Wangsness Electromagnetic Fields

V J- Sh
V>
X
J o
f~!
7 /
2 N D EDITION
*WB020991*
ELECTROMAGNETIC
FIELDS
ROALD K.WANGSNESS
P R O F E S S O R O F PHYSICS
UNIVERSITY OF ARIZONA
JOHN WILEY & SONS

NEW YORK
CHICHESTER
BRISBANE
A
TORONTO
SINGAPORE
OA
/y
Copyright 1979, 1986, by John Wiley & Sons, Inc.

All rights reserved. Published simultaneously in Canada.
Reproduction or translation of any part of
this work beyond that permitted by Sections
107 and 108 of the 1976 United States Copyright
Act without the permission of the copyright
owner is unlawful. Requests for permission
or further information should be addressed to
the Permissions Department, John Wiley & Sons.
Library of Congress Cataloging in Publication Data:
Wangsness, Roald K.
Electromagnetic fields.
Includes indexes.
1. Electromagnetic fields. I. Title.
QC665.E4W36 1986
537
85-24642
ISBN 0-471-81186-6
Printed in the United States of America
10 9 8 7 6 5 4 3 2 1
VECTOR FORMULAS
A (B X C) = (A x B) C
A X (B X C) = B(A C) - C(A B)
V X vu = 0
V { V X A) = 0
(A x B) (C X D) = (A C)(B D) - (A D)(B C)
d
du
dA
do
da
do
-(uA) = A + u
(1-30)
(1-48)
(1-49)
1-106)
-107)
d
dA
dB
~T~ (A B) = B + A do
(1-29)
do
-108)
do
dA
do
do
do
(A x B) = x B + A x
-109)
V ( + v ) = VM + V f
-110)
V ( u v ) = u V v + vV u
-111)
V (A B) = B X ( v X A) + A x ( v X B) + (B v ) A + (A v )B
-112)
V ( C r) = C
-113)
where C = const.
V - { A + B) = V - A + V - B
V
-114)
(wA) = A ( v ) + "(V A)
-115)
V (A X B) = B ( V X A) - A (V X B)
-116)
V X (A + B) = V XA + V X B
-117)
V X (wA) = ( v ) X A + w(V X A)
-118)
V X
(A X B) = ( v B)A - ( v A)B + (B V )A - {A V )B
V X ( v X A) = v ( v A) - v 2A
-119)
-120)
where
<95.
dB.
(A V )B ~ \ [ a
'
dy
dx
SB
dx
+ A.-
dBy
dy
+ Ay
SB
'
dy
dz
+ A.
+ A.
SB,
clz
dB.
Sz
(1-121)
VECTOR OPERATIONS
RECTANGULAR COORDINATES
du
du
du
V = x - I - y + z~
dx
dy
dz
V
dAx
=
dx
dAv
+
ay
IdA.
^
2 w
dA.
I dAx
~ - j r )
d2u
^
dx
d2u
^ 2
dy
(1-42)
dz
dAv\
^
=
(1-37)
A - j t
dA.\
-
ldAv
+
dAA
( ,
" { - i t
d2u
^ T1
dz
-43)
(1-46)
CYLINDRICAL C O O R D I N A T E S
dp
V A =
dp
(1-85)
JI
d(p
1- dA.
,
.
(pAp)
+ P
J l d A
V
du
du
1 du
P ^ - + Cp-
VK -
(p
d<p
dAA
dtp
dz
"
dA:
(1-87)
dz
I dA,
<p;
dz
dA.
A
+i 2
dp
1 ^
P d p ^ ^
"
d<p
(1-88)
i a /
V u = p dp\^
z
3h\
d2u
p2 dcp2
dp)
d2u
(1-89)
dz2
SPHERICAL C O O R D I N A T E S
v
" =
dw
a7
1 du
7 ^
I d ,
V A = (r2Ar)
r~dr
V X A =
1
<P
<9*
(1-101)
r sin 6 dtp
1
1
+ j-{smdAd)
r sin 6 do
d ,
r sin 6
(1-103)
dtp
dAe
<P'
r sin 6
dA,
sinf? dtp
dArA^
1
d I
du\
(1-104)
dd
1
r2s'm8
d I
du \
38 \
30)
1
r2s\n28
02u
3<p2
(1-105)
PREFACE
This is a textbook for use in the usual year-long course in electromagnetism at an
intermediate level given for advanced undergraduates. I have done my best to make it
as student oriented as possible by writing it in a systematic and straightforward way
without any sleight of hand, and minimum use of " I t can be shown t h a t . . . . " I have
also tried to make clear the motivations for each step in a derivation or each new
concept as it is introduced. At appropriate points I have pointed out the source of
many of the simple mistakes commonly made by students and have suggested how they
may be prevented. There is extensive cross referencing in the text so that there should
be n o doubt as to the detailed source of any specific result or its relation to the rest of
the subject; this will also make the book much more useful as a reference source well
after the course has been completed.
In preparing this revision, I have concentrated on adhering to these goals while
trying to improve the clarity and organization of the material, since my aim is to
produce a useful text with an adequate, rather than encyclopedic, coverage at a
reasonable mathematical level. Consequently, most of the changes are scattered
throughout the text. For example, I have improved the discussion of electric and
magnetic energies in the presence of matter in Sections 10-8, 10-9, and 20-6. In
addition, the beginning treatment of radiation has been simplified somewhat in
Sections 28-1 and 28-2, and the proof about the electric field in a cavity in a conductor
has been corrected in Section 6-1. The principal and most noticeable change has been
the addition of a completely new Chapter 27 on Circuits and Transmission Lines. I
hope that this will turn out to be useful and satisfactory to those who have expressed a
desire to have this material included.
The emphasis continues to be on the properties and sources of the field vectors, and
I hope that I have succeeded in making clear the shifts in concepts and points of view
that are involved in the change from action at a distance to fields. The overall treatment
is generally that of a macroscopic and empirical description of phenomena, although
the microscopic point of view is presented in the discussion of conductivity in Sections
12-5 and 24-8. However, Appendix B briefly surveys the microscopic origins of
electromagnetic properties, and it is written and organized so that, if desired, it can be
taken up section by section at an appropriate intermediate point. Thus Section B-l
could be discussed anytime after Section 10-7, and most of Section B-2 can follow after
Section 20-5 while the last part on ferromagnetism can follow Section 20-7; finally,
Section B-3 could be covered after Section 24-8 has been mastered and the student has
worked Exercise 24-28. Similarly, even more flexibility is possible since separate
sections of Appendix A that deals with the motion of charged particles can be studied
anytime after the corresponding force term involving E a n d / o r B has been obtained.
SI units are used throughout; in practice, this means we use MKSA units. It is
virtually certain, however, that at some time a student will encounter material in
Gaussian units and will need some guidance on what to do about it. This is the purpose
of Chapter 23 in which other unit systems are discussed, but only after the general
theory as given by Maxwell's equations has been systematically described. I have
written this chapter primarily in terms of the purely practical aspects of how to
vii
Viii
PREFACE
recognize an equation written in other unit systems, how to put it in more familiar
form if desired, and exactly what numbers one should put into an equation in
Gaussian units in order to get a correct answer.
In Chapter 9, the boundary conditions satisfied by an arbitrary vector at a surface of
discontinuity in properties are obtained in the general form involving its divergence and
curl. N o t only does this help the student by showing the importance of knowing these
particular source equations but it simplifies later discussion because, as each new vector
is defined, its boundary conditions can be found at once without having, in effect, to
rederive them each time.
I have added several new clarifying examples and exercises so that now there are
over 145 worked-out examples in the text. Virtually all of the standard ones are
included, many of them presented in more detail than is usually found, with particular
emphasis on the crucial stage of setting up the problem, since this is often what causes
students so much difficulty. By now, there are a total of 587 exercises included. Some
are numerical to give an idea of typical orders of magnitude, some are similar to
examples of the text, many refer to completely different situations, and some involve
extensions of the theory. Many of these exercises will be found to be suitable for use as
additional examples for classroom analysis. Answers to odd-numbered exercises are
provided except, of course, for those in which the answer is included in the statement of
the problem.
I have benefited over the years from discussions with and the questions of many
students and my colleagues; I am grateful for their contributions to the final character
of this book. In particular, I thank my colleagues C. Y. Fan, K. C. Hsieh, E. W.
Jenkins, and J. E. Treat for their useful and helpful comments. I am also grateful for
the suggestions and remarks given to me by H. Asmat, R. C. Henry, A. T. Mense, K.
Park, J. R. Ravella, and several reviewers.
Tucson, Arizona
January 1986
Roald K. Wangsness
CONTENTS
INTRODUCTION
1
VECTORS
E3
Us
ll-ll
Definition of a Vector
3
Addition
4
Unit Vectors
5
[O
Components
5
O
The
Position
Vector
7
IT3I
Scalar Product
8
EH]
Vector Product
9
im
Differentiation with Respect to a
Scalar
11
|1-9|
Gradient of a Scalar
12
I1-10I Other Differential
Operations
14
II-111 The Line Integral
15
I1-12I Vector Element of Area
17
I1-13I The Surface Integral
20
H-14I The Divergence Theorem
21
II -151 Stokes' Theorem
24
H-161 Cylindrical Coordinates
28
11-171 Spherical Coordinates
31
11-181 Some Vector Relationships
34
11-191 Functions of the Relative
Coordinates
35
11-201 The Helmholtz Theorem
37
2
C O U L O M B ' S LAW
12^1
F4l
51
Definition of the Electric

Field
51
pT>
G A U S S ' LAW
14-11
H ]
|4-3|
58
ixercises
[E
Derivation of
Gauss' Law
58
Some Applications of Gauss'
Law
60
Direct Calculation of V E
THE SCALAR POTENTIAL

Ml
5-2
15-41
lH
El
6-3
65
68 1ixercises
Definition and Properties of the

Scalar Potential
68
Uniform Spherical Charge
Distribution
72
Uniform Line Charge
Distribution
73
The Scalar Potential and
Energy
79
C O N D U C T O R S IN ELECTROSTATIC
FIELDS
83
401 IE
ixercises
40
Point Charges
Coulomb's Law
41
Systems of Point Charges
Continuous Distributions of
Charge
44
Point Charge Outside a
Uniform Spherical Charge
Distribution
46
THE ELECTRIC FIELD
Field of a Uniform Infinite Line

Charge
52
Field of a Uniform Infinite Plane
Sheet
53
What Does All of This
Mean?
55
Some General Results

Systems of Conductors
Capacitance
90
[Hi
83
88
43
7
ELECTROSTATIC ENERGY
rni
Energy of a System
Charges
98
Energy of a System
Conductors
100
Energy in Terms of
Field
101
Electrostatic Forces
Conductors
103
98 1ixercises
of
of
the Electric
on
IX
CONTENTS
ELECTRIC MULTIPOLES
110
|Exercisesl
The Multipole Expansion of the

Scalar Potential
110
The Electric Dipole Field
119
The Linear Quadrupole
Field
121
Energy of a Charge Distribution in
an External Field
123
12
9-3
10
Hl-61
Spherically Symmetric Solution of

Poisson's Equation
198
ELECTRIC CURRENTS
Origin of a Surface of
Discontinuity
132
The Divergence and the Normal
Components
133
ITT2l
emi
ESS
0 3
12-6
The Curl and the Tangential

Components
134
Boundary Conditions for the
Electric Field
136
Boundary Conditions for the Scalar
Potential
138
13
E ]
OH
fiT3l
Em]
Polarization
140
Bound Charge Densities
142
14
ED
The Electric Field within a

Dielectric
145
EM]
Uniformly Polarized Sphere

The D Field
151
148
EM]
Classification of Dielectrics
154
Em]
Linear Isotropic Homogeneous

(l.i.h.) Dielectrics
156
m \
Energy
161
Em
Forces
165
111-11
I11-2I
I11-3I
HL-4|
14-3
14-4
14-5
Uniqueness of the Solution of

Laplace's Equation
171
Method of Images
173
" Remembrance of
Things Past"
183
Separation of Variables in
Rectangular Coordinates
14-2
15
Current and Current

Densities
202
The Equation of Continuity
Conduction Currents
207
Energy Relations
205
211
A Microscopic Point
of View
212
The Attainment of Electrostatic
Equilibrium
213
2 1 7 1 xercises
The Force between Two Complete
Circuits
217
Two Infinitely Long Parallel
Currents
220
The Force between Current
Elements
222
225
[&
Definition of the Magnetic

Induction
225
Straight Current of Finite
Length
227
Axial Induction of a Circular
Current
229
Infinite Plane Uniform Current
Sheet
231
Moving Point Charges
233
THE INTEGRAL FORM O F AMPERE'S

Exercises
LAW
237
15-1
15-2
15-3
185
IKxercisesl
THE M A G N E T I C I N D U C T I O N
E13
SPECIAL M E T H O D S IN
Exercises
ELECTROSTATICS
171
202
AMPERE'S LAW
ELECTROSTATICS IN THE PRESENCE

Ifcxercisesl
OF MATTER
140
m
11
Separation of Variables in Spherical

Coordinates
190
mi
B O U N D A R Y C O N D I T I O N S AT A
SURFACE OF
[Exercises]
DISCONTINUITY
132
19^21
111-51
Derivation of the Integral

Form
237
Some Applications of the Integral
Form
242
Direct Calculation of
V X B
248
CONTENTS
E.xercisesl
16
THE VECTOR POTENTIAL

I16-H
H6-2I
II6-31
|16-4|
I16-5I
17
I18-2I
ll8-31
MAXWELL'S EQUATIONS
I21-1I
I21-2I
21-3I
I21-4I
I21-5I
22
Faraday's Law
263
Stationary Media
266
Moving Media
269
Inductance
277
MAGNETIC ENERGY
I18-1I
21
The Divergence of B
250
Definition and Properties of the
Vector Potential
251
Uniform Induction
254
Straight Currents
255
Infinitely Long Ideal
Solenoid
259
FARADAY'S LAW OF
Ifcxercisesl
INDUCTION
263
I17-1I
ll 7-21
I17-3I
I17-4I
18
250
22-3
284 Eixercises
23
Energy of a System of Free

Currents
284
Energy in Terms of the Magnetic
Induction
287
Magnetic Forces on Circuits
290
MAGNETIC MULTIPOLES
I19-1I
I19-21
I19-3I
l2(K2l
20-3
UMl
[203]
20-6
l2CV7l
[20^1
Poynting's Theorem
356
Electromagnetic Momentum
359
E.xercises
The Potentials in General

363
The Potentials for Linear Isotropic
Homogeneous Media
364
Gauge Transformations
366
SYSTEMS OF UNITS A GUIDE FOR

THE PERPLEXED
368
23-1
123^21
123^31
n o
19
348 IKxercisesl
The Displacement Current

348
Maxwell's Equations in General
Form
353
Maxwell's Equations for Linear
Isotropic Homogeneous
Media
355
SCALAR A N D VECTOR
POTENTIALS
363
l22^2l
XI
IKxercisesl
Origin of Other Systems of

Units
368
The Electrostatic and
Electromagnetic Systems
369
The Gaussian System
370
How to Cope with the Gaussian
System
371
2 9 7 |Exercises I
The Multipole Expansion of the

Vector Potential
297
The Magnetic Dipole Field
302
Filamentary Currents
303
Energy of a Current Distribution in
an External Induction
305
Magnetization
312
Magnetization Current
Densities
313
Uniformly Magnetized
Sphere
319
The H Field
321
Linear Isotropic Homogeneous
Magnetic Materials
327
Energy
333
Ferromagnetic Materials
338
Magnetic Circuits
342
24
PLANE WAVES
U S
n o
n o
24-7
24-8
25
375 Ifcxercisesl
Separate Equations for E

and B
375
Plane Waves in a Nonconducting
Medium
376
Plane Waves in a Conducting
Medium
382
Plane Waves in a Charged
Medium
388
Plane Wave in an Arbitrary
Direction
389
Complex Solutions and TimeAverage Energy Relations
391
Polarization
393
Are the Electromagnetic Parameters
of Matter Constants?
398
REFLECTION A N D REFRACTION OF
PLANE WAVES
405
|25-1|
The Laws of Reflection and

Refraction
405
[e;
xii
25-2
E Perpendicular to the Plane of

Incidence
411
HEU
25-31
E Parallel to the Plane of

Incidence
415
Magnetic Dipole Radiation

Linear Electric Quadrupole
Radiation
484
128^61
Antennas
I25-4I
Total Reflection
25-5
I25-6I
Energy Relations
420
Reflection at the Surface of a
Conductor
421
Continuously Varying Index of
Refraction
424
Radiation Pressure
425
25-71
|25-8|
26
418
29
29-1
129^21
12931
[293]
Boundary Conditions at the Surface

of a Perfect Conductor
430
26-2
Propagation Characteristics of Wave

Guides
431
433
Fields in a Wave Guide
m m
Rectangular Guide
26-5]
TEM Waves
441
Resonant Cavities
29-6
444
I27-3I
More Complicated
Situations
457
Transmission Lines
|27-4|
453
461
28-3
514
Electromagnetism in
Vacuum
518
Fields of a Uniformly Moving Point
Charge
523
|A-3l
Static Electric and Magnetic

Fields
538
A Time-Dependent Magnetic
Field
543
530
531
ELECTROMAGNETIC PROPERTIES OF
lExercises]
MATTER
546
B-l
Static Electric Properties
B-2
Static Magnetic Properties
lB-31
Response to Time-Varying
Fields
562
4 6 9 Exercises
Retarded Potentials
469
ANSWERS T O ODD-NUMBERED
EXERCISES
569
Multipole Expansion for

Harmonically Oscillating
Sources
472
Electric Dipole Radiation
Particle Mechanics
Static Electric Field

Static Magnetic Field
RADIATION
Hy]
[2^21
Historical Origins of Special

Relativity
494
The Postulates and the Lorentz
Transformation
499
General Lorentz Transformations,
4-Vectors, and Tensors
507
IA-1I
|A-2|
CIRCUITS A N D T R A N S M I S S I O N
ll'.xercisesl
LINES
449
Kirchhoff's Laws
449
The Series RLC Circuit
4 9 4 | Exercises I
M O T I O N OF CHARGED
lExercisesl
PARTICLES
530
435
27-3
I27-2I
477
482
487
SPECIAL RELATIVITY
HOI
26-1
[26^61
28
(n1>n2,6i>9e)
28-41
FIELDS IN B O U N D E D
I Exercises]
REGIONS
430
26-3
27
CONTENTS
INDEX
577
546
554
INTRODUCTION
. . . Faraday, in his m i n d ' s eye, s a w lines of f o r c e t r a v e r s i n g
all s p a c e w h e r e t h e m a t h e m a t i c i a n s s a w c e n t r e s of f o r c e
a t t r a c t i n g at a d i s t a n c e : Faraday s a w a m e d i u m w h e r e t h e y
s a w n o t h i n g b u t d i s t a n c e : Faraday s o u g h t t h e seat of t h e
p h e n o m e n a in real a c t i o n s g o i n g o n in t h e m e d i u m , t h e y
w e r e s a t i s f i e d that t h e y had f o u n d it in a p o w e r of a c t i o n
at a d i s t a n c e i m p r e s s e d o n t h e electric fluids.
A Treatise on Electricity
J. C. Maxwell,
and Magnetism
It has been more than 100 years since Maxwell wrote the above in the preface to his
now-famous book. His aim was to put the field concepts, which Faraday had been so
instrumental in developing, into mathematical forms that would be convenient to use
and would emphasize the fields as basic to a coherent description of electromagnetic
effects. At that time, it had been only slightly more than 50 years since Oersted and
Ampere had shown the relation between electricity and magnetismsubjects that had
been studied and developed completely separately over a long period. The emphasis
had been primarily on the forces exerted between electric charges and between electric
currents and the idea of shifting to electric and magnetic fields as the primary features
had little acceptance and was often, in fact, viewed with outright hostility.
As the title of this book indicates, times have changed, and our main interest here is
the study of the nature, properties, and origins of electromagnetic fields, that is, of
electric and magnetic vector quantities that are defined as functions of time and of
position in space. Forces, and associated concepts such as energy, have not disappeared
from the subject, of course, and it is desirable to begin with forces and to define the
field vectors in terms of them. Nevertheless, our principal aim is to express our
descriptions of phenomena in terms of fields in as complete a manner as we possibly
can. This emphasis on fields has proved to be extremely rewarding and it is difficult to
imagine how electromagnetic theory could have been developed to its present state
without it.
This book contains more material than is normally covered in the usual one-year
course; all of it, however, is of interest and value to a serious student of physics.
The points of view of all authors are generally not the same, and no book discusses
every detail of a given subject. Here is a short list of relatively recent books on
electromagnetism that are written at roughly the same level as this one.
D. M. Cook, The Theory of the Electromagnetic Field, Prentice-Hall, Englewood
Cliffs, N.J., 1975.
P. Lorrain and D. R. Corson, Electromagnetic Fields and Waves, Second Edition,
Freeman, San Francisco, 1970.
M. H. Nayfeh and M. K. Brussel, Electricity and Magnetism, Wiley, New York,
1985.
1
INTRODUCTION
J. R. Reitz, F. J. Milford, and R. W. Christy, Foundations of Electromagnetic

Third Edition, Addison-Wesley, Reading, Mass., 1979.
A. Shadowitz, The Electromagnetic Field, McGraw-Hill, New York, 1975.
Theory,
The following books discuss electromagnetism at a more advanced level:

J. D. Jackson, Classical Electrodynamics, Second Edition, Wiley, New York, 1975.
E. J. Konopinski, Electromagnetic Fields and Relativistic Particles, McGraw-Hill,
New York, 1981.
W. K. H. Panofsky and M. Phillips, Classical Electricity and Magnetism, Second
Edition, Addison-Wesley, Reading, Mass., 1962.
A. M. Portis, Electromagnetic Fields: Sources and Media, Wiley, New York, 1978.
J. A. Stratton, Electromagnetic Theory, McGraw-Hill, New York, 1941.
(Finally, a note on notation: in this book, the symbols = , = , = , ~ , = always
mean, respectively, equal to, approximately equal to, of the order of magnitude of,
proportional to, and different from.)
1
VECTORS
In the study of electricity and magnetism, we are constantly dealing with quantities that
need to be described in terms of their directions as well as their magnitudes. Such
quantities are called vectors and it is well to consider their properties in general before
we meet specific examples. Using the notation and terminology that has been developed
for this purpose enables us to state our results more compactly and to understand their
basic physical significance more easily.
1-1 I DEFINITION OF A VECTOR
The properties of the displacement of a point provide us the essentials required for our
definition. If we start at some point Pl and move in some arbitrary way to another
point P2, we see from Figure 1-1 that the net effect of the motion is the same as if the
point were moved directly along the straight line D from Pl to P2 as indicated by the
direction of the arrow. This line D is called the displacement and is characterized by
both a magnitude (its length) and a direction (from Pl to P2). If we now further
displace our point along E from P2 to still another point P3, we see from Figure 1-2
that the new net effect is the same as if the point had been given the single displacement
along F from P1 to P3. Accordingly, we can speak of F as the resultant, or sum, of the
successive displacements D and E, so that Figure 1-2 shows the fundamental way in
which displacements are combined or added to obtain their resultant.
A vector is a generalization of these considerations in that it is defined as any
quantity that has the same mathematical properties as the displacement of a point.
Thus we see that a vector has a magnitude; it has a direction; and the addition of two
vectors of the same intrinsic nature follows the basic rule illustrated in Figure 1-2.
Because of the first two properties, we can represent a vector by a directed line such as
those already used for displacements. A vector is generally printed in boldface type,
thus, A; its magnitude will be represented by |A| or by^4.
A scalar is a quantity that has magnitude only. For example, the mass of a body is a
scalar, whereas its weight, which is the gravitational force acting on the body, is a
vector.
Because of the nature of a vector as a directed quantity, it follows that a parallel
displacement of a vector does not alter it, or, in other words, two vectors are equal if
they have the same magnitude and direction. This is illustrated in Figure 1-3 where we
see that A = A'. Now we can investigate what mathematical operations we can perform
with and on vectors.
VECTORS
ADDITION
According to our basic rule we find that, if we take A and add B, we obtain the sum C
shown as a solid line in Figure 1-4. We also see that, if we take B and then add A, we
get the same vector C. Therefore addition of vectors has the commutative property that
C = A + B = B + A
(1-1)
By proceeding in the same manner, one can establish the associative property of
vector addition:
D = (A + B) + C = A + ( B + C) = (A + C) + B
(1-2)
and so on.
If we reverse a displacement such as D in Figure 1-1 by retracing it in the opposite
direction, the net effect is then no displacement; hence it is appropriate to define the
negative of a vector as a vector of the same magnitude but reversed in direction, for
then we should obtain A + ( A) = 0, as we would want. Then we can easily subtract a
vector by adding its negative:
A B = A + ( B)
(1-3)
Figure 1-4. The sum of t w o vectors does not depend

on the order in which they are added.
1-4 COMPONENTS
a
"V
Figure 1-5.
a is a unit vector in the direction of A.
Figure 1-6.
Unit vectors for rectangular coordinates.
The product of a scalar s and a vector, which we write as either sA or As, is then
merely the sum of s vectors A, or is a vector with a magnitude equal to | j | times the
magnitude of A, and is in the same direction as A if ^ is positive, and in the opposite
direction to A if s is negative.
UNIT VECTORS
A unit vector is defined as a vector of unit magnitude and will be written with a
circumflex above it, thus, e; since unit vectors are always taken to be dimensionless we
will have |e| = 1. If, for example, a unit vector & is chosen to have the direction of A,
then we can write
A = Aa.
and
A
a =
(1-4)
This point is illustrated in Figure 1-5.

A particularly convenient set of unit vectors can be associated with a rectangular
coordinate system. They are written x, y, z and are defined to be in the directions of the
x, y, and z axes respectively, as shown in Figure 1-6. In other words, each is in the
direction of increasing value of the corresponding rectangular coordinate. We also see
that any one of this set is perpendicular to each of the other two.
As we will see, it is often convenient and advantageous to define other unit vectors.
COMPONENTS
In order to proceed further, it is convenient to refer our vectors to particular coordinate

systems. From Figure 1-7, we see that we can write a vector A as the sum of three
properly chosen vectors, each of which is parallel to one of the axes of a rectangular
coordinate system; that is, A = Ax + Ay + A , . It is more useful, however, to write
each of these terms as the product of a scalar and the unit vectors of Figure 1-6. Thus,
VECTORS
Figure 1-7. A is the sum of the rectangular vector

components.
we write Ax = Axx, and so on, and the above expression becomes

A = Axx
+ Ay y + Azi
(1-5)
The three scalars Ax, Ay, Az are called the components of A; hence we see that a vector
can be specified by three numbers. The components can be positive or negative; for
example, if Ax were negative, then the vector A x of Figure 1-7 would have a direction in
the sense of decreasing values of x.
From Figure 1-7, it is seen that the magnitude of a vector can be expressed in terms
of its components as
A = |A| = (Al + Al + A])
1/2
(1-6)
In Figure 1-8, we illustrate the fact that A makes specific angles with respect to each
of the axes; these angles a, fi, y are called the direction angles of A and are measured
from the positive directions of their respective axes. Figure 1-9 shows the plane
containing both A and x and we see that Ax is given by Ax = A cos a. Combining this
with (1-6), we get
I = cos a =
(A2X + A2y + A2z)
1/2
(1-7)
where I x is called a direction cosine. Similar expressions hold for the other two direction
angles /? and y and their associated direction cosines l y and l z , so we see from (1-6)
and (1-7) that, if we know the rectangular components of a vector, we can calculate its
magnitude and direction.
y
Figure 1-8.
Definition of direction angles.
1-5 THE POSITION VECTOR
xL
Figure 1-9.
Ax is the x component of A.
If we now combine (1-4), (1-5), and (1-7), we find that the unit vector a can also be
written as
(1-8)
a = lxx + ly y + lzz
so that the components of a unit vector in a given direction are simply the direction
cosines associated with the direction. If we now apply the general result (1-6) to the
specific vector a, we get the important relation involving direction cosines
ll + I) + l] = 1
(1-9)
which can also be obtained from (1-7) and its analogues.

The addition of vectors illustrated in Figure 1-4 is easily expressed in terms of the
rectangular components. From Figure 1-10, we see that a component of the sum
C = A + B is given by the sum of the corresponding components, that is,
CX = AX + Bx
Cy = Ay + By
CZ = AZ + Bz
(1-10)
THE POSITION VECTOR
We now consider a simple specific example of a vector. As shown in Figure 1-11, the
location of a particular point P in space can be specified by the vector r drawn from
the origin of a suitably and conveniently chosen coordinate system; this vector r is
called the position vector of the point. In terms of the rectangular coordinate system of
Figure 1-6, the components of r are just the rectangular coordinates (jc, y, z) of the
point; thus we have
r = x x + yy + zz
(l-ll)
Similarly, another point P' with coordinates (x', y', z') will be located by its position
vector r' = x'x + y'y + z'z as shown in Figure 1-12. Now that we have located the two
points individually, we can describe the position of one with respect to the other by
drawing a vector from P' to P\ this vector R is called the relative position vector of P
VECTORS
Figure 1-11.
r is the position vector of the point P.
P'(x', y', z')
Figure 1-12. R is the relative position vector of P

with respect to P'.
with respect to P'. We see from Figure 1-12 that r ' + R = r so that
R = r - r'
(1-12)
Using (1-10) and (1-11), we can write R in component form as

R = (x - x ' ) x + (j> - y')y
+ (z - z')z
(1-13)
and therefore
R=
[(x-x>)2+
{ y - y ' f + (z-z')2Y
/ 2
(1-14)
because of (1-6). We will have frequent occasion to use these results. We note that the
relative position of P' with respect to P is given by a vector R' drawn from P to P'
and, in fact, R' = R.
Although we have not specified our coordinate system other than to say that it was
arbitrarily and conveniently chosen, once this choice has been made in a particular
case, the coordinate system is said to be "fixed in space" and its unit vectors x, y, and z
are to be taken as constant in direction as well as magnitude. In other words, the fixed
coordinate system that we use is simply one of the inertial reference frames familiar to
us from classical mechanics.
We now turn to multiplication of vectors; two types are defined.
rT6l
SCALAR P R O D U C T
We define the scalar product of two vectors as the scalar equal to the product of the
magnitudes of the vectors and the cosine of the angle between them, or
A B =
cos ^
Because of the notation the scalar product is also called the dot product.
(1-15)
1-7 VECTOR PRODUCT
F i g u r e 1-13.
The angle involved in the scalar product.
We see from Figure 1-13 that we can get a simple interpretation of the scalar
product: (2? cos 4^) ,4 = component of B along the direction of A times the magnitude
of A = (A cos <k)B = component of A along B times the magnitude of B.
It is clear from (1-15) that the order of terms does not change the scalar product,
that is,
A
B = B
(1-16)
and that if two vectors are perpendicular then A B = 0 and conversely. Furthermore,
the square of a vector can be interpreted as the vector dotted with itself; the result is
the square of its magnitude and we can write
A2 = A A = A2
(1-17)
If we knew only the rectangular components of A and B, it would be inconvenient to

calculate A B from (1-15) since this would necessitate finding the angle between A and
B. Fortunately, it is possible to express A B directly in terms of the rectangular
components. Since the angle between each pair of unit vectors defined in Figure 1-6 is
90, we easily find from (1-15) that
x - y = y- z = z- x = 0
(1-18)
x - x = y- y = z- z = l
(1-19)
and from (1-17) that

Writing A and B each in the form (1-5), we can multiply them together term by term to
get
A
B = ( A x x + Ayy + Azz)
=
A B
x x*
X +
A B
x y*
(Bxx
y +
+ Byy + Bzz)
A B
x A
Z + ...
and, after we use (1-18) and (1-19) to simplify the resulting nine terms, we find that
A
B = AXBX + AyBy
4- AZBZ
(1-20)
Suppose now that e is a unit vector in some specific direction. If we let Ae be the
component of A in this direction, we see from (1-15) that it can be found from
Ae = A e
(1-21)
VECTOR PRODUCT
This is also called the cross product and is written A X B. It is a vector perpendicular to
both A and B and its magnitude is defined as
| A X B| = AB s i n ^
(1-22)
Its direction is given by the following right-hand rule: if the fingers of the right hand
are curled in the sense necessary to rotate A through the smaller angle into coincidence
with B, the thumb points in the direction of A X B. This rule is illustrated in Figure
1-14.
10
VECTORS
Ax B
A sin
Figure 1-15.
Interpretation of the magnitude of a
.A cross product as an area.
If we look at the plane containing A and B shown in Figure 1-15, we can get a
simple interpretation of the cross product. We see f r o m the figure and (1-22) that the
magnitude of the cross product is equal to the area of the parallelogram with A and B
as sides.
F r o m the definition of the direction of a cross product shown in Figure 1-14, it is
evident that the order of terms is important, since it is seen that
B x A = - ( A X B)
(1-23)
If A a n d B are parallel, then it follows f r o m (1-22) that A X B = 0, and conversely.

In particular,
A X A = 0
(1-24)
F o r the unit vectors along the axes shown in Figure 1-6, if we use (1-22), the
right-hand rule, the facts that they are mutually perpendicular, and that the cross
product is perpendicular to both vectors, we find that
x X y = z
y X z = x
z X x = y
(1-25)
and, f r o m (1-24), that

xXx = yXy = zXz = 0
(1-26)
The vector product can also be conveniently written in terms of the rectangular
components. Proceeding similarly to what we did in obtaining (1-20), we write A and B
in the f o r m (1-5), multiply them together term by term, and use (1-23), (1-25), and
(1-26) to simplify the results. We find that
A X B = (AyBz
- AzBy)x
+ (AZBX
- AxBz)y
+ (AxBy
- AyBx)z
(1-27)
This can be written as an easily remembered determinant

A
y
A
A X B =
Bx
By
(1-28)
Bz
1-8
DIFFERENTIATION W I T H
RESPECT T O
A SCALAR
1 1
It will be left as exercises to verify that

Ax
A ( B X C) = (A X B ) C = Bx
cx
Bz
(1-29)
and that
A X ( B X C) = B(A C) - C(A B )
(1-30)
In (1-29) we see that the dot and the cross can be interchanged without affecting the
value of this triple scalar product; hence the parentheses are not really needed. In the
triple vector product (1-30), however, the parentheses are important because (A X B) x
C = - C X (A X B) from (1-23).
Division of vectors is not defined.
1-8
DIFFERENTIATION WITH RESPECT T O A SCALAR
Suppose that A is a continuous function of some scalar variable a, so that we can write
A = A(CT). This is equivalent to the three scalar equations: Ax = Ax(o), Ay = Ay(a),
Az = Az(a). If a is changed to a + ACT, A can generally change in both magnitude and
direction as shown in Figure 1-16. The change in A is AA = A(CT + ACT) A(CT). We
can then define the derivative of the vector A with respect to the scalar CT as follows:
dA
d a
lim
AA
A0 0 A C T
AAo
A(CT + ACT) A(CT)
lim
(1-31)
ACT
AO
0 - . O
0
This process has yielded another vector from a vector. Familiar examples of (1-31) are
the velocity and acceleration of a particle that are successive derivatives of the position
vector with respect to the time.
If A is written in terms of its rectangular components, then, since these unit vectors
are constant, we see from (1-31) that the components of the derivative are the
derivatives of the respective components, and we have
dA
da
dA,
da
-X
dA}
da
y+
A
dA^
da
(1-32)
Once we have defined the derivative of A, we can proceed to its differential dA,
which is to represent an infinitesimal change in A. This is obtained, in effect, by
multiplying (1-32) by da to give
dA = dAxx
A(A
+ dA y + dA, z,
(1-33)
+ AA)
Figure 1-16. A A is the change in the vector corresponding to the

change Ao of the scalar.
12
VECTORS
Applying this to the position vector of (1-11), we get

dx = dx x 4- dy y + dz z
(1-34)
[T-9l GRADIENT OF A SCALAR
Suppose we have a scalar quantity u that is a function of position so that we can write
u = u(x, y, z). Such a situation is called a scalar field. An example would be the
temperature at each point in a room. At some other point, which is displaced by ds
from the first, the value of the scalar will have changed to u + du (Figure 1-17). In fact,
du
du
du
du = dx + dy + dz
dx
dy
dz
(1-35)
where we should remember that the derivatives are evaluated at the original point, that
is, du/dx = (du/dx)P,
and so on. Although we have written the displacement as ds, it
is clearly the change dx in the position vector of the point in this case, so that
ds = dx x + dy y + dz z
(1-36)
according to (1-34). Comparing (1-35) and (1-36) with (1-20), we see that we can also
write du as the scalar product of ds and the vector
du
du
du
V u = x + y + zdx
ay
dz
(1-37)
du = ds Vm
(1-38)
so that
The vector that we have obtained in this way and that is written in terms of its
rectangular components in (1-37) is called the gradient of u and is also often written
grad u. We can regard (1-38) as the general definition of V u since it is written in a
form that is independent of a particular coordinate system. In other words, the gradient
is that quantity that will give the change in the scalar when it is dotted with the
displacement.
In order to understand the meaning of the gradient, let us consider Figure 1-18, in
which is indicated a series of surfaces each of which is made up of those points for
which u has the same value; in other words, these are surfaces of constant u, with
corresponding values ult u2, u3... . Now a displacement such as ds;, which takes us to
\
U
+ du
P(x, y, z)
Figure 1-17. Change in a scalar function of

position u as a result of the displacement ds.
Figure 1-18. Surfaces of constant u.

The gradient is perpendicular to such
a surface.
1-9 GRADIENT OF A SCALAR
13
a point on the same surface, does not take one to a point where u has changed.
Therefore, dul = dsl v = 0. Comparing this with (1-15), we see that V u and dsx
are perpendicular; thus v u is perpendicular to a surface of constant u, as is also shown
in Figure 1-18.
Now let us consider displacements of constant magnitude ds0 from a given point but
of varying directions such as ds', ds", and dsof
Figure 1-19. We see from (1-38) and
(1-15) that the change in u resulting from any of these displacements is given by
du = ds(} | V w | cos ^ and is different for each displacement only because of the different
value of the angle ^ between it and the fixed direction of V . We now see that du will
be a maximum when cos ^ = 1 or ^ = 0 so that V u and the corresponding displacement are parallel. In other words, the direction of the gradient is also the direction in
which the scalar has its maximum rate of change, and its magnitude is just that
maximum rate of change of u.
A unit vector n that is perpendicular to a given surface at a given point is called the
normal vector and is illustrated for a surface u = const, in Figure 1-20. But we have just
seen that V u is also perpendicular to the surface so that n and V u are parallel. Thus,
by using (1-4), we can write
h
Vu
= T^T
|V|
C1"39)
In this case, ft will also give the direction in which u is increasing.
Example
Suppose we consider a two-dimensional case where u = y2 kx where k = const. If
we write y2 = kx + u we see that the surfaces of constant u are curves in the xy plane.
In fact, they are parabolas as illustrated for k = 1 and some specific values of u in
Figure 1-21. Substituting this expression for u (with k = 1) into (1-37), we get
V
x + 2yy
14
VECTORS
so that, from (1-6), |Vw| = (1 + 4 y 2 ) 1 / 2 and therefore

x + 2yy
,/2
n =
(1-40)
(i + V )
according to (1-39). In using this result, we of course have to use the value of y
corresponding to a point on the curve for the given u. For example, let us evaluate n for
the point P of Figure 1-21 where the parabola corresponding to u3 = 2 crosses the
positive y axis. Here xP = 0, and yj = xP + w3 = 2 so that yP = \/2. Substituting this
into (1-40), we get
= j x + y2 y. This vector, with its negative x component and
considerably larger positive y component is also shown in the
figure.
1 - 1 0 OTHER DIFFERENTIAL OPERATIONS
It is quite possible that the components of a vector can also depend on position, so
that, for example, Ax(x, y, z), and so on. Then A will generally change as we go from
one point to another, in magnitude or direction or both, and we can write A =
A(x, y, z) = A(r). In the last term we have used a compact notation that we find
convenient for expressing A as a function of the coordinates of the point through its
position vector. A vector whose value is thus given at every point in space is called a
vector field. We now consider some specific ways in which it can change with position.
Looking back at (1-37), we see that Vw can be interpreted as the product of u and
the del operator given by
d
1-11 THE LINE INTEGRAL
15
It will be shown in Section 29-3 that this somewhat abstract operator has the same
mathematical properties as the displacement of a point and hence can be treated as a
legitimate vector. We can now perform two differential operations of interest by using
our two forms of multiplication of vectors.
Using (1-20) and (1-41), we get
dA,
Bj_4,
V A =
+
+
(1-42)
dx
By
dz
This scalar product is called the divergence of A and is often written div A.
Using (1-27) and (1-41), we obtain
V X
A = x
dA,
By
9A*
+y
BA
d_Ay
y
Mi
+ i
dx )
\ 3x
Bx
dz
" \ Bz
Bz ]
3 A,
dy
(1-43)
This vector product is called the curl of A and is often written curl A. It can also be
conveniently written as a determinant as in (1-28):
X
d
y
3
z
3
dx
Ax
By
dz
/K
V x A =
(1-44)
The significance of the names "divergence" and "curl" will become clearer as we use
them in situations in which they naturally arise.
Another operator of great interest and utility is the Laplacian:
V2 = V
dx'
Br
(1-45)
Bz'
For example, if applied to a scalar,

V u =
B2U
dx'
d2u
d2u
By
Bz'
while the expression V 2A represents the three equations in which v

of the three rectangular components of A, that is,
V2A
d2Ax
Bx2
d2A,
By2
B2A,
Bz2
(1-46)
2
operates on each
(1-47)
with similar expressions for the other two components.

We also note two useful results involving del. Because of (1-24), the curl of a
gradient is zero:
V x V = 0
(1-48)
Similarly, the divergence of a curl is zero,
V ( v x A) = 0
(1-49)
because of (1-29) and (1-24).

We now turn to some integrals involving vectors. Although many possibilities can be
imagined, two particular ones are of interest to us and we take them up in turn.
1-11 THE LINE INTEGRAL
Let us imagine starting at some given initial point ^ ( x , , yt, zt) and moving to a given
final point P f ( x f , y{, z f ) along a specific given curve C (a "line" or " p a t h " ) as shown
16
VECTORS
in Figure 1-22. The whole traversal of this path can be treated as the vector addition of
a sequence of infinitesimal displacements ds along C. We assume that there is a vector
field A so that its value can be found at all points along the way. At each intermediate
step, we evaluate A, multiply its component along ds with the magnitude of ds, and
add all of these quantities. The result is called the line integral of A along C and is
given by
fÂ cos ^ds
>1
= J ( a cos ^ ds = J f A ds
c
c
(1-50)
Perhaps the most familiar example of a line integral is that of work done on a particle;
in that case, A would be a force acting on the particle.
If the path of integration follows a closed curve, for example, a circle, then the initial
and final points will coincide. We write the line integral for this case as
<> A ds
This integral is sometimes called the circulation of A; it may or may not be zero
depending upon A as we will see.
If r is the position vector of each point on C, then ds = dr and we can use (1-20)
and (1-34) to write
j A ds = j (Ax dx + Ay dy + Az dz)
(1-51)
In using (1-51), we must be sure to take into account the fact that dx, dy, and dz
cannot be varied independently because the coordinates x, y, z are related by the
equation describing the path. Similarly, the expressions for Ax = Ax(x, y, z) etc. must
also take this interdependence into account. These considerations are best illustrated by
looking at a specific case.
i
Example
Let A = x2x + y2y + z2z and let us choose as our path that part of the parabola
y2 = x between the origin (0,0,0) and the point (2,^2,0)-, this curve is exactly the
parabola illustrated for u2 = 0 in Figure 1-21. Here z = const., so that dz = 0 and the
integrand of (1-51) becomes simply
Axdx
+ Ay dy = x2 dx + y2 dy
1 - 1 2 VECTOR ELEMENT O F AREA
17
We can write this as a function of one variable by using the equation of the path. Since
y2 = x, 2 y dy = dx or dy = dx/2\[x and y2 dy = \\J~x dx\ thus we get
f A ds = f ( x 2 + jv/x) dx = [ f x 3 + j x 3 / 2 ] q = | ( 4 + ]/2)
JR
JN
[i-12 VECTOR ELEMENT OF AREA

Before we consider integrals similar to the above but instead representing sums over
given areas, we will find it helpful to look in some detail at the representation of area in
vector terms. In Figure 1-23 we show an infinitesimal element of area da that has some
particular orientation with respect to the coordinate axes. We also see that a direction
can be associated with this area, that is, the unit vector ft, which is normal to the
surface. Thus, we can associate a vector d a with this element of area and write it as
da = da h
(1-52)
by following the general form of (1-4). It is clear, however, that there is some ambiguity
in this definition because we could equally well have chosen ft to be in the opposite
direction and it would still be perpendicular to the area element da. Hence we need to
supplement (1-52) with a convention that will tell us what to do; there are two cases to
be considered.
First, da may be part of an open surface, that is, it is bounded by a closed curve C;
a page of this book is an example of such an open surface. In this case, the first step is
to choose a sense of traversal around the bounding curve; once this is done, curl the
fingers of the right hand in the sense of traversal and then, by convention, the direction
in which the thumb is pointed is to be taken as the direction of ft. This right-hand rule
is illustrated in Figure 1-24; note how the direction of n would be reversed if C were
traversed in the opposite sense.
Second, da may be part of a closed surface. In this case, there is no bounding curve
C but the surface divides a volume into an inside and an outside; the surface of a ball
is such an example. Here the direction of ft is always chosen to point from the inside to
the outside. This is illustrated in Figure 1-25 where the corresponding directions of the
outward normal are shown for several points.
W h e n (1-52) is combined with an expression for ft of the form (1-8), d a can be
written in component form as
da = dax x + da v y + da, z
Figure 1-23.
area.
(1-53)
The normal vector for an element of
18
VECTORS
Figure 1-24.
Definition of the direction of the normal vector.
Figure 1-25.
surface.
Various outward normals for a closed
where
dax = lx da
day = ly da
daz = lz da
(1-54)
and where l x , lY, I, are the components of ft, that is, its direction cosines. Our
experience with multiple integrals has made us familiar with the use of dxdy as an
element of area in the xy plane, and it is clear that this expression is somehow related
to one of the components of da. In order to find this relationship, let us consider Figure
1-26, which shows a rectangular area element da of sides b and c so that da be. The
plane of the area is parallel to the y axis so that ft is parallel to the xz plane and makes
an angle y with the z axis; Figure 1-27 shows a side view looking along the y axis
toward the origin. The figures also show the projection of the area onto the xy and yz
planes; these projections are rectangles of area dxdy and dy dz respectively; it is also
evident from the figure that dy = c. The projection of da onto the xz plane is the line
labeled b in Figure 1-27. The other two direction angles of n are obtained by
comparing these two figures with Figure 1-8 and are seen to be a = 90 y and
/? = 90, so that the direction cosines are l x = siny, / = 0, and l z = cosy. Substituting these values into (1-54) and using Figure 1-27, we find that daz = da cos y =
(b cos y ) c = dxdy, which is just the projection of the area da onto the xy plane, that is,
the projection perpendicular to the corresponding coordinate axis. Similarly, we find
dax = dy dz while day = 0 for this special case.
Now suppose that nz were negative so that y > 90 while everything else was kept
the same; this is illustrated in Figure 1-28. Comparing with Figure 1-26, we see that the
projections on the xy and yz planes will still be rectangles of respective areas dxdy and
dy dz. Now, however, cos y will be negative so that da z = da cos y = dx dy. On the
other hand, since a = y 90, cos a will be positive and dax = dy dz as before; day is
still zero.
These considerations can be generalized to a situation in which ft makes arbitrary
angles with all axes. The magnitude of a given component of da along a given axis will
be equal to its projection on the coordinate plane perpendicular to the axis and will be
1 - 1 2 VECTOR ELEMENT O F AREA
19
Figure 1-26. Determination of the components

of a vector area element.
Figure 1-27. Side view of situation

in Figure 1-26.
Figure 1-28. An area element with a

negative z component.
given, in rectangular coordinates, by the product of the corresponding differentials.

Since we always treat these differentials as positive as written, the actual component
will be obtained by multiplying this product by a plus or a minus sign depending on the
sign of the corresponding component of ft. Thus we have succeeded in finding a
representation of the rectangular components of an element of area and we can write
dax = dydz
day = dzdx
daz = dxdy
(1-55)
where the plus sign is to be used for a given component if the direction angle of ft with
the corresponding axis is less than 90, while the negative sign is used when the
direction angle is greater than 90.
20
VECTORS
|1-13|THE SURFACE INTEGRAL
Consider a surface S\ as shown in Figure 1-29, we can divide S into vector elements of
area da as discussed in the previous section. We assume the existence of a vector field A
so that its value can be found at all points of S. At each element of area, we evaluate A,
multiply its component in the direction of da by the magnitude of da, and add all of
these quantities. The result is called the surface integral of A over S and is given by
f A cos
da = f A ft da = f A da
JS
JS
JS
(1-56)
This integral is also called the flux of the vector A through the surface S. We have
written (1-56) with a single integral sign for convenience, but it actually represents a
double integral.
If the surface is a closed surface, it is useful to indicate this explicitly by writing the
integral as
The value of this integral may or may not be zero, depending on A as we will see.
Using (1-20) and (1-53), we can write (1-56) in terms of rectangular coordinates as
f A da = Jf (Ax dax + A da
s
s
+ Az daz)
(1-57)
In using (1-57), we would need to use (1-55) and then be sure to take into account the
fact that dx, dy, and dz cannot be varied independently because the coordinates JC, y, z
are related by the equation describing the surface S. This interdependence must also be
taken into account in determining the limits of integration and in the writing of the
components such as Ax = Ax(x, y, z). Let us illustrate these points with a specific case.
Example
Let A = yzx + zxy + xyz and let us choose our area S to be that part of a circle of
radius a centered at the origin and in the first quadrant of the xy plane as shown in
Figure 1-30; the equation of the circle is x2 + y2 = a2. Here the area is perpendicular
to the z axis so that ft can be taken as either z or z. Let us arbitrarily choose ft = z;
then the only component of da is daz = dxdy as obtained from (1-55) with the plus
sign since the direction angle y = 0, and (1-57) becomes
f A da = Jf x y dxdy
s
s
(1-58)
1 - 1 4 THE DIVERGENCE T H E O R E M
21
(a2 -x2)Y>
Figure 1-30. Area of integration for the example of Section

1-13.
The element of area is shown more darkly shaded in the figure. We choose to evaluate
this integral by first integrating over y while keeping x constant; this will add up the
contributions from the less darkly shaded strip; the upper limit of integration on y
corresponds to the location of the point P and is found from the equation of the circle
for this fixed value of x, that is, yP = (a2 x 2 ) 1 / 2 . After this is done, we integrate over
all possible values of x, summing the contributions from similar strips, and thereby
covering the whole area S. Thus, (1-58) becomes
(A
Js
d*=
("xdx(("*-x')'/2ydy=
j0
Jo
= \ Jf
x(a2
rxdx{iy>Yf-'>'/2
o
- x 2 ) dx =
i[itf 2 x 2 -
i x 4 J o = ja4
m
o
We now turn to two important theorems involving the types of integrals we have just
discussed.
1 - 1 4 THE DIVERGENCE THEOREM
Consider a volume V enclosed by a surface S. Gauss' divergence theorem states that

<\ da = f V A dr
Ts
Jy
(1-59)
The integrals are taken over the total surface S and throughout the volume V whose
volume element is dr. Again, for convenience, we have written the volume integral with
a single integral sign although in reality it is a triple integral. Since is a closed surface,
the unit normal ft used for da is the outward normal according to our convention of
Section 1-12 as shown in Figure 1-25.
This theorem relates a surface integral of a vector to the volume integral of its
divergence. The surface integral depends only on the values of A on the surface,
whereas the volume integral requires a knowledge of V A (but not of A itself)
throughout the whole volume.
We will prove this theorem by direct evaluation. In rectangular coordinates, the
volume element is
4
dr dx dy dz
(1-60)
22
VECTORS
and, using (1-42), we can write the volume integral as a sum:

r
r dAx
f dA
r dA
f V A dr = I dxdydz + j dxdydz + / dxdydz
Jy
Jy
OX
Jy
dy
Jy
dz
(1-61)
Consider the first integral. Our first step will be to integrate over x while keeping y and
z constant at the values y0 and z0. Thus we will be summing up the contributions from
a rod of cross section dydz. This rod is shown in Figure 1-31 where its projection on
the yz plane is also indicated. The rod intersects the surface S at the points P1 and P2
and thus defines two elements of area dax and da2 of S whose directions are shown.
(For clarity, the rest of V and S are not shown in the figure.) The coordinates of the
points P1 and P2 will be (jcx, yQ, z 0 ) and (x2, y0, z0) respectively where x1 and x2 are
found as the values that satisfy the equation defining the surface S; thus x1 and x2 are
the limits of integration for x. In this step, then, 3Ax/dx
will be a function only of x
since y and z are constant, so that (dAx/dx)
dx = dAx.
Therefore, the first integral in (1-61) becomes
2
/ j dydz f
J f [A X(x2,
y0, z 0 ) - Ax(xlf
y0, z 0 ) ] dydz
(1-62)
X\
In the integrand of this result, the term in brackets is the difference in Ax when
evaluated at the points P2 and Pv which we can write as AX(P2) AX(PX). From
(1-55) and Figure 1-31, we see that da2x = dydz while dalx = dydz because the
angles made by these areas with the x axis are less than 90 and greater than 90
respectively. The integrand in (1-62) thus can be written as
AX(P2)
da2x - Ax(Px)(-dalx)
= Ax2da2x
+ Axldalx
(1-63)
which is just equal to the total contribution to the surface integral of Ax dax arising
from the areas da2 and da1 intercepted on the surface S by this rod. Thus, when we
perform the integrations over y and z in (1-62), we will be summing the contributions
of all rods of this type; the contribution from each rod will be A x dax from its share of
the area, so that the final result will be the surface integral of Axdax over the whole
area S. In other words, we have found that
f
d A
dy dz = <f>Ax dax
In the same manner, the last two integrals in (1-61) will be found equal to
<j)Ay day
Figure 1-31.
and
<>AZ daz
Volume used in deriving the divergence theorem.
(1-64)
1 - 1 4 THE DIVERGENCE THEOREM
23
respectively, so that, if we add these to (1-64), substitute into (1-61), and use (1-20), we
find that
which is exactly (1-59) and proves the theorem.

We have proved this only for a volume b o u n d e d by a single surface, but we can
easily extend the proof to a region bounded by several surfaces, such as a hollow ball.
Figure 1-32 shows a volume V surrounded by two surfaces St and S2; two representative outward normals to the volume are shown as ft and n / . We now imagine a plane
intersecting the volume and dividing it into two volumes V2 and V1; the trace of this
plane is shown by the dashed lines AB and CD. The volume V2 is now enclosed by a
single surface composed of those portions of S2 and
to the left of the plane plus the
plane surfaces of intersection shown by AB and CD; the outward normal to the new
part of the bounding surface is shown as ft2. Similar remarks apply to V1; its
corresponding outward normal is n x . Applying (1-59) to each of these volumes and
adding, we obtain
f
VX+V2
V A dr = f A du + f A da. + f
J
Sl
S7
JABCD
A -h2da+
f
JABCD
A -hxda
In the last two integrals, the normals are oppositely directed so that at each point of
ABCD, ft 2 = ftjwhile the values A and da are the same; thus these integrals cancel
each other and we are left with
which is the same as (1-59) because the total volume is Vx + V2 and the total bounding
surface is Sx + S2.
This proof can obviously be generalized to an arbitrary number of bounding
surfaces by introducing as many intersecting surfaces like ABCD as would be necessary.
If we now apply the divergence theorem to a particular simple situation, we can get a
useful and illuminating result. Consider a point P at the center of a small volume AF.
If AV is very small, V A will be nearly constant throughout the volume so that it is
11
B
Figure 1-32. A volume bounded

by t w o surfaces.
24
VECTORS
easily seen that the volume integral in (1-59) can be written as

f V A dr = ( V A ) p AF
jAK
where ( V A)P is the average value of V A in the vicinity of P. Putting this into
(1-59) and dividing by AF, we get
<V-A>P- ~^
'RFA
(1-65)
If we now let AF > 0, while P remains at the center, the average value of v A in the
vicinity of P becomes the value of V A at P\ writing this simply as V A, we have
V A =
1
1
lim -<T)A da
AK-*0 A V J s
(1-66)
This important expression for v A is in a form independent of a particular coordinate

system, in contrast to (1-42), and hence can be taken as a general definition of the
divergence of a vector in the same sense that (1-38) is the general way of defining a
gradient. The result (1-66) also gives us a much better understanding of the significance
of the divergence since we see that it is a measure of the outward flux of the vector
through a small area about the point in question.
It is possible to start with the definition (1-66) and, by evaluating the flux of A
through the surface around a volume AxA_yAz, find the expression for v A in
rectangular coordinates; the result is, of course, (1-42).
1-15 STOKES' THEOREM
Consider a surface S enclosed by a curve C. Stokes' theorem states that

( j ) \ ds = J ( v X A) da
c
s
(1-67)
and hence relates the line integral of a vector about a closed curve to the surface
integral of its curl over the enclosed area. Since 5 is an open surface, the direction of
da in (1-67) is determined from the chosen sense of traversal around C according to
(1-52) and the right-hand rule illustrated in Figure 1-24; it is essential that this sign
convention be used in any application of (1-67).
It is interesting to note that (1-67) does not require S to have any particular shape
other than that it be bounded by C; thus there are many possibilities in choosing the
surface. In general, the values of the integrand ( v X A) da will be different for
the points on all these surfaces, but (1-67) tells us that the sum of all these terms will be
the same, since the line integral depends only on the values of A along the common
perimeter. This point is illustrated somewhat schematically in Figure 1-33. Suppose, for
S"
!
\
/
/
5
c
\I / /
y
J
Figure 1-33.
curve.
Surfaces with the same bounding
1-15 STOKES'THEOREM
25
simplicity, that C is a closed curve lying in a plane such as a circle. S could be taken as
the plane area enclosed by the circle; when viewed from the side, C and S would both
appear as the line shown. The dashed lines represent traces of other possible surfaces
S', S", and so on, all of which have C as a boundary, and the integral of ( v X A) da
over any of them will give the same result.
We prove this theorem by direct evaluation of the surface integral by using our
previous expressions in rectangular coordinates. Using (1-20) and (1-43), we find that
f
c{
dA
J ( V X A) rfa = / s (
r
+
dA
day -
( dAz
da
L\~Sy *
f(
dA
da, j + / J
dAz
da
~ ~d^ ")
dA
da, -
daxf
\
(1_68)
where we have grouped the terms by components of A. Consider the first integral that
we designate by Ix. We evaluate it by first integrating over a strip of width dx, which is
parallel to the yz plane, and a distance x from it. Then, by integrating over x, we sum
up the contributions from all of the strips into which 5 can be subdivided. Initially, we
assume S to be simple enough that we can choose the orientation of our axes so that
both y and z increase as we go from the beginning to the end of the strip. This
situation is illustrated in Figure 1-34, which also shows the projections of the strip on
the xz and xy planes as an aid to understanding the orientation of the surface. Px and
P2 are, respectively, the initial and final points of our integration, that is, they are the
points of intersection of the strip with the bounding curve C and their coordinates
satisfy the equation describing C. The area element da is shown at an intermediate
stage of the integration; we see that the direction angles of da have values such that a
Figure 1-34.
Area used in deriving Stokes' theorem.
26
VECTORS
and y are less than 90, while /? > 90. (A pencil held normal to a piece of cardboard
oriented in the same sense as the shaded strip will quickly make this clear.) Thus,
according to (1-55), day = dxdz and daz = dxdy and we can write
rp
= ~ Lstrips
, .
d x
( 3A
/ J
1 7
dA
dy +
1 7
\
dz
(1
"69)
In the term in parentheses, dy and dz are not independent because y and z are related
by the equation for S and the value of x involved. Since this integrand is being
evaluated on the strip for which x = const., and hence dx = 0, we can add
( dAx/3x)
dx = 0 to it to give it a quickly recognizable form:
3AX
dAx
dAx
dx + dy
+ dz = dA
7
dx
dy
dz
As a result, (1-69) becomes
I
dx
= ~ F
strips
FPL
P1
dAx =
F
[ A
^strips
( P
) -
AX{PX)}
dx
(1-70)
If we turn again to Figure 1-34, and consider the displacements dsx and ds2 along C at
the corresponding limits, we see that, at Px, dsx has a positive x component so we can
write dsXx = dx, while at P2, ds2 has a negative x component with the result that
ds2x = dx. Consequently, the integrand of (1-70) can be written as
-Ax(P2)(~ds2x)
+ Ax(Px)(dslx)
= Ax2ds2x
+ Axldslx
(1-71)
which is just equal to the total contribution to the line integral of Ax dsx arising from
the displacements ds2 and dsx intercepted on the curve C by this strip. Thus, when we
perform the final integration over x in (1-70), that is, adding up the contributions of all
of the strips, the contribution from each strip will be Ax dsx from its share of the
bounding curve; the final result will be the line integral of Ax dsx over the whole curve
C. In other words, we have found that
(1
"72)
Similarly, the last two integrals of (1-68) can be shown to have the respective values
<,Aydsy
and
^>Azdsz
Substituting these results, along with (1-72), into (1-68), we find that
(VX A). da
/,
= <f){Ax dsx + Ay dsy + Az dsz) =
ds
which is exactly (1-67) and proves the theorem.

The theorem can be extended to the case of a surface bounded by more than one
curve by a method similar to that we used for the divergence theorem. An example of
such a situation is shown in Figure 1-35. Note the direction of traversal of the inner
bounding curve; this sense is chosen to keep the area of interest to one's left as one
moves along the curve and is seen to be equivalent to the right-hand rule illustrated in
Figure 1-24. We would divide S into surfaces each bounded by a single curve by
introducing as many pairs of coincident lines as we would need; two such pairs are
shown dashed in the figure. Then Stokes' theorem can be applied to each of these
surfaces and the results added. The contributions to the line integrals from the
introduced lines would cancel since they are traversed in opposite directions and the
1-15 STOKES'THEOREM
Figure 1 -35.
27
An area bounded by t w o curves.
final result would again be (1-67). A similar procedure will also dispel any remaining
qualms relating to our assumption that the surface in Figure 1-34 was "flat enough" for
us to orient our axes so that y and z always increased as we integrated along the strip.
If the surface is very convoluted, we can divide it into pieces each of which is flat
enough so that, if necessary, we can choose a different set of axes for each piece which
will satisfy the requirements of Figure 1-34; such a division is illustrated in Figure 1-36.
When we apply the theorem to each piece and add the results, the contributions from
the dividing lines will cancel each other, and we will again obtain (1-67).
Consider a point P at the center of a small area A ah. When (1-67) is applied to this
case, (V X A) ft will be nearly constant over the area, so that, much as in the last
section, we can write
where the left-hand side is the average value near P of the component of V X A in the
direction of n according to (1-21). Now if we let A a > 0, the average value near P
becomes the value at P, or
(V X
A) n =
(1-73)
which gives the component of V X A in a given direction in terms of the line integral
of A about a small area normal to this direction. Thus we can take (1-73) as a general
definition of the component of the curl in a given direction. If we do this for three
Figure 1-36. Division

of an area for a general
proof of Stokes' theorem.
28
VECTORS
mutually perpendicular directions (such as x, y, and z), we will get the components of
V X A in each of these directions and thus obtain the whole vector v X A. When this
procedure is carried out for rectangular coordinates, the result is, of course, (1-43).
1-16 CYLINDRICAL COORDINATES

Up to now, we have used only rectangular coordinates with their constant unit vectors.
However, many problems are more conveniently stated and worked in other systems of
coordinates, and we want to see what happens to many of our results. We need to deal
with only two of the more important ones.
The first is cylindrical coordinates in which the location of a point P is specified by
the three quantities p, <p, 2 whose definition is illustrated in Figure 1-37; this figure also
shows the position vector r of the point along with three new unit vectors that we will
define shortly. We see that when r is projected onto the xy plane, p is the length of this
projection while <p is the angle that it makes with the positive x axis; z is the same as
the corresponding rectangular coordinate. The relation between the cylindrical and
rectangular coordinates of P is seen from the figure to be
x p cos <p
y P sin <p
z = z
(1-74)
so that
p = (x2 + y2)1/2
tan<p=^
(1-75)
We can now define a set of three mutually perpendicular unit vectors as follows:
first, z is the same as the rectangular z; second, p is chosen to be in the direction of
increasing p and is perpendicular to z so that p is parallel to the xy plane; finally, <p is
defined to be perpendicular to both of these and in the direction shown. We see that <p
1-16 CYLINDRICAL COORDINATES
29
is perpendicular to the semiinfinite plane <p = const., and its direction is therefore in
the sense of increasing <p. These unit vectors are shown at the location of P in order to
emphasize that they are functions of P in the sense that if P is displaced, both p and cp
change their directions, although z does not change. Thus, these unit vectors are not all
constants, in contrast to x, y, and z.
Since p,
and z are mutually perpendicular unit vectors, they satisfy relations
analogous to (1-18), (1-19), and (1-25):
A
p - p
p
A
\ /
$ = $
p X f
(p-<p =
= z
A
z = z
A
Z- Z =
v /
<p X
p = 0
z = p
A
(1-76)
z X p = 9
The rectangular components of p and $ are found from inspecting Figure 1-37; it is
helpful to imagine them projected onto the xy plane, and the results are
p = cos px + sinpy
= sin<px + cos <py
(1-77)
These equations can be solved for x and y to give their components in cylindrical
coordinates:
x = cos pp sin <pq>
y = sin <pp + cos (pep
(1-78)
By differentiating (1-77), we can find explicitly how p and cp vary as P is displaced:
dp
da>
a n d
(1
"79)
Since p, 9 , and z are mutually perpendicular, we can express any vector A in terms
of its components along these directions; by analogy with (1-5), we write A in the form
A = App +
+ Azz
(1-80)
For the particular case of the position vector r, we see from Figure 1-37 that
r = pp + zz
(1-81)
which can also be obtained from (1-11) by substitution of (1-74) and the use of (1-77).
We can also find the differential of dx from (1-81) and (1-79):
dx = dp p + p dp + dz z, = dp p + p dtp <p + dz z
(1-82)
so that its components in the directions of increasing p, cp, and z are dp, p dtp, and dz,
respectively. These component displacements are shown in Figure 1-38, and we see that
they correspond to the distance moved by P resulting from the change of any one
coordinate while the other two are held fixed. The shaded volume element has as sides
just the components of dx given by (1-82). Therefore, the element of volume in
cylindrical coordinates is given by
dr = p dp dtp dz
(1-83)
We also see from this figure that the areas perpendicular to the unit vectors are p dtp dz,
dp dz, and p dp dtp, so that the components of an element of area da are given by
dap=
pdtpdz
da(p=
dp dz
da2=
pdpdtp
(1-84)
where the proper sign is to be chosen in the same manner as for (1-55).
N o w we can go on to find what our differential operators become when expressed in
this system. If u = w(p, tp, z), then
du
du
du
du = dp + dtp + dz
dp
dtp
dz
and, on comparing this with (1-82) and the general definition of the gradient given in
(1-38), and noting that ds = dr in this case, we find that the expression for the gradient
30
VECTORS
Figure 1-38.
Volume element in terms of cylindrical coordinates.
in cylindrical coordinates is
du
v =
dp
1 du
du
p dtp
dz
v-j- +v
(1-85)
so that the del operator can be written as

d
I d
(1-86)
TP
p ^
^
By applying (1-86) to (1-80), we can find corresponding expressions for the divergence
and curl. In doing this, however, we must remember that the unit vectors are not
constant and we have to take (1-79) into account; in addition, we need to use (1-76).
We illustrate this process for v A:
__
dA
p-
P\
dA.
+z
dz
dp
dp
P +
dA
<
"
dtp
i 1
dAz
dp
3A_
dtp
<p +
dz
1 dA
A
+
dp_
9
^ I
^
.
<p +I
dtp
+ 2
dA,
p *
1 / 3_A_
+i <p
dA r
% ^
dp
+ < p
H ^
-A
= 9
dtp
Ml
dz
+ Ml
dz
-*
+ A
9Az
1 7
1 - 1 7 SPHERICAL C O O R D I N A T E S
which is usually written as
1d
.
V A -
i
(pA
p op
) H
1 dA
dA,
f~
p dtp
31
(1-87)
dz
In the same manner, the expression for the curl becomes

lldAz
V
dA
(1-88)
while V V turns out to be
1
V u
p dp \
1 d2u
du
dp
P2 dcp2
d2u
(1-89)
dz"
where u is a scalar function of position.

It is very important to remember that (1-86), (1-87), (1-88), and (1-89) cannot be
obtained from the corresponding expressions in rectangular coordinates as given by
(1-41), (1-42), (1-43), and (1-46) by the simple replacement of x, y, z by p, <p, z.
Similarly, (1-44) and (1-47) can only be used for rectangular coordinates; see (1-120)
for the definition of v 2A for other coordinate systems. You would be surprised at how
often these mistakes are made.
1-17 SPHERICAL COORDINATES

In this system, the location of a point P is specified by the three quantities r, 6, <p
shown in Figure 1-39. We see that r is the distance from the origin and thus the
magnitude of the position vector r, 6 is the angle made by r with the positive z axis,
while cp is again the angle made with the positive x axis by the projection of r onto the
xy plane. The relations between the rectangular and spherical coordinates are seen to
be
x = r sin 6 cos <p
y = r sin 6 sin <p
z = r cos 6
(1-90)
so that
2
r = (x
+y
2 1/2
+ z )
tan6 -
/ 2 _|_ 2 'J 1 / 2
z
tan<p =
x
(1-91)
We define a set of mutually perpendicular unit vectors ?, 0, and <p in the sense of
increasing r, 6, and <p, respectively, as shown in Figure 1-39; we see that as the location
of P is changed, all three of these vectors also change. They satisfy relations analogous
to (1-76):
f . f = 6- e = $ - $ = i
r - 0 = 0 - q> = <p-r = O
r xS = $
0X$ = r
(1-92)
$ X ? = 0
Their rectangular components are found from inspection of the figure to be

r = sin 6 cos <px + sin 0 sin <py + cos dz
= cos 6 cos <px + cos 6 sin <py sin Oz.
= sin <px + cos <py
(1-93)
32
VECTORS
Figure 1-39.
Definition of spherical coordinates.
and therefore
x = sin 6 cos <pr + cos 0 cos cp0 sin <p
(1-94)
z = cos Or sin Oh
By differentiating (1-93), we obtain
dr
~d0
30
~d6
= 0
= r
dr
= sin flcp
o
~de
= 0
d(p
(1-95)
sin dr cos 00
A vector A will be written in component form as

A = Arr + AeQ + Av$
(1-96)
r = rr
(1-97)
The position vector is

and its differential is found with the use of (1-93) and (1-95) to be
dr = dr r + r dd 0 + r sin 6 dcp <p
(1-98)
These component displacements are shown in Figure 1-40. Since the shaded volume
element has as sides just the components of (1-98), the volume element will be
dr = r2 sin Q drdd dtp
(1-99)
1 - 1 7 SPHERICAL C O O R D I N A T E S
33
r s i n 8 dtp
Figure 1-40. Volume

element in terms of
spherical coordinates.
r s i n 6 dip
while the components of an element of area da will be given by

dar=
+ r2smddddcp
dag = rsinddrdcp
dciy = rdrdd
(1-100)
If u = u(r, 9, <p), then

du
du
du
du = dr + -~t d6 + dcp
or
do
dcp
so that the gradient as obtained from (1-38) and (1-98) is
du
1 du
1
du
Vu = r
dr
(1-101)
h 0 7 7 + cp-
r dO
r sin 6 dtp
showing that the del operator is

d
d
A i
r + 0 + $
r sin 0 dcp
r dd
dr
(1-102)
If we now proceed to use (1-102) and (1-96) in the same manner as in the last section,
taking account of (1-92) and (1-95), we find:
1
d
1
d
1
dAw
\
2( ,
V A = (r
Ar)
+ :~ "T7 ( sin 6A g ) + :
(1-103)
2
r sin 6 dd
r sin 6 dcp
dr
V x A
do
r sin 0
A
<P
+
(sin 0AfV )
dA
dAe
dtp
0
+ -
^
r sin0 dcp
dr
A\<p)
(1-104)
34
VECTORS
Remarks similar to those following (1-89) also apply here. You cannot get the correct
spherical coordinate expressions for (1-102) through (1-105) f r o m the corresponding
ones in rectangular coordinates by simply replacing the symbols x, y, z by r, 6, cp.
1-18 SOME VECTOR RELATIONSHIPS
We list below some general results that are useful in evaluating and simplifying m a n y
of the expressions encountered when using vectors; we have already seen examples in
(1-29) and (1-30). The first set of these expressions can be verified by straightforward,
but sometimes tedious, calculations. Such verifications are most easily done in rectangular coordinates since the unit vectors involved are constants. For completeness, it
should be noted that V 2 A in (1-120) can be resolved as we did in (1-47) only for
rectangular components; for other coordinate systems, (1-120) should be taken as a
definition of V 2A to be evaluated from the rest of the identity.
(A X B) (C X D ) = (A C ) ( B D ) - (A D ) ( B C)
d
du
- ( * A ) - - A
d
(1-107)
dB
B + A
da
da
dX
d
(A X B) =
da
da
v ( + v) =
dA

d\
(A B) =
(1-106)
(1-108)
da
dB
X B + A X
(1-109)
da
+ Vf
V M
(1-110)
v ( w ) = V v + vVu
(1-111)
v ( A B) = B X ( v X A) + A X ( v X B) + (B V )A + (A v )B
(1-112)
v(c
0 = C
where
C = const.
(1-113)
V (A + B)
V A + V B
(1-114)
v (MA)
A- ( v ) + m(V A)
(1-115)
V (A X B) = B- ( v X A) - A- ( v X B)
(1-116)
V X A + V X B
(1-117)
V x (u A)
( V ) X A + ( V X A)
(1-118)
X B)
<1
V X (A + B)
J>
t>
X A)
V )A -
( v B)A - ( v A)B + (B
(1-119)
(A V )B
(1-120)
V ( V A) - V A
where
BB.
I
dB x
(A V )B = x Ax +A
dx
+y
>
+ z A,
dx
dx
+ A7
+
dz
dy
dB y
dy
d^
A,
dy
+ A,
+ A,
dBy_
dz
dB^
dz
J*
(1-121)
If we take A to be a constant, but otherwise arbitrary, vector in (1-115) and (1-118),

while B is taken as constant in (1-116), and use (1-59) and (1-67), we obtain three
1-19
F U N C T I O N S O F THE RELATIVE C O O R D I N A T E S
3 5
interesting and occasionally useful integral theorems:

<uda = f Vudr
JS
Jy
(ax
da = - f (v
JS
Jy
(1-122)
X A) dr
(1-123)
(uds=
J V X da
(1-124)
c
s
We can obtain an alternative way of looking at the meanings of the gradient and curl
from (1-122) and (1-123) by considering a small volume AK and proceeding as we did
to get (1-66); we find that
V=
V XA =
1
lim - ( Ih u d a
AK-0 A V s s
(1-125)
1 t
lim -<T) da X A
AK-0 A V j s
(1-126)
Finally, if we integrate (1-115) over a volume and use (1-59), we get

<uA da = f [A- ( V ) + ( V A)] dr
Ss
(1-127)
Jy
Consider the particular case in which u = Bx, then (1-127) becomes

(f>Bx(A -da)=
JS
[ [A VBX + Bx(v
Jy
A)] dr
(1-128)
Since there are similar expressions involving the other two components of B, we
recognize that (1-128) will also lead us to
(b(A da) = f [(A V )B + B( v A)] dr
To
Jv
(1-129)
1-19| FUNCTIONS OF THE RELATIVE COORDINATES

As we proceed, we will find that we are constantly dealing with functions that depend
only on the differences of the coordinates, that is, they are functions solely of the
combinations x x', y y', and z z'. We see from (1-13) that these combinations
are simply the components of the relative position vector R, hence the name "relative
coordinates" for x x' and so on. Functions of this type have properties that will
enable us to simplify much of our later work and it is convenient to consider them now.
Let / be such a function; / could be a scalar or a component of a vector. Since /
depends only on the relative coordinates, we write it variously as / ( R ) = f ( x x', y
y', z z') f ( X , Y, Z ) where X x x', Y = y y', Z = z z'. Using the chain
rule of differentiation, we find that
df
df dx
df
df
df dx
df
j
!hc ~ J x
and
~ Jx
dx' ~ dx dx' ~
dx
so that
df( R)
dx
df{ R)
dx'
(1-130)
36
VECTORS
with similar expressions for the y and z derivatives. Following (1-41), we can define a
del operator V ' in terms of the primed coordinates as
d
V ' = x
dx'
1-
d
>-z
dz'
dy'
v(1-131)
'
If now, we calculate the gradient of / according to (1-37) and use (1-130) and (1-131),
we get
V/(R) = - V ' / ( R )
(1-132)
which shows us that when we are dealing with functions of the relative
V and V ' operators can be interchanged provided that the sign is also
If the vector A is a function only of the relative coordinates, A(R),
^ V (R), and vl z (R), we can apply (1-130) and its analogues to (1-42)
obtain results similar to (1-132):
coordinates the
changed.
so that Ax (R),
and (1-43) and
V A(R) = - V ' A(R)
(1-133)
V X A(R) = v ' X A(R)
(1-134)
The Laplacian operator defined in (1-45) is unchanged, however:

V2/(R) = V/2/(R)
(1-135)
Example
The magnitude of the relative position vector R as given by (1-14) is an important

example of the kind of function we have been discussing. By direct differentiation of
(1-14) we find that
dR
x x'
dR
to"
"
( W 3 6 )
with similar expressions for the y and z derivatives. If we combine these results with
(1-41), (1-131), and (1-13), we get
R
= R-
vR=-V'R=
(1-137)^
where R is the unit vector in the direction of the relative position vector. Similar results
are found for functions of R:
Sg(R)
3R
dg
dg
dx
dR dx
dR
so that
V g ( R ) = -V'g(R)
dg(R)
(1-139)
and, in particular,
V(R")
= -V'(Rn)
= n
An especially important case of (1-140) corresponds to n = 1:

/ 1 \
v
U ) The x component of (1-141) is

d I 1 \1
'
dx\R
( 1 \
~
R
(1_141)
' U r
d ( 1 \
/ i
dx' \ R i
(jc JC')
i
R
(1-142)
1-20
THE H E L M H O L T Z T H E O R E M
3 7
If we differentiate this once more, we get

d2 ( 1 \
dx2[ R) ~
1
R3
3 ( x x')2
R5
O"143)
with the use of (1-136). There will be similar expressions for the y and z second
derivatives; if we add them to (1-143), use (1-14) and (1-45), we find that
3
3 R2
t -t
r
R3
R5
I 1 \
v
=
\R)
so that, if we also recall (1-135),
2
V 2 ( ) = V , 2 f ) = 0
O * 0 )
(1-144)
R)
\R.
We have added the parenthetical term in (1-144) to remind us that all of our
calculations from (1-141) on were made with the implicit assumption that R = 0 so
that we were dealing with finite quantities. If we recall that V 2 = V V, we see from
(1-141) that (1-144) can also be written as
R
R
V - ( 172 \ == VV - I - 3 I = 0
^ )
' U
(**)
(1-1 4 5)
(R*
(1-146)
0)
It will be helpful later to know that

V
1 ^ 1 == - V
x V| 1 = 0
f )
" v x v ( S
(1-147)
which follows from (1-141) and (1-48).

1-20 THE HELMHOLTZ THEOREM
We will not prove this theorem at this time but simply quote it as an aid to
understanding the motivations for many of the procedures we will be following. We
will, in effect, prove it eventually, but piece by piece.
The theorem deals with the question of what information we need to calculate a
vector field. Basically the answer is that if the divergence and curl of a vector field are
known everywhere in a finite region, then the vector field can be found uniquely. We
consider a vector field F = F(x, y, z) = F(r) and assume that the functions v F = b(r)
and V X F = c(r) are given to us everywhere in a finite volume V, that is, they are
known functions of position. Then, if we define the following two functions,
1 r b{r') dr'
$ (Vr ) = [
(1-148)J
V
'
4irJv\rr'l
w
1
4TT
c(r') d t '
J V |r - r ' |
the theorem tells us that F can be found from

F = F(r) = - V<5 + V X s/
(1-149)
'
(1-150)
In these expressions, |r r ' | is seen to be the magnitude of the relative position vector
as given by (1-12) and shown in Figure 1-12.
Since F can be found from a knowledge of them, its divergence and curl are often
called the sources of the field. The point r where we evaluate F is called the field point,
38
VECTORS
while the point r' where the sources are evaluated for purposes of integration is called
the source point, dr' is then a volume element at the location of the source point.
Similarly, the derivatives in (1-150) involve the components of the field point r. The
functions $ and s/ are called scalar and vector potentials, respectively, because F is
obtained from them by differentiation. For reasons that will become clear later, the
divergence is known as the "charge" source while the curl is called a " c u r r e n t " source.
If there are sources at infinity, so that they are not all confined to a finite volume, the
above theorem is not absolutely correct, because one must also include certain surface
integrals involving F. We will handle such situations by special methods when we
encounter them.
M u c h of the content of modern electromagnetism involves various electric and
magnetic field vectors. Consequently, much of our effort will be devoted to finding
expressions for their divergence and curl f r o m f u n d a m e n t a l experimental results; this
theorem shows us why we want to know them. W h e n the complete set of these source
equations are found, they are called Maxwell's equations and form the fundamental
description of the electromagnetic field.
EXERCISES
I really understand the theory;
I just can't work the problems.
Anonymous
1-1
Verify Equation 1-2 by graphical methods.
1-2 Given the two vectors A = 2x 3y 4z

and B = 6x + 5y 4- z, find the magnitudes and
angles made with the x, y, and z axes for A + B
and A B.
1-3 Find the relative position vector R of the
point P{2,-2,3)
with respect to P ' ( 3,1,4).
W h a t are the direction angles of R?
F i n d the unit vector normal to each point of the

surface of these ellipsoids.
1-10
tion.
Verify (1-48) and (1-49) by direct calcula-
1-11 D o the example of Section 1-11 by integrating over y rather than x and thus show that
the same result is obtained.
1-4 Given the two vectors A = x + 2y + 3z and

B = 4x 5y 4- 6z, find the angle between them.
Find the component of A in the direction of B.
1-12 F i n d the surface integral of r over a surface

of a sphere of radius a and center at the origin.
Also find the volume integral of V r and comp a r e your results.
1-5 Given the vectors A = 2x 4- 3y 4z and

B = - 6x - 4y + z. Find the component of A X
B along the direction of C = x y + z.
1-13 Given the vector field A = xyx + yzy +

zxz. Evaluate directly the flux of A through the
surface of a rectangular parallelepiped of sides
1-6
Verify (1-29) and (1-30).
1-7 Show that A - ( B x C ) equals the volume of

a parallelepiped if A, B, and C are the vectors
representing the three edges with a common
corner.
1-8 A family of hyperbolas in the xy plane is
given by u = xy. Find v w . Given the vector
A = 3x + 2y 4- 4z, find the component of A in
the direction of V u at the point on the curve for
which u = 3 and for which x 1.
1-9
The equation giving a family of ellipsoids is

Figure 1-41. Rectangular parallelepiped with
a corner at t h e origin.
EXERCISES
39
a, b, c with origin at one corner and edges along

the positive directions of the rectangular axes as
shown in Figure 1-41. Evaluate / V A dr over
the volume of this same parallelepiped and compare your results.
1-20 Given A = ar 4 68 4 c<p where a, b, c are

constants. Is A a constant vector? Find V A and
V X A. Find the rectangular and cylindrical components of A, expressing them in terms of x, y, z
and p, 9 , z, respectively.
1-14 Calculate directly the line integral $A d s

of the vector A = _yx + xy around the closed
path in the xy plane with straight sides given by:
(0, 0) - (3, 0) ^ (3, 4) - (0, 4) - (0, 0). Also
calculate the surface integral of V X A over the
enclosed area and show that (1-67) is satisfied.
1-21 Find V r for the position vector r expressed in rectangular, cylindrical, and spherical
coordinates, thus showing that the same result is
obtained in all cases.
1-15 Given the vector field A = x2yx 4- xy2y

+ a V ^ c o s a x z where a, a, [3 are constants.
Evaluate directly the line integral of A over the
closed path in the xy plane shown in Figure 1-42.
The straight portions are parallel to the axes, and
the curved portion is the parabola y2 = kx where
k = const. Evaluate the surface integral of V X A
over the area S enclosed by C and compare your
Figure 1-42.
1-15.
1-22 Given A = app + b(p + czz where a, b, c

are constants. Find <f>\ da over the surface of a
right circular cylinder of length L and radius p 0 .
The axis of the cylinder is along the positive z
axis and the origin is at the center of the lower
circular face. Also find / V A dr over the volume
enclosed by the cylinder and compare your results.
1-23 Given the vector A = 4 r + 3 8 2<p, find
its line integral around the closed path shown in
Figure 1-43. The curved portion is the arc of a
circle of radius r0 centered at the origin. Also find
the surface integral of V X A over the enclosed
area and compare your results.
Path of integration for Exercise
1-16
Verify (1-88) and (1-89).
1-17
Verify (1-103), (1-104), and (1-105).
1-18 G i v e n A = a x 4 by 4 cz where a, b, c are

constants. Is A a constant vector? Find the cylindrical and spherical components of A, expressing
them in terms of p, <p, z and r, $, <p respectively.
1-19 Given A = ap + bSp + cz where a, b, c are
constants. Is A a constant vector? Find V A and
V X A. Find the rectangular and spherical components of A, expressing them in terms of x, y, z
and r, 6, <p, respectively.
Figure 1-43.
1-23.
1-24
Path of integration for Exercise
Verify (1-115) and (1-118).
1-25 By applying the divergence theorem to the

special case in which A is a constant but otherwise arbitrary vector, show that the total vector
area of a closed surface is zero, that is, <f>da = 0.
Similarly, show that <f>ds = 0. D o these results
surprise you?
1-26 Verify (1-122), (1-123), and (1-124) by using
the method suggested in the text.
The phenomena nowadays associated with the term "electricity" go back many
centuries to observations that when certain naturally occurring materials were rubbed,
they acquired the ability to exert forces on other objects. A typical material of this sort
is amber (elektron in Greek). The process is called "electrification by friction" or
" triboelectrification" and, in order to describe the altered state of the matter, one says
that it has become "charged" or has an "electric charge" on it.
After many experiments and much thought, it was finally concluded that electrification by friction did not represent a process of creation of electric charge, but rather a
separation of two types of charge that were originally present in equal amounts in the
uncharged "neutral" material. These two types of charge are arbitrarily called "positive" and "negative." Positive charge is defined as that which is left on a glass rod after
it has been rubbed with a silk cloth; since the process is one of separation, the silk cloth
will be left with a negative charge equal in magnitude to that on the glass rod. As
implied by these remarks, and confirmed by all subsequent experiments, electric charge
is conserved in the sense that net charge cannot be created or destroyed; we will put
this fundamental experimental result in quantitative terms in Chapter 12.
The forces between electric charges can be forces of repulsion as well as of
attraction. The first quantitative investigation of the dependence of these forces on the
magnitudes of the charges and the distance between them was made by Coulomb in
1785 and the result is known as Coulomb's law.
I2=T1 POINT CHARGES

We use the symbol q to represent electric charge. In a general situation, the charge of
an object will be distributed in some manner on or throughout it and the force between
two objects will depend on these distributions as well as on the total amount of each
charge. As a result, it is convenient to begin with the case of a point charge in which it
is assumed that all of the charge is located at a geometrical point in space. This is
obviously an idealization, but can be approximated very well in the laboratory by
making any distances of separation that are involved very large compared to the
dimensions of the charged objects.
In order to make further progress one has to be able to compare the magnitudes of
two point charges ql and q2- This can be done by introducing another arbitrary point
charge q, putting it at a fixed distance R from qv, and measuring the resultant force F,
on q; this is illustrated in Figure 2-la. Then qx is removed and replaced by q2 at the
same distance R from q\ the new force F2 on q can then be measured as indicated in
Figure 2-16. Since both q and R are the same in the two cases, the difference in the
forces can only be due to the difference in the numerical values of the charges qx and
q2, and it is natural to ascribe the magnitudes of the forces as being directly
proportional to the magnitudes of qx and q2. Accordingly, we can define the ratio of
their magnitudes as equal to the ratio of the magnitudes of the forces they produce on
40
2-2 C O U L O M B ' S LAW

(")
(bj
.
$2
.
1
41
> F,
>
Figure 2-1. Comparing charges by comparing

t h e forces they exert.
the arbitrary charge q\ thus, we get

l<7il
|FX|
I #2 I
1^2 I
(2-1)
for q and R both constant.

Once this procedure for comparing magnitudes has been established, one can
proceed to the study of how the force between two point charges depends on their
relative sizes. In addition, one can now assign absolute values to the charges by
choosing a charge of unit magnitude in some arbitrary but convenient way, and by
using (2-1) with the numerical value |# u n i t | = 1.
This basic experimental law refers to the situation illustrated in Figure 2-2 in which we
have two point charges q and q' separated by a distance R\ we assume the charges to
be fixed in position and that there is no other matter, that is, the charges are situated in
a vacuum. The force on q due to q' will be written as q , q . Thus, q' is being treated as
the origin of the force, and we can refer to q ' as the "source" and its location given by
r' as the "source point." Since q is the charge for which the force is to be found, we say
it is at the "field point" located by r. Then, according to (1-12), R is the relative
position vector of q with respect to q' and is seen to be directed from the source point
to the field point as is its corresponding unit vector R. Thus, using (1-12) and (1-4), we
have
R = |r - r ' |
R = r - r'
R
- =
R
,(2-2)^
In terms of all of these quantities, Coulomb's law says that

"
477*6 o R1
(2-3)
^ '
42
so that the force is proportional to the product of the charges and to the inverse square
of the distance between them; it is seen to be similar to gravitation in these respects.
The factor l/47re 0 is a constant of proportionality whose numerical value will
depend on the system of units being used; it is written in this form for later
convenience. We will be using exclusively the International System of units (SI, for
Systeme /nternational d'Unites), which is essentially the same as the MKSA system.
This means that distance is measured in meters, mass in kilograms, time in seconds,
force in newtons, energy in joules, and so on. Charge is defined in this system in terms
of electric current which is rate of flow of charge. The unit of current is called an
ampere, while the unit of charge is given the name coulomb and is defined by 1
coulomb = 1 ampere-second. We defer giving the precise definition of the ampere in
terms of the magnetic forces between currents until Section 13-2; in the meantime, we
can still take the coulomb as a known charge unit for our purposes. Other systems of
units that are used in electromagnetism are discussed in Chapter 23.
The significance of this to us is that the units of all of the physical quantities in
Coulomb's law have been already chosen so that the constant of proportionality must
be found by experiment; this necessity is analogous to that of finding the gravitational
constant appearing in the law of gravitation. The result turns out to be that
12
e 0 = 8 . 8 5 X 10
(coulomb)2/newton-(meter)2
= 8.85 X 10" 1 2 f a r a d / m e t e r
(2-4)
The constant e 0 is called the permittivity of free space, and is generally written in the
last form from which we see by comparison of both forms that 1 farad = 1
(coulomb) 2 /joule. It is also useful to note that
1
477-0
= 9 X 10 9
meter
farad
(2-5)
to an accuracy that will be sufficient for our purposes.

From (2-3), we see that if qq' > 0, so that both charges are of the same sign, then
F,
is in the direction of R, that is, the force is repulsive as is seen from Figure 2-2.
On the other hand, if qq' < 0, so that the charges have opposite signs, F ,
is in the
opposite direction to R, that is, the force on q is one of attraction toward q'. This is
often summarized by the statement that "like" charges repel one another, while
" unlike" charges attract.
We can also write Coulomb's law completely in terms of R by combining (2-2) and
(2-3) to give
(2
"6)
If we wanted the force of q on q', which we write as q ,, the only change necessary
would be to use the relative position vector of q' with respect to q, that is, R' = r ' r,
so that
q'q R'
=
"
>
We see from (2-2) that R' = R and since the magnitude of each is equal to |r r ' | ,
we see from (2-6) and (2-7) that
F,_,=
- F ,
(2-8)
which shows that the Coulomb forces are equal and opposite even though the
individual charges may differ greatly in magnitude.
2-3
SYSTEMS O F
POINT CHARGES
4 3
We are assuming a static situation, that is, the charges are at rest at fixed positions.
This means that in order for q to be in equilibrium there must be an additional
mechanical force F m on it so that the net force will be zero; in other words, we must
have
V - , + F,. = 0
(2-9)
Similar remarks apply to q'.
[2-31
SYSTEMS OF P O I N T CHARGES
Now suppose that, in addition to q, there are a number N of point charges distributed
at fixed positions throughout otherwise empty space. We designate each charge by qt
and its position vector by r, where i = 1 , 2 , . . . , N. This situation is illustrated in Figure
2-3; for clarity, the individual position vectors are not shown, but the unit vectors R,
corresponding to the relative positions of q with respect to the qt are shown. Each of
these charges can exert a force on q, F^
which will be of the general form given by
(2-3) or (2-6). The experimental facts of the superposition properties of forces are
already familiar to us from mechanics; hence F ? , the total force on q, will be given by
the vector sum of the individual forces so that
N
(=1
qq R
, = 1 49TC0 Jt-i
qq R
i = 1 477 0 Kj
where
R, = r - r,.
R, = |r - r,|
R, = ^
(2-11)
The last form in (2-10) is often more convenient to write as a starting point in solving
problems, whereas we will usually use the form expressed in terms of the unit vectors
for general discussions. Equation 2-10 expresses the fact that the total force can be
found as the sum of the individual forces between pairs that are calculated from
Coulomb's law as if the other charges were not present. Again we assume that the
individual charges are at rest, and are kept at rest, by mechanical forces of some sort as
may be required.
Figure 2-3. Calculation

of total force due t o more
than o n e charge.
44
Example
If we express all positions in rectangular coordinates, we can easily write down an

explicit form for (2-10). Using (1-13) and (1-14), a n d noting that the various charges are
designated by the subscripts i rather than by primes, we find that (2-10) becomes
y
q
Wi
^ 477
Aire0
1-1
[(* - *,-)* + (y - n)y

2
+ (z - 2,)z]
\( - Xf)\ + (y
t - y.) \ + (z
< - zj)\21j3/2
[(x
In a sense, (2-12) provides a simple recipe for solving problems, since once the values of
all the charges and their positions in rectangular coordinates are given, all that remains
is to substitute these numbers into (2-12) and to simplify the result as much as possible.
[2-4]
C O N T I N U O U S DISTRIBUTIONS OF CHARGE
We o f t e n encounter situations in which the other charges are so close together

compared to the other distances of interest that we can regard them as being
continuously distributed, much as we can treat a glass of water, on a laboratory scale,
as a continuous distribution of mass by neglecting its molecular structure. We can deal
with such a case by considering a region of the charge distribution that is so small that
the charge within it can be written as dq' a n d treated as a point charge; this is
illustrated in Figure 2-4. W e can still use (2-10), b u t now the sum will become an
integral over the complete charge distribution so that
f
q
c dq'YL
= 4 Z R J - h -
(2
-13)
where (2-2) continues to be applicable.

If the charges are distributed throughout a volume, we can introduce a volume
charge density p, which is defined as the charge per unit volume and hence will be
measured in coulombs/(meter) 3 . (We will write this charge density as pch in the
infrequent cases in which it might be confused with the p of cylindrical coordinates.)
Then the charge contained in a small source volume dr' will be given by
dq' = p ( r ' ) dr'
(2-14)
2-4 C O N T I N U O U S DISTRIBUTIONS OF CHARGE
45
as shown in Figure 2-5a, and (2-13) will become

Fq =
47re0 Jy
p(r')R^T'
We have written p = p(r') because, in general, the volume density can vary with the
location of the source point; the integral in (2-15) is to be taken over the total volume
V containing the charge distribution.
W h e n we say that d r ' is a " s m a l l " volume, we mean that it is small on a
macroscopic, laboratory scale. On the other hand, it must b e large on a microscopic,
atomic scale so that it will contain many atoms a n d / o r molecules. Only in this way can
we treat p as a continuously varying function of position. If dr' were m a d e comparable to, or smaller than, atomic sizes, then at most locations dr' would contain no
charge a n d p would be practically always zero. T h e charge density would be different
f r o m zero only when d T ' included an electronic or nuclear charge, but this would lead
to such widely fluctuating values for p that it would no longer be a useful concept.
Similarly, the charges can often be idealized to lie on a surface or along a line. W e
introduce analogous charge densities: the surface charge density a defined as the charge
per unit area, and the linear charge density X defined as the charge per unit length; they
will have units of coulomb/(meter) 2 and c o u l o m b / m e t e r , respectively, and can also
vary with position in general. From these definitions, we get
dq' = a(r') da'
or
dq' = X(r') ds'
(2-16)
as indicated in Figure 2-5b and c. For such cases, (2-13) will become
q
a ( r ' ) R da'
A ( r ' ) R ds'
F,9 = T ~ f
Aut.0 Js'
F,q = y
f
Awt0 Ju
0*2
R
(2-17)
(2-18)
where (2-17) is to be integrated over the total surface S' on which there is a surface
distribution while (2-18) covers the whole line L' occupied by a linear distribution of
charge.
Finally, if all the possibilities we have discussed are simultaneously present, the total
force on q would be obtained from the sum of all forces due to the various charge
distributions, that is,
Fq = F ( / (from points) + F^(from volumes)
+ F ? (from surfaces) + F (? (from lines)
(2-19)
A very important word of warning: you can avoid a lot of trouble, wasted time, and
wrong answers by remembering, and following, these two simple, although^ almost
trivial sounding rules(1) always draw the relative position vector, and hence R, from
the source point to the field point; (2) never write the location of a source point as r or
dq' ~ pdr'
dq' = oda'
dq' = \ds'
IT7]
Hdr'
(a)
(b)
(c)
Figure 2-5. Definition

of various charge densities.
46
(x, y, z) and so on, but instead use r', or (x', y', z') or some sort of label, such as was
done in (2-10) and (2-11).
P O I N T CHARGE O U T S I D E A U N I F O R M SPHERICAL CHARGE DISTRIBUTION
As an example of the effect of a continuous charge distribution, we will evaluate (2-15)

for a case in which q is located outside a sphere containing a uniform distribution of
charge, that is, for which p = const. We choose the origin at the center of the sphere of
radius a and let q be on the z axis so that z > a; the situation is shown in Figure 2-6 in
which only one octant of the sphere is pictured. We use spherical coordinates to
describe the source point r' and for doing the integration. Figure 2-7 shows the plane
containing the z axis, r', and R. We see that r = zz and r' = r'r'*from (1-11) and
(1-97), and therefore R = zz r'r' according to (2-2); it then follows f r o m (1-17),
(1-19), (1-92), (1-15), and Figure 2-7 that
R2 = z2 -f r'2 2zr'z r' = z2 + r'2 2zr'cos8'
We see, in fact, that this value of R2 is exactly that given by the law of cosines as
applied to Figure 2-7. If we now obtain R f r o m these results, we find that (2-15)
becomes
(2-20)
where p has been removed from under the integral sign because it is constant.
Figure 2-6. Point charge

outside a uniform spherical charge distribution.
2-5
P O I N T C H A R G E O U T S I D E A U N I F O R M SPHERICAL C H A R G E D I S T R I B U T I O N
47
Because f ' is not constant during the integration, it will be convenient to find
in
terms of its rectangular components. Following (1-21), we dot each side of (2-20) with
z, and use (1-19), (1-93), and (1-18); then, after writing dr' in the form (1-99), we get
qp
qz
r'cosd')r'2
[2* fir f{z
sind'dr'dd'
dq/
( z 2 + r'2 2z/*'cos 0 ' ) 3 / 2
4ttc 0 Jq Jo *4
The integration over dcp' can be performed at once and gives 2m. We do the integration
over dd' next. For this purpose, it is convenient to introduce a new variable ju, = cos 6'.
Then du = sin0' dd', and, if / = / ( c o s d') is a function of cos 0', we make the
indicated substitutions and get the general and useful result that
= Jf1 /(ju) d\s.
-i
f / ( c o s 6') sind'dd'
o
(2-22)
When this is done, and the 2 m from the integral over dcp' is included, (2-21) becomes
*
s l
i % *
d r
2t0J0
J-i
( 2
2 3 )
( z 2 + r'2 2zr'n)
The integral over jti can be found from tables to be

1 ( z - r'
(z/x - r')
z 2 ( z 2 + r'2 - 2zr'fi)1/2
-l
I2 \ |z - r'l
t
+
z + r'
|z + r'\
(2-24)
where we have written the terms [(z r')2]1/2, which appear, as \z r'\ in order to
emphasize that we must be sure that we get a positive value for the square root. In our
case, we have assumed that q is outside the sphere so that z > a; since r' < a, we will
always have z > r' so that |z r'\ = z r'. N o w |z + r'\ = z + r' since we have
taken z to be positive and r' always is. When these are substituted into (2-24), the
integral over ju, in (2-23) is seen to be just 2 / z 2 , which is constant as far as the
integration over r' is concerned, so that (2-23) becomes
"
QP ra 2,
r' dr'
o2
VP**3
2
3t
oz
(2-25)
48
Before we discuss this result, let us find the remaining components.

If we dot (2-20) with x, and use (1-93), we see that the resulting integrand will be
proportional to cos <p' and therefore
qX = * ' Fq ~ f^cos 'P' d(P' = 0
Similarly,
F
qy
, ~ J ^
S i n (
P '
<P'
The fact that these two components vanish is a consequence of the "symmetry" of the
situation as we can see with the use of Figure 2-8. The charge element contained in dr'
will produce its contribution d F' to the total force and this contribution will have a
horizontal component. Corresponding to dr', however, is another volume element dr"
that is the reflection of dr' in the line r, and hence is at the same distance R from q.
The equal charge contained in dr" will produce a contribution dF" to the total force.
Since d" and dF' have the same magnitude, we see that the horizontal component of
d" will be equal and opposite to that of dF'. Thus, when the contributions of this
pair are added, the horizontal forces will cancel, although the vertical ones will not.
Since all charge elements in the sphere can be paired off in this way, the total force will
have n o net horizontal components F and Fqy while Fqz will not vanish, as we found
above. As we will see, symmetry considerations such as these can often simplify our
work and one should try to be aware of and look for them.
Since only the z component is different from zero, the total force will be in the z
direction and
=
qpa3 z
( 2
"
2 6 )
from (2-25) and (1-5). We see that if q > 0 and p > 0, then F ? is directed away from
the sphere as expected since q would be repelled by all of the positive charges;
similarly, if p < 0,
is directed toward the sphere, that is, the force on q is attractive.
We can write (2-26) in an interesting and instructive form if we express it in terms of
EXERCISES
49
t h e t o t a l c h a r g e Q' c o n t a i n e d w i t h i n t h e sphere. F r o m (2-14), w e get

Q' = f dq' = (p dr'
= p f
dr'
= ^Tra^p
(2-27)
sphere
since p is c o n s t a n t ; w h e n this is used t o e l i m i n a t e p in (2-26), w e find t h a t

F
qQ'z
, = T^-T
9
(2-28)
47TC QZ
W e see f r o m F i g u r e 2-7 t h a t z is t h e d i s t a n c e f r o m t h e c e n t e r of t h e s p h e r e to q, so
t h a t , o n c o m p a r i n g (2-28) w i t h (2-3), w e find t h a t this u n i f o r m s p h e r e of c h a r g e acts as
if it w e r e a single p o i n t c h a r g e l o c a t e d a t t h e c e n t e r of t h e s p h e r e as f a r as its effect o n a
c h a r g e l o c a t e d o u t s i d e the s p h e r e is c o n c e r n e d . A s w e will see later, this is n o t the case
if t h e p o i n t of interest is l o c a t e d i n s i d e t h e s p h e r e .
A c t u a l l y , o u r result is even m o r e g e n e r a l t h a n it first a p p e a r s . T h e l o c a t i o n of q w a s
t a k e n t o b e a l o n g t h e z axis f o r c o n v e n i e n c e in e v a l u a t i n g t h e integral. A s we previously
f o u n d f r o m F i g u r e 2-7, t h e p o s i t i o n v e c t o r of q w i t h respect t o the c e n t e r of t h e s p h e r e
is r = z z so that |r| = r = z a n d r = z, w h i c h e n a b l e s u s t o r e w r i t e (2-28) f o r a n y
l o c a t i o n of q in t e r m s of its spherical c o o r d i n a t e s as
F
qQ'r
, = TT7T
I
4ire0r:
(2-29)
[This also follows f r o m (1-90) a n d (1-93) since t h e l o c a t i o n of q in F i g u r e 2-7

c o r r e s p o n d s t o the special case 6 = 0.]
IEXERCISESI
2-1 Two point charges q' and q' are on the x
axis with coordinates a and a, respectively.
Find the total force on a point charge q located
at an arbitrary point in the xy plane.
2-2 Four equal point charges q' are located at
the corners of a square of side a. The square lies
in the yz plane with one corner at the origin a n d '
its sides parallel to the positive axes. Another
point charge q is on the x axis at a distance b
from the origin. Find the total force on q.
2-3 Eight equal point charges q are located at
the corners of a cube of edge a, which has the
location and orientation of the figure shown in
Figure 1-41. Find the total force on the charge at
the origin.
2-4 Repeat the calculation of Section 2-5 for the
case in which q is outside the sphere but below it,
that is, z is negative and \z\ > a. Show that your
result is consistent with (2-26) and (2-29).
2-5 Repeat the calculation of Section 2-5 for the
case in which q is inside the sphere (z < a) to
show that F = (<7pz/3c 0 )z.
2-6 A sphere of radius a contains charge distributed with constant volume density p. Its center
is on the z axis at a distance b from the origin
with b > a. A point charge q is located on the y
axis at a distance c from the origin where c > b.
Find the force on q.
2-7 A line charge of length L with A = const,
lies along the positive z axis with its ends located
at z = z 0 and z 0 + L. Find the total force on this
line charge due to a uniform spherical charge
distribution with center at the origin and radius
a < Zq .
2-8 The surface of a sphere of radius a is charged
with a constant surface density a. What is the
total charge Q' on the sphere? Find the force
produced by this charge distribution on a point
charge q located on the z axis for z > a and for
z < a.
2-9 Two line charges of the same length L are
parallel to each other and located in the xy plane
as shown in Figure 2-9. They each have the same
50
Figure 2-9.
2-9.
The t w o line charges of Exercise
linear charge density X = const. Find the total

force on II due to I.
2-10 The line charge I in Figure 2-9 now has a
linear charge density X = Ay2 where A is a constant. What are the units of At What is the total
charge on I? Find the total force due to I on a
point charge q placed on the x axis at x = a.
2-11 Charge is distributed over the surface of a
circle of radius a lying in the xy plane with
origin at the center. The surface density is given
in cylindrical coordinates by a Ap2 where A is
a constant. What are the units of A1 What is the
total charge on the circle? Find the force produced by this charge distribution on a point charge
located on the z axis.
3
THE ELECTRIC FIELD
Coulomb's law is an example of what is known as an "action at a distance" law. It
provides us with a straightforward way of calculating the force on a given charge when
the relative position with respect to the source charge is known. Coulomb's law does
not purport to describe how the first charge " k n o w s " the other one is present. If, for
example, the position of the source charge is changed, the force on the first charge will
also be changed and again given by Coulomb's law. The implication is that this change
will occur instantaneously, but again there is no suggestion as to how this altered state
of affairs is brought about. As a result of these and similar considerations, it has been
found convenient and useful to make a mental division of the interaction between the
two charges into two aspects: first, one assumes that the source charge produces
"something" at the field point, and, second, that this "something" then interacts with
the charge at the field point to produce the resultant force on it. This "something,"
which acts as a kind of intermediary between the two charges, is called the electric field
and is what we discuss now.
m \
DEFINITION OF THE ELECTRIC FIELD
If we look again at (2-10), we see that q is a common factor of all of the terms so that
can be written as a product of q and a quantity that is independent of q but does
depend on the values of all of the other charges and their locations with respect to q.
This quantity is called the electric field E; thus, we write (2-10) in the form
F, = <7E
(3-1)
where
(3-2)
Equation 3-1 provides us with the definition of the electric field and we see that we can
interpret it as a quantity such that, when it is multiplied by a point charge, the result is
the force on the point charge. It also follows from (3-1) that E will be measured in
newtons/coulomb. Equation 3-2 then provides us with a prescription for calculating E
at the location r (the "field point") for a given distribution of point charges; we are still
using (2-11) of course. We note that q is not included among the source charges in
(3-2), that is, we do not envision a charge exerting a force on itself.
If the source charges have a continuous distribution, we can combine (3-1) with our
previous results (2-15), (2-17), and (2-18) to get corresponding expressions for E:
(3-3)
(3-4)
(3-5)
51
52
THE ELECTRIC FIELD
If all locations are given in rectangular coordinates, we get an explicit expression for E
from (2-12):
N
li
[(* - *,)* + ( y - y,)$ + (z - zt)z]
47r \(
\2 . /
x2
,
,213/2
, = 1 4 7 r 0 [ ( * - x,) + (y - y,) + (z - Z,) j
^ ^
(3"6)
Finally, if all of the possibilities we have discussed are simultaneously present, we

see f r o m (3-1) and (2-19) that the total E at a given point will be obtained as the vector
sum of the contributions from all of the various charge distributions producing the
field.
If the charge distribution is simple enough, E can be easily calculated by a
straightforward integration. We look at two such examples; one for a linear distribution
of charge, and the other for a surface distribution. A f t e r we have done this, we discuss
in more detail the significance of what we have done.
FIELD OF A UNIFORM INFINITE LINE CHARGE
We assume that A = const, and choose the z axis to coincide

distribution as shown in Figure 3-1. We choose the origin so that the
lie in the xy plane for convenience; then we have r = pp and
R = pp z'z and R2 p 2 + z' 2 . We also see f r o m the figure that
case, a n d therefore (3-5) becomes
E =
A
477-e,
/*oo (pp
/.
z'z) dz'
3/2
(p2 + z'2)
App
/:
47re0 J-<
with the charge

field point P will
r' = z'z so that
ds' = dz' in this
dz'
,2^/2
(P 2 + z'2)
(3-7)
The last form is obtained because the z component of the integral vanishes since the
integrand is an odd function of z' and p is a constant with respect to the integration
variable z'. The integral in (3-7) is found to be
J
2 1/2
p (p* + 2 ' )
1[
(r
(-1)]
~~2
(3-8)
so that the final result is

E =
2 77"0P
(3-9)
3-3
FIELD O F A U N I F O R M
INFINITE PLANE SHEET
5 3
Figure 3-2. A cylinder of constant electric field magnitude

for a uniform infinite line charge.
Thus the electric field has only a radial component. It is directed away from the line
charge if X > 0, as it should be since a positive charge q would be repelled, while it is
directed radially inward if X is negative. The magnitude of E varies inversely with the
distance p from the line charge.
Since (3-9) is independent of the angle <p, we see that a surface of constant
magnitude of E will be a cylinder of radius p with the line charge as the axis of the
cylinder. A portion of this cylinder is shown in Figure 3-2 in which are also indicated
some directions of E for X > 0 on a circle formed by the intersection of a plane
perpendicular to the z axis and the cylinder.
3-3
FIELD O F A U N I F O R M INFINITE PLANE SHEET
We assume the surface charge density a to be constant on an infinite plane which we

take to be the xy plane. It is convenient to choose the z axis to pass through the field
point P. We will use rectangular coordinates for integration. We see then from Figure
3-3 that r = zz and r ' = x'x + y'y. Since the element of area is da' = dx' dy', (3-4)
becomes, with the use of (2-2), (1-13), and (1-14):
E =
f<x>
roo
( x ' x y ' y + z z ) dx'
(x'2 + y'2 + z2)
J - o o ' - oo
1/?
dy'
(3-10)
We see at once that Ex = Ey = 0 since the x and y terms in the integrand are odd
functions of x' and y', respectively. Therefore, (3-10) reduces to
E =
The integral over y'
oz
/-oo
/
zdx'
/>oo
4^0^00
is identical in form
dy'
oo (.x' 2 + y'2
"f- z2)
(3-11)
to that in (3-7) and from (3-8) equals
54
T H E E L E C T R I C FIELD
Figure 3-3. Calculation of t h e field due

t o a uniform infinite plane sheet.
2/(x'2
+ z2); (3-11) thus becomes

CTZ
fOO
2 ? T C q jJ - m
cc
zdx'
x'
(3-12)
where the plus sign is to be used for z > 0, while the minus sign is used when z < 0. It
is often convenient to write (3-12) as
(3-13)
which automatically gives the correct signs for E.

We see from (3-12) that E is always directed away from a positively charged plane
(a > 0), and is always directed toward the plane when a < 0. These directions
correspond to those of the force on a positive charge q placed at some point. It is
interesting to note that the magnitude of E is independent of position, that is, E has the
same value no matter how near one is to the plane or how far away one is from it; this
arises essentially from the fact that no matter where the field point is located, there is
always an infinite amount of charge " visible" to it. These properties of E are indicated
in Figure 3-4 which is drawn for a > 0, and shows an edge on view of the charged
a > 0
Figure 3-4. Electric field d u e to a uniform infinite

plane sheet.
3-4
W H A T D O E S ALL O F T H I S M E A N ?
5 5
plane. The dashed lines are traces of planes above and below the charged plane and
parallel to it. The figure looks the same if it is turned upside down, as it should since
our original choice of direction for positive z was completely arbitrary. Similarly, the
appearance is the same if one views it looking from behind the page rather than toward
the page. In other words, our result (3-12) is completely consistent with the basic
"symmetry" of the source charge distribution. E also changes discontinuously in
direction as one passes through the charged plane; if one passes from below to above,
for example, the total change is
(above) E z (below) = a / e 0 as found from (3-12).
W H A T DOES ALL OF THIS MEAN?
We managed rather easily to introduce an auxiliary quantity that enabled us to divide

the interaction between two charges into conceptually different parts. We did this by
defining a new vector field E and we have ways that, in principle, allow us to find it at
any point once the charges that are its sources are given. It is natural, however, to ask if
there is really anything useful to be gained in doing this.
One can quite easily adopt the point of view that this is done merely for mathematical convenience, if for no other reason than one saves some writing by not having to
carry the symbol q along in all of the equations, but can calculate E first and then
insert q as a last step by means of (3-1). We can thus regard the calculation of E as
merely providing us with a sort of contingency statement distributed throughout space
in the sense that E(r), combined with (3-1), tells us what would happen if we were to
put a point charge q at r.
On the other hand, our formulas (3-2) through (3-6) enable us to calculate the
electric field at r whether or not there is a charge there to be subject to a force. This fact
provides us with a strong temptation to make a conceptual leap and regard E as an
actual physical entity in its own right. Most of these ideas originated with Faraday, and
he felt that the presence of charges actually changed the physical properties of space,
and that E was a manifestation of this altered state. For him, the electric field was a
very real physical quantity.
If we adopt the attitude that E is a physical quantity, the question then naturally
arises as to how one would measure it. At first glance, this appears to be very simple:
one merely puts a point charge q at rest at the point r of interest, measures the force
on it, and then (3-1) tells us that E(r) = F Q / q . A possible problem now arises with the
recognition that the presence of q now subjects the source charges qi of (3-2) to new
forces as given by (2-8) and they will no longer be in equilibrium, although eventually
equilibrium will be reestablished. In the idealized case in which we can assume the qt to
be rigidly attached to fixed positions, the new electrical force can be compensated by a
new mechanical force that is produced without deformation of the support; then (2-9)
as applied to a given qt will still hold, but the positions of the ql will not have changed.
The value of E as given by (3-2) will be exactly what it was before q was introduced
and E will be correctly given by (3-1). In a real case, however, the new values of the
mechanical forces required for equilibrium can generally only be obtained by deformation of the support, such as by the bending of a rod or by a stretching or compression
of a spring; thus, when the new equilibrium configuration is attained, the positions of
the q{ will have been changed, the value of (3-2) will generally be different, and the net
result will be that the very act needed to measure the preexisting field has altered it. (In
addition, as we will see later, if the source charges are associated with conductors, they
will ordinarily need to move about on the conductor in order to come into mutual
equilibrium, again leading us to the conclusion that the field can be different.) This
problem of changing what you are trying to measure is not unique to electromagnetism,
56
THE ELECTRIC FIELD
of course, and one usually tries to solve it in the same general way, that is, by
minimizing the disturbance as much as possible, while still being able to get a
measurable effect. In order to apply this idea to E, one imagines the charge on which
the force is to be measured to be very small and then try to go to the limit in which it
approaches zero; if we use Sq to denote this "test charge," and 5 F the measured force
on it, then we would require that E be determined by
SF
E(r) =
lim 8q>0 bq
(3-14)
Although the use of the electric field may be regarded as a convenient artifice if one
is dealing solely with electrostatics, when one comes to handling other problems,
particularly time dependent ones, such as that briefly alluded to in the introductory
paragraph of this chapter, it has been found to be virtually impossible to do this
without an extensive use of vector fields. We find it useful to define several other vector
fields as we proceed, and whether we want to regard them as real physical quantities or
not, we will certainly treat them as if they were. We want to study their properties
extensively as well as applications of them; among these properties are their differential
source equations, that is, their divergence and curl. We already know that the sources of
the electric field are charges of any type, but we want to restate this in the form of
explicit expressions for v E and V X E; we obtain them, along with other information, in the next two chapters.
EXERCISES
3-1 Two point charges q and - q are located on
the y axis at y = a and -a, respectively. Find E
for any point in the xy plane. For what points, if
any, will Ex = 0?
3-2 Four point charges are located at the corners
of a square in the xy plane. Their values and
locations are as follows: <7,(0,0); 2q,(0,a)\
3q,(a,0)\
- 4q,(a, a). Find E at the center of
the square.
3-3 Consider a cube
and orientation of the
is a point charge q at
at ( a , a,0). Find E at
(a, b, 0). Find the rectangular components of E

produced at the point (0, c,0).
3-8 T w o infinite plane sheets with equal constant surface charge density a are parallel to the
xy plane and located as shown in Figure 3-5.
Find E for all values of z.
of edge a with the location

figure in Figure 1-41. There
each corner except for that
the empty corner.
3-4 Repeat the calculation of Section 3-2 for a

general field point r = pp + zi and thus show
that the same result is obtained. Is this physically
reasonable?
3-5 Repeat the calculation of Section 3-3 for a
general field point (x,y,z)
and thus show that
the same result is obtained.
3-6 Repeat the calculation of Section 3-3 using
cylindrical coordinates for the source point.
3-7 A uniform infinite line charge is parallel to
the z axis and intersects the xy plane at the point
Figure 3-5. The t w o infinite plane sheets of

Exercise 3-8.
3-9 T w o infinite plane sheets have equal and
opposite constant surface charge density a. They
are parallel to the xy plane and located as shown
in Figure 3-6. Find E for all values of z.
3-10 The circular arc of radius a shown in Figure 3-7 lies in the xy plane and has a constant
linear charge density X and center of curvature at
the origin. Find E at an arbitrary point on the z
EXERCISES
57
P(p,
Ip, z)
Figure 3-6. The t w o infinite plane sheets of

Exercise 3-9.
/
- 7 .
'
Figure 3-8. The finite line charge of

Exercise 3-11.
The disc lies in the xy plane with center at the
origin. Show that the electric field at a point on
the z axis is given by
E = z
Figure 3-7. The circular arc of charge of
Exercise 3-10.
axis. Show that when the curve is a complete
circle your answer becomes
E =
Xazz
2 e 0 O 2 + z2)
3-11 Charge is distributed with constant linear

charge density X on the line of finite length
shown in Figure 3-8. Find E at P. With the aid of
the distances R2 and R1, express E in terms of
the angles a 2 and a x shown. Find E for the
special case for which L2 = Lx = L and P is in
the xy plane.
3-12 Charge is distributed with constant surface
charge density a on a circular disc of radius a.
2e r
(a2
+ z
2\l/2
)
(3-15)
What does this become as a -* oo?

3-13 An infinitely long cylinder has its axis
coinciding with the z axis. It has a circular cross
section of radius a and contains a charge of
constant volume density pch. Find E at all points,
both inside and outside the cylinder. Hints: use
cylindrical coordinates for integration; for convenience, choose the field point on the x axis
(Will this be general enough?); you will probably
need this definite integral
o A
(A B cos t) dt
2AB cos t + B
if A2 > B2
if A2 < B2
(3-16)
4
GAUSS' LAW
We can anticipate from the general definition of the divergence given in (1-66) that the
flux of E through a closed surface, that is, its surface integral, will be worth investigating. As we will see, the inverse square nature of Coulomb's law makes it possible to
evaluate this flux over a surface of arbitrary size and shape.
DERIVATION OF GAUSS' LAW
We want to show that

(e da =
I
C
?, =
e
0 inside
(4-1)
where Qin is the net charge contained within the volume enclosed by the arbitrary
closed surface S. Now from (3-2) we quickly obtain
r
Se
<
Ts
1-
4770
r R*
SS
R,
da
There are two cases to consider.

1. qt is inside S (Figure 4-1). If we recall our general discussion in connection with
Figure 1-26, we see that
R, J a
' ,
R2
da c o s ^
R2
area
~ =
_L t o R ,1
-=
R2
= dQ
(4-3)
where dQ, = element of solid angle subtended at q, by the area da. In order to evaluate
the integral in (4-2), we consider a sphere S0 of radius R0 with qt as the center. This
same solid angle dQ, will intercept the area da0 on this sphere; as we see f r o m the
figure, d a 0 is parallel to R, so that if we apply (4-3) to this case, dSl also equals
da0/R2Q. Therefore, the integral in (4-2) can equally well be written as
da0
j ) d S l = ( f ) Y =
J
s0 R o
t
477'R2q
d a 0 = ^ - = 47T
So
Ko
(4-4)
since R0 is constant for all points on the surface of the sphere. Thus, the total solid
angle subtended by any surface about a point within it is 47r, and we can write
R, da
2 = 477Ri
(if r, is inside S)
(4-5)
2. qx is outside S (Figure 4-2). Here we consider the two elements of area da1 and
da 2 of S that are cut out by the same solid angle dU but that are on opposite sides of
S. Their distances from qt will be written as Ra and Rn. As before, we will have
R,dax
R)i
58
dax cos %
4-1
but now, since
DERIVATION OF GAUSS' LAW
5 9
*s negative,
> 77-/2, so that cos

R, c/a,
da9 cos Sis
Rh
*ti
= -dto
and therefore
R, da,
^2
Ka
R, d a 7
(4
*6)
Ki2
so that the net contribution of these two elements of area to the integral in (4-2) is zero.
Since all of the elements of area of S can be paired off in this way, all of their
contributions to the integral will mutually cancel, and therefore
R,, da
(J) = 0
Ts
R-i
(if r, is outside S)
(4-7)
Hence, from (4-2), (4-5), and (4-7),

r
$ E '
S
d a
z=r
4T7Cq
^
/. R, <ia
1
qtf
ni
=c
inside S Rj
0
inside
0,n
h =
(4-g)
60
GAUSS' LAW
and we have proved Gauss' law as stated in (4-1), and which is now seen to be a direct
consequence of the inverse square law of force between point charges.
We can note here some of the interesting implications of this result. Any charges
outside the surface do not affect the value of the integral, although their values and
locations can clearly affect the particular value of E at each point on the surface.
Similarly, the integral depends only on the total value of the charges inside the surface
and thus is independent of their specific locations inside; however, if they are placed at
new locations, the value of E at a particular point on the surface can be expected to be
changed, but the value of the complete integral will again be unaffected. Since the result
given in (4-1) is a simple sum, we see that each charge contributes independently to the
total flux of E through S; therefore, a given point charge q has a total flux of E equal
to q/(. 0 through any closed surface surrounding it.
If we now assume the charges within S to be continuously distributed with density
p, we can use (2-14) and write
<2i = / P d r
JV
where V is the total volume enclosed by S. If we also apply the divergence theorem
(1-59), we can write (4-8) as
(4-9)
Since this result holds for any arbitrary volume V, it will be true for an infinitesimal
one, and we can equate the integrands to give
(4-10)
This important result is one of Maxwell's equations and is equivalent to Coulomb's law
of force between point charges. (Later, we will find it important to recall that the charge
density p includes all charges from any source whatsoever since the electric field is
produced, by its definition, by all charges.)
fh2l
S O M E APPLICATIONS O F G A U S S ' LAW
If the charge distribution has sufficient symmetry, Gauss' law provides a simple and
convenient way of calculating the electric field. The principal problem is that of
choosing a suitable closed surface of integration. The main things to look for are
surfaces on which E has a constant magnitude, and surfaces for which E is either
parallel to or perpendicular to the surface both for ease in integration and for avoiding
difficulties with an unknown dependence of E on position. This process is most easily
demonstrated with examples, and we will look at a sample of each of our types of
continuous charge distribution.
Example
Uniform infinite line charge. We assume A = const, and again let the line charge
coincide with the z axis to be specific; we will use cylindrical coordinates. For an
infinite line, it doesn't make any difference where you are located along it since the
charge still extends to infinity in both directions; hence we conclude that E cannot
depend on z. Similarly, there is nothing to distinguish one value of <p from another
since the charge distribution looks the same no matter from where you view it
4-2 SOME APPLICATIONS O F G A U S S ' L A W
61
perpendicular to z. Therefore, E must also be independent of <p and, at most, can only
depend on the distance p from the line; thus we conclude that E = E(p).
N o w suppose that E had a component parallel to the line, that is, Ez(p) # 0. In this
example, the choice of positive or negative direction for the z axis is completely
arbitrary, that is, there is no real distinction between " u p " and "down." But if Ez were
different from zero, this fact would distinguish up from down; hence we must conclude
that Ez = 0. Similarly, there is no reason to prefer the sense of increasing <p over the
sense of decreasing <p, so that E^ = 0. Therefore, we conclude from the symmetry of
the situation that E can only be radial and that we can write E = Ep{p)p. Thus, a
surface of constant magnitude of E is an infinitely long cylinder of radius p whose axis
coincides with the line charge; a finite portion of this cylinder chosen to be of length L
is shown in Figure 4-3, and we note that its outward normal n c is exactly p, and
therefore parallel to E. All of this suggests that we use the curved surface of this
cylinder as part of our surface of integration. We can now get a closed surface of
integration by choosing the rest of it to consist of the two circular cross sections of
radius p shown, thus making our final surface of integration a right circular cylinder.
The outer normals to these upper and lower surfaces are shown as n u and n, and are
seen to equal z and z, respectively. Although Ep has an unknown dependence on p
on these circular areas, E is perpendicular to the vector areas and hence its contribution
to the flux will vanish.
After this lengthy discussion, the actual evaluation of the surface integral is almost
anticlimactic. We can write it as the sum of integrals over the curved surface, and the
upper and lower faces (which we designate as c, u, and /). We also remember that
Ep(p) is constant on the curved surface since p is constant. Therefore, (4-1) becomes, in
Figure 4-3. Calculation of t h e field

of a uniform infinite line charge.
62
G A U S S ' LAW
this case,
<e da = Jep(p)p
= Ep(p)
hc da + J Epp huda
+ jEpp
t
I da + 0 + 0 = Ep(p)2-7TpL =
n, da
Gin
^
C
where we used (1-76) and (2-16), the latter to find the total charge inside the cylinder
which is that on the line of length L. We see that when we solve the above for Ep, the
arbitrary length L cancels and we get Ep(p) = X/27rpt 0 so that
E = p
2 7re0p
(4-11)
which is exactly the result (3-9) that we previously obtained by direct integration.
Example
Uniform infinite plane sheet. Here we assume that a = const., and, for definiteness, that
the sheet coincides with the xy plane. A n edge on view of the situation is shown in
Figure 4-4. Since the choice of the origin as well as of the orientation of the x and y
axes is completely arbitrary in this example, E must be independent of x and y.
Similarly, there is no basic distinction here between left and right, or into or out of the
paper, so that E cannot have any components parallel to the sheet; thus E has only a z
component which, at most, can be a function of z, the distance f r o m the sheet. There is
also no real difference here between up and down, so that E must always point away
E
n
) \
J*
z
)
r
n
D<
Q S
D <
k
AaS
/
Nf
/
1
\
Figure 4-4.
Calculation of t h e field of a uniform infinite plane sheet.
4-2 SOME APPLICATIONS O F G A U S S ' L A W
63
from the sheet or always toward it depending on the sign of a. Therefore, E must have
the form E = E(z)z with the upper sign for z > 0 and the lower for z < 0; E(z) can
be positive or negative, the figure being drawn for the case in which E(z) is positive.
These considerations suggest choosing the closed surface of integration to be that of
a right cylinder extending an equal distance D above and below the sheet, with faces
parallel to the sheet and of area A a; the outward normals are shown as n = z and
H/ = z in the figure. E has the constant magnitude E(D) on these faces. The curved
surface connecting these faces will have its sides perpendicular to the sheet; one of its
outward normals is shown as n c and we see that n c z = 0 for all parts of the curved
surface. We also see that Qin is the charge intercepted on the sheet by the cross section
of the cylinder and, hence, equals a Aa.
Therefore, in this case, (4-1) becomes:
(^)E
da
= JE{D)Z
da +
J E ( D ) ( Z ) ( Z)
da +
JE(Z)Z
hcda
Qin
a Aa
Aa + 0 = 2 E ( D ) Aa = =
e
o
F r o m this we find that E(D) = a/2e0 and has turned out to be actually independent
of D\ therefore, the electric field for this sheet is given by
= E(D)
Aa + E(D)
E = tz
2<f0
(4-12)
which is exactly the result (3-12) that we obtained by direct integration.
Example
Spherically symmetric spherical charge distribution. The charge distribution is contained

within a sphere of radius a. We do not necessarily assume the charge density to be
constant, but we do assume it to be independent of angle so that at most p p(r); this
is what "spherically symmetric" means. F r o m this we can conclude that E itself is
radial with a magnitude independent of angle and therefore can be written in the form
E = Er(r)f.
Since the magnitude of E is constant on a sphere of radius r, we choose such a
sphere as our surface of integration; its outward normal is also f. Therefore, the left
side of (4-1) becomes, with the use of (1-92):
(e da = ( f > E r ( r ) r f da = Er(r)(
da = 4irr2Er(r)
(4-13)
and therefore
E
1 4
4tTt 0 r 2
p(r')
(4-15)
where
Qin=
Y(r)
dr'
and where V(r) is the volume of the sphere of radius r. There are two cases to consider.
1. Outside the sphere of charge, r > a. Here p ( r ' ) = 0 if r' > a, and the volume of
integration in (4-15) reduces to V(a), the total volume of the charge distribution, so
that
p{r')dT'=Q
0in = /
V(a)
(4-16)
64
GAUSS' LAW
where Q is the total charge contained in the sphere. Then (4-14) becomes
E , ( r ) -
(r>a)
(4-17)
and the electric field outside is the same as if all of the charge were a point charge
located at the center of the sphere. We recall f r o m (3-1) and (2-29) (where the total
charge was written Q') that this is what we f o u n d by direct integration for a sphere of
constant charge density, but now we see that this conclusion is generally true for any
sphere of charge as long as the charge distribution is spherically symmetric.
2. Inside the sphere of charge, r < a. In this case, we can use (1-99) to write (4-15) as
Qin = f2" [ f p(r')r'2
'o 'o '0
sin 6' dr' d6' dq>' = 477 f p(r')r'2dr'

J
(4-18)
which, when substituted in (4-14), gives the field inside as

Er(r)
/>r
? / p(r')r'2dr'
*o r Jo
(r < a)
(4-19)
We cannot proceed any further until we know the explicit f o r m of p(/')
Example
As a special case of the previous example, we assume p = const., that is, we have a
uniformly charged sphere. Then the integral in (4-19) becomes
f pr'2dr'
o
= p Jf r'2 dr' = \pr3

o
so that
Er(r)
(r < a)
(4-20)
We can express this in terms of the total charge Q by using (2-27); the result is
Er(r)--^,
4tre 0 a
(r<a)
(4-21)
which shows that the field increases linearly with distance as we proceed out f r o m the
zero value at the center; outside, it varies inversely with the square of the distance f r o m
the center according to (4-17). We note that both (4-17) and (4-21) give the same value
Figure 4-5. Field of a uniform sphere

of charge of radius a as a function of
distance r from the center.
4-3
DIRECT CALCULATION O F
6 5
Q/Aire.0a2 at the surface of the sphere of charge where r = a\ thus the electric field is
continuous across the surface. These results for the uniformly charged sphere are shown
in Figure 4-5.
Other spherically symmetric charge distributions p ( r ) will give a different dependence of Er on r inside the sphere, but exactly the same as shown in Figure 4-5 outside
the sphere when expressed in terms of the total charge contained within the sphere
r = a.
u
DIRECT CALCULATION OF V ' E
Although we have succeeded in our aim of finding the divergence of E as given by

(4-10), it was obtained somewhat indirectly, and it is of some interest to see how this
same result can be obtained by a direct approach which starts from our basic defining
equation for E. We find it convenient to start with the expression (3-3) appropriate for
a continuous charge distribution. If we dot both sides of (3-3) with v , we get
(4-22)
Now the derivatives in V are all taken with respect to the field point coordinates x, y,
and z, which is where we want to know the value of v E(r). The definite integral in
(4-22) is taken over a specific source volume V; the limits of integration will be either
constants or, at worst, functions of the source point variables, that is, primed variables.
(In the example of Section 1-13, we have already seen a two dimensional case in which
the limits involved the variables of integration.) Therefore, the only dependence of the
integral on r can be in the integrand, and, since the definite integral can be regarded as
the limit of a sum, we can use (1-114) as generalized to a sum of more than two terms
and interchange the order of differentiation and integration. If we then use (1-115) and
note that p(r') and dr' are constants as far as V is concerned since they involve only
primed source point coordinates, we find that (4-22) becomes
(4-23)
But now we see from (1-145) that the integrand is zero as long as R + 0, so that any
contribution to (4-23) must come from the region corresponding to R = |r r'| = 0,
that is, from the immediate neighborhood of the field point. Thus, in order to evaluate
(4-23), with its problem of an infinite integrand at a point, we use a limiting process; we
integrate over a small but finite volume AT' around the field point P and then we can
go to the limit AT' - 0. It is easiest, but not necessary, to choose AT' to be a sphere of
radius R about P as shown in Figure 4-6; then we can eventually let R -* 0. We note
that, as usual, R is drawn toward the field point, that is, toward the center of the
sphere. Since |r r'| = 0 within this sphere, p(r') is nearly constant and equal to p(r)
so that we can take it out of the integral; this procedure will become exact as R -> 0. If
we do this, we get
(4-24)
We can now use (1-133) to express the derivatives in the integrand in terms of the
primed coordinates; once this has been done (including the change in sign), we can use
the divergence theorem (1-59) to convert (4-24) into a surface integral over the surface
66
G A U S S ' LAW
AS'
Figure 4-6.
point.
AS'
Small sphere surrounding the field
e n c l o s i n g AT':
( - R ) da'
V
4770 'AC
J&S'
R'
(4-25)
N o w , if w e recall (4-3), we see t h a t t h e i n t e g r a n d is j u s t dil, t a k i n g i n t o a c c o u n t t h e fact

t h a t t h e a n g l e b e t w e e n R a n d da' is 1 8 0 as seen f r o m F i g u r e 4-6; t h e r e f o r e , since P is
i n s i d e A S ' , t h e integral in (4-25) will b e e x a c t l y 4m a c c o r d i n g t o (4-5), a n d (4-25)
becomes
V E =
p(r)
(4-26)
w h i c h is j u s t w h a t we f o u n d b e f o r e . W e n o t e t h a t this d e r i v a t i o n r e e n f o r c e s o u r
r e c o g n i t i o n t h a t p m u s t b e e v a l u a t e d at t h e p o i n t w h e r e we k n o w E(r) a n d w a n t the
v a l u e of V E at this s a m e p o i n t .
EXERCISES
4-1 The rectangular parallelepiped of Figure
1-41 with a > b > c is filled with charge of constant density p. A sphere of radius 2a is constructed with its center at the origin. Find the flux
<f>E da through the surface of this sphere. What
is the flux when the center of the sphere is at the
c o r n e r ( a , b, c)?
4-2 A sphere of radius a has its center at the
origin and a charge density given by p = Ar 2
where A = const. Another sphere of radius 2 a is
concentric with the first. Find the flux <fE du
through the surface of the larger sphere.
4-3 The infinite line charge of Figure 4-3 is
surrounded by an infinitely long cylinder of radius
p 0 whose axis coincides with the line charge. The
surface of the cylinder carries a charge of constant surface density a. Find E everywhere. What
particular value of a will make E = 0 for all

points outside the charged cylinder? Is your
answer reasonable?
4-4 A charge of constant volume density has the
form of a slab of thickness a. The faces of the
slab are infinite planes parallel to the xy plane.
Choose the origin midway between the faces and
find E everywhere.
4-5 A sphere of radius a has a charge density
that varies with distance r from the center
according to p = Ar>/2 where A = const. Find E
everywhere.
4-6 T w o concentric spheres have radii a and b
with b > a. The region between them (i.e., a < r
< b) is filled with charge of constant density. The
charge density is zero everywhere else. Find E at
all points and express it in terms of the total
EXERCISES
charge Q. D o your results reduce to the correct

values as a -* 0?
4-7 An infinitely long cylinder has a circular
cross section of radius a. It is filled with charge of
constant volume density pch. Find E for all points
both inside and outside the cylinder. Are your
results consistent with (4-11)?
4-8 Two infinitely long coaxial cylinders have
radii a and b with b > a, as shown in Figure 4-7.
The region between them is filled with charge of
volume density given in cylindrical coordinates by
p(7, = A p" where A and n are constants. The
charge density is zero everywhere else. Find E
67
everywhere. For what values of n and a should

your results reduce to those of Exercise 4-7? Do
they?
4-9 The region between the infinitely long
coaxial cylinders of Figure 4-7 is filled with charge
whose volume density in cylindrical coordinates is
pch = Ae~ap. Find E everywhere. Under what
simple circumstances should your results here and
for Exercise 4-8 reduce to the same values for E?
D o they?
4-10 The average electrostatic field in the earth's
atmosphere in fair weather has been found experimentally to be approximately given by E =
EQ(Ae~az -I- Be ,iz)z. All of the empirical constants are positive and z is the height above the
(locally flat) surface. Find the average charge
density in the atmosphere as a function of height.
What is its sign? (Over land, a typical value for E0
is = 250 newtons/coulomb.)
4-11 A certain electric field is given by E =
E0(p/a)3p
for 0 a is
given by Er = 2 A cos 0/r3, Ee = A sin 9/P, Ev
= 0 where A = const. Find the volume charge
density in this region.
5
Up to now, our description of electrostatic effects has been done completely in terms of
the vector field E. By rewriting our expression for E, we will see that we will be able to
express substantially the same information in terms of a scalar field that will be much
more convenient for many purposes.
fîl
DEFINITION A N D PROPERTIES OF THE SCALAR POTENTIAL
O u t basic defining equation for the electric field is given by (3-2). If we use (1-141) to
replace R { / R j by V ( 1 /Rt), and then use (1-110) to write the sum of the derivatives
as the derivative of the sum, we find that (3-2) becomes
(5_1)
E(r)
= - E - V ( ) = - V E ~
W
i 4ire0
\R,
^ 4
where, as usual, R t = |r r,|. Thus, we see that if we define
*(') = E
= 1 4-7TQRi
(5-2)
E(r)=-v*(r)
(5-3)
V X E = 0
(5-4)
we can write
so that
according to (1-48). The scalar field <p is called the scalar potential or the electrostatic
potential. The unit in which the scalar potential is measured is called a volt", we see from
(5-3) that the electric field can also be stated in volts/meter, which, in fact, is the most
commonly used unit. Combining this with our previous unit of newton/coulomb for E,
we find that 1 volt = 1 joule/coulomb.
Since the curl of the electrostatic field is zero everywhere according to (5-4), we find
from Stokes' theorem (1-67) that
E ds = 0
(5-5)
where C is an arbitrary closed path. This result shows us explicitly that the electrostatic
field is an example of what is known as a conservative field.
In summary, we have found that the vector electrostatic field can be written as the
negative gradient of the scalar potential, and (5-2) provides us with a method of
calculating
at any desired field point r once we are given the values and locations of
the source point charge distributions. For example, if everything is given in rectangular
coordinates, we can use (1-14) to write (5-2) as
4>(r) = <p(x, y, z) = E
7
'
i.i4n0[(x-ij
^
~T
+ { y - y i ) + (z -
^TTT?
i) \
(5"6)
5-1
D E F I N I T I O N A N D P R O P E R T I E S O F THE S C A L A R P O T E N T I A L
69
Since <> is a scalar quantity, it is generally easier to proceed indirectly by evaluating the
sum (5-2) and then finding E by differentiation by (5-3), rather than by the direct
process of evaluating the vector sum (3-2); this is one reason why <f> is of practical
interest.
If the source charges have a continuous distribution, we can proceed exactly as we
did in Sections 2-4 and 3-1 by using (2-14) and (2-16) to write (5-2) in integral form;
the results are
1
( p(r') dr'
4t7C 0 j y
47T0Q JJSS''
477
<#>(r) = ^
(5
a(r') da'
RR
f X(r') ds'
(
5
44TTC
TT ^
JT'
0 Ju
"7)
K
R
where the integrals are taken over all volumes, surfaces, or lines containing the source
charges or, in fact, over all space, since in regions where there is no charge the
respective charge densities will vanish. Finally, if all of the possibilities we have
discussed are simultaneously present, we see from (5-1) that the total <f> at a given point
will be the scalar sum of all of the contributions from (5-2) and (5-7) through (5-9) and
the total E at a given point can be found as the negative gradient of this total scalar
potential.
There is a certain amount of ambiguity remaining in our results, however, which
arose when we went from (5-1) to (5-2). Suppose that, instead of (5-2), we had chosen
to define <p as
+M -
L
+
, = i 4tte 0 i?,
<5"10)
where C is a constant but otherwise completely arbitrary. When this expression is

substituted into (5-3), we get exactly the same field E as given by (5-1) since v C = 0. In
other words, the scalar potential in principle always includes an arbitrary additive
constant and we can assign this constant any convenient value without changing any
essential features of a given problem. It is inconvenient to carry this constant along in
all one's equations so that the usual convention is to simply choose C = 0, thus bringing
us back to (5-2) and the equations that follow from this. In order to see the significance
of this choice, let us assume that all of the charges in (5-2) occupy a finite region in
space; then we can always choose our field point r to be so far away from all of the
charges that all of the Rt -* oo and <j>(r) > 0. In other words, with the choice of (5-2),
the scalar potential will vanish "at infinity." While this is the standard conyention and
the one we will generally use, there are some situations where it is convenient to choose
other points, such as the surface of the earth, to have a value of zero for the potential; if
we do this, we will mention it explicitly.
There is another useful way of expressing the relation between E and <p. Let us
consider the line integral of E between an initial point Pt at r : and a final point Pf at r 2
as illustrated in Figure 1-22:
J E ds = J - V< ds = -
d<j> = - (<#>2 - <f>i) = - [<>(r2) - <K r i)]
where we have used (5-3) and (1-38). We can write this, using (5-5), as
A<f> = </>(r2)
$(ri)
~J f E
1
=J f ~E ds
2
(5-11)
70
giving us a relation between the change in the scalar potential A<> and the line integral
of the electric field. Since the result depends only on the values of <p at the end points,
the value of the integral is independent of the path. We recall that this is another way of
saying that the field E is a conservative field. [If the path is closed, r 2 =
and (5-11)
gives (5-5) again.] We note that A<f> given by (5-11) is independent of any arbitrary
additive constant that might be included in the definition of <f> since this constant will
cancel when the difference is evaluated. We can use (5-11) to find the difference in
potential between two points if the field E has been found by other methods, such as
those illustrated in the last two chapters, since the known value of E can be used and
the integral evaluated over any convenient path as it is independent of the path which
may be chosen. As we will see, this procedure is sometimes useful when the charges are
assumed to extend to infinity so that <p may not go to zero at infinity, or even becomes
infinite there as well as at various inconvenient points, or in other situations in which
(5-7) through (5-9) are difficult to apply.
A surface on which <f> is constant is called an equipotential surface. We saw in Section
1-9 that the gradient of a scalar is normal to a surface of constant value of the scalar
and in the direction of the maximum rate of change of the scalar. As a result, we see
from (5-3) that the direction of E will be perpendicular to the equipotential surface and
in the sense of decreasing <f>. This is illustrated in Figure 5-1, in which the equipotential
surfaces are shown as solid curves and the dashed lines are drawn to indicate the
direction of E at each point for the case in which <#>3 > <j>2 >
(A line that is defined
to be tangent to the direction of E at each of its points is called a "line of force" or a
"line of E." It is conventional to draw them so that the density of the lines is
proportional to the magnitude of E, that is, the numerical value of E is greater in
regions where the lines are closer together than it is in regions where the lines are
farther apart.) This mutually orthogonal relationship illustrated in the figure enables us
to obtain a visualization of the situation by sketching the field lines once we have
solved for <j> so that the equipotential surfaces can be found; the converse way of
sketching the shape of the equipotentials from the known lines of E is also useful.
Example
Single point charge. Let us illustrate some of these ideas with the simplest possible
example, that of a point charge Q located at r'. In this case, (5-2) reduces to the single
Figure 5-1. Equipotential surfaces (solid)

and lines of electric field (dashed).
5-1 D E F I N I T I O N A N D P R O P E R T I E S O F THE SCALAR P O T E N T I A L
71
term
w
(5
- ^
-12)
where R = |r r ' | . When this is combined with (5-3) and (1-141), the result is
E(r)
"4 ^
<5 13)
"
which is, of course, in agreement with what (3-2) becomes for a single charge. The value
of 4> given by (5-12) as a function of the distance R from Q is shown in Figure 5-2 for
both possible signs of Q. The equipotential surfaces are found by solving (5-12) for R
and letting <j> have a definite value; the resulting equation is
R
4W0<
(5
-14)
so that these surfaces correspond to R = const., that is, they are spheres centered on
the charge Q. This situation is shown in Figure 5-3 in which we have assumed Q to be
positive so that <j>3 > <#>2 >
According to Figure 5-1, E must be perpendicular to
these spheres and in the sense of decreasing <p. Therefore E must be directed radially
outward from Q, in agreement with (5-13).
In principle, an equation like (5-7) provides us with the complete solution to a

problem in electrostatics. As we will see, however, some problems are not stated in such
a way that (5-7) can be used directly, or it is too difficult to do so. For example, we may
know the charge distribution only within a finite region of space, and it is unknown
outside this region, but the values of 4> on the bounding surfaces of the region may be
given instead. Then (5-7) cannot generally be used directly; however, another approach
Figure 5-2. Potential as a function

of distance R from a point charge Q.
Figure 5-3.
charge.
Equipotential spheres due to a point
72
THE S C A L A R P O T E N T I A L
can be formulated. If we combine (5-3) with (4-10), we can eliminate E and we find,
with the use of (1-45), that
V24> =
c
o
(5-15)
In other words, the scalar potential satisfies this differential equation that is known as
Poisson's equation. In regions where p = 0, (5-15) simplifies to Laplace's equation:
V2< = 0
(5-16)
We will defer the discussion of the detailed use of these equations and the justification
of such an approach until Chapter 11.
Finally, we can point out here that our source equations for E, (4-10) and (5-4),
along with the other results (5-3) and (5-7), agree completely with the Helmholtz
theorem quoted in (1-148) through (1-150).
We consider next a few more elaborate examples of the calculation of the scalar
potential for continuous charge distributions.
U N I F O R M SPHERICAL CHARGE DISTRIBUTION
This is exactly the same charge distribution we considered in Section 2-5, that is, a total
charge Q contained in a sphere of radius a with constant charge density p = 2>Q/4ira^.
We again use the coordinate system shown in Figures 2-6 and 2-7; the point in these
figures showing the location of q is now to be interpreted as the field point where we
want to find <>. As before, R2 = z 2 + r'2 2 z r ' c o s 0', so that (5-7) becomes
P
rim r-" ra
r'2 sin 6' dr' d0' dw'
A=
f
( f
47re0 A) A) 'o ( z 2 4- r'2 2zr' cos 0')
where we have used (1-99) and taken the constant value of p outside of the integral.
The integration over dcp' can be performed at once and gives 2tt. If we again let
ju, = cos 0', and use (2-22), we find that (5-17) becomes
dp
<f> = f"r'2dr'
f1 2
2
2e 0 Jq
'-1
' - l (z
(z2 +
+ r'r'z
2zr'a)
Izr'nY
ttj
(5-18)
The integral over ju. can be found from tables to be

(z2 + r'2 - 2zr'ix)1/2
zr'
'
-
= f(\z
zr
+ r'\ -
\z r'\)
(5-19)
There are now two cases to be considered.

1. Outside the sphere. Here z > a\ since we are assuming z to be positive for
simplicity, and since r' < a, we have z > r', so that the parentheses in (5-19) become
(z + r') (z r') = 2r' and the integral over n equals 2 / z . Inserting this into (5-18)
and integrating over r', we find the potential at a point outside the sphere, 4>i/7 to be
pa3
Q
- T
3e 0 z ~ 747re0z
(5
-20>
2. Inside the sphere. Here z < a, so that r' can be both greater than or less than z.
When z < r' < a, (5-19) becomes
7E(z + r') - (r' - z ) ] =
zr
r
5 - 3 U N I F O R M LINE C H A R G E D I S T R I B U T I O N
73
while, when r' < z < a, (5-19) is 2 / z as before. Since the integral over n has these
different values for different ranges of r', we have to evaluate the integral over r' as the
sum of two integrals each with an integrand appropriate for the range of r' involved;
thus, our expression for the potential inside, <,, as obtained f r o m (5-18) becomes
P / cz
- 271 i / '
ca
2
dr
'-7
2 \
'
dr
e
(
87T0a I
6c 0
'
3
- 4 |1
a
(5-2!)
We note that (5-20) and (5-21) give the same value of the potential, Q/4we0a, at the
surface of the sphere where z a.
These results can be easily generalized to an arbitrary location. The field point was
taken to be along the z axis for convenience in evaluating the integral. The position
vector of the field point is thus r = zz so that | r | = r = z and r = z; this enables us to
write our results for any field point in terms of its spherical coordinate distance r f r o m
the origin, that is, f r o m the center of the sphere. Therefore, (5-20) and (5-21) yield these
general expressions for the potential at any point outside of or inside of the sphere:
pa3
Q
*.(') = 7 = - T 3 e0r
47re0/-
(5-22)
(3a2 - r 2 ) = ^ ( 3
6e 0
87T0a y
j]
a J
(5-23)
As a check on our calculation of <j>, we can see if it gives the known electric fields. If
we insert (5-22) and (5-23) into (5-3), and use (1-101), we find the electric fields outside
and inside the sphere to be, respectively,
pa3r
E
Or
=
w 0r
Je
pri
e , -
3c o
44vT
^ 70r*
(5
"24)
Qrr
=
~ r ~ ^
4<7T0a
2 5
and these agree with the results (4-17), (4-20), and (4-21), which we previously obtained
by using Gauss' law.
According to (5-22) and (5-23), constant values of <j> correspond to constant values
of r; in other words, the equipotential surfaces are concentric spheres with centers at
the origin (i.e., at the center of the charge distribution). Figure 5-4 is a plot of the
potential as a function of r; the negative slope of this curve gives the electric field Er
that we have previously shown in Figure 4-5 and it is instructive to compare these two
figures with this in mind.
U N I F O R M LINE CHARGE DISTRIBUTION
As an example of the use of (5-11) to find the potential, let us consider the uniform
infinite line charge for which we previously found the field as given by (3-9) to be
E = ( \ / 2 7 r e 0 p ) p . Since E has only a p component, we see f r o m (5-3) and (1-85) that
is independent of <p and z, that is, it is a function only of p. This already shows us that
the equipotential surfaces will be cylinders whose common axis coincides with the line
charge since <p = const, leads to p = const. Since we can use any convenient path, let us
74
Figure 5-5. Path of integration used t o find

potential difference due t o a uniform line
charge.
choose the simple one shown in Figure 5-5; the path lies in a plane perpendicular to the
charge distribution that intersects this plane at the origin. We will integrate along the
straight line in the constant direction p from the initial point 1 to the final point 2
whose distances from the charge are p1 and p 2 , respectively. We see from the figure that
ds = p dp in this case so that the integrand in (5-11) becomes E p dp = Epdp from
(1-21), and therefore (5-11) becomes
/*P2 ^ dp
A
4>(p2) ~ 4>(Pi) = ~ f z
= ~
Pi 2tre 0 p
2tre 0
ln
I Pi \
(
\ p2 j
(5-26)
This method gives us the difference in potential between the two points, and we would
like to find the absolute value for <j>. The logarithm term in (5-26) can be written as
[( In p 2 ) ( In p x )] and, since any additive constant will cancel on the left side, we
conclude from comparing the two sides of (5-26) that </> has the form
<KP) = -
2tt 0
lnp+C
(5-27)
where C is some constant. Now as p - oo, ln p - oo so that if we try to preserve our

convention that
vanish at infinity, we conclude that C must also be infinite. The
source of this difficulty is that in this case the charge is not finite and extends to
infinity, whereas our expectation that <f> should vanish at infinity arose from (5-2) and
the assumption that all of the charges were contained within a finite volume, as was the
case in the previous example of the sphere. Even though we may be uncomfortable with
it at this stage, the fact remains that the <f> of (5-27) does give the correct field E by
5-3
U N I F O R M LINE C H A R G E D I S T R I B U T I O N
75
means of (5-3). It is sometimes convenient to put (5-27) into another form; if we

introduce a new constant p 0 by writing C = X In p0/2m0,
then (5-27) becomes
^>(p) =
2 77" (
In
7)
(5-28)
W h e n p = p 0 , <(P0) = 0 so that the physical significance of p 0 is seen to be that it is

the distance at which we have chosen
to vanish, since we cannot make it do so at
infinity. The derivative of <f> as obtained from (5-28) and that gives E is unaffected by
the choice of p 0 , as is the potential difference that will still be given by (5-26).
Since these effects arose from our assumption that the line charge was already
infinitely long, it is helpful to look at the problem afresh by starting with a uniform line
charge of finite length for which we will get a finite result as we will see, and then seeing
what happens as we let it get very long. W e will also obtain in this way a plausible way
of choosing p 0 . Let us consider the charge distribution shown in Figure 3-8 for
X = const., and calculate <t> at the general point P ( p , <p, z). Since the charge distribution coincides with the z axis, r ' = z% while r = pp + zz according to (1-81). Therefore, R = pp + (z z')z so that R = [p 2 + (z z ' ) 2 ] 1 / 2 and, since ds' = dz' in this
case, (5-9) becomes
X
rr,
dz'
<t> =
(5-29)
1/2
47Te0 - L i [p 2 + ( z Z') 2 ]
The integral is easily evaluated with the use of tables and the result is
1/2
<\>
47TCr
z + Ly + [ p 2 + ( z + Lj) J
In
(5-30)
1/2
z - L2 + [p 2 + ( z - L2)2\
[It will be left as an exercise to verify that when (5-30) is used in (5-3), the same E is
obtained as was found by direct evaluation in Exercise 3-11.] N o w let us assume that
the line charge is very long, although not yet infinite, that is, we assume that L2 p,
L2 | z | , Lx p, and Lx s> | z | , so that we can get an approximate expression for <f>
for situations in which our field point is not too close to the ends of the charge
distribution nor too far away from it in comparison with its length. The numerator of
the argument of the logarithm in (5-30) gives us no trouble and is approximately 2 L 1
since we can write z + Lx = L1 and can similarly neglect p in comparison with Lv We
have to be more careful with the denominator since just omitting z and p makes it
vanish; accordingly we expand it in a power series in an appropriate small quantity and
keep the lowest order nonvanishing term. We use the expansion
(1 + x)
1/2
1 + \x \x2
(5-31)
although here we will require only the first two terms and the upper sign. If we factor
out L2 z, the denominator becomes
l/2\
(L
-Z) -1 +
1 +
- L,
(L2-ZY
Substituting this into (5-30), along with 2L1

<f>
477*6 (
In
p2 \ ^
1 + 7
L
/
- LA-
4 L '2I L
-1 +
1 + 1 +
2 L\
2L,
for the numerator, we get
2tt t
In
(4 L 2 L J
1/2
(5-32)
76
which is exactly of the form (5-27) or (5-28). Thus we are led back to the same general
expression for <f> as we approach the limit of an infinite line. We note that (5-32) is
independent of z, which is reasonable since if we d o not get too near the ends of this
very long line charge, a displacement parallel to the line leads one to a point where the
charge distribution will appear essentially unaltered.
If we now go further and let both L2 and Lx get infinitely large in (5-32), we again
get an infinite value for <f>. Hence this approach has not made the infinities inherent in
our problem disappear, of course, but we have a better understanding of how they have
arisen.
There is one interesting case in which these ambiguities in <fi disappear because of
the n a t u r e of the system.
Example
Two parallel oppositely charged lines. Let us consider a system comprised of one long
uniform line charge of density A and the other of density A. We assume the same
values of L2 and L1 for them as shown in Figure 5-6. The distance p in (5-32) began as
the cylindrical coordinate variable but is clearly the perpendicular distance f r o m the
line charge to the field point P; these distances for this case are labeled p + and p_ in
the figure. If we assume that L2 and L1 are already large enough, we can use (5-32).
Thus the individual potentials due to these two line charges are
0-
<#- =
2 7T f
(4 L2L,)
In
2lTf
In
1/2
(4L2L1)
1/2
so that the total potential at P is

0 = <#> + + <f>-
"
2 ttc i
in'"-
(5-33)
and the terms depending on L2 and Ll have canceled. Now we can let L2 and Lx
each go to infinity and we will be left with the same unambiguous result given in (5-33).
In a specific case, p + and p_ have to be evaluated in terms of the particular coordinate
system that is being used. Let us illustrate this for the case in which the line charges are
Figure 5-6. Two parallel oppositely

charged lines.
5-3
U N I F O R M LINE C H A R G E D I S T R I B U T I O N
77
taken to be parallel to the z axis, the line of length 2 a between them lying along the x
axis, the origin is chosen midway between them, and P lies in the xy plane as shown in
Figure 5-7. We will use polar coordinates p and <p to specify the location of P.
Applying the law of cosines, we find that p\= a2 + p2 lap cos <p and p2_ a 2 + p 2
+ lap cos <p, so that (5-33) can be written
+ p2 + lap cos <p
_2 \
<p(p, <p) =
4TT(
ln
P+
In
4 7TC I
+ p 2 lap cos Xa(p2
Em/dz = 0.
The equipotential surfaces <f> = const, are given by (5-34) as
n2
= const.
P+
(5-37)
This equation is more easily interpreted in rectangular coordinates; we see by inspecting Figure 5-7 that it becomes
+ a)2 + y' 4o<t>/x
e
(x a) + y2
(jc
which, with a little algebra, can be written in the form

\ 2
(x a cothi)) + y' - (
\ sinh tj
(5-38)
where 17 = l-rre0<p/\. This is just the equation of a circle with radius a/sinhrj and with
center displaced a distance a coth 77 along the x axis. In other words, the equipotential
surfaces are cylinders with axes parallel to the z axis whose intersections with the xy
plane are the circles given by (5-38). These circles are shown as the solid curves in
Figure 5-8. We note that the circles whose centers lie on the positive x axis correspond
to <f> > 0, while those with centers on the negative x axis correspond to <j> < 0. We also
note that the yz plane (x = 0) is the equipotential surface for <J> = 0; this is easily seen
Figure 5-7. Geometry for

the potential of t w o line
charges parallel to the z
axis.
78
Figure 5-8. Equipotentials (solid) and lines of electric field (dashed) for t w o oppositely
charged lines parallel to the z axis.
to be the case since we see from Figure 5-7 that each point on the y axis has p + = p_,
making <p = 0 according to (5-33).
The lines of E are perpendicular to these equipotentials as we know and are shown
as dashed lines in Figure 5-8. We can use our results for this case to illustrate how one
can find an explicit equation for these curves. As we mentioned before, a line of E is
defined in order to be tangent to the direction of E at the corresponding point.
Therefore, if d s , , represents a small displacement along a line of E (line of force), it is
necessarily parallel to E at that point; this is shown in Figure 5-9. Therefore, we can
write
ds/f=kE
(5-39)
since they have the same direction and where A: is a constant of proportionality with
appropriate dimensions. We can use (5-39) to obtain a differential equation for the
curve describing the line of E.
In order to apply (5-39) to the particular case worked out above, we note that d s
will equal dx as expressed by (1-82) where r is the position vector of each point on the
curve. Since the line of E lies in the p, <p plane, dz 0 and (5-39) becomes
dpp + p dcpy = k(Epp
+ Ev$)
Equating components, we get

dp = kEp
and
p dcp kEv
(5-40)
5 - 4 THE S C A L A R P O T E N T I A L A N D E N E R G Y
79
and we can now eliminate k by dividing the first equation by the second. The result is
1 dp
ED
-P- dcp
T
(5
T~
Ey
~41>
The right-hand side of (5-41) will be a known function of p and <p, which, in principle,
will enable us to integrate this equation to find p as a function of <p, that is, the
equation of the line of force.
In our case, if we substitute (5-35) and (5-36) into (5-41), we find that
1 dp
(p 2 a2) cos<P
p dfp
(p 2 + a2) sin <p
or
(p2 + a2) dp
cos <p dcp
p (/ p 1 a 2 \)
sm<p
This equation can be integrated with the help of tables to give

ln
/ p 2 - a2\
\
= ln sin <p + ln K
I
where the constant of integration is written in terms of the dimensionless constant K.

