most certainty possi-
ble in experimental philosophy is achieved when, and only when, propositions
are "deduced from phenomena." The latter involves deduction or induction
or causal simplification from generally accepted facts established by observa-
tion.
2. However, on this interpretation one is allowed not only to consider
propositions not "deduced from phenomena," that is, hypotheses, but to
make weak inferences to them in cases in which "deductions" have not been
achieved. But we must recognize that such inferences are weak, and we must
continue to search for phenomena from which the propositions in question
can be "deduced."
3. One sort of non-"deductive" inference to hypotheses is illustrated in
Queries 28 and 29 in the Opticks in the discussion of the particle and wave
theories of light. Here Newton seems to be making a (weak) inference to the
particle theory on the grounds that it explains a range of optical phenomena.
In section 5 above this argument is reconstructed probabilistically in such a
way as to reflect, at least in part, Newton's discussion in Queries 28 and 29, as
well as his general methodology.
4. We cannot suppose, as Newton seems to, that "deductions from phe-
nomena" will always yield the maximal certainty possible in experimental
philosophy. In the case of induction, for example, such certainty is not guar-
anteed simply by observing positive instances of an inductive generalization
and no negative ones. What we can say is that, granted certain minimal
assumptions, an increase in the number of positive instances will increase the
strength of such inferences toward maximality. Finally, assuming that proba-
bilistic explanatory reasoning of the type constructed in section 6 can be
valid, we may say this: If probabilistic arguments are not construed as "de-
ductive," then we cannot suppose, as Newton seems to, that only "deductions
from phenomena" can generate the highest certainty possible in experimental
philosophy. *
*For very helpful suggestions I am indebted to Robert Rynasiewicz, Doren Recker, Robert
Kargon, and Alan Shapiro.
This page intentionally left blank
ESSAY 3
Light Hypotheses
1. INTRODUCTION
At the beginning of the nineteenth century Thomas Young published papers
that defended the wave theory of light against the Newtonian particle theory.
Following this there occurred a lengthy and sometimes heated dispute be-
tween particle theorists and wave theorists which, it is alleged, stemmed from
deep divisions over scientific methodology. Particle theorists, it is said, partic-
ularly British ones, used the method of induction whereas wave theorists
employed the antithetical method of hypothesis. Thus Geoffrey Cantor
writes:
Although in the eighteenth century almost every British natural philosopher ac-
cepted without question the corpuscular interpretation of Newton's writings on
optics, by the 1830s most British natural philosophers had rejected Newton's cor-
puscular theory in favor of the wave theory of light. Intimately bound up with this
scientific "revolution" in optical theory was a change in scientific methodology: the
replacement of the method of induction by the method of hypothesis.1
According to Cantor, nineteenth-century particle theorists "followed the

[eighteenth-century] common-sense philosophers in considering induction to
be the proper scientific method" (p. 111), and in rejecting or limiting a
reliance on hypotheses; nineteenth-century "supporters of the wave theory,
unlike its objectors, championed the method of hypothesis" (p. 114).
Larry Laudan has also emphasized a change in methodology between the
late eighteenth and early nineteenth centuries. The wave theory of light re-
1. Geoffrey Cantor, "The Reception of the Wave Theory of Light in Britain: A Case Study
Illustrating the Role of Methodology in Scientific Debate," Historical Studies in the Physical
Sciences 6 (1975), p. 109. Emphasis mine. See also his book Optics after Newton (Manchester,
1983), pp. 177-186.
69
quired an imperceptible luminiferous ether, which eighteenth-century induc-

tivists rejected as an untestable hypothesis. Speaking of the reception of ether
theories in Scotland, Laudan writes:
The primary reason for opposition to ether theories was the widespread acceptance
among Scottish philosophers and scientists of a trenchant inductivism and empiri-
cism, according to which speculative hypotheses and imperceptible entities were
inconsistent with the search for reliable science.2
Laudan goes on to argue that nineteenth-century defenders of the wave theo-

ry used a form of the method of hypothesis (which I shall consider in this
essay).
In 1803 Henry Brougham, a defender of the particle theory of light, wrote
a scathing review of Thomas Young's "Bakerian Lecture on the Theory of
Light and Colors." Brougham begins by saying:
As this paper contains nothing which deserves the names either of experiment or
discovery, and as it is in fact destitute of every species of merit, we should have
allowed it to pass among the multitude of those articles which must always find
admittance into the collections of a Society which is pledged to publish two or three
volumes every year.3
Brougham's principal objection to Young's paper is that it is not based on

inductions from experiments but involves simply the formulation of hy-
potheses to explain various facts. And Brougham writes:
A discovery in mathematics, or a successful induction of facts, when once complet-

ed, cannot be too soon given to the world. But . . . an hypothesis is a work of
fancy, useless in science, and fit only for the amusement of a vacant hour. . . . (p.
451). It is scarcely possible to conceive a wider difference than that which subsists
between the philosophy of Newton and the philosophy of Dr. Young. While the
former utterly rejects hypotheses, and asserts that our stock of facts upon the
subject of the ether is insufficient; the latter says that we have enow [sic] of experi-
ments, and that we only require to have a stock of hypotheses, (p. 455)
In this review Brougham defends the Newtonian particle theory on the

grounds that it is inductively supported by experiments, while he rejects
2. Larry Laudan, "The Medium and its Message," in G. N. Cantor and M. J. Hodge, eds.,
Conceptions of the Ether (Cambridge, England, 1981), p. 170. Unlike Cantor, Laudan's principal
aim in this paper is to compare the methodology of those who defended the wave theory in the
nineteenth century with eighteenth- (rather than nineteenth-) century inductivist critiques of that
theory (and of ether theories generally). Since my main interest in what follows is (like that of
Cantor) in the debate between wave and particle theorists in the nineteenth century, Laudan's
claim that is of special concern to me is that nineteenth-century wave theorists utilized a method
of hypothesis.
3. Edinburgh Review 1 (1803), p. 450.
LIGHT HYPOTHESES 71
Young's defense of the wave theory on the grounds that it employs an unac-
ceptable method of hypothesis.
In what follows I propose the following:
1. To give an account of the method of hypothesis (or of various such
methods).
2. To argue, contrary to Cantor and Laudan, that in their actual practice,
as well as in their reflections on this practice, nineteenth-century wave theo-
rists such as Young, Fresnel, Lloyd, and Herschel typically employed a meth-
od that is significantly different from the method of hypothesis.
3. To argue that this method contains not only an explanatory component
present in the method of hypothesis, but an "independent warrant" compo-
nent that is not. For wave theorists the strategy for supplying independent
warrant is an eliminative one that can be justified by appeal to probabilistic
and inductive considerations. Of particular interest in this justification will be
probability considerations introduced in section 5 that pertain to the intro-
duction of auxiliary hypotheses.
4. To give an account of what nineteenth-century British particle theorists
such as Brougham meant by "induction" and how they utilized this method in
developing the particle theory. In doing this some attention will need to be
given to Newton's ideas, which exerted considerable influence on later particle
theorists.
5. To argue that there are strong similarities, if not identities, between the
inductivism of British particle theorists and the methodology of wave theo-
rists, and that the important debate is over particles versus waves, not meth-
odologies. Accordingly, the present case will not support a form of relativism
that states that fundamentally different theories employ fundamentally dif-
ferent methodologies.
2. THE METHOD OF HYPOTHESIS
On this method one proposes an hypothesis to explain observed phenomena.

If the hypothesis, if true, would correctly explain those phenomena, one can
claim support for it, even if the hypothesis postulates unobserved or unob-
servable entities and processes. Here is a simple use of this method by David
Hartley, in defending the hypothesis that an ether exists:
Let us suppose the existence of the aether, with these its properties, to be destitute
of all direct evidence, still, if it serves to explain a great variety of phenomena, it
will have an indirect evidence in its favour by this means.4
4. David Hartley, Observations on Man, His Frame, His Duty, and His Expectations, 2nd
ed., 2 vols. (London, 1791), vol. 1, p. 15. Quoted by Laudan, op. cit., p. 161.
More generally, the method of hypothesis contains the following idea:
Basic method of hypothesis: The fact that hypothesis h if true would correctly
explain observed phenomena O1, . . . ,On constitutes at least some reason to think
that h is true.
Frequently, especially in the mathematical physics developed by both wave

and particle theorists, the explanations consist of deductive derivations of C\,
. . . ,On from h. If so, we get the modern hypothetico-deductive viewpoint.
However, the method of hypothesis is also close to the idea of retroduction
introduced by Peirce at the end of the nineteenth century and developed in the
middle of the present century by N. R. Hanson. According to Hanson, it
involves an inference of the following form:
Some surprising phenomenon P is observed.

P would be explicable as a matter of course if h were true.
Hence, there is reason to think that h is true.5
Now, according to Laudan, by the 1830s an important change had oc-

curred in the method of hypothesis. Prior to this, hypothesists were willing to
conclude that there is some reason to think an hypothesis true if it explains
phenomena that have already been observed and for which the explanation
was proposed in the first place. (This would make early versions of the
method akin to Hanson's form of retroduction.) By the 1830s a requirement
was instituted that the hypothesis has to explain states of affairs significantly
different from those it was invented to explain. This can be accomplished if
the hypothesis can predict (and explain) some new and as yet unobserved
phenomenon or some known phenomenon that did not prompt the hypothe-
sis in the first place.6 Such a view is expounded by William Whewell, who
speaks in this connection of a consilience of inductions. We might formulate
it like this:
Method of hypothesis with consilience: Let h be some hypothesis proposed initially

to explain O1 (and nothing else). The fact that h if true would correctly explain O1,
5. N. R. Hanson, Patterns of Discovery (Cambridge, England, 1958), p. 72.

6. See Laudan, op. cit., p. 175. At least one exception to Laudan's thesis is Huygens, who in
1690, in the Preface to his Treatise on Light, defends a form of the method of hypothesis that
involves deriving or explaining not only known facts but new ones as well. He writes:
. . . here [in his Treatise] the principles are verified by the conclusions to be drawn from them . . . especial-
ly when there are a great number of them, and further, principally, when one can imagine and foresee new
phenomena which ought to follow from the hypotheses which one employs, and when one finds that
therein the fact corresponds to our prevision. (Christian Huygens, Treatise on Light, reprinted in Robert
Maynard Hutchins, ed., Newton, Huygens (Chicago, 1952), p. 551.)
LIGHT HYPOTHESES 73
. . . ,On (where these are significantly different) provides some reason to think that
h is true.
In point of fact, Whewell's version of the method seems even more com-
plex than this. Whewell notes that formulations like those above presuppose
that
the hypothesis with which we compare our fact [is] framed all at once, each of its
parts being included in the original scheme. In reality, however, it often happens
that the various suppositions which our system contains are added upon occasion
of different researches.7
In modifications toward a true theory, Whewell notes:
all the additional suppositions tend to simplicity and harmony; the new supposi-
tions resolve themselves into the old ones, or at least require only some easy
modification of the hypothesis first assumed: the system becomes more coherent as
it is further extended.8
Perhaps, then, Whewell would have espoused the following:
"Dynamical" method of hypothesis with consilience: Let h1 be some hypothesis

proposed initially to explain O1 but not O2, . . . ,On, where the latter are different
in kind from each other and from O1; let h2, . . . ,hk be hypotheses added to h1 to
explain O2, . . . ,On. The fact that h1, . . . ,hk if true would correctly explain O1,
. . . ,On, and in addition would correctly explain On+1, . . . ,On+p (different facts
that did not prompt h1t . . . ,hk), provides some reason to believe h1t . . . ,hk,
provided that h2, . . . ,hk are "natural" extensions of h1, so that h1, . . . ,hk has
"coherence."
These three formulations of the method, although by no means identical,

have in common the basic idea that the fact that an hypothesis if true would
correctly explain phenomena counts as some reason for believing that hypoth-
esis. There may be restrictions on the kinds of phenomena explained (e.g.,
they should be different in kind from ones that prompted the hypothesis in
the first place). And there may be restrictions on the additions to the hypothe-
sis required for subsequent explanations. But there is no requirement that the
hypothesis in question, or any subsequent one, be inductively inferable from
any observations. More generally, there is no requirement that there be any
independent warrant for the hypotheses introduced, that is, any reason for
7. William Whewell, The Philosophy of the Inductive Sciences (New York, 1967), vol. 2, p.
68. Emphasis in original.
8. Ibid., p. 68. Emphasis in original.
believing such hypotheses other than the explanatory ones hypothesists men-
tion.
3. WHAT METHOD DID NINETEENTH-CENTURY WAVE

THEORISTS EMPLOY IN PRACTICE?
In publications setting forth arguments for their theory, a strategy wave theo-
rists typically used is this:
1. Start with the assumption that light is either a wave phenomenon or a

stream of particles.
According to the wave theory, light consists of a wave motion or pulse trans-
mitted through some medium; the medium itself may be composed of parti-
cles that vibrate rather than exhibit translational motion. According to the
particle theory, light consists of discrete particles emanating from luminous
bodies; these particles are subject to forces obeying Newton's laws of motion;
if no such forces are acting the particles move in straight lines with constant
finite velocity.
The assumption that light is either a wave or particle phenomenon is made
on the grounds that these are the two main theories that have been proposed
by the physics community, or on the grounds of some empirical consideration
regarding motion, or both. Taking the former line Young writes:
It is allowed on all sides, that light either consists in the emission of very minute
particles from luminous substances, which are actually projected, and continue to
move with the velocity commonly attributed to light, or in the excitation of an
undulatory motion, analogous to that which constitutes sound, in a highly light
and elastic medium pervading the universe; but the judgments of philosophers of
all ages have been much divided with respect to the preference of one or the other
of these opinions.9
Humphrey Lloyd, after a few preliminaries, begins the body of his 1834
report on the present state of physical optics as follows:
The first property of light which claims our notice is its progressive movement.
Light we know, travels from one point of space to another in time, with a velocity
9. Thomas Young, A Course of Lectures on Natural Philosophy and the Mechanical Arts
(London, 1845), p. 359. Also taking the former line, Fresnel, like Young, begins by noting that
the particle and wave theories represent "the two systems which have up till now divided scientists
with respect to the nature of light." A. Fresnel, "Memoir on the Diffraction of Light" (1816),
reprinted (in part) in Henry Crew, ed., The Wave Theory of Light (New York, 1900). For parts of
this material not in Crew, I have used a translation provided by Laurence Selim.
LIGHT HYPOTHESES 75
of about 195,000 miles a second. The inquiry concerning the mode of this propaga-
tion involves that respecting the nature of light itself.
There are two distinct and intelligible ways of conceiving such a motion. Either
it is the self-same body which is found at different times in distant points of space;
or there are a multitude of moving bodies, occupying the entire interval, each of
which vibrates continually with certain limits, while the vibratory motion is com-
municated from one to another, and so advances uniformly. Nature affords numer-
ous examples of each of these modes of propagated movement; and in adopting
one or the other to account for the phenomena of light, we fall upon one or other
of the two rival systems, — the theories of Newton [particle theory] and of Huygens
[wave theory].10
Lloyd's assumption that light is either a wave phenomenon or a particle

phenomenon is based on the observation that light travels in space from one
point to another with a finite velocity, that both particle and wave theories
can account for this movement, and that in nature one observes motion from
one point to another occurring by the motion of a body and by the motion of
vibrations through a set of bodies.
Herschel, another defender of the wave theory, begins his account of phys-
ical optics as follows:
Among the theories which philosophers have imagined to account for the phenom-
ena of light, two principally have commanded attention; the one conceived by
Newton ... in which light is conceived to consist of excessively minute molecules
of matter projected from luminous bodies. . . . The other hypothesis is that of
Huygens . . . , which supposes light to consist, like sound, in undulations or
pulses, propagated through an elastic medium.11
Although Herschel recognizes that other theories have been proposed, he

notes that "these are the only mechanical theories which have been ad-
vanced." He seems to suppose that these are the most plausible theories.
2. Show how each theory explains various observed optical phenomena (e.g.,
rectilinear propagation, reflection, refraction, diffraction, dispersion, etc.).
Young, for example, begins with diffraction ("when a portion of light is

admitted through an aperture and spreads itself in a slight degree in every
direction"):
In this case it is maintained by Newton that the margin of the aperture possesses an
attractive force, which is capable of inflecting the rays. . . . In the Huygensian
10. Humphrey Lloyd, "Report on the Progress and Present State of Physical Optics," Reports
of the British Association for the Advancement of Science (1834), pp. 297-298.
11. J. F. W. Herschel, "Light," Encyclopedia Metropolitana (1845), vol. 4, p. 439.
system of undulation, this divergence or diffraction is illustrated by a comparison

with the motions of waves of water and of sound, both of which diverge when they
are admitted into a wide space through an aperture. . . . 12
Young continues by noting other observed optical phenomena and indicating

whether or how the two leading theories account for them. Exactly similar
strategies are followed by Fresnel, Lloyd, and Herschel.
3. Show that the particle theory, in explaining one or more of the observed
optical phenomena, introduces improbable hypotheses, while the wave
theory does not.
Immediately after introducing the particle explanation of diffraction at an

aperture as caused by an attractive force exerted at the margin, Young writes:
But there is some improbability in supposing that bodies of different forms and of
various refractive powers should possess an equal force of inflection, as they ap-
pear to do in the production of these effects; and there is reason to conclude from
experiments, that such a force, if it existed, must extend to a very considerable
distance from the surfaces concerned, at least a quarter of an inch, and perhaps
more, which is a condition not easily reconciled with other phenomena.13
Fresnel offers a similar argument against the forces introduced by the

particle theory to explain diffraction. He presents experiments on the basis of
which he concludes that
The phenomena of diffraction do not at all depend upon the nature, the mass, or
the shape of the body which intercepts the light, but only upon the size of the
intercepting body or upon the size of the aperture through which it passes. We
must, therefore, reject any hypothesis which assigns these phenomena to attractive
and repulsive forces whose action extends to a distance from the body as great as
that at which rays are inflected.14
12. Young, op. cit., pp. 359-360. Young's wave explanation of diffraction, which is rather
sketchy, is perhaps this: waves (e.g., those of sound and water) diverge when admitted into a wide
space through an aperture. Light is a wave motion. That is why it diverges similarly. It remained
for Fresnel to give a quantitative wave-theoretic explanation of this and other diffraction phe-
nomena. In one such phenomenon light bends around obstacles into the shadow as well as away
from it. Newton (to whom Young refers in the quotation above) observed the external fringes but
not the internal ones. This was one of his reasons for rejecting the wave theory (see section 7).
13. Ibid. Young proceeds to show how in explaining numbers of other phenomena the
particle theory, by contrast to the wave theory, employs dubious hypotheses. This strategy is also
evident in one of his earliest papers ("Outlines of Experiments and Inquiries Respecting Sound
and Light," Philosophical Transactions of the Royal Society, 1800), in which he also argues in
favor of the wave theory, though with somewhat less assurance (see pp. 613-616).
14. Fresnel, op. cit., p. 99.
LIGHT HYPOTHESES 77
Earlier in this prize essay on diffraction Fresnel presents an argument against

Newton's introduction of "fits of easy reflection and transmission" of light
particles as an auxiliary hypothesis to explain Newton's rings. He concludes:
Not only is the hypothesis of fits improbable because of its complexity, and diffi-
cult to reconcile with the facts in its consequences, but it does not even suffice in
explaining the phenomenon of the colored rings, for which it was imagined.15
Fresnel goes on to argue that Newton's rings, as well as diffraction, can be

explained as natural consequences of the wave theory, without introducing
improbable hypotheses.
Lloyd notes that discoveries of Bradley and Roemer lead to the conclusion
that the velocity of light is the same
whatever be the luminous origin: the light of the sun, the fixed stars, the planets
and their satellites, being all propagated with the same swiftness. This conclusion
must be allowed to present a formidable difficulty in the theory of emission.16
The difficulty is that if light consists of particles (with mass), then a massive
object such as a fixed star should exert a force on them that "would be
sufficient to destroy the whole momentum of the emitted molecules, and the
star would be invisible at great distances" (p. 300). Lloyd asserts that the only
way to explain the fact that the velocity we observe is the same for all lumi-
nous bodies is to adopt an hypothesis of Arago "that the molecules of light
are originally projected with very different velocities: but that among these
velocities is but one which is adapted to our organs of vision, and which
produces the sensation of light" (pp. 300-301). Lloyd notes that such a suppo-
sition has some support from discoveries of invisible rays of the spectrum.
But he concludes that the supposition is not "easily reconciled with hy-
potheses which we are able to frame respecting the nature of vision" (p. 301;
for additional discussion of this point, see my "Light Problems: Reply to
Chen," Studies in History and Philosophy of Science 21 (1990)). By contrast,
Lloyd asserts, the fact that the observed velocity of light is the same for all
luminous bodies is explained without difficulty by the wave theory:
This uniformity of velocity, on the other hand, is a necessary consequence of the

principles of the wave-theory. The velocity with which vibratory movement is prop-
agated in an elastic medium depends solely on the elasticity of that medium and on
its density; and if these be uniform in the vast spaces which intervene between the
material bodies of the universe, (and it is not easy to suppose it otherwise) the
velocity must be the same, whatever be the originating source. (p. 301)
15. Ibid., sec. 7.

16. Lloyd, op. cit., p. 300.
4. Conclude that the wave theory is (very probably) true, while the particle
theory is (very probably) false.
At the beginning of his paper Lloyd makes explicit the strategy he will follow:
To take, in the first instance, a rapid survey of the several leading classes of optical
phenomena which the labours of experimental philosophers have wrought out in
such rich profusion, and afterward to examine how far they are reducible to one or
other of the two rival theories which have alone advanced any claim to our consid-
eration. This is, in fact, the only way in which the truth of a physical theory can be
established; and the argument in its favor is essentially cumulative.17
Having seen how the two theories explain (or fail to explain) various optical
phenomena, and having argued that the particle theorists introduce hy-
potheses in these explanations that are improbable, whereas wave theorists do
not—or do not to such a great extent—it is concluded that the wave theory is
probably true.
4. ANALYSIS OF THIS METHOD

The method described in the previous section is a type of "eliminative" one
that consists of four parts. Schematically:
1. Assume that either theory T1 or theory T2 is correct, and give grounds for
such an assumption.
2. Show how T1 and T2 explain various observed phenomena.
3. Show that T2 in explaining one or more of these phenomena introduces
improbable hypotheses, whereas T1, does not.
4. Conclude that T1 is probably true.
Let us examine the three steps leading to the conclusion, beginning with the
second.
Step 2 introduces the idea of explanation. To conclude that T1 is probably
true, a theorist using this strategy must show that T1 is capable of explaining
certain phenomena (by producing the explanations). Following the method of
hypothesis, we might say that such a theorist seeks a theory T that will satis-
fy a
Basic explanatory condition: T if true would correctly explain observed

phenomena O1t . . . ,On.
17. Ibid., p. 295.

LIGHT HYPOTHESES 79
In the light of the discussion in section 2 this explanatory condition, which is

associated with the "basic" method of hypothesis, could in principle be
broadened to include more sophisticated features of the "dynamical method
of hypothesis with consilience." For example, it might be required that theory
T explain phenomena different in kind from those that first prompted the
theory, and that hypotheses added to T for this purpose be "natural" ones that
result in a "coherence" among the theoretical assumptions.18
Despite the fact that wave theorists satisfied one or the other explanatory
conditions of the method of hypothesis, there is a fundamental difference
between the method they employed and the method(s) of hypothesis in sec-
tion 2. In the latter case, when the explanatory condition is satisfied, it is
concluded that there is some reason to think the theory is true. But with the
method described in the previous section this is not a sufficient condition,
only a necessary one. For each of the rival theories may satisfy the explana-
tory condition. The particle theory (no less than the wave theory), together
with auxiliary assumptions needed, would, if true, correctly explain certain
observed phenomena. Indeed, it might be the case that these auxiliary as-
sumptions are "natural" extensions of the particle theory resulting in a "co-
herent" set of assumptions, and that if the enlarged theory is true it will
correctly explain phenomena different from those prompting the theory in the
first place. If so, more sophisticated "explanatory conditions" will be satis-
fied.
For example, in explaining diffraction a particle theorist introduces the
auxiliary hypothesis that the margins of the aperture exert an attractive force
on the particles of light that is capable of causing the bending of their path.
This is a "natural" extension of the particle theory, since the latter seeks to
explain optical phenomena in terms of particles subject to forces. The objec-
tion to this explanation offered by nineteenth-century wave theorists is not
that it introduces ideas foreign to the particle theory, ones that render the set
of theoretical assumptions "incoherent," but that it introduces hypotheses
that are improbable, given observations made in this case and others.
(Fresnel's objection to the Newtonian attractive force at the margins of the
aperture in diffraction is based on experiments and observations that show
that diffraction does "not at all depend upon the nature, the mass, or the
shape of the body which intercepts the light.")
If the satisfaction of an explanatory condition is not sufficient, what else is
necessary? Here we must examine steps 1 and 3.
In step 1 wave theorists begin with the assumption that light is either a
particle or a wave phenomenon, that is, that one or the other of these two
theories is true. This is not simply assumed for the sake of argument to see
18. These additional features of an explanation are emphasized by Lloyd, op. cit., pp. 349-
350.
what follows. Rather, arguing in the manner of the previous section, wave
theorists are committed to the idea that it is likely that one theory or the other
is true, and reasons are given for thinking that this is so. What sorts of
reasons are these?
Lloyd, we recall, argues "there are two distinct and intelligible ways of
conceiving" the observed motion of light, and that "in adopting one or the
other [modes of propagated motion] to account for the phenomena of light"
we obtain the particle or the wave theory. Accordingly, a part of Lloyd's
reasoning is explanatory: both theories will explain the fact that light exhibits
motion. But to this claim Lloyd adds importantly that "nature affords numer-
ous examples of each of these modes of propagated movement." Had others
been observed, presumably further ways of conceiving the propagation of
light would have been noted. Accordingly, if, as seems plausible, Lloyd means
that these are the only modes of communication known (or perhaps even just
the most prevalent ones), then, in addition to the explanatory reason, there is
an inductive one: light is observed to be communicated from one point to
another in a finite time; the modes of communication observed in nature
consist in the motion of bodies from one point to another and in the vibratory
motion of a medium; so it is reasonable to assume that light is either a
particle or a wave phenomenon.
Sometimes (as with Young) the claim that light is either a particle or a wave
phenomenon is defended by saying that these are the two leading theories
proposed by physicists. If, as I am assuming, this is to be understood as
providing some reason not simply to examine these theories but to think that
one or the other is correct, then, even in this case, the reasoning involves some
inductive steps: The fact that these are the leading theories proposed by
physicists, together perhaps with an implicit appeal to the reputations of
those supporting each theory (e.g., Newton vs. Huygens) and to their other
successes, provides some reason to think that one of these theories is true.
(Certainly a particle theorist such as Brougham had no qualms about defend-
ing his theory, at least in part, by appeal to the success of his authority
Newton. And, as noted in Essay 1, Young in an 1802 paper also appealed to
the authority of Newton in defense of some of the assumptions of the wave
theory.)
Since the claim that light is either a particle or a wave phenomenon is not
simply assumed, but grounds are given for it, and since it is recognized that
the resulting accounts are not the only possible ones,19 just the most likely, it
19. An alternative theory, which had some defenders until the early years of the nineteenth
century, was that light is produced by the rectilinear motion, rather than the vibrations, of a
fluid. According to most of these theorists this fluid consists of particles. However, unlike the
particle theorists, fluid theorists refused to offer mechanical explanations of the interactions of
light and matter, i.e., explanations in terms of attractive and repulsive forces obeying Newtonian
Chapter 4
Two-dimensional
In Chapter 3 we discussed the properties of one-dimensional electromag-

netic waves. These waves depend only on one spatial coordinate, viz. the
x3 -coordinate. We now consider the more general case of wave motion that
depends on the two spatial coordinates x1 and x3 , but that has no variation
in the x2 -direction. In this case, we assume that the permittivity, the con-
ductivity and the permeability are independent of the x2 -coordinate. Hence,
ε = ε(x1 , x3 ), σ = σ(x1 , x3 ), µ = µ(x1 , x3 ) . (4.1)
We further assume that the sources of the electromagnetic wave field are
also independent of x2 , hence
Ĵ = Ĵ (x1 , x3 , s), K̂ = K̂(x1 , x3 , s) . (4.2)
Then, the generated electromagnetic field will be independent of x2 , viz.,
Ê = Ê(x1 , x3 , s), Ĥ = Ĥ(x1 , x3 , s) , (4.3)
and, hence, ∂2 = 0. With this, Eqs. (2.54) - (2.59) separate into
∂3 Ĥ2 + (σ+sε)Ê1 = −Jˆ1ext , (4.4)

−∂1 Ĥ2 + (σ+sε)Ê3 = −Jêxt , 3 (4.5)
∂3 Ê1 − ∂1 Ê3 + sµĤ2 = −K̂2ext , (4.6)
82 two-dimensional electromagnetic waves
and
−∂3 Ê2 + sµĤ1 = −K̂1ext , (4.7)

∂1 Ê2 + sµĤ3 = −K̂3ext . (4.8)
−(∂3 Ĥ1 − ∂1 Ĥ3 ) + (σ+sε)Ê2 = −Jˆ2ext . (4.9)
In the first set of equations (4.4) - (4.6) the field components Ê1 , Ê3 and Ĥ2
occur, while in the second set of equations (4.7) - (4.9) the field components
Ĥ1 , Ĥ3 and Ê2 occur.
Parallel polarization
When K̂1ext , K̂3ext and Jˆ2ext are equal to zero in all space, the only non-
zero field components are Ê1 , Ê3 and Ĥ2 . Since the electric field vector
is parallel to the (x1 , x3 )-plane of observation, this electromagnetic field is
called to be parallelly polarized. Since the magnetic field vector is parallel
to the x2 -direction, this electromagnetic field is also denoted as the case
of H-polarization. The system of equations (4.4) - (4.6) are the governing
electromagnetic field equations. In a source-free domain, we observe that
electric field components Ê1 , Ê3 follow from the magnetic field component
Ĥ2 as
Ê1 = −(σ+sε)−1 ∂3 Ĥ2 , (4.10)

−1
Ê3 = (σ+sε) ∂1 Ĥ2 . (4.11)
In those parts of the medium where σ, ε and µ are constants, substitut-

ing these expressions for the electric field components into the source-free
counterpart of Eq. (4.6) results in the two-dimensional modified Helmholtz
equation for the magnetic field component Ĥ2 in a homogeneous domain,
viz.,
∂1 ∂1 Ĥ2 + ∂3 ∂3 Ĥ2 − (σ+sε)sµĤ2 = 0 . (4.12)

plane waves in a homogeneous medium 83
Perpendicular polarization
When Jˆ1ext , Jˆ3ext and K̂2ext are equal to zero in all space, the only non-
zero field components are Ĥ1 , Ĥ3 and Ê2 . Since the electric field vector is
perpendicular to the (x1 , x3 )-plane of observation, this electromagnetic field
is called to be perpendicularly polarized. Since the electric field vector is par-
allel to the x2 -direction, this electromagnetic field is also denoted as the case
of E-polarization. The system of equations (4.7) - (4.9) are the governing
electromagnetic field equations. In a source-free domain, we observe that
magnetic field components Ĥ1 , Ĥ3 follow from the electric field component
Ê2 as
Ĥ1 = (sµ)−1 ∂3 Ê2 , (4.13)

−1
Ĥ3 = −(sµ) ∂1 Ê2 . (4.14)
In those parts of the medium where σ, ε and µ are constants, substitut-

ing these expressions for the magnetic field components into the source-free
counterpart of Eq. (4.9) results in the two-dimensional modified Helmholtz
equation for the electric field component Ê2 in a homogeneous domain, viz.,
∂1 ∂1 Ê2 + ∂3 ∂3 Ê2 − (σ+sε)sµÊ2 = 0 . (4.15)
4.1. Plane waves in a homogeneous medium
In this section we discuss the important properties of plane electromag-

netic waves. These waves exist in a homogeneous medium. The properties
of such a medium are characterized by the spatially independent constants
σ, ε and µ. Without loss of generality, we may assume that the propagation
direction of a plane wave is in the (x1 , x3 )-plane, and as such it is a partic-
ular example of two-dimensional wave motion. It can be verified that, in a
source-free domain, Eqs. (4.10) - (4.15) admit solutions of such a form that
the electromagnetic field components depend on the spatial coordinates x1

and x3 through the factor exp(−γ1 x1 − γ3 x3 ), in which γ1 and γ3 are com-
plex constants. We call an electromagnetic wave of this type a plane wave.
In case γ1 = 0, the wave only depends on the x3 -coordinate and the theory
of Chapter 3 of one-dimensional waves applies. To study the properties of
plane waves, we write
Ê = ê(s) exp(−γ1 x1 − γ3 x3 ) , (4.16)

Ĥ = ĥ(s) exp(−γ1 x1 − γ3 x3 ) , (4.17)
in which ê = ê(s) and ĥ = ĥ(s) are spatially independent vectors. Substi-

tuting the Ĥ2 -component of Eq. (4.17) into Eq. (4.12) and the Ê2 -component
of Eq. (4.16) into Eq. (4.15), we observe that non-zero solutions exist if the
complex propagation vector γ = γ1 i1 + γ3 i3 satisfies
γ · γ = γ12 + γ32 = (σ + sε)sµ . (4.18)
For the one-dimensional waves with a propagation vector in the i3 -direction

only, Eq. (4.18) is equivalent to Eq. (3.8).
Steady-state propagation properties

For a single frequency component of angular frequency ω = 2πf , it is
customary to separate the complex propagation vector into a real vector and
and an imaginary vector according to
γ(jω) = α(ω) + jβ(ω) , (4.19)
where
α = α1 i1 + α3 i3 = attenuation vector (Np/m),

β = β1 i1 + β3 i3 = phase vector (rad/m).
. . . . . . . . . .
.... ... .... ... .... ... .... ... .... ...
... . ... .. ... . ... . ... .
.. ... ... .. ... ... ... ... ... ...
... . .
. .
. .
. .
.
... ... .... .... .... .... .... .... .... ....
.. ... .... .... .... ....
........................................................................................... .....
...
.....
..
.....
..
.....
..
.....
..
....... .. . . . .
....... .... ...
... .
.......................... β ..
... ....
. ..
.
. ..
... ....
.
. ..
... ....
.
. ..
... ....
.
. ..
... ....
.. ... .. ... . .. . ..
α ..
... .....
.. ..
... ...
... ..
... ...
... ..
... ....
... ..
... ....
...
... ... ... ... ... ... ... ... ... ...
non-uniform plane wave in a lossy medium
... ... ... ... ...

....... .......... ....... ....... .......... ....... ....... .......... ....... ....... .......... ....... ....... ........... ...
... ... ... ... ...
... ... ... ... ...
... ... ... ... ...
.. ... .
.
.
.
.
.
.
.
. ...
.
........................................................................................ ....... ........ ....... ....... ........ ....... ....... ........ ....... ....... ........ ....... ....... ........... ...
. .
. .
. .
.
... .. .. .. .. .. ...
... ... ... ...
.....
...
β ...
...
...
.
.
...
.
.
...
.
.
..
...
...
... ... .. ... ... ... .. .
......... .
....... ........ ....... ....... ........ ....... ....... ........ ....... ....... ......... ....... ....... ........... ...
. .
......
. .... .... .... .... ....
α
non-uniform plane wave in a lossless medium
.... .. .... ... .... .. .... .. ....

... .. ... . ... .. ... .. ...
.. .. .. .. ..
... ...
. ... ..... ... ...
. ... ...
. ...
... ... ... ... ...
... ... ... .. ... ... ... ... ...
.. ..
.................................................................................................... ... .. ... .. ... .. ... .. ...
.. .. ... ... ... ... ...
α β ...
...
...
.. ...
...
...
.. ...
...
...
.. ...
...
.... ...
...
... .. ... .. ... .. ... .. ...
... .. ... .. ... .. ... .. ...
... ... ... ... ... ... ... ... ...
. . . . .
... .. ... .. ... .. ... .. ...
uniform plane wave in a lossy medium
... ... ... ... ...

... ... ... ... ...
.. .. .. .. ..
... ... ... ... ...
... ... ... ... ...
... ... ... ... ...
...
..................................................................................... .... .... .... .... ....
.. .. .. .. .. ..
... ... ... ... ...
β ...
...
...
...
...
...
...
...
...
...
... ... ... ... ...
... ... ... ... ...
.. .. .. .. ..
.. .. .. .. ..
uniform plane wave in a lossless medium
Figure 4.1. Planes of equal phase (−−−−−) and planes of equal amplitude (− − −).
Substitution of Eq. (4.19) in the propagation factor exp(−γ1 x1 − γ3 x3 ) leads

to
exp(−γ1 x1 − γ3 x3 ) = exp(−α1 x1 − α3 x3 ) exp(−jβ1 x1 − jβ3 x3 ) . (4.20)
Since | exp(−γ1 x1 − γ3 x3 )| = exp(−α1 x1 − α3 x3 ) and arg[exp(−γ1 x1 − γ3 x3 )]

= −β1 x1 −β3 x3 , the first factor on the right-hand side determines the ampli-
tude of the propagation factor and is denoted as the attenuation factor, while
the second factor on the right-hand side determines the phase of the propa-
gation factor and is denoted as the phase factor. The planes α1 x1 + α3 x3 =
constant are the surfaces of equal amplitude, while the planes β1 x1 + β3 x3 =
constant are the planes of equal phase.
Substitution of Eq. (4.19) into Eq. (4.18) leads to
α · α + 2jα · β − β · β = (σ + jωε)jωµ , (4.21)
or
α · α − β · β = −ω 2 εµ , (4.22)
2α · β = ωσµ . (4.23)
Since ωσµ ≥ 0 we observe that the angle between the vectors α and β is less
than or equal to 12 π. In a lossless medium we have σ = 0 and then either
the case of α = 0 and β · β = ω 2 εµ occurs, or the case of α = 0 and β = 0
with α · β = 0 occurs. In the latter case, the planes of equal amplitudes are
perpendicular to the planes of equal phase. A plane wave is called uniform
when the vectors α and β have the same direction. A plane wave is called
non-uniform when the vectors α and β have different directions (see Fig. 4.1)
The time average Poynting’s vector of a plane wave in the frequency
domain is given by
∗

ST = 12 Re Ê × Ĥ = S 0 exp[−(γ1 +γ1∗ )x1 − (γ3 +γ3∗ )x3 ] , (4.24)
in which
∗
S 0 = 12 Re ê × ĥ . (4.25)
4.1.1. Uniform plane waves
When the plane wave is uniform, the vectors α and β have the same
direction. Let s be the unit vector in the direction of α and β. Then, we
may write
α = αs, β = βs, γ = γs , (4.26)
in which (cf. Eq. (4.19))
γ = α + jβ (4.27)
is the propagation coefficient, α is the attenuation coefficient and β is the
phase coefficient. Substitution of γ = γs into Eq. (4.18) yields, with s·s = 1,
γ 2 = (σ + jωε)jωµ , (4.28)
which is identical to Eq. (3.8). The uniform plane wave is a one-dimensional
wave in the direction of s; the vector s = {s1 , 0, s3 } can have an arbitrary
orientation with respect to the chosen coordinate system (in Chapter 2 we
have either s = i3 or s = −i3 ).
The phase factor exp(−jβ1 x1 − jβ3 x3 ) is periodic in the direction of
s = s1 i1 + s3 i3 with period
2π
λ= . (4.29)
β
This spatial period λ is denoted as the wavelength of the plane wave.
Parallelly polarized plane wave

In the case of parallel polarization, Ê1 , Ê3 and Ĥ2 are the non-zero
electromagnetic field components. Substituting Eqs. (4.16) - (4.17) into
Eqs. (4.10) - (4.11), it follows that the spatially independent components
ê1 , ê3 , ĥ2 and γ satisfy
γ3
ê1 = ĥ2 (4.30)
σ + sε
and
−γ1
ê3 = ĥ2 . (4.31)
σ + sε
A particular linear combination of Eqs. (4.30) - (4.31) yields
γ1 ê1 + γ3 ê3 = γ · ê = 0 . (4.32)
ê......
............
.......
.....
..... ....
s
.....
..... ..............
..... .......
..... .....
..... ..
......
..... ....
..... .....
..... ĥ .....
r
f
.....
....
..... .. ...........
i.1 .......
..... .
. .
. .....
.....
........ ......
..... .
....... x
.... . .
...... ..
.....
... .......
.... .......
.......
.. ......
... ..... .
.... ..
... ... ...
... ....... ....
.......
r ... .......
.................................................................
i3
O
Figure 4.2. The field components of a parallelly polarized,

uniform plane wave propagating in the (x1 , x3 )-plane.
When the plane wave is uniform, we may use γ = γs = γs1 i1 + γs3 i3 . Then,
Eqs. (4.30) - (4.32) can be written as
ê1 = Z ĥ2 s3 , (4.33)

ê3 = −Z ĥ2 s1 , (4.34)
and
s1 ê1 + s3 ê3 = s · ê = 0 , (4.35)
where we have used the expression for the wave impedance

1
γ sµ 2
Z(s) = = . (4.36)
σ + sε σ + sε
For a parallelly polarized, uniform, plane wave, we observe that the electric
field strength ê = {ê1 , 0, ê3 }, the magnetic field strength ĥ = {0, ĥ2 , 0} and
the direction of propagation s = {s1 , 0, s3 } form a mutually perpendicular,
right-handed triad (see Fig. 4.2).
Further, in the frequency domain, the Poynting vector S 0 defined in
Eq. (4.25) becomes
∗

S0 = 1
2 Re ê × ĥ = 12 Re −ê3 ĥ∗2 i1 + ê1 ĥ∗2 i3
1 ∗
= 2 Re [Z(jω)] ĥ2 ĥ2 s . (4.37)
Hence, the power flow of the parallelly polarized, uniform, plane wave is in
the direction of propagation.
Perpendicularly polarized plane wave

In the case of perpendicular polarization, Ê2 , Ĥ1 and Ĥ3 are the non-
zero electromagnetic field components. Substituting Eqs. (4.16) - (4.17) into
Eqs. (4.13) - (4.14), it follows that the spatially independent components ê2 ,
ĥ1 , ĥ3 and γ satisfy
−γ3
ĥ1 = ê2 (4.38)
sµ
and
γ1
ĥ3 = ê2 . (4.39)
sµ
A particular linear combination of Eqs. (4.38) - (4.39) yields
γ1 ĥ1 + γ3 ĥ3 = γ · ĥ = 0 . (4.40)
When the plane wave is uniform, we may use γ = γs = γs1 i1 + γs3 i3 . Then,
Eqs. (4.38) - (4.40) can be written as
ĥ1 = −Y ê2 s3 , (4.41)

ĥ3 = Y ê2 s1 , (4.42)
and
s1 ĥ1 + s3 ĥ3 = s · ĥ = 0 , (4.43)
where we have used the expression for the wave admittance

