W±- William Henry Hunt

i
c

-
THEORETICAL PHYSICS
 This book is in
THE ADDISON-WESLEY SERIES IN PHYSICS
THEORETICAL PHYSICS

Thermodynamics, Electromagnetism,
Waves, and Particles

by

F. WOODBRIDGE CONSTANT
Jarvis Professor of Physics
Trinity College

ADD IS ON-WES LEY PUBLISHING COMPANY, INC.

READING, MASSACHUSETTS, U.S.A.
 Copyright © 1958
ADDISON-WESLEY PUBLISHING COMPANY, INC.

Printed in the United States of America

ALL RIGHTS RESERVED. THIS BOOK, OR PARTS THERE¬

OF, MAY NOT BE REPRODUCED IN ANY FORM WITH¬
OUT WRITTEN PERMISSION OF THE PUBLISHERS.

Library of Congress Catalog Card No. 54-5728

Second printing—August, 1961

 To
Francis Weston Sears
in appreciation

t
 i

PREFACE

The trend in education is toward a broader, more liberal curriculum for

undergraduate science majors; at the same time our graduate schools find
it necessary to raise the level and increase the scope of their programs.
Beginning graduate students frequently have difficulty with their theo¬
retical courses, partly because of a lack of familiarity with the mathematical
techniques that must be used, and partly due to failure to comprehend the
broad principles underlying physics and the deductive manner of deriving
relationships from these laws.
This volume has been designed for use in junior, senior, or first-year
graduate courses in theoretical physics. Undergraduate science majors
may take a year of general physics followed by intermediate courses in,
say, mechanics, heat, and electricity, or they may take one two-year
course based on a text such as Sears and Zemansky’s University Physics.
They are then ready for two, three, or four semesters of theoretical physics,
taken concurrently with mathematics courses beyond the differential and
integral calculus. The author’s previous volume. Theoretical Physics—
Mechanics, together with this book, cover the known fundamental laws of
our physical world. These laws are stated both in our own language and
in the symbolic language of mathematics. All terms are explicitly defined
and postulates are clearly stated. The resulting derivations are carried
out and explained step by step. Illustrative, worked examples are included
for the sake of interest and to relate the theory to concrete reality. The
rationalized mks-coulomb system of units is developed and used through¬
out. It is presumed that by the time the student is ready to study elec¬
tromagnetic theory he has had some previous course in theoretical
mechanics and so is familiar with vector differentiation and integration,
and the differential equations met in the study of oscillations, elastic
waves, and fluid dynamics.
Certain chapters may be omitted, such as those on statistical mechanics,
optics, and wave mechanics, without serious loss to the other chapters.
The book may be adapted to a semester course, a year course, or, when
combined with theoretical mechanics, a one and one-half to two-year
course.
Although the topics are varied, they are all related. Principles and
methods developed in mechanics are applied in thermodynamics, elec¬
tromagnetism, optics, and wave mechanics, d he theory of electromagnetic
waves leads to the theory of light. The principle of relativity is closely
connected with all branches of physics. The quantum theory has had a
similar unifying influence. Only from a study of all the basic laws of
vii
Vlll PREFACE

physics can one realize the unity of physics and its broad philosophical
aspects, a truly stimulating picture and a liberal education! More ad¬
vanced and specialized study may appropriately follow.
The author wishes to express sincere appreciation to Professor Francis
W. Sears for his careful reading of the manuscript and many helpful
suggestions.

July 1957 F. W. C.
 i

CONTENTS

Part I. Thermodynamics

Chapter 1. Thermodynamic Variables and the State of a System 1

1-1 Introduction. 1
1-2 Thermodynamic systems. 2
1-3 State of a system: thermodynamic variables. 3
1-4 Thermodynamic processes. 4
1-5 Definitions of volume, pressure, and temperature. 5
1-6 Work. 8
1- 7 Heat.10

Chapter 2. The Laws of Thermodynamics.12

2- 1 Internal energy of a system.12
2-2 The first law of thermodynamics.12
2-3 The second law of thermodynamics.15
2-4 The Carnot cycle.16
2-5 Carnot’s theorems.18
2-6 Kelvin’s absolute scale of temperature.19
2-7 Entropy.22
2-8 Entropy changes and irreversibility.27
2-9 Entropy and availability of energy.31
2- 10 Summary.32

Chapter 3. Thermodynamic Relations.34

3- 1 Mathematical statement of the combined first and second laws . 34
3-2 Differentiation in thermodynamics.34
3-3 Some thermodynamic coefficients.37
3-4 Gibbs’ functions and Maxwell’s relations.38
3-5 The Clausius-Clapeyron equation.39
3-6 The Gibbs-Helmholtz equation.41
3-7 Ideal gases .42
3-8 The Joule-Kelvin or porous-plug experiment.46
3-9 Equations of state.50
3-10 Properties of a van der Waals gas.51
3- 11 Gibbs’ phase rule.55
*

Chapter 4. Kinetic Theory.59

4- 1 Introduction.59
4-2 Molecular hypotheses for an ideal gas.59
4-3 Distribution of molecular velocities with respect to direction. . 60
IX
X CONTENTS

4-4 Concept of pressure in kinetic theory.61

4-5 Concept of temperature in kinetic theory.66
4-6 Molecular hypotheses for a van der Waals gas.68
4-7 The law of atmospheres.70
4-8 Determination of Avogadro’s number.71
4- 9 A survey of further applications of kinetic theory.73

Chapter 5. Classical Statistical Mechanics.76

5- 1 Introduction.76
5-2 The state of a gas.76
5-3 Phase space.77
5-4 Fundamental hypotheses.78
5-5 Probability.79
5-6 Boltzmann’s distribution law.81
5-7 Maxwell’s distribution law.85
5-8 Energy distribution.87
5-9 Evaluation of the constants B and /3.87
5-10 Experimental verification of Maxwell’s speed distribution . 90
5-11 Entropy and probability .90
5- 12 The equipartition of energy principle.93

Chapter 6. Quantum Statistics.98

6- 1 Limitations of the Maxwell-Boltzmann statistics.98

6-2 The quantum principle.100
6-3 Hypothesis that particles are indistinguishable.102
6-4 Bose-Einstein statistics.103
6-5 Quantum statistics of a gas.106
6-6 Statistics of a photon gas. Planck’s radiation law .... 108
6-7 Fermi-Dirac statistics.112
6-8 Electron gas.114
6-9 Summary and comparison.117

Part II. Electromagnetism

Chapter 7. Fundamental Laws of Electromagnetism .... 123

7-1 Introduction.123
7-2 General outline.123
7-3 Historical summary.124
7-4 Fundamental concepts.126
7-5 Forces and fields.129
7-6 The fundamental laws of electromagnetism for free space . . 132
7-7 Summary.135
 CONTENTS XI

Chapter 8. Coulomb’s Law. Electrostatics . . 1.137

8-1 Coulomb’s law of force between charged bodies.137

8-2 Electric intensity.138
8-3 Gauss’ law.139
8-4 Divergence and curl of E.141
8-5 Electric potential.142
8-6 Potential due to point charges: multipoles.148
8-7 Potential due to continuous charge distributions.152
8-8 Poisson’s and Laplace’s equations.152
8-9 The theory of images.160
8-10 Conjugate functions.164
8- 11 Energy stored in an electrostatic field.168

Chapter 9. Ampere’s Law. Steady Magnetic Fields .... 171

9- 1 Introduction.171
9-2 Ampere’s law of force between current elements.171
9-3 Moving charge equivalent to a current element.173
9-4 The magnetic field. Definition of B.173
9-5 Magnetic induction and force due to a current circuit. . . . 175
9-6 Magnetic forces on current circuits.177
9-7 The circuital form of Ampere’s law.179
9-8 The curl of B.182
9-9 The magnetic scalar potential.185
9-10 The divergence of B.188
9-11 The vector potential. 189

Chapter 10. Faraday’s Law. Electromagnetic Induction . . . 196

10-1 Definition of electromotive force.196

10-2 Motional electromotive force.198
10-3 Magnetic flux.201
10-4 Faraday’s law.203
10-5 Lenz’s law.208
10-6 General expression for the curl of E.208
10-7 Relation between E and the vector potential.209
10- 8 Mutual inductance and self-inductance.210
10- 9 Energy stored in a magnetic field.214

Chapter 11. Maxwell’s Postulate and Electromagnetic

Equations.218
11- 1 The field equations based on the laws of Coulomb, Ampere, and
Faraday.218
11-2 The equation of continuity.220
11-3 Maxwell’s postulate.221
11-4 Physical interpretation of Maxwell’s postulate.223
11-5 Maxwell’s displacement current.225
Xll CONTENTS

11-6 Maxwell’s electromagnetic equations.226

11-7 The wave equations.226
11-8 Relation between E and H in a plane wave.228
11-9 Solution of the wave equation: waveguides.230
11-10 Relation between the vector and scalar potentials.233
11-11 Retarded potentials.235
11-12 Energy flow and Poynting’s theorem.236
11- 13 Radiation from an accelerated charge.239

Chapter 12. Electromagnetic Properties of Material Media . 243

12- 1 Introduction.243
12-2 Ohm’s law: the relation between J and E.244
12-3 Kirchhoff’s laws.245
12-4 Electric polarization.246
12-5 Polarization charges.248
12-6 Gauss’ law for dielectrics.250
12-7 The postulate relating D to E.251
12-8 Capacitance.254
12-9 Work required to charge a capacitor, and the energy stored in an
electric field.255
12-10 Intensity of magnetization.256
12-11 Circuital form of Ampere’s law for magnetized media . . . 259
12-12 The postulate relating B to H.260
12-13 Boundary conditions for magnetic media.264
12-14 Maxwell’s equations for material media.264
12- 15 Summary of treatment of dielectrics and magnetic media . . 266

Part III. Waves, Particles, and Relativity

Chapter 13. Physical Optics.271

13- 1 Reflection and refraction of electromagnetic waves .... 271

13-2 Energy reflected and refracted at normal incidence .... 274
13-3 Amplitude relations for oblique incidence.276
13-4 Fermat’s principle.278
13-5 Principle of superposition; group velocity.281
13-6 Huygens’ principle.283
13-7 Green’s theorem.284
13-8 Kirchhoff’s diffraction formula.286
13- 9 Derivation of Fermat’s principle based on the modified \
Huygens principle. 292*

Chapter 14. The Special Theory of Relativity.295

14- 1 Inertial systems.295

14-2“ The postulates of the special theory of relativity . . . . . 296
14-3 Space-time and field transformations.297
 CONTENTS xiii

14-4 The Michelson-Morley experiment.301

14-5 Relativity derivation of the Lorentz space-time transformation . 304
14-6 The Lorentz-Fitzgerald contraction.306
14-7 Simultaneity and relativity.307
14-8 The apparent slowing of moving clocks.308
14-9 The transformation and addition of velocities.309
14-10 Relativistic dynamics and the variation of mass with velocity . 311
14- 11 Einstein’s mass-energy relationship.315

Chapter 15. Wave Mechanics.320

15- 1 Introduction.320
15-2 The dual nature of light.322
15-3 The dual nature of atomic particles.326
15-4 The analogy between ordinary mechanics and geometrical optics. 329
15-5 Schroedinger’s equation.332
15-6 Solution of Schroedinger’s equation.334
15-7 Simple harmonic oscillators.334
15-8 The hydrogen atom.337
15-9 Penetration of a potential barrier.339
15-10 The operator approach.342
15-11 The uncertainty principle.345

References.350

Answers to Problems ..352

Index.360

*
 ' ■

,
 Part I
THERMODYNAMICS
Including
KINETIC THEORY
and
STATISTICAL MECHANICS

«
 CHAPTER 1

THERMODYNAMIC VARIABLES AND THE STATE

OF A SYSTEM

1-1 Introduction. Thermodynamics is that branch of physics which

deals with the relations between heat and work. The purpose of the sub¬
ject is to describe the behavior of a macroscopic object, say a liter of air
or 100 grams of water, under a change of temperature, pressure, or phase.
We shall not be interested in listing a great many experimental facts;
we shall, instead, take the theoretical approach and develop a theory to
correlate the observed facts. This means that we must first seek a minimum
number of general laws which summarize the experimental observations,
then we shall express these laws as equations, so that through mathematical
exploitation we can apply them to predict what will happen in any given
experiment. The advantage of this approach is that one does not have to
remember a long list of experimental facts, but just the few general laws.
On the other hand, the theoretical physicist must have not only a certain
mathematical training and dexterity, but also a very clear understanding
of the terms used, terms which are represented by symbols in the mathe¬
matical expressions. Quantities must be defined in terms of operations
that can be performed or clearly pictured. In some ways the procedure
is like learning a new language. First one must build a vocabulary (a
collection of definitions) and learn the rules of grammar (the “laws”),
then one can apply this knowledge to reading the literature and journals
written in the language studied.
In thermodynamics, then, we commence with definitions of the terms
employed—pressure, temperature, heat, work, etc. In Chapter 2, the
fundamental laws of thermodynamics (two in this case) will be stated and
discussed. In Chapter 3, we shall express the laws mathematically and
derive various relations from them. We shall then discuss (also in Chapter
3) the experimental significance of the derived relations.
This was the method followed in the preceding volume, “Theoretical
Physics—Mechanics” (hereafter referred to as Volume I). There the
definitions of velocity and acceleration formed a kinematical preliminary
to mechanics, while the fundamental laws were, of course, Newton’s laws
of motion. After a discussion of the motion of a single particle, the theory
was extended to a group of particles by applying Newton’s laws to each
member of the group and deriving certain theorems, such as the conserva¬
tion of momentum, which apply to the motion of the group as a whole.
2 THERMODYNAMIC VARIABLES [CHAP. 1

When the positions of the particles in a group could be assumed to be

related in some systematic way, as in a rigid or elastic body, it was found
unnecessary to follow the motion of each individual particle. In hydro¬
dynamics the fluid was taken to have a definite velocity, pressure, and
density at a given point; these quantities, and not the positions of the
many particles, were taken as the variables in the discussion.
A knowledge of Newtonian mechanics and the methods just referred
to is a great help in the study of thermodynamics. For one thing we may
turn back to mechanics for the definitions of many of the terms which we
shall introduce—volume, pressure, work, etc.—and for such quantities
we shall use principally the mks (meter-kilogram-second) system of units
defined in Volume I. However, in what is called classical thermodynamics
no hypotheses are made about the structure of matter. The state of a gas,
for example, is not defined in terms of the positions and velocities of the
component particles, but in terms of the pressure and volume, or the pres¬
sure and temperature of the gas. Newton’s laws are thus not directly
applicable and instead we must start with new empirical laws, the so-
called “laws of thermodynamics.” Relations between pressure, volume,
temperature, etc. will then be derived.
It is possible, on the other hand, to make some hypothesis about the
structure of matter (such as molecules in random motion) and then to
apply Newton’s laws to the individual particles. This method follows the
tendency, so popular in the past century, of seeking a mechanistic explana¬
tion of all physical phenomena. This may be done in two ways. First,
the properties of matter in the bulk may be predicted by deriving expres¬
sions such as those relating the pressure and internal energy of a gas to
the mass and speed of its molecules. This is the method of kinetic theory,
discussed in Chapter 4. Second, recognizing the impossibility of following
individually millions of particles in random motion, we may, because of
this randomness of motion, apply statistical methods to find, for example,
what fraction of the molecules of a gas have speeds between v and v + dv.
This is the subject of classical statistical mechanics, discussed in Chapter 5.
In Chapter 6 we shall take up those modifications, based on the quantum
theory, which must be made when classical mechanics is applied to very
small-scale systems.
Through such a combined study of classical thermodynamics, kinetic
theory, and statistical mechanics will the most complete insight into the
subject be attained.

1-2 Thermodynamic systems. By the term system will be meant the

quantity of matter in a certain region of space on which our attention is
to be focused. The boundary surface need not remain fixed. For example,
the system might be the gas contained in a cylinder with a movable piston,
1-3] STATE OF A SYSTEM: THERMODYNAMIC VARIABLES 3

or it might be the air inside a tire jdiat is being inflated. However, we shall
usually deal with systems containing a definite quantity of matter; these
are called closed systems and they are characterized by a constant mass,
although not necessarily by a constant volume. Such systems may ex¬
change energy, but not matter, with the surroundings (i.e., adjoining
systems). Thus 18 kilograms (one kilogram-mole) of water may be frozen,
partly vaporized, or entirely vaporized and still constitute the same closed
system. Henceforth systems referred to will be assumed to be closed
unless otherwise stated.

1-3 State of a system: thermodynamic variables. In mechanics the

position of a particle may be specified by giving its three coordinates,
while the velocity is known when we are given the three velocity com¬
ponents. These quantities together determine what might be called the
state of the particle. In a more complex system with n degrees of freedom,
a set of n generalized coordinates and n generalized momenta may be in¬
troduced to specify the state of the system. In thermodynamics the state
of a system is specified in terms of certain quantities characteristic of the
system as a whole, such as the volume, pressure, temperature, internal
energy, etc. These quantities are called thermodynamic variables or state
variables. They take the place of coordinates and momenta.
When an isolated system is left to itself the temperature and pressure
at a given point are found to approach steady values as time goes on.
When these steady values have been reached the system is said to be in a
state of thermodynamic equilibrium. We shall generally be interested in
systems that are in thermodynamic equilibrium. We postulate that in
such a state the temperature, pressure, and other thermodynamic variables
of a homogeneous system are the same at all points. Should a system con¬
tain more than one phase in equilibrium (e.g., a liquid in equilibrium
with its vapor), then we shall assume the temperature and pressure to be
the same at all points and each phase to be homogeneous.
In the case of a hoop of radius r rolling down an inclined plane, the
position can be given in terms of 0, the angle turned through, or s, the
distance traveled, one being related to the other by the geometrical rela¬
tion s = r6. Thus s or 0, but not both, need be specified, and the body is
said to have one degree of freedom. Similarly, in thermodynamics we find
that if we consider a system containing a certain quantity of a chemically
homogeneous gas, then at a given temperature and pressure there is only
one possible valufe that the volume may have. Thus the volume must be
functionally related to the temperature and pressure, so that only two of
the three are independent. This relation is called the equation of state of
the gas. We may say that, thermodynamically speaking, a homogeneous
gas has two degrees of freedom. When additional variables are introduced,
4 THERMODYNAMIC VARIABLES [CHAP. 1

such as the internal energy, the entropy, etc., each new variable must be
functionally related to any two of the others. One of the problems of
thermodynamics is to find these relations.
I *

1-4 Thermodynamic processes. Any change which causes the thermo¬

dynamic variables of a system to have a new set of values is called a
process. There are two kinds of processes, reversible and irreversible. A
reversible process is one in which the system passes through a succession of
equilibrium states, or states that depart only infinitesimally from equi¬
librium. Such a process must take place slowly in a controlled manner so
that, so far as the system is concerned, each step may be reversed and
the system taken back through the same intermediate states to the initial
state. If these conditions are not satisfied the process is called irreversible.
Reversible processes can only be idealized. Although it is not hard to
picture mentally a reversible process, when it comes to experiment it is
found that all natural processes are irreversible to a greater or lesser extent.
During an irreversible process definite values cannot be assigned to the
thermodynamic variables, but, as the initial and final states are taken to
be equilibrium ones, the thermodynamic variables always undergo definite
changes in their values. These changes in the thermodynamic variables
are dependent only on the initial and final states and not on whether the
process is reversible or irreversible, nor do they depend on what might be
termed the “path” (sequence of events) followed from the initial to the
final state. If the initial and final states are infinitesimally separated, the
changes in the thermodynamic variables may be considered to be exact
differentials, represented by the symbol d (e.g., dp for change in pressure).
The integral of an exact differential is dependent on the limits of integra¬
tion, but not on the path taken between these limits.

Example 1. A gas is allowed to expand to twice its volume at constant

temperature. If the process occurs by opening a stopcock and letting the gas rush
into an evacuated vessel of the same volume as that of the original gas, then the
process is irreversible. If the gas is contained in a cylinder with a movable
piston at the top and the weight holding the piston down is suddenly reduced,
the process is again irreversible. To picture the reversible process one must
imagine the piston to move very slowly, the weight on it being gradually reduced
until the desired final state is reached. In this way we may in our minds replace
an irreversible expansion with a reversible one, the two processes occurring be¬
tween the same initial and the same final states. Although the reversible process
is an ideal one, it is a useful concept in thermodynamics.
In the case of the slowly moving piston we may note that if friction is present
and heat is developed at the walls, then so far as the walls (or surroundings of the
gas) are concerned, the process is not reversible. A system may undergo a re-
1-5] DEFINITIONS OF VOLUME, PRESSURE, AND TEMPERATURE 5

versible process while its surroundings experience an irreversible one. Only if

a slowly moving piston is frictionless will both the gas and its surroundings
undergo reversible processes.

Example 2. A beaker of water is heated from a temperature T\ to a tempera¬

ture T2. If the process occurs by supplying heat from a burner, the bottom
part of the water will heat up first. The reversible process may be pictured as
one in which the beaker is placed successively on a series of plates, each plate
infinitesimally hotter than the previous one, the last plate being at the tempera¬
ture T2, as illustrated in Fig. 1-1. Since this requires an infinite number of steps
it could never be done practically.

Tx + dT T i + 2 aIT

Fig. 1-1. Steps in a reversible temperature change.

1-5 Definitions of volume, pressure, and temperature. We shall first

define some of the more familiar thermodynamic variables.
(a) Volume. Volume is defined in geometry. Since it has the dimensions
of length cubed, volume may be measured in cubic meters (m3, the mks
unit), in cubic centimeters (cm3), in cubic feet, etc. At a given tempera¬
ture and pressure the volume of a substance depends on how much mass
of the substance is taken; if we double the mass we have twice the volume.
Such quantities that depend on the mass taken are called extensive, as
contrasted with quantities like pressure and temperature which are called
intensive. Unless specified otherwise, we shall consider one kilogram-mole
of the substance, this being a mass in kilograms equal to the molecular
weight. (A gram-mole is a mass in grams equal to the molecular weight.)
Volume will be denoted by the symbol V. A process in which V does not
change {dV = 0) is called an isochoric process.
(b) Pressure is defined in mechanics as force per unit area. Pressure is
a scalar, the direction of the force due to pressure depending on the direc¬
tion of the normal to the surface against which the fluid exerts this force.
Pressure may be measured in newtons per square meter (n/m2, the mks
unit), in terms of the height of a mercury barometer (cm of Hg), in atmos¬
pheres, etc. Pressure will be denoted by the symbol p. A process in which
p does not change {dp — 0) is called an isobaric process.
(q) Temperature. This is not defined in mechanics but, like force, it is a
quantity about which we have some sensory idea. We all recognize the
difference between hot and cold. Now we want an exact operational
definition of temperature. For commercial purposes this is done by as-
6 THERMODYNAMIC VARIABLES [CHAP. 1

suming a linear relationship between (1) temperature and the height to

which a liquid rises in a thermometer stem, or (2) temperature and the
resistance of a wire, or (3) temperature and the pressure exerted by a gas
at constant volume. But how can we be sure that any of these relations is
linear? To check back by using a temperature scale so defined would be
to employ circular reasoning. We can compare scales based on (1), (2),
and (3), and we can compare scales using different liquids, wires, and
gases. The results of such experiments indicate that the scales are not all
the same, but depend on the substance chosen. In other words, none of
the above relationships is strictly linear relative to a temperature scale
based on one of the others. However, an interesting experimental fact
appears in the case of (3), the so-called gas thermometer. Call ps/pi
the ratio of the pressures produced by a given volume of gas at two fixed
temperatures, such as the boiling point of water (subscript s) and the
melting point of ice (subscript i) at atmospheric pressure. Suppose that
Ps/Pi is measured for gas A using successively 750, 550, and 350 milli¬
grams of the gas in the bulb of the thermometer. With a smaller amount
of gas the density is lower and the pressures are smaller, but the ratio
Ps/pi drops only slightly and nearly linearly with pi, as shown in Fig. 1-2.
While pi cannot be reduced to zero experimentally, the curve, which is
so nearly linear, can be extrapolated to find the value (1.36609) of ps/pi
when pi =0. If a similar procedure is followed with a second gas B the
result is as shown in the figure, namely ps/pi has the same value when
extrapolated to pi = 0. This ratio appears to be independent of the gas
used. Later we shall see that a gas at very low pressure approximates an
“ideal” gas very closely, and that if the latter could be used in a gas ther¬
mometer it would give us an absolute scale of temperature such as the
Kelvin scale. Hence we shall designate temperature on the gas ther¬
mometer scale (defined below) by the symbol T, the one used later for
the Kelvin absolute scale, and we shall refer to a temperature as so many
degrees Kelvin (°K). The symbol t will be used for centigrade temperature.
We assume, then, a direct linear
proportionality between the pres¬
sure p of a constant volume of gas
and the absolute temperature T,
provided the pressure is sufficiently
low, and write

w- =
■1 /
lim (ir) > (1-1)
Pf-> 0 \Pf'

where the subscript / refers to a _

fixed point, that is, an easily repro¬
ducible state. Figure 1-2
1-5] DEFINITIONS OF VOLUME, PRESSURE, AND TEMPERATURE 7

To complete the definition of the gas thermometer temperature scale

we must add to Eq. (1-1) a second postulate which will fix the size of the
interval that we shall call one degree. The degree interval has been defined
as follows:
(a) Method before 1954- The interval between the melting point of ice,
T{, and the boiling point of water, Ts, at normal sea-level atmospheric
pressure, was arbitrarily set equal to 100°. Then

Ts - T i = 100°. (1-2)

Equation (1-1) defines the ratio of Ts to as

Ts
lim (1-3)
Ti p —*o

The experimental determination of limp _>o (Ps/pi) has been discussed;

the best value is 1.36609 ± 0.0004. When we take

^ = 1.36609 ± 0.0004 (1-4)

i

and combine this with Eq. (1-2), we get

Ti = 273.15° ± 0.03°,
(1-5)
Ts = 373.15° ± 0.03°.

(b) Method since 1954. The temperature and pressure at which ice,
liquid water, and water vapor coexist in equilibrium is called the triple
point of water. On the temperature scale just defined the temperature Tt
of this triple point was determined to be 273.16°K. Since 1954 the interval
between 0°K and Tt has been arbitrarily set equal to 273.16°. This means
that Eq. (1-2) is replaced by

Tt = 273.16°. (1-6)

All other temperatures are given by the relation

£ = lim(4 d-7)
Tt Pt-^o\Pt/

where pt is the pressure of the gas at the temperature Tt and the right
side of Eq. (1-7) is subject to experimental determination.
This currently accepted definition of the Kelvin scale requires only one
fixed point, while method (a) involved two.
8 THERMODYNAMIC VARIABLES [CHAP. 1

The centigrade scale is defined by taking the functional relationship

t = T + b,

where b is a constant, and letting

U = 0°C, ts = 100°C. (1-8)

Evidently (see Eqs. 1-5),

t = T - 273.15. (1-9)

A process in which temperature does not change is called isothermal.

As stated previously, the volume, pressure, and temperature of a gas
can all be measured while a gas is in a given state, and the values obtained
do not depend on how the gas attained the given state (the past history
of the gas does not matter). V, p, and T are thus thermodynamic variables.

1-6 Work. One way in which a system may exchange energy with its
surroundings is through work. If a force F produces an infinitesimal dis¬
placement ds, then the work done is defined as the scalar product F-ds.
For a finite displacement the work IE is /F*ds.
Let F be a force exerted by a system on its surroundings. If IE = /F-ds
is positive, the system does work on the surroundings, while if IE is negative
the surroundings do work on the system. When a gas in a cylinder expands
by pushing a piston up, IE is positive; when a gas is compressed, IE is
negative. The unit of work is the unit of energy. This unit in the mks
system is the joule, or newton-meter.
Suppose that a system is enclosed by a boundary of any shape, as in
Fig. 1-3, and that the pressure of the system is p. Consider a portion of
the surface of area A, so that the force exerted by the system against this
portion of the boundary is

F = pA.

If this portion moves out a distance ds, the work 5 IE done by the system is

5IE = pA ds.

But A ds — dV, the increase in the volume of the system. Hence we have

8W = p d V. (1-10)

It should be noticed that we have written 5 IE, rather than dIE, to

indicate that this is a small quantity which is not an exact differential.
1-6] WORK 9
i

Fig. 1-3. Work done by a substance

in expanding.

In other words, 8W is not the differential of a function of the thermo¬

dynamic variables (or of the state of the system). This will be explained
shortly.
For a finite change in volume the pressure will change and the work
done must be found by integration, that is,

W = fpdV. (1-11)

Since in any reversible process p and V have definite values at any

stage of the transformation, such a process can be represented by a curve
on a p-V diagram, as in Fig. 1-4. Here the shaded area represents the
work 8W done during the small volume change dV. The total work done
is represented by the area under the curve between the state a and the
state b. In going from a to b the work is done by the system (W is positive),
while in the reverse process from b to a the work is done on the system
(IF is negative).
In a cycle a system goes through such a series of changes as to return it
to its initial state, as shown by the curve with arrows in Fig. 1-5. Here the
area shaded represents the work done by the system and the area
shaded V77X the work done on the system. The net work per cycle done by
the system is the difference between these two areas, which is the area
bounded by the closed curve.
10 THERMODYNAMIC VARIABLES [CHAP. 1

If we compare the upper and lower paths between a and b in Fig. 1-5,
we see that the area under the curve (the work done) is greater for the
upper path. Thus the work cannot be expressed in terms of the initial
and final states only. It is then impossible to assign a definite value of W
to a given state.
To summarize:
(1) Work (W) around a closed path is not always zero.
(2) Work (W) depends on the path.
(3) Work (W) is not a thermodynamic function.
(4) BW is not an exact differential.

1-7 Heat. Since the time of Joule it has been recognized that when a
system takes in heat it is simply being given energy. In thermodynamics
we consider that a system may exchange energy with its surroundings in
two ways, (1) through work and (2) through heat. Heat then, like work,
will be measured in terms of the energy unit, the joule. In calorimetry
and engineering practice the usage of special heat units, the calorie and the
Btu, is well established, but such additional units are really unnecessary.
The calorie is just a certain fixed number of joules (it is now defined as
4.186 joules), just as the yard is defined as a certain fixed fraction of the
meter. (For contrast compare the case of foreign exchange where the ratio
of, say, the dollar to the franc varies from time to time and place to place.)
A system will be considered to “take in heat” when, as a result of a
difference in temperature between the system and its surroundings,
energy flows into the system. Heat is energy in transit. It will be more
completely defined in connection with the first law of thermodynamics.
Energy given to a system as a result of heat processes will be denoted by
the letter Q.
Let us return to Example 1 of Section 1-4, in which we considered a gas
expanding to twice its original volume at constant temperature. If the
gas expands by rushing into an evacuated vessel, it is found that to main¬
tain its temperature it need gain practically no energy through heat
processes (Q ^ 0). On the other hand, if the gas is allowed to expand
slowly (reversibly) by pushing up a piston, then the gas must be permitted
to gain energy through some thermal process or the temperature of the gas
will fall. The gas may “take in heat” (Q > 0) through conduction by
placing it in contact with surroundings at the temperature which is to be
maintained. Evidently Q, the heat taken in, depends on how the gas is
allowed to expand. If the reversible expansion process is represented by
a curve on a p-V diagram, then Q, like W, depends on the path followed
between two given states. Since the heat taken in depends on how the
system reaches a given state, it is meaningless to speak of the “heat in
the system” at that state. The term “heat” refers to a flow of energy and
not to stored energy.
 PROBLEMS 11

A system may get hotter by having work done on it, or by taking in

heat, or by a combination of these processes. What happens to the energy
gained by the system? While this question, along with the one concerning
how we measure Q, will be answered in the next chapter, it should be
stressed here that work and heat refer only to the exchange of energy be¬
tween the system and its surroundings.
To summarize:
(1) Heat taken in (Q) depends on the path.
(2) Heat taken in (Q) is not a thermodynamic function.
(3) 8Q, a small change in Q, is not an exact differential.

A process in which no heat is exchanged between the system and its

surroundings is called an adiabatic process.

Problems

1. Suppose that for a certain gas T' find r - T[. Is T' - T[ = 50° ?
ps = 1.4 atm when p; = 1.0 atm. If 5. Refer to method (b) of defining
further readings of ps and pi at lower the Kelvin temperature scale. Suppose
densities are not taken and if we as¬ that the value of Ft in Eq. (1-6) were
sume that TS/Ti = ps/pi, Ts — Ti = increased by 1%, how would Ts — T<
100°, find Ti and Ts. be affected?
2. If we assume Eq. (1-1) and Eq. 6. How is 0°K defined on the gas
(1-4), but in place of Eq. (1-2) we thermometer scale? Show how this
take Ts t- T i = 180°, then we obtain definition is independent of the fact
the Rankine or “Fahrenheit absolute” that all actual gases liquefy (and
scale, (a) Find T» and Ts on this scale, solidify) before 0°K is reached.
(b) Find absolute zero on the Fahren¬ 7. Compute the work done by a gas
heit scale. when expanding from a volume Va to
3. Suppose that we assume lim^^o a volume U6 if the process is (a)
(pjpi) = 1.366 and that there are isobaric, at constant pressure p, (b)
100° between the ice and steam points, isothermal, at constant temperature T.
but in place of Eq. (1-1) we take Assume the equation of state pV =
T'/T'i = limPi_*0 (p/pd2- (Primes RT, where R is a constant, (c) Repre¬
refer to this new scale of tempera¬ sent the work in (a) and the work in
tures.) Find T\ and T's. (b) graphically on a p-V diagram.
4. (a) Find limPi_>o(p/Pi) at T = 8. (a) Express a pressure of one
323°K (50°C) from Eq. (1-1). atmosphere (76 cm of mercury) in
Evaluate the square of this quantity, n/m2.
that is, limPi_>o(p323/p273)2- (b) Compute the work done due
(b) Set T'/T'i = limPi_»o(p/Pi)2, to expansion when 1 kg of water
as in problem 3. Use the value (volume = 1000 cm3) is converted
of T[ found in problem 3 and your into 1 kg of steam (volume = 1.67 m3)
result in (a) above to find the value of at atmospheric pressure, expressing
T' corresponding to 323°K. For this the answer in both joules and calories.
 CHAPTER 2

THE LAWS OF THERMODYNAMICS

2-1 Internal energy of a system. In Chapter 1 we saw that a system

may gain energy by having work done on it, or by taking in heat. We
must now consider the question of what becomes of such energy. Let us
recall that in mechanics when a person does work lifting an object the work
done is said to increase the potential energy (P.E.) of the object; if one
does work pushing an object to make it move faster, then the work done
is said to increase the kinetic energy (K.E.) of the object. It is therefore
logical to assume that when work is done on an isolated system in thermo¬
dynamics the energy of the system is correspondingly increased. In
kinetic theory one may say that the energy of the system may be increased
by causing its molecules to move faster (a gain in K.E.), or by causing the
molecules to move against intermolecular forces into a configuration of
greater P.E., or both. The K.E. plus the P.E. of the system is called its
internal energy and will be denoted by the letter XJ. This is an extensive
quantity.
In classical thermodynamics we do not try to examine the nature of
internal energy from the molecular viewpoint, but just say that internal
energy represents energy stored in the system. We shall be concerned
with the question of how the internal energy U is related to such other
thermodynamic quantities as the volume, pressure, temperature, work
done, and heat taken in.

Example. The temperature of a viscous liquid is to be raised 10°. It is found

that this may be done by (1) heating the liquid over a burner, (2) stirring the
liquid vigorously, or (3) by a combination of heating and stirring.
Each of these processes involves adding energy to the system, which suggests
that the 10° rise in temperature is associated with an increase in the internal
energy. In (1) heat is taken in and no work is performed, while in (2) work is
done on the system through the stirring, but no heat is taken in. Thus we are
led to infer that U is functionally related to T, but not to Q or W.

2-2 The first law of thermodynamics. This law is simply_a statement

of the conservation of energy principle as applied to thermodynamic
processes. These processes involve the exchange of energy between a
system and its surroundings through work, or heat, or both.
From the principle of conservation of energy we are led to say that when
there is a net transfer of energy into a system from its surroundings then,
12
2-2] THE FIRST LAW OF THERMODYNAMICS 13

since this energy cannot be destroyed, it must be stored in the system.

The energy gained must then represent AU, the increase in the internal
energy of the system. If the system takes in Q units of heat energy and
performs W units of work, Q — W is the net energy gained. The first
law then states that

(2-1)

This is the way we define change in internal energy. But of course the
first law must be more than a definition of AU. What the law further
states (and this is most important) is that AU is the same for all processes
leading from the same initial state to the same final state. In other words,
if the system is taken from a state a, specified by a definite temperature,
pressure, and volume, to another state b by two different paths (on a
p-V diagram), AU is the same for each path. If the two paths are denoted
by the subscripts 1 and 2, respectively, the first law of thermodynamics
may be written as

Uh - Ua = Qx - Wx = Q2 - W2. (2-2)

This gives the first law an experimental basis. While we cannot measure
U directly, we shall see that Q as well as W can be measured, and so we
can verify experimentally that Qi — Wj does equal Q2 W2 in Eq. (2-2).
An immediate consequence of the above is that we can think of there
being a definite internal energy associated with any given state of a system.
If Ua is the internal energy for state a and if, in going from state a to
state b, there is an invariable change AU, then for state b the internal
energy must have the definite value Ub = Ua + AU. This means that U
is a thermodynamic variable. As stated in Section 1—3, in the case of a
chemically homogeneous gas only two thermodynamic variables are
independent; in such a case U must be a function of any two other thermo¬
dynamic variables, say p and V, V and T, or p and T.
If Eq. (2-2) were not true we could take the system from state a to
state b by one path and back to state a by the other, ending with a different
internal energy for state a than at the start. Now if a system could have
its energy changed without any other apparent change, this would amount
to energy being created or destroyed, which contradicts experimental
observation.
When trying to understand the first law of thermodynamics it is helpful
to recall the significance of Newton’s second law of motion (F = ma).
Newton’s law introduced the concept of mass and stated that it is constant
for a given particle; defining the unit of mass arbitrarily, the equation
F — ma may be used to measure force. Similarly, the first law of thermo¬
dynamics introduces the new concept of change in internal energy and
14 THE LAWS OF THERMODYNAMICS [CHAP. 2

states that this change is constant for all paths between two given states,
as stated by Eq. (2-2); this equation may then be used to measure the
quantity Q.
For a process involving a very small change between neighboring states
we may call the change in internal energy, dU, an exact differential,
since U is a thermodynamic variable. Equation (2-2) may then be
written as

dU = 8Q - 8W, (2-3)

that is, the difference between two inexact differentials, 8Q and bW,
equals an exact differential.
Suppose that a system is taken from state a to state b by (1) giving the
system Q units of heat, but not letting any work be done (as when heating
a beaker of water over a burner). Then suppose that one may also bring
the system from state a to state b (2) without giving it heat, but by doing
a certain amount of mechanical work W on the system (see the example
at the end of the last section). In Eq. (2-2) Qi = Q, W\ — 0, Q2 = 0,
W2 = —IE, so that Q — W. The work IE may be measured, and hence
Q may be determined experimentally. This experiment was originally
done by Joule, and when carefully performed we find that 4.186 joules
will raise the temperature of 1 gm of water from 14.5°C to 15.5°C. When
the calorie was defined as the heat required to raise 1 gm of water from
14.5°C to 15.5°C, the above experiment was taken as determining the
mechanical equivalent of heat J to be 4.186 joules per calorie. Now the
calorie is regarded as merely another energy unit defined as 4.186 joules
(just as 1 inch equals 2.54 cm) and the above experiment is taken as a
measurement of what is termed the specific heat capacity c of water at 15°C.
The definition of c is

(2-4)

where m is the mass of substance involved and 8Q is the heat in joules

required to raise the temperature dT degrees. Thus for water at 15°C,
c equals 4186 joules per kilogram per degree centigrade. The letter C
will refer to the specific heat capacity in joules per kilogram-mole per
degree. Since SQ depends on the path taken, so must C. Hence we shall
later_distinguish between Cp, the specific heat capacity (per kilogram-
mole) at constant pressure and Cv, the specific heat capacity (per kilogram-
mole) at constant volume.
Let us consider the application of the first law of thermodynamics to a
cycle. Here the system returns to its original state. Hence Ub = Ua
since, according to the first law, U is a thermodynamic variable. From
2-3] THE SECOND LAW OF THERMODYNAMICS 15

Eq. (2-1) we see that Q — W = 0, or Q = W in this case. For a cycle,

the net flow of heat into the system equals the net work done by the
system.
According to the first law, energy is conserved in processes involving
heat and work, which means that it is impossible to construct a thermal
machine that will create energy out of nowhere. A machine that would do
this could run itself and is called a perpetual motion machine of the first
kind. The first law rules out such machines. The fact that no one has
succeeded in building such a machine constitutes one experimental basis
for the first law.

2-3 The second law of thermodynamics. There are many experimental

facts in thermodynamics which are not covered by the first law, for example
the following:
(1) When two bodies at different temperatures are brought in contact,
the hotter body cools down and the colder body warms up.
(2) The spontaneous development of a temperature difference in a
system originally at uniform temperature does not occur.
(3) The expansion of a gas into an evacuated vessel occurs spontane¬
ously, but not the reverse.
(4) In a cycle involving the taking of heat from a hot reservoir and
the development of work, not all of the heat taken from the hot reservoir
can be converted to mechanical energy.
(5) A storage battery will automatically discharge through a resistor,
but will not charge itself.
The above statements all refer to the direction in which natural processes
proceed, and we see that there are “one-way” processes in Nature. The
first law tells us nothing about this. The reverse of any of the above
statements could be consistent with the first law. For example, if 10 joules
of heat flowed naturally from a cold body into a hotter body, the internal
energy of the cold body decreasing by 10 joules and that of the hotter
body increasing by 10 joules, then the first law would be obeyed. Since
this process is not found to occur, another law is needed to cover this fact.
In expressing the laws of Nature we seek simple statements that include
as wide a range of observed facts as possible. Often a law can be expressed
in many alternative but equivalent ways; which way is best may be a
matter of personal preference. Listing sufficient statements like the above
might indicate the direction of natural processes in general, but this would
be far from concise. It would be helpful to find a thermodynamic variable
of an isolated system which, say, always increases during a natural process
and which equally decreases for the system when the latter (no longer
isolated) is made to return from the final to the initial state of the natural
process' Such a variable, the entropy, will later be found and used to indi-
16 THE LAWS OF THERMODYNAMICS [CHAP. 2

cate the direction of natural thermal processes (Section 2-8). At present

it is appropriate to base the second law, as we did the first, on the experi¬
mental impossibility of perpetual motion.
We find that a machine can be built to do work by taking in heat at a
higher temperature and giving up some of the heat at a lower temperature.
If another kind of machine could be built that did work by transferring
heat from a colder to a hotter body, then the second machine could be
used to replenish the heat reservoir of the first and the first machine could
be run indefinitely at no expense. The same result would follow if a case in
Nature could be found where heat flows of its own accord from a colder
to a hotter body. If a machine could be built to take heat from a
body and turn it all into work, then work could be obtained by cooling
the ocean; with this work a reservoir could be heated, to keep an ordinary
machine going at no expense. Such a device is called a perpetual motion
machine of the second kind. We may now take as our first form of the
second law the following:

I. A perpetual motion machine of the second kind is impossible.

From what has just been said we see that this statement is equivalent
to the following one, by Clausius:

II. It is impossible to construct a device which, operating in a cycle, has

the sole result of transferring heat from a cooler to a hotter body.

A third statement of the second law is the following, known as the Kelvin-
Planck law:

III. It is impossible to construct a device which, operating in a cycle, has

the sole effect of extracting heat from a reservoir and performing an equiva¬
lent amount of work.

2-4 The Carnot cycle. From the theoretical standpoint the most im¬
portant cycle is the Carnot cycle, which consists of four reversible processes,
alternately isothermal and adiabatic, as shown in Fig. 2-1. In this cycle

Fig. 2-1. Carnot cycle.

2-4] THE CARNOT CYCLE 17

. i "•
i............■■■..■■■■■■■I

■ . Gas \\

1
Wf
7-li-'-y X V %v

Insulation
m

Tx
(a) (b) (c)
Fig. 2-2. Steps in a Carnot cycle.

(1) the system is allowed to expand, taking in heat from a reservoir at a

constant temperature T2 (Fig- 2-2a), (2) the system is allowed to expand
further while insulated, its temperature falling from T2 to Tx (Fig. 2-2b),
(3) the system is compressed, giving up heat to a reservoir at a constant
temperature Tx (Fig. 2-2c) until (4) further compression while insulated
restores the system to its original state. Each step is assumed reversible.
Call Q2 the heat taken in during step (1) and Qx the heat rejected during
step (3), so that for the successive steps Q is equal to Q2, 0, —Qi, and 0,
respectively. Note that all the heat absorbed is taken in at a single tem¬
perature and all the heat rejected is given up at another constant tempera¬
ture. As explained in Section 1-6, the work W done by the system per cycle
is measured by the area enclosed by the cycle on a p-V diagram; this area
is shown shaded in Fig. 2-1.
From the first law, Eq. (2-1), we have

Ub - Ua = Q - W.

Flere Ub = Ua for the complete cycle and the net heat taken is

Q = 28Q = Q2-Qi,
so that here
W = Q2 - Qx. (2-5)

The thermodynamic efficiency 77 of a heat engine is defined as the ratio of

the work done to the heat taken in. Using this definition and Eq. (2-5),
we can write for the efficiency of a Carnot engine (an engine operating in
a Carnot cycle)
W Q2-Q1
18 THE LAWS OF THERMODYNAMICS [chap. 2

Multiplying rj by 100 will give the percent efficiency. Since, according to

the second law, Q i cannot be zero, the efficiency is always less than 100%.
The heat taken from the hot reservoir must be replenished during each
cycle and this is done at the expense of some fuel consumed, hence Q2 is
what we pay for. The work W obtained is what we want. The heat Qi
rejected is wasted, going out the exhaust or into a cooling system. Thus
the ratio W/Q2 is a measure of what we get to what we pay for. The work
W may not all be employed usefully (some will be used to overcome fric¬
tion) but this further loss in mechanical efficiency is not included in the
term “thermodynamic efficiency. ”
Since the steps of a Carnot cycle are reversible, the path may be trav¬
ersed in the opposite or counterclockwise direction. Then the sign of each
<5Q and that of W are reversed, so that work is now being done on the
system, with the net result that heat Q i is taken from a cold reservoir and
still more heat Q2 is given up to a hot reservoir. This is a Carnot refrigerator.
The circulating chemical constitutes the system, the motor supplies the
work W (which you pay for), the heat Qi is taken from the inside of
the box (which is what you want), and the heat Q2, which equals Qi + W,
is thrown out into the room. The ratio Qi/W thus measures the per¬
formance of the refrigerator (not its efficiency). A refrigerator does not
violate the second law (see form II) because it is not self-acting, but
requires an outside fuel-consuming device to run it.

2-5 Carnot’s theorems. In this section we shall derive two theorems

based on the second law of thermodynamics.
Let E be a Carnot engine and E' any other engine, each running between
the same two temperatures T2 and Since E is by definition reversible,
we may imagine it to run backwards as a refrigerator, the work W required
to drive it being just that supplied by E'. Let us say that E' takes heat
Q2 from the hot reservoir and gives up heat Qi to the cold reservoir,
while E gives up heat Q2 to the hot reservoir and takes in heat Qx from
the cold one. We then have

W = Q2 Qi = Q2 — Qi-

The net heat taken per cycle from the hot reservoir, AQ2, is then

AQ2 = Q2 Q2 — Q2 Qi T~ Qi — Q2 = Qi — Qi-

The net heat received per cycle by the cold reservoir is

AQ1 = Qi Qi = AQ2.

The two-engine combination is thus a device which operates in a cycle

2-6] kelyin’s absolute scale of temperature 19

and transfers the heat Q2 — Q2 from a hotter to a cooler body. If Q'2 > Q2,
the second law is not contradicted. However, if Q2 > Q2, then Q'2 — Q2
is negative and we have a device in which each cycle has the sole result of
transferring heat Q2 — Q 2 from the colder to the hotter reservoir, in
violation of the second law. The second law thus requires that

Q'2 > Q2,

which means that the efficiency W/Q2 of E must be equal to or greater

than the efficiency W/Q2 of E'. Thus, according to the second law, we
have the

Theorem. The efficiency of a Carnot engine is the greatest possible for any

engine operating between two given temperatures.

If E and E' are both Carnot engines, then either one can be run back¬
wards by the other. Following the reasoning above we can show, first,
that we must have Q2 > Q2, and then, when the roles of E and E' are
reversed, we must also have Q2 > Q2. Both conditions cannot be satisfied
unless
Q'2 — Q2,

and the efficiencies W/Q2 and W/Q2 are the same. We have proved the

Theorem. All Carnot engines, operating between two given temperatures,

have the same efficiency.

From these two theorems we may conclude that the efficiency of a Carnot
engine operating between two given temperatures is independent of the
working substance used and is a function only of the temperatures T2
and Tl

2-6 Kelvin’s absolute scale of temperature. Lord Kelvin sought a

temperature scale whose definition was independent of the properties of
any particular substance. The theorems of Carnot suggested a scale
based on the functional relationship between the efficiency of a Carnot
engine and the working temperatures of the engine.
Let the subscripts s and i refer, as before, to the boiling and freezing
points of water at atmospheric pressure.
(a) Kelvin first postulated that

Ts — T i — 100°, (2-7)

as on the centigrade scale.

 THE LAWS OF THERMODYNAMICS [CHAP. 2
20

(b) Let 77xoo be the thermodynamic efficiency of a Carnot engine that

takes in Qs units of heat at the temperature Ts and gives up Q{ units of
heat at the temperature rL\. We have

’Jioo — (2-8)

(c) Consider a Carnot engine which takes in heat Qs at the temperature

Ts and gives up heat Q at any lower temperature T. The efficiency for this
new cycle will be called t}x, so that

Qs — Q (2-9)
Qs
(d) Let
Vx Vx _
X = (Ts Ti) 100 (2-10)
”100 vioo

(e) Kelvin’s second postulate was that

T — Ts — X, (2-11)

where X is defined as in Eq. (2-10).

Examples. (1) If t]x = O.5?7100, then X = 50 and T is 50° lower than Ts,
i.e., halfway between the boiling and freezing points of water. This is a defini¬
tion of 50°C, and it does not depend on the thermal properties of any particular
substance.
(2) If 7]x = 1.4r7100, then X = 140 and T is 140° lower than Ts. T now corre¬
sponds to —40°C.
(3) The same formulas may be used to define T when Q > Qs. From the
second law, we see that T now cannot be less than Ts, for we must consider
our Carnot engine to be running backwards as a refrigerator, taking in heat Qs
and giving up more heat Q at a higher temperature. t]x will now be negative.
Suppose 7] x = — 0.5t7100, then X = —50 and T is 50° above Ts. This corresponds
to 150°C.

(f) Combining Eqs. (2-8), (2-9), (2-10), and (2-11), we can write

Ts - T _ Qs-Q
(2-12)
Ts - T{ Qs-Qi’

which is the functional relation between T and Q, and which is linear.

2-6] Kelvin’s absolute scale of temperature 21

Equation (2-7) determines the size of the degree interval. It remains

only to fix the zero point for the scale (or the value of Ts).
(g) Kelvin postulated the simplest linear relation between T and Q,
namely,

T « Q)
or

T = aQ, (2-13)

where a is a constant for a given value of Qs. This assumption amounts

to taking T — 0 when Q = 0. Actually, such an absolute zero has never
been attained experimentally, but it can be taken as a theoretical zero
point independent of the properties of any particular substance (e.g.,
water). We shall find methods of computing absolute zero from experi¬
mental data.

Example. Experimentally, we find that Qs/Qi = 1.366. From Eq. (2-13)

we have
Ts = aQs, Ti = aQi,

TJTi = Qs/Qi = 1.366.

When we use T= Ts — 100°, we find that

Ts = 373.15°,

as in Section 1-5. (Note that the value of the constant a is unimportant.)

An immediate consequence of Eq. (2—13) is that we may express the

efficiency of a Carnot engine in terms of its operating temperatures T2
and Ti, as follows: From Eq. (2-6) and Eq. (2-13) we have

Q2 — Qi _ T 2 — Ti (2-14)
v - Q2 t2

Example. A Carnot engine operates between the boiling and freezing points
of water at atmospheric pressure. Taking T2 = Ts = 373 and Ti = Ti = 273 ,
we find
100
0.268.
V ~ 373
22 THE LAWS OF THERMODYNAMICS [CHAP. 2

This represents an efficiency of only 26.8%, which no engine working between

these two temperatures can exceed. (In practice, the thermodynamic efficiency
of a steam engine is increased by raising T2, through the use of superheated
steam at high pressure.)
Suppose that in the present example Q2 = 3730 joules. Then, from Eq.
(2-13), Qi = 2730 joules. The work done per cycle is W = Q2 — Qi = 1000
joules and rj = 1000/3730 = 0.268, as before.

2-7 Entropy. For the four steps in a Carnot cycle we took Q to be

Q2, 0, —Q1, and 0, respectively. If we let 8Q refer to the heat taken in by
the system in an infinitesimal change, and if we divide each 8Q by the
corresponding Kelvin temperature and sum for the cycle, we get

-sp 8Q Q2 Op
(2-15)
2^ t ~ T2 Tx‘

From Eq. (2-13), T2 = aQ2, Tx = aQx, so that

8Q _ jl
(2-16)
T ~ a

for a Carnot cycle. In each step the heat SQ' given to the surroundings by
the system is 8Q' = —8Q. We see that

£f = 0 (2-17)

for the surroundings of a Carnot engine.

Equations (2-16) and (2-17) can be extended to the case of any re¬
versible cycle, such as the one represented by the smooth closed curve in
Fig. 2-3. The area enclosed by the curve may be divided into a number of
adjacent strips by drawing a series of adiabatic lines. The original cycle
can be approximated (as closely as desired) by taking the zigzag path
shown. This zigzag path follows isothermal and adiabatic lines alternately,
and thus encloses a series of strips, each of which represents a Carnot cycle.

-v -V
Fig. 2-3. Any reversible cycle may Figure 2-4
be divided into Carnot cycles.
2-7] ENTROPY 23

Equations (2-16) and (2-17) apply to each strip separately and hence to
the whole cycle, since going clockwise around all of the strips is equivalent
to going around the whole cycle. Replacing the summation sign by the
symbol for integration around a closed path, we have

/ (2-18)

for any reversible cycle.

Next consider a system which can be taken from state a to state b by
different reversible paths acb, adb, aeb, as shown in Fig. 2-4. From
Eq. (2-18),

acb bda acb bea

Hence

f 8Q_f 5Q_f 8Q
1 T ~ J T~ .IT’
acb bda bea

or

(2-19)
acb adb aeb

This proves the following

Theorem. The integral of bQ/T is the same for any reversible process
leading from a given initial state to a given final state.

Now if the value of a definite integral is independent of the path followed

and depends only on the limits of integration (i.e., the values of the coor¬
dinates at the terminals of the path), then the quantity being integrated
(the integrand) must be an exact differential of some function of the
coordinates. Hence we may write

^ = dS, (2-20)

where 8Q is the heat taken in during a very small reversible change. Here
S must be a function of the thermodynamic variables and hence is itself
a thermodynamic variable [see paragraph (2) below]. The function S is
called the entropy of the system. It is an extensive quantity. When a
24 THE LAWS OF THERMODYNAMICS [CHAP. 2

system is taken reversibly from a state a to a state b, integration of Eq.

(2-20) gives

(2-21)

This says, in words, that to evaluate the change in entropy of a system between
any two states we must take the system along a reversible path connecting the
states, divide the heat given to the system along an element of the path by the
corresponding Kelvin temperature, and find the integral of this quotient for
the whole path.
We see that while 8Q itself is not an exact differential, there are two ways
of obtaining an exact differential from 8Q. One is to subtract from 8Q
another inexact differential 8W and so obtain (according to the first law)
the exact differential dU. The other method is to divide 8Q by T and ob¬
tain (according to the second law and for a reversible process) the exact
differential dS.
In connection with the definition of entropy, the two following points
should be noted:
(1) We have defined only change in entropy, just as we define change in
potential energy in mechanics and change in the electric potential in elec¬
trostatics. In mechanics it is often convenient to arbitrarily assign the value
of zero to the potential energy at some given point (sea level, floor level,
etc.) and then to compute the potential energy at any other point relative
to the reference point. In this way values may be assigned to the potential
energy at different points. Similarly, here in thermodynamics, the entropy
of a given state of the system may arbitrarily be called zero. Thus in the
case of water, the entropy of the liquid at 0°C and a pressure of one at¬
mosphere is assumed to be zero when we wish to compile a table giving the
entropy of water over a wide range of temperatures and pressures. On the
other hand, Nernst has pointed out that there is justification for regarding
the state of zero entropy as the crystalline state at absolute zero. While
important to physical chemists, the absolute value of the entropy of a
system need not concern us here.
(2) Equation (2-20) may be used to compute changes in entropy for
reversible processes only. If S is to be a thermodynamic function and have
a certain value Sa for the state a and another definite value Sb for the state
b, then Sb — Sa must have the same value no matter how the system
goes from a to b, reversibly or irreversibly. We have seen that this is so
for all reversible processes. We now complete the definition of entropy
by saying that Sb — Sa is taken to have the same value for an irreversible
process as for a reversible one between the same two end states. To com¬
pute Sb — Sa for an irreversible process one of the following methods
may be used:
2-7] ENTROPY 25

i
(a) Devise a reversible process connecting the same two end states and
compute the entropy change for this process, using

56 - = (2-21)

(b) If the function relating S to, say, V and T is known, Sa may be

computed by substituting the values for Va and Ta in this function, then
by similarly computing Sb, Sb — Sa may be found.
(c) If a table of values for S has been computed, look up the values of
Sa and Sb in the table and compute Sb — Sa-

Example 1. A kilogram-mole of gas expands from a volume Va = V to a

volume Vb — 2 V at constant temperature T. Assume the gas to obey the
relation pV = RT (as for an ideal gas) and find Sb — Sa.

Solution. If such an ideal gas expands freely into an evacuated vessel no heat
is taken in or given out. However, this does not mean that the change in entropy
is zero, for free expansion is an irreversible process. We must replace such a
process with a reversible one. This may be done by imagining the gas to be in a
cylinder with a piston which is allowed to move slowly by gradually reducing
the pressure on its exterior.
In the free expansion process both W and Q are zero. Hence, from Eq. (2-1),
Ub — Ua = 0, or Ub = Ua. Since U is a thermodynamic function of, say, V
and T, Ub must equal Ua for the reversible expansion process as well. In the
reversible process, when the gas does the work 8W pushing the piston out, it
must be given, in order to maintain T (and U) constant, the heat

8Q = SW = p dV.

Using the relation pV = RT, we find

From Eq. (2-21) we obtain

R In 2.

Thus there must be a gain in entropy of R In 2 in the free expansion process as

well, even though throughout this process 8Q is zero.
26 THE LAWS OF THERMODYNAMICS [chap. 2

Example 2. Repeat Example 1, using the relation Sb — Sa = CV In (Tb/Ta) +

R In (Vb/Va), which holds for an ideal gas (see Eq. 3-39).

Solution. Here Tb = Ta and Vb = 2Va, hence we immediately obtain

Sb — Sa = R In 2,
as in Example 1.

Example 3. Compute the entropy of one kilogram-mole of steam at 100°C

and a pressure of 1 atm, taking the entropy of water at 0°C to be zero. Assume
for the specific heat capacity of water the constant value Cp = 75,000 joules/kg-
mole-deg, and for the latent heat required to change one kilogram-mole of
water into steam at the boiling point Lv = 40.7 X 106 joules/kg-mole.

Solution. As the first stage of the process we may consider the water to be
heated slowly in a reversible manner, as in Example 2 of Section 1-4. Since T
is not constant, we must integrate dS = 8Q/T. We thus obtain

373
= 75,000 In
273

= 23,600 joules/kg-mole-deg.

As the second stage of the process, we consider the water at 373°K to be

converted into steam at the same temperature. Since T is now constant, we have

AS2 = j,f 8Q

Lu
= ~T

40.7 X 106
373

= 109,100 joules/kg-mole-deg.

Taking S = 0 for water at 0°C, we have S = 23,600 joules/kg-mole-deg

for water at 100°C and S = 23,600 + 109,100 = 132,700 joules/kg-mole-deg
for steam at 100°C. In this way an entropy table for water could be prepared,
to avoid repeating calculations like the above.
2-8] ENTROPY CHANGES AND IRREVERSIBILITY 27

Example 4. One kilogram-mole of water at 273°K is mixed with one kilogram-

mole of water at 373°K, the final temperature being 323°K. Compute the in¬
crease in entropy of the 2 kilogram-moles together.

Solution. In the mixing process no heat is exchanged with the surroundings,

but this process is irreversible and so the fact that for it 8Q = 0 does not mean
that AS, the net change in entropy, is zero.
We may devise as a reversible process one in which the kilogram-mole of cold
water is gradually heated to 323°K, the kilogram-mole of hot water is gradually
cooled to 323°K, and then the two are allowed to mix. The last step involves no
change in the state of the system.
In heating the cold water reversibly the change in entropy is

ASi = Cp In ^
a

323
= 75,000 X In
273

= 12,670 joules/kg-mole-deg.

In cooling the hot water the change is

AS2 = Cv In ^
J- a

= 75,000 X In —

= —10,850 joules/kg-mole-deg.

The net gain in entropy of the 2 kg-moles of water is AS = 12,670 — 10,850 =

1820 joules/deg, whether the process is performed reversibly or irreversibly.

2-8 Entropy changes and irreversibility. In any process we consider

we may say that the entropy change of the universe is

A Su = AS + A S', (2-22)

where AS is the entropy change of the system undergoing the process and
A S' is the entropy change of the rest of the universe. The “rest of the
universe” may be considered to consist of two parts: (a) the immediate
surroundings with which the system exchanges energy and (b) the re¬
mainder of the universe, which is not affected during the process. If the
system whose p, V, and T are being observed is isolated from its surround¬
ings, then (b) will include all of the universe except the system. However,
28 THE LAWS OP THERMODYNAMICS [CHAP. 2

in general, the system together with its immediate surroundings may be

considered to constitute an isolated system during the given process.
Whatever occurs in such an isolated system does not (during the process)
affect the rest of the universe, so for such a system the entropy change is
the same as ASu. Therefore, when computing change in entropy, “uni¬
verse” and “isolated system” may be taken to mean the same thing.
Consider first a reversible process. Here, from the definition of reversi¬
bility, any interchange of heat between the system and its surroundings
must occur when the two are at very nearly the same temperature. If the
system gains a small amount of heat 8Q, we have

A S=^f A S'=-8-Q, ASU = 0.

The same conclusion holds when the system loses a small amount of heat
to its surroundings, as well as when there is no exchange at all. We have
the

Theorem. In a reversible process the entropy of the universe is unchanged.

Now let us turn to irreversible processes. All natural processes fall into
this class, since we can never completely eliminate friction or take the time
to perform an infinite number of infinitesimal reversible steps. We saw in
the examples of the last section that when an irreversible process occurred
in an isolated system, the system actually gained entropy. In the case of
a system isolated from its surroundings, the latter are not affected, and
so we have

AS > 0, AS' = 0, ASu > 0.

If the system is not isolated, we may, as previously explained, include the

immediate surroundings as part of a larger system (such as that composed
of the two moles of water in the last example). Reasoning as before,
we will always find, for an irreversible process, a net gain in entropy for
the system plus the immediate surroundings, and this will represent a net
gain for the universe as a whole.
To justify such a general statement, let us return to the second law
(Section 2-3). First take statement II. It tells us that in natural processes
heat passes from a hotter to a colder body, which is exactly what we con¬
sidered in Example 4 of the last section. We found there a net gain in
entropy, which represented a net gain for the universe.
When the heat Q'2 is taken from a reservoir at a temperature T2, not all
of this heat need pass to a cooler body. By means of an engine some of the
heat may be converted into the work W. Statement III of the second law
tells us that only part of Q'2 can be turned into work, the balance, Q\, is
2-8] ENTROPY CHANGES AND IRREVERSIBILITY 29

given to a cooler reservoir at temperature TV If the engine is not reversible,

we saw (Carnot’s first theorem) that its efficiency is less than or equal to
the efficiency (T2 — TQ/T2 of a Carnot engine. Suppose, then, that

Qa ~ Qi . T2 - Tj
Q'2 t2 ■

This leads to
Qi , __ T\
Q'2 t2 ’

Qi . Qi
Ti T2 ’

AS' = ^ > 0. (2-23)

J 1 1 2

Thus the cold reservoir gains more entropy than the hot reservoir loses
and there is a net gain in entropy for the surroundings of the engine. The
system within the engine returns to its original state during each cycle,
so that AS — 0. Hence

ASu > 0.

This holds for all actual engines, since we cannot build an engine without
some friction. As a result, some of the work W is spent against friction
and thus turned back into heat at a temperature lower than T2, which
results in lowered efficiency.
There are many natural processes in which a system or reservoir gains
internal energy at the expense of energy stored in some available form.
Examples include falling bodies, the release of a stretched film or spring,
the release of a compressed gas, the coming to rest of a pendulum or a golf
ball, the discharge of a battery, and other uses of chemical energy. In
these examples it may be noted that the source gives up available stored
energy, like the P.E. of a stone at the top of a cliff, or the P.E. of a stretched
string, or the K.E. of a moving object. Such energy is quite different from
the internal energy U, for the former is not a function of such thermo¬
dynamic variables as T and V, while U is. This available stored energy
may be used (1) to increase the available stored energy of some other
object, and (2)' to do work against friction. Natural processes always
involve some of (2). Doing work against friction produces a local rise in
temperature and results in a flow of heat into the surroundings, which,
as has been shown, means a net gain in entropy for the universe. Hence
in all the above processes ASu > 0. We thus have the
30 THE LAWS OF THERMODYNAMICS [CHAP. 2

Theorem. In an irreversible (i.e., natural) process the entropy of the uni¬

verse increases.

This leads us to the following fourth way of stating the second law:

IV. Natural processes always proceed in such a direction that, as a result

of such processes, the entropy of the universe increases.

The entropy of the universe is thus continually increasing. In an iso¬

lated system the final equilibrium state will be that of maximum entropy.

Example 1. A resistor has a heat capacity of 10 joules/deg. It is maintained

at 27°C by placing it in flowing water. The resistor is connected to a battery
from which, through an electric current, it takes 1200 joules. What is the change
in entropy of (a) the resistor, (b) the water, and (c) the universe?

Solution, (a) Since the T and p of the resistor are constant (its state is un¬
changed), AS = 0.
(b) The water (i.e., the surroundings) takes in the 1200 joules as heat at a
constant temperature of 300°K; hence

AS = "300” = 4 l°ules/deg-

(c) ASu = AS + AS' = 4 joules/deg.

Example 2. Repeat the preceding example for the case where the resistor is
thermally insulated from its surroundings.

Solution. The electrical resistance is like friction to the flow of the electrons
constituting the electric current. Thus the 1200 joules are added to the wire,
and raise its temperature 1200/10 = 120°. A reversible process leading to the
same end result would involve letting the resistor take in 1200 joules as heat,
and for such a process we obtain

/•420

AS / *2 10 dT
'300 T T

420
10 In 3.36 joules/deg,
300

AS' = 0,

ASu =AS -j- AS' = 3.36 joules/deg.

2-9] ENTROPY AND AVAILABILITY OF ENERGY 31

Example 3. A bullet whose K.E. is 1200 joules buries itself in the ground.
If the ground temperature is 27°C, find the entropy change of (a) the bullet, (b)
the ground, and (c) the universe.

Solution. This problem is exactly like Example 1 except that the available
stored energy used up is kinetic rather than chemical. We have (a) AS = 0,
(b) AS' = 4 joules/deg, and therefore (c) ASu = 4 joules/deg.

Example 4. An engine working between 127°C and 27°C takes in 1200 joules
of heat from the hot reservoir. Its efficiency is 80% of that of a Carnot engine
operating between these two temperatures. Find (a) W, (b) Qi, and (c) AS„ per
cycle.

Solution, (a) A Carnot engine working between 127°C and 27°C would have
an efficiency of 100/400 = £, or 25%. The actual engine then has an efficiency
of 0.8 X 25% = 20%. Q2 = 1200 joules. Hence

W
0.20,
1200

W = 240 joules.

(b) Qi = Q2 — W = 1200 — 240 = 960 joules.

(c) AS = 0 for one cycle.

960
AS' _-™“ +
400 300

= -3.0 + 3.2

= 0.2 joule/deg.

The —3.0 joules/deg is the entropy lost by the hot reservoir and the +3.2
joules/deg is that gained by the cooler reservoir. We thus have

ASU = AS + AS' = 0.2 joule/deg.

2-9 Entropy and availability of energy. We saw in the last section that
in natural processes there is a gain in the entropy of the universe and a loss
in the stored available energy in the universe. Thus an increase in entropy
is accompanied by a decrease in the energy available for work in the
universe. This energy is not lost or destroyed but, according to kinetic
theory, becomes associated with a random thermal motion of molecules,
atoms, or electrons, and cannot be put to our use. Kelvin first recognized
this 'principle of the degradation of energy. It leads us to a fifth statement of
the second law, as follows:
32 THE LAWS OF THERMODYNAMICS [CHAP. 2

V. A s a result of natural processes, the energy in the universe available for

work is decreasing.

Entropy is thus a measure of the unavailability of energy. Perhaps it is

unfortunate that entropy should turn out to be a negative concept. This
could be, and is, sometimes obviated by introducing the negative of entropy
(-S) as a new quantity, just as a minus sign is usually introduced when
defining such potential functions as electric potential or velocity potential. *
Later we shall obtain further insight into the nature of entropy; it will
be found that entropy is related to the disorder in the universe and to the
probability of a given arrangement of the particles of a system.

2-10 Summary. The laws of thermodynamics may be summarized as

follows:
1. The energy of the universe remains constant. 8Q — SW is an exact
differential dU. The internal energy U of a system is a function of its
thermodynamic variables (or state).
2. The entropy of the universe is increasing. 8Q/T, in a reversible process,
is an exact differential, dS. The entropy S of a system is also a function
of its thermodynamic variables (or state).

Problems

1. A system is taken from state a pute the specific heat capacity c in

to state b by allowing it to do 100 joules/gm-deg at each temperature,
joules of work and to take in 60 joules (b) Should one conclude from this that
of heat, (a) Find C/& — Ua, the change J is not constant? Explain.
in internal energy, (b) If the system 4. Explain why you would not invest
can be taken between the same two in the development of an engine for
states in such a way that W = 50 which the inventor claimed either of
joules, find Q, the heat taken in. (c) the following sets of data: (a) T2 =
Repeat (b) when W = 0. 400°K, Tx = 300°K, Q2 = 20,000
2. It is found that to raise 1 gm of joules, Qi = 15,000 joules, W =
ice 1°C requires 0.48 calorie at —20°C. 10,000 joules, or (b) T2 = 400°K,
Compute the specific heat capacity C Ti — 300°K, Q2 = 20,000 joules, Qx =
of ice at —20°C in joules/kg-mole-deg. 10,000 joules, W = 10,000 joules.
3. It is found that to raise 1 mole 5. A Carnot engine operates between
(18 gm) of water 1°C requires 18.09 T2 = 500°K and Tx = 300°K. Com¬
calories at 0°C, 18.00 calories at 27°C pute the increase in efficiency when (a)
and 18.03 calories at 67°C. (a) Com¬ T2 is increased 10% (or 50°), (b) Tx

*See Volume I, pp. 82 and 243.