From this result, we find our desired equation to be
p 2 a1 = Kap sin<p
(5-43)
For each value of K, we get a corresponding line of force. If we express (5-43) in

rectangular coordinates with the use of (1-74) and (1-75), we find that it can be written
x2 + {y - \Ka)2
= a2(\
+ \K2)
(5-44)
showing that the lines of E are also arcs of circles; a given circle has a radius
a( 1 + \K2)l/2
and its center displaced a distance \Ka along the y axis. These are the
curves that are plotted as dashed lines in Figure 5-8 and we see that, as expected, they
originate on the positive charges and terminate on the negative charges since it follows
from (5-43) that when <p = 0 or IT, p = + a o r , from (5-44), that x = a when y = 0.
THE SCALAR POTENTIAL A N D ENERGY
In general, changes in potential are related to changes in energy. Consider a point

charge q in equilibrium under the action of an electrostatic force F ? and a mechanical
force
m . This condition was described by (2-9) for only two charges, but can clearly
be extended to the case when more than one charge is acting on q. In this case, it is
appropriate to use (3-1), so that the condition of equilibrium becomes
Fq + F q, m - ^n E ~f- F q, m
or
F,,
=-Ê
(5-45)
Now let us imagine moving the charge infinitely slowly from an initial point r j to a final
point r 2 along some path. Under these conditions, the velocity will be essentially always
zero and constant so that the acceleration will be zero. The charge will then be always
80
in equilibrium, or nearly so, so that (5-45) applies. We assume this procedure so that we
can calculate the amount of reversible work in the thermodynamic sense that is done by
the external mechanical force, and by keeping the velocity zero, we can be sure that
there will be no dissipative or frictional effects involved. If we let Wx_,2 t>e the work
done by the external agent responsible for the mechanical force, we get
= Sy>m ' ds =
' ds =
~ *(ri)]
(5~46)
with the use of (5-11). In other words, the work done on the charge equals the value of
its charge times the change in potential. Under the circumstances we have assumed, the
work done can be equated to the change in potential energy AUe of the charge so that
(5-46) becomes
AUe =
r 2 ) - <#>(ri)] = q&<t>
(5-47)
We note that this change AUe is independent of any additive constant that may be
included in <p. Since the right-hand side of (5-47) already has the form of a difference, it
is natural to write the left-hand side in the same way, that is, AUe = Ue(r2) Ue(rx),
and by comparison, we can define the potential energy of a charge q at r, Ue(r), as
Ue( r) = q<t>( r)
(5-48)
As usual with potential energy, we can add any arbitrary constant to the right-hand
side of (5-48) without changing the difference, much as we could do for <p, that is, as in
(5-10). We will generally choose the form (5-48), however, as it has the convenient
property that when <f> vanishes at infinity, Ue does also. Since the energy Ue will be
measured in joules, we see again from (5-48) that the unit of 4>, the volt, will be equal to
1 joule/coulomb.
Example
Two point charges. Let us consider a system of two point charges q and Q separated by
a distance R as shown in Figure 5-10. The potential at the location of q is given by
(5-12) and when this is inserted into (5-48), we find that
v, -
qQ
;r
47re0R
(5-49)
This energy can be interpreted as the work that was required to bring q from infinity to
its location r while the charge Q was held fixed at r'. But, because of the symmetry of
EXERCISES
81
the expression, it can equally well be interpreted as the work required to bring Q from
infinity to r ' while holding q fixed at r. In other words, it is more appropriate to regard
Ue as the mutual potential energy of the system of the two charges rather than ascribing
it to one charge or the other. We return to this question of the potential energy of a
system of charges in detail in Chapter 7.
EXERCISES
5-1 Can the vector E = ( y z 2x)x 4- xzy +
xyz be a possible electrostatic field? If so, find the
potential <t> from which E can be obtained.
5-2 Could the vector A of Exercise 1-15 be
interpreted as a conservative electric field? If so,
find the potential as a

function of r.
5-12 Find < for all points outside and inside the
cylinder described in Exercise 3-13.
5-13 Consider the charge distribution of Exercise 3-10 and Figure 3-7. Find <f> at an arbitrary
point on the z axis. Why does your result give the
correct value of Ez but not of EX1
5-3 Two point charges q and q are located on

the z axis at z a and a, respectively. Find <#> 5-14 A sphere of radius a has a constant surface
at any point (x, y, z). Show that the xy plane is charge density a, but no volume charge density.
an equipotential surface and find its potential. Find <p for all points both outside and inside the
Explain this result.
sphere by using (5-8).
5-4 Consider the charge distribution of Exercise 5-15 An infinite sheet of charge of constant
3-2. Find the potential at the center of the square. surface density a coincides with the xy plane.
Why can't you find E at the center of the square Use the expression for E given by (3-13) and find
from your result?
a potential <#> from which E can be found. What
5-5 Consider a cube of edge a with the location are the equipotential surfaces in this case?
and orientation of the figure in Figure 1-41. There 5-16 Find the potential due to the sheet of the
is a point charge q at each corner. Find </> at the previous exercise by using (5-8).
center of the face for which x a.
5-17 Charge is distributed with constant surface
5-6 Within a certain region of space the electric charge density a on a circular disc of radius a
field E is constant. Show that a suitable potential lying in the xy plane with center at the origin.
for this case is <f> = E r + <#>0 where <#>0 is a Show that the potential at a point on the z axis is
constant. What are the equipotential surfaces?
given by
5-7 Obtain the results (5-22) and (5-23) by using
<t> = [(a2
+ zz)1/2
- \z\]
(5-50)
(5-11) and the fields given by (5-24) and (5-25).
zc 0
5-8 A sphere of radius a has a total charge Q
distributed uniformly throughout its volume. The
center of the sphere is at the point (A,B,C).
Find the potential <j> at any point (x, y, z) outside the sphere, and from this, find the rectangular components of E at that point.
5-9 A sphere of radius a has a charge density
that varies with distance r from the center
according to p = Ar" where A = const, and n >
0. Find
at all points inside and outside the
sphere by using (5-7) and express your results in
terms of the total charge Q of the sphere.
5-10 Find <j> for all points inside and outside the
sphere of charge of Exercise 4-5 by using (5-11).
Plot 4> as a function of r.
5-11 Find 4> at all points for the charge distribution of Exercise 4-6. Express your answer in terms
Verify that this gives (3-15). What does <f> become

as a becomes very large? What is the smallest
value of z for which the potential due to this disc
can be calculated as if it were a point charge
without making an error greater than 1 percent?
5-18 Show that = 0 and
<p = 90.
5-20 The parameter K, which characterizes a

given line of force in (5-43), can be related to the
magnitude E of the electric field where the
line crosses the y axis, that is, when a to another point on the x axis
for which jc = b, what is the change in potential?
5-22 Express <j> for the charge distribution of
Figure 5-7 in rectangular coordinates and use this
result to find Ex and Ev.
5-23 Consider the charge distribution of Exercise 5-3. How much work must be done by an
external agent to change the separation of the
charges from 2 a to al Illustrate this process on a

plot of Ue versus separation R.
5-24 Consider the line charge of Figure 3-8 with
Ll = 0. The charge per unit length is proportional to the cube of the distance from the origin.
Find the potential at a point P on the x axis for
which x = a. From your result, find Ez at P\ if
you cannot do this explain why not. How much
work must an external agent do to move a point
charge q very slowly along the x axis from P out
to infinity?
6
C O N D U C T O R S IN
ELECTROSTATIC FIELDS
Coulomb's law assumes that the region between the charges involved is a vacuum. We
are interested, of course, in what will occur when electric charges are in the presence of
matter, so that at least part of the region of interest to us is not a vacuum. We will not
consider this problem all at once, but in this chapter we restrict ourselves to the case in
which a particular class of matter known as conductors is present. Because of the
special properties of conductors, we will be able to deduce some interesting and
important consequences that arise from their presence.
6-1]
SOME GENERAL RESULTS
In general terms, a conductor can be defined as a region in which charges are free to
move about under the influence of an electric field. The most common example is that
of a metal in which the movable charges are " f r e e " electrons, but it is not the only case,
and we will not need to restrict ourselves here to such a specific example. We do,
however, assume that we are dealing with a completely static situation on the macroscopic scale.
Now, if an electric field were present in a conductor, the charges would move about,
and we would not have the static situation we are assuming. Therefore, we conclude
that E must be zero at all points within a conductor. It then follows directly from (5-3)
that <f> is constant within the conductor, that is, the conductor is an equipotential
volume. Therefore, we have
E(r) = 0
<p( r) = const.
inside a conductor
(6-1)
Now let us consider the situation at the surface of a conductor. The electric field in
the nearby vacuum region certainly can be different from zero. Suppose that E made an
angle with the surface as shown in Figure 6-1 a. We can imagine E resolved into a
component E, which is perpendicular (normal) to the surface, and a component E,,
which is parallel to the surface (the tangential component). Now, if E, were different
from zero, there would be a tangential force on the movable charges, they would move
parallel to the surface, and we would no longer have the static situation we are
assuming. Therefore, E, = 0, and the only possibility is that E
0; in other words, at
the surface of a conductor, E must be normal to the surface, as shown in Figure 6-1 b.
But, as was illustrated in Figure 5-1, the direction of E is perpendicular to the
equipotential surfaces, and therefore the surface of the conductor is an equipotential
surface. In summary, we have
E(r) * 0
E,(r) = 0
at surface of a conductor
(6-2)
<^ (r) = const.

83
84
C O N D U C T O R S IN ELECTROSTATIC FIELDS
(a)
(b)
Figure 6-1. (a) Hypothetical

direction of E at a conducting
surface. (b) Actual direction
of E.
(It m a y happen, of course, that at a particular point or points on the surface, the
normal component E may also be zero, but, in any event, it is the only component
that can be different from zero at the surface.)
N o w let us apply Gauss' law (4-1) to an arbitrary closed surface that is completely in
the interior of a conductor such as S shown dashed in Figure 6-2. Since E = 0 at all
points inside, it will be zero on each part of S and, in this case, (4-1) becomes
<f>E - da = 0 =
r
e
s
0
(6-3)
and therefore the total charge inside S, Qin, is zero. Because the surface S is completely
arbitrary, it can be deformed in any desired way, even to coincide with the bounding
surface of the conductor and Qm will still be zero. Thus, we can conclude that Qin = 0
everywhere inside. In other words, we have found that the net charge in the interior of a
conductor is always zero, so that whatever net charge is present on a conductor must
reside entirely on its surface. This was first shown experimentally by Faraday in his
famous "ice pail" experiment in which he measured the charge distribution on an
ordinary metal pail. We note that this result is a direct conclusion f r o m Gauss' law that
itself is a consequence of the inverse square n a t u r e of Coulomb's law. In fact, this
conclusion forms the basis for the most accurate experimental tests of the exactness of
the exponent 2 in Coulomb's law, since the experiments essentially look for deviations
from the prediction Qin = 0; the results show that the exponent is indeed equal to 2 to
within a few parts in 10 16 .
In Figure 6-3, we show the situation at the surface of separation between a
conducting region (1) and the vacuum region (2) where ft is the outer normal to the
conductor. We construct a small cylinder of cross-sectional area A a whose curved side
is parallel to n, so that its outward normal n c is parallel to the surface of separation.
The upper face of the cylinder is entirely in the vacuum region where E = 0 while its
lower face is entirely in the conductor where E = 0; we choose A a to be so small that E
is essentially constant over the upper face. We now apply Gauss' law to this small
6-1
Â
S O M E GENERAL RESULTS
85
- Aa
(2)
Figure 6-3. Calculation of the

electric field at the surface of
a conductor.
Si
(1)
cylinder, noting that Qin = oAa in this case; therefore

<E da = f
"\S
'top
E ft da + f
E n cda + f
"side
^bottom
E da = E ha =
ô
since E n = E = const, in the first integral, E n c 0 in the second, and E = 0 in the

third. Thus the value of E at the surface of a conductor is just o/e0, and, if we combine
this with (5-3) and (6-2), we can write
V<f> = ~
(6-4)
e
0
If a is positive, then E is directed away from the surface, while if a is negative, then E
is directed toward the conducting surface; these signs are consistent with the direction
of the force that would be exerted on a positive test charge placed near the conducting
surface. We note that E as given by (6-4) is exactly twice the value of E due to a plane
sheet of the same a that we found in (4-12) by a similar calculation. We can understand
this difference somewhat crudely by first deciding that the total flux of E per unit area
that can be produced by a given surface charge density is o / e 0 ; then, in the case of the
plane sheet, this total flux could be directed equally in the two directions away from the
charge, while for the conductor, since the electric field must be zero inside, only one
direction is available for this total flux. As we will find in subsequent examples, a is not
necessarily a constant but can vary with position on the surface; when this is the case,
E will also vary with location on the surface (but will always remain normal to the
surface). For example, suppose that the conductor were originally neutral, that is, it had
no net charge. Then, when a field is produced in its vicinity by moving other external
charges up from infinity, the conductor's movable charges will not generally be already
in equilibrium with respect to the new situation and will have to readjust their
locations. Thus, for a short time, there will be a flow of charges around the conductor
but equilibrium will eventually be established and then (6-1), (6-2), and (6-4) will all be
satisfied. But since it originally had no net charge, it must still have no net charge, that
is, <f>so da = 0, thus showing that a cannot be of the same sign everywhere and thus
cannot be a constant. On the other hand, an external agent might transfer a net amount
of charge to the conductor, that is, making it "charged." The surface charge density
now could be of the same sign everywhere, and may even be constant providing there is
sufficient symmetry.
^ a , surface =
Example
Isolated spherical conductor. Let us assume a net charge Q on a conducting sphere of

radius a. If it is otherwise isolated, there will be no reason for preferring one direction
over another so that we will have spherical symmetry. As a result, the charge Q will be
86
distributed uniformly over the surface with a constant density a = Q/Ama2, and the
electric field at the surface will be given by (6-4) as E = Q/Amtâ2. Furthermore, since
the field E at the surface is in the normal direction, it will be radial.
We can come to these same conclusions in another way. We have seen several times
that, because of the spherical symmetry, the charge Q can be treated as if it were a
point charge at the center of the sphere as far as effects outside the sphere are
concerned. The potential outside will be exactly that given by (5-22), that is, Q/4ve0r,
while the potential inside is constant and equal to the potential at the surface ( r = a)
according to (6-1) and (6-2); thus we obtain
Q
<t> = ~A
4770/<t> = ~A
^
(r > a)
(6-5)
(r < a)
The electric field outside will be given by (5-24) which becomes E = Qr/4TT(.0a2
when
evaluated at the surface; this agrees exactly with the result obtained in the previous
paragraph from (6-4). Inside the sphere, E = 0, because of the constant value of the
potential given by (6-5) for r < a.
u
We have not yet exhausted the possibilities of using Gauss' law. Now let us consider
a conductor with a cavity inside of it as shown in Figure 6-4. The total bounding
surface of the conductor now has two parts: an outer surface Sa, and an inner surface
St. Again let us consider an arbitrary surface S lying completely within the body of the
conductor so that E = 0 on every point of S. Then (6-3) applies in this case too so that
Qin = 0, that is, there can be no net charge contained within the surface S.
Now suppose that there is a charge inside the cavity, which we call Qcav. In order
that Qin = 0 for S, there must be an equal and opposite charge somewhere within S,
that is, on the conductor. But we found above that any charge on a conductor must
reside entirely on its surface, and therefore this charge Qt must be found on the inner
surface
Thus, we have Q{n = 0 = Qcav + Qt, or
Qi
G i n n e r surface
Qcav
(6-6)
which is a result that is always true.

If the conductor were neutral before Q cav was inserted, it must remain neutral
afterward; therefore a charge Q0= Q, = + Qcav will appear on the rest of the
surface, that is, on the outer bounding surface SQ as indicated in the figure. In this way,
a charge within the cavity would make its presence known to someone outside the
conductor by means of the induced charge Q0 produced on the outer surface and the
electric field produced by it.
Figure 6-4. Charges on a conductor with a cavity in

the interior.
6-1
S O M E G E N E R A L RESULTS
87
Let us consider the case Qcav = 0; then it will always be the case that the charge on
the inner surface S, is zero according to (6-6). (Q0, on the other hand, would be zero
only if the conductor were originally neutral and were kept that way.) Now consider a
closed surface S' lying entirely within the cavity as shown in Figure 6-5. Suppose that
S' is an equipotential surface whose potential <j>' is greater than </>,, the potential of S:,
which is also an equipotential surface by (6-2). Then, according to (5-11) and Figure
5-1, there will be lines of E generally directed from S' to St. Consequently, if we
evaluate the surface integral of E over another surface S" located between S' and S1,,
we will find that
< E da = 0
Js
and, in fact, the value of the integral will be positive since the cosine of the angle
between E and da will always be positive. But, by Gauss' law (4-1), this means that
there must be a positive charge somewhere within S contrary to our assumption that
<2cav = 0. Thus we see that <f>' cannot be greater than </>,. In a similar way, we will find
that <#>' cannot be less than <#>,. Therefore, S' must be an equipotential surface with the
same potential as St. But S' is completely arbitrary and can be made larger, made
smaller, or otherwise deformed in any manner; thus, all points within the cavity can be
evaluated in this way, and the net result is that the potential is constant and equal to 4>t
within the cavity so that E = 0 at each point in the cavity. In other words, if there is no
charge within a cavity inside a conductor, the cavity will be an equipotential volume
and the electric field will be zero everywhere within the cavity. In fact, the potential
within the cavity will be the same as that of the conductor because of (5-11).
These general conclusions are still valid, of course, even if there is another conductor
within the cavity as shown in Figure 6-6. As long as the inner conductor C' is
uncharged, there will be no electric field within the cavity and C' will always have the
same potential as C, the surrounding conductor. Thus, regardless of what charge may
be placed on or outside of C, of whatever sign or distribution, there will be no electric
field at C' and its movable charges will be unaffected. In the usual terminology, the
inner conductor will be completely shielded or screened; this principle of electrostatic
shielding has practical applications such as in the shielding of electronic components in
circuits by enclosing them in metal containers.
Figure 6-5. The surfaces 5' and S" are

entirety within the cavity. S" completely
surrounds S'.
Figure 6-6. The conductor C' is in

a cavity in the conductor C.
88
C O N D U C T O R S IN E L E C T R O S T A T I C FIELDS
SYSTEMS OF C O N D U C T O R S
Let us consider a system consisting of n conductors that we number in some manner,

that is, 1, 2 , . . . , 7 , W e assume that the conductors are charged with total charges
(?i, Q2,. ., Qj, .., Qn- We know that each charge will be on the surface of the
corresponding conductor so that they can be described by the respective surface charge
densities a x , a 2 , . . . , a - , . . . , a. Although we are not yet able to calculate all of these
quantities, we can expect in general that the o ' s will not be constant, but will usually
vary with position in some manner. In any event, the potential at any point P, which
we write as <f>P, can be found in principle from (5-8) and is
4770 JS'
_/=l 40 JSj
RJ
Here we have written the total potential as the sum of the contributions from each
conductor: Sj is the surface of the 7'th conductor, dcij the element of area of this
surface at the location r}, and Rj = \rp Tj\ is the distance from dcij to the field point
P at rP. These relations are illustrated in Figure 6-7, although the various position
vectors are not shown for simplicity. Now (6-7) is completely general, and, therefore, it
must hold in particular when P is chosen to be any point on the equipotential surface
of the z th conductor whose potential is <(; thus, we can write
"
<t>,= E 7
a. da /
--1
j = \ 477-e0 'sj Rji
/ = 1,2,...,)
(6-8)
where RJ( is the distance from the point ry of the 7 th conductor to the particular point
in question of the i th conductor. Since i could be chosen to be any of the conductors,
(6-8) really represents a system of n equations, each of which has n terms on its
right-hand side. [We note that the sum in (6-8) also includes the term 7" = i, that is, it
includes the integral over the surface of the conductor whose total potential we are
calculating.]
It will be convenient for our purposes to write (6-8) in terms of the total charges Qy
The average surface charge density of the 7th conductor (oj) will be just the total
charge divided by the total area, that is, (or) = Q,/Sr The actual charge density oJ at a
6 - 2 SYSTEMS O F C O N D U C T O R S
89
given point is not generally equal to the average but will be proportional to it; this
allows us to write
a,
(6 9)
= V,
"
where f j is a factor that describes how the actual charge density differs from the
average and itself will be a function of position on the surface of the y'th conductor.
Thus, Oj ~ Qj, and when (6-9) is substituted in (6-8) we get
+ . - - r " V /
y=1
t)Sj JSj Rji
(6-io)
These equations can be written in the form

*, = PijQj
0 = 1,2,...,*)
(6-11)
7-1
where
fjdaJ
ki
4ire:n0S;Js,
j SJ lxR,,
ji
<W2)
XV
thus showing that the potential of any conductor depends linearly on the charges of all
of the conductors, including its own. If (6-11) is written out explicitly, we get the set of
equations
4*1 = PllQl
+ P12Q2
+ + PlnQn
4*2 = P2lQl
P22Q2
+ + PlnQn
4> = PlQl
+ PrtlQl
(6-13)
+ + PnQn
T h e set of coefficients ptj defined by (6-12) are called coefficients of potential, and, in
general, their total number is n2. We note that (6-12) no longer contains any reference
to the potentials or charges so that the ptj represent purely geometric relationships. If
we knew the charge distributions so that the factors f j were known, then, in principle,
we could calculate all of the coefficients ptJ f r o m (6-12); we have not yet considered
m e t h o d s of solving electrostatic problems that would give us such information. Nevertheless, these coefficients are always measurable in principle, for we see from (6-11) that
=(
=P,j
(6-14)
so that ptJ is interpretable as the ratio of the change produced in the potential of the
/ th conductor to the change in the charge of the j th conductor when the charges on all
of the other conductors are kept constant.
The P j j have an interesting and useful symmetry property. If we eliminate f j in
(6-12) by using (6-9), we find that
1
r a, da,
PijQj
y
= 2
4^0
S, Rji
(6-15)
which also follows from equating the corresponding terms in the sums of (6-8) and
(6-11). W e also have
Qi = J/ i dai
s,
(6-16)
90
according to (2-16), and if we multiply these last two equations together, we get
1
r r a, a, da, da,
P u Q i Q j - j Jf j
V
4 7 7 o S/Sj
Rji
(6
"17)
If we interchange the indices i and j in (6-17), we get

1
o,o, da, da{
Now the order of integration is opposite for (6-17) a n d (6-18) so that, for example, in
(6-17) we would first integrate by keeping r, fixed a n d letting r, vary, and then finally
integrate again by letting r, vary, whereas these would be interchanged for (6-18).
Nevertheless, each double integral will include the contributions of all pairs of area
elements, and, since each specific pair always has the same distance of separation, we
will have the numerical equality that RtJ = RJt. Thus, we see that the integral of (6-18)
is exactly that of (6-17); therefore, the left-hand sides are also equal, making P t j Q i Q j =
PjiQjQi, or
Pji = Pij
(6-19)
The physical content of this symmetry property of the p's can be expressed according
to (6-19) and (6-11) as follows: if a charge Q on conductor j brings conductor z to a
potential <p, then the same charge Q placed on / would bring j to the same potential
<PSometimes, if the problem is simple enough to be readily solved, the ptj can be more
easily f o u n d from (6-11) than from (6-12).
Example
Isolated conducting sphere. This is an example we solved in the last section and the
potential at the surface is given by (6-5) to be 4> = Q/4ire0a.
For a single conductor
(6-11) reduces to 4> = pnQ and by comparison we see that
Pn ~
47T0a
6"20>
Thus, we were able to find the single coefficient of potential for this simple case and it
did turn out to depend on the geometric properties of the system.
CAPACITANCE
One of the earliest uses of conductors in electrostatics was for the storage of electric
charge; the conductor could be charged, for example, by giving it a definite potential by
means of a battery. For such an application, one would naturally be interested in the
"capacity" of the conductor for charge, in much the same sense that one would refer to
the capacity of a barrel in terms of how many apples it could hold. Nowadays, such
systems are usually called capacitors and the quantitative measure of their capacity is
called capacitance. Generally speaking, there are only two systems of conductors of
basic interest in this connection: a single isolated conductor, and a system of two
conductors with equal and opposite charges.
F o r an isolated conductor, the sum (6-11) reduces to the single term
= pnQ
(6-21)
6-3 CAPACITANCE
91
In this case, the charge is always directly proportional to the potential and the
capacitance C of a single conductor is defined as this ratio; thus
Q
1
C = =
0
/>11
(6-22)
and will be a definite property of the conductor and related to its geometry. As an
example, we can consider the sphere for which we found p u to be given by (6-20), so
that the capacitance is
^sphere =
4?rc
(6-23)
and is directly proportional to the radius. Since the units of e 0 were originally given as
f a r a d / m e t e r in (2-4), we see from (6-23) that the unit of capacitance is the farad; we
also see from (6-22) that 1 farad = 1 coulomb/volt, which is, of course, entirely
consistent with our previous finding of 1 (coulomb) 2 /joule given after (2-4).
When we now consider a system of only two conductors, the system of equations
(6-13) reduces to
01
02
PuQi
P12Q2
P2lQl
P22Q2
^ 24^
along with
i6'25)
P12 = P21
according to (6-19). Thus, the potential-charge relationships for this system will
generally require the knowledge of three quantities: pn, p22, and pl2. When the two
conductors are used as a capacitor, however, a somewhat more special situation is
envisionedone assumes the two conductors to be connected to each other by a
conducting path and the process of charging up this system is transfer of charge from
one to the other. Under these circumstances, the charge on one conductor will always
be equal and opposite to the charge on the other.
Accordingly, we take for our general definition of a capacitor the following: any two
conductors with equal and opposite charges Q and Q. Even so, it is not immediately
evident that one can even define a capacitance in this case. However, when we set
Qx = Q and Qz = -Q in (6-24), we get
^ - ( P u - P u ) Q
02
(P21
(6
26)
P 22^) Q
so that the difference in potential between the conductors is

A0 = <j>1 <f>2 = (/?n + P22 ~ P12 ~ P 21) Q
(6-27)
and we see that for a capacitor with equal and opposite charges, the charge and the
potential difference will always be proportional to each other so that we will be always
able to characterize the system by a single parameter. We call this quantity the
capacitance, C, and define it as
C = %A 4>
(6-28)7
by analogy with (6-22). U p o n comparing (6-28) and (6-27), and by using (6-25), we see
that the capacitance can be expressed in terms of the coefficients of potential as
C =
Pw
P21
P12
(6-29)
92
so that in this two conductor case, the capacitance is still essentially a property
reflecting the geometric relationships of the system.
Example
Spherical capacitor. Let us consider the two conductors shown in Figure 6-8; the
bounding surfaces are concentric spheres of radii a, b, and c. We call the inner
conductor 1, and the outer 2. It is assumed that 2 completely surrounds 1. The p's that
we need in order to find C can be obtained from the general relations (6-24) by
considering appropriately chosen special cases. First of all, let us suppose that Qx = 0
while <2 2 ^ 0, so that (6-24) becomes
0i = P12Q2
and
02 = P22Q2
(6-30)
From (6-6) it follows that Q2 is entirely on the outer surface of radius c; therefore, if
we use the appropriate radius of c rather than a in (6-5), the potential of 2 will be
given by
Qi
02 = ~~A
Airt0c
(6-31)
P22 = ~A
AttCqC
(6-32)
so that (6-30) gives
Since there is no charge within the cavity, our previous discussion in connection with
Figures 6-5 and 6-6 tells us that <j>1 = <j>2 so that, according to (6-30),
P12 = P22
(6-33)
for this particular system.

In order to find pn, let us now assume that Qx # 0 while Q2 = 0; the latter is the
net value on conductor 2 since the inner surface of radius b must have a charge Qx
according to (6-6). In this case, (6-24) yields
0i = PwQi
and
02 = P21Q1
(6-34)
We can relate these two expressions by using (5-11); the field in the vacuum region for
which a < r < b can be obtained from Gauss' law, and the result will again be that, in
6-3 CAPACITANCE
93
this region, Qt acts as if it were a point charge so that the field will be radial and given
by (4-17) with Q replaced by Qx. Combining all this, we get
$ 1 ~ <t>2
E ds = j
Erdr
= fh Qldr2 = - ^ - ( - - - )
Ja 4 t 7 t 0 r
Anre0 \ a
bJ
(6-35)
The left-hand side can be written as ( p l x p2i)Qi

because of (6-34); canceling Qx on
both sides and using (6-33) and (6-32), we finally get
1 / 1
P11 = P22 + 1
477-e0 \ a
1\
1 / 1
T = ^
bJ
4tt0 \ a
1 \
T "*
b
cj
(6-36)
Because of (6-33), the general expression for C given in (6-29) simplifies to

1
C =
P11-P22
477e0
4-irtn.ab
=
(
b
[ i _ i j
~ a
\ a
b I
with the use of (6-36) as well. Thus, we have been able to find the capacitance of this
particular capacitor by the evaluation of the general result (6-29) as stated in terms of
the coefficients of potential.
Nevertheless, in most simple cases, this is not the most convenient way to proceed.
W h a t is generally done is to find the potential difference f r o m a different and prior
solution to the problem. Ordinarily, this will include a knowledge of the electric field.
Then A<j> can be found by using (5-11) to integrate E over any convenient path between
the two conductors; when the two conductors f o r m a capacitor they are usually called
" p l a t e s . " Since we know from (6-27) that A(f> will be sure to turn out to be proportional
to the charge Q, the ratio of the two will give the capacitance according to (6-28); thus,
we have
=
(6-38)
* + plate
It is convenient to write the integral in this form because E will be generally directed
from the positively charged plate of higher potential to the negatively charged plate of
lower potential so that E
will be positive, thus giving the correct sign to A<f>. Let us
look at two examples from this point of view.
Example
Spherical capacitor. This is the same system shown in Figure 6-8 for which we found C
by the previous method. N o w let us assume the inner sphere of radius a to have the
positive charge Q\ then the inner sphere of radius b will have the charge Q in
agreement with (6-6). As before, the application of Gauss' law to a sphere of radius r
such that a < r < b shows us that E will be given by E = (Q/47re0r2)r
according to
(4-17). If we also integrate from the positive plate to the negative plate in the radial
direction so that ds = dr ?, then (6-38) becomes
t~
rb
f E ds = f
J+
Qdr
Ja 47re0r
( 1
I
477e0 \ a
1\
r )
b J
Q
C
(6-39)
which gives us the same result for C as we f o u n d before in (6-37); this time the result
has been obtained much more easily.
94
C O N D U C T O R S IN E L E C T R O S T A T I C FIELDS
Example
Parallel plate capacitor. This system consists of two conducting plates, each of area A,
which are parallel to each other and separated by a distance d that is small compared
to their linear dimensions. The plates need not be square, but any linear dimension
should be of the order of 4a so that, in effect, we are assuming that d
\j~A . A side
view of this capacitor is shown in Figure 6-9. Under these circumstances, the plates can
be treated as if they were of infinite extent, and we conclude in the same manner as for
the infinite sheet of Section 4-2 that E will be constant in magnitude and directed
normal to the plates as shown by the dashed lines in the figure. Since E is normal to the
conductor at the conductor, its magnitude is just that given by (6-4), that is, E = o/e0.
(This value for E is also exactly that found for the two plane sheets of Exercise 3-9 as
illustrated in Figure 3-6.) The simplest path of integration is clearly one along the
direction of E as shown by ds in Figure 6-9. Then E ds = Eds = (a/e0) ds and
(6-38) becomes
(6-40)
where we have written a = Q/A ; it is quite accurate to treat a as a constant in this way
because the plates are effectively of infinite extent, corresponding to a uniform surface
charge distribution. Solving (6-40) for C, we get the simple familiar result for the
capacitance
Incidentally, this system, as well as the spherical capacitor, shows us quite graphically
why capacitors are also often called "condensers." Because the field distribution is
confined to the finite region between the plates, one says that the electric field has been
"condensed" into this region, rather than extending throughout all space as has been
the case for many of our previous examples.
Now the plates in this case are not actually infinite, of course, and we have in effect
assumed that the electric field is constant and given by ( a / e 0 ) everywhere between the
plates and then drops off abruptly to zero at the edges. It will be left as an exercise to
show that this is impossible because of the conservative nature of the electric field. In
reality, the lines of E must curve outward somewhat as the edges of the plates are
approached, and they must also extend outward into the region beyond the plates as
indicated in Figure 6-10. However, if the plates are "large enough," we will not make
any appreciable error by neglecting these "edge effects"; it is customary to do so, and
we will continue to make this approximation.
Figure 6-9.
Parallel plate capacitor.
6-3
CAPACITANCE
95
Figure 6-10. General appearance of the

field of a parallel plate capacitor.
These two examples are probably enough to convince one that most of the problems
involving the calculation of capacitance by means of (6-38) must have a lot of
symmetry so that E can be easily found, usually by using Gauss' law. We will be able to
handle more complicated problems later when we will have discussed other systematic
ways of finding the potential ^ as a function of position.
EXERCISES
6-1 Suppose that the two conductors shown in
Figure 6-8 are originally uncharged. A charge Q
is then placed on the inner conductor of radius a.
Find the final static charge distribution. Find the
potential <}> for all values of r and plot your
result.
6-2 An infinitely long conducting cylinder of
radius a carries a total charge q, per unit length.
If p is the perpendicular distance from the axis of
the cylinder, show that the potential outside the
cylinder can be written
HP)
<7/ , I Po
ln
2 tre.
(p > a)
(6-42)
where p 0 = const. What is the potential inside the

cylinder? Verify that (6-4) is satisfied in this case.
Can one define an appropriate and unique pn for
this case?
6-3 The farad is actually an enormous unit of
capacitance. In order to illustrate this, treat the
earth as a conducting sphere of radius 6.37 X 10 6
meters and find its capacitance.
6-4 When the set of equations (6-13) is solved
for the charges, the result is another set of linear
equations of the form
Q, = L
(i = 1 , 2 , . . . , )
(6-43)
capacitance of this system can be expressed as

C =
c c
ll 22
12
n + C2 2 + 2c12
(6-44)
6-5 Using the results of the previous exercise,

find the coefficients c,- for the spherical capacitor
of Figure 6-8 and verify that they give the same
result (6-37) for the capacitance.
6-6 A capacitor C1 is charged resulting in a
potential difference A# between its plates.
Another capacitor C2 is uncharged. One plate of
C2 is now connected to a plate of Q by a
conductor of negligible capacitance; the remaining plates are similarly connected. For the resultant equilibrium state, find the charge on each
capacitor and the potential difference Aft between their respective plates.
6-7 The plates of two capacitors CX and C2 are
connected by conductors of negligible capacitance as shown in Figure 6-11a, that is, they are
connected in "parallel." If a potential difference
A<\> is now applied across the terminals T and T',
show that this combination is equivalent to a
single capacitor of capacitance CP = CX + C2.
Similarly, show that the equivalent capacitance of
the "series" combination shown in (b) can be
found from 1/Q. = ( 1 / Q ) + (1/C2).
6-8 Consider the spherical capacitor of Figure

6-8 for the case that a and b are nearly equal.
where the ctJ are combinations of the ptJ. If the Find an approximate expression for C and write
indices are the same, the c are called coefficients it in a form that explicitly involves the difference
of capacitance, and the ctJ with i = j are called S = b a where 8 <s: a or b. Interpret your recoefficients of induction. Find these coefficients sult with the aid of (6-41).
for the system of two conductors described by 6-9 The potential difference A<f> between the
(6-24) and verify that c21 = c12. Show that the plates of a spherical capacitor is kept constant.
96
C,
T'
O-
HHP
(b)
Figure 6-11. (a) Capacitors in

parallel. (b) Capacitors in series.
(a)
Show that then the electric field at the surface of

the inner sphere will be a minimum if a = \b.
Find this minimum value of E.
6-10 A capacitor is made from two infinitely
long conductors with coaxial cylindrical surfaces
as shown in Figure 6-12. Show that the capacitance of a length L of this system is given by
27TC0 L
C =
b
In | a
tion. Show qualitatively that this contradiction

will be removed for the general field lines shown
in Figure 6-10.
'
>
(6-45)
Why doesn't the result depend on c? Does the

inner conductor of radius a need to be a solid
cylinder?
1
1
' !
1
1
Figure 6-13. Assumed electric field for

Exercise 6-11.
6-12 A sheet of conductor of thickness t and
parallel faces of cross-sectional area > A is inserted between the plates of the capacitor of
Figure 6-9. The faces of the sheet are parallel to
the plates of the original capacitor. Show that the
capacitance is increased by
c0tA
AC
d( d t)
Why is this result independent of the distance of

the sheet from either of the original plates?
Figure 6-12.
plates.
Capacitor with coaxial cylindrical
6-11 Figure 6-13 illustrates the assumption we

made about E when we neglected edge effects for
the parallel plate capacitor, that is, we assumed
that E went abruptly to zero at the edges. Actually this is impossible because V X E = 0. Show
this by first calculating <E ds over the rectangular path shown, which is partly in the E
0
region and partly in the E = 0 region. Then show
that Stokes' theorem (1-67) leads to a contradic-
6-13 Suppose that a parallel plate capacitor has

rectangular plates but the plates are not exactly
parallel. The separation at one edge is d a and
d + a at the other edge where a
d. Show that
the capacitance is given approximately by
C =
enA
1+
3d2
where A is the area of a plate. (Hint: recall the

results of Exercise 6-7.)
6-14 Two conducting spheres have their centers
a distance c apart. The radius of one sphere is a,
the radius of the other is b. Show that when
EXERCISES
c a and c b, the capacitance of this system

will be given approximately by
1
C
4 TTCr
[Hints: imagine equal and opposite charges on

the spheres; how will one "look" at the location
97
of the other?; will (6-5) still be approximately

correct?]
6-15 Two infinitely long conducting cylinders
have their central axes parallel and separated by a
distance c. The radius of one is a and the radius
of the other is b. If c a and c b, find an
approximate expression for the capacitance of a
length L of this system.
We briefly touched on the question of energy as related to potential in Section 5-4
where we considered the potential energy of a single point charge and a pair of point
charges. Now we want to extend these ideas to a system containing an arbitrary
number of charges. We will calculate the potential energy Ue of this system, that is, the
amount of reversible work that would have to be done by an external agent in order to
produce the given configuration against the mutual conservative electrostatic forces of
the charges. After this has been done, we can apply it to a variety of situations, and we
will find it helpful to rewrite it in various ways.
ENERGY OF A SYSTEM OF CHARGES
Previously we found the expression (5-49) for the mutual potential energy of a pair of
point charges. If we now call these charges q, and
rather than q and Q, the potential
energy of this pair as given by (5-49) is
where Rtj is the distance between them and, as usual, Rt/ = |r. r,|. Now let us
consider a total number N of point charges in some given configuration so that the
individual charges and their distances from each other are all known. We number the
charges in some way so that i and j can both independently have the values
1 , 2 , . . . , N. The total potential energy of the system, Ue, will then be the sum of terms
like (7-1) and where the sum is taken over all the pairs of charges that can be formed
from the N charges:
v.-
UJ
(7-2)
all p a i r s
In this sum, we must exclude the case i = j for which a given charge is paired with itself
because that would imply that a charge would be exerting a force on itself, a possibility
that we already excluded in connection with (3-2).
It will be convenient to express (7-2) in terms of a double sum over the indices i and
j in which they independently cover their whole possible range of N values. However, if
we do that we will be counting each pair twice and must correct for this. For example,
consider the contribution of the pair q3 and q4 to (7-2) which is Ue34. In such a double
sum, this pair will occur once when i = 3 and j 4 and again when i = 4 and j 3
giving a total of Ue34 + Ue43 = 2U e 3 4 , according to (7-1), since R43 R34. Thus, we
will have to divide the result obtained from a double sum by 2; in this way, we can
write (7-2) as
1
uV = -J E
E U /L, Ls êij
e
i-l j = 1
j*i
i
Z2
i= l y=l
j*'
which is our desired result for the electrostatic energy.

98
qq
Am*
nR
(7-3)'
7-1
ENERGY O F A SYSTEM O F C H A R G E S
99
In particular, if all of the positions of the charges are given in rectangular coordinates, we can use (1-14) to get an explicit expression for Ue as
1
u
-~2
qq
2 E r
r
,
^
' = i y - i 4tT 0 [{ Xi - x j ) + (.y, - yj) + (z, - Zj) \
(7-4)
j+i
We can put (7-3) into another form by regrouping:

1
V. = xT.1,
2
,= 1
<7/
E T - V
^;-=1 4ff0n/y;
(7-5)
If we now recall (5-2), we recognize that the term in parentheses is just <, = 0,(r,), that
is, it is the scalar potential at the location of qt d u e to all of the other charges. Hence we
can write (7-5) as
N
Ue = \ H q M * i )
(7-6)
i=i
When dealing with the potential energy of a system, it is natural to ask oneself where
the energy is " s t o r e d " or "located." In mechanics, it is plausible to regard the potential
energy \kx2 of a stretched spring, for example, as being associated with the change in
relative positions of the strained portions of the spring so that the potential energy
would be " s t o r e d " at the locations of these parts of the spring. The form (7-6) that we
have obtained lends itself to a similar interpretation. Since the contribution to the sum
from a given charge involves only the value of the scalar potential at the location of
that charge, one can, if one wishes, associate a portion of the potential energy with this
charge and think of it as being "located" there.
If the charges are continuously distributed, we can rewrite the sum in (7-6) as an
integral by using (2-14) and (2-16). In this way, we get the following expressions for the
energies for volume, surface, and line charge distributions, respectively:
Ue=
\ Jf p(r)<l>(r) dr
v
Ue = \[o(r)<t>(r)
J
s
da
\jHM(r)ds
(7-7)
(7-8)
(7-9)
The integrals are to be taken over all regions containing charge of the particular type
involved.
If the charge distribution is such that it consists of some or all of these possibilities
considered above, then the total energy of the system will be a sum of quantities
obtained f r o m (7-6) and (7-7) through (7-9); for each part, the potential cf> must be
found as the resultant from all of the charges in all of their various distributions.
Actually, the integrals above can have their region of integration extended f r o m only
the volume actually occupied by the charges to all space, since in regions where there is
no charge, p = 0, and these regions will contribute nothing to (7-7). Thus, we can write
(7-7) equally well as
U. = \ f
'all space
with similar expressions for (7-8) and (7-9).
p(r)#>(r) dr
(7-10)
100
Example
Uniform spherical charge distribution. As an illustration of the use of (7-7), let us

consider the case of a sphere of radius a containing charge with constant volume
density p. We considered this example in Section 5-2 and the potential <f> is given
everywhere by (5-22) and (5-23). However, we need only the potential inside the sphere,
according to (7-7), and this is given by (5-23) as
* ( r ) = - ( 3 a 2 r2)
(7-11)
6e0
If we substitute this into (7-7), use (1-99), and take all of the constants (including p)
out of the integral, we get
2
Ue=
f 2 7 7j f
1 2 c 0 Jq
f ( 3 a 2 r2)r2
O jO
sin 9 drdd dcp
(7-12)
The integral over the angles gives 4 77, so that

r

.
4 77p2a ^
4
Ue = f (3a r - r ) dr =
3c0 0
15c 0
(7-13)
If we express this in terms of the total charge of the sphere by using Q = (4-rra3/3)p,
we get
<7
-i4)
which is smaller than the energy of two point charges Q that are separated by a
distance equal to the radius of the sphere.
ENERGY OF A SYSTEM OF C O N D U C T O R S
Because of their special properties some of our results can be written in simplified form
in the case of conductors. As we found in Section 6-1, the charge on a conductor is
confined entirely to its surface. In addition, the potential is constant on the surface,
according to (6-2), so that <f> can be taken out of the integral in (7-8). If we now apply
(7-8) to the /th conductor of our system of n which we considered in Section 6-2, we
find its energy to be given by
Uei = \$i Jf o, da, = 2 Q^,
s,
(7-15)
where
is its total charge as given by (6-16). The total energy of the system will be
given by a sum of terms like (7-15) so that
Ue = j t Q&i
/=i
(7-16)
The energy can also be expressed completely in terms of the charges if we use (6-11):
ue = \ t
PijQ.Qj
(7-17)
/-17=1
In the special case in which there are only two conductors, this becomes
Ue
with the use of (6-25).
\PwQ\
P\iQ\Qi
2P22Q1
(7-18)
7-3
E N E R G Y IN T E R M S O F THE ELECTRIC FIELD
101
These results can be written in an even more compact and useful form in the cases in
which the system can be described in terms of the single parameter, the capacitance C.
Example
Isolated conductor. In this case, (7-16) reduces to the single term \ Q<j>; if we also use
(6-22), we can write
Q2
e = iQ* = ^
= K02
(7-19)
If we apply this to a conducting sphere by using (6-23), we find its energy to be

u
e = r (-T)
2 1 477e0a
(7-20)
which is less than the energy given by (7-14) for the case in which the same charge is
distributed throughout the volume of the sphere rather than being confined to its
surface as is the case for (7-20).
Example
Capacitor. Here we have equal and opposite charges on the plates. Thus, if we let
Q} = Q and Q2 ~Q i n (7-18), and use (6-29) and (6-28), we find the general
expression for the energy of a capacitor to be
Q2
Ue = \QA<t>=
= \C(M)2
(7-21)
which is very similar to (7-19) for a single conductor. We will find this result to be very
useful in subsequent discussions.
ENERGY IN TERMS OF THE ELECTRIC FIELD
As we briefly remarked in the paragraph following (7-6), our energy expression easily
lends itself to the interpretation that the energy of the system is directly associated with
the charges and their positions. This is a point of view that is naturally consistent with
the action at a distance property of Coulomb's law with its emphasis on the charges
and relative positions. On the other hand, our interest is primarily in describing
phenomena in terms of fields, and we would like to do the same for the energy. It will
be appropriate for us to use (7-10) as our starting point.
We can write p = e0V E, because of (4-10), so that (7-10) can also be written as
ue= y/4>(v-E)</T
(7-22)
We can rewrite the integrand by using (1-115), (5-3), and (1-17):

<|>(V E) = - E - ( v ^ ) + V (<*>E) = E 2 + v (<#E)
(7-23)
When this is substituted into (7-22), we get

Ue = y Je2
dr + y Jv-
(<j>E) dr
(7-24)
and if we use the divergence theorem (1-59) to write the second integral as a surface
102
integral, we finally get

Ue = y Je2
dr + y <fi(4>E) da
(7-25)
Now we have to stop and consider the fact that our starting integral in (7-10) was to be
taken over all space; in order to do this, we shall find it best to start with taking V in
(7-25) to be an extremely large volume, so that S is its very large bounding surface.
Then we can let V become infinite; in this process, the surface S will also recede out to
infinity. We assume that the charge distribution is always contained within a finite
volume that, however, may be very large. Then V can always be chosen large enough
initially to enclose all of the charges. Now as we let V get very very large, S gets so far
away that the charge distribution will appear to be contained in such a small volume
that at any point on S the whole charge will seem to be a point charge a distance R
away. Thus, as far as the integral in (7-25) is concerned, as R -> oo, we see from (5-12)
and (5-13) that
ll - F
* - ^
(*E) - ^
(7-26)
so that the magnitude of the integrand is decreasing as R 3 . The surface of integration is

increasing, however, but it will be proportional to R2; thus, for very large R,
$(+E)-d*~
r 2
~ j
^ 0
(7-27)
so that when V in (7-25) increases to include all space, the surface integral will vanish
and the energy expression becomes simply
Ue= (
Ê
Jt
(7-28)
all s p a c e ^
where E 2 = E E. Thus, we have succeeded in expressing the energy completely in

terms of the electric field. [Although the charges no longer appear explicitly in (7-28),
they haven't disappeared, of course, because they are the sources of the electric field.]
The form of (7-28) as a volume integral, in which regions where E # 0 contribute to
the integral while those where E = 0 do not, lends itself to a simple and natural
interpretation: the electrostatic energy is distributed continuously throughout space with
an energy density ue given by
ue = i e o E 2
(7-29)
so that the total energy can be written

Ue=
uedr
(7-30)
âll space
The units of ue will be joule/(meter) 3 .

We can certainly make this interpretation but we are not forced to do so, nor is
(7-29) a unique possibility. For example, we could add to ue any quantity x whose
volume integral over all space is zero and still be in agreement with (7-28). However,
(7-29) is simple, plausible, and esthetically very appealing from the view of expressing
things in terms of fields. The accepted point of view is to adopt (7-29) as correct; we
will be able to justify this interpretation in more detail later, and, in fact, (7-29) turns
out to be useful and accurate even when time varying fields are considered.
The energy as calculated from (7-28) should, of course, agree with that obtained
from (7-6) or its variations. Let us check this in a particular case for which we have
already found E as well as Ue.
7-4
ELECTROSTATIC FORCES O N C O N D U C T O R S
103
Example
Isolated conducting sphere. We found the solution for this case in terms of the potential
4> given by (6-5). Thus the electric field produced by this system will be
Or
E =
r
47re 0 r 2
(r > a)
E = 0
(7-31)
(r < a)
so that there is no energy contained within the volume of the sphere, according to
(7-29), and the region of integration in (7-30) reduces to all space outside the sphere.
Using (7-31) in (7-29), the energy density ue is found to be
Q2
Ue
~ 327r 2 e 0 r 4
(r ~
ue = 0
a>
>
(7_32)
(r < a)
so that, with the use of (1-99), (7-30) becomes

Ue =
Q~
z2
tlm
32tT 0-'o
z*00 1
[ r2 s i n 0 drdO dtp =
Jo J a
Q2
%7T0a
(7-33)
in complete agreement with (7-20).
A useful application of (7-28) is for the calculation of capacitance. If one has found
E by other means, then one can evaluate (7-28). We know that in such a case Ue will
turn out to be proportional to Q 2 or <f>2 or (A<>)2, depending on what is given, so that
when Ue is found in this way, it can be immediately used in (7-19) or (7-21) to evaluate
C. This is often a more convenient procedure than is the use of (6-38).
Example
Parallel plate capacitor. This system was illustrated in Figure 6-9 and we found the field
to have the constant magnitude E = o/c0 between the plates and zero elsewhere.
Therefore,
u
i c
Q2
(7
-34)
and is constant within the volume Ad between the plates and zero elsewhere. Inserting
this into (7-30), we get
/
= juedr
M e (volume) -
r\
Q2
2e0A2 j
(Ad)
Q2d
Q2
2e0A
2C
(7-35)
which leads again to the value for C given by (6-41).
In general, two charged conductors will exert forces on each other and, in principle,
these forces can be calculated from Coulomb's law. It is often desirable to evaluate
these forces in a different way. A similar situation is often encountered in mechanics
when one finds it convenient to calculate force components as spatial rates of change of
104
E L E C T R O S T A T I C ENERGY
the potential energy. In addition, energy considerations provide one with another way
of looking at a situation and thus can help in broadening one's overall understanding.
We want to develop a similar approach to electrostatic forces; however, in this
section we restrict ourselves to conductors. Furthermore, we will not do this in
complete generality, but consider only the particular case of the parallel plate capacitor.
This will suffice to illustrate all of the general features of the problem in terms of an
easily visualized and simple system. It will turn out, nevertheless, that our final results
will be stated in such a manner that they can be seen to be actually of very general
applicability.
In Figure 7-1, we show a parallel plate capacitor of capacitance C. By means of the
switch S, the terminals T and T' can be connected by conducting wires to a battery B
that will provide a difference of potential
between the plates, thus charging the
capacitor to the value Q = CA<#>. Because of their opposite charges the plates will
attract each other so that there will be an electrostatic force Fe on the positive plate
with the direction shown. In order to have the system in equilibrium, Fe must be
balanced by an equal and opposite mechanical force Fm. Our aim is to find these forces
from energy considerations.
We let x be the separation of the plates. N o w let us imagine that this separation is
very slowly changed by the amount dx. Under these conditions, the work done on the
system by the mechanical force will be given by Fm dx and since this will be reversible
work, as we saw in Section 5-4, it will be equal to the change in the total energy dUt of
the whole system, that is, dUt = Fm dx. But since the acceleration is always nearly zero,
the capacitor plate will remain in equilibrium, or be only infinitesimally different f r o m
it, so that we will continue to have Fe= Fm, and therefore
dUt
dx
(7-36)
We have written Ut since the capacitor is not the whole system by itself; in general the
battery must be included because its energy may also change as a result of the work
done by Fm depending on whether the switch S is open or closed. In other words, the
capacitor is not necessarily an isolated system. In this sense, our problem is similar to a
situation often encountered in thermodynamics in which a thermodynamic system of
interest may be in contact with heat and work reservoirs. There the problem is one of
Q
T
F,m
c
x-
Figure 7-1. Forces on a plate of a

parallel plate capacitor.
7-4
105
trying to state the thermodynamic criteria of equilibrium, which refer to isolated

systems, in terms of a function characteristic only of the nonisolated system of interest,
that is, in terms of only part of the whole composite system, but that part with which
we are directly concerned. In our case, we want to write (7-36) in terms of the capacitor
only, if possible.
In our derivation of (7-36), we imagined a small relative displacement of the plates.
However, the force Fe is a definite and measurable quantity and hence must be
independent of the particular process that we visualize in order to do the calculation. In
this example, there are really only two different ways in which we can carry out the
infinitesimal change; that is, the switch is either open or closed. Thus, there are two
possibilities to be considered and it is instructive to discuss them separately. As we will
see, the details of what is going on differ in the two cases, but we should remember that
we must get the same value for the force as a final result.
1. Constant charge. Here we imagine that 5 has been closed until the capacitor is
charged, and then S is opened and kept open. This effectively removes the battery B
from the total system since it will be unaffected by the displacement dx; in other words,
the capacitor is isolated. In this case, dUt = dUe where Ue is the capacitor energy and
(7-36) becomes
F
e = - [
^ )
( G = const.)
(7-37)
Now Ue = Q2/2C according to (7-21), but since C is a function of the plate separation,
C = C(x), and therefore
e
dU
<
~ 2C1
d C
(?
"38)
which makes (7-37) become

Q1 dC
Fe =
2 C dx
(7-39)
[In this process, the potential difference between the plates must change in order that
Q = CA< remain constant. Since this leads to dQ = 0 = (dC)( A<p) + Cd(A<p), we find
d(A4>)
dC
-
=
A 4>
C
(7-40)
'
which says that the fractional change in potential difference is equal and opposite to
that of the capacitance. In fact, since C will decrease as x is increased, according to
(6-41) (with x replacing d), A<$> will increase as x is increased. The reason for this is
that with Q = const., a will also be constant, as will E = c / e 0 ; then, with the same E
applying over a greater separation, A(j> = J + Eds = Ex will also be greater.] Before we
discuss (7-39) in more detail, let us consider the other possibility.
2. Constant potential difference. This corresponds to keeping the switch S closed so
that A(j> = const. The battery is then part of the complete system, and the capacitor is
no longer isolated. In this case, the total energy change will be the sum of the energy
change of the capacitor, dUe, and that of the battery, dUB, so that
dUt = dUe+dUB
(7-41)
Here it is convenient to write Ue in the form Uc = \C(Acj>)2, which is also given by

(7-21). Then, we have
dUe=
j(dC)(A<j>)2
(7-42)
106
E L E C T R O S T A T I C ENERGY
[If dx is positive, dC will be negative, as noted above, so that the capacitor energy Ue
will decrease under these conditions, in contrast to its increase in the constant charge
case as described by (7-38).]
Since Q = C A</>, the capacitor charge will change a n d this change is given by
dQ = {dC)(A<f>)
(7-43)
As these charges pass slowly through the battery potential difference, work will be done
on them and the battery energy will change. We f o u n d in (5-46) that this work equals
the charge times the potential difference, and since any work done by the battery
represents a decrease in its energy, we get
dUB=
-dW=
= -(dC)(A<j>)2
-(dQ)(A<j>)
= - 2 dUe
(7-44)
with the use of (7-43) and (7-42). Thus, the energy change in the battery is always
opposite in sign to that of the capacitor and twice as great in magnitude. [If dx is
positive, we saw above that dUe will be negative, so that dUB as given by (7-44) will be
positive, that is, the battery energy will actually increase. The reason is that dQ will
also be negative from (7-43) since dC is negative; as these charges return through the
battery they will return to it the reversible work that the battery originally did on them
in the original charging process.]
When we substitute (7-44) into (7-41), we find that dUt = dUe in this case so that
(7-36) becomes
Fe = +
/ dU e
\ dx J
(A<> = const.)
(7-45)
which when combined with (7-42) and (6-28) gives

?dC
F
Q2
dC
- - W * - 2 C * *
(7
"46)
which, as it must be, is exactly the same as (7-39).

It is perhaps well to emphasize again that the basic equation for calculating the force
is (7-36) and that the apparent discrepancy in signs between (7-37) and (7-45) arises
from the fact that these expressions refer to different processes. In the case of (7-37), the
capacitor was an isolated system and it underwent the only energy change. In the case
of (7-45), the capacitor was no longer an isolated system but we were still able to
express the force in terms of the capacitor energy alone. This is all quite analogous to
the thermodynamic situation in which one can go f r o m the characterization of the
equilibrium of a system of interest by means of its internal energy to the use of its
Helmholtz or Gibbs function for the cases in which it is in contact with heat and work
reservoirs. This can be done by using only generalized properties of these reservoirs as
large featureless systems. We note that here too we did not need to know any of the
detailed internal workings of the battery; it was sufficient to know that it is a device
that is somehow capable of doing reversible work on charges.
As a matter of fact, when we review our work that led to (7-39) and (7-46), we see
that, except for the parenthetical illustrative remarks we made along the way, we really
did not use any detailed result that specifically required our system to be a parallel
plate capacitor other than that the displacement was parallel to the mechanical force so
that the element of work could be written as the simple product Fm dx. Now, however,
it is necessary to introduce these details.
If we let x be the plate separation, C = e0A/x, according to (6-41). Then dC/dx =
0A/X2 = C/x, and (7-39) becomes
F
Q2
= - f
2Cx
ue
= ~ x
(7-47)
7-4 ELECTROSTATIC FORCES O N C O N D U C T O R S
107
with the use of (7-21). This shows that Fe is negative, agreeing with the fact that the
oppositely charged plates of Figure 7-1 will attract each other. We found in (7-35) that
Ue can be written as the product of the energy density ue and the volume Ax between
the plates so that Ue = ueAx; putting this into (7-47), we find that
Fe = ueA
(7-48)
Since the total force is proportional to the area, it is convenient to introduce the force
per unit area, fe, as equal to the magnitude of this ratio; thus, we get
fe = - J - =
/
| j F e l
(7-49)
m
We see from Figure 7-1 that the direction of this force is outward from the conducting
surface so that we can write the force per unit area as
fe = fen = uen
(7-50)
where n is the outward normal to the conductor surface as shown in Figure 7-2. Thus
we have found that there is a tension or outward force per unit area on the conducting
surface that is numerically equal to the value of the energy density at the surface.
Although this result was obtained by considering the specific case of a parallel plate
capacitor, we can now show that it is actually of general validity. Consider a portion of
the conductor surface as shown in Figure 7-3. Inside the conductor where E is zero, the
energy density ue is also zero. Now let us imagine that a small portion of the conductor
surface of area A a is given a small displacement Ax perpendicular to the surface. The
volume of the region where ue = 0 has been increased by Aa Ax so that the overall
energy will have been changed by the amount AUe where
AUe = -ueAaAx
(7-51)
which is negative if Ax is positive. This energy change corresponds to a force AF e on

this area element given by AF e = AUJAx = ue Aa. Since AF e is proportional to the
area A a, we can once again introduce a force per unit area fe = AFJAa, which again
turns out to be equal to ue in agreement with (7-49). In order to determine its direction,
we recall from mechanics that the states of stable equilibrium for general systems
correspond to the configurations resulting in a minimum value of the potential energy,
that is, the natural tendency of systems is to "attempt" to decrease their potential
energy. As we saw above, a positive displacement Ax would lead to a decrease in the
energy Ue\ since this is what the system " w a n t s " to do, this will determine the direction
of the force to be outward from the conductor. In other words, we are led once more to
(7-50) that gives the electrostatic force per unit area as always a tension, that is, in the
Figure 7-2. Force per unit

area on the surface of a
conductor.
Figure 7-3. Calculation of the force on the

surface of a conductor. An element of the
surface is imagined displaced by Ax.
108
d i r e c t i o n of the o u t w a r d n o r m a l n. If t h e i n t e r n a l cohesive forces of the c o n d u c t i n g

m a t e r i a l are n o t large e n o u g h to c o u n t e r b a l a n c e this electrostatic force, the c o n d u c t o r
will d e f o r m . T h i s d e f o r m a t i o n will c o n t i n u e u n t i l t h e elastic f o r c e s t h a t are b r o u g h t i n t o
b e i n g will b e large e n o u g h to k e e p the s u r f a c e in a n e w e q u i l i b r i u m situation.
If w e c o m b i n e (7-49) w i t h (7-29) a n d (6-4), w e c a n express fe variously as
CT2
= \oE
ze0
fe = ue=^QE2=
(7-52)
w h e r e E a n d a m u s t b e evaluated at t h e p a r t i c u l a r p o i n t o n t h e s u r f a c e t h a t is b e i n g
considered.
If w e m u l t i p l y (7-50) b y da, w e will get t h e f o r c e o n this t o b e fe da = fen da = fe da
so t h a t t h e total force o n t h e c o m p l e t e s u r f a c e S of a given c o n d u c t o r will b e given b y
f / . d t -
^-fo2dz
2eQ
(7-53)
a n d t h a t c a n b e used once the s u r f a c e c h a r g e d e n s i t y h a s b e e n d e t e r m i n e d as a f u n c t i o n

of p o s i t i o n .
It is a very c o m m o n error t o t h i n k t h a t t h e f o r c e p e r u n i t a r e a is simply oE r a t h e r
t h a n \aE as correctly given b y (7-52). Also, o n e s h o u l d n o t f o r g e t t h a t (7-53) is a vector
e q u a t i o n . [ F o r example: w h a t is t h e total f o r c e o n t h e c o n d u c t i n g s p h e r e w h o s e electric
field is given b y (7-31)?]
EXERCISES
7-1 Consider a square of edge a. Starting at one
corner and proceeding counterclockwise, one puts
a point charge q at the first corner, 2q at the
next, then 3 q, and finally 4 q. Find Ue for this
charge distribution.
7-2 A point charge q is put at each corner of a
cube of edge a. Find the electrostatic energy of
this system of charges.
7-3 The expression for the energy of a capacitor
given by (7-21) can also be obtained in the following manner. Consider an intermediate stage when
the charge is q where 0 < q < Q. The potential
difference will be q/C. Find the work required to
increase the charge by dq. Then add all these
work increments from the initial uncharged state
to the final completely charged state and thus
obtain (7-21) again.
7-4 Find the energy of the charge distribution of
Exercise 5-9 by using (7-10). To what should your
j esult reduce when n = 0? Does it?
7-5 Find the energy of the charge distribution of
Exercise 5-17 by using (7-8).
7-6 Find the energy of a length L of the coaxial
cylinders of Figure 6-12 when they are used as a
capacitor with charge q/ per unit length by using
(7-8). Use your result to verify again the value
(6-45) for C.
1-1 Find the total gravitational energy of the

earth by treating it as a homogeneous sphere of
mass 5.98 X 10 2 4 kilograms and radius 6.37 X
10 6 meters. [The gravitational constant is G =
6.67 X 1 0 " 1 1 newton-(meter) 2 /(kilogram) 2 .] If a
uniform spherical charge distribution whose total
charge equaled the magnitude of the electronic
charge (1.60 X 1 0 " 1 9 coulombs) had the same
energy, what would its radius be?
7-8 Express the energy of a system of n conductors in terms of the potentials and coefficients of
capacitance and induction defined by (6-43). Show
that when there are only two conductors the
energy can be written as
Ue = 2 C 1 1 0 1 + c 1 2 0 1 0 2 + 2 C 2 2 0 2 (7-54)
and that when these two conductors are used as a
capacitor the energy is again given by (7-21) and
(6-44).
7-9 Show that when (7-28) is applied to the case
of the uniform spherical charge distribution the
result is again (7-14). What fraction of the total
energy is now regarded as being outside of the
sphere?
7-10 A charge Q is on the inner sphere of
Figure 6-8 and a charge Q on the outer sphere.
Find the energy of this system by using (7-28)
and thus show that the capacitance is (6-37).
EXERCISES
7-11 The coaxial cylindrical conductors of Figure 6-12 are used as a capacitor with charges per
unit length q: and q,. Find the energy of a
length L of this system by using (7-28) and thus
show that the capacitance is (6-45).
7-12 Use (7-28) to find the energy of the charge
distribution of Exercise 5-9 and verify that you
get the same result as for Exercise 7-4 and that
your answer reduces to the correct result when
n = 0. What fraction of the total energy is outside
the sphere?
7-13 Consider the two conductors of Figure 6-8.
There is a total net charge Qx on the inner
conductor and a total net charge 0 2 on the outer
conductor where 0 X
0 2 . Show that the total
energy of this system is given by
U =
8 mir
1
1
1
- " T + - IQ
a
20102
Qi
(Show that this reduces to the expected result

when 0 2 = 0].) Do this exercise in three different ways.
7-14 A parallel plate capacitor has plates of area
A. The lower plate is rigidly fastened to a table
top. The upper plate is suspended from a spring
of spring constant k whose upper end is rigidly
fastened. The plates are originally uncharged.
When the capacitor is charged to a final charge 0
and 0 show that the distance between the
plates changes by an amount Q2/2ke0A.
Is the
spring stretched or compressed?
7-15 A square metal plate of edge 20 centimeters is suspended from the arm of a balance so
that it is parallel to another fixed horizontal metal
plate of the same dimensions. The distance of
separation between the plates is 1.5 millimeters. A
difference of potential of 150 volts is now applied
between the plates. What additional mass must be
109
placed in the other arm of the balance so that the

suspended plate will retain its original position?
7-16 A battery is used to charge a parallel plate
capacitor to a potential difference A<#> and is then
disconnected. The separation between the plates
is now increased from d to ad where a is a
constant > 1. What is the ratio of the new energy
to the original energy? Is the energy increased or
decreased? Where does this energy change come
from or go? Verify your answer quantitatively.
7-17 There is a potential difference A<J> between
the coaxial conductors of Figure 6-12. Find the
magnitude of the force per unit area on the surface
of the inner cylinder. What is its direction? What
is the total force per unit length on it?
7-18 A parallel plate capacitor is formed by the
metal bottom and metal movable piston of a
cylinder with nonconducting walls. The cylinder
is airtight and is kept at constant temperature.
W h e n the capacitor is uncharged the separation
of the plates is d0 and the pressure inside the
cylinder is p0. When a potential difference A<j> is
applied between the plates, show that, if / is the
fractional decrease in plate separation ( / > 0), it
can be found from
/ ( 1
_ / )
it2/>o I\ "ot
7-19 Suppose the two coaxial cylindrical conductors of Figure 6-12 are kept at a constant
potential difference A<. The cylinders are very
long. N o w imagine the inner conductor to be
given a small displacement A * in the direction of
the common axis. Show that the inner conductor
will be attracted back to its original position by a
force which is given approximately by
r
___
( A ( fr) 2
In | a
8
ELECTRIC MULTIPOLES
Equation 5-2 provides us with a method by which we can, in principle, find the
potential due to any given distribution of charges at any field point of interest. Let us
suppose that the charges are contained in a finite volume of some reasonable size. The
volume need not be of any particularly simple shape but, in fact, may be quite
irregular. If we are near such a volume, we can expect that the values of the potential at
different points can be quite sensitive to the details of the charge distribution. However,
as we get farther and farther away, it seems clear that the finer details of the charge
distribution will become less and less important and that the potential will be
determined primarily by the larger scale variations in the locations and magnitudes of
the charges. The extreme case would occur when we are so far away that the charge
distribution would appear to act as if it were simply a point charge; we have already
used this in connection with (7-26) and (7-27). What we want to do now is to consider
this situation more carefully, and we will find that the effect of the charge distribution
can be characterized by a set of quantities that depend on different details of the charge
distribution, much as mechanical quantities such as the magnitude of the total mass
and the moment of inertia of a set of mass points depend on different features of the
mass distribution. These quantities are called electric multipoles and we will give them a
specific definition. These considerations will also be helpful to us when we face the
problem of describing the effects of matter in electrostatics, since, for our purposes, we
will regard a piece of matter as essentially a collection of electric charges with some sort
of distribution.
HT1
THE MULTIPOLE EXPANSION OF THE SCALAR POTENTIAL
The general situation is illustrated in Figure 8-1. We have a system of N point charges
qlf q2, , <7,, , qN located in some volume V. We choose an origin of coordinates 0
in some arbitrary way, but for convenience, either in or near V. The position vectors of
the charges are r1? r 2 , . . . , r , , . . . , rN. We want the potential 4> at the field point P whose
position vector with respect to this same origin is r; thus, P is in the direction r at a
distance r from 0. This potential is given by (5-2) as
( H )
where R,, = |r - r, |. If we introduce the angle 6i between the directions of r, and r, and
use the law of cosines, we see from the figure that
R. = (r2 + r2 - 2#rf cos 0,) 1 / 2
(8-2)
so that (8-1) becomes

<Kr)
110
1/2
22 t
a \
; = l 4tT 0 (r + r - 2rr, cos 6,)
2
(8~3)
8-1
Figure 8-1.
THE M U L T I P O L E E X P A N S I O N O F THE S C A L A R P O T E N T I A L
111
Geometry for calculating the potential due to a system of point charges.
However, we can do nothing more specific with this without knowing the exact details
of the charge distribution.
Now let us assume that P is far enough outside V so that it is farther from the
origin than any charge. Then r > rl for all i, the ratio ( r t / r ) is always less than unity,
and we can consider a power series expansion in this ratio. If we factor out r 2 from the
square root in (8-2), we can write
1
1
(8-4)
r ( l + t A) 1 / 2
where
rt
t = 2| I cos 6,
(8-5)
We now use the power series
33
-Z-t
(l + t) ( 1 / 2 ) = l + \ t + 18 1' 2 + -ht
+ .
(8-6)
16'
with the upper sign, to expand the square root in (8-4). We will keep all terms of the
order of ( r t / r ) 2 and neglect the rest; thus we do not need to use the fourth term in
(8-6), which involves t 3 and hence (r ; /V) 3 . Inserting (8-5) into (8-6), and dropping all
terms involving (r:/r)3 and (/-,/r) 4 , we find that
1
1
+
-2l^)cos0, +
- 2 1 7 ) cos et+ i-r
1/2
8
*~2
(1 + o
" c o s S , + 1 ^ ) 2 ( 3 cos
1)
r/
'
2\ r
If we divide this by r, according to (8-4), and substitute the result into (8-1), we find
that
= 1+
<p(r) =
4m
or
1
L <7, + ^ 7 r c r 22 E q,r, cos 0.

4 o ,=i
/=1
r :
4m0 '
(3 cos 2 Ql: l ) + . . .
i= 1
(8-7)
112
ELECTRIC M U L T I P O L E S
where we have indicated by the . . . that there would be other terms of higher order, but
ones which we have not calculated. The result (8-7) is called the multipole expansion of
the potential. The individual terms in the sum are called, respectively, the monopole
term, the dipole term, and the quadrupole term. We see that their dependence on the
field point distance r goes successively as 1 /r, 1 /r2, 1/r3, and so on, so that as we get
farther away from the charge distribution, the higher-order terms in the expansion
become less and less important. For convenience in later discussion, we write the sum
(8-7) in the form
0 ( r ) = 0A/( r ) + 0 o ( r ) + 0 e ( r ) +
(8-8)
Although it is not necessary for our further considerations in this chapter, it is of
some interest to note that the functions of the angles that have arisen in (8-7) are
known functions called Legendrepolynomials. If these functions are written P,(x\ they
are defined by the expansion
00
1
rTTT2=HPi(x)y'
(|X|<1,^<1)
(8-9)
(l 2xy +y)
i=o
so that they are the coefficients of y' in the sum. A few of them are
P0(x)
= 1
^(x) =x
P3(x)
= H5*3 - 3x),...
P2(x)
= i(3*2
- 1)
(8-10)
Once the first few are known, the others can be found by means of the recursion relation
that they can be shown to satisfy:
(/ + 1 )Pl+1(x) = (21 + 1 )xP,(x) - lP,^(x)
(8-11)
We also note that -P/(l) = 1If we compare (8-9) with (8-4) and (8-5), we see that we can make the identifications
y r y r a n c j x = cos 0,', both of these satisfy the parenthetical conditions given in
(8-9). Then we can write
1
1 00
/ r \'
= - ! > ,/ (v * * , ) -f
(8-12)
R,
' ~ 0
"W
so that (8-1) can be written in general as
N
I
oo i
r
E
q.r/P,(cos 0, )
<K ) = T z r D /.+1
/=0 r
i= 1
(8-13)
whose first few terms agree with (8-7), which we obtained by a straightforward
expansion. Although (8-13) gives us a complete expression for the general expansion of
tj>, we will need to consider only that part given by (8-7).
Our results are still somewhat inconvenient for our purposes because the quantities
appearing in the sums still involve both the location of the field point P and the
locations of the charges because of the appearance of cos Bt. It would be nicer if we
could write the terms of (8-7) and (8-8) in a form in which these two sets of variables
appeared separately and explicitly, preferably in the form of a product involving
something that depended on the location of the field point alone and something that
involved only the charge distribution. This will be done next, but we need a different
way of writing cos 6,. We see from Figure 8-1, (1-15), (1-97), (1-8), and (1-20) that
c o s 6 ,
<
/ = =
r*
/ r< \
" (-) =
*xi
'yy-
zzi
(8-14)
8-1
113
where lx, ly, and lz are the direction cosines of the position vector r of P, and x0 yn
and zi are the rectangular coordinates of the location of the charge qr It will be
convenient to look at the terms in (8-8) one by one.
1. The Monopole Term
The sum in the first term of (8-7) is easily identifiable. It is
N
X Qi = Qtotal = Q
/= 1
(8-15)
where Q is the net charge of the system. Thus, the monopole term has the form
MO = ~T
4tre 0 r
(8-16)
Since this is the dominant term in the potential, when we are very far away, we see that
the whole distribution will act as if it were a point charge as we have already concluded.
In this context, the net charge Q is called the monopole moment of the charge
distribution. In other words, the monopole moment is that feature of the charge
distribution that is important for the monopole term.
If the charges are continuously distributed, then the sum can be replaced by an
integral by means of (2-14) so that the monopole moment can be found from
Q = f p(r') dr'
Jy
(8-17)
where the integral is taken over the volume V' of the source charge distribution. For
surface and line distributions, there will be similar expressions obtained with the use of
(2-16).
2. The Dipole Term

If we insert (8-14) into the second sum in (8-7), we get
N
(=i
i=i
E q,n cos el = X ^(/A
+ iyyt + izZi)
= ix \
+ iy( E ^ j ,
+ M
= f-|Etf,r,j
(8-18)
Since the sum in parentheses in the last form involves only the properties of the charge
distribution and does not involve the location of the field point, it is an individual
property of the charge distribution only. It is defined as the dipole moment p of the
charge distribution; that is
N
p = E qpi
(=i
(8-19)
so that we can write

N
q,rcos 6,, = r p = lxPx

i=i
+ lyPy + lzPz
(8-20)
114
When this is inserted into (8-7), we find that the dipole term can be written in terms of
the dipole moment as
p ?
p r
(8 21)
"o ( r ) - 47T r
"
0
We note that (8-21) has the form of a (scalar) product of quantities, one of which
depends only on the location of the field point and one that depends only on the details
of the charge distribution.
If the point P is very far away and if the monopole moment Q vanishes, then (8-21)
will be the leading term in the expansion of 4> and the dipole moment p will be the
dominant feature of the charge distribution.
If the charges have a continuous distribution, then the sum (8-19) can be replaced by
an integral over the volume V so that p can be found from
P = J[ y p(r')r' dr'
(8-22)
and there will be similar expressions for surface and line distributions.
3. The Quadrupole Term

This term is more complicated, but it can be written in a desirable and convenient form
in a reasonably straightforward manner. If we use (8-14), we find that
r, 2 (3cos 2 0, - l ) = 3(r r,) 2 - rf
= 3(1Xx, + lyyt + lzzt)2
- r2(l2x + I2 + l2z )
(8-23)
In the last step, we have actually multiplied r 2 by 1 because of (1-9) and hence have
not altered its value. When we multiply out the square in (8-23) and group the terms we
obtain
r 2
/ (3 cos 2 0, - l ) = l2(3xf
- r, 2 ) + /2(3j>,2 - r,2)
+ / 2 (3z 2 - r2) + 6 l x l y x i y i + 6lyl7ylzl

+ 6lJxziXi
(8-24)
We now insert (8-24) into the sum in the third term of (8-7) after factoring the \ out of
it; we also divide the last three terms of (8-24) by noting that 6lxlyxiyi = 3lxlyxjyi +
3lvlxylxlWhen we do all of this, we find that the sum can be written in a nice
symmetrical form as follows:
E ^ , 2 ( 3 c o s 2 0 , - l)
i
= /2E4/(3*,2
- r?) + IJyY.qilxM
+ IJzYthlx.Zi
- r2) + iyizY,Q^y,z,
i
3z
+ hlyî iyi
z
- r,2)
(8-25)
We note that each term in this expression is a product of something that depends only
on the field point, that is, its direction, and a quantity that depends only on the details
of the charge distribution. Accordingly, we define a set of quantities Qjk, which are
8-1
115
called the components of the quadrupole moment tensor as follows:

N
Qjk=
r,2sjk)
(=i
( j , k = x,y,z)
In this expression, j and k can independently

Kronecker delta symbol defined by
1 if j
/* = L
0 if j
(8-26)
be x, y, or z; the symbol 8jk is the

= k
^
# k
(8-27)
Thus there are a total of nine Qjk defined by (8-26). For example,
Qxx = X9;(3*, 2 - r?)
QXy = T.^xiyi
(8-28)
Upon comparing (8-26), (8-28), and (8-25), we see that the last one can be more
compactly written as
X9/,2(3cos20,, - l)
i
= lxQxx + lxlyQxy + 1 xKQ xz

+ IylxQyx + lyQyy + 1yl zQ yz
+ l l
z xQzx + hlyQzy + 1zQzz
ljlkQjk
(8-29)
j x, y, z k = x, y, z
Finally, when (8-29) is inserted into (8-7), we find that the quadrupole term can be
written in terms of the quadrupole moment as
T0r
Z j=x
k = x,y,z
y z
''kQ"'
(8
-30)
If the point P is very far away and if both the monopole moment Q and the dipole
moment p are zero, then (8-30) will be the leading term in the expansion of 4> and the
quadrupole moment tensor Qjk will be the dominant feature of the charge distribution.
Sometimes it is convenient to express the quadrupole term explicitly in terms of the
coordinates of the field point rather than in terms of its direction cosines. We can do
this by using lx = x/r, ly = y/r, lz = z/r, according to (1-7) and (1-11), so that (8-30)
becomes
0
j =x,y,z
k = x,y,z
[The forms (8-30) and (8-31) are reminiscent of the result obtained in mechanics for the
kinetic energy of a rigid body when expressed in terms of the moments and products of
inertia.]
If the charges have a continuous distribution, then the sum in (8-26) can be replaced
by an integral, so that for a volume distribution we would have
Qjt = f p(T')(3j'k'
J
- r'%)
dr'
(8-32)
For example,
Qxx= f p(r')(3x'2
J yr
- r'2) dT'
Qxy=f
p{r')3x'y'dr'
J yr
There will be similar expressions for surface and line charge distributions.
(8-33)
116
Although there were a total of nine quantities Qjk defined by (8-26), there are
actually fewer independent ones. We see from (8-26) or (8-28) that Qyx = Qxy, and so
on; that is,
Qkj
Qjk
U * k )
(8*34)
Thus, the quadrupole moment tensor is an example of a symmetric tensor, and (8-34)
reduces the number of independent components to six. If we now sum up the diagonal
components, that is, those with j k, we find that
= E < ? , [ ( 3 4 - r,2) + ( 3 y f - r,2) + (3zf - r, 2 )] = 0
Qxx + Qyy +
since r xf + yf + z ; thus we have

Qxx + Qyy + Qzz
= 0
(8-35)
and this result, which is true regardless of the location of the origin or of the detailed
nature of the charge distribution, reduces the number of independent components to
five.
If the charge distribution has sufficient symmetry, the number of independent
components can be reduced even further. As an extreme example, let us consider the
case of axial symmetry, that is, the distribution has an axis of rotational symmetry such
as is the case for a cylinder, a cone, or an egg. Let us choose the z axis to be along this
axis and designate the elements in this case by Qjk. Then, for every charge q' at
( x y ' , z ' ) , say, there will be a charge of the same value at
x', y', z') and the
contribution of this pair to Qxy, for example, will be 3 q ' x ' y ' + 3 q ' ( - x ' ) y ' = 0. Since
all of the charges can be paired in this manner, the result will be that Qxy = 0. This
same argument holds for the rest of the off-diagonal elements so that
Q% = 0
( j * k )
(8-36)
and we are down to three components. But, because of (8-35), there are really only two
because
Qxx + Qyy +
= 0
(8-37)
Furthermore, because there is no real distinction between the x and y axes in this case,
for every charge at a given value x', there will be an equal charge at the same numerical
value for y' for the same value of /*'. Thus, the corresponding sums in (8-26) will be
equal so that
e;, = Q
(s-38)
When this is put into (8-37), we find that 2Q" x + Qazz = 0. Therefore, Qxx = Qyy =
\Q"Z and there is only one independent component of the quadrupole moment
characteristic of the charge distribution. Calling Qa = Q"z the quadrupole moment of
this axially symmetric charge distribution, we have
Ql = Qa
,
,
(8-39)
Qlx=Qayy=
-\Q
Under these circumstances the quadrupole term simplifies considerably. If we insert
(8-36) and (8-39) into (8-30), and use (1-9) to eliminate l x and l y , we find that 4>Q
becomes
Qa ( 3 / 2 - l )
Qa (3 cos 2 0 1)
= T ~
= T
n 1, "
(8-4)
y
i
4t70
4r
47re0
4r
where 6 is the angle made by the position vector r of the field point with the axis of
8-1
117
symmetry (the z axis). An example of such a situation as this is provided by atomic

nuclei. Although their properties must be described by quantum mechanics, the
treatment is quite analogous. It can be shown that if a nucleus can have a quadrupole
moment at all, it must necessarily have an axis of symmetry that is the direction of its
intrinsic angular momentum or "spin." Thus it is essentially the quantity Qa that is
found listed in tables of nuclear quadrupole moments; actually what is tabulated is
Qa/e where e is the electronic charge so that these nuclear quadrupole moments are
given as areas.
4. Effects of the Choice of Origin

The monopole moment Q given by (8-15) is a unique property of the charge distribution. Both (8-19) and (8-26) do depend on the absolute value of the r(, so that the dipole
moment and the quadrupole moment components are not generally unique properties
of the charge distribution but depend on the choice of origin as well. Under some
circumstances, however, they are independent of this choice, and we want to investigate
this in more detail.
Suppose that instead of choosing the origin at 0 in Figure 8-1, we choose a new
origin 0, which is obtained by translating our axes, without rotating them, by the
displacement a as illustrated in Figure 8-2. The position of q t with respect to this new
origin is r, and we see from the figure that a + rin = r, so that the old and new
position vectors are related by
r, = r, - a
(8-41)
If we insert these values into (8-19) in order to find the new value of the dipole moment
P, we get
P = Hqihn = Lfc r i
= P - 0a
(8-42)
which shows that the dipole moments as calculated with respect to these different
Figure 8-2. The new origin of coordinates is displaced by a from the old.
118
origins will generally be different. However, we now also see that the dipole moment
will be independent of the choice of origin and hence a unique property of the charge
distribution provided that the monopole moment vanishes, that is,
p = p
if Q = 0
(8-43)
In addition, the dipole term will be the leading term in the expansion (8-7). In order to
make Q = 0, we will need at least two charges in our distribution; this is a reason for
the name dipole.
Example
Two equal and opposite point charges. This is the simplest example for which Q
vanishes. The two charges and their locations are shown in Figure 8-3. From (8-19), we
find p to be
P = qr+- qr_= q(r+- r _ ) = ql
(8-44)
so that the dipole moment equals the product of the charge magnitude and the vector
displacement 1 from the negative charge to the positive charge. Although this charge
distribution is often thought of as the prototype of a dipole, and is often simply called a
"dipole," its higher-order moments will generally not vanish so that (8-8) will necessarily include more than the term <t>D(r).
Another simple example satisfying (8-43) is shown in Figure 8-4 and it is easily
verified with the use of (8-19) that the dipole moment is p = 2qi.
Similar results can be obtained for the quadrupole moment. It will be sufficient for
our purposes to consider only one component, Qxy, say. The x and y components of
(8-41) are xin = jc, ax and yin = _y, ay and when we insert these into (8-28), we
find that the new component Q"y is
Qxy
E<liX
i n
yin
= 3E?,(*/ ~
i
=
Qxy
) ( y i -
i
~
3ayPx
3 a
x P y +
<*
'
4 5
with the use of (8-19) and (8-15). Since there will be similar expressions for the other
components QJk, we can conclude that the quadrupole moment is also not generally a
unique property of the charge distribution. However, we do see from (8-45) that it will
be independent of the origin provided that both the monopole and dipole moments
vanish, that is,
Q"k = Qjk
if<2 = 0 a n d p = 0
(8-46)
In addition, the quadrupole term will be the leading term in the expansion (8-7).
i
n2q
0'
Figure 8-3. Two equal and o p p o site point charges have a dipole
moment in t h e direction
of I.
Figure 8-4. A charge distribution

that has a dipole moment.
8 - 2 THE ELECTRIC D I P O L E FIELD
-q
a |
2q
> <
I
I
119
(b)
Figure 8-5. Charge distributions that have quadrupole

moments.
(a)
In general, we can satisfy the two conditions in (8-46) only with at least four charges
in our distribution; this is a reason for the name quadrupole. The simplest examples are
usually obtained by arranging two dipoles so that the total dipole moment vanishes.
Two such examples are shown in Figure 8-5 and the constituent dipoles are indicated
by the dashed arrows. We note that Figure 8-5b is a variation of Figure 8-4 with 1 = 0;
we also note that it corresponds to our statement about needing four charges only if we
mentally divide the central charge 2q into two superimposed charges each of value q.
Some care is required in devising such illustrative arrangements. For example, if we
interchange the two charges at the top of Figure 8-5a, we find that the conditions of
(8-46) are no longer fully satisfied so that the resultant charge configuration will have
an origin-dependent quadrupole moment.
To summarize this long section, we have found that when we are sufficiently far
away from an arbitrary charge distribution, the potential produced by it can be
expanded in the form
<f>(r) =
( Q
4<ne0 r
p ?
2r
LjAQjk
+ '-J
(8-47)
as obtained from (8-8), (8-16), (8-21), and (8-30). Since only the location of the field
point appears in (8-47) through its distance from the origin and the direction of r, all of
the moments can themselves be thought of as being located at the origin, regardless of
the actual spatial extent of the source charge distribution.
So far we have concentrated on the form of the potential. Now we want to consider
some of its properties, as well as the electric field which it describes. It is convenient to
look individually at each contribution arising from (8-47); the total E will be the vector
sum of them all. The first term corresponds to a point charge at the origin, and since we
are sufficiently familiar with its properties we can proceed at once to the next term.
18-21 THE ELECTRIC DIPOLE FIELD
The potential we want to investigate is given by (8-21) and will be the predominant
term when Q = 0. Although <j>D was obtained by studying the potential far from the
charge distribution, it is convenient to study its properties by assuming that <j>D holds
everywhere in space. Such a field is commonly called the dipole field and is thought of
as being produced by a point dipole p located at the origin. This useful fiction can be
imagined to be the result of a limiting process applied to the equal and opposite charge
distribution of Figure 8-3. In this process, the separation 1 is decreased to zero while the
charge q is simultaneously increased so that the dipole moment, which is their product
according to (8-44), remains constant and equal to p. If we use spherical coordinates to
locate the field point P, and choose the z axis to be in the direction of p, we are led to
the situation shown in Figure 8-6. Using (8-21) and (1-15), we find that we can write the
120
Figure 8-6.
A point dipole at the origin and parallel to the z axis.
dipole potential as
p cos 6
. " ^
(8-48)
The equation giving the equipotential surfaces corresponding to <j>D = const, is thus
r ' - i - L - ' cos d = CD cos d
\ 47T0<t>D
(8-49)
where the constant CD characterizing a given surface depends on the value of <j>D. These
equipotentials are illustrated as the solid curves in Figure 8-7; the actual surfaces would
be generated by imagining this two-dimensional figure rotated about the z axis. Since
r2 in (8-49) must be positive, we see that CD must be positive for 6 < \ir where cos 6 is
positive; thus positive values of <j>D correspond to the equipotential curves in the upper
half of the figure. Similarly, the curves in the lower half of the figure correspond to
negative values of 4>D, since cos 6 is negative for 6 > \TT, so that CD must be negative
also.
The components of E can be found from (8-48), (5-3), and (1-101); the results are
p \ 2 cos 6
d<t>D
Er = dr
V 47reQJ
r3
(8-50)
1 dD ' p ^ sin 6
r dd
I 4?T0) r3
and E ~ d<pD/d(p = 0. These components have a different angular dependence but

both have the inverse cube variation with distance characteristic of the dipole field.
We can find the equation giving the line of E from our previous expression (5-39),
which expresses the fact that the line is parallel to E at each point on it. If we write
both dslf and E in spherical coordinates by using (1-98), we find that
dr = kEr
and
rdd = kEe
(8-51)
so that
dr
Er
2 cos 6
d In r
rd9
E0
sin#
d0
2d(\nsm9)
dd
which integrates to
In r = In sin2# + In KD = In ( KD sin2 6 )
8 - 3 THE LINEAR Q U A D R U P O L E FIELD
121
where KD is a positive constant of integration. Solving for r, we find the equation for
these curves to be
r = Kd sin2 6
(8-52)
so that each curve is characterized by a particular value of KD. When these curves are
plotted, we obtain the lines of E shown dashed in Figure 8-7. They are seen to be
consistent with the special cases of (8-50), which show us that Er = 0 when 6 = \ir,
and that Eg = 0 when 0 = 0 and v. Be sure to verify to your own satisfaction that the
directions of E as indicated by the arrows follow from the expressions for the
components given in (8-50), and that the lines of E are perpendicular to the equipotential surfaces.
THE LINEAR Q U A D R U P O L E FIELD
The general expression for the quadrupole potential as given by (8-30) can be quite
complicated depending on which components Qjk are different from zero. Consequently, we investigate only the special case in which the system has an axis of
122
symmetry so that the potential as given by (8-40) is

v
O"
t4ttc
~
(3 c o s 2 0 - 1)
4nr
"53>
Although many types of charge distributions can result in this potential, it is convenient
to regard the simple charge distribution of Figure 8-5 b as a prototype for this case since
it certainly has the necessary symmetry; Qa is actually negative for this situation as can
be seen from (8-39) and (8-26). As a result we can refer to (8-53) as a linear (or axial)
quadrupole field. The equation for the equipotential surfaces that are obtained by
setting <pQ = const, in (8-53) is
I
Qa
\ (3cos 2 # 1)
2
1
5
( W - l )
(8-54)
where the constant Cg characterizing a given surface depends on the corresponding

value of <g. For simplicity, only one of these is shown as the solid curve in Figure 8-8;
the others are of similar shape. Since (8-54) is independent of <p, the actual surfaces are
again generated by rotating these curves around the symmetry axis (the z axis in this
case.)
quadrupole.
8-4
ENERGY O F A C H A R G E D I S T R I B U T I O N IN A N EXTERNAL FIELD
123
The electric field components as found from (8-53), (5-3), and (1-101) are
I 3Q a \ (3 cos 2 6 - 1)
Er
cos 0 sin 6
Ee
\ 8t70 J
(8-55)
= 0
The field components decrease as the inverse fourth power of the distance to the field
point.
In order to find the equation of a field line, we again eliminate k from (8-51) by
division and use (8-55). In this way, we obtain
dr
Er
(3 cos 2 6 1)
rdO
Eg
2 cos 6 sin 6
which, when integrated, gives

r2 = Kq sin2 6 cos 6
(8-56)
where KQ is a constant of integration. Again for simplicity, we show only one of these
as the dashed curve in Figure 8-8. As was suggested for Figure 8-7, be sure to verify
that the directions of E as indicated by the arrows are consistent with the expressions
for the components given in (8-55) on the assumption that Qa is negative, as would be
the case for Figure 8-5b. We note again that, as they should be, the lines of E are
perpendicular to this much more complicated equipotential surface.
ENERGY OF A CHARGE DISTRIBUTION IN A N EXTERNAL FIELD
In the previous chapter, we discussed the mutual electrostatic energy of an arbitrary

distribution of charges that we obtained as the reversible work required to assemble
these charges against their mutual conservative forces of attraction and repulsion.
Under some circumstances, we are not interested in all of this energy, but only in part
of it. In order to see how this can arise, let us consider a situation in which the overall
charge distribution can be divided into two groups such that the charges in each group
occupy their own volume, and these volumes are some distance apart so that there is no
difficulty in distinguishing the groups. Such a situation is illustrated in Figure 8-9 where
I
124
the charges in one group are labeled ql0 and occupy the volume V0, while the other
group is simply labeled qi without an additional subscript, as is their volume V; we
have also labeled the groups as "external" and "system" for reasons that will become
clear shortly. Let us consider the single charge q, at its location r, with respect to the
origin 05 which, for later convenience, we choose to be within the system volume.
According to (5-48), we can write the energy of this charge as
Uei = q M ^ )
(8-57)
in terms of the total potential <#>(r;) at its location. Now (5-2) shows that <p is
determined by all of the charges and we can, in fact, divide the total sum into two
groups, one part representing contributions from the external charges ql0 and the other
part from the other charges in the system. Thus, we can write (f> = (f>s 4- <f>0 where, for
example,
N
Qm
foM -
(8-58)
1=1 477e0Ku
where N0 is the total number of external charges and R:l = |rt r,| where rI is the
position vector of ql0. The other part <j>s will be given by a similar expression. Inserting
this into (8-57), we find that the energy of q, can be written as the sum
Uei =
+ Qi<>o(Ti)
(8-59)
Now the first term represents the mutual energy of interaction among the charges of the
system and when we sum it up over all N charges of this type we will be led to a term
like (7-3). Thus this term will be the internal energy of the system and expresses the
amount of work required to assemble it. For cases in which we are now interested, the
system will be a definite entity in which the constituent charges will have fixed relative
spatial locations and this internal electrostatic energy will be a definite constant.
The second term in (8-59) then represents an energy arising from the interactions
between a system charge and the group of external charges. When this is summed over
i, we get the total energy of this kind and we can regard it as being the energy of the
charges of the system of interest due to the effect of the external field; if we call this
energy Ue0, we have
Ue0=
Ev+oM
i=i
(8-60)
and it is this interaction energy which is of interest here.

We can do a similar analysis for the external charges and reach similar conclusions.
However, in practical cases of interest, what is envisioned is that these external charges
are set up in the laboratory by charges on capacitor plates, or on other conductors, for
example, and one tries to study the system of interest by seeing how it responds to the
influence of the external source charges. In such a case, we generally do not study the
energy changes of the external charges so we need not concern ourselves further with
that. Thus, the energy term of value to us is that part of the total energy represented by
(8-60); consequently, in what follows, we restrict ourselves to it.
As it stands, (8-60) does not yet completely fit experimental reality. Generally, the
external sources are far enough away and the system is so small in spatial extent that <0
does not vary very much over the volume V. Consequently, it is sufficiently accurate to
expand <>0(r,) *n a power series about the origin and include only the first few terms of
the expansion. If we use rectangular coordinates, then the series expansion of 4>0 can be
8-4
125
written as
4>o(r,) = <#>o(*,
i ) = *o(0,0,0)
d<f> o
d<t>o
+ y
dx l0
'\
dy )0
+ Zi
dz
+ 2xiyi
dx
y Jo +
2 y
' Z \ dy dz JQ
(8-61)
where the subscript 0 on the derivatives indicates that they are to be evaluated at the
origin. It is convenient to consider the result of inserting (8-61) into (8-60) piece by
piece.
The first term in (8-61) is a constant, so that when it is put into (8-60) it can be taken
out of the summation and we find its contribution to Ue0 to be
Ue0M ~~ 0o(O'0'0)
6 0 0 (0)
(8-62)
where Q is the monopole moment as given by (8-15). Thus, this part of the interaction
energy is simply that of a single point charge equal to the total charge times the
external potential at the origin in harmony with (5-48).
The first bracketed term in (8-61) can be written as
T j ( V$o)o'
(8-63)
Er
because of (5-3) where E 0 is the external electric field. Since E 0 is a constant also, when
we put (8-63) into (8-60) we get this contribution to the energy as
UeOD ~
E0 E
- p E0
(8-64)
where p is the dipole moment as given by (8-19). Thus, this very important expression
gives us the interaction energy of a dipole in an external electric field.
By now, it seems clear that the remaining terms of (8-61) will lead to an energy
contribution involving the quadrupole moment components, although these terms are
not yet written in the most desirable form. Since the source charges are assumed to be
physically separate from the system, as in Figure 8-9, their charge density p 0 = 0
throughout the whole system volume V. Therefore, at all points within the system, $ 0
satisfies Laplace's equation according to (5-15) and (5-16); that is,
a y
dx2
d2<t>0
d2<t> o
dy2
dz2
= 0
(8-65)
Thus, we can add any multiple of (8-65) to the second-order terms in (8-61) without
changing their value; accordingly, we multiply (8-65) by (r, 2 /6), add this to what is
left of (8-61), and rearrange the terms by using the fact that mixed second partial
derivatives are equal, [e.g., (d 2 <j> 0 /dx dy)0 = (d2<f>0/dy dx) 0 ]; the result is that they
126
can be also written as

2 \ ( d2*\
( *, - r , )(
1 /S3 2
-> (
+ 3*.-.k
*X'(^l
d2<t
>
dx dy J 0
{3y
'
2 )
" "
' \ dx dz
( ^ ) o
M ) .
d2<
.
/
\
I d2*0 \ ,
2
t>0
2 \(
+ 3z,oc, ( I + 3Z.0/,- -rT- + ( 3 z , 2 - r , . )
dz'
dz dx
dz dy J0
(8-66)
If we now put (8-66) into (8-60) and use the definition of the quadrupole moment
components given in (8-26) and illustrated in (8-28), we find that the remaining
contribution to the energy in the external field becomes
U g 0 Q
d20\
I d2<j)0 \
Qxx\^T\
+ QJ - t t t I + e ,
dx'
dx dy J0
d2<j>0
+ Q
d24>o \
+ Q
dz dx In
dx dz JQ
l d2fj>0
I d24>0
+ Q yz
\ dydz
Jo
d2(j> o
d2<t>0 \
+ Qyy
dy dx J Q
I d24>o \
+ g J
J0
+ Qz
dz dy
(8-67)
dz2
or more compactly as
UeoQ
d2<P0
G jk I o
|
j ~ x , y , z k = x,y,z
O K
\ J
(8-68)
Jo
We can also write this in terms of the external electric field E 0 by noting that
(d<j> /dj) = E
because of (5-3) so that
0
0 j
ue0Q = - 7
/ dEn 0J_
E G
Jk
./ = *> >'> 2 k = x, y, .
(8-69)
dk ,0
which, when written out, becomes

1
U
eQ
6,
+ G yx
+ Gz
Ox
dx
dE, 0 y
dx
0*
(9^
+ Q yy \
d,0z
dx }o
dE, Ox
dE,
+ G *>-
dE,0 y
dy
Q ,
0z
+ Gz
dy /o
/o
0>-
+ G,
dE,
zr
dz
dz
dE,
0z
5z
Jo
(8-70)
showing us explicitly that the energy of a quadrupole moment in an external field

depends on the spatial derivatives of the electric field components.
If we combine (8-62), (8-64), and (8-69), we get our final expression for the energy of
the charge system of interest due to interactions with external source charges as
1 ^
/ dE 0 i
e0 = G<o() - P E o - 70 E g , * '
dk
j,k
+ ...
(8-71)
8-4
127
It is possible to carry this process even further by expanding the potential <0 in
terms of the multipole moments of the external source charges and thus getting a series
for the interaction energy in terms of the energy of a dipole in the field of a dipole, and
so on. We will not do this, however, but will leave the calculation of a few of the more
important of such terms as exercises.
We can write (8-71) somewhat more generally. As we noted after (8-47), the
moments can be regarded as being located at the origin, which in this case is the system
origin 0S of Figure 8-9. Since the system itself may be able to move, it is convenient to
refer its location to a coordinate system fixed in space. Thus, if we let r be the position
vector of 05 with respect to this fixed system of axes, then terms in (8-71) which are all
evaluated at 0S, the "location" of the system of interest, will be evaluated at r so that we
can finally write
1 ^
/ dE0i \
Ueo = <24>0(r) - P E 0 (r) - - E
+ ...
(8-72)
Now that we have obtained these results, let us look at some of the consequences of
the unfamiliar termsthe dipole and quadrupole energies.
1. The Dipole Energy
If we write it simply as UD, then the energy of a dipole p in an external field E 0 is
UD = p E 0 = pE0 c o s *
(8-73)
where * is the angle between them as shown in Figure 8-10. This would be an
appropriate form to use for a point dipole as we previously defined it. In Figure 8-11
we show this energy as a function of * . We see that the range of energy variation is
finite, and that the energy has a minimum at * = 0 when the two vectors are parallel.
This indicates that the tendency of the system is for its dipole moment to become
aligned with the external field. We can evaluate this effect in the following way. Since
the energy is a function of * , that is, UD = UD(yIr), we know from mechanics that there
will be a torque on p, and the component of this torque in the direction of increasing
* , T, will be given by
dUD
t = -
= -pE0 s i n * = - |p x E 0 |
(8-74)
Since T is negative, it means that the sense of the torque on p is such as to rotate p into
the direction of E 0 ; thus, if we use the definition of the direction of the cross product
given in Figure 1-14, we see that the vector torque T is in the direction of p X E 0 as
shown in Figure 8-12. Combining this with (8-74), the torque on p is given correctly in
magnitude and direction by
T = p X E0
(8-75)
128
Figure 8-11. The energy of a dipole

in an external field as a function of
the angle between their directions.
We note that T = 0 when * = 0 and IT SO that these values of the angle correspond to
equilibrium situations; since they correspond respectively to a minimum and a maximum value of the energy UD as shown in Figure 8-11, * = 0 is a position of stable
equilibrium, while * = tr is one of unstable equilibrium. In order to have equilibrium
at any intermediate angle, it will be necessary for an external agent to apply a
mechanical torque TM equal and opposite to the value of T given by (8-75) so that the
net torque on p will be zero:
T+
= 0
(8-76)
If we look back at (8-73), we see that if E 0 is not a constant, but a function of
position r, then it will be possible for a dipole to decrease its energy by moving to a
different position; in other words, we have the possibility of a nonzero translational
force F d on the dipole because of the external field. Again, from mechanics, this is given
by
Fd= -VUD = v ( p E 0 )
(8-77)
The fact that p is a constant enables us to put this into a more useful form. If we use
(1-112), we can write (8-77) as
F/) = E 0 X ( V X p) + p X ( v X E 0 ) + (E 0 V )p + (p V )E 0
(8-78)
The first and third terms vanish because p is a constant so that all of its spatial
8-4
129
derivatives are zero; the second term vanishes because E 0 is conservative and v X E 0
= 0 according to (5-4). Thus (8-78) reduces to
Fc=(p-V)E0
(8-79)
For example, if we write the x component of this with the use of (1-121), we get
8EQx
dE0x
3E0x
There are similar expressions for the other two rectangular components. The result
(8-79) thus verifies our expectation that a translational force on a dipole will exist when
the external field varies with position. [The torque given by (8-75) is present even in a
uniform external field.]
Now all of these results for the dipole moment have been obtained in a completely
general way and are applicable to the dipole moment of any type of charge distribution
of interest. Nevertheless, it is of value to see that they can also be obtained quite
directly and easily for the simple case of two equal and opposite charges shown in
Figure 8-3 that we can take as the prototype of a point dipole by letting 1 - 0 and
q -* oo so as to keep p = <7! constant. In order to emphasize this aspect, let us write the
separation between them as dr, as shown in Figure 8-13, so that we have r + = r _ + dr.
We also show the external field E 0 and the forces on the charges which are to be
calculated from (3-1).
Again, if 4>0(r) is the external potential, the energy will be obtained from (8-60) and
is found to be
ue0 = 4</>o(r + ) - 1^0(O
= 1dr
V<o = - p E 0
with the use of (1-38), (5-3), and (8-44). Thus, this result obtained for this special
situation agrees exactly with that found more generally in (8-64). The net force on the
system will be
F
net =
++
- = tf[Eo(r
) - E 0 ( r _ ) ] = qdE0
The x component of this can be found with the use of (1-38) and (8-44) to be
^net * = 1dE0x = <ldT Vo* = (p v )E0x
which is in agreement with (8-80) so that we will be led once more to (8-79). Similarly,
130
the resultant torque in a uniform field will be given by

T = r + X F + + r_X F _ = q(r+ r _ ) X E 0 = p X E,
and agrees exactly with (8-75).
2. The Linear Quadrupole Energy

As in the previous section, it will be sufficient for our purposes to look at the
interaction energy only for the case of a charge distribution with axial symmetry. If we
write it simply as Ug, and substitute the values for the moment components given by
(8-36) and (8-39) into the last term of (8-72), we get
TJa
Uq
Qa
dE0z
1 / dE0x
dz
2 1 dx
^ dE0y
dy
(8-81)
and where we remember that the derivatives are to be evaluated at the system location.
Again, since the external charges are all located outside the system so that p0 = 0, we
will have v E 0 = 0 according to (4-10). We can use this, along with (1-42), to express
(8-81) completely in terms of dE0z/dz\ the result is
- ^VLiOh*
- 1 - ^ 1
4 1 dz
(8-82)
where the last term follows from (5-3). Thus the energy term reduces to a very simple
compact form for this special type of charge distribution. We note that it depends only
on the variation along the axis of symmetry of that electric field component that is also
along the symmetry axis that we chose as the z axis. If Qa is positive, the energy will be
a minimum at that location where dE0z/dz has its greatest positive value. Thus, there
can be a translational force on this quadrupole given by
K -
- W S -
( dE0z \
1
d
- j e V v - >
(8 83)
'
EXERCISES
8-1 Carry out the expansion of 1/R, to include
all terms of order ( r , / r ) 3 and thus show directly
that the next term in (8-7) is correctly obtained
from (8-13) with / = 3. This part of the expansion
of <t> is called the octupole term.
8-2 A single point charge q is located at the
point (a, b, c). Find Q, p, and all components of
Qjk for this system. Which, if any, of these quantities are changed if a charge q located at the
origin is added to the system?
8-3 Assume the charges of Figure 8-3 to be on
the z axis with the origin midway between them.
Find <#> exactly for a field point on the z axis.
How large must z be in order that one can
approximate the exact potential on the z axis
with the dipole term to an accuracy of 1 percent?
How large must z be in order that one can
approximate E, on the z axis with the dipole

expression to an accuracy of 1 percent?
8-4 Evaluate (8-47) for the charge distribution
of Figure 8-4 and express $ in rectangular coordinates. Use the origin shown and take the negative
charges to be on the x axis and the positive
charge on the positive y axis.
8-5 Point charges are placed at the corners of a
cube of edge a. The charges and their locations
are as follows: 3 q at (0,0,0); ~2q, (a, 0,0);
- q , ( a , a , 0); q, (0, a, 0); 2q,(0,
a, a);
3q,(a,a,a)-,
4q,(a,0,a);
5q,(0,0, a). For this
distribution find the monopole moment, the dipole moment, and all components of the
quadrupole moment tensor. Show that your results satisfy (8-35). If it is possible to find a
different origin of coordinates for which the di-
EXERCISES
pole moment will vanish, where should this origin

be located?
8-6 Show that the charge distribution of Figure
8-5b leads to (8-40) and thus evaluate Qa for this
case.
8-7 A line charge of constant charge density X
and of length L lies in the first quadrant of the
xy plane with one end at the origin. It makes an
angle a with the positive x axis. Find Q, p, and
all of the Qjk. Express the quadrupole term of the
potential due to this charge distribution in terms
of the rectangular coordinates of the field point.
8-8 A sphere of radius a has a surface charge
density given in spherical coordinates by a =
a0 cos 6 where a0 = const, and the origin is at the
center of the sphere. Find Q, p, and all of the
Qjk. Express (8-47) for this charge distribution in
terms of the spherical coordinates of a field point
located outside the sphere.
8-9 Charge is distributed with constant volume
density p throughout the figure of Figure 1-41.
Find Q, p, and all of the Qjk. Interpret your
result for p. Then consider the special case in
which the volume is a cube of edge a and express
(8-31) in terms of the spherical coordinates of a
field point located outside the cube.
8-10 If one has a charge distribution whose
monopole moment is not zero, show that one can
always find an origin such that the dipole moment
will be zero. This point is called the center of
charge of the distribution.
8-11 Find the analogous expression to (8-45) for
Qxx-
8-12 A spherically symmetric spherical charge

distribution has the origin at its center. Show that
p and all of the QJk are zero.
8-13 A point dipole p is located at the origin,
but it has no special orientation with respect to
the coordinate axes. (For example, p is not parallel to any of the axes.) Express its potential at a
point r in rectangular coordinates, and find the
rectangular components of E. Show that E can be
written in the important and general form
E(r) =
t [3(p r)f - p]
4ttc0/"
(8-84)
Can E ever be zero, other than for r oo? If,

now, p is located at r', rather than at the origin,
how can E at r be obtained?
131
8-14 The ' constant K D in (8-52) that characterizes a given line of E can be related to the
magnitude E of the electric field for the point on
the curve corresponding to 0 = \tt. Find this
relation.
8-15 Evaluate (8-47) for the charge distribution
of Figure 8-5a and express the potential in terms
of the cylindrical coordinates of the field point.
Assume the figure is a square of edge a in the xy
plane, origin at the center, sides parallel to the
axes, and q in the first quadrant. Find the cylindrical components of E. Find the equipotential
curves in the xy plane and plot your result. Find
the equation for a line of E in the xy plane and
plot your result.
8-16 The constant KQ in (8-56) that characterizes a given line of E can be related to the
magnitude E of the electric field for the point on
the curve for which Er vanishes. Find this relation.
8-17 A point dipole p at r is in the field of a
point charge q located at the origin. Find the
energy of p, the torque on it, and the net force on
it.
8-18 The linear quadrupole of Section 8-3 is in
the field of a point charge q located at the origin.
Find the energy of the quadrupole and the translational force, if any, acting on it.
8-19 A point dipole p, is located at r, and
another point dipole p2 is at r2. Show that the
energy of p2 in the field of px is given by the
dipole-dipole interaction energy
Udd =
w?, [(Pi P 2 ) -
477Cq K
(Pi *)(Pz )]
(8-85)
where R = r2 . Does it make a difference if R

is drawn from 2 to 1? Find the force F2 on p2.
Evaluate F2 for these two special cases: (a) pj and
p2 are parallel to each other but perpendicular to
R; (b) P l and p2 are parallel to each other and
parallel to R.
8-20 The linear quadrupole of Section 8-3 is in
the field of a point dipole p located at the origin.
Find the energy of the quadrupole.
8-21 A linear quadrupole Q" is located at r and
is in the field of another linear quadrupole Q"
located at the origin. Both are of the type of
Section 8-3, that is, the axis of symmetry of each
is parallel to the z axis. Find the energy of Q".
B O U N D A R Y C O N D I T I O N S AT A
SURFACE O F D I S C O N T I N U I T Y
soon as we consider the possibility of matter being present and subject to the
influence of other charges, we realize that we may need to consider a situation in which
w e have two kinds of matter meeting at a common boundary. For example, we may
have a block of wax in contact with a sheet of glass, or glass in contact with a
conductor or vacuum, and so on. We can also expect that these different kinds of
matter will have different electromagnetic "properties," so that these properties will
change abruptly as we cross the surface of separation. As a result, it is quite possible
that our various fields will be different in the two regions and it will be useful to know
how they change as we go across this boundary. These changes, or lack of them as the
case may be, are often called "boundary conditions," and we want to investigate them
in general terms.
It turns out that these boundary conditions can be obtained for any vector in terms
0 f its divergence and curl, so that it is convenient for us to derive them now. In this
w e will have available a set of "ready-made" expressions that we can apply to
specific fields as we encounter them.
M]
0RIG,N
O F
S U R F A C E
O F
DISCONTINUITY
In Figure 9-1 we illustrate what we can call the "real situation" at a boundary between
two materials or media, which we label 1 and 2. What is plotted is the variation of some
a S yet unspecified electromagnetic "property" as a function of position x; the
"boundary" between medium 1 and 2 is indicated by the solid vertical line. As we go
far enough to the left, this property becomes constant and completely characteristic of
medium 1; similarly, if we go far enough to the right, the property again becomes
c o n s t a n t , but at a different value, so that there is no doubt that we are in medium 2.
jsjoW as illustrated, we expect that in an actual case this property will not change
abruptly, but continuously, although possibly very rapidly, in a thin region between the
two media; we call this region the "transition layer." This expectation is appropriate
for our constant assumption that the physical fields we are dealing with are continuous
and have continuous derivatives. Although we have indicated the extent of this
t r a n s i t i o n layer by the dashed lines and even assigned i t a thickness h, we really cannot
hope to know anything about the specific details of what is actually occurring within it.
Consequently, it is customary to replace this actual situation by an idealized situation,
w h i c h is illustrated in Figure 9-2. We do this by imagining the transition layer to shrink
to zero thickness, thus giving rise to a discontinuity in electromagnetic properties. As a
result, it is possible that the fields themselves can be regarded as possibly having
discontinuities in this limit as h - 0, and this is what we want to consider.
Let us consider a general vector field F(r). As in Section 1-20, we write its source
e q u a t i o n s in the form
V F = Z>(r)
and
v X F = c(r)
(9-1)
132
9 - 2 THE D I V E R G E N C E A N D THE N O R M A L C O M P O N E N T S
133
"Property"
"Property"
2
Surface of
discontinuity
Figure 9-1. The physical origin of a

transition layer between t w o media.
Figure 9-2. The idealized surface of

discontinuity between t w o media.
These will provide us with our desired information but we cannot apply them directly
to a discontinuity. What we will do is use these equations in the transition layer where
everything is continuous and then go to the limit h 0.
An important definition is given by Figure 9-3, which shows the direction of the
normal ft to the surface of discontinuity. As illustrated,
n = nfromito2
(9-2)
and we will always follow this sign convention and it is necessary not to forget it.
[9-2] THE DIVERGENCE A N D THE NORMAL C O M P O N E N T S
The divergence theorem (1-59) combined with (9-1) yields

(f>F da = ( V Fdr = f b{r) dr
Ts
Jy
Jy
(9-3)
We apply this to a small right cylinder of height h and cross-sectional area A a

constructed in the transition layer as shown in a side view in Figure 9-4. This is done so
that we will get contributions from both regions 1 and 2. The outward normals to the
faces are n 2 and n 1 while n w is one of the outward normals to the curved wall. We
choose the area A a to be small enough so that it will be a good approximation to take
F to be constant over these faces. (We contemplate that eventually we will let Aa -* 0
so that we will be obtaining relations valid at a point; in such a case, higher-order
corrections obtained from, say, a power series expansion of F will vanish since they
would be multiplied by A a in our final results, as we will see.) The surface integral in
(9-3) can then be written as
<f>F - da = F2 Aa 2 + F1 Aax + W
A
11 n
2
Figure 9-3. Definition of the normal to the surface

of discontinuity.
(9-4)
134
B O U N D A R Y C O N D I T I O N S AT A SURFACE O F D I S C O N T I N U I T Y
"ni
Figure 9-4. Volume used t o find boundary

condition from the divergence theorem.
where F2 and Fj are the values on the faces in the respective regions and W is the
contribution from the curved wall. W will have some finite value and we can write
W ~ h by the mean value theorem. We see from the figure that n 2 = n and fij = n so
that when we use (1-52), we can write (9-4) and (9-3) as
n (F2 - F j Aa + W=
fbdr
Jv
= (hb) Aa
(9-5)
where b should actually be written as its average value within the volume h Aa\ again,
however, we can take b to be approximately constant throughout this small volume that
we will eventually let go to zero.
Now we can carry out our limiting process of letting the transition layer shrink to
zero thickness by letting h -* 0 while we keep A a constant. Because W is proportional
to /?, W - 0 as h -* 0. On the other hand, we cannot generally be certain about the
behavior of the product hb, since it may well happen that b increases in such a way in
this process that lim A _ 0 (hb) remains finite. Upon doing this to (9-5), we see that we
are able to cancel A a from both sides of the result, and thus we get the following
expression applicable to the situation at the surface of discontinuity:
n- (F2 Fj) = lim (hb) = lim (hv F)
(9-6)
A-0
h->0
Since n F = F is the normal component of F, that is, that in the direction of the
normal, according to (1-21), we can also write (9-6) as
F2 Fl = lim (hb) = lim (hv F)
(9-7)
h->0
h>0
Since this difference may be different from zero, we have the possibility of a discontinuity in the normal components of the vector F.
R^3L
THE CURL A N D THE TANGENTIAL C O M P O N E N T S
Stokes' theorem (1-67) combined with (9-1) yields

(f
ds = Jf (v X F) da. ==J f c(r) da
(9-8)
T
c
s
s
We apply this to a small rectangular path constructed in the transition layer and
perpendicular to the surface of discontinuity as shown in Figure 9-5 so that the sides of
length As will give us contributions from both regions. The arrows indicate the sense
of integration around the path C; i2 and are unit vectors in their respective directions
9 - 3 THE C U R L A N D THE T A N G E N T I A L C O M P O N E N T S
135
*2
>
jt
!.4s
1
\
Figure 9-5.
Area used to find boundary condition from Stokes' theorem.
of integration and are parallel to the surface of discontinuity. The vector fV is the
normal to the area enclosed by the path and is parallel to the surface between 1 and 2;
thus n" is perpendicular to n, the normal to the surface. The figure also shows a
tangential vector I parallel to the plane of C defined so that l2 = ? and tx = t.
Therefore, n, t, and n' are a set of mutually perpendicular unit vectors and satisfy
relations analogous to (1-25), that is,
n' = n X ?
t = n' X n
t X n'
(9-9)
The vector area of the rectangle as given by (1-52) is then seen to be h'h As. Applying
(9-8) to this situation, we obtain
i F ds
>
'c
= F2 t2 As + Fi
As + if
= t - (F2 - F j As +
c h'h As
(9-10)
where F2 and Fx are the values of F in the respective regions, and iV is the contribution
to the line integral from the ends of the path. Again, strictly speaking, the values of
F2, F1? and c are average values, but because As and h As are so small, we can take the
vectors to be nearly constant. If we replace t by the middle expression in (9-9) and use
(1-29), we can write (9-10) as
n' [n X (F2 - F j - he] As + iT= 0
(9-11)
Now we again let the transition layer shrink to zero so that h > 0 while we keep As
constant. Similarly to before, W will be proportional to h and will vanish in this
process. When all of this is done, we find that As can be canceled from both sides of
what (9-11) becomes, and we are left with
n X (F2 - Fx) -
n'
A /
lim
o(M]
= 0
(9-12)
The orientation of our path of integration was completely arbitrary, so that n'
corresponds to an arbitrary direction in the surface. The only way in which (9-12) can
always be true under these circumstances is for the term in brackets to be zero; thus we
get
n
(F,
F j
lim ( h e )
hÔ
lim
h *0
[ * ( v
) ]
"
1 3
as our final result.

We can put (9-13) into a form that is more easily interpretable. Let us write F as the
sum of its component normal to the surface of separation (its normal component F)
and its component parallel to the surface (its tangential component F,), that is,
F = F + F, = Fh + F,
(9-14)
136
B O U N D A R Y C O N D I T I O N S AT A SURFACE OF D I S C O N T I N U I T Y
Consequently,
n X F = Fnh X n + n x F , = n X F ,
(9-15)
because of (1-24). Using this result, we see that (9-13) actually involves only the
tangential components of F:
n X (F2, - F 1( ) = Urn (Ac)
(9.I6)
We can, in fact, write our result even more explicitly in terms of the tangential
components. With the use of (1-23), (1-30), (1-17), and (9-14), we find that
(n X F) X n = F - Fnh = F,
(9-17)
Thus, if we cross both sides of (9-13) into h, and use (9-17), we get
F 2 , - F = lim [ h ( c X n)] = lim { h [ ( v X F) X n] }
h * 0
h > 0
(9-18)
and if this difference is different from zero, we have a discontinuity in the tangential
components of F.
If we combine these results with those of the last section, we see that we have
obtained a way of finding how the vector F at the surface of discontinuity changes as
we go across the bounding surface. Let us suppose that we know Fx. Our first step
would be to resolve it into its normal and tangential components Fln and F l r We can
find the normal component of F2 from (9-7), and its tangential component from (9-18).
Knowing these, we can then find the vector F2 by combining them according to (9-14).
Since, in principle, both the normal and tangential components can change as we go
across the bounding surface of discontinuity, we see that the vector F may have both a
different magnitude and a different direction on the two sides.
B O U N D A R Y C O N D I T I O N S FOR THE ELECTRIC FIELD
So far the only vector field we have considered is the electric field E, but we have
already found the information about it that we require, that is,
V XE = 0
and
v E =
e
(9-19)
according to (5-4) and (4-10). By comparison with (9-1), we see that in this case
p
c = 0
and
b =
(9-20)
0
Applying the first of these to (9-18), we see immediately that the tangential components
of E are unchanged, that is, they are continuous across the surface of discontinuity:
E
2, - E l f = 0
(9.21)
or, simply
2, = E u
(9-22)
since the tangential components are parallel to each other. [Later, we will see that
(9-21) is correct under all circumstances.]
If we use the value for b given in (9-20), we see that
1 r
ph A a
Aq
r
I bdr = I p dr =
= e
(9-23)
Jy
0
0
where Aq is the total charge contained within the volume h A a of this portion of the
9-4
B O U N D A R Y C O N D I T I O N S FOR THE ELECTRIC FIELD
137
transition layer. This charge Aq may occur because of the properties of the two media
as we will see in the next chapter, or it may be there because we originally put it there.
In any event, as we imagine the transition layer shrunk to zero thickness to produce our
idealized case of Figure 9-2, this total charge Aq must be conserved. In this limit, the
charge should then be described as a surface charge of some density a so that
Aq = oAa. Therefore, we get
Aq = oAa = Iv lim hp) Aa
h->0 '
so that
a = lim (hp)
h-> 0
(9-24)
and therefore
lim (hb) =
(9-25)
h-* o
e0
in this case. Inserting this into (9-6), we find the boundary condition satisfied by the
normal components of the electric field to be
h-(E2-E1)
= E2n-Eln
=-
(9-26)
Thus, there is a discontinuity in the normal components of E only if there is a surface

charge on the surface separating the two regions. (This will also turn out to be correct
under all circumstances.)
In Figure 9-6, we illustrate the application of (9-22) and (9-26) to finding the relation
between E j and E 2 ; the figure is drawn on the assumption that E j is known and that a
is positive. If a x and a 2 are the angles made with the normal, we see from the
construction of the figure that a 2 < a x ; in other words, the vector E is "refracted" as
we go across the boundary. In this case, its direction is changed so that it is more nearly
in the direction of n. If a were negative, E would be refracted away from the normal
and a 2 > a v
Let us now see how well what we have obtained in this section agrees with some of
our previous results for E.
F
= cF
E'ln
1n
Figure 9-6. Refraction of lines

of electric field at a charged
surface of discontinuity.
138
B O U N D A R Y C O N D I T I O N S AT A SURFACE O F D I S C O N T I N U I T Y
t ^ E2
N
1
Figure 9-7. Calculation of electric field for a uniform
sheet by using the boundary conditions.
'Ei
Example
Uniform infinite plane sheet. As given by (3-12), the electric field has the constant
magnitude of E = o/2e0 and is directed away from the charged sheet as shown for
positive a in Figure 9-7. In this case, the surface does not separate two regions of
different properties since we have assumed a vacuum on both sides, but our results are
actually applicable to any surface. We see that E 2 = Eh and E x = .En because of
(9-2) and our choice for labeling the regions. Since there are no tangential components,
(9-22) is automatically satisfied. Applying (9-26), we get
E2n-Eln
= E - ( - E ) = 2E = ?e
so that E = a/2e0 in exact agreement with (3-12). (We have, in fact, already remarked
on this result in the last sentence of Section 3-3.)
Example
Surface of a conductor. Here again there are no tangential components and the situation
is that shown in Figure 6-1 b. If we let the conductor region be 1, and the vacuum
region outside be 2, then n will be the outward normal to the surface of the conductor.
In this case, Eln = 0 because of (6-1), and E2n = E so that (9-26) becomes E2n Eln = E = a/e0, agreeing exactly with (6-4), which we obtained by other means.
Thus, these two examples agree with the general results we have obtained and,
incidentally, show us that the use of the boundary conditions in this way provides us
with a quick way of calculating the fields for simple enough situations.
51 B O U N D A R Y C O N D I T I O N S FOR THE SCALAR POTENTIAL
Since $ is a scalar field rather than a vector one, we cannot use the general results of
Sections 9-2 and 9-3 to discuss its behavior at a surface of discontinuity. We can,
however, easily obtain what we need with the use of (5-11). If we let 4>2 and (pl be the
values of <j> on each side of the transition layer, then we have
<f>2 4>i = f2E ds
J
(9-27)
where the integration is over any convenient path through the transition layer. If we
choose this path to be along n, then we can write (9-27) as
*2
-<#>!=
f
J
ds = - (E)h
(9-28)
EXERCISES
139
where ( E n ) is the average value of the normal component of E in the transition layer of
thickness h. Now in any real physical situation, the electric field will not become
infinite as h -* 0, since that would imply an infinite force on a point charge. In fact, as
we saw in (9-26), the most that could happen is that En would have a discontinuity, but
would otherwise be finite. Therefore, (En) will always be a finite quantity so that
lim^ _0(En)h = 0 and (9-28) will lead to
$2
(9-29)
Thus, the scalar potential must be continuous across a surface of discontinuity in

properties.
Although tp is continuous, its derivatives normal to the surface need not be. Because
of (1-21) and (5-3), we have
En n V)i = -
e
(9-31)
[While (9-29) is certainly correct for any real physical situations, one occasionally
deals with mathematical idealizations that do result in a discontinuity of 4> across a
surface. The most common example of this is a "dipole layer," which arises from
assuming the surface of separation to contain electric dipoles of the type we discuss in
the next chapter, but with an infinitesimal separation between the charges of opposite
sign. However, we will not find it necessary to consider such specialized cases.]
EXERCISES
9-1 The surface of separation between regions
1 and 2 is a plane whose equation is 2 x + y +
z 1. If E j = 4x + y 3z, find the normal and
tangential components of Ej.
9-2 The actual physical relation between E j and
E 2 cannot depend on our arbitrary choice of the
direction for n. Show that if the direction of the
unit normal vector were defined to be from region 2 to region 1, the relation between E 2 and
Ej shown in Figure 9-6 would be unaltered.
origin. Within the sphere, the electric field is
given by EL = ax + fly + yz where a, yS, y are
constants. There is a surface charge density on the
surface of the sphere given in spherical coordinates by a <T0 cos 6 where ct0 = const. Find E 2
at all points just outside the sphere and express it
in rectangular coordinates.
9-4 Show that the angles in Figure 9-6 are related by
tan a!
tan a-,
r-
l + /
\ e 0 Ex cos ! I
How are the angles related if Ex becomes very

large? Is this reasonable?
9-5 In Figure 9-8, we show two points P2 and
Pl located in the two media but just across the
surface from each other. The point P 3 is located
at the edge of the interface, wherever it may be.
Use (9-27) to find the potential difference between P2 and P3 by following a path of integration right along the bounding surface but just
inside region 2 as indicated by the curved arrow.
(Why is this all right?) Similarly, find the potential difference between /-*, and /*3 by staying
inside region 1 but next to the surface. From your
results, show once again that the potentials at P2
and Pj are equal, in agreement with (9-29).
Figure 9-8.
9-5.
Paths of integration for Exercise
|10
ELECTROSTATICS IN THE
PRESENCE OF MATTER
As we have remarked before, for our purposes, matter is to be regarded as a collection
of the positive and negative charges of its constituent nuclei and electrons. We assume
that an ordinary piece of matter is made up of atoms and molecules that have equal
amounts of positive and negative charges, that is, that have zero monopole moments. In
Chapter 6 we considered one class of matterconductorswhich we assumed to have
charges that were free to move about under the influence of fields. In this chapter we
assume that we are dealing with matter that is not a conductor. These latter types of
material are commonly called dielectrics and we refer to their charges as bound charges.
It is not always possible to make a clear-cut characterization of a given substance as a
conductor or dielectric (nonconductor), but it is still useful to have these two categories
available for a first rough classification scheme. Later we will be able to combine these
two properties in our overall description.
liFTl POLARIZATION
Although our primary aim is to develop a macroscopic empirical description of

electromagnetism, for a while we will use the microscopic picture of matter in a
qualitative manner in order to get an idea of how we should proceed. There are several
possibilities that come to mind when we consider the effect of an electric field on an
atom or molecule.
In the absence of a field, it may well happen that the molecule has its negative
electronic charge distributed symmetrically about the positively charged nuclei. In this
case, the molecule will have a zero dipole moment as can be easily seen from (8-22). If,
now, there is an electric field present, there will be forces exerted on these charges. The
positive charges will tend to be moved in the direction of the field, and the negative
charges will tend to be moved in the direction opposite to the field. Eventually, the
internal forces within the molecule will produce a new state of equilibrium, but the
molecular charge distribution will have been distorted from its originally spherically
symmetric form. Thus, the previously "coincident" positive and negative charges will
have had their overall "centers of gravity" shifted with respect to each other, and the
net result will be a new charge distribution with roughly the character of that shown in
Figure 8-3. But, as we know, this means that the molecule now will have a nonzero
dipole moment p, and probably other higher-order multipoles as well. We say that the
molecule now has an induced dipole moment and that it has become polarized.
Another possibility is that, because of their internal structure, some molecules
already have their positive and negative charge distributions separated so that they
have a dipole moment even in the absence of an electric field. Such molecules are called
polar molecules and the dipole moment is referred to as a permanent dipole moment; the
monopole moment is still zero, of course. An example of this type is the water
molecule, H 2 0 . Here the negative charges tend to cluster more around the oxygen
leaving the hydrogens positively charged; the net result is a charge distribution roughly
140
10-1
POLARIZATION
141
like that of Figure 8-4 (with the signs of the charges reversed) and with q approximately
equal to e, the magnitude of the electronic charge. In the absence of an electric field,
these permanent dipole moments will generally be randomly oriented so that the total
dipole moment of the whole piece of matter will still be zero. In the presence of a field,
however, there will be a torque on the dipole that will tend to rotate the dipole into the
direction of the field as we found in (8-75). This tendency can be expected to be
opposed by other internal forces of the system and, especially in the cases of gases and
liquids, by the temperature agitation of the molecules that will lead to collisions with
their tendency to make things random. Nevertheless, when equilibrium has been finally
attained in the presence of the field, we can expect that the permanent dipoles will have
been rotated on the average into some degree of alignment with the direction of the
field. The overall effect is that the matter will have a net dipole moment in the direction
of the field. Thus we have concluded again that the material will be polarized.
A final possibility is that the permanent dipole moments of some materials are
aligned to some extent even in the absence of an electric field. Such materials are called
electrets and are said to be permanently polarized. They are comparatively rare, and
not of as great commercial and technical importance as their magnetic analoguespermanent magnets.
As we noted above, we can reasonably expect the molecules, especially when
distorted by the field, to have higher-order multipoles, such as quadrupole moments,
associated with them in addition to their dipole moments. On the other hand, we know
from Chapter 8 that the contributions of these other multipoles to the potential and
field fall off more rapidly with distance from the molecule, and, in addition, their
contributions generally vary with angle in a more complex manner. Since we are trying
to get a macroscopic description in terms of the average of the microscopic behavior of
the constituents of matter, we will assume that, on the average, the dominant features
of matter that are of interest to us are simply those associated with the electric dipole
moments. Thus, all of our considerations lead us to the following:
Hypothesis
As far as its electrical properties are concerned, neutral matter is equivalent to an
assemblage of electric dipoles.
Although this is obviously an hypothesis, it clearly must have worked very well in the
development of this subject. Our next problem is to try to formulate this hypothesis in
a quantitative manner.
For this purpose, we define the polarization P as the electric dipole moment per unit
volume, so that the total dipole moment c/p in a small volume dr at r will be
d\> = P(r) dr
(10-1)
Thus the total dipole moment of a volume V of material will be

(10-2)
From its definition and (8-19), it follows that P will be measured in coulombs/(meter) 2 .
As implied by (10-1) and (10-2), P is generally expected to be a function of position
within the material. As in Section 2-4 when we introduced charge density, we are
assuming d t to be small on a macroscopic scale but large on a microscopic atomic
scale so that dr actually contains many molecules. This makes P obtainable as an
average over volumes large enough so that small fluctuations in the number of
molecules within a given volume element do not prevent us from treating P as a
142
ELECTROSTATICS IN THE PRESENCE O F MATTER
smoothly varying function of position. For example, if there are n molecules per unit
volume each of which has a dipole moment p, then P = p provided that all of the
dipoles are in the same direction.
Because of the way in which we were led to it, we expect there will be a functional
relation between P and E and eventually we shall have to consider how we can
determine P. For the time being, however, we simply take it as giving us a macroscopic
description of the material and we want to investigate the consequences of its existence.
BOUND CHARGE DENSITIES

Let us assume that we have a polarized object and we want to calculate the potential it
will produce at a field point r located outside the body as shown in Figure 10-1. The
dipole moment of the volume element dr' as given by (10-1) is dtf = P(r') dr' and its
contribution to the potential at r as obtained from (8-21) is
dtf R
P ( r ' ) R dr'
(10-3)
4-7T qR 2
477C0R
where, as usual, R = r r ' and corresponds to the r of (8-21). In order to find the total
potential, we integrate this over the volume V of the material and we get
, P(r') R ^ t '
1
<p(r) = /
-z = J y
4TTCqR
477t0
,
/ 1 \
[ P ( r ' ) V ' I dr'
Jv'
\ R
(10-4)
with the use of (1-141). This can be put into a better form if we use (1-115) to write the
integrand as
(10-5)
S'
Figure 10-1. Calculation of the

potential outside a polarized body.
10-2
B O U N D C H A R G E DENSITIES
143
If we insert this into (10-4) and use (1-59) and (1-52), we obtain
1
'
(-V'-P)JT'
4i(|l V
1
= i
I F
+
f -
4TT0 >V'
-R
IP
4ire 0 h ?
<
4TTC 0 ^ 5 '
where 5 ' is the surface bounding F ' and n' is the outer
in the figure. Upon comparing this with (5-7) and (5-8),
potential
that would be produced by a volume
throughout the volume and a surface charge density a b
' IR
P n" da'
^
dT'
(10-6)
normal to the surface as shown

we see that (10-6) is exactly the
charge density ph distributed
on the bounding surface where
P*=-V'- P
(10-7)
ab =
(10-8)
. ft' =
Pn
for then we would have

<J>(r) =
phdr'
R
oh da'
< "f V "
47Tq ~s
(10-9)
as we would expect. [In (10-8), Pn is the normal component of P evaluated on the

surface.]
What we have discovered, therefore, is that, as far as its effects outside itself are
concerned, the dielectric can be replaced by a distribution of volume and surface
charge densities that are related to the polarization P by means of (10-7) and (10-8).
The various steps that have led to our present conceptual scheme are summarized and
outlined in Figure 10-2. The total potential at the field point will then be given by
(10-9) plus the potential due to any other charges which might be present.
It is common practice to omit the prime in (10-7) and simply write
Pb ~
V P
(10-10)
with the understanding that the differentiations are made with respect to the source
point coordinates.
The subscript b appearing in (10-7), (10-8), and (10-10) reflects the fact that these
charge densities arise from the bound charges of the dielectric. Consequently, they are
usually referred to as the bound charge densities or the polarization charge densities.
Although we identified these quantities by means of a formal comparison of our
expression for the potential with its general form, it is possible to understand and
calculate the origin of these densities in a more direct "physical" manner; we will,
however, do this only qualitatively. As an extreme example, let us consider a piece of
material with uniform polarization produced by dipoles of the same magnitude all
pointing the same direction as shown in Figure 10-3. As in Figure 8-3, we can associate
a positive charge with an arrow head, and a negative charge of the same magnitude
with the tail of an arrow. Then, in a small volume within the material, as indicated by
the dashed curve, we will have, on the average, as many positive as negative charges so
Figure 10-2.
Conceptual scheme of replacing a dielectric by equivalent charge densities.
144
E L E C T R O S T A T I C S IN THE PRESENCE O F M A T T E R
/ >
Figure 10-3.
Origin of bound surface charges for a uniformly polarized dielectric.
Figure 10-4. Origin of bound volume charges in a

dielectric with nonuniform polarization.
that pb = 0 in agreement with (10-10). However, there will be no such cancellation on

the faces and this will produce a surface charge density of the signs shown; these signs
are in agreement with (10-8) since Pn is positive for the right side of the figure and
negative for the left side. Now let us suppose that P is not uniform, so that the
distribution of dipoles may be something like that of Figure 10-4 since the dipoles may
differ in both magnitude and direction. Let us consider a fixed volume of the material
as indicated by the dashed line. Before the matter was polarized, this volume will
contain many neutral molecules so that pb = 0. Now when the material is polarized
and the dipoles are formed, some charges will be displaced out of the volume while
others will be displaced into it. If P is not uniform, we see that it may well happen that
the volume may end up with more charge of one sign than of the other and there will be
a net bound charge density. This is exactly the effect described by (10-10), and, in fact,
considerations like these can be used as an alternative way of deriving pb = v P.
Suppose we have two polarized dielectrics meeting at a common boundary as
illustrated in Figure 10-5. Since each of them has a normal component at the surface,
they will produce a surface charge there given by (10-8); the net surface charge density
oh net will be the sum of these two terms so that
a
b, net = b2 + b\
^n2 + ^n\
^2 ' ^2
' "l
(10-11)
where n'2 and

are the outward normals to the respective media. If we now introduce
the unit vector n pointing from 1 to 2 as defined in (9-2), we see that
= n and
1 0 - 3 THE ELECTRIC FIELD W I T H I N A DIELECTRIC
145
n'j = n so that (10-11) becomes

ft-(P2-Pi)=
-a>net
(10-12)
This result is exactly what we get by applying our previous results (9-6) and (9-24) to
this case, which is described by (10-10); thus (10-7) and (10-8) are consistent with each
other.
As a further illustration of the internal consistency of our results, let us find the total
bound charge Qb of a polarized dielectric of finite extent. According to (2-14) and
(2-16), this will be given by
Qb
f Pb d t ' + < o b da' = f v ' Pdr'
Jy,
JS/
Jy,
= - j ) P n' da' + j ) P n' da' = 0
+ ( o b da'
JS,
(10-13)
with the use of (10-7), (10-8), and (1-59). Thus the total bound charge is zero. But this is
exactly what we would expect from our ideas of polarization as arising from the
separation of charge in our originally neutral medium rather than from the creation of
new charge. (In the case of permanent dipoles, the origin of polarization is due to
reorientation of already separated charge, but this does not involve charge creation
either.)
THE ELECTRIC FIELD W I T H I N A DIELECTRIC
Up to this point all of our results have been obtained by considering the potential, and
its corresponding electric field, at a field point in the vacuum outside the body of the
dielectric. The reason is that then there is no problem about what we mean by the
electric field in the sense of (3-14) since we can put a test charge there and measure
the force on it without any difficulty. However, if we look at the situation inside a
dielectric from this point of view, we cannot measure the force on a test charge without,
say, drilling a hole in the material in order to be able to insert the charge. But if we do
this, we can anticipate that we will alter the preexisting situation if for no other reason
than that we will remove some bound volume charges and introduce new surface
charges because we have created some new bounding surface. There are several ways in
which one can proceed to answer the question of what we mean by the electric field
inside the dielectric.
The simplest thing to do, and probably the best, is just to say that we calculate the
potential, and its associated electric field, inside the dielectric in exactly the same way
146
in which we do it for a point outside. In other words, we say that (10-6) or (10-9) can be
used to calculate <f> anywhere by definition. There is certainly nothing wrong with this
and it is a perfectly reasonable thing to do; it is completely consistent with our
conceptual replacement of the actual dielectric by a set of equivalent volume and
surface distributions of bound charge as shown in Figure 10-2. We have discussed this
approach before in Section 3-4, and we can point out again that our basic aim is to
develop a macroscopic description of electromagnetism that agrees with experiment. We
come back to this point below.
Another approach that can be used is to start with a microscopic picture and try to
determine the macroscopic equations as appropriate averages. This is quite a complicated program to carry out accurately for all of electromagnetism, and we shall have
to content ourselves with a brief outline of what is involved for electrostatics. On this
scale, most of the interior region will be vacuum and the electric field will be
determined by all of the nuclear charges and the atomic electrons. This electric field will
have enormous variations in it as we go from points very near these charges to points
which are comparatively far from them. At the same time, it will vary quite rapidly with
time because of the motion of the constituent charges. Thus, when we define our
macroscopic electric field, we will want to do so as an average both over the time and
over a volume large enough to contain a reasonable number of molecules but not so
large that we cannot consider it as infinitesimal on a laboratory scale. (This is similar to
the problem of defining average charge density as we discussed in Section 2-4; there we
saw that if we take too fine a subdivision, p can have very large fluctuations.) If we let e
be the microscopic electric field, we want it to have the same basic vacuum properties
that we have already found in (4-10) and (5-4), that is,
V e =
and
V Xe = 0
(10-14)
where pm is the charge density defined on a microscopic scale. Our macroscopic field E
would then be defined as
E = (e)
(10-15)
where ( e ) is an average over both time and space. Since the differential operators are
constant as far as this averaging process is concerned, we get from (10-14) and (10-15):
V X (e) = v X E = 0
V <e> = V E = (pm)
(10-16)
(10-17)
Now pm itself can be expected to have large spatial and temporal variations, but when it
is averaged we expect it to reduce to the average density of the displaced bound
charges; that is, ( p m ) = ph, so that
V E =
(10-18)
But (10-16) and (10-18) are equivalent to simply using (10-9) and (5-4) in the first place.
Thus this approach leads us to exactly the same conclusions we reached in the previous
paragraph.
This microscopic to macroscopic method of determining E in the material is
somewhat different from the question of what is the actual electric field seen by the
molecule, which is what will determine the actual molecular dipole moment produced.
This field will not necessarily be the same as the macroscopic E defined above as the
molecule itself contributes to the resultant field. This question is discussed in Appendix
1 0 - 3 THE ELECTRIC FIELD W I T H I N A DIELECTRIC
147
B; there we consider the question of calculating the electromagnetic properties of

matter as the resultant response of its constituent atoms and molecules, but it is not at
all necessary to worry about that now.
It is helpful at this point to verify that our definitions of <j> and E, which we have just
adopted, are in agreement with early experiments on dielectrics done by Faraday. We
do this only qualitatively now, but we will be able to give a quantitative discussion in
Section 10-7. We consider two simple experiments. We charge a capacitor to some
charge Q, and measure the potential difference A<0 when there is a vacuum between the
plates. The capacitance will be
Q_
Co
(10-19)
Aô
according to (6-28). Now we take this same capacitor, completely fill the region
between the plates with some "reasonable" dielectric like wax or oil, while keeping Q
constant, and then measure the potential difference A</>. The capacitance is now given by
c
(10-20)
A<
The experimental result is that A0, and therefore

(10-21)
C > C0
How can this result that the potential difference decreased be understood? Suppose, in
order to be specific, we consider a parallel plate capacitor. Then we can assume the
electric field to be directed from the positively charged plate to the negatively charged
plate as in Figure 6-9, and when we calculate A<j>0 by means of (6-38), we get
A<o = j
Er
d s = E0d
(10-22)
where E 0 is the electric field with a vacuum present and d is the plate separation.
Similarly, if E is the electric field with the dielectric between the plates, we get
A $ = Ed
(10-23)
and, since A<#> < A<0, while d is constant, we conclude that

E < E0
(10-24)
so that the electric field has been decreased by the presence of the dielectric. How could
this have come about? In Figure 10-6, we show the capacitor with dielectric between its
plates; the surfaces of the dielectric are shown slightly separated from the conductors
for clarity, but we really assume that the region is completely filled. If the dielectric is
uniformly polarized so that P = const., then pb = 0 by (10-10). Thus the only bound
charges will be surface charges given by (10-8), and they will have the indicated signs.
These surface charges will produce a field Eb with the direction shown; since the charge
Q
IP
MEl
I'
'+
Figure 10-6. Charges and fields

for a parallel plate capacitor.
148
Figure 10-7.
dielectric.
Cavity used to measure E in a
Q will still produce the field E0, the resultant field E is seen to be
E = E0 - Eh
(10-25)
so that E < Eq in agreement with (10-24) and experiment. Thus, this method of
evaluating the field inside the material by the use of equivalent bound charge agrees, in
this case at least, with experiment and our general ideas about the relation between
electric field and potential as expressed by (6-38).
Although we briefly alluded to the difficulty of actually measuring the electric field in
the dielectric by means of a test charge, one can still ask if there is a way of doing
itobviously we will have to cut a hole in order to insert the charge. It is possible to
devise such a scheme by means of a judicious use of the boundary conditions satisfied
by E; the relevant one is (9-21), which says that E, is continuous. The reasoning is as
follows: (1) somehow determine the direction of E in the dielectric (as we will see in
Section 10-6, this is easily done for a large class of dielectrics); (2) make a long narrow
hole in the material parallel to E as shown in Figure 10-7this cavity now has a
vacuum inside (such a cavity is usually called "needlelike"); (3) any bound surface
charges that might alter the fields are on the ends of the cavity since that is the only
place P has a normal component, and, since the ends are so far away and of such a
small area, we can neglect their effects; (4) since only tangential components are
involved, E t . = E where E c is the field in the cavity; (5) now insert a test charge 8q into
the cavity and measure the force 8 Fc on it. Then, by (3-14),
5F
E c = - j p, = E
8q
(10-26)
and the electric field in the dielectric can thus be found, in principle, from measurements made in the cavity. This scheme is known as the cavity definition of E.
10-4 UNIFORMLY POLARIZED SPHERE

As an example of the use of bound charges, let us consider a sphere of radius a that has
a constant polarization P. We choose the z axis to be in the direction of P and the
origin at the center of the sphere as shown in Figure 10-8; thus, P = Pz,. Since P is
constant, ph = 0 by (10-10). We see from the figure that the outer normal to the
dielectric n' = ?, and therefore there is a bound surface charge density which is found
from (10-8) and (1-94) to be
oh(6')
= Pcos6'
(10-27)
As we see, a h is not constant but varies in magnitude and sign with angle as indicated
in Figure 10-9. For simplicity, we will find the potential and electric field only for
points on the z axis and for positive z; in the next chapter, we solve this problem
completely by a quite different method.
We see from Figure 10-10 that
R = ( z 2 + a2 - 2 z a c o s 0 ' ) 1 / 2
(10-28)
and, therefore, if we use (10-27), (1-100) (with r = a), and (2-22), we find that
as
1 0 - 4 U N I F O R M L Y P O L A R I Z E D SPHERE
Figure 10-8.
sphere.
149
Figure 10-9. Equivalent charge

distribution of a uniformly
polarized sphere.
A uniformly polarized
Figure 10-10.
the axis.
Calculation of the potential at a point on
given by (10-9) is
1
^
r2n r* Pcos 6' a2 sin#' dd' dtp'
4i"0 o 'o ( z 2 + a2 2za cos 0')1/2
Pa2 ri
fi dp.
2
2<o
2e 0 J' i ( z + a 2zap,)1/2
(10-29)
The integral can be found from tables to be

( z 2 + a2 + zan)(z2
+ a2 2 zafi)1/2
3 z 2 a_ 2
1
-1
[(z 2 + a2)(\z + a\ \z a\) za(|z + a\ + |z a | ) ]

3 z2a2
and there are two cases to be considered.
(10-30)
150
1. Outside the sphere. Here z > a and \z a\ = z a; also \z + a\ = z + a since

both z and a are positive. Using these values, we find that (10-30) becomes 2 a / 3 z 2 and
when this is put into (10-29) we find the potential outside the sphere to be
Pa3
+.(*) =
(10-31)
r\Z
and therefore
d<t>0 2 Pa3
E
"
{ Z )
~ ~
<10
~ 2 ^ ?
"32)
These results become more understandable if we express them in terms of the total
dipole moment p of the sphere which, as found from (10-2), is
p = \-na3P%
(10-33)
so that (10-31) and (10-32) can also be written as

*<Z) - ^
(10-34)
*(*) -
(10-35)
Upon comparing these results with (8-48) and (8-50), and remembering that 6 = 0 and
r = z for a field point on the z axis, we see that they are exactly those of a point dipole
of total moment p. This makes us suspect that the field everywhere outside is a dipole
field corresponding to this total dipole moment; this turns out to be correct as we will
see in the next chapter.
2. Inside the sphere. Here z < a, so that \z a\ = a z; again, |z + a\ = z + a.
In this case, (10-30) becomes 2 z / 3 a 2 , which, when put into (10-29), gives the potential
inside the sphere as
<>l(z)=~
(10-36)
and therefore
d<j>i
P
" W = - a 7 = " 3 ^
<10"37)
We note that the electric field is constant; this may make us suspect that E is constant
throughout the whole interior of the sphere, and this turns out to be the case.
It will be left as an exercise to show that these same results also hold for negative
values of z, that is, Ezo is always in the positive direction, and Ezi is constant and
given by (10-37) for all z. In Figure 10-11, we illustrate the directions for E that we
have found; these directions can be understood from the source charge distribution
shown in Figure 10-9.
It is a worthwhile exercise to verify that these results are in agreement with the
general boundary conditions we previously obtained. We find from (10-31) and (10-36)
that the potential is indeed continuous at the surface of the sphere as we found in
(9-29), that is, when z = a, <j>0 = ^>( = Pa/3(.0. Along the z axis, Ez is a normal
component, and therefore we find that
12P\
P \
(7,(0)
10-5 T H E D FIELD
151
Figure 10-11. Electric fields on z axis due to a uniformly

polarized sphere.
with the use of (10-32), (10-37), and (10-27). This value for the difference in the normal
components is exactly what it should be according to (9-26).
10-5 THE D FIELD

We recall that when we defined E in Section 3-1, we pointed out that E is determined,
from its definition, by all charges of whatever origin or type. In Section 10-2, we
encountered a particular class of chargethe bound charge whose density is given by
(10-10) as pb = v P. It is conventional and convenient to divide charge into the two
broad classes of bound charge and free charge with corresponding densities pb and pj.
As we have seen, we regard the bound charges as arising from the constituents of
matter, and we generally have no control over their distribution. Free charges ace
essentially the rest of the charges. The name " f r e e " is given to them because we Van
control their distribution to a large extent by physically moving them about, by
spraying them onto or into material by electron beams, for example, and the like. This
class is also generally taken to include the mobile (free) charges of conductors. This
division does not always give us a clear-cut classification, but it is useful nevertheless.
Thus, we can write the total charge density as the sum of these two:
Ptotal =
P =
Pf +
Pb =
Pf -
* P
(10-38)
and when this is substituted into v E = p / e 0 , as we found in (4-10), we obtain

V ( e 0 E + P ) = Pf
(10-39)
The form of this equation, in which only the free charge density appears on the
right-hand side, suggests that it may be useful to define a vector field D(r) as follows:
D = e0E + P
(10-40)
152
ELECTROSTATICS IN THE PRESENCE O F MATTER
*
%
for then
Q
V D = pf
(10-41)
The vector D is often called the electric displacement or simply the displacement, or,
even more simply, the D field. The principal characteristic of D, and the primary reason
for its definition, is its property that its divergence depends only on the free charge
density. The dimensions of D are the same as those of P, and thus D will be measured
in coulombs/(meter) 2 . We can think of (10-41) as an expression of Coulomb's law for
the force between point charges plus the electrical effects of matter.
Now that we have (10-41), we can quickly find some of the properties of D. The
boundary condition satisfied by its normal component can be found from (10-41), (9-6),
(9-7), and (9-24) to be
n (D 2 - D J = D2n - Dln = of
(10-42)
where Of is the surface density of free charge. Thus, the normal component of D will be
discontinuous only if there is a free surface charge density; this is in contrast to E
whose normal component is discontinuous if there is a surface density of any kind of
charge.
Gauss' law for D is easily found from (10-41) and (1-59) to be
(f>D da=
f
J P f
v
d r = QfM
(10-43)
where Qf in is the net free charge contained within the volume V surrounded by the
closed surface S. We note both the analogy to and the contrast with the corresponding
result for E given by (4-1). Equation 10-43 can often be advantageously used for the
calculation of D for problems of sufficient symmetry, in much the same manner as we
did for E in Chapter 4.
Even though (10-41) contains only the free charge density, this does not mean that
free charges are the only source for D, since, according to the Helmholtz theorem of
Section 1-20, we have found only one of the source equations. The remaining one is
V X D . This is easily found by taking the curl of the definition (10-40) and using (5-4);
the result is
V XD = v X P
(10-44)
so that D can have sources in bound charges as well as in free ones. The boundary
condition satisfied by the tangential component of D can be found most easily from
(9-21) and (10-40) to be
D2, - D1( = P2, - Px,
(10-45)
It is also possible to devise a cavity definition for D in a manner analogous to that

for E as illustrated in Figure 10-7. The definition is based on the continuity of the
normal components of D in the absence of free surface charge, as stated in (10-42). For
this purpose, we imagine a cavity in the form of a small right cylinder whose height is
very small compared to the radius of its base. It is cut out of the dielectric so that its
base is perpendicular to D in the dielectric as shown in Figure 10-12. If we consider a
point near the center of the cavity, the edges will be too far away to affect the fields so
that D c (in the cavity) will be parallel to D (in the dielectric), and since, by construction, they have only normal components, D c = D. Then E c = D c / e 0 = D / e 0 since
P = 0 in the vacuum of the cavity. If we now imagine putting a small test charge Sq in
the cavity and measuring the force SFC on it, we will have 8FC = Sq E c = Sq D / e 0 , so
that D = 0(<5Ff/<5<7). Thus, in principle, we can find D in the dielectric from measurements made in the cavity.
1 0 - 5 THE D FIELD
153
ID
Figure 10-12.
dielectric.
Cavity used to measure D in a
Figure 10-13. D fields on z axis due to a uniformly

polarized sphere.
In spite of their apparent simplicity, (10-41) and (10-44) clearly will not be too useful
until we can relate E, P, and D better; we will do this in the next two sections.
Nevertheless, we can still briefly look at an example.
Example
Uniformly polarized sphere. This is the system we discussed in the last section for which
the results are summarized in Figures 10-8, 10-9, and 10-11. We would like to obtain
similar things for D; we will continue to consider only field points on the positive z
axis. Outside the sphere where there is no matter, P = 0, and we find at once from
(10-40) and (10-32) that
D
zo(
) =
0 zo
2 Pa3
=
3z:
(10-46)
The value of Dz inside the sphere as obtained f r o m (10-40) and (10-37) is

Dzi(z)
= e0Ezi
+ P=
-iP
+ P = f/>
(10-47)
which is seen to be independent of z and in the same direction as P. These results are
illustrated in Figure 10-13. Since there is no free charge on the surface, we expect the
154
E L E C T R O S T A T I C S IN T H E P R E S E N C E O F M A T T E R
normal components of D to be continuous there according to (10-42). We can verify

this a n d check our results at the same time. W e get, as expected, D2n Dln
= Dzo(a) - Dzi(a) = ( 2 P / 3 ) - ( 2 P / 3 ) = 0.
10-6 CLASSIFICATION OF DIELECTRICS
As we mentioned at the end of Section 10-1, we generally expect that there will be a
functional relation between the polarization and the electric field, that is, P = P(E) or
Px = PX(EX, E , Ez) and so on. Macroscopic descriptive electromagnetic theory does
not predict the form of these functions, but takes them as external information. F r o m
this point of view, these relations are left to be determined f r o m experiment or to be
calculated theoretically from the microscopic properties of matter in other branches of
physics such as statistical mechanics and solid state physics. This does not leave us in a
hopeless position, however, because a combination of experiment and general theory
shows us that most materials fall into easily classifiable groups, and this result can be
used to simplify our theory and make it more useful. It is desirable to do this step by
step in order to aid our understanding of the limitations of our final form.
1. Permanent Polarization
If E = 0, there are two possibilities for the value of P(0). If P(0) = 0, then the material
is polarized even in the absence of a field and, as we noted before, it is said to have a
permanent polarization and is called an electret. Although electrets do occur, we will not
consider them further in this section. The situation for which P(0) = 0 is more typical
and the sort of thing we expect when we think of the polarization as being produced by
the field; we shall generally use the term dielectric only for this latter case.
2. Nonlinear Dielectrics
Even with P(0) = 0, it is still possible that the relation between P and E can be quite
complicated. For most materials, however, this usually requires quite exceptional
conditions such as very large fields, or low temperatures, or both. Thus, it is found that
it is often sufficient to write P as a power series expansion in the components of E, that
is, we can write
p, -
+
j
(10-48)
where the indices i, j, and k take on the values x, y, and z; we note that this form
satisfies our assumption that P(0) = 0. The specific values of the coefficients a, y ,
will depend upon the particular dielectric involved. If the second-order or higher terms
in the components of E are required to describe the material adequately, the dielectric
is called nonlinear. It requires experiment to determine whether (10-48) is necessary in a
given case; for example, some ceramics fall into this category. We will not consider
nonlinear dielectrics further, but now restrict ourselves to cases for which only the first
term in (10-48) is required; such materials are called linear dielectrics.
10-6 CLASSIFICATION O F DIELECTRICS
155
3. Linear Dielectrics
In this case, the general expression relating the components of P to the components of
E can be written in the form
Px
ô(XxxEx
y = *o(XyxEx
XxyEy
Xxz^z)
+ XyyEy + XyzEz)
Pz = *o{XzxEx
+ Xzy
y + Xzz
(10-49)
z)
where the proportionality factors x / y are called components of the electric susceptibility
tensor. We have introduced the factor e 0 so that the xtJ will be dimensionless as can be
seen f r o m (10-40). In general, the x , , need not be constants but may be a function of
position within the material. The Xtj cannot depend on E because that would bring us
back to the nonlinear case of (10-48). We see from the form of these relations that P
will not be parallel to E even in linear dielectrics, nor will D generally be parallel to E.
This situation is quite common in crystals and leads to such phenomena as double
refraction. We can now proceed to our next simplifying assumption.
4. Linear Isotropic Dielectrics
We now assume in addition that at a given point the electrical properties of the
dielectric are independent of the direction of E; such a condition is known as isotropy.
Since one direction is completely equivalent to any other, P must necessarily be parallel
to E, the Xtj = 0 if / = j, and \ x x = xyy = Xzz> s o that (10-49) can be written in terms
of a single factor of proportionality as
P = Xee0E
(10-50)
where x e is called the electric susceptibility. Combining (10-50) with (10-40), we find
that
D = (1 + X e ) e o E =
eE
(10-51)
where
KE= 1 + XE
dielectric constant = relative permittivity
e = Ke0 = (absolute) permittivity
(10-52)
(10-53)
The quantities x<?>

and will characterize the electrical properties of the material
and are to be found by experiment; their values can be found in many tables. For all
known substances, xe i s positive for static fields and therefore Ke > 1. We see from
(10-51) that D and E are parallel in this situation. The equation D = eE is called a
constitutive equation; it is not a fundamental equation of electromagnetism but is
applicable only where it turns out to be applicable, so to speak.
In this case of linear isotropic dielectrics, we can also find the differential equation
satisfied by the scalar potential. Using (5-3), we can write (10-51) as D = e v<t>, and if
we substitute this into (10-41), and use (1-115) and (1-45), we obtain
V (eV4>) = V2< +
Ve = ~ Pf
(10-54)
as the equation we would have to solve for <f> since we still must allow for the possibility
that is a function of position, and we can go no further until this dependence is
known.
Our next simplification leads to such an important situation that we will allot it its
own section.
156
10-71 LINEAR ISOTROPIC HOMOGENEOUS (l.i.h.) DIELECTRICS

We now assume in addition that the electrical properties are independent of position;
such materials are called electrically homogeneous. Generally gases and liquids, as well
as many solids, fall into this category so that it is not such a special situation as one
might think. Then the quantities
e , and are constants', they are still characteristic
of the material, however. Equations 10-50 through 10-53 are still applicable and, in
addition, (10-54) simplifies to
V 2 j)= -
(10-55)
e
since now Ve = 0. By comparison with (5-15), we see that for l.i.h. dielectrics the
potential
again satisfies Poisson's equation but with e replacing e 0 and pj replacing
the total charge density p. (The bound charges of the material haven't disappeared, of
course, but their effect is now summarized in the factor e.) The fact that <p satisfies the
equation (10-55) also means that for l.i.h. dielectrics we can, with care, bodily take over
solutions we have previously found for the vacuum case and simply replace e by e 0 ; we
will do little of this, however. Also, if pf 0, then 4> will again satisfy Laplace's
equation v 2<j> = 0 in this region.
The boundary conditions at a surface of discontinuity can now be expressed
completely in terms of E. If we insert (10-51) into (10-42), and remember that (9-21) is
still applicable, we obtain
n-
( c
E
22 2 2
t x E j
1 1
af
'
E2, - E u = 0
(10-56)
But now we see that, even if oy = 0, the normal components of E will not generally be
continuous across the bounding surface separating two dielectrics, so that, as shown in
Figure 10-14, the direction of E can change at the boundary. Thus, the lines of E can be
refracted even in the absence of a free surface charge density and a 2 will be different
from
In a l.i.h. dielectric, the bound and free charge densities are related in a simple way,
as are the polarization and the displacement. If we eliminate E between (10-50) and
(10-51), and use (10-52), we find that
( . " 1)
P = D =
(10-57)
D
Figure 10-14. Electric fields at a boundary

between two dielectrics.
r
1 0 - 7 LINEAR I S O T R O P I C H O M O G E N E O U S (l.i.h.) DIELECTRICS
157
which shows us that P and D are also parallel a n d that | P | < | D | . N o w if we take the
divergence of (10-57) and use (10-10) and (10-41), we get
(k 1)
p =
p,
(10-58)
so that \ph\ < |py|. If we insert this result into (10-38), we find that the total charge
density in a l.i.h. dielectric can always be written as
P = = - ~0jLj
Ke
Ke ~ 1
(10-59)
which shows us that the total charge density is always less than the free charge density
since KE > 1. As a special case, we see that if pf = 0, then ph = 0, so that at any point
in a l.i.h. dielectric where there is no free charge density, there is no b o u n d charge
density either.
F r o m now on, we will assume that we are dealing with l.i.h. dielectrics in our
examples and exercises unless we explicitly specify otherwise. At this point, we are able
to discuss some examples quantitatively; we begin with the capacitor, which we first
considered qualitatively in Section 10-3.
Example
Parallel plate capacitor with charges constant. In Figure 10-15, we illustrate a capacitor
of total free charge Qf with a vacuum between the plates in ( a ) and with a dielectric
completely filling the region between the plates in (b). The directions of the various
field vectors are also shown. The vacuum value of the electric field was discussed in the
text immediately preceding (6-40) and was found to be E0 = oy/e 0 where oy = Qf/A is
the free surface charge density and A is the plate area. Since P 0 = 0, we find from
(10-40) that the displacement D0 is
D0 = e0E0 = af
(10-60)
N o w 2 / and oy are kept constant when the dielectric is put between the plates, so that
D will not be changed and will equal the vacuum value:
D = D0 = oy
(10-61)
This result is also consistent with (10-42) since the fields are zero inside the conducting
plate a n d therefore D2n Dln = D 0 = oy. T h e electric field has changed, however,
since we find from (10-51), (10-61), and (10-60) that
D
E* =
=
e
K
ec0
DN
i
(a)
(10-62)
-o,
(b)
Figure 10-15. Parallel plate capacitor of constant charge, (a) Vacuum between the plates.
(b) Dielectric between the plates.
158
and therefore E < E0 in agreement with (10-24). W e see now that the factor by which
the electric field is reduced is exactly equal to the relative permittivity. The potential
difference now is
r
A#> = /
EQd
A<^>q
E ds = Ed = - z - =
J+
KE
(10-63)
KE
with the use of (10-62) and (10-22). Thus, the potential difference is less than the
vacuum value by the same factor Ke so that A<#> < A<#>0 in agreement with the
experimental result. We can find the capacitance by expressing (10-63) in terms
of the total charge:
Ead
A<>- -
ofd
-
d
Kê0A
Qf
and therefore
Ke0A
C = -L^
= KeC0
(10-64)
where we have used (10-20) and (6-41) to identify the vacuum value of the capacitance
CQ. Thus, we see that the presence of the dielectric has increased the capacitance, in
agreement with (10-21), and the ratio of the capacitance with and without the dielectric
is exactly equal to the dielectric constant, that is, C/C0 = xe.
T h e polarization as found from (10-50) and (10-62) is
P =
Xee0E
= (Ke - 1 )e0E
(10-65)
With the further use of (10-62), the third form can be rewritten as
P = c0(E0-E)
(10-66)
so that
E = E0
P
e
(10-67)
which also follows from (10-40) and (10-61), that is D = e0E + P = D0 = e0E0. This
latter result has exactly the form we previously deduced in (10-25) and now we have a
quantitative expression for Eb. Since P is constant, V P = 0 and there are no b o u n d
volume charges in agreement with (10-58) since the only free charges are on the
surfaces of the conducting plates. However, there are b o u n d surface charges on the
dielectric surfaces. Their magnitude as found f r o m (10-8), (10-65), and (10-60) is
m = p ~
O,
'f
(10-68)
and the signs are exactly those already illustrated in Figure 10-6. We see f r o m the figure
that the sign of a h is always opposite to that of the immediately adjacent oy; taking this
into account we can write (10-68) as
a,
- 1
=
a,
7
(10-69)
which is completely analogous to the result previously found for volume charge
densities in (10-58). These bound surface charges, acting as two plane infinite sheets,
will produce the field Eb. W e can calculate the magnitude of Eh by using our previous
1 0 - 7 LINEAR I S O T R O P I C H O M O G E N E O U S (l.i.h.) DIELECTRICS
159
result (3-12) along with (10-68), and we find that

e
\L_M_L
+
zc
ze
(a0
.70,
This agrees exactly with (10-67), which we f o u n d by other means, and verifies our
previous analysis which led to the form E = E0 Eh given in (10-25). This field Eb is
often called the local field and one says that the resultant field E is the sum of the
vacuum field E 0 , produced by the free charges as if there were no matter present, plus
the local field E l o c = E^ produced by the b o u n d charges arising f r o m the polarization
of the dielectric.
Example
Capacitance in general. Although we obtained (10-64) by considering the specific case of
the parallel plate capacitor, the simplicity of this result, and the fact that C = KeC0 no
longer contains any characteristics of the parallel plate capacitor, suggests that this may
actually be a general relation that holds for any capacitor. This turns out to be the case.
If we combine (10-41) and (10-51), we get
V E = = ( )
(10-71)
since c is constant for a l.i.h. dielectric. We also have V X E = 0 since E is still

conservative. Suppose we are given a certain free charge distribution described by pf. If
there is no matter in the region of interest, we solve the pair of source equations to find
an electric field E 0 . N o w if we fill all of the region with a dielectric described by KC,
and if we keep py- unchanged, then we see from (10-71) that the problem is exactly the
same as the vacuum case we solved, except that the source charges are smaller
everywhere by the factor KE. Then, as we can see from (3-3) for instance, the field E we
now get will also be smaller by this same factor, that is,
is a general result. The potential difference A<p between the capacitor plates as given by
(6-38) will be
I-
c
E ds = /
J
o* ^
=
Ke
Ke
(10-72)
as in the special result (10-63). We note that this result is consistent with our comments
following (10-55).
Since Qf is the same in both cases, the capacitance will be
C = <
T ~ = I T 1 = " eA 0
A<f> A<f>0
(10-73)
'
thus showing in the general case that the capacitance of any capacitor is increased by
the factor n e when all the space between its plates is filled with a dielectric. The use of
this result provides a convenient way of measuring Ke.
If the dielectric is not homogeneous, or if not all of the space is filled with dielectric,
(10-73) is not generally true. One can often treat such problems by writing the potential
difference as
f
A<f> =
J+
r - D ds
E ds = I
J+
e
(10-74)
160
Then it is usually possible to express D in terms of the total free charge Qf by means of
(10-41), or by (10-43) if the problem has enough symmetry, so that when the integral is
evaluated, the capacitance can be found from (6-38). In such a case, the boundary
condition (10-42) often is of great help. M a n y of the exercises involve these considerations.
Example
Point charge in an infinite dielectric. Suppose a point charge q is embedded in a
dielectric as shown in Figure 10-16. The field of this charge will polarize the dielectric.
If the dielectric has a finite size, the bound charges on the surface will contribute to the
resultant field and the problem of calculating the field everywhere could be quite
complicated. However, if the dielectric is of infinite extent, the effect of any b o u n d
surface charges on the outer surface can be neglected and we can assume spherical
symmetry. Then we can write D = D{R)R and use Gauss' law for D as given by
(10-43). If we integrate over the sphere of radius R shown dashed in the figure, we get,
in the by-now-familiar way,
(^)D dsi = <>DR da R AmR2D <2/
in
= q
so that
D = - 2 r2
ATTR
and
E = -
AITCR
and, as expected, the electric field is just that of a point charge reduced by the factor
e / e 0 = Ke. If we now imagine a point charge q' placed at R, the force on it will b e
F
' - 'E " ^
(10
"76>
which is simply Coulomb's inverse square law with e replacing e 0 .