1
γ σ + sε 2
Y (s) = = . (4.44)
sµ sµ
For a perpendicularly polarized, uniform, plane wave, we see that the electric
field strength ê = {0, ê2 , 0}, the magnetic field strength ĥ = {ĥ1 , 0, ĥ3 } and
the direction of propagation s = {s1 , 0, s3 } form a mutually perpendicular,
right-handed triad (see Fig. 4.3).
...
s
...................
..
.....
.
..
......
.
.....
.....
....
x rf .....
.....ê
.......
....... ....
i.1 ......
......
....... .....
.....
.....
..... . .....
.... .... .. .....
....... ... ...
.... .
...
... .... .....
.....
..... .....
... .... .. .....
.... ..
.. ... ... .....
.... .. .... ..... .
.. .
...
... .... ................
...
.... .......
......
....... ..
ĥ
.......
r .. ......
.................................................................
i3
O
Figure 4.3. The field components of a perpendicularly polarized,

uniform plane wave propagating in the (x1 , x3 )-plane.
Further, in the frequency domain, the Poynting vector S 0 defined in

Eq. (4.25) becomes
∗
S0 = 2 Re ê × ĥ
1
= 12 Re ê∗ × ĥ = 12 Re ê∗2 ĥ3 i1 − ê∗2 ĥ1 i3
1 ∗
= 2 Re [Y (jω)] ê2 ê2 s . (4.45)
Hence, the power flow of the perpendicularly polarized, uniform, plane wave
is in the direction of propagation.
4.2. Interference of two plane waves
When in a linear medium two or more waves are present simultaneously,

the total value of a field strength in that domain is equal to the contributions
of the constituent wavefields (the principle of superposition). Even when the
constituent wavefields have a simple structure, the resulting wavefield may
depend on space and time in a complicated way. This phenomenon is called
interference. In this section we discuss the interference of two plane waves in
a homogeneous medium with constants σ, ε and µ. The first wave is denoted
as wave (1) with field strengths
interference of two plane waves 91
(1) (1) (1)

Ê = ê(1) exp(−γ1 x1 − γ3 x3 ) , (4.46)
(1) (1) (1) (1)
Ĥ = ĥ exp(−γ1 x1 − γ3 x3 ) . (4.47)
The second wave is denoted as wave (2) with field strengths
(2) (2) (2)

Ê = ê(2) exp(−γ1 x1 − γ3 x3 ) , (4.48)
(2) (2) (2) (2)
Ĥ = ĥ exp(−γ1 x1 − γ3 x3 ) . (4.49)
Since both waves are present in the same medium, we have
(1) (1) (2) (2)

[γ1 ]2 + [γ3 ]2 = [γ1 ]2 + [γ3 ]2 = (σ + sε)sµ . (4.50)
In the domain where interference occurs, the total field strengths are given
by
(1) (2)
Ê = Ê + Ê , (4.51)
(1) (2)
Ĥ = Ĥ + Ĥ . (4.52)
We now assume that both waves are uniform plane waves. Then
(1) (1) (1) (1)

γ1 = γs1 , γ3 = γs3 , (4.53)
and
(2) (2) (2) (2)
γ1 = γs1 , γ3 = γs3 , (4.54)
where
1
γ = [(σ + sε)sµ] 2 . (4.55)
i.1
...
..........
...
...
...
s(2) .....
...
... s(1) ....
...............
.......
... ...................
..... ... ..... .
.
..... ... ..
...
..... .... .....
..... .....
... .. ......
... ....... .. ....... ....
θ r θ
..... ..................................................................................................
.... i3
O
Figure 4.4. The propagation directions of the two uniform plane waves.
Without loss of generality we always can choose such a reference frame that
the two waves both propagate in the positive x1 -direction, while they have
opposite propagation directions in the x3 -direction (see Fig. 4.4). Let θ be
the angle between s(1) and i3 and π − θ be the angle between s(2) and i3 .
Then,
(1) (1)
s(1) = s1 i1 + s3 i3 = sin(θ)i1 + cos(θ)i3 , (4.56)
and
(2) (2)
s(2) = s1 i1 + s3 i3 = sin(θ)i1 − cos(θ)i3 . (4.57)
Substituting Eqs. (4.46) - (4.49) in Eqs. (4.51) - (4.52), we obtain

Ê = exp[−γ sin(θ)x1 ] ê(1) exp[−γ cos(θ)x3 ] + ê(2) exp[γ cos(θ)x3 ] , (4.58)

(1) (2)
Ĥ = exp[−γ sin(θ)x1 ] ĥ exp[−γ cos(θ)x3 ] + ĥ exp[γ cos(θ)x3 ] . (4.59)
In order to discuss the total electromagnetic power flow, we consider the

lossless case and restrict ourselves to the steady-state analysis.
4.2.1. Steady-state analysis: lossless case
In the lossless case with s = jω, we have

ω
γ = j , with c = (εµ)− 2 .
1
(4.60)
c
The time average Poynting vector is given by
∗

ST = 12 Re Ê × Ĥ , (4.61)
interference of two plane waves 93
in which we have to substitute the expressions for the electric and magnetic
field strengths of Eqs. (4.58) and (4.59). We arrive at

(1)∗ (2)∗
ST = 1
2 Re ê (1)
× ĥ + ê (2)
× ĥ

(2)∗
+ 12 Re ê(1) × ĥ exp[−2jω cos(θ)x3 /c]

(1)∗
+ ê (2)
× ĥ exp[2jω cos(θ)x3 /c] . (4.62)
The first two terms of the right-hand side of Eq. (4.62) represent the power
flows of the individual plane waves. In view of the obvious presence of the
third and fourth term we conclude that the superposition principle only
applies to the electromagnetic field strengths, but not to the power flow.
In order to quantify the electromagnetic power flow of two interfering
waves we consider the cases of parallel and perpendicular polarization sepa-
rately.
Parallelly polarized plane waves

Substituting Eq. (4.33) and (4.34) into Eq. (4.62), while observing that
in the lossless case,

1
µ 2
Z= (4.63)
ε
is real valued, we obtain

(1) (1)∗ (1) (2) (2)∗ (2)
ST = 1
2Z ĥ2 ĥ2 s + ĥ2 ĥ2 s

(1) (2)∗
+Z Re ĥ2 ĥ2 sin(θ) exp[−2jω cos(θ)x3 /c] i1 . (4.64)
Perpendicularly polarized plane waves

Substituting Eq. (4.41) and (4.42) into Eq. (4.62), while observing that
in the lossless case,

1
ε 2
Y = (4.65)
µ
is real valued, we obtain

(1) (1)∗ (1) (2) (2)∗ (2)
ST = 1
2Y ê2 ê2 s + ê2 ê2 s

(1) (2)∗
+Y Re ê2 ê2 sin(θ) exp[−2jω cos(θ)x3 /c] i1 . (4.66)
From the results of these cases of polarization, it follows that the power
flow in the i3 -direction is position invariant, but the power flow in the i1 -
direction is periodic in x3 with period
λ
λinterference = , (4.67)
2 cos(θ)
where λ is the wavelength of the uniform plane waves, see Eq. (4.29).
S1 T
....
........
..
...
...
...
...
..
.... ..... .... .... .... .... .... .... .... ..... ..
... ... .. ... .... ... .... ... .... ... .... ... .... ... .... ... ... ... .. ...
.. .. .. ... .. ... .. ... .. ... .. ... .. ... .. .. .. .. ... ..
... .. ... .. ... .. ... .. ... .. .. .. .. .. .. .. ... .. .. ..
... ... ....
.
. .. ....
. .. ....
. .. ....
. .. ....
. .. ....
. .. ....
. ... ....
. ... .....
. ...
.
.
.. .. .
. .. .
. .. .
. .. .
. .
.. .
. .
.. .
. .. .
. .. . .. .
... ... ... .. ... .. .. .. ... .. ... .. ... .. ... .. ... ... ... ..
... ... ... ..
..
... ... ...
... ... ... ... ... ... ... ... ... ... ... ... ... ...
... .. ... ... .. .. ... .. ... .. ... .. ... .. ... .. ... ..
... ..
.
. ... ...
.
. ... ..
. ... ..
. .
. ..
. .
. ..
. .
. ..
. ... ..
. ... ..
.
. .
. ..
.
.
... ... . ... . ... . ... . ... . ... . ... ...
.. .. ... .. ... ... ... ... ... .. ...
... .. ... .. ... .. ... ... ... ... ...
...
... .. ...
...
... .... ... ... ... .... ... ... ... ... ... ... ... ... ... ... .... ...
... .. ... ... ... ... ... .... ... .... ... .... ... .... ... .... ... .. ... ....
... .. ... ... ... .. . .
.
... .. .. .
.
... .. .. .
... .. .. .
... .. .. .
... .. .. .
... . .. .
. .
.
.. ... .. .. ... ..
... .. ... .. ... .. .. .. .. .. ... .. ... .. ... .. ... ... ... ...
.... .... ...... ....... ....... ....... ...... ...... ...... ....
............................................ x3
...................... .....................
λ
2 cos(θ)
Figure 4.5. The power flow in the i1 -direction as a function of x3 in case of

interference of two uniform plane waves propagating in directions given in Fig. 4.4.
reflection by an electrically impenetrable half-space 95
4.3. Reflection of a plane wave by an electrically

impenetrable half-space
When a wave is incident upon the boundary surface of an electrically

impenetrable object, the wave is reflected by this boundary surface. In the
case that the incident wave is a plane wave and the reflecting boundary
surface is a plane boundary of an electrically impenetrable, semi-infinite
medium, the reflected wave is a plane wave as well. Then, the pertaining
reflection problem can be solved with the theory of plane waves.
We take the Cartesian coordinate system in such a way that the plane
x3 = 0 coincides with the boundary plane of the electrically impenetrable
medium. The half-space −∞ < x3 < 0 consists of a homogeneous medium
with constitutive constants σ, ε and µ (see Fig. 4.6). The incident (plane)
wave is given by (cf. Eqs. (4.16) - (4.17))
i
Ê = êi exp(−γ1i x1 − γ3i x3 ) , (4.68)
i i
Ĥ = ĥ exp(−γ1i x1 − γ3i x3 ) , (4.69)
while the reflected wave is given by
r
Ê = êr exp(−γ1r x1 − γ3r x3 ) , (4.70)
r r
Ĥ = ĥ exp(−γ1r x1 − γ3r x3 ) , (4.71)
in which
(γ1i )2 + (γ3i )2 = (γ1r )2 + (γ3r )2 = (σ + sε)sµ . (4.72)
From this equation, it is observed that for given values of γ1i two solutions
of γ3i exist. In view of causality with respect to the incident wave we take
1
γ3i = (σ + sε)sµ − (γ1i )2 2
, (4.73)
where Re[γ3i ] ≥ 0. Similarly, in view of causality with respect to the reflected

wave we take
σ, ε, µ electrically
impenetrable
medium
.................
............
.... ........
.......
........
.......
.......
reflected wave ........
.
............................
i3
.........................
incident wave .. .
....... ...
........
.
...
.........
.
...
.......
......
x3 < 0 x3 = 0 x3 > 0
Figure 4.6. Reflection of a plane wave by a plane reflector.
1
γ3r = − (σ + sε)sµ − (γ1r )2 2
, (4.74)
where Re[γ3r ] ≤ 0. The total field in the half-space −∞ < x3 < 0 is obtained
as
i r i r
Ê = Ê + Ê , Ĥ = Ĥ + Ĥ . (4.75)
At x3 = 0 the electromagnetic field has to satisfy the boundary condition

that the tangential component of the electric field strength vanishes, i.e,
i r
lim i3 × (Ê + Ê ) = 0 , (4.76)
x3 ↑0
or
lim (Ê1i + Ê1r ) = 0 , (4.77)

x3 ↑0
lim (Ê2i + Ê2r ) = 0 . (4.78)

x3 ↑0
We now discuss the cases of parallel and perpendicular polarization sepa-

rately.
In the case of parallel polarization, Ê1 , Ê3 and Ĥ2 are the non-zero elec-
tromagnetic field components. We only have to meet the boundary condition
of Eq. (4.77) for Ê1 . Using Eq. (4.10) and (4.75), the boundary condition at
x3 = 0 becomes
lim ∂3 (Ĥ2i + Ĥ2r ) = 0 . (4.79)
x3 ↑0
Substituting the plane-wave representations of Eqs. (4.69) and (4.71) into

Eq. (4.79), we obtain
γ3i ĥi2 exp(−γ1i x1 ) + γ3r ĥr2 exp(−γ1r x1 ) = 0 . (4.80)
This equation can only be satisfied for all x1 , if
γ1r = γ1i , (4.81)
while from Eqs. (4.73) - (4.74) it follows that
γ3r = −γ3i . (4.82)
But, in that case, the boundary condition of Eq. (4.80) also requires
ĥr2 = ĥi2 . (4.83)
The two other non-zero vectors follow with the help of Eqs. (4.10) - (4.11)
as
êr1 = −êi1 , (4.84)

êr3 = êi3 . (4.85)
At this moment, the parallelly polarized reflected wave is determined com-

pletely.
zero electromagnetic field components. We only have to meet the boundary
condition of Eq. (4.78) for Ê2 . Substituting the plane-wave representations
of Eqs. (4.68) and (4.70) into Eq. (4.78), we obtain
êi2 exp(−γ1i x1 ) + êr2 exp(−γ1r x1 ) = 0 . (4.86)
This equation can only be satisfied for all x1 , if
γ1r = γ1i , (4.87)
while from Eqs. (4.73) - (4.74) it follows that
γ3r = −γ3i . (4.88)
The boundary condition of Eq. (4.86) also requires
êr2 = −êi2 . (4.89)
The two other non-zero vectors follow with the help of Eqs. (4.13) - (4.14)
as
ĥr1 = ĥi1 , (4.90)

ĥr3 = −ĥi3 . (4.91)
At this moment, the perpendicularly polarized reflected wave is determined

completely.
σ, ε, µ electrically
impenetrable
medium
.......
....... sr si
..................
............. .........................
... ....... ... .
........ ....... ...
............... ..........
θr ....
... ....... .............
. ... .........................................................................
.. .........
... ...... .
θi .
...... .... .. .
. .
i3
.......
.......
.......
Figure 4.7. Reflection of a uniform plane wave.
Uniform plane waves

When the incident wave is a uniform plane wave, we may write
1
γ i = γ1i i1 + γ3i i3 = [(σ + sε)sµ] 2 si , (4.92)
where si is the unit vector in the direction of propagation of the incident

wave. On account of Eqs. (4.81) - (4.82), or Eqs. (4.87) - (4.88), we may
also write
1
γ r = γ1r i1 + γ3r i3 = [(σ + sε)sµ] 2 sr , (4.93)
where sr is the unit vector in the direction of propagation of the reflected
wave. The reflected wave is also a uniform plane wave. The angle θi between
the vectors si and i3 is called the angle of incidence, while the angle θr
between the vectors sr and −i3 is called the angle of reflection (see Fig. 4.7).
From Eqs. (4.81) - (4.82), or Eqs. (4.87) - (4.88), we simply obtain
θr = θi . (4.94)
In the case of steady-state (s = jω), the total wave field, being the su-
perposition of the incident wave and the reflected wave, has the character of
a travelling wave in the i1 -direction and a standing wave in the i3 -direction.
This phenomenon is clearly observable when the medium in the half-space
−∞ < x3 < 0 is lossless. Then
γ1i = jωsi1 /c , γ3i = jωsi3 /c , (4.95)

γ1r = jωsi1 /c , γ3r = −jωsi3 /c , (4.96)
where c = (εµ)− 2 . Combining all the previous results, we conclude that the
1
total electromagnetic field is given by
Ê1 = −2jêi1 exp(−jωsi1 x1 /c) sin(ωsi3 x3 /c) , (4.97)

Ê2 = 0 , (4.98)
Ê3 = 2êi3 exp(−jωsi1 x1 /c) cos(ωsi3 x3 /c) , (4.99)
Ĥ1 = 0 , (4.100)
Ĥ2 = 2ĥi2 exp(−jωsi1 x1 /c) cos(ωsi3 x3 /c) , (4.101)
Ĥ3 = 0 , (4.102)
for the case of parallel polarization, and
Ê1 = 0 , (4.103)
Ê2 = −2jêi2 exp(−jωsi1 x1 /c) sin(ωsi3 x3 /c) , (4.104)
Ê3 = 0 , (4.105)
Ĥ1 = 2ĥi1 exp(−jωsi1 x1 /c) cos(ωsi3 x3 /c) , (4.106)

Ĥ2 = 0 , (4.107)
Ĥ3 = −2j ĥi3 exp(−jωsi1 x1 /c) sin(ωsi3 x3 /c) , (4.108)
for the case of perpendicular polarization.

reflection and transmission of a plane wave 101
4.4. Reflection and transmission of a plane wave

incident upon a plane interface
When a wave is incident upon the interface between two different media,
the wave is partly reflected by this interface and partly transmitted through
this interface. In the case that the incident wave is a plane wave and the
interface is a plane boundary between two semi-infinite media, the reflected
and transmitted waves are plane waves as well. Then, the pertaining reflec-
tion problem can be solved with the theory of plane waves.
We take the Cartesian coordinate system in such a way that the plane
x3 = 0 coincides with the interface between the two media. The half-space
−∞ < x3 < 0 consists of a homogeneous medium with constitutive constants
σ (1) , ε(1) and µ(1) and the half-space 0 < x3 < ∞ consists of a homogeneous
medium with constitutive constants σ (2) , ε(2) and µ(2) (see Fig. 4.8).
The incident (plane) wave is given by (cf. Eqs. (4.16) - (4.17))
i
Ê = êi exp(−γ1i x1 − γ3i x3 ) , (4.109)
i i
Ĥ = ĥ exp(−γ1i x1 − γ3i x3 ) , (4.110)
while the reflected wave is given by
r
Ê = êr exp(−γ1r x1 − γ3r x3 ) , (4.111)
r r
Ĥ = ĥ exp(−γ1r x1 − γ3r x3 ) , (4.112)
in which
(γ1i )2 + (γ3i )2 = (γ1r )2 + (γ3r )2 = (σ (1) + sε(1) )sµ(1) . (4.113)
In view of causality with respect to the incident wave we take
1
γ3i = (σ (1) + sε(1) )sµ(1) − (γ1i )2
2
, (4.114)
σ (1) , ε(1) , µ(1) σ (2) , ε(2) , µ(2)
.................
............
... ........
transmitted wave
....... ...............
........ ..................
....... ............. ....
....... .............
.............
reflected wave ........
. ....
...
...
...
..
............................
i3
.........................
incident wave .. .
....... ...
.......
.
...
..........
..
...
......
.
........
.......
.......
.......
.
...
..........
.
x3 < 0 x3 = 0 x3 > 0
Figure 4.8. Reflection and transmission of a plane wave by a plane interface

between two different media.
where Re[γ3i ] ≥ 0. Similarly, in view of causality with respect to the reflected

wave we take 1
γ3r = − (σ (1) + sε(1) )sµ(1) − (γ1r )2 2
, (4.115)
where Re[γ3r ] ≤ 0. The total field in the half-space −∞ < x3 < 0 is obtained
as
i r i r
Ê = Ê + Ê , Ĥ = Ĥ + Ĥ . (4.116)
The transmitted wave is given by
t
Ê = êt exp(−γ1t x1 − γ3t x3 ) , (4.117)
t t
Ĥ = ĥ exp(−γ1t x1 − γ3t x3 ) , (4.118)
in which
(γ1t )2 + (γ3t )2 = (σ (2) + sε(2) )sµ(2) . (4.119)
In view of causality with respect to the transmitted wave we take
1
γ3t = (σ (2) + sε(2) )sµ(2) − (γ1t )2 2
, (4.120)
where Re[γ3t ] ≥ 0.
At x3 = 0 the electromagnetic field has to satisfy the boundary conditions
that the tangential components of the electric and magnetic field strengths
have to be continuous, i.e.,
i r t
lim i3 × (Ê + Ê ) = lim i3 × Ê , (4.121)
x3 ↑0 x3 ↓0
i r t
lim i3 × (Ĥ + Ĥ ) = lim i3 × Ĥ , (4.122)
x3 ↑0 x3 ↓0
or
lim (Ê1i + Ê1r ) = lim Ê1t , (4.123)

x3 ↑0 x3 ↓0
lim (Ê2i + Ê2r ) = lim Ê2t , (4.124)

x3 ↑0 x3 ↓0
lim (Ĥ1i + Ĥ1r ) = lim Ĥ1t , (4.125)

x3 ↑0 x3 ↓0
lim (Ĥ2i + Ĥ2r ) = lim Ĥ2t . (4.126)

x3 ↑0 x3 ↓0
We now discuss the cases of parallel and perpendicular polarization sepa-

rately.
In the case of parallel polarization, Ê1 , Ê3 and Ĥ2 are the non-zero
electromagnetic field components. We only have to meet the boundary con-
ditions of Eq. (4.123) for Ê1 and Eq. (4.126) for Ĥ2 . Using Eqs. (4.10) and
(4.116) the boundary conditions at x3 = 0 become
1 1
lim ∂3 (Ĥ2i + Ĥ2r ) = lim ∂3 Ĥ2t , (4.127)
x3 ↑0 σ (1) + sε(1) x3 ↓0 σ (2) + sε(2)
lim (Ĥ2i + Ĥ2r ) = lim Ĥ2t . (4.128)
x3 ↑0 x3 ↓0
Substituting the plane-wave representations of Eqs. (4.110), (4.112) and

(4.118) into Eqs. (4.127) - (4.128), we obtain
γ3i γ3r
ĥi exp(−γ1i x1 ) + ĥr exp(−γ1r x1 )
σ (1) + sε(1) 2 σ (1) + sε(1) 2
γ3t
= ĥt exp(−γ1t x1 ) , (4.129)
σ (2) + sε(2) 2
ĥi2 exp(−γ1i x1 ) + ĥr2 exp(−γ1r x1 ) = ĥt2 exp(−γ1t x1 ) . (4.130)
Both equations can only be satisfied for all x1 , if
γ1t = γ1r = γ1i (4.131)
while from Eqs. (4.114), (4.115) and (4.120) it follows that
γ3r = −γ3i , (4.132)
1
γ3t = (σ (2) + sε(2) )sµ(2) − (γ1i )2 2
. (4.133)
But, in that case, the boundary conditions of Eqs. (4.129) - (4.130) also
require
γ3i i γ3r r γ3t

ĥ 2 + ĥ 2 = ĥt , (4.134)
σ (1) + sε(1) σ (1) + sε(1) σ (2) + sε(2) 2
ĥi2 + ĥr2 = ĥt2 . (4.135)
Let us now introduce the reflection coefficient R = R (s) and the transmis-
sion coefficient T = T (s) via the relations
ĥr2 = R ĥi2 , (4.136)

ĥt2 = T ĥi2 . (4.137)
Substituting the latter two expressions into Eqs. (4.134) - (4.135), divid-
ing the resulting two equations by the non-zero quantity ĥi2 , and using
Eq. (4.132), we arrive at the system of two algebraic equations,
γ3i γ3i γ3t

− R = T , (4.138)
σ (1) + sε(1) σ (1) + sε(1) σ (2) + sε(2)
1 + R = T , (4.139)
from which R and T are obtained as
γ3i γ3t
(1) (1)
−
R = σ +i sε σ (2) + sε(2) , (4.140)
γ3 γ3t
+
σ (1) + sε(1) σ (2) + sε(2)
γ3i
2
T = σ (1) + sε(1) . (4.141)
γ3i γ3t
+
σ (1) + sε(1) σ (2) + sε(2)
At this moment, the parallelly polarized reflected and transmitted wave are
determined completely.
In the special case of normal incidence, we have γ1i = 0. Then, we have
arrived at the reflection and transmission coefficients of the one-dimensional
waves (cf. Eqs. (3.95) derived from a magnetic field analysis), viz.,
Z (1) − Z (2)
R = , (4.142)
Z (1) + Z (2)
2Z (1)
T = , (4.143)
Z (1) + Z (2)
in which
1
(1) γ (1) sµ(1) 2
Z = (1) = (4.144)
σ + sε(1) σ (1) + sε(1)
and 1
(2) γ (2) sµ(2) 2
Z = (2) = (4.145)
σ + sε(2) σ (2) + sε(2)
are the wave impedances in medium (1) and (2), respectively.
zero electromagnetic field components. We only have to meet the boundary
conditions of Eq. (4.124) for Ê2 and Eq. (4.125) for Ĥ1 . Using Eqs. (4.13)
and (4.116) the boundary conditions at x3 = 0 become
lim (Ê2i + Ê2r ) = lim Ê2t , (4.146)

x3 ↑0 x3 ↓0
1 1
lim ∂3 (Ê2i + Ê2r ) = lim ∂3 Ê2t . (4.147)
x3 ↑0 sµ(1) x3 ↓0 sµ(2)
Substituting the plane-wave representations of Eqs. (4.109), (4.111) and

(4.117) into Eqs. (4.146) - (4.147), we obtain
êi2 exp(−γ1i x1 ) + êr2 exp(−γ1r x1 ) = êt2 exp(−γ1t x1 ) , (4.148)

i
γ3 i r
γ3 r t
γ3 t
ê2 exp(−γ1i x1 ) + (1) ê2 exp(−γ1r x1 ) = ê exp(−γ1t x1 ) . (4.149)
sµ (1) sµ sµ(2) 2
Both equations can only be satisfied for all x1 , if
γ1t = γ1r = γ1i , (4.150)
while from Eqs. (4.114), (4.115) and (4.120) it follows that
γ3r = −γ3i , (4.151)
1
γ3t = (σ (2) + sε(2) )sµ(2) − (γ1i )2 2
. (4.152)
But, in that case, the boundary conditions of Eqs. (4.148) - (4.149) also
require
êi2 + êr2 = êt2 , (4.153)

γ3i i γ3r r γ3t t
ê 2 + ê 2 = ê . (4.154)
sµ(1) sµ(1) sµ(2) 2
Let us now introduce the reflection coefficient R⊥ = R⊥ (s) and the trans-
mission coefficient T⊥ = T⊥ (s) via the relations
êr2 = R⊥ êi2 , (4.155)

êt2 = T⊥ êi2 . (4.156)
Substituting the latter two expressions into Eqs. (4.153) - (4.154), divid-
ing the resulting two equations by the non-zero quantity êi2 , and using
Eq. (4.151) we arrive at the system of two algebraic equations,
1 + R⊥ = T⊥ , (4.157)
γ3i γ3i γ3t
− (1) R⊥ = T⊥ , (4.158)
sµ(1) sµ sµ(2)
from which R⊥ and T⊥ are obtained as
γ3i γ3t
−
µ(1) µ(2)
R⊥ = i , (4.159)
γ3 γ3t
+
µ(1) µ(2)
γ3i
2
µ(1)
T⊥ = . (4.160)
γ3i γ3t
+
µ(1) µ(2)
At this moment, the perpendicularly polarized reflected and transmitted

wave are determined completely.
In the special case of normal incidence, we have γ1i = 0. Then, we have

arrived at the reflection and transmission coefficients of the one-dimensional
waves (cf. Eqs. (3.85) derived from an electric field analysis), viz.,
Y (1) − Y (2)
R⊥ = , (4.161)
Y (1) + Y (2)
2Y (1)
T⊥ = , (4.162)
Y (1) + Y (2)
in which 1
σ (1) + sε(1) σ (1) + sε(1) 2
Y (1) = = (4.163)
γ (1) sµ(1)
and 1
(2) σ (2) + sε(2) σ (2) + sε(2) 2
Y = = (4.164)
γ (2) sµ(2)
are the wave admittances in medium (1) and (2), respectively.
4.4.1. Uniform plane waves in the frequency domain
We restrict ourselves to the analysis in the freqency domain with s = jω.

When the incident wave is a uniform plane wave, we may write
1
γ i = γ1i i1 + γ3i i3 = [(σ (1) + jωε(1) )jωµ(1) ] 2 si , (4.165)
where si is the unit vector in the direction of propagation of the incident

wave. On account of Eqs. (4.131) - (4.132), or Eqs. (4.150) - (4.151), we may
also write
1
γ r = γ1r i1 + γ3r i3 = [(σ (1) + jωε(1) )jωµ(1) ] 2 sr , (4.166)
where sr is the unit vector in the direction of propagation of the reflected
wave. The reflected wave is also a uniform plane wave. The angle θi between
the vectors si and i3 is called the angle of incidence, while the angle θr
between the vectors sr and −i3 is called the angle of reflection (see Fig. 4.9).
From Eqs. (4.131) - (4.132), or Eqs. (4.150) - (4.151), we simply obtain
θr = θi . (4.167)
This relation is identical to Eq. (4.94) of Section 4.3 dealing with the reflec-
tion by an electrically impenetrable half-space. This relation is called Snell’s
law of reflection.
Subsequently, we investigate the conditions under which the transmitted
wave is a uniform plane wave. Let us assume that
1
γ t = γ1t i1 + γ3t i3 = [(σ (2) + jωε(2) )jωµ(2) ] 2 st , (4.168)
where st is the unit vector in the direction of propagation of the transmitted

wave. From Eq. (4.131), or Eq. (4.150), it follows that
1 1
[(σ (1) + jωε(1) )jωµ(1) ] 2 si1 = [(σ (2) + jωε(2) )jωµ(2) ] 2 st1 . (4.169)
In the general case, real values of si1 and st1 do not meet this equation, and
in general the transmitted wave is a non-uniform plane wave. However, for
lossless media, Eq. (4.169) can be satisfied for real values of si1 and st1 .
Lossless media
In the case of lossless media, we introduce the (real-valued) index of
refraction n(1) of medium (1) as
1
n(1) = c0 [ε(1) µ(1) ] 2 , (4.170)
and the index of refraction n(2) of medium (2) as

1
n(2) = c0 [ε(2) µ(2) ] 2 . (4.171)
Then, we have (cf Eq. (4.169))
n(1) si1 = n(2) st1 . (4.172)
The angle θt between the vectors st and i3 is called the angle of transmission
(see Fig. 4.9). From Eq. (4.172) we obtain Snell’s law of transmission (also
called Snell’s law of refraction),
n(1) sin(θi ) = n(2) sin(θt ) . (4.173)

ε(1) , µ(1) ε(2) , µ(2) ε(1) , µ(1) ε(2) , µ(2)

n(2) ..
n(1)
....... ..
.....
...... ......
. .
1 ....... .......
.......
. ...... ......
.
.....
.. ..
.... n(2) n(2) t
...... 1 ....... ....... ....... ....
. n(1) s
... ...
. . .
...... ......
si
..
n(2) t ... n (1) ......
sr....................... .
... ...
.
.................... ............................. n(1)
s sr........................ ..............
......... ... ...........
...
................. si
. .. ...... ..
.... ....................
. . ... . ... ................. ..... .. ......... . .....
... ...... . ... ...... ......... .....
θr .. .............
.. ..
θt θt
. ........ ..
............... ... ..... ........
..... θr .
. ... ..
.. .... . ..
.... . ..............................................................................
... .... . .. ..
. ..
.. ...
.
...
.... .
...
.........
......... ..........................
......... .
..
.. ...... ...
....... .......................................................................
... ...
...
.. θi ...... i3 .
.
.. .
.
..
...
..
θi
...
... .......
....
.......
.
.
..
i3 .
...
... ....
...
..
.. .. ... ....... ...
...... .... ... .. . . ...... ...
.
..
. .
.
.....
... .....
..
... ....... ... .
. . .. .. . .. ...... ..
.
......
... .
...... ....... ..... .. ..... ....... ....
... ..
. .... .. .
.
...... . ...... . ..
.
...... . ......
.. . . . . .......
....... ....... ....... ....... ...... .......
(a) n(1) < n(2) (b) n(1) > n(2)
Figure 4.9. Reflection and transmission of a uniform plane wave.
Further, see Eq. (4.133) or (4.152),
⎧ ⎫1
1 ⎨ n(1)
2⎬ 2
γ3t = jω ε(2) µ(2) 1 − (2) sin(θ )
i
2
, (4.174)
⎩ n ⎭
has to be imaginary valued. This means that
n(2)
0 ≤ sin(θi ) ≤ for n(1) > n(2) , (4.175)
n(1)
which yields a restriction of admissible values of the angle of incidence θi .

When we are dealing with a uniform transmitted wave, the expressions for
the reflection and transmission coefficients become
1 1
µ(1) µ(2)
cos(θi ) −
2 2
ε(1) ε(2)
cos(θt )
R = 1 1 , (4.176)
µ(1) 2 µ(2) 2
ε(1)
cos(θi ) + ε(2)
cos(θt )
1
µ(1) 2
2 ε(1)
cos(θi )
T = 1 1 , (4.177)
µ(1) 2 µ(2) 2
ε(1)
cos(θi ) + ε(2)
cos(θt )
for parallel polarization, and
1 1
ε(1) ε(2)
µ(1)
2
cos(θi ) − µ(2)
2
cos(θt )
R⊥ = 1 1 , (4.178)
ε(1) 2 ε(2) 2
µ(1)
cos(θi ) + µ(2)
cos(θt )
1
ε(1) 2
2 µ(1)
cos(θi )
T⊥ = 1 1 , (4.179)
ε(1) 2 ε(2) 2
µ(1)
cos(θi ) + µ(2)
cos(θt )
for perpendicular polarization. These reflection and transmission coefficients

are known as the Fresnel reflection and transmission coefficients. For loss-
less, dielectric media, the reflection coefficients R and R⊥ as a function of
the angle of incidence θi are presented in Fig. 4.10.
1.0 ..
...
1.0 .....
.......
... . ......
.....
|R | ...
..... |R⊥ |
.
. .. .
. . ...
. .... ...
.
... . ... ....

.......... . ..
6 6
.
.. . .. ...
....... . ... ....
......
0.75 ........
....
0.75 .
.
.
.
.
. ..
... ...
.
.
...... . ...
. ..
..
...... . . ..
........ .
. ... ...
..... . ... ...
..... . . ..
....... .
. ... ..
.
ε(2) ....... . ε (2) . ..... ...
.
ε(1)
=8 ... .
ε(1) . . . .
=8
. .
..
.... ...
0.5 ..... 0.5 ..... ..
..
. . . . . .... . . . . . . .
. .
. . ..... . .
... ..
. . .
. . ..... ...
. . ..... . .... ...
... .. . .. .. ...
.
. ... .
.. .. . ...... ..... ...
. 4 ..
....
.
. . .. ...... ....
. ... .. . ...... ...... ...... .... ....
...... ...... ...... .... . ....
.. ......
......
......
.
.
4 ... . .
... ... . .
......
.
....
...... .. .. . .....
. . ......
0.25 ......
...... .
.
... .. .
... ..
.
0.25 .........
.......
...... 2
..... ... ... . ... .
......
.............................. ..
..... 2 . .
. .. . .
.
. .
. ..............................
...........
........
....... .... .... ... .
...... .. . . .
...... ... .... . ... .
......
..... ..... ... .
..
.... . . .. .
..... ...... ... .. . .
....... .... .
0 0
0 30 60 90 0 30 60 90
- θi - θi
1.0 .
.
...
.. .
.
.... 1.0 .
.
.
...
.
...
...
(1)
.
. ..
.
. (1) . .
.
.. ..
.
|R | ε ε
.
... ...
ε(2)
.
.
= 8 4 ....
....
...
....
2 |R⊥ | ε(2)
.
.
= 8 4 .
..
.
...
...
...
2
. . .
. . ..
.
... ... . .
... ...
6 .
. .
...
...
...
6 .
.
.
...
...
....
0.75 . ... .... 0.75 . .
.
. . .. . ... ...
.... ... . . ...
.
... . ... ..
. ... .... ..
. ..
.
.
. .
. . .. . . ...
. ... ... ... ...
. .. ... . . ...
. ..
. . ... ..
.
. .
.. ... . .
... ...
. ... . ..
0.5 .... .. 0.5 . ... ...
. .
.
. ..
.
.
. .
. ..
.
.
. .
. ... ... ... ...
. . . ... .
... ...
. .. .. ..
.
. .
.
..
.
. . ...... ...
.
. .. ... ..... ...
...... ...... . . .. ... ...... .....
...... . . .. ...
...... . ... ..
...
.. ... .... ..
.
....
0.25 .... .
. ....
.
...
.
..
...
..
0.25 ......
.....
.....
. . .. ... ..
.
. .
.. .. .
..... .......
..................... . . .. . ............
........ . .. ...
.............. . ...
.. ............. ..
. .. .
...
.. . ... ...... .
...
.. .... .... ...
. ..... ... ...
...
0 .
0
0 30 60 90 0 30 60 90
- θi - θi
Figure 4.10. The reflection coefficients as a function of the angle of incidence θ i .

The media are lossless (σ (1) = σ (2) = 0) and pure dielectric (µ(1) = µ(2) = µ0 ).
Brewster angle
It is possible that the Fresnel reflection coefficient vanishes for a partic-
ular value of the angle of incidence. The pertaining angle of incidence is
i
called the Brewster angle. In the case of parallel polarization, this angle θB
follows from
1 1
µ(1) 2
µ(2) 2
cos(θ ) − i
cos(θt ) = 0 (4.180)
ε(1) ε(2)
and Snell’s law of refraction, Eq. (4.173). The result is
⎛ ⎞12
µ(2) ε(2)
(1) − ε(1)
i
tan(θB )= ⎝µ ⎠ , for parallel polarization . (4.181)
(1) µ(2)
ε
ε(2)
− µ(1)
For dielectric media, where µ(1) = µ(2) = µ0 , this Brewster angle follows
from
1
i ε(2) 2
(4.182)
tan(θB ) = , for parallel polarization .
ε(1)
In this last case, we also have that θi + θt = 12 π, with θi = θB

i . In the case
i
of perpendicular polarization the Brewster angle θB follows from
1 1
ε(1) 2
ε(2) 2
cos(θ ) − i
cos(θt ) = 0 (4.183)
µ(1) µ(2)
and Snell’s law of refraction, Eq. (4.173). The result is
⎛ µ(2)
⎞12
ε(2)
(1) − µ(1)
i
tan(θB )= ⎝ε ⎠ , for perpendicular polarization . (4.184)
µ(1) ε(2)
µ(2)
− ε(1)
For dielectric media, where µ(1) = µ(2) = µ0 , this Brewster angle does not
exist; the right-hand side of Eq. (4.184) is not real valued.
Total reflection
If n(1) > n(2) and the angle of incidence is larger than the critical angle,

i n(2)
θ > θci = arcsin , (4.185)
n(1)
Eq. (4.174) has to be written as
⎧ 2 ⎫1
1 ⎨ n(1) ⎬2
γ3t = ω ε(2) µ(2) 2 sin(θ i
) − 1 . (4.186)
⎩ n(2) ⎭
In this case γ3t is real valued, while on account of Eq. (4.131), or (4.150), γ1t is
imaginary valued; hence, the transmitted wave is a non-uniform plane wave
(Fig. 4.12). The planes of equal phase are perpendicular to the interface
between the two different media, while the planes of equal amplitude are
parallel to the interface. Since we are dealing with lossless media, the planes
of equal amplitude are perpendicular to the planes of equal phase. Moreover,
since σ (1) = σ (2) = 0, γ3i is imaginary valued and γ3t is real valued, we
observe from Eq. (4.140) that |R | = 1, while from Eq. (4.159) we observe
that |R⊥ | = 1. The reflected wave has an amplitude identical to the one of
the incident wave; only the phase is different. This phenomenon is called
total reflection (see Fig. 4.11). There exists an electromagnetic field in the
half-space 0 < x3 < ∞, but there is no time average power transported in
the i3 -direction. This is easily observed from (cf. Eq. (4.25))
t t∗ t∗
Re[(Ê × Ĥ ) · i3 ] = Re[(êt × ĥ ) · i3 ] exp(−2γ3t x3 ) (4.187)
and (cf. Eqs. (4.37) and (4.45))

⎧
⎪
⎪ γ3t t t∗
⎪
⎪ Re ĥ ĥ = 0 , for parallelly polarized waves,
⎪
⎪ jωε 2 2
t∗
⎨
Re[(êt × ĥ )·i3 ] =
⎪
⎪
⎪
⎪ γ t
⎪
⎪ 3 t t∗
⎩ Re ê2 ê2 = 0 , for perpendicular polarized waves.
jωµ
(4.188)
ε(1) , µ(1) ε(2) , µ(2) ε(1) , µ(1) ε(2) , µ(2)
....
.......
...... ....... ....... .......
......
.
s........r. ....... ...... ....... ....... ....... .......... si
s.r.................... n(2) t
(1) s .................. .......
.....
...
.si
................. ....
..
.....
............
.....
... ....... .. ..... ..
..... ..
..
........... .......
...
. .... n . .
.
....... . ...
... ...... ..
... . ..... ..... .
.......... . . .
..... ... ... ..
.
. .
.
. .. ... ... ... ... ...
.. .....
....
.
.... . . .... ... ... ... ..
.. .....
..... ..... .... ... .... ..
..... .. ..... ... ... .. ...
... ..... .... ......... ... .. ... ... .... .
. ..
. ..... .. .....
..... .. ..... .. . ... ..... ...
. . ... ... .
... ............................................................................
. . .. ..
. ... . ... . ..
.. ..
. .. ..
.. .. ...............................................................................
.... .
...
..
..
.
.....
..
.. i3 ...
. ...
.. .
.... .
.
..
i3 .
.
..
... ..... . ..
.. ... ... . ..
..
.. . ... .. ...
..... . . .... . .
... ....
.. ... (2)
.
..... ...
. ...
..
.
(2)
.
.....
....
... n ....... ....
... ..
..... .....
.
... n ....... ....
... ..
.....
..... ...... (1) . .. .. (1) .
.....
. ......
. . n ..
...... ........ n ..
......
...... ...
.. . ....... ...
..
.. ...
..... .... ....... ....... .... ....... .......
1 ..
1
...
n(2) n(2)
(a) si1 = n(1)
(b) si1 > n(1)
Figure 4.11. Total reflection of a uniform plane wave, when n(1) > n(2) .
γ1t .....
............
γ1t
... ..... ...
...... ... ......
........ ... ........
....... ... ... ...
............ ...
. ..... ... ... ...
.....
..... .... ...
... ....
..... .. ... ..
.....
..... ... ... ...
..... .... ... ....
..... .. ... ..
..... ...
..... ... ... ...
..... .... ... ....
..... .. ... ..
..... ...
.....
..... ... ... ...
..... .... ... ....
..... .. ... ..............................................................................
..... ...
..... ... .....
.....
.....
.....
.....
...
...
...
...
γ3t
... ............................. ...... .............................
............... ..............
......... ..
θi = θci ..........
. ..
i3 θi > θci
..
...
...
.
i3
..... ...
..... ...
..... .
......
. ...
...... ...
..... ...
..... ...
..... ....
.....
.
.
..
....... ...
..... ...
..... ...
..... ...
......
. ...
.....
... ...
..... ...
...
....
.
..
...
...
...
Figure 4.12. Planes of equal phase ( ) and planes of equal amplitude ( )

for critical incidence (θ i = θci , γ1t is imaginary valued, γ3t = 0) and for
above critical incidence (θ i > θci , γ1t is imaginary valued, γ3t is real valued).
Exercise 4.1
A parallelly polarized plane wave in free space has an electric field strength
E = (3i1 − 4i3 ) sin[3π × 109 t − 2π(4x1 + 3x3 )] .
(a) Find the frequency and the angle of propagation of this wave.
(b) Find the frequency domain electric field strength.
( c ) Find the frequency domain magnetic field strength.
(d) Find the time domain magnetic field strength.
Exercise 4.2
Given the electric fields of two parallelly polarized uniform plane waves,
the total field of which is given by Eq. (4.58) with γ = jω(ε0 µ0 )1/2 and
|ê(1) | = |ê(2) |. Show that this total electric field corresponds to a wave
standing in the x3 -direction and propagating in the x1 -direction. Hint: use
the fact that s · ê = 0.
Exercise 4.3
A uniform perpendicularly polarized, steady state wave is incident from a
lossless halfspace upon the plane boundary surface x3 = 0 of a perfect con-
ductor. The incoming wave is uniform, and its electric field is expressed as
i
Ê = E0 exp[−(γ1i x1 + γ3i x3 )]i2 , for x3 < 0. Draw the amplitude of the total
i r
electric field, Ê = Ê + Ê , as a function of x3 .
Exercise 4.4
A uniform plane wave in free space, having an electric field strength
√ √
E i = ( 12 i1 − 12 3i3 ) E0 cos[6π × 109 t − 10π( 3x1 + x3 )] ,
is incident on the interface x3 = 0 between free space and a lossless dielectric
medium, σ (2) = 0, ε(2) = 1.5ε0 and µ = µ0 .
(a) Does this electric field correspond to a parallelly or a perpendicularly
polarized wave, why?
(b) Find the angle of incidence θi , the angle of transmission θt , and the
vectors si and st along the directions of propagation of the incident
and the transmitted wave, respectively.
( c ) Find the expressions for the reflected and transmitted electric field
strength.
(d) Draw the electric field vector and the direction of propagation of the
incident, reflected and transmitted electric field in the configuration.
Exercise 4.5
A perpendicularly polarized, plane electromagnetic wave is incident from
medium (1) with ε(1) = 4ε0 , σ (1) = 0 and µ(1) = µ0 , upon the plane boundary
of medium (2) with ε(2) = ε0 , σ (2) = 0 and µ(2) = µ0 , at an angle of incidence
θi = π3 with the i3 -axis, i3 being the normal of the plane interface.
(a) What is the critical angle?
1
(b) Find γ1i and γ3i in terms of γ0 = jω(ε0 µ0 ) 2 .
( c ) Find γ3t in terms of γ0 .
(d) Find in medium (2) the distance d at which the electric field strength
decays a factor of exp(−1).
( e ) Find the reflection coefficient R⊥ and write this coefficient as R⊥ =

|R⊥ | exp(j arg[R⊥ ]).
Exercise 4.6
If you are under water looking towards the water surface, only a small circle
in the water surface is light while the rest of the surface is dark. Explain
why there is only a cone of light connected with the circle. Calculate the
angle θ of the cone (see Fig. 4.13). For optical frequencies the constitutive
parameters of water are ε = 1.77ε0 , σ = 0 and µ = µ0 .
air
..... ...
..... .....
..... .....
.....
..... ..
......
. water
..... ...
.....
.....
θ
..... .................
..... .....
..........
.
Figure 4.13. The cone of light.