 PROBLEMS 33

is decreased 10% (30°), (c) T\ is de¬ tropy change per cycle of each reser¬
creased 50°. voir and that of the universe.
6. A Carnot refrigerator is operated 10. A 1-kg stone at 27°C falls 102
between 250°K and 300°K and requires meters into a lake whose temperature is
250 joules of work per cycle. Find (a) 27°C. Find the entropy change of (a)
the heat absorbed from the low- the stone, (b) the lake, and (c) the
temperature reservoir, (b) the heat universe.
ejected at the higher temperature, (c) 11. Repeat problem 10 for the case
the coefficient of performance. where the stone is lowered reversibly
7. A refrigerator operates between into the lake. Compare the loss in
250°K and 300°K, requiring 250 joules available energy in the two cases.
of work per cycle, but it has only one- 12. Compute the entropy per kilo¬
half the coefficient of performance of gram-mole of ice at —20°C relative to
the Carnot refrigerator in problem 6. that of water at 0°C, using C = 37,500
Find the heat absorbed from the cold joules/kg-mole-deg and, for the heat
box and the heat ejected into the room, to melt ice, L/ = 6 X 106 joules/kg-
per cycle. mole.
8. Compute the net gain in entropy, 13. A kilogram-mole of an ideal gas
ASU) per cycle (a) in problem 6, (b) in is compressed reversibly at constant T
problem 7. to half its original volume. Compute
9. A reversible engine takes in 2000 the change in entropy of (a) the gas,
joules from a reservoir at 400°K; it (b) the surroundings, and (c) the
then gives up Q[ joules to a reservoir at universe.
300°K and Qi joules to a reservoir at 14. Repeat problem 13 for the cycle
200°K, doing 750 joules of work per consisting of a reversible compres¬
cycle, (a) Compute the efficiency and sion to half volume followed by a
compare with that of a Carnot engine free expansion back to the original
working between 400°K and 200°K. volume.
(b) Find Q[ and Qi. (c) Draw the cycle 15. Draw a Carnot cycle on a T-S
on a p-V diagram, (d) Find the en¬ diagram.
 CHAPTER 3

THERMODYNAMIC RELATIONS

The various terms used in thermodynamics have been defined and the
fundamental laws have been stated; it remains to express the laws in
convenient mathematical form and to derive certain relations from them.
In this chapter we shall not make an exhaustive survey of such relations,
but rather content ourselves with the derivation of some of the more
important ones. In this way the mathematical technique employed in
thermodynamics will be illustrated.

3-1 Mathematical statement of the combined first and second laws.

From the first law, we have

dU = 8Q - 8W. (3-1)

From the second law, we have

8Q = T dS, (3-2)

where 8Q refers to the heat absorbed in a reversible process occurring

between two adjacent states. From the definition of work, we saw that

8W = pdV. (3-3)

Substituting Eqs. (3-2) and (3-3) into Eq. (3-1), we obtain

dU = T dS — p dV. (3-4)

Equation (3-4) is based on and expresses the content of both the laws
of thermodynamics. This is the fundamental equation of thermodynamics,
just as (d/dt)(mv) = F is the fundamental equation of mechanics. Note
that Eq. (3-4) involves only thermodynamic variables and exact differ¬
entials whose values do not depend on whether a process is reversible or
irreversible.

3-2 Differentiation in thermodynamics. We know from experience that

when dealing with a given amount of a homogeneous substance the pres¬
sure, volume, and temperature are related. For example, one kilogram-
mole of nitrogen at 27°C and a pressure of 1 atm will always be found to
34
3-2] DIFFERENTIATION IN THERMODYNAMICS 35

occupy a volume of 22.4 m3. Thus we can write

V = fi(p, T), (3-5)

which says that V is a function of p and T. This equation is called the

equation of state for the system. We can equally well solve the equation
for p in terms of V and T, or for T in terms of V and p, obtaining

P = h(V, T), (3-6)

and
T = h(V,p). (3-7)

The above three equations all refer to the same functional relationship,
and so the equation of state can be expressed in the general form

F(p, V, T) = 0. (3-8)

It is frequently necessary, in thermodynamics, to find the form of the

function F for a given system. We shall consider this problem later.
We have seen that U and S may be considered to be thermodynamic
variables. They must, then, be functionally related to p, V, and T, for,
as stated in Section 1-3, a single substance has only two thermodynamic
degrees of freedom. Of the five variables p, V, T, S, and U, we can take
any two as the independent ones. In forming partial derivatives of a func¬
tion with respect to one of the chosen variables, we must state which
other independent variable is to be held constant. This is conveniently
done by use of a subscript. For example, {dV,/dp)x will mean the rate of
change of V with respect to p when T is held constant; here p and T are
taken to be the independent variables. On the other hand, (dV/dp)s
will mean the rate of change of V with respect to p when S does not change,
p and S now being the independent variables. These two derivatives are
not the same.

Example 1. For an ideal gas we shall find that pV = RT and S = So +

Cv In (pVy), where R, So, Cv, and y are constants. We then have

RT V
V =
V p

Let us consider S constant, as we differentiate the equation for it with respect to

p. We then get
Cv
0 = V7 + y pV7 1 i
pV^
36 THERMODYNAMIC RELATIONS [CHAP. 3

which leads to

Unless 7 is unity,

In a process involving a very small change in the variables, we can

write

^ - ©/»+ (jr\dT (3"9)

when p and T are taken as the variables. The validity of this equation
may be proved by letting dT = 0 and dividing by dp, or by letting dp = 0
and dividing by dT] either procedure leads to an expression identical with
our definition of the partial derivative.
When p and S are considered to be the independent variables, we can
write

Example 2. For the ideal gas of Example 1, we have

(?z)
\dp/T
= _Z
p’
(d_Z)
\dT/p
_ zp _ zT
Hence
V V
dV = --dp-Y-dT.

Similarly, we find

Introduction of additional functions, such as U, opens up further pos¬

sibilities in differentiation, e.g., (dU/dT)p, (dU/dT)v, (dU/dT)s, etc.
However, the mathematical procedure is the same as that which we have
just been following.
3-3] SOME THERMODYNAMIC COEFFICIENTS 37

3-3 Some thermodynamic coefficients. Of the many possible partial

derivatives in thermodynamics we shall call attention to a few that are
frequently referred to and measured experimentally.
In Section 2-2 (Eq. 2-4), we defined specific heat capacity. Let 8Q
now refer to the heat taken in by a kilogram-mole of a substance. At
constant pressure the specific heat capacity per kilogram-mole may be
written as

= (3-u)

It should be noted that taking a process to occur between two adjoining

equilibrium states and at constant pressure is equivalent to specifying the
path on a p-V diagram. In such a case, 8Q does have a definite value and
hence the ratio in Eq. (3-11) does also. (In an irreversible process, such
as the free expansion of a gas, the system does not pass through a succession
of equilibrium states; the 8Q for such processes will not concern us in what
follows.) Now we can write SQ = T dS, and Eq. (3-11) becomes

Similarly, the specific heat capacity at constant volume is defined as

Cy = (3-13)

In the case of solids and liquids there is not a great deal of difference
between Cv and Cv, but for gases Cp is considerably the greater. This is
because at constant pressure a gas expands and does work, this work
requiring the addition of more heat for a given increase in T. The ratio of
specific heats 7 is defined as

7 = C7fi- (3-14)

In Example 3 of Section 2-7 we have already encountered a quantity L

called the latent heat of transformation. This stands for the heat per kilo¬
gram-mole absorbed or liberated during a change of phase. We shall let
Lf, Lv, and Ls refer to the latent heat of fusion (solid-liquid), vaporization
(liquid-vapor), and sublimation (solid-vapor), respectively. If we define

(3-15)
38 THERMODYNAMIC RELATIONS [CHAP. 3

and call V i and V2 the respective volumes for the two states, we have

L = l(V2 - 70. (3-16)

Two other important coefficients are

/3 = i(«Z)
77l =
) = the coefficient of expansion,
V \dTj„
and

k = -L(~) the compressibility.

V \dp )t

3-4 Gibbs’ functions and Maxwell’s relations. Let us return to

Eq. (3-4):
dU — T dS — p dV. (3-4)

Here U is considered as a function of S and V, or U = U(S, V). We

expand dU as we did dV in Eqs. (3-9) and (3-10), and write

dU 8U
dU = dS dV. (3-17)
,dS/ v \d V/ s

Comparing Eq. (3-17) with Eq. (3-4), we see that

T =
<£),- -’AS).-
Further differentiation yields

dT d2U (dp\ d2U

dV/s dVdS \dS/v

Since in second partial derivatives the order of differentiation does not

matter, we have

(fvt = - (iX ■ <3->8>

which is Maxwell’s first relation.
The following are three other useful thermodynamic functions which,
along with U, are sometimes referred to as Gibbs’ Functions:

A — U — TS, the Helmholtz free energy,

G = U — TS + pV, the Gibbs free energy,
H = U + pV, the enthalpy.
3-5] THE CLAUSIUS-CLAPEYRON EQUATION 39
i
From each of the above we may obtain an equation like Eq. (3-18)
relating partial derivatives. The procedure is similar in each case. For
example, let us start with A. We have

dA = dU — d(TS)

= dT-dSr~ pdV - SdT — JAetS'

= -pdV — SdT.

Here we consider A = A(V, T) and have

Taking second derivatives of A, we get

(3-19)

which is Maxwell’s second relation.

Similarly, we have

dG = dU - d(TS) + d(pV)

= -SdT + V dp,

(3-20)

Maxwell’s third relation, and

dH = dU + d(pV) = T dS + V dp,

(3-21)

Maxwell’s fourth relation.

3-5 The Clausius-Clapeyron equation. If we refer to the expression for

dG (preceding Eq. 3-20), we see that in a process for which both dT and
dp are zero dG = 0, or G is a constant. Such a process corresponds to a
change in phase. When we have two phases in equilibrium the addition
of heat (dS > 0) causes a change in volume (dV), as more of the substance
melts (vaporizes, sublimes). Throughout such a process T and p are very
nearly constant.
40 THERMODYNAMIC RELATIONS [CHAP. 3

Equation (3-19) contains the ratio of dS to dV in an isothermal process.

If we apply this equation to a change in phase, we can interpret (ST/dp) v
as the rate of change of the temperature of the melting (boiling, sublima¬
tion) point with pressure. Using Eqs. (3-15) and (3-16), Eq. (3-19)
becomes

(3-22)

which is the Clausius-Clapeyron equation. The following are some useful

applications of this equation:
(1) By observing whether V2 is greater than Uj or V\ is greater than
V2, we can predict the effect of a change of pressure on the melting (or
boiling) point of a substance. In the case of water, V\ > V2 in the transi¬
tion from solid to liquid, L is positive, and hence (dp/dT)v < 0, which
means that an increase in pressure (dp positive) causes a lowering of the
melting point (dT negative). For the change from liquid water to steam
V2 > Vi, L is positive, and (dp/dT)v > 0. This leads to the result that
an increase in pressure raises the boiling point (as in a pressure cooker).
(2) By measuring (dp/dT)y, L, V2, and V\ experimentally, we may
compute T, the temperature on the Kelvin scale at which the change in
phase occurs. We see that by applying this procedure to the boiling point
of water at a pressure of one atmosphere, we can calculate Ts. The result
is that Ts is found to be approximately 373° above absolute zero.
(3) Since Ts may be found by other methods, we can use Eq. (3-22) to
compute V2, the volume (per kilogram-mole) of saturated steam at the
boiling point temperature.

Example. When water boils at atmospheric pressure it is found that

(dp/dT)v = 2.68 cm of mercury per degree C, the latent heat Lv = 9.7 X 106
cal/kg-mole, T = 373°K, and Vi = 18,000 cm3/kg-mole. Compute V2.

Solution. This example illustrates the care that must be taken, in thermo¬
dynamic problems, to express all quantities in the same system of units. Let us
use the mks system.
2.68 cm of mercury = 0.0268 m of mercury.
Density of mercury = 13.6 gm/cm3 = 0.0136 kg/cm3 = 13,600 kg/m3.
g = 9.8 m/sec2.
(dp/dT)v = (0.0268) X (13,600) X (9.8) = 3570 newtons/m2-deg.
Lv = (9.7 X 106) X (4.186) = 4.06 X 107 joules/kg-mole. Hence

V2 - Vi
3-6] THE GIBBS-HELMHOLTZ EQUATION 41

Since Fi = 0.018 m3/mole, it may be neglected, and we have

V2 = 30.5 m3/kg-mole = 30,500 cm3/gm-mole.

3-6 The Gibbs-Helmholtz equation. Up to this point we have con¬

sidered only systems for which the work bW could be expressed as p dV.
There are, however, other ways in which systems can do work, or have
work done on them. For example, when a force F stretches a wire an
amount dx, the work done by the wire is bW = —Fdx, hence here —F
takes the place of p, and x that of V, in the various thermodynamic
relations, such as those of Maxwell. Another interesting example is the
thermodynamic theory of magnetism, where we find bW = —B dm, B
being the flux density (analogous to —p), and m being the magnetization
(analogous to F). A third example, which we shall now consider more
fully, is the application of thermodynamics to a reversible cell, such as
a storage cell.
Suppose that a quantity de of electricity passes through a reversible
cell C in the direction of discharge, the volume of the cell remaining
constant. The work done by the cell is

bW = S de, (3-23)

where S is the electromotive force (work per unit charge) of the cell at
the temperature T of the experiment. In this case we see that 8 takes the
place of p, and e that of V. Equation (3-4), which expresses the combined
first and second laws, now becomes

dU = TdS - 8 de (3-24)

and the increase in the free energy A is

dA = dU - d(TS) = —8> de — SdT.

Taking e and T as the variables, we have

Taking second derivatives, we get

(3-25)

which is the analog of Maxwell’s second relation.

42 THERMODYNAMIC RELATIONS [CHAP. 3

Now dQ = dU + 8W — dU + £ de, so that

(as) =l(zQ\
\ de)x T V de)t
1_
(3-26)
T

If we substitute this in Eq. (3-25) and rearrange terms, we obtain

(3-27)

which is the Gibbs-Helmholtz equation for a reversible cell. Here — (d U/de) t

represents the loss of energy by the cell per unit charge passing through it.
Since the temperature is constant, this loss of internal energy comes about
here through a release of stored chemical energy. As current passes through
the cell a chemical reaction takes place, energy is released, and this energy
enables the current to pass through an external resistance and develop
heat, or to run a motor and do work. If the chemical reaction took place
by itself, the internal energy released would appear as the heat of the
reaction. The heat of reaction in joules/mole is

Heat of reaction = Fv
de / t

(3-28)

where v is the valence and F is the Faraday constant (about 96,500

coulombs), the quantity of electricity passing through the cell per gram-
equivalent of the reacting substances. Hence, by measuring the electro¬
motive force and its temperature coefficient, the heat of the chemical
reaction can be calculated. As electromotive force measurements may be
performed very accurately with a good potentiometer, this method can
be a precise one for finding heats of reaction.

3-7 Ideal gases. Here we shall define an ideal gas as one which obeys

(1) Boyle's law: At constant temperature

pV = a, (a constant), (3-29)
3-7] IDEAL GASES 43

(2) Joule’s law: U depends on T only, or

U = U(T). (3-30)

From the kinetic theory viewpoint this amounts to neglecting the volume
occupied by the molecules themselves, and to assuming no intermolecular
forces. At low pressures and high temperatures actual gases closely
approach this ideal.
Let us consider p as p(V, T) and write

*= + (l)/r- (3-31)
From Eq. (3-29), V dp + p dV = 0 at constant T, or

(dp\ _ p_
(3-32)
\d V) t V

From Maxwell’s second relation, we get

(eA =(ds\
\dT/v \dV/t

i(sq\
~ T\dV/T

From the first law of thermodynamics and Eq. (3-30),

5Q = dU + pdV — + pdV,

so that

(ifi)
\;i V/t
= V-
Hence
(8p\ = P (3-33)
\dTJv T

Substituting for the partial derivatives in Eq. (3-31), we obtain

dp — — ~ydV + ipdT,

dp ,dV_dT
(3-34)
p + V T
44 THERMODYNAMIC RELATIONS [CHAP. 3

If we integrate and call the constant of integration In R, we get

In p + In V = In T + In R,
In (pV) = In (RT),
pV = RT, (3-35)

which is the equation of state for an ideal gas. When V refers to a kilogram-
mole, R is the gas constant per kilogram-mole. It is found experimentally
that R has the same value for one kilogram-mole of any ideal gas.

Example. Compute R, given that when p = 76 cm of mercury and T =

273°K (the so-called “standard conditions,” STP), then V = 22.4 m3/kg-mole.

Solution. Changing to mks units, as in the example at the end of Section 3-5,
we have

p = (0.76) X (13,600) X (9.8) = 1.013 X 105 n/m2,

V = 22.4 m3/kg-mole.
Hence
(1.013 X 105) X (22.4)
R
273

8310 joules/kg-mole-deg.

From Eq. (3-35) it follows that at constant volume p is proportional

to T. In Section 3-8 we shall show that an actual gas approaches an ideal
gas at low pressures. Thus we see that the definition of T in Section 1-5
coincides with that of the T which we are now using, namely, the Kelvin
temperature.
Equation (3-29) is the expression for an isothermal process involving
an ideal gas, and the slope of an isothermal on a p-V diagram is given
by Eq. (3-32).
For an ideal gas, the equation for an adiabatic process (dS = 0) is also
important. While this equation may be obtained in a number of ways,
we shall give a derivation that will illustrate further the mathematical
manipulation of partial derivatives. Taking S = S(V, p), we have

for an adiabatic process. Dividing by dV yields

3-7] IDEAL GASES 45

V\ ,= _ (dS/dV)p _
&V) s (dS/dp) v

The value of the right side of this equation will not be altered if we multiply
the numerator and denominator by equal quantities, such as those in
Eqs. (3-19) and (3-20), Maxwell’s second and third relations. Using also
the definitions of Cp, Cv, and y given in Eqs. (3-12), (3-13), and (3-14),
we get

(dp) 0dS/dV)p(dV/dT)p(dS/dV)T
\dV/s (dS/dp)v(dp/dT) y(dS/dp)T

0dS/dT)p (dS dp)

(dS/dT)v Vau dsJr

The slope of an adiabatic on a p-V diagram is y times that of an isothermal

for an ideal gas.
Using Eq. (3-32), we now have

(dv\ _ _yp
\dV/s v ’
dp dV

In p = —y In V + In b,

pVy = b (a constant), (3-37)

at constant S, for an ideal gas. The value of y will be discussed later.

We have noted that S may be regarded as a function of, say, V and p.
We shall now find this relation-for an ideal gas. We have

pdV
(3-38)
T

from Eq. (3-4). From Eq. (3-30),

46 THERMODYNAMIC RELATIONS [CHAP. 3

since at constant volume 8W = 0 and all heat given to the gas goes to
increase its internal energy. From Eq. (3-35),

P R
T V
Hence we have
dT_ dV_
dS = Cv R
T V

Since Cy and R are constant, integration between states a and b shows that

Sb- Sa= [ dS =: Cy In yjy- + R In > (3-39)

Ja -La, * a

which relates S to T and V. To eliminate T, we use Eq. (3-35), and write

-feMft)-
Then we have

Sb-Sa = Cv ln(^) + (Cv + R) In (y^) • (3-40)

\Pa/

From Eqs. (3-38) and (3-35),

r _ T(as'\ dU , fdV\
Lp ~ 1 \dTs dT^V \dT/p

— Cy d- R- (3-41)
Thus we have
J
Sb — Sa = Cv In (3-42)
nte)’
which may be reduced to the form given in Example 1 of Section 3-1.

3-8 The Joule-Kelvin or porous-plug experiment. Now that we have

discussed the properties of an ideal gas, we must consider how closely
actual gases come to being ideal and in what manner they differ from
being so. In an earlier attempt to do this, Joule simply allowed a gas to
expand freely into an evacuated vessel, and measured the temperature at
the start and finish of the process. Because of the relatively large heat
capacity of the surroundings, Joule was unable by this method to detect
any significant effects.
3-8] THE JOULE-KELVIN OR POROUS-PLUG EXPERIMENT 47

Joule and Thomson (who later became Lord Kelvin) later improved
on the free-expansion experiment by throttling the expansion so as to
permit a continuous flow of gas from a region of higher pressure p i to one
of slightly lower pressure p2. This was done by separating the two regions
by a plug of porous material, as shown in Fig. 3-1 (a). We see that a
thermometer is placed on each side of the plug to record the respective
temperatures. The device is thermally insulated, and after the experiment
has run for some time a steady state is reached in which no heat flows
from the gas to the wall or from the wall to the gas. Any temperature
difference between T\ and T2 is then due only to the expansion of the gas.

(b)

Fig. 3-1. The porous-plug experiment.

In this experiment it is important that p\ and p2 be kept constant.

If the right side of the apparatus is open to the atmosphere, p2 will remain
at atmospheric pressure, but pi must be maintained artificially. For
theoretical purposes we may imagine pistons, inserted as shown in Fig.
3-1 (b), being moved to the right at the correct speeds for keeping px and
p2 constant. If we call the piston area A, forces of pxA and p2A, respec¬
tively, must act against the pistons.
Suppose that a mole of gas with initial volume Vx moves through the
plug, expanding on the other side to volume V2. Let X\ and X2 be the
distances the pistons move, respectively. Then the work done by the
mole of gas as it expands is

8W = p2AX2 p\AX\,
48 THERMODYNAMIC RELATIONS [CHAP. 3

and since AX2 = V2 and AXi — Vwe have

SW = p2V2 - p1V1 = d(pV).

Since the system is isolated, 8Q = 0 and

dU = -SW = -d(pV).

If we refer back to the enthalpy function H = U + pV, we see that

here
dH = dU + d(pV) - 0,

that is, the experiment involves a process occurring at constant enthalpy.

Then
dH = T dS + V dp = 0.

Regarding S = S(T, p), we may write

Hence

Dividing through by dp yields

(3-43)

where the subscript H means that this equation only holds when dll = 0,
as in this experiment. From Eqs. (3-12) and (3-20), we have

Hence Eq. (3-43) becomes

(3-44)

which is the porous-plug equation. Much useful information about real

gases can be obtained from this equation, as follows:
3-8] THE JOULE-KELVIN OR POROUS-PLUG EXPERIMENT 49
i
(1) Consider first how an ideal gas would behave in the porous-plug
experiment. For such a gas,

pV = RT,

Since Cp is not zero, (dT/dp)jj must be. Hence an ideal gas would show
no heating or cooling. Conversely, if a gas shows no heating or cooling in
the porous-plug experiment, we have

dV_dT
(p constant),
V ~ T

V oc T at constant pressure.

Such a gas would make a perfect constant-pressure gas thermometer, for

such a thermometer would measure true degrees Kelvin.
(2) If Cp(dT/dp)h is measured experimentally for an actual gas at
various temperatures and pressures, we can determine to what extent the
behavior of such a gas deviates from that of an ideal gas, and obtain an
empirical equation of state. For example, over a considerable range of
temperatures it is found that, for most gases,

where K is a small constant having different values for different gases.

From Eq. (3-44), we now have

When this equation is multiplied through by dT and divided by T2, it

may be put in the form

d (V\ _ TdV - V dT dT
(p constant).
dT \ T/ T2 F4
50 THERMODYNAMIC RELATIONS [CHAP. 3

Upon integration, we get

Y = — 3^ + C (p constant).
I i

Let us use the further experimental fact that as T increases actual gases
obey more and more closely the ideal gas equation

V _ R
T ~ v

When T is large the K/3T3 term is very small, hence we must have

V
and

pV=RT-|j|- (3-45)

Equation (3-45) is the equation of state for most actual gases. Since this
equation is based on experimental data, it is closely obeyed by most gases
over the temperature range covered by the data. From Eq. (3-45) we see
that as p approaches zero {p —> 0) the behavior of an actual gas approaches
that of an ideal gas, as represented by pV = RT and the resulting pro¬
portionality between T and p at constant volume. This is the justification
for the definition of the Kelvin temperature scale given in Section 1-5.
(3) The value of T, the Kelvin temperature, at any fixed point may be
computed by measuring all the other quantities in Eq. (3-44), substituting
these values in the equation, and solving for T. By this method we may
find T{.

3-9 Equations of state. We have already met the following two equa¬
tions relating p, V, and T:

pV = RT (ideal gas)
and
pV = RT — K' ^ (actual gas),
where

K' = — is a constant.

Many other empirical equations of state have been proposed. One that
fits experimental data extremely well over a wide range of p, V, and T is
3-10] PROPERTIES OF A VAN DER WAALS GAS 51

the following Beattie-Bridgman equation:

pV = ~V + B)( 1 (3-46)
where

c
A = Aa (l - fj, B = £„ (l - ,
y^3 ’

and A0, ai Bo, b, and c are constants having different values for different
gases. Notice the resemblance of this equation to Eq. (3-45) when
A = B = 0. The more adjustable constants it contains, the better will
an empirical equation fit the facts.
A more systematic way of representing the behavior of an actual gas
is by means of a power series expansion of pV, such as the following:

PV = A + j + y-2 + ---, (3-47)

where A, B, C, . . . are functions of T and are called the virial coefficients.

Values of these coefficients for a given gas at various temperatures may be
put in tabular form for convenience.
One of the most interesting equations of state is the van der Waals
equation:
(?+A)(F -6) = RT’ (M8)
where a and b are constant for a given gas. Since this expression contains
only two adjustable constants, it cannot be expected that it will fit the
facts as well as Eq. (3-46), or Eq. (3-47). However, it has two important
points in its favor: it has a semi theoretical basis, and it applies to the
liquid as well as to the gaseous phase. We shall show in the next chapter
that the term a/V2 arises because of intermolecular forces and that the
term b is proportional to the volume (per mole) occupied by the molecules
themselves. For a gas under standard conditions of temperature and
pressure (STP) these terms are relatively small, i.e.,

y-2«p, b « V.

3-10 Properties of a van der Waals gas. Writing the van der Waals
equation in powers of V, we get

F3 - (b + —) V2 + - V - — = 0. (3-49)
V P / V V
52 THERMODYNAMIC RELATIONS [CHAP. 3

This expression is a cubic in F, and hence for a given substance (i.e.,

given values of a and b), at a given pressure and temperature, it must have
either one or three real roots. We shall define the critical point as that
intermediate state whose pressure pc and temperature Tc are such as to
make all three roots equal. Thus, above the critical point Eq. (3-49)
has one real root, and below the critical point it has three different real
roots. At the critical point, where the three roots are all the same, say Fc,
the equation must have the form (F — Fc)3 = 0, or

F3 - 3FCF2 + 3F2cF - Fc = 0,

while Eq. (3-49) becomes

Since these two equations must be identical, we have

ah
3FC = 5 + f 3F2 — —— > vi =
Pc Pc Pc

Solving for the critical constants, we get

8a
Pc 2752 ’ Vc 35, Tc (3-50)
2TRb

In Fig. 3-2 are sketched (a) two isothermals above Tc, (b) two isother¬
mals below Tc, and (c) the isothermal for Tc, all as given by Eq. (3-49).

Fig. 3-2. Isothermals for a van der Waals gas and vapor.
3-10] PROPERTIES OF A VAN DER WAALS GAS 53
1

We see that below Tc a given T and p may correspond to three values of V,

as at l, m, and n. The region between l and n actually corresponds to the
transition from the liquid to the vapor state and, experimentally, a sub¬
stance normally follows the dotted line path from l to n, that is, V increases
at constant p and T as heat is added at the boiling point. As T is raised,
the transition from liquid to vapor corresponds to smaller increases in V,
until at T = Tc there is no transition at all. The point C corresponds to
the critical point. Above this temperature there is no apparent liquid-
vapor transition, and the substance is usually regarded as a “gas.” At
very high temperatures the gas becomes very nearly ideal and the iso¬
thermals become rectangular hyperbolas.
Let us return to Eqs. (3-50). If we measure pc, Vc, and Tc, we may use
equations (3-50) to determine a, 6, and R. But A is a universal constant,
and so we have three equations and only two unknowns. This is another
indication that the van der Waals equation represents a simplification
and hence an approximation of the true behavior of a substance. Since
it is more difficult to measure Vc than pc or Tc, we solve Eqs. (3-50) for
a and b as functions of pc, Tc, and R. In terms of these measurable quanti¬
ties, we have
27R2T2c _ RTc
(3-51)
64pc ’ Spc

By determining a and b in this manner we obtain useful information about

intermolecular forces and the sizes of molecules (see Chapter 4).
It is interesting to calculate how a gas that obeys the van der Waals
equation will behave in the porous-plug experiment. Suppose a > 0,
6 = 0. Then we have

Substituting this in Eq. (3-44), we find that

(dT) _ RT - VV + a/V
Lp \dpjH V - «/^2

2a
pV
54 THERMODYNAMIC RELATIONS [CHAP. 3

(neglecting the term a/V2, since it is small compared with p). This means
that (dT/dp)n is positive so that, with dp negative for an expansion, dT
must also be negative. Thus the gas cools.
In a similar manner, when a = 0, b > 0, we get

that is, the gas warms up during expansion.

We see that the molecular size as well as the intermolecular force
determines the temperature change of an expanding gas. With a > 0 and
b > 0 the gas may either heat up or cool down, depending on whether
b > 2a/pV or 2a/pV > b.

Example. For hydrogen, the values of the constants a and b in the van der
Waals equation are a — 25 X 103 n-m4/(kg-mole)2, b = 2.66 X 10-2 m3/kg-
mole, and we may take R = 8300 joules/kg-mole-deg.
Let us first estimate the relative magnitudes of the a/V2 and b terms at, say,
p — standard atmospheric pressure and T = 300°K. We have seen that at
one atmosphere p = 1.01 X 105 n/m2. As a first approximation for V, we take

RT 8300 X 300
V 24.7 m /kg-mole,
V 1.01 X 105

a 25 X 103
72
■ 41 n/m2.
610

To this approximation (which here is a very good one),

~2 « P> b «

To find the critical point we substitute in Eq. (3-50) and get

Pc = = 13 X 105 n/m2,

T c = 36 = 8 X 10 2 m3/kg-mole,

8a 8 X 25 X 103
Tc 33.5°K.
27 Rb 27 X 8300 X 2.66 X IQ-2

The critical pressure is about 13 atm. Hydrogen at 300°K is far above its
critical temperature, which means that at ordinary temperatures (1) it cannot
be liquefied, (2) it will behave very nearly like an ideal gas.
3-11] GIBBS’ PHASE RULE 55

To determine the behavior of hydrogen in the porous-plug experiment we

must compute 2a/pV and compare it with b. At 300°K and p = 105n/m2,
we have

2a 2 X 25 X 103
2.02 X 10 2 m3/kg-mole.
pV ~ 105 X 24.7

Since b = 2.66 X 10 2 m3/kg-mole it is the larger factor and the gas will
warm up.
If we take 2a/pV = 2a/RT and equate this to b, we get the inversion tempera¬
ture Tq, below which the gas cools on expansion. For hydrogen

2a 50 X 103
226°K.
Rb ~ 8300 X 2.66 X IQ-2

Thus to liquefy hydrogen it must first be precooled to below 226°K, then, by

allowing it to expand, it may be cooled further. However, a pressure of over 13
atm must be maintained if liquefaction is to be at all possible.

3-11 Gibbs’ phase rule. We have seen that for a single substance in
one of its phases (solid, liquid, or vapor) there are only two independent
variables or “degrees of freedom.” When we consider two phases of the
same substance in equilibrium, such as ice and water, or water and water
vapor, we find that at a given pressure the temperature is fixed, that is,
there is but one degree of freedom left. Similarly, it is found that there is
just one temperature and pressure, called the triple point, at which ice,
water, and water vapor can exist together in equilibrium, that is, there are
no degrees of freedom for such a system.
It is an experimental result that for a system consisting of a single
chemical substance
n = 3 — </>, (3-52)

where <j> is the number of phases present in equilibrium, and n is the number
of degrees of freedom. If we add another substance to our system, we may
then have an additional degree of freedom. In a mixture of two gases the
proportion of one gas to the other may be considered as a variable, in ad¬
dition to T and p, giving such a system three degrees of freedom. Dissolv¬
ing a little salt in some water is another example, the salt concentration
constituting a third degree of freedom.
However, when two different substances do not mix, react, or form a
solution, the substances may as well be considered as separate systems
having no relation to each other.
When two substances react chemically to form a third, we cannot regard
each as an independent constituent in a mixture of the three. We designate
56 THERMODYNAMIC RELATIONS [CHAP. 3

by c the smallest number of independently variable constituents needed

to define the composition of every phase in the system. We say that a
system then has c components. With c = 1 we saw that n = 3 — <t>.
Each additional component means one additional degree of freedom, so
that, in general, we have

n = 2 + c — </>, (3-53)

which is known as Gibbs’ phase rule. This expression reduces to Eq. (3-52)
as a special case when c = 1.

Example 1. A beaker of water in air has a bell jar placed over it. The air
under the jar then becomes saturated with water vapor. If we consider air as a
single substance, since its composition is fixed, we have c = 2 (H2O, air), 4> = 2
(liquid water, air and vapor mixture), and so n = 2. The variables may be the
temperature and total pressure, or T and the amount of air. The latter could be .
varied by connecting the bell jar to a pump.

Example 2. If, in Example 1, the air is kept at atmospheric pressure but

alcohol is added to the water, then c = 3 (H2O, C2H5OH, air), 4> = 2 (mixture
of liquids, mixture of vapors and air), so that n = 3. The variables could be
T, p, and the relative concentration of liquid alcohol to water.

Example 3. Suppose that some ice is also in equilibrium with the water and
alcohol mixture in Example 2. Now c = 3 as before, but (j) = 3 (solid ice, liquid
mixture, vapors and air), and so n = 2. One variable is p, since the pressure of
the atmosphere changes. The other variable may be the alcohol concentration,
which then determines T.

Example 4. Salt (NaCl) is gradually added to a beaker containing ice and

water in equilibrium under a bell jar. The jar also contains air at atmospheric
pressure. We then have c = 3 (H2O, NaCl, air). If all the salt dissolves, </> = 3
(solid ice, liquid solution, mixture of air and vapor), and so n = 2. Thus at a
given pressure and salt concentration the temperature is determined. If, how¬
ever, the solution is saturated and salt remains undissolved, c = 3, <p = 4,
and n = 1. The added phase is the solid salt. If the pressure is taken as the
variable, the concentration of the solution and the temperature are fixed. Such
a “freezing mixture” has a temperature of about —18°C (or 0°F) at normal
atmospheric pressure.
 57
i

Problems

1. Solve the van der Waals equation 9. An ideal gas is taken around a
for p as a function of V and T. Carnot cycle (see Fig. 2-1). At the
2. Given that pV + a/V = RT temperature T2 the volume increases
and U = (3/2)RT — a/V, where a is from Va to Vb, then the volume in¬
constant. Find (a) (dU/dT)v and (b) creases adiabatically from Vb to Vc,
(dU/dT)p. decreases isothermally (at temperature
3. (a) Show that the coefficient of T1) to Va, and decreases adiabatically
expansion /3 = l/T for an ideal gas. back to Va- Find the work done by the
(b) Show that for a gas obeying the gas and the heat taken in along each
van der Waals equation step, in terms of the F’s and T’s. Show
that the efficiency is (T2 — TT+TV
= RV\V - b) 10. Prove that, in general, (a)
P RTV3 - 2a(7 - b)2 0BT/dV)s = (Cv - Cp)/fiVCv, (b)
(dT/dp)s = k(Cp - Cv)/0Cp, and (c)
4. (a) Show that the compressibility
(dp/dV)s = y/kV.
k = 1/p for an ideal gas. (b) Show
11. Taking Cv as constant, show that
that for a gas obeying the van der
(a) Ub — Ua = Cv(Tb — Ta) for an
Waals equation
ideal gas, and (b) Ub — Ua —
V2(V - b)2 Cv{Tb — Ta) + a/Va — a/Vb for a
k ~ RTV3 - 2a(V - 6)2' van der Waals gas.
12. Show that for a van der Waals
5. If a wire expands on being heated, gas Sb - Sa = Cv In (Tb/Ta) +
show that if a tension is applied R In (Fb — b)/(Va — b).
adiabatically the temperature falls. 13. Show that for a van der Waals
6. Find T;, the Kelvin temperature gas
of melting ice, when p = 1 atm, given 2a(V - b)2
that an increase in pressure of 1 atm Cp — Cv = R
RTV3
depresses the melting point 0.0073°C,
Vi = 0.0196 m3/kg-mole for ice, V2 = 14. (a) Suppose that with a certain
0.0180 m3/kg-mole for water, and Lf = gas in the porous-plug experiment it is
1.44 X 106 cal/kg-mole. observed that Cp(dT/dp)H — —k,
7. If the electromotive force for a where A; is a positive constant. Find
certain cell is found to be given by the relation between V and T for such
£ = 1.0000 + 0.00015 (t — 15), where a gas. (b) Compare the behavior of
t is the temperature in °C, find the such a gas with that of one whose
heat of reaction at 25°C. equation of state is p(V — b) = RT.
8. An ideal gas undergoes the follow¬ 15. Compute T(dV/dT)p — V for a
ing processes: (a) V = k, (b) pV = k, van der Waals gas. (Neglect second-
(c) pV2 = k, (d) p/V = k, (e) pT =k, order small terms containing a2, ab,
where A; is a constant. If the initial or b2, and for first-order small terms
state is given by Pa = 1 atm, Va = containing a or b consider pV = RT
24.6 m3, Ta = 300°K, and if in the to be a good enough approximation.)
final state pb = 2 atm, find Vb and Tb Show that the inversion temperature
for each process. (below which the gas cools and above
58 THERMODYNAMIC RELATIONS [CHAP. 3

which it warms on expansion) is atm, Tc = 5.25°K), and (b) C02 (pe =

To = 2 a/Rb. 73 atm, Tc = 304°K).
16. Approximating to the same ex¬ 19. Find c, <f>, and n for a system
tent as in problem 15, find the virial composed of a thermos containing solid
coefficients A, B, and C in Eq. (3-47) carbon dioxide (dry ice) and liquid
for a van der Waals gas. alcohol, the top of the bottle being
17. Show that for a van der Waals open to the air. (The system may be
gas (dp/dV)r and (d2p/dF2)r are both considered to be under a bell jar.)
zero (that is, the isothermal has a 20. Find c, </>, and n for the system
horizontal tangent and point of inflec¬ P2CI5 (solid), P2CI3 (liquid), and Cl2
tion) at the critical point. (gas), given that
18. Compute the van der Waals con¬
stants a and b for (a) He (pc = 2.3 P2C15 P2CI3 + Cl2.
 CHAPTER 4

KINETIC THEORY

4-1 Introduction. From the two laws of thermodynamics we have

been able to derive many relations between heat, work, the various thermo¬
dynamic variables, and their derivatives. In this manner a great many
experimental facts were correctly predicted in a qualitative way. When
we came to numerical calculations we had to assume, in addition to the
fundamental laws, certain experimental data, such as that obtained from
the porous-plug experiment or from simultaneous measurements of p, V,
and T. Measured values of specific and latent heats, and of the van der
Waals constants a and b, were also assumed in some of the worked examples.
We cannot compute these quantities from expressions derived theoretically
without introducing additional fundamental hypotheses.
In kinetic theory these additional hypotheses concern the structure of
matter. While they cannot be stated as simple “laws” backed by direct
experimental observation, these hypotheses play the same role as such
laws, and their validity is based upon the prediction of results that can be
checked experimentally. In addition to a molecular hypothesis, kinetic
theory also assumes Newton’s laws of motion, and applies these laws of
mechanics to the individual molecules of a system.

4-2 Molecular hypotheses for an ideal gas. In classical thermo¬

dynamics we defined an ideal gas as one which obeys Boyle’s law and
Joule’s law (U depends only on T). We shall show that the following funda¬
mental molecular hypothesis, stated by Daniel Bernoulli in 1738, is an
equivalent definition:
A finite volume of an ideal gas consists of a very large number of molecules
of negligible size, these molecules being in random motion and colliding like
hard elastic spheres with each other and with the walls of the container. It
is assumed here that the molecules exert no forces on one another except
during a collision, so that between collisions they must move in straight
lines.
The molecules of a given substance are taken to be all alike, each with
a mass m, which is proportional to the molecular weight. Hence if we take
one kilogram-mole of any gas, we shall always have the same number of
molecules. For example, the molecular weight of oxygen is 32 and that of
hydrogen is 2. If we take equal numbers of molecules of each, the oxygen
molecules together will weigh 16 times as much as all of the hydrogen
ones. Conversely, 32 kg of oxygen will contain the same number of
molecules as 2 kg of hydrogen. This common number of molecules in a
59
60 KINETIC THEORY [CHAP. 4

kilogram-mole is a universal constant called Avogadro’s number, N0. (When

referred to the gram-mole, Avogadro’s number is one-thousandth as great.)
The molecular hypothesis assumes that No is so large that even in a
small element of volume dV of the gas there are enough molecules to form a
representative sample. Later N0 will be shown to be equal to 6.03 X 1026
molecules per kg-mole, and we shall see that this assumption is valid.*
Letting n stand for the number of molecules per unit volume, we have

Mol. wt.
m (4-1)
Ao
and

No
n = (4-2)
V

where V is the volume per mole.

4-3 Distribution of molecular velocities with respect to direction. If

the molecules in a gas are in random motion, as many must be moving
in one direction as in any other. For convenience, it is customary to
consider a fairly narrow range of directions. Consider a solid angle ft with
its vertex at the origin, as in Fig. 4-1. The number of molecules in a given
volume moving in directions contained within ft must be independent of
the orientation of ft. A solid angle of 2ft may be considered equivalent to
two solid angles, each equal to ft. Hence there must be twice as many
molecules whose directions of motion are contained within 2ft. Thus the
number dn of molecules per unit volume whose motional directions axe
contained within a small solid angle
dft is proportional to that solid angle.
A solid angle is measured by the
area it subtends on a sphere of unit
radius with center at the vertex of
the angle. Thus the 'total solid angle
about a point is 47t. We then have

dn __ dft
n 47t

77
dn = — dft. (4-3)

*As stated in the Example of Section 3-7, the volume V of a kg-mole of any
ideal gas under standard conditions is 22.4 m3, so that Nq/V is of the order of
1025 molecules/m3, or 1019 molecules/cm3.
4-4] CONCEPT OF PRESSURE IN KINETIC THEORY 61

4-4 Concept of pressure in kinetic theory. We now apply mechanical

concepts and laws to explain how a gas exerts pressure on the walls of its
container. In mechanics, pressure is defined as force per unit area, and
force is equal to the rate of change of momentum. When a molecule of
mass m and speed v approaches a wall normally, the momentum of the
molecule is mv, toward the wall. After an elastic collision with the wall the
molecule’s momentum is mv away from the wall. This represents a change
in momentum of 2mv. If we consider the more general case of a molecule
approaching the wall at an angle d with the normal, the change in mo¬
mentum AM is
AM = 2mv cos d. (4-4)

During many such collisions per unit time, the molecules striking a given
area undergo a certain average rate of change of momentum. From New¬
ton’s third law, the wall experiences an equal and opposite rate of change
of momentum, just as though a certain average force were pushing against
the wall. The pressure of the gas against the wall is taken to be this
average force per unit area. By following this line of reasoning we shall
proceed to derive an expression for the pressure in terms of m, n, and v.

Figure 4-2

A direction in space can be specified by using the spherical coordinates

6 and <£, where d is the angle measured down from the polar axis and <p is
the angle measured around the polar axis from a fixed plane, as in Fig. 4-2.
(On the earth wh take the latitude as 90° — d, and the longitude is <f>.)
Consider the solid angle dtt within which d varies from d to d -f- dd and <f>
varies from 4> to </> + d(f>. This solid angle will intercept the small area
(r sin d d4>) X (r dd) — r2 sin d dd d<j> on a sphere of radius r. This area
is shown shaded in Fig. 4-2. On a sphere of unit radius this area is just
62 KINETIC THEORY [CHAP. 4

sin 0 dd d<fr, so from the definition of a solid angle we have

dO = sin d dd d<t>. (4-5)

I «

The number of molecules per unit volume with directions between 0

and 0 -f- dd, and between <f> and 4> + d<j> is found by substituting Eq. (4-5)
in Eq. (4-3). Calling this number d2ne<*,, we have

d2no<t, = ~ sin 0 dd d<f>. (4-6)

47r

Consider an element of the wall of the gas container, of area dA.

Let the normal to the wall be the polar axis and construct the oblique
cylinder shown in Fig. 4-3. This cylinder is slanted in the 0,^-direction
and its length is v dt. Assuming for the moment that all the molecules
have the same speed v, we see that any molecule in the cylinder moving
in the 0,</>-direction will strike area dA in the time dt. If a molecule in the
cylinder has a direction between 0 and 0 + dd, and </> and <f> + d4>, it has a
chance of striking dA and, as we let dd and d4> approach zero in the limit,
this chance approaches certainty. However, molecules in the cylinder with
other directions will miss dA. Thus the cylinder contains all the mole¬
cules of speed v and directions between 0 and 0 + dd, 4> and </> + d<t>, which
hit dA in the time dt.
Since the oblique cylinder has a volume dA v dt cos 0 (base times altitude),
it contains

AN = sin 0 dd d<t>^ X (dA v dt cos 0)

molecules with directions between 0 and 0 + dd, <t> and 0 + d(j>. Each of
these AN molecules experiences a momentum change AM given by Eq.
(4-4). The contribution of these molecules to the pressure is found by

Figure 4-3
4-4] CONCEPT OP PRESSURE IN KINETIC THEORY 63

multiplying AN by AM and dividing by dA X dt. Calling this contribution

to the pressure d2pH, we have

AN AM
mnv2 sin 9 cos2 6 dd d<f>.
dA dt 2tt

The total pressure pv (for molecules all of speed v) is found by integrating

for all possible directions on the inside of the wall, that is, for values of 9
between 0 and ir/2 and values of 4> between 0 and 2x. We thus have

r2ir /'7r/2

— mnv2 / / sin 9 cos2 9 d9 d<fi

Z-k Jo Jo

x/2

mnv~ I sin 9 cos2 9 d9

o

\rnnv1. (4-7)

To be completely general, we now remove the restriction that all the

molecules have the same speed v. If v varies from molecule to molecule,
so that only m is constant, we may proceed as follows.
Let nv dv be the number of molecules per unit volume with speeds be¬
tween v and v -f dv, and let dp be the contribution to the pressure made by
the molecules with speeds between v and v + dv. We may now write

dp = \mnvv2 dv.

Integration with respect to v then yields

p = nvv2 dv.

The average value of v2 is defined as

dv,
n

so that we have «
p = \mnv2. (4-8)

In Eq. (4-8) v* means the average value of the squares of the different
speeds, or the mean square speed, while the square root of this quantity is
64 KINETIC THEORY [CHAP. 4

called the root mean square (rms) speed. The rms speed vrm3 must be dis¬
tinguished from the mean or average speed v. Let us put dnv for nv dv,
the number of molecules per unit volume with speeds between v and v + dv.
Then, by definition,

j2 dnv, (4-9)

') dnv, (4-10)

and the rms speed vTms = Vp2. Note that if we were dealing with only a
few molecules we would replace the above integrals with summations over
the individual molecules.

Example 1. Suppose that the speeds of ten individual molecules are 1, 1,

2, 2, 2, 3, 3, 4, 6, and 10 m/sec, respectively. Find v2, ^v2, and v.
Solution. Here we must use summations, or

_ l2 T l2 T 22 + 22 T 22 + 32 + 32 + 42 + 62 + 102
V 10

= ^ = 18.4(m/sec)2,

vims = V 18.4 = 4.29 m/sec.

On the other hand, we find that

l~bl~t_2T2-j-2-|-3 + 3T4-|-6Tl0
v = --
10

34
= — = 3.40 m/sec,

which shows that the rms and mean speeds may be quite different.

Let us return to Eq. (4-8). We note that ran, the mass per molecule
times the number of molecules per unit volume, is the mass per unit
volume or density of the gas. If we call the density p, we have

V = hpv2- (4-11)
4-4] CONCEPT OF PRESSURE IN KINETIC THEORY 65

Now we can measure p and p for a gas and use Eq. (4-11) to calculate the
rms speed of its molecules in a given state. We have

®rms — y/ 3 p/p. (4_12)

The velocity of sound in a gas is given by*

vs = V yp/p. (4-13)

Since sound is transmitted through a gas as a result of molecular motions,

we would expect that the speed of sound might approach, but not exceed,
the magnitude of molecular speeds. This is just what kinetic theory
predicts, for experimental values of y fall in the range 1 < y < 1.7, so
that vs is a little smaller than vTms. This lends qualitative support to the
kinetic theory and its hypotheses.
A more quantitative check on kinetic theory is provided by using
molecular beam methods, as described in the next chapter. In such experi¬
ments we can measure the actual distribution of molecular velocities and
compute vrms, obtaining values in close agreement with those calculated
from Eq. (4-12).

Example 2. For most gases we find that when p = 1.01 X 105n/m2 (one
atmosphere) and T = 273°K (0°C), then V = 22.4 m3/kg-mole. Find the
rms speed under these conditions for (a) hydrogen, (b) oxygen, (c) air
(p = 1.29 kg/m3). Compare (c) with vs for air.

Solution, (a) One kg-mole of H2 = 2 kg. Hence

" = dt ks/n'8'

I'rms = V 3p/p

= V(3.03 X 105) X 11.2

= 1840 m/sec.

(b) For oxygen 1 kg-mole = 32 kg, hence its density is 16 times that of
hydrogen. Since rrms varies inversely as the square root of the density, the rms
speed for oxygen must be one-fourth that for hydrogen at the same T and p.
Thus, for oxygen,
1840
Vrms = — /
7— = 460 m/sec.
4

*See Volume I, p. 224, Eq. (12-38).

66 KINETIC THEORY [CHAP. 4

(c) For air we have as an average (the nitrogen molecules have a greater vims
than the oxygen molecules)
3.03 X 105
^rms —
1.29

= 485 m/sec.

The speed of sound in air under standard conditions is found to be 331 m/sec,
which is somewhat smaller than prms, as expected.

4-5 Concept of temperature in kinetic theory. Consider a kilogram-mole

of an ideal gas, the volume being V. From Eqs. (4-2) and (4-8), we have

1 N0 —z

V = 3 -y mv2,

or

pV = £ Nomv*. (4-14)

The average kinetic energy (K.E.) of translation of a molecule of the gas

is _
K.E./molecule = fmw2.

(In this chapter K.E. will refer only to translational kinetic energy.)
We are postulating the molecules to be hard elastic spheres with no forces
between them. The internal energy U of such a gas can only be kinetic.
When the gas takes in heat (8Q > 0) at constant volume (5 W = 0), its
internal energy must increase (dU > 0) and so its molecules must move
faster. Experimentally we observe a rise in temperature for any gas when
its internal energy is increased. The average K.E. per molecule, \mv2,
and the temperature T go up and down together. At constant temperature
%mv2 is constant and hence, according to Eq. (4-14), pV is also constant,
and the gas obeys Boyle’s law.
That an ideal gas also obeys Joule’s law follows from the assumption of
no intermolecular forces. What happens if we increase V at constant T?
Moving the molecules farther apart does not increase the internal P.E.
when the molecules do not attract one another. Keeping T constant means
that there is no change in the internal K.E. Hence dU — 0, and U is a
function of T only.
Since the kinetic theory definition of an ideal gas is equivalent to that
in thermodynamics, we can now use the equation of state derived for an
ideal gas, namely,
pV = RT.
4-5] CONCEPT OF TEMPERATURE IN KINETIC THEORY 67

Comparing this with Eq. (4-14), we get

0mv2 = RT,

\mV2 = ~§~Q T. (4-15)

Thus the average K.E. per molecule is proportional to the absolute tem¬
perature. The kinetic theory concept of temperature is that it is a measure
of the average kinetic or thermal energy of the molecules.
Let us set

k = ~, (4-16)
iV o

where the important constant k is called Boltzmann’s constant, or the gas

constant per molecule, just as R is the gas constant per kilogram-mole.
From Eq. (4-15) we see that for an ideal gas

K.E./molecule = § kT. (4-17)

Or, if we multiply through by N0, we get

K.E./kg-mole = f RT. (4-18)

If a gas obeys the equation pV — RT, computing yrms from Eq. (4-15)
will lead to the same results as when we used Eq. (4-12). Let us, however,
assume the value of N0, compute k, and find the K.E./molecule.

Example. Given R = 8310 joules/kg-mole-deg and iVo — 6.03 X 1026 mole-

cules/kg-mole. Find (a) k, and (b) \mv2 at T = 273 K.

Solution.
R 8310
(a) k ~ Nh ~ 6.03 X 1026

= 1.38 X 10-23 joule/molecule-deg.

* " __
(b) imv2 = f kT

= 1.5 X 1.38 X 10"23 X 273

= 5.65 X 10 21 joule/molecule.
68 KINETIC THEORY [CHAP. 4

In atomic physics the amount of energy associated with one particle is

much less than a joule, so that a smaller unit of energy, the electron volt,
is generally used. One electron volt (ev) is defined as the energy acquired
by a particle bearing the electronic charge e (=1.6 X 10-19 coul) when
it is accelerated through a potential difference of 1 volt (=1 joule/coul).
Thus
1 ev = 1.6 X 10 19 joule,
and

5.65 X 10-21
K.E./molecule
1.6 X 10-19

= 3.5 X 10~2 ev at 0°C.

Hence, at about room temperature, the average kinetic energy of a gas

molecule is about 1/30 of an electron volt. This will be found to be reason¬
able.

4-6 Molecular hypotheses for a van der Waals gas. The van der Waals
equation of state differs from that for an ideal gas in two correction terms:
b, which was added by Clausius, and a/V2, which was later included by van
der Waals.
In his correction term, Clausius allowed for the finite size of the mole¬
cules. Let us call the radius of a single molecule r and assume that all the
molecules are alike and are hard elastic spheres. Then in a collision between
two molecules the closest the two centers can get must be 2r, as in Fig.
4-4(a). The center of the unshaded molecule cannot move everywhere
in the containing vessel, for in this particular collision it finds itself ex¬
cluded from a spherical volume of radius 2r about the other molecule, as
shown in Fig. 4-4(b). This excluded volume is (4/3)7r(2r)3 = (32/3)7rr3,
and it is the unavailable volume per pair of molecules; for if the center of
the unshaded molecule is kept out of this volume, this will ensure that the
center of the shaded molecule will, at the same time, be kept at a distance
of at least 2r from the molecule with which it is colliding.
In one kilogram-mole there are
N0/2 pairs of molecules. The total
volume excluded because of the
finite size of the molecules is

-2 r—
6 = ^xf
(a) (b)
4 3
= 4W0 X g TrrL (4-19)
Fig. 4-4. Molecular collision.
4-6] MOLECULAR HYPOTHESES FOR A VAN DER WAALS GAS 69

If we exclude this volume from the actual volume V, then we can consider
the molecules as point masses moving in the reduced volume V — b.
Hence in place of pV = RT, we write

p(V - b) = RT, (4-20)

where, according to Eq. (4-19), b is equal to four times the volume of all
the molecules in V.
Van der Waals attempted to allow for possible intermolecular forces in
his correction term. At close distances the internal structure of the mole¬
cules plays a part in determining the equation of state of a gas. If a mole¬
cule consists of positive and negative charges which are not symmetrically
distributed, and if these charges can be caused to shift their relative posi¬
tions in the molecule, then one molecule will have an electric field around
it and this field will act on a neighboring molecule. Since unlike charges
attract, the unlike charges on adjacent faces of two molecules will cause an.
attraction between them.
As a molecule reaches the wall of the container there are no other mole¬
cules ahead of it, but it is attracted to those in the layer immediately be¬
hind. This pull from behind results in the molecule striking the wall with
less momentum than would otherwise be the case, so that the pressure p
which we observe is really less than that for an ideal gas at the same V
and T. Before we can use the equation pV = RT for an actual gas we
must add a term to the observed pressure p. This correction term will be
proportional to the number of molecules per unit area in the surface layer
(on which the other molecules pull back) and to the number of molecules
per unit area in the next layer (these being the molecules doing the pulling).
Both quantities are proportional to N0/V, hence the reduction in pressure
is proportional to 1/E2. Calling the equivalent ideal gas pressure p', we
write for the observed pressure p,

, a
V — V y2 ’

or

P = V+y-2-

If we add the pressure correction to Eq. (4-20), we get

(p + -jh)(V-b)= RT, (4-21)

which is the van der Waals equation.

70 KINETIC THEORY [CHAP. 4

By measuring p, V, and T simultaneously for a number of states of the

gas, we can calculate a and b. From these values of a and b we can gain
some knowledge of the strength of the intermolecular attractive force and
of the size of the molecule.

Example. It is found for hydrogen that b = 0.0266 m3/kg-mole. Estimate

from this the radius of the hydrogen molecule.

Solution. If we use Eq. (4-19) and let No = 6.03 X 102G molecules/kg-mole,

as in the previous example, we get

3 _b_
167T No

= 2.64 X to-30 m3,

r = 1.38 X 10 10 m.

This value agrees well with those computed by other methods, thus lending
further experimental support to the kinetic theory.

4-7 The law of atmospheres. In the earth’s atmosphere the density

decreases with height. However, since the rate of decrease gets smaller
with increasing height, it is impossible to say just where the “top of the
atmosphere” is.
At any height y above the earth’s surface let the density be p, the pres¬
sure p, and the acceleration of gravity g. At this height y consider a layer
of the air of thickness dy and unit area. The weight of this layer (or the
force of gravity on it) will be pg dy. At the height y + dy the pressure
p + dp will be p less the weight of this layer. Hence

p + dp = p — pg dy,

dp = —pg dy. (4-22)

If m is the mass of a single molecule, we have

mN o
p = mn =
y
so that

If we assume pV = RT, the gas law, this expression becomes

4-8] DETERMINATION OF AVOGADRO’S NUMBER 71

dp __ _ mg No
(4-23)
p RT V

The pressure at any height is found by integration. Letting p = p0 at

y — 0, we get

(4-24)

To integrate the right side of Eq. (4-24) we must know how T depends
on y, or the “lapse rate. ” In the case of the earth’s atmosphere T is found
to vary with altitude (usually decreasing), up to the beginning of the
stratosphere, the bottom of which varies from a distance of 5 miles above
the earth’s poles to 10 miles above the equator. Up to this level the total
decrease in T is about 25%, but above this point T is constant up to much
greater heights.
We shall now, however, consider a simpler problem and assume that T
and g are constant for all values of y. Integration of Eq. (4-24) yields

—(mgNol RT)y
P = Poe (4-25)

This equation is known as the law of atmospheres, and for small differences
in altitude the pressure of the atmosphere is actually found to vary some¬
what in this exponential manner.

4-8 Determination of Avogadro’s number. We have seen that the

constant No is an important one in kinetic theory, and we now discuss
how it may be measured experimentally. Conveniently, this may be
done by several independent methods, as follows:
(a) Colloidal suspensions. Perrin suspended microscopic beads of a
resinous substance in a liquid of slightly lower density and then, with the
aid of a microscope, measured the concentration of the beads as a function
of depth.
In the case of the atmosphere at uniform temperature, the density, and
hence the concentration of, say, oxygen molecules, is proportional to the
pressure. Letting c be the concentration at the height h, Eq. (4-25) be¬
comes
—(mgNolRT)h
c = Coe (4-26)
72 KINETIC THEORY [CHAP. 4

Perrin realized that this same formula could be applied to the con¬
centration of beads in his liquid suspension, and with much greater accuracy
than when applied to the atmosphere. First, the value of mg was so much
larger for one of his beads than for a molecule that he needed to vary h
very little in order to get a large variation in c. Second, a uniform tempera¬
ture could be maintained very precisely. Third, the beads were large
enough so that their diameters could be measured and, knowing the
density of the resinous substance, the mass of the bead could be calculated.
By subtracting from this value the mass of the displaced liquid, the effec¬
tive value of the term mg in Eq. (4-26) could be found, and the expression
could then be used to compute N0.
(b) Faraday’s constant and charge on the electron. In electrolysis experi¬
ments it is found that it takes 1000 F coulombs of electric charge to
deposit one kilogram-mole of a univalent substance like silver, where F is
the Faraday constant defined in Section 3-6. Each ion carries an electronic
charge e, while the charge 1000 F is carried by N0 ions. Thus

N0e = 1000 F,

N0 = 1000 F/e.

F can be measured directly, while e may be independently determined

(e.g., by Millikan’s oil-drop method). Taking the experimental values
1000 F = 9.65 X 107 coul/kg-mole and e = 1.6 X 10“19 coul, we get

No = 6.03 X 1026 molecules/kg-mole.

This method is one of the most accurate for finding N0.

(c) X-rays. The wavelength of monochromatic x-rays can be measured
by use of a ruled grating, and the same x-rays can be used to determine
the grating space of a simple crystal. When the distances between planes
of ions are known, the number of ions in a given volume can be com¬
puted, and from the density the number of molecules per kilogram-mole
can be determined.
(d) Radioactivity. The number of alpha particles a given source emits per
minute can be observed directly. If, over a period of time, these alpha
particles are collected in a tube that has been evacuated and sealed, a
gradual accumulation of helium gas will result. The number of helium
molecules (atoms) in a given volume at the observed pressure and tem¬
perature, and hence the number in one kilogram-mole, can be computed
on the basis of the elapsed time.
(e) There are other methods for finding N0. For example, one is based
on thermionic emission, another on the scattering of light, and still others
are based on kinetic theory (see the following section). The main point is
that methods representative of many different branches of physics lead to
4-9] A SURVEY OF FURTHER APPLICATIONS OF KINETIC THEORY 73

the same value (within limits of error) for N0. That some of these values
are based on kinetic theory is additional verification of that theory.

4-9 A survey of further applications of kinetic theory, (a) Specific heats.

Since for an ideal monatomic gas the internal energy U is entirely transla¬
tional kinetic energy, we have from Eq. (4-18)

U = %RT.

At constant volume all heat taken in goes to increase U, hence

dU= 3
Cv
dT ~ 2
From Eq. (3-41),

Cv = Cv + R = | R.
Hence
Cp 5
7 Cv 3'

This value for y is in good agreement with that found experimentally for
monatomic gases (He, Ne, A, Kr, Xe). Other gases have different values
for 7 and this is attributed to their molecules having other forms of
kinetic energy (rotational, vibrational) in addition to that of translation.
We would expect the molecules of a monatomic gas to have the closest
resemblance to the hard elastic spheres postulated for an ideal gas, and
experimental results support this view.
(b) Transport phenomena. Molecular theory may be used to explain the
following three properties of a gas: viscosity, thermal conductivity, and
diffusion.
Viscous stresses arise when different layers of a flowing gas move with
different speeds. Then, because of thermal motions, some of the molecules
in a faster moving layer drift into an adjacent layer where the forward
velocity is less, transporting forward momentum to the slower moving
layer. Conversely, some molecules of the slower moving layer drift into
the faster moving layer and decrease the momentum of the latter. This
transfer of momentum, which is equivalent to internal friction, can be
computed from kinetic theory considerations.* One interesting result
is the prediction that at constant temperature the coefficient of viscosity
is independent of the pressure, a result which can be verified experi¬
mentally.

*See Sears, Thermodynamics, pp. 258-263.

74 KINETIC THEORY [CHAP. 4

In a similar manner, kinetic theory explains thermal conductivity as a

transport of energy resulting from molecular motions. The molecules in
a hotter region have greater average energies than those in a colder part
of the gas.
Kinetic theory explains diffusion as a transport of mass resulting from
molecular motions. A denser region will lose more molecules than it will
gain in this way. The density of a gas becomes uniform throughout its
volume when the gas reaches the equilibrium state.

Problems

1. Use the value of Avogadro’s num¬ suming (a) that the bullets are em¬
ber to calculate the mass of a molecule bedded in the wall, and (b) that the
of (a) H2, (b) 02. bullets rebound with 40% of their
2. Compute the number of mole¬ original horizontal velocity.
cules per cm3 in a gas at (a) standard 7. Compute i>rm8 for oxygen mole¬
conditions, and (b) at 0°C and a pres¬ cules at (a) 27°C and a pressure of 1
sure of one-millionth of an atmosphere atm, (b) 327°C and 1 atm, and (c)
(a fair vacuum). 327°C and 0.25 atm.
3. Show that Eq. (4-7) may be de¬ 8. Compute the temperature at
rived easily if one assumes that on the which a gas molecule would have a
average one-third of the molecules are mean K.E. of translation of one elec¬
moving parallel to each of the three tron-volt.
rectangular axes X, Y, Z. 9. Show that the number of mole¬
4. Find v and ^rras for a group of 10 cules of gas striking the walls of a con¬
molecules when (a) each has a speed of tainer per unit area, per unit time, per
5 m/sec, (b) five have speeds of 3 m/sec unit solid angle, and in a direction mak¬
and five have speeds of 7 m/sec, (c) ing an angle 6 with the normal, is
their speeds are 1, 2, 3, 4,..., 10 m/sec, nv cos 6/4-jt.
respectively. 10. Show that the total number of
5. Suppose that the speed distribu¬ molecules of a gas striking a unit area
tion for a group of particles is given by of a container wall per unit time, in¬
Nv dv = 106 v dv up to v = 500 m/sec, cluding those coming from all direc¬
but Nv dv = 0 for v > 500 m/sec. tions, is nv/4.
Find (a) the total number of particles, 11. A vessel of volume V contains
(b) v, and (c) vrmB. Plot a graph of Nv no molecules of a gas per unit volume.
VS. V. A small hole of area A is made in the
6. In one minute 240 machine gun container wall, (a) Neglecting leakage
bullets strike a vertical wall. Each back into the vessel from outside, show
bullet has a mass of 0.01 kg and a that the number of the original mole¬
horizontal velocity of 2.5 km/sec. Find cules left inside decreases exponen¬
the average force against the wall, as¬ tially with time, (b) In terms of V, A,
 PROBLEMS 75

and v, find the time required for only 15. In a colloidal suspension experi¬
no/e molecules to be left. (Here e is the ment it would be practicable to have
base of natural logarithms.) c = co/2 when y = 2 mm and T =
12. Find the volume and radius of an 300°K. Further, the suspended beads
oxygen molecule, given that b in the should be at least 10~5 m in diameter.
van der Waals equation equals 0.032 Suppose that their density is 800
m3/kg-mole for oxygen. kg/m3. Show that the effective value
13. Taking 29 kg/kg-mole as the of mg must be greatly reduced by sus¬
average molecular weight for air and pension of the beads in a liquid of very
T — 260°K, find (a) the height y at nearly the same density.
which, according to Eq. (4-25), p = 16. Assume that the coefficient of
po/2, and (b) p/po at the top of Mt. viscosity pv is a function of the mass,
Everest (29,000 ft). (Experimental ob¬ radius, and speed of the gas molecules.*
servation indicates that p = po/2 at Take pv = kmxrvvz, where k is a di¬
about 18,000 ft.) mensionless constant. From dimen¬
14. Below the stratosphere we may sional analysis find the powers x, y,
take T = To — ky (k is a constant) to and z, and from this find the depend¬
represent the variation in temperature ence of pv on T.
with height in the earth’s atmosphere.
(a) Integrate Eq. (4-24) for this case.
(b) Taking the observed lapse rate *pv may be briefly defined as the
k = 6°K/km and To = 280°K (aver¬ viscous force per unit area per unit
age sea-level temperature at New velocity gradient. For a more complete
York), compute y when p = po/2. definition see Volume I, pp. 229-230.
 CHAPTER 5

CLASSICAL STATISTICAL MECHANICS

5-1 Introduction. In Chapter 4 we saw that the molecules of a gas have

a mean square speed and energy of translation proportional to the absolute
temperature, and that by their impacts on the walls of a container the
molecules produce a pressure. It is evident that the molecules must also
strike one another, and in such collisions there will usually be a transfer
of kinetic energy (see Vol. I, p. 137). As a result we would expect that, at
any instant, some molecules will be moving faster than the average and
some slower, so that there will be a velocity distribution. If we want to
develop a theoretical expression for this velocity distribution, we must
introduce additional hypotheses.
We shall now employ statistical methods in working out the energy or
velocity distribution of the molecules. The larger the number of events or
individuals in the group considered, the more nearly correct are statistical
predictions. From the size of Avogadro’s number we have seen that in even
a very small volume of matter there are ordinarily many, many molecules.
While it is obviously impossible to follow the motions of all the individual
molecules, the situation is ideal for the application of a statistical treat¬
ment.
As at the start of any theory, certain new concepts must be defined. This
will now be done.

5-2 The state of a gas. For simplicity we shall at present consider a

system composed of N molecules of a monatomic gas. The state of such a
system may be specified in a number of ways, such as by giving one of the
following:
(a) The values of two thermodynamic variables, say V and T.
(b) The average number of molecules per unit volume and the rms speed
of the molecules, i.e., n and yrms.
(c) The number of molecules in any given element of volume and the
fraction of these with speeds between v and v + dv, or energies between E
and E + dE.
(d) The approximate position (within a certain volume element) and
approximate velocity components of each molecule at a given instant.
(e) The exact position (x, y, z) and velocity components (vx, vy, vz)
of each molecule at a given instant.
We recognize (a) as the specification used in thermodynamics and (b)
as that used in kinetic theory. From Eqs. (4-2) and (4-15) they are seen
76
5-3] PHASE SPACE 77

to be equivalent. As we pass from these to (c), (d), and (e), the descrip¬
tions become more and more specific, (e) being too much so for practical
purposes. In this chapter we shall be primarily interested in (c), which we
shall call a macroscopic description of a state. However, to find (c) we shall
also have to refer to (d), which will be termed a microscopic description of
a state. Abbreviating, we say that (c) defines a macrostate and (d) a micro¬
state of the system.

5-3 Phase space. We can easily picture an element of volume dx dy dz

in three-dimensional space. Likewise we could plot the velocity com¬
ponents vx, vy, vz in three mutually perpendicular directions and picture
in this velocity space a three-dimensional volume element dvx dvy dvz.
As a mathematical concept we may imagine a six-dimensional space in
which dx dy dz dvx dvy dvz is an element of volume and a point corresponds
to a set of values for x, y, z, vx, vy, and vz. This six-dimensional space
will be called phase space and an element of volume of this space will be
termed a cell. It is assumed that our phase space is restricted to include a
range of coordinates consistent with the dimensions of the gas container
and a range of velocities consistent with the total energy of the gas.
The position and velocity of a molecule can be exactly represented by a
point in phase space. If we could plot such a point for every molecule we
would have fulfilled specification (e) of Section 5-2.
Let us divide phase space into cells which we shall number 1, 2, 3, etc.
To define a microstate we must state to which cell each molecule tem¬
porarily belongs. To define a macrostate we need only tell how many
molecules are in each cell, such as nx molecules in cell 1, n2 in cell 2, n3 in
cell 3, etc.
Many different microstates may
correspond to the same macrostate.
For example, suppose that we iden¬
ade ni = 3
tify the molecules as a, h, c, etc.; then
Fig. 5-1 shows a certain microstate
corresponding to the macrostate for
be n2 = 2
which nj = 3 , n2 = 2, n3 — l,etc.
If we interchange any two molecules
from different cells, say a and b, we
will have a different microstate but n3 = 1
the same macrostate. If we inter¬
change two molecules in the same
cell, say a and e, we will have the
same microstate as well as the same
macrostate, because here we are not
considering any subdivision of a cell. Fig. 5-1. Cells in phase space.
78 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

Since the molecules of a gas are in constant motion, the system is con¬
tinually changing of its own accord from one microstate to another and,
almost as frequently, from one macrostate to another. However, the more
microstates corresponding to the same macrostate, the more likely the
latter is to occur.

Example. As an analogy, consider two coins, a penny (p) and a dime (d),
that are tossed into the air simultaneously. Call falling “heads” equivalent to
being in cell 1 and “tails” to being in cell 2.
There are three macrostates, as follows: (1) ni = 2, n2 = 0, (2) m = 1,
712 = 1, (3) ni = 0, ri2 = 2, that is, two heads, a head and a tail, and two tails.

Now the first and third macrostates can each occur in only one way, p\di and
P2^2, respectively; there is only one microstate for each of these macrostates.
However, two microstates, p\d,2 and p2di, correspond to the second macrostate.
If the coins are not “loaded,” heads and tails are equally likely to occur and the
second macroscopic state should turn up twice as often as either of the others.
In a large number of throws the percent difference between the predicted and
observed ratios is small.