This result (10-76) is the basis for a common statement that the presence of a
dielectric decreases the forces between charges by the factor e / e 0 = KE. We see now,
however, that this is true only for a l.i.h. dielectric of infinite extent, or one so large that
the b o u n d surface charges will not affect the field a n d we can safely assume spherical
symmetry. In fact, it is possible for the force on q' to actually increase under the right
circumstances.
The polarization of the dielectric can be found f r o m (10-57) and (10-75) to be
P =
(ic - l ) # R
^
Ke47tR2
V e
(10-77)
V
7
1 0 - 8 ENERGY
161
and is directed radially from q. According to (10-58), the bound charge density should
vanish. It follows from (10-10), (10-77), and (1-145) that this is indeed the case.
10-8 ENERGY
We recall our result (7-10) for the energy of a system of charges:
P<t) dT
*4=2 /
(10-78)
all s p a c e
When we obtained this expression, we calculated it as the amount of reversible work

required to assemble a given configuration of charges, and we did not make any
distinction between free and bound charges, nor was it necessary to do so. The
implication was that this energy could be retrieved from the system as reversible work,
and in that sense it is appropriate to regard it as potential energy as we did. Now when
we consider charges in the presence of matter, we have to ask ourselves: what is the
useful definition of energy? Again, this must be the work required for us to put together
an assembly of charges over which we have control and thus energy that can, in
principle, be retrieved from the system. But the only charges over which we have any
sort of control are the free charges and it is only these in a broad sense that we can use
to store or extract energy, as in the process of charging or discharging a capacitor in a
reversible manner. Hence, the energy of interest to us is that of the free charge
distribution in the presence of matter. Failure to remember the restrictive nature of this
definition leads to, and has led to, a considerable amount of confusion and argument
about energy relations when matter is involved.
In order to appreciate the limitations of our final result, it will be useful to begin
again and (5-48), which gives the energy of a point charge in terms of the potential,
provides a convenient starting point. We assume that we have some preexisting
situation and increase the free charge density in a volume element dr by the amount
8pf. Then 8pf dT is small enough to be treated as a point charge and, by (5-48), the
energy of the free charge distribution will change by
8Ue = <f> 8pf dr
(10-79)
The bound charges of matter are included, of course, since </> is the potential resulting
from all charges; for example, it could be evaluated by substituting (10-38) into (5-7).
As usual, we are assuming rigid mechanical supports or constraints on the charge
positions so that no purely mechanical work is done. If we now integrate over all space,
we get the total change in the electrostatic energy:
8Ue = f
<}> 8pf dr
(10-80)
all s p a c e
During this process, D will have changed and by (10-41), we initially have V D = pf
and, finally, v (D + SD) = Pf + 8pp so that by using (1-114), we get v <SD = 8pf.
Then (10-80) becomes
8Ue = f
4>(v-8D)dT
(10-81)
âll space
If we now compare this expression with (7-22), we see that they are of exactly the same
form with SD replacing | e 0 E . The arguments that led us from (7-22) to (7-28) can be
applied in a completely analogous manner, so that (10-81) can also be written as
8Ue = f
till s p a c e
E 8D dr
(10-82)
162
Finally, if we take the zero of energy to correspond to D = 0, we can obtain the total
energy Ue by integrating this from the initial state of D = 0 to the final value:
Ue=
fE-8BdT
(10-83)
all s p a c e * 0
In general, (10-83) cannot be evaluated further until the dependence of E upon D is

known,, and this could be quite complicated. However, for the important case of linear
isotropic dielectrics, we can use (10-51) to write E S D = D 8D/e = ( D 2 / 2 e ) so
that
/D
i0 >
rD / D 2 \
D2
= i > h 7
- I 7 - ^
and thus (10-83) becomes

Ue = f
E dr
(10-84)
all s p a c e
and gives the total reversible work done on the free charges. W e can interpret (10-84)
just as we did (7-28) by introducing an energy density ue given by
D2
ue = D E = i f E 2 =
(10-85)
for linear isotropic dielectrics.

Example
Spherical capacitor. We imagine all space between the plates of the capacitor shown in
Figure 10-17 to be filled with a dielectric of permittivity e. The total free charge on the
capacitor is Q. As usual, D will be radial because of the spherical symmetry, so that | D |
will b e constant on the surface of the dashed sphere of radius r. Then, if we apply
(10-43) to this sphere, we get, as in the last section:
(f)D da = Airr2D = Qf
J
s
so that D = Q/Airr2
in
= Q
and (10-85) gives

U
Q:
=
,4
3277 2tr
(10-86)
-Q
/
h \
11!
/
Figure 10-17. Spherical capacitor with dielectric b e t w e e n the plates.
10-8 ENERGY
163
This applies to the region between the spheres; ue = 0 elsewhere since the fields are
different from zero only in the region occupied by the dielectric. Inserting (10-86) into
(10-84), and using (1-99), we find the total energy of the capacitor to be
r2 7T fw rb 1
Q2 / 1
1\
2
f
f
f
r
sine
drdd
dtp
=
f
- - - (10-87)
J
j
327r t o q 'a r
87n \ a
bJ
If there were a vacuum between the plates, the energy Ue0 would be given by (10-87)
with e replaced by e 0 . Therefore, since e / e 0 = Ke, we see that
Ue =
Q^
Ue=
UeO
(10-88)
so that Ue < Ue0 and the total energy is decreased by the presence of the dielectric in
this case in which Q is kept the same.
In (7-21), we found that the energy of a capacitor is equal to Q2/2C. This result was
obtained by using very general properties of the charge distributions on conductors,
and thus is generally true whether there are dielectrics present or not; this same result
was also obtained by using only general properties of work and potential difference in
Exercise 7-3. Thus, we can still apply energy methods to the calculation of capacitance,
and this is sometimes more convenient. In this example, if we equate (10-87) and (7-21),
we find the capacitance to be
4 nreab
c
= T
= KeC0
(10-89)
o a
with the use of (10-53), and where C 0 is the vacuum value as given by (6-37). As it
should, this specific result agrees with our general conclusion about the effect of a
dielectric on capacitance as expressed in (10-73).
In general, and as we just saw in the last example, the presence of a dielectric can
alter the values of D and E everywhere so that the total energy as given by (10-84) can
be expected to change. The exact amount of this change will normally depend on the
particular manner in which the process is carried out; for example, charges kept
constant, potentials kept constant, only part of space filled with the dielectric, more
than one dielectric used and so on. As a result, the general problem of energy involving
dielectrics can be quite complex, and it is not always possible to make an unambiguous
assignment of the energy change to specific parts of the system. There is one comparatively simple case, however, in which it is possible to ascribe the energy change to the
dielectric itself, and we shall consider it as an illustration.
Let us assume that initially we have vacuum everywhere and some distribution of
charges which result in the fields E 0 and D 0 = e 0 E 0 throughout all space. The initial
energy UeQ can then be evaluated from (10-84) as
Ue0=\f
' a l l space
D 0 E 0 dr
(10-90)
Now let us assume that we keep the source charges fixed in values and locations and
introduce a dielectric of volume V into this preexisting field E 0 . (This is in contrast to
the previous example where all of the region containing the already existing field was
filled with a dielectric.) As we know, the presence of a dielectric will generally change
the values of E and D everywhere and the new energy Ue can be found from (10-84) by
using these new values of the fields. The change in the energy, Ue Ue0, in this case,
can be ascribed entirely to the presence of the dielectric, and if we call it Ueb, we have
Ueb=Ue-Ue0=\(
(D
' a l l space
E - D0
E 0 ) dr
(10-91)
164
It can now be shown that under these conditions, (10-91) can be written as an integral
over the volume V of the dielectric only. Since this is a rather lengthy calculation, we
have to leave it to an exercise and simply quote the final result, which turns out to b e
U
eb
~ i f
J 1/
E 0 dT
(10-92)
Since this involves only the volume of the dielectric, it is reasonable to consider it to be
localized in the dielectric and to represent the energy of the dielectric. Thus we can
introduce an energy density for these b o u n d charges, ueb, given by
ueh=
E0 = - i x e e 0 E E0
(10-93)
where we also used (10-50). These expressions (10-92) and (10-93) are thus appropriate
for a situation in which the polarization is thought of as being produced by the field.
In all of our discussion so far, we have assumed that Ke is constant during these
processes. It turns out that many dielectrics have a. Ke that depends on the temperature,
as we will see in Appendix B, so that here we are effectively assuming isothermal
processes. This is consistent with our emphasis on UE as being related to the reversible
work, so that it is really more analogous to the Helmholtz function or free energy of
thermodynamics, but for systems at constant temperature, there is no distinction
between their changes. With this in mind, we can look on (10-92) as a contribution to
the internal energy of the dielectric system.
We should also remind ourselves of the distinction between these energies and the
interaction energy of a charge distribution and an external field, which we discussed in
Section 8-4. In particular, we obtained (8-64) for the energy of a dipole. If we want to
apply this to a polarized material, we will have to assume that the polarization is either
a permanent one or that the external field is small enough so that it does not affect P
appreciably. Then the dipole moment of a small volume will be given by (10-1) and the
external interaction energy as obtained from (8-64) can be written as
dUe^t=
P ' Eext d r
(10-94)
where we write the external field as E e x t , rather than E 0 that we have been using
recently to represent vacuum values. The total interaction energy will then be obtained
by integrating (10-94) over the dielectric to give
*4,ext= - / P - E
e x t
^T
(10-95)
For example, if E e x t does not vary much over the volume, we can take it out of the
integral and use (10-2) to get UE EXT = p E e x t in agreement with (8-64).
Example
Energy of a capacitor in general. We have already seen the effect of a dielectric on a
capacitor in the special case by which we obtained (10-88). Let us now briefly look at
the general case. The energy is general is given by (7-21). In addition, we have obtained
the general relation (10-73) for the effect on the capacitance. Therefore, if the charge Q
is kept constant, the energy UE with the dielectric between the plates will be UE =
Q2/2C
Q2/2KcC0
= U0/NE where UQ is the vacuum energy, so that the decreased
energy given by
U0
UE =
(Q = const.)
is a completely general result, in agreement with (10-88).
(10-96)
10-9 FORCES
165
N o w let us suppose, instead, that the potential difference A<j> is kept constant.
Combining (7-21) with (10-73), we find that Ue = jC(A<p)2 = \iceC0(A<p)2 = KeU0 SO
that
Ue = KeU0
(A< = const.)
(10-97)
and the energy of the capacitor has increased by the factor KE, in contrast to (10-96).
The reason for this is that, with the increase in capacitance, the charge has also been
increased and the battery had to do work on these free charges in order to separate
them. In addition, it had to polarize the dielectric and the net effect of all of these
changes is just that described by (10-97).
10-9 FORCES
W h e n a dielectric is polarized, the resultant b o u n d charge densities will have forces on
them due to the electric field. This general subject of forces on dielectrics, and on
conductors in the presence of dielectrics, is really quite complex and it is easy to get an
incorrect answer. Generally speaking, the only satisfactory way of handling these and
similar problems is to use "energy methods," that is, by comparing the energy of the
initial configuration of the system with that of the final configuration. Usually such
problems fall into two classes: (1) the system of interest is completely isolated and thus
its total energy is conserved or (2) the system is n o t isolated and one has to consider the
possibility of energy transfer between it and external energy sources such as batteries;
thus the system energy will not be conserved, although that of the system plus battery
will be. Ordinarily, these two classes will correspond to our previous cases of constant
free charge or constant potential difference.
Because of these considerations, we will content ourselves here with discussing only
two of the simplest situations, although other c o m m o n examples will be found in the
exercises. Furthermore, we will not discuss the balance between electrical and mechanical forces that is required to maintain mechanical equilibrium within the system, or to
p r o d u c e a new one when electric fields are applied. If the dielectric is not rigid, it will
generally deform under the influence of these electrical forces. This phenomenon is
called electrostriction; it is generally a small effect, but can be of interest and
importance under some circumstances.
Example
Average surface force on a dielectric. Let us consider the particular case of a capacitor.
As we saw in (10-96), the energy of the capacitor for constant charge will be decreased
when the dielectric is present. Since the general tendency of systems is to reduce their
energy, we see that the capacitor will " w a n t " to have the dielectric in place. In other
words, there should be a force on the dielectric whose direction is such as to pull it into
the region between the plates.
In order to be specific, we consider a parallel plate capacitor with square plates of
side L so that A = L2. W e also assume we have a solid slab of dielectric of the correct
size to just fit between the plates. We neglect edge effects and assume the field to be
different from zero only in the region between the plates. In Figure 10-18, we show a
side view in which the dielectric is partway between the plates. With the dielectric
completely in place, the total change in energy can be found from (10-96) and is
AUe =Ue-U0=
(10-98)
166
iI
Figure 10-18.
dielectric.
Force on a
We let ( F ) be the average force on the dielectric; the total displacement of the
dielectric is L, so that using (7-37) we get
<f> = -
AUe
{ Ke - 1 \ U0
) T
(10 99)
"
which is positive, showing that the dielectric will be pulled into the region between the
plates. If we express this in terms of the original constant energy density ue0, we find
from (10-84) and (10-85) that U0 = ue0 (volume) = ue0L2d so that
N o w Ld is the area of the face of the dielectric u p o n which ( F ) is acting, so that if

( f a ) is the average force per unit area we get
(fa)=
(10-100)
as expressed in terms of the initial vacuum situation.
Example
Capacitor immersed in a fluid dielectric. As an example of the force on a conductor in
the presence of a dielectric, we consider a plate of a parallel plate capacitor for which
all of the region where the field is different f r o m zero is occupied by a dielectric. In
Section 7-4, we evaluated this by looking at the energy changes produced by and the
force required to change the plate separation by a small amount. Our treatment was
general enough so that we can use the same results a n d need only supplement them
with our knowledge of the effect of a dielectric on capacitance as given by (10-73), that
is, C = KeC0. If the charge is kept constant, the total force on the plates in the presence
of a dielectric will still be given by (7-39), and we get
Q2
dC
Q2
dUeC0)
Feo
in terms of the force

when there is a vacuum between the plates. Thus the force on
the plates has been reduced by the factor ne. In using C = KeC0 in this way, it is
essential that the dielectric completely fill the region between the plates, as will be
illustrated by some of the exercises. This means that we are in effect assuming the
dielectric to be a fluid and that the capacitor is immersed in the fluid. Then, if the
separation is decreased, the fluid can move out f r o m between the plates, while if the
separation of the plates is increased, the fluid can move in to fill the region and there is
also fluid available for this purpose.
EXERCISES
167
If the potential difference is kept constant, the middle f o r m of (7-46) is appropriate

to use, and we get
, dC
7d(icC0)
Fe = i(A<#>)2 = l(Afl)2
d x
= K eFe0
(A <j> = const.)
(10-102)
and, in this case, the force on the plate is increased by the factor tce.
Even though we realize that there are different processes involved, it still may seem
somewhat paradoxical that the presence of the dielectric decreases the total force in one
case while increasing it in the other. If we recall (7-37) and (7-45), we see that we should
be able to understand these results somewhat more by looking at the total field energy.
W e found in (10-85) that the energy density is given by ue = j D E. If the free
charge Q is kept constant, then D will be constant and we find that ue = D 2 / 2 e =
u
eo/Ke The use of (10-84) then gives Ue = Ue0/ice and the force will decrease by the
factor KE as in (10-101). Similarly, if the potential difference A< is kept constant, then E
will b e constant, and we are led to u e = f e E 2 = k e u e 0 . Thus, in this case, we will find
that Ue = KeUe0 which corresponds to the increase in the force described by (10-102).
[We n o t e that these results for the capacitor energy are exactly those we found by a
somewhat different approach as given by (10-96) and (10-97).]
F u r t h e r study as to exactly how these mechanical forces are affected by the presence
of a dielectric shows that the pressure distribution within the fluid is altered in the
presence of an electric field. The resulting pressure changes can be shown to account in
detail for the changes in the force on the conductor.
EXERCISES
10-1 The permanent dipole moment of the water Similarly, find the net gain due to motion of
molecule is about 6.2 X 10" 30 coulomb-meter. negative charge of density p+ (Why?) and averWhat is the maximum polarization possible for age displacement R_. Combine the results for
water vapor at 100C and atmospheric pressure? all faces found in this way and thus show that
10-2 Electrostatics began with the observation the net gain of charge per unit volume is ph =
that charged objects attracted small pieces of - V - P.
matter. Describe qualitatively how a charged body 10-5 Find the potential <p and Ez on the axis
can exert a net force on neutral matter and show produced by the uniformly polarized sphere disthat the force will be attractive as observed.
cussed in Section 10-4 for negative values of z.
10-3 A slab of material has parallel faces. One Show that your answers are consistent with the
coincides with the xy plane while the other is results found for z > 0 and with Figure 1 0 - 1 1 .
given by z = t. The material has a nonuniform 10-6 Find the total positive bound charge of the
polarization P = P(1 + az)z where P and a are uniformly polarized sphere of Figure 10-8.
constants. Find the volume and surface densities
of bound charge. Find the total bound charge in a 10-7 A sphere of radius a has a radial polarizacylinder of the material of cross section A and tion given by P = ar" r where a and n are consides parallel to the z axis and thus verify directly stants and n > 0. Find the volume and surface
densities of bound charge. Find E outside and
that ( 1 0 - 1 3 ) holds for this case.
inside the sphere. Verify that your results for E
10-4 Consider a parallelepiped of volume
satisfy the appropriate boundary conditions. Find
A x A j A z fixed within a dielectric. If p + is the <j> outside and inside the sphere. Skctch your
average positive bound charge density, and R + is
results for E and <j>.
the average displacement of these charges when
the material is polarized, show that the net gain of 10-8 Repeat Exercise 10-7 for the case n = -1.
positive charge through the faces parallel to the Also show that n = - 2 is not a possibility because of ( 1 0 - 1 3 ) .
yz plane will be
d
10-9 A cube of edge la has its faces perpendic- ~(p+R+x)
AxAyAz
ular to the xyz axes and the origin at its center. It
168
is uniformly polarized in the z direction. Find E

at the center of the cube.
10-10 A spherical cavity of radius a is inside a
very large dielectric that is uniformly polarized.
Find E at the center of the cavity.
10-11 A cylinder of length 2 L has its axis along
the z axis and a circular cross section of radius a.
The origin is at the center of the cylinder that is
uniformly polarized in the direction of the axis,
that is, P = Pi where P = const, (a) Find the
bound charge densities ph and afl. (b) Find the
electric field for all points on the z axis for which
z > 0. (c) Verify that your results in (b) satisfy the
boundary condition at z = L. (d) From the result
of (b), find E at the origin, (e) Sketch the result of
(d) as a function of a/L. For what value of a/L
does E at the origin have its maximum magnitude
and what is its value? Is this reasonable?
10-12 An infinite cylinder of circular cross section of radius a has its axis along the z axis. It is
uniformly polarized transverse to its axis, that is,
P = Fx where P = const. Find E at any point on
the axis of the cylinder. (It is easiest to choose the
field point at the origin.)
10-13 Which of the results (10-55) through (1059) are still valid if the dielectric is linear and
isotropic but nonhomogeneous?
10-14 Show that the angles in Figure 10-14
satisfy the relation icf, cot aj = n e7 cot a 2 . Suppose region 1 is polyethylene for which Kel = 2.30,
while region 2 is a glass for which n e2 = 4.00.
Find a 2 if
= 3 6 . Show that the ratio of the
bound surface charge density from region 1 to the
total bound surface charge density is independent
of angle and evaluate it for this case.
10-15 Find the expressions analogous to (10-58)
and (10-59) when the dielectric is not homogeneous and thus show that there can be a bound
charge density even in the absence of free charge.
What additional conditions are necessary to have
Pfy = 0 when pf = 0?
10-16 Two point charges q and q are initially
in a vacuum and are separated by a distance a.
Then a slab of dielectric of thickness d < a is
inserted halfway between them, with the faces of
the slab perpendicular to the line connecting the
charges. Show qualitatively that the total force on
q will be increased.
10-17 A point charge q is located at the center
of a dielectric sphere of radius a. Find D, E, and
P everywhere and plot your results. What is the
total bound charge on the surface of the sphere?
10-18 An infinite dielectric with dielectric constant Kc has a spherical cavity of radius a in it.
There is a point charge q at the center of the
cavity. Find ph and ah. Find the total bound
charge on the surface of the cavity. How can you
reconcile your results with (10-13)?
10-19 The infinitely long coaxial conductors of
Figure 6-12 have the space between them filled
with a dielectric for which
is given in cylindrical coordinates by ap" where a and n are
constants. There is free charge \ j per unit length
on the inner cylinder. Find D, E, and ph everywhere between the conductors. Under what circumstances will the magnitude of E be constant?
What will be the corresponding values of D and
P/,?
10-20 An infinite slab of dielectric of thickness t
and parallel faces has a constant surface density
of free charge cy on one of its surfaces. Find E
everywhere. What is the surface density of bound
charge on the face which has no free charge on it?
10-21 An infinite line charge with a constant
free charge
per unit length coincides with the
z axis. It is coaxial with a dielectric cylinder of
radius a whose dielectric constant varies along
the axis according to Ke = a + fiz where a and (i
are constants. Find D, E, P, and pb at all points
within the cylinder. Is your result for ph consistent with the results of Exercise 10-15?
10-22 How does the introduction of a l.i.h. dielectric into all regions of a general system affect
the values of the coefficients of potential, capacitance, and induction? [These coefficients are defined in (6-11) and (6-43).]
10-23 A slab of dielectric of thickness t is inserted into a parallel plate capacitor of plate
separation d and plate area A as shown in Figure
10-19. The surfaces of the slab are parallel to the
plate surfaces. Find D, E, and P as functions of
x, and plot your results. (Express them in terms
of Q.) Find the capacitance of this system. Verify
that your result for C reduces to the proper
values when t = 0 and t = d.
10-24 Reconsider the example that led to (10-73)
for the case in which the potential difference is
kept constant while the dielectric is put between
the plates. Find what happens to E, D, and Q and
thus show that (10-73) is still correct.
10-25 The region between the plates of a parallel plate capacitor is filled with a dielectric for
which Ke varies linearly with distance from the
value Kel at one plate to /ce2 at the other. Find
EXERCISES
169
-Q
Figure 10-19.
Capacitor in Exercise 10-23.
the capacitance. As a check on your result, show

that C reduces to the correct expression when Ke
is constant. Also show that the result is independent of whether Kel is greater than or smaller
than Kp2.
10-26 The region between the plates of the
spherical capacitor of Figure 10-17 is filled with a
dielectric for which n e varies according to
1 +
r a
where Kea and a are constants. Find the capacitance. Does your result reduce to the correct
value for a = 0?
10-27 The coaxial cylindrical capacitor of Figure 6-12 has two different dielectrics between its
plates. The value of ne is Kel for a < p < p0, and
KC,2 for p0 < p < b. Find the capacitance of a
length L of this system.
10-28 The region between the plates of the
spherical capacitor of Figure 10-20 is filled with
two l.i.h. dielectrics with permittivities shown.
The total volume is divided exactly into halves by
a plane that passes through the common center of

the spherical conductors. Show that the capacitance is given by C = 2n(e1 + e 2 )a6/(Z> a).
10-29 The spherical capacitor of Figure 10-17
now has two different dielectrics between its
plates. The value of ice is Kel for a < r < r0 and
Ke2 for r0 < r < b. Find the capacitance of this
system by finding the total energy of the fields
between the plates.
10-30 The coaxial cylindrical capacitor of Figure 6-12 has a dielectric between its plates for
which the dielectric constant varies as Ke = K0p"
where K0 and n are positive constants. Find the
capacitance of a length L of this system by
finding the energy in the fields between the plates.
10-31 A l.i.h. dielectric sphere of radius a also
contains a free charge of constant density p0.
Find D and E for all points inside the sphere.
Find the total energy of the fields inside the
sphere. What is the total bound volume charge?
Find the potential <j> due to this system for all r.
What is <p a t the center of the sphere?
10-32 The parallel plate capacitor of Figure 1018 has square plates of edge L. When the dielectric is in a distance x, show that the capacitance
as a function of jc is given by C(x) = (c 0 L/d){ L
+ (Kf - l j x ] .
10-33 The parallel plate capacitor of Figure 1018 has square plates of edge L. It is connected to
a battery of potential difference A<f>0 when there is
no dielectric between the plates. While it is kept
connected to the battery, the dielectric is inserted
halfway into the region between the plates. The
battery is then disconnected. After this, the dielectric is inserted until it fills the whole region
between the plates. Find the final values of A<p,
Q, and E.
Figure 10-20.
Capacitor in Exercise 10-28.
10-34 Consider the parallel plate capacitor of

Figure 10-18. Find the total force on the face of
170
the dielectric as a function of x. Does F(x)

increase or decrease as x increases? Show that
when you average your expression for F(x) over
x, the result is (10-99). Hint: recall Exercise 10-32.
the liquid is given by

P/8h(t>2
X
"
a2) In( b/a)
2 e 0 ( A <t>Y
(Do not forget the battery.)

10-38 Derive (10-92). It is best to proceed in the
following way. First, write the integrand in (10-91)
as E (D D 0 ) + E 0 (D D 0 ) + (E D 0 E 0
D). Then, with the use of (5-3), (1-115), (10-41),
(1-59), and arguments analogous to those which
led to (7-27), you should be able to show that the
first two terms in the newly written integrand
each contribute zero to (10-91). Now show that
the third term is zero in the vacuum outside the
dielectric.
Finally, the further use of (10-40) as
10-36 A parallel plate capacitor with square
applied
to
(D
0 , E 0 ) and (D, E) will lead to (10-92).
plates of edge L and separation d is given a
charge Q and disconnected from the battery. It is 10-39 We noted at the beginning of Section 10-9
then placed vertically with one edge immersed in that the forces on dielectrics can be ascribed to
a reservoir of liquid dielectric of mass density p7. electric forces acting on the bound charges.
Neglect edge effects and show that the liquid will Therefore, it may seem puzzling that we found a
force on the dielectric of Figure 10-18 that is in
rise to a height h within the capacitor where
the x direction while the electric field, which was
2 1/2
assumed
for simplicity to be like that of Figure
4 K - i) Q
h =
1 +
10-156,
is
perpendicular to the direction of this
- 1
e
2(*f~ 1)
o gP/L5
resultant force. The origin of the force can be
found in the very thing we neglected, namely the
10-37 Two coaxial conducting cylinders like "edge effects" or curvature of the actual electric
those of Figure 6-12 are lowered vertically into a field lines as illustrated in Figure 6-10. Show
reservoir of liquid dielectric of mass density ph If qualitatively that the net force on the bound
a potential difference A<f> is applied, the liquidsurface charges as produced by the curved field
will rise to a height h between the plates. Neglect lines will really be in the x direction as we found
edge effects and show that the susceptibility of from energy considerations.
10-35 Two square conducting plates of length L
on a side are placed a distance d apart to form a
parallel plate capacitor. They are kept at a potential difference A</> while a slab of dielectric of
thickness t < d that is also of length L on a side
is inserted parallel to the edges of the plates.
Neglect edge effects and find the average force
with which the dielectric is drawn between the
plates.
11
SPECIAL M E T H O D S IN
ELECTROSTATICS
U p to now, we have concentrated on finding the scalar potential 4> by integration over
a given distribution of source charges by using (5-7), for example, and then obtaining
the electric field from E = V<#>. As we noted in the paragraph following (5-14), some
problems are stated in such a way that this is not feasible so that it is desirable to have
other methods available. Such an approach is to regard the problem of finding the
potential as that of solving the partial differential equation satisfied by <p. This is
Poisson's equation, as given by (5-15):
V2 = -
e
(H-2)
If the relevant charge densities are zero, both of these reduce to Laplace's equation
V24> = 0
(11-3)
Because of the comparative simplicity of Laplace's equation we will concentrate on

solving it. We will, however, discuss one example of finding the solution of Poisson's
equation in Section 11-6.
Over the years, many methods have been devised for solving these equations. Some
of these methods are very general and systematic, while others are extremely specialized
and of limited applicability. We will consider samples of both types. The motivation
and justification for much of what we will do is provided by an important theorem that
we consider next.
IfTil UNIQUENESS OF THE SOLUTION OF LAPLACE'S EQUATION
We want to show that, if we have found a solution of Laplace's equation that satisfies
the given boundary conditions (i.e., it reduces to the correct preassigned values on all
points of the surface surrounding the region), then this solution is unique. The term
" b o u n d a r y conditions" is now being used in a different sense from that of Chapter 9.
There it referred to the necessary behavior of fields at a bounding surface of discontinuity between two media. Here we are assuming that we are dealing with a region that is
surrounded by a surface for which the numerical value of the potential is given or
known at all points; that is, we do not k n o w the details of the source charge
distribution outside this region, but we do know the potential produced by it on the
surface. F o r example, part of the boundary may be a conductor kept at a potential of
12 volts by a battery. Then, whatever <j> may be elsewhere, it must reduce to the value
12 volts whenever the field point is located anywhere on the conductor. Sometimes the
b o u n d a r y involved is very far away, that is, at infinity.
171
172
S P E C I A L M E T H O D S IN E L E C T R O S T A T I C S
It will be sufficient for our purposes to deal with a somewhat special case. We
assume that, in addition to satisfying Laplace's equation, < is constant on all points of
the bounding surface S:
<j> = const.
on boundary
(11-4)
Now if we use (1-115), (1-45), (1-17), and (11-3), we find that
= (V<P)2
v (<pv) = v<T> v<T> +
(n-5)
which, when used in (1-59), leads to

j (v<p) 2 dr = / v
(<j>V<p)dT =
da =
E da = 0
(11-6)
because of (11-4) and Gauss' law (4-1) since Qin = 0. As the integrand in the first
integral of (11-6) is a sum of squares, and thereby intrinsically positive, the integral can
be zero only if the integrand itself is zero everywhere; therefore
2
w ' - U r J
M<M2
U ? J
<-7>
according to (1-37). The expression in (11-7) is again a sum of squares, and therefore
the individual terms must be zero, or
d<f>
dcf>
dtp
dx
dy
dz
so that <p = const. But, since </ = const, on the boundary, we see that
<p = const.
everywhere
(11-8)
Now it is easy to prove our uniqueness theorem. We let ^ ( r ) be a solution of (11-3)

that satisfies the given boundary conditions. We also assume that there is another
distinct solution <f>2(r) satisfying these same boundary conditions. We want to prove
that <t>i and <p2 are identical. For this case, we let ^> = <p1 - <p2. Then v 22 = 0 because of (11-3); therefore 1 = 4>2 so that <j> = 0 on the boundary. Since zero is a constant,
this satisfies all of the conditions of the previous result; thus (11-8) shows that 2
everywhere
(11-9)
which is what we were to prove.

Sometimes the boundary conditions are stated in terms of the components of the
field, that is, in terms of the derivatives of the potential rather than its absolute value.
One can prove analogous uniqueness theorems for Laplace's equation that are applicable in such situations; an example is given in Exercise 11-1. Similarly, one can prove a
uniqueness theorem for Poisson's equation, although we do not find it necessary to do
so.
The significance of this result (11-9) from a practical point of view is that, once we
find a solution of Laplace's equation, by any means whatsoever, which satisfies given
boundary conditions, we know that it is the only solution and we do not need to
consider the possibility that there are others. "Any means whatsoever" can include
systematic methods, shrewd guesses, lucky guesses, or simply remembering past solutions and giving them a new twist.
11-2 M E T H O D O F IMAGES
173
Example
N e a r the end of Section 6-1, we considered conductors with cavities as illustrated in
Figures 6-5 and 6-6. If there is no net charge inside, we concluded that the cavity (and
any conductors therein) forms an equipotential volume with the same potential as the
surrounding conductor so that the electric field is zero everywhere within the cavity.
But, since the inner surface of the surrounding conductor has a constant potential,
(11-4) is satisfied and thus the result (11-8) applies to these cases, showing us once
again that the cavity (and any conductor within it) is an equipotential volume.
The first methods of solving (11-3) that we consider are quite interesting, but also
rather specialized.
fVh2\ METHOD OF IMAGES

We recall that Coulomb's law led us to the expression (5-2) for the potential of a system
of point charges, where the contribution of each charge is proportional to 1/R where R
is the distance from the charge to the field point. Therefore, such an expression must
necessarily satisfy Laplace's equation; we also see this explicitly f r o m (1-144). In other
words, the sum of individual potentials f r o m a set of point charges is automatically a
solution of Laplace's equation. This fact is the basis for the method of images. Here the
aim is to find a set of fictitious charges (image charges) that, together with any actual
charges that may be present, will enable us to satisfy the b o u n d a r y conditions and thus
obtain the unique potential function. That is, we try to write the potential in the form
<(,-
"
(11-10)
and find the right combination as best we can. T h e basic idea is that the image charges
will somehow simulate the behavior of the other source charges or of the bulk material;
accordingly, the image charges are chosen to be located outside the region in which we
are trying to find <f>. This method is best illustrated by specific examples.
Example
Point charge and a semiinfinite plane grounded conductor. As shown in Figure 11-1, the
point charge q is at a distance d from the yz plane that is the surface of a conductor
occupying all space to the left of this plane,, that is, for all negative values of x. The
other half of space is a vacuum. The boundary condition is that cf> = const, at x = 0
according to (6-2). For simplicity, we take this constant value to be zero (the conductor
is " g r o u n d e d " ) ; if the actual value is a different constant, we can simply add it to our
final result. Thus, our boundary condition is
4>(0,y,z)
= 0
(11-11)
for all y and z. We will try to use (11-10) to satisfy (11-11) with only a single image
charge q' also located on the x axis a distance d' into the conductor (and thereby
replacing the conductor as far as the vacuum region is concerned). Since the coordinates of q and q' are (d, 0,0) and ( - d0,0),
respectively, we find that (11-10) together
with (5-6) gives
*(x'
z )
' ^ r
[(x -
V + z2]'/2
[(* + d ' f l y * + z 2 ] ' / 2 }

(11-12)
174
Ay
QJ
(x, y, z)
Figure 11-1.
charge.
Point charge and a semiinfinite plane grounded conductor, q' is the image
When this is combined with (11-11), we see that we must satisfy the condition
I
2
(d
+y
+ z )
l/2
4'
(d'
+y
+ z2)1/2
(11-13)
Clearly this will always be satisfied if d' = d and q' = -q. Hence q' is as far
" b e h i n d " the boundary as q is in " f r o n t " of it, so that the term "image" charge fits
very well; we note that the charge sign changed in this processthis is characteristic. If
we now put these values we have just found into (11-12), we obtain the unique
expression for the potential
y
'
z ) =
^ k { [(, -
"
[<*+)2 + V + ^ r )
(11
"14)
and it is the complete solution to the problem. We use (11-14) only for x > 0; for
x < 0, <{> has the same value zero that it has at the surface of the conductor as we know
from (6-1).
We can now find the electric field components from (5-3):
d
dz
qz
47rc
o\[<X
(x + d)
(x - d)
~
+ y2
3 / 2
[(* +
z2Y/2
1
+
+ y2
y2 + ;2J3/2
l
3
d),\2 +, y 2 +i z 2l\ /2
[(x + d)
+ y2 + z2]3/2
1
[(x + d)2 + y2 + z22\
p/2
(11-15)
175
W e can check our solution by seeing whether these have the correct properties. At the
surface of the conductor, E and Ez are tangential components and should vanish
according to (6-2); we see at once f r o m (11-15) that this is correct, for we find that
Ey{0, y,z) = Ez{0, y, z) 0. At the conductor, the normal component of E is
n = n - E = x - E = x and we find that
E.-EJ
o , y , z ) - ~ -7-^777572 = "
27T0{d + y + z )
27re0/<0
(11-16)
where R0 is the distance from q to the corresponding point on the plane x = 0 as is

also shown in the figure. But we know that a nonvanishing En means that there is a
surface density of charge (free charge, in this case), and we find f r o m (11-6) and (6-4)
that it is given by
r(y, z) =
122
- ' (d
2-7T
-577
3 2
,+2
y2 1+, 2z2\) /
v(11-17)
'
This surface charge is said to have been induced by the point charge q. We see that ay
is not constant over the plane; its maximum magnitude of q / l T r d 2 occurs at the origin
directly below q, and oy - 0 as y and z go to infinity. W e can find the total charge
induced on the yz plane by combining (11-17) with (2-16) and (1-55):
qdu t/oo
-00 ROC
roc
dydz
uy u
q ind = ~ 7T" j J
j J
/ 2 22
2 \ 3/2
7T - OO -00
- OO( f( /D ++y2V ++ z27) \
qd
uu F<x>
roo
dz
uz
17 J
J2 . 2 =
IT J - oo D + Z
(H"18)
where we have used our previous results (3-7), (3-8), and (3-12) to evaluate the integrals.
Thus the total induced charge turns out to be equal and opposite to the inducing
charge, a n d hence equal to the image charge as we should have expected since the
image charge simulates the overall behavior of the conductor.
In order to find the force on q, we need the value of E at its location. W e cannot,
however, use the first terms in the braces of (11-15) since they represent the contribution d u e to q itself, as we see from (11-12), and would mean that q is exerting a force
on itselfa possibility we have constantly excluded. Putting the coordinates of q,
(d, 0,0), into the remaining terms of (11-15), we find that Ey = Ez = 0 and
Ex = q/l6m0d2,
so that the force on q is
F
- ^
q2
167Tt0d
(U-19)
and is directed toward the conductor. This clearly represents the resultant force of
attraction between q and the induced surface charge oy as can be verified by direct
integration of (2-17). If we rewrite (11-19) as
q2
4we0(2d)
-x
(11-20)
we see that it is exactly in the form of Coulomb's law for the attraction between q and
the image charge q, as they are separated by the total distance 2d.
T h e equation giving the equipotential surfaces in the region x > 0 is obtained by
setting (p equal to a constant in (11-14), and the equipotential curves in the xy plane
are then found by setting z = 0. In terms of the distances R and R' of Figure 11-1, the
equation of the equipotential surface as found f r o m (11-14) is just
1
R'
4m0<j>
q
= const.
(11-21)
176
Some of these curves are shown as the solid lines in Figure 11-2. The dashed lines are
the lines of E. Their equation can be found from (5-39) a n d (11-15).
Example
Point charge and a grounded conducting sphere. See Figure 11-3. We use spherical
coordinates with origin at the center of the sphere of radius a and the z axis is chosen
to pass through the location of q at a distance d f r o m the center. The boundary
condition is that the potential be zero on the surface of the sphere, that is,
*(fl,0,<p) = 0
(11-22)
We try to solve this problem with a single image charge q' at a distance d' f r o m the
center; we require d' < a so that q' will be outside of the vacuum region. The
potential at any field point P as obtained from (11-10), the law of cosines, and an
177
inspection of the figure is

/
/\
I
4 +
l'
47TC0 \ R
R'
4m0
(r2 + d2 2rd cos 6)1/2
( r 2 + d'2 2 r d ' cos 6)1/2

(11-23)
When this is combined with (11-22), we obtain the condition
q
2
(a
+ d
q'
l/2
- lad cos 6)
(a
+ d'
lad'
cos 0)1/2
= 0
(11-24)
from which we must find the two unknowns q' and d'. Since this will, in general,
require two equations, and since (11-24) must hold for all values of 0, we can get these
equations by using two particularly useful values of 0. We note that we can get rid of
the square roots in the denominator by making each term in the parentheses form the
square of something, and we see that this will occur for the values 0 and it for 6. When
these are put into (11-24), we obtain the two equations
q
q'
+
= 0
(H-25)
d a
a d'
q
q'
+
= 0
(11-26)
d + a
a + d
after using the fact that d > a > d'. Solving these, we find that
a
q ' = - - q
and
d'=
a2
--
(11-27)
In this case, the image charge again has the opposite sign to the inducing charge but is
178
not of the same magnitude, and, in fact, \q'\ < \q\. Substituting these results into
(11-23), we find the potential that satisfies the boundary conditions, and thus gives the
correct value everywhere outside the sphere, to be
4>
(a/d)
=
477e
o \ (r2 + d2 2rd cos 0)l/2
Jr 2 + (a2/4)2
2r(a2/d)
cos # ] 1 / 2
(11-28)
The electric field components can now be found from E = v<f> and (1-101); the
nonvanishing ones are
q
E =
((r-dcos0)
R-
4tre 0
Ea =
- (a2/d)
(a/d)[r
qd sin 6
/3
{a/dY
477
cos 0]
(11-30)
R'3
At the surface of the sphere, Ee(a) = 0 as it should since it is a tangential component;

however, Er(a) # 0, and since this is a normal component at the surface, there is a
surface density of charge given by
q(d2
<.,()-c0l,
Anra{a
+ d
a2)
2adcos0)
"372
(11-31)
We can show again that the total induced charge is equal to the image charge. With the
use of (2-16), (1-100), and (2-22), we get
q(d2
<?ind=
q(d2
a2) riv rT
a2sin0d0d(p
TZ:;
Jo Jo
A IT a
(a2 + d2 2adcos0)3/2
- a2)a
2
n
dfi
J_{T
2 2 +, d,22 2adn)
.^.^3/2
^ i a(a
(11-32)
The integral can be found from tables and is

1
ad {a'
1 / 1
2ad\i)
1/2
ad \ | d a |
1
|d +
In this case, d > a, and (11-33) becomes 2/[d(d2 a2)] so that

a
a = q
*7ind
(11-33)
(11-34)
as we thought.
The charge q will be attracted toward the sphere by a force that can be found as the
Coulomb force between q and its image q'. Their distance of separation is D = d d'
= (d2 a2)/d, so that, according to (2-3),
F = -
adq2 z
4ire0(d2
azy
(11-35)
If d s> a, the force varies approximately as the inverse cube of the distance of
separation.
The general appearance of the equipotentials and lines of E for this case are shown
in Figure 11-4.
Figure 11-4.
179
Equipotentials and electric field lines for the system of Figure 11-3.
Example
Point charge and an insulated uncharged conducting sphere. This is a variation on the
previous example. The sphere is assumed to be originally neutral, and is no longer kept
at a definite potential. In the presence of q, the sphere must still have zero net charge
because it is no longer connected to something f r o m which it can obtain charge as was
the case in the last examples. It must also be a n equipotential volume. We begin by
introducing the same image charge q' = (a/d)q
at the same position as the last
example; this will make the surface of the sphere have constant (zero) potential. But, in
order to keep the sphere neutral, we must p u t another charge q" = q' = (a/d)q
somewhere within the sphere. The only place we can put it and keep the sphere's
surface an equipotential is at the center. Thus, we are led to the system of three charges
shown in Figure 11-5. Since this charge distribution satisfies all of the requirements, it
will lead to the correct potential and field everywhere outside the sphere. The calculation of most of the features of this example will be left as exercises, but we can easily
find the final potential of the sphere. If we combine the results shown in the figure with
180
(7)*
"ft)'
a/
Figure 11-5. Point charge q and

its images for an insulated uncharged
conducting sphere.
(6-5), we find that
j"
(U
* * - - < " > - 4 ^ - 4 ^
"36)
which, interestingly, is the potential that would b e produced by q at the location of the
center of the sphere if the sphere were not present at all.
Example
Point charge and semiinfinite plane dielectric. The situation here is similar to that of
Figure 11-1 except that the shaded region for all negative x is filled with a l.i.h.
dielectric rather than with a conductor. As in the previous example, the potential does
not have a preassigned value. This time the b o u n d a r y conditions at the surface x = 0
are those satisfied by the components of E as given by (10-56) with oy = 0:
n- ( c 2 ^ 2
E
2y
ii)
= E
^1 y
o^2* tEXx 0
E = E
(11-37)
"^lz
We have taken region 1 as the dielectric ( x < 0) a n d region 2 as the vacuum ( x > 0)
corresponding to the direction of ft shown.
In the vacuum, we try the same set of charges as in Figure 11-1, but we will assume
at once that d' = d. Thus, as in (11-12), we get
[(* -
f + V + z2]1/2
[ ( x + d f ly*
+ ^]
1 / 2
(U 38)
"
which leads to field components

1
47re
( x - d)q
o \ [ ( x - d)2 + y2 +
( x + d)q'
Z
3 / 2
[(X
d)2 + y2 + z 2 ] 3 / 2
9'
1
< n
"40)
and where E2z is obtained by replacing the y in the front of (11-40) by z.

W e cannot use this same set of charges to calculate the potential in the dielectric
since q' is located there, and our fictitious charges should be located outside the region
181
4>y
Figure 11-6. Image charge used to calculate the

potential in the dielectric.
in question. Let us press our luck as m u c h as possible and try using a single point
charge q" located at the position of q itself as shown in Figure 11-6. This will produce
a potential in the dielectric given by
q"
(11_41)
^ = r ,
^ ; r r v ?
47re 0 [(x - d) + y + z ]
with the corresponding field components
El x =
(x d )q"
4tt 0 [ ( x d)2 + y2 + z 2 ]
3/2
yq'
3/2
4 776 o [ ( x d)2 + y2 + z2]
(11-42)
(11-43)
and, again, Elz is obtained by replacing y by z in the numerator of (11-43). N o w we

are ready to try to satisfy the boundary conditions on the field components.
T h e first equation of (11-37) is 0E2x(0, y, z) = eElx(0, y, z) and leads to
c 0 ( - 4 + <l')
(-<7")
(11-44)
while the second equation of (11-37) is Ely{Q, y, z ) = - E ^ O , y, z) and yields

q + q' = q"
as does E2z(0, y, z) = Elz(0,
e / e 0 = Ke, we find that
(11-45)
y, z). When we solve (11-44) and (11-45), and recall that

- 1
q' =
K
+ 1
q"
2q
+ 1
(11-46)
which means that we have managed to solve the problem completely.

Thus we are led to the two sets of image charges shown in Figure 11-7. The box
around the region number indicates that the charges shown are to be used to calculate
<j> in that region. The complete solution to this problem is thus obtained by inserting the
values of q' and q" given by (11-46) into (11-38) through (11-43).
We can find the force on the inducing charge q at x = d from (11-39) and (11-40)
by using only the contribution from q', and it is seen to be
qq'x
k - 1
q 2x^
F = qE(d,0,0)
=
(11-47)
JC + 1 / \6'7Trid2
4iT0(2d)
182
-d
Ke
\Ke
~+1/
])
Ke +
(a)
Figure 11-7.
dielectric.
2?
<b)
Charges for finding the potential (a) outside the dielectric and (b) inside the
Figure 11-8. Equipotentials and lines of E for a

point charge and a semiinfinite dielectric.
11-3 "REMEMBRANCE O F THINGS PAST"
183
and, as before, is the same as the Coulomb force of attraction between q and q', the
induced image charge in the dielectric. Physically, this is just the force of attraction
between q and the bound surface charges on the dielectric. We can find their density by
using (10-8), (10-50), (10-52), (11-42), and (11-46):
b(y>
) = p\n = (Ke ~ l)eOi*(0>
y, z) =
K
K
-l\
qd
2
e + 1 / lir^d .+
y2 + z 2 ) 3 / 2
(11-48)
and is negative as we expect. By comparing (11-17) and (11-18) with (11-48), we see
that the total bound charge induced on the dielectric surface will again equal the image
charge q'.
T h e equipotential curves and lines of E for this system are shown in Figure 11-8. As
expected, we see that the lines of E are refracted as they cross the bounding surface of
the dielectric.
"REMEMBRANCE OF THINGS PAST"

Sometimes a solution found for a previous problem can, with appropriate interpretation, be adapted and used to solve what at first appears to be a completely new
problem. We will illustrate this with two examples, both of which can be related to an
example from Chapter 5.
Example
Uniform infinite line charge and a semiinfinite plane grounded conductor. We consider an
infinitely long line charge of constant charge X per unit length that is a distance a f r o m
and parallel to the surface of a grounded conductor occupying half of all space. If we
take the line to be parallel to the z axis, the surface of the conductor to be the yz plane,
and choose the jc axis to pass through the line charge, we get the situation in the xy
plane shown in Figure 11-9. The boundary condition is that the potential be constant
and equal to zero on the conducting surface, that is, <f>(0, y, z) = 0. Because of the
similarities of Figures 11-9 and 11-1, as well as the identity of this b o u n d a r y condition
with (11-11), we can make a good guess that the appropriate image charge to use in this
case is another infinite line charge of density X located at x = a. But this is exactly
the charge distribution shown in Figure 5-7, for which the potential is given everywhere
by (5-34), and whose equipotentials and lines of E are illustrated in Figure 5-8. (We
Ay
Figure 11-9. Uniform infinite line charge perpendicular

to the page and parallel to a semiinfinite plane grounded
conductor.
184
recall that the equipotential surfaces are actually cylinders whose axes are parallel to
the z axis and, in fact, these axes lie in the xz plane.) As we noted in the discussion
following (5-38), the yz plane (x = 0) is the equipotential surface for
= 0, which is
exactly the boundary condition (11-11) that we have to satisfy for this problem.
Therefore, the solution is given by our previous results. However, we can use (5-34)
only in the unshaded vacuum region of Figure 11-9, that is, (77/2) < <p < t t / 2 . Thus,
the equipotentials and lines of force for this example are given by the curves in the right
half of Figure 5-8. In other words, we already have the complete solution available for
this example.
We can go even further with what we found for the complete system of Figure 5-7.
Suppose that one of the equipotential cylinders were replaced by a solid conductor
occupying the volume enclosed by the cylinder. T h e surface of the conductor would be
an equipotential as required and would have the potential corresponding to the surface
it replaced. The electric field will be normal to the conductor as required since the lines
of E are already normal to the equipotential cylinder. There will be a surface charge on
the cylinder given by (6-4), but, as we can easily see, the total charge per unit length on
the cylinder will still be A (assuming a surface in the right half of Figure 5-8). Consider
a Gaussian surface of integration just outside the conductor. The value of E will still be
found f r o m (5-35) and (5-36) so that the surface integral of Gauss' law (4-1) will be the
same as if the line charge were there. But since G a u s s ' law equates the surface integral
to the total charge inside divided by e 0 , regardless of its distribution, the total charge
per unit length will be A and this will be the total charge on the conductor's surface. In
other words, outside the conductor, nothing has changed, and we can still use (5-34)
everywhere else. [Inside, things have changed, of course: the electric field is now zero
and <p is constant, according to (6-1).] Similar remarks apply to the left-hand part of
Figure 5-8. We now are ready to consider our next example.
Example
Capacitance of two parallel cylindrical conductors. Consider two infinitely long conducting cylinders whose axes are parallel. F o r simplicity, we assume that they have the same
radius A\ their axes are separated by a distance D as shown in Figure 11-10. If we
identify these cylinders with the appropriate equipotentials of Figure 5-8, they will
carry charges of A and + A per unit length. O u r problem now is to relate these
dimensions to our previous results and to find the potential difference. In (5-38), we
found the radius of a circle of Figure 5-8 to be a/sinh rj and the location of its center
with respect to the origin is a coth 17 where 17 = 2 7re 0<f>/A. Therefore, if <p is the
potential of the right-hand cylinder of charge A, we have
a
(11-49)
sinh j]
D
= a coth 7] = A cosh 17
(11-50)
Figure 11-10. Cross-sectional view of t w o parallel

cylindrical conductors.
1 1 - 4 S E P A R A T I O N O F V A R I A B L E S IN R E C T A N G U L A R C O O R D I N A T E S
185
Solving the last equation for tj, we find the potential of the cylinder to be
x
"
> _
2^
J
C O S h
"
\
\ 2 l )
(11
"51)
Since the potential of the other cylinder is 4>, the potential difference between the two
will b e
X
J D \
cosh
-
\ 2A )
A<j> = 2<f> =
7T0
(11-52)
As the charge on a length L of this system will be XL, we see, f r o m (6-28), that the
capacitance for a length L will be
XL
7rea L
C
~K$
c o s h - 1 (Z>/2 J 4)
<n"53>
If A <sc D, as would be appropriate for two thin wires, we find f r o m the series
expansion c o s h - 1 u In2m ( 1 / 4 u 2 ) that we can approximate (11-53) as
7TqL
C
TTq L
2
ln(D/A)
(A/D)
In(D/A)
54
which is a commonly found expression. [This result also agrees with that of Exercise
6-15 when applied to this case.]
We now want to consider some of the more systematic methods for solving Laplace's
equation.
111-41 SEPARATION OF VARIABLES IN RECTANGULAR COORDINATES

W h e n (11-3) is expressed in rectangular coordinates with the use of (1-46), it becomes
d2d>
d2<f>
+
a?
d2d>
a?
"
(n
"53)
We will try to solve this by assuming a solution in the form of a product of quantities
each of which is a function of a single variable:
4>(x, y, z) = X(x)Y(y)Z(z)
(11-56)
U p o n substituting this into (11-55) and dividing by XYZ,

2
1 dX
1 dY
X dx
Y dy
we obtain
1 dZ
+ Y2
Z dz
= 0
(H-57)
which can also be written as

1 d2X
X dx2
1 d2Y
Y dy2
1 d2Z
Z dz2
(H-58)
The left-hand side of (11-58) is a function only of x and y, while the right-hand side is
a function only of z. But (11-58) says that both sides must always be equal for any and
all values of x, y, and zquantities
that can be varied independently. This is only
possible if both sides are equal to the same constant, which we write as y2. Thus, we
have
1 d2Z
(U 59)
186
Repeating this process, we find that

1 d2X
1 d2Y
x * * - " -
I T S ' *
Q2
2
(11
"60)
where a 2 and fi2 are also constants. These constants are not independent, for when we
substitute (11-59) and (11-60) into (11-57), we find that
a 2 + fi2 + y 2 = 0
2
(11-61)
We have seen that d X/dx
= a X and this can be integrated at once to give

JT(jc) = axeax + a2e'ax
(11-62)
where ax and a2 are constants of integration. Similarly,

r(>>) = bxe^
Z(z)
yz
= cxe
+ b2e~Py
(11-63)
yz
+ c2e~
(11-64)
The product of these three functions will be a solution of (11-55), provided that (11-61)
is satisfied. Because of this condition, a 2 , fi2, and y 2 cannot all be positive or all
negative; this means, in turn, that the constants a, fi, and y cannot all be real or all
imaginary so that if one is real, at least one of the others must be imaginary.
Consequently, at least one of the functions X, Y, and Z must vary exponentially with
its argument, while at least one will vary sinusoidally.
Evidently, there are many combinations of a, fi, y that will satisfy (11-61), and each
combination will give a solution, so that there are many possibilities. At the same time,
the constants of integration ax, a2, bx,...
may themselves depend on the particular
values of a, fi, y, so that we should write them in the form ^ ( a ) , a2(a), bx(fi),....
Since Laplace's equation is a linear equation, a sum of solutions of the form (11-56) will
also be a solution; if we add up all the possibilities, we see that we can write the most
general solution to Laplace's equation in rectangular coordinates in the form
4>(x, y,z)=L
M)e*x
yz
[cx(y)e
+ a2(a)e-x]
yz
+ c2( y)e~ ]
[bx(fi)e^
+ b2{fi)e~^}
(11-65)
where the sum is to be taken over all values of a, fi, y that satisfy (11-61). Since there
undoubtedly are an infinite number of combinations of a, fi, y that meet this requirement, we see that our general solution contains an infinite number of constants of
integration. The basic idea in the application of (11-65) is that these constants must be
determined so that 4> will satisfy the given boundary conditions. Once this has been
done, the problem is completely solved, and we know from the theorem of Section 11-1
that the solution obtained for 4> will be unique. In using this method, then, one would
start with (11-65), and satisfy the boundary conditions step by step in a systematic
manner; no guesswork should be required. As usual, all of this is best illustrated with a
specific example.
Example
We consider a region bounded by: (1) a semiinfinite conducting plane at i = 0
occupying that half of the yz plane corresponding to positive y (thus, 0 < y < oo, - oo
< z < oo); (2) a similar plane at x = L; and (3) the strip between them in the xz plane
(thus, 0 < x < L). The region defined in this way is shown in Figure 11-11 <3, and its
projection on the xy plane is shown shaded in Figure 11-116.
187
Figure 11-11. (a) Two semiinfinite conducting planes parallel to the yz plane. ( b ) Their
projection on the xy plane.
Let us assume the following boundary conditions:

at x = 0
<(0, y, z) = 0
(11-66)
at x = L
y, z) = 0
(11-67)
at y = oo
<f>(x, oo, z) = 0
at y = 0
(11-68)
4>(x,0, z) = / ( x )
(11-69)
where / ( x ) is some given function, that is, the potential on strip 3 varies with x only in
some predetermined manner. This set of b o u n d a r y conditions could correspond to the
semiinfinite sheets being grounded conductors, while strip 3 in the xz plane is
maintained in the manner described by / ( x ) by a suitable arrangement of batteries;
(11-68) corresponds to the usual requirement that the potential vanish at an infinite
distance from the source charges.
N o w since the region of interest extends over the complete range of z, and there is
no dependence on z in the boundary conditions, this situation is actually independent
of z and is really a two-dimensional problem so that <p = <p(x, y). Therefore, we see
from (11-65) that y = 0 is the only allowable value, and (11-65) reduces to
<P =
U M e
a2 + /?2 = 0
a x
+ A2(ot)e"x]
[b^P)e^
+ b2(fi)e~^]
(11-70)
where we have set AX(A) = a 1 (a)[c 1 (0) + c 2 (0)], and so on. [Actually, if y = 0, then
(11-59) will integrate to Z ( z ) = c[ + c'{z but, in order to have our solution independent of z, c'{ must be zero, which again will result in the form given by (11-70).] Since
a2 + fi2 = 0, we have a2 = ft2, so that a = ifi, where i = / T , and there is only
one independent quantity to be summed over and we can write
# = E [AAW*
+ A2(J8)e-ifix]
[b1(j3)e^
+ b2(fi)e'^]
(11-71)
188
This reduction of (11-65) to (11-71) followed solely from the fact that the problem is
really independent of zwhich is a " b o u n d a r y condition" in a generalized sense. Now
we turn to the boundary conditions that give actual values of <#>.
By comparing (11-68) and (11-71), we see that if fi is positive, the term varying like
ePy cannot appear because it does not vanish at infinity so that bx{fi) must be zero. On
the other hand, if fi is negative, b2(fi) must be zero. In both cases, the remaining term
will vary as e ~ ^ y , so that there is really only one possible form. For definiteness, we
choose fi > 0; thus we must have by(fi) = 0, and then (11-71) becomes
4> = E b2(fi)[Ax(fi)eiPx
/}>o
+ A2(fi)e-^]e-^
(11-72)
[/? cannot be zero, since then (11-68) will not be satisfied.] If we define b2(fi)A1(fi)
Ap, b2(fi)A2(fi)
= Bp, then (11-72) can be written more simply as
</>= E (Apeiflx
+ Bpe-Vê-to
(11-73)
/3>0
If we now use (11-66), we get

* ( 0 , y) = 0 = L
(Afi + Bfi)e-Py
(11-74)
/J>0
The only way in which this can be zero for arbitrary y, since e"f3y is always positive, is
for each term in the sum to be zero; thus Ap + Bp = 0, so that Bp = Ap. The term in
the parentheses in (11-73) can be then written as Ap(e'^x e'^x)
= 2iAp sin fix and
<j> is now
= 2iApSmfixe~py
(11-75)
/8> 0
(Since 4> must be a real quantity, we see that Ap itself must be an imaginary constant.)
Applying (11-67) to (11-75), we have
4>(L, y) = 0 =
fiLe~^y
2iAp sin
(11-76)
/3 > 0
which shows us that sin fiL = 0 and therefore fiL = HIT or

MT
fi = ~
(11-77)
where n is a positive integer (since fi is positive). Thus our sum over fi is actually a sum
over n and it is convenient to write it that way; if we set 2iAp = An, and use (11-77), we
can write (11-76) as
/ n TTX \
4>(x,y) = E ^ s i n ^ J e ~ ^ L
(11-78)
and all that remains to be found are the constant coefficients An. For this purpose, we
have one boundary condition remaining.
Putting y = 0 into (11-78), and setting the result equal to f{x), according to (11-69),
we get
4>(x, 0) = f ( x ) =
/ MTX \
(H-79)
which shows that our remaining problem is that of expanding f ( x ) in a Fourier series
(using only the sine terms). This is done by using the first of the following results that
are easily proved by direct integration
rL sinj mirx
_ \ sin(n7rx \ A fL cosTMITX\
_ cos / MTX\ i,: = 1 t8 (11_80)
/ ( r") ("r) "/
( H ('~r)'
2 ""
189
that is, these integrals are zero when m = n, and each equals \L for m = n, as we see
from (8-27). This result (11-80) is generally known as the orthogonality and normalization property of the trigonometric functions; when this integral is zero, the functions
are said to be "orthogonal." Combining this result with (11-79) enables us to "pick
o u t " any of the coefficients we desire as follows: we multiply both sides of (11-79) by
sin{mirx/L)
and integrate over the range L of x a n d we get
/ mmx \
//.
/ mirx \
/ mrx \
rr
r )
d x
s i n
( )
L8mn
s i n
( )
d x
= -LAm
since each term in the sum is zero except when n = m. Changing the index back to n
and solving for An, we obtain
A
"
2 rL
t nirx\
LJ0 ^ ( x ) s i n ( ~ z r J
d x
(11-81)
Once / ( x ) has been given, we can determine the coefficients An by carrying out the
integration indicated in (11-81), and then the An can be put into (11-78). The result will
be the correct unique value for the potential f r o m which <|>, and then E, can be found at
any point by evaluating the sum.
Example
Special case. In order to have a definite example, let us now assume that / ( x ) =
<p0 = const. This would correspond to keeping the whole strip in the xz plane at a
constant potential, as would be the case if it were a conductor connected to a single
battery. (It would have to be insulated f r o m the conducting walls by a thin strip of
dielectric on each edge.) Anyhow, putting this into (11-81), we get
A
n=
2 rt s i n I nrrx\ d x
T
$0
)
L, J0
\ L, j
2
% (1 cos nv)
YITT
(11-82)
N o w (1 cos MT) = 2 if n is an odd integer, and 0 if n is even, so that An = 0 if n is

even, a n d An = <f>0(4/mr) if n is odd. Thus, the general solution (11-78) as applied to
this special case becomes
4
1
I nmx \
* ( x , y) = * 0 - - s i n
(11-83)
77
odd "
\ L j
This sum cannot easily be further simplified, and in order to find 4> at a definite point
the sum will generally have to be evaluated numerically. Ordinarily, this is not a great
problem for not many terms in the sum will be needed to achieve a reasonable accuracy
because the successive terms get less and less important, both because of the factor 1 /n
and of the appearance of n in the exponential term.
W e can now use (11-83) to calculate the electric field at any point f r o m E =
and we get
4
/ n 7TX \
, = - * 0 7 E cos
(11-84)
L
4
L
while, of course, Ez=
nodd
E
nodd
sin
L,
1 mrx \
(11-85)
d<f>/dz = 0. W e note that these expressions have the correct
190
units of volts/meter. We also see that E = 0 at jc = 0 and L as it should because it is

a tangential component at these conducting surfaces. Ex does not necessarily vanish
there because it is a normal component and, in fact, by using (6-4), we could find the
surface charge density as a function of position on these conducting surfaces.
SEPARATION OF VARIABLES IN SPHERICAL COORDINATES

If we use (1-105), we can write (11-3) as
1 d I d<f> \
1
d (
1
d2=<p
r
+ :
sin 0
+
ttt
0
r dr\
dr )
r s i n ^ 90 \
36 J r sin # dtp
(11-86)
The general solution of this equation can be found by the same systematic application
of separation of variables that we have illustrated for rectangular coordinates. The
results are more complicated; consequently, for our purposes, it will be sufficient to
restrict ourselves only to those cases for which <> is independent of the angle <p, that is,
to systems with axial symmetry. There are still a large number of situations which fall
into this category, however. If <j> = <p(r, 0), (11-86) simplifies somewhat to
d (
Tr{r
d
= 0
(11-87)
By analogy to (11-56), we now look for a solution of the form

<f>(r,0) = R(r)T(0)
(11-88)
If we substitute this into (11-87), divide the result by the product RT, and equate the
resulting function of r alone to that for 0 alone, we get
1 d 1
dR \
r2
=
R dr\
dr J
1
T
d (
dT\
sin 0
= const. = K
d0 \
d0 J
sm0
(11-89)
since each term is a function of a different independent variable so that they must be
separately equal to the same constant K. If we set each in turn equal to AT, we get two
equations, one for R, and one for T. Doing this for the first term and carrying out the
differentiation, we find that R must satisfy the equation
d2R
,
dr
+ 2r
dR
KR = 0
dr
(11-90)
In order to make a start at solving this, we try a solution of the form R = ar' where a
and / are constants; upon substitution into (11-90), the result is that [/(/ + 1) A'JT?
= 0. Since we do not want R to be zero, as this would make <p = 0 everywhere, we see
that we must have
K=
1(1+1)
(H-91)
Equating this to the second term of (11-89), we find that the equation for T becomes
1
d j
dT,\
sin 0 rsin 0 d0 \
d0
+ / ( / + ! ) 7) = 0
(11-92)
where we have added the subscript / to indicate the association of the solution with the
constant. N o w since Tl is part of the physical quantity <p, it must be a reasonable
function. By this we mean that it should be finite, single valued, and continuous over
(he whole range of 0. It would take us too far afield to investigate this, but it can be
shown that this is possible only if / is a positive integer, including zero; thus
7 = 0,1,2,3,...
(11-93)
1 1 - 5 S E P A R A T I O N O F V A R I A B L E S IN S P H E R I C A L C O O R D I N A T E S
191
As a matter of fact, the T, can be identified with the Legendre polynomials we have
already encountered in Chapter 8. If we refer back to Figure 8-1, we see that 0, and ri
are just the corresponding spherical coordinates of the location of qt with respect to the
fixed direction f shown; Rt is the corresponding distance from the source point to the
field point and r of the figure will be a definite constant. Accordingly, in (8-12), we can
write 1 /rl+1
as a constant G,, and then drop the index i on 0, and r, to get an
expression in the same notation we are using here, that is, we can write
= E G,r'P,(cos0)
K
i
1=0
(11-94)
We already know from (1-144) that this is a solution of Laplace's equation and
therefore cf> = 1 / / ? , must be a solution of (11-87). If we now substitute (11-94) into it,
we obtain
E g /
1= 0
/(/+
v
1
d I
dP,
s i n 0 ^
sin 0 dO \
dO
1 )P,+
' '
= 0
(11-95)
In general, this sum can be zero only if each term is zero since r is arbitrary, so that the
bracketed term must be zero for each /. Upon comparing this with (11-92), we see that
7} and Pt satisfy the same differential equation and thus Tt can be taken as, at most, a
constant times P/. We will absorb any such constant into the factor R(r) of (11-88) and
simply let T,(0) = P,(cos0). Then we can write (11-88) as R/(r)P,(cos 0), and since
there will be a solution of this form of the linear differential equation (11-87) for each
possible /, we can write the general solution of (11-87) in the form
00
Yl Ri(r)pi(cosO)
/=o
4>{r,6)=
(11-96)
We still have to find the general form of R h however.

The equation satisfied by Rh as found from (11-90) and (11-91), now is
d2R,
r -tdr
dR,
+ 2r /( / + 1)/?/ = 0
dr
(11-97)
where / satisfies (11-93). We now try to solve this equation more generally than before
by assuming a form R, = atr" where a, is a constant and n is an integer, and, upon
substitution of this into (11-97), we find that we must have n(n + 1) /(/ + 1) = 0.
This equation has the two solutions n = I and (/ -I- 1), so that the general solution of
(11-97) has the form
= A,rl
R,(r)
+ -7/1
(11-98)
where A, and Bt are constants of integration. Substituting this into (11-96), we finally
get the general form of the solution to Laplace's equation for an axially symmetric
situation as
00
4>(r,6) = E [A/
/=o\
B, \
+
r
P/(cs0)
(11-99)
As we saw in (8-10), the first few Legendre polynomials are

Po(cos0) = 1
P1(cos0) = cos#
P 2 ( c o s 0 ) = | ( 3 c o s 2 6 1) (11-100)
and higher-order ones can be found from the recursion relation (8-11):
( / + 1)P/ + j(cos 0) = (21+ l ) cos 0P/(cos 0) - IP^^cosd)
(11-101)
192
In the last section, we found that the orthogonal properties of trigonometric

functions as expressed in (11-80) were a great help in finding the expansion coefficients
in rectangular coordinates. The Legendre polynomials have a similar property and it is
equally helpful. It can be shown that
fJ-P/(cos 0)Pm(cos
6) sin 6 d6 = f _ P / ( f i ) P
(ix)dfi=^rj8
l m
(11-102)
where we have used (2-22). As an application of this property, we can derive a result
that will be quite useful to us in what follows. Consider the sum
00
Z C;?,(cos0) = O
(11-103)
1= 0
where the C, are constants. Since the sum must be zero for any arbitrary value of the
angle 6, it seems plausible that this can be the case only if each term in the sum is itself
zero, that is, if all of the Ct are zero. We can easily show that this is the case. In
(11-103), we let cos 6 = ju, multiply through by Pm(ix) d/x, integrate over ju. from - 1 to
+ 1 , and use (11-102); in this way we get
"
since every term in the sum is zero except when I = m. Therefore, Cm = 0 for all m, as
we suspected. Changing the index from m back to /, we have shown that
00
if C,P,(cos 0) = O,
/=o
then C, = 0
(11-104)
Now let us consider some specific examples.

Example
Grounded conducting sphere in a previously uniform field. We assume that initially we
have a completely uniform field E 0 and then a conducting sphere of radius a is inserted
into the region and kept at zero potential. The field E 0 could be produced, for example,
by a parallel plate capacitor whose plate separation is very large compared to the size of
the sphere. We choose the 2 axis in the direction of E 0 , so that E 0 = E0z, and take the
origin at the center of the sphere as shown in Figure 11-12. Even before we have found
the solution for this case, we can anticipate that the lines of E, which will finally result,
will have the general form shown by the lines in the figure since they must come in
normal to the conducting surface. This situation clearly has axial symmetry, making a. In order to do this, we need to determine the boundary conditions.
At the surface of the conducting sphere, <j> is zero, so that one of our conditions is
that
<p(a,0) = O
(11-105)
At the other extreme, when we get far away from the sphere, we expect the original
uniform field to be unaffected by the presence of the sphere, that is, as r - 00,
E(r,6) -* E0z. Thus, in this limit, /dz which gives <t> = - E0z as the limiting expression. Thus, our
condition at the far boundary is
4>(r,0)
-*
E0z = E^rcosd
(11-106)
Figure 11-12.
193
Lines of E for a grounded conducting sphere in a previously uniform field.
Putting r = a in (11-99) and using (11-105) gives us

E
t A
1=0
A a
i'
~77T
CI
(11-107)
-P/(cos0) = 0
N o w (11-104) shows that the term in parentheses is zero, so that B, = a2l+ 'A,,
and (11-99) thus becomes
oo
<#> =
a2,+ 1\
7TT
/(cos0)
(11-108)
1= 0
Applying (11-106) to this, we get

YtA,rlP,{co&0)
/=o
= Eorcos0
= E0rP x(cos 6)
(11-109)
with the use of (11-100). If we separate out the term in the sum for 1=1,
convenience in writing, (11-109) takes the form
(A1 + E0)rPx(cos
9) + E Ajr'P,(cos#)
iî
= 0
for
(11-110)
Since this must hold for all large r, it will be true in particular for a definite constant
value r0. But then (11-110) is exactly of the form (11-104), which shows us that
A1 = E0, and At = 0 (/ ^ 1) thus completing the evaluation of the coefficients.
Consequently, the sum in (11-108) reduces to the single term for / = 1, and our final
unique expression for (/-, 0) = -E0rcos6
a3Encos 6
-I^
(11-111)
Having obtained this result, we have completely solved the problem, and <| and E can
be found at any desired field point.
194
As we see from (11-106), the first term in (11-111) is just that corresponding to the
original uniform field. The second term must represent that part of the potential that
arises because of the presence of the sphere. By comparing it with (8-48), we see that it
is a dipole term and the dipole moment of the sphere is
p = 4m0a3E0
(11-112)
Thus, the conducting sphere has acquired a dipole moment proportional to the original
field and has been, in effect, polarized. The ratio of the induced dipole moment to the
applied field is called the polarizability a and we see that
a
= L. = 4 0 a 3 = 3e0Vs
E
o
(11-113)
where V is the volume of the sphere. This dipole moment must have its origin in the
surface density of free charge on the sphere, which we can easily find. The radial
component of E is
Er=
d<}> I
~ Tr = 1
2a3 \
- j r Eoôs0
(11-114)
At the surface of the sphere, this is the normal component, and we can get the surface
charge density from (6-4) as
af(0) = e0Er(a,0)
= ( 3 e 0 0 ) cos 6
(11-115)
which is seen to be proportional to cos 0. This result has the same form as the charge
density given by (10-27) and illustrated in Figure 10-9. The surface charge has opposite
signs on the two hemispheres, and it is this overall charge separation that gives rise to
the dipole moment. [It is easy to verify directly that when p is calculated by using
(11-115) in the surface integral form of (8-22), one gets (11-112); this same result was
also obtained in Exercise 8-8.] The sphere was originally neutral before insertion, and
we can see if it still has zero net charge by combining (11-115) and (2-16); we get
Qf total = Ja/da=
f2" f
( 3 e 0 0 c o s 0){a2
sin0d6d(p)
= 0
and thus the sphere is still neutral.

The remaining component of E is
E
>=
1 d<f>
=
a3 \
<n-116)
We see that Ee(a, 0) = 0 as it must since it is a tangential component at the surface of

the sphere.
Since <j> as given by (11-111) has an angular dependence arising solely from the
/ \ ( c o s 0) term of (11-99), we can conclude that the sphere has a dipole moment only.
In other words, the charge distribution (11-115) is such that it produces no monopole
moment, as well as zero quadrupole and higher moments; this was also verified directly
for the quadrupole moment in Exercise 8-8.
The lines of E as found from (11-114) and (11-116) are illustrated in Figure 11-12.
Example
Dielectric sphere in a previously uniform field. This example is the same as the last except
that we have an uncharged dielectric sphere rather than a grounded conducting one.
The boundary conditions are somewhat different because the sphere is no longer at a
1 1 - 5 . S E P A R A T I O N O F V A R I A B L E S IN S P H E R I C A L C O O R D I N A T E S
195
definite potential nor is <j> necessarily constant within it. It will be convenient to look at
the regions outside the sphere (r > a) and inside the sphere (r < a) separately; we will
label the potentials and fields applicable for these regions with the subscripts o and /,
respectively.
At large distances, the field again has to become uniform, so one boundary condition
is exactly the same as (11-106):
<t>o(r,0) - -E0r
r> oo
cos 8
(11-117)
At the surface of the sphere, (11-105) no longer applies. However, the sphere's surface
is a surface of discontinuity between the dielectric and the vacuum, and we know that
the tangential components of E and the normal components of D are continuous there.
These are written in terms of E only in (10-56). In this case, we have oy = 0, and since
Er and Ee are the normal and tangential components, respectively, we can write (10-56)
in terms of the derivatives of 4> as
0
dr )r=a
l_f<M
r dO J r = a
- 1 - & )
\
dr Jr=a
=
r 36 J r=a
(11-118)
(11-119)
where region 2 is taken to be the outside, while 1 is the inside. Finally, we have one
more condition which we haven't met before. It is a "boundary condition" of a
generalized kind, and deals with conditions at the origin (r = 0). By assumption, there
are no free point charges within the sphere, so that there will be none at the origin.
Since this is the only way in which the potential could become infinite there, we must
also require that
4> is finite for r = 0
(11-120)
Rather than going about solving this problem in the same way as the last example,
we will try to shorten our work by taking advantage of our experience. We expect the
dielectric sphere to become polarized by the field and acquire a dipole moment, so it is
plausible that once again only the 1=1 term of (11-99) will survive. Accordingly, let us
assume the potential outside to be given by
<j>0 = [ ~ A 0 r +
cos#
(11-121)
where Aa and B0 are constants. Since the boundary condition (11-118) will thus have a
cos 6 on the left-hand side, it seems reasonable to expect the same on the right-hand
side; accordingly, for the potential inside the sphere, let us choose the same general
form as (11-121). Thus we also write
(j>i = ^ A j r + j | cos 0
(11-122)
where At and Bi are constants. If we can manage to find these four constants by
satisfying all of the boundary conditions, we know that we will have found the unique
solution to the problem.
Combining (11-117) and (11-121), we see that Aa = E0 and we now have
4 > o = ^ - E 0 r + ~ jcos#
(11-123)
We also see that if Bi ^ 0 in (11-122), the 1/r2 term will make <t> > oo as r - 0 which
is not allowable by (11-120); thus this term cannot appear for the inside form of the
196
potential, so we must really have B{ = 0 making (11-122) reduce to

4>t = -Afcosd
(11-124)
Substitution of these expressions into (11-118) and (11-119) results in the following two
equations:
2B a
E0 +
= k eAt
a
-E0a
B0
a
= -A,a
since e/0 = xe. Solving these, we find that At = 3E0/(K

(ks + 2)]a3E0, so that our potentials are
+ 2) and B0 [(Ke 1 ) /
/ K 1 \ aôcos 6
( ^ i ^ |
^
* = -E0rcos+
"(rT2)
r C
S #
(11-125)
(11-126)
and the problem is completely solved.

The field inside the sphere is E, = V<J>, = Etz, a n d since E 0 = E0z, we have
E , . - ( ^ ) e
(11-127)
and we see that |E,| < | E 0 | since KE > 1. Thus the electric field is constant, parallel to
the original external field, but smaller than it. Consequently, the sphere will be
uniformly polarized, and the polarization as obtained f r o m (10-50), (10-52), and
(11-127) is
p=
(7T4)3cEo
(11 128)
"
and the magnitude of the total dipole moment as obtained by multiplying P by the
volume of the sphere is
I K.-W
P = 477 j
+ 2
a\0E0
(11-129)
This enables us to write (11-125) as

<
f> = E00r cos 6 +
Vo
p cos 6
j
4tt 0 /- 2
(11-130)
where the second term agrees exactly with (8-48) as we would expect.
We can rewrite (11-127) in an interesting and instructive way by subtracting E 0 f r o m
both sides. We find that E ( E 0 = [(1 x e ) / ( K e + 2)]E 0 = P / 3 e 0 by using (11-128),
and therefore
E, = E 0 -
(11-131)
which shows that the resultant field in the interior can be written as E, = E 0 + E ] o c ,
that is, as the sum of the original external vacuum field and a local field E l o c that is
proportional to the polarization and opposite to it. [We previously found the similar
result (10-67) for the case in which the dielectric had the form of a slab with parallel
faces.] The source of this internal field is, of course, the b o u n d charges appearing on the
surface of the dielectric where there is a discontinuity in P, and, f r o m (10-8), their
197
density is given b y a f e = P - n' = P - f = P cos 6\ this is a distribution just like (10-27)

and Figure 10-9.
In writing (11-130) as we did, we were able to write the potential outside explicitly as
the sum of that due to the external field and the contribution of the polarized sphere.
We can get a similar result for <f>, by combining (11-126), (11-127), and (11-131) to give
(11-132)
There is an interesting special case of our results. If we let Ke * oo, we find that
E, -* 0, p -> 47T0a3E0, <f- 0, and <p0 -* (11-111); these are just the results corresponding to the conducting sphere of the last example as we see from (6-1), (11-105),
and (11-112). Thus, we see that, as far as electrostatic effects are concerned, a
conductor acts like a material of infinite dielectric constant.
The lines of E found for this system are shown in Figure 11-13.
Example
Sphere with uniform permanent polarization. Suppose we have a sphere with uniform
polarization P = Pz where P = const, but with n o external field present. This is exactly
the system we considered in Section 10-4 and for which we found <j> and E along the z
axis only. The form in which we have written (11-130) and (11-132) enables us to write
down the complete solution at once, for we can set E0 = 0, interpret p and P as the
permanent dipole moment and polarization, respectively, and we get
(11-133)
The first expression shows us that the field everywhere outside the sphere is a dipole
field as we suspected from our results for the z axis. Similarly, the second expression
shows us that the electric field is uniform everywhere inside the sphere and is, in fact,
=
Figure 11-13.
"
Lines of E for a dielectric sphere in a previously uniform field.
198
given by
E, = - t-*o
'
in agreement with (10-37) that we previously found to hold on the z axis.
(11-134)
Since so many local fields turn out to be proportional to the polarization and
opposite to it, it is common practice to write E l o c in the general form
(n-135)
Eloc =
where N is a dimensionless constant called the depolarizing factor. We have already

found two of them. They are N = 1 for an infinite slab with parallel faces and N = \
for a sphere, as found in (10-67) and (11-134). The result of Exercise 11-26 will show
that N = ^ for a cylinder, as could also have been surmised from Exercise 10-12.
rrrgj SPHERICALLY SYMMETRIC SOLUTION OF POISSON'S EQUATION
If we combine (11-1) with (11-55) or (11-86), we will have Poisson's equation expressed
in rectangular or spherical coordinates. For illustrative purposes in this section,
however, we will restrict ourselves to completely spherically symmetric situations so
that
is a function of r only, that is,
= 4>(r). Then p will necessarily be a function
only of r also, and we get
1 d fr 22d<t> \
2 j i
~ I=
dr\
dr j
pO)
(11-136)
Example
Sphere with uniform charge density. Let us consider a sphere of radius a containing
charge of constant density so that p = const, inside, while p = 0 outside. We want to
find 4> everywhere by using (11-136). Outside the sphere, this equation becomes
d(r2d<pa/dr)/dr
= 0, which can be integrated twice to give
* A r ) - A . + -f
(11-137)
where Aa and B0 are constants of integration. [This is also what we get from (11-99) by
dropping all the angle-dependent terms, that is, using only / = 0.] Inside the sphere,
(11-136) becomes
- L ^ ) dr\
dr J
- p
e0
which can be easily integrated twice to give

*<(r)=
Pr~
- 7 +Ai+
6 0
~
r
(11-138)
where A i and Bt are constants. Now all that remains to be done is to evaluate the
constants of integration from the boundary conditions.
Since all of the charges are contained within a finite volume, we want < to vanish at
infinity, according to our discussion in connection with (5-10). Thus, as r > 00,
EXERCISES
199
<j)a > 0, and we see from (11-137) that Aa = 0 and therefore <f>0 = Ba/r. Since there are
no point charges at the origin, (11-120) still is applicable and shows us, that, in
(11-138), Bt = 0. We know from (9-29) that <j> is continuous at r = a, so that
4>0(a) = 4>i(a) which gives A{ = (BJa) + (pa2/6e0),
making ^>, now have the form
6e 0
(a2 - r2) +
a
(11-139)
Finally, since there is no surface charge, En = Er is continuous at r = a, according to

(9-26), so that -(d<t> 0 /dr) r = a = ~(d4>i/dr)r=a,
which leads to (BJa1)
= (pa/3e0)
3
or Ba = pa /3e0. Putting this into (11-137) (with Aa = 0) and (11-139), we obtain
< >
t o(r) ~ T
j0r
4>,(r) = 7~(3fl2
oe 0
- r2)
which is exactly what we obtained in (5-22) and (5-23) when we discussed this same
example by other methods.
EXERCISES
11-1 Reconsider the theorem of Section 11-1,

and show that if the normal component of E is
preassigned on all points of the bounding surface,
then 4>, and <2 need not be equal but can differ
at most by a constant.
11-2 Verify by direct integration that (11-19) is
the resultant of the Coulomb force between q and
the induced charge described by (11-17).
11-3 A point charge q is located in the xy plane
near two grounded conducting planes intersecting
at right angles as shown in Figure 11-14. The z
axis lies along the line of intersection of the
planes. Find and justify the image charges that,
together with q, will give the potential at all
points in the vacuum region a > 0, y > 0,
-cc<z<oo.
Find <j>(x,y,z) in the vacuum
region. Find Ey(x, y, z). Verify that Ev vanishes
on the conducting plane for which it is a tangential component. Find the surface charge density af
induced on the plane for which Ey is an appropriate component to use. What is the sign of
a/1 (Hint: recall multiple image formation in plane
mirrors from geometrical optics.)
11-4 Suppose that the angle between the conducting planes of Figure 11-14 is 60 rather than
the 90 shown, and that q is along the line
bisecting the angle (i.e., set a = b and then change
the angle). Find and justify the image charges
that, together with q, will give <f> at all points in
the vacuum region. What is the direction of the
resultant force on ql
11-5 Suppose that q in Figure 11-1 is replaced
by a point dipole p = py. Find the force on p.
Ay
Figure 11-14.
Geometry for Exercise 11-3.
11-6 Suppose that the point where the x axis

intersects the middle equipotential curve of Figure 11-2 is midway between the charge and the
conductor. Find <j> and E at this point.
11-7 Verify that (11-35) can also be obtained by
using (11-29) and (11-30).
11-8 Using the spherical coordinate system of
Figure 11-3, find <j> at all points outside the
sphere for the charge distribution of Figure 11-5.
Show that your result gives (11-36). Find E and
the force on q. Find ay and show that the total
charge induced on the sphere is zero.
11-9 The conducting sphere of Figure 11-3 is
now insulated and has a total charge Q on it.
Find the potential of the sphere and the force
on q.
11-10 After (11-48), we saw that the total bound
surface charge on the dielectric was equal to the
image charge. How do you reconcile this result
with (10-13)?
200
11-11 For the system of Figure 11-9, find the

surface density of induced charge on the conductor, the total induced charge per unit length parallel to the line charge, and the force per unit length
on the line charge.
11-12 Show that the attractive force on a length
L of one of the cylinders of Figure 11-10 is given
by
F - -
irt0L( A<)
l[cosh~
(D/2A)]2(D2
- 4A2)1/2
11-13 A long wire of circular cross section of

radius A is strung on poles at a height h above
the ground. Neglect the sag of the wire and the
curvature of the earth, and find the capacitance
per unit length of this system. Also find the force
attracting a unit length of the uniformly charged
wire toward the earth.
11-14 Show that the potential produced by an
infinite line charge of constant density X at a
distance D from the parallel axis of a circular
conducting cylinder of radius A is that which
would be produced by the actual line charge and
an image line charge X that is a distance A2/D
toward the real charge from the center of the
cylinder. What is the potential of the cylinder?
11-15 For the system of Figure 11-11, find the
surface charge density on the face x = 0.
11-16 Verify that Ex of (11-84) is zero for y = 0.
[Hint: it may help to recall that cos u = ^(V"
these functions, or appropriate derivatives of

them, are related, if in fact they are. Find the
general form of X(x) and interpret the corresponding electric field.
11-20 (a) Calculate directly /L x P2(fi) dp for
I = 0,1,2, and thus show that (11-102) is correct
for these cases.
(b) Show that when the differential equation (1192) satisfied by P, is expressed in terms of the
variable /x = cos 9, it becomes
d
dp.
+ /(/+
1)P, = 0 (11-140)
(c) The orthogonality property expressed by (11102) for 11= m can be shown to be a consequence
of (11-140) and the fact that P, is finite and has
finite derivatives at p = 1 as follows: multiply
(11-140) by Pm\ write down the differential equation satisfied by Pm(p) and multiply it by P,\
subtract these two expressions and integrate the
result over p from 1 to +1, integrating by parts
as necessary.
11-21 Find the equation for the lines of E corresponding to (11-111), that is, for the conducting
sphere in a previously uniform field.
11-22 Suppose that instead of (11-106) we had
required that at large distances $ -* E0r cos 6
+ <f>0 where <#>0 is constant, since this will still
yield a uniform field. Find <f> under these conditions. If your solution is different from (11-111),
how can you interpret the result?
11-23 A spherical cavity of radius a is within a
11-17 This is a two-dimensional problem. Conlarge
grounded conductor. A charge q is placed
sider a square in the xy plane with corners at
within
the cavity at a distance b from the center.
(0,0), (a,0), (a, a), and (0, a). There is no charge
Find <f> at all points within the cavity by using
nor matter inside the square. The sides perpendicspherical coordinates with origin at the center and
ular to the y axis have a potential of zero. The
side at x = a has the constant potential <#>0, while z axis passing though the location of q. Find E at
that at x = 0 has the constant potential <j>0- all points within the cavity. Find E at the center
of the cavity. Find the surface charge density
Find <#> (x, y) for all points inside the square.
induced on the wall of the cavity. What is the
Find E at the center of the square and evaluate
total induced charge on the wall?
the ratio of E to (<j>0/a) at this point to four
significant figures.
11-24 Solve the two-dimensional form of
Laplace's
equation expressed in plane polar coor11-18 Find the potential <#> at all points within a
dinates
(p,
<p) by separation of variables. Thus,
cube of side L with the location and orientation
show
that
the
general solution has the form
of Figure 1-41. There is no charge nor matter
within the cube. The potential on the face z = L
tf> = A + B In p
has the constant value </>0, and the potential on all
B
other faces is zero. Show that, to four significant
m
+
H
MmP
+
figures, <#> at the center of the cube is 0.1667 <f>0.
P
m=l\
11-19 Show that the solution of Laplace's equaX (C m cos m<p + Dm sin mtp)
(11-141)
tion can be written as a sum of terms each of the
form X(x) + Y( y) + Z(z). Be sure to show how
where m is a positive integer and must be single 11-29 A circular ring of radius a lies in the xy
plane with origin at the center. It has a constant
valued.)
11-25 An infinitely lcJng grounded cylindrical linear charge density X on its circumference. Find
conductor with circular cross section of radius a <f>(r, 0), expressed as a series in the P,(cos 6), for
has its axis coinciding with the z axis. It is placed all r. (See previous exercise.)
in a previously uniform electric field E 0 = 0 x, 11-30 A system of two concentric spheres has
so that its axis is perpendicular to E 0 . Find <#> for inner radius a and outer radius b. The region
all points outside the cylinder. Find the surface between them is filled with a spherically symmetcharge density and show that the cylinder remains ric charge distribution of volume density p =
p0(r/a)n where p0 = const, and n > 0. The inner
neutral.
11-26 The cylinder of the previous exercise is sphere is kept at a constant potential <f>1, while
now a dielectric rather than a conductor. Find <#> the outer is at a constant potential <#>2 Find <> for
everywhere. Find E and P inside and thus verify a < r < b by using (11-136).
11-31 Two infinite conducting planes are paralthat the depolarizing factor is \ in this case.
lel
to the xy plane. One of them is located at
11-27 A circle of radius a lies in the xy plane
z
=
0 and is kept at a constant potential <j>0. The
with its center at the origin. The semicircular part
of the boundary for jc > 0 is kept at the constant other, at constant potential <j>d, has z = d. The
with charge with
potential <#>0; the other semicircle for x < 0 is region between them is filled
2
volume
density
p
=
p
(z/d)
.
Solve Poisson's
0
kept at the constant potential <j>0. Find <#> for all
equation
to
find
<p
for
0
<
z
<
d.
Find the surface
points within the circle. Find E at the center of
charge
density
on
each
plate.
the circle.
11-28 Although (11-99) gives the form of the 11-32 Consider the coaxial cylinders of Figure
general solution for a situation with axial symme- 4-7. The inner cylinder is kept at constant potentry, it is not always easy to find the coefficients A, tial <f>a and the outer at <j>h. There is a cylindrical
and B/ by setting up the problem for general sheath of constant charge density in the region
values of 0. Sometimes the following approach between them and vacuum elsewhere. In other
can be used. Since (11-99) is true for all values of words, the volume charge density pch is: zero for
0, it must be true, with the same coefficients, for a < p < Pj, Acq for pj for a < p < b.
axis. Then r = z, and Pt(cos 9) = P/(l) = 1, and 11-33 Here are some miscellaneous results for a
system of source charges occupying a finite re(11-99) reduces to
00
gion.
You can easily prove them with the use of
/
B \
* ( * ) = M / * ' + -z / T r )
(z>0)
(1-111), (1-122), (1-123), (1-141), and (3-3):
i~o\
I
(11-142)
Thus, if one can solve for <f> on the z axis, it
should be possible to find At and B, by comparison with (11-142), and when these coefficients are
put back into (11-99), the result is an expression 11-34 Occasionally, one will find in other books
for <#> which is correct for all 6. Sometimes the expressions for the image charges used for the
identification must be made by expanding <j>(z) example of the point charge and semiinfinite diin a power series, and the coefficients found by a electric that are different from the ones we obterm by term comparison. As an example of this tained. As shown by the e0 in the denominator of
procedure, consider the uniform line charge of (11-41), we used vacuum properties in the region
finite length of Figure 3-8, for which the potential occupied by the dielectric. Another point of view
is given by (5-30). Assume that L2 =
= L,
is to assume in effect that the dielectric fills all
and show that when r > L, 4> can be written in space as far as calculations involving q" are
concerned. Thus, one can still use (11-41) but
the form
+l
with
an c in the denominator instead. Show that,
X ^ (L\' P,(cos0)
if
this
is done, q" is found to be ice times that
('.) - T
"77TTT
given in (11-46). Also show that the potential and
where the sum is taken over only even values of / fields in the dielectric will nevertheless be exactly
including I 0.
the same as before, as must be the case.
12
ELECTRIC CURRENTS
Electrostatics deals with relations among electric charges that are at rest. The next
principal division of electromagnetism that we will considermagnetostaticsinvolves
forces between moving charges. A flow of electric charge is called an electric current,
and, in this chapter, we want to devise useful ways of describing these currents in
general. We will also discuss some aspects of a particular class of currents, those in
conductors.
12-1
CURRENT A N D CURRENT DENSITIES
Suppose that we have someone stationed at a point P observing charges passing that
point; these charges may be traveling along a metal wire or may be simply a beam of
charged particles moving through space. In any event, we assume that in a time interval
At, our observer finds that a charge Aq has passed P. Then we can define the average
current ( / ) during this interval as the average rate of flow of charge
Aq
</> - -
(12-1)
Later, we will be more specific about how Aq or ( I ) could be measured but if, for
example, the current were due to a flow of protons, each of charge e, then the
measurement process could be imagined as being simply one of counting. Thus, if N
protons passed by in this interval, Aq = Ne and ( I ) = Ne/ At. As is implied by (12-1),
the "direction" or sense of the current is defined as that of the flow of positive charge.
If the moving charges had a negative charge, for example if they were electrons, the
direction of ( I ) would be opposite to their direction of motion. The reason for this is
quite simple. Suppose the region around a point originally were neutral, that is, had
equal amounts of positive and negative charge. Then if negative charges were leaving
this region, it would acquire an excess of positive charge which is the same net effect as
if positive charges were coming into the region.
If the flow of charge is not uniform in time, we can define an instantaneous current I
as the instantaneous rate of flow of charge:
/ - ^
at
(12-2)
For some time, we will be concerned with currents that are constant in time so that
I = const, and ( I ) = I. These are called steady currents or stationary currents and
describe a uniform rate of flow of charge.
As we noted in Section 2-2, the unit of charge is actually defined in terms of the unit
of current which is called an ampere, so that, according to (12-1), 1 coulomb = 1
ampere-second. The ampere itself is defined in terms of the force between currents and
we will give it a precise definition in the next chapter.
It is often convenient to think of the current as traveling along a geometric curve, as
illustrated in Figure 12-1 where the arrow head indicates the direction of / and ds is a
displacement along the line in the sense of I. This idealized situation could usefully be
202
12-1
Figure 12-1.
C U R R E N T A N D C U R R E N T DENSITIES
203
A filamentary current.
taken to describe the flow of charge in a very thin wire or in a beam of small
cross-sectional area. Such currents are called filamentary currents. However, we will
also need to consider situations in which the flow of charge is distributed throughout a
volume or on a surface, and we want to have suitable descriptions of them. We can do
this by introducing the current densities.
The first of these is the volume current density J . Its direction is that of the direction
of flow of charge and its magnitude J is given by the current per unit area through an
area set perpendicular to the flow, or it is charge per unit time per unit area. We can
simultaneously illustrate this definition and obtain a useful relation by considering the
situation of Figure 12-2. Let us find the charge Aq, which, in a time A t, has passed
through the small area A a on the left, which is perpendicular to J. By (12-1),
Aq = (I)At = (J)AaAt
since ( J ) is average current per unit area. But all of the
charge that has passed through A a is contained within the volume AT of the cylinder of
length A/, so that, if we use (2-14), we also have Aq = p AT = p A/ A a where p is the
volume charge density. Equating these two expressions for Aq, and canceling the
common factor A a, we find that ( J ) = p(Al/At)
= p(v) where (v) is the average
speed of the charges. Now this relation clearly holds instantaneously, as well as on the
average, and since the direction of J is defined as that of the direction of flow, which is
v, we can write
J = pv
(12-3)
If the moving charges are of different types with densities p, and corresponding
velocities v(, then we see that, in At, the charge of type i that has passed through will be
Aqi = p,|v ( | A a At. Then the total of all kinds will be A q = -p-|v-| At Aa, and gives us
the natural generalization of (12-3) in the form
J = Ep,V,
(12-4)
By comparing these last two results, we see that we can still use (12-3) in the general
case by taking p as the total volume charge density and v as an average velocity,
weighted by the densities p,, much as in the calculation of the velocity of the center of
mass of a collection of mass points.
N o w suppose that J and an element of area d a are not parallel as is shown in Figure
12-3. We can find the charge that has passed through da in a time dt in a manner
similar to that we used for Figure 12-2. This time the total charge would be that
A/
Figure 12-2. Calculation of

v o l u m e current density.
204
ELECTRIC C U R R E N T S
dl.
1
1
J
da
Figure 12-3. The current

density and element of area are
not parallel.
Figure 12-4. Calculation of total

rate of f l o w of charge through t h e
area S.
contained in the cylinder of slant height dl and volume dl cos ^ da and would be given
by dq = p dl cos ^ da = pv cos ^ da dt = pv da dt = J da dt with the use of (12-3).
Therefore, the rate of flow of charge through d a will be
)
dt
= J da
(12-5)
) through da
If we consider an arbitrary surface S as shown in Figure 12-4, we can find the total rate
at which charge is flowing through it by adding up the contributions of all the elements
da as given by (12-5). Thus we get
( - )
\
dt
J through S
J/ J
da
(12-6)
which is sometimes called the flux of charge. In (12-6), S can be either an open surface
or a closed surface.
If, for some reason, the moving charges can be thought of as being constrained to
flow on a surface, we can define a surface current density K. Its direction is that of the
direction of flow of charge and its magnitude K is defined as equal to the current per
unit length through a line lying in the surface and oriented perpendicular to the flow.
This definition is indicated schematically in Figure 12-5<a. In Figure 12-56, we illustrate
a situation where K is not at right angles to the line ds; the unit vector t is drawn at
right angles to ds and lying on the surface, as does K, so that it is a tangential vector.
By using methods analogous to those by which we obtained (12-3) and (12-5), we can
show that
K =
|K t | ds
( - )
\ dt
(12-7)
(TV
(12-8)
) through ds
where a is the surface charge density.

Similarly, for a filamentary current, one obtains
I = X) v |
(12-9)
where X is the linear charge density of the flow.

In subsequent discussions, we will be constantly dealing with what is known as a
current element and it is convenient to introduce it here. For the filamentary current of
12-2 THE E Q U A T I O N O F C O N T I N U I T Y
205
(a)
Figure 12-5.
line ds.
|p
(a) Definition of surface current density K. (b) K is not perpendicular t o the
(a)
Figure 12-6.
(b)
Calculation of equivalents t o a current element.
Figure 12-1, it is defined simply as the product Ids. In order to get an equivalent to
this for distributed currents, we consider Figure 12-6 a in which dA is the small
cross-sectional area of the filament. Since the flow is normal to the cross section, we see
from (12-5) that I = J dA so that Ids = J dA ds = J dr where dr is the volume of the
element shown shaded. Since J and ds are parallel in this case, we get Ids = J dr.
Similarly, for a surface current, we get the correspondence Ids = K da where da is the
area dlds of the element shown shaded in Figure 12-66. Thus we have the following
equivalent expressions for current elements:
Ids = Jdr = Kda
(12-10)
These results also show the reasons for the names "volume" and "surface" current
densities.
1 2 - 2 THE E Q U A T I O N O F C O N T I N U I T Y
In the introduction to Chapter 2, we mentioned that all experiments indicate that net
charge is conserved. We can express this fundamental law of conservation of charge in a
convenient quantitative manner in terms of the quantities we have just introduced.
Suppose that, in Figure 12-4, the surface S is a closed stationary surface bounding a
volume V. Now the total rate at which charge is flowing out through the surface S must
206
equal the rate at which the total charge within V is decreasing, since the total must be
constant. Therefore, if Q is the total charge within V, we find from (12-6), (2-14), and
(1-59) that
=
dt
^ J ' DA = - J f P dr = J f
dr = f V Jdr
Js
dt v
v dt
Jv
(12-11)
We were able to go from the third expression to the fourth as we did because V is a
volume of constant shape and size so that any limits of integration which may be
involved in the definite integral over V are independent of the time; in addition, p may
be a function of position as well as of time. Combining the last two expressions in
(12-11), we get
f J c
( 1 M 2 )
Since charge is conserved at all points, not at just part of a given volume, the integral
must hold for any arbitrary volume, including an arbitrarily small one located anywhere. Thus (12-12) can be always true only if the integrand is zero everywhere so that
dp
V J + = 0
dt
(12-13)
This important result is called the equation of continuity and is a mathematical

expression of the fundamental experimental result that net charge is conserved. We can
note here that processes like "pair production" and "annihilation" of, say, electrons
and positrons, do not violate this result because the net charge is constant since equal
amounts of positive and negative charge are "created" or "destroyed" in these
phenomena. Similar results are found to hold in more complicated reactions in nuclear
and high energy physics where large numbers of particles are produced; in all cases, the
net charge is conserved.
If we now combine (12-13) with (9-6) and (9-24), we get the boundary condition
satisfied by the current density at a surface of discontinuity as
da
h (J 2 - J x ) = j2n - Jln = - -yt
(12'14)
Physically, this condition expresses the fact that if more charge is brought up to the
surface than is taken away, charge will necessarily accumulate there, and conversely.
In the special case of steady currents, in which everything is constant in time, dp/dt
and do/dt will both be zero, and these last two results simplify to
V J = 0
h-(J2-J1)=/
2 n
-/1 = 0
(12-15)
(12-16)
Since all charge is conserved, p and J are clearly the total charge density and current
density, respectively. Now let us consider their component parts. We begin with the
bound charge whose density is pb. Now in the process of polarizing a material, the
bound charges will generally be moving, as we saw in Section 10-1, so that we can
define a bound charge current density Jb. Since the process of polarization involves
only separation of bound charges, or reorientation of dipoles, bound charges are
necessarily conserved, as shown by (10-13). Thus we must have a separate equation of
continuity for bound charges, that is,
dph
V Jh + = 0
(12-17)
12-3 C O N D U C T I O N C U R R E N T S
207
If we write ph = v P, as given by (10-10), we also have

d
dP
and, since this must be true everywhere, we can write

dp
J. - -
(12-18)
thus identifying the bound current density. This current is often called the polarization
current density and is a consequence of the process of polarization.
Since total charge is conserved, according to (12-13), and bound charge also is, as
shown by (12-17), the free charge must be conserved too and we can write
dpf
if H = 0
dt
(12-19)
which will lead to

ft (J / 2 - J / x ) = Jf2n - Jfln = -
(12-20)
in exactly the way by which we obtained (12-14). In the special case of steady currents,
these become
V Jy = 0
and
ft-
(jy 2 J } i ) = 0
(12-21)
Since free charges, and thus free currents, are the ones over which we have some
control, they are usually the ones in which we are most interested, and, consequently,
we will be concentrating on them for some time. Free currents are often classified
further into the two broad categories of conduction currents and convection currents,
although the distinction between them is somewhat ill defined. Generally speaking,
conduction currents include the motion of charges in conductors, that is, materials that
already contain mobile charges because of their intrinsic nature. The most common
example is that of currents in metals, although we can include, in this class, currents in
semiconductors and solutions of electrolytes. In the last case, positive and negative ions
resulting from the formation of the solution provide the ready-made carriers. On the
other hand, convection currents are usually associated with the motion of charged
particles in physical streams through otherwise empty space such as beams of ions,
electron beams in vacuum tubes, charged particles in the solar wind, and the like. This
class can also be taken to include the physical motion of macroscopically charged
bodies, such as we would get by moving around a piece of glass that had been
positively charged by electrification by friction. The class of most importance to us,
however, is that of conduction currents and we now consider that in more detail.
12-3
C O N D U C T I O N CURRENTS
As a convenient prototype of this case, we can think of the current in a metallic wire. In
(6-1), we found that E = 0 in the interior of a conductor for a completely static
situation. Now, with moving charges in the wire, we no longer have a static situation,
although it may be steady, so that it may well be that E = 0 within the conductor. In
fact, the very motion of the charges implies that there are forces on them that in turn
implies a nonzero value of E. Consequently, the conductor can no longer be expected to
be an equipotential volume.
We find experimentally that if we apply an initial difference of potential to a
conductor there will be currents in it, but if it is then left alone, the currents will
208
eventually cease to exist, and the conductor will have attained the state of electrostatic
equilibrium whose properties we have discussed in Chapter 6. It is also found that we
can maintain a constant current in a conductor by means of a constant potential
difference only if we also continuously supply energy to the system f r o m some external
source. Thus work is somewhere being done on these moving charges as they circulate
in the closed paths of the ordinary circuit. If the total work W is done on a charge q as
it goes around a closed path, the ratio of the two is called the electromotive force $, or
simply the emf, so that, with the use of (3-1), we have
S ^ = <F
ds = T(F>E ds
q
qTc
c
(12-22)
But, as we know from (5-5), the conservative electric field with which we are familiar
cannot do any net work on the charge in such a case as this, so that somewhere within
the circuit there must be a source or sources of a nonconservative electric field E c ; then
(12-22) can be written as
= <^>Enc ds
(12-23)
(Since electric field can be measured in volts/meter, we see that the unit for emf is the
voltthe same as for potential and potential difference.) Later we will discuss how
some of these nonconservative electric fields can be produced, but, for now, it is
sufficient to note that the most common and familiar of these sources are batteries. A
battery does work on a charge that passes through it and the source of this energy is
essentially due to chemical reactions of one sort or another within the battery; thus, in
a sense, a battery is analogous to a p u m p that can do work on a fluid and raise it, for
example, against the conservative gravitational field. The battery is an example of a
localized source of a nonconservative field so that E n c in (12-23) is different f r o m zero
only when the path of the charge is within the battery and E n c = 0 elsewhere on the
circuit. In this case, it is possible to speak of the emf of the source itself as a specific
quantity, and its value will be obtained from (12-23) as
= /
E . c d>
(12-24)
^source
For simplicity, therefore, we restrict the rest of our discussion to those conducting
regions that are free of nonconservative electric fields, so that we can write E = v<
and hence v X E = 0, as long as we stay outside of the batteries.
Since E will exert forces on the moving charges, there should exist some functional
relation between
and E, that is, we expect to be able to write SF = Jy(E). We will also
assume for now that J^(0) = 0, thereby excluding superconductors f r o m our considerations. This relation between JF and E can be quite complex and will depend on the
material. We dealt with a similar situation in Section 10-6 when we discussed the
relation between P and E for dielectrics, and we could construct a similar classification
scheme. However, we pass immediately to the case of a linear isotropic conductor, that
is, we assume that we can write
J7=aE
(12-25)
where the factor of proportionality a is called the conductivity. (This is a standard

notation and should not lead to any confusion; in those very few cases where we have
surface charge density and conductivity in the same expression, we write the surface
charge density as a c h .) Equation 12-25 is another example of a constitutive equation, as
is D = cE, and only experiment can decide whether it is suitable for a given material or
not. In (12-25), a is to be independent of the field E, although it may still depend on
12-3 C O N D U C T I O N CURRENTS
209
position, and other variables such as temperature. The unit of a is called 1 (ohmm e t e r ) - 1 and since
is in amperes/(meter) 2 and E in volts/meter, we easily find that
1 ohm = 1 volt/ampere.
When (12-25) is applicable, the boundary condition (12-21) for a steady current can
be written completely in terms of E:
ft- (<t 2 E 2 - A J E J = 0
(12-26)
We will still have ft X (E 2 E x ) = 0, as well, since V X E = 0 in all of the regions we
will consider. Thus, we have a situation similar to that we found for dielectrics in
Figure 10-14 in which the lines of E are refracted as we cross a bounding surface
between two media of different conductivities.
If the material is homogeneous as well, then a = const., that is, independent of
position. In other respects, a is a characteristic of the material and is to be found from
experiment or else calculated from the atomic properties of matter by means of the
theories used in other branches of physics; for our purposes, we take it as a given
quantity. Not all materials are linear, isotropic, and homogeneous, but the assumption
of (12-25) with a = const, holds very well for metals and solutions of electrolytes, for
example.
If we have a l.i.h. conductor and steady currents, we can combine (12-21), (12-25),
(5-3), and (1-45) to give V Sj = 0 = V (<*E) = <xV E = ov2<j>- In other words,
V 2<t> = 0 and the potential still satisfies Laplace's equation. This result provides one
with an experimental method of solving Laplace's equation by setting up the required
boundary values of <j> on the boundaries of a conducting region for then, by measuring
the magnitude and direction of the current density J f , one can find the values of E from
E = Jf/o.
The relation Jf = a E for a l.i.h. conductor is equivalent to the macroscopic empirical
relation known as Ohm's law and, in fact, is often called the microscopic form of
Ohm's law. We can see how this comes about by analyzing the situation depicted in
Figure 12-7. This is a portion of a uniform conductor of length / and cross-sectional
area A carrying a total current I. If |A<#>| is the magnitude of the potential difference
between the ends, then the magnitude of E = | A<f>|/7 according to (5-3) and (1-38)
since V< is a constant in this region of constant current and dimensions. Similarly, the
current density is
= I / A by (12-6), if we assume the current to be distributed
uniformly over the cross section, which turns out to be a very accurate approximation.
Substituting these into (12-25), we get I/A = ct| A<f>|//, or / = (oA//)|A<|>|. Thus, we
have a relation between these macroscopic quantities and we see that the current is
proportional to the potential difference, (or conversely). This is usually written in the
form
| A<^> I
/ =
(12-27)
R
where the proportionality factor is
R =
oA
(12-28)
|A0|
J/
Figure 12-7. Portion of a conductor carrying

a total current /.
210
Equation 12-27 is the empirical result discovered by Ohm and is known as Ohm's law,
while the quantity R is called the resistance, and is measured in ohms (i.e.,
volts/ampere). Thus we have shown the equivalence between (12-25) and (12-27) for a
l.i.h. conductor and have, simultaneously, obtained (12-28) as a method of calculating
the resistance. The reciprocal of the conductivity, 1 /a, is called the resistivity and is
very often written as p!
As we see from the figure, JF is directed longitudinally along this uniform conductor,
as is E since they are parallel. Thus, at the surface, E will be tangential, and since the
tangential components of E are continuous by (9-21), there will be a tangential field
outside the conductor which is given by E = J { / o . This is in marked contrast to the
static case where not only was E = 0 inside the conductor, but it was necessarily
normal to the surface, as we saw in (6-2) and Figure 6-1 b.
We conclude this section by using some of our results to obtain an interesting and
somewhat unexpected relationship.
Example
Relation between resistance and capacitance. Suppose we have two conductors of some
shape. We consider two ways in which we can put them to use.
1. As a capacitorlet us fill the region between these conductors with a l.i.h.
dielectric of permittivity c and put equal and opposite charges on them as shown in
Figure 12-8a. We want to find the capacitance. From (6-38), we know that we can find
the potential difference by evaluating the integral
= f E - d s
(12-29)
over any convenient path between the plates. We can write the free charge Q on the
positive plate as an integral over its surface S:
Q = Jf of da = J f eE da = eJ f E da.
(12-30)
s
s
s
with the use of (2-16), (10-56), and (6-1). Putting these into (6-38), we find that the
capacitance can be written as
C =
eJ/ E da
s
E ds
/;
(12-31)
2. As a resistancenow, instead of the dielectric, let us fill the region between the
plates with a l.i.h. conductor of conductivity o as shown in Figure 12-8b. We also
maintain the same potential difference A< between the plates by keeping each plate at
(a)
Figure 12-8.
(b)
Two conductors used (a) as a capacitor and ( b ) as a resistance.
12-4
ENERGY RELATIONS
211
the same potential as was used before. As we saw above, <f> satisfies Laplace's equation
under these circumstances, and since the boundary conditions are exactly the same in
the two cases, we know from the uniqueness theorem of Section 11-1 that <f>(r) will be
identical for each. In other words, the potential difference will again be given by (12-29)
with exactly the same values of E at each point on the path of integration. The total
current I passing between the plates can be expressed as a surface integral over the
same upper plate with the use of (12-6) and (12-25); the result is that
I J if da = j a E da = a J E da
(12-32)
s
s
s
When we put (12-29) and (12-32) into (12-27), we find the resistance of this system to
be given by
E
f E ds
J+
d a
Comparing (12-31) and (12-33), we see that 1 /Ro

i?C = o
(12-33)
= C / e or
(12-34)
showing that these two properties of the system are not independent but are in fact
related in this simple way. [This relation (12-34) is reminiscent of a typical thermodynamic result as it expresses a relation between macroscopic properties of a system
without giving an indication of the absolute value of either.] This very general result
also can be used as a means of measuring C indirectly, since resistance measurements
are comparatively easy to perform with an ammeter and a voltmeter, while a direct
electrostatic measurement of C is generally much harder to do.
12-4 ENERGY RELATIONS
In the previous section, we mentioned the experimental fact that the maintenance of a
steady current in a conductor is possible only if energy is constantly being supplied to
the system. But in a steady situation, where things remain constant in time, there
cannot be an accumulation of electrical energy either. Consequently, what is supplied
as electrical energy must be converted into another form of energy, and, in fact, it is
observed that the electrical energy expended appears as heat produced within the body
of the conductor. We can put this into quantitative form in the following way. We saw
in (5-45) and (5-46) that the work done by the electric field on a charge Aq is
A W= Aq A<>
| where A4> is the change in potential. Therefore the rate at which
electrical work is being done on the system is AW/At = (Aq/At) A(j> = IA<f> by
(12-1) since we have a steady current. In the steady state, this must also be the rate of
conversion of energy into heat. If we let w be the rate of production of heat per unit
volume, then we find by using the dimensions used in Figure 12-7, where the volume is
Al, that
Since 3 f and E are parallel, we can also write this as

J
f
w = JJf E = oE2 2 =
o
,(12-35)
212
with the use of (12-25). The quantity w is also referred to as the power "dissipation"
per unit volume. Since our unit, of power is 1 watt = 1 joule/second, w will be
measured in watts/(meter) 3 .
I 2-5| A MICROSCOPIC POINT OF VIEW
This result that we have just obtained is perfectly general since it follows directly from
a combination of the macroscopic laws of conservation of energy and the meaning of
current and potential difference. Nevertheless, its origin may seem obscure and one
may feel the need of an "explanation" in terms of the microscopic average behavior of
the moving charges in the conductor. Related to this question of " w h y " is there a
production of heat is the problem of " w h y " does a conductor have a resistance in the
first place. Although we leave a more comprehensive discussion of these questions to
Appendix B, we can easily enough get a semiquantitative answer to both of these
questions from a microscopic picture.
First of all, if the electric field produced the only force on the mobile free charges,
they would have a constant acceleration given by a = F q / m = qE/m where m is the
mass of the charge carrier. But a constant acceleration will result in an indefinitely
increasing velocity, which is not observed. A steady current means, according to (12-3),
a constant velocity and hence zero acceleration, that is, a zero net force. Therefore, the
electrical force, which is in the direction of motion of the charges, must be balanced, at
least on the average, by another force directed opposite to the motion. In order to get
an idea of the origin of this force, let us consider the specific case of a metal where the
free charge carriers are electrons each of charge e. They move about, not in
completely empty space, but among the ions of the metal that are arranged in the
regular array of the crystal. The electrons can certainly collide with these ions (as well
as with each other), and, when they do, their velocity will be changed. Between
collisions, they will be accelerated by the electric field, but on collision, the result of this
process will be abruptly changed. Thus, it is the collisions that are the origin of the
other force, and it is the average effect in which we are interested. By analogy with
similar effects involving "friction" in mechanics, we shall try to describe the overall
effect of collisions as their giving rise to a force proportional to the velocity and
opposite to it; thus we write the mechanical force as E q m = v where is an
appropriate proportionality factor. The net force will then be the sum of the electrical
and mechanical forces, or
F
net =
m a =
- e E
(12-36)
We see that this equation of motion allows a situation in which the acceleration a = 0;
then the velocity will be a constant, \ d , which is found from (12-36) to be
This velocity is commonly called the drift velocity, in mechanics it is usually known as
the terminal velocity, as, for example, in the case of an object falling near the earth and
also subject to the viscous (resistive) drag of the air. If n is the number of electrons per
unit volume, the free charge density will be pf = ( e), and if we substitute this, along
with (12-37), into (12-3), we find the value of the steady free current to be
(12-38)
But this is exactly of the form of Ohm's law given by (12-25) and, by comparison, we
1 2 - 6 THE A T T A I N M E N T O F E L E C T R O S T A T I C E Q U I L I B R I U M
213
see that not only have we accounted for its general form from this simplified
microscopic view, but we also have obtained an expression for the conductivity:
ne2
o =
(12-39)
Qualitatively, this result has reasonable properties. It is proportional to the number

density of carriers, and inversely proportional to the "frictional" term , so that the
smaller the effect of collisions, the larger will be the current for a given E and vice
versa. We also note that it is proportional to the square of the charge so that it is
actually independent of the sign of the charge of the carriers. We cannot go any further
in trying to calculate a until we can calculate . This, however, involves the detailed
theories of solid state physics and the analysis of collisions that we cannot pursue
further here. On the other hand, one can use measured values of CT to evaluate the
factor .
We have just seen how the overall effect of collisions can account for the observed
finite value of the current. We can also see how the existence of collisions can account
qualitatively for the conversion of electrical energy into heat described by (12-35). Not
only does a collision change the electron velocity in direction, but by changing its
magnitude, also changes the kinetic energy of both the electron and the struck ion.
Thus there is a transfer of energy from the ordered motion of the electron produced by
the field to the disordered random vibrations of the ions constituting the crystal. As we
know from thermodynamics, this type of process is an irreversible one, so that the net
effect of the collisions will be to increase the disordered energy of the metal crystal. But
it is just this increase in disordered motion that is associated with the production of
heat. As we mentioned, this is exactly what is observed, and we can see now that it too
is a necessary consequence of the same microscopic processes that result in a finite
conductivity.
12-6
THE A T T A I N M E N T O F ELECTROSTATIC EQUILIBRIUM
As we know from Chapter 6, if we put some free charge on or within a conductor, the
system will not generally be in equilibrium and it will have to readjust itself by means
of currents until the final electrostatic equilibrium state is attained for which all the
charge is on the surface and the conductor is an equipotential volume. We do not,
however, know any of the details of this process, nor do we know how long it will take
in a typical case. We can use some of the results we have just obtained to give us an
idea of the nature of this process; it will involve a nonsteady situation since the currents
involved will eventually become zero.
We consider a l.i.h. conducting dielectric for which we can write Jf = oE and
D = cE. Since the free charge density can be changing in time, we use (12-19) and the
above relations, along with (10-41), to get an equation involving pf only:
dpf\
-
{ oD \
V(E)-V( 7 j -
a
7
v . D -
p ,
so that pf satisfies the differential equation
dt
-Pf
(12-40)
= Pf( 0)e-'^
(12-41)
which has the solution

P / ( 0 = PffO)e-l/t
214
where T = e/a and pf{0) is the initial value. This result says that the free charge density
in a l.i.h. material can only decrease! [One can make it increase, of course, but only by
means other than conduction processes described by J p o E such as, for example,
injecting a pulse of electrons that come to rest within the body of the material.
However, after the termination of the pulse, the charge density that was produced will
then decay exponentially as described by (12-41).]
Thus, as long as conduction processes are the only mechanism by which equilibrium
(p f = 0) is produced, we see that the approach to equilibrium will be one of exponential
decay. The charge density will decrease by a factor of l / e in time r = e / o . This
particular behavior is often given the name of relaxation; hence r is called the
relaxation time, and its value gives us an estimate of the time required for attainment of
equilibrium.
Actually, the simple result we have obtained is only applicable to a poorly conducting dielectric. For a very good conductor such as a metal the situation turns out to be
much more complicated. 1 In this case, the fields change so rapidly in time that one has
to include magnetic effects as well and some of the phenomena that will not be
discussed until Chapters 17, 21, and 26. One finds that the shape and size of the
conductor affect the time taken to attain final equilibrium and that this time is generally
much longer than simply T. We can get a rough intimation as to why this can be the
case by estimating the value of T for a typical metal. For most metals, c = e 0 , and for
copper a 5.8 X 10 7 /hm-meter, s o that r = e/o (8.85 X 1 0 ^ 1 2 ) / ( 5 . 8 X 10 7 ) ~
10" 1 9 second. This is a very short time, but it is also much smaller than the average
time between collisions, which is about 10 14 second, as can be deduced from
experiment and a more detailed form of (12-39). Since we have assumed collisions to be
the physical origin of a finite conductivity, it is unreasonable that effects depending on
conductivity can occur on such a much shorter time scale, and hence that a simple
expression like (12-41) can really describe the situation in metals.
If the bound charges in a dielectric do not have the equilibrium distribution
described by all of the results of Chapter 10, they too will relax in some manner toward
equilibrium. Here, however, conduction is not generally available as a mechanism, or is
negligible, and the primary processes of relaxation usually involve thermal agitation in
one way or another. This is a separate and large field of study all by itself and we are in
no position here to consider its details, and must simply content ourselves with the
observation that these phenomena exist.
EXERCISES
12-1 At a given instant, a certain system has a

current density given by J = A(x 3 x + y3y + z 3 z)
where A is a positive constant, (a) In what units
will A be measured? (b) At this instant, what is
the rate of change of the charge density at the
point ( 2 , - 1 , 4 ) meter? (c) Consider the total
charge Q contained within a sphere of radius a
centered at the origin. At this instant, what is the
rate at which Q is changing in time? Is Q increasing or decreasing?
12-2 Verify (12-7), (12-8), and (12-9).
12-3 A total charge Q is distributed uniformly
throughout a sphere of radius a. The sphere is
then rotated with constant angular speed to about

a diameter. Assume that the charge distribution is
unaffected by the rotation, and find J everywhere
within the sphere. (Express it in spherical coordinates with the polar axis coinciding with the axis
of rotation.) Find the total current passing through
a semicircle of radius a fixed in space with its
base on the axis of rotation.
12-4 A dielectric sphere of radius a is uniformly
polarized. It is rotated with constant angular speed
co about the diameter parallel to P. Assume that P
is unaffected by the rotation and find the currents. Plot your result as a function of the ap-
' H . C. Ohanian, Am. J. Phys , 51, 1020-1022 (1983).
EXERCISES
O-
i?,
R2
-AAA/WV
215
T'
-o
(a)
Figure 12-9.
(a) Resistors in series. ( b ) Resistors in parallel.
propriate coordinate(s). Find the total current

passing through a semicircle of radius a fixed in
space with its base on the axis of rotation.
12-5 A sphere of radius a centered at the origin
is made of a l.i.h. conducting material. The potential on the surface is maintained at the values
given in spherical coordinates by 4>0 cos 0 where
</>0 = const. Find the free current density Jf everywhere inside.
12-6 Two "resistors" of resistances Rx and R 2
are connected by conductors of negligible resistance to each other and to terminals T and T' as
shown in Figure 12-9a, that is, they are connected
in "series." If a potential difference A<j> is now
applied across the terminals, show that this combination is equivalent to a single resistor of resistance Rs Rl + R 2- Similarly, show that the
equivalent resistance of the "parallel" combination shown in (b) can be found from 1 /Rp =
(1 / * , ) + ( l / * 2 ) .
12-7 A long straight wire carries a steady current I. The current is distributed uniformly over
the circular cross section of radius a. Consider a
length / of the wire that has a resistance R. Find
the electric field in the vacuum region just outside
the wire and express it in terms of the quantities
given. (Overall, this portion of the wire is neutral.)
12-8 Consider a material with a nonzero conductivity. Assume the material to be linear and
isotropic in both its dielectric and conducting
properties. Assume steady free currents, (a) If the
material is inhomogeneous, show that there will
be a volume density of free charge in the material
given by
P/
= Jrv(-)
(12-42)
(b) Show that, at a surface of discontinuity between two materials, there will be a free surface
charge density given by
a
fch
(12-43)
(Since all complete circuits will have interfaces
between different materials, such as connections

between wires and battery terminals, there will be
accumulations of charge within the system that
will be constant in time. Thus, the above results
show us that it is these distributions of electric
charge which are the sources of the electric field
in systems carrying steady currents.)
12-9 Two large parallel plane conducting plates
are a distance d apart. The region between them
is filled with two l.i.h. materials whose surface of
separation is a plane parallel to the plates. The
first material (with properties ox and e : ) is of
thickness x, while the second material (cr 2 ,e 2 )
has thickness d x. There is a steady current
between the plates that are kept at constant
potentials of <j>1 and 4>2 Find the potential at the
surface of separation of the two media, and the
free surface charge density there.
12-10 The region between the coaxial cylinders
of Figure 6-12 is filled with a l.i.h. conducting
material. If a potential difference A<J> is applied
between the cylinders, find the current between
the cylinders for a length L of the system.
12-11 Consider an infinite slab of conducting
dielectric of thickness d and with plane parallel
faces. Two metal cylinders, each of radius A, have
their axes parallel and a distance D apart. These
cylinders pass through the dielectric with their
axes normal to the plane surfaces. If a potential
difference A<#> is applied between these metal
"electrodes," find the current I that will pass
from one to the other.
12-12 A very long wire of radius a is suspended
near the bottom of a very deep lake. Assume the
lake to have a plane bottom that is a very good
conductor. The wire is parallel to the bottom and
at a height h above it. If the conductivity of the
water is a, find the resistance between the wire
and the bottom for a length L of the system.
12-13 Show that when (12-35) is integrated over
the total volume of a uniform conductor, the total
rate of production of heat can be written as I2R.
216
12-14 Show that, when a complete circuit includes a localized source of emf , the steady
state current / will be given by I = S/R where R
is the total resistance of the whole circuit. (See
previous exercise.)
12-15 Show by direct integration that when a
capacitor is charged by a battery of emf f, the
amount of heat developed in the circuit is equal
to the final electrostatic energy of the capacitor.
12-16 Two square metal plates each 0.1 meter
on a side are parallel to each other and separated
by a distance of 1 0 ' 2 meters. The region between
the plates is filled with water of conductivity 10" 3
(ohm-meter)" l . If a potential difference of 150
volts is maintained between the plates, find the
rate of change of temperature of the water. (Neglect end effects and heat losses to regions outside
the water.)
12-17 A copper wire 2.5 millimeters in diameter
is carrying a current of 10 amperes. Assume one
free electron per copper atom and find the magnitude of the drift velocity. [Copper has a density of
8.92 grams/(centimeter) 3 , an atomic weight of
63.5 grams/mole, and Avogadro's number is 6.02

X 10 23 /mole.]
12-18 In Section 12-5, we discussed a microscopic picture that associated the existence of
resistance in a metal with collisions between the
mobile electrons and the metal ions. Based on this
viewpoint, what do you think will be the effect on
the resistance if the temperature of the conductor
is increased?
12-19
time.
Verify that ( / a has the dimensions of
12-20 Find the relaxation time for a glass with a

resistivity of 1012 ohm-meter and a relative permittivity of 4.0.
12-21 The region between the plates of a capacitor is filled with a material of finite conductivity
with resultant total resistance R. Show that the
charge on the plates will decrease to l/e of its
initial value in a time RC where C is the capacitance. [Note the similarity between this result and
(12-34) and (12-41). But, of course, different
processes are being describedor are they the
same?]
i j
*
Ni
AMPERE'S LAW
Interestingly, the general subject that we nowadays call magnetism also began with the
observation that certain naturally occurring minerals could attract other materials. The
name presumably derives from the association of these objects with the ancient cities of
Magnesia in Asia Minor since they could be found nearby. For centuries, the subject
was thought to be independent of electricity. During this time, however, people did
learn about the magnetic properties of the earth and invented the magnetic compass.
The first indication of the connection between electricity and magnetism occurred in
1819 when Oersted somewhat accidentally discovered that an electric current could
exert forces on a magnetic compass needle. Ampere heard of Oersted's result and
quickly found that an electric current could also exert a force on another electric
current. He began a systematic study of these forces, and, by means of a series of
ingenious and elegant experiments during the period 1820-1825, he was able to deduce
the form of the basic law of force between electric currents. About 50 years later,
Maxwell described Ampere's work as being " o n e of the most brilliant achievements in
science."
These forces between steady currents are what we will take as our starting point, and
we leave to a later chapter the description of magnetic materials as well as the effect of
matter in general. Since currents are involved, the charges are not at rest but moving.
However, as we saw in the last chapter, the charges are moving with constant average
speed in steady currents. Under these circumstances, the forces involved will also be
constant in time; thus we are justified in referring to this subject as magnetostatics.
13-1
THE FORCE BETWEEN T W O COMPLETE CIRCUITS
In Section 2-2 where we introduced Coulomb's law, we dealt with individual point
charges. One might expect, by analogy, that here we would similarly consider "little
pieces" of the current and study the forces between them. The natural thing to think of
in this context is the current element of (12-10) since it involves both the magnitude of
the current and its local direction. In the laboratory, however, one must, of necessity,
deal with complete closed circuits for steady currents, so what one has to do is to take,
as a fundamental experimental law, that which describes the total force between two
complete circuits. Once this has been done, one can try to deduce from this, as best one
can, a formula that gives the force between the hypothesized current elements. This is
what we will do.
We consider two idealized complete circuits C and C , carrying steady filamentary
currents / and I', respectively. The situation we have in mind is shown in Figure 13-1.
What we want is the total force on C due to C', which we write as F c , _ c . The circuits
are idealized in the sense that the batteries are not included in the diagram. It is
assumed that the necessary batteries are some distance away and the conducting leads
from them are twisted closely together, since one of Ampere's first experiments showed
that two oppositely directed currents close together produced no effect on another
current. In spite of what was said above, the result for the total force is written in terms
of the current elements Ids and I' ds' shown; their location with respect to an
217
218
AMPERE'S LAW
arbitrary origin is given by their position vectors r and r' and the relative position
vector is defined as usual by R = r - r', that is, it is drawn from the "source point" to
the "field point." (A comparison of Figure 13-1 with Figure 2-2 is instructive.) We
assume a vacuum throughout the region not occupied by currents.
The basic experimental law that gives the total force on C due to C' can be
written as
(13-1)
and is what we will take as Ampere's law. We note that it has the form of a double line
integral, each of which is taken over the corresponding circuit. It strains one's credulity,
of course, to have it asserted that a formula of the generality implied by (13-1) could
have been deduced from a very few experiments on circuits of special and simple
shapes as was done by Ampere. Equation 13-1 evidently represents a generalization
from results that have been found for very many special cases.
The factor n 0 / 4 v that appears in (13-1) is a constant of proportionality whose
numerical value depends on the system of units that is being used; the dimensions of ju,0
are seen to be those of force/(current) 2 . In the SI unit system which we are using (in
effect, MKSA units), jx0 is defined to have precisely the value
jli0 =
X 10
newton/(ampere) 2
= 47t X 1 0 " 7 h e n r y / m e t e r
(13-2)
This constant ju,0 is called the permeability of free space, and is generally written in the
last form from which we see, by comparison of both forms, that 1 henry = 1
joule/(ampere) 2 . Since the newton and meter are already determined by other independent definitions, we see that this definition of jli0 fixes the unit of current that is all that
remains undefined in (13-1). Thus, (13-2) essentially gives us the definition of the
ampere and thence that of the coulomb.
Figure 13-1. Relations bet w e e n t w o circuits as used

t o state Ampere's law.
13-1 THE F O R C E BETWEEN T W O C O M P L E T E CIRCUITS
219
If there are several circuits that can interact with C, the force of each will be given by
an expression like (13-1), and the total force on C will be the vector sum of them:
Fc = F t o W < , c - E r c . - c
(13-3)
Although we can imagine generalizing (13-1) and (13-3) to include distributed currents
with the use of (12-10), it will be convenient to continue with filamentary currents only.
The integrand in (13-1) is more complicated from a directional point of view than is
Coulomb's law as expressed by, say, (2-15), since the integrand depends on the relative
orientation of the three quantities Ids, I' ds', and R. We also note that the quantities
relating to the two circuits appear unsymmetrically in the integrand. This observation
may be somewhat disquieting because it may make us wonder that, if the roles of C
and C' were interchanged, the force on C' due to C might not be equal and opposite to
that on C due to C', as we would expect from Newton's third law as applied to these
macroscopic complete circuits. However, this asymmetry is only apparent as we can see
by rewriting our result. Using (1-30), (1-141), and (1-38), we find that we can write
Ids*.
(I' ds' X R)
/
= H'ds'\ds
\~~
R \1
II'R(ds-ds')
R2
R1
II'R(ds
A-V|i
II' ds'
( 1 \
= -Il'ds'dc
ds')
2
R
II'R(ds
ds')
(13-4)
where we have written dc(\/R)

to indicate the differential change in (1 / R ) resulting
from a displacement along C. Inserting (13-4) into (13-1), we find that the total force
can also be written as
iXr,II'
/ 1 \
r r
{ds
^s')R
where, in the first term, we integrate over C first and then over C'. The integral over C
has the form of an integral over a closed path of the differential of a scalar, and since
the final and initial points coincide, we have
and the first term of (13-5) thereby vanishes, leaving us with

u0//'
t
(ds ds')K
<
as another expression that will give us the same total force. We can, in fact, regard
(13-6) as another version of Ampere's J a w in vacuum. In this form, the two circuits
appear more symmetrically except for R, which has a definite sense. If we were to use
(13-6) to calculate the force on C' due to C, we would get
r
r (ds' ds)R'
a*
d " )
where R' = r' r. Since R' = \r' r| = R, we have R' = R; also ds' ds = ds ds'
by (1-16). Thus, on comparing (13-6) and (13-7), we see that
Fc^c= -Fc-c
(13-8)
as is to be expected for the total forces.
220
AMPERE'S LAW
This calculation shows us that there is a lot of ambiguity in the expression for the
total force on a complete circuit that, after all, is all one can really measure in the
laboratory. In fact, we could write down an innumerable set of " versions" of Ampere's
law by simply adding to the integrand of (13-1) any function that vanishes when
integrated over a complete circuit. Long experience has shown, however, that nothing is
to be gained from this and the formulation given in (13-1) is the most useful and is to
be preferred for many reasons as we will see.
We are assuming a static situation, that is, the circuits are at rest at fixed positions.
This means that in order for C to be in equilibrium there must be an additional
mechanical force F c m on it so that the net force will be zero; in other words, we must
have
Fc^c+Fc,m =
(13"9)
Similar remarks apply to C'.

As we said before, Ampere's law (13-1) is a generalization of results from many
special cases. As usual, in order to get any use out of this law one has to evaluate it for
special cases in order to obtain results that can be checked relatively easily in the
laboratory. As an illustration of the use of (13-1), we will apply it to a particularly
simple and important case.
13-2
T W O INFINITELY L O N G PARALLEL CURRENTS
We consider two infinitely long straight circuits carrying currents / and I'. They are
parallel to each other and a distance p apart. In order to be specific, let us use
cylindrical coordinates and choose the z axis to coincide with the source current / ' as
shown in Figure 13-2. These currents must each be part of a closed circuit, of course;
we assume that the closing portions he along curves at infinity (such as large semicircles) and hence are so far away that their contribution can be neglected because of
the R2 in the denominator of (13-1). We see from the figure that r = pp + zz and
r' = z'z so that R = r r' = pp + (z z')z and R2 = p2 4- (z z')2. We also see
that ds dr = dzz and ds' = dr' = dz' z since both p and p are constants in this
Figure 13-2. Calculation of force bet w e e n t w o infinitely long parallel currents.
1 3 - 2 T W O I N F I N I T E L Y L O N G PARALLEL C U R R E N T S
221
case. Using these results and the fact that R = R / R , we find that
ds X ( d s ' X R)
dzdz' z X (z X [pp + (z z')z] }
R2
R3
p dz dz' p
[P2 + ( z - z ' ) 2 ] 3 / 2
(13-10)
with the use of (1-76) and (1-26). Substituting this into (13-1), we obtain
-.11 Ulf sOO
V-oII'PP
ifOO fOO
/oo
F/"' r
ucdz'
dz /
7 2
^77
oo J - oo ^p + ( z z ' ) j
4v
(13-11)
where the limits of integration will cover the circuits C and C'. If we make the
substitution / = z ' - z so that dt = J z ' since z is kept constant during the integration
over z', we find that the integral over z' becomes
dt
/:
00 (p 2 + t2)3/2
P2
according to (3-7) and (3-8). Therefore, (13-11) reduces to

U, II'p roo
V
'/ ^
2 77p ^ oo
n
(13-12)
If we were now to carry out the integration over z, that is, over C, we would get an
infinite force. We can, nevertheless, still get something useful from our result by noting
that the integrand is independent of z, so that the force on a length dz will be given by
d F = ([I0II'P/2TTP)
dz. Thus, if we introduce a force per unit length, f c , we can write
f
=
dz
M q / /
= -
2 up
i Parallel \
\ currents /
(13-13)
We see that f c has constant magnitude, is proportional to the product of the currents,
is inversely proportional to the separation p, and is directed perpendicular to the
currents. In fact, since the coefficient of p in (13-13) is negative, we see from Figure
13-2 that this is a force of attraction of C toward C'.
In Figure 13-2, we assumed that I and I' were in the same direction. If, now, I and
/ ' were oppositely directed, then either ds or ds' would be of opposite sign to that it
previously had since the line element is defined to be in the same sense as that of the
current. Then we see from (13-10) that the double cross product would have its sign
changed and (13-13) would become
l c
P
~
/antiparallel\
\ currents /
2up
which is one of repulsion on C.