Exercise 4.7
Calculate the critical angle θci and the Brewster angle θB i of an air/glass
interface. At optical frequencies, glass is a lossless dielectric with ε = 2.25ε0 .

Exercise 4.8
Show that for perfectly dielectric media and parallelly polarized waves the
expression of the Brewster angle is indeed given by tan(θBi ) = ( ε(2) ) 12 (see
ε(1)
Eq. (4.182)).
Problem 4.1
A uniform plane wave is incident on a magnetically impenetrable wall at an
angle of incidence of θi = π6 . Give for both the parallelly and perpendicularly
polarized waves
(a) the boundary conditions at the interface
(b) the expressions for the total electric and magnetic field strengths.
Problem 4.2
A parallel polarized plane wave is incident from medium (1) on the plane
boundary between medium (1) and medium (2). Both media are perfect
dielectrics. You know that for an angle of incidence larger than the critical
angle no time averaged power is transmitted into medium (2); also the re-
flected power in medium (1) is zero for an angle of incidence equal to the
Brewster angle. Now imagine a situation where the Brewster angle is larger
than the critical angle. Then the incident wave will not be reflected, because
of the Brewster angle of incidence, and will not be transmitted, because of
post-critical incidence. Solve this paradox.
Problem 4.3
A sinusoidally varying uniform plane wave is incident on a plane interface
between two lossless dielectric media at an angle larger than the critical
angle, θi > θci .
(a) Find the Fresnel reflection and transmission coefficients.
(b) Find the phase shift of the reflected wave at the interface.
( c ) Sketch the amplitude of the transmitted wave as a function of the

distance to the interface.
Problem 4.4
Indicate in the four graphs of Fig. 4.10 the Brewster angles and the critical
angles. Give the expressions for the Brewster angles, both for parallel and
perpendicular polarization, in case ε(1) = ε(2) while µ(1) = µ(2) .
Problem 4.5
A parallelly polarized plane uniform wave in free space (n(1) = 1) is incident √
at an angle θi on a slab of lossless dielectric with index of refraction n(2) = 3
and thickness d = 3 m, while on the other side there is another lossless
dielectric medium that extends to infinity with index of refraction n(3) = 3
(see Fig. 4.14). Determine the angle of incidence in case the wave is totally
transmitted through both interfaces.
n(1) n(2) n(3)
d -
Figure 4.14. A three-media configuration.
Problem 4.6
A lossless dielectric slab, with thickness d = λ4 m and ε = 2.25ε0 , is on one
side coated with an electrically perfectly conducting material. A uniform
plane wave in free space (n(1) = 1) impinges on the dielectric slab at normal
incidence (see Fig. 4.15).
n(1) n(2)
d -
Figure 4.15. The slab coated with a perfect conductor.

Problem 4.6 (continued)

(a) Give the general expressions for the plane waves in the two regions.
(b) Give the boundary conditions at the two interfaces.
( c ) Find the electric and magnetic field strengths in the two regions by sub-
stituting the general expressions of the field strengths into the bound-
ary conditions at both interfaces.
Problem 4.7
A parallelly polarized plane uniform wave in a lossless dielectric medium (1)
with ε(1) = 12ε0 is incident at an angle θi = π6 on a lossless dielectric slab
with index of refraction n(2) = 2 and thickness d = 1 m, while on the other
side there is another lossless dielectric medium that extends to infinity with
index of refraction n(3) (see Fig. 4.14). Determine the maximum index of
refraction n(3) , for which the wave is totally reflected back into medium (1).
Chapter 5
Electromagnetic Rays in a
Two-dimensional Medium
In Chapter 3 we have seen that a planar source with a uniform current

distribution, when it is placed in a homogeneous medium, will only generate
plane electromagnetic waves. Further we have discussed in Chapter 4 that
at a plane boundary between two different homogeneous regions, a plane
incident electromagnetic wave causes a plane reflected and eventually a plane
transmitted electromagnetic wave. If we are dealing with either a nonplanar
source, a non-uniform current distribution, an inhomogeneous medium, or
a curved boundary – or combinations of these –, the electromagnetic waves
that occur are in general not plane.
In this chapter, we will first introduce the concepts of a wavefront, a ray
trajectory, and an electromagnetic ray by means of the s-domain analysis
of electromagnetic waves in a homogeneous medium. Subsequently, we will
derive an approximate solution for the electromagnetic field that propagates
in a weakly inhomogeneous medium. The applied approximation is valid in
the limit s → ∞ and is known as the ray approximation. The resulting
approximate solution in a source-free domain is called the electromagnetic
ray. The wave propagation is very similar to the one of rays defined in
geometrical optics.
In this chapter we will restrict the analysis to the two-dimensional case,
i.e., we assume that the permittivity, the conductivity, and the permeability
122 electromagnetic rays in a two-dimensional medium
are independent of the x2 -coordinate (cf. Eq. (4.1)), and that the sources
of the electromagnetic wave field are also independent of x2 (cf. Eq. (4.2)).
Then, the generated electromagnetic field will be independent of x2 (cf.
Eq. (4.3)). From Chapter 4 we know that in this case the resulting total
electromagnetic field may be decomposed in two constituents, viz. the paral-
lelly and perpendicularly polarized fields. Therefore, we separately consider
the two cases of parallel and perpendicular polarization.
5.1. Homogeneous, lossless medium
For the introduction of several notions we first consider the case of a

uniform plane electromagnetic wave in a homogeneous, lossless medium.
According to Eqs. (3.38), (4.16), (4.17), and (4.18), the electric and the
magnetic field strengths in this case satisfy
s
Ê = ê(s) exp[− (s1 x1 + s3 x3 )] , (5.1)
c
s
Ĥ = ĥ(s) exp[− (s1 x1 + s3 x3 )] . (5.2)
c
If we introduce the so-called eikonal L as
c0
L(x1 , x3 ) = (s1 x1 + s3 x3 ) , (5.3)
c
these equations may be written in the form
s
Ê = ê(s) exp[− L(x1 , x3 )] , (5.4)
c0
s
Ĥ = ĥ(s) exp[− L(x1 , x3 )] . (5.5)
c0
In the time domain these functions correspond to
L(x1 , x3 ) L(x1 , x3 )
E = e[t − ] H[t − ], (5.6)
c0 c0
L(x1 , x3 ) L(x1 , x3 )
H = h[t − ] H[t − ]. (5.7)
c0 c0
homogeneous, lossless medium 123
For a given time instant t, the electric and magnetic field strengths in all
points of the planes given by
L(x1 , x3 ) (s1 x1 + s3 x3 )
= = constant (5.8)
c0 c
have an equal value. The plane
L(x1 , x3 ) (s1 x1 + s3 x3 )
= =t (5.9)
c0 c
is in a certain respect special: here the values of the electric and the magnetic
field strengths are given by E = e(0) and H = h(0), respectively. This
implies that at this plane the disturbance of the source has just arrived. For
this reason the plane given in Eq. (5.9) is called the wavefront. In Fig. 5.1
the wavefront is depicted for some time instants. Obviously, plane waves
have plane wavefronts. These move away from the source.
L(x1 , x3 ) L(x1 , x3 ) L(x1 , x3 )

= t(1) = t(2) = t(3)
c0 c0 c0
(1) (2) (3)

s- s- s-
Ray
trajectory
Wavefronts
Figure 5.1. A sequence of wavefronts and a ray trajectory for a uniform

plane electromagnetic wave in a homogeneous medium
(t(1) < t(2) < t(3) ).
If we take a certain point on the wavefront at t = 0 and keep track of this

point during the propagation of the wavefront through space, we obtain a line
that is known as a ray trajectory. By definition, the tangent in every point of
a ray trajectory coincides with the vector s in the propagation direction. In
Fig. 5.1 an example of a ray trajectory has been drawn. The ray trajectories
belonging to plane waves are straight lines, and (provided that the medium
is isotropic) these lines are perpendicular to the wavefronts. The solutions
of the form of Eqs. (5.4) and (5.5) are called electromagnetic rays.
Knowing Eqs. (5.4) and (5.5), there exists an alternative way to find the
value of the electric and the magnetic field strenghts at the wavefront. This
method is based on Tauber’s theorem for the Laplace transformation, which
states that
lim sfˆ(s) = f (0) , (5.10)

s→∞
and, more general
lim sfˆ(s) exp(st1 ) = f (t1 ) , (5.11)

s→∞
provided that the waves arrive at t = 0 and t = t1 , respectively, and the

functions on the right-hand side exist. In view of this theorem, we may
write
s
lim sÊ exp[ L(x1 , x3 )] = lim sê(s) = e(0) , (5.12)
s→∞ c0 s→∞
s
lim sĤ exp[ L(x1 , x3 )] = lim sĥ(s) = h(0) . (5.13)
s→∞ c0 s→∞
Note that the results are equal to the results obtained above. However, from
the latter two equations it may be seen that the time domain wavefield at
the wavefront follows from the behavior of its complex frequency domain
counterpart for s → ∞.
5.2. Parallel polarization
In this section we consider electromagnetic waves in an inhomogeneous

medium with ε = ε(x1 , x3 ), σ = σ(x1 , x3 ), and µ = µ(x1 , x3 ). Assuming
parallel polarization 125
that the sources of the electromagnetic field are also independent of x2 ,

the parallelly polarized field forms an independent wavefield constituent, for
which Ê1 , Ê3 , and Ĥ2 , are the non-zero electromagnetic field components.
In a source-free domain they satisfy the field equations (cf. Eqs. (4.4) - (4.6))
∂3 Ĥ2 + (σ+sε)Ê1 = 0 , (5.14)

−∂1 Ĥ2 + (σ+sε)Ê3 = 0 , (5.15)
∂3 Ê1 − ∂1 Ê3 + sµĤ2 = 0 . (5.16)
In analogy with Eqs. (5.4) and (5.5), we now investigate a solution of the
form
s
Ê1 = ê1 exp[− L(x1 , x3 )] , (5.17)
c0
s
Ê3 = ê3 exp[− L(x1 , x3 )] , (5.18)
c0
s
Ĥ2 = ĥ2 exp[− L(x1 , x3 )] . (5.19)
c0
In the inhomogeneous situation, the eikonal L = L(x1 , x3 ) is a more intricate

function of position than in the homogeneous case. Moreover, the amplitudes
ê1 = ê1 (x1 , x3 , s), ê3 = ê3 (x1 , x3 , s), and ĥ2 = ĥ2 (x1 , x3 , s), will now also
depend on the position. In order to investigate the spatial behavior of ê1 ,
ê3 , ĥ2 , and L, we substitute Eqs. (5.17) - (5.19) into Eqs. (5.14) - (5.16) and
divide by the non-zero factor exp(− cs0 L). We arrive at
s
∂3 ĥ2 − (∂3 L)ĥ2 + (σ+sε)ê1 = 0 , (5.20)
c0
s
−∂1 ĥ2 + (∂1 L)ĥ2 + (σ+sε)ê3 = 0 , (5.21)
c0
s s
∂3 ê1 − (∂3 L)ê1 − ∂1 ê3 + (∂1 L)ê3 + sµĥ2 = 0 . (5.22)
c0 c0
For most inhomogeneous media it is impossible to solve these equations ex-

actly, and only approximate solutions may be obtained. Since the wavefront
is the most interesting part of a propagating field, it makes sense to derive
an approximation that focusses on the behaviour at the wavefront. It fol-
lows from Tauber’s theorem that the behavior at the wavefront is found by
letting s → ∞. In that case we may neglect in Eqs. (5.20) - (5.22) the terms
without the factor s and retain the terms with the factor s. In the resulting
equations we divide by s/c0 . Then we obtain
−(∂3 L)ĥ2 + c0 εê1 = 0 , (5.23)

(∂1 L)ĥ2 + c0 εê3 = 0 , (5.24)
−(∂3 L)ê1 + (∂1 L)ê3 + c0 µĥ2 = 0 . (5.25)
Away from the wavefront the approximation remains valid as long as ê1 , ê3 ,
ĥ2 , and L, are slowly varying functions in space. This requires that ε, σ, and
µ, are slowly varying functions of position. The resulting approximation is
called the ray approximation, and the corresponding approximate solutions
of the form of Eqs. (5.17) - (5.19) are called electromagnetic rays. In each
point of a domain where the ray approximation is valid, Eqs. (5.23) - (5.25)
are linear algebraic equations for ê1 , ê3 , and ĥ2 , in which ∂1 L and ∂3 L have to
be determined in such a way that non-zero solutions exist. In order to derive
the condition to be satisfied by the partial derivatives of L, we eliminate ê1
and ê3 . Substituting Eqs. (5.23) and (5.24) into (5.25) yields
[(∂1 L)2 + (∂3 L)2 − c20 εµ]ĥ2 = 0 . (5.26)
For a non-zero solution, ĥ2 = 0, L has to satisfy the non-linear partial

differential equation
(∂1 L)2 + (∂3 L)2 = c20 εµ , (5.27)
which is known as the eikonal equation.
Uniform electromagnetic rays

For the interpretation of the electromagnetic rays of the form of Eqs.
(5.17) - (5.19), we recall that we have assumed that ê1 , ê3 , ĥ2 , and L, are
slowly varying functions in space. Then, the electromagnetic field of the
form of Eqs. (5.17) - (5.19) can locally be considered as a plane wave, of
which the amplitude and propagation direction are slowly varying in space.
As in the case of uniform plane waves, for general uniform rays the equation
L(x1 , x3 )
=t (5.28)
c0
parallel polarization 127
defines the wavefront at time instant t, and connection of corresponding

points at consecutive wavefronts yields a ray trajectory. In inhomogeneous
media, the wavefronts and ray trajectories are in general curved (see Fig. 5.2).
As long as the medium is isotropic, the wavefronts and the ray trajectories
are perpendicular to each other. In each point the tangent to a ray trajec-
tory coincides with the unit vector s in the propagation direction. Since
∇ L = ∂1 Li1 + ∂3 Li3 is a vector orthogonal to the surface L = constant, the
unit vector s = s1 i1 + s3 i3 along the ray and in the direction of propagation
is directed along the vector ∇ L. Therefore, we write
∂1 L = ns1 or s1 = n−1 ∂1 L , (5.29)

−1
∂3 L = ns3 or s3 = n ∂3 L , (5.30)
L(x1 , x3 )
= t(3)
L(x1 , x3 ) (2) c 0
=t
c0
L(x1 , x3 )
= t(1)
c0
s(1) (2)
s-
1
s(3)
q
Ray
trajectory
Wavefronts
Figure 5.2. A sequence of wavefronts and a ray trajectory for a

uniform electromagnetic wave in an inhomogeneous medium
(t(1) < t(2) < t(3) ).
in which n = n(x1 , x3 ) is the index of refraction. Using Eqs. (5.29) - (5.30),

we rewrite Eqs. (5.23) - (5.24) as
−ns3 ĥ2 + c0 εê1 = 0 , (5.31)

ns1 ĥ2 + c0 εê3 = 0 , (5.32)
so that
s1 ê1 + s3 ê3 = s · ê = 0 , (5.33)
while from Eq. (5.27) it follows that
1
n = c0 (εµ) 2 . (5.34)
The index of refraction n is independent of the propagation direction s of

the particular ray. This is a consequence of the isotropic character of the
medium. Equations (5.31) - (5.33) have the same form as Eqs. (4.33) - (4.35).
This leads to the conclusion that the relations between ê1 , ê3 , ĥ2 , s and n
are identical to the ones pertaining to the plane wave. This fact is recognized
by stating that an electromagnetic ray behaves locally as a plane wave.
5.3. Perpendicular polarization
In this section we consider electromagnetic waves in an inhomogeneous

medium with ε = ε(x1 , x3 ), σ = σ(x1 , x3 ), and µ = µ(x1 , x3 ). Assuming
that the sources of the electromagnetic field are also independent of x2 , the
perpendicularly polarized field forms an independent wavefield constituent,
for which Ê2 , Ĥ1 and Ĥ3 are the non-zero electromagnetic field components.
In a source-free domain they satisfy the field equations (cf. Eqs. (4.7) - (4.9))
−∂3 Ê2 + sµĤ1 = 0 , (5.35)

∂1 Ê2 + sµĤ3 = 0 , (5.36)
−(∂3 Ĥ1 − ∂1 Ĥ3 ) + (σ+sε)Ê2 = 0 . (5.37)
In analogy with Eqs. (5.4) and (5.5), we now investigate a solution of the
form
perpendicular polarization 129
s
Ê2 = ê2 exp[− L(x1 , x3 )] , (5.38)
c0
s
Ĥ1 = ĥ1 exp[− L(x1 , x3 )] , (5.39)
c0
s
Ĥ3 = ĥ3 exp[− L(x1 , x3 )] . (5.40)
c0
In the inhomogeneous situation, the eikonal L = L(x1 , x3 ) is a more intricate

function of position than in the homogeneous case. Moreover, the amplitudes
ê2 = ê2 (x1 , x3 , s), ĥ1 = ĥ1 (x1 , x3 , s), and ĥ3 = ĥ3 (x1 , x3 , s), will now also
depend on the position. In order to investigate the spatial behavior of ê2 ,
ĥ1 , ĥ3 , and L, we substitute Eqs. (5.38) - (5.40) into Eqs. (5.35) - (5.37) and
divide by the non-zero factor exp(− cs0 L). We arrive at
s
−∂3 ê2 +(∂3 L)ê2 + sµĥ1 = 0 , (5.41)
c0
s
∂1 ê2 − (∂1 L)ê2 + sµĥ3 = 0 , (5.42)
c0
s s
−∂3 ĥ1 + (∂3 L)ĥ1 + ∂1 ĥ3 − (∂1 L)ĥ3 + (σ+sε)ê2 = 0 . (5.43)
c0 c0
For most inhomogeneous media it is impossible to solve these equations ex-
actly, and only approximate solutions may be obtained. Since the wavefront
is the most interesting part of a propagating field, it makes sense to derive
an approximation that focusses on the bahaviour at the wavefront. It fol-
lows from Tauber’s theorem that the behavior at the wavefront is found by
letting s → ∞. In that case we may neglect in Eqs. (5.41) - (5.43) the terms
without the factor s and retain the terms with the factor s. In the resulting
equations we divide by s/c0 . Then we obtain
(∂3 L)ê2 + c0 µĥ1 = 0 , (5.44)

−(∂1 L)ê2 + c0 µĥ3 = 0 , (5.45)
(∂3 L)ĥ1 − (∂1 L)ĥ3 + c0 εê2 = 0 . (5.46)
Away from the wavefront the approximation remains valid as long as ê2 , ĥ1 ,
ĥ3 , and L, are slowly varying functions in space. This requires that ε, σ, and
µ, are slowly varying functions of position. The resulting approximation is
called the ray approximation, and the corresponding approximate solutions
of the form of Eqs. (5.38) - (5.40) are called electromagnetic rays. In each
point of a domain where the ray approximation is valid, Eqs. (5.44) - (5.46)
are linear algebraic equations for ê2 , ĥ1 , and ĥ3 , in which ∂1 L and ∂3 L
have to be determined in such a way that non-zero solutions exist. In order
to derive the condition to be satisfied by the partial derivatives of L, we
eliminate ĥ1 and ĥ3 . Substituting Eqs. (5.44) and (5.45) into (5.46) yields
[(∂1 L)2 + (∂3 L)2 − c20 εµ]ê2 = 0 . (5.47)
For a non-zero solution, ê2 = 0, L has to satisfy the non-linear partial

differential equation
(∂1 L)2 + (∂3 L)2 = c20 εµ , (5.48)
which is known as the eikonal equation.
Uniform electromagnetic rays

For the interpretation of the electromagnetic rays of the form of Eqs.
(5.38) - (5.40), we recall that we have assumed that ê2 , ĥ1 , ĥ3 , and L, are
slowly varying functions in space. Then, the electromagnetic field of the
form of Eqs. (5.38) - (5.40) can locally be considered as a plane wave, of
which the amplitude and propagation direction are slowly varying in space.
As in the case of uniform plane waves, for general uniform rays the equation
L(x1 , x3 )
=t (5.49)
c0
defines the wavefront at time instant t, and connection of corresponding

points at consecutive wavefronts yields a ray trajectory. In inhomogeneous
media, the wavefronts and ray trajectories are in general curved (see Fig. 5.2).
As long as the medium is isotropic, the wavefronts and the ray trajectories
are perpendicular to each other. In each point the tangent to a ray trajec-
tory coincides with the unit vector s in the propagation direction. Since
∇ L = ∂1 Li1 + ∂3 Li3 is a vector orthogonal to the surface L = constant, the
unit vector s = s1 i1 + s3 i3 along the ray and in the direction of propagation
is directed along the vector ∇ L. Therefore, we write
ray trajectories 131
∂1 L = ns1 or s1 = n−1 ∂1 L , (5.50)

−1
∂3 L = ns3 or s3 = n ∂3 L , (5.51)
in which n = n(x1 , x3 ) is the index of refraction. Using Eqs. (5.50) - (5.51),

we rewrite Eqs. (5.44) - (5.45) as
ns3 ê2 + c0 µĥ1 = 0 , (5.52)

−ns1 ê2 + c0 µĥ3 = 0 , (5.53)
so that
s1 ĥ1 + s3 ĥ3 = s · ĥ = 0 , (5.54)
while from Eq. (5.48) it follows that
1
n = c0 (εµ) 2 . (5.55)
The index of refraction n is independent of the propagation direction s of

the particular ray. This is a consequence of the isotropic character of the
medium. Equations (5.52) - (5.54) have the same form as Eqs. (4.41) - (4.43).
This leads to the conclusion that the relations between ê2 , ĥ1 , ĥ3 , s and n
are identical to the ones pertaining to the plane wave. This fact is recognized
by stating that an electromagnetic ray behaves locally as a plane wave.
5.4. Ray trajectories
By definition, a ray trajectory is a curve for which its tangent coincides

with the unit vector s in the propagation direction of the uniform electro-
magnetic ray (see Fig. 5.3). The trajectory comes into being by following
the ray in its propagation direction. Let
x = x(l) = x1 (l)i1 + x3 (l)i3 (5.56)

the parametric representation of an electromagnetic ray, in which the para-

1
meter l is the arclength along the ray (dl = [(dx1 )2 + (dx3 )2 ] 2 > 0). Then,
the unit vector τ along the tangent of the curve of Eq. (5.56) is given by
τ = ∂l x = ∂l x1 i1 + ∂l x3 i3 . (5.57)
Since Eq. (5.56) represents the ray trajectory, we simply have s = τ , hence,
s1 = ∂l x1 , s3 = ∂l x3 . (5.58)
With either Eqs. (5.29) - (5.30) or Eqs. (5.50) - (5.51) we find
n ∂l x1 (l) = ∂1 L , (5.59)
n ∂l x3 (l) = ∂3 L . (5.60)
Differentiation of these equations with respect to l, while using the relation,
∂l L = (∂l x1 )∂1 L + (∂l x3 )∂3 L

= s1 ∂1 L + s3 ∂3 L
= n(s1 s1 + s3 s3 ) = n , (5.61)
τ =s
........
t ......................
................................................................
................. ..........
...
..
. .
............ .......... ........
.
........ .. .......
..
..
........
. ..
.
.
......
......
...
......
. ..
. .....
.....
....
.
.
....
.
..
.
.
.....
.
.
.
.
.
.
.
.
. .....
....
....
L = L(xQ )
... . ....
....
. .
. .... ....
.
.... . .... .....
.
...
.
.. ..
. .... ......
.
.. .
. ... ......
.
... .. ... ..........
Q t
..
.. .
. ... .....
..... ...
....
....
....
...
..
x ......
.......
...... .....
...
.
.....
. ..
.
. .
......... ...
...
.
.... .. ..
.... .. ...... ...
..
. .
. ...... ...
...
..
... .
.. ...
... .
L = L(xP ) ..
.....
.
.
.
.
.
.
.
.
..
...
...
...
...
...... . .
...... ....
. .
.
. ...
...... ... .. ...
...... . . ...
tP
...... .... ..
. ...
....... .
. ..
. .
.. .......... ..
.
. ...... .
.
.... ...... ..
.
..
. ...... .
.
.
.. ....
... ...
... ...
...
... ... t
...
...
..
.
.
O
..
....
.
.
Figure 5.3. The ray trajectory of a uniform electromagnetic ray.

which follows from either Eqs. (5.29) - (5.30) or Eqs. (5.50) - (5.51), yields
∂l [n∂l x1 (l)] = ∂1 n , (5.62)

∂l [n∂l x3 (l)] = ∂3 n . (5.63)
This is the differential equation of the electromagnetic ray trajectory. The

coefficients in this equation depend only on the spatially dependent index of
refraction. By specification of the starting position and the starting di-
rection, this differential equation of the second order prescribes the ray
trajectory completely. The ray trajectory can be determined analytically,
numerically or graphically.
Finally, we remark that in a homogeneous domain the index of refraction
is constant, and, hence,
∂l ∂l x1 (l) = 0 , (5.64)
∂l ∂l x3 (l) = 0 , (5.65)
with the solution
x1 (l) = x1 (0) + s1 (0)l , (5.66)
x3 (l) = x3 (0) + s3 (0)l , (5.67)
in which x0 = x1 (0)i1 + x3 (0)i3 is the starting position of the ray and
s0 = s1 (0)i1 + s3 (0)i3 is the starting direction of the ray. Equations (5.66)
- (5.67) represent a straight line. Hence, in a homogeneous domain, the ray
trajectories of uniform, electromagnetic rays are straight lines.
5.4.1. Ray trajectories in a horizontally layered medium
In a horizontally layered medium, the medium parameters vary only in

the vertical direction i3 , hence,
ε = ε(x3 ), µ = µ(x3 ), n = n(x3 ) . (5.68)
For a horizontally layered medium we have
∂1 n = 0 , (5.69)
and Eq. (5.62) becomes

∂l [n∂l x1 (l)] = 0 , (5.70)
or equivalently
n ∂l x1 (l) = constant (5.71)
along the ray trajectory. Since ∂l x1 = s1 , we may write
n s1 = constant (5.72)
along the ray trajectory. Let us now introduce θ as the angle between i3 and
s, and θ0 as the angle between i3 and s0 (see Fig. 5.4), then s1 = sin(θ),
and we obtain
n(x3 ) sin(θ) = C0 , (5.73)
i.3
....
.......
i.3 ....
...
....
........ .......
... ......
θ= π 1
2
u
..
...
.... s ..
...
...
...
...
...
...
...
...
...
.
. .
shor
...................................................................
i3 .......... θ . .
.............
... .................................
.
...
...
...
...
..............
. .............
..........
u
........
....... ...
.
.
........ .
...
............
.
... ...
......
.... s0 .....
...
.. .
........
θ0
... .................
........... ................
..
......
u ... ..........
.
.
...
.
......
......
......
...
......
......
......
.....
......
..
......
.
.....
.....
.....
x3;hor
..... x3
x3;0
i.3
....
.......
.....
...
..
...
r ..........................................
.... i1
O
Figure 5.4. The ray trajectory of a uniform electromagnetic ray

in a horizontally layered medium.
in which
C0 = n(x3;0 ) sin(θ0 ) , (5.74)
denotes the trajectory constant of the ray. Equation (5.73) is known as

Snell’s law for a horizontally layered medium. A second result follows from
Eq. (5.63) as
∂l [n cos(θ)] = ∂3 n , (5.75)
where we have used that ∂l x3 (l) = s3 = cos(θ).

In Figs. 5.5 – 5.7, a number of trajectories of uniform electromagnetic rays
have been plotted. These trajectories have been computed, by a numerical
solution of Eqs. (5.73) and (5.75).
For a chosen value of the starting position and the starting direction, the
trajectory can have a point where θ = 12 π; in such a point the trajectory has
a horizontal tangent. Let x3 = x3;hor the value of the vertical coordinate,
where the tangent is in the horizontal direction (see Fig. 5.4), then substi-
tution of θ = 12 π and x3 = x3;hor into Eq. (5.73), with use of Eq. (5.74),
yields
n(x3;hor ) = n(x3;0 ) sin(θ0 ) = C0 . (5.76)
Since 0 ≤ sin(θ0 ) ≤ 1, Eq. (5.76) is only valid if n(x3;hor ) ≤ n(x3;0 ). After

passing the point with horizontal tangent, the ray bends in the direction of
increasing n. According to Eq. (5.73) the angle θ can keep the value of 12 π
and the vertical coordinate x3 can keep the value x3;hor , but Eq. (5.75) shows
that it is only possible if the level x3 = x3;hor coincides with the vertical level
where ∂3 n = 0. Such a level can only be reached asymptotically, unless the
trajectory starts in a horizontal direction at this level. In Fig. 5.7, this last
situation is illustrated for a Gaussian profile of the refraction index.
The theory of electromagnetic rays in a horizontally layered medium
finds its application in, e.g., the analysis of propagation of radio waves in the
atmosphere over such a small distance that the curvature of the atmosphere
can be neglected.
x..3 x..3
.. .
......... ..........
.... ....
.. .. ◦ ◦
4 ...
...
...
..
... 4 θ = 15 ... 0
...
...
θ = 30
...
.. ......
.0
.... θ = 45 .......
.......
.
0
◦
... . ......
... ... ... ... .... .......
... ... ... . .. ...... ............
.
... ..... ..........
3 ...
...
...
..
.. 3 ...
... ................. .......................... ...................... 0
θ = 60 ◦
...
...
...
..
..
..
..
...
...
...
s ...........................
..........
.........
.......
......
..... .
.........
.......
..........
... .. ... ..... ....

.. ..... ....
2 ... ..
. 2 ... ..... .....
... .. ... ..... .....
... ..
. ...
......
...... ..
........
.. ...
... . ... .......
.......... .......
... .. ... ...................................
... .. ...
1 ... 1
... ... ...
... .
.
. ...
... ... ...
... .. ...
....................................................................................
... n .........................................................................................................................................................................................................................................................................................................
... x1
0 1 2 0 2 4 6 8 10
x..3 x..3
... ...
........ ........
... ...
.... .... θ = 15 ◦
0 . ◦
4 ..
...
...
..
... 4 ..
...
...
θ = 30 ...
..
..
..... 0
.....
... ..
. ....
... ... ... . ...
.
.... ... .... ... ....
.....
.. ... .. ... .......
3 ...
...
..
.. 3 ...
...
... .....
.. ....... .................
..............................
........ .
...........
....... θ = 45 ◦
.. . .
... .. ... ... .... ...... ......
...... ...... 0
...
..
... ................................................................ ..... ...............................................
.. ..........
s
. ........
... .. ... ..................... ....... ....
. .
...... ..
.. ........
.
2 ...
..
2 ... ... ....... ..... ....... .....
.......
.......
... ..
.. ... ......... ......... ..............
. .
........ .
.........................................
θ = 60 ◦
.
... .. ... ...... .....
.
.. ....... ...... 0
... . ... ......... .......
... ..
.. ... .......................................
... ... ...
1 ... ... 1 ...
... .
.
. ...
... ... ...
... .. ...
..................................................................................... .......................................................................................................................................................................................................................................................................................................... x1
.. n ..
0 1 2 0 2 4 6 8 10
x..3 x..3
.... ....
........ ........
.... .... ◦ ◦
..
...
..
...
θ = 15 0 θ = 30 0
4 ...
..
... 4 ...
..
.
...
....
..
.... θ = 45 .
.......
....... 0
◦
... ... ... . .
. ...
...
... ... ... ... .... .......
... ... ... ..
. .
....
. ...
........
.
... ... ... ... .....
. .....
.......
3 ...
...
..
..
..
3 ...
...
... ..... ......... ................................
... ... ..... ....... ......... θ = 60 ◦
0
.. ... ........ ............ ....... ....
... ..
.. ... . ...... .......
... .. ... .. ... ..... ...... ..... ......
... ..
.. ... .... ... ..... ........ .....
. ..
.......
.. .. ... .... ..... ....
2 ...
... ..
. 2 ...
... .. .... ....
...................
.....
..... .....
.....
...
... ...
..
.
..
s
...
...
...........
......
.......
........
................ .................. .............
......
... .. ... ......
... .. ...
1 ...
...
... 1 ...
... .
.
. ...
... ... ...
... .. ...
....................................................................................... x1
n ...........................................................................................................................................................................................................................................................................................................
0 1 2 0 2 4 6 8 10
Figure 5.5. Ray trajectories for a piecewise linear profile of the refraction index;
n = 1 + 12 (x3 − 1) when 1 ≤ x3 < 2, n = 32 − 12 (x3 − 2) when 2 ≤ x3 ≤ 3.
x..3 x..3
.. .
......... ..........
.... ....
.. .. θ = 15 ◦
... ... 0
4 ...
..
... 4 ... .
...
θ = 60
...
..... 0 ◦
... ... ... ... .....
... ... ... .. ....
... ... ... ..
. ...
......
. ..
... .. ... .. .....
3 ...
..
..
. 3 ... ..
.. ....
.....
...
...
... ..
...
. ...
...
...
s .. ......
........
... .. ...
..
...
2 ...
...
..
..
..
2 ...
...
... .. ...
... ..
.. ...
... .. ...
... .. ...
1 ...
...
.
... 1 ...
...
...
... ... ...
... .. ...
....................................................................................
... n .........................................................................................................................................................................................................................................................................................................
... x1
0 1 2 0 2 4 6 8 10
x..3 x..3
... ...
........ ........
... ...
.... .... θ = 15 ◦
0
4 ..
...
...
..
... 4 ..
...
...
.
..
θ = 60
..
..
... 0 ◦
... . ...
... ... ... .. ...
.... ... .... .. ...
... ...
.. .. .. .. ...
3 ...
... ..
..
. 3 ...
...
..
.. ...
...
....
. ...
... .
.
.. ... ... .........
... .. ... .. ...
2
...
... ...
..
.. 2
...
... s .. .......
......
... ..
.. ...
... .. ...
... ..
.. ...
... .. ...
... .. ...
1 ... .
... 1 ...
... ... ...
... ... ...
... .. ...
.................................................................................... .......................................................................................................................................................................................................................................................................................................... x1
.. n ..
0 1 2 0 2 4 6 8 10
x..3 x..3
.... ....
........ ........
.... ....
.. .. ◦ ◦
4 ...
...
..
... 4 θ = 15 ... 0
...
θ = 30 .
...
0.
θ = 45..
.. ...
...
0
◦
... ... ... ... . .
..
... ... ... .
.. ... ...
...
... ... ... .. ..
. ..
... .. ... .. .. ..
...
3 .. 3 .. ....
... . ... .. ....
... .. ... .. ... ....
... ..
.
... ... ...
. ...
.....
.
. ..
... .. ... ... .... .......
... .. ... ... .... .........
...
2 ...
...
.. 2 ...
...
... ... ......
.............................................................
s
.. . . .......
..
... .. ... ............. ......
... ..
.. ... ......
.....
... .. ... .....
... .. ...
1 ... .
... 1 ...
.....
.....
.....
... ... ... .....
... ... ... .....
..... ◦
... .. ... θ = 60 .
...................................................................................... ....................................................................................................................................................0 x1
n ...........................................................................................................................................................
0 1 2 0 2 4 6 8 10
n = 32 − 12 (x3 − 1) when 1 ≤ x3 < 2, n = 1 + 12 (x3 − 2) when 2 ≤ x3 ≤ 3.
x..3 x..3
.. .
......... ..........
.... ....
.. ..
θ = 49 ◦
... ... .. 0
4 ...
... ..
..
.. 4 ...
... .
....
....
.
... .. ... ....
... .....
... .. ... .....
... ..
.. ... .....
3 ... .. 3 ... ........
.
... .. ... .....
... ..
... ......
.. .......
... .. ... ........
...
...
..
.. ... ..... ........................
......................................................................................................................................................................... 0
θ = 49.45 ◦
2 ...
... ..
..
2 ...
... .............................................................................................
.................... ...........
... .. ... ............. ........
... ..
.. ... .
....
............ .......
..... ......
s
... .. ...
.. ..... .....
... ... ..... .....
1 ...
...
..
..
..
..
1 ...
...
.....
.....
.....
... .. ... .....
...
..
.. ... θ = 50 .....
.. ◦
................................................................................... .........................................................................................................................................................................................................................0
... n ........................................................................................ x1
0 1 2 0 2 4 6 8 10
Figure 5.7. Ray trajectories for a Gaussian profile of the refraction index;
n = 32 − 12 exp(−|x3 − 2|2 ).
5.4.2. Ray trajectories in a radially layered medium
In a radially layered medium, the medium parameters are only functions

of the radial direction
1
r = (x21 + x23 ) 2 (5.77)
in the (x1 , x3 )-plane, hence,
ε = ε(r), µ = µ(r), n = n(r) . (5.78)
For a radially layered medium we have

x1
∂1 n = ∂r n(r) , (5.79)
r
x3
∂3 n = ∂r n(r) , (5.80)
r
and Eqs. (5.62) - (5.63) become
x1
∂l [n∂l x1 (l)] = ∂r n(r) , (5.81)
r
x3
∂l [n∂l x3 (l)] = ∂r n(r) . (5.82)
r
Multiplying Eq. (5.81) with x3 and Eq. (5.82) with −x1 , and adding the
results, we arrive at
x3 ∂l (n∂l x1 ) − x1 ∂l (n∂l x3 ) = 0 , (5.83)
which can be rewritten as
∂l (x3 n∂l x1 − x1 n∂l x3 ) = 0 , (5.84)
or equivalently
n(x3 ∂l x1 − x1 ∂l x3 ) = constant (5.85)
along the ray trajectory. Since ∂l x1 = s1 and ∂l x3 = s3 , we may write
n(x3 s1 − x1 s3 ) = constant (5.86)
along the ray trajectory. Let us now introduce φray as the angle between i3
and s, and φ as the angle between i3 and x, then
s1 = sin(φray ) , s3 = cos(φray ) , (5.87)

x1 = r sin(φ) , x3 = r cos(φ) , (5.88)
and
x3 s1 − x1 s3 = r sin(φray − φ) = r sin(θ) , (5.89)
where θ as the angle between x and s. Let us further introduce θ0 as the
angle between x0 and s0 (see Fig. 5.8), we obtain
r n(r) sin(θ) = C0 , (5.90)
in which
C0 = r0 n(r0 ) sin(θ0 ) , (5.91)
denotes the trajectory constant of the ray. Equation (5.90) is known as
Snell’s law for a radially layered medium. A second result follows from a
multiplication of Eq. (5.81) with x1 and Eq. (5.82) with x3 , and adding the
results, we arrive at
x1 ∂l (n∂l x1 ) + x3 ∂l (n∂l x3 ) = r∂r n(r) . (5.92)

x
r.....
.....
..
...
....
........
.. ....
θ s
........
xang
.. ...
...............
t .................
....................................................................
........... ....... rang
x0 ......
.
.
.. ..
.......
. .
. .
..
..
..
.
.
.
.
........
.......
...... ......
.....
.........
.
...
. ..
....... ...
....... .. ..... .....
...... ..... ..............
r0 ..... s0 ......
......
... .......
. t θ= 1π ..
......... ..
....... .
... ...... ......... ....
2
... .... . ......
................ ..
...
...
... θ0
............................
.
...... .
..
.
..
.....
.....
.....
....
.... ....... .
... ...... . ..
........... sang
t ..
... ..... . ..
.. .......
...
.......
... .. r .......
. ...
..... .... .
. .
....
.
....
.
. .
... .
... ..
.
.
.... rang
.... .. ....
....
..
....
.
.... ...
.. r0...
..
..
..
....
..
.... .. . .....
... ...
... ... .....
.
... ..
.
... .
.
. .
..
.. ... . ....
... ..
.. ..
... ... ....
.. .
....
.
... ..... .. .
.. .....
........
..... t
O
Figure 5.8. The ray trajectory of a uniform electromagnetic ray

in a radially layered medium.
The left-hand side of Eq. (5.92) can be rewritten as
∂l (x1 n∂l x1 + x3 n∂l x3 ) − n[(∂l x1 )2 + (∂l x3 )2 ]

= ∂l (x1 ns1 + x3 ns3 ) − n(s21 + s23 )
= ∂l [r n cos(θ)] − n . (5.93)
The right-hand side of Eq. (5.92) can be rewritten as
r∂r n(r) = ∂r [rn(r)] − n(r) . (5.94)
Combining the results of the last three equations, we obtain
∂l [r n(r) cos(θ)] = ∂r [rn(r)] . (5.95)

For a chosen value of the starting position and the starting direction, the
trajectory can have a point where θ = 12 π; in such a point the trajectory
has a tangent in the angular direction. Let r = rang the value of the radial
coordinate, where the tangent is in the angular direction (see Fig. 5.8), then
substitution of θ = 12 π and r = rang into Eq. (5.90), with use of Eq. (5.91),
yields
rang n(rang ) = r0 n(r0 ) sin(θ0 ) = C0 . (5.96)
Since 0 ≤ sin(θ0 ) ≤ 1, Eq. (5.96) is only valid if rang n(rang ) ≤ r0 n(r0 ).
This means that in the starting position the trajectory has to start in the
direction of decreasing r n. After passing the point with horizontal tangent,
the ray bends in the direction of increasing r n. According to Eq. (5.90) the
angle θ can remain at the value of 12 π and r can remain at the value rang ,
but Eq. (5.95) shows that it is only possible if the value r = rang coincides
with the surface where ∂r (r n) = 0. Such a surface can only be reached
asymptotically, unless the trajectory starts in a horizontal direction at this
surface.
n0
1
r2 2
n(r) = n0 2 − 2
.....
........
a
.....
.................................................... .......
........... ........ ..............
......................................................................................................................................................................................................................................................................................
.
........ ............
.....
.........
......
q
.................
.......
..... ........
..... .....
.............................................................................................................................................................................................................................................................. .......
......
.....
.....
. ............... ......
...... ...........
.............
...... ....
...
... ...... ...
. ............. .
.
..................................................................................................................................................................................................................................................... ............
.....
...
...
...
.. .................. . .....
. ................... .......... ....... ...
... ............ ... ....... ....... ...
. . . . . .
. ........... .. . ..
.. .
................................................................................. ................................................................................................................................................................... ............ .. .
.. ......................... ......... ........... ...... .....
... .................... ......... ...... ... ..
................. ......... ...... .. ..
.. ....... .... .
............. ....... ..... .. ...
. .. .. ..
....................................................................................................................................................................................................................................................................................................
....................................................................................................
Or
.
.... .....................................
................................
... .....................................................................................................
.................................................................................................................................................................................................................................................................................................... ... .. ........ .................. .... ...
.
... . .... .
.
... .
. .... .......... .......... ..... ....... ....... ... ...
............. . ...... .... ... .
.. ........................ .......... ........... ...... ....
............................................................................................................................................................................................................................................................ .......... . ....
...
...................... ... ......... ..... ...
.
.. .......
... .............. ........ ....
.... ...
. . .
...
. ... ..
. .... ..
...
...
... . . ................ ........ ..... ...
................................................................................................................................................................................................................................. ...... .....
. ..
... ......... ..... ...
... .......... ...... ...
........... ...... ...
.... .............. ...... ....
.............................................................................................................................................................................................................................................................. ...... ....
..... ....... ....
...... ........ .....
...... ......... ......
... ............ .......
...................................................................................................................................................................................................................................................................................... ............
........... ..
...................................................
Figure 5.9. Luneberg’s lens of radius a.

