Foundations of Applied

Electromagnetics

KAMAL SARABANDI
 FOUNDATIONS
OF APPLIED
ELECTROMAGNETICS

Kamal Sarabandi
The University of Michigan
Copyright © 2022 Kamal Sarabandi

This book is published by Michigan Publishing under an agreement with the author. It is made
available free of charge in electronic form to any student orinstructor interested in the subject
matter.

Published in the United States of America by

Michigan Publishing
Manufactured in the United States of America

ISBN 978-1-60785-819-5

The Free ECE Textbook Initiative is sponsored by the ECE

Department at the University of Michigan.

The image used in the chapter title pages was taken by the James Webb Space Telescope,
courtesy of NASA.
 I dedicate this book to my family:
to my wife Shiva and our sons Arya and Sina
for their love and support,
and to the memory of my parents Abbasali and Jaleh,
who instilled in me a passion for science and engineering
iv

Preface

Field theory is one of the fundamental pillars of electrical engineering, with many threads
interwoven into numerous areas of science and technology. Classical electromagnetic theory
may be considered a mature field of science, but because of its importance in wireless
transmission of data and energy, it has remained an area of intense research and development
for almost two centuries. In recent years, the interest in the field of applied electromagnetics
has been fueled by the demand for high data-rate wireless communication, everywhere and at
any time. The implementation of such wireless systems relies on innovation in miniaturized
wideband and multiband antennas for handheld devices, vehicles, and infrastructures such
as base stations and wifi networks. In addition, knowledge of the characteristics of wave
propagation and wave interaction with terrain, vegetation, and manmade structures in urban
environments is an essential and critical tool for the proper design of wireless communication
networks. Moreover, over the past decade we have witnessed a boom in investment focused
on the rapid development of autonomous vehicles. Sensors envisioned to enable autonomous
functionality include short-range and long-range sensors with different modalities. Some of
these sensors are based on electromagnetic waves, such as millimeter-wave radars to provide
the range, direction, and velocity of objects present in traffic scenes. For these systems
to function with a high degree of reliability, we need to use applicable wave propagation
and scattering models together with highly sophisticated directional antennas, all of which
requires in-depth understanding of electromagnetic wave theory. In addition, traditional
applications of field theory in areas such as microwave remote sensing, military systems,
biomedical applications, and space exploration are active and ongoing. Graduate students
interested in such exciting fields of research need a strong foundation in field theory, and that
was my motivation for writing this book on classical electromagnetics but with an eye on its
modern applications.
This book is the outgrowth of my class notes for an entry-level graduate course on
electromagnetic theory at the University of Michigan and was inspired by my own research on
radar remote sensing, antenna theory, electromagnetic wave propagation, and more recently
on bioelectromagnetics.
Any textbook based on a field with more than 200 years of history draws very heavily from
the work of an enormous number of scientists and engineers. This book is not an exception and
I found it impossible to provide a comprehensive list of references to the original contributors
for most topics included in this book.
Publishing this book would not have been possible without the help and encouragement
of several colleagues and students. I have to first thank my dear colleague and former advisor
Professor Fawwaz Ulaby, who has carefully reviewed and provided valuable comments and
suggestions for improving what is presented in this book. I am also indebted to Mr. Richard
Carnes, who spent a significant amount of effort typing and formatting the book. I have also
to mention the contributions of my students Abdelhamid Nasr, Aditya Varma Muppala, and
Behzad Yektakhah for helping with some of the figures and computations presented in the
book.
K AMAL S ARABANDI
A NN A RBOR , AUGUST 2022
CONTENTS v

Contents

Preface iv

1 Electromagnetic Fields 1
1-1 The Field Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1-2 Faraday’s Law for a Moving Surface in a Time-Varying Magnetic Field . . . 10
1-3 Ampère’s Law for a Moving Surface in a Time-Varying Electric Field . . . . 13
1-4 Constitutive Relations: Macroscopic Properties of Matter . . . . . . . . . . . 14
1-5 Kramers-Krönig Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1-6 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1-7 Drift Current in Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1-8 Hall Effect in Conducting Media . . . . . . . . . . . . . . . . . . . . . . . . 35
1-9 Generalized Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2 Electromagnetic Concepts, Tools, and Theorems 62

2-1 Equivalent Magnetic Charge and Current Densities . . . . . . . . . . . . . . 63
2-2 Image Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2-3 Method of Images for Other Problems . . . . . . . . . . . . . . . . . . . . . 71
2-4 Polarization Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2-5 Stored Electromagnetic Energy . . . . . . . . . . . . . . . . . . . . . . . . . 83
2-6 Flow of Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
2-7 Superposition Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2-8 Uniqueness Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2-9 Equivalence Principle for Electromagnetic Sources . . . . . . . . . . . . . . 92

3 The Electromagnetic Potentials and Radiation 107

3-1 Electromagnetic Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3-2 Time-Harmonic Electromagnetic Waves . . . . . . . . . . . . . . . . . . . . 122
3-3 Time-Harmonic Retarded Potential . . . . . . . . . . . . . . . . . . . . . . . 131
3-4 Far-Field Distance Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3-5 Small Loop of Current: A Hertzian Magnetic Dipole . . . . . . . . . . . . . 139
3-6 Wire Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
3-7 Equivalent Circuit for Receiving Antennas . . . . . . . . . . . . . . . . . . . 147
vi CONTENTS

4 Formal Solutions to Maxwell’s Equations and Their Applications 161

4-1 Formal Solution of the Helmholtz Equation . . . . . . . . . . . . . . . . . . 162
4-2 Solution of the Helmholtz Equation for a Complex Medium . . . . . . . . . . 168
4-3 Integral Equations for Electromagnetic Fields . . . . . . . . . . . . . . . . . 176
4-4 Integral Equation Formulation Based on Equivalent Sources . . . . . . . . . 178
4-5 Integral Equation Formulation for Dielectric Scatterers . . . . . . . . . . . . 181
4-6 Reciprocity Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
4-7 Applications of the Reciprocity Theorem . . . . . . . . . . . . . . . . . . . . 189
4-8 Babinet’s Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

5 Electromagnetic Plane Waves 227

5-1 Plane-Wave Propagation in Homogeneous Media . . . . . . . . . . . . . . . 228
5-2 Polarization of Plane Waves . . . . . . . . . . . . . . . . . . . . . . . . . . 238
5-3 kDB Coordinate for Plane Waves in Bianisotropic Media . . . . . . . . . . . 247
5-4 Transverse Electric (TE) and Transverse Magnetic (TM) Field Solutions of
the Helmholtz Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
5-5 Plane-Wave Reflection at the Interface between a Dielectric Medium and a
Good-Conducting Medium . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
5-6 Wave Propagation in an Inhomogeneous Medium: Geometric-Optics Approx-
imation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
5-7 Plane-Wave Reflection and Transmission from a Half-Space Uniaxial Medium 275
5-8 Plane Waves in Layered Media . . . . . . . . . . . . . . . . . . . . . . . . . 280
5-9 Plane-Wave Propagation in a Negative-Index Medium . . . . . . . . . . . . . 290
5-10 Negative Refractive-Index Lens . . . . . . . . . . . . . . . . . . . . . . . . 292

6 Cartesian Wave Functions: Guiding Structures and Resonators 313

6-1 The Dielectric Plate Waveguide . . . . . . . . . . . . . . . . . . . . . . . . . 315
6-2 Guided Waves on Impedance Surfaces . . . . . . . . . . . . . . . . . . . . . 321
6-3 Practical Realization of Reactive Impedance Surfaces . . . . . . . . . . . . . 324
6-4 Isotropic Reactive Impedance Surfaces . . . . . . . . . . . . . . . . . . . . . 330
6-5 The Rectangular Waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . 335
6-6 Transmission-Line Circuit Model for Waveguides . . . . . . . . . . . . . . . 357
6-7 Other Modal Solutions for Rectangular Waveguides . . . . . . . . . . . . . . 363
6-8 Modal Expansion of Field Quantities . . . . . . . . . . . . . . . . . . . . . . 367
6-9 Calculus of Variations for Estimation of Resonant Frequencies in General
Cavities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382

7 Cylindrical Wave Functions and Their Applications 404

7-1 Wave Functions in the Cylindrical Coordinate System . . . . . . . . . . . . . 406
7-2 The Circular Dielectric Waveguide . . . . . . . . . . . . . . . . . . . . . . . 421
7-3 Green’s Functions Solutions for Some Canonical Problems . . . . . . . . . . 436
7-4 Scattering from a Metallic Circular Cylinder . . . . . . . . . . . . . . . . . . 452
7-5 Integral Representation of Bessel Functions . . . . . . . . . . . . . . . . . . 460
7-6 2-D Green’s Function for Homogeneous Media in the Presence of a Metallic
Wedge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
CONTENTS vii

7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic Wedge . . . . . . . 474

8 Spherical Wave Functions and Their Applications 502

8-1 Wave Functions in the Spherical Coordinate System . . . . . . . . . . . . . . 504
8-2 Wave Transformation to Spherical Wave Functions . . . . . . . . . . . . . . 533
8-3 Multipole Representation of Spherical Waves . . . . . . . . . . . . . . . . . 538
8-4 Plane-Wave Scattering from Spheres . . . . . . . . . . . . . . . . . . . . . . 542
8-5 Wave Propagation in a Conical Waveguide . . . . . . . . . . . . . . . . . . . 553
8-6 Biconical Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 560
8-7 Other Spherical Waveguides . . . . . . . . . . . . . . . . . . . . . . . . . . 568

A Properties of Complex Functions 586

A-1 Cauchy–Riemann Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 587
A-2 Conformal Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587
A-3 Branch Cut and Branch Point . . . . . . . . . . . . . . . . . . . . . . . . . . 588
A-4 Cauchy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
A-5 Cauchy Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 590
A-6 Poles and Residues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 590
A-7 Jordan’s Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591

B Method of Steepest Descent 592

B-1 Saddle Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593
B-2 Integration along the Steepest Descent Path . . . . . . . . . . . . . . . . . . 594

C Useful Vector Identities, Operators, and Coordinate Transformations 596

Index 601
 Chapter 1
Electromagnetic Fields

Chapter Contents Objectives

Overview, 2 Upon learning the material presented in this chapter, you

1-1 The Field Equations, 2 should be able to:
1-2 Faraday’s Law for a Moving Surface
in a Time-Varying Magnetic Field, 10 1. Explain the basic physics behind Maxwell’s
1-3 Ampère’s Law for a Moving Surface equations.
in a Time-Varying Electric Field, 13
1-4 Constitutive Relations: Macroscopic 2. Understand the relations between the field
Properties of Matter, 14 intensities and the flux densities in a material
1-5 Kramers-Krönig Relations, 28 medium (known as the constitutive relations)
1-6 Boundary Conditions, 30 and their behavior as a function of time and
1-7 Drift Current in Metals, 34 frequency.
1-8 Hall Effect in Conducting Media, 35
1-9 Generalized Coordinates, 36 3. Incorporate complexities associated with field
Chapter Summary, 44 discontinuities across abrupt boundaries between
Problems, 46 dissimilar media.

4. Express the expanded form of Maxwell’s equa-

tions and relevant differential operators in both
standard and arbitrary coordinate systems.

1
2 Chapter 1 Electromagnetic Fields

Overview
All material media—from an isolated atom to an entire galaxy of stars and planets—radiate
electromagnetic waves, all the time! We are constantly getting bombarded by a spectrum
of EM waves, including light waves, microwaves, and radio waves. Electromagnetics, the
bidirectional interaction between the electric and magnetic fields, is at the core of what makes
electronic circuits function, communication systems transfer data, and computer systems
process information. This book is about how electromagnetic fields interact with material
media, including reflection by and refraction across boundaries between electromagnetically
dissimilar media, transmission across boundaries, absorption by lossy media, and scattering
within inhomogeneous media.
Chapter 1 starts with an examination of the four famous equations known as Maxwell’s
equations, in both differential and integral forms. Maxwell’s equations are then formulated
for the condition when the electric and magnetic fields are in the presence of a time-varying
(moving) surface, and also used to define the constitutive properties of material media, namely
the electrical permittivity and magnetic permeability, in both homogeneous and anisotropic
materials. These foundational formulations will serve to facilitate the treatments of the various
EM-related topics covered in forthcoming chapters.

1-1 The Field Equations

Electromagnetism, as the structure of the word implies, encompasses certain laws of physics
that define the interrelationships between the electric and magnetic fields. Electromagnetic
phenomena are observed only when the two field quantities are time-varying. The nature of
the interaction between the time-varying electric and magnetic fields was first discovered by
Michael Faraday (1791–1867) and later formulated into mathematical expressions by James
Clerk Maxwell (1831–1879). Faraday’s significant discovery was based on experimental
observations and Maxwell’s formulation was based on mathematical deduction. Faraday’s
extensive experimental work was motivated by the belief that every cause and effect has its
converse. That is, if electricity can produce a magnetic field, a phenomenon discovered by
Oersted, then, conversely, a magnetic field should be able to produce an electric field.
The fundamental laws of electricity and magnetism are encapsulated by Maxwell’s
equations:*
∂D
∇×H = J+ (modified Ampère’s law), (1.1a)
∂t
∂B
∇×E = − (Faraday’s law), (1.1b)
∂t
∇·D = ρ (Gauss’ law for electricity), (1.1c)
∇ · B = 0 (Gauss’ law for magnetism). (1.1d)

* James Clerk Maxwell, A Treatise on Electricity and Magnetism, Constable and Co., London, 1873.
1-1 The Field Equations 3

where

H = H(r,t) (magnetic field intensity, A/m),

E = E(r,t) (electric field intensity, V/m),
D = D(r,t) (electric flux density, C/m2 ),
B = B(r,t) (magnetic flux density, W/m2 ),
ρ = ρ (r,t) (volumetric charge density, C/m3 ),
J = J(r,t) (current density, A/m2 ),

and r is the position vector for an ordinary point in the medium. Here ordinary point refers
to a point wherein within its immediate neighborhood the physical properties of the medium
are continuous. In other words, the small medium around r is considered to be homogeneous.
If the physical properties of a medium change abruptly as a function of location, the vector
field quantities may also do the same. That is, the transition of field vectors across a surface
where change in material properties is abrupt may be discontinuous. The nature of these
discontinuities will be investigated in detail in later sections.
Equations (1.1a) and (1.1b), in addition to the law of conservation of charge, constitute
the necessary and sufficient set of equations required for determining the field quantities.
The phenomena of electricity and magnetism are governed by the presence of static
and moving charges. Basically, charges are considered the source of electromagnetic fields
without which the field quantities (E, H, D, B) cannot exist.

1-1.1 Charge Density

The quantized nature of charges is well established. A charge quantum is equal to the absolute
value of the charge of an electron, namely

electron charge e = −1.6 × 10−19 C .

However, Maxwell’s equations describe large-scale phenomena, i.e., the macroscopic element
of volume must contain a large number of atoms and molecules, in which case a macroscopic
particle can contain any real number of charges. If the amount of charge contained in volume
element ∆ V is ∆ q, then the volumetric charge density is defined as
∆q
ρ = lim . (1.2)
∆ V →0 ∆ V

In the strictest sense, Eq. (1.2) does not define a continuous function of position because ∆ V
cannot approach zero without limit.

1-1.2 Current Density

From a macroscopic point of view, any ordered motion of charge constitutes a current.
Hence, a current is represented by a vector quantity whose direction is the direction of
4 Chapter 1 Electromagnetic Fields

|J| = density of streamlines J(r,t)

J(r,t)
= tangent to the streamlines Streamlines of current traced
|J|
by the movement of charges

Figure 1-1: Streamlines of electric current in a medium, where the intensity of the current
density is represented by the density of the streamlines and the unit vector tangent to the lines
specifies the direction of flow.

motion of the charges and its magnitude is proportional to the velocity and number density
of the charges. To define a current vector more precisely, we do so in terms of the current
density (distribution) defined over the region of space under consideration. Current density
(distribution) is characterized by a vector field J, as depicted in Fig. 1-1. To better quantify
current density, let us consider a differential surface ∆ S whose unit normal is denoted by n̂,
as shown in Fig. 1-2. If ∆ I represents the total current crossing the differential area ∆ S, then
J is defined such that
∆I = J·∆S . (1.3)

u
J

∆ S = ∆ S û
Figure 1-2: The diagram depicts a surface S intersecting streamlines of an electric current
flowing through that surface. The current density J is defined as the normalized current flowing
through a differential surface. Here, û denotes the direction of the velocity vector associated
with J.
1-1 The Field Equations 5

By extension, the total current crossing an arbitrary surface S can be computed from
Z
I = J · ds . (1.4)
S

1-1.3 Point Form of Law of Conservation of Charge

The law of conservation of charge states that the net value of charge in a closed system remains
constant. This means that if there is a certain number of positive charges and a certain number
of negative charges in an enclosed medium, nothing can be done to create an excess amount
of either kind of charge nor to annihilate only one type of charge. To change the value of net
charge in the closed system, charges will have to be either removed from or brought into the
system.
Now suppose the surface S is a closed surface. Define n̂ as a unit normal to the surface
drawn in the outward direction. According to the law of conservation of charges,
$
d d
I= J · ds = − Q = − ρ dυ , (1.5)
S dt dt V

where Q is the total charge enclosed and V is the volume enclosed by S. If the surface is
stationary, then $
∂ρ
J · ds = − dυ . (1.6)
∂t
V
According to the divergence theorem,
$
J · ds = ∇ · J dυ , (1.7)
V

so that in the limit where V is very small, denoted as ∆ V ,

$
∇ · J dυ = ∇ · J ∆V .
∆V

This equivalence leads to the following definition for the divergence operation:

∆ 1
∇ · J = lim J · ds . (1.8)
∆ V →0 ∆ V

Using Eq. (1.7) in Eq. (1.6) leads to the following relation for volume V :
$ $
∂ρ
∇ · J dυ = − dυ . (1.9)
∂t
V V
6 Chapter 1 Electromagnetic Fields

Since Eq. (1.9) has to be valid for any arbitrary volume, it follows that

∂ρ
∇·J = − , (1.10)
∂t

which is known as the “law of conservation of charge in the neighborhood of a point,” and
also known as the equation of continuity.

1-1.4 Interdependence of Maxwell’s Equations

Equations (1.1c) and (1.1d) of Maxwell’s equations (Gauss’ laws) are not independent.
Noting that for any vector A,

∇·∇×A = 0 , ∀A (1.11)

and then applying the divergence operator to both sides of Eq. (1.1a) leads to

∂
∇·J+ ∇·D = 0 . (1.12)
∂t
Substituting Eq. (1.10) into Eq. (1.12) gives

∂
(∇ · D − ρ ) = 0 ,
∂t
which implies that (∇ · D − ρ ) must be a constant. It makes sense for this constant to be zero
as we do not expect a nonzero electric flux to exist at any point in the medium in the absence
of charges. Hence,
∇·D = ρ , (1.13)
which is Gauss’ law for electricity.
Also, from Eq. (1.1b), applying the divergence operator to both sides and then using
Eq. (1.11) leads to
∂
∇·B = 0 ,
∂t
from which it can be concluded that
∇·B = 0 . (1.14)
In summary, Faraday’s law, the modified form of Ampère’s law, and the equation of
continuity constitute a sufficient set of equations to characterize both the electric and magnetic
field quantities completely.

1-1.5 Integral Form of Maxwell’s Equations

To develop the integral-form equivalents of the four Maxwell equations defined by
Eqs. (1.1a)–(1.1d), we need to use the basic definition of the curl and divergence operators.
The curl of a vector field quantity A whose components and their first derivatives are
1-1 The Field Equations 7

continuous is defined by I
A · dℓℓ
(∇ × A) · n̂ = lim (1.15)
∆s
∆ s→0

All three components of ∇ × A can be obtained once n̂ is aligned with the coordinate unit
vectors. Here, it is important to note that the directions of n̂ and the differential length dℓℓ are
chosen so that they follow the right-hand rule. (That is, if dℓℓ is along the right-hand fingers,
then n̂ would be along the right-hand thumb.)
Stokes’ theorem is a natural extension of the curl definition. For a regular closed contour C
and any arbitrary surface S bounded by C over which the components of A and their first
derivative are continuous, Stokes’ theorem states that
" I
∇ × A · ds = A · dℓℓ . (1.16)
C
S

Upon applying a surface integral to both sides of Eq. (1.1a) and then incorporating Eq. (1.16)
into the result, it can be easily shown that
I "  
∂D
H · dℓℓ = J+ · ds . (1.17)
C ∂t
S

Here, ∂ D/∂ t has the unit of A/m2 and is also known as the displacement current. If S is
stationary, then I "
d
H · dℓℓ = I + D · ds , (Ampère’s law) (1.18)
C dt
S
!
where I = J · ds is the total conduction current passing through the surface defined by C.
In fact, Eq. (1.18) is the physical formulation from which Eq. (1.1a) is derived. This
formulation also allows for the surface S and contour C themselves to be time-varying. In a
similar manner the integral form of Faraday’s law can be obtained and is given by
I "
d
E · dℓℓ = − B · ds , (Faraday’s law) (1.19)
C dt
S

which states that the induced voltage around a closed contour is equal to the negative time
rate of change of the flux linking the surface.
Applying a volume integral to both sides of Eq. (1.1c) leads to
$ $
∇ · D dυ = ρ dυ = Q . (1.20)
V V
8 Chapter 1 Electromagnetic Fields

Next, application of the divergence theorem gives:

D · ds = Q . (Gauss’ law for electricity) (1.21)

Similarly it can be shown that

B · ds = 0 , (Gauss’ law for magnetism) (1.22)

which can be interpreted as “the flux lines of the magnetic flux density are continuous.”
In summary, Maxwell’s equations in integral form are given by

I "
d
H · dℓℓ = I + D · ds (Ampère’s law), (1.23a)
C dt
S
I "
d
E · dℓℓ = − B · ds (Faraday’s law), (1.23b)
C dt
S

D · ds = Q (Gauss’ law for electricity), (1.23c)

B · ds = 0 (Gauss’ law for magnetism). (1.23d)

Example 1-1: Deriving Circuit Laws from Maxwell’s Equations

Show that the well-known circuit laws can be derived from Maxwell’s equations under a
quasi-static condition, which refers to a condition where the time rates of change of field
quantities are very slow (compared with f d/c, where d is a typical dimension of the circuit
or circuit element, f is the frequency, and c is the speed of light).
Solution: Kirchhoff’s current law at an electrical node or junction in a circuit, which states
that the sum of electric currents entering a node should add up to zero, is a direct result of
the law of conservation of charge. We can use Faraday’s law to prove Kirchhoff’s voltage
law under quasi-static conditions. Starting from the integral form of Faraday’s law, as given
by Eq. (1.23b), we note that the magnitude of the magnetic flux density is many orders of
magnitude smaller than the magnitude of the electric field E (which will be shown later). We
also note that under the quasi-static condition the time rate of change of the linking flux that
appears on the right-hand side is very small. Hence, it is concluded that
I
E · dℓℓ = 0 .
C
1-1 The Field Equations 9

This equation indicates that the sum of voltage drops over any closed loop in the circuit must
add up to zero, which is the statement of Kirchhoff’s voltage law.

Example 1-2: Coulomb’s Law of Electricity

Show that Coulomb’s law of electricity is not independent of Maxwell’s equations.

Solution: Coulomb’s law was derived experimentally by Charles-Augustine de Coulomb in
1785. This law provides the formula for the force on a test charge created by another charge
and is considered the first law of electricity. It is important to realize that this law is not
independent of Maxwell’s equations, and that in fact it can be derived from Gauss’ law of
electricity. To show this, let us consider a point charge q centered at the origin of a Cartesian
system in a medium with permittivity ε , as shown in Fig. 1-3. At an observation point r on
the surface of sphere S, the electric flux density D, according to the integral form of Gauss’
law, must satisfy
D · ds = q .

Because of the spherical symmetry of the problem, D = Dr r̂ and Dr is constant on S, which

leads to
4π r2 Dr = q ,

z
D

ds
r
q
y

Figure 1-3: A point charge q at the origin creates a uniform electric flux density D on the surface
of a sphere S with radius r.
10 Chapter 1 Electromagnetic Fields

from which the electric field of a point charge, consistent with Coulomb’s law, is obtained:

q
E= r̂ .
4πε r2

1-2 Faraday’s Law for a Moving Surface in a Time-Varying

Magnetic Field

Figure 1-4 displays a moving surface S at two points in time, S(t) at time t and the same
surface at an incrementally later time (t + ∆ t), denoted S(t + ∆ t). The surface exists in a
medium containing a time-varying magnetic flux density B(r,t). As we will demonstrate in
this section, the configuration in Fig. 1-4 induces an induction voltage difference Vi composed
of two components, a transformer induction component Vit due to the time-varying magnetic
flux density and a motional induction component Vim due to the moving surface.
The integral of the electric field intensity over a closed contour on the left-hand side of
Eq. (1.19) is equivalent to a voltage. Hence, the right-hand side of the equation represents a

C2 = C(t + Δt)

Δll1 =
u Δt
dll S1 = S(t)
C1 = C(t)
ΔS

C2 = C(t + Δt)

Δll1 =
u Δt S2 = S(t + Δt)

dll
C1 = C(t)
ΔS

Figure 1-4: A moving surface in a magnetic field B.

1-2 Faraday’s Law for a Moving Surface in a Time-Varying Magnetic Field 11

voltage quantity, which we denote by Vi . Thus,

I
E · dℓℓ = −Vei , (1.24)
C

with "
d
Vei = B · ds . (1.25)
dt S(t)

Next, we use the basic definition of the time derivative to express Vi as a sum of two terms:
" " !
1
Vei = lim B(t + ∆ t) · ds − B(t) · ds . (1.26)
∆ t→0 ∆ t S(t+∆ t) S(t)

Noting that for small ∆ t

∂ B(t)
B(t + ∆ t) ≈ B(t) + ∆t ,
∂t
Eq. (1.26) becomes
""   " #
1 ∂ B(t)
Vei = lim B(t) + ∆ t · ds − B(t) · ds , (1.27)
∆ t→0 ∆ t S(t+∆ t) ∂t S(t)

which can be rearranged into two components:

Vei = Vit +Vim , (1.28)

with "   "

1 ∂B ∂B
Veit = lim ∆ t · ds = · ds (1.29)
∆ t→0 ∆ t S(t+∆ t) ∂t S(t) ∂ t

and (" " )

1
Veim = lim B(t) · ds − B(t) · ds . (1.30)
∆ t→0 ∆ t S(t+∆ t) S(t)

For the moving surface S in Fig. 1-4, every point on the contour moves a distance

dℓℓ1 = u ∆ t , (1.31)

where u is the local velocity vector at that point. As the contour moves, it traces a surface on
its side. We denote this side surface as ∆ S. Next we denote the enclosed surface comprising
the top surface S(t), the bottom surface S(t + ∆ T ), and the side surface ∆ S as S0 . Hence,
Eq. (1.30) can be rewritten as
 " 
1
Veim = lim B(t) · ds − B(t) · ds . (1.32)
∆ t→0 ∆ t S0 ∆S
12 Chapter 1 Electromagnetic Fields

The differential surface ds for the side surface ∆ S is given by

ds = dℓℓ × ∆ ℓ 1 = dℓℓ × u ∆ t ,

which leads to
" I
1
lim B(t) · ds = B(t) · dℓℓ × u(t)
∆ t→0 ∆ t ∆s C
I
= [u(t) × B(t)] · dℓℓ . (1.33)
C

Also,
$
1 1
lim B · ds = lim ∇ · B dυ
∆ t→0 ∆ t S0 ∆ t→0 ∆ t
" "
1
= lim (∇ · B)u ∆ t · ds = (∇ · B)u · ds . (1.34)
∆ t→0 ∆ t

Since ∇ · B = 0, the integral term given by Eq. (1.34) vanishes, and Eq. (1.28) simplifies to

I " I
∂B
E · dℓℓ = − · ds + (u × B) · dℓℓ . (1.35)
C S ∂t C

The first term is referred to as the transformer induction voltage Veit , and the second term is
known as the motional induction voltage Veim .
It is very important to note that in order to arrive at a nonvanishing motional induction
term, the contour of the surface S has to be moving. That is, if the contour is fixed and S(t)
is still a function of time (such as a fluctuating surface), there will be no motional induction.
Physically the contour of the integral (circuit) has to “cut” the flux lines of the magnetic flux
density in order to generate “electromotive potential.”
This phenomenon can be explained physically using the Lorentz force F on a charged
particle q:
F = qE + qu × B . (1.36)

Since the electric field E is defined in terms of force per unit charge, the equivalent induced
electric field due to moving charges can be defined as
Femf
Eemf = = u×B
q
and
I I
Vemf = Eemf · dℓℓ = (u × B) · dℓℓ . (1.37)

Equation (1.37) is consistent with Eq. (1.35). This equivalence indicates that the Lorentz force
1-3 Ampère’s Law for a Moving Surface in a Time-Varying Electric Field 13

F F

++
u + u
+q ++

u
Vemf

u
__
B
_ __

(a) (b)

Figure 1-5: (a) A positive charge q moving in a magnetic field B, and (b) a metallic rod
containing many charges moving in a magnetic field, which produces an electromotive voltage
Vemf across the bar.

equation is not independent of Maxwell’s equations.

Figure 1-5(a) depicts the Lorentz force on a moving charge in a magnetic field, and
Fig. 1-5(b) shows a conductor moving in the same magnetic field. The Lorentz force acts
on free electrons in the conductor and pushes them downward as shown in the figure. Note
that the movement of charges in the conductor is confined to within the conductor itself. For
a static magnetic field, a dc voltage difference between the two ends of the conductor builds
up, and can be computed using Eq. (1.37).
In a medium with distributions of charges and currents, Eq. (1.36) can be used to
determine the force per unit volume. For a differential volume ∆ V in a medium with charge
density ρ and current density J, the charge in ∆ V is ∆ q = ρ ∆ V and J = ρ u. The force per
unit volume can be calculated from
F
f= = ρE + J × B . (1.38)
∆V

1-3 Ampère’s Law for a Moving Surface in a Time-Varying

Electric Field
The sequence of steps that were applied to Faraday’s law in the preceding section can also be
applied to the integral form of Ampère’s law defined by Eq. (1.23a). The process leads to
I " " I
∂D
H · dℓℓ = I + · ds + (∇ · D)u · ds − (u × D) · dℓℓ .
S ∂t S C
14 Chapter 1 Electromagnetic Fields

Noting from Eq. (1.13) that ∇ · D = ρ , the above expression becomes

I " " I
∂D
H · dℓℓ = I + · ds + ρ u · ds − (u × D) · dℓℓ , (1.39)
C S ∂t S C

in which I represents the conduction current, ρ u represents the drift current, ∂ D/∂ t
represents the displacement current, and (u × D) represents the motional current.
Before proceeding to the next section, let us pose the following question: If we were to
consider Eq. (1.35) as the correct form of Faraday’s law, how would it reduce to its differential
form given by Eq. (1.1b)? To answer the question, let us rewrite Eq. (1.35) as
I "
∂B
[E − u × B] · dℓℓ = − · ds . (1.40)
C S ∂t
Now if we apply Stokes’ theorem, as given by Eq. (1.16), to the left-hand side to convert the
contour integral into a surface integral, the process leads to
"  
∂B
∇ × (E − u × B) + · ds = 0 . (1.41)
S ∂t
Since S is arbitrary, the integrand must vanish. Hence,

∂B
∇ × E − ∇ × (u × B) = − . (1.42)
∂t
As will be shown later, the magnetic flux density B is many orders of magnitude smaller than
the electric field intensity E, and unless u is close to the speed of light, the second term in
Eq. (1.42) can be ignored compared with the first term. As a result, the point form of Faraday’s
law reduces to
∂B
∇×E = − . (1.43)
∂t

1-4 Constitutive Relations: Macroscopic Properties of Matter

As was shown previously in Section 1-1.4, out of the four Maxwell’s equations only the
modified Ampère’s and Faraday’s laws (Eqs. (1.1a) and (1.1b)) are truly independent.
Considering the fact that it is the charges and currents that are the sources of electromagnetic
fields E, D and H, B, we need to impose two additional constraints, in addition to the two
independent Maxwell’s equations, in order to ensure that the overall system of equations is
indeed determinate.
In the most general case, we may consider the relationship among the four field quantities
to be of the form

D = D(E, H) (1.44a)
and
B = B(E, H) , (1.44b)
1-4 Constitutive Relations: Macroscopic Properties of Matter 15

where D(·) and B(·) are some general vector functions dependent upon the material in which
the field vector quantities are established.
In the treatment of Maxwell’s equations considered henceforth, we shall confine our
attention to only the small-signal condition wherein the constitutive relations given by
Eqs. (1.44a) and (1.44b) are linear. That is, if D1 and B1 are induced by E1 and H1 , and D2
and B2 are induced by E2 and H2 , then D1 + D2 and B1 + B2 are induced by the combination
of E1 + E2 and H1 + H2 .
A material medium is considered electromagnetically homogeneous if the constants of
proportionality relating D and B to E and H are independent of position within the medium.
Conversely, the medium is considered inhomogeneous if the constants of proportionality are
position-dependent.
In the absence of any material (vacuum), the constitutive relations are very simple and are
given by
D = ε0 E , B = µ0 H , (1.45)
where

ε0 = 8.85 × 10−12 (farad/m) (free-space permittivity)

and
µ0 = 4π × 10−7 (henry/m) (free-space permeability).

Note that the unit for the D is farad/m × volt/m = C/m2 . Similarly, the unit for the magnetic
flux density is henry/m × ampere/m = W/m2 (magnetic flux per unit area).
The next simplest pair of constitutive relations belongs to materials known as isotropic
homogeneous materials. If the physical properties of the medium are the same along all
directions (as seen by E and H) the medium is considered isotropic. Hence D is parallel to E
and B is parallel to H at every location within the medium:

D = εE and B = µH , (1.46)

where ε = εr ε0 and µ = µr µ0 .

εr = relative permittivity (dimensionless)

and
µr = relative permeability (dimensionless).

Usually εr and µr are quantities larger than unity. In a homogeneous isotropic material,
the applied electric field slightly polarizes the atoms or molecules in a direction parallel to
the direction of the applied electric field. That is, the center of mass of the electron cloud
around the nucleus shifts slightly in a direction opposite to that of the applied electric field,
thereby forming a small electric dipole. The electric field formed by these small dipoles are
in the opposite direction to that of the applied field, as shown in Fig. 1-6, and therefore the
total electric field in the medium is smaller than the applied electric field. In the presence of
external electric and magnetic fields, the material gets polarized and magnetized, respectively.
16 Chapter 1 Electromagnetic Fields

S S S
External B
N N N
magnetic field
H Total magnetic field
S S S in the medium
N N N
Magnetized molecule

__ __ __
External D
++ ++ ++
electric field
E Total electric field
__ __ __ in the medium
++ ++ ++

Polarized molecule

Figure 1-6: Magnetized and polarized molecules in the presence of external applied electric and
magnetic fields. The total field in such a medium is weaker than the external applied field in the
absence of the material.

The in a medium can be written as

D = ε0 E + P , (1.47)

where P is known as the polarization vector and constitutes a dipole moment per unit volume.
If the electric field intensity is not very high, the polarization vector is related linearly to the
electric field by P = ε0 χe E, where χe is the electric susceptibility of the medium. Hence, D
can be expressed as
D = εE , (1.48a)
with
ε = ε0 (1 + χe ) .

In a similar manner the magnetic flux density in a magnetic medium, under linearity
conditions, can be written as
B = µ0 H + µ0 M , (1.49)
with M = χm H, where χm is the magnetic susceptibility of the medium. The quantity M is
known as the magnetization vector and provides a measure of the induced magnetic dipole
moment per unit volume. The manifestation of magnetic dipoles in a medium can be attributed
to the electron orbital motion, as well as electron or nuclear spin. An unbalanced net effect
of such equivalent atomic currents flowing on a closed loop can give rise to magnetic dipoles
within the material. Such an equivalent atomic current loop is characterized by its magnetic
moment m. The number of these magnetic moments per unit volume (N) is the magnetization
1-4 Constitutive Relations: Macroscopic Properties of Matter 17

vector M = Nm. From Eq. (1.49) we can show that

B = µH (1.50a)

and
µ = µ0 (1 + χm ) . (1.50b)

In general, magnetic materials can be categorized as:

1. Paramagnetic (µr > 1), which refers to magnetic materials with relatively small
magnetic dipoles and which also do not retain permanent magnetism.

2. Diamagnetic (µr < 1), which refers to magnetic materials with induced magnetic
moments that are parallel to but in the opposite direction to the applied magnetic field.
Such materials are repelled by a permanent magnet. Most materials that are known as
nonmagnetic, such as water or gold, are in reality diamagnetic but with a very weak
response. For such materials, the magnetic susceptibility is a small negative number.

3. Ferromagnetic (µr ≫ 1), which refers to spontaneous magnetization in subdomains.

Such materials are highly nonlinear and are characterized by hysteresis (a time-varying
phenomenon related to material memory). Above its Curie temperature, the material
becomes paramagnetic.

The phenomenon of ferroelectricity has been observed for a variety of materials; for
example, barium titanate (BaTiO3 ) shows a ferroelectric behavior marked by hysteresis,
nonlinearity, and very large values for εr .

1-4.1 Anisotropic Media

For certain materials, their electrical and/or magnetic properties may behave differently along
different directions. In such media, vector B may not be parallel to vector H, and vector D
may not be parallel to vector E. Consequently, the relationships between D and E and between
B and H should be expressed in terms of permittivity and permeability tensors:
 
εxx εxy εxz
D = ε ·E , ε = ε0 ε r = ε0 εyx εyy εyz  (1.51a)
εzx εzy εzz
and
 
µxx µxy µxz
B = µ ·H , µ = µ0 µ r = µ0 µyx µyy µyz  . (1.51b)
µzx µzy µzz

A material is called anisotropic if either or both of its permittivity and permeability is/are
tensor quantities. It should be noted that the tensor elements are functions of the coordinate
system. For materials with axes of symmetry, the permittivity or permeability tensors become
symmetric matrices. In that case the matrix (matrices) is (are) diagonizable with real
18 Chapter 1 Electromagnetic Fields

eigenvalues and eigenvectors that are orthogonal. By rotating the Cartesian coordinate system
to align it with the eigenvectors’ directions, the tensor becomes diagonal:
 
εx 0 0
ε = ε0  0 εy 0  (biaxial material). (1.52)
0 0 εz

Such a material is known as . If there is invariance in any coordinate plane, two of the entries
of Eq. (1.52) become equal (say εx = εy , εz ). The material is then called uniaxial.

1-4.2 Bianisotropic Media

A bianisotropic material exhibits magnetization when subjected to an externally applied
electric field and exhibits polarization when subjected to an externally applied magnetic field,
and these cross-field components are in addition to the direct magnetization and polarization
components caused by the external fields. The constitutive relationships for these materials
are given by

D = ε ·E+ζ ·H (1.53a)

and

B = ξ ·E+µ ·H . (1.53b)

1-4.3 Dispersive Materials

Except for free space, for which neither the permittivity nor the permeability are functions of
either time t or the oscillation frequency ω , all other media, strictly speaking, are frequency-
dependent. This is due to the fact that all charges that interact with the field quantities have
finite masses, which leads to a time delay between the timing of the applied external fields
and the formation of the polarization and magnetization vectors.
Under the small-signal approximation (linear), the polarization vector (dipole moment per
unit volume) is linearly proportional to the applied electric field. For an isotropic medium,

P = ε0 · χe E , (1.54)

where χe is called the electric susceptibility. If χe is assumed to be a real constant and

independent of frequency, the relationship given by Eq. (1.54) essentially ignores the delay
between E and P. The correct way of thinking about a linear relationship between P and E is
in terms of a time-domain convolution. If χe (t) represents the true impulse response, then
Z t
P(t) = ε0 χe (t − τ ) E(τ ) d τ . (1.55)
−∞
1-4 Constitutive Relations: Macroscopic Properties of Matter 19

The integral has been truncated at t by imposing the causality condition; i.e., χe (t) = 0 for
t < 0. Hence,
Z t
D(t) = ε0 E(t) + ε0 χe (t − τ ) E(τ ) d τ
−∞
Z ∞
= ε0 E(t) + ε0 χe (τ ) E(t − τ ) d τ . (1.56)
0

Taking the Fourier transform of both sides of Eq. (1.56) gives

Z +∞ Z +∞  Z ∞ 
iω t
D(t) e dt = ε0 E(t) + χe (τ ) E(t − τ ) d τ eiω t dt ,
−∞ −∞ 0

which is equivalent to
Z ∞ Z +∞
iωτ
D(ω ) = ε0 E(ω ) + ε0 χe (τ ) e E(t − τ ) eiω (t−τ ) dt d τ
0
| −∞ {z }
E(ω )

and which leads to  Z ∞ 

iωτ
D(ω ) = ε0 1 + χe (τ ) e dτ E(ω ) . (1.57)
0
Therefore, the complex permittivity is given by
 Z ∞ 
iωτ
ε (ω ) = ε0 1 + χe (τ ) e d τ = ε0 [1 + χe (ω )] . (1.58)
0

According to Eq. (1.58), the permittivity of any material (excluding free space) is, in general,
a complex function of frequency. If a narrow pulse were to propagate in such a medium, it
can be shown that the pulse would spread in time and space as it travels in the medium.
A similar behavior can be shown to be true for magnetic materials:

M(ω ) = χm (ω ) H(ω ) . (1.59)

The dependence of the permittivity ε (ω ) and the permeability µ (ω ) on the frequency ω

may vary significantly among different materials. For most nonpolar materials, ε and µ
can be approximated by constant quantities at microwave and millimeter-wave parts of the
spectrum. Near molecular resonances, however, the variations of the constitutive parameters
with frequency must be taken into account carefully.

1-4.4 Conducting Media

Another element of the constitutive relations is the relation between the current density J and
the electric field in the medium E. In the absence of magnetic fields, a conducting medium is
characterized by
J = σE (Ohm’s law), (1.60)
20 Chapter 1 Electromagnetic Fields

where J is the conduction current (to be differentiated from any externally impressed source
current), E is the electric field in the medium, and σ is the conductivity of the medium.
Equation (1.60) is known as the point form of Ohm’s law. In the presence of a magnetic field,
the direction of the conduction current is no longer parallel to E. This phenomenon is known
as the Hall effect, which will be revisited later in this chapter.
Within a conducting medium (σ > 0) there can be no accumulation of charges. If an initial
charge density ρ0 (r) is established in a conducting medium, then according to the equation of
continuity given by Eq. (1.10), we have

∂ρ
∇·J = − , (1.61)
∂t
and Ohm’s law (Eq. (1.60)), we have

∂ρ
∇·σE = − . (1.62)
∂t
Next, using Gauss’ law (Eq. (1.1c)),
σ σ
σ∇·E = ∇·D = ρ , (1.63)
ε ε
and then combining it with Eq. (1.62) leads to

∂ρ σ
+ ρ =0 (1.64a)
∂t ε

⇓
ρ = ρ0 e−(σ /ε )t . (1.64b)

That is, at any point in the medium, charges vanish exponentially. Basically, the positive
and negative charges recombine, or move away from each other to be accumulated at the
surface of the bounded conducting medium. As charges move in a conducting medium to
redistribute themselves over the surface, some of the stored energy is dissipated into heat.
After equilibrium has been established, the stored electric energy is lower than that of the
cumulative initial condition of the charges in the medium.

1-4.5 The Lorentz Dielectric Model

A classical model for the complex dielectric constant of materials was developed by H. A.
Lorentz at the turn of the twentieth century. The model is based on a simple oscillatory
mechanical system in which bound electrons are allowed to move around stationary ions under
the driving force of applied electromagnetic fields. In this model each molecule is considered
to be independent of the other modules within the material medium. That is, the motion of
1-4 Constitutive Relations: Macroscopic Properties of Matter 21

F = qE
Equilibrium
L
m

S d0

Figure 1-7: An equivalent mechanical model describing the motion of bound electrons around
ions subjected to an external electric field E.

an electron belonging to a particular molecule does not influence the motions of the other
electrons and vice versa. Also, under the small signal approximation, the electrostatic force
acting on displaced electron clouds around an ion is described by a linear relation. Electron
collision is described by a damping coefficient in an equivalent mechanical system, as shown
in Fig. 1-7. The equation of motion for an electron cloud of mass m under the influence of the
local electric force field E is given by

d2L dL
m 2
+ d0 + SL = q E(t) , (1.65)
dt dt
where L is the displacement of the electron cloud’s center of mass from the position of
equilibrium, q is the charge of the moving electron cloud, S is the spring constant representing
the linearized electrostatic force, and d0 is the damping coefficient. Assuming the excitation
to be time-harmonic, i.e., E(t) = Re[Ee e −iω t ], the phasor form of displacement is then given
by
(q/m)E e
e=
L , (1.66)
ω02 − ω 2 − iγω
where ω02 = S/m is the natural (resonant) frequency of the system and γ = d0 /m is the
damping factor. Next, assuming there are N independent polarized molecules per unit volume,
then the polarization is given by

ωp2
e = Nq L
P e= e,
ε0 E (1.67)
ω02 − ω 2 − iγω

where ωp2 = Nq2 /mε0 is defined as the plasma frequency. Recalling that Pe = ε0 χe E
e and
ε = ε0 (1 + χe ), the dielectric constant of the medium is then given by
!
ωp2
ε = ε0 1 + 2 , (1.68)
ω0 − ω 2 − iγω
22 Chapter 1 Electromagnetic Fields

20
Real partofofεrε
Realpart
Imaginary partofofεεr
Imaginarypart
15

10

ε′r
ε′r , ε′′r 5
ε

ε′′r
0

−5
−5

−10
−10
0 10 20 30 40 50 60 70 80 90 100
Frequency(ω)
Frequency (ω)

Figure 1-8: The real and imaginary parts of the dielectric constant of a material, according to
the Lorentz model given by Eq. (1.69) with ω0 = 50 rad/s, ωp = 70 rad/s, and γ = 5 rad/s.

whose real and imaginary parts can be written as

!
′
ωp2 (ω02 − ω 2 )
ε = ε0 1+ 2 , (1.69a)
(ω0 − ω 2 )2 + γ 2 ω 2
!
ωp2 γω
ε ′′ = ε0 . (1.69b)
(ω02 − ω 2 )2 + γ 2 ω 2

It is interesting to note that near resonance (ω ∼ ω0 ) the real part of ε can become less than
unity or even a negative number. The maximum value of ε ′′ occurs at around ω = ω0 and
its peak value is approximately equal to ωp2 /γω0 . Figure 1-8 shows plots of εr′ and εr′′ as a
function of frequency ω for some chosen values of ω0 , ωp , and γ .
At frequencies much smaller than the resonant frequency (ω ≪ ω0 ), it can be easily shown
that
!
ω 2
p
ε ′ ≈ ε0 1 + 2 , (ω ≪ ω0 ) (1.70a)
ω0
and
1-4 Constitutive Relations: Macroscopic Properties of Matter 23

γωp2 ω
ε ′′ ≈ ε0 , (ω ≪ ω0 ) . (1.70b)
ω04

These expressions indicate that the real part of the dielectric constant is independent of
frequency and the loss factor, ε ′′ /ε ′ , due to dielectric dispersion, is very small and increases
linearly with frequency. Also, at frequencies well above the resonant frequency, approximate
expressions for the real and imaginary parts of the dielectric constant can be obtained and are
given by
!
ω 2
p
ε ′ ≈ ε0 1 − 2 (ω ≫ ω0 ) (1.71a)
ω
and
γωp2
ε ′′ ≈ ε0 . (ω ≫ ω0 ) . (1.71b)
ω3
It is interesting to note that in the limit as ω → ∞, the real part of the permittivity approaches
that of free space and the imaginary part vanishes.
Some materials may have multiple resonances, for which their spectral dielectric behavior,
based on the Lorentz model, can be represented by
N
X ε0 ωp2j
ε1 = ε0 + , (1.72)
j=1
ω02j − ω 2 − iγ j ω

where ω0 j , ωp j , and γ j are the jth resonant frequency, plasma frequency, and damping factor
respectively.

1-4.6 The Drude Model for Metals

Conductors are materials with very high conductivity (σ ≫ 1). As will be shown later, the
material conductivity can be expressed in terms of a frequency-dependent imaginary part of
the dielectric constant given by
′′ σ
εcond = . (1.73)
ω
Hence, a standard dielectric model used for metals with finite conductivity is expressed as
 
σ
εmetal = ε0 1 + i (low frequencies). (1.74)
ωε0
This model is based on Ohm’s law and is valid at low frequencies ( f < 100 GHz). At higher
frequencies, a better approach to describe the spectral behavior of conductors is the Drude
model, which can be derived from the Lorentz dielectric model introduced in the previous
subsection. Good conductors contain a large number of electrons at the top of the energy
distribution and they can be easily excited and moved around within the conduction band.
These electrons are essentially free electrons and not bound to specific ions in the material
lattice. As a result, the spring constant S in the equivalent mechanical model shown in Fig. 1-7
24 Chapter 1 Electromagnetic Fields

can be set to zero, which in turn sets ω0 = S/m = 0 in Eq. (1.68), simplifying it to
!
ωp2
ε = ε0 1 − 2 . (1.75)
ω + iγω

The explicit expressions for the real and imaginary parts of the dielectric constant are given
by

ωp2
εr′ = 1 − (1.76a)
ω2 + γ2
and (Drude model for high frequencies)
ωp2 γ
εr′′ = . (1.76b)
ω (ω 2 + γ 2 )

At low frequencies where ω ≪ γ and ωp ≪ γ ,

εr′ ≈ 1 (1.77a)
and (Drude model for low frequencies)
ωp2
εr′′ = , (1.77b)
ωγ

which is consistent with the expression for εmetal at low frequencies. A direct comparison
between these expressions reveals the relationship between the conductivity of a metal and its
plasma frequency and damping factor:

ε0 ωp2 Nq2
σ= = S/m. (1.78)
γ mγ

It is important to note that the conductivity of a metal at low frequencies is independent of

frequency.

Example 1-3: Collision Frequency of Copper

It is interesting to examine how some of the parameters in the Drude model can be estimated
experimentally. Starting from Eq. (1.78), find the collision frequency γ for copper.
Solution: Let us consider copper at low frequencies and try to estimate an empirical value for
the collision frequency γ . Assuming one free electron per atom and noting that the molecular
1-4 Constitutive Relations: Macroscopic Properties of Matter 25

weight of copper is Mr = 63.54 g and its density is d = 8.92 g/cm3 , the number density of
electrons per unit volume can be calculated using Avogadro’s number (A = 6.022 × 1023 ):

Ad 6.022 × 1023 × 8.92 × 106

N= = = 8.43 × 1028 m−3 . (1.79)
Mr 63.54
The conductivity of copper at microwave and lower frequencies is measured to be
σ = 5.8 × 107 . Using the charge and mass of an electron (q = 1.6 × 10−19 and
m = 9.1 × 10−31 ) in the Drude conductivity equation given by Eq. (1.78), the estimated
collision frequency of copper is found to be

Nq2 8.43 × 1028 × (1.6 × 10−19 )2

γ= = = 4.08 × 1013 s−1 .
mσ 9.1 × 10−31 × 5.8 × 107

1-4.7 Dielectric Relaxation Model for Polar Molecules

Many substances in nature are made up of molecules with a net charge of zero, and yet they
exhibit a nonzero electric dipole moment. By definition, a polar molecule is a molecule with
one end being slightly positively charged and the other end being slightly negatively charged.
Permanent dipole moments are formed in atoms with different electronegativity values like
carbon monoxide. The positive charge of the dipole comes from the protons in the carbon
nucleus and the negative charge of the dipole comes from the excess electrons orbiting mostly
around the oxygen atom. There are also molecular structures in which a significant dipole
moment is created from the asymmetry of the bond atoms. The water molecule is one such
example. In the atomic structure of a water molecule, the positively charged hydrogen atoms
are positioned in an asymmetric manner with respect to the negatively charged oxygen atom.
The lines connecting the oxygen and hydrogen atoms form an angle of about 104.5◦ (instead
of 180◦ for a symmetric structure), giving rise to a permanent electric dipole moment, as
shown in Fig. 1-9. In the presence of an applied external electric field E, an electric dipole
with dipole moment p = qd experiences a torque given by

T = p×E . (1.80)

˚
Figure 1-9: The atomic structure of a water molecule generates a permanent dipole moment P.
26 Chapter 1 Electromagnetic Fields

Here q refers to the equivalent charge and d is the displacement vector. Assuming the electric
field is along the z-axis and denoting the angular momentum of the dipole by I, the equation
of motion is given by
∂ 2θ
I 2 = pE sin θ , (1.81)
∂t
where θ denotes the instantaneous angle between P and E. The solution of this nonlinear
differential equation depends on the time variation of the electric field E. For a static electric
field, eventually the dipole will align itself with the electric field and consequently, per
Eq. (1.80) the torque disappears. For a small deviation from the equilibrium condition, the
sine function can be replaced with its argument, which leads to the second-order differential
equation
∂ 2θ
I 2 = pE θ . (1.82)
∂t
The natural resonant frequency of the dipole around its equilibrium can easily be calculated
from r
1 pE
f0 = . (1.83)
2π I
For a polar substance, the equation of motion should also include a damping factor due to
the collision, which results in a damped oscillation of the dipoles around the equilibrium
condition. For substances in solid or liquid form, the frequency of collision is high and the rate
of change of angular velocity with time is very small. This leads to overdamped oscillation, a
situation where the dipole “relaxes” to its equilibrium after an initial perturbation.
In an unbiased small region of a medium with polar molecules having random
orientations, the net dipole moment can be assumed to be negligible. The sudden application
of an electric field, in the form of a step function, causes the medium to get polarized. For
such a medium, the polarization vector is composed of two components: The first component
is due to the quick motion of electron clouds in the direction opposite of the applied electric
field, and the second component is due to the rotation of the permanent dipole moments of the
polar molecules. Since the molecules themselves are much heavier than the electron clouds,
this latter motion is far slower than the motion of the electron clouds. Denoting the static
susceptibility associated with the electron clouds by χ0e and that of the permanent dipoles
by χ0d , and noting that the polarization vector due to the dipole moments reaches the steady-
state value in an overdamped manner, the time-domain step response of the polarization
vector can be written as

Pstep (t) = ε0 E0 [(χ0e − χ0d )e−t/τ + χ0d ] u(t) , (1.84)

where τ is the relaxation time of the overdamped permanent dipoles and u(t) denotes a
step function. Here, it is assumed that the polarization due to the electron clouds happens
instantaneously (compared with the relaxation time of the dipoles). Hence the impulse
response is given by
 
χ0d − χ0e −t/τ
P(t) = ε0 E0 e u(t) + χ0e δ (t) , (1.85)
τ
1-4 Constitutive Relations: Macroscopic Properties of Matter 27

ε 0d

ε 0e

t=0
Figure 1-10: The step-function response of polarization in a medium with polar molecules.

where δ (t) denotes a delta function. The step-function response of the polarization as a
function of time, as given by Eq. (1.84), is shown in Fig. 1-10.
Upon comparing the expression for P(t) with the expression given by Eq. (1.54), we
obtain the following expression for the susceptibility χe :
χ0d − χ0e −t/τ
χe (t) = e u(t) + χ0e δ (t) . (1.86)
τ
Taking the Fourier transform gives
χ0d − χ0e
χe (ω ) = χ0e + . (1.87)
1 − iωτ
Incorporating the preceding expression into Eq. (1.58) leads to the following expression for
the normalized complex dielectric constant of the medium:
χ0d − χ0e
εr (ω ) = 1 + χ0e + . (1.88)
1 − iωτ
For water molecules, the relaxation time has been determined experimentally to be
τ = 8 × 10−12 , and the susceptibilities associated with the electron cloud and the permanent
dipole moment are also found to be χ0e = 4.27 and χ0d = 76.5, respectively. Using these
values, the real and imaginary parts of the relative dielectric constant of water were computed
as a function of frequency and then plotted in Fig. 1-11.
It is interesting to note that once water molecules crystallize into ice at low temperatures,
28 Chapter 1 Electromagnetic Fields

80
Real part of εr
70 Imaginary part of εr

60
ε′r
50

ε′r , ε′′r 40

30

20
ε′′r
10

0
9 10 11 12 13
10 10 10 10 10
Frequency (ω)

Figure 1-11: The real and imaginary parts of the relative dielectric constant of water at 20 ◦ C.

the dielectric constant becomes drastically different from what is shown in Fig. 1-11. Dipoles
of water molecules in solid form cannot respond to the applied electric field, and hence the
susceptibility associated with the permanent dipoles cannot contribute to the polarization
vector (P). For ice, the susceptibility χeice associated with the electron clouds is 2.155 at
microwave and lower frequencies. Hence εr′ice = 3.155 and the imaginary part is very small.
An interesting experiment is to place an ice cube in a styrofoam cup with a cap, keeping
it in a freezer to ensure that no part of the ice cube has any surface water formed on it. Then
the cup is placed in a microwave oven for a minute of time at normal power. After exposure
to the microwave power, remove the cap and observe the ice cube in its initial form. The same
amount of water in the cup would boil (make sure you don’t have the cap on).

1-5 Kramers-Krönig Relations

The structure of the expression for the complex permittivity of materials given by Eq. (1.58)
is not arbitrary. It can be shown that the real and imaginary parts of the complex permittivity
are related to each other. In addition to the causality condition that led to Eq. (1.58), we
further postulate that the system of charges in the material is unconditionally stable. That
is, the impulse response χe (t) → 0 as t → ∞. This stipulation guarantees that the Fourier
transform χe (ω ) has no poles in the upper half-plane of the complex ω -plane corresponding
to Im(ω ) ≥ 0. Equation (1.58) also implies that χe (ω ) → 0 as ω → ∞ in the upper half-plane.
Now consider the contour integral
I
χe (ω ′ )
I= ′
dω ′ , (1.89)
C ω −ω
1-5 Kramers-Krönig Relations 29

[ω′]
ω′ complex plane

ω
[ω′]

Figure 1-12: Complex ω ′ -plane and closed contour C.

where the closed contour C is composed of the real axis and a semicircle in the upper half-
plane, as shown in Fig. 1-12, whose radius approaches infinity. Since there are no poles and
branch cuts associated with the integrand of Eq. (1.89) within the contour C, according to
Cauchy’s theorem I = 0 (see Appendix A). Also, according to Jordan’s lemma, the integral
over the semicircle
> of infinite radius vanishes. Defining the principal value of the integral,
denoted by , as the integral over the real axis except at ω = ω ′ , and evaluating one-half of
the residue at ω = ω ′ , it can be shown that
? +∞
χe (ω ′ )
−iπχe (ω ) + ′−ω
dω ′ = 0 . (1.90)
−∞ ω

Representing the real and imaginary parts of the complex susceptibility as χe (ω ) =

χe′ (ω ) + iχe′′ (ω ), it can easily be shown that
? +∞ ′′ ′
′ 1 χe (ω )
χe (ω ) = dω ′ (1.91a)
π −∞ ω ′ − ω
and
+∞
χe′ (ω ′ )
?
1
χe′′ (ω ) = − dω ′ . (1.91b)
π −∞ ω′ − ω

Clearly, Eqs. (1.91a) and (1.91b) show that the real and imaginary parts of the susceptibility
function are related to each other through a pair of integral transforms known as Kramers-
Krönig relations. Using Eq. (1.58), a similar relationship between the real and imaginary
30 Chapter 1 Electromagnetic Fields

parts of the permittivity (εr (ω )) can be obtained and are given by

1 +∞ εr′′ (ω ′ )
?
εr′ (ω ) − εr∞ = dω ′
π −∞ ω ′ − ω
and
+∞
εr′(ω ′ ) − εr∞
?
1
εr′′ (ω ) =− dω ′ .
π −∞ ω′ − ω

where εr∞ is the value of the permittivity at infinite frequency. Strictly speaking, εr∞ = 1
according to Eq. (1.58).

1-6 Boundary Conditions

The Maxwell’s equations presented and discussed in the preceding sections are valid in the
neighborhood of a point in a medium where all constitutive parameters (ε , µ , σ ) are constant
functions of position. At interfaces between different media with different constitutive
parameters, the sudden discontinuity across the interface may cause the field quantities to
become discontinuous. To address this mathematical difficulty, we assume there is a very
thin hypothetical layer of width ∆ z in which the constitutive parameters vary continuously
between their values in the two distinct media (see Fig. 1-13), and then we evaluate the field

E1, H1, B1, D1

μ1, ε1, σ1, ... S Interface
surface
Medium 1 μ2, ε2, σ2, ...
E2, H2, B2, D2
Medium 2

ε1

ε2

z
Dz

Figure 1-13: An abrupt change in the values of constitutive parameters is approximated by

inserting a thin layer transitioning the discontinuous parameter in a continuous manner.
1-6 Boundary Conditions 31

Region 1 n̂1
∆a

∆l

Region 2

Figure 1-14: A small cylindrical volume between two homogeneous media. The base of the
cylinder is chosen to be small enough so that the interface can be assumed flat locally.

in the limit as ∆ z → 0. The reason for introducing this fictitious layer is to satisfy the condition
that the field quantities and their derivates are continuous throughout the space of interest. It
is only under such a continuity condition that theorems such as the divergence theorem and
Stokes’ theorems can be applied. To derive the solution to Maxwell’s equations for a medium
composed of two or more homogeneous media, we need to derive relationships between the
field quantities across the media interfaces. To establish such relationships, let us consider a
thin cylindrical volume between the two media as shown in Fig. 1-14, with the cylinder base
being parallel to the interface (locally) and having a very small area ∆ a and height ∆ ℓ.
Consider a field vector F whose components and their derivatives are continuous within
the volume and on the surface of this cylinder. For this field vector, the following relations
hold:
$
∇ · F dυ = F · ds (1.92a)
∆V ∆S
and
$
∇ × F dυ = − F × ds , (1.92b)
∆V ∆S

where ∆ V and ∆ S refer respectively to the volume and surface of the thin cylinder. Equation
(1.92a) is a direct statement of the divergence theorem and Eq. (1.92b) can be derived from
the divergence theorem (see Appendix C). Applying the form of Eq. (1.92b) to the electric
field E in Faraday’s law (Eq. (1.1c)) leads to
$ $
∂B
∇ × E dυ = − E × ds = − dυ . (1.93)
∂t
∆V ∆S ∆V
32 Chapter 1 Electromagnetic Fields

In the limit as ∆ ℓ → 0, ∆ V → 0, and as a result the right-hand side of Eq. (1.93) goes to
zero. Then we have

lim E × ds = (E1 × n̂1 + E2 × n̂2 ) ∆ a = 0 ,
∆ ℓ→0
∆S

or
n̂ × (E1 − E2 ) = 0 , (1.94)

where n̂ = n̂1 = −n̂2 and n̂1 is the unit vector pointing toward medium 1. Equation (1.94)
states that the tangential component of the electric field intensity is continuous across a
boundary interface between two media. Applying Eq. (1.92b) to the modified Ampère’s law,
given by Eq. (1.1a), we have
$ $ $
∂D
∇ × H dυ = − H × ds = dυ + J dυ . (1.95)
∂t
V ∆S ∆V ∆V

Following the same recipe using in connection with the derivation leading to Eq. (1.94), as
the cylinder shrinks into a circular disk, the contributions from the side wall of the cylinder as
well as the volume integrals on the right-hand side of Eq. (1.95) vanish, resulting in

n̂ × (H1 − H2 ) = 0 (tangent component of H is continuous). (1.96)

An exception to this boundary condition is when a volumetric current density J collapses into
a surface current density Js at the interface. In that case
$
lim J d υ = Js ∆ a ,
∆ ℓ→0
∆V

and the boundary condition given by Eq. (1.96) should be modified to

n̂ × (H1 − H2 ) = Js . (1.97)

The boundary condition for electric and magnetic flux densities can be obtained from Gauss’
law in conjunction with the divergence theorem given by Eq. (1.92a). Upon substituting
Gauss’ magnetic law, as in Eq. (1.1d), into Eq. (1.92a) and shrinking ∆ ℓ → 0, we have
$
∇·B = B · ds = 0
∆V ∆S
and

lim B · ds = (n̂1 · B1 + n̂2 · B2 ) ∆ a = 0 , (1.98)
∆ ℓ→0
∆S
1-6 Boundary Conditions 33

which leads to

n̂1 · (B1 − B2 ) = 0 (normal component of B is continuous). (1.99)

Similarly, substituting the electric flux density for F into Eq. (1.92a) and using Gauss’ law for
electricity (Eq. (1.1c)), we get
$
D · ds = ρ dυ .
∆S ∆V

As the cylinder shrinks to a disk we can again show that

(normal component of D is continuous

n̂ · (D1 − D2 ) = 0 (1.100)
across boundary with no surface charge).

For time-varying field quantities, it should be noted that the continuity of the tangential
components of E and H at the interface between two media also implies the continuity of
the normal components of B and D at the interface. That is, if the continuity of the tangential
components of E and H are imposed at the boundary, there will be no need for imposing the
continuity of the normal components of B and D. In the case where there is a surface charge
density at the interface, i.e., $
lim ρ d υ = ρs ∆ a ,
∆ ℓ→0
∆V
then Eq. (1.100) must be modified to

(continuity of normal component of D

n̂ · (D1 − D2 ) = ρs (1.101)
in presence of surface charge).

If the second medium is a perfect electric conductor (σ2 = ∞), then E2 = 0 and D2 = 0
throughout the entire volume of the perfect conductor medium, and therefore Eqs. (1.94),
(1.97), and (1.101) reduce to
 
n̂ × E1 = 0
 n̂ × H1 = Js 
 
 n̂ · B1 = 0  (Medium 2 = perfect conductor).
n̂ · D1 = ρs
34 Chapter 1 Electromagnetic Fields

1-7 Drift Current in Metals

As noted earlier, there is a large number of free electrons in the conduction band of good
conductors, which facilitates the flow of electric currents without significant resistance. One
important fact to keep in mind is that the net charge on the surface of good conductors, as they
carry electric current, is zero. Another interesting feature is the speed with which the current
flows in conductors. In air, electromagnetic fields propagate at the speed of light, 3 × 108
m/s. Consider an air-filled coaxial cable along the z-axis, supporting an electromagnetic field
propagating along the z-direction with its magnetic field oriented along φ̂ φ. This magnetic
field will generate z-directed surface currents on both the inner and outer conductors, which
can be calculated from Js = n̂ × H. Since H is propagating at the speed of light, the electric
current on the metal must be moving at that speed. But is it really feasible for electrons in
the conduction band to achieve such a high velocity? The answer is obviously no, so how
do we reconcile these facts? As is shown in the following example, the charges that make
up the surface current move at a much lower velocity known as the drift velocity. For good
conductors the drift velocity is on the order of 10−5 m/s, which depends on the current density
and is inversely proportional to the number of free electrons. The apparent contradiction can
be resolved by realizing that if an electron moves slightly, the change in the fields as a result
of that movement propagates at the speed of light and can act on the other electrons down the
line very quickly. The movement of electrons themselves is in fact hampered by collisions.
The drift velocity ud associated with electrons refers to the average velocity of the electrons
in the conductor. For materials with finite conductivity, the field can penetrate to some extent
into the conductor and a volumetric current, instead of a surface current, is formed near the
surface. If ρ represents the charge density of the moving charges near the surface, the drift
current, according to Eq. (1.39), is given by

J = ρ ud . (1.102)
√
As will be shown later, the penetration depth in a good conductor is given by δ = 1/ πσ µ0 f ,
where f is the frequency. Assuming that the current density is uniform over this thin layer,
the volumetric current, in terms of the surface current, assuming the conductor is a perfect
conductor, is given by
Js n̂ × H
J= = .
δ δ
Hence, using Eq. (1.102), the drift velocity can be computed from
√
|n̂ × H| |n̂ × H| πσ µ0 f
ud = = . (1.103)
δρ ρ
It follows that the drift velocity increases with increasing the field intensity H, the frequency f ,
and the metal conductivity σ .
1-8 Hall Effect in Conducting Media 35

Example 1-4: Drift Velocity

Find the drift velocity in copper at 100 MHz, assuming that the tangential magnetic field
intensity is 1 A/m and that there is one free electron available per copper atom.
Solution: The conductivity of copper is σ = 5.8× 107 S/m. To find the charge density, we first
need to find the number N of copper atoms per cubic meter, from which the charge density
can be obtained using
ρ = eN .
√
From Eq. (1.79), N = 8.43 × 1028 m−3 . Using δ = 1/ πσ µ0 f , the skin depth of copper
at f = 100 MHz is calculated to be δ = 6.6 × 10−6 m, and the charge density is ρ = eN =
1.6 × 10−19 × 8.43 × 1028 = 1.349 × 1010 C/m3 . Hence,
H 1
ud = = −6
= 1.12 × 10−5 m/s.
δρ 6.6 × 10 × 1.34 × 1010

1-8 Hall Effect in Conducting Media

In a conducting medium, the flow of electric current is influenced by an external magnetic
flux density B. This phenomenon is known as the Hall effect, discovered by Edwin H. Hall in
1879. As discussed earlier, the electrons in a conducting medium move with drift velocity ud
and thus experience the Lorentz force once the conductor is exposed to an external magnetic
flux density B. The electromotive electric field that is perpendicular to ud is given by

E⊥ = ud × B . (1.104)

Calculating the drift velocity from Eq. (1.102) and substituting into Eq. (1.104), we have
J×B
E⊥ = . (1.105)
ρ
There is also a component of the electric field inside the conductor that is parallel to J given
by Ohm’s law:
J
Ek = . (1.106)
σ
Hence the direction of the electric field is tilted away from the direction of the current by

|E⊥ | |E⊥ | σ
θ = tan−1 ≈ = |B| .
|Ek | |Ek | ρ

If we now consider a metallic strip carrying a current density J and having a width of w and a
magnetic flux density perpendicular to the metallic strip surface, the voltage measured across
the strip is given by Z w
|J| |B|
VH = E⊥ · dℓℓ = w. (1.107)
0 ρ
36 Chapter 1 Electromagnetic Fields

For example, using a copper strip with charge density ρ = 1.349 × 1010 C/m3 (see Example
1-4) at a frequency with skin depth 20 µ m carrying 1 A of current, the Hall voltage is
measured to be
I IB
VH = Bw = = 3.7 × 10−6 B V.
w δρ δρ
Hence if B = 1 mT, the Hall voltage induced is VH = 3.7 nV.

1-9 Generalized Coordinates

An important feature of vector calculus is that the equations expressing the physical
relationships take the same form independently of any particular system of coordinates.
However, to express the field quantities associated with a particular solution of the physical
system, it is necessary to do so using a specific coordinate system. Often, depending on the
geometry of the boundary between the two media, a particular coordinate system is deemed
the most appropriate for a given configuration. The most commonly used coordinate system
is the Cartesian coordinate system, in which a position vector specifying a point in three-
dimensional space is given by
r = x x̂ + y ŷ + z ẑ . (1.108)
The differential length in this system is simply given by

dr = dl = dx x̂ + dy ŷ + dz ẑ .

We should note that in the general case, when the differential length is applied to a contour or
path, the differential increments dx, dy, and dz may not be necessarily independent.
In a generalized coordinate system, the position of a point is defined by the coordinate
triplet (u1 , u2 , u3 ). Let U1 = f1 (x, y, z), U2 = f2 (x, y, z), and U3 = f3 (x, y, z) be three
independent, single-valued, and continuously differentiable functions of the rectangular
coordinates x, y, z. In general, there exist three single-valued functions

x = X (u1 , u2 , u3 ) , y = Y (u1 , u2 , u3 ) , and z = Z(u1 , u2 , u3 ) , (1.109)

that provide an ordered set (x, y, z) in terms of (u1 , u2 , u3 ). The functions U1 , U2 , U3 are called
general or curvilinear coordinates. Functions of the form

U1 = constant, U2 = constant, and U3 = constant

define coordinate surfaces. The intersection of two coordinate surfaces creates a coordinate
curve. The intersection of two coordinate curves or three coordinate surfaces defines a point.
At any point, unit vectors in coordinate surfaces tangent to coordinate curves define the unit
vectors û1 , û2 , and û3 at that point. Henceforth, we shall only consider orthogonal coordinate
systems; that is, at any point in space
(
1 i= j,
ûi · û j = (1.110)
0 i, j,
1-9 Generalized Coordinates 37

û1 × û2 = û3 , û2 × û3 = û1 , and û3 × û1 = û2 . (1.111)

1-9.1 Differential Length in the General Coordinate System

In the general coordinate system, the differential length dr is defined as

∂r ∂r ∂r
dr = du1 + du2 + du3 . (1.112)
∂ u1 ∂ u2 ∂ u3
It is obvious that ∂ r/∂ u1 is parallel to û1 (tangent to u1 curve). Similarly, ∂ r/∂ u2 and
∂ r/∂ u3 are parallel to û2 and û3 , respectively. The configuration of an orthogonal generalized
coordinate system showing the constant coordinate surfaces, coordinate curves, and the unit
vectors is shown in Fig. 1-15.
Let us denote
∂r
= hi ûi , i = 1, 2, 3
∂ ui
where  
 ∂r 
hi =   .
∂ ui 
Hence
3
X
dℓℓ = dr = hi dui ûi . (1.113)
i=1
The three scaling factors hi can easily be obtained by noting that

∂r ∂X ∂Y ∂Z
= x̂ + ŷ + ẑ , (1.114)
∂ ui ∂ ui ∂ ui ∂ ui

U3-curve û3
U1-surface

U2-surface

û1 P

U1-curve û2
U3-surface U2-curve

Figure 1-15: Orthogonal generalized coordinate system showing constant curves and surfaces.
38 Chapter 1 Electromagnetic Fields

which leads to
" 2  2  2 #1/2
∂X ∂Y ∂Z
hi = + + . (1.115)
∂ ui ∂ ui ∂ ui

Also,
1 ∂r
ûi = . (1.116)
hi ∂ ui

Example 1-5: Scaling Factors and Unit Vectors

Derive the expressions for the scaling factors and unit vectors in the spherical coordinate
system.
Solution: Consider a spherical coordinate system with u1 = r, u2 = θ , u3 = φ , and

X = r sin θ cos φ , (1.117a)

Y = r sin θ sin φ , (1.117b)
Z = r cos θ . (1.117c)

Using Eq. (1.115),

hr = [sin2 θ cos2 φ + sin2 θ sin2 φ + cos2 θ ]1/2 = 1 , (1.118a)

hθ = [r2 cos2 θ cos2 φ + r2 cos2 θ sin2 φ + r2 sin2 θ ]1/2 = r , (1.118b)
hφ = [r2 sin2 θ sin2 φ + r2 sin2 θ cos2 φ ]1/2 = r sin θ . (1.118c)

Also,
1 ∂r
û1 = r̂ = = sin θ cos φ x̂ + sin θ sin φ ŷ + cos θ ẑ , (1.119a)
hr ∂ r
1 ∂r
û2 = θ̂
θ= = cos θ cos φ x̂ + cos θ sin φ ŷ − sin θ ẑ , (1.119b)
hθ ∂ θ
1 ∂r
û3 = φ̂
φ= = − sin φ x̂ + cos φ ŷ . (1.119c)
hφ ∂ φ

1-9.2 Differential Area and Volume

When dealing with electric and magnetic fields in various types of media, we often encounter
surface and volume integrals. Depending on the problem at hand, the choice of which
coordinate system to use may prove critically important. In general, the differential areas
1-9 Generalized Coordinates 39

in constant surfaces are given by

dS1 = ds1 û1 = h2 du2 û2 × h3 du3 û3 = h2 h3 du2 du3 û1 , (1.120a)
dS2 = ds2 û2 = h1 h3 du1 du3 û2 , (1.120b)
dS3 = h1 h2 du1 du2 û3 . (1.120c)

Also, in a general coordinate system the differential volume is given by

d υ = h1 du1 û1 · h2 du2 û2 × h3 du3 û3

= h1 h2 h3 du1 du2 du3 . (1.121)

1-9.3 Gradient of Scalar Functions

The gradient operator acts on a scalar field quantity and specifies the rate of change of that
quantity with respect to distance. For a differentiable scalar field ψ , the differential d ψ in a
Cartesian coordinate system, is given by

∂ψ ∂ψ ∂ψ
dψ = dx + dy + dz ,
∂x ∂y ∂z
which can be cast as the dot product of two quantities:
 
∂ψ ∂ψ ∂ψ
dψ = x̂ + ŷ + ẑ · (dx x̂ + dy ŷ + dz ẑ) . (1.122)
∂x ∂y ∂z
We recognize the quantity in the second parenthesis as the differential length along which the
change in ψ is measured. The quantity in the first parenthesis is defined as the gradient ∇ψ
of scalar function ψ . Hence, Eq. (1.122) can be rewritten as

d ψ = ∇ψ · dℓℓ .

In a similar fashion, we can show that d ψ in a general coordinate system may be written as

∂ψ ∂ψ ∂ψ
dψ = du1 + du2 + du3
∂ u1 ∂ u2 ∂ u3
 
1 ∂ψ 1 ∂ψ 1 ∂ψ
= û1 + û2 + û3 · (h1 du1 û1 + h2 du2 û2 + h3 du3 û3 ) .
h1 ∂ u1 h2 ∂ u2 h3 ∂ u3
(1.123)
P3
Since dℓℓ = i=1 hi dui ûi , it then can be concluded that
3
X 1 ∂ψ
∇ψ = ûi . (1.124)
hi ∂ ui
i=1
40 Chapter 1 Electromagnetic Fields

u
P
u

Figure 1-16: The curvilinear differential volume and surface used in evaluation of the divergence
of a vector function.

1-9.4 Divergence of Vector Functions

The divergence of a continuous and differentiable function F is defined as

P
i F · ∆ Si
∇ · F = lim . (1.125)
∆ V →0 ∆V
At a specified point in space, the divergence of a vector field quantity represents the
normalized outflux of the field from a very small volume surrounding the point. In a
generalized coordinate system, referring to Fig. 1-16 it can be shown that
X

F · ∆ Si = F1 h2 h3 |u1 +du1 /2 − F1 h2 h3 |u1 −du1 /2 du2 du3
i

+ F2 h3 h1 |u2 +du2 /2 − F2 h3 h1 |u2 −du2 /2 du3 du1


+ F3 h1 h2 |u3 +du3 /2 − F3 h1 h2 |u3 −du3 /2 du1 du2 , (1.126)

where F1 , F2 , and F3 are the components of F along dimensions U1 , U2 , and U3 , respectively.

Noting that

∂ (F1 h2 h3 )
F1 h2 h3 |u1 +du1 /2 − F1 h2 h3 |u1 −du1 /2 = du1 (1.127)
∂ u1
and that similar formulations apply to the second and third terms in Eq. (1.126),
X  
∂ ∂ ∂
F · ∆ Si = (h2 h3 F1 ) + (h3 h1 F2 ) + (h1 h2 F3 ) du1 du2 du3 . (1.128)
∂ u1 ∂ u2 ∂ u3
Also, d υ = h1 h2 h3 du1 du2 du3 . Hence,
1-9 Generalized Coordinates 41

3
X  
1 ∂ h1 h2 h3
∇·F = Fi . (1.129)
h1 h2 h3 ∂ ui hi
i=1

1-9.5 Curl of a Vector Function

Using Eq. (1.15) with the unit normal n̂ replaced with n̂i , the curl ∇ × F of a vector function
F can be expressed as
I
X
F · dℓℓ F · ∆ℓ
ûi · ∇ × F = lim ∆ S = lim , (1.130)
∆ S→0 ∆S ∆ S→0 ∆S
where ∆ S is the surface whose unit normal is ûi . Basically, the curl of a vector field quantity
at a point is the normalized circulation of the field quantity along three orthogonal small
contours circling the point. Let’s first consider ∆ S in the u2 –u3 plane as shown in Fig. 1-17,
with û1 as its unit normal.
According to Fig. 1-17,
X

F · ∆ ℓ = F3 h3 du3 |u2 +du2 /2 − F3 h3 du3 |u2 −du2 /2 − F2 h2 du2 |u3 +du3 /2 + F2 h2 du2 |u3 −du3 /2
 
∂ (h3 F3 ) ∂ (h2 F2 )
= − du2 du3 . (1.131)
∂ u2 ∂ u3
Also, ∆ S = h2 h3 du2 du3 ; therefore
 
1 ∂ (h3 F3 ) ∂ (h2 F2 )
û1 · ∇ × F = − . (1.132a)
h2 h3 ∂ u2 ∂ u3

u
P
u

∆ S = h2 h3 du2 du3
Figure 1-17: A small contour in the constant U1 -surface used to compute a component of the
curl of a vector field.
42 Chapter 1 Electromagnetic Fields

Similarly, by permutation it can be shown that

 
1 ∂ ∂
û2 · ∇ × F = (h1 F1 ) − (h3 F3 ) (1.132b)
h3 h1 ∂ u3 ∂ u1
and
 
1 ∂ ∂
û3 · ∇ × F = (h2 F2 ) − (h1 F1 ) . (1.132c)
h1 h2 ∂ u1 ∂ u2
Finally,
3
X
∇×F = (∇ × F) · ûi
i=1
 
h1 û1 h2 û2 h3 û3 
 
1  ∂ ∂ ∂ 

=  . (1.133)
h1 h2 h3  ∂ u1 ∂ u2 ∂ u3 
 
h1 F1 h2 F2 h3 F3 

Example 1-6: Spherical Coodinate System

Determine ∇ × F in the spherical coordinate system.

Solution: From Eq. (1.117) in Example 1-5,

h1 = hr = 1 , h2 = hθ = r , h3 = hφ = r sin θ .

Application of Eq. (1.133) gives

 
1 ∂ ∂
∇×F = 2 (r sin θ Fφ ) − (rFθ ) r̂
r sin θ ∂θ ∂φ
 
1 ∂ ∂
+ (Fr ) − (r sin θ Fφ ) θ̂ θ
r sin θ ∂ φ ∂r
 
1 ∂ ∂
+ (rFθ ) − φ.
(Fr ) φ̂
r ∂r ∂θ
1-9 Generalized Coordinates 43

Upon further expansion of the derivatives involved, the expression can be simplified to
 
1 ∂ ∂
∇×F = (sin θ Fφ ) − (Fθ ) r̂
r sin θ ∂ θ ∂φ
 
1 1 ∂ Fr ∂
+ − (rFφ ) θ̂ θ
r sin θ ∂ φ ∂r
 
1 ∂ ∂
+ (rFθ ) − (Fr ) φ̂φ. (1.134)
r ∂r ∂θ

1-9.6 Laplacian of a Scalar Field

The Laplacian is a differential operator applied to scalar field quantities. The Laplacian of a
differentiable scalar field quantity ψ is defined by

∇2 ψ = ∇ · ∇ψ . (1.135)

Using Eqs. (1.124) and (1.129), it can be shown that

      
2 1 ∂ h2 h3 ∂ ψ ∂ h3 h1 ∂ ψ ∂ h1 h2 ∂ ψ
∇ ψ= + + . (1.136)
h1 h2 h3 ∂ u1 h1 ∂ u1 ∂ u2 h2 ∂ u2 ∂ u3 h3 ∂ u3

1-9.7 Curl of Curl of a Vector Field

When analyzing field quantities, we often encounter the need to evaluate ∇ × ∇ × F. We adopt
the following identity:
∇ × ∇ × F = ∇∇ · F − ∇2F , (1.137)
where ∇2 F = ∇ · (∇F).
Given a vector function F, the quantity on the left-hand side of Eq. (1.137), namely
∇ × ∇ × F, can be readily computed by applying the curl operator twice. Similarly,
∇∇ · F = ∇(∇ · F) can be computed by applying the divergence operator to F, which would
produce a scalar function, and then applying the gradient operator, resulting in a vector
function. Such procedures, however, do not apply to the last term in Eq. (1.137) because
∇F represents a gradient operator applied to a vector function, which is inconsistent with the
rules of vector algebra. The term ∇2 F can be computed indirectly by computing the two other
terms in Eq. (1.137) and then applying

∇2 F = ∇∇ · F − ∇ × ∇ × F . (1.138)

In Cartesian coordinates the process leads to

∇ · ∇F = ∇2 Fx x̂ + ∇2 Fy ŷ + ∇2 Fz ẑ . (1.139)
 44 Chapter 1 Electromagnetic Fields

Summary
Concepts
• Maxwell’s equations can be expressed in • When considering the full range of the
integral and differential form (with the latter electromagnetic spectrum, all materials exhibit
also known as the point form of Maxwell’s a dispersive behavior (the permittivity and
equations). permeability are functions of frequency). This
• Gauss’ laws of electricity and magnetism can be is due to the fact that the charges or permanent
derived from the combination of Ampère’s law dipoles (electric or magnetic) have finite masses
and Faraday’s law.ă and cannot react instantaneously to the applied
• Faraday’s law (∇ × E = −∂ B/∂ t) and the fields.
modified Ampère’s law (∇ × H = J + ∂ D/∂ t), • Electromagnetic boundary conditions define the
in addition to the law of conservation of relationships between field quantities across a
charge, which is expressed by the equation of discontinuity in material properties between two
continuity (∇ · J = −∂ ρ /∂ t), are the necessary adjacent media. These include the continuity of
and sufficient equations required for solving for the tangential components of the electric and
the field quantities E, D, B, and H. magnetic fields (Et and Ht ) and the continuity
• In the general case, the electric flux density D of the normal components of the electric and
and the magnetic flux density B in a material magnetic flux densities (Dn and Bn ).
medium are each related to both the electric field • Differential operators such as the gradient,
intensity E and the magnetic field intensity H divergence, and curl assume different mathe-
in that medium. The relationships depend on matical forms in different orthogonal coordi-
the degrees of polarization and magnetization nate systems. The specific choice of which
of the molecules comprising the material. Under coordinate system to use depends on the
the small-signal approximation, however, these geometry of the boundary value problem under
relations can be expressed in linear form. consideration.
 SUMMARY 45

Important Equations
Point form of Maxwell’s equations: Kramers-Krönig relations:
∂B 1 +∞ εr′′ (ω ′ )
?
′
∇×E = − εr (ω ) − εr∞ = dω ′
∂t π −∞ ω ′ − ω
∂D
1 +∞ εr′ (ω ′ ) − εr∞
?
∇×H = +J ′′
∂t εr (ω ) = − dω ′
∇·D = ρ π −∞ ω′ − ω
∇·B = 0 Boundary conditions:
Dielectric-dielectric interface:
Law of conservation of charge:
∂ρ n̂ × (E1 − E2 ) = 0
∇·J = − (equation of continuity) n̂ × (H1 − H2 ) = Js
∂t
n̂ · (D1 − D2 ) = ρ
Ohm’s law: n̂ · (B1 − B2 ) = 0
J = σE
Dielectric-metal interface:
Integral form of Maxwell’s equations: n̂ × E = 0
I " I n̂ × H = Js
∂B
E · dℓℓ = − · ds + u × B · dℓℓ n̂ · D = ρs
C S ∂t C
I " " I n̂ · B = 0
∂D
H · dℓℓ = · ds + I + ρ u · ds − u × D · dℓℓ
C S dt S C

D · ds = ρ
S

B · ds = 0
S
 46 Chapter 1 Electromagnetic Fields

Important Terms Provide definitions or explain the meaning of the following terms:
anisotropic divergence theorem induced electric field ordinary point
bianisotropic drift current inhomogeneous paramagnetism
biaxial electric field intensity isotropic homogeneous point form of Ohm’s law
boundary conditions electric flux density material polarization vector
conduction current electric susceptibility Kramers-Krönig relations Stokes’ theorem
constitutive relations equation of continuity Laplacian surface charge density
coordinate curve Faraday’s law law of conservation of surface current density
coordinate surface ferromagnetism charge transformer induction
curl Gauss’ law for electricity magnetic field intensity uniaxial
current density Gauss’ law for magnetism magnetic flux density volumetric charge density
diamagnetism gradient magnetization vector
dispersive material Hall effect modified Ampère’s law
displacement current homogeneous motional induction

PROBLEMS
Definitions and Identities

1.1
(a) Prove the divergence theorem:
$
∇ · A dV = A · dS ,
V S

for a differential vector field A and a continuous surface S that bounds a volume V .
(Hint: Start from the general definition of divergence.)
(b) Prove Stokes’ theorem: " I
∇ × A · dS = A · dℓℓ ,
S C
for a differential vector field A and a continuous surface S bounded by a continuous
contour C. (Hint: Start from the general definition of curl.)

1.2 Prove the following vector identities in a generalized coordinate system:

(a) ∇ · ∇ × A = 0 for all A. (Hint: Start with the divergence theorem applied to an arbitrary
volume V and then use Stokes’ theorem.)
(b) ∇ × ∇ψ = 0 for all ψ . (Hint: Start with Stokes’ theorem applied to an arbitrary surface
S and note that d ψ = ∇ψ · dℓℓ.)
PROBLEMS 47

1.3

(a) Suppose C is the closed boundary around an arbitrary surface S. Prove that for any
scalar function T , " I
∇T × dS = − T dℓℓ .
S C

(b) Suppose V is a volume bounded by an arbitrary surface S. Prove that

$
∇T d υ = T dS .
V S

(Hint: Start with Stokes’ theorem and the divergence theorem assuming A = T C, where
C is a constant vector.)

1.4 Construct a nonconstant vector function that has vanishing divergence and curl. Can you
find a method for finding such vectors?

1.5

(a) Prove the “bac-cab” vector identity given by Eq. (C.1) in Appendix C in Cartesian
coordinates:
A × (B × C) = (A · C)B − (A · B)C . (1.140)

(b) Is the identity given by Eq. (1.141) always valid? If not, find the conditions under which
the following identity is valid:

A × (B × C) = (A × B) × C . (1.141)

1.6 Consider at time t a surface S1 bounded by contour C1 . Let u be the instantaneous

velocity of element dS of the surface. The surface S1 together with the contour C1 may change
shape as time elapses so that u need not be constant for all elements of S1 . At time (t + ∆ t), S1
and C1 become S2 and C2 , as shown in Fig. P1.6. Use the general integral form of Ampère’s
law: I "
d
H · dℓℓ = I + D · ds ,
C dt
S
to show that
I " " I
dD
H · dℓℓ = I + ρ u · ds + · ds − (u × D) · dℓℓ .
C S S dt C
48 Chapter 1 Electromagnetic Fields

C2 = C(t + Δt)

Δll1 =
u Δt
dll S1 = S(t)
C1 = C(t)
ΔS

C2 = C(t + Δt)

Δll1 =
u Δt S2 = S(t + Δt)

dll
C1 = C(t)
ΔS

Figure P1.6: Surface and contour for Problem 1.6.

Fields and Currents

1.7 Suppose a magnet is approaching a conductive loop at a constant velocity, and the loop
is connected to a voltmeter as shown in Fig. P1.7.
(a) Qualitatively, plot the magnetic flux through the loop as a function of time.
(b) Qualitatively, plot the voltage measured by the voltmeter as a function of time.
(c) Does the size of the loop relative to that of the magnet have any effect on the results
obtained in parts (a) and (b)?
(d) Now imagine that you replace the voltmeter with a resistor and repeat the experiment.
Plot the current through the loop as a function of time.
(e) What is the effect of this current on the magnet?

1.8 As shown in Fig. P1.8, a dc current of 50 mA passes downward along the negative z
axis, enters a thin spherical shell of radius 0.03 m, and at θ = 90◦ enters a plane sheet. Write
expressions for the current sheet densities in the spherical shell and in the plane.
PROBLEMS 49

S N

Figure P1.7: A permanent magnet is traveling at a constant velocity toward a loop, passes
through the loop’s center, and continues its travel on a straight line. The loop is attached to a
voltmeter that measures the induced voltage.

Figure P1.8: Geometry of a conducting hemispherical shell attached to a ground plane. A

constant current I enters the hemispherical shell at its top.

1.9
(a) The electric flux density inside a cube is given by D = (5 + z)x̂ + (3xy) ŷ, as shown in
Fig. P1.9. Find the total electric charge enclosed inside the cubical volume.
(b) Repeat part (a) for D = (5t + xt) x̂ + (2t 2 + 2) ẑ and calculate the total displacement
current passing through the cubical surface.

1.10
di
(a) Starting from Maxwell’s equations, derive the circuit law for inductors, υ = L dt .
dυ
(b) Starting from Maxwell’s equations, derive the circuit law for capacitors, i = C dt .
(c) The “quasi-static condition” refers to a condition where the time variation of the fields
is very slow, (i.e., ∂ /∂ t ≈ 0). Use Faraday’s law to prove Kirchhoff’s voltage law under
quasi-static conditions.
50 Chapter 1 Electromagnetic Fields

z
5m

1m y
2m

Figure P1.9: A cubical volume in a medium with an established flux density.

1.11 Show whether or not each of the following fields satisfies Maxwell’s equations (and is
therefore a valid electromagnetic field) in a source-free (J = 0, ρ = 0) homogeneous region
of space. Find the corresponding magnetic field.
1
(a) E = cos(ω t − k|r − r′ |) ẑ, for r , r′ .
4π |r − r′ |
(b) E = sin(α x) cos(ω t − β z) ŷ and α 2 + β 2 = ω 2 µε .

1.12
(a) Consider the magnetic flux density given in Cartesian coordinates by
B0
B(x, y) = (x x̂ − y ŷ) (Wb/m2 ).
a
Show that this field obeys Maxwell’s equations in free space and sketch the magnetic
field lines. (Hint: The field lines are given by the differential equation dy/dx = By /Bx .)
(b) Suppose the electric field E and the magnetic flux density B are given in spherical
coordinates by
q
E(r,t) = − u(υ t − r)r̂ (V/m)
4πε0 r2
and
B(r,t) = 0 (Wb/m2 ) ,

where u(x) is the step function defined as

(
1 x>0,
u(x) ≡
0 x<0.

Show that these fields satisfy all of Maxwell’s equations, and determine ρ and J. (Hint:
The derivative of the step function u(x) is the Dirac delta function δ (x).)
PROBLEMS 51

1.13 A square wire loop of side a is placed at distance s from a very long straight wire
carrying a current I, as shown in Fig. P1.13.

s I

Figure P1.13: A square loop with a side parallel to a long wire carrying current I.

(a) Find the magnetic flux through the square loop.

(b) If someone now pulls the loop upwards, directly away from the straight wire, at
speed υ , what electromotive force Vemf is generated? In which direction (clockwise
or counterclockwise) does the current flow in the loop?
(c) What if the loop is pulled to the right, instead of upwards, at speed υ ?
(d) Now let us assume a slowly varying current I(t) is flowing through the wire. Repeat
part (b) for this case.
1.14 A wire loop in the shape of an isosceles right triangle with side of length a is placed at
distance s from a very long straight wire carrying a current I, as shown in Fig. P1.14.

s I

Figure P1.14: An isosceles right triangle in the vicinity of a long wire carrying current I. The
tip of the triangle is at a distance s from the wire.

(a) Find the magnetic flux flowing through the wire loop.
(b) If someone now pulls the loop upwards directly away from the straight wire, at
speed υ , what electromotive force Vemf is generated? In which direction (clockwise
or counterclockwise) does the current flow in the loop?
(c) What if the loop is pulled to the right, instead of upwards, at speed υ ?
(d) Now let us assume a slowly varying current I(t) is flowing through the wire. Repeat
part (b) for this case.
1.15 A rectangular loop of wire of linear resistance ρ (Ω/m) is placed in a time-varying
magnetic flux density B(t). If a voltmeter is connected across the loop using the same wire as
shown in Fig. P1.15, what voltage does the voltmeter show?
52 Chapter 1 Electromagnetic Fields

l1 l2

V h

B(t)

Figure P1.15: A rectangular wire loop with a linear resistivity ρ is placed in a medium
supporting a time-varying magnetic flux density B(t). A voltmeter placed at a distance l1 from
one end is measuring the induced voltage.

1.16

(a) A moving rod is rotating with ω = 0.4 rad/s inside a medium with a background
magnetic field of B = (3t + 4) ẑ, as shown in Fig. P1.16(a). The radius of the semicircle
is R. Find Vemf at t = 3 s. Assume that θ is 60◦ at t = 3 s.

(b) The semicircle (radius = R) shown in Fig. P1.16(b) is moving with constant velocity
u = v0 x̂ inside a medium with constant magnetic field B = B0 ẑ. Find the induced Vemf .

B = (3t + 4) ẑ
+
ω = 3 rad/s
B = B0 ẑ u
V
θ y
V _
+ _
x
(a) (b)

Figure P1.16: (a) Configuration of a semicircular wire attached to a moving wire pivoted at
the center and moving with a constant angular velocity. A time-varying magnetic flux density
perpendicular to the surface of the semicircle is present. (b) A semicircular wire moves with a
constant velocity in a constant magnetic field perpendicular to the surface of the semicircle.
PROBLEMS 53

1.17 An air-filled coaxial cable has an inner conductor of radius a and an outer conductor
of radius b (with b ≪ λ , where λ is the wavelength), as shown in Fig. P1.17. A time-varying
current flows along the inner conductor in the z-direction:

Jac = e−100(1−r/a) cos(ω t − kz) ẑ , 0 < r < a. (A/m2 ).

Contour C

a
x

z b

Figure P1.17: Configuration of a hollow coaxial cable with inner conductor of radius a and thin
outer cylinder of radius b. The coaxial line is carrying a sinusoidal time-varying current and a
hypothetical uniform volumetric charge density with a constant velocity. This charge density is
uniformly distributed over the cross section of only the inner conductor.

Additionally, charged particles with a fictitious volumetric density ρ0 (C/m3 ) are

distributed uniformly over the cross section of the inner conductor, moving with a constant
velocity uẑ. These charges only exist on the inner conductor.
(a) Determine the total current flowing on the inner conductor (units should be A).
(b) Determine the total linear charge density on the inner conductor (units should be C/m).
(c) Using the cylindrical symmetry of the problem, determine the magnetic field H0 in the
region a < r < b.
(d) Calculate the radial electric field.
(e) Assume that now your reference contour moves with velocity u = υ0 ẑ, where υ0 is
comparable in magnitude to u but much smaller than the velocity of light c. Compute
the magnetic field H.

Material Properties

1.18 Starting from the units of permittivity and electric field intensity, and using the
constitutive relations between E and D in a homogeneous isotropic medium, show that the
unit for the flux density is indeed C/m2 .
54 Chapter 1 Electromagnetic Fields

1.19 The electric field in a certain region of space is given by E(r,t) = ŷ10x2 e−α t u(t),
where u(t) is the step function.
(a) Find D(r,t) if ε (r,t) = 3ε0 y2 δ (t).
(b) Find D(r,t) if χe (r,t) = 3y2 u(t).
(c) Find D(r,t) if  
320
ε (r,t) = ε0 0 0 0 δ (t) .
040

(d) Chirality is a natural state of symmetry. Many natural substances are chiral materials,
such as DNA and many sugars. A chiral material is bi-isotropic and has the following
constitutive relation:  
∂H
D = ε E−χ .
∂t
Using Faraday’s law, find D for ε = 7ε0 and χ /µ = 10.

1.20 In a given anisotropic medium, the electric permittivity tensor in a certain coordinate
system V is given by  
7 −2 2
ε rV = −2 4 −1 .
2 −1 4
Let W be the coordinate system where ε rW is diagonal. Find ε rV and the unitary transformation
Q that describes the transformation from V to W . What type of material is the medium?
Hint: Find the eigenvalues (roots of the equation |ε rV − λ I| = 0) and then their corresponding
eigenvectors {Vn } of ε rV . Note that if the eigenvectors of ε rV do not correspond to unique
eigenvalues, they may not be mutually orthogonal. In this case, a set of orthogonal unit vectors
{Un } can be found using the Gram-Schmidt process. The transformation matrix Q can be
−1 t
constructed with the three orthonormalized vectors Q = [U1 U2 U3 ], and will satisfy Q = Q .
−1
The diagonal matrix is computed as ε rW = Q ε rV Q, which has the eigenvalues of ε rV as its
nonzero elements.

1.21 The permittivity for a linear, temporally dispersive medium, with conductivity σ and
electric susceptibility χe , is given by
 Z ∞ 
iσ iωτ
ε (ω ) = ε0 1 + + χe (τ ) e d τ .
ωε0 0

Derive the Kramers-Krönig relation for this medium. Hint: Evaluate the integral
I
ε (ω ′ ) − ε∞
dω ′
ω′ − ω
PROBLEMS 55

over an appropriate closed contour C using Cauchy’s residue theorem, which states that if
f (z) is singular at {zn }N(n=1) , then

I N
X
f (z) dz = 2π i lim (z − z0 ) f (z0 ) if {zn }Nn=1 are inside the contour C ,
C z→z0
n=1
I N
X
f (z) dz = π i lim (z − z0 ) f (z0 ) if {zn }Nn=1 are on the contour C,
C z→z0
n=1
and
I
f (z) dz = 0 if {zn }Nn=1 are outside the contour C.
C

1.22 Consider a rectangular cuboid volume of height d and base w × w containing a

homogeneous material with permittivity ε = ε0 εr . Suppose the electric field is perpendicular
to the base and is given by E = Ez ẑ, as shown in Fig. P1.22(a).

d
E = Ez ẑ
ε
w
y
w
x

Figure P1.22(a): A rectangular cuboid section of a homogeneous medium supporting a constant

electric field. The top and bottom surfaces perpendicular to the electric field can be viewed as a
parallel-plate capacitor.

(a) Noting that the displacement current in the material is defined as JD = ∂ D/∂ t, find the
equivalent capacitance of the cuboid.
(b) Now consider a medium with a periodic arrangement of dielectric slabs (parallel to the
x–z plane) as shown in Fig. P1.22(b). Consider a rectangular cuboid whose width is
w = w1 + w2 . Following a similar procedure as in part (a), find the effective dielectric
constant of this medium.
(c) If the periodicity is now along the z-direction as shown in Fig. P1.22(c), choosing the
height of the cuboid to be d = d1 + d2 , find the effective dielectric constant of the
medium.
56 Chapter 1 Electromagnetic Fields

E = Ez ẑ

Figure P1.22(b): A rectangular cuboid section of a priodic medium supporting a constant

eletcric field. The two surfaces perpendicular to the electric field can be viewed as a parallel-
plate capacitor.

E = Ez ẑ

Figure P1.22(c): A rectangular cuboid section of a priodic medium supporting a constant

eletcric field. The two surfaces perpendicular to the electric field can be viewed as a parallel-
plate capacitor.
PROBLEMS 57

1.23 To find the effective dielectric constant of a medium composed of molecules with
polarizability p and with a number density of N molecules per unit volume, the following
procedure can be followed. Consider the diagrams shown in Fig. P1.23. Part (a) shows
the molecules individually in a background dielectric ε0 , and part (b) shows an equivalent
homogeneous dielectric medium with permittivity ε and an isolated single molecule within a
void sphere.

ε0
ε Eloc
E

(a) (b )

Figure P1.23: The configuration of a heterogeneous medium composed of a background

medium with permittivity ε0 and including molecules with dipole moment p. This medium can
be viewed as a homogeneous medium with effective permittivity ε .

The constitutive relation D = ε E can also be expressed as D = ε0 E + P, where P is the

polarization vector. In the case of induced dipoles, the polarization P is proportional to the
polarizability per unit volume N p, where N is the number of dipoles per unit volume and p is
the polarizability per dipole. The polarization is given by

P = N pEloc ,

where Eloc is the electric field in the void due to the average electric field in the effective
medium. Under the quasi-static approximation, it can be shown that the local electric field is
given by
P
Eloc = E + .
3ε0
Find the effective dielectric constant of the medium (ε ). This is the well-known Clausius-
Mossotti (or Lorentz-Lorenz) formula.
1.24 A 3D array of identical metallic spheres rests in a background material with an electric
field E = E0 . The diameter of each sphere is R and the separations between adjacent spheres
in the x, y, and z directions are ax , ay , and az , respectively. The periodic structure can be
viewed as an equivalent dielectric medium. Find the relative permittivity of this structure.
Is the relative permeability a function of the radius of the spheres? (Hint: A small metallic
sphere in a medium under the influence of background electric field E gets polarized and
behaves like a dipole with dipole moment p = 4π0 ε ER3 .)
58 Chapter 1 Electromagnetic Fields

Boundary Conditions

1.25 Prove that at the interface between two dielectrics, the relations n̂ × (E1 − E2 ) = 0
and n̂ × (H1 − H2 ) = 0 are sufficient boundary conditions. That is, n̂ · (D1 − D2 ) = 0 and
n̂ · (B1 − B2 ) = 0 are satisfied automatically. (Hint: Use Cartesian coordinates with a
dielectric interface in the x–y plane.)
1.26 Prove that ∂ Dn /∂ t = 0, where Dn is the normal component of the electric flux density,
on the surface of a perfect magnetic conductor. (Hint: Start from Ampère’s law and use
Stokes’ theorem for asmall surface on the magnetic surface.)
1.27
(a) Using the equation of continuity, establish the boundary condition for volumetric
current density J at an interface between two media that can support surface charge
density ρs .
(b) If you also include surface current density Js at the surface, how will the boundary
condition change?
1.28 A magnetic sphere with relative permeability µr and relative permittivity εr = 1 rests
in free space. A static magnetic field of H0 ẑ lies within this sphere. Assume no currents or
charges exist on the surface of the sphere. Find the magnetic field on the outside surface of
the sphere Hext (θ ), as shown in Fig. P1.28.

ε0, μ0 Hext(θ)
εr y
μr
H0 ẑ
x

Figure P1.28: A spherical magnetic material with permeability µ = µ0 µr and permittivity ε = ε0

supporting a constant magnetic field inside.

1.29 Two dielectric media are in contact as shown in Fig. P1.29. An electric field E1 is
incident from medium 1 upon the interface at an angle α1 measured from the common normal.
(a) If both media are lossless dielectrics with permittivities ε1 and ε2 , find the magnitude
and direction of E2 in medium 2.
(b) If both media are lossy dielectrics with permittivities and conductivities (ε1 , σ1 ) and
(ε2 , σ2 ):
PROBLEMS 59

α1 E1

Medium 1

Medium 2

E2 α2

Figure P1.29: The interface between two dielectric media.

(i) Find the magnitude and direction of E2 in medium 2. (Hint: There are no surface
currents at the discontinuity.)
(ii) Find the surface charge density at the interface.
(iii) Compare the results with part (a).
(iv) What happens if medium 2 is an insulator (σ = 0)?

Generalized Coordinates

1.30 In cylindrical coordinates, we can define the following:

p
r = x2 + y2 ,
y
φ = arctan ,
x
z=z.

(a) Find the inverse relation for x(r, φ , z), y(r, φ , z), and z(r, φ , z).
(b) Find the scaling factors hr , hφ , and hz (also known as “metrical coefficients”).
(c) Derive expressions for each of the following operations in cylindrical coordinates.
Confirm your results with those available in Table C-2 in Appendix C.

(i) ∇ψ
(ii) ∇ · F
(iii) ∇ × F
(iv) ∇2 ψ = ∇ · ∇ψ
60 Chapter 1 Electromagnetic Fields

1.31 Let two fixed points P1 and P2 be located at x = c and x = −c along the x-axis, and let
r1 and r2 be the distances of a variable point P in the x–y plane from P1 and P2 , respectively.
Let the axes of the coordinate system be defined as

u1 = ξ , u2 = η , u3 = z ,

where
r1 + r2 r1 − r2
ξ= and η= .
2c 2c
Note that ξ ≥ 1 and −1 ≤ η ≤ 1.
(a) Find expressions for x and y in terms of ξ and η .
(b) Let z = 0. Plot a few ξ = constant curves and η = constant curves to develop a sense
for the coordinate system. (Note: Beware of sign ambiguity; your plots should cover all
four quadrants.)
(c) Find the scaling factors hξ , hη , and hz .
(d) Define each of the following operations in this coordinate system:

(i) ∇ψ

(ii) ∇ × F

(iii) ∇ · F

(iv) ∇2 ψ

1.32 Oblate spheroidal coordinates (µ , ν , φ ) belong to a system of curvilinear coordinates

in which two sets of coordinate surfaces are obtained by revolving the curves of the elliptic
cylindrical coordinates. They are related to Cartesian coordinates by the expressions

x = a cosh µ cos ν cos φ ,

y = a cosh µ cos ν sin φ ,
z = a sinh µ sin ν ,

where a is a constant, and µ ∈ (0, ∞), ν ∈ [−π /2, +π /2], and φ ∈ [0, 2π ).
(a) Show that the coordinate surfaces are:

x2 + y2 z2
for µ = constant: + = cos2 ν + sin2 ν = 1 ,
a2 cosh2 µ a2 sinh2 µ
x2 + y2 z2
for ν = constant: + = cosh2 µ − sinh2 µ = 1 .
a2 cos2 ν a2 sin2 ν

(b) Find the scaling factors hµ , hν , hφ .

PROBLEMS 61

(c) Define each of the following operations in oblate spherical coordinates:

(i) ∇ψ
(ii) ∇×F
(iii) ∇·F
(iv) ∇2 ψ

1.33 Bispherical coordinates (ξ , η , β ) are related to Cartesian coordinates by the

expressions

a sin η cos β
x= ,
cosh ξ − cos η
a sin η sin β
y= ,
cosh ξ − cos η
a sinh ξ
z= ,
cosh ξ − cos η

where a is a constant parameter and −∞ < ξ < +∞, 0 < η < π , and 0 < β < 2π .
(a) Show that:
(i) the coordinate surfaces ξ = constant represent spheres:
 2
2 2 2 a
x + y + (z − a coth ξ ) = .
sinh ξ

(ii) the coordinate surfaces η = constant represent spindle-shaped surfaces of

revolution around the z-axis:
p  2
2 2 2 2 a
( x + y − a cot η ) + z = .
sin η

(iii) the coordinate surfaces β = constant are semi-infinite planes diverging from the
z-axis.
(b) Find the scaling factors hξ , hη , hβ .
(c) Show that the coordinate surfaces are orthogonal to each other.
(d) Let x = 0. Plot a few ξ = constant curves and η = constant curves in the y–z plane to
develop a sense for the coordinate system.
(e) Derive the expression for each of the following operations in bispherical coordinates:
(i) ∇ψ
(ii) ∇×F
(iii) ∇·F
(iv) ∇2 ψ = ∇ · ∇ψ
 Chapter 2
Electromagnetic Concepts,
Tools, and Theorems

Chapter Contents Objectives

Overview, 63 Upon learning the material presented in this chapter, you

2-1 Equivalent Magnetic Charge and Current should be able to:
Densities, 63
2-2 Image Theory, 68 1. Understand the concept of “fictitious” magnetic
2-3 Method of Images for Other Problems, 71 charges and magnetic currents and how to
2-4 Polarization Current, 80 incorporate them into Maxwell’s equations.
2-5 Stored Electromagnetic Energy, 83
2-6 Flow of Energy, 86 2. Examine the duality relations between the
2-7 Superposition Principle, 90 electric and magnetic fields in a material
2-8 Uniqueness Theorem, 90 medium.
2-9 Equivalence Principle for Electromagnetic
Sources, 92 3. Apply the concept of images to charge and
Chapter Summary, 97 current sources to simplify certain types of
Problems, 99 boundary value problems.

4. Learn about energy flow and the law of conser-

vation of energy for electromagnetic waves, as
well as the conditions that make the solution to
Maxwell’s equations unique.

62
2-1 Equivalent Magnetic Charge and Current Densities 63

Overview
While undeniably elegant in their mathematical structure, Maxwell’s equations and their
associated constitutive relations do not lend themselves to straightforward solutions for the
electric and magnetic field intensities present in a material medium. In this chapter we
explore the use of certain concepts and theorems that will prove helpful in (a) developing
intuition about the behavior of the electric and magnetic fields and (b) transforming a complex
problem into a simpler version for which we have a readily available solution, or (c) to simply
be able to exploit certain symmetries that exist among the field quantities. For example,
we will introduce a new concept wherein “artificial” magnetic charges and currents are
created to make Maxwell’s equations totally symmetric, from which we derive a new duality
principle for the electric and magnetic fields. The concept of electromagnetic energy flow
and the Poynting theorem are also presented in this chapter to demonstrate how the law of
conservation of energy is indeed applicable to electromagnetic fields. The conditions required
to make the solution of Maxwell’s equations in a homogeneous region of space unique are
derived using the Poynting theorem. The concept of equivalent surface sources is presented to
show that knowledge of the field quantities outside a source region cannot be used to uniquely
identify the sources that produced those fields.

2-1 Equivalent Magnetic Charge and Current Densities

We are all familiar with the concept of electric charge density ρ and electric current density J.
It is often convenient when solving a certain class of electromagnetic problems involving
magnetic materials to introduce the concept of hypothetical or equivalent distributions of
magnetic charge density ρm and magnetic current density Jm .
The foundation for introducing such hypothetical distributions can perhaps be found in
the definition of the magnetization vector M for a magnetic material. From Eq. (1.49) and
B = µ H, the magnetization vector for a magnetic material with magnetic permeability µ is
given by  
1 µ
M= B−H = − 1 H = (µr − 1)H . (2.1)
µ0 µ0
Here, µ is the magnetic permeability of the magnetic material. To derive expressions for
the hypothetical magnetic charge density ρm and magnetic current density Jm , we create
an equivalent nonmagnetic medium with permeability µ0 and then create artificial magnetic
charges and currents in that medium so as to generate a magnetization vector equal to that of
the real magnetic material. To that end, we employ Faraday’s law in the magnetic medium:

∂B ∂H
∇×E = − = −µ0 µr . (2.2)
∂t ∂t
64 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

Next, we employ a mathematical divide to break up the quantity on the right-hand side of
Eq. (2.2) into two terms:

∂H ∂H
∇ × E = −µ0 − µ0 (µr − 1)
∂t ∂t
∂H ∂M
= −µ0 − µ0 , (2.3)
∂t ∂t
where we used Eq. (2.1) to replace the second term with the time derivative of the
magnetization vector. Equation (2.3) consists of two terms, one representing the relationship
between E and H in a nonmagnetic medium (with permeability µ0 ), and a second term
proportional to ∂ M/∂ t. This second term has the units of V/m2 , and therefore can be
interpreted as an equivalent magnetic current distribution:

∂H ∂M
Jm = µ0 (µr − 1) = µ0 . (2.4)
∂t ∂t
Using Eq. (2.4), Eq. (2.3) can be rewritten as

∂H
∇ × E = −µ0 − Jm . (2.5)
∂t
Next, we use Gauss’ law for magnetism

∇·B = 0 . (2.6)

Now consider µ0 as the permeability of an equivalent medium and then seek for a condition
that allows the magnetic field H to remain unchanged throughout the region. Equation (2.6)
can be written as
∇ · (µ0 H) = −µ0 (µr − 1)∇ · H = −µ0 ∇ · M . (2.7)
The quantity ρm = −µ0 (µr − 1)∇ · H = −µ0 ∇ · M, which has the units webers/m3 , can be
interpreted as the equivalent magnetic charge density. The preceding expression can be
rewritten as
ρm
∇ · (H) = . (2.8)
µ0
In conclusion, Maxwell’s four equations can now be written in the alternative form

∂H
∇ × E = −µ0 − Jm , (2.9a)
∂t
∂E
∇×H = ε +J , (2.9b)
∂t
ρm
∇·H = , (2.9c)
µ0
ρ
∇·E = . (2.9d)
ε
2-1 Equivalent Magnetic Charge and Current Densities 65

It should be emphasized that the notion of equivalent magnetic current density is only
applicable to time-varying field quantities. That is, if ∂ H/∂ t = 0, then Jm = 0. We also note
that in this treatment the magnetic polarization current Jm and magnetic charge density ρm
are themselves functions of the unknown magnetic field, but as will be shown later this
formulation is useful for setting up an integral equation for the unknowns. Here the four parts
of Eq. (2.9) are provided to demonstrate how hypothetical distributions of magnetic charge
and current densities can be made part of Maxwell’s equations. It should be noted that the
choice of µ0 as the background permeability was arbitrary. In fact any permeability can be
assumed as the background material simply by modifying Jm and ρm . For example, if we
desire a background permeability µ̃r and try to maintain the same H as that of the original
problem, the equivalent magnetic current and charge densities become

∂H
Jm = µ0 (µr − µ̃r ) (2.10a)
∂t
and
ρm = −µ0 (µr − µ̃r )∇ · H . (2.10b)

Hence, in general Maxwell’s equations may be written as

∂H
∇ × E = −µ − Jm , (2.11a)
∂t
∂E
∇×H = ε +J , (2.11b)
∂t
ρ
∇·E = , (2.11c)
ε
ρm
∇·H = , (2.11d)
µ

where µ = µ0 µ̃r .

2-1.1 Duality Relations

The set of equations given by Eq. (2.11a)–Eq. (2.11d) is now symmetric. In fact Eq. (2.11a)
and Eq. (2.11b) are duals of one another—and the same is true of Eq. (2.11c) and Eq. (2.11d),
in the sense that if we make the following substitutions:

E −→ H , (2.12a)
H −→ −E , (2.12b)
µ −→ ε , (2.12c)
ε −→ µ , (2.12d)
J −→ Jm , (2.12e)
Jm −→ −J , (2.12f)
66 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

ρ −→ ρm , (2.12g)
ρm −→ −ρ , (2.12h)

then Eq. (2.11a) becomes Eq. (2.11b) and Eq. (2.11b) becomes Eq. (2.11a). Also, Eq. (2.11c)
becomes Eq. (2.11d) and vice versa. This correspondence is known as the duality principle.
The significance of the duality principle is that if for a medium with parameters µ1 , ε1 , the
solutions (E1 and H1 ) are known for an excitation J, then the solution for excitation Jm = J
for a medium whose permeability is µ2 = ε1 and permittivity is ε2 = µ1 can be obtained using
the duality relations given by
E2 = −H1
and
H2 = E 1 .
The duality relations given by Eq. (2.12) are not unique. In fact, for most practical situations
these may not be appropriate. For example, the dual of free space with µ0 = 4π × 10−7 H/m
and ε0 = 8.85 × 10−12 F/m becomes a medium having a permittivity ε = 4π × 10−7 F/m and
a permeability µ = 8.85 × 10−12 H/m. It is certainly unlikely to encounter such a medium. A
more practical set of normalized duality relations is given by
r
µ
E −→ H, (2.13a)
ε
r
ε
H −→ − E, (2.13b)
µ
r
ε
J −→ Jm , (2.13c)
µ
r
µ
Jm −→ − J, (2.13d)
ε
r
ε
ρ −→ ρm , (2.13e)
µ
r
µ
ρm −→ − ρ. (2.13f)
ε
p
Note that in this case the units are preserved (the unit of µ /ε is ohms). Substituting the
duality relations given by Eqs. (2.13a)–(2.13f) into Eq. (2.11a) leads to
r  r  r
µ ∂ ε µ
∇ × H = −µ − E + J,
ε ∂t µ ε
or equivalently
∂E
∇×H = ε +J ,
∂t
2-1 Equivalent Magnetic Charge and Current Densities 67

which is Eq. (2.11b). Similarly, Eq. (2.11a) can be obtained from Eq. (2.11b) under the duality
relation given by Eqs. (2.13a)–(2.13f), and Eq. (2.11c) can be obtained from Eq. (2.11d) and
vice versa.
An important characteristic of the duality relations given by Eqs. (2.13a)–(2.13f) is that
the medium parameters are not changed.

2-1.2 Modifications to Boundary Conditions

Introducing the concept of magnetic current and charge densities into Maxwell’s equations
necessitates modification of the boundary conditions derived earlier in Section 1.6. In
the presence of a surface magnetic current density Jsm and a surface magnetic charge
density ρsm , the tangential electric field and the normal component of the magnetic flux
density across a surface boundary are no longer continuous. New boundary conditions can
be obtained by following the procedure outlined in Section 1.6 or by applying the duality
relations given by Eq. (2.12). Application of the duality approach to

n̂ × (H1 − H2 ) = Js
and
n̂ · (D1 − D2 ) = ρs

renders

n̂ × (E1 − E2 ) = −Jsm (2.15a)

and
n̂ · (B1 − B2 ) = ρsm . (2.15b)

The duality relations can also be used to introduce the concept of magnetic conductors and
perfect magnetic conductors. Starting from the point form of Ohm’s law, namely J = σ E, and
then applying the duality relations given by Eq. (2.12), we get

Jm = σm H , (2.16)

where σm represents the notion of magnetic conductivity. Furthermore, in the same way that
we define a perfect electric conductor whose conductivity σ = ∞, the notion of a perfect
magnetic conductor with σm = ∞ can be introduced as well. Inside a perfect magnetic
conductor, H = B = 0, and therefore on the surface of a perfect magnetic conductor

n̂ × H = 0 , (2.17a)
n̂ × E = −Jsm , (perfect magnetic conductor) (2.17b)
n̂ · B = ρsm . (2.17c)

These relations constitute the boundary conditions for a perfect magnetic conductor.
68 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

+q +q
n̂ n̂
d ˆ
n E=0 d ˆ
n E=0

PEC d E2 E1
-q E
(a) (b)

Figure 2-1: (a) A positive charge above a perfect electric conductor (PEC) with a planar
interface and (b) its equivalent system of charges in the equivalent homogeneous medium.

2-2 Image Theory

In certain circumstances involving boundary value problems, it is more convenient to solve
the problem in an equivalent homogeneous medium with appropriate modifications of sources
and objects than to solve the problem in its actual medium. One simple example where this
technique can be applied is a half-space medium composed of a perfect electric conductor with
a planar interface and a homogeneous dielectric half-space above it, as shown in Fig. 2-1(a).
First, let us consider a stationary positive point charge in the dielectric medium above the
perfect conductor. Knowing that the tangential electric field at the interface must be zero
(because n̂ × E = 0), and using Coulomb’s law, it can easily be shown that by placing a
negative charge of the same magnitude as that of the point charge at the mirror-image point
of the charge in the upper medium, as well as removing the conducting medium altogether,
leads to an electric field distribution in the upper medium that indeed satisfies the condition
n̂ × E = 0 along the interface points. The equivalent configuration is shown in Fig. 2-1(b).
Since the electric field in the upper-half space, including across the boundary, is the same
for both configurations, according to the uniqueness theorem (which will be introduced later),
the two configurations are identical so far as the upper space is concerned.

2-2.1 Electric Current Filament Parallel to Interface (Fig. 2-2)

Next, let us consider the case of an electric current filament (a current confined in a very thin
piece of wire) present in an upper-half space above a perfect conductor (Fig. 2-2(a)). The

Dq u I Dll Dq u I Dll
Image theory

PEC -D q u I Dll

(a) (b)

Figure 2-2: Image of an electric current over a perfect electric conductor (PEC).
2-2 Image Theory 69

current consists of a filament of length ∆ ℓ and intensity I. This current can be viewed as the
motion of a point charge ∆ q moving along the direction of ∆ ℓ with velocity u. Denoting the
time that it takes ∆ q to traverse the length of the wire by ∆ t, then u = ∆∆ℓt , and since I = ∆ q/∆ t,
it then follows that
I ∆ℓ = ∆q u .
The image of the moving charge ∆ q in a planar perfect conductor is −∆ q and it moves in
the same direction. The image theory techniques allow us to remove the perfect conductor and
to replace it with an image current (−∆ q u) flowing in a direction opposite to the direction
of the original current (Fig. 2-2(b)). Hence, the image of an electric current over a perfect
electric conductor and flowing in a direction parallel to the interface is the same current at the
mirror image point flowing in a direction parallel to the interface but in the opposite direction.
In general, the current filament can be oriented in any direction relative to the planar interface.
Such a current can be divided into two orthogonal components, one parallel to the interface
and the other perpendicular to the interface. Hence, we also have to consider the current
filament oriented perpendicular to the interface.

2-2.2 Electric Current Filament Normal to Interface (Fig. 2-3)

v v
Dq I Dll I Dll
Image theory Dq

PEC -Dq
I Dll
v
(a) (b)

Figure 2-3: The image of a current filament normal to a PEC is the same current at the mirror
image point along the same direction.

Application of image theory to the configuration shown in Fig. 2-3(a) leads to removal of
the perfect conductor and replacing it with charge −∆ q at the image location of charge ∆ q,
and as ∆ q moves upwards, −∆ q has to move downward to maintain image location. Using
the equivalent moving charge concept for current filament, it is now obvious that the image of
a current flowing normal to the interface is the same current at the mirror image point flowing
in the same direction as the original source.

2-2.3 Magnetic Current Filament Parallel to Interface (Figs. 2-4 and 2-5)
In many radiation and scattering problems, magnetic current sources are encountered and
therefore it is also important to obtain the image of magnetic currents on planar metallic
surfaces. In finding the image representation of magnetic currents the duality principle is
used. In addition, we use the fact that the normal component of time-varying magnetic
fields vanish on the surface of a perfect electric conductor. Applying the duality principle
70 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

Im Dll H

Im Dll

Figure 2-4: Magnetic duality applied to the electric charge configuration shown in Fig. 2-2.

(Fig. 2-4) to the problem of two parallel electric currents of equal magnitude and opposite
direction, it becomes evident that two parallel magnetic current filaments flowing in opposite
directions produce a vanishing tangential magnetic field and a normal electric field along a
plane that bisects the medium between the two sources. Conversely, if the directions of the
magnetic currents are the same, the normal component of the magnetic field and the tangential
component of the electric field will vanish. The latter is the boundary condition for a perfect
electric conductor (PEC). Hence, it can be concluded that the image of a magnetic current
filament parallel to a planar PEC is a magnetic current filament that flows in the same direction
as the original source and located at the mirror image point (Fig. 2-5).

PEC
Im Dll

(a) (b)

Figure 2-5: Equivalent problems (image theory) for a horizontal magnetic current over a planar
perfect electric conductor.

2-2.4 Magnetic Current Filament Normal to Interface (Fig. 2-6)

To find the image of a perpendicular magnetic current filament over a planar PEC, we can
use a similar procedure based on the duality principle applied to two normal electric currents
of equal magnitude along the same line and the related boundary conditions. Using the same
argument as before, it can be shown that the image of a normal magnetic current filament is
a normal magnetic current with the same magnitude at the mirror image point flowing in the
opposite direction. The fields produced by this configuration of the magnetic filament currents
satisfy the boundary conditions in the perpendicular bisector plane of the line connecting the
source and its image.
2-3 Method of Images for Other Problems 71

Im Dll Im Dll

PEC
Im Dll

(a) (b)

Figure 2-6: Equivalent problems (image theory) for a vertical magnetic current over a planar
perfect electric conductor.

I Dll Im Dll I Dll Im Dll

I Dll Im Dll I Dll Im Dll

PMC I Dll Im Dll

I Dll Im Dll
(a) (b)

Figure 2-7: Image theory for elementary electric and magnetic currents over a planar perfect
magnetic conductor (PMC).

2-2.5 Electric and Magnetic Current Filaments over a Planar Perfect

Magnetic Conductor (Fig. 2-7)
The dual of the perfect electric conductor is the perfect magnetic conductor (PMC) for which
n̂ × H = 0. Also, for time-varying fields, the normal component of the electric fields on PMC
is zero. To find the images of electric and magnetic currents over perfect magnetic conductors,
the duality relations can be applied to the appropriate cases of currents over a perfect electric
conductor. Figure 2-7 shows the images of electric and magnetic sources over a PMC. It
should also be mentioned that objects in the upper-half space above a planar PEC or PMC are
also mirror-imaged. The reason for this is that objects, whether being conductor or dielectric,
can be viewed as surface or polarization current distributions and can be imaged using image
theory. For example, the equivalent of the combination of a metallic object and a current
source above a PEC, as shown in Fig. 2-8(a), consists of the source, the object, and their
images, as shown in Fig. 2-8(b). The method described here is based on the electrostatic
concept for charges. However, the results are equally valid for time-varying fields. This can be
easily verified once we derive the fields radiated by short current filaments (Hertzian dipole).

2-3 Method of Images for Other Problems

In the previous section, image theory was developed for perfect electric and magnetic planar
surfaces of infinite extent. The method of images can be developed for a few other objects in
the electrostatic or quasi-static regime. For example, for a metallic sphere or a half-space
72 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

I Dll PEC I Dll PEC

PEC PEC
I Dll

(a) (b)

Figure 2-8: Equivalent problems for a current source and an object above a perfect electric
conductor plane.

dielectric, the image of an external point charge can be obtained using electrostatic field
analysis. That is, the value(s) and the location(s) of image charge(s) are sought for in order
to establish an equivalence between the electrostatic field produced by the original boundary
value problem and those produced by the original point charge and its image charge(s).