We can sum up the qualitative nature of these results by saying that parallel ("like")
currents attract, while antiparallel (" unlike") currents repel', this is sort of "opposite"
to the behavior that we recall for the electrostatic forces between point charges.
A particular case of (13-14) that is of interest is that for which the currents are
equal; when 1 = 1', it becomes
f cc
- ^
2irp
(V1 3
. 1 5 7)
This arrangement can be attained in practice by making the very long straight parallel
lines C and C' in Figure 13-2 portions of the same circuit by connecting the ends at
222
AMPERE'S LAW
infinity to close the circuit. In this case one is certain to have equal and opposite
currents so that (13-15) is applicable. But now, with n 0 defined by (13-2), there is
nothing arbitrary left in (13-15) so that the absolute value of I in amperes can be f o u n d
from a determination of the mechanical quantities f c and p. In other words, (13-15)
can be used to measure / ; this is essentially the method that is actually used to
determine the value of the ampere. Once any current I has been found in this manner,
the value I' of any other can be found in principle by using it in this arrangement so
that either (13-13) or (13-14) describes the measurable force per unit length.
13-31 THE FORCE BETWEEN CURRENT ELEMENTS
Although it is often unwise to draw conclusions about an integrand from the nature
and value of a definite integral, we have done it before, for example in getting (7-29),
and we will be doing it again. The form of (13-1) is such that it can be written as
Fc^c=$J>cdFe,ê
(13-16)
and the integrand

u 0 IdsX
(I' ds' X R)
( m 7 )
TP
given a natural interpretation as the force exerted on the current element Ids by the
current element I' ds'. We could call this still another version of Ampere's law and, in
fact, this is very often what is done. As we pointed out, it cannot be checked directly
with experiments on current elements but it does allow a verifiable generalization to
moving point charges and it is consistent with all that we shall do later. In this form,
(13-17) is more nearly the analogue of Coulomb's law that we would expect to find,
including the inverse square dependence on the distance R between the elements.
In contrast to Coulomb's law, however, dFe,_,e is not generally along the line
connecting the elements, the direction of which is given by R. As a consequence,
Newton's third law is not satisfied by (13-17), although we saw it to hold for the overall
circuits in (13-8). In order to show this, let us calculate the force exerted on I'ds'
by
I d s ; as we saw before, we can get it by interchanging the primed and unprimed
quantities in (13-17) and by replacing R by R. In this way, we find that
u 0 I'ds'X
- " 4 ?
(Ids
X R)
(13
"18)
where R = r r ' still. If we add (13-17) and (13-18), we find, with the use of (1-30),
that
dFe,_e
+ dF_,
Mo II'
(ds' x ds)]
(13-19)
4TTR"
which is generally different from zero, in contrast to (13-8). This sum will be zero only
if ds' and ds are parallel as in Figure 13-2, or if R is perpendicular to the plane formed
by ds' and ds.
If we do the same thing with our more symmetrical form (13-6), we would get still
another possible form for the force between current elements, namely
d
H0 [(Ids)
4 TT
(I'ds')]R
,
(13-2)
R'
which is directed along the line connecting the elements. It is also easy to verify that
EXERCISES
223
this form satisfies the third law, that is

d^
4- d ^ e , = 0
(13-21)
Although (13-17) and (13-20) will lead to the same value for the total force between
complete circuits, they can give different results when applied to elements. As an
example, suppose the current I of Figure 13-2 were directed into the page while / ' is
kept the same. Then we would get a picture for these "crossed" currents like that of
Figure 13-3. In this case, ds and ds' are perpendicular so that (13-20) gives d'e, e = 0,
while from (13-17) we find that de,_>e = 0. The reason for this discrepancy is that in
going from (13-1) to (13-5) we, in effect, lost a term that is zero for a complete circuit
but is not zero for the integrands; this was shown explicitly in (13-4).
In spite of the more symmetrical appearance of (13-20), its direction along the line
connecting the elements, its desirable property expressed in (13-21), in contrast to
(13-19), why is it that (13-17) is the accepted form for the force between current
elements rather than (13-20)? Putting aside future comparison with experiment for the
moment, we can see that (13-20) has an inherent problem that makes it unsuitable for
our purposes. Because of its proportionality to ds ds', (13-20) involves the cosine of
the angle between the elements. Thus it cannot be written in the form of a product of
the element Ids and something that is independent of the element, that is, a field.
Since this product property is essential to the development of a theory based on fields,
we abandon (13-20). On the other hand, (13-17) does possess the form needed to enable
us to introduce a new field, as we will immediately proceed to do, and hence we adopt
the form (13-1) along with (13-17) as our fundamental law of magnetostatics.
EXERCISES
13-1 Apply (13-6) to the system of Figure 13-2

and thus show directly that (13-12) is again obtained.
13-2 In order to get an idea of the magnitude of
magnetic forces, find the force per unit length
between very long equal and opposite 10 ampere
currents separated by 1 centimeter. Find the ratio

of this to the weight per unit length of the copper
wire described in Exercise 12-17.
13-3 Consider the two infinitely long straight
currents shown in Figure 13-4. I' coincides with
the y axis. / is parallel to the yz plane, is at a
%
224
AMPERE'S LAW
13-5 An infinite plane current sheet coincides

with the xy plane. The surface current density is
K' = K'y where K' const. A very long wire
carrying a current I is parallel to the y axis and
intersects the positive z axis at a distance d from
the origin. Find the force per unit length on the
wire.
Figure 13-4.
The t w o currents of Exercise 13-3.
distance p from it, crosses the x axis aL y = z = 0,

and makes the angle a with the xy plane as
shown. Show that the force on / of C due to / ' of
C' is cot ax.
13-4 Consider the two circuits shown in Figure
13-5. All currents lie in the same plane. C' is
infinitely long. The sides of the rectangle of length
b are parallel to C'. Find the total force on C. Is
it attractive or repulsive?
/'/ \
I
\
Figure 13-5.
The circuits of Exercise 13-4.
13-6 A current / ' is distributed uniformly over

a very long cylinder of circular cross section of
radius a. The axis of the cylinder coincides with
the z axis. / ' is in the direction of positive z. A
very long current / is along a line parallel to the
z axis; it is also in the direction of positive z. /
crosses the positive * axis at a distance d from
the origin. Find the force per unit length on the
current I. [You will probably need to use (3-16).]
13-7 A circular loop of radius a lies in the xy
plane with its center at the origin. It carries a
current / ' that circulates counterclockwise as seen
when looking back at the origin from positive
values of z. A very long current I is parallel to
the x axis, is going in the positive x direction,
and intersects the positive z axis at a point a
distance d from the origin. Find the total force
on the circuit C carrying I.
13-8 Two circles, each of radius a, carry currents circulating in the same sense. One, with
current /', lies in the xy plane with center at the
origin. The other, with current /, is parallel to the
xy plane with its center on the positive z axis at a
distance d from the origin. Find the force on the
circuit with current I. Express your answer in
terms of the angles y and <{/ of cylindrical coordinates, but leave your answer in integral form.
13-9 Four very long straight wires each carry
current of the same value I. They are all parallel
to the z axis and intersect the xy plane at the
points (0,0), (a, 0), (a, a), and (0, a). The first
and third have their currents in the positive z
direction; the other two have currents in the
negative z direction. Find the total force per unit
length on the current corresponding to the point
(a, a).
14
Ampere's law is another example of an "action at a distance" law. At the beginning of
Chapter 3, we discussed the similar property of Coulomb's law and we concluded that
it would be useful to divide it into two parts by introducing the electric field as a kind
of intermediary to describe the interaction between the charges. We now want to do the
same thing for the force between currents. For historical reasons, the field that we will
use for this purpose is called the "magnetic induction"; the term "magnetic field" is
generally used for a different vector field that we define later when we include the
effects of matter.
14-1 DEFINITION OF THE MAGNETIC I N D U C T I O N
We recall that we were able to go from Coulomb's law to the electric field because we
saw in Section 3-1 that the force on a charge could be written as the product of the
charge and another quantity that we called the electric field. We now see that (13-1) can
be rewritten in a somewhat similar fashion as
F
f
=
/ Mo /- I' ds' X R \
i
[ ^ i , ^ j
<i4
-i)
The factor in parentheses is independent of the current element I d s at the position r,

but it does depend on the distribution of the other current elements with respect to
I ds. If we represent the quantity in parentheses by B(r), and use (1-137), we have
u0
B(r)
= ~(f>
w
4tt- Tc>
I' ds' X R
R
a0 r I' ds' X R
= i r i
^
4?r T c ,
R3
v(14-2)
and
F c , _ c = (lds
X B(r)
(14-3)
The vector field B, which we have defined in this way, is called the magnetic induction;
sometimes it is also known as the magnetic flux density or simply as the B field. In
addition, (14-2) is generally known as the Biot-Savart law.
We see from (14-3) that the unit of B will be 1 newton/(ampere-meter). This
combination is not normally used, however, and the standard unit for B is generally
written in one or another of two forms: 1 tesla = 1 weber/(meter) 2 . By comparing this
with the above we see that 1 weber = 1 j o u l e / a m p e r e = 1 volt-second.
By means of this procedure, then, we have introduced another vector field B that we
can calculate at any field point r by means of (14-2) even if there is no current element
there to have a force on it. Again, as for E, one can regard this as all simply a
mathematical convenience in that the calculation of B provides one with a kind of
distributed contingency statement that tells us what the force on a circuit C would be if
it were put there. On the other hand, one can, as we discussed in Section 3-4, regard B
as a real physical entity in itself.
225
226
If there is more than one source circuit, we can find B for a given one, C ( , by means
of (14-2), and then the resultant induction at r will be given by the vector sum of the
individual contributions:
ota'
B r
()
Mo r h
^4^%
/
R2
i
R,-
..
(14-4)
where R, = r r, in terms of the position vector r( of the current element I, ds, of the
/th circuit. We note that Ids of C is not included in (14-4), that is, we do not envisage
a current element exerting a magnetic force upon itself.
Although (14-3) was written in terms of the total force on the complete circuit C, it
is natural to interpret the integrand as giving the force rfF on the current element I d s
located at r:
d = Ids x B(r)
(14-5)
This force is perpendicular to both the current element and the magnetic induction, is
zero when they are parallel, and has a maximum magnitude when they are perpendicular, as we see from (1-22). This directional property of d F is illustrated in Figure 14-1,
which corresponds to the definition of the cross product shown in Figure 1-14.
In a similar manner, we can interpret the integrand of (14-2) as the contribution
J B ( r ) to the total induction produced by the current element I'ds'
located at the
source point r ' :
ju0 I' ds' X R
u 0 I' ds' X R
d B(r)
=
=
=
(14-6)
w
2
v
7
4tt
R
4Tt
R3
This expression is another version of the Biot-Savart law. The directional relations
given by (14-6) are shown in Figure 14-2. We see that dB is perpendicular to the plane
formed by ds' and R, has its maximum magnitude at a field point located on a line
perpendicular to ds', and is zero at a field point located directly forward of or behind
ds'. There is a convenient right-hand rule that qualitatively describes the situation of
Figure 14-2: point the thumb of the right hand in the direction of I' ds' and then the
fingers will curl in the correct general sense of the direction of d B.
Everything we have done up to now has been stated in terms of filamentary currents.
However, as we know, there are many situations that arise in which it is convenient to
describe them in terms of currents distributed throughout a volume or over a surface.
We can easily adapt our results to such cases by using the equivalents for current
elements that we found in (12-10). For example, if the source currents are described by
'ds
dr
Figure 14-1.
induction B.
Force o n a current element d u e t o an
1 4 - 2 S T R A I G H T C U R R E N T O F FINITE L E N G T H
227
dB
'ds' figure 14-2. Geometrical relations between a

current element and the magnetic induction it
produces.
the volume current density J ' ( r ' ) , the results analogous to (14-2) and (14-6) will be
B(r) =
Mo r J ' ( r ' ) X R
/
dr'
R2
477 J V
Mo I X R dr'
d B =
R2
4 77
where the integral is to be taken over the whole volume V
Similarly, (14-3) and (14-5) become
F = / J ( r ) X B ( r ) dr
d = J X B dr
(14-7)
(14-8)
containing currents.
(14-9)
(14-10)
where F is the total force on all of the currents contained within the volume V whose
distribution is described by J(r).
Using (12-10) again, we get corresponding expressions when surface currents are
involved:
u0 r K'(r') X R
B
(r) = ^ /
da'
(14-11)
4 7 7 J'S",
R2
s
dB =
Mo K X R da'
4 77
R2
F = JK(r)
d = KXBda
X B ( r ) da
(14-12)
(14-13)
(14-14)
Finally, if all of the possibilities we have mentioned are simultaneously present, the
total B at a given point will be the sum of all of the various contributions from (14-4),
(14-7), and (14-11), and it will be this resultant B that would be used in (14-3), (14-5),
(14-9), (14-10), (14-13), or (14-14) as would be appropriate.
W e have not made any use of any of our previous classifications of currents that we
discussed in Section 12-2 and indeed it is a basic hypothesis of our subject that the
magnetic induction B is produced by all currents f r o m any source whatsoever, and later
we will find it necessary to introduce currents of types different from any we have yet
encountered.
As in the case of the electric field, an equation like (14-2) can be regarded as a
" r e c i p e " for calculating B once the current distribution is given, and now we want to
consider a few examples of such direct calculations.
14-2 STRAIGHT CURRENT OF FINITE LENGTH
Let us consider a steady filamentary current I' and find the B produced by a finite
length of it. If we choose the origin where the perpendicular from the field point P
228
intersects the current and use cylindrical coordinates, we get the situation shown in
Figure 14-3a. Evidently this is not a complete circuit but a portion of one. Cases like
this are useful, however, because more complicated complete circuits can frequently be
thought of as being composed of parts such as this; then, once we can find B for each
part, the resultant can be found by vector addition, as implied, for example, in (14-4).
We see from the figure that r = pp and r' = z'z so that R = r - r' = pp - z'z,
R2 = p2 + z'2, and ds' = dz'z. Therefore, we find that ds' X R = p dz'z X p = p dz'y
and then, since
is constant, (14-2) becomes
dz'
Mo I'PV
4 77
= <p
Vol'
4 77p
H0I'p$
,22 \ V
-Ly ( p 2 + z' )
L-,
( fp-2 + L\)
a/2
,2\l/2
p2(p2 + z'2)
47T
L,
( P 2 + L\)
1/2
-L,
(14-15)
This result can also be written in terms of the angles a 2 and a 1 whose positive sense is
defined in Figure 14-36 for we see from the figure that
Mo 7 '
B = 
47rp
( s i n a 2 + sino^)
(14-16)
Thus, the induction is always perpendicular to the plane formed by the current and the
position vector of the field point.
We can obtain B for ati infinitely long straight current by letting L2 - oo and
TT/2
and a !
77/2) and the result is that
oo (or, correspondingly, a 2
B
2 77p
<p
(14-17)
We can use this expression to compare this present approach with that of the last
chapter. If we have an infinitely long current parallel to the z axis, which passes
through P, as is the case in Figure 13-2, then the force on a length ds of it can be
found by combining (14-17), (14-5), and (1-76) to give
[i 0/'q>
n0II'dzp
d F = Idz z X
=
2irp
Z7rp
which is exactly what we found by the direct calculation that led to (13-13).
(a)
Figure 14-3.
Calculation of B due t o a straight current of finite length.
c
is
1 4 - 3 AXIAL I N D U C T I O N O F A C I R C U L A R C U R R E N T
229
The magnitude of B as given by (14-17) varies inversely with the distance from the
current. The surfaces of constant B are thus cylinders whose axes coincide with the
current. The radius of a given cylinder is determined from p = nQI'/2TTB. Since B is in
the <p direction, the lines of B in a plane perpendicular to these cylinders, and hence to
/ ' , will be circles. Thus, we get a picture for B due to an infinitely long straight current
like that shown in Figure 14-4; the figure is drawn so that / ' is perpendicular to the
page and coming out of it.
14-3] AXIAL I N D U C T I O N OF A CIRCULAR CURRENT
As another example of a filamentary current, let us consider a current I' circulating

along the circumference of a circle of radius a as shown in Figure 14-5. In order to be
specific, we assume the circle to he in the xy plane with the origin at its center. We
Figure 14-5. Calculation of axial

induction of a circular current.
230
choose our field point along the z axis so that it is on the line normal to the area
enclosed by the current.
We see that r = zz while r' = x'x + y'y = a (cos <p'x + sin <p'y); we are using the
polar angle <p' to locate the source point, but expressing its position vector in terms of
the constant rectangular unit vectors. Hence, R = a cos <p'x a sin <p'y 4- zz so that
R2 = a2 + z2. In addition, ds' = dr' = a dq>'( - sin <p'x + cos p'y) and therefore we
get
ds' X R = ci dtp' [z(cos (p'x + sin<p'y) + az]
with the use of (1-28). Inserting these results into (14-2), we find that
( )
H0I'a r2 [z(cos<p'x + sinqp'y) + az] dtp'

~A
/
7~1
^
4tt -/0
(a2 + zl)
lL0I'a2
^
Z
>,( z2 i z2X3/2 ( 1 4 " 1 8 )
2 (a + z )
where the x and y components have integrated out to zero. Thus, the induction is
completely along the z axis as seems evident f r o m the symmetry of the situation
depicted in Figure 14-5. At the center of the circle (z = 0), the induction becomes
simply
Bcenter =
V-o1'.
T T
2a
(14-19)
At large distances from the circle, that is, for z a, (14-18) approximates to
8(z) =
PoI'd2
,
,
(14-20)
and varies inversely as the cube of the distance z. In this respect, it is similar to the
electric field of a dipole as we saw in (10-35), for example. This similarity is not
accidental, as we will learn in Chapter 19. Now let us consider one of the applications
of our result.
Example
Axial induction of an ideal solenoid. Suppose we have a cylinder of length L and circular
cross section of radius a. If we wind a wire uniformly around it to give us a total of N
" turns," the resultant device is called a solenoid and is shown in Figure 14-6 (only a
I
H
I
OOOOOO
Figure 14-6.
I
h
I
Cross section of an ideal solenoid with its axis along t h e z axis.
14-4
INFINITE PLANE U N I F O R M C U R R E N T SHEET
231
few of the turns are shown in cross section.) If the wires are very thin and wound right
next to each other, we can, to a first approximation, neglect the pitch of the winding,
and we can consider it as equivalent to a set of N circular filamentary currents of radius
a. (This is often referred to as an " i d e a l " solenoid.) The value of B at a point P on the
axis can then be obtained by summing up the contribution of each circular turn as is
given by (14-18). Consider a small portion dz0 of the length located a distance z0 f r o m
one end. If n = N/L
is the number of turns per unit length, then there will be
dN = ndz0 circular rings in this small portion, each approximately the same distance
z = zp z0 from the field point P. Hence, f r o m (14-18), their contribution to the
magnitude of B at P will be
dB=
anI'a2n
dzn
5
+ ( z , - z0) J
(14 21)
"
^rrr<^tv
2[a2
according to (14-4). The total value of B at P will then be

fi0I'a2n naz
dz00
Po*
rL
"
2
fi0iI'nanu
'o 2 [ a 2 + ( z , - z 0 ) 2 ] 3 / 2
\ [a2 + (L - z P f ] 1 / 2
dz'
rL-z
,7 P
("2 + z'2)3/2
(a2 + z2)
where we have let z' = z0 zP and used the value given for the integral in (14-15).
This result can also be expressed very simply in terms of the angles
and a 2 defined
in the figure; we see that we have
B = ^ j u 0 / ' ( c o s a2 + c o s o ^ )
(14-23)
If the solenoid is infinitely long, then both a x and a 2 approach zero and (14-23)
simplifies to
B = n0nl'
(14-24)
and is independent of the location of P.
14-4 INFINITE PLANE UNIFORM CURRENT SHEET

As an example of a continuous current distribution, let us consider an infinite plane
sheet o n which there is a constant surface current density K'. We will find the value of
B at a n arbitrary field point with the use of (14-11). Let us assume the sheet to coincide
with the xy plane and that K' is in the y direction so that K' = K'y where K' const.
As we see from Figure 14-7, r = xx + yy + zz, r' = x'x + y'y, R = (x x')x + (y
y')y + zz, R2 = ( x x')2 + (y y')2 + z 2 , a n d da' = dx' dy'. Therefore, K' X R
= K'y X R = K'[zx + (x' x)z] and (14-11) becomes
[i0K'
4w
=
iinK'
,oo
oo
,oo
[ z x + (x' x)z\
oc J(x x')2
dx' dy'
+ (y y')2
,oo ,oo (zx + X'z) dX'dY'

[
f
477" '-oo^-oo (X'2 + Y'2 + Z 2 )
+ z2]3/2
(14-25)
where X' = x' x and Y' = y' y. We see at once that the z component will vanish
because the integrand is an odd function of X'. T h e n the integral reduces to exactly the
232
da''
f (x, y, z)
Figure 14-7. Calculation of

B d u e t o an infinite plane
uniform current sheet.
same form as that of (3-11), and if we use the results given in (3-12) and (3-13) by
simply replacing A/T 0 by /j.0K' and Z by X, we obtain
B =
(14-26)
Thus, we see that B has a magnitude independent of the location of the field point, is
parallel to the current sheet and at right angles to the direction of the current, and has
opposite signs on the two sides of the sheet. This is summed up in Figure 14-8 showing
the lines of B as drawn in the xz plane so that K' is perpendicular to the plane of the
page and coming out of it.
K ' (out)
9*
^
Figure 14-8. Lines of B due

t o an infinite plane uniform
current sheet perpendicular
t o t h e page.
14-5 M O V I N G P O I N T C H A R G E S
233
14-5| M O V I N G P O I N T CHARGES
If w e w r i t e t h e v o l u m e c u r r e n t d e n s i t y as the product p'v' given by (12-3), then (14-7)
becomes
u0
B(r) = ^
( p ' v ' X R d r '

^
477 J V'
(14-27)
N o w l e t u s a s s u m e t h a t t h e c h a r g e s described by p' are contained within a very small

v o l u m e . T h e n r ' will b e p r a c t i c a l l y the same for all volume elements dT' and we can
t a k e r ' = c o n s t . , t h u s m a k i n g R a n d R 2 constants also. If we assume in addition that
all of t h e c h a r g e s h a v e t h e s a m e velocity v', we can remove all of these constant factors
f r o m o u t of t h e i n t e g r a l a n d (14-27) becomes
B
u0 v' X R ,
ju0 q'\' X R
( r ) = ^ j - / P' dr' - ^
4tt
R
Jy
4TT
R
(14-28)
w h e r e q' is t h e t o t a l c h a r g e . B u t u n d e r these conditions, q' can be regarded as a point

c h a r g e , s o t h a t ( 1 4 - 2 8 ) gives t h e magnetic induction produced by a moving point
charge.
B y c o m p a r i n g ( 1 4 - 2 8 ) w i t h (14-6), we see that this value of B is just the same as that
p r o d u c e d b y a current element
I' ds' = q'y'
(14-29)
In o t h e r w o r d s , w e h a v e f o u n d t h a t a moving point charge is the equivalent of a current
e l e m e n t a n d ( 1 4 - 2 9 ) gives t h e q u a n t i t a t i v e connection. Having obtained this result, we
c a n e a s i l y a d a p t s o m e of o u r previous results. For example, if we combine this
e q u i v a l e n c e w i t h ( 1 4 - 5 ) a n d w r i t e t h e force as F rather than d F, we find that there will
b e a m a g n e t i c f o r c e o n a m o v i n g p o i n t charge given by
F m a g = qy X B
(14-30)
w h e r e B is t h e v a l u e of t h e i n d u c t i o n at the location of the charge. Similarly, we can use

( 1 3 - 1 7 ) t o e x p r e s s t h e m a g n e t i c f o r c e on one point charge q moving with velocity v due
to a n o t h e r p o i n t c h a r g e q' of velocity v ' and we get
iir^qq' v X (v' X R)
F i n a l l y , if w e a d d ( 1 4 - 3 0 ) t o (3-1), which gives the electric force on q, we find the total
electromagnetic force to be
F = q(E + v X B)
(14-32)
T h i s i m p o r t a n t r e s u l t is u s u a l l y called the Lorentz force.
S i n c e all of o u r d i s c u s s i o n of magnetism u p to now has been based on steady
c u r r e n t s , w e c a n e x p e c t t h a t (14-28) a n d (14-31) are applicable for constant velocities,
t h a t is, z e r o o r n e g l i g i b l e acceleration. It does turn out that the fields produced by
a c c e l e r a t e d c h a r g e s a r e d i f f e r e n t f r o m these and lead to phenomena usually described
as " r a d i a t i o n . " I n a d d i t i o n , even if the charges are moving with constant velocity, we
c a n u s e ( 1 4 - 2 8 ) a n d (14-31) o n l y if the velocities involved are "small." Although it is
n o t a t a l l e v i d e n t a t this p o i n t t h a t this should be the case, nor even what is to be
c o n s i d e r e d " s m a l l , " w e c a n a n t i c i p a t e some of the results of Chapter 29 and simply
a s s e r t t h a t w e r e q u i r e t h a t | v | ^ c where c is the speed of light in vacuum. [Although
we h a v e b e e n d e a l i n g o n l y w i t h i n d u c t i o n s that are constant in time, it is worth noting
t h a t t h e v a l u e of B a t t h e fixed p o i n t r as given by (14-28) is actually time dependent
since b o t h R a n d R will c h a n g e as q' moves.]
234
W e have found that the sources of B are currents of any type. Just as we did for E in
Chapters 4 and 5, we would also like to express this information in terms of the
differential source equations of the induction. In other words, we want to evaluate
V X B and V B, and this is what we will obtain, along with other information, in the
next two chapters.
EXERCISES
14-1 Find the magnetic induction produced by
the currents of Figure 13-4 at the point on the x
axis that is midway between them.
14-2 Suppose the field point P of Figure 14-3 is
located at an arbitrary value of z rather than at
z = 0 as shown. Show that the value of B there
can still be written in the form (14-16) where a!
and of2 are measured positive below and above,
respectively, from the perpendicular dropped from
P to the line of
14-3 Two infinitely long straight currents are
parallel to the z axis. One of them, carrying a
current /,, intersects the xy plane at the point
(jCj, ^ ) ; the other, with current / 2 , intersects the
xy plane at (x2, y2)- Find the resultant B produced by them at any field point (x, y, z).
14-4 A square of edge a lies in the xy plane
with the origin at its center. Find the value of the
magnetic induction at any point on the z axis
when a current / ' circulates around the square.
Show that your result gives the value 2/2 p.0I'/ira
for the induction at the center.
14-5 An ideal solenoid of length L and N turns
is wound with a square cross section, that is, each
turn is a square of edge a. Find the induction
produced at the center of the solenoid when there
is a current / ' in the windings. What does this
become for an infinitely long solenoid?
14-6 Two circular loops, each of radius a, have
planes parallel to the xy plane and centers on the
z axis a distance d apart. Each carries the same
current / ' circulating in the same sense. Choose
the origin midway between them and find the
axial field B.(z). What is Bz at the origin? Show
that dB./dz vanishes when evaluated at the midway point. Show that the second derivative of B,
will also vanish there provided that d a. What
is B,(0) under these conditions? Show that
d3B,/dz3 also vanishes at the origin. An arrangement like this with d = a is called a Helmholtz
coil and is used to produce an approximately
constant induction over a small region.
14-7 The current shown in Figure 14-9 follows
along the arc of a circle lying in the xy plane with
Hgure 14-9.
The current of Exercise 14-7.
center of curvature at the origin. Find B at any

point on the z axis. As a check on your result,
show that it reduces to (14-18) under appropriate
conditions.
14-8 A wire is wound in a helix of pitch angle a
on the surface of a cylinder of radius a so that N
complete turns are formed. If the wire carries a
current / ' , show that the axial component of the
induction produced at the center of the helix is
L(n0NI'/a)(
1 + 772N2 tan2 a )

with the xy plane. Its surface current density is
K' = K'y where K' = const. Another infinite
plane current sheet is parallel to the xy plane and
intersects the positive z axis at z = d. The second
current density is K' = K'y. Find B everywhere.
14-10 Repeat the preceding exercise with the
surface current of the second sheet now given by
K' = K'y.
14-11 A circular dielectric disc of radius a has a
uniform surface charge density a on it. It is
rotated with constant angular speed u about an
f
EXERCISES
axis that is normal to the surface of the disc and

passes through its center. Assume that the charge
distribution is not altered by the rotation and find
B at an arbitrary point on the axis of rotation.
What is B at the center of the disc?
14-12 A sphere of radius a contains a total
charge Q distributed uniformly throughout its
volume. It is set into rotation with constant angular speed co about a diameter. Assume that the
charge distribution is not altered by the rotation
and find B at the center of the sphere.
14-13 A dielectric sphere of radius a is uniformly polarized. It is rotated with constant angular speed u about the diameter parallel to the
polarization. Assume that the polarization is not
affected by the rotation and find B at the point
where the axis of rotation intersects the surface of
the sphere, that is, at the "north pole" of the
sphere. What is B at the center of the sphere?
235
14-14 An infinitely long cylinder of circular cross

section of radius a carries a current I' that is
uniformly distributed over the cross-sectional area.
The axis of the cylinder coincides with the z axis
and / ' is in the positive z direction. Choose a
field point on the x axis and find B for all values
of x, both inside and outside of the cylinder.
14-15 In the circuit shown in Figure 14-10, the
curved lines are semicircles with common center
C. The straight portions are horizontal. At this
instant a point charge q located at C has a
velocity v directed vertically downward. Find the
magnetic force on q.
14-16 Figure 14-11 shows a short thick solenoid.
The N turns of the winding uniformly occupy the
region between the two coaxial cylinders of radii
a and b. Neglect the pitch of the windings and
show that the axial component of the induction
Figure 14-10.
The circuit of Exercise 14-15.
-21-
Figure 14-11.
14-16.
The solenoid of Exercise
236
(14-7) for B can be transformed into the form
produced by a current / ' is given by
u0 r V ' X J ' ( r ' )

B(r) = ~ f
dr'
flgNI'
/J
I
f 4 + [ ^ + (/ + z ) ] ' |
X (I + z) In I
\
^17?
\ a + [a2 + (/+ z) ] /
+ ( /
2 ) l n
[
a + [a2 + (Î - z ) J
J^
14-17 Show that if all of the distributed source

currents occupy a finite volume V , the expression
This result emphasizes the role of the nonuniformities of the source currents in producing the
induction. We will not generally find it convenient to use this form.
14-18 A closed filamentary circuit C carrying a
steady current I is in a region where B is constant. Show that the total force on C is zero. Now
suppose that C is a plane circuit and the plane
surface is parallel to the direction of B. Show that
in this case there will be a nonzero torque on C
that will be parallel to the plane of C.
15
THE INTEGRAL FORM
O F AMPERE'S LAW
The first differential source equation that we want to consider is v X B. The general
definition of the curl of a vector given by (1-73) suggests that we consider the line
integral of B about some closed path.
115-11 DERIVATION OF THE INTEGRAL FORM

We will show that
(^B ds = ii0Ienc
(15-1)
where the integral is taken about an arbitrary closed path C and 7 enc is the total current
passing through the area enclosed by the curve C. The path C can be any closed curve
and need not coincide with any real circuit. The expression (15-1) is known as the
integral form of Ampere's law, it is also very commonly referred to as Ampere's
circuital law.
For simplicity, we initially assume that B is produced by a single filamentary circuit
C' carrying a current I' so that B is given by (14-2). Inserting this expression into the
left-hand side of (15-1), we find that
jTo
fo-f' I
ds
'
( r f s ' X R)
Mo 7 ' J
(~ds
X ds') R
*
(I")
where we used (1-29) to interchange the dot and the cross in the integrand. If we now
recall f r o m Section 1-7 that the magnitude of the cross product equals the area of the
parallelogram with the two vectors as sides and then, in addition, recall (4-3), we see
that we should analyze (15-2) in terms of solid angles.
The general situation is illustrated in Figure 15-1. Suppose we consider a given point
P on the path of integration C. The source circuit C' will subtend some total solid
angle 2 at that point. Now in carrying out the integration over C, we give the point P a
series of successive displacements, one of which is ds. After P has been displaced by
ds, the source circuit C' will generally have a different aspect as viewed from P, so that
the solid angle which C' subtends at the new position of P will have changed to a new
value 2' = 0 + dti. Thus, dSl = change in solid angle subtended by C' at P as a result
of the displacement of P by ds. But we can also obtain the same relative change by
imagining P held fixed and giving an equal and opposite displacement ( ds) to every
point of C. (You can easily convince yourself of this by holding this book at a
reasonable distance from your eyes and then comparing the visual result of: (1) keeping
the b o o k held fixed and moving your head back, and (2) keeping your head fixed and
moving the book an equal amount away from your head.) T h u s we can also say that
dSl = change in solid angle produced by keeping P fixed and displacing every point of
C' by ds; the new position and orientation of C are also shown in the figure. N o w
we see that ds X ds' = da = shaded area with sides ds and ds', so that, according
237
238
THE INTEGRAL F O R M O F AMPERE'S LAW
to (4-3), the term under the integrals of (15-2) is just da R / R 2 = solid angle
subtended at P by da. change in solid angle subtended at P resulting from the
displacement of ds' by ds. Therefore, when we carry out the integration over C' in
(15-2), we are summing up all of the contributions from all of the d s ' of C ' , so that we
have
f
( ds X ds') R
<pa~
<15"3)
where, again, dSl is the change in solid angle observed at P due to a displacement of P
by ds, but it is calculated in (15-3) by an equivalent summation over C' based on
holding P fixed and displacing C'. Inserting (15-3) into (15-2) and carrying out the
integration over C, we get
r
V-nl' f
Un/'
(b rfs = ( dL = - - f Afi
JC
Am Jc
47T
(15-4)
where AS2 is the total change in the solid angle subtended by C ' at the various points
on C when summed over the closed path C. It turns out that there are two cases to be
considered.
15-1
D E R I V A T I O N O F THE INTEGRAL F O R M
239
1. The path C does not link the circuit C'. In this case, the relative orientations are
like those shown in Figure 15-2. Here, if we start at P, then when we come back to P
after completing the loop C, the final solid angle has the same value that it had initially,
so that AS2 = 0 and (15-4) becomes
d)B ds = 0
c
(15-5)
2. The path C does link the circuit C ' . In this case, the path of integration encloses
the source current and we have a situation like that shown in Figure 15-3. It is easier to
see what the change will be if we pick our initial point A just above the surface S'
enclosed by C' and the final point A' just below the surface S' and then go to the limit
as A' and A become coincident. First, we note that the direction of the normal n' to the
surface enclosed by C' is determined from the direction of / ' by the standard
right-hand rule shown in Figure 1-24.
(a) At the initial point A. We see from Figure 15-4a that the angle between da' and
R is
= 90 S where 5 is very small and positive. Then the solid angle subtended
by da' at A as given by (4-3) will be da' cos Â/R2
= da' cos (90 S)/R2 and is
positive. Now as A is taken closer to the surface S', 8 - 0, all of the contributions to
the solid angle subtended by C ' at
are positive, and, since A then "sees" one half of
all space, we get
^for A =
înitial
= +27r
(15-6)
(In order to convince yourself of this, imagine your eye located at A just above this
page, and then let your eye approach near coincidence with the page.)
(b) At the final point A'. As we see from Figure 15-4b, the angle between da' and R
is * A . = 90 + e where e is positive and very small. Then the solid angle subtended by
240
<b)
Figure 15-4.
Situation near (a) the initial point A and ( b ) the final point A'.
da' at A' will be da' cos (90 + t)/R2

and will be negative. Then, as A' nears the
surface, c * 0, all of the contributions to the solid angle will be negative, and since A'
will " see" one half of all space, we get
^ f o r A'
^ final
~ 277
(15-7)
Combining (15-7) and (15-6), we find that the total change in solid angle for this
case will be
=
Qfinal ~
înitial =
( ~ 2 t t )
(2ff) =
~4w
(15-8)

$B-ds
= [i0I'
(15-9)
and equals /x0 times the current passing through the surface enclosed by the curve C.
Comparing (15-5) and (15-9), we see that the value of this line integral of B is different
from zero only if the path of integration encloses a current.
We also see that if the sense of integration around C were reversed, then A and A'
would be interchanged as initial and final points, AS2 would equal +477, and the value
of the integral would be n0I', that is, the current I' would be counted as negative.
We can summarize these sign results by means of Figure 15-5. Once the sense of
integration about C has been chosen, this determines the positive sense of its normal ft,
as in Figure 1-24. Then, if a current / ' passes through the surface enclosed by C in this
same sense, as in (a) of the figure, it will make a positive contribution n0I' to the
integral, while if it passes through C in the opposite sense, as in (b), it contributes
/Xq/' to the integral. (Note that Figure 15-5(2 corresponds to Figure 15-3.)
If there is more than one source current so that B will be given by an expression like
(14-4), then each current /, will contribute either 0 or /x 0 A depending on whether /,
/'(negative)
/'(positive)
(b)
Figure 15-5. Sign conventions for current as related

t o sense of integration
about C.
15-1
D E R I V A T I O N O F THE INTEGRAL F O R M
241
passes through C or not, as well as upon its sense of doing so, and we will get
C
/1 0 /, = Moênc
(15-10)
enclosed
where 7 enc is the net current enclosed by the path of integration. Thus we have obtained
the integral form of Ampere's law that we initially stated in (15-1). [We note that this
result has a certain "similarity" to Gauss' law (4-1).] There are some general implications of this result that we can point out. Any currents not enclosed by the path of
integration do not contribute to the value of the integral, although they certainly can
affect the value of B at a particular point. Also, the location of the enclosed currents
does not affect the value of the integral, although if they are moved about within C
they can affect the value of B at particular points on the path of integration.
We can now put (15-10) into another useful form by expressing 7 enc in terms of the
current density J by means of (12-6). (We will drop the prime on the source current
density now for simplicity of notation.) If we do this, and use Stokes' theorem (1-67),
we obtain
(/)B ds = ju,0J f J da = J f (v X B) da
'c
s
s
(15-11)
where S is the open surface enclosed by C. Since C is arbitrary, (15-11) also holds for
any very small area, and we can equate the integrands to get our desired source
equation
V X B = /x 0 J
(15-12)
This fundamental result is equivalent to Ampere's law of force between complete

circuits. We see at once that, in contrast to the electrostatic field, the magnetic
induction is not a conservative field since V X B i s not always zero.
Furthermore, we can use (15-12) to find the boundary conditions satisfied by the
tangential components of B at a surface of discontinuity. If we insert (15-12) into
(9-13), we obtain
ft X (B 2 B j ) = lim (fi0hJ)
h0
(15-13)
In order to interpret this result, we consider the small area element in the transition
layer that is perpendicular to the flow as shown in Figure 15-6 (Compare to Figure 9-5.)
The total current AI through this area as given by (12-6) is AI = Jh A s. As the
transition layer is shrunk to zero thickness, so that h -* 0, this total current will be
squeezed into a surface current of density K and the constant total can be written as
AI = K As, as seen from Figure 12-5 a. Thus, we see by comparing the two expressions
for A I that hJ -> K, and since the two vectors are in the same direction, we get
K = lim (hJ)
h~>0
(15-14)
ftX (B 2 - B x ) = n 0 K
(15-15)
and (15-13) becomes
As
Th
Figure 15-6.
Small area element in the transition layer.
242
which is the required boundary condition. As in (9-18), we can use the alternative form
written directly in terms of the tangential components:
B 2 , - B 1 ( = fi 0 K X
ft
(15-16)
where the relation between the normal and tangential vectors is that illustrated in
Figure 9-5.
15-2 SOME APPLICATIONS OF THE INTEGRAL FORM

M u c h as we were able to do with Gauss' law in Section 4-2, we can use (15-1) to
calculate B fields if the problem has sufficient symmetry. The principal problem is that
of choosing a suitable closed path for integration. The main things to look for are
curves on which B has a constant magnitude, and curves for which B is either parallel
to or perpendicular to the sense of traversal of the path, both for ease in integration
and for avoiding difficulties with an unknown dependence of B on position. We
illustrate this process by considering a few standard examples.
Example
Infinitely long straight current. Let us assume the current I to be uniformly distributed
over the circular cross section of radius a of an infinitely long cylinder as shown in
Figure 15-7. If we review the general dependence of the direction of B due to a current
as shown in Figure 14-2 and consider the general " s y m m e t r y " of this problem, we
conclude that B lies in the plane perpendicular to the direction of I, is everywhere
tangent to the dashed circle of radius p shown, and its magnitude can depend on p at
most and must be independent of z and <p. In other words, B has the general form
B = By(p)$. These considerations are just as valid whether the field point is inside the
cylinder or is outside of it as happens to be shown in the figure. Thus, for any value of
I
Figure 15-7. Portion of a long
straight current.
15-2 S O M E A P P L I C A T I O N S O F THE INTEGRAL F O R M
243
p, let us choose as our path of integration a circle of radius p and go around it in the
same sense as the assumed direction of B. Then we will have ds = p dtp a, and C clearly encloses the total current I; thus
7 enc = I and (15-18) becomes
B
<P(P) = ^
(P > a)
(15-19)
By comparing this result with (14-17), a n d recalling that we have dropped the prime on
the source current, we see that the induction outside an infinitely long straight current
is the same as if the current were a filamentary one going along the axis of the cylinder.
2. Inside the cylinder. Here p < a, and now C does not enclose all of the current but
only that fraction equal to the circular area enclosed by C divided by the total
cross-sectional area of the cylinder. Thus / e n c / / = 7ip2/Tra2 so that 7 enc = I(p2/a2)
and (15-18) becomes
b
<p(P) =
^
2ira
(P < a )
(15-20)
[Both of these results (15-19) and (15-20) agree with those found for Exercise 14-14, but
have been obtained much more easily here.]
W e also note that both of our expressions give the same value Bv(a) = fi 0 / / 2 it a at
the surface of the cylinder so that B v is continuous across it. This is in agreement with
(15-16) since B has only tangential components in this case and there is no surface
current K on the cylinder. The dependence of B v on p is illustrated in Figure 15-8.
Example
Infinite uniform plane current sheet. In Figure 15-9 we show an edge-on view of the
infinite plane sheet with K out of the page. W e assume that | K | = K = const.; this
current distribution could be approximated by m a n y closely packed wires all parallel
244
THE I N T E G R A L F O R M O F AMPERE'S LAW
~ ~ l
I
J
*>\
I1
K (out)
Figure 15-9. Path of integration for

finding B due to an infinite uniform plane
current sheet.
and all carrying the same current. Since there is no preference for angles above or
below the horizontal, nor in or out of the page, we conclude, after looking at Figure
14-2 again, that B is perpendicular to K, is parallel to the plane of the sheet, and is
oppositely directed on the two sides of the sheet as shown. B may still depend on the
normal distance D from the sheet, however. Accordingly, we choose the rectangular
path of integration shown dashed, with two horizontal sides each of length / and each
the same distance D from the sheet; these are connected by two vertical sides of length
2D. O n the horizontal sides, B is parallel to ds, so that B ds = B(D) ds and B(D) is
constant. O n the vertical sides, B is perpendicular to ds by construction so that
B d s = 0 and there is no contribution to the integral f r o m these sides; this is an
advantage since the form of B(D) is unknown. Since | K | is current per unit length, we
have 7 enc = Kl for this path and (15-1) becomes
<j)B ds = 2Bl = iu0Ienc = fx0Kl
so that
B = \ti0K
(15-21)
which agrees exactly with what we found in (14-26) by direct integration. Thus we have
found once again that the magnitude of B is independent of distance f r o m the current
sheet.
This result is also in agreement with the boundary condition (15-16) for we see f r o m
Figure 15-10 that it becomes B 2 t B l r = B\ ( Bi) = 2Bi = ^ 0 K X n = n0K% so
that B = jn0K as above.
Example
Infinitely long ideal solenoid. We assume that the turns of the cylindrical coil are so
closely wound and that the wires are so thin that the pitch of the windings can be
neglected. Then, in effect, we have a current sheet of surface density K circulating
around the cylinder as indicated in Figure 15-11. If there are n turns per unit length
and I is the current in the windings, the current per unit length will be nl so that
K=nl
(15-22)
If, in addition, we assume that the solenoid is infinitely long, we can conclude that B
> ^
/*N
n
K(out)
Figure 15-10. Geometry used to verify the

boundary conditions on B.
/C(out)
245
/f(in)
'1
I
I
jv
/ ^
Figure 15-11. Path of integration for

finding B due to an infinitely long ideal
solenoid.
will be parallel to the axis of the cylinder and with the sense shown. In the figure, B, is
the induction inside the solenoid, while B 0 is the value outside; both must be
independent of z since the solenoid is infinitely long, making one value of z as good as
another. Again we choose a rectangular path of integration C shown dashed; the
vertical sides are of length I and the sense of integration is parallel to B, and
antiparallel to B c . The vertical side inside is a distance d from the current sheet and the
outside vertical side is a distance D from it. The horizontal sides are of length d + D;
since J s is perpendicular to B on them, B ds = 0 and they will not contribute to the
integral. Furthermore, 7 enc = Kl = nil, then (15-1) becomes
(^)B ds = Btl BJ = ja0Kl = [iQnIl
so that
Bt -Ba
= ti0K = w l
(15-23)
We see that this result is independent of both d and D\ therefore, (15-23) is also in
agreement with (15-16) as well if we choose region 1 to be outside of and region 2 to be
inside of the solenoid.
Since (15-23) is independent of D, it will also be true as D -> oo. But then, if we
imagine looking back at the solenoid from a great distance away, we will see the
oppositely directed currents essentially superimposed and Ampere's experimental result
tells us that they will produce a negligible effect. Thus, for D -> oo, B0 0 and (15-23)
gives for this case:
Bi = V o K = Mo" 7
(15"24)
If we substitute this back into (15-23), which is possible since it holds for any D, we
now find that
Bc = 0
(everywhere)
(15-25)
Furthermore, since the results (15-23) and (15-24) are also independent of d, we can
then conclude that
B, = HqKz, = [x0nlz
(everywhere)
(15-26)
246
In ot,her words, for an infinitely long ideal solenoid, the induction is confined entirely to
the interior and is uniform across the cross section; it is also independent of a, the
radius of the cylinder. [Our result (15-24) agrees with our previous one given in (14-24);
in the latter case, however, we found this value only on the axis, but now we see that it
is true for all interior points.]
Example
Toroidal coil. Here we have a current / in a conductor that is wound uniformly, that is,
with constant pitch, around a torus as in Figure 15-12 where only a few turns are
shown. The cross section could be circular or rectangular, for example. (Such an
arrangement can also be imagined to b e obtained f r o m taking a solenoid of finite
length, and bending it into the shape of a doughnut, thus joining its ends.) We cannot
learn everything about B from (15-1) for this case, so we will consider only circular
paths of radius p which are centered at the center of the torus and are in the plane
which includes the central axis of the torus. For any such circular paths, ds = p dcp <p
and (15-1) becomes
(b ds = <^B ypd<p = By - 2ttp = Moênc
so that
(15-27)
Figure 15-12.
produces.
Toroidal coil and paths of integration used to calculate the induction it
247
since, from the symmetry, B will depend only on p and hence will be constant on such
a path.
Now for a path inside the circle defined by the inner boundary of the torus, such as
C\, I enc = 0 a n d By = 0. For a path within the torus, such as C 2 , / e n c = NI where N is
the total number of turns and (15-27) gives
fx0NI
(15
"28)
Finally, for a path completely outside the torus, such as C 3 , we find that / e n c = NI
NI = 0, according to Figure 15-5, since for every current I out of the page there is
necessarily an equal one going into the page. Therefore, B = 0 only within the torus
itself.
If the dimensions of the cross section of the torus are small compared to its central
radius, then p = const, for all paths C 2 and B^ is approximately constant across the
cross section.
These calculations do not tell us anything about Bp and Bz; they have to be found
by a direct calculation from the Biot-Savart law (14-2). We will not do this, but content
ourselves with the observation that they turn out to be of the order of magnitude of
B y / N ; thus they can be made negligible by using very many turns, that is, by using
closely packed windings. We can, however, at least make these comments plausible by
considering the path of integration C 4 shown in Figure 15-13. This is a path completely
surrounding the torus and lying in the plane perpendicular to its axis, that is, in the
plane <p = const.; the trace of C 4 is also shown in Figure 15-12. For any finite pitch
angle, corresponding to a finite number of turns N, we see that exactly one turn will
pass through the plane enclosed by C 4 ; thus 7 enc = I. The value of ds as given by (1-82)
with dtp = 0 is ds = dp p + dz z, so that (15-1) becomes
j) B ds = j) (Bp dp + Bz dz) = p0I
Q
(15-29)
C4
Since this is different from zero, we see that Bp or Bz or both must be different from
zero on at least some portions of C 4 . Furthermore, the value of this integral is N times
smaller than that for B , as we see from (15-28), so that for comparable dimensions of
the paths, the magnitude of Bp a n d / o r Bz will be roughly of the order of B^/N as we
stated above.
/
Figure 15-13. Path of integration in a plane perpendicular
t o the axis of the torus.
248
3 1 DIRECT CALCULATION OF V X B
Although we have already
somewhat indirectly. It is
operating directly on our
expression (14-7) in terms
this expression, we get
found the curl of B to be given by (15-12), we obtained it

of interest to see that we can also get this same result by
basic defining equation for B. It is convenient to use the
of a continuous distribution of current. Taking the curl of
J'(r) X R
Mo
f V x
dr' (15-30)
47T v
R1
R
where we were able to go from the second to the third term because the limits of
integration depend on source point coordinates ( x y ' , z') while V involves derivatives
with respect to the field point coordinates ( x , y, z). If we now use (1-119), we find that
the quantity under the integral becomes
V X B(r) = V X
f
4tt J y
R
J ' ( r ' ) V 7
R1
J'(r') X R dr'
1
R
^2 |[V- J'(r')l +
R
R^
V J'(r') - [J'(r') V ] |
R
" [J'(r')- V ] | - ^
R
J'(r')
R
~R*
(15-31)
where the second and third terms are zero because any component of v operating on
J ' gives zero since J' depends only on the components of r'. Now that we have (15-31),
we can use (1-132) to change from v to v ' in the second term of the right-hand side in
order to get an integrand that involves the primed coordinates in such a way that we
can perform the integration; we leave the first term unchanged. Doing this, and
inserting the result in (15-30), we find that
R
R
rfT +
f J'(r')
'
(15-32)
~R}
47T J V'
R*
We can show that the second integral is zero by rewriting it with the use of (1-129):
V XB =
/>' v , ] (f)
j,)dr'+i[j->Yr
(i5 33)
Since we are considering only steady currents, V ' J ' = 0 by (12-15) and the volume
integral vanishes. Now the source current distribution occupies a finite volume;
consequently, the bounding surface S' can always be made large enough so that
J'(r') = 0 on all points of S' and the surface integral also vanishes. Thus, as we said,
the integral in (15-33) is zero, and (15-32) reduces to
R
Mo
- /J J ' ( r ' )
d t'
(15-34)
R2
IT V'
The x component of the right-hand side of this equation has exactly the same form as
(4-23), with ix0J; replacing p/e0; the result of integrating (4-23) was found to be (4-26)
so that we have
V X B =
(15-35)
dr' = Mo/x'O")
F"
477 IJ V'
with similar expressions for the y and z components of (15-34). Adding these results,
and finally dropping the prime on J ' in order to agree with the standard notational
convention, we get
V X B = n0[jx(r)x
+ /^(r)y + Jz{r)z] = [x0J(r) = ( v X B) a t r
which is exactly what we found before as given by (15-12).
EXERCISES
249
EXERCISES
with its center at the origin. Find the solid angle
A subtended by this circle at a point on the
positive z axis.
15-2 Consider the induction B produced by an
infinitely long straight current I. Choose a convenient closed path C in the plane perpendicular
to the current and which does not enclose the
current. Show by direct integration that (15-5) is
correct in this case.
15-3 Apply (15-1) to a circle of radius p < a
whose plane is perpendicular to the z axis of the
solenoid of Figure 15-11 and thus show that
Bv = 0 in the interior.
15-4 Apply (15-1) to a circle of radius p > a
whose plane is perpendicular to the z axis of the
solenoid of Figure 15-11 and show that, if the
windings have a finite pitch, the average value of
By on the circle equals fi()I/2irp.
15-5 Find another current distribution that will
produce the same B^ in a torus as that which we
found in (15-28).
15-6 A toroidal coil of N turns is wound on a
circular cross section of radius b. If the radius of
the central axis is a, that is, this is the distance
from the center 0 to the center of the cross
section, find the ratio b/a necessary in order that
the total deviation in B across the cross section
be not more than 2 percent of the value at the
center.
15-7 Consider the infinitely long coaxial cylindrical conductors shown in Figure 6-12. The inner
conductor carries a total current I in the z direction, while the outer conductor carries a current /
in the z direction. Assume the currents to be
uniformly distributed over their respective cross
sections. Find B everywhere and plot your results
as a function of p.
15-8 A certain B field is given in cylindrical
coordinates by B = 0 for 0 < p < a, B =
(jtt 0 //27rp)[(p 2 - a2)/(b2 - a2)]y for a < p <
b, and B = (p,0I/2irp)(p for b < p. Find the
current density J everywhere. How could you

produce such a B?
15-9 Consider a very long cylindrical beam of
charged particles. The beam has a circular cross
section of radius a, a uniform charge density pch,
and the particles have the same constant velocity
v. Find B inside and outside of the beam and
express your result in terms of these given quantities.
15-10 Two infinitely long coaxial cylindrical
surfaces have the z axis as their common axis.
The inner surface of radius a carries a surface
current Kj = Aj<p, and the outer surface of radius
b carries a surface current K 2 = AT2<p. Both Kl
and K2 are constant. Find B everywhere.
15-11 In Figure 15-14, we show a portion of a
current-free region in which B is uniform. Show
that such an induction cannot drop abruptly to
zero, as is indicated schematically by the location
of the arrows, by applying (15-1) to the dashed
rectangular path shown. Show qualitatively that a
change in B of the general nature shown in Figure
6-10 will be compatible with (15-1).
B *
Figure 15-14. Hypothetical lines of B for

Exercise 15-11.
15-12 Show that (15-12) can be obtained by a

direct calculation of the curl of the expression for
B given in Exercise 14-17.
16
THE VECTOR POTENTIAL
The remaining differential source equation of the induction that we need is v B. This
is much more easily obtained than was V X B. After we have f o u n d it, we will find that
we can express much the same information as before in terms of a new vector field that
we will be able to introduce.
-1 THE DIVERGENCE OF B
The direct approach is quite easy in this case. If we use our definition of B as stated in
terms of a filamentary current in (14-2), we find that
Mo
V B = v
I'ds'
X R
2
~R
477 ->c
M 7 ' r
(
~J~
4 7TP J r ' '
ds' X R
R2
(16-1)
since once again the limits of integration depend at most on r' while V operates on the
components of r. We can use (1-116) to rewrite the integrand as
ds' X
R
R
= 1^2
"(VXrfs')
ds'
R
v
(16-2)
Now ds' is a function only of the components of r', and hence is a constant as far as v
is concerned; thus V X ds' = 0. Also, (1-147) gives V X ( R / R 2 ) = 0; therefore, (16-2)
is identically zero and (16-1) becomes exactly
V B = 0
(16-3)
It is clear that this is true in the case of more than one filamentary source when B
would be given by (14-4); the case of distributed currents will be left as an exercise.
N o w that we have (16-3), we can immediately obtain the b o u n d a r y condition
satisfied by the normal components of B at a surface of discontinuity and we find from
(9-6) a n d (9-7) that
n (B 2
B x ) B2n
5i = 0
(16-4)
so that the normal components of B are always continuous.

This important result V B = 0 is one of Maxwell's equations. We will find that we
will have no occasion to change it, even when we consider nonstatic fields, so that
(16-4) as well will be generally correct. We can get an interpretation of (16-3) by
comparing it with the analogous equation for the electric field as given by (4-10),
namely V E = p / e 0 . Here p is the net charge density and arises because electric
charge exists in individual units of opposite sign so that it is possible to have p = 0 by
having an excess of one sign or another. But we see f r o m (16-3) that such a situation
cannot arise in the magnetic case and therefore there cannot be individual units of
magnetic charge analogous to electric charge. Such magnetic charges are called magnetic monopoles. Although their existence has been shown to be compatible with the
requirements of quantum mechanics, every experimental search that has been made for
them has been unsuccessful. Consequently, until any such time that magnetic monopoles have been unambiguously found to actually exist, we must continue to write
V B = 0 and remember that the sources of B are only and always currents.
250
16-2
D E F I N I T I O N A N D PROPERTIES O F THE V E C T O R P O T E N T I A L
251
If we combine (16-3) with the divergence theorem (1-59), we find that

<f>B da = 0
(16-5)
which says that the flux of B through any closed surface is always zero. If we express
this in terms of the lines of B, (16-5) says that there are no net lines of B through any
closed surface, or, that there are always as m a n y lines leaving the surface as entering it;
hence the general situation will always be like that shown in Figure 16-1. This is in
contrast to the electric case, where lines of E can begin or end on charges, as shown, for
example, in Figure 8-7.
Nevertheless, we will find it very useful to consider the surface integral of B over a
surface that is not closed. If we let 3> be the magnetic flux through a surface S, then by
definition, we shall have
$ = Jf B da
s
(16-6)
and this can be different from zero. Since B is measured in webers/(meter) 2 , the unit of
flux will be a weber.
DEFINITION AND PROPERTIES OF THE VECTOR POTENTIAL

If we compare (16-3) with the general vector theorem (1-49) that says that the
divergence of a curl is always zero, we are led to suspect that we should be able to write
B(r) = v X A(r)
(16-7)
The vector field A(r) introduced in this way is called the vector potential. We will see
that in many ways it is a kind of analogue to the scalar potential ^ We find from (16-7)
that the unit of A will be 1 w e b e r / m e t e r = 1 volt-second/meter.
If (16-7) is to be useful, it should be possible to find an explicit way of calculating A
in terms of a given distribution of source currents. We do this by the direct approach of
showing that our defining equation for B can actually be written in this form. If we use
(1-141) and (1-118), we find that
ds' X R
/1\
/ ds' \
^2 = - r f s ' X v | ) = V X | )
R
\R j
\ R }
( v X ds')
R
(ds'\
= v X ( ) (16-8)
\ R
since v X ds' = 0 as before. Inserting this into (14-2), we get
when we use the fact that v operates only on the components of r. We see that (16-9)
252
THE V E C T O R P O T E N T I A L
does have the form of (16-7) and by comparison we can define A for this filamentary
current as
A(r) -
un
f I'ds'
477 *C' R
(16-10)
If there is more than one filamentary source current, then we can apply (16-8) to each
term of (14-4) and the resultant A will be given by
A
- E ^ ^
. 4 77 JC, Rt
(16-11)
where R, = r r,. Thus, we have succeeded in showing that B can be written in the
form (16-7) and, in addition, we have simultaneously found how to calculate A. We
note that the contribution of each current element /, ds, to A is in the direction of the
element itself.
If the currents are distributed, rather than filamentary, we can use our equivalents
(12-10) to adapt (16-10), and we find that the vector potential produced by volume and
surface currents will be given, respectively, by
un
J(r') dr'
H0
K(r') da'
while that produced by a moving point charge would be obtained by our combining
(14-29) and the integrand of (16-10) to give
, .
Mo <7V
v
( ,
A(r) =
(16-14)
477 R
[As in Section 14-5, we would use (16-14) only for the case jvj <a: c.]
The general idea behind all of this is that we would use the above expressions to find
A and then find B by differentiation according to (16-7), so that the procedure would
be, in this sense, analogous to that we have used in electrostatics.
It is of interest to find the divergence of A. Dotting v into (16-10), we obtain
and, if we use (1-115), (1-141), and (1-16), we find that

^-J
= ds' v | i j + ~~(V ds') =
ds'
(16-16)
since V ds' = 0 for the usual reason. If we insert (16-16) into (16-15), and use (1-67),
we get
V A =
4i
But the integrand is always zero by (1-48) and therefore

V A = 0
V'XV'i^
d a'
(16-17)
While (16-17) was obtained for only a single filamentary current, we see that this will
still hold for more than one source current when A would be given by (16-11).
Although we have an integral form (16-12) that enables us to calculate A once the
current distribution is known, it is useful to obtain the differential equation satisfied by
16-2
D E F I N I T I O N A N D PROPERTIES O F THE V E C T O R P O T E N T I A L
253
A in case the information is given in other forms, as we found to occur so often for 4>,
as for example in Chapter 11. If we take the curl of both sides of (16-7), use (1-120),
(16-17), and (15-12), we get
VXB = VX(VXA) = V(V-A)-V
A = V 2A = jx 0 J
so that
V2A=ju0J
(16-18)
which, in rectangular coordinates, is

V2AX
= -FI0JX
V 2AY FI0JY
V 2AZ = [X0JZ
(16-19)
By comparing this with (11-1), we see that each rectangular component of A satisfies
Poisson's equation.
If we now change our point of view somewhat, we can regard (16-7) and (16-17) as
being the two differential source equations of A in the sense of the Helmholtz theorem
of Section 1-20. But then we can immediately find the boundary conditions satisfied by
the components of A at a surface of discontinuity in properties. We find f r o m (16-17)
and (9-6) that
ft" ( A 2 A x ) = A2N ALN = 0
(16-20)
so that the normal components are continuous. If we now insert (16-7) into (9-13), we
obtain
NX (A 2 - A x ) = lim (/zB) = 0
h > 0
(16-21)
since, whatever else B may do in the transition layer, we certainly expect B to remain
finite as the thickness of the transition layer is reduced to zero, thus making /zB > 0 as
h -> 0. Thus, the tangential components of A are also continuous. But when all of the
components of a vector are continuous, the vector itself is continuous across the
surface, and therefore we conclude that
A 2 = AJ
(16-22)
in complete analogy to the continuity of the scalar potential $ as expressed in (9-29).

The magnetic flux $ can also be expressed in terms of the vector potential. If we put
(16-7) into the definition (16-6), and use Stokes' theorem (1-67), we find that
$ = j ( v X A) da = (^)A ds
(16-23)
showing that the flux can be written as the line integral of A about the curve bounding
the surface for which $ is desired. Sometimes (16-23) offers an alternative way of
calculating A if O can be easily found from a previous solution and if the problem has
enough symmetry so that a convenient path C can be devised.
Quantities that are called "potentials" in physics generally have an ambiguity
associated with their absolute value because they are ordinarily introduced as functions
from which other functions of interest can be obtained by differentiation. We saw this
in the case of the scalar potential that was undetermined up to a scalar additive
constant as we discussed in connection with (5-10). A similar, but more complex,
situation also holds for the vector potential. If we review how we went f r o m (16-9) to
(16-10) with the use of (16-7), and then remember the general vector result (1-48) that
the curl of a gradient of a scalar is always zero, we see that we could add the gradient of
an arbitrary scalar to (16-10) and still get the same induction. In order to make this
explicit, let us assume that A is a "suitable" vector potential, that is, it gives the correct
value of B when we use it in B = V X A. N o w suppose that A1" is another function
254
THE V E C T O R POTENTIAL
given by
A + (r) = A ( r ) + V x ( r )
(16-24)
where x ( ) is a scalar field, but is otherwise arbitrary. The induction

to Af as obtained from (16-7) and (1-48) will be
Bt =
V X
= V X A + V X V x = V X A = B
corresponding
(16-25)
Thus the two inductions will be identical. However, the property that v A = 0
followed directly from our original straightforward definition (16-10), and led us to the
very desirable continuity condition (16-22) as well as the simple differential equation we
obtained in (16-18). Consequently, it would seem very reasonable that any vector
potential we wish to use should satisfy this same condition, so that, in addition to
asking that A+ give the correct B, we require that A t also satisfy (16-17):
V Af = 0
(16-26)
This will clearly lead to some restriction on our choice of x- We can see what this will
be by substituting (16-24) into (16-26) and using (16-17) and (1-45): V A1" = 0 =
V A + v V x = V 2 x> a n < 3 therefore,
V2X = 0
(16-27)
In other words, we do not want to use just any scalar function in (16-24), but we
require in addition that x be a solution of Laplace's equation. [The expression (16-24) is
an example of what is known as a gauge transformation, and we consider them again in
more detail in Chapter 22 after we have completed the development of the general
theory. The requirement (16-26), and hence (16-27), leads to the so-called Coulomb
gauge.]
The vector potential is not as useful in magnetostatics as one might expect by
analogy with the scalar potential in electrostatics. One of the problems is that, even in
many apparently simple cases, A cannot be expressed in a closed form, or at least one
that is reasonably familiar. It is in the discussion of time dependent problems that the
vector potential is most useful. Nevertheless, we consider some specific examples to
illustrate the general features that we have just considered.
31 UNIFORM INDUCTION
If B has been found by other means, then it is often possible to use (16-7) in
conjunction with (16-17) to obtain A. As an extreme example, let us consider a uniform
induction given by B = Bz where B = const. Writing (16-7) in rectangular coordinates
with the use of (1-43), we get the three equations
dy
dz
dz
dx
dx
dy
(16-28)
which we propose to solve by inspection. We can satisfy the first two by taking Ax and
A Y to be both independent of z, while Az is at most a function of z; however, if we keep
(16-17) in mind it is probably simpler just to take Az = const. Then we see that the
remaining equation in (16-28) will be satisfied if (a) Ax = 0, Av = Bx, or, ( b )
Ax= - By, Ay = 0, or, (c) Ax = - \By, Ay = \Bx, and so on. We note that (c) is
just one half the sum of (a) and (b). W e also see that these three solutions all satisfy
the condition v A = 0. Clearly, there are many other possibilities, but these are
enough to illustrate the point that there is a lot of possible arbitrariness in A. In order
to emphasize that these are different functions, yet all give the same B, the projections
16-4 STRAIGHT CURRENTS
(a)
Figure 16-2.
(b)
255
(c)
Various vector potentials that describe the same uniform induction.
of A on the xy plane for an arbitrary origin are illustrated as the corresponding parts
( a ) , ( b ) , and (c) of Figure 16-2; part (c) is obtained most easily by noting that
A2 = A2 + A2V = \B2(x2
+ y2) which is the equation of a circle.
Physically, the ambiguity in A in this case reflects the fact that a uniform B can be
produced within a given region in a variety of ways. For example, it could be produced
in the interior of a solenoid as in (15-26), or on one side of an infinite current sheet as
in (15-21), or in the region between two current sheets as in Exercise 14-9, and so on.
Thus the symmetry properties of A will generally reflect the corresponding symmetry of
the source distribution as would be found by direct calculation from (16-12), for
example.
A very common and useful expression for the vector potential describing a uniform
induction is based upon (c), which can be written in vector form as
A = jB X r
(B = const.)
(16-29)
In this form, A is independent of the original choice of direction for B.
In Section 14-2, we found the value of B produced by the straight current of finite
length shown in Figure 14-3. Now let us consider the same current distribution in terms
of the vector potential. As before, ds' = dz'z, R2 = p2 + z'2, and if we now call the
current I, (16-10) becomes
A
< ^
A = z
= z
A
dz'
/V J"o
* "
477" Z-i. ( p 2 + Z ' 2 ) 1 / 2
477
In
(P2
+ L2)1/2
{p2 + L\)l/2
= z
+ L2
In
4v
( p 2 + L\)
+ Lx
(Pj + Z , 1 ) 1 / 2 - Z . 2
(16-30)
- L,
where we have used (5-29) and (5-30) with z = 0. Before we go on to discuss the
properties of this A, let us check our result by showing that it gives the correct B. If we
combine (16-30) with (16-7) and (1-88), noting that Az{p) is the only nonzero
256
component of A, we find that Bp Bz = 0, while

dA,
B
dp
MoJP I
4"
\ (p2 + Lj)1/2$
Mo I
Airp
(p2 + L i ) 1 / 2
p2 + L \ f
+
/ 2
(P2 + L$
( p 2 + l\)1/2[{p2
- L2
+ L\)W2
+ Li
(16-31)
1/2
which is just what we found before in (14-15).

The dependence of Az on p as given by the logarithm in (16-30) is exactly that which
we found for the scalar potential of a line charge in (5-30), and therefore will have some
of the same difficulties. For example, if we try to go to the limit of an infinitely long
current by letting L2 and Ll become infinite, we will find according to (5-32) that
A = z
Mo 7
In
(4 L2LX)
i TT
1/2
(16-32)
This shows the dependence of A on p for a very long straight current but will go to
infinity for an infinitely long one. As we did for <#> in (5-28), we can choose another zero
for our vector potential by writing
A
A
= z
2<tt
In (PO
\ p
(16-33)
so that p0 is the distance from the infinitely long current at which we choose A to be
zero, since we cannot make it do so at infinity. However, there is also a special case in
which the ambiguities in A disappear because of the n a t u r e of the system.
Example
Two antiparallel long currents. Let us consider a
positive z direction and a parallel current I,
two currents are a distance la apart as shown
5-6.) Assuming L2 and Ll to be already large
system comprised of a current I in the

that is, in the negative z direction; the
in Figure 16-3. (Compare with Figure
enough so that we can use (16-32), we
257
find the total potential at P to be

<4Z-,L,)12
277 |
- In
1/2
(4 L2L,)
= z
477
In |
(16-34)
where p + and p_ are the distances f r o m the respective currents. Since L2 and L , no
longer appear in this expression, we can let them go to infinity and we get the
unambiguous finite result (16-34) for these oppositely directed currents; we note that
we would have gotten this same result if we h a d started with (16-33) instead.
In a specific case, p + and p_ have to b e evaluated in terms of the particular
coordinate system that is being used. In order to illustrate this, we assume our currents
that are parallel to the z axis, also lie in the xz plane, with I intersecting the jc axis at
a, and I at a. Then the situation will correspond exactly to that shown in Figure
5-7 with the symbol A replaced by I, and we can use (5-34) to write down the vector
potential for this case as
A /
x
.Mo I
A ( p , < p ) = z In
477
a2 + p 2 + lap cos <p

a2 + p 2
lap cos (p
(16-35)
The surfaces of constant AZ will then be the same as those given by (5-37) and (5-38)
with [X0I replacing X / c 0 , that is, AZ = const, corresponds to
= e4jrAi//fio1
= const.
(16-36)
(16-37)
P+
or, in rectangular coordinates,
/
\ 2
( x a coth 7]) + y
sinh 7j
where, now, tj = 1TTAZ/^QI.

Thus, the surfaces of constant AZ are cylinders with axes
parallel to the z axis whose intersections with the xy plane are the circles described by
(16-37), and, in fact, are exactly the circles shown as solid curves in Figure 5-8. As
before, the circles whose centers lie on the positive x axis correspond to AZ > 0 while
those with centers on the negative jc axis correspond to AZ < 0; the yz plane (x = 0 or
<p = 9 0 ) is the surface on which A, = 0.
The induction B can now be calculated f r o m (16-35) and (16-7) with the use of
(1-88); the results are that
B
1 dA,
=
H0Ia(a2
+ p2) sin cp
P d<P
DAZ
dp
B, = 0
n0Ia(p2
a2) cos <P

2
77p + p
(16-38)
(16-39)
(16-40)
so that B lies completely in the xy plane. We note that the dependence on position of
BP is like that of EV of (5-36), while that of BV is like that of EP in (5-35); in fact, we
can connect these two fields in a formal sense by B = (jii 0 0 //A)z X E as is easily
verified with the aid of (1-76). This means, however, that the lines of B for this case are
perpendicular to the lines of E shown as the dashed lines of Figure 5-8; but the curves
that are perpendicular to the lines of E in this figure are exactly the solid equipotential
circles of the same figure. Thus, the lines of B are themselves circles of this same type
described by (16-36) and (16-37) and are as shown in Figure 16-4
258
Figure 16-4.
currents.
Lines of B produced by t w o antiparallel long straight
We can also come to this same conclusion by finding the lines of B directly. The
direction of a line of induction will be given by an equation of the same kind as (5-39),
that is,
(16-41)
dsti k B
where A: is a constant. When this is written in terms of the coordinates p and tp in the
xy plane, we get the analogue of (5-41):
1 d
P
Bp
- = -f-
(16-42)
P dip
In this specific case, if we substitute (16-38) and (16-39) into this equation, we find that
1 dp
( a 2 + p2 ) sin p
p dtp
(P 2
a2) cos <p
or
(p 2 - a2) dp
2
p(a + p )
sin cp d(p
cos cp
(16-43)
This equation can be integrated with the help of tables to give

p2 + a2 Jfap cos <p
(16-44)
where J f is a dimensionless constant whose value characterizes a given line of

induction. If we express (16-44) in rectangular coordinates, it becomes
(x - \Xaf
+ .y2 = a2{\JT2
- l)
(16-45)
showing that the lines of B are circles with centers at \ J f a on the x axis and with a
radius a ( \ j f 2 1) 1 / 2 , which again leads us to Figure 16-4.
16-5
INFINITELY L O N G IDEAL S O L E N O I D
259
16-5 INFINITELY LONG IDEAL SOLENOID

Since we have found the values of B produced by this system as given in (15-25) and
(15-26), we will use them as an example of the calculation of A by means of (16-23). In
Figure 16-5 we show an end-on view of the solenoid with the equivalent surface current
K of (15-22) circulating around it; the axis of the solenoid (z axis) is out of the page as
is B. According to (16-13), each current element K da' gives a contribution to A in its
own direction, and if we consider the respective contributions dA1 and dA2 of two
symmetrically located ones K j da[ and K 2 da'2 as shown in Figure 16-6, we see that the
p components will cancel and we will be left with only a q> component. Hence we
conclude that A has only a <$> component, that is, the lines of A are circles in this plane,
(16-46)
A = Ayip)^
since A must be independent of z for this infinitely long solenoid and is independent
of <jp by symmetry. Thus, a suitable path of integration C to be used in (16-23) is a
circle of radius p for which ds = p dtp
in this way, we get
<> A ds = (^) A q>p dtp = InpAy
= <f>
so that
A j p )
<P
=
(16-47)
2 77p
1. Inside the solenoid. Here p < a, and since B, is uniform, according to (15-26), the
flux as obtained from (16-6), (1-52), and (1-53) is
$ = JBtz
s
da z. ii0nl J da = p 0 /77p 2
s
(16-48)
which, when substituted into (16-47) gives

Av(p) = jn0nlp
(p < a)
(16-49)
2. Outside the solenoid. Here p > a, and since B 0 = 0, the flux enclosed is the
constant value obtained from (16-48) for p = a, that is, $ = ju0/77-a2. When this is
inserted into (16-47), we get
/ 2 \
^<p(p)
= lVo
nI
Figure 16-5.
(P>a)
(16-50)
End-on view of a long ideal solenoid.
260
\ /
/ K 2 d"2
Figure 16-6. The resultant vector potential of

the t w o symmetrically located current elements
has only a <p component.
Figure 16-7. Vector potential and magnetic induction of a long ideal

solenoid as functions of the distance p from the axis.
We note that both (16-49) and (16-50) give the same value \n0nla when p = a; this is
in agreement with the continuity property of A as expressed by (16-22).
In Figure 16-7, we show both A and Bz plotted as a function of p in order to
contrast their behavior. This result is interesting because it shows us that outside the
solenoid where the induction is zero, the vector potential is different from zero. This
illustrates quite clearly that what really counts f r o m this point of view is not the
absolute value of A but the way in which it changes as given by B = V X A. In order to
see what is happening here, we find from (1-88) that with A of the form (16-46), the
only possible nonzero component of B is Bz= d(pAv)/p
dp. Inside the solenoid,
pAy = \ii0nlp2 from (16-49), so that Bzi = n0nl in agreement with (15-26). Outside,
however, (16-50) gives pA = \fiQnIa2 = const., so that Bz = 0 as we know.
It is also of interest to note that the circular lines of A, which we have found to hold
for this case, correspond qualitatively to Figure 16-2c, which we deduced as a possible
description for a uniform induction.
EXERCISES
261
EXERCISES
16-1 We obtained the results (16-3), (16-7), and
(16-17) by assuming filamentary currents. Show
that these same results follow if one assumes
distributed steady currents.
16-2 Apply (16-5) to a small cylinder in the
interior of an infinitely long ideal solenoid. Assume the axis of the cylinder to coincide with the
solenoid axis, and thus show that Bp = 0.
16-3 A certain induction has the form B =
(ax/y2)x
+ (fiy/x2)y
+ f ( x , y, z)z where a
and /3 are constants. Find the most general possible form for the function f(x,y,z).
Find the
current density J and verify that it corresponds to
a steady current distribution.
16-4 Show that (16-29) is a suitable vector
potential for any constant induction, not just one
in the z direction. Also show that (16-29) satisfies
(16-17).
16-5 Find the x that transforms each of the
vector potentials of Section 16-3 into the others
and verify that each such x satisfies (16-27).
16-6 Repeat the calculation that led to (16-30)
for a field point with coordinates (z, p) rather
than simply (0, p). Show that the A that you
obtain gives the same B as found for Exercise
14-2.
16-7 We always have V B = 0; also, V X B =
0 at a point where there is no current. These
equations tell us how certain derivatives of B are
related. These relations can be used to obtain
approximate expressions for B in cases where the
general case is too hard to solve exactly while
special cases are easy to do. As an example,
consider the circle carrying a current. We found
the induction along the axis quite easily in (14-18).
If one sets up (14-2) for a general field point, one
finds an integral which must be expressed in
terms of elliptic functions. One can approximate
this integral for points near the z axis by making
a power series expansion of the integrand for
small p, but we consider another approach. Beginning with (14-18), we write a Taylor series
expansion for Bp as Bp(p, z) = Bp(0, z) +
(0Bp/dp)0p.
Use (14-18) and V B = 0 to
evaluate this approximation for Bp at a point off
the axis but near the axis. Similarly, find an
approximate expression for B,{p, z).
with the origin at its center. A current / traverses
this circle in the sense of increasing polar angle
cp'. Find an expression for A produced at an

arbitrary field point (x, y, z). Write it in terms of
its rectangular components and express it as an
integral over <p'. D o not evaluate the integral.
N o w assume that the field point is on the z axis
and evaluate the integral to find A at any point on
the axis. Now go back to the general expression
for A, find integral expressions for the components of B, and then show that if the field point is
on the axis, (14-18) is obtained.
16-9 A square of edge 2 a lies in the xy plane
with the origin at its center. The sides of the
square are parallel to the axes, and a current /
goes around it in a counterclockwise sense as seen
from a positive value of z. Find A at all points
within the square. What is A at the center?
16-10 Find A produced at any point on the z
axis by the current on the circular arc shown in
Figure 14-9. Why won't your result give the correct value of B as found in Exercise 14-7?
16-11 An infinitely long cylinder has a circular
cross section of radius a and its axis along the z
axis. A steady current / is distributed uniformly
over the cross section and is in the positive z
direction. Use (16-23) to find A everywhere. If A
outside the cylinder is written in the form (16-33),
what is A on the z axis?
16-12 A wire is wound in a helix of pitch angle
a on the surface of a cylinder of radius a so that
N complete turns are formed. If the wire carries a
current I, show that the axial component of the
vector potential produced at the center of the
helix is
( / i 0 / / 2 f f ) In | Nit tan a + [ l + ( NM tan a )
] j
Show that this is the same as that which would be

produced by a current I in a wire of length equal
to that of the cylinder and going parallel to the
axis along the outer surface of the cylinder. Show
why this should be the case.
16-13 Find A for the coaxial cylinders carrying
equal and opposite currents as described in Exercise 15-7. Express your answer in terms of A 0 ,
the value on the axis. If it is possible to make
A = 0 outside of the outer cylinder, do so, and
then find the corresponding value of A 0 .
with the xy plane. The current density has constant magnitude K and is in the positive y direction. Find the vector potential A everywhere. If
262
THE V E C T O R P O T E N T I A L
you cannot make it vanish at infinity, express it in

terms of its value at the current sheet.
16-15 A sphere of radius a contains a total
charge Q distributed uniformly throughout its
volume. It is set into rotation about a diameter
with constant angular speed . Assume that the
charge distribution is not altered by the rotation
and find A at any point on the axis of rotation.
16-16 Express the constant Jf" in (16-44) in
terms of the magnitude of B when it crosses the
positive x axis at its greatest distance from the
origin. What would be the value of A, whose
constant magnitude corresponds to this same
circle?
16-17 The results that we obtained for the two

parallel infinitely long lines carrying oppositely
directed currents are not as accidental as one may
think. Consider any current distribution for which
the currents are only in the z direction. Show that
in this case, one can always write B = V Az x z,
and thus show that the lines of B will always be
parallel to the surfaces of constant Az. Apply this
result to the two antiparallel infinitely long currents and show that it is consistent with the
directions of B shown in Figure 16-4.
16-18 Physically, one would expect the flux $
to be independent of gauge. Why? Show that this
is indeed the case.
17
FARADAY'S LAW
OF INDUCTION
The overall general results that we have obtained so far can be summarized by the four
differential source equations for the vector fields
V E =
f
V x E = 0
V B = 0
(17-1)
V X B = JU0J
as given by (4-10), (5-4), (15-12), and (16-3). In addition, we have (12-13), which
expresses conservation of charge, and (14-32), which gives the force on a point charge
in terms of the electric field and the magnetic induction:
dp
V-J + = 0
at
F = # ( E + v X B)
(17-2)
T h e equations (17-1) form two completely independent sets, one for E and one B,
thereby implying no connection between the two field vectors. Faraday felt, or
suspected, that there really was a relation between these fields and tried many
experiments to verify this. About 1831, he finally succeeded in doing so, but only for
the case in which things were changing in time, that is, for a nonstatic situation. This
effect was also found independently by Henry, but it is usually referred to as Faraday's
law of induction or simply as Faraday's law.
N o w all of the equations in (17-1) were obtained from a study of static fields. Only
in our derivation of the equation of continuity did we explicitly consider, for a short
while, cases that were time dependent. In order to make progress when things are
varying in time, we will do the only thing which is at all reasonable and assume that the
equations (17-1) are still valid for nonstatic cases unless we are forced by experiment to
modify them. In this chapter, we will see that one of them must be changed to
encompass new circumstances.
17d] FARADAY'S LAW

We will consider substantially the same experimental situation as did Faraday, although it will be somewhat simplified. Suppose that we have a closed circuit C formed
by a conducting wire as shown in Figure 17-1. We also assume the presence of an
induction B so that there will be a flux O through the surface S enclosed by C. If we
arbitrarily choose a sense of traversal about C as indicated by the arrow, this will define
the direction of the element of area d a by Figure 1-24 and we can then find O from
(16-6); $ will be positive for the choice shown in the figure. We assume that there are
no batteries or other sources of emf in the circuit.
If the flux through C is constant so that d<S>/dt = 0, then it is found that there is no
current in the circuit. Faraday found, however, that if the flux through C is not
constant, so that d$>/dt = 0, then there is a current produced in C. This current is said
to be " i n d u c e d " by the changing flux and is accordingly called an induced current. It
263
264
FARADAY'S LAW O F I N D U C T I O N
Figure 17-1.
The flux through this circuit is positive.
turns out that the numerical value of the current depends on the resistance of the
circuit so that it is more convenient to express the quantitative result of experiment in
terms of the induced emf &ind, that is, the work done per unit charge as we defined it in
(12-22). It is found that
d<t>
(17-3)
ind
dt
which is Faraday's law. Since S\nA will be measured in volts, it is also often called the
induced voltage or just the voltage; another name given to it is induced electromotance.
The flux $ can change for a variety of reasons: the induction B may be changing in
time, the circuit may be altering in shape or size so that the area enclosed by it may be
changing, the circuit may be moving by translation or rotation so that it may enclose
different values of B, or a combination of these effects may be occurring. It has been
found by means of many experiments that (17-3) is valid regardless of the origins of
d$/dt. We can also note that B does not have to be different from zero at all parts of
the surface 5" and it could well be zero on portions of S; the result (17-3) still turns out
to be applicable. Furthermore, Faraday's law is still found to hold in the form (17-3)
even in the presence of matter. Since we will not systematically consider the magnetic
effects of matter until Chapter 20, we will assume materials we consider until then to be
"nonmagnetic," that is, to have the same magnetic properties as a vacuum.
The negative sign is included in (17-3) to represent the "direction" or "sense" of the
induced emf as compared to the original arbitrarily chosen sense of traversal about C.
This sense is often more conveniently described by Lenz' law. the induced emf (and
hence the induced current) has such a sense as to oppose the change that is producing it.
The key words to remember are oppose and change; in other words, it is not the
absolute value of O, nor its sign, which is of importance here but the way in which it
is changing. (When stated in this way, Lenz' law is seen to be an example of
Le Chatelier's principle, which describes the response of a system in a state of stable
equilibrium to external effects which tend to alter that state.) In order to illustrate the
application of Lenz' law let us consider for simplicity a circuit that is fixed in shape,
size, and position so that $ can change only if B changes. This is shown in Figure 17-2
and, to be specific, let us choose our indicated sense of traversal about C so that O is
positive. Now we first suppose that |B| is increasing; then d<t>/dt will be positive, and
*?ind will be negative by (17-3). This means that ?ind will be opposite to our original
choice of traversal about C and hence will be in the " sense" of the double arrowheads
of (a) of the figure; this will also be the direction of the induced current / i n d as shown.
[<îrid is not necessarily to be thought of as localized, as may be implied by the position
17-1
(a)
FARADAY'S LAW
265
(b)
Figure 17-2. The direction of the induced current when the

magnitude of B is (a) increasing and (b) decreasing.
,
' ^
înd
B (out)
\
V
x
>-
'
Figure 17-3. The conducting rod moves to the right

with constant velocity.
of its symbol, since (17-3) describes the overall effect on the macroscopic circuit.] Let us
see how our conclusions fit Lenz' law. Since O is increasing, the induced current will
" w a n t " to oppose this by trying to decrease the flux by producing lines of B opposite to
the positive sense of the normal to the area, that is, " i n t o " the circuit. The general
directions of the lines of B produced by / i n d as obtained f r o m the right-hand rule
following (14-6) are shown by the dashed curves a n d we see that they really do tend to
decrease the flux through C. Thus, the two statements lead to the same qualitative
conclusion. In part (b) of the figure we illustrate the results that we obtain when |B| is
assumed to be decreasing. Then, $ is still positive but decreasing so that d<b/dt is
negative and S i n A is positive, that is, in the same sense as our originally chosen sense of
traversal about C. The direction of Imd now is such as to produce lines of B that will
tend to increase the flux through the enclosed area, that is, to oppose the change, in
accord with Lenz' law.
Before we go on to restate Faraday's law in terms of fields, let us look at an example
that we can easily analyze quantitatively.
Example
In Figure 17-3, we show a conducting rod that rests on another conductor bent to form
a U-shaped figure. An external agent moves the rod to the right with constant velocity
v. We assume the presence of a constant induction B that is normal to and out of the
page. T h e circuit C will then be the rectangle of sides I and .x:. Let us choose the sense
of traversal about C to be counterclockwise, so that the positive sense of the area is out
of the paper. Then the flux through C as found f r o m (16-6) is
O = J(B da. = Blx
s
(17-4)
266
The induced emf as obtained from (17-3) is

d$!
dx
=
=
m
~~di~
~ ~dt
~ ~m
<17"5>
and is negative, that is, in the clockwise sense. This will also be the direction of / i n d as
shown. To check this against Lenz' law, we see that 4> increases as x increases. Thus
I ind will " w a n t " to decrease 3> by producing lines of B that, within the circuit, will be
directed into the page, in agreement with the sense of / i n d found above and shown in
the figure. The magnitude of the induced emf is constant since v is constant. But, as x
increases, more conductor is included in the circuit so that the resistance will increase.
Thus, the induced current will decrease as the area enclosed by the circuit increases.
F r o m (12-23), we see that we can interpret the existence of the induced emf as
indicating the presence of a nonconservative induced electric field E i n d along the wire so
that we can also write (17-3) in the form
,
d<&
<f)Eind ds = -
(17-6)
In (9-21) we saw that the tangential components of the electric field are continuous, and
we will see in the next sections that this is still correct. Thus, the electric field just
outside the wire will be the same as that inside so that (17-6) also holds for a path
immediately adjacent to and outside of the circuit. In fact, (17-6) no longer contains any
specific characteristics of the wire, and, in light of the last sentence, it is natural to
assume that (17-6) represents a general physical law relating the induced electric field to
the changing flux and thus will apply to any closed curve whether or not there is a
circuit there to show an induced current. Furthermore, at any point in space, we can
write the total electric field E as the sum of a conservative part E c and a nonconservative part E i n d , that is, E = E c + E i n d and then we have
(f>E ds = (f)Ec ds + <E
(17-7)
jnd ds
'c
*c
-rc
But the first integral on the right is zero for the conservative field, as we know from
(5-5), so that when (17-7) and (17-6) are combined we get an expression relating the
total electric field E and the changing flux:
d$>
d r
c
(f>E ds
B da
(17-8)
Jc
dt
dt Js
We take (17-8) to be our final statement of Faraday's law of induction as stated in
terms of the field vectors and it holds for any closed path C and for any manner by
which $ can change. Because of the various possibilities for changing flux, it is
desirable to divide our further discussion of (17-8) into two broad categories depending
on whether the medium is at rest or is moving.
STATIONARY MEDIA
If the various parts of the region of interest to us are not moving, then the bounding
curve C will not be instantaneously changing its shape or size, so that 3> can be
changing in time only because B is changing in time, that is, B = B(r, t). Then if we
also use (1-67), (17-8) becomes
r
d /
r BB
r
<T>E ds =
/ B da = da = / ( v x E) da
Tc
dt Js
s ot
's
17-2 STATIONARY MEDIA
267
and therefore
dB ,
V X E H da = 0
ot
(17-9)
But since (17-8), and hence (17-9), holds for any arbitrary path of integration with its
corresponding surface S, including an infinitesimal one, the integrand of (17-9) must be
zero everywhere, and we get
dB
V X E = - ot
(17-10)
as a differential statement of Faraday's law for stationary media. With (17-10), we

finally have a relation between electricity and magnetism in terms of the field vectors.
W e can state (17-10) in still another way. W e still have V B = 0 and therefore
B = V X A as in (16-7). Inserting this into (17-10), we get V X E = <9(V X A ) / d t =
- V X ( dk/dt),
or
v x
d A
E +
=0
(17-11)
But we also know from (1-48) that a quantity whose curl is zero can be written as the
gradient of a scalar so that we can conclude from the above that E + ( dk/dt)
= V<P,
or
E=-V<t>~
dk
(17-12)
showing that, in general, E depends on b o t h a scalar and a vector potential. In a static

case, where dk/dt = 0, (17-12) reduces to E = V< and we are back to a conservative electric field; in general, however, we can anticipate that (f> will not always be
exactly the same as the scalar potential of electrostatics.
Since V X E can now be different from zero, we have to reinvestigate the behavior of
the tangential components of E at a surface of discontinuity, since our previous result
(9-21) depended o n v X E = 0. If we p u t (17-10) into (9-18), we get
E 2 , E l f = lim { h
h 0
dB
, X n
dt
As the transition layer is shrunk to zero thickness, we certainly expect that dB/dt will
remain finite so that as h - 0, h( dB/dt) -> 0; thus the right hand side of the above
equation is zero and we find that
E2r = Ex,
(17-13)
which shows that the tangential components of the electric field are still continuous, in
agreement with the parenthetical remark following (9-22). We will have no reason in
what follows to alter this conclusion.
Example
Fixed loop in an alternating induction. As an example of a stationary system involving a
real circuit, let us consider the rectangular loop of sides a and b shown in Figure 17-4.
We choose the z axis to lie in the plane of the loop and parallel to the side a; the origin
is chosen at the center. We let the plane of the loop make an angle <p with the yz plane
so that the normal h to the plane lies in the xy plane and makes the same angle <p with
the x axis. Furthermore, let us assume the presence of an induction B directed along
the x axis as given by B = \B0 cos (ut + a), that is, it is spatially constant over the
268
Figure 17-4. A fixed rectangular loop in

an alternating induction.
area of the loop but is oscillating harmonically in time with a phase angle a that
depends on the choice of the zero for /. Here co = 2ttv is the circular (or angular)
frequency and is measured in ( s e c o n d ) - 1 or r a d i a n s / s e c o n d ; v, on the other hand is the
ordinary frequency, that is, number of oscillations per unit time and will be measured in
hertz where 1 hertz = 1 ( s e c o n d ) - 1 . Then the flux through the loop as found from
(16-6) is
$ = J/ B ft da = B0 cos B0ab cos <p sin (co/ + a )
(17-14)
(17-15)
which is seen to be proportional to the frequency of oscillation and, since it varies as

sin(co/ + a), is 90 out of phase with B, which varies as cos (co/ + a). For example,
<find is a maximum when the flux is zero but changing at its maximum rate. If the loop
were formed from a single wire wound into N turns, each turn would have an induced
emf given by (17-15) so that the total emf for the coil would be N times as great, since
it corresponds to total work per unit charge when taken around the whole circuit and is
therefore additive. Hence it would be given by NcoB0ab cos a)
(17-16)
17-3 M O V I N G MEDIA
269
Figure 17-5. The induced electric

field as a function of distance from
the axis of a cylinder of radius a
containing an alternating induction.
where B0 = const. In other words, B is spatially constant over the area of the circle but
harmonically oscillating in time; we would visualize this as being produced by an
infinite ideal solenoid with an alternating current in the windings. Applying (17-10), we
get v X E = o}B0 sin(w/ + a)z. The cylindrical symmetry of this problem and our
previous experience leads us to expect that E will lie in the xy plane and have the form
E = Ey(p)q>, that is, be tangent to circles of radius p. Accordingly, we choose such a
circle as a path of integration, and then, for any p,
ds
<ficEviP
p dcpy = 2-TTpEy
= / ( V X E) da = coi? 0 sin(co/ + a)J da,
(17-17)
with the use of (1-67) and (1-53). The area integral equals nrp2 if p < a, and has the
constant value tto2 if p > a because of (17-16). Substituting these into (17-17), we find
that
E v = - w 0 p sin (co/ + a )
(p < a)
(17-18)
(p > a)
(17-19)
<p =
sin (<o/ + a )
In Figure 17-5, we show the maximum value of \E \, that is, its amplitude, as a
function of p. As in the last example, E is 90 out of phase with B, depending as it
does on the rate of change of B, rather than on its absolute value. Induced electric fields
produced in this general way are the basis of operation of the charged particle
accelerator known as the betatron.
m
j 117-31 MOVING MEDIA

|
!
;
i
Many of the more interesting and practical applications of Faraday's law arise when a
part or all of the circuit or "medium" is moving. If we imagine our path of integration
C to pass through specific points of the medium, then, as these points move, C will be
carried along and $ can change for that reason; simultaneously, B may be changing in
time. Therefore, in order to evaluate d<$>/dt, we have to compare the flux passing
through the final surface enclosed by the final shape of C to that enclosed by the initial
270
shape of C; that is, we need to evaluate

</<*>
dt
A$
1
lim = lim
Ar-0
At
f
B ( t + At) d a ( t + At) - f B(t)
S(t + At)
S(t)
A/>o A t
da(t)
(17-20)
In Figure 17-6, we show the initial and final locations of the bounding curve C. The
element ds of C(t) is displaced an amount vAt as a result of the motion. In the
process, it sweeps out the element of area
das = ds X vAt
(17-21)
shown shaded. We also see that | J a s | is a portion of the area Ss of the "side"
connecting C(t) and C(t + At) so that the total area
S^ = S(t)
+ Ss + S{t + Lt)
(17-22)
is the bounding area of a closed volume through which the total flux of B is zero,
according to (16-5). In order to evaluate the first term in the brackets of (17-20), we
apply (16-5) to the volume of Figure 17-6 at the time t + At. Therefore
(I B(r + At) da = 0 = f
J
B(r + At) da(t)
S(t)
+ Jf B(t + At) das+

Ss
f
B(t + At) da(t + At)
S(t + AO
(17-23)
where the minus sign in the first term arises because da(t) points into the volume while
the surface integral was defined in (1-56) in terms of the element of area being positive
outward. In addition, the values of B are all evaluated at t + At, while the areas are
written in terms of their values which define the shape of the figure in Figure 17-6.
Since we are going to the limit A? - 0 eventually, it is appropriate to think of
expanding B in the power series
5B
B(t + At) = B ( t ) + At
+ ...
dt
C(t + At)
Figure 17-6. The bounding curve C at a time t and a time t + At

later. The line element ds sweeps out the area das.
271
and we explicitly keep only terms of first order in At. If we now substitute this
expression into only the first and second integrals of (17-23), use (17-21), (1-23), and
(1-29), we obtain
f
B(t + At) da(t + At) -
S(t + At)
B(t)
da(t)
S(t)
= Atl
d B
da(t)
1/5(0
,
+ <
Ot
C(t)
[B(0
x v
(0] "
+ terms of the order of ( A t ) 2
(17-24)
where we were able, after substitution of (17-21), to write the integral over
in (17-23)
as an integral over C because all of the terms involved in it are evaluated on C(t), the
initial bounding curve. If we now substitute (17-24) into (17-20) and let At > 0, the
terms originally of the order of (At)2 and higher will vanish and we are left with
d
,dB
=
dt
^s dt
da +((B
*c
X v) ds
(17-25)
The first term is familiar to us by now as arising from the time variation of B, so that
the second term is that arising because of the motion. If we now substitute (17-25) into
(17-3) and (17-8), we will get the induced emf and the line integral of the electric field
in the moving system. If we label these quantities with a prime, and use (1-23) again, we
find that
<?' = & E ' ds = - J f da +((v
jc
s ot
Jc
x B) ds
(17-26)
which can be written with the use of (1-67) as

,
, dB
/ v X ( E ' - v X B) da = - / da
Js
Js dt
(17-27)
dB
V X ( E ' - v X B) = -
dt
(17-28)
so that
since (17-27) holds for any arbitrary bounding curve.

It is important to remember that these last three numbered equations contain
quantities that are referred to, and would be measured in, different systems. The
primed quantities are those that would be observed by someone in the moving system
and hence at rest with respect to it. On the other hand, the quantities v, B, and dB/dt
are those that would be measured by an observer at rest in the unprimed system of
coordinates (which we will often call the laboratory system); this follows since our
derivation based essentially on Figure 17-6 was all written from the point of view of
one watching the moving curve C.
Keeping these remarks in mind, we can now obtain an interpretation of the
argument of the curl in (17-28) by considering the force on a point charge q as
described by these two hypothesized observers. F r o m the point of view of the stationary
observer, what one has is a charge q moving with velocity v in an induction B and the
force on it will be given by (17-2) as F = q(E + v X B) where E is the electric field in
this system. For the observer moving along with the primed system, the charge q is at
rest, and therefore the only possible electromagnetic force that could be written
according to (17-2) is F ' = q & where E ' is the electric field in the moving system. If we
restrict ourselves to cases of zero or negligible relative acceleration, as we have been
272
consistently doing, then both of these systems can b e regarded as inertial systems, and
we k n o w from mechanics that the two forces will b e the same, that is, F' = F. Equating
these expressions for the forces, we find the fields in the two systems to be related by
E' = E + v X B
(17-29)
N o w we see that the argument of the curl in (17-28) is really the electric field in the
unprimed system so that the equation can be written as
SB
V x E = - at
(17-30)
which is exactly the same as (17-10) that we found f o r the case of stationary media. In
other words, Faraday's law of induction, when written in this way, has a form that is
independent of the motion of the medium. [Again, when we consider relative velocities v
that are comparable to the speed of light in a vacuum in Chapter 29, we will find that
(17-29) will need to be modified somewhat; our conclusions with respect to (17-30),
however, will not be altered.]
The terms depending on v in (17-26) and (17-29) are usually known as " m o t i o n a l "
terms, while the others that depend on the time variation of B are often known as
" transformer" terms. Some problems are most easily solved and understood by looking
at things from the point of view of an observer in the moving system, so that one uses
expressions for the motional emf and motional electric field given by
^=$E'
-ds=$(vXB)-ds
(17-31)
E'm = v X B
(17-32)
N o w let us look at a few examples of this type.

Example
W e begin by looking again at the system illustrated in Figure 17-3. We have already
analyzed this by using Faraday's law in the macroscopic form (17-3) and found the
overall induced emf to be given by (17-5). Since B is constant in time in this example,
the emf must be given completely by the motional term. As the sliding rod is the only
part of the system that is moving, we see f r o m (17-32) that it is the only place where
E ' = 0. W e also see that E ' is directed along the moving rod as shown in Figure 17-7
and has the constant magnitude E' = Bv since B and v are perpendicular. We also
show the direction of ds that corresponds to the original choice of positive sense of
traversal around the circuit on which the calculation of $ given in (17-4) was based.
Then E ' and ds are oppositely directed and E ' ds = E'ds Bvds and (17-31)
gives
- f
Bvds = Blv
(17-33)
rod
in agreement with (17-5). Since the direction of the induced current will be that of E ' ,
we see from Figures 17-7 and 17-3, that 7 ind will have a clockwise sense that we found
before to be consistent with Lenz' law. Although we got the correct value of the
induced emf in (17-5), we were not able to "localize" it, but now we see its origin can
be ascribed completely to the situation within the moving conducting rod. Furthermore,
in this example, we can interpret the origin of the induced current as due to the
magnetic force produced on a charge moving in an induction as seen by the stationary
observer.
273
ds
B(out)
E'
Figure 17-7. The electric field seen by an observer

moving with the rod of Figure 17-3.
Example
As a variation on the above example, we consider a conducting rod of length I moving
with constant velocity perpendicular to the rod a n d also perpendicular to a constant B
as shown in Figure 17-8. In this case, we do n o t have a complete circuit and there
cannot be a circulating induced current. In fact, in the final steady state of the system,
there can be no current at all in the rod and therefore E ' = 0 according to (12-25). But
then (17-29) gives
E = -v X B
(17-34)
so that the stationary observer sees an electric field of magnitude E = Bv directed

upward along the rod as shown in Figure 17-9 a n d that is drawn f r o m the observer's
point of view. N o w this electric field must come from somewhere, and since we are
dealing with a homogeneous material, it can only come f r o m surface charges on the
ends, a n d they will have the signs shown in the figure. Furthermore, the unprimed
observer will conclude that there is a potential difference A^> between the ends of the
rod that can be found from (5-11) and (17-34) to be
A<f>= f
E ds = El = Blv
(17-35)
' +
B(out)
v x B
l igure 17-8. A moving conducting

rod not part of a circuit.
Figure 17-9. The situation as

seen by a stationary observer.
274
FARADAY'S LAW OF I N D U C T I O N
E' *
-E'
E' #
= 0
Figure 17-10.
The situation as seen by the observer moving with the rod.
[We note that this has the same numerical value as the motional emf given by (17-33)
for the previous example, but now we are talking a b o u t a completely different concept.]
W h a t is the origin of the surface charges? During the original stages of the motion,
the movable charges of the conductor were subject to the magnetic force q\ X B that is
directed downward along the rod. This will lead to a separation of the charges, the
positive ones moving toward the bottom of the rod and the negative ones to the top.
But these separated charges will produce an electric field pointing upward that will tend
to decrease the total force on a given charge in the interior. Finally, enough charges will
be separated so that the electric field E produced by them will lead to an upward force
that just balances the downward magnetic force. This is the final equilibrium state
described by (17-34).
N o w the number of charges at the ends will be the same for both observers, since it
involves only counting that they can do equally well. Therefore, the moving observer
sees the same distribution of charges on the ends that we found above. But since
E ' = 0 in this system, the rod will be seen as an equipotential volume and therefore the
potential difference between the ends will be A^/ = 0, in contrast to (17-35), and is
consistent with the requirement that there is no current in this system. However, these
surface charges will produce an electric field E ' = 0 in the whole region outside the rod,
so that the final view of things will be qualitatively like that shown in Figure 17-10; the
values of E ' at the surface can be found in terms of the surface charge density from
(6-4). Thus, this example has been valuable in showing us that different observers will
not necessarily describe a given physical situation in exactly the same way when stated
in terms of fields, but their relative descriptions will depend on their relative motion.
If the rod is a dielectric, rather than a conductor, there will be no movable charges
which can be separated to lead to a situation like Figure 17-9 and it is quite possible for
E ' to be different from zero. In this case, the moving dielectric can become polarized;
several examples of such situations will be considered as exercises.
Example
Rotating loop. Let us consider again the loop with the dimensions and orientation
shown in Figure 17-4. However, we now assume that B = B$l is constant in time, while
275
the loop is rotating as a rigid body about the z axis with constant angular speed to so
that <p is a function of the time with the form = Jf B ft da = B 0 a b cos t + cp0)
s
(17-36)
Substituting this into (17-3), we find the induced emf to be

$ind uB0ab sin ( u>t + <p0)
(17-37)
Again, if the loop were formed from a single wire wound into N turns, each turn would
have an emf given by (17-37) so that the total emf of the coil would be N times as
great, that is,
ind NuB0ab
sin (cor + <p0)
(17-38)
What we have discussed here is a rudimentary form of an electric generator, and it is

Faraday's law applied in this manner that accounts for most of the electrical energy
produced in the world today.
Although a comparison of the induced emfs given by (17-15) and (17-37) shows
obvious similarities in form, the physical mechanisms involved are quite different, one
being produced by a changing induction in a stationary coil, while the other arises in a
coil moving with respect to a constant induction.
It is worthwhile looking at this same example in terms of the motional electric field.
In Figure 17-11, we show the plane of the loop at some instant. The angular velocity of
rotation is given by w = coz while r is the position vector of a point on the loop with
respect to the origin at its center. The arrows indicate the positive sense of traversal
about the loop that we have been using so that the normal n of Figure 17-4 is here out
of the paper. For rigid body rotation, the velocity of any point is given by v = u X r,
and with B given by B = B$i, we find the motional electric
field
from (17-32),
(1-30), and (1-20) to be
E'm = v X B = (B r)co =
BqXCJZ
(17-39)
We now must insert this expression into (17-31) and integrate around the loop. We see
276
FARADAY'S LAW OF INDUCTION
(
*
from Figure 17-11 that the ds for the horizontal portions of length b are all
perpendicular to z so that, on them, Y,'m ds ~ z ds = 0 and they will not contribute
to the integral. On the two vertical sides of length a, we have ds = dz z, and if we let x,
and xr be the x coordinates of the left- and right-hand sides, respectively, we find that
when (17-39) is substituted into (17-31), we get
= f
(a/2)
V 2
( B 0 X / T O dz) + f " / 2
J
( B 0 x r t o dz) ccB0a(x/
- xr)
(17-40)
~(a/2)
Upon referring back to Figure 17-4, we see that x r = \b sin<p while x, = \b sinqp;
thus x! xr = b sincp and (17-40) becomes
= uB0ab sinqp = uB0ab sin (tot + cp0)
(17-41)
exactly as we found in (17-37). We now see, however, that while there is a motional
electric field associated with every point on the loop as given by (17-39), only those
portions that are parallel to the angular velocity contribute to the induced emf. On the
other portions, the induced electric field is always perpendicular to the displacement
and hence can do no work on a charge as it passes around the loop.
Example
Homopolar generator. This interesting device was invented by Faraday. Consider a
plane circular conducting disc of radius a rotating with constant angular speed w in a
uniform B that is perpendicular to the plane of the disc. This is shown in Figure 17-12
where both B and u have been chosen to be into the paper. There are sliding contacts
at the central axle at 0 and at a point P on the rim of the disc. The circuit is completed
with conducting wires and a resistance R . At a point with position vector r with respect
to the origin at the center, the velocity will be v = u X r and will be tangential as
shown. Therefore, there will be a motional electric field found from (17-32) to be
E ^ = ( t o X r ) x B = uBr
(17-42)
and is thus seen to be radially outward. When we put this into (17-31), we see that the
only contribution to the induced emf will come from that part of the disc between 0
B (in)
(in)
Figure 17-12.
tor.
Homopolar genera-
*
C
C
f
Qr
17-4
INDUCTANCE
277
and P and since ds dr, we will get

=
f E'm ds = f <x>Brdr = \u>Ba2

J
o
o
(17-43)
The resultant current in the external circuit will have the direction shown since it
corresponds to the general sense of YL'm. Although devices like this can be built and they
do operate, they are generally impractical because of the large sizes and high rotational
speeds necessary to get a reasonable value of the emf. If, instead of this arrangement,
an external emf, such as a battery, is used to produce a current through the system by
means of the contacts, then the disc will rotate a n d one has a homopolar motor.
IT>4l INDUCTANCE
Although we began with Faraday's law in the f o r m (17-3) appropriate to the system as
an overall unit, we rather quickly converted it into an expression involving fields. For
many applications, it is convenient to deal with properties of the system as a whole,
much as capacitance proved to be a useful concept, and we now want to revert to that
attitude. We begin by looking again at magnetic flux ; in this way, we will be led to
another important geometrical property of a system known as inductance.
In Figure 17-13, we show two filamentary circuits C, and Ck with their respective
currents I j and I k . The line elements dsy and dsk are located by their position vectors
tj a n d rk with respect to an arbitrary origin 0 a n d RJk = |r r j . N o w the current Ik
will produce an induction BA at each point of the surface Sj enclosed by C, and thence
a flux $ k j in Cj that can be found from (16-6). For our purposes, however, it will be
more useful to express the flux in terms of the vector potential by means of (16-23).
Therefore, if A ^ ) is the vector potential produced by circuit Ck at the point r- of the
Figure 17-13.
circuits.
Positions of current elements of t w o filamentary
278
circuit Cj, then the flux through C, due to Ck will b e given by (16-23) and (16-10) as
A t (r,.) d S j - J Z < f > < $
Cj
477- Jc/ck
KJk
(17-44)
We see that the flux through Cj is proportional to the current Ik in Ck. If we give this
factor of proportionality the symbol Mjk and call it the mutual inductance of circuits j
and k, we can write
=
(17-45)
Mo r r dsj ' dsk
, - . .
(17
* -
"46)
It can be seen from (17-46) that the mutual inductance is a purely geometrical factor
relating the sizes and relative orientation of the two circuits and can be found in
principle by evaluating the double integral over the two circuits. Since n 0 was defined
to have the units of h e n r y s / m e t e r in (13-2), we see that the unit of mutual inductance
will be 1 henry.
Similarly, if we calculate the flux in Ck due to the current I j in C,, we will find that
*j~k = MkJIj
(17-47)
R
4 7tJCJcj
kj
<17-48)
But since dsk dsj = dsy dsk, Rkj = Rjk = |r, r fc |, and, by using arguments similar to those used in connection with (6-19), we see that the order of integration does not
affect the value of the double integral, so that
MkJ = Mjk
(17-49)
This tells that the flux through Cj by a current I0 in Ck equals the flux through Ck by
the same current I 0 in C,.
It can be seen from (17-46) that the mutual inductance can be either positive or
negative depending on the choices made for the senses of traversal about Cy and Ck
since they can be made arbitrarily and independently and this will of course be
reflected in the signs of the fluxes.
The overall induced emf arising in one circuit due to a changing current in another
can be conveniently written in terms of mutual inductance; if we substitute (17-45) into
(17-3), we get
d
h
^ = " ^ - 7 -
(17"50>
since the circuits are assumed to be at rest so that Mjk is a constant. The emf induced
in Ck can be obtained from (17-50) by interchanging the indices j and k.
If there is more than one circuit producing the resultant B throughout Cy, then it
follows from (14-4) or (16-11) that the total flux will be the sum of terms like (17-45):
% -
(17-51)
k
Similarly, the total induced emf in C, produced by other circuits can be obtained f r o m
(17-51) and (17-3) and is
d<t>,
^
dlk
/
(H-52)
Although (17-46) provides a recipe by which the mutual inductance can be calculated in principle, a direct application of it often leads to complicated integrals.
17-4
INDUCTANCE
279
Example
Two coaxial parallel rings. In Figure 17-14, we have a circle C, of radius a lying in the
xy plane with origin at the center. Another circle Ck of radius b is parallel to the xy
plane and its center is on the z axis at a distance c f r o m the origin. The relevant
position vectors are seen to be r = a cos <pyx + a sin <p-y, rk = b cos
+ b sin <pky +
cz a n d therefore
dSj = d\j = ad<pj( sin <pjk + cos <p7y)
dsk = drk = bdq>k( sin ( p ^ + coscp^y)
Rjk = I*'j ~ f/tl = [c2 + a2 + b2 - lab cos
- <pj]1/2
dSj dsk = ab cos (<p, - <p*) Jp,When these are substituted into (17-46), the result is that
Mjk
l*0ab r2v ,2*
cos (cpj - cpk) d<pjd<pk
477- Jq Jq
[c2 + a2 + b2 lab cos (<pj qp^)]1,72
Although it was easy enough to write (17-53) down, it cannot b e expressed in terms of
elementary functions but requires the use of elliptic functions, so we leave (17-53) as it
is.
In m a n y cases the problem of finding the mutual inductance can be handled more
easily by using the definition of it as flux per unit current as expressed by (17-45). This
method thus generally requires the calculation of 0 from a previous knowledge of B
and the use of the defining equation (16-6).
Figure 17-14.
Two coaxial parallel rings.
280
FARADAY'S LAW OF I N D U C T I O N
Example
Short coil wound over a long solenoid. Suppose we have an infinitely long solenoid with
ns turns per unit length. We now suppose that another coil is wound tightly about the
solenoid so that there are a total of Nc turns occupying a total length lc as shown in
Figure 17-15. The number of turns per unit length of the coil will then be nc = Nc/lc.
In this case, we know how to find B if there is a current I s in the windings of the
solenoid; the uniform value inside as given by (15-26) is B,, = fi0nsIsz.
If the coil is
tightly wound on the solenoid surface, we can take the cross-sectional area of it to be
approximately the same as that of the solenoid, S. Then since Bs is normal to the plane
of S, the flux per turn of the coil will be $ t u r n = BsS = fi0nsSIs
so that the total flux
through the Nc turns will be $ c = Nc$turn (i0nsNcSIs.
Substituting this into (17-45),
we find the mutual inductance of this system to be
Mcs = = n0nsNcS
= nQnsnclcS
(17-54)
Since Msc = Mcs, according to (17-49), we can use this result to find the flux 4>s
produced in the solenoid when there is a current I c in the coil. It will be given by
= MsJc = Mo nsnclcSIc.
m
It is possible, of course, for the circuit C, to produce a flux / ; that passes through
itself. The coefficient of proportionality that arises in this case is called the self-inductance or, often, simply the inductance. If we use Figure 17-16 and proceed in the same
manner by which we obtained (17-44) through (17-46), we find that
(17-55)
% v = LJi
L
n =
Mo 1 1
477 Jc/c,
d s
ds'
(17-56)
;J
Here now the double integral is taken twice over the same circuit and dsj and ds'j are
separate line elements of C,. As indicated in (17-56), we will sometimes suppress one or
more of the indices on the self-inductance when it will not lead to any confusion. By
comparing Figures 17-16 and 14-2, we see that because of the sign conventions we are
consistently using, 3>7 } will always be positive so that the self-inductance L / / as given
by (17-55) will always be a positive quantity.
If the current I- is changing, then by (17-3), there will be an emf induced in the
circuit because of its own changing flux. This is usually called the self-induced emf or
the back emf and will be given by
dl,
dL
L^r
=
L
(17-57)
^j, self
JJ
dt
dt
J
Finally, if there are other circuits about, the total flux in C, will be the sum of
(17-51) and (17-55), and the total emf induced in it will be the sum of its self-induced
Figure 17-15. Short coil wound

over a long solenoid.
EXERCISES
281
a n d mutually induced emfs. Thus, we can write

= Y.MjkIk
k
(17-58)
(17-59)
al
k
where the index k takes on the value j as well as the values used to label the other
circuits and with the understanding that when the indices are the same we use the
notation
MJJ = LJJ- Lj
(17-60)
Although (17-56) has an appealing appearance and is useful in formal calculations, it
will not work for true filamentary currents of zero thickness in the sense that it will
always give an infinite value for the self-inductance. This occurs because in the double
integration, as soon as d s j and ds'- coincide, R j j = 0 and the integral diverges. The
physical reason is that near such a wire |B| ~ 1 / p and goes to infinity as p - 0,
leading to an infinite flux. For more realistic circuits, for which one is really dealing
with distributed currents, rather than trying to develop a generalization of (17-56), one
ordinarily goes back to (17-55) to find the self-inductance as the ratio of flux to current,
or uses the energy methods that we will discuss in the next chapter. Consequently, we
consider only one simple example of the application of (17-55).
Example
Ideal solenoid. As in the previous example, the induction is given by B = ix0nl and is
u n i f o r m over the cross section of area S. Therefore, the flux per turn will be
BS = fj.0nSI. If we consider a length / of the solenoid, there will be nl turns and the
total flux will be n0n2SH. Substituting this into (17-55), we get L = <b/I or
L = n0n2lS
(17-61)
We note that L ~ IS, which is the volume of this portion of the solenoid. Also L ~ n2\
this arises because B ~ n and the number of turns is also ~ n.
EXERCISES
17-1 Choose the positive sense of traversal about
C to be the opposite of that shown in Figure 17-2
and then show that (17-3) leads to the same
directions for the induced emf for the two cases
discussed in the text.
17-2 In a certain region, the induction as a

function of time is given by B = B0(t/T)Z where
B{) and T are constants. Find A. Assuming the
scalar potential to be zero, find E from the above
result for A. Use this E to evaluate the left-hand
282
side of (17-8), and thus show directly that it

equals the right-hand side.
induced emf between the ends of the rod when

the final steady state has been attained.
17-3 Suppose the current in the infinitely long

straight circuit C' of Figure 13-5 is given by
/ = I0e~x' where / 0 and X are constants. Find
the induced emf that will be produced in the
rectangular circuit of this same figure. What is the
direction of the induced current?
17-11 A conducting fluid is flowing with speed v

in a horizontal channel of depth w and width / in
a region where the vertical component of the
induction due to the earth is Bd. Two metal
electrodes are placed opposite each other on the
vertical walls. The electrodes are rectangular, of
dimensions a and b, and are at the same distance
d from the bottom of the channel with the long
edges horizontal. Find (a) the resistance of the
fluid column contained in the rectangular parallelepiped between the electrodes, (b) the induced
emf between the electrodes, (c) the resultant current if the electrodes are connected by a wire of
negligible resistance, and (d) the numerical values
of (a), (b), and (c) for the case of sea water
(o = 4/ohm-meter) if Bd = 5.5 X 10""5 tesla, w
2 meter, 1=5 meter, a = b = 0.5 meter, v = 3
meter/second, d = 0.25 meter.
17-4 An infinitely long straight wire carrying a

constant current / coincides with the z axis. A
circular loop of radius a lies in the xz plane with
its center on the positive x axis at a distance b
from the origin. Find the flux through the loop. If
the loop is now moved with constant speed v
parallel to the x axis and away from / , find the
emf induced in it. What is the direction of the
induced current?
17-5 Find the emf induced in the loop of Figure

17-4 when it is rotating so that <p = ut + <p0
while simultaneously B is oscillating at the same
frequency, that is, B = xB0 cos (ut + a). Is it 17-12 Suppose that the circular loop of Exercise
possible to choose the constants <p0 and a in such 17-4 is now rotated so that its center traces out a
circle of radius b with center at the origin while
a way that the induced emf is always zero?
the plane of the loop remains parallel to the z
17-6 A certain closed circuit has a total resis- axis. In other words, the loop has an angular
tance R. If the total flux through it changes from velocity o = uz. Find E' in the moving loop.
3>, to <Ey, show that the magnitude of the total Find $E' ds around the loop. Explain your
charge that will flow through the circuit is given results from the point of view of someone in the
by Q = ($y - 3>,)/R. Note that this result is fixed laboratory system.
independent of the total time required or of the
17-13 Two small nonmagnetic dielectric needle
particular manner in which O varies with time.
shaped dust particles are moving with velocity v
17-7 A uniform induction B is parallel to the at right angles to a uniform B. The volume of
axis of a nonmagnetic cylinder of radius a and each particle is V and the distance R between
dielectric constant Ke. If the cylinder is rotated them is much larger than their linear dimensions.
about its axis with a constant angular velocity w The particles are oriented with their axes on the
parallel to B, find the polarization produced within same line that is perpendicular to both v and B.
the cylinder and the surface charge on a length I Find the force that the particles will exert on each
of it.
other, in addition to gravitation. Is it attractive or
17-8 A spherical shell of radius a rotates about repulsive?
the z axis with constant angular velocity w. It is 17-14 An electromagnetic "eddy current" brake
in a uniform induction which is in the xz plane consists of a disc of conductivity a and thickness
and at an angle a with the axis of rotation. Find d rotating about an axis passing through its center
the induced electric field at each point on the and normal to the surface of the disc. A uniform
sphere.
B is applied perpendicular to the plane of the disc
17-9 Find the value of S'm produced by a homopolar generator of 1 meter diameter, rotating with
an angular speed of 3600 revolutions/minute in
an induction of 0.1 tesla.
over a small area a2 located a distance p from

the axis. Show that the torque tending to slow
down the disc at the instant its angular speed is co
is given approximately by auB2p2a2d.
17-10 A metal rod of length / is rotated about

an axis passing through one end with constant
angular speed to. If the circle swept out by the
rod is perpendicular to a uniform B, find the
17-15 A very thin conducting disc of radius a

and conductivity a lies in the xy plane with the
origin at its center. A spatially uniform induction
is also present and given by B =
cos (ut + a)z.
EXERCISES
Find the induced current density Jj produced in

the disc.
17-16 The previous two exercises involved currents that are produced in conductors as a result
of Faraday's law and they are given the name
"eddy currents." If they arise from a time varying
induction, it is possible to obtain a differential
equation describing their properties. Begin by
combining (17-10) and (12-25), and then by
applying (1-120) show that, for steady currents,
the eddy currents satisfy the equation v 2 J/ =
an0(dJf/dt).
17-17 The henry is a fairly large unit of inductance. In order to get an idea of its size, apply
(17-61) to a solenoid 1 meter long and 2 centimeters in diameter that has a total of 600 turns.
17-18 Use (17-46) to find the mutual inductance
of the circuits shown in Figure 13-5. Is your
answer consistent with the result of Exercise 17-3?
(Hint: assume the straight portion goes from L
to L where L is very large and go to the limit
L -* oo as late in your calculation as possible.)
17-19 The two infinitely long antiparallel currents and the rectangle of Figure 17-17 all lie in
the same plane. The sides of length b are parallel
to the directions of the currents. Find the mutual
inductance between the circuit of the oppositely
directed currents and the rectangle. Verify that in
the appropriate limit your result reduces to that
of the previous exercise.
283
17-21 A toroidal coil of N turns has a central

radius b and a square cross section of side a.
Find its self-inductance.
17-22 Show that, as might be expected, when
the dimensions of the cross section of a toroidal
coil are small compared to the central radius, the
results of the two previous exercises give the same
approximate self-inductance when expressed in
terms of the cross-sectional area S.
17-23 Two infinitely long hollow cylindrical
conductors are parallel to each other. The radius
of each is A and the parallel axes are separated
by a distance D so that in cross section they are
as shown in Figure 11-10. They carry equal and
opposite currents /. Use the results of Exercise
16-17 and following (16-40) to show that the
self-inductance of a length / of this system is
simply related to the capacitance of the same
length / and then apply this relation to find L.
17-24 Two infinitely long coaxial cylindrical
conducting shells carry oppositely directed currents of equal magnitude. If the radius of the
inner shell is a and that of the outer is b, find the
self-inductance of a length I of this system.
17-25 In Figure 17-14, assume that c a and
c b and show that (17-53) becomes approximately 7rn Q a 2 b 2 /2c 3 .
17-26 Consider a situation where one can assume that all of the flux that passes through a coil
1 also passes through a second coil 2, and conversely. For example, we could have a very long
solenoid of nl turns per unit length and crosssectional area S1 over which is wrapped another
/>'
very long solenoid of n2 turns per unit length and
area S2 such that S2
as would be the case
for very thin wires. Similarly, we could have one
<
d
>
set of toroidal windings wrapped tightly over
another set on the same torus. Show that under
these conditions, the mutual inductance Mn is
-Drelated to the individual self-inductances Lt and
L2 by | M l 2 \ = y!-\L2 . This relation gives an
upper limit to the mutual inductance of the two
coils. In a real case, however, not all of the flux
produced within a given coil will pass through the
other, as would certainly be the case for the
Figure 17-17. The currents and rectangle of
circular loops of Figure 17-14. Then the mutual
Exercise 17-19.
inductance would be less than the value given
17-20 A toroidal coil of N turns has the central above. It is customary in this case to write the
where k is some
radius of the torus equal to b and the radius of its relation as \M12\ = k^LxL2
circular cross section is a. Show that its self- number less than unity and is called the coefficient of coupling.
mductance is p0N2[b (b2 a2)1/2].
18
M A G N E T I C ENERGY
00
iN
In Chapter 7, we evaluated the electrostatic energy of a system in terms of the

reversible work required to establish a given configuration of charges. It also takes work
to produce a given set of currents in circuits and our aim here is to find it and thus be
able to assign a magnetic energy to the system. In the magnetic case, however, we do
not have the analogue of the point charge of electrostatics so we have to proceed in a
different way. After we have obtained a result in terms of the overall parameters of the
system, we will find that we will be able to express the energy in terms of the magnetic
induction and to think of it as being distributed throughout space.
8-1
ENERGY OF A SYSTEM OF FREE CURRENTS

We begin by considering a system of currents in a group of circuits, so that what we
will actually be evaluating is the energy of free currents, specifically conduction
currents. W e want to calculate the reversible work required to start from an initial
situation in which all of the currents are zero and to attain the situation in which the
current in the jth circuit has its final value I j where j = 1, 2 , . . . , N with N being the
total n u m b e r of circuits. As we know from Chapter 13, there will be forces of attraction
and repulsion between these circuits that we will assume to be all perfectly rigid and in
fixed positions so that we need not be concerned with the mechanical energy of possible
deformation or motion of the circuits. Furthermore, we know from (12-35) that the very
existence of the currents implies a conversion of electrical energy into heat. Although
this energy must be supplied by the external sources, such as batteries, it is not part of
the reversible work required to establish the currents and hence we can exclude it from
consideration here.
Suppose that we are at an intermediate stage of the process where the current in each
Cj is given by its if so that the currents have not attained their final values Ij. Any
change in any of the currents in a time interval dt will produce a change dQ>j in the flux
through Cj and, from (17-3), an induced emf in it. The external source must supply an
emf equal to this value but of opposite sense in order to maintain the system. During
this time, a charge dqj = ij dt will have passed through it, so that the work done by the
external source will be dWext j = <fext dqj = $jM&ijdt
= ijdQj. Thus we have, as a
general expression,
d W ^ u j = ijd&j
(18-1)
as the reversible work done by the external agent and hence the change in magnetic
energy. Summing up over all of the currents, we find the total work to be
N
dWext = dUm = Z i f d Q j
(18-2)
7= 1
which we now write as the increment dUm of the magnetic energy.

Since the rigid circuits are constant in shape and relative configuration, we see f r o m
(17-46) that the inductances will be constant, and then, according to (17-58), the only
284
18-1
E N E R G Y O F A S Y S T E M O F FREE C U R R E N T S
285
way the fluxes can change is by way of current changes so that

d*j = Mjk dik
k=\
(18-3)
If we combine (18-2) and (18-3), we obtain

dUm=
E E Mjkijdik
j =1A: = 1
(18-4)
which must now be summed up as the currents are increased from their initial values of
zero to their final values I j . There are a variety of ways in which one can imagine this
process being carried out. For example, we could bring the currents to their final values
one at a time, we could increase them all simultaneously, we could increase some first
and the rest later, and so on. In any event, the final result for the energy must depend
only on the final state and be independent of the process by which it was attained as is
familiar to us from thermodynamic discussions of state functions. Accordingly, we will
use a simple scheme for increasing all the currents simultaneously, and leave as an
exercise the task of showing that the same result is obtained by changing only one
current at a time.
Let us assume that at any instant t, each current is the same fraction f of its final
value, so that j .(*) = f { t ) I j where f ( t ) is independent of j and ranges from its initial
value of zero to its final value of unity. Then dik = Ikdf and (18-4) becomes
dUm = Y . Y , M j k I j I k f d f
j *
(18-5)
Summing these changes from the initial to the final state, and choosing our zero of
energy such that \Jm = 0 for the initial state of n o currents, we find the magnetic energy
to be given by
um =
L E A f j i ' , 4 J0 ' f d f 4 z e
j k
y=l*=l
(18-6)
We see that (18-6) is independent of the precise manner in which / changed in time
and, in that sense, is independent of the process; this is, of course, always subject to the
conditions that these changes are slow enough so that the process can be considered to
be reversible.
We can rewrite (18-6) in terms of the currents and fluxes by using (17-58) again to
give
=
(18-7)
J
where <1> is the total flux through C- from all sources including itself.
Example
Two circuits. With N = 2, we can write (18-6) with the use of (17-60) and (17-49) as
Um =
= \Lxll
+ MnIJ2
+ MX2IXI2
4- M2lI2Ix
+ \L2q
+ M22/|)
(18-8)
The first term depends only on circuit 1 and represents the work needed to establish the
current I x against its own self-inductance and can properly be described as the
"self-energy" of Lx; similar remarks apply to the third term. Thus, for a single isolated
286
inductance L with a current I, we can write its energy as

Um = LU2
(18-9)
The middle term of (18-8) involves b o t h currents and arises f r o m the fact that a
changing flux in one circuit produces an induced emf in the other; this term can be
taken to represent the "interaction energy" of the two circuits. [It is interesting to
compare (18-6) through (18-8) with the results (7-16) through (7-18) which we found for
the case of conductors in an electrostatic field.]
Our results so far have been appropriate for filamentary currents and now we would
like to rewrite them so as to cover the important cases where we have distributed
currents. If we use (16-23) to rewrite the flux in (18-7), we obtain
Um = ^ Z $ M r j ) - I j d s
j cj
(18-10)
where A(ry ) is the total vector potential at the location r7 of the current element I j ds y
of Cj. N o w the integral over Cj gives the contribution to Um f r o m all of the current
elements of Cy, and then the sum over j finally produces the sum of the contributions
from all of the current elements of the whole system under consideration. Therefore, if
we use the first of the equivalents in (12-10), remember that we are dealing with free
currents, and integrate over all regions containing currents, we can rewrite (18-10) as
u
m = \ JyJ/(r)
A(r) dr
(18-11)
Actually, we can extend the region of integration to cover all space, since in regions in
which there are no currents, J j = 0, and they will contribute nothing to (18-11); thus we
can finally write the magnetic energy in the f o r m
Um = \ [
J/(r) " A(r) dr
(18-12)
^ all s p a c e
U p o n comparing this result with (7-10), we see that the vector potential plays much the
same role in determining the energy of a system of currents as the scalar potential does
for charges. The result (18-12), as does (18-7), has a form appropriate to the attitude
that the energy is associated with the currents and is " located" at their positions.
If there are surface currents, we can use the other equivalent of (12-10) to write their
contribution to the energy as
U
- \ f
^
M r ) A(r) da
(18-13)
all s p a c e
Example
Infinitely long ideal solenoid. In (15-22), we found that this system can be described in
terms of a surface current K = nl circulating about its outer surface so that, in accord
with Figure 16-5, we can write KF = nly. Similarly, we f o u n d in (16-46) and after
(16-50) that the vector potential on the surface (where
= 0) is A = \n0nla$
where
we write the radius as a. If we insert these into (18-13) in order to obtain the energy of
a length / of the solenoid, so that the area of integration is 2 veil, we find that the
integrand is constant and therefore
u
rn= 2 / ( 2 M o " 2 / 2 5 )
da =
2MAj2^2')//2
(18-14)
03
O
18-2 ENERGY IN T E R M S O F THE M A G N E T I C I N D U C T I O N
287
If we compare this with (18-9), and note that the cross-sectional area is S = ira2, we see
that the inductance is L = n0n2Sl, exactly as we found in (17-61) by calculating the
flux.
18-2 ENERGY IN TERMS OF THE MAGNETIC INDUCTION

As we briefly noted after (18-12), its form is appropriate to interpreting the energy as
associated with and located at the currents. This is a point of view that is consistent
with the action at a distance property of Ampere's law with its emphasis on the current
elements and their relative orientations. Our primary interest, however, is in the field
description of phenomena and we now want to express the energy in these terms.
W e can use (15-12) to write JF = ( v X B)/N0 since we are only considering free
currents at present; then (18-12) becomes
U
(18-15)
- * - / ( V X B ) - A J T
Z f l 0 >
We can rewrite the integrand by using (1-116), (16-7), and (1-17):

A ( V X B) = B ( V X A) V (A X B) = B 2 V (A X B)
(18-16)
If we substitute this into (18-15) and use the divergence theorem (1-59) to write one of
the resulting volume integrals as a surface integral, we find that
Um = - i - (b2
2fl0 Jy
dr -
XB)-Ja
llLa'S
(18-17)
N o w our starting integral (18-12) was to be taken over all space so that in (18-17) we
should regard F a s a very large volume and S its extremely large bounding surface.
Then as we let V become infinite, 5 will recede out to infinity. We assume that the
current distribution will always be contained in a finite volume. Thus, as 5 gets very far
away, the whole current distribution will appear to be contained within a very small
volume a distance R away. If we look back at (16-12) and (14-7), we see that, as
R -* oo, the worst that we can expect to occur is that
A -
A X B
B - i
(I8"18>
~ i l
so that the magnitude of the integrand is decreasing as R3. As we will see in the next
chapter, however, this estimate is overly pessimistic and, in fact, at large distances the
current distribution will appear as a closed loop for which we will find that
The surface of integration is increasing as R2, so that, for very large R

ft
A X B) J a ~ ~
R2 ~
(18-20)
' OO
Thus, when the volume V in (18-17) is increased to include all space, the surface
integral will vanish by (18-20) and the expression for the energy becomes simply
B2
u
- "
/
âll space 41* o
d j
1 8
"
2 1
>
[We note that even if (18-18) were applicable, the surface integral would still vanish as
it did in the corresponding electric case (7-27).]
288
The result (18-21) has the same general form as the electric result (7-28) and we can
give it the same sort of interpretation, that is, the magnetic energy is distributed
continuously throughout space with an energy density um given by
U
B2
^ro
(18-22)
so that the total magnetic energy can be written as

um=
umdr
(18-23)
âll space
The units of u m will be joule/(meter) 3 .

As discussed after (7-30), we are not forced to interpret our expression (18-21) in this
way, but it is the natural thing to d o and turns out to be very useful as well as
consistent with all that we will do later.
We also recall that (7-28) could be usefully applied to the calculation of capacitance.
Similarly, we can use (18-21) for the calculation of inductance by combining it with
(18-9). Thus, if we have found B by other means, we can use it to calculate the energy;
then we know that Um will turn out to be proportional to I2 so that L can be found as
L = 2 U m / I 2 . As usual, the method is best illustrated with examples.
Example
Infinitely long ideal solenoid. We have already used this to illustrate (18-13). Using
(15-26) and (15-25) in (18-22), we find the energy density to be um = \(i0n2I2
= const,
inside the solenoid and um = 0 outside. If we consider a length / of the solenoid of
cross-sectional area S, the volume will be SI and (18-23) becomes
Um = / 2^0n2I2dr
= ~ii0n2I2Sl
in complete agreement with (18-14); thus we get the same self-inductance L = n0n2Sl
as we have found twice before.
Example
Coaxial line. In Figure 18-1, we show the same system previously shown in Figure 6-12,
that is, two coaxial cylindrical conductors; the inner conductor has radius a while the
outer conductor has an inner radius b and an outer radius c. We now assume that the
conductors carry equal and opposite total currents of magnitude / uniformly distributed over the cross sections, while there is a vacuum in the region between them. W e
choose the z axis in the direction of the current in the inner conductor. The required
values of B can be found by using the integral form of Ampere's law (15-1). By now we
recognize from the symmetry of the situation that B has the form B =
so that
it is appropriate to choose as a path of integration a circle of radius p. Then the
evaluation of (15-1) is exactly like that of (15-17) so that, for any value of p, we can use
(15-18) and say that
(p) =
(18-24)
and all that remains is the evaluation of 7 enc for a given path. It will be convenient to
treat the four regions labeled 1, 2, 3, 4 on the diagram separately; they correspond
18-2
ENERGY IN T E R M S O F THE M A G N E T I C I N D U C T I O N
289
Figure 18-1. A coaxial line

carrying equal currents in
opposite directions.
respectively to the inner conductor, the space between conductors, the outer conductor,
and all space outside of the system. In this way, we will get a clearer idea of how each
region contributes to the total self-inductance of the system.
1. (0 < p < a) Here, Ienc/I = { ^ / n a 2 ) = p2/a2 and (18-24) yields
Mo'P
iv - 5 ^
(18.25)
as in (15-20). Since B 2 = B2 in this case, the corresponding energy density found from
(18-22) is
Mo j 2 P 2
57T a
<18
"26)
The volume element is dr = p dp dtp dz by (1-83), and, if we integrate over a length /

of the system, we find the energy associated with this part to be
Uml = j' j2" j
o 2 9A P dP dv
<S1T a
dz
10 77"
(18-27)
We note that this result is finite because the value of B goes to zero as p > 0 rather
than becoming infinite there.
2. (a < p < b) Here 7 enc = 7, and proceeding exactly as above, we obtain
B<f2
(18
~ ~2ttp
(18-29)
87r2p2
Um 2 =
H0II2
47t
"28)
In
(b
\a
(18-30)
290
3. (b 
Ic4
1P
(18-32)
/i0//:
2
r 2 C2 + p2
In -
2 2
47r(c 6 )
- - ( c 2 - 62)(3c2 - 62)
(18-33)
4. (c < p) Here, 7 enc = 7 - 7 = 0. Therefore, 2? 94 , w m4 , and UmA are all zero.

We can thus obtain the total energy by adding (18-27), (18-30), and (18-33), and
when we combine this with (18-9), we find the total self-inductance of the length / of
this system to be
11
L p.()/ (
\ 8*r
1
(b\
1
+ In I +
xc 4 l n (
2tt
\a/
2t7(c 2 - Z>2) L
b 2
)(
3 c 2
ft2
(18-34)
which has a form clearly indicating the contribution of each region to the total. In most
practical situations, the middle term in (18-34) is the principal source of the inductance.
I B-3 MAGNETIC FORCES ON CIRCUITS

As we know, two current carrying circuits will in general exert forces on each other and,
in principle, these forces can be calculated from Ampere's law. However, as we saw for
electrostatic forces in Section 7-4, it is often desirable to express forces in terms of
energy changes. This is what we now want to do for magnetic forces.
For simplicity, we consider only two circuits, since this will be sufficient to illustrate
all of the general features involved. The general situation will then be just that shown in
Figure 13-1, which was our starting point in magnetism. As we saw in (13-9), in order
to have this system in equilibrium, the magnetic force F m on C must be balanced by an
equal and opposite mechanical force F m e c h due to an external agent, which could include
springs, braces, and the like.
N o w let us imagine that the position vector r of each point of C is slowly changed
by the same amount dr; thus the whole circuit is being translated by this amount b u t is
not rotated. The other circuit C' is held fixed, as are the batteries responsible for the
maintenance of the currents. (Consequently, we will never be able to treat these circuits
as being completely isolated systems.) U n d e r these conditions, the work done by the
mechanical force will be reversible work and will b e equal to the change dUt in the total
energy of the complete system, that is, dUt = F m e c h dr. But with zero acceleration, or
nearly so, the circuit C will remain in equilibrium, or be only infinitesimally different
from it, so that we will continue to have F m = F m e c h and then we can also write
dUt=
Fm'dr
(18-35)
Comparing this with (1-38), we see that

F m = -VU,
(18-36)
which is a result analogous to the one-dimensional one of (7-36). Although these are
18-3 MAGNETIC FORCES O N CIRCUITS
291
fundamental results, we would like, if possible, to relate the magnetic force to changes
in the magnetic energy only. We begin by noting that the total energy change will be
the sum of the magnetic energy change, dUm, and that of the batteries, dUB, so that
dU( = dUm + dUB
(18-37)
As we saw for the analogous electrostatic case in Section 7-4, it is instructive and useful
to visualize two possible conditions under which this displacement is carried out.
1. Constant currents. When one circuit is moved relative to the other, the fluxes
through them will generally change; these will produce induced emfs and in order to
keep the currents constant, the batteries will have to do work against these emfs or will
have work done on them depending upon the signs of the flux changes. We can use
(18-7) to write the energy of the currents I and / ' as Um = \(I$> + /'<&'). Then, if I
and / ' are constant, we have dUm = \(Id<& -I- I' d<&'). We previously found the work
required from the external sources (the batteries) in (18-2); however, we must remember
that in this case the currents are already at their final values so that the dUm of (18-2) is
not the same as the dUm we are discussing here. Since any work done by the batteries
represents a decrease in their energy, we get
dUB=
dWext = -(IdQ>
+ I'd')
= - 2 dUm
(18-38)
which is opposite in sign to that of the circuits and twice as great in magnitude.
Substituting this into (18-37), we get dUt = dUm in this case so that (18-36) becomes
Fm = ( v U m ) j
(constant currents)
(18-39)
where the subscript I on the gradient reminds us that all currents are to be kept
constant while evaluating the derivatives.
Since the force described by (18-39) is in the same direction as the gradient of Um, its
tendency will be to increase the magnetic energy of the system. Thus, the equilibrium
state with constant current (Fm = 0) will correspond to a maximum of the magnetic
energy. Then we see from (18-7) that the constant current circuits will tend to adjust
themselves by translation so as to enclose as much flux as possible. In the next chapter,
we will see that a similar conclusion can be drawn with respect to possible rotations.
In terms of the present notation, the energy expression (18-8) can be written
Um = \LI2 + Mil' + \L'I'2 where M is the mutual inductance. Since we are considering only displacements that are rigid translations, the shapes of the circuits will not
change so that the self-inductances will also be constant according to (17-56); therefore,
M is the only quantity affected by the displacement and (18-39) gives
Fm = II'vM
(18-40)
For example, the jc component would be given by

F
= II'
dM
dx
(18-41)
Before we discuss this in more detail, let us consider the other possibility.
2. Constant flux. As we mentioned above, when the circuits are displaced, the fluxes
could be expected to change if nothing were done to prevent this. However, if the
currents are appropriately adjusted during this process, one could carry it out while
keeping O and 0 ' constant. Then, the corresponding magnetic energy change obtained
from (18-7) will be
dUm = \{<$> dl + <t>' dl')
(18-42)
We also see from (18-1) that this will require n o energy change in the battery so that
dUB = 0; the batteries' energy will nevertheless decrease because of the constant
k
292
irreversible conversion of energy into heat as described by (12-35), but we are not
interested in that here. Thus, for this case, (18-37) becomes dU\ = dUm so that (18-36)
takes the form
Fm = ( v f / m ) 0
(fluxes constant)
(18-43)
[We recall the similar results (7-45) and (7-37) that we previously obtained.]
Now (18-43) must lead to the same expression (18-41) since the overall physical
situation is the same, only the calculational schemes differ. Unfortunately, (18-8) is
expressed in terms of the currents, while in order to use (18-43) effectively, we want an
expression for Um in terms of the fluxes. Applying (17-58) and (17-60) to this case, we
get 3> = LI + MI' and <J>' = MI + L'l'. Solving these for the currents, we find that
Z/4> -
/ =
r2
LL' - M
Substitution of these into Um = ^(/<& + / ' $ ' ) >
1 / 1
t -L'O2
2
LL' - M \2
U=
m
- M $ + L5>'
=
(18-44)
v
LL' - M2
'
obtained from (18-7), yields
1' =
as
1
\
- M$><$>' + -Lib'2]
2
/
(18-45)
7
Now, in consonance with (18-43), we can differentiate this expression while realizing
that everything in it except M is constant; we find that
( )
=
rfML'O2
2
I dx )9
(LL' M )
- (LL'
+ M2)<&<$>' + ML$/2]
dx
8M
= -II'-
dx
with the use of (18-44). Therefore, the .x component of (18-43) will be

f
I dx U
Sx
exactly as we found in (18-41) as an application of (18-39); since there will be similar

expressions for the y and z components, we will again be led to (18-40) as a general
result for the force on the circuit C.
Example
Ampere's law. Although (18-40) has a different appearance, it must express the same
thing as did Ampere's law. We can easily see that this is the case. Substituting the form
of (17-48) appropriate to the present case into (18-40), we obtain
r
" . - ^ r
ds ds'
<18
4 r ~
"46)
Now the rigid translation of C that we have envisaged will not affect the line elements
ds and ds', nor the limits of the double integrals; in fact, the only variable as far as the
V operator is concerned is R. Thus we can interchange the differentiation and
integration and use (1-141) to obtain
IL.JI'
- -îiS
r
ds
ds (v1
/x0II'
'> ( r ) = -
r r (ds ds')R
which is exactly the version of Ampere's law expressed by (13-6) and, as we know, is
equivalent to (13-1) for calculating the total force on C.
18-3 M A G N E T I C FORCES O N CIRCUITS
293
Example
Two interpenetrating long solenoids. Suppose we have two long ideal solenoids, one of
which extends into the other a distance x as illustrated in Figure 18-2. We will assume
the windings are thin enough so that their cross-sectional areas are approximately equal
to the same value S. In this case, the solenoids are not infinitely long and we cannot
assume that the values of B produced by them in the overlapping region of interest here
are the same as those found in the interior far from the ends. In fact, as we will see
later, the lines of B near the end tend to diverge quite rapidly from the axial direction
as indicated schematically in Figure 18-3, and thus will not all pass through the distant
turns of the other solenoid. Nevertheless, we can neglect these " e n d effects," provided
that jc is already large enough, and find the mutual inductance by considering only the
flux contained within the overlapping region; we can therefore use a previous result
based on Figure 17-15. Let the inner solenoid have n turns per unit length and carry a
current / ; the values for the outer are n' and I'. The mutual inductance can then be
obtained from (17-54) and is M = fi0nn'Sx. Using this in (18-41), we find the force to
be
Fmx = ^nn'SII'
SBB'
Mo
(18-47)
where the last form, expressing it in terms of the inductions produced by each solenoid,
was obtained by using (15-24). What will be the direction of this force? If I and I'
both circulate about their respective solenoids in the same sense, the flux produced in
one by the other will be positive and M will be positive by (17-45). Then dM/dx
will
be positive, Fmx will be positive, and the inner one will be attracted into the outer one.
This is consistent with our qualitative statement following (13-14) that parallel currents
attract each other. If / and / ' circulate in opposite senses, then the flux produced by
one in the other will be negative, according to our sign conventions, making M
negative, and then dM/dx
and Fmx will both be negative. Therefore, the inner
solenoid will be repelled by the outer, again consistent with our qualitative statement
that " unlike" currents repel each other. But all of this is also contained in (18-47) if the
currents are assigned relative signs according to their relative sense of circulation. Thus,
if they are in the same sense, II' will be positive, as will Fmx, while if they are
circulating in opposite senses, II' < 0 and Fmx < 0 in agreement with all of the above.
We can also express this same result quite nicely in terms of the inductions by writing
(18-47) as
Fx -
B
Mo
B'
(18-48)
which will automatically give the correct sign to Fmx.
Figure 18-2. Two interpenetrating long solenoids.
294
-a-
Figure18-3. General appearance

of lines of B near the end of a
solenoid.
Previously, by considering the force on a plate of a charged parallel plate capacitor,

we f o u n d in (7-50) that there was a force per unit area on the conductor that was
directed outward from the surface (a tension) and was numerically equal to the energy
density at the surface. We can find a similar result here by looking at the specific
example of what is, perhaps, the nearest analogue we have to the oppositely charged
plates of a parallel plate capacitor.
Example
Two oppositely directed current sheets. In Figure 18-4, we show an edge-on view of two
long parallel planes carrying surface currents of constant density K but in opposite
senses, one directed out of the page and one into it. The width of the sheets is w and jc
is the distance between them as measured in the sense shown. If the sheets are very long
and if w jc, then we can neglect the edge effects and treat them as if they were
infinite plane sheets. In this case, we know from (14-26) that the B produced by each
sheet has magnitude i0K, is perpendicular to the direction of K and hence is in the
B = 0
K(out)
K(in)-
B = 0
Figure 18-4.
Two oppositely directed current sheets.
EXERCISES
295
plane of the page, and has opposite directions on the two sides of the sheet. These
inductions are shown as the dashed arrows in the figure and are labeled by their source
current ("out" or "in"). The resultant induction B = B c + B( is shown as the solid
arrow and we see that B ^ O only in the region between the sheets where it has the
constant magnitude fi 0 K. Putting this into (18-22), we find the energy density to be
um = 2^0K2 = const, between the plates and zero everywhere else. Therefore, the
magnetic energy of a length I of this system, and corresponding volume wlx, as
obtained from (18-23) is
Um=
j iMo K2dr
= jn0K2wlx
(18-49)
The magnetic force on the upper sheet, which corresponds to an increase in x, as

obtained from (18-39) is then found to be
dU
1
m = ~ ^ x = ~n0K2(wl)x
(18-50)
We see that this force is in the positive x direction and thus is repulsive, as is to be
expected for these oppositely directed currents. We also note that Fm is proportional to
wl, the area of the sheet, so that if we introduce a force per unit area, fm, we can write
1
|FJ
p o *
<18-51'
-<-
and we see that its magnitude is exactly equal to the magnetic energy density. In this
case, x is the normal to the surface directed from the B # 0 region to the B = 0 region
which enables us to write the force per unit area in vector form as
I=/*-*
(18-52)
The direction of the force is such as to tend to move the sheet into the B = 0 region
and hence may be properly described as a pressure.
Combining all of the relations involved, we can write fm variously as
f
- -
w r
K 2
K B
( 1 8
-53)
where K and B are evaluated at the position where we want fm. These results are all
very similar to what we found in the electrostatic case as described by (7-49), (7-50),
and (7-52).
We have assumed throughout that the circuits are completely rigid. If the internal
forces of the material are not large enough to counterbalance these magnetic forces, the
conductors comprising the circuits will deform. This deformation will generally continue until the new internal elastic forces produced will be large enough to produce a
new equilibrium configuration of the systems.
EXERCISES
18-1 The magnetic force on a current element
ids is ^ F = ids X B by (14-5). Suppose that
every element of ds is given the same displacement dr while i is kept constant. Find the work
done by the external agent and thus show directly
that it is idO where d$> is the change in flux.
18-2 Consider a single self-inductance L at an

intermediate stage where its current is i with
0 < i < I. Find the work required to increase the
current by di. Then add all these work increments
from the initial state of zero current to the final
state of current / and thus obtain (18-9) again.
296
18-3 Instead of the method we used to sum up

(18-4), evaluate Um in the following way. Beginning with all currents equal to zero, increase il
from 0 to fl while keeping all the others zero.
Then, with
at its final value, increase i2 from 0
to f2 and so on for each in turn. In this way,
%
show that (18-6) is again obtained.
18-4 Use the fact that the energy of two circuits
as given by (18-8) must be positive to show that
| My 21 < yjLlL1 as was discussed in another way
in Exercise 17-26.
18-5 A self-inductance L, a resistance R, and a
battery of emf Sb are all connected in series. Use
energy considerations to show that the current /
satisfies the differential equation L(di/dt) + Ri
= Sh. Now assume that i = 0 and the battery is
switched out of the circuit. Solve the resulting
equation and find the relaxation time of this
system.
18-6 Consider a vacuum situation where E and
B each have the same numerical values in their
appropriate set of units. In other words, E = x
volt/meter and B = x tesla. Find the ratio um/ue
of their respective energy densities and evaluate
its numerical value.
18-7 A long cylindrical nonmagnetic conductor
of radius b has a coaxial cylindrical hole of radius
a drilled along it, that is, it is like Figure 18-1
with conductor in region 2 and everything else a
vacuum. It carries a current I distributed uniformly over the cross section. Find the magnetic
energy associated with the induction in a length /
of the conductor.
18-8 A closely wound toroidal coil of N turns
has the central radius of the torus equal to b and
the radius of its circular cross section is a. Find
its magnetic energy when there is a current I in
the windings and show that this leads to the same
self-inductance as found in Exercise 17-20. If
: / > , show that L becomes approximately the
same as that of a long ideal solenoid of length
2-nb. Is this reasonable?
18-9 Under what conditions, if any, will the
contribution of region 1 of Figure 18-1 to the
self-inductance of the coaxial line be greater than
that of region 2?
18-10 A common method of making coaxial lines
is to use a very thin outer conductor, that is, the
radii c and b of Figure 18-1 are nearly equal.
Show that under these conditions the contribution of the outer conductor to the self-inductance
becomes approximately (p0l/8irb)(c b) and
thus show that (18-34) is consistent with the result

found for Exercise 17-24.
18-11 Generalize (18-39) and (18-40) to the case
in which there are more than two circuits in the
system.
18-12 Use (18-41) to find the force on C of
Figure 13-5 and verify that the result is the same
as obtained for Exercise 13-4.
18-13 The circular loop of Exercise 17-4 now
has a current / ' in it that is in the sense required
to make M positive. Find the force on the loop.
18-14 Find the force on the circle of radius b
shown in Figure 17-14. Leave your answer in
integral form but verify that it reduces to the
result of Exercise 13-8 under appropriate conditions.
18-15 Use the result of Exercise 17-25 to find
the force on the circle of radius b of Figure 17-14
when the two circles are far away from each
other.
18-16 A long flexible spring of length / hangs
vertically with its upper end rigidly fastened. It
has n turns per unit length and the radius of its
circular cross section is a. A mass m is fastened
to the lower end of the spring. A current I passed
through the spring will enable it to support the
weight without its being stretched or contracted.
If the mass of the spring is neglected show that
I = (\/na)(2mg/'n\LQ)l/2.
Do this in two ways by
using both (18-39) and (18-43).
18-17 A long thin coil of length /, cross-sectional area 5, and n turns per unit length carries
a currcnt /. It is placed along the axis of a large
circular ring of radius a which is carrying a
current
If S is the displacement of the center
of the coil from the center of the ring along the
coil axis, find the force on the coil as a function
of S.
18-18 A circular loop of wire of self-inductance
L and radius r carries a current I. Show that the
force tending to increase its radius is \I2(dL/dr).
If T is the tension in the wire for which it will
break, show that the wire will break unless T is
greater than ( I 2 / 4 m ) ( dL/dr).
18-19 What vacuum value of the induction will
result in a magnetic pressure of one atmosphere?
18-20 A very long thin cylindrical shell of radius
a carries a current / in the direction of its axis.
Find the force per unit area on the shell. Does
this force tend to explode the shell or to collapse
it? What is the total force on a length / of the
shell?
19
M A G N E T I C MULTIPOLES
In Chapter 8, we saw how the scalar potential at a point outside of a finite charge
distribution could be described by the various multipole moments of the system. Each
multipole moment depends on a particular aspect of the detailed charge distribution. In
this chapter, we want to do a similar thing for an arbitrary current distribution.
FHTil THE MULTIPOLE EXPANSION OF THE VECTOR POTENTIAL

The general situation is illustrated in Figure 19-1; compare with Figure 8-1. We have a
current distribution J(r') contained in some volume V. We choose an origin 0 in some
arbitrary way, but near or within V. We want to find the vector potential A at the field
point P with position vector r, that is, in the direction r and at a distance r from 0.
This potential is given by (16-12) as
u0
a
J(r') d r '
( 1 9 4 )
- / k . - 4 -
where
R = |r r'| = ( r 2 + r'2 2rr' cos 0')1/2
(19-2)
follows from the law of cosines applied to the figure.

As before, we assume that P is far enough outside V so that r > r' for any portion
of V'. Then we can use the expansion given by (8-12) to write (19-1) in the form
00
1
A(r) = Z -7TT / J(i"') r ' / ^ > /( c o s #') d r '
477- / = 0 r
J y
U
which is the general multipole expansion of the vector potential.

few terms with the use of (8-10) and (8-14), we get
(19-3)
If we write out the first
A(r) = A M ( r ) + A D ( r ) + A e ( r ) + . . .
=
f J(r') d r ' +
J J(r')(r r') d r '
4vTr JV
47rr Jy
+
*r ' ) 2 ~ r ' 2 ] d T ' + ...
(19-4)
where the terms in the sum are, respectively, the monopole term, the dipole term, and
the quadrupole term. As in the electrostatic case, their dependence on the field point
distance r goes successively as 1 / r , 1/r2, 1 /r \ and so on, so that as we get farther
away from the current distribution, the higher-order terms in the expansion become less
and less important. These terms still involve b o t h the field point P and the source
points because of the term ? x' ~ cos 6', and we want to write them as a product of
something that depends only on the location of the field point and something
characteristic only of the current distribution. It will be convenient to discuss these
terms separately.
297
298
Figure 19-1.
at P.
Geometry for the calculation of the vector potential
1. The Monopole Term

In the electrostatic case, the monopole term turned out to be proportional to the
monopole moment of the charge distribution, that is, its net charge. In the magnetic
case, however, there are no magnetic charges, as we discussed following (16-4), and this
leads us to suspect that the monopole term may actually be zero. Let us consider the x
component of the integral in AM. If we let A in (1-115) be replaced by J , while u is
replaced by x', we find that
V ' ( x ' J ) = J x + x ' V ' J = J x = Jx
since V ' J = 0 for stationary currents. Then, if we also use the divergence theorem
(1-59), we obtain
f Jx dr' = f V ' ( * ' J ) dr' = < x'J da!
Jy,
Jy,
JS>
Since the currents are localized, J = 0 on each point of S' so that the last integral
vanishes and
/ JxdT'
Jy
= 0
There will be similar results for the y and z components, and when these are combined,
we get
f J(r') dr' = 0
Jy
(19-5)
for localized stationary currents. Since this integral is always zero

Aw(r) = 0
(19-6)
19-1 THE M U L T I P O L E E X P A N S I O N O F THE V E C T O R P O T E N T I A L
299
thus verifying our conjecture and showing that the leading term in the expansion of the
vector potential is always the dipole term. [This also justifies our previous use of (18-20)
rather than (18-18).]
While the proof of (19-5) is formally correct, it is also helpful to obtain this same
result in a different way. In a steady current distribution, the currents follow closed
paths, and we can think of the charges moving in filamentary tubes of currents much as
shown in Figure 12-6a. Then for the y'th such loop of corresponding current /, we see
f r o m the equivalent in (12-10) that its contribution to the first integral of (19-4) will be
the sum of all its current elements I j d s , that is, $ I j d s j . Summing u p over all of these
closed filamentary currents into which we have divided the whole current distribution,
we again get
f
J ( r ' ) dr' = / , ( dsj = 0

j
q
since the integral over d s j is just the sum of the successive displacements of a point and
hence is the net displacement, and when a point is taken around a closed path, its net
displacement is zero so that $ d s j 0; this was also proved more formally in Exercise
1-25.
2. T h e Dipole Term
For convenience, let us give the symbol
3
to the integral in the dipole term:
= f J ( r ' ) ( r r ' ) dr'

Jy
(19-7)
The process of writing this as a product in which the field point is somehow separated
out is somewhat involved and it is easier to deal with a scalar quantity. If we let C be
an arbitrary constant vector, then we can form the scalar product C S>. After doing
this, we divide the integrand into two equal parts, add and subtract the quantity
!,(J ?)(C r') under the integral, and we find that we can write the result as the sum of
two integrals:
C . = / " (C J ) ( f r - r ' ) d r ' = ( C - # ) + + (C 0 ) _
Jy
(19-8)
(C 9)+=
(19-9)
where
(C
| / [(C J ) ( f r ' ) + ( J r ) ( C r ' ) ] dr'

Jy
= \ f
J yf
[(C J ) ( f r ' ) - ( J f ) ( C r ' ) ] dr'
(19-10)
and we consider these separately.

N o w C is a constant by definition, and f is a constant with respect to source point
derivatives, that is, with respect to V ' . Therefore, it follows from (1-113) that
V'(C r') = C
and
V'(rr') = f
(19-11)
Using these last results, along with (1-111), we find that the bracketed part of the
integrand of (19-9) can be written as
J [C(r r ' ) + r ( C r ' ) ] = J [(f r ' ) v ' ( C r ' ) + (C r ' ) V ' ( ? r ' ) ]
= J V ' [ ( C - r')(? r')]
(19-12)
300
MAGNETIC MULTIPOLES
N o w the last term has the form J V

where i f is the scalar (C r')(f r ' ) ; thus, by
using (1-115) and (12-15) for steady currents, we can write (19-12) further as
J v ' ^ = V ' - ( S T J ) - y ( v ' J ) = V ' [(C r ' ) ( f r ' ) j ]
(19-13)
which is our final form for the integrand of (19-9). If we n o w substitute (19-13) into
(19-9) and use (1-59), we finally get
(C 9)+
= \<f> (C r')(r r ' ) ( J da')

*S'
(19-14)
where S' is the surface enclosing V. However, V encloses all of the currents so that
J = 0 at all area elements da' of S', and therefore
( C - 9 )
= 0
(19-15)
The terms in the brackets in (19-10) can be written as

C [J(? r ' ) - r ' ( J f ) ] = C [r X ( J X r ' ) ]
(19-16)
with the use of (1-30) and (1-16). If we put this into (19-10), and use (19-15), we find
that (19-8) becomes
C 3 = \ j
C [r X ( J X r ' ) ] dr' = C
[f X ( J X r ' ) ] dr'^j
(19-17)
Since C is a completely arbitrary vector, (19-17) can be always true only if 3> equals the
quantity in braces. Furthermore, r is constant as far as the integration over the primed
variables is concerned and can be taken out of the integral. Doing this, we find that
= t X \ j
J X r'dr'
= (2/
'
dr,]j
X r
(19-18)
Finally, if we combine (19-18), (19-7) and (19-4), we find that the dipole term can be
written as
a D (n
A
r) =
Mo
477-/"'
\ ( r' X J ( r ' ) dr'

J y/
X R
(19-19)
which has the desired form of a product of a quantity referring only to the location of
the field point and something depending only on the properties of the current
distribution.
T h e quantity in brackets is given the symbol
m = \ f r' X J ( r ' ) dr'
Jy
(19-20)
and is called the magnetic dipole moment of the current distribution. [In some books, m
is defined as fi 0 times the right-hand side of (19-20); this is comparatively rare
nowadays, but it is a point to keep in mind.] This definition enables us to write the
dipole term as
JU0 M X ?
FI 0 M X R
A * ( r ) = T T - T 2T - = " T :T "
477 r
4tt
r
(19~21)
As shown in Figure 19-2, A D is perpendicular to the plane formed by m and r and its
magnitude will be AD = n0m sin^/477/- 2 . The appearance of s i n ^ may be a little
surprising since, by analogy with the electrostatic case as given by (8-48), one might
have expected to find cos ^ in a potential called a " d i p o l e " potential. However, when
we come to calculate the induction in the next section we will see that the analogy is
really quite exact when fields are compared.
19-1
THE M U L T I P O L E E X P A N S I O N O F THE V E C T O R POTENTIAL
301
Figure 19-2. Relation between a magnetic dipole and the

vector potential it produces.
Fortunately, it will be sufficient for our purposes to consider only the dipole term, so
we will not discuss the quadrupole term of (19-4) any further.
3. Effect of the Choice of Origin

The magnetic moment m defined in (19-20) depends on the absolute value of r' and, if
we choose a different coordinate system, r' will certainly change and perhaps m will
too. W e recall that in the electrostatic case, we f o u n d in (8-43) that the electric dipole
m o m e n t was independent of the choice of origin and hence was a unique property of
the charge distribution provided that the electric monopole moment vanished. We saw
in (19-6) that the monopole term always is zero in the magnetic case so we suspect that
m may be independent of the choice of origin as well. Suppose that instead of choosing
our origin at 0 of Figure 19-1, we choose a new origin 0, which is obtained by
translating our axes without rotating them, by the displacement a as illustrated in
Figure 19-3. Then we see f r o m the figure that the old and new position vectors of J are
related by a + r'n = r' or r'n = r' a. If we insert this into (19-20) to find the value of
the magnetic dipole moment referred to the new axes, we get
(r' a)
J dr' = \ f r'
J dr' \a
f J dr' = m
J [//
y
Figure 19-3. The new origin is displaced
by a from the old.
302
since the integral multiplying a is zero by (19-5). Thus, m = m, and the magnetic
dipole m o m e n t is always a unique property characteristic of the current distribution
and any convenient choice of coordinates can be used for calculating it.
19-21 THE MAGNETIC DIPOLE FIELD

The expression \ D given by (19-21) will be the predominant term in the vector
potential when the field point is sufficiently far away from the current distribution. In
order to study its properties it is convenient, as it was in Section 8-2, to assume that
(19-21) holds everywhere in space. Then we can call it the magnetic dipole field and
think of it as being produced by a fictional point dipole m located at the origin. Later
we will consider specific current distributions that can be thought of in this way, but for
now we concentrate on finding the induction B produced. W e use spherical coordinates
to locate the field point P and choose the z axis in the direction of m; this gives the
situation shown in Figure 19-4. Then m = m i and (19-21)'becomes
.
Pom z
/x
AD(r) = -7I
xw ra
4irr'
ôm
in 0
2
4 7rr"
(19-22)
so that the only nonzero component of X D is A D ( p . The curves of constant values of

A D(p are given by
r2
H0M
sin 0 = CD sin 6
4TTADtp
(19-23)
where the constant CD characterizing a given curve depends on the value of AD(p. These
curves are shown as the solid lines in Figure 19-5; however, the direction of A D is into
the page on the right half of the figure and out of the page for the left half.
The induction is found from B = V X A D . Using (19-22) and (1-104), we find the
components of B to be
B.
r sin 0 86
1
-7J?
(sin 6Ad)
Vo
( r A
"'
Pom
4tt
sin 0
2 cos 0
(19-24)
47T
while B v = 0. Thus B lies in the same plane as m and the field point, while A D is
perpendicular to this plane.
C o m p a r i n g (19-24) with the corresponding results (8-50) for the electric dipole field,
we see that they are proportional to the magnitude of their respective dipole moments
Figure 19-4.
dipole.
Calculation of the field produced by a magnetic
19-3 FILAMENTARY C U R R E N T S
303
Figure 19-5. Field of a magnetic dipole. Lines of B are dashed. Solid

lines are those of constant magnitude of the vector potential that is
directed into the page for the right half and out of the page for the left
half.
and they have the same dependence on angle and distance. In fact, we can easily see
that they are simply related by B = n0e0(m/p)E.
Consequently, the lines of B, which
are shown as the dashed lines in Figure 19-5, will be given by an equation of exactly the
same form as (8-52) and will be just like the dashed lines of E shown in Figure 8-7.
19-3 FILAMENTARY CURRENTS

So far we have concentrated on distributed currents. We can easily adapt our results to
filamentary currents by using (12-10) in order to replace Jdr' by Ids'. Thus, we see
from (19-20) that the magnetic dipole moment of a single filamentary circuit C' will be
given by
1 r
m = ( r' x ds'
(19-25)
2 Ja
and any convenient choice of coordinates can be used to evaluate the integral.
Example
Plane filamentary current. If the circuit lies in one plane, the integral can be given a
simple and useful interpretation. Choosing the origin in this same plane, we get what is
304
MAGNETIC MULTIPOLES
shown in Figure 19-6. N o w r' X ds' is perpendicular to the plane of the loop, and the
magnitude |r' X */s'| equals the area of the parallelogram with r' and ds' as sides as
we f o u n d from Figure 1-15. Comparing these two figures, we see that the shaded
triangular area in Figure 19-6 formed by lines f r o m the origin to the extremities of ds'
is just half the area of the parallelogram, that is, if da' is the area of the shaded region,
then \t' X ds' = da' and the integral in (19-25) becomes
r' X ds' = j da' = S = S n
(19-26)
where S is the total vector area enclosed by the current and n is the normal to its
surface given by the usual convention defined in Figure 1-24. Therefore (19-25)
becomes simply
m = IS = ISA
(19-27)
so that the magnitude of the dipole moment of a plane filamentary current is just the
product of the encircling current and the area enclosed by it. We note that the result is
independent of the shape of the circuit.
Example
Plane circular ring. If the ring is a circle of radius a, then m = lira2. Then at large
distances, its axial induction Bz can be found f r o m (19-24) with 0 = 0 and r = z. The
result is that Bz = j u 0 I a 2 / 2 z 3 exactly as we found in (14-20).
Plane circuits like these are often called current loops or current whirls and will
produce the dipole vector potential and induction when one is far away f r o m them.
[Nearby, of course, the details of the distribution become more important and one has
to go back to the exact expression (19-1).] Thus, as a prototype for a point magnetic
dipole, we can take a very small current whirl with its magnetic moment perpendicular
to its area as shown in Figure 19-7.
Figure 19-7. . The magnetic dipole moment of a current loop is

perpendicular to the area of the loop.
19-4
E N E R G Y O F A C U R R E N T D I S T R I B U T I O N IN A N E X T E R N A L I N D U C T I O N
305
If the system consists of a number of filamentary circuits, then a dipole moment m y

can be f o u n d for the y'th one by means of (19-25) and the dipole moment of the whole
system will b e the vector sum of these terms:
r
m = m, =
j
(19-28)
In particular, if they are all plane circuits, then each n i j will have the form (19-27) and
the total will b e
m = ^ra;- =
j
(19-29)
j
Example
Ideal solenoid. In the example of Section 14-3, we saw that an ideal solenoid of N turns,
for which the pitch of the windings could be neglected, could be treated as a set of N
parallel circular current loops each carrying the current I in the same sense. Thus, if S
is the cross-sectional area, the dipole m o m e n t of each will be I Si by (19-27) where z is
in the direction of the axis. Then all of the m y of (19-29) will be parallel and the total
dipole m o m e n t of the solenoid will be N times that of a single loop, that is,
m = NSIz
= nlSIi
(19-30)
Consequently, w h e n we are far away f r o m the solenoid, the induction it produces will
be given by (19-24) with m = NSI. (This is essentially the justification for Figure 18-3.)
19-4 ENERGY OF A CURRENT DISTRIBUTION IN AN EXTERNAL INDUCTION

In Section 8-4, w e gave a fairly elaborate discussion of the energy relations involving
two distinct g r o u p s of charges ("system" and "external") and concluded that, for many
purposes, o n l y p a r t of the energy was of practical interest, namely the interaction
energy of (8-60). W e now want to consider a similar situation for the magnetic case.
We assume t h a t the currents involved can be divided into two easily distinguishable
groups. O n e of these current distributions forms a distinct physical entity (the "system"
of interest) t h a t is subject to the influences of the other current distribution (the
" e x t e r n a l " sources) that produce an induction B 0 (r) at each point of the system. Again
disregarding t h e internal energies of the two groups, we are interested in that part of
the energy t h a t we can call the magnetic interaction energy Um0. Our immediate
problem then is t h a t of finding the appropriate expression for Um0. N o w in the case of
two circuits, w e saw that the interaction energy was given by the middle term of (18-8)
and was M ^ I ^ I j
where, by (17-45),
is the flux through 1 produced by
2. Therefore, if w e a d a p t this to the present case by letting / be the system current and
4>() be the flux p r o d u c e d by the external sources, we can write the interaction energy as
U m0 =
(19-31)
We can also express this in terms of the external induction B 0 with the use of (16-6):
Um0 = l(J B 0 ( r ) da
s
(19-32)
where S is t h e a r e a enclosed by the system current.

N o w that w e h a v e (19-31) and (19-32), we can easily generalize them. If we let I j be
the current in t h e y'th circuit of our system of interest and O 0y is the flux through it
306
produced by the external sources, the interaction energy for this part will be I ft 0j so
that the total interaction energy will be
u m o = Hlftoj
j
= / , / B 0 ( r 7 ) d*j
s
j
J
(19-33)
Although we shall not need to use it here, it is worth transforming UmQ into a form
suitable for distributed currents. In order to do this, we must introduce current
elements and we can do this by using (16-23) to write the fluxes in terms of the
externally produced vector potential A 0 , and we find that
U m0 = I , $ * o ( * j ) - t j d S j
j cJ
(19-34)
We can now go from the filamentary current description to the one in terms of
distributed currents in exactly the same way we went f r o m (18-10) to (18-11). In this
way, we obtain a commonly found expression for the magnetic interaction energy
Um0 = j v J ( r ) ' A 0 ( r ) dr
(19-35)
where the integral is taken over the whole volume containing the currents J of the
system of interest.
As in Section 8-4, we will now restrict ourselves to the important case in which the
external sources are so far away and the system is so small in spatial extent that B 0 does
not vary much over the current distribution of the system. Then, as a first approximation, we can take B 0 to be a constant and take it out f r o m under the integral. Thus,
if we set every r in (19-33) equal to the same value r that is " t h e " position vector of
the system, (19-33) becomes
U mQD = B 0 (r)
= B0(r) ( X ^ S , ) = m B0(r) = m
B 0 (19-36)
where m is the magnetic dipole moment of the whole system by (19-29). Thus this
dipole interaction energy has the form of a product of the dipole moment characteristic
of the system and the external induction.
Although we will not do it because of its complexity, it is evident that we can get
further approximations to Um0 by expanding the components of B 0 in a power series,
much as we did for the scalar potential in (8-61). By analogy with (8-70), we can expect
the next term in the expansion of the interaction energy to involve products of the
spatial derivatives of the external induction with appropriately defined components of
the magnetic quadrupole moment.
If B 0 depends on position, then it is possible for a dipole m to change its energy by
moving to another point; in other words, there can be a nonzero translational force h
on the dipole. Since the system is being treated as a definite physical entity, we should
assume it to be characterized by a constant current distribution. Then the appropriate
expression to use for finding the force is the constant current expression of (18-39), and
when we put (19-36) into it, we obtain
F0
VU
M Q
V(m-B0)
(19-37)
where the derivatives are taken with respect to the components of r, the position vector
of the dipole. The fact that m is constant enables us to put this into a more useful form.
If we use (1-112), we can write (19-37) as
F D = B 0 X ( v X m) + m X ( v X B 0 ) + (B 0 V )m + (m v )Bq
(19-38)
19-4
E N E R G Y O F A C U R R E N T D I S T R I B U T I O N IN A N E X T E R N A L I N D U C T I O N
307
The first and third terms vanish because m is constant. By assumption, all of the
external sources are located elsewhere, so that there are no source currents J 0 at the
location of m. Then v X B 0 = 0 by (15-12), and (19-38) reduces to
d = (m V )B 0
(19-39)
showing us, as we suspected, that the translational force depends on the external
induction being a function of position. We note that (19-37) and (19-39) are completely
analogous to the corresponding results (8-77) and (8-79) that we found for the electric
dipole; in addition, (19-39) is consistent with the result of Exercise 14-18 in which it
was shown that there is no net force on a filamentary circuit in a uniform external
induction.
There is one place, however, in which the analogy between the electric and magnetic
cases apparently breaks down and that is in the difference in sign of the interaction
energies that are given by (19-36) and (8-73) as m B 0 and p E 0 respectively. In
fact it is commonplace to find a quantity called the energy of a magnetic dipole in an
external field which is given by
Up = m B 0 = mB0 c o s ^
(19-40)
where ^ is the angle between m and B 0 . Why is this? In essence, the use of (19-40)
represents the result of a complete change in point in view. What one is trying to do in
this case is to write the expression for the force in complete analogy to that used in
mechanics, namely, as the negative gradient of a potential energy. In other words, the
aim is to write (19-37) in the form
Vd=~VU
(19-41)
and then interpret Up as a potential energy of the usual kind. We can easily see
that the choice given in (19-40) works since (19-41) becomes F^, = v ( m B 0 ) =
V ( m - B 0 ), which leads us right back to the correct expression for the force. Hence it is
appropriate to think of Up as the energy of the dipole in this sense; Up is often called
the orientational energy in order to stress its role as a potential energy. Similarly, we can
change the signs of (19-31) through (19-35) and reinterpret them in this manner. In this
way, for example, the tendency of a circuit to readjust itself to enclose as much flux as
possible, which we found after (18-39), can be seen from the negative of (19-31), /4>0,
as corresponding to the general tendency of systems to try to attain equilibrium by
minimizing their potential energy. In Figure 19-8, we show Up as given by (19-40) as a
Figure 19-8. Orientational

energy of a dipole in an external
induction as a function of the
angle b e t w e e n their directions.
308
function of ty. The range of energy variation is finite with the minimum at ^ = 0
corresponding to the stable equilibrium state in which m and B 0 are parallel and a
maximum at ^ = 7t that describes the unstable equilibrium state for which m and B 0
are oppositely directed.
A potential energy like Up, which depends on the angle, implies the existence of a
torque T on the system. Because of the similarity of (19-40) and (8-73), we can simply
change the symbols in (8-75) to obtain the i m p o r t a n t expression for the torque on a
magnetic dipole in an external induction:
T = M X B0
(19-42)
This torque will be present even in a uniform B 0 in contrast to

again this is
consistent with the result of Exercise 14-18. We note as before that in order to have
equilibrium at an intermediate angle
this magnetic torque must be balanced by an
equal and opposite mechanical torque T M E C H so that
T+ W
= 0
(19-43)
N o w all of the results in this section have been obtained in a completely general way
and are applicable to the dipole moment of any type of current distribution of interest.
Nevertheless, it is of value to see how they can be obtained directly by considering the
forces on our prototype of a point magnetic dipole, that is, a small current loop.
Example
Small current loop in an external induction. We consider a filamentary current / in a
circle of radius a lying in the xy plane with origin at the center as shown in Figure
19-9. Its dipole moment as obtained f r o m (19-27) is
m
lira2
(19-44)
The force on the loop in the external induction B 0 is given by (14-3) as

F = / ^ s X
B 0 (r)
(19-45)
It will be convenient to work with rectangular coordinates, because of the constant unit
vectors, but we will express the components in terms of the azimuthal angle <p for ease
in integration over C. Then we have r = xx + yy a cos <px + a sin <py so that ds =
dr = ad<p( sinqox + cosqpy) and
ds X B 0 = a d<p [x J B 0z cos <p + yi? 0z sin<p z(B0xcos
cp +
sin<p)]
(19-46)
1 9 - 4 E N E R G Y O F A C U R R E N T D I S T R I B U T I O N IN A N E X T E R N A L I N D U C T I O N
309
if (1-28) is used. Since the loop is very small and B 0 is slowly varying over it, it is
appropriate to expand the components of B 0 in a power series about the origin and
keep only the terms that will give us a nonvanishing first approximation. Thus we can
write
dB,0 z
dB,0 z
dB,0 z
dB,02
o,(r) = BOz(0) + x
+ y\
~ B0M
+ a COS <p
dx
dy
dx
(19-47)
where we will remember that the derivatives are evaluated at the originthe location of
the dipoleand hence are constant as is B0z(0). There is no term z(dB0z/dz)
in
(19-47) because z = 0 since the loop lies in the xy plane. Substituting (19-47) into
(19-46), we find the x component to be
dB,0 z
(ds X B 0 ) x = a dtp I B0z( 0) cos <p + a
dx
COS <p +
dB,Oz
dy
cos <p sin <p

(19-48)
The integration around the complete circuit C required by (19-45) is attained by

integrating over <p from 0 to 2it. If we substitute (19-48) into (19-45), and use the
appropriate integrals from the set
/"2tt
f2ir .
f2n
.
I cos (ds
dB,
w,
dx
dB,
dx
(19-50)
where we also used (19-44). Proceeding in exactly the same way, we find the remaining
two components of the force to be
Fy = m2
dB,Oz
(19-51)
dy
dB,Ox
F = m,
dx
dB,0y
dy
(19-52)
These do not yet look like the components of (19-39) but we have not used all of the
information available to us. Since the external source currents are some distance from
the loop, J 0 = 0 at the loop and V X B 0 = 0 by (15-12), and we find from (1-43) that
(dB0z/dx)
= (dB0x/dz)
and (dB0z/dy)
= (dB0 /dz). In addition, V B 0 = 0 by
(16-3) so that, according to (1-42), (dB0x/dx)
+ (oB0y/dy)
= (dB0z/dz).
Substituting these relations into (19-50)-(19-52), and recalling that m has only a z component
by (19-44), we find that the force components can be written as
dB 0 x
F
Fy
F=m,
dz
dB,0y
dz
dB,Oz
dz
= (m v)B0x
= (m v)B0y
=* ( m v ) f l 0 2
310
MAGNETIC MULTIPOLES
with the use of (1-41) and (1-20). But these expressions are just the rectangular
components of F = (m v )B 0 , which is exactly (19-39).
A direct evaluation of the torque can be carried out in a similar manner. The force
on a current element is given by (14-5) as dF = I ds X B 0 so that the torque on it will
be dr = r X dF = r X (Ids X B 0 ). Summing this up over the whole circuit, we get
T = J dr = I(f)[r X (ds X BQ)]
(19-53)
F r o m (1-30), we find that the term in brackets can be written as ds(r B 0 ) - B 0 (r ds).
Since ds = dr, we have r ds = d(\r r) = d(\r2)\
if we use the first-order approximation that B 0 is constant, the contribution of the second term to (19-53) will be
- I$B0d(^r2)
= IB0j>d(\r2)
= 0 since the integral of the differential of a scalar
over a closed path vanishes as we saw, for example, after (13-4). If we substitute the
remainder of the bracketed term into (19-53), and use (1-124), we obtain
T = lj)(r
BQ) ds = if^da
X v ( r B0)
(19-54)
Since we are taking B 0 to be a constant, it is just like C in the first part of (19-11) so
that v ( r B 0 ) = B 0 . Inserting this into (19-54), taking B 0 out of the integral, and using
(19-27), we get
T = ( / /
da j X B 0 = / S X B 0 =
B0
exactly what we found in (19-42) from very general considerations.
EXERCISES
19-1 A constant current I follows a closed path
wrapped around a cylinder of radius a. In cylindrical coordinates, the position vector of a point
on this circuit is given by r = ap + b sin ntpz
where b is a constant and n is a positive integer
> 2. Find the magnetic dipole moment m of this
current distribution.
19-2 A plane circuit carrying a current 1 is
constructed in the xy plane as follows. Cylindrical coordinates are used. Starting at the origin
for (p = 0, we have p = p0rp" where p0 is a constant and n > 1. Thus a spiral is formed. This is
continued until a value <p0 of the angle is attained. Then the current follows a straight line
back to the origin. Find the magnetic dipole
moment of this current distribution.
19-3 A circular cylinder of radius a and length /
has a total charge Q distributed uniformly
throughout its volume. It is rotated about its axis
with constant angular velocity u>. Assume that the
charge distribution is not affected by the rotation
and find the magnetic dipole moment of this
system.
19-4 A dielectric sphere of radius a has a constant surface charge density a on all parts of its
surface. It is rotated about a diameter with constant angular velocity w. Assume that the charge
distribution is not affected by the rotation and
find the magnetic dipole moment of this system.
19-5 A point dipole m is located at the origin,
but it has no special orientation with respect to
the coordinate axes. (For example, m is not parallel to any of the axes.) Express its potential A at a
point r in rectangular coordinates, and find the
rectangular components of B. Show that B can be
written in the form
B(r)
4 nr
[3(w)r -
m]
(19-55)
and compare with (8-84).