5-4 Fundamental hypotheses. In addition to the postulate that a gas

is composed of molecules that are in motion and behave like very small
elastic spheres, we make four other hypotheses, as follows:
(a) Let us take all of the cells in phase space to be of the same size and
postulate that all microstates corresponding to possible macrostates are equally
probable.* This corresponds to the assumption in the example above that
the coins are not loaded. It follows that the probability of the occurrence
of a given macrostate is proportional to the number of microstates that
correspond to that macrostate.
(b) The equilibrium state of a gas corresponds to the macrostate of maximum
probability. It has been pointed out that the state of a gas is continually
changing, and as it does so the most probable state will occur most often.
That the gas will actually be in this most probable state, or in one only
slightly different, nearly all of the time, follows from statistical theory.
When dealing with such a large number of particles and events as in the
present case, the probability function has a very sharp maximum. In
other words, appreciable departures from the most likely distribution are
so improbable as to be of no practical consequence. For example, when
repeatedly tossing a coin the most probable distribution is 50% heads and
50% tails; in two tosses one might get all heads, in ten tosses 6 heads would
not be unusual, but in a million tosses the percent variation from the 50-50
ratio would be very small.

*This postulate follows from Liouville’s theorem; for a derivation, based on

the laws of mechanics, see Joos, Theoretical Physics, pp. 550-551.
5-5] PROBABILITY 79

(c) The number of molecules is constant. This expresses the principle of

conservation of matter. If the total number of molecules is N, we have (for
a closed system)

N = nx + n2 + n3 + • • • = a constant. (5-1)

(d) The total energy is constant. This follows from the principle of con¬
servation of energy. In this chapter we are not considering processes in¬
volving a change from one state to another, but rather we are investigating
the velocity and energy distribution for a given state. If we call the total
energy V and the energy per particle Ex for cell 1, E2 for cell 2, E3 for cell
3, etc., we have

U = Exnx + E2n2 + E3n3 + • • ■ = a constant. (5-2)

5-5 Probability. We call the number of microstates corresponding to

a given macrostate the thermodynamic 'probability of that macrostate, and
designate it as W. To find an expression for W we use the hypothesis of
Section 5-4(a) and reason as follows:
Suppose that there are nx molecules in cell 1, n2 in cell 2, etc., and N
in all. We can choose the first molecule to go into cell 1 in N ways, the
next in (AT — 1) ways, the third in (N — 2) ways, etc. After cell 1 is filled
we choose from the remaining molecules to fill cells 2, 3, etc. until we come
to the last position in the last cell, which we can fill in only one way (using
the remaining molecule). This gives us N(N — l)(iV — 2) • • • (1), or AM,
ways of filling the cells. But AM is not W because in arriving at this number
we counted separately distributions in which the same molecules fell into
the same cell, but in different sequence. There are nx! permutations of the
molecules in cell 1 which were counted separately above, but which really
correspond to one microstate. So we must divide AM by nx\, n2\, etc. We
then have
N\
W = (5-3)
nx\n2\n3\ ■ ■ ■

Example. Suppose that there are only two cells and four molecules, a, b, c,
and d. Consider the possible macrostate defined by m = 3, n2 = 1. Figure
5-2 (a) shows the four corresponding microstates. Figure 5-2 (b) shows the same
microstate as the first microstate in Fig. 5-2(a). Thus W = 4. By Eq. (5-3),
80 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

Cell 1 bed acd abd abc Til = 3 dbc

Cell 2 a b c d n2 = 1 a

(a) (b)

Figure 5-2

Other possible macrostates and their corresponding probabilities are as follows:

4!
m = 4, n2 = 0, W = ~ = 1,

ni = 2- n2 = 2, W = JL = 6)

m = l,n2 = 3, W = ■— = 4,

ni = 0, n2 = 4, W = = 1.

The reader should check the above values of IF by listing the different micro¬
states for each case. Note that 0! = 1, for an empty cell presents no alterna¬
tives when distributing molecules. Observe also that the most uniform distribu¬
tion (rii = ri2 = 2) is also the most probable.

Later we shall find it convenient to consider a group of adjacent cells,

calling such a group a zone of cells. Since the cells are of equal size, the size
of a zone is proportional to the number of cells that it contains. Compared
with a zone of one cell, a zone with two cells might cover twice the range
of one coordinate or velocity component and so include twice as many
molecules, regardless of the form of the distribution function. (E.g., our
population contains just about twice as many people between 65.0 and
65.2 years of age as it does people between 65.0 and 65.1 years of age.)
Thus, so far as the size of the zone alone goes, the probability of finding a
given molecule in a zone of 9i cells is proportional to gt. Since this proba¬
bility can be determined before considering the energy per particle (E •)
for that zone, we call gt the a priori probability for the zone- it depends
on the size of the zone only. ’ p
5-6] Boltzmann’s distribution law 81

5-6 Boltzmann’s distribution law. If we refer back to Section 5-4, we

see that hypothesis (a) was used to derive Eq. (5-3). We shall now use
the other three hypotheses.
Suppose that the numbers n2, n3, etc. characterize the equilibrium
state of a gas. Our problem is to find these numbers in terms of the energy
or velocity associated with the corresponding cells. For instance, cell i
might be one containing particles of low energy and cell j one with high-
energy particles; we are interested in whether n; > rij, or nj > tt;.
As the gas molecules move around, the n/s will continually change in
many different ways, but will always keep values close to those for the state
of maximum W. Let drii be the variation of n4-, at any instant, from its value
for the state of maximum probability. The three hypotheses referred to
above require that the Srii’s be subject to the following conditions: 5 IF = 0
because IF is a maximum, 8N = 0 because iV is a constant, and 8 U = 0
because U is a constant. We proceed to express these conditions mathe¬
matically.
From Eq. (5-3) we see that In IF is a simpler function than IF itself.
Since In IF will be a maximum when IF is, we shall let 8(In IF) = 0.
Now we have

In IF = In N! — In rii! — In n2! • • • .

Fortunately our n/s are all large numbers and so we can use a mathe¬
matical approximation known as Stirling’s theorem, namely,

In x! = x In x — x, (5-4)

if x is large. Proof of this theorem is based on the small percentage differ¬

ence between the sum

In 1 —f— In 2 —|— In 3 —}— • • • T- In x,

which is the true value of In x\ and

is the area under the step curve in
Fig. 5-3, and the integral

/ X

In x dx = x In x — x + 1,

which is the area under the smooth

curve in Fig. 5-3. (The integration
is done by parts and the final 1 is
neglected, since x is large.) Fig. 5-3. Stirling’s theorem.
82 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

Using the above approximation, we get

In W — N In N — N — fti In + fti — n2 In ft2 + ft2 ■ • •

= N In N — N — £[»* In ft; — ft;].

%

Next we let 5(ln W) = 0, differentiating with respect to fti, n2, etc.,

but regarding N as constant. We then have

5 (In W) = — Z[ln ft;5ft; + ft;5(ln ft;) — 5ft;]

i

= — Z j^ln ft;5ft; + ft; 5ft; — 5ft;J = 0,

or
Z hi ft;5ft; = In fti5fti + In n25n2 + In n3dns • • • = 0. (5-5)
i

Now we must have 5N — 0. Equation (5-1) gives

5ft i -j- 5ft2 U 5ft3 -f- • • ■ = 0, (5—6)

which simply states that if some cells gain molecules, they do so at the
expense of other cells that lose the same molecules.
The final condition is SU = 0. We differentiate Eq. (5-2), noting that
the ft/s may change, but not the Ejs (which are characteristic of the cells).
Hence
^iSfti + E25n2 + E35ft3 + • • • = 0. (5-7)

Equations (5-5), (5-6), and (5-7) must all be satisfied at the same time,
and they are independent of one another. The equations may be com¬
bined and yet have their independence specified mathematically if we
multiply each by a separate arbitrary constant and then add them. Since
the right side of the resulting equation is zero, we can divide through
by one of the arbitrary constants. This amounts to multiplying one
equation by unity, a second by the ratio of two arbitrary constants,
and the third by another such ratio. These two ratios are the only two
independent constants introduced. This method is known as Lagrange’s
method of undetermined multipliers. Let us, then, multiply Eq. (5-5) by
unity, Eq. (5-6) by the constant X, and Eq. (5-7) by the constant 0.
Adding these three equations, we get

(In fti + h + /3E1)8n1 -f (In ft2 + X + /3E2)Sn2

+ (In ft3 -f X -f 0E3)5ft3 H-= 0. (5-8)

5-6] boltzmann’s distribution law 83
i

The Sn/s in Eq. (5-8) must satisfy the two condition equations (5-6)
and (5-7), otherwise the Sn/s may vary arbitrarily. If we consider Sn3,
Sn4, Sn5, ■ • • as independent, Eqs. (5-6) and (5-7) determine Sni and Sn2
in terms of the other Sn/s. Now let us assign to X a value which makes

In 7i\ —f- X —(— /3Ei = 0,

and then give j8 a value such that

In 712 X (3E2 = 0.

Equation (5-8) now becomes

(In n3 —|— X (— /3E3)8n3 -|- (In ^4 —j— X —f— fiEf)8n4 + • • • = 0.

In this equation the Sri/s are all independent. Since we can make all of
them zero except Sn3, we see that the coefficient of Sn3 must vanish, and
so on for the other terms. Hence W (or In IE) is a maximum, subject to
the stated conditions, for those values of n\, n2, n3, • • • for which

In rii X d- fiEi — 0 (5-9)

for any value of i. Solving for w*, we get

If we introduce the constant A — e x in place of X, we have

m = Ae~0Ei, (5-10)

which is Boltzmann’s distribution law, giving the number of molecules per

cell as a function of the energy associated with a particle in that cell. Here
A and /3 are the same for all cells.
According to Boltzmann’s law, the number of molecules per cell decreases
exponentially with the energy Ei associated with the cell, as shown in Fig. 5 4.

Fig. 5-4. Boltz¬

mann’s distribution
law.
84 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

Low-energy cells will contain more particles than high-energy cells, with
cells of zero energy being the most populated. This does not mean, how¬
ever, that more molecules have zero energy than any other energy, for we
shall find that there are more cells corresponding to high than to low energies
and the most probable energy is greater than zero. Thus if we take a zone
of Qi cells all corresponding to about the same energy Ei, the number of
particles in this zone, which we shall call Ni, is given by

Ni = Agie~pE\ (5-11)

We shall see that for a small energy range dE, gi is proportional to Ej*.
From Eq. (5-10) we s^e that all cells corresponding to the same energy
but to different positions in the container are equally populated. This is
true for all values of E{. If the potential energy due, say, to a gravita¬
tional field, may be neglected, it follows that Ei is independent of x, y, and
z and thus the molecules are uniformly distributed in coordinate space.
This is what we would expect, and it is a consequence of our hypothesis
that all microstates are equally probable.

Example. Suppose that there are three cells and m = 5, n2 = 3, n3 = 2,

Ei = 0, E2 = 2 joules per particle, E3 = 4 joules per particle. Let Sn3 = —2
and find Sni and Sn2, assuming N and V to be constant. Find the initial and
final probabilities.
Solution. We have

N — ni T n2 T- n3 = 10,

U = Eim + E2n2 -j- E3n3 = 6 —)— 8 = 14 joules.

If 5n3 - —2, two molecules leave the cell of highest energy for cell(s) of
lower energy. Since these two molecules give up energy, others must gain
energy. Hence molecules in cell 1 gain energy and enter cell 2. Since N and U
remain constant, we must have

Sni = —2, Sn2 = +4, dn3 = —2.

The new distribution is n\ - 3, n' - 7, «' =0. As a check we note that

n'i + n2 -)- n3 = 10,

E\m + E2n2 E3n3 — 14 joules.

5-7] maxwell’s distribution law 85

Now I 10!
w 5! 3! 2!
2520,

10!
W' 120.
3! 7! 0!

The original distribution is much the more probable one and also bears closer
resemblance to the Boltzmann distribution.

5-7 Maxwell’s distribution law. We shall now derive an expression for

the number of molecules with speeds between v and v + dv. Since this
number will be proportional to dv (see last paragraph of Section 5-5), we
shall call it Nv dv. Here Nv is the distribution function with regard to speed.
Let us look for the fth zone of cells corresponding to the range in speed
from v to v + dv. Let gv represent the number (previously called gf)
of cells in the zone, or the a 'priori probability for the zone. Then Nv dv
must be the number (Nf) of particles in the zone, each having an energy
close to Ei = Ev = %mv2. We may use Eq. (5-11) and write Boltzmann’s
law as
Nv dv = Agve~pEv. (5-12)

Our problem now is to express gv in terms of v and dv. From the definition
of gv, we have
gv = number of cells in the fth zone

volume in phase space of the fth zone

size of one cell

To obtain the volume in phase space of the fth zone we must integrate an
element of volume dx dy dz dvx dvy dvz throughout that zone, or

ffffff dx dy dz dvx dvy dvz

gv = -“-’

where r stands for the volume of one cell.

The zone of cells to which gv refers must include all possible values of
the coordinates. In computing the volume of the zone we can integrate
at once with respect to x, y, and z, getting

fffdxdydz = V,

where V is the volume occupied by the gas. This leaves integration with
respect to vx, vy, and vz. As we care only about the resultant speed v
and not about the velocity components, we can proceed as follows.
86 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

Imagine vX) vy, and vz plotted along three mutually perpendicular axes.
About the origin describe a sphere of radius v. The surface of this sphere
will contain all points for which v2 + v2 + vz = 1,2 ■ Now describe a
second sphere of radius v -j- dv, concentric with the first. The volume
between these two spheres will contain all points in velocity space for
which the speed is between v and v + dv. This volume is Airv2 dv. Hence

Iff dvx dvy dvz = 4tv2 dv.

The volume of our zone in phase space is F(47rv2 dv). Since the volume of
one cell is r, we have

gv = irZ
T
(5~13)

and

Ev = \mv1.

If we substitute these last two expressions back into Eq. (5-12) and let
B = 4ttA V/t, we get
Nvdv = Bv2e~?mv2'2dv, (5-14)

which is Maxwell’s distribution law for molecular speeds. For very small
values of v the exponential term is near unity and Nv is proportional to v2.
For very large values of v the exponential term is the more important one
and Nv is proportional to e~av2. Thus when we plot Nv against v, as in
Fig. 5—5, the curve first rises parabolically, later reaches a maximum, and
finally falls exponentially to zero. The maximum corresponds to the most
probable speed, designated as vp. As the area under the curve to the right
of the maximum is greater than to the left, the mean speed v is greater
than vp and wrms > v > vv. (See problems 7 and 8 at the end of this
chapter.)

Fig. 5-5. Maxwell’s speed distribution.

6-9] EVALUATION OF THE CONSTANTS B AND |3 87
i

5-8 Energy distribution. Let Ne dE be the number of molecules with

energies between E and E + dE. From E = frav2, we have

where dv is the range in speed corresponding to the range in energy dE.

Then
Ne dE = Nv dv.

Substituting in Eq. (5-14), we get

,r j.(2E\ —fiE ( mV/2 dE

NEdE = B{-)e [ze)
or
Ne dE = B'Ell2e~PE dE, (5-15)

where B' is a new constant independent of E. When Ne is plotted against

E, one gets a curve similar to that in Fig. 5-6, showing that the most
probable energy is greater than zero. This curve is somewhat different
from that in Fig. 5-5 because Eq. (5-15) contains Ei as a factor, while
Eq. (5-14) contains a v2 factor.

Fig. 5-6. Maxwell-Boltzmann energy distribution function.

5-9 Evaluation of the constants B and /3. We have seen that each
distribution law contains two constants, such as B and /3 in Eq.. (5-14),
which are related to the two undetermined multipliers in Section 5-6.
These multipliers were introduced because of the conditions that the total
number of particles N and total energy U are constant for a given state.
88 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

These same conditions may be used to evaluate the constants.

We have
/» oo

N = / Nv dv
Jo

2 — pmv /2
= B vze dv.
Jo

For convenience, let us define the definite integral Jn as

n -/3mu 12
Jn — dv.
Then
N_
B = (5-16)
J2

showing that B is proportional to N.

Similarly, the total energy must be

U = / imv2Nv dv

Bm / V*e~emv2i2dv
2 Jo

Nm
© (5-17)

The following values of some of the J’s are taken from standard integral
tables:

Jo — T>X/2-jr/l3m, J1 = ——, J2 =

J3 ~ $2m2 > J4 — f'V/327r/P5m5.

When we substitute for J2 and /4 in Eqs. (5-16) and (5-17), we get

B = 2NVWmz/2TT,
5-9] EVALUATION OF THE CONSTANTS B AND (8 89
i

From Eq. (4-17), we have

U = N$kT).

Let us equate the two expressions for U and solve for /3. We then find that

/?• = (5-18)
kT'

In terms of the temperature, Boltzmann’s law becomes

Ni = Agie~Ei/kT, (5-19)

and Maxwell’s law for the speed distribution is

Nvdo = Bv2e~mv2'2kT dv,

(5-20)

The effect of the temperature on the speed distribution of a given number

of molecules is shown in Fig. 5-7, where graphs of

Nv = (4N/V^)(m/2kT)3l2v2e~mv2l2kT

vs. v have been drawn for three different temperatures, Tx < T2 < T3.
The higher the temperature, the more energy the molecules possess and
the more the higher speeds are favored. However, the areas under all
three curves are the same (N constant).

Fig. 5-7. Distribution function at three different temperatures, T3 > T2 > T\.
90 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

5-10 Experimental verification of Maxwell’s speed distribution. One

of the most direct experimental checks on the above theory was made by
Zartman and Ko, using the apparatus diagrammed in Fig. 5-8. A beam
of molecules emerging from the container 0 into an evacuated region is
collimated by the slits Si and S24, C is a rotating cylinder with a slit at S3.
As the slit S3 passes in front of the beam, a pulse of molecules is admitted
into the cylinder, traveling across its diameter and striking the glass plate
G, which is attached to the inner wall. The slower molecules will reach G
later than the faster ones, that is, the less v is for a molecule, the farther
to the left will be the point where G is struck. The faster molecules will
hit G more to the right.
Zartman and Ko chose to use
a beam of metal molecules which
would stick to the glass plate G.
Their container 0 was an oven in
which the temperature of the metal c
could be raised until the vapor
pressure was sufficiently high. The
emerging beam then produced a
deposit whose density varied across
G according to the speed distribution
of the molecules in the oven. The
glass plate was examined with a
Si
microphotometer and the density
distribution measured. The results
were in good agreement with the
speed distribution predicted by
a 0

Fig. 5-8. Experiment of Zartman

Eq. (5-20). and Ko.

5 11 Entropy and probability. Consider two completely independent

systems, each in equilibrium, designated by the subscripts a and b, re¬
spectively. Since entropy is an extensive quantity, the entropy S of the
two systems together must equal the sum of their separate entropies, or

& — Sa + Sb. (5-21)

Let Wa be the probability or chance that a certain event will occur and
let Wb be the probability of occurrence of another independent event.
Then the probability W that both of these events will occur is W = WaWi).
Suppose, for example, that one person tosses a coin while another draws
a card from an ordinary complete deck. The result of repeated coin tossing
is that heads appear close to half the time, so we say that the chance that
a person will toss heads one particular time is Wa = Similarly, the
5-11] ENTROPY AND PROBABILITY 91

chance of drawing a king from the card deck is Wb = j3. So the prob¬
ability that one person will toss heads while another person draws a king
is 2V) i-e-, both of these events will occur together on the average only
once in twenty-six times.
Let us return to the two independent systems in equilibrium. If Wa
now represents the probability that the first system is in the state of
entropy Sa, while 1T& is the probability that the second system is in the
state of entropy Sb, then the probability that both of these events will
occur together is
W = Wa X IT&. (5-22)

We have seen that the state of equilibrium is both the state of maximum
entropy and of maximum probability, and this suggests a relationship be¬
tween entropy and probability. The simplest relationship consistent with
Eqs. (5-21) and (5-22) is a logarithmic one. Therefore we shall take

S = /c In W, (5-23)

where k is a universal constant which, it will develop, is the same as Boltz¬

mann’s constant (Section 4-5). To see that this relation satisfies the con¬
ditions mentioned above, we note first that when IT is a maximum In W
and S will be also. Furthermore, since

Sa = k In Wa, Sb = k In Wb,

we have
S = Sa + Sb = k In (WaWb) = k In IT,

or Eq. (5-23) is consistent with Eqs. (5-21) and (5-22).

To evaluate k we must combine the results of statistical mechanics with
those of classical thermodynamics.
In this chapter we have, up to the present, considered only systems in
equilibrium and the small variations about that state resulting from the
continual motion of the molecules. Variations in the quantities nx, n2, n3,
etc. resulting from such statistical fluctuations were called 8nx, 8n2> 8n3,
etc. Now we must give our attention to a process involving a small change
from one equilibrium state to another slightly different one. For such a
process we shall use the differential symbol d, as in thermodynamics. If,
as a result of the process, the number of molecules in cell 1 changes, the
change will be called dnx, etc.
In the process we shall consider a small amount of heat 8Q is added
to a gas at constant volume, thereby moving some of the molecules from
cells of lower energy to cells of higher energy. At constant volume the
increase in internal energy, according to thermodynamics, is
92 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

du = 8Q = TdS. (5-24)
From Eq. (5-2), we have

dU — Ei drii -)- E2 dn2 ~!~ E3 dn^ -f- ■ ■ *

— ^ '.E j drii, (5-25)

i

and from Eq. (5-23),

dS = k d(In W). (5-26)

We showed in Section 5-6 that

In IF = iV In iV — N — [rii In rii — nr],

i
so that (for N constant)

d(ln W) = — X) In rii drii,

i
and (see Eq. 5-26)
dS = — In m drii. (5-27)
i

Since we are assuming the system to be in thermal equilibrium during

the process, we may substitute for rii the Boltzmann expression, Eq. (5-10),
which yields
dS = — In (Ae pEi) drii
i

~ — &E(In A — $Et) drii

i

= In drii + kfiJ^Ei drii.

From Eq. (5-1),

X) drii = 0
i

when N is constant. Using Eq. (5-25), we then have

dS = k(3 dU. (5-28)

If we compare this expression with Eq. (5-24), we see that

5-12] THE EQUIPARTITION OF ENERGY PRINCIPLE 93

The k in Eq. (5-29) is the universal constant introduced in Eq. (5-23).

When we compare Eq. (5-29) with Eq. (5-18), where k was defined as the
Boltzmann constant of Chapter 4, we see that the expressions for /3 are
the same, so therefore the two k’s must be identical. Equation (5-29) is
a general one, for in deriving it we did not need to assume a system com¬
posed of monatomic gas molecules moving in three-dimensional space (an
assumption made in Sections 5-7, 5-8, and 5-9, but not in Section 5-6
or the present one).

5-12 The equipartition of energy principle, (a) Three degrees of freedom.

We have seen that for molecules free to move in three dimensions the
average K.E. of translation per molecule is

K.E./molecule = § kT.

(b) Two degrees of freedom. Suppose that we had molecules which, like
marbles in a flat dish, could move only in a two-dimensional plane. Equa¬
tion (5-19) would still apply, but in deriving the velocity distribution we
would proceed as follows.
In place of the two concentric spheres in velocity space, one of radius v
and the other of radius v + dv, take two concentric circles of radii v and
v -j- dv, respectively, these circles being in the -plane containing all possible
velocities. Then, for the fth zone,

ffdvxdvy = 2-kv dv,

the area between the two circles, and

We can take Ni = Nv dv and E = \mv2 as before, so that Eq. (5-19)

becomes 9

Nv dv = Cve

where C is independent of v.
For such a velocity distribution we have

K.E./molecule =
94 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

(c) One degree of freedom. Imagine molecules which, like marbles in a

tube, have only one degree of freedom. Here

V , —mv 12kT
gv = — dv, N v dv = De dv,

rx i 2 —mv2/2kT j
>mv e
J0 \mve dv
and K.E./molecule = ncc
00 _—tod2/ 2kT
So e dv

Jz
\m
Jo
= UT.

(d) / degrees of freedom. The three previous cases all indicate that the
average K.E. per molecule is %kT per degree of freedom. We shall show that
this is indeed so whenever the energy of a molecule is a quadratic function
of the velocity components, as is usually the case. Then each degree of
freedom acquires the same average K.E., namely, fkT.
Suppose, for example, that q is the coordinate associated with one of the
degrees of freedom, vq the corresponding velocity component, and %mv\
the K.E. associated with this velocity component. (Here m is a constant,
such as the mass or moment of inertia of the molecule.) To find the number
of molecules with a given velocity component between vq and vq -f dvq we
must take a zone of cells in a phase space of 2/ dimensions, this zone in¬
cluding all possible values of the coordinates and other velocities, but just
the exact variation dvq of vq. Then Nv dv will be proportional to dvq. For
the energy associated with this sheet of cells we may write

Eq = E -j- \mv\,

where E' is the average energy corresponding to the other degrees of free¬
dom. E is independent of vq, so that when integrating with respect to vq
we may consider e~E'lkT as a constant. We then have for the average
value of \mv\, that is, the average K.E. per molecule associated with the
velocity component vq,

So \mv\e E /kTe—mvql2kT dv.

E„ =
So e-E'lkTe-mvy2kT dvQ

I™ f00 ,.2
2m jo vqe — mvllZkT dv.
1
£ _ J2
~ST^~mvl^kT dvq = tm Jo

= ikT.
5-12] THE EQUIPARTITION OF ENERGY PRINCIPLE 95

Thus, according to the Maxwell-Boltzmann statistics, the mean K.E.

'per molecule associated with each degree of freedom is %lcT. This is the
principle of the equipartition of energy.

Example. Suppose that the molecules of an ideal diatomic gas are rigid, but
free to rotate about either of two mutually perpendicular axes, these axes being
at right angles to the axis of the molecule. Find Cv, Cv, and y for such a gas and
compare with experimental values of y for diatomic gases.

Solution. We have a gas with five degrees of freedom per molecule, three of
translation and two of rotation. From the equipartition of energy principle the

K.E./molecule = - kT,
Zi

and the

K.E./mole = - RT.
Zi

Hence

U = - RT,

dU 5
Cv =
dT 2 ’

Cp = Cv + R — 2

7
&-S-
At 15°C the values of y found experimentally are 1.41 for H2, 1.404 for N2,
1.401 for 02, 1.404 for CO, 1.41 for HC1, all in excellent agreement with the
theoretical value. _

The Maxwell-Boltzmann theory predicts that Cp, Cv, and 7 should be

independent of temperature. Such, however, is not the case. For example,
for hydrogen at very low temperatures (—200°C or less) we find Cv = § R,
q = 7 _ 5? jugt as for a monatomic gas. These results suggest that
at very low temperatures the rotational energy is frozen and when
energy is added to the gas the only result is to increase the translational
energy of the molecules. At room temperature, however, intermolecular
collisions cause frequent changes in both the rotational and translational
energy of a molecule. This temperature effect will now be briefly explained.
The translational K.E. of a molecule may vary continuously; for ex¬
ample, it may change from zero to a value only infinitesimally greater
96 CLASSICAL STATISTICAL MECHANICS [CHAP. 5

than zero. Such is not the case with the rotational energy. A study of
molecular spectra indicates that the rotational energy of a molecule is
limited to a discrete set of values, say 0, Ex, E2, E3 ■ • • . This is explained
theoretically by the quantum theory (Chapter 15). The successive allowed
values of rotational energy differ by finite amounts. As a typical example
we may take HC1, for which Ex is about 2.5 X 10-3 electron volts (ev).
At the end of Section 4-5 we found that at 0°C the average K.E. of trans¬
lation of a molecule is 3.5 X 10~2 ev; since this latter value is over ten
times E\, it is quite possible that in a collision between two molecules of
HC1 gas at 0°C rotational energy will be excited at the expense of transla¬
tional energy. In other words, relative to the mean translational K.E.,
the allowed rotational energy states are so close together that they may
be regarded as practically continuous, in which case the rotational degrees
of freedom must share in the equipartition of energy.
At very low temperatures, however, the mean translational K.E., which
is proportional to T, may be much less than Ex, the first allowed value of
the rotational energy. In this case the population of the first rotational
state, which is proportional to e~EllkT, must be very small. The rota¬
tional energy then will not contribute to the specific heat.
A nonrigid molecule containing n atoms (here considered as mass points)
must possess 3n degrees of freedom. Three of the degrees of freedom are
associated with the translational K.E. In general, a complex molecule has
three degrees of freedom of rotation (corresponding to rotation about three
mutually perpendicular axes), although for diatomic molecules the number
is only two, as stated in the previous example. This leaves (3n — 6) or
(3n — 5) degrees of freedom associated with the vibrations of nonrigid
molecules. To a first approximation these vibrations are simple harmonic
oscillations. Since a simple harmonic oscillator possesses an average
-potential energy equal to its average kinetic energy, if a vibrational degree
of freedom shares in the thermal energy, the mean total energy associated
with this vibrational degree of freedom must be kT rather than \kT.
According to quantum mechanics the energy of an oscillator is also
limited to a series of discrete values (Chapter 6), and it turns out that for
molecules these energy states are much farther apart than are the rota¬
tional ones referred to earlier. The result is that at room temperature the
population of any vibrational state (other than that of zero energy) is very
small, and so the vibrational degrees of freedom do not contribute to the
specific heat.
It is interesting to observe that throughout physics the internal structure
of particles does not play a role until energies are reached which are
comparable to the energy levels involved in the structure. This is the
reason why the hard-sphere model of a molecule works as well as it does
even though we know it to be a great simplification.
 97
i

Problems

1. In tossing a coin, let us consider 11. At 0°C find the number of oxy¬
that “heads” are in cell 1, and “tails” gen molecules per thousand with speeds
are in cell 2. Suppose that ten coins greater than 1000 m/sec.
are tossed, (a) Compute W for each pos¬ 12. Show that the number of gas
sible distribution from all heads to all molecules with ^-components of veloc¬
tails, (b) If the ten coins are tossed at the ity between vx and vx -f- dvx is equal to
same time, what is the percent proba¬ (N/^n){m/2kTyl2e-™xl2kT dvx.
bility that seven or more will land alike? 13. From Eq. (5-15) find, in terms
2. Given three cells and five parti¬ of kT, the most probable energy Ev.
cles. Find W for (a) n\ = 5, ri2 = 14. Zartman and Ko used in their
713 = 0, (b) Ul = 4, 712 = 1, 713 = 0, experiment a beam containing a mix¬
(c) 711 =3, 712 = 2, 713 = o, (d) ture of Bi and Bi2 molecules issuing
711 = 3, 712 = 713 = 1, (e) ni = 712 = from an oven at 850°C. (a) Find ?;rms
2, 713 = 1. for each of the two kinds of molecules,
3. Compute the percent error made (b) Taking the cylinder of Fig. 5-8 to
in taking In N! = N \n N — N when be 10 cm in diameter and rotating at
(a) N = 4, (b) N = 8, (c) N = 16. 3600 rpm, find where, relative to when
4. Refer to the Example at the end the cylinder is at rest, the Bi and Bi2
of Section 5-6. If 5ri2 = 2, find 5ni, molecules with rms speeds would strike
bnz, and W'. the glass plate G.
5. Given 100 particles and 10 cells 15. Consider polyatomic molecules
of equal energy. Find In W for (a) the with three rotational degrees of free¬
most probable distribution, (b) the dom. Find Cv, Cp, and y. List some
least probable distribution. experimental values of y for polyatomic
6. Given 100 particles and 10 cells. gases.
Find the ratio of W for the distribution 16. A nonrigid diatomic molecule
7ll = 712 = • • • = 778 = 10, 719 = 13, possesses six degrees of freedom (3 for
ni0 = 7; to the W for a completely each atom), one being vibrational. As¬
even distribution. suming simple harmonic vibrations,
7. Show that the most probable the K.E. and P.E. associated with the
speed for a molecule is given by vp = vibration are equal. If the vibra¬
y/2kT/m. Find the ratio vv/v rms* tional energy is in thermal equilibrium
8. Show that the mean speed is with that associated with translation
given by v = y/8kT/irm. Find the and rotation (as is actually the case
ratio v/vima. at high temperatures), find Cv, Cp,
9. At 0°C compute vp, v, and vTma for and y.
O2 molecules. 17. Compute the value of Cv for a
10. At 0°C find the number of solid, assuming that the molecules can
oxygen molecules per kg-mole with oscillate only three-dimensionally
speeds between (a) 100 and 101 m/sec, about fixed points. (The result,
(b) 300 and 301 m/sec, (c) 500 and Cv = 3R, is the law of Dulong and
501 m/sec, (d) 700 and 701 m/sec. Petit, which is in good agreement with
Using these values and that of vv from experiment at high temperatures.)
problem 9, plot Nv vs. v.
 CHAPTER 6

QUANTUM STATISTICS

6-1 Limitations of the Maxwell-Boltzmann statistics. The development

of statistical mechanics is typical of many branches of theoretical physics.
A theory is proposed and at first it may seem to be in very good agreement
with experiment. However, new experimental techniques and discoveries
tempt physicists to extend the scope of a successful theory to cover a
wider range of phenomena. Here the theory may meet with partial success,
but the chances are that it will completely fail to explain some of the new
observations. Of course the theory is still good for what it was originally
designed to explain, so it usually turns out to be a special case of a more
general (and probably a more complicated) theory. The older theory
serves as a starting point for the newer one, and the original theory must
be re-examined to determine which of its hypotheses should be eliminated
and what new ones introduced.
The Maxwell-Boltzmann (abbreviated henceforth as “M-B”) statistics
worked well when applied to the translational and rotational energies of
gas molecules at room temperature. Then as physicists were able to extend
measurements to very low temperatures, variations in the values of
Cv, Cp, and y were observed which the theory could not predict (see end
of Chapter 5). With the discovery of the electron about 1900, it seemed as
though the free electrons in a metal should behave in many ways like gas
molecules and the term electron gas was introduced. The M-B theory was
unsuccessful when applied to an electron gas. For one thing, the electrons
in a metal at room temperature do not contribute to the specific heat
capacity; Cy is approximately 3R for both metals and nonmetallic crystals,
yet, according to the M-B theory, Cv for metals should be 3R (see problem
17 of Chapter 5) + §R (if the number of free electrons is equal to the
number of atoms, as is approximately the case). Furthermore, later
measurements of photoelectric and thermionic emission indicated that the
Boltzmann distribution law did not hold for an electron gas.
Just after the discovery of the electron came that of the photon, the
elementary particle of light energy. Attempts were made to extend the
M-B theory to a “photon gas,” that is, radiation contained in a certain
space, such as the interior of a furnace. Here the old theory failed to ex¬
plain correctly the distribution of energy with respect to frequency
In examining these failures of the M-B theory we see that the theory
breaks down when applied to electrons and photons. These particles
have very much less mass than do gas molecules. It is also for just such
98
6-1] LIMITATIONS OF THE MAXWELL-BOLTZMANN STATISTICS 99

very light particles that the postulates of the quantum theory are
most significant. This suggests incorporating the fundamental principle
of quantum mechanics (Heisenberg’s uncertainty principle) into the
statistical theory of electrons and photons. At the same time, when
dealing with such minute particles, it became apparent that all electrons
are alike and indistinguishable, as are photons of the same energy. In fact,
should not this idea even be extended to gas molecules?
The study of the spectra of atoms with several valence electrons led to
two further discoveries, (1) the spin of the electron, and (2) Pauli’s exclu¬
sion principle. It was found that the electron possesses a spin angular
momentum of %(h/2r) and, for a given situation of the electron, two
orientations, opposite to each other, are possible for the spin vector.
To the other quantum numbers that specify the state of an atomic electron
a spin quantum number was added. The Pauli principle asserts that no
two electrons in an atom or molecule can have the same set of quantum
numbers, i.e., that no two electrons in a given system can be in the same
quantum state. Fermi and Dirac recognized that the Pauli principle
should also apply to the free electrons in a metal and they developed
quantum statistics for such particles (see Section 6-7). The so-called
Fermi-Dirac (F-D) statistics were extended later by Sommerfeld, with
highly successful results.
Further experimental evidence has indicated that the F-D statistics
are obeyed by any group of particles for which the individual spin is (a)
the same as for the electron (this includes protons, neutrons, positrons,
and n-mesons), or (b) an odd integral multiple of the electron’s spin. Parti¬
cles whose spins are zero or an even integral multiple of the electron s
spin are not subject to the Pauli principle, and so a group of such particles
must obey a different set of quantum statistics, that known as the Bose-
Einstein (B-E) statistics (see Section 6-4). Photons and x-mesons fall into
this category.
The question of which statistics gas molecules obey is answered by count¬
ing the number of particles (protons, neutrons, and electrons) in the
complex molecule; if this number is odd, the F-D statistics apply, while
if the number is even, the B-E statistics must be used. Actually, the
number is odd in only a few rare cases (e.g., monatomic beryllium vapor),
so we shall henceforth consider only molecules that obey the B-E statistics.
In a complex particle the spins of its components seem to be lined up
parallel or antiparallel, so that an even number of such components, each
with a spin of %(h/2x), gives a resultant spin of 0, 1, 2, • • • times (h/2x),
while an odd number of components gives a resultant spin of 2, 2>
times (h/2x). If we consider this resultant spin to determine the statistics
and then apply the rules of the preceding paragraph, we see that “even”
molecules should obey the B-E and “odd” molecules the F-D statistics.
 100 QUANTUM STATISTICS [CHAP. 6

6-2 The quantum principle. In an attempt to explain the distribution

of energy with respect to frequency for radiation inside a cavity, Planck
was led (in 1900) to make an important new hypothesis. He pictured the
radiation as being emitted and absorbed by simple harmonic oscillators
with various natural frequencies, emission of radiation corresponding to a
decrease and absorption to an increase in the energy and amplitude of an
oscillator. Planck’s fundamental hypothesis was that for an oscillator of
frequency v the energy cannot vary continuously, but is limited to the discrete
set of values 0, hv, 2hv, • • • , nhv • • • , h being a universal constant known
as Planck’s constant. Thus the total energy (kinetic plus potential) of such
an oscillator must be

E = nhv, n = 0, 1, 2, 3, • • • (6-1)

Let F = kx be the restoring force acting on an oscillator of mass m

and amplitude A. Then the energy equation* is

E = \mx2 + \kx2 = %kA2, (6-2)

and the frequency is given by

(6-3)

In Eq. (6-2) we may express the kinetic energy in terms of the mo¬
mentum px = mx instead of x. We then obtain

or

(6-4)

snuwii in rig. o-i. me area of the ellipse will be

(6-5)

energy equation states that the sum of the

ntion equals the total energy, which, for
leie equal to the potential energy at the
\j ui UllG 6 wing,
6-2] THE QUANTUM PRINCIPLE 101

Fig. 6-1. Phase space orbit of linear Fig. 6-2. Quantized phase paths for
oscillator. linear oscillator.

We now introduce Planck’s hypothesis, Eq. (6-1), and put

E = \ kA2 = nhv.

Since n must be an integer, this equation determines a discrete series of

amplitudes that the oscillator may have, namely,

A = \/2nhv/k.

Using Eq. (6-3) to eliminate v, we have

A2 = —=■ (6-6)
7rv mk

If we put this expression for A2 in Eq. (6-5), we get

px dx = nh, n = 0, 1, 2, 3, • • • . (6-7)

In Fig. 6-2 are shown several ellipses corresponding to consecutive values

of n. Note that the area between successive ellipses is always equal to h. Hence
as the oscillator emits and absorbs energy it may be pictured in the co¬
ordinate-momentum plane as jumping from one ellipse to another ellipse
which is separated from the previous one by the area h. Each ellipse
corresponds to a certain energy or state of the oscillator.
In Chapter 5 the state of a particle was specified by giving the cell in
phase space containing the particle. Instead of plotting the coordinates
and velocity components, let us now plot the coordinates and momentum
components to picture a particle in what will still be called phase space.
Then, for a linear oscillator, phase space becomes the x — px plane, and
cells are small areas in this plane.
According to Planck’s hypothesis there is no point in specifying the state
of an oscillator closer than the area h between ellipses in the phase plane,
for the oscillator never stops at points between two ellipses. However, if
 102 QUANTUM STATISTICS [CHAP. 6

we took our cells to be several times larger than h, they would include more
than one possible energy state. Therefore, for one-dimensional oscillators
the size of the cell should be h. This suggests that for six-dimensional
phase space the size of the unit cell should b eh3. In the last chapter we
assumed all cells to be of the same size, but we did not specify this size.
Now we see that quantum theory tells us what this size should be.
Planck’s ideas were extended by Einstein, who reasoned that since
energy is emitted and absorbed in units of hv, radiation itself must be com¬
posed of energy bundles, called photons or quanta. For light of frequency v
the energy per photon is taken to be hv. This hypothesis found excellent
confirmation in measurements of the photoelectric effect. These experi¬
ments also led to an accurate determination of h, namely,

h — 6.6237 X 10~34 joule-sec.

Equation (6-7) later became the basis of the Bohr theory of the hydro¬
gen atom, where it was generalized to apply to each degree of freedom.
In classical mechanics the laws of physics place no limitation on how
precisely we may specify the position and momentum of a particle. This
position and momentum, and hence the state of the particle, can be
represented by a point in phase space. In quantum mechanics, however,
we find that if a coordinate x is measured to within an uncertainty Ax,
then by the very process of this measurement an uncertainty Apx is introduced
into the corresponding momentum such that Ax Apx ~ h.
The above is a statement of Heisenberg's uncertainty or indeterminacy
principle. It is a generalization of the conclusion reached for the harmonic
oscillator on the basis of Planck’s quantum hypothesis, and it leads to
a confirmation of our postulate that for particles moving in three-dimen¬
sional coordinate space, the size of the unit cell in the corresponding
phase space should be h3. The state of a particle cannot be specified more
precisely than by saying that the point representing its position and
momentum lies somewhere within such a cell.

6-3 Hypothesis that particles are indistinguishable. We saw in the last

chapter that the size of the unit cell made no difference, so long as it was
small. This is not so, however, when we introduce the postulate that the
particles in a system cannot be distinguished from one another.
If the particles are all alike, interchanging particles between cells as
well as within a cell (see Fig. 5-2) must be considered not to alter the micro¬
state. Corresponding to each macrostate discussed in Chapter 5 there
wouid then be only one microstate, or IF - 1. Since these macrostates
all have the same probability, and since they represent the most detailed
description of a state, compatible with the uncertainty principle, we must
6-4] BOSE-EINSTEIN STATISTICS 103

change our viewpoint and our terminology. These former macrostates

will now be considered as our new micro states.
Let us again define a zone of cells as that group of unit cells which
corresponds to a range dE in the energy per particle, or to a range dv in
the speed of a particle. The ith zone will be taken to include all particles
with energies between E and E + dE, or speeds between v and v + dv.
The volume of such a zone in coordinate-momentum phase space will be
given by
////// dx dy dz dpx dpy dpz,

where the integration is taken to include all the cells in the given zone.
Let gz represent the number of unit cells and Ni the number of particles
in the ith zone. While dv is to be considered small compared with the
range in speed of the particles in the whole container, it is still possible to
consider it large enough for a zone to contain a great many cells and
particles. We may then regard <7; and Ni as large numbers and apply
Stirling’s theorem to gd. and Nil
The numbers N1, N2, N3, etc., which specify the number of particles
in each zone, now define a macrostate. The number of ways in which the Ni
particles of the ith zone can be distributed among its g{ cells will be desig¬
nated as Wi, and the thermodynamic probability of the macrostate will
be defined as
W = Wi ■ W2 ■ W3 ■ • • = IIWi,
i
where the product extends over all zones of phase space.

6-4 Bose-Einstein statistics. We shall proceed to find the equilibrium

distribution of N particles assuming the postulates of the last two sections
plus those of Section 5-4. Just as we did for the M-B statistics, we take
the variations in the Ni’s, resulting from the thermal motion of the particles,
to be subject to three conditions: (1) 5(ln W) = 0, because W is a
maximum for the state of equilibrium; (2) 5N = 0, because N is constant;
and (3) SU = 0, because U is constant.
We have seen in discussing Maxwell’s distribution law (Section 5-7) that
the number of particles in the ith zone is proportional to the a priori
probability gz of that zone. It is simplest to commence by assuming zones
all of the same size g and then, when finding Ni; to multiply the distribu¬
tion function so obtained by gjg. So let us for the moment say that the
ith zone has g cells. Although the cells are considered to have an identity
and the particles not, it will help us to see how Wi is computed if we call
the cells 1, 2, 3, • • • up to g, and the particles a, b, c, • • • up to N{. For
example, a possible arrangement would be
 104 QUANTUM STATISTICS [CHAP. 6

c b, d a, e
(6-8)
Cell 1 2 3 4

which can also be represented by the sequence Ic236d4ae • • • . The number

of such arrangements or sequences can be computed as follows. The
sequence must start with the number of a cell and there are g such numbers;
the remaining numbers and letters, of which there are (g -\- Ni — 1),
can be arranged in any order, which means in (g + Ni — 1)! ways;
altogether, then, we can write

g(g + Ni — 1)! (6-9)

sequences such as the above. Included among these would be

b, d c a, e
(6-10)
Cell 3 2 1 4

which is the same as (6-8), since the same cells contain the same particles.
The cells (with their respective particles) can be interchanged like this in g\
ways, all of which correspond to the same microstate, so that when we
come to counting different arrangements we must divide (6-9) by g!.
Since the particles are now to be regarded as indistinguishable, an
arrangement such as

b c, d a, e
(6-11)
Cell 1 2 3 4

is also the same, as (6-8), because interchanging two identical particles

cannot make a difference. The particles can be interchanged so as to give
Ni\permutations. Hence in finding W{ we must also divide (6-9) by NI
Note that the cells are not alike (they refer to different positions and
momenta), so that

c b, d a, e
(6-12)
3 2 1 4

is different from (6-8).

6-4] BOSE-EINSTEIN STATISTICS 105

We see then that

w N j — 1)!
g\Ni\

(g + Nj - l)!
(6-13)
(g - i)!^<!

When we combine these arrangements with each of those for all other
zones, we get

w = n Wi = n (g(g+-Nii)!tff!
i i
- 1)!
‘
(6-14)

As before, it is easier to deal with In W than with W. We then have

In W = L (.9 + Ni — 1)1 — In (0 — 1)'- — In JV<!].

i

As in the derivation of the Boltzmann distribution law (Section 5-6), we

apply Stirling’s theorem. Then, letting each of the Ni s change by a small
amount SN{, we find 5 (In W) and set this equal to zero. These mathe¬
matical steps (see problem 4 at the end of this chapter) lead to

g -\- N j l
2> Ni
8N i = 0. (6-15)

(The —1 can be neglected compared with g + Ni.)

Taking N and U constant yields

L SNi = 0, (6-16)

£ Ei 8Ni = 0, (6-17)

as in Section 5-6. We now apply Lagrange’s method of undetermined

multipliers and multiply Eq. (6-15) by unity, Eq. (6-16) by —X, and Eq.
(6-17) by —18. Upon adding the equations thus obtained, we get

8Ni = 0.

Since each 8Ni is independent, we must have

106 QUANTUM STATISTICS [CHAP. 6

or

(6-18)

and we see that for any zone the ratio of the number of particles to the
number of cells in the zone is independent of g. If the zones have different
numbers of cells, say <7; in the fth zone, then Ni will be equal to Qi times
the ratio given in Eq. (6-18). We then obtain

Qj
Ni = ex+PEi — 1 ’
(6-19)

which is the Bose-Einstein distribution function. It differs from the Boltz¬

mann function of Eq. (5-11) in that it contains the —1 in the denominator.
If the —1 can be neglected as small compared with the exponential term,
then we need only let e~x —- A. to obtain the Boltzmann distribution of
Eq. (5-11).
Just as for the M-B statistics, the conditions ^ N{ = N and E{N{ =
U determine the constants X and (3. By following the same procedure as in
Section 5-11, i.e., putting dS = dU/T = k d(ln W), we again find (see
problem 5 at the end of this chapter) that

/3 ^ ji' (6-20)

6-5 Quantum statistics of a gas. Einstein realized that the postulates

leading to Eq. (6—19) should hold for gas molecules. Since at room temper¬
atures gas molecules obey the M-B statistics very well, it must mean that
in this case the —1 in the B-E expression can be neglected, or ex+pEi A> 1.
Since this is true for all energies Ei} including those near zero, ex must be
large, or

A = ^ « !• (6-21)

We shall compute A for a gas at room temperature and show that con¬
dition (6—21) is actually satisfied. Since such a gas is found to obey the
M-B statistics, we may write

Ni — Agie El,kT. (6-22)

As in the derivation of the Maxwell velocity distribution, we let Ni =

Nv dv, the number of molecules with speeds between v and v -f dv. This
6-5] QUANTUM STATISTICS OF A GAS 107
i

is the number of molecules in the fth zone in phase space; Qi, the number of
unit cells in this zone, is equal to the volume of the zone divided by h3.
To find the volume of this zone we must first remember that it includes
all possible values of the coordinates so that (refer to the sextuple integra¬
tion in Section 6-3)
/// dx dy dz = V,

V being the volume of the gas. In momentum space the zone will include
the volume between two spheres, one of radius p and the other of radius
p -f- dp, where
p — mv

is the scalar magnitude of the momentum of any molecule with the speed
v, and
dp = m dv.

The volume of our zone in momentum space is then

fff dpx dpy dpz = 47i-p2 dp

= 4:ttvi3v2 dv.
Hence

gi = gv = ^ SSfSff dpx dpy dpz dx dy dz

47rm3T 2
v dv.
h*

Now for this zone Ei — Ev — \mv2, so Eq. (6 22) becomes

47rm3FA 2 — mv2/2kT
Nv dv = v e dv, (6-23)
h3

which is the Maxwell volocity distribution law, Eq. (5-14), with the
constant of proportionality
47rm3 V A
B = (6-24)
h3

The only difference between the above expressions and those derived in
Section 5-7 is the presence of the m3 term, which enters because our phase
space is a momentum-coordinate rather tlian a velocity-coordinate one.
The size of the unit cell, h3, appears also, but we shall see that A is pro¬
portional to h3, so that actually Planck’s constant cancels out and its
magnitude here is immaterial.
 108 QUANTUM STATISTICS [CHAP. 6

Using the condition Ni = N, we replace the summation by an

integration and have

(^m3VA\ 1 /8 fffcaya
N = Nv dv = BJ 2
~ \ h* ) 4 \ m3
or
Nh3
A = (2TmkT)-312. (6-25)
V

For nitrogen at standard conditions N = 6 X 1026 molecules/kg-

Example.
mole, V = 22.4 m3, T = 273°K, m = 28/(6 X 1026) = 4.67 X 10“26
kg/molecule, k = 1.38 X 10~23 joule/molecule-degree, h = 6.6 X 10~34 joule-
sec. When we substitute these values in Eq. (6-25), we find

A « 2X 10-7,
so that A <5C 1.

Let us now consider the case when A might be much larger than in this
example, so that ex = 1/A is not large enough to allow us to neglect the — 1
term in Eq. (6-18). We see that A increases as (1) T decreases, and (2)
m decreases. The effect of decreasing the temperature is twofold because
of the T ^2 factor and because V decreases with T at constant pressure.
The best gas to use in testing for a departure from the M-B theory is
helium, for it is one of the lightest gases and it can be cooled to very low
temperatures before it condenses. Low-temperature measurements with
helium have shown departures from the M-B theory in the direction in¬
dicated by Eq. (6—18) j the gas is then termed degenerate. Thus we can
regard the M-B statistics as a special case of the B-E statistics, this special
case including gas molecules at all but very low temperatures.

6-6 Statistics of a photon gas. Planck’s radiation law. Suppose that a

cavity of volume V contains radiation in equilibrium with the walls at the
temperature T. Let uv dv represent the radiant energy per unit volume
in the frequency range between r and „ + dv. Our problem now is to find
uv as a function of v and T.
This problem was first solved by Planck. He found an empirical formula
that fitted the experimental data on the energy-frequency distribution in
radiation from a cavity. He then attempted to derive this formula theo¬
retically , picturing the radiation in an enclosure as being reflected back
and forth between the walls and also being emitted and absorbed by
oscillators in the walls.
In a stretched string of length L only certain waves can exist as standing
waves namely, those of wavelength 2L, 2L/2, 2L/3, etc. Similarly, in a
three-dimensional enclosure only certain wavelengths or frequencies will
6-6] STATISTICS OF A PHOTON GAS. PLANCK’S RADIATION LAW 109

be reinforced by reflection and form standing waves. From purely geo¬

metrical considerations,* we find that the number of such possible fre¬
quencies between v and v + dv is 4ttv2 dv/c3 per unit volume, where c
is the speed of light. For light this number is doubled because, for a given
frequency, light waves (being transverse) can be polarized in two inde¬
pendent ways (circularly clockwise and circularly counterclockwise).
Letting Nvdv be the number of modes of vibration in unit volume between
the frequencies v and v dv, we have

Nv dv — 8irv2 ^ • (6-26)
c3
If we were to assume the equipartition of energy, we would assign an
average K.E. of \kT and an average total energy of kT to each member of
a group of linear oscillators in thermal equilibrium. Each wave is
statistically equivalent to an oscillator, so we would then associate an
energy kT with each mode of vibration in Eq. (6-26). However, realizing
that this would not lead to his empirical law, Planck made the assumption
stated in Section 6-2, that energy variation is limited to a discrete set of
values.
Let us use the Boltzmann distribution law, Eq. (5-10), with /3 = 1 /kT,
and, following Planck, impose the restriction that E can only have one of
the following values: 0, hv, 2hv, • — . The average energy associated
with the frequency v is then

_
E n=0 nhvAe
oo a 7 —7i hv I kT

CJv ~ E:=o Ae-^'kT (6-27)

With the substitution x = hv/kT, this becomes

Ev = hv

e~x + 2e~2x + 3e~3x +

= hv
1 + e~* + e~2x H-

e-x(l - e~xr2
= hv
(1 - e~x) —i

= hv
1 — e~x
1
= hv
ex — 1

hv (6-28)
ehvjkT — 1

*See Joos, Theoretical Physics, pp. 572-574.

no QUANTUM STATISTICS [CHAP. 6

Planck used this quantum partition of energy in place of kT and obtained

u, dv = E.N. dv = ■, (6-29)

the law which he had already found empirically and which is known as
Planck’s radiation law.
We shall now follow the method of Bose and show that Planck’s law can
be obtained from quantum statistics without recourse to such classical
concepts as oscillators or standing waves. We shall, however, assume
Einstein’s hypothesis that light is composed of photons such that the

Energy/photon = hv (6-30)
and the
Momentum/photon = hv/c. (6-31)

The last relation follows from Einstein’s famous relationship E = me2

(see the theory of relativity), so that for a photon the momentum me =
E/c — hv/c.
The statistics of a photon gas are identical with the Bose-Einstein
statistics of Section 6-4 except that Eq. (6-16) no longer applies, that is,
SN = Yli SNi is no longer necessarily zero. To understand this point, we
must remember that the symbol <5 refers to a variation from the state of
equilibrium to a slightly different state. This variation must be one that
could actually occur during the thermal collisions of the particles with each
other and with the walls of the container. During such collisions the gas
molecules cannot disappear, nor may they be created. On the other hand,
a photon may, on striking a wall, be absorbed and so disappear, while new
photons may be emitted by the walls through the process of radiation.
Thus photons may be created (emitted) and destroyed (absorbed).
We shall still impose the restriction that the only variations from equilib¬
rium to be considered are those for which 8U = 0. The total energy may
remain constant, even though the number of photons does not, through
processes such as the following. A single photon of frequency 2v and energy
2hv may be absorbed by the walls at the same time that two photons, each
of frequency v and energy hv, are emitted. The reverse process is also
possible. We shall investigate the distribution of photons with respect to
frequency, subject to the condition that U is constant.
Reviewing Section 6-4, we then have the conditions 8(In W) = 0 and
dU = 0, but not 8N = °- The omission of Eq. (6-16) amounts to taking
X = 0 and

A = ~= 1 (6-32)
6-6] STATISTICS OF A PHOTON GAS. PLANCK’S RADIATION LAW Ill
1
for a photon gas. In place of Eq. (6-19), we have

N‘ = JSifzri - (6-33)
where again /3 = 1/kT.
We now let Ni be the number of photons in the volume V, and nv dv
the number per unit volume, between the frequencies v and v + dv, or

Ni — Vnv dv.

We compute gi as before, except for the factor of 2 that must again be

introduced, corresponding to the two independent directions of polariza¬
tion. Writing
2V 2 i \
gi = -p- (4ttp dv),

we now let p = hv/c, dp = h dv/c, so that

ffi = v dv-

Finally, the photons of frequencies close to v all have practically the same
energy
Ei = Ev = hv, (6-34)

so that Eq. (6-33) becomes

87TV dv
_
nv dv = (6-35)
g3 ghvIkT — 1

This is the number of photons per unit volume with frequencies between
v and v + dv. It will be seen that Eq. (6-35) differs from (6-26), which
was based on the classical concept of standing waves and did not contain
the temperature.
The energy density of the radiation for the frequency range v to v + dv
is the product of nv dv and the energy per photon, or

7 \ Ti 8irhv dv /{%
Uv dv = (n„ dv)E, = —ehv/kT _ 1» (b-ob)

which is identical with Eq. (6-29), Planck’s radiation law. It is interesting

to note that while the product of the two factors Ev and N, dv (or nv dv) is
the same, the factors themselves are different in the two derivations (that
of Planck and that based on the Bose-Einstein statistics).
The fact that A = 1 for a photon gas means that for such a “gas” the
M-B approximation will never hold, and only quantum statistics can be
used.
 112 QUANTUM STATISTICS [CHAP. 6

6-7 Fermi-Dirac statistics. As stated in Section 6-1, the statistics of an

electron gas differ from the Bose-Einstein statistics in two respects:
(a) Electrons possess a spin. Hence, two electrons with the same co¬
ordinates and momenta (within the limits of the uncertainty principle)
but with opposite spins can be distinguished as being in different states.
These two electrons must then be regarded as being in different cells in
phase space. Each of our previous cells of volume h3 must now be sub¬
divided into two cells, each corresponding to the same position and mo¬
mentum, but to opposite spins. Thus we postulate that the size of the unit
cell is h3/2.
(b) While in the B-E statistics no restriction was made as to the number
of particles that could be in the same cell, we now postulate that no cell
can contain more than one particle. This hypothesis is an expression of Pauli’s
principle, which states that no two electrons in the same system can be in
identical states. Application of this principle to the electrons in an atom
led to a successful explanation of the periodicity, shell structure, and
chemical- properties of the elements. The results of applying Pauli’s
principle to an electron gas are equally good.
Consider again zones of cells (each of size h3/2) and, for simplicity, let
each zone contain the same number of cells. Let g denote the number of
cells per zone and assume g 1. Let IV; be the number of particles in the
*th zone. (From Pauli’s principle, A* cannot be greater than g.) Call W{
the number of distinguishable ways the IV; particles may be arranged among
the g cells, with never more than one particle to a cell, and set

W = Wx • W2 ■ W3 ■ • • = UWi.
i

In computing W; the easiest method is to divide the cells into two groups,
group A including all occupied cells (one particle each), and group B all
unoccupied cells (no particles). There will be IV; cells in group A and
g — Ni in group B. Let us number the cells 1, 2, 3, etc. We might have
a distribution such as

Group A Group B

□ □□-
(6-37)
Cell 2 3 6 1 4 5
_j L_
■ y _j
Ni cells (9 — Nf) cells
6-7] FERMI-DIRAC STATISTICS 113

If we try rearranging the cells, we find $! possible distributions in all.

However, a rearrangement involving interchanging two cells in the same
group is not to be considered as distinguishable from the original arrange¬
ment, for the order in which we list the cells in a group is immaterial.
Thus the distribution

Group A Group B

(6-38)

Cell 6 3 2 1 5 4

is the same as (6-37). The cells in group A may be changed about in Nil
ways and those in group B in (g — Ni)! ways. Hence

Q' (6-39)
Ni\(g - Ni)\ ’

ni Nil(g gl- Ni)!

(6-40)

Having found the expression for IT, we proceed as we did with the M-B
and B-E statistics. Variations in the AVs from their values for the state
of equilibrium are taken to be subject to the three conditions (1) 8W -
5 (In W) = 0 because IT is a maximum, (2) 8N = 0 because N is constant,
and (3) bU = 0 because U is constant.
From Eq. (6-40), we have

In IT = £ [In g\ — In Nil — In (g — Ni)!]

i

Using Stirling’s theorem, this becomes

In IT = £ [g In g - g - Ni In Ni + N<
%

— (g — Ni) In (g — Ni) + g — Ni]

= £ [g In g - Ni In Ni - (g - Ni) In (g - Ni)].
 114 QUANTUM STATISTICS [CHAP. 6

If we let Ni vary by SN{, but keep g constant, the three conditions hsted
above become

S(ln W) = E In SNi = 0,
i %

SN = X SN i = 0,
i

8U = X EibNi = o.
i

As in Section 6-4, we multiply the first equation by unity, the second

by —X, and the third by —/?, and then add, which yields

g - Nj
E Ni
8N i = 0.

Since the SNi s are independent,

g - Nj
X - pEi = 0,
Ni

Ni = 1
g ex+?Ei 1

The number of particles in a zone of gi cells is given by

^i e^+PEi + 1 ’ (6-41)

which is the Fermi-Dirac distribution function. It differs from the B-E

function of Eq (6-19) in that (1) g{ is the number of cells of size h3/2
(rather than h ) in the z'th zone, and (2) the denominator contains +1
(rather than —1). As in the case of the B-E statistics, the F-D statistics
are ind1stinguishable from the M-B when ^ » 1. Thus for gas molecules
at ordinary temperatures any one of the three statistics may be used - they
all apply equally well. ’
Again the relation dS = dU/T = k d(ln W) leads to

^ = kf' (6-42)

6-8 Electron gas If we take e* » 1 and proceed as in Section 6-5,

the final expression for A = 1/e* will be the same as Eq. (6-25) except for
a factor of one-half due to the change in the size of the unit cell, i.e.,

Nh
A 2~y (ZTrmkT) -3/2 (6-43)
6-8] ELECTRON GAS 115

For an electron gas consisting of the free electrons in a metal, this formula
will lead to a much larger value of A than we computed before because (1)
m is much smaller, and (2) N/V is much greater. The mass of a nitrogen
molecule is 28 X 1840 times that of an electron. In a metal there is ap¬
proximately one free electron per atom, but the density of atoms in a solid
is several thousand times that of the molecules in a gas at standard condi¬
tions. Thus even at room temperature we cannot assume A « 1 for the
free electrons in a metal. Such an electron gas is said to be completely
degenerate, that is, at laboratory temperatures it never obeys classical
statistics.
To picture the distribution for an electron gas, let us start with absolute
zero. As the temperature of an ordinary gas is decreased toward abso¬
lute zero, its molecules crowd more and more into the cells of very low
energy, approaching the condition where they would all be in the cell of
lowest (zero) energy. However, because of Pauli’s principle, electrons can¬
not crowd into one or a few cells. The distribution leading to the lowest
total energy (the one we would expect to be associated with absolute zero)
is that where the cells of lowest energy are all solidly filled, each with one
particle. Enough of these cells from zero energy up to, say, Em must be
occupied to accommodate all of the electrons, as shown in Fig. 6-3(a).
All cells of higher energies (A; > Em) will be empty. Such a distribution
is shown by the solid rectangular curve in Fig. 6-4. Thus an electron gas
possesses a so-called zero-point energy at absolute zero.

Et

Ni
9i

T = 0°K Tx t2>t1
(a) (b) (c)
Fig. 6-4. Fermi-Dirac distribution
Fig. 6-3. Variation in cell distribu¬
tion with temperature, for electrons. function.

Picture a metal near absolute zero being given enough energy to raise
its temperature one degree. Very little of this energy will be transferred
to the electrons in the metal, for the average thermal energy of the mole¬
cules C’kTj is not enough to disturb those electrons which are not near the
 116 QUANTUM STATISTICS [CHAP. 6

top of the filled cells. Such electrons would have to jump all the way up
to the unfilled cells above Em. However, a little energy might be absorbed
by electrons in those cells just below the highest one filled, raising them to
nearby unfilled cells, as shown in Fig. 6-3 (b) and (c). As the temperature
rises, more and more of the “knee” of the curve in Fig. 6-4 will be rounded
off, as shown by the dotted curves, which begin to resemble the exponential
shape corresponding to the Boltzmann distribution. However, suppose
that even at room temperature the rounding-off affects only a very slight
portion of the zero-point rectangular curve. This would mean that only a
negligible fraction of the free electrons share in the thermal motion, and so
their contribution to the specific heat would be small. The fact that the
specific heat of a solid at ordinary temperatures does not depend on whether
it possesses free electrons can thus be explained by the F-D, but not by
the M-B, statistics. Detailed calculations show that only at temperatures
of the order of 106 degrees will the electrons packed in the lower occupied
energy states be loosened and so contribute to the specific heat.
The solid packing of the energy states below E{ = Em corresponds math¬
ematically to X < 0 in Eq. (6-41). This would mean that ex < 1,
f^>1. Approximate calculations based on Eq. (6—43), assuming
one free electron per atom, also lead to A ]> 1. If we put

X = -EJkT,
Eq. (6-41) becomes
9i
Ni = (6-44)
e(E~Em)lkT 4 1
- ’

which, we shall see, corresponds to Fig. 6-4. Let N{ refer to just one cell
so that gt 1. If T is small, or 1/kT is large, the exponential term is
practically zero when E{ < Em and nearly infinite when Ei > Em; there¬
fore if we plot Ni vs. Ei, we will obtain a flat curve with a sharp knee at Em
as in Fig. 6-4. ’

The actual energy distribution depends on the distribution per cell

and on the number (gi) of cells between, say, E and E + dE. As we found
for the M-B statistics, this latter factor is proportional to E\ as in Eq
(5 15), so that we now have for the number of electrons with energies be¬
tween E and E E dE
El
Ne dE * ~e(E~Em)lkT I dE- (6-45)

for ^6 TY distribution Unction Ne has been plotted against E

for 0 K and two higher temperatures. It will be seen that this curve also
has a sharp knee and that at the highest temperature shown only a small
percent of the electrons have been thermally disturbed.
For an electron to escape through the surface of a metal it must be able
6-9] SUMMARY AND COMPARISON 117
1

Fig. 6-5. Energy distribution function in F-D statistics.

to perforin a certain amount of work, known as the work function of the

metal. The thermionic emission of electrons from a hot metal will depend
on how many electrons in the metal have energies equal to or greater than
the work function, and this depends on the distribution function. Calcula¬
tions based on the F-D distribution show the best agreement with measured
values. The above is also true in the case of photoemission, an effect in
which the electrons receive from photons sufficient additional energy to
escape from a metal, even at low temperatures.

6-9 Summary and comparison. The points distinguishing the three

sets of statistics, M-B, B-E, and F-D, are shown in Table 6-1, and the

Table 6-1

M-B B-E F-D

Molecules distinguishable a/

Uncertainty principle V v/
Pauli’s exclusion principle v/

postulates assumed in each are checked. The applicability of each is shown

in Table 6-2.
Table 6-2

Near 0°K 300°K 106 degrees

Gas molecules B-E M-B M-B

Photons B-E B-E B-E

Electrons F-D F-D M-B

To illustrate further the essential differences between the F-D, B-E,

and M-B statistics, let us consider a numerical example and compute the
thermodynamic probability W according to each.
118 QUANTUM STATISTICS [CHAP. 6

Example. Given two particles to be divided among three cells, two of the
cells being in zone 1 and one in zone 2. Compare, using each of the statistics, the
probability for (a) Ni = 2, N2 = 0 with that for (b) Ni = 1, N2 = 1, and
that for (c) Ni = 0, N2 = 2.

Solution. We have g\ = 2, <72 = 1.

92!_
(1) Fermi-Dirac: W = W\ W2 = .. ,,-:——
Ni\(gi — N1)! ■A/2! (92
— IV2)!
Zone 1 Zone 2
2! 1!
(a) Wa
2!0! ' 0!1!

2!
(b) Wb
Tm

(c) N2 — 2 is impossible, since zone 2 has only one cell and two particles
cannot be in the same cell.
Wc = 0.

Wa : Wb: Wc = 1:2:0.

(2) Bose-Einstein: W = W1W2 = _HI . (.92 ~l~ N2 1)!

(gi - l Ni
)\ \ (92 - l)!iV2!

3! 0!
(a) Wa
1!2! ' 0!0!

2! 1!
(b) Wb
111! ’ 0!1!

1! 2!
(c) Wc
1!0! ’ 0!2!

Wa : Wb: Wc = 3:2:1.
6-9] SUMMARY AND COMPARISON 119
i

(3) Maxwell-Boltzmann: Here we must consider not zones but cells of equal
size. If there are n\ particles in the first cell of zone 1 and m in the second,

N\
W = £
7ii!ri2hV2!

the summation covering all the ways in which n\ + 112 = IV i.

(a) If Ni = 2, we can have n\ = 2, ri2 = 0, or m = 1, »2 = 1, or

ni = 0, U2 = 2. Hence

ab

a b
2! 2! 2!
Wa = 4
2!0!0! + 1!1!0! + 0!2!0!
b a

ab

(b) If IVi = 1, we can have ni = 1, U2 = 0, or n\ = 0, ri2 = 1.

(c) If Ni = 0, we must have n\ = n2 = 0.

2! ab
Wc = 1
0!0!2!

Wa :Wb-We = 4:4:!.
 120 QUANTUM STATISTICS [CHAP. 6

Problems

1. Suppose that with an electron 9. From Eq. (6-36) show that the
microscope the position of a particle energy per unit volume integrated for
with respect to the X-axis is determined all frequencies is proportional to T4.
to within an accuracy of 10-10 m (1 A). Given f™x3dx/(e* - 1) = tt4/15,
Find the velocity range Av x per cell in show that the proportionality constant
phase space for particles of mass (a) is a = 87t5A:4/15c3/i3.
9 X 10-31 kg (electron), (b) 2 X 10~24 10. From Eq. (6-36) find the ex¬
kg (large molecule), (c) 3 X 10-15 kg pression for u\ d\, the energy density
(small oil drop). between the wavelengths X and X -f- dX.
2. Given 4 particles to be divided Differentiate and show that when u\
among 2 zones, one of 3 cells and one is a maximum the product XT' is inde¬
of 2 cells. For B-E statistics find W pendent of T (Wien’s displacement
for each macrostate from Ni = 4, law).
N2 = 0 to N1 = 0, N2 = 4, using 11. Show that the product XT7 in
both the formula and block dia¬ problem 10 is approximately equal to
grams. he/5k.
3. Repeat problem 2 for F-D statis¬ 12. Experimentally we find 7.64 X
tics. 10~16 joule/m3(°K)4 for the constant
4. Show in detail the derivation of a in problem 9, and 2.88 X 10~3 m-°K
Eq. (6-15) and Eq. (6-18). for the product XT of problems 10 and
5. Show that /3 = 1/kT for the (a) 11. From these values determine h
B-E statistics, (b) F-D statistics. and k.
6. Find the temperature at which 13. Use Eq. (6-43) to compute A
A = 0.5 for an ideal gas composed of for an electron gas, taking m = 9 X
helium molecules. (Take p = 1 atm.) 10~31 kg and N/V = 6 X 1028 free
7. Show that if the summations in electrons/m3, T = 300°K.
Eq. (6-27) are replaced by integrations 14. In Eq. (6-44) let gi = 1, T =
and n is considered to vary contin¬
300°K, and Em = 9.0 electron volts
uously from zero to infinity, then (tungsten). Find for (a) =
Ev = JcT.
8.9 ev, (b) Ei =. 9.0 ev, and (c) E{ =
_ 8. Show that Eq. (6-28) reduces to
9.1 ev. Note the sharpness of the knee
X, = kT if hv/kT « 1.
of the distribution curve.
 1

i t

Part II
ELECTROMAGNETISM

%
 CHAPTER 7

FUNDAMENTAL LAWS OF ELECTROMAGNETISM

7-1 Introduction. As with most branches of physics, the theory

of electromagnetism has developed in an inductive-deductive manner.
A wide variety of experimental facts has been observed and recorded. A
mere statement of all these facts would be bewildering, but it has been
found that Nature operates in a manner that can be described in terms of
a limited number of “laws, ” and that what makes the behavior of our world
seem complicated is the fact that these laws apply in such a wide variety
of circumstances. The discovery of these laws has been the work of men
of genius. In an elementary course it is advisable to trace some of the steps
leading inductively from experimental observations to general laws; in a
theoretical treatment we commence with a statement of these laws and
then show how they may be applied deductively to specific problems.