The theory of electromagnetic rays in a radially layered medium finds

its application in, e.g., the analysis of propagation of radio waves in the
atmosphere over such a distance that the curvature of the atmosphere can
be assumed to be constant.
An interesting application is the lens of Luneberg. In our two-dimensional
case, the lens is a dielectric cylinder that is radially layered. The index of
refraction increases from n0 , the index of refraction of the embedding, to
√
2 n0 . In Fig. 5.9, some ray trajectories have been plotted. These trajecto-
ries have been computed by a numerical solution of Eqs. (5.90) and (5.95).
The lens of Luneberg focusses a beam of parallel rays in one point at the
surface of the lens. Conversely, this lens is used to make a beam of parallel
rays from a concentrated source located in a point at the surface of the lens.
The location of the source is important. This is illustrated in Fig. 5.10,
where the source is located at a distance 14 a from the surface of the lens. We
then observe that the beam of parallel outgoing rays is lost.
n0
1
r2 2
n(r) = n0 2 − 2
......
.......
a
.....
..................................................... .......
........... ........ ..............
.....
..................
........
..........................................................
............
.........
... q
..................
......
......... ........ .....
........... ....... .........
........ ..... ................................................. ....... ....
. .. . .
.. ............................................. .............. ...... ....
................ .... ........... ...... ...
............. ... ....................................................................
.......... ..... ...
.................. ......... ..... ..
................................................................................... .................
.............
........
........ ........
....
............ ........................................... .
.. ..... ............. ........... .........
................................................. .......................................................................
.....
. . ..
.................. . . . ......
... .
......................................................... ..... . . . . . . . ..
.. ......................... .............. ......... .........
.......... . .... .. ........ ... ..... ... .. ...
................ ..................... ...
................. .
.............. .............. ...................
................................................................................................................................ ............ ......... ..... ....
................................................. ...........................................................................................................
...
.. ............................................................................................................. ............................... ........... ................
..................... .......... ..
. .
...
.. . . . . . . . . . . . . ........
.. ......
...... .. . ........................ ......................................
Or s
... ........ ... ............... ...................
.................................................................................................................................................................................................................................................................................. ..........................................................................................................................................................
. . . . ........ . .......... ..........
.
.. .
.................................................................................
.... . . .
....
... .. . . . .. . . ........ .. ... .......................................................................... source
................................................. ......................................................................................................... .................... ..
........... ...... .... ....
. .
................................................................................................................................. ............................................................................................... ................ ..................................
....................... . .... ... ... .. .... .
..................... ........................................................................... .................. ........ ....... .....
................................ ....... ................................................... ..................................................................... ....... .......... ........... ..........
........................................ ... .. ...........
............. .................... ............ ....... .........
............. ............................ ...
............... ...... .
............. ....... .
..................... .......................................................................................................... ....... .......
........ ........
.............
..................... ..... ..... ........ ..... ...
...................... ................. ... .
.. ..
..... ....
.
........ .............................. ................................... .. ... .
........ .... .......... ...... ....
........... ....... .........
.......... ........ .....
................. .......... ......
...... ............................................................. .....
....... ..
......... .......
............ .........
................................................
Figure 5.10. Source in front of the Luneberg’s lens.

Exercise 5.1
Given the profile for the index of refraction of a piecewise linear medium,
n = 2 − 12 (x3 − 1) when 1 ≤ x3 ≤ 3. Sketch in Fig. 5.11, where the ‘bullets’
denote the starting positions, the ray trajectories for the starting angles
θ0 = 45◦ , 60◦ , 90◦ and 120◦ . Whenever relevant, calculate the value of x3;hor
where the ray has a horizontal tangent and the angle θexit with which a ray
enters one of the homogeneous halfspaces.
Exercise 5.2 √
Given the starting position x3;0 = 2 2 − 1 in a medium with the profile
indicated in Fig. 5.5, what is the range of starting angles for which the
ray trajectories describe undulating paths that are confined to the region
1 < x3 < 3?
Exercise 5.3
Given a medium with the profile indicated in Fig. 5.5. Does each starting
position within the region 1 < x3 < 3 possess an associated range of starting
angles such that the ray trajectories are confined to the range 1 < x3 < 3? If
so, give the range of starting angles for all starting positions, if not, explain
your answer.
Exercise 5.4
For a medium with a profile as given in Fig. 5.6, find the starting positions
within the region 1 < x3 < 3 for which there exist rays with a horizontal
tangent at some value of x3 = x3;hor .
Exercise 5.5
For a medium with the profile indicated in Fig. 5.6, find the starting angle
belonging to each starting position within the region 1 < x3;0 < 3 for which
there exists a ray that reaches a horizontal tangent asymptotically at x3 = 2.
Problem 5.1
Given the profile for the index of refraction of a piecewise radially layered
medium, n(r) = 3 − r when 1 ≤ r ≤ 2. Consider a ray that comes from the
region r < 1 and enters the region 1 ≤ r ≤ 2 with a starting angle θ0 . Give
the range of starting angles for which the ray remains trapped.
x..3 x..3
.. .
......... ..........
.... ....
.. ..
... ...
4 ...
...
..
... 4 ...
...
...
... ... ...
... ... ...
... .... ...
3 ... ... 3 ...
... ...
s
...
... ... ...
...
... ... ...
... ... ...
2 ...
...
...
...
...
2 ...
...
... ... ...
... ... ...
... ...
... ...
... ...
1 ...
...
.... 1 ...
...
...
... ... ...
... .. ...
.....................................................................................
... n .........................................................................................................................................................................................................................................................................................................
... x1
0 1 2 0 2 4 6 8 10
x..3 x..3
... ...
........ ........
... ...
.... ....
4 ..
...
...
..
... 4 ..
...
...
...
... ... ...
.... ... ....
.. .... ..
3 ...
...
...
...
3 ...
...
... ... ...
...
... ...
s
...
... ... ...
2 ...
...
...
...
...
2 ...
...
... ... ...
... ... ...
... ...
... ...
... ...
1 ... .... 1 ...
... ... ...
... ... ...
... .. ...
.................................................................................... .......................................................................................................................................................................................................................................................................................................... x1
.. n ..
0 1 2 0 2 4 6 8 10
x..3 x..3
.... ....
........ ........
.... ....
.. ..
... ...
4 ...
..
... 4 ...
... ... ...
... ... ...
... ... ...
... .... ...
3 ...
...
...
...
3 ...
...
... ... ...
...
... ... ...
... ... ...
2 ...
...
...
... 2 ...
...
...
...
...
...
...
s
...
...
... ...
... ...
1 ... 1 ...
... .... ...
... ... ...
... ... ...
... .. ...
......................................................................................
n ........................................................................................................................................................................................................................................................................................................... x1
0 1 2 0 2 4 6 8 10
n = 2 − 12 (x3 − 1) when 1 ≤ x3 ≤ 3.
Chapter 6
Transmission Lines
In Chapter 3 we have discussed the propagation of a transverse electro-

magnetic (TEM) wave along a parallel-plate waveguide and we have shown
that it can be seen as a transmission line. Transmission lines serve to guide
the propagation of energy from one point to another, viz., from the source
to the load. Some more common examples of transmission lines are a pair
of parallel wires and coaxial cables. Transmission lines consisting of more
than two conductors are called multiconductor transmission lines. In elec-
tronic systems they are present in the form of bundles of wires, while in
power transmission systems they are present in the form of high-voltage
three-phase transmission lines above the earth or cables in the ground.
The transmission lines, to be discussed in the present chapter, are uni-
form lines in the sense that the cross-section of the line does not change along
the line. We shall assume that the lines are in a homogeneous medium, i.e.,
the medium in the domain D surrounding the conductors is homogeneous.
We assume that we have a transmission line consisting of N inner conductors
with outer boundaries Cn , n = 1, 2, · · · , N , and one outer conductor with in-
ner boundary C0 , see Fig. 6.1. This is the most general representation of a
multiconductor transmission line. In a particular case the outer conductor
may not be present. We assume the conductors to be perfect or in other
words electrically impenetrable. In this chapter we will investigate the prop-
agation properties of transverse electromagnetic waves (TEM) waves along
this transmission line.
146 transmission lines
.....
.....
.....
.....
. ........
....
.....
.....
.....
... ......
.....
.....
.....
.....
.
.. .....
.....
.....
.....
.....
...
.....
....
.....
.....
.....
..
.......
..... ...........................................
..... ........... .....
............. ..... ................
........ .....
.....
C0
.......
.....
......
.
.
..
.... ..
......
......
......
...... i3 ..
.
....
.
. .... ...... ................. .....
... ...
i3 .... .....
C2
..
...
.
.... .
....
. . ..... .................. . ..
......
.
... .... ........ .... ..... ....
...
...
...
.....
............................ ...........
..... ....
...... .....
.....
..... τ .......
........
.... ........
.....
.....
.....
..... ..... ...... ....
...
...
....
... τ
.....
. .
.............
... .... .........
.... ...
.....
.... . .. .
...
...
.
....
.....
.
..
.. ... .. ...... ..... ... .....
... ... . . ..... ... .....
...
...
....
..... .. ........... ......... ...
..
......
.
......
......... ...........
... ..... ... ....
C1 i 3
... ..... ... ....
... .......... ..... ... .....
... ..... ...............
. . .....
... ............................
. . . .
... ....... ..
......
.
... .. ...... ..... .. .. ...
... ... ........ ..... .... .....
.. ..... .... .....
...
... .... τ .........
............. .........
..... ...
.. .....
.....
D
... ... . .. .
..
.
... ... . .
... . ...... . . .
... ... .....
...
..........
..
... .....
... ....
..... .....
.... ...... ...... ... .....
....
.... .......... .......... .. .........
.
..... ....... .. ...
..... .. ......
..... ........
.....
...... ...........
.
...... ..
..
...... ....
....... .....
........ .....
......... ......
........... .... ..........
..............................................
Figure 6.1. The multiconductor transmission line.
6.1. TEM-waves
We start our analysis of the transmission lines with the frequency-domain

field equations in the source-free domain D between the conductors, viz.
−(∂2 Ĥ3 − ∂3 Ĥ2 ) + (σ+sε)Ê1 = 0 , (6.1)

−(∂3 Ĥ1 − ∂1 Ĥ3 ) + (σ+sε)Ê2 = 0 , (6.2)
−(∂1 Ĥ2 − ∂2 Ĥ1 ) + (σ+sε)Ê3 = 0 , (6.3)
∂2 Ê3 − ∂3 Ê2 + sµĤ1 = 0 , (6.4)

∂3 Ê1 − ∂1 Ê3 + sµĤ2 = 0 , (6.5)
∂1 Ê2 − ∂2 Ê1 + sµĤ3 = 0 . (6.6)
tem-waves 147
At the boundaries Cn , n = 0, 1, 2, · · · , N , of the conductors these equations

have to be supplemented with the boundary conditions that the tangential
component of the electric field strength vanishes, viz.,
τ1 Ê1 + τ2 Ê2 = 0 on Cn . (6.7)
Here, τ = {τ1 , τ2 , 0} is the unit vector tangent along the boundaries Cn .
We now assume that a transverse electromagnetic (TEM) wave propa-
gates along the transmission line in the longitudinal i3 -direction. A TEM
field structure is one in which the electric and magnetic field vectors at each
point in space have no components in the direction of propagation. The elec-
tric and magnetic field vectors at each point in space lie in a plane transverse
to the direction of propagation. Hence, Ê3 = 0 and Ĥ3 = 0. In Section 3.6,
we have considered a special case of a TEM-wave, where the field vectors
are independent of the position in the transversal plane, the (x1 , x2 )-plane.
In a general TEM-wave, the field vectors are not necessarily independent
of position in this transversal plane. Therefore, we generalize the results of
Eqs. (3.128) - (3.127) as
Ê1 = e1 (x1 , x2 ) V̂ (x3 , s) , (6.8)

Ê2 = e2 (x1 , x2 ) V̂ (x3 , s) , (6.9)
Ê3 = 0 , (6.10)
ˆ 3 , s) ,
Ĥ1 = h1 (x1 , x2 ) I(x (6.11)
ˆ 3 , s) ,
Ĥ2 = h2 (x1 , x2 ) I(x (6.12)
Ĥ3 = 0 , (6.13)
in which e1 = e1 (x1 , x2 ), e2 = e2 (x1 , x2 ), h1 = h1 (x1 , x2 ) and h2 = h2 (x1 , x2 )

are real functions of the transversal coordinates x1 and x2 . Each has the
dimension m−1 . Hence, the function V̂ = V̂ (x3 , s) is expressed in Volt
and Iˆ = I(x
ˆ 3 , s) is expressed in Ampère. The transversal vector functions
e = {e1 , e2 , 0} and h = {h1 , h2 , 0} are normalized such that

(e × h) · i3 dA = (e1 h2 − e2 h1 ) dA = 1 . (6.14)
(x1 ,x2 )∈D (x1 ,x2 )∈D
The TEM-wave of the type represented by Eqs. (6.8) - (6.14) satisfies the
partial differential equations (6.1) - (6.6) and the boundary conditions of
Eq. (6.7). To investigate this, we substitute the representations of Eqs. (6.8)
- (6.13) into Eqs. (6.1) - (6.7). This yields
h2 ∂3 Iˆ + (σ+sε)e1 V̂ = 0, (6.15)
−h1 ∂3 Iˆ + (σ+sε)e2 V̂ = 0, (6.16)
ˆ
−(∂1 h2 − ∂2 h1 ) I = 0 , (6.17)
−e2 ∂3 V̂ + sµh1 Iˆ = 0 , (6.18)

e1 ∂3 V̂ + sµh2 Iˆ = 0 , (6.19)
(∂1 e2 − ∂2 e1 ) V̂ = 0. (6.20)
(τ1 e1 + τ2 e2 )V̂ = 0 on Cn . (6.21)

Subsequently, we apply the method of separation of variables. As Eqs. (6.15)
- (6.20) have to hold for all x1 , x2 ∈ D and for all x3 , it follows that we arrive
at a system of differential equations for V̂ = V̂ (x3 , s) and Iˆ = I(x ˆ 3 , s),
∂3 Iˆ + γY0 V̂ = 0 , (6.22)
∂3 V̂ + γZ0 Iˆ = 0 , (6.23)
with
1
Z0 = , (6.24)
Y0
while the transversal functions have to satisfy the equations
−γY0 h2 + (σ+sε)e1 = 0 , (6.25)

γY0 h1 + (σ+sε)e2 = 0 , (6.26)
∂ 1 h2 − ∂ 2 h1 = 0 , (6.27)
γZ0 e2 + sµh1 = 0 , (6.28)

−γZ0 e1 + sµh2 = 0 , (6.29)
∂ 1 e2 − ∂ 2 e1 = 0 , (6.30)
tem-waves 149
and the boundary conditions
τ1 e1 + τ2 e2 = 0 on Cn . (6.31)
Note that γY0 and γZ0 play the role of the constants of separation. In order
that e1 and h2 satisfy Eqs. (6.25) and (6.29), and that e2 and h1 satisfy
Eqs. (6.26) and (6.28), we require that
1
γ = [(σ + sε)sµ] 2 with Re(γ) ≥ 0 . (6.32)
Then, Eqs. (6.25) - (6.26) and (6.28) - (6.29) may be replaced by

1
σ + sε 2
h1 = − Z0 e2 , (6.33)
sµ

1
σ + sε 2
h2 = Z0 e1 . (6.34)
sµ
Substitution of these expressions for h1 and h2 into Eq. (6.27) yields
∂ 1 e1 + ∂ 2 e2 = 0 . (6.35)
This relation expressing that the vector field e is divergence-free has to

be satisfied together with Eqs. (6.30) and (6.31). Eq. (6.30) is satisfied
identically if a two-dimensional scalar potential function ϕ = ϕ(x1 , x2 ) is
introduced such that
e1 = −∂1 ϕ , e2 = −∂2 ϕ . (6.36)
On account of Eq. (6.35) the function ϕ has to satisfy the two-dimensional

Laplace equation
∂1 ∂1 ϕ + ∂2 ∂2 ϕ = 0 . (6.37)
...........................................................................
............. ..........
.......... ........
.............. .......
....... .......
.
. ......... ......
......
.... .... .....
.....
....... .....
... ....
D
.
. ....
...
. ...
.... ...
... ...
. ...
... ...
... ........... ..................... .................. ....
.......... ...
.... ..... ..... ...
C2 C1 C0
..... ..
..... ..
... ........
... .....
. ... ...... ... ...
..
.... ..... ...
.. .
.
.
. ...
.. ...
... ... .. .... .. .
... ... . . . ...
... ... .
. .
..... ...
. .. .
....
.
......
... ..... .. .
. .. .... .. .
.. ... . .....
.
........ ... . .. ..
........ ... . .. .. ....
... ................................. ....... ................................. ....... ........
...
...
...
τ . .
τ .
...
.. . τ
..
... ...
... ...
... ...
.....
..... ......
.
..... .....
..... .....
...... ......
...... ......
.......
........ ....
........
......... .....
........... .........
............... ...........
.................................................................
Figure 6.2. Multiply-connected cross-sectional domain D of a multiconductor

transmission line. The boundary of D is C0 + C1 + C2 .
The boundary conditions of Eq. (6.31) require ∂τ ϕ = 0 (see Eqs. (1.48) -

(1.49)), or
ϕ = constantn on Cn . (6.38)
The TEM-wave can only exist if the two-dimensional Laplace equation (6.37)
supplemented with the boundary conditions of Eq. (6.38) has a non-zero
solution. It can be proved that this is indeed the case if the cross-sectional
domain D is multiply-connected. The boundary curve of D then consists
of a number of non-intersecting curves C0 + C1 + · · · + CN , where N ≥ 1
(Fig. 6.2).
Once the solution of this two-dimensional potential problem has been
solved, we can calculate e1 and e2 using Eq. (6.36). Subsequently, the
expressions for e1 and e2 , and the expressions for h1 and h2 that follow
from Eqs. (6.33) - (6.34), are substituted in the normalization integral of
Eq. (6.14). This finally determines the so-called characteristic impedance
Z0 of the transmission line. Once the characteristic impedance Z0 and its
inverse, the characteristic admittance Y0 , have been determined, we can
study the propagation properties of the TEM-wave by investigating the
transmission-line equations (6.22) - (6.23).
parallel-plate waveguide 151
6.2. Parallel-plate waveguide
As a first example, we solve the potential problem in the two-dimensional

transversal plane of a parallel-plate waveguide (see Fig. 6.3). In view of
the invariance of the configuration and the boundary conditions in the x2 -
direction, we may assume that the potential function is independent of x2 .
Hence, ϕ is a function of x1 only and
∂2 ϕ = 0 . (6.39)
Then, the two-dimensional Laplace equation is replaced by a one-dimensional

Laplace equation
∂1 ∂1 ϕ = 0 , 0 < x1 < a , (6.40)
where a is the width of the parallel-plate waveguide in the x1 -direction. The
boundary conditions become
lim ϕ = 0 , (6.41)
x1 ↑a
lim ϕ = 1 . (6.42)
x1 ↓0
.............................................................................................. w ................................................................................................
x1 = a
E
.....
........
...
..
...
....
..
×f............................................ H
.
D
S
... i1
..........
...
r .... i2
............................. x1 = 0
O
Figure 6.3. Cross-section of a parallel-plate waveguide.

By convention the constants of Eq. (6.38) are simply chosen as 0 and 1,

respectively. In order to satisfy Eq. (6.40), ϕ must be a linear function
of x1 . With the boundary conditions of Eqs. (6.41) - (6.42) we obtain as
solution
x1
ϕ=1− . (6.43)
a
Hence, from Eq. (6.36) it follows that
1
e1 = , e2 = 0 , (6.44)
a
and from Eq. (6.33) - (6.34) we obtain

1
σ + sε 2 1
h1 = 0 , h2 = Z0 . (6.45)
sµ a
Substituting Eqs. (6.44) and (6.45) in the normalization integral of Eq. (6.14)
and integrating over the domain {(x1 , x2 ) ∈ D; 0 < x1 < a, 0 < x2 < w},
where w is a characteristic length in the x2 -direction (see Fig. 6.3), we obtain
the characteristic impedance as

1
sµ 2 a (6.46)
Z0 = ,
σ + sε w
which is identical to the one of Eq. (3.140).
6.3. Coaxial line
As a second example, we solve the potential problem in the two-dimen-

sional transversal plane of a coaxial line (see Fig. 6.4). Let the x3 -axis of the
Cartesian coordinate system be the axis of the coaxial line. In view of the
rotational symmetry of the configuration, we may assume that the potential
function is a function of the radial coordinate
1
r = (x21 + x22 ) 2 (6.47)
coaxial line 153
..............................................................
........... .........
......... ........
............. .......
.
...... ......
.......... ......
.....
..
..... .....
.....
....... ....
... ....
D
.
. ...
...
. ...
.
.. ....
....... ....... ....... . ..
.
.........
.
.
E ...
...
.... . . ...
. .. ...
...
...... .......
×f
... .. . . .. ...
... .... ...
.... ...
.
.......... .. . . . .
...
....... ...
...
...
...
.. ...... ........ ......... ...
...........
.. S .
..
........
. . ...
...
... ...
r i
..........
... . .... 1.....
.
.
.
..... . ....
.....
. . .
... .....
H ...
...
rO
.. . ... ... ..
.. .. ..... ... ..
.
......... .
.. .
. ...
... . . .. .. . ..
... ... ...
...
............................... ...
. .
. ..
.. .. ... ...... ... . . ..
a i
... . .... . .. . .
.
... ........... .. 2 ..
. ..
... . . . . .
.. .......... .... .
..... ... ...
... ...... ................... ...... ...
... ....... .........
... ........ ..................... ...
.. ..
... ...
.......... .. .
...
... ...... ...... .
....... . ..... ...
...
...
... ...............
....... b .......
....... ....... ...... ...
...
..
...
.......... .
....
... .
... ...
... ...
.... ...
....
..... ......
.
..... ....
..... .....
...... ......
...... ......
.......
........ . ...........
......... .......
............ .........
.........................................................
Figure 6.4. Cross-section of a coaxial line.
only. The general solution, ϕ(x1 , x2 ) = ϕ(r), of the two-dimensional Laplace

equation is obtained as
ϕ = A + B ln(r) , a<r <b, (6.48)
where a is the radius of the inner conductor and b is the radius of the outer
conductor of the coaxial line. Further, A and B are arbitrary constants.
To prove that ϕ satisfies the two-dimensional Laplace equation (6.37), we
compute
x1 x2
∂1 ϕ = B 2 , ∂2 ϕ = B 2 , (6.49)
r r
and
1 x2 1 x2
∂1 ∂1 ϕ = B( 2 − 2 41 ) , ∂2 ∂2 ϕ = B( 2 − 2 42 ) , (6.50)
r r r r
hence, the two-dimensional Laplace equation, ∂1 ∂1 ϕ+∂2 ∂2 ϕ = 0, is satisfied.
The constants A and B follow from the boundary conditions (cf. Eqs. (6.41)
- (6.42)
lim ϕ = 0 , (6.51)
r↑b
lim ϕ = 1 , (6.52)
r↓a
as
ln(1/b) 1
A= , B= . (6.53)
ln(a/b) ln(a/b)
Hence, the solution of the pertaining two-dimensional potential problem is
obtained as
ln(r/b)
ϕ= . (6.54)
ln(a/b)
Hence, from Eq. (6.36) it follows that
1 x1 1 x2
e1 = , e2 = , (6.55)
ln(b/a) r2 ln(b/a) r2
and from Eq. (6.33) - (6.34) we obtain

1
1
σ + sε 2 Z0 x2 σ + sε 2 Z0 x1
h1 = − , h2 = . (6.56)
sµ ln(b/a) r2 sµ ln(b/a) r2
Substituting Eqs. (6.55) and (6.56) in the normalization integral of Eq. (6.14)
1
and integrating over the domain {(x1 , x2 ) ∈ D; a < (x21 + x22 ) 2 < b}, (see
Fig. 6.3), we obtain the relation

1
σ + sε 2 Z0 1
(e1 h2 − e2 h1 ) dA = dA = 1 .
(x1 ,x2 )∈D sµ [ln(b/a)]2 (x1 ,x2 )∈D r2
(6.57)
Since
b
1 1
dA = 2π dr = 2π ln(b/a) , (6.58)
(x1 ,x2 )∈D r2 r=a r
we arrive at the characteristic impedance as

1
sµ 2 ln(b/a) (6.59)
Z0 = .
σ + sε 2π
Commercial coaxial lines are available with Z0 specified. Neglecting the

losses, for example a 50 Ω or a 75 Ω line.
propagation properties 155
6.4. Propagation properties
After the determination of the characteristic impedance of the transmis-

sion line, the propagation of the TEM-wave in the x3 -direction is governed
by Eqs. (6.22) - (6.23), viz.,
∂3 Iˆ + γY0 V̂ = 0 , (6.60)
∂3 V̂ + γZ0 Iˆ = 0 , (6.61)
with Y0 = Z0−1 and γ = [(σ+sε)sµ] 2 . Similar to the analysis in Section 3.6,

1
these equations can be represented by a circuit equivalent of infinitesimal

length dx3 (see Fig. 6.5).
Eliminating Iˆ from these two differential equations of the first order we
end up with the second order differential equation,
∂3 ∂3 V̂ − γ 2 V̂ = 0 , (6.62)
for the electric potential V̂ . An analysis similar to Sections 3.1 and 3.6 yields
the general solution
ˆ 3 , s)
I(x ˆ 3 +dx3 , s)
I(x
e ...................................... r ............................... e
... ..
...... ......
......... .........
....
γZ0 dx3 ....
... ...
... ...
... ...
. .
V̂ (x.3 , s) γY0 dx3 V̂ (x3 +dx3 , s)

... . ...
... ...
... ...
... ...
... ...
.......... ..........
........ ........
. .
e r e
Figure 6.5. The circuit equivalent of a transmission line of infinitesimal length dx3 .
V̂ (x3 , s) = v̂ + (s) exp(−γx3 ) + v̂ − (s) exp(γx3 ) , (6.63)

ˆ 3 , s) = î+ (s) exp(−γx3 ) + î− (s) exp(γx3 ) ,
I(x (6.64)
where the s-dependent constants v̂ + , v̂ − , î+ and î− are determined by the
electromagnetic source strength at the beginning of the transmission line
and the terminating load at the end of the transmission line. On account of
Eq. (6.60) these constants have to satisfy the constraining relations
î+ = Y0 v̂ + , î− = −Y0 v̂ − . (6.65)
The functions v̂ + (s) exp(−γx3 ) and î+ (s) exp(−γx3 ) represent a wave trav-
eling in the positive x3 -direction, while v̂ − (s) exp(γx3 ) and î− (s) exp(γx3 )
represent a wave traveling in the negative x3 -direction.
6.4.1. Two-conductor transmission line
To demonstrate the coupling of a transmission line to a source and a load,

we now consider a transmission line consisting of two conductors (N = 1).
The parallel-plate waveguide and the coaxial line are examples of a two-
conductor transmission line. We assume that the transmission line has a
length L. At x3 = 0 it is driven by a source and at x3 = L it is terminated
by a load impedance ZL (see Fig. 6.6). Let V̂ (x3 , s) be the voltage differ-
ence between the two conductors and I(xˆ 3 , s) the electric current through a
conductor. Substituting Eq. (6.65) in Eqs. (6.63) - (6.64), we obtain
V̂ (x3 , s) = v̂ + (s)[exp(−γx3 ) + Γ exp(γx3 )] , (6.66)

ˆ 3 , s) = Y0 v̂ + (s)[exp(−γx3 ) − Γ exp(γx3 )] ,
I(x (6.67)
where Γ = Γ(s) is the reflection coefficient of the reflected wave at x3 = 0,

i.e.,
v̂ − (s)
Γ(s) = + . (6.68)
v̂ (s)
ˆ s)
I(0, ˆ s)
I(L,
.... ...
...................................... ..... ... .. .......................................
.... ...
....... ........
......... .........
...
...
.. incident wave ...
.... ......................................................... ....
.. ..
... ...
... ...
... ...
transmission line
V̂ (0, s) V̂ (L, s) ZL
.. ...
...
Z0 , L .. ...
...
... ...
... ...
... .......................................................... ...
... ...
.. ..
..........
.......
reflected wave ..........
......
.
.... ....
..... .... ..
... ....
x3 = 0 x3 = L
Figure 6.6. The terminated transmission line.
At x3 = 0 we also define an input impedance of the transmission line. This

input impedance Zin = Zin (s) is obtained as
V̂ (0, s) 1+Γ
Zin = = Z0 . (6.69)
ˆ s)
I(0, 1−Γ
The reflection coefficient is determined by the terminating load at x3 = L,
where the boundary condition
ˆ s)
V̂ (L, s) = ZL I(L, (6.70)
applies. Substituting Eqs. (6.66) - (6.67) into this condition, and rearranging
some terms, we arrive at
(1 − ZL Y0 ) exp(−γL) + (1 + ZL Y0 ) exp(γL) Γ = 0 . (6.71)
Using Y0 = Z0−1 , the reflection coefficient is obtained as
ZL − Z0
Γ= exp(−2γL) , (6.72)
ZL + Z0
while, from Eq. (6.69), the input impedance is obtained as
ZL + Z0 tanh(γL)
Zin = Z0 . (6.73)
Z0 + ZL tanh(γL)
We now determine the reflection coefficient and the input impedance for
various load impedances.
• Short-circuit load (ZL = 0)

We find Γ = − exp(−2γL) and Zin = Z0 tanh(γL).
• Open-circuit load (ZL = ∞)

We find Γ = exp(−2γL) and Zin = Z0 / tanh(γL).
• Matched load (ZL = Z0 )

We find Γ = 0 and Zin = Z0 . There is no reflected wave and the input
impedance is equal to the characteristic impedance.
6.4.2. Lossless transmission line: steady-state analysis
We now restrict ourselves to the lossless case (σ = 0) and discuss the

results for one angular frequency (s = jω, ω ≥ 0). We then arrive at
ZL − Z0
Γ= exp(−2jkL) (6.74)
ZL + Z0
and
ZL + jZ0 tan(kL)
Zin = Z0 , (6.75)
Z0 + jZL tan(kL)
where
1 ω
k = ω(εµ) 2 = . (6.76)
c
In the lossless case the characteristic impedance becomes real. For the
1
parallel-plate waveguide we then have Z0 = ( µε ) 2 wa and for the coaxial line we
1
have Z0 = ( µε ) 2 ln(b/a)
2π . We subsequently determine the reflection coefficient
and the input impedance for various load impedances.
• Short-circuit load (ZL = 0)

We find Γ = − exp(−2jkL) and Zin = jZ0 tan(kL). The input impe-
dance is either inductive or capacitive, depending on kL. Notice that
in the limit of kL 1, which implies low frequency or a line that is
small compared to the wavelength, Zin ≈ jZ0 kL, and the terminated
transmission line behaves like an inductor.
• Open-circuit load (ZL = ∞)

We find Γ = exp(−2jkL) and Zin = −jZ0 / tan(kL). The input im-
pedance is either capacitive or inductive, depending on kL. Notice
that in the limit of kL 1, which implies low frequency or a line
that is small compared to the wavelength, Zin ≈ −jZ0 /(kL), and the
terminated transmission line behaves like a capacitor.
• Matched load (ZL = Z0 )

We find Γ = 0 and Zin = Z0 . There is no reflected wave and the input
impedance is equal to the characteristic impedance.
The time average of the power transmitted through the cross-section of

the transmission line is obtained as

∗

1
2 Re (Ê × Ĥ ) · i3 dA = 12 Re V̂ Iˆ∗ , (6.77)
(x1 ,x2 )∈D
where we have used Eqs. (6.8) - (6.13) and the normalization integral of
Eq. (6.14). Substitution of the expressions for V̂ = V̂ (x3 , jω) and Iˆ =
ˆ 3 , jω), i.e.,
I(x
V̂ (x3 , jω) = v̂ + (jω)[exp(−jkx3 ) + Γ exp(jkx3 )] , (6.78)

ˆ 3 , jω) = Y0 v̂ + (jω)[exp(−jkx3 ) − Γ exp(jkx3 )] ,
I(x (6.79)
into the right-hand side of Eq. (6.77), leads to

1
2 Re V̂ Iˆ∗ = 12 Y0 |v̂ + |2 − 12 Y0 |Γ|2 |v̂ + |2 = 12 Z0 |î+ |2 − 12 Z0 |Γ|2 |î+ |2 . (6.80)
It shows that in the lossless case the power flow in the transmission line is
equal to the difference of the power flow of the incident wave and the power
flow of the reflected wave.
6.4.3. Transients on lossless transmission lines
Up to here, we have assumed a single operating frequency. In some

applications, however, pulsed signals rather than single-frequency signals
will be used. Let the transmission line be terminated with a resistive load
ZL and be driven by an impulsive source at x3 = 0 (see Fig. 6.7). The
internal resistance of the generator equals ZS . Note that both ZL and ZS
are assumed to be real. To reconstruct the transient time behavior of the
transmitted waves, the s-domain equations (6.66) - (6.67) with Eq. (6.72),
have to be transformed back to the time domain. This transformation is
most easily carried out for a lossless medium, i.e., if σ = 0. Then, the
s-domain propagation coefficient can be written as
s
, with c = (εµ)− 2 > 0 ,
1
γ= (6.81)
c
while Z0 = Y0−1 is a real constant, independent of s. We then arrive at

x3 2L−x3
V̂ (x3 , s) = v̂ + (s) exp(−s ) + ΓL exp(−s ) , (6.82)
c c

ˆ 3 , s) = Y0 v̂ + (s) exp(−s x3 ) − ΓL exp(−s 2L−x3 ) ,
I(x (6.83)
c c
I(0, t) .. ...
I(L, t)
....................................... ......... .. .......................................
...... ...
... ....
.......... ..........
. .. . ..
... ...
.. ..
.... ...
ZS ..
...
...
...
... ...
.... ....
transmission line
V (0, t) V (L, t) ZL
. ...
... Z0 , L . ...
...
VS ...
...
...
...
...
...
...
...
...
...
.......... ..........
........ ........
.
... ..
...
..... .... .
... ...
x3 = 0 x3 = L
Figure 6.7. An impulsive source connected to the transmission line.

where
ZL − Z0
ΓL = (6.84)
ZL + Z0
is the reflection coefficient of the reflected wave of a transmission line of zero
length. Applying the shift rule of the Laplace transformation, the corre-
sponding time domain result is
x3 2L−x3
V (x3 , t) = v + (t − ) + ΓL v + (t − ), (6.85)
c c
x3 2L−x3
I(x3 , t) = Y0 v + (t − ) − ΓL v + (t − ) . (6.86)
c c
Equations (6.85) and (6.86) show that on the transmission line a wave prop-
agates in the positive x3 -direction with wave speed c. At x3 = L, this wave
is reflected with reflection coefficient ΓL and propagates with wave speed c.
Once v + (t) is known the present problem is completely determined. This
quantity follows from the transient time behavior of the source.
When Zin = Zin (s) is the input impedance of the transmission line and
ZS is the internal resistance of the source, then from Fig. 6.7 we observe
that
Zin (s)
V̂ (0, s) = V̂S (s) , (6.87)
Zin (s) + ZS
where V̂S (s) is the Laplace transform of the time-dependent voltage VS (t) of
the source. Equation (6.87) cannot be transformed back to the time domain
by simple inspection. Therefore, we use Eqs. (6.66), (6.69) and (6.87) to
obtain
V̂ (0, s) Z0
v̂ + (s) = = V̂S (s) . (6.88)
1+Γ Z0 + ZS + (Z0 − ZS )Γ
Now in the first factor of the right-hand side of Eq. (6.88) only Γ =
ΓL exp(−2sL/c) contains an s-dependence. We therefore expand the first
factor of the right-hand side of Eq. (6.88) in terms of powers of Γ. We write
1 Z0
v̂ + (s) = V̂S (s) , (6.89)
1 − ΓS ΓL exp(−2sL/c) Z0 + ZS
in which
ZS − Z0 ZL − Z0
ΓS = , ΓL = . (6.90)
ZS + Z0 ZL + Z0
Since ZS , ZL and Z0 are positive real constants (independent of s), we

observe that |ΓS | ≤ 1 and |ΓL | ≤ 1. Further, in view of causality we only
consider those values of s for which Re(s) > 0, hence | exp(−2sL/c)| < 1.
Then, it is allowed to expand v̂ + (s) as
∞
+ Z0 n 2nL
v̂ (s) = V̂S (s) (ΓS ΓL ) exp(−s ) . (6.91)
Z0 + ZS n=0
c
Applying the shift rule of the Laplace transformation, the corresponding

time domain result is
∞
Z0 2nL
+
v (t) = (ΓS ΓL ) VS (t −
n
) . (6.92)
Z0 + ZS n=0
c
Together with Eqs. (6.85) and (6.86) the transient time behavior of the
propagation of waves on the transmission line is described completely. In
order to interpret the different terms of Eq. (6.85) and (6.92) we rearrange
the different terms as
Z0 x3
V (x3 , t) = [ + VS (t − )
Z0 + ZS c
2L−x3
+ΓL VS (t − )
c
x3 +2L
+ΓS ΓL VS (t − )
c
4L−x3 (6.93)
+ΓL ΓS ΓL VS (t − )
c
x3 +4L
+ΓS ΓL ΓS ΓL VS (t − )
c
6L−x3
+ΓL ΓS ΓL ΓS ΓL VS (t − )
c
+ ··· ].
Suppose that at t = 0 the source generates a pulse VS (t). At the beginning,

the wave does not know whether or not the transmission line is matched
at the other end. The apparent input impedance at x3 = 0 is then equal
to Z0 , the characteristic impedance of the transmission line. Hence, the
voltage of the wave at the beginning of the transmission line is equal to
Z0 VS (t)/(Z0 + ZS ). This pulse travels along the transmission line with wave
speed c and at t = L/c it arrives at the load. A reflection occurs with a re-
flection coefficient ΓL . The reflected pulse travels with wave speed c toward
the source and at t = 2L/c it arrives at the source, where a second reflec-
tion takes place with a reflection coefficient ΓS . The reflected pulse travels
again to the load with a voltage ΓS ΓL Z0 VS (t)/(Z0 + ZS ). This process is
continued. In Fig. 6.8 we illustrate these phenomena by plotting the voltage
V (x3 , t) at the middle, x3 = 12 L, of the transmission line as a function of
time.
VS (t) Z0 = 50 Ω L = 0.6 m
..
.......
........
1 ...
....... ZS = 150 Ω ZL = 10 Ω
.. ... ...
... .... ....
..
...
...
.. ..
... ...
... ....
c = 3 × 108 m/s
.. .... ....
.. ..
... ....
... ...
.... ....
.. ..
... ....
... .... ......................................................................
0 .. ..
.. ...
... .
... 2
................................................ t in (ns)
Z0 +ZS 1
Z0 ..V ( 2 L, t)
......
.........
...
....
..
...
...
1 .
......
... ....
... ...
..... ....
. ..
... ....
... ...
.... ....
.. ...
... ...
... ...
.... .... .... ..
.....
.. .. .. .. .. ..
.. .. .. ... ...... .. .. .......
.................................................... .... ......................................................................... .... ......................................................................... .... ............................................................................. ........................................................................................ .........................................
0 .. ..
.. .. ... ..
... ...
.
.. ...
.. .
... ..
2 .. ... ..
... ..
. 4 .
. .
.
.....
.. 6 8 10
... ...
... ....
... .. ................................................................................... t (in ns)
... ...
.. ..
... ...
......
Figure 6.8. The source voltage VS (t) and the voltage V ( 21 L, t).
Exercise 6.1
(a) Calculate the inner radius of the outer conductor b of a lossless coaxial
transmission line with Z0 = 50 Ω, ε = 1.5ε0 and an inner conductor
with outer radius a = 4.11 mm.
(b) Find the time-averaged power flow in the coaxial transmission line in
case of a matched load.
( c ) If the coaxial transmission line has a prescribed maximum voltage of

100 V, then what is the maximum |E| in the cable?
Exercise 6.2
For the circuit shown in Fig. 6.7, Z0 = 50Ω, ZS = 30Ω, ZL = 20Ω, L = 2 m,
c = 1 × 108 m/s and VS = 1 V for 0 ≤ t ≤ 1 ns, otherwise VS = 0 V.
(a) Calculate v + (t).
(b) Give the expressions for V ( 12 L, t) and I( 12 L, t) and sketch them as a

function of t.
( c ) Calculate the total energy WS delivered by the source.
(d) Calculate the energy W + that is at the beginning transmission line.
( e ) Calculate the energy WL absorbed by the load.
Exercise 6.3
For a two-wire transmission line as depicted in Fig. 6.7, the following data
are available: VS (t) = V0 H(t) V, H(t) denotes the unit step function starting
at t = 0; Z0 = 100 Ω. The voltages at the beginning and at the end of the
line are measured for the first 5 µs, they are
! !
100 V, 0 < t < 4 µs, 0 V, 0 < t < 2 µs,
V (0, t) = V (L, t) =
90 V, 4 < t < 5 µs, 75 V, 2 < t < 5 µs.
Find the values of V0 , ZS , ZL and the travel time T = L/c.