19-6 Suppose the coaxial circles of Figure 17-14
are so far apart that they can be treated as
dipoles. Find their mutual inductance and compare the result with that of Exercise 17-25.
19-7 Two circular loops of radii a and b lie in
the same plane. Assume that the distance c between their centers is so large that the dipole
approximation is appropriate and find their mutual inductance.
EXERCISES
19-8 (a) A small rectangular loop of sides a and

b, carrying a current /, lies in the xy plane with
center at the origin. Find A at a point r where
r a and r b. (b) Using the result of (a), find
B at this same point, (c) A point dipole M = Mx
is on the positive y axis a distance c from the
origin where c a and c b. Find the force
and torque on M due to the loop. Express all of
your answers in rectangular coordinates.
19-9 By using (19-35) show that there is no
monopole contribution to the interaction energy
analogous to that found in (8-62).
19-10 A point dipole
is located at r, and
another point dipole m2 is at r2. Either by direct
calculation or by transcription of the corresponding results for the electric case, show that the
potential energy of m2 in the induction of m1 is
given by the dipole-dipole interaction energy
U
^mi '
~ 3^mi '
' ^)1
(19-56)
where R = r2 rx. Similarly, find the force F2 on

m2.
19-11 Two dipoles mx and m 2 are in the same
plane. The direction of
is held fixed, while m 2
is free to rotate in the plane. If m, and m 2 make
respective angles
and a 2 with R, find the
relation satisfied by the angles when equilibrium
is attained.
19-12 A circular ring of radius a lies in the xy
plane with the origin at its center. It carries a
current I -...ux uniting clockwise when viewed from
311
a point on the positive z axis. A point dipole

m = mz is on the z axis. Find the z component
of the force on m.
19-13 Consider the plane rectangular loop shown
in Figure 17-4. Assume that the external B shown
is uniform and that there is a constant current /
about the loop in the positive sense determined
by the direction of n. Find the torque that would
have to be exerted by an external agent in order
to prevent the loop from rotating.
19-14 How much work would an external agent
have to do on the loop of the previous exercise in
order to reverse the direction of n, that is, increase cp by 1 8 0 ? Evaluate this in two different
ways.
19-15 We could have begun the multipole expansion of the vector potential with the expression ( 1 6 - 1 1 ) that expresses A in terms of filamentary currents. Show that, if this is done, one again
obtains (19-21) with m given by (19-28).
19-16 By analogy with Figure 8-56, one can
devise a simple model for the prototype of a
magnetic quadrupole that consists of two small
parallel loops with currents circulating in opposite senses and that are separated by a small
distance. In particular, consider two magnetic dipoles of equal dipole moments mz located
at z = + a; the total dipole moment of this system is zero. Show that, at large distances, the
vector potential is given approximately by A =
<(>6n0ma sin 0 cos 0/AirP. Calculate the components of B and show that they are completely
analogous to the linear electric quadrupole expressions given by (8-55).
20
MAGNETISM IN THE
PRESENCE OF MATTER
U p to this point, we have assumed that when we have considered magnetic forces
involving currents the regions involved either were vacuum or contained " nonmagnetic"
conductors. Now we want to generalize our considerations to include matter of any
type. As it was in Chapter 10, it will be helpful to consider the microscopic picture of
matter as being a collection of atoms and molecules, that is, charged particles, in order
to get an idea of how to proceed.
1 MAGNETIZATION
Previously, we assumed that atoms and molecules were composed of positive and
negative charges and were, on the whole, electrically neutral. We now extend this
picture by assuming that at least some of these charges are not at rest but are in
continuous motion. Presumably these charges are moving in closed paths, the nature of
which will be determined by the resultant structure of the atomic and molecular
systems. From a great distance, these moving charges will appear as current whirls or
magnetic dipoles. Such permanently circulating currents are often called Amperian
currents since Ampere first postulated their existence in order to account for the
magnetic properties of matter. We can now envisage several possibilities.
If B = 0, it is possible that the charges are circulating in such a manner that the net
currcnt whirl vanishes, that is, the dipole moments of the individual charges, when
added vectorially by (19-28), combine to give zero. Now let us suppose that things are
arranged so that B # 0, say by using external currents. We know from (19-42) that this
will result in torques being exerted on the dipoles tending to align them in the direction
of the induction. Under the influence of these torques, it may well happen that the
paths of the moving charges will be changed enough so that in the resultant new
configuration the net dipole moment will now be different from zero. Such a moment
can be appropriately described as having been induced by the B field and we say that
the matter has been magnetized.
In the absence of an external induction, it may be that the atom or molecule already
has such a structure that the dipole moment associated with it is different from zero so
that it has a permanent magnetic dipole moment. (Although it is not essential for our
considerations now, it is of interest to note that such permanent dipoles cannot be
accounted for on the basis of classical mechanics and electromagnetism alone, but are
associated with the intrinsic angular momentum or " s p i n " of these systems that is itself
essentially a consequence of quantum mechanics. However, it is not part of our task
here to account for such permanent dipoles, but merely to accept their existence and to
describe them in a macroscopic manner so that they can be included in our basic
equations.)
Even if the material contains permanent dipoles, if B = 0, it is still possible that they
will be randomly oriented so that the total dipole moment of the piece of matter
involved is zero, that is, it is unmagnetized. If, now B # 0, there will be a torque on
312
20-2 M A G N E T I Z A T I O N C U R R E N T DENSITIES
313
these dipoles that will tend to rotate them into alignment with the resultant induction.
Generally, this aligning tendency will be offset by the randomizing processes associated
with thermal agitation, such as collisions, but we can still expect that the net effect will
be to produce a net dipole moment in the direction of the field and the material will
again have become magnetized.
Some materials have the property that, even in the absence of a B field, the
permanent dipoles are at least partially aligned and the material is said to be
permanently magnetized or to be a permanent magnet. Such cases are generally of more
technological importance than the corresponding electric cases.
As we saw in the last chapter, we can reasonably expect the molecules, especially
when distorted by the presence of B, to have higher-order multipole moments whose
effects would be described by (19-3). However, since the contributions of these
higher-order terms to A and B drop off much more rapidly with distance and depend
on angles in a more complicated way than those of dipoles, we assume that, for the
purposes of describing the average properties of matter, the dominant features of
interest to us are those associated with the dipole moments. Thus, all of our considerations lead us to the following:
Hypothesis
As far as its magnetic properties are concerned, neutral matter is equivalent to an

assemblage of magnetic dipoles.
We now have to put this hypothesis into quantitative terms. For this purpose, we
define the magnetization M as the magnetic dipole moment per unit volume, so that the
dipole moment dm 'm a. small volume dra.tr will be
dm = M(r) dr
(20-1)
Thus the total dipole moment of a volume V of material will be

m
total = / ^ M ( r )
dT
(20-2)
From its definition and (19-20), we see that the unit of M will be 1 ampere/meter.
As usual, the definition (20-1) implies that (/T is large enough to include enough
material so that M can be regarded as a smoothly varying function of position; at the
same time, d r must be small on a macroscopic scale. Because of the way in which we
were led to introduce M, we expect that there will be some functional relation between
M and B. We will consider this in more detail later, but for now, we take M as part of
our macroscopic description of the material and we want to investigate the consequences of its existence.
2 0 - 2 MAGNETIZATION CURRENT DENSITIES
Let us assume that we have a magnetized object

potential produced by it at a field point r outside of
The dipole moment of the volume dr' as given by
contribution to the vector potential at r as obtained
[i0 M ( r ' ) X R dr'

r
(-^-J)
4tt
R
4tt
R
where, as usual, R = r r ' and corresponds to the r of (19-21) since m there was
dA =
Mo dm! X R
and we want to find the vector

the body as shown in Figure 20-1.
(20-1) is dm/ = M ( r ' ) dr' and its
from (19-21) is
314
M A G N E T I S M IN THE PRESENCE O F M A T T E R
S'l
Figure 20-1. Calculation of the vector

potential at a field point outside a
magnetized object.
assumed to be at the origin. The total potential is obtained by integrating (20-3) over
the volume V of the material and is
ju0 M ( r ' ) X R dr'
r
A(r)
fi0
I 1 \
M(r,)
*'(*)'*'
(20
-4)
with the use of (1-141). W e can now use (1-118) to rewrite the integrand as
v' X M
/ 1 \
M
' U r ~ s
/ M
v
'
(20-5)
( *
If we now substitute this into (20-4) and use (1-123) and (1-52), we get
AM -
4 77" JI
v'
Oil. f
Air Ji/'
'y
( V ' X M ) dr'
R
( V ' X M ) dr'
Mo
47T
M X n' da'
Mo
^
47T
RR
477-^5'
- V' X
M
~R
dT'
(20-6)
where S' is the surface bounding the volume V of the material and n' is the outward
normal as shown in Figure 20-1. U p o n comparing this with (16-12) and (16-13), we see
that this is exactly the vector potential that would be produced by a volume current
density Jm distributed throughout the volume and a surface current density Km on the
bounding surface where
Jm = V ' X M
K
(20-7)
= M X n'
(20-8)
for then we would have

A(r)
M0 f J m l r M
477 J v
R
dT
MO Km(r>) da'
4v ^s'
R
i.
(20-9)
as we would expect. (Compare with Equations 10-7 through 10-9.)

W h a t we have found, therefore, is that, as far as its effects outside itself are
concerned, the material can be replaced by a distribution of volume and surface
current densities that are related to the magnetization M by means of (20-7) and (20-8).
315
The various steps which have led to our present conceptual scheme are summarized and
outlined in Figure 20-2. (Compare to Figure 10-2.) The total vector potential at the
field point will then be given by (20-6) plus that due to any other currents which might
be present.
It is common practice to omit the primes on (20-7) and (20-8) and to simply write
Jm = V X M
and
Km = M X n
(20-10)
with the understanding that the differentiation is with respect to source point coordinates and n is the outward normal. We note that M X n is tangent to the surface since
it is perpendicular to n, as it should be to represent a surface current; this also shows
that it is the tangential component of M that leads to the surface current. (It is essential
not to forget that K m is determined by the value of M X n at the surface.)
The subscript m on these quantities reflects the fact that they are usually referred to
as magnetization current densities. They are also often called Amperian current densities
or bound current densities', we will not use the latter term, however, since in (12-18) we
have already associated it with the displacement of bound charges as reflected in the
polarization current density.
We have obtained these results by making a formal comparison of the expression
(20-6) with its general form (20-9). It is also possible to understand and calculate these
currents in a direct " physical" manner; we will, however, do this only qualitatively. As
an extreme example, let us consider a piece of material with uniform magnetization as
produced by dipoles of the same magnitude all aligned, that is, circulating in the same
sense as shown in an end-on view in Figure 20-3. If we consider the immediate
Matter
Assemblage
of d i p o l e s
described
by M
Equivalent volume
and surface
current densities
J m and K m
Figure 20-2. Conceptual scheme for replacing matter by equivalent

current densities.
Figure 20-3. Origin of

magnetization surface
currents for a uniform
magnetization.
316
M A G N E T I S M IN T H E P R E S E N C E O F M A T T E R
neighborhood of an interior point, as indicated by the dashed line, we see that the
current due to one whirl in one direction is canceled by the oppositely directed currents
in adjacent whirls. Thus, in the interior of a uniformly magnetized material, the
magnetization current is zero in agreement with (20-7). However, at the surface there
are no adjacent currents to produce a cancellation, and, since the currents in the whirls
are all circulating in the same sense, the result in effect is that of a current K m
circulating on the surface as indicated by the arrow. Since M is directed out of the page
in this case, the direction of K m as given by (20-8) is exactly the same as that just
deduced by consideration of the current whirls.
Now let us suppose that M is not uniform throughout the material; for example, the
current of each whirl could be different. Then, if we reconsider the region enclosed by
the dashed line, we see that there will no longer generally be a complete cancellation by
the oppositely directed adjacent currents and there will be a resultant current in the
interior. This is exactly the effect described by (20-7) and, in fact, this picture can be
used in a quantitative manner as an alternative way of deriving J m = v ' X M. There is
still a surface current K m in this case of nonuniform M of course.
Suppose that we have two magnetized materials meeting at a common boundary as
illustrated in Figure 20-4. Since each of them has a tangential component at the surface,
they will each produce a surface current given by (20-8). The net surface current density
will be the sum of these two terms so that
Km = K, + K2 = M , X
+ M 2 X ft'2
(20-11)
where
and n'2 are the outward normals to the respective media. If we now introduce
the normal n drawn from 1 to 2 by our usual convention, we see from the figure that
n', = n and n ' 2 = n so that (20-11) becomes
Kmnet = ( M > - M
) x n = n x (M2 - M,)
(20-12)
As an illustration of the internal consistency of our results, let us consider a piece of

magnetized material of finite volume and find the total rate at which bound charge is
being transferred across a plane passing through the material as shown in Figure 20-5a.
Let S be the area of the plane intercepted by the material, n the normal to it, and C its
bounding curve as shown in (b) of the figure. The total rate at which charge is being
transferred across the plane, that is, from the shaded part of the volume on the right to
the unshaded volume on the left, is given by (12-6), (12-8), and Figure 12-5 as
^
= f j
da + <f>Km I ds
(20-13)
where I is drawn perpendicular to ds as required. Substituting (20-10) into this, and
M,
2
i i "1
1
FM,
2
Figure 20-4. Boundary between t w o
magnetized materials.
317
Plane
(b)
(a)
Figure 20-5.
plane.
Calculation of total bound current passing across the
using (1-67) and (1-29), we find that

Y
=</>M ds +<^j(M X n') -lds=<j)
M ds + ^ > M - (n' X i) ds
(20-14)
where n' is the normal to the surface of the material at the point where K m is to be
evaluated. But we see from the figure that (n' X t) ds = ds so that the two integrals in
(20-14) cancel and dq/dt = 0. Thus the magnetization currents do not transfer net
charge. This is just what would be expected, for otherwise a piece of magnetized
material would "spontaneously" have its charges separated, which is incompatible with
our picture of magnetization as arising from a reorientation of existing current whirls.
The induction produced outside the material can be found from (20-6) by using
B = v X A. Alternatively, we can use the current densities as found from (20-7) and
(20-8) in (14-7) and (14-11) to give
,
w
M. r
AwJv
ft,
R
, K>')xR-fa4tt Ts>
R2
}
7
So far all of our results have been obtained by considering the vector potential, and
its corresponding induction, at a field point in the vacuum region outside the matter.
Here there is no difficulty in knowing what B is and we could imagine measuring it by
finding the force on a moving point charge or the torque on a small current loop. What
about the situation inside the material? We cannot easily measure the torque on a
current loop without first drilling a hole into the material in order to insert the loop and
we can anticipate that this may alter the situation since in the process we will certainly
remove some volume currents and introduce some new surface currents. We faced a
similar problem in the electric case and discussed it at length in Section 10-3. We could
go through the same arguments that we used there, with appropriate changes of terms
and symbols, and we will be led to exactly the same conclusion: the only practical and
reasonable thing to do is to be consistent with our conceptual replacement of the
matter by the set of equivalent magnetization current densities and assume by definition
that (20-9) and (20-15) can be used to calculate A and B anywhere.
Nevertheless, we can still ask if it is possible to give a cavity definition of B in the
same sense that we did for E in (10-26) and D after (10-45). In other words, we want to
be able to cut a hole of appropriate shape within the material so that we can insert, say,
318
Figure 20-6.
Cavity used t o measure B in the material.
n
Km
M
Figure 20-7. Uniformly
magnetized cylinder, (a)
Side view. (b) End view.
a small current loop into the cavity and find the value of B in the material from a
measurement of the torque on the loop in the vacuum inside the cavity. As before, we
can use our knowledge of the boundary conditions satisfied by B; the relevant one is
that the normal components of B are continuous as given by (16-4). But this is precisely
the same boundary condition that we used to obtain the cavity definition for D after
(10-45) so that we see exactly what is to be done. The appropriate cavity is thus a short
right cylinder cut in the material with its base perpendicular to the direction of B there
as shown in Figure 20-6. Since, by construction, only normal components are involved,
we see that the value of B in the cavity equals that in the material, that is, Bc = B.
Finally, let us consider a simple example illustrating our results so far. We discuss a
more complicated one in the next section.
Example
Infinitely long uniformly magnetized cylinder. Suppose we have an infinitely long
cylinder of circular cross section that has a uniform magnetization M parallel to the
axis of the cylinder as shown in Figure 20-7a. We can find the magnetization currents
from (20-10). Since M = const., we find that K = V X M = 0 and there are no
volume currents. Figure 20-7 shows an outward normal n, which is perpendicular to M,
and we see that the surface current has a constant magnitude Km = M and circulates
about the cylinder in the sense shown in (b) of the figure. But a surface current like this
is exactly the equivalent of an infinitely long ideal solenoid as illustrated in Figure
15-11 and for which we have already calculated the induction. The magnitude of B
inside is given by (15-24) and therefore in this case we have Bt = n0Km = ii0M = const.,
while B0 = 0 outside. Since B( is in the axial direction, as is M, we can finally write
B, = fi0M
as the complete solution to the problem.
(20-16)
20-3 U N I F O R M L Y M A G N E T I Z E D SPHERE
319
2 0 - 3 UNIFORMLY MAGNETIZED SPHERE
Let us now consider a sphere of radius a that has a constant magnetization M. We

choose the z axis in the direction of M and the origin at the center of the sphere as
shown in Figure 20-8; thus M = Mi. Since M is constant, J m = 0 by (20-7). We see
from the figure that the outer normal n' = ?' and therefore the surface current density
as found from (20-8) is
K
= M z X ?' = M sin 6
(20-17)
with the use of (1-94) and (1-92). Thus, in this case, the surface currents are along the
lines of "latitude" on the sphere with a magnitude that is zero at the "poles" and a
maximum at the "equator" as shown in Figure 20-9. If we insert (20-17) into (20-15),
and use (1-100) to write da' = a 2 sin 0' d0' d(p', while R = R/R, we find B to be given
in general by
f-v (<p' X R ) sin 2 8' d0' dcp'
477"
'O M)
RJ
(20-18)
For simplicity in this example, we evaluate (20-18) only for a field point on
the positive z axis; later, we will solve this problem completely by a quite different method. We see from Figure 20-10 that r = zz, r' = ar', R j.z or', R =
( z 2 + a2 2za cos 6')l/2
and <p' X R = z sin 6'r' + (z cos 6' o)0' = a sin 0'z +
(z a cos 0')(xcos <p' + y sin <p') with the use of (1-94), (1-92), and (1-93). If we insert
these into (20-18), we see that the x and y components will vanish on integrating over
cp' because of (19-49) so that B has only a z component, as is evident from the
symmetry of the current distribution shown in Figure 20-9. Therefore, the only
nonvanishing component is
(z) =
H0Ma4TT
/JQf2 /fQ
J
sin 6' d6' dcp'

( z 2 + a2 2za cos 6')3/2
(20-19)
The integration over <p' gives a value of 2v; if we use sin2 0' = 1 cos 2 0', and change
/k
Figure 20-8.
sphere.
/ \
Uniformly magnetized
Figure 20-9. Equivalent

surface currents of uniformly
magnetized sphere.
320
Figure 20-10.
on the axis.
Calculation of the induction at a point
the variable of integration by means of (2-22), we find that this becomes

Bz{z)
lx0Ma3
=
fi
/'
^-- i
J
(l - ii2) dn
(20-20)
( z 2 + a2 2za[i)3//2
The integral can be found with the use of tables to be

2 ( z 2 + a2 2zan)
3 z3373
a
1/2
z2 + a2 + zafi +
3z2a2([x2 - 1)
2 ( z 2 + a2 2zafi)
-l
{(z 2 + a2)[\z + a\ \z a\\ za[\z + a\ + |z a\\)

(20-21)
3z3a3
and, as usual, there are two cases to be considered.
1. Outside the sphere. Here z > a, so that \z a\ = z a and (20-21) becomes
4 / 3 z 3 which, when put into (20-20), gives the induction outside the sphere to be
2unMa3
= lz,
<20-22)
This result becomes more understandable if we express it in terms of the total dipole
moment of the sphere that from (20-2) is
rn _ 4Tra M
(20-23)
un 2m
* ( * ) = T2" 3
477 Z
(20-24)
so that (20-22) can also be written as
Upon comparing this result with (19-24), and remembering that 8 = 0 and r = z for a
field point on the z axis, we see that (20-24) is exactly that of a point dipole of total
moment m.
2. Inside the sphere. Here z < a, so that |z a\ = a z and (20-21) becomes
4 / 3 a 3 which, when put into (20-20), gives the induction inside the sphere to be
Bzi(z)
= i/i0 M
(20-25)
which is constant and parallel to the magnetization. The numerical factor of f is
2 0 - 4 THE H FIELD
321
/ ,
Figure 20-11. Magnetic induction on axis due to a

uniformly magnetized sphere.
different from unity as found for the cylinder in (20-16) even though both of these
systems have uniform magnetization; the difference in numerical factors is evidently
due to the difference in geometry.
It will be left as an exercise to show that these same results also hold for negative
values of z, that is, Bzo is always in the positive direction and Bzi is constant and given
by (20-25) for all z. In Figure 20-11 we illustrate these directions for B that we have
found. W e recall that these results are similar to those we found for E for the uniformly
polarized sphere in Section 10-4. This leads us to suspect that similar things can be said
about this problem, that is, that the induction everywhere outside is a dipole field
corresponding to the total dipole moment given by (20-23), while everywhere inside, B,
is constant and given by f j u 0 M . This actually turns out to be the case as we will show
in a quite different manner later in this chapter.
At the surface of the sphere (z = a), (20-22) gives Bzo(a) = /u0M = Bzi{a)\ this is
as it should be since both of these are normal components and should be equal by
(16-4).
20-4 THE H FIELD

We recall that when we defined the magnetic induction in Section 14-1, we pointed out
that B, by its definition, is determined by currents of all kinds. Thus, in v X B = ju 0 J,
given by (15-12), J represents the total current density. In (20-10), we found a current
density Jm = V X M, which is associated with the presence of matter and, as we found
it useful to do for charges in Section 10-5, it is convenient to divide currents arising
from moving charges into the two broad classes of magnetization currents and free
currents described by the respective densities J m and
The magnetization currents are
to be associated with the constituents of matter and generally speaking we do not have
322
any real control over them. As we discussed near the end of Section 12-2, free currents
are those over which we can exert some control, for example by sending currents
through wires with the use of batteries or by using convection currents in the form of
streams of charged particles. Thus we can write the total current density as the sum of
these two:
Jfotal = J
J/ +
(20-26)
Inserting this into (15-12) and using (20-10), we find that V X B = n 0 J = ^ 0 ( J /

V x M) or
vx
K~mH
(20-27)
The form of this equation, in which only the free current density appears on the
right-hand side, suggests that it will be useful to define a new vector field H(r) by
H =
B
Mo
(20-28)
for then (20-27) becomes

V X H = Jf
(20-29)
The vector H is called the magnetic field or sometimes the H field. The principal
characteristic of H and the primary reason for its introduction is that its curl depends
only on the free current density. The dimensions of H are the same as those of M and
thus H will be measured in ampere/meter. We can think of (20-29) as expressing
Ampere's law of force between current elements plus the magnetic effects of matter.
Now that we have defined H we can easily determine some of its properties. The
behavior of its tangential components at a surface of discontinuity in properties can be
found by inserting the source equation (20-29) into (9-13) and (9-18) and using the
analogue of (15-14) for free currents; the result is that we can express the boundary
conditions in the two equivalent ways
n X ( H 2 H x ) = K7
H2r-H1, = K/Xn
(20-30)
(20-31)
where
is the free surface current density. Thus, there will be a discontinuity in the
tangential components of H only if there is a free (e.g., conduction) surface current
density. This is in contrast to B whose tangential components will be discontinuous if
there is a surface current density of any kind as we saw in (15-15) and (15-16).
The integral form of Ampere's law for H can be found by using (20-29), (1-67), and
(12-6); the result is that
ds = j j
da = If<enc
(20-32)
where If enc is the net free current passing through the surface S enclosed by the
arbitrary path of integration C.
Equation 20-29 can be used to devise a cavity definition of H so that one can
contemplate finding the value of H in the material from measurements done in the
vacuum inside an appropriately shaped cavity. We assume that there are no free
currents within the material. Then v X H = 0 and its tangential components are
continuous by (20-31). This is exactly the type of boundary condition that was the basis
of the cavity definition of E given by (10-26) and we see at once what we should do. As
shown in Figure 20-12, we imagine a long needlelike cavity cut into the material with its
2 0 - 4 T H E H FIELD
Figure 20-12.
323
Cavity used to measure H in the material.
axis parallel to the direction of H there. Thus, by construction, only tangential

components are involved, and since they are continuous, the magnetic field in the cavity
H ( . equals that in the material, H c . = H. Now the induction Bc. in the cavity can be
measured by determining the force on a moving charge or the torque on a small current
loop. Since there is no matter in the cavity, M c = 0, and we see from (20-28) that
H , = Bc/(i0 and therefore, the field in the material is given by
B
H = Hc =
(20-33)
Mo
As the Helmholtz theorem of Section 1-20 tells us, we need one more differential
source equation for the magnetic field, that is, its divergence. This is easily found by
combining (20-28) with (16-3) and we see that V B = 0 = V [m 0 (H + M)] =
/ x 0 ( v H + v - M) and therefore
V H = - V M
(20-34)
showing us that H can have sources associated with the Amperian currents of matter as
well as free currents. We will return to this point later.
The boundary conditions satisfied by the normal components of H can be most
easily obtained from the fact that the normal components of B are continuous, and
when we substitute (20-28) into (16-4) we find that
n (H2 - HO = - n
(M2 - M,)
(20-35)
or
Hln-Hln=
{M2n Mln)
(20-36)
In spite of the apparent simplicity of our results, particularly (20-29), they are not
too useful at this point since we have to be able to relate the three vectors B, M, and H.
Nevertheless, we can look again at some of our previous examples.
Example
Infinitely long ideal solenoid. Here we have free currents producing the fields. Since
there is a vacuum everywhere else, M = 0 and H = B//x 0 . Thus we can use our
previous results (15-25) and (15-26) and we see that H 0 = 0 while inside the solenoid
B,
H,- = = niz
(20-37)
Mo
where n is the number of turns per meter and z is along the solenoid axis. (This result is
basically the reason that H is often given in the units of ampere-rwrm/meter.) Here we
have a discontinuity in the tangential components of H; if we take n to point from the
inside (1) to the outside (2), and write Kf = nl$ from (15-22) and Figure 15-11, (20-31)
becomes H 2 ; H u = 0 H , = /<p X n = niz and therefore H, = niz in agreement with (20-37); we note that the use of the boundary conditions in this way provides
us with a quick way of finding H , in this case.
Example
Infinitely long uniformly magnetized cylinder. Here there are no free currents. We solved
this problem for B above with the result that B 0 = 0 while B, = ju,0M. Outside the
324
cylinder, M 0 = 0 and therefore H a = 0. Inside the cylinder, we find from (20-28) that
H , = (B,/ju 0 ) M , = M M = 0. Thus H = 0 everywhere in spite of the resemblance of this case to the ideal solenoid as we found before. The real difference between
these two examples is that the first one has free currents while the second one does not.
Since M is uniform, V M = 0 and V H = 0 by (20-34); M does not have any
normal components either so that our results are also consistent with (20-36). [If the
cylinder were not infinitely long, there would be a discontinuity in the normal
components of M at the ends where the surface separates the matter from the vacuum.
Then, from (20-36), we can expect there would be sources of H and that H + 0 for a
cylinder of finite length.]
Example
Uniformly magnetized sphere. There are also no free currents in this example. In the
previous section, we found the values of B along the axis by a direct calculation from
the magnetization currents. We can use these results to find H by means of (20-28).
Outside the sphere, M c = 0 and (20-24) and (20-22) give us
H
B
= T r
1 2m
^
^
2 Ma3
-
(20
-38)
Inside the sphere, M , = Mi and (20-25) yields

Hzi{z)
= Mo
- \M
(20-39)
Thus, in this case, H =# 0 both inside of and outside of the sphere and, in fact, is
oppositely directed to both B and M inside the sphere as illustrated in Figure 20-13. At
the surface of the sphere (z = a), (20-38) gives Hzo(a) = M so that there is a
H0
Figure 20-13. Magnetic field on axis due to a uniformly

magnetized sphere.
2 0 - 4 T H E H FIELD
325
discontinuity in the normal components of H. But if we take n = z in this case, (20-36)

becomes H2n - Hln = Hzo(a) - Hzi(a) = fM \M) = M = - ( 0 - M) =
Mln) exactly as it should.
Both of these last two examples were characterized by the absence of free currents
and the presence of a uniform magnetization, yet in one case H = 0 while H = 0 for
the other. What is the real difference between them? Although v M = 0 in the interior
for both of these examples, they do differ in that there are clearly discontinuities in the
normal components of M for the sphere, while, as we remarked above, this is not the
case for the infinitely long cylinder, and the fact that H # 0 for the sphere must
somehow be related to this difference; we would like to be able to describe this
somewhat more systematically and conveniently than by (20-36) alone. In addition,
there is the nagging resemblance between these examples and those we discussed for the
electrostatic case in Chapter 10, particularly for the uniformly magnetized and polarized
spheres, and one feels that these problems are somehow related. This is, in fact, the
case.
In the absence of free currents, our results (20-29), (20-31), (20-34), and (20-35)
become
V X H = 0
V H = v M
H2t-Hlt = 0
n-(H2-H1) =
fl"M1
^ " >
where, for simplicity, we take region 2 to be a vacuum so that M 2 = 0. Similarly, in the
absence of free charge, our electrostatic results (5-4), (9-21), (10-39), (10-10), (9-26), and
(10-12) can be written in the form
VX(e0E)0
.. v ( e
( oE2r) ~~ ( c o E i , ) = 0
E ) = - V - P = Pb
n ( e 0 E 2 e 0 E i ) = n Pi o b
We also recall that P and M each represent their corresponding dipole moment per unit
volume. U p o n comparing these two sets of equations, we see that by analogy we can
define magnetic analogues of charge densities by
P m
=-V- M
am = n
(20-42)
so that two of the equations of the set (20-40) become

V H = pm
n - ( H 2 - H 1 ) = am
(20-43)
and now look just like the corresponding members of (20-41). Thus we can call pm the
volume density of magnetic charges ("poles") and om the surface density of magnetic
charges ("poles") and regard them as the source of the magnetic field H and they will
play the same role as did the corresponding electric charges in the calculation of f 0 E .
As we saw in Section 16-1, there are actually n o magnetic charges so that the densities
defined by (20-42) represent fictitious charges', however, this does not prevent them
from being useful in discussing this restricted type of problem.
This analogy can be pushed even further. W e recall that the fact that v X E = 0
followed from our being able to write E in terms of the electrostatic scalar potential in
(5-3), that is, E = V<t>. Similarly, the fact that V X H = 0 here enables us by using
(1-48) to introduce a magnetic scalar potential <pm so that
H =
(20-44)
If we now substitute this into (20-43) and use (1-45), we find that <t>m satisfies Poisson's
equation
V
^ = - P
(20-45)
326
while, in regions where pm is zero, it satisfies Laplace's equation

V 2<$>m = 0
(20-46)
2
in analogy with (11-1) and (11-3), if the former is written as v (t0<P) = ~p = ~ph
since we are only considering cases in which
= 0. The uniqueness theorem of Section
11-1 that we found for <j> will also hold for <m, since the only requisite was that the
function involved satisfy Laplace's equation.
Thus what we have found is that many magnetostatic problems can be solved exactly
as we did some electrostatic problems, and therefore many of the methods we
developed in Chapter 11 can also be applied here. Furthermore, if we have already
solved the analogous electrostatic problem, we can simply take over the solution by
making the replacements e 0 E - H, e0m, P - M, and so on. For example, we
can write down the integral from which <j>m can be found from a given distribution of
magnetization by changing (5-7) and (5-8) in this way:
, v
~
=
f Pm( r ')
477 K'
1
dr
477 JV'
'
+
V ' - M J T '
CT
477 JS'
1
m(r') da'
+ /
477 JS'
( n ' M ) da'
(20-47)
Example
Uniformly magnetized sphere. Let us reconsider this example from this new point of
view. Since M = Mi = const., pm = V M = 0 and there is no volume pole density.
However, since n in (20-42) is the n' = r ' of Figure 20-8, we see that there is a generally
nonzero surface pole density given by
am Mi r' = Mcos 6'
(20-48)
and we see that this is exactly of the form found in (10-27) for the bound surface charge
density of the uniformly polarized sphere, and which is illustrated in Figure 10-9. Now
we can see that the sources of H in this example can be ascribed to the discontinuities
in the normal components of M at the surface of the sphere and, on comparing Figures
20-13 and 10-11, we can see why H , is oppositely directed to M in the interior. In the
last example of Section 11-5, we found the complete solution (11-133) for the uniformly
polarized sphere, and we found that the electric field everywhere outside was a dipole
field corresponding to the total dipole moment of the sphere, while it was uniform
everywhere inside and given by (11-134) as E, = ( P / 3 e 0 ) . Now we see that exactly
the same things can be said about the uniformly magnetized sphere: outside, the field is
a dipole one corresponding to the moment of (20-23) and the magnetic field vectors
everywhere inside will be constant and equal to
B, =
H;. = - } M
in agreement with the results (20-25) and (20-39) found for points on the axis.
(20-49)
As we saw in (11-135), many internal or local electric fields turn out to be

proportional to and opposite to the polarization. It is clear that similar results will be
found in magnetostatic problems of similar type, and, accordingly, it is common
practice to write
H J o c = jVmM
(20-50)
where Nm is a dimensionless constant called the demagnetizing factor. We see from
(20-49) that Nm = j for a sphere. It will be left as exercises to show that Nm = 1 for an
20-5
LINEAR I S O T R O P I C H O M O G E N E O U S M A G N E T I C M A T E R I A L S
327
infinite slab with parallel faces and Nm = \ for a cylinder, exactly as in the electrostatic
case.
Historically, the approach to magnetostatic problems that we have just outlined and
illustrated was quite extensively applied so that poles were used in many calculations;
this is essentially the way in which H got its early emphasis and acquired the name
magnetic field, which one might reasonably have expected to have been given to B.
However, it is important to remember that the use of poles works consistently only in
the absence of free currents and, of course, there is no experimental evidence for the
existence of magnetic poles. From our present-day point of view, all magnetic effects
are ultimately due to currents, and the use of poles in the way we have illustrated is a
sometimes useful fiction, and applicable only in limited cases, generally in connection
with calculations involving permanent magnets. In principle then, any problem that can
be worked by using poles can be done by using magnetization currents, as we saw in
several examples; it may seem harder, but it can be done. (In spite of the remark above
about the necessity of the absence of free currents, it is possible to extend the use of the
magnetic scalar potential to cases involving filamentary free currents. It turns out,
however, that <f>m then depends on the solid angle subtended by the current at the field
point. Since solid angles can be multiple-valued functions, this causes problems in the
use of 4>m, and it quickly loses much of the simplicity implied above. Consequently, we
will not pursue its use any further.)
Let us now consider one more simple example that will lead us naturally into our
next major topic.
Example
Infinitely long straight constant current. We let the free current / coincide with the z axis
and be in the positive z direction. We assume initially that there is no matter present.
By the familiar symmetry arguments, H will have the form H = Hv(p)
and we can
use Ampere's integral form (20-32) by integrating over a circle of radius p in the xy
plane. Thus,
ds = 2<npH= I< enc = I s o that Hv = I/Imp and therefore
H = -Jq,
2 77p
and
B =
2 irp
(20-51)
since M = 0. Now, as an extreme example, let us suppose that all space is filled with an
obviously magnetic material such as iron. (We need to assume the presence of a thin
insulating material around the wire since iron is a conductor.) What will happen? If we
keep the free current I unchanged, and if the material is homogeneous enough so that
the axial symmetry is still present, then we can apply (20-32) in exactly the same way
and we will get precisely the same value for H as given by (20-51). In other words, in
this case, H is unaffected by the presence of the matter. We can, however, expect that
M will now be different from zero and then (20-28) shows us that the value of B will be
altered, considerably as a matter of fact. It is also clear that we cannot evaluate B
exactly until we know how the magnetization depends on the fields, and it is this that
we now need to consider.
20-5 LINEAR ISOTROPIC H O M O G E N E O U S MAGNETIC MATERIALS

If we recall the manner in which we were led to introduce the magnetization in Section
20-1, we should assume the existence of a functional relation between M and B, that is,
we would write M = M(B) and expect that the exact form of this relation would
depend on the material and will have to be found by experiment. As logical as this
328
might seem, it is not what is usually done; instead, one begins by writing a relation
between M and H, that is, M = M(H). There are essentially two reasons for this. One is
primarily historical, in that at first greater importance was attached to H than to B and
this seemed to be the reasonable thing to do. A much more practical reason is
illustrated by the previous example where we saw that H was not altered in the
presence of matter provided that we kept the free current constant and that we had an
appropriate geometry. Since this is a desirable situation to have in the laboratory where
one can alter the free current at will, it is convenient to consider M to be a measurable
function of H.
There still remains the problem that the functional form of M(H) must be found for
each material, either by experiment or by calculations done in other branches of
physics. Fortunately, as was the case for dielectrics as discussed in Section 10-6, it is
possible to classify most magnetic materials into groups in a way that can be used to
simplify our theory and to make it more useful.
If H = 0, and M(0) = 0, then the material is magnetized even in the absence of an
external field. It is said to have a permanent magnetization and is called a permanent
magnet. Many materials for which M(0) = 0 have a dependence of M on H that is very
nonlinear and sometimes the relation is not even single valued. Many such materials are
of very important practical interest, and, in fact, very often the materials that can fall
into the first class also fall into the second. We do not consider these cases further in
this section, but will come back to them in Section 20-7.
Rather than repeating the various simplifying classification steps outlined in Section
10-6, we proceed immediately to the simplest possible case, that of a linear isotropic
homogeneous magnetic material for which the magnetization is proportional to and
parallel to the magnetic field so that we can write
M =
Xm
(20-52)
Here
is called the magnetic susceptibility and is a constant characteristic of the
material. For virtually all materials that fall into this category, |x J 1, and, in
contrast to the electric case, xm c a n be of either sign. If xm >
the material is called
paramagnetic, while if x m < 0, it is said to be diamagnetic. As is discussed in detail in
Appendix B, all materials have a diamagnetic contribution to their susceptibility
arising from the altered orbital motion of their constituent electrons that is produced
by an applied field. In paramagnetic materials, however, this is overwhelmed by the
much larger paramagnetic susceptibility due to permanent magnetic dipole moments.
Combining (20-52) with (20-28), we obtain
B
= MoC1 + X m ) H = Kmn 0 H =
(20-53)
where
KM = 1 + XM
relative permeability
fx = Kmju.0 = (absolute) permeability
(20-54)
(20-55)
Thus, in this common, but not universal, case, B and H are parallel. The relation
B = jiiH, or H = B/jti, is another example of a constitutive equation and is not a
fundamental equation of electromagnetism, in contrast to (20-28) which is.
In this case, we can easily relate M and B by eliminating H between (20-52) and
(20-53) and we find that
M =
Xm
-B = t
,
B
1
KmMo
I + XmMo
so that M is also a linear function of B.
(20-56)
20-5
LINEAR I S O T R O P I C H O M O G E N E O U S M A G N E T I C MATERIALS
329
Since ju. is a constant, we see that V - B = 0 = V - (/xH) = ju,v H so that

V H = 0
(20-57)
V M = 0
(20-58)
and, therefore, from (20-52):

which is, of course, consistent with (20-34) and, according to (20-42), shows that there
can be no volume density of magnetic charges within a l.i.h. magnetic material.
The free and magnetization current densities are also simply related in a material
such as this. If we take the curl of (20-52) and use (20-10), (20-29), and (20-54), we find
that
J m = X m J f = ( * m - 1)J,
(20-59)
and the total current density as given by (20-26) becomes

J = (1 + X J J / = K J /
(20-60)
Since x m c a n be of either sign, we see that the magnetization current can be oppositely
directed to the free current; the total current, however, will always be in the same
direction as the free current since KM is positive. We also see that when L = 0, both Jm
and J are zero, so that at any point in a l.i.h. magnetic material where there is no free
current, there is no magnetization current either. If Jy = 0, then v X H = 0 which,
together with (20-57), shows us that there are no sources of H within the material in the
absence of free currents.
When (20-53) is applicable, the boundary conditions at a surface of discontinuity in
magnetic properties can be written in terms of a single vector, either B or H as one
chooses. Using (16-4) and (20-30), we can write them as
n- (B 2 B j ) = 0
fix
/ B,
\ 1*2
B, L\
= Kf
Mi /
(20-61)
or, equally well, as

N ( J U
J ^ H J )
= 0
n x ( H
- H
= K/
(20-62)
We see from these relations that, even if K f = 0, the lines of B (and H) will generally
have different directions on the two sides of the bounding surface, that is, they will be
refracted at the boundary.
Example
Infinitely long ideal solenoid. As usual, we assume the solenoid has n turns per unit
length, a cross section of area S, and a (free) current I in its winding. We now assume,
in addition, that the interior is filled with a l.i.h. magnetic material of relative
permeability Km. In this case, it would be appropriate to use the integral form of
Ampere's law for H given by (20-32). We could proceed with this equation exactly as
we did in the example of Section 15-2, and we will again obtain (20-37), that is, the
magnetic field is uniform across the cross section and is given by
H = nil
(20-63)
while it is zero outside. Thus, H is unaffected by the presence of the matter. The
induction B is now given by (20-53) and is
B = Kmn0U
= Kmii0nlz.
(20-64)
and has been increased by the factor Km as we see by comparison with (15-26).
330
Consequently, the flux per turn given by BS has been increased by this same factor and
will equal Kntfx0nSl. If we now consider a length /, with its nl turns, the total flux will
be KmfxQn2Sll, so that the self-inductance L = O / / , as given by (17-55), will be
L
o" 2 lS = K m L 0
(20-65)
where we have used (17-61) to identify the self-inductance L0 when the interior has a
vacuum in it. Thus, we see that the presence of the matter has increased the self-inductance and the ratio L/L0 = KM is just equal to the relative permeability of the material.
Example
Self-inductance in general. The simplicity of the final result L = kmL0, which we have
just obtained, and its independence of the details of the specific system we considered,
suggest that this may actually be a general relation that holds for any self-inductance.
We can see that this is actually the case. Suppose we are given a certain distribution of
free currents described by J p and we assume a vacuum everywhere. Then the source
equations for H 0 will be V X H 0 = Jy and V H 0 = 0 from (20-29) and (20-57) and
we can imagine solving them for H 0 . The corresponding induction will be B 0 = ju. 0 H 0
and the flux through the circuit will be O 0 = fB0-da
= j u 0 / H 0 da. Now let us
suppose that all space occupied by fields is filled with a l.i.h. magnetic material of
relative permeability <cm; this requirement is necessary in order that there be no surfaces
of discontinuity on which the normal component of M is different from zero, as this
would introduce sources of H other than the free currents. Now if we keep J f
unchanged, the equations to be solved will be exactly the same as before, and therefore
H will be unchanged, that is, H = H 0 . But we see from (20-53) that, under these
conditions, B will be increased by the factor KM, so that B = KmJu,0H = K m i u 0 H 0 =
/cmB0. Then when we integrate over the same surface the new flux will be <& = Kmo
and the inductance L = 4>/7 = /cm3>0/7; thus
L = KmL0
(20-66)
will again be obtained, but now as a general result.
Example
Coaxial line. In. Figure 20-14, we assume the infinitely long cylindrical conductor of
radius a (region 1) to be nonmagnetic and to have a total current I uniformly
distributed over its cross section. The coaxial outer conductor of radius b is, for
simplicity, assumed to have an infinitesimal thickness and to carry a current I
uniformly distributed over its surface. Region 2 between the conductors is filled with a
l.i.h. nonconducting magnetic material of permeability ju, = Km\u0. Region 3 includes all
space (vacuum) outside the system. We want to find all of the field vectors everywhere.
Because of the cylindrical symmetry, H will, as usual, have the form H = H y ( p ) $
and we can use the integral form of Ampere's law (20-32) to calculate it. But this
system is just like that of Figure 18-1 with c = b and the present region 3 corresponding to region 4 of the previous figure. We previously worked this out in the vacuum case
and we can simply take over our results by dividing them by fi 0 in order to get H, or
one can simply repeat the same calculations. Therefore, we find from (18-25), (18-28),
and following (18-33) that
Ip
lira2
,pz
277p
*>-
<20-67)
20-5
LINEAR I S O T R O P I C H O M O G E N E O U S M A G N E T I C M A T E R I A L S
331
Figure 20-14. Coaxial line

carrying equal oppositely
directed currents.
These are shown as a function of p in Figure 20-15. As a check, we will see if the
necessary boundary conditions are satisfied. Since H has only tangential components at
any of the surfaces, the relevant one here is (20-31). We see from the figure that H v is
continuous at p = a as it should be since there is no free surface current density there.
It is also seen that
is discontinuous at p = b and this must be due to the fact that
Kf is different from zero there. We find from Figures 20-14 and 12-5 that
(20-68)
2 77 b
If we go from region 2 to 3. then n = p and (20-31) becomes H 3 , ( b ) ~ H 2,(b) =
X p = (I/2irb)^>
and the
-Hvl(b)v
= ~(I/2irb)$
= Kf X n = (I/2irb)z
boundary condition at p = b is satisfied and the origin of the discontinuity in Hrf is
clear.
We can now find B from (20-53) and (20-67) by using /Xj = ju,3 = ju0 and /x2 = ju, and
we get
p0Ip
fil
B <P 2, =
3 = 0
(20-69)
2 77p
2ira'
K,
Hr
i
2ixa
2nb
Figure 20-15. Magnetic field produced
by coaxial line as a function of distance
P from the axis.
332
These are shown as a function of p in Figure 20-16. We see that Brf also has a
discontinuity at p = b, but it is different from the value ii0I/2irb, which it would be if
there were a vacuum in region 2. In addition, B v has a discontinuity at p = a, while
did not. These differences must be due to the existence of magnetization surface
currents, but we cannot verify this quantitatively until we have found M. (Actually, we
know that B v must satisfy the boundary conditions since
does, as we saw, but it will
nevertheless be instructive to work out the details.)
We can find M from (20-67) and (20-52) and by using Xmi = Xm3 = 0 and
X m 2 = X. a n d w e get
1= 0
<p2=J^
^ 3 = 0
(20-70)
These results are shown as a function of p in Figure 20-17. We see that M has
discontinuities at both a and b, and therefore this must result in net magnetization
surface currents K m according to (20-12). Before we calculate them, however, we can
also check our results in other ways. From (20-70) and (1-87) we find that V M 2 =
p~1(dMtp2/d<P)
= 0 as it must by (20-58). Since there is no free current in the
nonconducting region 2, we know from (20-59) that Jml = V X M 2 given'by (20-10)
must be zero; we can verify this from (20-70) and (1-88) and we find that Jm2p ~
-dM 2/dz = 0, Jm2<p = 0, and Jm2z = p~1[d{pM2v)/dp\
= p~l[d(xmI/2^)/dp]
= 0.
20-6
ENERGY
333
In order to calculate the surface currents, we must remember that n in (20-10) is the
outer normal to the region containing the matter. Therefore, at p = a, n = p and we
get
K2(<0 - A f , 2 ( a ) x ( - p ) -
(20-71)
Lua
At p = b, n = p and we find that

K . , W - W ? x i = - | ^ 4
(20-72)
On the other hand, in order to apply the boundary condition on B, which is given by
(15-16) and the analogue of (20-26) as
B 2 , B u = ju0K X n = fx0Ky X n + n0Km
X n
(20-73)
we must remember that in this equation n is the unit normal drawn from region 1 to 2,
and so on. Therefore, if we always go from the lower-numbered region to the
higher-numbered region, we see from Figure 20-14 that we can always use n = p in
(20-73).
At p = a, Kf = 0 and we find that (20-73), (20-71), (20-54), and (20-55)
give B 2 , - Bj, = n0Kml(a)
X n = 0 i 0 x m I / 2 i T a ) z X p = [fi0(Km ~ l)I/2ira]q> =
[(jLt ii0)I/2na]$
in exact agreement with (20-69) and Figure 20-16. At p = b,
however,
is not zero but is given by (20-68), so that (20-73) applied here becomes
B 3 , - B 2r = (ii0I/27rb)(-z
- xOTz) X p = ~[n0(l + xm)I/2vb]$
= -(fiI/2-irb)$,
which also agrees exactly with the values given in (20-69) and shown in Figure 20-16.
At this point, we have managed to squeeze almost everything out of this long
example that we can, and we have seen that all of the relevant quantities were
calculable and satisfied, in this specific case, all of the applicable general results that we
have previously obtained.
20-6
ENERGY
We recall our result (18-12) for the magnetic energy of a system of free currents
Um=\f
J f A d r
<- 'all space
(20-74)
In the electric case, we concluded after a lengthy discussion following (10-78) that the
useful definition of energy was that associated with the free charges since it represented
the reversible work that could be put into and retrieved from the system by the
manipulation of charges over which we have control. In the same manner, we can
conclude that the energy of the free current distribution is the suitable quantity to call
the magnetic energy of the system since it represents the reversible work associated with
currents that we can control, and we can still use this definition in the presence of
matter.
In order to appreciate the limitations of our final results, it will be useful to begin
again and (18-1), which gives the work required from an external source to produce a
small change in flux, provides a convenient starting point because it was based only on
a general application of Faraday's law. If we let the changes in magnetic energy and
flux be written as 8Um and
for convenience, we find from (18-2) and (16-23) that
sum-
E v * y = L'd
j
j
(20-75)
334
where SAy is the corresponding change in the vector potential on the y'th part of the
system. When matter is present, the magnetization currents contribute to the flux
because of the induction they produce, as described for example by (20-15). As usual,
we are assuming rigid mechanical supports or constraints on the circuits so that no
purely mechanical work is done.
We can now rewrite the sum in terms of the current density in exactly the same way
that we went from (18-10) to (18-11) and we get
8Um = f j
Jy
8 A d r = ( ( V X H ) SA dr
Jy
(20-76)
with the use of (20-29). In this expression, V is the total volume occupied by free
currents. We can, however, extend this to include all space since regions where Jf = 0
will not contribute to the integral. Thus, we get
8Um
= [
( v X H ) SA dr
, space
Jail
'all
(20-77)
But V X A = B by (16-7), so that V X (A + SA) = B + SB and therefore, by (1-117),

V X SA = SB. We can now rewrite the integrand by using (1-116) and (16-7):
(V X H ) SA = H - ( v X SA) - v (SA X H ) = H SB - v (SA X H )
(20-78)
If we compare this with (18-16) and recall how we went from (18-15) to (18-21), we see
that we could proceed in exactly the same way and we will obtain
8Um = f
SB dr
(20-79)
all space
Finally, if we take the zero of energy to correspond to B = 0, we can obtain the total
energy Um by integrating this from the initial state of B = 0 to the final value:
Um=
fBJrl SB dr
(20-80)
all space 0
In general, (20-80) cannot be further evaluated until the dependence of H on B is

known, and this could be quite complicated; we will briefly look at such a situation in
the next section. However, for the important case of linear isotropic magnetic materials,
we can use (20-53) to write H S B = B S B / | U , = S(B 2 /2JLI) so that
2
/-B /( BB 2 \ ^
/B
22
B
and thus (20-80), which gives the total reversible work on the free currents, becomes
Um=
Bdr
(20-81)
all space
which we can interpret in the usual way by introducing the magnetic energy density
um=\
(20-82)
umdT
(20-83)

v=
(
all space
[Compare with (10-84) and (10-85).] If we use (20-53), we can also write this as
B2
Um =
2^
1
=
2M
which, of course, reduces to (18-22) for a vacuum when ju = ju0.
(20
~84)
20-6 ENERGY
335
In Section 18-2 we saw that one of the useful applications of the energy expression
was the calculation of self-inductance. If we have found the fields by other means, we
can evaluate (20-83) and by equating it to (18-9), L can be found from L = 2 U m / I 2 .
Example
Coaxial line. Let us consider the same system as shown in Figure 20-14 for which we
have already found H and B. For simplicity, we shall calculate only the contribution of
the region containing the matter (region 2) to L. The energy density in this region as
found from (20-82), (20-67), and (20-69) is
U
m2
T-^<p2-<p2
ul2
n 2~2
(20-85)
077 p
Since dr = p dp dtp dz by (1-83), the total energy contained in a length / of this region
will be
n. L
2w rh til2
lll2l
t b\
l^ ln u)
so that its contribution to the self-inductance will be

u/
f b\
L2 = In 277 \a J
<20 86)
"
(20-87)
Upon comparing this with the middle term of (18-34), we see that the ratio of this to
the vacuum value is L2/L20 = ju/Mo = l c i n complete agreement with (20-66) for this
case in which the free currents have been kept constant.
The presence of matter will generally change the values of B and H everywhere and
thus the magnetic energy given by (20-81) can also be expected to change. As we
discussed for the electric case after (10-89), this change will generally depend on the
process by which the matter is introduced into the field, and a general discussion can be
quite complex. There is, however, one special case in which the energy change can be
directly identified with the matter, and we consider it as an illustration.
We assume that initially we have vacuum everywhere and some distribution of free
source currents that produce the fields H 0 and B 0 = ju.0H0 throughout space. Now let
us assume that we keep the free source currents fixed in value and location and
introduce a material of volume V into this preexisting field B 0 . If we let B and H be the
new values of the fields that arise because of the presence of the matter, the change in
the energy of the system as obtained from (20-81) will be given by
Um=um
- U0-\(
(H
B - H0
B 0 ) dr
(20-88)
âll space
and can be ascribed entirely to the presence of the matter. As was done in Exercise
10-38 for the analogous electric formula (10-91), it can be shown that (20-88) can be
written as an integral over the volume V of the matter only. The calculation is quite
long, so we shall simply quote the final result, which turns out to be
(20-89)
Since this involves only the volume of the matter, it is reasonable to consider the energy
to be localized in the matter and to represent the energy of the matter. Thus we can
336
introduce an energy density umm associated with the matter given by

umm = i M B 0 = i x m l " o H H 0
(20-90)
We note the similarity of these results to the electrostatic results (10-92) and (10-93),
except for the difference in sign. These expressions are appropriate for situations in
which the magnetization is produced by the field, as is implied by (20-52).
On the other hand, we have to remember the difference between these energies and
the (potential) interaction energy of a permanent dipole in an external induction given
by (19-40). If we consider a small volume d t , its dipole moment will be given by (20-1),
and the interaction energy will be
du'D = du'm
ext
= - M B ext dr
(20-91)
if we write the external induction as B ext rather than the B 0 of (19-40). The total
interaction energy can then be obtained by integrating (20-91) over the volume V of the
material to give
-/MBtrfT
Jy
(20-92)
For example, if B ext does not vary much over the material, we can take it out of the
integral and use (20-2) to get
>ext = m B ext in agreement with (19-40).
When a material is magnetized, there will generally be forces on the magnetization
currents due to the induction. If the material is not rigid, it can deform under the
influence of these forces. This effect is called magnetostriction', it is generally a small
effect and we will not consider it further and assume completely rigid materials.
As we saw in Section 18-3, energy considerations are often useful in discussing
magnetic forces. Because of the complexity of this general subject, we illustrate these
ideas with the following single example.
Example
Permeable rod in a long solenoid. Suppose we have a long solenoid of n turns per unit
length, total length /, circular cross section of area S, and carrying a current /
circulating as shown in Figure 20-18. We also suppose that a cylindrical rod of the
same cross section and of permeability /x is inserted into the solenoid a distance z. We
assume that the current in the solenoid is kept constant throughout and we would like
to find the force on the rod. As we saw in connection with the example of the
interpenetrating solenoids of Section 18-3, we can get a reasonably simple solution only
by neglecting the "end effects" that will be associated with the diverging field lines near
the ends of the solenoid and rod as was indicated in Figure 18-3. Since \xm\ ^ 1> the
/(out)
I
I
I
,
I
z
>
!r
I
i
i
i
Figure 20-18. Permeable rod
in a long solenoid.
20-6 ENERGY
337
magnetic field will be practically unaffected by the presence of the rod and to a first
approximation can be taken from (20-63) as H = niz. The energy density um0 in the
unoccupied vacuum region of the solenoid of length I z and volume (/ z)S as
obtained from (20-84) is um0 = \\x0n2I2.
Similarly, the energy density umM in the
volume zS occupied by the matter will be umM = \fin2I2.
Inserting these constant
values into (20-83), we find that the total energy of this configuration is
Um(z)
= \n2I2S[iiz
+ p0(l
- z)]
(20-93)
Since the currents are kept constant, we use (18-39) to calculate the force:
Fm=(vt/m)/=i(/x-Mo)2^2^
= hxmôn2I2Sz
(20-94)
There are many interesting aspects to this result.

Since everything else in (20-94) is positive, the sign of m depends on the sign of xmIf the material is paramagnetic so that x > 0 the rod will be attracted into the
solenoid, while if it were diamagnetic with a negative susceptibility, it would be
repelled. [This is in contrast to the analogous electrostatic result (10-99), which showed
that a dielectric slab would always be attracted in between the capacitor plates; this
simply expresses the fact that all electric susceptibilities are positive for static fields.]
We can understand this effect of the sign difference qualitatively. If the material is
paramagnetic, M will be parallel to H by (20-52) and the associated surface currents
K m will be circulating in the same sense as I, as we saw in Figure 20-7b. Thus we have
two sets of parallel currents and these will attract each other, as noted following
(13-14), in agreement with the sign found f r o m (20-94). On the other hand, in a
diamagnetic material, M will be opposite in direction to H, its corresponding surface
currents will circulate oppositely to the sense of / , and these "unlike" currents will
repel each other.
While the force Fm is independent of both / and z, it is proportional to the
cross-sectional area S; thus we can introduce a force per unit area f, as
fw = Y
= i ( / x - M o )n2I2i
(20-95)
This can be put into a more interpretable form by expressing it in terms of the energy
densities in the two regions and we get
= (umM
- um0)z
(20-96)
In other words, the magnitude of the force per unit area is just equal to the difference in
energy densities of the two regions, and its direction is such as to tend to move the
material so as to increase the total magnetic energy of the system. This is consistent
with all of our previous results relating to the general tendencies of systems to " try" to
increase their magnetic energy.
Furthermore, (20-96) is in agreement with the idea expressed after (18-52) (and
found from a somewhat different type of example) that magnetic forces per unit area
are properly described as pressures. In effect, we found in (18-52) that we can associate
a force per unit area at the surface of a given region as the product of the energy
density within that region and the outward normal to the region. In this case, we would
then write a pressure term at the interface for the vacuum region as f m 0 = um0( z)
while that for the matter would be fmM = umM( + z). Then the resultant force per
unit area on the bounding surface between the two would be fm = imM + f m 0 =
( u mM ~~ u mo)^ which is just (20-96) again.
338
20-7|
FERROMAGNETIC
MATERIALS
Up to now, we have considered in detail only l.i.h. media for which we can write
B = juH where n is a constant characteristic of the material. Although many materials
can be described quite well by this relation, it turns out that quite a few of the
technologically important ones have a markedly different behavior, and in this section
we consider briefly some of their principal characteristics.
These materials are usually called ferromagnetic materials since they are typified by
the neighboring metals iron, cobalt, and nickel. There are, however, many alloys and
nonmetals that also fall into this category. Although specific properties vary from
material to material, and must be found by experiment, the general features are very
similar and we shall concentrate on these. Essentially, there is no simple relation
between B and H for these cases.
It is also commonly found that the magnetic properties depend on the previous
history of the particular sample. Nevertheless, one can get a sample into a state where it
possesses a certain regularity of behavior. This is usually accomplished by the frequent
reversal of an applied field of decreasing magnitude, and we assume that this has been
done. (The reason for this particular prescription will be apparent shortly.)
What one generally wants to do is to apply an external H field, produced by external
free currents, and then to measure B as a function of H. In order to accomplish this, it
is clear that the sample has to be of such a shape that B and H can be measured
a n d / o r calculated conveniently and unambiguously. In particular, it is desirable that H
can be found from a knowledge of the applied external currents only. If, for example,
one were to use a long solenoid with an iron core as the sample, there would be
discontinuities at the ends and one would have to concern oneself with the effect of the
normal components of M, that is, of the surface density of " poles." But, as we saw in
(20-42), these depend on the value of M, which is connected to the relation between B
and H by B = J L I 0 ( H + M). Since this would complicate the problem enormously, we
want to choose a sample without any ends, and this immediately leads us to think in
terms of a ring or torus. If the torus is tightly wound about the material, then, as we
saw in the last example of Section 15-2, we can to a good approximation assume H to
be confined completely to the interior and it can be found from the free current I in
the N turns of the toroidal coil. Such a device is called a Rowland ring.
We can use the integral form of Ampere's law given in (20-32) to find H. The
calculation is substantially identical to that which led to (15-28) and was illustrated in
Figure 15-12. For a circular path of radius p, ds = p dcpy and (20-32) becomes
(^)H ds = JHy(p)p
d c n c = NI
(20-97)
so that
(20-98)
We let a be the radius of the circular cross section of the ring and b be the central
radius of the torus, and we shall assume that b a. In this case, H ^ i p ) will be
approximately constant and equal to
H =
NI
277b
(20-99)
Then both M and B will also be approximately constant across the cross section of
area S = rra2, and the flux through S will be
$ = SB
(20-100)
20-7
F E R R O M A G N E T I C MATERIALS
339
The measurement can then be carried out in a typical manner. The controllable
external quantity I is changed by a small amount, and thus, by (20-99) the magnetic
field will change by AH = NAI/lvrb
and can be calculated from the known properties
of the toroid. There will be corresponding changes in the induction, A B, and the flux
AO = SAB. The latter quantity can be measured by wrapping another coil about the
torus and finding the total charge AQ c that passes through it as a result of the induced
current. If R c is the resistance of this circuit, we find from (12-2) and (17-3) that the
magnitude of A Qc is
,
,|<find| dt
1 fd<t>
A$
Qc
'induccd
Rc
r J
dt
Rc
which agrees with the result of Exercise 17-6 and gives us A B = RCAQC/S
thus
enabling us to evaluate A B. In this way, the curve describing the induction B as a
function of H can be obtained as the result of a series of small steps.
When a procedure like this is carried out with a monotonically increasing H, the
result typically is a B versus H curve with the general appearance shown in Figure
20-19. A curve such as this is called a magnetization curve; this name is also often given
to an M versus H curve, which will contain substantially the same information. We see
at once that the relationship shown is very nonlinear. Eventually, as H - oo, M
approaches a constant value Ms called the saturation magnetization since the dipoles
become fully aligned. Then the relation B = ju 0 (H + M) - jx0H + ja 0 M s = n0H +
const, and the B H curve becomes linear with constant slope /i 0 . For many materials,
the values of H required to reach this point are impractically large.
Although when we defined ju, = B/H in (20-53), we had linear systems in mind, we
can stretch our definition of n to this case by continuing to use this same equation.
When ju, is found from a curve like Figure 20-19, the result is like that shown in Figure
20-20. The maximum is generally of the magnitude of several thousand, although it can
vary widely from material to material; the ratio J U . / J L I 0 also approaches the limiting value
of unity as H becomes very large. Thus, ju as defined in this way is far from being a
constant, but it can still be a useful and convenient way of describing the material. (We
can also note here that many ferromagnetic materials are crystals, and it is found that
the magnetization curves are generally different for the various directions along which
H is applied.)
N o w suppose that, instead of increasing H indefinitely as we assumed for Figure
20-19, we go only to the maximum value Hl (and corresponding Bx) shown in Figure
Figure 20-19. Typical B versus H

curve for a ferromagnetic material.
Figure 20-20. Relative permeability

versus H for a ferromagnetic material.
340
20-21, and then start to decrease H. We find, as indicated by the directions of the
arrowheads, that the curve does not retrace its path, and B does not decrease as rapidly
as it originally increased. A general behavior like this is known as hysteresis (from the
Greek " to be late" or " to fall short"). Thus, we see that the relation between B and H
is not only nonlinear, but it is not single valued either. If we continue decreasing H
until H = 0, we find that B = 0 as shown. This value Br that remains is called the
remanent induction or remanence or retentivity. In fact, in order to reduce B to zero, the
field H must now be applied in the reverse direction so that B = 0 when H = Hc.
This value Hc is called the coercive force or coercivity. If we now continue this process
of decreasing H to Hlf and then reverse direction and increase it back to +HX, the B
versus H curve becomes the closed curve known as a hysteresis loop.
If we had gone out to H2 > H1 before starting the reversal, we would find as shown
in Figure 20-22 that we would get a different hysteresis loop with different remanent
induction and coercive force; similarly for the value H3 of the maximum. In other
words, the whole B H space can be filled with hysteresis loops and a given numerical
value of /x does not have a great significance unless the corresponding conditions are
more closely specified. We see that the tips of these hysteresis loops trace out the
monotonic magnetization curve of Figure 20-19; we also see why we prescribed
the application of alternating fields of gradually decreasing magnitude in order to get
the sample to an initial state where B = 0 when H = 0. Finally, we can also note that
if we are at a value such as HA and then go through a small cycle in H starting there,
we will trace out a little hysteresis loop like that illustrated.
In systems in which hysteresis occurs, there is also an irreversible conversion of
energy into heat when the system is taken around a complete cycle. In the present case,
it must correspond to a conversion of magnetic energy into heat; this is in addition to
the heat produced because of the existence of a conductivity and that is described by
(12-35). We can use our previous general result (20-79) to demonstrate this effect.
If we let 8wm be the energy required from the external source per unit volume of the
material, we can write the integrand of (20-79) as 8wm dr and we have
8wm = H SB
(20-101)
This is equal to the shaded area shown in Figure 20-23 since we are dealing only with
20-7 F E R R O M A G N E T I C MATERIALS
341
Figure 20-22. Various possible hysteresis loops. PM indicates the

general state of a permanent magnet.
Figure 20-23.
Interpretation of H8B as an area.
342
isotropic materials where B and H are parallel. If we now imagine taking the system
around one cycle of the hysteresis loop, the work done per unit volume will be the sum
of terms like (20-101):
Wm
SB = <
'cycle
H8B
(20-102)
cycle
This is numerically equal to the area enclosed by the hysteresis loop so that wm is
different from zero and positive. Since the initial and final states of the system are the
same, the magnetic energy, which is a function only of the state, will not have changed,
and therefore the work per unit volume given by (20-102) must represent an irreversible
conversion of work into heat. This result can also be written in a form involving only
the material. Since, by (20-28), B always equals /x 0 (H + M), we have
8wm = H
SB = j u 0 5 ( f H 2 ) + ju 0 H S M
(20-103)
When this is inserted into (20-102), the first term will give zero, as we saw, for example,
after (13-4), so that finally
wm= <
'cycle
ju 0 H 5 M
(20-104)
The last term of (20-103), /x 0 H 5M, represents the work supplied to a unit volume of
the material and provides the starting point for the description of magnetic systems in
thermodynamics.
20-8 MAGNETIC CIRCUITS
By rewriting some of our results from the last section, we can obtain something that is
both suggestive and useful. We will continue to write B = nH with whatever value of n
is appropriate; then from (20-97), (20-99), and (20-100), we obtain
$ =
jaSNI
ITTI)
= t
NI
NI
r =
(L/fiS)
3%
1 ,
= (H ds
where / = 2nb is the length of the path C around the toroid and
/
<=
IxS
(20-105)
(20-106)
We now recall the result of Exercise 12-14 in which, by using energy considerations,
one could show that in a complete circuit of total resistance R containing an emf $ , the
steady-state value of the current is
1
i
(f>E-ds
(20-107)
R
R Jc
where R = 1/oA by (12-28) where A is the cross-sectional area of the conductor of
conductivity a. Comparing these expressions, we see that there is a resemblance, and,
by analogy, these systems are called magnetic circuits. Again, by analogy, there is a
corresponding terminology; one can define a magnetomotive force or mm/..# as
-i
H ds = NI
and & is called the reluctance. Then (20-105) can be written as

Ji
<D =
(20-108)
(20-109)
20-8 MAGNETIC CIRCUITS
343
so that it looks like (20-107). Also from (20-106) we see that the permeability ju plays
the same role in determining the reluctance as the conductivity a does for the
resistance.
The correspondence between the two cases is not exact. In the electrical circuit, the
moving charges comprising the current do not leave the conductor, while in a general
magnetic circuit, it is quite possible for the lines of B to be out in the space surrounding
the material. This flux "leakage" can result in large quantitative errors in the application of (20-109). Even so, this method of discussing magnetic circuits does prove useful,
particularly in aiding us to understand the overall features of the situation.
The main use of these ideas is in the design and construction of magnets. These are
of two principal general classes: electromagnets in which the magnetic field is produced
by a current in the windings, and permanent magnets in which one uses a ferromagnetic material that will have an induction present even in the absence of free currents.
We will look at each of these, but, for simplicity, we consider only those of a toroidal
shape.
Since these magnets are generally built to produce a magnetic field for experimental
purposes, one needs a place to put the sample to be studied. The natural thing to do is
to cut out a piece of the magnet material, thus making a " g a p " of length / 0 as shown in
Figure 20-24. We assume a small gap length; this is done in order that we can neglect
the "fringing" properties of B discussed in Exercise 15-11 and illustrated for the electric
field in Figure 6-10. In this way, we can assume that all of the lines of B present in the
material are in the gap as well so that, by (20-100), we can take $ to be constant and
need not be concerned with a possible change in the effective cross-sectional area S.
Example
Electromagnet. If we let Hi be the magnetic field in the material of length I l0 where

/ = Irrb as before, and H0 the value in the gap of length / 0 , we find that the mmf of
(20-108) becomes
C^)H ds = Hj(l - /) + H0l0 = NI
(20-110)
We assume that the surfaces forming the gap are cut so that they are perpendicular to
B. Since the normal components of B are continuous, we have Bt = B0 = $ / S ;
'/
\
>
Figure 20-24. Toroidal magnet with a
gap of length / 0 .
344
furthermore, 77, = Bt/n, while H0 = B0/fi0.

in terms of $ and NI as
<D
Thus we find that (20-110) can be written
(I ~ la)
lo
+
US
n0S
= NI
(20-111)
which again has the form (20-105) with the total reluctance of the circuit now given by
# =
U-Io)
pS
.
+
lo
St s +
H0s
(20-112)
In other words, these reluctances add, just as do resistances in series. We can also see
that we can generalize the idea of reluctance to a circuit of varying properties so that
we will have
ds
(20-113)
St
-i
If the gap is small (/ 0
MS
/), then (20-112) can be approximated as

I
lo
nS
n0S
(20-114)
Furthermore, if the material is something like iron with ju. = 5000juo, it is often possible
to neglect the first term and approximate St once again to get
'o
(20-115)
Mo S
so that the corresponding approximation to (20-111) would be

$ = NI-
Sfi0
(20-116)
ln
This last approximation is not always valid since it depends both on the value of /i and
on the relative values of the lengths / and / 0 ; it should really be checked before being
used. In any event, (20-115) shows us that most of the reluctance is due to the vacuum
(or air) in the gap, that is, the iron is more "permeable." Also since Bt = B0,
/ / , / / / = iu.0/fj, so that H, : H0 and (20-110) becomes Jt = NI H0l0 showing that
essentially all of the magnetomotive force producing the flux is applied "across" the
gap, even though the gap is physically very small.
Example
Permanent magnei. In this case, 7 = 0. The expression (20-110) is still valid, however,
and we find that
/ 0 77 0
l0
o
(20
'117)
showing that the magnetic fields in the two regions are oppositely directed. In the gap,
M = 0 and H0 = B0/fi0 so that H and B are in the same direction in the gap, but since
Bj = B0, H , and B, are oppositely directed within the magnet. Thus, a permanent
magnet can be represented by a point in the second quadrant of a B 77 curve as is
the point labeled PM in Figure 20-22.
EXERCISES
In addition, Bt = B0 = n0H0
Mi =
= fi0(Hi
+ Mt) and we find that
Bl
MoU
'_
345
'o)
= (l + y)
Mo \
w
(20-118)
and is in the same direction as Bt, as is to be expected. The change in sign of H, which
occurs at the surface separating the material from the vacuum, results from the
discontinuity of M at this surface, that is, it arises from the surface density of " p o l e s "
as described by ( 2 0 - 4 2 ) and ( 2 0 - 4 3 ) . Since M , and H , are oppositely directed in the
material, H , is often described as a "demagnetizing"
field.
EXERCISES
20-1 As we will see in Appendix B, molecular

permanent magnetic dipole moments are typically
of the magnitude of the Bohr magneton fie =
eh/Airme = 9.27 X 10~ 24 ampere-(meter)2 where
h is Planck's constant and me is the electron
mass. Suppose each molecule of an ideal gas had
a permanent moment of n e . Find the maximum
possible magnetization for this gas at 100 C and
one atmosphere.
20-2 A large sheet of material of thickness d
has parallel faces that are perpendicular to the z
axis. It is magnetized in such a way that M =
A/(l + otz)z. where M and a are positive constants. Sketch the lines of M. Find Jm and Km.
Repeat this for the case where M is given by
M( 1 + ax)z.
20-3 A cube of edge a has the location and
orientation shown for the figure in Figure 1-41.
The material within the cube has a magnetization
given by M = (M/a)yx + (M/a)xy
where
M = const. Find the current densities J, and Km
and sketch their directions.
20-4 A cylinder of length / and circular cross
section of radius a has its axis along the z axis. It
is magnetized such that M = Mx where M =
const. Find J, and K, and sketch their directions.
origin. Its magnetization is nonuniform and given
by M = (az 2 + fi )z where a and ft are constants. What are the units of a and /?? Find the
magnetization current densities J, and
as
expressed in spherical coordinates.
20-6 Verify that the magnitudes and directions
given for B, by (20-22) and (20-25) also hold for
negative values of z and thus agree with Figure
20-11.
20-7 A cylinder of length / and circular cross

section of radius a has its axis along the z axis
and the origin at the center of the cylinder. It has

a uniform magnetization M = Mi. Find B and H
for all points on the z axis. Then use your results
to discuss the limiting case of a very thin disc,
that is, / a. Find B outside at a point z /.
Find B and H inside the disc.
20-8 An infinitely long cylinder of radius a has
its axis along the z axis. Its magnetization is given
in cylindrical coordinates by M = M0(p/a )2<p
where M0 = const. Find Jm and Km. Verify that
the total charge transferred is zero. Find the
values of B and H everywhere, both inside and
outside the cylinder.
20-9 A sphere of radius a has the origin at its
center. It is magnetized in such a way that, in
spherical coordinates, M = M(r)i and M(0) is
finite. Find Jm and Km. If there are no free
currents anywhere, find B inside the sphere. Find
H inside the sphere. Find H at the surface of the
sphere and just outside of it.
20-10 In order to use (20-33), one must be able
to measure B. If one wants to do this by using a
moving point charge and (14-30), one measurement of the force is not enough to determine B.
Show that, if two measurements are made using
mutually perpendicular velocities vx and v2, and
the forces are found to be F, and F 2 , respectively,
then B can be found from
B = - ^ { v22l X v, + vjF2 X v2
2qv\v\
+ [vi (F2 X v 2 )]vj
+ [v2- (Fj X v,)]v 2 }
20-11 Since there are no magnetic poles, one
should expect that the total pole strength (net
magnetic charge) of a finite piece of magnetized
material would be zero. Show that this is the case.
346
20-12 Show that the total magnetic dipole moment of a system as given by (20-2) can be
written in terms of poles as
Show that then each component of H satisfies

Laplace's equation, for example, V 2 / / v = 0.
20-14 Find the pole distribution for the cylinder

of Exercise 20-4. Find <!>, everywhere. Show that
the demagnetizing factor is \ for this case. (Assume that I is infinite.)
20-19 Lines of B can be " refracted" at a surface

of discontinuity. Assume that there are no free
surface currents on the boundary and that the
angles giving the directions of B are measured
from the normal as in the analogous electric case
shown in Figure 10-14. (a) Show that the law of
refraction is m l cot
= Kml cot a2. (b) If |x,|
<sc 1 for both media, then
= fi2 ji0 and the
angle of deviation 8 = a2 will be very small.
Find an approximate expression for 8. Evaluate
this result for , = 4 5 , region 1 a vacuum and
region 2 a paramagnet for which Xm2 2.2 X
10 5 (e.g., aluminum), (c) If |x,| is not very
small, the result of (a) must be used. As extreme
examples, evaluate 8 for aj = 4 5 , region 1 a
vacuum, and: (i) x m i ~ 0.95, a nearly "ideal
diamagnet" (Note: if x, = ~1> then B = 0); (ii)
Xmi~ 500, a moderate ferromagnet.
20-15 A spherical shell of radii a and b is

uniformly magnetized so that M = Mz for a < r
< b. Choose the origin at the center of the sphere
and find <f>m at all points on the positive z axis.
Find Hz for positive values of z and sketch its
behavior as a function of z. Verify that Hz satisfies
the appropriate boundary conditions.
20-20 Find the expressions analogous to (20-59)

and (20-60) when the material is not homogeneous and thus show that there can be a magnetization current density even in the absence of a
free current density. Similarly, find an expression
for the volume density of magnetic poles for this
situation.
20-16 Consider a cylinder of length I and circular cross section of radius a with a permanent
magnetization M. A uniform external induction B
is also present. Assume that M is unaffected by
the presence of B and find the total torque on the
cylinder. Use the result of Exercise 20-12 to express the torque in terms of the pole distribution
and show that the result can be interpreted
as saying that there is a force on a "point pole"
qm given by <7mB. If there is vacuum everywhere
else, show that B due to another point pole q'm
is given by B = fi0q'mR/47rR2, and thus deduce Coulomb's law for poles in the form
= Mo<7, n < A / 4 ^ R 2 20-17 When you look up susceptibilities in tables, you will not generally find numerical values
of xm- Instead, one usually finds either the mass
susceptibility xm mass o r
molar susceptibility
These
Xm,molarare defined so that x m ,,n.sJ f and
Xm,molarH give the magnetic moment per unit
mass and per mole, respectively. Find how x m ' s
related to each of these; you will need to use the
symbols for the mass density d and molecular
weight A as well.
20-21 A sphere of radius a and relative permeability K m is placed in a previously uniform magnetic field H 0 = H0i. Find H everywhere.
( Pmr
Jv
<amrda
Js
and compare with (8-22), for example. [Hint:

recall (19-8) and (19-11).]
20-13 A very large sheet of material of thickness
d has parallel faces that are perpendicular to the
z axis. It is uniformly magnetized with M = A/z.
Find the corresponding pole distribution, and
using this find H everywhere. What is the demagnetizing factor in this case? Is your result compatible with those corresponding to the last sentence
of Exercise 20-7?
20-18 Suppose that we have a static situation

involving a l.i.h. material with conductivity a.
20-22 A spherical shell of radii a and b has a

relative permeability nm for a < r < b. There is
vacuum everywhere else. It is placed in a previously uniform magnetic field H 0 = H0i.
Show that the "shielding factor," that is, the ratio
Ht/H0 where Ht is the magnetic field in the
cavity, is given by 9/cm[(K, + 2)(2ic, + 1)
2(k, - 1 )2{a/bf]
20-23 A very long straight current I is parallel
to the plane surface of a semiinfinite l.i.h. magnetic material and a distance d from it. Show that
the magnetic field in the vacuum region can be
obtained as the resultant of this current and an
image current / ' = I{nm - l)/(nm + 1) located
a distance d inside the material. Find the current
I", which will give the correct value of H within
the material if it is located at the same position as
/. What is the force per unit length between I
and the material? Is this force attractive or repulsive?
20-24 In our discussion of the coaxial line, wc
assumed that the inner conductor (region 1) was
nonmagnetic. Suppose that it is actually diamag-
EXERCISES
netic with susceptibility x m j < 0 while everything

else is unchanged. Find H, B, M, and J, for all
points of region 1 and plot your results. Verify
that the appropriate boundary conditions are
satisfied at p = a.
20-25 Consider the same coaxial line as discussed at the end of Section 20-5, except that
region 2 between the conductors is filled with a
nonhomogeneous material such that (cm = n(p/a)
where = const. Find H and B within this region
and the contribution L2 of a length / of this
region to the self-inductance.
20-26 Consider the same coaxial line as discussed at the end of Section 20-5, except that
region 2 between the conductors is filled with two
l.i.h. materials as follows. For a < p < c, the
relative permeability is Kml, while it is Kml for
c < p < b. Find the contribution L2 of a length /
of this region to the self-inductance.
20-27 A paramagnetic liquid of mass density d
is contained in a U-tube of circular cross section
of radius a. When a field H 0 is applied to one
arm of the U-tube, the liquid level is found to rise
a distance h into the region where H 0 = 0. Assume that x m ^ 1. neglect all end effects, and
find an expression from which the susceptibility
can be found.
20-28 A very long straight wire carries a current
I. Another long wire of relative permeability Km
and circular cross section of radius a is parallel to
the first and at a distance b away from it where
b a. Show that there will be an attractive force
per unit length between them given approximately by ( K - 1 ) f i a 1 / 2 ( K + l)nb3.
20-29 An electromagnet is made in the form of
a torus with a rectangular cross section. The
m
3 4 7
width of the cross section is 2 centimeters, while

the inner and outer radii are 7 and 8 centimeters.
The coil has 1000 turns and has a gap cut into the
core of length 0.25 millimeters. When the flux
produced in the ring is 2.60 X 10 4 webers, the
permeability of the core is 4250fi0. Find the induction, the reluctance, the current in the windings, the magnetic field in the gap and in the iron
core, and the fraction of the total magnetomotive
force that is "across" the gap.
20-30 Show that the energy stored in a toroidal
magnetic circuit can be written in the form \3t<f>2.
20-31 Find the expression giving the force attracting one of the faces forming the gap in an
electromagnet toward the other face. Evaluate
this force for the case described numerically in
Exercise 20-29, and then express the force per
unit area in atmospheres.
20-32 We noted before the last example of Section 20-6 that forces on magnetic materials can be
ascribed to magnetic forces acting on the bound
magnetization currents. Therefore, it may seem
puzzling that we found a force on the permeable
rod of Figure 20-18 that is in the z direction since
both the magnetic field and induced magnetization were assumed to be uniform and along this
direction; this would make the magnetic forces
perpendicular to the solenoid axis. The origin of
the net force can be found in the very thing we
neglectedthe "edge effects" or curvature of the
actual lines of induction as illustrated in Figure
18-3. Show qualitatively that the net force on the
bound surface currents as produced by the curved
field lines will really be in the z direction and that
the sign of the force agrees with that deduced
from (20-94) found from energy considerations.
21
MAXWELL'S E Q U A T I O N S
Since our last summary at the beginning of Chapter 17, we have added Faraday's law
and defined the vector H as a result of including the magnetic effects of matter. Thus, at
this point, our overall results can be summarized by the equations
V D = pf
v B = 0
V X E =
3B
(21-1)
V x H = j7
as given by (10-41), (16-3), (17-30), and (20-29). These field vectors are related to each
other and to matter through (10-40) and (20-28): D = c 0 E + P and H = (B/ju 0 ) M.
In addition, we have the equation of continuity (12-19), which describes the conservation of free charge:
v *
J/ +
dpf
~dt
(21
"2)
As we pointed out previously, we assume that these equations are still correct when the
fields can vary with time even though, with the exception of Faraday's law, the
equations were obtained from action at a distance laws that describe time-independent
situations.
A great contribution of Maxwell to electromagnetic theory was his first pointing out
that two of the equations of (21-1) are incompatible with charge conservation as
described by (21-2), and then showing how this situation could be resolved by the
introduction of yet another "current."
TTT1 THE DISPLACEMENT CURRENT
As stated in (1-49), the divergence of the curl of any vector is always zero. If we
calculate the divergence of the curl of H as given in (21-1) and use (21-2), we obtain
( V X H ) = V - Jf=
(21-3)
Since we can generally expect to have ( d p f / d t ) # 0, we see that we have a fundamental

contradiction between (21-3) and the requirement of (1-49).
In order to remedy this situation, Maxwell assumed, in effect, that the equation for
V X H was not yet complete, but must have another "current density" in it, that is, we
assume that it should really be of the form
V X H = Jf + Jd
(21-4)
where J(j is the additional necessary term. Our aim now is to find 3d.
If we substitute (21-4) into (1-49), and use (21-2) and v D = pf, we obtain
dp/
V - ( V X H ) = 0 = V - J / + V - J ^ = - ++ vV J J d == - v
7
' " - a 7
348
D + v
Jd
21-1
so that
THE D I S P L A C E M E N T C U R R E N T
0
dD
-
I
=
~dt
349
(21-5)
since we can interchange the order of partial differentiation. Therefore, whatever Jd is,
it must satisfy (21-5). The simplest assumption, of course, is that the term in the
parentheses is zero; accordingly, Maxwell assumed that
h =
dD
(21-6)
dt
so that (21-4) can also be written

dD
V X H = Jf +
(21-7)
and is now completely consistent with (21-2). We can also note that (21-5) could be
satisfied in general by writing Jd = (dD/dt)
+ V X G where G is some arbitrary
vector; however, it has never yet been found necessary to do this and it would be
needlessly complicating to carry this extra term along.
This new current density Jd was called the displacement current by Maxwell. It has
important consequences and is needed to make our later results agree with experiment,
and, as we will see in Chapter 24, it is essential for the existence of electromagnetic
waves.
Now that our form for v X H has been changed, we must investigate how this will
affect some general results that depend on it. First, the integral form of Ampere's law
for H is obtained by combining (21-7) with Stokes' theorem (1-67), and it becomes
^ H ds =
J (J
+ J^) da =
JJ
da +
= I f , enc + I d , enc
da
(21"8)
where / rf>enc is the total displacement current enclosed by the path of integration. When
D is constant, dD/dt = 0 and (21-8) properly reduces to our previous expression
(20-32).
Another important result that depends on V X H is the boundary condition for the
tangential components of H at a surface of discontinuity in properties. If we put (21-7)
into our general result (9-13), we obtain
3D
n X ( H 2 H x ) = lim h I fJ , +
h->0
dt
(21-9)
As before, we can use the analogue of (15-14) for free currents to get limA 0(h 5 f ) = K^.
Furthermore, just as we did preceding (17-13), we can expect that dH/dt will remain
finite as the transition layer is shrunk to zero thickness so that, as h > 0, h( dD/dt) - 0
and (21-9) will become n X ( H 2 H x ) = K^, which is exactly (20-30) again. Therefore
the introduction of the displacement current has not changed this boundary condition
so that the tangential components of H will change only as a result of the existence of
free surface currents.
Example
Charging capacitor. In Figure 21-1, we have a side view of a parallel plate capacitor
with circular plates of radius a. For simplicity, we assume a vacuum between the plates,
and, as usual, we also assume that the separation between the plates is so small
350
MAXWELL'S EQUATIONS
-3>z
Figure 21 -1. Side view of a

parallel plate capacitor being
charged.
compared with the radius that the electric field E can be taken to be uniform and
confined entirely to the region between the plates. The electric field is given by (6-4) as
E = (oy/e 0 )z = (q/e Q ira 2 )i. in terms of the surface density cy, the total charge q, and
the area ma2 of a plate. Therefore, in the region between the plates,
D = e 0 E = ^-z
Tra
(21-10)
These results are those corresponding to the electrostatic case in which the charge
has a constant value and is uniformly distributed over the surfaces of the plates. N o w
let us assume that the capacitor is being charged at a constant rate dq/dt. We can
imagine this being done by means of a current / = dq/dt that is traveling in a long
thin wire coinciding with the z axis, as illustrated. However, we want to continute to
treat this case as if the charge is still uniformly distributed over the plates so that we
can continue to use (21-10). In other words, we assume that the charge is essentially
instantaneously distributed over the plate surface. W h a t this means, in effect, is that the
capacitor is being charged at such a low rate that any appreciable amount of charge
transferred to the plate is added in a time large compared to the relaxation time of the
conductor that we discussed in Section 12-6. We saw there that this time could well be
as short as 10~ 1 4 second for a good conductor so that we will have no difficulty in
assuming that / is small enough so that the charge distribution is nearly always the
same as that of the equilibrium electrostatic case, that is, we are dealing with a
quasistatic process.
Since q is not constant, the displacement current as obtained from (21-6) and
(21-10) is
3D
I
(
>
Thus, we have the equivalent of a uniform current density between the plates as shown
in Figure 21-2; we use the word "equivalent" because there is no transfer of real charge
in the vacuum region between the plates, but, by (21-7), there is a source term for the
magnetic field. There is a numerical equivalence, however, for we find from (21-11) and
(12-6) that the total "current" between the plates is
Id = f j d d* = f - ^ z da = /
*S
(21-12)
S HQ
and is thus equal to the total real current bringing charge up to the capacitor.
We can use (21-8) to calculate H. With circular plates, and the current I along the z
axis, we will have axial symmetry about the z axis and we can conclude that H has the
general form H = H v ( p ) . Therefore, choosing a circular path of integration of radius
p centered on the z axis and perpendicular to it, we will find that the line integral in
21-1
THE D I S P L A C E M E N T C U R R E N T
351
Figure 21 -2. The displacem e n t current b e t w e e n t h e

plates of a parallel plate
capacitor being charged.
(21-8) becomes H^lirp as usual so that

H^P)
^f,enc ^ d, enc
2irp
(21-13)
It will be helpful to divide space into the four regions shown in Figure 21-3. Region 1 is
the volume between the capacitor plates, and 2 is the remainder of that enclosed by the
two parallel planes which coincide in part with the plates. Regions 3 and 4 are the rest
of space; / is different from zero within them.
In regions 3 and 4, If enc = I while Id enc = 0 and we get
/ / 3 ( p ) = Hv4<p(p)
4*
For a path in region 2, I^ e n c = 0 while I d

"*2<P)
enc
2^
I
lirp
(21-14)
= / by (21-12) so that
( p > a )
(21-15)
Although (21-14) and (21-15) look alike, they were found by using different sourcesa
real current and a displacement current, respectively. Finally, for region 1 where p < a,
Figure 21-3. Regions used for

calculating t h e magnetic field.
352
we have I/ cnc = 0 and l d ^ J Id = Id,enc/I

us another familiar-looking expression:
,(p)
hf>
2
lira
= up2/2
IP
= p2/a:
(p ^ a )
2 it a'
(21-16)
If we first consider values of p > a, we see from the above results that as we go from
region 3 to 2 and then from 2 to 4, the tangential components of H are continuous as
we expect from (20-31) since there are no free currents on these imaginary boundaries.
On the other hand, if the displacement current had not been included in this particular
calculation for this axially symmetric case, we would have concluded that H = 0 in
region 2 (and 1 as well); this would have led to a discontinuity in the tangential
components of H across the boundary between 3 and 2 where there is no real current in
direct contradiction to all of our experience and expectations. Maxwell also considered
this problem and it was arguments such as these that helped persuade him of the
necessity of the existence of the displacement current.
If we now consider values of p < a, there is a discontinuity in the tangential
components of H as we go from region 3 to 1, say, and we find from (21-16) and
(21-14) that the difference is
H
-H
2 ?rp \ a '
(21-17)
1 Ml-
and, according to (20-31), this must be equal to KfX n which in this case is
K^X n,
= Ky X z. In order to verify this, we need to know K^,
which we can find
/'
with the aid of Figure 21-4. In a time A/, a charge Aq = I At is brought to the plates.
By our assumption, this is distributed "instantaneously" and uniformly over the plates.
By symmetry, Ky must be radial and therefore has the form Kj = Kj(p)p. As shown in
the figure, the surface current through the circle of radius p must be distributing the
charge that is to go to the remainder of the plate outside of the shaded circle of radius
p. Therefore, in a time At, the total charge transferred across this circle as obtained
from Figure 12-5a is
jKf
ds At = Kj At Inp = Aa^ma2
Aq
na
( 7ta
7rpz)
np2) = lAt\l
Figure 21-4. Calculation of surface current density

o n t h e plate of a capacitor being charged.
2 1 - 2 M A X W E L L ' S E Q U A T I O N S IN G E N E R A L F O R M
and therefore Kf (//27Tjo)[l (p2/a2)]

K
353
so that
(
Thus, with the use of (1-76), we find that Kf X z = Kfp X z =

=
2 2
(I/277p)[(p /a )
l]q>, which is exactly the right-hand side of (21-17). Thus the
boundary conditions are precisely satisfied, and we can point out again that this would
not have been the case if we had failed to include the displacement current as a source
of H.
If J j = 0, then V x H = dH/dt and there is still a possible source of H. In this

form, there is a sort of analogy between it and Faraday's law given in (21-1), which one
can state somewhat crudely by saying that Faraday's law describes an electric field as
being produced by a changing magnetic field vector, while (21-7) describes a magnetic
field as being produced by a changing electric field vector.
21-2 MAXWELL'S EQUATIONS IN GENERAL FORM

At this point, finally, our present knowledge of macroscopic descriptive electromagnetism is complete and can be summarized by the equations (21-1) as extended in
(21-7):
(21-19)
v D - pf
dB
(21-20)
V X E = -
V B = 0
(21-21)
3D
V x H = Jff +
dt
(21-22)
These equations are known as Maxwell's equations and are assumed to be always true.
[These actually correspond to a total of eight scalar equations, since (21-20) and (21-22)
each have three components.]
It is worthwhile reminding ourselves of the physical content of these equations.
Equation 21-19 summarizes Coulomb's law of force between point charges plus the
electrical effects of matter, while (21-20) represents Faraday's law of induction, and is
also compatible with Coulomb's law for static fields. The third member of the group
(21-21), is a consequence of Ampere's law of force between currents and also reflects
the fact that free magnetic charges are not known to exist. Finally, the last one, (21-22),
includes Ampere's law of force between currents plus the magnetic effects of matter
plus conservation of free charge; the last follows from the fact that the equation of
continuity (21-2) can be derived from (21-22), (21-19), and (1-49), and thus need no
longer be written separately.
These differential source equations must of course be supplemented by the defining
equations (10-40) and (20-28), which relate pairs of the field vectors along with the
description of matter in terms of the corresponding volume densities of dipole moments:
D = e0E + P
(21-23)
H =
B
Mo
(21-24)
354
MAXWELL'S EQUATIONS
Although the boundary conditions at a surface of discontinuity can always be

obtained from Maxwell's equations and our general results from Chapter 9, it is usual
and convenient to list them separately. They are given by (10-42), (9-16) and (17-13),
(16-4), and (20-30) and (20-31):
n (D 2 D x ) = a f
nX(E2-E1)=0
(21-25)
or
E 2 , = Elt
n (B2 B x ) = 0
n X ( H 2 - H j = Kf
(21-26)
(21-27)
or
H2f - H ^ ^ X n
(21-28)
where we remember that n in these equations is always drawn from region 1 to region 2.
(Note here that many books, particularly more advanced ones, drop the subscript /
used to denote free charges and currents and require that the reader remember that this
is what is meant; thus one commonly finds all of these equations written in terms of the
symbols p, J, a, and K. For convenience, however, we will not do this but continue
using the subscript.)
These equations all describe the behavior of the field vectors. The basic connection
between the fields and their effect on charged particles is described by the Lorentz force
on a point charge q as given by (14-32):
F = ?(E + v X B )
(21-29)
where v is the velocity of the charge.

It is often convenient to have Maxwell's equations expressed in terms of only two
vectorsone electric and one magnetic. For example, if we use (21-23) and (21-24) to
eliminate D and H from (21-19) through (21-22), we obtain Maxwell's equations
expressed only in terms of E and B:
V E = ( p f - V P)
(21-30)
V XE = -
at
V B = 0
(21-31)
(21-32)
/
<9E
dP \
V X B = juJ Jf + V X M + e 0 + I
(21-33)
Two of these equations are unchanged from before and the other two are no longer
quite as compact since the properties of the matter now appear explicitly. Nevertheless,
their form is such as to be easily understood. The term in the parentheses of (21-30) is
clearly the total charge density written as the sum of the free and bound charge
densities as we found in (10-38). Similarly, the term in parentheses of (21-33) represents
the total current density J t o t . The first two parts are clearly the free current density and
the magnetization current density of (20-10). The last two terms together are the
displacement current density that now is seen to have two contributions: the first,
e 0 ( d E / d t ) , is present even if there is no matter and is often called the vacuum
displacement current density and the last term dP/dt we met before in (12-18) as the
polarization current density associated with the motion of bound charges. Unless P and
M represent permanent polarization and magnetization, there is an additional dependence on E and B in these equations in that we still have the possibility that P = P(E)
and M = M(B) and until these functional relationships are known, these equations will
be of limited utility.
21-3 MAXWELL'S E Q U A T I O N S FOR LINEAR I S O T R O P I C H O M O G E N E O U S MEDIA
355
The boundary conditions can also be expressed completely in terms of E and B and
we find that the ones that do change in form become
n (E 2 - E x ) = -
[a f - n (P 2 - P,)]
n X (B 2 - B j - ju0 [K, + n X ( M 2 - M x ) ]
(21-34)
(21-35)
in agreement with our previous results (10-12) and (20-12).

In a similar manner, one can write the basic equations in terms of the pairs (E, H),
(D,B), and (D, H) if it is desirable and the terms that arise can be interpreted in the
same way.
The so-called integral forms of Maxwell's equations are obtained by combining the
divergence theorem (1-59) and Stokes' theorem (1-67) with the appropriate members of
(21-19) through (21-22) and the results are
(T)D da = I pfdr
Js
Jy
(21-36)
p
r 5B
(f)E ds = - / da
Tc
Js dt
(21-37)
<^)B </a = 0
(21-38)
r
r
H ds = J / Jf
'c
s
dn +
s dt
da
(21-39)
A little reflection will convince one that all of the general results that we have found
in the previous chapters can be obtained from the equations that have been summarized in this section, often by working backward from the way in which we originally
got them.
Finally, we can point out that Maxwell's equations are linear differential equations.
Consequently, if two or more electromagnetic fields that are solutions of them are
known, the sum of these fields will also be a solution. This is often referred to as the
superposition property of the electromagnetic field or as the superposition principle.
MAXWELL'S EQUATIONS FOR LINEAR ISOTROPIC HOMOGENEOUS MEDIA

An important special and simplifying situation occurs when we are dealing with linear
isotropic homogeneous media. For simplicity, we restrict ourselves to the case where the
materials are l.i.h. in all of their properties. Then we can use the constitutive equations
(10-51) and (20-53) to write
D = eE
B = juH
(21-40)
In addition, the free current would be given by (12-25) as Jf = crE. The quantities e, n,
and a are scalar constants, characteristic of the material. Now it may happen that there
are free currents present other than those due to the conductivity of the material; for
example, the material could have beams of charged particles incident on it from an
external source. Thus, if we let J / be the free current density arising from sources other
than conductivity, we can add this to o E and write the total free current as
J / = a E + J/
(21-41)
Although (21-41) represents a general possibility that should be kept in mind, we
356
MAXWELL'S EQUATIONS
actually have little need of it in our applications, but we will continue nevertheless to
include jy for completeness.
If we insert these expressions into (21-19) through (21-22), we can, for example,
express them in terms of E and B only as
V E
x E
Pf
e
(21-42)
dB
B =
V
V
(21-43)
~ ~~dt
(21-44)
dE
B = juaE + juJ/ + jue
'
at
(21-45)
and the corresponding boundary conditions that are changed in form become
n ( e 2 E 2 - C i E j = af
B-j,
fi 2
Bi,
=
Ml
(21-46)
X n
(21-47)
Similarly, if the equations are expressed completely in terms of E and H, we obtain

V E =
V X E = -II
(21-48)
(21-49)
at
V H = 0
(21-50)
dE
V X H = ffE + J / +
~~d7
(21-51)
n ( h 2 H 2 - jlijHJ = 0
(21-52)
along with
In contrast to the equations of the previous section, these results are not always true,
but they are often true.
The integral forms of Maxwell's equations can be obtained for this restricted case in
the same manner by which one gets (21-36) through (21-39). The general superposition
principle also applies to these materials, of course.
|~^4l POYNTING'S THEOREM

We are now in a position to reconsider energy for these general fields described by
Maxwell's equations. We recall our previous result of (12-35) that the rate at which
electromagnetic energy is dissipated into heat per unit volume is given by w = Jf E.
Therefore, if we integrate this expression over an arbitrary volume V, the total rate of
loss of electromagnetic energy 'W will be
f wdr=
Jy
Jf
JrEdT
(21-53)
and we would now like to express this in terms of the field vectors. If we use J f as given
21-4
POYNTING'S THEOREM
357
by (21-22), we find that we can also write

t
c
dD
/ E ( v X H) dr / E dr
Jy
*v
dt
(21-54)
Combining (1-116) with (21-20), we can write the first integrand as

dB
E ( V X H) = H ( V X E) - V (E X H) = - H - V (E X H)
dt
If we now insert this into (21-54), use the divergence theorem (1-59) and transfer the
result to the other side of the equation, we obtain
r (
dD
dB \
f
- / E + H \ dr = iT+(f)(E
Jk\
dt
dt J
*s
X H) da
(21-55)
which is known as Poynting's theorem.

So far, our treatment has been perfectly general. However, it is easiest to interpret
this result if we assume completely l.i.h. media, and since we will apply it only to these
cases anyhow, we assume this to be the case. Using (21-40), we get E- (dD/dt) =
e E - ( d E / d t ) = d(\tE2)/dt\
similarly, we find that H - ( d B / d t ) = d[(B2/2n)]/dt.
Substituting these into (21-55), recalling that we are dealing with a volume V with fixed
boundaries so that we can interchange the order of differentiation and integration, and
reintroducing the integral expression for iV from (21-53), we finally get
d
I1 2
B2\
- f - e E + \dr=
dt Jy\ 2
2fi /
f Jf E dr + (f> (E X H) da
ss
(21-56)
Now the integrand on the left-hand side is just the sum of the expressions (10-85) and
(20-84) that we previously found to be the electric and magnetic energy densities,
respectively. If we now make the reasonable assumption that they can be interpreted in
exactly this same way when the fields vary with time, then the integrand will be the
total electromagnetic energy density
" = ue + um = 2 e 2
2/i
(21-57)
and the integral will be exactly the total electromagnetic energy contained in the
volume V. Consequently, the left-hand side of (21-56) represents the rate of decrease of
this total energy. Where does it go? The first term on the the right-hand side is the rate
at which this energy is being converted into heat. If we now invoke conservation of
energy, then whatever energy is not converted into heat must pass out from the volume
V across its bounding surface S. Since the remaining term has just this form of an
integral over the surface, we can interpret
((E X H) <ia = ( - j - )
S
dt
(21-58)
J through S
as being the rate of flow of energy across the bounding surface. The next thing we can
do is to give a similar interpretation to the integrand of (21-58); thus
S = E XH
(21-59)
will be taken to be the rate of flow of electromagnetic energy per unit area. In other
words, it is an energy current density or power flux and will be measured in watts/
(meter) 2 . The quantity S is called the Poynting vector, its direction is the same as that
of the instantaneous flow of energy, that is, it " points" in the direction of energy flow.
Although we have abstracted this interpretation of the integrand frojH the physical
358
MAXWELL'S EQUATIONS
meaning of the whole integral, it is certainly a plausible interpretation and it is one

which has turned out to be consistent and useful, particularly in its application to
time-dependent solutions of Maxwell's equations.
Even though most of the applications of this result are made to fields that change in
time, we can consider an example to see that this interpretation of S gives reasonable
and consistent results when applied to a simple case involving steady-state fields, that
is, time-independent ones.
Example
Cylinder with constant current. Consider a portion of a long straight cylindrical
conductor of length I and radius a as shown in Figure 21-5. We assume a constant
current in the z direction, which is distributed uniformly across the cross section so that
Jy = J j z = const. We want to find S at the surface just outside of the conductor.
This is similar to the situation depicted in Figure 12-7 and, in the paragraph
following (12-28), we found the electric field just outside the cylinder to be equal to the
constant value inside and given by E = J f / a making it parallel to the axis as shown. In
(15-19) we have the value of B outside the cylinder so that, in this case, H just outside
will be H = B/ju 0 , or
/
H
since Jj = l/ira1.
that
(21 60)
2
2Jfa*
"
Substituting these expressions into (21-59), and using (1-76), we find
Jf
S =
1
X
J?a
Jfa
(21 61)
a
2
*
~ 27^
"
We see, as is also clear from the figure, that S is normal to the surface and directed
radially inward so that there is a steady flow of energy into the conductor as S is
constant. Now the element of area da is directed outward from the surface; also, S is
y
Figure 21-5. Fields and Poynting vector for a cylinder
carrying a constant current.
21-5
ELECTROMAGNETIC M O M E N T U M
359
parallel to the ends of the cylinder. Therefore, the total rate at which energy is flowing
into the volume is given by
J2
(f)S da = S f da = S(2iral)
J
J
= (ita2l)
a
j 2
(volume)
a
(21-62)
with the use of (21-61) and where ira2l is the volume of the conductor. If we compare
this with (12-35), we see that (21-62) says that the total rate at which energy is flowing
into the conductor is exactly equal to the rate at which energy is being dissipated into
heat within the volume. This is exactly what is required for the steady-state situation we
have assumed. The ultimate source of this energy is some device, like a battery, which
maintains the potential difference between the ends of the conductor by continuously
doing work on the charges as they pass through the complete circuit. Our description
here involving S depicts the energy as being transferred from the source by means of
the fields that it establishes throughout space as a result of the charge and current
distributions within it. Finally, the energy flow passes normally through the surface of
the conductor in just the right amount to be transformed into heat. Thus the
interpretation given to the Poynting vector provides a complete and internally consistent description of the processes involved in this case.
21-5 ELECTROMAGNETIC MOMENTUM

We have just seen how energy density and energy flow can be ascribed to the
electromagnetic field. It is also possible to associate momentum with it and it is, in fact,
gratifying that this is so. We recall that the Coulomb forces between two point charges
are equal and opposite, that is, they are in agreement with Newton's third law. On the
other hand, we found in (13-19) that the forces between current elements as given by
(13-17) are not generally equal and opposite so that these forces do not satisfy the third
law. As we know from mechanics, it is the fact that the forces between the point masses
of a composite system are equal and opposite that is the basis for the initial derivation
of the conservation of linear momentum of an isolated system. Since this momentum
conservation property is such a basic and far-reaching general principle, it would be
extremely disturbing if our fundamental magnetic force law (13-17) somehow forced us
to conclude that conservation of momentum was no longer a valid general principle in
the presence of electromagnetic fields. Our aim now is to see how it can be retained.
For simplicity, however, we do this only for the vacuum case.
We consider a distribution of free charges and currents occupying a volume V and
described by the densities pj and Jy. If we let rfF be the force on the charges and
currents in a volume dr, we can combine (21-29) with (2-14) and (12-3) to get
d = (pfE + if X B) dr
(21-63)
Thus, we see that we can introduce a force per unit volume f v by means of
fy =
= P/E + Jf X B
(21-64)
Furthermore, the total electromagnetic force on our whole system can be obtained by
summing (21-63) over the whole distribution to give
F = J f v d r = j (PfE + JfX B) dt
(21-65)
and now, in accordance with our usual custom, we want to express this completely in
terms of the fields.
360
Setting P and M each equal to zero in (21-30) and (21-33), and using the results, we
find that the integrand of (21-65) can be written as
1
3E
(21-66)
f y = t 0 E ( V E) + ( v X B) X B - e:0
dt X B
Mo
We can improve the symmetry of this expression between E and B by adding zero to it
as written in the form
dB
1
e n E x
q E X ( v X E) + B ( V B) = 0
dt
Mo
which follows from (21-31) and (21-32). Doing this, and using (1-23), we get
Q
fv + o - ^ ( E X B) = e 0 [ E ( v E) - E x ( v X E)]
+ [B( v B) - B X ( v X B)]
(21-67)
Mo
In order to see what we have, let us calculate the x component of the first bracketed
term on the right; we find in a straightforward manner that it can be written as
d (
1 2 \
S
d
[ E ( v E) E X ( v x E)] - St -jE ) + ~^(E,E)
+ -^(E,E, )
and there will be a similar result for the term involving B. This can be written more
compactly if we introduce a set of quantities Ttj defined as
1
7)j = c01 E^j
\
+
-B2Su ,J
2
Mo
(21-68)
with the use of the Kronecker delta symbol defined in (8-27) and where the indices i
and j can independently take on the values x, y, and z. Using this definition, we see
that the x component of (21-67) can be written in the form
+ C
~dt
dTx>
dx
dT xy
dy
dT xi
dz
V X
(21-69)
where the vector X = TxJi + T y + Txzz. (The quantities TtJ are called the components of the Maxwell stress tensor, we will, however, not be using them again after this
section.) If we now integrate (21-69) over the volume V and apply the divergence
theorem (1-59), we get
+ * o ^ ( E X B)
dr = J V X dr = ^ X da
(21-70)
Now let us extend the region of integration to include all space which is possible since
pf and Jf are zero outside of V. Since all of the sources pf and Jf are within a finite
region, at large distances R, we will, at worst, have E ~ l/R2 and B 1/R3 as we saw
in (7-26) and (18-19), while the surface area of integration goes as R2. Now the
components of 7]- involve products of the field components, and we see from (21-69)
that X will decrease much more rapidly with R than the surface area of integration will
increase, so that its surface integral will vanish and (21-70) becomes
all space .
d
5/
dr
X da = 0
(21-71)
'all space
We will get the same values for the y and z components of this integral so that the
EXERCISES
361
vector integral vanishes and thus

(
all space
ft, + e 0 ( E X B) dr = F + /
vt
Ot 'all
marp
all space
c 0 (E X B) dr = 0
(21-72)
where we have used (21-65) to reintroduce the total force on the charges and currents.
Now F acts on the charge and current carriers, that is, on the matter in the field; the
matter has mechanical properties and we know from mechanics that F equals the rate
of change of the momentum of the matter, that is, F = dpmatter/ dt. Substituting this
into (21-72), and noting that there is no longer any distinction between a partial and
total time derivative in the second term since the integration is over all spacfe, we find
that (21-72) can be written as
d
dt
Pmatter
+ /
all space
0 (E X B) dr
= 0
(21-73)
so that the quantity in the brackets is a constant. But this is exactly the way in which we
would expect a statement of the conservation of the total momentum of the whole
system to be expressed. Therefore, we can interpret the integral as being the momentum
of the electromagnetic field and write
Pem field
C0E X B dr
âll space
(21-74)
Then it is only natural to interpret the integrand as being a momentum density g:

g = e 0 E X B = n0eQS
(21-75)
where S is the Poynting vector since B = ju 0 H in this case.

If we recall from mechanics that the angular momentum I of a point mass with linear
momentum p is defined by I = r X p where r is its position vector, we see that by
analogy we can define an angular momentum density 3? associated with the electromagnetic field by
s e = r X g = 0 r X (E X B)
(21-76)
EXERCISES
21-1 A parallel plate capacitor consists of two
circular plates of area S with vacuum between
them. It is connected to a battery of constant emf
<?. The plates are then slowly oscillated so that
they remain parallel but the separation d between them is varied as d d0 + d1 sin to/. Find
the magnetic field H between the plates produced
by the displacement current. Similarly, find H if
the capacitor is first disconnected from the battery and then the plates are oscillated in the same
manner.
21-2 An infinitely long cylindrical capacitor of
radii a and b (b > a) carries a free charge Ay per
unit length. The region between the plates is filled
with a nonmagnetic dielectric of conductivity a.
Show that at every point in the dielectric, the
conduction current is exactly compensated by the
displacement current so that no magnetic field is

produced in the interior. Find the rate of energy
dissipation per unit volume at a point a distance
p from the axis. Find the total rate of energy
dissipation for a length I of the dielectric and
show that this is equal to the rate of decrease of
the electrostatic energy of the capacitor.
21-3 Express the general form of Maxwell's
equations completely in terms of each of the
following pairs: (E, H), (D, B), and (D, H). In
other words, find the analogues of (21-30) through
(21-33) for each of these pairs.
21-4 Express the general boundary conditions
completely in terms of each of the following
pairs: (E, H), (D, B), and (D, H). In other words,
find the analogues of (21-34), (21-26), (21-27), and
(21-35) for each of these pairs.
362
21-5 Starting with the Lorentz force (21-29) and

Maxwell's equations for vacuum, derive Coulomb's law (2-3).
21-6 Assume that a certain distribution of free
charges and currents (p l5 Jj) results in an electromagnetic field Ej, B,, Hj, and D1( that is, they
are solutions of Maxwell's equations. Similarly,
another distribution (p 2 ,J 2 ) produces E 2 ,
H 2 , and D 2 . Verify the superposition property
directly by showing that E == Ej + E 2 , and so on,
are also solutions of Maxwell's equations for the
total source distributions pl + p2 and Jj + J 2 .
21-7 Find the form of Maxwell's equations in
terms of E and B for a linear isotropic but nonhomogeneous medium.
21-8 Here are some genuine electromagnetic
fields, that is, solutions of Maxwell's equations in
a vacuum. (Later, we will see how they can be
obtained.)
Ev = - H0n0u ( ) sin | ) sin ( kz - w / )
/ / v = / / ( ) k ( j sin | ) sin ( kz - u>t)
lJTX\
Hz = //() cos I I cos {kz cot)
All other components are zero and 0 < x < a.

/7 0 , a, k, and to are constants and fi0e0o}2 = k2
(a) Verify by direct substitution that

these satisfy Maxwell's equations for a vacuum
containing no free charges or currents, (b) Find
the displacement current, (c) Find the Poynting
vector, (d) At x = 0, there is a perfectly conducting wall coincident with the yz plane. Assume
that all fields are zero for negative x and find the
free surface charge density and free surface current density as a function of position on this wall.
+
(77/a)2,
21-9 Find the Poynting vector on the bounding

surface of region 1 of Figure 21-3. Find the total
rate at which energy is entering region 1 and then
show that it equals the rate at which the energy of
the capacitor is changing.
21-10 Suppose that the very long coaxial line of
Figure 20-14 is used as a transmission line between a battery and resistor. The battery of emf S
has its two terminals connected to the two conductors at one end of the line, while, at the other
end, the two conductors are connected by a resistor of resistance R. Find S in the region between
the conductors. Show that the total power passing
across a cross section of the line equals S 2 / R
and interpret. Will the direction of energy flow
change if the connections to the battery are interchanged?
21-11 A conducting sphere of radius a is uniformly magnetized with a magnetization of magnitude M. It also has a net charge Q. Find the
total angular momentum of the electromagnetic
field of this system.
21-12 Consider a l.i.h. medium where pf and Jy
are both zero in the region of interest. Show that
Maxwell's equations are invariant to the transformation
E' = c [ E c o s a -I- ( f a )
l/2
Bsina]
B' = c [ - (/u.e)1/2Esin a + Bcos a]
(21-77)
where C is a dimensionless constant and a is a

constant but arbitrary angle. In other words, if E
and B are solutions of Maxwell's equations, show
that then E' and B' are too. Consider the special
case a = m/2, and thus show that, in this sense,
the fields E and B can be interchanged. Illustrate
this last result with a simple sketch. Is the relation
we found after (16-40) compatible with these results? This property is often called the duality
property of the electromagnetic field.
22
SCALAR A N D VECTOR
POTENTIALS
In the previous chapter, we summarized electromagnetism in terms of the field vectors
themselves. We recall, however, that much of our previous work was simplified by the
introduction of the scalar and vector potentials, and it is natural to wonder if they can
also be usefully applied to these more general situations. It does turn out to be possible
to formulate the subject in an analogous way. These more general potentials are found
to be of the most use in discussing the production of fields from given sources and
actually we will not make extensive use of them until Chapter 28. Nevertheless, it is
convenient to discuss them in general terms at this point.
-1 THE POTENTIALS IN GENERAL

We begin by noting that the equation V B = 0, which we originally used to introduce
the vector potential A in (16-7), is still unchanged in form as shown in (21-21).
Therefore, we can always satisfy (21-21) by writing
B(r, t) = V X A(r, / )
(22-1)
The difference between (22-1) and (16-7) is that now both B and A are assumed to be
possible functions of time t as well as of position r.
If we substitute (22-1) into (21-20), we obtain V X E = d{ V X A ) / d t = V X
( dA/dt) so that
/
dA \
v x E + 1=0
(22-2)
Because of (1-48), the term in parentheses can always be written as the gradient of a
scalar, that is, E -+- (dA/dt) = V<, or
dA
E
= - V < ^ -
(22-3)
This is, of course, of the same form as (17-12) since they both were derived in exactly
the same way. Again, we must assume that the scalar potential 4> is generally a function
of both position and time.
Therefore, two of the set of Maxwell's equations will always be satisfied if E and B
are written in the form (22-1) and (22-3). The remaining two members of the set should
tell us how A and <j> are to be found from the sources. If we substitute these expressions
for E and B into (21-30) and (21-33), we obtain
d
V2^> + V - A =
dt
d2A
I
v
V-A + ^
1 ,
( p / - V - P)
d<j>\
0
- ^
(
0
U + V X M
(22-4)
dP\
+ -
363
364
SCALAR A N D VECTOR POTENTIALS
with the use of (1-45) and (1-120). If we can now find solutions of these equations that
satisfy appropriate boundary conditions, then we know from the way in which they
were obtained that E and B as found from this A and <f> will automatically be a solution
of all of Maxwell's equations.
A difficulty with these two equations is that both A and <$> appear in each equation,
that is, the equations are not separated. In addition, if P and M do not represent
distributions of permanent dipole moments that would be unaffected by the fields, then
we have to assume that P and M are functions of E and B and thus of A and <j>.
Consequently, unless we have a specific dependence given to us, we cannot make any
further progress with the general equations. Nonetheless, (22-1), (22-3), (22-4), and
(22-5) are a set of equations that are assumed to be always true.
We can still draw some conclusions about boundary conditions at a surface of
discontinuity in properties. The arguments that we used in connection with (16-21),
which followed from B = v X A, are still valid when B is a function of the time, since
we only required that B be finite. Since (22-1) still holds, we can conclude that the
tangential components of A are always continuous:
A 2 , = A1,
(22-6)
Now we do not yet have a specific expression for v A; thus we cannot be sure of the
behavior of the normal components of A, but it is not unreasonable to expect that they
will behave as in (16-20) and also be continuous.
In Section 9-5, we concluded that the electrostatic scalar potential was continuous.
Now, however, (22-3) is applicable rather than simply E = V<, but if we review our
previous arguments, we see that if we add to them the usual requirement that dA/dt
remain finite as the transition layer is shrunk to zero thickness we will again find that 9
is continuous:
4*2 = $1
(22-7)
Finally, as a sort of check on our results, if we revert to the static case in which all
time derivatives are zero, and use V A = 0 as given by (16-17), we find that (22-4) and
(22-5) become v 2<f> = p/e0 a n d V 2A = ju 0 J, which are exactly our previous results
given by (11-1), (10-38), and (16-18). Thus our generalized scalar and vector potentials
properly reduce to the previous ones that were applicable to static situations.
2 2 - 2 THE POTENTIALS FOR LINEAR ISOTROPIC HOMOGENEOUS MEDIA
As in Section 21-3, considerable simplification can be attained if we are dealing with

completely l.i.h. materials, so that the constitutive equations (21-40) are applicable as is
(21-41) for the free current density. It is somewhat more convenient to begin again with
the forms of Maxwell's equations written for this case in (21-42) through (21-45). The
middle two will again be satisfied if we use (22-1) and (22-3), and when these
expressions for E and B are substituted into (21-42) and (21-45), we get
V A -
dA
dA
V 2< + V =
Ot
2
dA
I
~ VI V A +
We can now give (22-8) a form more like

terms and we find that we can also write
d<t>
d2^>
d
V * + J,
pf
(22-8)
d(j>
\
I=
(22-9)
(22-9) by adding and subtracting appropriate

it as
/
d(j>
\
pf
V A + e + pot
- - 7
(22-10)
22-2 THE POTENTIALS FOR LINEAR I S O T R O P I C H O M O G E N E O U S MEDIA
365
These equations are the ones that we have to solve for A and <p. However, they each
contain both A and (f> and it would be desirable if we could somehow separate them.
We recall from the Helmholtz theorem of Section 1-20 that the vector A will not be
completely determined until we have specified both v X A and v A everywhere. So
far, we have only specified V X A by way of (22-1); consequently, we are free to choose
V A in any convenient way. Comparing (22-9) and (22-10), we see that a simple thing
to do is to require that
d<j>
V A + litI- jno(j> = 0
(22-11)
at
for then
dA
d2\
- fia - Pe-gp;
VA
dip
~/iJ/
d2<j>
pf
- - 7
(22-12)
(22.13)
The requirement (22-11) is known as the Lorentz condition. The advantage of this
particular choice is that not only does it make the equations for A and <j> independent
of each other, but that the four scalar equations that result (one from each component
of A and one for <#>) are all differential equations of the same general form so that, in
this sense, we are left with only one general problem to solve.
The formulation we have developed here thus anticipates that the general procedure
to be used in solving an electromagnetic field problem by means of the potentials would
be as follows. We assume that the external sources plus boundary conditions are given,
that is, we assume that the functions pf(r, t) and J/(r, t) are known. Then (22-12) and
(22-13) are to be solved for A and <#>, making sure that they satisfy the Lorentz
condition (22-11). When this has been done, E and B are found from (22-1) and (22-3)
and the appropriate boundary conditions are satisfied as well. We know that E and B
as found in this way will satisfy Maxwell's equations and thus will represent the
complete solution. If we desire, we can then find D = eE and H = B/ju; we can also
calculate the polarization and magnetization everywhere from P = (x e l ) e 0 E and
M = [(, - l)/Kmn0]B.
Actually the most important and convenient situations for which these potentials are
used are for a nonconducting medium, and, since this will also be true for us, let us see
to what our results reduce in this case. When a = 0, 3f =
by (21-41), and (22-11)
through (22-13) become
d _ _Pf
V 2 < # > - ^ 2^ T = - ~
dt
e
These are often written more compactly by defining the D'Alembertian
(22-16)
operator as
d
= V2-M^-
so that we obtain
2A = -jLtJ,1
(22-17)
Jt
n2<t> = - ^
e
which are reminiscent of the Poisson equations (11-2) and (16-18).
(22-18)
366
SCALAR A N D VECTOR POTENTIALS
If we assume that d<p/dt

stays finite within the transition layer, we find from (22-14)
and (9-7) that the normal components of A are continuous across a surface of
discontinuity, that is,
A2, - Aln - 0
(22-19)
which, when combined with (22-6), shows us that A itself is continuous,

A 2 = AX
(22-20)
just as it was in the static case as found in (16-22).

22-3 GAUGE TRANSFORMATIONS
Near the end of Section 16-2, we saw that the requirement B = v X A still leaves some
ambiguity in the vector potential, and we also saw, in (16-24) and (16-25), that a new
potential A* defined by
At = A + V x
(22-21)
gives the induction B equally well. We see that this will still be the case here where we
would now expect the scalar x to be a function of time as well as of position, so that we
write x = X ( r , 0 - However, E now depends on the vector as well as the scalar
potentials, and of course we want to obtain the same electric field. If we substitute
(22-21) into (22-3), we get
d ,
,
(
K - - v* - ^(At - vx) - - v ( *
dx \
dtf
<9At
- v*t -
(22-22)
where we have set

<?X
<t>+ = <#> -
(22-23)
Thus, the form of (22-3) is preserved if we also change the scalar potential in this way.
In other words, we get the same fields E and B whether we use the set of potentials
(A, <j>) or (A*, $ ) . The kind of transformation defined by (22-21) and (22-23) is called a
gauge transformation', thus, we have found that Maxwell's equations are invariant to a
gauge transformation or to a "change in gauge." There are certain restrictions we want
to put on x, however, because it is desirable that both sets of potentials satisfy the
Lorentz condition. If we substitute (22-21) and (22-23) into (22-14), we find that
dtf
V A+ + j u e - V 2 X +
d2x
(22-24)
so that A+ and <j>* will also satisfy (22-14) provided that

d2X
V 2 x - J " ^ y = D2x = 0
(22-25)
[In the case of static fields, (22-14) and (22-25) reduce to our previous results (16-26)
and (16-27).]
It is also easily verified that under this assumption A1" and <$ satisfy the same
differential equations as do A and <:
2A+ = -iiJf
U2tf = -
(22-26)
When one deals with potentials that fulfill all of these requirements, it is said that the
Lorentz gauge is being used.
EXERCISES
367
EXERCISES
22-1 Show that (22-8) and (22-9) can also be
obtained by starting from (22-4) and (22-5) rather
than by going back to Maxwell's equations first.
22-2 Show that, if the free charge and current
distributions and the polarization and magnetization are all given functions of position and time,
then the general equations satisfied by the potentials can be separated by using the form of the
Lorentz condition appropriate for a vacuum. Find
the differential equations satisfied by A and <j>
under these conditions.
22-3 Find the equations satisfied by the potentials for a linear isotropic but won homogeneous
medium.
22-4 Find the boundary conditions satisfied by
the normal components of A from the Lorentz
condition (22-11) involving conductivity.
22-5 Consider a l.i.h. nonconducting medium. If
one requires that V A = 0, one is said to be
using the Coulomb gauge. Find the differential
equations satisfied by A and <f> for this case.
22-6 Show that, in a l.i.h. nonconducting region

where pf = 0 and Jf = 0, the fields E and B can
be found completely from a vector potential, that
is, 4> const, (zero is usually chosen). What two
equations determine A?
22-7 Consider a region where pf = 0, Jf = 0,
M 0, while P is a given function of position and time. Show that (22-4) and (22-5)
can be satisfied and the electromagnetic field obtained from a single vector ire such that A =
ju.0e0( dme/dt),
<\> = - V "ne and V 2 n e c
2
2
Mo o( d %./dt )
= P / c 0 . This vector tre is called
the Hertz vector or polarization potential. Express
E and B completely in terms of ire, and P also if
it leads to any simplification.
22-8 This exercise is similar in spirit to the
preceding one. Assume that pf = 0, Jf = 0, P = 0,
while M is a given function of position and time.
Show that the electromagnetic field can now be
obtained from a single vector it,, which satisfies
the equation V 2 ir, - f i 0 c 0 ( d 2 n m / d t 2 ) /i 0 M.
Find expressions for A, <j>, E, and B in terms of
23
SYSTEMS OF UNITS A
G U I D E FOR THE PERPLEXED
We have been using exclusively SI units which, for our purposes, are the same as those
of the rationalized MKSA system, and we will continue to do so. This means that our
unit system is based on four arbitrarily chosen and defined quantitiesthe meter,
kilogram, second, and ampere. Primarily because of historical reasons based on the
originally separate development of electricity and magnetism, other systems of units
have been used and continue to be used. This is particularly the case in more advanced
treatments of physics, especially in subjects like quantum mechanics and its application
to the microscopic properties of matter. As a result, a student whose training has been
conducted entirely in terms of MKSA units very often faces the problem of answering
two questions: In what system of units is this equation written? What numbers do I put
in it in order to work this problem? This chapter provides some guidance in how to
answer these questions. Consequently, we do not give an exhaustive discussion of
various unit systems but primarily try to indicate how they originated, the effects on the
forms of our basic equations, and what to do about it. Thus, in a sense, this short
chapter is a digression, but, since we have just finished the task of writing electromagnetism in its most general and fundamental form, this is a useful point at which to
consider these questions.
23-1 ORIGIN OF OTHER SYSTEMS OF UNITS

In order to see how different systems can occur, it is sufficient for our purposes to
consider two basic experimental resultsone electrostatic and one magnetostatic.
From Coulomb's law (2-3), we know that the magnitude of the force between two point
charges has the form
f = c
qq'
. j t
f23"1)
where Ce is a constant of proportionality whose numerical value will depend on the

units that are used. We previously chose to write Ce = l/477e 0 . Similarly, the magnitude of the force per unit length between two parallel currents as found from Ampere's
law is given by (13-13) and (13-14) and can be written as
d F
'
-T - 2Cm
dz
p
(23-2)
where Cm is another constant of proportionality that we previously wrote as Cm =

jti0/477\ In addition, all systems of units use the definition of current given by (12-2):
/ = dq/dt.
If one always uses the same set of mechanical units in them, then the two forces
involved will have the same dimensions and we see that the combinations Ceqq'/R2
368
23-2 THE ELECTROSTATIC A N D E L E C T R O M A G N E T I C SYSTEMS
and CmII'
369
also must have the same dimensions so that the ratio

j r =
(23-3)
must have the dimensions of (distance/time) 2 , that is, c has the dimensions of a speed.
The value of this ratio has been measured many times and the experimental result is
that
c = 3 X 10 8 meters/second
(23-4)
which is the same as the measured speed of light in a vacuum. The best value of c as
presently known is 2.99792458 X 10 8 meters/second, but the value given in (23-4) is
accurate enough for us here. (As we will see in the next chapter, the agreement between
this ratio and the speed of light is not accidental.) We have already taken this numerical
result into account in the values we gave for Ce and Cm and we find that Ce/Cm =
( 4 f f 0 ) - 1 / ( / x 0 / 4 w ) = (ju.0co) 1 = (9 X 1 0 9 ) / ( 1 0 ~ 7 ) = (3 X 10 8 meters/second) 2 with
the use of (2-5) and (13-2); this is in agreement with (23-3) and (23-4). In other words,
we have the fundamental result that
Moeo =
(23-5)
for the MKSA system that we are using.

The various systems of units used in electromagnetism essentially differ in the way in
which these constants are chosen. It is clear that either Ce or Cm can be chosen
arbitrarily, but then the value of the other is fixed by the requirement of (23-3).
All other systems of any interest are based on the use of the CGS system in which
everything is expressed in terms of the arbitrary choice of three fundamental units for
length, mass, and time; these are, respectively, the centimeter, gram, and second. The
mechanical units are then found in the usual way from their definitions. Thus, the force
unit is 1 gram X 1 centimeter/(second) 2 = 1 dyne. The unit of work or energy will be
the product of a unit force and a unit displacement: 1 dyne-centimeter = 1 erg. The
unit of power will be 1 erg/second, and so on.
Another distinction that is made between unit systems concerns whether they are
"rationalized" or "unrationalized." What this means, in effect, is that for a rationalized
system there are no numerical factors of 4it appearing in Maxwell's equations, while, as
we will see, they do appear if an unrationalized system is used. Equations 21-19
through 21-22 show that we are using rationalized MKSA units. (The use of a
rationalized system does not make An disappear, rather it simply means that
'S are
found elsewhere in results found from Maxwell's equations; thus, the choice of which
type to use is somewhat a matter of taste.)
23-2] THE ELECTROSTATIC AND ELECTROMAGNETIC SYSTEMS
Suppose that you felt that Coulomb's law was a fundamental result that was
place to start in defining a system of units for electromagnetism. Your
inclination would be to give this equation as simple an appearance as possible;
certainly be done by choosing Ce = 1. Then, from (23-3), Cm must be taken to
and (23-1) and (23-2) would become
qq'
dF
211'
F - - ~ 2r
= ~~2~
(esu)
R
dz
c2p
This procedure leads one to the electrostatic system of units (esu). We see from
the best
natural
this can
be 1 / c 2
(23-6)
the first
370
SYSTEMS O F U N I T S A G U I D E FOR THE PERPLEXED
expression in (23-6) that two equal unit charges a distance 1 centimeter apart will repel
each other with a force of 1 dyne; the unit of charge defined in this way is called a
statcoulomb (from electrostatic). The unit of current will then be 1 statcoulomb/second
= 1 statampere, and that of potential difference 1 erg/statcoulomb = 1 statvolt. One
can continue this process and define a statfarad, statohm, and so on, and in this way
develop a consistent and complete description. At some point, however, one has to
decide how to define B and relate it to E; this is done in this system by writing
Faraday s law as \ / X El
5 B o r , equivalently, the Iôrentz force as ^(E "f- v X B).
In this pure form, however, this system is seldom used anymore and we will not
describe it further. Nevertheless, it should be pointed out that it is very common to find
quantities measured in this system but not given in statamperes, statfarads, and so on,
but merely stated as being so many "electrostatic units" or just so many "esu."
Now suppose that you were more interested in and had more experience with
magnetostatics than in electrostatics; then you might feel that the expression (23-2) is a
better starting point and you would like to simplify it as much as possible. This can be
done by choosing Cm = 1 so that Ce = c2 and (23-1) and (23-2) become
c2qq'
F =
R
dF
dz
2II'
p
(emu)
(23-7)
Such a procedure leads to the electromagnetic system of units (emu). We now see from
the second expression in (23-7) that two very long equal parallel unit currents 1
centimeter apart will attract each other with a force of 2 dynes/centimeter; the unit of
current defined in this way is called an abampere (from absolute). The unit charge will
be 1 abcoulomb = 1 abampere-second and one can continue this process to get abvolts,
abfarads, and the like; it is also very common to use simply the terminology of
"electromagnetic units" or "emu." Again the definitions of E and B are related by
writing V X E = dB/dt or F = q(E + v X B). In this pure form, the electromagnetic system is practically never used. What is still very much used, however, and the
system which one needs to be able to deal with, is that which we consider next.
23-3 THE GAUSSIAN SYSTEM

This is an unrationalized CGS system that is mixed in the sense that electric quantities
are measured in electrostatic units while magnetic quantities are measured in electromagnetic units. For our purposes, it will suffice simply to quote the form that the basic
equations assume for this system. Maxwell's equations in general are
V D = 477pf
v x E =
1 3B
c dt
V B = 0
477
1 <9D
V x H = J ,1 +
c
c dt
(23-8)
where the various field vectors are related by

D = E + 4ttP
H = B - 4ttM
(23-9)
and the Lorentz force is

F = $ ( E + ~ X BJ
(23-10)
[It follows from (23-8) that the equation of continuity still has the form v J\f +
(dpf/dt)
= 0.]
23-4
H O W T O C O P E W I T H THE G A U S S I A N SYSTEM
Where they are applicable, the various constitutive equations are written
D = eE
H = B/ju
Jy = o E
M
P = XeE
= Xm H
371
(23-11)
so that
e = 1 + 4nxe
fi = 1 + 4t7X,
(23-12)
Expressions involving the potentials are easily seen to become

B = v XA
E=-v<#>
1 <9A
c at
(23-13)
while energy formulas of interest are

1
u = (E
77
D + B
H)
c
S = (EXH)
47T
(23-14)
where the first holds for linear media.

We see from the above that all of the field vectors E, D, B, H, P, and M have the
same dimensions', this, of course, has not kept people from giving different names to the
units. This is most prevalent with respect to magnetic quantities and the common usage
is as follows: B, gauss; H, oersted; M, oersted (but see the next section); , 1
gauss-(centimeter) 2 = 1 maxwell.
It also follows from (23-9) that in a vacuum, D = E and H = B. The quantities e, /x,
Xe, and Xm a r e a " dimensionless; we discuss their numerical values in the next section.
Furthermore, it is not uncommon to find a "modified" Gaussian system being used.
It is just like the one we have described except that, while charge is still measured in
statcoulombs (esu), current is measured in abamperes (emu). What this does, in effect,
is to replace any symbol for current by c times that symbol (e.g., Jf -> cif). The only
one of Maxwell's equations that is affected by this is Ampere's law, and then the
equation of continuity, which become
1 3D
V X H = 4tt J , +
'
c at
1 dp,
V J, + - V = 0
c at
(23-15)
Finally, the Heaviside-Lorentz system is simply a rationalized Gaussian system; when

this is used, the effect is to replace every 47T in the equations (23-8) through (23-12) by
unity; for example, one gets V D =
and D = E + P. The factors of c still remain,
however.
If a particular author does not state the unit system that is being used, one can
generally deduce what it is by looking at the form of some familiar results, preferably
Maxwell's equations.
23-4 H O W TO COPE WITH THE GAUSSIAN SYSTEM
In principle, any desired result in the Gaussian system can be derived by starting with
Maxwell's equations (23-8) and using the expressions (23-9) through (23-13) as required. This is not always convenient and it is desirable to have a method that will
enable one to transform a given result written in the Gaussian system into the
corresponding one in the MKSA system and vice versa. Table 23-1 provides a recipe
for doing this. In order to use this table, one replaces a symbol in the column labeled
by the system in which the formula is written by the symbol or combination listed for
the other system. Symbols representing essentially mechanical quantities are unchanged.
372
SYSTEMS O F UNITS A
Table 23-1.
G U I D E FOR THE PERPLEXED
Conversion of symbols in e q u a t i o n s
Symbols representing essentially mechanical quantities (length, mass,

time, force, work, energy, power, etc.) are not changed (nor are
derivatives). To convert an equation written in the MKSA system to the
corresponding one in the Gaussian system, replace the symbol listed
under the column labeled MKSA by that listed under Gaussian. The
entries can also be used to convert a Gaussian equation to an MKSA
one by going from right to left in the table.
Quantity
Gaussian
MKSA
Capacitance
Charge
Charge density
Conductivity
Current
Current density
Dielectric constant
Dipole moment (electric)
Dipole moment (magnetic)
Displacement
Electric field
Inductance
Magnetic field
Magnetic flux
Magnetic induction
Magnetization
Permeability
4 itCqC
(4T70)1/2<7
p,(a,
A)
(4t7t
1 / 2
p , ( a , A)
47r0CJ
(4770)1/2/
J,(K)
(4 f f t 0 ) 1 / 2 J> (K)
P
m
D
(4T70)1/2P
( 4 w / f i
(
1 / 2
1 / 2
/ 4 T 0
(4 W 0 )- 1 / 2 E
(47T 0 )- 1 L
(4^0)"1/2H
H
0
B
M
( ^ o / 4T7-)1/2B
(4Ô)1/2M
(1)
then
(2)
Permeability (relative)
Permittivity
t1
f1
(1) Ket0 , then
(4 W o ^ P
(4TT0)
(4T70)_1P
(2) Ke -
Polarization
Resistance
Resistivity
Scalar potential
Speed of light
Susceptibility
Vector potential
(4 77" o ) ~" 1 /
e
(Mo o)"
1 / 2
x e ,(x OT )
A
c
4 ^X e >(Xm)
(^0/4w)1/2A
Example
Let us transform V D = pf as given by (21-19). Using the table, we get
V [(E 0 /477-) 1/2 D] = (4T7E0)1/2py, which reduces to V D = 4irp f as quoted in (23-8).
Example
Let us transform the expression for the Poynting vector given in (23-14) into MKSA
form. Since S is an energy flow, it is not changed and, by going from the Gaussian
column to the MKSA one, we get
(Wo)
47T
which is exactly (21-59).
s
[(47T 0 ) 1/2 EX (4T7JU0)1/2H] = E X H
23-4 HOW TO COPE WITH THE GAUSSIAN SYSTEM
373
The use of Table 23-1 may occasionally lead to a wrong result when applied to a
Gaussian system equation that has been worked out for a vacuum. The reason is that,
since D = E and B = H in this case, there is a tendency to use these symbols
interchangeably which may lead to ambiguities since the conversion factors listed in
Table 23-1 are different for the members of each of these pairs. For example, in such a
situation it is quite common to find the equation connecting the field with the vector
potential written as H = V X A, and we quickly find from the table that this will not
transform directly back to the corresponding MKSA equation B = v X A, although it
does lead to /i 0 H = V X A which, for a vacuum, is all right.
The problem of converting the form of an equation is different from that of
converting the numerical values of a given physical quantity from one unit system to
another. For example, the data may be given numerically in Gaussian units and it is
necessary to insert their equivalent values into an MKSA formula, or conversely. For
this purpose, one requires a numerical conversion table and Table 23-2 is adequate for
most purposes. The entry in each row gives the same amount of the given quantity
expressed in different units; that is, the terms in each row are equal. The various factors
of 3 arise from writing c = 3 X 10 8 meters/second; this does not apply to powers of
10. Other needed conversions can be obtained in the usual manner by multiplying by
unity as expressed by the appropriate ratio and canceling units; for example, one can
write 1 = 10 3 grams/1 kilogram.
Although H and M are both measured in ampere/meter in the MKSA system, we
see that the conversion factors to oersted are different for each of them; this is a
consequence of the 4tr in (23-9), and similar remarks apply to D and P. It is not
unusual to find magnetization stated in gauss rather than in oersted; in the overwhelming majority of such cases, the author really means "oersted" and one can proceed by
Table 23-2.
Conversion table for numerical values
Quantity
MKSA
Length
Mass
Time
Force
Work, energy
Power
Capacitance ( C )
Charge ( q)
Charge density ( p )
Conductivity ( a )
Current ( / )
1 meter (m)
1 kilogram
1 second
1 newton
1 joule
1 watt
1 farad
1 coulomb
1 coulomb/m 3
1 (ohm-m) - 1
1 ampere
Current density (J)

Displacement (D)
Electric field (E)
Inductance ( L )
Magnetic field (H)
Magnetic flux (O)
Magnetic induction (B)
Magnetization (M)
Polarization (P)
Potential (<j>)
Resistance ( R )
1 ampere/m 2
1 coulomb/m 2
1 volt/m
1 henry
1 ampere/m
1 weber
1 weber/m 2 = 1 tesla
1 ampere/m
1 coulomb/m 2
1 volt
1 ohm
Gaussian
2
10 centimeters (cm)
10 3 grams
1 second
10 5 dynes
107 ergs
107 ergs/second
9 X 1011 statfarads
3 X 10 9 statcoulombs
3 X 103 statcoulomb/cm 3
9 X 10 9 (statohm-cm)~ 1
3 X 10 9 statamperes
= 10~ 1 abamperes
3 x 105 statampere/cm 2
1277 X 105 statvolt/cm
3 X 10 4 statvolt/cm
I X 10 11 stathenrys
477 x 10 3 oersted
10 8 maxwells
10 4 gauss
10" 3 oersted
3 X 105 statvolt/cm
statvolt
j X 10 11 statohms
374
SYSTEMS O F U N I T S A G U I D E FOR THE PERPLEXED
changing the name and using the factor given in the table for M. Occasionally, the
author really means "gauss"; this will normally signify that he or she has in mind an
MKSA definition of magnetic dipole moment equal to ju0 times the expression (19-20);
this would make the relation among the magnetic vectors take the form B = ju 0 H + M
rather than (21-24). In that case, it would be appropriate to measure B and M in the
same units; this is a rare situation, however.
When one looks up numerical values of the quantities permeability, dielectric
constant, and susceptibilities, one often finds them given in Gaussian terms. The
numerical relations among these parameters in the two systems are:
"cMKSA
I
V
)
0 / MKSA
" m MKSA
I
I
\ Mo / MKSA
XMKSA
^^XGaussian
where the last relation holds for both x e
an
^Gaussian
(23-16)
^Gaussian
(23-17)
(23-18)
d Xm- (Also see Exercise 20-17.)
EXERCISES
23-2 Show that 1 statcoulomb/(centimeter) 2 =

1 statvolt/centimeter. Also show that 1 statfarad
= 1 centimeter, and that 1 statohm = 1 s e c o n d /
centimeter.
s h o w that 1 Gaussian unit of inductance = 1

stathenry = 1 (second) 2 /centimeter. N o w use
(23-8) and (23-9) to find the self-inductance of a
length I of an infinitely long ideal solenoid of
cross-sectional area S, n turns per unit length,
and v a c u u m inside. Verify that your result is
consistent with Table 23-1 and the above result
for its dimensions.
23-3 Show that all of the equations (23-8)

through (23-13) can be obtained by applying Table 23-1 to the corresponding M K S A equations.
2 3 - 7 U s e (23-8) and (23-11) to derive Poynting's

theorem for a linear isotropic medium and thus
s h o w the suitability of the results quoted in (23-
23-4 Beginning with the equations stated in

Gaussian form, derive the differential equations
satisfied by A and <j> and the Lorentz condition
for a l.i.h. medium. Verify that they are the same
as those obtained with the use of Table 23-1.
14).
23-1 Using (23-6), express the dimensions of a

statcoulomb in terms of centimeters, grams, and
seconds. Similarly, use (23-7) to do the same for
an abampere.
23-5 U s e (23-8) and (23-9) to obtain the capacitance of a parallel plate capacitor of plate area A
and separation d with vacuum between the plates.
Verify that your result is consistent with Table
23-1 and Exercise 23-2.
23-6 U s e (23-8) to show that the induced emf
will be written in Gaussian units as <$ =
c~ l(d$/dt).
If self-inductance is also defined
in the usual way by $= L(dl/dt),
show that
the analogue of ( 1 7 - 5 5 ) must be L = <t>/cl. Then
23-8 A s a simple numerical exercise in the use

of Table 23-2, suppose that H and M are parallel
and that they have the values a a m p e r e / m e t e r
and ft a m p e r e / m e t e r , respectively, where a and
ft are numbers. Find B in w e b e r s / ( m e t e r ) 2 , B in
gauss, H in oersted, and M in oersted. Show that
w h e n the values just found for H and M are put
into (23-9), the same value of B is obtained as
f o u n d by direct use of the conversion factor for B
itself.
23-9 Verify that
rect. ( Hint: as in
specific numerical
tity and carry out
(23-16) through (23-18) are corthe previous exercise, choose a

value for the appropriate quanall of the conversions.)
24
PLANE WAVES
In principle, we regard any solution of Maxwell's equations that satisfies given
boundary conditions as a possible electromagnetic field that would be produced by the
appropriate distribution of charge and current. In practice, however, one does not go
about trying to solve Maxwell's equations indiscriminately, but rather looks for
solutions of a desired type, or for those thought to be suitable for the particular
situation of interest at the moment. We first discuss several examples of this latter
procedure for time-dependent fields, putting off for a while the question of how we
would produce these fields if we so desired.
In the rest of this book, except for Section 25-7 and portions of Appendix B, we
assume that we are dealing with media that are linear isotropic and homogeneous in all
properties.
124-11 SEPARATE EQUATIONS FOR E AND B

Let us begin by assuming that there are no external free charges and currents in the
region of interest, that is, we take p f = 0 and J / = 0. Then (21-42) through (21-45)
become
V E = 0
V x E =
(24-1)
<9B
(24-2)
dt
V B = 0
(24-3)
dE
V X B = fxa E + JNE
(24-4)
at
We can eliminate one of the fields in the following way. If we take the curl of (24-2),
use (1-120), (24-1), and (24-4), we find that
VX(VXE) = V(V-E)-V2E= - v
dE
d
E = - V x B
dt
d E
d2E
dE
v 2 E - /iff -
= 0
(24-5)
In exactly the same manner, v X ( v X B) will lead us to

V
dB
~di ~
tl
d2 B
~dtY =
ti
{2A-6)
We see that E and B separately satisfy the same equation. Thus, if \p(r, t) is any of the
376
PLANE WAVES
six rectangular components of E and B, we find that

d2\p
d\p
= 0
(24-7)
so that, in effect, we have only one scalar equation to solve.