7-2 General outline. The fundamental laws of physics are those which
do not involve the special properties of a particular substance. In electro¬
magnetism such laws must be those that apply to charged bodies situated
in empty space, although the experiments leading to the discovery of these
laws were usually conducted in air. Fortunately, the presence of air makes
very little practical difference; therefore we shall divide our treatment of
electromagnetism into two parts and consider (1) the laws for charged bodies
in vacuo, or in a gas such as air at atmospheric pressure, and (2) the
modifications in these laws when the medium is a liquid or solid possessing
particular electric or magnetic properties. The first part, (1), dealing with
general laws, will receive the most attention; we shall include in the present
chapter a brief statement of these laws, and then we shall devote a separate
chapter to the development and applications of each (Chapters 8-11).
The second part, (2), will comprise Chapter 12, further development of
this more complicated branch of electromagnetism being left to more ad¬
vanced texts.
After briefly sketching the historical development of electromagnetism,
we shall try to present the subject in the most logical manner possible.
The historical growth of electromagnetism was much like putting together
the different sections of a jigsaw puzzle and then finally fitting these
sections together to form the complete picture. Now, of course, we can
start with a look at the picture as a whole and then make a detailed study
of its parts.
123
 124 FUNDAMENTAL LAWS OF ELECTROMAGNETISM [CHAP. 7

7-3 Historical summary. The Greeks had some knowledge of electricity;

for instance, they knew about the electrification of amber by friction.
However, the Greeks did not conduct experimental tests to find out how
Nature behaves. Instead they speculated about why things happen. This
led them more into the realm of philosophy than science and caused them
to make guesses, most of which were wrong.
The Romans gave little time to science, and in the centuries of the
Middle Ages an interest in learning was barely kept alive. Science as we
think of it today dates from about the middle of the sixteenth century,
that is, from the time of Copernicus and Galileo. Scientific attention was
first directed toward simple mechanical problems. A study of the more
complicated phenomena of electricity and magnetism did not get under
way until about the year 1600, when Sir William Gilbert, court physician
to Queen Elizabeth, published his book De Magnete, in which he described
his experiments in electrostatics and magnetism.
The work of Gilbert was followed by the discovery of two kinds of
electrification, by Franklin’s investigation of lightning, and by other
qualitative studies. Credit for the first quantitative work in electricity
and magnetism must go to Coulomb who, just before 1790, discovered the
law governing the force between static charges. This general principle,
which we call Coulomb s law, coordinated a large number of observed facts
and opened the way for a theory of electrostatics. Coulomb discovered a
similar law for the forces between magnetic poles and from this developed,
in a parallel manner, the theory of magnetostatics.
With the discovery of the voltaic cell, a source of electric current became
available, and the properties of such currents were then investigated. A
most important result was the observation by Oersted in 1819 that current
m a wire will affect a compass needle placed over the wire. This was the
first conclusive proof of an interrelationship between electricity and
magnetism. Oersted’s discovery was soon followed by quantitative experi¬
ments by Ampere and others. Ampere was a good experimentalist who was
also able to find the general principle that explained the facts which he
observed. This fundamental law, Ampere’s law, was first published in
1822; it expresses the force of interaction between short sections of two
current-carrying conductors. From this expression the force between
complete current circuits can be computed. Ampere’s law was expressed
m a different form by Biot and Savart, who found an expression for the
orce on a magnetic pole due to the current in a neighboring conductor.
A second interrelationship between electricity and magnetism was dis¬
covered by Faraday in 1831. Faraday reasoned that if electric currents
produced magnetic effects, then magnetism should produce electric effects
He was led, after considerable experimentation, to the discovery of electro¬
magnetic induction. Like Ampere, he summarized his results in the form
7-3] HISTORICAL SUMMARY 125

of a general principle which we call Faraday’s law. From this law we can
compute the electromotive force induced in a circuit when the time rate
of change of the magnetic flux through the circuit is known. Faraday re¬
garded electric and magnetic forces as due to fields, and he pictured these
fields by drawing so-called lines of force to represent the direction and
intensity of a field.
Maxwell, in 1864, brought the theory of electromagnetic fields to its
culmination. Fie summarized, in the form of differential equations, the
laws governing electric and magnetic fields. These electromagnetic equa¬
tions embody the laws of Coulomb, Ampere, and Faraday, and they are
compatible with the continuity 'principle that electric charge can neither
be created nor destroyed. In his attempt to make his equations consistent
with that expressing the continuity principle, Maxwell introduced a term
called the displacement current into one of his equations. This term was
not predicted by the laws of Coulomb, Ampere, and Faraday, but its inclu¬
sion in his theory led Maxwell to the belief that electromagnetic waves
should exist. He computed the speed of such waves to be that of light and
so concluded that light and electromagnetic waves must be the same thing.
The experiments of Hertz on the properties of electromagnetic waves
confirmed Maxwell’s theory.
Since Maxwell’s time the point of view taken by physicists toward elec¬
tromagnetism has changed. First, since isolated magnetic poles have not
been observed, all magnetic effects are attributed to charges in motion,
in line with a suggestion made by Ampere. The magnetism of a piece of
iron is regarded as due to small “Amperian currents” (which actually may
be spinning electrons) in the iron. This viewpoint eliminates the need for
such a concept as the magnetic pole, and as a result it makes Coulomb’s
law for the force between poles unnecessary. We are left with the following
basic principles: (1) the law of force between static charges, (2) the law of
force between moving charges (or currents), (3) electromagnetic induction,
and (4) the displacement current postulate of Maxwell.
Another modification of the nineteenth century viewpoint is due to the
theory of relativity. The basic principle of this theory is that the laws of
physics which one would observe to hold true on, say, a rocket moving
with a high but constant velocity, would be the same as the laws that have
been discovered on or relative to the earth. This means that the laws of
Coulomb, Ampere, Faraday, and Maxwell must be such that from them
alone we cannot detect such a thing as absolute motion; in the above case
of the earth and rocket, one is no more “at rest” than the other. If we
change from the usual laboratory reference system (x, y, z, t) to a coor¬
dinate-time reference system (x'; y', z', t') moving with the rocket, Max¬
well’s equations should have the same form as before, i.e., they should be
invariant, according to the relativity theory. We shall find in Chapter 14
126 FUNDAMENTAL LAWS OF ELECTROMAGNETISM [CHAP. 7

that this is so. The relativity principle plays an important part in the
theory of rapidly moving charged particles. The expressions derived in
relativistic electrodynamics are more general (and more complicated) than
those encountered in the nonrelativistic, or “classical ” theory.
In the past thirty years quantum mechanics has also made its impact on
electromagnetic theory. This subject is treated briefly in the last chapter.
We shall now proceed to develop the theory of electromagnetism as
based on the laws of Coulomb, Ampere, Faraday, and Maxwell.

7-4 Fundamental concepts. We wish to state our laws in mathematical

form, that is, as equations relating mathematical symbols. These symbols
stand for physical concepts whose definitions relate them to an experimental
measurement, or to other concepts previously defined. We must commence
electromagnetic theory by introducing and defining the new concepts that
appear in the laws of electromagnetism; then these laws may be intelli¬
gently presented.
In mechanics it is customary to take three concepts as fundamental
(usually length, mass, and time are chosen) and to express all others in
terms of these three. The three corresponding fundamental units are
arbitrarily defined, and from these all other mechanical units are derived.
In the mks system, adopted for this book, the meter, kilogram, and second
are defined, while the unit of force, the newton, is derived from F = ma,
so that 1 newton = 1 kg-m/sec2.
Corresponding to each fundamental concept we may introduce a dimen¬
sion. In both the mks system and the cgs system, the fundamental di¬
mensions are length (L), mass (M), and time (T). We then say that the
dimensions of force are LMT~2. The dimensions of all other quantities
may be found from their definitions. The dimensions of a quantity serve
as a tag that helps us to identify the quantity whenever it appears in the
theory. In terms of L, M, and T we find that such different mechanical
quantities as force and work have different dimensions and units. This
helps us to keep them straight. One exception is the case of torque and
work, both of which have the dimensions of L2MT~2, but this single
ambiguity does not usually bother us. When we come to the law of
gravitation,
p = n

where G is here the gravitational constant and F is the force of attraction

between two mass particles a distance r apart, we find that G must have
dimensions, since, in a physical equation, the dimensions of each side must
be the same.^ Since the dimensions of force are LMT~2, those of G must
be L M T (Here the assumption is made that inertial and gravita¬
tional mass are identical.)
7-4] FUNDAMENTAL CONCEPTS 127

Although physicists are becoming increasingly aware of the fact that we

are not obliged to take exactly three concepts as fundamental in mechanics,
it is generally agreed that this number is probably best from the point of
view of clarity and convenience. Suppose, for example, that we decided to
introduce only two fundamental dimensions in mechanics, length and time.
From Newton’s second law, F = ma, and the law of gravitation, F =
Gmim2/r2, we have two expressions that relate force to mass. If Ave again
assume inertial mass to be the same as gravitational mass, the dimensions
of ma must be the same as those of Gm2/r2. Since the dimensions and unit
of mass have not yet been defined in this length-time system, we are free
to postulate that G is a dimensionless quantity equal to unity. The
dimensions of mass become those of ar2 (acceleration times distance
squared), or L3/T2. There is nothing wrong with this; it does not affect
the physical content of any equation. On the other hand, this two-dimen¬
sional system has practical disadvantages. For one thing, it does not permit
us to choose arbitrarily a unit of mass of convenient size. Also, this system
leads to such dimensions as L4/T4 for force, L5/T2 for moment of inertia,
etc.; these are more awkward than the conventional dimensions and would
probably prove more difficult for students to understand.
When we turn to electromagnetism we are confronted with the necessity
of choosing the system of dimensions that will also lead to the greatest
clarity and so be the easiest to comprehend. At the present time the most
popular system among physics teachers is the four-dimensional one, in
which the fundamental quantities and dimensions are taken to be length
(L), mass (M), time (T), and charge (Q). This is the system that will be
used in this book, but before we develop it further let us first consider two
other systems that have been used a good deal in the past and are still
encountered in the literature. These are three-dimensional systems, each
based on cgs mechanical units.
In the electrostatic and the electromagnetic systems the unit of charge is
a derived unit and the dimensions of all electromagnetic quantities may
be expressed in the form LXMVTZ. Unfortunately, the exponents x, y, and
z are not always integers. There are also several ambiguities; for example,
capacitance in the electrostatic system, inductance in the electromagnetic
system, and length are all measured in centimeters. Finally, since in these
systems a fundamental constant similar to G is taken to be unity (see
problems 6 and 7 on page 136), the size of the unit of any electromagnetic
quantity is fixed and related to the centimeter, gram, and second. We are
not, say, free to choose for current a unit of convenient size. For this reason
a separate “practical system” was introduced, with such familiar units as
the ampere, the volt, and the ohm. The relationship between each such
practical unit and its counterpart in the electrostatic and electromagnetic
systems had to be introduced. For all these reasons, these systems proved
128 FUNDAMENTAL LAWS OF ELECTROMAGNETISM [CHAP. 7

confusing and difficult to learn. The four-dimensional system that we shall

use does not possess these disadvantages. The two cgs three-dimensional
systems are related to our four-dimensional one in much the same way as
the length-time system (with G = 1) is related to the length-mass-time
system (with G not dimensionless and not unity) in mechanics.
Let us now consider the four-dimensional system to be used in this book.
If we start with length, mass, time, and one electrical quantity, then
all other electromagnetic quantities can be expressed in terms of these
four. We may take any one electromagnetic concept to be the basic one.
Perhaps it would seem best to choose, say, electrical resistance, since a
standard unit of resistance (the ohm) can be stored away and easily main¬
tained along with a standard meter bar and a standard kilogram; this would
lead to the mkso (meter-kilogram-second-ohm) system. However, we
shall see that our fundamental laws are concerned with electric currents
and charges, and so we shall choose electric current or charge as our fourth
fundamental concept. Which of these two we take is immaterial, since the
statement current is time rate of flow of charge defines one concept in terms
of the other. If we let the symbol for current be I and that for charge q,
we have

(7-1)
dt

As indicated earlier, we are going to consider electric charge to be the fourth

fundamental concept (length, mass, and time are the other three), but we
shall take the unit of current as the new basic unit. This unit of current
is called the ampere and it is defined with the aid of Ampere’s law (see
Section 7-6). The corresponding unit of charge, the coulomb, then follows
from Eq. (7-1), as follows: one colurnb is that quantity of charge passing
any point in a circuit in unit time when the current in the circuit is steady and
is equal to one ampere. We may turn this statement around and say that
an ampere is one coulomb per second. This system is called the mksc
(meter-kilogram-second-coulomb) system.
Since the discovery of the electron we have come to regard a charged
body as one which possesses an excess or a deficit of electrons* and an
electric current is thought of as a stream of moving charges, usually elec¬
trons. We could take the charge on the electron as the basic unit, but
although this would have advantages in atomic physics, it would lead to a
unit of current that would be far too small for common use in electrical
engineering and daily life. It seems more practical to use the freedom of
choice available in a four-dimensional system to pick a unit of current
of convenient size.

It is conventional to take the charge on

on the
the elect
electron to be negative, so that a
deficit of electrons corresponds to a positive charge.
7-5] FORCES AND FIELDS 129

The concept of a point charge is (like that of a mass particle) a hypo¬

thetical one, approximated experimentally by a small charged sphere
whose radius is much less than its distance from other objects.
Another convenient concept is that of a current element. Picture a small
segment of a conductor carrying a steady current 7. The length of such a
segment is taken in the direction of the positive current and designated
as ds. (Boldface type will be used to denote vector quantities.) The
product I ds is called a current element.

7-5 Forces and fields. In mechanics, Newton’s laws are the general
ones, whereas Hooke’s law describes a property of a certain class of media.
Newton’s laws of motion are used to determine the accelerations and the
resulting motion of a system once the forces are known. However, physics
must also list and describe the different kinds of forces encountered in
our world. In mechanics we consider gravitational forces (including weight),
which are of a general nature, and forces such as friction, elasticity, and
viscosity, which are associated with specific objects or media. In addition
to such mechanical forces there are forces of a general nature that act only
on charged bodies (at rest or in motion), forces that are termed electro¬
magnetic. Finally, within the nuclei of atoms a third universal type of
force, associated with mesons, is believed to account for the strong forces
holding nuclear particles together.
At present we are concerned with electromagnetic forces. We must de¬
scribe when they occur, how they act, and in what manner they depend
on other physical quantities. For example, we must state what force one
charged body exerts on another and how this force depends on the posi¬
tions, velocities, and charges of the two bodies (as well as on the intervening
medium, if it is not equivalent to a vacuum). Like the law of gravitation
and Hooke’s law, our electromagnetic laws (1) postulate the existence of
new types of forces, and (2) lead to formulas relating the forces to other
quantities.
Forces of friction, elasticity, and viscosity involve contacts between
bodies. Electromagnetic forces, however, resemble gravitational forces
in that they seem to involve “action at a distance. ” Two charged bodies
A and B may interact even when separated by a vacuum, just as the
earth and sun interact gravitationally despite the miles of empty space
between them. We shall find it preferable to replace the action-at-a-
distance concept with the field point of view. We say that surrounding a
charged body A is a field, and that if the other charged body B is placed
in this field, then A’s field acts (exerts a force) on B; similarly, B’s field
acts on A, so that A and B interact mutually, in agreement with Newton’s
third law. This concept of fields, with lines of force to represent them,
may at first seem to be merely a dodge to make action at a distance seem
130 FUNDAMENTAL LAWS OF ELECTROMAGNETISM [CHAP. 7

more plausible. However, we shall see that an electromagnetic field can

exist after the charges producing it have been neutralized or have ceased
to radiate, as in the case of radiowaves from a distant transmitter. Energy
is propagated by such fields and so they must have physical reality. We
shall see that an electromagnetic disturbance travels with the finite speed
of light, so that we must conceive of fields as both surrounding and em¬
anating from charged bodies.
We may now say that our electromagnetic laws (1) postulate the exist¬
ence of new types of fields and (2) describe how such fields depend on the
positions, velocities, and charges of the bodies present.
Experimental results suggest that there are two kinds of electromagnetic
fields. When a charged body at rest experiences a force which a neutral
body at the same position would not experience, we say that the body is
in an electric field. When an additional force acts on a charged body when
the body is in motion,* we say that a magnetic field is present. It is conven¬
ient to introduce two vector quantities to represent these two kinds of fields.
(a) When a small static charge q' experiences a force F when placed at a
certain point in an electric field, the ratio

E (7-2)

defines the electric intensity at the given point. The direction of E is that
of the force on a positive charge located in the field. From this definition
the mksc unit for E is seen to be the newton per coulomb (abbreviated to
n/coul). Since F is found to be proportional to q', this definition makes the
value of E independent of q'.
(b) When a small charge q' moves in a magnetic field with a velocity v',
the force F acting on it is found to be perpendicular to v/ and to depend
on (1) the charge q , (2) the speed v', and (3) the direction of motion.
The force has a maximum value Fmax when the velocity v' lies in a certain
plane, while the force is zero when v' is perpendicular to this plane. The
value of Fmax is found to be proportional to q' and to v'. (We shall see that
these statements are all contained in Ampere’s law.) The vector B rep¬
resenting the magnetic field is defined as follows. The magnitude of B is
taken to be
o F
1 max

B = w (7-3)

and the direction of B is chosen normal to that plane in which v' must lie
m order for the force to be a maximum. This means that when v' is

Any relativistic change in the electric force is here ignored.

7-5] FORCES AND FIELDS 131

parallel to B the force will Ipe zero. While the positive and negative
directions of B will be distinguished later, the convention usually followed
is indicated in Fig. 7-1.* B is called the magnetic induction, or the magnetic
flux density. From Eq. (7-3) we see that the mksc unit for B is the newton-
second per coulomb-meter (n-sec/coul-m). However, it is customary to
define a new mksc unit, the weber, as

„ , „ newton-meter-second
1 weber — 1 --,
coulomb

so that the unit for B may also be

called the weber per square meter
(w/m2). We associate webers with
magnetic lines of force, or flux, and
webers per square meter with flux Fig. 7-1. Direction of B.
concentration, or flux density.
Since Fmax is proportional to q'v', the definition given leads to a value
for B that is independent of the charge on, or speed of, the test body used
to measure B.

Example. An electron (q' = —1.6X 10-19 coul) at rest experiences a force of

24.0 X 10-17 n to the east; when moving with a speed of 107 m/sec an additional
force is found to act, this force being zero when the charge moves vertically. The
magnetic force has a maximum value of 8.0 X 10-17n northward when the
charge is moving eastward. Find E and B.

Solution. From Eq. (7-2), we have

—17
_ 24.0 X 10— _ 1500 n/coul to the west.
1.6 X 10-19

From Eq. (7-3) we see that

8.0 X 10
B = 5 X 10 5 w/m2.
1.6 X 10-19 X 107

The direction of B must be vertical and would conventionally be taken as up¬

ward. (If the charge were positive the forces would be reversed in direction.)

*This convention is to take B to be in the direction of advance of a right-

.handed screw rotating from the direction of Fmax (when the charge q is positive)
to the direction of v'.
 132 FUNDAMENTAL LAWS OF ELECTROMAGNETISM [CHAP. 7

7-6 The fundamental laws of electromagnetism for free space. Before

we take up one law at a time, expressing it in different forms and applying
it to problems, we shall state all four laws together and in their most
logical order. The reasons are as follows:
(1) The interrelationship between electricity and magnetism and the
general framework of electromagnetic theory may be better understood in
this way.
(2) The ampere (in terms of which we define the coulomb) was chosen as
our fundamental electrical unit, yet it cannot be defined until we come to
Ampere’s law. It is, however, most convenient to commence the study of
electromagnetism with its simplest branch, electrostatics, and then to dis¬
cuss current effects (Ampere’s law) at a considerably later date. Since we
do not want to have to assume a unit of charge, with the promise that it
will be defined many pages later, Ampere’s law and the definitions of the
ampere and coulomb are given here at the start.

(a) Ampere’s law. As in Fig. 7-2, let I ds and I' ds' be two current ele¬
ments a distance r apart, and let r j be a unit vector directed from the
first current element toward the second. Assume the currents to be steady.
Ampere’s law states that the two current elements will interact, the force
d~F on I' ds' due to / ds being given by

jj■/ ds x (ds x
1 ^2 (7-4)

where Cx is a constant of proportionality. The force is represented by a

second-order differential, since the right side of the equation contains the
product of two differentials (ds and ds').
We have decided to measure distances in meters and force in newtons
If a value is assigned to C,, then I and 1> must be expressed in terms of a
unit of such size that d F will come out to be the number of newtons ob¬
served experimentally. We shall define the ampm as the unit of current
corresponding to
Ci — 10 7 newton/ampere2. (7-5)
This definition of the ampere is actually equivalent to a former one
1 amp 0.1 abamp) based on the concept of magnetic poles, for the
7-6] LAWS OF ELECTROMAGNETISM FOR FREE SPACE 133

factor of 10-7 is just the following fraction: (F in newtons/F in dynes)

divided by (7 in amps/7 in abamps)2. The international ampere, related
to the amount of. silver deposited by a coulomb in electrolysis, is defined
so as to make it agree, within experimental error, with the ampere as
defined above.
To simplify matters, take the special case where the current elements
are parallel to one another and side by side, so that rx is perpendicular to
both ds and ds'. Then the vector product (ds x rx) is a vector perpendic¬
ular to ds (and also to ds') having the magnitude of ds, (see Fig. 7-2).
Hence for this case

d2F = 1 o~7ir > (7-6)

and the force is attractive if the currents are in the same direction. We can
say that the ampere is that steady current which will produce an inter¬
action of 10~7 newton when present in each of two parallel current ele¬
ments of such dimensions and separation that dsds'/r2 is unity. We see
in this way how Eq. (7-6), and hence Eqs. (7-4) and (7-5) together,
define the ampere.
One coulomb is that quantity of charge which in one second passes any
section of a circuit carrying a steady current of 1 ampere.
(b) Coulomb’s law for charges. As in Fig. 7-3, let q and q' be two point
charges a distance r apart, and let rj be a unit vector directed from q
toward q'. Assume the charges to
be at rest. Coulomb’s law states
that the two charges will interact,
the force F on q' due to the presence
of q being given by

Fig. 7-3. Coulomb’s law.

F -- (7-7)

where C2 is a constant of proportionality.

We have chosen to express q and q' in coulombs, or ampere-seconds,
r in meters, and F in newtons. For the two sides of the equation to be
equal, C2 must have a particular value. We cannot assign this value, but
must determine it experimentally. In this manner it has been found that
to a close approximation

C2 = 9 X 109 newton-meter2/coulomb2. (7-8)

This is a positive constant; in other words, we find that like charges repel.
For two charges, each of one coulomb, one meter apart, the force would
be 9 X 109 newtons, a tremendous force. However, while currents of 1
 134 FUNDAMENTAL LAWS OF ELECTROMAGNETISM [CHAP. 7

to 10 coul/sec are not excessive, a static charge on a small pith ball will
measure about 10-8 coulomb. For atomic particles q is of the order of
10-19 coulomb, although r may then be as small as 5 X 10-11 meter
(as in the hydrogen atom). From

F = 9 X 109 & (7-9)

we see that F is never really large.

(c) Faraday’s law. This law states qualitatively that a changing magnetic
field induces an electric field.
Quantitatively, this law may be expressed as follows. Consider any

dS

surface S (Fig. 7 4) and let /\E • dX be the line integral of the electric
intensity around the periphery A of this surface. Take an element dS
of the surface, in the normal direction. The positive normal is determined
by the way a right-handed screw would advance when turning in the same
sense as that in which the line integral is taken. Now form the surface
integral of the time derivative of the magnetic induction, namely,
JXs (dB/dt) • dS. Then Faraday’s law states that

/*•*=-//“•■»• p-io)

The left side of this equation is a quantity defined as the electromotive

force (emf) around the chosen path, while the right side will be called the
time rate of change of the magnetic flux through the surface S.
Equation (7-10) does not include the case of an emf induced in a moving
conductor, with which we are not concerned in electromagnetic field theory,
owever, by the use of Ampere’s law and the conservation of energy

show1faterFaraday S ^ be extended to cover this case, as we shall

It may be noted that all quantities in Eq. (7-10) have been previously
e ned (see Eqs. 7 2 and 7-3) and yet no new constant appears. The
7-7] SUMMARY 135

equation states a fundamental interrelationship between electric and mag¬

netic fields.
(d) Maxwell’s displacement current postulate. This principle may be
stated in several equivalent ways. The term “displacement current” refers
to Maxwell’s original explanation, but it is better to express this fourth
law in a form comparable with Eq. (7-10).
Let us take a surface S, as before. Maxwell’s postulate may be expressed
as follows:

rJB-dx
x
= ff*-da’
s
(7'u)
when there is no net current through the surface S. This equation states
that a changing electric field induces a magnetic field. Maxwell’s equation
has been confirmed directly and, as was said earlier, it has received a
great deal of indirect confirmation because it led to the prediction of
electromagnetic waves.
Any experiment based on Maxwell’s postulate may be used to find the
ratio CyC'i- Since we arbitrarily chose C\ to be 10 7 n/amp2, such
measurements serve to evaluate C2 in mksc units. In this way we obtain
the value of 9 X 109 n-m2/coul2 for C2. If we assigned a different value
to Ci, the value of C2 would also change, but the ratio C2/C\ would be the
same. This ratio is the important quantity; its value is close to

— = 9 X 1016m2/sec2.
C1

Let us call the square root of this ratio c. Then

c = VC2/Ci = 3 X 108 m/sec, (7-12)

which is the speed of light in empty space. This is one of our universal
constants. It makes its appearance in electromagnetic theory regardless of
what system of units is chosen. We are encountering here a constant of
Nature that also appears in relativistic mechanics, but not in Newtonian
mechanics.

7-7 Summary. The following points should be noted:

(1) Electromagnetic theory is based on four fundamental laws, namely,
those of Ampere, Coulomb, Faraday, and Maxwell.
(2) It is convenient to choose as fundamental concepts length, mass,
time, and one electrical quantity which, in the mksc system, is taken to be
electric charge.
 136 FUNDAMENTAL LAWS OF ELECTROMAGNETISM [CHAP. 7

(3) The force on 8, charge in an electromagnetic field may be due to the

presence of an electric intensity E and (if the charge is moving) a magnetic
induction B.
(4) The unit of current (or charge) is defined from Ampere’s law by
assigning a value to the constant of proportionality (Ci).
(5) The constant (C2) in Coulomb’s law is then fixed by experimental
observation; there is a relationship between C\, C2, and c, the speed of
light, so that measurement of the last quantity determines C2.

Problems

1. From the definitions in Sections and putting C1 = 1 dyne/(abamp)2,

7-5 and 7-6 derive the mksc dimen¬ where the abamp is the unit of current.
sions of (a) E, (b) B, (c) Ci, (d) C2, (a) How many amperes equal 1 abamp?
(e) c. (b) What value must C2 have in this
2. With current-carrying coils a system? (c) Find the magnitude and
magnetic field whose B = 10~3 w/m2 units of (C2/Ci)1/2 in the electromag¬
may be readily established. If a netic system and compare with Eq.
charged particle is moving at right (7-12). (d) The em unit for B, which
angles to such a field with a speed of is called the gauss, may be encountered
106 m/sec, what electric field would in the literature. How many w/m2
exert an equal but opposite force? equal 1 gauss? (Hint: 1 n = 105
Give relative directions of the two dynes.)
fields.
8. The electrostatic system of units (es
3. Find the ratio of the electric force units: also called statunits) is defined
to the gravitational force between two by using cgs mechanical units and
electrons, each with a mass of 9 X putting C2 = 1 dyne-cm2/statcoul2,
10~31 kg and a charge of 1.6 X 10“19 where the statcoulomb is the unit of
coul.
charge, (a) How many statcoulombs
4. If we put C2 = l/(47re0) and e0/uo equal 1 coul? (b) What value must Ci
= 1/c2, find the values of e0 and mo- have in this system? (c) Compute
5. Explain why the factor ds' X (cyci)1/2.
(ds X rfr/r2 in Eq. (7-4) is dimension¬
9. Calculate in joules/coul (volts)
less.
the emf per turn induced in a coil
6. Suppose that a new unit of through which B is uniform, parallel to
charge, the franklin, is chosen, such
the axis of the coil, and increasing at
that 1 franklin = 10 coul, 1
the rate of 10~7 w/m2-sec, given that
franklin/sec = 10 amp. Using’mks
the radius of the coil is (a) 10 cm
mechanical units (or an mksf system), (b) 1 m.
find the values that C2 and Ci would
10. Compute $B • dX around a cir¬
have in this system. Evaluate C2/Ci.
cular path of 10 cm radius when the
7. The electromagnetic system of units
electric intensity E is uniform, parallel
(em units; also termed ab units) is de¬
to the axis of the path, and increasing
fined by using cgs mechanical units
at the rate of 104 n/coul-sec.
 Chapter 8

COULOMB’S LAW. ELECTROSTATICS

8-1 Coulomb’s law of force between charged bodies. This law may be
stated in words as follows:
Bodies with like charges repel and with unlike charges attract one another;
for point charges the force of interaction is proportional to the product of the
charges and inversely proportional to the square of their distance apart.
Expressed mathematically [see Section 7-6(b)], Coulomb’s law states
that the force F on a point charge q' due to the presence of another point
charge q is

(8-1)

where C2 is the constant of proportionality, r the distance of separation,

and ti a unit vector directed from q toward q'. Unit vectors are introduced
to give the proper direction to a vector quantity; since the magnitude of a
unit vector is unity in any system of units, it is dimensionless (see problem
5 of Chapter 7). Thus the magnitude of F is

E = (8-2)

We have seen that when q and qf are expressed in coulombs (as defined
by Ampere’s law), and when r is in meters and F is measured in newtons,
then for free space (a vacuum) C2 has a measured value close to
9 X 109 n-m2 /coul2. If we had chosen a larger unit in terms of which
to express charge, then, for the same two charges, q and qf would be smaller
numbers; to preserve the equality of C2qq' /r2 to F, C2 would have to have
a larger value (see problem 6 of Chapter 7). Similarly, a smaller unit of
charge would result in a smaller value for C2. We prefer to take the coulomb
as our unit of charge because the units derived from it (ampere, volt, ohm,
farad, etc.) are those used in common practice.
It is found convenient to replace C2 by another constant «o, defined by

1
C2 (8-3)
4tT €q

The reason for introducing the factor 4ir here is to avoid its appearance
later on, particularly when we come to Maxwell’s electromagnetic equa¬
tions. From the numerical value of C2 given above we obtain for e0 a
measured value, in the mksc system, close to
137
138 coulomb’s law. electkostatics [chap. 8

1 1 coul2
e° ~ 4^C ~2 ~ 36x X 109 n-m2 '

e0 is called the 'permittivity of free space. Henceforth we shall write l/4xe0

for C2. Coulomb’s law then becomes

1 (8-4)
47reo r2

8-2 Electric intensity. As stated in Section 7-5, the electric intensity E

at any point is defined as the force per unit charge that would act on a small
test charge located at that point, that is,

(8-5)

where q' is the test charge and F the force on it due to the presence of
other charges.
At a distance r from a charge q the force on another charge q' is given
by Eq. (8-4), so that the electric intensity at q' due to q is

1 q_
E = (8-6)
4x6(3 r2 r

where rx is a unit vector directed away from q. Equation (8-6) gives the
field due to a point charge only; it is not a definition of E or a general
expression, such as is Eq. (8-5).
If the electric field is due to the presence of more than one charge, each
exerts its own force on the test charge. Since forces add vectorially, so also
do the fields of the separate charges, and we have for the resultant field

E = 1 V Si r (8-7)
4x6(3 4^ r2
u l l
ll’

where the subscript i refers to any one of the charges. If the charges are
spread over surfaces, or occur as a continuous (macroscopically speaking)
volume distribution, then the summation in Eq. (8-7) is replaced by an
integration. The integration of a vector function must be performed on
one component at a time; this gives us the components of the resultant
vector. Actually, however, we seldom use this method in practice, for we
shall find others that are better. It is assumed that the reader is familiar
with an elementary text in which examples of the application of Eq. (8-7)
are given.*

For example, Sears and Zemansky, University Physics, Chapter 25.

8-3] gauss’ law 139

8-3 Gauss’ law. The electric field of a point charge and the gravitational
field of a mass particle are both central, inverse square fields. Because of
this close analogy, many of the expressions and theorems applying to
electric fields may be found from the corresponding ones derived for
gravitational fields. Charge corresponds to mass, electric intensity to
gravitational field, and l/47re0 to minus the gravitational constant.* The
reader who is familiar with the gravitational expressions may use this
analogy to arrive at the equations which will now be derived directly from
Coulomb’s law.
Consider any surface S which surrounds a volume r in which is located
a point charge q (Fig. 8-1). At any element of area dS let 6 be the angle
between the direction of dS and that of the electric intensity E due to q.
From Fn. fS-fil. we have s

= 47rg,
Fig. 8-1. Gauss’ law.

where dtt is the solid angle subtended by dS at q and 4tt is the complete
solid angle about a point. If more than one point charge is present within
S, we will find that
(8-8)
s

where Em represents the total net charge contained within S. If the

volume r contains a continuous charge distribution, we let p denote the
charge density, or net charge per unit volume. Equation (8-8) then becomes

(8-9)
s T

While Eqs. (8-8) and (8-9) state what is generally called Gauss’ law for
electrostatics, this “law” should be considered a theorem derived from
Coulomb’s law, or an alternate form of Coulomb’s law.

*See Yol. I, Sections 5-10 and 5-11, where Gauss’ law is stated in Eq. (5-28).
 140 COULOMB S LAW. ELECTROSTATICS [CHAP. 8

With the aid of Gauss’ law we can quickly derive expressions for the
electric field in the neighborhood of charged spheres, cylinders, and
infinite planes, as we shall briefly show.
Illustration 1. Charged sphere. Suppose that a charge q is uniformly
distributed over the surface or volume of a spherical body. We take for
the surface S in Eq. (8-9) a sphere of radius r outside and concentric with
the charged sphere. From symmetry considerations, the vector E must be
constant in magnitude over S and directed radially outward if q is positive,
so that E • dS = EdS. We then have

€qE • dS = eoE / / dS — eoF(47rr )

Since the right side of Eq. (8-9) is just q, we get

E = (8-10)
4ir«o r2

Thus, outside the sphere, the field is the same as it would be if the spherical
distribution were replaced by a point charge q at the center. (Compare
the analogous gravitational theorem.)
Illustration 2. Charged cylinder of infinite length. Let the charge per unit
length be constant and call it X. Take the surface S to be cylindrical, con¬
centric with the charged cylinder, and of unit length. Then from Gauss’
law the field outside the charged cylinder at a distance r from the axis is
given by

/ / e0E • dS = e0F(2-7rr) = X,

1
E = (8-11)
2-7reo r

Illustration 3. Field between oppositely charged parallel plates. Two

parallel conducting plates are frequently given equal and opposite charges,
eit er y connecting a battery across them, or by charging one plate and
grounding the other. We shall assume that the plates are very large com¬
pared with their distance of separation, so that end effects may be neglected.
In practice, this may be ensured by placing guard rings outside the edges
of the plates. Let the charge per unit area on one plate be <r and on the
other -a From symmetry considerations, we may assume that * is the
same at all points on a plate, so that the field between the plates is uniform,
n ig. 8 2 lets be a closed surface shaped like a pillbox and enclosing
an area A of the positive plate. A field inside a conductor will cause the
free charges in the conductor to move. Here we assume that we are
8-4] DIVERGENCE AND CURL OF E 141

considering a static problem, so that there is no field inside the plates.

For the surface of our pillbox within the conducting plate JJe0E • dS
will be zero. The cylindrical side of the pillbox may be made as small as
we wish. Then the only contribution to the surface integral comes from
the area A of the pillbox which is in the region between the plates. Call
the field in this region E. We then have

£oE • dS = toEA.
s

The charge enclosed within the

pillbox is a A. From Gauss’ law
we then have

eo EA = <jA,

E = —• (8-12)
eo Fig. 8-2. Electric field between
charged plates.
The field is directed as shown in
Fig. 8-2.
Gauss’ law is also employed in elementary texts to show that (1) the
charge on a solid conductor resides entirely on the surface, and (2) if the
conductor is hollow and empty, there can be no net charge on the inner
surface.

8-4 Divergence and curl of E. The divergence theorem in vector

analysis states that for any vector function A

where r is any volume enclosed by the surface S. If we apply this theorem

to the left side of Eq. (8-9), this equation becomes

(e0 is a constant and can be put before or after V). Since the limits of
integration are the same for each side of this equation, we may equate the
integrands and obtain
V • E = p/eo- (8-13)
 142 coulomb’s law. electrostatics [chap. 8

The divergence of E at any point is thus proportional to the charge density

in the immediate neighborhood of that point. The field emanates from
any positive charge and converges toward a negative one. In a region
where p is zero any field present will merely cross a closed surface, entering
on one side and leaving from another. This, of course, is just what we ex¬
pect from Gauss’ law, from which Eq. (8-13) was derived.
In mechanics it is shown* that if the force F acting on a particle is
central (i.e., toward or away from a certain point) and is a function of r
only (r being the distance from the center of attraction or repulsion),
then V x F = 0 and the force field is said to be irrotational and conserva¬
tive. Since the electric field E* due to a point charge is also central and
a function of r only, it follows that V x E; = 0. (See problem 6 at the
end of this chapter.) Then for a field due to any number of charges, we have

V x E = V x £ E, = £ v x Ei = 0. (8-14)
i i

Therefore, in general, in electrostatics the curl of the field is zero, and the
field is again referred to as irrotational and conservative. Given the value
of p for all points in free space, Eqs. (8-13) and (8-14) define E and so are
equivalent to Coulomb’s law.

8 5 Electric potential. In common with all conservative fields, the

electric intensity is derivable from a scalar potential function. We shall
now introduce this function, which is called the electric potential and
is designated by the symbol V.
When a small test body with a positive charge q' is placed in an electric
field, a force F = qfE will act on the body. If the charge moves with the
field, work will be done by this force, just as work is done by the force of
gravity when a mass particle falls. In such cases we say that the body
has moved to a position of lower potential energy. The work AW done by
the field when a test charge qf is given a displacement ds in an electric
field E is
AW = F • ds = q'E • ds.

Since ^7 x F - V x E — 0, F is a conservative force, which means

that AW must be the same for all paths leading from some position a to a
neighboring ^sriion 6 Therefore we may express this work as an exact
fW’ We f1ollow the same terminology as in mechanics and call
dW the change in the potential energy of the test body. We now define

™tin mX/
ratio —dW/q ' , that
iny+WhlCh
is, the deCtriC v°tential at b exceeds that at a as the

*See Vol. I, p. 119.

8-5] ELECTRIC POTENTIAL 143

This is called the potential difference (p.d.) between a and b.

When the positions a and b are not adjacent, but a finite distance apart,
we obtain

(8-15)

for the potential difference. Since dV is independent of path, Vb Va

is also. V is a state function, just as entropy was found to be in thermo¬
dynamics.
Difference in potential has been defined, but not the absolute value of V.
We may arbitrarily take some convenient point (say at infinity), or a
certain conducting body (such as the earth), to be at zero potential. If
Va = 0, then

Let us compute the work per unit charge performed by an agent when a
charge is taken around a closed path in the field, say from the point a
back to a. From Eq. (8-15), we have

dV = -
f E • ds,

where £ refers to integration around a closed path. Now since dV is an

exact differential,

Since E • ds need not be zero, we must have

/ E • ds = 0,

or the line integral of the electric intensity around any closed path is zero
in*electrostatics. This is just equivalent to the previous statement that
an electrostatic field is irrotational.
 144 coulomb’s law. electrostatics [chap. 8

Let the components of E be Ex, Ey, and Ez. Take the case where ds is
in the X-direction, so that E • ds = Ex dz. For this displacement dV =
—Ex dx. Since this represents a change in V when y and 2 are held con¬
stant, we have

u
dx = u.
In a similar manner, we find that

«Z__p p
dy ~ " dz - E‘-

These last three equations may be combined in vector form as

E = — VF, (8-16)

where V is the vector differential operator

. d d
V = 1
dx Jaj + k5’
and i, j, k are unit vectors in the three principal directions.
Since electric potential has the dimensions of work divided by charge
the mksc unit is the joule/coulomb. This is also called the volt, which is
thus defined as
1 volt = 1 joule/coulomb.

Note that our unit of electric intensity may be expressed as

1 newton/coulomb = 1 joule/coulomb-meter = 1 volt/meter.

The volt/meter relates E to experimentally measurable quantities: we

actually determine an electric field not by getting its force on a test charge
but with the aid of Eq. (8-16). This is an example of how we use one
formula to define a quantity and another, derived from the first to
measure the quantity (see also the case of C2, e0, and c).
A surface over which V is constant is called an equipotential surface;
y giving the constant a succession of consecutive values, a set of such
surfaces is obtained. Then at any point in an electrostatic field, E is di¬
rected perpendicular to the equipotential surface through that point. Lines
representing the direction of E are called lines of force; lines of force and
equipotentia1 surfaces are mutually perpendicular, or orthogonal. This
leads to the use of conjugate functions in solving certain electrostatic
problems, as we shall see later. ttc
8-5] ELECTRIC POTENTIAL 145

In electrostatics all points in a conductor must be at a common po¬

tential; if not, a field — VF \Vould exist in the conductor and cause the
free charges to move. Since the surface of a conductor is an equipotential
surface, the electric intensity just outside is normal to the surface. Let
dn be an element of distance and tij a unit vector in the direction of the
outward normal; Eq. (8-16) then becomes

dV nx
E = — — (8-17)
dn
just outside a conductor.
We may use Gauss’ law to find a relation between E at a point just out¬
side a charged conductor and a, the charge per unit area on the conductor
in the neighborhood of the point. Let us refer back to Illustration 3 of
Section 8-3, where we took for the surface S in Gauss’ law a pillbox en¬
closing an area A of the conductor’s surface. We may follow the same pro¬
cedure here. In both cases there is no field inside the conductor, and just
outside the conductor the field is normal to the surface, therefore the re¬
sulting expression will be the same, namely,

a
E = — ni.
eo

However, in the present case a need not be uniform.

If we combine this relation with Eq. (8-17), we obtain

dV
a = — e o —— (8-18)
dn

This equation may be used to find the charge density at any point on the
surface of a conductor, once the potential function V is known.

Example 1. The electron volt has been defined (see Section 4-5) as the work
done in taking a particle bearing the electronic charge (1.6 X 10 coul) through
a p.d. of 1 volt. Thus

1 electron volt = 1.6 X 10 “19 coulomb volt

= 1.6 X 10~19 joule.

Example 2. Perpendicular to the Z-axis are three thin metal plates, A at

x = o B at * = 1 m, and C at i = 3 m. A is kept at ground potential, B at
a potential of 200 volts, and C at a potential of 100 volts, above that of A
Gee Fie 8-3) There is no charge between the plates, (a) Find the electnc field
between A and B and between B and C. (b) Find the charge density on each

plate.
146 coulomb’s law. electrostatics [chap. 8

7 = 0 V = 200 V = 100

Fig. 8-3. Relation between field and potential.

Solution. The fields are assumed to be uniform. Between A and B

AV 200
E = —200 volts/m.
Ax 1

This is to the left. Between B and C

F> (-100)
= -f- 50 volts/m
to the right.
From a — eoEn, we see that

2
cra = —200 eo coul/m ,
2
<rc = +50 eo coul/m .

B has a charge density of +200eo coul/m2 on its left side and one of —50eo
coul/m2 on its right side, or a net charge density

o\b = +150 eo coul/m2.

Example 3. An electric field is represented by E = axi — bj, where i and j are

unit vectors in the X- and F-directions, respectively, and a = 2 volts/m2,
b = 1 volt/m. (a) Find the electric potential V at any point, given V to be
zero at the origin, (b) Compute the work that would be done by the field if a
charge q = 10 8 coulomb were moved by an agent from the point x = 1 m,
y — 2 m to the point x = 2 m, y = —1 m. (c) Find the charge density p.
Solution, (a) We have Ex = ax, Ey = —6. From Eq. (8-16),

07 ax
—ax. V a possible function of y,
dx 2
8-5] ELECTRIC POTENTIAL 147
t

37
6, V = by + a possible function of x.
dy
Hence
V = —\ax2 + by.

The numerical constant is zero, since V = 0 at x = y = 0.

(b) At x = 1 m, y = 2 m we find Va = —1 + 2 - 1 volt, and at x = 2 m,
y - —1 m, Vb = —4 — 1 = —5 volts. The work done on the charge between
the two points (see Eq. 8-15) is

W = -q'(Vb - Va) = 6 X 1(T8 joule.

This same result can be obtained by evaluating

W = q' (Ex dx + Ev dy)

= 10 / (ax dx — b dy)

along any path connecting the two points. (The reader should check this.)
(c) From Eq. (8-13), we have

p = eoV • E =

= to a.

(Numerical values may be substituted for eo and a.) There is thus a uniform
positive space charge. This space charge alone does not determine the field, for
the charges on, or potentials of, the surfaces bounding the field must also be
given if the field is to be completely specified.
Example 4. An electric field is represented by the potential function V
b cos 6/r2, where b is a constant and r and 6 are polar coordinates. Find the
polar components of the electric intensity.

Solution. Since E = — V V, we have

dV _ 2b cos 6
Er =
dr r3

1 dV b sin 9
Ee =
r dd r3

This is the field due to an electric dipole, i.e., to two equal and opposite charges
very close together, centered at the origin. We shall develop the expression for
the dipole potential in the following section.
148 coulomb’s law. electrostatics [chap. 8

8-6 Potential due to point charges: multipoles. In computing the

potential at a point which is at a distance r from a point charge q, it is
customary to let V = 0 when r = oo. From Eq. (8-6), E is radial and a
function of r only, so that Eq. (8-16) becomes

E = —
dr
Hence
dr
^*2
4x60 r~ + C-

Since we take V = 0 at r = oc, we have (7=0 and

V = -J- (8-19)
47T€o r

This gives the electric potential due to a point charge only and it is not a
definition of or general expression for V.
The value of the potential concept becomes more apparent when we
consider a field due to several charges. Then, since potential is work per
unit charge, which is a scalar quantity, the potential due to a number
of charges is found by algebraic addition, i.e.,

V =
1 gj.}
(8-20)
47re0 4- Vi '

where n is the distance from the charge q:i to the point P(x, y, z) where V
is to be determined. Since V is a scalar quantity, the summation is much
easier to evaluate than in the case of E (Eq. 8—7). Hence it is often simpler
to first compute V and then to find E from E = —VF. This point will
now be illustrated.
Illustration. Potential due to a group of point charges located near the origin.
Suppose that we have a finite number of charges grouped about the origin
and that the distance of any charge q{ from the origin is much smaller
than the distance r from the origin to the point P(x, y, z) at which we
wish to compute V (see Fig. 8-4).
Let the coordinates oFr/,- be we then have for the distance
from qi to P

Ti = ^O ~ W2 + (V ~ Vi)2 + (2 - f;)2'. (8-21)

Using Eq. (8-20), we have for the potential at P

1
V =
47Te0
8-6] POTENTIAL DUE TO POINT CHARGES: MULTIPOLES 149
i

Remembering that &, Vi, f; are small compared with x, y, z, we may

expand this expression for V in a three-dimensional Taylor series about
the origin, as follows:

V = Vi

Ml "1 a2 /iV "l 1V

d2 ( 2

dfi yj 0 + .2 vj. 0 d + .2 W-
dr,*
Vi

1 d2 (iV
+ 2 W -dtjidrj i
(—)
Vi/Jo
%iVi

+ (8-22)
drudt; r) vrti
Vfi^li Vi.
01

where the subscript o means that following the indicated differentiations

all terms are evaluated at the origin (& = Vi = U =0)- For example,
referring to Eq. (8—21), we get for a few of the terms:

In this manner Eq. (8-22) finally yields

150 coulomb’s law. electrostatics [chap. 8

1 1
v=
4we0 Z ^ ^ 1Z + m 2 ?«< + w Z ms*

+ i(Sl2 - 1) Z ?*■*< + i(3m2 - !) Z ?«<

2(3R 1) ^ ] §iTi ~f" 3Z?R ^ ] Qi^iVi “l- 3VVfl ^ ' QiVi^i

3nl ^ ' Qzhi£i

(8-23)

where l = x/r, m = y/r, n = z/r are the direction cosines of the vector r.
The quantities indicated by the summation signs depend on the distribu¬
tion of the charges only and not on the location of P; these quantities are
properties of the charge distribution just as the total mass, moments of
inertia, and products of inertia are properties of a given distribution of mass
particles.
The first summation, Yli qi, is the total net charge. It is called the
monopole moment of the distribution and it is a scalar quantity.
The next three summation terms are called the three components,
mx, my, mz, of the dipole moment of the distribution; this is a vector with
the components

= £ QiU my='£ QiVi, mz = £ qtff. (8-24)

1 i i

If, for example, we have only two

charges, both on the X-axis, q at
+d/2 and — q at —d/2 (see Fig. 8-5),
then mx = qd, my — mz = 0, and

1 1
V = [lqd\
4tt €q r2

b cos 9
(8-25)
Fig. 8-5. Electric dipole.

where 6 is the angle between r and the X-axis and b is independent of the
coordinates of P. This is the potential function taken in Example 4 of the
last section. From it we easily obtained the field of a dipole.
Similarly, the third bracket contains six summations which are called
the components of the quadrupole moment of the distribution; this is a
8-6] POTENTIAL DUE TO POINT CHARGES: MULTIPOLES 151

symmetric dyadic, or tensor the second rank, The components of this

dyadic are defined as follows:

TUt.X - 'ffl'xy —— Wlyx

i i

E2
myy Q iV i. 'lUyz — mzy £ qmrfi, (8-26)
i

mz, = £ Qid mzx — nxxz — £ Qi£i£i’

Note that the potential due to the monopole moment varies as 1/r,
that due to the dipole moment as 1/r2, and that due to the quadrupole
moment as 1/r3. The term in 1/r4 would contain the octupole moment, etc.
In Fig. 8-6 are shown some charge distributions illustrating the physical
meaning of the electric moments that we have defined. In (a) only mxx is
not zero, that is, we have a linear quadrupole. In (b) we have a linear
quadrupole plus a monopole (2q). In (c) we have only mxy not zero, or
what is called a square quadrupole. In (d) the distribution has monopole,
dipole, and quadrupole moments.

Si

q — 2q 9

! A J

Linear quadrupole
(a) (b)

q . • 9
T -q
d- -X -x

1 • —q
I—d—~\
i
Square quadrupole

(c) (d)

Fig. 8-6. Quadrupoles.

152 coulomb’s law. electrostatics [chap. 8

8-7 Potential due to continuous charge distributions. The potential

due to a continuous distribution of charge (macroscopically speaking) is
obtained by replacing the summation in Eq. (8-20) by an integration. For
example, if the point P(x, y, z) is outside a region r in which the charge
density is a given function p of the coordinates, then the potential at P
due to this charge distribution is given by

(8-27)

where r is the distance from the volume element dr (containing the charge
p dr) to P. Even if the point P is within the volume r, so that there is a
charge density at the point of observation (where r = 0), Eq. (8-27) may
still be used. In this case, in the immediate neighborhood of P, we may
take for volume elements concentric spherical shells, so that dr = 47rr2 dr.
With p finite, as assumed, the integrand pdr/r = 4-rrpr dr is also finite,
no matter how small r is.
In a similar manner, the potential due to surface and line distributions
of charge may be computed by evaluating surface and line integrals, re¬
spectively.
The total potential V at a point in an unbounded region may be found
by adding the contributions due to all point-charge and continuous-charge
distributions. In a bounded region, however, the potential at a point P
cannot be calculated unless we also know the potential at every point on
the boundary. For example, within an empty hollow conductor we have a
region containing no charge, even at its boundary (the inner surface of the
conductor), yet, if the conductor is insulated, we may change the potential
at a point within this region by varying the charge on or near the outer
surface of the conductor. Problems involving boundary conditions will
now be considered.

8 8 Poisson’s and Laplace’s equations. If E = — V7 is substituted in

Eq. (8-13), we obtain

which is called Poisson’s equation. In a region where there is no volume

charge, so that p = 0, this equation becomes

V • vv = 0, (8-29)

which is Laplace’s equation. These same equations are encountered in

connection with the gravitational field and potential.* Laplace’s equation

*See Yol. I, Section 5-11.

8-8] poisson’s and laplace’s equations 153

is also met in many other branches of theoretical physics, such as the irrota-
tional flow of an incompressible fluid, steady heat flow, etc.; the scalar
function V has a different physical meaning in each case.
A typical electrostatic problem might be stated as follows. Certain con¬
ducting surfaces, each with a given size, shape, and position, are main¬
tained at stated potentials; we wish to find the electric intensity in the
region between these surfaces, and the charge density at any point on a
surface. If the space between the conductors is empty, we solve Laplace’s
equation; if a space charge exists in the region, we must be given the func¬
tion p and solve Poisson’s equation. In either case, the solution of the
partial differential equation (8-28), (or 8-29), will involve constants of
integration which must be determined from the boundary conditions.
These conditions are that V must reduce to the stated potential of a con¬
ductor when the coordinates are given values corresponding to points on
the surface of that conductor. Once the constants of integration have been
determined, we shall have found a particular solution or function V of the
coordinates which satisfies Poisson’s or Laplace’s equation, as the case
may be, and which reduces to the correct values of the potential at all
boundary surfaces. This solution will be mathematically and physically
unique, that is, for a problem such as this, which is completely defined, it
is the only solution. Once V has been found, we may differentiate V to find
the components of the electric intensity at a given point and, with the aid
of Eq. (8-18), the surface charge density on any of the conductors present.
The above theory is all based on Coulomb’s law, from which Poisson’s
equation Eq. (8-28) was derived.
The general solution of Poisson’s or Laplace’s equation is beyond the
scope of this text, but examples illustrating the solution of electrostatic
problems will be given in this and the following two sections of this chapter.
Illustration 1. One-dimensional fields. Consider a field in the X-direction
only, such as that between two parallel conducting plates, one at x = 0
and the other at x = L. Let the first plate be at a potential of V0 volts
and the other at zero potential. Let us find the electric intensity and surface
charge densities when (a) p = 0, (b) p = k, where k is a constant.
Solution. Since this is a one-dimensional problem, V is a function of x
only, and Eq. (8-28) becomes

d2V _ p
(8-30)
dx2 e0

(a) Letting p — 0, we obtain

d2V
dx2
= o, (8-31)
 [chap. 8
154 coulomb’s law. electrostatics

and, after integrating twice,

V = Ax + B.

The boundary conditions, which are V = V0 atx=0 and V = 0 at

x = L, lead to A = —V0/L and B = V0, so that our particular solution
of Eq. (8-31) becomes

V = - ^ x + V0. (8-32)

The electric intensity between the plates is

F dV Vo

which is, of course, a uniform field in the positive X-direction.

On the conductor at x = 0 the surface charge density is

eo To
1
L

while on the other conductor it is

(Note that for the right-hand conductor the outward normal is in the
negative X-direction.)
(b) Letting p = k, Eq. (8-30) becomes

d2V k
(8-33)
dx2 e0
Integration gives

V = - ~ x2 + Ax + B.
Z€0

Using the same boundary conditions as before, we find that A =

(kL/2e0) — V0/L and B = V0, so that our particular solution of Eq.
(8-33) is H

V 2e0 ^Z2 (8-34)

8-8] poisson’s and laplace’s equations 155

The electric intensity between the plates is

The field is no longer uniform, but it increases with x.

The respective surface charge densities now are

dV\ kL eoFo;

c?t\ kL cq Lq
eo^o
O’L
°L
—
“
+«0
+‘° U /,.1 " ~ V ~ T
2 L

Note that for the two plates together the net charge per unit area is
—kL, while the total space charge between these unit areas is -\-kL. In
a closed system, that is, one across whose boundary no field passes, the
total charge must be zero, according to Gauss’ law. In the present example
there is no field within either conducting plate, and so the surfaces of the
plates and the space between constitute a closed system with zero net
charge.
Illustration 2. Fields with zonal symmetry. Let us assume that the field
is symmetric about the X-axis and that p — 0. We write Laplace’s
equation in terms of spherical coordinates, omitting the 0-term, since V
is independent of 0. We then have

(8-35)

The solution of this equation is a series of zonal harmonics* namely,

V = A0Po + A\rP\ + A2r2P2 + • ■ •

(8-36)

where the P’s are Legendre polynomials, i.e.,

p0 = 1, Pj = cos 6, P2 = £(3 cos2 6 - 1), etc.

*gee Vol. I, Section 14-6; also see Margenau and Murphy, The Mathematics of
Physics and Chemistry.
156 coulomb’s law. electrostatics [chap. 8

To determine the two sets of constants we must be given the value of V

at the inner and outer boundaries of the region containing the field.
Suppose, for example, that we consider a grounded conducting sphere of
radius a, placed in a uniform external field E0. Let the uniform field be
in the X-direction and let the center of the sphere be the origin. The
uniform field will cause the free electrons in the sphere to be redistributed;
surface charges will appear and the field due to them will be added to the
original field, giving a field that is no longer uniform.
The boundaries of the field are r = a and r = oo. We are given that
V = 0 at r = a. At a great distance the sphere will produce a negligible
disturbance, so we have V = —E0x = —E0r cos 6 = —E0rPi when r
is very large. Substituting these conditions in Eq. (8-36) we get, respec¬
tively,

0 = A0P0 + AiaPi -f- A2a“P2 -}-•••

Bo

and

E0rP i — A 0Pq + A irP i -f-A 2i''2P2 T~ • • • , as r —> oo.

Each of these equations must be an identity for all values of the coordinate
6. The P’s are functions of 6, so that as d is varied, the P’s will be altered,
each in a different manner. Hence for either equation to be an identity,
each P must have the same coefficient on each side of the equation. Equat¬
ing coefficients of like P’s in each equation, we get

Ao + ~ = 0, Aia + ^A= 0, A2a2 + ^f=0, etc,

and

^-o = 0, Ai = Eq, A 2 = 0, etc.

All of the constants are zero except A1} and Px = -a3Ai = a3E0, so
for this particular problem Eq. (8-36) becomes

V = —E0r cos 6 -f- cos 6. (8-37)

The term E0r cos d represents the potential due to the uniform field and
the term a E0 cos 6/r2 is the potential due to the sphere. The latter term
is of the same form as for a dipole, Eq. (8-25).
8-8] poisson’s and laplace’s equations 157
1

The polar components of the field are

dV 2ad
Er =-r— = Eo COS 9 ~\-r- E0 COS 9,
or rA

Eg = — - = ” —Eq sin 9 + ^ E0 sin 9.

r o9 rA

This field is shown in Fig. 8-7.

Finally, the surface charge density on the sphere is given by

dV
o = — e0 \ l = 3e0E0 COS 9.
«.OT/r=a

Fig. 8-7. Sphere in uniform field.

Illustration 3. Fields depending on two or three cartesian coordinates.

In the case of a region bounded by parallel or orthogonal planes, we seek
a solution for V in cartesian coordinates. As an example of the type of
problem involved, we shall take a particular one involving only two vari¬
ables.
Consider the region between the two parallel semi-infinite planes shown
in Fig. 8-8. These planes are perpendicular to the X-axis and extend to
infinity in both the positive and negative Z-directions, but in only the
positive F-direction. The region we
are considering is bounded on the
bottom by the XZ-plane. Let us
assume that p = 0 between the
F= 0 y= o
planes. Then to find V we must
solve Laplace’s equation, which here
takes the form

(8-38)
dx2 dy2

Z
(From symmetry, V cannot depend
on z.) Fig. 8-8. Semi-infinite planes.
158 coulomb’s law. electrostatics [chap. 8

Equation (8-38) is a homogeneous partial differential equation in two

variables. Equations of this type are encountered and solved in mechanics
and in the study of heat flow,* examples being (1) the one-dimensional
wave equation, (2) the heat equation for one-dimensional variable flow,
and (3) the heat equation for a semi-infinite slab in a steady state. Only
for (3) does the differential equation have exactly the same form as Eq.
(8-38), i.e., it contains two second order derivatives. Of course, the de¬
pendent variable has a different physical meaning in our present problem
than it has in the examples referred to. However, the mathematical pro¬
cedure is similar in each case.
We integrate Eq. (8-38) by the method of separating the variables. This
means that we put
V= X(x)Y(y), (8-39)

where X(x) is a function of x only and Y(y) is a function of y only. Sub¬

stitution of Eq. (8-39) in Eq. (8-38) leads to

1 d2X _ 1 d2Y
(8-40)
X dx2 Y dy2 '

Since the left side of this equation is a function of x alone and the right
side is a function of y alone, x and y independent variables, each side must
be equal to the same constant, called the constant of separation. We shall
call this constant a; its value will be determined by the boundary conditions
of our particular problem. We may then write the two ordinary differential
equations
d2X d2Y
aX and —aY. (8-41)
dx2 dy2

Integration of these equations gives

X = Aie^x + A2e~^ax,
(8-42)
Y = Ble^v + B2e~^v .

If a > 0, A varies exponentially with x and F varies sinusoidally with y,

while if « < 0, X varies sinusoidally with x and F varies exponentially
with y. In our present problem we shall find that is imaginary or
a < 0.

*See Vol. I, p. 215, p. 266, and problem 15-6.

8-8] poisson’s and laplace’s equations 159

Substituting the solutions in Eq. (8-42) back in Eq. (8-39) and combin¬
ing constants, we get

V = Ae^xe^v + A'e^xe~^y
(8-43)
+ Be~raxe^~av + £'e-V^e-V=^
We are now ready to introduce the boundary conditions for our particular
problem, in order to determine the constants. Equation (8-38) contains
second order derivatives in both x and y, so that it was necessary to inte¬
grate twice with respect to each variable. Hence for boundary conditions
we must know the value of V at (a) two values of x for all values of y,
and (b) two values of y for all values of x.
Let us take our vertical planes to be grounded plates at x — 0 and
x = L, and let us take the bottom plane (y = 0) which closes the gap to
be an insulated plate which is held at the constant potential Vo- The
effect of this plane must decrease as y —> oc. Hence we have

(a) V = 0 when x = 0 and x = L,

(b) V = V0 when y = 0 and V —> 0 when y —> oo.

The second part of condition (b) tells us that_a must be negative, since
otherwise y/—a would be imaginary and e±v'~aV would oscillate in value
but not approach zero as y approaches infinity. Then, with y/ a real, we
must have A = B = 0. _
Putting a = —P2, where P is real, and i = y/—l, we have

V = A'eipxe~pv + B'e-ipxe~pv.

If we use the identity eix = cos x + i sin x and introduce new constants,
this may be written

V = e~Pv(C sin fix + D cos (3x).

Now from condition (a) we have D = 0 and

717T 1 O Q
sin PL = 0, P = -jj ’ n = 1, 2, 3, • .

Adding the solutions for each value of n in order to obtain a more general
solution, we get

V = Z Cne-n*v,L sin *) • (8-44)

160 COULOMB S LAW. ELECTROSTATICS [chap. 8

Application of our remaining condition, V = V0 when y = 0, leads to

the determination of the Cn by the Fourier series method.* Our particular
problem is then solved, for by differentiating V we may obtain the com¬
ponents of the electric field at any point between the plates, and the surface
charge density a[= —e0(dF/dn)} anywhere on the plates.
The method given above may be extended to a three-dimensional
problem involving cartesian coordinates.
If the problem is further complicated by the presence of a space charge
(p must be a known function of the coordinates), then we must solve
Poisson’s equation

V • VF =-- • (8-28)
«0

Since this is not a homogeneous differential equation, its solution is much

more difficult and is left to more advanced texts. However, we may note
that the solution may be considered to consist of two parts, namely, (1)
a general solution of the complementary equation, and (2) a particular
solution of Poisson’s equation. The complementary equation is Laplace’s
equation, solutions of which we have been considering.

8-9 The theory of images. One class of electrostatic problems involves

a charge in the neighborhood of a conducting sphere, cylinder, or infinite
plane that is maintained at a fixed potential. (See Figs. 8-9 through 8-12.)
In such problems it is possible to find a group of charges which, without
the conductor, will give a field whose potential function satisfies Laplace’s
equation and also the boundary conditions. In particular, this potential
function must take the value of the potential of the conductor at all points
corresponding to the surface of the conductor. Since the above conditions
are sufficient to determine the potential uniquely, this group of charges
must then give the same potential and field as that of the original charge
and conductor. Thus the original conductor-plus-charge problem may be
replaced by a simpler multicharge problem with the same field in the region
outside (but not inside) the conductor. The charges which, with the
original charge, give a field equivalent to that produced by the original
charge and conductor are called the images of the given charge in the
conductor. This substitution of image charges for a conductor is called
the method of images.
Illustration 1. Point charge and infinite grounded plane. As in Fig 8-9
let the charge 9 be at a distance d to the right of an infinite grounded
plane. We may take the plane to be at x = 0.

*See Vol. I, pp. 268-269.

8-9] THE THEORY OF IMAGES 1G1
i

Fig. 8-9. Point charge and grounded plane.

The boundary conditions are that V = 0 at x = 0 and V 0 at x — go

—

(the effect of the charge being negligible at a great distance). Both of

these conditions can be satisfied if the plane is replaced by a charge — q
at a distance d behind the plane and opposite q. The image charge — q is
at what would be the position of the mirror image of q, if the plane were
reflecting. The first boundary condition is satisfied because every point on
the surface x = 0 is equidistant from charges of +g and —q.
The potential due to two point charges, according to Eq. (8-20), is

1
V = (8-45)
4tT€0

where rq and r2 are the distances to the two charges. At the point P(x, y, z)

r\ = V(x — d)2 + y2 + z2, r2 = V (x + d)2 + y2 + z2.

Equation (8-45) becomes (with q' = —q)

g _1_ _1_
V = (8-46)
47T6o \/(x — d)2 + y2 + z2 VW+ d)2 + y2 + 22

We see that this satisfies both boundary conditions. Each term in Eq.
(8-45) or Eq. (8-46) satisfies Laplace’s equation and so V does also.
This must then be a solution of the problem.
For convenience, let the XF-plane include P and the charges. The
components of the field to the right of the plane (behind the plane there
is actually no field) are
dV q I (x ~ d)_(x + d) |

E* = “ “ 47T60 \[(x - d)2 + y2]3l2 [(x + d)2 + y2\ 3/2) ’

„ dV q (_y__ _y_ . 1
E* = _ ~du ~ 47T60 \[(X - d)2 + y2p/2 [{x + d)2 + y2?12J
(8-47)
162 coulomb’s law. electrostatics [chap. 8

The charge density induced on the plane at a distance y from the origin is

^ 60 U>Lo 2ir (d2 -f- ?/2)3/2 ' (8 48)

This distribution is shown in Fig. 8-10, in which the thickness of the

shaded part is proportional to a. Integration of a over the entire surface of
the plane shows that the total charge induced on the plane is —q, the image
charge (see problem 16 at the end of this chapter). The field is also shown
in Fig. 8-10, where the solid lines represent the actual field for the charge
and plane problem, while the solid and dotted lines together show the
field due to the two charges q and —q.

Fig. 8-10. Induced charge on grounded plane.

7- 0

~9- • 9

V= 0
9* • —9

Fig. 8 11. Image of point charge in two planes.

In a problem (Fig. 8-11) where the grounded plane is bent to form two
planes at right angles, we must introduce three image charges in order to
satisfy the boundary conditions (V = 0 for both the vertical and horizontal
planes) These image charges again resemble mirror images. Behind the
vertical plane we put an image charge -q as before; below the horizontal
8-9] THE THEORY OF IMAGES 163

plane we put another image charge —q, the same distance from the plane
as the given charge q; finally, a charge of -\-q, the image of the image
charges, must be located, as in Fig. 8-11. The field in the quadrant be¬
tween the planes is that of the four charges.
When two planes meet at any angle 0, a finite number of images is ob¬
tained only when 6 = w/n, where n is a positive integer.

Fig. 8-12. Charge and grounded sphere.

Illustration 2. Point charge and grounded sphere. Consider a point charge

q at a distance d from the center of a sphere of radius a < d, as in Fig. 8-12.
Let the center of the sphere be the origin.
The boundary conditions are V — 0 at r — a and V = 0 at r = go.
These may be satisfied by replacing the sphere with an image charge q'
located at a distance b from the center of the sphere on the line toward q,
provided that

q' = — % q and b = ^• (8-49)

* a d

To prove this, we shall first find the potential due only to q and q' and then
show that it satisfies the boundary conditions.
At any point P(r, 6) we have for the potential due to q and q'

V =
r~
4tt€o (-
Vi + r2/
-)>
as in Eq. (8-45). Substituting for r1 and r2 and using Eq. (8-49), we get

V =
1 __g_
€q
VV2 + d2 — 2r d cos 6
a _q_
d Vr2 + (o4/d2) — 2r(a2/d) cos 6 (8-50)

When r = a this reduces to V 0, and when r =

— go we again get V = 0,
so Eq. (8-50) satisfies our boundary conditions.
164 coulomb’s law. electrostatics [chap. 8

The components Er and Ee of the electric intensity and the surface

charge density on the sphere can be found by differentiating V. Here use
is made of the relations

The field obtained is that outside the sphere only (inside it is zero). The
total induced charge is q'. The force of attraction F between the charge
q and the grounded sphere is that between q and its image, or

q2 a 1
(8-51)
47re0 d (d — a2/d)2

For further applications of the image method the reader is referred to

other texts.*

8-10 Conjugate functions. Here we must consider only problems where

V is independent of one rectangular coordinate, or V = V(x,y), say.
Take any complex function/^), where

2 = x + iy, i = V—1.

This complex function can be broken up into a real part g(x, y) plus an
imaginary part ih(x, y), that is,

Six + iy) = g(x, y) + ih(x, y). (8-52)

If we let g(x, y) equal a successive series of constants, we will get a family

of curves. Similarly, letting h(x, y) have various constant values will give
us a second family of curves. We then have the

Theorem. The two families of curves, g(x,y) = a constant and h(x,y) = a

constant, are orthogonal.

To prove this theorem consider the gradients of g and h, that is, Vg

and Vh. The former will be normal to the family of curves g = a constant,
while the latter will be normal to the family of curves h = a constant.
The scalar product of these two vectors must equal the product of their

*See Jeans, The Mathematical Theory of Electricity and Magnetism.

8-10] CONJUGATE FUNCTIONS 165
f

x-components plus the product of their ^-components, or

dg dh
(8-53)
- %t dy dig
Since / is a function of 4-

dl=idl.
dy dx

If we substitute g + ih for / in this equation, we get

dg . dh . dg dh
dy 1 dy 1 dx dx

When we equate first the real and then the imaginary terms on each side
of the last equation, we obtain

dg dh dh dg
dy dx ’ dy dx

Substitution of these relations back in Eq. (8-53) shows that the scalar
product of the gradients is zero. The gradients, and hence the two
corresponding sets of curves, must intersect orthogonally.
We next prove the following

Theorem: The functions g(x, y) and h(x, y) each satisfy Laplace’s equa¬
tion.

From the last previous equation we have

-2 d2h
d2g d2h d g _
dy2 dydx dx2 dxdy
Since
d2h d2h
dxdy dydx '
we have
dg + —
n2 _

= 0.
dx2 ^ dy2

In a similar manner we can show that

d2h d2h
dx2 + dy2

Thus g and h satisfy Laplace’s equation. To this extent each represents a

possible potential function V(x, y).
166 coulomb’s law. electrostatics [chap. 8

Let A and B be constants that are given a succession of values. Then if

one of the two families of curves, g(x, y) — A and h(x, y) = B, represents
the traces on the WF-plane of the equipotential surfaces of an electrostatic
field, the other family of curves will represent the lines of force. This
follows because equipotential surfaces and lines of force are orthogonal.
Furthermore, any equipotential curve may be taken to be the trace in the
XF-plane of a conducting surface parallel to the Z-axis; in this case the
field to be considered is that between, or outside of, the conductors.
The systematic method of finding the appropriate function f(z) for a
particular problem is complicated and will not be taken up here. However,
various simple functions can easily be tried and the corresponding problems
described and listed. We shall give one example.
Illustration. Take f(z) — kz2 = k(x + iy)2, where k is a constant.
Expanding, we get/(z) = k(x2 + 2ixy — y2), so that

g{x, y) = k(x2 — y2), h(x, y) = 2kxy.