Example 2-1: Reverse Application of Image Theory

Use image theory to show that the x-component of the electric field generated by a constant
loop current centered at the origin of a Cartesian coordinate system and placed in the y-z plane
is zero. The geometry of the problem is shown in Fig. 2-9.
Solution: The electric current in the loop has circular symmetry. For every point on the loop
the following relations exist between the current and its counterpart at the opposite point with
respect to the x–y plane:
 π
Iz (θ ) = Iz θ +
2
and
 π
Iy (θ ) = −Iy θ + .
2
This pair of currents defines the relation between a current filament above a perfect electric
conductor and its image. Hence the equivalent for a loop of electric current is a half-loop over
a PEC surface in the x–y plane. Now if we rotate the y-z plane about the x-axis to form a new
Cartesian coordinate system xy′ z′ (see Fig. 2-9), the same argument can be used to prove that
the tangential electric field is zero on the x-y′ plane. From this argument it can be concluded
that the electric field produced by the current loop is always perpendicular to the x-axis and
has circular symmetry with respect to the x-axis.

2-3.1 Image of a Point Charge in a Grounded Metallic Sphere

Next, let us consider a metallic sphere of radius a centered around the origin of a Cartesian
coordinate system and a point charge q located on the z-axis at Z0 outside the metallic sphere,
2-3 Method of Images for Other Problems 73

x
z

Et
y

x
z′ z

y′

Et y
x

Figure 2-9: Reverse application of image theory is used for a wire loop carrying a constant
current I to show that the electric field generated from such a source is φ -directed.
74 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

z
q
Z0

q′
Z0′

θ
r
ϕ a
y

Figure 2-10: A point charge q outside a grounded metallic sphere of radius a. Using image
theory for a static electric field, the conducting sphere can be removed and an image charge
q′ = −qa/Z0 can be substituted at Z0′ = a2 /Z0 .

as shown in Fig. 2-10. Let us first consider a situation where the metallic sphere is grounded
(at zero potential). In this case, for a positive point charge, the sphere will be negatively
charged and all the positive charges from the sphere are pushed into the ground terminal.
Because of the φ -symmetry of the problem, if one can find a point image charge, it should
lie on the z-axis. Consider a point charge q′ on the z-axis at Z0′ inside the metallic sphere,
representing the image of q inside the metallic sphere. That is, by placing q′ on the z-axis at
Z0′ , we can remove the metallic sphere altogether, and therefore the fields produced by the
combination of q and q′ should be the same as the fields in the original problem insofar as
the field at points outside the sphere are concerned. It is emphasized that the image method is
applicable only if the sum of the fields from these two charges satisfy the boundary condition
on the surface of the metallic sphere. Choosing an arbitrary point on the surface of the sphere
at r = a(sin θ cos φ x̂ + sin θ sin φ ŷ + cos θ ẑ) and calculating the total electric potential there,
we have
q q′
Φ(r) = + . (2.18)
4πε |r − Z0 ẑ| 4πε |r − Z0′ ẑ|
2-3 Method of Images for Other Problems 75

This potential can be rewritten as

q/a q′ /Z0′
Φ(r) = + , (2.19)
4πε |r̂ − (Z0 /a)ẑ| 4πε |(a/Z0′ )r̂ − ẑ|

where r̂ = r/a. Inspection of this equation reveals that if

q q′
=− ′ (2.20a)
a Z0

and
Z0 a
= ′ =α , (2.20b)
a Z0
and realizing that
|r̂ − α ẑ| = |α r̂ − ẑ| = (1 − 2α cos θ + α 2 )1/2 ,
then it is obvious that the potential Φ(r) = 0 everywhere over the metallic sphere. Hence the
location of the image is at
a2
Z0′ = (2.21a)
Z0

and the value of the image charge is

a
q′ = −q . (2.21b)
Z0

2-3.2 Image of a Point Charge in an Isolated Metallic Sphere

Now we consider the case of a point charge outside of an isolated metallic sphere (Fig. 2-11).
In this case, placement of a point charge near the metallic sphere polarizes the sphere with a
surface charge density of opposite polarity concentrated near the charge and a surface charge
density of the same polarity concentrated on the other side of the sphere. For this problem,
according to the law of conservation of charge, the net charge on the sphere should add up to
zero. The boundary condition still mandates that the potential on the surface of the sphere be
a constant. From the previous problem, we concluded that the charge and its image canceled
each other’s electric potentials on the sphere’s surface. A point charge placed at the origin
creates a constant potential on the sphere’s surface and if placed anywhere else, the electric
potential on the surface of the sphere will not be constant. On the other hand, the total charge
of the image charges must add up to zero. Hence, the image of a point charge outside a metallic
sphere is composed of two equal point charges of opposite polarity, one located at Z0′ = a2 /Z0
with charge q′1 = −qa/Z0 and the other located at the origin with charge q′2 = qa/Z0 . Hence,
the polarized metallic sphere acts like a dipole with a dipole moment given by

a3
p = q′ Z0′ = q . (2.22)
Z02
76 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

z
q

q1 =

q2 = y

x
Figure 2-11: A point charge q outside an isolated metallic sphere of radius a. Also shown are
two point charges (image charges) q1 and q2 , which can be used to replace the metallic sphere
charges q, q1 , and q2 and produce the same field outside the sphere as the field generated by the
original configuration.

The overall electric field can be obtained from the superposition of the fields generated
by the three point charges, and is valid only outside or over the surface of the metallic sphere.
The surface charge density on the sphere, as dictated by boundary condition, can be calculated
from the normal component of the electric flux and is only a function of θ due to the azimuthal
symmetry of the problem. That is,

ρs (θ ) = ε Er (a, θ ) . (2.23)

Using Coulomb’s law, the explicit expression for ρs (θ ) is given by

"    2 #
q 1 − α cos θ a a α − cos θ 1
ρs (θ ) = − + .
4π a2 (1 + α 2 − 2α cos θ )3 Z0 Z0′ (1 + α 2 − 2α cos θ )3 α

Further simplification leads to

 
q 1 − α2 1
ρs (θ ) = + . (2.24)
4π a2 (1 + α 2 − 2α cos θ )3 α
Figure 2-12 shows a plot of the normalized surface charge density on an isolated metallic
2-3 Method of Images for Other Problems 77

0.5

−0.5
(4πa2/q)ρs

−1

−1.5
Z0 Z=0=2a
2a
−2 Z0 =0 3a
Z =3a

−2.5
0 20 40 60 80 100 120 140 160 180
θ (degrees)

Figure 2-12: Normalized surface charge density (4π a2/q)ρs due to a point charge q located at
Z0 = 2a and a second plot for a charge at Z0 = 3a.

sphere as a function of θ for a point charge located at Z0 = 2a, and another plot for a charge
located at Z0 = 3a.

Example 2-2: Force on a Charge outside a Metallic Sphere

Consider a positive point charge q outside a metallic sphere centered around the origin and
of radius a. Let’s assume the charge is located on the z-axis at z = Z0 , as shown in Fig. 2-11.
This charge is attracted towards the metallic sphere. What is the force on charge q?
Solution: The advantage of image theory is obvious in solving this problem. The equivalent
problem is realized by replacing the metallic sphere with the image charges −qa/Z0 and
qa/Z0 on the z-axis at a2 /Z0 and the origin, respectively. The force on q can easily be
computed from the fields generated by the two image charges. Using Coulomb’s law, the
electric field at z = Z0 is given by
   
qa qa
−
1  Z0 Z0 

E(Z0 ) =  2 + 
4πε  a2 Z02 
Z0 −
Z0
q a3 (2Z02 − a2 )
= (−ẑ) ,
4πε Z03 (Z02 − a2 )2
78 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

and therefore the attractive force is given by

q2 a3 (2Z02 − a2 )
F(Z0 ) = − ẑ .
4πε Z03 (Z02 − a2 )2

2-3.3 Image of a Point Charge over a Half-Space Dielectric Medium with a

Planar Interface
By inspection or by solving Poisson’s equation directly, it can be shown that the image of a
point charge q in a medium with permittivity ε1 above a half-space dielectric medium with
permittivity ε2 and a planar interface between the two media (Fig. 2-13) is a point charge q′
located at the mirror image point (below the interface and equal distance from the interface)
having a charge equal to
ε2 − ε1
q′ = −q . (2.25a)
ε2 + ε1
The equivalence is for a homogeneous medium having a permittivity equal to that of the
upper half-space (ε1 )when the observation point is in the upper half-space. However, if the
observation point is in the lower half-space, the image point is at the location of the source
and the image charge is
ε2 − ε1
q′′ = q . (2.25b)
ε2 + ε1
That is, the total charge is q + q′′ = 2ε2 /(ε1 + ε2 ) and located at the source location in a
homogeneous medium with permittivity ε2 . It can easily be shown that both the electric
potential and the normal component of the electric flux at any point on the interface right
above and right below the interface are equal, as required by the boundary conditions between
two dielectric media. This establishes the equivalence between the original problem and the
corresponding image problem.

Example 2-3: Force on a Charge above a Half-Space Dielectric

For the point charge in a dielectric medium composed of two half-space media with a planar
interface, as shown in Fig. 2-13(a), find the force on the charge. Is the force attractive or
repulsive? Determine conditions under which the forces are attractive or repulsive.
To find the force on charge q in the upper half-space with dielectric constant ε1 , image
theory as shown in Fig. 2-13(b) is used. Assuming the charge is located at z = Z0 , the electric
field generated by the image charge at z = Z0 is given by

q(ε2 − ε1 ) 1
E(Z0 ) = − ẑ ,
4πε1 (ε2 + ε1 ) 4Z02

and hence the force is

q2 (ε2 − ε1 ) 1
F(Z0 ) = − ẑ .
4πε1 (ε2 + ε1 ) 4Z02
2-3 Method of Images for Other Problems 79

z
E1(r)

ε x

ε
E2(r)

z
E1(r)

ε
x

ε ε −ε
ε +ε

z
2ε
ε +ε
ε
x

ε
E2(r)

Figure 2-13: (a) A point charge in a dielectric medium comprising two half-space media with a
planar interface. (b) The original charge and its image, placed in a homogeneous medium having
the same permittivity of the upper half-space, produce the same field as the original configuration
so far as the points in the upper half-space are concerned. (c) The original charge and its image,
placed in a homogeneous medium having the same permittivity of the lower half-space, produce
the same field as the original configuration so far as the points in the lower half-space are
concerned.
80 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

The force is attractive if ε2 > ε1 and is repulsive if ε2 < ε1 . It is also interesting to note that if
ε2 is negative then the force is repulsive if |ε2 | < ε1 and attractive if |ε2 | > ε1 . An interesting
condition is if ε2 = −ε1 , which makes the force infinitely large. In the neighborhood of
ε2 = −ε1 , this large force can change direction from attractive to repulsive and vice versa.

2-4 Polarization Current

To introduce the concepts of hypothetical magnetic current and magnetic charge densities,
it was demonstrated in Section 2-2 that the medium permeability can be replaced with a
material having an arbitrary permeability, provided that a volumetric magnetic current density
proportional to the time rate of change of the magnetic field is embedded throughout the
material. The same approach can be applied to the electric field in a dielectric medium.
Suppose we try to establish the same electric field E that exists in an equivalent hypothetical
medium with permittivity ε throughout a medium with a different permittivity ε0 . We start by
considering the modified version of Ampère’s law:

∂E
∇×H = ε +J , (2.26)
∂t
which can be rewritten as
∂E ∂E
∇ × H = ε0 + ε0 (εr − 1) +J . (2.27)
∂t ∂t
The middle term on the right-hand side of the preceding equation is equivalent to a
polarization current Jp :
∂E ∂P
Jp = ε0 (εr − 1) = , (2.28)
∂t ∂t
and an associated polarization vector

P = ε0 (εr − 1) E . (2.29)

Next we consider Gauss’ law:

∇·D = ρ ,
which can be rewritten as
ρ
∇·E = − (εr − 1)∇ · E
ε0
ρ 1
= − ∇·P ,
ε0 ε0
where use was made of Eq. (2.29). Given that the first term is ρ /ε0 , the second term is
equivalent to a polarization charge density

ρp = −∇ · P = −ε0 (εr − 1)∇ · E . (2.30)

2-4 Polarization Current 81

E, H E, H
E, H Jp, Jm

μ, ε μ0, ε0
μ0, ε0 μ0, ε0
Ω Ω

(a) (b)

Figure 2-14: (a) A background medium with permittivity and permeability ε0 and µ0 and an
object specified by domain Ω having permittivity ε and permeability µ . Certain sources outside or
inside the object establish electric and magnetic fields E(r) and H(r) throughout the space. (b) An
equivalent problem where the object is replaced with electric and magnetic polarization currents
within the Ω domain in a homogeneous medium with permittivity ε0 and permeability µ0 .

In general, a material with permeability µ and permittivity ε can be replaced with

vacuum (having permittivity ε0 and permeability µ0 ), and containing magnetic and electric
polarization current densities given by

∂E
Jp = ε0 (εr − 1) (2.31a)
∂t
and
∂H
Jm = µ0 (µr − 1) . (2.31b)
∂t
These polarization currents are distributed throughout the space where the material existed.
This equivalence is shown in Fig. 2-14, where an object with permittivity ε and permeability
µ enclosed within a background medium with permittivity ε0 and permeability µ0 is replaced
with a homogeneous medium with permittivity ε0 and permeability µ0 , and containing
distributions of electric and magnetic polarization currents across the region where the
object existed. It should be noted that a-priori knowledge of the electric and magnetic fields
throughout the object is needed to set up the equivalent currents. The original sources that
established the fields originally and the polarization currents in the homogeneous background
medium will produce the same electric and magnetic fields everywhere as those of the original
problem.
It should be noted that the polarization currents and charge densities satisfy the continuity
relations, which is a result of the law of conservation of charge. That is,

∂
∇ · Jp = ε0 (εr − 1) ∇·E
∂t
∂ ∂
=− [−ε0 (εr − 1)∇ · E] = − ρp .
∂t ∂t
82 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

Using duality, it can also be shown that

∂
∇ · Jm = − ρm . (2.32)
∂t

2-4.1 The Relationship between the Electric and Magnetic Polarization

Currents
An important point to consider is that the electric and magnetic polarization currents are not
independent of each other. Let us denote the domain in which the material exists by Ω and
define a step position function Ψ(r) as
(
1 r∈Ω,
Ψ(r) = (2.33)
0 r<Ω.

With this definition, Jp (r) and Jm (r) can be defined in the entire domain using

∂E
Jp (r) = ε0 (εr − 1) Ψ(r) (2.34a)
∂t
and
∂H
Jm (r) = µ0 (µr − 1) Ψ(r) . (2.34b)
∂t
Taking ∇× of both sides of Eq. (2.34a) yields
 
∂E ∂
∇ × Jp = ε0 (εr − 1) ∇Ψ × + Ψ(r) ∇ × E . (2.35)
∂t ∂t
But for r in Ω
∂H µ
Ψ(r) ∇ × E = −µ0 Ψ(r) − Jm = − Jm (r) .
∂t µ − µ0
Also,
∇Ψ = −n̂ δ (|r − r′ |) ,
where r′ is a position vector on the surface of Ω and n̂ is an outward unit normal to the surface
of Ω. Hence the first term in Eq. (2.35) can be interpreted as a surface current density given
by
Js = −n̂ × Jp δ (|r − r′ |)
The relationship between Jm and Jp is now given by

1   µ ∂
∇ × Jp + n̂ × Jp δ (|r − r′ |) = − Jm (r) . (2.36)
ε0 (εr − 1) µ − µ0 ∂ t
2-5 Stored Electromagnetic Energy 83

Using the duality relations, the relationship between Jp and Jm is given by

1   ε ∂
∇ × Jm + n̂ × Jm δ (|r − r′ |) = Jp (r) . (2.37)
µ0 (µr − 1) ε − ε0 ∂ t
Equations (2.36) and (2.37) establish the relationship between the magnetic and electric
polarization currents.

2-5 Stored Electromagnetic Energy

In the static or quasi-static regime, such that the time derivative of the field quantities can
be set to zero, the coupled Maxwell’s equations become decoupled. Thus, Faraday’s law and
Ampère’s law reduce to

∇×E = 0 (2.38a)
and
∇×H = J . (2.38b)

Since ∇ × E = 0 and ∇ × ∇Φ = 0 for any scalar function Φ, the electric scalar potential Φ(r)
is defined such that
E(r) = −∇ Φ(r) . (2.39)
The choice of negative sign is a convention that ensures the electric field direction points
toward regions of low potential from regions of high potential. Inserting Eq. (2.39) into Gauss’
law (Eq. (1.1c)) yields the Poisson equation for the scalar electric potential:

ρ (r)
∇2 Φ(r) = − . (2.40)
ε
The solution of the Poisson equation, subject to the appropriate boundary conditions, provides
the scalar potential function in the desired domain, and then using Eq. (2.39) the electric
field can be derived for any specified volumetric charge distribution. The simplest problem to
consider is a point charge q0 located at the origin of a coordinate system in a homogeneous,
unbounded medium with permittivity ε . This problem was solved in the preceding chapter
using Gauss’ law, from which the electric field of a point charge was found to be
q0
E(r) = r̂ .
4πε r2
Since the electric field has no dependence on the other spherical coordinate parameters
(θ and φ ), it is reasonable to expect the potential function Φ(r) to also be independent of
θ and φ . Using Eq. (2.39), it can be shown that
∂Φ
= −Er (r) ,
∂r
84 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

from which we can find the potential of the point charge to be

q0
Φ(r) = + Φ0 . (2.41)
4πε r
Here, Φ0 is a constant resulting from the integration. This constant can be determined to
be zero (Φ0 = 0) by specifying the potential to be zero at r = ∞. For a volumetric charge
distribution, Eq. (2.41) can be generalized to

ρ (r′ )
$
1
Φ(r) = ′|
dυ ′ .
4πε V |r − r

The amount of work required to bring another charge q from r = ∞ to location r can be
calculated from Z r Z r
W =− qE · dℓ = q∇Φ · dℓ .
∞ ∞
This is the potential energy stored in the system composed of two charges. Noting that
∇Φ · dℓℓ = (∂ Φ/∂ ℓ ) dℓℓ, it can easily be shown that

W = q Φ(r) . (2.42)

Next, consider N point charges, all in the same medium, at locations rn . The total electric
potential associated with all these charges at location rN+1 (rN+1 , rn , ∀n ∈ {1, 2, . . . N}) can
be obtained using the superposition principle:
N
1 X qn
Φ(rN+1 ) = .
4πε |rN+1 − rn |
n=1

Moreover, the potential energy required to bring an additional point charge qN+1 from infinity
to location rN+1 is given by
N
qN+1 X qn
WN+1 = .
4πε |rN+1 − rn |
n=1

Now the total potential energy can simply be calculated from

N
X N k
1 X X qn qk+1
W= Wk+1 = (2.43)
4πε |rk+1 − rn |
k=1 k=1 n=1
2-5 Stored Electromagnetic Energy 85

It is interesting to note that for the summand in Eq. (2.43), the indices are interchangeable
without affecting the result (symmetric), which allows us to rewrite Eq. (2.43) as

 
N X
X N
1 1 qn qk+1 
W=  . (2.44)
2 4πε |rk+1 − rn |
k=1 n=1
n,k

This representation for the potential can be easily extended to a continuous charge
distribution ρ (r) and can be written as

ρ (r) ρ (r′ )
$ $
1
W= dυ ′ dυ .
2 4πε |r − r′ |
One of the integrals is recognized as the potential function; that is,
$
1
W= ρ (r) Φ(r) d υ . (2.45)
2
Substituting for ρ from the Poisson equation given by (2.40),
$
−ε
W= Φ(r) ∇2 Φ(r) d υ .
2
This can be further modified by using the identity

∇ · (Φ∇Φ) = ∇Φ · ∇Φ + Φ∇2Φ ,

which leads to $ $
ε 2 1
W= |∇Φ(r)| d υ = ε |E(r)|2 d υ . (2.46)
2 2
In the derivation of Eq. (2.46), we used the divergence theorem and the fact that the potential
at infinity is zero: $
∇ · (Φ∇Φ) d υ = Φ∇Φ · dS = 0 .
S∞
Examination of Eq. (2.46) reveals that the quantity
1
we (r) = ε |E(r)|2 (2.47)
2
can be interpreted as the electric energy density (stored electric energy per unit volume). In
cases where the medium is anisotropic, it should be recognized that we = 21 ε E(r) · E(r) =
86 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

1
2 D(r) · E(r). Hence the electric energy density in its most general form is given by

1
we (r) = D(r) · E(r) . (2.48)
2

A similar procedure can be followed to derive a relationship for the magnetic energy
density by defining the magnetic scalar potential as the dual of the electric scalar potential.
Applying duality to Eq. (2.48) leads directly to

1
wm (r) = B(r) · H(r) . (2.49)
2

2-6 Flow of Energy

It is well established that electromagnetic waves carry power. Of the power radiated by
the sun, only a very small fraction reaches planet Earth, and consists of a wide spectrum
of electromagnetic waves. To understand how this power is carried by EM waves, in this
section we apply Maxwell’s equations to derive expressions for power density, power loss,
and stored energy. Let us consider a simple region V enclosed by a closed surface S, as shown
in Fig. 2-15.
The medium inside S is considered to be a homogeneous medium with constitutive
parameters µ , ε , and σ . Enclosed inside S are electric and magnetic current distributions

ε, μ, σ
S

Jm ds

V1
J
V2 V

Figure 2-15: An enclosed homogeneous region containing volumetric electric and magnetic
current distributions J and Jm , respectively.
2-6 Flow of Energy 87

J(r) and Jm (r). According to Maxwell’s equations

∂B
∇×E = − − Jm (2.50)
∂t
and
∂D
∇×H = + σE + J . (2.51)
∂t
Here J and Jm are the impressed electric and magnetic currents (source functions).
Application of the inner product on both sides of Eq. (2.50) by H and on both sides of
Eq. (2.51) by E, followed by subtracting the results leads to

∂D ∂B
E·∇×H−H·∇×E= E· + σE ·E + J·E + H · + Jm · H . (2.52)
∂t ∂t
Noting that
∇ · (E × H) = H · ∇ × E − E · ∇ × H ,

∂D 1 ∂
E· = (E · D) ,
∂t 2 ∂t
and
∂B 1 ∂
H· = (H · B) ,
∂t 2 ∂t
Eq. (2.52) can be rearranged to form

1 ∂
∇ · (E × H) + (E · D + H · B) + σ |E|2 = −J · E − Jm · H . (2.53)
2 ∂t
By applying the volume integral to both sides of Eq. (2.53) and then applying the divergence
theorem to the first term of the integrand, it can be shown that
$   $
∂ E·D H·B
E × H · ds + + dυ + σ |E|2 d υ
S ∂t V 2 2 V
$ $
=− J · E dυ − Jm · H d υ . (2.54)
V1 V2

Recalling that the units of J and Jm are A/m2 and V/m2 , respectively, and the units of E
and H are V/m and A/m, the unit of the integrals is watts. The right-hand side of Eq. (2.54)
is the total power delivered by sources J and Jm to the system. The unit of σ |E|2 is W/m3 ,
which represents the instantaneous power loss in the system. Hence

$
PL = σ |E|2 d υ (power loss) (2.55)
V
88 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

is the total ohmic loss in the system due to the medium’s finite conductivity σ . As shown in the
previous section, in electrostatic and magnetostatic regimes, the stored electric and magnetic
energies per unit volume can be computed from

1 1
we = E · D = ε |E|2 (electric energy density)
2 2
and
1 1
wm = H · B = µ |H|2 . (magnetic energy density)
2 2
Hence
$  
∂ E·D H·B
Ps = + dυ (stored power)
∂t V 2 2

can be interpreted as the instantaneous time rate of change of stored electric and magnetic
energies. The unit of quantity
S(t) = E × H (2.56)
is W/m2 . Considering that ds is an outward normal to surface S, the first term on the left-hand
side of Eq. (2.54) can be interpreted as the net flow of power leaving the system through the
closed surface S. The vector quantity S(t) is known as the time-dependent Poynting vector,
which may be interpreted as power flow density. Equation (2.54) states the law of conservation
of power for electromagnetic waves. It asserts that the power generated by the sources is
consumed by ohmic loss, radiated out, and/or getting used to store energy in the medium.
In antenna and scattering problems the integral

Prad = E × H · ds (2.57)

is used to calculate the power radiated by a system of sources.

Example 2-4: Flow of Energy in a Transmission Line

As shown in Fig. 2-16, in a coaxial line with an inner radius of a and an outer radius of b, and
filled with a material with permittivity ε and permeability µ , the electric and magnetic fields
are given by
V0
E(ρ , φ , z,t) = cos(ω t − kz) ρ̂
ρ
ρ ln(b/a)
and
V0
H(ρ , φ , z,t) = cos(ω t − kz) φ̂
φ,
2πη0 ρ
2-6 Flow of Energy 89

Figure 2-16: A coaxial transmission line supporting a traveling electromagnetic wave.

where η0 is the characteristic impedance of the line given by

r  
1 µ b
η0 = ln , (coaxial line)
2π ε a

and V0 is the amplitude of the voltage difference between the inner and outer conductors. Find
the time-average power carried by the coaxial line.

Solution: To find the time-average power crossing a surface along the coaxial line, say at
z = z0 , we need to find the power density S(r,t) at z0 :

S(r,t) = E(ρ , φ , z0 ,t) × H(ρ , φ , z0 ,t)

V02 cos2 (ω t − kz0 )
= ẑ .
2π ln(b/a) η0 ρ2
The total power crossing the coaxial line’s cross section at z = z0 is given by
Z b Z 2π
V02 cos2 (ω t − kz0 )
P(t) = ρ dφ dρ
a 0 2π ln(b/a) η0 ρ2
V02
= cos2 (ω t − kz0 ) .
η0
The time-average power (net power) at z0 is obtained by
Z T 2
1 V0
Pavg = cos2 (ω t − kz0 ) dt ,
T 0 η0
90 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

2π
where T = ω is one period of cos(ω t − kz0 ), and therefore

1 V02
Pavg = .
2 η0

2-7 Superposition Principle

The superposition principle, in general, can be applied to physical problems whose input and
output are governed by linear relationships. The curl, divergence, and gradient that are based
on spatial differentiations are all linear operators, and thus Maxwell’s equations, in general,
satisfy the condition for linearity. However, the constitutive relations that relate the electric
and magnetic flux densities to the corresponding electric and magnetic field intensities can be
nonlinear, depending on the material composition and the magnitude of the field intensities in
such media. As discussed in Chapter 1, under the small-signal approximation, the constitutive
relations are linear. Considering currents and charges as input and the electric and magnetic
fields throughout the region of interest as output, the superposition principle states that if
E1 (r) and H1 (r) are produced by J1 (r) and J1m (r) according to

∇ × H1 (r) = −iωε E1 (r) + J1 (r)

and
∇ × E1 (r) = iω µ H1 (r) − J1m (r),

and E2 (r) and H2 (r) are produced by J2 (r) and J2m (r) according to

∇ × H2 (r) = −iωε E2 (r) + J2 (r)

and
∇ × E2 (r) = iω µ H2 (r) − J2m (r),

then it follows that J1 (r) + J2 (r) and J1m (r) + J2m (r) produce E1 (r) + E2 (r) and
H1 (r) + H2 (r).

2-8 Uniqueness Theorem

As will be shown later, computation of field quantities generated from electric and magnetic
sources present in a homogeneous and unbounded medium can be obtained using electric and
magnetic vector and scalar potentials. The same potentials may be used for an inhomogeneous
medium made up of a finite number of homogeneous media. To solve such problems, we need
to specify the conditions required to make the solution in any given region unique. Let us
consider a region with constitutive parameters ε , µ , and σ , bounded by a closed surface S
and containing arbitrary electric and magnetic sources J and Jm . Suppose there are two sets
of solutions that satisfy Maxwell’s equations. Let us denote the first set by E1 and H1 and the
second set by E2 and H2 . Then, according to the superposition principle, E = E1 − E2 and
2-8 Uniqueness Theorem 91

H = H1 − H2 must satisfy the source-free Maxwell’s equations, that is,

∂H
∇ × E = −µ (2.58)
∂t
and
∂E
∇×H = ε + σ E. (2.59)
∂t
Applying the Poynting theorem to the source-free Maxwell’s equations given by Eq. (2.58)
and Eq. (2.59) leads to
$   $
∂ 1 2 1 2 2
ε |E| + µ |H| d υ = − σ |E| d υ − (E × H) · ds . (2.60)
∂t V 2 2 V S

If we assume n̂ × E = 0 or n̂ × H = 0 on S, then

(E × H) · ds = (n̂ × E) · H ds = − (n̂ × H) · E ds = 0 .
S

That is, if
n̂ × E1 = n̂ × E2 or n̂ × H1 = n̂ × H2 ,
then the surface integral vanishes. Also if n̂ × E = 0 over a portion of S and n̂ × H = 0 on
the rest, the surface integral vanishes as well. Under any of these conditions the right-hand
side of Eq. (2.60) is always a nonpositive quantity. As a result, if E1 = E2 and H1 = H2 at
t = 0, that is, E(t = 0) = H(t = 0) = 0, it can be concluded that E = H = 0 for all values of
t > 0 and at every point in the domain V . The conclusion is based on the argument that the
integrand of the left-hand side of Eq. (2.60) is always a positive quantity and its initial value
is zero, and hence it cannot have a negative time rate of change. So it is concluded that both
sides of Eq. (2.60) are identically zero.
The above-stated conditions are the necessary and sufficient conditions for making the
solution of Maxwell’s equations unique in region V . That is, the solution to Maxwell’s
equations is unique if the initial values of E and H are known throughout the region and
if the tangential component of E or the tangential component of H is known across a closed
surface, or the tangential component of E is known across part of the surface and the tangential
component of H is known across the remaining part of the surface.
If the medium is unbounded and the surface S approaches infinity, the assumption of
having a finite amount of conductivity is required to ensure that E and H approach zero on
the surface S. Alternatively, the application of the radiation boundary condition that will be
introduced later can be used to prove that the surface integral vanishes when the surface S is
sufficiently far away from the sources.
Oftentimes, when developing a solution involving partial differential equations, the total
tangential component of E or H is not known, but through the imposition of the boundary
conditions across the discontinuities, the uniqueness of the solution throughout the entire
region can be established. In certain cases n̂ × E = 0 for a perfect conductor or n̂ × H = 0 for
92 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

a perfect magnetic conductor, so the values of the tangential components are known per the
boundary conditions for those surfaces.

2-9 Equivalence Principle for Electromagnetic Sources

The electromagnetic fields established outside a source region can be reproduced by an infinite
number of combinations of multiple sources external to the desired region. To demonstrate
the validity of this statement, let us first consider a set of electric and magnetic current
distributions J and Jm confined within a volume V , as shown in Fig. 2-17. According to the
uniqueness theorem, the tangential components of the electric and magnetic fields on a closed
surface S are sufficient to determine the fields outside of S. For example, if we consider a
scenario where the fields inside S are zero and the fields outside of S are the original fields,
according to the boundary conditions for EM fields, we must establish tangential electric and
magnetic current sources on S given by

Js = n̂ × H (2.61a)
and
Jms = −n̂ × E , (2.61b)

in order for the original tangent electric and magnetic fields to remain the same. The equivalent
sources given by Eq. (2.61a) and Eq. (2.61b) are only one of many possible sources that we
can choose to satisfy the condition for finding the original fields exterior to S. In fact we can
place any arbitrary field functions for E and H interior to S so long as they are a solution to
Maxwell’s equation for the homogeneous medium. Suppose J′ and J′m produce E′ and H′ in
a medium with possibly different constitutive parameters µ ′ and ε ′ as shown in Fig. 2-18.
Another equivalent surface source can then be obtained from

Js = n̂ × (H − H′) and Jms = −n̂ × (E − E′) ,

E, H n̂ n̂

E, H
Zero
Desired field
J Jm V region Js = n × H
Source
region μ, ε μ, ε
S S Jms = −nˆ × E

Figure 2-17: Equivalent surface electric and magnetic currents Js and Jms , respectively, that can
produce a null field interior to S as well as the original fields outside of S (in a homogeneous
medium with parameters µ and ε ).
2-9 Equivalence Principle for Electromagnetic Sources 93

E', H'
J'm

J'
μ', ε'
E', H'

Figure 2-18: An arbitrary set of currents and material to establish another set of equivalent
sources on the boundary S.

J'm
E, H
μ', ε'
J' E', H' Js

μ, ε
Jms

Figure 2-19: Another set of equivalent surface currents that produce original fields outside S.
The medium is now inhomogeneous.

while maintaining the interior fields as E′ and H′ and the exterior fields as E and H, as shown
in Fig. 2-19. Note that we must keep the original sources and the constitutive parameters of
the media in the region for which we keep the fields. That is, for the equivalent problem,
sources J′ (r) and J′m (r) and fields E′ (r) and H′ (r) interior to S must be the same as those in
Fig. 2-18.
According to the uniqueness theorem, only the tangential component of the electric field,
or the tangential component of the magnetic field, on S is needed to reproduce the fields
everywhere in the region of interest. Therefore we should be able to provide the equivalent
source condition where only one of the two tangential fields is used. To demonstrate this,
consider a problem where surface S is a perfect electric conductor. On this surface, n̂ × E = 0.
To maintain the original tangential electric field on S, according to the boundary condition,
a surface magnetic current given by Jms = −n̂ × E is needed, where E is the original
electric field on the surface. Figure 2-20 shows the original and the equivalent configurations,
including sources both internal and external to surface S. We should note that the equivalent
electric surface current (Js = n̂ × H) will be automatically generated on the PEC surface S.
Similarly, surface S can be replaced with a magnetic conductor over which n̂ × H = 0.
To maintain the original tangential magnetic field on S, an electric surface current Js = n̂ × H
must be placed on S. In this case, only the tangential magnetic field on S is needed.
94 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

′
Jm ′
Jm

J′ J J′
Jm
PEC

Jms = −nˆ × E

(a) Original configuration (b) Equivalent configuration

Figure 2-20: (a) The original configuration includes sources internal and external to a closed
surface S, and (b) the equivalent configuration wherein external sources are kept unchanged but
internal sources are replaced with a PEC surface on S together with an impressed equivalent
magnetic surface current on the boundary S. In this case only the tangential electric field is used.

′
Jm ′
Jm

J′ J J′
Jm PEC

Js = nˆ × H
PMC Jms = −nˆ × E

(a) Original configuration (b) Equivalent configuration

Figure 2-21: (a) The original configuration includes sources internal and external to a closed
surface S and (b) the equivalent configuration wherein external sources are kept unchanged but
and internal sources are replaced with a PEC surface on a portion of S with an equivalent magnetic
surface current and a PMC surface on the complementary portion of S.

Another example is to consider a portion of S as being a perfect electric conductor covered

with a magnetic surface current and the complementary portion of S being a perfect magnetic
conductor with an electric surface current, as shown in Fig. 2-21.
To better understand how we use equivalent sources, we will explore electric circuits as
an analogous model. In a circuit composed of voltage and current sources and passive lumped
elements, one can enclose all active components in a block and the passive components in
2-9 Equivalence Principle for Electromagnetic Sources 95

Is = 0
I
Active Passive Zs V ZL
V I
block block

Figure 2-22: (a) Block diagram of a circuit with two blocks, one with active sources and one
without, and (b) its equivalent circuit with equivalent sources at the designated terminals.

V
Zs I ZL V ZL
Short
circuit

Figure 2-23: The equivalent voltage source with a short circuit at the terminal of the source
block. This is equivalent to having a PEC on S.

another block, as shown in Fig. 2-22(a). Equivalent voltage and current sources can be placed
at the terminal between the active and passive blocks, as shown in Fig. 2-22(b), similar to
the equivalent magnetic and electric surface current sources on surface S. Here V and I are
the voltage and current at the terminal, ZL is the equivalent input impedance of the passive
elements, and Zs is the source impedance. It can easily be shown that the equivalent sources
do not produce any voltage across Zs or any current through it. This is equivalent to null fields
within S. It can also be shown that placing a short circuit across Zs (similar to placing PEC
over S) disables the current source contribution to ZL and the equivalent source is what is
shown in Fig. 2-23. In this case only the voltage source (similar to magnetic surface current
on PEC) is sufficient to establish the original current through the passive block. Conversely,
by open-circuiting Zs (similar to placing PMC over S), only the current source is sufficient to
produce the original voltage across the load, as shown in Fig. 2-24.
The voltage source ensures that no current flows through Zs and keeps the voltage across
the load at V . Hence, the entire current from the current source flows through the load.
96 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

V
Zs I ZL I ZL
Open
circuit

Figure 2-24: The equivalent current source with an open circuit across the terminals of the
source block. This is equivalent to having a PMC on S.

Example 2-5: Equivalent Source

The electric and magnetic fields for a plane wave are given by

E(r) = E0 (− cos θ ŷ + sin θ ẑ) cos(ω t − k sin θ y − k cos θ z) (2.62a)

and
E0
H(r) = x̂ cos(ω t − k sin θ y − k cos θ z) . (2.62b)
η0
These fields are generated from certain sources in the z < 0 half-space. Find equivalent sources
on the x–y plane that generate the same fields at z > 0 half-space.
Solution: According to Eqs. (2.61a) and (2.61b), equivalent surface electric and magnetic
currents can be placed on the z = 0 surface (x–y plane):

Js (x, y) = ẑ × H(x, y, 0) (2.63a)

and
Jms (x, y) = −ẑ × E(x, y, 0) . (2.63b)

Using Eq. (2.62a) in Eq. (2.63a) and Eq. (2.62b) in Eq. (2.63b), we have

(1) E0
Js = cos(ω t − k sin θ y) ŷ
η0
and
(1)
Jms = −E0 cos(θ ) cos(ω t − k sin θ y) x̂ .

We can also place a PEC on the x–y plane and then just superimpose Jms (x, y) on it. Since
the ground plane is infinite, by applying image theory we can remove the ground plane and
double the magnetic current. In this case,
(2)
Jms = −2E0 cos(θ ) cos(ω t − k sin θ y) x̂
 SUMMARY 97

can produce fields given by Eqs. (2.61a) and (2.61b) in the upper half-space (z > 0).
If instead of PEC we use PMC on the x–y plane and use image theory, it can be shown
that
(2) 2E0
Js = cos(ω t − k sin θ y) ŷ
η0
can produce fields given by Eqs. (2.61a) and (2.61b) in the upper half-space.

Summary
Concepts

• The concepts of fictitious magnetic charge and • Electric and magnetic polarization currents are
current distributions are attributed to the magne- not independent of each other. That is, one can
tization vector M or the medium permeability µ . be produced from knowledge of the other.
The magnetic current distribution can only be • Maxwell’s equations predict that electromag-
defined for time-varying fields. netic waves can carry energy, in addition to
• Introduction of magnetic charge and current storing energy in material media. The Poynting
distributions into Maxwell’s equations makes theorem, derived from Maxwell’s equations,
them symmetric and results in a set of states that the law of conservation of energy is
relations between the electric and magnetic field applicable to electromagnetic fields.
quantities and their sources known as “duality • The solution to Maxwell’s equations is unique if
relations.” the fields satisfy Maxwell’s equations, the initial
• For boundary value problems involving elec- field values throughout the region of interest
tromagnetic sources (charges and current) and are known, and the tangential component of the
certain objects, such as an infinite planar electric field or the magnetic field over a surface
conducting object, an equivalent problem can that encloses the region of interest is known.
be created by replacing the objects with • It is shown that the electromagnetic fields
appropriate images of the specified sources. outside a region that encloses electric and/or
• Using the concept of electric and magnetic po- magnetic sources can be reproduced by many
larization currents for an object with parameters other equivalent sources. In other words, the
(µ , ε ), it is possible to construct an equivalent knowledge of fields outside the source region
configuration wherein the object is replaced cannot be used to uniquely specify the sources
with electric and magnetic polarization currents that produced them.
in the background medium while maintaining
the original fields.
 98 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

Important Equations
Equivalent magnetic charge density and current: Boundary conditions for a perfect magnetic
distribution for material with permeability µ0 µr : conductor:
ρm = −µ0 (µr − 1)∇ · H n̂ × H = 0
∂H n̂ × E = −Jsm
Jm = µ0 (µr − 1) n̂ · B = ρsm
∂t
Maxwell’s equations with inclusion of magnetic Polarization current:
sources: ∂E
Jp = ε0 (εr − 1)
∂H ∂t
∇ × E = −µ − Jm Stored electric and magnetic energy density:
∂t
∂E we (r) = 1
∇×H = ε +J 2 D(r) · E(r)
∂t
ρ wm (r) = 21 B(r) · H(r)
∇·E =
ε Poynting vector (power density):
ρm
∇·H =
µ S(r,t) = E(ρ ,t) × H(ρ ,t)

Duality relations: Equivalent surface currents over a surface enclosing

the original source with the assumption that fields
E → H, H → −E, J → Jm , Jm → −J are zero inside:
ρ → ρm , ρm → − ρ , µ → ε , ε → µ
Js = n̂ × H
Jms = −n̂ × E

Important Terms Provide definitions or explain the meaning of the following terms:
boundary conditions (generalized) image of a point charge over polarization current
duality relations a metallic sphere potential energy
electric energy density magnetic charge density potential function
equivalent source magnetic current density Poynting vector
flow of electromagnetic power normalized duality relations superposition principle
image of currents and charges over ohmic loss uniqueness theorem
a planar conductive surface perfect electric conductor (PEC)
image of a point charge over perfect magnetic conductor (PMC)
a dielectric half-space medium polarization charge density
PROBLEMS 99

PROBLEMS
Equivalent Charges and Currents

2.1 Consider a point charge Q placed at the center of a metallic sphere filled with a material
having permittivity ε . If we want to establish the same electric field inside the sphere but with
permittivity ε0 , what charge distribution do we have to establish?
2.2 The permittivity and permeability tensors for a particular medium are given by
   
3.1 0.5 0 1 0.1 0
ε r = 0.5 3.1 0 and µ r = 0 1 0  .
0 0 5 0 0 1.5

(a) For internal fields given by

x̂ + 2ẑ
E= √ cos(ω t) and H = ẑ sin(ω t) ,
5
what equivalent polarization and magnetic currents are necessary to treat this dielectric
medium as a free-space equivalent?
(b) Find the total electric and magnetic currents flowing through the depicted surface in
Fig. P2.2 at y = 5 m.

z
5m

1m y
2m

Figure P2.2: A block of an anisotropic material in a free-space background.

2.3 For a general bianisotropic medium, what equivalent polarization and magnetization
currents are needed to establish the same fields in a homogeneous medium with permittivity
ε0 and permeability µ0 ? Is it possible to find a solution where the bianisotropic medium
is replaced with only polarization or magnetization currents in a background medium with
permittivity ε0 and permeability µ0 ?

Duality

2.4 Consider a homogeneous medium with

D = εE + ξ H
100 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

and
B = ξ E + µH ,

which is called a Tellegen medium. Maxwell’s equations are solved for an electric current
density J and a magnetic current density Jm , and the fields are found to be (E1 , D1 , B1 , H1 ).
The problem is then solved again with J replaced with Jm /η0 and Jm replaced with −η0 J,
and the fields are found to be (E2 , D2 , B2 , H2 ). Determine the constitutive parameters
pfor the
dual problem. Note that η0 is the free-space wave impedance and is defined by η0 = µ0 /ε0 .
2.5
(a) Prove that Maxwell’s equations are invariant under the following duality transforma-
tions:

E′ = E cos α + cB sin α ,
1
B′ = B cos α − E sin α ,
c
q′e = qe cos α + cε0 qm sin α ,
1
q′m = qm cos α − qe sin α ,
cε0
where qe and qm are electric and magnetic charges (the same relations are true for the
√
charge densities), c = 1/ µ0 ε0 , and α is an arbitrary angle in “E–B plane.” Hint:
Show that Maxwell’s equations hold for the prime field quantities, and note that in the
presence of fictitious magnetic charges we have

∂B
∇ · B = ρm and ∇ × E = −Jm − .
∂t
(b) Then prove that  
B
F = qe (E + u × B) + qm − ε0 u × E
µ0
is invariant under the above duality transformation.

Image Theory

2.6 Consider a 2-D space, infinite along the z-direction, defined by the perpendicular
intersection of semi-infinite PEC and PMC planes, as shown in Fig. P2.6. Let a point source
J exist in this space. If the origin is defined by the intersection of the two planes, then the
expression for the current is given by
ŷ − x̂
J = J0 √ δ (x − a, y − a, z).
2

(a) Draw the equivalent image currents.

PROBLEMS 101

PMC a I

a
y

x PEC

Figure P2.6: A filament current placed inside a corner reflector. One side of the reflector is made
from a semi-infinite perfect magnetic conductor and the other side is made from a perfect electric
conductor.

(b) Suppose J is replaced with Jm , a magnetic current density. Draw the equivalent image
currents.
2.7 Consider a 2-D space, infinite along the z-direction. Two PEC planes (both perpendicular
to the x–y plane) intersect at an angle α , as shown in Fig. P2.7. Current J exists in that space
and is given by the following expression:

J = ŷJ0 δ (x − a, y − b, z).

C
PE
J
y
α

x PEC

Figure P2.7: Configuration of a filament current placed inside of a metallic wedge with interior
angle α .

(a) Suppose α = 45◦ . Draw the equivalent image currents.

(b) Suppose α = 44◦ . Draw the equivalent image currents.
(c) Comparing your answers to parts (b) and (c), what can you generalize about applying
image theory to wedges with angle α = 360◦ /N or α , 360◦ /N (N = integer)?
102 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

2.8 Consider a 2-D space, infinite along the z-direction, and defined by two pairs of parallel
PEC and PMC planes, as shown in Fig. P2.8. Let a point source Je exist in this space. The
expression for the current is given by
 
ŷ + x̂ a b
Je = J0 √ δ x − , y − , z .
2 2 2

y
PEC
b

PMC Je PMC
x
PEC a

Figure P2.8: A filament current inside a rectangular region defined by two vertical PMC walls
and two horizontal PEC walls.

(a) Draw the equivalent image currents.

(b) Suppose Je is replaced with a magnetic current density Jm . Draw the equivalent image
currents.
2.9 Consider a loop with radius a that has N turns and carries a current I. If the loop is
located at the origin and is oriented so that its axis lies along the z-direction, the magnetic
field due to the loop can be expressed as

NIa2
H= (r̂ 2 cos θ + θ̂
θ sin θ ).
4R3
Use the modified Lorentz law, I
F = NI dℓℓ × B ,
C
and image theory to calculate the force on the loop if it is held above a perfect electric
conductor at height z0 . Assume that the current is time-variant enough to be levitated, but
static enough to treat it as such.
2.10 Consider an arbitrary planar current sheet in the x–y plane of a Cartesian coordinate
system (Js (x, y) = jsx (x, y) x̂ + jsy (x, y) ŷ). Use image theory to prove that the magnetic field
generated by this current on the plane that contains the current sheet (x–y plane) is normal to
the plane; i.e., H(r) = Hz (r) ẑ.
2.11 Consider an arbitrary planar magnetic current sheet in the x–y plane of a Cartesian
x (x, y) x̂ + jy (x, y) ŷ.) Use image theory to prove that the
coordinate system (Jms (x, y) = jms ms
PROBLEMS 103

electric field generated by this current on the plane that contains the current sheet (x–y plane)
is normal to the plane; i.e., H(r) = Hz (r) ẑ.
2.12 What constitutes the image of a line charge with charge density ρl C/m, outside a
grounded metallic cylinder at y = y0 , as shown in Fig. P2.12? The line charge is parallel to
the cylinder axis (z-axis).

y0 ρl

a
x

Figure P2.12: A geometrical metallic cylinder and a line charge located at y = y0 outside the
cylinder. You are asked to find the image source for this problem.

2.13 A point charge q is placed on the z-axis outside of a metallic hemisphere of radius a that
sits on an infinite planar conductor in the x–y plane. Use image theory to find the equivalent
charges. What if the charge is moved to a different location, say in the y-z plane at θ = θ0 and
r = z0 ?

Flow of Energy

2.14 A thick wire of radius a carries a constant current I uniformly distributed over its cross
section. A narrow gap in the wire of width w forms a parallel-plate capacitor, as shown in
Fig. P2.14.
(a) Find the electric and magnetic fields in the gap region as a function of r (distance from
the axis) and time.
(b) Find the electromagnetic energy (Uem ) and Poynting vector S in the gap. Show that

∂ Uem
−∇ · S = .
∂t
104 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

w
I I
a

Figure P2.14: A cylindrical metallic wire carrying a time-varying current I with a narrow gap
that acts as a capacitor. The current density in the cross section of the wire is uniform.

(c) Determine the total energy in the gap as a function of time. Calculate the total power
flowing in the gap. (Integrate the Poynting vector over the appropriate surface.)
(d) Check that the input power is equal to the rate of increase of energy in the gap. Discuss
your results.
2.15 A coaxial cable consists of an inner conductor, a tubular insulating layer, and an outer
conductor. The insulating layer has a relative permittivity of εr . The electric field intensity
inside the insulating layer is given by
V0
E= cos(ω t − kz) ρ̂
ρ.
ρ ln(b/a)
Consider a piece of coaxial cable of length d where 0 < z < d, as shown in Fig. P2.15.

a b
Length d

Conductivity σ
εr

Figure P2.15: Configuration of a coaxial transmission line filled with a dielectric material
having a relative permittivity εr .

(a) First assume that the conductors are perfect electric conductors (σ = ∞).
(i) Find the magnetic field intensity H.
(ii) Find the surface currents on the inner and outer conductors Jsa and Jsb ,
respectively.
(iii) Find the time-dependent Poynting vector S(t), and calculate the power flow
passing through the cross sections at z = 0 and z = d.
(iv) Find the stored electric and magnetic energy inside the piece of coaxial cable.
PROBLEMS 105

(v) Show that conservation of energy is maintained.

(b) If the conductors have a finite conductivity σ at high frequency, the surface resistivity
is given by r
ω µ0 E
ρS = and JS = .
2σ ρS
We denote the attenuation factor as α = Pl /(2Pavg ), where Pl is the dissipation power
per unit length and Pavg is the time-average power flowing through the cross section.
We can assume in this part of the problem that the length d is an integer multiple of
the wavelength λ ; i.e., d = N λ = 2π N/k. Note: k ≫ α . Use proper approximations to
simplify your derivations.
(i) Find the power dissipating on the inner and the outer conductors.
(ii) Find the attenuation factor α .
Then the electric field can be modified to
V0 1
E= · cos(ω t − kz) e−α z ρ̂
ρ.
ln(b/a) ρ

(iii) Find the power flow passing through the cross sections at z = 0 and z = d.
(iv) Show that conservation of energy is maintained.
(v) Discuss how to choose the ratio a/b to minimize the attenuation factor α .

Uniqueness Theorem

2.16 Let the tangential E-field and tangential H-field on a closed surface S be related by an
impedance matrix Z(r) such that

n̂ × E = Z(r) · (n̂ × H) ,

where n̂ is the outward normal to the surface S bounding the region of interest V and
 
ŝ1 z11 ŝ1 ŝ1 z12 ŝ2
Z(r) = .
ŝ2 z21 ŝ1 ŝ2 z22 ŝ2

ŝ1 and ŝ2 are unit vectors tangential to the surface S such that ŝ1 , ŝ2 , and n̂ form right-
handed orthogonal coordinates and zi j are real constants. Denote E = E1 ŝ1 + E2 ŝ2 + En ŝn
and H = H1 ŝ1 + H2 ŝ2 + Hn ŝn .
(a) Show that

E1 = −z21 H2 + z22 H1
and
E2 = z11 H2 − z12 H1 .
106 Chapter 2 Electromagnetic Concepts, Tools, and Theorems

(b) Determine the condition for elements zi j such that the uniqueness theorem holds.
2.17 A Tellegen medium is a homogeneous medium with constitutive relations

D = εE + ξ H
and
B = ξ E + µH .

Consider a special case of a Tellegen medium with ε = Mq, ξ = q, and µ = q/M.

(a) Show that in the limit as q → ∞ the constitutive relations simplify to D = MB and
H = −ME. (Hint: Field quantities must remain finite.) Assume q → ∞ for the rest of
this problem.
(b) Show that the Poynting vector vanishes at all points within such a medium. Using the
Poynting theorem, show that for time-varying cases, the stored field energy densities
are zero in the absence of sources within the medium.
(c) Consider an interface between such a medium and free space. Show that the following
boundary conditions must hold at the interface:

n̂ × (H + M E) = 0
and
n̂ · (D − M B) = 0 .

Medium 1 Medium 2
(Free space) ε = Mq
ε0 , µ0 ξ =q
µ = q/M
(d) Consider an electric charge Q1 at a distance d from a half-space region (Medium 2).
Show that the above boundary conditions can be satisfied if the medium is replaced
with two overlaying magnetic and electric charges Q2 and Q2m . Find the values of Q2
and Q2m . (Hint: Use Coulomb’s law and its dual. Also assume that the interface is at
the x–y plane, Q1 is placed at r1 = d ẑ, and Q2 and Q2m are placed at r2 = −d ẑ.)
(e) Using the uniqueness theorem and the results of part (d), show that image theory can
be used to find the solution for an electric current source J in front of Medium 2. Find
the equivalent image currents.
(f) Using duality and the results of part (e), find the equivalent image currents for a
magnetic current source Jm . Note that the media should not change under the duality
transformations.
(g) Comment on your results for the special cases of M = 0, M = ∞, and M η0 = 1.
 Chapter 3
The Electromagnetic Potentials
and Radiation

Chapter Contents Objectives

Overview, 108 Upon learning the material presented in this chapter, you
3-1 Electromagnetic Potentials, 108 should be able to:
3-2 Time-Harmonic Electromagnetic Waves, 122
3-3 Time-Harmonic Retarded Potential, 131 1. Calculate the radiated electromagnetic fields in a
3-4 Far-Field Distance Criterion, 137 bounded homogeneous medium, given any
3-5 Small Loop of Current: A Hertzian Magnetic arbitrary distribution of electric and magnetic
Dipole, 139 currents.
3-6 Wire Antennas, 144
3-7 Equivalent Circuit for Receiving 2. Work with the reduced form of Maxwell’s
Antennas, 147 equations for time-harmonic sources and fields.
Chapter Summary, 150
Problems, 152 3. Apply the complex Poynting theorem to evaluate
the power radiated by antennas of known current
distributions.

4. Understand the notion of equivalent current for

transmitting and receiving antennas and the
far-field criteria for small and large antennas.

5. Explain the equivalence between a small current

loop and a magnetic Hertzian dipole
perpendicular to the loop surface.

107
108 Chapter 3 The Electromagnetic Potentials and Radiation

Overview
The intent of this chapter is to develop a procedure for solving Maxwell’s equations to
obtain explicit solutions for the electric and magnetic fields in a homogeneous, isotropic, and
unbounded medium. In general, the solution to the four-dimensional (space-time) coupled
Maxwell’s equations, even for a homogeneous and unbounded medium, cannot be derived
directly for arbitrary source functions (electric and magnetic currents). To facilitate the
solution process, electromagnetic potentials, including electric and magnetic scalar potentials
as well as magnetic and electric vector potentials, are defined from which the electric and
magnetic fields can be evaluated. It is also shown that such potentials satisfy the wave
equation common to all of the aforementioned potential functions. To develop an intuitive
approach for the solution of wave equations, the solution for the scalar electric potential for
a point charge varying with time dependence f (t) is derived and shown to take the form of
√
f (t − r/up )/r, which represents a wave traveling in space with wave velocity up = 1/ µε .
This result is extrapolated to arrive at general solutions for all potentials associated with
arbitrary charge and current distributions, which provides a recipe for field computation for
any antenna, so long as the current distributions are known. The notion of the far field for
an antenna is introduced, and it is shown that simple expressions can be obtained for the
field quantities in terms of current distributions. To further simplify Maxwell’s equations, the
notion of time-harmonic electromagnetic waves is introduced to reduce the dimensionality
of the equations from four (time plus three spatial variables) down to only the three spatial
variables. Time-harmonic expressions for the time-average Poynting vector and the stored
electric and magnetic energy densities are also obtained and used in the derivation of the
complex Poynting theorem, which establishes the law of conservation of energy for time-
harmonic fields. Using the time-harmonic expressions for the potentials, we are able to
examine the radiation characteristics of basic antennas, such as the Hertzian dipole and wire
antennas. Also, an equivalent circuit model for an antenna is introduced and shown to be
equally applicable in both the transmitting and receiving modes.

3-1 Electromagnetic Potentials

The coupled differential equations provided by Maxwell’s equations must be solved for a
given set of impressed currents and charges, subject to a set of specific boundary conditions.
To arrive at a solution, the set of Maxwell’s equations must be simplified as much as possible
without losing any embedded information or relation between the EM fields and the associated
currents and charges. A viable approach to that end is to make use of electromagnetic vector
and scalar potentials. To start, we consider a homogeneous unbounded medium that includes
volumetric electric current and charge densities, J and ρ , respectively. At every ordinary point
in the medium of interest, the following relationships are applicable:

∂B
∇×E = − , (3.1a)
∂t
∂D
∇×H = +J , (3.1b)
∂t
∇·B = 0 , (3.1c)
3-1 Electromagnetic Potentials 109

∇·D = ρ . (3.1d)

Starting from Eq. (3.1c) and noting that ∇ · ∇ × A = 0 for any arbitrary vector A, it can be
concluded that
B = ∇×A (3.2)
for some hitherto unknown vector potential that will be referred to as the magnetic vector
potential. It should be noted that Eq. (3.2) is not unique; that is, vector A′ = A + ∇ψ also
satisfies Eq. (3.2) since ∇×∇ψ = 0 for all scalar functions ψ . To uniquely define the magnetic
vector potential, we need to not only satisfy Eq. (3.2), but to also specify ∇ · A.
Substituting Eq. (3.2) into Eq. (3.1a) gives
 
∂A
∇× E+ =0, (3.3)
∂t
which implies
∂A
E+ = −∇Φ
∂t
for some scalar function Φ, known as the electric scalar potential. If A and Φ can be
determined, then the magnetic flux density B can be obtained from Eq. (3.2) and the electric
field can be obtained from
∂A
E = −∇Φ − . (3.4)
∂t
For a simple homogeneous and isotropic medium we have

D = εE and B = µH ,

and therefore the electric flux density and magnetic field can be obtained from
 
∂A
D = −ε ∇Φ + (3.5a)
∂t
and
1
H = ∇×A . (3.5b)
µ
Substituting Eqs. (3.5a) and (3.5b) into Eq. (3.1b) leads to

1 ∂Φ ∂ 2A
∇ × ∇ × A = −ε ∇ −ε 2 +J ,
µ ∂t ∂t
or equivalently
∂ 2A ∂Φ
∇ × ∇ × A + µε 2
+ µε ∇ = µJ . (3.6)
∂t ∂t
110 Chapter 3 The Electromagnetic Potentials and Radiation

Also, substituting Eq. (3.5a) into Eq. (3.1d) leads to

∂A ρ
∇2 Φ + ∇ · =− . (3.7)
∂t ε
Noting that for any vector A, ∇ × ∇ × A = ∇∇ · A − ∇2A, Eq. (3.6) can be rewritten as
 
2 ∂ 2A ∂Φ
−∇ A + µε 2 + ∇ ∇ · A + µε = µJ . (3.8)
∂t ∂t
As noted earlier, we still need to specify ∇ · A. For simplicity we define

∂Φ
∇ · A = −µε . (3.9)
∂t
This relationship is known as the Lorenz gauge condition. Hence

∂ 2A
∇2 A − µε = −µ J . (3.10)
∂ t2

Also, substituting Eq. (3.9) into Eq. (3.7) leads to

∂ 2Φ ρ
∇2 Φ − µε 2
=− . (3.11)
∂t ε

Equations (3.10) and (3.11) represent wave equations of similar form, except that A is a vector
and Φ is a scalar. In a Cartesian coordinate system,

∇2 A = ∇2 Ax x̂ + ∇2 Ay ŷ + ∇2 Az ẑ .

Hence, Eq. (3.10) represents three scalar wave equations similar to Eq. (3.11), given by

∂ 2 Ax
∇2 Ax − µε = −µ Jx , (3.12a)
∂ t2
∂ 2 Ay
∇2 Ay − µε = −µ Jy , (3.12b)
∂ t2
∂ 2 Az
∇2 Az − µε = −µ Jz . (3.12c)
∂ t2
If the solution to Eq. (3.11) is known, as will be demonstrated in future sections, the solution
to Eq. (3.12) can be obtained by replacing ρ with µε Ji and Φ with Ai for i = x, y, or z. The
relation between A and Φ is also determined by the gauge condition of Eq. (3.9).
3-1 Electromagnetic Potentials 111

3-1.1 Hertz Vector Potential

The idea behind developing a vector potential different from the vector potential A introducted
earlier is to see if a single potential can be defined so that all field quantities can be derived
from it directly. Hertz defined such a vector potential Π such that

∂Π
A = µε . (3.13)
∂t
Hence, according to Eq. (3.9), the scalar electric potential in terms of the Hertz vector
potential is given by
Φ = −∇ ·Π Π. (3.14)
From Eq. (3.2),
∂Π
B = µε ∇ × , (3.15a)
∂t
or equivalently
∂Π
H = ε ∇× , (3.15b)
∂t
and using Eq. (3.4),
∂ 2Π
Π − µε
E = ∇∇ ·Π . (3.16)
∂ t2
It is now obvious that both E and H can be directly computed once Π has been characterized.
To find an equation for Π, we note that from Eqs. (3.15a) and (3.16), the modified Ampère’s
law can be written as
 
∂ ∂ ∂ 2Π
∇ × H = ε ∇ × ∇ ×Π = ε ∇∇ ·Π Π − µε 2 + J , (3.17a)
∂t ∂t ∂t
which can be compacted into
 
∂ ∂ 2Π J
Π + µε 2 = ,
∇ × ∇ × Π − ∇∇ ·Π (3.17b)
∂t ∂t ε

Π = −∇2Π ,
or, using the vector identity ∇ × ∇ × Π − ∇ · ∇Π

Z
∂ 2Π 1
∇2Π − µε 2
=− J dt , (3.18)
∂t ε

which indicates that the Hertz vector potential also satisfies the vector wave equation.
112 Chapter 3 The Electromagnetic Potentials and Radiation

3-1.2 Potentials for Magnetic Sources

Let us consider a scenario where only magnetic current and charge distributions are present.
For such a case, Maxwell’s equations can be written as

∂B
∇×E = − − Jm , (3.19a)
∂t
∂D
∇×H = , (3.19b)
∂t
∇·D = 0 , (3.19c)
∇ · B = ρm . (3.19d)

Obviously Eq. (3.19a) is the dual of Eq. (3.1b), Eq. (3.19b) is the dual of Eq. (3.1a),
Eq. (3.19c) is the dual of Eq. (3.1c), and finally Eq. (3.19d) is the dual of Eq. (3.1d). The
electric vector potential Am is now defined as the dual of the magnetic vector potential A,
which is defined as the dual of Eq. (3.2) and is given by

D = −∇ × Am . (3.20)

Also, the magnetic scalar potential is defined as Φm , the dual of the electric scalar potential Φ.
Applying the duality relation to Eq. (3.4) gives

∂ Am
H = −∇Φm − . (3.21)
∂t
Following a similar procedure, it can be easily shown that the electric vector potential and the
magnetic scalar potential satisfy the following wave equations:

∂ 2 Am
∇2 Am − µε = −ε Jm (3.22a)
∂ t2

and
∂ 2 Φm ρm
∇2 Φm − µε 2
=− , (3.22b)
∂t µ

which are the duals of Eqs. (3.10) and (3.11), respectively.

Using the duality relations, we define the magnetic Hertz vector potential as Πm , which
satisfies the following wave equation:

Z
∂ 2Π m 1
∇2Π m − µε =− Jm dt . (3.23)
∂ t2 µ
3-1 Electromagnetic Potentials 113

The electric and magnetic fields can be obtained in terms of Π m by applying the duality
relations to Eq. (3.15a) and Eq. (3.16), respectively:

∂ Πm
E = −µ ∇ × (3.24a)
∂t
and
∂ 2Π m
Πm − µε
H = ∇∇ ·Π . (3.24b)
∂ t2

3-1.3 Electric and Magnetic Fields in Terms of Potentials: The Most General
Case
Consider a homogeneous medium with constitutive parameters µ , ε and time-varying electric
and magnetic current distributions J and Jm . As was shown in previous sections, the electric
and magnetic fields can be obtained from the electromagnetic potentials when either electric
or magnetic current distributions exist. Using the linearity property of Maxwell’s equations,
the solution to the simultaneous existence of both electric and magnetic sources can be
obtained using the superposition principle. Suppose A and Φ are the solutions for the
magnetic vector and electric scalar potentials generated by J and ρ , and Am and Φm are
those generated by Jm and ρm , then the total magnetic and electric fields are given by

1 ∂ Am
H= ∇×A− − ∇Φm (3.25a)
µ ∂t
and
1 ∂A
E = − ∇ × Am − − ∇Φ . (3.25b)
ε ∂t
Similarly, there are also electric and magnetic Hertz vector potentials associated with J
and Jm . Using the superposition principle, the total electric and magnetic fields in terms of Π
and Π m are given by
 
∂ 2Π ∂ Πm
E = ∇∇ ·Π Π − µε 2 − µ ∇ × (3.26a)
∂t ∂t
and
 
∂ 2 Πm ∂Π
H = ∇∇ ·Π Πm − µε + ε ∇ × . (3.26b)
∂ t2 ∂t

3-1.4 Solution of Wave Equation (Retarded Potential)

To demonstrate the application of potentials, we consider the simple example of a point charge
at the origin. According to Eq. (3.11) the potentials generated by the charge must satisfy

∂ 2Φ ρ
∇2 Φ − µε =− , (3.27)
∂ t2 ε
114 Chapter 3 The Electromagnetic Potentials and Radiation

where in this case

ρ (r) = Q δ (r) , (3.28)
where Q is the magnitude of the charge and δ (r) is a delta function specifying that the charge
is located at the origin of a coordinate system. First it is noted that the problem has spherical
symmetry; that is, Φ(r, θ , φ ) = Φ(r). Therefore ∂∂Φ ∂Φ
θ = ∂ φ = 0 and
 
2 1 ∂ 2 ∂Φ
∇ Φ= 2 r . (3.29)
r ∂r ∂r

By introducing the regular function ψ (r) = rΦ(r), it can easily be shown that

1 ∂ 2ψ
∇2 Φ = . (3.30)
r ∂ r2
Equation (3.27) can then be written as

∂ 2ψ ∂ 2ψ Q
2
− µε 2
= − r δ (r) . (3.31)
∂r ∂t ε
Excluding the location r = 0, ψ must satisfy

∂ 2ψ ∂ 2ψ
− µε =0. (3.32)
∂ r2 ∂ t2
√
Any function of the form f (t − r µε ) satisfies Eq. (3.32) because it can be shown that

∂2 f
= µε f ′′ (3.33a)
∂ r2
and
∂2 f
= f ′′ . (3.33b)
∂ t2
The magnitude of the function at a distance r + ∆ r at a later time t + ∆ t is
√
ψ (r + ∆ r, t + ∆ t) = f (t + ∆ t − (r + ∆ r) µε ) .
∆r √1
So if ∆t = µε , then
ψ (r + ∆ r, t + ∆ t) = ψ (r,t) . (3.34)
This indicates that ψ (r,t) is a propagating signal that has a propagation velocity up = √1µε .
That is, the scalar potential
1 √
Φ(r,t) = f (t − r µε ) (3.35)
r
is a wave. To determine f (t, r), let us examine the quasi-static behavior of the electric
potential. Coulomb’s equation for the potential due to a point charge Q(t) can be used and
3-1 Electromagnetic Potentials 115

is given by
Q(t)
Φ(r,t) = . (3.36)
4πε r
By comparing Eqs. (3.35) and (3.36), it is obvious that

Q(t)
f (t) = . (3.37)
4πε
Hence in general, for a time-varying function
 
r
Q t−
up
Φ(r,t) = , (3.38)
4πε r
if the point charge is located at r′ , then
 
|r − r′ |
Q t−
up
Φ(r,t) = . (3.39)
4πε |r − r′ |

For a distribution of charge density ρv (r′ ,t), the potential can be obtained from

ρv (r′ , t − |r − r′ |/up ) ′
$
1
Φ(r,t) = dr . (3.40)
4πε v |r − r′ |

Equation (3.40) was obtained somewhat heuristically, but it can also be obtained from
Eq. (3.27) rigorously. Comparing Eq. (3.27) with Eq. (3.12), and noting the similarity of
the differential equations, the solution for the vector magnetic potential is deduced to be

J(r′ , t − |r − r′ |/up ) ′
$
µ
A(r,t) = dr . (3.41)
4π v |r − r′ |
The corresponding electric and magnetic fields can be obtained from Eqs. (3.4)To demonstrate
the application of and (3.5b), respectively.

3-1.5 Radiated Field from an Infinitesimal Current Element

Let us consider a very small electric current filament located at the origin of a Cartesian
coordinate system with an orientation along dℓℓ, as shown in Fig. 3-1. The volumetric current
distribution for this current can be mathematically represented by

J(r,t) = I0 dℓℓ δ (r) f (t) . (3.42)

For this current, the retarded vector potential is found to be

 
µ I0 dℓℓ r
A(r,t) = f t− . (3.43)
4π r up
116 Chapter 3 The Electromagnetic Potentials and Radiation

r d
y
x

Figure 3-1: A very small electric current filament at the origin of the Cartesian coordinate
system carries current I0 f (t). Also shown in the expanded picture is the current filament and
the instantaneous charge accumulations at both ends of the wire.

According to Eq. (3.5b)   

I0 dℓℓ r
H= ∇× f t− , (3.44)
4π r up
and since dℓℓ is a constant vector
  
I0 1 r
H= ∇ f t− × dℓℓ . (3.45)
4π r up

Also noting that ∇ f [g(r)] = f ′ ∇g (for a single variable scalar function f ), ∇r = r̂, and
∇(1/r) = (−r̂/r2 ), it can be easily shown that
     
I0 ′ r 1 r 1
H= f t− + f t− dℓℓ × r̂ . (3.46)
4π up up r up r2

The second term corresponds to the quasi-static solution (wherein f ′ ≈ 0), which is the
familiar Biot-Savart law that provides the magnetic field intensity generated by a small
segment of wire carrying an electric current I0 .
As we will show later, the first term inside the square bracket, which varies as 1r , is the
propagating component of the field that carries with it electromagnetic energy. It is interesting
to note that the radiated magnetic field is proportional to the time derivative of the excitation
current function. This indicates that a current that does not vary with time cannot radiate an
electromagnetic field.
The electric field can be obtained from
∂A
E = −∇Φ − , (3.47)
∂t
3-1 Electromagnetic Potentials 117

which requires knowledge of the electric scalar potential. The electric scalar potential can be
obtained from the gauge condition given by Eq. (3.9). Using Eq. (3.43), we have
 
µ I0 f (t − r/up ) dℓℓ
∇·A = ∇·
4π r
 
µ I0 ′ 1 1
=− f (t − r/up ) + f (t − r/up ) 2 r̂ · dℓℓ . (3.48)
4π up r r

From the gauge condition we have

∂Φ ∇·A
=− , (3.49)
∂t µε
which leads to    Z 
I0 r 1 1
Φ(r,t) = f t− + f dt 2 r̂ · dℓℓ . (3.50)
4πε up up r r
To compute E(r,t), ∇Φ is needed, which after some algebraic manipulation is shown to be
given by
" R !  R  #
I0 f′ 2f 2 f dt f f dt
∇Φ = −r̂ + + (r̂ · dℓℓ) + + 2 ∇(r̂ · dℓℓ) , (3.51)
4πε0 u2p r up r2 r3 up r r

and therefore
" R !  R  #
I0 f′ 2f 2 f dt f f dt µ I0 f ′
E= r̂ + + (r̂ · ℓ
dℓ ) − + ∇(r̂ · ℓ
dℓ ) − dℓℓ .
4πε u2p r up r2 r3 up r r2 4π r
(3.52)
It can also be shown that ∇(r̂ · dℓℓ) is proportional to 1r . Hence the terms proportional to f ′
are the only terms that survive in far distances from the source. In fact, after some algebraic
manipulation, it can be shown that
1
∇(r̂ ·ℓℓ) = r̂ × (dℓℓ × r̂) . (3.53)
r
Denoting dℓℓ = r̂ × (dℓℓ × r̂) + (dℓℓ · r̂)r̂ and recalling that 1/u2p = µε , it can be easily shown
that
 R  p R !
I0 f f dt I0 µ /ε f f dt µ f ′
E= + 3 (r̂ ·dℓℓ)r̂ − + + r̂ × (dℓℓ × r̂) . (3.54)
2πε up r2 r 4π r2 ε r3 r
R
Noting that I0 f (t) dt represents the instantaneous accumulated charge Q(t) at the end of
the filament, the terms with coefficient 1/r3 represent the field contribution from the electric
dipole. The other terms, 1/r2 and 1/r, are the electrodynamic terms. Specifically, terms
proportional to 1/r2 result from the Bio-Savart component of the magnetic field. As noted
earlier, at far distances from the source, such that only the 1r term remains, the electric field
118 Chapter 3 The Electromagnetic Potentials and Radiation

does not have a radial component and is proportional to the time derivative of the current.
Note that in such a “far-field” region, the electric field is proportional to the time derivative of
the current source as well. In the far-field region where 1/r ≫ 1/r2 ≫ 1/r3 ,

√
I0 µε f ′ (t − r/up )
H(r,t) = dℓℓ × r̂ (3.55a)
4π r
and (far-field region)
µ I0 f ′ (t − r/up )
E(r,t) = (dℓℓ × r̂) × r̂ . (3.55b)
4π r

It is obvious that no radiation takes place along the axis of the current filament (dℓℓ); that
is, both E and H are zero in the far-field region along the dℓℓ direction, and E and H are
perpendicular to each other and to the direction of propagation r̂. It is also interesting to note
that the ratio of the magnitude of E to that of H is independent of position and time and the
constant of proportionality is given simply by
r
|E| µ
= =η . (3.56)
|H| ε
This quantity has the unit of ohms (V/A) and is known as the characteristic impedance of the
medium.

Example 3-1: Electric Field of an Electric Dipole

Derive an expression for the electric field generated by a static dipole with dipole moment
P = Q dℓℓ centered at the origin, as shown in Fig. 3-2.
Solution: The electric scalar potential associated with the positive and negative charges is
given by  
Q 1 1
Φ(r) = − . (3.57)
4πε r1 r2
Noting that for very small dℓ, we can approximate r1 and r2 as
dℓℓ · r̂
r1 ≈ r − (3.58a)
2
and
dℓℓ · r̂
r2 ≈ r + , (3.58b)
2
and then conclude that
 
1 1 dℓℓ · r̂
≈ 1+ (3.59a)
r1 r 2r
3-1 Electromagnetic Potentials 119

Observation
point
r1
+Q r
r2
−Q y

Figure 3-2: An electric dipole with dipole moment P = Q dℓℓ centered at the origin of a
coordinate surface.

and
 
1 1 dℓℓ · r̂
≈ 1− . (3.59b)
r2 r 2r
Hence the electric scalar potential is given by
Q 1
Φ(r) = (dℓℓ · r̂) . (3.60)
4πε r2
The electric field can be obtained from
 
Q −2 1
E(r) = −∇Φ(r) = − ℓ
(dℓ · r̂) + ℓ
∇(dℓ · r̂) . (3.61)
4πε r3 r2
Using Eq. (3.53),  
Q 2 1
E(r) = ℓ
(dℓ · r̂) − r̂ × ℓ
(dℓ × r̂) . (3.62)
4πε r3 r3
R
Recognizing that Q = I0 f (t)dt, the field of the dipole is recognized as the two terms in
Eq. (3.54) that vary as 1/r3 .

3-1.6 E and H Fields Due to a Distribution of Electric Currents

The expressions for the magnetic and electric fields given by Eqs. (3.46) and (3.54) can be
interpreted as the spatial impulse response of the medium (solution to an impulse function of
position δ (r)). Consider a time-varying current filament at r′ , as shown in Fig. 3-3. According
120 Chapter 3 The Electromagnetic Potentials and Radiation

Observation
Source
point
point
r - r′
r′ r
r′ dll

Figure 3-3: An infinitesimal time-varying current at r′ .

to Eq. (3.46) the magnetic field produced by this current is given by

 
I0 f ′ (t − |r − r′ |/up ) f (t − |r − r′ |/up ) (r − r′ )
dH = + ℓ
dℓ × , (3.63)
4π up |r − r′ | |r − r′ |2 |r − r′ |

which is obtained from Eq. (3.46) upon substituting |r − r′ | for r and (r − r′ )/(|r − r′ |) for r̂.
Now let us consider an arbitrary distribution of line currents along a curve defined by C, as
shown in Fig. 3-4. The current distribution is denoted by I(r′ ,t). Using superposition, the

z
Observation point

r C

r′
y

Figure 3-4: A wire carrying a time-varying current specified by curve C in a general coordinate
system.
3-1 Electromagnetic Potentials 121

magnetic field at the observation point is given by

Z  ′ ′ 
1 I (r ,t − |r − r′ |/up ) I(r′ ,t − |r − r′ |/up ) ′ (r − r′ )
H= + ℓ
dℓ × . (3.64)
4π C up |r − r′ | |r − r′ |2 |r − r′ |

where I ′ (·) = ∂∂t I(·). It is obvious that for this current distribution, the electric field can be
obtained from Eq. (3.55b) in a similar manner. Duality relations can be used to find the electric
and magnetic fields generated by a magnetic current distribution.
Instead of a line current, if the source is a surface or a volumetric current distribution, then
we need to make the following substitutions for I(r′ ,t) dℓℓ′ :

I(r′ ,t) dℓℓ′ Js (r′ ,t) ds′ for surface current, and
′ ′ ′ ′
I(r ,t) dℓℓ J(r ,t) dv for volumetric current.

For example, the magnetic field for a volumetric current distribution is given by
$  ′ ′ 
1 J (r ,t − |r − r′ |/up ) J(r′ ,t − |r − r′ |/up ) (r − r′ ) ′
H= + × dv , (3.65)
4π V up |r − r′ | |r − r′ |2 |r − r′ |

where J ′ (·) = ∂∂t J(·). It is obvious that for a magnetic current distribution the electric field
can be obtained from Eq. (3.65) using the duality relation.
The far-field expressions, which are valid at observation points far from the source, are
much easier to compute. The far-field criterion is a condition that depends on the dimensions
of the current source and the oscillation frequency. This will be discussed in more detail in
Section 3-4. For the time being, let us assume that the observation point is in the far-field
region of an electric current distribution J(r,t) and a magnetic current distribution Jm (r,t).
Using the expression given by Eq. (3.55a), Eq. (3.55b) in conjunction with the duality and
superposition principles, we can find the total radiated electric and magnetic fields in a
homogeneous nondispersive medium with constitutive parameters ε and µ using
√ $ ′ ′
µε J (r ,t − |r − r′ |/up )
H(r,t) = ′ |2
× (r − r′ ) dv′
4π V1 |r − r
$  ′ ′ 
ε Jm (r ,t − |r − r′ |/up )
+ ′ |3
× (r − r ) × (r − r′ ) dv′
′
(3.66a)
4π V2 |r − r
and
$  ′ ′ 
µ J (r ,t − |r − r′ |/up )
E(r,t) = ′ |3
× (r − r ) × (r − r′ ) dv′
′
4π V1 |r − r
√ $ ′ ′
µε Jm (r ,t − |r − r′ |/up )
− × (r − r′ ) dv′ . (3.66b)
4π V2 |r − r′ |2

The expressions given by Eqs. (3.66a) and (3.66b) are the basic time-domain equations
governing antenna radiation. Since |r| ≫ |r′ |, for simplicity (r − r′ ) is often approximated
by r in the denominator and in the vector cross product terms of the integrands of Eqs. (3.66a)
122 Chapter 3 The Electromagnetic Potentials and Radiation

and (3.66b). However, this approximation cannot be used in the delay term |r − r′ |/up of the
argument of function f ′ for fast time-varying functions.

3-2 Time-Harmonic Electromagnetic Waves

In the preceding sections, it was shown that electromagnetic field quantities and their related
potentials can be expressed in terms of coupled partial differential equations in a general four-
dimensional space-time vector space. It is often convenient to reduce the dimensionality by
resorting to Fourier transformations. That is, the temporal behavior of field quantities can be
removed by expressing the fields in the frequency domain and solving the resultant equations
in the frequency domain first. In the time-domain treatment of Maxwell’s equations, we have
so far assumed the medium to be nondispersive. That is, the permittivity and permeability
of the medium were assumed to be constant. In Section 1-4.3, it was shown that except for
vacuum, all materials are dispersive and thus ε and µ are functions of time. To circumvent
this difficulty, we can resort to frequency-domain solutions of Maxwell’s equations, which
are discussed in this section. Once the solution in the frequency domain is obtained, using the
linearity properties of Maxwell’s equations and inverse Fourier transformations, the temporal
behavior of the field quantities can then be reconstructed.
Strictly speaking, Fourier transformations can be applied to “energy” functions, i.e.,
functions that are square integrable. Basically, for a field quantity E(r,t) defined at all points
in space we require Z +∞
|E(r,t)|2 dt < ∞ .
−∞
For simplicity, let us assume that all field quantities and source functions are square integrable
functions of time and their Fourier transforms are well defined. The Fourier transform of
E(r,t) is defined as Z +∞
e ω) =
E(r, E(r,t) eiω t dt , (3.67)
−∞
e ω ) is in general a complex vector field quantity. Conversely, the temporal field
where E(r,
e ω ) through an inverse Fourier transformation given by
quantity can be obtained from E(r,
Z +∞
1 e ω ) e−iω t d ω .
E(r,t) = E(r, (3.68)
2π −∞
Note that the Fourier transform definitions given by Eqs. (3.67) and (3.68) are different from
the conventional definitions, which are the complex conjugate of what we have here. Our
definition facilitates integration on the upper half of the complex ω -plane and is consistent
with the conventions used in the literature on optics and physics. Equation (3.68) clearly
demonstrates that an arbitrary field quantity can be expressed in terms of the superposition of
e ω ) e−iω t . Here E(r,
time-harmonic field quantities of the form E(r, e ω ) in general is a complex
quantity.
3-2 Time-Harmonic Electromagnetic Waves 123

Confining our attention to field quantities that are real functions of time and space, i.e.,
E∗ (r,t) = E(r,t), it can be easily shown that

e ∗ (r, ω ) = E(r,
E e −ω )

and Z +∞
1 e ω ) e−iω t ] d ω .
E(r,t) = Re[E(r, (3.69)
π 0
Hence any time-domain field quantity can be expressed in terms of the superposition of real
time-harmonic functions of the form
e ω ) e−iω t ] .
e(r,t) = Re[E(r, (3.70)

The complex quantity

E(r, e r (r, ω ) + iE
e ω) = E e i (r, ω ) (3.71)
is referred to as a vector phasor. Similarly, the other time-harmonic fields and sources can be
obtained from their frequency-domain counterparts through
e ω ) e−iω t ] ,
h(r,t) = Re[H(r, (3.72a)
e ω ) e−iω t ] ,
d(r,t) = Re[D(r, (3.72b)
e ω ) e−iω t ] ,
b(r,t) = Re[B(r, (3.72c)
J(r, ω ) e−iω t ] ,
j(r,t) = Re[e (3.72d)
and
ρ (r,t) = Re[ρe(r, ω ) e−iω t ] . (3.72e)

Like any field function, time-harmonic field quantities must satisfy Maxwell’s equations.
Consider Faraday’s law:
∂ b(r,t)
∇ × e(r,t) = − . (3.73)
∂t
Substituting the phasor forms we have

e ω ) e−iω t ] = − ∂ Re[B(r,
∇ × Re[E(r, e ω ) e−iω t ] . (3.74)
∂t
∂
The ∇×, ∂t , and Re[·] are all linear operators and are commutative; hence

Re{[∇ × E(r, e ω )]e−iω t } = 0 .

e ω ) − iω B(r, (3.75)

Since Eq. (3.75) must be valid for all values of t, it follows that the quantity in the square
bracket must vanish. Hence
∇ × E(r, e ω) .
e ω ) = iω B(r, (3.76a)
124 Chapter 3 The Electromagnetic Potentials and Radiation

Similarly, the modified Ampère’s law yields

e ω ) = −iω D
∇ × H(r, e +e e ω) .
J(r, ω ) + σ E(r, (3.76b)

Gauss’ law provides

e ω ) = ρe(r, ω )
∇ · D(r, (3.77a)
and
e ω) = 0 .
∇ · B(r, (3.77b)

Constitutive relations also take on a rather simple form. As mentioned previously, the
permittivity and permeability of matter are in general functions of frequency and the
relationships between the electric and magnetic flux densities and their corresponding
field intensities in the time domain involve mathematical convolutions. Hence, the Fourier
transform of the flux densities can be expressed in terms of the product of Fourier transforms
of the constitutive parameters and the field intensities. That is,

D(r, e ω)
e ω ) = ε (ω ) E(r, (3.78a)
and
B(r, e ω) ,
e ω ) = µ (ω ) H(r, (3.78b)

where ε (ω ) and µ (ω ) are in general complex permittivity and permeability tensors.

For simplicity we drop the parameter ω from the time-independent Maxwell’s equations.
Allowing generalized time harmonic electric and magnetic sources, the time-independent
Maxwell’s equations assume the following form:

∇ × E(r) = iω µ H(r) − Jm (r) , (3.79a)

∇ × H(r) = −iωε E(r) + J(r) , (3.79b)
∇ · B(r) = ρm (r) , (3.79c)
∇ · D(r) = ρ (r) . (3.79d)

Also, the continuity relations are given by

∇ · J(r) = iω ρ (r)
and
∇ · Jm (r) = iω ρm (r) .

For media with finite conductivity where σ , 0, the conductivity can be absorbed into the
complex permittivity. From Eq. (3.76b),
 
iσ
∇ × H = −iω ε E + σ E + J = −iω ε + E+J . (3.81)
ω
3-2 Time-Harmonic Electromagnetic Waves 125

Comparing Eq. (3.81) with Eq. (3.79b), the quantity ε + iσ /ω can be viewed as an effective
permittivity of the medium:
iσ
εeff = ε + . (3.82)
ω
That is, the medium conductivity adds to the imaginary part of the effective permittivity of
the medium.

3-2.1 Vector and Scalar Potentials

As mentioned earlier, the time-harmonic potentials can also be defined in terms of their
respective phasor quantities. It is then easy to show that

B(r, ω ) = ∇ × A(r, ω ) (3.83a)

and
E(r, ω ) = −∇Φ(r, ω ) + iω A(r, ω ) . (3.83b)

The gauge condition given by Eq. (3.9) assumes the following form:

∇ · A(r, ω ) = iω µε Φ(r, ω ) , (3.84)

from which we obtain

iω
Φ(r, ω ) = −
∇·A , (3.85)
k2
√
where k = ω µε is known as the medium’s wave number. Hence,

iω
E= ∇∇ · A + iω A . (3.86)
k2
In a similar manner, the phasors must satisfy the following frequency-domain wave equations:

∇2 A + k 2 A = − µ J (3.87a)
and
ρ
∇2 Φ + k 2 Φ = − . (3.87b)
ε
Duality relations can be applied to the phasor quantities as well. Hence, in the presence of
magnetic sources, the electric vector potential and magnetic scalar potential must satisfy

∇2 Am + k2 Am = −ε Jm (3.88a)
and
ρm
∇2 Φ m + k 2 Φ m = − . (3.88b)
µ
126 Chapter 3 The Electromagnetic Potentials and Radiation

In general, when both electric and magnetic sources are present,

1 iω
H= ∇ × A + 2 ∇∇ · Am + iω Am (3.89a)
µ k
and
iω 1
E= 2
∇∇ · A + iω A − ∇ × Am . (3.89b)
k ε

3-2.2 Hertz Potentials

For time-harmonic fields, the phasor forms of the Hertz vector potentials are denoted by
Π (r, ω ) and Π m (r, ω ). These satisfy the following wave equations:
−i J
∇2Π + k2Π = (3.90a)
ωε
and
−i Jm
∇2Π m + k2Π m = . (3.90b)
ωµ
Also, the field quantities, in terms of the Hertz vector potentials, are given by

Πm + k 2 Π m
H = −iωε ∇ × Π + ∇∇ ·Π (3.91a)
and
Π + k2Π + iω µ ∇ × Π m .
E = ∇∇ ·Π (3.91b)

3-2.3 Poynting Vector for Time-Harmonic Fields

Consider a time-harmonic electromagnetic wave given by

e(r,t) = Re[E(r) e−iω t ] (3.92a)

and
h(r,t) = Re[H(r) e−iω t ] . (3.92b)

The Poynting vector for this set of electric and magnetic fields is given by

s(r,t) = e(r,t) × h(r,t)

= Re[E(r) e−iω t ] × Re[H(r) e−iω t ] . (3.93)

As mentioned earlier, the phasors E(r) and H(r) are, in general, complex quantities that can
be written in the following forms:

E(r) = Er (r) + i Ei(r) (3.94a)

3-2 Time-Harmonic Electromagnetic Waves 127

and
H(r) = Hr (r) + i Hi (r) . (3.94b)

Substituting Eq. (3.94) in Eq. (3.93), and after simple algebraic manipulation, it can be shown
that

s(r,t) = [Er (r) cos ω t + Ei(r) sin ω t] × [Hr (r) cos ω t + Hi (r) sin ω t]
= Er (r) × Hr (r) cos2 ω t + Ei(r) × Hi (r) sin2 ω t
+ [Er (r) × Hi (r) + Ei (r) × Hr (r)] sin ω t cos ω t . (3.95)

The time-average Poynting vector representing the net power density going through at a given
point in space can be obtained from
Z t+T
1
Savg (r) = s(r,t) dt
T t
1
= [Er (r) × Hr (r) + Ei (r) × Hi (r)] , (3.96)
2
2π
where T = ω is a period of the sinusoidal functions. On the other hand, we note that

Re[E(r) × H∗ (r)] = Er (r) × Hr (r) + Ei (r) × Hi (r) . (3.97)

Comparing Eq. (3.96) with Eq. (3.97), it is obvious now that

1
Savg (r) = Re[E(r) × H∗ (r)] , (3.98)
2
which states that the time average value of the Poynting vector can be obtained directly from
the cross product of the phasor vector field quantities.
In general, the complex Poynting vector or complex power flow density is defined as
1
S(r) = E(r) × H∗ (r) , (3.99)
2
from which the time-average power density can be obtained:

Savg (r) = Re[S(r)] . (3.100a)

Also
1
Im[S(r)] = Im[E(r) × H∗ (r)] (3.100b)
2
is related to the flow of reactive power density. In fact this power does not flow out, but
oscillates back and forth through a boundary.
128 Chapter 3 The Electromagnetic Potentials and Radiation

3-2.4 Energy Density for Time-Harmonic Fields

The stored electric energy per unit volume for a time-harmonic field can be directly computed
from
1
we (r,t) = d(r,t) · e(r,t)
2
1
= Re[D(r) e−iω t ] · Re[E(r) e−iω t ] . (3.101)
2
The complex phasors for the electric flux density and the electric field intensity have real and
imaginary vector components denoted by D(r) = Dr (r) + iDi (r) and E(r) = Er (r) + i Ei (r),
respectively. Upon substituting these expressions in Eq. (3.101), it can be shown that
1
we (r,t) = Dr (r) · Er (r) cos2 (ω t) + Di (r) · Ei (r) sin2 (ω t)
2

+ (Dr (r) · Ei (r) + Di (r) · Er (r)) cos(ω t) sin(ω t) . (3.102)

The time-average electric energy density is given by

Z T
1 1
wavg
e (r) = we (r,t) dt = [Dr (r) · Er (r) + Di (r) · Ei (r)] , (3.103)
T 0 4

where T = 2π /ω . Noting that for an isotropic medium

Dr (r) = Re[ε E(r)] = ε ′ Er (r) − ε ′′ Ei (r) (3.104a)

and
Di (r) = Im[ε E(r)] = ε ′ Ei (r) + ε ′′ Er (r) , (3.104b)

it can be shown that the time-average electric energy density can be computed from

ε′
ε′
wavg
e (r) = |Er (r)|2 + |Ei (r)|2 = |E(r)|2 , (3.105)
4 4
where |E(r)|2 = E(r) · E∗ (r).
Following a similar procedure, it can be shown that the time-average magnetic energy
density can be computed from

µ′

wavg
m (r) = |Hr (r)|2 + |Hi(r)|2
4
µ′
= |H(r)|2 . (3.106)
4
3-2 Time-Harmonic Electromagnetic Waves 129

3-2.5 Complex Poynting Theorem

The complex Poynting theorem can be obtained from the time-harmonic version of Maxwell’s
equations given by Eqs. (3.79a) and (3.79b). Starting from

∇ · (E × H∗) = H∗ · ∇ × E − E · ∇ × H∗

and using Eqs. (3.79a) and (3.79b), it can be shown that

∇ · (E × H∗) = iω µ |H|2 − Jm · H∗ − iωε ∗ |E|2 − J∗ · E , (3.107)

where the medium conductivity is absorbed into the imaginary part of the complex
permittivity.
Integrating over a volume V and applying the divergence theorem to the left-hand side, it
can be shown that
$ $  
1 ∗ 1 ∗ ∗ 1 2 1 ∗ 2
(E × H ) · ds = − [E · J + Jm · H ] dv + iω µ |H| − ε |E| dv .
2 S 2 V V 2 2
(3.108)
The complex time-average input power is given by
$
1
c
Pin = − (E · J∗ + Jm · H∗ ) dv . (3.109)
2 V

The time-average stored electric field energy (a positive real quantity) is

$
1 1 ′ 2
we = ε |E| dv . (stored electric energy) (3.110a)
2 V 2

The time-average stored magnetic field energy is

$
1 1 ′ 2
wm = µ |H| dv , (stored magnetic energy) (3.110b)
2 V 2

where ε ′ and µ ′ are the real parts of the complex permittivity and permeability of the medium.
The time-average dissipated power is

$
1
Ploss = (ω µ ′′ |H|2 + ωε ′′ |E|2 ) dv . (dissipated power) (3.111)
2 V

c ), we have
By denoting the left-hand side of Eq. (3.108) as the “complex” radiated power (Prad
c
Pinc = Prad + Ploss + i2ω (we − wm ) . (3.112)
130 Chapter 3 The Electromagnetic Potentials and Radiation

Explicitly,

Re[Pinc ] = Re[Prad
c
] + Ploss (3.113a)
and
c
Im[Pinc ] = Im[Prad ] + 2ω (we − wm) . (3.113b)

Equation (3.113a) is a statement of the law of conservation of energy and Eq. (3.113b) is
the extension of the same law for the reactive power.

3-2.6 Time-Harmonic Polarization Currents

As described in Section 2-4, a medium with permittivity ε and permeability µ can be replaced
with polarization currents in free space. Suppose the fields in the medium are time-harmonic
fields given by e(r,t) and h(r,t); then the electric and magnetic polarization currents are given
by

∂
Jep (r,t) = ε0 [(εr − 1) e(r,t)] (3.114a)
∂t
and
∂
Jmp (r,t) = µ0 [(µr − 1) h(r,t)] . (3.114b)
∂t
The frequency-domain equivalents of Eqs. (3.114a) and (3.114b) are given by

Jep (r) = −iωε0 (εr − 1) E(r) (3.115a)

and
Jmp (r) = −iω µ0 (µr − 1) H(r) . (3.115b)

Using ε = ε ′ − iε ′′ , we have

Jep (r) = ωε0 εr′′ E(r) − iωε0 (εr′ − 1) E(r) . (3.116)

The first component (in-phase) is known as the dissipative current and the second component
is known as the reactive current, which together with the displacement current in the
equivalent free space constitutes the total displacement current −iωε0 εr′ E(r). The quality
of a dielectric material can be determined in terms of the relative magnitude of the reactive
current to that of the dissipative current, which is independent of the magnitude of the electric
field and known as the quality factor

displacement current density ε′

Q̆ = = ′′r . (3.117)
dissipative current density εr
The reciprocal of this quantity is known as the dielectric loss tangent. It should be noted
that the quality factor definition given by Eq. (3.117) is consistent with the definition for the
3-3 Time-Harmonic Retarded Potential 131

quality factor of a capacitor filled with the same material, defined by

maximum of stored energy
Q̆ = ω
average dissipated power
1
2 ε ′|E|2 ε′
=ω = . (3.118a)
1 ′′
2 ωε |E|
2 ε ′′

For magnetic materials (µ , µ0 ), an analogous magnetic quality factor Q̆m can be defined as

µ′
Q̆m = . (3.118b)
µ ′′

3-3 Time-Harmonic Retarded Potential

The magnetic vector potential produced by a time-harmonic filament current with

j(r,t) = δ (r) dℓℓ Re[I0 (ω ) e−iω t ] , (3.119)

noting that f (t) = Re[I0 (ω ) e−iω t ], can be obtained from the Fourier transformation of the
time-domain solution obtained in Section 3-1.5. Noting that the Fourier transform of f (t − t0 )
is given by
F[ f (t − t0 )] = F(ω ) eiω t0 = I0 (ω ) eiω t0 , (3.120)
and recalling that t0 = r/up , the phasor vector of the corresponding magnetic vector potential
is given by
µ I0 eikr
A(r) = dℓℓ , (3.121)
4π r
where, as before, k = ω /up .
Placing the current filament at location r′ , the corresponding magnetic vector potential is
given by
′
µ I0 eik|r−r |
A(r) = dℓℓ . (3.122)
4π |r − r′ |
Also, using Eq. (3.46), the phasor magnetic field is
  ik|r−r′ |
I0 k2 1 e (r − r′ )
H(r) = −i + ℓ
dℓ × , (3.123)
4π k|r − r′ | k|r − r′ | |r − r′ |
and from Eq. (3.54), the phasor electric field is given by
′  
I0 k2 eik|r−r | 2 i2 (r − r′ ) · dℓℓ
E(r) = η + (r − r′ )
4π k|r − r′ | k|r − r′ | k2 |r − r′ |2 |r − r′ |2
   
1 i (r − r′ ) (r − r′ )
− + −i × dℓℓ × . (3.124)
k|r − r′ | k2 |r − r′ |2 |r − r′ | |r − r′ |
132 Chapter 3 The Electromagnetic Potentials and Radiation

|r −r'|
Source

r'
r
0

Figure 3-5: The geometry of a source at location r′ and an observation point at location r far
from the source.

For a volumetric current distribution with a phasor representation J(r), the magnetic field can
be obtained from Eq. (3.59b):
$   ik|r−r′ |  
k2 1 e ′ (r − r′ )
H(r) = −i + J(r ) × dv′ . (3.125)
4π V k|r − r′ | k|r − r′ | |r − r′ |
The far-field expression can be obtained by assuming that

k|r − r′ | ≫ 1 .

Accordingly, the following approximations can be made (see Fig. 3-5):

|r − r′ | ≈ r − r′ · r̂ (3.126a)

for the phase term and

|r − r′ | ≈ r (3.126b)
for the magnitude terms. In the far-field region, the expression given by Eq. (3.125) can be
approximated as

ikeikr
$
′
H(r) = − [J(r′ ) × r̂]e−ikr ·r̂ dv′ . (3.127a)
4π r V

The corresponding electric field is given by

ikη eikr
$
′
E(r) = − [J(r′ ) × r̂] × r̂ e−ikr ·r̂ dv′ . (3.127b)
4π r V
3-3 Time-Harmonic Retarded Potential 133

Careful examination of Eqs. (3.127a) and (3.127b) reveals that

E(r) = η H(r) × r̂ , (3.128)

which indicates that the electric field is perpendicular to the magnetic field and both are
perpendicular to the direction of propagation r̂. This type of field configuration is known as a
transverse electromagnetic wave. It is also interesting to note that the ratio of the electric field
magnitude to that of the magnetic field is a constant, known as the characteristic impedance
of the medium η .

Example 3-2: Radiated Power and Antenna Directivity of a Hertzian Dipole

Consider a Hertzian dipole with I0 ∆ ℓ = I0 ∆ ℓ ẑ at the origin of a Cartesian coordinate system.

Find the complex Poynting vector and total power radiated by the Hertzian dipole antenna.
Defining antenna directivity along a given direction in the far-field region as the ratio of the
power density in that direction to the power density had the radiated power were to be radiated
uniformly, find the directivity function of the Hertzian dipole.
Solution: Using Eqs. (3.123) and (3.124) with r′ = 0 and ẑ × r̂ = sin θ φ̂
φ, the explicit forms
of the magnetic and electric fields are given by
 
I0 ∆ ℓk eikr 1
H(r) = −i + sin θ φ̂
φ (3.129)
4π r kr
and
     
I0 ∆ ℓk eikr 1 i 1 i
E(r) = η 2 + cos θ r̂ + −i + + sin θ θ̂
θ , (3.130)
4π r kr (kr)2 kr (kr)2

where use was made of the relations ẑ · r̂ = cos θ and r̂ × (ẑ × r̂) = − sin θ θ̂
θ. The complex
Poynting vector can now be evaluated from
1
S(r) = E(r) × H∗ (r)
2
       
η I0 ∆ ℓk 2 −i 1 i 2
= 1+ sin 2θ θ
θ̂ + 1 + sin θ r̂ . (3.131)
2 4π r kr (kr)2 (kr)3
To calculate the total radiated power, we consider a spherical surface with radius r and
evaluate the time-average net power flow through the surface. This can be done by evaluating
Z 2πZ π
Pr = Re[S(r)] · r̂ r2 sin θ d θ d φ
0 0
Z π
η |I0 ∆ ℓ|2 k2 η |I0 ∆ ℓ|2 k2
= (2π ) sin3 θ d θ = . (3.132)
2 (4π )2 0 3 4π
134 Chapter 3 The Electromagnetic Potentials and Radiation

It is interesting to note that this calculated radiated power is independent of the radius of the
sphere, as expected. Another important fact to highlight is that the radiated power is produced
entirely by the far-field terms that are proportional to kr1 . The directivity of the dipole, per the
definition given in the problem statement, can be evaluated from

Re[Sr (r)] 3
D(θ , φ ) = 2
= sin2 θ . (short dipole) (3.133)
Pr /(4π r ) 2

The directivity pattern is independent of φ , as expected, and assumes its maximum value at
θ = 90◦ .

Example 3-3: Stored Electromagnetic Energy and Quality Factor for a

Hertzian Dipole

Find the time-average stored electric and magnetic energy supplied by the Hertzian dipole
and determine its quality factor Q̆.
Solution: Suppose the antenna and its sources can be enclosed within a sphere of radius a.
In our example a can be as small as ∆ ℓ/2. The total time-averaged stored energy can be
calculated by integrating Eq. (3.105) and Eq. (3.106) over an infinite volume, excluding the
small sphere of radius a around the antenna. As noted in the previous example, the far-
field components of the electric and magnetic fields are solely responsible for the radiated
power. These fields fill all space, and under steady-state conditions, the source has existed
indefinitely and the energy delivered and stored in the space outside the antenna region is
infinite. However, excluding the far-field terms, only fields contributing to the reactive power
are retained. The time-average stored electric energy per unit volume, excluding the far-field
component, can be obtained from Eq. (3.130) and is given by
       
1 ′ η I0 ∆ ℓk 2 1 1 1
wavg
e (r) = ε 4 1+ 2
cos θ + 1 + 2
sin θ
4 4π k2 r4 (kr)2 (kr)2
   
1 ′ η I0 ∆ ℓ 2 1 1
= ε 1+ (3 cos2 θ + 1) . (3.134)
4 4π r4 (kr)2
The total electric energy stored outside the antenna can be computed from
Z ∞ Z 2π Z π
We = wavg 2
e (r) r sin θ d φ d θ dr
a 0 0
  Z ∞ 
1 ′ η I0 ∆ ℓk 2 1 1
= ε 8π + dr
4 4π a (kr)2 (kr)4
 
µ k(I0 ∆ ℓ)2 1 1
= + . (3.135)
8π (ka) (ka)3
3-3 Time-Harmonic Retarded Potential 135

The time-average stored magnetic energy per unit volume can be evaluated from Eq. (3.129):

1 ′ (I0 ∆ ℓk)2 1 1
wavg
m (r) = µ sin2 θ . (3.136)
4 (4π )2 r2 (kr)2
The total magnetic energy stored outside the antenna is therefore given by
Z ∞ Z 2π Z π
Wm = wavg 2
m (r) r sin θ d φ d θ dr
a 0 0
µ k(I0 ∆ ℓ)2 1
= . (3.137)
8π 3ka
Now the quality factor of the dipole can be computed from

ω (We +Wm )
Q̆dipole =
Pr
 
η (I0 ∆ ℓk)2 4 1
+
8π (3ka) (ka)3
=
η (I0 ∆ ℓk)2
3 4π
 3
1 3 1
= + . (3.138)
ka 2 ka

The maximum value a can assume is ∆ ℓ/2, and noting that for a Hertzian dipole, k∆ ℓ/2 ≪ 1,
the Hertzian dipole quality factor can be approximated as
 3
3π 3 λ
Q̆dipole ≈ , (3.139)
2 ∆ℓ

where λ = up / f = 2π /k is the wavelength.

Example 3-4: Equivalent Circuit for a Transmitting Hertzian Dipole

An antenna connected to a source presents an equivalent circuit to which the source power is
delivered. Using the expressions for the radiated power and reactive power, find the equivalent
circuit for the Hertzian dipole.
Solution: The radiated power Pr delivered to space is lost power, similar to the power
delivered to a resistor. Also referring to Eq. (3.135) the stored reactive energy near the dipole
is predominantly electrical or capacitive. It is also noted from Eqs. (3.132) and (3.135) that the
dissipative (radiation) power and the reactive energy are proportional to I02 . This indicates that
the same current is flowing through the equivalent resistor and capacitor and therefore these
two elements are in series. Figure 3-6 shows the equivalent circuit for the Hertzian dipole.
136 Chapter 3 The Electromagnetic Potentials and Radiation

Equivalent circuit
Hertzian for Hertzian dipole
dipole
Cd
Δl
Rd

Figure 3-6: A source connected to a Hertzian dipole and its equivalent series RC circuit.

1
Noting that the power delivered to the equivalent resistor is 2 Rd I02 , the equivalent radiation
resistance of the Hertzian dipole can be calculated from
 2
2Pr 2πη ∆ℓ
Rd = 2 = . (3.140)
I0 3 λ

In free space, η = 120π and therefore the radiation resistance is

 2
2 ∆ℓ
Rd = 80π (Ω) . (short dipole)
λ

Note that the radiation resistance is a function of frequency (or wavelength) and decreases
rapidly with decreasing frequency. The equivalent capacitance is obtained from the reactive
power ω (We + Wm ). Comparing this reactive power to the reactive power stored in an
equivalent capacitor, I02 /(2ωCd ), the dipole equivalent capacitance is given by

I02
Cd =
2ω 2 (We +Wm )
1
=  2  . (3.141)
∆ℓ 4 1
πωη +
λ (3ka) (ka)3

For a Hertzian dipole, k ∆ ℓ/2 ≪ 1, so choosing a = ∆ ℓ/2 and then ignoring 1/(ka) compared
with 1/(ka)3 , a very simple formulation for the capacitance is obtained:

π
Cd = ε ∆ℓ . (short dipole) (3.142)
2

It is interesting to note that this capacitance is not a function of frequency, and thus the
antenna’s capacitor acts like a lumped capacitor. For free space, ε = 361π × 10−9 F/m, and
3-4 Far-Field Distance Criterion 137

the explicit expression for the Hertzian dipole series capacitance is given by

Cd = 13.89 × 10−12 ∆ ℓ (F) . (3.143)

3-3.1 Simplification of Differential Operators in the Far-Field Region

In the far-field region where fields are in the form of transverse electromagnetic waves with
spatial functions of the form eikr /r, such as in Eqs. (3.127a) and (3.127b), the differential
operators can be simplified by replacing the operator ∇ with ik r̂. For example, in the far-field
region, the magnetic field intensity can be derived from
1 1
H(r) = ∇ × A(r) = ikr̂ × A(r) , (3.144)
µ µ
and we can use Eq. (3.122) to obtain an explicit expression for the far-field approximation for
the magnetic field of a Hertzian dipole:
′
!
1 µ I0 eik|r−r |
H(r) = ikr̂ × dℓℓ
µ 4π |r − r′ |
′
ikI0 eik|r−r |
= r̂ × dℓℓ
4π r
ik
= (Aθ φ̂ φ − Aφ θ̂ θ) . (3.145)
µ
Also, Eq. (3.86) in the far-field is reduced to
iω
E(r) = ∇∇ · A + iω A
k2
iω
≈ 2 (ikr̂)(ikr̂) · A(r) + iω A(r)
k
= iω [A(r) − Ar (r) r̂] = iω (Aθ θ̂
θ + Aφ φ̂
φ) . (3.146)

When both electric and magnetic sources are present, duality and superposition can be invoked
to obtain the following far-field expressions for the electric and magnetic fields:
ik
E(r) = iω (Aθ θ̂
θ + Aφ φ̂
φ) − (Amθ φ̂
φ − Amφ θ̂
θ) (3.147a)
ε
and
ik
H(r) = (Aθ φ̂ θ) + iω (Amθ θ̂
φ − Aφ θ̂ θ + Amφ φ̂
φ) . (3.147b)
µ

3-4 Far-Field Distance Criterion

The far-field distance for electromagnetic sources was described qualitatively earlier. To
quantify a measure for the far-field distance, we start with small antennas like the short
138 Chapter 3 The Electromagnetic Potentials and Radiation

Hertzian dipole. Consider the magnetic field given by Eq. (3.123). Requiring the first term
in the expression to be much larger than the second term, i.e., requiring the Biot-Savart’s
component of the magnetic field to be much smaller than the radiation term, leads to
1
≪1,
k|R|

where R = |r − r′ |. If we adopt a factor of 10 for the ratio of the radiation component to the
nonradiation component, then

10 10λ
R= = = 1.6λ
k 2π
is considered to be the far-field distance for a Hertzian dipole. This criterion can be used for
all antennas whose dimensions are small compared to the wavelength. However, for antennas
whose dimensions are comparable to or larger than the wavelength, the far-field criterion can
be obtained by examining the phase difference between the radiated fields emanating from
the center and the edge of the antenna. For an antenna with a maximum dimension of D, as
shown in Fig. 3-7, the phase difference between the radiated fields emanating from the center
and the edge of the antenna can be computed from
   
D2
∆ φ = k(Re − R0 ) ≈ k R0 1 + 2 − R0
8R0
1 D2
= k . (3.148)
8 R0
Requiring this phase difference to be less than π /8, the far-field distance is given by

2D2
R0 ≥ . (3.149)
λ

e ≈

Figure 3-7: Geometry of an antenna with maximum dimension D and an observation point at
distance R0 from the antenna center.
3-5 Small Loop of Current: A Hertzian Magnetic Dipole 139

This criterion ensures that the variations of the field with distance can be expressed as eikr /r
with a maximum phase error of less than 22.5◦ .

3-5 Small Loop of Current: A Hertzian Magnetic Dipole

As was noted earlier, magnetic dipoles observed in certain materials are the result of electron
orbital motion, or electron or nuclear spin. The motion of electrons along a closed loop can
generate a magnetic field and a permanent magnetic dipole moment. This permanent magnetic
dipole moment can react to an external magnetic field excitation. In this section we show that
when the loop radius is small (compared with the wavelength as well as with the observation
distance from the loop center), the electromagnetic field produced by the loop is the same
as what a Hertzian magnetic dipole produces. To establish this equivalence, we consider a
small circular loop, carrying a time-harmonic electric current, with radius a centered in the
x–y plane of a Cartesian coordinate system, as shown in Fig. 3-8. The current distribution can
be expressed mathematically as

J(r) = I0 δ (ρ − a) δ z φ̂
φ. (3.150)

The magnetic vector potential generated by this current loop is given by

Z 2π ′
µ I0 eik|r−r |
A(r) = (− sin φ ′ x̂ + cos φ ′ ŷ)a d φ . (3.151)
4π 0 |r − r′ |

Noting that r′ = a(cos φ ′ x̂ + sin φ ′ ŷ) and that

r = r(sin θ cos φ x̂ + sin θ sin φ ŷ + cos θ ẑ) ,

z
Observation point

r |r - r′|
a

r′
y
x

Figure 3-8: Geometry of a circular loop antenna of radius a and an observation point at distance
R0 from the antenna center.
140 Chapter 3 The Electromagnetic Potentials and Radiation

it follows that q
|r − r′ | = r2 + a2 − 2ra sin θ cos(φ − φ ′ ) . (3.152)
Using a Taylor series expansion of the integrand around a = 0 yields
′  
eik|r−r | eikr eikr 1
≈ + −ik + sin θ cos(φ − φ ′ ) a . (3.153)
|r − r′ | r r r
After inserting Eq. (3.153) into Eq. (3.151) and noting that the first term will integrate to zero
and
Z 2π
sin φ ′ cos(φ − φ ′ ) d φ ′ = π sin φ (3.154a)
0
and
Z 2π
cos φ ′ cos(φ − φ ′ ) d φ ′ = π cos φ , (3.154b)
0

we obtain the following closed-form expression for the magnetic potential:

 
µ I0 π a2 eikr 1
A(r) = −ik + sin θ φ̂
φ. (3.155)
4π r r

The electric field can be obtained from Eq. (3.86); however, noting that A(r) has only one
component along φ̂ φ but it is not a function of φ , it is obvious that ∇ · A(r) = 0. Therefore,
 
iω µ kI0 π a2 eikr 1
E(r) = −i + sin θ φ̂
φ. (3.156)
4π r kr
Also, the magnetic field can be derived from
1
H(r) = ∇ × A(r) ,
µ
which yields
     
−ik2 I0 π a2 eikr 1 i 1 i
H(r) = 2 + cos θ r̂ + −i + + sin θ θ̂ .
θ (3.157)
4π r kr (kr)2 kr (kr)2

The fields for a Hertzian magnetic dipole, with a current moment I0m ∆ ℓ, can be obtained
from Eqs. (3.129) and (3.130) using the duality relations. These expressions are exactly the
same as Eqs. (3.156) and (3.157) if we set

I0m ∆ ℓ = −ikηπ a2 I0 . (3.158)

Hence, we emphasize here that a small current loop acts like a Hertzian magnetic dipole.
Considering that I0m ∆ ℓ = −iω qm ∆ ℓ, the equivalent magnetic dipole moment of a current
3-5 Small Loop of Current: A Hertzian Magnetic Dipole 141

loop is given by
pm = qm ∆ ℓ = µπ a2 I0 (Wb m). (3.159)
In the literature, the quantity m = π a2 I0 is referred to as the magnetic moment of the loop.
In a medium with N magnetic moments per volume, the magnetization vector M = Nm A/m,
and as was shown before in Eq. (1.49) of Chapter 1, the magnetic flux density can be obtained
from B = µ0 H + µ0 M.
Given the equivalence between a small current loop and a Hertzian magnetic dipole, the
radiated power and the equivalent circuit model for a small loop antenna can be easily obtained
using duality. The dual of Eq. (3.132), using Eq. (3.158), is given by

η |I0m ∆ ℓ|2 k2 η k4 A2 |I0 |2

Pr = = , (3.160)
3 4π 12π
where A = π a2 . The radiation resistance, therefore, is

 2
8π 3 η A
Rloop = . (small loop antenna) (3.161)
3 λ2

The dual of the equivalent circuit of Fig. 3-6 is an inductor in series with Rloop . The equivalent
inductance can be calculated from
ω (We +Wm )
Lloop = 1 2
. (3.162)
2 ω |I0 |

The expressions for We and Wm can be obtained by applying duality to Eqs. (3.137) and
(3.135), respectively. In the application of the duality principle, the quantity I0 ∆ ℓ should
be replaced with −ikηπ a2 I0 . Choosing the radius of the enclosed volume to be the same as
the radius of the loop, and after some simple algebraic manipulations, the loop inductance is
found to be
π
Lloop = µ a . (3.163)
4
In most low-frequency applications, multiturn loops are used quite frequently. In a
multiturn loop antenna with N turns, the current contributing to the radiation is NI0 , whereas
the terminal current is I0 . Hence the radiation resistance and loop inductance are given by

 
8π 3 η A 2
Rloop = N 2 (3.164a)
3 λ
and (N-turn loop antenna)
π
Lloop = N 2 µa . (3.164b)
4
142 Chapter 3 The Electromagnetic Potentials and Radiation

Example 3-5: Magnetic Dipole Moment of Hydrogen Atoms: A Classical

Approach

Use the approximate Bohr’s model to determine the resulting magnetic moment m for a
hydrogen atom. Assume the orbital radius of the electron is a = 5.3 × 10−11 m.
Solution: According to Bohr’s model, a single electron orbits around the nucleus, which
contains a single proton. The motion of the circling electron constitutes an electric current in
the opposite direction of the electron motion. If the velocity u and the circumference of the
electron orbit 2π a are known, the equivalent electric current can be obtained from
eu
I0 = . (3.165)
2π a
Using classical mechanics, the electron velocity can be estimated by balancing the
electrostatic force, between the electron and proton, against the centrifugal force of the
electron. That is,
e2 me u2
= ,
4πε a2 a
from which the electron velocity is found to be
e
u= √
4πε me a
1.6 × 10−19
= √
4π × 8.86 × 10−12 × 9.1 × 10−31 × 5.3 × 10−11
= 2.2 × 106 m/s.

Here me = 9.1 × 10−31 kg is the mass of an electron. Hence the equivalent electric current of
a single orbiting electron in a hydrogen atom is

I0 = 1.05 mA ,

and its magnetic moment is calculated to be

m = π a2 I0 = 3.1415 × (5.3 × 10−11 )2 × 1.05 × 10−3

= 9.26 × 10−24 (A m2 ).

It should be noted that Bohr’s model is not an accurate model for atoms with more than
a single electron. One contradiction to the above model is that a loop current radiates
electromagnetic fields, and therefore the electron should lose its momentum and fall into
its nucleus. Bohr explained this contradiction by stating that the orbital radius of an electron
around its nucleus is quantized with the smallest radius being a = 5.3×10−11 , which is known
as the “ground state.” Hence the electron is not allowed to get closer to the nucleus, and larger
orbits can only be achieved if the electron receives specific quantized energy levels from
3-5 Small Loop of Current: A Hertzian Magnetic Dipole 143

the electromagnetic waves (photons). Because of its magnetic dipole moment, it seems like
hydrogen is a paramagnetic element. However, hydrogen in its natural form appears as an H2
molecule with two electrons having opposite spins, thereby producing a net zero magnetic
dipole moment. Monoatomic hydrogen can exist only at extremely high temperatures.

3-5.1 Equivalence between Small Current Loops and Short Magnetic Dipoles:
An Alternative Method
In the previous section, the equivalence between a small loop carrying an electric current
and a short magnetic dipole perpendicular to the loop surface was established by comparing
the radiated fields from both. In this section an alternative approach based on Maxwell’s
equations is presented. Let us denote the field solutions to Maxwell’s equations generated by
an impressed electric current distribution J(r) as E1 (r) and H1 (r). In this case,

∇ × E1(r) = iω µ H1 (r) (3.166a)

and
∇ × H1(r) = −iωε E1 (r) + J(r) . (3.166b)

By taking the curl of Faraday’s law and then using the modified Ampère’s law, it can easily
be shown that
∇ × ∇ × E1(r) − k2 E1 (r) = iω µ J(r) . (3.167)
Now let us consider another problem for which the fields are produced entirely by a magnetic
current distribution Jm (r). We denote these fields as E2 (r) and H2 (r):

∇ × E2(r) = iω µ H2 (r) − Jm (r) (3.168a)

and
∇ × H2(r) = −iωε E2 (r) . (3.168b)

Similar to the previous case, by taking the curl of Faraday’s law and using Ampère’s law, it
can be shown that
∇ × ∇ × E2(r) − k2 E2 (r) = −∇ × Jm (r) . (3.169)
If we require the field solutions to be the same, while maintaining the same boundary
conditions, then the following relation between the sources must apply:

∇ × Jm (r) = −iω µ J(r) . (3.170)

Consider a small current loop carrying current I and a uniform magnetic current density Jm of
height ∆ ℓ flowing in a direction perpendicular to the loop with its extent confined to the loop’s
cross section as shown in Fig. 3-9. Applying a surface integral on both sides of Eq. (3.170)
and then applying Stokes’ theorem, we have
I
−iω µ J(r) ds = Jm (r) dℓ . (3.171)
S C
144 Chapter 3 The Electromagnetic Potentials and Radiation

Jm S
Δl
y

Figure 3-9: A circular current loop with a small radius, a cross-section area A, and a short
uniform magnetic current distribution confined within the loop. This small uniform magnetic
current distribution can be approximated as a short magnetic current filament Jm A = Im at the
center of the loop.

Consider S to be a small rectangular surface perpendicular to the surface that contains the
current loop and has its contour inside the loop, as shown in Fig. 3-9. It then can be shown
that
−iω µ I = Jm ∆ ℓ. (3.172)
Multiplying both sides of the above equation by the area of the loop A and noting that the
total magnetic current is Im = Jm A, the equivalence between the loop current and the magnetic
current filament is found to be
Im ∆ ℓ = −ikη IA , (3.173)
which is consistent with Eq. (3.158). It is noted that the equivalence of the current loop and
a short uniform magnetic current distribution within the loop is exact. Applying the duality
principle, it can also be shown that a magnetic current loop carrying current Im is equivalent
to a short Hertzian dipole if the following relation holds:
ik
I ∆ℓ = Im A . (3.174)
η

3-6 Wire Antennas

In the previous sections, we presented the procedures for computing the fields radiated by
arbitrary electric and magnetic sources of known distributions. But how are these currents
established in space? The currents and voltages generated by active circuits at a desired
frequency are used to derive electric current distributions using conductors with very high
conductivities. The shapes of these conductors connected to these sources dictate the current
distribution function in space. Engineering the desired current distribution function using
metallic wires and surfaces, which are referred to as antennas, is a complicated task.
3-6 Wire Antennas 145

Figure 3-10: An open-ended two-wire transmission line supporting a standing-wave current

distribution. Shown on the right side is a dipole antenna formed by bending the wires of the
transmission line at the antenna terminal points to form a dipole antenna that supports a current
distribution similar to that along the transmission line.

However, for certain antenna geometries like straight wires and loops, the current distributions
can be accurately estimated. For example, along a dipole antenna of arbitrary length, the
current distribution can be approximated from what would have existed on an open-ended
transmission line if the dipole wires were straightened along the feeding transmission line, as
shown in Fig. 3-10.
Assuming the length of the dipole is ℓ, the current distribution is approximated by
I0
I(z) = sin[k(ℓ/2 − |z|)] , (3.175)
sin(kℓ/2)
where I0 is the current at the antenna terminal. This equation ensures that the current at the end
of the wires on the dipole goes to zero as expected. This current can be used in Eqs. (3.127a)
and (3.127b) to find the electric and magnetic fields in the far-field region. When ℓ ≪ λ , the
current distribution on the dipole can be approximated by a linear current distribution with
the current being I0 at the terminal and going to zero at the end. In this case, the radiated
fields can be shown to be proportional to the average current on the antenna (I0 /2) and hence
the radiated and reactive powers are reduced by a factor of four compared with what was
calculated for a Hertzian dipole with a constant current. In this case the radiation resistance
and capacitance are

 2  2
πη ℓ 2 ℓ
Rshort dipole = = 20π (3.176a)
6 λ λ
and (short dipole with linear current distribution)
Cshort dipole = 2πε ℓ . (3.176b)
146 Chapter 3 The Electromagnetic Potentials and Radiation

The half-wave dipole antenna is the most commonly used antenna configuration. The
length of this antenna is ℓ = λ /2, and according to Eq. (3.175) the current distribution on this
antenna is of the form

I(z) = I0 cos(kz), − λ /4 ≤ z ≤ λ /4 . (3.177)

For this case, ẑI(z′ ) dz′ can be substituted for J(r′ ) dv′ in Eq. (3.127b) to find the electric field
Z λ /4
−ikη I0 eikr ′
E(r) = (ẑ × r̂) × r̂ cos(kz′ ) e−ikz cos θ dz′
4π r −λ /4
Z λ /4 ′ ′
−ikη I0 eikr eikz + e−ikz ikz′ cos θ ′
= sin θ θ̂
θ e dz
4π r −λ /4 2
−iη I0 eikr cos( π2 cos θ )
= θ.
θ̂ (3.178a)
2π r sin θ
Similarly, the magnetic field is calculated to be

−iI0 eikr cos( π2 cos θ )

H(r) = φ.
φ̂ (3.178b)
2π r sin θ
The time-average power density can now be easily calculated:

 2
|I0 |2 η cos( π2 cos θ )
Savg (r) = 2 2 , (λ /2 dipole) (3.179)
8π r sin θ

from which the total radiated power can be evaluated:

Z 2πZ π  cos( π 2
|I0 |2 η 2 cos θ )
Pr = sin θ d θ d φ . (3.180)
8π 2 0 0 sin θ
The integral can be evaluated numerically to yield

Pr = 36.56|I0 |2 , (3.181)
1
and noting that Pr = 2 Rr |I0 |2 , the radiation resistance for a half-wave dipole is

Rr = 73.13 Ω . (λ /2 dipole) (3.182)

The directivity function defined as

4π r2 Savg (r)
D(θ , φ ) =
Pr
3-7 Equivalent Circuit for Receiving Antennas 147

◦
Half-wave dipole

◦ ◦

◦
◦

Figure 3-11: Radiation pattern of a half-wave dipole in the elevation plane (as a function of θ ).

for a half-wave dipole is given by

 2
cos( π2 cos θ )
D(θ , φ ) = 1.64 . (λ /2 dipole) (3.183)
sin θ

The radiation pattern of the half-wave dipole is shown in Fig. 3-11. As expected, there is no
dependence on φ and therefore the radiation pattern is uniform in the azimuthal plane. It is
also a maximum along θ = 90◦ , its directivity is 1.64, and the 3-dB beamwidth is 78◦ .

3-7 Equivalent Circuit for Receiving Antennas

It was shown previously that an antenna in transmission mode is viewed by a source as
an equivalent load impedance Zt , which is a combination of resistive and reactive lumped
elements. Now the question is: Can this equivalent circuit be used for the same antenna
when operating in the receiving mode? Electric current is excited on an antenna exposed to a
uniform incident electric field. Basically, the ambient electric field is the force function on the
free electrons on the surface of a metallic antenna. The receiving antenna is usually connected
to a receiver circuit that may be composed of a low-noise amplifier, mixer and a detector. If
we represent the input impedance of the receiver by ZL and the current at the terminal of the
antenna by IL , the receive input voltage is VL = ZL IL . The Thévenin-equivalent circuit of the
148 Chapter 3 The Electromagnetic Potentials and Radiation

Thevenin
Receiver Receiver equivalent
E(r)
circuit circuit circuit
IL ZA
Voc

Figure 3-12: A receiving antenna connected to a receiver circuit and its equivalent circuit model.

receiving antenna is composed of a voltage source Voc in series with a source impedance ZA ,
as shown in Fig. 3-12, where Voc is simply the open-circuit voltage that appears at the antenna
terminals if the receiver circuit is disconnected.

3-7.1 Equivalence of Antenna Impedance in Transmission and Reception

Modes
In what follows, it will be shown that the receive antenna impedance ZA is the same impedance
had the antenna been used as a transmitter. To show this equivalence, we consider three
experiments. In experiment 1, as shown in Fig. 3-13(1), we consider the antenna in a receiving
mode connected to a load impedance ZL and then we measure a current Ir that goes through
it. According to the Thévenin equivalent circuit of the antenna, the measured current is given
by
Voc
Ir = . (3.184)
ZA + ZL
In the second experiment, the same antenna is connected to a source with open-circuit voltage
Vt and a source impedance equal to ZL , and the current going through the source (It ) is
measured. Since the antenna is in transmitting mode, it presents a load impedance Zt to the
source and the current provided by the source is computed using
Vt
It = .
Zt + ZL
In the third experiment, the antenna is both exposed to the incident electric field and, at the
same time, connected to the source. In this experiment we adjust the open-circuit voltage of
the source to be
Vt = −Voc .
3-7 Equivalent Circuit for Receiving Antennas 149

E(r) E(r)
It Ir + It
Ir ZL
ZL
ZL Vt −Voc

Figure 3-13: Configurations of three experiments to show that the antenna equivalent impedance
in transmission and reception modes are the same. (1) An antenna in reception mode connected to
a load impedance ZL ; (2) the antenna of experiment 1 in transmitting mode connected to a source
with source impedance ZL ; (3) the same antenna in simultaneous reception and transmission
modes when the source voltage is adjusted to be −Voc .