This last result, however, does not mean that we can take any arbitrary six solutions
of (24-7), call them EX, EV,...,
BZ in any manner and regard the result as a possible
electromagnetic field. The reason is that the fields must still satisfy Maxwell's equations
given above and this imposes restrictions on E and B. This has occurred because we
differentiated Maxwell's equations to obtain (24-5) through (24-7) and the process of
differentiation always results in some loss of information. Nevertheless, we can continue in this manner, but we will eventually have to satisfy (24-1) through (24-4) as well.
It will now be convenient to simplify things a bit more.
24-2 PLANE WAVES IN A NONCONDUCTING MEDIUM

First, let us assume that the medium is nonconducting so that a = 0; then (24-7) takes
the form
2
V yp
d2^
(24-8)
" e ^ - 0
which is known as the three-dimensional wave equation. Although we will come back to
(24-8), let us look for a function \p of the special type \[/ = \p(z, t). In other words, for
any t, xp is independent of x and y so that, for each value of z, xp is constant on the
corresponding infinite plane parallel to the xy plane; one such plane is indicated in
Figure 24-1. In this case, (24-8) becomes
d2\p
d2xp
(24-9)
/it 2 = o
dz2
fr
which is the one-dimensional wave equation.
In essence, the term wave is used to describe a pattern (or form) that propagates,
that is, travels along. Then a function / ( z , t) = / ( z vt) with z and t appearing only
in the combination z vt where v is a constant can be taken to be a wave traveling in
the positive z direction with velocity v. (This is called a plane wave since / is constant
on the plane z = const.) We can verify the above with the aid of Figure 24-2. Consider
the value of / at t = 0 and z = 0: / ( 0 ) = / 0 . At a later time / 0 , / will again equal the
| \p = const, i
Figure 24-1. A p l a n e of constant t|// w h i c h is parallel t o

t h e xy plane.
2 4 - 2 P L A N E W A V E S IN A N O N C O N D U C T I N G M E D I U M
=0
7o
Figure 24-2.
/
4f
/Vo
<
1
1
//
v
\
^
\
^v
2
20
377
2<T
\
The shape at t = 0 has m o v e d a distance z0 in a t i m e t0.
same value / 0 at a position z0 such that the argument is again zero, that is,
f(z0 ~ vt0) = /o = /(0), so that
z
o ~
(24-10)
vt
In other words, the particular feature f fQ, which appeared at z = 0 and t = 0, now
has moved to the new position z 0 given by (24-10), that is, to a point reached by
traveling with constant velocity v. Similar remarks apply to the other values of / at
t = 0, and the net result is that the whole picture has been shifted in the direction of
positive z by the amount vt0. Similarly, one can show that g(z + vt) represents a wave
of arbitrary shape moving in the direction of negative z, that is, to the left, with speed v
and hence velocity v.
We now want to show that a completely general solution of (24-9) can be written in
the form
\p(z, t) = f ( z - vt) + g(z + vt)
(24-11)
where / and g are arbitrary functions. In order to do this, let w = z vt\ then we can
write / = f(w) and we find that
df dw
= = dw dz
dw
dz
df
d2f
and
dz2
d2f dw dj_d2f
dw2 dz
dw2
while
df
dt
df dw
dw dt
df
dw
andH
d2f
2
dt
d2f
dw
dw
d^f_
dt
dw1
Substituting these into (24-9), we get

d2f
which shows that f ( z
vt) is a solution provided that v2[xe = 1, or

v =
1
v^7
(24-12)
Similarly, g(z + vt) can be shown to be a solution of (24-9) with v given by the same
expression (24-12). Thus we see already that our electromagnetic fields can have the
form of plane waves propagating in the z direction and the speed of the wave is
determined solely by the electromagnetic properties of the medium as given by the
product ju.
378
PLANE W A V E S
**"
<4
It is usual to write (24-12) in a different manner; if we use (10-53), (20-55), and

(23-5), we find that we can write
v = -
(24-13)
n = ]/KeKm
(24-14)
where n is called the index of refraction of the medium and is a dimensionless quantity.
For a vacuum, n = 1 and the speed of the wave becomes simply c, which, as we
remarked following (23-4), is the same as the measured speed of light in a vacuum. This
result was first obtained by Maxwell and was taken by him, and others since then, as
strong evidence for believing that light waves are actually electromagnetic waves. Since
Maxwell's time, much more evidence has been accumulated in support of this ideaso
much, in fact, that nowadays it is universally accepted. Furthermore, the concept of
electromagnetic waves has been extended both experimentally and theoreticaliyi beyond
merely that of describing visible light.
\
Substituting (24-12) into (24-9), we find that we can now write the wave equation in
the form
^
d2^
1 d
(24-15)
v2 dt2
dz
Although we know that (24-11) represents a general solution of this equation, it is too
general for our purposes, and we want to consider solutions of a more specific form. In
order to get an idea of the type we want to study, let us begin again with (24-15) and
try the method of separation of variables that we found so helpful in Sections 11-4 and
11-5; in other words, let us try the form \p(z, t) = Z(z)T(t).
Substituting this into
(24-15), dividing by ZT, and proceeding in the usual manner, we get
1 d2Z
1 d2T
r- =
COnSt. =
2
Z dz2
V2T dt
so that we have the two differential equations
d2Z
dz2
+ k2Z = 0
d2T
T + w2r = 0
dt
and
(24-16)
where
CO2
k2 = - y
(24-17)
The general solutions of (24-16) can be written as Zk(z) = ake'kz + fike~,kz and
Tk(t) = ykeiut + 8ke~iuI where ak, fik, yk, and 5^ are constants and i = i/ 1 . If we
multiply these two results together we get a solution to the wave equation for this
particular value of k and the corresponding co given by (24-17). Since there may well be
many possible values of k, and since (24-15) is a linear differential equation, the sum of
solutions each of this form will also be a solution. Thus, if we sum the products ZkTk
over all possible values of k, we get the general solution to the wave equation, and we
find that it has the form
*(z, 0
+ Pkyke-*kz-u,)]
=
k
+ E [akyke^kz
+ ut)
+ pk8ke-'(k2
+ l)
"]
(24-18)
The arguments of the exponential terms are seen, with the use of (24-17), to all be of
^
CiC
24-2
P L A N E W A V E S IN A N O N C O N D U C T I N G M E D I U M
379
the form kz + <x>t = k[z + (co/k)t] = k(z + vt) so that \p actually has turned out to
look like (24-11), and, in fact, the first sum in (24-18) can be identified with / while the
second sum represents g. Thus we see that we have found that the solution of the
one-dimensional wave equation can be written as a superposition of sinusoidal plane
waves.
Actually, we can reduce the complexity of (24-18) even more. Consider the term that
involves e'(kz_<0,); for reasons that will become clear shortly, we will always take co to
be a positive quantity. Now if k is also positive, this form represents a plane wave
traveling in the direction of positive z with speed v = co/k. If k is negative, we can
write k = |fc|, and this takes the form e~,(-Wz + u , ) , which is a plane wave in the
direction of negative z with speed v = ic/\k\. Therefore, as far as the first term in each
sum of (24-18) is concerned, we need only concern ourselves with the single form
e i(kz-ut) a n c j t ^ e s j g n Qf (positive or negative) will tell the sense of travel of the wave.
The other exponential terms in (24-18) are simply the complex conjugates of those we
have just discussed and still have the appropriate form. Accordingly, for our purposes it
will be sufficient to study the behavior of a single prototype plane wave of a particular
value of k (and corresponding to) that we can take to be of the form
xP(z, t) = xp0ei(kz~ut)
(24-19)
where \p0 is a constant; the general solution can be constructed as a sum of terms
found from this as shown in (24-18). [The form (24-19) is the normal way of
representing a plane wave in physics; however, in some other books, the exponential
term is written as e '( u, ~ kz '> > that is, as the complex conjugate of (24-19). In addition, the
notation j rather than i for
1 is often used. Generally speaking, the results can
usually be compared by making the replacement of i by i or + j as the case may be.]
Since ip is to be a component of E or B, it is a physical quantity and hence must be
real. However, it is more convenient to deal with exponential functions than with sines
and cosines, so that it is normal practice to continue to write \p in the form (24-19) and
to supplement this with a convention: If \p (or E or B) is written or found in complex
form, then the real part is taken to be the physically significant solution, that is,
^ physical =
(24-20)
(This convention is possible because the wave equation is linear so that the real part of
tp and its imaginary part are separately solutions.) Since it will usually turn out that the
amplitude \p0 is itself complex, it will be useful to have explicit forms for the real part.
If we write \p0 in terms of its real part \p0R and its imaginary part \pQ] as
&0
R +
(24-21)
and use
e'u = cos u + i sin u
e~'u = cos u i sin u
(24-22)
/
cos
: z 03
then we find that (24-19) becomes \p = (ipo/t + /i/ 0 /)[ (^
0 + ' sin(/cz co?)L
so that
Re \p = \p0R cos (kz cot) \p0I sin (kz cot)
(24-23)
Another useful way of writing a complex number is in terms of a real amplitude \p0a
and phase angle
(24-24)
We see that these various quantities are related by
^0R
,
^0a
/ ,2
COS
$0a ~ {^PoR
,+ , 2
^0/
\J/2
0/)
^0a S ^ n ^
_ t o i
tan ^
OR
(24-25)
380
PLANE WAVES
When (24-24) is substituted into (24-19) it becomes ip = 4'oa e ' (kz
u + ) so
" ^
that
Re \J/ = tj/0a cos (kz cot + &)
(24-26)
which is often a very convenient form to use. A complex amplitude \p0 thus means that
there exists a phase factor & in the sinusoidal plane wave.
Equation 24-26, for example, shows us that a sinusoidal plane wave is periodic
both in space and in time. The spatial period A is called the wavelength, and the
temporal period T is related to the frequency v by T = 1 /v. These quantities can be
related to k and w as follows. Since the period of a cosine is 2 it, the function in (24-26)
will repeat itself whenever its argument changes by 2tt. Thus, for a fixed t, A| (kz cot
+ #)| = 2tt = |&Az| = |A:|X so that \k\ = 2tt/\.
Similarly, at a fixed position,
A | ( k z cot + ft)I = 2tt = coAr = coT and co = 2m/T = 2itv. Hence the relations are
2 -tt
\k\ =
2 TT
a = 2ttv =
(24-27)
From (24-17), we have | A: | = co/v and when (24-27) is combined with this, we get the
familiar result for a sinusoidal plane wave
v = v\
(24-28)
-1
The quantity k is called the propagation constant and is measured in (meter) ; co is

the (angular) frequency and is in (radian)/second while v is in hertz as we previously
discussed before (17-14). As a result of all of these relations, there is a large number of
different ways in which expressions involving a plane wave can be written; we will
generally continue with (24-19) for economy in notation.
The argument of the exponential, kz cot, is called the phase of the wave, so that
the velocity v is the rate at which a definite value of this phase travels; accordingly, v is
known as the phase velocity.
The three characteristics of the wave (k, co, v) are related by our previous result
(24-17). Now v is determined by the properties of the medium as given by (24-12); thus,
the remaining two (k and co) cannot be chosen independently. Since co (or v) is usually
determined by the source which is producing the wave, it is customary to regard the
frequency as the independent variable characterizing the wave, so that k becomes the
dependent variable of this set.
Since we have decided to concentrate on \f/ in the form (24-19), all of the
components of E and B will be of this form as well, that is, for a sinusoidal plane wave
we will have
E = Eoe'^-"0
B = BQeiikz~ut)
(24-29)
where E 0 and B0 are appropriate constant vector amplitudes that must be related so as
to satisfy Maxwell's equations. These forms enable us to simplify the writing of these
equations. Since \f/ is a function of z and t only, we find from (24-19) that
d\b = d\b = 0
d\l>= ik
d\b
24 3
ir
ir
ax
oy
~r
+
oz
at
( " )
and therefore V E = dEz/dz = ikEz and V B = ikBz. Similarly, we find that V X E

= ik( Eyx + Exy) with a corresponding expression for V X B; also dB/dt = ;wB.
Substituting these expressions into (24-1) through (24-4), setting a = 0, and using
(24-12), we find that Maxwell's equations for a plane wave propagating in the z
direction in a nonconducting medium become
kEz = 0
kBz = 0
w
/
^
(24-31)
k(-Eyx
+ E j ) = UB
k(-Byx
+ B j ) = --yE
24-2
P L A N E W A V E S IN A N O N C O N D U C T I N G M E D I U M
381
These can be written more compactly and usefully by noting that Ez = z E and
-Eyx + EJ = zX E with similar expressions for B; thus we can write (24-31) as
kz E = 0
kz B = 0
kz X E = coB
kz X B = - 4 e
(24-32)
v
and we remember that the direction of propagation is given by + z, corresponding to
the sign of k.
Suppose that k = 0; then we see that it is possible to have Ez = 0 and Bz # 0. But
we see from (24-17) that, in this case, co = 0 as well, so that we actually have a static
situation. While this is certainly possible, it is not our primary interest here.
For a nonstatic case, co # 0 and therefore k = 0. Then we see from the first two
equations of (24-32) that we must have Ez = 0 and Bz = 0. Thus, both E and B have
no component in the direction of propagation, that is, both E and B are perpendicular
to it; in other words, the electromagnetic plane wave is a transverse wave. We see from
kz X E = <oB that B is perpendicular to E, which also follows from the last equation of
(24-32). This all means that these vectors form a mutually perpendicular set; this is
illustrated for positive k in Figure 24-3. As also shown in the figure, the Poynting
vector S = E X H = (E X B)/ju is itself in the direction of propagation. If we solve the
third equation (24-32) and use (24-17), we find that
k
1
B = z X E = z X E
(24-33)
co
v
showing the magnitudes of the fields to be related by
Ifcl
1
n
IB | = | E | = | E | = y 7 | E | = | E |
(24-34)
c o v
c
so that, for a vacuum, | B | / 1 E | = 1 / c .
We see from (24-29) that the above relations also hold for the vector amplitudes E 0
and B 0 since the exponential factors will cancel from both sides. In addition, if E 0 has
the complex form, E 0 = E0ae'9, then these relations hold for the real amplitudes E 0 a
and B 0 a as well. Consequently, if we were to take the real parts of our solutions, as we
did to get (24-26), we will obtain
E
reai
o a c o s (kz - cot + &)
B real = B 0 a cos (kz - cot + # )
(24-35)
Since the arguments of the cosines are the same in both cases, the two fields are said to
be in phasethey become zero together, reach their maximum together, and so on.
This is illustrated for positive k and a definite time in Figure 24-4, that is the figure
shows a "photograph" of the fields as a function of distance along the direction of
AE
Figure 24-3. Relations a m o n g t h e field vectors,

direction of propagation, and energy f l o w for a plane
transverse w a v e .
382
PLANE WAVES
Figure 24-4.
Fields in a plane transverse w a v e at a given time.
travel. For an observer at a fixed point (z = const.), the time dependence of the fields
described by (24-35) can be visualized by imagining this picture moving in the direction
of the positive z axis with speed v. [For simplicity, we have assumed that E (and B)
always lies in the same plane, that is, it is linearly polarized; we return to this in Section
24-7.]
Although we haven't explicitly used the last relation of (24-32), that is, kz X B =
( w / u 2 ) E , it is easily verified that it is consistent with (24-33).
24-3 PLANE WAVES IN A CONDUCTING MEDIUM

We now assume that a # 0, but we will continue to assume that pf = 0 and J j = 0. The
appropriate equation for any component of E and B is now (24-7), which becomes
d2\b
dip
d2\p
= 0
(24-36)
when ip = *K-z, 0 - We again try to solve this equation with a plane wave of the form
(24-19), and when this is substituted into (24-36), we find that k and co must now be
related b y the dispersion relation
k2 = co2jue + iiiOfxa
(24-37)
rather than (24-17). Since we are only interested in fields that vary harmonically in
time, that is, are not damped in time, we require co to be real as well as positive.
Therefore, (24-37) can be satisfied only if A: is a complex quantity; accordingly we
assume k to have the form
k = (a + ifi)
(24-38)
where a and /3 are real and positive. Substituting this into (24-37), we obtain
a2 ft2 + 2/ayS = co2/! e
/COJLICT
(24-39)
Since the real and imaginary parts must be separately equal, we get
a
/3 2 = c o V
(24-40)
2 a / ? = cojuer
(24-41)
24-3
P L A N E W A V E S IN A C O N D U C T I N G ' M E D I U M
383
where we have assumed /i, e, and a to be real; we return to this point in Section 24-8.
These can be solved for a and /? and the results are
a
co
1/2
He
1 +
cot
+ 1
(24-42)
1/2
1 + I I - 1
coe
(24-43)
(We note that when a = 0, these reduce to a =

= co/v and f3 = 0 as they
should.)
T hese results are often written in terms of a dimensionless parameter Q, which, for
reasons we will consider later, is defined by
coe
Q =
(24-44)
so that
a = co
&=
/X
2
W
^T
1/2
1 +
l +
Q2
+ 1
(24-45)
1/2
(24-46)
Q1
Furthermore, it will often be useful to write k in the form

k = \k\ein
(24-47)
so that if we use (24-22) and (24-38), we get k = a + ifi = |&|(cos2 + i sin 2), and
therefore,
a = | k | cos 2
j3 = | k \ sin 2
(24-48)
| A: | = ( a 2 +
tan 2
= Q
= co^/juT 1 +
l +
1/4
(24-49)
= /L + Q2
Q:
(24-50)
Now that we have all of these fairly complicated results, let us see what their
significance is, and, in particular, what are the physical consequences of a complex
propagation constant k.
Substituting k = a + ifi into (24-19), we find that it becomes
* = -PzpKaz-ut)
(24-51)
since i2 = 1. A function of this form is called a damped traveling wave since its
amplitude is not constant, but decreases with distance in the direction of propagation
because of the factor e~^ z . The origin of this term arises from the existence of the
conductivity since /3 = 0 only when a = 0. Evidently it must be related to the loss of
energy of the wave because of the resistive dissipation of energy into heat, and we will
consider this in more detail shortly. Comparing the form of the propagation term with
/ of (24-11), we find that the velocity in this case is given by
co
I [1 + (1/Q2)]1/2 + 1
(24-52)
384
PLANE WAVES
and we see that in general v < (juc)~ 1 / 2 or < ^nonconductor because of (24-12). Thus
another effect of the conductivity of the medium is to make the wave -travel" - more
slowly. The wavelength can be found in the same manner by which we got (24-27), and
we find that
27t
x =
2m-v
=
"V
2m I
=
^ \ [ i
2
+ (i/e2)]1/2 + i /
/
(24
"53)
which is, of course, in agreement with (24-28). Since v has decreased, we see that
^ < ^nonconductor s o that the same frequency wave has its wavelength shortened as
compared to what it would be in a nonconducting medium with the same value of fie.
Since Q as given by (24-44) depends on the frequency co, (24-52) shows that the wave
velocity v is no longer a constant but is also a function of frequency, that is, we have
v = u(co). This phenomenon is called dispersion and hence a conducting medium is an
example of what is known as a dispersive medium. In order to see the principal
consequence of this, let us suppose that we are dealing with a more complicated field
that is written as a superposition of sinusoidal traveling waves as is, say, the first sum of
(24-18). Since k is no longer a constant but a function of co, each component wave in
the sum will travel at a different speed, so that, at a later time, the numerical value of
the superposition at a given z will be different, that is, the superposition will have been
changed in form. This is often described by saying that the superposition will alter its
shape as it travels along.
From the damping factor e~^ z in (24-51), we find that the distance of travel in which
the amplitude decreases by a factor of \/e is Az = 1/yS. This is generally written as 8
and is called the attenuation distance or skin depth and is a convenient measure of the
damping, since one can think of the wave as having essentially disappeared after
traveling a few values of 5. It is possible to write this in a variety of ways with the use
of (24-41), (24-52), (24-53), and (24-28):
1
2a
2
47t
2
- =
(24-54)
fi
juaco
jaav
juctcoA
fiav A
where X is the wavelength in the medium.
For completeness, we should check to see that all of our conclusions are also
appropriate for a wave traveling in the direction of negative z. Substituting k =
- ( a + i[i) into (24-19), we find that it becomes
^ = ^^gfiZg-Kaz + ut)
(24-55)
and, on comparing this with (24-11), we see that it does represent a wave traveling in
the correct sense with the same speed v = co/a as found in (24-52). Furthermore, since
the values of z are now decreasing, the exponential factor efiz will decrease as the wave
progresses, so that it still represents an attenuation.
There is another interesting and often very useful way of describing and thinking
about our results. If we make a purely formal comparison between (24-19), (24-11), and
(24-13), we see that we can always define a "wave velocity" V and an "index of
refraction" N by
<0
CO
e~iQ
(24-56)
k
a + ifi
1*1
( c/3\
c
(c/S
c
ck
/ ca\
+ j(
= - + /
=
CO /
V
\ CO
~ V "
l
c/
= n' -+ - 1 f M
l <0 /
(24-57)
24-3
P L A N E W A V E S IN A C O N D U C T I N G M E D I U M
385
where n' = c/v is what one would call the usual index of refraction since it reduces to
the value appropriate for a nonconductor as given by (24-14). Thus we can say that an
absorbing medium (j8 # 0) can be described as having a complex index of refraction,
with the appearance of the imaginary part being associated with attenuation.
So far, everything we have concluded about the effects of conductivity has been
based on the behavior of a single component ip. Now let us turn to the relation between
the fields E and B. If we review our work from (24-29) on, we see that nothing
depended on k being real; therefore, Maxwell's equations in the form (24-32) are still
valid, except for the last one which came from (24-4) and that now takes the form
H x B = (jueo) + /ju.cr )E
(24-58)
Therefore, the fields are still transverse to the direction of propagation, while
k
B = - z X E
(24-59)
CO
again by (24-33). Thus all that is required is to see if (24-58) is satisfied. If we substitute
(24-59) into it, use (1-30) and z E = 0, we find that it becomes ( k 2 jueto2 i/xau)E
= 0, which is satisfied with E = 0 because of the dispersion relation (24-37). Now k is
complex, and if we use the form (24-47) in (24-59), we find that the fields are now
related by
B =
1*1
r,
e'az X E
CO
(24-60)
In order to see the significance of the phase factor e'Q, we will find the real parts of E
and B. If we again write E 0 in (24-29) as E0ael&, where E 0 a is real, and use (24-38) as
well, we get
E = E0oe-*V("-"'+d)
B =
I Arl
% x E0ae~^zei(az^"'+^ + a)
(24-61)
03
so that
E
real =
real
Ikl
03
O
X
S
E
0ae
((XZ ~
03t +
COS
&)
( a Z ~ W/ + # + 2)
(24-62)
and we can write the amplitude of B real as B0a = (\k\/03)Z

X E 0 f l . Comparing these
expressions with (24-35), we see that in a conducting medium E and B are no longer in
phase and the phase difference 2 is positive according to (24-50). This means that E
and B no longer reach their maxima and minima together, nor do they vanish together.
We can compare them quantitatively. Suppose the phase, that is, the argument of the
cosine, has the definite value P at a given position z0 and a time tE for the electric
field, that is, P = az0 cotE + ft. Then, the induction will have reached the same
relative value at this point at a time tB when the phase has the same value, so that
P = az0 U3te + t> + 2. Equating these expressions, we find that the times are related
by
tB
tE +
2
03
> tE
(24-63)
which shows that B reaches its maximum value, for example, at this point at a later
time than does E; in other words, B lags E as a function of time. On the other hand, at
a given time t0, the locations at which B and E have the same phase are seen to be
386
PLANE W A V E S
E
Figure 24-5. Fields in an

attenuated plane w a v e in
a c o n d u c t o r at a given time.
related by
zE ~
< zE
(24-64)
so that we can speak of B leading E as a function of position. This effect, along with the
attenuation, is illustrated schematically in Figure 24-5, which shows the situation at a
given time and for positive k. (This figure should be compared with Figure 24-4.)
Another effect that follows from (24-62) is an alteration in the relative magnitudes,
and, with the use of (24-49), we find that
B-L . M .
I
real I
* r
(24.65)
The value of this ratio is greater than the value -/jit applicable to a nonconductor, as is
seen from (24-34). Thus the magnitude of B compared to that of E is relatively greater
in a conductor than in a nonconductor.
[Actually, in a purely formal sense, one can justifiably assert that things have not
changed at all. If we use (24-56) in (24-32), we can write B = (k/u>)z X E = (z X E)/V
and when we compare this with (24-33), we see that in both cases the (complex) B is
simply equal t o z X E divided by the appropriate "wave velocity."]
All of our results so far have been exact, although somewhat complicated. It is
convenient and useful to obtain approximations to these results that are applicable to
two extreme limiting cases. This process is most easily done in terms of the dimensionless parameter Q defined in (24-44). First of all, the definition is not frivolous as Q can
be seen to have a simple physical interpretation. For an electric field varying harmonically in time, that is, proportional to e~~iut, the magnitude of the displacement current
will be \dD/dt\ = \tdE/dt\ = cw|E|, while the magnitude of the conduction current
will be |Jy| = a | E | . If we now multiply the numerator and denominator of (24-44) by
|E|, we find that we can also write
\dto/dt\
Q - L-7V7-L
(24-66)
I J /l
so that Q can be regarded as a measure of the relative importance of the displacement

current and the conduction current. (Also see Exercise 24-29.) It is useful to use the
24-3
P L A N E W A V E S IN A C O N D U C T I N G M E D I U M
387
value of Q to characterize a material as an "insulator" or as a "good conductor" for

the given value of the frequency. We see from (24-44) that for small conductivity
a n d / o r high frequencies Q will be large, and we can say that we are dealing with an
insulator. On the other hand, for large conductivity a n d / o r low frequencies, Q will be
very small and we are dealing with a good conductor; this is usually the situation
applicable to metals. Accordingly, we can take our limiting cases as corresponding to
2 1 and Q ^ 1, and we shall approximate our previous results to get formulas that
are adequate and somewhat easier to use. We begin with the case more nearly like that
of the previous section.
1. "Insulator" (Q 1)
In this case, l/Q2
1, and we can use the approximation (1 + JC)" = 1 + nx for
x c 1. In this way, we find that, if we keep either the first-order correction term to
something that stays finite for a nonconductor, or the first-order term for something
that would vanish, our previous results become approximately
+ 57S I
P= s
\k\ = < o / ) i c | l +
/mc _ a / M 1 / 2
" 2(7)
lanSJ =
2TT
= | i - ^ i
/_
(24 67)
(24-68)
1
8Q-
flr = ^( 1 + ^ l
(24 70)
and, of course, there are only small differences between the results and those of the last
section.
2. "Good Conductor" (Q
1)
Proceeding in a similar manner, we find the approximate expressions to be (with

correction terms to order Q only):
\k\ = (ixau)1/2
tan 2 = 1 -Q
x=2
i Q
4^r{ -j )
(24-72)
(24
-">
It is actually common practice to approximate these even further by dropping the
388
PLANE WAVES
correction terms and we get

\1/2
/ 1
a =/3 = l-noui
(24-75)
tan 0 = 1, so that 2 = or 45
(24-76)
\ 1/2
"nonconductor
(24-77)
(24-78)
Comparing (24-74) and (24-70), we see that | B | is relatively much larger than | E | in a
good conductor as compared to an insulator, while the phase difference is virtually
always T T / A according to (24-76). Similarly, (24-77) shows that the wave speed is very
much smaller in the case of a good conductor, while (24-78) shows that the penetration
depth (attenuation distance) is of the order of a wavelength in the conductor, which in
turn is very much smaller than in an insulator.
A numerical example is helpful here. A typical metal such as copper has a = 6 X 10 7
(ohm-meter) - 1 and e = e 0 . Thus we find from (24-44) that
Q =
2"jn>e
( 2iTn .
o- j1p = 9 .X. 10~
^ i1y9f = in-18.
= /^
10
so that in order to have Q <s: 1, we must have v

10 18 hertz. This means that this
approximation is good all the way up in frequency until we get to the ultraviolet region,
and there we may run into quantum-mechanical effects in any case. In other words, the
case Q <c 1 really fits ordinary metals very well.
If we use this same value of a and take ju. = ju,0 in (24-78), we find that the
penetration depth 8 becomes (6.5 X 10 2)/vl/2
meter. At low frequencies, this is quite
large so that there is very little attenuation. But by the time we get to the lower limit of
the microwave region (c = 3 x 10 9 hertz), we find that 8 1 0 " 6 meter, so that the
fields are essentially different from zero only within this small distance; this is the
origin of the term "skin depth."
24-4 PLANE WAVES IN A CHARGED MEDIUM

Now that we have some experience, we can consider a more complicated situation. Let
us assume that there is a free charge distribution p}, within our medium; we still assume
that J f = 0 since a nonzero value corresponds to such a rarely occurring case. Thus,
instead of (24-1), we need to use v E = pf/e by (21-42), so that generally v E = 0.
The remaining equations (24-2) through (24-4) are unchanged. Now, when we eliminate
B by the same method we used before, we find that instead of (24-5), we obtain
5E
V E - ixa.
d2E
= V ( V E)
(24-79)
On the other hand, the equation satisfied by B alone is still (24-6):

V
since v B = 0 always.
lia
3B
~dt' ~
lie
d2B
~dtT
(24-80)
24-5
P L A N E W A V E IN A N A R B I T R A R Y D I R E C T I O N
389
We again consider the case in which the fields are functions only of z and t so that
the only nonzero spatial derivatives are those with respect to z. Then (24-79) reduces to
32E
dz2
d2E
dE
rH o dt
32EZ
T
dz
dt2
(24-81)
The x component of this equation is

d2Er
dz
dEr
fio
dt
d2E
dt
= 0
and there will be a similar equation for Ey. Thus the equations for Ex, Ey, and the three
components of B as obtained from (24-80) are still of the form (24-7). T h e only one
which is different is the z component of (24-81):
d2E,
dz
dEz
d2E,
d2E,
dt
dt
dz'
/ l a fie
which reduces to
d ia
dE
E,z + dt\e
dt
- 0
(24-82)
and shows that the term in parentheses is at most a function of z. If we let this function be a(z), then the general solution of the resulting equation is Ez = (e/a)
[a(z) + b(z)e~a,/t].
This component has one part that decays exponentially to zero
with the same relaxation time e / o of (12-41) and is certainly not a wave. The other part
represents a static field that can be a function of position; while this is a possibility, it
is of no interest for a study of wave propagation and, rather than carry it along, we
simply take Ez = 0. Thus, we conclude that a traveling wave in a charged l.i.h. medium
is still transverse and all of the relevant components satisfy the equation (24-7).
Consequently, all of our conclusions of the last two sections apply in this case also;
hence we need consider it no further and can continue to use (24-1) through (24-4) for
traveling waves.
24-5 PLANE WAVE IN AN ARBITRARY DIRECTION
For simplicity, we have considered only plane waves traveling in a specific direction
that we chose to call the z axis. For further use, it is helpful to generalize our results in
order that we can describe a plane wave traveling in an arbitrary direction with respect
to a given set of coordinate axes.
Previously we wrote \p = \p(z, t) so that \p was constant on all points of an infinite
plane perpendicular to the z axis as shown in Figure 24-1. W h a t we want now is for \p
to depend only on the time t and the distance f of a given plane f r o m the origin, so that
\p = \p(, t). This is illustrated by the edge-on view of the plane shown in Figure 24-6
where the orientation of the plane is described by its normal n and where r is the
position vector of an arbitrary point on the plane. We see from the figure that the
constant distance is given by
= n r
(24-83)
so that this is the equation of the plane. Thus, for a plane wave of propagation constant
k, we have
^
(24-84)
Now if we always choose the normal n to be in the direction of propagation of the
390
ZD %
tH ~
PLANE W A V E S
wave, we can take k to be positive and define a propagation vector k by

k = An = &k
(24-85)

= xp 0 e i ( k - r -" l )
(24-86)
as our representation of a plane wave traveling in the direction k (with corresponding

unit vector k = n). If, for example, we refer this to a specific set of rectangular axes, we
can write
k = kx x + kyy + k2z
(24-87)
and therefore
^
ôeH(kxx
+kyy+k:z)-
u t
(24-88)
Since \p can be a component of E or B, the generalization of (24-29) will be

E = E 0 e , ' ( k "-"' )
B = B0<?'(k -"'>
(24-89)
and now we need to see what Maxwell's equations become for this more general way of
writing our fields.
With the assumed form (24-88), the analogues of (24-30) are now
dxp
aJ "
dip
i k A
d\p
" ' ^
Tz
d\p
i k A
~si ~ ~i<c4,
(24 90)
"
which means that the del operator is equivalent to the substitution v = /k = ikk and
Maxwell's equations (24-1) through (24-4) become
k E = 0
k B = 0
k X E = (oB
k X B = (jueco + /jUa)E
^ ^
which are exactly the first three of (24-32) plus (24-58) with kz replaced by k = A:k as
we would expect. Thus, we still have the fields related by
k
B = -k XE
(24-92)
CO
where k and w are connected by the dispersion relation (24-37) so that all of the results
of the previous sections can be taken over to this general case simply by replacing z by
k and kz by k r.
Later it will be helpful to have an explicit relation between H and E, which can be
found from (24-92) since H = B/ju, and we get
H =
/X CO
\ k \
k X E = e'Qk X E
(24-93)
JU.C0
with the use of (24-60). In the important special case of a nonconducting medium, we
24-6 C O M P L E X S O L U T I O N S A N D TIME-AVERAGE ENERGY RELATIONS
can write \k\/u> = l/v
391
= -file by (24-17) and (24-12) so that we also have

k
H =
Hv
/ e \ 1/2
k X E
k X E =
\n)
Z
(24-94)
where
HT r - K f - ^ r *
is called the wave impedance. The quantity Z 0 = (jUo/ o) 1 / 2
known as the impedance of free space.
24-6
COMPLEX SOLUTIONS A N D
M oc
377 ohms is
TIME-AVERAGE ENERGY RELATIONS
As has already been amply illustrated, it is often convenient to obtain solutions in

complex form. However, we must always remember to take the real part of the solution
in order to get the physical quantity of interest as we did in (24-62), for example. Since
it is sometimes inconvenient to find the real parts, it is desirable to rewrite some of our
previous results so that the complex solutions can be substituted into them directly. We
will, however, do this only for fields that have a sinusoidal time dependence, that is, are
proportional to e~'ut. It is important to remember that while this is applicable to plane
waves, it is not restricted to them.
Let E c , B(., H c , and D c represent solutions that have been obtained in complex form.
Then the physically real electric and magnetic fields are
E = Re ( E c ) = Re (
<
H = R e ( H j = Re ( H 0 e ~ ' " ' )
(24-97)
and that we previously wrote as E r e a l and so on. The quantities E 0 and H 0 are
functions only of r. If we now write them in terms of their real and imaginary parts, so
that
^0
E/?
+ i'Ey
Hq = h
M ,
where E ^ , E / 5 H w , and H 7 are all real, then (24-96) becomes

E = Re [(E,; + / E 7 ) ( c o s cor i sin cor)]
= E^ co s cor 4- E 7 sin cor
(24-98)
while (24-97) yields

H = H zjcos cor + H 7 sin cor
(24-99)
Substituting (24-98) and (24-99) into (21-59), we find that the Poynting vector is given
by
S = E X H = (Efl x H ) cos 2 cor + ( E ; X H 7 ) sin 2 cor
+ [ ( E r X H 7 ) + (E 7 X
)] sin cor cos cor
(24-100)
which is generally a varying function of time.

In many situations, we are not primarily interested in the instantaneous value of the
energy flow, often because it fluctuates too rapidly to be followed by our measuring
instruments. The time average of the energy flow, ( S ) , is generally of much more
significance. We see from (24-100) that this will be given by
(S) = i[(E x H R ) + (E, x H,)]
(24-101)
392
PLANE W A V E S
since
(cos 2 co/) = (sin 2 <o/) = \
(sin oit cos uit) 0
(24-102)
These last results follow from (19-49) by dividing the appropriate integrals by 2 it in
order to average over a period of the trigonometric functions. Equation 24-101 is a
fundamental result, but it can be written in a more convenient form.
Let H * = complex conjugate of H c = H c with / replaced by i = H J e ' " ' =
(Hj,
Now consider
E C X H * = [(ER
=
+ LEJE-'-'X (H* -
[( E /?
/H,)*'"']
H ^ ) + (E f X H 7 ) ] + i[(Ey X H a ) (E/j X H 7 ) ]
(24-103)
Comparing (24-101) and (24-103), we see that

(S) = ^Re(Ec X H*)
(24-104)
Thus we see that we have succeeded in expressing the time average value of the
Poynting vector entirely in terms of these two complex solutions of Maxwell's equations, so that we can calculate it directly without the necessity of first finding the real
parts.
We can also carry through the same type of calculations for the average energy
densities. The results are
(u,)
- (ilE2)
<> =
/ JMH2\
JCEC E *
-
(24-105)
H* -
B*
(24-106)
Now that we have obtained these important results, we no longer need to distinguish
between real and complex fields for these purposes. Accordingly, in what follows,
although we will continually be dealing with complex solutions of Maxwell's equations,
we will not write them as E c and H c , but simply as E and H and so on. If we actually
need the real parts for specific purposes, we will point this out explicitly.
Example
Energy relations for a plane wave. We first consider the case a = 0. Substituting (24-94)
into (24-104), using (24-95), (1-30), and k E = 0 from (24-91), we obtain
( S > =
1 / e \1/2
R e
2 U )
[
(~
kXE
1 / e \1/2
*)] = 2 ( ~ )
(E"E*)k
(24-107)
since E E * is real. If we now use (24-89) and (24-94), we find that (24-107) can also be
written as
l/c\1/2
1 ( a\1/2
A
< S > _
U)
| E O | 2 K
"
2 ( 7 )
|HO1
(24-108)
since E E* = E 0 E * = | E 0 | 2 , and so on for R We see from this result that not only
is the energy flow in the direction of propagation k, as we already knew, but also that it
is proportional to the square of the amplitude of either E or H.
In the same way, we find that the energy densities are
<ue) = i e E E * = 4 | E 0 | 2 = | j u | H 0 | 2 = (um)
(24-109)
so that the average densities are equal. The total average energy density then becomes
(> = (ue)
+ ( O
= 2|E0|2 = iHHol2
(24-110)
24-7 POLARIZATION
393
which enables us to write (24-108) as

<S> = - ^ t k = (u)v k = <>v
(24-111)
with the help of (24-12). Thus, the average energy current is the product of the average
energy density and the velocity of the wave. This is in accord with the analogous result
for current density J = pv given by (12-3) and that of fluid kinematics where the mass
flow per unit area is the product of the mass density and the fluid velocity.
In a conducting medium where a = 0, we must use (24-93), and, when this is
substituted into (24-104), we obtain
<S> = ^ R e [ E x
Zfxco
(k X E * ) e " ' ' " ]
|A:) cos S2 2 W
ae~2l3t
.
- H:
e- |E|2k -
|E0|2k =
2juco
2/xco
e-WS
2/xv
|E0|2k
(24-112)
with the use of (24-22), (24-48), (24-52), (24-89), (24-85), (24-83), and (24-38), and
where f now measures the distance in the direction of propagation. Similarly, the
average total energy density becomes
a2
<"> = 2 e ~ W \ E 0 \ 2 = 2~ IE 0 1 2
2 /xco
'|
2jav
(24-113)
with the additional use of (24-49) and (24-40). We see that in this case ( S ) = ( u ) u k
just as given by (24-111) for the nonconducting medium.
Both ( S ) and ( u ) are seen to be proportional to e _ 2 / 3 ? so that they decrease with
twice the attenuation factor of the fields; this is a result of the fact that they both are
proportional to the square of the amplitudes. This energy is lost because of the resistive
heating of the material arising from its conductivity.
24-7 POLARIZATION
So far we have obtained many of our results f r o m the assumed form (24-89) without
being very specific about E 0 and B 0 other than that they are constants, that they will be
related by B 0 = (k/io jk X E 0 as found from (24-92), and that both lie in the plane
perpendicular to the direction of propagation. The character of the wave depends on
the nature of the amplitudes in this plane; we can concentrate on E 0 since B 0 can
always be found from it.
In order to simplify our discussion somewhat, let us assume that our axes have been
chosen so that the positive z axis is the direction of propagation; the transverse plane is
then the xy plane, but we will not assume any specific orientation of the x and y axes
with respect to the amplitude; for example, neither is chosen to be along the direction
of E 0 . Since E 0 lies in this plane, we can resolve it into its components and write
E 0 = E0xx
+ E0y y
(24-114)
Since E0x and E0y are both generally complex numbers, we can write them in the form
of (24-24) as
E0x = Exe^
E0y = E2e'*>
(24-115)
so that E as given by (24-29) becomes

E = ( E ^ x + E2ei^y)e'(,cz~ul)
(24-116)
394
PLANE WAVES
For simplicity, we assume k to be real; we will see that this will not affect our basic
conclusions. Then, taking the real parts of (24-116), we find the components of the
electric field to be
E = E, cos (kz cot + #,)
'
(24-117)
Ev = E2cos(kz
cot + &2)
The description of the electric field now depends on the relative values of the
amplitudes (E1, E2) and phases ( # 1 ; # 2 ) .
Since E1< Ex < Ex and E2 < Ey < E2, according to (24-117), the tip of the
electric field vector must always lie within the dashed rectangle shown in Figure 24-7;
we note that now we are using E to represent the physical electric field, and not the
complex expression of (24-116). The direction of propagation as given by the z axis is
out of the page.
Suppose we imagine ourselves to be at a definite location. Then as time goes on, we
see from (24-117) that the components of E will vary so that E itself changes and the tip
of the E vector will trace out a path of some sort within the dashed rectangle. We can
find this path or "orbit" by eliminating (kz cot) f r o m (24-117). We find that
E
= cos {kz cot) cos &1 sin {kz ut) sin

E
\
= cos {kz cot) cos # 2 sin (kz cot) sin d 2
E2
and therefore
sin
EVy
Z E sin ft
2
cos
Z Y2 cos &
EXX
EX
E~i
vy
sin (kz w?)sin(d1
d2)
Squaring each of these expressions, and then adding corresponding sides, we obtain
2 / 7 7 \ / Z 7 \
c o s ( # ! &2) + I
Figure 24-7. The components of the electric

field in the plane transverse to the direction
of propagation.
I = sin 2 (i^j # 2 )
(24-118)
Figure 24-8. An elliptically polarized

electric field.
24-7 POLARIZATION
395
This second-degree equation is generally that of an ellipse (since both Ex and Ey stay
finite) and the electric field is said to be elliptically polarized. Thus the path traced out
by its tip could be like that shown in Figure 24-8. The magnetic induction B will also be
elliptically polarized since B is always perpendicular to E, and its ellipse will be rotated
by 9 0 with respect to that for E.
T h e values of the principal axes of the ellipse and its orientation with respect to the
axes clearly depend on the amplitudes Ex and E2 and the relative phase of the two
components, since (24-118) depends only on the absolute value of the phase difference
| # i t> 2 |. It will be helpful to consider some special cases.
1.
~ #2 = 0
(Ey/E2)]2
In this case, (24-118) reduces to [(EyE^

Ex
= 0 of
Ev
(24
~ T
"U9)
This is the equation of the straight line lying along the diagonal shown in Figure 24-9.
The tip of E always lies on this line and the field is said to be linearly polarized. This
can also be seen f r o m (24-117) since, when d j = &2, the two components are in phase,
that is, they reach their maxima and minima together, and become zero together; a
relation like this leads to the straight line shown in the figure.
2. | d , - d 2 | = tt
N o w (24-118) reduces to [(Ex/E{)
+ (Ey/E2)]2
= 0 or
In this case, the field is again linearly polarized but, since Ex and Ey always have
opposite signs, the line traced out by E lies along the other diagonal as shown in Figure
24-10.
3. | d , - d 2 | = i r / 2
Here (24-118) becomes
Figure 24-9. A linearly polarized

electric field with zero phase
difference between the x and y
components.
()
_1
<24m)
'
Figure 24-10. A linearly

polarized electric field with a
phase difference of n between
the * and y components.
396
PLANE W A V E S
which is an ellipse with the major and minor axes along the coordinate axes as shown in
Figure 24-11. A special case of this arises when E1 = E2 = E0 and (24-121) becomes
the equation of the circle E* + Ey = Eq\ the field is then said to be circularly
polarized.
As we just saw, the shape of the ellipse is independent of the sign of the phase
difference
&2. However, the sense in which the p a t h is traced out does depend on
the sign, and this is what we now want to consider. It is convenient to introduce the
phase difference A explicitly by writing
$2 = A
(24-122)
and introducing the abbreviation for the phase associated with Ex

P = kz at +
(24-123)

Ex = El cos P
Ey = E2
COS (
P A)
(24-124)
It will b e helpful in analyzing these expressions to refer to Figure 24-12 in which we

have plotted a portion of the cosine terms as a function of P. T h e solid line is cos P so
that A = 0 for it. The dashed line shows cos ( P A) for A > 0 and we see that, as a
function of P, it lags the solid line. The dotted line shows c o s ( P A) for negative A,
y
Figure 24-11. An elliptically polarized electric field

with a phase difference of
between the x and
y components.
-4
-4
-1
Figure 24-12. cos (P A)

for various values of A.
24-7 POLARIZATION
397
and it is seen to lead the A = 0 curve as a function of P. If we now consider (24-124) as

a function of increasing P, then, if A > 0, E lags Ex, that is, it reaches its m a x i m u m
after Ex, becomes zero after Ex, and so on. This behavior is shown in Figure 24-13 and
since the z axis is perpendicular to this plane a n d out of the page, we see that the ellipse
is traced out in a counterclockwise sense when viewed opposite to the direction of
propagation (and hence clockwise when viewed in the direction of propagation, that is,
looking out from the page). If A is negative, then Ey leads Ex a n d the sense of rotation
is opposite to the above, as shown in Figure 24-14.
N o w , as seen from (24-123), P varies with b o t h z and /, but in different ways. As a
result, it is convenient to consider them separately.
Suppose / is a definite value, that is, we consider what one would observe as a
function of position z, much like viewing a photograph of the wave. Then as z
increases, P increases and the figures are directly applicable. Thus, if A > 0, then
looking in the direction of propagation (out of the page), the electric field will be seen
to rotate clockwise, while it will b e counterclockwise if A is negative. These are shown
for a circularly polarized wave in Figure 24-15. By using the standard right-hand rule to
define senses of rotations, we see that (a) corresponds to a positive sense of rotation,
and (b) to a negative sense of rotation; a wave like (a) corresponding to A > 0 would be
P increasing
P increasing
A > 0
A < 0
Figure 24-13. The ellipse is traced in

a counter-clockwise sense when
viewed opposite to the direction of
propagation.
Figure 24-14. The sense of rotation

is clockwise when viewed opposite
to the direction of propagation.
y
(b)
Figure 24-15. A circularly polarized wave of (a) positive helicity and
(6) negative helicity.
M
398
PLANE WAVES
described as a wave of positive helicity, while one for a negative phase difference would
have negative helicity.
On the other hand, consider an observer at a definite position (z = const.). As t
increases, P decreases according to (24-123). If we now review the way in which Figures
24-13 and 24-14 were constructed with the aid of Figure 24-12, we see that if we replace
"P increasing" by "P decreasing" in these figures, we will also have to reverse the sense
of rotation about the ellipses. Then the observer, looking into the page, and thus
opposite to the direction of propagation, will see E trace out its ellipse in a clockwise
sense for positive A and counterclockwise for A < 0. (In other words, the observer is
looking toward the source, which is the most reasonable thing to do.) The polarizations
in these cases are called right-handed (positive) and left-handed (negative), respectively.
Thus, from either point of view, the correlation has turned out to be the same; a
positive phase difference A corresponds to a positive sense of rotation, and a negative
phase difference to a negative sense of rotation.
24-8 ARE T H E E L E C T R O M A G N E T I C P A R A M E T E R S O F M A T T E R C O N S T A N T S ?
We have found that the propagation properties of plane waves can be related to the
parameters describing the medium, that is, ju, e, and a. As a specific example, we found
in (24-24) that the index of refraction of a nonconducting medium is given very simply
by n = \jKeKm. This relation works very well for many materials, but as a contrary
example, let us consider the case of water. If we look up these quantities for water in
tables, we find that xm 1 and Ke = 80, so that n = 9. But it is well known that the
index of refraction of water for light is given very closely by f 1.33. The solution to
this apparent difficulty lies in the fact that our macroscopic formulation of electromagnetic theory gives no indication of the values to be expected for Ke and n m but must
rely on experiment to obtain them. In other words, what is actually done in practice is
to take the results predicted by Maxwell's equations as always being correct, and then,
by making them agree with experiment, deducing the values to be assigned to /i, e, and
a. In this way, it is found that these parameters are not really constant for a given
material but usually have a strong dependence on frequency. (We are not concerned
here with a possible dependence on things like temperature and pressure.) In the above
example, we used the value of the dielectric constant for water for the static case,
whereas the appropriate one to use is that corresponding to the_ very high frequency of
light waves.
The atomic nature of matter is the ultimate reason for this variation with frequency;
the atomic charges that are polarized by the fields possess inertia that makes their
response to the electromagnetic forces depend on frequency. Furthermore, a combination of the effects of inertia and damping forces can introduce a phase difference
between the applied force and the response of the system and, as a result, it turns out
that it is appropriate to write the parameters as complex numbers. These effects are
considered from a detailed atomistic point of view in Appendix B. For purposes of
illustration now, however, we can restrict ourselves to a simple case which we have
already considered from a microscopic point of view, namely the conductivity of a
material.
In Section 12-5, we discussed the microscopic origin of a finite conductivity in terms
of the equation of motion (12-36) describing the net force on an electron of charge e
as the sum of the electrical force and a mechanical "frictional" force described by an
overall parameter :
F net = ma = -eE
- v
(24-125)
We found the static value of the conductivity, which we now call a 0 , to be given by
24-8 ARE THE ELECTROMAGNETIC PARAMETERS OF MATTER CONSTANTS?
399
(12-39) as
ne2
ct0 =
(24-126)
where n is the number of electrons per unit volume. We now want to apply this same
picture to the case in which the material is subject to the time varying electric field of a
plane wave.
Substituting (24-89) into (24-125), we obtain
d\
ma = m
= - e E 0 e ' ( k r - " " ) - v
(24-127)
The equivalent of a steady-state solution here is one in which the response at a given
location has the same time variation as the forcing function; accordingly we try to solve
this equation by assuming a solution of the form v = \0e~ loU where v0 is independent of
time. When this is substituted into (24-127), we obtain m(dv/dt)
= imcov = eE
v so that
v =
-eE
(24-128)
imco
The free current density can be found in the same way as was used in (12-38):
J1/ = 1 P/V = n(-e)v
ne2 E
(24-129)
imoi
Since, by definition of the conductivity, this would be written as 3f = a E, we see that

the conductivity is
a = o(cj) =
ne2
imu
(ne2/)
1 i(mco/)
on
7
1 iyo0mu/ne
ry
)
(24-130)
with the use of (24-126). First of all, we see that the conductivity has turned out to be a
function of the frequency and the frequency dependent term involves both the inertia
of the charge carriers ( m ) and the resistive forces (); we also note that a reduces
properly to a 0 for the static case co 0. Second, we see that a is complex, and by now
we recognize that this means that the current density and the electric field are no longer
in phase. [In fact, by analogy with mechanics, we recognize that (24-125) is simply the
equation of motion of a damped " h a r m o n i c oscillator" with no restoring force, as is
appropriate for unbound electrons, and we recall that then the displacement and
velocity are no longer generally in phase with the applied oscillating force.]
We can find the real and imaginary parts of the conductivity by multiplying the
numerator and denominator of (24-130) by the complex conjugate of the denominator
and we get
or-mco
On
a = aR + ioj
1 + /(
(24-131)
ne
1 + (o0mco/ne2y
In order to investigate the physical significance of the complex conductivity a little
more, let us assume that we are at a fixed location r so that E in (24-89) and hence if
can be written in the form
E = E'0e~io>t
Jf=
(aR + ioj) E'0e~iu'
(24-132)
For simplicity, let us assume that E'0 is real; then the physical quantities of interest are
400
PLANE WAVES
given by the real parts of (24-132) and, with the aid of (24-22), they are found to be
E r e a i = E' 0 cos
J
/ real
c o s ut
of s i n Oit ) E q
(24-133)
Thus the real part of a complex conductivity gives rise to a component of the current
that is in phase with the applied electric field and it is the imaginary component a f that
describes a component of the current that is completely out of phase with the field. [It is
easily seen from (24-131) that both aR and oL are positive quantities.]
N o w we can investigate the effect of a complex conductivity on the propagation
characteristics of the medium. The dispersion relation (24-37) did not depend on the
quantities being real, nor did the use of (24-38). However, when (24-39) is combined
with (24-131), we find that now we have a2 fi2 + 2 i a f i = co2ju,e + icon(oR + iaf), so
that, instead of (24-40) and (24-41), a and ft are related by
a2 fi2 = co2fxe ufiar
2 afi = cojaaR
(24-134)
We could go ahead and solve these for a and /? as before, but there is an easier way.
We can write the first expression as a 2 ft2 = co2/x[e (ct7/co)] and upon comparing
with (24-40) and (24-41), we see that (24-42) and (24-43) can still be used by simply
making the replacements o * o R and c -> c (ct 7 /cj).
This connection between the permittivity and the conductivity is sometimes emphasized by writing the dispersion relation (24-37) in the form
k2 =
(24-135)
so that, as far as wave propagation is concerned, the existence of a conductivity can be

handled simply by adding the (generally complex) term io/io to e.
Rather than continuing with these general results, let us consider two extreme
limiting cases that can be characterized by the relative size of the parameter , which is
a measure of the "friction" in the system, that is, of the overall effect of collisions.
1. "Large Friction" (mto / ^
1 or u : ne2 / o0m)
We see from (24-130) that this means that a = <t0 = const, so that the conductivity is
essentially always real and equal to the static value. This is found to hold very well for
metals. Using the numbers quoted for copper in Exercise 12-17, we find that n = 8.5 X
10 28 (meter)~ 3 . Using a0 6 X 10 7 ( o h m - m e t e r ) - 1 and 1.60 X 1 0 " 1 9 coulomb and
9.11 X 1 0 " 3 1 kilogram for the electron's charge and mass, we find that this requirement
on the frequency v is v <c 6 X 10 12 hertz. This condition is satisfied until we are well
into the microwave region, so that metals can justifiably be treated as having a constant
and real conductivity as we implied in Section 24-3.
2. "Small Friction" (moo / s> 1 or to ne2 / a0m)

Here we can neglect the 1 in the denominator of (24-130), which yields
and the conductivity is a pure imaginary, a R = 0, and the current density and the
electric field are exactly 90 out of phase as shown by (24-133). Such a situation as this
arises in an ionized gas with low particle density or plasma as it is known. Because of the
EXERCISES
401
tenuous nature of the medium, the effect of collisions is very small so that is very
small. Putting (24-136) directly into the dispersion relation (24-37), we find that
k2 = v ( l - - ^ - j ) = < o V ( l \
mcco J
\
co j
(24-137)
where we have written

co
ne2
=
(24-138)
me
and where vp = COp/2TT is called the plasma frequency. Actually, for such a medium,
jli ju0 a n d e = e 0 and we can approximate these expressions very well by
co2 \
-
co j
co =
ne2
me
(24-139)
Because of the minus sign in the expression for k2, there are two separate cases to be
considered.
If co > coP, then k2 > 0 so that k is real. Thus we have an u n d a m p e d traveling wave
of the f o r m (24-84) with a phase velocity found f r o m (24-139) to be
(24
"140)
which is greater than the value c, which would apply to a nondispersive medium with
otherwise vacuum properties.
If co < coP, then k2 < 0 so that A: is a pure imaginary. In this case, we can write k as
f o u n d f r o m (24-139) in the form
k = ifip = / ( c o / c ) [(cop/co) 2 l ] 1 / 2
(24-141)
and then a solution of the form (24-84) becomes

xf = x p o e - ^ e - i u '
(24-142)
This is not a traveling wave, b u t a sinusoidally oscillating field component whose

amplitude is exponentially decreasing with distance. If the plasma is large enough, then
\p - 0, and the field will not penetrate the medium. Hence one can speak of a plasma
as being an example of a "high-pass" filter in the sense that the frequency must be
greater than the plasma frequency in order to get propagation rather t h a n simply
attenuation.
Although one would not describe the conduction electrons in the body of a neutral
metal as a plasma of low particle density, they otherwise fit the requirements of this
case for a high enough frequency to outweigh the effect of collisions, and it is of interest
to evaluate the order of magnitude of vp for a metal. Using the same numbers that we
have already used for copper, we find from (24-139) that vP = 2.6 X 10 16 hertz. W h a t
this extremely high value of vP tells us is that we cannot expect a typical metal to
become " t r a n s p a r e n t " to electromagnetic waves until we are well into the ultraviolet
region of the spectrum, a result that is completely consistent with our other conclusions
about the behavior of metals. Other examples of plasmas will be considered in the
exercises.
EXERCISES
24-1 The form (24-11) was verified to be the

solution of (24-9) by direct substitution. Similarly,
separation of variables in z and t was used to get

the explicit form (24-18). Another approach to
402
PLANE W A V E S
(24-9) is the following. Define a new set of variables by = z + vt and 17 = z vt and express
(24-9) in terms of and rj. If v is again given by
(24-12), show that the resulting equation has a
solution of the form \p = /(tj) + g().
24-2 Phenomena that arise as a result of the
superposition principle, that is, a sum of solutions
is also a solution, are often characterized by the
term interference. As an example, consider two
plane waves \pl and \p2 that are each of the form
(24-19) with the same values of k and w, that is,
they are traveling in the same direction with the
same velocity. Suppose, however, that they have
different amplitudes and phases, so that xp(n =
xpê'^1 and xp02 = &2ae'^2- Find their sum xp =
+ \j/2. Find the real parts of xpt, xp2, and xp. In
the special case in which they have equal amplitudes (xplcl = <p2a = â)' show that Re if =
2\pa cos ! ( $ ! # 2 ) c o s [ ^ z ut + 2(^1 + ^2)]
and interpret the result. If |#x #2I = w> what is
Re xp? Explain how this happened by considering
the relation between the individual waves in the
superposition.
2 4 - 3 / A s another example of interference, suppose that xp 2 f the previous exercise is now
propagating in the negative z direction. Show
that if the waves have equal real positive amplitudes that Re xp = 2xpa cos kz cos u>t. This is an
example of a standing wave. Plot Re xp as a function of z for a convenient value of t, and as a
function of t for a convenient value of z.
24-4 As was noted after (24-53), a superposition
of plane waves traveling in a dispersive medium
generally changes its form as it progresses. As an
extreme example, consider two waves of equal
real amplitudes and with almost the same propagation constants and frequencies that are traveling in the positive z direction, so that their sum
has the form
^
^êi{(k + Uk)z-(u + du)t\
+
A- t/Ar
)/]
Show that
Re xp = 2xp0 cos [( dk) z ( du) / ] cos (kz ut)
What is the spatial period (wavelength) X of the

main wave? What is the wavelength Xm of \pm?
Find the ratio A,/A. Sketch (24-143) at t = 0,
and at a slightly later time, and thus verify that
the form of Re ip has changed. Identify which
physical feature of your sketch travels with the
phase velocity v = u/k and which with the group
velocity vG. Why do you think the name "group"
velocity was given to (24-144)? (This specific result is an example of the more general phenomenon known as beats.)
24-5 A plane electromagnetic wave traveling in
a vacuum is given by E = yE0e'(kzwhere
E0
is real. A circular loop of radius a, N turns, and
resistance R is located with its center at the
origin. The loop is oriented so that a diameter lies
along the z axis and the plane of the loop makes
an angle 6 with the y axis. Find the emf induced
in the loop as a function of time. Assume that
a <=c X. (Why?)
24-6 A particle of charge q and mass m is
traveling with velocity u in the field of a plane
electromagnetic wave in free space that itself is
traveling in the z direction. Find the force on the
particle. What does this become for the special
case in which the particle is traveling in the same
direction as the wave? What is the direction of the
force? Under what conditions (if any) will the
force vanish in this case?
24-7 A representative value for the conductivity
of sea water is a 4 (ohm-meter) -1 . Take t = c 0
and 11 = n0 and find Q, v, 8, the magnitude of
the ratio E/cB, and 12 for each of these frequencies v in hertz: 10 2 , 107, 1010, and 1015. (They
correspond roughly to " power," " radio," " microwave," and "light.") The result obtained in this
way for 8 for light is obviously ridiculous. Why?
What is the probable reason for this erroneous
value having been obtained?
24-8 Find the index of refraction n of a conductor as a function of Q. Show that, in the limit of a
good conductor, your result reduces to n2 =
KeK,/2Q.
24-9 The solar constant, which is the average

(24-143) rate of incidence of radiation at the earth's atmo2
where \pm is called the "modulation." Thus sphere from the sun, is 1340 watts/(meter) . As(24-143) has the form of a wave with average suming that the radiation is a linearly polarized
propagation constant k and average frequency w. plane wave, find the amplitudes of E and B.
Its amplitude is not constant, but itself is a wave 24-10 Verify (24-105) and (24-106).
and travels with the group velocity vG
24-11 Find the ratio ( u m ) / ( u e ) for a plane
dui
wave in a conducting medium. Then find the
vG =
(24-144)
approximate expressions for this ratio for the
= xpm cos (kz tot)
EXERCISES
limiting cases of an insulator and a good conductor.

24-12 A plane wave travels in the positive z
direction in a conductor with real conductivity,
(a) Find the instantaneous and time average power
loss per unit volume due to resistive heating for
any z. (b) Find the total power loss per unit area
between z = 0 and z - oo. (c) Find the time
average Poynting vector at any z. (d) Compare
the value of your result for (b) with the magnitude of your result in (c) evaluated at z = 0. Is
your answer reasonable? Explain.
24-13 Consider a plane wave that is a superposition of two independent orthogonal plane
waves and has the form E = xEae'(k*~*"+^
+
yEpe' (k! ~ wt+9 i> ) where k is real. Find (S) and
show that it equals the sum of the average Poynting vectors for the components.
24-14 Consider a plane wave that is a superposition of two independent plane waves with
parallel electric fields and has the form E =
+
+ kEfiei(kz-u,
W where k is
real. Find (S) and show that it is not equal to the
sum of the average Poynting vectors for each
component. (This is a result of "interference.")
24-15 Consider a superposition of plane waves
all traveling in the same direction in a nonconducting medium. Thus, the real field will
be a sum of terms like (24-35), that is, E =
E0itcos(A:z ukt + &k). Show that the timeaverage electric energy density is the sum of the
average energy densities associated with each
component.
24-16 In Exercise 22-6, we found that in a l.i.h.
nonconducting region where there are no free
charges or free currents, E and B can be found
from a vector potential A alone so that we can
take <f> = 0. (a) What are the two equations that A
must satisfy? (Actually, there are four scalar
equations.) (b) Find the solution of these equations for which A is a plane wave traveling in the
k direction, (c) Find E and B from this A, and
verify that they are the same as those found in the
text, that is, they satisfy the set of equations
(24-91) with or = 0.
24-17 A certain plane wave has a propagation
vector k = 314x + 314y + 444z (meter) -1 . Assume it is traveling in a vacuum and find the
wavelength, frequency, and angles made with the
xyz axes.
24-18 Given the plane wave E = y[(3 4- 2i')x +
(3 -I- Ai)y]e'(k: ^n where y is a positive constant
403
and k is positive and real. Find: Elt E2, d l 5 #2, A,

and the angle 9 between the major axis of the
ellipse and the x axis. Does the wave have positive or negative helicity? Is its sense of polarization right-handed or left-handed?
24-19 Show that the electric field of a righthanded circularly polarized wave can be written
in the form E + = E0(x iy)ei(kz~"'+9).
What is
the corresponding expression if it is left-handed?
24-20 Show that a linearly polarized wave can
be written as a superposition of right- and lefthanded circularly polarized waves of equal amplitudes. Show that ( S ) for the linearly polarized
wave equals the sum of the average Poynting
vectors of each of its circularly polarized components.
24-21 Show that for a plasma for which e = t 0
the plasma frequency in hertz can be found from
vP = 8.97n 1/2 . As examples of other important
plasmas, find vP for (a) a typical gas discharge for
which n 1018 (meter) -3 ; and (b) the ionosphere where n 1010 (meter) -3 .
24-22 In an ionized gas, there are movable ions
present as well as electrons. Why were we able to
neglect their contribution to the conductivity and
hence to the plasma frequency?
24-23 If a material has a conductivity that is a
pure imaginary, such as is the case in (24-136),
show that the average rate of energy dissipation
per unit volume, (w), is zero. Is this result compatible with the assumption that collisions have
been effectively neglected?
24-24 For a system with a conductivity given by
(24-136), find the total current density J, = ctE +
(dD/dt) and show that it is less than it would be
in the absence of conduction electrons. Explain
how this occurs.
24-25 The plasma frequency was defined in a
purely formal manner in (24-138), but it can be
understood as a natural frequency of the system.
This same expression can also be derived in these
two ways, (a) Show that when an oscillatory solution for pj proportional to e~IOp' is combined
with V + ( dpj/dt) = 0, if = ffE, and V E
= Py/c, the result is again (24-138). (b) Suppose
that the equal number of positive ions and electrons are displaced relative to each other by a
small distance x. Find the force on one of the
resulting surface charge densities that is produced
by the other, and show that the corresponding
equation of motion is that of a simple harmonic
oscillator of frequency u P given by (24-138).
404
PLANE WAVES
24-26 (a) Find the group velocity % defined in

(24-144) for a system with the dispersion relation
of (24-137). Plot vG as a function of u for > uP.
Show that the product of the phase and group
velocities is vvG = (fie)'1, (b) Show that if the
dispersion relation has the form k 2 = /(w 2 ), then
the product vvG will be constant provided that /
has the form f = Aca2 + B where A and B are
constants.
relaxation time of a conducting medium as found

in (12-41) and the temporal period T of the
oscillation, (b) Express 8/\ as a function of Q
only and show that when Q 1, this reduces to
Q = TT8/X. (c) Use this latter result to justify the
following convenient way of remembering the
physical meaning of Q: "The value of Q equals
the number of cycles that a (nearly) free oscillation will persist."
24-30 A study of Figure 24-12 might lead one to

wonder if our conclusions about leading or lagging might be changed if A turns out to have a
value outside the range shown. Show that, for the
discussion of elliptical polarization, one can restrict
A to the range
to -n so that, if A as
<
V.
found from (24-122) falls outside, one can add or
24-28 Consider the most general possibility in subtract appropriate multiples of 2 IT to bring it
which a medium has all complex parameters, that within this range. Illustrate this with a simple
is, fi = fiR + i f i j , e = cR + ic,, and a = aR + iaf.
sketch of a convenient example.
Show that k can still be written in the form
24-31 Consider two separate electromagnetic
(24-38) with a and ft real. Find a and ft. Now
fields described by the pairs (Ej, Hj) and
consider a strictly nonconducting material, and
( E 2 , H 2 ) . Assume that the medium is l.i.h. in all
show that there can still be attenuation (ft = 0)
properties, that J'( = 0, and that these two fields
and that there will still be a phase difference
are sinusoidal functions of time with the same
between E and B (ft/a
0).
frequency, that is, are proportional to
Show
24-29 The symbol Q was originally introduced that it follows from Maxwell's equations that
as an abbreviation for "quality factor." Although V (Ei X H 2 - E 2 X Hj) = 0. This result is
we found a useful interpretation of Q in (24-66), known as Lorentz' lemma and shows that these
there are two other ways of writing it that are otherwise independent electromagnetic fields are
helpful, (a) Show that Q measures the ratio of the related in this sense.
24-27 Assume that a system is described by
(24-137) and that w > u P . Find ( u e ) , (u m ),
X u m)/i u e)> ( " ) ' < s >. and show that (S) =
(w)
where vL, can be interpreted as the speed
of energy flow. Show that vc < vu < 0 0 1/2
25
REFLECTION A N D REFRACTION
O F PLANE WAVES
In the last chapter, we considered plane waves of infinite extent, so that they could be
thought of as traveling in an infinite medium. Eventually in any real situation, the wave
will encounter another medium with different electromagnetic properties; for example,
a light wave in air may impinge on a piece of glass. The incident wave will generally be
able to enter the second medium to some extent and we want to find out what the
overall situation will be when the resultant steady state has been attained. We will find
that we can solve this problem by using the boundary conditions that the field vectors
have to satisfy at a surface of discontinuity in properties, and we recall that these
boundary conditions were obtained directly f r o m Maxwell's equations.
25-11 THE LAWS OF REFLECTION AND REFRACTION

We assume that the boundary between the two media is an infinite plane surface. We
further assume that there are no free charges or free surface currents on the boundary
surface. The boundary conditions given by (21-25) through (21-28) then reduce simply
to (1) the normal components of D and B are continuous, and (2) the tangential
components of E and H are continuous. It will be seen soon that we need only the
second set for our purposes; it will be left as an exercise to show that no additional
information would be obtained by using the first set involving the normal components.
The most general situation we can visualize would therefore correspond to an
incident wave traveling in a medium 1 of electromagnetic parameters
e l5 o : )
incident on a medium 2 described by (ju 2 , c 2 , o 2 ). Although the general solution can be
easily enough obtained when both media are conducting, the complexity of the results
tends to obscure the understanding and interpretation of what is happening. Consequently, we assume for now that both media are nonconductors so that all propagation
vectors are real; we briefly consider the effect of conductivity in Section 25-6. We can
take advantage of the experience of others and begin at once with the knowledge that
the boundary conditions can be satisfied only by assuming the existence of three waves:
the incident wave in medium 1, a reflected wave also in 1, and a transmitted wave in
medium 2; we use the subscripts i, r, and t, respectively, to label quantities referring to
these waves.
Using (24-89), we can write the electric fields of these waves as
E, = E 0 ,e' ( k E, = E 0 ,e' (k <
Er = E0re'(k'
""rt)
(25-1)
where
(25-2)
405
406
REFLECTION A N D REFRACTION O F PLANE W A V E S
and where we have introduced the phase velocity a n d index of refraction of the
corresponding medium with the use of (24-17) and (24-13); we note that we have m a d e
no special assumptions about the frequencies. Similar equations can be written down
for the magnetic fields.
In (25-1), the origin of time (t = 0) is arbitrary; r represents the position vector of a
given point with respect to an arbitrary origin. F o r simplicity, however, we choose the
origin to lie in the surface of separation so that the position vector xB of a point in the
bounding surface also lies in the surface. These quantities are all illustrated in Figure
25-1 where we also show the normal n to the b o u n d a r y drawn f r o m medium 1 to
medium 2 according to our standard convention.
The total electric field at a given point in medium 1 will b e E x = E, + E r , while that
in 2 is E 2 = E ( . N o w at any point on the boundary where r = xB, we must have
1 tang
2 tang
[EofC''0"-"-"'0 + E o / i ( k ' r ^ u ' ] t a n g = [E0,e'(k<r*
']tang
(25-3)
and this is to be true for all values of t and all possible rB where t and rB can b e
independently varied. Regardless of the relative values of the amplitudes E 0 , it is clear
that we cannot satisfy (25-3) unless the values of the exponents are all equal. Thus, we
see first of all that to, = ur = cof = co so that all three waves must have the same
frequency to; then (25-2) simplifies to
(25-4)
Similarly, we must have k, r B = k r rB = k , rB; it is helpful to subtract these in
pairs to get the two independent equations
(k,. - k r ) rB = 0
(k, - k , ) rB = 0
(25-5)
which shows that these differences in the propagation vectors must be perpendicular to
r B , and hence to the bounding surface; this is also true for the pair k r and k , since we
also have ( k r k , ) rB = 0. The first equation of (25-5) relates the incident and
reflected propagation vectors so we can call it the law of reflection; similarly, the second
is the law of refraction since it relates the incident a n d transmitted (refracted) waves.
Figure 25-1. Relations at a plane

boundary between two media.
25-1 THE LAWS O F REFLECTION A N D REFRACTION
407
T h e plane defined by n and the incident vector k ; is called the plane of incidence and
is shown in Figure 25-2. The angle 8t m a d e by k ( with the normal is the angle of
incidence. W e can resolve k, into its components normal and parallel to the surface of
discontinuity by defining a unit tangential vector T, which is parallel to the surface and
in the plane of incidence; thus we can write
k,. = kin n + kiTt
(25-6)
Other vectors that do not necessarily lie in the plane of incidence can b e written in
c o m p o n e n t form by introducing a third unit vector n X t , which is also in the surface
between 1 and 2 and is perpendicular to b o t h n and t as shown in Figure 25-3. Thus,
we can write the other two propagation vectors in the form
kr = krnn
+ krJr + krcn X t
k , = ktnh
+ kt7t + ktcn
X t
(25-7)
while
rB = rBS +
(25-8)
since it lies entirely in the surface and thus has n o normal component.
Substituting (25-6) through (25-8) into (25-5), a n d remembering that the unit vectors
are mutually perpendicular so that n T = 0, a n d so on, we obtain
(*.> - k)rBr
- krcrBc
= 0
(kiT - ktT)rBT
- ktcrBc
= 0
(25-9)
Since rB is arbitrary, (25-9) must also be true for the special case for which rBl. = 0
while rBc = 0, and then (25-9) yields krc = k[c 0 so that k r and k , have n o
c o m p o n e n t s perpendicular to the plane of Figure 25-2. In other words, all three
propagation vectors lie in the same plane, that is, they are all in the plane of incidence.
N o w , since rBr can be different f r o m zero, we see f r o m (25-9) that
krT = kiT
ktT = kiT
(25-10)
so that all of the k's have the same tangential components. We can regard (25-10) as a
second version of the laws of reflection and refraction. N o w we can learn something
a b o u t the normal components.
1
k
Figure 25-2.
Figure 25-3.
This is the plane of incidence.
The three mutually perpendicular unit vectors.
408
Remembering that krc = 0, we find f r o m (25-7), (25-4), (25-10), and (25-6) that
k;n = kf - kfT = kf - kfT = kf - kfT = k f n and therefore krn = kin.
If the plus
sign were chosen, then we see f r o m Figure 25-2 that the reflected wave would be going
toward the surface in the same way as the incident wave, whereas it should be tending
away from it as in Figure 25-1. Thus we must have krn = kin. Similarly we find that
k f n = kf kfT = k\ kfT. Therefore, the third versions of the laws of reflection and
refraction are
krn -
kl
= k\ - kl
(25-11)
N o w k, is a vector of magnitude k x by (25-4) and we see f r o m Figure 25-2 that

kjn = kx cos 6t
kiT = k1 sin 0,
(25-12)
and therefore krn = kl cos 6,, krT = k1 sin 6, by (25-11) and (25-10). Since k r also
has the magnitude kv we can define an angle of reflection 6r by
krn = kxcos6r
krT = kl sin0 r
(25-13)
as shown in Figure 25-4a. By comparing these different expressions for the components
of k r , we see that
=
(25-14)
so that the angle of reflection equals the angle of incidence. Thus, this ancient and
well-known law of optics is seen to be a direct consequence of Maxwell's equations.
For the transmitted wave, we find f r o m (25-10), (25-11), (25-12), and (25-4) that
ktT = k1 sin 0,
(25-15)
2
kf
k\ k\ sin 6, = k\
1 1 1 sin 2 6,
n-
(25-16)
These results are correct for any values of the ratio n1/n2 and of the angle of incidence.
U n d e r m a n y circumstances, it is useful to continue by defining an angle of refraction
0t for k, of magnitude k 2 by means of
klT = k2 sin 0t
ktn = k2 cos 6t
(25-17)
as is also illustrated in Figure 25-4Z>. Combining (25-17) with the law of refraction as
given in (25-15) and (25-16), along with (25-4), we find that they lead to
n x sin &i = n 2 sin 0t
cos 2 d, = 1 ( ) sin 2 0,
n2
Figure 25-4.
refraction.
(a) 6r is the angle of reflection, (b) 0, is the angle of
(25-18)
(25-19)
25-1 THE LAWS O F REFLECTION A N D REFRACTION
409
which are internally consistent since it is seen that cos 2 0t = 1 s i n 2 0 r The first
equation (25-18) is known as Snell's law of refraction and was discovered experimentally by him; it is seen to relate the angles of incidence and refraction to the indices of
refraction of the two media, and is also seen to be a consequence of Maxwell's
equations via the b o u n d a r y conditions obtained from them.
Although (25-17) defines two quantities sin 0, and cos 0t, they do not always
correspond to a real physical angle 6 t . Let us see how this comes about and what we
should d o about it. W e find from (25-18) that sin#, = (nl/n2)
sin 0t. If nx < n2, then
sin 6t < sin 0, < 1 and cos 2 0, = 1 sin 2 6, > 0; thus 0t is a real angle and 0t < 0r This
case is illustrated in Figure 25-5; note that k2/ki
= n 2/nl > 1. If nx> n2, then we
still have sin0 r = (n1/n2) sin 0t > sin 9i so that 0t > 6t in general while k2/kx
< 1.
However, we should also have sin 0, < 1 so that 0, will be a real angle only as long as
sin 8I < n 2/nl < 1. It is useful to define the critical angle 0C by
sin0=
"i
(25-20)
and we can write

sin0 t =
sin 0,
sin 0C
(! > n2)
(25-21)
Therefore, for 6i < 6C, sin 6t < 1 and 0, > 8,. This case is illustrated in Figure 25-6 (We
note that the larger of the two angles 0, a n d 0, is always in the medium of smaller index
of refraction.) If 8i = 0C, then sin 8t = 1 a n d 0t = 90, and the refracted wave is parallel
to the interface as shown in Figure 25-7; we still have k2 < kx.
Since 6l can be freely chosen by us, it is certainly possible to make 6t > 8C. Then
(25-21) shows that sin 8t> 1 while cos 2 0t = 1 sin 2 0t < 0, which is not possible for a
real angle and no figure like any of the last three can be drawn. (The quantity 0, can
still b e interpreted as an angle as long as it is complex and of the f o r m 0t = jv + ix. To
see how this comes about, we use Equation 24-22 to evaluate the sine as sin 0t = (e'6<
e~'e,)/2i
= [ e ' ( w / 2 ) ~ * e~'<-"/T> + x\/2i = \(e~~x + ex) = cosh x > 1 as required.)
Figure 25-5. The three propagation vectors as related by the laws of

reflection and refraction when n, < n2.
410
, > 2 ; 0(
<
, > 2 ; 0, 9C
6C
Figure 25-6. The three propagation

vectors as related by the laws of
reflection and refraction when n, >
n2 and the angle of incidence is less
than the critical angle.
0t = 90"
Figure 25-7. The three propagation

vectors when the angle of incidence
equals the critical angle.
Thus in this case it is not fruitful to continue with 6 t ; we can, however, go back a n d
evaluate the components of k , in terms of 6 i by means of (25-15) and (25-16) since
these are always valid. Consequently, there is no difficulty in handling this case in
principle. It turns out to be convenient to discuss these two b r o a d classes separately,
and we put off the detailed consideration of the case ( n l > n2; 6i > 6C) until Section
25-4.
All of our results so far have been obtained solely f r o m the exponential propagation
terms in the single boundary condition (25-3), and we still have to find the necessary
relations among the fields themselves. N o w E, must lie in a plane perpendicular to k,,
but it need have n o special orientation with respect to the plane of incidence of Figure
25-2. T h e general situation for an arbitrary orientation of E, for an arbitrary angle of
incidence is shown in Figure 25-8. The shaded plane is perpendicular to k, and
therefore contains E,; this plane is also perpendicular to the plane of incidence defined
by k, a n d n. We see that E ( can be written as the vector sum
E/ = E , + E ,
(25-22)
where E, is the component perpendicular to the plane of incidence and E ^ is the

c o m p o n e n t parallel to (in) the plane of incidence. It turns out that these two components behave differently at the plane of separation between the two media, so that it is
desirable to consider the two cases separately. After we have f o u n d how each of these is
affected, we can recombine the results to find the total reflected and transmitted fields,
that is, we will be able to write
E r = E rJ _ + E r ( (
E , E,j_ -HE,,,
(25-23)
2 5 - 2 E PERPENDICULAR T O THE PLANE O F I N C I D E N C E
411
Figure 25-8. Definition of the components of E(- parallel to and

perpendicular to the plane of incidence.
25-2 E PERPENDICULAR T O THE PLANE OF INCIDENCE
Since we have no a priori knowledge about the relations among the various electric
fields, let us assume for definiteness that they are all directed out of the page at this
particular instant and location on the bounding surface; if our assumption about the
relative directions is incorrect, the final result will show us by means of its sign. The
corresponding directions of H can be found f r o m the condition that E X H is in the
direction of propagation k. In this way, we get the situation shown in Figure 25-9; for
clarity, the E's and H ' s are drawn at some distance f r o m the interface between the two
media, but we remember that they actually are to represent the situation at the surface
where we apply the boundary conditions.
Since the electric fields are all parallel to the surface and to each other, they are all
tangential components so that E l t a n g = E 2 t a i l g becomes simply
Et + Er = E,
(25-24)
All of the magnetic fields are in the plane of incidence for which T is the unit tangential
412
REFLECTION A N D REFRACTION OF PLANE W A V E S
Figure 25-9. E, is perpendicular

to the plane of incidence.
vector and the other boundary condition H l t a n g = H 2 t a n g becomes
(H,. + H r ) t = H , T
(25-25)
This last result can be expressed in terms of the electric fields by using
k
H =
and
T=
E,
E,
(25-26)
as obtained from (24-94) and an inspection of Figure 25-9, while recalling that E, is out
of the page. Therefore, the tangential component of any of the magnetic fields has the
form
H
t = ^ ( k
X E) ( E ; X n) = - | ( k n)
(25-27)
with the use of (1-106), k E, = 0 (they are perpendicular), and E E, = , (they are
parallel). The appropriate value of the impedance is Zx = ( f i i / t i ) l / 2 for the incident
and reflected waves in medium 1, while it is Z 2 = ( H 2 / e 2 ) 1 / 2 f r H , . Taking this into
account, and substituting (25-27) into (25-25), we get
z, i
[,(k,. n) + Ekr
A)] = y - ( k , n)
z,2
(25-28)
Since Et is arbitrary, the significant quantities are the two ratios Er/E[ and Et/En
and they can now be found from the two equations (25-24) and (25-28). The results are
( E\
''
,L
Z 2V
, ( k , n) - Z l,V( k . n)
'
'
'
'
Z 2 ( k : n) + Z ^ k , ii)
E,\
E,
2Z2("k,n)
Z 2 ("k, A) + Z , ( k , ft)
/V
(25-29)
(25-30)
where use has been made of k r n = k, n as f o u n d f r o m (25-4) and the law of

reflection given in (25-11). These ratios are the general solution for this case and are
valid for any values of n1/n2 and 0,. In order to emphasize this, we can express them
completely in terms of 0, by using k, n = kin/kx
= cos 0, from (25-12) and k , n =
2 5 - 2 E PERPENDICULAR T O THE PLANE O F INCIDENCE
ktn/k2
413
as found from (25-16) and we obtain

/ Er\
Z 2 cos 0, Z j [ l ( i / 2 ) 2 sin 2 0 , ] 1 / 2
\ Ei I
Z2 cos 0, +
[ l (nl/n2)2
I Et \
\
sin 2 0 , ] 1 / 2
2Z2cos0,
Z 2 cos 0 , + Z j [ l ( ! / 2 ) 2 s i n 2 fy]1/2
(25-31)
(25-32)
In the special case of normal incidence (0, = 0), these reduce to

I Er\
Z2 Z1
/ j.
I Et\
2Z2
\ Ei j
+ Zx
Z2
(25-33)
Zl
It is also very common to find these results written in terms of both the angle of
incidence 0, and the angle of refraction 6r Since 0, is real only for nl<n2ornl>n2
and 0, < 0C, the following expressions are restricted to these cases. Then (25-29) and
(25-30) become
IE
y Et:
Z 2 cos 0, - Z , cos 0,
i \
Z2 cos 0, + Zl cos 0,
y Et
==
2 Z 2 cos 0,
Z2 cos 0, + Z x cos 0,
These are often put in other forms with the use of

f u , e 7 \4 1 / 2
Z\
1 P1
Z2
V /x2e! /
Mi"?
^ w
M2 i
Mi s i n 0
'
fi2sin0t
(25-35)
which follows from (24-95), (24-12), (24-13), and (25-18). Dividing the numerator and
denominator of each expression in (25-34) by Z 2 and using (25-35), we find that the
ratios can also be written as
/M
\ Et J
fi 2 tan 0, /i 1 tan 0,
lEA
n2 tan 0, +
\ Et
tan 0,
2/i2tan0,
(25 36)
ju2 tan 0, + /i1 tan 0,
The results for normal incidence as found f r o m (25-33) are

fi2n1
- filfj2
/,\
2nini
M 2 "l + Ml"2
\ il
(25-37)
M2"l+Ml2
These are seen to be compatible with (25-36), since, as 0, and 0, both approach zero,
Snell's law (25-18) becomes approximately
n1
sin 0,
0,
tan 0,
sin 0,
0,
tan 0,
(25-38)
In the very important special case n1 = jx2, which includes nonmagnetic media, these
can also be written in general as
\ El j
sin(0, 0,)
I Et\
sin (0, -I- 0,)
\ E( J
(ni/n2)
sin 20,
sin (0,. + 0,)
(25-39)
while, for normal incidence, they reduce to

(Er\
E
\ ij
(n2
i\
\ 2 + "l /
I E,\
\Eilx
The equations in (25-39) are usually called FresneVs
2 + " l
(25-40)
equations', he derived them from
414
REFLECTION A N D REFRACTION O F PLANE WAVES
an elastic solid theory of light long before the development of Maxwell's equations and
the electromagnetic theory of light. (It is perhaps not too surprising that he was able to
do this, since, if an elastic wave in one medium is incident u p o n another medium of
different properties, there will be b o u n d a r y conditions that must also be satisfied. F o r
example, if a wave is traveling on a stretched string, one would have to require that the
displacement b e continuous at the junction for otherwise the strings would separate.)
T h e general form of the results obtained f r o m Fresnel's equations as a function of 0,
are shown in Figures 25-10 and 25-11 for the case nx < n2. W e note that the ratio
(Er/E>> is negative for all values of 0,; this can also be seen f r o m (25-39), since when
nx < n2, 0, > 0,, so that sin(0, 0,) > 0. This means that E r is actually opposite to the
direction we assumed; in other words, the reflected wave has its phase changed by 180
u p o n reflection on a medium of greater index of refraction. On the other h a n d ,
(Et/Et) is always positive, so that the transmitted wave undergoes no change in
phase. (Thus, to represent the actual situation, Figure 25-9 should really have the
direction of H r reversed, while E r is directed into the page; the rest of the figure will be
unchanged.)
If nx > n2, but 0, < 0C, then 0, < 0 as in Figure 25-6; thus sin(0, 0,) will be
negative so that (Er/Ei)
in (25-39) will always b e positive, as will (EJE^ as
before. T h u s there is no phase change u p o n reflection against a medium of lower index,
as long as the angle of incidence is less than the critical angle.
-1.0
90
Figure 25-10. Ratio of

reflected to incident
electric field as given
by Fresnel's equations
for n, < n 2 .
1.0

transmitted to incident
electric field as given by
90 Fresnel's equations for
n, < 1*2-
2 5 - 3 E PARALLEL T O THE PLANE O F I N C I D E N C E
415
It follows from (25-39), and is shown in Figure 25-11, that the transmitted wave is
always zero for grazing incidence (6 i = 9 0 ) .
25-3
E PARALLEL T O THE PLANE OF INCIDENCE

We proceed in a manner similar to that of the last section. Let us assume that all of the
magnetic fields are into the page; this leads to a situation as depicted in Figure 25-12.
F r o m H = k X E / Z and k E = 0, as well as an inspection of the figure, we find the
analogue to (25-26) to be
E = ZH X k
and
t =
H ,1
(25-41)
so that
E t = -ZH(k
n) = - ( k n)
(25-42)
since k H ( = 0 (they are perpendicular), and the values of E a n d H are related by the
factor Z.
Consequently, the continuity of the tangential c o m p o n e n t s of E ( E j T = E 2 T)
leads to
,.(k,. n) 4- Er(kr
n) = ( E , - 2?,)(k f n) = E,(k,
n)
(25-43)
with the use of (25-11) and (25-4). Since the H ' s are always tangential components and
parallel, H l t a n g = H 2 t a n g yields
Ht + Hr = H t =
z, i
= Y
z, 2
(25-44)
Equations 25-43 and 25-44 can now b e solved for the two ratios Er/Ei and Et/Er
D o i n g this, and using a fair amount of algebra and trigonometry, we can obtain all of
the following results in the same way as used in the last section.
c
Figure 25-12. E, is parallel to the
plane of incidence.
416
REFLECTION A N D REFRACTION OF PLANE W A V E S
The general solutions applicable to all values of nx/n2

Zx (k, n) - Z 2 ( k , A)
and 0, are:
Z ^ o s f l , . - Z2[ 1 - (ni/n2)2
sin2 0,]1/2
z x cos 0, + Z 2 [ l ( i / " 2 ) 2 sin 2 0 , ] 1 / 2
Z i ( k , n) + Z 2 ( k , n)
(25-45)
{Et\
1
2 Z 2 ( k , n)
III
2 Z 2 c o s 0,
z x cos 0, + Z 2 [ l ( n x / n 2 ) 2 sin 2 0,]
' ") + Z 2 ( k , n)
(25-46)
When 0, is a real angle (nx < n2 or nx > n2 and 0, < dL), these become
I Er\
\ Ej J
Zx cos 0, Z 2 cos Bt
jux sin 20, ju.2 sin 20,
Zx cos 0, + Z 2 cos 0,
/x1 sin 20, 4- ju2 sin 20,
2Z-, cos 0,
4/x 2 cos 0, sin 0,
Zx cos 0, + Z 2 cos 0,
ju1 sin 20, + ju,2 sin 20,
(25-47)
(25-48)
and, in addition, when ju.2 = jli2, we get the other two Fresnel equations
2 cos 0, sin 0,
tan(0, - 0,)
//n
tan (0, + 0,)
sin(0; + 0,)cos(0, 0,)
(25-49)
The various expressions for normal incidence are

Hxn2 - n2nx
-
Z x -I Z 2
2 Z,
Zx + Z 2
jUi2nx
2[i2nx
n
}ixn2 + l^2 i
and
and
-)
2 +
2nx
n2 +
( M l = /X2)
(25-50)
(Mi =
(25-51)
^2)
With these last results, we have an apparent contradiction in sign. If we look again at
Figures 25-2 and 25-8, we see that for normal incidence when k, and n are in the same
direction, there is no definite plane of incidence so that there is no distinction between
E being perpendicular to or parallel to the plane of incidence and the results of both
cases should reduce to the same thing. Now (25-51) does agree with the corresponding
expressions in (25-33), (25-37), and (25-40), but the various forms for ( E r / E , ) all differ
by a minus sign. This is a result of our original sign conventions for the two cases. In
Figure 25-9, we assumed E r and E, to be parallel and thus treated Er/Ei as a positive
number; for nx < n2, it actually turned out to be negative as we saw. On the other
hand, the opposite directions of E r and E, were taken into account right from the start
in Figure 25-12, so that the ratio Er/Et was also treated as positive. In other words, the
sign difference reflects the difference in our treatment of the two, but each case is
internally consistent and thus consistent with each other.
Plots of the ratios as given by the Fresnel equations (25-49) for nx < n2 are shown in
Figures 25-13 and 25-14. Looking at Figure 25-14 first, we see that it is similar to
Figure 25-11 in that there is no phase change of the transmitted wave, and it again
vanishes at grazing incidence. Since this is also what happened in the previous case, we
conclude that electromagnetic radiation is always completely reflected from a surface of
a dielectric at grazing incidence. (You can verify this for yourself right now by looking
at a light source by reflection from a rough sheet of paper and then gradually increasing
the angle of incidence to 90.)
2 5 - 3 E PARALLEL T O THE PLANE O F I N C I D E N C E
417

reflected to incident
electric field as given
by Fresnel's equations
for n, < n2.
-1.0
Figure 25-14. Ratio of transmitted to incident electric

field as given by Fresnel's
equations for rt, < n2.
W e see, moreover, from Figure 25-13, that we have a case that did not occur before,
that there is an angle 6p for which the reflected wave is zero. The condition for this is
seen f r o m (25-49) to correspond to the denominator of (Er/Et)^
becoming infinite, or
6 p -I- d t p = 90, where 6 [p is the corresponding angle of refraction. The law of refraction
(25-18) then becomes n1 s i n 0 p = n2 sin# ?/ , = n2 s i n ( 9 0 6p) = n2cosOp, so that
tan 0 =
i
(25-52)
a result known as Brewster "s law. The angle 8p is known as the polarizing angle for the
following reason. Consider an incident wave that is unpolarized in the sense that its
electric field has E x and Ey components that are both different from zero. If this wave
is incident at the polarizing angle 6 p , then only the component E L will be reflected, so
that the reflected wave will have only such a component a n d hence will be polarized.
The polarization of light by reflection was first discovered experimentally in just this
way by Malus who was observing sunlight reflected f r o m the windows of the Luxembourg Palace in Paris, and for us it is further evidence for the electromagnetic wave
nature of light.
418
We also see from Figure 25-13, that, when Gi > 6p, the ratio (Er/E,) is negative.
This means that the phase of the reflected wave is suddenly reversed, that is, changed
by 180, as the angle of incidence passes through the polarizing angle.
4j TOTAL REFLECTION (rr, > n 2 , 6, > 0C)

We will see the reason for this n a m e shortly. In this case, 6, is not a real angle, but we
already have results expressed completely in terms of 6 t and they can b e used directly.
It will be helpful in interpreting our results if we b e a little more specific about our
coordinate system. Let us choose the xy plane to be the surface of separation and the
xz plane to be the plane of incidence. This is shown in Figure 25-15 where the y axis is
out of the page. We see that with this choice, kT = kx a n d kn = kz, so that the incident
electric field will have the form
L? F'(K/T ut)
T? e i[kAx sin 0, + z cos 0,) wr]
(25-53)
/ *^0ie
~
i
with the use of (25-12). The transmitted wave will be given by
- cor)
E , = E 0 ,e' (k <
(25-54)
and we have ktx = k,T = kt sin0, from (25-15). T h e quantity in brackets in (25-16) is
now negative so it is preferable to rewrite kf in the f o r m
2
k
K
= k2 = k k2
tz
2
ni
sin 2 6t - 1
kl
sin 6,
= -K2
sin 0,
(25-55)
. The plus sign in

with the use of (25-20). Therefore, ktz = iK, making e iklzz __ e TKzKz
the exponent would lead to a wave that increases indefinitely as the wave travels in the
positive z direction, so we reject it and write ktz = iK where
1/2
K = k,
I sin Oi
"2
tl20S
sin
sin Qr.
1/2
- 1
(25-56)
with the further use of (25-4). Substituting these various results into (25-54), we get the
expression for the transmitted electric field in the m e d i u m of lower index of refraction;
KZgi(k
E, = Ew
- "^^c sin 0, ut)
(25-57)
n,e
This is a wave traveling in the x direction, that is, parallel to the surface of separation,
but with its amplitude decreasing in the z directionperpendicular to the direction of
Figure 25-15. Coordinate system used in discussing

total reflection.
2 5 - 4 TOTAL REFLECTION (n-, > n2,
propagation. The speed of the wave is

co
c
J
v-,
1x ~ k sin 61 ~ n sin 0,
l
1
(c/n2)
sin 6C
sin 0,
( i / i ) sin 0,
0, > 0 C )
419
(25-58)
where v2 is the normal speed of a wave in this medium. Since t > 6c> 2x < V2 and this
wave travels more slowly than a usual plane wave; (25-57) is not a plane wave in the
sense that its value is not constant on a plane perpendicular to the direction of
propagation. Furthermore, the wave speed v2x is not a constant that depends only on
the properties of the medium, but it depends on the angle of incidence as well.
T h e amplitude decreases exponentially with increasing depth of penetration, and
decreases by a factor of \/e in a penetration d e p t h 8Z given by
(*2/2tt)
Sz
(25-59)
[("i/"2)2sin20, - l ] V 2
where \ 2 = 2ir/k2 = 2mc/n2<jo is the wavelength of a normal plane wave in this

medium. This wave that we have described is o f t e n called the evanescent wave and its
properties are indicated very schematically in Figure 25-16. Although not indicated in
the figure, this wave can also be characterized by saying that the planes of constant
amplitude are perpendicular to the planes of constant phase. N o w let us turn to the
reflected electric field.
Using (25-56), we find that (25-31) and (25-45) can be written as
( Er\
EI
Z2 cos 0, - iZl ( K/k2)

Z2cos
/ E\
Zl cos 0, - iZ2 ( K/k2)

Zx cos 6t + iZ2(K/k2)
+ iZx(K/k2)
(25-60)
W e n o t e that both of these have the same general form:

Er
~E,
(A2 + B2)
A - iB
=
A + iB
1 2
/ e~i<p
(A2 + B2)1/2e
up
: p-2i<p
(25-61)
where tan <p = B/A. Therefore, in b o t h cases, \Er/Ei\ = 1 so that the reflected wave
has the same amplitude as the incident wave. A s we will see in the next section, the
reflected energy also equals the incident energy; hence the n a m e total reflection for this
case.
T h e appearance of the factor e ~2,<p in the ratio means, however, that there is a
relative shift in phase of 2<p of the reflected wave with respect to the incident wave. The
angles <p as found from (25-60), (25-56), and (25-35) are given by
tan <px =
tan<p,, =
Zi(K/k2)
Z 2 cos 0,
Z2(K/k2)
Zx cos 0,
/ fixn2 \ [ ( 1 / 2 ) 2 s i n 2 ^ . - l ]
\ /x 2 i j
(25-62)
cos 0,
/ /*2"i \ [Oi/n2)2sin2di
\ /x1 2
1/2
- l]
COS
0,
1/2
(25-63)
Figure 25-16. Schematic

behavior of the evanescent
wave.
420
REFLECTION A N D REFRACTION OF PLANE WAVES
We see that these angles are different and, in fact,

'
a n
^ -
\2
I
U X
tan,p
(25-64)
Thus, for the case n1 = ji2, t a n < p < tancpy since n 2 < n \ and therefore cp < <P|(.
Because of the similarity of the plane containing E, in Figure 25-8 to the plane shown
in Figure 24-7, we conclude that, if a linearly polarized wave is incident at an angle
greater than the critical angle, its perpendicular and parallel components will be
reflected with different phase shifts and the resultant reflected wave will be elliptically
polarized and can be analyzed by the methods we used in Section 24-7.
-5
ENERGY RELATIONS
N o w we are in a position to see how the incoming energy flow is divided between the
reflected and transmitted waves. The average incident power flow will be given by (S, )
and it will have a component normal to the bounding surface and one parallel to the
surface. W h a t will be evidently of physical interest is what happens to the normal
component since it gives what is reflected back into the original medium. Therefore, we
define a reflection coefficient for power R as the magnitude of the ratio of the normal
components:
<Sr> n
<S,> n
(25-65)
n = t(e//01/2|E|2(k-n)
(25-66)
R =
Now we find from (24-107) that
<S>
Since the incident and reflected waves are in the same medium, ( c / m ) 1 / 2 has the same
value ( f i / M i ) 1 / 2 for each of them; in addition, | k r n | = |k, n | = cos 6r = cos 6 i by
the law of reflection (25-11) and (25-14). Taking all of this into account, we find f r o m
(25-65) and (25-66) that
2
R =
(25-67)
and since we have these ratios already, we can easily evaluate R.

In this way, we find from (25-34) and (25-47), for example, that
R =
Rn =
Z2 cos 0t ZL cos 61
Z 2 cos 6i + Zx cos dt
Z j cos 0, Z 2 cos 6,
Z x cos 6t + Z2 cos 9t
(25-68)
(25-69)
and similar expressions can be found for all of the various ways in which we wrote the
ratios of the fields. At normal incidence and ju,x = /a 2 , for example, both (25-40) and
(25-50) give
2
normal
n 2 + nx
If we consider a specific case for which n2/nY
(Mi = M2)
(25-70)
= 1.5, which is typical for glass and air,
2 5 - 6 REFLECTION AT THE SURFACE OF A C O N D U C T O R
421
we find that (25-70) gives R n o r m a l = 0.04 so that about 4 percent of the incident energy
flow is reflected at normal incidence.
Similarly, we can define a transmission coefficient for power T by
T =
(S,) A
Z j cos 6\ E,
<S,> n
Z , cos 0,
(25-71)
where we have used (25-66), (25-12), (25-17), and (24-95). For example, the cases
corresponding to (25-68) and (25-69) can be found from (25-34) and (25-48) to be
T =
4Z, Z , cos 6: cos 0,
'
j
( Z 2 cos 8t + Zx cos 6,)
(25-72)
4Z, Z 2 cos 6, cos 6.

7
( Zx cos 0, + Z 2 cos 6, f
?3
while, for normal incidence (0, = 0, = 0), (25-51) and (25-71) give
4j
[
normal
(n
+ nj
v2
(Mi = Ma)
(25-74)
since Z : / Z 2 = n2/n1 in this case by (25-35).

It is easily verified that all three of these pairs satisfy the relation
R + T = 1
(25-75)
which can also be seen to hold for the most general cases described by (25-29), (25-30),
(25-45), and (25-46). This result expresses the fact of energy conservation since it
essentially says that all of the incident energy appears either in the reflected wave or in
the transmitted wave so that none is lost during the process.
In the particular case of total reflection, the relation (25-61) shows us at once that
R = | Er/Et | 2 = 1 so that all of the incident energy is reflected as we previously
remarked. Therefore, we know that T must be zero for this case, and it will be left as an
exercise to verify this in detail. In some respects, this seems to be a paradoxical
situation. The transmitted fields are certainly different from zero, so that the average
energy densities ( u e ) and (,) are different from zero, and it seems strange that there
can be energy in the second medium, when none can be transmitted into it. Our
discussion of this case has actually been restricted to the steady-state in which all of the
waves do not change their character with time so that quantities like R and T are
constants. If we imagine how this system was started, say, by turning on a beam, then
the initial transient behavior must have been different from the final steady state to
which the system eventually settles down. It is these beginning processes that we must
regard as the source of the original energy transmitted into the second medium, and we
are not in a position to try to consider this in detail.
25-6
REFLECTION AT THE SURFACE O F A C O N D U C T O R
An important situation arises when a wave in a dielectric is incident upon a conducting

medium such as a metal. We know from Section 24-3 that the transmitted wave will be
attenuated because of resistive heat losses, but we do not yet know how much of the
incident energy will enter the conductor.
It will be sufficient for our purposes to consider only normal incidence; in this case,
we do not need to distinguish between incident fields parallel to or perpendicular to the
plane of incidence. Again, let us choose a specific coordinate system like that of Figure
422
25-15 for which the xy plane coincides with the surface of separation. We assume that
all electric fields are along the x axis and thus are led to Figure 25-17. T h e unit
propagation vectors are seen to b e
k, = k = k, = z
(25-76)
and the electric fields will thus have the form

Ey = x E 0 i e ^ z - u ^
Er = *E0re
kiZ ut)
(25-77)
E, = x E 0 t e ^ z - u t )
where
n1w
k
k2
0i2
^2
N2OJ
(25-78)
by (25-4), (24-38), and (24-57); a 2 and /?2 can b e f o u n d for the second medium f r o m
(24-42) a n d (24-43) or (24-45^1 and (24-46). The magnetic fields can be f o u n d f r o m the
general relation H = (k/fxo))k X E given by (24-93), and, with the use of (25-76), we
find that
H, = y
kx
-yE.
co
H,
H,
(25-79)
ju, 2co
The b o u n d a r y conditions giving the equality of the tangential components of E and

H hold for z = 0 and all t. When the common factor e~'ut is canceled f r o m b o t h sides
of the result obtained by using (25-77) and (25-79), we are led to the two equations
E0i + Eor ~ E0,
kt
(^0/
fl1<0
k~
0r)
JU,205
(25-80)
"0t
which, when solved for the two ratios, lead to

Or
E0I
- (Mi/M2)(fe2Ai) _
1 + (n1/n2)(k2/k})
- (Pi/P-2)(c/nia)(a2
1 + (Mi/M2)(^2AI)
+ ifi2)
1 + ( / x 1 / M 2 ) ( c / 1 w ) ( a 2 + i(32)
0t
E0i
2
1
+ ( M i / / * 2 ) ( c / i i ) ( a 2 + iPi)
(25-81)
(25-82)
Since b o t h of these ratios are complex, the reflected a n d transmitted electric fields each
have their phase shifted with respect to the incident field.
A<2' e2' 2
Ml, 6l
/
kEi
kr
Figure 25-17. Coordinate system used in discussing reflection at the

surface of a conductor.
2 5 - 6 REFLECTION A T THE S U R F A C E O F A C O N D U C T O R
423
These can be written in an interesting m a n n e r by introducing the complex index of

refraction N2 by means of (25-78) and we find that
E0r
"l - (Ml//*2)^2
E0i
0i
2n
0,
(Ml/M2 ) ^2
nx + ( î/M2 ) ^2
(25-83)
which are exactly of the form previously f o u n d in (25-37) and thus shows us how the
complex index of refraction can really b e used in a physically significant way. These
ratios are often found written in terms of the real indices of refraction by writing
N2 = n2 + i(cfi2/oi), as in (24-57), so that we get
[i - ( m i / M 2 ) 2 ] -
'Or
E0i
E0l
[i + (MI/M2)"2]
+ *(MI/M2)(<#>A>)
2rti
E0i
["i + (/* I/M2)"2]
. . . OA.
(25-84)
(25-85)
+ '(MI/>2)M2A0
In m a n y cases of interest, these can be further approximated by taking jiij = fi 2 .

T h e reflection coefficient for power can now b e found f r o m (25-67) and (25-84) and
is
R
r
l"l
(MI/M2)"2J
T2
(M1/M2)
2
2
(cfii/u)
2 5
8 6
[1 + ( / i : / M 2 ) 2 ] 2 + (Ml / M 2 ) 2 ( ^ 2 / " ) 2
As a special case, we see that as /?2 - oo, i? - * 1. This Umiting behavior tells us that
waves that are very strongly absorbed are very strongly reflected. A n excellent example
of this behavior is provided by the optical properties of gold. This metal appears
yellowish by reflection so that if a sufficiently thin sheet is illuminated by white light,
the yellow to red portions will be practically all absorbed. As a result, the transmitted
light will contain only the green to blue portions of the spectrum, so that the same sheet
will a p p e a r greenish when viewed by transmission. A similar effect can be observed for
copper.
All of our results so far in this section are exact, but it is useful to approximate R for
an i m p o r t a n t limiting case.
Example
"Good conductors," (i.e., metals). W e f o u n d in (24-75) that a2 fi2~ (|jii 2 a 2 co) 1 / 2 ,
a n d therefore n2 = caû
= cfi2/u> = c(n2o2/2u>jL/2
= (Km2Ke2o2/2io 2)1/2 =
(Km2Ke2/2Q2)1/2
1 since Q <: 1 for a good conductor. Therefore, we can write
R
f"I
~ (Mi/m2)"232
+ (FTI/Vifnl
1 + [1 -
[i + (Mi / M 2 ) 2 ] 2 + (Mi / > 2 ) 2 " i
(ft2i/Mi"2)]2
8?^
1 + [1 + ( M 2 " I / M I " 2 ) ] 2
Since n x ~ 1 for a typical dielectric, and if [ j l j / P i is also of order unity, we can take
Iu2nl//u.1n2 = x
1 and expand (25-87) in a power series in x; keeping only the first
nonvanishing term in x, we get R = [1 + (1 x ) 2 ] / [ l + (1 + x)2] (2 2x)/(2 +
2x) = (1 x)/(l
+ x) = (1 x)(l x ) = 1 2x, so that for a good conductor
R
2a2nl
Ml"2
= 1 - 2
( k iK
\
ml
\1//2/
/
2o)e0 \1/2
\
(25-88)
where we also used (24-14) to simplify the result. This mildly complicated expression
has the general form that R = 1 (small quantity) so that R will be close to unity for
good conductors. Thus, essentially all of the energy will b e reflected as is commonly
424
observed to be the case for metals. What little energy does flow into the good conductor
is quickly dissipated by the heat loss associated with the induced currents.
One can find results for an arbitrary angle of incidence on a conducting medium in a
manner similar to that we used for normal incidence. They turn out to be considerably
more complicated, but for a good conductor one finds that the transmitted wave travels
nearly normal to the bounding surface. In the limit a -> oo, its direction of travel is
normal to the surface so that the effect is the same as if it arose from normal incidence.
7 CONTINUOUSLY VARYING INDEX OF REFRACTION

The two media that we have discussed were both assumed to be uniform so that their
properties are independent of position. There are other possibilities that can occur. For
example, the incident wave may be incident on a system composed of a series of
parallel slabs, each slab with its own properties. Then one would have to satisfy the
boundary conditions at each interface; a typical example of this will be illustrated by
an exercise. Another possibility is that the second medium may have an index of
refraction that varies continuously with position, and we want to look into a few of the
consequences very briefly.
As a specific example, consider the case of a plasma for which the index of refraction
can be found from (24-140), (24-139), and (24-13) to be
e2 \1/2
me,,w
(25-89)
where we have now written the number density of electrons as ne. We see that when we
have true propagation (to > coP), then nP < 1. Now suppose that ne is not a constant
but varies with position; then n P will also be a function of position.
The ionosphere provides an example of this situation. The ionosphere is a portion of
the earth's atmosphere where the air molecules have been ionized by radiation from the
sun. Generally speaking, the density of electrons ne increases with height so that nP
decreases with height. The ionosphere is not a stable feature of the earth's atmosphere
in the sense that its thickness, amount of ionization, and location of its lower boundary
can vary considerably with time, often within a matter of hours. Consequently, we
consider only a simplified situation, which, however, is sufficient to illustrate the
principal effects.
Let us assume that there is a definite boundary between the plasma and an
unionized medium; again we choose the bounding surface as the xy plane with the z
axis upward as shown in Figure 25-18. (The reflected wave is not shown.) We assume
that ne increases with increasing z > 0, thus making nP decrease with z while
remaining less than unity. A wave of direction k ; is traveling in the air of index na and
is incident on the boundary with an angle of incidence 0 t . Now Snell's law (25-18) says,
in essence, that if we follow the transmitted wave, then the quantity n sin 6 = const.,
that is, it is "conserved." Therefore, if 0P(z) is the angle made with the z axis by the
transmitted wave at a distance z into the plasma, we will have
nasir\0i
= nP(z)
sin 0P(z)
= const.
(25-90)
Since na>l>nP(z),
and nP is decreasing, we see from this equation that 9P(z)
increases with distance of penetration. In other words, the direction of k, is turning
continuously toward the horizontal, and the transmitted wave follows the curved path
shown. Eventually, it will attain a maximum height zm where k, is horizontal, and then
2 5 - 8 R A D I A T I O N PRESSURE
425
it will turn back and follow the symmetrical p a t h shown, finally returning to the air at
the same angle 6i and thence back to earth. (In radio transmission, the distance
between the point of origin and return is called the "skip distance," and it can be many
h u n d r e d s of kilometers.) In effect, the incident wave has been "reflected" from the
plasma.
At the distance zm, 0P(zm) = 90 so that (25-90) becomes na sin 0, = nP(zm).
Combining this with (25-89), we find that
M O
^ T - ( l
- nliirfe,)
(25-91)
which can be used to evaluate n e at this particular value of zm.
25-8
RADIATION PRESSURE
In Section 21-5, we saw that one can associate m o m e n t u m with an electromagnetic field
in a vacuum, and the momentum density as given by (21-75) is g = ju.0e0S. If we now
want to apply this to plane waves, we will as usual be interested primarily in the time
average of this quantity, which will be
<g> - / V o < s > =
<S>
<25-92)
We k n o w from the example of Section 24-6 that for a plane wave ( S ) is different from
zero and is in the direction of propagation; therefore ( g ) will also be different from
zero and in the direction of propagation.
F o r situations involving reflection, the incident wave will be bringing m o m e n t u m up
to the bounding surface at a certain rate, while the reflected wave will be carrying it
away at another rate. As a result, the m o m e n t u m of the electromagnetic field in
426
medium 1 will be changed, and we know f r o m mechanics that this means that medium
2 has " e x e r t e d " a force on it. But by Newton's third law, the force on medium 2 will be
equal and opposite to this; hence we conclude that an incident electromagnetic field
exerts a force on the medium from which it is being reflected. This effect generally goes
under the name of radiation pressure and we can easily calculate it f r o m results we have
already obtained. (This origin of radiation pressure is strongly reminiscent of that of
the pressure exerted by a gas on the confining walls of its container in the kinetic theory
of gases, and our calculation here will be recognized as very much like that used in
kinetic theory.)
As shown in Figure 25-19a, let us consider the incident fields contained in an
oblique cylinder of cross-sectional area A A and sides c At where A? is a small time
interval. T h e volume of the cylinder is AT = c At A A cos 0,. In this time interval, all of
the m o m e n t u m carried by the fields that is within this volume will be incident on the
b o u n d i n g surface. Since the normal component of m o m e n t u m is of interest for finding
the n o r m a l force, we see that, in the time interval, the initial normal m o m e n t u m as
f o u n d by using (25-92) is
Gin = <g;> n
( S , ) ft A t A A cos 0,
AT =
(25-93)
In Figure 25-196, we have a similar picture for the reflected radiation, and since the
distance of travel c At is the same, the m o m e n t u m of the fields in this volume is that
associated with the same incident fields. Thus, the final normal m o m e n t u m of the fields
in m e d i u m 1 is
Gfn~
( S r ) ft A t A A cos 0,
(25-94)
which is a negative quantity, since the angle between ( S r ) and n is greater than 90.
The change in the normal m o m e n t u m of the fields in medium 1 then is AG l n = Gfn
Gin = (At AA cos 0 , / c ) [ ( S r ) ft ( S , ) ti] which, if we explicitly take the signs into
account, becomes
AC
A t A A cos 0,
!
[ ! < $ > L +
L<S ( > | ]
(25-95)
If we let F.e m be the average force on the fields in medium 1, and F 2 be the force on
medium 2, then F 2 = F e m . Furthermore, f r o m mechanics, the change in normal
m o m e n t u m during the time interval At is AG l n = em ft Ar = F 2 n At = F2n At,
(b)
(a)
Figure 25-19. Calculation of radiation pressure, (a) Incident wave.
(b) Reflected wave.
EXERCISES
427
and therefore
F2~
A A cos 6:
' |<S,> - | + |(S,> A|]
c
(25-96)
This is a positive quantity and therefore is a force on medium 2, that is, in the direction
of n. Since this is proportional to the area A A, we can define a pressure P(d,) as
F2n/AA
and we get
COS 6 i:
P(6,)
c
=
[l(S,)-A| + l<s,> - n|]
(1 + R) cos Oj
| (S,) |
-|<S,> A | = (1 + R ) cos Oj
c
c
(25-97)
with the use of (25-65) and (1-15). This is sometimes more conveniently written in
terms of the average energy density since | ( S , ) | = ( , ) c by (24-111) for an incident
wave in a vacuum; thus (25-97) becomes
P(6,)
= (1 + R)cos26i(ui)
(25-98)
T h e maximum value of P corresponds to 6, = 0 and R = 1, that is, normal

incidence on a perfect conductor ( a -> oo), and we see that
- 2 - ^ 7 ^ = *<,)
c
(25-99)
On the other hand, consider a complete absorber of radiation for which R = 0; this is
not a conductor, but rather an idealized perfectly " b l a c k " object. Then, at normal
incidence, P b l a c k = | < S , ) | / c = (ut) = \Pmax.
T h e numerical values of the radiation pressure are really quite small for the usual
situations encountered. F o r example, the value of the incident solar radiation at the
earth, as quoted in Exercise 24-9, is 1340 w a t t s / ( m e t e r ) 2 . Inserting this into (25-99), we
find that the corresponding maximum radiation pressure is 8.9 X 10 6 n e w t o n s / ( m e ter) 2 = 8.8 X 10 11 atmospheres. In spite of its very small value, the radiation pressure
was measured for light and (25-99) thus verified as early as 1903; the fact that comets'
tails always point away f r o m the sun was once ascribed completely to the pressure of
the solar radiation, but it is now thought to m a k e only a partial contribution to this
effect.
EXERCISES
for the
25-1 Show that using the facts that the normal 25-4 The expression tan 0 = n2/nl
components of D and B are continuous gives no polarizing angle found in (25-52) involved the
new information for the case in which E is per- assumption that /nx = /u,2. Consider the general
pendicular to the plane of incidence. Repeat when case in which media 1 and 2 are nonconductors
but otherwise have arbitrary properties, (a) Show
E is parallel to the plane of incidence.
that ( E r / E / ) y = 0 when 0, is such that tan2 0, =
25-2 Find the expressions analogous to (25-29),
n\n})[n 2 2 nl(n\ i ) ] - 1 . ( b ) Find the
(25-30), (25-45), and (25-46) for the ratios Hr/H)
corresponding expression for tan2 0, in order that
and II,/H,.
(Er/E,) = 0. (c) Now assume that medium 1 is
25-3 Show for the case nx > n2 that the polariz- a vacuum and show that a polarizing angle is
possible for the parallel case only when Kc2 > Ki2.
ing angle is less than the critical angle.
428
(d) Find the corresponding requirement for the

perpendicular case when medium 1 is a vacuum.
25-5 Find the ratios ( E , / E : ) and (EJE^)^ for
the case of total reflection and express them in
terms of the angle of incidence. Show that each
can be written in the general form Ate~'v' and
find the ratio (tan <pfX/tan <p,y). If the incident
wave is linearly polarized, what kind of polarization will the transmitted electric field have in
general?
25-6 Evaluate (S,) for the case of total reflection and then show that (S,) fi = 0 as required
for T to be zero.
25-7 (a) For the case of total reflection, show
that
surface separating two media. (This is not a plane

wave but, if the wavelength is very small compared to both a and b, it can be a good approximation to ascribe plane wave properties to
it.) (a) What are the dimensions of the transmitted beam? (b) What is the total incident power
Pt in terms of (S,)? the transmitted power Pt in
terms of (S,)? (c) If we define a transmission
coefficient T' = Pt/Pt, how does this one compare with the definition of T used in the text?
25-10 Show that for a good conductor, E0r/E0i
= 1[1 2-rr(fi 2 /f l i)(^2/\)0- ~ 0 ] where S2 is
the skin depth in the conductor and Aj is the
wavelength in the incident medium. For the
remainder of this exercise, assume a perfect con-
v
(/*i"2/M2"i)[(M2"i/>i"2) 2 - x ] c o s 0 [ ( n 1 / " 2 ) 2 sin20,. - l ]
tan ( <p - <p) =
r\ 2 ,1 ' 2 a
l ( i / 2 ) - l j sin 8t
(b) An incident wave is traveling in a nonmagnetic medium of index of refraction 4.5. The
second medium is a nonmagnetic glass of index
1.5. The incident wave is linearly polarized such
that the components of E 0 / perpendicular and
parallel to the plane of incidence are equal. Find
0, such that the totally reflected wave is circularly
polarized, (c) Find the minimum value of the
index of refraction of the incident medium for
which it is possible to obtain a circularly polarized
wave by one total reflection against this glass and
find the corresponding angle of incidence.
25-8 Verify that the pairs (25-29), (25-30) and
(25-45), (25-46) lead to expressions for R and T
that satisfy (25-75).
25-9 In Figure 25-20, we show a beam of rectangular cross section a x b incident on a plane
Figure 25-20.
25-9.
The incident beam of Exercise
1/2
ductor (a2 * oo). Assume E0i to be real for

simplicity, and find EX real , that is, the total physical electric field in the incident medium and show
that it is a standing wave with a node at the
conducting surface. Similarly, show that Hj real is
a standing wave with an antinode at the surface.
Does Hj reaI lead or lag EX real in time and by how
much? Find ( u e ) and (um) and show that the
total energy density (u) is independent of position. Find Sj and show that it too is a standing
wave with a node at the surface. Find ( S j ) and
interpret the result. Show that there must be a
surface current density Ky and evaluate it.
25-11 Find the lowest-order approximation to
the ratio E0t/E0i for a good conductor. If there is
a phase difference between E, and E ; , evaluate it,
and tell whether E, leads or lags E, in time.
25-12 Show that for a good conductor the reflection coefficient can also be written in the approximate form R 1 4w(/x 2 /ju, 1 )(5 2 /X 1 )
where S2 is the skin depth in the conductor and
X, is the wavelength in the incident medium.
25-13 Evaluate T for normal incidence on a
good conductor by using the basic definition given
in (25-71) and show that the result is in agreement
with (25-75) and the result of the previous exercise.
25-14 See Figure 25-21. A plane wave traveling
in a medium of impedance Zx is normally incident at z = 0 on a second medium of impedance
Z 2 . The second medium has thickness L and
behind it is another medium of impedance Z 3 ,
which occupies the rest of space, (a) Show that
the ratio of "the reflected and incident electric field
EXERCISES
Figure 25-21.
The waves and media of Exercise 25-14.
amplitudes in the incident medium is given by

EQr
^0/
Z2(Z3
429
earth is 5.98 X 10 24 kilograms, its radius is 6.37
Zj) cos k2L - /'( Zf Z X Z 3 ) sin k2L
Z 2 ( Z 3 + Z x ) cos k2L i(Z2
+ Z1Z3) sin k2L
(You will need the five resultant waves indicated,

that is, the situation is in its final steady state.
Remember: the boundary conditions must be
satisfied at both boundaries simultaneously.) (b)
Show that if Zx
Z 3 , the reflected wave will be
zero when L equals an odd multiple of a quarter
wavelength in medium 2 and Z2 = (Z1Z3)1/2.
(c)
Find the corresponding conditions for zero reflected wave when Zx = Z3 # Z 2 . (d) Light of
wavelength 5 X 10 7 meters is normally incident
in a vacuum upon a large slab of nonmagnetic
glass of index of refraction 1.5. If the glass is to
be coated with a layer of nonmagnetic material in
order that the light not be reflected, find the
required index of refraction and minimum thickness of the coating.
25-15 A curve that is tangent to the direction of
k at each point is called a ray, thus Figure 25-18
shows such a curve, (a) If the distance measured
along the ray is s, show that it follows from
(25-90) that dOP/ds = ( t a n 8 P / n P)(dn P/ds) =
(sin 6P/nP)(dnP/dz).
(b) Show that this result
is also consistent with the general appearance of
the ray as shown in Figure 25-18. (c) Why, when
6P becomes 90, doesn't the ray continue horizontally, rather than bending back down as
shown? (d) Find the dependence oi nP on z that
will result in the path in the plasma being an arc
of a circle of radius a.
25-16 Assume that the earth absorbs all of the
solar radiation incident upon it, and find the ratio
of the total force due to radiation pressure to the
gravitational force of the sun. (The mass of the
X 10 6 meters, and the average distance between

the earth and the sun is 1.49 X 1011 meters.)
25-17 The radiation pressure on a boundary that
is perfectly reflecting for any angle of incidence 0,
is found from (25-98) to be P(0,) = 2 cos 2 0, ( , ) .
Now suppose that one side of the surface is
isotropically irradiated by incident waves of the
same frequency and equal values of ( , ) . "Isotropically" means that all angles of incidence 6:
occur and with equal likelihood, that is, the probability that k, will be in an element of solid angle
dQ: is proportional to dl:. Find the total radiation pressure on the surface and express it in
terms of the total energy density <M)tot a t the
surface, that is, (w) t o [ includes both the incident
and reflected waves.
25-18 In our derivation of the concept of radiation pressure, we used only very general concepts
of momentum and the meaning of a change in
momentum. It is also possible to approach this
from a microscopic point of view by investigating
the forces exerted by the transmitted electromagnetic field on the charged particles within the
second medium. It is difficult to carry this out
except for special cases, but when it is done it is
concluded that radiation pressure arises from the
magnetic force on the charges moving under the
influence of the electric field. A general statement
like this should be independent of the particular
model that was used. Review our discussion of
electromagnetic momentum in Section 21-5 and
convince yourself that the conclusion quoted
above is essentially correct.
FIELDS IN B O U N D E D
REGIONS
In the two preceding chapters, we have considered time-dependent solutions of
Maxwell's equations in the form of plane waves of infinite extent so that they
necessarily exist in unbounded regions. In more realistic cases, we can expect there to
be boundaries (walls) of some sort about the region in which we wish to know the
fields. In such situations, it is fairly clear that the solutions cannot generally be plane
waves with definite values of the fields over an infinite plane since the fields will have to
satisfy boundary conditions at the limits of the region as well as being solutions of
Maxwell's equations.
As soon as we start thinking about bounded regions, it is evident that there can be
many possibilities, both in the shape of the region and in the materials comprising the
boundaries. Accordingly, it will be desirable to restrict ourselves to only a few cases.
We assume that all bounding surfaces are surfaces of perfect conductors, as this will
simplify the boundary conditions considerably. (This assumption is analogous to that
used in mechanics in the study of vibrating strings and membranes, where one assumes
the bounding regions to be perfectly rigid.) The first problems we will consider will
concern the possible transfer of electromagnetic energy along a wave guide, that is, a
tube with open ends. (From everyday experience, we a l r e a a y k n o w that this is possible
from the simple fact that we can see through long straight pipbs.) We will also consider
a resonant cavity, that is, a box of some shape enclosed by perfectly conducting walls.
26-T] BOUNDARY CONDITIONS AT THE SURFACE O F A PERFECT C O N D U C T O R

By the term perfect conductor, we mean one for which a -* oo, or, more precisely, one
for which the ratio Q = eu/o -+ 0. We have already seen in the paragraph following
(24-78) that <2
1 for common metals even at very high frequencies so that Q = 0
should be a good first approximation for metallic boundaries. Since we will be
considering only fields that vary harmonically in time, that is, are proportional to e ,u ",
we see f r o m the form of our general superposition given by (24-18) that we can study
the fields in the conductor by considering the behavior of their plane wave components.
In Section 25-6, we covered the case in which the field was normally incident on a
conducting surface. At the end of that section, it was pointed out that, in the limit of
infinite conductivity, the wave in the conductor travels normal to the surface regardless
of the angle of incidence. Combining these results with (24-61), (24-54), and (24-84), we
see that a component of the superposition for the electric field in the conductor will
have the general form
E t = E 0 T e ~ s / s e i ( a S - u , + 9)
(26-1)
where f is the distance of penetration into the conductor and S is the skin depth. The
subscript r indicates that E T is parallel to the surface of the conductor since it is
transverse to the direction of propagation n of Figure 25-3. Thus, at the surface, E T will
be a tangential component. We found in (24-78) that 5 = (2//xcrw) 1/2 for a good
430
26-2
P R O P A G A T I O N CHARACTERISTICS OF WAVE GUIDES
431
conductor so that 5 -* 0 as a -* oo. Therefore, we see from (26-1) that E T -* 0 as

a -* oo for any value of = 0, that is, the electric field is zero at any point in a perfect
conductor. Since the tangential components of E are always continuous, according to
(21-26), we see that E t
= 0 just outside of the surface. In other words, E has no
tangential component at the surface of a perfect conductor so that E must be normal to
the surface.
The value of B inside the conductor is given by (24-92) to be B = (k/u)k X E T so
that B will also be transverse and e~i/s
since k = n here. Consequently, the
transverse component of B inside will also vanish as a -* oo. Since B has no normal
component, the continuity of the normal components as given by (21-27) shows that
Bnorm = 0 just outside the conductor. Thus, at the surface of a perfect conductor and
outside of it, B has no normal component, that is, it must be tangential to the surface.
Similarly, all of the components of D and H inside the conductor will be e~^ / s and
will vanish as a - oo.
Since these conclusions hold for each component of the superposition of (24-18), we
see that all of the field vectors will be zero inside a perfect conductor. This enables us
to simplify the general boundary conditions given by (21-25) through (21-28). If we
choose medium 1 to be the conductor and medium 2 to be the region adjacent to it, we
can set D r , E x , B 1; and H j all equal to zero, and, if we drop the subscript 2, the
boundary conditions become
n D = oy
E = 0
ft
B = 0
ft
H = Ky
(26-2)
where n is the unit normal vector pointing outward from the surface of the conductor
(since our convention requires that n always be drawn from region 1 to region 2). To
repeat: at the surface of a perfect conductor, E is normal to the surface and B is
tangential to the surface. To put it another way, E has no tangential component while B
has no normal component. It is possible to have a finite surface current density Ky even
though E t a n g = 0 since the conductivity is infinite. In fact it is precisely these surface
currents that keep H t a n g from being zero at the surface, even though it is zero inside.
Similarly, the surface charge density af arises because D norm can be different from zero
at the surface while it is zero inside the conductor.
2f?2\ PROPAGATION CHARACTERISTICS OF WAVE GUIDES
Figure 26-1 shows a wave guide that extends indefinitely in the z direction and of
arbitrary and constant cross section in the xy plane. We assume that the bounding
walls are perfect conductors and that the interior is filled with a l.i.h. nonconducting
medium described by ju and e. If xp is any component of E or B, we know from (24-7)
and (24-12) that it satisfies the equation
(26-3)
where v2 = 1 /[ie and v would be the speed of a plane wave in the medium. We will try
to find a solution of (26-3) in the form of a wave traveling in the z direction, that is,
along the axis of the guide, by assuming that \p has the form
\p(x, y, z, t) = \p0(x, y)eiik*z
ul)
(26-4)
We note that this is not a plane wave since the amplitude \p0 is not a constant but can
vary across the cross section. The quantity k g is the guide propagation constant and can
432
FIELDS IN B O U N D E D R E G I O N S
be written as
2m
(26'5)
g = ~r~
where Xg is the guide wavelength, that is, the spatial period along the guide.
If we substitute (26-4) into (26-3) and cancel the common exponent, we find that the
equation satisfied by the amplitude is
l i s
TP"
"*"
(26
"6)
where
kl - kl - k\
to
2w
kQ = - =
v
X0
(26-7)
(26-8)
We see from (24-17) and (24-27) that, if tp0 were a plane wave of circular frequency to,
then k0 and X0 would be the propagation constant and wavelength respectively; for
this reason, A0 is often called the free space wavelength when there is a vacuum inside
the guide. It is often useful to write kc in the form
277'
k, -
(26-9)

1
1
1
pAc- " w
A0 ~ w
Ag
(26
~10)
It is not enough that we be able to solve (26-6) because the corresponding

component or components must also satisfy (26-2). It is not generally possible for these
boundary conditions to be satisfied simultaneously for any arbitrary value of kc but
26-3
F I E L D S IN A W A V E G U I D E
433
only for specific values of kc and corresponding definite values of Xc. Hence, even
before we consider a specific problem, we can expect to find that only certain values of
kc are allowed or are the eigenvalues of the system. What is the physical significance of
these allowed values that we are sure to encounter?
Suppose we consider a definite mode of the guide, that is, a particular value of kc. If
we also assume a given frequency co, then kg will be given by (26-7) as k2g = kl k2.
There are two possibilities of interest. If k0 > kc, so that A0 < Xc by (26-10), then
kg> 0 and kg will be real and we will have wave propagation along the guide. On the
other hand, if k0 < kc, and hence X0 > Ac, then k2 < 0 and kg will be a pure
imaginary which we can write as kg = i\kg\. Substituting this into (26-4), we find that
\p will have the form ip = *p0e~^k^ze~'ul, which is not a wave but rather a harmonically
oscillating "disturbance" whose amplitude is steadily decreasing as we go along the
guide in the sense of increasing z. Therefore, we have found for a wave guide that we
will get wave propagation only if k0 > kc, or X0 < Xc. For this reason, Xc is called the
cutoff wavelength for this particular mode.
We can also state this result in terms of a cutoff frequency coc defined by
kc = y
(26-11)

k2s = (o>2 - u2)
(26-12)
v
Then wave propagation is possible only if co > <oc, that is, if the applied frequency is
greater than the cutoff frequency. Thus, a wave guide will also act as a high-pass filter
in the same sense that we found for a plasma following (24-139). In fact, we see on
comparing (26-12) with (24-137) that the dispersion relation for a wave guide is
identical in form with that of a plasma and with the cutoff frequency for a definite
mode corresponding to the plasma frequency ioP.
26-3 FIELDS IN A WAVE GUIDE
In order to continue, we must use Maxwell's equations. We assume that there are no
free charges or currents within the guide so that (21-48) through (21-51) become
V E = 0
V H = 0
dH
<9E
(26-13)
V X H - t
v x E _ _ _
Since each component of E and H is assumed to have the general form of (26-4), we
will have
E = {x, y)e'(k*z-ut)
(26-14)
H = J f ( x , y)ei(k*z-"n
(26-15)
so that, if Ea is any component of E, we will have dEa/dz = ikgEa and dEa/dt =

iuEa, with similar expressions for the components of H. It is important to remember
that, although the amplitudes and
depend only on JC and y, we are not assuming
the fields to be transverse, that is, to have no z component. In other words,
has the
general form
{x, y) = <?x(x, y)\
where the separate components
ways.
+ v(x,
y)y + <?z(x, y)z
(26-16)
6y, and <z can depend on x and y in different
434
Substituting (26-14) and (26-15) into (26-13), and canceling the common exponential
factor, we obtain the eight scalar equations:
d$x
dSy
^ + y- + ikssS2 = 0
(26-17)
dx
dy
d3#>x
dJ? v
- + ^ + i ks ^ z = 0
ox
oy
(26-18)
(26-19)
- ik agS,,y
dy
d$z
ik&X
dSy
dx
d<?x
dx
dy
w v
djez
ikvJ?v
g y
dy
dJCy
dx
d*x
dx
dy
(26-20)
/coju.^
(26-21)
iu(.'x
d3fz
ikv8Jtx -
iuifiJfy
ioiee y
= tooec
(26-22)
(26-23)
(26-24)
Solving for Sx by eliminating 3^y between (26-20) and (26-22), and using (26-7) and
(26-8), we find that
i (
d.
k
+ r
1
(26-25)
x
221 g T
kc\
dx
dy
Similarly, we find from (26-19) and (26-23) that
*,-Ti
i /
kc \
dSz
d3fz
dy
dx
(2"26)
and that
i (
* - = \-"TZ
d6\
ij
i I
dsz
djr,
s - ^ \
<26"27)
dJTA
/
(26
-28)
also follow from the pairs (26-19)-(26-23) and (26-20)-(26-22), respectively. These
results show that all four of the transverse components of $ and Jri? are independent of
each other and can be found from the derivatives of the two longitudinal components;
this should simplify our work.
However, in obtaining (26-25) through (26-28), we used only four of the eight
Maxwell equations. If we substitute (26-25) and (26-26) into (26-17), we find that it will
be satisfied provided that
d2S'z
2
^3 z
<26"29'
which must be the case as we already know from (26-6). This same result is easily seen
26-4 RECTANGULAR G U I D E
435
to follow from (26-24). Similarly, (26-18) and (26-21) will be satisfied if

d23<e2
i p -
d2J>TZ
i p -
(26
"30)
which again agrees with (26-6).

Since each of the expressions (26-25) through (26-28) has the form of a sum, it is
possible to regard the general transverse components as a superposition of two
independent waves, one corresponding to S z # 0 and
= 0, and the other to
= 0
and 3VZ # 0. Accordingly, it is convenient to separate the solutions into these two
groups and consider them separately. If Sz = 0, then lies entirely in the xy plane and
is perpendicular to the direction of propagation. Such a wave is called a transverse
electric or TE mode. Similarly, the case
= 0 corresponds to a transverse magnetic or
TM mode. (If both Sz and JFZ are zero, the corresponding mode is called transverse
electromagnetic or TEM; this case requires a separate treatment that we defer to a later
section.)
We can usefully summarize our results in the form of a step-by-step procedure for
solving wave-guide problems.
For a TE mode: (1) set S z = 0 and find 3tCz as the general solution of (26-30) or its
equivalent when written in another appropriate coordinate system such as cylindrical
coordinates; (2) substitute this expression for 3^z into (26-25) through (26-28) to find
the transverse components of the amplitudes; (3) substitute these results into (26-14)
and (26-15) to find the fields E and H; (4) make sure the expressions for E and H (or
equivalently $ and 3V) satisfy the two boundary conditions of (26-2) that E t a n g = 0
and H n o r m a l = 0 (note that the tangential component Ez is already zero for this case);
(5) take the real parts of the resulting expressions for E and H if it is desired to have the
physical fields; and (6), the surface charge density and surface current density can now
be found, if wanted, from the remaining boundary conditions n D = ay and n X H =
Kj where they are of course evaluated on the bounding surface. [As we will see, it is
sometimes more convenient to do step (4) before (2).]
For a TM mode, the same basic procedure is followed with 3VZ = 0 and S'z found
from (26-29). In this case, <f, is generally different from zero within the guide, but since
it is a tangential component at the surface, it must be made to vanish there.
When these procedures are carried out, the fields obtained will automatically satisfy
Maxwell's equations since they are used to obtain the basic results. Furthermore, the
eigenvalues k c , which characterize the various modes, will also be obtained during the
process for the following reasons. When the general solution of (26-29) or (26-30) is
obtained, it will involve constants of integration. It will then be found that the
boundary conditions can be satisfied only for certain values of some of these constants,
and they in turn will determine the allowed values of kc. (As we will see in detail
shortly, the ways in which the kc are found are strongly reminiscent of what occurred
when we solved Laplace's equation in Section 11-4.)
It will be left as exercises to show that the results of this section can be written in an
even more compact and elegant form before it is necessary to specify the shape of the
cross section of the guide. For our purposes, however, it is desirable to proceed at once
to the consideration of a specific and important example.
26-4 RECTANGULAR GUIDE
This guide has a rectangular cross section of sides a and b, which we take to be located
in the xy plane as shown in Figure 26-2. We see from (26-29) and (26-30) that for either
type of mode we have to solve an equation of the form of (26-6). We use separation of
436
^ Figure 26-2.
variables and write *p0(x, y) = X(x)Y(yy,

11-4 lead us to the separated equations
1 d2X
**
1 d2Y
^
Cross section of a rectangular wave guide.
then the same arguments as used in Section
'
m >
- -
(26
>
"31)
so that ( d 2 X / d x 2 ) + k\X = 0 and ( d 2 Y / d y 2 ) + k2Y = 0 where

k\ + k\ = k2
(26-32)
and Ar| is also a constant. Solving these in terms of trigonometric functions, we find
that
x
Po(x' y) = ( Q sin kx + C 2 cos kxx){C3 sin k2y + C^cosk2y)
(26-33)
where the C ' s are constants of integration. This expression for \p0 contains a total of
six constants, and, depending on the mode of interest, we identify (26-33) with either
or S'.
i
Example
TE modes. Here we set <$z = 0 and write

3fz = ( C j sin k^x + C 2 cos fc1x)(C3 sin k2y + C 4 cos k2y)
(26-34)
Since Ji?z is a tangential component, it need not vanish at the boundary so we will not
learn much from (26-34) by itself. On the other hand, S x and $ will be tangential
components at appropriate parts of the boundary and must vanish there, so it will be
useful to find them next. (Of course, 3^x and
would be normal components at
parts of the boundary and must vanish there, so that we could deal with them instead.
As we will see, however, we will end up with the same results.)
With S z = 0, we find that when (26-34) is substituted into (26-25) and (26-26) we
obtain
iioiik 2
Sx =
J ( Q s i n & J J T 4- C2 cos A:JX)(C 3 cos k2y C4s.in k2y)
(26-35)
kc
icofik1
^(C1cosA:1x C 2 sin A: 1 x)(C 3 sin k2y + C4cos k2y)
(26-36)
We see from Figure 26-2 that S x will be a tangential component and must therefore
vanish at y = 0 and y = b. Similarly, <y must be zero at x = 0 and x = a. Evaluating
for the zero values of x and y first, we get
<?x(x,0) = 0 =
( y ) = o
iuukjC-,
j ( Q sin ktx + C2cos kjx)
rC,,
ZcjuAiiC]
72(^3
k
s n k
iy
Qcos
k2y)
(26-37)
(26-38)
26-4 RECTANGULAR GUIDE
437
so that we must have C3 = 0 and Cl = 0. Therefore, at this stage, (26-34), (26-35), and
(26-36) have simplified to
3^z C 2 C 4 cos kxx cos k2y
(26-39)
i(A}[xk2
~2 C2C4 cos kxx sin k2y
Sx=
/cojuA:1
Sy = 2 C2C4 sin kx cos k2y
(26-40)
(26-41)
kc
We still have boundary conditions to satisfy at the two remaining faces. We see from
(26-40) that the requirement #x(x, b) = 0 leads to sin k2b = 0 so that k2b = nm where
n is an integer. Similarly, Ja, >0 = 0 gives the condition that kxa = mm with m an
integer. Thus we have found that
mm
nm
kx =
k2 =
(26-42)
a
b
so that (26-32) shows that the allowed values oi k2 are
m
k2 mn = wm:
c = *c
K
k
2
+
(26-43)
The cutoff wavelengths and frequencies can now be found by using (26-43) in (26-9)
and (26-11). The corresponding guide propagation constants and wavelengths can be
found from (26-7) and (26-5):
Lm
kg
77
\2
(26-44)
I +
The only quantity left undetermined is the arbitrary amplitude C2C4 of 3VZ. If we set
C2C4 = H0, then we can use (26-42) to write (26-39) through (26-41) more explicitly.
Furthermore, we can use (26-39) in (26-27) and (26-28) to find the remaining amplitudes. When all this is done, we find that the amplitudes of a general TE mode in a
rectangular guide become
iuix ( nm \
/ mmx \
/ nmy
IH 0 cos I
I sin
(26-45)
2
k
b
-
<?> =
icoju / mm
k
H0 sin
mmx
cos
nmy
, = 0
yU?
,k
st
mm
*rl \ a
Hq
sin
(26-47)
mmx
a
cos
nmy
~b~
st ' nm - H cos mmx sm nmy

0
k2\ v b ,
a
mmx
nmy
= H0 COS
cos
a
ik
(26-46)
(26-48)
(26-49)
(26-50)
where kc and kg are found from (26-43) and (26-44).

As a check on our results, we see from Figure 26-2 that J^ x will be a normal
component at x = 0 and x a and should vanish there; we see from (26-48) that this
is indeed the case. Similarly, (26-49) shows that JriP = 0 at y = 0 and y = b as it
should. [This is actually not such a surprise because the result of one of the exercises is
that the transverse components in a TE mode are related by
X <F
trans = (k^/cofi)!
TRANS
438
F I E L D S IN B O U N D E D R E G I O N S
and therefore are perpendicular. Thus, if <^trans is made normal to the surface, JC t r a n s
will automatically be tangential as found above. A similar result holds for the TM
modes.]
We recall that we still have to multiply each of these amplitude factors by the
propagation term and take the real part to get the physical fields. For example, when
(26-45) and (26-14) are combined, we get
(26-51)
Since i e~' ( 1 / 2 ) , r , the exponential factor can be written exp i[kgz (tot + \tt)\
which, when compared with (26-50) shows that Ex leads Hz in time by 90. Similarly,
(26-48) and (26-49) show us that Hx and Hy will lead Hz by 90 while Ey will lag Hz
by this same amount according to (26-46).
So far we have not specified m and n other than to say that they are integers. First
of all, we see from (26-45) through (26-50) that if m a n d / o r n is negative, all of the
field amplitudes are unchanged. Thus we can restrict m and n to be positive or zero. If
m and n are both zero, (26-50) gives 3t?z = H0 = const., while the other components are
apparently zero, while (26-44) gives kg = k0 = u>/v. Thus our field would seem to be of
the form Hz = H0e'(kz~u,), which would be a longitudinal field of only one component
traveling along the axis of the guide with the speed v characteristic of a plane wave.
However, k 2 = 0 by (26-43) and, since we divided by k 2 to get (26-25) through (26-28),
we cannot actually use them in this case but must go back to Maxwell's equations
(26-13). Then we would get V H = 0 = dHz/dz = ik0Hz so that k0 = 0 and hence
co = 0; this makes Hz = const., which is also seen to be consistent with the rest of
(26-13) with E = 0. Thus, for m = n = 0, the solution is simply a constant magnetic
field along the axis of the guide; while this is certainly possible, it is of no interest here.
In summary, m and n can be restricted to m > 0 and n > 0 but not m = n = 0.
Thus a given TE mode of propagation can be characterized by the pair of independently assigned integers m and n. It is customary to describe it as a TE m mode.
Since the general TE case is quite complicated, let us consider only a particularly
simple case in more detail.
Example
TEl0 mode. Let us assume that a > b and look for the mode with the smallest kc and
hence the largest cutoff wavelength \ c . We see from (26-43) that this will correspond to
m = 1 and n = 0, that is, to the TE 1 0 mode. Then kcW = <n/a so that \ c = 2a and
coc-= irv/a by (26-9) and (26-11). The cutoff wavelength thus is twice the longest
dimension of the cross section; in other words, it is not possible to "squeeze" a
"bigger" wave into the guide. The guide propagation constant is found from (26-44)
and (26-8) to be
(26-52)
The field amplitudes as obtained from (26-45) through (26-50) are
(26-53)
(26-54)
(26-55)
26-4
RECTANGULAR GUIDE
439
while S x = S z = 0 and J f y = 0. Inserting these into (26-14) and (26-15), assuming for
simplicity that H0 is real, and taking the real parts of the resulting expressions, we find
the only nonzero field components to be
Ey = H0cofi ( ) sin ( ] sin ( k z tot)
' \v I
\ a /
Hx = H0kg^ j s i n | j sin (kgz - ut)
( 7TX \
Hz = H0 cos I I cos ( kgz tot)
(26-56)
(26-57)
(26-58)
We see that the values of Ey are independent of y; hence the electric fields are
straight lines with constant magnitude at a given value of x but with a magnitude that
does vary with x and is a maximum at the center where x = \a. The lines of Hx are
straight with their maximum value at the center as well. The value of Hz, on the other
hand, is zero in the center and has opposite signs on the two sides of the center. In
Figure 26-3, we show Ey and Hx in the xy plane at a given time and where it is also
assumed that H0 and sin(fc g z - co/) are both positive. The solid lines are those of Ey
while Hx is shown by the dashed fines; the arrows underneath indicate a variation of
Hx with position. If we assume that we are at a position z where cos (kgz tot) is also
positive, then, as also indicated in the figure, the lines of Hz as found from (26-58) will
be out of the page in the left half of the figure and into the page in the right half since
the z axis is out of the page. At a distance of half a guide wavelength further down the
guide, kgz tot will have changed by kg Az = (Iv/Xg)(Xg/2)
= ir so that the directions of all of the fields in Figure 26-3 will be reversed as is seen from (26-56) through
(26-58).
Since E and Hx are each proportional to sin (kgz cot) while Hz varies as
c o s ( k g z tot), Ey and Hx will both be zero when \HZ\ has a maximum. Similarly,
\EV\ and | H x \ have maximum values when Hz = 0. The relationships of the fields at a
given time are also illustrated in Figure 26-4, which shows a section of the guide
parallel to the xz plane; the y axis is out of the page. In this case, the dashed lines
show the resultant vector H in this plane; the situation depicted in Figure 26-3 could
then correspond, for example, to a plane perpendicular to the z axis at the location
indicated by the arrows labeled A.
Since our fields are actually waves, their behavior in time can be visualized by
imagining that one is located at a given point and that these figures are traveling by in
the positive z direction. They will be moving with the guide speed vg found from
(26-52) to be
to
(26-59)
8
kg
(,rv/ato)2J1/2
from which it follows that

the speed of a plane wave in the material within the
guide; for a vacuum, we would have vg > c.
Equations 26-57 and 26-58 show that Hx and Hz generally will have different
amplitudes at a given location in the guide. If we recall our discussion in Section 24-7,
we see that this means that the total magnetic field H will be elliptically polarized when
regarded as a function of time.
Since there are nonzero normal components of D and tangential components of H,
there must be free surface charges and currents on the walls of the guide according to
(26-2), and these relations can be used to calculate them from our results. For example,
440
H, = 0
H2[
out)
Figure 26-3.
A
Fields of a TE10 mode in the cross section at a given time.

Hz
Ht = 0
=0
I
/
\
^
Hz(in)
I
I
\ ~(
iI i l
i
i i
! - V
'! /
i
^
\
i
i ii
|
I
I
\
Ev (in)
Figure 26-4.
T
)
i
Y
!
'
/
i
'y (Out)
Fields of a TE10 mode in the xz plane at a given time.
on the face y = 0, the outward normal to the conductor is n = y and we find that
afh-Ti
= (.Ey=
H0u>ijl| j sin | j sin (kgz
u>t)
(26-60)
while on the face y = b where n = y, oy has the same value but is of opposite sign.
There are no surface charges on the faces perpendicular to the x axis since Ex = 0. We
see from (26-60) that a f oscillates in time, and, in particular, its sign changes.
Physically, this occurs because of the flow of free charge on the surfaces, that is, there
must be surface currents. These can be calculated in a s i m i l a r f a n n e r from
= n X H
and (26-57) and (26-58).
^
Example
TM modes. In this case, we set 3VZ = 0 and set Sz equal to the expression for t(/0 given
in (26-33); hence (26-32) is again applicable. This case is actually simpler because Sz
26-5 TEM WAVES
441
can be a tangential component and must vanish for x = 0 and a and y = 0 and b.
Proceeding in the same manner as before, we see that we must now have C2 = C4 = 0
while kv k2, and kc are once again given by (26-42) and (26-43). Thus the TE and T M
modes of a rectangular wave guide have the same set of cutoff wavelengths; the field
configurations can be expected to be different however. Setting CjC3 = E0, we find that
(26-33) gives the starting point for the T M calculation to be
/ mmx \
( nmy \
Sz = E0 sin I
j sin I J
(26-61)
which can now be used to calculate the rest of the field amplitudes by means of (26-25)
through (26-28). We note that m = n = 0 makes Sz, and then all of the other field
components, zero; thus there is no T M ^ mode. Furthermore, if m = 0 or n = 0,
<z = 0 and all of the fields are zero. Thus, it is not possible to have a T M m 0 or TM 0
mode, in contrast to the TE case.
It is possible, of course, to have wave guides with cross-sectional shapes that are not
rectangles; for example, ones that are circular. Such cases can be discussed in the same
systematic manner that we have illustrated for the rectangular guide. The mathematical
details are different (and generally more difficult) but the principles are exactly the
same. We will not do this, however, but instead consider a different type of mode.
26-5 TEM WAVES
As we noted in the paragraph following (26-30), a TEM mode is one for which both S'z
and
are zero. Looking at (26-25) through (26-28), we are tempted to conclude that
all of the components are also zero so that such a mode is not possible. We did,
however, divide through by k2c to get these expressions, so it is best to go back to
Maxwell's equations themselves as given by (26-17) through (26-24). Setting Sz = 0 and
J^ z = 0 in them, we find that they reduce to
a# y
a#
= 0
(26-62)
+
* = 0
dx
dy
dy
dx
dJTy
dJTx
dx
dy
= 0
dJTx
dx
dJ?y
dy
= 0
we
JTX = - - 6 y =y - S y
COJU
kg
je y -
ka
W/LI
60
- <t x
kg
(26-63)
(26-64)
(26-65)
From both (26-64) and (26-65), we find that kg/cofi = coe/A:g, so that
k lX = co2jiie = j
= kl
-L
V
(26-66)
by (26-8) and (24-12); then (26-7) show that kl = 0. Since kg = k0, the TEM wave will
propagate along the guide with the same speed v as a plane wave would; since kc = 0,
the cutoff wavelength is infinite and a T E M wave can have any frequency or wavelength.
Because $ only has components transverse to the direction of propagation, it has th-^
form < = <fxx + # y and there will be a similar expression for
; then (26-64) and
442
(26-65) can also be written simply as

k
j p = -5-2
co/x
(26-67)
where Z = (/i/e) 1 / 2 is the plane wave impedance. Comparing (26-67) with (24-94), we
see that a TEM wave has characteristics similar to those of an unbounded plane wave.
For these reasons, TEM waves are also known as principal modes.
If we substitute (26-64) and (26-65) into the first equation of (26-63), we find that it
gives the first equation of (26-62); the same is true for the second equation in each set.
In other words, (26-63) will be satisfied if (26-62) is. Recalling that Sx and <y are
functions only of jc and y by (26-14), we introduce a scalar function <f>(x, y), and
define Sx = d$/dx and Sy = d4>/dy, that is,
S=
-V4>
(26-68)
We then find that the second equation of (26-62) is satisfied because d2<j>/dy dx =
d2(p/dx dy while the first one becomes
d2<t> d2$
IP + J? ~0
<26"69)
which is just the two-dimensional form of Laplace's equation. Therefore, if we take any
solution of the appropriate two-dimensional electrostatic potential problem that gives
an electric field normal to the perfectly conducting bounding surface, call this field S,
and then find Jff from (26-67), we can use the results in (26-14) and (26-15) to give a
possible TEM mode for the system. The two-dimensional field pattern obtained in this
way will then travel down the guide with the plane wave speed v characteristic of the
medium.
Things are somewhat more complicated, however. If is normal to the surface, then
the gradient of the scalar potential <f> is also normal to the surface by (26-68), and we
saw in Section 1-9 that then <j> must be constant on the surface as illustrated in Figure
1-18. In other words, the conducting surface must be an equipotential surface. But we
found in Section 11-1 that a solution of Laplace's equation, which has the same value
on all points of the boundary, has the same constant value everywhere within the
region. If this is the case, V<p = 0 so that <? = 0 and
= 0 by (26-68) and (26-67) and
there will be no TEM wave after all. Thus a hollow pipe type of wave guide as shown in
Figure 26-1 cannot have a TEM mode.
There can be a TEM mode, however, if the guide consists of at least two conductors
such as the two parallel wires of Figure 11-10 or the coaxial cylinders of Figure 6-12.
The reason is that, with there being more than one part to the bounding surface, the
potential <f> can have a constant value on one part of the surface and a different
constant value on the other portion or portions. In such a case, 4> need not be constant
within the region between the conductors and & = V<p can be different from zero,
thus making a TEM mode possible.
To illustrate these ideas, let us consider the simplest possible TEM mode of a very
important type of guide.
Example
Coaxial line. This consists of two coaxial cylinders of different radii a and b shown in
Figure 26-5. We use plane polar coordinates p and is independent of angle, and we see that
26-5 TEM WAVES
443
Figure 26-5. Fields in the

principal mode of a coaxial
line.
then
4> = A + B In p
(26-70)
where A and B are constants. Inserting this into (26-68) and using (1-85), (26-67), and
(1-76), we get
B
B
6=
p
$
(26-71)
p
Zp
so that has only a radial component while Jd? has only a cp component as is shown
in the figure. Inserting these results into (26-14) and (26-15), and using (26-66), we find
the fields of the wave to be given by
B
B
E =
H =
Qei(k0z-ut)
(26-72)
Zp
The potential difference between the two conductors can be found from (5-11) to be
A0 = J*E ds = - B ] n ^ e * k z - a , ) = L^ 0 e i { k z ~ u , )
which has a maximum value A<f>0 = B \n(b/a),
(26-73)
k$ Q e K k z ~ u t ) ^
(26-74)
pln(b/a)
The current on the inner conductor can be found from (21-39), if we note that when
the path of integration is chosen to be in the plane of the cross section ( dD/dt) da. = 0,
so that
r
(hH- ds=
2 7tB
e^0z-^,)
= i = I
i(k z w,)
(26-75)
444
where the maximum value of the current is 7 / ~ 2vB/Z. (There will be an equal
and opposite current on the surface at p b, since, if we choose the path C to lie in the
conductor for p > b, the line integral of H must vanish since H = 0 within the perfect
conductor, and then no net free current can be enclosed by the path.) Using (26-75), we
can write the magnetic field in terms of the current as
r i(k0z-wt)
H =
<p
(26-76)
27Tp
We see from (26-73) and (26-75) that the current and potential difference are in
phase so that their ratio is constant and given by
A
/^q
277
If
l n ib-\-Z,
(26-77)
y CI
where Zc is called the characteristic impedance of the coaxial line.
I >-6| RESONANT CAVITIES
Suppose that we were to take a portion of length L of the wave guide of Figure 26-1
and cover the open ends with perfect conductors. In this way, we would obtain a
volume enclosed by perfectly conducting walls. We can no longer expect to find the
fields in the form of traveling waves because additional boundary conditions have been
introduced and, at the very least, there will be reflections at the new boundary surfaces
that will arise from the necessity of satisfying the boundary conditions. Recalling
Exercises 25-10 and 24-3, we would now expect the fields to be in the form of standing
waves corresponding to certain definite and characteristic frequencies of oscillation.
Then the total fields can be regarded as a superposition of these normal modes. (This is
very analogous to the mechanical problem of the vibrating string. The introduction of
perfectly rigid supports at the ends of a finite length of string makes the normal modes
take the form of standing waves arising from the superposition of incident and reflected
traveling waves. An arbitrary displacement of the string can then be obtained as an
appropriate superposition of these standing waves.)
Systems of this type are known as resonant cavities or cavity resonators. Evidently
they can be of a wide variety of shapes and sizes, but the general principles for finding
the fields are exactly the same and are most easily demonstrated by means of a specific
example.
Let us consider a closed region with perfectly conducting rectangular walls of sides
a, b, c and with the origin at one corner as shown in Figure 26-6. The cavity is filled
with a l.i.h. nonconducting material described by ix and c. Each field component still
satisfies the wave equation (26-3) but we can no longer assume a solution of the form
(26-4). It is evident from the treatment that led to (24-16), however, that the time
variation can always be separated out in the form e - " 0 ' . Therefore, instead of (26-4), we
assume that \j/ can be written as
*(r,/) = * 0 ( r ) e-f"'
(26-78)
which, when substituted into (26-3), leads to the equation that ip0 must satisfy:
v V o +*o*o = 0
(26-79)
where kl = (u/v)2 again. (This result is actually valid for any kind of coordinate
system and is often called the wave equation without the time.)
We can solve (26-79) in rectangular coordinates by separation of variables. If we
write \p0 = X(x)Y(y)Z(z),
and proceed in the same manner as we went from (26-6) to
2 6 - 6 R E S O N A N T CAVITIES
445
Figure 26-6. A rectangular

resonant cavity.
(26-33), we find that

^ 0 ( r ) = (Cx sin kxx + C2 cos k1x)(C3
sin k2y + C4 cos k2y)
x(C5 sin k3z + C6 cos k3z)
(26-80)
where
k\ + k\ + k\ = kl
(26-81)
and where the C ' s are constants. This can also be verified by direct substitution of
(26-80) into (26-79). Combining this form with (26-78), we can write one of the
components of the electric field as
Ex = (C x sin kxx + C 2 cos kxx)(C3
sin k2y + Q c o s k2y)
x ( C 5 sin k3z + C6cos k3z)e~'"
(26-82)
Now Ex will be a tangential component and must therefore vanish at the faces y = 0
and b and z = 0 and c. We see that this requires that C4 = C6 = 0 and that
MT
k 2 ~ ~r~
b
3=
pTT
(26-83)
where n and p are integers. At this point, therefore, we have

Ex = (C{ sin kxx + C2 cos kxx) sin k2y sin k3ze~'ul
(26-84)
where C[ = ClC3C5 and C2 = C2C3C5. Similarly, we find that

Ey sin /r 1 x(C 3 ' sin k2y + C4' cos k2y) sin k3ze~'ut
Ez = sin k^x sin k2y(C$
sin k3z + C6' cos k3z)e~,oU
(26-85)
(26-86)
and
kx=
WITT
(26-87)
Substituting (26-83) and (26-87) into (26-81), and using (26-8), we find that the possible
frequencies of oscillation in this cavity are given by
co \ 2
I =77"
V
m\2
7/
in
+
+ I-
(26-88)
446
We note that if any two of the integers m, n, p are zero, then all of the components of
E will be zero, as will all components of H as well.
The electric field must still satisfy Maxwell's equations and, in particular, we must
have V E = 0. When (26-84) through (26-86) are substituted into this, we obtain
(kxC2 + k2Cl + k3C) sin k1 x sin k2y sin k3z
+ [(A^C/ cos kxx sin k2y sin k3z) + (k2C3 sin kxx cos k2y sin k3z)
+ (k3C$ sin kxx sin k2y cos k3z)\
= 0 (26-89)
A little thought will convince one that we cannot possibly have this expression always
equal to zero for all values of x, y, and 2 (which can be chosen independently) unless
C; = C3 = C5' = 0 and kxC{ + /c2C4 + k3Cl = 0. If we now let C{ = Ex, C4' = E2,
and CI = 3 , we find that the last condition can be written as
klE1 + k2E2 + k3E3 = 0
(26-90)
while the expressions for the field components finally become

Ex = E1 cos kxx sin k2y sin k3ze~,ut
(26-91)
Ey = E2 sin kxx cos k2y sin k3ze~
lut
(26-92)
Ez = E3 sin kxx sin k2y cos k^ze~,u"
(26-93)
so that Ex, E2, and E3 are seen to be the maximum values of the respective
components.
If we define a vector k with components kx, k2, and k3, and a vector E 0 with
components Ex, E2, and E3, (26-90) can be written as
k E0 = 0
(26-94)
which is very much like the result found for a plane wave in (24-91). Thus, for a given
mode, E 0 must be perpendicular to the vector k = (w7r/a)x + (mr/b)y + (pir/c)z..
The magnetic field can be found from v X E = }i(d\\/dt)
= z'w/xH. For example,
we get
9EZ
dE
ioyaH = = (k2E3 -^kÊ2) sin kxx cos k2ycos k3ze
ay
oz
,
(26-95)
Since k2E3 k3E2 is the x component of k X, E 0 , it is desirable to define a vector H 0

by
1
H 0 = k X E0
(26-96)
C0JLI
for, if we let its rectangular components be Hx, H2, and // 3 , then we can write (26-95)
as
Hx iHx sin kxx cos k2y cos k3ze
(26-97)
Again we note the similarity of (26-96) to the plane wave result given by the first
expression of (24-93). Thus we can associate a set of mutually perpendicular vectors k,
E 0 , and H 0 with each standing wave mode of the cavity.
Similarly, we find the other two components of H to be
Hy = iH2cos kxx sin k2y cos k3ze~'"'
(26-98)
Hz = iH3cos kxx cos k2y sin k3ze
(26-99)
'"
We see that Hx = 0 at x = 0 and x = a, that is, at the walls for which it is a normal
EXERCISES
447
component; similarly, Hv and Hz vanish at y = 0 and b and z = 0 and c, respectively. Thus, the boundary conditions on H have been automatically satisfied once E
was made to satisfy its own boundary conditions. Furthermore, it is easily verified that
the two remaining Maxwell equations that we have not yet used are satisfied, that is,
V H = 0 and V X H = e(<9E/dt). [One needs to use k H 0 = 0 as well as (26-96),
(26-94), and (26-81), the last of which can also be written as k k =
= co2/i>2 = co2jue.]
Each component of E varies as
while the components of H are proportional to
ie~
= e~'t"'+(1/2),rK Thus, the electric and magnetic fields are not in phase in these
standing waves but instead H leads E by 90.
A given k corresponds to a given mode, that is, a given set of integers m, n, p
according to (26-83) and (26-87). Now (26-94) tells us that the vector E 0 must be
chosen to be perpendicular to k. However, there are two independent mutually
perpendicular directions along which E 0 can be chosen and still be perpendicular to k.
Thus, for each possible value of k, there are two possible independent directions of
polarization of E 0 , so that there are two distinct modes for each allowed frequency
given by (26-88). This property is known as degeneracy and is a fundamental and
important feature of electromagnetic standing waves. If a, b, c are all different, then the
various frequencies given by (26-88) will generally be different. However, if there are
simple relations among the dimensions, it is possible that different choices of the
integers will give the same frequency so that we will also have degeneracy, but arising in
a different manner. As an extreme example, consider a cube for which a = b = c so
that(26-88) reduces to
(^)
( a ) ( 2
+/?2
(26-100)
Thus, all combinations of integers that have the same value of m2 + n2 + p2 will have
the same frequency and the modes will be degenerate.
EXERCISES
26-1 Let <fT = <fxx + tf y y be the transverse component of S\ with a similar expression for
For a TE mode, show that JPT = (ikg/kl)
V ,
= (z X ST~)/Ze where Ze = ( k 0 / k g ) Z =
(Xg/\0)Z.
Find the corresponding relations for a
TM mode.
26-2 Using the results of the previous exercise,
show that any solution of (26-6) that vanishes on
the boundary of the guide will lead to a TM
mode. Similarly, show that a solution of (26-6) for
which n V'/'o = 0 on the boundary will lead to a
TE mode.
26-3 Suppose we define a group velocity vG for
propagation along a wave guide by vc du/dkg
as in (24-144). Find % as a function of and
compare with v. Show that vcvg v2 = 1 / f i t and
compare with the results of Exercise 24-26.
26-4 A rectangular wave guide with vacuum inside has a 8 centimeters and b = 6 centimeters. If a wave of frequency v = 4 X 10 y hertz is
to be propagated down the guide, what modes are
possible?
26-5 Consider a wave guide of square cross section. What conditions must the edge a satisfy in
order that it be possible to propagate a TE 1 0
mode but not T E n , T M n , or higher mode?
26-6 The interior of a hollow rectangular wave
guide is filled with a l.i.h. nonmagnetic nonconducting dielectric. Show that the cutoff frequencies are smaller by a factor of a 1 / 2 than if the
interior were a vacuum. If one wanted to build
the dielectric filled guide to operate in the same
manner at a given frequency as the vacuum case,
that is, to keep the cutoff frequencies the same,
should it be made larger or smaller and by what
factor should the dimensions a and b be changed?
26-7 For a TE 1 0 mode in a rectangular guide:
(a) find ( u ) ; (b) find ( S ) ; (c) find the spatial
average values ( ( ) ) and ( ( S ) ) from integrating
your previous results for the time averages over
the cross section of the guide; (d) define a speed
of energy flow va by ( ( S ) ) = ((u))v,jZ
(is this
reasonable?) and compare va with the group
velocity found in Exercise 26-3.
448
26-8 Find the surface current density

for a
TE10 mode in a rectangular guide. Sketch the
current at a given time over at least one spatial
period along the guide.
26-9 Consider a rectangular guide with vacuum
inside operating in the TE10 mode. Find the time
average electrical force per unit area (f ( .) on the
face y = 0.
26-10 For a rectangular guide, assume H0 to be
real, and find all of the real field components for
a TE 20 mode. Repeat for a TE U .
26-11 Find the general expressions for the real
field components in a TM mode of a rectangular
guide (assume E0 to be real).
26-12 Find the time average Poynting vector for
a general TM mode in a rectangular guide. By
integrating over the cross section of the guide,
find the total power transmitted along the guide.
26-13 The rectangular wave guide satisfies the
conditions of Exercise 22-6 for describing an electromagnetic field completely in terms of a vector
potential A. Find A for a TE mn mode and verify
that it satisfies V A = 0.
26-14 Find that superposition of a TE mn and a
TMm(l wave in a rectangular guide that will make
E transverse to the y direction, that is, E = 0
while all of the other components of E and H are
different from zero.
26-15 Consider a cylindrical wave guide, that is,
one with a circular cross section of radius a and
with the z axis coinciding with the axis of the
cylinder. Write (26-3) in cylindrical coordinates,
assume that the dependence of ^ on z and t is
given by (26-4), and find the differential equation
satisfied by ^ 0 . Solve this equation by separation
of variables and show that i(/0 has the form
\p0 = CmJm (kcp)e"'"f> where Cm is a constant and
m is an integer. [Hint: i/>0 must be a single-valued
function of <p, for example, ip0(q> + 2m) =
v//()(<jp).] Furthermore, Jm(17) satisfies Bessel's differential equation
d2J,
1 dJm
di]
1] di\
r +
\ rf
m1
so that Jm is a Besselfunction. Find the condition

which determines the kc for a TM mode corresponding to the value m.
26-16 Find (S) for the mode of the coaxial line
discussed in the text. Find the total power transmitted along the line and express it in a form that
involves Zc.
26-17 Find the surface charge density on the
inner conductor for the mode of the coaxial line
discussed in the text. Show that the surface current density on the inner conductor equals the
surface charge density times the velocity of the
wave.
26-18 Consider a length / of the coaxial line
with inductance L and capacitance C. Show that
Z( = ( L / C ) 1 / 2 . (Note that the result is independent of /; accordingly, L and C are often described as the quantities per unit length.)
26-19 Show that there is no TEM mode for a
coaxial line that can be obtained from any angular dependent term of (11-141), that is, for a <j>
given by <f> = (Ampm + Bmp m)(Cm cos my +
Dm sin mcp).
26-20 Show that the integers m, n, p for the
rectangular cavity can be restricted to be positive
or zero. [Do not forget (26-94).]
26-21 Consider a cubical cavity of edge a. Show
that the lowest frequency is threefold degenerate
with respect to the integers m,n,p. Find all
components of E and H for the mode m = n = 1,
p = 0. Sketch the field lines at a given instant.
Find (S) and the total electromagnetic energy
within the cavity.
26-22 Consider a cylindrical cavity of length /
and circular cross section of radius a. The axis of
the cylinder lies along the z axis and the two
faces are at z = 0 and z = I. Find the possible
frequencies of oscillation for this cavity. (You do
not have to find all of the field components to do
this. Also see Exercise 26-15.)
26-23 The Hertz vector -nc was introduced in
Exercise 22-7. Show that a TM mode can be
described by a Hertz vector of the form ir e i and
find 7Te for a rectangular guide with a vacuum
inside.
27
CIRCUITS A N D
TRANSMISSION LINES
Up to now, our primary interest and emphasis have been on describing electromagnetic
phenomena in terms of fields. Our approach has often been one of obtaining results by
considering macroscopic situations and then rewriting things in field terms, for example, in the way we went from (18-6) to (18-21). On the other hand, in many practical
applications, such as in communications work, the emphasis is on the macroscopic
arrangements of devices such as resistors, capacitors, and inductances known as
circuits. Here the primary interest is in the currents and potential differences (often
called voltages) within these systems, and the field aspects are generally not considered.
Because of its practical importance, the study of circuits has been highly developed,
and there are many books devoted to it. Consequently our brief discussion must be very
limited in scope.
Our first aim is to find appropriate macroscopic equations for these systems. As we
will see, they will generally turn out to be differential equations. There are a variety of
possible approaches, but the simplest and most common one is to use a set of equations
known as Kirchhojf s laws. These basic statements about circuits actually predate
Maxwell's equations and are closely related to the conservation laws for charge and
energy.
It is often convenient to divide these systems into two broad classes. In one, the
various circuit components can be considered to be distinct physical entities connected
together by wires whose electromagnetic properties are neglected. Such systems can be
described as having lumped parameters. In the other class, the construction is such that
the parameters cannot be assigned to particular circuit portions but are generally and
collectively associated with every individual parthence, they are said to be distributed
parameters. We will consider examples of both classes.
-1 KIRCHHOFF'S LAWS
We begin by considering steady currents as would be appropriate for direct current

(DC) circuits. We know from (12-15) that this means that V J = 0 so that
da = 0
(27-1)
by the divergence theorem. Let us apply this to the situation shown in Figure 27-1
where we assume that all of the currents I x , I 2 , . . . , / are filamentary, as on wires, and
the closed surface S surrounds a junction P at which the currents flow toward or away.
We see that, in this case, (27-1) reduces to
t l j - o
(27-2)
j- 1
when the currents are assigned appropriate algebraic signs. This result is known as
Kirchhojf s first law or the junction theorem. It says that the algebraic sum of the
449
450
C I R C U I T S A N D T R A N S M I S S I O N LINES
currents at a junction is zero since charge cannot accumulate there in a steady-state

situation and is clearly expressive of the conservation of charge.
If we take a charge around a complete loop in a circuit, then the net change in
potential must be zero since the potential is a single-valued function and the initial and
final points coincide. In other words, the sum of all of the potential changes encountered (or " voltages" Vj) must add to zero:
L Vj = 0
(27-3)
This is Kirchhoffs second law or loop theorem and is related to conservation of energy
because potential differences are defined in terms of work done per unit charge.
If there is resistance in the circuit, we know from Sections 12-3 and 12-4 that there
will be energy dissipated as heat and the circuit must include a nonconservative source
of emf, such as a battery, to maintain the steady current.
The typical use of Kirchhoffs laws in DC circuits is to obtain enough independent
equations to solve for the unknowns of the system. These are generally the currents if
the emfs and resistances are given. One usually proceeds by assigning the directions of
the currents in some arbitrary manner and using (27-2) and (27-3). The principal
difficulties encountered are in assigning the correct signs to the voltages. This is also the
main problem when the currents depend on the time and are just as easily illustrated
for that.
Our discussion of time dependent currents (AC circuits) in this chapter will be
restricted to "slowly" varying currents. To see what "slowly" means, consider the
particular Maxwell equation (21-22):
dv
V XH = J +
(27-4)
at
(The subscript / has been dropped on J since we are only concerned with free currents
anyhow.) A basic approximation in circuit theory is to neglect the displacement current
dD/dt. If we take the divergence of both sides of what remains we get V (V X H) =
0 = V J. Thus, in this approximation, (27-1) is still valid and we will be led again to
(27-2), except that now it also holds for the instantaneous values of the time dependent
currents:
Elj(t) = 0
(27-5)
What is the physical significance of the neglect of the displacement current? As we

will see in the next chapter, the inclusion of this term leads to radiation of electromag-
27-1
KIRCHHOFF'S^tAWS
451
netic waves which, from our present point of view, is an additional energy loss that we
are neglecting. It will be shown that this means that we are assuming that / A where /
is the maximum linear dimension of the system and X the wavelength. Consequently,
any two equal and opposite current elements I d s will have approximately the same
phase since they are much less than a wavelength apart and, at large distances, their
fields will cancel, that is there will be no radiation and associated energy loss. From
another point of view, it means that the circuit is so small in physical size that the time
needed for the propagation of electromagnetic signals can be neglected.
In order to evaluate these limitations on the theory, we can calculate the wavelength
from A = c/v. We find that for a "power" frequency of 60 hertz, and for a typical
radio frequency of 10 6 hertz, the corresponding wavelengths are 5 X 10 6 meters and
300 meters. However, at a microwave frequency of 1010 hertz, the wavelength is only 3
centimeters and here we clearly have to be concerned. Thus, we conclude that standard
AC circuit analysis cannot be slavishly used at very high frequencies but it certainly
should be adequate for power and most communications systems.
The principal effect of time dependence on KirchhofFs second law is the existence of
induced emfs as described by Faraday's law (17-3), namely înd ~ d<S?/dt. Thus we
will be dealing with four sources of voltages: nonconservative emfs arising from
generators, power supplies, or batteries; potential differences across resistors; potential
differences across capacitors; and induced emfs associated with mutual or self-inductance. Since we are neglecting propagation times, all quantities can be evaluated
throughout the circuit at the same time. The fact that there will be no net change of
potential when a charge is taken around a complete loop will then still hold and we are
led once more to (27-3), but now it applies to instantaneous values:
T.VjU)-0
(27-6)
J
The problem of giving the correct signs to the terms in (27-6) is best handled by
assuming a direction for each current involved, a sign for each dl/dt, and a sense of
traversal around the loop. If one does this, and is consistent with the assumptions
made, then careful attention to the situation at each element will enable one to evaluate
the voltages involved in an unambiguous way. Let us illustrate all this with an example.
Example
RL circuit. Figure 27-2 shows a battery of emf S\ a resistance R, and an inductance L

connected in series. The circuit also contains a switch S. To discuss the general case, we
assume that the switch is closed, there is a current I in the sense shown, and dl/dt > 0.
Let us imagine starting at P and traversing the loop in the same sense as I. We know
from (12-27) that the magnitude of the potential difference across R is JR, and from
(12-25) that the direction of E is in the same sense as I. Therefore, the potential is
Figure 27-2.
An RL circuit.
452
decreasing as we traverse R, so that the change, found from the final value minus the
initial value, will be negative; this gives the value VR = IR for the voltage. We know
from Lenz' law that the emf induced in L has a sense that tends to oppose the change.
Because we have assumed that dl/dt > 0, the voltage VL will tend to decrease / and
hence will have a sense opposite to I; thus, with the use of (17-57), we conclude that
VL = L(dl/dt). The battery terminal marked + is, by convention, the one of higher
potential. This means that a charge passing through it has work done on it, thereby
increasing its potential. Thus, the corresponding voltage is VB = . Inserting these
results into (27-6), we obtain VR+ VL+ VB = 0 = IR L(dl/dt)
+ , or
L
dl
1- RJ =
(27-7)
dt
This first-order differential equation with constant coefficients can be used to find the
current as a function of time, once the initial conditions are known.
The general solution of the homogeneous form of (27-7), that is, with zero on the
right-hand side, is I = Ke~Rt/L
where AT is a constant of integration. A special
solution of the inhomogeneous equation, that is, (27-7) as it stands, is clearly I = /R.
Thus, the general solution, which is the sum of these two, is
<0
/ ( / ) = + KeR,/L
(27-8)
R
All that remains is the determination of K.
For example, let us assume that the switch S1 is initially open and, at t = 0, it is
suddenly closed so that the emf is instantaneously applied to the circuit. The initial
condition for the current is then 7(0) = 0. When this is substituted into (27-8), we find
that K = /R, so that the current at any subsequent time is
I
(t)=^{\-e~Rt/L)
(27-9)
Figure 27-3 shows the current as a function of time; the dimensionless variable t/T is
used where T = L/R is called the time constant or relaxation time of this system. (We
have already found this value of T in Exercise 18-5 where a different initial condition
was used.) We note that as t -* oo, the current approaches the final steady-state value
27-2
T H E SERIES R L C C I R C U I T
453
of S/R. Under these conditions, dl/dt -* 0 and hence VL -* 0 so that there is no

voltage across the inductance and the current is determined solely by the resistance of
the circuit.
27-2 THE SERIES RLC CIRCUIT
We now consider the more elaborate circuit shown in Figure 27-4. A capacitance C has
been added, and now the applied emf can be a function of the time so that & = (t).
As before, we assume that / has the sense shown and that dl/dt > 0. We will still have
VR = IR, VL = -L(dl/dt),
while we can also write the "generator" voltage as
VG = <(t). Assuming the charges on the capacitor plates to be q as shown, and
taking q to be positive, we find from (6-28) that Vc = A<#> = q/C. (We note that this
will be correct in magnitude and sign even if q actually turns out to be negative.)
Inserting these into (27-6), we get VR + VL + Vc + VG = 0 = IR L(dl/dt)
(q/C) -I- , so that

dl
q
L + RI + = (t)
(27-10)
dt
C
Since we are primarily interested in the current, we can differentiate this with respect to
the time, and use / = dq/dt to obtain
d2I
dl
I
d
LT2 _l
+_ R + =
(27-11)
dtf ^
dt
C
dt
Thus, the application of KirchhofFs second law has resulted in a second-order differential equation with constant coefficients.
We note that (27-11) is much like the mechanical equation of motion of a damped,
forced, simple harmonic oscillator. Consequently, we can expect to get analogous
behavior. For example, we recall that if we suddenly apply a periodic force to such an
oscillator, the initial response (displacement) will be nonperiodic. However, after a long
time, the displacement is found to be periodic in time with the same frequency as the
applied force. The nonperiodic part of the solution that is eventually damped out is
called a transient, while the periodic part that persists is called the steady state. It will
be convenient to look at these two types of behavior separately.
1. Transient Response
A transient (damped) response can arise in general for any form of ( t ) , not only for a
periodic one, since it corresponds to the solution of the homogeneous form of (27-11).
\
Figure 27-4.
A series RLC circuit.
454
As a simple illustration of such a case, let us assume that the external emf is provided
by a battery so that = const.; then dS/dt = 0 and (27-11) becomes
L
d2I
dl
(27
c =
"12)
The standard procedure is to look for a solution of exponential form; therefore we try a
solution I = aeyt where a and y are constants. When this is substituted into (27-12), it
gives [Ly2 + Ry -I- (1 /C)]aey' = 0. Since we need a = 0 in order to have a current at
all, the term in brackets must be zero. This requirement determines y that is found to
be
7
R \2
2L
2LJ
1/2
= - 8
~ Tc
(27-13)
so that there are two possible values of y. We write them as y + and y_ corresponding,
respectively, to the plus and minus signs in (27-13). For each possible y, there will be
an associated constant a, so that the general solution of (27-12) has the form
= a + ey+t + a_ey
l(t)
= (a + eSt + a_e-St)e~Rt/2L
(27-14)
The only remaining unknowns are the constants a + and a _ that can be found from the
initial conditions.
As an example, let us assume that, at t = 0, the capacitor is uncharged and the
previously open switch is quickly closed. The corresponding initial conditions are that
7 = 0 and q = 0. First we see that 7(0) = 0 = a + + a_. We can get another independent equation from (27-10) and we find that L(dl/dt)t=0
= . When (27-14) is
differentiated, evaluated at t = 0, and combined with the last result, we obtain
L(a + y + + a_y_) = . Solving the two equations for the a's, we find that a + = a =
/[L(y+ y_)] = /2L8 and therefore
7 ( 0
2l5
"
S ,
'
/ 2 L
(27"15)
The nature of this solution now depends on 8 that in turn depends on the relative
values of R, L, and C.
For example, a common situation corresponds to R being very small or zero. In this
case, the quantity under the square root in (27-13) will be negative, 6 will be imaginary,
and it is more convenient to set 8 = z<o where
=
-LC- ( \-2 LF
1/2
(27-16)
Then (27-15) becomes

7(0 =
$
L co
sin u n t
(27-17)
This is a current that oscillates with the natural circular frequency co but with an
amplitude that decreases exponentially with time. Figure 27-5 shows this current as a
function of time. This behavior clearly corresponds to the " underdamped" case for a
mechanical system. Similarly, situations can be found corresponding to the overdamped
and critically damped cases as will be shown in the exercises.
2 7 - 2 T H E SERIES R L C C I R C U I T
455
2L
R
j2n
3lL
"ir
/ 4TT
Figure 27-5. Underdamped current

as a function of time.
2. Steady-state Response
Now let us assume that the applied emf is itself an oscillating function of the time and,
in particular, has the form
{t)
= <?0 cos cat
(27-18)
where <f0 = const. We also assume that the switch in Figure 27-4 has been closed for a
long time before t = 0 so that any transient currents will have long since been damped
out. Now we want to find a special solution of the inhomogeneous form of (27-11).
We note that we can write = Re[0e'ut] and, since (27-11) is linear and has real
coefficients, the real and imaginary parts of any solution will separately be solutions of
the differential equation. Hence, as in Section 24-2, it will be convenient to deal with
complex quantities and, by convention, take the real part to be the physical current.
Accordingly, we write the emf as
( 0 = <oQeiwt
(27-19)
and assume <f0 to be real. (A sinusoidal time dependence written as

is general
usage in circuit theory in contrast to e - ' " ' that is common practice in the rest of
physics. Therefore we will also use this notation, but only in this chapter.) We now try
a solution of the form I = I0e'ut where I0 is a constant that itself may turn out to be
complex. Substituting this into (27-11) and using (27-19), we obtain
[ o)2L + iuR + ( 1 / C ) ] V * " = ico0eiu'
so that, upon canceling the common factor e'", we find that we can write I0 in the
form
-
(27-2)
where
Z = R +
j = R + iX
(27-21)
Z is called the impedance and X the reactance. If we also write X = XL + Xc, then
XL = 01L and Xc = 1 / w C are the inductive reactance and capacitive reactance,
respectively.
456
The physical significance of a complex impedance is that the steady-state current is

not in phase with the applied emf. We can show this explicitly by writing Z = R + iX
= \Z\e'9 and, by equating real and imaginary parts, we find that
| Z | = (R2 + X2)1/2
oiL
1/2
(JOC
coL (1/coC)
X
tan & = =
R
(27-22)
(27-23)
and therefore
/ ( / ) = 70<?"" =
(27-24)
\Z\
so that the real quantities are

( t ) = <f0 cos cot
l(t) =
cos (cot #)
(27-25)
\ Z \
We see that, if # > 0, then the current lags the emf in time, whereas it is in phase with
or leads if # = 0 or & < 0, respectively. These cases correspond to XL being greater
than, equal to, or less than Xc. For given circuit parameters, we see from (27-23) that
# = 0 for the frequency
(27-26)
Wn =
J l c
Although this is not the same as the natural frequency

that we found in (27-16), the
two will be very nearly the same if the resistance is small enough.
From the form of (27-20), we see that impedance is a generalization of resistance.
However, the impedance is a function of the applied frequency and hence is not
completely characterized by the electromagnetic parameters of the system. This means
that the steady-state current will have different values depending on the frequency. The
frequency for which the current has its maximum value is the resonance frequency and
corresponds to the minimum value of | Z | . We see from (27-22) that this corresponds to
X = 0, which shows that the resonance frequency is exactly co0 that we previously
found to correspond to the current being in phase with the applied emf. At resonance,
the current has the simple value of S/R.
The resistance determines the " sharpness" of the resonance, that is, how quickly the
current decreases from its value at resonance as the frequency is changed from to0. It is
convenient to describe this by introducing the dimensionless quantity Q defined as the
ratio of the inductive reactance at resonance to the resistance:
Q=
1 /LW2
UQL
R
(27-27)
~R\~C
The impedance magnitude can then be written as

IZI = R 1 +
Qu
<02
\ 2
W,l
(x>
1/2
(27-28)
Figure 27-6 shows the current amplitude as a function of co/co0 for a few selected
values of Q. One sees quite clearly how the sharpness of the resonance is decreased as
the resistance is increased in order to make Q smaller. If Q is large enough, we can get
an approximate relation between it and the broadness of the resonance curve. Because
(27-28) is fairly complicated, it will suffice simply to ask at what frequency will the
27-3 M O R E C O M P L I C A T E D SITUATIONS
457
Q = 6.0
0
Figure 27-6.
2.0
1.0
Resonance curves for various values of Q.
second term in the square root equal unitythen the current will be down by a factor y/2
from its maximum. Setting (<2co/co0)2[l (co^/co2)2] = 1, we can solve for (co/co0)2,
and we find that (co/co0)2 = 1 + ( 1 / 2 Q 2 ) + (1/<2)[1 + (1/4> 2 )] 1 / 2 which, for Q
large enough, gives (co/co0)2 1 + (l/>). If we introduce the magnitude of the
corresponding frequency change | Aco |, assume that the resonance is fairly sharp, and
use the plus sign, we get
so that
^ ~ 21 Aco |
(27-29)
Now 21Aco| can be taken as an approximation to the total "width" of the resonance
curve so that we get this simple relation between Q and the width: the larger the Q of a
circuit is the sharper will be the resonance and conversely. (We will leave to an exercise
the question of how this Q is similar to the Q we introduced in Section 24-3.)
The most general solution for the current when the emf is oscillatory will be given by
the sum of the solutions to the homogeneous and inhomogeneous forms of the basic
differential equation; in other words, I(t) will be given by the sum of (27-14) and
(27-24). The constants of integration a+ and a_ will have to be evaluated from the
applicable initial conditions and, of course, will not necessarily have the values which
led to (27-15).
27-31 MORE COMPLICATED SITUATIONS
Impedance as a generalization of resistance has proved to be a very useful concept. As

might be expected, it is possible to combine impedances much as can be done with
resistances and the results are quite analogous.
458
C3
In Figure 21-la, we have two impedances connected in series so that the same
current / passes through them. By (27-20), the individual voltages will be Vx = ZXI and
V2 = Z2I so that the total voltage across the combination is V = VY + V2 = (Zl +
Z2)I = Z^l where
+ Z2. Thus, the combination is equivalent to a single
impedance of value Z eff . This can clearly be generalized to more than two, so that the
equivalent impedance of a group connected in series is
Zs = Zx + Z2 + Z3 + ...
(27-30)
We can analyze the parallel combination shown in Figure 27-76 in a similar way.
The voltages V1 and V2 are equal to the same value V by (27-6) so that the individual
currents are Ix = Vl/Z1 = V/Zx and I2 = V2/Z2 = V/Z2. The total current I as
found from (27-5) is I = Ix + I2 = [ ( 1 / Z j ) 4- ( 1 / Z 2 ) ] F = K/Z e f f . Thus, for parallel
connections, the impedances add in reciprocals, and the effective impedance is given by
1
= + + +
7
^71
^72
(27-31)
^73
(We note that these are exactly the same combining rules found for pure resistances in
Exercise 12-6.)
The only real hazard in using these results is in forgetting that impedances are
complex numbers so that the real and imaginary parts are added separately. For
example, (27-30) would be written as
Z, = (Rx + R2 4- R3 + ...)
+ i(Xx + X 2 + X3 + . . . )
(27-32)
+ (Xi + X2 + X3 + ...)2]1/2
(27-33)
that has the magnitude

| Z J = [(Rx + R2 + R3 + ...)2
Because of (27-31), it has been found to be convenient to define the admittance Y as

the reciprocal of the impedance. Thus, we have
1
1
R - iX
y - ~ - -TTT^ = T T
T - V I = G-iB
(27-34)
Z ~ R + iX ~ R2 + X7
The real and imaginary parts of the admittance, G and B, are called the conductance
and susceptance, respectively. When this is done, then (27-20) can be written as
I0 = Y<f0
(27-35)
which is sometimes useful.