Putting each of these equal to a set of constants yields

k(x2 — y2) = A, 2kxy = B. (8-54)

Equation (8-54) represents two sets of hyperbolas that intersect

orthogonally.
Let us take 2kxy to be the potential function, or

V = 2 kxy. (8-55)

Then the equation 2kxy = B will, as we vary B, give us the equations for
different equipotentials. These will be equilateral hyperbolas, as shown in
Fig. 8-13, and they may be thought of as traces in the XF-plane of equi¬
lateral hyperbolic cylinders extending parallel to the Z-axis.
Now consider the equipotential for which B = 0; this is represented by

xy = 0 or x = 0, y = 0.

This corresponds to the YZ- and XZ-planes. Suppose that we take these
planes to be conducting. Taking V = 0 for these planes simply means
that potentials at all other points are given relative to the potential of these
planes. We may now investigate the field in the quadrant between these
planes, for which x, y, and B are all positive. In order to have a bounded
region, we may think of this field as extending out to another equipotential
surface which could be another conductor at the potential of that surface.
This would be a hyperbolic cylinder, and we can take it as far from the
origin as we wish.
8-10] CONJUGATE FUNCTIONS 1G7
<

Fig. 8-13. Field near corner of conducting planes.

Equation (8-55) satisfies Laplace’s equation, d2V/dx2 -f- d2V/dy2 = 0,

and it satisfies the boundary conditions. We may then differentiate to find
the electric field and surface charge density on the conductors:

dV
= -2t*'

(8-56)
dV
Ey= - -2 kx,
dy

(av\ 2e0ky for the vertical plane,

C° \dx/x=o

(J =
(d_v) —2 t0kx for the horizontal plane.
€° \dy/v=o

Note that the charge density and field increase with distance from the
corner of the planes outward.
The direction of the field could also be found from the first theorem we
derived, which tells us that the two sets of curves given by Eq. (8-54)
are orthogonal. Since we let 2kxy = B represent our equipotentials,

k(x2 - y2) = A (8-57)

must represent the lines of force. The lines of force are hyperbolic and, for
the quadrant we are considering, asymptotic to the line x — y, as shown
by the dotted lines in Fig. 8-13. To prove that Eqs. (8-56) and (8-57) both
168 coulomb’s law. electrostatics [chap. 8

lead to the same field, let us write the differential equation for a line of
force. Since an element of a line of force has, by definition, the direction of
the field, the components dx and dy of such an element must satisfy the
equation
dx _ d y
(8-58)
Ex ~ Ey '

Substituting for Ex and Ey from Eq. (8-56) and integrating, we get

dx _ dy
or x dx = y dy ,
y ~ x

2
x a constant,

which agrees with Eq. (8-57).

The problem is still not completely defined unless we know the value of k.
Here is where the potential at the other boundary enters in. For example,
if we take this boundary to be given by xy = 20 m2 and specify its poten¬
tial as V = —100 volts, then from Eq. (8-55) we must have k = —2.5
volts/m2. Now both the magnitude and direction of the field are deter¬
mined, as well as the charge density (it is positive in this example) on the
YZ- and XZ-planes.

8-11 Energy stored in an electrostatic field. To set up an electrostatic

field requires work, for unlike charges must be separated. The conserva¬
tion of energy principle tells us that the work done in producing the field
must be stored as potential energy which can be recovered when the unlike
charges are allowed to neutralize one another.
We may think of this potential energy as being associated with the
positions of the charges, or as being stored in the region containing the
field. Actually it seems immaterial which viewpoint we take when we con¬
sider that energy is not measured in a static situation, but only in a dy¬
namic one. However, we know that energy can be transported through
space by electromagnetic waves, so it is a useful concept to think of energy
as residing in an electrostatic field.
To find an expression for the energy density, or energy per unit volume,
of an electrostatic field, consider a uniform field between parallel plates a
distance l apart. Let us start with the plates uncharged, so that E — 0,
then let us gradually transfer electrons from one plate to the other until
we have built up the field to its final value E. We now compute, for this
chaiging process, the total work done per unit area of plates. Suppose
that when the charge density on the positive plate is ^ (« is to increase
from 0 to 1), we transfer a small charge da = a da from the negative to
 PROBLEMS 169

the positive plate. At this instant the field (force per unit charge) will be
acr/e0 (see Eq. 18-12), so that the work done on da will be

dW = (air/ e o)(a da)(l).

The total work done, or energy supplied to the system, in building up the
charge a will be
2 7 f1
W = — / a da
Co Jo
1 a2 3 4l
2 eo
= hoE%
since E = a/e0. Since the field is uniform, we may picture this energy W
as being stored throughout the region (of volume Z) between unit areas of
the plates. We then have

Energy density of electric field = %e0E2. (8-59)

Problems

1. Show that it is only because is uniformly distributed throughout a

Coulomb’s law is an inverse square law spherical volume of radius R, the
that when we draw lines of force charge per unit volume being p, use
radiating out from a point charge, Gauss’ law to find the electric intensity
the density of the lines in a given region inside the volume, at a distance r from
is a measure of the intensity of the field the center.
in that region. fm)Using Gauss’ law, find the total
2. A thin positively charged ring of clWges on the inner and on the outer
radius a has a charge q uniformly dis¬ surfaces of a hollow conductor, when a
tributed around the circumference. charge q is placed on the conductor
Find by integration (see Section 8-2) and a net charge of q' is placed in the
the electric intensity at a point on the cavity.
axis of the ring and at a distance x 6. Show mathematically that V X
from the center. E; = 0, where E,- is the electric field
3. Find by integration the electric due to a point charge gq.
intensity to the right of a thin infinite 7. Show that if the law of force be¬
plane which has a charge density of a tween charges were other than an in¬
on each side. Check, using Gauss’ law. verse square one, then inside an empty
4. (a) Using Gauss’ law, find the uniformly charged spherical shell, at a
electric intensity inside a charged point other than the center, the electric
spherical shell, (b) If a positive charge intensity would not be zero. In such a
170 coulomb’s law. electrostatics [chap. 8

case a charge would be drawn from the 15. Two conducting spheres are con¬
outer to the inner surface of a charged centric, the radius of the inner being
spherical conductor. The fact that no a and that of the outer b. One sphere
such charge is found on the inner sur¬ is grounded and the other is at a po¬
face was shown by Cavendish to be a tential Vo■ Find the charge on each
sensitive experimental proof of the when the grounded sphere is (a) the
inverse square law. outer one, (b) the inner one.
C'SyTwo horizontal metal plates are 16. By integration of the charge
1 cm apart in air. An oil drop bearing density a, find the total charge induced
a charge of one electronic unit is bal¬ on a grounded plane when a charge q
anced between the plates when the is placed in front of the plane.
p.d. between them is 800 volts. Find 17. (a) Find the image charges when
thnmass of the drop. a charge q is at equal distances d from
(T9. JAn electric field is represented by two grounded planes meeting at 60°.
E^= ay, Ev = ax, where a = 100 (b) Find the force on the charge q.
volts/m2. Find (a) the potential func¬ 18. A conducting sphere of radius a
tion V, taking V = 0 at the origin, is at a potential Vo. The center of the
(b) the work done by the field when sphere is at a distance d/2 from an in¬
10-8 coul is taken from x — —1 m, finite grounded plane. Neglecting
y = —2 mtoz = 2 m, y = 3 m, and terms smaller than a2/d2, find the
(c) the charge density p at any point. image charges and the charge on the
10. If the electric potential for a sphere.
field is V = ax/y, where a is constant, 19. The centers of two conducting
find Ex, Ey, and the charge density at spheres are a distance d apart. One
any point P{x, y). sphere, of radius a, is at a potential Vo,
11. (By direct application of Eq. the other, of radius b, is grounded.
(8^=20), find the electric potential at the Find the charges on each sphere, neg¬
point P{r, 0) due to an electric dipole lecting terms smaller than (a) a/d and
consisting of a charge -\-q at x = d/2, b/d, (b) a2/d2, ab/d2, and b2/d2.
y = 0 and a charge — q at x = —d/2, 20. Consider the complex function
y = 0. Use polar coordinates and take f(z) = kz = k(x iy). To what elec¬
r^>d. trostatic problem does it lead, if we
12>Find directly the potential at follow the procedure of Section 8-10?
Pfe-tf) due to the quadrupole shown 21. Repeat problem 20, taking/(z) =
in (a) Fig. 8-6(a), and (b) Fig. 8-6(c). kz112.
13. Solve Poisson’s equation, given 22. A sphere of radius a bears a
that the field is in the X-direction and charge q. (a) Find the work done in
P = kx, where k is constant. Take as
charging the sphere, (b) From Eq.
boundary conditions V = 0 at x = 0, (8-59), by integration, find the energy
V = Fq at x — L. stored in the field surrounding the
14. Solve (so far as possible) La¬ sphere.
place’s equation for a region bounded
by a rectangular box. Introduce ap¬
propriate boundary conditions.
 CHAPTER 9

AMPERE’S LAW. STEADY MAGNETIC FIELDS

9-1 Introduction. From the outline given in Chapter 7 we recall that .

magnetic fields act on, and are produced by, moving charges, or current
elements. Since moving charges and current elements have direction and
so are vector quantities, whereas a static charge has only scalar properties,
the discussion of magnetic fields is more complicated than that of electro¬
statics. However, in each case we start with an experimental law and de¬
velop our theory logically from it; the two procedures have much in com¬
mon so far as method goes, although not in the resulting expressions.
Therefore we shall commence by outlining the steps which we shall take
in the present chapter and comparing them with the corresponding ones
taken in the development of electrostatics.
(1) We start with an experimental law, the law of force between current
elements (Ampere’s law). This corresponds to the law of force between
static charges (Coulomb’s law) in electrostatics.
(2) In terms of the (magnetic) force postulated by Ampere’s law, we
define the magnetic field vector B. This corresponds to the definition of
the electric field E in terms of the force postulated by Coulomb’s law.
(3) Using the definition of B together with Ampere’s law, we compute
the magnetic field due to a given distribution of current elements or cur¬
rent circuits. In electrostatics we used the definition of E and Coulomb’s
law to compute the field due to a given distribution of charges.
(4) With the aid of the vector B we find an “integral statement” of
Ampere’s law, just as we found in Gauss’ law an “integral statement” of
Coulomb’s law.
(5) We examine the properties of the vector B, finding expressions for
its curl and divergence, just as we found that V x E = 0 and V • E =
p/€q in electrostatics.
(6) We look for potential function(s) from which B may be derived,
just as we found that E could be derived from the electrostatic potential
function V.
If these steps are kept in mind, the logical development of the theory of
steady magnetic fields will be much clearer.

9-2 Ampere’s law of force between current elements. As stated in

Section 7-6(a) (see Fig. 7-2), two current elements I ds and I' ds' a distance
r apart interact with a force given by

d‘f = cor - - (f—^ • (9-1)

171
172 ampere’s law. steady magnetic fields [chap. 9

We assume that 7 and /' are steady currents. The direction of the unit
vector rj is from one current element (say I ds) toward the other (/' 7s'),
on which the force is to be calculated. * When 7s and 7s' are parallel or
antiparallel, the two current elements interact on each other with equal
and opposite forces, and when the currents are in the same direction the
interaction is attractive, as explained in Chapter 7.
It must be admitted that Eq. (9-1) cannot be verified by direct experi¬
mental testing because in practice we deal with complete current circuits,
not with idealized current elements. This objection may be answered as
follows.
We shall see that Eq. (9-1), which can be termed a “point statement”
(since current elements have small spatial dimensions), is more general
than any of the formulas that give the force between whole circuits, since
the latter statements depend on the size and shape of the circuits. Our aim
is to express a fundamental law of nature in its most general form and then
to apply it to special cases. If from our general expression we derive ones
that can be tested experimentally and which, when so tested, are always
in agreement with fact, then our original “law” can be considered to have
the necessary validity. The statement of the law of gravitation in terms
of two idealized point masses and its subsequent extension to finite bodies
is a case in point. A statement such as Eq. (9-1) may be called a shrewd
“guess,” based on hindsight, but after all it is due to such guesses that
theoretical physics has progressed and men like Newton and Planck have
become famous.
We have seen that the factor C\ is arbitrarily chosen, thus determining
the unit of current or charge. The ampere, and hence the coulomb, are
defined by taking

Ci = 10“7 newton/ampere2. (9-2)

If, as with Coulomb’s law, we wish to introduce explicitly a factor of 47t,

we can change to another constant mo = 47t C\ and write Eq. (9—1) as

where mo is the magnetic permittivity of free space. Its value is

Mo = 47r X 10-7 newton/ampere2. (9-4)

Throughout this chapter the notation will generally be as follows: primed

symbols refer to current elements or charges on which the magnetic field acts,
corresponding unprimed symbols refer to current elements or charges that give
rise to the magnetic field.
9-4] THE MAGNETIC FIELD. DEFINITION OF B 173

In important expressions later; the factor 4tt will not appear if we use p0
rather than C i.

9-3 Moving charge equivalent to a current element. Equation (9-1)

gives the force on a current element. Since B was defined (Section 7-5) in
terms of a moving charge, we must consider the relationship between a
current element and a moving charge.
When there is a current I' in a conductor, such as a piece of wire, we
picture free electrons moving between a corresponding number of station¬
ary positive ions. This means that in a current element we have the
relative motion of a certain amount of charge, say q'. Assume that this
charge has an average forward velocity v'. Call the cross section of the
wire A and the density of the free charge p. Then the charge crossing any
section per unit time is pAv'. This is the current I' present in the wire.
Hence we have

I' ds'= pAv' ds' = pA ds'v',

the direction of the forward velocity being the same as that of ds'. Since
pA ds' = q', we obtain
I'ds' = q'v', (9-5)

which means that the force on a current element I' ds' is the same as it
would be on a charge q', at the same location and moving with a velocity
v', provided that I' ds' = q'v'.
We may now say that for a charge q', moving with the velocity v', the
force due to a current element I ds, according to Ampere’s law, is

dF = to J , V- X (ds X £t) ;
4tt r2

where r is the distance from the current element to q' and rx is a unit
vector in that direction.

9-4 The magnetic field. Definition of B. In Chapter 7 we stated that

when a moving charge experiences a force which a stationary charge (at
the same point) would not experience, then a magnetic field is present. In
other words, surrounding a current element is a magnetic field which is
capable of acting on a moving charge placed in that field. We shall use
this field point of view in place of the action-at-a-distance concept.
A definition of the magnetic induction B was given in Section 7-5 (b),
but we shall now define it more explicitly. From Eq. (9-6) we see that,
all other quantities remaining unchanged, the magnitude of dF will depend
on the direction of v', being a maximum when v' is in the plane perpendic-
174 ampere’s law. steady magnetic fields [chap. 9

ular to (ds x rx), as in Fig. 9-1. Taking v' to be in the plane giving maxi¬
mum dF, we define the magnitude of the magnetic induction at q' as

ldB| = dF^, (9-7)

where we use the symbol dB in place of B because we are dealing with a

current element, involving the infinitesimal ds. The direction of the mag¬
netic induction at q' is taken to be such that the vector product v' X dB
has the direction of dF. From Eq. (9-6) this implies that dB has the di¬
rection of (ds x r:). The above definitions of the magnitude and direction
of dB are consistent with the statement that

dB = I ds * ri, (9-8)
4t r2 v

as first proposed by Biot in 1820. Equation (9-8) obviously gives dB the

desired direction; that it gives the magnitude defined by Eq. (9-7) may
be seen by letting v' be perpendicular to ds x rx (and hence perpendicular
to dB), in which case Eq. (9-6) becomes

MO t / / ds X Ti
dFn = — la v = gV|dB|
4tt *

Direction of ds X and dB

Fig. 9-1. Magnetic force on moving charge.

As stated in Section 7-5, or as may be seen from Eq. (9-7), the mksc
unit for magnetic induction is the n-sec/coul-m, the n/amp-m, or the
kg/coul-sec. If we define

1 weber = 1 newton-meter-second/coulomb,

we may express B in webers/(meter)2, or w/m2. The weber is introduced

because B is a field vector and it is sometimes helpful to picture it as
being represented by lines of flux. These lines are taken to have the
direction of B at any point and to represent by their concentration the
relative strength of the field. Thus where B = 8 w/m2 we might draw 8
lines of flux per square meter, in which case webers would mean lines of
flux. In the following chapter we shall see that the total flux (or webers)
crossing a surface is an important physical quantity.
9-5] MAGNETIC INDUCTION AND FORCE 175

9-5 Magnetic induction and force due to a current circuit. We shall

now extend our theory from consideration of a current element to that of
a complete circuit.
Since dF in Eq. (9-6) is found to have all the properties of force, and
since forces add vectorially, we may integrate Eq. (9-6) over all the
current elements of the entire circuit of which I ds is a part. Then, since
I, q', and v' are constant, we get

F = g IqW * / 0-9)
s

for the force on a charge q' moving with the velocity v', due to the circuit
containing the current I.
Since dB is proportional to (ds X ri)/r2, which, according to Eq. (9-9),
is integrable, we may integrate Eq. (9-8) and so get

» = g1 / S
(9“10)
for the magnetic induction due to a current circuit.
Substituting Eq. (9-10) into Eq. (9-9), we obtain

q'v' x B (9-11)

for the magnetic force on a moving charge due to the field B. This is also
referred to as the Lorentz force. Since this equation follows from our defini¬
tion of magnetic induction and is consistent with Ampere’s law, it may be,
and frequently is, taken as defining B.
Illustration 1. Field due to a linear current circuit. Consider a current I
in a long straight wire, the return portion of the circuit being so far away
that its effect can be neglected. Let us compute B at an external point P
whose perpendicular distance from the
axis of the wire is R. Referring to Fig.
9-2, in which 6 is the angle between
direction of the current and that of iq,
we see that

ds X ri ds sin 6
into the paper.

Let s be the distance of the element

ds from the point 0, where OP is
perpendicular to the wire. Then R/s Fig. 9-2. Magnetic induction near
= tan 6, or straight current circuit.
176 ampere’s law. steady magnetic fields [chap. 9

S = R cot d, ds — —R esc2 d dd, r — R esc d.

If we take 6 as the independent variable, its limits being 0 and tt for a

very long wire, Eq. (9-10) becomes

B = ~ ~t into the paper at P. (9-12)

Zir K

This is called the law of Biot and Savart.

The lines of flux representing B form closed circles about the wire, as
given by the following right-hand fist rule: If one imagines the wire to be
grabbed in the right fist with thumb pointing in the direction of the positive
current, then the fingers will curl in the direction of the lines representing the
magnetic induction, or field. (To avoid confusion, no other hand rules will
be given.)
Illustration 2. Magnetic deflection of a beam of charged particles. Suppose
that we have a beam of particles, each with a mass m, a charge e, and a
velocity v (primes have been dropped here). Let the beam be moving
through a region where the magnetic induction has a uniform value B.
From Eq. (9-11) the force on each particle is

F = ev x B.

Let B\\ and Bx be the components of B parallel and perpendicular to v,

respectively. Then B\\ will not contribute to the force and may be dis¬
regarded. But due to Bx a force evB_L will act at right angles to the direc¬
tion of motion and to that of B. As a result, the motion in the direction
of B will remain unchanged, but the vector v will rotate in a cone about the
direction of B. In other words, the path of the beam will be a helix (spiral)
with its axis parallel to B. Charged particles traveling in the earth’s
magnetic field are found to follow just such paths.
When v is perpendicular to B we have F = evB. Since this force is at
right angles to v, the motion is circular. From Newton’s second law
(F = ma) and the expression for centripetal acceleration, we get

e v
m BR (9-13)

which is an extremely useful equation when we want to determine the

mass or speed of an atomic (or subatomic) particle.
9-6] MAGNETIC FORCES ON CURRENT CIRCUITS 177

In cloud tracks we can observe the motion of an individual particle, or

we can define with slits the motion of a beam of charged particles, and
from the shape of the path verify the existence of magnetic forces as postu¬
lated by Ampere’s law.

9-6 Magnetic forces on current circuits. The magnetic force acting

on a moving charge is given by Eq. (9-11). To find the magnetic force on
a finite section, or the whole, of a current circuit, we return to the original
statement of Ampere’s law, namely,

_ Mo jji ds' X (ds X ri) ; (9-3)

— r2

which gives the force of interaction between two current elements. We

wish to integrate this expression with respect to both current circuits (7
and I').
First, integrating with respect to s, we get for the force on the current
element I' ds',

d¥ = ? W ds' x • (9-14)
47r J
S
r2

Using Eq. (9-10), this becomes

dF = V ds' X B, (9-15)

where B is the magnetic induction at ds' due to the circuit with the current
7. If more than one circuit contributes to the field at ds', we may sum
Eq. (9-15) for each.
Second, to find the force on the I' circuit between, say, the points a
and b, we integrate Eq. (9-15) with respect to s’:

F = 7' f ds' x B. (9-16)

Ja

Illustration 1. Force between parallel current circuits. Consider two paral¬

lel wires a distance R apart and carrying the currents 7 and 7', respectively.
B is the magnetic induction at ds' due to the current 7. Assume the wires
to be very long; then B is constant along the wire carrying 7' and its value
is given by Eq. (9-12), namely,
J
B _ 71° :L perpendicular to ds'.
2ir R

Integrating Eq. (9—16) for a length l, we have

 178 ampere’s law. steady magnetic fields [chap. 9

F = 7

F _ no (2II'\
2 X 10~7 (9-17)
l 4tt \ R )

the force being attractive when the currents are in the same direction and
repulsive when the currents are opposed.
Equation (9-17) leads to a simplified definition of the ampere. We see
that when I = T = 1 amp and R = 1 meter, then the force per meter is
2 X 10 7 newton. We could use Eq. (9-17) to measure current, letting
I = I' and measuring R and the force on a movable section of one wire;
this would be an absolute determination of the current, i.e., it would be
independent of any electrical standards. While this experiment can be
performed, it is found in practice that it is more accurate to measure the
force between current-carrying coils, for example with the Kelvin current
balance, deriving the appropriate formula from Eq. (9-16). The instru¬
ments used for such measurements are called electrodynamometers.
Illustration £. Torque on a rectangular current circuit in a uniform mag¬
netic field. Consider a rectangular coil suspended so that its plane is parallel
to the magnetic induction B, as in Fig. 9-3. Let the length of the coil
perpendicular to the field be l and its width w, and suppose that there is
a current /' in the coil.
We use Eq. (9-16) to compute the force on each side of the coil. For the
top and bottom the force is zero, since ds' and B then are parallel and
ds X B = 0. For the side ab, which is perpendicular to B, the force per
turn will be /7£ normal to the plane of the coil. On the opposite side cd the
force will have the same magnitude, but the opposite direction. The
result is a torque L whose magnitude is given by

L = NI’lBw = NI'AB, (9-18)

where A is the area of the coil and

N the number of turns.
If the coil is allowed to turn, the
torque due to the field will decrease
and the restoring torque of the
suspension fiber will increase until South pole

equilibrium is reached. Letting 6 be

the angle turned through and k the
stiffness constant for the fiber, we
then have

L = NI'AB cos 6 = he. (9-19)

Fig. 9-3. Galvanometer coil.
9-7] THE CIRCUITAL FORM OF AMPERE’S LAW 179

If N, A, and k are known and 0 is measured, we have a means of measuring

B in terms of Without knowing N, A, or k, Eq. (9-19) can still be used
to compare currents; in most moving-coil galvanometers the factor cos 0
is nearly unity and I' is proportional to 0.

9-7 The circuital form of Ampere’s law. In the remainder of this

chapter we shall discuss some of the interesting and important properties
of a steady magnetic field. We turn first to step (4) in the outline given
in Section 9-1, that is, we shall seek an “integral statement” of Ampere’s
law.
Let us find the line integral of B around a closed path X; we designate this
integral as >\B • dX, where dX is an element of the path.
Suppose that the current loop shown in Fig. 9-4 produces a magnetic
induction B and subtends a solid angle 12 at the point P. If we displace P
by a distance dX, the solid angle will change by an amount d!2. The same
change d!2 will result if, instead of moving P, we displace the current loop
by — dX, that is, by the same distance in the opposite direction. In the
latter event the change in the solid angle will be the sum of all the ele¬
mentary changes brought about as each element ds of the current circuit
is displaced by — dX and thus sweeps out a small parallelogram of surface
—dX X ds. If this surface lies with its plane normal to the vector r (from
ds to P), then it will subtend at P a solid angle equal to its area, |—dX X ds|,
divided by r2. In general, the solid angle will be (—dX x ds)'i\/r2,
where r1 is a unit vector in the direction of r, and this may be written as
—dX-(ds X Ti)/r2, since the dot and cross are interchangeable in a triple
vector product. Integrating this expression around the current loop, that
is, with respect to s, we have for
the change in 12

d!2 = -dX- j (9-20)

S

From Eq. (9-10) the integral in

Eq. (9-20) is just B(47t/Vo-0- When
we make this substitution and solve
for B • dX, we get

B • dX =» - ^ d!2. (9-21)
47T

The minus sign means that we are

taking 12 to be positive on that side Fig. 9-4. Solid angle subtended by

of the current loop for which the current circuit.

 180 ampere’s law. steady magnetic fields [chap. 9

field B is directed away from the loop; on the other side ft will be con¬
sidered to be negative.
To find the line integral of B around a closed path X we let dX be an
element of such a path* and, using Eq. (9-21), obtain

f B • dX = - ^ j> dft. (9-22)

x

We see that the value of the line integral depends on the total change in
the solid angle ft. Two cases arise, as follows:
(a) The path of integration is not linked with the current circuit. Take,
for example, the path a shown in Fig. 9-5. If we go around this path, ft
alternately increases and decreases, the net change for the whole path be¬
ing zero. Hence

current links the path).

x
'\
- \
\

j Path a
:>{ \

1
q / Path b

Fig. 9-5. Paths of integration for • dX.

(b) The path of integration links the current circuit. Let us follow the
path b in Fig. 9 o. Starting in, or just above, the plane of the current coil
we have initially ft = 2,r (half the solid angle about a point). As we move
up and away from the coil, ft decreases, becoming zero at the point Q.
Passing around to the negative side of the coil, ft decreases further be¬
coming negative. As we approach the plane of the coil and our starting
point, ft approaches the value -2r. Thus for the whole path the net change
m ft is —4tt. Hence &

J B-dX — Mo/ (if current I links the path),

x

Hn this and following chapters dX will refer to an element of any path whereas
ds will refer to an element of a conducting circuit. UP ’ wneieas
9-7] THE CIRCUITAL FORM OF AMPERE’S LAW 181

Should the path of integration link more than one current circuit, we would
find

j> B • dX = mo X sum of currents linking path. (9-23)

x

This last statement is sometimes called the circuital form of Ampere’s law.
Take any surface S whose periphery is the path of integration X. Then
the sum of the currents linking the path is the total current passing through
S. In terms of the current density J (the current per unit area), this total
current through S is //s J ■ dS. Hence we obtain

(9-24)
x s

Just as Gauss’ law enabled us to obtain quickly the expression for the
electric intensity in problems involving a high degree of symmetry, so we
find that the circuital form of Ampere’s law offers, in some cases, the
easiest method of finding the expression for B.
Illustration 1. Field due to a linear current circuit. Let us return to the
problem of the very long straight circuit discussed as Illustration 1 of
Section 9-5. We wish to find B at the external point P, which is a distance
R from the circuit. In Eq. (9—23) we take for our path X a circle of radius
R passing through P and with center at the wire, as shown in Fig. 9-6.
From symmetry considerations B must have the same magnitude at all
points on this path. From Ampere’s law we know that B cannot be radial,
but must be circular about the wire, so that B and dX are parallel. Hence

and Eq. (9-23) becomes

I

2-kRB = mo 1}

as before. The integration in the

present method is simpler than when
we used Eq. (9-10). Fig. 9-6. f B •dX for linear current.
 182 ampere’s law. steady magnetic fields [chap. 9

Illustration 2. Toroidal solenoid. Consider a doughnut-shaped core

wound uniformly with a wire carrying a current I (Fig. 9-7). Let there
be N turns in all, or N/l turns per unit length, where l is the mean cir¬
cumference around the interior. In applying Eq. (9-23) we take for our
path A this circular path of length l. From symmetry considerations we
again know that B is constant and in the direction of the path, so that

N turns
Since the path A threads the current
N times, we have, from Eq. (9-23),

Bl = hqNI,

(9-25)
B =

As l increases, a toroidal solenoid

becomes more and more similar to
a very long straight solenoid. Hence
Eq. (9-25) also gives B for a long
Fig. 9-7. JIB around toroidal
straight solenoid with NI/l ampere- solenoid.
turns per meter.

9 8 The curl of B. F rom Eq .(9-24) we may easily obtain the expression

for V x B. We employ Stokes’ theorem in vector analysis, which states
that for any vector function B

B • dl = I (V X B) ■ dS,

where S is a surface whose periphery is A. Equation (9-24) thus becomes

(V x B) . dS = //MoJ . dS.

since both surface integrals refer to the same surface, we may equate the
integrands, and obtain

V X B = MoJ. (9-26)

Thus the curl of B is zero only where there is no current density.

I the current density J is due to the motion of a charge density p with a
mean velocity v (all quantities refer to the immediate neighborhood of the
9-8] THE CURL OF B 183

<
point where V x Bis being evaluated), then

J = pv (9-27)
and
V x B = mo pv (9-28)
for a steady flow of charge.
Illustration 1. The magnetic field outside a linear current circuit. We shall
again consider the field around a long straight current-carrying wire. We
saw that outside the wire the field B is directed in circles around the wire.
This might lead one to conclude that the field has a curl, but we shall
show that V x B = 0 outside the wire.
Let the Z-axis be along the wire in the direction of the current, or out
from the page in Fig. 9-8.

Fig. 9-8. Magnetic induction outside linear current.

From Eq. (9-12) we have for the field outside and at a distance r from
the axis of the wire

Since the field is directed circularly about the wire, as shown, at the point
P(x, y) we have

B* V
~ J
R_mol v_
B r - 2tt r2

By MO I
Bvj n o
B ~ 2tt r2

Bz = 0.
 184 ampere’s law. steady magnetic fields [chap. 9

Since Bx and By are independent of z, and Bz is zero, V X B reduces to

k (dBy _ dBx\
V x B
\dx dy /

Putting b IXqI/2tt and ,2 = x2 -f y2, we get

dBy
dx
5 (
bx
dx '\X2 + V2)
)
1 - b
x2 -f- y2
2 bx2
{x2 + y2)2 J

dBx by \i . b 2 by2
* (
dy dy \kX2 + V2) X2 _|_ y2 {pc,2 + y2)2
Thus

The same conclusion follows from Eq. (9-26), since outside the conductor
J is everywhere zero.
We have here a field whose curl is zero at all points outside the con¬
ductor. In this region the field does not rotate or turn about itself (it
possesses no local eddies), but the field does circulate about the conductor.
The line integral of B around a closed path is ny0I, where n is the number
of times that the path passes around the conductor. The reason that the
line integral around any closed path is not zero is that V X B is not zero
everywhere, as we shall show in the next illustration.
Illustration 2. The magnetic field inside a linear current circuit*. Take
the same conductor as in the last illustration and call its radius a. Let
us assume that the current density J is constant throughout, so that

Substituting into Eq. (9-26), we have

V X B = I.
7r a2 (9-29)

Thus within the conductor the field is rotational, i.e., it turns about itself
just as do two adjacent points on a rotating turntable.
Let us see how B depends on r in this case. Applying Eq. (9-23) to a
circular path of radius r (r < a), we have

•It is true that we shall not be computing the field in air or in vacuo, but this
makes no d.fference so long as the conductor is a nonmagnetic one like, say!
9-9] THE MAGNETIC SCALAR POTENTIAL 185

(9-30)

Thus inside the wire B increases with r, while outside B is inversely pro¬
portional to r. For a circular field whose magnitude is proportional to rn,
the curl is zero only when n = —1. Consider the present case where
n = +1. We have

as in Eq. (9-29).
Corresponding to the theorem that circular fields with magnitudes pro¬
portional to rn have zero curl only when n = — 1, there is the theorem
that radial fields with magnitudes proportional to rn have zero divergence
only when n = —2. For the electric field due to and outside of a charged
sphere |E| °c 1/r2 and so V-E = 0; but in general V-E = p/e0 and
V x B = n0J, according to the laws of our physical world.

9-9 The magnetic scalar potential. In a region such as the inside of a

conductor carrying a current we have seen that, since J is not zero, neither
is V x B = 0. In such cases we cannot hope to introduce a scalar potential
function as we did in electrostatics. However, if we have a current-carrying
conductor surrounded by empty space, then outside of the conductor
V x B = 0 and we can write B = —V7m, where Vm is a scalar potential
function.
Referring back to Eq. (9-21), let dl have the components dx, dy, and
dz. Then

dtt = f2 dx + f2 dy + ^ dz = Vfl • d'X.

dx dy dz

If we put this in Eq. (9-21), we get

B • dX = - ™ ■ dl,

or

B = - ^ Vfi = (9-31)
 186 ampere’s law. steady magnetic fields [chap. 9

where

Vm = lT n (9-32)

is the scalar magnetic 'potential.

In charge-free space outside a current-carrying conductor we can com¬
pute Vm from Eq. (9-32) and then B from B = —Wm. Nevertheless,
the magnetic field represented by B is not a conservative one because we
cannot say that the line integral of B around any closed path is zero-
f B • dX is not zero when the path is linked with the current. Consequently,
the line integral of B between any two points depends on the path followed
between the two points. It is thus futile to try to assign absolute values of
Vm to various points in the field, or to associate Vm with any concept of
potential energy. The following illustration deals with the most important
application of the magnetic scalar potential.

Fig. 9-9. Small current circuit.

Illustration. Equivalence of a small current circuit and a magnetic dipole.

Consider a current I present in a small plane circuit of area A. Let the
vector r, from the center of A to some point P(x, y, z), make an angle 9
with the normal to the loop, as in Fig. 9-9. Assume that r is large compared
with the dimensions of A. Then at P the magnetic scalar potential, ac¬
cording to Eq. (9-32), is

Vm = m = £2 / (Acosd\
47T 4-7T \ r2 /

From B = —Wm, we have

B — _ dFm 2/i0IA cos 6

dv 4z7T 7*3

(9-33)
Be — — 1 dFTO _ hqIA sin 6
T SO 7*3
9-9] THE MAGNETIC SCALAR POTENTIAL 187

Comparing these expressions with those in the last example of Section

8-5, we see that, as regards dependence on r and 9, the magnetic field due
to the small current element is similar to the electric field of an electric
dipole.
Suppose that we imagine a fictitious quantity, which we shall call a
magnetic pole, from which a magnetic field diverges if the pole is positive
and toward which it converges if the pole is negative, the field due to
a pole varying according to the inverse square law. Then we may define a
magnetic dipole as two poles of equal magnitude and opposite sign which
are a distance apart that is small compared with r, the distance from the
dipole to P(x, y, z). If we think of such a dipole as situated at the location
of our small current circuit, the axis of the dipole coinciding with the
normal to the surface of the loop, then by adjusting the strength of the
dipole we can arrive at the same field with the dipole as we actually have
from the current circuit.
I

Fig. 9-10. Large current circuit, equivalent to elementary loops.

A large current circuit may be pictured as being made up of a great

number of small elementary loops, as shown in Fig. 9-10. If the same
current I circulates in each small loop in the same direction, the cui rents
will cancel each other on all interior boundaries and so be equivalent to the
current I in the large circuit. Thus the magnetic field of the large circuit
must equal that of all the small loops. Since the field of each elementary
loop is the same as that of a magnetic dipole at its center, with its axis
normal to the surface, the field of the large circuit is the same as that of a
uniform surface layer of magnetic dipoles, the surface being bounded by
the circuit. Such a layer is termed a double layer, since it amounts to a
layer of positive poles adjacent to a parallel layer of negative poles.
While isolated magnetic poles have not been discovered, the magnetic
dipole has reality in that a small bar magnet behaves like one. Although
we prefer to think of the field of such a magnet as having an electrical origin
(since it is due to spinning electrons within the metal) it may be more
convenient for some purposes to adopt the dipole concept. Such a physical
model is justified if it serves a purpose and if we do not believe in its reality
to the extent of letting it mislead us where it does not serve a purpose.
 188 ampere’s law. steady magnetic fields [chap. 9

9-10 The divergence of B. We have adopted the Amperian approach

to electromagnetism and have assumed that all magnetic fields can be
attributed to the motion of electric charges. We may consider B to be the
sum or integral of fields such as those given by Eq. (9-8), namely,

= fioI ds x rx
47r r2

This equation tells us that lines representing the direction of the field dB
are circles about the axis of the current element. Such lines, then, form
closed curves. These lines do not start at (diverge from) any point, nor
do they stop at (converge toward) any point. Thus the field dB is source-
free or solenoidal, that is, its divergence is zero. Furthermore, the sum of
any number of such fields will be solenoidal. We shall now prove mathe¬
matically that V • B = 0.
Let a = (uq/4ltt)(I ds) and b = rj/r2. Then dB = a x b. For any
two vectors such as a and b, we may write

V • (a x b) = b ■ (V x a) — a • (V x b). (9-34)

Here V is an operator involving space differentiation with respect to

the coordinates of the point P where the field is computed. As P moves,
the distance r and the direction of rx may change, but n0, /, and ds remain
constant. Hence
V x a = 0.

The vector function b is in the radial direction with a magnitude inversely

proportional to r2. This is the same type of vector function as E, the electric
field due to a point charge. We saw that V x E was zero and so, in a
similar manner,

V x b = 0.

Thus the right side of Eq. (9-34) is zero, and since a X b = dB, we
have V • dB = 0. Since

V • B = V • j> dB = j> v • dB,

8 $

and V • dB = 0, it follows that

V • B = 0. (9-35)

Equation (9-35) is one of the fundamental equations of field theory and

it will appear later as one of Maxwell’s electromagnetic equations: Com¬
paring this with V-E = P/e0, Eq. (8-13), we see that if we attempt to
9-11] THE VECTOR POTENTIAL 189

attribute magnetic fields to magnetic “poles, ” analogous to electric charges,

then the volume density pm of such poles must be everywhere zero. The
traditional approach interpreted this as meaning that north and south
poles always appear in pairs. The Amperian approach does not need to
introduce the concept of poles.

9-11 The vector potential. It is shown in vector analysis that a vector

function whose divergence is always zero can be expressed as the curl
of another vector function.* Therefore we know that we can write

B = V X A, (9-36)

where A is called the vector potential. We shall now look for the expression
for A at the point P, given the current circuit(s) producing the field B.
First we note that if ds has the components dsx, dsy, and dsz, and
r = (x2 + ij2 + z2)q then

= li (~r x ds) = ^ (ds x r)

ds x rx
^2

ri is a unit vector in the direction of r. As in the previous section, V

operates differentially on r, but not on ds.
We can thus write Eq. (9-10) as

b = s7/v><(t)'
S

Since the differentiation involved in V and the integration around the

*See Vol. I, p. 33.

 190 ampere’s law. steady magnetic fields [chap. 9

circuit are independent of each other, the curl can be taken outside of the
integral sign. We then have

B = V x A,

where

(9-37)

is the vector 'potential which we have been seeking.

The vector function A given by Eq. (9-37) is not the only one whose curl
is equal to B. If to A we add a term A' such that A' = V<f>, where $ is a
scalar function, we see that

VX(A + A') = VXA + VXV$=VXA=B,

since the curl of the gradient is always zero. Therefore we must not
consider A to be completely determined by Eq. (9-37). A somewhat
analogous situation is met in electrostatics, where we found that we could
express the electric intensity as E = — VV, where V is the electrostatic
potential. While the potential due to a point charge is given by
V = g/4xe0r, it is still possible to add a numerical constant, say C, to V,
since V( V T~ C) = —V V = E. This constant is determined by the
boundary conditions for a given problem.
The vector potential may also be expressed in terms of the current
density J, or in terms of the volume density p and velocity v of the moving
free charges producing the magnetic field at P. Suppose that the current
element I ds has a cross section A and a volume dr = Ads. Then

I ds = JA ds = JA ds = J dr,

since ds is in the direction of J. Hence we can write Eq. (9-37) as

A = (9-38)

where the volume integral is taken over all regions containing a current
density. Since J = pv, we also may write

A =
(9-39)

While p and v may vary throughout space, we must make the restriction
that they are independent of time, so that we have a steady field. In the
case of a single charge, the field that it produces at a point P will vary as
9-11] THE VECTOR POTENTIAL 191

the charge moves along; such a field will not be steady. Our expressions
will give the instantaneous field at P to a good approximation if v « c,
where c is the speed of light.
Illustration 1. Vector potential for a straight section of a current circuit.
Referring to Fig. 9-11, in which the Z-axis is taken to coincide with the
part of a current circuit giving a field at P(0, y, 0), we see that here
ds will always be in the Z-direction. Then Ax = Av = 0, ds — dz,
and r = V22 + y2- Consider the contribution to Az at P due to that
segment of wire extending from z — zx to z = z2, namely,

rz2
A = I dz
4tt JZl Vz2 + y2

= ^ [In (z2 + Vz$ + y2) - In (zx + Vzf + y2)]. (9-40)

Fig. 9-11. Straight section of current circuit.

Let us try to find the field at P due to this straight segment of a current
circuit, using B = V x A, Ax = Ay = 0, and Eq. (9-40) for Az. Since
A depends on y only, we have

B = V x A

The field at P is perpendicular to the plane of the figure. Differentiating

Eq. (9-40), we get

y/Vif + y2 y/V z2 + y 2
dAz _ no I • (9-41)
BX
dy 4tt 22 + Vz2 + y2 zl + Vz2 + y2
 192 ampere's law. steady magnetic fields [chap. 9

It is interesting to see what happens to this expression if we let z2 —» oo

and zi —» — oo, as in the case of a very long straight current circuit. As
z2 oo, the first term in the square bracket approaches zero. As
Z\ —» — oo, the second term approaches the value —2/y and we get an
expression for Bx in agreement with Eq. (9-12), with R = y. The evalua¬
tion of this second term is a bit tricky; we must remember to regard all
square roots as positive and z\ as negative. We then have

Lim _ZJ/_
Zl~>— oo (zi + Vzf -f y2)Vz\ + y2

= Lim
zi- zx(zf + y2)1'2 + z\ + y2

= Lim _zfitt_ —y
—oo
z1V^1[r+ Mv2/zj) + • • •] y2 —\y2 + y2

2
y

Illustration 2. Vector 'potential and magnetic field for a (slowly) moving

charge. For a charge q moving with the constant velocity v (see Fig.
9-12) we may put qv for I ds or pv dr, provided that the charge is moving
with a speed considerably less than that of light (so that relativity effects
may be neglected). Let us again find B from the vector potential A. Here

^ __ Mo gy
4ltv r (9-42)

We put B = V x A, remembering that V operates differentially only

on the 1/r factor, since q and v are independent of the coordinates. We
then get

4tt \ r)

We may follow the same procedure

as in Section 9-11 and obtain

TTVV__VXrl
v x r ~ r2 ’ Fig- 9-12. Slowly moving charge.

where rx is a unit vector from the charge to the point P where we are getting
the field. Thus
Moq v x rx
4tt f2 (9-43)
 PROBLEMS 193

gives the magnetic induction due to a slowly moving charge. As a check we

may observe that the same result can be obtained directly from Eq. (9-10),
putting qv for I ds.

Problems

1. Calculate the force on each of two 6. If the magnetic induction is

current elements, both 1 cm long and 10_2w/m2 in the X-direction, find the
carrying 10 amp, if one is at the origin force on an electron whose velocity is
and directed along the X-axis, while 107 m/sec in (a) the X-direction, (b)
the other is (a) at x = 0, y = 0.1, the F-direction, (c) the F-direction, (d)
z = 0 and in the X-direction, (b) at in the XF-plane at 45° to the X-axis,
x = 0, y = 0.1, z = 0 and in the in¬ (e) a direction making equal angles
direction, (c) at x = 0.1, y = 0.1, with the three axes.
^ = 0 and in the X-direction, (d) at 7. Using Eq. (9-10), find an expres¬
x — 0.1, y = 0, 3 = 0 and in the X- sion for the magnetic induction at a
direction, (e) at x = 0.1, y — 0.1, point P on the axis of a single circular
3 = 0.1 and in the F-direction. (Co¬ turn of wire carrying a current I. Let
ordinates are in meters.) R be the radius of the circular turn and
2. Find dB, due to the current ele¬ let x be the distance from its center to
ment at the origin, at each of the four the point P.
points in problem 1. 8. Find the expression for B on the
3. Find B at each of the four points axis of a straight solenoid of radius R,
in problem 1 if there is a current of 10 length l, and N turns, carrying a cur¬
amp along the X-axis. rent I, at a point a distance x from
4. Show that the magnetic force on one end and l — x from the other
a moving charge q' due to another end.
slowly moving charge q is 9. Find the expression for B at the
center of a square current circuit of side
„ mo f v' X (v X ri) 2a and N turns carrying a current I.
F = — q q-s->
4-7T r2 10. A beam of electrons passing at
where v' and v are the respective right angles to a magnetic field whose
velocities, r is the separation, and ri a flux density B equals 2 X 10_3w/m2
unit vector from q toward q'. is bent in a circular path of 10 cm
5. A wire whose cross section is radius, (a) Find the speed of the elec¬
1 mm2 carries a current of 10-6 amp. trons. (b) If the electrons are originally
Assume that the wire contains 8.0 X moving in the X-direction and B is in
1028 free electrons per cubic meter, the F-direction, find E, the electric
(a) What is the average drift velocity field whose force on the electrons will
of the free electrons in the wire? (b) balance that of the magnetic field.
What length of the current-carrying 11. Prove the cyclotron principle
wire would constitute a current ele¬ that, for a particle of constant mass
ment equivalent to an electron moving and charge, the time taken to complete
with a speed of 107 m/sec? (For elec¬ a circular path of radius R at right
trons e = —1.6 X 10-19 coul.) angles to a magnetic field of flux
194 ampere’s law. steady magnetic fields [chap. 9

density B is independent of the speed uniform magnetic field of induction B,

of the particle. then the torque on the conductor is
12. Suppose that in the fusion of ttR2IB.
hydrogen to form helium a temperature 16. Suppose that for the very long
of 107 degrees Kelvin is attained, (a) coaxial line shown in Fig. 9-13, there
Find the rms speed of a deuteron (H2) is a current I out of the page along the
of mass 3.3 X 10~27 kg at this tempera¬ inner cylindrical conductor and into
ture (see Chapter 4). (b) Compute the the page through the outer cylindrical
strength of the magnetic induction B shell. Assume each conductor to have
that would bend such particles in a a uniform current density. From Eq.
circular path of 1 m radius. (9-24), find B for (a) r < a, (b)
13. A magnetron consists of a a < r < b, (c) b < r < c, (d) r > c.
straight filament of radius a surrounded
by a coaxial cylindrical plate of radius
b. The tube is placed in a uniform
magnetic field B, which is parallel to
the axis of the tube. Let the potential
of the plate be V volts above that of
the filament. Electrons (charge e and
mass to) from the filament are acceler¬
ated toward the plate and, at the same
time, are deflected from a radial path
because of the force of the magnetic
field. As B is increased, a critical point
is reached such that electrons emitted
with zero initial velocity spiral about
the filament and just fail to reach the
plate. If B is further increased, the
plate current falls rapidly. Show that
the critical value of B is given by
17. Find the current density J at a
^2 _ 8b2 Vm point P(x, y, z) in a field given by
(b2 — a2)2e Bx = k/y, By = k/x, where k is a
constant with the proper mksc dimen¬
14. A galvanometer, fitted with sus¬ sions.
pension, mirror, telescope and scale, 18. Compare the torque per unit
has a coil with 50 turns and an area of electric field (L/E) on an electric di¬
6 cm2. The coil is suspended by a pole with the torque per unit mag¬
fiber whose stiffness constant k = 3 X netic field (L/B) on a current circuit.
10“8 n-m/radian and the coil lies Remember that electric moment is de¬
parallel to magnetic field whose B = fined as qd, where d is the charge
0.1 w/m2. Find the current sensitivity separation. What might we call the
in amp/mm deflection on a scale 1 m magnetic moment of a small current
away. element?
15. A circular turn of wire of radius 19. Find the scalar potential Vm for
R carries a current I. Show that when a point on the axis of a circular circuit
the plane of the coil lies parallel to a of radius R, at a distance x from its
 PROBLEMS 195

center, then find B = —VFm. Check distant R\ from one wire and R2 from
against the answer to problem 7. the other.
20. Show that at a point outside a 22. Find the vector potential A, and
current circuit the vector potential A from it the components of B, at the
satisfies Laplace’s equation. point P(x, y, z) for a small square cir¬
21. Two long straight parallel wires cuit of side 2a and current I, taking the
carry a current I in opposite directions. circuit to be in the XF-plane, with
Find the vector potential A at a point center at the origin, and a <3' r.
 CHAPTER 10

FARADAY’S LAW. ELECTROMAGNETIC INDUCTION

In Chapter 9 we assumed that we were dealing with magnetic ,fields which

did not change with time, or whose time dependence could be ignored.
Such steady magnetic fields are analogous to the fields of static charges
in electricity. Hence Chapter 9 might have been entitled “magnetostatics. ”
In the present chapter we shall discuss the effects of changing magnetic
fields.
From the work of Oersted, Ampere, and others, Faraday knew that
electric currents could produce magnetic fields. He then began to look for
the converse effect, i.e., currents produced by magnetic fields. While he
found that a steady magnetic field passing through a stationary coil of wire
connected to a galvanometer produced no current, he did observe a
temporary current in the coil when the magnetic field linking it was
changed. We now call this effect electromagnetic induction, and currents
produced in this way are termed induced currents.
Faraday proceeded to investigate all the possible ways of producing in¬
duced currents and then he summarized his results in the law that bears his
name. There are several different methods of inducing a current in a
conducting circuit. First, the circuit may be rigid but moving in the mag¬
netic field in such a way that the surface integral of B over a surface
enclosed by the conductor (this is called the total flux through the circuit)
is changing with time. Second, the circuit may be deformable and its area
may be changing in such a way that the total flux through the circuit is
varying. Third, the circuit may have constant area but include a changing
path in a moving conductor, whose motion is in a direction other than
that of B. Fourth, the circuit may be stationary, but the magnetic induc¬
tion B directed through a surface enclosed by the circuit may vary with the
time. Combinations of these methods are also possible. In general, we may
classify the above processes into (a) those involving motion of all or part of
the circuit and (b) those involving a time rate of change of B. We shall find
that in (a) the effects may be explained in terms of Ampere’s law and the
Lorentz force, Eq. (9-11), but that in (b) we are encountering something
quite apart from either Ampere’s or Coulomb’s law.

10-1 Definition of electromotive force. Let us call the electric fields due
to stationary charges electrostatic fields. We saw in Chapter 8 that if we
imagine an agent to take a small test charge around any closed path X in
such a field, then the total work done by the field on the charge will be zero,
196
 I
10-1] DEFINITION OF ELECTROMOTIVE FORCE 197

since

/
x
Es • dT. = 0, (10-1)

where Es is the electric intensity (force per charge) in such an electrostatic

field. If a conductor is placed in an electrostatic field, the charges in the
conductor that are free to move will redistribute themselves in such a way
that at all points within the conductor the electric field will be zero.
At ordinary temperatures the free charges in a conductor encounter re¬
sistance when they move through the conductor, which means that when a
steady current is maintained electrical energy is continually being dissi¬
pated; a rise in the temperature of the conductor reminds us that this
energy has become part of the internal energy of the conductor. This
irreversible process is called the Joule heating effect, and it requires the
presence of a device whereby energy may be taken from some outside
source and continually converted into electrical energy. Such a device
must produce an electric field E' whose line integral around a closed
circuit is not zero. This must be so in order that work may be done by
the charges against resistance forces. The field E' will be called a non-
electrostatic field; a source of such a field is said to be the seat of an electro¬
motive force. The electromotive force (emf) around a closed circuit s will be
designated as 8, and is defined by

(10-2)

This represents the work per charge done by the field; 8 is expressed in
volts.
In general, the electric field E may be partly electrostatic and partly not.
If
E = Es + E',

we see from Eq. (10-1) that <f Es • ds = 0 and

8 = f E • ds (10-3)

is equivalent to the definition given in Eq. (10-2).

Let us consider some sources of emf. The outside source of energy in a
photocell is light, in a thermocouple it is heat, and in a battery it is chemical
energy. In a generator, work is performed when a conductor is moved
across a magnetic field and this work is converted into electrical energy.
The expression for the resulting emf may be derived from Ampere’s law.
This emf is called a motional electromotive force.
198 faraday’s law. electromagnetic induction [chap. 10

10-2 Motional electromotive force. Let us start with a simple example.

In Fig. 10-1 let ab be a movable conductor of length l parallel to what we
shall take to be the Z-axis. To eliminate the effect of weight, picture the
conductor riding on two frictionless horizontal tracks extending in the F-
direction. Assume that a uniform magnetic field, of induction B, extends in
the Z-direction (out from the paper in Fig. 10-1). Now let us suppose that
the conductor is moved at a speed v in the positive F-direction. The free
charge q in the wire, according to Ampere’s law, will experience a force F
given by Eq. (9-11), which tells us that the force is

F = qvB (10-4)

in the plus Z-direction. (Primes have been dropped here.) Imagine a

reference system moving with the conductor, so that in this system the
free charge in the wire is a static charge experiencing the force F. A mag¬
netic force does not act on a static charge, so an observer in this moving
reference system would say that an electric force, represented by an electric
field, must be present. Whether this force is termed electric or magnetic
depends on the choice of reference system to which it is referred. The im¬
portant thing is that there is a force acting on the charge q and that this
force can be used to drive a current around the circuit. The contact points
at a and b will correspond electrically to the terminals of a battery. From
the definition of electric intensity, we have for the effective electric field

E = — = vB in the Z-direction.
q

This effective field exists only in the moving arm ab of length l. From Eq.
(10-3), we have for the emf around the circuit

g = f E • ds = vlB, (10-5)

Since vl is the area swept out per unit time, which we shall call dA/dt, we
may write

(10-6)

Note that if the entire circuit were moving in the F-direction with the
speed v, the galvanometer arm cd would experience an effective E from c
toward d just equal to that in ab and, as a result, the S for the circuit
would be zero. The galvanometer G would then show no deflection
Therefore what we detect in the case where only the arm ab moves is the
relative motion of ab with respect to the rest of the circuit.
Figure 10-2 shows the same circuit as in Fig. 10-1; as before, ab is the
moving arm. The induced current I, which is in the counterclockwise
10-2] MOTIONAL ELECTROMOTIVE FORCE 199
*

Fig. 10-2. Electric generator.

sense, is also indicated. Since 8 is the work per unit charge done by the
field as charge flows around the circuit, the time rate of doing work by the
field, or the power P, is

P = 8 % = 8/. (10-7)
at

Since this work is expended only when the conductor is moving across the
magnetic field, the source of the energy must be the agent moving the
wire, as we shall now show.
From Chapter 9, we know that a conductor of length l carrying a cur¬
rent I across a uniform magnetic field B experiences a force which, to
distinguish it from the force F in Eq. (10-4), we shall call F'. From Eq.
(9-16), we have

F' = I ds x B = IIB
J a

in the direction of the vector ds X B, which in Fig. 10-2 is down, or in the

negative F-direction. Since F' is opposite to the direction of motion, mov¬
ing the wire with the speed v involves expending power by the moving
agent (say one’s hand) at the rate

P = F'v = IlBv. (10-8)

Equating the two expressions for P, Eqs. (10-7) and (10-8), we find again

which shows that the derivation of this equation based on Ampere’s law
and the concept of an electric field existing in a conductor that is moving
across a magnetic field are consistent with the conservation of energy
principle. This, then, is the theory of the electric generator, through which
mechanical energy is converted into electrical energy.
200 faraday’s law. electromagnetic induction [chap. 10

If we choose to move the conductor in the opposite direction, as in Fig.

10-3, the directions of the induced electric field and the emf S will be re¬
versed. To maintain the current 7 in the same sense as before, we must in¬
sert into the circuit a battery or generator whose emf 8' is opposite to, and
greater than, the 8 due to the moving conductor. In this case the force F'
will be the same as before (Fig. 10-2), because ds and B are unchanged, but
the conductor is now moving in the direction of F'. When an object moves
in the direction of an applied force it receives mechanical energy. The
source of this energy must be in the electrical circuit, for the current I
is moving against the induced emf. This energy source is the battery or
generator inserted elsewhere in the circuit, of which the moving conductor
is a part. This is the principle of the electric motor, through which electrical
energy is converted into mechanical energy.
The simple problem that we have been considering serves to illustrate
the principles involved. We may now generalize for the case where v
is the velocity of any element ds of a circuit, and v, ds, and B are not
necessarily mutually perpendicular. In general, we have, in place of Eq.
(10-4),

F = qv x B, (10-9)

so that

E = v x B (10-10)

From the definition of emf,

8 = J'E • ds,

and we now have

8 = fv x B • ds = <£ ds • v x B.

(10-11) Fig. 10-3. Electric motor.

In a triple scalar product the dot and cross may be interchanged, or

ds • v x B = ds X v • B. Let u§ write

ds'
v =
dt

where ds' is the displacement of the element of length ds in the time dt.
We then have

ds x ds'
-I dt
B. (10-12)
10-3] MAGNETIC FLUX 201

Since the area of a parallelogram is the cross product of the base times the
side, the surface dS' swept out by the element ds in the time dt is dS' =
ds x ds'.* Therefore we finally get

8 = (10-13)

as our general expression for the motional emf.

10-3 Magnetic flux. We found it convenient to picture an electric

field by drawing lines of force to show the direction of the field, and to
represent the strength of the field by the concentration of these lines. We
may do the same for a magnetic field. Let us arbitrarily take B lines of
force per unit area through a surface whose normal is in the direction of B.
(Because of the small numerical value of B when it is expressed in mksc
units, lines of fractional value must be imagined.) In any event, we define
the total magnetic flux crossing a surface S as

4> = B • dS. (10-14)

s

The weber is evidently our unit of flux.

Let us recall that we may change the flux through a closed surface essen¬
tially in two ways, namely, (a) by moving the surface, and (b) by varying
the field through a fixed surface. In the first case we may (1) move the
whole of the circuit to a position where <t> has a different value, or (2)
alter the size or shape of the circuit so as to vary the flux through it, or (3)
both move and alter the circuit.
In case (a) above whatever we do involves motion of the periphery of
the circuit, and so we may apply Eq. (10-13), namely,

8 = (10-13)

From the definition of <h in Eq. (10-14), the flux threading a circuit is given

<t> = [[b • dS,

where S is any surface bounded by the circuit. Here we must consider dS

*In a case such as the Faraday disk generator (problem 3 at end of chapter),
where the circuit has a moving arm and yet a fixed area, dS' does not refer to
area added to, or subtracted from, that of the circuit.
202 faraday’s law. electromagnetic induction [chap. 10

as an element of such a surface, If the periphery is changing but B is steady

(time-independent),

(10-15)
dt

where dS/dt represents the time rate of change of the surface element dS.
However, dS'/dt is the area swept out per unit time by an element ds of
the periphery. To see how dS/dt and dS'/dt are related, let us consider the
situation shown in Fig. 10-4. According to Section 10-1, 8 is counterclock¬
wise. Let us take this as the 'positive sense for integration around the circuit.
It is convenient to relate this positive sense around the circuit to a positive
direction through the circuit. Since we are using only right-handed axes
and rules, we shall take the positive direction for S and dS to be that of
the advance of a right-handed screw rotating in the positive sense. If we
apply this rule to Fig. 10-4, we see that the positive direction for S is that
of B, out from the page. Motion of the moving element ds is such that

dS'
—rr = ds X v
dt

is also in the positive direction.

However, dS/dt is negative, since
the positive surface is being reduced
in magnitude. We then have

dS
dt dt

If we combine Eqs. (10-13) and Flux out through the

circuit is decreasing
(10-15), we get
d4> Fig. 10-4. Direction of induced emf.
(10-16)
dt

for the motional emf. The induced emf is equal to minus the time rate of
change of magnetic flux through the circuit. ,
The example just discussed covers most cases involving motional emf.
The Faraday disk generator, referred to in the footnote at the end of the
last section (see also problem 3 of this chapter), belongs in a little different
category, one involving accelerated (rotational) motion of a conductor;
foi such cases Eq. (10—16) is still valid provided d^/dt is interpreted as the
time late of cutting flux. If accelerated motion is not involved, we must
speak only of the motion of one part of the circuit relative to another
part and not of motion relative to the magnetic field, so that the expression
“cutting flux” has no meaning for unaccelerated motion.
10-4] Faraday’s law 203

10-4 Faraday’s law. This law is simply a generalization of Eq. (10-16)

to cover both (a) moving circuits and (b) fields varying with time. As has
been stated, only the application to time-dependent fields is a law of
Nature quite apart from other laws such as Ampere’s.
Faraday’s law states that the emf induced in a circuit equals the negative
of the time rate of change of magnetic flux through the circuit, or

whatever the reason for the flux change.

Using Eq. (10-14), we may write Faraday’s law as

where S refers to any surface bounded by the circuit.

Case (a). S varies with time, but B does not. This case has already
been discussed. We shall give a simple illustration, one that demonstrates
a convenient method for measuring B.
Illustration. Search coil; measurement of B. Suppose that a coil of N
turns, each of area A, lies with its normal parallel to a uniform magnetic
field of induction B. The coil is then rotated 90° (so that B is perpendicular
to the normal) in a time At. Here the direction but not the magnitude of
S varies. Calling the total flux change A<f>, the average induced emf is

__M> _ _ NAB
8 ~ ~ At ~ At

Note that NA is the effective area of the whole circuit comprised by the
coil. Suppose that N = 200, A = 10 cm2 - 1(T3 m2, B = 0.2 w/m2,
and At = 0.1 sec. Then A4> = 4 X 10~2 weber and the average induced
emf would be 0.4 volt. This numerical example brings out an interesting
point, namely, that no new quantity is introduced and defined by Faraday’s
law. While constants of proportionality Cx and C2 were necessary when
stating Ampere’s and Coulomb’s laws, no such constant appears in Fara¬
day’s law. This is because Coulomb’s law was our introduction to a new
branch of physics (electricity) and Ampere’s law served as our starting
point in magnetism, while Faraday’s law states an interconnection between
magnetism and electricity.
In this experiment we usually measure the total charge q passing a point
in the circuit while the coil is turned, rather than S. This may be done
by connecting the coil to a calibrated ballistic galvanometer.* A ballistic

*See Vol. I, pp. 92-93; also Page, Introduction to Theoretical Physics.

 204 faraday’s law. electromagnetic induction [chap. 10

galvanometer is a moving-coil type of galvanometer with a relatively

long period, obtained by increasing the moment of inertia of the coil.
If the period is long compared with the time At during which the induced
emf S exists, the deflection of the galvanometer is proportional to q.
Calibration involves measuring the deflection due to the passage of a
known charge, such as that on a capacitor of known capacitance with a
known potential difference between its plates (see Section 12-8). Pre¬
cautions must be taken to ensure the same amount of damping during
calibration and use.
The instantaneous emf is

Now assume that the current in the circuit is proportional to the emf
(Ohm’s law), or

7 = I’
where R is a constant called the resistance of the circuit. We shall see (Chap¬
ter 12) that this assumption holds for metallic circuits. Since I = dq/dt,
we have
T. dq d<f>
K -r- = -— •
dt dt

We may integrate each side of this equation with respect to the time, for
the interval during which the coil is turned. Then

R/t * = Rjd<l = - /f * = - A*.

Calling the total induced charge q, we get

_ [j NAB
q~ Jdl = = --R-- (10-20)

If q, N, and R can be measured, B can be determined. According to Fara¬

day’s law, this same equation will apply if the coil is jerked out of the field
into a position where no flux passes through it—a more convenient pro¬
cedure in some cases than rotating the coil.
Case (b). B varies, with time, but S does not. This case involves (1)
moving a magnet relative to the circuit, (2) moving a second current circuit
relative to the circuit in question, and (3) varying the current in a second
circuit m whose field the given circuit is located. For all these processes
Faraday found his law to hold true.
10-4] Faraday’s law 205
i

Equation (10-18) now becomes, with S constant,

(10-21)
s

Note that we write dB/dt rather than dB/dt because B is a function of the
coordinates as well as the time, and we want here only its variation with
time at a fixed point, not its variation due to moving the circuit to where
B has a different value.* We shall henceforth use the dot notation for par¬
tial time derivatives, writing Eq. (10-21) as

8 = — B-dS. (10-22)

To Faraday the emf given by Eq. (10-22) and the motional emf of Eq.
(10-6) were both illustrations of the same physical law, namely, that an
emf is present in a circuit whenever the magnetic flux linking the circuit is
changing. This approach has the advantage of simplification, one of the
aims of theoretical physics, but at the same time it obscures an important
point. In the case of a steady magnetic field (B = 0), an emf can exist
only if a moving conductor is present. If the conductor is removed, does
the emf exist? The answer is no; the nonelectrostatic field in the moving
conductor can be said to exist only when there are moving charges present
on which the magnetic field can exert a Lorentz force. Motional emf was
attributed to an effective electric field, but the existence of such a field
depends on the relative motion between the conductor and the observer.
Such is not the case for the nonelectrostatic field that gives rise to the emf
in Eq. (10-22), for a changing magnetic field gives rise to a nonconservative
electric field, whether or not a conductor is present. This important point
was first recognized by Maxwell. It is here that Faraday s law makes its
contribution to the theory of electromagnetic fields.
So far in this chapter we have con¬
sidered 8 to be the line integral of
the electric field around a closed
conducting path, or circuit, s. If
the emf is due to a changing mag¬
netic field, as in Eq. (10-22), we
may replace the path s around the
circuit by a path X fixed in space, js increasing
as in Fig. lO-o. j?lG ^q_5 gmf arounq fixed path.

*See Yol. I, p. 240.

206 faraday’s law. electromagnetic induction [chap. 10

We take the following as a general definition of emf:

S >*
/
x
E • dl (10-23)

is the emf around any path X. We now combine Eqs. (10-22) and (10-23)
to obtain

(10-24)
X s

where S is a surface whose periphery is X. This is the integral statement of

Faraday’s law given in Chapter 7. It applies to any closed path, whether
or not a conducting wire coincides with it. While this equation holds only
for a fixed path, in the theory of electromagnetic fields this restriction is no
limitation because we seek only expressions for the field, irrespective of its
effects on charges, and a path such as X is a purely mathematical one.
Illustration 1. The betatron. This is an interesting application of Fara¬
day’s law in atomic physics. We have seen that an electron may be ac¬
celerated in an electrostatic field, while a steady magnetic field may be used
to deflect, but not to speed up, a beam of charged particles. We have just
seen that a changing magnetic field produces an electric field, so here we
have another method of speeding up charged particles. In the betatron
this principle is used to accelerate electrons to very high energies.
Electrons from a hot filament are injected into an evacuated doughnut¬
shaped tube of mean radius R. An alternating magnetic field is applied
parallel to the axis of the tube. This field serves two purposes: (1) an
emf tangential to the tube is produced by the changing magnetic flux, thus
speeding up the electrons, and (2) a radial force (see Ampere’s law) acts
on the electrons as they move across the magnetic field, a force which may
be used to keep the electrons in the circular path of the tube. To accom¬
plish this the magnetic induction must be a certain function of r, the dis¬
tance from the axis, so the field must be produced by placing the tube
between specially shaped poles of an electromagnet. Of course, the elec¬
trons can be kept in the tube and accelerated in one direction during only
one-quarter of each cycle of the alternating field.
From Faraday’s law, we have

where e is the electronic charge. This is the work done on an electron

during one revolution in the tube. This work must equal the tangential
force F times the circumference 2ir R, or
10-4] FARADAY S LAW 207

eS __ e d&
F =
2irR 2tvR dt

From Newton’s second law,

F = it(mv)-
so that
d , s e d<t>
- (m«) _ ^ ^

d(mv) = d<f>.
2irR

The condition that the electrons keep in the circular orbit of radius R
is (Eq. 9-13)
mv
eB =
~R

Differentiating this equation, we have

d(mv)
e dB = d(mv) = eR dB.
R

Equating the expressions for d(mv) yields

= eR dB,
ZttR

which, when integrated between the limits zero and <t>, and zero and BR,
respectively, gives us, for the flux out to r = R,

<!> = 2itR2Br.

Note that if the field had everywhere the same value BR that it has at
r — the flux through the orbit would be only irR2BK. Hence the field
must be some function B(r) of the distance r from the axis of the tube.
We have
rR

$= B(r) • 2irr dr = 2ttR2Br,

Jo

which may be satisfied by having

B(r) = 7 Br-

Since the field must decrease with increasing r, an electromagnet with

tapered poles is employed.
 208 faraday’s law. electromagnetic induction [chap. 10

Illustration 2. Diamagnetism. The atoms of a substance contain electrons

which move about in orbits inside their atoms. Application of a changing
magnetic field to the substance, with the consequent flux change through
the electron orbits, will result, according to Faraday’s law, in an induced
emf in each such orbit. Because of the negative sign of Eq. (10-17), this
induced emf will oppose currents in the positive sense and aid currents in
the negative sense; in either case the effect will be that the electron currents
will change in such a way as to produce less and less magnetic flux of their
own in the direction in which the applied magnetic field is increasing, or
more and more flux in the opposite direction. Assuming that such orbits
possess perfect conductivity, this effect will persist after the applied field
has reached a constant value. The result is that an attempt to “magnetize”
the substance in a given direction results in its becoming “magnetized” in
a sense opposite to the applied field. This is called diamagnetism and it is a
universal property of all substances, although in some it is masked by
other magnetic properties.

10-5 Lenz’s law. This is a theorem concerning the direction of the

induced emf and it is a consequence of the minus sign in Faraday’s law.
Lenz’s law states that the direction of the induced emf is such as to oppose the
cause producing it. The justification for the statement may be seen from the
example of the moving conductor in Section 10—1 and the discussion of
diamagnetism in the last section. Were this law not true, Nature would
be in a state of unstable equilibrium—a slight displacement of a current-
carrying conductor in a magnetic field, in the direction of the force on the
conductor, would result in an increase in the current and an increased
force.
Statistically speaking, a state of unstable equilibrium is much less likely
to occur than one of stable equilibrium. Hence we would expect to find,
and actually do, that an undisturbed physical system is usually in a state
of stable equilibrium, which means that when such a system is disturbed it
will exhibit a tendency to return to its equilibrium state. This general
principle, of which Lenz’s law is one example, was stated by Le Chatelier
as follows:

When a system in equilibrium is disturbed, the equilibrium is displaced in

the direction which tends to undo the effects of the disturbance.

This principle of Le Chatelier is sometimes called the “law of the cussed¬

ness of Nature. ” It is, however, fortunate for us that it exists, or we would
live in a chaotic world!

10-6 General expression for the curl of E.

Let us return to the integral
statement of Faraday’s law, which was
10-7] RELATION BETWEEN E AND THE VECTOR POTENTIAL 209

E • dl = - 11 B • dS. (10-24)
X s

When we apply Stokes’ theorem in vector analysis to the left side of this
equation, we get

(V X E) • dS = — JJ B ■ dS.
s s

Since both integrals refer to the same arbitrary surface S, we may equate
the integrands, obtaining
V x E = —B. (10-25)

This is the differential, or point statement, of Faraday’s law. We shall see

that it is perfectly general and forms one of Maxwell’s electromagnetic
equations. It reduces to Eq. (8-14) in the special case of a steady (or static)
field.