Exercise 6.4
In a steady-state situation, for what value(s) of L will the transmission line
terminated by a short circuit load behave like a capacitor? Conversely, for
what value(s) of L will the open circuit load behave as an inductor?
Exercise 6.5
The Voltage Standing Wave Ratio (VSWR) is a measure for the effectiveness
of the matching load at the end of a transmission line. In the steady-state
situation it is found as VSWR = |V |max /|V |min = (1 + |Γ|)/(1 − |Γ|). A loss-
less, two-wire line is terminated with a television set A giving a VSWR=3.
When A is replaced by the set B, it is meaured that VSWR=1.5. Which
of the two sets is better matched to the transmission line? Explain your
answer. What is the amplitude of the reflection coefficient of the set B?
Problem 6.1
Given a parallel-plate transmission line, made up of perfect conductors, of
width a = 1 mm and characteristic length w = 10 cm, with a perfect di-
electric medium between the plates, µ = µ0 . The voltage along the line,
for a uniform plane wave propagating along the line, is given by V (x3 , t) =
10 cos(3π × 108 t − 3πx3 ). Find
(a) the electric field strength E1 (x3 , t) of the wave
(b) the magnetic field strength H2 (x3 , t) of the wave
( c ) the electric current I(x3 , t) along the line
(d) the power flow P out (x3 , t) along the line.
Problem 6.2
Consider the two-wire transmission line of Fig. 6.7 but now the source is
connected with the transmission line via a switch at x3 = 0. VS = 100 V,
ZS = 40 Ω, Z0 = 60 Ω and T = L/c = 1 µs. If the switch is closed at t = 0 s,
find the value of ZL for the following situations
(a) V ( 12 L, 1.6µs) = 75 V
(b) V ( 13 L, 2.5µs) = 44 V
( c ) V ( 34 L, t → ∞) = 80 V.
Problem 6.3
A lossless transmission line is short circuited at the far end. A variable
frequency voltage generator is connected as its input and the current drawn
is measured. It is found that the current reaches a maximum value for
f = 500 MHz while it reaches a minimum for f = 505 MHz. Determine
whether the current drawn would be a maximum, minimum or neither for
the following frequencies
(a) f = 450 Mhz
(b) f = 394 Mhz
( c ) f = 335 Mhz.
Chapter 7
Electromagnetic Waveguides
At microwave and optical frequencies of the electromagnetic spectrum,

the transmission of signals takes place through electromagnetic waves that
carry the signal from one point to another. The simplest kind of waves in
this category are the one-dimensional and two-dimensional plane waves. As
we have seen in Chapters 3 and 4, each of these electromagnetic plane waves
has a unique propagation coefficient, so we can say that each of them pro-
vides a single wave channel for transmission. A similar observation can be
made for a transverse electromagnetic (TEM) wave that propagates along a
transmission line (see Chapter 6). Apart from the latter case, the electro-
magnetic wave that constitutes the transmission channel propagates in an
unbounded, homogeneous isotropic medium along a straight line, while its
intensity (or amplitude distribution) is uniform on the plane transverse to
this line.
In this chapter we discuss the problem of transmitting electromagnetic
power from one point to another in such a way that the intensity of the wave
is limited to a finite cross-section and that the intensity of the wave is not
uniform on the plane transverse to the direction of propagation. Similar to
the transmission lines, one can say that waveguides are configurations that
are designed to carry energy or information, along a specific trajectory (not
necessarily straight), from one point in space to another. As the shortest
path joining two different points in space is the straight line, the simplest
type of waveguide is a cylindrical one, i.e., a structure that is uniform in a
fixed direction in space. When a waveguide is used in practice, for example
168 electromagnetic waveguides
as a transmission channel in a communication system, it will of course many

times be necessary to deviate from the assumed uniformity and guide the
wave along bends and corners. In this chapter we will restrict the discussion
to cylindrical waveguides.
The cylindrical waveguide has one direction in space in which it is uni-
form; this direction is called the axial or longitudinal direction. Any plane
perpendicular to the axial direction will be called a transverse plane. In each
transverse plane the configuration is the same. As regards their properties in
the transverse plane, the cylindrical waveguide can be classified into two cat-
egories: (a) closed waveguides and (b) open waveguides. A closed waveguide
contains a closed cylindrical wall across which no transfer of electromag-
netic energy can take place (Fig. 7.1). Energy or information is in this case
carried by an electromagnetic wavefield that is confined to the interior re-
gion. In view of the manufacturing process, there is a lower bound on its
cross-sectional dimensions and this puts an upper limit on the frequencies
of operation of the electromagnetic waves that can be sent along the wave-
guides of the closed type. In practice, waveguides of the closed type are
used up to frequencies in the gigahertz region and the walls are made of
highly conducting metal. Some examples of closed waveguides are shown in
Fig. 7.2.
axial
direction
exterior
domain
electromagnetically
impenetrable wall
interior
domain
. transverse
plane
Figure 7.1. Closed waveguide.

electromagnetic waveguides 169
metallic wall
metallic wall
(a) (b)
metallic outer
conductor
metallic inner
conductor
(c)
Figure 7.2. Closed waveguides: (a) waveguide with rectangular cross-section;

(b) waveguide with circular cross-section; (c) coaxial waveguide.
In an optical communication system, the frequencies of operation are

an order of magnitude higher and hence the use of closed waveguides is
out of the question. For this reason, open waveguides should be used as
waveguiding structures. Since the electromagnetic field in the cross-section
extends to infinity, it should be operated in such a mode that the electro-
magnetic field in the transverse plane shows an exponential decay away from
the cross-sectional structure. There are no fundamental lower limits due to
the manufacturing process on the cross-sectional dimensional sizes and many
types are made of readily available material (glass). Fig. 7.3 shows the basic
configuration of an open waveguide. To transmit signals in the range of op-
tical frequencies over long distances from one point to another, one usually
employs a uniformly cylindrical, open waveguide of bounded cross-section.
Such a waveguide is commonly designated as an optical fiber. The free space
surrounding the fiber forms an unbounded part of its cross-section. If the
electromagnetic properties of the bounded part of its cross-section vary con-
tinuously with position in the cross-section, we speak of a graded-index fiber,
axial
direction
inhomogeneously
distributed
. penetrable
material
. transverse
plane
Figure 7.3. Open waveguide.
axis axis
mantle
fiber core
. .
.
(a) (b)
Figure 7.4. Open waveguides: (a) Circularly cylindrical graded-index fiber;

(b) circularly cylindrical multistep-index fiber.
electromagnetic waveguides 171
free space free space
layer . layers .
.
. .
substrate substrate
(a) (b)
Figure 7.5. Open waveguides: (a) planar graded-index fiber;

(b) planar multistep-index fiber.
while, if those properties are piecewise constant, we speak of a (multi)step-

index fiber (Fig. 7.4). In almost all cases one aims at a configuration that
is rotationally symmetric and deviations from the rotational symmetry (in
practice often unavoidable) are considered as imperfections.
For local processing of signals in an integrated-optics system, one of-
ten employs a planar waveguide configuration. This consists of a (semi-
infinite) substrate on which a film is deposited. A semi-infinite superstrate
of free space completes now the open waveguide. Again, we distinguish the
planar graded-index waveguide and the planar (multi)step-index waveguide
(Fig. 7.5).
In this chapter we will discuss the simplest closed-waveguiding structure,
the parallel-plate waveguide (see also Section 3.6), and the simplest open
waveguiding structure, the dielectric slab waveguide.
7.1. Parallel-plate waveguide
As we have seen in Section 3.6, the parallel-plate waveguide consists of

two parallel perfectly conducting plates a distance a apart (see Fig. 7.6).
Between these plates a homogeneous and lossless medium is assumed with
constitutive constants ε and µ. In contrast to the TEM-wave that we dis-
cussed in Section 3.6, we will now investigate whether an electromagnetic
wave of more general character can propagate in the parallel-plate wave-
guide. We now assume that the wavefield varies in the x1 -direction and is
independent of x2 , viz.,
Ê = Ê(x1 , x3 , s) , Ĥ = Ĥ(x1 , x3 , s) . (7.1)
The problem is then two-dimensional and in Chapter 4 we have seen that the
frequency-domain field equations (2.54) - (2.59) separate into two indepen-
dent sets of equations: one in which only the field components Ê1 , Ê3 and
Ĥ2 occur (parallel polarization) and one in which only the field components
Ĥ1 , Ĥ3 and Ê2 occur (perpendicular polarization).
In a source-free domain, we find the following sets of equations (s = jω,
ω ≥ 0, and σ = 0, cf. Eqs. (4.10) - (4.15))
• Parallel polarization
∂1 ∂1 Ĥ2 + ∂3 ∂3 Ĥ2 + ω 2 εµĤ2 = 0 , (7.2)
Ê1 = −(jωε)−1 ∂3 Ĥ2 , (7.3)

−1
Ê3 = (jωε) ∂1 Ĥ2 . (7.4)
In literature we often speak in this case of H-polarized waves or trans-
verse magnetic waves or TM-waves.
• Perpendicular polarization
∂1 ∂1 Ê2 + ∂3 ∂3 Ê2 + ω 2 εµÊ2 = 0 , (7.5)
Ĥ1 = (jωµ)−1 ∂3 Ê2 , (7.6)

−1
Ĥ3 = −(jωµ) ∂1 Ê2 . (7.7)
Here, we often speak in this case of E-polarized waves or transverse
electric waves or TE-waves.
x1 = a
ε, µ
propagation direction
....
......................................................................................
...
.....
i1
........
r
....
........................... i3
.... x1 = 0
O
Figure 7.6. A parallel-plate waveguide.
TE-waves
We will now first discuss in detail the case of E-polarized waves or TE-
waves. Since we are interested in the propagation of an electromagnetic wave
in the positive x3 -direction we will assume a solution of Eq. (7.5) of the type
Ê2 (x1 , x3 , jω) = Ê(x1 , jω) exp(−jk3 x3 ) , 0 < x1 < a . (7.8)
Substitution of Eq. (7.8) into the two-dimensional Helmholtz equation (7.5)

yields
∂1 ∂1 Ê + k12 Ê = 0 , (7.9)
in which
k12 = ω 2 εµ − k32 . (7.10)
Equation (7.9) constitutes a second-order differential equation with general

solution
Ê = Â exp(jk1 x1 ) + B̂ exp(−jk1 x1 ) , (7.11)

in which the frequency-dependent constants Â and B̂ follow from the bound-

ary conditions at the perfectly conducting plates. At these plates we have
the boundary conditions
lim Ê2 = 0 =⇒ lim Ê = 0 , (7.12)

x1 ↑a x1 ↑a
lim Ê2 = 0 =⇒ lim Ê = 0 . (7.13)

x1 ↓0 x1 ↓0
From Eqs. (7.12) - (7.13) we see that
Â exp(jk1 a) + B̂ exp(−jk1 a) = 0 , (7.14)

Â + B̂ = 0 , (7.15)
which leads to B̂ = −Â and
2j Â sin(k1 a) = 0 . (7.16)
For a non-zero value of Â, the latter equation can only be satisfied if
k1 a = mπ , for m = 0, 1, 2, · · · . (7.17)
Substitution of this result into Eq. (7.10) leads to the dispersion relation
mπ 2
k32 = ω 2 εµ − ( ) , m = 0, 1, 2, · · · . (7.18)
a
Introducing the wavenumber
1
k = ω(εµ) 2 , (7.19)
we find from Eq. (7.18) the propagation constants
⎧ 1
⎪
⎪ mπ 2 2 mπ
⎪
⎪ k −(
2
) when k ≥ ,
⎪
⎨ a a
k3 ≡ k3;m = (7.20)
⎪
⎪ 1
⎪
⎪ mπ 2 mπ
⎪
⎩ −j ( ) −k 2 2
when k < .
a a
With the superposition principle, the corresponding electromagnetic field

distributions then follow directly from Eqs. (7.6), (7.7), (7.8) and (7.11) as
∞
mπ
Ê2 = 2j Âm sin( x1 ) exp(−jk3;m x3 ) , (7.21)
m=1
a
∞
−k3;m mπ
Ĥ1 = 2j Âm sin( x1 ) exp(−jk3;m x3 ) , (7.22)
m=1
ωµ a
∞
−mπ mπ
Ĥ3 = 2j Âm cos( x1 ) exp(−jk3;m x3 ) . (7.23)
m=1
jωµa a
Note that the terms pertaining to m = 0 vanish identically. When 1 ≤

m ≤ ka/π, we have a finite set of propagating waves in the parallel-plate
waveguide. Each member of the set propagating in the positive x3 -direction
is denoted as a TEm -mode (m > 0). The frequency-dependent constants Âm
are determined by the electromagnetic source strengths. It is noted that in
contrast to the TEM-wave discussed in Section 3.6, the field distribution in
the transverse plane is not uniform. For ka/π < m < ∞, we find an infinite
number of non-propagating or evanescent waves. The x3 -dependence of the
non-propagating modes is given by
mπ 2 1
exp(−jk3;m x3 ) = exp{−[( ) − k 2 ] 2 x3 } . (7.24)
a
These modes decay in the positive x3 -direction. From Eqs. (7.21) - (7.22)
we find for these modes the time-averaged power flow density,
∗

1
2 Re (Ê × Ĥ ) · i3 = 1
2 Re −Ê2 Ĥ1∗
m m
∗

k3;m mπ
= 2 Re |Âm |2 [sin( x1 )]2 exp(−2jk3;m x3 )
ωµ a
= 0, (7.25)
since k3;m is imaginary. This result is typical for an evanescent wavefield.
TM-waves
Secondly, we will discuss in detail the case of H-polarized or TM-waves.
Since we are interested in the propagation of an electromagnetic wave in the
positive x3 -direction we will assume a solution of Eq. (7.2) of the type
Ĥ2 (x1 , x3 , jω) = Ĥ(x1 , jω) exp(−jk3 x3 ) , 0 < x1 < a . (7.26)
Substitution of Eq. (7.26) into the two-dimensional Helmholtz equation (7.2)

yields
∂1 ∂1 Ĥ + k12 Ĥ = 0 , (7.27)
in which
k12 = ω 2 εµ − k32 . (7.28)
Equation (7.27) constitutes a second-order differential equation with general
solution
Ĥ = Â exp(jk1 x1 ) + B̂ exp(−jk1 x1 ) , (7.29)
in which the frequency-dependent constants Â and B̂ follow from the bound-

ary conditions at the perfectly conducting plates. At these plates we have
the boundary conditions
lim Ê3 = 0 =⇒ lim ∂1 Ĥ = 0 , (7.30)

x1 ↑a x1 ↑a
lim Ê3 = 0 =⇒ lim ∂1 Ĥ = 0 , (7.31)

x1 ↓0 x1 ↓0
where Eq. (7.4) has been used. From Eqs. (7.30) - (7.31) we see that
Â exp(jk1 a) − B̂ exp(−jk1 a) = 0 , (7.32)

Â − B̂ = 0 , (7.33)
which leads to B̂ = Â and
2j Â sin(k1 a) = 0 . (7.34)
For a non-zero value of Â, the latter equation can only be satisfied if
k1 a = mπ , for m = 0, 1, 2, · · · . (7.35)
Substitution of this result into Eq. (7.28) leads to the dispersion relation
mπ 2
k32 = ω 2 εµ − ( ) , m = 0, 1, 2, · · · . (7.36)
a
Introducing the wavenumber
1
k = ω(εµ) 2 , (7.37)
we find from Eq. (7.36) the propagation constants
⎧ 1
⎪
⎪ mπ 2 2 mπ
⎪
⎪ k −(
2
) when k ≥ ,
⎪
⎨ a a
k3 ≡ k3;m = (7.38)
⎪
⎪ 1
⎪
⎪ mπ 2 mπ
⎪
⎩ −j ( ) −k 2 2
when k < .
a a
With the superposition principle, the corresponding electromagnetic field

distributions then follow directly from Eqs. (7.3), (7.4), (7.26) and (7.29) as
∞
mπ
Ĥ2 = 2 Âm cos( x1 ) exp(−jk3;m x3 ) , (7.39)
m=0
a
∞
k3;m mπ
Ê1 = 2 Âm cos( x1 ) exp(−jk3;m x3 ) , (7.40)
m=0
ωε a
∞
−mπ mπ
Ê3 = 2 Âm sin( x1 ) exp(−jk3;m x3 ) . (7.41)
m=0
jωεa a
When 0 ≤ m < ka/π, we have a finite set of propagating waves in the

parallel-plate waveguide. Each member of the set propagating in the positive
x3 -direction is denoted as a TMm -mode (m ≥ 0). The frequency-dependent
constants Âm are determined by the electromagnetic source strengths. For
ka/π < m < ∞, we find an infinite number of non-propagating or evanescent
waves. The term pertaining to m = 0 leads to the TM0 -mode with field
components
Ĥ2 = 2Â0 exp(−jkx3 ) , (7.42)

1
µ 2
Ê1 = 2Â0 exp(−jkx3 ) , (7.43)
ε
Ê3 = 0 , (7.44)
1
the latter mode being a TEM-mode with propagation constant k = ω(εµ) 2 ,
see Section 3.6.
7.2. Propagation properties of modes in a

parallel-plate waveguide
For a particular mode in the parallel-plate waveguide, the x3 -dependence

is given by
exp(−jk3;m x3 ) = exp(−γm x3 ) , (7.45)
where γm = jk3;m . With
γm = αm + jβm , (7.46)
see Eq. (3.46), where
αm = attenuation coefficient of the mode (Np/m),

βm = phase coefficient of the mode (rad/m),
we have from Eqs. (7.18) - (7.20), or Eqs. (7.36) - (7.38),

mπ 2
( ) = k 2 + αm
2
− βm
2
, (7.47)
a
0 = αm βm . (7.48)
1
In Eq. (7.47), k = ω(εµ) 2 with real constants ε and µ is independent of m.
Equations (7.47) and (7.48) yield two possibilities, viz.,
mπ 2 1
αm = 0 , βm = [k 2 − ( ) ]2 ≥ 0 , (7.49)
a
when 0 ≤ m ≤ ka/π, referring to propagating modes, and
mπ 2 1
αm = [( ) − k 2 ] 2 > 0 , βm = 0 , (7.50)
a
when ka/π < m < ∞, referring to non-propagating modes. The expressions
of Eqs. (7.49) - (7.50) can be cast into a different form by introducing the
cut-off angular frequency ωc,m of the m-th mode through
mπ
(εµ)− 2 .
1
ωc,m = (7.51)
a
propagation of modes in a parallel-plate waveguide 179
Introducing of the normalized quantities

ω αm a βm a
ωm = , αm = , βm = , (7.52)
ωc,m mπ mπ
we rewrite Eq. (7.49) and Eq. (7.50) as
1
αm = 0 , β m = (ω 2m − 1) 2 , (7.53)
when 1 ≤ ω m < ∞, and

1
αm = (1 − ω 2m ) 2 , βm = 0 , (7.54)
when 0 ≤ ω m < 1. These results lead to the dispersion curve of the m-th
mode in a lossless parallel-plate waveguide shown in Fig. 7.7.
. ..
αm .
.
. .....
...
....
.
. ..
. ....
βm .
.
. ...
....
. ..
. ....
6 .
.
.
....
...
. ..
2.0 .
.
. ....
...
....
.
. ..
. ....
...
.
. ...
. .
. ....
. ...
. ....
.
.
.
....
αm
. ...
. ....
.
.
. ....
1.0 ........................
..........
........
....... .
.
. ...
....
...... .
...... ...
..... . .
.....
......
...
...
βm
. .....
. ...
. ... ..
. ... ..
. ... ...
. ...
. ...
. ... ..
. ... ...
. ... ..
...
. ....
. ...
. ..
..
0 .
0 1.0 2.0 - ωm
" #$ % ........
evanescent propagation region
region
Figure 7.7. Dispersion diagram for the normalized

attenuation and phase coefficients.
Plane-wave decomposition
Each mode can be written as a superposition of two plane waves. For a
TEm -mode the electric field strength may be written as (cf. Eq. (7.21))
Ê2;m = Âm [exp(jk1;m x1 − jk3;m x3 ) − exp(−jk1;m x1 − jk3;m x3 )] , (7.55)
and for a TMm -mode the magnetic field strength may be written as (cf.
Eq. (7.39)
Ĥ2;m = Âm [exp(jk1;m x1 − jk3;m x3 ) + exp(−jk1;m x1 − jk3;m x3 )] , (7.56)
where
mπ 1
k1;m = , k3;m = [k 2 − k1;m
2
]2 . (7.57)
a
For propagating modes k3;m is real-valued and each plane-wave constituent
of Eqs. (7.55) - (7.56) is a uniform one. The direction of propagation of such
a uniform plane wave is determined by the vector {k1 , k3 } ≡ {k1;m , k3;m } for
a wave propagating in the positive x1 -direction and the positive x3 -direction
and by the vector {k1 , k3 } ≡ {−k1;m , k3;m } for a wave propagating in the
negative x1 -direction and the positive x3 -direction. These directions of prop-
agation can be constructed in the (k1 , k3 )-plane (see Fig. 7.8). Each prop-
agating mode becomes non-propagating or evanescent in the x3 -direction,
when the operating frequency becomes below the cut-off frequency for that
mode. In Fig. 7.8, this phenomenon is also illustrated in the (k1 , k3 )-plane.
Finally, in Fig. 7.9, we present the amplitude distributions of three guided
TEm -modes, m = 1, 2, 3.
propagation of modes in a parallel-plate waveguide 181
k1
.....
.........
..........
...
....
..
...
...
evanescent ...
...
modes ... 4π ......
..
...
...
... a ....
...
......... ...
.... ...
.
.... ...... ..
. .. .
.. ..............................................................
........
k3 = {k1;3 , k3;3 }
...........
....... . ...... ..... .
..
3π .... .........
........
... . .
. ... ... ......
... ....... ... ........... ............
... ..
........ a .
. . ......
... ....
.... . .
.
. ...
. ......
..... .
. .. .....
... . .
..... .
. ... .....
.....
... ... .
. .. .....
... ..
.............. .
.
. . .
. .....
... .... ......... .
. .. ....
k2 = {k1;2 , k3;2 }
.
... ..... .
. .. ....
... ..
. .
...... 2π .. .
. .
. ...
.
... ....... . ... ............
..
. . a .
. .. ....... .....
.
propagating ..
.
.
.. .
.....
..
.....
.
.
.
.
....
...
..
.. ........ . ...
...
... ..... ... ...... ...
modes .. ..... ... ... .....
.. ..... . ........ ...
...
...
...
.
..
.
.
.
.
k ......
.......
.....
.
.
.
.
.
..
. ..
.
.
....
.
.....
. ...
........ ...
...
...
. π .... .
k1 = {k1;1 , k3;1 }
.. ..... .
. .
. . . ...
... . ..... . . .. ....................
... .... .
. ..
. ...... .
......
... .
.
.....
..... a ... .
.
.
.
. .
......
. ..
.... .
. ... .
.......... ...
... ..
.
.....
. .
.
. ..
. ..
......
.... ............ ...
..... . .. ...
... .. ..... .... .... ...... ...........
... .
... ..... .
. .. ..
......
. .... . . .
............
. ...
...
... .
.
.
.....
. . ... .......
.
. . ................. ..
........ . ..... .. .. ..... .... ...... ..
... . . ..
..... .... ..... .......... . . ..
. ...
... ..
.. ...................... ..
........................................................................................................................................................................................................................................................................................................................................................................................................................
.
..
O k3
Figure 7.8. The (k1 , k3 )-plane for the lossless parallel-plate waveguide.
x1 x1 x1
6 6 6
TE1 -mode TE2 -mode TE3 -mode
........... .............. ...........................
.......... ................ .........
........ ............ ..............
......
.... ................
.................... .......
.............................................
.. .................. ..................
.............................
...... ........... .........................
........... .
..... .....
.... ..
. . .............. .........
........... ................... ...........................
- - -
Ê(x1 , jω) Ê(x1 , jω) Ê(x1 , jω)
Figure 7.9. The amplitude distributions of three guided TE-modes.

7.3. Dielectric slab waveguides
In this section we will discuss the simplest open-waveguiding structure:

the planarly layered, dielectric waveguide or dielectric slab waveguide. This
class of waveguides plays an important role in integrated optics. The di-
electric slab waveguide or thin-film waveguide consists of a homogeneous
dielectric slab or thin film sandwiched between the semi-infinite substrate
D(1) and the semi-infinite superstrate D(3) . The numbering of the domains
is illustrated in Fig. 7.10. The nomenclature and material properties of the
various domains are listed in Table 7.1. When the slab is homogeneous
(the case we discuss here), we refer to the waveguide as a step-index slab
waveguide, otherwise we are dealing with a graded-index slab waveguide.
When sub- and superstrate have the same electromagnetic properties, i.e.,
n(1) = n(3) , we have a symmetrical slab waveguide.
Like before we will now investigate whether a time-harmonic electro-
magnetic wave can propagate in the slab waveguide. We assume that the
wavefield varies only in x1 -direction and is independent of x2 , viz.,
Ê = Ê(x1 , x3 , jω) , Ĥ = Ĥ(x1 , x3 , jω) . (7.58)
In Section 7.1 we have seen that two types of waves can exist independently:
E-polarized waves or TE-waves (perpendicular polarization) with field com-
ponents Ĥ1 , Ĥ3 and Ê2 that satisfy Eqs. (7.5) - (7.7), and H-polarized waves
or TM-waves (parallel polarization) with the field components Ê1 , Ê3 and
Ĥ2 that satisfy Eqs. (7.2) - (7.4).
Table 7.1. The various subdomains and their electromagnetic properties.
subdomain x1 -coordinate permittivity permeability refractive index
D(3) a < x1 < ∞ ε(3) µ0 n(3)

D(2) 0 < x1 < a ε(2) µ0 n(2)
D(1) −∞ < x1 < 0 ε(1) µ0 n(1)
dielectric slab waveguides 183
superstrate ε(3) , µ0 D(3)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x1 = a
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . .
.
.
.
.
.
.
.
.
.
. . propagation direction . . . . . . . . . . . .
slab .
.
.
.
.
.
.
.
.
ε(2) , µ 0
.
.
.
.
.
.
....
....................................................................................
.. . .
.
.
.
.
.
.
.
.
.
.
.
. D(2)
. . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.......
.
. ....
i1 .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
r i3
. . . . . . . . . . . . . . . . .. . . . . . . . . . . . . .
...
. . . . . . . . . . . . . . . . ........................... . . . . . . . . . . . . . .
.... x1 = 0
O
substrate ε(1) , µ0 D(1)
Figure 7.10. A dielectric slab waveguide.
TE-waves
We will now first discuss in detail the case of E-polarized waves or TE-
waves. We are looking for a solution of the equations, cf. Eqs. (7.5) - (7.7),
∂1 ∂1 Ê2 + ∂3 ∂3 Ê2 + ω 2 ε(i) µ0 Ê2 = 0 , in D(i) , i = 1, 2, 3 , (7.59)
Ĥ1 = (jωµ0 )−1 ∂3 Ê2 , (7.60)

−1
Ĥ3 = −(jωµ0 ) ∂1 Ê2 , (7.61)
together with the boundary conditions for the tangential field components
Ê2 and Ĥ3 at x1 = a and x1 = 0. We will further assume that the wave
we are looking for will propagate in the positive x3 -direction and will have
a nonuniform, transverse intensity distribution that is possibly restricted to
the slab region. That means we will require that the wavefield vanishes for
|x1 | → ∞. In this respect it seems reasonable to assume a solution of the

type, see Eq. (7.8),
Ê2 (x1 , x3 , jω) = Ê(x1 , jω) exp(−jk3 x3 ) , −∞ < x1 < ∞ . (7.62)
Substitution of this representation into the two-dimensional Helmholtz equa-

tion (7.59) yields
(i)
∂1 ∂1 Ê + (k1 )2 Ê = 0 , (7.63)
in which
(i)
(k1 )2 = ω 2 ε(i) µ0 − k32 , i = 1, 2, 3 . (7.64)
Equation (7.63) constitutes a second-order differential equation. In each
subdomain D(i) , with i = 1, 2, 3, the general solution can be written as
(i) (i)
Ê = Â(i) exp(jk1 x1 ) + B̂ (i) exp(−jk1 x1 ) . (7.65)
(i) 1
In order to satisfy the condition at |x1 | → ∞, k1 = (ω 2 ε(i) µ0 −k32 ) 2 is chosen
(i)
such that Im(k1 ) ≤ 0, and in addition we take B̂ (1) = 0 and Â(3) = 0. The
other frequency-dependent constants Â(i) and B̂ (i) follow from the boundary
equations at x1 = a and x1 = 0, where
lim Ê2 = lim Ê2 =⇒ lim Ê = lim Ê , (7.66)
x1 ↑a x1 ↓a x1 ↑a x1 ↓a
lim Ĥ3 = lim Ĥ3 =⇒ lim ∂1 Ê = lim ∂1 Ê , (7.67)

x1 ↑a x1 ↓a x1 ↑a x1 ↓a
and
lim Ê2 = lim Ê2 =⇒ lim Ê = lim Ê , (7.68)
x1 ↓0 x1 ↑0 x1 ↓0 x1 ↑0
lim Ĥ3 = lim Ĥ3 =⇒ lim ∂1 Ê = lim ∂1 Ê . (7.69)

x1 ↓0 x1 ↑0 x1 ↓0 x1 ↑0

(2) (2) (3)
Â(2) exp(jk1 a) + B̂ (2) exp(−jk1 a) = B̂ (3) exp(−jk1 a) ,
(2) (2) (2) (2) (3) (3)
k1 Â(2) exp(jk1 a) − k1 B̂ (2) exp(−jk1 a) = −k1 B̂ (3) exp(−jk1 a) ,
Â(2) + B̂ (2) = Â(1) ,
(2) (2) (1)
k1 Â(2) − k1 B̂ (2) = k1 Â(1) . (7.70)
These four equations containing the four unknowns Â(2) ,B̂ (2) , B̂ (3) and Â(1)
have a nonzero solution only if the determinant of the matrix of their coef-
ficients is zero. In this manner we find an equation which is known as the
eigenvalue equation or characteristic equation, because it leads to specific
values or eigenvalues for the still unknown propagation coefficients k3 . In
the lossless guiding structure, we expect those solutions of Eq. (7.70) for
(1) (3) (2)
which k1 and k1 are both negative imaginary and k1 is positive real.
Therefore, it is advantageous to introduce the quantities
(1) 1
κ(1) = jk1 = [k32 − ω 2 ε(1) µ0 ] 2 > 0 ,
(2) 1
κ(2) = k1 = [ω 2 ε(2) µ0 − k32 ] 2 > 0 ,
(3) 1
κ(3) = jk1 = [k32 − ω 2 ε(3) µ0 ] 2 > 0 . (7.71)
Then, the characteristic equation which must be satisfied if a solution of
Eq. (7.70) is to be obtained, is found as

exp(jκ(2) a) exp(−jκ(2) a) − exp(−κ(3) a) 0

(2)
κ exp(jκ(2) a) −κ exp(−jκ a) −jκ(3) exp(−κ(3) a)
(2) (2) 0
=0.

1 1 0 −1

κ(2) −κ (2) 0 jκ(1)
(7.72)
By straightforward algebra, Eq. (7.72) reduces to the form
κ(1) κ(3)
(2)
+ (2)
tan(κ(2) a) = κ (1)κ (3) . (7.73)
κ κ
1 − (2) (2)
κ κ
To obtain the relevant values of k3 , it is advantageous to use the addition
formula
tan(u) + tan(v)
tan(u + v) =
1 − tan(u) tan(v)
and to rewrite Eq. (7.73) as

κ(1) κ(3)
κ(2) a = arctan + arctan + mπ, m = 0, 1, 2, · · · , ∞ .
κ(2) κ(2)
(7.74)
This equation, in which the quantities κ(i) are related to the propagation
coefficient k3 through Eq. (7.71), is called the dispersion equation. Its solu-
tions
k3 ≡ k3;m (ω) , m = 0, 1, 2, · · · , ∞ ,
form an innumerable set of real and complex numbers. The real values k3;m
that satisfy the inequalities in Eq. (7.71) form a finite set with
max[k (1) , k (3) ] < k3;m < k (2) , (7.75)
where
1
k (i) = ω(ε(i) µ0 ) 2 , i = 1, 2, 3 , (7.76)
with m = 0, 1, 2, · · · , M −1. Each value k3;m in Eq. (7.75) can be considered

as the propagation constant of the m-th guided mode of the thin-film wave-
guide. The electromagnetic field distribution corresponding to each k3;m is
denoted as the TEm -mode (m = 0, 1, 2, · · · , M−1) of the slab waveguide. The
(i)
quantities κ(i) corresponding to k3;m are denoted as κm . From Eq. (7.65) it
then follows that
⎧ (2) (3)
⎪
⎪
⎪
cos(κm a−ψm ) exp[−κm (x1 −a)] exp(−jk3;m x3 )
⎪
⎪
⎪
⎪ when a < x1 < ∞ ,
(1) ⎪
⎨ (2)
Âm cos(κm x1 −ψm ) exp(−jk3;m x3 )
Ê2;m =
cos(ψm ) ⎪
⎪
⎪ when 0 < x1 < a ,
⎪
⎪
⎪
⎪
(1)
cos(ψm ) exp(κm x1 ) exp(−jk3;m x3 )
⎪
⎩
when − ∞ < x1 < 0 ,
(7.77)
in which
(1)
κm
ψm = arctan (2)
. (7.78)
κm
The magnetic field strength then follows from Eqs. (7.60) - (7.61) as
⎧
⎪
⎪ −k3;m (2) (3)
⎪
⎪ cos(κm a−ψm ) exp[−κm (x1 −a)] exp(−jk3;m x3 )
⎪
⎪ ωµ
⎪
⎪
0
⎪
⎪
⎪
when a < x1 < ∞ ,
⎪
⎪ −k
(1)
Âm ⎨ 3;m (2)
cos(κm x1 −ψm ) exp(−jk3;m x3 )
Ĥ1;m = ωµ0
⎪
cos(ψm ) ⎪
⎪
⎪ when 0 < x1 < a ,
⎪
⎪
⎪
⎪ −k 3;m (1)
⎪
⎪ cos(ψm ) exp(κm x1 ) exp(−jk3;m x3 )
⎪
⎪ ωµ0
⎪
⎩
when − ∞ < x1 < 0 ,
(7.79)
and
⎧
⎪
⎪
(3)
κm
⎪
⎪
(2) (3)
⎪
⎪
⎪
⎪ jωµ 0
⎪
⎪ when a < x1 < ∞ ,
⎪
⎪
⎪
⎪
(1) ⎪
⎨ κm
(2)
(2)
Ĥ3;m =
Âm sin(κm x1 −ψm ) exp(−jk3;m x3 )
cos(ψm ) ⎪ jωµ0
⎪
⎪
⎪
⎪ when 0 < x1 < a ,
⎪
⎪
⎪ (1)
⎪ −κm cos(ψ ) exp(κ(1) x ) exp(−jk x )
⎪
⎪
⎪
⎪ m m 1 3;m 3
⎪
⎪ jωµ0
⎩
when − ∞ < x1 < 0 .
(7.80)
From Eqs. (7.77) and (7.80) it follows that the time-averaged power flow
density in the x1 -direction,
∗ ∗
1
2 Re[(Ê m × Ĥ m ) · i1 ] = 12 Re[Ê2;m Ĥ3;m ]=0, (7.81)
in each of the three domains. The time-averaged power flow density in the
x3 -direction follows from Eqs. (7.77) and (7.79) as
∗ ∗ ]
1
2 Re[(Ê m × Ĥ m ) · i3 ] = 12 Re[−Ê2;m Ĥ1;m
⎧
⎪
⎪ k3;m (2) (3)
⎪
⎪ [cos(κm a−ψm )]2 exp[−2κm (x1 −a)] , a < x1 < ∞ ,
⎪
⎪ ωµ0
⎪
⎪
⎪
⎪
Â(1) ⎪ 2
⎨
m k3;m (2)
= [cos(κm x1 −ψm )]2 , 0 < x1 < a ,
cos(ψm ) ⎪ ⎪ ωµ0
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪ k3;m
⎪
⎩
(1)
[cos(ψm )]2 exp(2κm x1 ) , −∞ < x1 < 0 .
ωµ0
(7.82)
From Eqs. (7.81) and (7.82) we observe that the time-averaged power flow
of a guided mode is directed in the longitudinal direction and is confined to
(1)
the slab region. The frequency-dependent constant Âm is determined by
the electromagnetic source strength.
TM-waves
We will now first discuss in detail the case of H-polarized waves or TM-
waves. We are looking for a solution of the equations, cf. Eqs. (7.2) - (7.4),
∂1 ∂1 Ĥ2 + ∂3 ∂3 Ĥ2 + ω 2 ε(i) µĤ2 = 0 , in D(i) , i = 1, 2, 3 , (7.83)
Ê1 = −(jωε(i) )−1 ∂3 Ĥ2 , i = 1, 2, 3 , (7.84)

(i) −1
Ê3 = (jωε ) ∂1 Ĥ2 , i = 1, 2, 3 , (7.85)
together with the boundary conditions for the tangential field components
Ĥ2 and Ê3 at x1 = a and x1 = 0. We will further assume that the wave
we are looking for will propagate in the positive x3 -direction and will have
a nonuniform, transverse intensity distribution that is possibly restricted to
the slab region. That means we will require that the wavefield vanishes for
|x1 | → ∞. In this respect it seems reasonable to assume a solution of the
type, see Eq. (7.26),
Ĥ2 (x1 , x3 , jω) = Ĥ(x1 , jω) exp(−jk3 x3 ) , −∞ < x1 < ∞ . (7.86)
Substitution of this representation into the two-dimensional Helmholtz equa-

tion (7.83) yields
(i)
∂1 ∂1 Ĥ + (k1 )2 Ĥ = 0 , (7.87)
in which
(i)
(k1 )2 = ω 2 ε(i) µ0 − k32 , i = 1, 2, 3 . (7.88)
Equation (7.87) constitutes a second-order differential equation. In each
subdomain D(i) , with i = 1, 2, 3, the general solution can be written as
(i) (i)
Ĥ = Â(i) exp(jk1 x1 ) + B̂ (i) exp(−jk1 x1 ) . (7.89)
(i) 1
In order to satisfy the condition at |x1 | → ∞, k1 = (ω 2 ε(i) µ0 −k32 ) 2 is chosen
(i)
such that Im(k1 ) ≤ 0, and in addition we take B̂ (1) = 0 and Â(3) = 0. The
other frequency-dependent constants Â(i) and B̂ (i) follow from the boundary
conditions at x1 = a and x3 = 0, where
lim Ĥ2 = lim Ĥ2 =⇒ lim Ĥ = lim Ĥ , (7.90)

x1 ↑a x1 ↓a x1 ↑a x1 ↓a
1 1
lim Ê3 = lim Ê3 =⇒ lim ∂1 Ĥ = lim ∂1 Ĥ , (7.91)
x1 ↑a x1 ↓a x1 ↑a ε(2) x1 ↓a ε(3)
and
lim Ĥ2 = lim Ĥ2 =⇒ lim Ĥ = lim Ĥ , (7.92)

x1 ↓0 x1 ↑0 x1 ↓0 x1 ↑0
1 1
lim Ê3 = lim Ê3 =⇒ lim ∂1 Ĥ = lim (1) ∂1 Ĥ . (7.93)
x1 ↓0 x1 ↑0 x1 ↓0 ε(2) x1 ↑0 ε

(2) (2) (3)
Â(2) exp(jk1 a) + B̂ (2) exp(−jk1 a) = B̂ (3) exp(−jk1 a) ,
(2) (2) (3)
k1 (2) k1 (2) k1 (3)
(2)
Â(2) exp(jk1 a) − (2) B̂ (2) exp(−jk1 a) = − (3) B̂ (3) exp(−jk1 a) ,
ε ε ε
Â(2) + B̂ (2) = Â(1) ,
(2) (2) (1)
k1 k1 k1
(2)
Â(2) − (2) B̂ (2) = (1) Â(1) . (7.94)
ε ε ε
These four equations containing the four unknowns Â(2) , B̂ (2) , B̂ (3) and
Â(1) have a nonzero solution only if the determinant of the matrix of their
coefficients is zero. In this manner we find an equation which is known as
the eigenvalue equation or characteristic equation, because it leads to specific
values or eigenvalues for the still unknown propagation coefficients k3 . In

the lossless guiding structure, we expect those solutions of Eq. (7.94) for
(1) (3) (2)
which k1 and k1 are both negative imaginary and k1 is positive real.
Therefore, it is advantageous to introduce the quantities
(1) 1
κ(1) = jk1 = [k32 − ω 2 ε(1) µ0 ] 2 > 0 ,
(2) 1
κ(2) = k1 = [ω 2 ε(2) µ0 − k32 ] 2 > 0 ,
(3) 1
κ(3) = jk1 = [k32 − ω 2 ε(3) µ0 ] 2 > 0 . (7.95)
Then, the characteristic equation which must be satisfied if a solution of

Eq. (7.94) is to be obtained, is found as

exp(jκ(2) a) exp(−jκ(2) a) − exp(−κ(3) a) 0

(2) (2)
κ κ jκ(3)
(2) − (2) exp(−jκ(2) a) − (3) exp(−κ(3) a) 0
ε(2) exp(jκ a) ε ε
=0.