In this case, because of the opposing voltage sources of equal amplitude and phase, no current
will flow through ZL . Using the superposition principle, the total current through ZL is

Ir + It = 0 .

That is,
Voc −Voc
+ =0. (3.185)
ZA + ZL Zt + ZL
This can be true only if
ZA = Zt , (3.186)
which indicates that the impedance of an antenna is exactly the same for both transmitting and
receiving modes. It should be emphasized that the above equation is valid for any arbitrary
choice of ZL .
 150 Chapter 3 The Electromagnetic Potentials and Radiation

Summary
Concepts

• Scalar and vector potential functions are intro- of the complex permittivity of the medium to
duced as tools to calculate fields generated by create an effective complex dielectric constant.
electromagnetic sources (electric and magnetic • Simple expressions for the electric and magnetic
current and charge distributions). fields generated from small current filaments,
• It is shown that the scalar and vector potentials known as Hertzian dipoles, are obtained. Using
satisfy the wave equation, which can be easily the superposition principle, these results are
solved for a homogeneous and unbounded expanded to include the field expressions for
medium. arbitrary time-harmonic sources.
• Using the superposition principle, the solution • The notion of equivalent circuits for transmit-
to the wave equation for point sources (retarded ting and receiving antennas in terms of radiation
potential) is used to find the potentials and the resistance and conductance are obtained, and it
fieds for any arbitrary distribution of electric and is shown that the same equivalent circuit model
magnetic current distributions. can be used for an antenna when operated in
• The notion of time-harmonic electromagnetic transmitting or receiving modes.
waves is introduced to reduce the dimension- • The concept of far-field distance for antennas is
ality of Maxwell’s equations. Expressions for introduced. At distances larger than the far-field
time-average power density and stored electric distance the dependence of field expressions
and magnetic energy density are obtained for with range is of the form 1r eikr .
time-harmonic waves. • The equivalence between a small current loop
• It is shown that the conductivity of a conducting and a magnetic Hertzian dipole, perpendicular
medium can be included in the imaginary part to the loop surface, is demonstrated.
 SUMMARY 151

Important Equations
Magnetic field in terms of magnetic vector potential: Time-average Poynting vector:
1
H(r) = ∇ × A(r) Savg (r) = 12 Re[E(r) × H∗ (r)]
µ
Complex power flow density:
Electric field in terms of potentials: S(r) = 12 E(r) × H∗ (r)
∂ A(r)
E(r) = −∇ Φ(r) − Time-average electric and magnetic energy
∂t
densities:
Lorentz gauge condition: ε′ µ′
∂ Φ(r) wavg
e (r) = |E(r)|2 , wavg
m (r) = |H(r)|2
∇ · A(r) = −µε 4 4
∂t
Time-average electric and magnetic dissipated
Wave equation: power densities:
∂ 2 A(r) avg
Ploss = 12 ω (ε ′′ |E|2 + µ ′′ |H|2 )
∇2 A(r) µε = −µ J(r)
∂ t2
2 Dielectric quality factor:
∂ Φ(r) ρ (r2 )
∇2 Φ(r) µε = − ε′
∂ t2 ε Q̆ = ′′r
εr
Electric and magnetic fields in terms of
Antenna directivity:
dielectric and magnetic Hertz potentials:
  Savg (r, θ , φ )
∂ 2Π ∂ Πm D(θ , φ ) =
E = ∇∇ ·Π Π − µε 2 − µ ∇ × Prad /(4π r2 )
∂t ∂t
  Quality factor of an antenna:
2
∂ Πm ∂Π
H = ∇∇ ·Π Πm − µε + ε∇ × ω (We +Wm )
∂ t2 ∂t Q̆ant =
Prad
Magnetic vector potential in terms of a Far-field expansion for field in terms of vector
volumetric electric
$current distribution: potentials:
µ J(r′ , t − |r − r′ |/up ) ′ ik
A(r,t) = dr E(r) = iω (Aθ θ̂ φ) − (Amθ φ̂
θ + Aφ φ̂ φ − Amφ θ̂
θ)
4π v |r − r′ | ε
ik
Characteristic impedance of a homogeneous H(r) = (Aθ φ̂ θ) + iω (Amθ θ̂
φ − Aφ θ̂ θ + Amφ φ̂
φ)
µ
medium: r
|E| µ Far-field distance:
= =η R ≥ 1.6λ for small antennas
|H| ε
2D2
Time-harmonic Maxwell’s equations: R≥ for large antennas
λ
∇ × E(r) = iω µ H(r) − Jm (r)
Relations between a current loop and Hertzian
∇ × H(r) = −iωε E(r) + J(r)
magnetic dipole:
∇ · B(r) = ρm (r)
I0m ∆ ℓ = −ikηπ a2 I0
∇ · D(r) = ρ (r)
Directivity of a Hertzian dipole and a λ /2 dipole:
Effective complex permittivity of a conducting
DHertzian = 23 sin2 θ
medium:  
εeff = ε + iσ /ω cos( π2 cos θ )
Dλ /2 = 1.64
sin θ
 152 Chapter 3 The Electromagnetic Potentials and Radiation

Important Terms Provide definitions or explain the meaning of the following terms:
antenna directivity gauge condition retarded potential
antenna equivalent circuit Hertz vector potential time-average electric energy density
antenna quality factor Hertzian dipole time-average magnetic energy density
characteristic impedance Hertzian magnetic dipole time-average Poynting vector
complex power flow density magnetic Hertz vector potential time-harmonic fields
dielectric loss tangent magnetic moment of current loop transverse electromagnetic wave
dielectric quality factor magnetic scalar potential vector phasor
effective permittivity magnetic vector potential wave equation
electric scalar potential quasi-static wave number
electric vector potential radiation resistance wire antenna
far-field distance receiving antenna

PROBLEMS
Electromagnetic Potentials

3.1 Consider a region defined by the perpendicular intersection of two semi-infinite PMC
and PEC planes. Let a filament current source I(t) exist in this space, as shown in Fig. P3.1.
Suppose the PEC and PMC surfaces coincide with the coordinate surface of a Cartesian
coordinate system and the expression for the current is given by
ŷ − x̂
I(t) = I0 √ δ (x − a, y − a, z) f (t) .
2

PMC a I

a
y

x PEC

Figure P3.1: A small current filament placed inside a corner reflector whose one side is made
from a semi-infinite PMC surface and whose other side is made from a semi-infinite PEC surface.
PROBLEMS 153

(a) Draw the equivalent images and write the expression for the image currents.
(b) Find the overall magnetic vector potential in the first quadrant.

3.2 A fundamental relation used to solve scattering problems states that the total E-field is
equal to the sum of the incident and scattered fields:

Et = Ei + Es .

Consider a thin vertical cylinder (along the z-axis) with relative permittivity εr , height L, and
small area A. The polarizability tensor of such a cylinder is given by
 
2
0 0
 εr + 1 
 
P=  0
2 .
 0
εr + 1 
0 0 1

Remember that the normalized polarizability tensor of a material relates the internal electric
field to the incident electric field:
Eint = P · Ei .
Let the incident field be given by
ŷ + ẑ
Ei = √ cos(ω t − kx) .
2

(a) Find the equivalent polarization current so that the medium may be treated as free space.
(b) Find the magnetic vector potential in the surrounding space due to the polarization
current.
(c) Give an expression for the scattered (and thus total) electric field in terms of A. You do
not need to simplify this expression.

Fields Radiated by Current Sources (Time Domain)

3.3 Evaluate ∇(r̂ · dℓℓ) for any arbitrary differential line element dℓℓ. Show that it is
proportional to 1/r.

3.4 Image theory has been proven to be valid for electrostatic fields using Coulomb’s law.
Prove the validity of image theory for a horizontal infinitesimal current source above an
infinite PEC sheet in the electrodynamic case.
Starting from the following electric field expression:
 R  p R !
I0 f f dt I0 ε /µ f f dt µ f ′
E= + 3 (r̂ · dℓℓ)r̂ − + + r̂ × (dℓℓ × r̂) ,
4πε up r2 r 4π r2 ε r3 r
154 Chapter 3 The Electromagnetic Potentials and Radiation

prove that the tangential electric field component vanishes everywhere on the surface. (Hint:
You can assume that the PEC sheet lies in the x–y plane and that the current is along the x̂
direction, and position your test point on the z-axis at z = z0 .)
3.5 Let the current on a thin metallic wire of length L be given by
L L
I = ẑ I0 cos(ω t), − <z< .
2 2
(a) Find the magnetic vector potential A in the surrounding space due to the wire.
(b) Find the far-field electric and magnetic fields from the magnetic vector potential.
(c) Compute the Poynting vector S(t) = E × H and its time average hSi.
3.6 Use the Poynting theorem to derive the expression for the radiated power from an
infinitesimal current source (Hertzian dipole).

J(r,t) = ẑ I(z) δ (x) δ (y) cos(ω t) ,

where I(z) = I0 (1−2|z|/∆ L) for ∆ L/2 < z < ∆ L/2. Find the radiation resistance of a Hertzian
dipole. Calculate this for ∆ ℓ = 0.01λ .
3.7 Derive the general expression for a time-dependent magnetic field along the z-axis of an
arbitrary loop located in the x–y plane and centered around the origin.
(a) Assume that the current is uniform along the loop [I = I0 f (t)].
(b) Assume that the current varies with φ as I = I0 | sin φ | f (t).

Time-Harmonic Fields

3.8 One of the most common and elementary types of antenna is the half-wavelength dipole.
Evaluate the fields produced by this dipole in the far field, and the corresponding total radiated
power, its radiation resistance, and its directivity.
The antenna consists of two thin wires, each of length h = λ /4, as shown in Fig. P3.8.
The signal to be transmitted is fed into the space between the rods. The current distribution
can be written as
(
I = I0 cos(ω t) cos(kz) ẑ −λ /4 < z < λ /4,
I = 0, otherwise.

(a) Give the phasor form for this current source.

(b) Find the far-field vector potential and radiated electric and magnetic fields.
(c) Find the Poynting vector S = E × H∗ due to these fields and determine the radiation
pattern for this antenna. The radiation pattern is simply |S|/[max(|S|)]. Sketch the
radiation pattern.
(d) Find the total radiated power P by integrating the Poynting vector over a spherical
surface. You may need to perform a numerical integration.
PROBLEMS 155

λ/4
I

λ/4

Figure P3.8: Configuration of a λ /2 dipole antenna fed by a two-wire transmission line with
terminal current I.

(e) The radiation resistance for such an antenna is defined as 2P/I02 . Find the radiation
resistance.
(f) The directivity of an antenna is defined by 4π r2 max(|S|)/P. Find the directivity for
this antenna.
3.9 Consider the dipole described in Problem 3.7. Now, place it in front of an infinite ground
plane as shown in Fig. P3.9. Find the radiation pattern (electric field distribution in the far field
as a function of θ and φ ) and plot it for (a) h = λ /4 and (b) h = λ /2.

PEC

Figure P3.9: Geometry of the λ /2 dipole of Problem 3.8 placed in front of an infinite ground
plane at a distance h.

3.10 The magnetic vector potential of a short dipole positioned at the origin can be derived
as
µ I0 eikr
A= dℓℓ .
4π r
156 Chapter 3 The Electromagnetic Potentials and Radiation

Two short vertical dipoles are placed above a perfect electric conductor (PEC) at a height h
and separated by a distance s, as shown in Fig. P3.10.

(−s/ 2, h) h (s/ 2, h)
x
−s/ 2 PEC s/ 2

Figure P3.10: Two infinitesimal current filaments with separation s placed at a height h above
an infinite ground plane. The direction of the currents is perpendicular to the ground plane.

(a) Derive an expression for the total magnetic vector potential in x–y plane. Simplify the
result for when the observation point is in the far-field region of the source.
(b) What is the magnetic vector potential in the far-field region when the perfect conductor
is replaced with a perfect magnetic conductor?

φ I0 cos(ω t) and radius a (with a ≪ λ ), as

3.11 For a small circular loop of current I = φ̂
shown in Fig. P3.11:

r̂
ϕ̂
θ̂

ϕ̂
ϕ̂′ ρ̂
ρ̂′

Figure P3.11: Configuration of a small loop carrying a constant current I.

(a) Determine the magnetic vector potential A.

PROBLEMS 157

(b) Use this vector potential to derive the electric field E and magnetic field H and the
time-average complex Poynting vector S = 12 (E × H∗ ) in the far field.
(c) Find the total power P radiated by the small loop by integrating the Poynting vector
over a spherical surface centered at the origin with radius R (such that R ≫ λ ).
(d) The radiation resistance for such an antenna is defined as 2P/I02 . Find the radiation
resistance for the small current loop. Hint:
Rπ
(i) 0 cos(nφ ) e jz cos π d φ = π jn Jn (z)
(ii) J1 (z) ≈ z/2 if z ≪ 1.

3.12
(a) For a small loop of radius r0 carrying current I1 and centered at the origin, as shown in
Fig. P3.12, show that the far-field magnetic vector potential is given by

−ik0 µ0 (π r02 I1 )
A= sin θ eik0 r φ̂
φ.
4π r
(b) Using this expression, find the far-field electric and magnetic fields.
(c) If a short dipole of length dl carrying current I2 is also placed along the z-axis at the
origin, choose the radius of the loop r0 so that the total far-field radiation pattern is
everywhere circularly polarized (Eθ and Eφ equal but 90◦ out of phase).

r0 I2
y

I1
x

Figure P3.12: Geometry of a composite source made of a constant current loop and a vertical
Hertzian dipole in the loop center and perpendicular to the surface of the loop.

3.13
(a) Using your results from Problem 3.12, show that the far-field electric and magnetic
fields of a small circular loop of radius a (with a ≪ λ ) carrying a time-harmonic
current I is the same as that of a short magnetic dipole of current Im , provided that
Im dl = −iω µ (π a2 )I.
(b) Now consider a small circular loop located at the origin excited by a time-harmonic
current
φ δ (r − a) δ (z) I0 cos(ω t) ,
I = φ̂
158 Chapter 3 The Electromagnetic Potentials and Radiation

a y
I N turns R
b
x

Figure P3.13: Configuration of a small loop carrying a uniform current used as a radio
transmitter and a solenoid with N turns used as the receiver located at a distance d from the
transmitter.

as shown in Fig. P3.13. A solenoid of N turns and radius b is placed in the far field of
the excited small loop and is loaded with a resistor R.

(i) Assuming that the height of the solenoid is much smaller than the wavelength,
calculate the inductance of the solenoid.

(ii) Find the electromotive force in the solenoid.

(iii) Calculate the power delivered to the load.

3.14 The electric and magnetic fields of a plane wave are given by

E = x̂ E0 e− j(ω t+kz)
and
E0 − j(ω t+kz)
H = ŷ e .
η
This plane wave is normally incident from z < 0 on a square aperture of side length a in an
infinite PEC ground screen at z = 0 (see Fig. P3.14). Assume that the field in the aperture
is identical to the field of the plane wave with the screen absent (this is called the Kirchhoff
approximation). Use the equivalence principle to find the corresponding sources. Find the
electric vector potential Am , and from that, determine the electric and magnetic fields in the
far field for z > 0.
PROBLEMS 159

kˆ
x E
H

Figure P3.14: A rectangular aperture in a planar metallic plate excited by a plane wave
illuminating the aperture from the lower half-space.

3.15
(a) Starting from the general definition of the stored magnetic energy per unit volume:
1
wm = B(r,t) · H(r,t),
2
prove that the time-average magnetic energy density is
1 ′ 2
wavg
m = µ |H| .
4
#
(b) Prove that the time-average dissipated power is Ploss = 1
2 (ω µ ′′ |H|2 + ωε ′′ |E|2 ) dv.

3.16 A cavity is a closed metallic structure that confines electromagnetic fields. The electric
field inside the rectangular cavity shown in Fig. P3.16 is given by
πx πy
E = − jE0 sin sin ẑ .
l a

(a) Find the magnetic field distribution inside the cavity, assuming free-space permittivity
and permeability for the medium inside the cavity.
(b) Find the electric and magnetic stored energy in the cavity.
(c) Now assume that the metallic walls of the waveguide are lossy with finite
conductivity σ . The electric field and the surface current on the surface of a lossy
conductor are related by the following equation:

n̂ × E = Zs n̂ × Js ,
160 Chapter 3 The Electromagnetic Potentials and Radiation

where Zs is the surface impedance given by

r
ωµ
Zs = (1 + j) .
2σ
The loss due to the finite conductivity of the metal is equal to the total power passing
through the surface penetrating into the metal and given by
"
1
P= Re(E × H∗ ) · ds .
2
Compute the total loss in the cavity for the given field distribution.
(d) Calculate the quality factor of the cavity defined as
We +Wm
Q̆ = ω .
Ploss

Figure P3.16: The geometry of a rectangular cavity with walls made from metallic plates of
finite conductivity.
 4 Chapter
Formal Solutions to Maxwell’s
Equations and Their Applications

Chapter Contents Objectives

Overview, 162 Upon learning the material presented in this chapter, you
4-1 Formal Solution of Helmholtz Equation, 162 should be able to:
4-2 Solution of Helmholtz Equation for a
Complex Medium, 168 1. Derive expressions for the electromagnetic
4-3 Integral Equations for Electromagnetic potentials in a homogeneous medium containing
Fields, 176 any number of objects with different constitutive
4-4 Integral Equation Formulation Based on parameters.
Equivalent Sources, 178
4-5 Integral Equation Formulation for Dielectric 2. Understand the behavior of the potentials and
Scatterers, 181 fields at distances far away from the sources,
4-6 Reciprocity Theorem, 184 known as the radiation condition.
4-7 Applications of the Reciprocity Theorem, 189
4-8 Babinet’s Principle, 205 3. Set up the integral equation formulations for
Chapter Summary, 216 fields and potentials in a heterogeneous medium.
Problems, 221
4. Explain the concept of reciprocity for electro-
magnetic waves and its applications in antenna
and scattering problems.

5. Understand the importance and applications of

Babinet’s principle in antennas and scattering
problems.

161
162 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Overview
The focus of this chapter is to use Maxwell’s equations to develop rigorous solutions for
electromagnetic fields in complex media formed from a heterogeneous mixture of different
objects of finite extent embedded in a homogeneous, isotropic, and unbounded medium, in
the presence of known electric and magnetic charge and current distributions. We begin by
applying the scalar Green’s function in a homogeneous unbounded medium to derive wave
equations in response to a spatial impulse function. The Green’s function is simply the spatial
impulse response of the wave equation. Green’s theorem is then used to assist in solving
the wave equation for the electromagnetic potentials in complex media. Also, the notion
of a radiation condition is discussed to express the spatial behavior of the electromagnetic
potentials and fields at observation points far from their sources. These solutions are used as
a foundation for developing integral equations for the electromagnetic potentials and fields.
The process leads to multiple formulations, including the extinction theorem and the electric-
field integral equation (EFIE) and magnetic-field integral equation (MFIE) for dielectric and
metallic objects, using equivalent sources on the surfaces of the objects. These equations
can be solved numerically for a wide range of scattering and antenna problems. Also in
this chapter, the reciprocity theorem, which establishes a symmetrical relation between two
different sources and their corresponding fields, is introduced. Applications of the reciprocity
theorem to antennas are discussed, and it is shown that the radiation pattern of an antenna
is the same in both transmission and reception modes. Also the concept of effective area of
a receiving antenna is introduced, and its relationship to the antenna directivity is derived.
Finally, Babinet’s principle for planar metallic structures is introduced, and the relations
between the fields of metallic sheets with an opening (aperture) and those forming the
complementary metallic sheet are presented. Babinet’s principle is used to relate the radiation
pattern, polarization, and input impedance of a planar antenna and those for its complementary
geometries. Antennas using self-complementary geometries, which makes them frequency-
independent, are discussed as well.

4-1 Formal Solution of the Helmholtz Equation

This section presents a formal procedure for deriving the solution of the scalar and vector
wave equations given by Eqs. (3.87a) and (3.87b). The results presented previously for the
retarded potential were obtained somewhat heuristically and are only valid for an unbounded
homogeneous medium. The formal solution presented in this section can, in principle, be
obtained for far more complicated media. We begin with the derivation of the spatial impulse
response of the scalar wave equation for an unbounded homogeneous medium, which is
known as Green’s function. Consider the following scalar wave equation with an impulse
excitation at the origin:
∇2 g(r) + k2 g(r) = −δ (r) . (4.1)
The source function (point source) is assumed to have spherical symmetry. In the context of
generalized functions, h(r) satisfies the following integral identity:
$
h(r′ ) δ (r − r′ ) dr′ = h(r) , (4.2)
V
4-1 Formal Solution of the Helmholtz Equation 163

which is valid for any continuous function h(r). Here dr′ represents the differential volume
in a general coordinate system. For example, in the Cartesian, cylindrical, and spherical
coordinate systems, dr′ assumes the following definitions, respectively:

dr′ = dx′ dy′ dz′ , (4.3a)

′ ′ ′ ′ ′
dr = ρ d ρ d φ dz , (4.3b)
dr′ = r2 sin θ ′ d θ ′ d φ ′ dr′ . (4.3c)

To ensure that Eq. (4.2) is satisfied for any continuous function, when expressing δ (r) in
terms of variables of a specific coordinate system, appropriate coefficients must accompany
the expanded forms. For example, for the three standard coordinate systems, the following
expanded forms for δ (r) are appropriate:

δ (r) = δ (x) δ (y) δ (z) , (4.4a)

1
δ (r) = δ (ρ ) δ (φ ) δ (z) , (4.4b)
ρ
1
δ (r) = 2 δ (φ ) δ (θ ) δ (r) . (4.4c)
r sin θ
To express spherical symmetry explicitly, one can use

δ (r)
δ (r) = , (4.5)
4π r2
since Z
$ R Z 2πZ π
δ (r) 2
δ (r) dr = r sin θ d θ d φ dr = 1 .
V 0 0 0 4π r2
Using Eq. (4.5) in Eq. (4.1) and noting that g(r) is only a function of r, the following second-
order differential equation is obtained:
 
1 ∂ 2 ∂ g(r) −δ (r)
r + k2 g(r) = . (4.6)
r2 ∂ r ∂r 4π r2
Equation (4.6) can be rearranged to yield

d2 δ (r)
[r g(r)] + k2 [r g(r)] = − . (4.7)
dr2 4π r
The general solution of Eq. (4.7) is found to be

eikr e−ikr
g(r) = A1 (k) + A2 (k) . (4.8)
r r
164 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Noting that eikr represents an outward-going wave and e−ikr an incoming wave, it is concluded
directly that A2 = 0. Hence,
eikr
g(r) = A1 (k) . (4.9)
r
Basically, since it is assumed that there are no sources in the medium except for the point
source at the origin, no incoming wave is allowed. To determine coefficient A1 (k), we integrate
Eq. (4.1) over a spherical volume of radius R and centered at the origin:
$ $
2
∇ · ∇ g(r) dv + k g(r) dv = −1 . (4.10)
V V

Applying the divergence theorem to the first term leads to

$
2
∇ g(r) · ds + k g(r) dv = −1 , (4.11)

and upon inserting Eq. (4.9), we have

   Z R
∂ eikR 2 2
A1 4π R + k 4π A1 reikr dr = −1 . (4.12)
∂R R 0

In the limit as R approaches zero the left-hand side approaches −4π A1 , and therefore
1
A1 = . (4.13)
4π
Hence,
eikr
g(r) = . (4.14)
4π r
To generalize the solution, let us assume that the point source is located at r′ . In this case the
function representing the point source is expressed by δ (r − r′ ) and simply through a change
of variable to R = r − r′ , a differential equation similar to Eq. (4.6) can be obtained. Denoting
the Green’s function by g(r, r′ ) and emphasizing that the observation point is at r and the
source point is at r′ , we have
′
eik|r−r |
g(r, r′ ) = . (4.15)
4π |r − r′ |
Because of the built-in symmetry of Eq. (4.15), the expression for g(r, r′ ) would remain
unchanged if we were to interchange the locations of the source and observation points. That
is,
g(r, r′ ) = g(r′ , r) . (4.16)
This is a statement of reciprocity, which will be explored in more detail later on in this chapter.
4-1 Formal Solution of the Helmholtz Equation 165

4-1.1 Green’s Theorem

Green’s theorem provides the necessary tool to construct the solution for any arbitrary source
function once the point spread function (Green’s function) is known. Let V be a closed region
of space bounded by a regular surface S. Let Φ and ψ be two scalar functions of position that
together with their first and second derivatives are continuous in V and on S. By applying the
divergence theorem to function ψ ∇Φ, it can be shown that
$
∇ · (ψ ∇Φ) dv = (ψ ∇Φ) · ds . (4.17)
V S

Recalling that ∇Φ · n̂ = ∂∂Φn for a unit normal vector n̂, and that ∇ · (ψ ∇Φ) =
ψ ∇2 Φ + ∇ψ · ∇Φ, Eq. (4.17) can be written as
$ $
2 ∂Φ
∇ψ · ∇Φ dv + ψ ∇ Φ dv = ψ ds . (4.18)
V V ∂n

Equation (4.18) is Green’s first identity. For the special case of ψ = Φ and ∇2 Φ = 0,
Eq. (4.18) reduces to $
2 ∂Φ
|∇Φ| dv = Φ ds . (4.19)
V S ∂n
Equation (4.19) can be used to prove the uniqueness theorem for static problems, which states
that the solution to Laplace’s equation in a region V is unique if the potential or its normal
derivative is known on a closed surface bounding the region.
Now interchanging ψ and Φ in Eq. (4.18), we have
$
2 ∂ψ
[∇Φ · ∇ψ + Φ ∇ ψ ] dv = Φ ds . (4.20)
V ∂n
Subtracting Eq. (4.20) from Eq. (4.18) yields
$  
2 2 ∂Φ ∂ψ
[ψ ∇ Φ − Φ ∇ ψ ] dv = ψ −Φ ds . (4.21)
V S ∂n ∂n
Equation (4.21) is Green’s second identity for scalar functions, which is often referred to as
Green’s theorem.

4-1.2 Solution of the Scalar Helmholtz Equation

In this section the solution to the scalar wave equation for an arbitrary source function
is obtained through the application of and Green’s function. Recall that Green’s function
satisfies
∇2 g(r, r′ ) + k2 g(r, r′ ) = −δ (r − r′ ) , (4.22)
166 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

and the scalar potential satisfies

ρ (r)
∇2 Φ(r) + k2 Φ(r) = − (4.23)
ε
for a homogeneous region V bounded by a closed surface S. According to Green’s theorem
 
  ∂ g(r, r′ )
$
′ 2 2 ′ ′ ∂ Φ(r)
g(r, r ) ∇ Φ(r) − Φ(r) ∇ g(r, r ) dv = g(r, r ) − Φ(r) ds .
V ∂n ∂n
(4.24)
Using Eqs. (4.22) and (4.23) in Eq. (4.24), the left-hand side of Eq. (4.24) simplifies to
$  
ρ (r)
−k2 g(r, r′ ) Φ(r) − g(r, r′ ) + k2 Φ(r) g(r, r′ ) + Φ(r) δ (r − r′ ) dv
V ε
 $
 ′ 1

Φ(r ) − ρ (r) g(r, r′ ) dv r′ ∈ V ,
ε V
= $

 1
− ρ (r) g(r, r′ ) dv r′ < V .
ε V

Interpretation of the case where r′ < V will be given later in this chapter. Hence, using
Eq. (4.15) for g(r, r′ ) the potential function, when the observation point is within V , can
be obtained from
$ ik|r−r′ |
" ik|r−r′ | ik|r−r′ |
!#
1 e 1 e ∂ Φ(r) ∂ e
Φ(r′ ) = ρ (r) dv + − Φ(r) ds .
4πε |r − r′ | 4π S |r − r |
′ ∂n ∂ n |r − r′ |
(4.25)
It is more common to use r as the observation point and r′ as the source point. Interchanging
r and r′ and using the reciprocity relation g(r, r′ ) = g(r′ , r), Eq. (4.25) is often written as

 ′ 
∂ Φ(r′ )
$
1 ′ ′ ′ ′ ′ ∂ g(r − r )
Φ(r) = ρ (r ) g(r, r ) dv + g(r, r ) − Φ(r ) ds′ ,
ε V S ∂ n′ ∂ n′

(4.26)
where r is the observation point, r′ is the source point and n̂′ is an outward unit vector on
surface S. It is noted here that Φ(r′ ) and ∂ Φ(r′ )/∂ n′ can be interpreted as equivalent surface
sources. In the absence of charge within volume V , the equivalent surface sources generated
by the exterior sources, is the reason for the nonvanishing potential Φ(r) inside S. Basically,
Φ(r′ ) and n̂′ · ∇Φ(r′ ) act as equivalent surface sources for observation points within S.

4-1.3 Radiation Condition

Let us assume that the charge density ρ (r′ ) is confined to finite volume V . Also assuming
that the medium is homogeneous, we can extend the closed surface to infinity. In this case
the surface contribution generated entirely by the volumetric charge density ρ (r′ ) is expected
4-1 Formal Solution of the Helmholtz Equation 167

r
R = r − r′
n̂′
r
r′

Figure 4-1: A volumetric charge density confined within volume V and equivalent surface
sources (∂ Φ/∂ n and Φ) on S that approaches infinity. The contribution from the equivalent
surface sources is expected to vanish at observation points far from surface S.

to vanish when the surface S is infinitely far away from the observation point and the source
function ρ (r′ ). Let us examine this equivalent surface source contribution more carefully by
evaluating  
′ ′
′ ∂ Φ(r ) ′ ∂ g(r, r )
I= g(r, r ) − Φ(r ) ds′ (4.27)
S ∂ n′ ∂ n′
as r′ → ∞, as shown in Fig. 4-1.
In this case the scalar Green’s function can be simplified to
′
eik|r−r |
g(r, r′ ) ≈ .
4π r′
Choosing a spherical surface for S, we have n̂′ = r̂′ and
′
!   ′
∂ g(r, r′ ) ∂ eik|r−r | ik 1 eik|r−r |
≈ ′ = − − ′ + ′2 .
∂ n′ ∂r 4π r′ r r 4π

In this case the surface integral is approximated by

" ik|r−r′ |   ik|r−r′ | #
e ∂ Φ(r′ ) −ik 1 e
I≈ ′ ′
+ Φ(r′ ) ′
+ ′2 ds′ . (4.28)
S 4π r ∂r r r 4π
168 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

We can ignore the 1/r′ 2 term compared with all other 1/r′ terms, and noting that ds′ = r′ 2 dΩ′ ,
we have "   ik|r−r′ | #
∂ Φ e
I≈ r′ ′
− ikΦ dΩ′ .
S ′ ∂ r 4 π
The only way to ensure that the contribution due to I vanishes is to require that
 ′ 
′ ∂ Φ(r ) ′
lim r − ik Φ(r ) =0. (4.29)
r ′ →∞ ∂ r′
This is known as the radiation condition, which was first introduced by Sommerfeld. It is
interesting to note that the free-space Green’s function satisfies the radiation condition.
The radiation condition for the electric scalar potential can be extended to the magnetic
vector potential in a similar form, as the magnetic vector potential is also a solution to a
similar wave equation. This condition is given by
 ′

′ ∂ A(r ) ′
lim r − ik A(r ) = 0 , (4.30)
r ′ →∞ ∂ r′

and the magnetic scalar potential φm and the electric vector potential Am (r) satisfy similar
equations. All these radiation conditions enforce the behavior of potentials as a function of
distance to be of the form eikr /r in the far-field region. This property was used to derive
the far-field expressions for the electric and magnetic fields as given by Eqs. (3.147a) and
(3.147b). Using these equations, it can easily be shown that the radiation conditions for the
electric and magnetic fields take the following forms:

lim r [E(r) + r̂ × η H(r)] = 0 (4.31a)

r→∞
and  
E(r)
lim r H(r) − r̂ × =0. (4.31b)
r→∞ η
Equations (4.31a) and (4.31b) indicate that in the far-field region the electric field is
perpendicular to the magnetic field and both are perpendicular to the direction of propagation.
In addition these equations
p indicate that the ratio of the electric field to the magnetic field is a
constant equal to η = µ /ε .

4-2 Solution of the Helmholtz Equation for a Complex Medium

The formulations provided in the previous section can be extended to include inhomo-
geneities. For example, let us consider a homogeneous medium with a finite number of
inclusions (dielectric or metallic) of finite extent, as shown in Fig. 4-2. Suppose the volumetric
charge density exists only in the background medium.
Initially let’s assume the background medium to be finite and bounded by a closed
surface S′ . The composite closed surface consisting of the surfaces of the inclusions and the
enclosing surface S′ encloses the homogeneous surrounding medium denoted by V . According
4-2 Solution of the Helmholtz Equation for a Complex Medium 169

V
n̂
ε2,μ2 S3

S2 ε3,μ3
n̂
ε1,μ1

σ=

8
n̂ S4
n̂
S' r

8
Figure 4-2: A complex medium composed of a number of homogeneous segments.

to Eq. (4.25),
$  ′ ′ 
1 ′ ′ ′ ′ ∂ Φ(r ) ′ ∂ g(r, r )
Φ(r) = ρ (r ) g(r, r ) dv + P
g(r, r ) ′
− Φ(r ) ′
ds′ .
ε1 V ′
S + Si ∂ n ∂ n

Now letting S′ approach infinity, the surface integral contribution according to the radiation
condition vanishes, and therefore
$  ′ ′

1 ′ ′ ′ ′ ∂ Φ(r ) ′ ∂ g(r, r )
Φ(r) = ρ (r ) g(r, r ) dv + P g(r, r ) − Φ(r ) ds′ (4.32)
ε1 V Si ∂ n′ ∂ n′

for any observation point in V . Here it should be emphasized that contributions from surfaces
connecting S′ and other inhomogeneities (shown by dashed lines in Fig. 4-2) cancel each
other, as the adjacent surfaces have unit normals opposite to each other. Note that in the
absence of inclusions, $
1
Φi (r) = ρ (r′ ) g(r, r′ ) dv′ .
ε1 V
The surface integral
 ′ ′

′ ∂ Φ(r ) ′ ∂ g(r, r )
Φs (r) = P
g(r, r ) ′
− Φ(r ) ′
ds′
Si ∂ n ∂ n
170 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

can be interpreted as the scattered potential. Here it is assumed that the observation point is
within V and outside the volume containing the scatterers. In this case, the total potential Φ(r)
is the superposition of the incident potential and the scattered potential:

Φ(r) = Φi (r) + Φs (r) . (4.33)

In general we can write

(
Φ(r) r ∈V ,
Φi (r) + Φs (r) = (4.34)
0 r <V .

Equation (4.34) is a statement of the extinction theorem. It indicates that when the observation
point is within a scatterer (r < V ), the equivalent surface sources on the scatterers are formed
in such a way that the potential produced by them completely eliminates the incident potential
everywhere inside the scatterers. This is physically a necessity, as for example if one (or more)
of the scatterers is a perfect conductor, the potential is zero (or a constant) on and within the
scatterer, and hence the surface charge density ρs (r) = n̂·D(r) = −ε n̂·∇Φ(r) must be formed
in such a manner as to cancel out the incident potential that was present in the absence of the
scatterer.
Also, in the case of dielectric scatterers, the potential inside a dielectric scatterer with
dielectric constant ε j and closed surface S j can be obtained using Green’s second identity in
a similar manner and is given by
 (
′ ′ 
′ ∂ Φ j (r ) ′ ∂ g j (r, r ) ′ Φ j (r) r ∈ Vj ,
g j (r, r ) ′
− Φ j (r ) ′
ds = (4.35)
Sj ∂ n ∂ n 0 r < Vj .

Here we are assuming that there are no charge distributions inside this scatterer and that
′
1 eik j |r−r |
g j (r, r′ ) =
4π |r − r′ |
is the scalar Green’s function for the medium with dielectric constant ε j . Again, according to
the extinction theorem, the equivalent sources on the surface assume a distribution such that
they produce no potential with wave propagation constant k j outside the scatterer. This is a
physical requirement as the wave outside must have a different propagation constant (k).
It should be noted that when the observation point is on the surface, it is possible that
r′ = r, and this makes the scalar Green’s function and its gradient on the surface singular.
This singularity can be removed for smooth surfaces by first deforming the surface around the
observation point and then evaluating the surface integral contribution around the deformation
as the deformation approaches zero. To further elaborate on this point, let’s assume that the
observation point r is on S j . We deform this surface around r by placing a hemispherical
surface of radius a centered at r and then allowing a → 0. This configuration is shown in
Fig. 4-3, where S j is deformed by introducing a vanishing hemispherical surface Ss around
4-2 Solution of the Helmholtz Equation for a Complex Medium 171

Ss
n̂′ a
Sj

r′ r

Figure 4-3: This figure shows a portion of the interface between the jth scatterer and the
background medium. To extract the contribution from the singularity of the integrand, the surface
is deformed by adding a small hemispherical surface with radius a around the singularity point r
on surface S j . The contribution from this hemispherical surface is calculated analytically in the
limit as a → 0.

the observation point. The scattered potential can be calculated from

"  
∂ Φ(r′ ) ′ ∂ g(r, r )
′
Φs (r) = lim g(r, r′ ) ′
− Φ(r ) ′
ds′
a→0 S j −Ss ∂ n ∂ n
 #
∂ Φ(r ′) ∂ g(r, r ′) 
+ g(r, r′ ) ′
− Φ(r′ ) ′
ds′ .
Ss ∂ n ∂ n

On Ss , the scalar Green’s function and its normal derivative are

eika
g(r, r′ ) =
4π a
and
 
∂ g(r, r′ ) eika 1
=− ik − .
∂ n′ 4π a a

where the negative sign is due to the fact that n̂′ is pointing inward on Ss . Also noting that
ds′ = a2 sin θ d θ dΦ, the contribution from the first integrand of the surface integral on Ss
172 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

vanishes as a → 0 and
  Z Z
′ ∂ Φ(r )
′ ′
′ ∂ g(r, r ) ′ −Φ(r) π /2 2π
lim g(r, r ) − Φ(r ) ds = sin θ d θ dΦ
a→0 Ss ∂ n′ ∂ n′ 4π 0 0

−Φ(r)
= .
2
Hence Eq. (4.34) should be modified to

 ′ ′  
Φ(r) r ∈V ,
′ ∂ Φ(r ) ′ ∂ g(r, r )
Φi (r) + −−− g(r, r ) − Φ(r ) ds = Φ(r)/2 r on S j , (4.36)
P
Si ∂ n′ ∂ n′ 

0 r <V ,

and Eq. (4.35) becomes


 ′ ′  Φ j (r)
 r ∈ Vj ,
′ ∂ Φ j (r ) ′ ∂ g j (r, r ) ′
−−− gi (r, r ) − Φ j (r ) ds = 12 Φ j (r) r on S j , (4.37)
∂ n′ ∂ n′ 
0 r < Vj .

Here −−− represents the principal-value integral.
Equation (4.36) can be used as an integral equation for metallic objects in order to find the
potential in the space on and around the scatterer for an arbitrary incident potential generated
by a known charge distribution ρ (r) outside the metallic scatterer. In the quasi-static regime,
assuming the scatterer is at potential zero (Φ(r) = 0, r on S) and solving for ρs (r), Eq. (4.36)
can be used to form the following integral equation, known as the Fredholm integral equation
of the first kind:
 

1
Φi (r) = −Φs (r) = −−− g(r, r′ ) ρs (r′ ) ds′ , (4.38)
ε S
(metallic scatterers)

where r is on the surface of the scatterer S. Equation (4.38) is to be solved for ρs (r), and once
that is found by placing it in the top part of Eq. (4.36), the potential everywhere else (r ∈ V )
can be obtained.
For dielectric objects a coupled integral equation can be formed in terms of the unknown
equivalent source on the surface using the boundary conditions

Φ(r) = Φ j (r) (4.39a)

and
ε n̂ · ∇Φ(r) = ε j n̂ · ∇Φ j (r) . (4.39b)
4-2 Solution of the Helmholtz Equation for a Complex Medium 173

These boundary conditions for the scalar potential are derived in Section 4-2.2. Using these
in Eqs. (4.36) and (4.37), the desired integral equations are obtained:
 ′ ′ 
′ ∂ Φ(r ) ′ ∂ g(r, r ) Φ(r)
Φi (r) + −−− g(r, r ) ′
− Φ(r ) ′
ds′ = (4.40a)
Sj ∂n ∂n 2
and
 ′ ′ 
ε ′ ∂ Φ(r ) ′ ∂ g j (r, r ) Φ(r)
−−− g j (r, r ) ′
− Φ(r ) ′
ds′ = . (4.40b)
Sj ε j ∂n ∂n 2
(dielectric scatterers)

These equations can be solved for Φ(r) and ∂ Φ(r)

∂ n by discretizing the surface into a number
of finite facets and casting the equations in terms of matrix equations that can be solved
numerically.

4-2.1 Solution for the Vector Helmholtz Equation

For a homogeneous unbounded medium, the magnetic vector potential satisfies

∇2 A + k 2 A = − µ J .

In a Cartesian coordinate system the three components of A are related to the corresponding
components of the volumetric current density by
     
Ax Ax Jx
∇ Ay + k Ay = µ Jy  .
2  2   (4.41)
Az Az Jz

This constitutes three scalar Helmholtz equations similar to Eq. (4.23). Hence,
′
eik|r−r | ′
$
µ
A p (r) = Jp (r′ ) dv ,
4π V |r − r′ |
where the subscript p can be x, y, or z. Assembling all three components, the magnetic vector
potential for a homogeneous unbounded medium can be calculated from
ik|r−r′ |
$
µ ′ e
A(r) = J(r ) dv′ .
4π V |r − r′ |
For a complex medium composed of discrete metallic or dielectric scatterers the statement for
the magnetic vector potential takes the following form:

 ′ ′
 
A(r) r ∈V ,
′ ∂ A(r ) ′ ∂ g(r, r ) ′
Ai (r) + −−− g(r, r ) − A(r ) ds = A(r)/2 r on S j , (4.42)
P
Sj ∂ n′ ∂ n′ 

0 r <V ,
174 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

where Ai (r) is the magnetic vector potential due to the sources in the absence of the scatterer
and S j represent the closed surface of the jth scatterer. Also inside each scatterer, assuming
no interior current distribution is present, we can get
 ′ ′ 
′ ∂ A j (r ) ′ ∂ g j (r, r )
−−− g j (r, r ) ′
− A j (r ) ′
ds′
Sj ∂ n ∂ n


A j (r) r ∈ Vj ,
= A j (r)/2 r on S j , (4.43)


0 r < Vj .

For electrodynamics, using Eqs. (4.42) and (4.43), together with enforcing the continuity
of the tangential electric field [E(r) = iω A(r) − ∇Φ(r)] and magnetic field [H(r) =
1
µ ∇ × A(r)] at the interfaces between the dielectric media as well as using the Lorentz gauge
condition [∇ · A(r) = iω µε Φ(r)], leads to a set of coupled integral equations. Derivation
of these equations is somewhat lengthy and will not be shown here. Instead another set of
integral equations will be derived using the field equations and Green’s second identity.

4-2.2 Boundary Conditions for Electromagnetic Potentials

In previous chapters we discussed the importance of boundary conditions at interfaces
between dielectric and metallic media. In this section we examine how such boundary
conditions can be extended to electromagnetic potentials. Let us begin with the electric scalar
potential Φ(r) and the magnetic vector potential A(r), and their relationships to E(r) and
B(r):

B(r) = ∇ × A(r) (4.44a)

and
E(r) = iω A(r) − ∇ Φ(r) . (4.44b)

Considering a differential volume dv between two dielectric media, as shown previously in

Fig. 1-14 and repeated here as Fig. 4-4, and applying Eq. (1.92b), leads to
$ $
B(r) dv = ∇ × A(r) dv = − A(r) × ds ,
∆v ∆v ∆S

and hence in the limit as ∆ ℓ → 0 it can be shown that

n̂ × [A1 (r) − A2 (r)] = 0 , (4.45)

indicating the continuity of the tangential magnetic vector potential.

To examine the continuity of the normal component of the magnetic vector potential, the
Lorenz gauge condition
∇ · A(r) = iω µε Φ(r) (4.46)
4-2 Solution of the Helmholtz Equation for a Complex Medium 175

can be used together with the divergence theorem:

$ $
∇ · A dv = A · ds = iω µε Φ(r) dv . (4.47)
∆v ∆S ∆v

Referring to Fig. 4-4, in the limit as ∆ ℓ → 0, the right-hand side of Eq. (4.47) vanishes, and
we get
n̂ · [A1 (r) − A2 (r)] = 0 . (4.48)
In view of Eqs. (4.45) and (4.48), the magnetic vector potential is shown to be continuous
across the interface between two dielectric media; that is,

A1 (r) = A2 (r) , r∈S.

To show the continuity of electric scalar potential across two dielectric media, we recall
that
n̂ × (E1 − E2 ) = 0 . (4.49)
By combining Eqs. (4.45) and (4.49) in Eq. (4.44b), it becomes obvious that

n̂ × (∇Φ1 − ∇Φ2 ) = n̂ × ∇(Φ1 − Φ2 ) = 0 . (4.50)

For Eq. (4.50) to hold, it must be true that

Φ1 (r) = Φ2 (r) , r∈S. (4.51)

That is, the potential function across the surface must be continuous. Otherwise ∇(Φ1 − Φ2 )
becomes infinite. Equation (4.50) can also be independently derived noting that ∇ × ∇Φ = 0
and using Eq. (1.92b) for ∆ v of Fig. 4-4.

Region 1 n̂1
∆a

∆l

Region 2

Figure 4-4: A small cylindrical volume between two homogeneous media. The base of the
cylinder is chosen to be small enough so that the interface can be assumed flat locally.
176 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

The boundary condition for the normal component of the electric flux density is given by

ε1 n̂ · E1 (r) = ε2 n̂ · E2 (r) . (4.52)

Using Eq. (4.44b) and noting the continuity of the magnetic vector potential, it can easily be
shown that
∂ Φ1 ∂ Φ2
ε1 = ε2 . (4.53)
∂n ∂n
where as before, n̂ is a unit vector normal to the interface pointing towards medium 1.
Using duality, the continuity of the magnetic scalar potential and electric vector potential
across dielectric interfaces can be shown. Also, applying duality to Eq. (4.53), we can show
that
∂ Φm1 ∂ Φm2
µ1 = µ2 . (4.54)
∂n ∂n
Now let us consider a situation where the second medium is a perfect electric conductor.
According to Eq. (4.44b), since E2 (r) inside the conductor is zero, it then follows that
A2 (r) = 0 inside the conductor as well. Hence, on the surface of a perfect electric conductor

A1 (r) = 0 , r∈S.

Also, on the surface of a perfect conductor we have

n̂ × E1 (r) = 0 . (4.55)

Using Eq. (4.44b) for E1 (r) and noting that n̂ × A1 (r) = 0, Eq. (4.55) simplifies to

n̂ × ∇ Φ1 (r) = n̂ × (∇Φ1 · t̂ t̂ + ∇Φ1 · n̂ n̂)

∂ Φ1
= n̂ × t̂ =0, (4.56)
∂t
where t̂ is a unit vector tangential to the conductor surface. Equation (4.56) indicates that
∂ Φ1 /∂ t = 0, and thus the electric scalar potential on the surface of the conductors must be a
constant.

4-3 Integral Equations for Electromagnetic Fields

In this section we consider the derivation of integral equations for electromagnetic fields
applicable to antennas and scattering problems. In the previous chapter we demonstrated how
fields radiated from antennas with known current distribution can be evaluated. However,
in practice the determination of the current distribution on antenna surfaces cannot be
performed in a straightforward manner. Also, in scattering problems where the goal is to
determine the electromagnetic fields scattered by specific objects when illuminated by specific
sources, the solution cannot be easily realized except for a very limited number of objects
with simple geometries. In principle, integral equation formulations with computer-aided
numerical analysis can be used to solve for the current distribution on an antenna of arbitrary
4-3 Integral Equations for Electromagnetic Fields 177

shape, and the same procedure can be used to determine the fields scattered by an arbitrary
object. The formulation is derived directly from the fields’ wave equation instead of the
potentials. Let us consider electric and magnetic sources in a homogeneous and unbounded
medium with a finite number of metallic or dielectric objects, as was shown earlier in Fig. 4-2.
Maxwell’s equations in the background medium with permittivity ε1 and permeability µ1 and
with electric and magnetic sources J(r), ρ (r), Jm (r), and ρm (r) are given by

∇ × E(r) = iω µ1 H(r) − Jm (r) , (4.57a)

∇ × H(r) = −iωε1 E(r) + J(r) , (4.57b)
1
∇ · E(r) = ρ (r) , (4.57c)
ε1
1
∇ · H(r) = ρm (r) . (4.57d)
µ1
Taking ∇× of both sides of Eq. (4.57a) and using Eq. (4.57b), it can be shown that

∇ × ∇ × E(r) = k12 E(r) + iω µ1 J(r) − ∇ × Jm (r) . (4.58)

Noting that ∇ × ∇ × E(r) = ∇∇ · E(r) − ∇2E(r) and using Eq. (4.57c), we have
1
∇ × ∇ × E(r) = ∇ρ (r) − ∇2 E(r) . (4.59)
ε1
Also, using the equation of continuity (∇ · J(r) = iω ρ (r)) in Eq. (4.59), we have
1
∇ × ∇ × E(r) = ∇∇ · J(r) − ∇2 E(r) . (4.60)
iωε1
Finally, using Eq. (4.60) in Eq. (4.58), we obtain the following wave equation for the
electric field E:
∇2 E(r) + k12 E(r) = −̥
̥(r) , (4.61)
where
1
̥ (r) = iω µ1 J(r) − ∇ × Jm (r) − ∇∇ · J(r) (4.62)
iωε1
is the excitation vector for the electric field in the background medium. Application of duality
leads to the following wave equation for the magnetic field:

∇2 H(r) + k12 H(r) = −̥

̥m (r) , (4.63)

where
1
̥ m (r) = iωε1 Jm (r) + ∇ × J(r) − ∇∇ · Jm (r) . (4.64)
iω µ1
178 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Next, by applying Green’s second identity, as defined by Eq. (4.20), to Eq. (4.61), the
following equations for the electric field are obtained:
$  ′ ′ 
′ ′ ′ ′ ∂ E(r ) ′ ∂ g(r, r )
̥ (r ) g(r, r ) dv + P g(r, r ) ′
− E(r ) ′
ds′
v Si ∂ n ∂ n
(
E(r) r∈v,
= (4.65)
0 r<v,

where g(r, r′ ) is the free-space scalar Green’s function. Applying duality to Eq. (4.65), the
integral equation for the magnetic fields can be obtained from
$  ′ ′

′ ∂ H(r ) ′ ∂ g(r, r )
̥m (r′ ) g(r, r′ ) dv′ + P g(r, r ) ′
− H(r ) ′
ds′
v Si ∂ n ∂ n
(
H(r) r ∈V ,
= (4.66)
0 r <V .

The first integral expression in Eqs. (4.65) and (4.66) can be viewed as the incident electric
and magnetic fields. These are the fields that would exist in the absence of the scatterers. As
shown previously, when the observation point is on the surface, the singularity associated with
the scalar Green’s function can be removed from the surface integrals, and thus Eqs. (4.65)
and (4.66) can be written as

 ′ ′
 
E(r) r∈V ,
′ ∂ E(r ) ′ ∂ g(r, r ) ′ 1
Ei (r) + −−− g(r, r ) − E(r ) ds = 2 E(r) r on Si ,
P
Si ∂ n′ ∂ n′ 
0 r<V ,
(4.67a)
and

 ′ ′
 
H(r) r∈V ,
′ ∂ H(r ) ′ ∂ g(r, r ) ′
Hi (r) + −−− g(r, r ) − H(r ) ds = 21 H(r) r on Si ,
P
Si ∂ n′ ∂ n′ 
0 r<V ,
(4.67b)

where Ei (r) and Hi (r) represent the first integrals in Eqs. (4.65) and (4.66).

4-4 Integral Equation Formulation Based on Equivalent Sources

To derive an alternative integral equation based on surface fields using electromagnetic
potentials, let us start with the expression for the field quantities in terms of magnetic
potentials A and Am , as given by Eqs. (3.89a) and (3.89b):
iω 1
E= ∇∇ · A + iω A − ∇ × Am (4.68a)
k2 ε
4-4 Integral Equation Formulation Based on Equivalent Sources 179

and
iω 1
H= 2
∇∇ · Am + iω A + ∇ × A , (4.68b)
k µ
with
$
A(r) = µ J(r′ ) g(r, r′ ) dv′ (4.69a)
v
and
$
Am (r) = ε Jm (r′ ) g(r, r′ ) dv′ , (4.69b)
v

where g(r, r′ ) is the free-space scalar Green’s function. To derive an explicit expression for
the fields in terms of the currents, the following observations are in order:

∇ · [J(r′ ) g(r, r′ )] = J(r′ ) · ∇g(r, r′ ) (4.70a)

and
∇g(r, r′ ) = −∇′ g(r, r′ ) , (4.70b)

where ∇′ is the gradient operator acting on the prime coordinate variables.

Also, note that
∇ × [Jm (r′ ) g(r, r′ )] = −Jm (r′ ) × ∇g(r, r′ ) . (4.71)
First we evaluate ∇∇ · A using Eqs. (4.70a) and (4.70b):
$
∇∇ · A(r) = −µ ∇ J(r′ ) · ∇′ g(r, r′ ) dv′ . (4.72)
v

Using the following identity:

∇′ · [J(r′ ) g(r, r′ )] = J(r′ ) · ∇′ g(r, r′ ) + ∇′ · J(r′ ) g(r, r′ ) ,

and employing the divergence theorem, we get

$
′ ′ ′
∇ · [J(r ) g(r, r )] dv = g(r, r′ ) J(r′ ) · n̂′ ds′ . (4.73)
v S

The surface S here identifies a closed surface confining the volumetric current J(r′ ).
Obviously no currents are flowing outside S, and therefore n̂′ · J(r′ ) = 0 along the surface.
This renders the right-hand side of Eq. (4.73) equal to zero, and therefore
$
∇∇ · A = µ ∇′ · J(r′ ) ∇g(r, r′ ) dv′ . (4.74)
v
180 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Using Eqs. (4.74) and (4.71) in Eq. (4.68a), the explicit expression for the electric field is
given by
 ′  ′  
$ $
iη ′ ′ ′ ′
E(r) = 2
∇ · J(r ) ∇g(r, r ) + k J(r ) g(r, r ) dv + Jm (r′ ) × ∇g(r, r′ ) dv′ .
k v v
(4.75)
Applying duality to Eq. (4.75) gives an explicit expression for the magnetic field:
 ′ 
$
i
H(r) = ∇ · Jm (r′ ) ∇g(r, r′ ) + k2 Jm (r′ ) g(r, r′ ) dv′
ηk v
 ′ 
$
− J(r ) × ∇g(r, r′ ) dv′ . (4.76)
v

According to the equation of continuity,

∇′ · J(r′ ) = iω ρ (r′ )
and
∇′ · Jm (r′ ) = iω ρm (r′ ) .

Equations (4.75) and (4.76) form the basis for the integral equation for scattering and antenna
problems. To demonstrate this, let us consider a homogeneous medium with permittivity ε and
permeability µ . Also consider volumetric currents J(r) and Jm (r) as excitation. In the absence
of any objects in this homogeneous medium, these currents produce Ei (r) and Hi (r), and they
can be computed from Eqs. (4.75) and (4.76), respectively. In the presence of an object, these
fields generate a secondary set of fields that are referred to as the scattered fields. Hence the
total fields can be represented by

E(r) = Ei (r) + Es (r) (4.77a)

and
H(r) = Hi (r) + Hs (r) . (4.77b)

E(r) and H(r) are unknown, but assuming their values are known on the surface of the
scatterer, Ss , the field equivalence principle can be used by assuming zero fields inside the
scatterer and placing the following surface currents on the surface of the scatterer:

Js (r) = n̂ × H(r) , r ∈ Ss (4.78a)

and
Jms (r) = −n̂ × E(r) , r ∈ Ss . (4.78b)

Also, equivalent surface charges must be placed on the surface and are given by

ρs (r) = ε n̂ · E(r) and ρms (r) = µ n̂ · H(r) .

4-5 Integral Equation Formulation for Dielectric Scatterers 181

If these currents are known, they can be used in Eqs. (4.75) and (4.76) to find the scattered
fields. Hence the total fields are given by

E(r) =


Ei + iη k[n̂′ × H(r′ )] g(r, r′ ) − n̂′ · E(r′ ) ∇g(r, r′ ) + [E(r′ ) × n̂′ ] × ∇g(r, r′ ) ds′
Ss
(4.79a)
and
H(r) =
 
ik
Hi + [E(r ) × n̂ ] g(r, r ) − n̂ · H(r ) ∇g(r, r ) + [H(r ) × n̂ ] × ∇g(r, r ) ds′ .
′ ′ ′ ′ ′ ′ ′ ′ ′
Ss η
(4.79b)

This is valid when the observation point is outside the scatterer (outside Ss ). In cases where the
observation point is on the scatterer, the singularity of the Green’s function for the condition
r = r′ must be removed, similarly to what was shown in the previous section. In cases where
the observation point is inside the scatterer, the fields must be set to zero according to the
equivalence theorem. In general,


Ei (r) + −−− iη k[n̂′ × H(r′ )] g(r, r′ ) − n̂′ · E(r′ ) ∇g(r, r′ ) + [E(r′ ) × n̂′ ] × ∇g(r, r′ ) ds′


E(r) r outside Ss ,
= 12 E(r) r ∈ Ss , (4.80a)

0 r inside Ss ,
and
 
ik
Hi (r) + −−− [E(r ) × n̂ ] g(r, r ) − n̂ · H(r ) ∇g(r, r ) + [H(r ) × n̂ ] × ∇g(r, r ) ds′
′ ′ ′ ′ ′ ′ ′ ′ ′
η


H(r) r outside Ss ,
1
= 2 H(r) r ∈ Ss , (4.80b)


0 r inside Ss .

4-5 Integral Equation Formulation for Dielectric Scatterers

To examine how Eqs. (4.80a) and (4.80b) can be cast into integral equations, we consider a
dielectric object with permittivity ε2 and permeability µ2 placed in a homogeneous medium,
as shown in Fig. 4-5. As in the previous section, we consider known electric and magnetic
current sources J(r) and Jm (r) present in the background medium to produce the incident
fields Ei (r) and Hi (r). The fields outside the scatterer are denoted E1 (r) and H1 (r), and those
inside the scatterer are denoted E2 (r) and H2 (r). The expressions for E1 (r) and H1 (r) can
182 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

be obtained from Eqs. (4.80a) and (4.80b) by assigning subscript 1 to the fields, to the scalar
Green’s function, and to the medium parameters. Using a procedure similar to what led to
the derivation of Eqs. (4.80a) and (4.80b) for the exterior fields, the following equations are
obtained for E2 (r) and H2 (r):


−−− iη2 k2 [n̂′2 × H2 (r′ )] g2 (r, r′ ) − n̂′2 · E2 (r′ ) ∇g2 (r, r′ ) + [E2 (r′ ) × n̂′2 ] × ∇g2 (r, r′ ) ds′


0 r outside Ss ,
1
= 2 E2 (r) r ∈ Ss , (4.81a)

E (r)
2 r inside Ss ,
and
 
ik2
−−− [E2 (r ) × n̂2 ] g2 (r, r ) − n̂2 · H2 (r ) ∇g2 (r, r ) + [H2 (r ) × n̂2 ] × ∇g2 (r, r ) ds′
′ ′ ′ ′ ′ ′ ′ ′ ′
η2


0 r outside Ss ,
1
= 2 H2 (r) r ∈ Ss , (4.81b)

H (r)
2 r inside Ss .

Defining the tangential surface fields as unknown and applying the boundary conditions, we
have

n̂′1 × H1 (r′ ) = −n̂′2 × H2 (r′ ) = Js (r′ ) ,

E1 (r′ ) × n̂′1 = −E2 (r′ ) × n̂′2 = Jms (r′ ) ,
∇′s · Js (r′ )
ε1 n̂1 · E1 (r′ ) = −ε2 n̂′2 · E2 (r′ ) = ,
iω

ε
J
Jm
Ei(r)
Hi(r) E2
H2 Scatterer
r nˆ 2′
r′ ε
n̂′
E1(r′), H1(r′)

Figure 4-5: Geometry of the scattering problem for a dielectric object placed in a homogeneous
and unbounded medium illuminated by fields generated by J and Jm .
4-5 Integral Equation Formulation for Dielectric Scatterers 183

∇′s · Jms (r′ )

µ1 n̂1 · H1 (r′ ) = −µ2 n̂′2 · H2 (r′ ) = .
iω
Here ∇s · refers to a 2-D divergence operator acting on the surface of the scatterer. The electric-
field integral equations can be obtained by requiring the continuity of the tangential electric
fields at the boundary of the scatterer using Eqs. (4.80a) and (4.81a); that is,
1
− Jms (r)
2
 
iη1 ′
− n̂ × −−− iη1 k1 Js (r′ ) g1 (r, r′ ) + ∇s · Js (r′ ) ∇g1 (r, r′ ) + Jms (r′ ) × ∇g1 (r, r′ ) ds′
Ss k1
= n̂ × Ei(r) (4.82a)
and
1
− Jms (r)
2
 
′ ′ iη2 ′
+ n̂ × −−− iη2 k2 Js (r ) g2 (r, r ) + ∇s · Js (r ) ∇g2 (r, r ) + Jms (r ) × ∇g2 (r, r ) ds′
′ ′ ′ ′
Ss k2

=0. (4.82b)

Also, the magnetic-field integral equations can be obtained by enforcing the continuity of the
tangential magnetic fields at the boundary of the scatterer using Eqs. (4.80b) and (4.81b):

1 ik1
Js (r) − n̂ × −−− Jms (r′ ) g1 (r, r′ )
2 Ss η 1

i
+ ∇ · Jms (r ) ∇g1 (r, r ) − Js (r ) × ∇g1 (r, r ) ds′
′ ′ ′ ′ ′
k1 η1 s
= n̂ × Hi(r) (4.83a)
and

1 ik2
Js (r) + n̂ × −−− Jms (r′ ) g2 (r, r′ )
2 Ss η2

i
+ ∇ · Jms (r ) ∇g2 (r, r ) − Js (r ) × ∇g2 (r, r ) ds′
′ ′ ′ ′ ′
k2 η2 s
=0. (4.83b)

The magnetic-field integral equations (MFIE) given by Eq. (4.83) are duals of the electric-
field integral equations (EFIE). Hence, either EFIE or MFIE is sufficient to find the equivalent
surface currents.
The EFIE and MFIE take on a simple form when the scatterer is a metallic object. Noting
that n̂ × E = 0 and n̂ · H = 0 on metallic surfaces, as evident from Eq. (4.82a) by setting
184 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Jms = 0, the EFIE for metallic objects assumes the following form:
 
1 ′
′ ′
−iη1 k1 n̂ × −−− Js (r ) g(r, r ) + 2 ∇s · Js (r ) ∇g(r, r ) dS′ = n̂ × Ei(r) .
′ ′
(4.84)
Ss k1
(metallic object)

Also, using Eq. (4.83a) with Jms = 0, the following MFIE for metallic objects is obtained:

1
Js (r) + n̂ × −−− Js (r′ ) × ∇g1 (r, r′ ) dS′ = n̂ × Hi (r) . (4.85a)
2 Ss

(metallic object)

For metallic objects with no sharp edges, the MFIE formulation is preferred and provides a
more accurate numerical solution. For surfaces with large radii of curvature, the contribution
from the integral becomes negligible compared with n̂ × Hi (r) and can be ignored. This is
due to the fact that the dominant contribution of the integral comes from source points r′ that
are close to the observation point r. When r is close to r′ for smooth surfaces, ∇g(r, r′ ) is
almost tangential to the surface, and as a result Js (r′ ) × ∇g(r, r′ ) is almost perpendicular to
the surface and its cross product with n̂ is almost zero. Under such assumptions (large radii of
surface curvature),
Js (r) = 2n̂ × Hi (r) , (4.85b)
which is known as the physical-optics approximation.

4-6 Reciprocity Theorem

The notion of reciprocity was introduced in conjunction with the symmetry property of the
scalar Green’s function. In circuit theory, reciprocity manifests itself in the symmetry of the
impedance matrix of a passive linear network. Consider a two-port network, as shown in
Fig. 4-6. It can be easily shown that the short-circuited current at the second port (Ia ) excited
by a voltage source at the first terminal (Va ) and the short-circuited current at the first port (Ib )
excited by a voltage source at the second terminal (Vb ) satisfy the following relations:

IaVb = IbVa . (4.86)

The corresponding relationship in electromagnetics is provided by the Lorentz reciprocity

theorem. To find the reciprocity relation, let us consider two sets of excitations and their

Two- Two-
Va port Ia Ib port Vb
network network

Figure 4-6: The concept of reciprocity relationship for a two-port network.

4-6 Reciprocity Theorem 185

resulting fields in a bounded homogeneous medium specified by volume V and a closed

surface S. Suppose Ea and Ha are generated by electric and magnetic current sources Ja
and Jma and fields Eb and Hb are generated by current sources Jb and Jmb . According to
Maxwell’s equations,
(
∇ × Ea = iω µ Ha − Jma ,
(4.87a)
∇ × Ha = −iωε Ea + Ja ,
and
(
∇ × Eb = iω µ Hb − Jmb ,
(4.87b)
∇ × Hb = −iωε Eb + Jb .

Using the following vector identities

∇ · (Ea × Hb ) = Hb · ∇ × Ea − Ea · ∇ × Hb (4.88a)
and
∇ · (Eb × Ha ) = Ha · ∇ × Eb − Eb · ∇ × Ha, (4.88b)

and then subtracting Eq. (4.88b) from Eq. (4.88a) and using Eqs. (4.87a) and (4.87b), we have

∇ · (Ea × Hb − Eb × Ha ) = −(Ea · Jb − Ha · Jmb ) + (Eb · Ja − Hb · Jma ) . (4.89)

By applying the volume integral to both sides of Eq. (4.89), it can be shown that
$ $
(Ea × Hb − Eb × Ha )·ds = − (Ea ·Jb − Ha ·Jmb ) dv+ (Eb ·Ja − Hb ·Jma ) dv .
S V V
(4.90)
If we let S → ∞, the radiated fields in the far-field region become transverse electromagnetic
(TEM); that is, Ea,b ⊥ Ha,b ⊥ r̂, Ea = η Ha × r̂, and also Eb = η Hb × r̂. Hence,

(Ea × Hb − Eb × Ha ) · r̂ = Ea · (Hb × r̂) − Eb · (Ha × r̂)

1
= (Ea · Eb − Eb · Ea ) = 0 ,
η
which indicates that the left-hand side of Eq. (4.90) vanishes. In this case the reciprocity
relation takes the following form:
ha, bi = hb, ai , (4.91)
where $
ha, bi = (Ea · Jb − Ha · Jmb ) dv .
V
As an example, consider elementary sources

Ja = I0 ∆ ℓ a δ (r − r1 )
and
186 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Ea(r2)
Observation I0 Δlla
point Current
r2 source
r1

O
(a) Source at r1

Eb(r1)
I0 Δllb Observation
Current point
source r2
r1

O
(b) Source at r2

Figure 4-7: Geometry of two experiments with elementary sources to demonstrate the
reciprocity theorem in a homogeneous unbounded medium.

Jb = I0 ∆ ℓ b δ (r − r2 ) ,

as shown in Fig. 4-7, for which reciprocity dictates that

Ea (r2 ) · ∆ ℓ b = Eb (r1 ) · ∆ ℓ a .

That is, if the positions of the source and observation points are interchanged, the same field
is produced.
In situations where there are no sources within a finite and bounded medium, it follows
that
(Ea × Hb − Eb × Ha ) · ds = 0 , (4.92)
S
which is the statement of reciprocity for a source-free region (sources are outside S).

4-6.1 Generalized Form of the Reciprocity Theorem

The statement of reciprocity given by Eq. (4.91) is derived for a homogeneous medium.
Considering the symmetry properties discussed earlier for Green’s function in homogeneous
media, this statement of reciprocity may appear trivial. However, as will be shown in
this section, Eq. (4.91) can be extended to very complicated problems, such as a medium
composed of many homogeneous regions like the one shown in Fig. 4-8.
4-6 Reciprocity Theorem 187

μ1, ε1 μ2, ε2

μ4, ε4
μ3, ε3

Figure 4-8: An inhomogeneous medium composed of many homogeneous segments.

Without loss of generality, we consider a medium composed of only two homogeneous

media and we perform two experiments. In the first experiment (Fig. 4-9(a)) we assume that
sources Ja and Jma , which are confined within region 1, produce fields Ea and Ha . We consider
the observation point to be in region 2. In the second experiment (Fig. 4-9(b)) we place sources
Jb and Jmb only in region 2 and observe the fields Eb and Hb in the same region.
For the first experiment, Maxwell’s equations mandate

∇ × Ea = iω µ2 Ha (4.93a)
and
∇ × Ha = −iωε2 Ea , (4.93b)

because the sources are in region 1. For the second experiment

∇ × Eb = iω µ2 Hb − Jmb (4.94a)
and
∇ × Hb = −iωε2 Eb + Jb . (4.94b)

In this case, the sources are in region 2. As before, similar to Eq. (4.89), using Eqs. (4.93) and
(4.94) we can show that

∇ · (Ea × Hb − Eb × Ha ) = Ha · Jmb − Ea · Jb . (4.95)

Taking the volume integral of Eq. (4.94) in region 2 and using the divergence theorem leads
to $
(Ea × Hb − Eb × Ha ) · n̂2 ds = (Ha · Jmb − Ea · Jb ) dv , (4.96)
S∞ +S V2
where n̂2 is the unit normal to the closed-surface bounding region 2. The closed-surface
bounding region 2 in Fig. 4-9 is composed of S∞ and S, which encloses region 1. As shown
earlier, the contribution from S∞ vanishes.
188 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

n̂2
Ea,Ha μ1,ε1 S r

8
8
Ja,Jma n̂2
μ2,ε2
S

(a) Experiment 1

n̂

Jb,Jmb μ1,ε1

μ2,ε2 S
Eb,Hb

(b) Experiment 2

Figure 4-9: Fields and source points used in two experiments.

Next we consider Maxwell’s equations in region 1. For the first experiment,

(
∇ × Ea = iω µ1 Ha − Jma ,
(4.97)
∇ × Ha = −iωε1 Ea + Ja ,
4-7 Applications of the Reciprocity Theorem 189

for sources in region 1. For the second experiment when the sources are in region 2,
(
∇ × Eb = iω µ1 Hb ,
(4.98)
∇ × Hb = −iωε1 Eb .

Following a similar procedure, it can be shown that

$
(Ea × Hb − Eb × Ha ) · n̂1 ds = − (Hb · Jma − Eb · Ja ) dv . (4.99)
S V1

Now considering that −n̂2 = n̂1 = n̂ and for surface fields of region 2
(2) (2) (2) (2) (2) (2)
(Ea × Hb ) · n̂ = Hb · (n̂ × Ea ) = Hb · t̂ |n̂ × Ea | ,

where t̂ is a unit vector tangential to S given by

(2)
n̂ × Ea
t̂ = (2)
.
|n̂ × Ea |
Using the boundary conditions
(2) (1)
n̂ × Ea = n̂ × Ea
and
(2) (1)
Hb · t̂ = Hb · t̂

leads to the result

(2) (2) (2) (2) (1) (1) (1) (1)
(Ea × Hb − Eb × Ha ) · n̂2 ds = − (Ea × Hb − Eb × Ha ) · n̂1 ds .
S S

In view of Eqs. (4.96) and (4.99),

$ $
(Ha · Jmb − Ea · Jb ) dv = (Hb · Jma − Eb · Ja ) dv , (4.100)
V2 V1

which is the same statement as Eq. (4.91). Hence in general

ha, bi = hb, ai .

4-7 Applications of the Reciprocity Theorem

The reciprocity theorem has many applications in antenna theory and scattering problems. In
this section we examine a few of its important applications. For example, when the concept of
reciprocity was introduced in Section 4-6, we considered the relations between the voltages
and currents at the terminals of a two-port network, as shown in Fig. 4-6. Such relations
190 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

can also be proven to be true for the voltages and currents at the terminals of two antennas
in a communication link. That is, the environment around and in between two antennas
can be viewed as a passive circuit network. To show this, consider two metallic antennas
in a general isotropic medium at an arbitrary distance from each other. First, Antenna 1 is
excited by a voltage source Va and the second antenna is short-circuited. The short-circuit
current on the second antenna, denoted Ia , together with Antenna 1 generate fields Ea (r) and
Ha (r) everywhere in the medium. In a second experiment, the second antenna is excited by a
voltage source Vb at its port and the first antenna is short-circuited. The short-circuit current
on Antenna 1 is represented by Ib and the fields produced by the ensemble of antennas are
denoted Eb (r) and Hb (r). The configurations for these experiments are shown in Fig. 4-10.
In the application of reciprocity, we consider a composite surface that covers the surfaces of
the two antennas plus a large enclosing spherical surface that extends to infinity. As shown
before, the contribution from the spherical surface, which is in the far-field region of both
antennas, is zero. Also in this case the reciprocity condition for a source-free region given by
Eq. (4.92) must be used:

(Ea × Hb − Eb × Ha ) · ds = 0 , (4.101)
S1 +S2

where S1 and S2 are surfaces covering the surfaces of Antenna 1 and Antenna 2 and their
terminals, respectively, as shown in Fig. 4-10. Since n̂ × Ea (r) = n̂ × Eb (r) = 0 everywhere
on S1 and S2 except for the cylindrical surfaces around the antenna terminals, then Eq. (4.101)
simplifies to
(Ea × Hb − Eb × Ha ) · ds = 0 . (4.102)
Cyl1 +Cyl2

In the first experiment, the port of Antenna 2 is short-circuited and thus on cylinder 2
n̂ × Ea (r) = 0, and in the second experiment the port of Antenna 1 is short-circuited, which
enforces n̂ × Eb (r) = 0 on cylinder 1. Hence,

(Ea × Hb ) · ds = (Eb × Ha ) · ds . (4.103)
Cyl1 Cyl2

Near the terminals of the antennas on the cylindrical segments of S1 and S2 , for a very small
cylinder radius ρ0 according to Ampère’s law
Ib
Hb (ρ0 ) = φ
φ̂ (4.104a)
2πρ0
and
Ia
Ha (ρ0 ) = φ.
φ̂ (4.104b)
2πρ0
On the surface of cylinder 1,

(Ea × Hb ) · n̂ = [Hb (ρ0 ) φ̂ ρ] · Ea (ρ0 ) .

φ × ρ̂
4-7 Applications of the Reciprocity Theorem 191

n̂ × Ea = 0

n × Ea = 0
n̂
l ˆ = ρˆ
n Ia
Ia
Haϕ =
2πρ0
2πρ

n̂ × Eb = 0

n̂ × Eb = 0 l
ˆ = ρˆ
n
Ib
Hbϕ =
2πρ0

Figure 4-10: Configuration of two antennas used for transmission and reception. The reciprocity
theorem shows that the two ports of these antennas can be viewed as a two-port passive network.

Noting that Ea (ρ0 ) = ẑVa /∆ ℓ and Eb (ρ0 ) = ẑVb /∆ ℓ, and in view of Eq. (4.104a) and
Eq. (4.101), it follows that

(Ea × Hb ) · ds = −Va Ib . (4.105a)
Cyl1
192 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Similarly, using Eq. (4.104b), it can be shown that

(Eb × Ha ) · ds = −Vb Ia . (4.105b)
Cyl2

This proves the statement of reciprocity for a general two-port network, namely

Va Ib = Vb Ia . (4.106)

4-7.1 Equivalence of Antenna Patterns in Transmission and Reception

Equation (4.106) can be used to show that the radiation pattern of an antenna in a transmitting
mode is the same as its radiation pattern if the antenna were to be used as a receiver. The
radiation pattern of an antenna in transmitting mode is defined as the angular (θ , φ ) variation
of the normalized power density of the radiated fields in the far-field region at a fixed distance
from the antenna. On the other hand, the radiation pattern of the antenna in the receiving
mode is defined as the normalized angular variation of the power received by the antenna
if a transmitter at a fixed far-field distance and with fixed transmit power and polarization
configuration were to be moved along all possible directions relative to the receive antenna.
Figure 4-11 shows the configuration for radiation pattern measurements of an antenna in the
transmit and receive modes. Denoting the normalized radiation pattern in the transmitting
mode as Dt (θ , φ ) and that in the receiving mode as Dr (θ , φ ), we want to show that

Dt (θ , φ ) = Dr (θ , φ ) . (4.107)

Consider the geometry of the antenna under test centered at the origin of a spherical coordinate
system. This antenna is used as a transmitter and a short dipole placed in the far-field region
of the antenna is used as a receiver to measure the received power. In transmission mode, a
voltage Va is at the terminals of the antenna. This voltage induces a short-circuited current
Ia (θ , φ ) at the terminals of the dipole. The received power at each location of the dipole is
proportional to Ia2 (θ , φ ). In another experiment designed to measure the radiation pattern of
the antenna in the receiving mode, the dipole is attached to a source and the power received by
the antenna is measured as a function of the dipole position. In this case, a fixed Vb is attached
to the terminals of the dipole and a short-circuited current Ib (θ , φ ) flows through the terminals
of the antenna. The received power is proportional to Ib2 (θ , φ ). According to Eq. (4.106),

Va (θ , φ ) Vb (θ , φ )
= . (4.108)
Ia Ib
Note that Va and Vb are constant and can be made equal. Also noting that Dt (θ , φ ) is
proportional to Ia2 (θ , φ ) and Dr (θ , φ ) is proportional to Ib2 (θ , φ ), in view of Eq. (4.108) we
conclude that Eq. (4.107) holds.
4-7 Applications of the Reciprocity Theorem 193

b a Small dipole

a b

Antenna under test

Figure 4-11: Configuration of antenna pattern measurement. In the transmit mode a simple
antenna like a dipole is placed in the far-field of the antenna under test along the (θ , φ )-direction,
and then the received power is measured by the antenna under test as the dipole is moved over
θ –φ space. The same configuration can be used to measure the radiation pattern of the antenna
under test in receiving mode wherein the antenna under test becomes the transmitter and the
dipole becomes the receiver. In these experiments the dipole orientation is maintained in θ̂ θ or φ̂
φ,
as the dipole is moved.

Example 4-1: Application of the Reciprocity Theorem: Show That Tangential

Electric Currents near Metallic Surfaces of Arbitrary Shapes
Cannot Radiate Electromagnetic Fields

Using the reciprocity theorem, show that the field radiated by an electric current source near
and tangential to an arbitrary metallic surface is zero.

Solution: Consider a metallic surface S with a dipole of height δ h above it and located at rs .
Also consider an arbitrary observation point outside S, as shown in Fig. 4-12. Let us denote
the radiated field from the Hertzian dipole (I0 ∆ ℓ t δ (r − rs )) in the presence of the surface at
the observation point ro as E1 (ro ). Our objective is to show that

lim E1 (ro ) = 0 .
δ h→0
194 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

We showed previously that if surface S is an infinite planar surface, image theory can be
used. The image of a horizontal Hertzian dipole is a horizontal Hertzian dipole with the
opposite direction of current flow and located at the mirror image point. As δ h → 0 these
opposite currents will totally cancel each other and generate a null field everywhere. This
result cannot be directly extended to arbitrary surfaces as we do not know the image of a
tangential Hertzian dipole on arbitrary surfaces. To prove a tangential Hertzian dipole right
above any metallic surface cannot produce any fields outside the surface, we consider an
experiment wherein we place the Hertzian dipole at the observation point ro with an arbitrary
orientation (I0 ∆ ℓ a δ (r − r0 )) and then evaluate the field at the original source point rs , which
we denote as E2 (rs ). According to the reciprocity theorem,

E2 (rs ) · I0 ∆ ℓ t = E1 (ro ) · I0 ∆ ℓ a .

But on surface S
lim E2 (rs ) · ∆ ℓ t = 0 ,
δ h→0
since according to boundary conditions, the tangential electric field on the surface of a metallic
body must vanish. Hence, we conclude that the component of E1 (ro ) along any arbitrary
direction is zero, or simply E1 (ro ) = 0.
The result of this example can easily be extended to other similar problems. For example,
following the same procedure it can be shown that a Hertzian magnetic dipole placed on and
normal to a metallic surface of arbitrary shape cannot radiate electromagnetic fields. This
result and that of Example 4-1 can be used in conjunction with the duality principle to prove
that tangential magnetic Hertzian dipoles and normal electric Hertzian dipoles placed on a
perfect magnetic conductor cannot radiate electromagnetic waves regardless of the shape of
the magnetic conductor.

O
O
ro S
rs S

E1(ro)
t (r − ro)
t (r − rs) E2(rs)

Figure 4-12: A short Hertzian dipole radiating near a metallic surface S. Also shown is a
reciprocal configuration in which the source is moved to the observation position and the field
is evaluated at the source position of the original problem.
4-7 Applications of the Reciprocity Theorem 195

4-7.2 Effective Area for Receiving Antennas

The effective area Aeff of an antenna is defined as a fictitious area around the antenna that
allows for the antenna to collect the local incident wave power density and deliver it to
an impedance-matched transmission line connected to the antenna terminals. Based on this
definition,
Prec = Aeff (θ , φ ) Sinc , (4.109)
where Sinc is the time-average power density of the incident wave. In general, this effective
area may be a function of the direction of the incident wave, (θ , φ ). Our goal here is to relate
Aeff (θ , φ ) to the gain function, G(θ , φ ), of the antenna in the transmitting mode of operation.
To establish this relation we will make use of the reciprocity theorem. Here we assume that the
antenna is lossless and impedance-matched. Let us consider two experiments where initially
the antenna is used as a transmitter, generating a field at an observation point r = (r, θ , φ )
in the far-filed region. We denote this field as Ea (r). The voltage at the antenna terminals is
denoted Va . In the next experiment we use the same antenna as a receiver and measure the
short-circuited current Ib (θ , φ ) when illuminated by an infinitesimal current filament I0 ∆ ℓ
oriented to be parallel to Ea (r). We emphasize here that the short-circuited current is a
function of the infinitesimal current location. The current filament generates an electric field
in the vicinity of the antenna, which is represented by Eb , as shown in Fig. 4-13. Now we
apply the reciprocity theorem by considering a composite surface that includes the antenna
(surface S) and a spherical surface that extends to infinity (S∞ ). As before, the contribution of
S∞ vanishes due to the fact that fields far away from the antenna and the infinitesimal current
filament are TEM. According to Eq. (4.90), we have
$
(Ea × Hb − Eb × Ha ) · ds = − (Ea · Jb ) dv . (4.110)
S V

Since Eb × n̂ = 0 on S, Eq. (4.110) simplifies to

Va Ib (θ , φ )
× × 2πρ0 ∆ g = I0 ∆ ℓ Ea (θ , φ ) ,
∆g 2πρ0

noting that Ea = Va /∆ g, Hb = Ib /(2πρ ), the surface of the cylinder is 2πρ0 ∆ g, and that
Jb = I0 ∆ ℓ δ (r − r′ ),
Va Ib (θ , φ ) = I0 ∆ ℓEa (θ , φ ) . (4.111)
Assuming that the antenna is impedance-matched under both transmission and reception,
and denoting the radiation resistance of the antenna by RA , the radiated power can be
calculated from
|Va |2
Prad = .
2RA
Also, the power density at the observation point can be calculated in two ways, once based on
the radiated power and directivity and another based on the field power density; hence

|Va |2 |Ea (θ , φ )|2

Srad (θ , φ ) = G(θ , φ ) = .
2RA (4π r2 ) 2η
196 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Ea(r)

n̂ × Ea = 0

r(r,θ,ϕ)

E g

Expanded view of
the antenna terminal

(a) Antenna in transmission mode

Δl
Infinitesimal
n̂ × Eb = 0 current
filament
r(r,θ,ϕ)

n̂ × Eb = 0
nˆ = ρˆ
Ib
Hbϕ =
2πρ0
Expanded view of
the antenna terminal

(b) Antenna in reception mode

Figure 4-13: Configuration of a general antenna in transmission and reception modes. In the
transmission mode, a current I1 is excited at the antenna terminal and field Ea (r) is generated at
the observation point r = (r, θ , φ ). The same antenna is then illuminated by the fields generated
by current filament I0 ∆ ℓ parallel to Ea (r).
4-7 Applications of the Reciprocity Theorem 197

This provides the expression for the electric field in terms of the voltage

η |Va |2
|Ea (θ , φ )|2 = G(θ , φ ) . (4.112)
RA (4π r2 )
In the reception mode, the power density generated by the current filament is given by

η k2 |I0 ∆ ℓ|2
Sdipole = .
2(4π r)2
The received antenna power according to Eq. (4.109) is

η k2 |I0 ∆ ℓ|2
Prec = Aeff (θ , φ ) . (4.113)
2(4π r)2
This is the power delivered to a matched load connected to the receiving antenna. Referring
to the match-loaded antenna equivalent circuit shown in Fig. 4-14, the power delivered to the
matched load can be calculated from
|Voc (θ , φ )|2
Prec = . (4.114)
8RA
This same equivalent circuit can be used to predict the short-circuited current to be

Voc (θ , φ )
Ib (θ , φ ) = .
RA
Therefore,
RA |Ib (θ , φ )|2
Prec = . (4.115)
8
Substituting Eq. (4.115) into Eq. (4.113),

η k2 |I0 ∆ ℓ|2
|Ib (θ , φ )|2 = 4Aeff (θ , φ ) . (4.116)
RA (4π r)2
Using Eq. (4.111), we also have

|I0 ∆ ℓ|2 |Ea (θ , φ )|2

|Ib (θ , φ )|2 = . (4.117)
|Va |2
Now, substituting Eq. (4.116) and Eq. (4.112) into Eq. (4.117) provides a relationship between
G(θ , φ ) and Aeff (θ , φ ):

η k2 |I0 ∆ ℓ|2 |I0 ∆ ℓ|2 η |Va |2

Aeff (θ , φ ) = × G(θ , φ ) ,
RA (4π r)2 |Va |2 RA (4π r2 )
198 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

RA RA

Voc RA Voc

Figure 4-14: The antenna equivalent circuit of an antenna in the receiving mode when connected
to a matched load RA . Also shown on the right is the equivalent circuit of the antenna in receiving
mode when short-circuited at its terminals.

or
k2
Aeff (θ , φ ) = G(θ , φ ) .
π
Since k = 2λπ , a more familiar relation between antenna directivity and antenna effective area
is obtained:
λ2
Aeff (θ , φ ) = G(θ , φ ) . (4.118)
4π

4-7.3 Scattering Interaction between Two Adjacent Objects

In this section we examine the application of the reciprocity theorem to understand the
mechanics of electromagnetic scattering by objects of various shapes and electrical properties.
Oftentimes, scattering problems are solved for objects in isolation. However, when two or
more such objects are present simultaneously, the solution for the combination of objects
becomes rather complicated and computationally inefficient. In situations where the objects
are sufficiently far apart but can be in the near field of each other, the scattering solution can be
obtained in an iterative manner. To clarify this, let us consider two objects in a homogeneous
and unbounded background medium and illuminated by an incident wave characterized by
fields Ei (r) and Hi (r), as shown in Fig. 4-15. Let us denote the field scattered by object 1
in isolation as Es1 (r) and that by object 2 in isolation as Es2 (r). The scattered field Es2 (r)
can then be considered as a source illuminating object 1, and it causes a secondary scattered
field denoted Es12 (r) in Fig. 4-15. Conversely, the field scattered by object 1 (Es1 (r)) can be
considered a source of illumination for object 2, and it causes a secondary scattered field
denoted Es21 (r). This process continues, and higher-order scattered fields such as E121 (r),
E212 (r), and so on are generated. The fields in each scattering iteration become weakened due
to propagation path loss, and thus the multiple scattering process converges very quickly. In
many practical applications, scattering up to the second order is sufficient to capture important
effects such as depolarization or the dominant scattering mechanism. In what follows, the
4-7 Applications of the Reciprocity Theorem 199

s
Ei(r) E1
Hi(r) s
E12
Plane wave

s
E2

Figure 4-15: Adjacent objects in the near field of each other and illuminated by a plane wave.
In addition to the direct scattering components Es1 and Es2 , the close proximity of the two objects
leads to many multiple scattering components, such as Es12 .

reciprocity theorem is used as a way to analytically obtain the far-field expansion that includes
multiple scattering up to second order for two adjacent objects, based solely on the knowledge
of the first-order scattering of the individual objects.

4-7.4 Perfectly Conducting Objects

In this section we consider a simpler problem wherein the two objects are made from
a perfect conducting material. First, consider object 1 in isolation and illuminated by an
incident wave. Let us denote the surface current density on object 1 as J1 (r). To compute the
secondary scattered field from object 2, we use J1 (r) as the source of excitation for object 2.
This scenario is depicted in Fig. 4-16(b). Next we consider a scenario where object 1 or,
equivalently, surface current distribution J1 (r) is absent, and an infinitesimal current source
given by
Jd (r) = p̂ δ (r − r0 ) (4.119)
is placed at an observation point r0 in the far-field region of the objects. The fields produced
by the infinitesimal current source in the presence of object 2 are denoted Ed2 (r) and Hd2 (r),
as shown in Fig. 4-17. According to the reciprocity theorem (see Eq. (4.90)),
 + 

+
E1 (r) × Hd2 (r) − Ed2 (r) × H1 (r) ·ds = J1 (r)·Ed2 (r) ds− p̂ ·E+
1 (r0 ) . (4.120)
S2 S1

Here the field quantities with superscript “+” (E+ +

1 (r), H1 (r)) are the direct fields radiated by
J1 (r) (E1 (r), H1 (r)) plus the secondary scattered fields (Es21 (r), Hs21 (r)). According to the
s s

boundary condition on metallic surfaces, n̂ × E+ 1 = n̂ × Ed2 = 0, and therefore

(E+ +
1 × Hd2 ) · n̂ = (n̂ × E1 ) · Hd2 = 0 , r ∈ S2 (4.121a)
and
(Ed2 × H+ +
1 ) · n̂ = (n̂ × Ed2 ) · H1 = 0 , r ∈ S2 . (4.121b)
200 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

ε
(a) s
Ei(r) E1
J1(r)
Hi(r)

ε s
E1
J1(r)
(b) s
E21
Observation
s
point E2

Figure 4-16: (a) The incident wave illuminating object 1 in isolation induces a surface current
J1 (r) on the PEC object, and (b) this induced current becomes a source that generates the
secondary scattered fields (Es21 (r), Hs21 (r)) from object 2 at the observation point.

Object 1
ε
Ed2(r), Hd2(r)

Observation
point Object 2

p̂ (r − ro)

Figure 4-17: An infinitesimal current source with orientation p̂ placed at the observation point
r0 is illuminating object 2. The observation point is in the far-field region of the objects, and its
field can be locally approximated by a plane wave. Ed2 (r) and Hd2 (r) represent the sum of the
direct field from the source and the near field scattered by object 2.

As a result, the left-hand side of Eq. (4.120) vanishes and we have

+
p̂ · E1 (r0 ) = J1 (r) · Ed2 (r) ds . (4.122)
S1
4-7 Applications of the Reciprocity Theorem 201

As mentioned previously, E+ 1 (r0 ) represents the direct backscatter from object 1 (radiation
from J1 (r)) plus the secondary scattered field from object 2 (Es21 (r0 )); that is,

E+ s s
1 (r0 ) = E1 (r0 ) + E21 (r0 ) . (4.123)

Following a similar procedure, we can compute the first-order scattered field from object 2
plus the second-order scattered field generated by object 1 when illuminated by the first-order
scattered field of object 2:

+
p̂ · E2 (r0 ) = J2 (r) · Ed1 (r) ds , (4.124)
S2

where Ed1 (r) is the incident field from the dipole plus the scattered field from object 1
calculated on the surface of object 2. Thus E+ +
1 (r0 ) + E2 (r0 ) represents the total first-order
plus the second-order scattered fields from both objects at the observation point r0 . It must be
emphasized that the scattered field up to second-order scattering interaction is derived from
only the knowledge of the surface currents and the near-field scattered electric field of the
isolated objects.

4-7.5 Dielectric Objects

In this section we consider a scattering problem composed of two adjacent dielectric objects.
For simplicity let us assume the dielectric objects to be nonmagnetic (µ = µ0 and ε = ε0 εi ,
i = 1, 2). The incident wave, in the absence of object 2, creates a polarization current J1 (r)
within object 1. Recall that when the notion of polarization current is used, we are assuming
the entire medium is homogeneous and unbounded with permittivity ε0 and permeability µ0 .
This polarization current in the absence of the incident field and presence of object 2 induces
a polarization current J12 (r) in object 2. The total fields in this configuration, which represent
the first-order scattered field from object 1 and the second-order scattered field from object 2,
are denoted E+ +
1 (r) and H1 (r). As before, we conduct another experiment in which we place a
filament current at the observation point r0 in the far-field region of the scatterer and keep only
object 2. This excitation induces polarization current J2d (r) in object 2. The corresponding
electric and magnetic fields for this scenario are denoted Ed2 (r) and Hd2 (r). The expressions
for the polarization currents in object 2 are given by

J12 (r) = −ik0Y0 (ε2 − 1) E+

1 (r) , r ∈ V2 , (4.125a)
and
Jd2 (r) = −ik0Y0 (ε2 − 1) Ed2 (r) , r ∈ V2 . (4.125b)

Since the medium is homogeneous and unbounded (scatterers are replaced with polarization
currents), the reciprocity theorem mandates that
$ $ $
J1 (r) · Ed2 (r) dv + J12 (r) · Ed2 (r) dv = Jd2 (r) · E+ +
1 (r) dv + p̂ · E1 (r0 ) .
V1 V2 V2
(4.126)
202 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Using Eqs. (4.125a) and (4.125b) in Eq. (4.126), it can be shown that
$ $
J12 (r) · Ed2 (r) dv = Jd2 (r) · E+
1 (r) dv , (4.127)
V2 V2

and therefore $
p̂ · E+
1 (r0 ) = J1 (r) · Ed2 (r) dv . (4.128)
V1
Following a similar procedure by interchanging the roles of objects 1 and 2, it can be easily
shown that $
p̂ · E+
2 (r0 ) = J2 (r) · Ed1 (r) dv , (4.129)
V2

where E+ 2 (r0 ) is the first-order scattered field from object 2 plus the second-order scattered
field from object 1. Also J2 (r) represents the polarization current induced in object 2 in the
absence of object 1, and Ed1 (r) is the near-field first-order scattered field from object 1 due
to the infinitesimal current at the observation point. As before, E+ +
1 (r0 ) + E2 (r0 ) is the total
scattered field up to second-order mutual interaction.

Example 4-2: Radar Cross Section of a Dihedral Using the Reciprocity Theorem

Following the approach outlined for calculating scattering by two adjacent objects up to
second-order using the reciprocity theorem, calculate the scattered field in the backscatter
direction from two electrically large perpendicular metallic plates, known as a dihedral corner
reflector, as shown in Fig. 4-18. Assume the dihedral is illuminated by a plane wave given by
Ei (r) = E0 ŷ eik0 x . Also calculate the radar cross section of the dihedral corner reflector.

Solution: To find the scattered field up to second-order, we need to determine the surface
current induced on each plate in isolation when illuminated by the incident wave. Since the
plates are electrically large and planar (radii of curvature are infinite), the surface currents can
be computed from the physical-optics approximation:

J(r) = 2n̂ × Hi(r) . (4.130)

The expression for the incident magnetic field is given by

E0 ik0 x
Hi (r) = ẑ e . (4.131)
η0
Let us first consider plate 1, whose unit normal vector is given by
√ √
2 2
n̂1 = − x̂ − ŷ . (4.132)
2 2
4-7 Applications of the Reciprocity Theorem 203

W y

Ei(r)
Hi(r)
x
H

Figure 4-18: A dihedral corner reflector composed of two perpendicular planar metallic plates
with dimensions W × H, is illuminated by a plane wave. The edge of the dihedral coincides with
the z-axis. The goal is to compute the scattered field in the backscatter direction (−x-axis).

Also, on this object y = −x. Hence the surface current is given by

√ √ E0 −ik0 y
J1 (r) = ( 2 ŷ − 2 x̂) e . (4.133)
η0
We also need to find the electric field generated by an infinitesimal dipole at the observation
point r0 orientated along the y-axis (Jd (r) = ŷ δ (r − r0 )).
The far field produced by this small dipole can be obtained from Eq. (3.127b) and is given
by
eik0 r0
Ed (r) = −ik0 η0 [ŷ × (−r̂0 )] × (−r̂0 )e−ik0 r·r̂0 . (4.134)
4π r0
Since r̂0 = −x̂,
eik0 r0 ik0 x
Ed (r) = ik0 η0 e ŷ . (4.135)
4π r0
To find the scattered field from plate 2 illuminated by Ed (r), we use the geometric-optics
approximation. That is, the scattered field is simply the reflected field, given by Eq. (4.135),
on the metallic plate:
eik0 r0 ik0 y
E2dr (r) = −ik0 η0 e x̂ . (4.136)
4π r0
The total field from the dipole in the presence of plate 2 on plate 1 is given by

Ed2 (r) = Ed (r) + E2dr (r)

eik0 r0 ik0 x
= ik0 η0 (e ŷ − eik0 y x̂) . (4.137)
4π r0
204 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Recalling that x = −y for every point on plate 1, we have

eik0 r0 −ik0 y
Ed2 (r) = ik0 η0 (e ŷ − eik0 y x̂) . (4.138)
4π r0
Using Eq. (4.122), the first-order scattered field from plate 1 and the second-order scattered
field from plate 2 are related by
Z HZ √
2 W /2
ŷ · E+
1 (r0 ) = J1 (r) · Ed2 (r) dz dy
0 0
Z Z √
H/2 2 W /2
eik0 r0
= ik0 E0 2 (e−i2k0 y + 1) dy dz . (4.139)
4π r0 −H/2 0

After integration the scattered field given by Eq. (4.139) can be written as
√ eik0 r0 h √  √ i
+
E1y (r0 ) = ik0 E0 2 HW 1 + e−ik0 ( 2/2)W sinc k0 ( 2/2)W . (4.140)
4π r0
In a similar manner, we can find the current induced on plate 2 in the absence of plate 1 using
the physical-optics approximation, and noting that y = x on plate 2,
√ √ E0 ik0 y
J2 (r) = ( 2 x̂ + 2 ŷ) e . (4.141)
η0
The scattered field from plate 1 when illuminated by the fields of the infinitesimal dipole is
found to be
eik0 r0 −ik0 y
E1dr (r) = ik0 η0 e x̂ . (4.142)
4π r0
Hence the total field from the dipole in the presence of plate 1 on plate 2 is given by

eik0 r0 ik0 y
Ed1 (r) = ik0 η0 (e ŷ + e−ik0 y x̂) . (4.143)
4π r0
Using Eqs. (4.141) and (4.143) in Eq. (4.124), we find that
Z H/2 Z 0
ŷ · E+
2 (r0 ) = √ J2 (r) · Ed1 (r) dy dz
−H/2 −( 2/2)W
Z H/2 Z 0
eik0 r0 i2k0 y
= ik0 E0 2 √ [e + 1] dy dz .
4π r0 −H/2 −( 2/2)W

After performing the integration, we have

+
√ eik0 r0 √ √
E2y (r0 ) = ik0 E0 2 HW [1 + e−ik0W 2/2 sinc(k0W 2/2)].
4π r0
4-8 Babinet’s Principle 205

Hence the total scattered field up to second-order is

√ √ √ eik0 r0
Eys (r0 ) = 2 2 iAk0 E0 [1 + e−ik0W 2/2 sinc(k0W 2/2)] , (4.144)
4π r0
where A = W H is the area of each panel. Note that since W ≫ λ , it follows that
√
sinc(k0W 2/2) ≪ 1

and that
√ eik0 r0
Eys = 2 2 iAk0 E0 . (4.145)
4π r0
The radar cross section of a target is defined as

|E s |2
σ = lim 4π r2 .
r→∞ |E0 |2
Substituting Eq. (4.145) into Eq. (4.146), the backscatter radar cross section (RCS) of the
dihedral is given by
A2
σdihedral = 8π , (dihedral) (4.146)
λ02

where we have used k0 = 2π /λ0 .

4-8 Babinet’s Principle

In scattering and radiation problems that include planar metallic screens containing apertures
(slots), Babinet’s principle may prove useful in establishing a relationship to a dual scenario
in which magnetic or electric walls with appropriate sources are used in lieu of the slots. One
very useful application of Babinet’s principle pertains to establishing the relationship between
the input impedances of planar complementary antennas. Here, the term “complementary”
refers to another antenna structure in which those parts of the antenna made of metallic
surfaces are replaced with apertures, and vice versa. There are a number of steps that we need
to establish before we can show the relation between a planar antenna and its complementary
structure.

Lemma 1: Prove that the tangential magnetic field generated by a planar electric current sheet
vanishes in the plane that contains the electric current sheet.

Proof: To prove this lemma, let us consider an arbitrary planar electric current sheet J placed
in an arbitrary plane P, as shown in Fig. 4-19. To show that Ht = 0 in P, first we evaluate the
magnetic vector potential generated by the current distribution J, which is given by
"
A(r) = µ J(r′ ) g(r, r′ ) ds′ , (4.147)
206 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Ht = 0

Figure 4-19: An arbitrary planar surface current distribution contained in plane P.

where in free space

′
′ 1 eik|r−r |
g(r, r ) = .
4π |r − r′ |
The magnetic field can be obtained directly from Eq. (4.147) using
"
1
H(r) = ∇ × A = ∇ × [J(r′ ) g(r, r′ )] ds′ . (4.148)
µ
Noting that the differentiations in the curl operator are with respect to the unprimed coordinate
system
∇ × [J(r′ ) g(r, r′ )] = ∇g(r, r′ ) × J(r′ ) .
On the other hand,
  ′
′ 1 eik|r−r |
∇g(r, r ) = ik − (r − r′ ) . (4.149)
|r − r′ | |r − r′ |2

Evaluating H at a point in plane P guarantees that the vector (r − r′ ) lies in that plane. Also,
noting that the current vector J(r′ ) is in the P plane, it becomes clear that the integrand of
Eq. (4.148) is a vector perpendicular to P, and therefore

Ht (r) = 0, ∀r ∈ P .

Lemma 2: Prove that the tangential electric field generated by a planar magnetic current sheet
placed in an arbitrary plane P vanishes everywhere on P.
Proof: The proof of this lemma can be deduced directly from Lemma 1 by applying the
duality relations. These lemmas can also be proven using reverse application of image theory,
as described in Problems 2.10 and 2.11 of Chapter 2.
4-8 Babinet’s Principle 207

4-8.1 Complementary Theorem

Consider a planar metallic sheet consisting of a number of apertures (slots) of arbitrary shapes
excited by electric current sources all placed on one side of the metallic sheet, as shown in
Fig. 4-20(a). Now consider a second scenario in which identical sources are used as in the
first scenario, but the metallic sheet is removed and all apertures are replaced with perfect
magnetic conductor sheets, as shown in Fig. 4-20(b). Denote the fields of the first scenario by
E1 and H1 and those of the second scenario by E2 and H2 on the opposite half-space where
the sources are located. The complementary theorem states that

E1 + E2 = Ei (4.150a)
and
H1 + H2 = Hi , (4.150b)

in the source-free region where Ei and Hi are the field quantities generated by the same
source in the absence of all scatterers (electric and magnetic walls). Application of boundary
conditions leads to

n̂ × E1 = 0 on S1 (in Fig. 4-20(a)) (4.151a)

and
n̂ × H2 = 0 on S2 (in Fig. 4-20(a)). (4.151b)

The electric current source J on the left side of the metallic sheet induces a surface current
on the metallic surface. This current according to Lemma 1 does not produce any tangential
scattered magnetic fields across the apertures (on S2 ). Hence the total tangential magnetic
fields over the apertures are just the incident tangential magnetic fields produced by the source
in the absence of the electric conductor; that is,

n̂ × H1 = n̂ × Hi on S2 . (4.152)

Similarly, for the second scenario the electric current source J induces a magnetic current over
the magnetic walls that do not produce a tangential scattered electric field on S1 . Hence

n̂ × E2 = n̂ × Ei on S1 . (4.153)

Adding Eqs. (4.153) and (4.151a) yields

n̂ × (E1 + E2 ) = n̂ × Ei on S1 ,

and adding Eqs. (4.151b) and (4.152) yields

n̂ × (H1 + H2 ) = n̂ × Hi on S2 .

Since S1 extends to infinity, S1 + S2 in addition to a hemisphere at infinity can be considered

a closed surface for the source-free region. Since (E1 + E2 ) and (H1 + H2 ) satisfy Maxwell’s
equations and since we know the tangential electric field over a portion of the closed surface
208 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

PEC
J
Arbitrary-
shaped
apertures
(a)
E1
Arbitrary electric S1 H1
source
Metallic sheet
extended to
infinity

J
Apertures
S2
replaced E2 = E1 + E2s
(b) with perfect H2 = H1 + H2s
E2
magnetic
S2 H2
conductors

J
Ei Ei = E1 + E2
(c) Hi Hi = H1 + H2

Figure 4-20: Three configurations: (a) a planar metallic sheet (which acts like a perfect electric
conductor (PEC)) with arbitrary-shaped apertures excited by electric sources; (b) PEC is removed
and the apertures are replaced with magnetic conductor sheets; and (c) the source with no
scatterers present.

and the tangential magnetic field over the complementary portion of the surface, according to
the uniqueness theorem the fields in the source-free region are unique. However, noting that
the same tangential fields over S1 + S2 exist for the third scenario (Fig. 4-20(c)), where there
are no scatterers, then Eqs. (4.150a) and (4.150b) hold everywhere in the source-free region.

4-8.2 The Babinet Theorem

Consider a planar metallic sheet of infinite extent and a number of apertures of arbitrary
shape excited by electric current sources all on one side of the metallic sheet. The electric
and magnetic fields in the source-free region in this case will be denoted by E1 and H1 , as
shown in Fig. 4-21(a). In a second experiment, we consider instead the dual of electric current
sources as the source of the electromagnetic wave excitation for the metallic sheets instead of
4-8 Babinet’s Principle 209

PEC
J

E1

Arbitrary
electric current S1
H1

8
(a) Original configuration

ε
μ Jm PEC
S2
Dual magnetic Ems + Emi
current PEC
Hms + Hmi
S2

(b) Dual configuration

Figure 4-21: Two complementary configurations: (a) a planar metallic sheet with arbitrary-
shaped apertures excited by an electric current, and (b) the complementary configuration wherein
the apertures are replaced with PECs and excited by the dual magnetic current source.

the apertures in the first experiment, as shown in Fig. 4-21(b). Denoting the scattered fields
from these metallic sheets as Em m
s and Hs , the Babinet theorem states that
r
µ m
E1 = − H (4.154a)
ε s
and
r
ε m
H1 = E . (4.154b)
µ s
To prove Eqs. (4.154a) and (4.154b), we apply the complementary theorem to the first
experiment. That is, we maintain the same electric current source and replace the apertures
with perfect magnetic conductor sheets. This would produce the same electric and magnetic
fields in the right-hand side (source-free region) as shown in Fig. 4-20(b). Applying the duality
210 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

p
principle to this problem by simply replacing J with ε /µ Jm and the perfect magnetic
conductors with perfect electric conductors, we obtain the configuration of Fig. 4-21(b).
Under the duality relations, E2 and H2 of Fig. 4-20(b) must transform to
r
µ
E2 → (Hm m
s + Hi ) (4.155a)
ε
and
r
ε
H2 → − (Em + Emi ), (4.155b)
µ s
where Hm m m m
s + Hi and Es + Ei are the total magnetic and electric fields of the configuration
in Fig. 4-21(b), decomposed into the incident (in the absence of the conductors) and scattered
fields. The superscript “m” denotes that the fields are produced by the dual magnetic sources.
Substituting Eqs. (4.155a) and (4.155b) into Eqs. (4.150a) and (4.150b) and noting the duality
relations between the incident fields,
r
µ m
Ei → H (4.156a)
ε i
and
r
ε m
Hi → − E , (4.156b)
µ i
it becomes obvious that
r
µ m
E1 = − H (4.157a)
ε s
and
r
ε m
H1 = E . (4.157b)
µ s

4-8.3 Complementary Antennas

The Babinet principle provides a useful tool for understanding the concept and application of
complementary antennas. Babinet’s principle can be used in connection with planar metallic
antennas. To make the problem clear, let’s consider a simple antenna structure such as a dipole
antenna composed of two metallic strips fed by two thin wires (coplanar strips), as shown in
Fig. 4-22(a). Such an antenna is similar to wire dipoles in that it radiates most effectively when
the overall length of the dipole is about λ /2. Applying Babinet’s principle, a complementary
structure is formed as shown in Fig. 4-22(b). Here the areas around the metallic strips of the
dipole and its feed (aperture) are replaced with a metallic sheet, and the dipole strip and its
feed are left open as an aperture (slot). The antenna so formed is known as the slot antenna
and the feed is known as a coplanar waveguide (CPW).
Using Babinet’s principle, we first establish the relation between the input impedances of
these complementary antennas. At the terminal of the wire antenna, the voltage and current
4-8 Babinet’s Principle 211

a b

c
d
I I

J
I I

J Feed point

(a) Original configuration

a b

c
d
I I

ε
μ Jm
E E
Feed point
ε
μ Jm

(b) Complementary configuration

Figure 4-22: Two complementary configurations: (a) a planar dipole antenna composed of
metallic strips fed by a coplanar strip transmission line, and (b) a complementary antenna formed
from part (a) by replacing metallic sections with slots and vice versa.

can be obtained from

Z b
Vd = E1 · dℓℓ (4.158a)
a
and
Z d
Id = −2 H1 · dℓℓ . (4.158b)
c

In Eq. (4.158b) the symmetry of the problem (with respect to the plane containing the dipole)
212 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

is responsible for the factor of 2 in front of the integral. The current and voltage at the terminal
are produced by an elementary electric current source at the feed point connected to the
terminals of the coplanar strips. For the complementary problem, the feed is an elementary
magnetic current source placed adjacent to the center conductor of the CPW line, exciting two
opposite magnetic currents in the gaps of the CPW line. These magnetic currents excite the
slot antenna by producing a voltage and current at the terminals of the slot antennas, given by
Z b
Is = 2 Hm · dℓℓ (4.159a)
a
and
Z d
Vs = Em · dℓℓ . (4.159b)
c

Again, the symmetry of the problem explains the factor of 2 in front of Eq. (4.159a).
According to Eqs. (4.157a) and (4.157b),
Z b
Vd = −η Hm · dℓℓ
a
and
Z d
2
Id = − Em · dℓℓ .
η c

Assuming points a and c in Fig. 4-22, and the same for b and d, are very close to each other,
the input impedance of the dipole antenna can be computed from
Rb
Vd η 2 a Hm · dℓℓ η 2 Is
Zd = = R = . (4.160)
Id 2 d Em · dℓℓ 4 Vs
c

Recognizing the input impedance of the complementary slot Zs = Vs /Is , it is now easy to show
that
Zd Zs = η 2 /4 . (4.161)
Although Eq. (4.161) was obtained for a dipole antenna example, this result is valid for any
planar antenna and its complementary configuration, since no specific assumptions were made
about the operation or geometry of the dipole in deriving Eq. (4.161).
It should also be noted that the electric field radiation pattern of the dipole and the
magnetic field radiation pattern of its complementary antenna (slot antenna) are identical and
vice versa.

4-8.4 Self-Complementary Antennas

In most communication and radar systems, antennas play a very important role in directing the
transmitted power along specific directions of interest to the intended application. Across the
evolutionary history of wireless devices, multiple bands of operations have been added to each
device to enable different types of services and to support high data-rate communication. For
4-8 Babinet’s Principle 213

∞ ∞

∞ ∞

∞ ∞

∞ ∞
Figure 4-23: A planar antenna and its complementary structure. By rotating the antenna
counterclockwise by 90◦ , its complementary geometry is generated. Such an antenna structure
is self-complementary.

such devices, small antennas with wide bandwidth are required. The subject of frequency-
independent antennas that can provide infinite or, in practice, very wide bandwidth has
received significant attention in the recent past. For such antennas, it is desired that the input
impedance of the antenna be a real quantity and independent of frequency. In this section it is
shown that self-complementary antennas can, in principle, provide infinite bandwidth. Self-
complementary antenna structures refer to planar antenna geometries in which the metallic
portion of the antenna has exactly the same geometry as the slot or opening on the antenna
plane. Basically, for such antennas the complementary geometry can be obtained from the
original antenna by simply rotating the antenna by some fixed angle. To demonstrate this
concept, let’s consider the antenna geometry and its complementary structure shown in
Fig. 4-23. The complementary of the antenna geometry can be obtained from the antenna
structure itself by rotating the antenna by 90◦ . Since the geometry of the antenna and its
complementary are the same, then it is obvious that the input impedance of the antenna, ZA ,
and that of its complementary, ZC , are the same. Using Eq. (4.161), it is evident that

ZA ZC = ZA2 = η 2 /4 . (4.162)

Hence, the input impedance of a self-complementary antenna is given by

ZA = η /2 , (self-complementary antenna) (4.163)

which is independent of the antenna geometry and independent of frequency. It is noted

that in practice it is not possible to fabricate antennas with metallic sheets that extend to
infinity. Depending on the space available, the metallic sheets must be truncated. The size
of the truncated metallic sheets that define the antenna determines the low-end of the usable
frequency range, and the gap size for the antenna terminals determines the upper end of the
214 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

frequency range, yielding a wideband, but of finite bandwidth, antenna. Other geometries
for self-complementary antennas can also be considered. For example, a combination of a
dipole antenna and its complementary slot can form a self-complementary antenna, as shown
in Fig. 4-24. Also shown in the figure is an arbitrarily shaped self-complementary antenna.
As an example of a small antenna with a wide bandwidth, let’s consider a modified version
of the self-complementary dipole. The height of the dipole can be shortened by top-loading
it. Placement of an elliptical top-load provides multiple paths for the electric current on the
dipole to create a resonant condition and thereby allows for increasing the bandwidth of the
dipole inherently. Creating the complementary geometry and feeding the antenna, as shown in
Fig. 4-25, provides a very wide bandwidth, as shown in the same figure. This antenna is about
λm /5 in height, where λm is the wavelength at the minimum frequency. The lower frequency
is limited by the size of the ground plane, and the high frequency is limited by how faithful
the antenna near the feed is to the complementary geometry. One important drawback of self-
complementary antennas pertains to their radiation pattern. Although a self-complementary
antenna is impedance-matched over a very wide bandwidth, its radiation pattern can change
drastically over its bandwidth.

+
_

(a) Dipole antenna (b) Zigzag antenna

Figure 4-24: A self-complementary dipole antenna and a self-complementary “zigzag” antenna.

4-8 Babinet's Principle 215

cm

cm ≈

cm

(a) Top-loaded dipole antenna

−5

−10

−15

−20

−25

−30

(b) Frequency response of the antenna reflection coefficient

Figure 4-25: (a) A self-complementary top-loaded dipole antenna and (b) the antenna reflection
coefficient as a function of frequency. The antenna, with dimensions of about λm /4, provides a
bandwidth on the order of two octaves. Here, λm is the wavelength at the lowest frequency of
operation, corresponding to fm = 300 MHz.
 216 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Summary
Concepts

• The concept of Green’s function, which (MFIE), can be derived by applying the
presents the spatial impulse response of the continuity of the tangential electric field and
wave equation for a homogeneous unbounded the tangential magnetic field at the interface
medium, is presented. between the background medium and the
• Solutions of wave equations for the electro- scatterer, respectively.
magnetic potentials for arbitrary charge and • A fundamental theorem in electromagnetism
current distributions are obtained rigorously known as the reciprocity theorem is pre-
using Green’s function with the help of Green’s sented. This theorem establishes the relation-
second identity. ship between two different sources (electric
• The solution to wave equations based on and magnetic current distributions) and their
Green’s function for a homogeneous unbounded corresponding fields.
medium is extended to include complex media • The reciprocity theorem is applied to antennas
made up of a heterogeneous medium composed to show that the radiation pattern of an antenna
of a finite number of objects with different in the receiving mode is the same as its pattern
material properties embedded in an otherwise in the transmitting mode.
homogeneous unbounded medium. • The notion of effective area for a receiving
• The extinction theorem is introduced in the antenna is introduced and its relationship to
form of an integral equation, which states that its directivity function is derived using the
the equivalent sources (potentials or fields and reciprocity theorem.
their normal derivatives) on the surfaces of • Babinet’s theorem and the complementary
the scatterers are formed in such a way as to theorem that are applicable to planar conducting
produce potentials or fields inside the scatterers screens with arbitrary apertures are presented.
that can completely cancel out the incident • The concept of a complementary antenna,
potentials or fields inside the scatterers. which refers to a planar antenna for which the
• The extinction theorem is modified for an metallic and aperture areas are the opposite
observation point on the surface of the scatterers of those of the original antenna, is presented
to form integral equations for the unknown and shown to produce the same radiation
surface potentials or fields and their normal pattern as the original antenna with orthogonal
derivatives on the surfaces of the scatterers. polarization. Also the relationship between the
• Appropriate boundary conditions for the poten- input impedances of a planar antenna and its
tials at interfaces between two dielectric media complementary geometry is derived.
or between a dielectric medium and a metallic • The term “self-complementary antenna” refers
one are derived. to an antenna whose complementary geometry
• Integral equation formulations based on surface is the same as that of the antenna itself.
field equivalent sources are derived. Such antennas are shown to provide infinite
• It is shown that two independent integral equa- bandwidth (their input impedance is not a
tions, namely electric-field integral equations function of frequency).
(EFIE) and magnetic-field integral equations
SUMMARY 217

Important Equations

Free-space scalar Green’s function:

′
eik|r−r |
g(r, r′ ) =
4π |r − r′ |

Green’s first identity:

$ $
2 ∂Φ
∇ψ · ∇Φ dv + ψ ∇ Φ dv = ψ ds
V V ∂n

Green’s second identity:

$  
2 2 ∂Φ ∂ψ
[ψ ∇ Φ − Φ ∇ ψ ] dv = ψ −Φ ds
V S ∂n ∂n

Radiation condition for potentials and fields:

 
∂ Φ(r)
lim r − ik Φ(r) = 0
r→∞ ∂r
 
∂ A(r)
lim r − ik A(r) = 0
r→∞ ∂r
lim r [E(r) + r̂ × η H(r)] = 0
r→∞

lim r [nH(r) − r̂ × E(r)/η ] = 0

r→∞

Integral equations for scalar potential based on the extinction theorem:


 ′ ′
 
Φ(r) r∈V ,
′ ∂ Φ(r ) ′ ∂ g(r, r )
Φi (r) + −−− g(r, r ) − Φ(r ) ds = Φ(r)/2 r on S j ,
P
Si ∂ n′ ∂ n′ 

0 r<V ,

 ′ ′  Φ j (r)
 r ∈ Vj ,
′ ∂ Φ j (r ) ′ ∂ g j (r, r ) ′ 1
−−− gi (r, r ) − Φ j (r ) ds = 2 Φ j (r) r on S j ,
∂ n′ ∂ n′ 

0 r < Vj
218 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Important Equations (continued)

Boundary conditions for electromagnetic potentials at interface between two

dielectric media:

A1 (r) = A2 (r) , r∈S

Φ1 (r) = Φ2 (r) , r∈S
∂ Φ1 ∂ Φ2
ε1 = ε2 r∈S
∂n ∂n
Boundary conditions at interface between dielectric and metallic media:

A1 (r) = 0, r∈S
Φi (r) = constant, r ∈ S

Reciprocity theorem:

ha, bi = hb, ai
$
ha, bi = (Ea · Jb − Ha · Jmb ) dv
V

Reciprocity theorem for a source-free region:

(Ea × Hb − Eb × Ha ) · ds = 0

Two-port network reciprocity relations between voltages and currents at the

terminals of two antennas:
Va Ib = Vb Ia
SUMMARY 219

Important Equations (continued)

Electric-field integral equations:

1
− Jms (r)
2
 
′ ′ iη1 ′
− n̂ × −−− iη1 k1 Js (r ) g1 (r, r ) + ∇s · Js (r ) ∇g1 (r, r ) + Jms (r ) × ∇g1 (r, r ) ds′
′ ′ ′ ′
Ss k1

= n̂ × Ei(r)
1
− Jms (r)
2
 
′ ′ iη2 ′
+ n̂ × −−− iη2 k2 Js (r ) g2 (r, r ) + ∇s · Js (r ) ∇g2 (r, r ) + Jms (r ) × ∇g2 (r, r ) ds′
′ ′ ′ ′
Ss k2

=0

Magnetic-field integral equations:


1 ik1
J(r) − n̂ × −−− Jms (r′ ) g1 (r, r′ )
2 Ss η1

i
+ ∇ · Jms (r ) ∇g1 (r, r ) − Js (r ) × ∇g1 (r, r ) ds′
′ ′ ′ ′ ′
k1 η1 s
= n̂ × Hi(r)

1 ik2
J(r) + n̂ × −−− Jms (r′ ) g2 (r, r′ )
2 Ss η 2

i
+ ∇ · Jms (r ) ∇g2 (r, r ) − Js (r ) × ∇g2 (r, r ) ds′
′ ′ ′ ′ ′
k2 η2 s
=0.

Equivalence of antenna patterns in transmission and reception:

Dt (θ , φ ) = Dr (θ , φ )

Effective area for receiving antenna:

λ2
Aeff (θ , φ ) = G(θ , φ )
4π
 220 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Important Equations (continued)

Babinet’s theorem in reference to Fig. 4-21:

r r
µ m ε m
E1 = − Hs , H1 = E
ε µ s

Impedance relation between the input impedances of an antenna (Zd ) and its
complement antenna (Zs ):
Zd Zs = η 2 /4
Input impedance of a self-complementary antenna:

ZA = η /2

Important Terms Provide definitions or explain the meaning of the following terms:
antenna radiation pattern in receive Friis transmission formula radiation condition
and transmit modes Green’s first identity reciprocity condition for
antennas as two-port networks Green’s function two-port networks
Babinet’s principle Green’s second identity reciprocity theorem
boundary conditions for the potentials Green’s theorem reciprocity theorem for a
complementary antenna Helmholtz equation source-free region
complementary theorem incident potential scattered potential
effective area of a receiving antenna magnetic-field integral equation self-complementary antenna
electric-field integral equation (EFIE) (MFIE) singular point
extinction theorem physical optics current on metallic surface-field integral equations
Fredholm integral equation surfaces wave equation
frequency-independent antennas principal-value integral
PROBLEMS 221

PROBLEMS
Integral Equations

4.1 Follow a procedure similar to that used to derive Eq. (4.36) for electric scalar potential
where the observation point r is on the surface of the scatterer to show that the contribution
from the singularity of the scalar Green’s function in the surface integral of Eq. (4.65) is
1
2 E(r).

4.2 Consider a scatterer whose surface is denoted by S j , similar to what is shown in Fig. 4-3.
However, this surface has surface slope discontinuities at a finite number of points, as shown
in Fig. P4.2. If the observation point happens to be at an edge point, how would you change
Eq. (4.36) for this point?

α0
Sj

Figure P4.2: Surface of an object with a slope discontinuity.

4.3 Consider a metallic sphere of radius a centered at the origin of a Cartesian coordinate
system, as shown in Fig. P4.3. Suppose a plane wave propagating along the +x direction is
illuminating the metallic sphere. The electric field expression for the incident field is given by

Ei = E0 ŷ eik0 x ,

with
2π √
k0 = = ω µ0 ε0 .
λ0
Assuming a ≫ λ0 , use MFIE for conductors to show that the surface current is approximately
J(r) = 2n × H1 (r).
(a) Using this current only on the lit portion of the surface, find the scattered field.
(b) Note that the first-order physical-optics current also exists in the shadow area. Show
that the first-order physical-optics current produces a magnetic field that almost cancels
out the incident magnetic field in the shadow region.
222 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

z
y

Hi
Ei x
k̂i

Figure P4.3: A metallic sphere illuminated by a plane wave.

Reciprocity

4.4 Figure P4.4 shows two antennas with an obstacle. Let V1 be the voltage received at the
terminals of antenna (1) when a unit current source is applied at the terminals of antenna (2),
and let V2 be the voltage received at antenna (2) when a unit current source is applied at
antenna (1). Let V1i and V2i be the corresponding voltages when the obstacle is absent. Define
the scattered voltages as

V1s = V1 −V1i and V2s = V2 −V2i ,

and show that V1s = V2s .

(1)

Obstacle

(2)

Figure P4.4: Diagram for Problem 4.4, showing two antennas that are used as transmit and
receive pairs. In one experiment, antenna (1) is the transmitter and antenna (2) is the receiver. In
the second experiment, the roles of the two antennas are interchanged.
PROBLEMS 223

4.5 Consider a homogeneous anisotropic medium with permittivity and permeability tensor
elements ε and µ , respectively. Following a procedure similar to that used in the proof of the
reciprocity theorem, provide the necessary and sufficient conditions for fields to be reciprocal
in the anisotropic medium. (Hint: For two arbitrary vectors, the dot product a · b in matrix
notation can be written as aT b.)

4.6 Consider a homogeneous medium with the following constitutive relations:

D = εE + ξ H and B = ξ ∗E + µ H ,

where ξ is related to the Tellegen and Pasteur parameters χ and κ as follows:

√
ξ = µε (χ − jκ ) .

Find the necessary and sufficient conditions for the fields to be reciprocal in this medium.

4.7 Consider a typical communications system as shown in Fig. P4.7(a), in which the
antennas and the relative separation and orientation are completely arbitrary. The same
physical situation is shown in Fig. P4.7(b), except that the transmitter and receiver circuits
have been interchanged. We assume that the transmitter and receiver are both connected to the
antennas through a length of single-mode transmission line or waveguide. The only impressed
sources in the problem are located in the transmitter. The fields and currents inside V in each
of the two situations of the figure are denoted by (Ea , Ha , Ja ) and (Eb , Hb , Jb ), respectively.
By letting the surface S extend to infinity and applying reciprocity, show that Z12 = Z21 ,
where Zi j ’s are the Z-parameters of this two-antenna system. (Assume that the transmitter
and receiver and feed lines are shielded by a perfect conducting surface.)
    
V1 Z11 Z12 I1
= .
V2 Z21 Z22 I2

(Hint: You can assume that a coaxial cable connects the antenna to Tx/Rx. Then use the
fields distribution inside the coaxial cable (the tangential electric field Et is proportional to the
voltage V and the tangential magnetic field Ht is proportional to the current I).)

4.8 What is the effective area of a Hertzian dipole? If a Hertzian dipole carrying current
I0 is placed at the origin of a Cartesian coordinate system and oriented along the z-axis, and
θ direction, find
another one, used as receiver, is placed at location r and oriented along the θ̂
the power received by the receiver when the receiver is impedance-matched.

4.9 Consider a transmitter with power Pt connected to an antenna with gain Gt (θ , φ ) that is
impedance-matched. Another antenna with gain Gr (θ , φ ) at a distance R from the transmitter
is impedance-matched to a receiver. Show that the receiver power can be calculated from

λ 2 Gt Gr
Pr = Pt .
(4π R)2
This equation is known as the Friis transmission formula.
224 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

S Ra , Ha , Ja

2
1

Rx
Tx
Reference planes S2
S1

(a) (Tx, Rx) = (Antenna 1, Antenna 2)

S Rb , Hb , Jb

2
1

Tx
Rx
Reference planes S2
S1

(b) (Tx, Rx) = (Antenna 2, Antenna 1)

Figure P4.7: Diagram for Problem 4.7, showing a pair of antennas used as transmitter and
receiver in two complementary experiments.
PROBLEMS 225

4.10 Consider the dihedral corner reflector shown in Fig. 4-18 of Example 4-2. Applying
reciprocity, calculate the backscatter from the dihedral when it is illuminated by a plane wave
given by
Ei (r) = E0 (− sin φ x̂ + cos φ ŷ)eik0 (cos φ x+sin φ y) .

4.11 Find an expression for the radar backscatter cross section of the dihedral corner
reflector shown in Fig. 4-18, with the only difference being that the angle between the two
plates is α . Find out how the RCS decreases when α deviates from 90◦ .

4.12 Derive an expression for the scattered field for two adjacent objects in a case where one
object is made from a perfect electric conductor and the other object is made from a dielectric.