10-7 Relation between E and the vector potential. We saw that we may
put
B = V x A,

where A is the vector potential function in field theory. Then

d
B - (V x A) = V x A, (10-26)
dt

since we may interchange space and time derivatives. Substituting this in

Eq. (10-25), we have
^ ^ Jg _ _^ ^ ^

or
V x (E + A) = 0. (10-27)

Since the curl of (E + A) is zero, we may express E + A as the gradient

of a scalar function T and write

E + A = —V*. (10-28)

For steady fields A — 0 and E = —VF. We saw in Chapter 8 that

we could put E = -VF, where V is the electrostatic potential function.
Hence the ’F in Eq. (10-28) must reduce to this function V for an electro¬
static field. In general,
E = -VT - A, (10-29)
210 faraday’s law. electromagnetic induction [chap. 10

which suggests that when the field is time-dependent, we may think of

—V'k as giving the contribution to E due to the Coulomb field and of —A
as giving the contribution due to electromagnetic induction. Since A may
be expressed in terms of the current elements present (see Eq. 9-37), A
will contain terms involving the motion of such current elements or the
time rate of change of such currents, both of which factors, we have seen,
contribute to an induced field.

10-8 Mutual inductance and self-inductance. Suppose that there is

a steady current id in what we shall call circuit 1 (see Fig. 10-6).
At any point P in the neighborhood of this circuit the resulting value of
B will be proportional to id. Now let ix change with time, but so slowly
that the value of B at P at any moment is very nearly the same as it would
be if Ix had a constant value equal to its instantaneous value; we shall
find that time variations in a field are propagated with the speed of light,
which means that in the time it would take light waves to go from the cir¬
cuit to P, the current must not change significantly. Then with B °c id,
we will have B * dli/dt at all points.

Fig. 10-6. Mutual inductance.

Let P be a point on the surface S, whose periphery is X. From Eq.

(10-22) the emf S around X must be proportional to dl\/dt. Let us call
the proportionality constant — M; we then have

8
M = — (10-30)
dli/dt ’

where M is called the coefficient of mutual inductance between the closed

path X and circuit 1.
If we combine Eqs. (10-17) and (10-30), we obtain

M *—■ = — •
dt dt

Integrating with respect to the time between the limits 0 and and 0
and <F, respectively, we get

MI i = (10-31)
10-8] MUTUAL INDUCTANCE AND SELF-INDUCTANCE 211

Therefore the mutual inductance between the path X and circuit 1 is also
the magnetic flux through the path per unit current in circuit 1.
Usually the path X is taken to coincide with another conducting circuit,
which we shall call circuit 2; then the S in Eq. (10-30) is the induced
emf in circuit 2.
As a special case, we may let the path X coincide with circuit 1, in which
event <t> is the flux per unit current through circuit 1 due to its own current
and M is then called the coefficient of self-inductance, usually designated
by the symbol L. In this case S is the emf induced in circuit 1 as its own
current changes. Calling the induced emf Si, we have

Si
L (10-32)
dli/dt

If the current in circuit 1 is counterclockwise (see Fig. 10-6), the direction

of B within the circuit, according to Ampere’s law (see the right-hand fist
rule), will be out from the paper. This is also the direction of advance of a
right-handed screw turning with the current, so $ is positive. Then if
dli/dt is positive, the flux will increase in the positive sense and, according
to Faraday’s law, the induced emf will be in the negative sense, opposite
to I\ and dli/dt. Hence L is positive. (This result also follows from Eq.
10-31.) The induced emf Si is always opposite to the sense of dli/dt,
no matter which direction h has, and hence it is called the back emf.
In the case of two separate circuits, the direction of S depends on the
relative positions of the circuits.
The unit for M or L follows from its definition. From Eq. (10-30)
or Eq. (10-32), we see that in the mksc system we may call the unit the
volt-sec/amp, while according to Eq. (10-31), we may take it to be the
weber/amp. These two units are equivalent (see definitions of the volt
and weber). For convenience, a simpler name for the unit of inductance
has been chosen, namely, the henry, defined as

1 henry = 1 w/amp = 1 volt-sec/amp.

The name of this unit honors an American physicist, Joseph Henry, who,
while a schoolteacher at Albany Academy, discovered mutual inductance
independently of Faraday.
While M is usually calculated from Eq. (10-31) (in which $ is found from
Ampere’s law), we shall now derive a general expression which will show
that M depends only on the positions of two circuits and that it does not
matter which we call circuit 1 and which circuit 2. Let us recall that
B = V X A (see Eq. 9—37), and substitute for B in Eq. (10—22).
212 faraday’s law. electromagnetic induction [chap. 10

We find that

8 = — V X A • dS.

Using Stokes’ theorem, this becomes

8 A • dX, (10-33)

where
Mo m 11 ds
A =
4ir J r

and
Mo d/1 ds
(10-34)
47r df r

since circuit 1 is considered fixed. When we substitute Eq. (10-34) into

Eq. (10-33), we get

g _ _ Mo d/i i ds • d7> 1
4?r di J J r
X s

Dividing by d/i/df, we have, from the definition of M in Eq. (10-30),

^ = (10-35)

wheie, as in Fig. 10 7, ds and dX are two elements of length and r is the

distance between them. Because of the symmetry of the formula, we see
that it really makes no difference which circuit we call 1 (or s) and which
we call 2 (or X). The mutual inductance is the same, no matter which we
regard as the source of the changing current.

Fig. 10 7. Calculation of mutual inductance.

10-8] MUTUAL INDUCTANCE AND SELF-INDUCTANCE 213

For self-inductance, X and s are the same. While Eq. (10-35) then
appears to be an improper integral (since r = 0 when ds and d\ coincide),
a finite value for L results, except for a circuit of infinitely thin wire (a
physical impossibility).
Illustration. Toroidal solenoid with two windings. Consider the thin
air-core ring solenoid shown in Fig. 10-8. Let R be its mean radius, Ni
the number of turns in one winding, N2 the number of turns in the other
winding, and A the cross-sectional area within either winding. We assume
that each winding is uniform and thin (compared with R).

Fig. 10-8. Toroidal solenoid.

With a current 11 in the first winding, we will have a magnetic induction

within given by Eq. (9-25), namely,

d _ MqNiIi vqN\Ii
2tvR ~ l ’

where l is the mean circumference. As a result, the flux per turn of the
second winding will be BA, and the total flux threading, or passing through,
circuit 2, which comprises the second winding, will be

$ - N2BA = ^ N1N2AI1.

From Eq. (10-31), we have

M = ^ NxN2A. (10-36)

To find the self-inductance of one winding, say the first, we simply replace
N2 hy N i and obtain
Li = ff NjA. (10-37)

Similarly,
L2 = N\A. (10-38)
214 faraday’s law. electromagnetic induction [chap. 10

We may write these expressions in terms of the turns per unit length of
circumference for each winding. If we put

Ni N2
ni n2 =
l
we get
M Li = n0nlA. (10-39)
H0nln2-A, —r — noniA,
l l

If we imagine the radius R to increase indefinitely, the ring solenoid be¬

comes a long straight solenoid in the limit. Equations (10-39) then
apply to a straight solenoid for which end effects may be neglected.
In general, M is proportional to the product of the respective number of
turns and L varies as the square of the number of turns of the circuit in
question.
Note that in the case we have taken

M = V%L~2.

In general, when the windings are not thin and directly on top of one
another, M will be smaller, or M < \ZL\L2. The ratio M/\/LiL2 is
called the coefficient of coupling; its value must lie between 0 and 1.

Example. To help grasp the magnitude of the henry unit, take the following
numerical case. For a ring solenoid let IVi = 1000, N2 = 400, l = 40 cm, A =
3 cm2. We then find from Eq. (10-36) that

M = 4tt X 10-7 X 1000 X 400 X 3 X 10-4/0.4

= 3.77 X 10 ~4 henry
= 0.377 millihenrys.

10-9 Energy stored in a magnetic field. When we build up a current

in a circuit, an external source of emf must do work against the induced
back emf. Work is also expended in heating the wire, but such work need
not concern us here. The rate at which work must be done against the
back emf 8 is
P = -87,

where the positive sense is the direction of the current. Using Eq. (10-32)
we have
10-9] ENERGY STORED IN A MAGNETIC FIELD 215

If we integrate this expression between the limits 7 = 0 and 7 = 7, we

get for the total work W expended in building up the current in the circuit

W = \L12. (10-40)

Since the work represented by Eq. (10-40) may be recovered by letting

the current decrease back to zero, we must consider it as being stored some¬
where while the current is flowing. Analogous to the interpretation of the
electrostatic field (Section 8-11), we may think of this energy either as being
associated with the charges comprising the current, or as being stored in the
region containing the magnetic field resulting from the current. According
to the first viewpoint, the stored energy is in the conductor and in a
kinetic form, while in the field point of view we think of the energy as
traveling out into space while the field is being established and returning
to the wire when the field collapses. We shall adopt this latter viewpoint
for the same reasons as were given in Section 8-11. We shall see that time-
dependent fields travel through space with the speed of light, and that
such fields may transmit energy to a distant receiver. However, it is in¬
teresting to note that in circuit theory the self-inductance L plays a role
analogous to inertia, or mass, in mechanics.* Because of self-inductance,
work is required to increase the current in a circuit, just as it takes work to
increase the speed of a mass particle, or any object with inertia.
We shall consider the energy given by Eq. (10-40) as being stored in the
magnetic field and proceed to find an expression for the energy density, or
energy per unit volume. Although, as in Section 8—11, we shall take a par¬
ticular example, the resulting expression is generally valid.
Let us again take a thin toroidal solenoid. Here, as before, we assume a
negligible difference between the inner and outer circumferences, or, if we
wish to be exact, we consider a length l of a very long straight solenoid.
Now we must consider just one winding, say that with N turns on a mean
circumference l. According to Eq. (10-37), we have

L = y N2A.

Substituting this in Eq. (10-40) yields

w =.|| n2ai2.

The magnetic field exists only inside the winding; in this region of volume
lA, where A is the cross section, the magnetic induction is nearly uniform

*See Vol. I, Section 6-8.

216 FARAD AY’S LAW. ELECTROMAGNETIC INDUCTION [CHAP. 10

for a thin toroid and is given by Eq. (9-25), namely,

NI
B = ho -j- ■

Eliminating NI between the last two equations, we get

R1 2 3/ A
W = • (10-41)
2mo

If we assume a uniform field within the winding and postulate that the
energy W is distributed evenly throughout the volume IA, we have

1 B2
Energy density of magnetic field = - —- • (10-42)
2 no

This expression should be compared with Eq. (8-59), the corresponding

one for the electric field in empty space.

Problems

1. A closely wound coil of N turns 4. Two coils, with self-inductances

and area A rotates about a diameter of 0.2 henry and 0.3 henry, respec¬
which is perpendicular to a uniform tively, lie side by side. Their mutual
magnetic field of induction B. (a) If inductance is 0.1 henry. When the
the angular velocity is co, find an ex¬ current in the first coil is increasing
pression for the instantaneous induced counterclockwise at the rate of 100
emf when the normal to the coil makes amp/sec and that in the second is
an angle 6 with the direction of the increasing clockwise at the rate of 60
field, (b) For constant w, find the root amp/sec, what is the magnitude and di¬
mean square value of S. rection of the emf induced in each coil?
2. A straight wire of length l is 5. Same as problem 4 except that
rotating with angular velocity u in a the coils are coaxial, or one in front of
plane perpendicular to a magnetic field the other.
of induction B. Find, in terms of l, co, 6. A linear current circuit (see Sec¬
and B, the emf induced between the tion 9-5) carries a current I. A rec¬
ends of the wire when the axis of rota¬ tangular circuit has two sides of length
tion passes through (a) one end, (b) l parallel to the wire, one at a distance a
the center. and the other at a distance b from the
3. A metal disk of 10 cm radius wire, (a) Find the flux through the
rotates at 3000 rev/min perpendicular rectangular circuit, (b) What is the
to a magnetic field whose B = 0.1 mutual inductance between the two
w/m2. An external circuit is connected circuits?
by brushes to the center and rim of this 7. A circular coil of Ni turns and
disk. Find the induced emf. radius a lies near the center and in the
 PROBLEMS 217

plane of a large circular coil of N2 Aw = —eB/2m. (b) Find the change

turns and radius b. Find the coefficient in the product IA, where I is the
of mutual inductance, assuming b^> a. equivalent current for the orbit and A
8. A circuit consists of two con¬ is its area.
centric cylindrical shells of length l 11. Two long straight parallel wires
and radii a and b, respectively, con¬ carry a current I in opposite directions.
nected at each end by flat plates. The Find the induced electric field E at a
current I is carried in opposite direc¬ point (in the plane of the wires) dis¬
tions in the two cylinders and the two tant r\ from one wire and r2 from the
end sections. Take b > a. (a) Find B other when the current in the wires is
at a distance r from the axis, a < r <b, changing at the rate dl/dt. Use Eqs.
using the circuital form of Ampere’s (9-12) and (10-24).
law. (b) Compute L, the coefficient of 12. Show that the work required to
self-inductance, in terms of a, b, l, and establish a current 11 in circuit 1 and a
MO- current 12 in circuit 2 is W = \L\1\ +
9. Referring to problem 8, put the \L2I2 + MI1I2, where L1 and L2 are
expression for B in part (a) into Eq. the self-inductances and M is the
(10-42), to get an expression for the mutual inductance.
energy density of the field as a func¬ 13. Show that if the circuits in
tion of r. Integrate and obtain the problem 12 are rigid, but one circuit is
relation between the total energy of the displaced a small amount ds', then
field and a, b, l, I, and mo- Check your Ii^dM/ds' represents the component
result by using Eq. (10-40) and the in the direction of ds' of the force on
expression for L found in part (b) of the displaced circuit due to the other
problem 8. circuit.
10. Suppose that an electron of 14. Show that when two coils are
charge e and mass m is traveling in a connected in series, the self-inductance
circular orbit of radius R about a of the combination is L = L\ L2 ±
nucleus, the angular velocity being w 2M. Explain which connection leads
when there is no external field. A mag¬ to the plus and which to the minus sign
netic induction B is now applied paral¬ in the last term.
lel to o>. Assume that R does not 15. Find an expression for the self¬
change but that w changes by a small inductance of two coils of negligible
amount Aw, Aw <3C w. (a) Show that resistance connected in parallel.
 CHAPTER 11

MAXWELL’S POSTULATE AND ELECTROMAGNETIC

EQUATIONS

Maxwell correlated the expressions for electric and magnetic fields based
on the laws of Coulomb, Ampere, and Faraday. He found that, in general,
these expressions were not compatible with the equation expressing the
continuity principle that electric charge can neither be created nor de¬
stroyed. He saw that the field equations were consistent with the equation
of continuity for a steady state, but not when the charge density at any
point was changing with time. Maxwell found that to remove this incon¬
sistency in the case of a nonsteady state it was only necessary to add to
the equation for V X B another term, one involving the time rate of change
of the electric field E. With the addition of this new term, Maxwell’s
field equations became a concise summary of the laws of electromagnetism.
In the present chapter we shall consider Maxwell’s equations as they
apply to fields due to charges located in vacuo and (to a good approxima¬
tion) in air. These equations may be modified to include fields in media
with special electric or magnetic properties; in Chapter 12 we shall discuss
the macroscopic or nonquantum effects of such media. The great impor¬
tance of Maxwell’s work, however, was that it led him to the development
of the electromagnetic theory of light. We shall conclude this chapter
with the derivation, from Maxwell’s equations, of the wave equations for
fields in empty space.

11-1 The field equations based on the laws of Coulomb, Ampere, and
Faraday. Let us summarize the expressions which we have developed for
the divergence and curl of E and B.
From Coulomb’s law (Chapter 8), we found that

(11-1)

V x E = 0 (11-2)

for an electrostatic field (see Eqs. 8-13 and 8-14). We should realize that
these equations for a static field may not be sufficiently general to hold
true for fields varying with time. In fact, we found from Faraday’s law
that V x E is not zero at a point where the magnetic induction B is chang¬
ing with time, for the correct relation then (see Eq. 10-25) is

V x E = —B. (11-3)
218
11-1] maxwell’s field equations 219

Equation (11-2) is a special case of Eq. (11-3). Since we find experimentally

that electric fields arise only because of static charges or through electro¬
magnetic induction, we conclude that Eq. (11-3) is general. In regard to
Eq. (11-1), we also have reason to believe that it is valid generally. For
one thing, we know that an electric field diverges from a positive charge
and converges toward a negative one; no other such sources of divergence
and convergence of E are known. Furthermore, we shall see later that Eq.
(11-1) forms one of a consistent set of equations which together lead to the
prediction of electromagnetic waves, as experimentally observed.
Let us turn next to the magnetic field. Isolated magnetic poles are
not found in our physical world. This means that there are no sources
from which a magnetic field diverges and there are no sinks toward which
such a field converges, so that we must have

V-B = 0 (11-4)

everywhere. This was proved, in the case of steady currents, to follow

from Ampere’s law (see Eq. 9-35). According to Ampere’s law, lines of B
form closed circles about the axis of a current element or moving charge,
which is equivalent to saying that the field neither converges nor diverges.
This must also be true when the currents vary with time, for if we take the
divergence of each side of Eq. (11-3), we get

V•V X E = —V -B.

The left side of this equation vanishes because the divergence of the curl
of any vector is zero, and on the right side we may transpose the order of
the space and time differentiation. We thus find that

— (V • B) = 0, (H-5)
dt v

and since V • B is zero when the currents are steady, it must remain zero
even if the currents are not steady at a later time. All that has been said
about conduction currents applies as well to the so-called displacement
currents, to which we shall come presently. Equation (11-4) is therefore
true generally.
The expression for the curl of B derived from Ampere s law was

V x B = moJ (H-6)

(see Eq. 9-26), but we must remember that the magnetic field referred to
in this equation was assumed to be due to steady currents. We saw that
an electric field may originate from (1) static charges, and (2) changing
220 maxwell’s postulate [chap. 11

magnetic fields. Does this not suggest that there may be two possible
origins for a magnetic field, say (1) steady currents, and (2) changing
electric fields? We shall see that Maxwell answered this question affirma¬
tively. However, for the present we shall simply state that of the four
equations,
V*E = p/e0, V x E = —B,
(11-7)
V-B = 0, V x B = moJ,

the first three may be considered to hold for either a steady or a nonsteady
state, but that the last equation has been proved only for a steady state.

11-2 The equation of continuity. In the flow of fluids the principle of

conservation of matter may be stated in this way: if the net mass of fluid
crossing the surface bounding a closed volume is not zero, then the density
of the fluid within the volume must change with time in such a manner
that the time rate of increase of fluid within the volume equals the net rate
of flow of fluid into the volume. Designating the mass density at any point
by p, and the velocity of the fluid at that point by v, the mathematical
expression of this principle is

v • (pv) = ~ yt = -p. (11-8)

This is called the equation of continuity and it is derived in texts dealing

with fluid flow.*
Now we may state the principle of conservation of charge as follows:
if the net charge crossing the surface bounding a closed volume is not
zero, then the density of the charge within the volume must change with
time in such a manner that the time rate of increase of charge within the
volume equals the net rate of flow of charge into the volume. From the
similarity between this and the principle of continuity for fluids, we see
that Eq. (11-8) also holds for the case of charge conservation, provided
that p stands for charge density and v for the velocity of such charge at any
point. The derivation of the equation is as follows.
Let S be the surface enclosing a volume r and let dS be an element of
this surface. The direction of dS is taken to be that of the outward normal,
so that pv • dS represents the charge per unit time leaving r across dS.
The time rate of flow of charge into r, which we shall call dq/dt, is thus
given by

_ S

*See Vol. I, Section 14-4; also Page, Introduction to Theoretical Physics.

11-3] maxwell’s postulate 221

The rate of change of charge within the region r is the volume integral of
the rate of change of the charge density. From the principle of conserva¬
tion of charge this volume integral must also equal dq/dt, so that

dq
dt

Let us equate these two expressions for dq/dt and use the divergence
theorem of Gauss to transform the surface integral into a volume integral
over the same region r. We then have

-//^•dS = -///*•(*)*-///£*•
s

Since both volume integrals are taken throughout the same arbitrary
volume, we may equate integrands to obtain

_ V.(pv)=§J, (11-8)

which is the equation of continuity.

11-3 Maxwell’s postulate. We shall now show that Eq. (11-6) is

inconsistent with Eq. (11-8). If we take the divergence of Eq. (11-6) and
remember that the divergence of the curl is always zero, we get

V • V X B = moV • J = 0.

By definition, J = pv, so we have

V • (pv) = 0.

If we combine this result with Eq. (11-8), we are led to the conclusion that

This, of course, is a true statement for a steady state, and it was for such a
state that Eq. (11-6) was derived. However, we know that, in general, p
may vary with time; hence for such cases Eq. (11-6) cannot be valid.
Maxwell first investigated mathematically how one could alter Eq.
(11-6) so as to make it consistent with the equation of continuity. Taking
the time derivative of Eq. (11-1), we have
222 maxwell’s postulate [chap. 11

^ = |(V.e0E) = V.(eoE). (11-9)

Let us introduce a new field vector D which, for air or vacuum, is defined
as
D = e0E, (11-10)

where D is the electric displacement. Equation (11-9) becomes

~ = V • D. (11-11)

Putting this into Eq. (11-8), we obtain

V • (pv) = V- J = -V D,
or
V ■ (J + D) = 0. (11-12)

Since the divergence of the quantity (J -f D) is always zero, this quantity can
(see Vector Analysis) be put equal to the curl of another vector. Although
Eq. (11-12) alone does not completely define this other vector, in choosing it
Maxwell was guided by the knowledge that the resulting expression must
reduce to Eq. (11-6) for a steady state. Hence he assumed that Eq. (11-6)
must be replaced by the equation

J + D = v X (11-13)

Let us define a fourth field vector H as

13
H = (11-14)
MO

in vacuo or in air. H is called the magnetic intensity. Equation (11-13)

then becomes
V X H = J + D. (11-15)

The inclusion of the D term in Eq. (11-13) we shall call Maxwell's

postulate. This equation is consistent with the equation of continuity,
Coulomb’s law, and Ampere’s law (Eq. 11-13 reduces to Eq. 11-6 for a
steady state). In general, then, for a nonsteady state Eq. (11-13), or Eq.
(11-15), takes the place of Eq. (11-6), and Eq. (11-3) replaces Eq. (11-2).
Maxwell’s postulate is not the only possible way around the inconsist¬
ency of our previous equations;, e.g., taking J + D = V X (B-f G)/p0,
with G an arbitrary function, would work mathematically. However, only
11-4] PHYSICAL INTERPRETATION OF MAXWELL’S POSTULATE 223

with his particular postulate was Maxwell able to, derive his theory of
electromagnetic waves. We may consider the experimental observation
of such waves, with the properties predicted, as the experimental basis
for Maxwell’s postulate. Furthermore, we shall show that this postulate
has a reasonable physical interpretation.

11-4 Physical interpretation of Maxwell’s postulate. Maxwell’s original

explanation of his postulate was puzzling, so we shall first consider
another viewpoint.
Let us use Eq. (11-15) and form the surface integral of V xH over
the surface S whose periphery is X. We then have

JJ(V
s
X H) • dS = JJ
s
(J + D) • dS.

If we apply Stokes’ theorem to the left side of this equation, it becomes

^
x
H ■ dl = JJ
s
(J + D) • dS. (11-16)

Just as fx E • dl was called the electromotive force around the path X,

we may define the magnetomotive force (mmf) around a path as

mmf = J H • dX. (11-17)

x

Substituting this into Eq. (11-16), we get

mmf = JJ s
(J D) • dS. (11-18)

Equations (11-16) and (11-18) tell us that there are two ways of producing
a magnetic intensity, one with an ordinary current of density J, as observed
by Oersted and postulated by Ampere, and the other by means of a chang¬
ing electric displacement, as postulated by Maxwell. Since D 01 E for
air or a vacuum, we may say that a changing electric field results in a mag¬
netic field. This is the converse of Faraday’s discovery that a changing
magnetic field produces an electric field.
If we take the case where J = 0 everywhere, Eq. (11-16) becomes

<fu-dl= Jfv-dS, (J = 0), (11-19)

x s
224 maxwell’s postulate [chap. 11

which is equivalent to the statement of Maxwell’s postulate given in

Chapter 7 (Eq. 7-11) since C2/Ci = 1/Wo- While Eq. (11-16) is a formu¬
lation of Ampere’s plus Maxwell’s laws, this last equation expresses solely
Maxwell’s new contribution, the D term in Eq. (11-13). Comparing Eq.
(11-19) with Eq. (10-24), we see that they are similar in form, with H
analogous to E and B analogous to D, but that they differ in sign.
Direct observation of a magnetic field produced by a changing electric
field is difficult. We cannot look for an induced “magnetic current” because
there are neither free poles nor conductors for magnetic currents. Also, we
cannot maintain a constant value of D long enough to measure the resulting
magnetic field, as we would that due to a steady current.

Example. Two parallel plates are 1 cm apart. With air between the plates,
the maximum possible intensity is found to be 3 X 106 n/coul or volts/m; this is
called the dielectric strength of air, or the value of E at which the medium becomes
conducting. We may, then, let the p.d. between our plates increase to 30,000
volts. Suppose that this is accomplished in 1 sec. Then E = 3 X 106 n/coul-sec
and

6 = £0® = sL^io9 = 2-65 x 10~5 amp/m2-

Compare this with possible values for the J term in Eq. (11-15). In a copper
wire of 1 mm2 = 10-6 m2 cross section a current of 10 amp would not be
excessive, giving a value for J of 107 amp/m2, many times greater than the value
computed for D above.

We have seen that Maxwell’s postulate is not as susceptible to direct

experimental verification as is Faraday’s law. This is why it was the last
fundamental law of classical electromagnetism to be discovered.
If Maxwell’s postulate is the converse of Faraday’s law, then why is
there not a converse to Ampere’s law? Or, putting the question slightly
differently, why does Eq. (11-18) have two terms, while the corresponding
equation for emf,

emf = — If B • dS, (11-20)

s

has but one term? The answer is that the term missing in Eq. (11-20)
involves a current density of “magnetic current,” or a flow of magnetic
poles of one sign, and because, as has been pointed out, isolated poles of
one sign and “magnetic currents” due to them are not found in our physical
world, the term analogous to J in Eq. (11-18) and the converse of Ampere’s
law do not exist. We must face the fact that the fundamental role of
electric charges leads to a certain lack of symmetry in our equations.
11-5] maxwell’s displacement current 225

11-5 Maxwell’s displacement current. We shall discuss briefly Max¬

well’s original interpretation of the D term, which he introduced, because
this explanation led to the name “displacement current,” a name which
is still applied to this term.

1 -Q--

Switch

Fig. 11-1. Displacement current.

Imagine two parallel metal plates, each of area A, connected to a battery

and switch, as shown in Fig. 11-1. Suppose that the plates are sufficiently
close, or are surrounded by guard rings, so that the field E between them is
uniform. At some instant after the switch is closed the charging current
is I and the charges on the plates are +g and —q, respectively. We then
have

I =

and, from Eq. (8-12),

a
E =
£o

where a is the surface charge density. Combining equations, we obtain

I — A cq E,

Thus the value of D between the plates is equal to the current per unit
area of the positive plate. Maxwell pointed out that if one pictured a
current density D between the plates, then, since AD = I, one could
think of the current / as being continuous all the way around the circuit,
including the gap between the plates. With D representing a current density
between the plates, we may then attribute the resulting magnetic field to
 maxwell’s postulate [chap. 11
226

this current, according to Ampere’s law. The objection to this interpreta¬

tion is that it offers no physical explanation of just what the current be¬
tween the plates is. Maxwell suggested that a strain equivalent to a limited
displacement of charges took place within the medium between the plates
while the field was changing, and that this constituted the current between
the plates. Thus he introduced the term displacement current. Although
we will see in the next chapter that a displacement of charge may occur
in a dielectric, and that this then contributes to D, we now know that the
D term exists even when there is a vacuum between the plates; hence Max¬
well’s interpretation has been superseded by that given in the previous
section. We may say correctly that in any circuit, e.g., that in Fig. 11-1,
the sum of the conduction and displacement currents is the same for any
section of the circuit.

11-6 Maxwell’s electromagnetic equations. Let us summarize the

general equations for the divergence and curl of the field vectors. They are
f
*—
pi,
to

V D = P, V X E = —B,
(H-21)
V • B = 0, V X H = J + D,

where, in vacuo or in air, D = e0E and B = moH. Equations (11-21)

are called Maxwell’s electromagnetic equations; they express concisely our
experimental knowledge of electromagnetic fields as contained in the laws
of Coulomb, Ampere, Faraday, and Maxwell.

11-7 The wave equations. To obtain the wave equation for one of the
field vectors we must obtain an equation containing that vector only. Let
us take as our example the wave equation for E in empty space. In Eqs.
(11-21) we set p and J equal to zero, so that Maxwell’s equations take the
symmetric form pSj
V - D = 0, V X E = —B,
(11-22)
V • B = 0, V X H = D.

Now we take the curl of curl E, or

V X (V x E) = -V x B.

Expanding the triple vector product on the left side, we get

V(V-E) — V-VE = —V X B. (11-23)

11-7] THE WAVE EQUATIONS 227

From Maxwell’s first equation, <

V • D = V • (e0E) = e0V • E = 0,

so that V • E = 0, while from Maxwell’s last equation, we have

V X H = V X — = — V X B = D = e0E,
Mo MO
or
V X B = CgMO^-

The time derivative of this last equation yields

d2E
V X B — £0MoE — eoMo
dt2 ’

Substituting back into Eq. (11-23), we get

V • VE = e0MoE, (11-24)

the wave equation for E, as can be seen by comparing this equation with
the three-dimensional wave equations in the theory of elasticity,* where
it is shown that the ratio of the coefficient of the V • V term to that of the
d2/dt2 term is (at least for plane waves) equal to the square of the speed.
We conclude that the electric field E is propagated in infinite empty space
with a speed c which is given by

c=-4=- (H-25)
V £0M0

A similar wave equation may be derived for D, B, or H, the speed of

propagation being the same as for E.
Let us compute the value of c. From the definition of the ampere, we
have
= 4tt X 10~7 n/amp2.

From a measurement of the Coulomb force between known charges we

can obtain an experimental value for e0; this value is very nearly

1
eo —
coul2/n-m2.
4tt X 9 X 109

*See Vol. I, Chapter 12.

228 maxwell’s postulate [chap. 11

Substituting in Eq. (11-25), we get

c = 3 X 108 m/sec, (11-26)

which, within experimental error, is the speed of light. Maxwell’s equations

thus lead to the conclusion that electric and magnetic fields may be
propagated through empty space as waves whose speed is the same as
that of light. We shall now show that electromagnetic waves in free space
are transverse, just as are light waves.
Consider E to be propagated as a plane wave in the X-direction, so that
E and its components are not functions of y or z. From the first of Eqs.
(11-22), V • E = 0, we have

dEx BE^ BE z _
Bx By ' Bz

For the plane waves we are considering, BEy/By and BEZ/Bz are zero,
hence BEX/Bx is also zero. This means that Ex does not vary with x,
as it would for a longitudinal wave in the X-direction. Here Ev and Ez
may depend on x, but they are components of E perpendicular to the
direction of propagation. A similar argument follows for B from the
equation V • B = 0. Hence plane electromagnetic waves are transverse.
The similarity between the predicted properties of electromagnetic
waves and those of light led Maxwell to postulate that light and electro¬
magnetic waves are one and the same thing.
Hertz, about fifteen years after Maxwell published his theory, first
produced electromagnetic waves experimentally; such waves have since
been found to have all the properties predicted by Maxwell. The micro-
waves used in radar are a good example of this type of wave.

11-8 Relation between E and H in a plane wave. Let us return to the

plane waves in the X-direction, which were shown to be transverse.
Without loss of generality, we may take the F-axis to be the direction of
the vector E, so that Ex — Ez = 0 and Ey — E.
From the equation V x E = —B, we have

BE z
—Bx, (a)
II
1
£

By

BEX BE z
By, (b) (11-27)
Bz Bx

BEy dEx
—Bz. (c)
Bx By
11-8] RELATION BETWEEN E AND H IN A PLANE WAVE 229

Since Ex and Ez are zero, and, Ey can vary only with x and t for plane
waves in the X-direction, we here have Bx = By — 0. This means that
the only component of B that varies with time, and hence is being propa¬
gated as part of the wave, is Bz. Therefore, if the electric vector is in the
F-direction, then the magnetic vector is in the Z-direction. E and B
(or H) are mutually perpendicular and E X B, or E x H, is in the direc¬
tion of propagation of the wave (see Fig. 11-2).

/B = moH

/
Fig. 11-2. Propagation of plane electromagnetic wave.

We still have not used the fourth equation in (11-22), namely,

dH z _ dHy
= Dx, (a)
dy dz

dHx dH 2
= Dy, (b) (11-28)
dz dx

dHy dHx
= Dz- (c)
dx dy

With E (or D) in the F-direction and B (or H) in the Z-direction, Dx =

Dz = Hx = Hy = 0, and Hz is independent of y. Equations (a) and
(c) lead to 0 = 0, but equation (b) becomes

dHz _ dEy
dx €° dt

while equation (c) of (11-27) takes the form

dEy dH z
dx Mo dt
230 maxwell’s postulate [chap. 11

For plane waves traveling in the -j-X-direction with the speed c, both Ey
and Hz must be functions of (x — ct).* Since

| /(* - ct) = -c l }(x - ct), (11-29)

the two previous equations may be written as

dH g dEy dEy dH z
dx e°C dx ~dx~ ~ M°C ~dx '

These integrate to

Hz = e0CEy, Ey = HocH z.

Since c = 1/\AoMo and Ey = |E|, Hz = |H|, we have

€o |E |2 = M0|H|2. (11-30)

11-9 Solution of the wave equation: waveguides. The wave equation

for E in empty space may be written as

V • VE = ^ E. (11-31)
c2

This is a homogeneous partial differential equation, so we may, if we wish,

break the solution into the product of two parts, one a function of the
coordinates and the other a function of the time. Let us call the former
E*(x, y, z). For waves of a definite frequency v, the time factor may be
written as elut or e~lat, where i = V-1 and

a) = 2irv = 27r/period

is the so-called angular frequency. We shall take

E = E*(x, y, z)e~iat. (11-32)

We see that this satisfies the condition that, at any point, E must repeat
itself every time t increases by 2tt/w, or for each period of the vibration.
This assumes that we are dealing with harmonic waves; but any waveform
that repeats itself with a definite frequency may, by Fourier analysis, be
considered as a sum of harmonic waves.

*See Vol. I, p. 216.

11-9] SOLUTION OF THE WAVE EQUATION: WAVEGUIDES 231

Substituting Eq. (11-32) into Eq. (11-31), we ge't

I
2
V • VE* = - ^ E*. (11-33)

This equation involves the coordinates only, yet its solution may be very
difficult. For one thing, while V • V4> (where $ is a scalar function) is
simply the divergence of the gradient of 4>, we cannot regard V • VE in
the same way, because E is a vector and VE is a dyadic, or tensor. However,
in cartesian coordinates V • VE = d2E/dx2 + d2E/dy2 + d2E/dz2. Equa¬
tion (11-33) is sometimes called the Helmholtz equation, or the time-
independent wave equation.
If the direction of wave propagation is taken to be the Z-direction, then
it is best to go back to the form of the wave equation containing the time
explicitly and to build the solution as the product of a function of x and y
times a function of (oit — yz), where y is called the propagation constant.
Thus we might write for the X-component of E,

Ex = Af(x, y)e-i^z), (11-34)

where A is a constant and we have assumed a harmonic solution. For a

plane wave in empty space, traveling in the Z-direction, E does not depend
on x or y, so that the f(x, y) term in Eq. (11-34) would become unity in
this special case and substitution of Ex = Ae~'l<'ut~yz) into Eq. (11-31)
would lead to „
2 W
? =

In general, however, y does not equal w/c (that is, co/y ^ c), as the
following illustration will show, yet 2ir/y will always represent the wave¬
length X. Since v = w/2-k, the phase velocity \v will equal co/y; thus the
phase velocity is not always equal to c in space free of charge, although it
does equal c for electromagnetic waves in infinite empty space (without
walls).
Illustration. Rectangular wave¬
guide. Consider the hollow conduc¬
tor or waveguide of rectangular cross
section shown in Fig. 11-3. The
waveguide extends in the Z-direc¬
tion, has a width a in the X-direc-
tion and a height b in the F-di-
rection. We shall show that waves
of wavelength shorter than a cer¬
tain limit may be propagated down
the length of the pipe, i.e., in the
Z-direction.
232 maxwell’s postulate [chap. 11

The wave equation for E is

1 # *
V • VE = — E.
c2

Each component of E must satisfy such an equation, so that, using car¬

tesian coordinates, we have

d2Ex tfE* d^Ex _ d2Ex

(11-35)
dx2 dy2 dz2 c2 dt2

and similar equations for Ey and Ez. From what has been said, we may
assume that the solutions for Ex, Ey, and Ez will each contain the factor
e~Uat—ye) muitiplied by some function f(x, y). In finding the form of the
latter, we may turn back to the solution of the equation

d2V d2V
dx2 dy2

in Illustration 3 of Section 8-8. We will then see that it is reasonable to

assume here that

Ex = A i cos (fcix) sin (k2y)e l("i yz),

-Ey = A2 sin (kjx) cos (k2y)e~l(~ut~7Z), (11-36)

Ez = 0,

where Ax, A2, ki, and k2 are constants.

Let us see if the solutions in (11-36) will work. First, they satisfy the
wave equations for Ex, Ey, and Ez, provided that

k\ + k\ + t2 = (11-37)
0
where X0 is the wavelength in empty space (with no walls) of waves with
the frequency u/2t. Second, these solutions satisfy the boundary condi¬
tions
Ex = 0 at y = 0, Ey = 0 at x = 0,

conditions that must hold since we assume the walls to be perfectly

conducting and so incapable of supporting a field. Third, our solutions will
satisfy the boundary conditions

Ex = 0 at y = b, Ev = 0 at x — a,
11-10] RELATION BETWEEN VECTOR AND SCALAR POTENTIALS 233

provided that we choose v ,y 1

and ’ (11-38)
a b
where m and n are integers.
From Eqs. (11-37) and (11-38) we see that we must have

O
If 7 is imaginary, the eiyz factor will either increase or decrease indefinitely
with z, depending on the sign of y. E cannot increase indefinitely, since
that would involve a corresponding increase in energy. A decrease in E
with z involves attenuation of the waVe. Hence only those waves will be
propagated in the waveguide for which

iHs2+©2-
The largest value that X0 may have is called the cutoff wavelength. For
a > b, this is found by taking m = 1 and n = 0, giving 2a as the cutoff
wavelength.
We have been considering here only waves for which Ez = 0, that is,
transverse electric waves; the various integral values of m and n give us the
so-called TETOn modes for our waveguide. Similarly, we could have taken
Hz — 0 and obtained the transverse magnetic, or TMmre, modes. Further
discussion of waveguides and modes is left to a more specialized treatment
of this subject. However, the presentation just given has shown that with
a waveguide of laboratory dimensions, say 0.5 cm in width, wavelengths
longer than 1 cm cannot be transmitted without attenuation, while some¬
what shorter waves may be propagated if they properly “fit” the wave¬
guide; very much shorter waves, such as those in the optical region, are
easily transmitted with little effect from the walls.

11-10 Relation between the vector and scalar potentials. We shall

now investigate whether the potential functions A and F, which were
previously introduced, should be altered because of the addition of Max¬
well’s postulate. We had
E = — VF — A,
B = V X A.

Let us substitute these in Maxwell’s equations (11-21):

 maxwell’s postulate [chap. 11
234

The equation V X E = —B is satisfied, for since V X V4> is identically

zero, substitution for E and B leads to the identity V x A = V x A.
The equation V • B = 0 is also still satisfied, since V • V X A is identi¬
cally zero.
We next consider the equation V • D = p, or V • E = p/e0. Substituting
for E, we have
V • V4 4- V • A =-— ■ (11-41)
eo

Next we substitute the potentials in the remaining equation, V X H =

J + £>, to get

V X (V X A) = MoJ - + A).

When we expand the left side of this expression, it becomes

V(V • A) — V • VA = moJ — Moeo(^^ + A). (11-42)

We have seen that for empty space the field vectors satisfy a wave equation
such as Eq. (11-24), which means that when p = J = 0 the potentials
from which the field vectors may be derived must also satisfy a wave
equation with the same numerical coefficients. This is possible if

V-A=-e0Mo4r, (11-43)
or
V • A = ~ eoMo'f’-

With p = 0, Eq. (11-41) now becomes

V • V4 = e0Mo'E

and with J = 0, Eq. (11-42) now reduces to

V • VA = cqmoA,

so that 4r and A satisfy the correct wave equations.

When p 9^ 0, Eq. (11-41) becomes

V • V4 = eoPoE — — ? (11-44)
eo

which is called d’ Alembert’s equation. This more general equation contains

the Laplacian operator, and reduces to the wave equation when p = 0,
and to Poisson’s equation when T = 0 (as for a steady state). The function
A satisfies a similar equation.
11-11] RETARDED POTENTIALS 235

Illustration. Slowly moving point charge. Consider a charge q at the

origin to be moving with velocity v, where |v| « c. At the point P(x, y, z),
whose distance from the origin is r, we have

'P q 1
~ 4tT€0 (z2 + y2 + Z2)1/2 ’

assuming to be the electrostatic potential. is the time rate of change

of 'fr at P due to the motion of the charge and the consequent variation
in r. Since only the relative, but not the absolute, motion of q and P
matters, we will have the same result if we keep q fixed and let P move with
velocity (ix + jy + kz) = —v. Then

• ...
d'k ... d* ...
* “ & w + ai (y) + ^ (2)

q xx + yy + zz
4reo (x2 + y2 + *2)3/2 ‘

From Eq. (9-39),

A mo _SY_
4-7T (x2 + y2 + z2)l>2 '
so that

V •A =
Mo£
4t G
dx \(a;2 + y2 + z2)112
A ( -y
dy \(x2 + y2 + z2)112.

i(_=i_)
dz \(x2 + y2 + z2)112/

_ mo? xx + yy + zz
— 4tt (x2 + y2 + z2)3'2

Thus Eq. (11-43) is satisfied in this case.

11-11 Retarded potentials. We have seen that electromagnetic fields

travel in empty space with the speed of light. This speed is so great that
in ordinary laboratory experiments involving distances of a few meters
only, the time required for the field to travel from a charge to the point of
observation is practically nil. However, the speed of light is finite, and we
must realize that at a point P the field resulting from a moving charge is
determined by the position and velocity of the charge at a slightly earlier
time. If the charge was then at a distance r from P, the resulting field at
P at the time t is that field determined by the position and velocity of the
charge at the earlier time (t — r/c). Hence in our potential functions A
236 maxwell’s postulate [chap. 11

and v must be the velocity and r the distance from the charge q to P
at a previous instant, earlier by r/c than the time under consideration. Of
course, r/c varies with the distance from P to each respec tive charge. When
the potentials are computed on this basis they are called retarded 'potentials.

11-12 Energy flow and Poynting’s theorem. Let us return to Maxwell’s

equations (11-21). We shall dot H into the equation for V X E, then we
shall dot —E into the V X H equation, and finally we shall add the
resulting equations. The reason for this procedure is that, from vector
analysis, *

H • (V X E) — E • (V X H) = V • (E X H).

Since for air B = mqH and D = e0E, we have in this case

V • (E x H) = —moH • H — e0E • E — E • J,
or

V . (E X H) = - | ~ (e0E2 + MOH2) - E . J. (11-45)

This is the electromagnetic energy equation, as we shall show.

Let us integrate Eq. (11-45) throughout a volume r. We then have

jJJv.(E X K)dr = -
T
~ JJJ
T
\ (e0£2 + y0H2)dr

- JJJe
r
• J dr. (11-46)

Applying the divergence theorem to the left side, this expression becomes

JJ
s
(E X H) • dS = -|///
T
l (e0E2 + n0H2) dr - JJJE-Jdr,
T
(11-47)
where S is the surface surrounding r.
First consider a volume r for which E X H = 0 everywhere on its
surface, so that we have

- | /// \ (eoE2 + MOH2) dr= JJJe ■ J dr. (11-48)

*See Vol. I, p. 24.

11-12] ENERGY FLOW AND POYNTING’S THEOREM 237

The term E • J = E • pv represents the rate at which work is being done

by the field per unit volume and fff E • J dr is the rate at which work is
being done by the field on the moving charges in the volume r. The work
done on a charge may accelerate the charge or, if the charge encounters
resistance to motion, it may increase the energy of the conductor (the
Joule heating effect). Where does this energy come from? It cannot come
from outside the volume, for this would mean a flow of energy across the
surface S, and in Eq. (11-48) the term involving a surface integration
over S is missing. Therefore the work done by the field must be at the
expense of the energy stored in the field within the volume r. The left
side of Eq. (11-48) is expressed as the time rate of decrease of the integral
of ^(e0E2 + mo#2) throughout the volume, so the integrand must repre¬
sent the electromagnetic energy per unit volume. Thus the

Energy density of an electromagnetic field = %(e0E2 -f- mo#2)- (11-49)

This conclusion is in agreement with those reached for the electrostatic

field (Eq. 8-59) and for a steady magnetic field (Eq. 10-42).
Now take a region r in which J = 0 everywhere. How can we interpret
Eq. (11-47) in this case? If the right side of the equation is positive, so
that the energy stored in the field is decreasing, it must mean that energy
is flowing out across the boundary of the region. The left side of Eq. (11-47)
is in the form of a surface integral, so that the integrand must be the rate
of energy flow out per unit area of surface. Designating this vector as n, we
conclude that
n = E x H (11-50)

is the amount of energy per unit area and per unit time crossing a surface
whose normal is in the direction of II. (n is known as the Poynting vector.)
Equation (11-45) is the equation of, continuity for the flow and con¬
servation of electromagnetic energy. It is known as Poynting’s theorem.
We see that for energy to be propagated in an electromagnetic field we
must have both an electric and a magnetic field present, and these fields
must not be parallel. These conditions were satisfied in the case of plane
waves, for which we showed that E X H is a vector in the direction of
wave propagation. A static charge, however, does not produce a magnetic
field, and so in electrostatics there is no Poynting vector. For a charge
moving with constant velocity, the electric and magnetic fields move with
the charge, and the total energy stored in the field remains the same; at a
point in the neighborhood of the charge and not in the direction of motion,
E X H is not zero, but the flow of energy keeps pace with the motion of
the charge and no energy is radiated out into space. Only when a charge
is accelerated does it radiate energy (see Section 11-13).
 maxwell's postulate [chap. 11
238

Illustration 1. Plane wave. For a plane wave, E is perpendicular to H

and e0E2 = mo#2- Therefore we have

n = EH = J— E2,
\ MO
or since c = l/VeoMo>
n= c6qE2.

We must also have

II = energy density of the field X speed of field

= c X energy density of the field,

so that
€qE2 — energy density of the field.

Since e0E2 = mo#2 here, we can also write

%e0E2 + |mo#2 = energy density of the field.

This agrees with Eq. (11-49).

Illustration 2. Long straight current circuit. Consider a straight con¬
ductor of radius R carrying a steady current I. Let V be the potential
drop in a length l of the wire. Then inside the wire the electric field must be

V
E =
7 ’

in the direction of the current. We shall show in the next chapter that the
tangential component of E is the same on each side of an uncharged surface,
so that just outside our wire E is the same as it is inside. We saw that the
magnetic induction outside the conductor is directed in circles, according
to the right-hand fist rule of Section 9-5. Just outside the surface of the
wire,
MO I_ I
B = H =
2-k R 2tR

in the tangential direction. Just outside the wire E and H are perpendicular
and
VI
n = E x H =
2-71-Rl

into the wire. This is the rate of flow of energy per unit area. For the
length l the surface area is 2tRI, so that the rate at which energy enters
this section is just VI. But VI is also the rate at which energy is being
11-13] RADIATION FROM AN ACCELERATED CHARGE 239

taken from the source of emf ih the circuit and is being used to drive the
current through the length l of the wire. According to the Poynting view¬
point, this energy, which goes into raising the temperature of the system,
is supplied by the field outside the wire. However, in the source of emf,
say a battery, I and H have the same relative directions, while the direction
of E is reversed, so that n is outward. So, if we wish, we may picture energy
as flowing from the battery out into the field and then back into the wire.
Perhaps in this example a viewpoint built around the analogy between
the electrical circuit and a water system with a pump would have more
physical appeal, but there are other problems, involving electromagnetic
waves and radiation, in which the Poynting vector is a useful concept, as
the next section will illustrate.

11-13 Radiation from an accelerated charge. In Fig. 11-4 let a charge

q be at rest at the origin at the time t = 0 and let OPM be a line of force
of the electric field due to q. Now suppose that the charge is given an
acceleration a for a short time At, in which time it acquires the speed
v — a At and moves to the point Qx. Assume that v « c, c being the speed
of light. Now let the charge move with the speed v for a time t (t )>> At),
reaching the point Q2 at time t + At after the start. After leaving Qu
the field near the charge will have a new pattern, that associated with a
charge of speed v, rather than that of a charge at rest at 0. This change
in the field will be propagated outward with the speed c. At a distance
greater than c(t + At) the change will not have arrived at the time t -(- At;
thus at the point M the original line of force, with the direction PM, will
still exist. Close to the charge, however, this same line of force will be the
line Q2L, nearly parallel to OPM (if v « c), but emanating from Q2.

Fig. 11-4. Pulse radiated by accelerated charge.

«

Thus the line of force must have a kink LP, as shown. While LP may be
curved, it will be so short that we may treat it as a line segment, if At <$C t.
Therefore we see that Q2L = ct and OP = c(t + At), and Q2LPM is a
line of force.
 maxwell’s postulate [chap. 11
240

Let us describe a sphere of radius OP about 0 and one of radius Q2L

about Q2, as shown in the figure. With » « c, we may consider OQ2 small
compared with these radii, so that we may assume that the spheres are
concentric. The region between the two spheres contains the field due to
the acceleration of q. We are interested in the field E within this shell.
Let Er and Ee be the polar components of E. Since E is along LP, we
have
Ee LN
Er~ NP'

Since LN = OQ2 sin 8 = vt sin 8 (for ty> At), and NP — c At = cv/a,

then
Ee _ vt sin 8 _ at sin 8
Er c At c

We have Er = q/4:-jre0r2 at L, where r — ct, with only a slightly different

value for Er at P (since At « t). Thus we have

qat sin 8 qa sin 8

(11-51)
47T£0r2C 47r e0rc2

where we substitute r/c for t. This is the component of E that is being

radiated outward due to the acceleration of the charge.
Without loss of generality, we may wait until r — ct has become so great
that, at the point L or P, the wave is approximately a plane wave traveling
in the radial direction. Then we know from Section 11-8 that perpendicular
to Ee there must be a magnetic field H, and that according to Eq. (11-30),

qa sin 8
H = (11-52)
47r rc

(We have substituted 1/c for VeoMo-)

At the point P(r, 6) the Poynting vector is n = EeH, directed outward.
Hence we get
2 2 • 2 n
q a sm 6
(11-53)
1 ~ 16tr2e0r2c3

for the radiated energy per unit area and per unit time.
To find the total rate of radiation of the charge during its acceleration,
we must multiply Eq. (11-53) by the area 2xr2 sin 8 dd (the area of the
zone between 6 and 6 -(- dd on a sphere of radius r), and then we must
integrate with respect to 6. Calling this total rate of radiation dW/dt,
we get
 PROBLEMS 241
<
2 2
dW gV / ( sin3 ddd
q a
(11-54)
dt Sire0c3 70 6xeoC3

We see from this formula that the rate of radiation by an accelerated charge
increases with the square of its acceleration.
When electrons are stopped at the target of an x-ray tube a rapid de¬
celeration occurs and a part of the kinetic energy of the electron is trans¬
ferred to radiant energy, the rest being added to the internal thermal
energy of the target. (The correct quantitative explanation of this effect
requires the application of quantum mechanics.)
Another example of radiation by an accelerated charge is an oscillating
electric dipole, or an antenna in which the polarity of charge at the two
ends is opposite and continually reversing.

Problems

1. Derive the equation of continuity units, (a) Find the mksc unit for cr.
for the flow of (a) a fluid, (b) heat. (b) If an alternating electric field
Define all terms and state postulates represented by Eoeiat exists within the
assumed. copper, find the ratio of the conduction
2. Derive the wave equation for H to the displacement current density at
in empty space. a frequency of 1 megacycle/sec.
(Ik) (a) Derive the wave equation for 8. For a plane wave in free space,
E in a medium where p = 0, but show that the energy crossing unit area
J = aE, a being a constant (the me¬ in unit time equals the energy con¬
dium obeys Ohm’s law), (b) Taking tained in the volume c (the volume
E = E*e-“*, find the time-independ¬ swept out by unit area of the wave-
ent wave equation for this case. front in unit time).
4. Derive d’Alembert’s equation for 9. Suppose that for a plane wave in
the vector potential A. free space the Poynting vector is 0.06
5. For plane waves the ratio |E|/|H| watt/m2, (a) Find E and H. (b) Find
is called the wave impedance. Show that the energy density of the field. These
the mksc unit for wave impedance is will be effective (rms) values.
volts/amp and find its numerical value 10. Suppose that the wave in prob¬
for free space. lem 9 has a cross section of 105 m2 and
6. Show that for an air-core toroidal that it is emitted as a pulse, the trans¬
solenoid the ra^io of the mmf to the mitter radiating for 10-6 sec. Find
flux is 1/p.oA, where l is the circum¬ (a) the length of the pulse, (b) the total
ference and A is the area of cross sec¬ energy radiated, and (c) the power of
the transmitter during emission.
tion.
7. For copper, we find that J =oE, 11. Suppose that a charge q oscil¬
where a has the value 5.8 X 107 mksc lates about the origin, its displacement
242 maxwell’s postulate [chap. 11

at any time t being given by y — 13. (a) Write the wave equation for
A sin cot. Show that the energy radi¬ a scalar function u in spherical co¬
ated per cycle is W = q2A2co3/6e0c3. ordinates. (b) Show that u =
12. For the waveguide in Section Aei(-at~yr\/r, with y = oo/c, is a solu¬
11-9 the components of E were found tion. (c) What sort of wave does this
to be given by Eqs. (11-36). (a) solution represent?
Show that in these equations Ai = 14. Discuss the Poynting vector and
—Azki/hx. (b) Prove that H has a its physical interpretation in the case
longitudinal component Hz. of a charged bar magnet.
 f
CHAPTER 12

ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA

12-1 Introduction. In this chapter we shall discuss the behavior of

material media when placed in electric and magnetic fields and the re¬
sulting modifications in the equations which we have derived for air, or a
vacuum. From a practical point of view this is an extremely important
subject; however, in a theoretical exposition of the laws of the physical
world we are justified in treating this topic rather briefly, because we shall
have to introduce postulates that are valid only for a certain class of sub¬
stances, so that our resulting theory will not have the wide range of
applicability it would have if it were based on more general laws. For
example, we shall postulate the existence of relationships between certain
pairs of electromagnetic quantities. These relationships are founded on
experimental observations which indicate that the form of these so-called
constitutive relationships, and the numerical factors involved, vary from
substance to substance. It is also found that for a given substance a certain
relationship is valid only over a limited range, just as Hooke’s law holds
only up to the elastic limit. The relationships assumed may be called
“laws, ” but it must be understood that they are not general laws of Nature,
such as the laws of thermodynamics, or the laws of Coulomb, Ampere,
Faraday, and Maxwell; these “laws” for material media have the same
standing as Hooke’s law in elasticity and the analogous law in the theory of
viscous fluids.
Suppose that two physical quantities, represented by the symbols x
and y, are functionally related, and the relationship is expressed by
y _ jf the form of the function / varies from substance to substance,
the most general procedure would be to leave the form unspecified as we
develop our theory, and then at the end to substitute the definite functional
relationship appropriate to a given substance. The disadvantage of this
method is that the theory tends to be too complicated. Suppose, on the
other hand, that we find experimentally that y is proportional to, say, x
for a wide range of values of x and for a large class of substances. We may
then write y = kxn, where k is a constant. This simple relationship is
much easier to handle theoretically than the general expression y = fix).
As in the theory of elasticity, the simplest relationship one can assume
between two variables is a linear one. Such a relationship may be a first
approximation'to a more complicated one, but it is an interesting property
of our material world that linear relations are found experimentally to be
the most common and important ones. By observing how this case is
treated, the reader may learn how to handle the theory when the situation
is more complicated.
243
244 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

We shall postulate, then, that we are dealing with linear media. Unless
stated otherwise, we shall also assume that the medium is homogeneous
(same properties at all points) and isotropic (same properties in all direc¬
tions). Such linear, homogeneous, isotropic media are sometimes called
Class A media; they do not include crystals.
Three linear relationships between electromagnetic quantities will be
treated, namely, (1) that between the current density J and the electric
intensity E, (2) that between the electric displacement D and the electric
intensity E, and (3) that between the magnetic induction B and the
magnetic intensity H.

12-2 Ohm’s law: the relation between J and E. We may always write

J - cE, (12-1)
which is a definition of the electrical conductivity a. In general, a might be
some complicated function of E, or it might vary in an erratic way, like
the ratio of a person’s age to his weight, but henceforth we shall postulate
for the substances under consideration, that

the current density at a given point is proportional to the electric intensity

at that point, i.e., a is a constant independent of E.

This is one form of Ohm’s law. It is found to hold true for many common
materials, notably metallic conductors and electrolytes.
Suppose that we consider a portion of a conducting rod, or wire, whose
length is l and whose cross section is A. If the field inside the conductor is
longitudinal and of magnitude E, the current I will be

I = JA = *EA. (12-2)

Let V be the potential difference across the length l, so that E = V/l,

which yields

1 = (t) v■ (12~3)
We now define the resistance R of any section of a conductor as

V
R = j- (12-4)

In the mksc system R is measured in volts/amp, which, for simplicity,

are called ohms, so that

1 ohm = 1 volt/ampere.
12-3] kirchhoff’s laws 245

From Eqs. (12-3) and (12-4), we have

R = ~^r- (12-5)
aA

If, for a given conductor, a is independent of E or V, we see that R is also;

then Eq. (12-4) not only defines R, but states that V is proportional to I.
For a metallic conductor R is constant over a large range of currents. This
is a common way of expressing Ohm’s law, although the “point statement ”
given in Eq. (12-1) is more applicable to field theory. From Eq. (12-5)
we can show that resistances in series add directly, and that resistances in
parallel must be added reciprocally to get the reciprocal of the equivalent
single resistance, just as for heat flow.*

12-3 Kirchhoff’s laws. The main reason for introducing these well-
known laws here is to discuss their significance, rather than their applica¬
tions. Consider a network of current circuits containing sources of emf
Si, S2, S3, • • • and resistances R1} R2, R3, • ■ ■ • Call the respective currents
Iu I2j /3, • • • , where the subscript refers to a given branch of the network
(Fig. 12-1). Kirchhoff’s laws state that for steady (direct) currents:

Fig. 12-1. Kirchhoff’s laws.

I. At any junction the sum of the currents entering equals the sum of the
currents leaving that junction, and
II. Around any closed loop in the network the sum of the emf s equals the
sum of the products of current and resistance for each conductor in the loop.

The first law is a statement of the continuity principle, or the law of the
conservation of charge. With steady currents, charge cannot accumulate
anywhere and daw I must hold true. In the equation of continuity,

dp
V . (Pv) = V • J = — —>

*See Vol. I, problems 15-4 and 15-5.

246 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

we may put dp/dt = 0 for a steady state and get

V•J = 0 (12-6)

as a statement of Kirchhoff’s first law for any region containing steady

currents only.
The second law may be proved as follows. In general, the total field E
may be the sum of a conservative electrostatic field E., and a nonconserva¬
tive field E', so that we may write

E = Es + E'.
From Eq. (12-1), we have

E = i,

which yields
J
Es + E' = -
a

If we integrate this expression for a given loop in the network, we get

Es -ds + fE'-ds

The first integral on the left is zero, because the electrostatic field is con¬
servative. The second integral is the electromotive force 8 around the
loop, or the algebraic sum of the emf’s of whatever sources are present in
the loop. The integral on the right side may be transformed as follows:

where, according to Eq. (12-5), dR is the resistance of the element ds.

jf I dR is the algebraic sum of the products of the current and resistance
for each segment of the loop. By algebraic sum we mean that if an emf or
current directed clockwise is counted as positive, an emf or current di¬
rected counterclockwise must be considered as negative. The equation
expressing Kirchhoff’s second law is

fE • ds = fl dR. (12-7)

12-4 Electric polarization. We come now to the discussion of the

electrical properties of nonconducting media, i.e., good insulators, or
dielectrics. The molecules of any substance contain positive nuclei about
12-4] ELECTRIC POLARIZATION 247

which negative electrons continually circulate. However, in a dielectric the

electrons in a given molecule are bound to that molecule and can never
leave it, while in a conductor some electrons can and do escape from the
molecules in which they have been circulating and so become free electrons.
When a material is placed in an electric field, a rearrangement of the
charges takes place. The positive nuclei are displaced slightly in the direc¬
tion of the field E, while the bound electrons are displaced in the opposite
direction. Any free electrons present will continue to move through the
material so long as the field is present. In a given field the displacement of
a charge bound to a molecule ceases when the force on it due to the field
E is balanced by a restoring force due to the other charges. The net result
is the separation of equal and opposite charges and the formation of
di-poles throughout the medium. The charges in a dipole are held apart by
the external field and they return to their original positions when the field
is removed, just as a stretched spring returns to its original state when the
pull on it ceases. The analogy between a dipole and a spring may be carried
further. The displacement of the charges in a dipole is usually found to be
proportional to the applied electric field, just as the stretch of a spring is
(within the elastic limit) proportional to the applied force.
When dipoles not oriented at random are produced throughout a medium
the medium is said to be polarized. If the dipoles are produced because of
the opposite displacement in a field of positive and negative bound charges,
they are called induced dipoles. There are also what are termed permanent
dipoles, that is, molecules in which the positive and negative charges are
distributed so as to have a net electric moment even in the absence of an
external field. These so-called polar molecules are normally oriented at
random, but in an external field they tend to be aligned more and more
with the field as the intensity of the field increases. When the field is re¬
moved, the permanent dipoles return to a random orientation because of
thermal agitation, a factor which also limits the polarization when the
field is applied. Measurement of the polarization produced by a given field
yields information about the structure of the molecules concerned.
Suppose that a dipole consists of two charges -\-q and q and that the
vector d represents the displacement of the positive from the negative
charge, i.e., d is the vector from the position of — q to that of +q. The
electric moment m of the dipole is defined as the vector

m = qd. (12-8)

The polarization P of a medium is defined as the dipole moment per unit

volume, and it must be numerically equal to the vector sum of the moments
of all the dipoles in the unit volume. In an isotropic medium, such as is
postulated here, P is in the same direction as the applied field E. The re-
248 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

sultant of the dipole moments in any finite volume must then be in the
direction of E. If there are N dipoles per unit volume and on the average
each has a moment m# in the direction of E, then

P =: NmE. (12-9)

12-5 Polarization charges. Consider first a uniformly polarized medium

with polarization P. Within the volume of this medium the positive end
of one dipole will be next to the negative end of an adjacent one, and there
will be no net volume charge (Fig. 12-2). At a surface whose normal is in
the direction of P, an unbalanced positive charge will appear, with an equal
negative charge on the opposite face. Let us take an area A of the positive

Fig. 12-2. Surface polarization charge.

face and let this form part of a volume Ad extending a distance d straight
into the medium, where d is the length of a dipole. Then the electric mo¬
ment of this volume will be (Ad)P. According to Eq. (12-8), the charge q
that must be displaced a distance d from the charge — q in order to produce
the dipole moment (Ad)P = APd is q = AP. On the surface there will
appear an unbalanced positive charge per unit area equal to P. Letting <r'
be the surface charge density due to polarization, we have

</ = P-n!, (12-10)

where n! is a unit vector in the direction of the outward normal. Now

take the more general case where P may vary throughout the medium, as
in Fig. 12-3. We see that in this case there may be a volume polarization
density p'. Let S be a surface enclosing a volume r of the medium. Then
the surface integral of P • dS will, according to Eq. (12-10), represent the
charge q' leaving t across the surface S during polarization of the medium, or

P-dS. (12-11)
s
12-5] POLARIZATION CHARGES 249

W:
Q (±3 77—LA
a£3Wo
VI—J
<3 &M
x = 0

Fig. 12-3. Uniform volume polarization charge density.

From the conservation of charge principle, a net charge q' leaving a volume
that was originally neutral must leave a charge —q' inside the volume,
where
t (12-12)
—Q

If we combine Eqs. (12-11) and (12-12) and apply the divergence theorem
to the surface integral, we have

or, equating integrands,

p'=—V- P. (12-13)

Example. Find p when P is in the X-direction and has a magnitude given by

(a) P = a, (b) P = bx, (c) P = cy, (d) P = dx2, where a, b, c, and d are posi¬
tive constants with the proper dimensions.
Solution, (a) There is a uniform P and p —V • P = 0. (See Fig. 12-2.)
=

(b) Here P is away from the plane x = 0, increasing with x. p = —V • P =

-b. (See Fig. 12-3.)
(c) P is parallel to the X-axis, increasing with y. p = —V • P = 0.
(d) P is everywhere in the X-direction, increasing with x2. p' = V • P =
—2 dx, which is negative when x > 0, and positive when x < 0. (See
Fig. 12-4.)

Fig. 12-4. Variable volume polarization charge density.

250 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

12-6 Gauss’ law for dielectrics. Suppose that we wish to compute the
electric field at some point in a region that contains both free and polariza¬
tion charges. Two alternatives are open to us. We may (a) treat the polar¬
ization charges on an equal basis with the free charges and consider
the medium to be in other respects like air or a vacuum, or we may (b)
disregard the presence of polarization charges, but treat the medium as
possessing a different permittivity e than that of air or a vacuum (that is,
e ^ <e0). While (a) has the appeal of greater physical reality, it has the
disadvantages that one must first determine the polarization charges and
that the polarization charges further complicate the expressions. In
method (b) the results are simpler and more analogous to those for free
space. Values of e for different media have been measured and listed. We
shall show how method (b) follows from method (a).
Let us apply Gauss’ law to the surface S enclosing the volume r. This
theorem may be extended to media other than free space if we write it as

JJ e0E • dS = fff(p + p') dr. (12-14)

s T

The inclusion of the p' term takes care of the alteration in the field due to
the presence of a polarizable medium. Equation (12-14) reduces to Eq.
(8-9) for free space. Now for p we substitute — V • P and obtain

Next we apply the divergence theorem to the last term and transpose it,
which yields

JJ (e0E + P) -dS = JJJ pdr. (12-15)

We define the electric displacement D for a dielectric medium as

D = e0E + P. (12-16)

The definition of D for air or a vacuum was a special case (with P = 0) of

this more general definition. Equation (12-15) now becomes

(12-17)

which is Gauss’ law for dielectrics.

12-7] THE POSTULATE RELATING D TO E 251

We have eliminated p', but we have D appearing in place of E. To find

the electric intensity we must know how D and E are related, and this
relationship is a property of the medium. At this point we introduce the
restriction that the medium is “linear, ” or what we have termed a Class A
medium.

12-7 The postulate relating D to E. As stated earlier, we shall assume

a linear relationship between D and E. In most dielectric materials this
postulate is experimentally verified for fields of less than the dielectric
strength (defined in the example in Section 11-4). If we write

D = 6 E, (12-18)

this equation defines e, the electric 'permittivity of the medium, and the
statement that for a given medium e is a constant independent of E is our
restricted “law” for Class A dielectrics. This postulate is equivalent to
saying that the polarization is proportional to the applied field, for from
Eqs. (12-16) and (12-18) we have

P = D — €oE = (e £o)E.

When we put Eq. (12-18) into Eq. (12-17), Gauss’ law takes the form

eE • dS P dr, (12-19)

which differs from that for air or a vacuum only in that e replaces e0.
Finally, we define the dielectric constant Ke of the medium as
6
K, (12-20)
60

Gauss’ law may then be written as

Ke 6qE • dS — P dr, (12-21)

which is the most convenient form, since values of Ke for different media
are listed in tables. (For air Ke = 1.0006 and for a vacuum Ke is unity.)
Ke is a dimensionless quantity, while e is expressed in terms of the same
unit as <„ (coul2/n-m2), and D in coul/m2, as before
From Eq (12-21) we see that if we compare the field for a given distnbu-
tion of free charges in a dielectric (filling the whole region where there is a
fie'ld) with the field for the same charge distribution in free space, then,
252 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

since the right side of the equation is the same for each case, KeE at a
point in the dielectric must equal E at the same point when the medium
is air or a vacuum. Thus the presence of the dielectric reduces the field
to 1 /Ke of its value without the dielectric.
If we change the left side of Eq. (12-17) back into a volume integral,
using the divergence theorem, and equate integrands, we get

V •D — p, (12-22)

which is the more general form of Maxwell’s first electromagnetic equation.

Thus lines of D diverge only from free charges; if the surface of a dielectric
contains no free charge, lines of D will not start or stop at the surface.
In Fig. 12-5 let S be a surface shaped like a pill box, with the same area
in medium 1 and in medium 2. Using Eq. (12-17) and taking p = 0, we
find that the normal component of D is the same on each side of the un¬
charged surface, or
Dln = D2n. (12-23)

Since in a dielectric the field of a point charge is still central and a func¬
tion of the distance from the charge to the point where the field is com¬
puted, we again have
V X E = 0 (12-24)

for a static problem. If we take jfE • dX around the path X shown in

Fig. 12-5, we then find that the tangential component of E is the same on
each side of the surface, or
Eit = E2t. (12-25)

Illustration. Charged 'parallel plates with a dielectric filling part of the

intervening space. As in Fig. 12-6, the plates A and B contain the free
surface charge densities -fo- and —a, respectively. A parallel slab of di¬
electric, whose constant is Ke, fills part of the space between. If we apply
12-7] THE POSTULATE RELATING D TO E 253

Fig. 12-6. Gauss’ law applied to di- Fig. 12-7. Lines of D and E.
electrics.

Gauss’ law in the form of Eq. (12-21) to a surface Si enclosing an area A'
of the positive plate, and extending into the air space between the plates,
we find as before that in the air (Ke =1) the field Ea is
a
Ea
eo

since there is no field inside the plate. By extending S to S2, so as to let

its right side lie in the dielectric, we find that the field Ed in the dielectric is

Ed = §!=-• (12-26)
Ke e

Now take the surface S3 to include only an area of the dielectric s left
face, one side of S3 being in air (where the field is opposite to the normal),
and*one side being in the dielectric (where the field is outward). Using
Gauss’ law in the form of Eq. (12-14), we get

— eo Ea + eo Ed — <*'>
t

where a' is the surface charge density due to polarization on this face of the
slab. Substituting for Ea and Ed, we find that

(12-27)
1

Note that Dd = eEd = = e0Ea = Da. In Fig. 12-7 the solid lines
represent E, while the solid plus dotted lines represent D for the case
where Ke — 3.
254 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

12-8 Capacitance. It is of practical importance to know how much

charge can be placed on an insulated conductor before the electric intensity
just outside its surface reaches such a magnitude that the medium sur¬
rounding the conductor no longer behaves like an insulator. Then, as
more charge is given to the conductor, the conductor loses charge to its sur¬
roundings. To determine when this point will be reached we must know (1)
the size and shape of the conductor, (2) the size, shape, and location of
other conductors in the vicinity of the given conductor, (3) the dielectric
strength of the intervening medium, and (4) the potential difference be¬
tween the given conductor and each conductor in its vicinity. In place of
(4) it is sufficient to know the charges on each conductor.
In practice, two important special cases arise. First, there is the case
where the given conductor is isolated, that is, very far from any others.
We may define the potential V of this conductor as the potential difference
between it and a point at an infinite distance. Let the charge on the con¬
ductor be q when its potential is V. We define the capacitance C of the con¬
ductor as the ratio
C — y- (12-28)

The second important case is when there are two conductors in the vicin¬
ity of each other and they are given the charges -\-q and —q, respectively.
Let Vab be the potential difference between the conductors. The capaci¬
tance C of this combination, which is called a capacitor, is defined as

C = (12-28a)

It is common practice to put the charge q on one of the two conductors

while the other is connected to ground. The charge — q will then be
attracted from the earth to the grounded conductor.
When considering a system of several conductors, a consistent definition
of C will be obtained from Eq. (12-28a) if we postulate that all conductors
are grounded except the one with the charge q.
The value of introducing the concept of capacitance is that if q and Vab
alone are varied, C is constant up to the breakdown point of the surrounding
medium, for the following reason.
The electric intensity at any point in the field of a charge q is propor¬
tional to q. If this field results in the production of polarization charges,
these charges and the intensity of the field due to them will also be pro¬
portional to q, provided that the medium is a Class A dielectric. Since the
potential difference between two points in such a field is the line integral
of the electric intensity, we see that

Vab q,
12-9] ENERGY STORED IN AN ELECTRIC FIELD 255

or C is independent of q and Vab C does depend on the size, shape, and

■

location of the conductors present and the value of Ke for the intervening
medium.
From the definition of C we see that the mksc unit of capacitance is the
coulomb/volt. We again introduce another name for a derived unit and
define
1 farad = 1 coulomb/volt.