1 1 0 −1
κ(2) κ (2) jκ(1)
− (2)
(2)
0 (1)
ε ε ε
(7.96)
By straightforward algebra, Eq. (7.96) reduces to the form
ε(2) κ(1) ε(2) κ(3)

(1) (2)
+ (3) (2)
tan(κ(2) a) = ε κ(2) (1)ε (2)κ (3) . (7.97)
ε κ ε κ
1 − (1) (2) (3) (2)
ε κ ε κ
To obtain the relevant values of k3 , it is advantageous to use the addition
formula
tan(u) + tan(v)
tan(u + v) =
1 − tan(u) tan(v)
and to rewrite Eq. (7.97) as

ε(2) κ(1) ε(2) κ(3)
κ(2) a = arctan (1) (2) + arctan (3) (2) + mπ, m = 0, 1, 2, · · · , ∞ .
ε κ ε κ
(7.98)
This equation, in which the quantities κ(i) are related to the propagation
coefficient k3 through Eq. (7.95), is called the dispersion equation. Its solu-
tions
k3 ≡ k3;m (ω) , m = 0, 1, 2, · · · , ∞ ,
form an innumerable set of real and complex numbers. The real values k3;m
that satisfy the inequalities in Eq. (7.95) form a finite set with
max[k (1) , k (3) ] < k3;m < k (2) , (7.99)
where
1
k (i) = ω(ε(i) µ0 ) 2 , i = 1, 2, 3 , (7.100)
with m = 0, 1, 2, · · · , M−1. Each value k3;m in Eq. (7.99) can be considered as
the propagation constant of the m-th guided mode of the thin-film waveguide.
The electromagnetic field distribution corresponding to each k3;m is denoted
as the TMm -mode (m = 0, 1, 2, · · · , M − 1) of the slab waveguide. The
(i)
quantities κ(i) corresponding to k3;m are denoted as κm . From Eq. (7.89) it
then follows that
⎧ (2) (3)
⎪
⎪
⎪
⎪
⎪
⎪
⎪ when a < x1 < ∞ ,
(1) ⎪
⎨ (2)
Ĥ2;m =
cos(ψm ) ⎪
⎪
⎪ when 0 < x1 < a ,
⎪
⎪
⎪
⎪
(1)
cos(ψm ) exp(κm x1 ) exp(−jk3;m x3 )
⎪
⎩
when − ∞ < x1 < 0 ,
(7.101)
in which
(1)
ε(2) κm
ψm = arctan (1) (2) . (7.102)
ε κm
The electric field strength then follows from Eqs. (7.84) - (7.85) as
⎧
⎪
⎪
k3;m (2) (3)
⎪
⎪ cos(κm a−ψm ) exp[−κm (x1 −a)] exp(−jk3;m x3 )
⎪
⎪ ωε(3)
⎪
⎪ when a < x1 < ∞ ,
⎪
⎪
⎪
⎪
(1) ⎨ 3;m
k (2)
Ê1;m = ωε(2)
⎪
cos(ψm ) ⎪
⎪
⎪ when 0 < x1 < a ,
⎪
⎪
⎪
⎪ k3;m (1)
⎪
⎪
⎪ (1)
⎩ ωε
when − ∞ < x1 < 0 ,
(7.103)
and
⎧ (3)
⎪
⎪ −κm
⎪
⎪
(2) (3)
⎪
⎪ (3)
⎪
⎪ jωε
⎪
⎪ when a < x1 < ∞ ,
⎪
⎪
⎪
⎪
(1) ⎪
⎨ −κm
(2)
(2)
Ê3;m =
Âm sin(κm x1 −ψm ) exp(−jk3;m x3 )
cos(ψm ) ⎪ jωε(2)
⎪
⎪
⎪
⎪ when 0 < x1 < a ,
⎪
⎪
⎪
⎪
(1)
κm
⎪
⎪ (1)
⎪
⎪
⎪ jωε (1)
⎩
when − ∞ < x1 < 0 .
(7.104)
From Eqs. (7.101) and (7.104) it follows that the time-averaged power flow
density in the x1 -direction,
∗ ∗
1
2 Re[(Ê m × Ĥ m ) · i1 ] = 12 Re[−Ê3;m Ĥ2;m ]=0, (7.105)
in each of the three domains. The time-averaged power flow density in the
x3 -direction follows from Eqs. (7.101) and (7.103) as
∗ ∗ ]
1
2 Re[(Ê m × Ĥ m ) · i3 ] = 12 Re[Ê1;m Ĥ2;m
⎧
⎪
⎪ k3;m (2) (3)
⎪
⎪ [cos(κm a−ψm )]2 exp[−2κm (x1 −a)] , a < x1 < ∞ ,
⎪
⎪ ωε(3)
⎪
⎪
2⎪
Â (1) ⎪
⎨
m k3;m (2)
=
cos(ψm ) ⎪
[cos(κm x1 −ψm )]2 , 0 < x1 < a ,
⎪
⎪ ωε(2)
⎪
⎪
⎪
⎪
⎪
⎪ k3;m
⎪
⎩ (1)
[cos(ψm )]2 exp(2κm x1 ) , −∞ < x1 < 0 .
ωε(1)
(7.106)
From Eqs. (7.105) and (7.106) we observe that the time-averaged power flow
of a guided mode is directed in the the longitudinal direction and is confined
(1)
to the slab region. The frequency-dependent constant Âm is determined by
the electromagnetic source strength.
propagation properties of a dielectric slab waveguide 193
7.4. Propagation properties of guided modes in a

dielectric slab waveguide
It is noted that the propagation constant k3;m = k3;m (ω) is dependent

on the angular frequency ω for both TEm and TMm -modes, i.e., the guided
modes in a slab waveguide like the propagating modes in a parallel-plate
waveguide show waveguiding dispersion or geometrical dispersion. This dis-
persion is illustrated in Fig. 7.11, where the effective index of refraction
(mode index)
k3;m
neff ,m = (7.107)
k0
of the m-th guided mode is presented as a function of a/λ0 , in which

2π ω
λ0 = , k0 = , (7.108)
k0 c0
where k0 is the free-space wavenumber. Note that the effective index of a
mode is located in
max[n(1) , n(3) ] < neff ,m < n(2) . (7.109)
In terms of the phase velocity vφ,m of the m-th guided mode
ω
vφ,m = (7.110)
k3;m
we have
c0
neff ,m = . (7.111)
vφ,m
When the slab waveguide is used as a transmission channel, we often intro-
duce the group velocity as
1
vg,m = . (7.112)
∂ω k3;m
From Fig. 7.11 we observe that in general the guided (or confined) modes
have a cut-off frequency. Assuming that n(1) > n(3) , the cut-off frequencies
of the guided modes follow from the condition κ(1) (ω) = 0. Substitution of
k32 = ω 2 ε(1) µ0 = k02 (n(1) )2 in Eqs. (7.71) and (7.74) yields an equation from
which the cut-off frequencies of the TE-mode can be derived as
3.5 ..............................................................
............................
....................................................
...............................................
neff ,m n(3) = 1.0 .....
. ..
...
.
......................
..
...
.
.
. .
.
..
..
..
................................................
..............................
....................... ...........................................
n(2) = 3.5 .....
....................
..................... ......... .
.
......................................
............................................
6
. ..
... .... ...
....... ...... ........................
...... ...... .......................
n(1) = 3.2 ...... ......
...... ......
.......... .........
......... ........
........................ .........
..... ......... . .....................
..
..
.... .... .
.. .. . . . . .....................
..... ....... ....... ........ .......... .........
... . . .. .. ................... ....... . ...........................
.. .. ...... ...... .. .
... ... ......... .........
3.4 ... ...
... ...
...... ......
...... .....
...... ......
........ .......
....... .......
....... .......
TE 0....... ....... ....... ......... ....................
.. .. TE . .
1............ ............
..
.. ..
...... .......
...... .....
........
........ ....
... ...
... ...
. .
.... .....
TE 2............................. ....... .......
....... .......
.... ... ..
.... .......
... ...
...... ......
..... .....
TE 3 .
.....
...... .....
..................
... ... TM ... .... ..... ..... ..... ......
... ... 0 ... .... ..... ...... ..... .......
... ...
.. ...
... ...
... ..... 1 TM ..... ....
..... ........
...... .....
..... ..........
.
... ...
..
. .
... ....
.
. . ..
..... .....
.
TM 2 . ..
..... .....
..... ......
. . .
... ...
... ...
... ...
... ...
.... ....
.... ...
..... .....
..... .....
TM 3
3.3 .... ....
.. ..
.
.
.. ....
.
... ...
.
... ...
.. ... ..... ...
...... .........
..... .....
.... .... ... .... . . ... ..... ..
...... .........
.
. . ... ... .. .. .... ....
.. .. ... ... .... ....
.. .. ... ... ... ... ... ....
.. .
.. .. ... ... ... ... ... ....
... .... ... ... ..... ..... ...... ......
.. .. ... ..
. .. .. .. ..
.. .. .. ... ... ... ... ...
.. .. ... .. ... .. ... ...
... .. .. ... ... ....
... .... ... .... ..... ..... ..... ......
.. .. .. ..
.. ..
.. ... ... .....
.. .. ... ... .... ....
.. ... ... ... ... .... .... ..
... .... .... .... ..... ..... .... ...
3.2
0 0.5 1.0 1.5 2.0
- a/λ0
Figure 7.11. Dispersion curves for the guided modes of a GaAs film located on a
AlGaAs substrate (a = film thickness, λ0 = 2π/k0 = wavelength in free space).
! 1 &
c0 [(n(1) )2 − (n(3) )2 ] 2
ωc,m = 1 arctan 1 + mπ .
a[(n(2) )2 − (n(1) )2 ] 2 [(n(2) )2 − (n(1) )2 ] 2
(7.113)
Substitution of k32 = ω 2 ε(1) µ0 = k02 (n(1) )2 in Eqs. (7.98) and (7.95) yields an
equation from which the cut-off frequencies of the TM-mode can be derived
as
! 1 &
c0 [(n(1) )2 − (n(3) )2 ] 2 ε(2)
ωc,m = 1 arctan 1 + mπ .
a[(n(2) )2 − (n(1) )2 ] 2 [(n(2) )2 − (n(1) )2 ] 2 ε(3)
(7.114)
The mode with the lowest cut-off frequency (the TE0 -mode) is often referred
as the fundamental mode.
Like for the parallel-plate waveguide in Section 6.2 we show in Fig. 7.12,
(2)
the (k1 , k3 )-plane for three different TE-modes in a lossless, asymmetric,
dielectric slab waveguide. Further it is interesting to investigate the influence
of the operating frequency on the amplitude distribution of the guided mode.
(2)
k1
..
.......
.......
.. .... ..
...
....
...
...
...
...
...
...
...
...
...
...
...
...
.
... ..
. ..... ..
... ..
.........................................................
...... .. .
.. ..
.......... ....
...........
.........
...
... ....... ..... ... ........
.......
..
....... ... .......
.. ...... (2)
{k1;2 , k3;2 }
. ......
... ......
.. .
.
. ......
... .......
. .
. ......
... ...
. .
.
. ............. ..........
... .
...... .
.
. .
.. .....
... ..
............. .
.
. .... .....
.....
... ..... ........... .... ...
.. ....
.... ...... .....
..... .
. ... ....
.
... .. .
.
. .... ...
...
..
. ..... .
. .
. ...
. . . (2)
propagating {k1;1 , k3;1 }
.. .
..... .
. ... ...
.. ....... .
. ..
... ..... .. ... ...........
. . ... ......... ....
modes ..
. .....
.....
.
.
. ... .
.... ...
.. ...
k (2)
..
. . .
.
. ... . .
... . ...
... ..... . ....
... ... ..... .... ... ....... ...
... ..... ... ... ....... ...
...
... ..
.
.....
.. .
. .
.... ............ ...
... ....
.....
..... ..
. .... .......... ...
...
... ... ..... .
.
. .
... ........
. ...
..... .. .... (2)
{k1;0 , k3;0 }
... . .
. . .
... .. .
... .
.
..
......
..
.
.
.
. ...
..
...
.........
.
... ... . .
. .
.........................
.
... ...
.
.....
.. .
.
. .
...
.
...
...... ....
... ... ... ............ ..
...
.....
..... .... ..... .............
. .......
..........
.. .
.. ................ ..
...
..... .. ... ....... ................ ..
.... ..... .. .......... .............................. ..
... ......................... ..
.........................................................................................................................................................................................................................................................................................................................................................................................................................
..
O k3
(2)
Figure 7.12. The (k1 , k3 )-plane for the lossless dielectric slab waveguide.
In Fig. 7.13, we have shown the dispersion curve of a TE0 -mode and its
amplitude distribution for three different values of the operating frequency.
Close to cut-off, a mode stretches far out in substrate and superstrate. Its
effective index of refraction neff approaches the value n(1) . Far from cut-off,
the mode is entirely concentrated in the thin film. As can be expected, now
its effective index neff approaches the value n(2) .
In Fig. 7.14, we have shown the dispersion curves of the TEm -modes,
m = 0, 1, 2, and their amplitude distributions for a particular frequency of
operation.
It is finally noted that from Eqs. (7.77) and (7.101), it can be seen that
in the substrate and the superstrate a guided mode locally behaves as a
non-uniform plane wave. The planes of equal phase and equal amplitude are
perpendicular to each other.
n(2) ..............................
............................................................
.......................................
.................................
neff ,m n(3) = 1.0 .
...
...
...
..
.
..................
.......................
..
...........
n(2) = 3.5 .........
........
........
6 n(1) = 3.2 .
......
..
.........
......
.....
.....
.
.....
...
...
...
TE ...
0......
.
...
...
....
...
...
...
..
.
...
...
..
..
...
..
...
...
....
.
..
..
...
....
.
...
..
..
..
.
..
...
n(1) 0
ω1 ω2 ω3 ω4 4πc0 /a
- ω
ωc,0
x1 x1 x1 x1
...
...
6 ..... 6 .... 6 .... 6
... ..
.. ... ...
... ... ... ...
...
...
.....
......
...
...
...
.
.
.
.
. .. .
...
...
.
n(3)
....... ....... ... ...
........... ....... .......
..... ..... .........
. . . . . . . . ... . . . . . ...... . . . . ......... . . . . ................. . . . . . . . . . . .
. . . . . . . . ..... . . . . . . .... . . . . .
....... . .............
.... . . . . . . . . . . .
. .... . . .... . . ..... . ..... .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. ..... .
.
.
.
.
.
.
.
. ..... . .
.
.
.
.
.
...... .
.
.
.
.
.
.
.
. ........ .
.
.
.
.
.
.
.
.
.
.
.
. n(2)
. . . . . . . . . .... . . . . . ........ . . . . .............. . . . . . .......... . . . . . . . . . .
. ... .... .........
... ...... ........ .........
..
. ....... ....... .....
.
... ...
. ..
. ..
.
... ... ... ....
...
..
...
..
...
....
...
...
n(1)
..
..
.
...
- .
.
.. -
.
.
.. -
.... -
Ê(x1 , jω1 ) Ê(x1 , jω2 ) Ê(x1 , jω3 ) Ê(x1 , jω4 )
Figure 7.13. The influence of the operating frequency on the amplitude

distribution of the TE0 -mode.
n(2) .......................................
............................................................
..............................
.................................
neff ,m n(3) = 1.0 ..
....
.....
..................
....................... .......................
.......... .....................
n(2) = 3.5 .......
.........
..........
..............
................
..................
6 .
...
.. ....
......
.
.. ....
n(1) = 3.2 ......
.....
......
.........
..........
...........
............
...
..... ........ ...........
......
. .. ......... ..........
.
.
.
.... ............
...
..............
..... ..
... ...... .........
... ...... ........
... ...... .......
... .......
TE 0...... .
. ....... .
...........
.
...
TE 1............
... .....
......
......
... ....
... TE 2..............
.... .
.. ...
..
.. .....
... .... .....
... ... .....
... ... .....
..
. ....
. .......
... .. ....
... .....
... ... ....
.. ... ...
..
... .... .....
. ..
.. ... ...
... ... ...
... ... ...
.... ... .....
. ... ..
.. ...
.. ... ...
... ... ...
.... ... ....
. .. ..
... .. ...
.. .. ...
.. .. ..
..
. ... ....
.. .. ...
... ... .....
n(1) 0
ωc,0 ωc,1 ωc,2 ω 4πc0 /a
- ω
x1 x1 x1
.... 6 ... 6 6 ....
... .... ...
.. .. ..
... ...
....
0 TE -mode ..
...
..
1 TE -mode ....
TE -mode 2
n(3)
..
...... ...
...
. ..........
.......... ................
. ........................
.
... ...
. . . . . . . . ..
. ........... . . . ........ .. ... .. . . . . . . . . . . . . ................. . . . . . . . .
. . . . . . . . .
. .......... . . .
. ................. . . . . . . . . . ......................... . ... . . . . . .
. ... ................. . ........................ .
. . . . . . . . . . . . ................ . . . . . ................. . . . . . . . . (2)
. . . . . . . . . ..
.
...... . . . . . . .
.....
.................. . . . ... ...
... ......................... . . . . . . . . n
. . . . . . . . . ............ . . . . . . . . . ......... . . . . . . ...................................... . . . . . . . .
.. .. .
......... ....
............. ........
..... ....... .........
... .... .....
.
.....
.... .... ...
.
.. .. ...
....
... ...
...
...
...
n(1)
..
. - ....
. - ....
. -
Ê(x1 , jω ) Ê(x1 , jω ) Ê(x1 , jω )
Figure 7.14. The amplitude distribution of three guided TE-modes at

ω = ω = 2πc0 /a.
Exercise 7.1
A parallel-plate waveguide is used to guide TE-waves. The configuration is
depicted in Fig. 7.6 with ε = 4ε0 , µ = µ0 and a = 10 cm while the frequency
of operation is f = 1 GHz.
(a) Find the propagating and the evanescent modes.
(b) Give the expressions for the electric and magnetic field strengths of
the propagating mode(s).
Now consider a perfectly conducting wall positioned at x3 = 0 inside the

waveguide.
( c ) Find the total electromagnetic field in the waveguide.
Exercise 7.2
Find the maximum power that can be propagated in an air-filled parallel-
plate waveguide, with width a = 1 cm and characteristic length w = 5 cm,
without causing breakdown. In air breakdown occurs at approximately E =
2 × 106 V/m. Use a safety factor of 10.
Exercise 7.3
For frequencies above the cutoff frequencies, the propagation constants of
a parallel-plate waveguide are associated with the so-called ‘guided wave-
lengths’, λg,m , as k3;m = λ2π
g,m
for ω > ωc,m . Find the guided wavelengths of
the TEm -modes, m = 0, 1, 2, · · ·, with f = 300 MHz, c = 3 × 108 m/s and
a = 0.75 m.
Exercise 7.4
EHF electromagnetic waves, see Table 3.1, (also called millimeter waves) may
be guided by dielectric slab waveguides. Consider a slab, with ε = 10ε0 , in
free space. By selecting a proper thickness a it may be achieved that for
any frequency of the entire EHF frequency band only the TE0 -mode can
propagate in the slab. What is the range of possible values of a?
Exercise 7.5
Consider a symmetric dielectric slab waveguide in the following configura-
tion:
Superstrate : a < x1 < ∞, ε(1) ,
Slab : −a < x1 < a, ε(2) ,
Substrate : −∞ < x1 < −a, ε(1) .
Solve the following problems in case of TE-waves.

(a) Give the equations in the four unknowns Â(1) , Â(2) , B̂ (2) , B̂ (3) that arise
from application of the boundary conditions at the two interfaces. Use
(1) 1 (2)
the substitutions κ(1) = jk1 = [k32 − ω 2 ε(1) µ0 ] 2 > 0 and κ(2) = k1 =
1
[ω 2 ε(2) µ0 − k32 ] 2 > 0.
(b) Solve the equations for Â(2) and B̂ (2) in terms of Â(1) and also in terms
of B̂ (3) .
( c ) Use the results obtained in (b) to eliminate Â(2) and B̂ (2) and find
the two solutions with for each solution the corresponding relation
(2)
for κm a. Hint: One solution is found for m = 0, 2, 4, · · · (the even
solution), while the other is found for m = 1, 3, 5, · · · (the odd solution).
(d) Use the results in (b) and (c) to show that the electric field strength
is given by ⎧ (2) (1)
⎪
⎪ cos(κm a) exp[−κm (x1 − a)] exp(−jk3;m x3 ),
⎪
⎪
⎪
⎪ a < x1 < ∞,
⎪
⎨
(even) cos(κ(2) x ) exp(−jk −a < x1 < a,
Ê2;m = Ĉm m 1 3;m x3 ),
⎪
⎪
⎪
⎪ (2) (1)
⎪
⎪ cos(κm a) exp[κm (x1 + a)] exp(−jk3;m x3 ),
⎪
⎩
−∞ < x1 < −a,
with m = 0, 2, 4, · · ·, and
⎧ (2) (1)
⎪
⎪ sin(κm a) exp[−κm (x1 − a)] exp(−jk3;m x3 ),
⎪
⎪
⎪
⎪ a < x1 < ∞,
⎪
⎨ (2)
(odd)
Ê2;m = Ĉm sin(κm x1 ) exp(−jk3;m x3 ), −a < x1 < a,
⎪
⎪
⎪
⎪ (2) (1)
⎪
⎪−sin(κm a) exp[κm (x1 + a)] exp(−jk3;m x3 ),
⎪
⎩
−∞ < x1 < −a,
(even) (odd)
with m = 1, 3, 5, · · ·, and express the coefficients Ĉm and Ĉm in
(1)
terms of Âm .
Problem 7.1
For a parallel-plate waveguide find the relations for the phase velocity and
the group velocity, for propagating modes, and express them in terms of the
actual wave speed, c, the angular radial frequency of operation, ω, and the
modal cutoff angular frequency, ωc,m .
Problem 7.2
A dielectric slab is at x1 = 0 coated with a perfectly conducting foil, see
Fig. 7.15. Find the solution for the electric field strength in the configuration
in case of guided TE-waves in the slab. Hint: you may use your results in
Exercise 7.5 to your advantage.
ε(1) , µ0
x1 = a
ε(2) , µ0 propagation
.......
direction
................................................................
....
... i1
..........
p ... ... i3
.................... x1 = 0
Figure 7.15. A coated-slab waveguide.
Problem 7.3
In this chapter we have considered one open waveguide structure, viz., the
slab waveguide, and have found solutions for guided electromagnetic waves.
One can also investigate whether guided wave solutions exist in a two-media
configuration (see Fig. 7.16). In that case the guided wave is a surface wave
that propagates along the interface between two different dielectric media
and is evanescent in both directions away from the interface. We simply
assume a solution for TM-waves of the type
Ĥ2 (x1 , x3 , jω) = Ĥ(x1 , jω) exp(−jk3 x3 ), −∞ < x1 < ∞ ,
and
Ĥ = B̂ (1) exp(−κ(1) x1 ), 0 < x1 < ∞ ,
Ĥ = Â (2)
exp(κ (2)
x1 ), −∞ < x1 < 0 ,
ε(1) , µ0 .. i1
..........
p .... .. i3
..................... x1 = 0
ε(2) , µ 0 propagation
.......
direction
................................................................
....
Figure 7.16. A two-media configuration.

(1) 1 (2) 1
with κ(1) = jk1 = [k32 − ω 2 ε(1) µ0 ] 2 > 0 and κ(2) = jk1 = [k32 − ω 2 ε(2) µ0 ] 2 > 0,
where the condition Ĥ2 → 0 if |x1 | → ∞ is satisfied.
(a) Under what condition can there be a solution of this form?
(b) For an electron plasma the permittivity has a negative sign, when the
frequency is below the plasma frequency. When one of the two media
is such a plasma, give the solution for Ĥ2 and draw the amplitude of
the magnetic field as a function of x1 .
( c ) Investigate if there is a solution if ε(1) = ε(2) . If there is one, give the

solution.
Chapter 8
Excitation of
Two-dimensional
In Chapter 4 we have discussed the properties of two-dimensional electro-

magnetic plane waves in a homogeneous medium. In Chapter 3 we have seen
that an infinite, planar, electric-current sheet emits a one-dimensional plane
wave, which is a special case of a two-dimensional wave. We will discuss the
generation of two-dimensional waves by a planar electric-current sheet that
has a finite dimension in the i1 -direction and an infinite dimension in the i2 -
direction. This antenna emits a two-dimensional electromagnetic field that
can be composed through superposition of an infinite number of different
plane waves, both of uniform and non-uniform character. We consider the
cases that the electric current in the emitter flows either in the i1 -direction
or in the i2 -direction.
204 excitation of two-dimensional electromagnetic waves
8.1. The sheet emitter with a parallel electric

current
plane x3 = 0 (Fig. 8.1). Let the impressed electric current be uniform as a
function of x2 and flow from x1 = 12 a to x1 = − 12 a, then
!
−Iˆ∆ (x1 , s)δ(x3 ) , |x1 | < 12 a ,
Jˆ1ext = Jˆ2ext = 0 , Jˆ3ext = 0 , (8.1)
0, |x1 | > 12 a ,
and Iˆ∆ (in A/m) is the electric current per unit length (of the x2 -direction).
K̂1ext = 0 , K̂2ext = 0 , K̂3ext = 0 . (8.2)
The medium is assumed to be homogeneous and lossless with constitutive
constants ε and µ. Since the configuration is independent of x2 , the elec-
tromagnetic field is two-dimensional and in Chapter 4 we have seen that
the frequency-domain field equations (2.54) - (2.59) separate into two in-
dependent set of equations, cf. Eqs. (4.4) - (4.6) and Eqs. (4.7) - (4.9).
From Eqs. (8.1) - (8.2) we conclude that K̂1ext , K̂3ext and Jˆ2ext are equal to
zero, hence only a parallelly polarized electromagnetic field occurs. The non-
zero field components Ê1 , Ê3 and Ĥ2 satisfy the following set of equations
(s = jω, σ = 0, cf. Eqs. (4.4) - (4.6))
!
Iˆ∆ (x1 , jω)δ(x3 ) , |x1 | < 12 a ,
∂3 Ĥ2 + jωεÊ1 = (8.3)
0, |x1 | > 12 a ,
−∂1 Ĥ2 + jωεÊ3 = 0 , (8.4)
∂3 Ê1 − ∂1 Ê3 + jωµĤ2 = 0 . (8.5)
In Chapter 4 we have seen that in a domain outside the source distribution,
solutions of Eqs. (8.3) - (8.5) in the form of plane waves exist. We therefore
assume that the general solution consists of an infinite superposition of plane-
wave constituents
⎧
⎪
⎪ 1 ∞
⎪
⎨ 2π 1 (k1 , jω) exp(−jk1 x1 − jk3 x3 )dk1 , x3 > 0 ,
ê+
k1 =−∞
Ê1 (x1 , x3 , jω) =
⎪
⎪ 1 ∞
⎪
⎩ ê− (k1 , jω) exp(−jk1 x1 + jk3 x3 )dk1 , x3 < 0 ,
2π k1 =−∞ 1
(8.6)
sheet emitter with a parallel electric current 205
ε, µ ...........
.......... ..............
........
ε, µ
' (...................
.......
.......
......
... x = 12 a 1
H .....
...
. H
........ i1
....
O r ............................
...
...
...
... i3
...
..........
.......
' J ext .
x1 = − 12 a (
..... .......
....... .......
....... .......
............. .........
..
.
..... H H ..........
x3 < 0 x3 > 0

parallelly polarized electromagnetic waves.
⎧
⎪
⎪ 1 ∞
⎪
⎨ ê+ (k1 , jω) exp(−jk1 x1 − jk3 x3 )dk1 , x3 > 0 ,
2π k1 =−∞ 3
Ê3 (x1 , x3 , jω) =
⎪
⎪ 1 ∞
⎪
2π k1 =−∞ 3
(8.7)
and
⎧
⎪
⎪ 1 ∞
⎪
⎨ ĥ+ (k1 , jω) exp(−jk1 x1 − jk3 x3 )dk1 , x3 > 0,
2π k1 =−∞ 2
Ĥ2 (x1 , x3 , jω) =
⎪
⎪ 1 ∞
⎪
⎩ ĥ− (k1 , jω) exp(−jk1 x1 + jk3 x3 )dk1 , x3 < 0,
2π k1 =−∞ 2
(8.8)
where ⎧
⎨
1 1
(ω 2 εµ − k12 ) 2 , |k1 | ≤ ω(εµ) 2 ,
k3 = (8.9)
⎩ −j(k 2 − ω 2 εµ) 21 , |k1 | > ω(εµ) 2 .
1
1
Note that the definition of the square root of k3 takes care of the bounded-
ness of the integrals as |x3 | → ∞. This guarantees the boundedness of the
electromagnetic field components as |x3 | → ∞. From Eqs. (8.3) and (8.4) it
− − −
follows that the quantities ê+ + +
1 , ê3 , ê1 and ê3 are related to ĥ2 and ĥ2 as
k3 + −k1 +
ê+
1 = ĥ , ê+
3 = ĥ , (8.10)
ωε 2 ωε 2
−k3 − −k1 −
ê−
1 = ĥ , ê−
3 = ĥ . (8.11)
ωε 2 ωε 2
1
When x3 > 0 and |k1 | < ω(εµ) 2 we observe that the electromagnetic
field consists of a superposition of uniform plane waves propagating in the
1
{k1 , 0, k3 }-direction, while when |k1 | > ω(εµ) 2 the constituents are (non-
uniform) evanescent waves decaying exponentially in the x3 -direction.
The presence of the source distribution is accounted for by the application
of the excitation conditions
!
Iˆ∆ (x1 , jω) , |x1 | < 12 a ,
lim Ĥ2 (x1 , x3 , jω) − lim Ĥ2 (x1 , x3 , jω) =
x3 ↓0 x3 ↑0 0, |x1 | > 12 a ,
(8.12)
lim Ê1 (x1 , x3 , jω) − lim Ê1 (x1 , x3 , jω) = 0 , −∞ < x1 < ∞ , (8.13)
x3 ↓0 x3 ↑0
which is a consequence of Eq. (8.5). From Eqs. (8.13) and (8.6) it directly
follows that
ê− +
1 = ê1 , (8.14)
and from Eqs. (8.10) and (8.11) it follows that
ĥ−
2 = −ĥ2 .
+
(8.15)
We observe that the field components Ĥ2 and Ê3 are odd-symmetric func-
tions of x3 , while Ê1 is an even-symmetric function of x3 . Application of
Eqs. (8.12) and (8.15) to Eq. (8.8) leads to
∞ ! 1
ˆ |x1 | < 12 a ,
1 2 I∆ (x1 , jω) ,
ĥ+
2 (k1 , jω) exp(−jk1 x1 )dk1 =
2π k1 =−∞ |x1 | > 12 a .
0,
(8.16)
The left-hand side of this equation is a Fourier integral with respect to the
spatial variable x1 and with the transform parameter k1 . Fourier’s inversion
theorem yields
1
a
2
1ˆ
ĥ+
2 (k1 , jω) = 2 I∆ (x1 , jω) exp(jk1 x1 ) dx1 . (8.17)
x1 =− 21 a
With Eqs. (8.17), (8.15), (8.10) - (8.11) and (8.6) - (8.8) the electromagnetic
field is determined completely. Once Iˆ∆ is prescribed, we can calculate the
integral of Eq. (8.17) either analytically or numerically, and once ĥ+ 2 is de-
termined we can calculate the Fourier integral of Eq. (8.8) either analytically
or numerically. The other field components can be computed in a similar
way.
8.1.1. The far-field approximation
When the point of observation is far enough away from the emitter, i.e.,
1
r = (x21 + x23 ) 2 (8.18)
is large enough, we can approximate the representations of the field compo-

nents. Let us start with the expression for the magnetic field component
∞
1
Ĥ2 = 2 (k1 , jω) exp(−jk1 x1 − jk3 x3 )dk1 ,
ĥ+ (8.19)
2π k1 =−∞
when x3 > 0. In the case that r is large, the exponential function in the
integrand oscillates violently along the integration path, except in a point
where k1 x1 + k3 x3 is stationary. In such a stationary point the exponential
function does not vary rapidly in a small interval along the integration path.
In each part outside the stationary points the contribution to the integral
is negligibly small. Around the stationary point the contribution to the
integral is more significant. This fact is employed in Kelvin’s principle of
stationary phase. The stationary point k1 = k1s is the root of the equation
1

∂k1 [k1 x1 + k3 x3 ] = ∂k1 k1 x1 + (k 2 − k12 ) 2 x3 = 0 , (8.20)
where
1
k = ω(εµ) 2 (8.21)
has been used. Hence, this point k1 = k1s follows from
k1
x1 − 1 x3 = 0 (8.22)
(k 2 − k12 ) 2
as
x1 x1
k1s = k 1 =k . (8.23)
(x21 + x23 ) 2 r
Note that
x3 1
k3s = [k 2 − (k1s )2 ] 2 = k
. (8.24)
r
Only around this stationary point the integrand of Eq. (8.19) will contribute
significantly. Assuming that the function ĥ+ 2 varies much slower than the
exponential function, we obtain the approximation
∞
1 + x1
Ĥ2 ≈ ĥ (k , jω) exp(−jk1 x1 − jk3 x3 )dk1 . (8.25)
2π 2 r k1 =−∞
The argument of the exponential function has to be approximated with more

care. We expand the argument of the exponential function around the sta-
tionary point in a Taylor series

k1
k1 x1 + k3 x3 = k1s x1 + k3s x3 + 12 (k1 −k1s )2 ∂k x1 − 1 x3 +··· ,
1
(k 2 −k12 ) 2 k1 = k1s
(8.26)
where we omit higher order terms. Note that the term being linear in k1 −k1s
has vanished in the stationary point. Some algebraic manipulation yields
r3
k1 x1 + k3 x3 = kr − 12 (k1 −k1s )2 + ··· . (8.27)
kx23
Hence,
∞
exp(−jk1 x1 − jk3 x3 )dk1
k1 =−∞
∞
3
s 2 r
≈ exp(−jkr) exp 1
2 j(k 1 −k 1 ) dk1 (8.28)
k1 =−∞ kx23

1 ∞
2k 2 x3
≈ exp(−jkr) exp(jκ2 )dκ .
r r κ=−∞
The last integral can be calculated as follows

∞ 2 ∞ ∞
exp(jκ2 )dκ = exp(jx2 )dx exp(jy 2 )dy
κ=−∞ −∞ −∞

= 2
exp[j(x2 + y 2 )]dxdy
(x,y)∈IR
∞
= 2π lim exp[(j −δ)ρ2 ]ρdρ
δ↓0 ρ=0
−π
= lim = jπ . (8.29)
δ↓0 j −δ
Combining Eqs. (8.25), (8.28) and (8.29) we arrive at

1
x3 + x1 k 2
Ĥ2 ≈ ĥ2 (k , jω) exp(−jkr + j 14 π)
r r 2πr
1
when r = (x21 +x23 ) 2 → ∞ and x3 > 0 . (8.30)
This is the far-field representation of the magnetic field component. The

amplitude is directly related to the spatial Fourier transform of the electric
current through the emitter, by taking in Eq. (8.17) the quantity k1 = kx1 /r
as the transform parameter.
The electric-field components in the far field are obtained from Eqs. (8.10)
and (8.23) - (8.24) as

1
x3 + x1 k 2
Ê1 ≈ ê (k , jω) exp(−jkr + j 14 π)
r 1 r 2πr

2
1
1
x3 µ 2 x1 k 2
≈ ĥ+
2 (k , jω) exp(−jkr + j 14 π)
r ε r 2πr
1
when r = (x21 +x23 ) 2 → ∞ and x3 > 0 , (8.31)
and

1
x3 + x1 k 2
Ê3 ≈ ê3 (k , jω) exp(−jkr + j 14 π)
r r 2πr

1
1
x1 x3 µ 2 x1 k 2
≈ − 2 ĥ+
2 (k , jω) exp(−jkr + j 14 π)
r ε r 2πr
1
when r = (x21 +x23 ) 2 → ∞ and x3 > 0 . (8.32)
We observe that in the far field the relation
x1 Ê1 + x3 Ê3 = 0 (8.33)
holds, hence the electric field vector Ê = {Ê1 , 0, Ê3 } is perpendicular to the
direction of observation x = {x1 , 0, x3 }, see Fig. 8.2.
All the results pertaining to the electromagnetic field quantities in the
domain x3 < 0 can be derived in a similar way.
The time-averaged power flow density in the far field is given by
∗ ∗
1
2 Re[Ê × Ĥ ] = 1
2 Re[−Ê3 Ĥ2 i1 + Ê1 Ĥ2∗ i3 ]

2
1
x3 µ 2 k x1 x1 x3
≈ |ĥ+
2 (k , jω)|2 [ i1 + i3 ] ,
r ε 4πr r r r

2
1
x3 µ 2 k x1 x
≈ |ĥ+ (k , jω)|2 ,
r ε 4πr 2 r r
1
when r = (x21 +x23 ) 2 → ∞ , (8.34)
E
. ......
........
...
...
...
...
.............
........
S
(...................
...
.
.........
..........
.......
.......
H
.
......
..
... x
.......
.......
.......
.......
....... ......
. .θ ...... ....
.
.
.......
.......
..
...
................. .......
.. .. ............
..............
i1
.... .......
.......
.......... .......
.......
... ............
O r
............. .....
i3
......................
Figure 8.2. The orientation of the electromagnetic field vectors in the far field of
an emitter with parallel electric current.
where x/r = x/|x| is the unit vector in the direction x of observation. The
electromagnetic power flow is in this direction and decays with 1/r as a
function of r. In conclusion we observe that the electric and magnetic field
strengths have, in the far-field region, the structure of a cylindrical wave that
expands radially from the origin of the coordinate system. In the angular
direction (see Fig. 8.2), the field amplitudes depend on θ via
x1 x3
cos(θ) = , sin(θ) = . (8.35)
r r
The electric and magnetic field strengths in the far-field region are transverse
to the local direction of propagation and behave locally as if the wave were a
uniform plane wave, with exp(−jkr) = exp(−jk1 x1 −jk3 x3 ), traveling in the
direction {k1 = k cos(θ), 0, k3 = k sin(θ)} away from the source. The time-
averaged power flow vanishes in the i1 -direction (θ = 0). All field amplitudes
are proportional to ĥ+ 2 (k cos(θ), jω). Consequently, in the far-field region
only the spatial Fourier transform of the electric current through the emitter
(see Eq. (8.17)) at the subset of the transform parameter k1 = k cos(θ) is
”seen”.
To study the angular dependence of the far-field characteristic, the di-
rective gain is introduced. This directive gain D(θ) is defined as the power
flow in the observation direction, normalized to the angular-averaged power
flow in the far field, viz.,