4.13 Consider a dielectric sphere of radius a having a permittivity ε = 1.05ε0 placed above
a metallic plate in the x–y plane, as shown in Fig. P4.13. The plate has dimensions b × b, with

z
Ei(r) Hi(r)
ε0
Plane wave
y
1.05ε0
a
1.5a
x
b

Figure P4.13: Diagram for Problem 4.13, showing a dielectric sphere above a finite-size square-
shaped metallic plate illuminated by a plane wave.

b ≫ λ . The sphere center is on the z-axis at a height h = 1.5a. These objects are illuminated
by a plane wave given by

Ei (r) = E0 (cos θ x̂ + sin θ ẑ)eik(sin θ x cos θ z) .

For b > 6a, find the backscatter field from these objects. (Hint: Since the permittivity of the
sphere is very close to that of the background (ε0 ), the Born approximation can be used. Using
the Born approximation, the fields inside the dielectric sphere can be approximated by those
of the incident wave (in the absence of the sphere).)
226 Chapter 4 Formal Solutions to Maxwell’s Equations and Their Applications

Babinet’s Principle

4.14 Consider an electric source at z = −∞ that produces the following plane wave:

Ei = E0 eikz x̂ ,

and a metallic sheet in the z = 0 plane, with a square aperture of length L located at the origin.
The aperture extends from x = −L/2 to x = L/2 and from y = −L/2 to y = L/2. Using a
high-frequency approximation, the fields for z > 0 can be approximately estimated by the
following expression:
   
eikr 2 Lky Lkx
E = x̂ E0 L sinc sinc .
r 2r 2r
Assume that the source was replaced with a magnetic source and the ground plane with
the aperture was replaced with a square metal plate of the same size as the aperture. Use
Babinet’s principle to find the fields in the z > 0 half-space. The sinc(·) function is defined as
sinc(x) = (sin x)/x.
 5 Chapter
Electromagnetic
Plane Waves

Chapter Contents Objectives

Overview, 228 Upon learning the material presented in this chapter, you
5-1 Plane-Wave Propagation in Homogeneous should be able to:
Media, 228
5-2 Polarization of Plane Waves, 238 1. Have an intuitive understanding of the important
5-3 kDB Coordinate System for Plane Waves in properties associated with plane-wave
Bianisotropic Media, 247 propagation in a homogeneous isotropic medium,
5-4 Transverse Electric (TE) and Transverse including the propagation constant of the
Magnetic (TM) Field Solutions of the medium, wave polarization, and phase front.
Helmholtz Equation, 254
5-5 Plane-Wave Reflection at the Interface 2. Characterize the field in a source-free region,
between a Dielectric Medium and a when generated by sources outside that region, in
Good-Conducting Medium, 264 terms of a continuous spectrum of plane waves.
5-6 Wave Propagation in an Inhomogeneous
Medium: Geometric Optics Approximation, 3. Develop the plane-wave solution in an aniso-
266 tropic medium using the kDB system.
5-7 Plane-Wave Reflection by and Transmission
through a Half-Space Uniaxial Medium, 275 4. Explain the expansion of fields in terms of
5-8 Plane Waves in Layered Media, 280 transverse electric and transverse magnetic fields.
5-9 Plane-Wave Propagation in a Negative-Index
Medium, 290 5. Calculate plane-wave reflection and transmission
5-10 Negative Refractive-Index Lens, 292 at planar boundaries between a dielectric medium
Chapter Summary, 298 and another multilayer dielectric medium that
Problems, 303 may be an isotropic or anisotropic uniaxial
medium.

6. Understand the concept of negative-index media

and the characteristics of plane waves
propagating in such media.

227
228 Chapter 5 Electromagnetic Plane Waves

Overview
In this chapter, we examine the solution to Maxwell’s equations for a source-free
homogeneous medium. It is shown that the electric and the magnetic fields both satisfy the
vector wave equation. Considering an invariant direction for the electric field throughout
the medium, the vector wave equation is reduced to a scalar wave equation, which is
then solved using the method of separation of variables. The resultant solution is known
as a plane wave, as the magnitude and the phase of the electric and magnetic fields are
constant in planes perpendicular to the direction of propagation. For such waves, the electric
field is perpendicular to the magnetic field and both are perpendicular to the direction of
propagation. The notion of nonuniform plane waves is also introduced, for which the planes
of constant phase and constant magnitude differ. It is shown that the superposition of all
such planes also constitutes a solution to Maxwell’s equations, provided that the sum of the
square of the propagation constants of the components of these plane waves in the three
principal orthogonal directions are equal to the square of the wave number, k2 ; in other
words Kx2 + Ky2 + Kz2 = k2 = ω 2 µε . Explicit expressions for power and energy density of
plane waves are obtained. The notion of plane-wave polarization in terms of parameters
of a polarization ellipse is introduced. Then plane-wave solutions for homogeneous but
anisotropic media are examined. The dispersion relation that relates different components
of the propagation constant along different principal directions to the frequency and tensor
elements of permittivity is obtained for the general case, and examples for simpler uniaxial
media are discussed in more detail. The kDB coordinate system is presented for plane waves
in anisotropic media. This coordinate system facilitates simpler and more intuitive relations
between field components. The notion of ordinary and extraordinary waves in a uniaxial
medium is introduced in the kDB system, and simplified dispersion relations for them are
derived. To study the behavior of plane waves at planar boundaries between two dielectric
media, the concepts of transverse electric (TE) and transverse magnetic (TM) field expansion
is introduced. It is shown that all field components, in a source-free region, for transverse
magnetic fields can be obtained from an electric Hertz vector potential that has only one
component along a desired direction of propagation. Also, all field components for TE waves
can be generated from a magnetic Hertz vector potential that has only one component along
the direction of propagation. The reflection and transmission coefficients of an incident plane
wave at an interface between two media, as well as for multilayer media, are obtained using
TE and TM wave expansions. These results are extended to include calculation of plane-wave
reflection and transmission at planar interfaces between an isotropic and a uniaxial medium
with its optical axis parallel to the interface. Finally, plane-wave propagation in negative-index
media is examined and the concept of a perfect lens made up of a dielectric slab of negative-
index material whose index of refraction is the negative value of the index of refraction of the
surrounding medium is presented.

5-1 Plane-Wave Propagation in Homogeneous Media

In this section, we consider the solution of Maxwell’s equations in an unbounded, source-free,
homogeneous medium. Let us consider a medium with constitutive parameters ε , µ and free
5-1 Plane-Wave Propagation in Homogeneous Media 229

of currents and charges. For this problem, Maxwell’s equations take the following form:

∇ × E = iω µ H , (5.1a)
∇ × H = −iωε E , (5.1b)
∇·E = 0 , (5.1c)
∇·H = 0 . (5.1d)

In general ε and µ are complex quantities. The equations for the electric and magnetic fields
can be decoupled by first taking the curl of Eq. (5.1a) and then using Eq. (5.1b) in the resulting
equation:
∇ × ∇ × E = iω µ ∇ × H = ω 2 µε E . (5.2)
But
∇ × ∇ × E = ∇∇ · E − ∇2 E . (5.3)
Using Eq. (5.1c) in Eq. (5.3), Eq. (5.2) can be written as

∇2 E + k 2 E = 0 , (5.4)

where k2 = ω 2 µε . A similar procedure can be followed to show that

∇2 H + k 2 H = 0 . (5.5)

The quantity k is called the propagation constant of the medium. Equations (5.4) and (5.5)
indicate that both the electric and magnetic fields satisfy the vector wave equation. These
wave equations are separable in the Cartesian coordinate system and each represents three
scalar wave equations for the three components of the fields. That is,
   
Ex Ex
∇2 Ey  + k2 Ey  = 0 . (5.6)
Ez Ez

Once the solution for E is obtained, it is much easier to find the solution for H directly using
Eq. (5.1a) than to repeat the solution process with Eq. (5.5). That is,
1 i
H= ∇×E = − ∇×E , (5.7)
iω µ kη
q
µ
where η = ε .
The simplest form of solution that one can conceive for the electric field is of the
following:
E = E0 ψ (r) , (5.8)
230 Chapter 5 Electromagnetic Plane Waves

where E0 is a constant vector and ψ (r) is a scalar function of r. Upon substituting Eq. (5.8)
into Eq. (5.4), it can be easily shown that

E0 ∇2 ψ (r) + E0 k2 ψ (r) = 0 ,

which produces a nonvanishing solution if

∇2 ψ (r) + k2 ψ (r) = 0 . (5.9)

We assume that ψ (r) is separable; i.e.,

ψ (r) = X (x) Y (y) Z(z) , (5.10)

where X , Y , and Z are single variable functions of x, y, and z, respectively.

Substituting this solution into Eq. (5.9), we have
 2 
∂ ∂2 ∂2
+ + XY Z + k2 XY Z = 0 . (5.11)
∂ x2 ∂ y2 ∂ z2

For points where ψ (r) is nonzero, both sides of Eq. (5.11) can be divided by XY Z:
 
1 ∂2 1 ∂2 1 ∂2
X+ Y+ Z + k2 = 0 . (5.12)
X ∂ x2 Y ∂ y2 Z ∂ z2
The first term in the parenthesis is only a function of variable x, the second one is only a
function of variable y, and the last one is only a function of variable z. Hence for Eq. (5.12) to
be valid for all values of x, y, and z, each of the three terms in the parenthesis must be equal
to a constant; i.e.,

1 ∂2
X = −Kx2 , (5.13a)
X ∂ x2
1 ∂2
Y = −Ky2 , (5.13b)
Y ∂ y2
1 ∂2
Z = −Kz2 . (5.13c)
Z ∂ z2
Equation (5.12) is satisfied if

Kx2 + Ky2 + Kz2 = k2 = ω 2 µε . (5.14)

Finding solutions to the second-order differential equations given by Eq. (5.13) is

straightforward. The solutions are of the following form:

X (x) = eiKx x .
Y (y) = eiKy y ,
5-1 Plane-Wave Propagation in Homogeneous Media 231

Z(z) = eiKz z .

Hence
E(r) = E0 ei K·r , (5.15)
where

K = Kx x̂ + Ky ŷ + Kz ẑ
and
√
|K| = k = ω µε .

For lossless media wherein both µ and ε are real, the unit vector k̂ representing the direction
of propagation is
K Kx Ky Kz
k̂ = = x̂ + ŷ + ẑ
|K| k k k
= kx x̂ + ky ŷ + kz ẑ , (5.16)

where kx2 + ky2 + kz2 = 1 and k̂ denotes a direction for which the phase of the electric field in
planes perpendicular to that direction is constant. Basically

k̂ · r = const.

defines a plane perpendicular to k̂, as shown in Fig. 5-1.

In this plane, both the phase and magnitude of the electric field are constant, and that is
why this specific solution of Maxwell’s equations is known as a plane wave. Note that even

kˆ · r = const.

r
k̂

Plane wave

Figure 5-1: Phase front of a uniform plane wave is shown to be a plane across which the
magnitude and phase of the wave have constant values.
232 Chapter 5 Electromagnetic Plane Waves

for real values of ε and µ , Eq. (5.15) may represent a nonpropagating wave. For example, if

Kx2 + Ky2 > k2 ,

then q
Kz = i Kx2 + Ky2 − k2
and
E = E0 ei(Kx x+Ky y) e−Kz z .
In this case the wave attenuates exponentially along the z direction. The wave is then said
to be evanescent along the +z direction. The magnitude E0 is not entirely a free parameter.
Computing the divergence of the electric field shows that

∇ · E = E0 · ∇ei K·r = i E0 · K ei K·r = 0 ,

which indicates that

E0 · k̂ = 0 . (5.17)
Equation (5.17) states that the electric field of the plane wave is normal to the wave’s direction
of propagation. The corresponding magnetic field can be obtained from Eq. (5.7):
−i −i
H= ∇ × (E0 ei K·r ) = (i K × E0 )ei K·r ,
kη kη
or
k̂ × E0 i K·r
H= e , (5.18)
η
which indicates that H is perpendicular to both the electric field and the waves direction of
propagation. It is interesting to also note that
r
|E| µ
=η =
|H| ε
at all points in space.

5-1.1 Plane Waves in Lossy Media

In situations where either the permittivity ε or the permeability µ (or both) is complex, the
propagation constant √
k = ω µε = k′ + iα (5.19a)
becomes complex, with k′ and α representing the real and imaginary parts of k. A special
case of wave propagation in this case is when the unit vector k̂ is real. Such a plane wave is
referred to as a uniform plane wave. Consider a direction of wave propagation denoted by
spherical coordinate angles θ and φ for which

k̂ = kx x̂ + ky ŷ + kz ẑ = sin θ cos φ x̂ + sin θ sin φ ŷ + cos θ ẑ . (5.19b)

5-1 Plane-Wave Propagation in Homogeneous Media 233

In a plane perpendicular to k̂, where

k̂ · r = ξ = const. ,

the phase of E, given by ξ k′ , is constant. In this same plane the magnitude of the electric field,
given by
|E| = |E0 |e−αξ ,
is also constant. Hence, for a uniform plane wave whose direction of propagation is real, both
the magnitude and phase of the wave are constant in planes perpendicular to the direction
of propagation. In general, the planes of constant phase and magnitude are different. As an
example, let us consider a wave for which Kx and Ky are real and
q
Kz = k2 − Kx2 − Ky2 (5.20)

is complex (since k is complex). Then

Kz = Kz′ + iKz′′ .

For this wave, the electric field assumes the following form:
′ ′′
E = E0 ei K ·r e−Kz z , (5.21)

where now K′ = Kx x̂ + Ky ŷ + Kz′ ẑ.

Let’s define a unit vector

Kx x̂ + Ky ŷ + Kz′ ẑ
k̂′ = q .
Kx2 + Ky2 + Kz′ 2
Then in a plane where
k̂′ · r = ξ = const.
the phase of the wave is constant. Considering that the magnitude of the wave is given by
′′
|E| = |E0 |e−Kz z ,

it is obvious that the magnitude of the plane wave is constant in constant z-planes. Therefore
in general the constant magnitude and constant phase planes are different for lossy media.
Such plane waves are referred to as nonuniform plane waves.
In general, defining the complex phase as

φ = K·r , (5.22)

the vector quantity ∇φ specifies a vector that is normal to surfaces of constant φ . In cases
where K is real, ∇φ = K, which indicates that equiphase surfaces are planes perpendicular
to K. For a complex K vector where

K = K′ + i K′′ ,
234 Chapter 5 Electromagnetic Plane Waves

∇φ = K′ + i K′′ , indicating that equiphase surfaces are planes perpendicular to K′ and equi-
amplitude surfaces are planes perpendicular to K′′ . Assuming that the medium is lossless,

k2 = K · K = |K′ |2 − |K′′ |2 + 2i K′ · K′′ , (5.23)

which mandates a vanishing imaginary part. The imaginary part of Eq. (5.23) vanishes if
K′′ = 0 or K′ · K′′ = 0. If K′′ = 0, a regular uniform plane wave is obtained. Also, K′ · K′′ = 0
denotes a possible plane-wave solution in a lossless medium. That is, the equiphase and
equiamplitude planes are orthogonal, similar to the example considered earlier. This is known
as an evanescent wave. Lossy media can also support evanescent plane waves, but the
equiphase and equiamplitude planes are not necessarily orthogonal.

5-1.2 Expansion of a Field as a Continuous Spectrum of Plane Waves

In the previous section it was demonstrated that an elementary wave function of the form
E0 ei K·r is a solution to the homogeneous Helmholtz equation for arbitrary parameters Kx , Ky ,
and Kz , subject to the constraint

Kx2 + Ky2 + Kz2 = k2 .

That is, any two of the three parameters can be considered as free parameters that may take on
real numbers from −∞ to +∞. Hence, there are an infinite number of plane waves that satisfy
the wave equation for homogeneous and unbounded media. Since Maxwell’s equations are
linear, any linear combination of these plane waves is also a solution. That is, the electric field
produced by an arbitrary source, in a source-free region, may be written as
Z +∞ Z +∞
E(r) = E0 (Kx , Ky ) ei K·r dKx dKy (5.24a)
−∞ −∞
Z +∞ Z +∞  
= E0 (Kx , Ky ) eiKz z eiKx x eiKy y dKx dKy ,
−∞ −∞

and its corresponding magnetic field is given by

Z +∞ Z +∞
K × E0 (Kx , Ky ) i K·r
H(r) = e dKx dKy . (5.24b)
−∞ −∞ kη
Although the elementary field components of the electric and magnetic fields are orthogonal to
each other and their magnitude ratios are a constant (and equal to the characteristic impedance
of the medium), the same is not true for the overall field quantities. It should also be noted
that the choice of Kx and Ky as free parameters is arbitrary; i.e., any two pairs of Kx , Ky , Kz
could have been chosen. Also, as was shown before, the constant magnitude vector E0 is not
entirely arbitrary (see Eq. (5.17)). Equation (5.24a) indicates that in a source-free region the
fields produced by arbitrary electric or magnetic currents, outside the region of interest, can
be expressed in terms of the superposition of an infinite number of plane waves. It is also
interesting to note that Eq. (5.24a) represents a two-dimensional Fourier transformation , and
that its inverse can be used to find E0 (kx , ky ) in terms of E(r). Basically, taking the inverse
5-1 Plane-Wave Propagation in Homogeneous Media 235

Fourier transform of Eq. (5.24a), we have

Z +∞ Z +∞
E0 (Kx , Ky ) eiKz z = E(x, y, z) e−iKx x e−iKy y dx dy . (5.25)
−∞ −∞

Example 5-1: An Application of Plane-Wave Field Expansion

Show that by measuring the tangential electric field in a constant-z plane (z = z0 ) in the vicinity
of an antenna can be used to find the fields at every point where z > z0 . This approach can be
used to find the radiation pattern of antennas.
Solution: When characterizing the radiation pattern of an antenna, traditionally the antenna
under test (AUT) is used as a transmitter and another antenna is used as a receiver to
measure the power radiated by the AUT. The receiver antenna is moved around the AUT
at a constant distance and the recoded power provides the radiation pattern. For proper
measurements, the receiving antenna must be placed in the far-field region of the AUT in
an environment free of any objects that may perturb the radiated field. This often poses
challenging practical conditions for the radiation pattern measurement of large antennas.
Alternatively, by measuring the electric field in the vicinity of the antenna on a planar surface,
the field outside the surface in the source-free region can be determined. Figure 5-2 shows
the configuration for the problem. Let us denote the tangential electric field in the constant-z
plane by Et (x, y, z0 ) = Ex (x, y, z0 ) x̂ + Ey (x, y, z0 ) ŷ. Assuming that this field is measured and
known, we can find the tangential component of the field spectrum from the Fourier transform
of the measured tangential electric field:
Z +∞ Z +∞
E0t (Kx , Ky ) = e−iKz z0 Et (x, y, z0 )e−iKx x e−iKy y dx dy .
−∞ −∞

In practice the tangential electric field is not measured over the entire constant-z surface due
to the fact that fields decay fast for points away from the antenna and thus the limits of the
integrals can be truncated at some finite values. It was shown before for plane waves that
Gauss’s law of electricity mandates K · E0 = 0. This can be used to find the z-component
of E0 ; that is,
Kx E0x (Kx , Ky ) + Ky E0y (Kx , Ky )
E0z (Kx , Ky ) = − q .
k2 − Kx2 − Ky2

Once all three components of E0 (Kx , Ky ) have been determined, Eq. (5.24a) can be used to
find the electric field everywhere (z > z0 ). This is in agreement with the uniqueness theorem,
which states that the fields in a source-free region can be determined from knowledge of
the tangential electric field over a closed surface. The closed surface in this example can be
regarded as composed of the constant-z surface and an infinite half-sphere closing over the
constant-z surface.
236 Chapter 5 Electromagnetic Plane Waves

)
(x, y, z0
Et

z0
y z

Figure 5-2: Configuration for the antenna near-field measurement. The tangential component
of the electric field phasor generated by an arbitrary antenna is measured across a planar surface
near the antenna. This field is used to characterize the fields everywhere outside the surface in the
source-free region.

5-1.3 Power and Energy Densities

Consider an elementary plane-wave solution of a Helmholtz equation whose electric and

magnetic fields were found to be

E = E0 ei K·r
and
k̂ × E0 i K·r
H= e .
η
The stored electric energy per unit volume is given by
1 ′ 2 1 ′ ′′
we = ε |E| = ε |E0 |2 e−2 K ·r ,
4 4
and the stored magnetic energy per unit volume is

1 ′ 2 1 ′ |k̂|2 |E0 |2 −2 K′′ ·r

wm = µ |H| = µ e .
4 4 |η |2
5-1 Plane-Wave Propagation in Homogeneous Media 237

In a lossless medium, K′′ = 0, and noting that µ /η 2 = ε ,

1
wm = we = ε |E0 |2 . (lossless medium) (5.26)
4

That is, the stored electric and magnetic energies for a plane wave are independent of position
and are equal to each other. In contrast, for a lossy medium the stored energies are functions
of position and are not necessarily equal.
The complex power flow density can be obtained from the Poynting vector and is given
by
1 1 |E0 |2 k̂∗ − (E0 · k̂∗ )E0∗ −2 K′′ ·r
S(r) = E × H∗ = e . (5.27)
2 2 η∗
For a uniform plane wave in a lossless medium, the expression becomes much simpler and is
given by
|E0 |2
S(r) = k̂ . (5.28)
2η
Considering a uniform plane wave traveling along k̂, the amount of energy delivered to a
volume ∆ v along its direction of propagation over an interval ∆ t can be computed using

∆ W = (S · k̂ A) ∆ t . (5.29)

If we assume that the energy of the wave is propagating with velocity ue along k̂, then the
volume shown in Fig. 5-3 is filled with the wave energy if

∆ r = ue ∆ t . (5.30)

k̂
A

k̂ Δr

Figure 5-3: A fictitious volume in space intercepting an incident plane wave with its front and
back faces perpendicular to the direction of propagation. The power density carried by the plane
wave fills the volume with electromagnetic energy at the speed of light.
238 Chapter 5 Electromagnetic Plane Waves

For a differential volume ∆ ν , the total stored energy is (we + wm ) ∆ v = ∆ W . Hence

 
∆r
S · k̂ A = (we + wm ) ∆ v . (5.31)
ue
Computing the energy velocity from Eq. (5.31), we have
S
ue = . (5.32)
we + wm
For a uniform plane wave,

|E0 |2 /η 1 1
ue = 1 1
= = √ ,
2
2
ε |E0 | + 2 µ |H0 |2 ηε µε

indicating that the energy flows at the speed of light.

5-2 Polarization of Plane Waves

Let us consider a uniform plane wave in a lossless medium. In such a medium, the plane wave
can be fully characterized by the following five quantities:
1. Frequency (ω = 2π f ).
√
2. Propagation constant (k = ω µε ).
3. Direction of propagation (k̂i ).
p
4. Field intensity (|E0 | = E0 · E∗0 ).
5. Polarization (p̂ = E0 /|E0 |; a complex unit vector).
The time-domain expression for this plane wave can be expressed as
h i
E(r,t) = Re E0 e−i[ω t−k(k̂i ·r)] ,

where r is the position vector. The polarization of a plane wave can be described as the time-
varying behavior of the electric field vector at a given point in space. In the most general case,
as will be shown in this section, the tip of the electric field traces an ellipse with an angular
frequency ω , and the wave is called elliptically polarized. Without loss of generality, let us
assume that the direction of propagation is along the z-axis in a Cartesian coordinate system.
Since in a source-free region k̂i · E0 = 0, the electric field vector can be written as

E0 = E1 x̂ + E2 ŷ , (5.33)

where E1 and E2 are, in general, complex quantities. Hence, the polarization vector is given
by
E1 E2
p̂ = p x̂ + p ŷ . (5.34)
|E1 |2 + |E2 |2 |E1 |2 + |E2 |2
5-2 Polarization of Plane Waves 239

Since the reference of the time frame is arbitrary, E1 can be chosen as a real quantity and the
ratio of the electric field components can be expressed by
E2
= α e jδ , (5.35)
E1
where both α and δ are real quantities (α ≥ 0, |δ | ≤ π ). The time-varying expression for the
electric field can now be written in the following form:

E(z,t) = Ex x̂ + Ey ŷ = E1 {cos(ω t − kz) x̂ + α cos(ω t − kz + δ ) ŷ} .

By eliminating t from the expressions for Ex and Ey , the trace of the electric field vector can
be obtained. Noting that

cos(ω t − kz + δ ) = cos(ω t − kz) cos δ − sin(ω t − kz) sin δ ,

Ex
cos(ω t − kz) = ,
E1
and
s  2
Ex
sin(ω t − kz) = 1− ,
E1

it can easily be shown that

 2    2
Ex cos δ Ex Ey 1 Ey
−2 · + 2 = sin2 δ . (5.36)
E1 α E1 E1 α E1

Equation (5.36) represents an ellipse, as shown in Fig. 5-4. Note that when δ > 0 the tip of
the electric field moves clockwise on the ellipse as time proceeds. This wave, by definition,
is a left-hand elliptically polarized wave. Conversely, when δ < 0 the tip of the electric field
moves counterclockwise and the wave is a right-hand elliptically polarized wave. According
to the IEEE definition for handedness of elliptically polarized waves, a wave is called right-
hand polarized when the right-hand fingers of an observer, facing the receding wave, follow
the tip of the electric field, his/her thumb points towards the direction of wave propagation.
The definition of a left-hand polarized wave is analogous.
Two special cases are noteworthy: (1) δ = 0 and (2) δ = ± π2 and α = 1. In the first case
Eq. (5.36) reduces to an equation for a line given by
Ey
=α , (δ = 0 linear polarization)
Ex
for which the handedness is irrelevant. For the second case (δ = ±π /2), the ellipse
degenerates into a circle. The polarization unit vector of the wave is defined as

E Ex x̂ + Ey ŷ x̂ + α eiδ ŷ
p̂ = √ = p = √ ,
E · E∗ |Ex |2 + |Ey |2 1 + α2
240 Chapter 5 Electromagnetic Plane Waves

y
E2
E(x, y, t) A

O x
E1

Figure 5-4: Polarization ellipse for a plane wave having tilt angle ψ and ellipticity angle χ .

with the x-axis as the reference axis.

As shown in Fig. 5-4, the polarization ellipse can be expressed uniquely in terms of two
angles known as the ellipticity angle (χ ) and the tilt angle (ψ ). The ellipticity angle for an
ellipse is defined as
OB 1
χ = tan−1 = tan−1 ,
OA AR
where AR = OA/OB is known as the axial ratio. The ellipticity angle can vary from zero
for linear polarization to 45◦ for circular polarization. The tilt angle ψ can vary from zero
to 180◦ (0 ≤ ψ ≤ 180◦ ) and the ellipticity angle χ can vary from −45◦ to +45◦ . To
define the handedness of the polarization, the ellipticity angle is given a sign. The ellipticity
angle is defined as negative for right-hand and positive for left-hand elliptical polarizations,
respectively.
The orientation and ellipticity angles can be expressed in terms of α and δ and vice versa.
Assuming that α and δ for a wave are known, ψ and χ can be obtained from

tan(2ψ ) = tan(2γ ) cos δ , sin(2χ ) = sin(2γ ) sin δ , (5.37)

where
γ = tan−1 α .
Conversely, if ψ and χ are known for a wave, α and δ can be obtained from
tan 2χ
tan δ = , (5.38a)
sin 2ψ
cos 2γ = cos 2χ cos 2ψ , (5.38b)
and
5-2 Polarization of Plane Waves 241

Figure 5-5: A Poincaré sphere representing different polarization states of transverse

electromagnetic waves. There is a one-to-one relation between a point on the surface of a
Poincaré sphere and the polarization state of the wave. Points along the equator represent linear
polarizations, the north pole represents left-hand circular polarization (LHCP), and the south pole
represents right-hand circular polarization (RHCP).

α = tan γ . (5.38c)

Equations (5.38a) and (5.38c) show that when χ = +45◦ then δ = 90◦ and γ = 45◦ , which
implies that α = 1. Therefore χ = +45◦ is a left-hand circularly polarized wave. Similarly, it
can be shown that χ = −45◦ corresponds to a right-hand circularly polarized wave.
Since the polarization of a wave is fully characterized by two independent angles, the
polarization state of a wave can be uniquely represented by a point on the surface of a sphere,
as shown in Fig. 5-5. This sphere is known as the Poincaré sphere. Simple inspections reveal
that the linear polarizations are mapped along the equator of the Poincaré sphere and the left-
hand polarizations are mapped onto the upper hemisphere, whereas the right-hand polarized
waves are mapped on the surface of the lower hemisphere. The left-hand and right-hand
circular polarizations are, respectively, mapped on the north and south poles of the Poincaré
sphere.

5-2.1 Plane Waves in Anisotropic Media

As mentioned earlier, in certain materials, the directions of the electric field intensity and
the electric flux density differ due to some intrinsic material behavior known as electric
242 Chapter 5 Electromagnetic Plane Waves

anisotropy. In this case the electric flux density is related to the electric field intensity by
a dielectric permittivity tensor ε0 ε r :

D = ε0 ε r E . (5.39)

Similarly for a magnetically anisotropic medium, the magnetic flux density and the magnetic
field density are related by
B = µ0 µ r H . (5.40)
In this section we are interested in obtaining the possible plane-wave solutions for a source-
free homogeneous anisotropic medium. Let us begin with a medium that is only electrically
anisotropic whose relative permittivity tensor in Cartesian coordinates is given by
 
εxx εxy εxz
ε r = εyx εyy εyz  . (5.41)
εzx εzy εzz

Starting from Maxwell’s equations and following the same procedure that led to Eq. (5.2), we
can easily show that
∇ × ∇ × E − k2ε r E = 0 , (5.42a)
where k2 = ω 2 ε0 µ0 µr with µr being the relative permeability of the medium. A similar wave
equation for the magnetic field can also be obtained:
h −1 i
∇ × ε r ∇ × H − k2 H = 0 . (5.42b)

We are seeking a solution of the following form:

E = E0 ei K·r (5.43)

such that it satisfies Eq. (5.42a). To facilitate a tractable solution, let us consider the curl (∇×)
operator in the Cartesian coordinate system:
 
 x̂ ŷ ẑ 
 
∂ ∂ ∂ 
∇×E =   

∂x ∂y ∂z
E E E 
x y z
     
∂ Ez ∂ Ey ∂ Ez ∂ Ex ∂ Ey ∂ Ex
= − x̂ − − ŷ + − ẑ ,
∂y ∂z ∂x ∂z ∂x ∂y
5-2 Polarization of Plane Waves 243

which can be reconfigured in the form

 
∂ ∂
 0 −∂z ∂y  
  Ex
 ∂ ∂ 

∇×E =  0 −  Ey  . (5.44)
 ∂z ∂x Ez
 ∂ ∂ 
− 0
∂y ∂x
Now it is interesting to note that the ∇ × ∇× operator can easily be represented as the product
of two identical matrices, which yields
  2  
∂ ∂2 ∂2 ∂2
− ∂ z2 + ∂ y2 ∂x ∂y ∂x ∂z 
  2  
 2 2 2 
 ∂ ∂ ∂ ∂ 
∇ × ∇× =  − + . (5.45)
 ∂y ∂x ∂ x2 ∂ z2 ∂y ∂z 
  2 
 ∂2 ∂2 ∂ ∂2 
− +
∂z ∂x ∂z ∂y ∂ y2 ∂ x2
Substituting Eq. (5.45) into Eq. (5.42a) and using Eq. (5.43), the differential equation given
by Eq. (5.41) reduces to the following linear equation:
 2  
Ky + Kz2 − k2 εxx − (KyKx + k2 εxy ) −(Kx Kz + k2 εxz ) Ex
−(Ky Kx + k2 εyx ) Kx2 + Kz2 − k2 εyy −(Ky Kz + k2 εyz ) Ey  = 0 , (5.46)
−(Kx Kz + k2 εzx ) − (KyKz + k2 εzy ) Kx2 + Ky2 − k2 εzz Ez

where as before K = Kx x̂ + Ky ŷ + Kz ẑ. To obtain a nontrivial solution for Eq. (5.46), the
determinant of the dispersion matrix must vanish. This will provide a polynomial equation
for Kx , Ky , and Kz whose zeros provide relationships for Kz in terms of Kx and Ky .
Using Eq. (5.19b), the wave vector K = k k̂ can be written as

K = kn(α x̂ + β ŷ + γ ẑ) , (5.47)

with

α = sin θ cos φ ,
β = sin θ sin φ ,
γ = cos θ ,
√
where n is the index of refraction for a given mode, k = ω µ0 ε0 , and θ and φ denote
the spherical coordinates of a unit vector along the wave vector K in the (x, y, z) coordinate
system. The determinant of Eq. (5.46) vanishes when the following equation is satisfied:

An4 + Bn2 + |ε r | = 0 , (5.48)

244 Chapter 5 Electromagnetic Plane Waves

where |ε r | is the determinant of the permittivity tensor, and

A = εxx α 2 + εyy β 2 + εzz γ 2 + (εxy + εyx )αβ + (εxz + εzx )αγ + (εyz + εzy )β γ (5.49a)
and
B = (εxx εyy − εxy εyz )(γ 2 − 1) + (εxx εzz − εxz εzx )(β 2 − 1) + (εyy εzz − εyz εzy )(α 2 − 1)
+ [εyx εzy + εxy εyz − εyy (εxz + εzx)] αγ
+ [εxz εzy + εyz εzx − εzz (εxy + εyx )] αβ
+ [εyz εxz + εzx εxy − εxx (εyz + εzy )] β γ . (5.49b)

Since the determinant of Eq. (5.46) is zero, two components of the electric field can be
obtained in terms of the third component. Denoting the entries of the dispersion matrix by
di j with i = 1, 2, 3 and j = 1, 2, 3, from Eq. (5.46) we have
    
d11 d12 Ex d
= − 13 Ez ,
d21 d22 Ey d23

or equivalently,     
Ex −Ez d22 −d12 d13
= . (5.50)
Ey d11 d22 − d12 d21 −d21 d11 d23
Hence,
d12 d23 − d22 d13
Ex = Ez (5.51a)
d11 d22 − d12 d21
and
d21 d13 − d11 d23
Ey = Ez . (5.51b)
d11 d22 − d12 d21
Now Eq. (5.48) can be solved for n, with two possible solutions:
 q 1/2
B− B2 − 4A|ε r |
n1 =   (5.52a)
2A

and
 q 1/2
B+ B2 − 4A|ε r |
n2 =   . (5.52b)
2A

Note that we are eliminating two other solutions for which the real part of n is negative. Such
a medium is referred to as a double refracting medium, which can support two different types
of waves simultaneously. The electric field in such a medium can be written as

E = E1 eikn1 (α x+β y+γ z) + E2 eikn2 (α x+β y+γ z) .

5-2 Polarization of Plane Waves 245

The corresponding magnetic field is given by

1

H= K1 × E1 ei K1 ·r + K2 × E2 ei K2 ·r , (5.53)
ω µ0 µr
where K1,2 = kn1,2 (α x̂ + β ŷ + γ ẑ).
The case of a magnetically anisotropic medium can be studied using a similar procedure,
assuming the permittivity is scalar (ε = ε0 εr ) and the permeability is a tensor µ = µ0 µ r with
 
µxx µxy µxz
µ r = µyx µyy µyz  . (5.54)
µzx µzy µzz

By applying the duality to Eqs. (5.42a) and (5.42a), we obtain

∇ × ∇ × H − k2 µ r H = 0 (5.55a)
and
−1
∇ × (µ r ∇ × E) − k2 E = 0 , (5.55b)

where k2 = ω 2 µ0 ε0 εr . The dispersion matrix can be obtained from Eq. (5.46) by replacing
εi j with µi j . By applying duality, the rest of the problem treatment is similar to what was
presented for media with permittivity tensors.

5-2.2 Special Case: Homogeneous Isotropic Medium

In an isotropic medium, ε r = εr I, where I is the identity matrix. Even for this simpler case,
calculating the determinant of Eq. (5.46) and simplifying the result can be cumbersome. To
simplify the process, an approach based on eigenanalysis can be pursued. As shown earlier in
Eq. (5.44) for a plane-wave solution, ∇ × E can be represented in terms of an antisymmetric
matrix C given by  
0 −iKz iKy
C = −iKz 0 −iKx  .
−iKy iKx 0
−1
This matrix can be diagonalized and written as C = Q Λ Q , where Q is the matrix of
eigenvectors and Λ is a diagonal matrix whose entries are the eigenvalues of C. It can be
shown that  
0 q 0 0
 
Λ = 0 − Kx2 + Ky2 + Kz2 0 . (5.56)
 q 
0 0 Kx2 + Ky2 + Kz2
246 Chapter 5 Electromagnetic Plane Waves

It was also shown that ∇ × ∇ × E can be expressed in terms of a matrix equation whose
coefficient is given by C ·C. But
−1 −1 (2) −1
C ·C = Q Λ Q Q Λ Q =QΛ Q , (5.57)
(2)
where Λ is a diagonal matrix whose entries are simply the square of the eigenvalues of C.
Also, noting that
−1
ε r = εr I = εr Q I Q ,
the operator (∇ × ∇ × −ω 2 µε I) for the plane-wave solution can be expressed as
 
−ω 2 µε 0 0 −1
Q 0 −ω 2 µε + Kx2 + Ky2 + Kz2 0 Q , (5.58)
0 0 −ω 2 µε + Kx2 + Ky2 + Kz2

whose determinant is zero if

Kx2 + Ky2 + Kz2 = ω 2 µε ,
which is the expected dispersion relation for the homogeneous isotropic medium.

5-2.3 Special Case: Uniaxial Medium

Another case for which a nontrivial solution for the matrix equation (5.46) can be obtained is
for a uniaxial medium. For such a medium, the permittivity tensor is diagonal with two of its
diagonal elements being equal. Without loss of generality, let us assume that
 
εxx 0 0
ε r =  0 εxx 0  , (5.59)
0 0 εzz

with εxx , εzz . After much algebraic manipulation, the determinant of the matrix given in
Eq. (5.46) is found to be
 
∆ = −k2 (εxx k2 − Kx2 − Ky2 − Kz2 ) εxx (εzz k2 − Kx2 − Ky2 ) − εzz Kz2 .

Setting ∆ = 0, two dispersion relations are obtained. The first solution, given by

Kx2 + Ky2 + Kz2 = ω 2 µε0 εxx , (5.60)

is similar to the dispersion relation for an isotropic medium and the plane-wave solution for
this mode is referred to as an ordinary wave. The second solution provides the following
dispersion relation:
εzz 2
Kx2 + Ky2 + K = ω 2 µε0 εzz (5.61)
εxx z
5-3 kDB Coordinate for Plane Waves in Bianisotropic Media 247

and is known as an extraordinary wave. If the direction of wave propagation in the uniaxial
medium is determined by θ , φ in such a way that

Kx = k sin θ cos φ ,
Ky = k sin θ sin φ ,
Kz = k cos θ ,

then substituting these into Eqs. (5.60) and (5.61), for ordinary and extraordinary waves, we
have the following dispersion relations respectively:

k2 = ω 2 µε0 εxx (5.62a)

and
 −1
2 2 sin2 θ cos2 θ
k = ω µε0 + . (5.62b)
εzz εxx

5-3 kDB Coordinate for Plane Waves in Bianisotropic Media

In the preceding sections it was shown that electric and magnetic fields of the form of
E0 ei K·r and H0 ei K·r represent plane-wave solutions to source-free Maxwell’s equations in
a homogeneous unbounded medium. Substituting these solutions into Maxwell’s equations, it
can easily be shown that

K × E0 = ω B0 , (5.63a)
K × H0 = − ω D 0 , (5.63b)
K · D0 = 0 , (5.63c)
K · B0 = 0 , (5.63d)

where it was assumed that the electric and magnetic flux densities can also be expressed as

B = B0 ei K·r
and
D = D0 ei K·r .

As indicated by Eq. (5.63c) and Eq. (5.63d), it is obvious that D0 and B0 are perpendicular to
the K vector. The plane that contains D0 and B0 and is perpendicular to K will be referred to
as the DB plane. For isotropic media where D0 = ε E0 and B0 = µ H0 , the electric field E0
and magnetic field H0 are also perpendicular to K. However, for anisotropic media where, for
example, D0 = ε E0 , the electric field is no longer perpendicular to K.
Considering a coordinate system where one of the axes is along K, say the z′ -axis, then
B0 and D0 in an anisotropic medium would only have x′ and y′ components. Hence

D0 = D0x x̂′ + D0y ŷ′

248 Chapter 5 Electromagnetic Plane Waves

and
B0 = B0x x̂′ + B0y ŷ′ .

In this coordinate system let us denote the bianisotropic constitutive relations as

E =Γ D+ϑ B (5.64a)
and
H = χ D+ν B , (5.64b)

where Γ and ν are, for example, the impermittivity and impermeability tensors. Using
Eqs. (5.64a) and (5.64b) in Eqs. (5.63a) and (5.63b) and noting that K = k ẑ′ , it can be shown
that
ω
E0x ŷ′ − E0y x̂′ = B0 , (5.65a)
k
ω
H0x ŷ′ − H0y x̂′ = − D0 , (5.65b)
k
and
ω 
B0x = −E0y 

k 


= −[γ21 D0x + γ22 D0y + ϑ21 B0x + ϑ22 B0y ] ,
ω (5.66a)
B0y = E0x 

k 


= γ11 D0x + γ12 D0y + ϑ11 B0x + ϑ12 B0y ,
ω 
D0x = H0y 

k 


= χ21 D0x + χ22 D0y + ν21 B0x + ν22 B0y ] ,
ω (5.66b)
D0y = −H0x 

k 


= −[χ11 D0x + χ12 D0y + ν11 B0x + ν12 B0y ] .

Denoting ωk = u as a phase velocity and rearranging Eqs. (5.66a) and (5.66b) in matrix
notation leads to
     
u + ϑ21 ϑ22 B0x −γ21 −γ22 D0x
= (5.67a)
−ϑ11 u − ϑ12 B0y γ11 γ12 D0y
and
     
u − χ21 −χ22 D0x ν21 ν22 B0x
= . (5.67b)
χ11 u + χ12 D0y −ν11 −ν12 B0y

The dispersion relation, an equation for k in terms of the medium parameters, can be obtained
by first finding D0x and D0y in terms of B0x and B0y from Eq. (5.67a) and substituting it in
Eq. (5.67b). This way, a (2 × 2) matrix coefficient of B0x and B0y is obtained. For a nontrivial
solution, the determinant of this (2 × 2) matrix must be set to zero and this will provide the
5-3 kDB Coordinate for Plane Waves in Bianisotropic Media 249

dispersion equation for k. As was the case before in Eq. (5.46), finding the solution is not
trivial. However, in the kDB system approach we need to find the determinant of a (2 × 2)
matrix as opposed to a (3 × 3) matrix.
Let us first consider the simple case of an isotropic medium. For this problem, the
constitutive relations, Eqs. (5.64a) and (5.64b), take the following form:

E=γ D
and
H=ν B.

Also, Eqs. (5.67a) and (5.67b) simplify to

     
u 0 B0x 0 −γ D0x
= (5.68a)
0 u B0y γ 0 D0y
and
     
u0 D0x 0 ν B0x
= . (5.68b)
0u D0y −ν 0 B0y

Solving for D from Eq. (5.68b) and substituting in Eq. (5.68a) leads to
     
B0x 0 −γ 0 ν /u B0x
u =
B0y γ 0 −ν /u 0 B0y
  
γν /u 0 B0x
= .
0 γν /u B0y

In compact form,   
u − γν /u 0 B0x
=0. (5.69)
0 u − γν /u B0y
The determinant of Eq. (5.69) is zero if and only if
1
u2 = γν = ,
εµ
or simply √
k=ω µε
(recall that u = ω /k). Next we consider a uniaxial medium with constitutive relations

E=Γ D (5.70a)
and
H=ν B, (5.70b)
250 Chapter 5 Electromagnetic Plane Waves

where the impermittivity matrix takes the following form:

 
γ 0 0
Γ = 0 γ 0  . (5.71)
0 0 γz

In this coordinate system, suppose that the k̂ vector is along an arbitrary direction defined by
angles θ and φ :
k̂ = sin θ cos φ x̂ + sin θ sin φ ŷ + cos θ ẑ .
Let us choose a kDB coordinate system so that k̂ = ẑ′ and x̂′ is parallel to the x–y plane,
namely
k̂ × ẑ
x̂′ = = sin φ x̂ − cos φ ŷ .
|k̂ × ẑ|
Also,
ŷ′ = ẑ′ × x̂′ = cos θ cos φ x̂ + cos θ sin φ ŷ − sin θ ẑ .
It can easily be shown that any vector in the general coordinate system can be transformed to
the kDB system by the transformation matrix T given by

E′ = T E , (5.72)

where  
sin φ − cos φ 0
T = cos θ cos φ cos θ sin φ − sin θ  . (5.73)
sin θ cos φ sin θ sin φ cos θ
Conversely, a vector in the kDB coordinate system can be transformed back to the general
coordinate system by
−1
D=T D′ , (5.74)
where, since the transformation is unitary,
 
−1 t
sin φ cos θ cos φ sin θ cos φ
T = T = − cos φ cos θ sin φ sin θ sin φ  . (5.75)
0 − sin θ cos θ

Using Eqs. (5.72), (5.70), and (5.75), it can be shown that

−1
E′ = T Γ T D′ ,
5-3 kDB Coordinate for Plane Waves in Bianisotropic Media 251

and therefore
′ −1
Γ =T Γ T
 
γxx 0 0
=  0 γxx cos2 θ + γzz sin2 θ (γxx − γzz ) sin θ cos θ  . (5.76)
0 (γxx − γzz ) sin θ cos θ γxx sin2 θ + γzz cos2 θ

Using Eq. (5.76) in Eq. (5.67a) provides

     
u0 B0x 0 −γxx cos2 θ − γzz sin2 θ D0x
= , (5.77)
0u B0y γxx 0 D0y

and using Eq. (5.76) in Eq. (5.67b) gives

     
u 0 D0x 0 ν B0x
= . (5.78)
0 u D0y −ν 0 B0y

Solving for D from Eq. (5.78) and substituting that in Eq. (5.77) results in
    
2 B0x ν (γxx cos2 θ + γzz sin2 θ ) 0 B0x
u =
B0y 0 νγxx B0y
or   
u2 − ν (γxx cos2 θ + γzz sin2 θ ) 0 B0x
2 =0. (5.79)
0 u − νγxx B0y
A nontrivial solution for Eq. (5.79) can be obtained when either

u2 = νγxx k2 = ω 2 µεxx (5.80)

(recognizing that k2 = Kx2 + Ky2 + Kz2 ) or

u2 = ν (γxx cos2 θ + γzz sin2 θ ) , (5.81)

which can be written as

εzz
ω 2 µεzz = k2 cos2 θ + k2 sin2 θ . (5.82)
εxx
Also, noting that
k2 sin2 θ = k2 sin2 θ sin2 φ + k2 sin2 θ cos2 φ ,
Eq. (5.82) can be written as
εzz 2
Kx2 + Ky2 + K = ω 2 µεzz , (5.83)
εxx z
which is in agreement with Eq. (5.61), noting that here εxx and εzz are the permittivity values
and not the relative permittivity values. Equation (5.80) corresponds to a nontrivial solution
for Eq. (5.79) for which B0x = 0 and B0y , 0, and since the resulting dispersion relation
252 Chapter 5 Electromagnetic Plane Waves

corresponds to that of a plane wave in an isotropic medium with permittivity ε = εxx , it is

referred to as an ordinary wave. On the other hand, the dispersion relation Eq. (5.83) is a
solution of Eq. (5.79) for which B0x , 0 and B0y = 0. From Eq. (5.79) we also note that

ν k
D0x = B0y = B0y (5.84a)
u ωµ
and
ν k
D0y = − B0x = − B0x . (5.84b)
u ωµ
Hence for an ordinary wave, D0y = 0, and for an extraordinary wave, D0x = 0. For ordinary
waves
1
E0x = D0x
εxx
and
1
H0y = B0y .
µ
That is, the electric field is parallel to the electric flux density and the electric field is
perpendicular to the magnetic field intensity, and both are perpendicular to the direction of
propagation. The wave impedance is defined as
r
E0x µ D0x µ k µ
ηo = = = = .
H0y εxx B0y εxx ω µ εxx
For extraordinary waves, using Eq. (5.76), it can be shown that
 2 
cos θ sin2 θ
E0y = + D0y (5.85a)
εxx εzz
and
εzz − εxx
E0z = sin θ cos θ D0y . (5.85b)
εxx εzz
Also, since B0y = 0,
1
H0x = B0x .
µ
As shown by Eq. (5.85b), the electric field has a component along the direction of propagation
so long as θ , 0 or θ , π /2. It should be noted that θ = 0 corresponds to a case where the wave
is propagating along the optical axis (z-axis). In such a case, the dispersion relation Eq. (5.82)
reduces to that for the ordinary wave (k2 = ω 2 µεxx ), or when θ = π /2, Kx2 + Ky2 + Kz2 =
ω 2 µεxx , the direction of propagation is perpendicular to the optical axis (propagation along
±x-axis) and the dispersion relation simplifies to

k2 = ω 2 µεzz . (5.86)
5-3 kDB Coordinate for Plane Waves in Bianisotropic Media 253

In this case, E is also parallel to D:

1
E0y = D0y , (5.87)
εzz
and like ordinary waves, the electric field is perpendicular to the magnetic field and both are
perpendicular to the direction of propagation (±x-axis). The wave impedance is given by
√
E0y 1 k µεxx
ηe (θ = π /2) = − = = . (5.88)
H0x εzz ω εzz
Figure 5-6 shows the direction of propagation, the electric and magnetic flux densities as well
as the electric and magnetic fields associated with ordinary and extraordinary waves.
As mentioned earlier, the electric field of an extraordinary wave has a component along
the direction of propagation. This indicates that the direction of flow of energy and wave
propagation are not the same. The complex Poynting vector can be calculated from
1 1
S= E × H∗ = (E0y ŷ′ + E0z ẑ′ ) × H0x∗ ′
x̂
2 2
1
= (E0z B∗0x ŷ′ − E0y B∗0x ẑ′ ) . (5.89)
2µ

z z
z′ z′

k̂ x k̂ x

x′ x′

(a) Ordinary wave (b) Extraordinary wave

Figure 5-6: Configurations of field orientations in a uniaxial dielectric medium with optical axis
along ẑ: (a) the configuration for an ordinary wave where the direction of the electric flux density
and electric field intensity are parallel to each other and perpendicular to the plane formed by
the optical axis and the direction of propagation, and (b) the configuration for an extraordinary
wave where the electric field intensity and the electric flux intensity lie in the plane formed by the
optical axis and the direction of propagation but are no longer parallel to each other.
254 Chapter 5 Electromagnetic Plane Waves

p
Noting that B0x = − µ /εxx D0y , after some algebraic manipulations it can be shown that the
Poynting vector is given by
s  2  
1 1 cos θ sin2 θ εxx − εzz
S= + ẑ′ + sin θ cos θ ŷ′ |D0y |2 . (5.90)
2 µεxx εxx εzz εxx εzz

Here we assumed that the medium is lossless. It is obvious that the Poynting vector has a
component along ŷ′ , in addition to a component along ẑ′ (direction of wave propagation).
When θ = π /2 the Poynting vector has only the ẑ′ component. In this case, when both D0x and
D0y are nonzero, then both ordinary and extraordinary waves can coexist. In this case E0z = 0
for both ordinary and extraordinary waves and the medium is called birefringent, since the
phase velocities of the ordinary and extraordinary waves are different. The phase velocity for
the ordinary waves can be obtained from
ω ω 1
uop = = √ = √ , (5.91a)
k ω µεxx µεxx

and for the extraordinary wave, Eq. (5.82) leads to

  2 1/2
ω 1 cos θ sin2 θ
uep = = + . (5.91b)
k µ εxx εzz

For θ = π /2, the phase velocity is

1
uep = √ . (5.91c)
µεzz
The electric field propagating in the medium is given by
! !
D0x ω ′ D0y ω ′
E(r) = exp i o z + exp i e z . (5.92)
εxx up εzz up

5-4 Transverse Electric (TE) and Transverse Magnetic (TM)

Field Solutions of the Helmholtz Equation
In the previous section, a special solution of the Helmholtz equation was obtained. In that
solution, known as the plane-wave solution, the electric and magnetic fields of elementary
waves were found to be orthogonal to each other and to the direction of propagation. That
is, both the electric field and magnetic field are in the transverse plane with respect to the
direction of propagation, and thus this type of solution is referred to as a TEM wave. This
is just one form of many types (modes) of solutions that can satisfy the Helmholtz equation.
For example, a class of solutions can be constructed from elementary wave functions where
only the electric field or only the magnetic field is in the transverse plane with respect to an
arbitrary direction of propagation.
5-4 TE and TM Field Solutions of the Helmholtz Equation 255

In a source-free region, it was shown that the electric and magnetic fields can be obtained
from the electric and magnetic Hertz potentials, which satisfy Helmholtz equations

∇2Π + k2 Π = 0 (5.93a)
and
∇2Π m + k2 Π m = 0 , (5.93b)

as well as

Π + k2 Π + iω µ ∇ × Πm
E = ∇∇ ·Π (5.94a)
and
Πm + k 2 Π m .
H = −iωε ∇ × Π + ∇∇ ·Π (5.94b)

The TE and TM solutions can be constructed from magnetic and electric Hertz potentials,
respectively. Again, to obtain a simple solution we assume the Hertz potential (electric or
magnetic) has only one component along the direction of propagation (perpendicular to the
transverse plane). Without loss of generality let’s assume that the x–y plane is the transverse
plane. Hence TM modes are generated by an electric Hertz potential

Π = Πz ẑ (5.95)

and Πm = 0. Substituting Eq. (5.95) into Eq. (5.94a) gives

 
∂ Πz
E=∇ + k2 Πz ẑ . (5.96)
∂z
Equation (5.96) can be expanded by noting that the ∇ operator may be written as

∂
∇ = ∇t + ẑ ,
∂z
where ∇t is a two-dimensional “del” operator in the transverse plane given by

∂ ∂
∇t = x̂ + ŷ .
∂x ∂y
Hence Eq. (5.96) can be written as
 2 
∂ ∂ Πz 2
E= ∇t Π z + + k Πz ẑ . (5.97)
∂z ∂ z2
256 Chapter 5 Electromagnetic Plane Waves

Similarly, Eq. (5.94b) can now be expressed as

H = −iωε ∇ × (Πz ẑ)

 
∂
= −iωε ∇t + ẑ Πz × ẑ
∂z
= −iωε ∇t Πz × ẑ . (5.98)

According to Eq. (5.93a), Πz must satisfy the scalar wave equation

∇2 Π z + k 2 Π z = 0 ,

which can also be written as

∂2
∇2t Πz + Πz + k 2 Πz = 0 . (5.99)
∂ z2
Since we chose z as the direction of propagation, we are seeking a solution for Πz of the
following form:
Πz (r) = Ψ(x, y) e±iβ z . (5.100)
where β is the propagation constant along the z-axis. Substituting Eq. (5.100) into Eq. (5.99),
we get
∇2t Ψ + kc2 Ψ = 0 , (5.101)
where kc2 = k2 − β 2 . Equation (5.101) is a two-dimensional wave equation whose solution can
be expressed in terms of a wide range of elementary wave functions. For the time being, we
assume that appropriate wave functions satisfying Eq. (5.101) are known. Using Eq. (5.100)
in Eqs. (5.97) and (5.98) yields

E = (±iβ ∇t Ψ + kc2 Ψ ẑ)e±iβ z (5.102a)

and (TM wave)
H = −iωε (∇t Ψ × ẑ)e±iβ z . (5.102b)

Note that H does not have a z-component, and therefore field solutions for Eqs. (5.102a)
and (5.102b) are known as TM waves. The ratio of the electric field to the magnetic field
components in the transverse plane is independent of position and is known as the TM wave
impedance, given by
ẑ × Et β
ZTM = = . (5.103)
Ht ωε

Another set of modes can be generated from a magnetic Hertz potential Πm = Πmz ẑ. These
modes will be referred to as TE modes. As shown before, the set of fields obtained from a
magnetic potential is the dual of those obtained from the dual electric Hertz potential. Thus
5-4 TE and TM Field Solutions of the Helmholtz Equation 257

we define
Πmz (r) = Ψm (x, y) eiβ z , (5.104)
whose corresponding fields, using the duality relations, are given by

H = (iβ ∇t Ψm + kc2 Ψm ẑ)eiβ z (5.105a)

and (TE wave)
E = iω µ (∇t Ψm × ẑ)eiβ z (5.105b)

According to the duality relations, the wave admittance, obtained from Eq. (5.103), is given
by
ẑ × Ht β
YTE = − = . (5.106)
Et ωµ

5-4.1 TM Plane-Wave Reflection and Transmission at Planar Interface

Between Two Dielectric Media
In this section we consider inhomogeneous media that may be composed of different dielectric
or metallic bodies. We begin by first considering simple media whose constitutive parameters
are functions of only a single position variable.
As a first example, let’s consider an inhomogeneous medium composed of two
homogeneous half-space dielectric media with a planar interface, as shown in Fig. 5-7. In
this case the permittivity and permeability of the overall medium are only functions of z.
Assuming no field variations along the y-direction (∂ /∂ y = 0), for a TM wave in the upper
half-space Eq. (5.101) simplifies to
 2 
∂ 2
+ k1c Ψ1 = 0 , (5.107a)
∂ x2

μ1, ε1 y
x
μ2, ε2

Figure 5-7: Wave solution for a half-space dielectric medium.

258 Chapter 5 Electromagnetic Plane Waves

and for the lower half-space  

∂2 2
+ k2c Ψ2 = 0 , (5.107b)
∂ x2
where k1c2 = k2 − β 2 and k2 = k2 − β 2 . The possible solutions for Eqs. (5.107a) and (5.107b)
1 1 2c 2 2
are of the form e±ikc x . Assuming propagation in +x-direction and allowing propagation in the
±z-directions, for each half-space, Πz (r) can be obtained from Eq. (5.100):

Π1 (z) = (A1 e−iβ1 z + B1 e+iβ1 z )eik1c x

and
Π2 (z) = (A2 e−iβ2 z + B2 e+iβ2 z )eik2c x .

The field components can be obtained from Eqs. (5.102a) and (5.102b), and are given by
 
−iβ1 z B1 iβ1 z ik1c x
E1x = A1 β1 k1c e − e e , (5.108a)
A1
 
2 −iβ1 z B1 iβ1 z ik1c x
E1z = k1c A1 e + e e , (5.108b)
A1
 
−iβ1 z B1 iβ1 z ik1c x
H1y = −ωε1 k1c A1 e + e e . (5.108c)
A1
The fields in the second medium can be obtained by changing the subscript 1 to 2. Requiring
the continuity of tangential electric and magnetic fields at z = 0 provides the following
equations:
   
B1 ik1c x B2 ik2c x
A1 β1 k1c 1 − e = A2 β2 k2c 1 − e (5.109a)
A1 A2
and
   
B1 ik1c x B2 ik2c x
ε1 k1c A1 1 + e = ε2 k2c A2 1 + e . (5.109b)
A1 A2
Since Eqs. (5.109a) and (5.109b) must be valid for all x, then

k1c = k2c , (5.110)

which is known as the phase-matching condition. Consequently, Eqs. (5.109) simplify to

   
B1 B2
A 1 β1 1 − = A 2 β2 1 − (5.111a)
A1 A2
and
   
B1 B2
ε1 A 1 1 + = ε2 A 2 1 + . (5.111b)
A1 A2
5-4 TE and TM Field Solutions of the Helmholtz Equation 259

Now let’s assume these field components are generated by an incident wave

Hi (x, z) = ŷH0 eik1 k̂·r = ŷH0 eik1 (− cos θi z+sin θi x)

with H0 = −ωε1 k1c A1 , and

Ei (x, z) = −η1 H0 (sin θi ẑ + cos θi x̂)eik1 (− cos θi z+sin θi x) .

This wave excites a mode corresponding to β1 = k1 cos θi , k1c = k1 sin θi , and

q q
β2 = k2 − k1c = k22 − k12 sin2 θi .
2 2 (5.112)

Since the second medium is semi-infinite, no upward-going wave is excited, i.e., B2 = 0.

Taking the ratio of Eqs. (5.111a) and (5.111b),

β1 1 − B1 /A1 β2
= ,
ε1 1 + B1 /A1 ε2

from which B1 /A1 can be obtained and is given by

B1 ε2 /β2 − ε1 /β1
ΓTM = = . (5.113)
A1 ε2 /β2 + ε1 /β1
Hence the total magnetic field in the upper half-space is given by

H1y = H0 [e−iβ1 z + ΓTMeiβ1 z ]eik1 sin θi x . (5.114)

The quantity ΓTM , which is the ratio of upward-going (reflected) wave to incident wave, is
known as the magnetic reflection coefficient.
Total electric field:
β1
E1x = − H0 (e−iβ1 z − ΓTMeiβ1 z )eik1 sin θi x (5.115a)
ωε1
and
k1 sin θi
E1z = − H0 (e−iβ1 z + ΓTM eiβ1 z )eik1 sin θi x . (5.115b)
ωε1
Noting that β1 /ωε1 = η1 cos θi ,

η2 cos θt − η1 cos θi
ΓTM = − , (TM reflection coefficient) (5.116)
η2 cos θt + η1 cos θi
260 Chapter 5 Electromagnetic Plane Waves

where the transmission angle θt is related to the incidence angle θi by

s
β2 k2
cos θt = = 1 − 12 sin2 θi . (5.117)
k2 k2

To find the fields in the second medium, we note that the lower medium is semi-infinite and
cannot produce a reflected wave, that is, B2 = 0. Using Eq. (5.111b), A2 can be found from
ε1 A 1
A2 = (1 + ΓTM ) .
ε2
Hence
H2y = H0 (1 + ΓTM )e−iβ2 z eik1 sin θi x . (5.118)
The quantity H2y (z = 0)/H1y (z = 0) is called the magnetic transmission coefficient and is
given by
τTM = 1 + ΓTM . (5.119)

5-4.2 TE Plane-Wave Reflection and Transmission at Planar Interface

between Two Dielectric Media
To obtain the TE solution, the duality relations can be used. From Eqs. (5.114), (5.115a),
and (5.115b),

E1y = −E0 (e−iβ1 z + ΓTEeiβ1 z )eik1 sin θi x , (5.120a)

cos θi
H1x = − E0 (e−iβ1 z − ΓTEeiβ1 z )eik1 sin θi x , (5.120b)
η1
sin θi
H1z = − E0 (e−iβ1 z + ΓTEeiβ1 z )eik1 sin θi x , (5.120c)
η1
and from Eq. (5.118)
E2y = −E0 (1 + ΓTE)e−iβ2 z . (5.121)
The TE electric field reflection coefficient ΓTE can also be obtained using the duality relations:

η2 / cos θt − η1 / cos θi
ΓTE = . (TE reflection coefficient) (5.122)
η2 / cos θt + η1 / cos θi

Figure 5-8 shows the magnitude of the reflection coefficient for both TM and TE polarization
versus the incident angle for a case where ε1 = ε0 and ε2 = 4ε0 .
The well-known Snell’s law of refraction can be derived from Eq. (5.110) and is given by

n1 sin θi = n2 sin θt , (5.123)

5-4 TE and TM Field Solutions of the Helmholtz Equation 261

0.9 TM

0.8 TE
Magnitude of reflection coefficient

0.7

0.6

0.5
TE

0.4

0.3
TM
0.2

0.1

0
0 10 20 30 40 50 60 70 80 90
Incident angle (degrees) Brewster angle

Figure 5-8: Magnitude of the plane-wave reflection coefficient at a planar interface between two
dielectric media with ε1 = ε0 and ε2 = 4ε0 .

√ √
where n1 = ε1 µ1 and n2 = ε2 µ2 are respectively the indices of refraction of the upper and
lower media.

5-4.3 Brewster Angle

It is interesting to note that the angle at which the reflection coefficient for the TM wave is
zero can be obtained from Eq. (5.116):
s
η2 η2 k2
cos θi = cos θt = 1 − 12 sin2 θi .
η1 η1 k2

This angle is known as the Brewster angle θBTM , which can be explicitly expressed as

s !
ε2 (µ1 ε2 − µ2 ε1 )
θBTM = sin−1 . (5.124)
µ1 (ε22 − ε12 )
262 Chapter 5 Electromagnetic Plane Waves

In the case where the upper and lower half-spaces are nonmagnetic (µ1 = µ2 = µ0 ),
Eq. (5.124) takes a simpler form:

r 
−1 ε2
θBTM = sin . (nonmagnetized)
ε2 + ε1

The Brewster angle for TE polarization can be obtained from Eq. (5.124) using duality and is
given by

s !
µ2 (ε1 µ2 − ε2 µ1 )
θBTE = sin−1 . (for magnetic materials only) (5.125)
ε1 (µ22 − µ12 )

Note that the Brewster angle for TE polarization does not exist for nonmagnetic materials.
Another interesting observation is the phenomenon of total reflection for lossless media.
In situations where n1 > n2 , there exists a critical angle θi = θc for which θt = π /2 and for
θi > θc then θt is pure imaginary. In this case it is easy to show that the magnitudes of the
reflection coefficients for both polarizations are unity.

Example 5-2: Fermat’s Principle

Using Snell’s law of reflection and refraction, show that Fermat’s principle, also known as
the principle of least time, is valid for a medium composed of two half-space dielectric media
with a planar interface. Consider a point source as the source of radiation and an observation
point that could be in either medium.
Solution: Let us consider two dielectric media with indices of refraction n1 and n2 and a
point source at location S in medium 1 and two observation points, P1 in medium 1 and
P2 in medium 2. The geometry of the problem is shown in Fig. 5-9. As shown earlier,
the field radiated from a source can be expanded in terms of plane waves propagating in
different directions. Each plane wave can be represented by a ray indicating the direction of
propagation. The goal here is to show that the point on the surface responsible for reflecting
an incident ray that goes through P1 or P2 is a point that minimizes the travel time. Let us first
consider the point in medium 1 (P1 ). According to Snell’s law, the point q on the interface
between S and P1 is a point such that

n̂ · (k̂i + k̂r ) = 0 (5.126a)

and
t̂ · (k̂i − k̂r ) = 0 , (5.126b)

where k̂i and k̂r are unit vectors along the S–q and q–P1 lines respectively, n̂ is a unit normal
to the interface, and t̂ is a tangential unit vector to the interface. Now consider another point
5-4 TE and TM Field Solutions of the Helmholtz Equation 263

k̂r

lr
+δ
k̂i li
+
δl lr

lr
li i
n̂

q q′

l t + δl t
k̂t
lt

Figure 5-9: A source at S above a planar interface between two dielectric media having indices of
refraction n1 and n2 . According to Fermat’s principle, the optical path length between the source
and an observation point through reflection or refraction at a point on the interface is stationary
with respect to small variations in the path.

q′ on the interface that could be responsible for directing the reflected incident ray to P1 . We
want to show that the travel time through q is shorter than that for q′ , or q is the stationary
point on the surface. Since the speed of light is moderated by the index of refraction, the
travel time through a distance r is proportional to a parameter known as the optical path
length, defined by nr. Hence the minimum travel time corresponds to the minimum optical
path length. Representing the optical path length from S to q by n1 ℓi and that from q to P1
by n1 ℓr , for a point q′ near q the corresponding optical pathlengths may be expressed by
n1 (ℓi + δ ℓi ) and n1 (ℓr + δ ℓr ). The vector optical path between S and q′ and between q′ and P1
can be written as

n1 (ℓi + δ ℓi)(k\
i + δ ki ) = n1 ℓi k̂i + n1∆ (5.127a)
and
n1 (ℓr + δ ℓr )(k\
r + δ kr ) = n1 ℓr k̂r − n1∆ , (5.127b)

where (k\ \ ′ ′
i + δ ki ) and (kr + δ kr ) are unit vectors along Sq and q P, respectively. Also, ∆ is a
small tangential vector connecting q to q′ . Let us now carry out the dot product of Eq. (5.127a)
by k̂i and the dot product of Eq. (5.127b) by k̂r , and keeping terms only to the first order in δ ,
264 Chapter 5 Electromagnetic Plane Waves

we have

n1 ℓi + n1 δ ℓi = n1 ℓi + n1 ∆ · k̂i (5.128a)
and
n1 ℓr + n1 δ ℓr = n1 ℓr − n1 ∆ · k̂r . (5.128b)

Note that we have used the fact that to the first-order in δ

k̂i · (k\ \
i + δ ki ) = k̂r · (kr + δ kr ) = 1 .

Now, adding Eqs. (5.128a) and (5.128b) results in

(δ ℓi + δ ℓr ) = ∆ · (k̂i − k̂r ) . (5.129)

According to Snell’s law (see Eq. (5.126)) the right-hand side is zero, so the variational path
length difference δ ℓi + δ ℓr = 0. This indicates that the path SqP1 is stationary (corresponding
to the minimum travel time) with respect to small variations in path. This shows that Fermat’s
principle holds true for reflected rays.
Now let us consider point P2 in the lower half-space. The vector optical path between q′
and P2 can be written as

n2 (ℓt + δ ℓt)(k\
t + δ kt ) = n2 ℓt k̂t − n2 ∆ . (5.130)

Then we evaluate the dot product of Eq. (5.130) with k̂t and keep terms only to the first order
in δ :
n2 ℓt + n2 δ ℓt = n2 ℓt − n2 ∆ · kt ,
or simply
∆ · kt .
δ ℓt = −∆ (5.131)
Adding Eqs. (5.128a) and (5.131) gives

δ ℓi + δ ℓt = ∆ · (k̂i − k̂t) .

According to Snell’s law, t̂ · (k̂i − k̂t) = 0, which indicates that

δ ℓi + δ ℓt = 0 .

This indicates that the path SqP2 is also stationary with respect to small variations in the path,
proving Fermat’s principle for the refracted ray.

5-5 Plane-Wave Reflection at the Interface between a Dielectric

Medium and a Good-Conducting Medium

It can easily be shown that the electric field reflection coefficient of a plane wave reflected by
a perfect electric conductor has a magnitude of unity and a phase of 180◦ , independent of the
5-5 Reflection at the Interface between a Dielectric and a Good-Conducting Medium 265

incident angle or polarization. This result can be obtained from Eq. (5.122) by noting that for
a PEC with conductivity σ = ∞,


r v 
µ0 u µ0 
η2 = u
= u    =0. (5.132)
ε2 t σ 
ε0 1 + i 
ωε0 
σ =∞

However, for most practical situations, good conductors have very high, but not infinite
conductivity. Inside such conductors (σ ≫ ωε0 ) the electric and magnetic fields have some
finite values and, depending on whether we are considering TM or TE incidence, the magnetic
or electric fields are given by Eqs. (5.118) or (5.121). The propagation constant in the
conductor is given by
r
√ ω µ0 σ p
k2 = ω µ0 ε2 ≈ (1 + i) = π f µ0 σ (1 + i) , (5.133)
2
and the medium’s characteristic impedance is
r
π f µ0
η2 ≈ (1 − i) . (5.134)
σ
The refracted angle θt can be computed from
s s
k02 sin2 θi ω 2 µ0 ε0
cos θt = 1 − 2
≈ 1− 2 sin2 θi ≈ 1 ,
k2 iω µ0 σ /ω

where we used the defining relation σ ≫ ωε0 for a good conductor. This result indicates that
the transmitted wave propagates almost along
√ the unit normal to the interface. According to
Eq. (5.133), the attenuation constant α = π f µ0 σ , which means that after propagating a
distance of
1
δs = √ , (5.135)
π f µ0 σ
the field intensity reaches e−1 of its value at the interface. This distance is known as the skin
depth of the conductor, which decreases with increasing frequency and conductivity. Using
this definition in Eq. (5.134), the wave impedance of the conductor may be expressed as
1−i
ηc = . (5.136)
δs σ
Note that the resistive part of the wave impedance, Rc = 1/(δs σ ), is the dc resistivity of the
conducting sheet of the material with cross section of 1 m2 and thickness δs . The reactive part
of wave impedance represents an inductive impedance. It is interesting to examine the ratio
of the total tangential electric field to the total tangential magnetic field in the upper medium
at the interface (z = 0). Using Eqs. (5.115a) and (5.114) and using (5.116) for a TM wave, we
266 Chapter 5 Electromagnetic Plane Waves

have
E1x η1 cos θi (1 − ΓTM)
= = η2 cos θt ≈ ηc , (TM wave)
H1y 1 + ΓTM
where ηc is as given by Eq. (5.136). Since this ratio is independent of the incident angle,
it is valid for any arbitrary source above the conductor, because the fields of the source can
be expressed in terms of plane waves propagating in different directions and all these waves
would see the same surface impedance. It should be noted here that the same results hold
for a TE incident wave using Eqs. (5.120a), (5.120b), and (5.122). Also, if the surface is
curved instead of being planar, the incident plane wave at each point on the surface that has
a different incident angle would still experience the same surface impedance as long as the
radii of surface curvature are much larger than the skin depth. Hence, for electromagnetic
boundary value problems, good conductors can be modeled by a surface impedance given by
Eq. (5.136) without the need to examine the fields interior to the conductor.

5-6 Wave Propagation in an Inhomogeneous Medium:

Geometric-Optics Approximation
The problem of wave propagation in inhomogeneous media is of importance for many
practical problems, such as atmospheric wave propagation when the permittivity or index
of refraction of air can vary gently along the path of propagation. If such permittivity
variations as a function of space are slow with a scale much larger than the wavelength, an
approximate solution can be obtained. In an approximate formulation inspired by plane-wave
characteristics, a solution to Maxwell’s equations, known as the Eikonal equation, can be
obtained.
Let us consider a medium with permeability µ0 and inhomogeneous permittivity
ε (r) = ε0 ε1 (r). Inspired by the plane-wave solution in a homogeneous medium, we may
represent the solution to Maxwell’s equations in a source-free region in the following form:

E(r) = E0 (r) eik0 φ (r) (5.137a)

and
H0 (r) ik0 φ (r)
H(r) = e , (5.137b)
η0
√
where k0 = 2π /λ0 = ω µ0 ε0 , and for a lossless medium φ (r) is a real function that gently
varies with r. Note that the unit of H0 (r) is V/m due to the normalization used in Eq. (5.137b).
Applying Faraday’s law to the electric field given by Eq. (5.137a), we have

∇ × [E0 (r) eik0 φ (r) ] = ∇ eik0 φ (r) × E0 (r) + eik0 φ (r) ∇ × E0 (r)
H0 (r) ik0 φ (r)
= ik0 η0 e . (5.138)
η0
Simplifying Eq. (5.138), we get

ik0 ∇φ (r) × E0 (r) + ∇ × E0(r) = ik0 H0 (r) . (5.139)

5-6 Wave Propagation in an Inhomogeneous Medium 267

Also, the modified Ampère’s law using Eq. (5.137b) gives

 
H0 (r) ik0 φ (r) 1 1
∇× e = ∇eik0 φ (r) × H0 (r) + eik0 φ (r) ∇ × H0 (r)
η0 η0 η0
= −iωε0 εr (r) E0 (r) eik0 φ (r) . (5.140)

Upon simplifying Eq. (5.140) and recalling that ωε0 = k0 /η0 , we get

ik0 ∇φ (r) × H0 (r) + ∇ × H0(r) = −ik0 εr (r) E0 (r) . (5.141)

Equations (5.139) and (5.141) must be solved together to find the unknown functions.
For a homogeneous medium, we found k0 φ (r) = K · r (similar to Eq. (5.22)) and E0 and
H0 to be constant vectors independent of position, where K = k0 εr k̂i is the propagation vector
for a plane wave propagating along k̂i . We also found that

∇φ (r) = εr k̂i ,

indicating that K is perpendicular to a surface over which the phase front function is constant
(φ (r) = C).
At high frequencies, assuming that k0 is very large compared with the spatial variations
of E0 (r) and H0 (r), ∇ × E0 (r) and ∇ × H0 (r) can be ignored compared with other terms
in Eqs. (5.139) and (5.141). Therefore, at high frequencies where the wavelength is much
smaller than variations in εr (r), E0 (r), and H0 (r), Eqs. (5.139) and (5.141) can be expressed
as

∇φ (r) × E0 (r) = H0 (r) (5.142a)

and
∇φ (r) × H0 (r) = −εr (r) E0 (r) . (5.142b)

Also, using Gauss’ law of electricity we can show that

h i
∇ · ε0 εr (r) E0 (r) eik0 φ (r) = 0 ,

which can be expanded to show

h i
ε0 εr (r) eik0 φ (r) ∇ · E0 (r) + ε0 ∇εr (r) eik0 φ (r) + ik0 εr (r) ∇φ (r) eik0 φ (r) · E0 (r) = 0 .
(5.143)
Again ignoring other terms compared with terms having k0 as a coefficient, Eq. (5.143) is
simplified to
E0 (r) · ∇φ (r) = 0 . (5.144)
Similarly, Gauss’ law for magnetism provides

H0 (r) · ∇φ (r) = 0 . (5.145)

268 Chapter 5 Electromagnetic Plane Waves

Substituting Eq. (5.142a) into Eq. (5.142b), we get

∇φ (r) × [∇φ (r) × E0 (r)] = −εr (r) E0 (r) . (5.146)

Expanding Eq. (5.146) and using Eq. (5.144), we have

|∇φ (r)|2 = εr (r) = n2 (r) . (Eikonal equation) (5.147)

Equation (5.147) is the famous Eikonal equation that can be solved to find φ (r), which
represents the phase front of the wave in the inhomogeneous medium. There is some degree
of freedom in choosing E0 (r) so long as Eq. (5.144) is satisfied. Once E0 (r) is chosen, H0 (r)
can be obtained from Eq. (5.142a).
The Eikonal equation can be solved to find the ray trajectory in the inhomogeneous
medium. The ray trajectory represented by a parametric function r(s), with parameter s
denoting the arc length, can be used to find the direction of wave propagation k̂ that is tangent
to r(s). Noting that the direction of wave propagation at each point is perpendicular to φ (r),
we can write
dr(s) ∇φ (r)
k̂(r) = = . (5.148)
ds |∇φ (r)|
In view of Eq. (5.148), the Eikonal equation may be written as

∇φ (r) = k̂(r) n(r) . (5.149)

The solution to the Eikonal equation can be used to predict interesting phenomena such as
a mirage, which happens near the interface between a hot ground surface and air with an
increasing index of refraction away from the interface. The atmospheric ducting effect, which
happens at the interface between air and warm water bodies, is another example. The equation
for the ray trajectory in a medium with a slowly varying index of refraction can be derived by
first noting that    
d2r d dr dr
2
= = k̂ · ∇ = k̂ · ∇ k̂ , (5.150)
ds ds ds ds
according to the definition of gradient. Here we are using a unit vector along the ray path
(ŝ = k̂). Using Eq. (5.149) in Eq. (5.150), we have
 
d 2 r ∇φ (r) ∇φ (r)
= ·∇ . (5.151)
ds2 n(r) n(r)
Now consider the vector calculus identity

∇(U · V) = (U · ∇)V + (V · ∇)U + U × (∇ × V) + V × (∇ × U) , (5.152)

and choose U = V to further simplify Eq. (5.152) to

∇|U|2 = 2(U · ∇)U + 2U × (∇ × U) . (5.153)

5-6 Wave Propagation in an Inhomogeneous Medium 269

1
Substituting for U = n ∇φ (r) into Eq. (5.153), we have
 
∇φ (r) ∇φ (r) 1 |∇φ (r)|2 ∇φ (r) ∇φ (r)
·∇ = ∇ − ×∇× . (5.154)
n(r) n(r) 2 n2 n n(r)
Using Eq. (5.147) in the first term of the right-hand side of Eq. (5.154), in view of Eq. (5.151),
it can be shown that
 
d2r 1 ∇φ (r)
=− ∇φ (r) × ∇ ×
ds2 n(r) n(r)
   
1 1
=− ∇φ (r) × ∇ × ∇φ (r)
n(r) n(r)
 
1 2 1 1
=− |∇φ (r)| ∇ − ∇φ (r) · ∇ ∇φ (r) , (5.155)
n(r) n(r) n(r)

where we have used the fact that ∇ × ∇φ (r) = 0. Again using Eq. (5.147) in Eq. (5.155), this
equation can be simplified to
 
d2r ∇φ (r) 1 1
= ·∇ ∇φ (r) − n(r) ∇ . (5.156)
ds2 n(r) n(r) n(r)
Using the definitions given by Eqs. (5.148) and (5.149) in Eq. (5.156), the following
differential equation for the trace of rays in the inhomogeneous medium is obtained:
 
d2r dr dr
n(r) 2 + ∇ n(r) · = ∇ n(r) . (5.157)
ds ds ds
Also, noting that
dn(r) dr
= ∇ n(r) · ŝ = ∇ n(r) · ,
ds ds
Eq. (5.157) can be written in a more compact form as
 
d dr
n(r) = ∇ n(r) . (5.158)
ds ds
Equations (5.157) or (5.158) can be used to find the ray trajectory. For example, if n is
constant, then from Eq. (5.158) we have

d2r
=0,
ds2
which indicates that dr/ds is a constant and this represents a line along a fixed direction k̂.
270 Chapter 5 Electromagnetic Plane Waves

Example 5-3: Derivation of Snell’s Law

Derive Snell’s law from the Eikonal equation.

Solution: Consider a smooth interface between two media. Let’s assume that the local radii
of curvature are very large compared with the wavelength. Taking ∇× of both sides of
Eq. (5.149), we get
∇ × [k̂(r) n(r)] = 0 . (5.159)
For a small rectangular contour between the two media, as shown in Fig. 5-10, we have
I
n k̂(r) · dℓℓ = 0 . (5.160)
C

Letting the length of the sides perpendicular to the interface go to zero, and denoting the angle
between k̂1 and dℓℓ to be θi and the angle between k̂2 and dℓℓ to be θt , as shown in Fig. 5-10,
we have
n1 sin θi = n2 sin θt .

k̂1 z

i k̂1 i x̂ − i ẑ
x
n1
n2 t k̂2 k̂2 t x̂ − t ẑ

Figure 5-10: Proof of Snell’s law from the Eikonal equation. A small rectangular contour
between the two media is chosen to show that n1 sin θi = n2 sin θ2 .

The curl-free property of the ray vector can be used to study how a phase front proceeds as
it propagates in an inhomogeneous medium. Consider an initial phase front surface S1 (φ (r) is
constant on this surface), whose radii of curvature are large compared with the wavelength.
Suppose the rays emitted from this surface encounter the surface of a perfect conductor (Sc )
having a large radius of curvature as well. The reflected rays form a phase front S2 . As shown
in Fig. 5-11, using Eq. (5.160) on a closed contour C = p1 pp2 q2 qq1 p1 , we have
I
n(r) k̂(r) · dℓℓ = 0 .
C
5-6 Wave Propagation in an Inhomogeneous Medium 271

Figure 5-11: Phase-front propagation in an inhomogeneous medium based on the Eikonal

equation showing that the optical paths for all rays between two phase fronts are the same.

On this contour, k̂ is perpendicular to dℓℓ on S1 and S2 , and on the rest is either parallel or
antiparallel. As a result, it can easily be shown that
Z Z
n(r) dℓ = n(r) dℓ .
p1 pp2 q1 qq2

That is, S2 is formed in such a way that the optical path lengths for all rays between the two
phase fronts are the same.

Example 5-4: Bending of Rays in an Inhomogeneous Medium

Consider an inhomogeneous medium with index of refraction n(r) = n0 + α z. Starting with

a ray in the x–z plane having a direction k̂ = sin θ0 x̂ + cos θ0 ẑ at x = z = 0, find the ray
trajectory in the x–z plane using the Eikonal equation.
Solution: The starting point is Eq. (5.157), and noting that

∇ n(r) = α ẑ ,

Eq. (5.157) can be written as

 
d2r α dr dr α ẑ
2
+ ẑ · = . (5.161)
ds n(r) ds ds n(r)
272 Chapter 5 Electromagnetic Plane Waves

Finding an exact solution for Eq. (5.161) is not very easy even for this simple linear function
for the index of refraction. However, assuming that α /n0 ≪ 1, we can proceed with an
approximate solution. As shown earlier, when n is a constant the ray trajectory is a line, and
thus when δ = α /n0 ≪ 1, dr/ds may be expanded in terms of its Taylor series expansion
in δ :
∞  
dr X dr (i) i
= δ , (5.162)
ds ds
i=0

where (dr/ds)(i)is the ith-order representation of a ray direction at each point. Also, for all
values of z > 0,
α α
≤ =δ . (5.163)
n0 + α z n0
To the zeroth order in δ ,
d2r
=0
ds2
and
dr
= x′ (s) x̂ + z′ (s) ẑ = k̂ .
ds
If we were to keep the solution to the first order in δ , Eq. (5.161) simplifies to

d2r α ′ α
2
+ z (s) k̂ = ẑ . (5.164)
ds n n
This represents two equations:
α ′
x′′ (s) + z (s) x′ (s) = 0 (5.165a)
n
and
α ′ 2 α
z′′ (s) + [z (s)] = . (5.165b)
n n
Unfortunately this is still a nonlinear coupled differential equation. However, when n0 ≫ α z
(for relatively small values of z) these equations simplify to
α ′
x′′ (s) + z (s) x′ (s) = 0 (5.166a)
n0
and
α ′ 2 α
z′′ (s) + [z (s)] = . (5.166b)
n0 n0
To solve this nonlinear coupled differential equation for small values of z, x(s) and z(s) may
be expanded in terms of their Taylor series; that is,

x(s) = a1 s2 + a2 s + a3 (5.167a)
and
5-6 Wave Propagation in an Inhomogeneous Medium 273

z(s) = b1 s2 + b2 s + b3 . (5.167b)

But we know that (x(0), y(0)) = (0, 0), which renders the value of unknowns a3 and b3 to be

a3 = b3 = 0 .

We also know that


dr 
= k̂(0) = sin θ0 x̂ + cos θ0 ŷ
ds s=0
= a2 x̂ + b2 ŷ ,

and therefore

a2 = sin θ0
and
b2 = cos θ0 .

Substituting Eq. (5.167b) into Eq. (5.166b), we can find an expression for b1 :
α α
2b1 + (2b1 s + cos θ0 )2 = .
n0 n0
It is expected that z′′ (s) be small (proportional to α /n0 ); that is, b1 ≪ 1 and α /n0 ≪ 1. Hence,
α α
b1 ≈ (1 − cos2 θ0 ) = sin2 θ0 . (5.168)
2n0 2n0
Now using Eqs. (5.167a) and (5.168) in Eq. (5.166a), we have
 
α αs 2
2a1 + sin θ0 + cos θ0 (2a1 s + sin θ0 ) = 0 . (5.169)
n0 n0

Also, a1 is expected to be proportional to α /n0 , and therefore Eq. (5.169) to the first order in
α /n0 can be written as
α
2a1 + cos θ0 sin θ0 = 0 , (5.170)
n0
or
α
a1 = − cos θ0 sin θ0 . (5.171)
2n0
Finally, the ray trajectory can be obtained from

α s2
x(s) = − cos θ0 sin θ0 + s sin θ0 (5.172a)
2n0
and
274 Chapter 5 Electromagnetic Plane Waves

α s2 2
z(s) = sin θ0 + s cos θ0 . (5.172b)
2n0

Note that dr/ds = k̂(s) must be a unit vector; that is, we should be able to show that
  q
 dr 
  = (x′ (s))2 + (z′ (s))2 = 1 . (5.173)
 ds 

Using Eqs. (5.172a) and (5.172b) in Eq. (5.173), we can easily show that
  s  2
 dr 
  = 1+ α s2 sin2 θ0
 ds  n0
 
1 α 2 2 2
= 1+ s sin θ0 ≈ 1
2 n0

to the first order in (α /n0 ).

Figure 5-12 shows the ray trajectory for a medium with α /n0 = 0.05 and n0 = 1.01.

2.5
Zeroth-order
First-order
2

1.5
z(s)
1

0.5

0
0 1 2 3 4 5 6 7
x(s)

Figure 5-12: Ray trajectory calculated from the Eikonal equation for a medium with
n(r) = n0 + α z using a perturbation method for a small value of α /n0 = 0.05. Also shown is
the zeroth-order result where the ray trajectory is a line.
5-7 Plane-Wave Reflection and Transmission from a Half-Space Uniaxial Medium 275

5-7 Plane-Wave Reflection and Transmission from a Half-Space

Uniaxial Medium
As discussed in the previous section, uniaxial media can support two different types of waves,
namely ordinary waves and extraordinary waves, depending on the polarization of the electric
field with respect to the optical axis. Basically, when the electric field intensity and flux
density are perpendicular to the plane formed by the optical axis and the direction of wave
propagation, an ordinary wave with a transcendental equation similar to that for a plane wave
in an isotropic medium is generated. If the electric field and flux density are parallel to the
plane of the optical axis and of the wave propagation direction, an extraordinary wave is
generated. Consider two dielectric media with a planar interface in the x–y plane at z = 0. The
z > 0 medium is assumed to be isotropic with permittivity ε0 ε1 and permeability µ0 µ1 . The
lower half-space medium is chosen to be a uniaxial medium with its optical axis along the x
direction. The permittivity of the lower half-space is given by
 
εx2 0 0
ε 2 = ε0  0 ε2 0  , (5.174)
0 0 ε2

and its permeability is µ0 µ2 . A plane wave propagating along direction

k̂i = sin θi x̂ − cos θi ẑ

is impinging on the lower medium from above, as shown in Fig. 5-13. The plane wave’s

k̂i z

i ,ε ε
x

ε
2, ε 0 ε
ε

Figure 5-13: Geometry of plane-wave reflection from a uniaxial half-space medium with a
planar interface. Depending on the polarization of the incident wave, an ordinary wave (for
perpendicular polarization) or an extraordinary wave (for parallel polarization) is excited in the
lower half-space.
276 Chapter 5 Electromagnetic Plane Waves

electric and magnetic fields are given by

Ei (r) = E0 eik1 (sin θi x−cos θi z) (5.175a)

and
k̂i × E0 ik0 (sin θi x−cos θi z)
Hi (r) = e . (5.175b)
η1
Depending on the wave polarization (direction of E0 ), ordinary, extraordinary, or both types
of waves are generated in the lower half-space medium. Our goal here is to determine the
reflected and transmitted waves in the upper and lower media. To simplify the problem, we
first assume that E0 is along the y-axis (E0 = E0 ŷ), which excites an ordinary wave. In this
case, the mode of propagation is TE, whose fields can be generated from a magnetic Hertz
potential Π om = Πoj x̂. Note that here we are expanding the fields in terms of TE and TM to
x as opposed to z. This choice lends itself to a proper solution for uniaxial media with the
optical axis along the x-direction. Here the superscript “o” denotes ordinary wave. Since there
is no variation with respect to y, ∂∂y = 0 for all field and potential quantities is zero. The Hertz
vector potential must satisfy the wave equation

∇2 Πox j + k2j Πox j = 0 , (5.176)

where j = 1, 2 refers to the upper and lower media, respectively. The electric and magnetic
fields can be derived directly from Eqs. (3.91b) and (3.91a) by placing Π = 0 and Π m = Πox j x̂.
The explicit expressions for the fields are given by

∂ Πox j
Eoj = iω µ0 µ j ŷ (5.177a)
∂z
and
!
∂ 2 Πox j ∂ 2 Πox j
Hoj = 2 o
+ k j Πx j x̂ + ẑ . (5.177b)
∂ x2 ∂x ∂z

Using Eq. (5.176), Eq. (5.177b) can be simplified to

∂ 2 Πox j ∂ 2 Πox j
Hoj = − x̂ + ẑ . (5.178)
∂ z2 ∂x ∂z
The extraordinary wave mode is excited when a transverse magnetic (TM) wave perpendicular
to the x plane is incident upon the lower half-space. TM to x mode can be generated from an
electric Hertz vector potential that has only one component along the x-axis. That is,

Π e = Πex j x̂ , (5.179)

where the superscript “e” means extraordinary. Here we define the electric Hertz vector
potential slightly differently from before, because the permittivity can assume two different
values in the uniaxial medium:
Hej = −iωε0 ∇ × Π ej . (5.180)
5-7 Plane-Wave Reflection and Transmission from a Half-Space Uniaxial Medium 277

From Faraday’s law,

∇ × Eej = iω µ j Hej = ω 2 ε0 µ0 µ j ∇ × Πej , (5.181)
or equivalently,
∇ × [Eej − k02 µ jΠ ej ] = 0 . (5.182)
Hence, the electric field in terms of Hertz vector potential is given by

Eej = k02 µ jΠ ej − ∇φ j , (5.183)

where φ j is the scalar electric potential. Also, from Faraday’s law in a source-free region, we
have
∇ × Hej = −iω Dej . (5.184)
Upon substitution of Eq. (5.180) into Eq. (5.184), it can easily be shown that

Dej = ε0 ∇ × ∇ × Πej = ε0 [∇∇ ·Π

Πej − ∇2Π ej ] . (5.185)

From the constitutive relation D = ε E, we can write

Dej = ε0 [(εx j − ε j )Exej x̂ + ε j Eej ] . (5.186)

Using Eqs. (5.186) and (5.183) in Eq. (5.184), the following equation can be obtained:

∂φj
∇2Π ej + k02 εx j µ jΠ ej − ∇(∇ ·Π
Πej + ε j φ j ) − (εx j − ε j ) = 0. (5.187)
∂x
By defining ∇ · Π ej = −ε j φ j as a gauge condition, and noting from Eq. (5.179) that Π ej has
only an x component, the following wave equation for Πex j is obtained:

εx j − ε j ∂ 2 Πex j
∇2 Πex j + k02 εx j µ j Πex j + =0, (5.188)
εj ∂ x2
where we have used the gauge condition
e
1 ∂ Πx j
φj = − . (5.189)
εj ∂x
∂
Recalling that ∂y = 0 for all potentials, from Eqs. (5.180) and (5.183) we have

∂ Πex j
Hej = −iωε0 ŷ , (5.190a)
∂z
and
!
2 e 2 e
1 ∂ Πx j 1 ∂ Πx j
Eej = e
k0 µ j Πx j + x̂ + ẑ . (5.190b)
ε j ∂ x2 εj ∂x ∂z
278 Chapter 5 Electromagnetic Plane Waves

The solution to wave equations (5.176) and (5.188) for plane-wave excitation can be expressed
in terms of wave functions exp[i(kxo j x ± kzoj z)] and exp[i(kxe j x ± kze j z)] for the ordinary and
extraordinary waves. Upon substitution of these wave functions into the respective wave
equations, the following dispersion relationships are obtained:

(kxo j )2 + (kzoj )2 = k02 µ j ε j (5.191a)

and
1 e 2 1
(kx j ) + (ke )2 = k02 µ j . (5.191b)
εj εx j z j
Note that in the upper half-space we only have ordinary waves, and Eq. (5.191a) is used there.
At the interface between the upper medium and lower uniaxial medium (z = 0), the continuity
of the tangential electric field and magnetic field must be enforced. These relations provide
the following phase-matching condition:

kxo j = kxe j = k1 sin θi . (5.192)

Hence Eqs. (5.191a) and (5.191b) can be used to find

q
o
kz j = k0 µ j ε j − µ1 ε1 sin2 θi (5.193a)
and
r q
εx j
kze j = k0 µ j ε j − µ1 ε1 sin2 θi . (5.193b)
εj

The Hertz vector potential for ordinary and extraordinary waves in each region can be written
as
o o
Πox j = [Aij e−ikz j z + Arj eikz j z ]eik1 sin θi x (ordinary wave) (5.194a)
and
e e
Πex j = [Bij e−ikz j z + Brj eikz j z ]eik1 sin θi x , (extraordinary wave) (5.194b)

where the superscripts “i” and “r” refer respectively to incident and reflected waves. The
second region is semi-infinite, and thus there is no reflected wave; hence Ar2 = Br2 = 0. From
Eqs. (5.177a) and (5.178), the tangential components of electric and magnetic fields in the
upper half-space are given by
o o
o
Ey1 o
= ω µ0 µ1 kz1 [Ai1 e−ikz1 z − Ar1 eikz1 z ]eik1 sin θi x (5.195a)
and
o o
o
Hx1 o 2 i −ikz1 z
= (kz1 ) [A1 e + Ar1 eikz1 z ]eik1 sin θi x , (5.195b)

and the fields in the lower half-space are given by

o
o o i −ikz2 z
Ey2 = ω µ0 µ2 kz2 A2 e0 (5.196a)
5-7 Plane-Wave Reflection and Transmission from a Half-Space Uniaxial Medium 279

and
o z
−ikz2
o o 2 i
Hx2 = (kz2 ) A2 e0 . (5.196b)

Enforcing the continuity of the tangential electric and magnetic field components at the
interface between two adjacent media provides
o
µ1 kz1 (Ai1 − Ar1 ) = µ2 kz2
o i
A2 (5.197a)
and
o 2 i
(kz1 ) (A1 + Ar1 ) = (kz2
o 2 i
) A2 . (5.197b)

Noting that the reflection coefficient for the ordinary wave is ro = −Ar1 /Ai1 , it follows that
o − µ ko
µ1 kz2 2 z1
ro (θi ) = − o + µ ko , (5.198)
µ1 kz2 2 z1

o and ko are given by Eq. (5.193a). Explicitly,

where kz2 z1

q
√
o
µ1 µ2 ε2 − µ1 ε1 sin2 θi − µ2 µ1 ε1 cos θi
r (θi ) = − q . (5.199)
2 √
µ1 µ2 ε2 − µ1 ε1 sin θi + µ2 µ1 ε1 cos θi
(reflection coefficient for ordinary wave)

The transmission coefficient is defined as

o
µ2 kz2 Ai2
τ= o (5.200a)
µ1 kz1 Ai1

and
τ = 1 + ro (θ1 ) . (5.200b)

For an extraordinary wave, the expressions for the electric and magnetic fields in the upper
medium can be obtained from Eqs. (5.190a) and (5.190b), and are given by
 e e

e
Hy1 e
= −kz1 ωε0 Bi1 e−ikz1 z − Br1 eikz1 z eik1 sin θi x (5.201a)
and
 e e

e
Ex1 = k02 µ1 Bi1 e−ikz1 z + Br1 eikz1 z cos2 θi eik1 sin θi x , (5.201b)

and in the lower half-space,

e
e
Hy2 e
= −kz2 ωε0 Bi2 e−ikz2 z eik1 sin θi x (5.202a)
and
280 Chapter 5 Electromagnetic Plane Waves

 
µ2 ε2 − µ1 ε1 sin2 θi e
e
Ex2 = k02 Bi2 e−ikz2 z eik1 sin θi x . (5.202b)
ε2
Enforcing the boundary conditions at z = 0 provides
e
kz1 [Bi1 − Br1 ] = kz2
e i
B2 (5.203a)
and
µ2 ε2 − µ1 ε1 sin2 θi i
µ1 cos2 θi [Bi1 + Br1 ] = B2 . (5.203b)
ε2
Solving Eqs. (5.203a) and (5.203b) simultaneously gives the magnetic-field reflection
coefficient
q
√
ε2 εx2 cos θi − ε1 µ2 ε2 − µ1 ε1 sin2 θi
re (θi ) = q . (5.204)
√ 2
ε2 εx2 cos θi + ε1 µ2 ε2 − µ1 ε1 sin θi
(reflection coefficient for extraordinary wave)

Also, it can be shown that the magnetic-field transmission coefficient can be computed from

τ e (θi ) = 1 + re (θi ) . (5.205)

It is interesting to note that the phenomenon of no surface reflection at the Brewster angle still
occurs for the half-space uniaxial medium. Setting the numerator of Eq. (5.204) to zero, the
Brewster angle can be computed from

s !
−1 ε2 (εx2 − µ2 ε12 )
θB = sin . (Brewster angle) (5.206)
ε2 εx2 − µ1 ε13

Note that the permittivity and permeability values in Eq. (5.206) are all relative quantities and
are dimensionless. To illustrate with an example, the reflection coefficients for ordinary and
extraordinary waves are plotted in Fig. 5-14 as a function of incident angle using Eqs. (5.199)
and (5.204), respectively, for an arrangement in which µ1 = ε1 = 1 and µ2 = 1, ε2 = 4, and
εx2 having different values of 2, 4, and 6.

5-8 Plane Waves in Layered Media

In many high-frequency circuits and electromagnetic scattering from geophysical media,
layered dielectric problems are encountered. To gain some intuition about such problems,
we now examine the plane-wave interaction with a layered dielectric medium. Consider a
stratified dielectric medium with planar interfaces parallel to the x–y plane of a coordinate
system. Suppose the constitutive parameters of the nth layer are denoted by µn and εn , and
5-8 Plane Waves in Layered Media 281

0.9 r o, ε2 = 4
Magnitude of reflection coefficient

r e, εx2 = 2
0.8
r e, εx2 = 4
0.7 r e, εx2 = 6

0.6

0.5

0.4

0.3

0.2

0.1

0
0 10 20 30 40 50 60 70 80 90
Incident angle (degrees)

Figure 5-14: The magnitude of the reflection coefficient for a plane wave incident upon a
uniaxial medium with an optical axis parallel to the interface (x-axis), as a function of incident
angle. The upper half-space is isotropic with ε1 = µ1 = 1 and the lower half-space is uniaxial with
ε2 = 4 and three different values of εx2 = 2, 4, and 6.

its boundary with the (n + 1)th layer is at z = −dn . It is also assumed that the top (region 0)
and bottom (region N + 1) layers are semi-infinite, as shown in Fig. 5-15. Without loss of
generality, let us assume that a plane wave is illuminating the interface from above and whose

k̂i z

x
,ε
,ε
,ε

,ε
,ε

Figure 5-15: A layered dielectric medium composed of N + 1 dielectric media with parallel
planar interfaces illuminated by a plane wave from the upper half-space.
282 Chapter 5 Electromagnetic Plane Waves

direction of propagation is given by

1
k̂i = (k0x x̂ − β0 ẑ) , (5.207)
k0
√
where k0 = ω µ0 ε0 , k0x = k0 sin θi , and β0 = k0 sin θi . For TM waves all field quantities
can be expressed in terms of the axial component of the electric field (Ez ), and for TE waves
they can be expressed in terms of the axial component of the magnetic field (Hz ). Because of
the symmetry of the problem, ∂ /∂ y = 0 for all field components. It is also assumed that the
dependence on z in the nth layer is of the form e±iβn z .
For TE waves, Hnz can be obtained from the solution to Helmhotz’s equation
 2 
∂ 2 2


+ kn − βn Hnz = 0 , (5.208)
∂ x2
√
where kn = ω µn εn . Other field quantities can be expressed in terms of Hnz and are given by

1 ∂2
Hnx = Hnz (5.209a)
kn2 − βn2 ∂ x ∂ z
and
−iω µn ∂
Eny = Hnz . (5.209b)
kn2 − βn2 ∂ x
TE and TM waves are duals of each other under the standard duality relationship E → H,
H → −E, and µ ⇋ ε . Hence, once the field quantities for TE incidence are obtained, the
results for a TM excitation can be obtained using the duality relationships. For an incident TE
wave,
Eyi = E0 e−iβ0 z eik0 sin θi x . (5.210)
The phase-matching condition mandates that the x-dependence of all field quantities be the
same as that of the incident wave (eik0 sin θi x ). The total field quantity in region n, excluding its
dependence on x, is given by

Eny = An eiβn z + Bn e−iβn z , (5.211)

where An and Bn represent the coefficients of the wave components propagating along the
positive and negative z-directions, respectively. In the zeroth region it is recognized that

B0 = E0 (5.212a)
and
A0 = RTE E0 , (5.212b)

where RTE is the total electric field reflection coefficient for TE (perpendicular polarization)
waves. Since the bottommost region (namely the (N + 1)th region) is semi-infinite, there
5-8 Plane Waves in Layered Media 283

would be no upward-traveling wave in that region, i.e.,

AN+1 = 0 (5.213a)
and
BN+1 = T TE E0 , (5.213b)

where T TE is the total transmission coefficient.

Evaluating Hx and Ey from Eqs. (5.209a) and (5.209b), and then enforcing the continuity
of tangential electric and magnetic fields at the interface between the nth and (n + 1)th region
(at z = −dn ), we get

βn h −iβn dn i β
(n+1)
h i
An e − Bn eiβn dn = An+1 e−iβ(n+1) dn − Bn+1 eiβ(n+1) dn (5.214a)
µn µ(n+1)
and
h i h i
An e−iβn dn + Bn eiβn dn = An+1 e−iβ(n+1) dn + Bn+1 eiβ(n+1) dn . (5.214b)

There are (N + 2) regions, and therefore the number of unknowns is (2N + 2) (because B0
and AN+1 are known). On the other hand, there are (N + 1) boundaries and (2N + 2) equations
for the (2N + 2) unknowns. A simple procedure to solve for the unknowns is as follows:

1. Find Γ TE by recursively relating AN+1 /BN+1 to AN /BN (while keeping in mind that
AN+1 = 0), then relating AN /BN to AN−1 /BN−1 and so on until the ratio A0 /B0 is
reached.

2. Once A0 is found, relate An+1 , Bn+1 to An , Bn to find all coefficients.

Solving (5.214a), (5.214b) for An and Bn results in

1 n o
An e−iβn dn = 1 +Un(n+1) An+1 e−iβ(n+1) dn + rn(n+1)
TE
Bn+1 eiβ(n+1) dn (5.215a)
2
and
1  n TE o
Bn e+iβn dn = 1 +Un(n+1) rn(n+1) An+1 e−iβ(n+1) dn + Bn+1 eiβ(n+1) dn , (5.215b)
2
where
µn β(n+1)
Un(n+1) = ,
µ(n+1) βn
and
TE µn+1 βn − µn β(n+1)
rn(n+1) =
µn+1 βn + µn β(n+1)
284 Chapter 5 Electromagnetic Plane Waves

is the Fresnel reflection coefficient for a horizontally polarized wave. Taking the ratio of
Eqs. (5.215a) and (5.215b), we obtain
h i
An+1 −2iβ(n+1) d(n+1)
An −2iβn dn Bn+1 e e2iβ(n+1) (d(n+1) −dn ) + rn(n+1)
TE
e =h i (5.216)
Bn An+1 −2iβ(n+1) d(n+1)
e e2iβ(n+1) (d(n+1) −dn ) rTE +1
Bn+1 n(n+1)

This is a recurrence relation that can be used to find A0 /B0 starting from AN+1 /BN+1 = 0. In
order to solve for T TE and the rest of the coefficients, we rearrange Eqs. (5.214a) and (5.214b)
to obtain An+1 and Bn+1 in terms of An and Bn . In matrix notation we have
   −iβ d 
An+1 e−iβ(n+1) d(n+1) =TE A e n n
iβ(n+1) d(n+1) =F (n+1)n n iβn dn , (5.217)
Bn+1 e Bn e

=TE
where F (n+1)n is referred to as the forward propagation matrix, given by
" #
=TE 1
e−iβ(n+1) (d(n+1) −dn ) TE
r(n+1)n e−iβ(n+1) (d(n+1) −dn )
F (n+1)n = 1 +U(n+1)n , (5.218)
2 TE
r(n+1)n eiβ(n+1) (d(n+1) −dn ) eiβ(n+1) (d(n+1) −dn )

where
µ(n+1) βn
U(n+1)n =
µn β(n+1)
and
TE TE
µn β(n+1) − µ( n + 1)βn
r(n+1)n = −rn(n+1) = .
µn β(n+1) + µ(n+1) βn
Knowing A0 and B0 , Eq. (5.217) can be used to find An and Bn (n ∈ {1, . . . , N + 1}). For
TM-polarized waves, the duality relations can be used to find the desired quantities.
This formulation for calculating the reflection and transmission properties of plane waves
at interfaces between multilayer dielectric media can be used for the design of practical
devices in optics and electromagnetics. As discussed in Chapter 1, at very high frequencies
such as in the optical regime, the conductivity of metals may not be sufficiently high as to
provide surfaces with high reflectivity and minimal loss. However, using multilayer dielectric
materials, surfaces with very high reflectivity can be designed. Consider a multilayer periodic
√
layered medium composed of two different materials with indices of refraction n2 = µ2 ε2
√
and n3 = µ3 ε3 and thicknesses t2 and t3 , stacked upon each other N times and placed
√ √
between two half-space media with indices of refraction n1 = µ1 ε1 and n4 = µ4 ε4 (see
Fig. 5-16). Here µ and ε represent the relative permeability and permittivity of each layer.
Suppose thicknesses t2 and t3 are chosen such that

λ0
n2t2 = (5.219a)
4
and
5-8 Plane Waves in Layered Media 285

N layers of stacked dielectric pairs

±1

k̂i

ε ε ε ε ε ε

Figure 5-16: Periodic N-layer stack of dielectric media, each a quarter-wavelength thick, with
different indices of refraction, designed to exhibit total reflection.

λ0
n3t3 = , (5.219b)
4
where λ0 is the free-space wavelength. That is, the thickness of each layer is a quarter-
wavelength for λ in that medium. In this case it can be shown that the reflection of a plane
wave from region 1 is
( p p
−1 for µ2 /ε2 < µ3 /ε3 ,
lim R = p p (5.220)
N→∞ 1 for µ2 /ε2 > µ3 /ε3 .

This structure is known as a distributed Bragg reflector (DBR) or Bragg mirror. In fact, the
wave impedance as seen at the interface between region 1 and medium 2 is found to be
 2N
η2
Zin = η4 . (5.221)
η3
It is obvious that if η2 < η3 , then for a sufficiently large number of layers (N), one can make
Zin arbitrarily small, and hence R = −1. Also, when η2 > η3 , then Zin can be made very large,
rendering R = 1. These conditions correspond to PEC and PMC boundary conditions. This
structure is of course band-limited, as we chose each layer to be a quarter-wavelength thick.
The results obtained in Eqs. (5.220) and (5.221) can easily be extended to oblique
incidence, noting that the propagation constant and wave impedance in the jth layer are given
by

β j = k0 n j cos θ j ,
η TE
j = η j cos θ j ,
286 Chapter 5 Electromagnetic Plane Waves

and
ηj
η TM
j = ,
cos θ j

where s  
n1
cos θ j = 1− sin2 θ1
nj
and j = 1, 2, 3, or 4. In this case the layer thickness should be modified as well:

λ0
n jt j = .
4 cos θ j

Example 5-5: Plane-Wave Propagation in an Inhomogeneous Medium

(WKB Problem)

Consider an inhomogeneous medium characterized by permittivity function ε (r) = ε (z) and

permeability function µ (r) = µ , supporting a linearly polarized plane wave of the form

E(r) = ŷ Ey (z) eikx x .

Derive the wave equation for such a medium and find a possible solution for Ey (z), assuming
that ε (z) is a slowly varying function of z. That is, variation of ε (z) with respect to the
“average” wavelength at each point in the medium is slow.

Solution: The solution to this problem under the assumption that ε (z) is a slow-varying
function is known as the WKB solution, named after Wentzel, Kramers, and Brillouin. To
derive the wave equation for such a medium, we make use of Faraday’s law and the modified
Ampère’s law in a source-free region, i.e.,

∇ × E = iω µ H (5.222a)
and
∇ × H = −iω ε (z) E . (5.222b)

Taking ∇× of Eq. (5.222a) and then using Eq. (5.222b), we get the following vector wave
equation:
∇ × ∇ × E − ω 2µ ε (z) E = 0 . (5.223)
Note that
∇ × ∇ × E = ∇(∇ · E) − ∇2E . (5.224)
5-8 Plane Waves in Layered Media 287

Using Gauss’ law for electricity (∇ · D = 0), we have

  
D
∇(∇ · E) = ∇ ∇ ·
ε
   
1
=∇ ∇ ·D
ε
 
∇ε
= ∇ − 2 ·ε E
ε
 
∇ε · E
= −∇ . (5.225)
ε
Noting that
d ε (z)
∇ε = ẑ
dz
and E = ŷ Ey (z) eikx , it then follows that for this particular solution, ∇(∇ · E) = 0. As a result,
the wave equation (5.223) is simplified to

[∇2 + ω 2 µ ε (z)] Ey (z) eikx x = 0 . (5.226)

Since there are no variations with respect to y, it follows that ∂ /∂ y = 0, and therefore
Eq. (5.226) can be further simplified to
 2 
d 2
+ kz (z) Ey (z) = 0 , (5.227)
dz2
where
kz2 (z) = ω 2 µ ε (z) − kx2 . (5.228)
To proceed further, let us represent Ey (z) = F(z) ei φ (z) in its polar form for some slowly
varying magnitude function F(z). By substituting this polar form into Eq. (5.227), the
following equation is obtained:
 2
d 2 F(z) dF(z) d φ (z) d 2 φ (z) ∂ φ (z)
+ 2i + i F(z) − F(z) + kz2 (z) F(z) = 0 . (5.229)
dz2 dz dz dz2 dz

Since F(z) is a slowly varying function of z, d 2 F(z)/dz2 ≈ 0 compared with other terms in
Eq. (5.229). Separating the real and imaginary parts of Eq. (5.229), the following equations
for F(z) and φ (z) are obtained:
 2
d φ (z)
− kz2 (z) = 0 (5.230a)
dz
and
dF(z) d φ (z) d 2 φ (z)
2 + F(z) =0. (5.230b)
dz dz dz2
288 Chapter 5 Electromagnetic Plane Waves

From Eq. (5.230a),

d φ (z)
= ±kz (z) , (5.231)
dz
which can be solved to yield Z z
φ (z) = ± kz (z′ ) dz′ . (5.232)
z0

Now substituting Eq. (5.231) into Eq. (5.230b), we have

1 dF(z) 1 dkz (z)

=− . (5.233)
F(z) dz 2kz (z) dz
Equivalently,
d d  
[ln F(z)] = ln[kz (z)]−1/2 ,
dz dz
which can be solved to get
E0
F(z) = p .
kz (z)
Hence the possible solution to the wave equation takes the following form:

1  Rz Rz 
i k(z′ ) dz′ −i k(z′ ) dz′
Ey (x, z) = p E0+ e z0 + E0− e z0 eikx x . (5.234)
k(z)

Example 5-6: Reflection by and Transmission through a Dielectric Slab

Find expressions for the total reflection and transmission coefficients of a dielectric slab with
thickness t, permittivity ε1 , and permeability µ1 sandwiched between two half-space dielectric
media with planar interfaces and permittivities and permeabilities of ε0 , µ0 (top region) and
ε2 , µ2 (bottom region), as shown in Fig. 5-17. Assume that a plane wave propagating along
k̂i = (1/k0 )(k0x x̂ − β0 ẑ) is incident on the slab from the top region.
Solution: This is a special case of the general problem for multilayer dielectric media. The
electric field expression for a TE wave is given by

Ei = ŷ E0 eik0 sin θ x e−iβ0 z .

The starting point is Eq. (5.216). Recalling that for any layer n, An represents the magnitude of
the electric field of the wave propagating along +ẑ and Bn is associated with the wave moving
downward along −ẑ, in the present case A0 and B0 represent the fields in the top medium and
A1 and B1 pertain to the middle layer. Noting that A2 = 0 in the bottom region, application of
Eq. (5.216) for n = 1 yields
A1 −2iβ1 d1 TE
e = r12 . (5.235)
B1
5-8 Plane Waves in Layered Media 289

k̂i z

,ε
x
,ε
,ε
,ε

Figure 5-17: Dielectric slab with (µ1 , ε1 ) over a half-space dielectric medium with (µ2 , ε2 ).

Using Eq. (5.216) again for n = 0, we get

A1 −2iβ1 d0 TE
e + r01
A0 −2iβ0 d0 B1
e = . (5.236)
B0 A1 −2iβ1 d0 TE
e r01 + 1
B1

Choosing d0 = 0 and utilizing Eq. (5.235), the total reflection coefficient (RTE = A0 /B0 ) can
be obtained from Eq. (5.236), noting that d1 = t:

TE ei2β1 t + rTE
r12
RTE = 01
TE rTE ei2β1 t + 1
. (5.237)
r01 12

Also, using Eqs. (5.215a) and (5.215b) twice, it can be shown that
" #    " iβ1 t TE −iβ1 t # " TE
# " −iβ t #
A0 1 +U01 1 +U12 e r01 e 1 r12 A2 e 2
=
B0 2 2 TE eiβ1 t e−iβ1 t
r01 TE 1
r12 B2 eiβ2 t
   " iβ1 t TE rTE e−iβ1 t rTE eiβ1 t + rTE e−iβ1 t
#" #
1 +U01 1 +U12 e + r01 12 12 01 0
= .
2 2 TE eiβ1 t + rTE e−iβ1 t e−iβ1 t + rTE rTE eiβ1 t
r01 12 01 12 B2 eiβ2 t
(5.238)

Hence,   
1 +U01 1 +U12  i(β2 −β1 )t TE TE i(β1 +β2 )t

B0 = e + r01 r12 e B2 . (5.239)
2 2
290 Chapter 5 Electromagnetic Plane Waves

Defining the transmission coefficient as

B2 4ei(β1 −β2 )t
T TE = = TE rTE ei2β1 t )
, (5.240)
B0 (1 +U01 )(1 +U12 )(1 + r01 12

and recognizing that

2 1 −U01 TE TE
= 1+ = 1 + r01 = τ01 (5.241a)
1 +U01 1 +U01
and
2 1 −U12 TE TE
= 1+ = 1 + r12 = τ12 , (5.241b)
1 +U12 1 +U12
we have
TE τ TE ei(β1 −β2 )t
τ01
T TE = 12
TE rTE ei2β1 t
. (5.242)
1 + r01 12

5-9 Plane-Wave Propagation in a Negative-Index Medium

The constitutive parameters of various types of materials were discussed at some length
in Chapter 1. There, the dispersion behavior of permittivity and permeability of materials
associated with the spectral response of the induced dipole moments or magnetization vector
of molecules revealed the possibility for the real part of the permittivity or permeability
of the medium to assume negative values over a certain region of the spectrum. When the
real parts of permittivity (ε ′ ) or permeability (µ ′ ) are negative, the propagation constants
become imaginary, and the medium inhibits wave propagation over the frequency band
where ε ′ or µ ′ is negative. Such media are referred to as bandgap materials. However, in
situations where the real parts of both the permittivity and the permeability of the medium
are negative simultaneously, wave propagation is possible but with certain peculiarities that
deserve attention. Such media are referred to as double-negative (DNG) media. One of the
intriguing properties of a DNG medium is the fact that its index of refraction is negative, which
leads to the possibility that such a medium can potentially overcome the known limitations of
conventional lenses, thereby allowing for focusing the entire spectra of plane waves, including
the evanescent ones, which constitutes the realization of a perfect lens.
To examine some of these peculiarities, let us consider the behavior of a plane wave
propagating in a DNG medium. Similar to a source-free double-positive medium, a plane-
wave solution to the wave equation takes the following form:

E = E0 eik·r , (5.243)

where k, the propagation vector, must satisfy k · k = ω 2 µε . For the sake of simplicity let us
assume that k has only a z-component and that

E = E0 eikz (5.244)
5-9 Plane-Wave Propagation in a Negative-Index Medium 291

with a dispersion relation

k2 = ω 2 µε , (5.245)
or √ √
k=ω µ ε. (5.246)
To evaluate k properly, special attention must be given when evaluating the square-root
functions. As discussed in Chapter 1, the causality condition mandates that the imaginary
parts of µ and ε be positive quantities
√ and that the proper branch cut per the chosen time
convention e−iω t is such that −1 = i. Hence for media satisfying the causality condition,
ε and µ must fall within the upper half-plane of the complex µ and ε planes. As a result, for
an ε -negative medium, ε falls in the third quadrant√of the complex ε -plane, and the same can
√
be said about a µ -negative medium. Therefore, ε and µ are each in the first quadrant,
√
and the index of refraction n = µε again falls in the third quadrant. That is, the index of
refraction has a negative real part. This statement also indicates that

Re[k] = k′ < 0 .

Recall that the phase velocity is calculated from

d ′
(k z − ω t) = 0 ,
dt
or
dz ω
up = = ′ <0. (5.247)
dt k
That is, the phase velocity is a negative quantity unlike that for a double-positive medium that
has the same but positive values for µ and ε . The magnetic field can be obtained directly from
the modified Ampère’s law:
ik × E0 ik·r
H= e
iω µ
r
ε ikz
= ẑ × E0 e = H0 eikz . (5.248)
µ
p
But since both µ and ε are negative, the characteristic impedance η = µ /ε has a positive
real part and the direction of the magnetic field would be the same as what it would have
been for a double-positive medium. From Gauss’ laws for electricity and magnetism it can be
shown that E0 ⊥ H0 ⊥ k′ ẑ, similar to a double-positive medium. However, (E0 , H0 , k′ ) form a
left-hand system for a DNG, as opposed to a right-hand system for a double-positive system.
It is also interesting to examine the behavior of the time-average Poynting vector. As shown
in Eq. (5.27), the Poynting vector is given by
1
S(r) = Re[E × H∗ ]
2
 r ∗ 
1 ε ′′
2
= ẑ |E0 | Re e−2k z , (5.249)
2 µ
292 Chapter 5 Electromagnetic Plane Waves

p
where k′′ = Im[k]. Noting that Re[( ε /µ )∗ ] > 0, and from Eq. (5.249), it is evident that S
is in the positive z-direction, whereas the phase velocity is in the negative z-direction.
Next we consider a plane wave incident upon a half-space DNG medium. Let us consider
a plane wave propagating along k̂i = sin θi x̂ + cos θi ẑ in a medium with positive permittivity
and permeability ε1 and µ1 , respectively, incident on a DNG half-space with permeability ε2
and permeability µ2 , as shown in Fig. 5-18. In the lower half-space the propagation vector is
given by
k1 = k1 sin θi x̂ + k1 cos θi ẑ ,
√
with k1 = ω µ1 ε1 . At the interface the phase-matching condition mandates that

k2 = k1 sin θi x̂ + k2 cos θt ẑ ,

with k2 having a negative real value and θt being defined by Snell’s law:
n1
sin θt = sin θi ,
n2
and since n2 < 0, θt is negative, as shown in Fig. 5-18. Also shown in the figure is a double-
positive medium and the corresponding phase fronts. Since the wave impedance (TE or TM)
of a DNG medium is positive, the expressions for the reflection and transmission coefficients
are unchanged and are the same as those obtained for conventional double-positive media.

5-10 Negative Refractive-Index Lens

The characteristics of plane-wave propagation in double-negative or negative refractive-
index materials were discussed in the previous section. Here we examine qualitatively the
propagation of electromagnetic waves emitted from a point source in a double-positive
medium placed above a DNG half-space material with a planar interface, as shown in
Fig. 5-19. As shown in Section 5-2.2, fields emanating from an arbitrary source can be
expressed in terms of a continuous spectrum of plane waves that includes propagating and
evanescent plane waves. Propagating plane waves can be represented as rays coming off the
point source and then impinge on the interface between the two media. Each plane wave
refracts at the interface according to Snell’s law, but since the lower medium has a negative
index of refraction, the refracted rays form a linear caustic along a line perpendicular to the
interface that passes through the point source. In fact, rays on a constant incident angle θi cone
all pass through one point in the DNG medium. The location of these focal points depends on
the incident angle, and they all lie on the linear caustics as shown in Fig. 5-19. For example,
a ray that impinges on the interface at an angle θi refracts in a converging manner and meets
the perpendicular ray at a depth of H, as also shown in Fig. 5-19.
Using Snell’s law, and noting that
tan θi H
= , (5.250)
tan θt h
5-10 Negative Refractive-Index Lens 293

z Double-
negative
medium
k2 S2

t
,ε
x
S1 ,ε
i
k1
Double-
positive
medium

(a) Plane wave incident in a double-positive

medium upon a double-negative medium

z
Double-
positive
medium
S2
t k
2
,ε
x
S1
i ,ε
k1
Double-
positive
medium

(b) Plane wave incident in a double-positive

medium upon a double-positive medium

Figure 5-18: (a) A plane wave is incident from a double-positive medium (the lower half-
space) upon a planar interface with a double-negative medium (the upper half-space). The phase-
matching condition mandates that the tangential components of the k-vector (kx ) in both media
to be the same. However, the normal component of the propagation vector is reversed when
compared to a double-positive medium (as shown in (b)).
294 Chapter 5 Electromagnetic Plane Waves

Double- Point source

positive k1
medium h
i i
x

t t
H0 k2
Double-
negative Linear caustic
medium for
H

Figure 5-19: A point source in a double-positive medium above a double-negative medium. The
transmitted rays form a linear caustic in the lower half-space. Rays near normal incidence pass
through the caustic point H0 .

it can be shown that  

 n2  cos θt
H =   h . (5.251)
n1 cos θi
For very small values of incident angle (θi ≈ 0), the starting point of the caustic is located at
 
 n2 
H0 =   h . (5.252)
n1

For |(n2 /n1 )| > 1, the linear caustic starts from z = −H0 and goes to −∞. On the other hand, if
|(n2 /n1 )| < 1, the linear caustic starts from z = −H0 and goes to zero. An interesting condition
emerges when |(n2 /n1 )| = 1. In this case, the linear caustic collapses to a point and the half-
space DNG medium acts like a lens. This is due to the fact that θt = −θi , and as a result,
all refracted rays must go through the mirror-image point of the source. This particular lens
presents a number of advantages over traditional lenses, especially when

n2 = −n1 . (5.253)

In this case, the wave impedances for TE and TM waves in the upper half-space, computed
from Eqs. (5.103) and (5.106), are given by

β1
Z1TM = (5.254a)
ωε1
and
ω µ1
Z1TE = , (5.254b)
β1
5-10 Negative Refractive-Index Lens 295

q
where β1 = k12 − kρ2 and kρ2 = kx2 + x2y . Also, in the lower half-space we have

β2
Z2TM = (5.255a)
ωε2
and
ω µ2
Z2TE = , (5.255b)
β2
with q
β2 = k22 − kρ2 .

Noting that Re[k2 ] < 0, it can be shown that also Re[β2 ] < 0. For a lossless case and a
propagating wave (k22 > kρ2 ), considering the fact that ε2 < 0 and µ2 < 0, it is obvious that
the impedances given by Eqs. (5.254) and (5.255) are positive, and that

Z2TM = Z1TM (5.256a)

and
Z2TE = Z1TE , (5.256b)

recalling that k22 = k12 . The reflection coefficients at the interface are thus calculated to be

Z2TE − Z1TE
RTE = =0 (5.257a)
Z2TE + Z1TE
and
Z2TM − Z1TM
RTM = =0, (5.257b)
Z2TM + Z1TM

independent of the incident angle. That is, all power carried by the propagating waves is
transferred and focused. It is also interesting to examine the reflection coefficient of the
evanescent waves. For the evanescent waves,
q q
β1 = k12 − kρ2 = i kρ2 − k12 = iα , (5.258)

where α is a positive number. If the lower half-space is slightly lossy, then k22 is in the fourth
quadrant, and for kρ2 > k22 , we find
q
β2 = k22 − kρ2

in the third quadrant, and in the limit as k2 → −k1 we have

β2 = iα . (5.259)
296 Chapter 5 Electromagnetic Plane Waves

,ε h

ε ε t

,ε
t−h

Figure 5-20: Configuration of a perfect lens created by a negative-index material having

permittivity and permeability equal to the negative of those of the surrounding medium. It is
shown that the total reflection from the DNG slab is zero for all propagating and evanescent
waves emanating from a point source. Also, all radiating power appears at the image point with a
negative phase delay of 2β0t.

Thus, for the evanescent waves,

iα ω µ1
Z1TM = and Z1TE = , (5.260a)
ωε1 iα
and
iα ω µ1
Z2TM = − and Z2TE = − , (5.260b)
ωε1 iα
for which the reflection coefficient of the evanescent wave is found to be

RTE TM
eva = Reva = ∞ .

This indicates that evanescent waves build up very intensely near the interface.
This configuration for a lens is not very practical, as the focal point (image point) is
within the lens. However, a practical lens can be configured from a DNG material slab with
thickness t and planar interfaces having permeability and permittivity µ2 = −µ1 and ε2 = −ε1 ,
respectively, as shown in Fig. 5-20. In this configuration, for a source point at a height h < t, an
interior image point is formed inside the slab, which acts as a source for the second interface.
This interior image has a secondary image in the lower half-space at a distance of t − h below
the bottom interface. Using the results obtained in Example 5-4, the total reflection coefficient
can be obtained from Eq. (5.237). For propagating waves, as shown above
TE TM
r01 = r01 =0,
5-10 Negative Refractive-Index Lens 297

which when used in Eq. (5.237) renders

RTE = RTM = 0 .