This name honors the English physicist Michael Faraday, whose work in
connection with electromagnetic induction was discussed in Chapter 10.
Illustration. Parallel-plate capacitor. Consider the parallel plates de¬
scribed in the illustration at the end of Section 12-7. Let the separation of
the plates be l and the thickness of the dielectric slab be t. Then, as before,

Ea = — ’ Ed — j=-— ,
eo &eto

so that, if the negative plate is grounded, the potential of the other plate
will be
V = /E • <&. = Ea(l - t) + Edt

eo {l ~ + Kl.

If the plates each have an area A,

q = aA.

From Eq. (12-28) the capacitance of this capacitor will be

r_eo A . (12-29)
I — t A~ t/Ke

If the dielectric completely fills the space between the plates, so that t — l,
then C will be proportional to Ke.

12—9 Work required to charge a capacitor, and the energy stored in an

electric field. Suppose that we gradually build up the charge on a conduc¬
tor, bringing up a little charge at a time. When the charge has reached the
value aq (0 < a < 1), the potential will be aV, where q and V are the final
values of charge and potential, respectively. Let us now bring up the
additional charge dq = (da)q; the required work dW will be

dW = dq(aV) = qVa da.

256 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

Integrating from a = 0 to a — 1, we get for the total work to charge the

conductor (which may be the ungrounded plate of a capacitor)

W = qVa da

= \qV. (12-30)

If we use Eq. (12-28), this becomes

W = %CV2, (12-31)

which is analogous to Eq. (10-40) giving the work needed to build up a

current I in a circuit with self-inductance L.
In many respects a capacitor behaves like a spring, with charge corre¬
sponding to stretch, potential to stretching force, and 1/(7 to the stiffness
factor k of the spring. * The greater the charge on a capacitor, the greater
is the stored energy which will be released upon discharge.
From Eq. (12-29), when t = l we have C = Kee0A/l for a parallel-
plate capacitor with a uniform dielectric between the plates. For this case
V = El. Substituting these values in Eq. (12-31) gives

W = iKee0E2Al,

or, since Kee0 = e,

W = \tE2Al.

If we again take the view that the work done in charging the capacitor
goes into energy stored in the field, then, noting that the field is uniform
throughout the volume Al, we may say that

Energy density of electric field = \eE2. (12-32)

This expression reduces to Eq. (8-59) for free space.

12-10 Intensity of magnetization. Now we turn to the third main topic

of this chapter—the magnetic properties of substances. In some respects
we will find a close parallel between the magnetization of a medium and
its consequent contribution to the magnetic field on the one hand, and the
polarization of a dielectric and its contribution to the electric field on the
other; but again we must also recognize the essential differences between
electricity and magnetism, particularly that magnetism is now regarded
as an electrical phenomenon and is explained in terms of amperian currents.

*See Vol. I, Section 6-8.

12-10] INTENSITY OF MAGNETIZATION 257

We saw in Section 9-9 that the magnetic induction B due to a small

current circuit of area A, carrying a current I, was proportional to the
product I A) it was also pointed out that a large current circuit may be
thought of as built up out of many elementary, adjacent current loops,
each carrying the current I. Since, from the point of view of the resulting
field, the important factor related to the circuit is I A, we now define the
magnetic moment of a small current circuit as

Magnetic moment = IAnx, (12-33)

where nx is a unit vector normal to the plane of the circuit.

When a certain volume of a material is magnetized, we may follow
Ampere and picture permanently circulating amperian currents (or mag¬
netic dipoles) throughout the volume of the material. We define the
intensity of magnetization, M, as the total magnetic moment per unit
volume. If there are N amperian currents per unit volume, each with a
magnetic moment IA in the same direction, we have

M = NIAtlu (12-34)

which corresponds to Eq. (12-9) for the polarization P of a dielectric.

We saw that when the polarization varies throughout a medium, a
volume density of polarization charge p may arise. Similarly, when the
magnetization M is not uniform, a net current density of amperian currents,
which will be designated as J', may appear. We next seek the relation be¬
tween J' and the variation in M that corresponds to Eq. (12-13) relating
p' to the variation in P.
First consider a two-dimensional section of the medium, perpendicular
to M. Let us take all the amperian circuits to be squares of equal size, as
shown in Fig. 12-8. In Fig. 12-8 (a) there are equal currents in each circuit,
and as a result the currents cancel everywhere throughout the interior.
In Fig. 12-8(b) the amperian currents increase in strength toward the right,
with the result that they do not cancel on interior paths, but leave a net cur¬
rent, downward in the figure. Calling the plane of the figure the XF-plane,
M is outward, or in the Z-direction. In Fig. 12-8(b) Mz increases as we pro¬
ceed in the positive X-direction and the contribution to J' is in the negative
F-direction. Thus the component Jy is proportional to —dMJdx. In Fig.
12-8(c) we see that if Mz increases in the positive F-direction, we get a com¬
ponent of f' in the positive X-direction, or J'x « dMJdy. It does not matter
which directions we take for our axes, so long as we keep to a right-handed
system. Therefore, by changing x to y, y to 2, and 2 to x, we have
J'z oc —dMx/dy and J'y * dMx/dz. Similarly, J'x « —dMy/dz and
J'z a dMy/dx. Combining these results, we have
258 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

I' = 1' 7' = 1 'l' = 1

I' = 1 I' = 1 1' = 1

'
'
\
b-S
I' = 1 V = 1

II ’
X
(a)

I' = 2 I' = 3
^
o

II
II

’I' = 1 I' = 2 I' = 3

o
II

-►*>- -***-

0 x
(b)

Fig. 12-8. Amperian currents.

dMz dMy
J'x oc - — „ J
dy dz

dMx dMz
T' oc-— 7
dz dx

dMy dMx
J'z OC-^ — „ 7
dx dy
or
J' a v X M.

It remains to be shown that the proportionality factor is unity.

Let us consider a cube of the magnetized material of side l, with M
in the Z-direction. Then an amperian current I = IMZ circulating around
the side faces (Fig. 12-9) will, according to Eq. (12-34), give the required
12-11] CIRCUITAL FORM OF AMPERE’S LAW 259

Fig. 12-9. Amperian currents: three-dimensional view.

magnetization, since N — 1/Z3 and A — l2. If the magnetization changes by

dMJdy in unit distance in the F-direction, then around the adjacent similar
cube in this direction the amperian current will be l[Mz + (dMz/dy)l\,
or the change in amperian current will be l2(dMz/dy). It is this change
which constitutes the unbalanced current. It is not confined to the surface
which is common to both cubes. Actually the amperian currents are of
atomic size and we may consider the change in Mz to occur continuously
over an area A = l2 normal to the X-axis. Per unit area normal to the
X-axis an unbalanced amperian current of dMz/dy is distributed, giving
a current density J'x = dMz/dy, due to the variation of Mz with y. Simi¬
larly, the contribution to J'x due to the change in My in the F-direction
is —dMy/dz. Therefore, in general, we have

J' = V x M. (12-35)

12-11 Circuital form of Ampere’s law for magnetized media. Corre¬

sponding to Gauss’ law in electrostatics we have the ciicuital foim of
Ampere’s law, Eq. (9-24), for magnetostatics. The latter may be extended
to include magnetic media, as well as air or a vacuum, either (a) by adding
the term J' to the conduction current density J, or (b) by altering the
permittivity factor /xo to a value /x (/x X /xo) characteristic of the medium.
We start with method (a) and write

j) B • d'X = /xo JJ (J + JO ’ dS- (12-36)

x s

This amounts to saying that J' contributes to the magnetic induction in

the same way that J does. Substituting for J' from Eq. (12-35), we have
260 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

B • dl — y0 J■dS MO (V x M) • dS.

Next we apply Stokes’ theorem to the last term and transpose, getting

(12-37)

We define the magnetic intensity H for a magnetic medium as

H = — - M, (12-38)
MO

a definition in agreement with that given earlier for free space (where
M = 0). Equation (12-37) now becomes

H • d3i = JJ J • dS, (12-39)

x s

which is a more general form of Ampere’s law, good for any medium.
However, this equation is not of much practical value if we do not know
how H and B are related. At this point we shall introduce such a relation
and discuss its applicability.

12-12 The postulate relating B to H. From the magnetic viewpoint,

material media fall into three classes: (1) diamagnetic, (2) paramagnetic,
and (3) ferromagnetic. For (3) there is no functional relationship between
B and H, the value of B for a given H depending on the past history of the
specimen undergoing magnetization. We shall, therefore, limit our dis¬
cussion to diamagnetism and paramagnetism, for both of which B and H
are found to be linearly related (at least for moderate fields and ordinary
temperatures).
Let us put
B = mH. (12-40)

This equation defines y, the magnetic permittivity of the medium. The

statement that for a diamagnetic or a paramagnetic medium y is a constant
independent of H, is an experimental law of the restricted type. This
postulate is equivalent to saying that the intensity of magnetization M
is proportional to B or H, for from Eqs. (12-38) and (12-40) we have
12-12] THE POSTULATE RELATING B TO H 261

In terms of p and B Ampere’s law becomes

B • dX = n J j J • dS, (12-41)
x s

for a homogeneous medium.

Finally, we define the magnetic 'permeability Km as

Km = (12-42)
Mo

Ampere’s law may then be written as

j B • dk = Kmpoffj-dS, (12-43)
x s

which is a convenient form, since values of Km for different nonferromag¬

netic media are listed in tables. (Km equals 1.000004 for air and unity for
a vacuum.) Km is a dimensionless quantity, while p and mo are measured
in newtons/amp2, or webers/ampere-meter, B in webers/m2, and H and
M in amp/m, in the mksc system.
Illustration. Toroidal solenoid with a magnetic core. As in previous
illustrations, let us consider a thin ring solenoid of N turns, mean cir¬
cumference l, and cross-sectional area A, with current I in the winding.
However, now let us suppose that the core is a magnetic substance with
M ^ mo-
We find Eq. (12-36) a poor one with which to start because we do not
know either B or J’. Turning to Eq. (12-39), we take for the path X the
mean circumference l, along which the field is uniform, so that

j> H • dX = HI.
x

The right side of Eq. (12-39) is the total conduction current threading the
path, that is, NI. Hence we have

HI = NI,

H = —■ (12-44)

From this we see that for a toroidal solenoid, or a long straight solenoid, H
is equal to the ampere-turns per meter. We still cannot calculate B if m
262 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

is not known, but with a second winding connected to a ballistic galvanom¬

eter we might measure the change AB in B when the current in the primary
changes by AI. If, however the core material is nonferromagnetic, we may
use Eq. (12-43), which yields
I t

Bl = KmnoNI,

NI
B — Kmn0
l

Going back to Eq. (12-36), we have

Bl = mo NI MoJ' • dS,

or

JJj'.dS= (Km - 1 )NI. (12-45)

s

Within the core the field is uniform and the intensity of magnetization M
is constant in magnitude. Then J' = V x M is zero in the core. There¬
fore, the contribution to the left side of Eq. (12-45) must come from the
surface of the core, where the medium changes. We see that a current
circulating in the same sense as I, if Km > 1, or in the opposite sense if
Km < 1, and with the magnitude of (Km — 1) I per turn, will account
for the left-hand term.
It is instructive to compare this illustration with that at the end of Sec¬
tion 12-7. If we have two charged parallel plates, with a slab of dielectric
filling part (and air the rest) of the space between them, and if the free
charge density is a on one plate and — a on the other, the value of D is the
same in the dielectric as in the air; however, the value of E in the dielectric
(for which Ke > 1) is less than in the air. The polarization of the dielectric
produces on each surface of the dielectric slab a polarization charge opposite
in sign to that of the free charge facing it, with the result that the electric
field in the dielectric is weakened. In the magnetic case, on the other
hand, the magnitude of H is NI/l regardless of the medium, but the value
of B is greater in a core for which Km > 1 than in an air core with the same
number of ampere-turns per meter. The magnetization M adds to the
flux of magnetic induction when Km > 1.
Let us compare the values of Km for the three classes of magnetic media
mentioned earlier.
(1) Diamagnetism. We saw in Section 10-4 that when a magnetic field
is applied to a substance, the motions of the atomic electrons are altered
in such a manner as to produce magnetic moments in a sense opposite to
12-12] THE POSTULATE RELATING B TO H 263

that of the field. This effect, which is attributed to electromagnetic induc¬

tion, is called diamagnetism. While diamagnetism should occur in all
matter, it is a weak effect and may be obscured by other stronger ones
(paramagnetism, or ferromagnetism). In diamagnetic materials,

M < Mo, Km < 1,

so that M is opposite to H, but p is only slightly less than mo and Km only

a little less than unity. Such materials are strictly linear, or Class A
media; for any one of them p and Km are constant.
(2) Paramagnetism. In some substances the atoms or molecules possess
permanent magnetic moments, just as polar molecules occur in certain
dielectrics. In the absence of a field the individual moments are oriented
at random, but when a field is applied they are aligned to an extent limited
by thermal agitation. In magnetism this effect is called paramagnetism.
Paramagnetic materials are the magnetic analog of dielectrics with polar
molecules.
It is found that paramagnetic media are linear at ordinary temperatures.
For such media,
M > Mo, Km > 1.

The value of Km is temperature dependent. At very low temperatures the

disorienting effect of thermal agitation may be so small that a state of
complete alignment of the moments is approached and a saturation effect
encountered.
(3) Ferromagnetism. This is a property of a limited group of substances
including iron, cobalt, nickel, and certain alloys. For these B is not pro¬
portional to H, so that m and Km are not constant for a given material.
The magnetization is found to depend on the past history of the specimen
under study, as well as on the value of B or H. Such materials are called
ferromagnetic. In ferromagnetism the values for p and Km are frequently
much larger than in paramagnetism.
The explanation of ferromagnetism is based on a quantum-mechanical
phenomenon called exchange. According to this theory, which is beyond
the scope of this book, an alignment occurs between the spin directions of
the electrons in a given atom and the spin directions of the electrons in
those atoms nearest the given one. This results in a spontaneous mag¬
netization over a region containing many atoms. Such a region is termed
a domain. In1 an unmagnetized piece of iron the domains are oriented in
different directions and the resultant magnetic moment is zero; when a
magnetic field is applied to such a specimen, directions of spontaneous
magnetization in the direction of the field are favored over those in the
opposite direction.
264 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

12-13 Boundary conditions for magnetic media. Amperian currents

produce magnetic fields just as conduction currents do; both involve the
motion of electric charges. Hence whatever the cause of the magnetic in¬
duction, the divergence of B will be zero. As has been stated, lines of B
do not diverge from a magnetic source, nor do they converge toward a
magnetic sink, because magnetic poles do not exist as physical entities.
Therefore,
V • B = 0, (12-46)

whatever the medium. From this it follows that the normal component of
B is the same on each side of a boundary surface, or

Bln = B2n. (12-47)

The explanation is similar to that for Eq. (12-23), which followed from
the equation V • D = 0.
If we go back to Eq. (12-39) and use Stokes’ theorem to change the line
integral into a surface integral, we get

fj
s
(V x H) • dS = JJ]s’
• dS,

or, equating integrands,

V X H = J. (12-48)

This is the general form of the “point statement” of Ampere’s law. While
it applies to all media, it is valid only for a steady state (no displacement
currents). At a surface where J == 0, we find, in the same way as for
the electric case when V x E = 0, that the tangential component of H
is the same on each side of the surface, or

Hlt = H2t. (12-49)

12-14 Maxwell’s equations for material media. In Chapter 11 we

derived Maxwell’s equations (11-21) for a vacuum, and since Ke and Km
are very nearly unity for air, these may also be said to hold for air. Let us
investigate how these equations must be modified to apply to any material
medium.
Consider the first equation,
V • D = p.

In Chapter 11, D stood for e0E, so that this equation was equivalent to
Eq. (8-13), namely,
V •E = —•
12-14] maxwell’s equations for material media 265

From Eq. (12-22), we see that the equation V *D = p holds for any
medium so long as we understand that in general D equals eE and not e0E.
If we replace e0 with e in the equation for V *E, we get V -E = p/e.
Is this equation equivalent to V • D = p? The answer is “yes” only when
e is independent of the coordinates, such as at a point within a homogeneous
dielectric, but not at a point on its surface. (Why?) The correct general
form of Maxwell’s first equation is thus V • D = p.
Let us turn to Maxwell’s second equation,

V x E = -B.

This equation was derived from Faraday’s law. It is found exper¬

imentally that the induced electric field always depends on the time rate
of change of B rather than that of H*. The equation V x E = —B is
therefore perfectly general.
The third of Maxwell’s equations is

V -B = 0.

It was pointed out in the last section that this equation always holds true.
The fourth of Maxwell’s equations (11-21) was

V X H = J + b.

For a steady state this reduces to Ampere’s law expressed in the form

V x H = J.

In Chapter 11, H stood for B/p0. For magnetic media we get the same
equation provided that we take the relation between H and B to be
H = B/p rather than H = B/p0. What holds true in this respect for con¬
duction currents is also found to hold for displacement currents.
Maxwell’s equations for any medium then are:

V •D = p, V X E = —B,
(12-50)
V - B = 0, VxH = D|J.

From the definitions given for a, e, and p, we add to Eqs. (12-50) the
supplementary constitutive equations

* J = ctE, D = eE, B = pH, (12-51)

in which a, e, and p depend on the properties of the medium.

*For a discussion of the relative importance of B and H see Panofsky and

Philips, Classical Electricity and Magnetism, pp. 128-129.
266 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

For linear (Class A) media we may treat a, e, and p as constants and

proceed to derive the wave equation. For a medium in which p = J = 0,
the only difference the presence of the medium makes in the equations, as
compared with those for empty space, is that e replaces e0 and p replaces
PcJ. Thus in a linear medium other than empty space, when p = J = 0
we have for v, the velocity of propagation of electromagnetic waves,

1 (12-52)
v

wi1111■ for a plane wave we have, in place of Eq. (11-30),

eE2 = fiH2. (12-53)

We define the index of refraction n of a medium as

c
n (12-54)
v

that is, the ratio of the velocity of an electromagnetic wave in vacuo to its
velocity in the medium. For transparent media p generally equals p0, so
that in this case
n = Vc/ £o = y/K~e , (12-55)

a result in agreement with experiment for low frequencies where polariza¬

tion phenomena can keep step with the changing field.
The electromagnetic energy equation may again be derived from
Maxwell’s equations, using Eqs. (12-50). As before, the Poynting vector is

n = E x H, (12-56)

but for the energy density of an electromagnetic field the expression is now

Energy density of electromagnetic field = J(E • D +H-B). (12-57)

12-15 Summary of treatment of dielectric and magnetic media.

Dielectrics Magnetic media

B
V • e0E = p Vacuum (or air) —* V X —— J
Po
 PROBLEMS 267

Dielectrics Magnetic media

i

V • €oE = p ~\~ ?' Material vx — = J + J'

MO
media
p' = -V • P J' = V x M

V • («oE + P) = p
VX(^“M) = J

6oE + P = D ——M=H
Mo

V • D = p V X H= J
B
D = eE H
M

V • (eE) = p VX-=J
M

eo MO

B
V • Kee0E — p V X
KmPO
=J

Problems

1. Derive the expression for the re¬ 4. A point charge q is placed at the
sistance R equivalent to the resistances center of a small spherical cavity in an
Ri, R2, and R3 in parallel. infinite dielectric whose constant is Ke.
2. A circuit contains a resistance R, Find the induced polarization charge q'
self-inductance L, capacitance C, and on the wall of the cavity and show that
emf given by Soe1"*, all in series. Let I E in the dielectric is the same as it
be the current and q the charge on C, would be if q' were added to q and the
at any time l. (a) Write the equation medium were free space.
expressing the equality between the 5. A line of force representing E
applied emf and the sum of the poten¬ meets the interface between two di¬
tial differences across R, L, and C at electric media, of constants K1 and K2,
any instant. (b) Using I — dq/dt, respectively. The line is refracted,
obtain a differential equation for I and making an angle 61 with the normal in
compare it with the mechanical one for the first medium, and an angle 62 in
forced oscillations with damping. the second medium. Find an expres¬
3. How will the current I vary with sion relating di, 62, Ki, and K2.
the voltage V in a material for which 6. Two large parallel plates are 3 mm
(a) cr = <tq\/E/Eo, and (b) cr = apart and are maintained at a poten¬
aoE2 3/Eo, where op and Eo are con¬ tial difference of 600 volts. Between
stants ? the plates are two dielectric layers,
268 ELECTROMAGNETIC PROPERTIES OF MATERIAL MEDIA [CHAP. 12

one 2 mm thick with Ke = 3, and the permittivity varies with E according

other 1 mm thick with K e = 6. Find to the relation e = eo(l + 5e~E/E°),
(a) the capacitance per unit area of the where Eo is a constant, (a) How will
plates, (b) E and D in each medium, the capacitance C vary with V? (b)
(c) P in each medium, (d) a', the How will the work W required to
charge density on the interface be¬ charge the capacitor vary with V?
tween the two dielectrics. 10. The plates of a parallel-plate
7. Two parallel plates are 3 mm capacitor are 1 mm apart, each has an
apart and their potential difference is area of 1000 cm2, and they are kept
600 volts. Take the plates to be of unit connected across a 600-volt battery.
area and suppose that between half of Take the medium to be air (Ke = 1).
each plate the medium is air (Ke = 1), (a) What is the force between the
while between the other halves the plates? (b) If the plates are moved
medium is a dielectric with Ke = 2 apart until the separation is 2 mm,
(Fig. 12-10). Find (a) E and D in each how much mechanical work is done
medium, (b) the free charge density a and how much energy is returned to the
on each half of the plus plate, (c) the battery? (c) Check the conservation
capacitance C. (Neglect end effects.) of energy by also computing the change
in the energy stored in the field between
the plates.
11. Suppose that an electron is a
small spherical shell of mass m with a
charge e spread over its surface. Show
that the ratio of the magnetic moment
to the angular momentum of such an
8. Suppose that for negative values electron should be e/2m, whether the
of x the medium is a dielectric with electron is (a) moving in a circular
constant Ke, while for positive values orbit, or (b) spinning about a diam¬
of x the medium is air. A charge q is eter. (Experimental values for the
placed a distance d in front of the sur¬ above ratio are e/2m and e/m, re¬
face of the dielectric (the FZ-plane). spectively.)
Show that for x > 0 the field is the 12. A toroidal solenoid with 200
same as that in air due to the charge turns and a mean circumference of
q at x = d and another charge q1 = 1 m contains a core whose perme¬
—q(Ke — \)/(Ke-\- 1) located at the ability Km = 11 for H < 10 amp-
image point x = —d of q, and show that turns/m, but whose Km = 1 Ho/H,
for x < 0, the field is the same as it would with Ho = 100 amp-turns/m, when
be if the medium were entirely air with H > 10 amp-turns/m. Find H, Km,
a single charge q" = 2q/(Ke-\- 1) at B, and M when the current in the
x — d. (Show that at x = 0 the winding is (a) 0.02 amp, (b) 5 amp,
boundary conditions for the field are (c) 60 amp. (d) Find the ratio M/H in
satisfied and that the function for V each of these cases.
when x > 0 reduces to that for V 13. Derive Eq. (12-57).
when x < 0.) 14. Show that if e has different
9. Suppose that the medium be¬ values for different directions of E (as
tween the plates of a parallel-plate in a crystal), then e must be regarded
capacitor is not linear, but that its as a dyadic e and D = £• E.
 1

Part III
WAVES, PARTICLES, AND
RELATIVITY

#
 (

< >
 CHAPTER 13

PHYSICAL OPTICS

We have stated and developed the basic laws of thermodynamics and of

electromagnetism. In the present chapter our main concern will be with
formulating any additional fundamental law(s) that must be introduced
in explaining optical phenomena.
The subject of optics is usually divided into two parts, namely, (1)
geometrical, or ray, optics, and (2) physical, or wave, optics. Rays,
representing the direction of wave propagation and normal to wavefronts,
are convenient to use when we are not concerned with diffraction and
interference effects. This amounts to saying that in ray optics we assume
all apertures, lens diameters, etc., to be large compared with the wave¬
length. In discussing the refraction of light by prisms, or the formation
of images by mirrors and lenses, ray optics may be used quite satisfactorily.
However, except during emission and absorption, the essential property of
light is that it is a form of wave motion, hence wave optics must be regarded
as the more fundamental approach to the subject (the one which we shall
follow), with ray optics considered to be a special case.
The basic optical phenomena which we shall discuss in this chapter are
reflection, refraction, interference, and diffraction; dispersion and the finer
points of polarization are left to more specialized treatments. We shall first
show that the laws of reflection and refraction may be derived from rela¬
tions already developed in electromagnetism, although they also follow
from the general theory of wave propagation. When we come to inter¬
ference and diffraction we shall have to use the wave theory, for which the
fundamental postulate will be found to be a modification of Huygens
principle.

13-1 Reflection and refraction of electromagnetic waves. Electro¬

magnetic waves range in wavelength from the shortest gamma rays to the
longest radio waves, with visible light occupying only about one octave
of the whole spectrum, between wavelengths of about 4 X 10 7 m and
g x io-7 m. However, for simplicity we shall use the term light to
designate both visible and invisible electromagnetic radiations.
Since any small section of a wavefront or of a curved surface may be
treated as approximately plane, we shall consider the case of a plane wave
incident on a plane surface separating regions occupied by two different
media. The subscript 1 will refer to the medium containing the incident
wave, ex being the electric permittivity of this medium, Ml the magnetic
permittivity, vl the velocity of light in this medium, nx = c/vx the index
271
272 PHYSICAL OPTICS [CHAP. 13

of refraction, etc. Similarly, the subscript 2 will refer to the other medium
into which the light may penetrate. We shall take the interface between
the two media to be the plane x = 0.
The electric vector E has been shown to satisfy the wave equation

V • VE (13-1)

where

(13-2)

The magnetic vector H satisfies a similar equation, h or plane waves

proceeding in the X-direction the wave equation (13-1) becomes the
one-dimensional wave equation

<32E 1 d2E
(13-3)
dx2 V2 dt2

for which the solution was shown to be some function/(x ± vt). It has
been pointed out that any periodic wave may be considered to be built up
of harmonic waves, so we need only concern ourselves with the case where
the function / is a harmonic one, say

/ = Ae*rXx-vt)l\t (13_4)

where j = \/—l (in this chapter), and X is the wavelength. Changing this
to
Ae2* j(vt—x)/\
/ = (13-5)

makes no difference in steady-state problems (in which df/dt does not

appear), or if we are interested only in the real part of /. Introducing the
angular velocity w = 2ttv, where v = v/X is the frequency, we get

/ = AeMt-ix/v)]' (13_6)

Now the question arises as to how to write the corresponding function

for a plane wave proceeding toward an interface in a direction other than
the X-direction. We define the angle of incidence i as the angle between the
direction of propagation of the wave (the direction of an incident ray) and
the normal to the interface. Since the interface is the plane x = 0, we may
take the XF-plane to contain the normal and incident ray, as in Fig. 13-1.
This is called the plane of incidence. In Eq. (13-6) we must now replace
x with s, a distance along the ray. From the figure we see that s =
x cos i -(- y sin i, so that we now must write
13-1] REFLECTION AND REFRACTION 273

Fig. 13-1. Reflection and refraction.

(13-7)

for the incident wave.

For the reflected wave we call the angle between the ray and normal to
the interface the angle of reflection i'. From symmetry considerations we
see that the reflected ray could not favor any direction other than one in
the plane of incidence. Since its velocity has a component in the negative
X-direction, this wave may be represented by

(13-8)

where primes refer to the reflected wave.

Finally, suppose that some of the light penetrates into the second medium
as a refracted ray, making the angle r, called the angle of refraction, with
the surface normal. For this wave we write

A e30>2^ t~cos rJrV s'n V2^ (13-9)

At any point (0, y) on the interface the exponential factors in these

three functions must agree for all values of t and y, i.e., the functions must
agree in phase. If this were not so, no choice of amplitudes would satisfy the
boundary conditions at all times and at all places on the boundary. There¬
fore the coefficients of t and y must be the same in each exponential, or

COi = COi = W2, (13-10)

(13-11)

sm i sm r
(13-12)
274 PHYSICAL OPTICS [CHAP. 13

Equation (13-11) is the familiar law of reflection. If in Eq. (13-12) we put

Vi = c/wi, v2 = c/n2, we get

n\ sin i — n2 sin r, (13-13)

which is called Snell’s law of refraction.

13-2 Energy reflected and refracted at normal incidence. We turn

next to the relations between the amplitudes of the incident, reflected, and
transmitted waves. For simplicity we shall take the special case of normal
incidence. Since plane electromagnetic waves are transverse, with E
perpendicular to H and v in the direction of E X H, we may take E to be
in the F-direction, H in the Z-direction, and the wave to be propagated in
the X-direction. Then

Ei — |E ]. | = E\y, Hi = |H 11 = Hiz

are the amplitudes of the electric and magnetic vectors of, say, the incident
wave. (We shall continue to use primes for the reflected wave and the sub¬
script 2 for the refracted wave.)
In the last chapter (Eq. 12-25) we saw that at a boundary the tangential
component of E must be continuous. In medium 1 we must add the fields
due to the incident and the reflected waves, so we have

Ei -f- E'i — E2. (13-14)

Similarly, from Eq. (12-49),

Hi + H'i = H2. (13-15)

For plane waves (see Eq. 12-53) we found that

Ei I Mi
(13-16)
Hi ~ yjei ’

For the reflected wave either E or H must be reversed in phase, since the
direction of propagation, which is that of E x H, is reversed. If Ex and
H i are both positive, we have

(13-17)

For the wave transmitted in the second medium, we may write

E2
(13-18)
h2 ~
13-2] REFLECTION AND REFRACTION AT NORMAL INCIDENCE 275

Using these last three equations to eliminate the H’s in Eq. (13-15), we get
i

U3-19)

Between this equation and Eq. (13-14) we may eliminate E2, which yields

E'i _ Vei/m — V62/m2 _

(13-20)
■^'1 \/€l/Vl 4" V/62/m2
This is also equal to —H[/H1.
In most cases mi and M2 are very nearly equal to mo, the magnetic
permittivity of free space. Therefore in Eq. (13-20) we put mi = M2 = Mo-
Cancelling out the mo and using Eq. (13-2), we get

E\ _ ni — n2 Hi
(13-21)
Ei nx n2 Hi

for the so-called coefficient of reflection. We see that reflection occurs

without a change in phase for E when n\ > n2; in this case H undergoes
a 180° change in phase. Upon reflection, E changes phase when n2 > nx,
while H does not.
The incident energy per unit area of the boundary per unit time is pro¬
portional to E X H. The power per unit area is called the intensity, so that,
for a plane wave,

Intensity « EH, E2, or H2.

The ratio of the reflected power, and energy, to the incident power, and
energy, is
E'iH'i _ (nx — n2V
(13-22)
EiH} — \nx + n2)

In the same manner we may find the ratios E2/Ex and E2H2/EiHx.
The results are
E2 _ 2ni
E\ nx -f- n2
(13-23)
E2H2 __ 4win2
EXHX ~ (nx + n2)2 ‘

The last equation gives the fraction of the incident power, and energy, that
is transmitted into the second medium.
276 PHYSICAL OPTICS [CHAP. 13

Example. Given that n = 1.5 for ordinary glass. Find the percent of power
transmitted and the percent reflected for light incident normally from air to glass.

Solution. We have n\ = 1, n2 = 1.5, m — n2 = —0.5, n\-\- n2 = 2.5.

Hence
0
E2H2
= X 100% = 96% of power is transmitted,
ElHl O.Zu

e[h[
= ttp X 100% = 4% of power is reflected.
EiH! Zb

Thus a glass plate in air reflects only 4% of the light intensity falling normally
on its surface.

13-3 Amplitude relations for oblique incidence. We consider next the

theory of oblique incidence. We shall use the same sort of reasoning and
principles as were employed in the last section. We shall find that the
trigonometry which must be introduced makes the expressions more in¬
volved and that, in addition, the polarization of the waves is a factor which
must be considered. The important cases to consider are the following two:
(a) E is normal to the plane of incidence (and parallel to the interface), and
(b) E lies in the plane of incidence (and hence oblique to the interface).
We shall work out the theory for case (a) only and leave case (b) as a prob¬
lem.
Let us take the interface to be the plane x = 0 and Ei to be in the Z-
direction, as in Fig. 13-2. Then H: must lie in the plane of incidence with
H i sin i for its ^-component and —H x cos i for its y- or tangential com¬
ponent. E2 will also be in the Z-direction and H2 will have the tangential
component —H2 cos r. Let us take the case where E does not, and H does,

El

Ej

vi

Fig. 13—2. Oblique incidence: E normal to plane of incidence.

13-3] AMPLITUDE RELATIONS FOR OBLIQUE INCIDENCE 277

change phase upon reflection, as in the figure, so that the tangential com¬
ponent of Hi is H'i cos i.
Using the condition that the tangential component of E must be con¬
tinuous at the boundary, we have

Ex + E\ = E 2, (13-24)

and from the similar condition for H, we get

—Hi cos i + H'i cos i = — H2 cos r. (13-25)

For a plane electromagnetic wave in a medium where m = mo, we have

Hi = a/«i/mo Hi, H2 = H2, H'i = y/tx/mo

Substituting these values in Eq. (13-25), we find that

V~e[ (Ei — E'i) cos i = Ve2 E2 cos r. (13-26)

Between Eqs. (13-24) and (13-26) we may eliminate E2 and solve for E( in
terms of Ei. Let us at the same time put Vei/e0 — n-i and Ve2/e0 = n2.
The result is
n i cos i | j n i cos i i
= Ei
Ln2 cos r j [_n2 cos r J

Using Snell’s law, ni sin i = n2 sin r, this becomes

which may, by trigonometry, be written as

sin (r — i)
E i = Ei (13-27)
sin (r + i)

Note that when n2 > n\, and r < i, Ei and E\ have opposite signs, in¬
dicating a change in phase for E, as we found for normal incidence. The
case of grazing incidence, for which i = 90°, is interesting; for it we find
that E[ = Ei sin (r - 90°)/sin (r + 90°) = -Elt or the coefficient of
reflection equajs —1, which means that all of the light is reflected.
For case (b), where E lies in the plane of incidence, Fresnel found in a
similar way that
tan (r — i) _
E\ = Ei (13-28)
tan (r + i)
278 PHYSICAL OPTICS [CHAP. 13

An interesting situation arises in this case when r + i — 90°, that is, when
the incident and refracted rays are at right angles to each other. Then in
Eq. (13-28) the denominator becomes infinite and so E[ = 0. This means
that for light which is polarized so that E lies in the plane of incidence,
and which is incident at the angle i = 90° — r, no light is reflected and
all is transmitted. If unpolarized light is incident at this so-called polarizing
angle, then no reflected light will have its electric vector in the plane of
incidence, i.e., the reflected light will be completely polarized, with its
electric vector at right angles to the plane of incidence, or parallel to the
interface. It was this effect that led to the discovery of polarized light.
Equations (13-27) and (13-28) are known as Fresnel’s equations be¬
cause in 1823 Fresnel derived equivalent equations for transverse waves in
a hypothetical elastic ether. Of course, Fresnel did not know about elec¬
tromagnetic waves; what we called the electric vector E he thought of as a
displacement.

13-4 Fermat’s principle. The laws of reflection and refraction may be

concisely summarized in the form of a variation principle, first given by
Fermat.
The time required by a ray of light to traverse a path distance ds, in a
medium whose refractive index is n, is dt — ds/v = n ds/c. Since c, the
velocity of light in vacuo, is a universal constant, the time dt is always
proportional to n ds, the optical length of the path distance ds. We see that
in vacuo n ds is the same as ds, while in a denser medium it is greater than
ds. Fermat’s principle states that if we compare the optical length of the
actual path followed by light between the two given points, with the optical
length of a slightly varied path between the same two points, then the difference
This means that if we form
will be zero within first-order small quantities.
the first variation of the integral between two given points of n ds, then
this variation will be zero for the actual path, or

(13-29)

where the integration is taken between two points Q and P on the same
ray. It usually turns out that the optical length for the actual path is a
minimum, which also means that the optical path is the path of least time.
We shall prove this theorem for three cases, (a) a homogeneous medium,
(b) reflection, and (c) refraction.
(a) Homogeneous medium. Here n is a constant, so that

Q
13-4] fermat’s principle 279

Fig. 13-3. Fermat’s principle applied to reflection.

Fig. 13-4. Fermat’s principle applied to refraction.

In a homogeneous medium light rays are straight lines. Since a straight

line is the shortest distance between two points, the actual path is also the
path of minimum optical length.
(b) Reflection. This case may be reduced to (a) as follows. In Fig. 13-3
a ray of light from Q strikes the mirror MM at some point 0 and is reflected
through the point P. By the law of reflection, the path QO has the same
length as the distance Q'O, where Q' is the image of Q in the mirror, (it is
the same distance behind the mirror as Q is in front, and directly opposite
Q). Then the path length QOP equals the length Q'OP. Since the ray
actually travels in a single medium, the optical length is again n times the
path length and is a minimum when the path length is a minimum. If the
varied path is QO'P, then, since QO' = Q'O', the length of this path is
the same as that for Q'O'P, which is obviously greater than the straight
line distance Q'OP. Hence the optical length of the actual path is less than
that of the varied path.
(c) Refraction. Let the interface between two media (indicated by sub¬
scripts 1 and 2, respectively) be the plane a; = 0. In Fig 13-4, let QOP
be the actual ray between the points Q(xi, yi) and P(x2, y2)- e may a e
280 PHYSICAL OPTICS [CHAP. 13

the point 0, where the ray enters the second medium, to be our origin.
Let L be the optical length of this actual path. Then

L = niV x\ + y\ + n2V x2 + y\.

Suppose that the varied path crosses the interface at the point O', whose
coordinates are (0, by, bz). The optical length L' of this path will be

L' = 7iiV x\ + (yi — by)2 + bz2 + n2V x2 + (y2 — by)2 + bz2.

The variation bL in the optical length is L' — L. Expanding the radicals

by the binomial theorem* and neglecting terms in by2, bz2, and higher
orders, we get

5L = riiV x2 + y\ — 2yxby + n2Vx2 + y2 — 2 y2by

nxV x2 + y\ — n2V x2 + y\

n2y2
+ ,- sy
)
-

\V x2 + y22 V x2 + y2

= (—rii sin i + n2 sin r)by,

where i and r are the angles of incidence and refraction, respectively. (If
yi is positive, y2 must be negative, so that sin r = —y2/{x\ + y\)1/2.)
From Snell’s law, ni sin i = n2 sin r, so that

bL = 0,

which proves the theorem.

Later we shall sketch a derivation of Fermat’s principle based on
Huygens’ principle, using waves rather than rays.

1/2
*That is, Vx2 -J- y2 — 2yby = X^x2 -f- y2 1 2 yby
x2 + y2

= Vx2 + y2 terms with by , by , etc.

13-5] PRINCIPLE OF SUPERPOSITION; GROUP VELOCITY 281

13-5 Principle of superposition; group velocity/ According to the

principle of superposition, when two or more wave trains pass through the
same point, the resultant effect is the sum of the effects of the individual
waves. In computing this resultant effect, we must take account of the
respective amplitudes and phases of the individual waves. This principle
forms the basis of the theory of interference effects, such as the production
of standing waves. Here we shall consider another interference phenom¬
enon, that of group velocity, because of its application to wave mechanics.
The problem we shall be concerned with is the following. Two wave
trains, one of frequency v and the other of slightly different frequency
v -f- dv, are traveling in the same direction and are to be superimposed;
we want to find the behavior of the resultant wave train. The theory will
then be extended to a group of waves of continuously varying frequencies.
If each wave train has the same velocity of propagation v, here called
the phase velocity, then the waves may be added, as in the case of a Fourier
series, to give a new wave train with waves of a definite shape. This
resultant wave train will travel with the same velocity v and its shape will
not change with time. This leads to the phenomenon of beats.
Now let us consider the case where the phase velocity is a function of
the frequency, taking it to be v for the waves of frequency v, and v + dv
for waves of frequency v dv. This variation in v amounts to a variation
in the refractive index n = c/v with frequency, so that these two waves
would be refracted differently, leading to another optical phenomenon called
dispersion. Here, however, we are going to let the two waves be superim¬
posed in a given direction. At some instant t a crest of one wave train will
coincide with a crest of the other, giving a point of maximum disturbance,
but as time goes on these two crests will separate and the crest of the
first wave train will later coincide with a different crest of the second wave
train. This means that the point of maximum disturbance, or phase agree¬
ment, will not move with the same velocity as the phase velocity v. Since
signals are transmitted through the propagation of points of maximum
and minimum disturbance, the velocity of such points, called the group
velocity G, is of practical importance.
Let us assume that the two wave trains have the same amplitude, but
slightly different frequencies and phase velocities, so that we may represent
them by the equations

Ei = A cos aq(t — x/vi),

(13-30)
*
E2 = A cos w2(t — x/v2)

(see Eq. 13-6). Adding and using the trigonometric formula cos a+
cos /3 = 2 cos £(a + j8) cos £(« — 0), yields
282 PHYSICAL OPTICS [CHAP. 13

E = E\ + E2 = 2A cos |
-£)+?(■ -5)]
, x\ C02 / xy
cos t-)-[ t-)
L 2 V Vx) 2 \ v2 /_

We can write ui — co = 2ttv, o>2 = co dco = 2ir(v + dv), and o>i —

co2 = —dco. Also, since dco <<C co, we may say that, approximately, oq +
co2 = 2co. and, similarly, (coi/tq) + (W2A2) = 2co/p, where w — J(«q+i>2).
Then
dco 1 / C02
E = 2A cos co
T 2 \p2
Now
^ _ «i = d M,
P2 vx \vj
so that
„ 0 , dco d(u/v)
E = 2 A cos — cos CO (13-31)
dco

Equation (13-31) represents a wave, traveling with the phase velocity v,

whose amplitude is
dco X
2 A cos ?
T G_
where
dco dp
G =
d( co/p) d(v/v) ’
or
1 _ d(v/v)
(13-32)
G ~

We see that the amplitude itself is being propagated in the A-direction

with the velocity G, and since the amplitude velocity is the signal velocity,
this is the group velocity.
Signals can be sent only by initiating and ending a finite wave train. In
Fourier analysis such finite wave trains are analyzed as the superposition
of a band of frequencies. If the phase velocity v varies with frequency, we
see that the signal velocity will not be the same as the v of the carrier wave.
Usually d(v/v)/dv is nearly constant over the frequency range encountered,
so that when the waves are all superimposed the signal velocity will still
be G, as given by Eq. (13-32).

Example. For deep-sea waves, we find that v — \/~g\/2t, where g is the accel¬
eration due to gravity. Find the group velocity for waves with a small range of
frequency.
13-6] HUYGENS’ PRINCIPLE 283

Solution. Using the relation v = Av, we have

2
2 = gv V
2-jtv V g

1 d /v\ 47xv
G dv \v) g

Here the longer waves travel faster and, for a limited range in wavelength, the
group velocity is half the average phase velocity.

13-6 Huygens’ principle. We are familiar with the fact that sound
waves can pass around corners and that a speaker may be heard even when
he is not in the line of sight of the listener. Similarly, water waves bend
to some extent around the edge of a breakwater. This phenomenon, called
diffraction, is observed for light waves when they pass through apertures
not many times wider than the wavelength of the light, or when the edge of
the shadow cast by a sharp object is examined under a microscope. The
observed effects can be explained only in terms of wave theory, and in
doing so we must start with some law or postulate. This postulate is an
extension of the principle of superposition given in the last section, and it
was first stated by Huygens as follows.
Suppose that light travels out from a source and reaches all points on a
certain surface at the same time t. Then this surface is one of constant
phase for a given wavelength, and we call it a wavefront. We wish to
find the shape and position of this wavefront at a slightly later time
t + 8t. This may be done, according to Huygens, by assuming that at
l =z t every point on the original wavefront is a source of a secondary dis¬
turbance, which spreads out with the velocity v of the wave at that point,
forming spherical wavelets. After the time 8t these wavelets will have
radii of v8t and the envelope of these wavelets, Huygens postulated, will
be the new wavefront at the time t 8t. In justification of this procedure
it may be mentioned that points on different wavelets correspond to the
same phase, and that the envelope of these small spherical wavelets will
contain more such points of phase agreement than any other surface that
could be drawn. Thus, according to the principle of superposition, the
envelope will be4a surface of maximum reenforcement.
The laws of reflection and refraction may be derived from Huygens’
principle For example, in the case of a plane wave in medium 1 striking
the surface of a denser medium 2, as in Fig. 13-5, the wavelets will have
radii Vi8t = c8t/n\ in the first medium and v28t = c St/n2 in the second;
284 PHYSICAL OPTICS [CHAP. 13

Fig. 13-5. Huygens’ principle applied to refraction.

with n2 > rii, the wavelets travel more slowly in the second medium and
the new wavefront in that medium is more nearly parallel to the interface,
so that r < i. A quantitative treatment leads to Snell’s law.
There are, however, some objections to Huygens’ principle. First, it
leads to a backward as well as a forward wave because the secondary
wavelets have two envelopes, one on each side of the original wavefront.
Of course, we usually know which way the wave is progressing and so
can arbitrarily rule out the backward wave, but it is still a formal weak¬
ness of the theory. Second, Huygens’ principle does not tell us the relative
intensity at a point, in the case of diffraction.
Huygens’ principle was modified and extended to cover diffraction by
Young and Fresnel. They retained the idea that every point on a wave-
front may be regarded as a source of a secondary disturbance, but they
carried out the superposition of the secondary waves in a different way.
Rather than taking the envelope of the wavelets at the time t + 8t, they
considered the individual phase of each secondary wave as it passed, at
some definite time, through a given point. They postulated that the result
of adding these secondary waves, allowing for phase differences, determined
the actual disturbance at the point. This amounts to treating all the ele¬
ments of area of the original wavefront as interfering sources of light.
Simple applications of this method are given in elementary texts. Here
our task will be to derive a mathematical expression for diffraction which
embodies Fresnel’s modification of Huygens’ principle, but is even more
general in its approach. Before doing so, we must derive a purely mathe¬
matical theorem in vector analysis, which we shall need to use.

13-7 Green’s theorem. This is an extension of Gauss’ theorem. Accord¬

ing to the latter, if A is a vector function and S the surface surrounding the
volume r, then

(13-33)
s r
13-7] green’s theorem 285

Let us put A = uVv, where u and v are scalar functions of the coordinates.
We then have

uVv • dS (uVv) dr. (13-34)

In a similar manner, we let A = vVu, which yields

vVu•dS = V • (vVu) dr. (13-35)

s T

Now V • (uVv) = (Vu) • (Vv) + uV • Vv,

and V • (vVu) — (Vv) • (Vu) + vV • Vu,

as may be proved by expansion. We substitute these last two expressions

back into Eqs. (13-34) and (13-35) and subtract one equation from the
other, to obtain

vVu) • dS — Vv vV ■ Vu) dr, (13-36)

T

which is Green’s theorem.

In applying this theorem we shall take u and v to be the amplitude part
of functions U and V, each of which satisfies the wave equation, so that
77 2
v - vu = -zU,
c2
(13-37)
77 2 ..
v-w = \v,
cz
and
U = u(x, y, z)ejat and V = v(x, y, z)ejat. (13-38)

Combining Eqs. (13-38) and (13-37), yields the time-independent wave

equations
2 2
n w
V • Vu —-u

and (13-39)
nV
V • Vv =-5- v.
c2

Upon substituting these relations into the right side of Eqs. (13-36), we
find that
2 2
n co
(uV • Vv vV • Vu) dr (uv — vu) dr = 0.
286 PHYSICAL OPTICS [CHAP. 13

Hence, if u and v satisfy Eqs. (13-39), we have

(uVv — vVu) • dS = 0. (13-40)

Fig. 13-6. Derivation of the diffraction formula.

13-8 Kirchhoff’s diffraction formula. Suppose that light is being emitted

by a point source at Q in Fig. 13-6, and P is the point at which we wish to
compute the resulting disturbance at the time t. We shall take the source
to be monochromatic, emitting light only of the angular frequency w.
We take the function u in Eqs. (13-38) to represent the amplitude of the
disturbance at any point, and U will then have to satisfy the wave equa¬
tion. For v we shall take the function

e-jkr
v = (13-41)
r

where k = nw/c and r is the distance from P to a given point. The function
V = vejut will then also satisfy the wave equation (see problem 11-13);
while this is the solution for spherical waves spreading out from the point
r = 0, here we just consider v a mathematical function. Since v becomes
infinite at the point P, we must exclude this point from the region of integra¬
tion t in Green’s theorem. We therefore surround P with a very small
spherical surface S’ and take r to be the region between S' and some outer
closed surface S. Then the surface of integration in Green’s theorem must
include both S and S'. Remembering that the positive direction of the
normal is outward, we see that for S' the positive normal is toward P. Now
we may apply Eq. (13-40) to our present functions, and write
13-8] KIRCHHOFF S DIFFRACTION FORMULA 287

+ • dS' = 0. (13-42)
S'

The second integral, the one over the small spherical surface S' about P,
may be easily computed. Take the term

-jkr

-Vu • dS'. (13-43)

r

and assume u to be continuous, so that Vu is finite. Over this small

spherical surface, e~^kr/v will be constant and Vu nearly so. Hence the
expression (13-43) becomes
—jkr

- (Vu) p 4ttr2

and, as we let r approach zero, this term vanishes. Now consider the term

JJuV (^r) • dS'. (13-44)

As r —> 0, u up and dS' becomes 47rr2. Also, since e jkr/r is a function

of r only,

where rx is a unit vector in the r-direction, which means away from P

and opposite to dS'. The term (13-44) thus becomes

4:irUp.
Lim
r~*0

We may thus write Eq. (13 42) in the form

uf - b s
Vu — uV •dS. (13-45)
288 PHYSICAL OPTICS [CHAP. 13

Following Fresnel’s modification of Huygens’ principle, we want to ex¬

press the disturbance up at P in terms of the sum of all the secondary
wavelets contributing to up. However, it is not necessary to consider these
secondary wavelets as originating from an earlier wavefront. We may
generalize Huygens’ principle further, as follows. Take any closed surface
S surrounding P, as in Fig. 13-6. All of the light reaching P must cross this
surface. Therefore, we may consider each element of area dS of S to be the
source of a secondary wavelet starting out at the time t — r/v, where r
is the distance from dS to P, and reaching P at the time t. If S is not a sur¬
face of constant phase, then allowance must also be made for the differ¬
ences in time or phase of the light reaching the various points on S from
the source Q. If r' is the distance from Q to dS, then the secondary disturb¬
ance starting toward P from dS at the time t — r/v must be taken to have
the phase of the source at the time t — (r + r')/v. If it is understood
that we refer to retarded functions, as indicated, we may then find the
disturbance Mp at P by integrating the contributions from each element
of S. Looking back at Eq. (13-45), we see that it gives the disturbance at
P in terms of an integral over the surface S surrounding P. Hence this
equation is just the mathematical expression of the modification of Huy¬
gens’ principle outlined above. It is a special case of a formula first de¬
rived by Kirchhoff, and we shall call it Kirchhoff’s formula.
In diffraction problems, screens or other opaque objects block out a good
part of the light that might otherwise reach the point P from the source
at Q. In these cases we let the surface S include such screens, so that the
only part of S that can contribute to the disturbance at P is that which
coincides with apertures in the screens. For example, in the case of a
screen with a slit, the integration in Eq. (13-45) is carried out only over
the area of the slit. In performing this integration, we must know the values
of u and Vu at all points in the opening.
For the point source assumed, we may say that before the light from Q
strikes an obstacle it must consist of spreading spherical waves. For spheri¬
cal symmetry, we may write V • V in spherical coordinates and drop the
terms in d and <t>, so that the wave equation for spherical waves is

As may be proved by substitution (see problem 11-13), the solution of this

equation is

(13^6)

where to — 2-kv and k = u/v = 2ir/\. Taking out the time factor and
13-8] kirchhoff’s diffraction formula 289

remembering that we are letting r', rather than r, represent the distance
from Q, we have
A. — i kvr
u = — e , (13-47)
r
so that
_ , du , jk
Ae
—jkr'
V“ = r‘ 6? = 11

where r'x is a unit vector from Q toward dS. Similarly,

,—jkr
jk -jkr
rl y»2 r

When we substitute these expressions back into Eq. (13-45) we get the
terms r( • dS and rx • dS. Let r[ • dS = cos (n, r') dS, where (n, r') is the
angle between r( and the normal dS, the latter being outward on the side
away from P. Similarly, let rx • dS = cos (n, r) dS. We shall assume that

r' 7>> X and r » X,

so that 1/r2 « jk/r, and similarly for r'. This amounts to saying that
while the phase may not be considered constant over a narrow aperture,
the distances may. Substitution in Eq. (13-45) then yields

Up [cos (n, r) cos (n, r')] dS, (13-48)

which is Kirchhoff’s formula for the diffraction of light from a point source.
The factor j in Eq. (13-48) represents a constant phase shift of 90°.
This is unimportant, however, since we may start counting our time when¬
ever we wish. The quantity A depends upon the strength of the source.
The exponential factor is a phase factor; in it slight changes in r or r' may
mean a large change in phase, since k(r + r') = 2?r(r + r')/\ and X « r
or r'. However, the other factors may be treated as approximately constant
when the integration is taken across an aperture whose dimensions are small
compared with r and r'. In such cases, we may write

Up = #
2X TT
A
[cos (#, r) - cos («, r')] (/
JJ
dS. (13-49)

It should be pointed out that Eq. (13-48) is only an approximation. It

does not satisfy the differential equation (13—39), since it assumes that the
290 PHYSICAL OPTICS [CHAP. 13

disturbance at the aperture is what it would be if the screen did not exist
(Eq. 13-47), and that r and r' are large compared with A. Nevertheless,
Kirchhoff’s formula proves quite adequate for finding the disturbance up
at a point which is not too near the aperture.
Illustration 1. Variation with direction of the amplitude of the secondary
wavelets. As in Fig. 13-7, let Q be on the normal of a small aperture in a
screen, so thatr'x and dS are antiparallel and cos (n, r') = cos 180° = — 1.
Let 0 be the angle between iq and dS, so that

cos (n, r) — cos (n, r') = cos 0+1. (13-50)

Fig. 13-7. Variation of intensity with angle in diffraction.

When P is on the opposite side of the screen from the source and in line
with the source and aperture, cos 0=1 and the factor in Eq. (13-50) has
the maximum value of 2. On the other hand, if P is taken to be between
Q and the aperture and we consider only the secondary wavelets and not
the direct disturbance from Q, cos 0 = — 1 and the cosine factor is zero.
The backward wave is thus ruled out, and our theory also tells us how the
amplitude of the secondary wavelets varies with the direction, i.e., the
angle 0. The objections to the original Huygens’ principle have thus been
removed.
Illustration 2. Fresnel zones. We turn now to the phase factor e~ik(-r+r'\
Suppose that we have a monochromatic point source Q on one side of a
plane screen, with the point P on the other side, and that the straight line
connecting Q and P intersects the screen at 0 (Fig. 13-8). For the line
QOP the sum r + r' will have a minimum value which we shall call R.
Let us draw on the screen about the point 0 a curve such that for every
point on this curve r + r' = R + A/2, where X is the wavelength. Let us
next draw the curve for which r + r' = R + 2X/2, then the one for which
r + r' = R + 3X/2, and so on. These curves will be circles if the screen
13-8] kirchhoff’s diffraction formula 291

is perpendicular to QP, otherwise they will be the intersections of the

screen with a series of ellipsoids of revolution. These curves will divide the
screen into ring-shaped areas which are called Fresnel zones.
Now let us introduce an aperture into the screen. The disturbance at P
will depend on what zones are uncovered by this aperture. Let us call
an the magnitude of the contribution to Up due to the nth zone. The con¬
tributions due to successive zones will be 180° out of phase, or opposite
in sign. We shall consider the following cases.
(a) The screen is removed. All zones are uncovered and

up — a\ — «2 + a3 — «4 + • • ' • (13-51)

As n increases toward infinity the contributions of successive zones must

gradually decrease toward zero. We may group the terms in Eq. (13-51)
as follows:

(13-52)

Since the successive a’s differ only slightly in value, the quantities in
parentheses are very small and may be neglected, so that we have

ai
Up = T'
The disturbance at P for no screen is half that due to the first zone acting
alone. «
(b) The aperture in the screen uncovers just the first zone, then

Up = <Xi,

and the disturbance is twice that without a screen!

292 PHYSICAL OPTICS [CHAP. 13

(c) The aperture uncovers n zones of nearly equal size. If n is odd, we

may write
ai , /Ul I a3\ I | an

Up = A+ \2 - “2 + 27 + " • + V ’
while if n is even, we can put

Since a\ — (a2/2) ai/2 and the terms in parentheses contribute little,

we have
Up =

the plus or minus sign depending on whether n is odd or even. Hence

the disturbance at P equals half that due to the first and last zones to¬
gether. As successive zones are uncovered, the intensity at P will wax and
wane.
(d) Portions of n zones are uncovered. In this case the contributions
due to successive zones may change so rapidly that terms such as

/ Qre—1 ^n + l\
+
V 2
O-n
2 )
cannot be neglected. We must evaluate each an and use Eq. (13-51).
The contribution of a zone depends on (1) its area, (2) its distance, and (3)
the obliquity factor cos (n, r) — cos (n, r').

13-9 Derivation of Fermat’s principle based on the modified Huygens

principle. We can now obtain a more general proof of Fermat’s principle.
Again suppose that light travels by the actual path QOP from a source at
Q to the point P, 0 being any point on the path. Let us take the surface S
to include 0. Assume that the dimensions of any apertures present are so
large that diffraction may be neglected. Consider a pencil of rays going

Fig. 13-9. Fermat’s principle based on the superposition principle.

 PROBLEMS 293

from Q to P, but intersecting S at points O', 0", etc., all close to 0 (Fig.
13-9). We may, if we wish, imagine S to be a screen with a small hole
just permitting the pencil of rays to pass through. According to Fermat’s
modification of Huygens’ principle, all of these rays must reenforce one
another at P, which they will do if their paths all have nearly the same
optical lengths, thus assuring close phase agreement. This means that the
first variation in the optical length, as we go from the path QOP to the path
QO'P, or to QO"P, etc., must be zero, or

nds = 0,
as in Section 13-4.

Problems

1. Show that total reflection occurs the prism, then sin J(A -)- D) =
for ni > ri2 and sin i > U2/ni. Dis¬ n sin \A.
cuss for this case (a) the angle r, (b) 8. For ripples on the surface of a
the damping of the transmitted wave, liquid the phase velocity is given by
and (c) the Poynting vector for the v = (2tT/p\)1/2, where T is the sur¬
transmitted wave. face tension, p the density of the liquid,
2. Show that for reflection at normal and X the wavelength of the ripples.
incidence H[/Hi = —E[/E\. Find the group velocity for waves with
3. Derive Eq. (13-23) and show a small range of frequencies.
that the fraction of power trans¬ 9. How must the phase velocity
mitted plus the fraction reflected equals depend on the frequency for the group
unity. velocity to be infinite?
4. Show that for oblique incidence 10. By expanding both sides, prove
and E normal to the plane of incidence, that
E2 = 2Ei sin r cos I'/sin (r -f- i). V • (uVo) = (Vm) • (Vv) + uV • Vv.
5. Derive Eq. (13-28).
6. In terms of optical path, discuss 11. Plot up against 6 graphically,
the focusing effect of a double convex using polar coordinates with up as the
lens. radial coordinate.
7. A ray of light enters, passes 12. Consider a source emitting light
through, and emerges from a prism of wavelength X to be very far in front
whose vertex angle is A (see Fig. of a plane (that is, r' = °o), and let P
13-9). Inside tlje prism the index of be a distance l behind the plane, (a)
refraction is n and outside it is unity. Find, in terms of X and l, the outer
The ray travels parallel to the base in radii of the 1st, 2nd, 3rd, . . . , nth
the prism. Use Fermat’s principle to Fresnel zones on the plane, (b) How
show that if D is the angle through do the width and area of a zone depend
which the ray is deviated because of on n? Take nX <K l.
294 PHYSICAL OPTICS [CHAP, 13

13. Discuss the relative intensity at is a circular disk centered on the line
P in problem 12 if the zone pattern is QP, then, if the diameter of the disk is
gradually uncovered by moving to one small compared with r' and r, the value
side a semi-infinite screen that origi¬ of up is the same as if the disk were
nally covers the zonal plane in front not present.
of P' 16. Take Fresnel zones on a spheri¬
14. Take the zonal plane in problem cal surface of radius r' about Q, the
12 to be a screen with a circular hole distance QP being R. Show that
centered on the line QP. Discuss the Up = iai, or half the contribution
relative intensity at P as the radius from the first zone. Compute this con¬
of the hole is gradually increased. tribution by integration, showing that
15. Show that if the only obstacle up = Ae~jkR/R, as it should for a
between the source Q and the point P spherically spreading wave.
 CHAPTER 14

THE SPECIAL THEORY OF RELATIVITY

The fundamental laws of physics that we have considered so far all date
from before the year 1900, whereas the principle of relativity and the
quantum principle were only discovered in the present century. The
physics of the period prior to 1900 is frequently referred to as “classical
physics” and we have seen that it succeeded in discovering the laws and
building the theories in terms of which the ordinary occurrences of our
physical world can be explained. The reason that so much could be accom¬
plished without the use of the relativity and quantum principles, is that the
speed of light c is so great and Planck’s constant h is so small. In fact, if
the speed of light were infinite, those effects peculiar to the relativity theory
would not occur and, of course, if h were zero it (h) would be of no im¬
portance. As it is, the speeds of material objects in everyday life are very
much less than c and the uncertainties in their action* are much greater
than h. We shall see that when the expressions of the relativity theory or
quantum theory differ from those of classical physics, the former reduce
to the latter when c is relatively infinite and h is relatively zero, so that the
newer physical theories embrace the older ones as a special case; this case,
however, includes most common occurrences.

14-1 Inertial systems. By an inertial system we shall mean a frame of

reference, such as X-, Y-, Z-axes, relative to which Newton’s laws of motion
take their customary form. Since the earth is rotating it does not con¬
stitute a true inertial system, as witness the Coriolis acceleration of
winds, but in laboratory experiments the effects of the earth’s rotation are
negligible and the ground or walls of a building serve sufficiently well for
an inertial system. Newton believed an ideal inertial system to be one that
is absolutely at rest. A large group of so-called “fixed stars” are relatively
at rest, and these are sometimes taken as defining such an absolute system.
However, we shall see that the term “absolutely at rest” has no meaning,
since all motion is relative. The important point is that if we can find a
system S in which Newton’s laws hold true, then any system S which
moves with uniform linear motion relative to S will also be an inertial

*See Vol I p 193. The action between two states of a system is defined as the
time integral between the states of twice the kinetic energy; it has the dimensions
of the product of energy and time.
295
296 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

system. Hence an infinite number of inertial systems are possible and, at

least in mechanics, the question of which one is absolutely at rest is un¬
important.

14-2 The postulates of the special theory of relativity. We have noted

that Newton’s laws do not change their form (they are invariant) when a
transformation is made from one inertial system to another. * The exten¬
sion of this invariance property to the other fundamental laws of physics,
particularly those in electromagnetism, was one postulate made by Ein¬
stein in formulating his special theory of relativity. This theory is re¬
stricted to reference systems moving relative to one another with constant
linear velocity only. The subject of accelerated systems is dealt with in
the general theory of relativity, with which we shall not be concerned. In
his special theory Einstein also assumed that the speed of light in vacuo
is always the same, and hence is a universal constant. The two funda¬
mental postulates of the special theory of relativity may be stated as
follows:
I. The fundamental laws of physics have the same form for all inertial sys¬
tems (i.e., for all reference systems at rest or moving with constant linear
velocity relative to one another).
II. The speed of light in vacuo as measured relative to any inertial system
is always the same.

We shall see that the second postulate states what has been found to be
an experimental fact. The basis of the first postulate is that it leads to de¬
ductions which have also been found to be true. The two postulates to¬
gether tell us that there is no physical experiment that can be performed
relative to one inertial system only by means of which the uniform velocity of
that inertial system relative to another inertial system can be detected. This
is Einstein’s principle of relativity.
Suppose, for example, that we are enclosed in a large box without
windows, like a freight car or an elevator, and that this box can move
smoothly and quietly relative to the ground. If the motion is uniform,
we will not be able to tell when our box is moving. If, however, the motion
is accelerated, it would be immediately detectable, since, when a car speeds
up, we are accelerated backwards relative to the car. Similarly, when a
car is accelerated centripetally on a curved track, we appear to be thrown
toward one side of the car. In an elevator we can detect the speeding up
and slowing down by the apparent change in our weight, or the force of
gravity.

*See Vol. I, p. 67. We shall see that in relativistic mechanics the. invariance
of F = d{mv)/dt is maintained by postulating that mass varies with velocity.
14-3] SPACE-TIME AND FIELD TRANSFORMATIONS 297

14-3 Space-time and field transformations. Applying Einstein’s first

postulate to electromagnetism, we conclude that Maxwell’s electromagnetic
equations must be invariant when we make a transformation from one in¬
ertial system S to another system S'. Let x, y, z, and t refer to the positions
and time at which the field vectors E, B, etc. are measured in the system
S, while x', y', z', and t! refer to the same points and events in S'. At first
we might expect that t' = t, but we shall see later that this is not possible
except at one instant, say t = t' = 0. Now consider the equation

V X E = -B, (14-1)

the X-component of which, in the system S, is

dE z dEy 9BX
dy dz dt

The invariance principle requires that this expression must transform into
the equation
V' X E' = -B', (14-2)

the X'-component of which is

dE'z dE'y dB'x

dy' dz' ~ dt'

Similar transformations must hold for all of the electromagnetic equations.

Some time before Einstein announced his theory of relativity, H. A.
Lorentz tackled the problem of finding a set of transformation equations
between x, y, z, t and x', y', z', t', and between E and B, on the one hand,
and E' and B', on the other, such that Maxwell’s equations would be in¬
variant in form. As in Fig. 14-1, let
us take the axes X', Y', Z' of S' to
be parallel, respectively, to the axes
X, F, Z of S, the two origins coincid¬
ing at t= t' = 0. Let S' be moving in
the X-direction with uniform veloc¬
ity v relative to S. In accordance
with Newton’s laws and disregard¬
ing relativity, we would naturally
assume that the transformation
equations for the coordinates and
time are

x x vt, y y, (14-3) Fig. 14-1. Two inertial systems in

z> — z, t' = t, relative motion.
298 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

which are called the Newtonian (or Galilean) space-time transformation equa¬
tions. Referring to Fig. 14—1, we see that after a time t the Y'Z'-plane
has moved a distance vt, so that the rc’-coordinate of a point would appear
to be vt less than the ^-coordinate. Lorentz found, however, that with
these simple transformation equations the invariance of Maxwell’s equa¬
tions could not be maintained, so he abandoned Eqs. (14-3). Working on
the basis that Eq. (14-1) must transform to Eq. (14-2), and so on for each
of Maxwell’s equations, Lorentz arrived at the following expressions, which
we shall later derive from the principle of relativity:

. x — vt
x' — — j
Vl — v2/c2

y' — y, *' = 2, (14-4)

^ t — vx/c2
Vl — V2/c2

These are called the Lorentz space-time transformation equations. Setting

k =
VT~— v2/c2

Lorentz also deduced the relations

E’x = Ex, B' = B,

E'y = k(Ey — vBz), By-k{By VE Z/C ), (14-5)

E'z = k(Ez T vBy), B'z = k(Bz - vEy/c2),

which are the Lorentz field transformation equations. He also derived addi¬
tional transformation equations for charge and current density, subject
to the conservation of charge principle; but in order not to complicate
matters further, these equations will not be given here. In other words,
let us take the case where p = J = 0.
As for many expressions in physics, it is much easier to test for the
correctness of Lorentz’ equations than it was to derive them initially. If
the Lorentz transformations are correct, then Maxwell’s equations for
empty space,
V • E = 0, VxE = —B,
(14-6)
V • B = 0, V X B — e0/i()E,
14-3] SPACE-TIME AND FIELD TRANSFORMATIONS 299

should transform into

V'-E' = 0, V' x E' = —B',

(14-7)
V'-B' = 0, V' X B' = eoMoE',

where V' = i'd/dx' + j'd/dy' + k'd/dz' and B' = dB'/dt', etc. We shall
apply this test to the Lorentz equations by checking the first of Eqs.
(14-7); we shall leave the checking of the others to be done as problems.
We wish to show that Eqs. (14-4), (14-5), and (14-6) are consistent with
the equation
dE'x dE'y dE'z
(14-8)
dx' dy' dz'

If we solve Eqs. (14-4) for x, y, z, t in terms of x', y', z', t', we obtain
the relations

x = k(x' + vt'), V = y', z = z', t = k(t' + vx'/c2),

(14-9)

which show that motion of S' in the X-direction with velocity v is equiva¬
lent to motion of S in the X'-direction with velocity —v. Differentiating
Eqs. (14-9) yields

dx dx dx , dt _ kv
dy' dz' dt' ~ kV> dx' ~ c2

etc., so that
d _ d dx_ d_dy_.d_dz_,d_dt_
dx' ~~ dx dx' + dy dx' + dz dx' dt dx'
300 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

Using these relations and Eqs. (14-5), Eq. (14-8) becomes

£ ^ dEx V dEx
0.
dx ' c2 dt / \ dy dy / \ dz
I •

Since V • E = dEx/dx + dEv/dy + dEz/dz = 0, we have

dB z dBy 1 dEx
— CqMO Ex
dy dz c2 dt

which is the .XT-component of the fourth equation in (14-6). Hence Eq.

(14-8) is consistent with Eqs. (14-6) and the Lorentz transformations.
The invariance of Maxwell’s equations assures the invariance of the
laws of electromagnetism. For example, an observer in the system S' will
find that Coulomb’s law holds for two charges q[ and q'2 at rest in S', or

TV 1 ?1?2

F “ 4^ V5"11'

Consider the following situation. The system S' is moving relative to

the system S and the charge q' is at rest in S'. Suppose that q' is an isolated
charge and that observers in both S' and S investigate the field due to this
charge alone. The observer in S' will find that q' produces a radial and
spherically symmetric field E', but no magnetic field (B' = 0). The ob¬
server in S will observe a radial, but not spherically symmetric, electric
field E and, in general, a magnetic field B. To understand why this is so,
let us turn to Eqs. (14-5), put B'x = B'y = B'z — 0, and solve for
Ex, Ey, Ez, Bx, By, and Bz. We find that in the present example

Ex = E'x, Bx = 0,

Ey = kE'y, By — ~kvE’z / ^ ,

Ez = kE'z, Bz = kvEy/c2.

Suppose that each observer is at a point for which E'y = E'x and E'y ^ 0.
Then Ey = kEx and Bz ^ 0.
We find that the way in which an electromagnetic field is separated into
electric and magnetic parts depends on the motion of the observer’s frame
of reference.
We have seen that from the mathematical standpoint the Lorentz
transformations keep Maxwell’s equations invariant. There are, however,
other consequences. (1) We note that t' ^ t and that t' is a function of x
as well as of t. Thus the value of t' depends on where an observer in the
14-4] THE MICHELSON-MORLEY EXPERIMENT 301

system S' is located. In relativistic dynamics (unlike Newtonian dynamics)

space and time must be regarded as interrelated. (2) Newton’s law of
motion, which states that force equals time rate of change of the product
of mass and velocity, is no longer invariant under a transformation from
aS to S' unless the mass of a particle varies with its velocity. We shall
consider these points in more detail in subsequent sections. Next, how¬
ever, we shall present experimental proof of Einstein’s second postulate,
the principle of the constancy of the speed of light, and then we will show
how this postulate also leads to the Lorentz space-time transformation
equations.