∗ x
1
2 Re (Ê × Ĥ ) ·
r
D(θ) = 2π . (8.36)
1 ∗ x
1
2 Re (Ê × Ĥ ) · dθ
2π θ=0 r
Using Eqs. (8.34) and (8.35) we find
| sin(θ) ĥ+
2 (k cos(θ), jω)|
2
D(θ) = .
(8.37)
1 2π
| sin(θ) ĥ+ 2
2 (k cos(θ), jω)| dθ
2π θ=0
In Fig. 8.3, the directive gain is presented for the case that Iˆ∆ is a constant
function, i.e.,
Iˆ∆ (x1 , jω) = î∆ (jω) , |x1 | < 12 a , (8.38)
D(θ) D(θ)
...... ......
......... .........
a/λ = 0.1 ... .... ..
...
a/λ = 0.3 ... .... ..
...
.... ....
. . . .10 ..
.... . . . . . . . . 10 ..
.... . . . .
. . ... . . . . ... . .
. . ... . . . . ... . .
. . .. . . ..... . . .. . . .....
. .
. . ... . . . ...
. . .
.
.
. . ........... . .
.
.
.
. . . ...........
. . 1 .
. . . . . . 1 . .
.
. .
.. . . . . ... . . . . .
. .θ . .
. .
. . . . ... . . . . .
.
. .
.
θ
. . ..... ..
. .. . . .
.. ..... ..
. . . .
. . . .. . . . . .. .
. . . ......................... ... ............................ ..................... ... .....................
.......... . ....... . ..
...... . ...... . . . . . .
.
.......... . .... . ..........
. .....
. . . .
. . ...... .. .... .. .... . . . .... . . . . .
....... . ..... .. .... . . . .... . .
. . .. .. .
.
.
. .
0.1
.... .. ... .
.
.
. .
.
. .
.
.
.
. ....
.
.
. .
.... .. .
.
.
.
.
.
.
.... .. .... .
.
0.1 .
. .
.
.
.
.
. .
..
. ....
.
.
.
. . . .. . . . ..
. ... . . . . . .
. . .... .... .... . .
. . .
. ... . . .
... . .
. . .... .... .... . .
. . .... .
. .. .
.
. .
. . . . . . .
. .. .. . . .. .
.
. .
. . . . . . .
. .. . .
.. . . .. .......
.
..
. .. . . .. ......
. .
.. .
. . . . . ...... .. . . . . ... . . .. . . . ...... . . . . . . ... .
.
. . .
. .
. . .. . . . . .
. . .
... . .
. ........... .
. . . ..... . . . .
... . .
. ............ .
. . . ..... .
. .... . . . .
. . . .
. ...
.
. . . .... . . . .
. . . .
. ...
... .
. ... . . . ....... . . . . ... . . . ........ . . .
. ..
.. . . . . .... . . .
.
. . ...
. . . ... . . . .... . . .
.
. . .
.
.
. .
. . . . .. ... . . . . .. .. . . . . .. ... . . . . ..
. .. . . . . ... . .
. ... ... . . .
. .. . . .. .. . .. . . ...
. .. . . .
. .. . . . . . ... .... . . . . . .... . . . . . .. ... . . . . . . .
.
...
.... . . . .. . .. .. .. . . . .... . . ... . . . .. . .. .. .. . . . ...... .
.
.. . ....... . .. ..... .... . ... .. . . ....
. ..... ..
..... . . ..
. .... ..
. .... . .
. . ..... .
..... . . ...... . . ....... . .......... .
.
. . .
. . ..
...
. ..
. . .. . . . .. . ..
... ..... . .. .
. ...... . . ....... . .
.. . ........................... ..... ..
........................... . .. .. . ..................... ....... ..
.................... . ..
. . . .
. . . . . . . . . .
. .. .
. . .. . . .. .
. . .. .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . .
. . . . . . . .
.. . . .. .. . . ..
. . . . . .
. . . . . . . . . .
. . . . . . . . . . . .
..
D(θ) D(θ) ..
...... ......
....... .......
a/λ = 1 .. .... .. a/λ = 3 .. .... ..
.... ....
.... ....
10
. . . ...... . . . .
10
. . . ...... . . . .
. .
. . . ....
. .
. . . ....
. .
. . . . . .
. .... . . ..... . .... . . .....
. . .. . .... . . .. . ....
.
. . ... . . .......... .
. . ... . . ..........
.
.
. .
1
. . ... . . .
.
.
.
. .
. θ .
.
. .
1
. . ... . . .
.
.
.
. .
. θ
.
.
.
.. . .... . .. .
.
.
.
.. . .... . .. .
. . . ... . .. .
. . . ... . . . .
. . ... . .. . . ... . ..
. . . . . .
. .
. . . . . . . . .
. .
. . .
. . .
. . . . . .
. . .
. ................ 0.1 . . . . .. 0.1 . . .
. .... ........ . . . ... . . . .................................. .
.
. . . ... . . .
. . ..
. ..
......... . . . ....... .
. . .. . . .......
. . . ......
.
.
.
..
. . . . . .. .
.
.... . . ...... .
.. ..
. ........
. . . ..... . . ............... . . ..
. . . ............
.
. . . .
. . . . . . ..... .. ..... . . . ... . . ..... . ...... ..... .. ...... ....... . ....
... .. . .. . ... . . . . .. . . .
. ............................................................................................ .... ......................... ................................................................ .
.
. .... . . . . ...... ... ...... . . . . ... . ................... .............. .. .. ........... ....................
. .... . . . . ... ... ... . .
......... . . .
.
.... . ......... . . .......................................................... . .
.....
....
. ... . . . . . ... . . ... . ..
.
..
..
. .. .
.
..
.
..
.
..
...
.. ..
. ........ . . . . ..................... . .. .
. .
.
... . . .. ... . . . .. . ...... . . ................................... .................................. . ..
. . .
. ..
. ...
... .
. .
. . . ..... ...... . . .
. .
.. . . ............ . ... . ....... ................................................... .................................................................. .
.
....
.......
. ... . . . .
..... ....
. . . . . .
.... . . .
..
.
............................... ...... .. ...... .. ....
.. . . . . . . . .
.. . ... ..
.
....... ....... . ....... ................................. .
. . .. ... . ...
. ...
..... . .
. . ....
.... .
..
...........
.
. . . ... . . . ..................... . .. . . .................. . .
....... . . . ...... .. .. .... . ........ . . . . .. . . . .. .
. .... ....... . . . . . . . ..................................... .
. . . . . .
.
. . .................... . . . . .
. . . . . . . . .
. . . . . . . . . . . . . . . . . .
.. . . . . . . .. . . . . . .
.
.
. . . . .. .
.
. . . . ..
. .. . . .. . . .. . . .. .
. . . . . . . . . . . . . . .
. . . . .
. . . . . .
. . . . . .
. . . . . . . .
. . . .
. . . .
. . . . .
.
.. . . .. . .
. . . . . . . .
. . . . . . . . . .
. . . . . . . . . . . .
Figure 8.3. The directive gain of the sheet emitter with parallel electric current.
sheet emitter with a perpendicular electric current 213
so that Eq. (8.17) yields
sin( 12 k1 a)
ĥ+
2 = î∆ . (8.39)
k1
The directive gain is presented for various values of a/λ, where λ = 2π/k
is the wavelength. For values of a/λ → 0, we observe that ĥ+ 2 ≈ 2 a î∆
1
and D(θ) ≈ 2 [sin(θ)] . From Fig. 8.3, we observe that this approximation
2
holds up to a/λ = 0.3. For values of a/λ → ∞, we observe that ĥ+ 2 ≈

π î∆ δ[k cos(θ)] and D(θ) vanishes everywhere, except at θ = ± 12 π, where
it becomes infinite. Hence, for large values of a/λ the emitter radiates the
electromagnetic wavefield mainly in forward and backward directions.
8.2. The sheet emitter with a perpendicular

electric current
plane x3 = 0 (Fig. 8.4). Let the impressed electric current flow uniformly in
the x2 -direction. It is present at x3 = 0 from x1 = − 12 a to x1 = 12 a. Hence,
!
Iˆ∆ (x1 , s)δ(x3 ) , |x1 | < 12 a ,
Jˆ1ext = 0 , Jˆ2ext = Jˆ3ext = 0 , (8.40)
0, |x1 | > 12 a ,

and Iˆ∆ (in A/m) is the electric current per unit length (of the x1 -direction).
K̂1ext = 0 , K̂2ext = 0 , K̂3ext = 0 . (8.41)
The medium is assumed to be homogeneous and lossless with constitutive

constants ε and µ. Since the configuration is independent of x2 , the elec-
tromagnetic field is two-dimensional and in Chapter 4 we have seen that
the frequency-domain field equations (2.54) - (2.59) separate into two in-
dependent set of equations, cf. Eqs. (4.4) - (4.6) and Eqs. (4.7) - (4.9).
From Eqs. (8.40) - (8.41) we conclude that Jˆ1ext , Jˆ3ext and K̂2ext are equal
to zero, hence only a perpendicularly polarized electromagnetic field occurs.
The non-zero field components Ĥ1 , Ĥ3 and Ê2 satisfy the following set of
equations (s = jω, σ = 0, cf. Eqs. (4.7) - (4.9))
−∂3 Ê2 + jωµĤ1 = 0 , (8.42)

∂1 Ê2 + jωµĤ3 = 0 , (8.43)
!
−Iˆ∆ (x1 , jω)δ(x3 ) , |x1 | < 12 a .
−∂3 Ĥ1 + ∂1 Ĥ3 + jωεÊ2 = (8.44)
0, |x1 | > 12 a ,
In Chapter 4 we have seen that in a domain outside the source distribution,

solutions of Eqs. (8.42) - (8.44) in the form of plane waves exist. We therefore
assume that the general solution consists of an infinite superposition of plane-
wave constituents
⎧
⎪
⎪ 1 ∞
⎪
⎨ ĥ+ (k1 , jω) exp(−jk1 x1 − jk3 x3 )dk1 , x3 > 0,
2π k1 =−∞ 1
Ĥ1 (x1 , x3 , jω) =
⎪
⎪ 1 ∞
⎪
⎩ ĥ− (k1 , jω) exp(−jk1 x1 + jk3 x3 )dk1 , x3 < 0,
2π k1 =−∞ 1
(8.45)
ε, µ ..........
......... ..............
........
ε, µ
' '...................
.......
.......
.......
(x = 1 a
E ( 1 2 E
( .
........ i1
....
( ....
...r...........................
O( i3
(
(
' J ext x1 = − 12 a '
.......
....... .......
....... .......
....... ..........
..
.............
.
. E E ..........
.
x3 < 0 x3 > 0

perpendicularly polarized electromagnetic waves.
sheet emitter with a perpendicular electric current 215
⎧
⎪
⎪ 1 ∞
⎪
⎨ ĥ+ (k1 , jω) exp(−jk1 x1 − jk3 x3 )dk1 , x3 > 0 ,
2π k1 =−∞ 3
Ĥ3 (x1 , x3 , jω) = ∞
⎪
⎪ 1
⎪
⎩ ĥ− (k1 , jω) exp(−jk1 x1 + jk3 x3 )dk1 , x3 < 0 ,
2π k1 =−∞ 3
(8.46)
and
⎧
⎪
⎪ 1 ∞
⎪
⎨ ê+ (k1 , jω) exp(−jk1 x1 − jk3 x3 )dk1 , x3 > 0 ,
2π k1 =−∞ 2
Ê2 (x1 , x3 , jω) = ∞
⎪
⎪ 1
⎪
2π k1 =−∞ 2
(8.47)
where ⎧
⎨
1 1
(ω 2 εµ − k12 ) 2 , |k1 | ≤ ω(εµ) 2 ,
k3 = (8.48)
⎩ −j(k 2 − ω 2 εµ) 21 , |k | > ω(εµ) 21 .
1 1
Note that the definition of the square root of k3 takes care of the bounded-
ness of the integrals as |x3 | → ∞. This guarantees the boundedness of the
electromagnetic field components as |x3 | → ∞. From Eqs. (8.42) and (8.43)
− − −
it follows that the quantities ĥ+ + +
1 , ĥ3 , ĥ1 and ĥ3 are related to ê2 and ê2 as
−k3 + k1 +
ĥ+
1 = ê , ĥ+
3 = ê , (8.49)
ωµ 2 ωµ 2
k3 − k1 −
ĥ−
1 = ê , ĥ−
3 = ê . (8.50)
ωµ 2 ωµ 2
1
When x3 > 0 and |k1 | < ω(εµ) 2 we observe that the electromagnetic
field consists of a superposition of uniform plane waves propagating in the
1
{k1 , 0, k3 }-direction, while when |k1 | > ω(εµ) 2 the constituents are (non-
uniform) evanescent waves decaying exponentially in the x3 -direction.
The presence of the source distribution is accounted for by the application
of the excitation conditions
!
Iˆ∆ (x1 , jω) , |x1 | < 12 a ,
lim Ĥ1 (x1 , x3 , jω) − lim Ĥ1 (x1 , x3 , jω) =
x3 ↓0 x3 ↑0 0, |x1 | > 12 a ,
(8.51)
lim Ê2 (x1 , x3 , jω) − lim Ê2 (x1 , x3 , jω) = 0 , −∞ < x1 < ∞ , (8.52)
x3 ↓0 x3 ↑0
which is a consequence of Eq. (8.42). From Eqs. (8.52) and (8.47) it directly
follows that
ê− +
2 = ê2 , (8.53)
and from Eqs. (8.49) and (8.50) it follows that
ĥ−
1 = −ĥ1 .
+
(8.54)
We observe that the field components Ĥ3 and Ê2 are even-symmetric func-
tions of x3 , while Ĥ1 is an odd-symmetric function of x3 . Application of
Eqs. (8.51) and (8.54) to Eq. (8.45) leads to
∞ ! 1
1 ˆ
2 I∆ (x1 , jω) , |x1 | < 12 a ,
ĥ+
1 (k1 , jω) exp(−jk1 x1 )dk1 =
2π k1 =−∞ |x1 | > 12 a .
0,
(8.55)
The left-hand side of this equation is a Fourier integral with respect to the
spatial variable x1 and with the transform parameter k1 . Fourier’s inversion
theorem yields
1
a
2
1ˆ
ĥ+
1 (k1 , jω) = 2 I∆ (x1 , jω) exp(jk1 x1 ) dx1 . (8.56)
x1 =− 21 a
With Eqs. (8.56), (8.54), (8.49) - (8.50) and (8.45) - (8.47) the electromag-
netic field is determined completely. Once Iˆ∆ is prescribed, we can calculate
the integral of Eq. (8.56) either analytically or numerically, and once ĥ+
1 is
determined we can calculate the Fourier integral of Eq. (8.45) either ana-
lytically or numerically. The other field components can be computed in a
similar way.
8.2.1. The far-field approximation
When the point of observation is far enough away from the emitter, i.e.,
1
r = (x21 + x23 ) 2 (8.57)
the sheet emitter with a perpendicular electric current 217
is large enough, we can approximate the representations of the field compo-

nents. Using the method of stationary phase of Section 8.1.1, we obtain the
far-field representation

1
x3 + x1 k 2
Ĥ1 ≈ ĥ1 (k , jω) exp(−jkr + j 14 π)
r r 2πr
1
when r = (x21 +x23 ) 2 → ∞ and x3 > 0 . (8.58)
The amplitude is directly related to the spatial Fourier transform of the

electric current through the emitter, by taking in Eq. (8.56) the quantity
k1 = kx1 /r as the transform parameter.
The other field components in the far field are obtained from Eqs. (8.49)
and (8.23) - (8.24) as

1
x3 + x1 k 2
Ĥ3 ≈ ĥ3 (k , jω) exp(−jkr + j 14 π)
r r 2πr

1
x1 x1 k 2
≈ − ĥ+
1 (k , jω) exp(−jkr + j 14 π)
r r 2πr
1
when r = (x21 +x23 ) 2 → ∞ and x3 > 0 , (8.59)
and

1
x3 + x1 k 2
Ê2 ≈ ê2 (k , jω) exp(−jkr + j 14 π)
r r 2πr

1
1
µ 2 x1 k 2
≈ − ĥ+
1 (k , jω) exp(−jkr + j 14 π)
ε r 2πr
1
when r = (x21 +x23 ) 2 → ∞ and x3 > 0 . (8.60)
We observe that in the far field the relation
x1 Ĥ1 + x3 Ĥ3 = 0 (8.61)
holds, hence the magnetic field vector Ĥ = {Ĥ1 , 0, Ĥ3 } is perpendicular to

the direction of observation x = {x1 , 0, x3 }, see Fig. 8.5.
All the results pertaining to the electromagnetic field quantities in the
domain x3 < 0 can be derived in a similar way.
H
. ......
........
...
...
...
...
...............
........
S
'...................
...
.
...............
........
.......
E
.
......
..
.........
. x
.......
.......
.......
.......
....... ......
θ . .
...... ....
.
.
....
.......
..
...
................. .......
.. .. ............
..............
i1
.... .......
....
.......... .......
.......
......
.... ............
O r ..................................
.. i3
Figure 8.5. The orientation of the electromagnetic field vectors in the far field of
an emitter with perpendicular electric current.
The time-averaged power flow density in the far field is given by

∗ ∗
1
2 Re[Ê × Ĥ ] = 1
2 Re[Ê2 Ĥ3 i1 − Ê2 Ĥ1∗ i3 ]

1
µ 2 k x1 x1 x3
≈ |ĥ+
1 (k , jω)|2 [ i1 + i3 ] ,
ε 4πr r r r

1
µ 2 k x1 x
≈ |ĥ+
1 (k , jω)|2 ,
ε 4πr r r
1
when r = (x21 +x23 ) 2 → ∞ , (8.62)
where x/r = x/|x| is the unit vector in the direction x of observation. We

observe that the power flows in this direction and decays with 1/r as a
function of r. In conclusion we observe that the electric and magnetic field
strengths have, in the far-field region, the structure of a cylindrical wave that
expands radially from the origin of the coordinate system. In the angular
direction (see Fig. 8.5), the field amplitudes depend on θ via
x1 x3
cos(θ) = , sin(θ) = . (8.63)
r r
The electric and magnetic field strengths in the far-field region are transverse
to the local direction of propagation and behave locally as if the wave were
the sheet emitter with a perpendicular electric current 219
a uniform plane wave, with exp(−jkr) = exp(−jk1 x1 − jk3 x3 ), traveling

in the direction {k1 = k cos(θ), 0, k3 = k sin(θ)} away from the source. All
field amplitudes are proportional to ĥ+ 1 (k cos(θ), jω). Consequently, in the
far-field region only the spatial Fourier transform of the electric current
through the emitter (see Eq. (8.56)) at the subset of the transform parameter
k1 = k cos(θ) is ”seen”.
To study the angular dependence of the far-field characteristic, the di-
rective gain is introduced. This directive gain D(θ) is defined as the power
flow in the observation direction, normalized to the angular-averaged power
flow in the far field, viz.,

∗ x
1
2 Re (Ê × Ĥ ) ·
r
D(θ) = 2π . (8.64)
1 x∗
1
2 Re (Ê × Ĥ ) · dθ
2π θ=0 r
Using Eqs. (8.62) and (8.63) we find
|ĥ+
1 (k cos(θ), jω)|
2
D(θ) = .
(8.65)
1 2π
|ĥ+ 2
1 (k cos(θ), jω)| dθ
2π θ=0
In Fig. 8.6, the directive gain is presented for the case that Iˆ∆ is a constant
function, i.e.,
Iˆ∆ (x1 , jω) = î∆ (jω) , |x1 | < 12 a , (8.66)
so that Eq. (8.56) yields
sin( 12 k1 a)
ĥ+
1 = î∆ . (8.67)
k1
The directive gain is presented for various values of a/λ, where λ = 2π/k
is the wavelength. For values of a/λ → 0, we observe that ĥ+ 1 ≈ 2 a î∆ and
1
D(θ) ≈ 1. The emitter is then an isotropic radiator in the (x1 , x3 )-plane.

From Fig. 8.6, we observe that this approximation holds up to a/λ = 0.3.
For values of a/λ → ∞, we observe that ĥ+ 1 ≈ π î∆ δ[k cos(θ)] and D(θ)
vanishes everywhere, except at θ = ± 2 π, where it becomes infinite. Hence,
1
for large values of a/λ the emitter radiates the electromagnetic wavefield
mainly in forward and backward directions.
...
D(θ) ...
D(θ)
....... .......
....... .......
a/λ = 0.1 .. ... ..
.. a/λ = 0.3 .. ... ..
..
.... ....
.. ..
. . . . ..... . . . .
10
.
. . . . ..... . . . .
10 .
. . . ....
. .
. . . ....
. .
. . . . . .
. ... . . ..... . ... . . .....
. .. ... . ... . .. ... . ...
. ... . . ........... . ... . . ...........
. . 1 ..
. . . . . 1 .
. . .
.
. .
. ........
...... .
........... . .
....
...
....
...
...
...
..........
........
. .
.
θ .
. . .
. . ..............
. .. ...
... .
.
.. .. ...............
......... .
.
. ..
..........
. .
.
θ
. ...... . ........ . ........ ..
...........
..... . ... . ..... . ... .
. ..... . ... . .....
..
. ..... . ... . ..... ..
. . . ..
..... . .
. .
....
.
. . . . . . ..
..... . .
. .
..... . .
. . . .... . . .. . .... .
. . . ...... . 0.1 .
.
. . ... . . . . . ....... . 0.1 .
.
. . ..... . .
.
.
.
.. . . . . . .... . . . . ..... . .
.
.
... . . . . . .... . . . . ...... .
. .. . . .
. .
. . . ... . . .
.. . . .
. .
. . . .... .
.
. . . . .
. . .. .
. . . . .
. .
. .. . .. . .. . .. . .. . . ... . .. . .. . .. . ..... .
. .. . . . ... . . . .
. .. . .. . . . ... . . . .
.
....
... .. . .... ..
.
. ... . . . ... . . . . . . .... . . . ... . . . . .. .
. ...... . .. . ...... .
. ... . . .. . . . . .... . . .. . .... .
. . . . . . .
. . . .. . . . . . . .
.. . . . . . . .. ... . . . . . . . ..
. . . . . . ... . .
. ..
.. . . . . ..
..
.
. ..
...
. . . . . . . . ..... .
. .. .. . . . . .
. . .. . . . . . . .
. ..
... . . . . . .
. . .. . . .... . . . . . .
. . ... .
... . . .. . . . . .... . .... . . .. . . . . ..... .
. . . . . . ...
. . . . . . . ...
. . ....... . . .
... . . . . . ...... . . .
.... . . .
. . .... . . .. . . . . ..... . . .. .
. .... .... . . ..... .... .
.. . ..... . . .... . .. .. . ..... . . ..... . ..
. ..... .
....... . ......... . . ...... . . ......... .
. ....... ...
........ . .
.........
........... ...
.............. .
. . ............
................................
. . . . . . ...................................... . . .
. . . . . . . . .
. . . . . . . .
. . . .
. . . . . . . .
.. . .. .. . ..
. . . . . . . .
. . . . . . . . . .
. . . . . . . . . . . .
..
D(θ) D(θ)
..
...... ......
....... .......
a/λ = 1 ... .... ..
.. a/λ = 3 ... .... ..
..
.... ....
.. ..
10 . 10 .
. . . . ..... . . . . .
.
. . . . ..... . . . . . .
. . . .... . .
. . . .... . .
. ... . .
. ..... . ... .
. . .....
. .. ... . .... . .. ... . ....
. .
. . . .......... . .
. . . ..........
. . 1 ..
. . . . . 1 .. . .
.
. .
.. .
. . . ... . . .
.... . ..
. . θ
. .
. .
.. .
. . . ... . . .
.
.... . ..
. .
.
θ
. . .. . .. . . . .. . . . .
. . . . . .
. . . . . ... . . .. . . . . . ... . . ..
. . .
.
. . . . . . . .
. . . . .
. . .
.................................... . 0.1 .
.
. . ..
................................ . . . . . .
. . 0.1 ..... . . . . .
. ......... ...... . . ... . . ............ . .. ....... . . .. .
................ . .... . ............... . .. .
. ...... . . . .... ... .
. .. . .. . . .. .
...... . . . . .... .....
..
.. .
. . .....
. .. ...... .
. .
. . ...
.
. .... . ............ .... . ..... .. .... . .... ............ . .... . .
. .. . .. .... .. ... . . . ... . . . ..... . ........ .... . .. .. .. . .. ........ . ..... . . .
. . . . .... .
. ..
.. ..
. .... .
. . .. . . . . .
. .
.
. ............... . ................................... ................ ... .. .. . ............. . ................................. . ............... .... .......
......
.
...
. . . . . ..... ... .... . . . . ... . ................... .......... ..... ... .......... ....................
. .... . . . . ........... . . . . ... . ........ . . ............................ ....... .......... . .
.....
...
.............. . ..... ........................................ ..
. .... . . .... . . .
.
.
. . . . . .
................... ... . .. ..
. ...
... . . . .
. .... ..... . . ..
.
. . ..... . . ............................................................................. . ...
...
. ... . . . . .... .... . . .
.
... . . ............ . .......................................... ........................ .... ..................... ............................................ . ..
...... .
. ... . . . .... . . ...
. . . ... .
........................ . .
..... ................. . ... .... . ............... .......... ...................... . . . ..
...
. .
... . . . ... ..
.. .
.... . . . . ... .. ......
.. . . ... . .... ....... . .... .
. ... .
.... . . . .... .... . . ..
. . . .... ............ ...... ....... .. .. .............. . .
. ...... ..... . . . ....... . . . . .... ............ . . . ................ .
..... . . ..... . . . . . ............. . .
..
. ......................................... . ................................. . . . . .. . . . . .
. . . . . . . . . . . . . . . . .
. . . .
.. . .
. . . . .. .. . .
. . . . ..
. . . . . . . . . . . .
. .. .
. . .. . . .. .
. . .. .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . .
. . . . . . . .
. . . .
. . . . . . . .
.. . .. .. . ..
. . . . . . . . . .
. . . . . . . .
. . . . . . . . . . . .
Figure 8.6. The directive gain of the sheet emitter with perpendicular electric
current.
excitation of two-dimensional electromagnetic waves 221
Exercise 8.1
In the stationary phase method the argument of the exponential function is
expanded in a Taylor series around the stationary point k1 = k1s . Why does
the term linear in (k1 − k1s ) vanish?
Exercise 8.2
In the case of a constant impressed current Iˆ∆ (x1 , jω) = ı̂∆ (jω), |x1 | ≤ 12 a,
for what angles θ is the directive gain function D(θ) equal to zero?
Exercise 8.3
Assume that the impressed current flows uniformly in the x2 -direction and is
ext
present at x1 = 0, x3 = 0. Hence, Ĵ = ı̂∆ (jω)δ(x1 , x3 )i2 , where δ(x1 , x3 )
is the two-dimensional unit impulse operative at x1 = 0, x3 = 0. This implies
that the electric current sheet is degenerated to a line source.
(a) Give the expression for ĥ+1.
(b) Give in the far-field approximation the expressions for the electric and
magnetic field strengths in the frequency domain.
( c ) Draw the far-field radiation characteristic for the electromagnetic field
6 , n ∈ IN . Indicate
conform Fig. 8.5, for the angles of observation θ = nπ
the magnitude of the field strengths by the lengths of the vectors for
Ĥ and the radii of the circles for Ê.
(d) Calculate and draw the directive gain as a function of the angle of
observation θ.
Problem 8.1
Let for a current sheet the impressed perpendicular electric current be given
by Iˆ∆ (jω) = a−1 ı̂∆ (jω), for |x1 | ≤ 12 a. Show that in case we let a → 0
the same value for ĥ+1 is obtained as in Exercise 8.3(a). This is most easily
sin(k1 a/2)
obtained by taking lim ĥ+ +
1 with ĥ1 = k1 a ı̂(jω).
a→0
Problem 8.2
Assume the impressed current flows uniformly in the x1 -direction; it is
present at x3 = 0 for |x1 | ≤ 12 a and is given by Ĵ = −(1− 2|xa1 | )ı̂∆ (jω)δ(x3 )i1 .
ext
(a) Give the expression for ĥ+

2.
(b) Give in the far-field approximation the expressions for the electric and
magnetic field strengths.
( c ) Draw the far-field radiation characteristic for the electromagnetic field

6 , n ∈ IN . Indicate
conform Fig. 8.5, for the angles of observation θ = nπ
the magnitude of the field strengths by the lengths √ of the vectors for
Ê and the radii of the circles for Ĥ. Take a = 4 2λ.
Answers to Exercises
Exercise 1.1
(a) 0
(b) 0
( c ) (a · c)b − (b · c)a
(d) Replace one of the vector products by a new vector e Rewrite the
expression Substitute the original vector product for e Again
rewrite the expression to finally obtain (a · c)(b · d) − (a · d)(b · c)
( e ) Use the same procedure as in (d) to arrive at the result

[d · (a × b)]c − [c · (a × b)]d.
Exercise 1.2
Since c · a = (a × b) · a = 0 and c · b = (a × b) · b = 0, the vector c is normal
to both vectors a and b.
Exercise 1.3
A = d1 d2 is the area of the rectangle.
Exercise 1.4
V = d1 d2 d3 is the volume of the brick.
224 answers to exercises
Exercise 1.5
(a) x2 (b) x2
x1 x1
x3 x3
(c) x1 (d) x1
x3 x3
x2 x2
The vector fields are invariant in the direction of the coordinate axis that
points forward.
Exercise 1.6
(a) ∇ · v = −1 , ∇×v =0
(b) ∇ · v = 0 , ∇ × v = −i3
( c ) ∇ · v = −2 , ∇×v =0
(d) ∇ · v = 0 , ∇ × v = −2i2 .
Exercise 1.7
∇2 A = (∂12 + ∂22 + ∂32 )A1 i1 + (∂12 + ∂22 + ∂32 )A2 i2 + (∂12 + ∂22 + ∂32 )A3 i3 .
Exercise 1.8
∇ × E = −∇ × ∇V = 0, for all V (x) (see Eq. (1.32)).
Exercise 1.9
∇ · D = ∇ · (∇ × C) = 0, for all C(x) (see Eq. (1.33)).
answers to exercises 225
Exercise 1.10
Application of the chain rule of differentiation gives
x − x
(a) ∇d = − , ∇d = −∇d
|x − x|
x − x
(b) ∇d−1 = , ∇ d−1 = −∇d−1
|x − x|3
2
( c ) ∇ · ∇d = , ∇ · ∇ d = ∇ · ∇d
|x − x|
(d) ∇ · ∇d−1 = 0 , ∇ · ∇ d−1 = 0.
Exercise 1.11
Use Eq. (1.11) to obtain w × (ν × H) = (w · H)ν − (w · ν)H Substitute
w = ν × E (ν × E) × (ν × H) = [(ν × E) · H] · ν, since (ν × E) · ν = 0
Take the scalar product of the result with ν ν · [(ν × E) × ν × H)]
= (ν × E) · H = ν · (E × H)
Exercise 1.12
(ν · H)ν − ν × (ν × H) = (ν · H)ν − (ν · H)ν + (ν · ν)H = H
The term (ν · H)ν is the part of H that is normal to S; the term
−ν × (ν × H) = (ν × H) × ν is the part of H that is tangential to S.
Exercise 2.1
There are five unknown vectorial quantities in Maxwell’s equations in matter:
E, H, J , D and B, and hence fifteen unknown scalar quantities.
Exercise 2.2

Application of Gauss’ law yields x∈∂ D ext ν · J dA = −∂t x∈Dext ρ dV .
This equation states that the amount of electric current that is flowing out
of the domain Dext through its boundary ∂Dext equals the decrease per unit
time of the electric charge present in Dext .
Exercise 2.3
Following the reasoning of Section 2.6 the continuity of those components
that are differentiated in the i3 -direction must be enforced. Thus we must
enforce the continuity of H1 , H2 , E1 and E2 .
Exercise 2.4
Following the reasoning of Section 2.6 the continuity of those components
that are differentiated in the direction normal to the interface must be en-
forced. Thus ν · J + ∂t ν · D is continuous across S and ν · B is continuous
across S.
Exercise 2.5
(a) circular polarization
(b) circular polarization
( c ) elliptical polarization
(d) elliptical polarization
( e ) linear polarization.
Exercise 2.6
(a) J/m3 , J/m3
(b) W/m3 , W/m2
( c ) W/m2 , W/m2 .
An easy way to arrive at these results is to employ Poynting’s theorem and
to realize that in physics all terms of an equation have equal dimensions.
Exercise 2.7
∗ 1
ST = 12 Re[Ê × Ĥ ] = 1
2 Re[Ê1 Ĥ2∗ − Ê2 Ĥ1∗ ]i3 = ( µε00 ) 2 i3 .
1
E(t) = cos(ωt)i1 − sin(ωt)i2 , H(t) = ( µε00 ) 2 [sin(ωt)i1 + cos(ωt)i2 ],
1 1
S = E(t) × H(t) = ( µε00 ) 2 [cos2 (ωt) + sin2 (ωt)]i3 = ( µε00 ) 2 i3 .
Exercise 3.1
(a) The polarization of the wave is independent of the value of x3

Take x3 = 0 to simplify the problem Ê = i2 and Ê = −i1
the wave is circularly polarized (cf. Exercise 2.2 (a)).
After inversion the expression for the phase shows that wave propa-
gates in the positive x3 -direction.

Since Ê , Ê and the direction of propagation form a right-handed
triad, the polarization is right-handed (also called counter clockwise).
(b) Use the approach under (a) Circular, right-handed polarization

( c ) Use the approach under (a) Elliptical, right-handed polarization
(d) Use the approach under (a) Elliptical, left-handed polarization
( e ) Use step 1 under (a) Linear polarization No rotation.
Exercise 3.2
f = 1 MHz, µ = µ0 and γ = α + jβ = 0.04 + j0.1.
(a) The wave is attenuated by a factor of exp{−Re[γx3 ]} An attenua-
tion of exp(−π) occurs over a distance x3 = 25π m.
(b) The phase is determined by Im[γx3 ] A phase change of π rad occurs

over a distance x3 = 10π m.
( c ) A 1 MHz wave travels the distance of one wavelength in 1µs The

wavelength and hence the distance is equal to λ = 2π/Im[γ] = 20π m.
(d) |Ê|/|Ĥ| = |Z| and |Z| = | jωµ

γ | = 73.3 Ω.
( e ) Z = |Z| exp(jφ) φ = arctan( αβ ) = 0.381 rad (21.8◦).
Exercise 3.3
The velocity of electromagnetic waves in vacuum is c0 3 × 108 m/s
f = cλ0 The range is 3.75 × 1014 Hz (red light) - 1015 Hz (violet light), or
375 THz (terahertz) - 1 PHz (petahertz).
Exercise 3.4
1 1
2
δ = ( ωµσ ) 2 δCu = 9.3 mm and δAl = δCu ( 5.8
3.5 ) = 12 mm. Only in the
2
first skin depth there is a significant penetration of the electric field that
runs along the surface of the wire Only the first skin depth carries a
significant part of the electric current Using a radius larger than the skin
depth (δAl = 12 mm) would be a waste of (expensive) material since the
inner part will hardly contribute to the power transport.
Exercise 3.5
) 2 = (πf µσ)− 2 The skin depth is proportional to the inverse
1 1
2
δ = ( ωµσ
of the square root of frequency The lower frequency (f = 100 kHz)
determines the minimal thickness of the shield dAl = 5δAl = 1.3 mm and
dCu = 5δCu = 1.0 mm.
Exercise 3.6
I∆ (t) = Re[Iˆ∆ (jω) exp(jωt)] = sin(ωt) Iˆ∆ (jω) = −j A/m E1 (x3 , t) =
1 1
2 Z sin{ω[t − (εµ) x3 ]} and H2 (x3 , t) = 2 sin{ω[t − (εµ) x3 ]}.
1 2
1 2
Exercise 3.7
H(0, t) = 1.5 cos(3π × 108 t)i2 = Re[Ĥ(0, jω) exp(jωt)] Ĥ(0, jω) =
H0 i2 with H0 = 1.5 A/m and ω = 3π × 108 rad/s.
Assume that the field propagates in the positive x3 -direction
γ
Ĥ(x3 , jω) = H0 exp(−γx3 )i2 Ê(x3 , jω) = H0 σ+jω ε exp(−γx3 )i1
α+jβ β
E(x3 , t) = H0 exp(−αx3 ) Re{ σ+jω ε exp[jω(t − ω x3 )]}.
Substitution of the actual parameter values gives
1
(a) γ = jβ = jω(εµ) 2 E(x3 , t) = 188 cos[3π × 108 (t − 10−8 x3 )]i1 V/m
1
(b) ωε = 7.5 × 10−2 σ γ = α + jβ = 0 + jω(εµ) 2 The answer is
the same as in case (a)
1
( c ) ωε = 7.5 × 10−2 σ γ = α + jβ = ( ωµσ 2 ) (1 + j)
2 E(x3 , t) =
8 77 8 77
11.5 exp(−77x3 ){cos[3π×10 (t− 3π×108 x3 )]−sin[3π×10 (t− 3π×10 8 x3 )]}×
i1 V/m.
Exercise 3.8
Assume a unit amplitude of the incident electric wave, and apply the
electric field analysis
i i 1
Ê = exp(−γx3 )i1 Ĥ = ( µε00 ) 2 exp(−γx3 )i2 ,
r r 1
Ê = R exp[γ(x3 − 2L)]i1 Ĥ = −( µε00 ) 2 R exp[γ(x3 − 2L)]i2 .
Apply the boundary condition for Ê1 at x3 = L R = −1.
In free space and for steady state it follows that γ = jβ = j cω0
i r
Ê = Ê + Ê = {exp[−j cω0 x3 ] − exp[j cω0 (x3 − 2L)]}i1
= exp(−j cω0 L) {exp[−j cω0 (x3 − L)] − exp[j cω0 (x3 − L)]}i1
= −2j exp(−j cω0 L) sin[ cω0 (x3 − L)]i1 .
Apply inversion E = 2 sin[ω(t − cL0 )] sin[ cω0 (x3 − L)]i1 Since the
spatial behavior and the temporal behavior of E are decoupled, this is a
standing wave.
1
In the same way, the total magnetic field follows as H = 2( µε00 ) 2 ×
cos[ω(t − cL0 )] cos[ cω0 (x3 − L)]i2 the phase difference with the electric
field is π2 rad.
Exercise 3.9
i
(a) Apply Eqs. (2.54) - (2.56) with ∂1 = ∂2 = 0 and Ĵ = 0 Ê =
1
( µε00 ) 2 Ĥ0 exp(−γ (1) x3 )i1
r r 1
(b) Ĥ = R Ĥ0 exp(γ (1) x3 )i2 Ê = −( µε00 ) 2 R Ĥ0 exp(γ (1) x3 )i1 ,
t t 1
Ĥ = T Ĥ0 exp(−γ (2) x3 )i2 Ê = 2( εµ00 ) 2 T Ĥ0 exp(−γ (2) x3 )i1
( c ) From the modified Helmholtz equation and the principle of causality
1 1
it follows that γ (1) = s(ε0 µ0 ) 2 and γ (2) = 2s(ε0 µ0 ) 2 , i.e., in both cases
the root of γ 2 has to be chosen in such a way that Re(γ) ≥ 0
(d) Apply the boundary conditions 1 + R = T and 1 − R = 2T

R = − 13 and T = 23
t ∗t 1
( e ) s = jω ST = 12 Re[Ê × Ĥ ] = 12 Re[2( εµ00 ) 2 T T ∗ Ĥ0 Ĥ0∗ ×
1 1
exp(−γ (2) x3 − γ ∗(2) x3 )]i3 = ( εµ00 ) 2 |T |2 |Ĥ0 |2 i3 = 49 ( µε00 ) 2 |Ĥ0 |2 i3 .
Exercise 3.10
t
Employ the electric field analysis Ê = T⊥ exp(−γ (sea) x3 )i1 and
t
Ĥ = Y (sea) T⊥ exp(−γ (sea) x3 )i2 .
The medium properties of seawater are ε = 81ε0 , µ = µ0 and σ = 4 S/m
for f = 100 Hz it is found that in the sea ωσε = 8.9 × 106 1
1
γ (sea) = α + jβ = ( σµ20 ω ) 2 (1 + j) = 4 × 10−2 (1 + j) m−1 Y (sea) = γ (sea)
σ
=
50(1 − j) S.
Y (air) = Y0 = ( µε00 ) 2 = 2.6 mS.
1
−5 (1 + j).
(sea) = 5.3 × 10
2Y0
T⊥ = Y +Y
0
For the transmitted field it follows that
t ∗t
ST = 2 Re[Ê × Ĥ ]
1
∗ ∗(sea)
= 1
2 Re[T⊥ T⊥ Y exp(−γ (sea) x3 ) exp(−γ ∗(sea) x3 )]i3
∗(sea)
2 |T⊥ | Re[Y
2
= 1
] exp{−2 Re[γ (sea) x3 ]}i3
−7
= 1.4 × 10 exp(−8 × 10−2 x3 )i3 W/m2
x3 = 0 m gives ST = 1.4 × 10−7 i3 W/m2 ; x3 = 100 m gives ST =
4.7× 10−11 i 3 W/m2 .
t
Ĵ = σ Ê = 4T⊥ exp(−γ (sea) x3 )i1 x3 = 0 m gives Ĵ = 2.1 × 10−4 ×
(1 + j)i1 A/m2 .
Exercise 3.11
It could be included as a loss term in the inductance L The transmission
line equations, Eqs. (3.112) and (3.113), would then be
∂3 Iˆ + (G + sC)V̂ = Iˆ∆ (s)wδ(x3 ) ,

∂3 V̂ + (R + sL)Iˆ = 0 .
These are also known as the telegraph equations. The circuit parameter
R is of course the series resistance, while the parameter G is the shunt
conductance.
Exercise 4.1
(a) ω = 3π × 109 rad/s f = 1.5 GHz.
tan(θ) = ss13 = 43 θ = 0.927 rad (53.1◦ ).
(b) E = (3i1 − 4i3 ) sin[3π × 109 t − 2π(4x1 + 3x3 )] = Re[Ê exp(jωt)]

Ê = −j(3i1 − 4i3 ) exp[−j2π(4x1 + 3x3 )].
(c) The wave is parallelly polarized Ĥ = Ĥ2 i2 .