For evanescent waves, it can easily be shown that the reflection coefficients at the interface
TE = rTM = ∞. Rewriting Eq. (5.237) and recalling that rTE/TM = −rTE/TM ,
are infinite; i.e., r01 01 01 10
the total reflection coefficient for evanescent waves can be written as
 . 
TE/TM
1 r01 (1 − ei2β1 t )
TE/TM
Reva = .    .
TE/TM 2 i2β1 t
1 r01 −e

Now it becomes apparent that

RTE TM
eva = Reva = 0 . (5.261)
This indicates that the lens is also matched to all evanescent waves and that all active
and reactive power generated by the source appears at the image point. Now examining
the transmission coefficient for this lens, we can use Eq. (5.240), and we observe that for
propagating waves

µ0 β1 µ0 (−β0 )
U01 = = =1
µ1 β0 (−µ0 )β0
and
µ1 β2 (−µ0 )β0
U12 = = =1,
µ2 β1 µ0 (−β0 )
TE = rTE = 0. Thus the total transmission coefficient for propagating waves is given by
and r01 12

T TE = ei(−β0 −β0 )t = e−i2β0 t . (5.262)

For evanescent waves,

µ0 (iα )
U01 = = −1
−µ0 (iα )
and
−µ0 (iα )
U12 = = −1 ,
µ0 (iα )
TE = rTE = ∞. It can also be shown that
and r01 12

TE
(1 +U01 )r01 =2
and
TE
(1 +U12 )r12 =2,
 298 Chapter 5 Electromagnetic Plane Waves

and as a result, for evanescent waves the total transmission coefficient is given by
TE
Teva = e2α t . (5.263)

It is interesting to note that the transmitted evanescent waves are amplified through the DNG
lens.

Summary
Concepts

• The electric field and magnetic field satisfy • A plane-wave solution in anisotropic media is
the vector wave equation in homogeneous, facilitated by kDB coordinate transformation.
isotropic, and unbounded media. An example for a simple anisotropic medium,
• A class of solution to the wave equation is known as a uniaxial medium, is considered, and
known as plane waves, for which the vector it is shown that, depending on the polarization of
directions of the electric and magnetic fields are the electric field, two different waves (ordinary
invariant throughout space. and extraordinary) with different propagation
• The magnitude and phase of plane waves constants and characteristics can be supported.
are constant in planes perpendicular to the For ordinary waves, the electric field and
direction of propagation, and the ratio of the electric flux density are parallel to each other
electric field to the magnetic field is constant and both are perpendicular to the plane formed
throughout space. This constant is known as by the propagation vector and the optical axis.
the characteristic
p impedance of the medium For the extraordinary wave, D and E are no
η = µ /ε . longer parallel to each other, and both lie in the
• Due to the linearity of Maxwell’s equations, plane formed by the propagation vector and the
fields produced by arbitrary sources, outside the optical axis. In this case the electric field has a
source region, can be expressed in terms of the component along the direction of propagation.
superposition of an infinite number of plane • Another class of wave expansion, known as
waves propagating in different directions. transverse electric and transverse magnetic
• The polarization of a plane wave is defined as fields is presented. Transverse electric (TE)
the trace of the tip of an electric field as a fields are shown to be produced from a magnetic
function of time in a fixed plane perpendicular vector potential that has only one component
to the direction of propagation. In general the along a desired direction of propagation.
polarization of the wave is elliptical, clearly Transverse magnetic (TM) fields are obtained
identified by two angles, namely the ellipticity from an electric Hertz vector potential that
angle (χ ) and the orientation angle (Ψ). has a single component along the direction of
• A plane-wave solution can also be obtained propagation.
for propagation in anisotropic media, but the • TE and TM field expansions are used to find
dispersion relations relating components of the the reflection and transmission of incident plane
propagation vector to the medium wave number waves at the planar boundaries of layered
are complex. dielectric media.
 SUMMARY 299

Concepts (continued)
• TE and TM field expansions are used to find the reflection and transmission of incident
plane waves at the planar boundaries of layered dielectric media.
• Plane-wave reflection and transmission at the interface of a dielectric and a good
conductor are used to show that good conductors can be modeled by an impedance
boundary condition.
• For a uniaxial half-space a different gauge condition has to be defined to find the
reflected transmission coefficients for TE (ordinary wave) and TM (extraordinary
wave).
• Negative-index media or double-negative media (DNG) possess permittivity and
permeability values whose real parts are both negative. DNG media can support wave
propagation; however, it is shown that their wave number k has a negative real part.
Such media have negative phase velocities, and as a result, they demonstrate a negative
Doppler shift.
• A dielectric slab of a DNG medium whose index of refraction is equal to but the
negative of its surrounding medium is shown to perform like a perfect lens.

Important Terms Provide definitions or explain the meaning of the following terms:
axial ratio energy velocity phase-matching condition
bandgap material evanescent wave plane wave
birefringent extraordinary wave Poincaré sphere
Bragg mirror Fermat’s principle polarization unit vector
Brewster angle forward propagation matrix power density of plane waves
characteristic impedance good conductor principle of least time
continuous spectrum of plane waves impedance boundary condition skin depth
coordinate transformation matrix impermeability Snell’s law of reflection
critical angle impermittivity Snell’s law of refraction
DB plane kDB coordinate TEM wave
dielectric slab magnetic reflection coefficient tilt angle for polarization
dispersion relation magnetic transmission coefficient transcendental equation
distributed Bragg reflector (DBR) method of separation of variables transverse electric fields
double-negative media negative angle of refraction transverse magnetic fields
Eikonal equation negative refractive index uniform plane wave
electric reflection coefficient negative refractive index lens unitary transformation
electric transmission coefficient nonuniform plane wave wave polarization
ellipticity angle for polarization optical path length WKB solution
elliptic polarization ordinary wave
energy density of plane waves perpendicular polarization
300 Chapter 5 Electromagnetic Plane Waves

Important Equations

Wave equation:
∇2 E + k 2 E = 0
Plane-wave electric and magnetic fields:

E(r) = E0 ei K·r ,
k̂ × E0 i K·r
H= e
η
Dispersion relation for a homogeneous isotropic medium:

k2 = K · K = |K′ |2 − |K′′ |2 + 2i K′ · K′′

Plane-wave stored electric and magnetic energy density:

1 2
We =4 ε |E0 | ,

Wm = 14 µ |H0 |2

Plane-wave time-average power density:

|E0 |2
S(r) = k̂
2η
Polarization unit vector:

E x̂ + α eiδ ŷ
p̂ = √ = √
E · E∗ 1 + α2
Polarization tilt and ellipticity angles:

tan(2ψ ) = tan(2γ ) cos δ , sin(2χ ) = sin(2γ ) sin δ

Dispersion relation for ordinary and extraordinary waves in a uniaxial medium:

Kx2 + Ky2 + Kz2 = ω 2 µεxx ,

εzz 2
Kx2 + Ky2 + K = ω 2 µεzz
εxx z
SUMMARY 301

Important Equations (continued)

Transverse magnetic fields:

E = [±iβ ∇t Ψ + kc2 Ψẑ]e±iβ z ,

H = −iωε (∇t Ψ × ẑ)e±iβ z

Transverse electric fields:

H = [iβ ∇t Ψm + kc2 Ψm ẑ]eiβ z ,

E = iω µ (∇t Ψm × ẑ)eiβ z

TM wave impedance:
ẑ × Et β
ZTM = =
Ht ωε
TE wave admittance:
ẑ × Ht β
YTE = − =
Et ωµ
Magnetic field reflection coefficient:
η2 cos θt − η1 cos θi
ΓTM = −
η2 cos θt + η1 cos θi
Electric field reflection coefficient:
η2 / cos θt − η1 / cos θi
ΓTE =
η2 / cos θt + η1 / cos θi

Snell’s law of refraction:

n1 sin θi = n2 sin θt
Brewster angle:
s !
ε2 (µ1 ε2 − µ2 ε1 )
θBTM = sin−1 ,
µ1 (ε22 − ε12 )
s !
µ2 (ε1 µ2 − ε2 µ1 )
θBTE = sin−1
ε1 (µ22 − µ12 )

Eikonal equation:
|∇φ (r)|2 = εr (r) = n(r)
302 Chapter 5 Electromagnetic Plane Waves

Important Equations (continued)

Eikonal equation for ray trajectory:

 
d dr
n(r) = ∇ n(r)
ds ds

Skin depth:
1
δs = √
π f µε
Good conductor surface impedance:
1− j
ηc =
δs σ
Reflection coefficient for extraordinary waves:
q
√
e
ε ε
2 x2 cos θi − ε 1 µ2 ε2 − µ1 ε1 sin2 θi
r (θi ) = q
√
ε2 εx2 cos θi + ε1 µ2 ε2 − µ1 ε1 sin2 θi
Brewster angle for extraordinary waves:
s !
ε2 (εx2 − µ2 ε12 )
θB = sin−1
ε2 εx2 − µ1 ε13

Reflection and transmission coefficients for a dielectric slab:

TE ei2β1 t + rTE
r12
RTE = 01
TE rTE ei2β1 t + 1
,
r01 12
TE τ TE ei(β1 −β2 )t
τ01
T TE = 12
TE rTE ei2β1 t
1 + r01 12

Reflection from a DNG perfect lens:

RTE = RTM = 0 ,
RTE TM
eva = Reva = 0

Transmission from a DNG perfect lens:

T TE = e−i2β0 t ,
TE
Teva = e2α t
PROBLEMS 303

PROBLEMS

Plane Waves

5.1 Consider a plane wave propagating in a dielectric with a loss tangent of tan δ = ε ′′ /ε ′ .
Let the electric field be E0 eikx , where k is the complex propagation constant defined by
√
k = ω µε .

(a) Find the phase difference between the H-field and E-field.

(b) Find the resulting complex Poynting vector. How do you interpret the imaginary and
real parts?

5.2

(a) Can the wave-vector for a plane wave ever be complex in a lossless medium? If so, give
an example of such a situation and state what condition has to hold true between the
real and imaginary parts. If not, explain why.

(b) Give a physical interpretation for the magnitude and direction of the real part of the
wave-vector (K = K′ + i K′′ ). Do the same for the imaginary part.

(c) Prove that the stored electric and magnetic energies for a plane wave are equal in a
lossless medium.

(d) Consider a planar interface between two media, Region 1 and Region 2. For a
nonmagnetic medium, determine the conditions for a TM incident plane wave to have
zero reflectivity at oblique incidence. What about a TE plane wave?

5.3 A plane wave is normally incident from free space onto a half-space of a material
characterized by parameters ε and µ , as shown in Fig. P5.3.

(a) Calculate the transmission and reflection coefficients of the electric field when the
material is lossless.

(b) Prove power conservation for the incident, reflected and transmitted waves at the
interface.

(c) Calculate the transmission and reflection coefficients when the material has finite
conductivity σ .
(Hint: Assume that the incident plane wave has an electric field vector oriented along the
x-axis and is propagating along the positive z-axis, and then match the boundary conditions
for incident, reflected, and transmitted waves.)
304 Chapter 5 Electromagnetic Plane Waves

Ei

kˆ i ε0, μ0
x
ε1, μ1

Figure P5.3: A planar interface between two half-space dielectric media illuminated by a plane
wave of normal incidence.

5.4 A plane wave of angular frequency ω and wave number k propagates in a neutral,
homogeneous, anisotropic, nonconducting medium with µ = 1.
(a) Show that H is orthogonal to E, D, and K, and also that D and H are transverse but E
is not.
(b) Let D = ε E, where ε is a real symmetric tensor. Choosing the coordinate system
parallel to the eigenvectors’ directions, the tensor becomes diagonal:
 
ε1 0 0
ε =  0 ε2 0  .
0 0 ε3

Define K = k k̂, where the components of the unit vector k̂ along the axes are k1 , k2 ,
√
and k3 . If υ = ω /k and υ j = 1/ ε j , show that the components of E satisfy
3
!
X υ2
kj ki Ei + −1 Ej = 0 .
i=1
υ 2j

Write down the equation for the phase velocity u in terms of k j and υ j . Show that this
equation has two finite roots for u2 , corresponding to two distinct modes of propagation
in the direction of k̂.
5.5 Assume a TE plane wave obliquely incident from free space onto a half-space medium
with permittivity ε and permittivity µ0 , as shown in Fig. P5.5.
Show that
Pi = Pr + Pt ,
where Pi, Pr , and Pt are the time-averaged power in the incident, reflected, and transmitted
waves respectively. Hint: The power density toward or away from the interface is given by the
component of the Poynting vector in the direction normal to the interface.
PROBLEMS 305

Ei TE wave
Er
kˆ i z
Hi kˆ r Hr
i r ,ε
x
,ε
t kˆ t
Et
Ht

Figure P5.5: A planar interface between two half-space dielectric media illuminated by a plane
wave illuminating the lower half-space from the top medium at oblique incidence.

5.6 In this problem, you will derive the well-known physical optics (PO) formalism, used
for evaluating the scattered field from an object, assuming a plane-wave incidence on an
object. This is applicable only when the object (more specifically, its curvature) is much larger
than the wavelength of the incident field, so it is a high-frequency technique.
The idea here is as follows: Given an incident plane wave at a planar boundary (see
Fig. P5.6(a)), we can easily find the “scattered” (not really called scattering—simply called
reflected) field by matching boundary conditions at the planar interface. The fields are
computed in terms of Fresnel coefficients, and you will derive them in part (a).
Now, if the scatterer has a large curvature, at every point on the scatterer the surface can
be locally approximated by an infinite tangent plane. At each of those points, we can find
the scattered field using the Fresnel coefficients and tangent-plane approximation. Using the
equivalent surface currents over the entire lit region of the scatterer and using them to find the
reradiated fields gives us a good approximation to the total scattered field.
In part (e), you will implement PO to find the scattered field from a PEC cylinder. It
is important to note that PEC/PMC makes PO implementation much easier, as the Fresnel
coefficients just become constants (part (c)) and field calculation is simpler.
(a) Match boundary conditions to derive the Fresnel reflection coefficient for a plane wave
incident on a planar dielectric boundary. Derive the reflection coefficient for both TE
and TM cases of polarization. Figure P5.6(b) shows the state of polarization for each
case.
Hint 1: Fields in the TE mode are given by

Ei = E0 eik1 (x sin θi −z cos θi ) ŷ ,

1
Hi = E0 eik1 (x sin θi −z cos θi ) (x̂ cos θi + ẑsin θi ) ,
η1
Er = rTE E0 eik1 (x sin θi +z cos θi ) ŷ ,
306 Chapter 5 Electromagnetic Plane Waves

Ei TE wave Hi TM wave
Er Hr
kˆ i z kˆ i z
Hi ˆr
k Hr
Ei kˆ r Er

Region 1 i r i r ,ε
x x
Region 2 ,ε
t ˆt
k t kˆ t
Et Ht
Ht Et

(a)

y
Hi
kˆ i x
Ei

(b)

Figure P5.6: This figure shows the steps for calculating the PO equivalent current on a smooth
surface of objects with large radii of curvature using tangent plane approximation.

rTE
Hr = E0 eik1 (x sin θi +z cos θi ) (−x̂ cos θi + ẑsin θi ) ,
η1
Et = tTE E0 eik2 (x sin θt −z cos θt ) ŷ ,
and
tTE
Ht = E0 eik2 (x sin θt −z cos θt ) (x̂ cos θt + ẑsin θt ) .
η2
The Fresnel reflection coefficient is defined as
E0r
rTE =
E0i
and
H0r
rTM = .
H0i
PROBLEMS 307

Hint 2: Use Snell’s law: θi = θr , k1 sin θi = k2 sin θt .

Hint 3: Use duality to get rTM from rTE .
It should be noted that our definition of rTM is contrary to the usual convention: the H
field is generally assumed to reverse sign upon reflection.
(b) Show that rTE = −1 and rTM = 1 if medium 2 is a PEC.
(c) Given a surface (not necessarily planar) with a normal vector n̂, the following plane
wave is incident on the surface:
k̂ × E0
Ei = E0 eik·r , Hi = H0 eik·r , and H0 = .
η
Consider a typical point on the surface. Define a tangent vector

k̂ × n̂
t̂ = .
|k̂ × n̂|

1. Decompose the incident field into two components: one along the tangent t̂ and
the other one perpendicular to the tangent (k̂ × t̂).
2. Compute the reflected fields by applying the appropriate Fresnel coefficient to
each component of the incident fields.
3. Compute the following:

n̂ × Etotal = n̂ × (Ei + Er )
and
n̂ × Htotal = n̂ × (Hi + Hr ) .

(d) Use boundary conditions to show that at the surface of a PEC

Jms = 0 and Js = 2n̂ × Hi .

The problem of finding the scattered field is now equivalent to the problem of finding the
radiated field due to these currents in the homogeneous background medium (in the absence
of the scatterer). As such, scattered magnetic and electric fields are given by Eqs. (3.118a)
and (3.118b).
(e) Given a PEC cylinder of radius a and finite length ranging from z = −L/2 to z = L/2,
located at the center of the coordinate system, assume that the following plane wave is
incident on this cylinder, as shown in Fig. P5.6:

Ei = −ẑ E0 eikx
and
Hi = ŷ H0 eikx .
308 Chapter 5 Electromagnetic Plane Waves

Find an expression for Hs . You do not need to solve the integral; however, you should
simplify as many terms as you can. Show that the field scattered by PEC cylinders is
similar to the radiation pattern of a dipole. (Hint: Integrate only over the lit region of
the cylinder.)

5.7 Consider a perfect electromagnetic conductor (PEMC). The following boundary

conditions hold at the interface of a PEMC and free space:

n × (H +Y E) = 0
and
n · (D −Y B) = 0 .

In this problem you will find the reflection coefficients from such an interface.

Ei z Er Hr
Hi ki
kr Hr kr Er
θi θr x

Figure P5.7: A planar perfect electromagnetic conductor (PEMC) illuminated by a plane wave
at oblique incidenc.

(a) Assume a TE incident plane wave, as shown in Fig. P5.7. The incident field can be
written as Ei = E0 eik(x sin θ −z cos θ ) ŷ. Find an expression for the incident magnetic field
in the given coordinate system. (Hint: H = η1 k̂ × E.)
(b) The reflected field has both TE and TM components:

Er = ETE TM
r + Er
and
Hr = HTE TM
r + Hr .

Reflection coefficients RTE TM

TE and RTE are defined as follows:

ik(x sin θ +z cos θ )

ETE TE
r = RTE E0 e ŷ
and
 
E0 ik(x sin θ +z cos θ )
HTM
r = RTM
TE e ŷ .
η
PROBLEMS 309

1
Find an expression for ETM TE
r and Hr in the given coordinate system. (Hint: H = η k̂× E
and E = −η k̂ × H.)
(c) By satisfying the boundary conditions at the interface, find the reflection coefficients
and show that (RTM 2 TE 2
TE ) + (RTE ) = 1.
(d) Find the rotation angle (the angle between Er and Ei ). Comment on your results for the
cases of Y = 0, Y = ∞, and Y η = 1.
(e) What can you say about the reflection coefficients, RTM TE
TM and RTM , for a TM incident
field?
5.8
(a) A left-handed material is one that has both negative permittivity and negative
permeability. For a material with constitutive parameters (−µ , −ε ), calculate the
following:
1. The magnetic field vector if the electric field is given by E = x̂ E0 eikz .
2. The time-averaged Poynting vector.
3. What are the directions of phase velocity and energy flow?
(b) Now consider a planar interface between a right-handed medium with constitutive
parameters (µ , ε ) and the left-handed medium of part (a). For a TE plane wave at
oblique incidence from the right-handed medium, draw the rays along the flow of
energy.

z
H

μ, ε k̂ y
−μ, −ε

Figure P5.8: The geometry of a half-space left-handed material illuminated by a plane wave.

5.9 Ceramic tile absorbers. At lower microwave frequencies, and especially at upper VHF
bands, the design of absorbers for anechoic chambers based on conductive foams is rather
challenging due to the size limitations. At such frequencies, lossy µ –ε materials are used.
For example, combining high-µ material like Ferrite and high-ε material like barium titanate
having the same permittivity and permissivity (µ = ε ) and reasonable loss, a thin absorber
can be designed. Consider a metal-backed lossy µ –ε material µr = εr = 100 + i3.
(a) Choose the thickness of the tile so that the reflection coefficient would be less than
−25 dB at normal incidence.
310 Chapter 5 Electromagnetic Plane Waves

(b) Plot the plane-wave reflection coefficient as a function of incident angle for both
polarizations, and find the angular range for which the reflection is below −20 dB.
(c) To increase the angular response of the absorber, one more layer with the same
thickness can be added. Choose a value for the relative permeability and permittivity of
the second layer so as to get the maximum angular range for −20 dB reflection.
5.10 A thin dielectric slab of complex permittivity ε0 εr and thickness t can be modeled by
an impedance sheet, as shown in Fig. P5.10.

E s E Js
ε0 ε r s
E εr
E

Figure P5.10: Impedance sheet equivalent of a thin dielectric slab having a surface
impedance Zs .

(a) Since the dielectric is thin, using the boundary condition, show that

n̂ × E + = n̂ × E − , (5.264)

where E + is the electric field at a point just above the sheet and E − is the electric field
at the same point just below the sheet.
(b) Derive the tangential volumetric polarization current in the thin dielectric slab, and
collapse that to a surface electric current (Js ) to show that
ik0
− (εr − 1)t [(n̂ × E) × n̂] = Js ,
η0
or equivalently,
(n̂ × E) × n̂ = Zs Js , (5.265)
where Zs is the equivalent “sheet impedance” of the thin dielectric slab:
iη0
Zs = . (5.266)
(εr − 1)k0 t
Here k0 is the wave number in the background and η0 is the characteristic impedance.
Using the boundary condition for magnetic fields, it can be shown that

n̂ × (H + − H −) = Js . (5.267)

(c) Consider an infinite impedance sheet in the x–y plane of a Cartesian coordinate system
being illuminated by a plane wave from the Z > 0 half-space propagating in the
direction k̂i = sin θi x̂ − cos θi ŷ. Find the reflection and transmission coefficients for
TE and TM.
PROBLEMS 311

(d) Compare the results from part (c) with the results obtained from Example 5-4 in the
limit as k0 |ε |t approaches zero.

5.11 Consider an impedance sheet made from a thin conducting dielectric material with
εr = 1 + iσ /ωε . In this case the impedance sheet becomes a resistive sheet (Rs ). Choose the
conductivity and thickness so that the surface resistivity Rs = η0 . In the literature this sheet
is referred to as a Salisbury sheet. Calculate the plane-wave reflection coefficient at normal
incidence for this resistive sheet placed λ /4 above a perfect electric conductor, as shown in
Fig. P5.11.

k̂i

Salisbury Rs = η0
sheet
x
PEC

Figure P5.11: A Salisbury sheet placed above a perfect electric conductor at a height of λ /4.

5.12 The polarization matching factor for an antenna is defined as

PMF = p̂i · P̂a ,

where p̂i is the polarization unit vector of the wave and P̂a is the polarization vector of the
receive antenna. Find the polarization matching factor for an antenna with polarization

(a) linear vertical polarization,

(b) left-hand circular polarization,

(c) elliptical polarization with axial ratio AR = −2 and tilt angle 30◦ ,

when illuminated by a right-hand circularly polarized plane wave.

5.13 An elliptically polarized plane wave in air with axial ratio AR = 3 is obliquely incident
upon a dielectric half-space medium with εr = 3 and µr = 1. Choose the angle of incidence so
that the reflected wave is linearly polarized (along TE polarization).
312 Chapter 5 Electromagnetic Plane Waves

5.14 Consider a dielectric slab of thickness t and relative permittivity and permeability
(ε2 , µ2 ) between two media with relative constitutive parameters (ε1 , µ1 ) and (ε3 , µ3 ).
(a) Find the fields in each region for a normal incident plane wave from region 1 on the
slab.
(b) Find the ratio of the tangential electric field to the magnetic field at the interface
between regions 1 and 2. This is the input wave impedance. Show that when the slab
√
thickness t = λ0 /4n2 with n = ε2 µ2 and λ0 being the free-space wavelength, the wave
input impedance is
η2
Zin = 2 .
η3
(c) Use the result of part (b) to show the input impedance for a Bragg mirror as given by
Eq. (5.221).
5.15 Consider a uniaxial medium with a permittivity matrix
 
εxx 0 0
ε = ε0  0 εxx 0  .
0 0 εzz

(a) Find n1 and n2 as given by Eqs. (5.56) and (5.57) for a plane wave propagating in the
direction
k̂ = sin θ x̂ + cos θ ẑ .

(b) Find an expression for the electric and magnetic fields.

5.16 Repeat Problem P5.15 for a uniaxial medium with permittivity tensor
 
εxx 0 0
ε = ε0  0 εyy 0  .
0 0 εyy

5.17 Consider a homogeneous isotropic medium with constitutive parameters µ0 µ1 and

ε0 ε1 , as shown in Fig. 5-13. The lower half-space is a uniaxial medium with permittivity
matrix  
ε2 0 0
ε = ε0  0 ε2 0 
0 0 εz2
and permeability µ0 µ2 . Find the reflection and transmission coefficients for a wave impinging
on the interface between the two media from above at an oblique incidence angle θi .
 6 Chapter
Cartesian Wave Functions:
Guiding Structures and Resonators

Chapter Contents Objectives

Overview, 314 Upon learning the material presented in this chapter, you
6-1 The Dielectric Plate Waveguide, 315 should be able to:
6-2 Guided Waves on Impedance Surfaces, 321
6-3 Practical Realization of Reactive Impedance 1. Develop intuition about wave-guiding structures,
Surfaces, 324 other than the standard transmission lines, that
6-4 Isotropic Reactive Impedance Surfaces, 329 can support propagation of electromagnetic
6-5 The Rectangular Waveguide, 335 waves.
6-6 Transmission-Line Circuit Model for
Waveguides, 357 2. Understand the properties of guided electro-
6-7 Other Modal Solutions for Rectangular magnetic waves along dielectric slabs with finite
Waveguides, 363 thickness and metallic waveguides, such as cutoff
6-8 Modal Expansion of Field Quantities, 368 frequencies of different modes, dispersion
6-9 Calculus of Variation for Estimation of behavior of wave propagation, low attenuation
Resonant Frequencies in General rate properties, and so on.
Cavities, 382
Chapter Summary, 388 3. Realize the orthogonality properties of different
Problems, 393 modes supported by general waveguides and the
fact that power carried in a multimode wave-
guide is the sum of power carried by the
individual modes.

4. Calculate the resonant frequencies of a rectangu-

lar cavity and its quality factor when its walls are
made from good conductors.

5. Calculate the cutoff frequencies of waveguides

and cavities of general shape using a variational
formulation.

313
314 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Overview
In this chapter we study transverse electric and transverse magnetic guided wave structures
that are commonly used in microwave engineering. The chapter begins with a simple
nonmetallic dielectric structure known as a dielectric plate waveguide. It is shown that such
structures can support both TM and TE modes that can propagate along the dielectric slab
with exponentially decaying fields in the medium surrounding it, provided that the index of
refraction of the dielectric slab is higher than that of its surrounding medium. It is shown
that there are discrete modes of propagation with different propagation constants that are
dependent on the slab thickness and indices of refraction of the slab and its surrounding
medium. The concept of cutoff frequency for each mode is introduced as the frequency
below which that specific mode (with its specific field pattern) cannot propagate along the
dielectric slab. It is shown that for the dominant mode (TM0 ), there is no cutoff frequency. The
principle of nonmetallic waveguiding of a dielectric slab is a precursor for fiber optics, which
today is the most prevalent waveguide structure in use. Also, the guiding mechanism of the
dielectric slab or a modified metal-backed dielectric slab describes the coupling mechanism
between different transmission lines and antennas on printed circuit boards. The concept of
reactive impedance surface for guiding TE and TM modes is also examined and shown that
inductive surfaces can support TM waves and capacitive surfaces can support TE waves. To
realize reactive impedance surfaces, metallic surfaces with periodic grooves are introduced
and design procedures for creating inductive or capacitive surfaces, using transmission-line
theory, are presented. Also, a thin metal-backed substrate with periodic metallic patches on
top is shown to act as a reactive impedance surface with reactance values that can be controlled
by the substrate thickness, index of refraction, and the gap between the metallic patches.
Next, wave propagation inside closed metallic pipes of fixed cross-sectional geometry
along the pipe, known as metallic waveguides, is considered. Specifically, a waveguide with
rectangular cross section is considered. The method of separation is used to solve a two-
dimensional wave equation for both electric (TM) and magnetic (TE) Hertz vector potentials
having only one component along the waveguide axis. The field expressions for different
modes of propagation are expressed in terms of the solution to the 2-D wave equation,
known as the eigenfunction. It is shown that wave propagation in waveguides is, in general,
dispersive. That is, waves at different frequencies propagate with different velocities. Also,
the notion of , which represents the speed at which the envelope of a signal propagates in the
waveguide, is presented and shown to be different from the phase velocity. A perturbation
method that allows for computation of the attenuation rate in a rectangular waveguide is
presented as well. Transmission-line circuit models for each waveguide mode are introduced
as a means to simplify analysis of certain discontinuities along the length of waveguides. A
rectangular cavity structure made from a segment of a rectangular waveguide is also studied.
Finally, the properties of eigenfunctions and eigenvalues (cutoff wave number) of metallic
waveguides of arbitrary cross section are studied. It is shown that the eigenvalues of the
2-D Laplacian operator that defines the cutoff wave numbers of the waveguide are shown
to be always positive and real numbers, and their corresponding eigenfunctions are shown to
be orthogonal to each other within the waveguide cross section. Using this property of the
eigenfunctions, it is shown that the transverse electric fields for two TM or two TE modes or
for one TM and one TE mode are orthogonal. A variational formulation for the calculation
6-1 The Dielectric Plate Waveguide 315

of the eigenvalues of an arbitrary cross section waveguide is presented, and it is shown that
such a formulation can be used in a numerical method, known as the finite element method,
to find the eigenvalues and eigenfunctions. A similar procedure is presented for estimating the
resonant frequencies of a metallic cavity of arbitrary shape.

6-1 The Dielectric Plate Waveguide

In this example, an inhomogeneous medium with a step 1-D dielectric variation along the
x-axis is considered. The resulting geometry is a dielectric plate of thickness t, as shown in
Fig. 6-1. Let’s assume the dielectric plate has a permittivity and permeability given by ε2
and µ2 , respectively.

6-1.1 Transverse Magnetic (TM) Waves

Since we are seeking wave guidance along z, and the medium permittivity and permeability
variations are independent of z and y, for TM waves (see Eq. (5.91))

Πℓz (x, z) = ψℓ (x) eiβ z

with ψ satisfying
d 2 ψℓ 2
+ kℓc ψℓ = 0 . (6.1)
dx2
2 = k2 − β 2 . Note that the phase-matching
where subscript ℓ refers to regions 1, 2, or 3 and kℓc ℓ
condition is implicitly enforced by making β independent of the regions. Wave guidance along
the ±x-direction is desired to be suppressed, hence k1c and k3c are chosen as pure imaginary
quantities; in other words, k1c = k3c = iν . Solution of Eq. (6.1) can be obtained in terms of
two sets of even and odd functions.

TM Odd Modes
Considering the odd solutions, Πoℓz in the three regions is given by
t
Πo2z (x, z) = A sin(k2c x) eiβ z |x| < , (6.2a)
2

x
1 μ1, ε1
μ2, ε2 z
t
2
3 μ1, ε1

Figure 6-1: Geometry of a dielectric slab supporting wave propagation along the z-axis.
316 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

t
Πo1z (x, z) = Be−ν x eiβ z x> , (6.2b)
2
and
t
Πo3z (x, z) = −Beν x eiβ z x<− , (6.2c)
2
where
2
k2c = k22 − β 2 = ω 2 µ2 ε2 − β 2 (6.3a)
and
ν 2 = β 2 − k12 = β 2 − ω 2 µ1 ε1 . (6.3b)

The field components are obtained from Eqs. (5.102a) and (5.102b):
2 sin(k x) eiβ z

E2z = Ak2c 2c
Region 2,
H2y = iAωε2 k2c cos(k2c x) eiβ z

E1z = −Bν 2 e−ν x eiβ z
Region 1,
H1y = iBωε1 (−ν )e−ν x eiβ z
and

E3z = Bν 2 eν x eiβ z
Region 3.
H3y = iBωε1 (ν )eν x eiβ z

Continuity of tangential electric and magnetic fields must be enforced at x = ±t/2:

 t
2
Ak2c sin k2c = −Bν 2 e−ν t/2
2
and
 t
iAωε2 k2c cos k2c = −iBωε1 ν e−ν t/2 .
2
The ratio of the above two equations gives an equation relating k2c to ν :
k2c  t ν
tan k2c = . (6.4a)
ε2 2 ε1
Also adding the two equations given by Eq. (6.3), we get
2
k2c + ν 2 = ω 2 (µ2 ε2 − µ1 ε1 ) . (6.4b)

Equations (6.4a) and (6.4b) must be solved simultaneously to find ν and k2c , from which the
propagation constant β can be obtained. For a given solution (ν , k2c , β ), B can be obtained
from  2 
k2c t
B = −A sin k2c eν t/2 .
ν 2
6-1 The Dielectric Plate Waveguide 317

ε
y

Figure 6-2: Graphical solution for determining TM odd modes supported by a dielectric-
plate waveguide. The solution consists of the three intersection points between the blue curves
representing Eq. (6.5a) and the red curve representing Eq. (6.5b). It is shown that there is no
cutoff frequency for TM waves. In this example ε1 = ε0 , ε2 = 3ε0 , µ1 = µ2 = µ0 , and t = 1.8λ .

Defining new variables X = k2ct/2 and Y = ν t/2, Eqs. (6.4a) and (6.4b) can be written as
ε1
X tan X = Y (6.5a)
ε2
and
 t √ 2
2 2
X +Y = ω µ2 ε2 − µ1 ε1 . (6.5b)
2
Requiring µ2 ε2 > µ1 ε1 , the second equation in Eq. (6.5) represents a circle in the x–y plane.
Plotting both curves and finding the intersection of these curves (for Y > 0), the possible
solutions for Eq. (6.5) can be obtained graphically (see Fig. 6-2) . It is interesting to note
that independent of frequency (ω ) and the slab thickness (t) there is always a solution for k2c
and ν . That is, there is no cutoff frequency for TM odd mode propagation. A cutoff frequency
is defined as a frequency for which ν = 0 (no decay along x axis). Hence the cutoff frequencies
√
for a specific mode correspond to ν = 0 or β = ω µ1 ε1 , which leads to
t
tan k2c =0
2
318 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

or
t
k2c = nπ n = 0, 1, 2, . . .
2
or
ωt √
µ2 ε2 − µ1 ε1 = nπ n = 0, 1, 2.
2
As noted earlier, there is no cutoff frequency for TM0 . The cutoff frequency for TMn is given
by
TMo n
fc n = √ n = 1, 2, . . . (TM odd modes) (6.6)
t µ2 ε2 − µ1 ε1

TM Even Modes
For even modes of TM waves, the counterparts of Eqs. (6.2a) through (6.2c) are

Πe2z (x, z) = A cos(k2c x) eiβ z , (6.7a)

Πe1z (x, z) = Be−ν xeiβ z , (6.7b)
and
Πe3z (x, z) = Be+ν xeiβ z . (6.7c)

The field components can be obtained as before with the exception of exchanging sin(k2c x) for
cos(k2c x) and vice versa and keeping track of sign changes. Enforcing the boundary condition
at x = ±t/2 results in the following equation:
 
k2ct k2ct ε2 ν t
− cot = . (6.8)
2 2 ε1 2
Using the same definitions for X and Y as before,

 ε1
 − X cot X = Y , and
ε2  √ 2
 X +Y 2 = ω t µ2 ε2 − µ1 ε1 .
 2
2
Figure 6-3 shows the possible solutions for the even TM modes. The cutoff frequencies
correspond to  
k2c t
cot =0,
2
or
k2c t (2n + 1)
= π,
2 2
which is equivalent to
 
ωt √ 2n + 1
µ2 ε2 − µ1 ε1 = π, n = 0, 1, 2, . . . ,
2 2
6-1 The Dielectric Plate Waveguide 319

ε
y
ε

Figure 6-3: Graphical solution for determining TM even modes supported by a dielectric-plate
waveguide. In this example ε1 = ε0 , ε2 = 3ε0 , µ1 = µ2 = µ0 , and t = 1.8λ0.

or

TMen (2n + 1)
fc = √ n = 0, 1, 2, . . . (TM even modes) (6.9)
2t µ2 ε2 − µ1 ε1

Combining the even and odd modes, the cutoff frequencies for TM modes are given by

n
fcTMn = √ n = 0, 1, 2, . . . (all TM modes) (6.10)
2t µ2 ε2 − µ1 ε1

6-1.2 Transverse Electric (TE) Modes

As shown previously, TE modes are the dual of TM modes and the corresponding field
distributions can be obtained using the duality relations. As such, the dispersion relations
for TE modes are also the dual of their TM counterpart. For the TE modes with odd Πmz (x, z)
we have  
k2ct k2ct µ2 ν t
tan = , (TE odd modes) (6.11a)
2 2 µ1 2
320 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

and for even Πmz (x, z),

 
k2ct k2ct µ2 ν t
− cot = . (TE even modes) (6.11b)
2 2 µ1 2
Since Eq. (6.3) is unchanged under the duality relations, the cutoff frequencies obtained for
TM modes are the same as those for the TE modes. It is emphasized here that guided TE
and TM modes propagate unattenuated, independently of both the frequency and the slab
thickness.

6-1.3 Dielectric-Coated Conductor

The modal analysis of the previous section can easily be extended to solve the waveguide
mode analysis for the dielectric-coated planar metallic surface shown in Fig. 6-4. Such a
geometry is often encountered in printed circuit boards for high-frequency applications. The
boundary condition at x = 0 mandates that the tangential component of the electric field vanish
as x approaches the metal surface. Recalling that the electric field distribution of odd TM
modes is given by
2
E2z (x, z) = Ak2c sin(k2c x) eiβ z ,
the required boundary condition for these modes is automatically satisfied. Also, the electric
field of even TE modes is given by

E2y (x, z) = −iω µ2 Ak2c sin(k2c x) eiβ z ,

which also satisfies the boundary condition. For TM modes, the modes of propagation can be
obtained from Eq. (6.6), and for TE modes,
  
k2ct k2ct µ2 ν t 
− cot = , and 

2 2 µ1 2
 2   2   2 (6.12)
k2c t νt ωt 

+ = (µ2 ε2 − µ1 ε1 ) . 
2 2 2

μ1, ε1
μ2, ε2
t/2
z
σ=
8

Figure 6-4: Geometry of a metal-backed dielectric slab capable of supporting guided waves.
6-2 Guided Waves on Impedance Surfaces 321

6-2 Guided Waves on Impedance Surfaces

In the previous section, we examined TE and TM wave propagation modes on dielectric
slabs and dielectric-coated conductors. In the literature, the latter is referred to as substrate
modes or surfaces waves. In this section we intend to show that surface wave propagation can
also be supported by certain nonresistive materials known as reactive impedance surfaces.
Mathematically, such surfaces can be represented by an impedance boundary condition that
was first introduced by Leontovich for planar surfaces and surfaces with large radii of
curvature compared with the wavelength. For such surfaces the tangential electric field is
linearly proportional to the tangential magnetic field and the constant of proportionality is
known as the surface impedance. Denoting the surface impedance by Zs , the impedance
boundary condition for the electric and magnetic fields just above the surface is given by

n̂ × [n̂ × E(r)] = −Zs n̂ × H(r) . (6.13)

Here Zs can have both the real and imaginary parts but, as will be shown later, surface waves
can only propagate along impedance surfaces whose surface impedance is reactive. Using
the boundary condition given by Eq. (6.13), the solution to Maxwell’s equations can be
found using only fields exterior to the object with the impedance surface. This significantly
simplifies the solution. In practice, objects with impedance surfaces can only be realized in
approximate form. However, it should be emphasized that the impedance boundary condition
given by Eq. (6.13) should be independent of the wave attributes and is only a property of
the surface itself. To elaborate on this further, let us consider a planar interface between two
√ √
dielectric half-spaces with indices of refractions n1 = µ1 ε1 and n2 = µ1 ε2 , illuminated
by a plane wave from Region 1. For the TM and TE polarizations, referring to Eqs. (5.99) and
(5.102), it can be shown that the ratio of the the tangential electric field to tangential magnetic
field at the interface (z = 0) are given respectively by

Ex β2
= (TM) (6.14a)
Hy ωε2
and
Ey ω µ2
= . (TE) (6.14b)
Hx β2
Of course, these cannot be used as the surface impedance for the lower medium as the
two ratios are functions of wave polarization and incident angle. For an arbitrary source
in the upper half-space, the fields can be expanded in terms of a continuous spectrum of
plane waves with all propagation directions and polarizations, but because the interface will
present different surface impedances, in general the lower half-space cannot be modeled by
an impedance surface. However, under the condition that the the index of refraction of the
lower medium is much larger than that of the upper medium (n2 ≫ n1 ), then the application
of Snell’s law of refraction given by
n1
sin θ2 = sin θ1
n2
322 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

predicts that θ2 ≈ 0, independently of θ1 , and also β2 ≈ k2 . Under such a condition,

Eqs. (6.14a) and (6.14b) simplify to
r
Ex µ2
= = η2 (TM)
Hy ε2
and
r
Ey µ2
= = η2 , (TE)
Hx ε2
and the lower conducting medium can be modeled by an impedance surface ηc = η2 .
A metallic medium with high but finite conductivity presents itself as a medium with
an impedance boundary condition. Suppose the conductivity of the metallic medium is
represented by σ . According to Eq. (1.74), the dielectric constant of this medium is given
by
σ
εm = ε0 + i . (6.15)
ω
At frequencies where σ /(ωε0 ) ≫ 1, the metallic medium can be approximated by an
impedance surface having a surface impedance of
s r r
µ0 ω µ0 ω µ0
ηc = σ ≈ = (1 − i) . (6.16)
ε0 + i ω iσ 2σ

Hence, the surface is both resistive and inductive.

6-2.1 Impedance Surface Waveguide

In this section we demonstrate how an impedance surface can guide TM and TE waves,
through which we lay the foundation for the concept of a single-conductor transmission line.
Consider a planar impedance surface with surface impedance Zs placed in the y–z plane, as
shown in Fig. 6-5. Requiring wave propagation along z and exponential decay along x, the
scalar potential in the upper half space takes the following form:

ψ1 (x, z) = Ae−ν x eiβ z , (6.17)

z
Zs

Figure 6-5: A planar impedance surface in the y–z plane supporting TM and TE guided waves.
6-2 Guided Waves on Impedance Surfaces 323

where ν and β must satisfy

β 2 − ν 2 = k2 . (6.18)
For TM waves, the expressions for the tangential electric and magnetic fields can be obtained
from Eqs. (5.102a) and (5.102b) and are given by

Ez (x, z) = −ν 2 e−ν x Aeiβ z (6.19a)

and
Hy (x, z) = −iωεν Ae−ν xeiβ z . (6.19b)

The application of the boundary condition given by Eq. (6.13) to Eq. (6.19a) and
Eq. (6.19b) at z = 0 provides the following equation for ν :

ν = iωε Zs . (6.20)

This equation indicates that if the surface impedance is pure imaginary and inductive, i.e.,
if Zs = −iXi , then ν becomes a positive real number, which guarantees exponential decay
along x. In this case the propagation constant is given by

q
β= k2 + ω 2 ε 2 Xi2 . (TM waves) (6.21)

A similar analysis can be carried out for TE waves by using Eqs. (5.105a) and (5.105b) to
find the tangential electric and magnetic fields in the upper half-space:

Hz (x, z) = −ν 2 e−ν x Aeiβ z , (6.22a)

Ey (x, z) = iω µν Ae−ν x eiβ z , (6.22b)
and
iω µ
ν= . (6.22c)
Zs
Now, note that if the surface impedance is pure imaginary and capacitive, i.e, Zs = iXc , then ν
becomes a positive real number and TE wave propagation becomes possible. In this case the
propagation constant is given by

s  2
ωµ
β= k2 + . (TE even waves) (6.23)
Xc

In summary, impedance surfaces with purely reactive surface impedance can support surface
wave propagation. Surfaces with inductive surface impedance can only support TM waves
and surfaces with capacitive surface impedance can only support TE waves.
324 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Example 6-1: Surface Wave on a Lossy Conductor

Good conductors can be represented as impedance surfaces. Can good conductors support
surface waves?
Solution: From Eq. (6.16), the surface impedance of a good conductor is given by
r
ω µ0
Z s = ηc = .
iσ
This can be used in Eq. (6.20) for TM or Eq. (6.22c) for TE cases to find ν ; that is,
r r
ω µ0 ω µ0
νTM = iωε = ωε (1 + i)
iσ 2σ
and
r
iω µ ωσ
νTE = p =µ (−1 + i) .
ω µ /iσ 2 µ0

Since the decay factor along x is complex, there is some propagation along the x direction,
and hence wave propagation along the surface is not supported.

6-3 Practical Realization of Reactive Impedance Surfaces

As discussed in Section 6-2.1, surfaces with purely inductive surface impedance can support
TM surface waves, and surfaces with purely capacitive surface impedance can support
TE surface waves. There are a number of approximate approaches available to realize
surfaces with reactive impedances. One conventional approach is through the utilization of
a corrugated conductor. Let us consider a one-dimensional periodic surface, as shown in
Fig. 6-6, with periodicity in the z-direction and the period chosen to be much smaller than
the wavelength. Figure 6-6 shows three different periodic surfaces with different groove
geometries. Here it is assumed that w and p are much smaller than λ at the operating
frequency. For the TM case, the electric field has both z and x components and the z component
of the electric field can couple between the two sides of the corrugation geometry to form a
TEM wave inside the grooves.

6-3.1 Rectangular Groove

Considering the geometry shown in Fig. 6-6(a), each groove can be considered a parallel-plate
transmission line. This transmission line is filled with a dielectric of constitutive parameters
√
µ0 and ε0 , and as a result, the propagation constant is k0 = ω µ0 ε0 . To find the characteristic
impedance of this line, the capacitance and inductance per unit length of the line are required.
Since the field is uniform along y, a finite length s of the line along the y-axis can be used to
find such a parameter. Consider a segment of the transmission line of length ℓ, as shown in
Fig. 6-7. This forms a parallel-plate capacitor with capacitance
6-3 Practical Realization of Reactive Impedance Surfaces 325

x x

w w p
d z d z

(a) (b)
w
x

w d
d z

(c) (d)

Figure 6-6: Geometries for three different periodic surfaces made up of a perfect electric
conductor with period P ≪ λ . These surfaces can support TM surface waves.

ε0 sℓ
CL = (F). (6.24)
w
Hence the capacitance per unit length is given simply by
CL ε0 s
C= = (F/m) . (6.25)
ℓ w
For transmission line, the propagation constant in terms of the capacitance (C) and inductance
(L) per unit length is given by
√ √
k0 = ω LC = ω µ0 ε0 . (6.26)

z
x

y l

Figure 6-7: Geometry of a parallel-plate transmission line of length ℓ.

326 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Hence, the inductance per unit length of this line can be evaluated from
µ0 ε0 µ0 w
L= = (H/m) . (6.27)
C s
Using Eqs. (6.25) and (6.26), the characteristic impedance of the parallel-plate transmission
line is given by r r
L w µ0 w
Z0 = = = η0 . (6.28)
C s ε0 s
As depicted in Fig. 6-6(a), these transmission lines are short-circuited at z = 0. The input
impedance of the short-circuited line can be obtained from
ZL − iZ0 tan k0 d
Zin = Z0 , (6.29)
Z0 − iZL tan k0 d
with ZL = 0. Hence the impedance seen from the aperture of the corrugation is given by

V (d) iw
Zin = = − η0 tan(k0 d) . (6.30)
I(d) s
Noting that Z w
V (d) = Ez (d, z) dz ≈ w Ez (d) ,
0
and noting that the surface current density is proportional to Hy (x), the total current on the
metallic strip of width s, as shown in Fig. 6-7, is given by
Z s
I(d) = Hy (d) dy ≈ s Hy (d) ,
0

the surface impedance of the corrugated surface, in view of Eq. (6.30), is found to be

Ez (d) s V (d)
Zs = = = −iη0 tan(k0 d) . (6.31)
Hy (d) w I(d)

Equation (6.31) indicates that when k0 d < π /2, or equivalently, when d < λ0 /4, the reactive
impedance surface is inductive and can support TM wave propagation. It is noted that the
surface impedance is not a function of the aperture width (w). To support TE waves the
periodicity of the metallic walls should be along the y-direction. In this case, the electric
field (Ey ) can excite a TEM wave inside the grooves and, following a similar procedure, the
surface impedance can also be found to be

Ey (d)
ZsTE = = −iη0 tan(k0 d) .
Hz (d)

However, since the surface impedance has to be capacitive, π /2 < k0 d < π , or equivalently,
λ0 /4 < d < λ0 /2. Of course in this case the depth of the grooves is quite large.
Now considering the geometry of Fig. 6-6(b), we realize that over one period the surface
6-3 Practical Realization of Reactive Impedance Surfaces 327

(1) (2)
impedance takes on two different values: Zs = 0 and Zs = −iη0 tan(k0 d). Since the period
is much smaller than the wavelength, it is reasonable to assume that the effective surface
impedance is the weighted average between these two values; that is,
(2)
0 × t + Zs × w w (2) w
Zseff = = Zs = −i η0 tan(k0 d) . (6.32)
p p p

6-3.2 Triangular Groove

The computation of surface impedance for the triangular grooves shown in Fig. 6-6(c) is more
involved. As in the other cases, for a TM case, a TEM mode is excited inside the grooves,
and therefore the transmission line approach can be used to obtain the surface impedance.
As shown in Fig. 6-6(d), the triangular groove can be approximated by a staircase geometry
where at each step a parallel-plate transmission line can be visualized with a plate separation
of x
w(x) = p . (6.33)
d
The capacitance and inductance of this transmission line are now a function of position and
are given by Eqs. (6.25) and (6.27). To find the input impedance at the aperture of the groove,
transmission-line equations must be used directly. The telegrapher’s equation (considering
our time convention of e−iω t ) can be written as
−∂ V (x)
= iω L(x) I(x) (6.34a)
∂x
and
−∂ I(x)
= iω C(x) V (x) , (6.34b)
∂x
where L(x) and C(x) are respectively the inductance and capacitance per unit length of the
line. The negative signs on the left-hand side indicate propagation along the negative x-
direction. Taking the derivative of Eq. (6.34a) and then using Eq. (6.34b) renders

∂ 2V (x) L′ (x) ∂ V (x)

− + ω 2 L(x) C(x) V (x) = 0 . (6.35)
∂ x2 L(x) ∂ x
In a similar manner, taking the derivative of Eq. (6.34b) and using Eq. (6.34a), we have

∂ 2 I(x) C′ (x) ∂ I(x)

− + ω 2 L(x) C(x) I(x) = 0 . (6.36)
∂ x2 C(x) ∂ x
Upon substitution of the inductance and capacitance values and noting that
 
′ d ε0 s ε0 sd 1
C (x) = =−
dx (p/d)x p x2
and
328 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

d  pµ0 s  p
L′ (x) = x = µ0 ,
dx ds ds
Eqs. (6.35) and (6.36) simplify to

d 2V (x) 1 dV (x)
− + k02 V (x) = 0 (6.37a)
dx2 x dx
and
d 2 I(x) 1 dI(x)
+ + k02 I(x) = 0 . (6.37b)
dx2 x dx
Recalling that the Bessel differential equation is of the form
 
d2 f 1 d f n2
+ + 1− 2 f = 0 ,
dx2 x dx x
the solution to Eq. (6.37b) is the Bessel function of zeroth order (n = 0); that is,

I(x) = A J0 (k0 x) + B Y0 (k0 x) . (6.38)

Using Fig. 6.34b,

−k0
V (x) = [A J0′ (k0 x) + B Y0′ (k0 x)] . (6.39)
iω C(x)
The boundary condition mandates that

V (0) = 0 ,

which requires
B=0.
Noting that
J0′ (k0 x) = −J1 (k0 x) ,
the expression for the voltage inside the triangular grooves simplifies to
k0 A
V (x) = J1 (k0 x) . (6.40)
iω C(x)
Also, the expression for the current function is given by

I(x) = A J0 (k0 x) . (6.41)

Hence the input impedance at the aperture of the groove (x = d) is given by

V (d) k0 J1 (k0 d)
Zin = =
I(d) iω C(d) J0 (k0 d)
−ik0 w J1 (k0 d) w J1 (k0 d)
= = −iη0 .
ωε0 s J0 (k0 d) s J0 (k0 d)
6-3 Practical Realization of Reactive Impedance Surfaces 329

10

Inductive
5
J1(k 0 d )/[J0(k0 d )]
surface

Capacitive
−5 surface

−10
0 1 2 3 4 5 6
k0 d

Figure 6-8: Plot of the function J1 (k0 d)/[J0 (k0 d)] as a function of the triangular groove depth
(k0 d), used for calculating surface impedance.

As in Eq. (6.31), the surface impedance is

s V (d) J1 (k0 d)
Zs = = −iη0 . (6.42)
w I(d) J0 (k0 d)
Again the surface impedance of the triangular groove is not a function of aperture width (w).
This surface impedance is inductive for values of k0 d so long as

J1 (k0 d)
>0.
J0 (k0 d)
It can be shown that this happens when

k0 d < 2.405

or
d < 0.38λ .
Figure 6-8 shows a plot of J1 (k0 d)/[J0 (k0 d)] as a function of groove depth (k0 d) and
indicates the groove depth values when the surface impedance is inductive or capacitive. It
should also be noted here that the one-dimensional corrugated surfaces exhibit anisotropic
surface impedances.
330 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

μ0 μ1
ε0 ε1
(a)

z
E0

(b) i
μ0 , ε0
μ0 μ1
d ε0 ε1
t x

Figure 6-9: (a) Isotropic reactive impedance surface created by a periodic array of square
metallic patches on a thin metal-backed dielectric substrate, and (b) the same substrate illuminated
by a TM-polarized plane wave.

6-4 Isotropic Reactive Impedance Surfaces

In the previous section we showed that metals with one-dimensional surface corrugation can
support TM and TE surface waves when the electric field is perpendicular to the grooves’
walls. Also, to support TE waves we found that the depth of the grooves must be longer
than λ0 /4. For many applications, it is desirable to have reactive impedance surfaces that
are isotropic and compact in terms of surface thickness. The cost of fabrication is another
issue to consider. In this section we examine the design of capacitive and inductive reactive
impedance surfaces using printed circuit technology. The structure is composed of a periodic
arrangement of metallic patches over a grounded dielectric substrate with relative permittivity
ε1 and permeability µ1 . The configuration of this structure is shown in Fig. 6-9(a). The
metallic square patches form a 2-D periodic surface with a period p that is assumed to be
much smaller than a wavelength. To evaluate the performance of this structure, its interaction
with an incident plane wave at oblique incidence and TM or TE polarization is considered.
6-4 Isotropic Reactive Impedance Surfaces 331

Since the periodicity is much smaller than a wavelength, only reflected and refracted plane
waves are generated. Also, to analyze this structure, the behavior of the field over one period
can be considered, as depicted in Fig. 6-9(b). Referring to Section 5-4.2, the TM fields in
regions 0 (z ≥ d) and 1 (0 ≤ z ≤ d) can be written as

E0x (x, z) = E0 cos θi (e−ik0 cos θi z − Γ eik0 cos θi z )eik0 sin θi x , (6.43a)
E0 −ik0 cos θi z
H0y (x, z) = − (e + Γ eik0 cos θi z )eik0 sin θi x , (6.43b)
η0
E1x (x, z) = E1 cos θt (e−ik0 cos θt z − eik0 cos θt z )eik0 sin θi x , (6.43c)
and
E1 −ik1 cos θt z
H1y (x, z) = − (e + eik1 cos θt z )eik0 sin θi x , (6.43d)
η1
where we have imposed the phase-matching condition and the fact that E1x (x, 0) = 0 due to
the boundary condition at z = 0. Also, as before,
s
sin2 θi
cos θt = 1 − .
µ1 ε1

The x component of the incident field excites an electric field across the gap between the
patches, and the electric field produces the field expressions we have in Eq. (6.43c). Strictly
speaking, the electric fields given by Eqs. (6.43a)–(6.43d) are valid for values of z slightly
above or below the interface. Since the same voltage gap is exciting the electric fields on both
sides of the gaps, the continuity of tangential electric fields at z = d can be imposed; that is,

E0 cos θi (e−ik0 cos θi d − Γ eik0 cos θi d ) = −2iE1 cos θt sin(k1 cos θt d) . (6.44)

The incident field also excites an electric surface current on the metallic patches whose
continuity is maintained by the polarization current across the gap. In fact, the gap is acting
like a capacitor that facilitates the flow of the electric current. The voltage across this capacitor
is given by
V ≈ E0x (0, d) p , (6.45)
e the surface
where p is the period. Denoting the capacitance per unit length of the gap by C,
current density can be obtained from
e = −iωCep E0x (0, d) .
Jx = −iωCV (6.46)

The boundary condition mandates that

n̂ × (H0 − H2 ) = Jx . (6.47)
332 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Using Eqs. (6.43b) and (6.43d) in Eq. (6.47) provides a second equation for the unknowns:
E0 −ik0 cos θi d E1
(e + Γ eik0 cos θi d ) − 2 cos(k1 cos θt d) = −iωCep[−2iE1 cos θt sin(k1 cos θt d)] .
η0 η1
(6.48)
According to Eq. (6.13), the surface impedance is defined by
E0x
Zs = − . (6.49)
H0y

Using Eqs. (6.48) and (6.44), it can be shown that

η1 cos θt tan(k1 cos θt d)

ZsTM = −i . (6.50)
e η1 cos θt tan(k1 cos θt d)
1 − ωCp
This result is significant because even for a thin substrate where k1 cos θt d < 1, the surface
can be made to present either an inductive or a capacitive impedance. For example, assuming
k1 cos θt d < 1, if
ωCepη1 cos θt tan(k1 cos θt d) < 1 , (6.51)
the surface is inductive, and when

ωCepη1 cos θt tan(k1 cos θt d) > 1 , (6.52)

the surface is capacitive. At resonance,

ωCepη1 cos θt tan(k1 cos θt d) = 1 , (6.53)

and the surface acts as a perfect magnetic conductor. Strictly speaking, the surface impedance
should not be a function of incident angle. In practice, if µ1 ε1 ≫ 1, then cos θt ≈ 1 and the
requirement is satisfied.
The surface impedance presented by Eq. (6.50) can also be obtained by using a
transmission-line analogy. The lower medium can be represented by a short-circuited
transmission line of length d having a characteristic impedance of

Z1 = η1 cos θt (6.54)

and a propagation constant of

β1 = k1 cos θt . (6.55)
This transmission line is connected to a capacitor (C) that represents the gap between
two adjacent metallic patches and a semi-infinite transmission line having a characteristic
impedance and a propagation constant that are given by

Z0 = η0 cos θi (6.56a)
and
β0 = k0 cos θi . (6.56b)
6-4 Isotropic Reactive Impedance Surfaces 333

Zeq d

g p
μ0 μ1
ε0 ε 1
p Zin

Figure 6-10: The equivalent transmission-line representation for one period of metallic square
patches over a metal-backed dielectric substrate.

The configuration for this equivalent transmission line is shown in Fig. 6-10. The equivalent
impedance of this circuit can be obtained from two shunt impedances, namely the shunt
capacitor and the input impedance of the short-circuited segment of the transmission line.
The input impedance of the short-circuited transmission line is given by

Zin = −iZi tan β1 d = −iη1 cos θt tan(k1 cos θt d) . (6.57)

Hence the equivalent impedance can be obtained simply from

1


V (d) (−iZ1 tan β1 d)
= Zeq = −iω C
I(d) −Z1 tan β1 d + −i1ω C
Z1 tan β1 d
= −i .
1 − ωCZ1 tan β1 d

Noting that V (d) = E0x (d) p and I(d) = −H0y (d) p, it follows that

η1 cos θt tan(k1 cos θt d)

ZsTM = Zeq = −i . (TM mode) (6.58)
e η1 cos θt tan(k1 cos θt d)
1 − ωCp

Under the condition given by Eq. (6.51), this surface is inductive and can support a TM surface
wave in the upper medium with a propagation constant along x (in our example) given by
q
TM
β = ω 2 µ0 ε0 + ω 2 ε02 Xs2

with Xs = iZs .
Now for a TE incident wave with reference to Fig. 6-9(b), the field can be obtained
334 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

by applying the duality to Eqs. (6.43a)–(6.43d). However, duality would not apply to the
metallic patches unless the metallic patches were to be replaced with magnetic conductors.
Nevertheless, using the boundary conditions, continuity of the tangential electric field, and
discontinuity of the magnetic field (Eq. (6.47)), the surface impedance can be obtained. Note
that in this case the electric field is parallel to the surface. Equivalently, we can use the
transmission-line analogy, where the characteristic impedance of the equivalent transmission
lines must be modified to
η0
Z0 =
cos θi
and
η1
Z1 = .
cos θt
Using these, the surface impedance for this polarization takes the following form:

η1
tan(k1 cos θt d)
TE cos θt
Zs = −i . (TE mode) (6.59)
e η1 tan(k1 cos θt d)
1 − ωCP
cos θt

When k1 d ≪ 1, using the first term of the Taylor series expansion of the tangent function in
Eq. (6.59) leads to
η1 k1 d
Zs = −i .
1 − ωCepη1 k1 d
This surface impedance is inductive when ωCepη1 k1 d < 1. The interesting feature of this
polarization is that the surface impedance is independent of the incident angle, as it should
be.
An approximate formulation for the capacitance per unit length of coplanar metallic strips
can be found in the literature.* For a gap dimension of g, the capacitance value is given by
p
ε eff K[ 1 − (g/p)2 ]
Ce = , (6.60)
K(g/p)

where K(t) is the complete elliptic integral given by

Z π /2
dα
K(t) = p
0 1 − t 2 sin2 α
 
π 12 12 · 32 12 · 32 · 52
= 1 + 2 t2 + 2 2 t4 + 2 2 2 t6 + · · · . (6.61)
2 2 2 ·4 2 ·4 ·6

* S. Gevorgian and H. Berg, “Line capacitance and impedance of coplanar strip waveguides on substrates with
multiple dielectric layers,” 31st European Microwave Conference, 2001.
6-5 The Rectangular Waveguide 335

14

12 ε eff = 1
ε eff = 2
10
ε eff = 2.25
8 ε eff = 3
ε0
~

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
p

Figure 6-11: The normalized capacitance per unit length of coplanar strips over a substrate for
different values of εeff as a function of gap-to-strip ratio.

In Eq. (6.60), εeff is the effective dielectric constant formed by the substrate dielectric ε0 ε1
and the air ε0 on both sides of the square patches. Since the electric field is the same on both
sides of the gap it can easily be shown that
1 + ε1
εeff = ε0 .
2
Here we are assuming that the substrate thickness d is much larger than the gap g, so that the
capacitance value is not affected by the ground plane.
Figure 6-11 shows the normalized capacitance (C/ e ε0 ) per unit length of a coplanar strip
capacitor as a function of gap-over-height ratio (g/p) for different values of εeff .

6-5 The Rectangular Waveguide

In this section we consider wave propagation inside a hollow metallic pipe of rectangular
cross section, as shown in Fig. 6-12. In general, hollow, dielectric-filled, or partially filled,
closed metallic pipes of arbitrary cross section but invariant to the z-direction are known as
metallic waveguides. Here we consider a rectangular waveguide filled with a homogeneous
material having permittivity and permeability values denoted by ε and µ , respectively. The
boundary condition over the interior surface of the metallic pipe mandates that the tangential
component of the electric field be zero across the waveguide’s cross section.
For a TM wave, according to Eq. (5.102a),

Ez (x, y, z) = kc2 ψ (x, y) eiβ z ,

336 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

b z
x

a S

Figure 6-12: Geometry of a metallic rectangular waveguide. The cross section geometry is
invariant along the waveguide axis.

for a wave traveling along the +z-direction. Hence the boundary condition on the waveguide
surface S, namely Ez |x,y∈S = 0, where S encompasses the top and bottom surfaces as well as
the side surfaces of the structure shown in Fig. 6-12, mandates that

ψ (x, y) = 0 , x, y ∈ S , (TM mode) (6.62)

which is known as Dirichlet’s boundary condition. For a TE wave according to Eq. (5.105b),

n̂ × E = iω µ n̂ × (∇t ψm × ẑ)eiβ z .

Since n̂ · ẑ = 0, it can easily be shown that

∂ ψm iβ z
n̂ × E = −iω µ e ẑ .
∂n
Therefore for TE modes the boundary condition mandates that

∂ ψm
=0 x, y ∈ s . (TE mode) (6.63)
∂n
This is known as Neumann’s boundary condition.

6-5.1 TM Modes
Let’s now consider the TM solution for wave propagation in the rectangular waveguide. From
Eq. (5.101) in the Cartesian coordinate system, we are seeking a solution for
 2 
∂ ∂2 2
+ + kc ψ (x, y) = 0 , (6.64)
∂ x2 ∂ y2
6-5 The Rectangular Waveguide 337

subject to Dirichlet’s boundary condition given by

ψ (0, y) = ψ (a, y) = 0

on the vertical walls and

ψ (x, 0) = ψ (x, b) = 0
on the horizontal walls. A solution of Eq. (6.64) can be obtained using the method of
separation of variables. Assuming

ψ (x, y) = X (x) Y (y) , (6.65)

for continuous and differentiable functions of X (x) and Y (y), substituting Eq. (6.65) into
Eq. (6.64), and then dividing both sides by X (x) Y (y), results in

X ′′ (x) Y ′′ (y)
+ + kc2 = 0 . (6.66)
X (x) Y (y)
Since X is a function of only variable x and Y is a function of only variable y, the only possible
condition for Eq. (6.66) to be valid is when

X ′′ (x)
= −kx2 (6.67a)
X
and
Y ′′ (y)
= −ky2 . (6.67b)
Y
Hence kc2 = kx2 + ky2 provides the sufficient and necessary condition for the validity of
Eq. (6.66).
Equations (6.67a) and (6.67b) are ordinary second-order differential equations whose
solutions are given by

X (x) = A sin(kx x) + B cos(kx x) (6.68a)

and
Y (y) = C sin(ky y) + D cos(ky y) . (6.68b)

The boundary condition for ψ translates into the following conditions for X (x) and Y (y):

X (0) = X (a) = 0
and
Y (0) = Y (b) = 0 .

These conditions provide

B = D = 0, (6.69a)
338 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

sin(kx a) = 0 , (6.69b)
and
sin(ky b) = 0 . (6.69c)

Equations (6.69b) and (6.69c) provide discrete solutions (eigenvalues) for kx and ky :
mπ
kx = m = 1, 2, 3, . . .
a
and
nπ
ky = , n = 1, 2, 3, . . .
b
Therefore  mπ   nπ 
Ψmn (x, y) = E0 sin x sin y . (6.70)
a b
where E0 = AC. For each integer pair (m, n) corresponding to mode TMmn , the following
propagation constant is obtained:
r  mπ 2  nπ 2
βmn = k2 − − . (6.71)
a b
The corresponding guide wavelength (along the z-direction), defined by λg = 2π /β , is given
by r
1  m 2  n 2
(λgmn )−1 = − − . (6.72)
λ2 2a 2b
The field quantities for TM modes can be directly computed from Eqs. (5.102a) and (5.102b):
 mπ   mπ   nπ 
Ex (x, y, z) = iE0 βmn cos x sin y eiβmn z , (6.73a)
a a b
 nπ   mπ   nπ 
Ey (x, y, z) = iE0 βmn sin x cos y eiβmn z , (6.73b)
b a b
 
mπ 2  nπ 2  mπ   nπ 
Ez (x, y, z) = E0 + sin x sin y eiβmn z , (6.73c)
a b a b
 nπ   mπ   nπ 
Hx (x, y, z) = −iE0 ωε sin x cos y eiβmn z , (6.73d)
b a b
and
 mπ   mπ   nπ 
Hy (x, y, z) = iE0 ωε cos x sin y eiβmn z . (6.73e)
a a b
The wave impedance is defined by
q

2
mπ 2
ẑ × Et Ex Ey k2 − a − nbπ
ZTM = = = = .
Ht Hy −Hx ωε
6-5 The Rectangular Waveguide 339

Among the infinite number of discrete modes, only those for which βmn is a real quantity can
propagate. The rest are evanescent and do not carry energy. The TM cutoff frequency for the
mnth mode corresponds to a frequency for which

βmn = 0

and  mπ 2  nπ 2
k2 = (2π fc )2 µε = + ,
a b
or r   
(mn) 1 m 2 n 2
fc = √ + . (6.74)
2 µε a b
The lowest propagating mode corresponds to the smallest cutoff frequency. Assuming a > b,
the lowest propagating mode corresponds to m = 1, n = 1:
r
(11) 1 1 1
fc = √ 2
+ 2 . (6.75)
2 µε a b
(11)
Therefore, waveguide structures are high-pass devices that operate at frequencies f > fc .
The next propagating mode, in terms of cutoff frequency, is TM21 , and the mode after that
includes TM22 or TM31 depending on the ratio of b/a

6-5.2 TE Modes

The TE solution can easily be obtained by using Eqs. (5.105a) and (5.105b) and noting that
 mπ   nπ 
ψmn (x, y) = H0 cos x cos y , (6.76)
a b
which allows m or n to be zero. The expressions for the TE field quantities are given by
 mπ   mπ   nπ 
Hx (x, y, z) = −iH0 βmn sin x cos y eiβmn z , (6.77a)
a a b
 nπ   mπ   nπ 
Hy (x, y, z) = −iH0 βmn cos x sin y eiβmn z , (6.77b)
b a b
 
mπ 2  nπ 2  mπ   nπ 
Hz (x, y, z) = H0 + cos x cos y eiβmn z , (6.77c)
a b a b
 nπ   mπ   nπ 
Ex (x, y, z) = −iH0 ω µ cos x sin y eiβmn z , (6.77d)
b a b
and
 mπ   mπ   nπ 
Ey (x, y, z) = iH0 ω µ sin x cos y eiβmn z . (6.77e)
a a b
340 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Hence the lowest-order propagating wave in a rectangular waveguide is a TE10 mode (a > b)
whose cutoff frequency is given by

(10) 1
fc = √ . (6.78)
2a µε

The field expressions given by Eqs. (6.73) and (6.77) are plotted in Fig. 6-13 for four TM
and four TE modes in a rectangular waveguide with a = 2b. In this figure, the color indicates
the intensity of the total field, the solid lines indicate the direction of the electric field, and the
dashed lines show the direction of the magnetic field across the cross section of the waveguide.

6-5.3 k–β Diagram

In the preceding subsections, it was shown that for a given propagation mode the cutoff wave
number (kc ) is only a function of the waveguide’s dimensions a and b. For example, for the
mnth mode,
 mπ 2  nπ 2
2
kc(mn) = + .
a b
is a fixed quantity independent of frequency. The dispersion relation given by

k2 − β 2 = kc2 (6.79)

provides the relationship between β and k (or equivalently β and ω ). Equation (6.79)
describes a hyperbola curve in the k–β plane, as shown in Fig. 6-14. When kc > k, β becomes
pure imaginary, β = iα , and we have

k2 + α 2 = kc2 ,

which is a circle, and also shown in Fig. 6-14.

The phase velocity in any waveguide is obtained from
dz ω
up = = ,
dt β
which in this case is obviously a nonlinear function of frequency:
ω
up = p .
k2 − kc2
p
Since k2 − kc2 < k for a propagating mode, then

up > c .

That is, the phase velocity in waveguides is faster than the speed of light. Figure 6-15 shows
the β –k diagram for the first 7 modes of a rectangular waveguide with a = 2b.
6-5 The Rectangular Waveguide 341

TE10 Mode TM11 Mode

TE01 Mode TM12 Mode

TE20 Mode TM21 Mode

TE21 Mode TM22 Mode

Figure 6-13: The TE and TM field distributions across the cross section of a rectangular
waveguide. The color indicates the field intensity, the solid lines show the directions of the electric
fields, and the dashed lines show the directions of the magnetic fields.
342 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

k = ω/c

cβ c
tan γ = ω = u
p

kc

γ α circle

β
α

Figure 6-14: β –k diagram for propagation in a metallic waveguide.

6-5.4 Group Velocity

The group velocity ug is the velocity associated with the envelope of the traveling wave. For
a monochromatic signal the time-domain expression for the potential is given by

Πz (r,t) = Re[ψ (x, y) ei(β z−ω t) ] .

10

TEM
6
TE 10
k k c21 TE 20 , TE 01
4 k c11 TE 11 , TM11
TE 21 , TM21
k c20,01
2
k c10
0
1 2 3 4 5 6 7 8 9 10

Figure 6-15: β –k diagram for a rectangular waveguide with a = 2 and b = 1 for the first 7
modes of propagation. The cutoff wave numbers of the various modes are indicated along the
k-axis. Also shown is the β –k diagram in free space for a TEM wave.
6-5 The Rectangular Waveguide 343

0.8 ug
0.6 up
0.4

0.2
z (r,t)

−0.2

−0.4

−0.6

−0.8

−1
0 10 20 30 40 50 60 70 80 90 100
z

Figure 6-16: A time snapshot of a two-tone signal propagating inside a waveguide. The envelope
of the wave propagates with velocity ug , which is different from the phase velocity up .

To demonstrate the concept of group velocity in a waveguide, let us consider two mono-
chromatic signals at angular frequencies

ω1 = ω0 + ∆ ω and ω2 = ω0 − ∆ ω .

For simplicity we consider the case where the field distributions for both are the same across
the waveguide cross section. Hence the total field may be written as

Πz (r,t) = ψ (x, y) {cos [(ω0 + ∆ ω )t − (β0 + ∆ β )z] + cos [(ω0 − ∆ ω )t − (β0 − ∆ β )z]} .
(6.80)
Equivalently,
Πz (r,t) = 2 ψ (x, y) cos(ω0 t − β0 z) cos(∆ ω t − ∆ β z) . (6.81)
The first cosine term in Eq. (6.81) is the carrier, and the second one, which varies slowly with
time and space, is the envelope modulation of the wave, as shown in Fig. 6-16. As shown
earlier, the constant phase points in the z–t domain correspond to

ω0 t − β0 z = constant,
344 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

and hence
dz ω0 ω0
up = = = q . (6.82)
dt β0 k02 − kc2
Similarly, constant envelope points in the z–t domain can be obtained from

∆ ω t − ∆ β z = constant ,

from which the group velocity can be obtained and is given by

dz ∆ ω
ug = = .
dt ∆β
In the limit as ∆ ω → 0 and ∆ β → 0,
dω
ug = . (6.83)
dβ

Group velocity can also be expressed in terms of the phase velocity by noting that ω = up β ,
which leads to
d(up β ) dup
ug = = up + β . (6.84)
dβ dβ
Also noting that β = 2π /λg , Eq. (6.84) can be written as

dup
ug = up − λg . (6.85)
d λg

Since in a waveguide r
q
ω2
β= k2 − kc2 = − kc2 , (6.86)
c2
it can easily be shown that
up
dβ = dω ,
c2
and therefore
up ug = c2 , (6.87)

independent of the waveguide cross-section geometry and the mode of propagation.

6-5.5 Power Flow and Attenuation

Waveguides, like other transmission lines, can transfer power. However, as shown in this
section, waveguides can transfer power at much lower attenuation rates compared with other
transmission lines that support TEM modes like a coaxial line. The power flow inside a
waveguide can be calculated from the Poynting vector. The rate of energy flow in a rectangular
6-5 The Rectangular Waveguide 345

waveguide supporting a TMmn mode can be established by computing

Z a Z b 
1
Pmn (z) = Re (E × H ∗ ) · ẑ dx dy
2 0 0
Z a Z b 
1 ∗ ∗
= Re (Ex Hy − EyHx ) dx dy . (6.88)
2 0 0

Using the field expressions given by Eq. (6.73),

  Z aZ b
1 2 mπ 2  mπ   nπ 
Pmn (z) = |E0 | ωε βmn cos2 x sin2 y dx dy
2 a 0 0 a b
 nπ 2 Z aZ b     
2 mπ 2 nπ
+ ωε βmn sin x cos y dx dy . (6.89)
b 0 0 a b
Fortunately, the integrals can be evaluated analytically, and it can easily be shown that
Z aZ b  mπ   nπ  Z aZ b  mπ   nπ 
cos2 x sin2 y dx dy = sin2 x cos2 y dx dy
0 0 a b 0 0 a b
(
ab/4 m , 0 and n , 0 ,
= (6.90)
ab/2 m = 0 or n = 0 .

For a TM mode m , 0 and n , 0,

 
TM ab mπ 2  nπ 2
Pmn (z) = ωε + βmn |E0 |2
8 a b
ab 2
= kβmn kc,mn |E0 |2 . (TM mode) (6.91)
8η
The power carried by a TE mode can be obtained by using Eqs. (6.77a) and (6.77b) in
Eq. (6.88). Using Eq. (6.90), it can be shown that
(
1
TE ab m , 0 and n , 0 ,
Pmn (z) = 2
|H0 |2 η kβmn kc,mn × 2 (TE mode) (6.92)
4 1 m = 0 or n = 0 .

It is interesting to note that in situations where the waveguide is supporting two modes (or
more), say a TMmn and a TM pq , the total power carried by the waveguide is the superposition
of power carried by the individual modes; that is,

PTM = Pmn
TM TM
+ Ppq .
346 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

This is the direct result of orthogonality of modes in the waveguide’s cross section. For
rectangular waveguides,
Z a  mπ   pπ 
cos x cos x dx = 0 , m, p,
0 a a
and
Z b 
nπ   qπ 
sin y sin y dy = 0 , n,q,
0 b b
and because of these properties, the cross product of the fields of different modes does not
render a net flow of power along the waveguide. In practice, waveguides are made with metals
as opposed to perfect electric conductors. The finite conductivity of metallic walls causes
attenuation or loss of power as the wave propagates in the waveguide. Strictly speaking, in this
case the modes of propagation cannot be decomposed into TE or TM modes as Dirichlet’s or
Neumann’s boundary conditions are no longer applicable. However, an approximate solution
based on a perturbation method can be obtained if the conductivity of the metal is quite high.
With this approach, the waveguide walls are initially assumed to be PEC, and the surface
currents on the walls are obtained based on that assumption. Then, for good conductors, the
surface impedances given by Eq. (5.136) can be used to calculate the power loss:
I Z 
1 (1) ∗
PL = Re E · Js dℓ dz , (6.93)
2 c

where the contour integral is over the contour of the waveguide’s cross section and E(1) is the
first-order correction to the tangential electric field on a PEC. This electric field is related to
the tangential magnetic field through Eq. (6.13) with Zs = ηc . Noting that Js = n̂ × H,

n̂ × n̂ × E(1) = −ηc Js .

Hence
n̂ × E(1) = ηc n̂ × Js . (6.94)
In other words,
(1)
Et = ηc Js .
Substituting Eq. (6.94) into Eq. (6.93) leads to
I Z 
1 2
PL = Re ηc |Js | dℓ dz .
2 c

The power loss per unit length is given by

I
dPL 1
pL = = Re[ηc ]|Js |2 dℓ , (6.95)
dz 2 c
6-5 The Rectangular Waveguide 347

where
1
Re[ηc ] =
σ δs
is the surface resistivity of the metal. To find the attenuation rate in the waveguide, we note
that the power as a function of z can be written as

P(z) = P0 e−2α z , (6.96)

where α is the attenuation rate. The rate of power decrease in the waveguide, power loss per
unit length, is given by

dP(z)
− = 2α P0 e−2α z = 2α P(z) = pL . (6.97)
dz
So the attenuation rate for any mode can be obtained from
I I
2
|Js | dℓ |Ht |2 dℓ
1 pL
α= = = . (6.98)
2 P(z) 4σ δs P(z) 4σ δs P(z)

Example 6-2: Attenuation Rate for Dominant Mode in a Rectangular Waveguide

Find the attenuation rate in a rectangular waveguide supporting a TE10 mode.

Solution: The field expressions for a TE10 mode can be obtained from Eq. (6.77):

−iπ H0 β10 π 
Hx (r) = sin x eiβ10 z , (6.99a)
a a
 π 2 π 
Hz (r) = H0 cos x eiβ10 z , (6.99b)
a a
and
π π 
Ey (r) = iH0 ω µ sin x eiβ10 z . (6.99c)
a a
The unperturbed surface current along the walls can be found from Js = n̂ × H (r ∈ C). On
vertical walls at x = 0 and x = a,
 π 2
Js = x̂ × H = −ŷH0 eiβ10 z x=0 (6.100a)
a
and
 π 2
Js = −x̂ × H = −ŷH0 eiβ10 z x=a, (6.100b)
a
348 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

and on the horizontal walls

Js = ŷ × H
 π   π 2  π 
iπβ10
= H0 ẑ sin x + x̂ cos x eiβ10 z y=0 (6.101a)
a a a a
and
Js = −ŷ × H
 π   π 2  π 
iπβ10
= H0 −ẑ sin x − x̂ cos x eiβ10 z y=b. (6.101b)
a a a a
Using Eq. (6.95),
"   #
|H0 |2  π 4 πβ10 2  π 4
pL = 2 b+ a+ a
2σ δs a a a
"   #
|H0 |2  π 4 aβ10 2 a a
= b+ + . (6.102)
σ δs a π 2 2

From Eq. (6.92) we have

ab  π 2
TE
P10 (z) = |H0 |2 η kβ10 . (6.103)
4 a
Hence, the attenuation rate, using Eq. (6.98), is found to be
( "  #)
 π 2 a aβ10 2
2 b+ +1
a 2 π
α= .
σ δs abη kβ10
But  π 2
2
β10 = k2 − ,
a
so  π 2
2b + ak2
α= a nepers/m. (6.104)
σ δs abη kβ10
Note that near cutoff, where β10 → 0, the attenuation rate α → ∞. Figure 6-17 shows the
attenuation rate for a TE10 mode inside a waveguide with a = 2b = 2.28 cm, with walls made
from silver (σs = 6.14 × 107 S/m), copper (σc = 5.008 × 107 S/m), gold (σg = 4.1 × 107 S/m),
and aluminum (σAl = 3.54 × 107 S/m). It is shown that the attenuation rate increases near the
cutoff frequency, and as expected, metals with better conductivity provide lower attenuation
rates. In practice, the surface roughness of the walls increases the path length of the surface
current and causes the attenuation rates to be higher than what is shown in Fig. 6-17.
6-5 The Rectangular Waveguide 349

0.018
Silver
0.016 Copper
Gold
0.014
Aluminum
0.012
α (Np/m)

0.010

0.008

0.006

0.004
6 8 10 12 14 16 18
Frequency (GHz)

Figure 6-17: Attenuation rate inside a rectangular waveguide with a = 2b = 2.28 cm, computed
using Eq. (6.104). Four different conductivities were used in the calculation corresponding to
waveguide walls made from silver, copper, gold, and aluminum.

6-5.6 Rectangular Cavity

The term cavity resonator refers to a region enclosed by a metallic surface within which an
electromagnetic field can exist. Such fields can be predicted to exist, without any excitations,
at certain frequencies that depend on the shape and size of the cavity. Such frequencies are
known as the resonant frequencies at which the electric and magnetic energies are stored.
A rectangular cavity is a cavity made up of a rectangular waveguide terminated at both ends
with metallic surfaces, as shown in Fig. 6-18, where the short circuits are placed at z = 0 and
z = c. Similar to the fields in a rectangular waveguide, the fields can be decomposed into TM
and TE (with respect to the z-axis) modes. In this case, however, the waveguide segment of
length c can support both +z-going and −z-going waves. For TM waves, Ex (x, y, z) can be
written as (see Eq. (6.73)):
 mπ   mπ   nπ  ∂
Ex (x, y, z) = E0 cos x sin y [Aeiβmn z + Be−iβmn z ] . (6.105)
a a b ∂z
The boundary conditions at z = 0 and z = c mandate that

Ex (x, y, 0) = Ex (x, y, c) = 0 .
350 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

b z
x .

a c

Figure 6-18: Geometry of a metallic rectangular cavity.

Hence, A = B, and
ℓπ
sin(βmn c) = 0 βmn = ℓ = 1, 2, 3, . . .
c
According to Eq. (6.71) a nontrivial solution can be obtained when
 mπ 2  nπ 2  2
2 ℓπ
k = + + . (6.106)
a b c
That is, for frequencies at which Eq. (6.106) is satisfied, a resonant mode in the cavity can be
established. It is noted that the resonant wave number given by Eq. (6.106) is only a function
of the cavity’s geometrical parameters (a, b, c). If the material filling the cavity is lossless,
the resonant frequencies are all real; otherwise the resonant frequencies are complex so that
k2 = 4π 2 f 2 µε is real. The TM fields in the cavity are given by
   
ℓπ mπ   mπ   nπ  ℓπ
Ex (x, y, z) = −E0 cos x sin y sin z , (6.107a)
c a a b c
    mπ   nπ   
ℓπ nπ ℓπ
Ey (x, y, z) = −E0 sin x cos y sin z , (6.107b)
c b a b c
   
mπ 2  nπ 2  mπ   nπ  ℓπ
Ez (x, y, z) = E0 + sin x sin y cos z , (6.107c)
a b a b c
 nπ   mπ   nπ   
ℓπ
Hx (x, y, z) = −iE0 ωε sin x cos y cos z , (6.107d)
b a b c
and
 mπ   mπ   nπ   
ℓπ
Hy (x, y, z) = −iE0 ωε cos x sin y cos z . (6.107e)
a a b c
6-5 The Rectangular Waveguide 351

The field expressions for TE can be obtained from Eq. (6.77) and are given by
 nπ   mπ   nπ   
ℓπ
Ex (x, y, z) = iH0 ω µ cos x sin y sin z , (6.108a)
b a b c
 mπ   mπ   nπ   
ℓπ
Ey (x, y, z) = iH0 ω µ sin x cos y sin z , (6.108b)
a a b c
   
ℓπ mπ   mπ   nπ  ℓπ
Hx (x, y, z) = −H0 sin x cos y cos z , (6.108c)
c a a b c
    mπ   nπ   
ℓπ nπ ℓπ
Hy (x, y, z) = −H0 cos x sin y cos z , (6.108d)
c b a b c
and
   
mπ 2  nπ 2  mπ   nπ  ℓπ
Hz (x, y, z) = H0 + cos x cos y sin z . (6.108e)
a b a b c

According to the Poynting theorem for phasor quantities, the complex power in any region
is given by
c
Pinc = Prad + Ploss + i2ω (we − wm ) . (6.109)
In a cavity, there is no radiated power, and assuming no external sources and a loss-free cavity,
the conservation of complex power mandates that

i2ω (we − wm ) = 0 ,

or equivalently, at resonance in a general cavity we have

$
ε
we = wm = |E|2 dv
4 V
$
µ
= |H|2 dv . (6.110)
4 V

The time-domain expression for the Poynting theorem is

∂
Pin (t) = Prad (t) + Ploss (t) + (we + wm ) . (6.111)
∂t
The law of conservation of power in a lossless cavity provides the following condition:

∂
(we + wm ) = 0 . (6.112)
∂t
Equation (6.112) indicates that the total stored energy in a cavity is a constant function of
time. For a time-harmonic case, if we choose a time for which H(t) is zero, wm will be zero
as well. In this case the stored electric energy is maximum and twice its average value. In this
352 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

case the total stored energy can be computed from

$
ε
w = 2we = |E|2 dv . (6.113)
2
In practice, cavity walls are not perfect conductors and they can cause power loss. To
compute power loss in a cavity, we can use a perturbation solution similar to what was
used in Section 6-5.5 for the calculation of attenuation rates in waveguides. This power loss
calculation uses an equation similar to Eq. (6.93) and is given by
 
s 1 (1) ∗
PL = Re E · Js ds , (6.114)
2 s

(1)
where s represents the surface of the cavity. As before, Et is the first-order correction to the
tangential electric field over the cavity walls and is given by
(1)
Et = ηc Js = ηc n̂ × H(r) , r∈s.

Hence
 
1
PLs = Re 2
ηc |Ht (r)| ds
2 s

1
= |Ht (r)|2 ds (6.115)
2σ δs s

Equation (6.115) is valid for any cavity of arbitrary geometry. Of course, this expression
is proportional to the strength of the fields established inside the cavity. To compare the
performance of different cavities or different modes within a specific cavity, a different metric
must be used such that it is independent of field strength. One such metric is the quality factor
as defined in Eq. (3.149), which is a unitless quantity given by
maximum of stored energy
Q̆ = ω . (6.116)
average dissipated power
Using Eqs. (6.113) and (6.115) in (6.116), the cavity quality factor due to surface loss is given
by $
ωεσ δs |E|2 dv
V
Q̆s = . (6.117)
|Ht |2 ds
s
This expression for the quality factor assumes that the dielectric filling the cavity is lossless.
To consider the effect of dielectric loss as well, we note that the average power loss due to the
dielectric loss tangent can be computed from
$
1
PLd = ωε ′′ |E|2 dv . (6.118)
2 V
6-5 The Rectangular Waveguide 353

Using Eq. (6.118) in Eq. (6.116), the quality factor due to dielectric loss can easily be
computed and is given by
ε′
Q̆d = ′′ . (6.119)
ε
The total quality factor using Eq. (6.116) can be written as
ωW
Q̆ = ,
PLs + PLd

and thus it can easily be shown that

1 1 1
= + . (6.120)
Q̆ Q̆s Q̆d
According to Eq. (6.94), the eigenvalues of the wave equation subject to Dirichlet’s or
Neumann’s boundary conditions are real and positive quantities. These eigenvalues are a
function of the cavity dimensions only. Thus it is interesting to note that for Eq. (6.94) to
be true when the cavity is filled with lossy materials, the resonant frequency (ωmnℓ ) must be
complex so that
(ωmnℓ )2 µε = Cmnℓ , (6.121)
where Cmnℓ is a real and positive number. This establishes a relation between the real
and imaginary parts of the resonant frequency to those of the material index of refraction.
Representing the complex resonant frequency ωmnℓ = ωmnℓ ′ − iω ′′ with a negative imaginary
mnℓ √
part (ωmnℓ > 0) and the refraction index of the material by n = µε = n′ + in′′ , from
′′

Eq. (6.121) we have

′ ′′
(ωmnℓ − iωmnℓ )2 (n′ + in′′ )2 = Cmnℓ . (6.122)
Expanding Eq. (6.122) and setting the imaginary part of the left-hand side equal to zero, we
get the following relations:
′2 − ω ′′2
ωmnℓ mnℓ n′ 2 − n′′ 2
′ ω ′′ = . (6.123)
ωmnℓ mnℓ n′ n′′
In practice, resonators are filled with low-loss materials; that is, n′′ ≪ n′ , which results in a
′′ ≪ ω ′ . Under this assumption, Eq. (6.123) simplifies to
condition where ωmnℓ mnℓ
′
ωmnℓ n′
′′ ≈ . (6.124)
ωmnℓ n′′

For a nonmagnetic material, µ = µ0 , and with ε ′′ ≪ ε ′ , it can be shown that

√
n′ ε′ 2ε ′
≈ √ = , (6.125)
n′′ ε ′′ /(2 ε ′ ) ε ′′
354 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

where we use the first-order Taylor expansion of

√ √ ε ′′
ε ′ + iε ′′ ≈ ε ′ + i √ .
2 ε′

Recognizing ε ′ /ε ′′ = Q̆d as the dielectric quality factor, then

′
ωmnℓ
Q̆mnℓ
d = ′′ . (6.126)
2ωmnℓ

This result is true for an arbitrarily shaped cavity filled with slightly lossy material.
Due to the presence of losses in the cavity, the stored energy in the cavity decays as a
function of time once excitations are removed. Representing the field’s relaxation time by τ
or damping factor by γ = τ1 , the stored energy as a function of time may be represented as

W (t) = W0 e−2t/τ = W0 e−2γ t .

Note that the rate of decrease of stored energy must equal the power loss. That is,
dW
− = 2γW0 e−2γ t = 2γW = PL .
dt
In other words,
PL ω PL ω
γ= = = (6.127)
2W 2 ωW 2Q̆
and
2Q̆
τ= . (6.128)
ω

Example 6-3: Quality Factor of a Rectangular Cavity

Consider the rectangular cavity shown in Fig. 6-18 with a = 2.28 cm, b = 1.14 cm, and
c = 3 cm. Suppose this cavity is supporting the TE101 mode. Find the resonant frequency
and the quality factor for this cavity, assuming that the walls are made from copper with
conductivity σ = 5.8 × 107 S/m. Also determine the damping factor for this mode.
Solution: The field expression can be found from Eq. (6.108) for m = 1, n = 0, and ℓ = 1:
π π  π 
Ey (x, y, z) = iH0 ω µ sin x sin z , (6.129a)
a a c
π2 π  π 
Hx (x, y, z) = −H0 sin x cos z , (6.129b)
ac a c
and
6-5 The Rectangular Waveguide 355

 π 2 π  π 
Hz (x, y, z) = H0 cos x sin z . (6.129c)
a a c
Note that the electric field and magnetic field are 90◦ out of phase at every point inside the
cavity. This indicates that when the electric field is at its maximum, the magnetic field is zero
and vice versa. The time-average stored energy is computed from
$
ε
W = 2We = |E|2 dv
2 V
Z aZ bZ c
ε
= E 2 dx dy dz
2 0 0 0 y
ε   π 2 abc k2 µπ 2 bc
= |H0 |2 ω 2 µ 2 = |H0 |2 . (6.130)
2 a 4 8a
To calculate the power loss on the six walls of the cavity, Eq. (6.115) is used:
( Z Z 
|H0 |2 a c
π4      π 4    
2 π 2 π 2 π 2 π
PL = 2 sin x cos z + cos x sin x dx dz
2σ δs 0 0 (ac)2 a c a a a
Z bZ c  4 Z aZ b 4 )
π π  π π 
+2 sin2 z dy dz + 2 2
sin2 x dx dy ,
0 0 a a 0 0 (ac) a

which can easily be integrated to give

π 4 a3 c + c3 a + 2c3 b + 2a3 b
PL = |H0 |2 . (6.131)
σ δs 4a4 c2
Hence, using Eq. (6.117), the quality factor is given by

(σ δs )k03 η0 a3 bc3
Q̆s = . (6.132)
2π 2 (a3 c + c3 a + 2c3 b + 2a3 b)
The resonant frequency is given by
 π 2  π 2
k02 = + ,
a c
or r
1 1 1
f0TE101 = √ + = 8.263 GHz.
2 µ0 ε0 a2 c2
The quality factor is calculated from Eq. (6.132), which requires calculation of the skin depth:
1
δs = √ = 0.727 × 10−6 m.
π f µσ
356 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

This provides surface conductance σ δs = 42.16, and finally

Q̆s = 8200.

The relaxation time, which is inversely related to the damping factor, is obtained from
2Q
τ= = 316.04 nsec.
ω
This indicates that after 316.04 nsec an established field in the cavity reaches e−1 of its original
value. This corresponds to 2611 cycles of oscillation. The damping factor is given by
1
γ= = 3.16 × 106 Np/s.
τ

Example 6-4: Time-Average Stored Electric and Magnetic Energy in a

Rectangular Cavity

Find the expressions for the time-average stored electric and magnetic energy for TM modes
in a rectangular cavity with dimensions a, b, c, and show that they are equal.
Solution: The stored electric energy in the cavity can be calculated from
Z aZ bZ c
1
we = ε |E(x, y, z)|2 dx dy dz
4 0 0 0
Z aZ bZ c
1

= ε |Ex (x, y, z)|2 + |Ey (x, y, z)|2 + |Ez (x, y, z)|2 dx dy dz .
4 0 0 0

Using the field expression given by Eq. (6.109), and noting that the integrals of square of sine
or cosine functions over one period are equal to one-half of the period, it can be easily shown
that
(   2     )
ε abc ℓπ 2  m π  2 ℓπ nπ 2 mπ 2  nπ 2 2
we = · + + + |E0 |2 .
4 8 c a c b a b

Making use of Eq. (6.106), this expression can be simplified to

 
ε abc  mπ 2  nπ 2 2
we = + k |E0 |2
32 a b
εV 2 2
= k k |E0 |2 , (6.133)
32 mn
where V = abc is the cavity volume,
 
2 mπ 2  nπ 2
kmn = + ,
a b
6-6 Transmission-Line Circuit Model for Waveguides 357

Δz Z Δz
I(z) I(z + Δz) I(z) I(z + Δz)

V(z) V(z + Δz) V(z) Y Δz V(z + Δz)

(a) Transmission-line segment (b) Equivalent circuit model

Figure 6-19: A differential segment of (a) a transmission line supporting a TEM wave and (b) its
equivalent circuit.

and "  2 #
 mπ 2  nπ 2 ℓπ
k2 = ω 2 µε = + + .
a b c
Similarly, the stored magnetic energy can be calculated from
Z aZ bZ c
1
wm = µ |H(x, y, z)|2 dx dy dz
4 0 0 0
Z aZ bZ c
1

= µ |Hx (x, y, z)|2 + |Hy(x, y, z)|2 dx dy dz.
4 0 0 0

Inserting the magnetic field expressions from Eqs. (6.107d) and (6.107e),
 
µ abc 2 2  mπ 2  nπ 2
wm = ω ε + |E0 |2
32 a b
εV 2 2
= k k |E0 |2 . (6.134)
32 mn
which has the same expression we computed for we .

6-6 Transmission-Line Circuit Model for Waveguides

6-6.1 Transmission-Line TEM Mode

In general, circuit models are an easy way to understand the behavior of an electrical system.
Strictly speaking, circuit models are applicable for lumped elements and guiding structures
that support transverse electromagnetic waves. For example, for a transmission line, such as
a coaxial line, supporting a TEM wave, a circuit model for a short section of the line having a
differential length ∆ z is represented by a series impedance per unit length denoted by Z and a
shunt admittance for unit length represented by Y , as shown in Fig. 6-19.
358 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

The governing equations for the voltage and current on the line can easily be derived from
Kirchhoff’s voltage and current laws. They are given by

dV (z)
= −Z I(z) (6.135a)
dz
and
dI(z)
= −Y V (z) . (6.135b)
dz
The voltage and current satisfy one-dimensional wave equations given by

d 2V (z)
−Y Z V (z) = 0 (6.136a)
dz
and
d 2 I(z)
−Y Z I(z) = 0 . (6.136b)
dz
Here, the complex propagation constant is defined by
√
γ = YZ , (6.137)

and the characteristic impedance of the line is defined by

r
Z
Z0 = . (6.138)
Y
The transmission line’s series impedance per unit length Z is usually represented by an
inductance per unit length L in series with a resistance per unit length R, and therefore

Z = R − iω L , (6.139a)

and the shunt admittance per unit length is represented by a shunt capacitance per unit length
C and conductance per unit length G:

Y = G − iωC . (6.139b)

6-6.2 Waveguide TM Modes

Recall that we are using the time convention of e−iω t . As was shown in the previous section,
hollow conductors of infinite length having a constant cross section, namely waveguides,
cannot support TEM waves, and the mode of propagation is either TM and/or TE waves.
Considering a TM mode of propagation, according to Eqs. (5.102a) and (5.102b) the
transverse electric and magnetic fields are given by

∂ h i
iβ z
Et = ∇t ψ (x, y) e (6.140a)
∂z
and
6-6 Transmission-Line Circuit Model for Waveguides 359

h i
Ht = −iωε ∇t ψ (x, y) eiβ z × ẑ . (6.140b)

It is obvious from Eq. (6.140a) that ∇t × Et = 0, and therefore in the transverse plane a voltage
V can be defined so that
Et = −∇V . (6.141)
Comparing Eq. (6.141) with Eq. (6.140a) leads to

∂
V =− [ψ (x, y) eiβ z ] . (6.142)
∂z
Also, the longitudinal component of the electric field, according to Eq. (5.102a), is given by

Ez = kc2 ψ (x, y) eiβ z ,

which, together with Eq. (6.142), can be used to show that

∂ Ez
= −kc2V . (6.143)
∂z
Also, noting that −iωε Ez represents a longitudinal displacement current density and the
fact that 1/kc2 has a unit of m2 , the notion of a longitudinal current along the length of the
waveguide can be defined as
−iωε
I= Ez . (6.144)
kc2
Using Eq. (6.144) in Eq. (6.143), it is now obvious that

∂I
= iωε V . (6.145)
∂z
From Eq. (6.142), the rate of change of voltage along the length of the waveguide can be
obtained:
∂V β2
= β 2 ψ (x, y) eiβ z = 2 Ez . (6.146)
∂z kc
Using Eq. (6.144) in Eq. (6.146),
 
∂V β2 k2 − kc2 k2
= I= I = iω µ + c I . (6.147)
∂z −iωε −iωε iωε
Equations (6.147) and (6.145) are analogous to Eqs. (6.135a) and (6.135b) for the standard
model for a transmission line with
β2 k2
ZTM = = −iω µ − c (6.148a)
iωε iωε
and
YTM = −iωε . (6.148b)
360 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

μ ε/kc2 μ ε/kc2

ε ε

Figure 6-20: The transmission-line equivalent circuit model of a waveguide supporting a TM

mode.

The transmission-line equivalent circuit model for a TM mode in a waveguide is shown in

Fig. 6-20. The complex propagation constant of this waveguide transmission-line equivalent
is given by p
√
γTM = ZTMYTM = −β 2 = iβ , (6.149)
as expected, and the characteristic impedance of the TM wave is given by
r
TM ZTM β
Z0 = = , (6.150)
YTM ωε
which is in agreement with Eq. (5.103). It is clear that this transmission line is a high-pass
structure.

6-6.3 Waveguide TE Modes

The expressions given in the preceding subsection apply to TM waves. For TE waves, we can
apply the duality relations. Basically, by changing ZTM to YTE and YTM to ZTE, as well as ε to
µ and µ to ε , the following expressions are obtained for the elements of the equivalent circuit
for TE waves:

ZTE = −iω µ (6.151a)

and
kc2
YTE = −iωε − . (6.151b)
iω µ
The corresponding expressions for the complex propagation constant γTE and characteristic
impedance Z0TE are given by

γTE = iβ (6.152a)
and
β
Z0TE = . (6.152b)
ωµ
6-6 Transmission-Line Circuit Model for Waveguides 361

μ μ

μ/kc2 ε μ/kc2 ε

Figure 6-21: The transmission-line equivalent circuit model of a waveguide supporting a TE

mode.

d
d

C L

Figure 6-22: Metallic septa placed in a rectangular waveguide in a symmetrical manner with
a gap between them. The horizontal gap acts like a lumped shunt capacitor, and the vertical gap
behaves like a lumped shunt inductor.

Figure 6-21 shows the equivalent circuit model for a waveguide supporting a TE mode.
As will be shown in the following sections, there are discrete modes of propagation
for both the TE and TM modes. That is, for a fixed frequency above the dominant cutoff
frequency, there are specific discrete values for the propagation constant and the cutoff
frequency. Hence the transmission line models presented here are mode-dependent.
The concept of equivalent circuit for waveguides finds application in situations where
certain waveguide local discontinuities are present in the cross section of the waveguide.
Figure 6-22 shows two such discontinuities where metallic septa with edges perpendicular
362 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

and parallel to the electric field of a single-mode TE10 waveguide are placed in the waveguide.
The metallic septa perpendicular to the electric field shown in Fig. 6-22(a) represent a shunt
capacitive lumped element, whereas the septa in Fig. 6-22(b) behave as a shunt inductive
element. Figures 6-22(c) and (d) show the transmission-line equivalent circuit for each case.
The capacitance and inductance values of the shown septa must be obtained numerically using
an integral equation or a mode-matching method. For example, for a symmetrically positioned
capacitive septum with gap d, the susceptance is given by†
(   

4b πd Q2 cos4 π2bd
B = ωC = Y0 ln csc +

λg 2b 1 + Q2 sin4 π2bd
 2     )
1 b π d πd
+ 1 − 3 sin2 cos4 . (6.153)
16 λg 2b 2b

Also, for a symmetric inductive septum with gap d, the reactance is given by‡
 " !
a π d 3 1 πd
X = ω L = Z0 tan2 1+ p − 1 sin2
λg 2a 4 1 − (2a/3λ )2 a
 a 2  #
4 E(α ) − β 2 F(α ) E(β ) − α 2 F(β ) 1 2 πd
+2 1− × − sin .
λ π α2 β2 2 a
(6.154)

In Eq. (6.153),
1
Q2 = p −1 ,
1 − (b/λg )2
and in Eq. (6.154),    
πd πd
α = sin , β = cos .
2a 2a
It is noted that the expressions given by Eqs. (6.153) and (6.154) are approximate; for more
accurate results a full-wave numerical method must be used. Also, in Eq. (6.153) F(α ), E(α )
are complete elliptic integrals of the first and second kind, respectively; that is,
Z π /2
dφ
F(α ) = q
0 1 − α 2 sin2 φ
and
Z π /2 q
E(α ) = 1 − α 2 sin2 φ d φ .
0

† N. Marcuvitz, Waveguide Handbook, Peter Peregrinus Ltd., London, U.K., 1986.

‡ Ibid.
6-7 Other Modal Solutions for Rectangular Waveguides 363

6-7 Other Modal Solutions for Rectangular Waveguides

The transverse electric and magnetic field representations with respect to the direction of
propagation lend themselves to all waveguides and any arbitrary cross section. In certain
waveguide problems, it is not possible to express the modes in terms of only TE or TM modes.
For these problems, hybrid TE and TM modes have to be considered to arrive at appropriate
solutions. Alternatively, for rectangular waveguides it is possible to generate modes from
a single component of the Hertz vector potential that is perpendicular to the direction of
propagation. As before, we assume the direction of propagation is along the z-axis and the
cross section of the waveguide is invariant with respect to z. In this section we seek field
solutions generated from
Π (x, y, z) = ψ (x, y) eiβ z x̂ . (6.155)
A similar procedure that led to Eq. (5.101) can be used to show that ψ (x, y) must satisfy the
scalar 2-D wave equation
∇2t ψ (x, y) + kc2 ψ (x, y) = 0 , (6.156)
where kc2 = k2 − β 2 .

TM Fields
Using Eqs. (5.94a) and (5.94b), it can be shown that
 2 
∂ ψ
Ex = + k ψ eiβ z ,
2
(6.157a)
∂ x2
∂ 2 ψ iβ z
Ey = e , (6.157b)
∂x ∂y
∂ ψ iβ z
Ez = iβ e , (6.157c)
∂x
and

Hx = 0 , (6.158a)
Hy = ωεβ ψ eiβ z , (6.158b)
∂ ψ iβ z
Hz = iωε e . (6.158c)
∂y
As can be seen from Eq. (6.158), this choice for Hertz vector potential generates transverse
magnetic with respect to x fields (TMx ). For TMx modes, the boundary condition mandates
Ez to vanish on the waveguide walls. This is accomplished if

∂ ψ TMx (x, y)
=0 at x = 0, a and y = 0, b .
∂x
Hence  mπ   nπ 
ψ TMx (x, y) = cos x sin y . (6.159)
a b
364 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

The integer indices are m = 0, 1, 2, . . . and n = 1, 2, 3, . . . Using Eqs. (6.157) and (6.158), the
field expressions for TMx are given by
  mπ  2   mπ   nπ 
2
Ex = k − cos x sin y eiβ z , (6.160a)
a a b
 mπ   nπ   mπ   nπ 
Ey = − sin x cos y eiβ z , (6.160b)
a b a b
 mπ   mπ   nπ 
Ez = −iβ sin x sin y eiβ z , (6.160c)
a a b
and

Hx = 0 , (6.161a)
 mπ   nπ 
Hy = ωεβ cos x sin y eiβ z , (6.161b)
a b
 nπ   mπ   nπ 
Hz = iωε cos x cos y eiβ z . (6.161c)
b a b

TE Fields

To generate TEx fields, a magnetic Hertz vector potential that has only one component along
x is chosen. Noting the duality relations of the fields, the TEx fields can be evaluated from

Ex = 0 , (6.162a)
Ey = −ω µβ ψm eiβ z , (6.162b)
∂ ψm iβ z
Ez = −iω µ e , (6.162c)
∂y
and
 
∂ 2 ψm
Hx = + k ψm eiβ z ,
2
(6.163a)
∂ x2
∂ 2 ψm iβ z
Hy = e , (6.163b)
∂x ∂y
∂ ψm iβ z
Hz = iβ e . (6.163c)
∂x
For TEx modes the Ez = 0 boundary condition mandates that

∂ ψ TEx
=0 at x = 0, a and y = 0, b,
∂y
6-7 Other Modal Solutions for Rectangular Waveguides 365

and therefore  mπ   nπ 
ψ TEx (x, y) = sin x cos y . (6.164)
a b
The indices are m = 1, 2, 3, . . . and n = 0, 1, 2, . . . Also, the field expressions for TEx are given
by

Ex = 0 , (6.165a)
 mπ   nπ 
Ey = −ω µβ sin x cos y eiβ z , (6.165b)
a b
 nπ   mπ   nπ 
Ez = iω µ sin x sin y eiβ z , (6.165c)
b a b
and
  mπ  2   mπ   nπ 
Hx = k2 − sin x cos y eiβ z , (6.166a)
a a b
 mπ   nπ   mπ   nπ 
Hy = − cos x sin y eiβ z , (6.166b)
a b a b
 mπ   mπ   nπ 
Hz = iβ cos x cos y eiβ z . (6.166c)
a a b
Note that both TMx and TEx have longitudinal field components along the direction of
propagation. As a result, these field expressions may be appropriate for problems for which
TEz and TMz field expansions do not yield satisfactory solutions.

Example 6-5: Propagation Constant for a Partially Filled Waveguide

Consider a rectangular waveguide filled with two different dielectric materials with a planar
interface along the direction of propagation, as shown in Fig. 6-23. Find the propagation
constant for different possible modes.

Solution: Field solutions based on TMz or TEz modes do not yield a complete solution. For
this problem we examine the TMx and TEx field solutions. We also note that the phase-
matching condition at the interface between media 1 and 2 (at x = c) requires the same
propagation constant β in both regions. For TM to x, the solutions for the 2-D scalar potential
in regions 1 and 2 take the following form:
 nπ 
ψ1 (x, y) = A cos(kx1 x) sin y (6.167a)
b
and
 nπ 
ψ2 (x, y) = B cos[kx2 (x − a)] sin y , (6.167b)
b
366 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

b
,ε ,ε
x
a c

Figure 6-23: A partially filled rectangular waveguide.

which satisfy the boundary conditions


∂ ψ1 
=0
∂ x x=0
and

∂ ψ2 
=0.
∂ x x=0

The goal is to find kx1 and kx2 by enforcing the continuity of the tangential electric and
magnetic fields at the interface between the two dielectric media (x = c). From the dispersion
relations we have the following equations:
 nπ 2
2
kx1 + + β 2 = k12 = ω 2 µ1 ε1 (6.168a)
b
and
 nπ 2
2
kx2 + + β 2 = k22 = ω 2 µ2 ε2 . (6.168b)
b
This provides the following equation for kx1 and kx2 :
2 2
kx2 − kx1 = ω 2 (µ2 ε2 − µ1 ε1 ) . (6.169)

Using Eq. (6.157), the fields Ez (c, y) calculated in each of the two regions are given by
 nπ 
Ez1 (c, y) = −iβ kx1 A sin(kx1 c) sin y eiβ z (6.170a)
b
6-8 Modal Expansion of Field Quantities 367

and
 nπ 
Ez2 (c, y) = −iβ kx2 B sin[kx2 (c − a)] sin y eiβ z . (6.170b)
b
Also, using Eq. (6.158), the tangential magnetic fields in each region at the interface are given
by
 nπ   nπ 
Hz1 (c, y) = iωε1 A cos(kx1 c) sin y eiβ z (6.171a)
b b
and
 nπ   nπ 
Hz2 (c, y) = iωε2 B cos[kx2 (c − a)] cos y eiβ z . (6.171b)
b b
From Eqs. (6.170a) and (6.170b), we get

kx1 A sin(kx1 c) = kx2 B sin[kx2 (c − a)] , (6.172)

and from Eqs. (6.171a) and (6.171b) we get another equation for the unknowns:

ε1 A cos(kx1 c) = ε2 B cos[kx2 (c − a)] . (6.173)

Taking the ratio of Eqs. (6.172) and (6.173), another equation for kx1 and kx2 is obtained:
kx1 kx2
tan(kx1 c) = tan[kx2 (c − a)] . (6.174)
ε1 ε2
To find kx1 and kx2 , Eqs. (6.174) and (6.169) must be solved simultaneously. An analytical
solution for these equations does not exist and they have to be solved numerically.
A similar procedure can be followed for TE to x modes. Equations (6.162) and (6.163)
can be used to find the tangential electric and magnetic fields (Ez and Hz ) at x = c. Using these
two equations, the following characteristic equation for TEx can be obtained:
kx1 kx2
cot(kx1 a) = cot[kx2 (c − a)] . (6.175)
µ1 µ2
Now Eq. (6.175) must be solved together with Eq. (6.169) to find kx1 and kx2 . Then β can be
found from Eq. (6.168).

6-8 Modal Expansion of Field Quantities

In the previous section it was shown that in a closed region, the solution to Maxwell’s
equations can be expressed in terms of discrete modal wave functions that are solutions
to the scalar wave equation, subject to the Dirichlet or Neumann boundary conditions. For
waveguide problems the 2-D wave equation is given by

∇2t ψ + kc2 ψ = 0 , (6.176)

368 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

where as before kc2 = k2 − β 2 and ψ is a scalar function of two variables in the cross section
of the waveguide. For a rectangular waveguide it was shown that the solution of Eq. (6.176)
subject to ψ = 0 (TM modes) or ∂ ψ /∂ n = 0 (TE modes) on the surface of the closed boundary
provides a discrete set of solutions. Due to the linearity of Maxwell’s equations, any linear
combination of the modes is also a solution. In this section we demonstrate that for any
arbitrary cross section, the discrete modes constitute an orthogonal and complete set. That is,
any solution of Maxwell’s equations can be expressed in terms of these modes and the total
power can be expressed in terms of the superposition of the powers carried by the individual
modes.
The scalar wave equation can be viewed as an eigenfunction problem in which ψ is the
eigenfunction of the ∇2t operator with kc2 as its corresponding eigenvalue. As shown before
for a rectangular waveguide, the eigenfunctions for TM modes are given by
 mπ   nπ 
ψmn (x, y) = sin x sin y . (6.177)
a b
It can easily be shown that
Z aZ b   ′   ′ 
mπ   nπ  mπ nπ
hψmn , ψm′ n′ i = sin x sin y sin x sin y dx dy
0 0 a b a b
(
1
ab m′ = m, n′ = n ,
= 4 (6.178)
0 m′ , m or n′ , n .

That is, the eigenfunctions of TM modes are orthogonal in the cross-section domain. Note
that the inner product of functions ψmn and ψm′ n′ is defined as the integral of the product of
these functions in the cross section of the waveguide. The same is true for the eigenfunctions
of TE modes:  mπ   nπ 
ψmn (x, y) = cos x cos y . (6.179)
a b
The set of eigenfunctions given by Eq. (6.177) or Eq. (6.179) is complete. It is interesting to
note that the corresponding eigenvalues for both TE and TM modes, given by
 mπ 2  nπ 2
2
kmn = + ,
a b
are real and positive quantities.
These properties hold for waveguides with arbitrary cross-section geometries. To show the
orthogonality property, let’s consider two possible modes ψmn and ψm′ n′ . These must satisfy
the scalar wave equation (6.176), i.e.,

(∇2t + kmn
2
)ψmn = 0 (6.180a)
and
(∇2t + km
2
′ n′ )ψm′ n′ = 0 . (6.180b)
6-8 Modal Expansion of Field Quantities 369

Multiplying Eq. (6.180a) by ψm′ n′ , Eq. (6.180b) by ψmn , and then subtracting the resulting
equations, we have

ψm′ n′ ∇2t ψmn − ψmn ∇2t ψm′ n′ = (km2 ′ n′ − kmn

2
)ψmn ψm′ n′ . (6.181)

According to Green’s second identity

" I  
2 2 ∂ ψmn ∂ ψm′ n′
(ψm′ n′ ∇t ψmn − ψmn ∇t ψm′ n′ ) ds = ψm′ n′ − ψmn dℓ . (6.182)
S C ∂n ∂n

Since either ψ = 0 or ∂∂ψn = 0 on the surface of the waveguide, the right-hand side of
Eq. (6.182) vanishes, and therefore
"
2 2
(km′ n′ − kmn ) ψmn ψm′ n′ ds = 0 . (6.183)
S

In cases when m′ , m and n′ , n, then km′ n′ , kmn , which mandates the orthogonality condition:
"
ψmn ψm′ n′ ds = 0 . (6.184)
S

To prove that ψmn functions form a complete set (for all m, n), let us consider an arbitrary
continuous function f (x, y) in the domain S that satisfies the boundary condition. For the set
of eigenfunctions to be complete, there must exist coefficients amn such that
XX
f (x, y) = amn ψmn (x, y) . (6.185)
m n

Multiplying both sides of Eq. (6.185) by ψmn (x, y), integrating over the cross section, and
using the orthogonality properties of ψmn , the amn coefficients can be obtained from
"
f (x, y) ψmn (x, y) ds
S"
amn = . (6.186)
2
ψmn (x, y) ds
S

It can also be shown that the eigenvalues (kc2 ) are real and positive. Let’s consider the mnth
mode, for which we can write

(∇2t + kmn
2
)ψmn = 0 (6.187a)
and
∗ ∗
(∇2t + kmn
2
)ψmn =0, (6.187b)

where the superscript ∗ denotes the complex conjugate of the quantity. Multiplying
∗ and Eq. (6.187b) by ψ
Eq. (6.187a) by ψmn mn and subtracting the resulting equations, we
370 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

get
∗ ∗ ∗ ∗
ψmn ∇2t ψmn − ψmn ∇2t ψmn 2
= (kmn 2
− kmn )ψmn ψmn . (6.188)
As in the previous case, integrating both sides of Eq. (6.188) over the cross-sectional area of
the waveguide and then invoking Green’s second identity, we can show that
"
2 ∗ 2
(kmn − kmn ) |ψmn |2 ds = 0 .
S

Since the integrand is a positive quantity we conclude that

2 2 ∗
kmn = kmn , (6.189)
2 is a real quantity. To show that k2 is a positive quantity, Green’s first identity
proving that kmn mn
can be used. According to Green’s first identity,
" I
2 ∂ ψmn
(∇t ψmn · ∇t ψmn + ψmn ∇t ψmn ) ds = ψmn dℓ . (6.190)
S C ∂n
Dirichlet’s or Neumann’s boundary conditions force the right-hand side of Eq. (6.190) to go
to zero. Then, using Eq. (6.187a) in the left-hand side of Eq. (6.190), we get
"
(∇t ψmn · ∇t ψmn ) ds
2
kmn = " . (6.191)
2
ψmn ds

Assuming the eigenfunctions ψmn are real quantities, it then is obvious that both the numerator
and denominator of Eq. (6.191) are positive quantities, which indicates that kmn 2 is a positive

and real quantity. It should be pointed out that the assumption about eigenfunctions being
real functions is not necessary to prove that the eigenvalues are real and positive. In fact if
ψmn = ψmn ′ + iψ ′′ is complex, then its real and imaginary parts satisfy the wave equation
mn
2 is real. Following the above procedure that led to Eq. (6.191) for ψ ′ , then
(6.187a) since kmn mn
2 is positive. Alternatively, one can apply Green’s first identity to ψ
it can be shown that kmn mn
and ψmn∗ to show that
"
|∇t ψmn |2 ds
2
kmn = "S . (6.192)
|ψmn |2 ds

Another interesting property of the eigenfunctions is that not only are they themselves
orthogonal, but also their gradients (∇t ψmn ) are orthogonal. To show this, we can apply
Green’s first identity to two distinct eigenfunctions ψmn and ψm′ n′ :
" I
2 ∂ ψm′ n′
(∇t ψmn · ∇t ψm′ n′ + ψmn ∇t ψm′ n′ ) ds = ψmn dℓ .
S ∂n
6-8 Modal Expansion of Field Quantities 371

2 ψ ′ ′ for ∇2 ψ ′ ′ and using the orthogonality of ψ

Replacing −km ′ n′ m n t mn mn and ψm′ n′ and the fact
that the right-hand side is zero, it can easily be shown that
"
∇t ψmn · ∇t ψm′ n′ ds = 0 . (6.193)
S

6-8.1 Waveguide Fields Orthogonality

In the previous subsection we demonstrated the orthogonality of eigenfunctions for an

arbitrary waveguide. In this section, we show that the electric fields for two different TM, two
different TE, or one TM and one TE mode in the transverse plane of an arbitrary waveguide
are orthogonal. Let us first consider two different TM modes. Recalling from Eq. (5.102a)
that the transverse electric field can be computed from
iβ z
Emn
t (r) = iβ ∇t ψmn (r) e , (6.194)

it follows that the orthogonality of the transverse electric field is established provided that
∇t ψmn (r) is orthogonal to ∇t ψ pq (r). This is demonstrated in Eq. (6.193).
For two TE modes, the transverse electric field for any given mode can be obtained from
Eq. (5.105b) and is given by
iβ z
Emn
t (r) = iω µ [∇t ψmn (r) × ẑ]e . (6.195)

The orthogonality of the transverse electric field for two TE modes can be established if we
can show that "
[∇t ψmn (r) × ẑ] · [∇t ψ pq (r) × ẑ] ds = 0 . (6.196)
S
Expanding the integrand of Eq. (6.196), we can show that

[∇t ψmn (r) × ẑ] · [∇t ψ pq (r) × ẑ] = ẑ · [∇t ψ pq (r) × ẑ] × ∇t ψmn (r)
= ẑ · [∇t ψmn (r) · ∇t ψ pq (r) ẑ − ∇tψmn (r) · ẑ ∇t ψ pq (r)] .
(6.197)

However, ẑ · ∇t ψmn (r) = 0, and hence Eq. (6.196) simplifies to

"
∇t ψmn (r) · ∇t ψ pq (r) ds = 0 , (6.198)

which obviously holds according to Eq. (6.193).

Finally, to show that the transverse electric field of a TE mode is orthogonal to the
transverse electric field of a TM mode, we need to show that
"
TE TM
(∇t ψmn × ẑ) · ∇t ψ pq ds = 0 . (6.199)
S
372 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Consider the following identity:

TE TM TE TM TE TM
∇t · (ẑ ψmn × ∇t ψ pq ) = ∇t × (ẑ ψmn ) · ∇t ψ pq − ẑ ψmn · ∇t × ∇t ψ pq . (6.200)
TM = 0, and applying the surface integral on both sides of Eq. (6.200), and then
But ∇t × ∇t ψ pq
applying the divergence theorem, leads to
I "
TE TM TE TM
(ẑ ψmn × ∇tψ pq ) · n̂ dℓ = ∇t × (ẑ ψmn ) · ∇t ψ pq ds , (6.201)
C S

where n̂ is the unit normal to the waveguide contour pointing outward. The integrand of the
contour integral in Eq. (6.201) can be written as
TM TE TM TE
(ẑ × ∇t ψ pq ) · n̂ψmn = (n̂ × ẑ) · ∇t ψ pq ψmn .

Since n̂ × ẑ is tangential to the waveguide contour and ∇t ψ pq TM ∼ E pq (r) is normal to the

t
contour, it follows that "
TE TM
(∇t ψmn × ẑ) · ∇t ψ pq ds = 0 ,
S
and hence Eq. (6.199) is satisfied. Note that in this case, p can be equal to m and q can be
equal to n.

6-8.2 Attenuation Rate in Waveguides of Arbitrary Cross Section

The attenuation rate due to losses of metallic walls in a waveguide with arbitrary cross section
can be obtained from Eq. (6.98):
I
|Htan |2 dℓ
α= C . (6.202)
4σ δs P(z)

The expressions for Htan and P(z) depend on the mode of propagation. For a TMmn wave, the
expressions for fields inside the waveguide are given by Eqs. (5.102a) and (5.102b):
 
2
E = E0 iβmn ∇t ψmn + kmn ψmn ẑ eiβmn z (6.203a)
and
H = −E0 iωε (∇t ψmn × ẑ)eiβmn z . (6.203b)

The power carried by the waveguide can be obtained from

" "
1 ∗ 1 2
P(z) = Re (E × H ) · ẑ dz = |E0 | βmn ωε |∇t ψmn |2 ds , (6.204)
2 S 2 S
6-8 Modal Expansion of Field Quantities 373

where we have used the fact that ∇t ψmn · ẑ = 0. Using Eq. (6.192), we can show that
" "
2 2 2
|∇t ψmn | ds = kmn ψmn ds .
S

Thus Eq. (6.204) is simplified to

"
1 2
P(z) = |E0 |2 βmn ωε kmn 2
ψmn ds . (6.205)
2 S

The tangential magnetic field on the waveguide wall can be obtained from Eq. (6.203b).
Denoting the unit vector on the waveguide wall as n̂,

∂ ψmn
|Htan | = |n̂ × H| = |E0 ωε | .
∂n
As a result, the expression for the attenuation given by Eq. (6.202) can be written as
I  2
∂ ψmn
ωε dℓ
TMmn C ∂n
α = " . (6.206)
2 2
2σ δs βmn kmn ψmn ds
S

Similarly, the attenuation rate for a TEmn wave in the waveguide can be computed from the
field expressions given by Eqs. (5.105a) and (5.105b):
 
2
H = H0 iβmn ∇t ψmn + kmn ψmn ẑ eiβ z (6.207a)
and
E = H0 (iω µ ) [∇t ψmn × ẑ] eiβ z . (6.207b)

The power carried by the waveguide can be obtained in the manner shown in Eq. (6.204) and
is given by
"
1 2
P(z) = |H0 | ω µβmn |∇t ψmn |2 ds
2 S
"
1 2 2 2
= |H0 | βmn ω µ kmn ψmn ds . (6.208)
2 S

Also, " #
 2
2 2 2 2 ∂ ψmn 4
|Htan | = |n̂ × H| = |H0 | βmn + kmn (ψmn )2 , (6.209)
∂t
374 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

where t̂ is a unit vector tangential to the contour of the waveguide cross section. Hence,
I
 2 
βmn |n̂ × ∇t ψmn |2 + kmn
4 2
ψmn dℓ
α TEmn = " .
2 2
2σ δs βmn ω µ kmn ψmn ds
S

6-8.3 Expansion of 2-D Scalar Green’s Function in Terms of Eigenfunctions

It is somewhat obvious that a z-directed electric or magnetic current produces TM or

TE modes in a metallic waveguide. Solutions of the wave equation for arbitrary current
distributions can be obtained from the scalar Green’s function:

(∇2t + kc2 ) g(ρ ρ′ ) = −δ (ρ

ρ ,ρ ρ − ρ′ ) , (6.210)

where ρ = x x̂ + y ŷ and ρ ′ = x′ x̂ + y′ ŷ. Since the set of eigenfunctions is complete,

XX
g(ρ ρ′ ) =
ρ ,ρ Amn ψmn (ρ ρ) . (6.211)
m n

Also, XX
ρ − ρ′ ) =
δ (ρ Bmn ψmn (ρ
ρ) . (6.212)
m n
Using Eq. (6.186) it can be shown that

ψmn (ρ ′ )
Bmn = " . (6.213)
2
ψmn (ρ ) ds
S

Substituting Eqs. (6.211) and (6.212) into Eq. (6.210), we have

XX X X ψmn (ρ ′ ) ψmn (ρ )
Amn (∇2t + kc2 ) ψmn (ρ ) = − " .
2
m n ψmn (ρ ) ds
S

2 ψ , which leads to
But ∇2t ψmn = −kmn mn

1 ψ (ρ ′ )
Amn = " mn . (6.214)
2 − k2
kmn c 2
ψmn (ρ ) d ρ
S
6-8 Modal Expansion of Field Quantities 375

Replacing this value of Amn in Eq. (6.211), the scalar Green’s function that satisfies the
waveguide boundary condition is then given by
XX ψmn (ρ ′ ) ψmn (ρ )
g(ρ ρ′ ) =
ρ ,ρ " .
2 2 2
m n (kmn − kc ) ψmn (ρ ) d ρ
S

For example, for a rectangular waveguide carrying a TM mode, the scalar Green’s function
can be written as
 mπ   mπ   nπ   nπ 
′
∞
XX ∞ sin x sin x sin y sin y′
ρ ,ρ
g(ρ ′
ρ)= a  a b   b .
mπ 2  nπ 2 2 ab
m=1 n=1 + − kc
a b 4
It is also interesting to note that
X ∞
∞ X
4  mπ   mπ   nπ   nπ 
ρ − ρ′) =
δ (ρ sin x sin x′ sin y sin y′ .
ab a a b b
m=1 n=1

6-8.4 Calculus of Variations for Estimation of Eigenvalues

The method of separation of variables is available for only a small number of three-
dimensional cavity structures and cross-sectional waveguide geometries. For waveguides and
cavities of arbitrary geometries, approximate numerical methods must be used to find the
cutoff frequencies and resonant frequencies. The method of calculus of variations provides
one such method. As shown in Eq. (6.191), the eigenvalues of the scalar wave equation can
be obtained from "
∇t ψ · ∇t ψ ds
2 R"
kc = , (6.215)
2
ψ ds
R
where ψ is an eigenfunction defined in R, the domain specifying the waveguide cross section.
Here we want to show among all functions f that are continuous and have a continuous first
derivative in R, the function that minimizes the functional
"
∇t f · ∇t f ds
N( f ) R
= " (6.216)
D( f ) 2
f ds
R

2 .
is an eigenfunction of the ∇2t operator. The minimum is equal to the lowest eigenvalue kc1
Suppose there exists a solution ψ1 that minimizes the functional defined by Eq. (6.216).
Then for an arbitrary function g in the class of admissible functions and an arbitrarily small
376 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

constant δ ,
2
N(ψ1 + δ g) > kc1 D(ψ1 + δ g) . (6.217)
Since ψ1 is the solution to the minimum problem
2
N(ψ1 ) = kc1 D(ψ1 ) .