By using these results, it is often possible to analyze more complicated circuits by a
step-by-step process of combining individual parts as either series or parallel combinations.
Example
Series-parallel circuit. We want to find the effective impedance of the circuit of Figure
27-8. The resistance RL includes the resistance of the inductance as well as any other
( a )
(b)
Figure 27-7. (a) Impedances in series. (b) Impedances in parallel.
^
00
27-3 M O R E COMPLICATED SITUATIONS
459
WV
R
Figure 27-8.
A series-parallel circuit.
resistance there might be in this branch. RL and L are in series; they in turn are in
parallel with C, and this whole combination is in series with R. By (27-21) and (27-30),
the first pair have an impedance RL + ioiL. Since the impedance of C is i/wC, we
can use (27-31) and then (27-30) again to find the whole impedance of the circuit:
1
Z = R +
(RL + icoL)
= R +
R +
(-i/oiC)
+ iuL)
(Rl
1 + icoC(RL
Rl
+ ;uL)
+ iu[L{
u2LC)
1 2
CR\]
cO LC) + (O>CRL)2
(1
(27-36)
When this circuit is connected to a generator, one must add the generator impedance to
(27-36) to get the complete impedance of the system if one wants to find the current by
(27-24).
An interesting special case occurs when R L is small enough to be neglected. Then we
find from (27-36) that
\Z\ =
co22rL2
(1 - co2LC)'
1/2
(27-37)
Now, if co2 = COQ = 1/LC, we see that | Z | becomes infinite so that the current will be
zero. For frequencies different from this, the current will be different from zero and the
curve of current as a function of frequency will be qualitatively like an inverted form of
Figure 27-6. This is also a resonance phenomenon, but quite different from the previous
example; this behavior is sometimes called antiresonance.
n
Some circuits are deliberately constructed so that there is mutual inductance in the
system, for example by winding one coil over another as in the second example of
Section 17-4. Sometimes mutual inductance is present simply because of the proximity
of two coils so that the flux produced by one can penetrate the turns of the other. In
any case, one then has to take into account the induced voltage described by (17-50)
that, in our present notation, we can write as Vk^j = Mjk(dlk/dt).
When Kirchhoffs
law (27-6) is used, these voltages must also be included. This is ordinarily no problem
except for a sign ambiguity since we saw that mutual inductance can be a positive or
negative quantity. What one generally needs is a knowledge of the physical arrangement of the windings so that Lenz' law can be used to find the sense of the induced
voltages for the assumed current directions and signs of dlk/dt. Once one has done all
460
this in a consistent manner, the various loop equations can be written down and the
currents calculated.
Example
Coupled circuits. We assume the two series circuits of Figure 27-9 to be coupled by
means of the mutual inductance M of appropriate sign; for simplicity, we assume no
resistance or applied emf in either circuit. (In order to have currents in such a system at
all, we can assume that at one time there were emfs present but they have been
switched out.) The application of KirchhofTs law to each circuit is the same we used to
get (27-10) with the addition of VMl = M(dl2/dt)
and VM2 = M(dl1/dt)
where
M = M12 = M2l by (17-49). Including these in the equation that led to (27-10), and
setting R and $ equal to zero, we get two equationsone for each circuit:
dlx
dt
L2
+ M
dl2
dt
dl2
, dI^ 4" 42
+ M
C2
dt
dt
In order to have only currents involved, we differentiate again to obtain

W
L2
d%
d211
+ Mf2
2
ai~
ai
dt
dt
d2I2
d'ln
dt2
+ M
dA2IUl
2
dt2
+
+
A = 0
Cx
hh
= 0
(27-38)
c2
These equations are reminiscent of those applying to two coupled mechanical oscillators. This suggests that we try to solve them by looking for the normal modes and
normal frequencies of the system, that is, solutions in which both currents oscillate with
the same frequency. Accordingly, we try a solution of the form
Ix = aela-'
I2 = beiu'
(27-39)
where a and b are constants. Substituting these into (27-38), we find that they give
(Vl -
- u>lMCxb = 0
1
"
y
"
'
<o2MC2a + ( l idlL2C2)b = 0
(27-40)
This set of homogeneous equations has a nontrivial solution only if the determinant of
the coefficients is zero, that is, if
(1 -
1 - <L2C2)
= CyC2(LxL2
- JnM*CxC2
- M2)u>An - ( L i C j + L2C2)a>2 + 1 = 0
?2
"S r ^ 2
<t\
M
Figure 27-9.
Two coupled circuits.
(27-41)
2 7 - 4 T R A N S M I S S I O N LINES
461
This is a quadratic equation in co2 and the solutions are

,
( / . , , + i 2 C 2 ) [(L,C, + L2C2)2 - 4 C , C J ( L , L 2 - M 2 ) ] ' / 2

2ClC2(L,L2-M>)
(27 42)
"
The result of Exercise 17-26 shows us that LXL2 M2 > 0 so that there are two
possible positive values of <o2 that we call u>\ and coi according to which sign is used
for the square root. [As a simple check on our work, we can let M = 0 in (27-42), and
we find that co2 = 1/L 2 C 2 and co2_ = l/LlC1. In other words, the normal frequencies
are then just those of the individual uncoupled circuits.]
Since
and co2. are positive, upon taking the square roots, we find that there are
four possible real values of con we can use in (27-39), that is, co + and co_. As each
can have its own multiplicative constant, we see that each current can be written as the
sum of four terms:
L = axe'u+t 4- a2e~'u+l + a-,e"-' + a4e~lu~'
(27-43)
I2 = V,w+' + b2e~iu+t 4- b3eiu-t 4- b4e'"'
Thus each current generally has oscillatory components of both frequencies.
The eight constants of integration (ax, a2,..., b4) that have appeared are not all
independent as the equations (27-38) still have to be satisfied with the use of (27-43).
Actually, the ratio of each corresponding pair can be found from either equation in
(27-40) upon substitution of the appropriate <o2. Thus, there are only four independent
constants that can be evaluated from the initial values of the currents and charges. As
can be expected from looking at (27-42), the algebra involved will be "straightforward
but tedious," so we won't carry this further here but will leave the evaluation for a
simpler case to an exercise.
27-41 TRANSMISSION LINES
Up to this point, we have assumed that the circuits are small enough that the
instantaneous values of the voltages and currents can be assumed to be independent of
position. In many applications to power and communications systems, the circuit is so
large that this is no longer a legitimate assumption and the relevant quantities do
depend on position as well as on time. A typical example is shown in Figure 27-10.
There are two long parallel conductors of length /. They have uniform cross section,
although the cross sections need not be the same. The conductors need not be
physically separated as shown but one may surround the other as in the coaxial line of
Figure 18-1. At one end (x = 0), they are connected to an emf (t) = 0e'ul and an
impedance Z 0 . At the other end (x = I), they are connected to a load impedance Z,.
We want to find the steady state voltage and current distributions along the line.
Figure 27-10.
line.
Typical example of a transmission
462
Since the voltage and current will vary with position, we cannot assume that
parameters such as inductance and capacitance are localized at definite points but are
generally distributed uniformly between the ends. If we let R' and L' be the resistance
from all sources of loss and inductance per unit length of the pair, then the series
impedance of a length dx will be (R' + iosL') dx by (27-21). As there are two
conductors involved, there will also be a capacitance C' per unit length. Furthermore,
if there is imperfect insulation between the conductors, there can be a current leakage
between them that we can describe by a conductance G' per unit length. In other
words, we can describe the line schematically as a whole series of infinitesimal circuits
as shown in Figure 27-11. Since the admittance of a capacitor is Yc = 1 / Z C = 1 /Xc =
iu>C and admittances in parallel add, the (shunt) admittance between the conductors is
(G' + I U C ) dx.
Although we cannot apply KirchhofTs laws to the line as a whole, we can for an
infinitesimal portion of it. If I ( x ) is the current at x, then the current at x + dx is less
than this by the current across the line that, by (27-35), is YV(x) where V(x) is the
voltage between the conductors. Thus we get I(x + dx) = I(x) YV(x) = I(x)
(G' + ioiC')Vdx = I(x) + (91/dx) dx, so that
81
= (G' + iuC')V
(27-44)
ox
Applying the loop theorem to the loop closed off by the dashed lines of the figure, we
find that ~(R' + iuL')Idx V(x + dx) + V(x) = 0 so that
dV
= (R' + io>L')l
dx
(27-45)
By differentiating (27-45), and substituting from (27-44), we can get our equation
involving V only:
d2v
F = 0
(27-46)
where
y = a + i/3 = [(#' + iuL')(G'
+ ;wC')]1/2
(27-47)
Similarly, we find that

d2I
dx2
- y2I = 0
(27-48)
so that the voltage and current each satisfy differential equations of the same form.
Since we are looking only for steady-state solutions, we write these functions in the
form
V(x, t) = V0(x)eiou
I(x, t) = I0(x)ei"t
(27-49)
I
R1 dx
iAAA
Figure 27-11. Schematic representation of an infinitesimal portion of a transmission line.
463
and we find that the spatially dependent amplitudes also satisfy equations of the same
form:
d2fa
,v
d2la
" y o
PMO>
The general solution for I 0 can be written down at once as

/(*)
= axeyx + a2e'yx
(27-51)
where ax and a2 are constants of integration. Substituting this into (27-44), while using
(27-49), we find the general result for the voltage amplitude to be
V0(x)
= Zi( aleyx + a2e~yx)
(27-52)
where
Z. =
R' + iwL'
1/2
(27-53)
G' + iuC'
is called the characteristic impedance of the line and will be measured in ohms.
We can easily see the physical significance of these solutions. If, for example, we
combine the second part of (27-51) with (27-49), and use (27-47), we obtain a2elwt~yx
= a2e~axe'iu>'~Px\
Upon comparing this with (24-51), and taking account of the
difference in notation, we see that this is a damped wave traveling to the right (in the
sense of increasing x) with a speed v and wavelength X given by
CO
2 77
O - J
A- T
(27-54)
Similarly, the first term becomes a 1 e a x e'^ u t + ^ x ) that is a damped wave traveling to the
left (in the sense of decreasing x). Both waves have the same damping factor a and
speed y. We see from (27-47) that a and /? will generally be functions of frequency as
well as of the characteristics of the line. Thus the wave speed will also be a function of
frequency so that a transmission line is an example of a dispersive medium.
One can obtain a and ft from (27-47) in the same way that was used to obtain
(24-42) and (24-43). The results are
a = ^=r[{R'G'
f3 = ~{-(R'G'
- u>zL'C') + [(J?' 2 + co 2 L' 2 )(G' 2 + co2C"2)] 1 / 2 } 1 / 2 (27-55)

- <a2L'C') + [(/?'2 + O>2L'2)(G'2 + lo2C'2)]
1/2
} 1 / 2 (27-56)
We can find the constants of integration by considering conditions at the terminals.

At the input end (x = 0), we can apply (27-6) to the amplitudes and use a loop going
through Sq, Z 0 , and across the line. In this way we obtain S'Q Z 0 / 0 (0) F o (0) = 0
=
Z0(a1 + a2) Zi( a1 + a2) so that
(Z, - Z 0 K - (Z,. + Z0)a2
= -<?0
(27-57)
Similarly, at the load end, we have Z / / 0 ( / ) V0(l) = 0 and thus

(z,. + Z , ) ^ ! - (z,. - Z,)e-yla2
= 0
(27-58)
Solving these two equations for ax and a2, we obtain

<?0T,e-2yl
=
an
(z, + z 0 ) ( i - r 0 i > - 2 " )

<f0
(z, + z 0 ) ( 1 - r 0 i > - 2 * )
(27
"59)
(27-60)
464
where
Z, - Z0
r 00 =
Z, + Z 0
Z, - Z,
r =
'
Z, + Z,
(27-61)
'
are called the input and output reflection coefficients, respectively.

If we substitute these results into (27-51) and (27-52), we get the complete expressions for the current and voltage amplitudes:
tf20e~yxfl + I > - 2 y ( / ~ x ) ]
hoV
i * )J = T"
77 (z, + z 0 ) ( i - r 0 i > - 2 * )
0Z:e-yx[l
- T,e-2y{,-xA
(Z, + Z)(1 - r i > ^ )
< -
(27-62);
< 27 " 63 >
As a first example, let us consider an infinitely long hne; then e~2yl ~ e~2ai -* 0 as
I -* 00. Similarly, e~2y('~x) - 0 since I x will be positive; then (27-62) reduces to
/ 0 (jc) = 0 e _ Y V ( Z , + Z 0 ). The initial amplitude thus is / o (0) = <f 0 /(Z, + Z 0 ) and is
determined solely by the input impedance Z 0 and the characteristic impedance Z, of
the line. This shows that the line impedance acts in this case as if it were simply in
series with Z 0 . The net result will be a damped wave traveling along the line with speed
v given by (27-54).
Although (27-62) and (27-63) are exact, their physical significance is not too clear.
We can get a useful interpretation of these results by expanding the denominator in a
power series with the use of 1 / ( 1 y) = 1 + y + y2 + y3 +
In this way, we find
that (27-62) can be written as
I 0 (x) = ,-*7Ae-yx
(Z, +
+ Tie-2'-*
+ T0r2e-y(4t-x)
+ r f f i e -y(6'~x)
+ r0r,e-y(2,+x)
+ r02i>-?<4/+*>
+ ... ]
(27-64)
If we now multiply through by e'at to get the total current, we see that each term
represents a traveling wave. Thus the whole current is a superposition of traveling
waves, going in both directions; the amplitude of each component wave is proportional
to various powers of the reflection coefficients. The T's are now seen to give the
fraction of the incident amplitude that is reflected back into the line. The parenthetical
portion of each exponent gives the distance the corresponding wave has traveled. For
example, the sixth term in the expansion corresponds to a wave that has undergone a
total of five reflectionsthree at the load end and two at the input end. It has traversed
the line five times and is headed back to the input end and has gone a distance 1 x
from the load end. This shows us how the steady state is finally attained after a
sufficient number of reflections. If the load impedance equals the characteristic impedance so that Z/ = Z,, then T, = 0 and there are no reflections. In this case, (27-62)
reduces to exactly the same result as for an infinite line.
For many lines, the resistance and conductance per unit length are very small and
one can simplify our results by making some approximations. Let us assume that
co// R' and coC' G' and look for results with only first-order correction terms.
Rather than dealing with (27-55) and (27-56), it is more convenient to go back to
465
(27-47) and we get

y = a 4- ifi = /co/L'C'
,
/
= Wz/c l = iw/Z/C
1 -
i/l'
/'G'
^Z
^C
1/2
iG'
/*'
2 wZ/
i I R'
G'
2to I Z/
C'
2 aC'
so that
{UC
R'
L'
G'
+
w yfUC'
C'
(27-65)
and the wave speed is

v
(27-66)
{Uc'
by (27-54). We see that to this approximation both the wave speed and the attenuation
constant are independent of the frequency and that the wave speed is the same as if the
line had no resistance.
Similarly, the characteristic impedance obtained from (27-53) is
L'
C
1 +
/ ( G'
2 ( C'
R'
L'
z7
C7
(27-67)
since in many cases the line can be treated as resistanceless and then Z ( turns out to be
approximately a pure resistance.
Example
Two-wire line. A very common example of a transmission line consists of two parallel
wires as shown in cross section in Figure 11-10. From (11-53), (10-73), Exercise 17-23,
and (20-66), we get
D
7T
C' =
(27-68)
cosh 1
1
<7T
2A
cosh (D/2A)
and, if A
D, we can use
C' =
ix ID
Z/ = - l n v
\A
77C
\n(D/A)
(27-69)
where yl is the wire radius and D the distance between centers.

In either case, we find from (27-66) that v = 1 / y'juT and is exactly the same as the
speed of a plane wave for a medium characterized by ju, and e as given by (24-12). The
line impedance found from (27-67) is
1
IT
Z. = i/ cosh
'
77 V e
D \
. . .
\ 2^4
i
7T
nr
"
iD\
L N ( ^
/I
(27-70)
For simplicity, let us completely neglect the resistance of the line so that a = 0 and
assume that Z 0 = Z- so that r o = 0. Then (27-63) becomes
V0 =
- r,e-2if"+if*x]
466
We will also assume that Zt is real so that r, is too. If we multiply our last result by
e'wt and take the real part, we find the voltage to be given by
V = X [cos (cor fix) r , cos (cor + fix 2/31)]
(27-71)
Suppose that the load end of the line is open, Z, is infinite, r , = 1 and (27-71)
becomes
V = \S0[COS (cor fix) + cos (cor + fix 2fil)]
cosfi(x
/) cos (cor
fil)
(27-72)
This is a standing wave where the voltage oscillates sinusoidally at each point but with
an amplitude that also varies sinusoidally along the line. In particular, the open end at
x = I has a maximum value so that it is a voltage antinode. There will be points at
which the voltage is zero; these points are called nodes and their locations are given by
the condition that fi(x I) = n0ir/2 where nQ is an odd integer. The distance Ax
between adjacent nodes can be found since fiAx = An0IT/2 IT or Ax = IR/fi = A/2
showing that the nodes (and antinodes) are spaced a half wavelength apart.
Similarly, if the load end is short circuited so that Z, = 0, then 1 ^ = 1 and the
voltage as obtained from (27-71) is
V = (f0 sin fi(x I) sin (cor fil)
This is another standing wave with a voltage node at the short circuited end and with
the same half wavelength spacing between nodes (and antinodes). Thus, in these
respects, the two-wire line acts like an open or closed organ pipe.
Two parallel wires used in this way are often called Lecher wires.
m
Example
Coaxial line. Another very common transmission line is the coaxial line as illustrated in
Figures 6-12 and 18-1. From (6-45) and the principal portion of (18-34), we find that
2 7tc
C ' = 7
r
In ( B / A )
L' =
a
2TT
(b\
In \a)
so that once again the waves travel with the speed of light since u = 1 / \!L'C' = 1 /
The characteristic impedance is
We note that this is exactly the same impedance (26-77) that we found for the coaxial
line by treating it as a wave guide for the transmission of TEM waves.
EXERCISES
27-1 In the RL circuit of Figure 27-2, R = 4
ohms, L 1.5 henrys, and the battery emf is 12
volts. Find the current and the magnitudes of VR
and V, for these times in seconds after the switch
is closed: (a) 0.25, (b) 1.0, (c) 3.0.
27-2 In Figure 27-2, suppose that the inductance L is replaced by a capacitance C. Use
KirchhofFs laws to find a differential equation
from which the capacitor charge q can be found.

Before the switch is closed at t = 0, there is a
charge q0 on C. Find q for all later times. What
is the relaxation time of this system? Under what
conditions will the corresponding current have a
sense opposite to that shown in the figure?
27-3 Show that (27-10) can also be derived by
using conservation of energy in the sense that the
EXERCISES
time rate of change of electric plus magnetic

energy equals the power supplied by the emf less
the rate of resistive loss.
27-4 If (R/2L)2
> (1/LC), then S is real and
the series RLC circuit is said to be overdamped.
Find q as a function of time for the same initial
conditions that led to (27-15). What does q become after a very long time?
27-5 If ( R / 2 L ) 2 = 1/LC, then the series RLC
circuit is said to be critically damped. Since 5 = 0,
(27-14) reduces to I = (a++ a_)e~R,/2L
=
ae R,/2 L so that there is only one constant of
integration, whereas the general solution of a
second-order differential equation requires two.
Show that, in this case, the general solution of
(27-12) has the form / = (a + bt)e~R'/2L
where
a and b are constants. Find q as a function of
time for the same initial conditions that led to
(27-15). Show that | VR | has a maximum value at
a time 2 L/R after the switch is closed and find
its maximum value.
27-6 Consider an underdamped series RLC circuit for which =
cos ut with w =/= co. At
t = 0, the capacitor is uncharged and the previously open switch is suddenly closed. Find the
current for all later times. Does your solution
correctly reduce to (27-24) or (27-25) as t -* oo?
467
27-11 There is a real current / = I 0 cos (ut + e)

in a circuit consisting of a resistance R and
inductance L connected in parallel. Find the voltage V(t) across this circuit.
27-12 In the circuit of Figure 27-8, suppose that
R j is removed from the branch containing L and
put in that containing C. Find Z. Verify that
your result will correctly lead to (27-37) when RL
is negligible. Show that, when u2LC = 1, your
answer and (27-36) both give the same value of
|Z|.
27-13 Consider a special case of the coupled
circuits of Figure 27-9 where Lx= L2 = L and
Cx = C2 = C. The initial conditions at t = 0 are:
/ x = I 2 = 0, qx = q0, q2 = 0. (How could you
achieve this in practice?) Find the currents and
charges for all later times.
27-14 A generator providing the emf S0elu' has
an internal impedance Z 0 . It is connected in
series with a load whose impedance Z, can be
varied. Show that maximum time-average power
will be transferred to the load when Z, = Z<f.
27-15 Some lines are such that the resistance is
large compared to its series inductive reactance
and the capacitive reactance is large compared to
the leakage conductance. For this case, find the
lowest-order approximations for a, fi, v, and Z;.
27-16 By assuming that the line parameters

themselves are independent of frequency, one can
get a "distortionless" line, that is, the propagation
quantities are independent of frequency, provided
that L'G' = R'C'. Show this to be the case by
finding a, v, and Z,. (In practice, this generally
requires unacceptably large values of L' and/or
27-8 (a) Show that the equation determining
G'.)
the relative phase can be written as tan # =
Q(w/to 0 )[l (W 0 /M) 2 ]. Using this, show that, 27-17 It is often more convenient to express the
for a given d, the larger the value of Q, the voltage and current in terms of line input values,
smaller the corresponding value of will be. (b) that is, values at x = 0. Show that (27-62) and
If the voltage across the capacitor is regarded as a (27-63) can be written in the form
function of frequency, show that its maximum
F 0 ( Q ) c o s h [ y ( / - x) + ft]
amplitude occurs when w = to0[l (1 /2Q2)]1//2.
cosh(y/ + ft)
27-9 Show that Q also equals 2 <n times the
KQ(0) sinh [ y ( / x) + ft]
time-average of the total energy of the circuit
0
Z, cosh ( y / + ft)
divided by the average energy dissipated per cycle
when the system is at or very near resonance.
where tanhft = Zt/Zt. Also show that the effecHow does this result compare with the interpretative input impedance can be usefully written in
tion found for the Q defined in Chapter 24 as
terms of the line constants and load impedance as
given in the last sentence of Exercise 24-29?
K 0 ( 0 ) / / 0 ( 0 ) = Z , c o t h ( Y / + ft)
27-10 A series RLC circuit has an additional
inductance L" in parallel with the whole combi= Z, ( Z, + Z, tanh y / ) /
nation. Find the resultant impedance of this new
(Z/ tanhy/ -I- Z,).
system.
27-7 For a system described by (27-25), show
that the time-average power supplied by the emf
is (<?o / 2 1 Z | ) cos #. In this context, cos & is called
the power factor. What conditions will make the
power factor zero? Physically, what is occurring
to give rise to such a situation?
468
27-18 If Z i( and ZOI: are the input impedances

when the line is terminated by a short circuit
(Z, = 0) and an open circuit (Z, -> oo), respectively, show that the characteristic impedance is
given exactly by Z, = (ZSCZOC)1/2.
27-19 For the case of Lecher wires, show that
voltage nodes are current antinodes and vice versa.
27-20 In order to get an idea of magnitudes,
find the ratio b/a necessary to make Z, of a
coaxial line with vacuum inside equal to 60 ohms.
27-21 The time-average power transmitted past
a given point on a transmission line is ( P ) =
R.e(^/0J^*). (Do you agree?) If a2= 0 and Z, is
real, show that ( P )
(1 - |r,| ). Thus, maximum power transfer will occur when Z, = Z,.
27-22 For simplicity, assume that all impedances are real and y = ifi. Show that the depen-
dence of the voltage amplitude on position along

the line can be generally described by \VQ\2 =
A[1 + r,2 2T, cos2/?(/ *)] where A = const.
For this kind of standing wave, in which the
minima are not zero, a useful and measurable
quantity is the voltage standing wave ratio FSWR
defined by ^SWR = I VOlmax/IK)lminR
Show that
^SWR
i + ir,
i - irj
and
I/| =
^SWR
'SWR
+ 1
and that these are correct for both positive and

negative values of r,. (The second result provides
a convenient way of measuring | r71 and hence of
finding the load impedance.) Show that these
results agree with what we found for the Lecher
wire system. Explain the value found for FSWR
when r, = 0. Show that the spacing between adjacent maxima (and adjacent minima) is X/2.
28
RADIATION
In Chapters 24 through 26, we considered time-dependent electromagnetic fields in the
forms of traveling and standing waves. We did not concern ourselves with the problem
of how we could produce such fields if we so desired other than to note that it could be
done in principle with an "appropriate" distribution of charges and currents. Now we
want to consider some aspects of how electromagnetic fields in the form of waves can
be generated; this part of electromagnetism is usually given the name radiation. The
general subject of radiation covers a wide variety of problems and can become quite
complex. Accordingly, we will have to confine ourselves to a few representative,
reasonably simple, yet important examples.
If we recall our treatment of waves, we note that we managed very well by using
Maxwell's equations and dealing with the fields E and B (or H) themselves. That is, we
never found it necessary to use the general time-dependent vector and scalar potentials
that were discussed in Chapter 22. The production of electromagnetic radiation is,
however, much more easily handled in terms of these potentials A and <f>.
28-1
RETARDED POTENTIALS
We restrict ourselves to regions with vacuum properties so that we can set fie = /i0e0 =
1 / c 2 . If we simplify our notation somewhat by omitting the subscript / on the charge
and current densities, the equations determining the potentials are then found from
(22-14) through (22-16) to be
1 d2A
(28-1)
~d,?
1 d2<f>
^
p
T
1 d<$>
V A + ~2 - r - =
c dt
<28-2>
(28-3)
while the fields are obtained from

B = V XA
(28-4)
<9A
E=-V<?>-
(28-5)
according to (22-1) and (22-3). We recall that this set of equations is completely
equivalent to Maxwell's equations.
It will also be helpful for us to remember that we have found the solutions for the
potentials in the static case where all time derivatives are zero; these are given by
469
470
RADIATION
(16-12) and (5-7) and are
AM -
Ho , J(r') dr'
f477" /Jy
R
(28-6)
p(r') dr'
(28-7)
with R = |r r'|. We know that the solutions for the potentials in the time-dependent
case must reduce to (28-6) and (28-7) for a static situation. Unfortunately, we cannot,
for example, simply replace J(r') in (28-6) by J(r', t) because of the time derivatives in
the differential equation for A.
It is possible to solve (28-1) and (28-2) by elegant and general methods, but they are
quite complex. It is preferable for our purposes to use a simpler method that leads to
the correct results and is much more valuable in aiding one's understanding. We will
find the contribution to <f> from a charge element Aq = pAr' in a volume AT' small
enough so that Aq can be treated as a point charge. Then by summing the contributions of all the charge elements, we can get the total scalar potential. Now outside of
AT' there is no charge and (28-2) becomes the homogeneous wave equation
V2<j> -
1 d2<f>
c2 dt2
= 0
(28-8)
Because of the spherical symmetry associated with a point charge, <j>(r, t) can really
depend only on the relative distance R from the charge and not on angles, that is,
<f> = tp(R, t). Using (1-139), we find that V<#> = (d<j>/dR)R = (d<p/dR)(R/R),
and
then by successive use of (1-115), (1-139), and Exercise 1-21, we obtain
d<$> \
1 d
V = v v<l> = V
R +
R
R = V
R JR
R
R JR.
d
d<}>
R JR
JR
d24>
d<j>
JR
JR
d<j> \
2 dtf>
RJR

d2<f>
2 d<f>
JR2
RJR
1 d2>j>
^Ji
H2 JR
dR J
1 d2<(>
dr
= 0
(28-9)
[We note from (1-105) that this result is exactly what we would get by expressing v 2 in
spherical coordinates with R as the radial distance and with an angle-independent <.]
If we now let
x(R, t)
(28-10)
R
and substitute this into (28-9), we find that x satisfies the one-dimensional wave
equation
d2x
dR
"o 22
1 d2x
7
c2 dt2
(v 2 8 - 1 1 )
'
In (24-9), (24-11), and (24-12), we found the solution of this equation to have the form
X = f(R ct) + g(R + ct) where / and g are functions only of the combinations of
variables R + ct or, what amounts to the same thing, of the combinations t + (R/c);
28:,1
RETARDED POTENTIALS
471
then we find from (28-10) that <t> has the form

=
(28
.12)
Each term in (28-12) is a spherical wave traveling with speed c in the radial direction
with respect to the location of Aq and with an amplitude that is constant on a sphere of
radius R centered on the charge; / represents a wave traveling outward from Aq while
g is one traveling inward. Since (28-12) was obtained as the solution of the source-free
equation (28-8), / and g must be determined so that they correspond to the existence
of the charge Aq located at R = 0.
First of all, 4> in (28-12) is to be the potential due to the changing source. If the
function g were included as an ingoing wave, it would arrive at the source at a later
time than t, specifically a time interval R/c later, and the value of g would have to be
the correct value appropriate to a point very near to Aq. But this would mean that g
was already correct at an earlier time t at a distance R away. In other words, the
"effect" described by g would necessarily have to have occurred (or be known) before
the "cause" due to the charge. Since this is contrary to the logical order of these events,
we have to reject the term g on physical grounds and write <j> = f [ t (R/c)]/R.
Now
as we get in the immediate neighborhood of the nearly point charge Aq, that is, as
R -* 0, the potential must reduce to that of a point charge at its instantaneous value
and location, which as seen from (28-7), is Aq(t)/4<jre0R. Thus / [ / (R/c)] must
have a form such that, for R > 0, f ( t ) = Aq(t)/4ire0
or, if y is the general variable in
f , such that
My)
/(y) =
(28-13)
4irt0
Thus, the form of / , and hence of <j>, is finally determined to be
*(r-0 "
" t )
"
~ f )
At
'
(28
"14)
This result tells us that the scalar potential at the field point r at the time t depends on
the value of the charge at an earlier time t (R/c), and the time difference R/c is
exactly the time required for the spherical wave to reach the field point. Thus the charge
does not produce the potential instantaneously but a finite time, corresponding to the
travel time of the wave, is required for the effect to be felt at the field point. The time
t' = t (R/c) is usually called the retarded time.
N o w that we have the contribution to the scalar potential from an element of charge,
we can obtain the total scalar potential due to a charge distribution by summing
(28-14) over all of the volume V containing the charges; the result is
1
, p[r', t - (R/c)] dr'
*(', o = 7 /
47Tq Jy
K
(28-i5>
We see that each charge element must be evaluated at its own retarded time; these
times will generally vary from element to element since, for a given field point r, the
relative distance R = |r r'| will vary with the source point location r'. Thus (28-15)
depends on R explicitly because it appears in the denominator, and implicitly because
it appears in the time variable of the charge density p.
Each rectangular component of (28-1) will have the same form as (28-2); for
example, we have
1 d2Ax
V l J x
~~ci~dtr==
4 7 2
RADIATION
Each such equation will lead to a result like (28-15), so that when they are added
vectorially, we will find that A is given by
a0 ( J 1f r ' , t (R/c)
A(r, t) = ^ f
47t Jy
R
1 dr'
(28-16)
The results (28-15) and (28-16) are known as retarded potentials-, they are often
written in shortened form with brackets about the source term in the integrands to
remind one that they are to be evaluated at retarded times, that is, one often finds
Mo f
[J] dr'
4T7- Jy
[p]dr'
47Te 0 J V '
We note that (28-15) and (28-16) reduce correctly to (28-6) and (28-7) in the static case
when the charge and current densities are independent of time. [Potentials that would
arise from the g term of (28-12) are known as advanced potentials.]
^ 2 MULTIPOLE EXPANSION FOR HARMONICALLY OSCILLATING SOURCES

We will restrict ourselves to finding the fields produced at a point in a vacuum outside
of a finite region containing prescribed sources, that is, we assume that J ( r ' , /) and
p(r',t) are known functions of position and time. Furthermore, we consider only the
important case in which the sources vary harmonically in time with circular frequency
u so that we can write
J ( r ' , t) = 30(r')e~iut
p(r', t ) = p0(r')e~""
(28-18)
where J 0 and p0 can be complex functions as discussed in connection with (24-19).

Thus, if we write them in terms of a real amplitude and phase as
J0(r') = l|Jo(r')k'A
Po( r ') = |Po( r ')k , # c h
(28-19)
where j gives the direction of J 0 , then the real parts representing the physical currents
and charges will be
Jreal = J IJ01 COS ( Ut ~ &c )
preal = |p 0 | COs(o)/ $ c h )
Following our previous results, if we replace t by t (R/c)

Cd
2 7T
k =- =
(28-20)
in (28-18), and set
(28-21)
we find that (28-15) and (28-16) become

e-i"' r eikRp0(r')
4>(r,t) = /
^
47rc0 J v
R
ix0e~iu' , eikRJ0(r')
A(r, t) =
/
Y
4 tt Jy
R
dr'
(28-22)
dr'
(28-23)
We see that the potentials will also vary harmonically in time (as will E and B) and
with the same frequency as the sources. This will simplify our work somewhat as all
time derivatives can then be replaced by iu, that is
dxp
- = - i ^
where ip can be <f>, as well as any component of A, E, or B.
(28-24)
28-2
MULTIPOLE EXPANSION FOR HARMONICALLY OSCILLATING S O U R C E S
473
We can, in fact, also express everything in terms of the vector potential in this case
since the Lorentz condition (28-3) becomes V A (iu/c 2 )<j> = 0 so that
c2
4>=-iV-
03
(28-25)
If we now substitute (28-25) into (28-5), and use (28-4), (28-24), and (1-120), we obtain
E = -V<f> + /toA = / ( c 2 / ) [ v ( v A) + ( c o 2 / c 2 ) A ]
= /(c2/<o)[V X B + V2A + (co2/c2)A]
(28-26)
At points where we want to find E and B, the current density J = 0 so that (28-1)
becomes v 2 A + (co2/c2)A = 0; thus the last two terms of (28-26) vanish, and it
reduces to
c2
c
E = i V X B = i V X B
(28-27)
co
k
Actually this result is no surprise since it is just one of Maxwell's equations; thus, with
a = 0, (24-4) becomes V X B = j u 0 e 0 ( d E / d t ) = /(co/c 2 )E, which is exactly (28-27).
In summary, then, we have found that we need only consider the vector potential A as
given by (28-23); once A has been found, B is obtained from (28-4), and then E from
(28-27).
The integral in (28-23) is difficult to deal with in general, so that it has been found
useful to expand the integrand in a power series in much the same manner that we used
in discussing electric and magnetic multipoles in Chapters 8 and 19. We choose our
origin of coordinates at some arbitrary convenient location within the volume occupied
by the sources, and we are led to the situation shown in Figure 28-1 where r is the
position vector of the field point P. (Compare with Figures 19-1 and 8-1.) In addition,
we can write
r = |r - r ' | = ( r 2 + r'2 - 2 / r ' c o s 6')l/2
= r( 1 + t j ) 1 / 2
(28-28)
(r'\2
+ ( 7 )
(28-29)
by (19-2) and (8-4) where

v
/ r' \
= -2iJ cos 0'+
I r'\2
ij
2(r r ' )
We will assume that r is large compared with the dimensions of the source volume V,
and it will be sufficient for our purposes to keep only terms linear in r'/r. Then we can
Figure 28-1. Relations involved in the multipole expansion of the

vector potential.
474
RADIATION
write
2(r r ' )
t) = -
(28-30)
and, we can use the expansion (8-6) to write

1
1/
1 \
- = 7 ( l - ^ )
(28-31)
N o w let us consider the term e'kR, which we also want to expand in an appropriate
manner.
By (28-21), kR = 2NTR/X SO that this product essentially measures the distance R on
a wavelength scale. If we are dealing with a compact source, R ~ r, and if we then were
to expand elkR in powers of kR and keep only the first few terms, we would be
unnecessarily restricting ourselves to field points nearby the source. However, if D is
some typical dimension of the source volume V, the variations in R r will be of the
order of magnitude of D, and it is not unreasonable to assume that D
X; in this
way, we do restrict ourselves to small sources on a wavelength scale, but we need not
make any assumptions about r as compared to X. In other words, what we really want
to do is to expand the quantity e,k<-R~r) rather than simply elkR. Before we proceed
with this expansion, we want to consider another way of expressing our assumption
that D
X.
For harmonically oscillating systems with position vectors of the form r ' = r'Qe~io",
the velocities will be v' = dr'/dt = /'cor' and hence the maximum values of the
speeds will be v ~ COD. Assuming that i ) < X, we then find that v/c ~ IOD/C ~ D/X
<s: 1. In other words, our assumption, that all of the charges and currents can be
thought of as being contained within a compact volume that remains small compared
to the source-observer distance and to the wavelength during all of the times of
observation, is equivalent to assuming that we are dealing with the "nonrelativistic"
limit of "slowly moving" charges, that is, the charge speeds are small compared to c.
These restrictions to slowly moving charges apply to all of the material in this chapter
even though, in the last section, we will be able to find approximate means of dealing
with certain sources with dimensions comparable to a wavelength.
In order to expand e'kR appropriately, we need to express it in terms of tj. We can
do this by first writing kR = kr + k(R r) and then by using (28-28), we get
k(R r) = kr[( 1 + tj) 1 / 2 1] = \krr\. Inserting this into e'kR and then using the
expansion e" = 1 + u, and keeping only first-order terms, we get
gikR __ gikrgik(R-r)
= eikr{\
îkrî\krt\
+ \ikri})
(28-32)
Combining this with (28-31), we get the approximation

eikR
eikr
(1 ikr)tj >
(28-33)
Finally, inserting the expression for tj given by (28-30) into (28-33), we obtain
(
R
1 + ( 1 - ikr)\
/ r * r' \ ^
J
(28-34)
which, when inserted into (28-23) gives us our desired expansion of the vector potential
in the form
A = Aj + A n
(28-35)
28-2
MULTIPOLE EXPANSION FOR HARMONICALLY OSCILLATING SOURCES
where
u
A
'~
^11
e<(
475
kr u
44irr
r
- <)
lj yM
'
jit0(l - ikr)e'(kr~ut)
4irr2
) d T
'
f (f r ' ) J o ( r ' )
/ j//
<28"36'
(28-37)
since r is constant with respect to the primed variables of integration. (The subscripts
on the A's are temporary and have merely been chosen to reflect the inverse powers of r
that have naturally appeared.)
We note that all of the terms are proportional to e' ( k r ~" l ) so that they represent
spherical waves traveling outward from the origin with speed u/k = c. The amplitude
of each wave has a dependence on r that becomes more complicated with each
successive term in the expansion. Furthermore, each wave is proportional to an integral
of the source current amplitude over the source volume, although the field point still
remains in the integral of (28-37) since the integrand contains its direction ?. Because of
the use of expansions of R like those we used before, we suspect that these integrals are
related to various multipole moments of the source and, in fact, we note that (28-36)
and (28-37) contain exactly the first two integrals we found in (19-4) when we obtained
the multipole expansion of the magnetostatic vector potential. Accordingly, we will
investigate these integrals more closely with these ideas in mind.
The integral in (28-36) is what we called the magnetic monopole moment in Section
19-1 and we found in (19-5) that it was always zero. However, that was for steady
currents for which the equation of continuity reduced to v ' J = 0, which is not at all
the case here. With the use of (28-24), the general equation of continuity (12-13)
becomes v ' - J + (dp/dt)
= V ' J iup = 0 which, when combined with (28-18)
yields
V ' J 0 = /copo
(28-38)
and shows explicitly that the charge and current distributions are not independent.
N o w we find from (1-121) that (J 0 V ')r' = J 0 , so that when (1-129) is applied to the
source region it becomes
< r'(J 0 da') = f [J0 + r ' ( v ' J 0 ) ] dr' = 0
JSR
JY,
(28-39)
since the currents J 0 are zero on the bounding surface S'. Thus, if we also use (28-38),
we see that
f J0dr'=
J yf
f r ' ( v ' J 0 ) dr' icof p 0 (r')r' dr' = /wp 0

Jy
J yr
(28-40)
where p 0 is the (complex) electric dipole moment of the charge distribution amplitude
according to (8-22). Therefore, (28-36) actually represents an electromagnetic field
produced by a varying electric dipole moment and we can relabel A t and write it as
= -
IM0C P
0 eHkr-Ut)
(28-41)
4-nr
The integral appearing in (28-37) is exactly what we called 3) in (19-7). As before, it
is useful to write the integrand as the sum of symmetric and antisymmetric parts by
dividing the integrand into two equal parts and adding and subtracting a "suitable"
quantity. Thus, we get
(f r')J 0 = i [ ( ? r')J 0 - (f J 0 )r'] + [ ( * r')J 0 + (f J 0 )r']
= \{r'
X J 0 ) x r + i [ ( f - r')J 0 + (? J 0 )r']
(28-42)
476
RADIATION
with the use of (1-30) and (1-23). If we now insert this into (28-37), we find that A n can
be written as the sum
A111
=
H0(l - ikr)ei(kr-")
477/-'
- /r r' X J 0 dr'
L J V'
[I j
[(
. _^
X ?
_ ^
) r i ]
( 2 M 3 )
877-r'
By comparing the quantity in brackets in the first term of (28-43) with (19-20), we see
that it is exactly the magnetic dipole moment m 0 of the current amplitude distribution
so that we can write this whole first term as
u n ( l ikr)(m0
Amd =
^
X ?)
L e Hkr- u
(28-44)
The remaining integral in (28-43) will soon be seen to involve the electric quadrupole
moment so that we will want to rewrite it in terms of the charge density by means of
(28-38). Using J 0 = (J0 V') r '> r = V'(r * r') by (19-11) and (1-115), we find that the
integrand of the second term can be written as
{[(? . r ' ) J 0 ] V ' } r ' + r ' [ J 0 - V ' ( f r ' ) ]
= {[(? r')J 0 ] V ' }r' + r ' { v ' [(? r')J 0 ]}. ~ r'(r r ' ) ( v ' J 0 )
and when this is integrated over V
integral becomes
< r'(r r')(J 0 da') -
and (1-129) and (28-38) are used, the second
f r'(r r ' ) ( v ' J 0 ) dr' = -io> f r'(r r ' ) p 0 ( r ' ) dr'
Jy
J yf
(28-45)
since the surface integral again vanishes as in (28-39). Putting this into the second part
of (28-43), we find that it becomes
A
iu0co(l ikr)e'(kr~"")
-
//'
( f
' r')p(r')
dT
(28
'
"46)
so that we can also write A n = A m d + A e q . We see that the integral in (28-46) involves
second moments of the charge density amplitude and hence is related to its electric
quadrupole moment that we first discussed in detail following (8-22). We can now
rewrite this result in a more easily recognizable form.
Since B will be obtained by differentiation via B = v X A e q , we can add a constant
to the ? component of A e q without affecting the fields by (1-104). Accordingly, we add
f T r ' 2 f p 0 ( r ' ) dr'
J'v>3
I/' .1
to the integral appearing in (28-46), which then becomes
\ J f [3r'(f r') - r'2r\ p 0 (r') dr' =
5 y
J
(28-47)
where Q is defined as the vector equal to the integral. We recall from (8-32) that the
components of the quadrupole moment tensor Q a p for a charge density p 0 (r') are given
by
Qap = f (3a'0' - r'28afi)p0(r')
J I/'
dr' = f ( 3 r t f - r'28a/j)p0(r')
j X/'
dr'
(28-48)
28-3
ELECTRIC D I P O L E R A D I A T I O N
4 7 7
where a and /3 are independently taken to be the rectangular coordinates x, y, z. (We

are now using a and /? as indices as well as explicitly writing the components of r' as r'a
in order to avoid confusion in notation.) Let us also write r in terms of its rectangular
components (direction cosines) as ? = l x x + l y y + l z z. Now consider the sum
l
aQp=
a = x, y, z
'
- lar'2Safi)p0dTf
a = x, y, z
= / [3/>'(r r ' ) - r ' % ] p0 dr' = Qp

J yt
where (8-27) has been used. The components of the vector Q are thus given by
Qp=
LQafi
( f i = x,y,z)
(28-49)
a xy y, z
and can be found once the components of the quadrupole moment tensor of p0 are
known and the direction of r has been chosen. Now that we know how to find Q, we
can combine (28-47) and (28-46) to give
/ju0<o(l ikr)Q t l k r a , )
\ a _ . 2:
e
2A-nr
Aeq = -
(28-50)
Finally, collecting all of our separate results, we can write our vector potential as the
sum
A
= A ed + Amd + A eq
(28-51)
as given by (28-41), (28-44), and (28-50). It is convenient to consider separately the

electromagnetic fields arising from these terms. When we do this in the next sections,
we will find that they will lead to fields corresponding to outward-traveling waves that
have nonzero values of the time-average Poynting vector so that the source can be said
to be "radiating." These fields are conventionally called electric dipole, magnetic
dipole, and electric quadrupole radiation, respectively. In other fields of physics,
especially in nuclear physics, these are often given the corresponding symbolic designations of El, Ml, and E2 radiations.
28-3| ELECTRIC DIPOLE RADIATION

We want to consider the field described by A e d . It is most convenient to use spherical
coordinates and, to be specific, we assume that the dipole moment is along the polar
axis (z axis). Thus, we can write p 0 = p0z = p0(cos8r sin 00) by (1-94), and then
(28-41) becomes
Aed =
iu 0 up 0 e'( k r ~ u " )
477T
(cos Or sin 00)
(28-52)
which shows explicitly that, since A e d is parallel to p 0 , it has only an r and a 6

component and is also independent of <p.
The magnetic induction as found from (28-4) and (1-104) is
B = -
Hok2up0
4 77
kr
(krY
sin
(28-53)
478
RADIATION
Combining this with (28-27), we find the electric field to be

E = -
k3p0
2i
47ren
(kr)
+
(kr)
cos 0r
1
kr
(kr-)2
sin 00
(kr)
(28-54)
While these fields are quite complicated, they both have the form of waves traveling
outward with speed c from the source at the origin. The amplitudes of the various
components depend on r in different ways. The electric field has both an ? and a 0
component, each of which has a different dependence on angle and on distance from
the source. Thus, E is not proportional to z = cos 0r sin 00 and hence is not parallel
to the dipole p 0 , although it does lie in the plane defined by p 0 and r. Since B has only a
$ component, the lines of B are circles about the z axis, that is, about the axis of the
dipole. In addition, B is perpendicular to the plane containing p 0 and E; hence B and E
are perpendicular to each other. [Of course we still have to take the real parts of (28-53)
and (28-54) in order to get the physical fields.]
Although these results are exact, it is customary and convenient to consider what
they become in two extreme limiting cases corresponding to whether one is near to the
source or far from it on a wavelength scale. If r
A, so that kr
1, one speaks of
the "near zone," while if r A (kr 1), it is called the "far" or "radiation zone";
the case in which kr ~ 1 is called the "intermediate" or "induction zone" and the
complete expressions for E and B should be used. We consider the two extreme cases
separately.
1. Near Zone
Since kr ^ 1, we can set e'kr 1. Furthermore, since the kr terms in the amplitudes
all appear in the denominators, the highest power of kr will lead to the predominant
term. In this way, (28-53) and (28-54) lead to
icot
PO<?
(2 cos 0r + sin 00)
(28-55)
4m0riH0up0e
47rr2
sin 0<p
(28-56)
These can be put in a form that is even more understandable if we introduce the actual
dipole moment p given by
p = p Q e- i u t = p o e - i u t z
(28-57)
for then we have
E =
BN
4tT 0 r 3
(2 cos 0? + sin 00)
ju0 sin0 dp ^
47rr2
dt
Mo
47irA
dp
dt
(28-58)
X r
(28-59)
with the use of (28-24) and z X f = sin 0q> by (1-94) and (1-92). If we compare (28-58)
with (8-50), we see that E^ has exactly the form of an electric dipole field arising from
a dipole moment p located at the origin. In this case, of course, the dipole is oscillating
with time, but the field point is so close to it that the retardation effects are negligible
28-3
479
and the electric field follows the dipole moment in time as if the effects were
propagated instantaneously.
If we now compare (28-59) with (14-6), we see that B w has the same form as the
static induction produced by a current element I'ds' = dp/dt. This is also consistent
with the expression for the vector potential, since, with the use of (28-57) and kr
1,
A
Mo
- = 4 7 - r l i
/
(28"6)
which corresponds again to a current element dp/dt as is shown by (16-10).

This result that an oscillating dipole moment is equivalent to a current element can
be understood quite easily if we refer to the specific case of our prototype of an electric
dipole consisting of two equal and opposite charges q and q separated by a small
distance ds'. As we saw in (8-44), the dipole moment will be given by p = qds'. Now if
the charges vary periodically, there will be a current I' = dq/dt between the ends of
ds' and thus
dp
dq
= ds' = I'd s'
dt
dt
(28-61)
'
in agreement with the above. This is also consistent with our original assumption made
in the paragraph following (28-31) that the source dimensions are small compared to a
wavelength, that is, ds'
X in this case. Under these circumstances, it is appropriate
to assume that the current / ' is uniform over the length of the dipole.
2. Radiation Zone
Here we assume that kr 1, and we can no longer set e'kr 1. In this case, the
predominant terms in (28-54) and (28-53) will be those with the lowest power of kr in
the denominator, and we get the approximations
E, =
B/j
1c
sin0e i ( *'-"' ) 8
477-e0r
= ^ ^ P o sm0e , i k r - a t ) q>
4tjt
(28-62)
(28-63)
Both of these fields have amplitudes that vary only as 1 / r , and they have the same
dependence upon sin0. In addition, each is perpendicular to the direction of propagation to the field point as given by f and they are perpendicular to each other. In other
words, at large distances both of these fields become transverse waves. Since q> = ? X 0,
we see that these fields are related as follows:
1
Bjj = - r X E j
c
(28-64)
where we used (28-21) and ju0c0 = 1 / c 2 . Comparing this with (24-33) and remembering
that we are in a vacuum and r is the direction of propagation, we see that (28-64) is
exactly the relation between the fields characteristic of a plane wave in free space. These
relations are illustrated in Figure 28-2, which shows the fields at a given time.
Since both
and B R are proportional to sin 0, they both become zero in the
direction of the dipole p0 and have their maximum value for a given r in the direction
perpendicular to the dipole axis.
480
RADIATION
Figure 28-2. Fields at a given time in the

radiation zone of an oscillating electric
dipole.
Because of the relation between E A and

given by (28-64) and shown in the figure,
the Poynting vector S is seen to be radially outward. We can find its time average by
using (24-104):
<S> =
- i - Re(Efl X B*) = Sfr
(28-65)
Since this is positive, there is a net outward flow of energy, from the dipole; hence the
appropriateness of the name "radiation zone." This energy must be supplied by
whatever source is being used to keep the dipole oscillating with constant amplitude.
The magnitude Sf is seen to be proportional to 1/r2 that is the familiar inverse square
law of radiation intensity. (This has occurred because the individual fields vary as 1/r
in this region.) We also see that ? varies with angle as sin 2 6; this is illustrated in
Figure 28-3 that shows y as a function of 0 for a definite value of r. (Since Zf is
independent of <p, the figure can be imagined to be rotated about the dipole axis to give
the surface representing radiation intensity as a function of direction of propagation.)
Finally, we note that Sf is proportional to the absolute square of the dipole moment
and to the fourth power of the frequency; thus, all other things being equal, it is easier
to radiate at a higher frequency.
If we integrate (28-65) over a sphere of radius r, we can find the total rate at which
the dipole is radiating energy, that is, the power
Using (28-65) along with (1-100),
we get
0>= j(S)
da = J^J^r2
sin Q dd dtp =
(28-66)
We have calculated these results by using the approximate values for the fields in the
radiation zone. Physically, however, it is evident that there cannot be nonzero contributions to ( S ) from the terms that drop off with a higher power of r, since if they existed
in the near or intermediate zones, they also would have to be present far from the
source because there is no matter in the intervening region to absorb these contributions. This can also be verified directly. If we use the complete expressions (28-53) and
28-3
481
Figure 28-3. Magnitude of the Poynting vector as a function of angle

for electric dipole radiation.
(28-54) to calculate E X B*, and omit the various constants we find that
E
B* ~
1
(krY
sin 2 Or -
(kr)'
(krY
(krY J
sin 200
(28-67)
Since it is only the real part of E X B* that contributes to the time-average energy flow,
we see that R e ( E X B*) ~ sin 2 6r/(kr)2,
which involves only the 1 /kr terms in the
fields, that is, solely the contribution of the radiation zone. The rest of the terms in
(28-67) are imaginary. The other components of the physical fields do contribute to the
instantaneous value of S, but they give rise to parts that flow alternately away from and
toward the source as well as in the 0 direction and, on the average, are zero.
As we previously noted, the energy that appears as radiation must be supplied by the
external source producing the oscillating dipole. Thus, from the point of view of this
external source, the dipole acts as a dissipative element in much the same way as a
resistance dissipates electromagnetic energy into heat as we found in Section 12-4. It
has been found convenient to describe a radiating system in terms of an equivalent or
"radiation resistance" 3%. In Exercise 12-13 we saw that the total rate at which heat is
produced by a resistor 3$ is 3#I 2 . If, now, the current is oscillating in time, this becomes
3$\I 0 \ 2 cos 2 (oit + #), and, when this is averaged over time, with the use of (24-102), we
find the time-average rate of energy dissipation to be & = \&\I Q \ 2 = 3#I2ms in terms of
the magnitude of the peak current | / 0 | or the "root-mean-square" current I rms . Thus,
the resistance can be written in terms of power loss as
0>
2P
(28-68)
iôF
In order to apply these ideas to an electric dipole radiator, it is conventional to use

the prototype model of two charges separated by a distance ds'. Combining (28-57)
and (28-61), we find that I'ds' = I0e'lut ds' = iu>p0e~,ut,
so that |p 0 1 2 =
\IQ\2(ds')2/u>2.
Inserting this equivalence into (28-66), and using (28-68) to define the
radiation resistance, we find it to be given by
êd =
lx0o>2(ds'Y
6 7TC
ds'v
~Y
(28-69)
4 8 2
RADIATION
with the use of (28-21) and (24-95). [The value of the numerical factor in (28-69) is 790
ohms.] The radiation resistance can be taken as a measure of the efficiency of a
radiating system since we see from (28-68) that the larger the value of !%, the more
power will be radiated for a given peak current. (We recall, however, that we have
assumed that ds' < X.)
Example
Accelerated point charge. We know from (8-19) and Exercise 8-2 that a single point
charge can have a dipole moment with respect to an arbitrary origin and it is given by
p = qr'. If the charge position is oscillating in time, we will have r' = r0e~lut. In order
to use our previous results, let us assume that the charge is on the z axis so that
p qz' = qz0e~iut = p0e~'". It is also convenient here to deal with instantaneous
rather than time-average quantities. Now both (28-62) and (28-63) are proportional to
co2p0, which can be put in another form. The acceleration of the charge is a = d2z'/dt2
- co2z' so that we can write p = (qa/co2).
Then the term p0el(kr~"')
=
iu[ (r/c)]
2
p0e~ '~
= [/>]= (q/o> )[a] where both [p] and [a] represent these quantities
evaluated at the retarded time. Making these substitutions in (28-62) and (28-64), and
using (28-21), we find that we can write
ER =
q f a 1 sin 00
\
AvcqC r
BR =
q [ a 1 sin 0a>
3
47re0c3r
(28-70)
so that the instantaneous Poynting vector S = (E^ X B R ) / [ i 0

S =
q2[a\2
16 IT
sir? 0 r
. ,
e0c r
(28-71)
'
Integrating this over a sphere of radius r as in (28-66), we find the total instantaneous
rate of radiation due to this accelerated charge to be
~
q2W
6T7-0C3
(28"72)
Although these results were obtained for this special case of a harmonically
oscillating charge, their final form no longer contains this explicit assumption and, in
fact, they can be shown to be correct for a slowly moving point charge whose motion
otherwise is quite arbitrary. The commonly found expression (28-72) shows quite
clearly that a moving charge will not radiate unless it is accelerating and (28-72) gives
its total instantaneous rate of doing so.
!8-4 MAGNETIC DIPOLE RADIATION

Let us assume the magnetic dipole moment to be along the z axis so that m 0 = m0z.
Then (28-44) can be written as
A m d = ^ r ~ r 0 - ~ i k r ) e i ( k r - u ' ) sin0q>
477T
(28-73)
with the use of (1-94) and (1-92). We can now proceed exactly as before to find the
28-4
MAGNETIC DIPOLE RADIATION
483
fields from (28-4) and (28-27); the results are

B = -
k3m0
2i
:
4irc0c
(kr)2
1
+
E =
li0k2um0
kr
+
kr
cos 0t
(kr)
i
47r
(kr)'
sin0ei(kr
(kry
sin0)e'(fc,-u')
(28-74)
<P
(28-75)
(kr)~
If we now compare these with the electric dipole fields (28-53) and (28-54), we see that
the dependence on r and on 6 are the same in the two cases but that the electric field
and magnetic induction have been "interchanged" in a sense. The actual relations
among the field vectors are
m.
m0
Emd = - - ^ B e d
(28-76)
B md
E.ed
2
cp ~
PO
Put another way, the fields of one kind will be transformed into those of the other if we
make the replacements
m.
E -> cB
PO
(28-77)
[In the last two replacements, we recognize an example of the duality property of
electromagnetic fields found in Exercise 21-12 for they correspond to (21-77) with
A = 90 and C = 1. Since (21-77) was worked out for a source-free region, we were
not able to predict the appropriate replacement for p0.]
The lines of E given by (28-75) are circles about the polar axis that corresponds to
the direction of m 0 and thus E is perpendicular to the plane containing B and m 0 .
In the radiation zone, the fields are again transverse to the direction of propagation
and are
Br
= -
E =
k2m.r,
=- sin 0e'(fcr~&")8
4mnc
H 0 koim 0
477r
sin 0ei(kr
;
ot)*
<p
(28-78)
(28-79)
They also satisfy (28-64) again which is compatible with (28-77) since it becomes
E R / c = ? X B R , which leads back to (28-64) with the use of (1-30) and ? B^ = 0.
The relation between these fields is shown for a particular time in Figure 28-4. When
we compare this figure with Figure 28-2 we see that it corresponds to a rotation of
90 about the r direction as implied by (28-77).
The time-average Poynting vector and total radiated power flow are easily calculated
from (28-78) and (28-79) and it is found that
<S> = S?t =
0> =
u n co 4 lm n l 2
32v2c3r2
li0a4\m0\2
12TTC3
sin 0t
(28-80)
(28-81)
4 8 4
R A D I A T I O N
Figure 28-4. Fields at a given time in the

radiation zone of an oscillating magnetic
dipole.
which can equally well be obtained from (28-65) and (28-66) by making the replacement of p0 by m0/c. Thus magnetic dipole and electric dipole radiation are alike in
having the same proportionality to the fourth power of the frequency. They also have
the same sin 2 6 angular dependence so that Figure 28-3 applies to this case also (with p 0
replaced by m 0 ) .
Example
Current loop. In order to have a specific case, let us consider our prototype of a
magnetic dipole, that is, the small circular current loop of radius a for which the dipole
moment is ma11 as given by (19-27). If a
X, we can consider I to have the same
value at all points on the loop, and if I0 is its peak value, we have m0 = 7ra 2 I 0 . If we
substitute this into (28-81) and use (28-68), we find the radiation resistance to be
77-unco4tf4
i a \4
r U)
(28 82)
"
(The value of the numerical factor is 3.08 X 10 5 ohms.) Comparing this with (28-69),
we see that there is a difference in the frequency dependence of the radiation resistance
of our two model dipoles since 8%cA ~ a>2 while ^ m d ~ co4.
LINEAR ELECTRIC QUADRUPOLE RADIATION

The appropriate vector potential is given by (28-50). Since the situation in the radiation
zone (kr : 1) is of most interest, we go directly to it and approximate (28-50) by
u n a) 2 Q
(28
"83)
where we remember that Q is a function of the direction of propagation given by r.

Since there are such a large number of possibilities for Q a/3 and hence for Q, we will
consider only one simple but important example of electric quadrupole radiationthat
arising from a linear quadrupole.
28-5 LINEAR ELECTRIC QUADRUPOLE RADIATION
485
We previously saw that if a charge distribution has axial (rotational) symmetry

about the z axis then there is only one independent component of the quadrupole
moment tensor and the only nonvanishing values of Qaft are given by (8-39) as
Q", = Co
Qt, -Q%=
-\Qo
(28-84)
where we are using Q0 rather than the previous Qa for consistency of notation within
this chapter. A conventional prototype of such a case is the point charge distribution of
Figure 8-5b, which is repeated in Figure 28-5. The charge 2q is located at the origin
while the two charges q are on the z axis, each a distance a from the origin. It is
easily found from the point charge form of (28-48) given by (8-26) that if q = q0e' tu'
then in this case we have
Q0=-4q0a2
(28-85)
in agreement with Exercise 8-6. While the results of this section are applicable to the
charge distribution of Figure 28-5, they are not restricted to it but can be used for any
axially symmetric charge distribution described by (28-84). It is interesting to note,
however, that the charges of Figure 28-5 have neither an electric dipole moment nor
can they have a magnetic dipole moment. Thus the lowest-order radiation that they can
produce will be electric quadrupole radiation.
Combining (28-84) and (28-49), we find the components of Q to be 0R lRQRR so
that Qx = - \Q0lx, Qy= ~ iQ0ly, Qz = QJz a n d therefore
Q = 2 o ( - i ^ - 2 l ?y + 12*)
= ^ g o [ ( 3 c o s 2 0 l ) r sin 200]
(28-86)
where we have used (1-93) and (1-94) to express Q in spherical coordinates. Substitution of this Q into (28-83) yields the radiation zone vector potential for the linear
quadrupole:
Mo co
eq
Q
4877cr
3
( 3 c o s 2 0 l ) r sin20
(28-87)
The fields can now be found in the usual way from (28-4) and (28-27) while keeping
only the largest terms which are ~ 1/r since we are in the radiation zone with kr 1.
-1
Figure 28-5. Prototype charge distribution of a linear

electric quadrupole.
486
RADIATION
The results turn out to be

m n to 3 O n
* = J
sin20e'(
2
3277c r
//ioCO
Qo
Hirer
sin 2 0 e l ( k r - a t > 8
(28-88)
(28-89)
v
'
which again are transverse to the direction of propagation, are related by (28-64), and
will be described by a figure similar to Figure 28-2.
The time-average Poynting vector is found to be
1
Mo^lQnl 2
<S>
=
R
e
(
E
X
B
*
)
=
y
f
=
, ' - sin 2 20x
f
i
N
v
Rf
'
2/x0
*
204877 c r
(28-90)
'
which is proportional to the sixth power of the frequency. The angular dependence of
Sf is given by sin2 26 = 4 sin 2 $ cos 2 6 and is shown in Figure 28-6. The value of V is
zero at 6 = 0, 90, and 180, and has maxima at 6 = 45 and 135.
Integrating (28-90) over a sphere of radius r, we find the total radiated power to be
=
MoÎGOI2
'
960 77C
Figure 28-6. Magnitude of the Poynting vector as a function of angle

for linear electric quadrupole radiation.
(28-91)
'
28-6
ANTENNAS
4 3 7
ANTENNAS
U p to now we have assumed that the dimensions of our radiating system are small
compared to a free space wavelength, so that we could take the current in our simple
models to be uniform over the system. In practical cases, it is convenient to use
radiators whose dimensions are comparable to a wavelength; such a system is usually
called an antenna. In this case, the current distribution is not uniform over the length
and this must be taken into account. While antenna design is an art by itself, we can
illustrate the general principles involved by means of a few examples. However, we will
consider only examples involving electric dipole radiation.
The basic situation is illustrated in Figure 28-7. We have a straight conductor of
length 21 along the z axis. An oscillating current is produced and maintained in this
conductor usually by means of connections at its center to an external power supply.
We want to find the fields at a point P with coordinates r and 6; it is assumed that P
is far enough away so that we need deal only with the radiation zone. Since we found in
(28-61) that an oscillating dipole is equivalent to a current element, a natural approach
is to visualize the antenna as a series of current elements I'(z') dz'. Each element
produces an electric field d E and the total electric field at P will be the sum of these
because of the superposition property of electromagnetic fields. In other words, we are
dealing with an interference phenomenon.
Immediately after (28-68), we found the equivalence iup0 I0 ds', so that in this
situation we make the replacement p0 = ( / / ) / ' ( z ' ) dz' in (28-62) and use (28-21) to
obtain
dE = -
(28-92)
477/*'
where
r' = ( r 2 + z' 2 2rz' cosd)1/2
(28-93)
I'dz
z
0
Figure 28-7. Calculation

of the radiation field of a
linear antenna.
4 8 8
R A D I A T I O N
and 0' is the unit vector perpendicular to r'. The total electric field at the field point
will then be obtained by integrating (28-92) over the antenna.
We have already assumed that r X. We shall now also assume that the total
length 21 <sc r so that \z'\ : r. Under these conditions, we can approximate r' by
r' r z' cos 6
(28-94)
The amplitude of dE ~ 1/r' and we will not make much error by simply taking r' = r
here; we cannot do this in eikr', however, because even small changes in r' can make
large changes in the exponent. Finally, we neglect the differences between all of the 0'
and as well as between 6' and 6 so that, in effect, we are assuming that all of the
contributions to the total E are parallel to each other. Doing all of this, we finally get
our expression for the total electric field to be
(28-95)
Once E has been found, we can evaluate B = (r X E)/c according to (28-64) for these
fields in the radiation zone.
We can go no further until we have a definite current amplitude distribution I'(z')
along the antenna.
Example
Half-wave antenna. Clearly the current I' must be zero at the ends z' = + / of the
antenna, that is, the ends are current nodes. Stated in these terms, the problem is
reminiscent of the boundary conditions for a string rigidly fixed at both ends. One
would think then, that by proper excitation, one could obtain current distributions
analogous to the modes of oscillation of a vibrating string. We recall that the simplest
such mode is one with a displacement antinode at the center; in that case, the total
length of the string equals one half wavelength of the oscillation. By analogy, we
assume a similar situation here. Thus, 21 = \X so that / =
and we write the current
Figure 28-8. Assumed current amplitude

distribution for a half-wave antenna.
28-6
ANTENNAS
4 8 9
amplitude distribution as
I'(z')
j1TZ' \
( 2TTz' \
I0cos I
j = IQ cos I ~ I =I0coskz'
(28-96)
This distribution is illustrated in Figure 28-8. The limits of integration in (28-95) then
become + Â = ir/2k.
Such an antenna is called a half-wave antenna. It can be
obtained by using a conductor that is actually a half-wavelength long and excited at the
center or by bottom excitation of one that is one quarter-wavelength long and that is
perpendicular to a conducting plane such as the earth or a large metal sheet. In the
latter case, the image currents in the plane provide the other half of the antenna. It
turns out in practice that the existence of the radiation results in a current distribution
that is different from the simple one assumed in (28-96). The effect is small, however,
and we will not consider it further.
If we let J be the integral in (28-95), substitute (28-96) into it, change to the
dimensionless variable u = kz', and use (24-22), we find that it becomes
= IQ f"^2k e~lkz'cose
J
~ir/2k
= r / 2 e',ucose
k J -m/2
cos kz' dz'

cosudu
= f / 2
2 k J-v/2
I0 f sin [JTT(1 c o s # ) ]
+ *>] du
sin[|77-(l -I- c o s # ) ]
1 cos #
1 + cos #
2 cl
y cos (^77-cos # )
to sin2 #
(28-97)
Substituting this into (28-95), we obtain

Woplo cos ( 1 ^ cos
277T sin #
\ 2
flL(*-**)
(28-98)
'
which, interestingly, has an amplitude that is independent of the frequency. We see

from (28-97) that this occurred because the integration introduced a factor 1 /GO, which
canceled the to in (28-95). Since each element of the antenna does not radiate along its
axis, that is, for 6 = 0 or v, we expect E to vanish at these angles also. The angular
dependence found in (28-98) becomes indeterminate at these values of 6, but it is easily
verified by the use of L'Hopital's Rule that E * 0 as 8 * 0 or m.
The power flow is given by
(w S ) =
1
2/i0
| E |2
R e (v E X B * ) = -r =
'
2FI0c
aQc\IQ\2
. '
SIT R sin 8
cos 2
/ 1
\
- 7 7 c o s # r v(28-99)
!
\2
The angular distribution given by cos 2 (\TTCOS 6)/sin2 6 is shown as the solid curve of
Figure 28-9. The dashed curve shows the sin 2 6 distribution characteristic of a single
dipole. The two are qualitatively similar, although the antenna is seen to radiate even
more predominantly in the horizontal plane.
The total radiated power is
,
t v l / o l 2 (1- r c o s 2 ( J i r c o s # )
- ^ - J 0
I
smgdOdv
(28-100)
The integration over <p gives 2v. The integral over 8 cannot be expressed in terms of
elementary functions, but it can be put into the form of tabulated functions by defining
a new variable v by means of JTTCOS8 = ^(77- v) so that v = 77-(l c o s # ) . The
4 9 0
RADIATION
Figure 28-9. Radiation pattern of a half-wave antenna (solid) as

compared to that of a single dipole (dashed).
integral then becomes
cos 2 (^TTCOS 8) dO
sin 6
[2* sin 2 ( f u ) dv
-n p*
sin 2 6
IT J 0
(1
2 JQ
cos v) dv
v{2iT v)
l
1
(1 cos v)x dv l +
4 'o
\ v
2tt v
= I
riir (1 ~ cos v) dv
J0
r2w
(1 cos v) dv
JO
2TT
(28-101)
If we make the change of variable w = 2tt v in the second integral we see that it
equals the first so that (28-101) becomes
1 riw
(1
f2S
2 Jo
cos v) dv
v
1
= - 2.438)
2
as found from tables. Combining all of this, we find that (28-100) is

2.438ju 0 c|/ 0 | 2
877
(28-102)
The radiation resistance as found from (28-68) is 9$ = 2.438ju 0 c/4tt = 0.194Z o = 73.1
ohms.
The angular dependence of the radiated power shown in Figure 28-9 is independent
of the azimuthal angle <p; thus, for example, the antenna radiates equally in all
horizontal directions. It is often possible to enhance the radiation in a given direction
by utilizing the interference effects obtained by using more than one antenna. Such
combinations of antennas are called antenna arrays. The general principles used are
simple to describe but may be difficult to apply in detail.
Suppose we have a number n of half-wave antennas with their axes parallel to the z
direction. The electric field produced by any one of them at a field point P far away
from the array can be taken to be given by (28-98). Then the total electric field
'M. Abramowitz and I. A. Stegun, eds., Hanuoook of Mathematical
pp. 231 ff.
Functions, Dover, New York, 1965,
28-6
ANTENNAS
491
produced by the array will be a sum of terms of this form, that is, we will have
E = -
2tt
7 7
' cos ( 7 r c o s e J\ e ^ k r J - a t AJ
Vj sin 6j
\ 2
)
(28-103)
A
where, as in Figure 28-7, r- is the distance from the y th antenna to P, dj and 0, the
corresponding angle and unit vector, and h j the peak current of the (complex) current
distribution. One then chooses a convenient origin, usually near the center of the array,
so that the spherical coordinates of P are r, 6, and <p with respect to it. It is then
possible to express r, and 6j in terms of r, 6, <p, and the location of the 7 th antenna
with respect to the origin, much as was done to obtain (28-93). If the array is compact
enough so that its dimensions are small compared to r, then one can replace r by r in
the amplitude denominator, and take each 0y to be the same as the 0 for P, so that
A
E =
ik(r
J^-e'ikr-a,yb r~" cos ( -77cos 6,-)
J~ r )
J e
2irr
yî sm 0j
\2
)
(28-104)
We note the appearance of the term eik(r>~r) in the sum; this factor is very important in
the final result because even though r, r may be small in absolute value, k(rj r)
can have large variations so that each exponential term can be quite different in value.
The relative phasfes of the current distributions of the antennas are contained in the
amplitude terms hj because they can be written as \I0J\e'^, as was similarly done in
(28-19). We see from (28-104) that there are a large number of possibilities even for
such a special case as this. One can freely choose the number n of antennas in the
array, the relative magnitudes and phases of their currents (| Ioj\ and #,), and the
relative locations and spacings of the antennas (which will appear in the expressions for
Tj and dj). Because of the great variety of specific possibilities, we will not consider
antenna arrays in any more detail, but a few simple examples are considered in the
exercises.
So far we have only considered antennas as radiating sources. However, as every
user of a radio or television set knows, antennas are also used to detect electromagnetic
waves. (We considered this aspect briefly in Exercise 24-5 where the emf induced by a
plane wave in a small circular loop was calculated.) The question then naturally arises
as to how the receiving properties of an antenna are related to its radiative properties if,
in fact, there is any relation at all. The answer to this question is indicated by our final
topic in this chapter.
Example
A reciprocity theorem. Consider two length dsl and ds2 of two different conducting
wires as shown in Figure 28-10. Suppose the first has an oscillating current / in it; we
want to find the emf produced in the other. If we assume that they are far apart, the
electric field produced by the element I dsx can be obtained from (28-92), but it will be
helpful to write that expression in a more general form. First of all, we drop the primes
and replace I'(z') dz' by Ids^, second, (28-92) involves the assumption that the dipole
was oriented along the z axis so that we have Ids1 = Ids1 z. Now we write 8 = q> X r
= [(z X f ) / s i n 6] X f with the use of (1-92) and (1-94), so that Idsl sin 00 = I(ds1 X f)
X r. The electric field dE produced by the current element Ids{ at ds2 can thus be
written as
/unwf/l
dE =
{d%l X r) X f
(28-105)
4 77f
where [ / ] is the current evaluated at the retarded time. Now the tangential components
of this electric field are continuous across the surface of the wire by (21-26), so that the
4 9 2
RADIATION
Id s,
Figure 28-10.
Calculation of a reciprocity theorem.
emf induced in d s 2 as given by (17-7) is found to be

mnr/i r
dS=
dE
ds2
-- 1f(<3?Si X f ) X f1 ds2
4irr
ia nco I /1
= - [ ( r X dsx)
(r X ds2)]
(28-106)
with the use of (1-29) and (1-23). This result is completely symmetric in ds1 and ds2
and is unaltered if the sign of f is changed. This means that if the element d s 2 were
carrying the current I, exactly this same emf would be induced in the element dsv A
result of this type is called a reciprocity theorem.
What (28-106) tells us, then, is that the value of the induced emf or "received signal"
does not depend on whether the element ds is being employed as a radiator or as a
receiver. When extended to the whole antenna, this reciprocity theorem shows that the
antenna characteristics, such as angular dependence, are the same whether the antenna
radiates or absorbs radiation. For example, since a half-wave antenna does not radiate
in the direction of its axis (8 = 0) as we found from (28-98), an electromagnetic wave
traveling parallel to its axis will not induce an emf in it either. In this special case, it is
clear that this will be so since the incident electric field is transverse to the direction of
propagation and will be normal to the conductor forming the antenna. Thus, the
electric field will have no tangential component along the antenna length and cannot
produce an induced emf.
EXERCISESl
28-1 An infinitely long thin straight wire coincides with the z axis. It carries no current for
t < 0. At t = 0, a current I is applied over the
whole length of the wire, and it remains at this
constant value for all later times. At a point a
distance p from the wire find A, B, E, and S for
all times. Show that your results reduce to sensible values as t - oo. (Hint: recall Section 16-4.)
28-2 A system of charges and currents occupies
a fixed finite volume V. Show that in general
dp
r
(28-53) and (28-54). Then find S and show that it

has oscillatory components along r and
Finally
show that your expression for S reduces correctly
to (28-65).
28-6 Show that the fields in the radiation zone
due to an electric dipole can be written in the
form
"
dt2
- ~7,
where p is the total electric dipole moment of the

distribution.
28-3 Find the amplitude of the electric field at a
point in the equatorial plane that is 10 kilometers
from an electric dipole radiating one kilowatt.
28-4 Find the fraction of the energy radiated by
an electric dipole within 10 of the equatorial
plane.
28-5 Assume that p0 is real and find the general
physical fields E and B for an electric dipole from
E =
Mo
d2p
4wt
dt2
Mo
4wcr
d2 p
X r) X r
X fr
where the quantity in brackets is evaluated at the

retarded time.
28-7 Show that the electric dipole fields in the
radiation zone can be written in the form
Mo dr'
( ds' X r) X r
4 mr dt
dr
B,, = Mo
ds' X r
4m cr ~dt
EXERCISES
where the derivative of the current is evaluated at

the retarded time.
28-8 Show that the magnetic dipole fields in the
radiation zone can be written in the form
E =
B,
2
Mo d m
477cr dt2
Mo
4 irc2r
d2 m
dt2
X r
Xr
Xr
where [ d 2 m / d t 2 ] is evaluated at the retarded

time.
28-9 Show that all of the fields in the radiation
zone whether due to an electric dipole, magnetic
dipole, or linear electric quadrupole can be expressed in terms of the corresponding vector
potential as follows:
dA
dA
B* =
E* =
X r) X r
X r
~dt
~dt
where [DA/DT] is evaluated at the retarded time.
28-10 Find E and B in the near zone for the
linear electric quadrupole and compare your result with (8-55).
28-11 Show that the radiation fields of a slowly
moving point charge can be written in terms of
the retarded value [a] of the acceleration as:
E* =
<?[]
g{[a] X r) X r
47T0c2r
"r
4<7re0c3r
28-12 A commonly used approximate method of

deriving the fields in the radiation zone of a linear
electric quadrupole is based on the model illustrated in Figure 28-5. This can be interpreted
as two oppositely directed dipoles of amplitude
T q0a located at z = \a, respectively, and the
total field E# can be written as a superposition of
fields obtained from (28-62). Do this, and then by
using the approximation a c r, show that the
first-order result gives (28-89).
28-13 A simple model for calculating the fields
in the radiation zone from a magnetic quadrupole
can be devised by analogy with the previous
exercise and Exercise 19-16. Consider two magnetic dipoles of equal dipole moments m0z
located at z = b\ the total magnetic dipole
moment of this system is zero. Such a situation
could be obtained with two small current loops,
each parallel to the xy plane, and with oppositely
directed currents. The total induction B s can then
be written as a superposition of fields obtained
4 9 3
from (28-78). Do this, and then by using the

approximation b
r, find the fields BR (and
hence E w ) as the first-order result.
28-14 An antenna for which the total length
21 = X is called a full-wave antenna. Show that
two possible current amplitude distributions consistent with this are / ' = 70 sin kz' and / ' =
I0 sin k\z'\. Show that these are consistent with
the model of a full-wave antenna as equivalent to
two colinear half-wave antennas spaced a halfwavelength apart and excited with relative phases
of tT and 0, respectively. Find E and (S) for the
first case and thus show that the angular distribution is proportional to [sin (w cos 0)/sin 0 ]2. Make
a rough sketch of this angular distribution.
28-15 Two half-wave antennas lie in the yz
plane. They are parallel to the z axis and their
centers lie on the y axis at y = a and y = a.
Assume that a
r, and find the resultant electric field in the xy plane. Assume the currents to
have equal amplitude. Now assume that they are
a half-wavelength apart, and consider the two
cases in which their currents (1) are in phase and
(2) have a phase difference of it. Sketch the
magnitude of the electric field as a function of <p
for these last two cases.
28-16 If the source has both an electric dipole
moment and a magnetic dipole moment, the total
electric field in the radiation zone will be the sum
of (28-62) and (28-79) and we have the possibility
of interference. Show, however, that (S) = (S) ed
4- (S) m d , that is, the total radiated power does
not contain any interference terms. Will this same
result be true if it has both an electric dipole
moment and a linear electric quadrupole moment?
28-17 The Hertz vector ire was introduced in
Exercise 22-7. The Hertz vector corresponding to
a point electric dipole is asserted to be tte =
i(kr 6)t)
/4itt0r. Verify that this gives the corPo*
rect vector potential for an electric dipole.
28-18 Consider a system of charged particles for
which the ratio of charge to mass,
is the
same for each particle. Show that the total dipole
moment can be written as p =
R cm
where M is the total mass of the system and Rcm
is the position vector of the center of mass. Show
that, if there is no net external force on the
system, it will not produce electric dipole radiation. Similarly, show that, if there is no net external torque on this system, it will not produce
magnetic dipole radiation either.
SPECIAL RELATIVITY
The basic concern of the theory of relativity is the comparison of results obtained by
observers of physical phenomena who are moving with respect to each other. The
special theory, which was propounded by Einstein in 1905 and that is the only form we
will consider, deals with observers who have a constant velocity with respect to each
other. The term "relativity" refers to the assumed equivalence of these observers and
their respective coordinate systems for the description of phenomena. As we will see,
the origins of special relativity can be found in the structure of electromagnetic theory
so that we can usefully begin by considering how moving observers describe some
electromagnetic effects.
2 9 - 1 HISTORICAL ORIGINS OF SPECIAL RELATIVITY
When we consider, for example, the electric field vector E(r, /), we mean the value of
this quantity at a point r and a time t that are specified with respect to some particular
set of coordinate axes. The observer who is determining E at r and t is assumed to be at
rest with respect to these axes. In the paragraph following (1-14), we briefly alluded to
the question of what coordinate system was to be used and we concluded that it was
one of the inertial systems of mechanics. As we also know from mechanics, a
coordinate frame fixed to the surface of the earth is, for most purposes, a satisfactory
approximation to an inertial system.
In the discussion of Faraday's law for moving media in Section 17-3, however, we
did encounter the question of how observers moving with respect to each other would
describe the same phenomena and we saw that in general their descriptions would be
different. In particular, we found (17-29), which stated that
E' = E + v X B
(29-1)
was the way in which the fields in the two systems were related; then Faraday's law
V X E = dB/dt had a form that was independent of the motion of the medium. We
saw that in some cases it was easier to discuss a problem by going into the moving
system and dealing with the "motional" electric field E^ = v X B. What is of more
interest to us here, however, is that (29-1) shows that these two relatively moving
observers describe the same phenomena in different ways. As an extreme example, if
E = 0, then what appears to one observer as a purely magnetic induction B is seen by
the other as an electric field E'. Thus an electric field is not an absolute property of a
given point and time but also depends on the motion of the observer.
In order to obtain (29-1), we used the fact that F' = F where the forces being
compared are measured in two inertial systems, that is, they have no acceleration with
respect to each other and with respect to the primary inertial system. Let us remind
ourselves of how this situation comes about.
In Figure 29-1, we show two coordinate systems S and S' with rectangular axes
chosen to be parallel to each other for simplicity. The relative position of their origins is
given by R 0 = V/ so that their origins coincide at t 0 and V is the constant velocity
of S' with respect to S. The position vectors of a point P are r and r', so that
r' = r R 0 r Vt
(29-2)
494
29-1
H I S T O R I C A L O R I G I N S O F SPECIAL RELATIVITY
495
S'
0'
Figure 29-1. Two rectangular coordinate systems with parallel axes

of coordinates.
The velocities of the point P as measured in the two systems are therefore related by
the familiar result
V'
dr'
dt
dr
d R0
^ = v - V
dt
dt
(29-3)
Since V is constant, the accelerations are equal, that is, a' = dv'/dt = dv/dt = a.
Therefore, the forces on a point mass m located at P are also equal: F' = ma' = ma =
F. Thus, the basic law of mechanics has the same form in these two coordinate systems
moving with constant velocity with respect to each other. In other words, they are
equally suitable for use as a reference system. This result is known as the classical (or
Galilean) principle of relativity. The formula (29-2) that relates the two sets of
coordinates and time (t' = t) for a given "event" is known as the classical (or Galilean)
transformation. While (29-2) is perfectly general, it is more convenient to specialize it to
the case in which the relative motion is along the common xx' direction so that
V = Fx; then (29-2) can be written in component form as
x' = x Vt
y' = y
z' z
t' = t
(29-4)
where, for completeness, we have included the usual assumption that t' t. In this
case, (29-3) becomes simply
K = vx - V
v'y = vy
v'z = vz
(29-5)
Since (29-1) indicates that the expression for the field changes depending on the
coordinate system one is using, it is natural to wonder how a transformation of
coordinates like (29-4) will affect the equations determining the fields, that is, Maxwell's
equations. For our purposes, it will be sufficient to consider a consequence of Maxwell's
equations, that is, the wave equation
2
V 4>
1 d2i>
2 ~J7z
c at
(29'6)
4 %
S P E C I A L RELATIVITY
which is satisfied in a vacuum for any component of E or B. We know that (29-6) has
plane wave solutions with speed c. We will show that the form of this equation is not
preserved by the Galilean transformation (29-4) but that plane wave solutions are still
possible.
For simplicity, let us restrict ourselves to the case in which the direction of
propagation of the wave is along the x axis so that (29-6) reduces to
d24>
1 82xp
dx2
c2 dt2
(29-7)
Now let us assume that \p is written as a function of x' and t' by means of (29-4) so
that
dip
dx
2
and therefore d \(//dx
d\f/ dx'
d\p dt'
d\p
dx' dx
dt' dx
dx'
d \p/dx' .
d\p
dt
2
d i>
2L
dt

dip
dx'
d\}/
dt'
d \p
d2^
d2xp
y2Z_2 - 2V
1
dx'
dx' dt'
dt'2
When these results are substituted into (29-7), we obtain

(c
v
d2\(/
d2\p
d2\p
V )=
= j
2V
' dx'2
dt'2
dx' dt'
(29-8)
'
which is clearly different in form from the equation (29-7) satisfied in the unprimed
axes. Let us nevertheless try to find a plane wave solution of (29-8) of the form
t') = xpoeik'^'-u'n
(29-9)
where u' is the phase velocity. Substituting (29-9) into (29-8), we find that it is a
solution provided that u' = c V = ( c + V). In other words, if the wave is
traveling in the positive x' direction, it has the speed c V, while its speed is c + V
for propagation in the negative x' direction. These results are therefore consistent with
the particular case given by (29-5), and for a general direction of propagation, the wave
velocity in the primed system can be found from (29-3).
This calculation has shown us that if (29-4) and (29-7) are simultaneously valid then
electromagnetic effects will generally be different when observed from different coordinate systems moving with constant velocity with respect to each other and, in
particular, the speed of a plane wave in vacuum would not always have the value c.
Since (29-6) is a consequence of Maxwell's equations, these results imply that there can
be only one frame of reference in which Maxwell's equations have the form in which we
have been writing them and in which electromagnetic waves have the speed c. This
preferred reference system was generally identified with the primary inertial system of
mechanics, that is, one fixed with respect to the " fixed stars." This frame also had to be
endowed with electromagnetic properties so that it could propagate waves; this
augmented system was then called the ether.
Electromagnetism and the Galilean transformation taken together thus imply the
existence of only one system of reference in which Maxwell's equations are valid. On
the other hand, we have also seen that the whole idea of a preferred system is foreign to
mechanics because of the existence of the Galilean principle of relativity. However,
because the concept of equivalence of different coordinate systems has such a simplicity
29-1
H I S T O R I C A L O R I G I N S O F SPECIAL RELATIVITY
4 9 7
and attractiveness, we are led to consider yet another possibility, namely that there
exists a common principle of relativity valid for both mechanics and electromagnetism.
If this were to be the case though, then the Galilean principle of relativity based on the
Galilean transformation and Newtonian mechanics can not actually be the appropriate
one. (We are not considering the possibility that Maxwell's equations may need
alteration.)
The proper choice among these possibilities can only be made with the aid of the
results of experiment. Historically, one approach was to assume the validity of both
Maxwell's equations and the Galilean transformation, that is, the existence of a
preferred reference frame. Then a laboratory frame attached to the earth would be
moving with respect to the primary inertial system (ether), so that one could look for
effects associated with this motion and that could be calculated by transforming
Maxwell's equations into this system much as we did to obtain (29-8). We will consider
two of these experimentsone that does not involve wave propagation and one that
does.
Example
The Trouton-Noble experiment. In this experiment, a charged parallel plate capacitor
was suspended so that its plates were vertical and it could rotate about an axis parallel
to and between the plates. Because of the assumed motion of the earth with respect to
the ether, there will be a torque on the capacitor that tends to align the plane of the
plates parallel to their velocity V. We will simplify things somewhat by replacing the
charges on the plates by two point charges whose separation is the same as that of
the plates as shown in Figure 29-2. The magnetic force Fm on the positive charge due to
the negative charge is found from (14-31) to be
F
" "
li0q2\
X (V X R)
4 ^
<29-10)
and there will be an equal and opposite force on the negative charge. These two forces
will produce a torque T on the system given by
T
X F
" =
li0q2{\
R)(V X R)
4^5
(29
-U)
which has the direction shown in the figure for 6 > 90. (The sign of r is reversed for
6 < 90.) We see that T is parallel to the axis and tends to rotate the charges until the
v
Figure 29-2. The Trouton-Noble
experiment.
498
line R connecting them is perpendicular to V. The magnitude of the torque is

(29-12)
where Ue = q2/4ire0R
is the magnitude of the electrostatic energy of the system by
(5-49). We note that the effect measured by T is proportional to (V/c)2, which turns
out to be characteristic of these experiments. (The result of taking the actual distribution of the charges on the plates into account is to double the value given above for r.
Furthermore, the attractive electrostatic force is always along the direction of R and
hence does not contribute to the torque.)
Taking V to be the speed of the earth in its orbit, which is about 3 X 10 4
meters/second, one finds that the magnitude of the torque is large enough to be
observable. However, the most careful experiments failed to find such a torque, so that
this experiment provides no evidence in favor of a preferred reference frame for
electromagnetism.
Example
The Michelson-Morley experiment. This experiment was based on the result following
(29-9), which showed, in effect, that the velocity combination formula (29-3) applied to
electromagnetic waves as well as to point masses. If V is interpreted as the velocity of
S' with respect to the ether, then (29-3) says that the velocity of the wave will depend
on the motion of the medium and will be c only with respect to the ether. Because the
orbital speed of the earth is so small compared to c, it has not been feasible to make a
direct measurement of the wave velocity in various directions with respect to the earth's
surface in order to check (29-3). It is possible, however, to compare these directions and
look for this small effect by using the wave itself in a particular manner; this idea had
been suggested by Maxwell, and the experiment was first performed by Michelson and
Morley in 1887.
The experimental arrangement is illustrated in Figure 29-3. Light from the source L
is incident on the partially silvered glass plate M and divided into two beams that
travel in directions that are perpendicular to each other. Each beam travels the distance
/ to the mirror Ml or M2 and is reflected back over its original path. Part of the light
Figure 29-3. The Michelson-Morley

experiment.
29-2
THE POSTULATES A N D THE LORENTZ T R A N S F O R M A T I O N
4 9 9
from each beam then passes along the fourth arm of the device toward a detector D.
The beams have thus been recombined and any phase difference between them
resulting from their trips back and forth will produce interference effects observable as
a variable amplitude in the resultant beam. In order to calculate the phase difference,
we need to find the times taken by the beams to travel along their respective paths.
If we let the arm 1 be in the direction of motion of the apparatus, then, from before,
the speed of the light will be c V on the outward path and c + V on the return path.
Thus the time taken to pass to and fro along this arm is
I
211
V2\~l
21 (
= 7 (
1 +
V2\
^ )
(29 13
>
In the perpendicular direction along the second arm, the resultant relative light velocity
as found from (29-3) is (c 2 F 2 ) 1 / 2 , and the corresponding time for the round trip is
21
2It
V2 \
= t [
2 ? )
<29"14>
Thus the two times are not equal and their difference is
'
/ ( V\2
7(7)
<29 15)
"
Taking I = 30 meters and V = 3 X 10 4 meters/second as before, we find that At

10""15 second. This small time difference corresponds to a phase difference of 03At =
2mcAt/\
= 0.6(2t7) = 3.8 radians for visible light with a wavelength of 0.5 X 10~ 6
meters. This is a sizable relative phase difference and could result in detectable changes
as the apparatus is rotated to interchange the role of the two arms. However, no such
effects were observed by Michelson and Morley. Since then this experiment and
variations on it have been frequently repeated, particularly as more sensitive and
accurate equipment has become available. The results have always been the samethat
for electromagnetic waves there is no evidence to support the use of (29-3), which
results from combining the concepts of a preferred reference frame for electromagnetism with the Galilean relativity principle for mechanics.
While we have discussed only two experiments of this general type, there have not as
yet been any results of any kind that support a preferred frame for electromagnetism.
Accordingly, we are led to assume that there is a common principle of relativity for
both electromagnetism and mechanics. It cannot, however, be described by the Galilean
transformation given by (29-2) and (29-4). The necessary new concepts are provided by
the postulates of special relativity.
One should not get the idea, however, that the whole of special relativity stands or
falls on the results of a single "crucial" experiment, in particular, the Michelson-Morley
experiment. By now, special relativity has been extended and tested by means of such a
large number of interwoven experiments that it provides us with a single coherent and
useful description of a great variety of phenomena.
2 9 - 2 THE POSTULATES AND THE LORENTZ TRANSFORMATION
There is some doubt as to whether Einstein was strongly influenced by, or was even
aware of, the results of the experiments we have just described. There is no doubt,
however, that he thought deeply and hard about the questions we outlined in the last
section. The result was that in 1905 he proposed the equivalent of the two following
500
SPECIAL RELATIVITY
postulates:
I
All systems of coordinates are equally suitable for the description of physical
phenomena.
II
The speed of light in vacuum is the same for all observers and is independent of the
motion of the source.
The first postulate literally refers to all systems of coordinates, although we restrict
ourselves to those systems that have a constant velocity with respect to each other.
When we do this, we are discussing the special (restricted) theory of relativity. An
unrestricted discussion would correspond to the general theory of relativity, which we
do not consider.
The first postulate also asserts that there is a common principle of relativity for all of
physics, although we consider it only in terms of particle mechanics and electromagnetism. Another way of stating the first postulate is to say that no theory can contain
any reference whatsoever to an absolute velocity of translational motion of the coordinate system being used. We saw in the last section that Maxwell's equations combined
with the Galilean transformation did indicate a possibility of effects due to an absolute
velocity, but such predictions were not borne out by experiment. This could mean that
Maxwell's equations are themselves not completely correct, or it could mean that the
Galilean transformation is not correct. Einstein preferred the second possibility but
rather than postulating that Maxwell's equations were covariant (unchanged in form) to
the proper transformation of coordinates, he found it to be sufficient (and equivalent)
to state his second postulate in terms of the speed of light.
It is this second postulate that leads to the more unfamiliar results of special
relativity, including those that apparently violate "common sense." It also implies that
Newtonian mechanics is not the correct form for the exact representation of mechanical
phenomena, since the Galilean transformation representing the relativity principle
natural to mechanics is incompatible with the second postulate as we found after
(29-9). Thus we first have to find the transformation rules for the coordinates that will
satisfy the second postulate. Then we will have to find such laws of mechanics as will
satisfy the first postulate when these transformation rules are used. Finally, we will then
have to see how electromagnetism as described by Maxwell's equations fits with these
postulates. Whether these postulates are actually applicable to the description of
physical phenomena can only then be appraised by comparing predictions of experimental results with the actual results themselves.
Before we begin these tasks, however, it will be well to see that these postulates are
compatible with the results of the two experiments discussed in the last section. As far
as the Trouton-Noble experiment is concerned, we can invoke the first postulate to
allow us to calculate the effect in a coordinate system in which the charges are at rest.
Here the only force is the attractive electrostatic force and, as previously pointed out,
this will lead' to no torque and hence no rotation of the capacitor in agreement with
observation. In the Michelson-Morley experiment, the speed of the waves will be c for
both arms, making tx = t2 and thence At = 0, again in agreement with observation.
The transformation of coordinates, which is quantitatively in accord with the second
postulate, is called the Lorentz transformation. We consider two coordinate systems S
and S' in relative motion with velocity V along their common xx' direction. For a
given point event, an observer in S will assign the spatial coordinates and time
(jc, y, z,t), while those obtained by an observer in S' will be (x', y', z', t'). Thus our
transformation equations must enable us to calculate the position (x', y', z') and time
t' of the event, if it is known that the first observer assigns the values (x, y, z) and t to
29-2
THE P O S T U L A T E S A N D THE LORENTZ T R A N S F O R M A T I O N
501
it, and conversely. Our derivation will be somewhat simplified, but it will lead to the
correct result and will include all of the principal features of a more elaborate
treatment.
Suppose that, at the instant the two origins of the coordinate systems coincide, a
small pulse of light is produced at the origin. The second postulate requires that each
observer see the light propagating with the same speed c in all directions. In other
words, with respect to his or her own coordinate system, each observer must see the
wave front as a sphere centered on the corresponding origin of radius c times the time.
The equations of the wavefronts are thus
x2 + y2 + z2 c2t2 = 0
x'2 + y'2 + z'2 c2t'2 = 0
(29-16)
so that we must satisfy the identity

x'2 + y'2 + z' 2 - c2t'2 = x2 + y2 + z2 - c2t2
(29-17)
A n equivalent way of saying this is that the quantity on either side of (29-17) must be
invariant with respect to the transformation from one system to the other.
Since we have taken the x and x' axes in the direction of the relative velocity of S
and S', it is reasonable to assume that the coordinates transverse to the motion are
unchanged, that is, y' = y and z' = z; then (29-17) becomes
x'2 c2t'2 x2 c2t2
(29-18)
N o w there is only one relative velocity V. From the point of view of S, the origin 0'
must have the JC coordinate Vt as shown in Figure 29-4a. Similarly, the JC' coordinate of
0 as seen in S' must be Vt' as shown in ( b ) of the figure. Therefore, the
transformation equations must satisfy the conditions that
x = Vt
when x' = 0
(29-19)
x' = Vt'
when x = 0
We will try the simplest relations that will satisfy (29-19), that is, the linear equations
a:' = y ( * ~ Vt)
x y'(x'
+ Vt')
(29-20)
where y and y' are constants to be determined. We can now express t' in terms of the
unprimed quantities by eliminating x' from (29-20) and we find that
Putting these expressions for x' and t' into (29-18) and grouping terms, we get
+ 2 xt y2V -
yc2
i -
yy
(29-22)
S'
Vt
S'
Vf
0'
0'
x'
M
Figure 29-4.
observers.
(b)
Coordinates of the two origins as seen by the two relatively moving
5 0 2
SPECIAL RELATIVITY
In order that this will always be zero for all possible values of x and t, it is necessary
that the coefficients of x2, xt, and t2 individually vanish. Equating each of these
coefficients to zero, we find that
1
Y' = Y = 7
|l
T777
(29-23)
will satisfy (29-22).

Combining all of these results, we get the Lorentz transformation formulas
x' = y(x Vt)
x = y(x' + Vt')
y'y
z' = z.
y = y'
t'=
z = z'
yit
t = y\t'
j-*)
(29-24)
V
jx' ]
c'
(29-25)
where y is given by (29-23). The equations (29-25) can be obtained from (29-24) by
solving for the unprimed quantities in terms of the primed ones, or, equivalently, by
interchanging primed and unprimed symbols and changing the sign of V. as vyjpuld be
expected from the overall symmetry of the situation.
*
If we let V/c 0, so that y -* 1, we see that (29-24) reduces to the Galilean
transformation (29-4) that we now see as a first approximation to the Lorentz
transformation for finite values of V such that V/c
1.
It will be convenient to introduce a special symbol for the ratio of V to c
V
ft =
c
(29-26)
so that
(29-27)
(1 - fi2)1/2
while (29-24) becomes
x' y ( x
fict)
y'y
z' = z
t' = y^t
(29-28)
Actually, any linear transformation of coordinates and time that satisfies (29-17) is
called a Lorentz transformation. The one given in (29-28) is that particular one that
applies only to the case in which the relative motion of S and S' is along their common
x direction. In the next section, we consider more general Lorentz transformations but
for now we restrict ourselves to (29-28).
The first physical results implied by the Lorentz transformation that we will
investigate are all basically kinematic ones. The existence of the first three effects we
consider can be deduced qualitatively from the postulates alone, but for our purposes it
will be more efficient to calculate them by direct use of the Lorentz transformation.
Example
Relativity of simultaneity. Suppose that two events occur at the points x1 and x2 in the
S system and are simultaneous there so that t1 = t2. The times at which these events
are observed in S' as obtained from (29-28) are
=
<
(29-29)
29-2 THE POSTULATES A N D THE LORENTZ T R A N S F O R M A T I O N
5 0 3
so that the time interval between these events is

At' = t'2-t[
yfi
= - (*2 ~
* 0
(29-30)
This result shows that two events that occur at different points
and x2, and that
are simultaneous for an observer at rest in S, will no longer appear to be simultaneous
to an observer who is at rest in S'. In other words, simultaneity is not an absolute
property of a pair of events but is also dependent on the relative state of motion of the
observer.
Example
Time dilation. Suppose that we have a clock located at the point x x and that it is
emitting signals of some sort at regular intervals At, where At = t2 t1 = t3 t2, and
so on. The corresponding times in S' are given by (29-29), except that now x2 = x1
since the clock is fixed in S. Therefore, when seen from the system that is moving
relative to the clock, we find that these signals will be separated by the time intervals
At' = f2 t{ = t'3 t'2, and so on, given by
At' = yAt > At
(29-31)
since y > 1. Thus the time interval appears to be longer to the moving observer than it
does to the one at rest with respect to the clock.
Example
The Lorentz contraction. In principle, we measure a length by placing a measuring rod
along the distance to be measured and then finding the difference between the scale
marks that simultaneously coincide with the ends of the length of interest. Although
such a detailed specification appears to be trivial when the measuring rod and the
length are relatively at rest, it is essential when they are moving with respect to each
other.
Let L = x2 xl be the distance between two points as found with a scale that is at
rest with respect to them. The length as found by the moving observer who assigns
coordinates x'2 and x{ to the points is found from (29-28) to be
L' = x'2 x[ = y[x2
xl fic(t2 ^ ) ]
(29-32)
However, the values x'2 and x[ must have been determined simultaneously in S' so
that 12 = t[. This means that the time interval t2 tx in (29-32) must be given by
t2 tx (/3/c)(x2 x x ) = (yS/c)L according to (29-29). When this is substituted into
(29-32) and (29-27) is used, we find that
L' = y ( l - p2)L
= ^ = (1 - /32)1/2L
< L
(29-33)
Thus, as found by the moving observer, the length in the direction of motion will be
contracted by the factor 1 / y .
Example
Velocity addition formulas. The rectangular components of the velocity v' of a point as
observed in 5" are given by
dx'
dy'
dz'
504
We now want to find the components vx, vy, vz of the velocity of this same point when
its motion is referred to the system S.
Differentiating the first expression in (29-25) and the last in (29-28), and using
(29-26), we obtain
dx
l dx' dt'
dt' \
(29
dt'
- -
j
f$vx\
( l - - )
-35)
(29-36)
When these are combined so as to eliminate dt'/dt,
we find that
(
fivx
*-r2(;+ y) i -
which can be solved for vx; the result is that
Vx
'
O Q
1 + (Vv'x/c2)
We can now express (29-36) completely in terms of primed quantities by substituting

(29-37) into it and we get
~di
y[l + (Vv'x/c2)]
(29 38)
"

Vy
dy
dy' dt'
v'y
dt
dt' dt
v [ l + (Vv'x/c2)]
v* = -T
y[l
v'z
71
5TT
+ (Vv'Jc2)]
.
.
(29-40)
These last two relations, together with (29-37), enable us to convert velocity components observed in one system to those observed in another, and are the relativistic
analogues to the classical formulas (29-5) that are based on the Galilean transformation.
We see that these formulas are more complicated than the classical ones and, in
particular, v'x is involved in the formulas (29-39) and (29-40) used to convert y and z
components of the velocities. On the other hand, if /$ = V/c
1, we see that (29-37),
(29-39), and (29-40) become approximately vx = v'x 4- V, vy v'y, and vz v'z, which
are just (29-5) again. Thus, the simple addition of components that follows from the
Galilean transformation is an approximation appropriate to small relative velocities of
the two coordinate systems.
The relativistic velocity transformation formulas have the interesting property that
the sum of two velocities can never exceed c. This is most easily illustrated with an
extreme example. Suppose that V = c, while v'x = c also. The Galilean transformation
would yield vx = vx + V = c + c = 2c, while the Einstein formula (29-37) becomes
vx = (c + c)/[ 1 + (c2/c2)] = c as was asserted.
Example
The Doppler effect and aberration. The question of interest here is how the frequencies
and directions of propagation of a wave are related for the two relatively moving
29-2
THE P O S T U L A T E S A N D THE L O R E N T Z T R A N S F O R M A T I O N
5 0 5
observers. Suppose that a light source Q', at rest in S', emits a spherical wave. We
know from the results of the last chapter that this wave will be represented in S' by
y = ^Vw'-cy,')
(29_41)
where r' is the distance from Q', \p'0 is a constant amplitude, and the circular
frequency to' and propagation constant k' must be related by
to' = k'c
(29-42)
since the wave has speed c.

N o w suppose that there is an observer at rest in S at a point P in the xy plane. The
coordinates of P are (*, y) in S, and (x', y') in S'. The S observer will interpret the
wave as a spherical wave originating from a point source Q in that system, so that it
will be described by
= ^yei{kr-^
(29-43)
where
to = kc
(29-44)
since the wave also has speed c in S because of the second postulate.
A s shown in Figure 29-5, the direction of propagation from Q' to P can be
described by the angle 9' made with the x' axis, while 9 is the corresponding angle in
S. We also see from the figure that
r' = x' cos 6' + y' sin Q'
r = x cos 6 + y sin 6
(29-45)
N o w the numerical value of the phase of the wave (zero, for example) must be
independent of whatever system is being used to express it: thus the exponents in
(29-41) and (29-43) must be always equal. Furthermore, the coordinates and times must
be related by the Lorentz transformation (29-28). Equating these exponents, substituting from (29-42), (29-44), (29-45), and (29-28), we find that we must have
CO
x cos 6 + v sin 9
t\ to'
y(jc Pet) cos 9' + y sin 9'
I fix
t
(29-46)
This equality must hold for all possible values of x, y, and t; this is possible only if
their coefficients are separately equal.
Equating the coefficients of t on both sides of (29-46), we find that
to = io'y( 1 + /3cos9')
P (x, y)
(29-47)
5 0 6
which relates the frequencies and is the relativistic Doppler effect formula. We see that
if 0 c 1, then y = 1, and (29-47) reduces to the classical formula
to = to'(l + (3cos0')
(29-48)
Similarly, when the coefficients of x are equated, we find that co cos 6 = to'y(cos 0'
+ ft), which can be used to eliminate the frequencies from (29-47) with the result that
cos 0' + ft
cos 0 =
(29-49)
1 + ft cos 0
relates the directions of propagation and is called the aberration formula.
Finally, by equating the coefficients of y in (29-46), we obtain co sin 6 = to'sin#',
which is easily shown to be consistent with (29-47) and (29-49).
The formula (29-47) differs from the classical acoustical result (29-48) in that it
predicts a transverse effect. Suppose that the frequency is observed in S at right angles
to the motion, that is, 0 = 90. Then cos 0 = 0 and (29-49) gives cos 0' = ft. If this is
substituted into (29-47) and (29-27) is used, we find that to = co'(l ft2)l/2 < to'. This
predicted effect has been observed by studying the radiation from an electric discharge
in hydrogen and provides some experimental evidence for the basic correctness of the
relativity postulates.
In order to see the motivation for what we will do in the next section, it is useful to
look at some of our results from another point of view. The Lorentz transformation
applies to coordinate differentials too, and we find from (29-28) that
dx' = y (dx
ftcdt)
dy' = dy
dz' = dz
dt' = y^dt dx j
(29-50)
In addition, (29-17) shows us that the Lorentz transformation corresponds to the

equality
(dx)2
+ (dy)2
+ (dz)2
c2(dt)2
= (dx')2
+ (dy')2
+ (dz')2
c2(dt')2
(29-51)
Dividing out (dt) and (dt') , taking the square root, and using (29-34) and
(1-6), we find that the quantity dr given by
/
I
v 2 \ 1/2
v ,2 \ 1/2
dr = dt 1-^2
I
= d t ' \ ( 2 9 - 5 2 )
^
dz
dr
[i - (2a2)]1/2
vy
[l - (v2/c2)]1/2
il
"
^1
has the same value for all reference frames related by the Lorentz transformation. This
invariant quantity dr is called the proper time interval and we see from (29-52) that
dt = dt0 = the time interval measured in the system in which the point is at rest
(v = 0).
If we now divide both sides of each term in (29-50) by dr, we obtain
dt \
dy'
dy
dx' _ J d x
/3c
dr )
dr
dr
dr
I dr
(29-53)
dz
dz'
' dt
dt'
fi dx \
J
dr
dr
dr
dr
Using (29-52), (29-34), and comparing (29-53) with (29-28), we see that the four
quantities
dx
vx
dy
vy
^ [i - (2A2)]1/2
dt
dr
1
[l - (v2/c2)]1/2
(29 54)
29-3
GENERAL LORENTZ T R A N S F O R M A T I O N S , 4-VECTORS, A N D TENSORS
507
transform in exactly the same way as do jc, y, z, t. Therefore, since we know from
(29-25) that
x = y(x'
+ Vt')
t = y[t'
+ (V/c2)x']
we can immediately write the transformation equations for the analogous quantities
dx/dt
and dt/dr as
x
2
[i - ( V c ) ]
1
2
I/2
"
[l - (v /c )]
1/2
"
\ [1 - ( ' V c ) ]
/
1
2
_
2
,/2 +
1
[l -
[1 - ( v V c
V
1/2
(v' /c )]
) W
v'x
[l -
(^2
/ c
2)]V2
/
(29
"
55)
(29-56)
If we now divide (29-55) by (29-56), we obtain

4 +
""
1 + (Vvyc2)
which is exactly the velocity transformation formula (29-37). You should be sure to
verify that (29-39) and (29-40) can be obtained in a similar way.
Our use of (29-55) and (29-56) is quite different from the previous method of
obtaining (29-37), which involved direct differentiations of the transformation equations and subsequent elimination of terms. We were able to obtain (29-37) in this latter
manner because we knew how all of the quantities involved were affected by a Lorentz
transformation, that is, we knew their transformation properties.
The whole point of this example, then, is to show us that we should be able to write
the transformation formulas for certain quantities quite easily, provided that we know
something about their general properties. This turns out to be an extremely valuable
and efficient way of looking at things and is what we consider in the next section.
29-3 GENERAL LORENTZ TRANSFORMATIONS, 4-VECTORS, A N D TENSORS

Let us begin by considering the two rectangular coordinate systems shown in Figure
29-6. They have a common origin but differ by a rotation; that is, if we rotate one set of
axes as a rigid body, it can be made to coincide with the other. The exact rotation could
be described, for example, by specifying the direction cosines of the primed axes with
respect to the unprimed axes, or vice versa.
Figure 29-6. The two sets of axes are rotated with

respect to each other.
508
Consider a point P that has the coordinates x, y, z in one set of axes while its
coordinates are x', y', z' in the other set. Both of these different sets of numbers locate
the same point P. The distance r of P from the origin has the same numerical value
regardless of which coordinate system we are using. In other words, r is an invariant
and we must have
r2 = x2 + y2 + z2 = x'2 + y'2 + z'2
(29-57)
It will be convenient to introduce the notation

= x
x2 y
x3 = z
(29-58)
for then (29-57) can be written somewhat more compactly as

r2 = Y, x j = H xj2
y=i
7=1
(29-59)
The equations that relate the two sets of coordinates will be linear so that we can
write the three equations
x
j = E ajkxk
k-1
( j = 1,2,3)
(29-60)
that is, the three equations

x[ = a\\xl + a12x2 + allX3
*2 = a21xl + a22x2 + a23X3
(29-61)
*3 = a3lxl + a32x2 + CI 33 X j
The set of nine numbers ajk characterize the rotation relating the primed and unprimed
axes. It is also evident that these ajk cannot all be independent because the transformation equations (29-60) have not yet been made to satisfy the fundamental physical
requirement that the axes are related by a rotation, that is, that r2 be an invariant as
described by (29-59).
In this particular three-dimensional case, it is easy to identify the ajk. The position
vector of P in the primed system can be written as r' = x ' x ' + y'y' + z'z' by (1-11),
while for the unprimed system it is r = x x + yy + zz. Since they represent the same
physical vector (that is, displacement from the origin), we must also have r' = r. Then
we can use (1-21) to find JC' as
.x' = r' x' = r x' = x x ' x + >>x' y + z x ' z = lxxx + lxyy
+ lxzz
(29-62)
where lxx, lxy, lxz are the direction cosines of x' with respect to the x, y, z axes,
respectively, according to (1-8). [These direction cosines satisfy lxx + lxy + lx2 = 1 by
(1-9).] Hence by comparing (29-62) with (29-61), we see that an = lxx, a12 = lxy, and
<a13 = l x z . In the same way, we will find that the a2k are the direction cosines of y', and
the a3k are those of z'.
In Section 1-1, we defined a vector as any quantity with the same mathematical
properties as the displacement of a point. Now the position vector r certainly is such a
quantity and its components are exactly the coordinates of a point. Thus (29-60)
describes the effect of a rotation on the components of this particular vector. Therefore,
we can immediately say that if Ax, A2, and A3 are the rectangular components of any
vector A, then they will satisfy a relation like (29-60), that is,
^,-Y."jkAk
0 = 1,2,3)
(29-63)
k =\
where the ajk are the same set of coefficients that describe the effect of a rigid rotation
of axes on the coordinates of a point.
29-3
GENERAL LORENTZ T R A N S F O R M A T I O N S , 4-VECTORS, A N D TENSORS
5 0 9
While we could continue with this three-dimensional case, it is just as easy to carry
out a further discussion for a more general situation to which we now turn.
If we look back at (29-17) we see that this basic condition for the Lorentz
transformation can be expressed as requiring the invariance of the quantity s2 given by
s2 = x2 + y2 4- z2 c2t2 = x'2 + y'2 4- z'2 c2t'2
(29-64)
If we now introduce the notation

x1 = x
x2 = y
x3 = z
x4 = ict
(29-65)
we see that (29-64) can be written

j2
= E xl = x'2
p.=l
l
(29-66)
On comparing (29-66) and (29-59), we see that the most general Lorentz transformation
can be interpreted as a rigid rotation of axes in a four dimensional space with axes
jcj, x2, x3, x4 = ict. This interesting, although somewhat formal, result has many
far-reaching and useful consequences, as we will see.
We can write the transformation equations relating the two sets of coordinates as the
four general linear equations
4
K = E arvXv
v=l
(n = 1 , 2 , 3 , 4 )
(29-67)
If the transformation were not linear, it would give a preferred status to whatever origin
we happened to choose, or to some other point; but this would violate the first
postulate since it would provide us with an objective means of distinguishing one
coordinate system from another.
In order that (29-67) represent a Lorentz transformation, (29-66) must also be
satisfied. This will impose some requirements on the 16 coefficients a^v and we want to
find out what they are. Substituting (29-67) into (29-66), we get
I>A2 = E
X
E a\p.xp j ( E a\vx,)
>V P
X \ N
= E ( E ax/iax J x^x,
'
(29-68)
'
Since (29-68) must also equal H^x2, we see that we must have
E ^ A ^ X , = 8n*
(29-69)
x
with the use of (8-27) since the coefficient of x^x^ must equal 1 if v = ju. and 0 if v = ju,.
Now we could equally well have written our transformation in the form
4
= E Kxx'x
x=i
{v = 1 , 2 , 3 , 4 )
(29-70)
where the bvX are an appropriate set of coefficients. It is evident that the bvX must be
related to the a and we can determine this by substituting (29-70) into (29-67) to give
=
=
v
E ^ j E v ^ )
\
(29-71)
\
with the use of (8-27) again. Comparing the last two terms, we see that
EÂx =
(29-72)
We can solve (29-72) for the b's by multiplying both sides of it by a^, summing over ja,
510
SPECIAL RELATIVITY
and using (29-69):

= HbvX8pv
= z>
E fip,a..,Ax
!ivuv\
tiV
ft
Thus, bpX = aXp, so that we can say that
if x'^ =
then
= bpX
(29-73)
(29-74)
Furthermore, (29-72) can then be written entirely in terms of the a's by substituting the
result of (29-73) into it, and interchanging the indices X and v to give
= 5.
tip
(29-75)
Thus, there is only one set of coefficients describing the transformation and (29-69) and
(29-75) represent the conditions required by (29-66).
When the particular Lorentz transformation (29-28) that we have been using is
written in the notation of (29-65), it becomes
X
Y*1 +
x'j = x3
yx4
2 ~ X2
x'4 = yx4 i/3yxl
(29-76)
Upon comparing (29-76) with (29-67), we see that the coefficients a^ representing this
particular transformation can be written as the matrix
FLI>
a li
a 12
a 21
a 22
a 31
'32
'41
'42
a 13
14
a 23
a 24
a
a
33
34
4 3
44
Y
0
0
1
0
ifi Y
0
-i/By
(29-77)
It will be left as exercises to verify that the a u given in (29-77) satisfy the requirements
(29-69) and (29-75).
It is evident that a rigid rotation of axes in three-dimensional space will also keep the
expression x2 + y2 + z2 c2t2 invariant because of (29-57) and its independence of
the time. Thus such a physical rotation of axes should also be included in the group of
general Lorentz transformations described by (29-67).
At this stage, it will be useful to introduce some new quantities, mostly by analogy
with the three-dimensional case. We remind ourselves that an invariant is a quantity
whose numerical value does not change as the result of a Lorentz transformation.
Examples are j 2 of (29-66) and the proper time dr of (29-52).
As a generalization of what we found to hold for a vector A in (29-63), we define a
4-vector
as a set of four quantities ( AL, A2, A3, A4) whose transformation properties
are the same as those of the coordinates. That is, if the Lorentz transformation is
described by (29-74), then the components AM and A^ are related by
4
A
'a = E
4
a
*v A v
and
* = E avvp
..A'
(29-78)
where the coefficients a^v are exactly those appearing in (29-67) and ji = 1, 2, 3,4. Since
we saw that a rigid rotation of axes is also a Lorentz transformation, it follows from
this definition and (29-63) that the first three components (Alf A2, A3) of A must be
the components of an ordinary three-dimensional vector A.
The sum of two 4-vectors must be a 4-vector so that
C = ( A + B ) ,
in analogy with (1-10).
f-
(29-79)
29-3 GENERAL LORENTZ T R A N S F O R M A T I O N S , 4-VECTORS, A N D TENSORS
511
We can define a scalar product of two 4-vectors by generalizing the result obtained
for A B in (1-20):
Scalar product =
(29-80)
After reviewing the three-dimensional definition (1-15), we suspect that this scalar
product is an invariant. This is easily proven by using (29-80), (29-78), and (29-69):
L<b1
=
=
E w
* fl
=
'
v\
From (1-17), we are led to define the square of the "length" of a 4-vector as the
scalar product of the 4-vector with itself; thus
("length") 2 = Y.AI
f-
(29-81)
In addition to the coordinates, we have already encountered a 4-vector; it is the

4-velocity
whose components as found from (29-54) and (29-65) are
dx i
Ul
dr
vx
[i - (v2/c2)]1/2
dx-,
v.
U, = 1 =
, .
2 2 1/2
d-r
[l - (v /c )]
dx2
Ul
dr
vy
[i - (v2/c2)]1/2
dxA
UA =
ic
dr
(29-82)
[i - ( y / c 2 ) j i / 2
where v2 = v2 + v2 + v2. These four quantities form a 4-vector because we showed in

(29-55) and (29-56) that their transformation properties are the same as those of the
coordinates.
Inserting (29-82) into (29-81), we obtain the interesting result that the square of the
"length" of the 4-velocity is negative:
("length") 2 = U2 =
7,2
4 .
][
)2
_^
2 /
2 )
= -c^
(29-83)
A second-rank tensor T is defined as a set of 16 quantities such that each index

transforms in the same way as do the coordinate indices, that is, if (29-67) holds, then
T;, = Z , Z w , p T x
X
(29-84)
Two indices are needed to specify each component; hence the term "second rank."
Similarly, a 4-vector with its single index is seen to be a tensor of the first rank, while
an invariant scalar is a tensor of zero rank. It is possible to define tensors of higher rank
by a straightforward generalization of (29-84) but we will have no need of them.
A tensor is said to be symmetric if T^v = TPfl and therefore it has only 10
independent components. A tensor is antisymmetric if Tllfl = T^v and has only 6
independent components since the diagonal elements must be zero, that is 7 ^ = 0. A
tensor does not have to be symmetric or antisymmetric, but if it is, this symmetry
property is not changed by a Lorentz transformation; that is, if T^v is symmetric (or
antisymmetric), Tv is also symmetric (or antisymmetric). We can prove this for both
cases at once by writing T^v= TVfl and using the upper sign for the symmetric case
512
and the lower for the antisymmetric case. Using this in (29-84), we find that
T
v = Y.a\"v P T \ P =
(29-85)
Xp
Xp
If we now interchange /x and v in (29-84) and then interchange X and p, we obtain

^
= IX\^prXp =
Xp
'avpatlXTpX
(29-86)
Xp
Comparing (29-86) with (29-85), we see that T^v =

with the choice of signs being
the same as originally used, thus showing that the symmetry property has not been
changed.
An important example is
Fp, = A^B, - AvBtB
(29-87)
While we have written these 16 numbers as F^, this does not by itself prove that
is
actually a tensor, although if it is, it will be antisymmetric. In order to show that F
does indeed have the correct transformation properties, we substitute (29-78) into
(29-87) and we find that
=
'B'v
A B
= EanX<t*pA\Bp
', l
E^,papXApBx
Xp
= E^xavp(AxBp
Xp
- ApBx)
= E^^-p^P
Xp
(29-88)
Xp
which is exactly what is required by the definition (29-84). We also see that if
H, v = 1 , 2 , 3 then the components of F^ are exactly those of the usual three-dimensional cross product A x B as shown by (1-27). Accordingly, the cross product is seen
to be actually a tensor rather than a vector.
In a similar manner, one can show that the following quantities are 4-vectors:
B, = EAJVV.
c = E TÂ,
(29-89)
V
V
According to the chain rule for differentiating a function of given unprimed
variables with respect to different primed variables, one has
d
dx
dxv
dx
'p.
(29-90)
dx
as we have already used to obtain (29-8) from (29-7). We see from (29-74) that
dxydx'p a ^ , so that (29-90) becomes
d
7- E
8x'r
r
<29"91)
"dx
On comparing (29-91) with (29-78), we see that the four operators d/dx
transform
exactly like a 4-vector. This fact enables us to define a four-dimensional del operator
with the components
=
dxx ' dx2 ' dx3 ' dx
(29-92)
4
and use it in a fashion analogous to the three-dimensional operator v as written in

(1-41). [We note that this result justifies the statement following (1-41) that V has the
same properties as the displacement of a point, that is, it has vector characteristics.]
29-3 GENERAL LORENTZ T R A N S F O R M A T I O N S , 4-VECTORS, A N D T E N S O R S
513
If xp is an invariant, we can define a Gradient, //, as the 4-vector whose

components are dxp/dx^ much as in (1-37). Similarly, we can define
v- d A

Divergence:
(an invariant)
(29-93)
Four-dimensional Laplacian:
Curl:
d2
= E T~2
p. dxv
3AV
C =
*
dx^
2
c
dA
d2
2T1
dt
( a n invariant)
(an antisymmetric tensor)
dx
(29-94)
(29-95)
Divergence of a tensor:
X)
dT
(a 4-vector)
(29-96)
We note that the four-dimensional Laplacian is actually the vacuum d'Alembertian

operator of (22-17). Furthermore, if fi, v = 4 in (29-95), then we see from (1-43) that
the corresponding components of C ^ are the components of the three-dimensional curl,
V X A.
While all of this material in this section is very interesting, it is only natural to
wonder what is the point of discussing it. If we ask ourselves what the first postulate
means here, we see that it says, in effect, that there should be no way of making an
absolute distinction between two systems moving with constant velocity with respect to
each other. In addition, we found from the second postulate that observations made in
the two systems must be correlated by means of the Lorentz transformation. Combining these two results, we will see immediately that we can say that the two postulates
taken together require that the laws of physics when properly formulated must have
their form unchanged when subjected to a Lorentz transformation, that is, they must be
covariant with respect to the transformation. In order to show that this must be so, let
us see what the consequences would be if this were not so in a very simple case.
Suppose that a particular law under consideration had the general form & =
+ J(f.
Let us suppose also that when we referred everything to the primed system by means of
a Lorentz transformation, it became J5"' =
+ 3tf" 4- c' where e' depended on the
particular primed system involved. This equation is clearly not covariant and, in fact,
the very existence of the c' term would enable us to distinguish among the various
systems in an absolute way. Since this would violate the first postulate, we can see the
reason for requiring covariance with respect to a Lorentz transformation.
We have just seen that 4-vectors and second-rank tensors are covariant by their very
definition. Thus it is evident that, if we were to express all of our physical laws in
4-vector or tensor form, they would automatically be covariant with respect to a
Lorentz transformation and would thereby satisfy both the postulates. With this in
mind, we can see, in essence, what we will do next. We will look at some physical laws
and see if they can be written in terms of 4-vectors or tensors. If they already are or
easily can be, we need do nothing more because we know that they are already valid in
special relativity. If the laws are not covariant, yet are correct in the nonrelativistic case,
then our task is to try to generalize these results so that they can be expressed in
4-vector or tensor form and thus be compatible with special relativity. In this latter
situation, however, we will have to bear two points in mind. First of all, our
generalizations will always have to reduce to the known valid results in their nonrelativistic limit. Second, our generalizations will still need to be tested by experiment because
514
the process of generalizing to 4-vector or tensor form does not necessarily lead to a
unique result.
PARTICLE MECHANICS
While electromagnetism is our primary interest, the mechanics of mass points is so

important in physics and provides such a good illustration of what we just discussed at
the end of the last section, that we will briefly consider it in this section.
The basic equation of the nonrelativistic mechanics of a mass point subject to the net
force f is
dp
1 = (29-97)
where
p = m0v
(29-98)
is the linear momentum of the particle in terms of its velocity v and its inertial (or rest)
mass m0. Equation 29-97 is clearly not in 4-vector or tensor form because it has only
three components and the differentiation is with respect to the time which we know is
not an invariant scalar. Therefore these equations must be generalized in order to make
them satisfy the requirements of special relativity.
We begin with the momentum. While v is not a 4-vector, it is closely related to the
4-velocity U = dx^/dr. Hence a plausible generalization of (29-98) is to use the
invariant scalar m 0 and define the 4-momentum
as
P , = m0U^
(29-99)
We can write this in component form by using (29-82) and we get

m nV:
im nc
< y = l i 2
'
3 )
" h - ( ^ ) r
(29
"100)
We see that as v/c - 0, P, > m0Vj, which is (29-98) so that (29-99) appears to be a
reasonable choice. For the moment, we disregard the extra component P4 that we have
introduced in this way.
In order to get an equation of motion analogous to (29-97), we differentiate (29-99)
with respect to the invariant dT and define the 4-force
by
dPn
d2xtl
d ,
(29
"101)
to obtain the desired generalization of the Newtonian equation; the 4-vector F is also
called the Minkowski force.
In order to relate (29-101) to the ordinary force components, we can substitute for
dr from (29-52) and we get
2X1/2
- J,(meft)
(29-102)
and the first three components of this as found with the use of (29-82) are
v2\1/2_d_(
2
m0Uj
dt \ [! - (v2/c2)}1/2
'
= JJ
/,
(29-103)
where the f} must be the x, y, z components of the ordinary force f since (29-103) has
29-4 PARTICLE M E C H A N I C S
515
to reduce to (29-97) and (29-98) combined when v/c <: 1. Therefore the first three
components of the Minkowski force are related to the ordinary force by
A
[i - ( A ) ]
J = 7:
'
7-T7TTTT72
2
2
,/2
(29-104)
In practical calculations, one usually does not want to deal with the 4-force but
prefers to use just the three equations (29-103); thus it is common practice to write
d (
mnv
)
<29-105'
' - * { I T T T ^ j p }
and to call this the relativistic equation of motion. Furthermore, if one still wishes to
regard force as the rate of change of momentum, (29-105) can be written as f = dp/dt
where
p = my
and
ra0
m =
7777
(29-106)
[1 -c A 2 )] 1 / 2
The quantity m introduced in this way is then called the mass of the particle because of
the analogy with (29-98). If this procedure is followed, (29-106) provides the basis for
the common statement that the mass of a particle is no longer a constant but increases
as the speed increases.
On the other hand, it is not at all necessary to interpret these results in this way, and
doing so follows only from a natural desire to write momentum as always being the
product of the mass and the ordinary velocity v rather than as the rest mass m 0 times a
new function of the velocity. In fact, such an approach actually contradicts the basic
philosophy of the covariant approach of relativity because (29-106) is not in 4-vector
form. It is much more consonant with relativistic concepts to ascribe an invariant scalar
propertythe rest mass m 0 t o the particle and then define the 4-vector momentum as
the product of this scalar invariant with the 4-vector velocity, exactly as we did in
(29-99).
We have seen that our process of generalization led from the three component
equation (29-98) to (29-99), which has four components. We saw that the first three
components of P^ can be given an adequate interpretation and now we must look at the
"extra" one, which turns out to be not so unfamiliar after all.
Let us define the quantity W by
iW
P4 =
(29-107)
c
W =
m C
.1/2
2 1/2
[1 - (v /c )]
(29-108)
In order to interpret this, we see what it becomes when v/c <sc 1. Expanding the
denominator with the use of (8-6), we find that
W=
/
m0c II +
2
1)2
\
+ . . . I = m0c
1
+ ~m0v
+ ...
The second term can be recognized immediately for it is simply the kinetic
particle in Newtonian mechanics. Accordingly, it seems quite reasonable
general case to call W the total energy of the particle. We see that, if the
rest so that v = 0, the value of W is m0c2, which is called the rest
(29-109)
energy of the
in this more
particle is at
energy. It is
516
SPECIAL RELATIVITY
customary then to regard the total energy of the particle as being composed of two
partsan intrinsic part due to its rest mass (the rest energy) and the additional part,
which appears when the particle is moving. Thus, if we let T be the kinetic energy, we
can write
W = m0c2 + T
(29-110)
so that
r
^ ( [ i - ( ^
r -
( 2 9
"
m )
according to (29-108).
The fourth component of the 4-momentum thus has turned out to be simply
proportional to the particle energy. This shows that the linear momentum and energy
of a particle should not be regarded as different entities, but simply as two aspects of
the same attributes of the particle since they appear as separate components of the
same 4-vector.
This last remark can be illustrated quantitatively. Since P^ is a 4-vector it transforms
according to (29-78), and if we use the second form of this for the particular Lorentz
transformation of (29-77), we obtain
p x =y[p'
y
\
Pz = p;
+ W'\
c
I
W = y(W'
Py = P'
y
(29-112)
+ ficP^)
clearly showing that what appears as energy in one system appears as momentum in
another and, conversely, what appears as momentum in one system is energy in
another.
Example
Particle at rest in S'. Here v' = 0 so that P'x = P'y = P'z = 0 and W' = m 0 c 2 by
(29-100) and (29-108). Inserting these into (29-112) and using (29-26) and (29-27), we
find that
WlnV
WnC2
Px = 7
TT77
W = 7
T777
(29-113)
K
J
*
[1 - (V2/c2)]1/2
[l - (V2/c2)]1/2
and Py = Pz = 0. These results are exactly what we should expect by (29-100) and
(29-108) since, from the point of view of the observer in S, the particle is moving along
the x axis with speed V. Thus the particle has only (rest) energy in S', but has both
energy and momentum with respect to S.
m
We can now see the significance of the fourth component of the Minkowski force.
We find from (29-101), (29-107), and (29-52) that
dPA
F
' " 17
i dW
" 717
i
= c[i
- (Vc2)],/2
dW
(29
~n4)
and is proportional to the time rate of change of the energy, or to the rate at which the
force is doing work on the particle.
This can be seen in quite another way that provides further justification for the
interpretation of W as the energy. We find from (29-99) and (29-83) that
E^
V L u ^ = ~\c2
(29-115)
29-4
PARTICLE M E C H A N I C S
517
and, on differentiating this with respect to T and using (29-101), we obtain

dP.,
E^-rr - V i
a T
= 0
(29-116)
[Since the scalar product of P^ and

is zero, we can say that the 4-momentum and
4-force are always "perpendicular" by analogy with (1-15).] If we write out (29-116) in
detail and use (29-100), (29-104), and (29-114), we obtain
Jl
P,F
'
m0v f
~P
"
I
=
im0c
'F - ~
- (v2/c2)
\ I
\ [1 - (v2/c2)]1/2
/ i c [ l - (v2/c2)]1/2
dW\
I
so that
dW
= v f
(29-117)
This result shows explicitly that the time rate of change of W equals the rate at which
the force does work on the particle and, since this is the definition of rate of increase of
energy in mechanics, it is just what we should expect to find from our interpretation of
W.
The energy can also be expressed in terms of the linear momentum by writing out
(29-115) and using (29-107). The result is that
W2 = c2 Pj2 + (m0c2)2
7=1
= (Pc) 2 + (m0c2)2
(29-118)
which is a convenient starting point for the development of the Hamiltonian formulation of relativistic mechanics.
Other important aspects of mechanics that we have not yet discussed explicitly are
the conservation laws of momentum and energy. Since we have seen that it is no longer
adequate to consider momentum and energy as separate entities, it would seem that the
natural relativistic generalization would simply be the conservation of the 4-momentum. In fact, this is exactly what has been found to be correct experimentally, and, in
addition, this generalized conservation law holds for a system of particles, even when
the number of particles and their rest masses are different in the initial and final states.
The concept of 4-momentum conservation is particularly useful in the study of
collisions. In quantitative form this conservation law can be written as the four
equations
E P?(j) -
j= 1
P;(k)
k1
- 1,2,3,4)
where P*(j) is the jutth component of the 4-momentum of the y'th particle before the
collision (or general interaction); similarly the superscript a on the right labels the
values after the collision. This equation also provides for the number of particles to
change from N to N'. The fact that the sum of P4 is also conserved shows that the rest
energies and kinetic energies need not be individually conserved, although their sum
must be. In other words, rest mass and kinetic energy can be converted into each other.
This conservation law has been well verified experimentally, particularly in reactions
involving atomic nuclei and collisions of high-energy particles of various kinds. The
excellent agreement with experiment provides additional evidence for the basic correctness of special relativity.
518
SPECIAL RELATIVITY
c
29-5| ELECTROMAGNETISM IN V A C U U M
In contrast to mechanics, we will see that electromagnetism as described by Maxwell's

equations for a vacuum is already covariant with respect to Lorentz transformations.
We did not require this directly, although the second postulate did deal with the
invariance of one of the consequences of Maxwell's equations.
In principle we know from (29-91) how the differential operators in Maxwell's
equations transform, but we would also like the transformation properties of p, J, E, B,
A, and cj>.
We begin with the equation of continuity (12-13), which expresses the fundamental
property of conservation of charge. We certainly want this to be covariant, that is, we
want
dp
dp'
V - J + = 0 and v ' J ' + = 0
(29-119)
dt
dt'
If we introduce four quantities J^ by means of the respective equalities given by
(Jl3 J2, J-i, J4) = ( j x , Jy, jz, icp)
(29-120)
and use (29-65), we find that (29-119) can be written as

= 0
and
- ^ = 0
(29-121)
H
1* M
Comparison of this with (29-93) shows that it has the form of the divergence of a
4-vector, making us suspect that J' is itself actually a 4-vector.
Let us consider a volume element dY~ in a coordinate system S in which the charges
have a velocity v; the total charge in dV is p dY. Now let us consider a coordinate
system S0 in which the charges are at rest so that v0 = 0; this system is called the rest
system. In the volume element dY0 of S0, which corresponds to dY of S, the total
charge is p0dY0
where p 0 is the charge density in the rest system. It seems quite
unreasonable to expect that the basic elementary unit of chargethat on an electron or
a protonwill be changed merely because we observe it in another coordinate system.
Then since all charge is an integral multiple of this unit, as we mentioned following
(12-1), the determination of the total charge is essentially one of counting to find a
definite integer. Since an integer is an invariant, we conclude that it is reasonable to
assume that total charge is an invariant. This gives us
p0 dY0 = pdY
(29-122)
In this case, the relative velocity of the systems S and S0 is v. Since the dimensions
along the relative velocity are contracted according to (29-33) while the dimensions
transverse to the relative motion are unaffected, the two volumes are related by
dY=
/
v2\1/2
i l - - j
dY0
(29-123)
Combining this with (29-122), we obtain the transformation formula for charge
densities:
rPo
f)
P ~ T.
, 2 2
(29-124)
[l - {v /c )\
In (12-3) we found the current and charge densities to be related by J = pv. Using
this, along with (29-124) and (29-82), we find that
Po^x
/x
Pvx
r/ t2 , 2o\ll/2
P(A
[1 - ( " A ) ]
C
j
^
29-5
E L E C T R O M A C N E T I S M IN V A C U U M
519
Similarly, we find that Jy = p0U2, Jz = p0U3, icp = p0U4, so that (29-120) can be
written as
- PUf
(29-125)
This shows that J' is actually a 4-vector since it is the product of another 4-vector (the
4-velocity) and the scalar invariant p 0 (rest system charge density). This 4-vector is
called the 4-current. Thus the equation of continuity (29-121) has been correctly written
in covariant form.
The transformation properties of J^ are given by (29-78), or, to put it another way,
we see that Jx, J , J2, p transform like x, y, z, t, respectively. Therefore, for the
particular Lorentz transformation (29-24), we can immediately say that
j; = y ( j x - v p )
j; = Jy
(
V
pf = y^p--^Jxj
j; = jz
(29-126)
while the inverse equations

jx = y ( j ; + v p ' )
p = y|p' + ^ 7 ; J
(29-127)
follow from (29-25).
Example
Convection current. Suppose that S' is fixed within a material body that moves with
constant speed v relative to S. Then we can use (29-127) if we replace V by v. If we
now go to the nonrelativistic limit where v/c <s: 1 so that y = 1, then (29-127) gives
jx =
+ Vp'
p = p'
(29-128)
While the charge density remains constant, the current density observed in S is the sum
of Jx observed in S' and the convection current vp' due to the motion of the charge
density p' with respect to S\ we have already mentioned convection currents arising in
just this way at the end of Section 12-2.
If we define the four quantities
by
(A,, A2, A3, A4) = Âx, Ay, Az, y J
(29-129)
then, with the use of (29-94) and (29-120), the four equations (28-1) and (28-2) can be
combined into
^=
-/xo^
(29-130)
since
a2(i<f>/c) = ip/ce0
= -J4/c2e0=
-p0 J4
The Lorentz condition (28-3) then becomes simply

dA
L -T-t = 0
*
d x
(29-131)
Since (29-130) shows that operating upon A with an invariant yields a 4-vector, we see
that A is a 4-vector; it is called the 4-potential.
520
SPECIAL RELATIVITY
We see from (29-129) that Ax, Ay, Az, <j> transform like x, y, z, c21, respectively.
Therefore, for the particular Lorentz transformation (29-24), we can immediately say
that
A
'x
V
71*
y\Ax~
A
'z = z
A'y=Ay
$ = v(4> ~
(29-132)
VA
X)
Equations 29-130 and 29-131 are the covariant forms of Maxwell's equations as
written in terms of the potentials. Now we can turn to the fields E and B.
Since we have been obtaining B from B = v X A, it seems that it would be useful to
investigate the four-dimensional version of this as given by (29-95). Accordingly, we
define the antisymmetric electromagnetic field tensor
by
f =
J \IV
dA
dA.,
dx,.
dx,.
(29-133)
Then we find, for example, with the use of (29-129) that

/l2
dA-
dA1
dA,
dx
dx7
dx
dy
Br
Similarly,
/l4
dA,
dAl
i d<j>
1 dA
dxx
dxA
c dx
ic
iE
dt
since E = v<#> ( d A / d t ) . Continuing in this way, we find that E and B appear in

the components of f as follows:
0
Bz
-By
~BZ
Bx
f/iv =
~BX
_ i
(29-134)
r N1
r *1
0
c
J
Since we know how the components of a tensor transform, we can find the transformation properties of the field components; we will return to this shortly.
It can then be shown that all of Maxwell's equations as written in terms of the fields
are contained in the following covariant system of equations:
dfxP
dx
dfpv
dx,
dfvX
dx
dLfiv
dx
V0Ju
(29-135)
(29-136)
In order to see how this goes, let us consider the first component of (29-135), that is, the
form in which the index one does not appear; with the use of (29-134), we find that it is
dx-i
df42
dx7
df-23
dx*
= 0 = - -
i dE
1 dBr
h + +
dy
c dz
ic dt
29-5
E L E C T R O M A C N E T I S M IN V A C U U M
521
which can be rearranged as

dEz
dEy
dBx
dy
dz
dt
which is the x component of v X E = <9B/dt. Similarly, one can show that the
remaining three components of (29-135) give the y and z components of Faraday's law
as well as V B = 0.
If we set ju. = 1 in (29-136), and use (29-134) and (29-120), we get
dxx
dx2
-111 =
j
dx4
^0 1
dx3
dy
dz
ic dt \
which can also be written as

dBz
dB
1 dEx
which is the x component of V X B = [x0J -I- c~2(dE/dt),

that is, the vacuum form of
(21-33). The remaining two components of this equation are obtained for ju. = 2 and
[i3, while when [X = 4 is used in (29-136) we find as needed that the result is the
remaining Maxwell equation of the vacuum version of (21-30) through (21-33), that is,
V E = p/c0.
According to (29-84), the components of the electromagnetic field tensor will
transform as
/; = ZWrpfxp
(29-137)
\p
and we can use this in conjunction with (29-134) to obtain the transformation formulas
for E and B. We will do this only for the particular Lorentz transformation described
by (29-77).
As an example, we consider the 14 component of (29-137). Remembering that
fP\ = - A P . w e obtain
iE'
fU
Ylal\
Xp
4pf\p
l \ (
= a 1 1 ( a 4 1 / 1 1 + 0 4 4 / ^ ) + ^14(^41/41
= (ano44 - al4a41)fl4
= / l 4 =
"
4 l A l
^ 44-/a4 )
^44/44)
= (y2 - (32y2)fl4
iE
and therefore E'x = Ex. Similarly, we find that the 42 component of (29-137) leads to
iE'
/a'2 '
=
c
= a4J12
so that E' = y(E
IZ,aAXa2pfxp
\p
Yia4xa22fx2
\
+ <144/42 = (-ifiy)B,
ficBz) y(E
iE
+ yl
VBZ). By continuing this process, we find the
522
complete set of transformation formulas to be:

EL = E
B' =
E; = y(Ey-VBz)
B; = y[By
EL = y(Ez + VBy)
+ ~ - )
(29-138)
VEy_
B'z = y( B
c2
The inverse transformations are

EX = E'X
Ey = y(E;+
VB'Z), etc.
and are obtained by interchanging primed and unprimed quantities and changing the
sign of V.
Now that we have these results, we can see that we need no longer be restricted to
the case in which the relative velocity V of S ' with respect to 5 is along the x axis.
Since the orientation of the x axis is completely arbitrary, we can introduce the field
components parallel (||) and perpendicular ( ) to the direction of relative translation,
and we can write (29-138) in the general form:
E' =
B'
E
X BX)
x = y(E
r v ^x + V* ~
= y(E + V x B )
- B
BX =
,,
Y(Bx
- ^
X Ex)
c2
(29-139)
=y^B-^-xeJ
where we have E = E M + E x and B = B M + B x .