Use Eq. (4.6) with K̂2ext = 0 and s = jω Ĥ = jωµ0 ×
50π
exp[−j2π(4x1 + 3x3 )]i2 .
(d) H = Re[Ĥ exp(jωt)] = 5

120π sin[3π × 109 t − 2π(4x1 + 3x3 )]i2 .
Exercise 4.2
Parallel polarization s(1) · ê(1) = 0 and s(2) · ê(2) = 0 ê(1) = Ê (1) ×
[cos(θ)i1 − sin(θ)i3 ] and ê(2) = Ê (2) [cos(θ)i1 + sin(θ)i3 ].
|ê(1) | = |ê(2) | Ê (1) = Ê (2) = Ê0 (assuming their phases to be equal).
Subsitute in Eq. (4.58) and apply Euler’s formula
ω
Ê = 2Ê0 exp[−j sin(θ)x1 ]
c0

ω ω
× cos(θ) cos[ cos(θ)x3 ]i1 + j sin(θ) sin[ cos(θ)x3 ]i3
c0 c0
Propagating wave in the x1 -direction, standing wave in the x3 -direction.
Exercise 4.3
i
Assume for x3 < 0 the presence of the incident field Ê = E0 ×
r
exp[−(γ1i x1+γ3i x3 )]i2 and the reflected field Ê =E0 R⊥ exp[−(γ1r x1+γ3r x3 )]i2 .
The boundary condition Ê2 = 0 for x3 ↑ 0 must hold for all x1 γ1r = γ1i ,
γ3r = −γ3i and R⊥ = −1.
In case of a lossless medium and steady state it is allowed to write γ1i = jβ1i
and γ3i = jβ3i .
For x3 < 0 the total field is
Ê = E0 [exp(−jβ1i x1 − jβ3i x3 ) − exp(−jβ1i x1 + jβ3i x3 )]i2
= E0 exp(−jβ1i x1 ) [exp(−jβ3i x3 ) − exp(jβ3i x3 )]i2
= −E0 exp(−jβ1i x1 ) 2j sin(β3i x3 )i2 .
Propagating wave in the x1 -direction, standing wave in the x3 -direction

For x3 < 0 the amplitude of the total field is |Ê| = 2|E0 | | sin(β3i x3 )|
Graph of the amplitude vs. x3 looks like a rectified sine with height 2|E0 |
and with zeros at x3 = − nπ
βi
(n ∈ IN ).
3
Exercise 4.4
(a) Parallel polarization since the electric field is entirely in the plane of
propagation.
√ √
(b) si = 12 3i1 + 12 i3 sin(θi ) = si1 =√12 3 θi = π3 .
√ to Snell’s law, sin(θ ) = 1.5 sin(θ ) θ = 4 s =
i t t π t
According
√
1 1
2 2i1 + 2 2i3 .
(c) Parallel polarization Apply magnetic

√ field analysis.
E1 = 12 E0 cos[6π × 109 t − 10π( 3x1 + x3 )] = Re[Ê1 exp(jωt)]
√ √
Ê1i = 12 E0 exp[−j10π( 3x1 +x3 )] Ĥ2i = H0 exp[−j10π( 3x1 +x3 )]
1
with H0 = ( µε00 ) 2 E0 (the same result may be deduced starting with Ê3i ).
√
The reflected magnetic field is Ĥ2r = R H0 exp[−j10π( 3x1 − x3 )]
√ √
Ê1r = −√12 R E0 exp[−j10π( 3x1 − x3 )] and Ê3r = − 12 3R E0 ×
exp[−j10π( 3x1 − x3 )]
Apply, e.g., Eq. (4.176) R = −7.2 × 10−2 .
√
Perform inversion E r = −( 12 i1 + 12 3i3 ) E0 R cos[6π × 109 t −
√
10π( 3x1 − x3 )]. √
In a similar way it follows that E t = 13 3(i1 −i3 ) E0 T cos[6π×109 t−
√
10π 3(x1 + x3 )], in which T = 0.928.
(d) ...
...
...
er et
......... .
...... ............ ..... .....
....
.
.....
.......
... ........... ............
............
.. ...
π ........... .............. π
.
3 ...... ................ ..... 4
. ..
π ...............
3 ......
ei ..........
.......... ...........
...........
...
....
.
...
x3 = 0
Exercise 4.5
(2) (2) 1
(a) θci = arcsin( nn(1) ) = arcsin[( εε(1) ) 2 ] = arcsin( 12 ) = π
6 rad.
√
(b) γ1i = γ i sin(θi ) = γ0 3 and γ3i = γ i cos(θi ) = γ0 .
1 1
[−ω 2 ε(2) µ(2) − (γ i )2 ] 2 with Re[γ3t ] ≥ 0
( c ) γ3t = √ γ3t = (γ02 − 3γ02 ) 2 =
−jγ0 2.
(d) The wave is attenuated by a factor of exp{−Re[γ3t d]} √ An attenua-

tion of exp(−1) occurs over a distance x3 = 1/Re[γ3t ] = 12 2|γ0 |−1 .
√
1+j √2
( e ) Substitute the relevant quantities in Eq. (4.159) R⊥ = 1−j 2

|R⊥ | = 1 and arg(R⊥ ) = 1.91 rad (109◦ ).
Exercise 4.6
All the light that comes in from the air enters the water at angles smaller
than or equal to the critical angle; this is the reciprocal use of total reflection
at angles of incidence larger than the critical angle for plane waves from water
to air. The critical angle is θc = 0.85 rad (48.7◦ ).
Exercise 4.7
nglass
nair
= 32 The critical angle exists for a wave incident on the glass/air
interface and does not exist for a wave incident on the air/glass interface;
this is independent of the type of polarization.
glass
εglass
εair
= 2.25 and µµair = 1 The Brewster angle does not exist in case
of perpendicular polarization, whereas it exists for both the air/glass and
the glass/air interfaces in case of parallel polarization.
Interface Polarization Critical angle θci i

Brewster angle θB
Glass/air ⊥ arcsin( 23 ) —
arcsin( 23 ) arctan( 23 )
Air/glass ⊥ — —
— arctan( 32 )
Exercise 4.8
µ(2)
Take Eq. (4.181) and substitute µ(1)
=1
⎛ ⎞1 1 1
ε(2)
1−
2
ε(2) (ε(1) − ε(2) ) 2
ε(2) 2
i
tan(θB ) =⎝ ε(1) ⎠ = =
ε(1)
−1 ε(1) (ε(1) − ε(2) ) ε(1)
ε(2)
Exercise 5.1
x3;0 θ0 Ray trajectory
2.50 45◦ Ray enters the halfspace x3 > 3 with θexit = 62◦
Ray turns at x3;hor = 2.84 and enters the half-
60◦
space x3 < 1 with θexit = 147◦
Ray starts at x3;hor = 2.50 and enters the half-
90◦
120◦ Ray enters the halfspace x3 < 1 with θexit = 147◦
2.00 45◦
60◦
90◦
(Continued)
x3;0 θ0 Ray trajectory

1.50 45◦
60◦
90◦
Exercise 5.2
The extreme positions where a ray may have a horizontal tangent are
x3;hor = 1 and x3;hor = 3 For a horizontal tangent in between these
extreme positions it is found that n(x3;hor ) > 1 For√an undulating ray,
C0 = n(x3;hor ) sin( π2 ) > 1 C0 = n(x3;0 ) sin(θ0 ) = 2 sin(θ0 ) > 1
π 3π
4 < θ0 < 4 .
Exercise 5.3
(1) (2)
Yes: For any ray with 1 < x3;0 < 3 there exist a x3;hor and a x3;hor with
(1) (2) (1) (2)
1 < x3;hor < x3;0 < x3;hor < 3 and n(x3;0 ) > n(x3;hor ) = n(x3;hor ) > 1
For any 1 < x3;0 < 3 a range of θ0 may be found for which n(x3;0 ) sin(θ0 ) =
(1,2)
n(x3;hor ) sin( π2 ) > 1 arcsin( n(x13;0 ) ) < θ0 < π − arcsin( n(x13;0 ) ).
Exercise 5.4
For any 1 < x3;0 < 2 there exists a x3;hor in the range x3;0 < x3;hor < 2
such that n(x3;0 ) sin(θ0 ) = n(x3;hor ) sin( π2 ) may be solved for θ0 Any
starting position 1 < x3;0 < 2 has a ray with a horizontal tangent.
For any 2 < x3;0 < 3 there exists a x3;hor in the range 2 < x3;hor < x3;0
such that n(x3;0 ) sin(θ0 ) = n(x3;hor ) sin( π2 ) may be solved for θ0 Any
starting position 2 < x3;0 < 3 has a ray with a horizontal tangent.
Since ∂3 n = 0 there is no ray with a horizontal tangent for x3;0 = 2.
Exercise 5.5
For rays with a horizontal asymptote it must be possible to solve
n(x3;0 ) sin(θ0 ) = n(2) sin( π2 ) for θ0 For all 1 < x3;0 < 2 a solution
θ0 = arcsin[ n(x13;0 ) ] may be found, and for all 2 < x3;0 < 3 a solution
θ0 = π − arcsin[ n(x13;0 ) ] exists.

Since ∂3 n = 0 there is no ray with a horizontal tangent for x3;0 = 2.
Exercise 6.1
(a) b = 11.4 mm.
(b) P = 12 Re[V̂ Iˆ∗ ] = 12 Z0 |î+ |2 .
( c ) Use Eq. (6.55) for φ = 100 |E| = (e21 + e22 )1/2 = r ln(b/a)
100

|E| has a maximum when r = a |E|max = a ln(b/a) = 2.38 × 104
100
V/m.
Exercise 6.2
)∞ )∞ −8
(a) v + (t) = Z0Z+Z
0
S n=0 (ΓS ΓL )
n V (t−2nL )
S c = 5
8
3 n
n=0 ( 28 ) VS (t−4×10 n).
(b) V ( 12 L, t) = v + (t − 2c L
) + ΓL v + (t − 2L−L/2
c ) = 58 [VS (t − 10−8 )−
−8 −8 −8
7 VS (t − 3 × 10 ) + 28 VS (t − 5 × 10 ) − 196 VS (t − 7 × 10 ) + · · ·] and
3 3 9
I( 12 L, t) = Z10 [v + (t − 2c L
) − ΓL v + (t − 2L−L/2
c
1
) = 80 [VS (t − 10−8 )+
−8 −8 −8
7 VS (t − 3 × 10 ) + 28 VS (t − 5 × 10 ) + 196 VS (t − 7 × 10 ) + · · ·].
3 3 9
∞ 1 10−9
( c ) WS = 0 VS (t)I(0, t) dt = Z0 +ZS 0 VS2 (t) dt = 12.5 pJ.
10−9 Z0 10−9
(d) W + = 0 V (0, t)I(0, t) dt = (Z0 +ZS )2 0 VS2 (t) dt = 7.8 pJ.

(e) WL = 0∞ V (L, t)I(L, t) dt = 0∞ (1 + ΓL )v + (t − L 1
c ) Z0 (1 − ΓL )×

v + (t − Lc ) dt = Z10 (1 − Γ2L ) 0∞ [v + (t − Lc )]2 dt.
)∞ (2n+1)L 2
[v + (t − Lc )]2 = [ Z0Z+Z
n=0 (ΓS ΓL ) VS (t −
0 n
c )] = ( Z0Z+Z
0
)2 ×
)∞ (2n+1)L
S S
2n 2
n=0 (ΓS ΓL ) VS (t− c ) since VS (t− (2n+1)L
c )VS (t− (2m+1)L
c )=0
for all n = m.
∞ + L 2 Z0
2 )∞ (Γ Γ )2n ∞ V 2 (t− (2n+1)L ) dt =
0 [v (t− c )] dt = ( Z0 +ZS ) n=0 S L 0 S c
1−Γ2L
( Z0Z+Z
0
S
)2 1
1−(ΓS ΓL )2
× 10−9 WL = Z0
(Z0 +ZS )2 1−(ΓS ΓL )2
× 10−9 =
6.5 pJ .
Exercise 6.3
The 100 V signal travels in 2 µs from the beginning to the end T = 2 µs.
At the end of the line the sum of the incoming 100 V signal and the once
reflected signal gives a 75 V signal over the load A once reflected signal
L −Z0
of −25 V signal originates at the end ΓL = Z ZL +Z0 = − 4
1
ZL = 60 Ω.
At the beginning of the line the reflected −25 V signal and the twice
reflected signal gives a −10 V signal over the source A twice reflected
S −Z0
signal of 15 V originates at the source ΓS = ZZS +Z0 = − 5 ZS = 25 Ω.
3
The voltage divider existing of ZS = 25 Ω and Z0 = 100 Ω causes a 100 V

signal right after t = 0 s V0 = 125 V.
Exercise 6.4
To mimic a capacitor, the impedance at the beginning of the line must have
−j
the approximate form Zin ≈ ωC for some range of ω, say ω0 < ω < ω0 + ∆ω.
The imput impedance of a short-circuited transmission line may be writ-
jZ0
ten as Zin = cot(ωL/c) .
The approximation cot[ (1+2n)π
2 + x] ≈ −x holds for |x| 1 and n ∈ ZZ
The transmission line behaves like a capacitor for L = (1+2m)πc
2ω0 = λ40 + mλ
2 ,
0
with m ∈ IN , provided that ω0 (2m+1)π .

∆ω 2
Using an equivalent argumentation, it may be found that an open trans-

mission line behaves like an inductor for L = (1+2m)πc
2ω0 = λ40 + mλ2 , with
0
ω0 (2m+1)π .
m ∈ IN , provided that ∆ω 2
Exercise 6.5
−1
For a lossless transmission line, |Γ| = |ΓL | |ΓL | = VSWR
VSWR+1 |ΓL | = 2
1
for set A and |ΓL | = 5 for set B Set B is better matched to the trans-
1
mission line.
Exercise 7.1
1
(a) k = ω(εµ) 2 = 41.9 m−1 and πa = 31.4 m−1 k ≥ mπ a (propagating
modes) for m = 1, k < mπ
a (evanescent modes) for m = 2, 3, 4, · · · .
(b) The field strengths of the propagating TE1 -mode follows from Eqs.
(7.21) - (7.23) as
Ê2 = 2j Â1 sin(31.4x1 ) exp(−27.7jx3 ),

Ĥ1 = −7.02 × 10−3 j Â1 sin(31.4x1 ) exp(−27.7jx3 ),
Ĥ3 = −7.96 × 10−3 Â1 cos(31.4x1 ) exp(−27.7jx3 ).
( c ) At x3 = 0 the reflection coefficient of the electric field is R⊥ = −1

The total electric field is Ê2 = 4Â1 sin(31.4x1 ) sin(27.7x3 ) (cf. Ex-
ercise 4.3) Use Eqs. (7.6) and (7.7) Ĥ1 = −1.40 × 10−2 j Â1 ×
sin(31.4x1 ) cos(27.7x3 ), Ĥ3 = 1.59 × 10−2 j Â1 cos(31.4x1 ) sin(27.7x3 ).
Exercise 7.2
∗
P = S · ν dA = 12 w 0a Re[(Ê × Ĥ ) · i3 ]mode dx1 For a given |E0 | the
1
power is largest for the TM0 - or TEM-mode and equals P = 12 wa( µε ) 2 |E0 |2 .
For safety take |E0 | = 2 × 105 V/m P = 26.5 kW.
Exercise 7.3
k = 2πf 2π
c = λ λg,m = k2π =√ 1
λg,0 = 1 m and λg,1 = 1.34 m
g,m 1−4m2 /9
while for m = 2, 3, 4, · · · the modes are evanescent and we can no longer
speak of a guided wavelength.
Exercise 7.4
The EHF band ranges from 30 GHz to 300 GHz The cutoff frequency
must be chosen in such a way that fc,0 ≤ 30 × 109 Hz < f < 300 × 109 Hz ≤
fc,1 .
c0
Equation (7.113) with n3 = n1 and m = 0 gives fc,0 = 0 Hz and fc,1 = 6a
fc,0 ≤ 30×109 Hz is satisfied for any value of a while fc,1 ≤ 300×109 Hz
requires that a < 0.167 mm.
Exercise 7.5
In each of the regions the general solution is
(i) (i)
Ê = Â(i) exp(jk1 x1 ) + B̂ (i) exp(−jk1 x1 ).
Further it is known that Â(3) = 0 and B̂ (1) = 0.

(a) From the boundary conditions at x1 = a it follows that
Â(2) exp(jκ(2) a) + B̂ (2) exp(−jκ(2) a) = B̂ (3) exp(−κ(1) a),

κ(2) [Â(2) exp(jκ(2) a) − B̂ (2) exp(−jκ(2) a)] = jκ(1) B̂ (3) exp(−κ(1) a).
From the boundary condition at x1 = −a it follows that
Â(2) exp(−jκ(2) a) + B̂ (2) exp(jκ(2) a) = Â(1) exp(−κ(1) a),

κ(2) [Â(2) exp(jκ(2) a) − B̂ (2) exp(−jκ(2) a)] = −jκ(1) Â(1) exp(−κ(1) a).
(b) The first two equations of (a) yield

κ(2) + jκ(1) (3)
Â(2) exp(jκ(2) a) = B̂ exp(−κ(1) a),
2κ(2)
κ(2) − jκ(1) (3)
B̂ (2) exp(−jκ(2) a) = B̂ exp(−κ(1) a).
2κ(2)
The last two equations of (a) yield
κ(2) − jκ(1) (1)
Â(2) exp(−jκ(2) a) = Â exp(−κ(1) a),
2κ(2)
κ(2) + jκ(1) (1)
B̂ (2) exp(jκ(2) a) = (2)
Â exp(−κ(1) a).
2κ
( c ) From the first and third equation of (b) and from the second and fourth
equation of (b) it is found that
κ(2) − jκ(1) (1) κ(2) + jκ(1) (3)
Â exp(jκ(2) a) = B̂ exp(−jκ(2) a),
2κ(2) 2κ(2)
κ(2) + jκ(1) (1) κ(2) − jκ(1) (3)
Â exp(−jκ(2) a) = B̂ exp(jκ(2) a).
2κ(2) 2κ(2)
Addition of these equations yields Â(1) = B̂ (3) and
(1)
κm 1
κ(2)
m a = arctan( (2)
) + mπ, m = 0, 2, 4, · · · .
κm 2
Subtraction of these equations yields Â(1) = −B̂ (3) and
(1)
κm 1
κ(2)
m a = arctan( (2)
) + mπ, m = 1, 3, 5, · · · .
κm 2
(d) For m = 0, 2, 4, · · ·, part (c) shows that Â(1) = B̂ (3) Substitute

(2) (1)
in (b) Â(2) = B̂ (2) and Â(2) cos(κm a) = 12 Â(1) exp(−κm a)
Substitute in the general solution This results in Ê2;m for even m,
(even) (2) (1) exp(−κ(1)
m a)
in which Ĉm = 2Âm = Am (2) .
cos(κm a)
For m = 1, 3, 5, · · ·, part (c) shows that Â(1) = −B̂ (3) Substitute
(2) (1)
in (b) Â(2) = −B̂ (2) and Â(2) sin(κm a) = 12 j Â(1) exp(−κm a)
Substitute in the general solution This results in Ê2;m for odd m,
(odd) (2) (1) exp(−κ(1)
m a)
in which Ĉm = 2j Âm = −Am (2) .
sin(κm a)
Exercise 8.1
In the Taylor series expansion around the stationary point k1s the linear term
is proportional to the first derivative at k1s , which by definition is equal to
zero.
Exercise 8.2
Parallel electric current:
D(θ) = 0 if sin(θ) = 0 θ = 0 and θ = π.
sin[ka cos(θ)/2]
D(θ) = 0 if ĥ+ 2 [k cos(θ), jω] = k cos(θ) ı̂(jω) = 0 ka cos(θ)
2 = nπ
with n ∈ ZZ \{0} θ = arccos( nλ
a ) with |nλ| ≤ |a|.
Perpendicular electric current:
sin[ka cos(θ)/2]
D(θ) = 0 if ĥ+
1 [k cos(θ), jω] = k cos(θ) ı̂(jω) = 0 ka cos(θ)
2 = nπ
with n ∈ ZZ \{0} θ= arccos( nλ
a ) with |nλ| ≤ |a|.
Exercise 8.3
(a) Use Eq. (8.56) with a delta function in the integrand ĥ+1 (k1 , jω) =
1
2 ı̂(jω).
(b) Take together Eqs. (8.58) - (8.59) and substitute ĥ+ 1 Ĥ ≈

k 21
2 ı̂∆ (jω) ( 2πr ) exp(−jkr + j 4 ) [sin(θ)i1 − cos(θ)i3 ].
1 π
1
k 21
Take Eq. (8.60) and substitute ĥ+ 1 Ê ≈ − 12 ( µε ) 2 ı̂∆ (jω) ( 2πr ) ×
exp(−jkr + j π4 )i2 .
Note that the electric field strength and the magnitude of the mag-
netic field strength are not dependent on the angle of observation.
(c) θ=0
⊗
.........................
..
...........
.........
.
.
..
........ . ⊗ ..
.
⊗ ....
..
..
....... ...
.
.
... .... H .....
... .. . ........
.. ... ...
.. ..
⊗ ⊗
... .
.. .. ...
. . . . ......
....
.
........
. ....
.......
...
..
.
.
.
. ....
.
..... ..
.
E
... .......
.......
...
.. .
i1 ..
..
.......
.......
...
....... ... ........... .. .... .........
..
....... .. .... .. ....... ....
O p
. . .
θ = 32 π ⊗ ⊗ θ = 12 π
. . .
...................................... .
...
...
. ....... ....... ....... ....... .......
...... . .
.................. ....
.......
i3 ....... ....... ....... ...
.
...
.........
.... ....... ... .
.......
....... . ... .......
...... ... .. .......
...
..... . .. ... ... ....... ...........
....... ... . .. ....... .....
E⊗ ... . ... .
.
..
...
.. ⊗ .
... ...
........... .
... ... .... ..
.
. ...
... .......
H ⊗ ... .. ..........
.......
....... .
..........
.
..
⊗
.......
⊗ .........................
θ=π
∗ * +1
(d) The time-averaged power flow is 12 Re[Ê × Ĥ ] = |ı̂∆ (jω)|2 µε 2 16πr
k
×
[cos(θ)i1 + sin(θ)i3 ] D(θ) = 1 The directive gain is the unit
circle.
Bibliography
A. Einstein (1956), The Meaning of Relativity, Princeton University Press,

Princeton.
D.K. Cheng (1993), Fundamentals of Engineering Electromagnetics, Addison-
Wesley Publishing Company, Reading, Massachusetts.
R.L. Coren (1989), Basic Engineering Electromagnetics, Prentice-Hall, En-
glewood Cliffs, New Jersey.
C.T.A. Jonk (1988), Engineering Electromagnetic Fields and Waves, John
Wiley & Sons, New York.
J.D. Kraus (1992), Electromagnetics, fourth edition, McGraw-Hill, New
York.
S.V. Marshall and G.G. Skitek (1990), Basic Engineering Electromagnetics,
third edition, Prentice-Hall, Englewood Cliffs, New Jersey.
J.C. Maxwell (1873), A Treatise on Electricity and Magnetism, Clarendon
Press, Oxford.
N.N. Rao (1987), Elements of Engineering Electromagnetics, second edition,
Prentice-Hall, Englewood Cliffs, New Jersey.
L.C. Shen and J.A. Kong (1983), Applied Electromagnetism, Brooks/Cole
Engineering Division, Moneterey, California.
K.F. Sanders and G.A.L. Reed (1986), Transmission and Propagation of
Electromagnetic Waves, second edition, Cambridge University Press, Cam-
bridge.
E.T. Whittaker (1953), A history of the Theories of Aether and Electricity,
vol. I: Classical theories; vol. II: Modern theories (1900-1926), Nelson,
London.
Index
Admittance, 52 Characteristic impedance, 74, 150,

Angle 152, 154, 158, 159, 162
of incidence, 99, 108, 111, 113, Circuit equivalent, 74, 79, 155
114 Circuit parameter, 73
of reflection, 99, 108 Coaxial cable, 145
of transmission, 109 Coaxial line, 152, 156, 158
Angular direction, 211, 218 Compatibility relation, 25, 28
Angular frequency, 37, 54, 84 Conductance per unit length, 73
Antenna, 49, 203 Conductivity, 29, 34, 81, 121
Attenuation coefficient, 57, 84, 87, Conductor, 46
178 Conservation
Attenuation factor, 58, 86 of electric charge, 45
Axial direction, 168 of energy, 43, 44
Constitutive relation, 28, 31
Base vector, 2 Continuity
Boundary of normal component of S, 34
curved, 121 of tangential components of E,
plane, 95, 101, 121 34, 64, 68, 103
of tangential components of H,
Boundary condition, 32, 65, 69, 73,
34, 64, 68, 103
96–98, 103, 106, 147, 149,
Critical angle, 114
151, 153, 157, 174, 176,
Curl, 7, 13
183, 184, 188, 189
Current distribution
Boundary surface, 95
non-uniform, 121
Brewster angle, 113
uniform, 121
Bromwich inversion integral, 36
Cut-off angular frequency, 178
Cut-off frequency, 180, 193
Capacitance per unit length, 74 of TE-mode, 193
Cartesian coordinates, 4 of TM-mode, 194
Cartesian reference frame, 2 Cylindrical wave, 211, 218
Cartesian vector, 4
Causal field, 25 Dielectric property, 30
Causality, 29, 36, 52, 95, 101–103, Differentiation
162 with respect to parameter, 6
Characteristic admittance, 150 with respect to spatial coordi-
Characteristic equation, 185, 189 nates, 6
244 index
Dirac distribution, 33, 50, 204, 213 Electromagnetic theory, 2

Direction of propagation, 54, 56, Electromagnetic wave, 1
58, 60, 61, 83, 88–90, 92, Electromagnetic wave propagation,
99, 108, 109, 124, 126, 127, 49
130, 131, 180 Electromagnetic wave speed, 24
Directional derivative, 10 Energy
Directive gain, 211, 213, 219 stored in electric field, 42
Dispersion stored in magnetic field, 42
geometrical, 193 volume density of electric field,
waveguiding, 193 43
Dispersion curve, 179, 195 volume density of magnetic field,
Dispersion equation, 186, 191 43
Dispersion relation, 174, 177 Excitation, 1, 203
Distributed circuit, 75 Excitation condition, 53, 206, 215
Divergence, 7, 10
Far-field approximation, 207, 216
E-polarization, 83 Fast Fourier Transform, 36
E-polarized wave, 172, 173, 182, Fiber
183 graded-index, 169
Eigenvalue, 185, 190 step-index, 171
Eigenvalue equation, 185, 189 Field equations, 21, 31
Eikonal, 122, 125, 129 in frequency domain, 37, 146
Eikonal equation, 126, 130 Flux, 12
Electric charge, 22 Force, 21, 22
volume density of, 28 Fourier integral, 206, 207, 216
Electric current, 73 Fourier transform, 209, 211, 217,
volume density of, 25, 27 219
Electric-current sheet, 50, 203, 204, Fourier transformation, 37
213 Fourier’s inversion theorem, 207,
Electric field analysis, 53, 64, 69, 216
108 Frequency, 54, 56
Electric field strength, 22, 24, 88, Frequency domain, 36, 108
89 Fresnel coefficient
Electric flux density, 27 for reflection, 111, 113
Electric force, 23 for transmission, 111
Electric polarization, 27
Electric potential, 73 Gauss’ integral theorem, 12, 41
Electrically impenetrable half-space, Geometrical optics, 121
95, 109 Gradient, 6, 8
Electrically impenetrable object, 34 Group velocity, 193
Electromagnetic ray, 121, 124, 126, Guided wavelength, 198
129
parametric representation of, H-polarization, 82
132 H-polarized wave, 172, 175, 182,
uniform, 126, 130, 131 188
index 245
Heat, 42 Magnetic field analysis, 53, 65, 105

volume density of, 44 Magnetic field strength, 22, 24, 88,
Helmholtz equation, 173, 176, 184, 89
188 Magnetic flux density, 27
modified, 51, 74, 82, 83 Magnetic force, 23
Magnetic property, 30
Imaginary unit, 36 Magnetically impenetrable object,
Impedance, 53 35
Incident field, 62, 67 Magnetization, 27
Incident wave, 95, 101, 108, 121 Matter
Index of refraction, 109, 128, 131, properties of, 28
133, 135, 142 Maxwell’s equations, 31
effective, 193, 195 in matter, 25
Inductance per unit length, 74 in vacuum, 24
Infrared, 56 Maxwell, James Clerk, 2
Input impedance, 157, 158, 161 Medium
capacitive, 159 anisotropic, 29
inductive, 159 dielectric, 111, 113
Integrated-optics system, 171 heterogeneous, 29
Interface, 32, 62, 101 highly conducting, 59
Interference, 90 homogeneous, 29, 121, 122
International System of Units, 4 inhomogeneous, 29, 121, 124,
Isotropic radiator, 219 127, 128, 130
instantaneously reacting, 28
Laplace equation, 149, 151, 153 isotropic, 29, 124, 127, 128, 130,
Laplace transform, 161 131
Laplace transform parameter, 36 linear, 28
Laplace transformation, 36 locally reacting, 29
shift rule of, 60, 161, 162 lossless, 55, 59, 86, 92, 100,
Tauber’s theorem for, 124, 125, 109, 111, 122, 160
129 lossy, 57
Lens of Luneberg, 142 non-linear, 28
Line source, 221 parametrically affected, 28
Load, 156 passive, 28
matched, 158, 159 time invariant, 28
open-circuit, 158, 159 time variant, 28
short-circuit, 158, 159 Medium parameter, 133, 138
Load impedance, 156 Mode
Longitudinal direction, 168 evanescent, 180
Lorentz force, 23 fundamental, 194
Loss tangent, 80 guided, 186, 191, 193
Lumped circuit, 75 non-propagating, 175, 178, 180
propagating, 178, 180, 193
Magnetic current propagation properties of, 178,
volume density of, 25 193
246 index
TEm , 175, 180, 186, 193 Plane wave, 49, 83, 84, 90, 95, 101,
TEM, 178 121, 126, 128, 130, 131,
TMm , 177, 180, 191, 193 167, 180, 204, 214
Mode index, 193 evanescent, 206, 215
non-uniform, 86, 109, 114, 195,
Normal incidence, 105, 108 203, 206, 215
parallelly polarized, 87, 93
One-dimensional wave, 49, 56, 58, perpendicularly polarized, 89,
60, 61, 66, 87, 105, 108, 93
167, 203 uniform, 86–89, 91, 99, 108,
Operator 109, 122, 126, 130, 203,
curl, 7, 13 206, 211, 215, 219
del, 6 Plane-wave decomposition, 180
div, 7, 10 Plasma, 201
grad, 6, 8 Plasma frequency, 201
Laplacian, 18 Point charge, 21
nabla, 6 Polarization
of Hamilton, 6 circular, 39
Optical communication system, 169 elliptical, 39
Optical fiber, 169 left-handed, 41
Orientation linear, 39
of electric field, 54 parallel, 82, 87, 93, 97, 100,
of magnetic field, 54 103, 111, 113, 114, 122,
Origin, 2 124, 172, 182, 204
perpendicular, 83, 89, 93, 98,
Parallel wires, 145 100, 106, 111, 113, 114,
Penetration depht, 80 122, 128, 172, 182, 213
Permeability, 22, 24, 29, 35, 81, right-handed, 41
121 Polarization state, 39
relative, 30 Position, 2
Permittivity, 24, 29, 34, 81, 121, Power, 114, 159
201 dissipated into heat, 42
relative, 30 generated by sources, 43
Phase, 55 instantaneous, 41
Phase coefficient, 57, 87, 178 volume density of, 43, 44
Phase factor, 58, 86, 87 Power flow, 32, 89, 90, 92–94, 159
Phase vector, 84 Power flow density, 175, 187, 192,
Phase velocity, 193 210, 218
Physical laws, 4 Power relation, 41
Physical quantity, 2, 4 Poynting theorem, 41
Plane complex, 44
of equal amplitude, 58, 86, 114, instantaneous, 41
195 Poynting vector, 31, 43, 58, 61, 86,
of equal phase, 58, 86, 114, 195 88, 90, 92
of observation, 82, 83 complex, 44
index 247
Product Shielding effectiveness, 70

cross, 5 SI, 4
dot, 4 Skin depth, 59, 80
scalar, 4 Snell’s law
vector, 5 for horizontally layered medium,
Propagation, 1 135
Propagation coefficient, 51, 55, 57, for radially layered medium, 139
59, 62, 63, 67, 68, 87, 160, of reflection, 109
167, 185, 186, 190, 191 of refraction, 109, 113
Propagation constant, 174, 177, 178 of transmission, 109
Propagation factor, 58, 86 Source, 1, 26, 50, 81, 122, 156
Propagation laws, 2 nonplanar, 121
Propagation vector, 84 planar, 121
Sourcefree interface, 34
Radiation, 50 Spatial period, 56, 87
Ray approximation, 121, 126, 129 Specific value, 185, 190
Ray trajectory, 121, 124, 127, 130, Standard clock, 2
131 Standard experiments, 2, 4
in horizontally layered medium, Standard measuring rod, 2
133 Stationary phase, 207, 217
in radially layered medium, 138 Stationary point, 207
Receiver, 1 Steady-state, 38, 43, 54, 84, 92,
Reflected field, 62, 67 100, 158
Reflected wave, 95, 101, 108, 121 Stokes’ integral theorem, 15
Reflection, 62, 95, 101 Substrate, 171, 182
Reflection coefficient, 64, 66, 69, Superposition, 49, 90, 93, 100, 175,
104, 105, 107, 108, 111, 177, 180, 203, 204, 206,
156, 157, 161, 163 214, 215
reflection coefficient, 158 Superstrate, 171, 182
Refraction, 1 Surface impulse, 33
Relativity, 2, 24 Surface source, 33
Relaxation, 29
Relaxation time, 46 Taylor series, 208
Resistance, 79 TE-wave, 172, 173, 182, 183
Right-handed triad, 2 TEM-wave, 72, 145, 146, 167, 172,
175
Scalar function, 6 Time, 2
derivative of, 6 Time-invariant configuration, 36
Scalar potential, 149 TM-wave, 172, 175, 182, 188
Separation of variables, 148 Total reflection, 114
Sheet emitter Trajectory constant, 135, 139
parallel electric current, 204 Transient, 160
perpendicular electric current, Transient emission, 59
213 Transient field, 36
Shielding, 66 Transmission, 62, 101
248 index
Transmission coefficient, 65, 69, 104, Voltage standing wave ratio, 165
105, 107, 108, 111 VSWR, 165
Transmission line, 49, 145, 167
lossless, 158, 160 Wave
multiconductor, 145 evanescent, 175, 177
two-conductor, 156 non-propagating, 175, 177
Transmission line equations, 73, 79, not plane, 121
150 plane, see Plane wave
Transmission line equivalent, 73 propagating, 175, 177
Transmitted field, 63, 68 standing, 100
Transmitted wave, 102, 109, 110, travelling, 100
114, 121 Wave admittance, 52, 55, 62, 63,
Transverse electric wave, 172 68, 89, 108
Transverse electromagnetic wave, Wave impedance, 53, 55, 64–66,
72, 145, 167 69, 88, 106
Transverse magnetic wave, 172 Wave speed, 56, 60, 61, 161, 163
Transverse plane, 168 Wavefront, 121, 123, 125, 127, 129,
Triple product 130
scalar, 5 Waveguide, 167
vectorial, 5 asymmetric dielectric slab, 194
Two-dimensional medium, 121 closed, 168
Two-dimensional wave, 81, 121, 167, cylindrical, 167, 168
203 dielectric slab, 171, 182, 193
graded-index slab, 182
Uniform line, 145 open, 168, 169
Unit, 4 parallel-plate, 72, 145, 151, 156,
Unit impulse 158, 171, 172, 178
one-dimensional, 50, 204, 213 planar, 171
two-dimensional, 221 planar graded-index, 171
planar step-index, 171
Vacuum, 30 step-index slab, 182
Vector, 3 symmetrical slab, 182
addition of, 4 thin-film, 182
components of, 4 uniformly cylindrical, 169
length of, 5, 37 Wavelength, 56, 87, 94
multiplication of, 4 Wavenumber, 174, 177, 193
properties of, 4
subtraction of, 4
Vector calculus, 4
Vector function, 6
derivative of, 6
Vectorial position, 3
Velocity
of observer, 55
of point charge, 22
Textbooks on science and technology from VSSD
Business engineering, innovation, economics Electronics, physics
Creative Facilitation Design of High-performance Negative-feedback

A Delft approach Amplifiers
Marc Tassoul E.H. Nordholt
2005 / viii+144 p. / ISBN 90-71301-46-X 2001 / xvi+234 pag., ISBN 90-407-1247-6
http://www.vssd.nl/hlf/b005.htm http://www.vssd.nl/hlf/e022.htm
Fundamentals of Business Engineering and Electronic instrumentation

Management P.P.L. Regtien
A systems approach to people and organisations 2005 / xiv + 397 p. / ISBN 90-71301-43-5 / hardback
W. Ten Haaf, H. Bikker, D.J. Adriaanse http://www.vssd.nl/hlf/e008.htm
with contributions from J. in 't Veld and P.Ch-A.
Malotaux Electromagnetic Waves — An introductory course
2002 / xvi + 728 pp. / ISBN 90-407-2210-2 / hardback P.M. van den Berg and H. Blok
http://www.vssd.nl/hlf/b001.htm 2001 / X+244 p. / ISBN 90-407-1836-9
Macroeconomics
James K. Galbraith and William Darity, Jr.
2005 / xxii+505 p. / ISBN 90-71301-57-5
http://www.vssd.nl/hlf/b006.htm
Civil engineering, coastal engineering Materials science
Breakwaters and closure dams Fracture Mechanics

Bed, Bank and Shore Protection 2 M. Janssen, J. Zuidema, R.J.H. Wanhill
K. d'Angremond en F.C. van Roode 2002 / xii + 365 p. / ISBN 90-407-2221-8
2001 / xii+340 p. p. / ISBN 90-407-2127-0 http://www.vssd.nl/hlf/m004.htm
http://www.vssd.nl/hlf/f011.htm
From Polymers to Plastics
Ebb and Flood Channel Systems in the Netherlands A.K. van der Vegt
Tidal Waters 2006 / 268 p. / ISBN 90-71301-62-1
Johan van Veen http://www.vssd.nl/hlf/m028.htm
2002 / 32 p. / ISBN 90-407-2338-9
http://www.vssd.nl/hlf/f015.htm Mathematical Geodesy
Introduction to bed, bank and shore protection Series on Mathematical Geodesy and Positioning
Engineering the interface of soil and water http://www.vssd.nl/hlf/a030.htm
Gerrit J. Schiereck Adjustment Theory
2004 / xii+ 400 p. / ISBN 90-407-1683-8 P.J.G. Teunissen
http://www.vssd.nl/hlf/f007.htm 2003 / vi + 201 p. / ISBN 90-407-1974-8
Principles of river engineering Hydrography
The non-tidal alluvial river C.D. de Jong, G. Lachapelle, S. Skone, I.A.Elema,
P.Ph. Jansen (ed.) 2003 / x+351 pp. / ISBN 90-407-2359-1
1994 / XVI+509 p. / A4 / ISBN 90-407-1280-8
http://www.vssd.nl/hlf/f006.htm Dynamic Data Processing
P.J.G. Teunissen
The vertical motion of foundations and pontoons 2001 / x + 241 p. / ISBN 90-407-1976-4
Godfried Kruijtzer
2002 / 49 p. / ISBN 90-71301-64-8 Testing Theory
http://www.vssd.nl/hlf/f016.htm P.J.G. Teunissen
2000 / viii + 147 p. / ISBN 90-407-1975-6
Chemical engineering
Mathematics
Mass Transfer in Multicomponent Mixtures
J.A. Wesselingh and R. Krishna Mathematical Systems Theory
2006 / 329 p. / ISBN 90-71301-58-3 G.J. Olsder and J.W. van der Woude
http://www.vssd.nl/hlf/d004.htm 2008 / x+208 p. / ISBN 90-71301-40-0
http://www.vssd.nl/hlf/a003.htm
Transport phenomena data companion
L.P.B.M. Janssen and Numerical Methods in Scientific Computing
M.M.C.G. Warmoeskerken J. van Kan, A. Segal and F. Vermolen
2006 / 168 p. / ISBN 90-71301-59-1 2005 / xii+277 pp. / ISBN 90-71301-50-8
http://www.vssd.nl/hlf/c017.htm http://www.vssd.nl/hlf/a002.htm
Order in Space
JA.K. van der Vegt
2006 / 93 pp. / ISBN 90-71301-61-3
http://www.vssd.nl/hlf/a017.htm
Electromagnetic Waves – An Introductory Course
Electromagnetic waves appear in many forms and
their applications are extremely widespread. Without
exaggeration it may be said that our ability to employ
and manipulate electromagnetic waves forms one of
the reasons that communication plays such an
important role in society.
The macroscopic theory of electromagnetic waves
has been formulated by Maxwell in 1864. But the
mathematical-physical nature of the subject makes it
difficult for students to master even today. The
continuous stream of new college textbooks shows
that many teachers encounter this problem and
attempt to resolve it by presenting the theory in some
suitable form.
In the Electrical Engineering curriculum of the Delft
University of Technology, the teaching of
electromagnetic waves has been divided into three
stages: 1) a basic course on Electricity and Magnetism,
2) an introductory course on Electromagnetic Waves,
and 3) advanced courses on the application and
computation of electromagnetic waves. The current
book is written to facilitate the introductory course
on Electromagnetic Waves. It is assumed that students
are already acquainted with the basic phenomena
and notions of the electric and the magnetic field, and
that they know in which way Maxwell's equations
describe the electromagnetic field. Starting from
Maxwell's equations, this book deals with the
derivation of plane wave propagation, plane wave
reflection and transmission, electromagnetic rays,
waves in two-wire transmission lines, waves in planar
waveguides, and the excitation of electromagnetic
waves. As such, the aim of the book is to provide a
solid understanding of how the basic ingredients that
make up the more sophisticated applications follow
from Maxwell's equations.
The aim of the introductory course on Electromagnetic
Waves is to teach students to manipulate the
fundamental formulas in order to solve a problem at
hand. To focus on this skill and to overcome the
problem of having to learn many formulas by heart,
an outline of this book is available from the website
as a printable Repetitive Guide.
Published by VSSD.
ISBN 90-407-1836-9 EAN 9789040718366
URL on this book:

http://www.vssd.nl/hlf/e016.htm
E016

Electromagnetic Waves PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Electromagnetic Waves PDF

Uploaded by

Copyright:

Available Formats

Electromagnetic

M.D. Verweij, P.M. van den Berg, H. Blok

The course ”Electromagnetic Waves” oﬀers an introduction in the the-

that a description in terms of one-dimensional and two-dimensional waves

Delft, January 1999 H. Blok

Preface to the second edition

Delft, September 2001 M.D. Verweij

2 The Electromagnetic Field Equations 21

3 One-dimensional Electromagnetic Waves 49

3.2. Steady-state analysis . . . . . . . . . . . . . . . . . . . . . . . 54

4 Two-dimensional Electromagnetic Waves 81

5 Electromagnetic Rays in a Two-dimensional Medium 121

6 Transmission Lines 145

7 Electromagnetic Waveguides 167

8 Excitation of Two-dimensional Electromagnetic Waves 203

Answers to Exercises 223

Electromagnetic waveﬁelds have a wide range of applications: from com-

of the receiver will be neglected at all. These simpliﬁcations are dictated

length and an orientation, makes it a vectorial quantity, or a vector; nota-

Figure 1.2. Cartesian coordinates {x1 , x2 , x3 }.

The numbers {x1 , x2 , x3 } are denoted as the orthogonal Cartesian coordi-

1.1. Cartesian vectors and their properties

The mathematical framework of the theory of electromagnetic waves is

1.1.1. Addition, subtraction and multiplication of vectors

Vectors can be subjected to the algebraic operations of addition, sub-

v ± w = (v1 ± w1 )i1 + (v2 ± w2 )i2 + (v3 ± w3 )i3 . (1.2)

The product of the scalar ϕ and the vector v is given by

The scalar (dot) product of the vectors v and w is given by

The length of a vector v is denoted as

The vector (cross) product of the vectors v and w is given by

The scalar triple product of three vectors u, v and w is given by

u · (v × w) = u1 (v2 w3 −v3 w2 ) + u2 (v3 w1 −v1 w3 ) + u3 (v1 w2 −v2 w1 ) , (1.8)

The scalar triple product has the property

The vectorial triple product can be written as

u × (v × w) = (u · w)v − (u · v)w . (1.11)

As regards the diﬀerentiation of a vector, two cases have to be distin-

1.1.2. Diﬀerentiation with respect to a parameter

Let ϕ = ϕ(t) a scalar function and assume that ϕ is a diﬀerentiable

∂t v = (∂t v1 )i1 + (∂t v2 )i2 + (∂t v3 )i3 . (1.12)

Let ϕ = ϕ(t) be a diﬀerentiable scalar function of the parameter t and let

∂t (ϕ v) = (∂t ϕ)v + ϕ∂t v , (1.13)

1.1.3. Diﬀerentiation with respect to the spatial coordinates

Let ϕ be a scalar function and assume that ϕ = ϕ(x) = ϕ(x1 , x2 , x3 ) is a

∂1 v = (∂1 v1 )i1 + (∂1 v2 )i2 + (∂1 v3 )i3 . (1.17)

∂2 v = (∂2 v1 )i1 + (∂2 v2 )i2 + (∂2 v3 )i3 , (1.18)

∂3 v = (∂3 v1 )i1 + (∂3 v2 )i2 + (∂3 v3 )i3 . (1.19)

and in the curl operator, deﬁned as

We note that ∇ is a vector operator satisfying two sets of rules:

• partial diﬀerentiation rules, including diﬀerentiation of a product.

is also a twice diﬀerentiable function of x, we have the rules:

∇ · (∇ϕ) = (∇ · ∇)ϕ = (∂12 + ∂22 + ∂32 )ϕ , (1.31)

Subsequently, we present the rules for the spatial diﬀerentiation of a spatially

and when a is a constant vector:

dx = dx1 i1 + dx2 i2 + dx3 i3 . (1.44)

(∇ϕ) · dx = (∂1 ϕ)dx1 + (∂2 ϕ)dx2 + (∂3 ϕ)dx3 = dϕ , (1.45)

Figure 1.3. The gradient vector.

the change in the scalar function ϕ corresponding to a change in position

since we stay on the surface ϕ(x1 , x2 , x3 ) = constant. This shows that ∇ϕ is

When τ is the tangent along a surface ϕ = constant, we obtain

which is consistent with Eq. (1.46).

+ higher order terms . (1.50)

+ higher order terms . (1.51)

Figure 1.4. Elementary domain in three-dimensional space.