Noting that
N(ψ1 + δ g) = N(ψ1 ) + δ 2 N(g) + 2δ N(ψ1 , g)
and
D(ψ1 + δ g) = D(ψ1 ) + δ 2 D(g) + 2δ D(ψ1 , g) ,
the inequality in Eq. (6.217) can be written as
2
2δ [N(ψ1 , g) − kc1 D(ψ1 , g)] + δ 2 [N(g) − kc1
2
D(g)] ≥ 0 , (6.218)

where
"
N(ψ1 , g) = ∇t ψ1 · ∇t g ds
R
and
"
D(ψ1 , g) = ψ1 g ds .
R

2 D(g) > 0,
Note here that the second term in Eq. (6.218) is a positive quantity, i.e., N(g) − kc1
since g is not minimizing the functional. Since Eq. (6.218) must be valid for any arbitrary
positive or negative δ and any function g, then
2
N(ψ1 , g) − kc1 D(ψ1 , g) = 0 , (6.219)

Otherwise for a given function g, we could choose a sufficiently small δ with a proper sign so
that
2
2δ [N(ψ1 , g) − kc1 D(ψ1 , g)] + δ 2 [N(g) − D(g)] ≤ 0 .
Expanding Eq. (6.219), we get
" "
2
∇t ψ1 · ∇t g ds − kc1 ψ1 g ds = 0 . (6.220)
R R

According to Green’s first identity

" I
2 ∂ ψ1
(∇t ψ1 · ∇t g + g ∇t ψ1 ) ds = g dℓ . (6.221)
R C ∂n
Using Eq. (6.220) in Eq. (6.221), the following relation is obtained:
" I
∂ ψ1
2
g(∇2t ψ1 + kc1 ψ1 ) ds = g dℓ . (6.222)
R C ∂n
6-8 Modal Expansion of Field Quantities 377

for an arbitrary function g. Let us first assume that g is a function with a vanishing value at
the waveguide boundary C. Then
"
2
g(∇2t ψ1 + kc1 ψ1 ) ds = 0 (6.223)
R

for all arbitrary g functions with a vanishing value at the boundary. The integral given by
Eq. (6.223) can vanish if and only if
2
∇2t ψ1 + kc1 ψ1 = 0 . (6.224)

That is, the function that minimizes functional Eq. (6.216) is an eigenfunction of the ∇2t
2 as its eigenvalue. If g is not zero on the waveguide boundary, since the
operator with kc1
left-hand side of Eq. (6.222) vanishes according to Eq. (6.224), then

∂ ψ1
= 0 on the boundary. (6.225)
∂n
Hence ∂ ψ1 /∂ n = 0 is called a natural boundary condition. For the TM case where ψ must
be zero on the boundary the class of admissible functions ( f ) must also satisfy the boundary
condition f = 0 on the waveguide boundary. However, for the TE case where ∂ ψ /∂ n = 0 on
the waveguide boundary, the condition that ∂ f /∂ n = 0 on the boundary to be in the class of
admissible functions is not necessary, according to Eq. (6.225).
The next eigenvalue and the corresponding eigenfunction are obtained by minimizing the
functional given by Eq. (6.216) subject to the following constraint:
"
f ψ1 ds = 0 , (6.226)
R

which requires the orthogonality of the next eigenfunction with that of the first one, as
mandated by Eq. (6.224). Since we have added a constraint to the first minimization problem,
2 obtained from Eq. (6.216) subject to Eq. (6.226) must be at least as large as k2 , that is,
kc2 c1
2 2 . Now let us assume that this minimization problem has a solution ψ such that
kc2 > kc1 2

2
N(ψ2 ) = kc2 D(ψ2 )

and
D(ψ2 , ψ1 ) = 0 .
As before, for an arbitrary function g in the class of admissible function and an arbitrary
parameter δ we must have
2
N(ψ2 + δ g) ≥ kc2 D(ψ2 + δ g) . (6.227)

It should be noted that the set of admissible functions is constrained by

D(ψ1 , g) = 0 .
378 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

For a general function p in the class of admissible functions not subject to the orthogonality
constraint, we may write
g = p − cψ1 , (6.228)
where
D(ψ1 , p)
c= . (6.229)
D(ψ1 )
Substituting Eq. (6.228) into Eq. (6.227), the inequality can be written as
2
N(ψ2 + δ p − δ cψ1 ) ≥ kc2 D(ψ2 + δ p − δ cψ1 ) .

Expanding the functionals N and D and using the argument that p and δ are arbitrary, the
functional coefficient of δ must vanish for the inequality to hold. That is,
2 2
N(ψ2 , p) − kc2 D(ψ2 , p) − c[N(ψ2 , ψ1 ) − kc2 D(ψ2 , ψ1 )] = 0 . (6.230)

We know that D(ψ2 , ψ1 ) = 0 and that from the first variational problem for an arbitrary p
function
2
N(ψ1 , p) − kc1 D(ψ1 , p) = 0 . (6.231)
Choosing p = ψ2 , Eq. (6.231) gives the following relations:

N(ψ1 , ψ2 ) = 0 . (6.232)

Therefore, Eq. (6.230) reduces to

2
N(ψ2 , p) − kc2 D(ψ2 , p) = 0 . (6.233)

But Eq. (6.233) is exactly similar to Eq. (6.220), so with the help of Green’s first identity it
can be shown that
∇2t ψ2 + kc2
2
ψ2 = 0 ,
2 is an eigenvalue and ψ is an eigenfunction of the ∇2 operator. We can also
indicating that kc2 2 t
get the natural boundary condition
d ψ2
= 0 on C.
dn
as before.
This process can be continued to get the kth eigenvalue and eigenfunction. That is, we
minimize the functional given by Eq. (6.216) subject to
"
f ψ j ds = 0 for j = 1, 2, . . . , k − 1 .
R

2 ≤ k2 ≤ k2 ≤ · · · .
This procedure results in obtaining a sequence of eigenvalues kc1 c2 c3
6-8 Modal Expansion of Field Quantities 379

Example 6-6: Estimation of Cutoff Frequency

Use the variation principle by choosing an approximate eigenfunction ψ to find the

eigenvalues of the dominant mode in a rectangular waveguide using Eq. (6.215).
Consider a rectangular waveguide with cross-sectional dimensions a and b (a > b)
supporting a TE mode. The eigenfunctions for this problem were found to be
 mπ   nπ 
ψmn (x, y) = cos x cos y ,
a b
∂
which satisfy the boundary condition ∂n ψmn = 0 on the rectangular contour of the waveguide.
The eigenvalues are given by
 mπ 2  nπ 2
2
kc(mn) = + n, m = 0, 1, 2, 3, . . .
a b
The lowest eigenvalue corresponds to
 π 2
2
kc(10) = .
a
Let us choose f (x, y) = (x − a2 ) as an estimate for the first eigenfunction. Then

∇ f = x̂ ,

and using functional Eq. (6.216) an upper bound for the lowest eigenvalue can be obtained
from Z aZ b
dx dy
2 0 0 a 12
k̃c = Z a Z b   =  3 = 2 ,
a 2 2 a a
x− dx dy 3 2
0 0 2
which is about 20% higher than the lowest eigenvalue.

6-8.5 Numerical Solution

The finite element method offers an efficient numerical procedure for the calculation of the
eigenvalues of a waveguide of an arbitrary cross section.§ This technique is well known and
here only a brief discussion of the method is outlined. For homogeneously filled waveguides
the longitudinal component of the electric (TM case) or magnetic (TE case) field, denoted
by ψ , satisfies the homogeneous Helmholtz equation (∇2 + Kc2 )ψ = 0, where Kc is the cutoff
wave number. It is shown that the solution to the Helmholtz equation minimizes the following

§ Silvester,
P., “A general high-order finite-element waveguide analysis program,” IEEE Transactions
Microwave Theory and Technique, vol. MTT-17, no. 4, April 1969.
380 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

functional: "
∇ψ · ∇ψ ds
S
F(ψ ) = " , (6.234)
2
ψ ds
S

and the minimum is equal to the smallest eigenvalue λ = Kc2 . To find the minima of the
functional, the eigenfunction ψ is approximated by a piecewise linear function. In this
approximation the cross section of the waveguide is discretized into small triangular elements
with unknown values of ψ at each node of the elements. For the eth element the linear function
in terms of the node values (ψ ej ) can be represented by¶

3
X
ψ (x, y) = N ej(x, y) ψ ej ,
j=1

where N ej(x, y) constitute the linear basis function given in terms of the element’s nodes x–y
coordinates. That is,
1
N ej(x, y) = (ae + bej x + cej y) , j = 1, 2, 3 (6.235)
2∆ e j
and
ae1 = xe2 ye3 − ye2 xe3 , be1 = ye2 − ye3 , ce1 = xe3 − xe2 ,
where ∆ e is the area of the triangle given by
1 e e
∆e = (b c − be2 ce1 ) ,
2 1 2
and the expressions for ae2 , . . . , ce3 are obtained from Eq. (6.235) by cyclic interchange of the
subscripts (1 → 2, 2 → 3, 3 → 1).
Substituting the piecewise linear function into the functional defined by Eq. (6.234) and
searching for the minima by setting δ /δ ψ ej = 0, the following matrix equation is obtained:

ÃΨ̃ = Kc2 B̃ Ψ̃ , (6.236)

where the elements of à and B̃ are given by

1
Ãeij = (be be + cei cej )
4∆ e i j
and
∆e
B̃eij = (1 + δi j ) ,
12
and δi j is the Kronecker delta. Since some of the nodes are shared among the adjacent

¶ Jin, J., The Finite Element Method in Electromagnetics, New York: John Wiley & Sons, 1993.
6-8 Modal Expansion of Field Quantities 381

elements, the vector Ψ̃ for the unconnected elements must be related to the vector Ψ for
connected elements using a connection matrix C.|| In this case,

Ψ̃ = C Ψ ,

and the matrix equation takes the following form:

AΨ = Kc2 B Ψ , (6.237)

where A = Ct à C and B = Ct B̃ C. Equation (6.237) is recognized as the generalized

eigenvalue problem, which can be solved for Kc numerically using a standard method.**

Example 6-7: Cutoff Frequency of a General Waveguide

Use the formulation given by Eq. (6.237) for the trapezoidal waveguide shown in Fig. 6-24
to find the cutoff frequencies and first two modes’ field distribution in the cross section of
the waveguide. Such geometry for waveguides is encountered at millimeter-wave (MMW) or
sub-MMW bands where silicon micromachining must be used to circumvent the tolerance
limitation of traditional milling machines. Chemical etching of silicon wafers with a
specific crystal orientation shows different anisotropic etch rates for different planes of
the crystal. For example, anisotropic wet etching of a h100i oriented silicon wafer creates
trapezoidal grooves with an angle of 54.74◦ . Usually ethylenediamine pyrocatechol (EDP) or
tetramethylammonium hydroxide (TMAH) etchants are used in this process. The depth of the
groove is determined by temperature, concentration, and time of etching. For the waveguide
considered here, the larger base of the trapezoid is b = 2.4 mm and the groove depth is
h = 1.2 mm.

b = 2.4 mm 1st mode 2nd mode

h = 1.2 mm

54.74˚

Figure 6-24: Configuration of a trapezoidal waveguide and the electric field distribution in the
waveguide cross section for the first two modes.

Solution: The cross section of the waveguide is discretized by small triangular elements, as
shown in Fig. 6-24, and Eq. (6.237) is solved to find the eigenvalues from which the cutoff
|| Ibid.
** Wilkinson, J. H., The Algebraic Eigenvalue Problem, New York: Oxford University Press, 1965.
382 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

frequencies are obtained. The process leads to (listed in order)

fc = 84.6 GHz, 117.4 GHz, 165 GHz, 165.4 GHz, 175.5 GHz.

Also, the field distribution of the first two modes is shown in Fig. 6-24.

6-9 Calculus of Variations for Estimation of Resonant

Frequencies in General Cavities
In the previous section, we developed a procedure for estimating the eigenvalues and
eigenvectors for waveguides of arbitrary cross section. It was also shown that for cavities
made from a straight section of such waveguides, of length c along the z-axis, the resonant
2 , using
frequencies can be determined from the eigenvalues of the waveguide, kmn
 2
2 2 2 ℓπ
k = ω µε = kmn + . (6.238)
c
However, there are no exact solutions for a general cavity of arbitrary shape. For the best
estimation of the resonant frequencies, a stationary formula similar to Eq. (6.215) is needed.
To arrive at such a formulation, self-reaction of sources and fields inside a cavity can be used
to show that such a functional is stationary. Consider a general-shape cavity filled with a
material having permittivity and permeability ε and µ , as shown in Fig. 6-25. Initially let’s
assume there exists a volumetric current density Ja (r) inside the cavity that produces fields
Ea (r) and Ha (r). These fields must satisfy the modified Ampère’s law:

∇ × Ha = −iωε Ea + Ja . (6.239)

The boundary condition mandates that n̂ × Ea = 0 on the surface of the cavity. However, if
we consider a magnetic surface current Jmsa on the cavity walls, then

Jmsa = −n̂ × Ea , (6.240)

μ, ε
n̂

PEC walls

Figure 6-25: The geometry of an arbitrary-shaped cavity filled with a homogeneous and iso-
tropic material having parameters µ and ε .
6-9 Calculus of Variations for Estimation of Resonant Frequencies in General Cavities 383

indicating that the tangential electric field can assume nonvanishing values. Now evaluating
the self-reaction of sources and fields using Eq. (4.92), the surface integral vanishes and we
have $
ha, ai = [Ea (r) · Ja (r)] dv − [Ha (r) · Jmsa (r)] dS = 0 . (6.241)

Noting that within the volume of the cavity

1
Ha (r) = ∇ × Ea (r) , (6.242)
iω µ
Eq. (6.239) can be rewritten as
1
Ja (r) = ∇ × ∇ × Ea(r) + iωε Ea (r) . (6.243)
iω µ
Substituting Eqs. (6.242) and (6.243) into Eq. (6.241), we get
 
$
1
ha, ai = Ea (r) · ∇ × ∇ × Ea(r) − k2 Ea (r) · Ea (r) dv
iω µ v

1
+ {∇ × Ea(r) · [n̂ × Ea (r)]} dS . (6.244)
iω µ
To simplify Eq. (6.244), we can make use of

∇ · [∇ × Ea (r) × Ea (r)] = Ea (r) · ∇ × ∇ × Ea(r) − ∇ × Ea (r) · ∇ × Ea (r) (6.245)

and
∇ × Ea (r) · [n̂ × Ea (r)] = n̂ · {Ea (r) × [∇ × Ea(r)]} . (6.246)
Also, note that
$
∇ · [∇ × Ea(r) × Ea (r)] dv = − n̂ · [∇ × Ea (r) × Ea (r)] dS . (6.247)
v

The reason for the negative sign is that the unit vector n̂ used here is pointing towards the
inside of the cavity. Using Eqs. (6.245), (6.246), and (6.247) in Eq. (6.244), the self-reaction
expression takes the following form:

2
ha, ai = n̂ · {Ea (r) × [∇ × Ea(r)]} dS
iω µ
$
1
+ [∇ × Ea (r)] · [∇ × Ea (r)] dv
iω µ
k2
$
− [Ea (r) · Ea (r)] dv
iω µ
=0. (6.248)
384 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Equation (6.248) can now be used to find the cavity eigenvalue

$
[∇ × Ea (r)] · [∇ × Ea (r)] dv + n̂ · {Ea (r) × [∇ × Ea (r)]} dS
2 v S
k = $ . (6.249)
[Ea (r) · Ea (r)] dv
v

Of course, inside a cavity at resonance, we can have nonzero field quantities in the absence
of any sources like Ja (r) or Jmsa (r). In the absence of sources, Eq. (6.241) shows that the
self-reaction is zero. However, Eq. (6.249) can still be used to find the eigenvalues (k2 ) for
all modes in the cavity if Ea (r) is a corresponding field distribution for that mode. Equation
(6.249) also indicates that if we were to hypothesize an electric field distribution that could
result from certain sources in the cavity, that field distribution can approximately estimate
an eigenvalue that has the closest field distribution to the guessed field distribution. This
approximate estimate can be accurate if the function given by Eq. (6.249) is stationary with
respect to the chosen field function.
To prove the stationarity of the functional, let us assume that

Ea (r) = Ee (r) + δ g(r) , (6.250)

where Ee (r) is the exact field distribution for a resonant mode, g(r) is an arbitrary but
admissible function (n̂ × g(r) = 0 on the boundary), and δ is a small real number. In this
case the estimated value of k2 is a function of δ and can be written as
N[Ee(r) + δ g(r)]
k2 (δ ) = , (6.251)
D[Ee (r) + δ g(r)]
with the exact eigenvalue
N[Ee(r)]
k2 = k2 (0) = . (6.252)
D[Ee (r)]
Here N(·) and D(·) refer to the numerator and denominator functionals of Eq. (6.249). To
prove stationarity we need to show that

∂ k2 (δ ) 
=0,
∂δ  δ =0

which can be written explicitly as

 
∂N ∂D 
∂ k2  ∂δ D−N ∂δ 
=  . (6.253)
∂ δ δ =0 D2 
δ =0

Using Eq. (6.252) in Eq. (6.253), it can be shown that

 
∂N ∂D 
∂ k2  ∂δ − k2 ∂δ 
=  . (6.254)
∂ δ δ =0 D 
δ =0
6-9 Calculus of Variations for Estimation of Resonant Frequencies in General Cavities 385

But

∂ N 
$
=2 ∇ × g(r) · ∇ × Ee(r) dv
∂ δ δ =0 v

+2 n̂ · {g(r) × [∇ × Ee(r)]} ds
S

+2 n̂ · {Ee (r) × [∇ × g(r)]} ds . (6.255)
S

Noting that n̂ · {Ee (r) × [∇ × g(r)]} = ∇ × g(r) · [n̂ × Ee(r)] and the fact that tangential Ee (r)
on S is zero (n̂ × Ee (r) = 0), the third integral in Eq. (6.254) vanishes. Also, since g(r) is an
admissible function (n̂ × g(r) = 0), the second integral vanishes. Now noting that

∇ × g(r) · ∇ × Ee(r) = ∇ · {g(r) × [∇ × Ee (r)]} + g(r) · ∇ × ∇ × Ee(r)

and the fact that inside the source-free cavity

∇ × ∇ × Ee(r) = k2 Ee (r) ,

Eq. (6.255) reduces to


∂ N 
$
2
= −2 g(r) × [∇ × Ee (r)] · n̂ dS + 2k Ee (r) · g(r) dv . (6.256)
∂ δ δ =0 v

Again, the first term evaluates to zero, and we have


∂ N 
$
2
= 2k Ee (r) · g(r) dv . (6.257)
∂ δ δ =0 v

We also need to evaluate


∂ D 
$
=2 g(r) · Ee (r) dv . (6.258)
∂ δ δ =0 v

Using Eqs. (6.257) and (6.258) in Eq. (6.254) shows that


∂ k2 
=0,
∂ δ δ =0

which proves that the functional given by Eq. (6.249) is stationary. That is, the function that
minimizes Eq. (6.249) is a solution to the vector wave equation

∇2 E + k 2 E = 0

or
∇ × ∇ × E − k2 E = 0 .
386 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Example 6-8: Estimation of Resonant Frequency from the Stationary

Formulation

Consider a cavity of spherical shape with radius a. Using the stationary formula given by
Eq. (6.249) and hypothesizing an appropriate field distribution inside the cavity, find the
resonant frequency.
Solution: As will be shown later, using the method of separation of variables, an analytical
formulation for the field expressions can be obtained. The dominant mode has the simplest
field distribution and is generated from a TM to r potential (a set of fields in a spherical
coordinate system when the magnetic field does not have an r-component). The boundary
conditions on the cavity surface are as follows: Eθ = Eφ = 0 at r = a. As will be shown later,
these are satisfied when
ĵn (ka) = 0 ,
where ĵn (kr) represents the Schelkunoff spherical Bessel function of the first kind and nth
order. The dominant mode corresponds to the first zero of ĵ1 (ka); that is, when

ka = 2.744 .

So the exact solution for k is given by

2.744
k= .
a
As a guess, we are seeking a solution that is independent of φ and takes a simple form with θ
and r. We should also choose our field distribution so that it satisfies the boundary condition.
Let’s assume that the electric field is given by

Ea (r, θ , φ ) = Er r̂ + Eθ θ̂
θ,

with
 r 2
Er = cos θ
a
and
r 
Eθ = sin θ −1 .
a
Obviously Eθ (a, θ , φ ) = 0 at r = a. Using this field distribution, it can be shown that
  r   r 2 
sin θ
∇ × Ea(r) = 1−2 − φ.
φ̂
r a a

Noting that n̂ = r̂ and the fact that ∇ × Ea is along φ̂

φ, it can be shown that

r̂ · [Ea (r) × (∇ × Ea(r))] = |∇ × Ea (r)| Eθ (r) ,

6-9 Calculus of Variations for Estimation of Resonant Frequencies in General Cavities 387

but Eθ (a) = 0, which makes the second integral in the numerator of Eq. (6.249) vanish. Also,
using the fact that
Z π /2
4
sin3 θ d θ =
0 3
and
Z π
2
cos2 θ sin θ d θ = ,
0 3
the numerator of Eq. (6.249) is calculated to be
52
N= a = 1.1555a .
45
Also, the denominator is calculated to be
88 3
D= a = 0.1397a3 .
630
Hence from Eq. (6.249) we estimate
N 8.2717
k2 = = ,
D a2
or
2.876
k= .
a
This value is very close to the exact solution, k = 2.744/a, with less than 5% error.
 388 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Summary
Concepts

• A dielectric slab with planar interfaces and • Resonant structures, like rectangular cavities,
an index of refraction larger than that of can support multiple modes at a set of
its surroundings can support TM and TE discrete frequencies that depend on the physical
propagating waves inside the dielectric slab. dimension of the cavity.
• Different modes of propagation with different • In practice, cavities are made of good con-
field patterns can be supported depending on the ductors that have high but finite conductivity.
index of refraction contrast and the dielectric For such cavities initial energy established at
slab thickness. For each mode there is a cutoff each mode decays as a function of time and
frequency below which that mode cannot be this decay depends on the quality factor of the
established nor propagate. cavity.
• Impedance surfaces with purely reactive im- • A transmission-line circuit model for each mode
pedance can also support TE and TM surface in a rectangular waveguide can be created
waves. Inductive surfaces can support TM and used to analyze the effects of localized
waves and capacitive surfaces can support TE discontinuities in waveguides.
waves. • For general waveguides with arbitrary cross
• Reactive impedance surfaces can be engineered section, the cutoff wave number is always a
using metallic surfaces with periodic grooves positive real number that is only a function
or using a metal-backed substrate with printed of the geometrical shape and dimension of the
metallic patches on top. Both inductive and waveguide’s cross section.
capacitive reactive impedance surfaces can • The field generating potentials that satisfy
be realized with the metallic patches over a a 2-D wave equation subject to Dirichlet’s
relatively thin substrate. (TM) or Neumann’s (TE) boundary conditions,
• A metallic pipe of rectangular cross section, which are known as eigenfunctions, form an
whether it is hollow or dielectric-filled, can orthogonal and complete set of functions in the
support TE and TM waves. Depending on the 2-D domain of the waveguide’s cross section.
geometry of the cross section, different modes • To find the cutoff wave number for waveguides
with different field patterns and propagation of arbitrary geometry, the calculus of variations
constants may be supported by such structures. is used. A functional is defined from the
• The attenuation rate of wave propagation in eigenvalues and shown that the functions that
rectangular waveguides is substantially lower minimize that functional is an eigenfunction and
than that of ordinary transmission lines. that the value of the functional is an eigenvalue.
• Metallic waveguides are dispersive (phase • A numerical procedure, known as a finite
velocity is a function of frequency), and the element procedure, is presented to find the
speed signal envelope known as group velocity eigenfunctions and eigenvalues.
is different from the phase velocity inside the
waveguide.
SUMMARY 389

Important Equations

Transcendental equations for different modes of a dielectric-plate waveguide with

thickness t and constitutive parameters µ2 , ε2 in a background medium with
constitutive parameters µ1 , ε1
 

 k2ct 2  ν t 2  ω t √ 2

 + = µ 2 ε 2 − µ1 ε 1

 2 2 2

    

 ε1 k2ct k2ct νt

 tan = TM odd mode



 ε2 2 2 2

    
ε1 k2ct k2ct νt
− cot = TM even mode

 ε 2 2 2 2

    

 µ1 k2c t k2ct νt



 tan = TE odd mode

 µ2 2 2 2

    

 µ1 k2ct k2ct νt

− cot = TE even mode
µ2 2 2 2

Boundary condition for an impedance surface with surface impedance Zs :

n̂ × [n̂ × E(r)] = −Zs n̂ × H(r)

Propagation constant of a propagating surface wave (Zs = iXc ) on a capacitance

impedance surface: s  
TE 2
ωµ 2
β = k +
Xc
and inductive impedance surface (Zs = −iXi):
q
β TM = k2 + ω 2 ε 2 Xi2

Surface impedance of a metallic surface with periodic grooves of depth d supporting

surface waves:

Zs = −iη0 tan(k0 d) thin periodic blades

w
Zs = −i η0 tan(k0 d) thick corrugation with period P and groove width w
P
J1 (k0 d)
Zs = −iη0 triangular groove
J0 (k0 d)
390 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Important Equations (continued)

Surface impedance of metal-backed dielectric substrate with printed metallic

patches with perodicity P:

η1 cos θt tan(k1 cos θt d)

ZsTM = −i
e η1 cos θt tan(k1 cos θt d)
1 − ωCP
η1
tan(k1 cos θt d)
cos θt
ZsTE = −i
e η1 tan(k1 cos θt d)
1 − ωCP
cos θt
Propagation constant and guided wavelength for a rectangular waveguide with
dimensions a, b for mnth mode:
r  mπ 2  nπ 2
βmn = k2 − −
a b
r   
1 m 2 n 2
(λgmn )−1 = − −
λ2 2a 2b

Cutoff frequency for mnth mode:

r   
(mn) 1 m 2 n 2
fc = √ +
2 µε a b

Phase and group velocity:

ω ω
up = = p
β k2 − kc2
dω dup
ug = = up + β
dβ dβ

Relations between phase velocity and group velocity:

up ug = c2

Power loss per unit length of a waveguide made with metallic walls of
conductivity σ : I
1
PL = |Ht |2 dℓ
2σ δs
SUMMARY 391

Important Equations (continued)

Waveguide attenuation rate: I

|Ht |2 dℓ
α=
2σ δs Pmn
with
(
1
ab m , 0 and n , 0 ,
Pm,n = η kβmn kc,2 mn × 2
p 1 m = 0 or n = 0 ,
as power carried by TEmn mode, and
ab
Pm,n = kβmn kc,2 mn
8η
as power carried by TMmn mode.
Resonant frequency of a rectangular cavity with dimensions a, b, c:
s
1  m 2  n 2  ℓ 2
fmnℓ = √ + +
2 µε a b c

Quality factor of a cavity due to metallic loss:

$
ωεσ δs |E|2 dv
V
Q̆s =
|Ht |2 ds
s

Quality factor of a cavity due to dielectric loss:

ε′
Q̆d =
ε ′′
Total quality factor:
1 1 1
= +
Q̆ Q̆s Q̆d
Damping factor (γ ) and relaxation time of a cavity:
1 ω
γ= =
τ 2Q
392 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Important Equations (continued)

Total stored energy at a resonance:

$
ε
w = we + wm = 2we = |E|2 dv
2

Series impedance per unit length (Z) and shunt admittance per unit length (Y ) of
transmission-line equivalent model for a waveguide:

kc2 
Z = −iω µ − TM
iωε 
Y = −iωε

Z = −iω µ 
kc2 TE
Y = −iωε − 
iω µ
General waveguide:
2-D wave equation for eigenfunctions of a general waveguide:
2
∇2t ψmn + kmn ψmn = 0

with ψmn = 0 (TM) or ∂ ψmn /∂ n = 0 (TE) on the waveguide boundary.

Properties of eigenfunction ψmn :
"
ψmn (r) ψm′ n′ (r) ds = 0 orthogonality
S
"
∇ψmn (r) · ∇ψm′ n′ (r) ds = 0
S

2 ):
Properties of eigenvalues (kmn
2 2 ∗
kmn = (kmn ) >0 Eigenvalues are real and positive

Variational formulation: The function that minimizes the following functional is an

eigenfunction and the value of the functional is an eigenvalue:
"
∇t ψ · ∇t ψ ds
2 R"
kc =
ψ 2 ds
R

Variational formulation for a general cavity:

# 
[∇ × E a (r)] · [∇ × E a (r)] dv + n̂ · {Ea (r) × [∇ × Ea(r)]} dS
k2 = v # S

v
[Ea (r) · Ea (r)] dv
 PROBLEMS 393

Important Terms Provide definitions or explain the meaning of the following terms:
attenuation rate functional rectangular cavity
calculus of variations generalized eigenvalue problem rectangular waveguide
cavity damping factor group velocity stationary formulation
cavity relaxation time guided wavelength surface resistivity
circuit models for waveguides isotropic reactive impedance surface telegrapher’s equation
cutoff frequency k–β diagram TE mode
cutoff wave number metallic septa TE wave impedance
dielectric plate waveguide natural boundary condition TM mode
Dirichlet’s boundary condition Neumann’s boundary condition TM wave impedance
eigenfunction phase velocity waveguide
eigenvalue quality factor for cavities waveguide discontinuity
finite element reactive impedance surface waveguide mode

PROBLEMS
Waveguides

6.1 The TE mode fields in and above a metal-backed dielectric slab of thickness t/2 (see
Fig. P6.1) are given as follows:

 E1y = Aiω µ1 ν e−ν x eiβ z ,
In region 1: H = −Aiβ ν e−ν xeiβ z ,
 1x
H1z = −Aν 2 e−ν x eiβ z ,

 E2y = Biω µ2 k2c sin(k2c x) eiβ z ,
In region 2: H = −Biβ k2c sin(k2c x) eiβ z ,
 2x 2 cos(k x) eiβ z ,
H2z = Bβ k2c 2c

2 = k2 − β 2 , ν 2 = β 2 − k2 , and ε = ε = ε .
where k2c 2 1 1 2 0

Figure P6.1: The geometry of a metal-backed dielectric slab with planar interfaces and infinite
extent.
394 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

(a) Write down the equations that determine k2c and ν .

(b) Give the expression for the total z-propagating power in the waveguide.
(c) Assuming µ2 = 4µ1 and µ1 = µ0 , what is the minimum thickness t necessary to support
a propagating TE mode?
6.2 Consider a parallel-plate waveguide formed by conductors covering the planes y = 0
and y = b, as shown in Fig. P6.2.

Figure P6.2: A parallel-plate waveguide made from infinite metallic sheets separated by a
distance b and parallel to the x–z plane.

(a) Show that the field

 nπ y 
Ex = E0 sin e−γ z n = 1, 2, 3, . . .
b
defines a set of TEn modes, and show that the field
 nπ y 
Hx = H0 cos e−γ z n = 0, 1, 2, . . .
b
defines a set of TMn modes, where in both cases
r 
nπ 2
γ= − k2 .
b

(b) Show that the cutoff frequencies of the modes are

n
fc = √ .
2b εµ

(c) Investigate the possibility of propagating a TEM mode (try to derive it from either TM
or TE waves).
PROBLEMS 395

6.3 Consider a parallel-plate waveguide, as shown in Fig. P6.3. Regions 0 and 1 have
different dielectric constants ε0 and ε1 . Let the waveguide be infinite in the y- and z-directions.
Consider wave propagation in the +z-direction.

ε
ε

Figure P6.3: The geometry of a partially filled parallel-plate waveguide. The load is a dielectric
slab with planar interfaces parallel to the waveguide walls.

(a) Find the Hertz vector potential Π for TM waves in both regions 0 and 1.
(b) Find the electric fields in regions 0 and 1.
(c) Find the set of equations that can be used to compute the propagation constant inside
the waveguide.

6.4 Consider a rectangular waveguide whose side walls are made from a perfect magnetic
conductor, as shown in Fig. P6.4. The waveguide is filled with material 1 for z < 0 and with
material 2 for z > 0. A TMz wave is incident from the left (i.e., from material 1). Calculate
the reflection and transmission coefficient at the z = 0 interface.
Hint: For TMz modes, the Hertz vector potential is given by Π = Πz ẑ. In region 1,
Πz,1 = Ψ(x, y) (Aeiβ1 z + Be−iβ1 z ), and in region 2, Πz,2 = Ψ(x, y) Ceiβ2 z , where Ψ(x, y) satisfies
the scalar wave equation ∇2t + kc2 Ψ = 0.

6.5 Consider a metallic waveguide with dimensions a and b. We place a resistive sheet at
z = 0 with sheet impedance Zs , as shown in Fig. P6.5. The boundary conditions for resistive
sheets are provided in Problem P5.10 (see Eqs. (5.252), (5.253), and (5.255)). The waveguide
is supporting TE10 mode. For this mode, find the fields in region 1 and region 2 of the
waveguide.

6.6 Consider a rectangular waveguide with two horizontal PEC walls at y = 0 and y = b
and two vertical PMC walls at x = 0 and x = a. Derive the expressions for the appropriate
vector potential, electric and magnetic fields, cutoff frequencies, and wave impedance for TM
and TE waves. Identify the lowest order modes for each set. Does this structure support TEM
modes? (Hint: You may need to solve separately for kc = 0.)
396 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Figure P6.4: A rectangular waveguide with PMC walls. Half of the waveguide is filled with
a material with constitutive parameters µ1 , ε1 , and the other half is filled with a material with
constitutive parameters µ2 , ε2 .

s s

Figure P6.5: A thin resistive sheet (impedance sheet) placed in the cross section of the
waveguide acts as a discontinuity and reflects some of the incident wave.
PROBLEMS 397

6.7 Consider that the rectangular waveguide shown in Fig. P6.7 is partially filled with
dielectric material 2 from y = 0 to y = c and partially filled with material 1 from y = c to
y = b (a > b and 0 < c < b).

b
ε
c
ε x
0
a

Figure P6.7: A partially filled rectangular waveguide with a dielectric slab filling the lower part
of the waveguide.

(a) Show that TEz and TMz modes will fail to satisfy the boundary conditions, except for
the TEzm0 case. Solve for the TEzm0 mode and find the transcendental equation that
governs the propagation constant given by
µ2 µ1
tan(k2y c) + tan[k1y (b − c)] = 0 ,
k2y k1y
q q
where k1y = k12 − β 2 and k2y = k22 − β 2 .
(b) For the TEzm0 mode, what are the wave impedances in the two regions?
(c) Find the hybrid modes (having both Ez and Hz ) by solving TMymn modes, which have
electric Hertz potential in the y-direction: Π = ŷ ψ (x, y) eiβ z .
(d) Letting ε1 ≈ ε2 and µ1 = µ2 = µ , show that the cutoff frequency for the dominant TMxmn
mode is obtained by s
π ε2 (b − c) + ε1 c
ωc = .
a µε1 ε2 b

6.8 Consider a rectangular waveguide with a centered dielectric slab, as shown in Fig. P6.8.
The problem contains three homogeneous regions. We know we cannot find pure TEz or TMz
modes, but for such a structure, TEx or TMx modes can be introduced.
To solve for these modes, considering Πx and Πmx instead of Πz and Πmz .
(a) Derive the equations for the E and H fields using a procedure that led to Eqs. (5.93) and
(5.96).
398 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

ε0 ε1 ε0

x
a

Figure P6.8: A partially filled rectangular waveguide with a dielectric slab of thickness t filling
the middle part of the waveguide. The planar sides of the slab are parallel to the short side walls
of the waveguide.

(b) Find Πx , E, and H by applying the boundary condition.

(c) Show that the characteristic equations for determining the propagation constants of the
TEx modes are    
kx0 a−d kx1 d
cot kx0 = tan kx1
µ0 2 µ1 2
for the even mode, and
   
kx0 a−d kx1 d
cot kx0 =− cot kx1
µ0 2 µ1 2
for the odd mode. For TMx
   
kx0 a−d kx1 d
tan kx0 =− tan kx1
ε0 2 ε1 2
for the even mode, and
   
kx0 a−d kx1 d
tan kx0 = cot kx1
ε0 2 ε1 2
for the odd mode, where
 nπ 2
2
kx0 + + kz2 = k02
b
and
 nπ 2
2
kx1 + + kz2 = k12 .
b
PROBLEMS 399

6.9 Show that the TMz modes in the isosceles right-triangular waveguide shown in Fig. P6.9,
with base side a, can be found by specializing the results for a rectangular waveguide to the
square case b = a, and then finding the linear combinations of the eigenfunctions of the form
fmn = (ψmn + αψnm ) for an appropriate constant α that vanishes on the diagonal side defined
by y = x as well as the two sides defined by y = 0 and x = a. First show why fmn can be an
eigenfunction, and then find α by satisfying the boundary conditions. Then find the first five
modes of the waveguide. Do the same for TEz modes. Note: The sum of two eigenfunctions
is an eigenfunction only if their two eigenvalues are the same.

√2 a
a

x
z a

Figure P6.9: The configuration of a metallic waveguide with an isosceles cross section.

6.10 Consider a rectangular waveguide with dimensions a and b with a > b, as shown in
Fig. P6.10. Suppose the vertical side wall of the waveguide at x = a is a reactive impedance
surface with surface impedance Zs = −iXs . Find the propagation constant for TE modes.

y z
s

Figure P6.10: The geometry of a rectangular waveguide with three sides made from metallic
surfaces and one side made from a reactive impedance surface.

6.11 Suppose a general waveguide is operated at a frequency below its cutoff frequency.
Find the time-average net amount of stored energy within a segment of the waveguide of
length L in terms of the associated eigenfunction. (Hint: Apply the complex Poynting vector
to the volume defined by the waveguide walls and the surfaces defined by z = 0 and z = L.)
400 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

6.12 Following the procedure of Example 6-2, find the attenuation rote in a rectangular
waveguide TE20 and TE11 . Plot the attenuation rates for these modes for a rectangular
waveguide with a = 2b = 2.89 cm as a function of frequency.

Cavity

6.13 A rectangular cavity is made from a piece of copper (σ = 5.813 × 107 S/m) with
a = 4.755 cm and b = 2.215 cm, as shown in Fig. P6.13. The cavity is filled with polyethylene
(εr = 2.25, tan δ = ε ′′ /ε ′ = 0.0004). If the resonant frequency of TE101 mode is 5 GHz, find
the required length c and the Q̆ of the cavity.

b z
x

a c

Figure P6.13: A metallic rectangular cavity made from copper with finite conductivity.

6.14 Given a rectangular cavity of dimensions 2 cm × 3 cm × 1 cm:

(a) How
√ many electromagnetic
√ modes of wavelength λ exist in the cavity in the range
4/ 5 cm ≤ λ ≤ 8/ 13 cm? Find the wavelengths.
(b) Write down the expression for the electric field for each mode.
(c) Approximately how many modes are there in the range 0.01 cm ≤ λ ≤ 0.011 cm?
6.15 Consider an air-filled rectangular cavity with dimensions a, b, and c, and which
is partially filled with a dielectric material ε , as shown in Fig. P6.15. Find the resonant
frequencies for such a cavity.

Modal Expansion

6.16 Consider a standard rectangular waveguide, extending from z = 0 to ∞, as shown in

Fig. P6.16. Suppose we excite the waveguide with a field H = H0 x̂ at the z = 0 aperture. Find
the amplitudes of the different modes within the waveguide.
PROBLEMS 401

b
,ε
d

, ε εr c

Figure P6.15: A rectangular cavity partially filled with a dielectric material.

z
y

H = H0 x̂
x

Figure P6.16: A rectangular waveguide excited by a uniform magnetic field H = H0 x̂ for

0 < x < a.

6.17 Consider an arbitrary waveguide with eigenfunctions ψmn supporting a z-directed

current along the waveguide. Suppose the current distribution in the cross section of the
waveguide is given by Jz (ρ ρ ) eiβ z . Using Green’s function inside the waveguide, find the
resulting scalar potential ψ , noting that ψ satisfies
−i
∇2t ψ + kc2 ψ = Jz (ρ
ρ) .
ωε
Also, for the same waveguide, consider a current source of the following form:
 a  πy 
J(x, y) = I0 δ x − sin eiβ z ẑ .
2 b
Find the scalar potential ψ and the fields inside the waveguide.
402 Chapter 6 Cartesian Wave Functions: Guiding Structures and Resonators

Figure P6.18: Configuration of a junction between two rectangular waveguides with dissimilar
widths but the same height. The incident wave propagating in the +z direction gets partially
reflected at the junction. Away from transition on both sides of the junction only the dominant
modes can propagate, but higher-order modes are excited locally.

6.18 Consider a rectangular waveguide with a discontinuity at z = 0 that is terminated in

matched loads at both ends, as shown in Fig. P6.18. Assume that the propagating mode in
both sections z < 0 and z > 0 is TEZ10 . However, in the vicinity of the discontinuity, the fields
can be expressed as the superposition of TExmn modes. Assuming that Ey at the discontinuity
(z = 0) is the same as that of the incident wave, find the following:
(a) The amplitude of the excited TExmn modes at z = 0.
(b) The aperture susceptance due to this discontinuity, defined as
 ∗ 
P
Im[Ya ] = Im ,
|V |2
where P is the z-directed complex power at z = 0:
"
P= E × H∗ · dS ,
z=0

and V is the maximum voltage across the aperture:

Zb
V= Ey dy .
0
x=0, z=0
PROBLEMS 403

6.19 Suppose the material making a waveguide is such that the boundary condition on its
surface is given by
d ψmn
+ αψmn = 0 ,
dn
with α > 0. Show that potential ψmn that satisfies the 2-D wave equation
2
∇2t ψmn + kmn ψmn = 0
2 that are real positive numbers.
has eigenvalues kmn
6.20 Consider a waveguide with boundary condition
d ψmn
+ αψmn = 0 .
dn
2 and
If ψmn and ψm′ n′ are solutions to the 2-D wave equation with different eigenvalues of kmn
2
km′ n′ , show that "
ψmn ψm′ n′ ds = 0 .
S
 7 Chapter
Cylindrical Wave Functions
and Their Applications

Chapter Contents Objectives

Overview, 405 Upon learning the material presented in this chapter, you
7-1 Wave Functions in the Cylindrical Coordinate should be able to:
System, 406
7-2 The Circular Dielectric Waveguide, 421 1. Solve boundary-value problems associated with
7-3 Green’s Function Solutions for Some wave equations and objects whose boundary
Canonical Problems, 436 coincides with constant surfaces in a cylindrical
7-4 Scattering from a Metallic Circular coordinate system.
Cylinder, 452
7-5 Integral Representation of Bessel 2. Understand the properties of cylindrical wave
Functions, 460 functions and choose an appropriate set of such
7-6 2-D Green’s Function for Homogeneous functions for the problem at hand.
Media in the Presence of a Metallic
Wedge, 467 3. Determine TE and TM modes of propagation and
7-7 Asymptotic Evaluation of a Diffracted Field their cutoff frequencies in circular metallic and
from a Metallic Wedge, 474 dielectric (optical-fiber) waveguides.
Chapter Summary, 486
Problems, 493 4. Learn about other forms of cylindrical wave
propagation in the radial and circumferential
directions.

5. Express Green’s function in free space and in

circular and sectoral waveguides, as well as
external configurations such as metallic cylinders
and wedges, in terms of cylindrical wave
functions.

6. Use the integral representation of Bessel

functions in the asymptotic evaluation of fields
scattered by metallic wedges.

404
 405

Overview

This chapter deals with the solution of wave equations in a cylindrical coordinate system.
TE and TM wave solutions to two-dimensional boundary-value problems are obtained for
special cases where the boundary of the object coincides with the coordinate surfaces of a
cylindrical coordinate system. Initially the method of separation of variables is applied to the
homogeneous scalar wave equation expressed in terms of cylindrical coordinate variables. In
situations where such solutions are admissible, the solutions for the z-function and φ -function
are found to be harmonic functions, whereas the solution for the radial function is a Bessel
function. Depending on whether the desired domain includes the entire range 0 ≤ φ ≤ 2π
or not, the order of the Bessel functions may or may not be integer number. The properties
of cylindrical wave functions are presented so that a proper solution based on the domain
of interest can be chosen. A linear combination of cylindrical wave functions, in general,
represents the solution to the wave equation. Using the cylindrical wave function, the TE
and TM field expressions are obtained for waveguides with a circular cross section and the
properties of different modes of such waveguides are examined. It is shown that the different
modes of cylindrical waveguides are orthogonal and the total power carried by a wave in the
waveguide is the sum of the power carried by the propagating modes of the wave. Cylindrical
wave functions are also used to study propagating modes on circular dielectric waveguides,
which are known as optical fibers. It is shown that the dominant mode of propagation in
a circular dielectric waveguide is a hybrid TE and TM mode. That is, both modes must be
considered together when applying the boundary conditions to arrive at a solution. It is also
shown that for azimuthally symmetric modes (where fields are independent of φ ), the circular
dielectric waveguide can support TE and TM modes. Other metallic waveguide geometries,
such as angular-sector waveguides that can support Bessel functions of noninteger order, are
also considered. The resonant modes and quality factor for a cylindrical cavity are obtained
for both TE and TM modes, and an expression for the quality factor of the dominant mode
(TM010 ) due to ohmic losses of the walls is found. Other types of waveguides with propagation
directions along ρ and φ , known as radial waveguides and circumferential waveguides, are
also presented and the behavior of the fields for such waveguides, which are bounded by
two parallel metallic plates perpendicular to the z-axis, are studied as well. The responses
of the wave equation to a line source (Green’s function) in free space, inside a cylindrical
waveguide, or outside of a metallic cylinder or infinite wedge, are obtained and studied.
The expressions for Green’s functions in terms of a series expansion of eigenfunctions are
shown to converge poorly. Alternative solutions, referred to as the formal solution, which
are derived directly from the wave equation, are presented and shown to be computationally
more efficient. A solution based on Fourier transformation of the 2-D free-space Green’s
function is obtained, from which the addition theorem for Bessel functions is derived. Plane-
wave scattering from a metallic circular cylinder as well as a metallic wedge are presented
in terms of a series expansion of appropriate cylindrical wave functions. Then the integral
representation of Bessel functions is derived and used to derive an asymptotic solution in the
far-field region of a metallic wedge.
406 Chapter 7 Cylindrical Wave Functions and Their Applications

7-1 Wave Functions in the Cylindrical Coordinate System

In certain scattering and radiation problems where the current distributions’ or objects’
boundaries coincide with cylindrical coordinate surfaces, solutions may be best obtained using
a comparable cylindrical coordinate system. In a source-free and homogeneous region, the
scalar potential satisfies the Helmholtz equation given by

∇2 ψ + k 2 ψ = 0 . (7.1)

In cylindrical coordinates, Eq. (7.1) takes the following form:

 
1 ∂ ∂ψ 1 ∂ 2ψ ∂ 2ψ
ρ + 2 + 2 + k2 ψ = 0 . (7.2)
ρ ∂ρ ∂ρ ρ ∂φ2 ∂z
We assume that the scalar potential can be expressed in terms of the product of three functions,
each being a function of only ρ , φ , or z variables; i.e.,

ψ = R(ρ ) Φ(φ ) Z(z) . (7.3)

Equation (7.2) can be separated into three differential equations. By substituting Eq. (7.3) into
Eq. (7.2) and then dividing both sides by RΦZ, the process leads to
 
1 ∂ ∂R 1 ∂ 2Φ 1 ∂ 2Z
ρ + 2 + + k2 = 0 . (7.4)
ρR ∂ ρ ∂ρ ρ Φ ∂ φ 2 Z ∂ z2

In Eq. (7.4) (1/Z)(∂ 2 Z/∂ z2 ) is only a function of z, and the rest of the terms are independent
of z. This can only happen if
1 ∂ 2Z
= −kz2 (7.5)
Z ∂ z2
for some constant kz2 .
Equation (7.4) can now be rewritten as
 
ρ d dR 1 d2Φ
ρ + + kρ2 ρ 2 = 0 , (7.6)
R dρ dρ Φ dφ 2

where kρ2 = k2 − kz2 . Now, in Eq. (7.6), (1/Φ)(d 2 Φ/d φ 2 ) is only a function of φ , and the other
terms are independent of φ . In order for Eq. (7.6) to hold, the following must be true:

1 d2Φ
= −ν 2 (7.7)
Φ dφ 2
for some constant ν . As a result the following must be true:
 
d dR  
ρ ρ + (kρ ρ )2 − ν 2 R = 0 . (7.8)
dρ dρ

Clearly, the solution for R(ρ ) will depend on ν , and as we will see shortly, the form of the
7-1 Wave Functions in the Cylindrical Coordinate System 407

solution is in terms of Bessel functions of order ν . Hence, to emphasize that fact, we will
henceforth add a subscript ν to R.
Upon expanding Eq. (7.8) into

d 2 Rν dRν
(kρ ρ )2 2
+ (kρ ρ ) + [(kρ ρ )2 − ν 2 ]Rν = 0 , (7.9)
d(kρ ρ ) d(kρ ρ )

and then using a change of variable, namely ρe = kρ ρ , Eq. (7.9) can be modified into
"  2 #
d 2 Rν 1 dRν ν
2
+ + 1− Rν = 0 , (7.10)
d ρe ρe d ρe ρe

which is known as Bessel’s equation of order ν .

The Helmholtz equation is now separated into three second-order differential equations.
Solutions to Eqs. (7.5) and (7.7) are harmonic functions of the form

Z(z) = Aeikz z + Be−ikz z (7.11a)

and
Φ(φ ) = Ceiνφ + De−iνφ . (7.11b)

As noted earlier, Eq. (7.8) is a Bessel equation of order ν , whose solution can be expressed in
terms of a linear combination of real functions Nν (kρ ρ ) and Jν (kρ ρ ):

Rν (kρ ρ ) = AJν (kρ ρ ) + BNν (kρ ρ ) , (7.12)

where subscript ν has been added to R to emphasize its order, Jν (kρ ρ ) is the Bessel function
of the first kind and ν th order and Nν (kρ ρ ) is the Bessel function of the second kind and
ν th order. The solution of Eq. (7.8) can also be expressed as a linear combination of Hankel
(1) (2)
functions of the first kind (Hν (kρ ρ )) and Hankel functions of the second kind (Hν (kρ ρ ))
kind, defined by
(1)
Hν (kρ ρ ) = Jν (kρ ρ ) + iNν (kρ ρ ) (7.13a)
and
(2)
Hν (kρ ρ ) = Jν (kρ ρ ) − iNν (kρ ρ ) . (7.13b)

In situations where the domain of interest includes points whose φ coordinate spans the entire
domain [0, 2π ], it is necessary that

ψ (ρ , φ , z) = ψ (ρ , φ + 2π , z) . (7.14)

In other words, ψ must be periodic in φ if ψ is to be single-valued. In such a case we must

choose ν = n, where n is an integer and the possible solutions for φ are of the following form:

Φ(φ ) = Aeinφ + Be−inφ , (7.15)

408 Chapter 7 Cylindrical Wave Functions and Their Applications

1
n=0
n=1
n=2
0.5 n=3
n=4
Jn(x)
0

−0.5
0 5 10 15 20
x

Figure 7-1: Bessel functions of the first kind with integer order.

−1 n=0
n=1
n=2
Nn(x) −2
n=3
n=4
−3

−4

−5
0 5 10 15 20
x

Figure 7-2: Bessel functions of the second kind with integer order.

which is a periodic function with period 2π . In this case, the order of Bessel’s functions are
all integer. Figures 7-1 and 7-2 show the plot of Bessel functions of the first kind and the
second kind for integer orders. As shown, the Bessel functions of the first kind are regular
everywhere, whereas the Bessel functions of the second kind have singularity at x = 0.
To construct proper solutions for ψ , certain properties of Bessel functions must be taken
into account. Among the possible solutions presented, only Jn (kρ ρ ) are nonsingular at ρ = 0.
Also, the asymptotic behavior of Bessel functions as kρ ρ → ∞ is of interest to identify
7-1 Wave Functions in the Cylindrical Coordinate System 409

outgoing and incoming waves. The asymptotic behaviors of Bessel functions are given by
s  
2 2n + 1
lim Jn (kρ ρ ) ≈ cos kρ ρ − π , (7.16a)
kρ ρ →∞ π kρ ρ 4
s  
2 2n + 1
lim Nn (kρ ρ ) ≈ sin kρ ρ − π , (7.16b)
kρ ρ →∞ π kρ ρ 4
s
(1) 2
lim Hn (kρ ρ ) ≈ ei[kρ ρ −(2n+1)π /4] , (7.16c)
kρ ρ →∞ π kρ ρ
and
s
(2) 2
lim Hn (kρ ρ ) ≈ e−i[kρ ρ −(2n+1)π /4] . (7.16d)
kρ ρ →∞ π kρ ρ

These expressions are obtained from an asymptotic evaluation of the integral representation of
(1)
Bessel function. It is clear from the sign of the exponents in their expressions for Hn (kρ ρ )
(2)
that it represents an outgoing cylindrical wave whereas Hn (kρ ρ ) represents an incoming
wave.
The solution to the Helmholtz equations consists of a linear combination of the elementary
cylindrical wave functions for all discrete values of n and continuous values of kz . Hence, a
solution of the Helmholtz equation may be expanded in terms of the elementary cylindrical
functions as
+∞ Z +∞
X  
ψ (ρ , φ , z) = An (kz ) Jn (kρ ρ ) + Bn (kz ) Nn (kρ ρ ) einφ eikz z dkz . (7.17)
n=−∞ −∞

Noting that q
kρ = k2 − kz2 (7.18)
in Eq. (7.17), as kz varies from −∞ to +∞, kρ can become imaginary. For values of |kz | < k
the argument of the Bessel functions is real in a lossless medium and the elementary wave
functions can be decomposed into incoming and outgoing propagating cylindrical waves.
However, for the rest of kz values (|kz | > k), the argument of the Bessel functions becomes
pure imaginary, and the radial wave function become exponentially decaying. In these cases,
it is conventional to use the modified Bessel functions In and Kn defined by

Jn (iαρ ) = (i)n In (αρ ) (7.19a)

and
(1) 2 n+1
Hn (iαρ ) = (i) Kn (αρ ) . (7.19b)
π
Figures 7-3 and 7-4 show the modified Bessel functions of the first and second kind for integer
orders, which demonstrate the exponential decay behavior of the modified Bessel functions
410 Chapter 7 Cylindrical Wave Functions and Their Applications

20
n=0
n=1
15 n=2
n=3
n=4
In(x) 10

0
0 2 4 6
x

Figure 7-3: The modified Bessel functions of the first kind with integer order.

10
n=0
n=1
8
n=2
n=3
6 n=4
Kn(x)
4

0
0 1 2 3 4 5
x

Figure 7-4: The modified Bessel functions of the second kind with integer order.

of the second kind as their arguments increase. The large-argument approximation can be
obtained from Eqs. (7.16a) and (7.16c):
eαρ
lim In (αρ ) ≈ √ (7.20a)
αρ →∞ 2παρ
and
r
π −αρ
lim Kn (αρ ) ≈ e . (7.20b)
αρ →∞ 2αρ
7-1 Wave Functions in the Cylindrical Coordinate System 411

It should be emphasized here that the modified Bessel functions do not satisfy the Bessel
equation given by Eq. (7.10), but they instead satisfy the modified Bessel equation given by
  ν 2 
d 2 Rν 1 dRν
+ − 1 + Rν = 0 . (7.21)
dρ 2 Rν d ρ R

7-1.1 Some Useful Properties of Bessel Functions

As mentioned previously, the radial solution in the cylindrical coordinate system satisfies the
following ordinary second-order differential equation:
 
d 2 Rν 1 dRν ν2
+ + 1 − 2 Rν = 0 , (7.22)
d ρe2 ρe d ρe ρe

where ρe represents kρ ρ in Eq. (7.8). The series solution for Rν , where ν can be any real or
complex number, is given by
∞
X  ν +2m
(−1)m ρe
Rν (ρe) = , (7.23)
m! Γ (ν + m + 1) 2
m=0

where the gamma function is defined as

Z ∞
Γ (x) = t x−1 e−t dt , (7.24)
0

with the well-known property

1
Γ (x) =
Γ (x + 1) .
x
Since Eq. (7.22) is quadratic with respect to ν , a second solution for Eq. (7.22) can easily be
obtained simply by replacing ν with −ν . However, this cannot be done when ν is an integer.
In situations where ν = n, the gamma function Γ simplifies to Γ (ν + m + 1) = (n + m)!, and
the resulting solution is the Bessel function of the first kind given by
∞
X (−1)m  ρ n+2m
Jn (ρ ) = . (7.25)
m! (n + m)! 2
m=0

In this case J−n (ρ ) is not independent of Jn (ρ ); in fact,

J−n (ρ ) = (−1)n Jn (ρ ) . (7.26)

This can easily be shown by changing n to −n in Eq. (7.25) and changing index m to m′ + n
(noting that the reciprocal of the gamma function of a negative integer is zero). In general, the
Bessel function of the second kind is defined using
1
Nν (ρ ) = [Jν (ρ ) cos νπ − J−ν (ρ )], (7.27)
sin νπ
412 Chapter 7 Cylindrical Wave Functions and Their Applications

which is independent of Eq. (7.25) even for integer-order ν = n. Also, the following
recurrence relation is useful for computation of higher-order Bessel functions in terms of
the lower-order counterparts and vice versa:
2ν
Rν −1 + Rν +1 = Rν . (7.28)
ρ
To compute the derivatives, the following recurrence relations may be used:
dRν 1
= [Rν −1 − Rν +1] , (7.29a)
dρ 2
d
[ρ ν Rν (ρ )] = ρ ν Rν −1 , (7.29b)
dρ
d ν
[ρ Rν (ρ )] = −ρ −ν Rν +1 . (7.29c)
dρ
Another useful relationship is the Wronskian, given by
2
W = Jν (ρ ) Nν′ (ρ ) − Jν′ (ρ ) Nν (ρ ) = . (7.30)
πρ
(1)
A similar Wronskian can easily be obtained between Jν (ρ ) and Hn (ρ ):

(1) ′ (1) i2
Jν (ρ ) Hν (ρ ) − Jν′ (ρ ) Hν (ρ ) = . (7.31)
πρ
The recurrence relations for the modified Bessel functions are given by
2ν
Iν −1 − Iν +1 = Iν (7.32a)
ρ
and
2ν
Kν −1 − Kν +1 = − Kν , (7.32b)
ρ
and the derivatives are given by
dIν 1
= [Iν −1 + Iν +1] , (7.33a)
dρ 2
dKν 1
= − [Kν −1 + Kν +1] , (7.33b)
dρ 2
d
[ρ ν Iν (ρ )] = ρ ν Iν −1 (ρ ) , (7.33c)
dρ
and
7-1 Wave Functions in the Cylindrical Coordinate System 413

d
[ρ ν Kν (ρ )] = −ρ ν Kν −1 (ρ ) . (7.33d)
dρ

7-1.2 Orthogonality Relations

Unlike most orthogonal functions and polynomials, the orthogonality of Bessel functions
exists among the same kind and same order, but with differently scaled argument. This stems
from the fact that
Z a
a  ′ ′

Jν (k p ρ ) Jν (kq ρ ) ρ d ρ = 2 kq Jν (k p a) Jν (kq a) − kq Jν (kq a) Jν (k p a) (7.34)
0 k p − kq2

for arbitrary real numbers k p and kq . Denoting zeros of Jν (x) by xν p and zeros of Jν′ (x) by
xν′ p , where p is an integer index (p = 1, 2, 3, . . .), by choosing
xν p
kp = , (7.35a)
a
xν q
kq = , (7.35b)
a
or
x′ν p
kp = , (7.35c)
a
x′ν q
kq = , (7.35d)
a
the right-hand side of Eq. (7.34) vanishes when p , q. That is, the orthogonality relations for
Bessel functions of the first kind are given by
Z a 
xν p   xν q 
Jν ρ Jν ρ ρ dρ = 0 (7.36a)
0 a a
and
Z a  ′   ′ 
xν p xν q
Jν ρ Jν ρ ρ dρ = 0 . (7.36b)
0 a a
In cases where p = q, both the numerator and the denominator of Eq. (7.34) go to zero.
L’Hôpital’s rule must be applied to evaluate the value of the integral. Substituting x for k p a
and x′ for kq a, Eq. (7.34) can be evaluated from

a2 [x′ Jν (x) Jν′ (x′ ) − x Jν (x′ ) Jν′ (x)] a2 x Jν′ (x) [Jν (x) − Jν (x′ )]
lim′ ≈
x→x x2 − x′ 2 2x(x − x′ )
a2 x Jν′ (x) Jν′ (x) (x − x′ ) a2  ′ 2
= = J (x) . (7.37)
2x(x − x′ ) 2 ν
414 Chapter 7 Cylindrical Wave Functions and Their Applications

Now, using the recurrence relation

ν
Jν′ (x) = Jν (x) − Jν +1 (x) , (7.38)
ρ
and substituting xν p for x, it becomes obvious that

Jν′ (xν p ) = −Jν +1 (xν p ) . (7.39)

Therefore the norm of the Bessel functions of the first kind is given simply by
Z a 
xp  a2
Jν2 ρ ρ dρ = [Jν +1 (xν p )]2 . (7.40)
0 a 2
L’Hôpital’s rule can also be applied differently:

a2 [x′ Jν (x) Jν′ (x′ ) − x Jν (x′ ) Jν′ (x)] a2 x Jν (x) [Jν′ (x) + Jν′′ (x) (x′ − x) − Jν′ (x)]
lim′ ≈
x→x x2 − x′ 2 2x(x − x′ )
a2
≈− Jν (x) Jν′′ (x) . (7.41)
2
Directly from the Bessel equation we have
 
1 ′ ν2
Jν′′ (x) = − Jν (x) − 1 − 2 Jν (x) . (7.42)
x x

Now letting x = xν′ p , it can easily be shown that

Z a  ′  "  2 2 #
xν p a2 ν  2
Jν2
ρ ρ dρ = 1− ′ Jν (x′ν p ) , (7.43)
0 a 2 xν p

which represents the norm of Bessel functions scaled by xν′ p /a corresponding to Eq. (7.36b).

7-1.3 The Cylindrical Waveguide

Consider a metallic waveguide with circular cross-section of radius a, as shown in Fig. 7-5.
The solution for the field quantities in the source-free region of a circular waveguide can
be expressed in terms of TM and TE modes, for which we have respectively the following
boundary conditions:

ψ (ρ
ρ) = 0 , ρ on S , (TM mode)
and
∂ψ
=0, ρ on S . (TE mode)
∂ρ
7-1 Wave Functions in the Cylindrical Coordinate System 415

nˆ = ρ̂

Figure 7-5: A metallic circular-cylindrical waveguide of radius a.

For a scalar potential ψ that satisfies the 2-D Helmholtz equation in the cylindrical coordinate
system,  
1 ∂ ∂ψ 1 ∂ 2ψ
ρ + 2 + kρ2 ψ = 0 . (7.44)
ρ ∂ρ ∂ρ ρ ∂φ2
The method of separation of variables can be used to express ψ in terms of the elementary
wave functions in a cylindrical coordinate system. Since the domain of interest covers the
entire range of φ ([0, 2π ]), then (as noted earlier in conjunction with Eq. (7.14)) the order of
Bessel function ν must be an integer (n). Also, since the domain includes ρ = 0, only the
Bessel functions of the first kind are admissible. Hence a solution for ψ can be written as

sin(nφ ) ,
ψ (ρ , φ ) = An Jn (kρ ρ ) (7.45)
cos(nφ ) .

The choice of sin(nφ ) or cos(nφ ) is arbitrary, and one can be produced from the other
by rotating the x–y plane by 90◦ about the z-axis. Because of this circular symmetry of
the problem, there is a multiplicity of two for all eigenvalues. That is, the modes are
degenerate. For simplicity, we retain only the cosine function from hereon forward. Applying
the boundary condition for a TM mode requires that

Jn (kρ a) = 0 . (7.46)

Denoting the zeros of Jn by xnp , where p is the ascending order of the zeros of Jn , kρ can be
obtained from xnp
kρ = . (7.47)
a
The propagation constant in the waveguide is therefore given by
r  x 2
np
β = k2 − . (7.48)
a
The roots of the first few orders of Bessel function Jn (X ) are listed in Table 7-1. The TM
cutoff frequency for a specific mode and for a waveguide filled with homogeneous material
416 Chapter 7 Cylindrical Wave Functions and Their Applications

Table 7-1: Roots xnp of the first few orders of Bessel function Jn (x) for TM modes. Here, n
represents the order of the Bessel function and p represents the order of its zeros xnp .

p n
↓ 0 1 2 3 4
1 2.405 3.832 5.138 6.380 7.588
2 5.520 7.016 8.417 9.761 11.065
3 8.654 10.173 11.620 13.015 14.372
4 11.792 13.324 14.796

Table 7-2: Roots x′np of the first few orders of Jn′ (x′ ) for TE modes. Here, n represents the
order of Jn′ (x′ ) and p represents the order of its zeros x′np .

p n
↓ 0 1 2 3 4
1 3.832 1.841 3.054 4.201 5.317
2 7.016 5.331 6.706 8.015 9.282
3 10.173 8.536 9.969 11.346 12.682
4 13.324 11.706 13.170

having permittivity ε and permeability µ is given by

(np) xnp
kc = kρ =, (7.49a)
a
TM(np) √ xnp
2π fc µε = , (7.49b)
a
and
TM(np) xnp
fc = √ . (7.49c)
2π a µε

Applying the boundary condition for a TE mode, we need to have

Jn′ (kρ a) = 0 .

Denoting the roots of Jn′ (X ) by x′np , where again p is an ascending order of zeros of Jn′ (X ), kρ
is given by
x′np
kρTE = . (7.50)
a
The values of these roots for the first few orders of the Bessel functions (first kind) are
provided in Table 7-2. The cutoff frequency for TE modes of a homogeneously filled circular
7-1 Wave Functions in the Cylindrical Coordinate System 417

waveguide is given by

TE(np) x′np
fc = √ . (7.51)
2π a µε

The dominant mode of propagation in a circular waveguide corresponds to the smallest value
among the roots xnp and x′np . Referring to Tables 7-1 and 7-2, the dominant mode is TE11 .
The field distribution of the first few modes of TE and TM modes are shown in Fig. 7-6. The
order of propagating modes are TE11 , TM01 , TE21 , (TM11 , TE01 ), TE31 , etc. The eigenvalues
for TM11 and TE01 are the same, and thus these modes are degenerate.
The field distribution for a TM mode can be obtained from Eqs. (5.102a) and (5.102b),
and it is given by
x  x 
np np
Eρ (ρ , φ , z) = E0 iβ Jn′ ρ cos(nφ ) eiβ z , (7.52a)
a a
E0 iβ n  xnp 
Eφ (β , φ , z) = − Jn ρ sin(nφ ) eiβ z , (7.52b)
ρ a
 x 2  x 
np np
Ez (ρ , φ , z) = E0 Jn ρ cos(nφ ) eiβ z , (7.52c)
a a
E0 iωε n  xnp 
Hρ (ρ , φ , z) = Jn ρ sin(nφ ) eiβ z , (7.52d)
ρ a
and
x  x 
np np
Hφ (ρ , φ , z) = E0 iωε ′
Jn ρ cos(nφ ) eiβ z . (7.52e)
a a
The field expressions for TE modes can be obtained from Eqs. (5.105a) and (5.105b):
 ′   ′ 
xnp xnp
Hρ (ρ , φ , z) = H0 iβ ′
Jn ρ cos(nφ ) eiβ z , (7.53a)
a a
 ′ 
H0 iβ n xnp
Hφ (ρ , φ , z) = − Jn ρ sin(nφ ) eiβ z , (7.53b)
ρ a
 ′ 2  ′ 
xnp xnp
Hz (ρ , φ , z) = H0 Jn ρ cos(nφ ) eiβ z , (7.53c)
a a
 ′ 
H0 iω µ n xnp
Eρ (ρ , φ , z) = − Jn ρ sin(nφ ) eiβ z , (7.53d)
ρ a
and
 ′   ′ 
xnp xnp
Eφ (ρ , φ , z) = −H0 iω µ Jn′
ρ cos(nφ ) eiβ z . (7.53e)
a a
418 Chapter 7 Cylindrical Wave Functions and Their Applications

Figure 7-6: The field distribution of the first few modes of TE and TM waves in a cylindrical
waveguide. The color shows the field intensity, the solid lines indicate the electric field, and the
dashed lines indicate the magnetic field.
7-1 Wave Functions in the Cylindrical Coordinate System 419

Example 7-1: Attenuation Rate in Circular Waveguides

Find an expression for the attenuation rate α due to ohmic losses of the metallic surface of a
circular waveguide of radius a for TM and TE modes.
Solution: The expression for attenuation rate for a general waveguide is presented in
Eqs. (6.206) and (6.209), for TM and TE modes, respectively. Using the following
eigenfunctions in these expressions provides the attenuation rate:
x 
TM np
ψnp (ρ , φ ) = Jn ρ cos(nφ ) (7.54a)
a
and
 ′ 
TE
xnp
ψnp (ρ , φ ) = Jn ρ cos(nφ ) , (7.54b)
a

where as before xnp and x′np represent the zeros of Jn (x) and Jn′ (x), respectively. The following
integrals are needed in Eqs. (6.206) and (6.209):
I !2 Z !2 (
TM
∂ ψnp 2π TM
∂ ψnp  x 2 2π n=0,
np 2
dℓ = a dφ = a [Jn+1 (xnp )]
∂n 0 ∂ρ a π n,0,
(7.55a)
and
ZZ ah  2π
xnp i2 2
"
TM 2
(ψnp ) ds = Jn ρ cos nφ ρ d ρ d φ
S 0 0 a
(
a2 2 2π n=0,
= [Jn+1 (xnp )] (7.55b)
2 π n,0.

In the derivation of Eqs. (7.55a) and (7.55b), we used Eqs. (7.39) and (7.40), respectively.
Also, for the TE fields we need to evaluate
I Z 2π !
TE 2
2 1 ∂ ψnp
|n̂ × ∇t ψnp | dℓ = a dφ
0 a2 ∂φ
(
n2  ′
2 0 n=0,
= Jn (xnp ) (7.56a)
a π n,0,
I Z 2π
 2
(ψnpTE 2
) dℓ = Jn (x′np ) cos2 nφ a d φ
0
(
 ′
2 2π n=0,
= a Jn (xnp ) (7.56b)
π n,0,
and
420 Chapter 7 Cylindrical Wave Functions and Their Applications

" Z 2π Z a  
x′np 2 2
2
(ψnp ) ds = Jn cos nφ ρ d ρ d φ
S 0 0 a
"  2 # (
a2 n  ′
2 2π n=0,
= 1− ′ Jn (xnp ) (7.56c)
2 xnp π n,0.

Substituting Eqs. (7.55a) and (7.55b) into Eq. (6.206) provides the following expression for
the attenuation rate of TM waves in a circular waveguide:
ωε k
α TMnp = = r  x 2 . (7.57)
σ δs βnp a np
σ δs η a k2 −
a
√
This shows that in cases where k ≫ (xnp /a), the attenuation rate increases with f
(embedded in 1/δs ).
Similarly, using Eqs. (7.56a), (7.56b), and (7.56c) in Eq. (6.209), the expression for the
attenuation rate for TE waves is given by
 
x′np 2
βnp n2 a
α TEnp =  ′  "  2 # + "  2 #
xnp 3 n n
σ δs ω µ a 1− ′ σ δs ω µ a 1 − ′ βnp
a xnp xnp
 s  ′ 2  ′ 2 
xnp xnp
 n2 k2 − 
1  a a 
= " 
 2 #  + s 
 ′ 2
x  ′ 2  . (7.58)
n  np xnp 
σ δs η ka 1 − ′ a2 k 2−
xnp a a

It is interesting to note that for TE0p the attenuation rate simplifies to

!2
x′0p
1 a
α TE0p = v !2 . (7.59)
σ δs η ka u
u ′
tk2 − x0p
a

In situations where k ≫ x′0p /a, the attenuation rate decreases with frequency as 1/ f 3/2 . Hence
the attenuation from such modes can, in principle, go to zero as f → ∞. Note, however, that
in such cases the waveguide can support many modes, and maintaining the TE0p as the only
mode becomes very challenging in practice.
7-2 The Circular Dielectric Waveguide 421

125 μm
8.3 μm
d εd μd

Fiber
c εc μc
Cladding Outside jacket

Figure 7-7: A single-mode optical fiber with a core of radius a and index of refraction
√ √
nd = µd εd surrounded by a large cladding background with index of refraction nc = µc εc .
At λ = 1500 nm, a typical value for the index of refraction for a core made from doped silicon is
n ≈ 1.45, and for a cladding made from pure silica a typical value is nc ≈ 1.44.

7-2 The Circular Dielectric Waveguide

In this section we examine how electromagnetic waves are guided by a circular dielectric
rod with permittivity εd and permeability µd embedded in a uniform dielectric medium with
permittivity εc and permeability µc , as shown in Fig. 7-7. As in circular metallic waveguides,
we are seeking a solution such that all field quantities show a dependence with respect to z
of the form eiβ z , which indicates propagation along the z-axis. This structure, known as an
optical fiber, is the most prevalent structure for data transmission. Optical fibers are used
for long-distance (up to a few thousand kilometers) data transfer, and internet communication
networks, as well as applications in medicine (such as endoscopy) and other industrial sensors.
The dielectric waveguide modes can be determined by expanding the fields in terms of
cylindrical wave functions inside the core and in the cladding region. The fiber is designed so
that the wave is confined within the core and the cladding supports only evanescent waves.
The cladding around the fiber is chosen to be thick enough so that no fields can reach the
outside jacket. The scalar potential in the core region takes the following form:

ψI = Jn (kρI ρ ) einφ , (7.60)

where
(kρI )2 = kd2 − β 2 = ω 2 µd εd − β 2 . (7.61)
Since we desire the field to be confined within the core and around it, the potential chosen in
cladding must be of the following form:
(1)
ψII = Hn (kρII ρ ) einφ , (7.62)
422 Chapter 7 Cylindrical Wave Functions and Their Applications

where
(kρII )2 = kc2 − β 2 = −(β 2 − ω 2 µc εc ) = −ν 2 , (7.63)
(1) (1)
for some positive real number ν . The radial function Hn (kρII ρ ) = Hn (iνρ ) can be written
in terms of the modified Bessel function of the second kind (Kn ) as defined in Eq. (7.19b) and
shown in Fig. 7-4. It turns out that, in the most general case, the field expansion base on TE
to z or TM to z alone cannot satisfy all the necessary boundary conditions. Superposition of
TE and TM modes must be used to arrive at a satisfactory solution. In region 1 where r ≤ a,
the longitudinal electric and magnetic fields can be written as

EzI (ρ , φ ) = (kρI )2 An Jn (kρI ρ ) einφ eiβ z (7.64a)

and
HzI (ρ , φ ) = (kρI )2 Bn Jn (kρI ρ ) einφ eiβ z , (7.64b)

and in region 2, r ≥ a, these fields are expressed as

(1)
EzII (ρ , φ ) = (kρII )2Cn Hn (kρII ρ ) einφ eiβ z (7.65a)
and
(1)
HzII (ρ , φ ) = (kρII )2 Dn Hn (kρII ρ ) einφ eiβ z . (7.65b)

In view of Eqs. (5.98) and (5.101), the other field components take the following form:
 
nω µd
I I ′ I
Eρ (ρ , φ ) = iβ kρ Jn (kρ ρ ) An − Jn (kρ ρ ) Bn einφ eiβ z ,
I
(7.66a)
ρ
 
−nβ
I
Eφ (ρ , φ ) = Jn (kρ ρ ) An − iω µd kρ Jn (kρ ρ ) Bn einφ eiβ z ,
I I ′ I
(7.66b)
ρ
 
nωεd
I
Hρ (ρ , φ ) = Jn (kρ ρ ) An + iβ kρ Jn (kρ ρ ) Bn einφ eiβ z ,
I I ′ I
(7.66c)
ρ
and
 
βn
HφI (ρ , φ ) = iωεd kρI Jn′ (kρI ρ ) An − Jn (kρI ρ ) Bn einφ eiβ z . (7.66d)
ρ
In region 2 where r > a, the field expressions can be obtained in a similar manner and are
given by
 
II (1)′ II nω µc (1) II
II
Eρ (ρ , φ ) = iβ kρ Hn (kρ ρ ) Cn − Hn (kρ ρ ) Dn einφ eiβ z , (7.67a)
ρ
 
−nβ (1) II (1)′
EφII (ρ , φ ) = Hn (kρ ρ ) Cn − iω µc kρII Hn (kρII ρ ) Dn einφ eiβ z , (7.67b)
ρ
 
nωεc (1) II (1)′
HρII (ρ , φ ) = Hn (kρ ρ ) Cn + iβ kρII Hn (kρII ρ ) Dn einφ eiβ z , (7.67c)
ρ
7-2 The Circular Dielectric Waveguide 423

and
 
(1)′ β n (1) II
HφII (ρ , φ ) = iωεc kρII Hn (kρII ρ ) Cn − Hn (kρ ρ ) Dn einφ eiβ z . (7.67d)
ρ
The relations between An , Bn , Cn , and Dn for the various modes can be established by applying
the boundary conditions at the interface between the core and the surrounding medium
(cladding); that is,

EzI (a, φ ) = EzII (a, φ ) , (7.68a)

HzI (a, φ ) = HzII (a, φ ) , (7.68b)
EφI (a, φ ) = EφII (a, φ ) , (7.68c)
HφI (a, φ ) = HφII (a, φ ) . (7.68d)

Using Eqs. (7.64a) and (7.65a) in Eq. (7.68a) and Eqs. (7.64b) and (7.65b) in Eq. (7.68b), we
get
(1)
(kρI )2 Jn (kρI a) An = (kρII )2 Hn (kρII a) Cn (7.69a)
and
(1)
(kρI )2 Jn (kρI a) Bn = (kρII )2 Hn (kρII a) Dn . (7.69b)

Also, using Eqs. (7.66b) and (7.67b) in Eq. (7.68c) and using Eqs. (7.66d) and (7.67d) in
Eq. (7.68d), two more equations for the unknowns are obtained:

nβ nβ (1) II (1)′
Jn (kρI a) An + iω µd kρI Jn′ (kρI a) Bn = Hn (kρ a) Cn + iω µc kρII Hn (kρII a) Dn (7.70a)
a a
and
nβ (1)′ nβ (1) II
iωεd kρI Jn′ (kρI a) An − Jn (kρI a) Bn = iωεc kρII Hn (kρII a) Cn − Hn (kρ a) Dn . (7.70b)
a a
Equations (7.69a)–(7.70b) constitute four equations for the four unknowns. To arrive at a
nonzero solution, the determinant of the coefficients must vanish. That is,
 
 (kI )2 J (kI a) 0
(1)
−(kρII )2 Hn (kρII a) 0 
 ρ n ρ 
 
 (1) 
 0 (kρI )2 Jn (kρI a) 0 (kρII )2 Hn (kρII a) 
  = 0 . (7.71)
 nβ 
 a Jn (kρI a) iω µd kρI Jn′ (kρI a) − naβ Hn(1) (kρII a) −iω µc kρII Hn(1)′ (kρII a)
 
 
iωεd kI J ′ (kI a) − nβ Jn (kI a) −iωεc kII Hn(1)′ (kII a) nβ (1) II 
ρ n ρ a ρ ρ ρ a Hn (kρ a)

Alternatively, from Eqs. (7.69a) and (7.69b) we find that

Bn Dn
= (7.72)
An Cn
424 Chapter 7 Cylindrical Wave Functions and Their Applications

and
!2 (1)
An kρII Hn (kρII a)
= . (7.73)
Cn kρI Jn (kρI a)

Substituting these into Eqs. (7.70a) and (7.70b), the following transcendental equation
involving kρII and kρI is obtained:
" (1)′
#" (1)′
#
µd Jn′ (kρI a) µc Hn (kρII a) εd Jn (kρI a) εc Hn (kρII a)
ω2 − −
kρI a Jn (kρI a) kρII a Hn(1) (kII a) kρI a Jn′ (kρI a) kρII a Hn(1) (kII a)
ρ ρ
 !2 !2 2 
2 2 1 1  .
=n β − I (7.74)
kρII a kρ a

Using Eq. (7.19b), we can show that

(1)′ (1)′
Hn (kρII a) Hn (iν a) Kn′ (ν a)
(1)
= (1)
= −i .
Hn (kρII a) Hn (iν a) Kn (ν a)

It then follows that the transcendental equation given by Eq. (7.74) can be written as
    
µd Jn′ (u) µc Kn′ (υ ) kd2 Jn′ (u) kc2 Kn′ (υ ) 2 2 1 1 2
+ + =n β + , (7.75)
u Jn (u) υ Kn (υ ) uµd Jn (u) υ µc Kn (υ ) υ 2 u2

where u = kρI a and υ = ν a. We also recall that

(kρI )2 = kd2 − β 2
and
ν 2 = β 2 − kc2 .

Hence
u2 + υ 2 = a2 (kd2 − kc2 ) . (7.76)
Equations (7.75) and (7.76) are solved together to find kρI and kρII .
For azimuthly symmetric modes n = 0, and the transcendental equation given by
Eq. (7.74) factors into two equations given by

µd J0′ (u) µc K0′ (υ )

=− (7.77a)
u J0 (u) υ K0 (υ )
and
εd J0′ (u) εc K0′ (υ )
=− . (7.77b)
u J0 (u) υ K0 (υ )
7-2 The Circular Dielectric Waveguide 425

Using Eq. (7.29c) it can be shown that

J0′ (u) = −J1 (u)

and
K0′ (υ ) = −K1 (υ ) .

Hence the two transcendental equations defined by Eqs. (7.77a) and (7.77b) can be written as

µd J1 (u) µc K1 (υ )
=− (7.78a)
u J0 (u) υ K0 (υ )
and
εd J1 (u) εc K1 (υ )
=− . (7.78b)
u J0 (u) υ K0 (υ )
Equation (7.77a) (or Eq. (7.78a)) can be obtained from Eqs. (7.69b) and (7.70a) by setting
n = 0. However, to make sure Eq. (7.70b) is also satisfied, A0 must be zero and, according to
Eq. (7.69a), C0 must also be zero. This will constitute a TE wave in the fiber. Also, Eq. (7.77b)
(or Eq. (7.78b)) can be obtained directly from Eqs. (7.69a) and (7.70b) by setting n = 0. To
ensure that Eq. (7.70a) is also satisfied, we must have B0 = D0 = 0. This corresponds to a TM
wave in the fiber.
To examine the behavior of the cutoff frequencies of hybrid modes in optical fibers, we
consider an optical fiber with nc = 1.44 and nd = 1.45, and then solve the transcendental
equation (Eq. (7.76)) by varying kc a and the normalized propagation constant β /kc for
different values of n, as shown in Fig. 7-8. The significance of the results shown in Fig. 7-8
pertains to the bandwidth over which a single mode operation can be maintained. For the
dominant mode corresponding to n = 1 from kc a ≈ 7 to kc a ≈ 21, the fiber can only support
the dominant mode. This indicates that the dominant mode, for a fixed fiber radius, has a
bandwidth of almost 1.5 octaves.

7-2.1 The Cylindrical Cavity

By short-circuiting the two ends of a cylindrical waveguide, we form a cylindrical cavity, as

shown in Fig. 7-9. As in the case of the rectangular cavity, the fields can be expressed in terms
of TE and TM waveguide modes. In this case, tangential electric fields at z = 0 and z = ℓ must
vanish. The ρ -component of the electric field for a TM mode can be written as

Eρ (ρ , φ , z) = −E0 β kρ Jn′ (kρ ρ ) cos(nφ ) sin(β z) . (7.79)

Of course, the choice of sine function ensures that Eρ (ρ , φ , 0) = 0. To force Eρ (ρ , φ , z), to go

to zero as z approaches ℓ, we need
qπ
sin β ℓ = 0 β= , q = 0, 1, 2, 3, . . .
ℓ
426 Chapter 7 Cylindrical Wave Functions and Their Applications

nd = 1.45, nc = 1.44
n=1
1.006 n=2
n=3
1.005
n=4
β 1.004

kc 1.003

1.002

1.001

1
0 10 20 30 40 50 60 70
kc a

Figure 7-8: The normalized propagation constant of hybrid modes inside an optical fiber as
a function of kc a for different mode numbers n. Here the index of refraction of the cladding is
nc = 1.44 and that of the core is nd = 1.45.

Dielectric-filled

Metallic
l cavity

Figure 7-9: Configuration of a cylindrical cavity consisting of a metallic cylinder containing a

dielectric material.
7-2 The Circular Dielectric Waveguide 427

The fields can be generated from a potential function

x   qπ 
TM np
ψnpq = Jn ρ cos(nφ ) cos z , (7.80)
a ℓ
and are given by

E0 β n  xnp   qπ 
Eφ (ρ , φ , z) = Jn ρ cos(nφ ) sin z (7.81a)
ρ a ℓ
and
 x 2 x   qπ 
np np
Ez (ρ , φ , z) = E0 Jn ρ cos(nφ ) cos z . (7.81b)
a a ℓ
According to Eq. (7.70a),
 qπ 2  x 2
np
= ω 2 µε .
+ (7.82)
ℓ a
Hence, the resonant frequency of the npq mode is given by

r
TM 1 xnp 2  qπ 2
fnpq = √ + , (7.83)
2π µε a ℓ

where n = 0, 1, 2, . . . , p = 1, 2, 3, . . . , and q = 0, 1, 2, 3, . . . . The dominant mode is TM010 ,

which has a constant electric field E = Ez ẑ. For a TE mode, the field can be generated from
potential  ′ 
xnp  qπ 
TE
ψnpq = Jn ρ cos(nφ ) sin z . (7.84)
a ℓ
The expressions for the electric field components are given by
 ′   qπ 
H0 iω µ n xnp
Eρ (ρ , φ , z) = − Jn ρ sin(nφ ) sin z (7.85a)
ρ a ℓ
and
 ′   ′   qπ 
xnp ′
xnp
Eφ (ρ , φ , z) = −H0 iω µ Jn ρ cos(nφ ) sin z . (7.85b)
a a ℓ
The resonant frequency for TE modes is given by

s
 2  qπ 2
TE 1 x′np
fnpq = √ + . (7.86)
2π µε a ℓ

Here, the index q = 1, 2, 3, . . . , and thus the dominant resonant frequency is TE111 .
The quality factor due to the walls’ ohmic loss can be computed in a manner similar to
how the attenuation rates for the circular waveguides were estimated. The quality factor for
428 Chapter 7 Cylindrical Wave Functions and Their Applications

the dominant mode TM010 (this mode is dominant for d < 2a) also requires calculation of the
stored energy in the cavity. The field expressions for TM010 are given by
 x 2 x 
01 01
Ez (ρ , φ , z) = E0 J0 ρ (7.87a)
a a
and
x  x  x  x 
01 01 01 01
Hφ (ρ , φ , z) = E0 iωε J0′ ρ = −E0 iωε J1 ρ . (7.87b)
a a a a
The stored energy is given by
Z ℓ Z a Z 2π
ε
W = 2We = |Ez |2 ρ d φ d ρ dz
2 0 0 0
 4 Z ah 
ε 2 x01 x01 i2
= |E0 | 2π ℓ J0 ρ ρ dρ
2 a 0 a
π ℓε (x01 )4
= |E0 |2 [J1 (x01 )]2 . (7.88)
2 a2
Because the fields tangential to the lower and upper surfaces of the cylinder (at z = 0 and
z = a) are zero, no ohmic losses occur in those two surfaces. The power loss on the cylinder’s
walls is computed from

1
pL = |Htan |2 ds
2σ δs S
1  x 2  Z ah 
x01 i2

01
= |E0 |2 ω 2 ε 2 2π aℓ [J1 (x01 )]2 + 4π J1 ρ ρ dρ
2σ δs a 0 a
 x 2
01
|E0 |2 ω 2 ε 2 π a(ℓ + a)
= a [J1 (x01 )]2 . (7.89)
σ δs
In deriving Eq. (7.89) we used the following integral identity for Bessel functions:
Z    
2 ρ2  ′ 2 ν2 2
[Rν (ρ )] ρ d ρ = Rν (ρ ) + 1 − 2 [Rν (ρ )] (7.90)
2 ρ
and
1
J1′ (ρ ) = J0 (ρ ) − J1 (ρ ) . (7.91)
ρ
The quality factor for TM010 can now be evaluated using
ωW
Q̆ = . (7.92)
pL
7-2 The Circular Dielectric Waveguide 429

Substituting the expressions for the stored energy from Eq. (7.88) and the power loss from
TM = x /a, in Eq. (7.92)
Eq. (7.89), and noting that the cavity resonant wave number is k010 01
r
µ
σ δs x01
ε  σ δs η 1.202
Q̆ =  a =  a . (7.93)
2 1+ 2 1+
ℓ ℓ

Example 7-2: Circular Patch Antenna

Using modal analysis and appropriate boundary conditions, find the operating frequency of a
circular patch antenna formed from a circular metallic plate atop an infinite ground plane, as
shown in Fig. 7-10.

Circular patch PMC

PEC ,ε
h ,ε ε
PEC
PMC Coaxial feed line

Figure 7-10: Configuration of a circular patch antenna fed by a coaxial line. The circular patch
is placed on top of a substrate of height h and a relative dielectric constant ε1 . The dotted lines
represent equivalent perfect magnetic conductor (PMC) walls.

Solution: In practice, patch antennas are fabricated using a metallic-coated low-loss dielectric
substrate with a relative permittivity ε1 . The metallic circular plate is attached to the center
pin of a coaxial line at an appropriate distance away from the center of the patch, in order to
achieve impedance match to the transmission line. The current from the feed flows over the
top circular patch in a manner that satisfies the boundary conditions. One boundary condition
imposed on the surface current is the one that forces the current normal to the edge of the patch
to be zero at the edge. This is not a strict requirement since there are fringing electric fields
from the metallic patch to the ground. These fringing electric fields produce a polarization
current whose flow is perpendicular to the edge that allow a small amount of conduction
current to flow perpendicular to the edge. As a first-order approximation we can ignore the
fringing fields around the patch, and in this case the surface current normal to the edge is
approximately zero. As a result, the tangential magnetic field near the patch edge is also
zero. This constitutes an equivalent PMC wall around the edge of the patch and is a good
approximation when the height h of the patch above the ground plane is much smaller than
a wavelength. As a result we can consider the circular patch as a circular cavity with the top
and bottom faces as PEC and the side wall of the cylinder as PMC.
430 Chapter 7 Cylindrical Wave Functions and Their Applications

The boundary conditions for TM and TE waves on PMC are given respectively by

∂ ψ TM ∂ ψ TM 
= =0 (7.94a)
∂n ∂ ρ ρ =a
and

ψ TE ρ =a = 0 . (7.94b)

Hence, the TM fields can be generated from a potential function given by

 ′   qπ 
TM
xnp
ψnpq = Jn ρ cos(nφ ) cos z , (7.95)
a h

where x′np represent the pth zero of Jn′ (x). The TE fields on the other hand can be generated
from a potential function given by
x   qπ 
TE np
ψnpq = Jn ρ cos(nφ ) sin z . (7.96)
a h
TM , given by
The dominant resonant mode (according to Table 7-2) corresponds to ψ110
 
TM 1.841
ψ110 = J1 ρ cos φ , (7.97)
a
and the resonant frequency is given by

1.841 1.841c 0.879 × 108

f TM110 = √ √ = √ = √ . (7.98)
2π a µ0 ε0 ε1 2π a ε1 a ε1

In practice, the resonant frequency is slightly lower than what is calculated from Eq. (7.98)
due to the fringing electric field.

7-2.2 Angular-Sector Waveguide

In this section, we consider the problem of wave propagation in an angular section of a
metallic waveguide as depicted in Fig. 7-11. The region of interest is confined within a
metallic pipe whose surface is defined by

ρ ≤a, 0 ≤ φ ≤ φ0 , and −∞ < z < ∞ .

Similar to the cylindrical waveguide, the scalar potential must satisfy the 2-D Helmholtz
equation given by Eq. (7.22), subject to Dirichlet’s and Neumann’s boundary conditions for
TM and TE modes. For this problem the region of interest is not the entire domain but it does
include ρ = 0. Hence, the admissible solution for radial functions include the Bessel functions
of the first kind and noninteger order given by

ψ (ρ , φ ) = (An sin νφ + Bn cos νφ ) Jν (kρ ρ ) . (7.99)

7-2 The Circular Dielectric Waveguide 431

x
ϕ0 z
a

Figure 7-11: Geometry of a metallic angular-sector waveguide.

For the TM case, ψ (ρ , 0) = ψ (ρ , φ0 ) = 0, which enforces

Bn = 0
and
nπ
ν= , n = 1, 2, 3, . . . (7.100)
φ0
The boundary condition also mandates that ψ (a, φ ) = 0, and this condition gives a set of
discrete values for kρ given by
xν p
knp = , p = 1, 2, 3, . . . ,
a
where xν p is the pth zero of Jν (x). Hence the eigenfunctions are given by
  x 
TM nπ νp
ψnp (ρ , φ ) = Anp sin φ Jnπ /φ0 ρ . (7.101)
φ0 a
Explicit expressions for the field quantities can be obtained from Eqs. (5.102a) and (5.102b).
For the TE case for which ∂∂ψn = 0 on the waveguide boundary, the admissible solution is
given by    ′ 
TE nπ xν p
ψnp = Bnp cos φ Jnπ /φ0 ρ , (7.102)
φ0 a
where xν′ p is the pth zero of Jν′ (x). The expressions for the field quantities can be obtained
from Eqs. (5.105a) and (5.105b).

7-2.3 Large Angular-Sector Waveguide

The solution obtained for the angular-sector waveguide can be used to find the possible
field solutions in an unbounded angular-sector waveguide with the configuration shown in
Fig. 7-12. The domain of interest is now specified by

0<ρ <∞ and 0 ≤ φ ≤ φ0 .

432 Chapter 7 Cylindrical Wave Functions and Their Applications

x
ϕ0 z

Figure 7-12: Geometry of an unbounded angular-sector waveguide (interior wedge).

Considering this problem as the limiting case of the previous problem as a → ∞, wave
functions similar to Eq. (7.101) can be assumed. However, in this case the domain supports an
infinite number of modes with a continuous eigenvalue. To demonstrate that, let us consider
the behavior of Jν (kρ a) as kρ a → ∞. Using the large-argument expansion of Jν (kρ a), it can
be shown that s  
2 (2ν + 1)π
lim Jν (kρ a) = cos kρ a − . (7.103)
kρ a→∞ π kρ a 4
The eigenvalues (modes) correspond to zeros of Jν (kρ a) = 0. These can be obtained
approximately from Eq. (7.103) and are given by
 
ν 1
xνn = kρn a = n + − π, n = 1, 2, 3, . . .
2 4
The separation between two successive modes is given by
π
∆ kρn = kρn+1 − kρn = ,
a
which goes to zero as a → ∞.
The solution for the scalar potential can be expressed in terms of the superposition of all
possible modes:
  
TM
XX nπ xν p 
ψ (ρ , φ ) = lim Anp sin φ Jν ρ . (7.104)
a→∞
n p
φ0 a

For the unbounded angular-sector waveguide, the summation over n can be replaced by an
integral:
X∞  Z ∞
TM nπ
ψ (ρ , φ ) = sin φ An (kρ ) Jν (kρ ρ ) dkρ . (7.105)
φ0 0
n=1

The function ψ (ρ , φ ) given by Eq. (7.105) satisfies both the wave equation and the boundary
condition, and hence represents a general solution of the problem. The solution for the TE
7-2 The Circular Dielectric Waveguide 433

case can easily be constructed from Eq. (7.102) and is given by

X∞  Z ∞
TE nπ
ψ (ρ , φ ) = cos φ Bn (kρ ) Jν (kρ ρ ) dkρ . (7.106)
φ0 0
n=0

7-2.4 Radial and Circumferential Waveguides

The transverse electric and magnetic field expansion developed for uniform waveguides can
also be applied to structures confined by two planar conductors perpendicular to the z-axis.
In such cases a standing wave forms along the z-axis and propagation can occur in radial
or circumferential directions. Since a tangential electric field (transverse electric field) must
vanish on planes perpendicular to the z-axis, the appropriate boundary condition for the
potentials can be obtained from
 
TM ∂ 
TM  ∂ ψ TM 
Et = ∇t ψ  =0 =0 (7.107a)
∂z r∈z−const. plane ∂ z r∈z−const. plane
and

TE 

ETE
t = iω µ ∇t ψ r∈z−const. plane
=0 ψ TE r∈z−const. plane = 0 . (7.107b)

Radial Waveguide
The configuration of a sectoral horn is depicted in Fig. 7-13(a). The sectoral horn is made up
of two metallic plates at φ = 0 and φ = α and two horizontal metallic plates at z = 0 and

z z

l
y y

x x

(a) (b)

Figure 7-13: (a) A sectoral horn waveguide formed by z = 0 and z = ℓ metallic plates on the
top and bottom and φ = 0 and φ = α metallic plates. (b) A radial waveguide with azimuthal
symmetry.
434 Chapter 7 Cylindrical Wave Functions and Their Applications

z = ℓ. The TM radial waves supported by the sectoral horn can be generated from potential
 nπ   qπ 
(1)
ψ TM (ρ , φ , z) = Hnπ /α (kρ ρ ) sin φ cos z , n = 1, 2, 3, . . . , q = 0, 1, 2, . . .
α ℓ
(7.108)
as this potential satisfies the boundary conditions

ψ TM (ρ , 0, z) = ψ (ρ , α , z) = 0 (7.109a)
and
 
∂ ψ TM  ∂ ψ TM 
= =0. (7.109b)
∂ z ρ ,φ ,0 ∂ z ρ ,φ ,ℓ
p p
In Eq. (7.108), kρ = k2 − β 2 = k2 − (qπ /ℓ)2 . The choice of Hankel function of the
first kind indicates just an outward-traveling wave. In practice, for a sectoral horn of finite
dimensions between ρ = ρ1 and ρ = ρ2 , the modal solution may be written as

TM
   nπ   qπ 
ψnq (ρ , φ , z) = Anq Jnπ /α (kρ ρ ) + Bnq Nnπ /α (kρ ρ ) sin φ cos z . (7.110)
α ℓ
The unknown coefficients depend on the field distributions at ρ = ρ1 and ρ = ρ2 .
Also, for a TE wave, the following potential can be used:
 nπ   qπ 
TE (1)
ψnq = Hnπ /α (kρ ρ ) cos φ sin z , n = 0, 1, 2, . . . , q = 1, 2, . . . (7.111)
α ℓ
Note that for TM waves the lowest-order mode corresponds to q = 0, for which kρ = k0 ,
and thus there is no cutoff frequency for TM waves. On the other hand, for TE waves the
lowest-order mode corresponds to q = 1, for which the cutoff wave number is kcTE1 = π /ℓ.
A particular type of radial waveguide with azimuthal symmetry is shown in Fig. 7-13(b).
For this radial waveguide the lowest-order mode is TM0 mode, whose fields resemble that of
an infinite line source in free space:
 qπ 
(1)
ψ TM0q (ρ , φ , z) = H0 (kρ ρ ) cos z , (7.112a)
ℓ
where kρ = [k2 − (qπ /ρ )2 ]1/2 . For n = 0, the electric and magnetic fields are given by
(1)′
EzTM0 (ρ , φ , z) = k2 H0 (kρ ) , (7.112b)
and
r
(1)′ ε (1)
HφTM0 = iωε k H0 (kρ ) = −ik 2
H (kρ ) . (7.112c)
µ 1
7-2 The Circular Dielectric Waveguide 435

0.8
Real part
Imaginary part
0.6

0.4
Z0
η
0.2

−0.2

−0.4
0 2 4 6 8 10 12 14 16 18 20
kρ

Figure 7-14: The complex characteristic impedance of a TM0 mode in a radial waveguide as a
function of radial distance (kρ ).

ρ-direction, whose characteristic impedance is

In fact, this is a TEM wave propagating in the ρ̂
a function of ρ :
(1)
EzTM0 η H0 (kρ )
Z0 = = . (7.113)
−HφTM0 (1)
i H1 (kρ )
It can be shown that when kρ ≫ 1, the waveguide characteristic impedance is

lim Z0 = η .
kρ →∞

Figure 7-14 shows the complex characteristic impedance of the radial waveguide as a function
of kρ .

Circumferential Waveguide
The configuration for a circumferential waveguide is shown in Fig. 7-15. In this case,
φ. The geometry shown in this figure is formed by curved metallic
propagation is along φ̂
walls at ρ = ρ1 and ρ = ρ2 , and metallic plates at z = 0 and z = b. By setting ρ2 − ρ1 = a, the
configuration becomes an H-bent rectangular waveguide for a TE10 mode in a rectangular
436 Chapter 7 Cylindrical Wave Functions and Their Applications

Figure 7-15: Configuration of a circumferential waveguide formed by metallic plates at z = 0,

z = b, ρ = ρ1 , and ρ = ρ2 .

waveguide with a > b. The appropriate potential for a TM0 wave in this circumferential
waveguide can be written as

ψ TM0 = [Jν (kρ ) Nν (kρ1 ) − Jν (kρ1 ) Nν (kρ )] (Aeiνφ + B−iνφ ) . (7.114)

The choice of radial function is made to ensure that


ψ (ρ , φ , z)ρ =ρ1 = 0 , (7.115)

and coefficients A and B are to be determined depending on the input and output field
conditions at the ends of the bent. The boundary condition at ρ = ρ2 is satisfied when

F = Jν (kρ2 ) Nν (kρ1 ) − Jν (kρ1 ) Nν (kρ2 ) = 0 . (7.116)

This provides an equation for the angular propagation constant ν . Figure 7-16 shows the plot
of a function F representing the left-hand side of Eq. (7.116) as a function of ν for two cases,
one where ρ1 = λ and another where ρ1 = 1.1λ , ρ2 − ρ1 = 0.762λ in both cases. The zeros
of Eq. (7.116) are found to be ν = 6.53 and ν = 7 for ρ1 = λ and ρ1 = 1.1λ , respectively.

7-3 Green’s Functions Solutions for Some Canonical Problems

The cylindrical wave functions can be used to construct Green’s function for waveguides and
scattering problems. In what follows we will examine the solution for Green’s functions for
some canonical problems.
7-3 Green’s Functions Solutions for Some Canonical Problems 437

0.1

k 1= 2.2
0.05
k 1= 2
k 1= 2.2 7

F 6.53

−0.05
k 1= 2

−0.1

−0.15
0 1 2 3 4 5 6 7 8

Figure 7-16: The transcendental function F given by Eq. (7.116) as a function of ν . The zero
crossing of the function indicates the solution of Eq. (7.116). The two graphs correspond to ρ1 = λ
and ρ1 = 1.1λ , with ρ2 − ρ1 = 0.762λ in both cases.

7-3.1 Green’s Function for an Angular-Sector Waveguide

Consider a filament-line current along the z-axis located at ρ = ρ ′ within an angular-sector
waveguide, as shown in Fig. 7-17. The current distribution is J = δ (ρ ρ′ ) eiβ z ẑ, and it excites
ρ −ρ
a TM wave. The Green’s function for the problem can be constructed from the eigenvalues
and eigenfunction of the waveguide, and is given by
XX ψnp (ρ ′ ) ψnp (ρ )
g(ρ ρ′ ) =
ρ ,ρ  ,
2 − k2 )
(knp 2 ρ dρ dφ
ψnp
m n ρ

where kρ2 = k2 − β 2 . Noting that

Z φ0  
2 nπ φ0
sin φ dφ = (7.117a)
0 φ0 2
and
438 Chapter 7 Cylindrical Wave Functions and Their Applications

z-directed
electric
current

ρ′

Figure 7-17: Configuration of a line source in an angular-sector waveguide. The current

ρ − ρ ′ ) eiβ z ẑ.
distribution is J = δ (ρ

Z a  xν  a2 2
ρ Jν2 p
ρ dρ = J (xν ) , (7.117b)
0 a 2 ν +1 p
the Green’s function takes the following form:
     xν   xν 
nπ nπ ′
X∞ X ∞ sin φ sin φ Jmπ /φ0
p
ρ Jmπ /φ0
p
ρ′
4 φ 0 φ0 a a
g(ρ ρ′ ) = 2
ρ ,ρ 2 − k2 ) J 2
, (7.118)
a φ0 (knp ρ (mn/φ0 )+1 (xν p )
n=1 p=1

where
xν p nπ
knp = for ν = .
a φ0
Also, noting that s
2
lim Jν +1 (xν p ) = (−1)n ,
xν p →∞ π xν p
the Green’s function for the unbounded angular sector can be obtained from
∞    Z ∞
′ 2 X nπ nπ ′ α
g(ρρ,ρ
ρ)= sin φ sin φ 2
J
2 nπ /φ0
(αρ ) Jnπ /φ0 (αρ ′ ) d α .
φ0 φ0 φ0 0 α − kρ
n=1
(7.119)
The Green’s function for a circular waveguide with a vane at φ = 0 (see Fig. 7-18) can now
easily be obtained from Eq. (7.118) by letting φ0 = 2π and is given by
n  m  x  x 
′ np np ′
2 X∞ X ∞ sin φ sin φ Jn/2 ρ J m/2 ρ
g(ρ ρ′ ) = 2
ρ ,ρ 2 2 a a , (7.120)
a π (k 2 − k2 ) J 2 (x )
n=1 p=1 np ρ (n+3)/2 np

where xn is the nth zero of Jm/2 (x).

7-3 Green’s Functions Solutions for Some Canonical Problems 439

z-directed
electric r
current ϕ′
r′ ϕ

Figure 7-18: A circular waveguide with a metallic vane at φ = 0 and a line source at ρ ′ .

7-3.2 Formal Solution for Green’s Function of an Angular Sector

In the previous section, the solution for Green’s function for the angular-sector waveguide
was obtained from the formulation based on eigenfunctions. In this section the expression for
the Green’s function will be derived directly from the solution of the scalar wave function in
response to a delta function. The geometry of the problem is shown in Fig. 7-17. The Green’s
function must satisfy
   
1 ∂ ∂ 1 ∂2 2 ′ δ (ρ − ρ ′) δ (φ − φ ′ )
ρ + 2 + k ρ ρ
g(ρ ,ρ
ρ ) = − , (7.121)
ρ ∂ρ ∂ρ ρ ∂φ2 ρ

ρ,ρ
and for the TM case, g(ρ ρ′ ) must vanish on the waveguide surface (φ = 0, φ = φ0 , and
ρ = a). Using the method of separation of variables, we get
X
ρ′ ) =
ρ ,ρ
g(ρ g(ρ , ρ ′ ) Fn (φ , φ ′ ) ,
n

and we should note that

∞    
′ 2 X nπ ′ nπ
δ (φ − φ ) = sin φ sin φ . (7.122)
φ0 φ0 φ0
n=1

The appropriate solution for Fn (φ , φ ′ ) is similar to the one for δ (φ − φ ′ ). Hence

"    2 #
1 ∂ ∂ nπ 1 2 ′ δ (ρ − ρ ′)
ρ − + kρ g(ρ , ρ ) = − . (7.123)
ρ ∂ρ ∂ρ φ0 ρ2 ρ

Direct observation of Eq. (7.123) indicates that g(ρ , ρ ′ ) must be continuous at ρ = ρ ′ , but its
derivative with respect to ρ at ρ ′ is discontinuous and the level of discontinuity is −1/ρ ′ . This
is the reason for having a delta function on the right-hand side of Eq. (7.123). Let us denote
the solution for g(ρ , ρ ′ ) where ρ < ρ ′ by g< (ρ , ρ ′ ), and that for g(ρ , ρ ′ ) where ρ > ρ ′ by
440 Chapter 7 Cylindrical Wave Functions and Their Applications

g> (ρ , ρ ′ ). For region ρ < ρ ′ , g< must be regular. Hence

g< = A Jν (kρ ρ ) , (7.124)

but g> can be arbitrary in the region ρ > ρ ′ . Hence

(1)
g> = B Jν (kρ ρ ) +C Hν (kρ ρ ) , (7.125)

where as before ν = nπ /φ0 . Continuity of g at ρ = ρ ′ requires

g< = g> . (7.126)

Also,
∂ g> ∂ g< 1
− =− ′ . (7.127)
∂ρ ∂ρ ρ
The boundary condition at ρ = a enforces
(1)
g> (a) = B Jν (kρ a) +C Hν (kρ a) = 0 . (7.128)

From Eq. (7.126) and Eq. (7.127),


(A − B) Jν (kρ ρ ′ ) −C Hν(1) (kρ ρ ′ ) = 0 ,
(1) ′ 1 (7.129)
(A − B) Jν′ (kρ ρ ′ ) −C Hν (kρ ρ ′ ) = .
kρ ρ ′

Solving Eq. (7.129), we have

−Jν (kρ ρ ′ )
C= h i
(1) ′ (1)
kρ ρ ′ Jν (kρ ρ ′ ) Hν (kρ ρ ′ ) − Jν′ (kρ ρ ′ ) Hν (kρ ρ ′ )
and
(1)
−Hν (kρ ρ ′ )
A−B = h i.
(1) ′ (1)
kρ ρ ′ Jν (kρ ρ ′ ) Hν (kρ ρ ′ ) − Jν′ (kρ ρ ′ ) Hν (kρ ρ ′ )

Noting the Wronskian relationship

h i 2i
(1) (1) ′ (1)
W Jν (X ), Hν (X ) = Jν (X ) Hν (X ) − Jν′ (X ) Hν (X ) = (7.130)
πX
and the boundary condition given by Eq. (7.128), we can easily show that
π
C=− Jν (kρ ρ ′ ) ,
2i
(1)
iπ Jν (kρ ρ ′ ) Hν (kρ a)
B=− ,
2 Jν (kρ a)
7-3 Green’s Functions Solutions for Some Canonical Problems 441

and
(1)
!
πi (1) Jν (kρ ρ ′ ) Hν (kρ a)
A= Hν (kρ ρ ′ ) − .
2 Jν (kρ a)

Consequently,
∞ (1)
!
′ πi X (1) ′ Jν (kρ ρ ′ ) Hν (kρ a)
g< (ρ
ρ ,ρ
ρ)= Jν (kρ ρ ) Hν (kρ ρ ) − sin νφ sin νφ ′ (7.131a)
φ0 Jν (kρ a)
n=1
and
∞ (1)
!
πi X (1) Jν (kρ ρ ) Hν (kρ a)
g> (ρ ρ′ ) =
ρ ,ρ Jν (kρ ρ ′ ) Hν (kρ ρ ) − sin νφ sin νφ ′ , (7.131b)
φ0 Jν (kρ a)
n=1

where, as before, ν = nπ /φ0 .

By comparing Eqs. (7.131a) and (7.131b) with Eq. (7.118), it becomes obvious that the
expressions obtained from the latter procedure are much easier to compute.

7-3.3 Two-Dimensional Green’s Function for an Unbounded Homogeneous

Medium
In this section we consider the problem of field radiation from an infinitely long time-
harmonic filament current. Consider an infinitely long electric current Ieiβ z located at the
origin along the z-axis, as shown in Fig. 7-19. The current distribution is given by

J(r) = Ieiβ z δ (ρ
ρ) ẑ . (7.132)

This current produces only the z-component of the Hertz potential, which satisfies
−i iβ z
∇2 Π z + k 2 Π z = Ie δ (ρ
ρ) . (7.133)
ωε

Ieiβz
y
ϕ
x ρ

Figure 7-19: An infinite current filament along the z-axis carrying current Ieiβ z and an
observation point in the x–y plane indicated by ρ .
442 Chapter 7 Cylindrical Wave Functions and Their Applications

As previously stated, since the z-dependence of the current is of the form eiβ z , Π(ρ , φ , z) can
be written as
Πz (ρ , φ , z) = Ψ(ρ , φ ) eiβ z , (7.134)
and Ψ(ρ , φ ) is a solution to the following 2-D wave equation:
−i
∇2t Ψ(ρ , φ ) + kρ2 Ψ(ρ , φ ) = I δ (ρ
ρ) , (7.135)
ωε
where kρ2 = k2 − β 2 .
Using the method of separation of variables, ψ can be expressed in terms of the product
of two functions dependent on only φ and ρ , respectively. Because of the symmetry of the
problem, the solution is independent of φ . The dependence on ρ is of the form of Bessel
functions, and since the source is at the origin, an outgoing waveform must be chosen. An
admissible solution is then of the following form:
(1)
Ψ(ρ , φ ) = A H0 (kρ ρ ) .

Since the source is at the origin, a singular behavior of the field at ρ = 0 is expected. From
Eq. (5.102b), the magnetic field is obtained from
(1) ′
H(ρ , φ ) = −iωε kρ A H0 (kρ ρ ) ρ̂
ρ × ẑ
(1) ′
= iωε kρ A H0 (kρ ρ ) eiβ z φ̂
φ.

According to Ampère’s law,

I
lim Hφ (ρ , φ ) ρ d φ = I , (7.136)
ρ →0 C

because Z ρ Z 2π
lim iωε E · ẑ ρ d ρ d φ = 0 .
ρ →0 0 0
Using the small-argument expansion of the Hankel function, we have
 
(1) i2 2
lim H0 (X ) ≈ − ln ,
X→0 π γX
where γ = 1.78107. As a result it can be easily shown that

(1) ′ i2 1
H0 (kρ ρ ) ≈ . (7.137)
π kρ ρ

Using Eq. (7.137) in Eq. (7.136), we have

Z 2π
i2 1
iωε kρ A ρ d φ = −4ωε A = I ,
0 π kρ ρ
7-3 Green’s Functions Solutions for Some Canonical Problems 443

from which A is found to be

−I
A=.
4ωε
Hence the magnetic field produced by the filament current is given by
−ikρ I (1) ′
H(ρ , φ ) = H0 (kρ ρ ) eiβ z φ̂
φ. (7.138)
4
Also, the electric field is found from Eq. (5.102a) to be

−kρ2 I (1) iI β kρ (1) ′

E(ρ , φ ) = H0 (kρ ρ ) eiβ z ẑ − H (kρ ρ ) ρ̂
ρ. (7.139)
4ωε 4ωε 0
Now, positioning the current filament at ρ ′ , we have
−I (1)
Πz (ρ ρ′ ) =
ρ,ρ ρ − ρ ′ |) eiβ z ,
H (kρ |ρ (7.140)
4ωε 0
which can be used to find the Hertz potential for an arbitrary current distribution

ρ) eiβ z ẑ .
ρ, z) = J(ρ
J(ρ

The Hertz potential for this current distribution is given by

−eiβ z
"
Πz (ρ ) = ρ′ ) H0(1) (kρ |ρ
J(ρ ρ − ρ ′ |) ρ ′ d ρ ′ d φ ′ , (7.141)
4ωε S

where S specifies the region for the current distribution. The electric field is given by
" h i 
−kρ iβ z (1) ′ (1) ′ ′ ′ ′ ′ ′
E(ρ , φ ) = e kρ H0 (kρ |ρ − |) ẑ + iβ H0 (kρ |ρ − |) ρ̂ J(ρ ) ρ d ρ d φ ,
ρ ρ ρ ρ ρ ρ
4ωε S
p (7.142)
where |ρ ρ − ρ ′ | = (x − x′ )2 − (y − y′ )2 in the Cartesian coordinate system.
Also, the magnetic field can be computed from
ρ − ρ′) ′ ′ ′
"
−ikρ ′ ẑ × (ρ
H(ρ , φ ) = ρ′ ) H0(1) (kρ |ρ
J(ρ ρ − ρ ′ |) ρ dρ dφ . (7.143)
4 S ρ − ρ′|
|ρ

Note here that φ̂ φ = ẑ × ρ̂

ρ, as used in Eq. (7.138), depends on the origin of the coordinate
system, and thus in general its expanded form must be used. That is, once the origin is
translated to ρ ′ , φ̂
φ is changed accordingly. To track this change, φ̂
φ is first written as
ρ
φ = ẑ × ρ̂
φ̂ ρ = ẑ × ,
|ρ
ρ|

and then ρ is translated.

444 Chapter 7 Cylindrical Wave Functions and Their Applications

The far-field expressions for Eqs. (7.138) and (7.139) are given by
s
1
Hφ (ρ , φ ) ≈ kρ I ei(kρ ρ −π /4) eiβ z , (7.144a)
8π kρ ρ
s
kρ2 1
Ez (ρ , φ ) ≈ −η I ei(kρ ρ −π /4) eiβ z , (7.144b)
k 8π kρ ρ
and
s
kρ β 1
Eρ (ρ , φ ) ≈ η I ei(kρ ρ −π /4) eiβ z , (7.144c)
k 8π kρ ρ

which show that the equiphase and equiamplitude surfaces are cylinders (constant ρ ). Also,
in the far-field r
|E| µ
=η = ,
|Hφ | ε
and the fields are TEM.
The total power per unit length radiated by the filament current can be computed by
integrating the Poynting vector over the surface of a cylinder of height 1 m:
Z Z
1 2π 1
Prad = Re (E × H ∗) · ρ̂
ρρ d φ dz
2 0 0
( )
1 iπ kρ3 |I|2 h ′ i∗
(1) (1)
= Re ρ H0 (kρ ρ ) H0 (kρ ρ ) . (7.145)
2 8ωε

Noting that
(1) (1) ′ (1) (1)
H0 (kρ ρ ) [H0 (kρ ρ )]∗ = H0 (kρ ρ ) J0′ (kρ ρ ) − iH0 (kρ ρ ) N0′ (kρ ρ )
= J0 (kρ ρ ) J0′ (kρ ρ ) + N0 (kρ ρ ) N0′ (kρ ρ )
 
− i J0 (kρ ρ ) N0′ (kρ ρ ) − J0′ (kρ ρ ) N0 (kρ ρ ) ,

and using the Wronskian relation Jν (x) Nν′ (x) − Jν′ (x) Nν (x) = π2x , the radiated power per unit
length is given by
kρ2
Prad = |I|2 W/m. (7.146)
8ωε
It is interesting to note that the time-average radiated power is independent of distance from
the source and proportional to the square of the current.
7-3 Green’s Functions Solutions for Some Canonical Problems 445

y y

−I +I
x x
−I +I
−I +I

Figure 7-20: (a) The configuration of a two-dimensional dipole (line currents flowing in
opposite directions) and (b) the configuration of a two-dimensional quadrupole. Here δ , δ1 , and
δ2 are much smaller than a wavelength.

7-3.4 Multipole Representation of Cylindrical Wave Functions

In this section, it is shown that an nth-order cylindrical wave function of the following form:
(
(1) cos nφ even,
ψ (ρ , φ ) = Hn (kρ ρ ) (7.147)
sin nφ odd,

represents the Hertz vector potential of an ordered 2n-element array of closely spaced
z-directed filament currents of infinite extent flowing in the opposite direction of the adjacent
elements. It was shown in the previous section that the Hertz vector potential of an infinite
electric current filament is of the form of (see Eq. (7.134))
−I (1)
Π1z (ρ , φ , z) = H (kρ ρ ) eiβ z . (7.148)
4ωε 0
Now let us consider two closely spaced infinite current filaments carrying currents in opposite
directions, as shown in Fig. 7-20(a). The separation between the two current lines is denoted
by δ , which is much smaller than a wavelength. Obviously, as δ → 0, the fields emitted from
these two currents cancel each other. To ensure a nonzero field, the magnitude of current I can
be increased so that I δ is kept constant as δ becomes smaller. This configuration of currents
is called a two-dimensional dipole. Using superposition, the resulting Hertz vector potential
can be obtained from
   
2x 1 δ 1 δ
Πz = Πz x − , y − Πz x + , y . (7.149)
2 2
446 Chapter 7 Cylindrical Wave Functions and Their Applications

In the limit as δ → 0, Eq. (7.149) can be written as

∂ Π1z I δ iβ z ∂ (1)
Π2z x = −δ = e [H (kρ ρ )]
∂x 4ωε ∂x 0
kρ I δ iβ z (1)
= e H1 (kρ ρ ) cos φ . (7.150)
4ωε
As shown above, the Hertz vector potential of this dipole is entirely expressed by an even
cylindrical wave function of order n = 1. It can easily be shown that if the two current lines
are placed on the y-axis, the Hertz vector potential can be expressed by an odd cylindrical
wave function of order n = 1. Now let us consider the current configuration of Fig. 7-20(b).
This current configuration is know as a two-dimensional quadrupole. Using superposition,
and in the limit as δ1 and δ2 approach zero, the Hertz vector potential for the quadrupole is
given by
4 ∂ 2 Π1z ∂ Π2z x
Πz x,y ≈ δ1 δ2 = − δ2 . (7.151)
∂x ∂y ∂y
Using Eq. (7.150) in Eq. (7.151),

4 −Ikρ δ1 δ2 iβ z ∂ (1)
Πz x,y ≈ e [H (kρ ρ ) cos φ ] . (7.152)
4ωε ∂y 1
Carrying out the differentiation of Eq. (7.152) reduces to

4 kρ2 I δ1 δ2 iβ z (1)
Πz x,y ≈ e H2 (kρ ρ ) sin 2φ . (7.153)
4ωε
Equation (7.153) shows that the Hertz vector potential of a quadrupole line source is a
cylindrical wave function of order n = 2.
In general, a 2n-pole line source is defined as 2n current lines equally spaced around the
surface of a very small cylinder. It can be shown that the Hertz vector potential of the 2n-pole
line source is a cylindrical wave function of order n. Of course, duality can be applied to find
the magnetic Hertz potential of 2n-pole magnetic line sources.

7-3.5 Fourier Representation of the Two-Dimensional Green’s Function

The fields generated by a filament current can be represented in terms of a continuous
spectrum of plane waves. This representation can be obtained by expressing the scalar
potential function ψ (x, y) in Eq. (7.135) in terms of its Fourier transformation with respect
to variable x: Z ∞
1
ψ (x, y) = e (kx , y) eikx x dkx .
ψ (7.154)
2π −∞
Also, noting that Z ∞
1
δ (x) = eikx x dkx , (7.155)
2π −∞
7-3 Green’s Functions Solutions for Some Canonical Problems 447

Eq. (7.135) can then be written as

Z ∞    Z ∞
1 2 d2 2 1 −iI
−kx + 2 ψ e (kx , y) eikx x dkx =
e (kx , y) + kρ ψ δ (y) eikx x dkx .
2π −∞ dy 2π −∞ ωε
(7.156)
This equality holds if the integrands on both sides of Eq. (7.156) are equal; that is,
 2 
d 2 2 iI
2
+ (kρ − kx ) ψ e (kx , y) = − δ (y) . (7.157)
dy ωε
This is an ordinary second-order differential equation whose solutions are a harmonic function
of the following form:
(
A(kx ) eiky y y>0,
e (kx , y) =
ψ −ik y
(7.158)
B(kx ) e y y<0,

which represent waves traveling outward away from the source. Here
q
ky = kρ2 − kx2 ,

e (kx , y) and
and coefficients A(kx ) and B(kx ) are to be found by enforcing the continuity of ψ
the jump discontinuity of its derivative at y = 0. That is,

e (kx , 0+ ) = ψ
ψ e (kx , 0− ) (7.159a)
and
e (kx , 0+ ) ψ
ψ e (kx , 0− ) −iI
− = . (7.159b)
dy dy ωε
Solving Eq. (7.159) renders
−I
A=B= . (7.160)
2ky ωε
Substituting this solution into Eq. (7.154) provides the Fourier representation of the potential
function: Z ∞
−I 1
ψ (x, y) = q ei(kx x+ky |y|) dkx . (7.161)
4πωε −∞ k2 − k2
ρ x

The integrand represents a plane wave whose direction of propagation in the x-y plane is along
(kx /kρ ) x̂ + (ky /kρ ) ŷ. In Section 7-3.3 it was found that

−I (1)
p
ψ (x, y) = H0 (kρ x2 + y2 ) . (7.162)
2ωε
448 Chapter 7 Cylindrical Wave Functions and Their Applications

Comparing Eqs. (7.161) and (7.162), the following representation for this Hankel function is
obtained: Z ∞ √2 2
(1)
p 1 1
2
H0 (kρ x + y ) = 2 q ei(kx x+ kρ −kx |y|) dkx . (7.163)
π −∞ k2 − k2
ρ x

7-3.6 Series Solution for the 2-D Green’s Function of a Homogeneous Medium

In this section a series solution for the Green’s function for a homogeneous medium is
obtained in terms of cylindrical wave functions, following the procedure outlined for the
formal solution of Green’s function for the angular-sector waveguide. Considering the 2-D
scalar wave Eq. (7.121) and using the method of separation of variables, we have
∞
1 X ′
ρ,ρ
g(ρ ρ)= ′
g(ρ , ρ ′ ) eim(φ −φ ) . (7.164)
2π m=−∞

Note that in this case the entire φ domain is included in the region of interest. Denoting
the solution for g(ρ , ρ ′ ) by g< and g> for when ρ ≤ ρ ′ and ρ ≥ ρ ′ , respectively, similar to
Eq. (7.124),
g< = A Jm (kρ ρ ) .
The solution for g> must be chosen so that only outgoing wave functions are included:
(1)
g> = C Hm (kρ ρ ) .