We note that in the nonrelativistic limit, V/c <s: 1, y = 1 and (29-139) becomes
E' = E + V X B and B' = B. The first result is exactly (29-1) with v now appropriately
replaced by V and that we previously deduced by considering Faraday's law with the
help of the Lorentz force law.
These results show us clearly that the electric and magnetic field vectors E and B
actually are not independent and separate entities. The fundamental quantity is the
field tensor /
and the way in which it is resolved into electric and magnetic
components is determined by the relative motion of the observer. This can be
illustrated very well by considering two extreme examples.
Example
Pure electric case in S. Suppose that E = 0, but that B = 0. Then we find from (29-139)
that in S' one will have
E
I I = E
B ; =
E'
B
y E
JV
so that
B' = B'X =
V X E'
c
VXE'
(29-140)
since V X E'M = 0. Thus what appears to be purely an electric field to one observer is
seen as both an electric and a magnetic field to a second observer moving with respect
to the
first.
29-6
FIELDS O F A U N I F O R M L Y M O V I N G P O I N T C H A R G E
523
Example
Pure magnetic case in S. N o w suppose that E = 0 while B = 0. Then we find f r o m

(29-139) that in S ' one will have
B'n = B n
B'x=yB
E' n = 0
E' = yV
so that
E ' = E ' x = V X B'x = V X B'
(29-141)
and what appears to be a purely magnetic field for one observer will appear to be b o t h
an electric and a magnetic field to a relatively moving observer.
T h e transformation equations (29-138) or (29-139) sometimes make it easier to solve

certain problems because one can choose a coordinate system in which the answer can
be very easily found and then the desired results can be obtained by transforming back
to the actual system of interest. One will not get any results this way that could not be
obtained by solving Maxwell's equations, but when such a method is applicable it is
generally easier and faster. In the next section, we illustrate this approach with an
instructive and important example.
29-6 FIELDS OF A UNIFORMLY M O V I N G POINT CHARGE
We consider a point charge q moving with constant velocity v with respect to the
system 5 as shown in Figure 29-7. W e are interested in the fields associated with this
steady-state situation, that is, we will not consider the effects of the initial acceleration
of the charge during which we know f r o m (28-70) that it will radiate.
A good choice for S' should be the system in which q is at rest at the origin for then
the field in S' is just the electrostatic Coulomb field of a point charge and we have
qr'
E' =
t
B' = 0
(29-142)
47TtQr'*
where r' is the position vector of the point where the field is to be evaluated and
r>
= jc' 2 + y > 2 + z ' 2
(29-143)
Inserting (29-142) into (29-138), and using (29-143), (29-24), and (29-23), with v
replacing V, we find one of the field components in S to be
qx'
qy(x
4ire0r'3
s
S'
0'
~ vt)
4tt0 [ y 2 ( j c vt)2 + y2 + z 2 ]
3/2
(29-144)
X, X
Figure 29-7. The point charge q is moving

with constant velocity with respect to S.
524
where now
r
<29
u - ( U r
-145)
It will be convenient to express this in terms of the position of q with respect to S. The
S coordinates of q are (vt, 0,0), so that the relative position vector of the field point
with respect to the charge location is R = (x vt)x 4- _yy + zz. Then we can also write
(29-144) as
E
qyR>
^
^V2
4t7C0(Y2R2 + R2y + R2Z)~
(29-146)
The other two components of E can be found from the appropriate inverse forms of
(29-138) and the results are
E
y =
. =
4ire0(y2R2x
qyR y
h
^17?
+ R2y + R2)3/2
qyR
r:
"77377
4ttc 0 (Y 2 j R 2 + R2y + R2)
(29-147)
(29-148)
We see that these three results are the components of the single vector equation
E
"
/"TT7
4m0(v2*i + * I
777377
l)
(29-149)
The value of B could also be found from (29-138), but it is simpler just to use the
result (29-140) obtained for the pure electric case. Thus, when we take the interchange
of S and S' into account, we obtain
v x E
z
(29-150)
c
from which we could find the explicit components of B if we so desired. We see that the
lines of B are circles whose centers lie on the line of motion of the charge; this is
reasonable because of our previous recognition of the equivalence of a moving charge
to a current element.
Thus, we have found the exact solution to this problem in a comparatively simple
manner. We see that E is directed radially outward from the charge, while (29-150)
shows that B is perpendicular to the plane of v and E. This situation is illustrated in
Figure 29-8. The basic structure of the field will be the same for all times, and can be
obtained by simply translating this figure along the x axis with speed v.
It is helpful to express our results in terms of the distance R from the charge and the
angle 0 made by R with the direction of the velocity. We see from the figure that
Rx = R cos 0 while R2 4- R2 = R2 R2 = R2 sin 2 0 so that
B =
y2R2 + R2y + R2 = R2y2(l
- fi2 sm20)
(29-151)
where now (3 = v/c. Substituting (29-151) into (29-149), we obtain

E =
tf(l - 0 2 ) R
47rc 0 i? 2 (l - fi2 sin 2 0)
T-tj
(29-152)
which shows that the field is inverse square in its dependence on radial distance from
2 9 - 6 FIELDS O F A U N I F O R M L Y M O V I N G P O I N T C H A R G E
525
Figure 29-8. The fields of a uniformly

moving point charge.
the charge but that, at a given distance, its magnitude depends strongly on direction, in
contrast to the simple Coulomb field.
In order to illustrate the dependence of the magnitude upon direction, let us first
consider two extreme situations. Directly in front or behind (0 = 0 or 180), we have
while to one side (6 = 90), we find that

q
1
2
4ttcqR (1 - p 2 ) l / 2
Therefore, for a very rapidly moving charge for which fl = 1, we see that, at a given
distance, E n is very small, while E x is very large. (As /? 0, both of these
components become equal to each other and to the static Coulomb field.)
These effects are also illustrated in Figure 29-9, which shows the magnitude of E at a
given distance plotted as a function of angle 6 for
= 0.0, 0.5, and 0.9, that is,
E/[q/AIT(.QR2] is shown. We see again that for a rapidly moving charge the field is
concentrated in a small angle around the equatorial plane.
It is also desirable to look at this problem in terms of the potentials. The electrostatic field in S' described by (29-142) corresponds to the potentials
(29-153)
and therefore
= A'2 = A'3 = 0 and A\ = i<j>'/c by (29-129). Inserting these values
into the inverse form of (29-132), we find that the corresponding potentials in S are
$ = y<j>' =
4tj- 0 [y 2 (x vt)2 + y2 + z2]1/2
A = Axx
since Ay = Az = 0.
vy<f>' a
2
n0yqv
47T[Y2(X
vt)2 + y2 + z2]1/2
(29-154)
(29-155)
526
SPECIAL RELATIVITY
Figure 29-9. The magnitude of the electric field

as a function of angle for
three different values
of p.
If we also use (29-151), we find that they can be written as
<7
477-c0i?(l (32 sin 2 d)1/2
^
=
ft
o<?v
47ri?(l - ft2 sin 2 0 ) 1 / 2
(29-156)
(29-157)
where R is the distance from the charge to the field point.

We note that the vector potential has the direction of the charge velocity, and that,
when fi <&: 1, A = ( / x 0 / 4 i r ) ( q v / R ) in agreement with (16-14) and thus justifying the
remarks in Section 14-5 about the applicability of our previous results to slowly moving
charges.
Both of these potentials depend on 6 in the same way. Directly forward (or back) of
the charge where 6 = 0 (or 180), <f> = q/A-nt0R and equals the static value. However,
EXERCISES
527
Figure 29-10. The magnitude of

the scalar potential as a function
of angle for three different values of p.
when 6 = 90, <f> = q/[4-7rc0R(l j82)1/2] = y<static and is larger by the factor y. In
Figure 29-10, we illustrate the general angular dependence by showing ^/[q/4m0R]
as
a function of 0 for fi = 0.0, 0.5, and 0.9 corresponding to the values of Figure 29-9.
EXERCISES
29-1 Show by direct substitution of (29-24) into
(29-7) that the Lorentz transformation preserves
the form of the wave equation, that is, if d2\p/dx2
= d2\p/c2dt2, then d2^/dx'2 = d2^/c2 dt'2.
29-2 If the Lorentz contraction (29-33) is assumed to represent an actual physical contraction
of the moving rod, show that this will account for
the result of the Michelson-Morlev experiment.
29-3 The average lifetime of a /x-meson before

radioactive decay as measured in its "rest" system is 2.22 X 10" 6 second. What will be its average lifetime for an observer with respect to whom
the meson has a speed of 0.99 c? How far will the
meson travel in this time?
29-4 A rigid rod of length L makes an angle 9
with the x axis of the system in which it is at rest.
528
SPECIAL RELATIVITY
Show that, for an observer moving with respect to

the rod, the apparent length L' and angle 0' are
given by
L' = L [(cos 0/y)2
4- sin2 tf]1/2 tan 6' = y tan 0
29-5 Show that two successive Lorentz transformations corresponding to speeds V1 and V2 in the
same direction are equivalent to a single Lorentz
transformation with a speed V = (Vx + V 2 )/[l +
(V x V 2 /c 2 )]. Is this result compatible with (29-37)?
29-6 Show that the velocity transformation formulas can be written in vector form as v = [v',
+ V + (v' / y ) ] / [ l + (V v')/c 2 ] where v[, is the
velocity component parallel to the relative velocity V of S and S' while
is the component
perpendicular to V.
29-7 Find the transformation laws for the components of the acceleration, ax = dvx/dt, and so
on.
29-8 Fizeau measured the index of refraction of
light in flowing water. His result can be expressed
in terms of the phase velocity v in the moving
medium of index of refraction n and flow velocity
V as v = vQ + V[1 (1 /n2)] where v0 = c/n is
the phase velocity in the stationary medium. Show
that this follows from (29-37) for the case V/c
1.
29-9 If the angles 0 and 0' of Figure 29-5 are

both small and if
1, show that the angles are
approximately related by 6 = (1 fi)d' so that
their fractional difference is (6' 0 ) / 0 ' = fi =
V/c.
29-10 Show that
[l - (v'2/c2)]1/2/[l
- (v2/c2)]1/2
= y [ l + (fiVx/c)]
= { y [ l - (fivjc)]}-1
29-11 Verify that the

given in (29-77) satisfy
the requirements (29-69) and (29-75).
29-12 Show that, if
is an invariant for
any arbitrary 4-vector A , then
is also a 4-vector.
29-13 Show that the quantities defined in (29-89)
are 4-vectors.
29-14 The 4-acceleration is defined by
=
dU^/dr = d2xfi/dr2.
Show that its components
are related to the ordinary acceleration a and
velocity v by
ci 1 =
aA =
[l - ( v2/c2) j
(i/c)(\
a)
[1 -(v2/c2)]:
ar +
(Â2)(v-a)l
[i-(y<>)]/ etc.
29-15 If k and u are the propagation vector and

circular frequency of a plane wave in free space,
show that
= (k, ito/c) is a 4-vector. Use the
transformation properties of k^ to derive the
Doppler and aberration formulas (29-47) and (2949).
29-16 Derive the set of coefficients a ^ that
describe a general Lorentz transformation consisting of a 30 rotation about the y axis plus a
translation along the rotated x' axis with a constant speed V = \c.
29-17 Using the transformation properties of the
Minkowski force , find the transformation laws
for the ordinary force f and power dW/dt. (The
results of Exercise 29-10 should prove to be helpful.)
29-18 For many purposes outside the scope of
this book, electromagnetic radiation can be treated
as composed of small, localized "clumps" of radiation called photons. Show that, if we regard a
photon as a particle of zero rest mass and total
energy W = hv where h is Planck's constant, the
Doppler and aberration formulas (29-47) and (2949) can be obtained from the transformation laws
for the 4-momentum P.,.
H29-19 Two particles, each of rest mass jM 0 , are
connected by a compressed spring of negligible
rest mass. The particles are tied together with a
massless string and the whole system is at rest in
a coordinate frame S 0 . The string is then cut, and
the two particles fly off in opposite directions,
each with speed v0. What is the initial potential
energy of the system in S01 By explicitly transforming the velocities to another frame S, find the
final energy and momentum, and thus the initial
energy and momentum, in S. Then find the rest
mass of the initial system in the frame S, and
interpret the result.
29-20 (a) Show that the quantities E B and
E 2 - c 2 B 2 are invariants, (b) Evaluate these
quantities for a plane wave, (c) Show that the
statement that E and B are perpendicular has
absolute significance, that is, if they are perpendicular for one observer they will be perpendicular for all observers, (d) Show that a field
that is purely magnetic in one frame cannot be
transformed into one that is purely electric in a
different reference frame, and conversely.
29-21 A charge q is instantaneously at rest in
S' and subject to an electric field E' so that the
force on it is f' = qE'. Use the transformation
laws for the force found in Exercise 29-17 and for
the fields found in (29-138) to show that the force
EXERCISES
in 5" is exactly the Lorentz force f = q(E + V X

B).
29-22 Show that the relativistic equation of motion (29-105) can be written in the form
m
=
1/2
dt
[l - (v /c )]
(v f)
c2
and that if f is the Lorentz force on a point

charge, q(E + v X B), the right-hand side becomes q(E + v X B [(v E)/c 2 ]v}.
29-23 Show that the equation of motion of a
particle of charge q in an electromagnetic field
where the force is given by f = <?(E + v X B) can
be written as
d2xfl
o~^T
? L W
29-24 Find the explicit expression for B due to a

uniformly moving point charge in terms of R, the
relative position vector of the field point and the
location of the charge. Show that your result
reduces to (14-28) when fi
1.
29-25 Show that the potentials <j> and A found
for a uniformly moving charge in (29-154) and
(29-155) give the correct values of E and B.
29-26 Find the Poynting vector S for the uniformly moving point charge. [Use (29-152) for
simplicity.] Show that the net power radiated by
this charge is zero.
29-27 A point charge qx at rt has a constant
velocity Vj. A second point charge q2 is located at
r2 and has an instantaneous velocity v2. Find the
total electromagnetic force on q2 due to qx. Show
that when your result is applied to the system of
Figure 29-2 and if V/c
1, then the magnetic
force which you found reduces correctly to (29-10).
29-28 An infinitely long straight line charge with
linear charge density X coincides with the x axis
of system S. Find the fields E' and B' in a system
S' with speed V in the direction of positive x.
Compare your expression for B' with that of a
529
current-carrying wire as given by (14-17) and

show that they are in agreement. Note that E' is
not zero in the moving system. What is the physical difference between a current-carrying wire and
a line charge moving in the direction of its length
that accounts for the appearance of E'?
29-29 An infinitely long ideal solenoid is at rest
in the frame S' with its axis parallel to the y'
axis. It has n' turns per unit length and carries a
steady current I'. Find E and B inside and outside the solenoid for an observer in S for whom
the solenoid is traveling with constant velocity
V = V::. Sketch the lines of E and indicate the
nature of the charge distribution that must be
associated with such an electric field. Show that
this charge distribution is in qualitative agreement
with (29-127). Will an electric field be observed
by someone who sees the solenoid moving with
constant speed along the direction of the solenoid
axis?
29-30 A point electric dipole of moment pC is at
rest at the origin in S'. Find the scalar and vector
potentials produced by this dipole as observed by
someone in S. How do the components of the
dipole moment transform? Qualitatively account
for the fact that A ^ 0.
29-31 In (28-72), we found an expression for the
total rate of radiation from an accelerated charge.
It was proportional to the square of the retarded
value of the acceleration, that is, to [a]2 = [a] [a],
A plausible way of generalizing this would be to
replace the square of the actual acceleration by
the " square" of the 4-acceleration . In this way
we obtain
Use the results of Exercise 29-14 to express this in

terms of the ordinary acceleration and velocity
and verify that the result reduces to (28-72) in the
appropriate limit. (This general result is known as
Lienard's formula.)
M O T I O N OF CHARGED
PARTICLES
Our principal interest so far has been in the properties of the macroscopic electromagnetic field. We have found how it is produced by its sources and have learned how to
describe the overall effects of the presence of matter. An important use of E and B is to
affect the motion of charged particles. Such effects have applications in the design of
apparatus for producing particles with large kinetic energies that are then used in the
study of reactions in nuclear physics and in " high-energy" physics. Other applications
arise in astrophysics and geophysics, plasma physics in general, magnetohydrodynamics, the study of thermonuclear reactions, and the construction of devices using
beams of charged particles. We will, however, confine ourselves to only a comparatively
few simple and basic situations.
We assume that the net force f on a particle of charge q and position r is the Lorentz
force q(E + v X B) where v = dr/dt is the particle velocity. Equating this to the mass
m 0 times the acceleration, we get the equation of motion
W
= ?(E +
(A-l)
V X B )
Our use of m 0 shows that we will consider only nonrelativistic cases. In the most
general situation, E and B can each be a function of position and time. In many
applications, the particles used can interact with each other and with other particles,
primarily by collisions. The effect of collisions then must be somehow included in
(A-l), but we will not try to do so.
As we know from mechanics, we will generally also need to know the initial
conditions of the motion, that is, the initial position r 0 and velocity v0 at t = 0 in order
to get a complete solution of (A-l).
[Vll
STATIC ELECTRIC FIELD
We assume that E is independent of time. If B = 0, then (A-l) reduces to

m
dy
- = qE
(A-2)
If we now also assume that E is independent of position so that E is a uniform static

field, then dy /dt is constant. We can immediately integrate (A-2) twice to obtain the
complete solution
(A-3)
r = r(/) =
530
(2
+ V
+ ro
(A-4)
A - 2 S T A T I C M A G N E T I C FIELD
531
In order to apply these results to specific problems, we can, for example, write the
components of (A-3) and (A-4) in rectangular coordinates. Thus, we find vx(t) =
(q/m0)Ext
+ v0x, x ( t ) = (q/2m0)Ext2
+ v0xt + x0, with similar expressions for the
y and z components.
For other applications, it is often useful to write some quantities in terms of their
components parallel (||) and perpendicular ( ) to the field E; thus we could write
v = VM + v x
r = rn + r
(A-5)
Applying this to (A-3) and (A-4), we obtain

q
V|11, = E f + v0||11
v ==vVox
0
m0
r
"
q .+
2m~
"'
+ ro11
(A-6)
=
Vo f +
which show explicitly that only the component of v parallel to E is affected while its
component perpendicular to E remains constant. This problem is similar to that of a
mass point in a uniform gravitational field and, just as in that case, the path described
by (A-4) is a parabola.
If E is not uniform, the particle trajectories can be more or less complicated
depending upon the precise dependence of E on position. We will consider only an
integral consequence of (A-2) which involves energy changes.
If the particle moves from r x to r 2 , the work done on the particle by the field is
W1_t
= j
f ds = q j 2 E ds = qAcj>
r
i
i
= -q[<t>(*2) ~ (ri)]
(A-7)
by (5-11) where A<p is the change in the scalar potential. Now we also know from
mechanics that the work done by the net force equals the change in the kinetic energy
T = \mQv2
(A-8)
so that Wl_2 = AT = T2 Tv Equating this to (A-7), we find that T2 + q$(r2) = Tv

+ q<t>(rx) showing that
T + <7<>(r) = \m0v2
+ q4>{ r) = const.
(A-9)
which is just the statement of conservation of energy where q<j>(r) is the potential
energy of the particle in the field. This last result is of course in agreement with (5-48)
where we came to the same conclusion.
lA-2l STATIC MAGNETIC FIELD
We assume B = 0 and independent of time. If E = 0, then (A-l) becomes

dv
m0 = f = qv X B
dt
(A-10)
Since f and v are always perpendicular, f v = 0 and the magnetic induction does no
work on the particle; consequently the kinetic energy (A-8) will be constant.
If we use the resolution of v into parallel and perpendicular components given by
(A-5), we find that
d\
dv, 11|
d\,
q
q
'
v X B = v X B
dt
dt
dt
m0
mQ
532
M O T I O N O F C H A R G E D PARTICLES
since Vy X B = 0 by (1-22). Thus, we obtain

d \ d
-J = 0
dt
,
-
dt
q
= - v
m0
x B
(A-ll)
since v x X B is also perpendicular to B.
Now let us assume that B is independent of position so that we have a uniform

magnetic induction. We see then that vN = const., so that the particle moves with
uniform velocity along the direction of B. We see that d\/dt
is always perpendicular
to both v and B, and so v itself has a constant magnitude v . Consequently, the
magnitude of d\/dt
will be constant and equal to ( q / m 0 ) v B . This situation is
illustrated in Figure A-l, which is drawn for positive q and B out of the paper. We
recall that an acceleration of constant magnitude that is always perpendicular to the
velocity results in motion in a circular path. We can find the radius rc of the circle by
recognizing that d\/dt
must have the magnitude of the necessary centripetal acceleration v\/rc.
Equating this to ( q / m 0 ) v B, we get
mnv ,
r c - - g - (A-12)
which involves the particle properties only through the ratio of the charge to the mass.
We can also obtain the period of revolution Tc and the angular frequency COC by noting
that in one period the particle travels a total distance 2-7rrc so that Tc = (2vrc/vx);
then to c is given by U>C = 2TT/TC. Combining these expressions with (A-12), we find that
20
TcC =
qB
<o
=
cc
/ q \
\m0)
,
(A-13)
The name cyclotron frequency is given to coc. We see that both these quantities in (A-13)
are independent of both rc and v . The faster particles move in circles of greater
radius by (A-12), but all particles of the same value of q/m0 take the same time to go
around their path.
We can also write these results in another way. We let r c be the position vector of
the particle with respect to the center of the circular orbit G (for "guiding center").
Then we will have
r =
rG + rc
(A-14)
where rG is the position vector of G with respect to the origin of our fixed coordinate
system. We now see from the figure that the motion of the particle can be described as
a rigid rotation of r c about G with angular velocity to c . We know from kinematics that,
533
in such a case, we can write the rate of change of the vector v x in the form
dvl
= u r X v,
dt
(A-15)
Equating (A-15) with (A-ll), we find the vector angular velocity to be

<oc=-( ) b
(A-16)
Wo/
and whose sign agrees with Figure A-l since u c is into the page.
If V|| = 0, the total motion of the particle can be described as a constant velocity Vy
of the guiding center along the direction of B with the circular motion arising from v
superimposed on it. In other words, the motion will be a helix as shown in Figure A-2
for positive q and Vy in the same direction as B.
If we combine (A-8) and (A-5), we find that we can write
T
\\+
= \mov\
(A-17)
where
and T can be identified with those contributions to the total kinetic energy
associated, respectively, with the motion parallel to and perpendicular to B. We have
just seen that
and v are both individually constant for a uniform induction so that
the individual contributions are themselves constants:
Ty = \ m0v\ = const.
T = jmQv2 = const.
(A-18)
A point charge traveling around the circle of Figure A-l is equivalent to a current
element Ids = qv by (14-29) and thus will have a magnetic dipole moment m. We see
from Figure A-l that m will be directed into the page and hence in the direction of w c .
According to (19-27), m = IS = Iirr^ where / is the equivalent current. Since the total
charge q passes a given point on the perimeter each time around, and since each circuit
takes a time Tc, the equivalent current is I = q/rc = wcq/2ir. In this way we find that
m
1 (m0$m0v$
mQv\
( A
.19)
with the use of (A-12), (A-13), and (A-18). If we take the direction of m into account we
can write
ra
"M
c = T
" I tJ ~b^ ~ a1"
(A 20)
with the use of (A-16). We see that m is opposite to the direction of B and hence is
diamagnetic in the sense discussed following (20-52).
The magnitude of the magnetic flux enclosed by the circular orbit is
S> = Birr2 = 2 f l " | ^ ) ^ ; - = 2 7 r | ^ J m
(A-21)
and is, of course, a constant.
Figure A-2. The general motion in a uniform

induction is a helix.
534
Example
Rectangular coordinates. Although we know that the general motion in a uniform

induction will be a helix as in Figure A-2, it is instructive to see how this turns out in
detail in terms of a specific coordinate system. Let us use rectangular coordinates and
let B = Bi. where B = const. Then VJJ = vzz while v = vjL + vyy and we find that
( A - l l ) gives
dvz
= 0
dt
dvx
dt
qB
m0
dvy
=
dt
vyv
qB
m0
uY
(A-22)
From the first of these we obtain vz = vQz = const, so that z = z0 + v0zt.

If we differentiate the second equation of (A-22) and substitute from the third, we
get an equation for vx alone:
d \
dt2
^ d v
y =
- f r
mno dt
(A-23)
\\ rn
mo '' *
This has a solution

vx = v0x c o s (
)t
\ moj
(A-24)
corresponding, as we will see, to choosing v0y = 0. Since vx = dx/dt,

(A-24) once more to obtain
x
-*<,
'o(^)
S i n
we can integrate
()'
(A-25)
Now we can find vy directly from the second equation of (A-22) and (A-24):
vyv =
dvx
. / qB \
= 0x
v0x sin
t
qB dt
\m0 J
(A-26)
'
and therefore since v = dy/dt, we get

y = yQ + vtOx
/ mo
[qB
q B
cos Im \ \t , 1
\ o)
(A-27)
First of all, we see from (A-24) and (A-26) that v\ = vx + vy = v})x = const, in
agreement with (A-18). If we define y$ = y0 v0x(m0/qB),
then (A-27) can be written
somewhat more simply as
7-yi + oo, ( ^ ) c o s ( 0
<A-28>
We now find from (A-25) and (A-28) that

2
( x - x
+ ( ^ - ^ )
= | ^ j
(A-29)
which is the equation of a circle with center at ( x 0 , y) and a radius rc = m0u0x/qB =

m
o " / q B in agreement with (A-12) and Figure A-l. We note that the center of this
535
circle is not at the initial position in the xy plane, that is, at (x0, yQ) but that the y
coordinate of the center is y^ as given by (A-28) and (A-27).
We also see that all of the xy coordinates and velocity components oscillate with the
circular frequency qB/m0 in agreement with (A-l3).
Finally, we see that if x JC0 is increasing from zero, say, then y y{, will be
decreasing from rc. Thus the position vector in the xy plane, r c = (JC JC0)X + (y
y$)y, will be rotating clockwise as seen looking toward the xy plane from the direction
of the positive z axis, that is, opposite to the direction of B. But this is exactly the
rotation shown in Figure A-l so that we have verified in detail all of the previous
results by means of these specific calculations.
Example
Cyclotron. We have seen that the time rc required to go around the circular path is
independent of the radius of the path and the speed of the particle. This fact has been
used as the basis for the device, used to accelerate charged particles to large kinetic
energies, called the cyclotron. The principal features are shown in Figure A-3. A thin
circular hollow conducting cylinder is cut along a diameter and the two portions
separated by a small gap. A uniform B is normal to the plane of these "dees," while an
alternating potential difference A<f> of frequency coc is applied between them. A charged
particle that is produced within the gap will be accelerated toward one of the dees and
enter its interior. Inside it will be shielded from E but will travel in a semicircular path
whose radius is given by (A-12). It will reenter the gap at a time \TC later where it will
find that E in the gap has been reversed. It will be accelerated across the gap, gaining
kinetic energy ^|A<#>|, according to (A-7), and will enter the other dee. Since
is now
larger, its radius will be larger by (A-12), but it will travel its new semicircle in the same
time \TC and will reenter the gap to find E again reversed. It will again be accelerated
across the gap and will travel in a new semicircle of larger radius. The whole process
will continue in this manner, the particle gaining kinetic energy <7|AJ>| each time it
crosses the gap, and thus traveling a path of ever increasing radius. Finally, when the
radius is nearly equal to the radius R of the dees, the beam of particles can be extracted
with a deflecting field near an opening in the wall of one of the dees. Since /*Cmax = R,
we find from (A-12) that the maximum speed v m a x which the particle can be given is
q > 0
B (out)
Alternating
potential
difference
Figure A-3.
Schematic illustration of a cyclotron.
536
qBR/m0,
so that the final kinetic energy of a particle in the emerging beam is

7
1
max= 2mV
ma 2 =
q2B2R2
2m Q
(A"30)
which is proportional to B2 and to R2.

When the particle speed gets very large, the relativistic equation of motion (29-105)
has to be used. If we interpret this as corresponding to an increase in the particle mass,
we see from (A-13) that the cyclotron frequency will no longer be a constant but will
vary with the particle speed. Consequently, the frequency of the applied potential
difference must be correspondingly altered to keep " i n step" with the particles in order
to prevent them from finding an E of the wrong sign when they enter the gap, as this
would decelerate them and decrease their kinetic energy.
Example
Magnetic focusing. In Figure A-4 we show a particle that is projected at t = 0 from a

point P with an initial velocity v0, which makes the angle 6 with the direction of a
uniform B. It will travel in a helical path like that of Figure A-2. After a time r c , it will
have made one turn of the helix and will be at the point Q, which has the same
coordinates in the plane perpendicular to B as does P. We can also see this from (A-25)
and (A-27) since, when t = Tc = 2irm0/qB, the argument of the sine and cosine is 277and X(TC) = JC0 and _y(r c ) = _y0. During this time, the particle has traveled the
horizontal distance
1
= v\\Tc = 2^\\
U 0 COS
T ! ~ B
(A-31)
Now consider a group of particles for which P, v0 and 6 are the same but whose
azimuthal angles <p about the axis of B are different. Each of them will travel its own
helix, but they will all strike the line PQ at the same point Q. In other words, they have
been brought to a focus at Q, and one can call the distance I the focal length. We see
from (A-31) that a measurement of / could then be used as a way of measuring the
charge to mass ratio q/m0. It turns out that this arrangement does not lead to great
accuracy for such a measurement; the focusing property described does, however, prove
to be of use in other applications.
Each case in which B is not uniform generally has to be considered separately.

However, there is one situation of considerable practical interest that we can discuss
approximately. This is the one for which B has axial symmetry and is slowly varying
with position. Thus if we take the z axis of cylindrical coordinates to be the axis of
symmetry, we can write B(z, p) and assume that the variation of B with both z and p is
Figure A-4.
Magnetic focusing.
537
not great. This would correspond to lines of B like those of Figure A-5. We can expect
that the motion of the particle will still be approximately helical as indicated in the
figure.
There will now be a radial component Bp in addition to Bz. We can find its
approximate value by using V B = 0, which, with the use of (1-87), becomes
I d ,
dB
z
< A
For a slowly varying Bz, we can set dBz/dz

B
3 2 )
const., and integrate (A-32) to give

1
p = -
"
3B:
(A-33)
2 ~dz
since Bp(0) = 0 from Figure A-5, that is, from the way in which the z axis was chosen.
Using (1-76), we find that the z component of (A-10) becomes
dv,
= ~qvvBp
(A-34)
dt
In a slowly varying B, the radius of the helix will not change very much for each turn so
that we can take vp
v and then v^ v since v x and $ are oppositely directed
for positive q. Making this substitution into (A-34), along with (A-33), and using (A-12)
to express p as the radius of the orbit, we find that
m0
dv2
m
1 m0v2 8BZ
dBz
dBz
"~dt
~s7
with the use of (A-19). If we now multiply both sides of this by vz, we obtain
m
dVz
d( 1
d T " J, 2
\
m
8Bzdz
=
(A 35)
"
dBz
m
= ~ IT
(A
"36)
where dBz/dt is the total rate of change of Bz as seen by the particle as it moves along.
Since the magnetic force does no work on the particle, the total kinetic energy T is
constant and we find from (A-17) and (A-19) that
dTn
dt
d( 1
dt\2mVz
\ _
I
dT
dt
d
(mBz)
dt
(A-37)
to the same order of approximation that we have been using. Equating the right-hand
sides of (A-36) and (A-37), we find that d(mBz)/dt
= m(dBz/dt) + Bz(dm/dt) =
m(dBz/dt)
and therefore
dm
= 0 and m = const.
(A-38)
dt
Figure A-5. General motion in a

nonuniform B with axial symmetry.
538
In other words, the particle moves in such a way that its magnetic dipole moment is
approximately a constant. If we look again at (A-21), we see that an equivalent way of
saying this is that during the motion the particle orbit encloses the same amount of flux
<!>. Thus as it moves into regions of larger field magnitude, its radius will decrease. Since
$ is constant, we can also say that the orbit will enclose the same lines of B as is
indicated in Figure A-5.
We can integrate (A-37) and we find that
\m0vz
+ mBz = Tj( m B = const. =
(A-39)
if we take into account the opposite directions of m and B shown in (A-20). We see
from this that as the particle moves into a region of larger Bz, its longitudinal kinetic
energy will decrease, and if mBz becomes equal to <f, the particle will reverse its
direction and move back into the region of weaker field; this is in agreement with
(A-35) which indicates a force on the particle opposite to the direction of increasing Bz.
In other words, the particle will have been reflected. This fact is the basis of magnetic
mirrors, which are used in attempts to confine the ionized particles produced at the high
temperatures of fusion reactors.
During the process described above in which
will decrease, T will increase in
order to keep T constant. Thus the particle will go about its helix of decreasing radius
with greater and greater speed until the point of reflection is reached. This same
conclusion also follows from (A-38) and (A-19), since if B increases, T must also
increase in order to keep the ratio m constant.
f~3l
STATIC ELECTRIC A N D MAGNETIC FIELDS
We now assume that both E and B are different from zero. We also assume that they
are independent of position and time. The complete equation of motion (A-l) must
then be used. It is helpful to consider it in stages.
First we introduce the components of both v and E that are parallel and perpendicular to the direction of the magnetic induction. Then (A-l) becomes
dv
\\
w0 +
o~^~
^(eh +
+v-L
x B
,
(A-40)
since va X B = 0. This yields the two equations

m0
rfVy
dt
d\,
m0-^~
= ? E II
(A-41)
= q(E+yx
XB)
(A-42)
The solution to (A-41) is given by the parallel parts of (A-6) with Ey replacing E; thus
Vy is unaffected by B.
The effect of the remaining portion of E can be accounted for by writing
v =
+ v'
(A-43)
where vD will be chosen in a way that we will soon determine., We note that vD and v'
are each perpendicular to B. Using (A-43), we can write (A-42) as
m0
dy,
~
= q(Ex
+vD
B + v'
B)
(A-44)
A - 3 S T A T I C E L E C T R I C A N D M A G N E T I C FIELDS
539
and we would like to find \ D such that

E
+ v X B = 0
(A-45)
We can solve this for \ D by crossing B with it and using (1-30); the result is
B x E + B X ( v x B ) = 0 = B x E i + B2vD since \D B = 0. Thus we find that
B X Ex
B2
ExXB
B2
const.
(A-46)
Since vD is constant, we find that (A-44) reduces to

w
d\'
= q\' X B
(A-47)
which is exactly of the form of the second equation of ( A - l l ) and for which we know
the complete solution from the last section, that is, a circular motion like that of Figure
A-l.
The general motion in the simultaneous presence of uniform E and B thus is the
superposition of three parts. There is a constant acceleration along the direction of B
produced by E |( . There is a constant velocity in the direction of E x X B and hence
perpendicular to B. Finally, there is the circular motion about the axis of B with radius
and frequency given by (A-12) and (A-13) with the constant speed v' replacing v . The
velocity vD is known as the drift velocity, its magnitude is E/B and is independent of
both the charge and the mass of the particle.
For simplicity, we will concentrate on the combined drift and circular motion, that
is, on the projection of the motion on the plane perpendicular to B; E and vD are
both in this plane. This combined motion can be described in convenient geometrical
terms. The circular motion associated with v' has a radius r^ = mQv'/qB and will make
one complete rotation in the time r c = 2irm0/qB. During this time, the guiding center
G will have traveled a distance dG = VDTc = 2"!rm0Ex/qB2.
This is just the cir2
cumference of a circle of radius rG mQE/qB
= vD/coc. Thus the motion of G
corresponds to that of a circle of this radius that is rolling without slipping on a
horizontal line and with angular velocity ooc as illustrated in Figure A-6. Since the
circular motion described by v' is about G, we see that the net motion of the charge
will be the same as if it were rigidly fastened to the rolling circle at the point P a
distance r from the center G. This is exactly the situation that produces the curve
known as a cycloid. If rG = rG, then P is on the periphery of the circle and the curve
that P traces out is the ordinary cycloid shown in Figure A-7a. If rG > rG, then P is
outside of the circle and produces the prolate cycloid of ( b ) of the figure, while if
q > o
B (out)
w
(in)
Figure A-6. Calculation of the motion

in a uniform E and B.
540
(a)
(b)
(c)
Figure A-7.
Possible types of cycloidal paths.
c < rc> 'he curve is the curtate cycloid shown in (c). From our previous results, we
have
B
O o v'/qB)
( V
/ 2 I T )
(A-48)
Example
Rectangular coordinates. As we did in the last section, let us verify these results in terms
of a specific coordinate system. We assume that E = Ey and B = Bz. Thus Ey = 0 and
U|l = vz will be constant; we choose vz = 0 so that the motion is confined to the xy
plane. We also see from (A-46) that \D = vjx. = ( E / B ) x in this case. However, let us
begin anew with (A-l), which, for these fields, yields the two equations of interest:
dv
qB
dt
m,
dt
= {E
mr.
-Vy
(A-49)
- Bvx) = -uc(vx
- vD)
(A-50)
If we differentiate (A-49) and substitute from (A-50), we obtain

d
IF
dvy
=
which has a solution

Vx = vD + (v0x - vD) cos 0ict
(A-51)
A - 3 S T A T I C ELECTRIC A N D M A G N E T I C FIELDS
541
corresponding, as we will see, to choosing v0y 0. We can now find vy from (A-49)
and (A-51):
1 dv
vy
dt
= - ( v0x -
D)sin
(A-52)
Before we go on to find the coordinates, we can verify that these results for the velocity
components are consistent with our previous ones. The components of v' as obtained
from (A-43) are
v'x = vx-
vD
v'y = vy
(A-53)
so that (A-51) and (A-52) can also be written as

K = v'ox cos
c*
'y = - Kx sin coct
v'Qx = v0x - vD
(A-54)
The expressions for v'x and vy are just those expected for a velocity v' of magnitude
| v' | = v' = | fOJC t rotating in a clockwise sense with angular speed coc as in Figure A-6.
We can find the particle coordinates by integrating (A-51) and (A-52), and evaluating the constants of integration in terms of the initial position. The results are
x = x 0 + vDt +
(V0x
y = y0 4which can be rearranged as

,
D ,
(v0x vD)
d ) ,
y =y0+
(cos coct 1)
x = x0 + \ ^ c O ~
(VD~V
Ox)
u
0>C
, (vD-v0x)
0>c
sinto c /
x
(A-55)
rr\
(A-56)
K
sinc/
(1 - cos o)ct)
These are in the usual parametric form of the equation of a cycloid with associated
radii rG = Vd/OJc,
= \vD V0x\/OJc = v'/oic in agreement with previous results.
An interesting special case arises when v0x = vD for then (A-55) and (A-56) become
simply x x 0 + vDt, y = y0 showing that the particle travels in a straight line with
constant velocity equal to the drift velocity and hence is otherwise unaffected by the
fields. We see from (A-45) that this occurs because, with this initial velocity, the net
force is zero and remains so since there is no acceleration to change v from its initial
value of vD. We note that this straight line motion is independent of either q or m0. In
this sense, this field combination acts as a velocity selector since all paxticles traveling
horizontally must have the same definite velocity \ D .
m
Example
Magnetron. This is an example of a case in which B is uniform but E is not. (See Figure
A-8.) The system consists of two coaxial conducting cylinders of radii a and b with a
uniform induction B parallel to the axis. There is a potential difference A<#> between the
cylinders so that a radial field E is produced. Suppose a charge q is introduced at the
surface of the inner cylinder. The field E will accelerate it toward the outer cylinder.
Because of the B, however, the path will be curved and the particle may be turned back
before it can reach the outer cylinder. In this case there will be no current between the
cylinders. For a smaller B, the charge q can reach the plate and there will be a current.
542
Figure A-8.
Magnetron.
Thus a knowledge of the conditions for the transition between no current and current
will enable one to use this device to measure B. (In an actual magnetron, the inner
cylinder is a heated filament and the particles that are produced are electrons of charge
e. It is just as easy to analyze this problem for an arbitrary charge q, however.)
The variation of both E and <j> with position can easily be found by using Gauss' law
(4-1) or the general expression (11-141). As we will see, we actually do not need to
know these details. Since B does no work on the charge, the energy conservation
expression (A-9) is applicable. If we use cylindrical coordinates and note that </> = <f>(p),
by symmetry, (A-9) can be written as
where the velocity components can be found from (1-82). The constant can be
evaluated by setting p = a; if we assume that the particle is produced at the inner
cylinder with negligible velocity, its kinetic energy is zero and the constant in (A-57) is
just q<t>(a) and we have
(A-58)
We want the condition that the particle just misses the outer cylinder at p = b. If
this is the case, its path will be tangent to the surface of the cylinder so that
vp(b) = (dp/dt)b = 0 and (A-58) gives
(A-59)
We can get another expression for the quantity on the left by turning to the equation of
motion.
With v = upp +
E = Ep, and B = Bz, we find with the use of (1-76) and a
result from kinematics that the cp component of (A-l) is
= m0
d2
~dt
= -qBvp
.dp
= ~qB
A - 4 A T I M E - D E P E N D E N T M A G N E T I C FIELD
543
If we multiply both sides of this by p, we find that it can be written as

d
dt
2 I d(f>
= ~qBpvp = ~qB ( - P 2 |
(A-60)
In this form, it is perhaps more familiar. The term in brackets on the left-hand side is
the angular momentum of the particle about the z axis, while the middle expression is
the z component of the torque as found from r X f == pp X q[Ep 4- ( v p p -I- typ) X Bz]
using (1-76) again. Equating the first and third expressions of (A-60), we can integrate
the result to obtain
dtp
qB 2
p2 +
p = const.
(A-61)
dt
2 m0
We can evaluate the constant by considering the initial situation at the inner cylinder
p = a. Since we have assumed the particle to originate with negligible velocity, we have
(dcp/dt)a = 0 and the constant equals (qB/2m0)a2
so that (A-61) becomes
P2-J" = (p 2 - a 2 )
dt
2m 0
(A-62)
We note that since p > a, dcp/dt is negative for q > 0, which agrees with the path
shown in Figure A-8.
If we now set p = b in (A-62), we obtain
* ( ) = -^(b2
\ dt ) b
2m0
If we solve this for (dcp/dt)b
a2)
and substitute the result in (A-59), we find that

B2 =
8m0b2A<t>
q{b2 a2)
(A-63)
which relates the value of B to the value of the potential difference A< for which the
current between the cylinders just vanishes.
lA-4l A TIME-DEPENDENT MAGNETIC FIELD
Although a magnetic induction that is constant in time does no work on a charged

particle, a time-varying one can do so by means of the induced emf described by
Faraday's law. This fact has been used as the basis for the charged particle accelerator
known as the betatron; the name arises because it was originally used to produce
electrons of large kinetic energy and thus they were similar to the beta particles of
radioactivity.
The essential features of this device are shown in Figure A-9. A particle is
constrained to move in a circle of radius R that encloses a cylindrically symmetric
induction B = B,z, which is normal to the plane of the orbit. The induction B is varied
in time so that the work done on the particle in each complete circuit equals its charge
times the induced emf S'ind given by (17-6) as
,
d<&
d(Bz)
= $E, d ' d s - - (A-64)
where ( B z ) is the average value of the induction over the area enclosed by the orbit.
Because of the cylindrical symmetry we can write E =
and the integral becomes
544
CD
CO
Figure A-9. Essential aspects for describing the action

of a betatron.
2TTREY
Thus the magnitude of the induced electric field is

1
Ew = -R
*
2
d(B.)
(A-65)
dt
and it will produce a tangential force of magnitude

U
1 V
dt
mt 1
dVy
dt
(A-66)
We require that the particle be constrained to move in a circle of fixed R. This can
be done with an induction normal to the plane of the orbit and with a value BR, this is
known as the "guiding field." The required value of BR can b e obtained from (A-12)
and is BR = (m^/qR)
since v = v in this case. Since vv is changing by (A-66), BR
must be changing as well in order to keep R constant, and we find that it is necessary
that
dBK
m0 dvv
(A-67)
dt
qR dt
When we eliminate m0(dv^/dt)
between (A-66) and (A-67), we find that we must have
(dBR/dt)
= \(d(Bz)/dt),
which leads to
BR
2 (BZ) + const.
(A-68)
which is known as the betatron condition and must be satisfied at all times if the
betatron is to work. Normally, the two fields BR and (Bz) are changed simultaneously
with both starting from zero. In that case, the constant in (A-68) must be zero and the
condition simplifies to BR = \(Bz).
In other words, the induction at the particle
position must always equal one half the average induction enclosed by the circular
orbit.
EXERCISES
A-l At / = 0, a particle is at the origin with
initial velocity v0 = v(jk and enters a region of
uniform electric field E = Ey. Find the equation
of its path and its deflection d from the x axis
after it has gone a distance D along x. Find the
angle a made by v with the x axis at this point.
A-2 A particle moves in two dimensions in a

region where the scalar potential is <$> = jA(x2
y 2 ) where A is a constant with the same sign as q.
At t = 0, the particle is at rest at (x 0 , j 0 ). Find x
and y for subsequent times and sketch the resultant path of the particle.
EXERCISES
545
A-3 Suppose that B = 0 for the system of coaxial

cylinders of Figure A-8. Show that circular orbits
are possible when A<p + 0. Evaluate the kinetic
energy and show that it is independent of the
radius of the circular orbit.
out into a cone of very small angle. A long ideal

solenoid of 10 turns per centimeter has its axis
aligned with the axis of the cone. What current
will be required in the solenoid in order to refocus
the electrons within a distance of 10 centimeters?
A-4 A charged particle is traveling horizontally

with speed v0 when it enters a region of uniform
B directed vertically downward. Find the deflection d from its original direction of travel when
the projection of its position on its original direction is D where D < rc.
A-9 Suppose that charged particles are produced with negligible velocity at one plate of a
parallel plate capacitor of plate separation d.
There is a potential difference A<f> between the
plates and a uniform induction B is parallel to the
plates. Show that there will be no current between
the plates if A<p is less than \(q/m0) B2d2.
A-5 In one type of mass spectrometer a particle

of charge q is produced at rest and then accelerated through a potential difference A<. It then
enters a region of uniform B, which is perpendicular to its velocity. After traveling a semicircle,
the particle strikes a detector at a distance D
from its point of entrance. Find an expression
from which the mass m 0 can be found in terms of
the quantities given.
A-6 Suppose that Figure A-4 shows the xz plane
with B along the positive z axis and P is the
origin. Find the position vector r c of the guiding
center of this motion as a function of t.
A-10 The curves of Figure A-7 were drawn on

the assumption that q was positive. What changes,
if any, will there be in this figure if E and B are
unchanged but q is negative?
A-11 At / = 0, a charged particle at the origin
has a velocity v0z. It is in the presence of parallel
uniform fields E = Ex and B = Bx. Find the
coordinates x and y when the particle has reached
one half of its maximum value of z for the first
time.
A-12 A charged particle is in a region of uniform B and is also subject to a uniform gravitational field so that the gravitational force on it is
A-7 Suppose particles of the same q, m0, and
v0 are projected from P in Figure A-4 over a m 0 g. Show that it has a drift velocity equal to
range of angles 90 AO < 6 < 0O + AO. Their fo- m 0 ( g XB)/qB2. (Note that this drift velocity
cal points Q will vary over a corresponding range depends on the charge to mass ratio of the particle.)
of position A/. Find A/ and the ratio A/// 0
where /() is the focal length for the central angle A-13 At t = 0, a charged particle at the origin
0o. What is the largest allowable value of A0 if
has a velocity v0 = v^x. There is a uniform electric
60 10 and the spread A/ is required to be less field E = Ey present. Solve the relativistic equathan 5 percent of /0?
tion of motion (29-105) to find the coordinates for
later times. Verify that for small values of t, the
A-8 After being accelerated through a potential
difference of 100 volts, a beam of electrons spreads path of the particle is approximately a parabola.
B
ELECTROMAGNETIC
PROPERTIES OF MATTER
We have described the overall effects of the presence of matter on electromagnetic fields
in terms of parameters such as the electric susceptibility xe- From a purely macroscopic
point of view such parameters are assumed to be characteristic of a given material and
their numerical values are to be determined from experiment. In other words, the
formulation of the electromagnetic field that we have obtained does not attempt to
predict their values. The observed fact that different materials have different values of
these quantities indicates that they depend on the details of the microscopic structure of
the material, that is, on the atomic and molecular properties of its constituents as well
as on its state of aggregation, whether solid, liquid, or gas.
In this appendix we consider some of the principal features that are involved in a
microscopic approach; we have already seen an example of this in our brief discussion
of conductivity in Sections 12-5 and 24-8. Whole books can be, and have been, written
on these general subjects so that we will necessarily be very limited in what we can do.
In particular, while quantum mechanics is the appropriate way of dealing with atoms
and molecules, we can get sufficiently good results by using classical mechanics to
investigate the mechanical response of a collection of interacting charged mass points
to an applied electromagnetic field. The results we will obtain in this way will not be
completely accurate in all details, but are surprisingly good and will give us an
adequate understanding of what is going on.
STATIC ELECTRIC PROPERTIES
We want to evaluate the electric susceptibility x e defined in (10-50) for linear isotropic
materials by
P = X<E
(B-l)
This shows that we have to calculate the electric dipole moment per unit volume P. In
turn, P will be related to the average dipole moment of the molecules within the
volume.
At the beginning of Section 10-1, we discussed the possibility that a material could
become polarized by the production of an induced dipole moment. This induced
moment arises when an electric field applied to a neutral molecule with no permanent
dipole moment shifts the positions of the positive and negative charge distributions
with respect to each other. The relative shift is finite because the electrostatic forces on
the charges will be opposed by internal forces within the molecule and a new state of
mechanical equilibrium will be attained.
To be specific, let us consider a neutral molecule constructed from n atoms. The i th
atom has a positive nucleus of charge Z,e together with Z, electrons each of charge - e
where Z, is the atomic number. The total number of electrons will therefore be
Z, = Z : + Z 2 + . . . +Zn. If we let r( be the position vector of the /th nucleus and ry
546
B-1
S T A T I C ELECTRIC P R O P E R T I E S
that of the 7 th electron, the total dipole moment as found from (8-19) is
z,
P = L Zm - erJ
i-l
7=1
547
(B-2)
Since the molecule has no monopole moment, we know from (8-43) that we can use any
convenient origin for our coordinate system, for example, one at the center of mass; for
a monatomic molecule, we could choose the origin at the location of the nucleus so that
the first term of (B-2) would be zero. Since the charges can be assumed to be moving,
the electrons in particular traversing some "orbits," p can be a complicated function of
time. We are, however, concerned here only with time average values, so that our
interest is really in
(B-3)
'
so that
P = N(p)
(B-4)
where N is the number of molecules per unit volume.

If we let values in the absence of an applied field be designated by the subscript zero,
we can write (B-3) as
Po = 0 = Ez,<?(!,.>0 - e ( r y ) 0
i
(B-5)
If p0 # 0, the molecule has a permanent moment, while if p0 = 0 it does not and P as

given by (B-4) can arise only from induced moments. For example, in a monatomic
molecule, we can assume the electron charge distribution to be spherically symmetric
about the nucleus. Thus, if we choose the origin at the nucleus, we can take (r,) 0 = 0
and then (r.) 0 will be zero also because positive and negative values of xp yp Zj will be
equally likely so that (xj)0 = (yj)0 = (Zj)0 = 0. The result will be that p0 = 0 as we
expect.
In the presence of a polarizing field we can expect each (r) to be different from (r) 0 .
Since the shift arises from the balance between an electrostatic force and an internal
mechanical "restoring" force F m , we can write
= 0 = F + c,Ep
(B-6)
for the new equilibrium situation for any charge q as we did in (2-9). In (B-6),
is the
"effective" or "polarizing" electric field that is producing the displacement of the
charges. [It is sometimes called the local field, but we have already used this after
(10-70) to refer to the field produced by the bound charges.] It is not necessarily equal
to the macroscopic average field E in the material since the molecule in question is itself
contributing to E; we return to this point shortly.
Since Fm is to describe some sort of restoring force, we expect it to be a function of
the displacement ((r) (r) 0 ) and also that Fm(0) = 0. Furthermore, the displacement
is not expected to be large so that it is appropriate to expand Fm in a power series in its
argument and keep only the first term; this gives us
U < r > " ( r ) o ) = - K r > - <r>o)
(B-7)
where K is an effective "spring constant" for this charge q. Substituting (B-7) into
(B-6), we obtain
<r> - <r>o " f E
(B-8)
If we now substitute this into (B-3), using the appropriate charge in each case, and then
548
E L E C T R O M A G N E T I C PROPERTIES O F MATTER
use (B-5) along with our assumption that p0 = 0, we find that the average induced
dipole moment is
(P) = * ' - + E ^ ) E , = a E ,
> '
j
ij
(B-9)
and is linearly proportional to the polarizing field E . The constant of proportionality a

is called the polarizability. [We recall that in (11-112) we found a similar result for the
grounded conducting sphere when in a previously uniform field.] We see that when
(B-9) is substituted into (B-4), the polarization will be proportional to the inducing field
E^ as expected for a linear isotropic material.
The polarizability in (B-9) is expressed in terms of sums involving the various
"spring constants" of the molecule. We can also get a crude estimate of its value in a
somewhat different manner as follows. Let us consider a single atom and assume that
the electron charge distribution is not only spherically symmetric but uniform
throughout a sphere of radius a. We further assume that the effect of the field Ep is to
shift the negative charge distribution rigidly with respect to the nucleus so that the final
result is a relative displacement 1 between their centers as shown in Figure B-l. Then
the induced dipole moment will be (p) = (Ze) 1 by (8-44). The nuclear charge Ze can
then be considered to be in equilibrium under the influence of E and the electric field
at the surface of the sphere of uniform negative charge of radius /. Equating Ep to the
magnitude of this field as given by (4-21), we get E = (Ze)l/4ireQa3
= Ep and
3
therefore ( p ) = Zel = 4TT0a Ep. Thus the polarizability is
a = 47re 0 a 3
(B-10)
and is proportional to the volume of the atom. [We note that (B-10) is the same as was
found for the conducting sphere in (11-113).]
N o w we turn to the problem of evaluating the polarizing field E . We assume the
material to be homogeneous as well as linear and isotropic. We want the total electric
field at the location of the molecule except for that arising from the molecule itself. If
we think of the usual case of a material in a capacitor, the field E will have
contributions from the free charges on the plates, the bound surface charges on the
dielectric boundaries, and from all of the rest of the dipoles in the material. Thus we
can write
Ep = Ef+Eh
+ Ed
(B-ll)
[If we assume no free charge within the dielectric, there is no bound volume charge
density either by (10-58); the charges contributing to Ef and E 6 are illustrated for a
B 1 STATIC ELECTRIC PROPERTIES
549
parallel plate capacitor in Figure 10-6.] Now we know that the macroscopic average
field is produced by the free and bound surface charges only so that
is actually
E so that we can write
Ep = E + E
(B-12)
The dipole moment of each of the remaining dipoles contributes to E 0 . If we are very
far f r o m the molecule in question, the dipoles will appear to it to be continuously
distributed so that it would be appropriate to describe them in terms of the uniform
dipole moment per unit volume P. On the other hand, the other dipoles that are very
close on a microscopic scale will appear as discrete dipoles located at specific points; in
other words, they cannot be treated as a continuous distribution. This suggests that if
we d r a w a sphere of some appropriate radius R about the point in question, we can
divide space into two regionsoutside the sphere, the dipoles will be described by P
and will contribute E0 to E c , while inside the sphere, we treat them as individuals and
their contribution to E 0 w i l l be written as E 7 . Thus (B-12) becomes
E^ - E + E 0 + E ,
(B-13)
Figure B-2 shows this hypothetical sphere. It is evident that the proper value for R is
not unique since the distinction between "continuous" and "discrete" distributions is
somewhat fuzzy; on the other hand, a reasonable value for R can be decided upon. For
the material outside the sphere, ph = 0 by (10-58), and we see that it is completely
equivalent only to the b o u n d surface charges on the surface of the sphere arising from
the discontinuity of P there. The surface density of these b o u n d charges can be found
from (10-8) as ah P n' = PcosO'. This is exactly the surface charge density of
(10-27) that is illustrated in Figure 10-9. The field produced by it was found to have the
magnitude P / 3 e 0 in (10-37) and we see from the signs of the charge distribution of
Figure B-2 that it will have the same direction as P. Thus E0 = P / 3 e 0 and (B-13)
becomes
P
Ey - E + I- E ,
3e n
Figure B-2. Hypothetical spherical cavity used to calculate the

electric field due to the polarized material outside of it.
(B-14)
550
In the most general situation, E , can be different from zero, but we will confine
ourselves to the large number of cases for which E , = 0. Let us see how this can occur.
Equation 8-84 gives the electric field produced by a point dipole located at the origin.
If, instead, the dipole p^ is located at a point r k with respect to the origin chosen at the
so that the contribution of p^ to
center of our sphere, we can replace r in (8-84) by
the field at the center will be
1
E.
P*
4 77C,
3(p* Tk)xk
(B-15)
When we sum (B-15) over all the dipoles within the sphere, we get E ; . Doing this, and
noting that in a uniformly polarized material all molecules have the same average
dipole moment, we can let
= (p) and we get
E,
:)
 EV t + 3 E
k
k
k
477f
'
(B-16)
The x component of this is

Etx =
1
47Ttf
z x
/ \\ ^^
/ \ v 1 x k
~ \ ^ yk*k
kk
+ 3 </>,>
<#>,> + 3 < / > x > E - 5 + 3 (Py>T.-~r
k
(B-17)
If the dipoles are distributed isotropically, then for every molecule with coordinates
(X, y, z) there will be one with coordinates ( JC, y, z) or ( x, y, z) and so on.
Since there will actually be a large number of molecules within the sphere, the terms in
the last two sums will be positive and negative with equal likelihood for a given
magnitude of rk so that
z x
ykxk
kk = o
k
k
=E
Similarly, for a given rk, the squares of the coordinates will appear with equal
likelihood so that on the average (xk) = ( y k ) = (zk) and rk = (x%) + ( y l ) + ( z l )
= 3 ( x l ) . Since this will be true for each rk, when we sum up over the large number of
molecules, we get
Z 3 -
Er
k
so that (B-17) becomes E,x = 0. Similar results will be obtained for the y and
components. Thus Ey = 0 and (B-14) becomes simply
E +
E,
3cf
(B-18)
The reasoning that led us from (B-16) to (B-18) would be applicable if the dipoles were
distributed on a simple cubic lattice. Similarly, E 7 would be zero if the positions of the
dipoles were completely random as would be the case in a liquid or a gas. Hence there
are a great many situations for which we can expect (B-18) to be suitable.
If we now combine (B-18), (B-9), and (B-4), we find that
P = Afa|E +
I
o)
so that when we solve this for P and use (B-l) we find that
(Nct/tp)
=
X
' ~
3e
1 -
(NA/3<0)
(B-19)
(B-20)
B-1
S T A T I C ELECTRIC PROPERTIES
551
which relates the macroscopic susceptibility to the microscopic polarizability a and to

the number density N. Conversely, if we write xe = Ke ~ 1 by (10-52) and solve (B-20)
for a, we get
3c0(Ke-l)
a
(B 21)
N(KC + 2)
"
which is known as the Clansius-Mossotti relation. Since a is a constant, (B-21) tells us
that if we vary the density of the material the dielectric constant KE will also vary in
such a way that the quantity on the right-hand side will remain constant. On the other
hand, one can use measured values of KE to evaluate a; if this is then combined with
(B-10), one can find the atomic radius a. Values of a obtained in this manner agree
reasonably well with those obtained by other means.
Now let us suppose that the molecule has a permanent dipole moment, that is, p0 of
(B-5) is different from zero. We also considered this possibility in Section 10-1 and
concluded that, in the presence of a field, the torque on p0 described by (8-75) will tend
to align the permanent dipoles in the direction of the field. This tendency will be
opposed by randomizing processes like collisions associated with temperature agitation
of the molecules. In the final equilibrium state, however, we still expect there to be a net
dipole moment in the direction of Ep so that (p) = 0, but we also expect that (p) < p0.
In such a case, (p) can also be expected to have a dependence on the absolute
temperature T, and this, in turn, will make the susceptibility depend on temperature in
addition to any variation associated with a change in the number density N with T.
In order to find the average dipole moment for equilibrium conditions at a given
temperature, we must involve a result from statistical mechanics that says that the
probability of finding a system in a given quantum state of energy Wm is given by
e Pwm
(B-22)
m
where fi = 1/kT and k is Boltzmann's constant equal to 1.38 X 10~23 jouIe/K. The
denominator given in (B-22) is needed so that the total probability of finding the
system in some state is unity, that is, so that E m
= 1.
If \p is a dynamical variable of interest whose value in the m th state is xpm, then its
average value will be given by the usual definition of an average as
Pwm
(4*)
rn
y e~^Wm
*i ^
m
(B-23)
Strictly speaking, (B-22) and (B-23) apply to the quantum states of a single system of
interest. It can be shown that if the system consists of a collection of independent, or
nearly independent, molecules, then these results can also be applied to a single
molecule where Wm is now the energy of a molecule and ( i f / ) is the average value for a
given molecule. Furthermore, it can also be shown that in the limit where classical
mechanics is applicable, the sum in (B-23) can be replaced by integrals over the
momentum components and position coordinates. Thus, in rectangular coordinates,
(B-23) would become
f d v x dvy dvz dx dy dz
<+> = i f e (iw dvx dvy dvz dx dy dz
(B-24)
552
since dpx = d(m0vx) = m0 dvx, and so on, and the factors m30 would cancel from the
numerator and denominator. The integrals will generally be sixfold integrals with
appropriate limits; of integration, and, in order to use (B-24), we need to know both W
and ip as functions of (vx, v , vz, x, y, z). [For example, if we were dealing with an ideal
gas in which interactions between the molecules could be neglected, then W would be
simply the kinetic energy given by (A-8) as \m0v2 so that (B-22) would give 0 ~
2
^ dxciy
which is the familiar Maxwell velocity distribution funce-m0v /2kT
tion.]
In general, W will include potential energy as well as kinetic energy. Of particular
interest to us here is the interaction or orientational energy of a dipole in an applied
electric field; this is obtained from (8-73) as UD = pQEp cos 6 since now we are
dealing with a permanent dipole p0 in a polarizing field Ep and the angle between them
is now being written as 6. Then if we let W' be all of the rest of the energy, we can
write
W=W'~
p0Ep cos 0
(B-25)
and assume W' to be independent of 0. The quantity we are interested in here is the
component of p0 in the direction of
so we let \p = p0 cos 6. If we write the volume
2
element as dxdydz = r sin OdrdOdcp and insert pQ cos 6 into (B-24), we see that the
integrals involving W', r, <p, and each component of v will cancel from the numerator
and denominator and we will be left with
f p 0 cos de PPoE p cos e sin 6 dd
j1 p ^ e ^ ^ dfi
( p0 cos#) = --pn
=
H
( ePPaEpco&e sin 6 dd
[
dp
J
>o
-i
(B-26)
with the use of (2-22). Finally, if we let

7 = t>p0Ep =
(B-27)
we can write (B-26) as

(p0cosO)
ad
=Poln
i
du
/ 2z.s sum
i n h ^y\ \ =
fri
J ey,xdfi = Po In I
I
//pPo&
\ ,
0E
(B-28)
where
L(y)
1
= coth y
y
(B-29)
is called the Langevin function and measures the ratio of the actual average component
of p0 to its maximum value p0. The function L(y) is shown as a function of y in
Figure B-3. We see that L(y) - 1 as y becomes very large, that is, for large polarizing
fields and/or very low temperatures. In this case ( p 0 c o s d ) p0, which corresponds
to almost complete alignment of the permanent dipoles with the polarizing field. The
material is then said to be saturated.
The general behavior of ( p 0 c o s O ) is therefore not linear in Ep as we might have
expected. For very small values of y, corresponding to small Ep and/or high temperature, we can expand L{y) in a power series and we find that
y
- 37
(y 1)
(B-30)
Combining this with (B-28), (B-27), and (B-4), we find the magnitude of the polariza-
B-1
Figure B-3.
S T A T I C ELECTRIC P R O P E R T I E S
553
The Langevin function.
tion to be
P = N(p0cosO)
Npl
(B-31)
which is linear in E . The corresponding polarizability is

Po
(B-32)
3 kT
and is inversely proportional to the temperature. The result (B-31) is known as Curie's
law for dielectrics.
We can make some numerical estimates to see whether (B-30) is a reasonable
approximation. The value of p0 will be typically of the order of magnitude of the
electronic charge times some fraction of the radius a of a molecule. Since a is of the
order of 10" 10 meter, we get p0 ~ 10 30 coulomb-meter. At temperatures around room
temperature {T = 290 K), we find that y
1 provided that Pj p
k 1 /Pq
10
volts/meter. Since this is a very large field, we see that (B-31) is generally a very good
approximation.
Under these same conditions, the polarizability as found from (B-32) is ac ~ 10 40
farad-(meter) 2 . The polarizability associated with induced polarization given by (B-10)
is found to be a ~ 10~ 40 farad-(meter) 2 so that these quantities are of comparable
magnitude. Consequently, for polar molecules, it is appropriate to speak of a total
polarizability as the sum of these:
a-,
at = a + ac = a + Po
(B-33)
3kT
The expression comparable to the Clausius-Mossotti relation (B-21) would then be
3e 0 ( K e ~ 1)
= a + Po
N(Kc + 2)
3 kT
(B-34)
which is known as the Debye equation. If the quantity on the left-hand side is evaluated
from measured values and then plotted as a function of 1/T, we see that the result
554
should be a straight line whose intercept will give the induced polarizability a. Similarly
the slope of the line will give Po/3k from which pQ can be found.
Ferroelectric materials that can have a polarization in the absence of a field are most
conveniently discussed after the somewhat analogous magnetic materials are considered, and we do this at the end of the next section.
STATIC MAGNETIC PROPERTIES

Initially we consider only linear isotropic homogeneous magnetic materials so that we
are interested in finding the magnetic susceptibility x m defined in (20-52) as M = x,H.
For practically all materials that can be adequately described by this relation, it is
found that |x m | ^ 1 so that from (20-56) it will be almost equally accurate to use
Hence we write the defining equation for our purposes as
(B-35)
and within the material we generally just use H = B/ju,0 to relate the two fields.
In the last section, we introduced the polarizing field ~Ep as the field at the location
of the molecule due to all of the rest of the charges. Similarly we want to define a
magnetizing field H m (and hence Bw), which is the field affecting the molecule and
arises from all currents except those that we can ascribe to the molecule itself. The
arguments that led to (B-13) can be repeated almost verbatim in this case, so that we
can write down at once that
Hm = H + H 0 + H7
(B-36)
where H 0 is the magnetic field due to the continuous distribution of magnetization M
outside the sphere of radius R shown in Figure B-4. Since there is a discontinuity in the
normal component of M at the surface of this sphere and no free currents within the
material, we can find H 0 at the center either via the bound surface currents given by
(20-8) as K m = M X ft' or from the surface density of poles am = M ft' according to
Figure B-4. Hypothetical spherical cavity used to calculate the

magnetic field due to the magnetized material outside of it.
B-2
STATIC M A G N E T I C PROPERTIES
555
(20-42); the result will be the same either way. In this case, we see from the figure that
am = Mcos 6' and has the signs indicated; the direction of H 0 is thus the same as that
of M. This pole distribution is exactly that of (20-48) for which the magnitude of H 0 as
found from (20-49) is M/3; substituting this into (B-36) we get
M
H - H + + H,
(B-37)
and the corresponding induction would be B 0 = /x 0 M/3. [The apparent discrepancy in

sign between (B-37) and (20-49), as well as the difference in value of B inside the
spheres, arises from the fact that (20-49) was obtained for a sphere of magnetization M
for which M is different from zero inside, while in this case, we are interested in the
field inside an empty sphere within a uniformly magnetized material. This can be seen
by comparing Figures 20-8 and B-4; note that the pole distributions have the same
angular dependence in the two cases but opposite signs with respect to the direction of
Mthe angles 6' are defined differently in the two figures.]
Similarly, the arguments that showed that there are a great many cases for which
E 7 = 0 in the electric case can be equally well used in the magnetic case because of the
complete analogy between (8-84) and (19-55), which give the fields of the corresponding
dipoles. Consequently we restrict ourselves to cases in which H 7 can be taken to be zero
and, combining (B-37) with our approximation H = B / f i 0 , we can write
B = l 0 H - |I 0 (H + j )
= fo(L + J X ) H
(B-38)
as the induction acting on the molecule in question.

It was pointed out after (20-52) that, in contrast to the electric case, the susceptibility
can be negative. Such a material is called diamagnetic and we now show that it is due to
induced dipoles that arise from the altered orbital motion of the electrons when a field
is applied; as we will see, the induced moments are opposite in direction to the
inducing field.
Suppose an electron in an atom is circulating in a circular orbit of radius a and
angular speed w0 in the absence of a field. There will necessarily be a centripetal force
Ft. on it arising from internal interactions within the atom and whose magnitude is given
by
Fc =
a
= meu>la
(B-39)
where m e is the electron rest mass. In the presence of a field BM there will be an
additional force on it given by Fm = -ev X BM according to (14-30). If we assume that
v and Bm are perpendicular, for simplicity, the magnitude of this force will be evBm and
will be directed either radially inward or outward depending on the sense of v relative
to BM. This is shown in Figure B-5 for the case in which BM is out of the page; the
vector angular velocity W is seen to be out of the page in (a) and hence parallel to BM,
while it is into the page and antiparallel to BM in (b). If we assume for now that the
radius of the orbit remains constant, and thus Fc will as well, the new net centripetal
force can be written as
Fc + evBm = Fc eu>aBm = meoo2a
(B-40)
where the choice of signs corresponds to the sign of w as compared to Bm, that is, upper
sign for (a) and lower sign for (b) of the figure. We see that to must be different in the
presence of BM. Substituting for Fc from (B-39), we find that (B-40) leads to
e
co2 Wq = (co co0)(w-+ co0) = + u B
(B-41)
m
556
B m (out)
(b)
(a)
Figure B-5. Forces acting on an electron in an atom in the presence

of a magnetizing field.
It turns out that the frequency change Aw = w w0 is very small even for very large
values of BM. As a result we can approximate w + w0 by 2co0 and replace w by w0 in
the right-hand side; we therefore find that
e
Aw = w w0 = +
(B-42)
2m. m
so that w has been increased for parallel to BM and decreased if to is antiparallel.
The numerical value of eBm/2me is called the Larmor frequency.
Since we assumed a to be constant, a new value of w means that the electron speed
v = o) a has been changed. This means that the kinetic energy was changed so that work
must have been done on the electron. We now have to evaluate this work and see if it is
compatible with the above results. During the process of changing the induction from
zero to BM, there is an induced emf given by (17-6) to be
'ind = $ E m d d s = ~dt
IT
d_K
dt
a2
(B-43)
and corresponds to an induced electric field E i n d , which is clockwise in Figure B-5.

However, because of the negative charge of the electron, the corresponding force
eE i n d will be opposite to the induced field. Consequently, the work done on the
electron for each trip around its orbit will be e Kindl with the same choice of signs as
before. Since the number of times that it circulates in one second is the frequency
w'/27t, where w' is an intermediate value of the circular frequency, the rate at which
the changing induction is doing work on the charge is
dW
-*
u'e
-
1
7
dBm
ea
Again assuming that the radius a is constant, we can set this equal to the rate at which
/2 _
1,
the kinetic energy \m
2,.>
evev
2mea w' is changing so that
d<S
1 2 dBm
mea2u>'2 ' == mea w'- = ~ea w'
dt
2
dt
This shows that du>' = (e/2me)
Bm
Aw = f
in agreement with (B-42).
dBm and thus the total change in w will be

2m,
dBm
2m,
8-2
5 5 7
It is desirable to put these results in vector form. In Figure B-6 we show what is
essentially a side view of Figure B-5; parts (a) and (b) of the two figures correspond,
while (c) shows the direction of BM. Since (a) corresponds to the upper sign of (B-42),
the magnitude of to is increased and we see that A to is in the same direction as BM; the
lower sign of (B-42) indicates a decrease in the magnitude of to and (b) of Figure B-6
shows that Ato is again in the same direction as BM. Therefore we have the general
result that
A
"
2^B
(B_44)
for electrons in definite orbits. Now we need to relate these results to induced dipole
moments.
An electron traveling around either of the circles of Figure B-5 is equivalent to a
current element Ids = q\ = ev by (14-29) and will have a magnetic dipole moment
m. We see from Figure B-5 a that m will be directed into the page and thus opposite to
to; m is also opposite to to for (b) since m is out of the page in this case. According to
(19-27), m = IS = lira2 where I is the equivalent current magnitude. Since the total
charge e passes a given point on the perimeter each time around, and since each circuit
takes a time equal to the period T = 277/TO, the equivalent current is I = e/r = eu>/2m
and therefore m = \ea2co. Combining this with the fact that m and to are oppositely
directed we get
m = \ea2v>
(B-45)
for an electron in a definite orbit.
We are actually interested in the induced moment produced by BM, that is, the
change Am. Calculating Am from (B-45) and using (B-44), we get
e2 2
Ato = - a B m
Am = \ea
(B-46)
which we note is actually independent of the sign of the charge since it is ~ e2. Each
electron in the molecule will give a contribution like this where a2 has the appropriate
value. Now we obtained (B-46) by assuming that Bm was perpendicular to the plane of
the orbit; if BM were along the z axis, we would have a2 = x2 + y2 for the yth electron.
Hence if we sum up over all the electrons and average over all orientations, we find that
the average induced dipole moment for a molecule is
(^m)
\ wv" V 'J / I-m
4^
(<*;>
Aw
w0
Wq
Aw
(a)
(b)
(c)
Figure B-6. Changes in the angular velocity of an electron in an

atom resulting from a magnetizing field.
+ (yf>)
558
and if we multiply this by the number of molecules per unit volume N and use (B-38),
we find the magnetization to be
M=
+ (y,2>) H
4m,
(B-47)
. j
We will soon see that | x m | ^ 1 in this case so that H m H by (B-38) and then a
comparison of (B-47) with (B-35) shows that the susceptibility is
X--Ê(<*/>+<TF>)
e j
(B-48)
and is negative since all terms in this expression are positive. Thus the mechanism we
have been discussing actually does result in diamagnetism.
For an atom with isotropic charge distribution, we can use the result obtained after
(B-17) that ( x j ) = ( y f ) \rf for a given distance r} from the nucleus and when this
is put into (B-48) we finally obtain
PaNe2 __
uaNZe2
^ T < ' >
(B
"49)
where Z is the number of electrons.

As a numerical example, suppose we consider helium at standard temperature and
pressure (0 C and 1 atmosphere) for which TV = 3 X 10 25 (meter)" 3 . Also taking
Z = 2 and ( r 2 ) 10" 2 0 (meter) 2 , we find that xm~ ~ 3 X 10~ 9 . This is about equal
to what is observed, and also justifies our assertion that |x ml
The paramagnetic case for which \ m i s positive arises from the existence of
permanent magnetic dipoles. In (B-45) we found a magnetic dipole moment for an
electron in a definite orbit; if we multiply and divide this by the electron mass m e we
get
morb
2m,/
(m,ae 2)u,-~
2m
(B-50)
where 1 is its orbital angular momentum since mea2 is the moment of inertia of the
particle. The proportionality factor ( e/2me) is called the gyromagnetic ratio. Thus
(B-50) provides a connection between the magnetic and mechanical properties of an
electron and we see that m is always oppositely directed to 1. [If we were to repeat the
derivation that led to (B-45) for an arbitrary charge q, we see that we would get
m
orb = (?/2m ( 7 )l where m q is the particle mass; thus m and 1 are in the same direction
for a positive charge.] If we sum (B-50) over all the electrons in the molecule, we find
the total moment to be
|/- -7
L
2me ^j JJ
2me
m orb
(B-51)
where L is the total orbital angular momentum of the molecule.

Things are more complicated than this, however. It has been shown by quantummechanical methods that in addition to its orbital angular momentum, an electron has
an intrinsic angular momentum (or spin) s and associated magnetic moment m ; the
relation between them is twice that of (B-50):
m
,= -
(-)
(B-52)
If we sum (B-52) over all the electrons in a molecule, and add the result to (B-51), the
B-2
559
total magnetic dipole moment is

m = (L + 2S)
2 me
(B-53)
where S = L^-s . In order for the molecule to have a permanent dipole moment, this
quantity must be different from zero. Furthermore, it can be shown that under many
circumstances (B-53) can be written in terms of a total angular momentum J = L + S
as
m = -g-J
2 me
(B-54)
where g is a dimensionless quantity known as the Lande g-factor; g has a numerical

value of the order of unity and must be calculated by specifically quantum-mechanical
means. Since the numerical values of J can be shown to be multiples of h/2rr, where h
is Planck's constant, it is common practice to measure atomic dipole moments in units
of the Bohr magneton jue = eh/4irme = 9.27 X 10" 2 4 ampere-(meter) 2 .
Our primary interest in (B-54) is that it shows the possibility of a permanent
magnetic dipole moment m 0 . In the presence of the induction Bm, the moment will
have an interaction or orientational energy given by (19-40) as U= m 0 Bm =
fi0m0 Hm with the use of (B-38). If we compare this with (B-25), which led to
(B-28), we see that we could go through exactly the same calculation as before and we
would find the average component of m 0 in the direction of the magnetizing field to be
/
/)\
T i
\
(m
0 cos 6) = m0L I I
(B-55)
where L(y) is again the Langevin function (B-29). Similarly, for conditions for which
L(y) \y, the magnetization will be
M = N(m0 cos 0)
Na0m20
Hm
,
,
(B-56)
with a resultant susceptibility

X,
NnQm20
3kT
(B-57)
which is positive and thus corresponds to paramagnetism. This result is Curie's law for
magnetic materials.
As a numerical example, let us take m 0 to be a Bohr magneton and consider a gas at
standard conditions so that N 3 X 10 25 (meter)" 3 . We find that x m 3 X 10 7,
which is about what is observed. We note that this has a magnitude that is about 100
times greater than the comparable diamagnetic case. This is generally what occurs.
While we expect all materials to have a diamagnetic contribution to the susceptibility,
when they have a permanent moment as well, the corresponding paramagnetic susceptibility is so much larger that it is almost the whole susceptibility. We also note that
XM
1 as we have asserted; consequently, Hm in (B-56) can usually be simply
replaced by H as we see from (B-38).
One interesting property of ferromagnetic materials is that the permanent dipoles
can be aligned to some extent even in the absence of an external B field, and then they
are said to have a permanent magnetization. It is possible to see how this can occur by
means of a simple phenomenological theory introduced by Weiss. We assume that the
magnetizing field can be written in analogy with (B-38) as
Hm = H +\M
(B-58)
560
where we consider only the case where M and H are parallel. The coefficient X is to be
found by experiment; we will see shortly why we do not take X y as we might have
expected from (B-38). If we further assume that we can still use (B-55), then we can
write
M = Nm0L(y)
(B-59)
Unltlr,
y = - ^ r ( +
X M
(B-60)
If we solve (B-60) for M, it becomes

M=\-
kT
\
H
U - \Xn0m0j
X
(B-61)
Equations B-59 and B-61 are simultaneous equations that must be satisfied by M. If
both of these are plotted as functions of y, then the value of M is given by the point of
intersection of the two curves since M must lie on both of them. This is illustrated in
Figure B-7. As T is increased, the slope of the straight line of (B-61) is increased, and
we see from the figure that the point of intersection will correspond to a smaller value
of M.
If H = 0, it is still possible for the curves to intersect and give
0 as shown in
Figure B-8; the magnetization is then usually called the spontaneous magnetization Ms.
As T increases, the point of intersection moves down the curve L(y) so that MS(T)
decreases. We also see that Ms = 0 when the slope of the straight line is greater than or
equal to the initial slope of (B-59) that is |A^m0 according to (B-30). If Tc is this
limiting temperature, then ( k T c / X n 0 m 0 ) = \Nm0 from (B-61) so that
XuriNml
(B-62)
and
Ms = 0
for
T > Tc
(B-63)
NM.
Figure B-7. Calculation of the magnetization according to the Weiss theory.
B-2
r3
T2
561
7,
M,
Figure B-8. Calculation of spontaneous

y magnetization.
Tc is called the Curie temperature. The values of Ms as found by the method of Figure
B-8 are shown in Figure B-9. The overall behavior agrees surprisingly well with what is
observed when one considers that we are using a classical calculation rather than a
more exact quantum-mechanical one.
If y : 1, we can use (B-30) in (B-59), and when we substitute (B-60) into the result
and solve for M we find that
M =
Nn0m20H
3k(T-
Tc)
(B-64)
with the use of (B-62) as well. Thus, well above the Curie temperature, the magnetization is proportional to H\ in other words, the material has a "paramagnetic" susceptibility given by
Xm =
NPoml
3k(TTc)
(B-65)
which is known as the Curie- Weiss law and also agrees quite well with experiment.
Now we are in a position to evaluate the parameter X. For iron, m0 2fie and
Tc = 1043 K. The value of N as found from the molecular weight A, mass density d,
T_
TC
Figure B-9. Spontaneous magnetization

as a function of temperature.
562
and Avogadro's number NA is N = NAd/A 8.5 X 10 28 (meter) - 3 . Inserting these

into (B-62), we find that X = 4700. This is an enormous number and cannot be
understood on any classical calculation since we saw in the considerations that led to
(B-38) that we would expect a value of A = 1. Heisenberg showed that these large
values of X are specifically quantum-mechanical and electrostatic in origin; when the
electron spins are aligned parallel or antiparallel, the charge distribution is altered in
such a way as to favor parallel orientations that will lead to a spontaneous magnetization.
Hysteresis effects that were discussed in Section 20-7 have been found to be
associated with the existence of domains. In each domain, the permanent dipoles are
aligned, but the resultant magnetization of individual domains are themselves not all
parallel. When an external field is applied, there is a combination of rotation of
domains into alignment with the field and an increase in size of domains that are
generally parallel to the field and a decrease in those that are generally oriented in other
directions. The overall result is a magnetization curve of the type shown in Figure
20-19.
Many ferroelectric materials that can have a permanent electric polarization can be
discussed in terms of a Weiss-type theory with an appropriate change of notation.
Thus, instead of (B-18), one could write
= E + ( A e / e 0 ) P and the analogues of
(B-59), (B-61), (B-62), and (B-65) are found to be
P = NPaL(y)
(B-66)
e0E
(B-67)
(B68)
"
Npo
3 c a k ( T - Tc)
Actually, the Weiss theory does not work quite so well for ferroelectric materials as it
does for ferromagnetic ones; when it does seem to be reasonably applicable, the values
of the parameter Xe usually turn out to be much smaller than unity. Ferroelectric
materials also exhibit hysteresis properties and there is evidence for the existence of
domains in them.
B-3
RESPONSE TO TIME-VARYING FIELDS

In Section 24-8 we pointed out that, when detailed comparisons are made between the
predictions of Maxwell's equations and the results of experiment, it has been found
necessary to treat the electromagnetic parameters ju, e, and a as functions of the
circular frequency to rather than as simply constants. We also discussed the conductivity from a microscopic point of view and found in (24-130) that a was a complex
function of frequency. The conductivity expression involved both the inertia of the
charge carriers and a mechanical resistive force proportional to the velocity. This
general dependence of the parameters upon frequency is called dispersion and we now
want to consider its origins in more detail. We restrict ourselves to the case of the
response of a medium to a plane wave that is propagating within it. While this is not
the most general possibility, it will be sufficient to illustrate the basic aspects involved.
B-3
R E S P O N S E T O T I M E - V A R Y I N G FIELDS
5 6 3
We recall from (24-38) and (24-29) that the most general case of a plane wave
propagating in the z direction involves a complex propagation constant
k = a + ip
(B-70)
so that
E = E0ei(kz~u,)
= E0e-^zeâz'ul)
(B-71)
Thus, the real part of k determines the phase velocity while the imaginary part
describes an attenuation of the wave. The basic relation between k and the properties
of the medium was given in (24-135) as
,2 /
)
w2"m 1
ia \
k = w jti(e + i = r
k
H
c
toe
\
W/
\
o/
2
(B-72)
where we have used (10-53), (20-55), and (23-5). We also saw in (24-56) and (24-57)
that it is also possible to describe this situation in terms of a complex index of
refraction that we write here as
JT=
= n + if]
(B-73)
where n is the ordinary index of refraction [written in (24-57) as n'] and where rj will
describe attenuation through (B-70) and (B-71). Comparing (B-73) and (B-72), we see
that
n + ir\ = Om ,
la
1/2
coe0 .
For simplicity, we assume the material to be nonmagnetic and set
addition, it is nonconducting, a = 0 as well and (B-74) reduces to
n + it] = {K~6
(B-74)
= 1. If, in
(B-75)
We are continuing to write J f as complex even though the only source of attenuation
that we have considered so far has been conductivity; as we will see shortly, a
nonconducting medium can also lead to attenuation because of the existence of
damping of the motion of the charges. We will return to the more general case of a # 0
later in this section.
The form of (B-75) indicates what our approach will be. As in Section B-l, we expect
the incident fields to induce electric dipole moments because the charges will be
displaced from their equilibrium positions. From that we can find the polarization and
then the susceptibility XE frm (B-l); finally KE will be given by 1 + XE a n ^ this c a n be
substituted into (B-75) to find the overall response of the medium.
Because of the very large mass of the nucleus as compared to that of an electron, it
should be a good approximation to assume the nuclei to remain at rest; hence we need
only consider the behavior of the electrons. As a result this treatment is often referred
to as the electron theory of matter. If we let r be the displacement of the electron from its
equilibrium position, we can assume a mechanical restoring force Fm as we did in (B-7);
we write it as Fm = KT= meul r where to0 is the natural frequency of oscillation of
the charge. As we did in (12-36) and (24-125), we also assume a frictional damping
force of some sort proportional to the velocity, that is,
= - v = meyv where y is
now used as a measure of the damping. Finally, the fields arising from the wave
produce a Lorentz force that is given by (21-29) as F, = -e(Ep + v X B m ) where E^
and BM are the resultant inducing fields of (B-18) and (B-38). Setting the mass times the
564
E L E C T R O M A G N E T I C PROPERTIES OF MATTER
acceleration equal to the net force, we obtain the equation of motion

F
net =
e *
= -We"0r ~
m)
"
7 6
We will be considering only media for which n is of the order of unity; in that case,
we know from (24-34) that |B m | = |E | / c since the fields will not differ too much from
those in the incident wave. The ratio of the magnetic force to the electric force will then
be approximately
F
mag
-êlec
EVBM
- 1
C
and we can simplify (B-76) even further by neglecting the term v X Bm. The steady-state
displacement of the electron will then be parallel to E . The problem then reduces to a
one-dimensional one and if we let x be the displacement, we can write (B-76) as
mi
d2x
dF
+ v
dx
Tt
"x I " ~eE'"
-eEpel""~""
(B- 7 7 )
for a wave propagating in the z direction. Now the variation in kz over the molecule
will be of the order of 2ira/\ where a is the radius of the molecule. We now assume
that a A; since a ~ 10" 1 0 meter, this is satisfied even into the ultraviolet where one
begins to need a quantum-mechanical description anyhow. In other words, the electric
field will be approximately constant over the volume of the molecule located at z0. If
we absorb the constant factor e'kz into the amplitude Ep o, all that remains is the
sinusoidal time variation and (B-77) becomes
(B-78)
which is recognizable as the equation of motion of a forced damped harmonic
oscillator.
The only solution of (B-78) of interest to us is the steady-state displacement that we
can find by assuming a solution of the form x = x0e~'"' where x 0 is a constant. When
this is substituted into (B-78), we obtain
~(e/m
=
/
,ot
x = x0e-
co
)Ep0
e~,o3t
P
t
(B-79)
z'yco
The corresponding induced dipole moment in the direction of the field can be found
from the one-dimensional form of (B-3), remembering that here x is the displacement,
and is
p ex= p0e
lut
(ez/me)Ep0e-iu'
= 5
^
COQ co / yco
(B-80)
since we are assuming the nuclei to remain stationary. If there are n0 electrons of this
type in the molecule, their total contribution to the dipole moment will be
n
oP
n0(e2/me)Ep
2
1
:
COQ co /yco
(B_81)
since EP = Ep0e~'<. This result can be easily generalized to the situation in which there
is more than one type of electron since we do not expect them all to be in identical
situations. Thus if we let n k be the number of electrons in the molecule that are
B-3
565
characterized by the natural frequency cok and damping constant yk, we can sum (B-81)
over all types to get the total induced dipole moment:
2
nk(e
/me)
1
<P) = ,E I " "
.
k "k- w
= ocEp
(B-82)
where a is the polarizability according to the definition (B-9). We see that now a is
both complex and frequency dependent. Before we go on, we can compare this with the
static case (co = 0) and we find from (B-82) that
nke"
^
(<) - E
- E
l w
e k
^ e2
= E ^
where, in the last step, we went from a sum over types k to a sum over individual
electrons j. (The total number of electrons is Z, = Hknk.) Comparing this result with
(B-9), we see that they are the same if we recall that we are assuming the nuclear force
constants to be so very large that 1 /Kt 0.
The derivation of (B-21) from (B-20) still applies, of course, and if we combine
(B-21), (B-82), and (B-75), we get our basic result that
(n + i-g)2 - 1
Na
TV
(n + / T J ) 2 + 2
3e0
3c0
nk(e2/me)
k
u>\ - <o2 - iyku
where N is the number of molecules per unit volume. (We remember that the sum is
over a single molecule and hence is independent of N.) We could now find n and r\ by
separately equating the real and imaginary parts of each side of (B-83). As a qualitative
check on our results, we see that if all the yk = 0, the right-hand side is real and TJ = 0;
thus, as expected, it is the damping terms in the equations of motion that lead to
attenuation and energy absorption from the wave. Similarly, ij = 0 when to = 0; hence
there is no attenuation in the static case.
As a special case, let us consider a perfectly transparent material for which TJ = 0.
Then, for a fixed frequency co, (B-83) becomes
/n2 1 \ 1 =
~~2 ^ 77
\n + 2 J N
(w) =
3e 0
const.
(B-84)
v
'
This result is known as the Lorenz-Lorentz law and describes how the index of
refraction varies with the density of the material. It is found to be extremely accurate
for many materials, even to being approximately correct for the case in which a liquid
changes to its vapor phase. We note that in the static case (co = 0) where we can replace
n2 by Ke, (B-84) reduces to the Clausius-Mossotti relation (B-21). Instead of continuing
with the general result (B-83), we turn to a somewhat simpler situation.
For many cases, particularly gases, a turns out to be so small that n 1 and 17 = 0.
Consequently, \n + it]\ = 1 and we can approximate the denominator on the left
of (B-83) by 3. Then (n + / T J ) 2 = 1 + (Na/e0) and n + irj [1 + (Aâ/e 0 )] 1/2
1 + (Na/2e0) so that
N
nk(e2/mt,)
n + h = 1+ 2
y
2
2
*o k uk~ " ~ 'Ik"
(B-85)
We can now find n and TJ separately by multiplying the numerator and denominator in
the sum by (co^ - co2 + iyku) and equating real and imaginary parts on each side; the
566
results are
n =
2c
o k (u>l - co2) + (y^co)
N
nk(e2/me)ykw
1 - T
- E , - V - ,2
,
>2
2e
o k (oil - ) + (Y*w)
(B-86)
(B-87)
We can see the general nature of the frequency dependence more easily if we assume
only one type of electron with one set of constants n 0 , Wq, and y; then these results
reduce to
n
1 -
, -
Nn
o(el/lme*o)Wo2
~ 2" 2 )
/ 2
2\
t ^
( - ) + (yco)
(g - (02) + ( y )
(Tk fifi.
(B-88)
(B-89)
which are shown as functions of co in Figure B-10. The frequency region around co0
where n and TJ are both changing rapidly is called the region of anomalous dispersion.
In the general cases (B-86) and (B-87), the curves of n and TJ will consist of a
superposition of ones like those shown with anomalous dispersion near each of the coA,
and this is the general type of dispersion curve found for gases. By fitting curves like
these to experimental results, one can evaluate the quantities nk, o)k, and yk.
If we look again at (B-83) and assume TJ = 0 for simplicity, we see that in the more
general situation we will observe anomalous dispersion about the same frequencies, oik,
except for the quantity (n2 1 )/{n2 + 2) rather than for (n 1).
Now let us suppose that our medium also contains some electrons that are free to
move so that it will have conducting properties as well. We see from (B-74) that we can
handle this by simply adding (/a/coc 0 ) to KE. Thus, if (N + if])nc refers to the values
found in the nonconducting case, we find from (B-74), (B-75), and (B-83) that
Figure B-10. Behavior of the

real and imaginary parts of
the index of refraction near
a resonant frequency.
8-3
5 6 7
when Km 1
,
(n + i'i|) = (n + ir\)nc
1 + (2A r a/3e 0 )
ia
+
=
,
, +
co0
1 - {Na/3e0)
w0
ia
(B-90)
An electron that is free to move about corresponds to one not subject to a restoring
force so that cog = 0 for it. Thus its equation of motion as found from (B-78) will be of
the form
( d 2x
dx\
This is just the equation (24-127) that we previously used to find the conductivity where
= mey0. Consequently we can use the value of a given by (24-130):
a o(co)
t
1 - i(a0meu/N0e
ry
)
(B-91)
where N0 is now used for the number of free electrons per unit volume and the static
value of the conductivity a 0 is
N0e2
o ~ ~~7
N0e2
(B-92)
Jo e
Substituting (B-91) into (B-90), we obtain

1
/
. x2
(n + i-n) =
+ (2Na/3e0)
i(a0/ue0)
7
1 - i(o0meu/N0e
1 - (Na/3t0)
rv
)
(B-93)
as the complete expression from which to find n and t] separately. [Remember that a
as given by the sum in (B-83) is complex.]
For our further considerations, it will be sufficient to restrict ourselves to the case in
which \u\ = |Na/t{)\
1 so that the first term on the right-hand side of (B-93) can
be approximated by [1 + (2w/3)]/[l - (u/3)] = [1 + (2w/3)][l + (w/3)] = 1 + u =
1 + (Na/c0). Then if we multiply the numerator and denominator of each term in the
sum of (B-83) by the complex conjugate of its denominator, do a similar thing to the
last term on the right of (B-93), multiply out the left side of (B-93), and then equate real
and imaginary parts on each side, we find that
,
n
N
nk(e2/me){u>2 - co2)
-7) =1+
L-
-2
;
-J
0 k (o>l - W2) + (yku)

?
2n
= T
e
(co2 -
e 2 / m
CO 2 ) 2 +
k 0 >
(pqAÔ)
(yku)
(a2me/N0c0e2)
:
TTi
1 + [o0meu/N0e
)
(B-94)
( B 9 5 )
2 2
+ (a0me0)/N0e )
which now can be solved for n and TJ individually if desired. We will not do this, but
mostly content ourselves with some general remarks about these results.
Let us first consider the low frequency case, which we take as corresponding to
0 o / c o e o 1 , that is, the "good conductor" (Q
1 ) , which we discussed in detail
immediately following (24-70). Here the frequency is much lower than any natural
resonant frequencies of the bound electrons and their contribution to (B-94) and (B-95)
is negligible. The contribution of the conductivity to (B-94) is also small for small co
because the numerator is constant and small; on the other hand, the conductivity term
in (B-95) is very large. Thus these expressions take the form n 2 - r j 2 = l + 5 where S
is a very small quantity while 2nt] ( a 0 / c o e 0 ) 1 . This can only be the case if n and
5 6 8
TJ a r e b o t h large a n d a p p r o x i m a t e l y equal; thus n i] - ( A 0 / 2 w e 0 ) 1 / 2 . T h e n we see

f r o m (B-73) a n d (B-70) that the real a n d imaginary p a r t s of the p r o p a g a t i o n c o n s t a n t
are a p p r o x i m a t e l y equal a n d given by a - ft - ( c o / c ) n = ( ^ / x 0 a 0 w ) 1 / 2 , which is exactly
w h a t we f o u n d b e f o r e in (24-75) for the c o r r e s p o n d i n g case.
A s the f r e q u e n c y increases i n t o the i n f r a r e d a n d visible p o r t i o n s of the spectrum, it
b e c o m e s equal to at least s o m e of the n a t u r a l frequencies of the b o u n d electrons. T h e n
their r e s o n a n t c o n t r i b u t i o n s b e c o m e i m p o r t a n t a n d c o m p a r a b l e to or greater than that
of the free electrons and one has to deal with the very c o m p l i c a t e d general expressions
(B-94) a n d (B-95).
EXERCISH]
B-l The susceptibility of helium at one atmosphere and 140 C is x e = 6-84 X 10" 5 . Find the
polarizability a and use the result to estimate the
radius of a helium atom.
constant. Solve the equation of motion and show

that the solution describes a precession of m
about z, that is, m sweeps out a cone of constant
angle with an axis coinciding with the z axis.
B-2 For air at 0 C and one atmosphere, KL, = What is the angular frequency of the motion?
1.000590. Use the Clausius-Mossotti relation to What is the sense of rotation of the projection of
m on the xy plane when viewed from a point on
find Kc at 100 atmospheres and the same temper
ature. Find the percentage difference between your the positive z axis?
result and the measured value of 1.0548.
B-9 The index of refraction of water at 15 C
B-3 Integrate the equation of motion dl/dt =
torque, where the torque is due to the induced
electric field given in (B-43), and thus show that
the induced diamagnetic moment is again given
by (B-46).
B-4 It is found that x , ~ 2.2 X 10 5 for
aluminum at 20 C. Assume that this is completely paramagnetic in origin and find the corresponding permanent dipole moment m{}. How
many Bohr magnetons is this? [Aluminum has a
density of 2.7 grams/(centimeter) 3 and an atomic
weight of 27 grams/mole.]
B-5
Verify (B-66) through (B-69).
B-6 Show that the Weiss theory of ferromagnetism violates the third law of thermodynamics
by showing that (dMs/dT)T=0
= 0.
B-7 The susceptibility of barium titanate
(BaTiO,) is X e = ( 1 7 x 1 0 5 ) / ( T - 393) for T
sufficiently large. Use this to estimate Af for this
material.
B-8 Beginning with (B-54), show that the equation of motion of the magnetic dipole can be
written in the form ( d m / d t ) = (ge/2me)m
x B . Now assume that B = B0z where B0 is a
for a particular optical frequency is 1.3337; its

density is 0.9991 grams/(centimeter) 3 . Use the
Lorenz-Lorentz law to find n for water vapor at
0 C and one atmosphere by treating the vapor as
an ideal gas. Find the percentage deviation of
your result from the measured value of 1.000250.
B-10 Assume y : w0 and use (B-88) and (B-89)
to show that in a region of anomalous dispersion,
the maximum and minimum values of n occur at
frequencies where TJ is half its maximum value.
Find the relation between y and the half-width
Aw of TJ, that is, the frequency difference between
w0 and the frequency where TJ has half its maximum value.
B - l l Since the induced dipole in (B-80) is oscillating in time, it will radiate energy; this is called
scattered energy. Find the total rate at which p is
radiating by assuming a gas so that E Q can be
taken to be approximately the same as the incident electric field. Show that when to : w 0 , the
radiated power is proportional to 1 /A 4 where A is
the incident wavelength. This is called Ray lei gh
scattering and is used to account for the blue
color of the sky.
ANSWERS T O O D D - N U M B E R E D
EXERCISES
1-3
5x - 3y - z; 32.3, 120, 99.7
1-5
-14.4
x
1-9
-2k
1-13
\abc(a
1-15
l//J
\( x2
T2*
) h
(1/2)
y2
+
+ b + c)
(2k) [(8/21)
- (4k/5)\
1-19 N o ; a/p\ (b/p)z\ ( x 2 + y2ya/2)[{ax

+ (a cos 0 c s i n 0 ) 0 + by
1-21
1-23
vrn
1-25
No
2-1
2-3
2-7
[ ( * - a ) x + j/y]
W'
4,7e
- i.90(q2/4ne0a2)(x
3
z(\pa
L)/[3e0z0(z0
2
2-9
[ ( * + a)2 + y2\
+ y + z)
1/2
- 1
(a2
2-11
3/2
+ L)
\
1 +
llTCr
sin 6 + c c o s 8)r
[(xftf)x+jy]
+ y2\1/2
\ [(x - af
- by)x + (ay + bx)y] + ci\(a
coulomb/(meter) ; \ira A\z
+ 2z2)
2 y / 2
- 2z
3-1 (q/47rc0){[xx
+ (y - a)y][x 2 + {y - a ) 2 ] ( ^ 3 / 2 ) - [xx + (y + a)y]
2 (3/2)
2
[x + (y + fl) r
; Ex = 0 when x = 0 or y = 0
3-3
1.90(<7/47T 0 a 2 Xx + y z)
3-7
{\/2TTt0)[-ax
3-9
- ( o / f 0 ) z for a < z < a\ 0 for \z\ > a
+ (c - b)y]["2
+ (c - b)2]
3-11
(A/47re 0 p)[(sin a 2 + s i n a ^ p + (cos a 2 - cos a ^ z ] ; ( A L p ) / [ 2 7 r c 0 p ( L 2 + p " ) 1 / 2 ]
3-13
pcha2x/2(0x
4-1
pabc/e0
for x > a\ p c h x x / 2 c 0 for x < a
for both
4-3 Xp/2-jT0p, (p < p 0 ); (A + 2-rrpQa)p/2TTt0p,

charge per unit length on the cylinder
4-5
4-7
2Ar3/2r/le0,
( r < a); 2Aa1/2r/lc0r2,

2
p c h p p / 2 c 0 , (p < a); pcha p/2e0p,
4-9 0, ( p < a ) ; Af(p)p/c0a2p,

/ ( P ) = (?
e ' ) + a(ae
4-11
(p 0 < p); - A / 2 w p 0 ; yes, since 2irp0a
is the
(a < r)
(a < p); yes, since A = ira2pch
(a = -xyz
5.3
(q/4ir(0){[x2
+ y2
+ (z -
a)2]
+ x2 + const.
(l/2)
[ x 2 + _y2 + (z + a)2}
(l/2)
};
<#>(z = 0) = 0;
each point is equidistant from the equal and opposite charges
569
570
A N S W E R S T O O D D - N U M B E R E D EXERCISES
5-5
0.710(q/e0a)
5-9
<j>0 =
Q/47re0r;
5-11
p(b
p(b2
5-13
) / 2 t
(o/2c
5-17
(<J/2C
5-21
(X/irt
5-23
) / 3 (
4-
)\z\
)(|
b);
2)ne0a][(n
(p/3e
+
2
)[(3b
3) -
/2) -
\r
( r / a )
2
(a
n+ 2
/ r ) ] , (a
<
r <
b);
a)
z 2 ) _ ( 1 / 2 ) ; because
C =
(o/2e0)'\[z
a);
) \ n [ ( b-
>
r <
)(a
[ Q / A ( n +
r , (r
, (0 <
(Xaa/2ire
5-15
found only for the
<j>
value of x = 0
single
planes parallel to the charged sheet
C;
- 5 a
a ) / ( b +
a)]
q /87re0a
6-1
(b
<#>, =
on surface of radius
Q
<
r <
c,
for
- Q
c ) ; ( Q / 4 I T 0 r b c ) [ b c r(c
and
b,
for
b)],
(a
< r <
b);
a);
c22 =
Awe0(ab
a;
Q/A-nt0r,
(c
(Q/Air(.0abc)[bc
< r)\
a(c
Q/Amt^c,
b)],
(0 < r < a)
6-3
7.08 X 10" 4 farad
6-5
6-9
AA<t>/b
c12 =
c21 =
ATT0ab/(b
7-1
-0.918(^ 2 /c o a)
7-5
iro2a3/At0
7-7
2.25 X 10 32 joule; 6.14 X 10" 6 1 meter
7-9
5/6
7-11
( q
L / A ^
a);
C =
) \ n ( b / a )
1.81 X 10" kilogram
7-17
{ ^ / [ a \ n { b / a ) ] }
2
(p/ATTC. q )(z - \l )~ ',
8-5
Q =
9q\
6qa2,
Qyz
8-7
Q =
XL;
p =
1 5 q a 2 , Qzx
sin2
Qxx
a t (a/9)(Ax
+ sin ay);
( c o s ax
+
Qxx
3
Qxy = XL cos a sin a, Qvz = Qzx = 0; 4>q
a -
1),
Qzz
8-9
=
Q pabc;
p =
ih)Q{2a2
- b
\ Q ( a x
2
-llâ2,
~ ( \ ) X L
(\)XL3(3
21qa2;
=
2
\\L
z > 5/; z > 7.06/
+ 5y + 14z);
qa(Ak
p=
- , p; 0
8-3
Qxx
c a ) / ( b -
7-15
y y
be
2irQL[ln(c2/ab)}-1
6-15
by
c 2 ) , Qvv
+
=
cz)
y y
-%qa2,
Qzz
19qa2, Q
x y
5 y 4- 1 4 z )
( | ) \ L 3 ( 3 cos 2 a - 1),
(XL 3 /247re 0 r 5 )[3(x cos a + j s i n a ) 2 r 2 ]
^(position vector of center);
{\)Q(2b2
c2
) , Qzz
(\)Q(2c2
a2
b2),
Qxy = G)Qab,
Qyz
(\)Qbc,
Qzx
(pQca;
(3Qa2 sin 0/16ir(.Qr3)[sm
6 sin Q = a(p2
sin2cp)(p 2 + z 2 ) " 5 / 2 ; Ep = ap(3p 2 2z 2 )(sin2cp)
2
2
7/2
2
(p 4- z ) ~ , Ev = - 2 ap (cos 2 9 )(p + z 2 )~" 5 / 2 , Ez = 5 azp 2 (sin 2 cp)(p2 + z 2 ) " 7 / 2 ; p3 =
4
( < X / < $ > Q ) sin2(p; p = K(cos2<p)3
8-17
8-19
~(<7p r)/4TTC 0 r 2 ;
No;
r)/477C0/-2; ( V 4 7 R O ' ' 3 ) [ P
F 2 = (3/4TT- 0 /? ){[(P 1 p 2 ) -
3plP2R/A<Tre0R4;
8-21
( q p
-3plp2R/2'7T0R4
(3Q"Q2/6A-!TC0r9)(3r4
30z2r2
- 3(p * r)r]
5(p x R)(p 2 R)]R + (p 2 R)p x + (PI R ) p 2 } ;
+ 35z 4 )
9-1
E l n = 2x + y + z; E 1( = 2x - 4z
9-3
With A = a 0 z / e 0 a 2 , E 2 = ( a + Ax)x + (ft + Ay)y + (y + Az)z
9-5
Potential difference independent of path; follows from (9-22)
10-1
1.2 X 10
coulomb/(meter) 2
P(1 + at), -P,0
10-3
-aP,
10-5
<j>0 = - Pa3/3(.0z2,
10-7
4>,
571
ph = -a(n
Ezo = -2Pa3/3e0z3,
<f>, = Pz/3e0,
Ezi = - P / 3 e 0
+ 2 ) r " ~ \ ah = aa"\ E 0 = 0, E, = - P / e 0 ;
[ a / o ( w + l)](r" + 1 - a" + 1)
10-9
P/3c0
10-11 (a) ph = 0, ah top - - a f t bottom = P; (b) E = z ; = - ( / > / 2 e 0 ) { 2 + (z - L)

[(z - L ) 2 + a 2 ] ~ ( 1 / 2 ) - (z + L)[(z + L) 2 + a2]~(1/2) },(0 < z < L); = ( P / 2 e 0 )
{ - ( z - L)[(z - L) 2 + a2r(1/2)
+ (z + L)[(z + L ) 2 + a 2 ] " ' 1 / 2 ' } , (z > L);
(c) 2 = P/c0; (d) (0) = - ( P / c 0 ) { l - [1 + ( a / L ) 2 ] - ' 1 ^ ) ) . ( e ) | | m M = p / q
a ' L; yes, since like two infinite plane sheets
10-13
for
(10-56) and (10-57)
10-15 Ph = -[(Ke - 1 )pf/Ke] - (P V In Ke)/(Ke - 1);

ar,
~~ 1)5 ^ P
d VKe are perpendicular
P = ( P / A J - (P V In* e)/(Ke
10-17
D = qr/4<nr2, everywhere; inside: E = qr/4iTKeeQr2,

2
E = qr/4ire0r ,
P = (ne l)qr/4uKer2\
outside:
P = 0; (ne - l)q/ice
10-19 D = Xfp/lirp,
E = Xyp/(2wae 0 p" + 1 ), ph = n\f/(2irap"
changed, ph = Ay/277ap
+2
);
n=
1; D is un-
10-21 D = Ayp/2irp, E = Ayp[2n-c 0 p(a + fSz)]~l, P = (a + /Jz - l)Ayp[277-p(a + /?z)] _ 1 , ph

= 0, yes
10-23
D = Q / A ,
everywhere; in vacuum, E = Q/e0A,
P = (Ke - l)Q/KEA;
10-25
P =
0; in dielectric, E =
Q/K/.0A,
_1
C = Kei0A[Ke(d
- / ) + /]
[(Ke2 - Kel)t0A/d)[\n{Ke2/Kel)\-
10-27
1
C = 2nc0L
/ b
In
\ Po
el
10-29
4 77C
10-31
<t>i =
10-33
10-35
- -K 1
,i
0 ice2
D = ft,r/3; E = p 0 r/3/c f e 0 ; 277pâ 5 / 4 5 K /o; - 4 w ( K e - l)a 3 p 0 /3ic e ; <j>0 = p0a3/3c0r,

p0a2\
\ 30
1 /
1 +
r2 .
i -
| ; (2K + l)po = (ice + l)A^> 0 /2K e ; Q = 0(Ke + l)L2A(j>0/2d,

(c0L/2d)(Ke
- \)t(^) [Ke(d
E = (ice + l)A<j>0/2Ked
- t) + tr
11-3 q at ( a, b, 0), q at ( a, b,0) and (a, b,0); (q/4-n(.0)

{[(x - a)2 + (y - b)2 + z 2 ] ~ ( 1 / 2 ) - [(x + a)2 + (y - b)2 + z 2 ] ~ ( 1 / 2 ) +
[(x + a)2 + {y + b)2 + z 2 ]" < 1 / 2 ) - [(x - a)2 + (y + b)2 + z 2 ] ~ < 1 / 2 ) } ;
(q/4wQ)[(y
- b){[(x - a)2 + (y - b)2 + z 2 ] _ 3 / 2
2
- [ ( x + a) + (y - b)2 + z 2 ] ~ 3 / 2 } + (y + 6){[(x + a)2 + (y + b)2 + z2] 3/2
- [ ( x - a)2 + (y + b)2 + z 2 ] ~ 3 / 2 } ] ; Ey(0,y,z)
= 0;
2
2
2
3/2
af(x, 0, z ) = -(qb/2ir){[(x
- a) + b + z ]
[(x + a)2 + b2 + z 2 ] _ 3 / 2 }; opposite to q
11-5
-3p2x/64ire0dA
11-9
(477"c0)
11-11
[(<y/d) + (Q/a)]~, (qz/4m0)[(Q/d2)
\a/n(a2
- q(a/d)3(2d2
- a2)(d2
+ p 2 ); A; A 2 x/47rc 0 a
11-13
2 we () /tcosh" 1 (h/A)];
- wf 0 (A^>) 2 [cosh - 1 (h/A)Y
11-15
-(2e 0 <f>o/^)[sinh(^/L)]" 1
(h2
A2)
<1/2)
- a2)~2]
572
11-17
<J> = 2<#>0 T
nw(2x
( ) s i n ( - | sinh
oddVW
2\
\ a
- 1
nm
sinh
jJL
- \
o,
; E=
7T ~
( ô\
= -1.669
= - 4 | I ( - l ) * " " 1 ) / 2 sinh
a
V a
V
2
' n odd
2
11-19 X" + Y" + Z" = a 4- /? 4- y = 0; X(x) = \ax
4- Cx 4- C ; constant field plus one
varying linearly with x
1
11-21
ra2/(r3
11-23
With Rl = (r2 + b2 - 2rb cos 6f/2,
<t>
+ 2a3)
<7
40 I
E =
R-
sin 0 / 1
qb
[
477C
0n \\ /??
"1
0^(0) =
(qa/47r)[l
11-25
4<
11-27
\h
r =
1 \
( (r
/?-2 /
)][a
(a2/p2)]p
b c o s 6)
-zq(a
b )/4<ni0b
) b cos 0 ]
a ;
c o s qp
cos /n<p; (4<J>0/77-a)x
3/
a \5
ilT.
2Cr
cos
(r
< a):
3 ( r \4
<f> = <>0 + ( z / J ) { ^ - <>0 + (p0d2/
11-31
(Pod/12);
a(d)
= (e0/d)(<j>d
<f>0) -
1 2 e 0 ) [ l - ( z / < / ) 3 ] } ; a(0) = -(eQ/d)(<j>d
Ampere/(meter) 5 ; - 6 3 A \ - l 2 7 r A a 5 / 5 , decreasing
12-3
(3(?r sin #/47ra 3 )q>;
12-5
(o<j>0/a)z
QO)/2TT
12-7
1R/1
12-9
[ax(d - x ) ^ 4- o2x$2l/[at(d
12-11
77WA</>/cosh _1 (D/2A)
12-17
1.51 X 10~ 4 meter/second

-
- x) + a2x]; (o x c 2 -
- <h)Ai(d
IK'z
2
lAn.ll
FI
QH'AU
y-
IT
/>7T sin 2 <p dtp

FIR
/
'o d2 4- a 2 sin2cp
= -y/i0^' l
" "
13-9
fi0I2(x
14-1
( j n 0 / 7 r p ) [ - / s i n a y 4- (I cos a - 7')z]
14-3
( f i o / ^ H A f - O ' - J i ) x 4- (x - x x )y][(x - x^2
+ h[~{y
2
( J 2 + a 2\l/2
)
4- y ) / 4 w a
4- (y - j ^ ) 2 ] " 1
~ yi)* + (x - * 2 )y][(* - x2)2 + (y - y2)2]~1}
jh 0 /'{1 -
(2/TT)
2
14-7
(ix0I'a/2ir)(a
14-9
n0K'x,
a r c s i n [ a 2 / ( L 2 4- a 2 ) ] } ;
2
4- z ) ~
3/2
\iQnI'
( z sin a x 4- a a z )
(0 < z < d)\ 0, elsewhere

2
+ \a2)(z2
14-11
fi0oui[(z
14-13
f)n 0 Pwaz; 0
14-15
qvfiQI'(b
- a)/4ab,
- %) -
\p0d
12-1
14-5
' -2b ^ ") iU) p'(cose)
2er
13-7
r/a
P2(cos 8) + ( ) P 4 (cos 0) - .,
1 / / \2
13-5
'3/2.
4- ft 2abcos6]
11-29
(r > a):
[( b
+ a 2 - 2rZ> cos 0 ] 1 / 2 ;
Ri
cos <p\ cy = 2e0E0
m odd
Rl
; E ( r = 0) =
(-l)(m-1)/2
R2 = [(br/a)2
4irer
(ft / a
E0[l
<f> =
= K sin 2 6
4- a 2 ) " ( 1 / 2 ) | z | ] ; ^jn 0 awaz
horizontal and to the right
~ x) + 2*]
15-1
2TT[1 - z(z2+
5 7 3
a2)~(1/2)]
15-5 An infinitely long straight current NI perpendicular to the plane of the torus and passing
through the center 0.
15-7 B9 = n0Tp/2ira2, (0 < p < a); fi0I/2irp,
(b < p < c); 0, (c < p)
(a < p < by (ft0I/2irp)[(c2
- p2)/(c2
- b2)],
15-9 \n0pchv X p, (0 < p < a); !/i 0 p ch (a/p) 2 v X p, (a < p)

16-3 f(x, y, z) = -(az/y2) - (pz/x2) + g(x, y); ja0J = x[(2az/y 3 ) +
- y[(2y8z/x 3 ) + (dg/dx)]
(dg/dy)]
16-5 \{a
b) = ~xyB\
+ z[(2ax/y3) - (2Py/x3)\, V J = 0
x(a - c) = - \xyB\
\(b
- c) = \xyB
16-7 Bp(p, z) - (3p.0Ia zp/4)(a

+ z )'^ ;
fiz(p, z) - ( M o / a 2 / 2 X a 2 + z 2 r 3 / 2 { l + [$p2(a2
~ 4z2)(a2
+ z2)"2]}
16-9 4TTA/H0I = xln[(a + a + x)(/3 a x)/(S + a x)(y - a + x)] +

yln[(y + a + y)(a - a - y)/(P + a - y)(S a + y)\ where a 2 = (a + x) 2 + (a + y)2,
P2 = (a + x)2 + (a - y)2, y2 = (a - x ) 2 + (a + y)2, S2 = (a - x)2 + (a - y)2; 0
16-11 Az = (fi0I/4ir)[l - (p 2 /a 2 ) + 2In(p 0 /a)], (p < a); (ja 0 //2ir)ln(p 0 /p), (p > a);
^ ; (0) = (JLI0//477)[1 + 21n(p 0 /a)]
16-13 AZ= Ao - (fi0Ip2/4wa2), (0 < p < a);
~ (Mo / / 4 ' 7 )[ 1 + 2 In (p/a)], (a < p < by,
2
2
2
2
A0 - (n0I/47r)[l
+ 2\n(b/a)]
+ [P0I/4TT(C
- fc )][p - b
- 2c2 I n ( p / b ) \ , (b /277)ln[l + (a/d)]; clockwise
17-5
u>B0ab sin(2ut + <p0 + a)>
no
17-7 P = (Ke - l)e0uBpp; Qb = 2m(*e - 1 )e0cofia2/

17-9 4.71 volts
17-11 (a) l/aab] (b) vBdl\ (c) vaabBd\ (d) 5 ohms, 8.25 X 10"4 volts, 1.65 X 10~4 amperes
17-13 3 C [ ( K , - 1)VVB]2/27TR*',
attractive
0
17-15
ôu)B0p sin(oit + a)q>
17-17
1.42 X 10" 4 henrys
17-19
(ix0b/2w)ln[(a
17-21
(fi0N a/2Tr)
+ d)(d + D)/d(a
In [(2b + a)/{2b
- a)]
-1
17-23 LC = /i 0 C(/; = (/lo'AJcosh

18-5 i = i0e~,/fj where T = L/R
18-7
I2(nQl/l6ir)(b2
18-9
b < 1.28 a
- a2y2[(b2
+ d + D)]
(D/2A)
- a2){b2
- 3 a 2 ) + 4 a 4 \n(b/a)}
18-11 F,, = IjI k VjM jk

k*j
18-13 xv0TI'[l - b(b2 - a 2 r ( , / 2 ) ]

18-15
z(i'rTfiQa2b2IjIk/2cA)
18-17 - \n0irnSa2{[a2
18-19 0.504 tesla
19-1
Ima2i
19-3 i0a 2 w
19-7
77ja 0 a 2 ft 2 /4c 3
19-11 tan a2 = - { tanaj

19-13
IabB sin (pi
+ (\l - 5) 2 ]" 3 / 2 - [a1 + {\l + S) 2 ]~ 3/2 }
574
20-1
182 amperes/meter
20-3 Jm = (2M/a)z; K m : 0, (x = 0); - M z , (x = a); 0,(j> = 0); -Mz,

-(M/a)(xx
+ _yy), (z = 0); (M/a)(xx
+ yy), (z = a)
20-5
(y = a);
a in amperes/(meter) 3 , /? in amperes/meter; Jm = 0; K m = (aa2 cos 2 6 + /?) sin (?<p
20-7 fi, = i / i 0 M { ( z + /)[a 2 + (z + f / ) 2 ] " 1 / 2 - (z - /)[a 2 + (z - i / ) 2 ] " 1 / 2 } ,

H, = (Bz/n0) - M; for / : a: 5 z 0 = \fiQMla2(a2
+ z2) 3/2,
= n0Ml/2a = 0,
= A/[l - ( / / 2 a ) ] = - M
20-9
20-13
Jm = 0, K m = 0; B,. = 0, H, = - M , H 0 ( a ) = 0
pm = 0, am top M, am bottom M, H ( M, H 0 0, 1, yes
20-15 4>m = 0, Hz = 0, (0 < z < a); <J>m = ( M / 3 z 2 ) ( z 3 - a3), Hz = - ( A / / 3 z 3 ) ( z 3 + 2a 3 ),

(a < z < b);<f>m = (M/3z2)(b3
- a 3 ), Hz = (2M/3z3)(b3
- a 3 ), (6 < z)
20-17
20-19
Xm Xm,mass^> Xm Xm,molar(d/A)
(b) 5 = | ( x m 2 - X m i)sin2a 1 , 6.3 X 1(T 4 degrees; (c): (i) - 4 2 . 1 , (ii) 44.9
20-21 Outside: Hr = H0 cos 0{1 + [(<cm - 1 ) / ( k w + 2)](2a 3 /r 3 )},

= - H o s i n 0 { l - [(Km - l ) / ( K m + 2)](a 3 //- 3 )}, //,, = 0; inside: H = [3/(ic m + 2)]H 0
20-23
2 I / ( K
20-25
7/^2
20-27
Xm
+ 1); [(icm - l ) / ( * m + l ) ] ( / i 0 / 2 / 4 W ) ; attractive
V2""P> ^ 2
K i
l oI/2ira,
L2 = ic/x0(& - a)l/2na
= 2dgh/^H
20-29 1.30 tesla; 1.44 X 10 6 amperes/weber; 0.374 ampere; // 0 = 1.03 X 10 6 amperes/meter,

Hi = 247 amperes/meter; 0.689
20-31
(/x - ixQ)B(2S/2mi();
134 newtons; 6.63 atmospheres
21-1
\t0<ou>dlp cos o)t{d0 + c/j sin cof)~2<p; 0
21-3
V E = (pf - V P ) / e 0 . V X E = - / i 0 [ ( d H / 3 0 + (dM/dt)],
V H = - V M,
V X H = J/ + e0(dE/dt) + (dP/dt); V D = pf, V X D = -e0(dB/dt)
+ V X P,
V B = 0, V X B = ju0[Jy + ( d D / d t ) + v X M ] ; v D = p / ,
V X D = -fi0e0[(dU/dt) + (dM/dt)] + V X P, V H = - V M, V XH = J / + (dD/dt)

21-7 V E = (Py - E ve)/t, V x E
dB/dt, V B = 0,
V X B = n[J/ + o E + (dE/dt)]
21-9
21-11
2 i
+ (V/n X B)/ju.
dU/dt = qld/e0ira2
S = (ql/2(.0ir a )p\
= d(q2/2C)/dt
z(2/9)fL0a2QM
dI
<9? \
22-3
V 0 + V A + - A Ve
+ - ( V * ) - (V<0 =
22-5
v24> = -P/(r, t)/c; V2A fie(d2A/dt2)
p7
dA
d2A
/xt-y dt
dr
; V 2 A - /no-
e
/
e
e
( d<j>\ 1
V(V A) - fiaV4> - fiev^ j + "(VjO X (V X A) = -jnJ/
22-7
E = V ( V tte) - nQtQ(d\/dt )
= fiJj + /j,ev(d(j>/dt)
= V X ( V X i c ) - (PAoX
= MofoV
(9%/dt)
23-1 [statcoulomb] = (gram) 1/2 (centimeter) 3/2 (second) - 1 ; [abampere] =

(gram) 1/2 (centimeter) 1 / 2 (second) ~ 1
23-5
C = A/Awd
24-1
d2xp/dd T) = 0
24-5
= ira2NkE0 cos 9 sin cot; so that B = const. = B(z = 0)
24-7 Listed in the order Q, v(meters/second), 8(meter), E/cB, 2(radian). For v = 10 2 hertz:
I.39 X 10~ 9 , 1.58 X 10 4 , 25.2, 3.73 X 10~ 5 , tt/4 = 0.785; for 107: 1.39 X 10 4 , 5.00 X 10 6 ,
7.96 X 10" 2 , 1.18 X 10" 2 , 0.785; for 1010: 0.139, 1.48 X 10 8 , 2.70 X 1 0 - 3 , 0.371, 0.716; for
1015: 1.39 X 10 4 , 3 X 10 8 ,1.33 X 10~ 3 ,1.00 - 1.29 X 1 0 - 9 , 3.60 X 10~ 5 ; everyday observation
shows that light penetrates many meters into seawater; the static value for a was used
24-9
E0 = 1005 volts/meter, B0 = 3.35 X 10

2
1/2
1 + ( 1 / 2 Q ) , (Q 1); (1/Q)(1 + \Q2), (Q 1)
[1 + (1/Q )] ;
24-13
z (El + Ep)/2Z
24-15
<f) = E i E ^
24-17
A = 10
24-19
E _ = o(x + i f ) e n k z - u , + )
24-21
(a) 8.97 X 10 9 hertz, (b) 8.97 X 10 5 hertz
24-23
Yes
24-27
In the order asked for: ^ e | E 0 | 2 ,
|{(
meters, v = 3 X 10 10 hertz, a = /? = 60, y = 45
[ 1 - (TOP/W)2]}1/2|E0|2k,
24-29
tesla
24-11
Vu
^(/C2/JU,CO2)|E0|2, 1
1/2
- (uP/u)
2 l 2
= (iil)- [1 - (wP/o>) ] / [l
(u>2p/2u>2)],
|E012[1 -
- (CO2^2)]"1
(a) <2 = 2TT(T/T)\ (b) 2-nS/X = (1 + 0 2 ) 1 / 2 + Q
25-5 (Ej/Ej) x = 2 Z 2 cos 0, [Z 2 cos 0, + /(Z t Ay/c 2 )]~ 1; ( E t / E , ) n = 2 Z 2 cos

z ' ( Z 2 K / k 2 ) ] ' 1 ; ( Z , / Z 2 ) 2 ; elliptical
[Z x cos 0, +
25-7
(b) 0,. = 21.2 or 40.7; (c) ^ = 2.41 2 = 3.62, 9, = 32.8
25-9
(a) a, = a (cos <9,/cos 6,), ft, = ft; (b) P, = |<S,)|aft, P, = |<S,)|a,ft,; (c) T' = T
25-11
E0t/E0i
25-13
T = Re (.E0, //*, ) / R e (E 0J H) = 4TT ( ^ / ^ X V ^ i )
25-15
(c) d9p/ds = nP 1 \ dn,,/dz\
25-17
Ptot = j(")o
26-1
575
= 2ff(fi2/iii)(S2/X\)(1
<Tt = (ikg/k2.)V<?z,
- 0 ; , leads by 45
0; (d) nP(z)/nP(0)
= ana sin# ; [ P (0)z + ana sin#,]" 1
/ , = ( z X * T )/Z,, Zm = (kg/k0)Z
2 1/2
26-3
% = t>[l - ( ./&>) ]
26-5
(A/2) < a < ( A / i / 2 )
= (A0/A?)Z
< v. Like Exercise 24-26 with k -* kg, v - vg, and (fie)~l
- f 2
26-7 (a) (w) = ^jixJ| [2(oj<a/77t?)2 sin2 ( t r x / a ) + cos(2wx/a)]; (b) ( S ) = z\iio>k g

(ci/TT )21j2
sin2 (wjc/a); (c) ( < w = \jx(au/Trv)2 \ H012; S ) > = z^^kg(a/-!r)2\H0\2;
2
(d) V(j = v /vg = vG
26-9
\n0Hl(ua/-TTc)2
sin2 (-nx/a)y
26-11 WiihP = kgz wt: Hz = 0, Ex = E0(kg/k2)(miT/a)cos(mnx/a)sm(n'!Ty/b)smP,

Ey = - EQ(kg/k2.)(tnr/b)
sm(mTTx/a) cos(n7ry/b) sin P,
E. = E0 sin (mirx/a) sin (mry/b) cos P,
Hx = E(j(ui/kj){n-n/b)
s i n ( m i r x / a ) cos (niry/b) sin P,
Hv = EQ(ue/k2)(mTT/a)cos(mirx/a)sm(n'!Ty/b)sm
P
26-13 A = -(i/u)E,
so that with P = kgz - cot: Ax = -(n/k2.)(nTr/b)H0
cos (mirx/a) sin (nTry/b)elP,
Ay = (n/kl)(mir/a)H0
sin (mirx/a) cos (mry/b)
e'p, Az = 0
26-15
26-17
Of = eA^>0[a ln(ft/a)] lcos(k0z
- cot)
26-21 Ex = Ev = 0, E. = 3 sin(77-x/a) sin(7ry/a)e

'",
Hx = - i{7rE3/ufxa) sin (rrrx/a ) cos(7ry/a)e~'",
Hr iÊ^/ufia) cos (irx/a) sin(Try/a)e~,u',
H: = 0;(S) = 0; U = â 3 c| 3 | 2
26-23
TTe = Ez/k2
with E, given by (26-61) and (26-14)
27-1
I = 1.46, 2.21, 2.79 amperes; | VR\ = 5.84, 8.84, 11.17 volts; \ VL\ = 6.16, 3.16, 0.83 volts
27-5
q = CS{ 1 - [1 + (Rt/2L)\e~R,/2L}\
27-7
R = 0
27-11
u)RLI0cos(ut
(2/e)
+ c + <$>)/(R2 + U 2 L 2 ) 1 / 2 where tan<^ = R/uL
576
27-13 /j(r) = - 2%(u+ sinw + f + co_ sinco_/)',

? i ( 0 = i<7o( cos
+ e s W-0; q2(t) = 2%(~cos
M)] and w i = 1 /[C(L + M)]
a = 13 = (uR'C'/2)1/2,
27-15
hi*)
+l +
cos
2<7o(w+ s i n w + / sinco^r);
u
2
-0
where co = l/[C(L
v = ( 2 u / R ' C ' ) 1 / 2 , Z, = (R'/2wC')(
1 - /)
28-1 All fields are zero for t < (p/c). For / > ( p / c ) , and with F = [1 - ( p / c ? ) 2 ] 1 / 2 :
A = z ( / t 0 / / 4 * ) l n [ ( l + F ) / ( l - F)], B = <?(p0I/2vpF),
E = -%HI/2vtF),
2
2
2
S = p(ju. 0 / /47r pfF )
28-3
0.03 volts/meter
28-5
With P = kr-
E =
at:
2 sin P
2 cos P
/cVo
t- -I
r + 0 sin 6
rcos 6
47ren
(kr)
(kr)
2
fi0k <jp0
cos P
sin P
kr
(krY
cos 2 p
k upp
2
r
sin
9
I677-2en
(*r)'
cos 2 P
B =
477
sin 9
S =
0 sin 0 cos 0
28-13
1
k3r3
,5,5 /
FCV
~w
sin 2 P +
Bft = (i)iQ<Jm0b/4nc r)
k3r3
kr
3 3
k r
2 k5r5
cos P +
sin 2 P
sin P
Xkrf
2 cos 2 P
(AT)
(,k
sin20e' '"~"')0, E^ = -(/ju. 0 a) 3 m 0 6/47rc 2 r)sin2ê' <A: ''~ aJ ' ) <P
+ )]
28-15 With E 0 = (ifi0c\I0\/nr)ei[la-a'+^
^ i:
E[ = E 0 cos(j77 sin(p), E 2 = E 0 sin(^7r sin <p)
29-3
15.7 X 10" 6 seconds; 4.66 X 10 3 meters
29-5
Yes
29-7
a. =
With Tj = 1 + (Vv'x/c2): ax = a'JyV,

- (F/c 2 rj)fâ^]/y 2 rj 2
ay =
29-17 With 1 - (JivJcy. f j = /t - (V/c\)(vyfy

(</"/<"') [ ( r f f f / A ) - Vfx\/e
E = E 0 c o s [ k a sin .
2 2
(1/2)
= MosV[l - (V /c )]
, where Mos = M0/[\ - (vl/c2)]1/2
= M0 + [ ( P F ) 0 / C 2 ]
= (inertial rest mass) + (mass equivalent of initial potential energy)
29-27 Fon 2 = (y1?1<72/47r)[y2/?2 + R2 + R2Z]~3/2 X { ( R / e 0 ) + /x0v2 X (v, X R)} where R =
r2 - rx and yx = [1 - ( v \ / c 2 ) ] ( 1 / 2 )
29-29
Outside: B = 0, E = 0; inside: B = yju,0n'/'y, E = -yVfi0n'ri;
29-31
(q2/6ir0c3)[l
A-l
no
- ( t ; 2 / c 2 ) ] - 3 { [ a ] 2 - ([v] X [ a ] ) 2 / c 2 }
2
y = (qE/2m0vl)x ;
A-3 T = i Â</>[ln(6/)l"
c/ = qED2/2m0v\\
tana = qED/m0vl
A-5
mQ = qB2D2/SA<f>
A-7
A/ = 477(/M o y o /^B)sin0 o sin A0; A / / / 0 = 2tan0 o sin A0; A0 < 8.15
A-ll
x = ( i r 2 / 1 2 ) ( m 0 / q ) ( E / B 2 ) , y = 0.134(w 0 y 0 /^fB)
A-13 With A = (qE/m0c)[ 1 - (ugA 2 )] ( 1 / 2 ) : z = 0, y = ( c / 4 ) { [ l + ( ^ ) 2 ] 1 / 2 - 1},

x = ( v 0 / A ) In { A t + [1 + (At)2^/2)
= (v0/A) sinh" 1 (At)
B-l
a = 3.42 X 10~ 41 farad-(meter)2; a 6.75 X 10" 11 meter
B-7
2.3 X 1 0 - 3
B-9
1.000240; 0.001 percent
B-ll
(n0/12nc)(e2/me)2u4\Ep0\2[(<4
- u2)2 + (yco) 2 ]" 1
1,
QC
INDEX
A
Aberration, 504, 528
Acceleration, transformation of, 528
Action at distance, 51, 101, 225, 287
Addition of vectors, 4
Admittance, 458
Advanced potentials, 472
Air, dielectric constant of, 568
Aluminum, density of, 568
magnetic susceptibility of, 346, 568
Ampere, 42, 202, 218
Ampere's law, 218, 292
circuital form, 237
integral form, 237, 322
Amperian currents, 312, 315
Angle:
critical, 409
of incidence, 407
phase, 379, 456
polarizing, 417, 427
of reflection, 408
of refraction, 408
solid, 58
Angular frequency, 268
Angular momentum, 117
density, 361
Anomalous dispersion, 566
Antenna, 487
half-wave, 488
Antiresonance, 459
Area:
of closed surface, 39
cylindrical components, 29
rectangular components, 19
spherical components, 33
as vector, 17
Atomic weight, of copper, 216
Attenuation distance, 384
Avogadro's number, 216
Axis, symmetry, 116
B
Barium titanate, susceptibility of, 568
Battery, 208
energy change of, 106, 291
Beats, 402
Bessel function, 448
Betatron, 269, 543
Biot-Savart law, 225, 226, 247
Bohr magneton, 345, 559
Boltzmann constant, 551
Boundary conditions:
electric field, 136, 267, 354
general form, 354
general scalar potential, 364
general vector potential, 366
and Laplace's equation, 171, 195, 199
magnetic field vectors, 329, 349
normal components, 133
at perfect conductor, 430
scalar potential, 138
tangential components, 134
Bound charge, 143, 151
density, 143, 157
Brewster's law, 417
C
Capacitance, 90
of coaxial cylinders, 96
coefficients of, 95, 108, 168
effect of dielectric, 159
of parallel cylinders, 184
and resistance, 210
of sphere, 91
Capacitor, 90
cylindrical, 96
energy of, 101, 108, 164, 167
in fluid dielectric, 166
in parallel, 95
577
5 7 8
INDEX
Capacitor (Continued)
parallel plate, 94, 103
in series, 95
spherical, 93, 162
Cavity definition:
displacement, 152
electric field, 148
magnetic field, 322
magnetic induction, 317
Cavity resonator, 444
Center of charge, 131
Characteristic impedance, 444, 463
Charge:
bound,143,151
center of, 131
conservation, 40
electronic, 108
free, 151
induced, 175
as invariant, 518
linear density, 45
magnetic, 325
magnitude, 40
negative, 40
point, 40
positive, 40
spherical distribution, 46
surface density, 45, 137
test, 56
unit, 42
volume density, 44
Charge density:
bound,143, 157
free, 151, 157
linear, 45
polarization, 143
surface, 45, 137
total, 151, 157
volume, 44
Circuit:
coupled, 460
magnetic, 342
RL, 451
RLC, 453
series-parallel, 458
Circuital law, Ampere's, 237
Circular frequency, 268
Circulation, 16
Clausius-Mossotti relation, 551, 565
Coaxial line:
characteristic impedance, 444, 466
magnetic fields of, 330
self-inductance, 288, 296, 335
TEM mode, 442
Coefficient(s):
of capacitance, 95, 108, 168
of coupling, 283
of induction, 95, 108, 168
of potential, 89, 90, 93, 168
reflection, 420, 464
transmission, 421, 428
Coercive force, 340
Coercivity, 340
Components:
of area, 19, 29, 33
of vector, 5
Condenser, 94
Conductance, 458
Conductivity, 208
complex, 399
of copper, 214, 388
and permittivity, 400
of plasma, 400
of sea water, 282, 402
of water, 216
Conductor, 83
forces on, 103, 166
good,387
Conservation:
of charge, 40
of energy, 531
of 4-momentum, 517
Conservative field, 68
Constitutive equation, 155, 328, 355, 371
Continuity, equation of, 205, 518
Convection currents, 207, 322, 519
Conversion:
of numerical values, 373
of symbols, 371
Copper:
atomic weight, 216
conductivity, 214, 388
density, 216
Coulomb, 42, 218
Coulomb gauge, 254, 367
Coulomb's law, 41, 84, 152
in dielectric, 160
for magnetic poles, 346
INDEX
Coupled circuits, 460

Coupling, coefficient of, 283
Covariance, 513
Critical angle, 409
Critical damping, 467
Cross product, 9
Curie's law, 553, 559
Curie temperature, 561, 562
Curie-Weiss law, 561, 562
Curl, 15, 27, 35
cylindrical coordinates, 31
rectangular coordinates, 15
spherical coordinates, 33
Current:
Amperian, 312
average, 202
conduction, 207
convection, 207, 322, 519
displacement, 348
eddy, 283
filamentary, 203, 281
free, 321
image, 346
induced, 263
instantaneous, 202
magnetization, 321
stationary, 202
steady, 202
surface, 241
Current density:
Amperian, 315
bound,315
of energy, 357
in l.i.h. magnetic media, 329
magnetization, 315
polarization, 207, 354
surface, 204, 241
volume, 203
Current element, 204
equivalents, 205
forces between, 222
point charge as, 233
Current loop, 304, 308. 484
Current sheet, induction of, 243
Current whirl, 304
Cutoff frequency, 433
Cycloid, 539
Cyclotron, 535
Cyclotron frequency, 532
5 7 9
Cylinder, uniformly magnetized, 318,

323
Cylindrical coordinates, 28
D
D'Alembertian, 365, 513
Debye equation, 553
Degeneracy, 447
Del operator, 14, 36, 512
4-vector, 512
Delta, Kronecker, 115
Demagnetizing factor, 326
Demagnetizing field, 345
Density:
of aluminum, 568
angular momentum, 361
charge, 44, 45
of copper, 216
momentum, 361
Depolarizing factor, 198
Derivative of vector, 11
Diamagnetism, 328, 533, 555
Dielectric, 140, 154
and capacitance, 159
electric field within, 145
force on, 165, 170
internal energy, 164
linear, 155
nonlinear, 154
Dielectric constant, 155
of air, 568
Dielectric sphere, in uniform field, 194
Dipole, 118, 300
electric field of, 119, 131
energy in external field, 125, 306
force on, 128, 306
induced, 140,546
magnetic induction of, 310
magnetic orientational energy, 307
permanent, 140, 312, 547, 559
potential of, 114, 120, 300
radiation, 477, 482
torque on, 127, 128, 130, 308
Dipole-dipole interaction energy, 131,
311
5 8 0
INDEX
Dipole layer, 139

Dipole moment, 113, 300
of water, 167
Direction angles, 6
Direction cosines, 6
Disc, potential of, 81
Discontinuity, surface of, 132
Dispersion, 384, 562
anomalous, 566
Dispersion relation, 382, 400
Displacement, 152, 153
Displacement current, 348
of charging capacitor, 349
vacuum, 354
Displacement of point, 3
Dissipation of energy, 212
Distributed parameters, 449, 462
Divergence, 15, 24
Divergence theorem, 21
Domains, 562
Doppler effect, 504, 528
Dot product, 8
Double layer, 139
Double refraction, 155
Drift velocity, 212, 539, 545
Duality, 362, 483
E
Earth, mass and radius of, 108
Eddy currents, 283
Edge effects, 94, 170, 347
Einstein, 494, 499
Electret, 141, 154
Electric field:
boundary conditions, 136, 267, 354
cavity definition, 148
of circular arc, 56
definition, 51
in dielectric, 145
of dipole, 131
induced, 266
of infinite line charge, 52, 60
of infinte plane sheet, 53, 62
of long cylinder, 57
motional, 272, 494
of moving point charge, 523
nonconservative, 208
of parallel line charges, 77
as physical quantity, 55
of plane circular disc, 57
in rectangular coordinates, 52
of spherical distribution, 63
Electrification by friction, 40
Electromagnet, 343
Electromagnetic system of units, 370
Electromotance, 264
Electromotive force, 208
Electron, charge and mass of, 400
Electron theory of matter, 563
Electrostatic system of units, 369
Electrostriction, 165
Emf, 208
back, 280
induced, 264
motional, 272
self-induced, 280
transformer, 272
Emu, 370
Energy:
of capacitor, 101, 108, 164, 167
of conducting sphere, 103
conservation, 531
current density, 357
of dipole in external field, 125, 306
dissipation, 212
electrostatic, 98
of free charges, 161
hysteresis loss, 340
of ideal solenoid, 288
interaction, 124, 131, 164, 305
internal, 124
of isolated conductor, 101
kinetic, 516, 531
magnetic, 284
of magnetic circuit, 347
orientational, 307
of quadrupole in external field, 126
rest, 515
of self-inductance, 286
of spherical charge distribution, 100
of system of conductors, 100
Energy density:
in dielectric, 162
electromagnetic, 357
electrostatic, 102
magnetic, 288, 334
INDEX
magnetization, 336
polarization, 164
time average, 392
Equation of continuity, 205, 518
Equation of motion, 515, 529, 530
Equipotential surface, 70, 83
of dipole, 130
of linear quadrupole, 122
of parallel line charges, 77
Esu, 369
Ether, 496
Evanscent wave, 419
F
Farad, 42, 91
Faraday, 84, 147, 263
Faraday's law, 263
moving media, 269
stationary media, 266
Ferroelectricity, 562
Ferromagnetism, 338, 559
Field:
conservative, 68
demagnetizing, 345
dipole, 119, 131, 302
guiding, 544
linear quadrupole, 121
magnetic, 322
magnetizing, 554
polarizing, 547
Field point, 37
Filamentary current, 203, 281
dipole moment of, 303, 311
Fizeau, 528
Flux:
electric, 58, 60, 85
magnetic, 251
power, 357
of vector, 20
Flux density, 225
Flux leakage, 343
Focusing, magnetic, 536
Force:
on conductors, 103, 166
on current sheets, 294
density, 359
on dielectric, 165, 170
on dipole, 128, 306
on interpenetrating solenoids, 293

magnetic, 290
on quadrupole, 130
of system of point charges, 43
4-acceleration, 528
4-current, 519
4-force, 514
4-momentum, 514, 517
4-potential, 519
4-vector, 510
4-velocity, 511
Free charge, 151
density, 151, 157
energy of, 161
Free currents, 321
Free space:
permeability of, 218
permittivity of, 42
Frequency, 268, 380
angular, 268
circular, 268
cutoff, 433
cyclotron, 532
Larmor, 556
plasma, 401, 403
resonance, 456
Fresnel equations, 413, 416
Friction, electrification by, 40
G
Galilean relativity, 495

Galilean transformation, 495
Gauge:
Coulomb, 254, 367
Lorentz, 366
Gauge transformation, 254, 366
Gauss' divergence theorem, 21
Gaussian system of units, 370
Guass' law, 58, 60
and conductor, 84
forD, 152
Generator, 275, 276
^-factor, 559
Glass, resistivity of, 216
Gold, optical properties, 423
Gradient, 12, 35
5 8 1
5 8 2
INDEX
Gravitational constant, 108

Group velocity, 402
Guide speed, 439
Guiding center, 532
Guiding field, 544
Gyromagnetic ratio, 558
H
Half-wave antenna, 488
Heaviside-Lorentz system of units, 371
Heisenberg, 562
Helicity, 397, 398
Helium, susceptibility of, 568
Helix:
axial induction of, 234
as particle path, 533
vector potential of, 261
Helmholtz coil, 234
Helmholtz theorem, 37, 152, 253, 323
Henry, 218, 278
Hertz, 268
Hertz vector, 367, 448, 493
High-pass filter, 401, 433
Homopolar generator, 276
Homopolar motor, 277
Hysteresis, 340, 341, 562
I
Image charges, 173
of conducting sphere, 176, 179
of semiinfinite dielectric, 180
of semiinfinite plane conductor, 173
Image current, 346
Impedance, 455
characteristic, 444, 463
of free space, 391
load,461
in parallel, 458
in series, 458
wave, 391
Index of refraction, 378, 563
complex, 385, 563
Induced charge, 175
Induced dipole moment, 194, 546
Inductance, 277
of coil and solenoid, 280
mutual, 278
of parallel rings, 279
self-, 280
Induction:
coefficients of, 95, 108, 168
Faraday's law of, 263
magnetic, 225
of circular current, 229
of current sheet, 231, 243
of helix, 234
of ideal solenoid, 230, 244
line of, 258
measurement of, 345
of moving point charge, 233, 523
refraction of, 346
remanent, 340
of short solenoid, 235
of straight current, 227, 242
of toroidal coil, 246
Induction zone, 478
Inertial systems, 8, 494
Insulator, 387
Interaction energy, 124, 131, 164, 305
dipole-dipole, 131, 311
dipole in external induction, 306
magnetic, 305
magnetization, 336
Interference, 402, 403, 487
Internal energy, 124, 164
Invariant, 501, 508, 510
Ionosphere, 424
Iron, permeability of, 344
Isotropy, 155
Junction theorem, 449

K
Kinetic energy, 516, 531
Kirchhoff's laws, 449
Kronecker delta, 115
Laboratory system, 271
INDEX
Lande g-factor, 559

Langevin function, 552, 559
Laplace's equation, 72, 156, 254, 326,
346
rectangular coordinates, 185
and steady currents, 209
two-dimensional, 200
uniqueness of solution, 171
Laplacian, 15, 36
four-dimensional, 513
Larmor frequency, 556
Le Chatelier's principle, 264
Lecher wires, 466, 468
Legendre polynomials, 112, 191,200
Lenz' law, 264
Lienard's formula, 529
Light, speed of, 369
L.i.h. media, 156
Line charge:
electric field of, 52, 60
potential of, 74, 75
Line integral, 15
Line of force, 70
differential equation for, 79
dipole, 120
parallel line charges, 79
quadrupole, 123
refraction of, 156
Line of induction, equation of, 258
Load impedance, 461
Local field, 159, 198
Loop theorem, 450
Lorentz condition, 365, 519
Lorentz force, 233, 354
in Gaussian system, 370
Lorentz gauge, 366
Lorentz lemma, 404
Lorentz transformation, 502, 509
Lorenz-Lorentz law, 565
Lumped parameters, 449
M
Magnet, 313, 328, 343
fields of, 344
Magnetic charge, 325
5 8 3
Magnetic circuit, 342, 347

Magnetic dipole, 300
field of, 302
of filamentary current, 303
force on, 306
orientational energy, 307
radiation, 482
of solenoid, 305
torque on, 308
Magnetic energy, 284, 288, 333
Magnetic field, 322
boundary conditions, 322, 323, 349,
354
of cylinder, 323
of ideal solenoid, 323
of permanent magnet, 344
of short solenoid, 235
of sphere, 324, 326
Magnetic flux, 251
Magnetic focusing, 536
Magnetic induction, see Induction, magnetic
Magnetic mirror, 538
Magnetic quadrupole, 311, 493
Magnetic scalar potential, 325
Magnetic susceptibility, 328
Magnetization, 313
current densities, 315
currents, 321
energy density, 336
interaction energy, 336
permanent, 328
saturation, 339
spontaneous, 560
Magnetization curve, 339
Magnetizing field, 554
Magnetomotive force, 342
Magnetostriction, 336
Magnetron, 541
Malus, 417
Mass spectrometer, 545
Mass susceptibility, 346
Maxwell, 1,217,498
Maxwell's equations, 38, 60, 250, 348ff
general form, 353
integral form, 355
l.i.h. media, 355
relativistic form, 520
Maxwell stress tensor, 360
5 8 4
INDEX
Method of images, 173

Michelson-Morley experiment, 498, 500,
527
Minkowski force, 514
MKSA units, 368
Mmf, 342
Mode, 433,435,442
Molar susceptibility, 346
Momentum:
density, 361
electromagnetic, 359
as 4-vector, 514
Monopole, magnetic, 250, 298
Monopole moment, 113, 298, 475
Motor, homopolar, 277
Multipole expansion, 110, 297
Multipoles, 112, 297
Mutual inductance, 278
of coil and solenoid, 280
of parallel rings, 279
N
Near zone, 478
Negative charge, definition, 40
Neutral, definition, 40
Newton's third law, 219, 359
Normal vector, 13, 17
to surface of discontinuity, 133
O
Octupole term, 130
Ohm, 209
Ohm's law, 209, 212
Operator:
D'Alembertian, 365, 513
del, 14, 30, 33, 36, 512
Laplacian, 15, 31, 33, 36, 513
Orthogonality, 189, 192, 200
Overdamping, 467
P
Parallel plate capacitor, 94, 103
Paramagnetism, 328, 559
Permanent magnet, 313, 328, 343, 344
Permeability, 328
of free space, 218
of iron, 344
Permittivity, 155
and conductivity, 400
of free space, 42
Phase, 380
Phase angle, 379, 456
Phase velocity, 380
Photon, 528
Plane of incidence, 407
Plane sheet, electric field of, 53, 62
Plane wave, 376, 379
Plasma, 400, 424, 433
Plasma frequency, 401, 403
Point charge, 40
as current element, 233
radiation from, 482, 529
relativistic field of, 523
scalar potential of, 71, 526
systems of, 43
vector potential of, 252, 526
Point dipole, 119
Poisson's equation, 72, 156, 171, 253,
325
uniformly charged sphere, 198
Polarizability, 194, 548, 565
Polarization, 141
charge density, 143
current density, 207, 354
permanent, 154
of wave, 393
circular, 396
elliptical, 395, 420, 439
handedness, 398
linear, 382, 395
Polarization potential, 367
Polarizing angle, 417, 427
Polarizing field, 547
Polar molecule, 140
Pole, magnetic, 325, 346
Position vector, 7
differential form, 12, 29, 32
relative, 7
Positive charge, definition, 40
Potential:
scalar, 38, 68
advanced, 472
INDEX
charged disc, 81
coefficients of, 89, 90, 93, 168
continuity of, 139
dipole, 114, 120
and energy, 79
general, 363
line charge, 74, 75
magnetic, 325
of moving point charge, 526
multipole expansion, 110
parallel line charges, 76
point charge, 71
retarded, 472
spherical charge, 73
vector, 38, 251
advanced, 472
dipole, 300
general, 363
helix, 261
ideal solenoid, 259
multipole expansion, 297
point charge, 252, 526
retarded, 472
straight currents, 255
for uniform induction, 254
Potential energy:
of point charge, 80, 531
and scalar potential, 79
of two point charges, 80
Power factor, 467
Power flux, 357
Poynting's theorem, 356
Poynting vector, 357, 391
Pressure:
magnetic, 295, 337
radiation, 425
Product:
scalar, 8, 9, 511
vector, 9
Propagation constant, 380, 431
Propagation vector, 390
Proper time, 506
Q
Q, 383, 386, 404, 456, 467
and line width, 457
Quadrupole moment, 115
energy in external field, 126
field of, 121

force on, 130
linear, 116, 484
magnetic, 311, 493
nuclear, 117
radiation from, 484
Quality factor, 404
R
Radiation pressure, 425
Radiation resistance, 481, 484, 490
Radiation zone, 478, 479
Rationalized units, 369
Ray, 429
Rayleigh scattering, 568
Reactance, 455
Reciprocity, 491
Recursion relation, 112, 191
Reflection:
angle of, 408
law of, 406
by magnetic mirror, 538
total, 418
Reflection coefficient, 420, 464
Refraction:
angle of, 408
index of, 378, 385, 563
law of, 406
of lines of B, 346
of lines o f E , 137, 156
Relative coordinates, 35
Relative position vector, 7
Relativity:
Galilean, 495
of simultaneity, 502
special, 494
Relaxation time, 214, 296, 389, 452,
466
Reluctance, 342, 344
Remanence, 340
Remanent induction, 340
Resistance, 210
and capacitance, 210
radiation, 481, 484, 490
Resistivity, 210
of glass, 216
Resistors in series and parallel, 215
Resonance, 456
5 8 5
5 8 6
INDEX
Resonance (Continued)
frequency, 456
width of, 457
Resonant cavity, 444
Rest energy, 515
Rest system, 518
Resultant, 3
Retarded potentials, 472
Retarded time, 471
Retentivity, 340
Right-hand rules, 9, 17, 226
RL circuit, 451
RLC circuit, 453
Rowland ring, 338
S
Saturation, 552
Saturation magnetization, 339
Scalar, 3
Scalar field, 12
Scalar potential, see Potential, scalar
Scalar product, 8, 9, 511
Screening, 87
Sea water, conductivity, 282, 402
Self-inductance, 280, 330
coaxial line, 288, 296, 335
energy of, 286
ideal solenoid, 281, 330
toroidal coil, 283
Separation of variables, 185, 190, 200,
378, 435,444
Shielding, 87
Shielding factor, 346
Sign conventions, 10, 133, 240, 264
SI units, 42, 368
Skin depth, 384, 388
Skip distance, 425
Snell's law, 409, 413
Solar constant, 402
Solenoid, 230
dipole moment, 305
ideal, 230
induction of, 244
magnetic field of, 230, 235, 323
self-inductance, 281, 330
short, 235
vector potential of, 259
Solid angle, 58
Source point, 38
Speed of light, 369
Sphere:
capacitance, 91
of charge, 46, 198
conducting in uniform field, 192
dielectric in external field, 194
uniformly magnetized, 319, 324, 326
uniformly polarized, 148, 153, 197
Spherical capacitor, 162
Spherical charge distribution, 63, 72,
100
Spherical coordinates, 31
Spin, 117, 558
Standing wave, 402, 428, 466
Standing wave ratio, 468
Steady state current, 453, 455
Stokes' theorem, 24
Stress tensor, Maxwell, 360
Superposition, 355, 356, 362, 379, 402
Surface charge density, 45, 137
Surface current, 241
Surface integral, 20
Susceptance, 458
Susceptibility:
of aluminum, 346, 568
of barium titanate, 568
electric, 155, 546
of helium, 568
magnetic, 328, 554
mass and molar, 346
tensor, 155
Symmetry, 48, 55, 61, 116
T
TEM mode, 435, 441
TE mode, 435, 436
Temperature, Curie, 561, 562
Tensor:
electromagnetic field, 520
Maxwell stress, 360
quadrupole moment, 115
second rank, 511
susceptibility, 155
symmetric, 116, 511
Terminal velocity, 212
Tesla, 225
Test charge, 56
INDEX
Time constant, 452

Time:
dilation, 503
proper, 506
relaxation, 214, 296, 389, 452, 466
retarded, 471
TM mode, 435, 440
Torque, on dipole, 127, 128, 130, 308
Torus, 246
Transient response, 453
Transition layer, 132
Transmission coefficient, 421, 428
Transmission lines, 461
coaxial line, 466
two wire, 465
Triboelectrification, 40
Trouton-Noble experiment, 497, 500
U
Underdamping, 454
Units:
SI, 42, 368
systems of, 368
Unit vector, 5
cylindrical, 28
normal, 13, 17
rectangular, 5
spherical, 31
Unrationalized units, 369
V
Vector:
addition of, 4, 7
as area, 17, 39
component, 5, 9
curl, 15,27,31,33,35
definition, 3
derivative of, 11
divergence, 15, 24, 31, 33
flux of, 20
gradient, 12, 30, 33, 35
line integral of, 15
magnitude of, 6
negative of, 4
normal, 13, 17
product with scalar, 5

propagation, 390
scalar multiplication, 8
square of, 9
surface integral of, 20
unit, 5
Vector field, 14
Vector potential, see Potential, vector
Vector product, 9, 10
Vector relationships, 34
Velocity:
addition of, 495, 503
drift, 212, 539, 545
group, 402
phase, 380
terminal, 212
transformation of, 528
Velocity selector, 541
Volt, 68, 208
Voltage:
induced, 264
standing wave ratio, 468
Volume element, 21, 29, 32
W
Water:
conductivity, 216
dipole moment, 167
Watt, 212
Wave, 376
damped, 383
evanescent, 419
polarization of, 393
spherical, 471
standing, 402, 428, 466
transverse, 381
Wave equation, 376
time independent, 444
transformation of, 495, 527
Wave guide, 431, 435
Wave impedance, 391
Wavelength, 380
cutoff, 433
free space, 432
guide, 432
Weber, 225, 251
Weiss, 559, 562
5 8 7
C O N V E R S I O N OF SYMBOLS
INEQUATIONS
S y m b o l s r e p r e s e n t i n g essentially m e c h a n i c a l qua ntitie s (length, mass,
time, force, work, energy, power, etc.) arc not c h a n g e d ( n o r are
derivatives"!. T o convert an e q u a t i o n written in the M K S A system to the
c o r r e s p o n d i n g one in the G a u s s i a n system, replace the s y m b o l listed
u n d e r the c o l u m n labeled M K S A by that listed u n d e r G a u s s i a n . T h e
entries can also be used 10 convert a G a u s s i a n e q u a t i o n to an M K S A
one bv going from right to left in the table.
Quantity
MKSA
Capacitance
Charge
C h a r g e density
Conductivity
Current
C u r r e n t density
Dielectric c o n s t a n t
D i p o l e m o m e n t (electric)
D i p o l e m o m e n t (magnetic)
Displacement
Electric field
Inductance
M a g n e t i c field
M a g n e t i c flux
M. a g n e i i c induction
M agnetrzation
Pc r m e a b i l i t y
C
q
p, (a, \)
0
J
J, (K)
Gaussian
4 771,, C
(4 97 ( { ) ) s / y
(4770)1''"p,(0, A )
4 Tro0
(4T7 0 ) 1 / -/
(4T7C u ) 1/2 J,(K)

K
(4T7 0 ) 1 / 2 P
m
D
{4 IT/fi0)l/2m
(e0/4v)l/2D
(40)-,/2E
E
L
H
(4ô
( M o / 4 77 ) j / 2 <$>
(fL0/4v)l'2B
M
u
(457f0)
Polarization
Resistance
Resistivity
Scalar potential
Speed of light
Susceptibility
Vector p o t e n t i a l
Ktil
R
P
9
(P-rZi/)' >/~
r' H
ft(1)
then
(2) \ t . c
(4 77 e C) ) 1 -' - P
(4t7 0 ) lR
( 4 t t 0 ) " 'p
(4T7())
t/2
<]>
c
47r
(4w/MO)1/2M
(1)
then
(2)
P e r m e a b i l i t y (relative)
Peimittivity
/2
X,'(X,)
(w,)/'4T7)1/:A
VECTOR INTEGRAL FORMULAS

(hA ddi Jf V A dr
7s
v
<)A ds = J(V
c
s
(Divergence theorem)
(1-59)
(Stokes' theorem)
(1-67 J
X A) da
<uda = J Vudr
(1-122)
<F)A X da - ~ J{V
(puds
<UA
(112?)
= - j y u X da
da
7s
X A) dr
I A
J'v .
( v )
(1-124)
+
M(V
A)]
d T
<j)B(A da) = j [(A v ) B + B ( v A)] dr
i l i >n\
(1-129;
FORMULAS INVOLVING RELATIVE

COORDINATES
9f( R)
y(R)
5.x
(1-130)
<9.x'
V/(R) = - v 7 ( R )
(1-132)
V A(R) = - v ' A(R)
(1-133}
V X A(R) = - V ' X A(R)
(1-134)
v 2 /(R) ~ v /2 /(R)
(1-135)
vR
= -v'R
=
V
-(i
1
V 2J| iI = V "
K
R
R
(1-M7)
./ 1
R
= o
(i? * 0)
(1-141)
(1-144)

Wangsness Electromagnetic Fields

Uploaded by

Copyright:

Available Formats

Wangsness Electromagnetic Fields

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Wangsness Electromagnetic Fields

Uploaded by

Copyright:

Available Formats

V J- Sh

JOHN WILEY & SONS

Copyright 1979, 1986, by John Wiley & Sons, Inc.

Definition of the Electric

THE SCALAR POTENTIAL

Definition and Properties of the

THE ELECTRIC FIELD

Field of a Uniform Infinite Line

Some General Results

The Multipole Expansion of the

Spherically Symmetric Solution of

The Curl and the Tangential

The Electric Field within a

Uniformly Polarized Sphere

Linear Isotropic Homogeneous

Uniqueness of the Solution of

Current and Current

Definition of the Magnetic

THE INTEGRAL FORM O F AMPERE'S

ELECTROSTATICS IN THE PRESENCE

Separation of Variables in Spherical

Derivation of the Integral

THE VECTOR POTENTIAL

Energy of a System of Free

The Potentials in General

SYSTEMS OF UNITS A GUIDE FOR

The Displacement Current

Origin of Other Systems of

The Multipole Expansion of the

Separate Equations for E

The Laws of Reflection and

E Perpendicular to the Plane of

E Parallel to the Plane of

Magnetic Dipole Radiation

Boundary Conditions at the Surface

Propagation Characteristics of Wave

Static Electric and Magnetic

Static Electric Properties

Static Magnetic Properties

Multipole Expansion for

Static Electric Field

Historical Origins of Special

J. R. Reitz, F. J. Milford, and R. W. Christy, Foundations of Electromagnetic

The following books discuss electromagnetism at a more advanced level:

1-1 I DEFINITION OF A VECTOR

Figure 1-4. The sum of t w o vectors does not depend

a is a unit vector in the direction of A.

Unit vectors for rectangular coordinates.

This point is illustrated in Figure 1-5.

In order to proceed further, it is convenient to refer our vectors to particular coordinate

Figure 1-7. A is the sum of the rectangular vector

we write Ax = Axx, and so on, and the above expression becomes

(A2X + A2y + A2z)

Definition of direction angles.

1-5 THE POSITION VECTOR

which can also be obtained from (1-7) and its analogues.

THE POSITION VECTOR

r is the position vector of the point P.

P'(x', y', z')

Figure 1-12. R is the relative position vector of P