Requiring the continuity condition for g(ρ , ρ ′ ) at ρ = ρ ′ (similar to Eq. (7.126)) and the jump
discontinuity of derivation of g with respect to ρ (similar to Eq. (7.127)), it can be easily
shown that
π i (1)
A= Hm (kρ ρ ′ )
2
and
πi
C= Jm (kρ ρ ′ ) .
2
Hence
+∞
i X (1) ′
g< (ρ ρ′ ) =
ρ,ρ Hm (kρ ρ ′ ) Jm (kρ ρ ) eim(φ −φ ) (7.165a)
4 m=−∞
and
+∞
i X (1) ′
g> (ρ ρ′ ) =
ρ,ρ Jm (kρ ρ ′ ) Hm (kρ ρ ) eim(φ −φ ) . (7.165b)
4 m=−∞
7-3 Green’s Functions Solutions for Some Canonical Problems 449

Comparing Eq. (7.165b) with Eq. (7.140), it is now obvious that

 +∞

 X (1) ′


 Hm (kρ ρ ′ ) Jm (kρ ρ ) eim(φ −φ ) ρ ≤ ρ′ ,
(1)
H0 (kρ |ρρ − ρ ′ |) = m=−∞
+∞ (7.166)

 X (1) ′

 Jm (kρ ρ ′ ) Hm (kρ ρ ) eim(φ −φ ) ρ≥ ρ′ .

m=−∞

This result is known as the addition theorem for the Hankel function of the first kind and
zeroth order. A similar result can be obtained for the Hankel function of the second kind
(1)
and zeroth order by simply conjugating Eq. (7.166). Adding the addition theorem for H0
(2)
and H0 , we get the addition theorem for Bessel functions of the first kind and zeroth order:
+∞
X ′
J0 (kρ |ρ ′
ρ − ρ |) = Jm (kρ ρ ′ ) Jm (kρ ρ ) eim(φ −φ ) . (7.167)
m=−∞

7-3.7 2-D Green’s Function for a Homogeneous Medium in the Presence of a

Metallic Cylinder
The field solution for arbitrary sources present in a homogeneous medium next to a metallic
cylinder of radius a with its center placed at the origin of a Cartesian system can easily be
obtained if the Green’s function for the problem is known. Figure 7-21 shows the geometry
of the problem where an infinite filament current Ieiβ z is placed in parallel with the axis of

ρ′
z-directed
electric
current

Figure 7-21: The geometry for an infinitely long metallic cylinder and an infinitely long filament
current positioned at ρ ′ and directed along ẑ.
450 Chapter 7 Cylindrical Wave Functions and Their Applications

the metallic cylinder at location ρ ′ with cylindrical coordinates (ρ ′ , φ ′ ). In the absence of the
metallic cylinder, the current filament produces the potential ψi (ρ , φ ) given by
−I (1)
ψi (ρ , φ ) = ρ − ρ ′ |) eiβ z .
H (kρ |ρ (7.168)
4εω 0
This incident wave scatters off the metallic cylinder and produces the scattered potential
ψs (ρ , φ ) in such a way as to satisfy the boundary condition

ψi (a, φ ) + ψs (a, φ ) = 0 . (7.169)

The scattered potential can be expressed in terms of cylindrical elementary wave functions as
+∞
X (1) ′
ψs (ρ , φ ) = Am Hm (kρ ρ ) eim(φ −φ ) eiβ z . (7.170)
m=−∞

The choice of Hankel function of the first kind is necessitated by the fact that the scattered
potential is an outward-traveling wave. According to Eq. (7.166),
+∞
−I X (1) ′
ψi (a, φ ) = Hm (kρ ρ ) Jm (kρ a) eim(φ −φ ) eiβ z . (7.171)
4εω m=−∞

Using the boundary condition defined by Eq. (7.169), we have

+∞ 
X 
(1) I (1) ′
Am Hm (kρ a) − Hm (kρ ρ ′ ) Jm (kρ a) eim(φ −φ ) =0,
m=−∞
4εω

from which am can be found and is given by

(1)
I Jm (kρ a) Hm (kρ ρ ′ )
Am = (1)
. (7.172)
4εω Hm (kρ a)

Hence the total potential ψt (ρ , φ ) can be represented by

ψt (ρ , φ ) =
 !
+∞


 Ieiβ z X Jm (kρ a) (1) (1) ′

 (1)
Hm (kρ ρ ) − Jm (kρ ρ ) Hm (kρ ρ ′ ) eim(φ −φ ) ρ ≤ ρ′ ,
 4εω
m=−∞ Hm (kρ a)
!

 Ieiβ z X +∞
Jm (kρ a)

 (1) (1)
Hm (kρ ρ ′ ) − Jm (kρ ρ ′ ) Hm (kρ ρ ) eim(φ −φ )
′
ρ ≥ ρ′ .

 4εω (1)
m=−∞ Hm (kρ a)
(7.173)

Note that the reciprocity principle holds. That is, if the observation point (ρ , φ ) and the source
point (ρ ′ , φ ′ ) are interchanged, the same expression is obtained.
7-3 Green’s Functions Solutions for Some Canonical Problems 451

' = 0.75
105° 90° 75° '=
a = 0.5λ 120° 60°
' = 10.5
135° 45°

150° 30°

165° 15°
10
0
−10
180° 0°

−165° −15°

−150° −30°

−135° −45°

−120° −60°
−105° −90° −75°

Figure 7-22: Normalized total z-component of the electric field for a configuration where the
line source is at ρ ′ and is on the x-axis (φ ′ = 0) at different locations: ρ ′ = 0.75λ , 1λ , and 10.5λ ,
in the presence of a metallic cylinder of radius a = 0.5λ .

The expressions for the field quantities can be obtained from Eqs. (5.102a) and (5.102b).
Figure 7-22 shows the normalized total z-component of the electric field for a problem where
the line source is at ρ ′ and is on the x-axis (φ ′ = 0) at different locations ρ ′ = 0.75λ , 1λ , and
10.5λ , all in the presence of a metallic cylinder with radius a = 0.5λ . The observation point
is in the far-field region. For ρ ′ = 0.75λ , the source is λ /4 away from the cylinder surface
and, as expected, the incident field and the scattered field along the x-axis add up coherently.
Also, at φ = ±180◦ , the incident field and the scattered field cancel each other due to the
shadowing effect. For the two other cases, the incident and scattered fields are out of phase
along the x-axis (φ = 0) and interfere destructively.
452 Chapter 7 Cylindrical Wave Functions and Their Applications

θi

k̂i Ei

Figure 7-23: Geometry of a metallic circular cylinder illuminated by a plane wave.

7-4 Scattering from a Metallic Circular Cylinder

Consider a circular metallic cylinder of radius a whose axis coincides with the z-axis of the
coordinate system, as shown in Fig. 7-23. The cylinder is illuminated by a plane wave whose
propagation vector k̂i lies in the x–z plane with

k̂i · x̂ = − sin θi (7.174a)

and
k̂i · ẑ = cos θi . (7.174b)

Initially, we consider TM polarization for the incidence field, i.e.,

Ei = E0 (cos θi x̂ + sin θi ẑ)eik0 (− sin θi x+cos θi z) . (7.175)

The incident field is the total field in the medium in the absence of the metallic cylinder. In
the presence of the cylinder the incident field excites a current on the surface of the metallic
cylinder, which in turn reradiates elecromagnetic wave. This reradiated field is known as the
7-4 Scattering from a Metallic Circular Cylinder 453

scattered field, which together with the incident field constitutes the total field in the medium

Et (ρ
ρ, z) = Ei (ρ
ρ, z) + Es (ρ
ρ, z) . (7.176)

The boundary condition on the surface of the cylinder mandates that

ρ̂ × Et(aρ̂ , z) = 0 . (7.177)

The incident and scattered fields can be expanded in terms of elementary wave functions in
the cylindrical coordinate systems. Since the cylinder cross section does not change with z,
the scattered field maintains the same z-dependence as that of the incident wave

Es (ρ ρ) eiβ z ,
ρ, z) = Es (ρ

where β = k0 cos θi . The scalar potential ψ generating Es , in the region ρ ≥ a, can be written
as
+∞
X (1)
ψ (ρ
ρ) = An Hn (kρ ρ ) einφ , (7.178)
n=−∞

where kρ = k0 sin θi . The choice of the Hankel function of the first kind is to ensure an
outgoing cylindrical wave. Referring to Eq. (5.102a), the components of the scattered electric
field are given by
+∞
X (1) ′
Eρs (ρ , φ , z) = iAn β kρ Hn (kρ ρ ) einφ eiβ z , (7.179a)
n=−∞
+∞
X −An β n (1)
Eφs (ρ , φ , z) = Hn (kρ ρ ) einφ eiβ z , (7.179b)
n=−∞
ρ
and
+∞
X (1)
Ezs (ρ , φ , z) = An kρ2 Hn (kρ ρ ) einφ eiβ z (7.179c)
n=−∞

The incident wave can also be expanded in terms of cylindrical eigenfunctions. Using
+∞
X
e−ik0 sin θi x = (−i)n Jn (kρ ρ ) einφ , (7.180)
−∞
q
where kρ = k02 − β 2 = k0 sin θi , we have

+∞
X
Ezi = E0 sin θi eiβ z (−i)n Jn (kρ ρ ) einφ , (7.181a)
n=−∞
454 Chapter 7 Cylindrical Wave Functions and Their Applications

+∞
X
iβ z
Eφi = −E0 cos θi sin φ e (−i)n Jn (kρ ρ ) einφ , (7.181b)
n=−∞
and
+∞
X
Eρi = E0 cos θi cos φ eiβ z (−i)n Jn (kρ ρ ) einφ . (7.181c)
n=−∞

At the boundary between the cylinder and its surrounding medium,


(Eφi + Eφs )ρ =a = 0 .

This condition provides the following equation for the unknown An :

+∞
X +∞
X
inφ An β n (1)
−E0 cos θi n
(−i) Jn (ka) e sin φ = Hn (kρ a) einφ . (7.182)
n=−∞ n=−∞
a

Multiplying both sides of Eq. (7.182) by e−imφ and integrating the results over 0 to 2π provides
the following equation for Am :

mAm β (1) −E0 cos θi

Hm (kρ a) = [(−i)m−1 Jm−1 (kρ a) − (−i)m+1 Jm+1 (kρ a)] .
a 2i
Hence
 
−aE0 (−i)m Jm+1 (kρ a) + Jm−1 (kρ a)
Am = (1)
k0 Hm (kρ a) 2m
−aE0 (−i)m Jm (kρ a) (−i)m Jm (k0 sin θi a)
= = −E 0 . (7.183)
(1)
k0 Hm (kρ a) kρ a k02 sin θi Hm(1) (k0 sin θi a)

Similar results can be obtained using the boundary condition

(Ezi + Ezs )|ρ =a = 0 . (7.184)

Using Eqs. (7.179) and (7.181a) in Eq. (7.184) leads to

+∞
X +∞
X
inφ (1)
E0 sin θi n
(−i) Jn (kρ a) e =− An kρ2 Hn (kρ a) einφ . (7.185)
n=−∞ n=−∞

Multiplying both sides of Eq. (7.185) by e−imφ and integrating over φ in the range [0, 2π ]
gives
(1)
E0 sin θi (−i)m Jm (kρ a) = −Am kρ2 Hm (kρ a) .
Hence
(−i)m Jm (k0 sin θi a)
Am = −E0 , (7.186)
k02 sin θi Hm(1) (k0 sin θi a)
7-4 Scattering from a Metallic Circular Cylinder 455

which is consistent with Eq. (7.183).

In the far-field region, simpler expressions for the scattered field quantities can be
obtained. Noting that
(1) ′ 1 h (1) (1)
i
Hn (x) = Hn−1 (x) − Hn+1 (x) , (7.187)
2
and using the large-argument expansion of the Hankel function, we have
r
(1) ′ 1 2 i{x−[(2n+1)/4]π } iπ /2
lim Hn (x) ≈ e [e − e−iπ /2 ]
x→∞ 2 πx
r
2 i{x−[(2n+1)/4]π }
≈i e . (7.188)
πx

From Eq. (7.179) it is obvious that Eφs varies with ρ as 1/ρ 3/2 , whereas Eρs and Ezs decay as
1/ρ 1/2 . In the far-field region the Eφs -component can be ignored compared with the other two
terms. Hence, for points far away from the cylinder axis where kρ ρ = k0 sin θi ρ ≫ 1,
s +∞
!
2 X
i(kρ ρ −π /4) n inφ
s
Eρ (ρ , φ , z) = iβ kρ e iAn (−i) e eiβ z . (7.189)
π kρ ρ n=−∞

Recalling that β = k0 cos θi and kρ = k0 sin θi , and using Eq. (7.183) in Eq. (7.189),
s
2
Eρs (ρ , φ , z) = E0 cos θi ei(k0 sin θi ρ −π /4) eik0 cos θi z
π k0 sin θi ρ
+∞
X Jn (k0 sin θi a)
· (−1)n+1 (1)
einφ . (7.190)
n=−∞ Hn (k0 sin θi a)

Also, recalling that for any Bessel function Zn ,

Z−n = (−1)n Zn ,

Eq. (7.190) can be further reduced to

s
2
Eρs (ρ , φ , z) = −E0 cos θi ei(k0 sin θi ρ −π /4) eik0 cos θi z
π k0 sin θi ρ
∞
!
J0 (k0 sin θi a) X
n+1 Jn (k0 sin θi a)
· (1)
+2 (−1) (1)
cos(nφ ) .
H0 (k0 sin θi a) n=1 H n (k0 sin θi a)
456 Chapter 7 Cylindrical Wave Functions and Their Applications

z
k̂s

y
θi k̂i

x
Figure 7-24: Ray representation of a scattered field from a metallic cylinder illuminated by a
plane wave at oblique incidence. The scattered rays form a cone around the axis of the cylinder
with a cone angle equal to the incident angle.

In a similar manner,
s
2
Ezs (ρ , φ , z) = E0 sin θi ei(k0 sin θi ρ −π /4) eik0 cos θi z
π k0 sin θi ρ
∞
!
J0 (k0 sin θi a) X J (k
n 0 sin θi a)
· (1)
+2 (−1)n+1 (1) cos(nφ ) .
H0 (k0 sin θi a) n=1 Hn (k0 sin θi a)

The total scattered field in the far-zone is therefore

s
2
Es (ρ , φ , z) = (E0 sin θi ẑ − E0 cos θi ρ̂ ) ei(k0 sin θi ρ −π /4) eik0 cos θi z
π k0 sin θi ρ
∞
!
J0 (k0 sin θi a)X Jn (k0 sin θi a)
· (1)
+2 (−1)n+1 (1) cos(nφ ) . (7.191)
H0 (k0 sin θi a) n=1 Hn (k0 sin θi a)

The direction of propagation is along

k̂s = sin θi ρ̂ + cos θi ẑ ,

which indicates that the propagation of the scattered field is confined to the surface of a cone
with a generating angle θi , as shown in Fig. 7-24. The bistatic scattering echo width of a 2-D
7-4 Scattering from a Metallic Circular Cylinder 457

object is mathematically defined as

|Es |2
σ2 (φ ) = lim 2πρ , (7.192)
ρ →∞ |Ei |2
which can be interpreted as a fictitious width of an object that captures the incident power
density of the incident wave, and if it were to reradiate that power isotropically (uniform in all
φ -directions) it would produce the same scattered power density at the observation point as
does the scatterer. Using Eq. (7.191) in Eq. (7.192), the radar echo width of a metallic cylinder
for a TM polarized incident wave is given by

 ∞ 2
X 
4  n Jn (k0 sin θi a) inφ 
σ2TM (φ ) =  (−1) (1)
e  . (7.193)
k0 sin θi 
n=−∞ Hn (k0 sin θi a) 

The magnetic field in the far-field region is perpendicular to Es and k̂s and is given by
r
s k̂s × Es µ0
H (ρ , φ , s) = , η0 = .
η0 ε0
The solution for TE incidence can be obtained by applying the duality. The electric field is
given by  
∂ 1 ∂ψ
s
E (ρ , φ , z) = iω µ0 ψ ρ̂ + φ × ẑ eik0 sin θi z ,
φ̂ (7.194)
∂ρ ρ ∂φ
with ψ (ρ , φ ) given by
+∞
X (1)
ψ (ρ , φ ) = Bn Hn (kρ ρ ) einφ .
n=−∞

The components of the scattered electric field are given by

+∞
X in (1)
Eρs (ρ , φ , z) = iω µ0 Bn Hn (kρ ρ ) einφ eiβ z (7.195a)
n=−∞
ρ
and
+∞
X (1) ′
Eφs (ρ , φ , z) = −iω µ0 Bn kρ Hn (kρ ρ ) einφ eiβ z . (7.195b)
n=−∞

The incident wave for a TE polarized wave is given by

Ei (ρ , φ , z) = E0 ŷ eik0 (− sin θi x+cos θi z)

φ)eik0 (− sin θi x+cos θi z) .
= E0 (sin φ ρ̂ + cos φ φ̂
458 Chapter 7 Cylindrical Wave Functions and Their Applications

The boundary condition mandates that

+∞
X +∞
X (1) ′
E0 cos φ (−i)n Jn (k0 sin θi a) einφ = −iω µ0 Bn kρ Hn (kρ a) einφ . (7.196)
n=−∞ n=−∞

Multiplying both sides of Eq. (7.196) by e−imφ and integrating over [0, 2π ] gives
E0  (1) ′
(−i)m−1 Jm−1 (kρ a) + (−i)m+1 Jm+1 (kρ a) = −iω µ0 kρ Hm (kρ a) Bm
2
and
 
m Jm−1 (kρ a) − Jm+1 (kρ a) (1) ′
E0 (−i) = −ω µ0 kρ Hm (kρ a) Bm .
2
Noting that
Jm−1 (x) − Jm+1 (x)
= Jm′ (x) ,
2
the unknown coefficient Bm can be obtained from
(−i)m Jm′ (k0 sin θi a)
Bm = −E0 . (7.197)
(1) ′
ω µ0 k0 sin θi Hm (k0 sin θi a)
Substituting Eq. (7.197) into Eq. (7.195b),
+∞
X Jn′ (k0 sin θi a) (1) ′
Eφs (ρ , φ , z) = −iE0 (−i)n Hn (k0 sin θi ρ ) einφ eik0 cos θi z .
(1) ′
n=−∞ Hn (k0 sin θi a)

In the far field, using Eq. (7.188), Eφs (ρ , φ , z) is given by

s
2
Eφs (ρ , φ , z) = E0 ei(k0 sin θi ρ −π /4) eik0 cos θi z
π k0 sin θi ρ
+∞
X Jn′ (k0 sin θi a)
(−1)n einφ . (7.198)
(1) ′
n=−∞ Hn (k0 sin θi a)
The cylinder echo width can also be obtained from Eq. (7.192) and is given by

 2
X +∞ ′ 
4  n Jn (k0 sin θi a) inφ 
σ2TE (φ ) =  (−1) e  . (7.199)
k0 sin θi  (1) ′ 
n=−∞ Hn (k0 sin θi a)
7-4 Scattering from a Metallic Circular Cylinder 459

7-4.1 Low-Frequency Approximation

In situations where the cylinder diameter is much smaller than the wavelength; i.e.,
k0 sin θi a ≪ 1, the asymptotic expansion of Bessel functions can be used to find a closed
form solution for the radar echo width of metallic cylinders. Considering the fact that
limx→0 Jn (x) ≈ 0 for n , 0,

J0 (k0 sin θi a) ≈ 1
and
(1) 2
H0 (k0 sin θi a) ≈ 1 + i ln(0.8905k0 sin θi a) .
π
Hence, Eq. (7.193) is reduced to

1 πλ
σ2TM = . (7.200)
2 sin θi (π 2 /4) + ln2 (0.8905k0 a sin θi )

Also, noting that

lim J0′ (x) ≈ x ,

x→0

(1)′ 2 1
lim H0 (x) ≈ ,
x→0 π x
1
lim J1′ (x) ≈ ,
x→0 2
and
(1)′ 4 1
lim H1 (x) ≈ i ,
x→0 π x2
it can be shown that  2
π 2a 3 1
σ2TE = (k0 a sin θi ) + cos φ . (7.201)
sin θi 2

7-4.2 Numerical Simulations

To examine the scattering behavior of a metallic cylinder, the radar backscatter echo width of
a metallic cylinder illuminated by a plane wave at normal incidence is plotted in Fig. 7-25 as a
function of ka for both TE and TM polarizations. Here the echo width is normalized to its high
frequency value, π a. A similar plot is presented in Fig. 7-26, where the incident wave is now
at an oblique incidence θi = 45◦ . Figures 7-27 and 7-28 present the bistatic radar cross section
patterns for two metallic cylinder, one with ka = 1 and another with ka = 5, respectively, at
normal incidence (θi = 90◦ ) for both TE and TM polarizations. It is interesting to note that
the maximum radar cross section is observed in the forward direction. The scattered field in
the forward direction is nearly 180◦ out of phase with respect to the incident wave. In the
vicinity of the cylinder the scattered field totally cancels out the incident field, which results
in a shadow region near the forward scattering direction. Comparing Fig. 7-27 with Fig. 7-28
460 Chapter 7 Cylindrical Wave Functions and Their Applications

PEC cylinder backscattering for θi = 90°

10
1 TM exact solution TE low-frequency approximation
Normalized 2D RCS σ2 /(π a)

0
10

TM low-frequency
−1 approximation
10

TE exact TM
10
−2 solution TM low-frequency approximation
TE
TE low-frequency approximation
−2 −1 0 1 2
10 10 10 10 10
Normalized frequency ka

Figure 7-25: Normalized backscatter radar echo width of a metallic cylinder at normal incidence
for both TE and TM polarizations. Also shown are the low-frequency approximations. The
backscatter echo width approaches the high-frequency limit of σ2 = π a for values of ka ≥ 10.

shows that bistatic scattering echo width increases with increasing the radius of the cylinder
and the scattering lobes become narrower.

7-5 Integral Representation of Bessel Functions

The integral representation of Bessel functions is useful in a range of problems related
to scattering and antennas. In fact the large-argument approximation for the Bessel
functions given in Eqs. (7.16a)–(7.16d) are derived from the asymptotic evaluation of such
representations. It was shown that Bessel functions are solutions to the 2-D wave equation

∇2t ψ (ρ , φ ) + kρ2 ψ (ρ , φ ) = 0 . (7.202)

Following a procedure similar to that in Section 7-2.7, the solution to Eq. (7.202) can be
written as Z +∞
ψ (x, y) = A(kx ) ei(kx x+ky |y|) dkx , (7.203)
−∞
7-5 Integral Representation of Bessel Functions 461

PEC cylinder backscattering for θi = 45°

1 TM exact solution TE low-frequency approximation

10
Normalized 2D RCS σ2 /(π a)

0
10

TM low-frequency
−1 approximation
10

TE exact TM
−2 solution TM low-frequency approximation
10
TE
TE low-frequency approximation
−2 −1 0 1 2
10 10 10 10 10
Normalized frequency ka

Figure 7-26: Normalized backscatter radar echo width of a metallic cylinder at oblique
incidence (θi = 45◦ ) for both TE and TM polarizations. Also shown are the low-frequency
approximations. The backscatter echo width approaches the high-frequency limit of σ2 = π a
for values of ka ≥ 10.

q
(see Eq. (7.162)), where ky = kρ2 − kx2 . By performing the following change of variables,

kx = kρ cos γ ,
(7.204)
ky = kρ sin γ ,

and
x = ρ cos φ ,
(7.205)
y = ρ sin φ ,

Eq. (7.203) can be rewritten as

Z
ψ (ρ , φ ) = B(γ ) eikρ ρ cos(γ ∓φ ) d γ , (7.206)
C
462 Chapter 7 Cylindrical Wave Functions and Their Applications

90˚ 20 TM
120˚ 60˚ TE
10

150˚ 0 30˚

−10

φ = 180˚ (forward φ=0

180˚ 0
scatter direction) (backscatter direction)

210˚ 330˚

240˚ 300˚
270˚

Figure 7-27: Bistatic echo width of a metallic cylinder with ka = 1 at normal incidence for both
TE and TM polarizations.

90˚ 20 TM
120˚ 60˚ TE
10

150˚ 0 30˚

−10

180˚ 0

210˚ ˚ 330˚

240˚ 300˚
270˚

Figure 7-28: Bistatic echo width of a metallic cylinder with ka = 5 at normal incidence for both
TE and TM polarization.
7-5 Integral Representation of Bessel Functions 463

where B(γ ) = − sin γ kρ A(kρ cos γ ), a function to be determined. In Eq. (7.206), the negative
sign in the argument of the cosine function corresponds to y > 0 and the positive sign
corresponds to y < 0. The contour C corresponds to a path defined by Eq. (7.204). Another
change of variable is used as φ ′ = γ ∓ φ with d γ = d φ ′ as
Z
′
ψ (ρ , φ ) = B(φ ′ ± φ ) eikρ ρ cos(φ ) d φ ′ . (7.207)
C

If ψ (ρ , φ ) is to be represented by the product of two functions, each a function of only one

variable, that is, ψ (ρ , φ ) = F(φ ) R(kρ ρ ), then

B(φ ′ ± φ ) = F(φ ) g(φ ′ ) . (7.208)

Substitution of Eq. (7.208) into Eq. (7.207) shows that

Z
′
R(kρ ρ ) = g(φ ′ ) eikρ ρ cos(φ ) d φ ′ . (7.209)
C

Applying the method of separation of variables to Eq. (7.202) shows that

F(φ ) = aeiνφ + be−iνφ (7.210)

and  
d 2 R(ρe) 1 dR(ρe) ν2
+ + 1 − 2 R(ρe) = 0 . (7.211)
d ρe2 ρe d ρe ρe
That is, Eq. (7.209) must satisfy Eq. (7.211) with ρe = kρ ρ . Starting from Eq. (7.209),
Z
dR(ρe) ′
= i cos φ ′ g(φ ′ ) eieρ cos φ d φ ′ (7.212a)
d ρe C
and
Z
d 2 R(ρe) ′
= − cos2 φ ′ g(φ ′ ) eieρ cos φ d φ ′ . (7.212b)
d ρe2 C

Upon substitution of Eqs. (7.212a) and (7.212b) into Eq. (7.211), the following relation must
hold for Eq. (7.209) to satisfy the Bessel differential equation:
Z

′
g(φ ′ ) sin2 φ ′ ρe2 + i cos φ ′ ρe − ν 2 eieρ cos φ d φ ′ = 0 . (7.213)
C

Note that ′
d 2 eieρ cos φ ′ ′

′ 2
= −ρe2 sin2 φ ′ eieρ cos φ − iρe cos φ ′ eieρ cos φ .
dφ
464 Chapter 7 Cylindrical Wave Functions and Their Applications

As a result, Eq. (7.213) can be rewritten as

Z ′
!
d 2 eieρ cos φ ′
g(φ ′ ) ′ 2
+ ν 2 eieρ cos φ dφ ′ = 0 . (7.214)
C dφ

Also, we realize that

′ ′
!
d 2 eieρ cos φ 2
ρ cos φ ′ d g(φ )
′ d deieρ cos φ ρ cos φ ′ dg(φ )
′
g(φ ′ ) − eie
= g(φ ′ ) − eie
,
dφ ′ 2
dφ ′ 2 dφ ′ dφ ′ dφ ′

which can be used in Eq. (7.214) to arrive at

Z ′
!
d deieρ cos φ
′ ρ cos φ ′ dg(φ )
′
′
g(φ ) −e ie
dφ ′
C dφ dφ ′ dφ ′
Z  2 
′ d g(φ ′ )
+ eieρ cos φ ′ 2
+ ν 2
g(φ ′
) dφ ′ = 0 . (7.215)
dφ
Equation (7.215) is valid if
  
′ ′ dg(φ ′ ) ieρ cos φ ′ 
iρe sin φ g(φ ) + e  =0 (7.216)
dφ ′ end points

and
d 2 g(φ ′ )
′ 2
+ ν 2 g(φ ′ ) = 0 . (7.217)
dφ
A harmonic function can be chosen for Eq. (7.217) and can be written as
′
g(φ ′ ) = ce±iνφ .

The constant c is chosen as c = π1 e−iνπ /2 so that the resulting solutions are consistent with the
previous expressions for the Bessel function. To satisfy Eq. (7.216), the path of the integral in
the complex φ ′ plane must be chosen so that at the end points the requirement of Eq. (7.216)
is met. Finally, a general integral representation for the Bessel function can be written as
Z
e−iνπ /2
Zν (kρ ρ ) = eikρ ρ cos φ ±iνφ d φ , (7.218)
π C

subject to the following condition on contour C:



(kρ ρ sin φ ± ν )ei(kρ ρ cos φ ±νφ )  =0. (7.219)
end points
7-5 Integral Representation of Bessel Functions 465

For example, if we choose the contour on the complex φ -plane to be a segment of the real
axis from −π to π , then for ν = n (integer),
   
kρ ρ sin(π ) ± n ei(−kρ ρ ±nπ ) − kρ ρ sin(−π ) ± n ei(−kρ ρ ∓nπ ) = ±ne−ikρ ρ (e±inπ − e∓inπ )
= ±2ine−ikρ ρ sin(±nπ )
=0,

and then Eq. (7.219) is satisfied. In fact, if C is any segment of the real axis with length 2π ,
Eq. (7.219) is satisfied. This can be used to define the Bessel function of the first kind as
Z
(−i)n π ikρ ρ cos φ +inφ
Jn (kρ ρ ) = e dφ . (7.220)
2π −π

Other kinds of Bessel functions of arbitrary order can be obtained by requiring ei(kρ ρ cos φ +νφ )
to vanish at the endpoints on contour C in the complex φ plane. Let us define

φ = α + iβ ,

where α and β are respectively the real and imaginary parts of φ . In this case, the exponent
can be written as

ikρ ρ cos φ ± iνφ = ∓νβ + kρ ρ sin α sinh β + i(±να + kρ ρ cos α cosh β ) . (7.221)

If contour C is chosen so that (∓νβ + kρ ρ sin α sinh β ) is −∞ at the endpoints of the contour,
Eq. (7.218) is a solution to the Bessel differential equation. This can be the case if β → ∞
and α = (2nπ − π /2) for any natural number n. Also, if α = (2nπ + π /2) and β → −∞, the
exponent approaches −∞ again. So it seems that contours with endpoints [(2nπ − π /2) + i∞]
and [(2nπ + π /2) − i∞] will satisfy the requirement. Sommerfield defined Hankel functions
of the first kind and second kind as
Z
(1) e−iνπ /2 π /2−i∞ ikρ ρ cos φ +iνφ
Hν (kρ ρ ) = e dφ (7.222a)
π −π /2+i∞
and
Z 3π /2+i∞
(2) e−iνπ /2
Hν (kρ ρ ) = eikρ ρ cos φ +iνφ d φ . (7.222b)
π π /2−i∞

These contours are shown in Fig. 7-29 in the complex φ plane. Note that the actual functions
defining the shapes of these contours are not important so long as the endpoints are not
changed.
In the asymptotic evaluation of Eq. (7.222a) or Eq. (7.222b) when kρ ρ ≫ 1, the method
of stationary-phase approximation or the saddle-point technique is used. For such techniques
the contour path is chosen to go through the saddle point or the stationary phase point, near
which the variations of the phase of the integrand is slow. The saddle point (stationary phase
466 Chapter 7 Cylindrical Wave Functions and Their Applications

Complex ϕ plane

Figure 7-29: The complex φ plane and the appropriate contours for Hankel functions of the first
kind (C1 ) and Hankel functions of the second kind (C2 ).

point) is obtained from the solution


d 
(kρ ρ cos φ + νφ ) = 0 , (7.223a)
dφ φs
 
ν
φs = sin−1 ≈ 0, ±π , ±2π , . . . for ν ≪ kρ ρ . (7.223b)
kρ ρ

The contour chosen for Eq. (7.222a), C1 , is referred to as the steepest descent path, and it
goes through φs = 0. The contour chosen for Eq. (7.222b), C2 , goes through π , as shown in
(1)
Fig. 7-29. Applying the method of steepest descent path, the approximate values of Hν and
(2)
Hν for large values of kρ ρ are given by
s
(1) 2
Hν (kρ ρ ) ≈ ei{kρ ρ −[(2ν +1)/4]π } (7.224a)
π kρ ρ
and
s
(2) 2
Hν (kρ ρ ) ≈ e−i{kρ ρ −[(2ν +1)/4]π } (7.224b)
π kρ ρ
7-6 2-D Green’s Function for Homogeneous Media in the Presence of Metallic Wedge 467

For details, see Appendix B.

The integral representation for Bessel functions of the first kind and noninteger order can
be obtained from the definition
1 h (1) (2)
i
Jν (kρ ρ ) = Hν (kρ ρ ) + Hν (kρ ρ ) . (7.225)
2
Noting that contours C1 and C2 can be deformed to coincide anywhere and then follow along
α = π /2 to go to π /2 − i ∞, contour C3 for Jν (kρ ρ ) can be drawn as the one shown in
(2)
Fig. 7-29. It is noted that C2′ can be used in lieu of C2 to find Hν (kρ ρ ), and therefore C3′
can also be used to find Jν (kρ ρ ):
Z
e−iνπ /2
Jν (kρ ρ ) = eikρ ρ cos φ ±iνφ d φ . (7.226)
2π C3 ,C3′

7-6 2-D Green’s Function for Homogeneous Media in the

Presence of a Metallic Wedge
Another subject of practical importance is the scattering of electromagnetic waves by a
metallic wedge. This problem reveals the behavior of electromagnetic fields near objects with
sharp edges. The most general solution for an arbitrary current distribution can be obtained
if the solution for a line source is available. Let us consider a metallic wedge of angle α as
shown in Fig. 7-30, excited by a current source given by

PEC

Figure 7-30: Configuration of a metallic wedge with angle α excited by a z-directed line current
at (ρ ′ , φ ′ ).
468 Chapter 7 Cylindrical Wave Functions and Their Applications

I
J(r) = δ (ρ − ρ ′) δ (φ − φ ′ ) eiβ z ẑ . (7.227)
ρ
One side of the wedge face is on the x–z plane (φ = 0) and the other face is at φ = 2π − α .
The Hertz vector potential has only a z-component and it must satisfy

−i iβ z δ (ρ − ρ ′ )
∇2 Π z + k 2 Π z = Ie δ (φ − φ ′ ) . (7.228)
ωε ρ
As before, since the geometry of the wedge is invariant with respect to z, the Hertz potential
can be written as
Πz (r) = Ψ(ρ , φ ) eiβ z
and
−i δ (ρ − ρ ′ )
∇2t Ψ(ρ , φ ) + kρ2 Ψ(ρ , φ ) = I δ (φ − φ ′ ) . (7.229)
ωε ρ
The boundary condition for the TM to z field mandates that

Ψ(ρ , 0) = Ψ(ρ , 2π − α ) = 0 .

The method of separation of variables can be used in this case since the boundary condition is
applied to constant coordinate surfaces φ = 0 and φ = 2π − α . The domain of interest includes
0 ≤ φ ≤ 2π − α and 0 ≤ ρ < ∞. We divide the region of interest into two regions: 1) ρ ≤ ρ ′
and 2) ρ ≥ ρ ′ . In Region 1, which includes ρ = 0, the only admissible Bessel function is
the Bessel function of the first kind, Jν (kρ ρ ), and in Region 2 the admissible Bessel function
(1)
is Hν (kρ ρ ), which represents outgoing waves. The admissible solution with respect to φ is
sin νφ , as it satisfies the boundary condition ψ (ρ , 0) = 0. To satisfy the boundary condition
at φ = 2π − α , we must have
sin[ν (2π − α )] = 0 ,
which renders
mπ
ν= , m = 1, 2, 3, . . .
2π − α
For simplicity of carrying indices, let us define 2π − α = pπ for some real number p. In this
case the order of the Bessel function is ν = m/p. Also, recalling that functions sin( mp φ ) form
a set of complete orthogonal functions in the domain [0, pπ ],
∞
X    
′ 2 m ′ m
δ (φ − φ ) = sin φ sin φ . (7.230)
2π − α p p
m=1

The scalar potential that satisfies Eq. (7.229) in regions ρ < ρ ′ and ρ > ρ ′ are given
respectively by
X∞    
m ′ m
Ψ< (ρ , φ ) = Am Jm/p (kρ ρ ) sin φ sin φ (7.231a)
p p
m=1
and
7-6 2-D Green’s Function for Homogeneous Media in the Presence of Metallic Wedge 469

∞
X    
(1) m ′ m
Ψ> (ρ , φ ) = Bm Hm/p (kρ ρ ) sin φ sin φ . (7.231b)
p p
m=1

Equation (7.229) mandates that

Ψ> − Ψ< |ρ =ρ ′ = 0 (7.232a)

and

∂ Ψ> ∂ Ψ<  −iI
− = δ (φ − φ ′ ) . (7.232b)
∂ρ ∂ ρ ρ =ρ ′ ωερ ′

Using Eqs. (7.230), (7.231a), and (7.231b) in Eqs. (7.232a) and (7.232b), the following
equations for Am and Bm are obtained:
(1)
Am Jm/p (kρ ρ ′ ) = Bm Hm/p (kρ ρ ′ ) (7.233a)
and
 
(1)′ 2 −iI
Bm kρ Hm/p (kρ ρ ′ ) − Am kρ ′
Jm/p (kρ ρ ′ ) = . (7.233b)
(2π − α )ρ ′ ωε
The application of Wronskian

(1)′ (1) 2i
Jm/p (kρ ρ ′ ) Hm/p (kρ ρ ′ ) − Jm/p
′
(kρ ρ ′ ) Hm/p (kρ ρ ′ ) = (7.234)
π kρ ρ ′

provides compact expressions for Am and Bm from Eqs. (7.233a) and (7.233b):
−I (1)
Am = H (kρ ρ ′ ) (7.235a)
pωε m/p
and
−I
Bm = J (kρ ρ ′ ) . (7.235b)
pωε m/p
Hence the potential in each region is given by
∞    
−I X (1) ′ m ′ m
Ψ< (ρ , φ ) = Hm/p (kρ ρ ) Jm/p (kρ ρ ) sin φ sin φ (7.236a)
pωε p p
m=1
and
∞    
−I X ′ (1) m ′ m
Ψ> (ρ , φ ) = Jm/p (kρ ρ ) Hm/p (kρ ρ ) sin φ sin φ . (7.236b)
pωε p p
m=1

As expected, this solution satisfies the reciprocity requirement. The field expressions can be
derived from Eqs. (7.236a) and (7.236b) using Eqs. (5.102a) and (5.102b):

Ez (ρ , φ ) = kρ2 ψ (ρ , φ ) eiβ z , (7.237a)

470 Chapter 7 Cylindrical Wave Functions and Their Applications

∂
Eρ (ρ , φ ) = iβ ψ (ρ , φ ) eiβ z , (7.237b)
∂ρ
1 ∂
Eφ (ρ , φ ) = iβ ψ (ρ , φ ) eiβ z , (7.237c)
ρ ∂φ
∂
Hφ (ρ , φ ) = iωε ψ (ρ , φ ) eiβ z , (7.237d)
∂ρ
1 ∂
Hρ (ρ , φ ) = −iωε ψ (ρ , φ ) eiβ z . (7.237e)
ρ ∂φ
In specific,
 ∞    

 −Ikρ2 eiβ z X (1) ′ m ′ m


 pωε Hm/p (kρ ρ ) Jm/p (kρ ρ ) sin φ sin φ ρ < ρ′ ,
p p
Ez (ρ , φ ) = m=1
∞    

 −Ikρ2 eiβ z X ′ (1) m ′ m

 Jm/p (kρ ρ ) Hm/p (kρ ρ ) sin φ sin φ ρ > ρ′ .
 pωε p p
m=1
(7.238)
(1)
In the far-field region, as kρ ρ → ∞, the large-argument expansion of Hm/p (kρ ρ ) can be used:
s
(1) −2i
Hm/p (kρ ρ ) ≈ (−i)m/p eikρ ρ (7.239a)
π kρ ρ
and
s ∞    
−Ikρ2 −2i i(kρ ρ +β z) X m/p ′ m ′ m
Ez (ρ , φ ) = e (−i) Jm/p (kρ ρ ) sin φ sin φ ,
pωε π kρ ρ p p
m=1
(7.239b)

where p = 2 − α /π . Other field components can be derived in a similar manner.

As an example, Fig. 7-31(a) and (b) show the z-component of the electric field in the far-
field region for metallic wedges with interior angles of 30◦ and 90◦ when excited by a line
source placed at φ ′ = 165◦ and φ ′ = 135◦ , respectively, for three different values of ρ ′ . As
shown, the total electric field along the wedge surface is zero.

7-6.1 The TE Solution

Transverse electric fields are generated by a z-directed magnetic filament current (located at
ρ = ρ ′ and φ = φ ′ ) of the form
Im
Jm (r) = δ (ρ − ρ ′) δ (φ − φ ′ ) eiβ z ẑ . (7.240)
ρ
The required boundary conditions on the surface of the metallic wedge shown in Fig. 7-30
requires ∂ ψ /∂ n = ∂ ψ /∂ φ = 0 at φ = 0 and φ = 2π − α . This mandates the harmonic wave
function with respect to φ to be of the form of cos νφ with ν = m/p, where p = 2 − α /π .
7-6 2-D Green’s Function for Homogeneous Media in the Presence of Metallic Wedge 471

105° 90° 75° ' = /4

120° 60° ' = /2
135° 45° '=2

150° 30°

165° 15°
0 −10 −20
(a) α = 30˚ 180°
−30 0°

−165° −15°

−150° −30°

−135° −45°
−120° −60°
−105° −90° −75°

90° ' = /4
105° 75°
120° 60° ' = /2
135° 45° '=2

150° 30°

165° 15°
0 −10 −20
(b) α = 90˚ 180°
−30 0°

−165° −15°

−150° −30°

−135° −45°
−120° −60°
−105° −90° −75°

Figure 7-31: The z-component of the electric field in the far-field region for a metallic wedge
with angular extent (a) α = 30◦ and (b) al = 90◦ .
472 Chapter 7 Cylindrical Wave Functions and Their Applications

Also, noting that

∞    
′ 2 X m ′ m
δ (φ − φ ) = αm cos φ cos φ , (7.241)
pπ p p
m=0

where (
1
2 m=0,
αm =
1 m,0,
the appropriate scalar potential can be found by applying duality to Eqs. (7.236a) and
(7.236b):
∞    
−Im X (1) ′ m ′ m
Ψ< (ρ , φ ) = αm Hm/p (kρ ρ ) Jm/p (kρ ρ ) cos φ cos φ (7.242a)
pω µ p p
m=0
and
∞    
−Im X ′ (1) m ′ m
Ψ> (ρ , φ ) = αm Jm/p (kρ ρ ) Hm/p (kρ ρ ) cos φ cos φ . (7.242b)
pω µ p p
m=0

By applying Eqs. (7.242a) and (7.242b) to the dual of Eqs. (7.237e) and (7.238), all field
components can be obtained.

7-6.2 Plane-Wave Scattering from a Conducting Wedge

The behavior of the total field in the vicinity of a conducting wedge when illuminated by a
plane wave is of interest. To find the solution, consider the wedge shown in Fig. 7-32 when
illuminated by a TM plane wave given by
  ′ ′ ′
Ei (ρ , φ , z) = E0 cos θ ′ (cos φ ′ x̂ + sin φ ′ ŷ) + sin θ ′ ẑ e−ik0 sin θ cos(φ −φ )ρ eik0 cos θ z . (7.243)

ρ′ = cos φ ′ x̂ + sin φ ′ ŷ. The field

For this wave we define β = k0 cos θ ′ , kρ = k0 sin θ ′ , and ρ̂
′
ρ was obtained in Section 7-2.6 to be
generated by an infinite filament current located at ρ̂

−kρ2 I (1) iβ kρ I (1)′ ρ − ρ′

E(ρ , φ ) = ρ − ρ ′ |) eiβ z ẑ −
H0 (kρ |ρ ρ − ρ ′ |) eiβ z
H0 (kρ |ρ (7.244)
4ωε 4ωε ρ − ρ′|
|ρ

(see Eq. (7.139)). By placing the current far away from the origin such that kρ ρ ′ ≫ 1, the
large-argument expansion of the Hankel function can be used to find an approximate field
expression in the vicinity of the origin. Noting that

ρ − ρ ′ | ≈ ρ ′ − ρ cos(φ − φ ′ ) ,
|ρ
7-6 2-D Green’s Function for Homogeneous Media in the Presence of Metallic Wedge 473

x x

kρ = k sin θ΄ ˆi
k

y
y

Figure 7-32: Configuration of a metallic wedge illuminated by a plane wave of oblique

incidence.

and using the large-argument expansion of Hankel functions,

s
(1) ′ −2i ikρ ρ ′ −ikρ ρ cos(φ −φ ′ )
H0 (kρ |ρ
ρ − ρ |) ≈ e e (7.245a)
π kρ ρ ′
and
s
(1)′ −2i ′ ′
ρ − ρ ′ |) ≈
H0 (kρ |ρ ieikρ ρ e−ikρ ρ cos(φ −φ ) . (7.245b)
π kρ ρ ′

Also, for ρ ′ ≫ ρ , we have

ρ − ρ′
ρ′ .
≈ −ρ̂ (7.246)
ρ − ρ′ |
|ρ
Substituting Eqs. (7.245a), (7.245b), and (7.246) into Eq. (7.244) and comparing the resultant
equation with Eq. (7.243), we recognize that if
s
kkρ I −2i ikρ ρ ′
E0 = − e , (7.247)
4ωε π kρ ρ ′

then the field of a distant filament current of infinite extent resembles a local plane wave near
the tip of the wedge. Hence the total field (incident plus diffracted) can be obtained from
(1)
Eq. (7.238) for ρ < ρ ′ using the large-argument expansion of Hν (kρ ρ ′ ). That is,
∞
X    
4 sin θ ′ ik cos θ ′ z m m ′
Ezt = E0 e (−i)m/p
Jm/p (kρ ρ ) sin φ sin φ . (7.248)
p p p
m=1
474 Chapter 7 Cylindrical Wave Functions and Their Applications

Other field components can be derived from Eq. (7.237). For example,
∞
X    
t 4i cos θ ′ ik cos θ ′ z m/p ′ m m ′
Eρ = E0 e (−i) Jm/p (kρ ρ ) sin φ sin φ (7.249a)
p p p
m=1
and
′ ∞
X    
4i cos θ ′ eik cos θ z m m m ′
Eφt = E0 (−i)m/p
Jm/p (kρ ρ ) cos φ sin φ . (7.249b)
p kρ ρ p p p
m=1

It is noted that at normal incidence for which θ ′ = π /2, both Eρt and Eφt vanish.

7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic

Wedge
In this section we derive a simplified expression for the fields diffracted by a metallic
wedge illuminated by a plane wave. The total field expression for the TM wave is given
by Eq. (7.248), and that for the TE wave can be obtained by using duality and the potential
expression given by Eq. (7.242a). The goal here is to evaluate the infinite summation in these
expressions for cases where kρ ρ ≫ 1. Let us denote the infinite summation by
∞     
TM 1 X m/p m ′ m ′
S TE = αm (−i) Jm/p (kρ ρ ) cos (φ + φ ) ∓ cos (φ − φ ) , (7.250)
2 p p
m=0

where, as before, (
1/2 m=0,
αm =
1 m,0.
The negative sign between the two terms inside the square brackets is used for the TM solution
and the positive sign is used for the TE solution. We have used the following identities to arrive
at Eq. (7.250):
       
m ′ m ′ m m ′
cos (φ + φ ) − cos (φ − φ ) = 2 sin φ sin φ
p p p p
and
       
m ′ m ′ m m ′
cos (φ + φ ) + cos (φ − φ ) = 2 cos φ cos φ .
p p p p
Introducing a new variable
γ± = φ ± φ′ , (7.251)
Eq. (7.250) can be written as
TM
S TE = s(kρ ρ , γ + ) ∓ s(kρ ρ , γ − ) , (7.252)
7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic Wedge 475

where
∞
X  
αm m/p m
s(kρ ρ , γ ) = (−i) Jm/p (kρ ρ ) cos γ
2 p
m=0
X∞ X1 ∞
1 m m
= Jm/p (kρ ρ )ei p [γ −(π /2)] + J (kρ ρ )e−i p [γ +(π /2)] (7.253)
4 4 m/p
m=1 m=0

To evaluate Eq. (7.253) asymptotically, we will make use of the integral representation of
Jm/p (kρ ρ ) given by Eq. (7.248):
Z
1 m
Jm/p (kρ ρ ) = ei[kρ ρ cos η + p (η −π /2)] d η , (7.254)
2π C3

where the contour of integration C3 is the one shown in Fig. 7-29. Also noting that

Jν (kρ ρ ) = (−1)ν J−ν (kρ ρ ) ,

Eq. (7.254) can also be written as

Z
1 m
Jm/p (kρ ρ ) = ei[kρ ρ cos η − p (η +π /2)] d η . (7.255)
2π C3′

Substituting Eqs. (7.254) and (7.255) into the conjugate of Eq. (7.253) and changing the order
of summation and integration leads to
Z ∞
X Z ∞
X
1 m
(γ + η ) 1 m
(γ + η )
∗
S (kρ ρ , γ ) = ikρ ρ cos η
e ei p dη + eikρ ρ cos η e−i p dη .
8π C3 8π C3′
m=1 m=0
(7.256)
In the first integral of Eq. (7.256) on contour C3 , η has a positive imaginary part, and as a
result the exponential terms in the summation have exponentially decaying terms. Hence
∞
X m
(γ + η ) ei(γ +η )/p 1
ei p = γ η
=− . (7.257)
m=1
1−e i( + )/p 1 − e γ +η )/p
−i(

In the second integral of Eq. (7.256) on contour C3′ , η has a negative imaginary part, and
therefore the exponential terms in the summation have exponentially decaying terms as well.
Therefore ∞
X m 1
e−i p (γ +η ) = . (7.258)
m=0
1 − e γ +η )/p
−i(

Substituting Eqs. (7.257) and (7.258) into Eq. (7.256) gives

Z
∗ 1 eikρ ρ cos η
S (kρ ρ ), γ ) = dη . (7.259)
8π C3′ −C3 1 − e−i(γ +η )/p
476 Chapter 7 Cylindrical Wave Functions and Their Applications

The integration path is along C3′ and in the opposite direction of C3 . It is also noted that
    
γ +η γ +η
1 ei( γ + η )/(2p) 1 cos 2p + i sin 2p
−i( γ + η )/p
= i(γ +η )/(2p) −i( γ + η )/(2p)
=    
1−e e −e 2i sin γ +η 2p
 
1 γ +η 1
= cot + . (7.260)
2i 2p 2
Recognizing Z
1 ikρ ρ cos η
e d η = J0 (kρ ρ ) − J0 (kρ ρ ) = 0 , (7.261)
C3′ −C3 2π
and in view Eqs. (7.260) and (7.261), Eq. (7.259) can be written as
Z  
1 γ +η
S∗ (kρ ρ , γ ) = cot eikρ ρ cos η d η . (7.262)
16π i C3′ −C3 2p

According to Cauchy’s theorem,

I   X
γ +η
cot eikρ ρ cos η d η = 2π i Rn , (7.263)
C 2p n

where C is a closed contour composed of C3′ , C2 , −C3 , and C2′ . Here R represents the residue
of the integrand evaluated at its poles:
   
γ +η ikρ ρ cos η
Rn = lim (η − ηn ) cot e .
η →ηn 2p

Poles of cot[(γ + η )/2p] occur at

γ + ηn
= nπ , n = 0, ±1, ±2, . . . ,
2p
from which ηn can be obtained:
ηn = 2pnπ − γ .
Noting that  
γ +η
cos 2p
lim (η − ηn )   eikρ ρ cos η = 2peikρ ρ cos(2pnπ −γ ) , (7.264)
η →ηn γ +η
sin 2p

the poles captured by the closed contour are those for which

−π ≤ 2pnπ − γ ≤ π ,

or equivalently,
|2pnπ − γ | ≤ π . (7.265)
7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic Wedge 477

Therefore
Z  
1 γ +η 1 X
∗
S (kρ ρ , γ ) + cot eikρ ρ cos η d η = Rn
16π i C2 −C2′ 2p 8 n
p X ikρ ρ cos(2pnπ −γ )
= e U (n − |2pnπ − γ |) , (7.266)
4 n

where U (·) is the unit step function, which takes a value of unity when its argument is positive
and zero when its argument is negative. The right-hand side of Eq. (7.266) represents plane
waves corresponding to the incident wave and reflected waves from the faces of the metallic
wedge. For n = 0 and γ = φ − φ ′ , the conjugate of the corresponding term for the electric field
given by Eq. (7.248) takes the following form:
− ′ ′
Ez0 = sin θ ′ E0 eik cos θ z e−ikρ ρ cos(φ −φ ) U (π − |φ − φ ′ |) .

This represents the incident wave and is present only in the region where φ < π + φ ′ . As
shown in Fig. 7-33, the half-plane φ = π + φ ′ represents the shadow boundary. Basically, the
wedge shadows the incident wave for φ > π + φ .
Now, examining the case of n = 0 and γ = φ + φ ′ gives another contribution of the poles
of the integrand:
+ ′ ′
Ez0 = − sin θ ′ E0 eik cos θ z e−ikρ ρ cos(φ +φ ) U (π − |φ + φ ′ |) . (7.267)

This represents the reflected wave from the top face of the wedge, and it only exists in the
region where φ < π − φ ′ . The plane φ = π − φ ′ represents the reflection boundary. Finally,
we need to consider the case of n = 1 and γ = φ + φ ′ . For this case,
+ ′ ′
Ez1 = − sin θ ′ E0 eik cos θ z e−ikρ ρ cos(φ +φ −2pπ ) U [(φ + φ ′ ) − (2p − 1)π ] , (7.268)

which represents a reflected wave from the lower face of the wedge. This reflected wave exists
only when
φ ≥ (2p − 1)π − φ ′ . (7.269)
Note that in this case φ ′ > φ − α = (p − 1)π . The plane φ = (2p − 1)π − φ ′ also represents
the reflection boundary for the lower face of the wedge. In general, the contribution from the
residues predicts the incident wave and either a reflected ray from the top surface or a reflected
ray from the bottom face, depending on the values of (φ + φ ′ ).
Next we consider the contribution from the integral in Eq. (7.266). Consider
Z  
1 γ +η
IC2 ,C2′(ρ , γ ) = − cot eikρ ρη d η . (7.270)
16π i C2 ,C2′ 2p
478 Chapter 7 Cylindrical Wave Functions and Their Applications

Figure 7-33: A metallic wedge with interior angle α illuminated by a plane wave at an angle φ ′ .
Also shown are the shadow boundary condition and reflection boundary.
7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic Wedge 479

This integral can be evaluated approximately for kρ ρ ≫ 1 using the saddle-point method. The
saddle point of the integrand can be obtained from the solution of
d
(i cos η ) = 0 ,
dη
which are found to be ηs = ±π . The positive solution lies on C2 and the negative solution lies
on C2′ . The second derivative of the exponent evaluated at the saddle point is given by

d2 
2
(i cos η ) =i.
dη η =±π

In conditions in which the saddle point is away from the poles,

"s  #
1 2π −i(kρ ρ +π /4) γ ±π
IC2 ,C2′ = − e cot .
16π i kρ ρ 2p

Hence
s     
1 2π −i(kρ ρ +π /4) γ +π γ −π
IC2 − IC2′ = − e cot − cot . (7.271)
16π i kρ ρ 2p 2p

It further is noted that

   
    sin γ2p
−π
− γ2p
+π
2 sin πp
γ +π γ −π
cot − cot =     =−    .
2p 2p sin γ2p
+π
sin γ2p−π
cos πp − cos γp

The diffracted components of Eq. (7.252), excluding the geometric optics term obtained from
the contributions from the poles, are given by
s  
TM
TE i 2π i(kρ ρ +π /4) π
Sd = e sin
8π kρ ρ p
 
 1 1 
·
    ∓     . (7.272)
π φ +φ ′ π φ −φ 
′
cos − cos cos − cos
p π p p
480 Chapter 7 Cylindrical Wave Functions and Their Applications

Referring to Eq. (7.248), the diffracted z-component of the electric field is given by
 
π
i(k
e ρ ρ − π /4) sin p
d ′ ik cos θ ′ z
Ez = p sin θ e E0
2π kρ ρ p
 
 1 1 
·
    −     . (7.273)
π φ −φ ′ π φ +φ 
′
cos − cos cos − cos
p p p p
For the TE case, a similar expression for the z-component of the magnetic field can be
obtained. It is noted that when p = 1 (corresponding to α = π , i.e., a flat plane), the diffracted
field is zero as expected. It is also interesting to note that when p = 1/M for some integer M,
again the diffracted field vanishes (M = 2, 3, 4, . . . ). These correspond to interior wedges with
angles of π2 , π3 , π4 , . . . . The diffracted field becomes indeterminate for values of φ − φ ′ = π ,
φ + φ ′ = π , and 2pπ − (φ + φ ′ ) = π near shadow and reflection boundary conditions.
In summary, the z-components of the total electric and magnetic fields can be written as

Ezt = Ezd + EzGO (7.274a)

and
Hzt = Hzd + HzGO , (7.274b)

where EzGO and HzGO are geometric optics terms composed of the incident and reflected fields.
Oftentimes the diffracted field components are expressed in terms of diffraction coefficients
defined by
′ ρ +cos θ ′ z)
eik(sin θ
Ezd (ρ , φ , z) = E0 sin θ ′ √ Ds (ρ , φ , φ ′ ) (7.275a)
ρ
and
′
eik(sin θ ρ +cos θ ′ z)
Hzd (ρ , φ , z) = H0 sin θ ′ √ Dh (ρ , φ , φ ′ ) . (7.275b)
ρ

The diffraction coefficients for soft boundary conditions (TM) and for hard boundary
conditions (TE) are given by
 
π
e−i π /4 sin p
′
D s (ρ , φ , φ ) = √
h 2π k sin θ ′ p
 
 1 1 
·
    ∓     . (7.276)
π φ −φ ′ π φ +φ 
′
cos − cos cos − cos
p p p p
As an example, the exact solution for the total field given by Eq. (7.248) is computed and the
7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic Wedge 481

5
Exact solution
4
Asymptotic solution

3
Ds
(a) TM
2

0
0 50 100 150 200 250 300 350

Exact solution
4 Asymptotic solution

3
Dh
(b) TE
2

0
0 50 100 150 200 250 300 350

Figure 7-34: Diffraction coefficient for a metallic wedge with α = π /6 when illuminated by
(a) a TM plane wave and (b) a TE plane wave, both for φ ′ = π /4 and θ ′ = π /2.

geometric optics terms are subtracted. The resultant field is the diffracted component from
which the diffraction coefficients for the TM case are computed. For this example, the wedge
angle is chosen to be α = π /6 and the angles specifying the direction of incidence are chosen
to be π ′ = π /4 and θ ′ = π /2. The diffraction coefficients based on the asymptotic evaluation
and the exact solution (using ρ = 10λ ) are compared in Fig. 7-34, where, except at shadow
and reflection boundaries, excellent agreement exists between the two solutions.
Of particular interest is the diffraction from half-metallic planes. Half-planes can be
modeled as a wedge with angle α = 0, for which p = 2. The diffraction coefficient for half-
482 Chapter 7 Cylindrical Wave Functions and Their Applications

planes can be obtained from Eq. (7.276) with p = 2, and is given by

 
half−plane ′ i2 sin(φ ′ /2) sin(φ /2)
Ds (φ , φ ) = √ (7.277a)
2π k sin θ ′ cos φ + cos φ ′
and
 
half−plane −i2 cos(φ ′ /2) cos(φ /2)
Dh (φ , φ ′ ) = √ . (7.277b)
2π k sin θ ′ cos φ + cos φ ′

Example 7-3: Edge Diffraction

Calculate the backscatter echo width of a half-plane illuminated by a TM incident wave.

Solution: Let us consider a TM plane wave:

′ x+sin φ ′ y)
E i = E0 ẑ e−ik0 (cos φ .

The diffracted or scattered field is given by

eikρ
E d = ẑ E0 √ Ds (φ , φ ′ ) .
ρ

In the case of backscatter, φ = φ ′ and the diffraction coefficient is simplified to

 
i 1
Ds (φ , φ ) = √ −1 .
2 2π k cos φ
The backscatter echo width is defined as
 2
|E d |2 λ 1
σ2D = lim 2πρ i 2 = −1 .
ρ →∞ |E | 8π cos φ

The echo width assumes its minimum at φ = 0 (σ2D (0) = 0) and is singular at φ = π /2 and
φ = 3π /2 due to the inaccuracy of the diffraction coefficient at the reflection boundary. The
echo width at edge-on incidence (φ = π ) is

λ
σ2D (π ) = .
2π
7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic Wedge 483

Example 7-4: Diffraction from a Slit in a Metallic Screen

Consider a single slit of width w in a planar metallic plate placed in the x–z plane, as shown
in Fig. 7-35. Suppose the slit is being illuminated by a plane wave given by
′ x+sin φ ′ y)
Ei = ẑ e−ik0 (cos φ . (7.278)

Find the diffracted field in the far-field region by ignoring multiple diffraction between the
two edges of the slit. Assign coordinate (ρ , φ ) to the observation point.

k̂i

−w/2 w/2

Figure 7-35: The geometry of a single metallic slit illuminated by a TM plane wave.

Solution: The diffracted field from the half-plane PEC 1 can be evaluated using

(1) (1) eikρ1 (1)

Ed (ρ , φ , φ ′ ) = E0 √ Ds , (7.279)
ρ1
(1)
where E0 is the incident field at the edge of the half-plane PEC 1 given by
′
E0i = e−i(k0 w/2) cos φ .

The diffraction coefficient can be calculated from Eq. (7.277a) noting that

φ1′ = φ ′ (7.280a)

and in the far-field region

φ1 = φ (7.280b)
484 Chapter 7 Cylindrical Wave Functions and Their Applications

and
w
ρ1 ≈ ρ − cos φ . (7.280c)
2
Similarly, the diffracted field from half-plane PEC 2 can be evaluated using

(2) (2) eikρ2 (2)

Ed (ρ , φ , φ ′ ) = E0 √ Ds . (7.281)
ρ2

For this half-plane, the incident field at the edge is

(2) ′
E0 = ei(k0 w/2) cos φ

and

φ2 = π − φ , (7.282a)
φ2′= π −φ , ′
(7.282b)
w
ρ2 = ρ + cos φ . (7.282c)
2
Using Eq. (7.280) in Eq. (7.277a), we have

(1) i2 sin(φ ′ /2) sin(φ /2)

Ds = √ . (7.283)
2π k cos φ + cos φ ′
Also, using Eq. (7.282) in Eq. (7.277a), we get

(2) −i2 cos(φ ′ /2) cos(φ /2)

Ds = √ . (7.284)
2π k cos φ + cos φ ′
The total diffracted field can now be calculated from
eikρ  (1) ′ (2) ′

Ed = ẑ √ Ds e−i(k0 w/2)(cos φ +cos φ ) + Ds ei(k0 w/2)(cos φ +cos φ )
ρ
−2ieikρ
= ẑ p
2π kρ
′ ′
!
cos(φ ′ /2) cos(φ /2)ei(k0 w/2)(cos φ +cos φ ) − sin(φ ′ /2) sin(φ /2)e−i(k0 w/2)(cos φ +cos φ )
· .
cos φ + cos φ ′
(7.285)
7-7 Asymptotic Evaluation of a Field Diffracted by a Metallic Wedge 485

It is interesting to note that when φ = π − φ ′ , the expression for the total diffracted field at the
reflection boundary, as given by Eq. (7.285), remains finite. That is,
 
−2ieikρ  ′
 cos(φ /2) cos( φ /2) − sin(φ ′ /2) sin(φ /2) 

lim ′ Ed = ẑ p    
φ →π −φ 2π kρ  φ + φ′ φ − φ′ 
2 cos cos
2 2
−ieikρ 1
= ẑ p .
2π kρ sin φ′
486 Chapter 7 Cylindrical Wave Functions and Their Applications

Summary
Concepts
• The method of separation of variables is applied to the Helmholtz equation in
cylindrical coordinates and the resulting solutions form the cylindrical wave functions.
• The z-dependent and φ -dependent solutions for the Helmholtz equation are harmonic
functions and the ρ -dependent solution is a Bessel function whose order depends on
the harmonic number of the φ -dependent function.
• If the domain of interest includes the entire range in φ (0 < φ ≤ 2π ), the order of the
Bessel functions has an integer value. Otherwise, the order of the Bessel functions must
be a noninteger.
• The solution to the wave equation consists of a linear combination of cylindrical wave
functions.
• A metallic waveguide whose surface coincides with the constant surfaces in a
cylindrical coordinate system, such as a circular or angular sector waveguide, supports
TE and TM modes whose cutoff frequencies are expressed in terms of zeros of the
Bessel functions or their derivatives.
• A segment of a cylindrical waveguide can be capped on both sides to form a cylindrical
cavity.
• A circular dieletric waveguide is shown to support hybrid TE and TM modes similar
to dielectric plate waveguides. These waveguides have a much lower attenuation rate at
optical frequencies and are known as optical fibers.
• Optical fibers can support pure TE or TM modes, but only in cases where the fields
have azimuthal symmetry (n = 0).
• Other types of waveguides supporting cylindrical wave functions with propagation in
radial and circumferential directions can be used to analyze practical problems, such as
an H-bent rectangular waveguide.
• Green’s functions for both interior cylindrical waveguide problems as well as exterior
problems such as metallic cylinders and wedges are derived and expressed in terms of
a series summation of appropriate cylidrical wave functions.
• The expression for the 2-D Green’s function when the source is displaced from the
origin can be expressed in terms of a Hankel function of zeroth order. When solving
the Green’s function directly, a series solution in terms of a summation of cylidrical
wave functions, referenced to the origin of the coordinate system, is also derived. The
comparison between the two solutions results in the addition theorem for the Hankel
function of the zeroth order.
• The integral representation for Bessel functions of arbitrary order is derived in
combination with the saddle-point method and used to find the asymptotic behavior
of the scattered field from metallic wedges in the far-field region.
SUMMARY 487

Important Equations

Bessel equation: "  2 #

d 2 R 1 dR ν
2
+ 2
+ 1− R=0
d ρe ρe d ρe ρe
Hankel functions of the first and second kind in terms of Bessel functions of the first
and second kind:
(1)
Hν (kρ ρ ) = Jν (kρ ρ ) + iNν (kρ ρ )
(2)
Hν (kρ ρ ) = Jν (kρ ρ ) − iNν (kρ ρ )

Solution of wave equation in terms of cylindrical wave functions:

+∞ Z
X +∞  
ψ (ρ , φ , z) = An (kz ) Jn (kρ ρ ) + Bn (kz ) Nn (kρ ρ ) einφ eikz z dkz
n=−∞ −∞

Definition of modified Bessel functions:

Jn (iαρ ) = (i)n In (αρ )

(1) 2 n+1
Hn (iαρ ) = (i) Kn (αρ )
π
Gamma function: Z ∞
Γ (x) = t x−1 e−t dt
0
Recurrence relations for Bessel functions of any kind:
2ν
Rν −1 + Rν +1 = Rν
ρ
Wronskian relationships:
2
Jν (ρ ) Nν′ (ρ ) − Jν′ (ρ ) Nν (ρ ) =
πρ

(1) ′ (1) i2
Jν (ρ ) Hν (ρ ) − Jν′ (ρ ) Hν (ρ ) =
πρ
488 Chapter 7 Cylindrical Wave Functions and Their Applications

Important Equations (continued)

Orthogonality of Bessel functions:

Z a 
xν p   xν q 
Jν ρ Jν ρ ρ dρ = 0
0 a a
Z a  ′   ′ 
xν p xν q
Jν ρ Jν ρ ρ dρ = 0
0 a a

The norm of Bessel functions:

Z a 
xν p  a2
Jν2 ρ ρ dρ = [Jν +1 (xν p )]2
0 a 2
Z a  ′  "  2 2 #
xν p a2 ν  2
Jν2
ρ ρ dρ = 1− ′ Jν (x′ν p )
0 a 2 xν p

Scalar potential function for circular waveguides:

(
x  sin(nφ )
np
ψ (ρ , φ ) = An Jn ρ (TM)
a cos(nφ )
 ′  (
xnp sin(nφ )
ψ (ρ , φ ) = An Jn ρ (TE)
a cos(nφ )

Cutoff frequency for circular waveguides:

TM(np) xnp
fc = √ (TM)
2π a µε
TE(np) x′np
fc = √ (TE)
2π a µε
SUMMARY 489

Important Equations (continued)

Attenuation rate in circular waveguides:

ωε a k
α TMnp = = r
 x 2  x 2 (TM)
σ δs βnp x2np np np
2
σ δs η a k −
a a
 s  ′ 2  ′ 2 
xnp xnp
 n2 k2 − 
1  a a 
α TEnp = "  2 #    ′ 2 + s  (TE)
x  ′ 2 
n  np xnp 
σ δs η ka 1 − ′ k 2−
xnp a a

Transcendental equations for optical fiber:

! !
J ′ (kI a) K ′ (kII a) J ′ (kI a) K ′ (kII a)
µd n ρ n ρ ε d n ρ n ρ
kρII + kρI kρII + kρI
µc Jn (kρI a) Kn (kρII a) εc Jn (kρI a) Kn (kρII a)

β 2 n2 [(kρI )2 + (kρII )2 ]2
=
a2 kc2 (kρI )2 (kρII )2
(kρI )2 + (kρII )2 = kd2 − kc2

Cutoff frequency for cylindrical cavity:

r
TM xnp 2  qπ 2
1
fnpq = √ +
2π µε a ℓ
s
 ′ 2  
TE 1 xnp qπ 2
fnpq = √ +
2π µε a ℓ

Quality factor for TM010 mode in a cylindrical cavity:

r
µ
σ δs x01
ε  σ δs η 1.202
Q̆ =  a =  a
2 1+ 2 1+
ℓ ℓ
490 Chapter 7 Cylindrical Wave Functions and Their Applications

Important Equations (continued)

Modal solution for a radial waveguide:


 nπ  qπ 
TM (1)
ψnq (ρ , φ , z) = Hnπ /α (kρ ρ ) sin φ cos z (TM)
α ℓ
n = 1, 2, 3, . . . , q = 0, 1, 2, . . .
 nπ   nπ 
TE (1)
ψnq = Hnπ /α (kρ ρ ) cos φ sin z (TE)
α ℓ
n = 0, 1, 2, . . . , q = 1, 2, . . .

Green’s function for an angular sector waveguide:

∞
!
X (k ρ ′ ) H (1) (k a)
π i (1) J ν ρ ν ρ
g< (ρ ρ′ ) =
ρ,ρ Jν (kρ ρ ) Hν (kρ ρ ′ ) − sin νφ sin νφ ′
φ0 Jν (kρ a)
n=1
∞ (1)
!
π i X (1) J (k
ν ρ ρ ) H ν (k ρ a)
g> (ρ ρ′ ) =
ρ,ρ Jν (kρ ρ ′ ) Hν (kρ ρ ) − sin νφ sin νφ ′
φ0 Jν (kρ a)
n=1

Hankel function in terms of plane waves:

p Z ∞ √2 2
(1) 1 1
2
H0 (kρ x + y ) =2 q ei(kx x+ kρ −kx |y|) dkx
π −∞ k2 − k2
ρ x

Addition theorem for Hankel function:

 +∞

 X (1) ′


 Hm (kρ ρ ′ ) Jm (kρ ρ ) eim(φ −φ ) ρ ≤ ρ′
(1)
ρ − ρ ′ |) = m=−∞
H0 (kρ |ρ +∞

 X (1) ′

 Jm (kρ ρ ′ ) Hm (kρ ρ ) eim(φ −φ ) ρ ≥ ρ′

m=−∞

Green’s function for a metallic cylinder:

ψt (ρ , φ ) =
 !
iβ z X+∞

 Ie Jm (kρ a) (1) (1) ′

 Hm (kρ ρ ) − Jm (kρ ρ ) Hm (kρ ρ ′ ) eim(φ −φ ) ρ ≤ ρ′

 4εω (1)
m=−∞ Hm (kρ a)
!

 Ie iβ z X+∞
J (k a)

 m ρ (1) (1)
Hm (kρ ρ ′ ) − Jm (kρ ρ ′ ) Hm (kρ ρ ) eim(φ −φ )
′
ρ ≥ ρ′

 4εω (1)
m=−∞ Hm (kρ a)
SUMMARY 491

Important Equations (continued)

Radar echo width:

 ∞ 2
4 X Jn (k0 sin θi a) 
 inφ 
σ2TM (φ ) =  (−1)n (1)
e  (TM)
k0 sin θin=−∞ Hn (k0 sin θi a) 
 +∞ 2
4 X Jn′ (k0 sin θi a) inφ 

σ2TE(φ ) =  (−1)n e  (TE)
k0 sin θi n=−∞ (1) ′ 
H (k sin θ a)
n 0 i

Integral representation of Bessel function of the first kind:

Z
(−i)n π ikρ ρ cos φ +inφ
Jn (kρ ρ ) = e dφ
2π −π

Integral representation of Hankel functions:

Z π /2−i∞
(1) e−iνπ /2
Hν (kρ ρ ) = eikρ ρ cos φ +iνφ d φ
π −π /2+i∞
Z 3π /2+i∞
(2) e−iνπ /2
Hν (kρ ρ ) = eikρ ρ cos φ +iνφ d φ
π π /2−i∞

Total field of a metallic wedge in response to a line source:

Ez (ρ , φ ) =
 ∞    

 −Ikρ2 eiβ z X (1) ′ m ′ m


 pωε Hm/p (kρ ρ ) Jm/p (kρ ρ ) sin φ sin φ ρ < ρ′
p p
m=1 (TM)
∞    

 −Ikρ2 eiβ z X ′ (1) m ′ m

 Jm/p ρ(k ρ ) H (k ρ ) sin φ sin φ ρ> ρ′
 pωε m/p ρ p p
m=1

Far-field plane-wave scattering from a metallic wedge:

∞
X    
4 sin θ ′ ik cos θ ′ z m m ′
Ezt = E0 e (−i)m/p
Jm/p (kρ ρ ) sin φ sin φ
p p p
m=1
 492 Chapter 7 Cylindrical Wave Functions and Their Applications

Important Equations (continued)

Diffraction coefficients for a wedge:

 
π
e−iπ /p sin p
′
D (ρ , φ , φ ) =
s √
h 2π k sin θ ′ p
 
 1 1 
·
     ∓     
π φ − φ′ π φ + φ′ 
cos − cos cos − cos
p p p p

Important Terms Provide definitions or explain the meaning of the following terms:
2-D Green’s function for external problem Hankel function of the first kind
(metallic circular cylinder and wedge) Hankel function of the second kind
addition theorem harmonic functions
angular-sector waveguide hybrid mode
Bessel equation integral representation of Bessel function
Bessel function of the first kind modified Bessel function of the first kind
Bessel function of the second kind modified Bessel function of the second kind
bistatic echo width multipole
circumferential waveguide optical fiber
cladding order of Bessel function
core orthogonality of Bessel functions
cylindrical cavity radial waveguide
cylindrical waveguide reflection boundary
cylindrical wave functions saddle point
diffraction coefficients for metallic wedge saddle-point technique
elementary wave functions shadow boundary
formal solution for 2-D Green’s function stationary phase point
Fourier representation of 2-D Green’s function steepest descent path
gamma function two-dimensional dipole
Green’s function for angular-sector waveguide Wronskian relationship
H-bent rectangular waveguide
PROBLEMS 493

PROBLEMS
7.1 Show that ψ (ρ , z) = ln(ρ ) eikz satisfies the scalar wave equation.

Circular Waveguides

7.2 Consider the coaxial waveguides shown in Fig. P7.2.

,ε ,ε

(a) (b)

Figure P7.2: Configuration of two coaxial waveguide structures with inner and outer radii of a
and b. For configuration (b) a metallic plate is connecting the inner and outer cylinders.

(a) Write down the scalar potential for the proper TM and TE modes inside the coaxial line
shown in Fig. P7.2(a).
(b) If a PEC plate is inserted between the inner and outer conductors at the position φ = 0,
as shown in Fig. P7.2(b), write down the TE and TM modes, and find the lowest TM
and the lowest TE modes.
7.3
(a) For a given radius a, what are the TE and TM modes for a circular waveguide with
angular extent 0 < φ < φ ′ ? (This structure is called the Pacman waveguide.) Give the
potentials only.
(b) Now let φ ′ = 270◦ and simplify the form of the potentials. Using information on the
roots of Bessel functions, find the lowest-order mode.
(c) Find the potentials and fields for TE and TM modes inside a waveguide with a
semicircular cross section (φ ′ = 180◦ ).
(d) Find the TM Green’s function inside the waveguide of part (c). This Green’s function
can be used to find the fields due to an arbitrary electric current source placed along the
axis of the waveguide. Find the scalar potential that exist in the waveguide due to the
following source:  π  ikc z
J(r, φ ) = I0 δ φ − e ρ̂ ρ.
2
7.4 Consider a circular metallic waveguide with radius b, as shown in Fig. P7.4. Inside the
waveguide there exists a concentric dielectric cylinder of radius a < b with parameters εd , µd .
494 Chapter 7 Cylindrical Wave Functions and Their Applications

Except for azimuthly symmetric modes (∂ /∂ φ = 0), TEz or TMz modes cannot be supported.

d , εd
,ε

Figure P7.4: A circular metallic waveguide loaded with a concentric dielectric cylinder having
permittivity εd and permeability εd .

(a) Find the transcendental equations and the cutoff frequencies for symmetric modes.
(b) Find the transcendental equations for the general case.
7.5 For a circular cylinder of radius a composed of a conductor with conductivity σ , show
that the attentuation rate can be evaluated from
1
α TE = p
σ δs η a 1 − ( fc / f )2
and
"  2 #
TM 1 n2 fc
α = p + .
1 − ( fc / f )2 (x′np )2 − n2 f

7.6 For the dielectric-covered conducting rod shown in Fig. P7.6, find the transcendental
equations for the TEz and TMz modes when ∂ /∂ φ = 0. Find transcendental equations for
the symmetric modes. Find also the transcendental equations for hybrid modes in the general
case.
7.7 Consider the example of an optical fiber shown in Fig. 7-8. Consider a frequency of
operation for which kc a = 25.
(a) How many modes can propagate in this fiber?
(b) If a portion of the power is carried by the first mode and a portion is carried by the
second mode, what is the phase delay between two signals over 100λc of propagation
(λc is the wavelength in the cladding).

Cylindrical Cavity

7.8 Consider a terminated by conductors at z = 0 and z = d, with d = 0.25a.

PROBLEMS 495

d , εd
,ε
y

Figure P7.6: The geometry of a metallic cylinder of radius b covered uniformly with a dielectric
layer with outer radius a and constitutive parameters εd and µd . This structure can support guided
surface waves.

(a) Find the three lowest-order modes for TE and TM waves, respectively, and identify the
overall lowest-order mode.
(b) For the lowest mode, write down the fields and calculate the quality factor Q for good-
conducting cavity walls. The quality factor is given by
ωW
Q= ,
Pd
where W is the total energy stored in the cavity and Pd is the power dissipated on the
walls, which can be calculated using
1
Pd = |JS |2 ,
2σ δS
where σ is the conductivity and dS is skin depth of the conducting material.
7.9 In Example 7-2, we considered the problem of a circular patch antenna. The operating
frequencies of this antenna were determined by finding the resonant frequencies of a
cylindrical cavity with metallic top and bottom surfaces and a perfect magnetic conductor
on its side. Repeat the problem for a patch antenna whose geometry is shown in Fig. P7.9.
Assume that the substrate thickness is h and the dielectric constant of the substrate is ε0 ε1 .
7.10 Consider a cylindrical cavity with radius a and height ℓ. This cavity is loaded by a
dielectric rod with constitutive parameters µd and εd and radius b concentric with the cylinder,
496 Chapter 7 Cylindrical Wave Functions and Their Applications

, ε εr

Figure P7.9: Geometry of a Pacman patch antenna over a substrate with height h and dielectric
constant ε0 εr .

as shown in Fig. P7.10. Find the transcendental equation from which the cavity resonant
frequencies can be obtained.

,ε d , εd l

Figure P7.10: A partially loaded cylindrical cavity. The resonant frequency is a function of the
dielectric parameters and its radius.

7.11 Consider a cylindrical cavity with radius a and height ℓ. This cavity is loaded by a
dielectric rod having radius a, height d, and constitutive parameters µd and εd , as shown in
Fig. P7.11. Find the transcendental equation from which the cavity resonant frequencies can
be obtained.

Radial Waveguide

7.12 Consider the radial waveguide shown in Fig. P7.12. This waveguide is partially
filled with a dielectric slab of thickness d having material parameters εd and µd . Find the
PROBLEMS 497

,ε
l
d , εd

Figure P7.11: A partially loaded cylindrical cavity. The resonant frequency is a function of the
height of the dielectric disc and its permittivity and permeability.

transcendental equations and the field expressions in this waveguide for both TMz and TEz
modes.

l y
d , εd

x
Figure P7.12: A dielectric-loaded waveguide. A dielectric slab is placed between two parallel
metallic plates.

7.13 Consider the dielectric radial waveguide shown in Fig. P7.13. The thickness of this
dielectric slab is d and its constitutive parameters are εd and µd .
(a) Find the transcendental equations.
(b) Find the field expressions for both TMz and TEz .

Multipole

7.14 Consider the quadrupole shown in Fig. P7.14. Find the corresponding Hertz vector
potential and the resulting electric and magnetic field.
498 Chapter 7 Cylindrical Wave Functions and Their Applications

d , εd
y

Figure P7.13: A radial dielectric waveguide, which is basically a dielectric slab of infinite
extent.

−I

+I x

−I

+I

Figure P7.14: A two-dimensional quadrupole along the y-axis. Note that the separation between
the line currents δ is much smaller than a wavelength.

7.15 Consider the octopole shown in Fig. P7.15. Find the corresponding Hertz vector
potential and the resulting fields. The line currents are equally spaced around a small cylinder.

Radiation and Scattering

7.16 Consider a uniform cylindrical z-directed current sheet of radius a, as shown in

Fig. P7.16. Find the electric field throughout the region and radiated power per unit length

J(ρ , φ ) = Js δ (ρ − a) ẑ .
PROBLEMS 499

−I
+I x

−I
−I +I

Figure P7.15: A two-dimensional octopole arranged symmetrically around the surface of a

fictitious small cylinder.

Js
,ε
y

Figure P7.16: The geometry of a uniform cylindrical current sheet.

7.17 Solve Problem 7.16 for a case where the surface current is an arbitrary function of φ :

J(ρ , φ ) = Js (φ ) δ (ρ − φ ) ẑ .

7.18 Consider a uniformly distributed cylindrical z-directed current sheet of radius a

concentric with a dielectric cylinder of radius b < a with constitutive parameters εd and µd ,
as shown in Fig. P7.18. Find the far-field electric field, assuming that

J(ρ , φ ) = Jz δ (ρ − a) ẑ .
500 Chapter 7 Cylindrical Wave Functions and Their Applications

Js
,ε y

d , εd
x

Figure P7.18: A uniform cylindrical current sheet around a dielectric cylinder of radius b and
parameters µd and εd .

7.19 For Problem 7.19 now consider a case where b > a.

7.20 Consider an infinite z-oriented circular metallic cylinder of radius a illuminated by

a TM-polarized plane wave whose propagation vector lies in the x–z plane, as shown in
Fig. P7.20.
(a) Compute and plot the backscattering radar cross section (RCS) of this cylinder as a
function of frequency over the corresponding wavelength range 0.05a ≤ λ ≤ 50a for
normal incidence (θi = 90◦ ).
(b) Compute and plot the RCS of the cylinder in the x–y plane for oblique incidence at
θi = 50◦ for λ = 0.5, λ = 5a, and λ = 50a.

7.21 Consider the two concentric dielectric cylinders shown in Fig. P7.21. The parameters
of the inner dielectric cylinder of radius a1 are ε1 and µ1 . The outer cylinder has a radius
a2 > a1 with constitutive parameters ε2 and µ2 . This object is illuminated by a plane wave
whose electric field is parallel to the cylinder axis. Find the electric field in each region and
the cylinder echo width.

7.22 Consider a half-plane metallic sheet on the positive x-axis (φ = 0) being illuminated
by a plane wave whose direction of propagation towards the edge is denoted by φ ′ . Assuming
the z-polarized incident plane wave intensity is E0 , show that the surface current density on
PROBLEMS 501

θi

k̂i Ei

Figure P7.20: A metallic cylinder illuminated by a TM-polarized plane wave at oblique

incidence.

i
k̂i Hy

1 , ε1
Ezi
2 , ε2

Figure P7.21: A two-layer concentric dielectric cylinder illuminated by a plane wave

propagating along the −x-axis.

the half plane is given by

∞
2iE0 X nφ ′
Jz = n(−i)n/2 Jn/2 (kx) sin .
ωµx 2
n=1
 8 Chapter
Spherical Wave Functions
and Their Applications

Chapter Contents Objectives

Overview, 503 Upon learning the material presented in this chapter, you
8-1 Wave Functions in the Spherical Coordinate should be able to:
System, 504
8-2 Wave Transformation to Spherical Wave 1. Solve boundary-value problems associated with
Functions, 533 wave equations and objects whose boundaries
8-3 Plane-Wave Scattering from Spheres, 542 coincide with constant surfaces in a spherical
8-4 Wave Propagation in a Conical coordinate system.
Waveguide, 553
8-5 Biconical Structures, 560 2. Understand the properties of spherical wave
8-6 Other Spherical Waveguides, 568 functions composed of special functions, such as
Chapter Summary, 571 Legendre and associated Legendre functions and
Problems, 579 spherical Hankel functions, and apply them to
solve propagation and scattering problems.

3. Determine the resonant frequency of metallic

cavities formed by constant coordinate surfaces
in a spherical coordinate system and those of a
dielectric sphere.

4. Express a plane wave in terms of a summation

of spherical wave functions.

5. Calculate the radar cross section of metallic and

dielectric spheres.

6. Characterize the fields, and associated wave

functions, in conical waveguides and in the
region between two cones sharing the same axis
(biconical structures).

502
 503

Overview

In this chapter we consider the solution of wave equations in a source-free, homogeneous

medium, in a spherical coordinate system. We start by showing that under the Lorenz gauge
condition, none of the spherical-coordinate components of the magnetic vector potential A
satisfy the wave equation. To circumvent this difficulty, a magnetic vector potential that has
only a radial component (Ar ) is considered, and a new gauge condition is introduced so that
a scalar potential defined by Ar (r, θ , φ )/r would indeed satisfy the scalar wave equation. It
is shown that such a choice renders a transverse (with respect to r̂) magnetic set of fields
(TM-to-r). Application of the duality principle leads to an electric vector potential that
has only a radial component (Amr ) and a set of fields TE-to-r. Superposition of TM-to-r
and TE-to-r can be used to generate any arbitrary set of fields. The solution to the scalar
wave equation for the scalar potential proceeds by applying the method of separation of
variables. It is shown that such a procedure can produce harmonic functions with respect
to the variable φ , Legendre and associated Legendre functions as regarding variable θ , and
spherical Bessel functions with respect to r. The type and order of such functions depend
on the spatial domain of interest. The properties of these functions are presented with some
detail to facilitate the choice of the proper solution for the problem at hand. Orthogonality
of spherical harmonic functions formed from the product of harmonic functions (sine or
cosine) and the corresponding associated Legendre functions are demonstrated. The spherical
wave functions are used for canonical problems associated with dielectric or metallic objects
whose boundaries coincide with constant coordinate surfaces in the spherical coordinate
system. The formulation is applied through examples to illustrate how to compute the resonant
frequencies and the corresponding fields in a spherical metallic cavity. For the dominant
mode of the resonance, an expression for the quality factor of a metallic spherical cavity
with finite conductivity is obtained as well. A similar approach is used to find the resonant
frequencies of two concentric metallic spheres, and the results are applied to estimate the
resonances that occur at very low frequencies in the atmosphere between the Earth’s ground
and its ionosphere. Also, spherical dielectric resonators are studied using the spherical
wave functions, and it is shown that, in general, the resonant frequencies of such objects
are complex even when the dielectric sphere and its surrounding dielectric material are
lossless. It is shown that certain higher-order modes can provide very high quality factors.
To study scattering problems from spherical objects, expansion of plane waves in terms
of the spherical wave function is presented. Also, the addition theorem for spherical wave
functions is presented. The solution for bistatic scattering from metallic and dielectric spheres
is considered and analytical solutions for the scattered fields and the radar cross section for
such scatterers are obtained. Then metallic conical structures are considered. For example,
modal expansion of fields inside a conical waveguide is presented and solutions for the order
of the associated Legendre function are obtained as a function of the cone angle and azimuthal
harmonic orders. Similarly, field modal expansions for biconical structures, made from two
metallic cones sharing a common axis, are considered and their applications as antennas are
considered. Finally, we examine other spherical waveguide structures formed by the region
between two conical surfaces and two constant φ -planes.
504 Chapter 8 Spherical Wave Functions and Their Applications

8-1 Wave Functions in the Spherical Coordinate System

In previous chapters, we examined how electromagnetic waves can be expanded in terms of
wave functions in Cartesian and cylindrical coordinate systems. It was shown that using a
constant coordinate direction, we can expand the fields in TE and TM modes. For spherical
structures, such an expansion does not lend itself to simple construction of solutions for
Maxwell’s equations. To arrive at an appropriate scalar potential, we need to revisit the
magnetic vector potential A and the scalar electric potential Φ. According to Eqs. (3.6) and
(3.7), we have the following relations between A and Φ:

∇ × ∇ × A − k2A − iω µε ∇Φ = µ J (8.1a)
and
ρ
∇2 Φ − iω ∇ · A = − . (8.1b)
ε
Previously we used the Lorenz gauge condition

∇ · A = iω µε Φ

so as to simplify Eq. (8.1a) into the standard form of the wave equation

∇2 A + k 2 A = − µ J ,

which can be decomposed into three separate scalar wave equations when using the Cartesian
coordinate system. The procedure, however, does not work in the spherical coordinate system
because the vector components of ∇2 A do not decompose into a Laplacian (∇2 ) for any vector
components of A. Let us assume that the magnetic vector potential and the excitation current
have only radial components, i.e., Aφ = Aθ = 0. In this case,

A = Ar r̂ and J = Jr r̂ . (8.2)

Substituting Eq. (8.2) into Eq. (8.1a), the following scalar equations are obtained:
   
1 ∂ ∂ 1 ∂2 2 ∂
2
sin θ + 2 2 2
+ k Ar = −iω µε Φ − µ Jr , (8.3a)
r sin θ ∂ θ ∂θ r sin θ ∂ φ ∂r
 
1 ∂ ∂
Ar − iω µε Φ = 0 , (8.3b)
r ∂θ ∂r
and
 
1 ∂ ∂
Ar − iω µε Φ = 0 . (8.3c)
r sin θ ∂ φ ∂ r
Equations (8.3b) and (8.3c) suggest the use of a different gauge condition, namely

∂
Ar = iω µε Φ . (8.4)
∂r
8-1 Wave Functions in the Spherical Coordinate System 505

Application of Eq. (8.4) leads to satisfactory conditions for Eqs. (8.3b) and (8.3c), and
Eq. (8.3a) reduces to
 2   
∂ 1 ∂ ∂ 1 ∂2 2
+ sin θ + + k Ar = −µ Jr . (8.5)
∂ r2 r2 sin θ ∂ θ ∂θ r2 sin2 θ ∂ φ 2
Noting that in the spherical coordinate system
   
1 ∂ ∂ 1 ∂ ∂ 1 ∂2
∇2 = 2 r2 + 2 sin θ + 2 2
r ∂r ∂r r sin θ ∂ θ ∂θ r sin θ ∂ φ 2
and   
1 ∂ 2 ∂ Ar 1 ∂2
r = Ar ,
r2 ∂ r ∂r r r ∂ r2
Eq. (8.5) can be represented by
Ar µ
(∇2 + k2 ) = − Jr , (8.6)
r r
which is a scalar wave equation for Ar /r. By applying the duality relations, a similar equation
for the radial component of the electric vector potential can be obtained and is given by
Amr ε Jmr
(∇2 + k2 ) =− . (8.7)
r r
The field quantities can now be obtained from scalar potentials ψ = Ar /r and ψm = Amr /r,
where both ψ and ψm are solutions to the wave equation. Recalling that

E = −∇Φ + iω A , (8.8)

and in view of Eq. (8.4) ,

−1 ∂
E= ∇ Ar + iω Ar r̂ , (8.9)
iω µε ∂ r
which can be expanded to provide the field components
 2 
−1 ∂ 2
Er = + k Ar , (8.10a)
iω µε ∂ r2
 2 
−1 ∂
Eθ = Ar , (8.10b)
iω µε r ∂ r ∂ θ
and
−1 1 ∂2
Eφ = Ar . (8.10c)
iω µε r sin θ ∂ r ∂ φ
506 Chapter 8 Spherical Wave Functions and Their Applications

Also, the magnetic field in terms of A is given by

1
H= ∇×A ,
µ
which, upon expansion, gives

1 ∂ Ar
Hθ = (8.11a)
µ r sin θ ∂ φ
and
1 ∂ Ar
Hφ = − . (8.11b)
µr ∂ θ
Using the duality relations, the field quantities generated from ψm = Amr /r are given by
 2 
−1 ∂ 2
Hr = + k Amr , (8.12a)
iω µε ∂ r2
−1 ∂2
Hθ = Amr , (8.12b)
iω µε r ∂ r ∂ θ
−1 1 ∂2
Hφ = Amr , (8.12c)
iω µε r sin θ ∂ r ∂ φ
−1 ∂ Amr
Eθ = , (8.12d)
ε r sin θ ∂ φ
and
1 ∂ Amr
Eφ = . (8.12e)
εr ∂ θ
Superposition must be used to combine Eqs. (8.10), (8.11), and (8.12) when both electric and
magnetic sources are present. Equations (8.10), (8.11), and (8.12) produce TM- and TE-to-r̂
modes.

8-1.1 Spherical Wave Functions

In this section, we consider the method of separation of variables for the solution of wave
equations in the spherical coordinate system. Consider the wave equation

(∇2 + k2 ) ψ (r, θ , φ ) = 0 , (8.13)

and assume that the solution can be expressed in terms of the product of three continuous and
differentiable functions:
ψ (r, θ , φ ) = R(r) Q(θ ) F(φ ) .
8-1 Wave Functions in the Spherical Coordinate System 507

By substituting this solution into Eq. (8.13) and then dividing both sides by RQF/r2 sin2 θ , it
can be shown that the process leads to
   
sin2 θ d 2 dR sin θ d dQ 1 d2F
r + sin θ + + k2 r2 sin2 θ = 0 . (8.14)
R dr dr Q dθ dθ F dφ 2

Here, only (1/F)(d 2 F/d φ 2 ) is a function of φ , and the other terms in Eq. (8.14) are not
functions of φ . Hence,
1 d2F
= −µ 2 , (8.15)
F dφ 2
for some constant µ . Substituting Eq. (8.15) into Eq. (8.14) and dividing the resulting equation
by sin2 θ yields
   
1 d 2 dR 1 d dQ µ2
r + sin θ + k2 r2 − 2 = 0 . (8.16)
R dr dr sin θ Q d θ dθ sin θ
Choosing ν (ν + 1) as the separation constant, we can show that
 
1 d dQ µ2
sin θ − 2 = −ν (ν + 1) (8.17)
sin θ Q d θ dθ sin θ
and  
1 d 2 dR
r − ν (ν + 1) + k2 r2 = 0 . (8.18)
R dr dr
Hence, the three differential equations for R, Q, and F are given by:
 
d 2 dR
r + [(kr)2 − ν (ν + 1)]R = 0 , (8.19a)
dr dr
   
1 d dQ µ2
sin θ + ν (ν + 1) − 2 Q=0, (8.19b)
sin θ d θ dθ sin θ
and
d2F
+ µ 2F = 0 . (8.19c)
dφ 2
The solution to Eq. (8.19c) consists of harmonic functions of the form

F = Aeiµφ + Be−iµφ .

As noted previously, if the range for φ includes the entire domain [0, 2π ], µ has to be an
integer for ψ to be a single-valued function.
The differential equation given by Eq. (8.19a) closely resembles a Bessel equation, and
its solutions, known as spherical Bessel functions , are expressed in terms of ordinary Bessel
functions:  π 1/2
R(kr) = Zn+1/2 (kr) , (8.20)
2kr
508 Chapter 8 Spherical Wave Functions and Their Applications

where Zn+1/2 (kr) represents an ordinary Bessel function of order (n+ 21 ). The spherical Bessel
functions behave in a manner similar to their counterpart; i.e., jn (kr) and nn (kr) represent
standing-wave-type solutions whose large argument expansions are given by
 
1 (n + 1)π
lim jn (kr) ≈ cos kr − (8.21a)
kr→∞ kr 2
and
 
1 (n + 1)π
lim nn (kr) ≈ sin kr − . (8.21b)
kr→∞ kr 2

In general, jn (kr) and nn (kr) can be expressed in terms of polynomials in 1/kr multiplied by
sin(kr) and cos(kr). Starting from

sin(kr) sin(kr) cos(kr)

j0 (kr) = , j1 (kr) = − , (8.22a)
kr (kr)2 (kr)
cos(kr) cos(kr) sin(kr)
n0 (kr) = − , n1 (kr) = − − , (8.22b)
kr (kr)2 (kr)
and using the recurrence formula

(2n + 1)
Rn+1 (kr) = Rn (kr) − Rn−1 (kr) , (8.23)
kr
all higher-order spherical Bessel functions can be obtained. The following also provides
recurrence relations for the derivatives of the spherical Bessel functions:
d 1
Rn (r) = [n Rn−1 (r) − (n + 1) Rn+1 ] , (8.24a)
dr 2n + 1
d n+1
(r Rn (r)) = rn+1 Rn−1 (r) , (8.24b)
dr
and
d −n
(r Rn (r)) = −r−n Rn+1 (r) . (8.24c)
dr
It can also be shown that
d n
Rn (r) = −Rn+1 (r) + Rn (r)
dr r
and
d n+1
Rn (r) = Rn−1 (r) − Rn (r) .
dr r
Similar to the definition of ordinary Hankel functions, the spherical Hankel functions are
defined as
(1)
hn (kr) = jn (kr) + inn (kr) (8.25a)
8-1 Wave Functions in the Spherical Coordinate System 509

1
n=0
n=1
n=2
0.5 n=3
n=4
jn (kr)
0

−0.5
0 5 10 15 20
kr

Figure 8-1: The spherical Bessel function of the first kind for the first five orders.

−1
n=0
n=1
−2 n=2
nn (kr)
n=3
−3 n=4

−4

−5
0 5 10 15 20
kr

Figure 8-2: The spherical Bessel function of the second kind for the first five orders.

and
(2)
hn (kr) = jn (kr) − inn (kr) . (8.25b)

Figures 8-1 and 8-2 show the spherical Bessel functions of the first kind and second kind for
different orders.
The spherical Hankel function of the first kind represents an outward-going spherical
wave and the spherical Hankel function of the second kind represents an incoming spherical
510 Chapter 8 Spherical Wave Functions and Their Applications

wave:

(1) ei[kr−π (n+1)/2] eikr

lim hn (kr) ≈ = (−i)n+1 (8.26a)
kr→∞ kr kr
and
(2) e−i[kr−π (n+1)/2] e−ikr
lim hn (kr) ≈ = (i)n+1 . (8.26b)
kr→∞ kr kr
The only spherical Bessel functions that have finite values at r = 0 are jn (kr), and therefore
these are the only admissible radial functions if the domain of the problem includes the origin.
The series solutions for the spherical Bessel
√ functions can be obtained from Eq. (7.23) by
noting that Γ (ν + 1) = ν Γ (ν ), Γ ( 12 ) = π , and using the duplication formula it can shown
that   √
1 π Γ (2ν )
Γ ν+ = (2ν −1) , (8.27)
2 2 Γ (ν )
which leads to ∞
X (−1)m (n + m)! 2m
n n
jn (r) = 2 r r . (8.28)
m!(2n + 2m + 1)!
m=0

Also, Nn+1/2 (r) = (−1)n−1 J−n−1/2 (r), and therefore

( n ∞
)
1 X Γ (2n − 2m + 1) 1 X (−1)m+n Γ (m − n)
nn (r) = − n n+1 r2m + r2m . (8.29)
2 r m! Γ (n − m + 1) 2 m! Γ (2m − 2n)
m=0 n=m+1

Another useful formula is the Wronskian of the spherical Bessel functions, which can be
directly obtained from the Wronskian of the Bessel functions:
1
jν (r) n′ν (r) − jν′ (r) nν (r) = (8.30a)
r2
and
(1)′ (1) i
jν (r) hν (r) − jν′ (r) hν (r) = . (8.30b)
r2

Equation (8.19b) is the generalized Legendre’s equation, and its solutions are known as
the associated Legendre functions. Solutions of Eq. (8.19b) can be expressed in terms of
µ
the associated Legendre functions of the first kind Pν (cos θ ) and the associated Legendre
µ
functions of the second kind denoted by Qν (cos θ ). If θ = 0 and θ = π are in the domain of
µ
the problem, the only admissible solution is Pν (cos θ ).
When µ = 0, the associated Legendre equation given by Eq. (8.19b) reduces to the
ordinary Legendre equation given by

d2y dy
(1 − x2 ) 2
− 2x + ν (ν + 1)y = 0 , (8.31)
dx dx
8-1 Wave Functions in the Spherical Coordinate System 511

where the substitution x = cos θ is used in Eq. (8.19b) with µ = 0. The range 0 ≤ θ ≤ π in the
spherical coordinate system corresponds to the range −1 ≤ x ≤ 1. The solution to Eq. (8.31)
can be obtained in terms of an infinite series given by

X N  
(−1)m (ν + m)! 1 − x m sin νπ
Pν (x) = −
(m!)2 (ν − m)! 2 π
m=0
X∞  
(m − 1 − ν )!(m + ν )! 1 − x m
· , (8.32)
(m!)2 2
m=N+1

where N is the largest integer smaller or equal to ν . It should be noted here that when ν is a
noninteger, Pν (x) and Pν (−x) become independent of each other and both satisfy Eq. (8.31).
However, if ν = n is an integer, the second term in Eq. (8.32) vanishes and the series solution
becomes finite. Using the binomial expansion
m  
X
m m
(1 − x) = (−x)k ,
k
k=0

and rearranging the terms,

N2
X (−1)m (2n − 2m)!
Pn (x) = xn−2m , (8.33)
2n m! (n − m)! (n − 2m)!
m=0

where N2 = n/2 or (n − 1)/2, depending on whether n is even or odd, respectively.

Consequently,
Pn (x) = (−1)n Pn (−x) . (8.34)
That is, the Legendre polynomials of odd degrees are odd functions and those of even degrees
are even functions. Equation (8.34) clearly indicates that Pn (x) and Pn (−x) are no longer
independent. Another solution for the Legendre equation, known as the Legendre function of
the second kind, is given as

π Pν (x) cos πν − Pν (−x)

Qν (x) = . (8.35)
2 sin νπ
This solution when ν approaches an integer exists and provides a solution independent of
Pn (x). However, it should be noted that Qn (x) become infinite at x = ±1 (corresponding to
θ = 0 and θ = π ). The series solution for Legendre functions of the second kind is given by
  Xn  
1 1+x (−1)m (n + m)! 1−x m
Qn (x) = Pn (x) ln − g(n) + g(m) , (8.36)
2 1−x (m!)2 (n − m)! 2
m=1

where
n
X 1
g(n) = .
k
k=1
512 Chapter 8 Spherical Wave Functions and Their Applications

A compact representation of the Legendre polynomials is given by

1 dn 2
Pn (x) = (x − 1)n ,
2n n! dxn
with

P0 (x) = 1 , (8.37a)
P1 (x) = x , (8.37b)
1 2
P2 (x) = 2 (3x − 1) , (8.37c)
and
P3 (x) = 1
2 (5x2 − 3x) . (8.37d)

Also,
 
1 1+x
Q0 (x) = ln , (8.38a)
2 1−x
 
x 1+x
Q1 (x) = ln −1 , (8.38b)
2 1−x
 
3x2 − 1 1+x 3x
Q2 (x) = ln − , (8.38c)
4 1−x 2
and
 
5x3 − 3x 1+x 5x2 2
Q3 (x) = ln − + . (8.38d)
4 1−x 2 3
The associated Legendre functions for integer values µ = m can be expressed in terms of the
Legendre functions as shown below:

d m Pn (x)
Pnm (x) = (−1)m (1 − x2 )m/2 (8.39a)
dxm
and
2 m/2 d m Qn (x)
Qm m
n (x) = (−1) (1 − x ) , (8.39b)
dxm
and they satisfy the following differential equation:
2  
2 d y dy m2
(1 − x ) 2 − 2x + n(n + 1) − y=0, (8.40)
dx dx 1 − x2

where y can be Pnm (x) or Qm m

n (x). We note from Eq. (8.39a) that Pn (x) = 0 for m > n. Figures
8-3 and 8-4 show the first few orders of Legendre functions of the first and second kind of
integer orders, respectively. Figure 8-5 shows the associated Legendre function of the first
kind and third order for m = 0, 1, 2, 3.
8-1 Wave Functions in the Spherical Coordinate System 513

1
P1
P2
P3
0.5 P4

Pn (x) 0

−0.5

−1
−1 −0.5 0 0.5 1
x

Figure 8-3: The first few orders of Legendre functions of the first kind of integer orders.

4
Q0

3 Q1
Q2

2 Q3

Qn (x) 1

−1

−2
−1 −0.5 0 0.5 1
x

Figure 8-4: The first few orders of Legendre function of the second kind of integer orders.
514 Chapter 8 Spherical Wave Functions and Their Applications

10

0
P3m (x)
P30
−5
P31

P32
−10
P33

−15
−1 −0.5 0 0.5 1
x

Figure 8-5: The associated Legendre function of the first kind and third order for m = 0, 1, 2, 3.

The general solution of scalar potential can be expressed in terms of the superposition of
elementary spherical Bessel functions:
XX
ψ (r, θ , φ ) = (an jn (kr) + bn nn (kr)) Pnm (cos θ ) eimφ , (8.41)
n m

for integer values of m and n. The superposition can also be expressed in terms of integrals in
ν and µ . The field quantities are expressed in terms of Ar and Amr in Eqs. (8.10) and (8.12),
and Ar and Amr are in turn expressed in terms of rψ and rψm . Sometimes, it is convenient to
introduce another kind of spherical Bessel function, which was first introduced by Schelkunoff
and defined as r
π kr
ẑn (kr) = krzn (kr) = Z (kr) . (8.42)
2 n+1/2
These spherical Bessel functions satisfy the condition
 2 
d n(n + 1)
+1− ẑn (kr) = 0 . (8.43)
d(kr)2 (kr)2

8-1.2 Orthogonality Properties

To represent an arbitrary wave function in a source-free region in terms of an elementary
spherical wave function similar to Eq. (8.41), orthogonality relationships must be established
among the elementary functions. For example, it can be shown that the Legendre polynomial
Pn (cos θ ) forms a complete orthogonal set in the θ -domain [0, π ]. Similarly, the two-
8-1 Wave Functions in the Spherical Coordinate System 515

dimensional elementary functions

o
Unm (θ , φ ) = Pnm (cos θ ) sin mφ (8.44a)
and
e
Unm (θ , φ ) = Pnm (cos θ ) cos mφ (8.44b)

form a complete orthogonal set over the domain θ ∈ [0, π ] and φ ∈ [0, 2π ] and are sometimes
referred to as spherical harmonic functions. These are orthogonal functions on the surface
of a unit sphere, and thus the orthogonality integral needs a weighting function (sin θ ).
To demonstrate orthogonality of functions Pn (cos θ ), known as zonal harmonics, we
consider two specific elementary wave functions:

ψ1 (r, θ ) = jn (kr) Pn (cos θ )

and
ψ2 (r, θ ) = jm (kr) Pm (cos θ ) ,

which are solutions to the wave equation (8.13) for µ = 0 and ν = n. Applying Green’s second
identity to ψ1 and ψ2 , we get
$  
2 2 ∂ ψ2 ∂ ψ1
(ψ1 ∇ ψ2 − ψ2 ∇ ψ1 ) dv = ψ1 − ψ2 ds .
S ∂n ∂n
Using Eq. (8.13), it can be shown that the left-hand side is zero. Considering surface S to be a
sphere of radius r, we have
Z 2π Z π  
∂ ψ2 ∂ ψ1
ψ1 − ψ2 ds = 0 . (8.45)
0 0 ∂n ∂n
Since on this surface the spherical Bessel functions are constant, it can be shown that
Z π
2 ′ ′
2π kr [ jn (kr) jm (kr) − jn (kr) jm (kr)] Pn (cos θ ) Pm (cos θ ) sin θ d θ = 0 .
0

It is now obvious that when n , m,

Z π
Pn (cos θ ) Pm (cos θ ) sin θ d θ = 0, m,n, (8.46)
0

which proves that Pn and Pm are orthogonal. Noting that

1 dn 2
Pn (x) = (x − 1)n ,
2n n! dxn
it can be shown that Z π
2
[Pn (cos θ )]2 sin θ d θ = . (8.47)
0 2n + 1
516 Chapter 8 Spherical Wave Functions and Their Applications

Note that, for any function f (θ ) defined in the domain θ ∈ [0, π ],

∞
X
f (θ ) = an Pn (cos θ ) ,
n=0

where Z π
2n + 1
an = f (θ ) Pn (cos θ ) sin θ d θ .
2 0
To show orthogonality of spherical harmonic functions we consider the following:

ψ1 = Unm (θ , φ ) jn (kr) (8.48a)

and
ψ2 = U pq (θ , φ ) j p (kr) . (8.48b)

Obviously ψ1 and ψ2 satisfy the wave equation (8.13). Again, using Green’s second identity
for a spherical surface, it can be shown that Eq. (8.45) is valid for the new functions ψ1 and
ψ2 given in Eq. (8.48). Here the superscript for the spherical harmonic functions Umn o or U e
mn
is suppressed to show an orthogonality relationship that is valid for both.
Direct substitution of Eq. (8.48) into Eq. (8.45) results in
Z 2π Z π
kr2 [ jn (kr) j′p (kr) − jn′ (kr) j p (kr)] Unm (θ , φ ) U pq (θ , φ ) sin θ d θ d φ = 0 .
0 0

The term outside the integral vanishes for n = p, and hence for n , p, we have
Z 2π Z π
Unm (θ , φ ) U pq (θ , φ ) sin θ d θ d φ = 0 , (8.49)
0 0

which is valid for both even and odd functions. Since

Z 2π
sin(mπ ) cos(qφ ) d φ = 0 ,
0

it follows that Z 2π Z π
o e
Unm (θ , φ ) U pq (θ , φ ) sin θ d θ d φ = 0 ,
0 0
independently of n, p, m or q. However, in cases where both functions in Eq. (8.49) are either
even or odd, and since
Z 2π Z 2π (
0 m,q,
sin(mφ ) sin(qφ ) d φ = cos(mφ ) cos(qφ ) d φ = (8.50)
0 0 π m=q,
8-1 Wave Functions in the Spherical Coordinate System 517

it follows that even when n = p, unless m = q, Eq. (8.49) will still vanish. For n = p and
m = q, it can be shown that
Z 2π Z π
Unm (θ , φ ) U pq (θ , φ ) sin θ d θ d φ
0 0

 4π

 m = 0 (only even functions),
 2n + 1
= (8.51)

 2π (n + m)!

 m,0.
(2n + 1)(n − m)!
Proof of this result is based on integration by parts using

d m Pn (x)
Pnm (x) = (−1)m (1 − x2 )m/2 . (8.52)
dxm
e as a function of θ and
Figure 8-6 shows the plot of a set of spherical harmonic functions U4m
φ for m = 0, 1, 2, 3, and 4.

8-1.3 Fourier-Legendre Expansion

In many problems for which a set of data or a function is known on the surface of a constant
sphere, the orthogonality of spherical harmonic functions can be used to express the data or
the function in terms of a series expansion. Usually such series can be truncated by keeping
only a small number of terms depending on the smoothness of the function with respect to θ
and φ . This way significant compression of data can be achieved. Consider a function f (θ , φ )
on the surface of a sphere. This function may be expanded in terms of spherical harmonic
functions given by
∞ X
X n
e o
f (θ , φ ) = [anm Unm (θ , φ ) + bnm Unm (θ , φ )] . (8.53)
n=0 m=0

According to Eq. (8.44), Eq. (8.53) may also be written as

∞ X
X n
f (θ , φ ) = [(anm cos mφ + bnm sin mφ ) Pnm (cos θ )] . (8.54)
n=0 m=0

To find the unknown coefficients amn and bmn , we can multiply both sides of Eq. (8.54)
e (θ , φ ) sin θ or U o (θ , φ ) sin θ and integrate with respect to θ and φ over 0 to π and
by U pq pq
0 to 2π , respectively. Using the orthogonality relations obtained in the previous section, the
following expressions for anm and bnm are obtained:
Z Z
2n + 1 2π π
an0 = f (θ , φ ) Pn (cos θ ) sin θ d θ , (8.55a)
4π 0 0
518 Chapter 8 Spherical Wave Functions and Their Applications

e e

z z
−

− −

x y x y
− − −
− − −

e
2 e

z z
− −
−
−

x y x y
− − −
− − −

z
−
−

x y
− −

Figure 8-6: The plot of spherical harmonic functions for n = 4 and m = 0, 1, 2, 3, and 4.

Z 2π Z π
2n + 1 (n − m)!
anm = f (θ , φ ) Pnm (cos θ ) cos mφ sin θ d θ d φ , (8.55b)
2π (n + m)! 0 0
and
8-1 Wave Functions in the Spherical Coordinate System 519

Z 2π Z π
2n + 1 (n − m)!
bnm = f (θ , φ ) Pnm (cos θ ) sin mφ sin θ d θ d φ . (8.55c)
2π (n + m)! 0 0

8-1.4 Orthogonality of Spherical Harmonic Functions for Vector Fields

Other useful orthogonality relations for spherical harmonic functions applicable to vector
fields can also be obtained. These relationships can be derived from the application of
the Lorentz reciprocity theorem for a source-free region. Consider a homogeneous region
bounded by a spherical surface of radius r0 . The Lorentz reciprocity for two sets of vector
fields generated by two different sources outside the spherical region can be written as
 a 

E (r) × Hb (r) − Eb (r) × Ha (r) · ds = 0 . (8.56)
S

Suppose the fields are generated from an electric current source that has only r̂ component.
This, as shown before, creates TM-to-r fields. The potentials in this case can be represented
by

Aar (r) = ĵn (kr) Unm (θ , φ )

and
Abr (r) = ĵq (kr) U pq (θ , φ ) .

Unm and U pq here can be either even or odd functions of φ . It is noted that ds in Eq. (8.56) is
along r̂, hence only the θ and φ components of the fields are sufficient in the calculation of
Eq. (8.56); that is,
Z 2π Z π
 a b 
(Eθ Hφ − Eφa Hbθ ) − (Ebθ Hφa − Eφb Haθ ) sin θ d θ d φ = 0 . (8.57)
0 0

Using Eqs. (8.10b) and (8.10c) together with Eqs. (8.11a) and (8.11b), Eq. (8.56) can be
written as
1  ′ 
2 2
ĵn (kr0 ) ĵ p (kr0 ) − ĵn (kr0 ) ĵ′p (kr0 )
iω µ ε r0
Z 2π Z π  e,o e,o e,o e,o 
∂ Unm ∂ U pq 1 ∂ Unm ∂ U pq
· sin θ + dθ dφ = 0 . (8.58)
0 0 ∂θ ∂θ sin θ ∂ φ ∂φ
If n , q, the orthogonality relation takes the following form:
Z 2π Z π  e,o e,o e,o e,o 
∂ Unm ∂ U pq 1 ∂ Unm ∂ U pq
I= sin θ + dθ dφ = 0 . (8.59)
0 0 ∂θ ∂θ sin θ ∂ φ ∂φ
Also, according to Eq. (8.50), if m , q, it can be easily shown that the integral will go to zero
even when n = p. If n = p and m = q, then using integration by parts and Eq. (8.51), it can be
520 Chapter 8 Spherical Wave Functions and Their Applications

shown that 

 4π n(n + 1)
 m=0,
2n + 1
I= (8.60)

 2π n(n + 1) (n + m)!
 m,0.
2n + 1 (n − m)!

Example 8-1: Expansion of Spherical Delta Function

Find the Fourier-Legendre expansion of f (θ , φ ) = sin1 θ δ (θ ) δ (φ ) and plot the approximate

representation of the function by truncating the summation over n to values of N = 5, 10, 15,
and 20.
Solution: Since f (θ , φ ) is a product of delta functions of θ and φ , the integrations given by
Eqs. (8.55a)–(8.55c) can be carried out easily, and the results are given by
2n + 1 2n + 1
an0 = Pn (1) = ,
4π 4π
2n + 1 (n − m)! m
anm = P (1) ,
2π (n + m)! n
and
bnm = 0 .

Hence,
N
XX n
1 2n + 1 (n − m)! m
δ (θ ) δ (φ ) ≈ αm P (1) ,
sin θ 4π (n + m)! n
n=0 m=0
where (
1 m=0,
αm =
2 m>0.
Figure 8-7(a) shows the normalized truncated expansion of spherical delta function in φ = 0
cut as a function of θ for four different values of truncation numbers N = 5, 10, 15, and 20.
Figure 8-7(b) shows the 3-D representation of the truncated spherical delta function for
N = 20.

8-1.5 The Spherical Cavity

To demonstrate the application of the TE-to-r and TM-to-r modes representation of field
quantities, let’s consider the problem of a spherical cavity. Figure 8-8 shows a metallic sphere
of radius a filled with a homogeneous material having permittivity ε and permeability µ . First,
considering TE-to-r modes,
(
cos mφ ,
Amr = ĵn (kr) Pnm (cos θ )
sin mφ .
8-1 Wave Functions in the Spherical Coordinate System 521

At ϕ = 0 degrees
0
N=5
N = 10
Normalized amplitude (dB) −5 N = 15
N = 20

−10

(a) −15

−20

−25

−30
0 50 100 150
θ (degrees)

3D pattern (dB), N = 20

z
(b)

−
−

0
y − − x

Figure 8-7: Plots of the truncated expansion of spherical delta function.

522 Chapter 8 Spherical Wave Functions and Their Applications

a
y

μ, ε

Figure 8-8: A metallic sphere filled with a homogeneous material having permittivity ε and
permeability µ .

The tangential electric fields are Eθ and Eφ , which must vanish on the metallic surface (r = a).
Referring to Eq. (8.12), the boundary condition is satisfied if

ĵn (ka) = 0 .

Denoting the zeros of Schelkunoff’s spherical Bessel functions ĵn (x) by xnp , where subscript
p represents the order of infinitely many roots of ĵn (x), it follows that

xnp n = 1, 2, 3, . . .
k=
a p = 1, 2, 3, . . .

At a resonance, the potential function Amr is given by

 
r m cos mφ ,
Amr = ĵn xnp Pn (cos θ )
a sin mφ .

For TM-to-r modes,


cos mφ
Ar = ĵn (kr) Pnm (cos θ ) m≤n.
sin mφ

According to Eq. (8.10), the tangential electric field on the sphere surface vanishes if

ĵ′n (ka) = 0 .
8-1 Wave Functions in the Spherical Coordinate System 523

Table 8-1: Ordered zeros of ĵn (x).

p n
↓ 1 2 3 4 5 6 7 8
1 4.493 5.763 6.988 8.183 9.356 10.513 11.657 12.791
2 7.725 9.095 10.417 11.706 12.967 14.207 15.431 16.641
3 10.904 12.323 13.698 15.040 16.355 17.648 18.923 20.182
4 14.066 15.515 16.924 18.301 19.653 20.983 22.295
5 17.221 18.689 20.122 21.525 22.905
6 20.371 21.854

Denoting the roots of jn′ (x) by x′np , k is obtained from

x′np
k= .
a
The resonant frequency for TE and TM modes are given respectively by
xnp
( f )TE
mnp = √ (8.61a)
2π a µε
and
x′np
( f )TM
mnp = √ . (8.61b)
2π a µε

Since m and n are independent, there are many degenerate modes. First a set of degenerate
modes exist through the choice of cos mφ (even mode) or sin mφ (odd mode), then for a given
n there exist (n + 1) degenerate modes corresponding to Pnm (cos θ ) for m ≤ n. That is, they all
have the same resonant frequency.
It is noted that n = 0 is not admissible since P0 is a constant function of θ , and there
is no variation with respect to φ either since m must be zero. According to Eqs. (8.10) and
(8.12), the corresponding magnetic and electric fields are zero, and thus no resonance can be
established.
Tables 8-1 and 8-2 provide the ordered zeros of the spherical Bessel functions and those
of the derivative of spherical Bessel functions, respectively. It is noted that the lowest-order
resonant frequency (dominant mode) is TMm,1,1 , for m = 0 or 1. The next resonant frequencies
are for modes TMm,2,1 , TEm,1,1 , TMm,3,1 , and so on.
It is interesting to examine the distribution of the fields of the dominant mode in a
spherical cavity. For TM0,1,1 the fields can be computed from Eqs. (8.10) and (8.11) using

Ar (r, θ , φ ) = ĵ1 (kr) P1 (cos θ ) = ĵ1 (kr) cos θ , (8.62)

524 Chapter 8 Spherical Wave Functions and Their Applications

Table 8-2: Ordered zeros of ĵ′n (x).

p n
↓ 1 2 3 4 5 6 7 8
1 2.744 3.870 4.973 6.062 7.140 8.211 9.275 10.335
2 6.117 7.443 8.722 9.968 11.189 12.391 13.579 14.763
3 9.317 10.713 12.064 13.380 14.670 15.939 17.190 18.425
4 12.486 13.921 15.314 16.674 18.009 19.221 20.615 21.894
5 15.644 17.103 18.524 19.915 21.281 22.626
6 18.796 20.272 21.714 23.128
7 21.946

and are given by

i n(n + 1)
Er = ĵ1 (kr) cos θ , (8.63a)
ω µε r2
−ik ′
Eθ = ĵ (kr) sin θ , (8.63b)
ω µε r 1
and
1
Hφ = ĵ1 (kr) sin θ , (8.63c)
µr
where we have used Eq. (8.43) to show that
 2 
d 2 n(n + 1)
+ k ĵn (kr) = ĵn (kr) . (8.64)
dr2 r2

In Eqs. (8.63a)–(8.63c), k = 2.744/a. The time-average stored energy inside the cavity can be
computed from $
1
W = 2Wm = µ |H(r, θ , φ )|2 dv . (8.65)
V 2
Noting that
Z π Z 1
4
sin3 θ d θ = (1 − x2 ) dx = , (8.66)
0 −1 3
it follows that Eq. (8.65) can be written as
Z a
4π 2
W= ĵ1 (kr) dr . (8.67)
3µ 0
8-1 Wave Functions in the Spherical Coordinate System 525

Using the integral identity

Z a
 2 a n 2 o
ĵ1 (kr) dr = ĵ1 (ka) − ĵ0 (ka) ĵ2 (ka) , (8.68)
0 2

and for ka = 2.744, Eq. (8.68) is evaluated to be 1.14/k. Hence,

4π × 1.14
W= .
3µ k
The ohmic loss on conducting walls of a cavity can also be computed from the tangential
magnetic field, and is given by

1
pL = |Hφ |2 ds
2σ δs S
Z Z
1 1  2 2π π 3
= ĵ1 (ka) sin θ d θ d φ
2σ δs µ 2 0 0
4π 1  2
= 2
ĵ1 (2.744) .
3σ δs µ
Noting that
sin(x)
ĵ1 (x) = − cos(x) ,
x
it is found that ĵ1 (2.744) = 1.063. The quality factor of the spherical cavity can be calculated
from
ω W ωσ δs µ × 1.14
Q̆ = = = σ δs η (1.088) ,
pL k × (1.063)2
where, as before, σ is the conductivity of the metal and δs is the skin depth.

Example 8-2: Ground Ionosphere Cavity

At low frequencies, the ground acts like a very good conductor and can be approximated as a
perfect conductor. The ionosphere, which contains free electrons (but with zero net charge),
can also be treated as a perfect conductor at low frequencies (below the plasma frequency).
Hence the ground and the ionosphere can be viewed as two concentric metallic spheres having
radii represented by Rg (for the ground surface) and Ri (for the ionosphere), respectively, as
shown in Fig. 8-9. Find the resonant modes of this cavity.
Solution: The modes that can exist in the spherical shell between the ground and the
ionosphere are either TE-to-r or TM-to-r. Considering TE-to-r modes first, the admissible
spherical wave functions that constitute Amr are given by
(
(1) cos mφ ,
Amr = [an ĵn (k0 r) + bn ĥn (k0 r)]Pnm (cos θ ) (8.69)
sin mφ .
526 Chapter 8 Spherical Wave Functions and Their Applications

Ionosphere modeled
ve as a PEC shell
wa g
M in
cy E ghtn
n li
ue
f r eq d by
w- ate
Lo ener
g

Conducting Earth
modeled as a PEC

Figure 8-9: Configuration of the Earth and ionosphere modeled as a pair of concentric metallic
spheres. The resonant fields in the cavity generated by lightning persist at the low frequencies
listed in Table 8-3.

The tangential electric fields are given by

1 ∂ Amr (r, θ , φ )
Eθ (r, θ , φ ) = − (8.70a)
ε0 r sin θ ∂θ
and
1 ∂ Amr (r, θ , φ )
Eφ (r, θ , φ ) = . (8.70b)
ε0 r ∂θ
The boundary condition mandates that

Eθ (Re , θ , φ ) = Eθ (Ri , θ , φ ) = 0 (8.71a)

8-1 Wave Functions in the Spherical Coordinate System 527

and
Eφ (Re , θ , φ ) = Eφ (Ri , θ , φ ) = 0 . (8.71b)

These boundary conditions are satisfied if

Amr (Re , θ , φ ) = Amr (Ri , θ , φ ) = 0 ,

or equivalently,
(1)
an ĵn (k0 Re ) + bn ĥn (k0 Re ) = 0 (8.72a)
and
(1)
an ĵn (k0 Ri ) + bn ĥn (k0 Ri ) = 0 . (8.72b)

A nontrivial solution can be found if the determinant of the coefficients of Eqs. (8.72a) and
(8.72b) are zero. This provides the following transcendental equation:
(n) (1) (n) (n) (1) (n)
ĵn (k0 Re ) ĥn (k0 Ri ) − ĵn (k0 Ri ) ĥn (k0 Re ) = 0 . (8.73)
(n) √
Distinct values of fn for which k0 = 2π fn µ0 ε0 satisfy Eq. (8.73) are the resonant
frequencies. In general, Eq. (8.73) must be solved numerically; however, noting that the
ionosphere’s height above the ground ranges from 80 km to 650 km and that this is
significantly smaller than the Earth’s radius, an approximate solution for the resonant
frequencies can be obtained. Recall that the Schelkunoff spherical Bessel functions satisfy
 2 
d n(n + 1)
+ 1 − ẑn (kr) = 0 . (8.74)
d(kr)2 (kr)2

For the problem at hand, Re ≤ r ≤ Ri and 1/(kr)2 can be approximated by 1/(kRe )2 . Hence,
Eq. (8.74) reduces to the ordinary second-order differential equation given by
 2  
d n(n + 1)
+ 1− Zen (kr) = 0 , (8.75)
d(kr)2 (kRe )2
which has a solution of the form
( s
sin(Un kr) , n(n + 1)
Zen (kr) = with Un = 1− (n)
. (8.76)
cos(Un kr) , (k0 Re )2

Therefore Eqs. (8.72a) and (8.72b) can be modified to

(n) (n)
an sin(Un k0 Re ) + bn cos(Un k0 Re ) = 0
and
(n) (n)
an sin(Un k0 Ri ) + bn cos(Un k0 Ri ) = 0 .
528 Chapter 8 Spherical Wave Functions and Their Applications

Again requiring that the determinant of the coefficient be zero, we have

(n) (n) (n)
sin(Un k0 Re ) cos(Un k0 Ri ) − cos(Un Re ) sin(Un k0 Ri ) = 0 ,

or
(n)
sin[Un k0 (Ri − Re)] = 0 . (8.77)
Denoting the height of the ionosphere by h = Ri − Re, we have

(n) ℓπ
k0 Un = , ℓ = 1, 2, . . . (8.78)
h
Substituting Eq. (8.76) into Eq. (8.78), we get
s  2
c n(n + 1) ℓπ
f n,ℓ = + . (8.79)
2π R2e h

There are (2n + 1) degenerate modes corresponding to m ≤ n and the choice of sin mφ or
cos mφ . The lowest TE-to-r mode corresponds to n = 0 and ℓ = 1:
01 c
fTE r
= . (8.80)
2h
For h = 100 km, fTE 01 = 1.5 kHz.
r
Lower-order modes can be excited by TM-to-r modes. These modes are generated by
(
(1) m cos mφ ,
Ar (r, θ , φ ) = [an ĵn (kr) + bn ĥn (kr)]Pn (cos θ ) (8.81)
sin mφ .

The tangential electric fields Eθ and Eφ can be derived from Eq. (8.16). These tangential fields
vanish at r = Re and r = Ri if
 
∂ A(r, θ , φ )  ∂ A(r, θ , φ ) 
 =  =0. (8.82)
∂r r=Re ∂r r=Ri

Using an approximation for the spherical Bessel functions similar to that in Eq. (8.76),
Eq. (8.82) is satisfied if
(n) (n) (n) (n)
anUn k0 cos(Un k0 Re ) − bnUn k0 sin(Un k0 Re ) = 0 (8.83a)
and
(n) (n) (n) (n)
anUn k0 cos(Un k0 Ri ) − bnUn k0 sin(Un k0 Ri ) = 0 . (8.83b)

Requiring the solution to be nontrivial provides a similar equation to Eq. (8.77):

(n)
sin[Un k0 (Ri − Re)] = 0 ,
8-1 Wave Functions in the Spherical Coordinate System 529

Table 8-3: Resonant frequencies for TM-to-r modes of the cavity between the ionosphere
and the Earth’s surface.
n 0 1 2 3 4 5 ···
f (Hz) 0 10.45 18.27 25.89 33.36 40.86 ···

from which we find

(n) ℓπ
k0 Un = , ℓ = 0, 1, 2, 3 .
h
We note that in this case ℓ = 0 is admissible, since the solution for Zen (kr), in view of
Eq. (8.83), is given by
(n)
Zen (kr) = C cos[k U (r − Re )]
0
for some constant C. The resonant frequency has the same form as Eq. (8.79). However, we
note that since h ≪ Re , the dominant set of modes corresponds to
ℓ p
f n,0 = n(n + 1) .
2π Re
Assuming that the Earth’s radius is Re = 6400 km, the resonant frequencies for TM-to-r
modes are given in Table 8-3. Note that these frequencies are independent of the ionosphere’s
height, which varies from time to time. For the modes in which the electric field has a radial
component, the electric field is a constant function of r.
Lightning is a frequent event around the Earth with a frequency of about 40–50
occurrences per second. At low frequencies, lightning is the source of background electric
noise whose spectrum peaks around the resonant frequencies reported in Table 8-3.

8-1.6 Dielectric Resonators

Microwave and millimeter-wave resonators, such as metallic cavities, are commonly used in
filter and oscillator designs as an alternative to resonators made from lumped-circuit elements
because of their high quality factor. Another type of resonator is a dielectric resonator
that is made entirely from low-loss dielectric material. Such a resonator can provide a
higher quality factor than their metallic counterpart, particularly at frequencies in the high
microwave and millimeter-wave bands. The spherical dielectric resonator is one of the few
geometries for which an exact analytical formulation can be obtained. Consider a dielectric
sphere of permittivity εs and permeability µs with radius a placed in a background medium
with constitutive parameters εb and µb , as shown in Fig. 8-10. We are seeking possible
electromagnetic field solutions to this source-free problem. The potential function in regions
inside and outside the dielectric sphere must satisfy the wave equation given by Eq. (8.13)
with the appropriate wave number for each region. As discussed earlier, a solution to the wave
equation given by Eq. (8.13) can be expressed in terms of TM-to-r and TE-to-r. Considering
first the TM-to-r solution, we recognize that the magnetic vector potential inside and outside
530 Chapter 8 Spherical Wave Functions and Their Applications

μb, εb

μs, εs a
y

Figure 8-10: Geometry of a dielectric sphere of radius a and constitutive parameters µs and εs
in a background medium with parameters µb and εb .

the sphere must take the following forms:

(
sin mφ ,
Asr (r, θ , φ ) = an ĵn (ks r) Pnm (cos θ ) r≤a (8.84a)
cos mφ ,
and
(
(1) sin mφ ,
Abr (r, θ , φ ) = bn ĥn (kb r) Pnm (cos θ ) r≥a, (8.84b)
cos mφ ,

where
√
ks = ω µs εs (8.85a)
and
√
kb = ω µb εb . (8.85b)

Choice of the sine or cosine functions in Eqs. (8.84a) and (8.84b) is arbitrary due to the
symmetry of the sphere. The boundary condition mandates the continuity of tangential electric
and magnetic fields. According to Eq. (8.10), tangential electric fields (Eθ and Eφ ) across the
boundary (r = a) are continuous if
 
1 ∂ Asr  1 ∂ Abr 
= . (8.86)
µs εs ∂ r r=a µb εb ∂ r r=a
8-1 Wave Functions in the Spherical Coordinate System 531

Similarly, according to Eq. (8.11), the continuity of the tangential magnetic fields is satisfied
if
1 s 1 b 
Ar |r=a = A . (8.87)
µs µb r r=a
Substituting Eqs. (8.84a) and (8.84b) into Eq. (8.86) renders
ks ′ kb (1)′
an ĵn (ks a) = bn ĥn (kb a) . (8.88)
µs εs µb εb
Also, substituting Eqs. (8.84a) and (8.84b) into Eq. (8.87) provides another equation:
1 1 (1)
an ĵn (ks a) = bn ĥn (kb a) . (8.89)
µs µb
Taking the ratios of Eqs. (8.88) and (8.89), the unknown coefficients an and bn are eliminated
and we arrive at the following transcendental equation:
r r (1)
εs ĵn (ks a) εb ĥn (kb a)
= . (TM modes) (8.90)
µs ĵn′ (ks a) µb ĥ(1)′
n (kb a)

The solution to Eq. (8.90) reveals all the resonances of the spherical dielectric resonator.
It turns out that the solution to this transcendental equation does not have any real roots. The
roots are, in general, complex resonant frequencies, and therefore, even for lossless dielectric
materials, only finite quality factors can be achieved. If the material properties of the dielectric
sphere or those of the background medium are lossy, the resonator will present a lower quality
factor. Note that for any value of n, all the m ≤ n modes are degenerate. The solution for
TE-to-r can be obtained using a similar procedure. Noting that TE waves are the dual of TM
waves, the transcendental equation for TE modes can be obtained from Eq. (8.90) by changing
ε to µ and µ to ε . That is,
r r (1)
µs ĵn (ks a) µb ĥn (kb a)
= . (TE modes) (8.91)
εs ĵn′ (ks a) εb ĥ(1)′
n (kb a)

To examine the behavior of the natural resonant frequencies of spherical dielectric

resonators, an example is considered: a dielectric sphere with radius a, permittivity εs = 9ε0 ,
and permeability µs = µ0 placed in a medium with parameters εb = ε0 and µb = µ0 . Solving
Eq. (8.90) for TM modes gives complex solutions for ks a. Figures 8-11(a) and 8-11(b) show
the TM and TE solutions for different values of n in the complex ks a plane. It is shown that
for certain high values of n, the imaginary part of (ks a) is quite small, which can result in a
high quality factor. Referring to Eq. (6.126), the quality factor for each mode can be obtained
from
ω′ Re[ksTEn ,TMn a]
QTEn ,TMn = = . (8.92)
2ω ′′ 2 Im[ksTMn ,TEn a]
Figures 8-12(a) and 8-12(b) show the quality factor of the spherical dielectric resonator
for TM and TE modes, respectively. It is shown that the quality factor can get as high as 108
532 Chapter 8 Spherical Wave Functions and Their Applications

27

24

21

18

15
[ksa]
12

3 TM1 TM3 TM5 TM7

TM2 TM4 TM6 TM8
0
0 −0.1 −0.2 −0.3 −0.4 −0.5 −0.6 −0.7 −0.8
[ksa]
(a) TM modes

27

24

21

18

15
[ksa]
12

3 TE1 TE3 TE5 TE7

TE2 TE4 TE6 TE8
0
0 −0.1 −0.2 −0.3 −0.4 −0.5 −0.6 −0.7 −0.8
[ksa]
(b) TE modes

Figure 8-11: The complex zeros of transcendental equations (a) Eq. (8.90) (for TM) and
(b) Eq. (8.91) (for TE) in the complex ks a plane for a dielectric sphere with εs = 9 and µs = µ0 in
a background medium with εs = ε0 and µs = µ0 . The zeros are obtained for different values of n,
noting that for each value of n there are many zeros.
8-1 Wave Functions in the Spherical Coordinate System 533

TM1 TM3 TM5 TM7

TM2 TM4 TM6 TM8
log10 (QTMn)

−
−
−

[ksa]
−
−

[ksa]
−
−
−

(a) TM modes

TE1 TE3 TE5 TE7

TE2 TE4 TE6 TE8
log10 (QTEn )

−
−
−

[ksa]
−
−
−
−

[ksa]
−

(b) TE modes

Figure 8-12: The quality factor of the spherical dielectric resonator for (a) TM and (b) TE
resonant modes calculated from Eq. (8.92) in the complex ks a plane. Note that the quality factor
increases for certain zeros for higher values of n.
534 Chapter 8 Spherical Wave Functions and Their Applications

Figure 8-13: The magnetic field distribution (normalized to the maximum value) of one of the
TM8 modes with the highest Q corresponding to m = 0.

for the modes considered in this plot. For higher values of n the quality factor can get higher,
but it should be noted that the quality factor is eventually limited by the loss tangent of the
dielectric constant of the sphere. Figure 8-13 shows the magnetic field distribution for TM8
mode inside the sphere and its surroundings for m = 0.

8-2 Wave Transformation to Spherical Wave Functions

To solve boundary value problems in a spherical coordinate system, it is often convenient to
express fields in terms of spherical wave functions. In this section, a few useful examples of
such transformations are considered.

8-2.1 Spherical Wave Function Expansion of a Plane Wave

In scattering problems, the excitation wave is often a plane wave. For scatterers, such as
spheres, cones, wedges, etc., whose geometries coincide with the coordinate surfaces, field
expansion in terms of spherical wave functions has been proven to be advantageous. Without
loss of generality, let’s consider a plane wave that is propagating along the z-axis and is given
by
Ei = E0 eikz .
8-2 Wave Transformation to Spherical Wave Functions 535

Expressing z = r cos θ in the spherical coordinate system, we note that the function is
independent of φ and is finite at the origin:
∞
X
ikr cos θ
e = an jn (kr) Pn (cos θ ) .
n=0

To find the unknown coefficients an , we make use of the orthogonality properties of Pn (cos θ )
given by Eqs. (8.46) and (8.47). It can easily be shown that
Z π
2an
jn (kr) = eikr cos θ Pn (cos θ ) sin θ d θ (8.93a)
2n + 1 0
and
X∞
(−1)m (n + m)!
n
jn (kr) = 2 (kr) n
(kr)2m . (8.93b)
m! (2n + 2m + 1)!
m=0

To reduce the calculation and evaluate an explicitly, the trick is to differentiate both sides of
Eq. (8.93a) n times with respect to r and then evaluate the result at r = 0. Using the series
expansion of jn (kr), it is easy to show that

dn  n
n 2 (n!)
2
j (kr) = k
drn
n  (2n + 1)!
r=0

and 
d n ikr cos θ 
e  = (i)n kn cosn θ .
drn r=0
Also noting that Z π
2n+1 (n!)2
cosn θ Pn (cos θ ) sin θ d θ = ,
0 (2n + 1)!
the coefficient an is given by
an = (i)n (2n + 1) .
Hence ∞
X
eikr cos θ = (i)n (2n + 1) jn (kr) Pn (cos θ ) . (8.94)
n=0
Conversely, using Eq. (8.93a),
Z π
(i)−n
jn (kr) = eikr cos θ Pn (cos θ ) sin θ d θ , (8.95)
2 0

which provides a useful formula for direct computation of jn (kr).

536 Chapter 8 Spherical Wave Functions and Their Applications

8-2.2 Addition Theorem for Spherical Wave Functions

The transformation of wave functions from a spherical coordinate system to another one is
encountered in some radiation or scattering problems. The solution to a wave equation for a
point source at the origin must satisfy

∇2 ψ + k2 ψ = −δ (r) . (8.96)

This solution was obtained in Section 4-1 and is referred to as the scalar Green’s function:

eikr
ψ (r, θ , φ ) = . (8.97)
4π r
Noting that
(1) eikr
h0 (kr) = ,
ikr
it follows that
ik (1)
ψ (r, θ , φ ) = h (kr) . (8.98)
4π 0
Now if the source point is moved to r′ , the potential function must satisfy

∇2 ψ (r, r) + k2 ψ (r, r′ ) = −δ (r − r′ ) , (8.99)

and the solution is found to be

ik (1)
ψ (r, r′ ) = h (k|r − r′ |) . (8.100)
4π 0
The solution to Eq. (8.99) can also be written in terms of a summation of spherical wave
functions with reference to the origin of the coordinate system and can be written as
XX
ψ (r, r′ ) = amn Rn (kr) Pnm (cos θ ) eimφ , (8.101)
n m

where Rn (kr) is a representation of a spherical Bessel function and amn ’s are coefficients that
depend on θ ′ and φ ′ . Noting that in a spherical coordinate system

δ (r − r′ )
δ (r − r′ ) = δ (θ − θ ′ ) δ (φ − φ ′ ) ,
r2 sin θ ′
it is possible to express the angular dependence of this function in terms of zonal harmonics:
∞ X
X n
δ (θ − θ ′ ) δ (φ − φ ′ ) = bnm Pnm (cos θ ) eimφ . (8.102)
n=0 m=−n
8-2 Wave Transformation to Spherical Wave Functions 537

The unknown coefficients bnm ’s can easily be obtained using the orthogonality properties of
the zonal harmonics and are given by

(2n + 1)(n − m)! ′

bnm = sin θ ′ Pnm (cos θ ′ ) e−imφ . (8.103)
4π (n + m)!
Hence,
∞ n
sin θ ′ X X (2n + 1)(n − m)! m ′ ′
δ (θ − θ ′ ) δ (φ − φ ′ ) = Pn (cos θ ′ ) Pnm (cos θ ) ei(m−m )φ .
4π m=−n
4π (n + m)!
n=0

By substituting Eqs. (8.101) and (8.102) into Eq. (8.99), it can be shown that
 
d 2 dRn (kr)
 
r + (kr)2 − n(n + 1) Rn (kr) = −δ (r − r′ ) , (8.104)
dr dr
with
(2n + 1)(n − m)! m ′
anm = Pn (cos θ ′ ) e−imφ . (8.105)
4π (n + m)!
The solution to Eq. (8.104) is an appropriate spherical Bessel function. For r < r′ , the function
must be regular, and for r > r′ the spherical Bessel function must be an outward-traveling
wave, or equivalently

Rn< (kr) = cn jn (kr) r ≤ r′ (8.106a)

and
(1)
Rn> (kr) = dn hn (kr) r ≥ r′ . (8.106b)

Rn (kr) must be continuous at r = r′ ; that is, Rn< (kr′ ) = Rn> (kr)) and r2 [dRn (kr)/dr] must
have a jump discontinuity at r = r′ . In other words,
 
dRn> (kr)  dRn< (kr)  1
 −  =− 2 . (8.107)
dr r=r ′ dr r=r ′ r′
These two equations can be used to find the unknowns from
(1)
cn jn (kr′ ) = dn hn (kr′ ) (8.108a)
and
(1)′ 1
k hn (kr′ ) dn − k jn′ (kr) cn = − . (8.108b)
r′ 2
Using the Wronskian relations,

(1)′ (1) i
jn (kr′ ) hn (kr′ ) − jn′ (kr′ ) hn (kr′ ) = , (8.109)
(kr′ )2
538 Chapter 8 Spherical Wave Functions and Their Applications

the unknown coefficients are found to be

(1)
cn = ik hn (kr′ ) (8.110a)
and
dn = ik jn (kr′ ) . (8.110b)

As a result, Eq. (8.101) can be written as

ψ (r, r′ ) =
 ∞ n

 X X (2n + 1)(n − m)! (1) ′


 hn (kr′ ) jn (kr) Pnm (cos θ ′ ) Pnm (cos θ ) eim(φ −φ ) r ≤ r′ ,
ik n=0 m=−n (n + m)!
∞ n
X X (2n + 1)(n − m)!
4π  (1) ′


 jn (kr′ ) hn (kr) Pnm (cos θ ′ ) Pnm (cos θ ) eim(φ −φ ) r ≥ r′ .
(n + m)!
n=0 m=−n
(8.111)

Comparing Eq. (8.111) with Eq. (8.104), the expression for the addition theorem is obtained. It
turns out that associated Legendre polynomials satisfy the same differential equation whether
m is positive or negative. In order for the orthogonality coefficients to work for all positive or
negative m,
(n − m)! m
Pn−m (x) = (−1)m P (x) .
(n + m)! n
With this definition it can be shown that
(2n + 1)(n − m)! m (2n + 1)(n − |m|!) |m| |m|
Pn (cos θ ′ ) Pnm (cos θ ) = Pn (cos θ ′ ) Pn (cos θ ) .
(n + m)! (n + |m|!)
As a result,
(1)
h0 (k|r, r′ |) =
 ∞ n
 XX (2n + 1)(n − m)! (1)′

 αm jn (kr) hn (kr) Pnm (cos θ ′ ) Pnm (cos θ ) cos[m(φ − φ ′ )]

 (n + m)!

 n=0 m=0


 r ≤ r′ ,
 X∞ X n

 (2n + 1)(n − m)! (1)


 αm hn (kr) jn′ (kr) Pnm (cos θ ′ ) Pnm (cos θ ) cos[m(φ − φ ′ )]

 (n + m)!

 n=0 m=0
r ≥ r′ ,

where αm is Neumann’s number defined as

(
1 m=0,
αm =
2 m>0.
8-3 Multipole Representation of Spherical Waves 539

z z z

I dl I dl
I dl
y y y
x x x

z z

I dl
I dl

y y
x x

Figure 8-14: (a) The geometry of a monopole, (b) a dipole, (c) a different dipole, (d) another
different dipole, and (e) a quadrupole.

8-3 Multipole Representation of Spherical Waves

In this section we show that spherical wave functions of order nm, such as
(
(1) cos mφ ,
ψnm (r) = hn (kr) Pnm (cos θ )
sin mφ ,

represent by the magnetic vector potential of an ordered 2n-element array of closely spaced
small Hertzian z-directed current elements flowing in the opposite direction of their adjacent
elements. An isolated z-directed Hertzian current element is referred to as monopole and two
adjacent z-directed current elements are referred to as dipole. Figure 8-14 shows the geometry
of a monopole, three different dipoles, and a quadrupole. Higher-multipole sources can be
realized by induction.
The fields generated by a small current filament were studied extensively in Chapter 3.
The magnetic vector potential for such a current was found to be

µ I dℓ eikr
Az (r) = (8.112)
4π r
540 Chapter 8 Spherical Wave Functions and Their Applications

Recalling that
(1) eikr
h0 (kr) = ,
ikr
it follows that
iµ kI dℓ (1)
Az (r) = h0 (kr) . (8.113)
4π
This corresponds to a spherical harmonic function of n = 0 order. Also, the resulting magnetic
field can be obtained from Eq. (3.123) and is given by
 
k2 I dℓ 1 eikr
H(r) = −i + sin θ φ̂
φ. (8.114)
4π kr r

Since the current is r-directed at θ = 0, the fields can also be generated from Ar (TM-to-r).
Using Eq. (8.11b), we have
1 ∂ Ar
Hφ (r) = − . (8.115)
µr ∂ θ
Comparing Eqs. (8.114) and (8.115), the expression for Ar is found to be
 
µ k2 I dℓ 1
Ar (r) = −i + eikr cos θ . (8.116)
4π kr
Using Eqs. (8.22a), (8.22b), and (8.24a), it can easily be shown that
 
(1) i
ĥ1 (kr) = − 1 + eikr . (8.117)
kr
Also, from Eq. (8.37b),
P1 (cos θ ) = cos θ ,
therefore, the expression for Ar given by Eq. (8.116) can also be expressed by

iµ k2 I dℓ (1)
Ar (r) = ĥ1 (kr) P1 (cos θ ) . (8.118)
4π
Note that in a source-free homogeneous region (with r > dℓ), in general, Ar (r) can be written
as (
X∞ X n
(1) m cos mφ ,
Ar (r) = anm ĥn (kr) Pn (cos θ ) (8.119)
n=0 m=0
sin m φ .
Comparing Eqs. (8.118) and (8.119), it is noted that

0 n , 1 and for all m ,
anm = iµ k2 I dℓ (8.120)
 n = 1, m = 0 .
4π
Now let’s consider Fig. 8-14(b) for a dipole arrangement along the z-axis. The overall
magnetic vector potential is the superposition of the vector potentials from each current source
8-3 Multipole Representation of Spherical Waves 541

separated by a small distance δ . Representing this magnetic vector potential by A2z z , it can
easily be shown that

A2z z = A1z (x, y, z − δ /2) − A1z (x, y, z + δ /2)

∂ A1z µ kI dℓ δ ∂ (1)
≈ −δ = −i h (kr) . (8.121)
∂z 4π ∂z 0
Noting that
∂ (1) (1) ′ ∂r (1) ′
h0 (kr) = k h0 (kr) = k h0 (kr) cos θ ,
∂z ∂z
and recognizing that
(1) ′ (1)
h0 (kr) = −h1 (kr) ,
the vector potential for the dipole becomes

iµ kI dℓ δ (1)
A2z z (r) = h1 (kr) P1 (cos θ ) . (8.122)
4π
This corresponds to a spherical wave function with n = 1 and m = 0. In a similar manner, the
magnetic vector potential for the dipole arrangement of Fig. 8-14(c) can be written as

∂ A1z −iµ kI dℓ δ ∂ (1)

A2z x (r) ≈ −δ = h (kr)
∂x 4π ∂x 0
−iµ kI dℓ δ (1) ′
= h0 (kr) sin θ cos φ . (8.123)
4π
Noting that
sin θ = P11 (cos θ ) ,
Eq. (8.123) can be written as

iµ kI dℓ δ (1)
A2z x (r) = h1 (kr) P11 (cos θ ) cos φ , (8.124)
4π
and this corresponds to a spherical wave function with n = 1 and m = 1 for an even function
of φ .
In a similar manner for the dipole configuration of Fig. 8-14(d), the vector potential is
given by
2 iµ kI dℓ δ (1)
Az y (r) = h1 (kr) P11 (cos θ ) sin φ , (8.125)
4π
which corresponds to a spherical wave function with n = 1 and m = 1 for an odd function
of φ .
For the quadrupole of Fig. 8-14(e), the magnetic vector potential can be obtained by
following a similar procedure. Again assuming the separations between the opposing current
elements are much smaller than the wavelength, the magnetic vector potential for the
542 Chapter 8 Spherical Wave Functions and Their Applications

quadrupole can be obtained from

4 ∂ 2 A1z (r)
Az y,z (r) = δ1 δ2 , (8.126)
∂y ∂z
or alternatively
4 ∂ A2z z
Az y,z (r) = −δ2 . (8.127)
∂y
Noting that

∂ h (1) i z ∂ h
(1)
i
(1) ∂ z
h1 (kr) cos θ = h1 (kr) + h1 (kr)
∂y r ∂y ∂y r
h i  
z ∂r ∂ (1) (1) −yz
= h (kr) + h1 (kr)
r ∂y ∂r 1 r3
 
yzk (1) ′ 1 (1)
= 2 h1 (kr) − h1 (kr) . (8.128)
r kr
Using Eq. (8.24c), it can be shown that

(1) ′ 1 (1) (1)

h1 (kr) − h (kr) = −h2 (kr) . (8.129)
kr 1
As a result from Eq. (8.127), the magnetic vector potential for the quadrupole of Fig. 8-14(e)
can be obtained from

4 iµ k2 I dℓ δ1 δ2 (1)
Az y,z (r) = h2 (kr) sin θ cos θ sin φ . (8.130)
4π
Using Eqs. (8.37c) and (8.39a), it can be shown that

P21 (cos θ ) = −3 sin θ cos θ . (8.131)

As a result, Eq. (8.130) can be written as

4 −iµ k2 I dℓ δ1 δ2 (1)
Az y,z (r) = h2 (kr) P21 (cos θ ) sin φ . (8.132)
12π
In Eq. (8.132) it is shown that the magnetic vector potential of a quadrupole placed in the y–z
plane is entirely presented by a spherical wave function with n = 2 and m = 1. In fact, any
spherical wave function of order n represents the magnetic vector potential of a multipole of
2n-element array of z-directed current filaments that are closely spaced.
It is interesting to note that if one sets the currents and separations of a collection of
multipoles to commensurate the coefficients obtained in Example 8-1 for a truncated expansin
of spherical delta function, such an electrically small array can, in principle, produce a highly
directive radiation pattern. However, in practice this is quite limited due to the mutual coupling
8-4 Plane-Wave Scattering from Spheres 543

among the elements and the need for exponentially growing level of currents for high-order
multipoles.

8-4 Plane-Wave Scattering from Spheres

Spherical objects are among the very few geometries for which analytical solutions can be
found for the field scattered in response to plane-wave excitation. In this section we consider
plane-wave scattering from metallic and dielectric spherical objects.

8-4.1 Scattering from a Metallic Sphere

We are now in a position to tackle the problem of electromagnetic wave scattering from
spherical objects. Consider a metallic sphere centered at the origin of a Cartesian coordinate
system and having a radius a, as shown in Fig. 8-15. Also, consider a plane wave propagating
in the +z direction. Without loss of generality, let’s assume the electric field is polarized along

a
θ

ϕ y

k̂

Hi

Ei

Figure 8-15: Geometry of a metallic sphere illuminated by a plane wave.

544 Chapter 8 Spherical Wave Functions and Their Applications

the x̂ direction. In this case,

Ei = E0 x̂eikz (8.133a)
and
E0 ikz
Hi = ŷe . (8.133b)
η
The incident field can be expressed in terms of a superposition of TE- and TM-to-r waves.
The r-component of the incident electric and magnetic fields are found to be
cos φ d ikr cos θ
Eir = Ei · r̂ = E0 cos φ sin θ eikr cos θ = E0 e
−ikr d θ
and
E0 sin φ d
Hir = Hi · r̂ = (eikr cos θ ) .
η −ikr d θ
Using Eq. (8.34), it can easily be shown that
d
Pn (cos θ ) = Pn1 (cos θ ) ,
dθ
and using the expansion given by Eq. (8.94), we have
∞
E0 cos φ X n
Eri = (i) (2n + 1) jˆn (kr) Pn1 (cos θ ) (8.134a)
−i(kr)2
n=1
and
∞
E0 sin φ X n
Hri = (i) (2n + 1) ĵn (kr) Pn1 (cos θ ) , (8.134b)
−i(kr)2 η
n=1

where ĵn (kr) is the Schelkunoff spherical Bessel function defined in Eq. (8.32).
According to Eq. (8.10), the corresponding magnetic vector potential that can generate
the incident wave is found to be
∞
X
E0 (i)n (2n + 1)
Air = cos φ ĵn (kr) Pn1 (cos θ ) ,
ω n(n + 1)
n=1

where we have used Eq. (8.43) to show that

 
d2 n(n + 1) 2
ĵn (kr) = −k ĵn (kr) .
dr2 r2
8-4 Plane-Wave Scattering from Spheres 545

In a similar manner the electric vector potential is determined to be

∞
X
E0 (i)n (2n + 1)
Aimr = sin φ ĵn (kr) Pn1 (cos θ ) .
ωη n(n + 1)
n=1

The scatterer is symmetric with respect to φ ; however, as shown above, the incident field
has φ -dependency. Using Eq. (8.41) and noting the orthogonality sine and cosine functions,
the scattered potential functions can be written as
∞
X
E0 (1)
Asr = cos φ an ĥn (kr) Pn1 (cos θ )
ω
n=1
and
∞
X
E0 (1)
Asmr = sin φ bn ĥn (kr) Pn1 (cos θ ) .
ωη
n=1

Equation (8.10) can be used to find the electric field. For a metallic sphere, boundary
conditions mandate that the tangential electric field vanish at the sphere’s surface (r = a).
That is,

(Eθi + Eθs )r=a = 0
and

(Eφi + Eφs )r=a = 0 .

This is satisfied if  
∂ i ∂ s 
A + A =0 (8.135)
∂ r r ∂ r r r=a
and 
(Aimr + Asmr )r=a = 0 . (8.136)
Equation (8.135) provides

(i)n (2n + 1) jˆn′ (ka)

an = − . (8.137)
n(n + 1) ĥ(1)′
n (ka)

Also, Eq. (8.136) provides

(i)n (2n + 1) jˆn (ka)

bn = − . (8.138)
n(n + 1) ĥ(1)
n (ka)

The scattered field can easily be computed by combining Eqs. (8.10) and (8.12):
∞
X (1)
ω µε E0 ĥn (kr) 1
Ers = cos φ n(n + 1)an Pn (cos θ ) , (8.139a)
(kr) (kr)
n=1
546 Chapter 8 Spherical Wave Functions and Their Applications

∞
!
X (1)′ (1)
ĥn (kr) ĥn (kr) Pn1 (cos θ )
Eθs = −E0 cos φ ian sin θ Pn1′ (cos θ ) + bn , (8.139b)
(kr) (kr) sin θ
n=1
and
∞
!
X (1)′ (1)
ĥn (kr) Pn1 (cos θ ) ĥn (kr)
Eφs = −E0 sin φ ian + bn sin θ Pn1′ (cos θ ) . (8.139c)
(kr) sin θ (kr)
n=1

The scattered magnetic field components are also found in a similar manner:
∞
X (1)
ω µε E0 ĥn (kr) (1)
Hrs = sin φ n(n + 1)bn Pn (cos θ ) , (8.140a)
(kr)η (kr)
n=1
∞
!
X (1)′ (1)
E0 ĥn (kr) ĥn (kr) Pn1 (cos θ )
Hθs = − sin φ ibn sin θ Pn1′ (cos θ ) + an , (8.140b)
η (kr) (kr) sin θ
n=1
and
∞
!
X (1)′ (1)
E0 ĥn (kr) Pn1 (cos θ ) ĥn (kr)
Hφs = cos φ ibn + an 1′
sin θ Pn (cos θ ) . (8.140c)
η (kr) sin θ (kr)
n=1

In the far-field region, only the θ and φ components of the scattered field remain, noting the
(1)
large-argument expansion of ĥn given by Eq. (8.26). Also, noting that
(1)′
lim ĥn (kr) ≈ (−i)n eikr ,
kr→∞

the field expressions in the far-field region are given by

∞  
eikr X Pn1 (cos θ )
Eθs = −iE0 cos φ n 1′
(−i) an sin θ Pn (cos θ ) − bn (8.141a)
kr sin θ
n=1
and
∞  
eikr X Pn1 (cos θ )
Eφs = −iE0 sin φ n
(−i) an 1′
− bn sin θ Pn (cos θ ) . (8.141b)
kr sin θ
n=1

Equation (8.141) can now be used to find the radar cross section of the sphere. Denoting Eθ
as the co-polarized and Eφ as the cross-polarized field components, the bistatic radar cross
section of a metallic sphere can be computed from

|Eθs |2
σc (θ , φ ) = lim 4π r2
r→∞ |E0 |2
  2
λ 2 cos2 φ X Pn1 (cos θ ) 
∞
n 1′
=  (−i) an sin θ Pn (cos θ ) − bn  (8.142a)
π  sin θ 
n=1
and
8-4 Plane-Wave Scattering from Spheres 547

|Eφ |2
σx (θ , φ ) = lim 4π r2
r→∞ |E0 |2
 ∞  2
λ 2 sin2 φ X n Pn1 (cos θ ) 1′ 
=  (−i) an − bn sin θ Pn (cos θ )  . (8.142b)
π  sin θ 
n=1

Metallic spheres are often used as calibration targets for radar systems. In such approaches
the measured backscatter power by a radar from a target with unknown radar cross section
is compared with that from a metallic sphere placed at the same range and location as the
unknown target. This radar cross section measurement procedure is known as the method of
substitution. Metallic spheres are ideal calibration targets due to their symmetry, and hence
they do not need careful positioning and alignment. Equations (8.142a) and (8.142b) can be
used to calculate the backscatter radar cross section of a metallic sphere. The backscatter
direction is specified by θ = π . For the co-polarized component of the scattered field we need
to set φ = 0, and for the cross-polarized component φ = π /2. To simplify the expression for
the backscatter using Eq. (8.31), we note that Pn (−1) = (−1)n , and using Eq. (8.40),

dPn (x) n(n + 1)

lim =− (−1)n . (8.143)
x→−1 dx 2
Also, from Eq. (8.39a),
p dPn (x)
Pn1 (x) = − 1 − x2 . (8.144)
dx
Hence
Pn1 (cos θ ) n(n + 1)
lim = (−1)n . (8.145)
θ →π sin θ 2
We also need to simplify sin θ Pn′ (cos θ ) as θ → π . Using Eq. (8.40), it can be shown that
p dPn1 (x) 1 1 1 dPn (x) n(n + 1)
lim 1 − x2 = lim √ Pn1 (x) = lim − = (−1)n .
x→−1 dx x→−1 2 1−x2 x→−1 2 dx 2
(8.146)
Hence, in backscatter,
 2
λ 2 X n 
∞

σc =  (i) n(n + 1)(an − bn ) , (8.147)
4π  
n=1

σx = 0 ,

but
(1) (1)′
(i)n (2n + 1) ĵn′ (ka) ĥn (ka) − ĵn (ka) ĥn (ka)
an − bn = (1)′ (1)
n(n + 1) ĥn (ka) ĥn (ka)
(i)n (2n + 1)
= (1)′ (1)
, (8.148)
n(n + 1) ĥn (ka) ĥn (ka)
548 Chapter 8 Spherical Wave Functions and Their Applications

where the Wronskian relations for the Schelkunoff spherical Bessel functions are used.
Substituting Eq. (8.148) into Eq. (8.147), the following equation for the radar cross section of
a metallic sphere is obtained:
 2
λ 2 X (−1)n (2n + 1) 
∞
σc =  (1)′ (1)
 . (8.149)
4π  ĥn (ka) ĥn (ka) 
n=1

For small values of ka (low-frequency approximation), the following approximation can be

used:
1
lim ≈ −ka (8.150a)
ka→0 ĥ(1) (ka)
1
and
1
lim ≈ (ka)2 , (8.150b)
ka→0 ĥ(1)′ (ka)
1

and the summation can be terminated at n = 1, as higher terms are proportional to higher
powers in ka. Hence, when a ≪ λ ,

9λ 2
σcLF = (ka)6 . (8.151)
4π
This equation indicates that the radar cross section of a metallic sphere is proportional to λ14
and is used to explain why the sky is blue. At high frequencies where the radius of metallic
spheres are large compared with the wavelength, using the physical-optics approximation, it
can be shown that the co-polarized radar cross section of a metallic sphere is equal to the area
of its great circle:
σcPO = π a2 . (8.152)
Figure 8-16 shows the normalized radar cross section (σc /(π a2 )) of a metallic sphere as a
function of the normalized radius (a/λ ). Also shown are the low-frequency approximation
based on Eq. (8.151) and the high-frequency approximation based on Eq. (8.152). The low-
frequency formula given by Eq. (8.151) provides accurate results when a/λ ≤ 0.1. The
oscillatory behavior of the RCS around the physical-optics approximation is due to creeping
waves. Basically the physical-optics contribution emanates mainly from the specular point
on the surface, which is the point closest to the source. The rays at the shadow boundary
travel along the surface in the shadow area and emerge from the other side, propagating in
the backscatter direction and can cause constructive or destructive interference. The excess
pathlength for the creeping wave is (2 + π )a, and when this is equal to a wavelength,
constructive interference occurs. This indicates that the spacing between the peaks or valleys
is ∆ (a/λ ) ≈ 0.2. The creeping waves shed energy as they travel along the surface. Hence, the
larger is the radius, the weaker is the signal that emerges from the other side, and hence the
level of the peaks in the RCS as a function of a/λ decreases with increasing a/λ . Figure 8-17
shows the ray representation of the creeping waves around the surface of the metallic sphere.
8-4 Plane-Wave Scattering from Spheres 549

10 1
Exact
Low-frequency
High-frequency

10 0

σc /πa2

10 −1

10 −2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
a /λ

Figure 8-16: The normalized backscatter radar cross section of a metallic sphere as a function
of the normalized radius (a/λ ). Also shown are the low- and high-frequency approximations.

Creeping wave

Specular reflection
(physical-optics solution)

Figure 8-17: The ray representation of creeping waves around a metallic sphere. The creeping
waves in backscatter interfere with the reflected ray from the specular point and cause the ripple
observed in the RCS shown in Fig. 8-16 as a function of a/λ .
550 Chapter 8 Spherical Wave Functions and Their Applications

8-4.2 Scattering from a Dielectric Sphere

Electromagnetic scattering from a dielectric sphere can be obtained following a similar

procedure. Consider the geometry of the problem given in Fig. 8-15. In this case, however,
we assume that the sphere is a homogeneous dielectric material with permittivity εs and
permeability µs . In addition to the fields outside the sphere we need to consider the interior
fields that can be generated from the following TM- and TE-to-r potentials:
∞
X
E0
Adr = cos φ cn ĵn (ks r) Pn1 (cos θ ) (8.153a)
ω
n=1
and
∞
X
E0
Admr = sin φ dn jn (ks r) Pn1 (cos θ ) , (8.153b)
ωη
n=1
√
where ks = ω µs εs . The incident and scattered potentials have the same form as those for the
metallic sphere. To find the four sets of unknown coefficients an , bn , cn , and dn , the boundary
condition relationship can be used. Basically,
 
(Eθi + Eθs )r=a = Eθd r=a , (8.154a)
 
(Eφi + Eφs )r=a = Eφd r=a , (8.154b)
 
(Hθi + Hθs )r=a = Hθd r=a , (8.154c)
and
 
(Hφi + Hφs )r=a = Hφd r=a . (8.154d)

Using Eqs. (8.10)–(8.12), it can easily be shown that Eq. (8.154) is satisfied if

∂  ∂ d 
(Air + Asr )r=a = A ,
∂r ∂ r r r=a
 
(Air + Asr )r=a = Adr r=a ,
∂  ∂ d 
(Aimr + Asmr )r=a = A ,
∂r ∂ r mr r=a
and
 
(Aimr + Asmr )r=a = Admr r=a .

These give four equations for the four unknowns, which can be solved simultaneously to find
p p
(i)n (2n + 1) − εs /ε0 ĵn′ (ka) ĵn (ks a) + µs /µ0 ĵn (ka) ĵn′ (ks a)
an = p (1)′ p (1)
, (8.155a)
n(n + 1) εs /ε0 ĥn (ka) ĵn (ks a) − µs /µ0 ĥn (ka) ĵn′ (ks a)
p p
(i)n (2n + 1) − εs /ε0 ĵn (ka) ĵn′ (ks a) + µs /µ0 ĵn′ (ka) ĵn (ks a)
bn = p (1) p (1)′
, (8.155b)
n(n + 1) εs /ε0 ĥn (ka) ĵ ′ (ks a) − µs /µ0 ĥn (ka) ĵn (ks a)
n
8-4 Plane-Wave Scattering from Spheres 551

p
(i)n (2n + 1) i εs /ε0
cn = p (1)′ p (1)
, (8.155c)
n(n + 1) εs /ε0 ĥn (ka) ĵn (ks a) − µs /µ0 ĥn (ka) ĵn′ (ks a)
and
p
(i)n (2n + 1) −i µs /µ0
dn = p (1) p (1)′
. (8.155d)
n(n + 1) εs /ε0 ĥn (ka) ĵn′ (ks a) − µs /µ0 ĥn (ka) ĵn (ks a)

In the limiting case as the sphere dielectric εs approaches that of a PEC (εs → ε0 (1 + i∞)),
an and bn given by Eqs. (8.155a) and (8.155b) reduce to those given by Eqs. (8.137) and
(8.138) for a metallic sphere. It is also interesting to note that when
εs µs
= , (8.156)
ε0 µ0
the expressions given by Eqs. (8.155a) and (8.155b) lead to

an = bn . (8.157)

The significance of this result is that in the backscatter direction, for which the co-polarized
backscatter radar cross section can be computed from Eq. (8.147), the backscatter goes to
zero independently of the sphere radius or its relative permittivity and permeability, so long
as εs /ε0 = µs /µ0 . In this case, a dielectric sphere becomes invisible to a monostatic radar.
In the general case, the backscatter can be computed from Eq. (8.147) using the an and bn
coefficients given by Eqs. (8.155a) and (8.155b). As an example, the normalized backscatter
RCS values of dielectric spheres with µs = µ0 and two different values of εs = 1.6ε0 and
εs = 2ε0 , were evaluated as a function of the normalized radius (a/λ ) and are shown in
Fig. 8-18. The highly oscillatory behavior of the RCS, particularly at higher frequencies (large
a/λ ), is due to internal resonances of the dielectric sphere. The oscillatory behavior damps
out as the dielectric of the sphere becomes lossy.
A low-frequency approximation for the scattering expressions for dielectric spheres can
be obtained by expanding the expressions for an and bn in terms of a truncated power series
in (ka). This can be done by noting that
 
n n! n+1 n+1 2
ĵn (ka) ≈ 2 (ka) 1− (ka) + · · · (8.158a)
(2n + 1)! (2n + 1)(2n + 3)
and
(1) i (2n)! 1
ĥn (ka) ≈ n + ··· (8.158b)
2 n! (ka)n
Simple expressions can be obtained when µs = µ0 , namely
 
1 εs
a1 ≈ − 1 (ka)5 + · · · , (8.159a)
30 ε0
a2 ≈ 0 , (8.159b)
552 Chapter 8 Spherical Wave Functions and Their Applications

10 1

10 0

σc /πa2 10 −1

Exact, εs = 1.6
−2
10 Low Freq., εs = 1.6
Exact, εs = 2
Low Freq., εs = 2
−3
10
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
a /λ

Figure 8-18: The normalized backscatter RCS of two dielectric spheres, one with εs = 1.6ε0
and another with εs = 2ε0 , as a function of the normalized radius a/λ . Also shown are plots of
the low-frequency approximation.

 
εs − ε0 3 3 (εs − ε0 )(εs − 2ε0 ) 5
b1 ≈ i (ka) + (ka) + · · · , (8.159c)
εs + 2ε0 5 (εs + 2ε0 )2
and
1 εs − ε0
b2 ≈ (ka)5 + · · · (8.159d)
18 2εs + 3ε0
For small values of ka, the first term of b1 becomes the only significant term. Using b1 only
in Eq. (8.76) yields the backscatter RCS for the low-frequency approximation
 
 εs − ε0  2
σcLF = (π a2 ) 4   (ka)4 . (8.160)
εs + 2ε0 
The computed RCS based on Eq. (8.160) is compared with the results based on the exact
solution in Fig. 8-18. Good agreement is observed for a/λ < 0.1.
For magnetodielectric spheres, it turns out that to the lowest order in (ka), a1 becomes
 
µs − µ0
a1 ≈ i (ka)3 ,
µs + µ0
8-5 Wave Propagation in a Conical Waveguide 553

and as a result the low-frequency backscatter RCS for a magneto-dielectric sphere is given by
 
LF 2
 εs − ε0
 µs − µ0 2
σc = (π a ) 4  −  (ka)4 . (8.161)
εs + 2ε0 µs + 2 µ0

8-5 Wave Propagation in a Conical Waveguide

Consider a metallic cone with a cone angle θ0 and whose axis coincides with the z-axis. We
are interested in the fields that can be supported inside the cone portrayed in Fig. 8-19. For an
infinite cone, outward-propagating TM waves can be generated from
(
m (1) cos mφ ,
Ar (r, θ , φ ) = Pνnm (cos θ ) ĥνnm (kr) (8.162)
sin mφ ,

where m is an integer, since the domain of interest includes φ ∈ [0, 2π ]. The spherical Hankel
function of the first kind is chosen for the radial function, since only outward-going waves
are of interest. The associated Legendre function of the first kind is used, as θ = 0 is included
in the domain of interest. Boundary conditions require that Eφ = Er = 0 at θ = θ0 . Using
Eq. (8.10a) or Eq. (8.10c), it becomes evident that

Pνmnm (cos θ0 ) = 0 (8.163)

Figure 8-19: Geometry of a metallic conical waveguide with cone angle θ0 .

554 Chapter 8 Spherical Wave Functions and Their Applications

θ = 10˚ θ = 20˚
1 1
m = 0 (/1) m = 0 (/1)
m = 1 (/15.7) m = 1 (/12.1)
m = 2 (/1.28e+03) m = 2 (/872)
0.5 m = 3 (/9.51e+04) 0.5 m = 3 (/7.45e+04)
Pνm(cos θ0)

Pνm(cos θ0)
0 0

−0.5 −0.5

−1 −1
0 20 40 60 80 0 20 40 60 80

θ = 30˚ θ = 40˚
1 1
m = 0 (/1) m = 0 (/1)
m = 1 (/9.87) m = 1 (/8.89)
m = 2 (/800) m = 2 (/680)
0.5 m = 3 (/5.76e+04) 0.5 m = 3 (/5.66e+04)
Pνm(cos θ0)

Pνm(cos θ0)

0 0

−0.5 −0.5

−1 −1
0 20 40 60 80 0 20 40 60 80

Figure 8-20: The plot of transcendental equation Eq. (8.163) for different cone angles θ0 = 10◦ ,
20◦ , 30◦ , and 40◦ , as a function of ν for different values of m. The zeros of the order νnm specify the
different modes that can be supported in a conical waveguide. The plots shown here for different
modes are normalized to their maximum value (noted in the legend) for better viewing.

for the boundary conditions to be satisfied. Equation (8.163) is the transcendental equation
that provides the admissible values of the order of Legendre function of the first kind for
different values of m (νnm ). Once values of νnm are found from Eq. (8.163), a potential function
constructed from the superposition of modes given by Eq. (8.162) can be used to express a
field inside the conical waveguide:
∞ X ∞
(
X (1) cos mφ ,
Ar (r, θ , φ ) = anm Pνmnm (cos θ ) ĥνnm (kr) (8.164)
m=0 n=1
sin mφ .

Here n is an index for the ordered zeros of Eq. (8.163), and the superscript m in ν m indicates
the zeros of Eq. (8.163) for different values of m. Note that νnm ≥ m. Figure 8-20 shows plots of
the transcendental equation for different values of θ0 and m. The zeros of these functions (νnm )
are extracted and plotted in Fig. 8-21 as a function of θ0 for different values of m. Considering
8-5 Wave Propagation in a Conical Waveguide 555

40
35 m=0
30 m=1
m=2
25 m=3
ν1m 20
(a) 15
10
5
0
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚

60
m=0
50 m=1
40 m=2
m=3
ν2m 30
(b)
20
10
0
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚

80
70 m=0
60 m=1
m=2
50
m=3
ν3m 40
(c) 30
20
10
0
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚

Figure 8-21: The zeros of the TM transcendental equation given by Eq. (8.163) are extracted
and plotted as a function of the cone angle θ for multiple values of m. (a) shows the first zero ν1m ,
(b) shows the second zero ν2m , and (c) shows ν3m .

the fact that  

d2 2 (1) ν (ν + 1) (1)
2
+ k ĥν (kr) = ĥν (kr) , (8.165)
dr r2
the field components can be obtained from Eq. (8.10):
∞ ∞
(
−1 X X (1) cos mφ ,
Er = 2
amn νnm (νnm + 1) ĥνnm (kr) Pνmnm (cos θ ) (8.166a)
iω µε r sin mφ ,
m=0 n=1
556 Chapter 8 Spherical Wave Functions and Their Applications

∞ ∞
(
k sin θ X X (1)′ cos mφ ,
Eθ = amn ĥνnm (kr) Pνm′nm (cos θ ) (8.166b)
iω µε r sin mφ ,
m=0 n=1
∞ X ∞
(
−k X (1)′ − sin mφ ,
m
Eφ = mamn ĥνnm (kr) Pνnm (cos θ ) (8.166c)
iω µε r sin θ cos mφ ,
m=0 n=1
∞ X ∞
(
1 X (1) − sin mφ ,
Hθ = mamn ĥνnm (kr) Pνmnm (cos θ ) (8.166d)
µ r sin θ cos mφ ,
m=0 n=1
and
∞ ∞
(
sin θ X X (1) cos mφ ,
Hφ = amn ĥνnm (kr) Pνm′nm (cos θ ) (8.166e)
µr sin mφ .
m=0 n=1

For the TE-to-r modes, the potential Amr takes a similar form to Eq. (8.162). However, the
transcendental equation obtained from Eq. (8.12e) requires that

Pνm′nm (cos θ0 ) = 0 . (8.167)

Once the values of νnm are obtained for different values of m from Eq. (8.167), the field
quantities can be obtained from the dual of Eq. (8.166). Figure 8-23 shows νnm values as a
function of cone angle θ0 for the TE case. Note that ν = 0 is a solution to Eq. (8.167), but
this solution renders null fields inside the conical antenna. In other words, conical antennas
cannot support TEM waves, as expected.
A truncated conical waveguide can be used as a horn antenna. Such antennas are usually
fed by a circular waveguide. However, it should be noted that modal analysis for such antennas
is cumbersome. Figure 8-23 shows the geometry for a conical horn antenna with θ0 = 25◦ and
a height of 10 cm, connected to a circular waveguide with a diameter of 3 cm, supporting a
TE11 mode. The resulting radiation pattern in the far field and the input reflection coefficient
as a function of frequency are also shown in this figure. Numerical computation is carried out
using an FEM solver.

Example 8-3: Metallic Cone Excited by an External Electric Current

Consider a metallic cone whose axis coincides with the z-axis of the Cartesian coordinate
system shown in Fig. 8-24. The cone is defined by a constant θ0 surface. The cone is excited
by a spherical surface current that extends between θ1 and θ2 :
(
J0 (θ ) φ̂
φ θ1 ≤ θ ≤ θ2 ,
Js (r) = (8.168)
0 otherwise.

Find an expression for the electric field.

Solution: Since the excitation current is independent of φ and the cone is also symmetric with
respect to φ , all field quantities are independent of variable φ ; i.e., ∂ /∂ φ = 0. Noting that a
8-5 Wave Propagation in a Conical Waveguide 557

40
35 m=0
30 m=1
25
ν1m 20
m=2
m=3
(a) 15
10
5
0
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚

60
50 m=0
40 m=1
ν2m 30
m=2
m=3
(b) 20
10
0
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚

80
70 m=0
60
50 m=1
ν3m 40 m=2
(c) 30
20 m=3
10
0
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚

Figure 8-22: The zeros of the TE transcendental equation given by Eq. (8.167) are extracted and
plotted as a function of the cone angle θ for multiple values of m values. Part (a) shows the first
zero ν1m , (b) shows the second zero ν2m , and (c) shows ν3m .

φ-directed current should produce an electric field that has a φ component. Noting that for a
φ̂
TM-to-r field, according to Eq. (8.10c), Eφ = 0, it can be concluded that this current can only
produce TE-to-r modes, with electric field components given by Eqs. (8.12d) and (8.12e):

1 ∂ Amr
Eθ = − =0 (8.169a)
ε r sin θ ∂ φ
and
1 ∂ Amr
Eφ = . (8.169b)
εr ∂ θ
We partition the domain of interest into two regions corresponding to r < a and r > a, and
for each region we provide an appropriate solution for Amr (r, θ ) in terms of the modes of the
558 Chapter 8 Spherical Wave Functions and Their Applications

(a)

−10
Reflection Coefficient

−20

(b) −30

−40

−50
8 9 10 11 12
Frequency (GHz)

20
H-plane
E-plane
Total gain at 10 GHz (dBi)

10

(c) −10

−20

−30
−90 −70 −50 −30 −10 10 30 50 70 90
(degrees)

Figure 8-23: (a) A conical horn antenna fed by a circular waveguide supporting a TE11 mode,
(b) the input reflection coefficient as seen by the circular waveguide as a function of frequency,
and (c) the far-field radiation pattern in the E- and H-planes at 10 GHz.
8-5 Wave Propagation in a Conical Waveguide 559

J0(θ) ϕ̂

a θ0
y

Figure 8-24: Geometry of a metallic cone excited by a spherical current sheet flowing in the
φ-direction with a uniform distribution with respect to φ but variable along θ . The radius of the
φ̂
spherical current sheet is a.

conical waveguide. For r < a,

∞
X
A<
mr (r, θ ) = aνn Pνn (cos θ ) ĵνn (kr) , (8.170a)
n=1

and for r > a,

∞
X (1)
A>
mr (r, θ ) = bνn Pνn (cos θ ) ĥνn (kr) . (8.170b)
n=1
The transcendental equation from which the values of νn are determined is given by
Eq. (8.166) for m = 0 :
Pν′ n (cos θ0 ) = 0 , (8.171)
which enforces the condition Eφ (θ0 ) = 0 on the surface of the metallic cone. Once the values
of νn are determined, Eqs. (8.170a) and (8.170b), in conjunction with the applicable boundary
conditions, must be used to find the unknowns aνn and bνn . Referring to Eq. (8.169b),
continuity of Eφ at r = a requires

A< >
mr (a, θ ) = Amr (a, θ ) ,

which provides the following equation relating aνn to bνn :

(1)
aνn ĵνn (ka) = bνn ĥνn (ka) . (8.172)
560 Chapter 8 Spherical Wave Functions and Their Applications

The tangential component of the magnetic field across the surface current must be discon-
tinuous at r = a:
Hθ (a+ , θ ) − Hθ (a− , θ ) = J0 (θ ) . (8.173)
Using Eq. (8.12b) for Hθ and substituting the expressions in Eq. (8.173), we have
∞
X
k ∂ (1)′
− Pνn (cos θ ) [bνn ĥνn (ka) − aνn ĵν′ n (ka)] = J0 (θ ) . (8.174)
iω µε a ∂θ
n=1

Using Eq. (8.172) in Eq. (8.174) and the Wronskian for the spherical Bessel functions,
Eq. (8.174) can be simplified to
∞
1 X ∂ bνn
Pνn (cos θ ) = J0 (θ ) . (8.175)
µε a ∂θ ĵν n (ka)
n=1

Multiplying both sides of Eq. (8.175) by (∂ /∂ θ ) Pνm (cos θ ) sin θ and integrating with respect
to θ from 0 to θ0 , we have
Z θ2
1 bνn ∂
Cν = J0 (θ ) Pν (cos θ ) sin θ d θ , (8.176)
µε a ĵνn (ka) n θ1 ∂ θ n

where we have used the orthogonality relations of the Legendre function

Z θ2 (
∂ 0 m,n,
Pνn (cos θ ) Pνn (cos θ ) sin θ d θ =
θ1 ∂θ Cνn m=n,

with  
νn (νn + 1) ∂ 2 Pνn
Cνn = − sin θ Pνn (cos θ ) .
2νn + 1 ∂ θ ∂ ν θ =θ0
After finding bνn from Eq. (8.176) and aνn from Eq. (8.172), the field quantities can then be
obtained.

8-6 Biconical Structures

Biconical structures are formed from two cones of arbitrary angles with a common axis and
coinciding apexes at the origin. Figures 8-25(a) and Fig. 8-25(b) show two types of such
structures, one structure having one cone in the upper half-space and the other cone in the
lower half-space, and the other structure having both cones in the same half-space. Assuming
the conical surfaces are metallic and the region of interest is between the surfaces of the cones,
such structures can support propagating waves. The outward-traveling TM fields supported by
these structures can be generated from
(
(1) cos mφ ,
A(r, θ , φ ) = ĥν (kr) [Aν Pνm (cos θ ) + Bν Qm
ν (cos θ )] (8.177)
sin mφ .
8-6 Biconical Structures 561

(a) (b)

Figure 8-25: Biconical structures with the same axis: (a) external cones and (b) coaxial cones.

The choice of fraction order for Legendre functions stems from the fact that θ = 0 and/or
θ = π are excluded from the domain of interest. The boundary condition mandates that Er =
Eφ = 0 at θ = θ1 and θ = θ2 . Using Eq. (8.10a) or Eq. (8.10c), it can be shown that the
required boundary condition at θ = θ1 is satisfied when

Aν = Qm
ν (cos θ1 ) (8.178a)
and
Bν = −Pνm (cos θ1 ) . (8.178b)

Imposing the other boundary condition at θ = θ2 provides the following transcendental

equation:
Pνmn (cos θ2 ) Qνmn (cos θ1 ) − Qνmn (cos θ2 ) Pνmn (cos θ1 ) = 0 . (8.179)
This equation must be solved for ν to determine the fields supported by the biconical structure.
The subscript n here denotes the nth zero of Eq. (8.179) for a given ν .
For TE-to-r modes, a similar expression for the electric vector potential can be written.
However, the boundary condition requires that

1 ∂ Amr 
Eφ = =0. (8.180)
ε r ∂ θ θ =θ1 and θ2
562 Chapter 8 Spherical Wave Functions and Their Applications

Hence the expression for the electric vector potential takes the following form once the
boundary condition at θ = θ1 is satisfied:
 m (cos θ ) m (cos θ )
(
(1) dQ ν 1 dP 1 cos mφ ,
Amr (r, θ , φ ) = hν (kr) Pνm (cos θ ) − Qνm (cos θ ) ν
d θ1 d θ1 sin mφ .
(8.181)
Enforcing the boundary condition at θ = θ2 , the following equation for ν is obtained:

dPνm (cos θ2 ) dQm

ν (cos θ1 ) dQm m
ν (cos θ2 ) dPν (cos θ1 )
− =0. (8.182)
d θ2 d θ1 d θ2 d θ1
It is interesting to note that m = 0 when ν = 0, and therefore the derivative with respect to φ
is zero for all field quantities. Also, we can show that
 2 
d 2 (1) ν (ν + 1) (1)
2
+ k hν (kr) = hν (kr) , (8.183)
dr r2

and for ν = 0, Er (r, θ , φ ) = 0. Therefore in this case the TM electric field has only a
θ -component and the TM magnetic field has only a φ -component. That is, the tangential
components of the electric field on the surfaces of the cones, independent of the cone angles,
are naturally zero. This constitutes a TEM wave with E = Eθ θ̂ θ and H = Hφ φ̂ φ, and the
direction of propagation is along r̂. The magnetic vector potential for TM-to-r in this case is
written as
(1) 1 1 + cos θ
A00 ikr
r = ĥ0 (kr) Q0 (cos θ ) = −ie × ln
2 1 − cos θ
= −i ln[cot(θ /2)] eikr . (8.184)

Using Eq. (8.184) in Eqs. (8.10b) and (8.11b), the expression for the electric and magnetic
fields of the TM00 (TEM) mode are given by

−iω eikr
Eθ (r, θ ) =
sin θ kr
and
−iω eikr
Hφ (r, θ ) = .
η sin θ kr
Plots of the functions given by the right-hand side of Eq. (8.179) are shown in Fig. 8-26
for the biconical structures shown in Fig. 8-25(a) when θ1 = π − θ2 , for different values of
θ1 as a function of ν . The zeros of this function specify the modes of propagation in the
biconical structure. For each value of θ1 = π − θ2 , the zeros of the transcendental equation
are also functions of m, as shown in the plots. Each function is normalized to its maximum
value so that the zero crossing can be easily observed for all curves.
The zeros of the TM transcendental equation defined by (8.179) were extracted and are
8-6 Biconical Structures 563

Pνm(cos θ1) Qνm(cos θ2) − Qνm(cos θ1) Pνm(cos θ2)

Pνm(cos θ1) Qνm(cos θ2) − Qνm(cos θ1) Pνm(cos θ2)

θ1 = 20˚, θ2 = 160˚ θ1 = 30˚, θ2 = 150˚
1 1
m = 0 (/3.47) m = 0 (/2.63)
m = 1 (/44) m = 1 (/28.8)
m = 2 (/1.07e+04) m = 2 (/6.23e+03)
0.5 m = 3 (/2.95e+06) 0.5 m = 3 (/1.47e+06)

0 0

−0.5 −0.5

−1 −1
0 5 10 15 0 5 10 15

θ1 = 40˚, θ2 = 140˚ θ1 = 50˚, θ2 = 130˚

Pνm(cos θ1) Qνm(cos θ2) − Qνm(cos θ1) Pνm(cos θ2)

Pνm(cos θ1) Qνm(cos θ2) − Qνm(cos θ1) Pνm(cos θ2)

1 1
m = 0 (/2.02) m = 0 (/1.53)
m = 1 (/23.9) m = 1 (/19.2)
m = 2 (/5.83e+03) m = 2 (/4.27e+03)
0.5 m = 3 (/1.33e+06) 0.5 m = 3 (/1.01e+06)

0 0

−0.5 −0.5

−1 −1
0 5 10 15 0 5 10 15

Figure 8-26: Plots of the transcendental equation given by Eq. (8.179) for multiple combinations
of cone angles as a function of ν . The zeros of the function for different values of m indicate a
mode of propagation. The plots shown here for different modes are normalized to their maximum
value (noted in the legend) for better viewing.

plotted as a function of the cone angle θ1 = π − θ2 for different values of m, and are shown in
Fig. 8-27.
It is interesting to note that the characteristic impedance of the TEM wave defined as
Eθ
Z0 = =η
Hφ

is independent of r, θ , and φ . When exciting a coaxial biconical structure at its tip by a

coaxial transmission line, as shown in Fig. 8-28, the input impedance can be calculated by
determining the voltage and current near the tip of the cones. The voltage between the two
564 Chapter 8 Spherical Wave Functions and Their Applications

5
4
3
ν1m
2 m=0
m=1
1 m=2
m=3
0
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚
θ1 (= 180˚ − θ2)

9
8 m=0
7 m=1
6 m=2
ν2m 5 m=3
4
3
2
1
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚
θ1 (= 180˚ − θ2)

14
m=0
12
m=1
10 m=2
ν3m 8 m=3

6
4
2
10˚ 15˚ 20˚ 25˚ 30˚ 35˚ 40˚ 45˚ 50˚ 55˚ 60˚ 65˚ 70˚
θ1 (= 180˚ − θ2)

Figure 8-27: The zeros of the TM transcendental equation given by Eq. (8.179) were extracted
and plotted as a function of the cone angle θ1 = π − θ2 for multiple values of m.

cones near r = 0 is given by

Z θ2  
iω ikr cot(θ2 /2)
V = V (θ1 ) −V (θ2 ) = − lim E0 (r, θ ) r d θ = − lim e ln
r→0 θ1 r→0 k cot(θ1 /2)
 
−iω cot(θ2 /2)
= ln ,
k cot(θ1 /2)
where we have used the fact that
1 d
− = [ln(cot θ /2)] .
sin θ dθ
8-6 Biconical Structures 565

Coaxial transmission line

Figure 8-28: Geometry of a coaxial biconical antenna fed by a coaxial transmission line
supporting TM00 .

Also, the current flowing over the inner cone can be calculated from
Z 2π
−i2π
I = lim Hφ (r, θ ) r sin θ d φ = .
r→0 0 µ
Hence the input impedance of the biconical structure is given by
 
V η cot(θ1 /2)
Zin = = ln .
I 2π cot(θ2 /2)

Example 8-4: Coaxial Biconical Antenna

For an air-filled coaxial biconical antenna, plot θ2 in terms of θ1 so that the antenna is matched
to a coaxial cable with each of the following characteristic impedances: 50 Ω, 75 Ω, and
100 Ω.
Solution: For an air-filled antenna, η = 120π , and hence
 
120π cot(θ1 /2)
Z0 = Zin = ln .
2π cot(θ2 /2)

Solving this equation for θ2 yields

h i
−1 −Z0 /60
θ2 = 2 cot e cot(θ1 /2) .
566 Chapter 8 Spherical Wave Functions and Their Applications

150

100

θ2
Z in = 50
50
Z in = 75
Z in = 100

0
0 10 20 30 40 50 60
θ1

Figure 8-29: The variation of θ2 as a function of θ1 for a coaxial biconical antenna with cone
angles θ1 and θ2 , for the geometry depicted in Fig. 8-28, for achieving a matched input impedance
to a coaxial line for lines with characteristic impedances of Z0 = 50 Ω, 75 Ω, and 100 Ω.

Figure 8-29 shows plots of θ2 versus θ1 for Z0 = 50 Ω, 75 Ω, and 100 Ω. We observe that
θ2 increases with increasing characteristic impedance of the line. It is noted that the antenna
produces a wider beam in elevation angle as θ2 − θ1 increases.
Choosing an input impedance of 75 Ω can provide a wide-angle cone. Choosing θ2 = 65◦
and using Fig. 8-29 for Zin = 75 Ω provides an interior cone angle θ1 ≈ 20◦ . This biconical
structure is truncated for a radial length of 3 cm to form a biconical antenna fed by a coaxial
line. A numerical method based on FEM is used to simulate the interior field pattern, the input
reflection coefficient, and the far-field radiation pattern, which are shown in Figs. 8-30(a),
8-30(b), and 8-30(c), respectively in the frequency range of 10–20 GHz.. The cone angle
predicted analytically provides a useful approach to ensuring a very good impedance match
to the coaxial line feeding the conical antenna. The 3-dB beamwidth is about 35◦ , which is
smaller than the angular range between the two cones (θ2 − θ1 = 45◦ ).
8-6 Biconical Structures 567

(a)

−
−
dB(S(1,1))

−
(b) −
−
−
−

Frequency (GHz)
dB(GainTotal)

(c)

−
− − − − − −
θ (degrees)

Figure 8-30: (a) Configuration of a biconical antenna fed by a 75-Ω coaxial line and the resultant
field distribution. (b) The input reflection coefficient as a function of frequency, which shows a
good impedance match over a wide bandwidth. (c) The far-field radiation pattern in a constant
φ -plane with omnidirectional characteristic and a 3-dB beamwidth in elevation of about 35◦ . The
angles of the cones are predicted from the results shown in Fig. 8-29.
568 Chapter 8 Spherical Wave Functions and Their Applications

Figure 8-31: Geometry of a θ –φ spherical waveguide. The domain of interest is between

θ1 -constant, θ2 -constant, φ = 0-constant, and φ0 -constant surfaces.

8-7 Other Spherical Waveguides

Other types of spherical waveguides can also be realized by considering a metallic structure
formed by φ -constant surfaces and θ -constant surfaces, as shown in Fig. 8-31. The θ -constant
surfaces are conical surfaces with generating angles θ1 and θ2 . For a TM-to-r mode, the
appropriate vector potential takes the following form:
(1)  µ µ 
A(r, θ , φ ) = ĥ0 (kr) aν Pν (cos θ ) + bν Qν (cos θ ) (sin µφ +Cµ cos µφ ) , (8.185)

because the desired domain in φ is [0, φ0 ] and the domain in θ is [θ1 , θ2 ]. In this example we
consider an outward-propagating mode only. On φ -constant surfaces, the boundary condition
requires Er and Eθ to vanish at φ = 0 and φ = φ0 . According to Eqs. (8.10), this requires

A(r, θ , 0) = A(r, θ , φ0 ) = 0 . (8.186)

Enforcing Eq. (8.186) renders the following equations:

Cµ = 0 (8.187)
and
mπ
sin µφ0 = 0 µ= m = 1, 2, . . . (8.188)
φ0
On θ -constant surfaces, Er and Eφ must be zero. Referring to Eqs. (8.10) again, their boundary
condition is satisfied if
A(r, θ1 , φ ) = A(r, θ2 , φ ) = 0 . (8.189)
8-7 Other Spherical Waveguides 569

Similar to Eqs. (8.110a) and (8.110b), Eqs. (8.189) is satisfied for θ = θ1 if

mπ /φ0
aν = Qν (cos θ1 ) (8.190a)
and
mπ /φ0
bν = −Pν (cos θ1 ) . (8.190b)

In order for Eq. (8.189) to be true at θ = θ2 , we must enforce

mπ /φ0 mπ /φ0 mπ /φ0 mπ /φ0
Qνnm (cos θ1 ) Pνnm (cos θ2 ) − Pνnm (cos θ1 ) Qνnm (cos θ2 ) = 0 . (8.191)

Equation (8.191) is the transcendental equation for the possible modes in the spherical
waveguide shown in Fig. 8-31. Here νnm refers to the nth zero of Eq. (8.191) for µm = mπ /φ0 .
For TE-to-r mode, a transcendental equation similar to Eq. (8.182) can be obtained and is
given by
d mπ /φ d mπ /φ0 d mπ /φ0 d mπ /φ
Qνnm 0 (cos θ1 ) Pνnm (cos θ2 ) − Pνnm (cos θ1 ) Qνnm 0 (cos θ2 ) = 0 .
d θ1 d θ2 d θ1 d θ2
(8.192)
A truncated θ –φ spherical waveguide in the radial direction (r = R0 ) forms a θ –φ
spherical horn. This horn antenna can be fed by a conventional rectangular waveguide
or a ridged waveguide at its tip. To examine the performance of such a horn antenna, a
transition from a rectangular waveguide to a θ –φ spherical waveguide is designed, as shown
in Fig. 8-32(a). Also shown in Fig. 8-32(a) are the dimensions of a truncated horn antenna for
operating at X-band (8–12 GHz). The co-polarized and cross-polarized radiation patterns (at
10 GHz) of this antenna, for both the E-plane and the H-plane, are shown in Fig. 8-32(b). It is
shown that the roll-off of the radiation pattern is not symmetric with angle. This type of horn
seems to be appropriate as a feed for an offset reflector antenna.
570 Chapter 8 Spherical Wave Functions and Their Applications

(a)

30
E-plane Co-pol
20 H-plane Co-pol
E-plane X-pol
H-plane X-pol
10

-10
Gain (dB)

-20
(b)
-30

-40

-50

-60

-70

-150 -100 -50 0 50 100 150

(degrees)

Figure 8-32: The geometry and simulated radiation pattern of a truncated θ –φ spherical
waveguide (horn) antenna. A transition to a rectangular waveguide is designed for impedance
matching.
SUMMARY 571

Summary
Concepts
• In the spherical coordinate system, the vector wave equation for both the magnetic and
the electric vector potentials under the Lorentz gauge condition do not yield a scalar
wave equation for any component of either vector potential.
• Using a different gauge condition, it is shown that a scalar potential defined by
ψ (r, θ , φ ) = Ar (r, θ , φ )/r satisfies the scalar wave equation if the magnetic vector
potential has only an r component. This vector potential (A = Ar r̂) generates a set
of fields for which the magnetic field has no r component. This set of fields is known
as TM-to-r fields.
• Using the duality principle, it is shown that if the electric vector potential has only an
r component (Am = Amr r̂), a set of fields can be generated for which the field has no r
component. This set of fields is known as TE-to-r fields.
• The method of separation of variables is used to solve the scalar wave equation.
This leads to three separate differential equations, one for each of the three spherical
coordinate variables. The function of the φ variable is a harmonic function, that of the
θ variable is a Legendre or associated Legendre function whose order is a function of
the domain of interest, and finally the function of the r variable is a spherical Hankel
function.
• The characteristics of the spherical wave functions are studied to guide the selection of
proper functions suited for the domain of interest.
• Application of the method of separation of variables to problems such as the
determination of eigenvalues (resonances) and eigenfunctions (fields for source-free
regions) of spherical cavity and dielectric resonator are presented.
• To study plane-wave scattering by metallic and dielectric spherical objects, expansion
of plane waves in terms of spherical wave functions is presented. Using such an
expansion, the scattering of plane waves by spherical objects is computed and analytical
solutions for the bistatic scattering cross sections of such targets are obtained.
• The characteristics of wave propagation in conical waveguides are studied. The field’s
modal expansion is obtained by first solving a transcendental equation and then using
the solution to provide the order of the associated Legendre functions and the spherical
Hankel functions.
• Field modal expansion in the region between two metallic cones sharing a common axis
(known as a biconical structure) is presented. The characteristics of such fields depend
on the order of the associated Legendre functions of the first and second kind, which
are found from a transcendental equation obtained by imposing boundary conditions.
• Other spherical waveguide structures formed by four metallic surfaces that coincide
with constant θ (cones) and constant φ (planes) surfaces are also considered.
572 Chapter 8 Spherical Wave Functions and Their Applications

Important Equations

New gauge condition:

∂
Ar = iω µε Φ
∂r
Wave equation for TM-to-r:
Ar µ
(∇2 + k2 ) = − Jr
r r
Wave equation for TE-to-r:
Amr ε Jmr
(∇2 + k2 ) =−
r r
Electric field components for TM-to-r:
 
−1 ∂2 2
Er = + k Ar ,
iω µε ∂ r2
 2 
−1 ∂
Eθ = Ar ,
iω µε r ∂ r ∂ θ
−1 1 ∂2
Eφ = Ar
iω µε r sin θ ∂ r ∂ φ

Magnetic field components for TM-to-r:

1 ∂ Ar
Hθ = ,
µ r sin θ ∂ φ
1 ∂ Ar
Hφ = −
µr ∂ θ
SUMMARY 573

Important Equations (continued)

Electric and magnetic field components for TE-to-r:

 2 
−1 ∂ 2
Hr = + k Amr ,
iω µε ∂ r2
−1 ∂2
Hθ = Amr ,
iω µε r ∂ r ∂ θ
−1 1 ∂2
Hφ = Amr ,
iω µε r sin θ ∂ r ∂ φ
−1 ∂ Amr
Eθ = ,
ε r sin θ ∂ φ
1 ∂ Amr
Eφ =
εr ∂ θ
Separation of variables:
Ar Amr
or = ψ (r, θ , φ ) = R(r) Q(θ ) F(φ )
r r
The differential equations for R, Q, and F:
 
d 2 dR
r + [(kr)2 − ν (ν + 1)]R = 0 ,
dr dr
   
1 d dQ µ2
sin θ + ν (ν + 1) − 2 Q=0,
sin θ d θ dθ sin θ
d2F
+ µ 2F = 0
dφ 2

Harmonic functions for F(φ ):

F = Aeiµφ + Be−iµφ

Spherical Bessel function for R(r):

 π 1/2
R(r) = zn (kr) = Zn+1/2 (kr)
2kr
Recurrence formula for spherical Bessel functions:

(2n + 1)
zn+1 (kr) = zn (kr) − zn−1 (kr)
kr
574 Chapter 8 Spherical Wave Functions and Their Applications

Important Equations (continued)

Large-argument expansion for spherical Bessel functions:

(1) ei[kr−π (n+1)/2] eikr

lim hn (kr) ≈ = (−i)n+1 P
kr→∞ kr kr
(2) e−i[kr−π (n+1)/2] e−ikr
lim hn (kr) ≈ = (i)n+1
kr→∞ kr kr
Wronskian relations for spherical Bessel functions:
1
jν (r) nν′ (r) − jν′ (r) nν (r) =
r2
(1)′ (1) i
jν (r) hν (r) − jν′ (r) hν (r) = 2
r
Legendre function of the first kind:

X N  
(−1)m (ν + m)! 1 − x m sin νπ
Pν (x) = −
(m!)2 (ν − m)! 2 π
m=0
X∞  
(m − 1 − ν )!(m + ν )! 1 − x m
·
(m!)2 2
m=N+1

Legendre function of the second kind:

  Xn  
1 1+x (−1)m (n + m)! 1−x m
Qn (x) = Pn (x) ln − g(n) + g(m)
2 1−x (m!)2 (n − m)! 2
m=1

Orthogonality properties of zonal harmonics:

Z π
Pn (cos θ ) Pm (cos θ ) sin θ d θ = 0, m,n
0
Z π
2
[Pn (cos θ )]2 sin θ d θ =
0 2n + 1
SUMMARY 575

Important Equations (continued)

Orthogonality properties of spherical harmonic functions:

Z 2π Z π
Unm (θ , φ ) U pq (θ , φ ) sin θ d θ d φ
0 0

 4π

 m = 0 (only even functions),
 2n + 1
=

 2π (n + m)!

 m,0
(2n + 1)(n − m)!

Resonant frequency of a spherical cavity for TE-to-r and TM-to-r modes:

xnp
( f )TE
mnp = √ ,
2π a µε
x′np
( f )TM
mnp = √
2π a µε

Transcendental equation for determination of resonant modes for two concentric

spherical cavities:
(n) (1) (n) (n) (1) (n)
ĵn (k0 Re ) ĥn (k0 Ri ) − ĵn (k0 Ri ) ĥn (k0 Re ) = 0 .

Transcendental equation for determination of resonant modes for TM and TE

modes of a spherical dielectric resonator:
r r (1)
εs ĵn (ks a) εb ĥn (kb a)
= , (TM modes)
µs ĵn′ (ks a) µb ĥ(1)′
n (kb a)
r r (1)
µs ĵn (ks a) µb ĥn (kb a)
= (TE modes)
εs ĵn′ (ks a) εb ĥ(1)′
n (kb a)
576 Chapter 8 Spherical Wave Functions and Their Applications

Important Equations (continued)

Plane-wave expansion in terms of spherical wave functions:

Z π
2an
jn (kr) = eikr cos θ Pn (cos θ ) sin θ d θ ,
2n + 1 0
X∞
(−1)m (n + m)!
jn (kr) = 2n (kr)n (kr)2m
m! (2n + 2m + 1)!
m=0

Addition theorem for spherical wave functions:

(1)
h0 (k|r, r′ |) =
 ∞ n

 XX (2n + 1)(n − m)! (1)′


 αm jn (kr) hn (kr) Pnm (cos θ ′ ) Pnm (cos θ ) cos[m(φ − φ ′ )]

 (n + m)!

 n=0 m=0

 r ≤ r′ ,
 X∞ X n

 (2n + 1)(n − m)! (1)


 αm hn (kr) jn′ (kr) Pnm (cos θ ′ ) Pnm (cos θ ) cos[m(φ − φ ′ )]

 (n + m)!

 n=0 m=0
 r ≥ r′

The backscatter radar cross section of a metallic sphere:

(1) (1)′
(i)n (2n + 1) ĵn′ (ka) ĥn (ka) − ĵn (ka) ĥn (ka)
an − bn = (1)′ (1)
n(n + 1) ĥn (ka) ĥn (ka)
(i)n (2n + 1)
= (1)′ (1)
n(n + 1) ĥn (ka) ĥn (ka)

Low-frequency backscatter radar cross section of a metallic sphere:

1
lim , ≈ −ka ,
ka→0 ĥ(1) (ka)
1
1
lim ≈ (ka)2
ka→0 ĥ(1)′ (ka)
1

High-frequency backscatter radar cross section of a metallic sphere:

9λ 2
σcLF = (ka)6
4π
SUMMARY 577

Important Equations (continued)

The backscatter radar cross section of a dielectric sphere:

p dPn1 (x) 1 1 1 dPn (x) n(n + 1)
lim 1 − x2 = lim √ Pn1 (x) = lim − = (−1)n
x→−1 dx x→−1 2 1−x2 x→−1 2 dx 2

with an and bn given by

 
(Eθi + Eθs )r=a = Eθd r=a ,
 
(Eφi + Eφs )r=a = Eφd r=a

Transcendental equations for TM-to-r and TE-to-r modes for wave propagation in
a conical waveguide:
(
(1) cos mφ ,
Ar (r, θ , φ ) = Pνmnm (cos θ ) ĥνnm (kr)
sin mφ ,

∞ ∞
(
−1 X X m m (1) m cos mφ ,
Er = 2
amn νn (νn + 1) ĥνnm (kr) Pνnm (cos θ )
iω µε r sin mφ ,
m=0 n=1
∞ ∞
(
k sin θ X X (1)′ cos mφ ,
Eθ = amn ĥνnm (kr) Pνm′nm (cos θ )
iω µε r sin mφ ,
m=0 n=1
∞ X ∞
(
−k X (1)′ − sin mφ ,
Eφ = mamn ĥνnm (kr) Pνmnm (cos θ )
iω µε r sin θ cos mφ ,
m=0 n=1
∞ X ∞
(
1 X (1) − sin mφ ,
Hθ = mamn ĥνnm (kr) Pνmnm (cos θ )
µ r sin θ cos mφ ,
m=0 n=1
∞ ∞
(
sin θ X X (1) cos mφ ,
Hφ = amn ĥνnm (kr) Pνm′nm (cos θ )
µr sin mφ .
m=0 n=1
578 Chapter 8 Spherical Wave Functions and Their Applications

Important Equations (continued)

Transcendental equations for TM-to-r and TE-to-r modes for wave propagation in
a conical waveguide:

Aν = Qνm (cos θ1 ) ,
Bν = −Pνm (cos θ1 )

1 ∂ Amr 
Eφ = =0
ε r ∂ θ θ =θ1 and θ2
Input impedance of a biconical antenna:
 
η cot(θ1 /2)
Zin = ln , θ2 < θ1
2π cot(θ2 /2)

Transcendental equations for TM-to-r and TE-to-r modes for a θ –φ spherical

waveguide:
mπ /φ0
aν = Qν (cos θ1 ) ,
mπ /φ0
bν = −Pν (cos θ1 )

mπ /φ0 mπ /φ0 mπ /φ0 mπ /φ0

Qνnm (cos θ1 ) Pνnm (cos θ2 ) − Pνnm (cos θ1 ) Qνnm (cos θ2 ) = 0 .
 PROBLEMS 579

Important Terms Provide definitions or explain the meaning of the following terms:
θ –φ spherical waveguide radar cross section (RCS)
addition theorem for spherical wave functions RCS of dielectric spheres
associated Legendre functions of the first kind RCS of metallic spheres
associated Legendre functions of the second kind Schelkunoff spherical Bessel functions
biconical structure spherical Bessel functions of the first kind
coaxial biconical antenna spherical Bessel functions of the second kind
concentric spherical cavity spherical cavity
conical horn antenna spherical dielectric resonator
conical waveguide spherical Hankel functions of the first kind
dipole spherical Hankel functions of the second kind
Fourier-Legendre expansion spherical harmonic functions
Legendre functions spherical wave functions
Legendre polynomial transverse electric to r (TE-to-r)
monopole transverse magnetic to r (TM-to-r)
multipole Wronskian for spherical Bessel functions
quadrupole zonal harmonic functions

PROBLEMS

8.1 Find the Fourier-Legendre expansion of

1
f (θ , φ ) = δ (θ − θ0 ) δ (φ − φ0 )
sin θ
and plot the truncated approximation of f (θ , φ ) for N = 10.
8.2 Consider a function defined by
(
1 0 < θ ≤ π /2 ,
f (θ , φ ) =
0 π /2 < θ ≤ π .

Determine the Fourier-Legendre coefficients of f (θ , π ).

8.3 Consider an array of a dipole and a quadrupole as shown Fig. 8-14(c) and Fig. 8-14(e).
Choose the ratio of the currents flowing on the dipole and quadrupole so as to maximize the
front-to-back ratio of the radiated field.
8.4 A quadrupole is formed from two dipoles, as shown in Fig. P8.4. Find the magnetic
vector potential of such a source. Assuming the observation point is in the far-field region,
find the far-field expression for the resulting electric and magnetic fields.
580 Chapter 8 Spherical Wave Functions and Their Applications

I dl

y
x

Figure P8.4: A quadrupole placed along the z-axis. The separations between current elements
(δ1 and δ2 ) are much smaller than a wavelength.

Cavity

8.5 Consider a hemispherical metallic cavity of radius a.

(a) Determine the dominant TM-to-r mode and the expression for its resonant frequency.
(b) Find the Q of the dominant mode.
(c) Compare Q of part (b) with that of a spherical cavity.

8.6 For a spherical wedge cavity of radius r with PEC boundaries as shown in Fig. P8.6, find
the scalar potentials for TE-to-r and TM-to-r and the corresponding resonance frequencies.

Figure P8.6: A spherical sector cavity formed from two φ -constant planes and an r-constant
surface.
PROBLEMS 581

8.7 Consider a hemispherical dielectric with radius a, permittivity εd , and permeability µd

placed on an infinite ground plane. Find the natural resonant frequencies of such a dielectric
resonator.

8.8 Consider a cavity lying between concentric conducting spheres at r = a and r = b (with
b > a). Show that the characteristic equation for TM-to-r modes is given by

ĵn′ (kb) n̂′n (kb)

= ,
ĵn′ (ka) n̂′n (ka)
and for TE-to-r modes it is given by

ĵn (kb) n̂n (kb)

= .
ĵn (ka) n̂n (ka)

8.9 Consider a spherical cavity of radius b with PMC boundary partially filled with a
medium of permittivity εr and permeability µr for 0 < r < a. A PEC plate is inserted inside
the cavity at φ = φ0 and 0 < r < b, as shown in Fig. P8.9. Assuming that φ0 = 0, find the
characteristic equations for TM-to-r and TE-to-r modes.

ε0, μ0
εr, μr
y
x

Figure P8.9: The configuration of a spherical cavity of radius b made from a PMC having a
concentric dieletric sphere of radius a and constitutive parameters εr and µr . The cavity also has
a φ -constant metallic septum.
582 Chapter 8 Spherical Wave Functions and Their Applications

Scattering

8.10 Consider a conducting sphere of radius a enclosed in a dielectric sphere with radius b,
as shown in Fig. P8.10. Suppose this object is illuminated by an x-polarized plane wave
propagating along the +z-axis. Find the scattered electric and magnetic fields.

ε, μ
y
x

kˆ i
Ei Hi

Figure P8.10: The geometry of a dielectric-coated metallic sphere illuminated by a plane wave
propagating along the +z-axis. The radius of the metallic sphere is a and that of the dielectric
sphere is b.

8.11 Consider a radially directed electric current filament (short-dipole) near a metallic
sphere. The goal in this problem is to find the radiated field in the far-field region using
reciprocity. Suppose a small dipole is placed at the observation point in the far-field region.
The field of this dipole can be approximated, locally, by a plane wave. The scattered field
produced by the plane wave at the location of the original source can be computed. According
to the reciprocity theorem, the field of the short dipole near the metallic sphere can be
computed from the plane-wave scattered field. Derive the expression for the radiated field
from a radially directed small dipole in the presence of a metallic sphere of radius a.
PROBLEMS 583

8.12 Consider a conducting sphere of radius a. A loop of uniform current I = I0 φ̂

φ and radius
R > φ is placed concentric with the metallic sphere, as shown in Fig. P8.12. Show that the
radiation field can be obtained from
∞
η I jkr X 2n + 1 −n
Eφ = e j An Pn1 (0) Pn1 (cos θ ) ,
r 2n(n + 1)
n=1

where
(1)′ Jˆn (ka) N̂n′ (kR) − N̂n(ka) ĵn′ (kR) (1)
A−1
n = Ĥn (kR) − Ĥn (kR) .
Jˆn (ka) N̂n (kR) − N̂n (ka) Jˆn (kR)
Hint: Find the electric and magnetic fields from the electric vector potential and then apply
the boundary and discontinuity conditions.

Figure P8.12: A metallic sphere of radius a is excited by a concentric loop of constant electric
current of radius R.

8.13 Find the scattered field for a dielectric sphere with parameters εs and µs and radius a,
placed above an infinite ground plane whose center is at a height of h > a above the ground
plane. This object is illuminated by a plane wave given by

Ei = x̂ E0 eik sin θi y−cos θi z .

Hint: Use image theory and ignore the mutual scattering between the sphere and its image.
584 Chapter 8 Spherical Wave Functions and Their Applications

Conical Structures

8.14 For a biconical waveguide with a flare angle θ , as shown in Fig. P8.14, calculate the
electric vector potential Amr and the magnetic vector potential Ar . The dominant mode of this
structure is a TEM. Show that Ar for the TEM mode is defined as
(1)  
(2) θ
Ar = Q0 (cos θ ) ĥ0 (kr) = ln cot (∓i)e±ikr .
2

Figure P8.14: A metallic biconical structure with θ2 = π − θ1 .

8.15 Consider a coaxial biconical waveguide. Assuming the conductivity of the metal used
to make a biconical waveguide suppoorting a TEM mode is σ , show that an expression for
the attenuation rate as a function of radial distance is given by
1 csc θ1 + csc θ2 1
α= .
2ησ cot θ1 /2 r
log
cot θ2 /2
PROBLEMS 585

8.16 Consider an infinitesimal current filament I dℓ placed at the tip of a metallic cone
specified by exterior angle θ1 , as shown in Fig. P8.16. Find the radiated electric field for this
problem.

I dl
y
x

Figure P8.16: A metallic cone excited by a z-directed infinitesimal current at its tip.
APPENDIX A:
PROPERTIES OF COMPLEX
FUNCTIONS

In this appendix important theorems for complex functions are reviewed. To start, basic
definitions and properties of complex functions are provided. In general, a function of
complex variable z within a domain Ω in the complex z-plane is a collection of ordered
complex pairs (z, w), where the first complex number (z) is the independent variable and
the second complex number (w) is known as the dependent variable. In complex notation, this
relation between the dependent and independent variables is given by

w = f (z) . (A.1)

Function f (z) with domain Ω is said to be continuous at z0 ∈ Ω if f (z0 ) , ∞ and for any
positive real number ε > 0 and for all z in the neighborhood of z0 for which

| f (z) − f (z0 )| < ε , (A.2)

one can find a real and positive number δ such that

|z − z0 | < δ . (A.3)

In this case it can be shown that

lim f (z) = f (z0 ) . (A.4)
z→z0

A function w = f (z) is said to be differentiable in some ε -neighborhood of a point z0 if the

following limit exists and is not infinite:

f (z0 + h) − f (z0 )
f ′ (z0 ) = lim . (A.5)
h→0 h
Here h, in general, is a complex number and the requirement for differentiability mandates that
the same limit should exist independent of the path in the complex plane along which h → 0.
Since this definition for differentiation is similar to that for real functions, all differentiation
properties, such as linearity and the rules for products and ratios of differentiable functions,
apply. Also, the chain rule applies.

586
A-1 Cauchy–Riemann Conditions 587

A-1 Cauchy–Riemann Conditions

Consider a differentiable function w = f (z) at z0 = x0 + iy0 . This function may be expressed

in the following manner:
f (z) = u(x, y) + iυ (x, y) , (A.6)
where u and υ are real functions of two variables. The Cauchy–Riemann theorem states that
if f is differentiable at z0 , the partial derivatives ∂∂ ux , ∂∂ uy , ∂∂υx , and ∂∂υy exist at (x0 , y0 ), and

∂ u(x0 , y0 ) ∂ υ (x0 , y0 )
= (A.7a)
∂x ∂y
and
∂ u(x0 , y0 ) ∂ υ (x0 , y0 )
=− . (A.7b)
∂y ∂x
A function w = f (z) is said to be analytic in some ε -neighborhood of z0 if and only if
it is differentiable in that neighborhood. For example, polynomial functions are analytic
everywhere, and rational functions are analytic everywhere except for points where the
denominator becomes zero.
It is interesting to note that if f (z) is analytic, using Eqs. (A.7a) and (A.7b), then it can be
shown that

∇2t u(x, y) = 0 (A.8a)

and
∇2t υ (x, y) = 0 , (A.8b)

where
∂2 ∂2
∇2t = +
∂ x2 ∂ y2
is a 2-D Laplacian operator.
It is noted that analyticity at one point guarantees that the function has all higher-order
derivatives, and therefore all partial derivatives of u(x, y) and υ (x, y) exist and are continuous.

A-2 Conformal Mapping

A function w = f (z) can be considered a mapping from z = x+ iy in z-plane into w = u + iυ in

w-plane. If f (z) is analytic in domain Ω and f ′ (z0 ) , 0 for a point z0 in Ω, then the mapping
is said to be conformal, as it preserves angles between any two curves crossing at z0 in the
z-plane and those of the mapped curves at w0 = f (z0 ) in magnitude and sense. Small segments
of curves passing through z0 are rotated by arg[ f ′ (z0 )] and a curve length is scaled by | f ′ (z0 )|.
It is also interesting to note that the Laplacian operator is unchanged under a conformal
mapping. If φ (x, y) represents a real function of x and y that satisfies Laplace’s equation
588 Appendix A Properties of Complex Functions

(∇2t φ (x, y) = 0), then it can be shown that by conformal mapping from z-plane to w-plane
 2 
∂ 2φ ∂ 2φ 1 ∂ φ ∂ 2φ
∇2t φ (u, υ ) = + = + =0. (A.9)
∂ u2 ∂ υ 2 | f ′ (z)|2 ∂ x2 ∂ y2

A-3 Branch Cut and Branch Point

In the definition of complex functions, multiple-valued functions are not allowed. Functions,
in general, can be one-to-one and many-to-one. Functions like ez or z2 are examples of many-
to-one functions. For such functions the inverse cannot be defined uniquely, as they become
multiple-valued functions. However, if we restrict the domain of many-to-one functions in
such a way as to make them one-to-one, then an inverse exists. For example, noting that

ez+i2nπ = ez , (A.10)

it is obvious that f (z) = ez is a many-to-one function. But if we restrict the domain of the
function so that −π ≤ Im(z) ≤ π , then the function becomes one-to-one. The inverse of this
function is
g(z) = f −1 (z) = ln z = ln |z| + iφ , (A.11)
where φ = arg(z) and π ≤ φ ≤ π . Depending on what restrictions can be placed on φ , to make
the function single-valued, many logarithm functions can be defined. Based on the above
definition, the logarithm function is discontinuous on the negative real axis. Also, at z = 0 the
function is undefined; otherwise ln(z) is analytic. The negative real axis where the function is
discontinuous is called a branch cut, and z = 0 is called a branch point. It should be noted
that the branch cut depends on the restriction put on φ or definition used for φ to make the
function single-valued. In fact the branch cut does not have to be a straight line and can be an
arbitrary curve passing through the branch point.
Other logarithm functions can be defined from

ln(z) = ln |z| + i(φ + 2nπ ) , n = 0, ±1, ±2, . . . (A.12)

The function for n = 0 is referred to as the principal branch. Now consider the nth value of
the logarithm function, the range of argument of z(φ ) must satisfy

(2n − 1)π < φ < (2n + 1)π . (A.13)

The lines corresponding to φ = (2n − 1)π are called the branch cuts, as shown in Fig. A-1.

A-4 Cauchy’s Theorem

A region in the complex z-plane is called a simply connected domain if every enclosed
contour within this domain encloses only points that are in the domain. Figures A-2(a) and
A-2(b) show a simply connected and multiply connected domains in the complex z-plane.
Cauchy’s theorem states that for a complex function f (z) that is analytic in a simply connected
A-4 Cauchy’s Theorem 589

y
Complex z-plane

x
Branch cut Branch point

Figure A-1: Configuration of a branch cut and the branch point of a logarithm function.

y y
Complex z-plane Complex z-plane

x x

(a) (b)

Figure A-2: Configuration of (a) a simply connected domain and (b) a multiply connected
domain in a complex z-plane.
590 Appendix A Properties of Complex Functions

domain Ω and for any simple closed contour C in Ω, we have

I
f (z) dz = 0 . (A.14)
C

A-5 Cauchy Formulas

Consider a complex function f (z) that is analytic within and on a simple contour C. Then for
any point z within C, we have
I
1 f (ζ ) d ζ
f (z) = , (A.15)
2π i C+ ζ − z
where the direction of integration is in the positive sense (counterclockwise). The result of the
first Cauchy formula given by Eq. (A.15) can be used to show that
I
′ 1 f (ζ ) d ζ
f (z) = . (A.16)
2π i C+ (ζ − z)2

By differentiating Eq. (A.16), the second derivative of f (z) can be obtained from
I
2 f (ζ ) d ζ
f ′′ (z) = . (A.17)
2π i C+ (ζ − z)2
In general, it can be shown that
I
(n) n! f (ζ ) d ζ
f (z) = . (A.18)
2π i C+ (ζ − z)n+1

In fact, Eq. (A.18) can be used to prove that if f (z) is an analytic function in a neighborhood
around z0 , then f (z) has all derivatives at z0 .

A-6 Poles and Residues

The pole of a complex function f (z) is defined as a point at which the function is singular (the
function becomes infinite). However, the function is analytic in a neighborhood of this point.
In this neighborhood the function can be expanded in terms of its Laurent series given by
∞
X M
X bm
f (z) = an (z − z0 )n + . (A.19)
(z − z0 )m
n=0 m=1

The second term in this expansion is called the principal value of f (z) about z0 . In situations
where the principal part of f (z) about point z0 has an infinite number of terms (m = ∞), the
point is called an essential singular point.
A-7 Jordan’s Lemma 591

The coefficient b1 is known as the residue of f (z) at z0 and is given by

Z
1
b1 = f (z) dz , (A.20)
2π i C
where C is a small closed cntour around z0 and the direction of integration is in the positive
sense (counterclockwise).
The residue theorem states that for an analytic function f (z) within and on a contour C
except for a finite number of singular points z1 , z2 , z3 , . . . , zn interior to C, the integral
I n
X
f (z) dz = 2π i Ki , (A.21)
C+ i=1

where Ki is the residue associated with the singular point zi .

A-7 Jordan’s Lemma

Jordan’s lemma is often used in connection with the evaluation of integrals of function on the
real axis. In such problems the real axis is closed using a semicircle of radius R (CR ) centered
at the origin in the upper or lower half-plane in the complex z-plane to form a closed contour
and then let R go to infinity. If the value of the integral over CR is known, by using the residue
theorem, the value of the integral over the real axis can be obtained. This theorem applies to
integrands of the form f (x) eiax for a positive parameter a. For a continuous function f (z) on
CR that satisfies
lim | f (Reiφ )| = 0 , (A.22)
R→∞
Jordan’s lemma states that Z
IR = lim f (z) eiaz dz = 0 . (A.23)
R→∞ C
R

In this case,
Z ∞ n
X
f (x) eiax dx = 2π i Ki eiazi . (A.24)
−∞ i=1

where Ki is the residue of f (z) at a finite number of nonreal singular points of f (z) in the
upper half-plane.
APPENDIX B:
METHOD OF STEEPEST
DESCENT

In radiation and scattering problems we often encounter a certain type of integral of the form
Z
I = g(z) eλ h(z) dz . (B.1)
C

With some appropriate assumptions, this integral can be evaluated approximately using the
method of steepest descent. Here, C represents a fixed path in the complex z-plane on which
g(z) and h(z) are analytic in some region Ω that includes C. Assuming the integral exists
for Re[λ ] > 0 and that its value does not change if the contour C is deformed, the method
of steepest descent consists of changing the path of integration to a new path over which the
integrand changes most rapidly.
Let us assume that
h(z) = u(x, y) + iυ (x, y) , (B.2)
and consider a point z0 = x0 + iy0 in the domain Ω. A direction away from z = z0 along which
u(x, y) decreases from u(x0 , y0 ) is called a direction of descent, whereas if u(x, y) increases
from u(x0 , y0 ), the direction is called a direction of asscent. Consider a path C′ in Ω that starts
at z0 and ends at z1 . If the tangent to C′ is always in a direction of descent (ascent), then C′
is called a path of descent (ascent). A path of steepest descent (ascent) is a path on which
variations δ u of u are maximum. Noting that for

δ h = h(z) − h(z0 ) = δ u + iδ υ , (B.3)

it follows that
|δ h|2 = |δ u|2 + |δ υ |2 . (B.4)
To maximize variations of u for a fixed |δ h|2 , variations of υ must be set to zero:

δυ = 0 . (B.5)

Hence the path of steepest descent or ascent that goes through z0 = x0 + iy0 is given by

υ (x, y) = υ (x0 , y0 ) . (B.6)

Now consider the exponent function of the integrand of Eq. (B.1) evaluated along the steepest
descent or ascent path
eλ h(z) = eλ [u(x,y)+iυ (x0 ,y0 )] . (B.7)

592
B-1 Saddle Point 593

The absolute value of h(z), obviously, takes its maximum variation to lower values than |h(z0 )
if Eq. (B.6) is chosen to be the steepest descent path, or to higher values than |h(z0 )| if
Eq. (B.6) is chosen to be the steepest asscent path.

B-1 Saddle Point

As mentioned earlier, the path of the contour in Eq. (B.1) can be changed if g(z) and h(z)
are analytic functions. The question remains as to which new contour can provide an accurate
approximation of Eq. (B.1). In this section, it is shown that if the new contour passes through
a point z0 for which
h′ (z0 ) = 0 , (B.8)
the steepest descent path that goes through z0 provides the maximum variation of eλ h(z) to
lower values. The point z0 that satisfies Eq. (B.8) is known as the saddle point of function
h(z). The reason for choosing the name “saddle point” for z0 is that h′ (z0 ) = 0 does not
specify a maximum or minimum for |h(z)|, u(x, y), or υ (x, y).
To show that for an analytic function, |h(z)| does not assume its maximum or minimum
at z0 where h′ (z0 ) = 0, let us consider the Cauchy integral formula
I
h(z) dz
= 2π i h(z0 ) (B.9)
C z − z0

for a point z0 inside C. If C is a circle centered at z0 with radius ρ , then on C

z − z0 = ρ eiφ . (B.10)

Using Eq. (B.9), it can be shown that

Z  Z 2π
1  2π 
≤ 1
|h(z0 )| = h(z) i d φ |h(z0 + ρ ′ eiφ )| d φ . (B.11)
2π  0  2π
0

If M represents the maximum value of |h(z)| on the circle, then the inequality in Eq. (B.11)
can be written as Z 2π
1
|h(z0 )| ≤ M dφ = M (B.12)
2π 0
for all values of z0 inside C. This result can easily be generalized to any noncircular contour C.
Equation (B.12) shows that |h(z)| assumes its maximum on the contour and not inside the
contour. This is true for all z0 points inside the contour including a point for which h′ (z0 ) = 0.
For a contour in which h(z) has no zero values, 1/h(z) is analytic, and therefore |1/h(z)|
has no maximum inside C. As a result, |h(z)| has no minimum value inside C. For this function
z0 for which h(z0 ) = 0 represents a saddle point, as shown in Fig. B-1. This theorem also holds
for the real and imaginary parts of h(z). That is, the points of zero derivative of h(z) are saddle
points for u and υ functions:

u(x0 , y0 ) < umax on C

594 Appendix B Method of Steepest Descent

Figure B-1: For an analytic function h(z) = u(x, y) + iυ (x, y), the point z0 for which h′ (z0 ) = 0
does not present a maximum or a minimum of |h(z)|, u(x, y), or υ (x, y). It simply represents a
saddle point for |h(z)|, u(x, y), or υ (x, y). The maximum and minimum of such functions in a
simply connected domain occur on its boundary.

and
υ (x0 , y0 ) < υmax on C .

B-2 Integration along the Steepest Descent Path

Considering the integral represented by Eq. (B.1) and for contour C shown in Fig. B-2, we
first identify the saddle point z0 for function h(z) for which

dh(z) 
=0.
dz z=z0

Let us also assume that 

d 2 h(z) 
= Aeiα
dz2 z=z0
and
lim g(z) eλ h(z) = 0 ,
z→C∞

where C∞ is a circle with a very large radius ρ → ∞. Then according to Cauchy’s theorem,
Z Z
g(z) eλ h(z) dz + S(z) eλ h(z) dz = 0 .
C SDP

The integral on the steepest descent path (SDP) can be evaluated by expanding h(z) in terms
of its Taylor series expansion and only retaining the second term of expansion in the exponent.
B-2 Integration along the Steepest Descent Path 595

Complex z-plane

Figure B-2: The path of integration method (C) for the saddle point method is deformed to the
steepest descent path (SDP) noting that the integrand contributionon C∞ is zero. The integration
on SDP is approximated by retaining only the second term of the Taylor series expansion of h(z)
and then using the Fresnel integral value.

After some algebraic manipulations, the approximate value of the integral is found to be
Z √
λ h(z) − 2π
g(z) e dz ≈ p g(z0 ) eλ h(z0 ) e−i(α −π )/2 . (B.13)
C λ |h′′ (z )|
0
APPENDIX C:
USEFUL VECTOR IDENTITIES,
OPERATORS, AND COORDINATE
TRANSFORMATIONS

Table C.1: Useful vector identities.

A · (B × C) = B · (C × A) = C · (A × B) (C.1)

A × (B × C) = (A · C)B − (A · B)C (C.2)

∇(φ ψ ) = φ ∇ψ + ψ ∇φ (C.3)

∇ · (ψ A) = A · ∇ψ + ψ ∇ · A (C.4)

∇ × (ψ A) = ψ ∇ × A − A × ∇ψ (C.5)

∇(A · B) = A × (∇ × B) + B × (∇ × A) + (A · ∇)B + (B · ∇)A (C.6)

∇ · (A × B) = B · (∇ × A) − A · (∇ × B) (C.7)

∇ × (A × B) = A(∇ · B) − B(∇ · A) + (B · ∇)A − (A · ∇)B (C.8)

1
(A · ∇)A = 2 ∇|A|2 + (∇ × A) × A (C.9)

∇ × ∇φ = 0 (C.10)

∇ · (∇ × A) = 0 (C.11)

∇ · ∇φ = ∇2 φ (C.12)

∇ × (∇ × A) = ∇(∇ · A) − ∇2A (C.13)

596
 597

Table C.2: Gradient, divergence, curl, and Laplacian operators.

Cartesian Coordinates Cylindrical Coordinates Spherical Coordinates
(x, y, z) (ρ , φ , z) (r, θ , φ )
∂f ∂f ∂f ∂f 1∂f ∂f ∂f 1∂f 1 ∂f
Gradient ∇ f x̂ + ŷ + ẑ ρ+
ρ̂ φ+
φ̂ ẑ r̂ + θ+
θ̂ φ
φ̂
∂x ∂y ∂z ∂ρ ρ ∂φ ∂z ∂r r ∂θ r sin θ ∂ φ
1 ∂ (r2 Ar )
r2 ∂ r
∂ Ax ∂ Ay ∂ Az 1 ∂ (ρ Aρ ) 1 ∂ Aφ ∂ Az 1 ∂ (Aθ sin θ )
Divergence ∇ · A + + + + +
∂x ∂y ∂z ρ ∂ρ ρ ∂φ ∂z r sin θ ∂θ
1 ∂ Aφ
+
r sin θ ∂ φ
     
∂ Az ∂ Ay 1 ∂ Az ∂ Aφ 1 ∂ ∂ Aθ
− x̂ − ρ
ρ̂ (Aφ sin θ ) − r̂
 ∂y ∂z  ρ ∂φ
 ∂z  r sin θ  ∂ θ ∂φ 
∂ Ax ∂ Az ∂ Aρ ∂ Az 1 1 ∂ Ar ∂
Curl ∇ × A + − ŷ + − φ
φ̂ + − (rAφ ) θ̂ θ
∂z ∂x ∂z ∂ρ r sin θ ∂ φ ∂r
     
∂ Ay ∂ Ax 1 ∂ (ρ Aφ ) ∂ Aρ 1 ∂ ∂ Ar
+ − ẑ + − ẑ + (rAθ ) − φ
φ̂
∂x ∂y ρ ∂ρ ∂φ r ∂r ∂θ
 
1 ∂ 2 ∂f
r
r2 ∂ r ∂r
   
∂2 f ∂2 f ∂2 f 1 ∂ ∂f 1 ∂2 f ∂2 f 1 ∂ ∂f
Laplacian ∇2 f + + ρ + 2 + + 2 sin θ
∂ x2 ∂ y2 ∂ z2 ρ ∂ρ ∂ρ ρ ∂ φ 2 ∂ z2 r sin θ ∂ θ ∂θ
1 ∂2 f
+ 2 2
r sin θ ∂ φ 2

Coordinate Transformation
Consider a field vector F that can be expressed in Cartesian, cylindrical, and spherical
coordinates as

F = fx x̂ + fy ŷ + fz ẑ ,
F = fρ ρ̂ φ + fz ẑ ,
ρ + fφ φ̂
F = fr r̂ + fθ θ̂
θ + fφ φ̂
φ.

The components of F in these coordinates systems are related to each other at a given point
p = (x x̂ + y ŷ + z ẑ) = ρ ρ̂
ρ + z ẑ = r r̂ as shown below:

   √x √ y2

0 f 
fρ x2 +y2 x +y2
   √y  x
Cartesian to cylindrical: fφ = − 2 2 √x 0  f y 
x +y x2 +y2
fz 0 0 1 fz
598 Appendix C Useful Vector Identities, Operators, and Coordinate Transformations

    
fx cos φ − sin φ 0 fρ
Cylindrical to Cartesian:  f 
y =  sin φ cos φ 0   f φ
fz 0 0 1 fz
 
√ x √ y √ z
  x2 +y2 +z2 x2 +y2 +z2 x2 +y2 +z2  
fr  √  fx
 yz − x2 +y2 
Cartesian to spherical:  fθ  =  xz  
 √(x2 +y2 )(x2 +y2 +z2 ) √(x2 +y2 )(x2 +y2 +z2 ) √x2 +y2 +z2 )  fy
fφ   fz
√−y √x 0
x2 +y2 x2 +y2
    
fx sin θ cos φ cos θ cos φ − sin φ fr
Spherical to Cartesian:  fy  =  sin θ sin φ cos θ sin φ cos φ   fθ 
fz cos θ − sin θ 0 fφ

Integral Identities
Surface integral for an arbitrary open surface bounded by the contour C:
" I
∇ × A · ds = A ·ℓℓ
S C
" I
n̂ × ∇φ ds = φ ds
S C

Volume integral for an arbitrary volume bounded by closed surface S:

$
∇ · A dv = A · ds
V S
$
∇ × A dv = − A × ds
V
$
∇φ dv = φ ds
V

Proofs of Integral Relations

Proofs of some of the useful integral identities are provided here:

a. Show that $
∇ × F dv = − F × ds .
V
Proof: Starting from the vector calculus identity

∇ · (G × F) = F · ∇ × G − G · ∇ × F
 599

and assuming G is a constant vector, we have

$ $
∇ · (G × F) dv = −G · ∇ × F dv = (G × F) · ds ,
V V

but
G × F · ds = G · F × ds .
Since G is an arbitrary vector, it is now obvious that
$
∇ × F dv = − F × ds . (C.14)
V

b. Show that $
∇ψ dv = ψ ds .
V S
Proof: In a similar manner, we first note that

∇ · (ψ G) = G · ∇ψ + ψ ∇ · G .

Then for some arbitrary constant G,

∇ · (ψ G) = G · ∇ψ . (C.15)

Taking the volumetric integral from both sides of Eq. (C.15), we have
$ $
∇ · (ψ G) dv = G · ψ ds = G · ∇ψ dv .
V V

Hence $
∇ψ dv = ψ ds . (C.16)
V S

c. Show that, for a differentiable scalar function ψ and a surface S having C as its contour,
Z I
n̂ × ∇ψ ds = ψ dℓℓ . (C.17)
S C

Proof: According to Stokes’ theorem, for a constant vector a,

Z I I
∇ × (ψ a) · ds = ψ a · dℓ = a · ψ dℓℓ .
ℓ (C.18)
S C C

Also,
∇ × (ψ a) = ∇ψ × a + ψ ∇ × a = ∇ψ × a , (C.19)
600 Appendix C Useful Vector Identities, Operators, and Coordinate Transformations

since ∇ × a = 0. Using Eq. (C.19), we have

Z Z Z
∇ × (ψ a) · ds = (∇ψ × a) · n̂ ds = a · (n̂ × ∇ψ ) ds . (C.20)
S S S

From Eq. (C.18) and Eq. (C.20),

Z I
(n̂ × ∇ψ ) ds = ψ dℓℓ .
S C
Index

Note: Page numbers in italics refer to Bessel function of the second kind, 407–
occurrences in the end-of-chapter problems. 409, 411, 487
bianisotropic, 18, 99, 247–254
θ –φ spherical waveguide, 569, 570, 578 biaxial, 18
2-D Green’s function for external problem, biconical structure, 503, 560–566, 584
449–451, 467–474 birefringent, 254
bistatic echo width, 456, 460, 462
addition theorem, 405, 449, 490 boundary conditions, 30–33, 67, 70, 74–76,
addition theorem for spherical wave func- 78, 83, 91–93, 98, 106, 303, 305,
tions, 503, 535–538, 576 307–310, 389, 395, 397–399, 403,
angular-sector waveguide, 405, 430–433, 583
437–441 boundary conditions for the potentials, 173–
anisotropic, 2, 17–18, 85, 99, 223, 241–245, 176, 218
247, 304, 329, 381 Bragg mirror, 285, 312
antenna directivity, 133–134, 146, 147, 151, Brewster angle, 261–264, 280, 301, 302
154, 155, 162, 198
antenna equivalent circuit, 108, 135–137, calculus of variations, 375–379, 382–387
141, 147–149, 197, 198 cavity damping factor, 354, 356, 391
antenna quality factor, 134–135, 151 cavity relaxation time, 354, 356, 391
antenna radiation pattern in receive and characteristic impedance, 118, 133, 151,
transmit modes, 192–194, 219 234, 265, 291, 310, 324, 326, 332,
antennas as two-port networks, 189, 191, 334, 358, 360, 435, 563, 565, 566
218 circuit models for waveguides, 314, 357–
associated Legendre functions of the first 362
kind, 510, 512–514, 553 circumferential waveguide, 405, 433, 435–
associated Legendre functions of the second 436
kind, 510, 513 cladding, 421, 423, 426, 494
attenuation rate, 314, 344, 347–348, 352, coaxial biconical antenna, 563, 565–566
372–374, 391, 400 complementary antenna, 162, 205, 210–212
axial ratio, 240 complementary theorem, 207–208
complex power flow density, 127, 151, 237
Babinet’s principle, 162, 205–214, 226 concentric spherical cavity, 575
bandgap material, 290 conduction current, 7, 14, 429
Bessel equation, 407, 411, 414, 487 conical horn antenna, 556, 558
Bessel function of the first kind, 407–409, conical waveguide, 503, 553–560, 577, 578
411, 413–416, 430, 449, 465, 468, constitutive relations, 14–28, 90, 99, 106,
487, 491 124, 223, 248, 249, 277

601
602 INDEX

continuous spectrum of plane waves, 234– drift current, 14

235, 292, 321, 446 duality relations, 65–67, 71, 83, 98, 112,
coordinate curve, 36 113, 121, 125, 140, 206, 210, 257,
coordinate surface, 36, 119, 152, 405, 406, 260, 282, 284, 319, 320, 360, 364,
468, 503, 533 505, 506
core, 421, 423, 426
critical angle, 262 effective area of a receiving antenna, 162,
curl, 6, 7, 41–43, 90, 143, 206, 229, 242 195–198, 219, 223
current density, 3, 63, 100, 104, 130, 173, effective permittivity, 125, 151
199, 331, 359, 382, 500 eigenfunction, 314, 368–371, 374–375, 377–
cutoff frequency, 314, 317–320, 339, 340, 380, 392, 399, 401
348, 361, 375, 379, 381–382, 390, eigenvalue, 314, 338, 353, 368–370, 375–
394, 395, 397, 399 382, 384, 392, 403
cutoff wave number, 314, 340, 342, 379 Eikonal equation, 266, 268, 270, 271, 274,
cylindrical cavity, 405, 425–430, 489, 494– 301, 302
496
electric energy density, 85, 88, 106, 108,
cylindrical wave functions, 409, 421, 436,
128, 236–238, 300
445–446, 448, 450, 487
electric field intensity, 3, 10, 14, 16, 32, 53,
cylindrical waveguide, 405, 414–420
63, 90, 104, 124, 128, 238, 241,
DB plane, 247 242, 253, 275, 340, 341, 418, 500
diamagnetism, 17 electric flux density, 3, 9, 15, 16, 32, 33, 49,
dielectric loss tangent, 130, 303, 352, 533 58, 90, 109, 124, 128, 176, 241,
dielectric plate waveguide, 314–320 247, 252, 253, 275
dielectric quality factor, 130, 151, 354, 531, electric reflection coefficient, 228, 260, 264,
533, 534 282, 301, 303
dielectric slab, 228, 288–290, 302, 310, 312 electric scalar potential, 108, 109, 111–113,
diffraction coefficients for metallic wedge, 117–119, 125–126
480–483, 492 electric susceptibility, 16, 18, 54
dipole, 108, 117, 134–136, 139, 142–147, electric transmission coefficient, 228, 303
151, 154–157, 192–194, 201, 203, electric vector potential, 108, 112, 125–126,
204, 210–212, 214, 215, 223, 308, 151, 158
445, 446, 539–541, 579, 582 electric-field integral equation (EFIE), 162,
Dirichlet’s boundary condition, 336, 337, 183, 219
346, 353, 367, 370 elementary wave functions, 409, 415, 450,
dispersion relation, 228, 246–248, 251, 252, 453
278, 291, 300 elliptic polarization, 238–240, 311
dispersive material, 18–19, 122 ellipticity angle for polarization, 240, 300
displacement current, 7, 14, 130, 359 energy density of plane waves, 228, 236–
distributed Bragg reflector (DBR), 285 238, 300
divergence theorem, 5, 8, 31, 32, 46, 47, 85, energy velocity, 237, 238
87, 129, 164, 165, 175, 179, 187, equation of continuity, 6, 20, 44, 45, 58,
372 177, 180
double-negative media, 290–294, 296, 298, equivalent source, 71, 92–97, 162, 166, 167,
302 170, 172, 178–181
INDEX 603

evanescent wave, 232, 234, 290, 292, 295– Green’s theorem, 162, 165–166
298 group velocity, 314, 342–344, 390
extinction theorem, 162, 170, 217 guided wavelength, 390
extraordinary wave, 228, 247, 252–254,
275, 276, 278–280, 300, 302 H-bent rectangular waveguide, 435
Hall effect, 20, 35–36
far-field distance, 118, 137–139, 145, 151, Hankel functions of the first kind, 407, 434,
154, 156, 157 449, 450, 453, 465, 466, 487
Faraday’s law, 2, 6–8, 10–14, 31, 49, 54, 63, Hankel functions of the second kind, 407,
83, 123, 143, 266, 277, 286 449, 465, 466, 487
Fermat’s principle, 262–264 harmonic functions, 405, 407, 447, 464
ferromagnetism, 17 Helmholtz equation, 162–176, 234, 236,
finite element, 315, 379, 380 254–264, 282, 379, 406, 407, 409,
flow of electromagnetic power, 86–90, 104, 415, 430
105, 133, 344–348 Hertz vector potential, 111, 126, 228, 255,
formal solution for 2-D Green’s function, 276–278, 314, 363, 395, 397, 441,
405, 439–441, 448 443, 445, 446, 468, 497, 498
forward propagation matrix, 284 Hertzian dipole, 108, 133–138, 145, 151,
Fourier representation of 2-D Green’s func- 154
tion, 446–448 Hertzian magnetic dipole, 139–144, 151
Fourier-Legendre expansion, 517–520, 579 homogeneous, 2, 3, 15, 31, 50, 55, 57, 68,
Fredholm integral equation, 172 99, 106, 151, 187, 228–238, 245–
frequency-independent antennas, 162, 213 246, 300, 397, 441–444, 448–451,
Friis transmission formula, 223 467–474
functional, 375–380, 382, 384, 385, 392 hybrid mode, 363, 397, 405, 425, 426, 494

gamma function, 411, 487 image of a point charge over a dielectric

gauge condition, 110, 117, 125, 151, 174, half-space medium, 78–80
277, 503, 504, 572 image of a point charge over a metallic
Gauss’ law for electricity, 2, 6, 8, 9, 20, 32, sphere, 72–78
33, 80, 83, 124, 235, 267, 287, 291 image of currents and charges over a planar
Gauss’ law for magnetism, 2, 6, 8, 32, 64, conductive surface, 68–71
267, 291 impermeability, 248
generalized eigenvalue problem, 381 impermittivity, 248, 250
good conductor, 264–266, 302 incident potential, 170, 172
gradient, 39, 43, 90, 170, 179, 268, 370 inhomogeneous, 2, 15, 90, 93, 168, 187,
Green’s first identity, 165, 217, 370, 376, 257, 266–274, 286–288
378 integral representation of Bessel function,
Green’s function, 162, 165, 167, 168, 170, 405, 409, 460–467, 491
171, 178, 179, 181, 182, 184, 186, isotropic homogeneous material, 15, 108,
217, 221, 493 109, 162, 245–246, 300, 312
Green’s function for angular-sector wave- isotropic reactive impedance surface, 330–
guide, 437–441, 448, 490 335
Green’s second identity, 165, 170, 174, 178,
217, 369, 370, 515, 516 k–β diagram, 340
604 INDEX

kDB coordinate, 228, 247–254 multipole, 538–542

Kramers-Krönig relations, 28–30, 45, 54
natural boundary condition, 377, 378
Laplacian, 43, 504 negative refractive index, 292–298
law of conservation of charge, 3, 5–6, 8, 45, negative refractive index lens, 292–298
75, 81 Neumann’s boundary condition, 336, 346,
Legendre functions, 503, 510, 512, 513, 353, 367, 370, 430
560, 561 nonuniform plane wave, 228, 233
Legendre polynomial, 511, 512, 514, 538 normalized duality relations, 66

magnetic charge density, 63–65, 98 ohmic loss, 88, 525

magnetic current density, 63, 65, 67, 80, optical fiber, 405, 421, 425, 426, 489, 494
100, 101, 102, 143 optical path length, 263, 271
magnetic field intensity, 3, 35, 63, 90, 104, order of Bessel function, 405, 407, 408, 412,
116, 124, 137, 252 415, 416, 468
magnetic flux density, 3, 8, 10, 12, 14–16, ordinary point, 3, 108
32, 35, 50, 51, 67, 90, 109, 124, ordinary wave, 228, 246, 252–254, 275,
141, 242, 247, 253 276, 278, 279, 300
magnetic Hertz vector potential, 112, 113, orthogonality of Bessel functions, 413–414,
151, 228, 255, 256, 276, 364, 539– 488
542, 544, 562, 579, 584
magnetic moment of current loop, 139–141 paramagnetic, 17, 143
magnetic reflection coefficient, 228, 259, perfect electric conductor (PEC), 68–72, 93,
280, 301 94, 101, 102, 104, 152, 153, 156,
magnetic scalar potential, 108, 112, 125– 176, 208, 210, 225, 264, 311, 325,
126 346
magnetic transmission coefficient, 228, 260 perfect magnetic conductor (PMC), 67, 71,
magnetic vector potential, 108, 109, 112, 92, 94, 97, 98, 100–102, 152, 156,
113, 115, 125–126, 131, 139, 151, 194, 207, 209, 332, 395, 429, 495
153–157 phase velocity, 314, 340, 343, 344, 390
magnetic-field integral equation (MFIE), phase-matching condition, 258, 278, 282,
162, 183, 219 292, 293, 315, 331, 365
magnetization vector, 16–18, 63, 64, 141, physical-optics current on metallic surfaces,
290 221
metallic septa, 361 plane wave, 158, 199, 200, 202, 203, 221,
method of separation of variables, 228 222, 225, 226, 228–298, 321, 330,
modified Ampère’s law, 2, 6, 14, 32, 80, 83, 446, 447, 452, 456, 459, 472–474,
111, 124, 143, 190, 267, 286, 291, 477, 478, 481–483, 490, 500, 501,
382, 442 503, 533–535, 543, 582, 583
modified Bessel function of the first kind, Poincaré sphere, 241
410, 487 point form of Ohm’s law, 19, 20, 45, 67
modified Bessel function of the second polarization charge density, 80
kind, 409, 410, 422, 487 polarization current, 65, 71, 80–83, 98, 99,
monopole, 539 130–131, 153, 201, 202, 310, 331,
motional induction, 12 429
INDEX 605

polarization unit vector, 239, 300, 311 saddle-point technique, 465, 479
polarization vector, 16, 18, 21, 26, 28, 57, scattered potential, 170, 171
80, 238, 239, 300, 311 Schelkunoff spherical Bessel functions, 514,
potential energy, 84, 85 522, 527, 544, 547
potential function, 83, 85, 108, 166, 175, self-complementary antenna, 162, 212–215,
427, 430, 446, 447, 488, 522, 529, 220
536, 545, 554 shadow boundary, 477, 478
power density of plane waves, 228, 237, 300 singularity point, 170, 171
Poynting vector, 88, 98, 103, 104, 106, 126– skin depth, 265, 266, 302
127, 133, 154, 157, 237, 253, 254, Snell’s law of reflection, 262, 270–271
291, 303, 304, 309, 344, 399, 444 Snell’s law of refraction, 260, 262, 270–
principal-value integral, 172 271, 301
principle of least time, 262 spherical Bessel functions of the first kind,
507, 509
quadrupole, 539, 541, 542, 579, 580 spherical Bessel functions of the second
quality factor for cavities, 352–356, 391 kind, 507, 509
quasi-static, 114, 116 spherical cavity, 503, 520–529, 575, 580,
581
radar cross section (RCS), 503, 546–549, spherical dielectric resonator, 503, 529, 531,
551, 576, 577 533, 534, 575
radial waveguide, 405, 433–435, 490, 496, spherical Hankel functions of the first kind,
497 508, 509, 553
radiation condition, 162, 166–169, 217 spherical Hankel functions of the second
radiation resistance, 136, 141, 145, 146, kind, 508, 509
154, 155, 157 spherical harmonic functions, 503
RCS of dielectric spheres, 551, 552, 577 spherical wave functions, 503, 506–585
RCS of metallic spheres, 546–549, 576 stationary formulation, 382, 386–387
reactive impedance surface, 314, 321, 323– stationary phase point, 465
335, 399 steepest descent path, 466, 592–595
receiving antenna, 147–149 Stokes’ theorem, 7, 14, 31, 46, 47, 58, 143
reciprocity condition for two-port networks, superposition principle, 84, 90, 113, 121,
192, 218 149
reciprocity for a source-free region, 190 surface charge density, 33, 34, 58, 59, 67,
reciprocity theorem, 162, 184–205, 218, 75–77, 170
223, 519, 582 surface current density, 32, 34, 58, 82, 199,
reciprocity theorem for a source-free region, 331, 500
186, 218 surface resistivity, 347
rectangular cavity, 314, 349–357, 391, 400,
401 TE mode, 314, 319–320, 334, 336, 340,
rectangular waveguide, 314, 335–357, 361, 345, 349, 360–368, 371, 374, 379,
363–368, 375, 379, 390, 395–402 389, 393, 394, 397, 399
reflection boundary, 477, 478, 480–482, 485 TE wave impedance, 395, 397
retarded potential, 113–115, 131–137 telegrapher’s equation, 327
TEM wave, 254
saddle point, 465, 479 tilt angle for polarization, 240, 300, 311
606 INDEX

time-average electric energy density, 128 wave number, 125, 228, 304, 310, 350, 429,
time-average magnetic energy density, 128, 434, 529
135, 159 wave polarization, 228, 276
time-average Poynting vector, 108, 127, waveguide, 314–320, 322–324, 335–367,
151, 157 371–374, 381–382, 389–392, 393–
time-harmonic fields, 108, 122–131, 151, 395, 397, 399–403
154–159 waveguide discontinuity, 396, 402
TM mode, 314, 318–324, 326, 333, 336– waveguide mode, 320
340, 345, 346, 349, 356, 358–361, wire antenna, 108, 144–147, 154
363–364, 368, 371, 374, 375, 389, WKB solution, 286–288
391, 394, 395, 397, 399 Wronskian for spherical Bessel functions,
TM wave impedance, 395 510, 547, 560, 574
transcendental equation, 275 Wronskian relationship, 412, 440, 444, 469,
transformation matrix, 250 487
transformer induction, 10, 12
transverse electric fields, 228, 254–264, zonal harmonic functions, 515, 536, 574
301, 304
transverse electric to r (TE-to-r), 520, 525,
529, 531, 544, 561, 569, 572, 573,
575, 577, 578, 580, 581
transverse electromagnetic wave, 133, 137
transverse magnetic fields, 228, 254–264,
301, 304
transverse magnetic to r (TM-to-r), 519,
520, 522, 525, 528, 529, 540, 544,
557, 562, 568, 572, 575, 577, 578,
580, 581

uniaxial, 18, 312

uniform plane wave, 231–234, 237, 238
uniqueness theorem, 68, 90–93, 105, 106,
165, 208, 235
unitary transformation, 250

vector phasor, 123, 127, 131

volumetric charge density, 3, 35, 53, 58,
108, 166–168

wave equation, 108, 110, 111, 125, 126,

151, 162, 165, 168, 177, 228, 229,
234, 242, 256, 276–278, 286–288,
290, 300, 314, 353, 358, 363, 367,
368, 370, 374, 375, 385, 392, 395,
403, 405, 432, 442, 460, 487, 493,
503–506, 515, 516, 529, 535
 Foundations of Applied Electromagnetics
Electromagnetics is credited with the greatest achievements of physics in the 19th century.
Despite its long history of development, due to its fundamental nature and broad base,
research in applied electromagnetics is still vital and going strong. In recent years electromag-
netics played a major role in a wide range of disciplines, including wireless communication,
remote sensing of the environment, military defense, and medical applications, among many
others. Graduate students interested in such exciting fields of research need a strong founda-
tion in field theory, which was part of the motivation for writing this book on classical electro-
magnetics but with an eye on its modern applications.

KAMAL SARABANDI is the Fawwaz T. Ulaby

Distinguished University Professor and the Rufus
S. Teesdale Professor of Engineering serving in
the Department of Electrical Engineering and
Computer Science at the University of Michigan.
His research interests include microwave and
millimeter-wave radar remote sensing, electro-
magnetic wave scattering and propagation,
antennas, meta-materials, and bio-electromag-
netics. Dr. Sarabandi served as a member of
NASA’s Advisory Council for two consecutive
terms (2006-2010) as well as the President of the
IEEE Geoscience and Remote Sensing Society
(2015-2016). His contributions to the field of elec-
tromagnetics have been recognized by many
awards including the Humboldt Research Award,
the IEEE GRSS Distinguished Achievement Award,
the IEEE Judith A. Resnik medal, the IEEE GRSS
Education Award, and the NASA Group Achieve-
ment Award. He is a Fellow of IEEE and of the
American Association for the Advancement of
Science (AAAS), and a Fellow of the National Acad-
emy of Inventors. He is a member of the National
Academy of Engineering.