14-4 The Michelson-Morley experiment. To make the meaning of this

experiment clearer, the following analogy is frequently presented.
Suppose that in Fig. 14-2, A and B are two points on opposite banks of a
stream which is flowing in the X-direction with the speed v relative to the
ground. Let l be the width of the stream and let C be a point a distance l
upstream from A. We imagine a race between two identical swimmers,
each of whom swims with the speed c in still water, one swimmer being told
to swim across the stream to B and back to A, while the other must swim
upstream to C and then back to A. How will the times of the two swimmers
compare?
Consider the first swimmer, who must cross the stream. In order to
allow for the drift current v, he must aim diagonally upstream so that the
component of his effective velocity across the stream, say ce, is only one
component of the vector c, the other being equal and opposite to the current
v. Then

ce = Vc2 — V2.

The distance across and back is 21,

hence the time ty for this swimmer
will be

h = 21 - = —a (14-10)
Vc2 — v2 c Fig. 14-2. Race between two swim¬
mers in a flowing stream.

where k = l/\/l — w2/c2> as before.

The other swimmer makes slow progress upstream, his effective speed
being c — v, bht he has the current with him on the return leg, when his
resultant speed is c + v. His total time £2 wiH thus be

2 Ic 2 ir
(14-11)
h c — v C + V
302 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

We must take v < c, or the second swimmer could not reach the point C.
Then k > 1 and hence t2 > tx. The first swimmer wins, but the differ¬
ence in times, involving only terms in the square and higher powers of v/c,
will be small if v « c. Note that by measuring tx, t2, and l we can de¬
termine the drift velocity v, if we know c.
Let us assume that the two swimmers are accurately timed. Suppose
(this is a purely hypothetical case) that the two times were found to be
exactly the same, what explanations might be offered? We might conclude
that (1) the ratio v/c is zero and k = 1, within the limits of measurement,
or (2) that the distance from A to C was inaccurately measured and that it
is actually less than the distance across the stream. Conclusion (l)Jmplies
either no appreciable current (v = 0), or very fast swimmers (c —» oo).
Conclusion (2) would necessitate a shortening of the distance AC to l/k,
where l is the width of the stream; with l/k in place of l in the expression
for t2, we see that we would have t2 = tx.
The reason that tx and t2 are different is because the current, which
carries the swimmers along, affects them differently. In the same way,
sound waves are carried by the air and, relative to the ground, travel
faster with the wind than across or against the wind. In the time of
Maxwell much thought was given to the question of whether there was
any medium which carried light. Some physicists suggested that there
was such a medium, pervading all space, which they called the ether, and
they regarded it as an absolute frame of reference. If such an ether existed,
the earth could not always be at rest relative to the ether, since in six
months’ time the motion of the earth in its orbit is reversed. Suppose that
the earth is moving through the ether, then this relative motion, called
ether drift, should affect the motion of light just as the current of the stream
was seen to alter the effective speeds of the two swimmers.
Maxwell suggested a crucial experiment for testing the ether theory.
This experiment was performed by Michelson, who was shortly aided by
Morley, and it is called the Michelson-Morley experiment. The apparatus
employed was essentially a Michelson interferometer floating on a pool of
mercury. The interferometer, diagrammed in Fig. 14-3, is the optical
analog of the case of the stream and the two swimmers. A beam of light
from a source Q strikes a half-silvered glass plate G, so that part of the light
is reflected to the mirror M x and part passes through G and is reflected by
the mirror M2. Upon returning to G, the two beams are again partly re¬
flected and partly transmitted. The transmitted part of the beam from
Mi joins the reflected part of the beam from M2 and the combined light
passes into the telescope T. The optical lengths l of the two paths are made
the same. Then if there is an ether drift v parallel to GM2, the light from
M2 will be, with respect to phase, a little behind that from Mx. A pattern
of interference fringes is observed through the telescope T.
14-4] THE MICHELSON-MORLEY EXPERIMENT 303

The crucial part of the experiment comes when the whole apparatus is
rotated 90° about a vertical axis, so that any ether drift originally parallel
to GM2 will now be parallel to GM i, causing a shift in the relative phases of
the two combining beams. During the rotation the observer must watch
for a resulting shift in the interference pattern seen through the telescope.
Taking for v the earth’s orbital velocity of 30 km/sec, Michelson and
Morley computed for their apparatus a shift of about half a fringe, ac¬
cording to the ether theory. Actually they found no significant shift, al¬
though they believed that a shift of 1/100 of a fringe could have been
observed. This experiment has been repeated many times since Michelson
first attempted it in 1880, and it is now generally agreed that the earth’s
motion through space cannot be detected as a motion through the ether.
As in the case of the two swimmers, the null result in the Michelson-
Morley experiment may be interpreted as indicating (1) that the drift
current is effectively zero, or (2) that distances parallel to the drift current
are effectively shorter by a factor 1/k than they appear to be to an ob¬
server moving with the apparatus. As we have said, the earth moves in dif¬
ferent directions at different times of the year, and therefore if there were
an ether the earth could not always be at rest relative to it. Conclusion (1)
then necessitates giving up the ether drift hypothesis. The Irish physicist
Fitzgerald suggested possibility (2), which amounts to the following.
Suppose that an observer measures the length of an object to be V when
he is moving with the object, but he finds the same length to be l = V/k
when the object is moving longitudinally with a speed v relative to the
observer In the Michelson-Morley experiment the observer is moving
with the apparatus and he finds that the V of the longitudinal path is equal
to the l of the transverse path. An observer at rest relative to the fixed
stars would say that the longitudinal path is shorter. Lorentz, who was
304 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

trying to work out a theory of matter based on the dynamics of a spherical-

shell-type of electron, came to the same conclusion. In fact, a longitudinal
contraction of an object when moving relative to the observer is a conse¬
quence of the Lorentz space-time transformation equations, as will be
shown later. Since any measuring stick placed alongside the Michelson
apparatus will undergo the same relative contraction, an observer moving
with the apparatus could measure the motion of the apparatus relative to
his reference system only and not relative to the ether. So whatever con¬
clusion we draw, we see that the concept of an absolute ether is useless,
since an ether drift can never be observed.
On the positive side, we may adopt Einstein’s first postulate and accept
the principle of the invariance of the electromagnetic equations and the
resulting Lorentz-Fitzgerald contraction. Or we may simply say that the
results of the Michelson-Morley experiment prove that the speed of light
is independent of the motion of the source or the observer. This conclu¬
sion is Einstein’s second relativity postulate and, of course, it must be con¬
sistent with his other postulate, the invariance principle. To show that
this is so, we shall now derive the Lorentz space-time transformation and
the resulting Lorentz-Fitzgerald contraction, starting from Einstein’s
postulate of the constancy of the speed of light.

14-5 Relativity derivation of the Lorentz space-time transformation.

Let the inertial system S' be moving relative to S with the speed v in the
X-direction. Let the axes coincide at the time t = t' = 0 and, starting
at this time, let a spherical wave spread out from the origin. At the time
t = t this wave will reach some point P whose coordinates are x, y, z. Let
the wave reach P at the time t' in the S' system and call the coordinates
of P in this system z', y', z'. Then we have

_ (x2 + y2 + zy12
c~ t

x2 + y2 + z2 — c2t2 — 0. (14-12)

Similarly, we may write

x'2 + y’2 + z'2 — c2t'2 = 0, (14-13)

where the c is not primed since, according to Einstein’s postulate, c is the

same in S' as in S.
From symmetry considerations it seems reasonable to try putting y = y'
and z = z'. Combining Eqs. (14-12) and (14-13), we then get

x2 — c2t2 = x'2 — c2t'2. (14-14)

14-5] LORENTZ SPACE-TIME TRANSFORMATION 305

For the transformation equation relating x' to x we assume

x' = K(x — vt), (14-15)

where K is independent of x and t. The reasons for this “guess” are (a) a
linear relation is the simplest and should be tried first, and (b) the relation
must reduce to that given by the Newtonian transformation x' = x — vt,
when »«c. We shall see that the linear relation is sufficiently general to
satisfy all conditions and that it does reduce to the Newtonian relation
when v <3C c.
Since motion is relative, we may also regard S as moving in the — in¬
direction with the speed v, in which case we must write

x = K'{x' + vt').

Eliminating x' between this equation and (14-15) and solving for t' yields

t' = K (14-16)

In Eq. (14-14) we substitute for x' from Eq. (14-15) and for t' from Eq.
(14-16), to obtain

c2t2 K2(x2 - 2xvt + v2t2)

2 xt
+ c2K2
v V KK'J v2 V

Since this equation must be an identity, it must hold for all values of x
and t. This means that the coefficients of the x2, xt, and t2 terms must
vanish separately, or

K‘
0 it
KK h) = 0’
(v2 - c2)KK' + c2 = 0,

and
-c2 - K2v2 + K2c2 = 0.

From the last two equations we see that

1
K = K' = k.
y/1 — w2/c2
306 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

These values of K and K' also satisfy the first conditional equation.
Equations (14-15) and (14-16) then become

x' = k(x — vt),

% •

t' — k(t — vx/c2),

which are the same as the Lorentz equations (14-4). Thus these equations
are consistent with each of Einstein’s postulates for the special theory of
relativity.

14-6 The Lorentz-Fitzgerald contraction. Suppose that we have a rigid

meter stick which when at rest in the laboratory is checked against a
standard meter bar. Call the laboratory, or ground, the system S. For
the system S' we shall take a very fast train which is moving in the X-
direction with the speed v. Let the meter stick ride on the train with its
length parallel to the direction of motion. An observer C on the train
checks the meter stick against a standard bar that he is carrying, and finds
the length of the stick still to be one meter.
Observers A and B on the ground set up telescopes along the tracks so
that they can observe where, relative to S, the two ends of the stick are at
some time t. The ground observers must first check their clocks with each
other to be sure that they agree on the time t. They then say that they will
observe the two ends of the moving meter stick at the same time t, or
“simultaneously. ” Observer A finds the front end of the moving stick to
be passing the point xa on the ground at the same time that B finds the
other end passing the point xb. Substituting in the Lorentz equation
x' k(x — vt), we have for the conditions above
=

x'a = k(xa — vt),

x'b = k(xb — vt),

where the t is the same in each case. Subtraction yields

x'a — x'b = k(xa — xb),

or

Xa — Xb = ^ (x'a — Xb) = (x'a — x£) V1 — V2/c2. (14-17)

Now (x'a — x'b) is the length of the stick as checked by the observer on
the train, and hence is exactly one meter. Then the length measured by
the ground observers is less than one meter by the factor 1 /It.
14-7] SIMULTANEITY AND RELATIVITY 307

Since motion is relative, the system S may be thought of as moving

relative to S' with the velocity —v. Therefore we may turn the experiment
around and have observers on the train measure, at their same time t',
the length of an object on the ground. The result would be

x'a — x'b = (xa — xb) Vl — (—v2)/c2

= (xa — xb)/k.

We see that this apparent shortening effect is reciprocal. The measured

length of an object is less when the object and the observer are in relative motion
than when they are at rest with respect to each other. This is the Lorentz-
Fitzgerald contraction.

14-7 Simultaneity and relativity. The Lorentz-Fitzgerald contraction

becomes less mysterious when we understand that two events which are
simultaneous in one inertial system may not be simultaneous in another.
Let us return to the case of the ground observer A who makes a measure¬
ment at x = xa at the time t, while his friend B makes an observation at
x — xb at the “simultaneous” time t. Let the third observer C, traveling
as before on a fast train, note the times t'a and t'b when, according to his
clock, observers A and B make their respective measurements. From the
Lorentz equation f = k(t — vx/c2) we have

t'a = k(t — vxjc2),

(14-18)
t'b = kit — vxb/c2).

If we take xa > xb, we see that tb > t'a. Observer C then claims that
by his clock B was a little later than A in making his observation. C might
then argue that since B looked at the moving meter stick later than A did,
B would see the back end at a point farther along the track than he would
have if he had looked at his end just when A was looking at the front end.
The result of B’s lateness would be to make the back end position seem
nearer that of the front end, i.e., the length would come out short of its
true value. . 7 . ,,
We see that two events, at xa and xb, which occur simultaneously in the
system S, are not observed to be simultaneous in the system S'. This theorem
is also a reciprocal one.
308 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

Solution. We have x'a — x'b = 1 m and

l - = vT^o^ = 0.8.

Hence from Eq. (14-17) we get

Xa — Xb = (x'a — x'b)/Jc = 0.8 m.

Subtracting the first equation in (14-18) from the second gives

t'b — t'a = ^ (xa — Xb) = \ X (1.0 m).

c c

The ground observers report that the moving stick appears to them to be 0.8 m
long. In communicating this information to the moving observer, the ground
observers hold up a stick of their own which they have cut to be 0.8 m long. The
moving observer looks at this stick and concludes that it is 0.8/k = 0.64 m long
and that the ground observers missed the length of the moving stick by 0.36 m.
The moving observer explains this as the result of B’s reading being made
(y/c2)( 1.0 m) sec later than A’s reading. The moving observer points out that in
this time the back end of his stick would advance a distance

2
v(t'b -t'a) =^X(lm) = 0.36 m,
c*

which is what the moving observer believes to be the error made by the ground
observers.

14-8 The apparent slowing of moving clocks. In the last section we

took xa > xb) ta = tb, and found that t'b > t'a. Now we shall take xa = xb
and tb > ta.
Let observer A at x = xa flash one signal at the time t = ta and another
signal at t = tb. The signals might be one minute apart by A’s clock.
Suppose that a group of observers in the moving system S' have synchro¬
nized their clocks and that the observer passing A at the time of the first
flash looks at his clock and records the time t'a of the flash. Another ob¬
server in S' similarly records the time t'b of A’s second flash. From the
Lorentz equations, we have

t'a = k(ta — vxa/c2),

t'b = k(tb — vxa/c2),

where x = xa in each case. Subtraction yields

t'b ~ t'a = k(tb — ta). (14-19)

14-9] THE TRANSFORMATION AND ADDITION OF VELOCITIES 309

Since k > 1, the time between A’s flashes is found to be longer by ob¬
servers in the S' system than it is according to A (who is in the system S).
Conversely, due to the relativity of motion, the time between two events
occurring at the same place in S' is found to be longer by observers in *S
than by observers in S'. We may sum up by saying that a clock will be
found to run more and more slowly the greater the relative motion between the
clock and the observer.
This effect becomes important in the case of high-energy particles of low
mass, such as mesons, whose speeds relative to the ground may be only
slightly less than that of light. These particles are observed to have short
lifetimes, but with their high speeds they may travel considerable distances
during their lifetimes. Since the speed of a meson gradually decreases as it
loses energy in the earth’s atmosphere, the velocity of the ground relative
to the meson also changes. Hence for a given meson the lifetime listed is
that for a system in which the meson is at rest. Calling this lifetime to,
the lifetime r observed in the earth system, relative to which the meson
is moving, may be considerably greater than r0- The apparent distance
traveled by the meson will be the value of v dt integrated over the apparent
lifetime r.

Example. A meson has a speed v = 0.6c relative to the ground. Find how far
the meson travels relative to the ground if its speed remains constant and the
time of its flight, relative to the system in which it is at rest, is 2 X 10-8 sec.

Solution. Since v = 0.6c, l/k = 0.8, k = 5/4. Relative to the earth the time
of flight At is
At = k X 2 X 10~8 = 2.5 X 10~8 sec.

The apparent distance traveled is then

Ax = v At = 0.6 X 3 X 108 X 2.5 X 10-8

= 4.5 m.

14-9 The transformation and addition of velocities. As before, take

the system S' to be moving in the X-direction with the speed v relative
to S, or S to be moving in the X'-direction with the speed —v relative to
S'. Consider an object which is moving relative to both systems, calling
the velocity components Vx, Vy, Vz relative to S, and Vx, Vy, Vz relative
to S'. From the definition of velocity, we have
i

v, _ dv'
V = * V"~W’
» dt

Differentiating the Lorentz equations (14 4), we find

310 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

dx' = k(dx — vdt), dt' — k(dt — vdx/c2).

Let us use these expressions to find the relation between Vx and Vx. We
get
dx' _ dx — v dt _ (dx/dt) — v Vx — v
V' = --
y X - 7./
=
-
dt' dt — vdx/c2 1 — vdx/c2 dt 1 — vVx/c2

In this manner, we obtain

Vx
V' =
1 — vVx/c2 ’

y _ -Yjl- (14-20)
k(l-vVx/c2)’

V' =
v2
k( 1 - vVx/c2)

The corresponding equations for Vx, Vy, Vz in terms of V'x, V'v, V'z
follow in the usual way, i.e., by changing the sign of v and interchanging
primed and unprimed quantities. For example,

V'x + v
(14-21)
~ 1 + vV'Jc2 '

Suppose now that a bullet is fired from a train in the forward direction
(Fig. 14-4). The muzzle velocity of a bullet is measured relative to the
gun, and since the gun moves with the train, we may let Vx be the muzzle
velocity. Then Vx is the velocity of the bullet relative to the ground. We
shall show that Vx cannot exceed c so long as V’x and v are less than c.

Fig. 14-4. Addition of velocities.

We may write Eq. (14-21) as

Vx =
(V'x/c )+ (y/c)
c
(1 - V'x/c)(l - v/c)
c.
LI + (v/c)(V'x/c) J 1 + (v/c){V'x/c)

(14-22)

Since each factor in brackets on the right is less than unity, we must have
Vx < c.
14-10] RELATIVISTIC DYNAMICS AND THE VARIATION OF MASS 311

Example. Given v = 0.9c and V'x = 0.9c, find Vx.

Solution. From Eq. (14-22), we find that

(0-l)(0.1)
Vx 1 0.9945c,
1 + 0.81 J °
whereas according to the Galilean transformations the answer would be Vx = 1.8c.

Since the relativistic composition of velocities makes it impossible to

build up a velocity greater than c out of two velocities each less than c,
we see that the velocity of light in vacuo is the maximum attainable velocity
for a material object.

14-10 Relativistic dynamics and the variation of mass with velocity.

According to the relativity principle, the laws of mechanics, as well as the
laws of electromagnetism, must remain invariant under a transformation
from one inertial system S to another system S'. This means that in any
inertial system we must find that when one or more forces act on a system:

I. The net work done by the forces equals the gain in kinetic energy of the
system* and
II. If the net force is zero, the momentum must remain constant.

Statement I is the principle of the conservation of mechanical energy,

and II is the conservation of momentum principle. Both follow from
Newton’s laws of motion, which must be invariant. If we write Newton’s
second law as
F = ma, (14-23)

with m regarded as constant, it is invariant under a Galilean transforma¬

tion but not under the Lorentz transformation. Since momentum appears
to be a more fundamental concept than acceleration, we write Newton’s
second law in the form
F = x (mv), (14-24)
dt

and we now inquire whether this may have the property of invariance
under a Lorentz transformation if we consider the mass m to vary with
the velocity The relation which we shall derive was first arrived at from
the study of the motion of charged particles in electromagnetic fields, a
subject called electrodynamics. We shall find it easier to derive the relation

*Work done by the earth’s gravitational force when a body is raised and work
done by frictional forces must be considered negative.
312 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

by applying the conservation of energy and momentum principles to a

simple hypothetical experiment first devised by Tolman.
In the system S' let two identical balls A and B approach each other at
the same speed along parallel lines, as in Fig. 14-5(a). Let the balls collide
in a perfectly elastic collision, so that kinetic energy is conserved. Then,
as is shown in mechanics, the conservation of energy and momentum re¬
quire that the direction of motion of each ball must be deflected through
the same angle 6. We may choose our X'-axis so that it makes equal angles
with all lines of motion, as shown in the figure. If the original velocity
components of B in the system S' are V'x and —V'y, respectively, the
corresponding components of A must be — V'x and -\-Vy. After the colli¬
sion the ^'-components will be unchanged, while the ^'-components will
be reversed.
Now take the inertial system S to be moving relative to S' with the
speed v = V'x in the negative X'-direction. Then, relative to S, ball A
will have no velocity along the X-axis, but it will appear to move in the
positive F-direction with a velocity Vay before impact, reversing its
motion after impact (Fig. 14-5b). Relative to S, ball B will have the veloc¬
ity component — Vby in the F-direction before impact and +Vby after¬
wards, while its ^-component Vx will remain unchanged. Since Vx must
exceed V'x, ball A rebounds more obliquely in S than in S'. The changes
in velocity components in S and S' may be summarized as follows:

In S' In S

A B A B

— V'
V x —>
7 V'
vX V' -4 V' 0^0 vx » vx
—

V y
r ' —> —V'
V y - V'y -> V'y Vay * Vay V by * V by

Letting v = V'x, we use the velocity transformations (14-20) to relate

corresponding y-components in the two systems. With Vax = 0, Vhx — Vx,
we get
T7> V ay T77 Vby

k ’ Vy ~ k(l - vVx/c*) ‘

Dividing the second equation by the first yields

lM= i
(14-25)
Vay C2

As viewed in S, each ball reverses its ^/-component of velocity upon im¬

pact. Equation (14-25) then tells us that ball B suffers a smaller change
in its ^-component of velocity than does A. If each ball had the same
14-10] RELATIVISTIC DYNAMICS AND THE VARIATION OF MASS 313

Fig. 14-5. Tolman’s hypothetical impact experiment.

mass, then the conservation of momentum principle would be violated,

so we must now assume that while the balls were taken to be identical,
this only means that, in a given inertial system, they have the same mass
when at rest or when they both have the same speed. In system S' the
balls have equal speeds in our experiment, but in S ball A has the speed
Va = Vay, while ball B has the speed Vb = (7f + Vlv)112. We shall
show that the conservation of momentum principle is satisfied if we take
the respective masses to be

m0 m0
ma = mb (14-26)
\/l — V2/c2 Vl - \E2/c2

where m0 is the mass of either ball as measured in the system in which it

is at rest. We call ra0 the rest mass.
Let us square each side of Eq. (14-25), multiply it through by the ratio
(mb/ma)2, and substitute from Eqs. (14-26) on the right side of the re¬
sulting equation. We then obtain

(mbVby\2 _ [ 1 - Vl/c2
W-Ll -(VI+V1J/C*
$

To show that this is unity, we must relate Vx to Vx v, using Eq. (14 21),
which gives us
17 -

~ 1 +V2/c2’
314 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

so that
2 2
VVs 2vJ C — V
= 1 -
c2 4- v2 c2 -1- v2

V by
“ (x " ^ S+
c2 + v2> '

V2
vX
,
+
V2
V by ~
4v'c4 + (c2 -
—
(C2
V*)2vi
+ P2)2
and

(rnbVby\2 _ 1 - Vl/c2
\ma Vayj C2 -V
+ yy2/
4»V + (cz -
1 -
c2(c2 -j- V2)2

(c2 - y2)(c2 - p2)2

c2(c2 + y2)2 — 4p2c4 — (c2 — w2)2F2

(c2 - y2)(c2 - P2)2

= 1.
(c2 — F2)(c2 — P2)2

Therefore, by Eqs. (14-26),

WaVay — m^Vby-

In the system S the ^-component of the momentum of A changes by

—2maVay, while the y-component of the momentum of B changes by
+2rribVby. There is no change in the ^-component of momentum during
impact. If we assume Eqs. (14-26), momentum is conserved during the
collision in both system S and system S'.
While we have shown, in the experiment considered, that mass must
vary with velocity as in Eq. (14-26), this relationship is quite general.
Although further treatment is beyond the scope of this book, the above
discussion serves to show why it is necessary to say that the mass of an
object varies with its speed.
Let us now call v the speed of a particle and m0 its rest mass in the
system S. Then, relative to an observer stationary in S, the apparent
mass m of the particle is given by
mo
m (14-27)
V1 — v2/c2

Since this is the mass that satisfies Eq. (14-24), it is also called the mo-
mental mass. From Eq. (14-27), we see that as v —> c, m —> oo, which
14-11] einstein’s mass-energy relationship 315

again tells us that no material object can attain a speed equal to or greater
than the speed of light.
Since objects in everyday life do not have speeds at all close to that of
light, their variation of mass with velocity is not measurable. However,
in the case of such small particles as the electron, proton, deuteron, etc.,
it has been possible to verify Eq. (14-27) experimentally. A beta particle
from a radioactive source, or an electron accelerated to a high energy in a
betatron, may have a speed sufficient for its apparent mass to be several
times its rest mass. For the heavier proton to show a mass increase of the
same proportion as that for an electron, the energy of the proton must
be 1840 times as great; nevertheless our powerful proton accelerators,
such as the synchrotron and cosmotron, reach energies well past the
threshold beyond which mass variation must be taken into account.

14-11 Einstein’s mass-energy relationship. We have seen that relativity

requires that Newton’s law of motion must be expressed as in Eq. (14-24)
and that at the same time we must take m to vary in accord with Eq.
(14-27). With m variable, Eq. (14-24) may be written as

d , \ dv dm , dm . nn\

F = df(mv> = mdt + vlu = ma + vdT' (14_28)

We see that the equation F = ma does not hold at high speeds if we take
m _ mo> nor is it valid in general if we take m to be the apparent mass.
In fact, if v and a are in the same direction, we find (see problem 6 at end
of this chapter) that
m0 (14-29)
F = a,
(1 - v2/c2)3'2

where m0/( 1 — v2/c2)3'2 is called the longitudinal inertial mass. Inertial

mass is defined as the ratio F/a. On the other hand, if F is perpendicular
to v, the inertial mass is found to be the same as the apparent mass given
by Eq. (14-27). _ ,
From the conservation of energy principle, we take the kinetic energy of
a moving body to be the work done by the force that accelerates the body
from rest. If we retain the definition of work as force times distance, and
assume linear motion, we have

d ds
K.E. = \ F ds = I % (mv) ds - I o dt
“ (mv) dt dt
v=0
dt

= / v f (mv) dt = I vd(mv).
.1 n dt
316 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

Substituting for m from Eq. (14-27) and integrating, we get

m0v
K.E.
V1 — v2/c2

1
C2(l _ V2/C2)3I2
v dv
-V1 — v2/c2

rv
/ V dv 2 1
= m0
Jo (1 — P2/c2)3/2 “ m°C (1 - P2/C2)1/2

m0c (14-30)
(a/ 1 Vc5

This is the relativistic expression for kinetic energy. If the first term is
expanded by the binomial theorem, we find that

K.E. = tuqc'
2 , 3 ^
c2t8c4 '"O'

which reduces to the familiar \m{)v2 of Newtonian dynamics when v « c.

Equation (14-30) may be rewritten as

K.E. = me2 — m0c2 = (m — mf)c2. (14-31)

Thus the kinetic energy of a moving body equals its gain in mass times c2.
We may also say that the apparent mass of a body increases linearly with
its kinetic energy, so that an increase in mass is an indication and measure
of the gain in kinetic energy. It is also found* that an increase in the
potential energy of a system of particles is accompanied by a similar in¬
crease in mass equal to the gain in energy divided by c2. Therefore we may
say, in general, that the gain (or loss) in the energy of a system is equal to the
gain (or loss) in its apparent mass multiplied by c2.
We may go one step further and interpret the term m0c2 in Eq. (14-31)
as the rest energy of a body whose rest mass is m0. This rest energy may

*See Richtmeyer, Kennard, and Lauritsen, Introduction to Modern Physics,

5th ed., pp. 69-70.
14-11] einstein’s mass-energy relationship 317

be regarded as a form of internal energy inherent in the nature of the

particles out of which matter is composed. Then

me2 = rest energy -j- kinetic energy

= total energy.

If here we let E stand for the total energy, we arrive at Einstein’s famous
principle of the equivalence of mass and energy.

E = me2. (14-32)

The value of any theory is measured by its success in predicting new

results. In this respect Einstein’s theory has been outstanding. In nuclear
physics the equivalence of mass and energy has been put to the test
repeatedly and it has always been confirmed. With the ability to measure
the masses of atomic particles to a high degree of accuracy, nuclear physi¬
cists have been able to predict the energy changes accompanying nuclear
and particle transmutations, and they have also been able to verify their
predictions experimentally. The whole subject of nuclear energy (popularly
called “atomic energy”) illustrates the usefulness of the above principle.
The energy exchanges involved in chemical reactions must also be accom¬
panied by corresponding mass changes, but in this relatively low-energy
field the mass changes are too small to be detected experimentally.
While chemical reactions and most nuclear ones involve a rearrangement
of atomic or subatomic particles, the electron-positron and the proton-
antiproton reactions are exceptions. In the latter two cases, physicists
apply the mass-energy equivalence principle in various ways. Some say
that when an electron and positron are annihilated, their rest mass is
converted into energy in the form of radiation called gamma rays; however,
we shall see in the next chapter that gamma rays may be considered to be
photons or light particles, which, because of their energy, carry with them
the momental mass originally associated with the electron and positron.
It would seem preferable to regard energy as a property of mass, or mass
as a property of energy, the two being inseparable. The basic conservation
principle then is the conservation of mass-energy.

Example. An electron and positron which are practically at rest come to¬
gether and annihilate each other, producing two photons of equal energy. Find
the energy and equivalent mass of each photon.
Solution. The rest mass m0 of an electron is 9.1 X 10"31 kg. This is equivalent
(in the mks system) to the energy
318 THE SPECIAL THEORY OF RELATIVITY [CHAP. 14

E = m0c2 = 9.1 X 10 31 X 9 X 1016 kg-m2/sec2

= 81.9 X 10-15 joule

81.9 X 10~15 joule

1.6 X 10-19 joule/ev

= 5.12 X 105 electron volts.

The positron has the same rest mass and energy. Since the combined energy is
shared equally by two photons, each photon will have an energy of 512,000 ev.
While the rest mass of a photon has no meaning, since a photon cannot be
thought of as existing in a state of rest, we may say that during its existence each
of the above photons will have an “associated mass” equal to 9.1 X 10-31 kg.

Problems

1. Show that all four of Maxwell’s length, what should the fringe or phase
equations (14-6) are invariant under shift in the interference pattern be,
the Lorentz transformations. according to the ether theory, when the
2. In the system S take E to be in apparatus is rotated 90° ?
the F-direction and B to be in the Z- 5. A certain young lady decides on
direction. For the two cases when (a) her twenty-fifth birthday that it is
E > cB and (b) cB > E, find the time to slenderize. She weighs 200 lb.
velocity v < c of the system S' in She has heard that if she moves fast
which the field will appear to be either enough she will appear thinner to her
entirely electric or entirely magnetic. stationary friends, (a) How fast must
For each case state whether it is E' or she move to appear slenderized by a
B' that is zero. factor of 50% ? (b) At the speed found,
3. Referring to the two swimmers in what will her mass appear to be (to her
Section 14-4, take the current v — stationary friends)? (c) If she main¬
3 ft/sec and the speed c of each swim¬ tains her speed until the day she calls
mer in still water to be 5 ft/sec. Find her twenty-ninth birthday, how old
(a) the times t\ and <2 when AC = will her stationary friends claim she is
BC = 100 ft, (b) the distance AC for according to their measurements?
which, with BC = 100 ft, the swim¬ 6. Calling the ratio F/a the inertial
mers would have the same times. mass mi, show that, according to rela¬
4. In the Michelson-Morley experi¬ tivistic dynamics, (a) mi = too/(1 —
ment take l — 11 m and v/c = 10~4 v2/c2')3/2 when F is in the direction of
(the ratio of the earth’s orbital velocity the velocity v, and (b) to; = mo/(l —
to the speed of light), (a) Neglecting v2/c2)1/2 whenF is normal to v. (These
terms smaller than v2/c2, show that values of mi are called the longitudinal
the time difference <2 — £1 = lv2/(?. and transverse inertial masses, respec¬
(b) For light of 5 X 10-7 m wave¬ tively.)
 PROBLEMS 319

7. (a) Define the momentum p of a weighs close to 1 kg. If the sphere cools
body, according to relativistic dy¬ 1200°C, what is the equivalent loss in
namics. (b) Show that p2 = m2c2 — mass?
TOqC2. 13. An atomic mass unit (amu) is
8. For what energy in electron volts defined as 1/(6.02 X 1026) kg. (a)
will the ratio m/mo for an electron be Find the energy equivalence of 1 amu.
(a) 1.1, (b) 2, (c) 10, (d) 100? (b) If a deuteron (H2) and a triton
9. Repeat problem 8 for protons in¬ (H3) combine in a fusion process to
stead of electrons. form an alpha particle (He4) and a
10. For what energy in electron volts neutron (n), find the percent loss in
will the ratio v/c for an electron be (a) mass if the respective masses are
0.10, (b) 0.50, (c) 0.90, (d) 0.99? 2.0147, 3.0170, 4.0039, and 1.0090 amu.
11. Assume an electron to be a (c) Compute in kilowatt-hours the
charged spherical shell of radius a and energy released per kilogram in (b).
charge e. (a) Find the work W to put 14. Suppose that a plane electro¬
a charge e on a sphere of radius a. (b) magnetic wave is incident normally on
Equate this work to the rest energy a plane absorbing wall. Use Eq.
tooc2 and get an expression for a in (14-32) to show that the radiation will
terms of e, mo, c, and eo. (c) Substitute exert a pressure equal to the Poynting
numerical values and find a for an elec¬ flux divided by the speed of light. Com¬
tron. (d) Similarly, find a for a proton. pute this pressure for sunlight (for
12. A red-hot sphere of iron (speci¬ which the Poynting flux = 1400
fic heat capacity = 0.11 cal/gm-°C) watts/m2).
 CHAPTER 15

WAVE MECHANICS

15-1 Introduction. The laws of classical physics apply very well to the
behavior of objects that we can see with our eyes. However, in modern
atomic physics we talk about and experiment with particles too small to
be seen even under the highest-powered microscope. Physicists have be¬
come so familiar with the properties of, say, the electron, that the existence
of this particle seems just as real as that of a tennis ball. It is often con¬
venient to form a mental picture, or model, of something that we cannot
see. Such a model helps us to assimilate the known facts about the object
in question. Rutherford’s planetary model of the atom is a good example.
A model may also suggest and lead to the discovery of new properties
possessed by the object. Important discoveries have been made in this
way. However, we must guard against taking a model too literally, since
a model may also suggest properties that the object in question does not
actually possess. For example, physicists talked so much about the
planetary model of the atom that some began to think that the electrons
in an atom really were tiny spheres circulating in definite orbits; it seemed
to them that with a more powerful microscope one might actually be able
to see an electron occupying a definite position at a given time, just as in
the case of the familiar tennis ball. We shall see that this conclusion is
misleading, and that we are on safe ground only when we speak of the
observable properties of the electron.
In describing most of the properties of atomic particles a new universal
constant, not needed in classical physics, must be introduced. This
constant is h, Planck’s constant, and its introduction forms the basis for
the quantum principle. We saw in Chapter 6 how this constant was
originally introduced by Planck to explain cavity radiation. We also
saw how this postulate led, in the case of a linear oscillator, to discrete
energy levels and to a recognition of the futility of specifying the state
of the oscillator, in px — x momentum space, closer than to within an
area equal to h. This latter result, it was pointed out, is a special case of
Heisenberg’s uncertainty principle.
A law of physics can usually be expressed in many different, but equiv¬
alent, ways. Heisenberg’s principle is the most concise statement of the
quantum principle, so that in Planck’s work we have the essence of
the quantum theory. Unfortunately, the uncertainty principle is not
the easiest starting point in quantum theory, and Planck did not work out
the mechanics for applying this principle to atomic systems in general.
This was eventually done by Heisenberg about twenty-five years later.
320
15-1] INTRODUCTION 321

Meanwhile, the quantum theory progressed in another direction. Ein¬

stein postulated that if an oscillator emits and absorbs energy by jumping
between discrete energy states, then the light emitted and absorbed must
also consist of bundles of energy called photons or quanta. Only in terms
of photons could the photoelectric effect, and later the Compton effect,
be explained. Einstein’s postulate was also incorporated into the Bohr-
Sommerfeld and later theories of spectra as follows. It was postulated that
a series of discrete energy levels is possible for an atom, just as for Planck’s
oscillators, but that the values of the allowed energies are characteristic
of each kind of atom. When an atom undergoes a transition from an initial
state of energy Ei to a final state of lower energy Ef, the atom must emit
the energy Ei — Ef. According to the quantum theory, this energy ap¬
pears as a radiated photon with the energy hv, where v is the frequency
of the emitted light and

hv — Ei — Ef. (15—1)

By 1924 there still remained the problem of finding a logical theory that
would account for the discrete energy levels of an atom and that would
lead to the numerical calculation of these energies. The postulate of the
Bohr-Sommerfeld theory, fpdq= nh, was arbitrary, and it was found to
apply only to one-electron atoms. The solution of the problem developed
independently along two different lines, one being the matrix mechanics
of Heisenberg and the other the wave mechanics of Schroedinger, Born,
and others. These two approaches to the problem were found to be
equivalent, and since wave mechanics is easier to understand, we shall
use this approach here.
Wave mechanics is a theory. Like kinetic theory, it is based on funda¬
mental laws plus certain additional postulates. The validity of these
postulates must be determined by how successful the theory is. A good
theory should lead to (1) the correct functional relationship between
quantities involved, (2) the correct numerical values for any constants in
the formulas, and (3) the prediction of new experimental discoveries.
Wave mechanics has accomplished all of these things.
In presenting a theory, a choice must be made between a historical ap¬
proach and the most logical approach. The latter was chosen in mechanics,
thermodynamics, and electromagnetism. Here we shall follow the historical
development of our theory, as we did in the last chapter, because it is
instructive to see how the new theory grew out of and is related to con¬
cepts and theories presented earlier, and this approach may make the
theory seem more reasonable than if it were presented in the abstract.
We shall also see that just as relativistic mechanics includes Newtonian
mechanics as a special case, so ivave mechanics reduces to ordinary me-
322 WAVE MECHANICS [CHAP. 15

chanics for particles of sufficiently large mass. Newtonian mechanics has

been a highly successful theory and, because of its greater simplicity,
is still preferred when we are dealing with nonrelativistic and nonatomic
problems. However, the theories that preceded the theory of Newton and
the laws of thermodynamics were not special cases of our present theories,
nor were these earlier theories successful ones; they were mainly mislead¬
ing, and so are not worthy of consideration here.

15-2 The dual nature of light. In discussing light, we may say that the
three basic processes are emission, 'propagation, and absorption. In emis¬
sion and absorption we are concerned with how light energy is produced
at the expense of some other form of energy, and vice versa. Actually, we
do not detect the presence of light until it falls on some absorbing object
such as the human eye, a strip of film, or a photocell, in which cases the
light energy is transformed into an electric nerve pulse, chemical energy,
and the energy of photoelectrons, respectively.
Photoelectric and other measurements indicate that light energy is
emitted and absorbed in quantum units, rather than continuously, but
note that the emission and absorption processes have nothing to do with
determining what path the light will follow as it travels from its source
to its sink. However, suppose that we use a system of slits to limit the
path of the light and that we observe along which paths the greatest
amount of energy will travel. Then we can explain or predict our results
correctly by means of the wave theory. In wave theory, we must intro¬
duce the concept of wavelength X and velocity of propagation c, both of
which may be determined experimentally. The frequency v associated
with light of wavelength X is found from the wave relation v = c/X.
We see, then, that the photon theory is used to explain emission and
absorption, and the wave theory to explain the propagation of light.
These theories complement each other, since they apply to different phe¬
nomena and describe different aspects of light. Using both the corpuscular
and the wave descriptions tells us more about the nature of light than we
could learn if we adopted only one viewpoint.
It is tempting to think of photons as maintaining their identity along
the path between where we observe them to be emitted and where we see
them absorbed, but such a postulate has no experimental basis. When
we use the wave theory to calculate the intensity of the light falling on a
certain area of a screen, we are computing how many photons will be
received there in a given time; this in turn may be interpreted as finding
what fraction of a large number of emitted photons will reach the given
area, or the probability that one photon will do so. While we may also
think of the intensity as being proportional to the square of the magnitude
of the electromagnetic field vectors, these quantities are not directly
15-2] THE DUAL NATURE OF LIGHT 323
i
measurable for light in the range from the near infrared on up to gamma
rays because the frequencies are so high. Conversely, in radiowaves the
energy per photon is too small to measure, and so the photon theory
cannot be employed. We do not use both theories at the same time.
The important difference between classical wave theory and the photon
viewpoint is that the former is concerned with continuously varying
quantities, while the latter deals with discontinuous processes. For ex¬
ample, according to classical physics, the intensity of an electric field E
can be decreased gradually from some finite value to zero. A photon,
on the other hand, cannot be subdivided; “less than one photon” must
mean “no photon.” We shall see later that the indivisibility of the photon
is related to Heisenberg’s uncertainty principle. The completely causal
relationship between past and future events which is assumed in classical
theory is replaced in the quantum theory by a statistical relationship
giving the probability that a certain future event will occur.
It was realized that there must exist borderline phenomena for which
an equally good explanation can be found by either the wave or the photon
theory. It was also realized that a subject can never be completely under¬
stood if we look at only one aspect of it at a time. The bridging of the two
theories was accomplished by the correspondence principle, which was
first proposed by Bohr. This principle states that quantum expressions
must revert to the corresponding classical ones in the limiting case for
which the contribution of one photon to the total energy may be regarded
as infinitesimally small. In this case the average value predicted for a
quantity by the quantum theory becomes the stated value by the classical
theory. As an example, we may refer back to Eqs. (6-27) and (6-28)
where, if hv « kT, so that the summation may be replaced by an integra¬
tion, the classical expressions are obtained.
If a photon possesses energy, we may, if we wish, speak of the mass as¬
sociated with, or equivalent to, that energy. We shall call this “associated
mass” of the photon m. Note, however, that a photon does not have a rest
mass since there is no such thing as a photon at rest. A photon either
moves with the speed of light, or it has been absorbed and no longer exists.
Einstein postulated that the energy E of a photon is

E = hv, (15-2)

where v is the frequOncy of the light. The “associated mass must then be

Since a photon travels with the speed of light, it must possess a momentum
p given by
324 WAVE MECHANICS [chap. 15

hv h
p = me = — = — (15-4)
c A

This last result may be obtained in another way. As the speed of a par¬
ticle approaches that of light, its momentum and energy both become
very large, but the ratio of the momentum to the energy approaches the
finite value 1/c. Thus

This relation has been found to hold for electromagnetic waves. If we

apply it to the photon and use Eq. (15-2), we immediately obtain p =
hv/c.
We shall see that the relationships for the energy and momentum of a
photon are experimentally supported, as in the Compton effect, for example
(see below).
Illustration 1. The photoelectric effect. This effect constitutes experi¬
mental confirmation of Eq. (15-2). When light of frequency v strikes a
metal, electrons are ejected if hv exceeds a quantity W which is character¬
istic of the metal. W is called the work function and it is interpreted as the
work required to remove the electron from the metal. If hv > W, the
electrons are ejected with a kinetic energy given by

EE. = hv - W. (15-5)

Applying the conservation of energy principle, we see that hv must be the

energy of the incident photon.
Illustration 2. The Compton effect. When high-frequency radiation, such
as a beam of x-rays, strikes an electron, the electron may receive the whole
energy of a photon, as in the photoelectric effect, or the electron may recoil
with only a fraction of this energy. In the latter event, the photon is
scattered with reduced energy and frequency, the whole process corre¬
sponding to a collision between particles of generally unequal mass. If
the energy E of a photon is expressed in electron volts (ev), its “associated
mass” is
E _ E(ev) X 1.6 X 10 19joule/ev
c2 (3 X 108 m/sec)2

= 1.78 X 10~36E kg.

This equals the rest mass of an electron (ra0 = 9.1 X 10 31 kg) when
E is about 512,000 ev, so that for x-ray photons of about 100,000 ev m is
less than one-fifth of ra0. It is shown in mechanics that in an elastic impact
between a moving and a stationary particle very little K.E. is transferred,
15-2] THE DUAL NATURE OF LIGHT 325

even in a head-on collision, unless the two masses are of the same order of
magnitude. Accordingly, we might expect that the Compton effect would
be of little practical significance in the case of low-energy photons corre¬
sponding to the visible and ultraviolet regions; this is verified experi¬
mentally.
As in Fig. 15-1, consider a photon of energy hv to strike an electron of
low kinetic energy. Let the electron recoil at the angle 9 with the speed v
and let the photon be scattered at the angle <j> with the reduced energy
hv'. From the principle of conservation of energy, we have

1
hv = m0c2 + hv'. (15-6)
y/1 — v2/c2

The relativistic expression for kinetic energy must be used, because, for the
recoil electrons, v may be an appreciable fraction of c.
K.E. = me2 — m0c2

Fig. 15-1. Compton collision between photon and electron.

Let us apply the conservation of momentum principle to the collision.

Since momentum is a vector, we write two equations, expressing the con¬
servation of momentum for both the X- and F-directions. Using Eq.
(15-4) for the momentum of the photon and the relativistic expression
for the momentum of the electron, we have

hv moV a 1 hv' (15-7)

—--cos 9 4- — cos (t>.
c Vl — v2/c2 c

mnV . „ hv' .
0 = ---sin 9-sin <f>. (15-8)
VT — v2/c2 c

Between these last three equations we may eliminate 9 and v, solving for
/ in terms of v and </>. We now introduce the wavelengths X = c/v and
326 WAVE MECHANICS [CHAP. 15

X' = c/v', since these are what we measure, and find (see problem 3 at
end of chapter) that

X' — X = (1 — cos #). (15-9)

m0c

The factor h/m0c, called the Compton wavelength, has a numerical value
of about 0.024A, where 1A = 10-10m. For monochromatic x-rays
scattered through 90°, we find that part of the beam undergoes just this
change in wavelength.
The kinetic energy of the recoil electrons is given by

K.E. = hv — hv'.

Suppose, for example, that X = 0.124A. Then

he 6.62 X 10 _34 joule-sec X 3 X 108 m/sec

hv ~ “ 0.124 X 10“10 m

= 1.6 X 10-14 joule = 105 ev.

Such x-rays are obtained from a tube operating at over 100,000 volts. If
X' — X = 0.024A, X' will be 0.148A, or about 6/5 of X. Then hv' will be
about 5/6 of hv, or about 83,000 ev, so that the K.E. of the recoil electron
will be about 17,000 ev. The coincident production of a scattered photon
and recoil electron has been observed by means of Geiger counters.

15-3 The dual nature of atomic particles. Louis de Broglie first pro¬
posed in 1924 that a material particle such as an electron might have a dual
nature, just as light does. In the study of light the wave properties, in¬
volving X, v, and c, are the more familiar ones. In terms of these properties
we have defined the mechanical properties of the light corpuscles as follows:

E = hv, V = \- (15-10)
A

For a material particle it is the other way around; we are used to speaking
of mass m, momentum p, and energy E. These are measurable quantities.
Therefore when de Broglie postulated that such a particle may also have
wave properties, he defined the associated wavelength X and frequency v
in terms of m, p, and E. In doing so, de Broglie was guided by the relations
for light (Eqs. 15-10), and so he postulated that associated with a particle
are waves of wavelength X and frequency v, given by

h E
X = v (15-11)
V h '
15-3] THE DUAL NATURE OF ATOMIC PARTICLES 327

De Broglie developed a relativistic theory, taking E = me2, so that the

phase velocity, or velocity of propagation, u, of the associated waves was
taken to be
E me2 c2
(15-12)
U V p mv v

Since v < c for a material particle, c2/v > c. This need not concern us
because u is neither the velocity of a material particle nor, as will be shown
later, the group velocity of a set of waves. In fact, the waves associated
with a particle are not to be regarded as being mechanical or electro¬
magnetic, but rather “probability waves.” By this we mean that the in¬
tensity of the waves at any point will be taken as giving the fraction of a
large number of similar particles, emitted with the same initial velocity,
that will reach a given area in unit time, or the probability of one particle
reaching that area. The waves are thus a device for computing the prob¬
ability that a particle will behave in a certain way. De Broglie argued
that in the case of light one cannot think of a unit of energy without
associating with it a wavelength and frequency, and that therefore material
particles, like protons, must be accompanied by phase waves.
The behavior of a material particle is usually computed from the laws
of mechanics. Is there, then, any need for a new method? If so, what are
its advantages? De Broglie suggested that wave mechanics might stand
in relation to particle mechanics as wave optics does to geometrical optics.
In geometrical optics light is assumed to travel through a homogeneous
medium in straight lines or rays. To explain the details of the diffraction
pattern caused by an obstacle, one must resort to wave optics and be con¬
cerned with distances of the order of magnitude of one wavelength. WTen
we compute the wavelength X = h/mv associated with an ordinary object,
it turns out to be insignificantly small, but for an electron the mass m
is very much less than for ordinary objects and h/mv is found (see example
below) to be comparable to x-ray wavelengths. Since x-rays are diffracted
by crystals, why should not such electrons be, too? This question was
answered by Davisson and Germer who, while studying the scattering of
electrons from metals, found that electrons are indeed diffracted by crystals
just as x-rays are, and in accordance with the relation X h/mv.
De Broglie also realized that in dealing with atomic systems, where the
dimensions are also comparable to x-ray wavelengths, wave mechanics
might be able to explain things that ordinary mechanics could not. For
example, according to the Bohr theory of the hydrogen atom, only those
electron orbits are allowed for which the angular momentum pe is an in¬
tegral multiple of H/2t, or

(15-13)
328 WAVE MECHANICS [CHAP. 15

This whole number rule was empirical when it was first proposed, the argu¬
ment in its favor being that it worked. De Broglie suggested, by way of
explanation, that if waves accompany an electron around its orbit, then
a sort of resonance might occur when the circumference of the orbit equals
an integral multiple of the electron’s wavelength. This idea, which was
later refined by Schroedinger, amounted to saying that the allowed
quantum orbits correspond to the various possible modes of vibration, or
standing waves, in a string. For such a string, waves are reflected back
and forth with constructive interference only when the wavelength X is
such that the complete distance down the string and back is an integral
number n of wavelengths. Here the restriction that n must be a whole
number has a physical basis, hence it was appealing to carry the idea over
to the case of the atomic orbit and its electron waves.
Consider a circular orbit of radius r. Taking X = h/p, where p is linear
momentum, de Broglie put the orbital circumference 2-kr equal to nX, or

2-7rr = n\ = n— >
V
h
Pe = rp = n

which is the same as Eq. (15-13).

While de Broglie’s ideas proved most important, he did not develop
them into an exact theory, as we shall attempt to do in the following
sections.
Illustration. Find the de Broglie wavelength of an electron whose
kinetic energy equals V electron volts (eV joules), using the nonrelativistic
expression for kinetic energy.
We have
\mv2 — eV,

mv = V2 meV,

X _ h _ h
mv \/~2meV

Taking h = 6.6 X 1Q~34 joule-sec, m = 9.1 X 10~31 kg, and e = 1.6 X

10~19 coul, we get
. 1.23 X 10~9
X =-—-m.
Vv
Thus for 1600-volt electrons, X = 0.3 X 10~10 m = 0.3A, which is of
the order of the separation of the planes in a crystal lattice.
15-4] analogy: ordinary mechanics and geometrical optics 329

15-4 The analogy between ordinary mechanics and geometrical optics.

If a projectile fired with a given energy is to reach a certain target within
its range, the gun must be aimed correctly. If we compare the dynamical
path actually followed with all neighboring paths that the bullet might
be imagined to be constrained to take, with the same energy, from gun to
target, we find that the actual path is the one of least action* This variation
principle may be written as

sf 2Tdt=0, (15-14)
Jc

where T is the kinetic energy, t is the time, and C and D refer to the
terminals of the path.
In optics a beam of light, say from a flashlight, must also be properly
aimed to strike a given target. According to Fermat’s principle (see Sec¬
tion 13-4), if the optical path is compared with all neighboring paths that
might be traversed, at the speed appropriate to each medium, from source
to target, it will be found that the actual path is the one traversed in the
least time. This principle may be written as

rD

8 - = 0, (15-15)
Jc u

where ds is an element of path length and u is the velocity of propagation

at ds, so that ds/u is an element of time.
Hamilton first recognized that both the dynamical path in mechanics
and the optical path in optics may be singled out by the aid of a variation
principle. Schroedinger proceeded to develop the analogy further.
Equations (15-14) and (15-15) differ in their integrands. For better
comparison, we shall first change Eq. (15-14) from a time integration to
one with respect to the path length s, as in Eq. (15-15). Here we shall
follow Schroedinger’s nonrelativistic treatment and put

E — T + V,

where E is the total energy of the particle and V its potential energy at a
given point. As we have said, E is assumed to be constant, or the system
to be conservative, but V may be a function of the coordinates. Neglecting
relativity enables us to take
9 ds

*See Vol. I, Section 10-5.

330 WAVE MECHANICS [CHAP. 15

If we substitute this in Eq. (15-14), we get

8
[D mv ds = 0. (15-16)
Jc
To eliminate v, we write

v = y/2T/m = y/2 (E — V)/m

and substitute in Eq. (15-16), getting

-D

8 V/2m(E — V) ds — 0. (15-17)
Jc

The principle of least action in the form of Eq. (15-17), and the principle
of least time, Eq. (15-15), become identical if we postulate that

V2m{E - V) = -,
u
(15-18)

where C is not a function of the coordinates but may depend on E. Here u

is to be regarded as the phase velocity of the waves associated with a
particle. These waves are being introduced as a postulate in the hope that
by computing their behavior we may be able to gain information about
that of the particle. We are free to define the phase velocity and frequency
of these waves to suit our convenience. By choosing u in accordance with
Eq. (15-18), we turn a mechanical problem into an optical one, where the
principle of least time, which summarizes the laws of ray optics, tells us the
correct path from the point where a particle is emitted to the point where
it is absorbed. The only difference is that now our particles are material
ones rather than photons.
We see that u is a function of V, so that in a field of force the phase
velocity varies from point to point, just as the velocity of light changes
with the index of refraction of the medium through which the light is
passing. The force field refracts the particle waves.
Another important point is that the u defined by Eq. (15-18) is depend¬
ent on E, so that the waves associated with particles of different energy
have different phase velocities and are refracted differently.
We must now decide how to choose the frequency v of the particle waves.
Schroedinger was guided by de Broglie, and assumed that

(15-19)

only here we must remember that E stands for T + V and not for me2.
15-4] analogy: ordinary mechanics and geometrical optics 331

The wavelength of the waves follows from X = ufv, so that only the
constant C remains to be determined.
Condition (15-18) ensures the right path for our particle, but it does not
ensure that the particle will traverse the path with the correct speed v
at every point. Fortunately, C may be chosen in such a way as to make
this so.
From Eqs. (15-18) and (15-19) we see that u varies with the frequency.
This is dispersion. In a dispersive medium the group velocity G is not
the same as the average phase velocity u of a set of waves of varying fre¬
quency. We have seen that group velocity is identified with signal velocity,
which suggests that we should try to make the group velocity of our
particle waves the same as the actual velocity v of the particle. This im¬
plies that we must take our waves to have a range of frequency Av about
the frequency v = E/h. From Eq. (13-32), we have

where we now write u for the phase velocity. Putting v — E/h and using
Eq. (15-18), we get

I _ A (E\ = A.
\EV2m(E - V) (15-20)
G ~ dE\u) ~ dE L C

The velocity v may be expressed in a similar way, for

1 m_rn_
v ~ mv y/2m(E — V)

= A [v/2m(£l — F)]. (15-21)

(This step may be easily verified by using E — V — T — mv2/2 and

checking the indicated differentiation.) We see that to make G = v we
need only choose
C = E

♦ __= (15-22)
V2m{E - V) mv P

As in Eq. (15-12), u equals the energy divided by the momentum of the

particle, but here our E has the nonrelativistic meaning E = T + V.
332 WAVE MECHANICS [CHAP. 15

The wavelength A of the particle waves may now be found. We have

u _ E h h _ h_
~ ^ 7~T
p E p mv

just as de Broglie postulated.

15-5 Schroedinger’s equation. Like de Broglie, Schroedinger realized

that wave mechanics might give us more detailed information about, say,
an electron in an atom, than we could ever get from classical mechanics,
just as wave optics tells us things about diffraction that ray optics never
can. Schroedinger’s important contribution was to carry out this idea in a
practical way. He reasoned that since the starting point in diffraction
theory is the differential wave equation, we must then look for a corre¬
sponding equation for wave mechanics. Since waves have been associated
with a particle, it seemed logical to commence with the wave equation
for these waves.
We saw in Chapter 11 that if E is a function of the coordinates and time,
and if E is propagated through space with the velocity u, then E must
satisfy the wave equation

V • VE

We also saw that for waves of a definite frequency v we could put

E = E*0, y, z)e±2*ivt.

Substituting this in the wave equation, we obtained the time-independent

amplitude equation

V • VE* + E* = 0.

This equation and the boundary conditions determine the amplitude factor
E* at any point.
When we consider the application of this procedure to the de Broglie
waves, we must note that the latter are not physical waves in the sense
that elastic and electromagnetic waves are, but that they are waves which
we invent and study mathematically. Nevertheless, it seems like a good
“guess” to assume that they satisfy equations similar to those for E above.
However, another problem arises in connection with the fact that the total
energy E of our particle will be constant only for steady-state problems.
This means that when de Broglie waves are propagated, the frequency v,
which according to Eq. (15-19) equals E/h, will not necessarily be con¬
stant. Since the relationship between the amplitude equation and the
wave equation depends on the time factor e±2*ivt, it will vary with the
15-5] schroedinger’s equation 333

total energy. This suggests that we start with just the amplitude part
of our wave function and with the amplitude equation.
Let y, z) be the amplitude function for our de Broglie waves. We
postulate that it must satisfy the differential equation

4 2 2
V • V* + * = 0. (15-23)

For these waves we must put v = E/h and give u the value chosen in Eq.
(15-22). Equation (15-23) then becomes

V • V* + (E - V)ip = 0, (15-24)

which is Schroedinger’s equation (without the time).

To find the corresponding time-dependent equation we must eliminate E.
Without loss of generality we may assume simple periodic waves and put

* = *(z, y, z)e~2*ivt, (15-25)

where SI' is the complete wave function. (The minus sign in the exponent
is preferred only for relativistic reasons.) Since v = E/h,

d* . 2iriE
— 2 TlvV —-T—
dt h
or
h 34' _ ih d’L
E'ir
2wi dt 2ir dt
so that we have
4:irm d'f' 8ir2mV^ _ ^ (15-26)
V • V’F
~Hh ~dt h2 _ :

which is Schroedinger’s time-dependent equation.

In the examples that follow we shall be concerned only with finding
the amplitude function and so we shall start with and integrate Eq.
(15-24). The solution of Eq. (15-26) is left to more specialized texts.
Since a particular problem may be defined by giving the field of force,
or the potential function V from which the field may be derived, we see
that by introducing this particular V in Eq. (15-24) we will get a differen¬
tial equation characteristic of the problem. Thus, for the simple harmonic
oscillator, we put V = kx2/2, while for the hydrogen atom we take
y — e2/4:ireQr, For each such problem we will have a different
Schroedinger equation to integrate. Since this equation contains a term
defining the force field, it is not surprising that from its solution much
334 WAVE MECHANICS [CHAP. 15

may be learned about a given problem. In fact, Schroedinger’s equation is

the quantum-mechanical equivalent of the Newtonian equation of motion.

15-6 Solution of Schroedinger’s equation. We have seen that in solving

a differential equation for a particular problem, an important step is the
introduction of boundary conditions for the purpose of evaluating the
constants of integration. A hydrogen atom has no boundary, so when
Schroedinger’s equation is solved for this problem the solution must be
limited in other ways. Schroedinger had the genius to find these conditions.
He reasoned that \p must have some physical meaning. For other waves
in our physical world the amplitude always varies continuously apd, at a
particular time, has only one value at a given point in space. Furthermore,
wave amplitudes squared usually represent intensities, and so must give
a finite integral when integrated throughout space. Hence Schroedinger
postulated that solutions of Eq. (15-24) must be single-valued, continuous,
and such that
Iff I’/'|2 dx dy dz
infinite. We shall see that this postulate leads to discrete energy levels for
an atomic system. This comes about because in solving Eq. (15-24) we
find that Schroedinger’s conditions cannot be satisfied unless the quantity
E, which is a parameter in the equation, is restricted to a set of values
Ei, E2, E3 • • • . These are referred to as eigenwerte, or 'proper values.
Since Schroedinger’s equation is a homogeneous, partial differential
equation, it may be solved by the method of separating the variables. This
method has been treated in connection with solutions of the heat equation,
the wave equation, and Laplace’s equation. For each proper value of E,
a solution, called the wave function, is found for f. Schroedinger thought
that, in the case of an electron, \f\2 must represent the density of charge
in a given region. However, physicists have come to prefer an alternative
interpretation given by Born, according to which \x//\2 dx dy dz measures
the probability of finding the particle in a given element of volume.
For illustration we shall now solve Schroedinger’s equation for some
familiar problems.

15-7 Simple harmonic oscillators. Simple harmonic oscillators are de¬

fined by the potential function V = %kx2, or by the Schroedinger equation

d2\p r2m
87
(E — \kx2)\J/ = 0. (15-27)
dx2 h2
We first simplify this equation by putting

87v2mE 4:ir2mk 7

h2 = a, ~h^ = b-
15-7] SIMPLE HARMONIC OSCILLATORS 335

which yields

(15-28)

The solution of Eq. (15-28) is not simple, but fortunately it had already
been worked out by mathematicians before Schroedinger’s equation was
known. We shall sketch the solution, emphasizing only those points
which help illuminate Schroedinger’s theory.
As a solution of Eq. (15-28) we try

* = e-q2'2 £ (15-29)
s=0

and substitute this into Eq. (15-28):

(g2 - 1 )2>sgs - 2q(ax + 2a2q + 3a3g2 H-)

+ (2a2 + 6 a3g + 12a4g2 + • • •)

This expression must be an identity, i.e., the coefficient of each power of

q must vanish separately. Putting

and equating to zero the coefficients of q°, g\ q2, and qn, we find that

2 — p P a2, •
a2 = — ^ «o> a3 =
3-2
ai, a4
4-3

2 n — ft (15-30)
dn-
a„+2 — (n _)_ 2)(n + 1)

This is the only mathematical restriction on the solution of (15 29). We

may take either a series of even or a series of odd powers of g with coeffi¬
cients decreasing according to Eq. (15-30). However, Schroedinger s
postulate requires that ^0asxorg^», Asg->co,e -*0,
while the power series rapidly increases. Note that

2n 2
Lim n
n—>°° n2
336 WAVE MECHANICS [CHAP. 15

while if we expand
n+2
i L.
e9 = 1 + q*2 +
2! (n/2)! 1 [in + 2)/2]!
+
— 1 b2q~ T~ q4 • + bnqn + &n+2 Qn h2 +
then

Lim (%h) = Lim i —— \ = - ,

n—>oo \ On
/ 7i >oo \7Z
— 2/ 71/

which shows that as q —► oo, ^ behaves like

e~i2n . e<z2 = /I2>

which —4oo. This makes our solution appear physically impossible until we
consider the possibility that the infinite series might break off after, say,
the nth term. For this to happen, we must have an+2 — 0, then am+4,
an+6; etc., will all vanish as well, and our series will become a polynomial
Hn>2 known as a Hermite 'polynomial. These polynomials, multiplied by
e~q 12, approach zero as q —» oo. They also give solutions that are single¬
valued and continuous.
Evidently, the important condition is that

2n - )8 = 0, (15-31)

where n is an integer, since this will make an+2 = 0 in Eq. (15-30). If

we substitute in Eq. (15-31), first putting (a/y/b) — 1 for /3 and then
using the expressions for a and b, we find that the energy E must satisfy
the condition

2t E = (n + n = 0, 1, 2, 3, • • • .

The mechanical frequency / of the oscillator is

hence we get
E = (n + £)hf, to = 0, 1, 2, 3, • • • . (15-32)

The allowed integral values of n lead to a discrete set of equally spaced

energy levels. The spacing hf of these levels agrees with that postulated
by Planck (Eq. 6—1), and with that derived by the Bohr-Sommerfeld
quantum theory, but the factor £ in Eq. (15-32) is new. From a study in
molecular spectroscopy of certain vibration-rotation bands, we find ex¬
perimental confirmation of the existence of the £ factor in Eq. (15-32).
15-8] THE HYDROGEN ATOM 337

The wave functions

in = Cne-q2,2Hn{q) (15-33)

have a wavelike character, the number of maxima plus minima being equal
to n. Note the resemblance, in this respect, to the modes of vibration of a
string. If we plot l^l2 as a function of x, it is found that when n is large
\in\2 is a maximum in the neighborhood of ±A, where A is the amplitude
of the oscillator in classical mechanics. An oscillator such as a pendulum is
observed to spend more time near the ends of its range of swing than
elsewhere, for then its velocity is less, so that, for large values of n, Born’s
probability interpretation of \\J/\2 seems reasonable. However, for small
values of n the situation is different. For example,

\io\2 = Cle~q,
which has the form of the error function, with a maximum at q = x = 0.
This does not mean that Born’s postulate is incorrect, for when n = 0 the
energy of a visible oscillator, like a spring or a pendulum, would be so
small that its amplitude would be too minute to observe. For an atomic
oscillator, such as a molecule, the important fact is that its spectral fre¬
quencies, and hence energy levels, can be observed, and that wave mechan¬
ics predicts these correctly.

15-8 The hydrogen atom. Here we shall take V = — e2/4jre0r, which

suggests that spherical coordinates are most appropriate. Therefore we
write V • V in terms of r, 0, <f>, and Schroedinger’s equation becomes

1 M
sin2 6 d(j>2)

872m
+ (15-34)

Just as when solving Laplace’s equation in spherical coordinates, we

commence the solution of this equation by putting

i = R(r)(d(0)$(i). (15-35)

Substituting this into Eq. (15-34) and multiplying through by r2 sin2 6/

#0$, we find ttiat the only term containing 4, is

1 d2®
4> d<j)2 '

*See Vol. I, Section 14-6.

338 WAVE MECHANICS [CHAP. 15

We may then put this equal to a constant of separation, say —a2, and
obtain

The solution of this equation is a function of the form

t —ia<f>

We now introduce Schroedinger’s postulate that our solution must be

single-valued. For a given r and 6, <j> and <f> + 27r, </> + 47t, etc., all refer to
the same point in space. Hence adding 2t to </> must not change the value
of <f>. This means that a must be an integer, or

a = 0, 1, 2, 3, • • • .

A quantity such as a, which is restricted to integral (or half-integral)

values differing by unity, is called a quantum number. While these numbers
were introduced almost empirically in the earlier Bohr-Sommerfeld theory,
we have seen, in this and the previous section, how they arise in wave
mechanics as a result of Schroedinger’s postulate that \p must be single¬
valued, continuous, and finite. Further integration of Eq. (15-34) leads to
the introduction of two other quantum numbers, usually designated as l
and n. Since the development is very lengthy, it is left to more specialized
texts.* The final result is that the energy E of the electron must be
restricted to the proper values

n = 1, 2, 3, • • • . (15-36)
8e2h2n2

If we combine this equation with Eq. (15-1), the correct spectral fre¬
quencies for hydrogen are obtained.
Since Eq. (15-36) was also derived by Bohr, the real success of wave
mechanics does not become apparent until we investigate a more com¬
plicated problem, such as that of the normal helium atom. Here the
Bohr-Sommerfeld theory failed, but if the methods of wave mechanics
are extended by the application of perturbation theory, f the correct energy
levels are obtained for helium.

*See, for example, Pauling and Wilson, Introduction to Quantum Mechanics,

Chapter V.
fSee Vol. I, Section 6-6.
15-9] PENETRATION OF A POTENTIAL BARRIER 339

15-9 Penetration of a potential barrier. Let us take the one-dimensional

problem where the potential function is defined by V = V0 between
x = 0 and x = L, and V = 0 for all other values of x, as shown in Fig.
15-2. Between x = 0 and x = L, which is the region labeled 2, we have
a potential barrier. If a particle of energy E approaches this barrier from
region 1 on the left, classical mechanics tells us that the particle will not
penetrate the barrier if E < V0■ Wave mechanics, however, predicts
that the particle may penetrate to region 3, the probability of its doing
so being greater the smaller (F0 — E) and L are. If E > F0, classical
mechanics predicts that the particle will always pass across the barrier,
while wave mechanics states that this is not a certainty.

Fig. 15-2. Particles reflected and transmitted by a potential barrier.

The wave-mechanical procedure is to write three Schroedinger equations,

one for each region:

*1 = o, (15-37)
ax2 n2

- (E - V,M'2 = 0, (15-38)

ax2 h2 ^ = °' (18_39)

For the solutions of these equations, we may write

= Aieikx + Bie~ikx,

*2 = A2e‘“ + B2e-“x, (15-40)

. ikx
$3 = e ,
where

i = vn, k = y V2mE, l = ~ V2m(E - Vj).

In writing the expression for ^3, we assume that no wave travels back
from infinity in region 3. The amplitude of *3 is made unity because we
are interested only in relative amplitudes.
340 WAVE MECHANICS [CHAP. 15

The A’s and B’s are determined by conditions at the two boundaries
x = 0 and x = L. One condition is that \p must not change discontinu-
ously at a boundary, or

’/'i = ^2 at x = 0, — ^3 at x — L.

The other condition is that d\p/dx must be continuous at a boundary, other¬

wise d2\j//dx2, and hence \p, would then be infinite. We thus must have

= (#h and (dh) = (dp) .

\dx/x=o \dx/x=o \dx / X=L \dx / x=l

The conditions at x — 0 give us the relations

Ai + Bi — A2 + B 2,

k(A, - B0 = l(A2 - B2),

from which we obtain

a* “ K1+i) A>+l(i ~ I)®2’ (16-41)

Bl + + D®2

Similarly the conditions at x = L lead to

eikL = + B2<r<ZJ1,

/ceaL = l(A2eilL - B2e~ilL),

or

A2=K1+1) •**■“
(15-42)

B2 = | (i _ *) e«*+M

The term .die1**, when multiplied by the time factor e~2irlvt, represents
waves in region 1 moving in the plus X-direction, or toward the barrier.
We then want to compare Ai (the amplitude of these waves) with unity
(the amplitude of the waves in region 3);

;)(1 + *)e.— +
Al~4 1 + i-
15-9] PENETRATION OF A POTENTIAL BARRIER 341

Finally, to compare the probability that a particle is in region 1 with

the probability that it is in region 3, we compute \AX\2. Putting

l = ~ V2m(V0 - E) = ik',

where k' is real when F0 > E, we have

■ i2 =1 ~ g ft ~ ff + g ft + iTcosh 2k'L■ <15-43)

When 2k'L is large, cosh 2k'L behaves like e2k’L. The important factor
then is the product of the width of the barrier and VFo ~~ E.
In a radioactive nucleus a particle has a finite chance of escaping in a
given time, even though the total energy E of the particle may be con¬
siderably less than the value of V0 at the top of the nuclear barrier (Fig.
15-3). As a result, we observe that in a sample of radioactive material a
certain fraction of the nuclei decay in any given time. This fraction is
called the decay constant. While the potential barrier for a nucleus is not
the simple rectangular one that we assumed above, the essential features
are the same. When the decay constants for different radioactive sub¬
stances are measured and compared with the corresponding energies of the
emitted particle, we find a relationship qualitatively similar to the one
above between l/|Ai|2 and y/V0 — E, namely (when 2k'L » 1),

_ -oVFo—e
oc e >

where a is approximately constant.

Fig. 15-3. Potential vs. distance for a-particle and nucleus.

342 WAVE MECHANICS [CHAP. 15

This “leak” effect is encountered elsewhere in atomic physics, for ex¬

ample, in the escape of electrons through the surface of a metal. The
explanation of this effect, forbidden in classical mechanics, is a notable
triumph for wave mechanics. The phenomenon is analogous to the rapidly
damped transmitted wave in total reflection (see problem 1 of Chapter 13).

15-10 The operator approach. We shall now consider briefly a more

formalized method of presenting quantum mechanics. Postulates will be
introduced simply because they lead in a systematic way to a method that
works.
By an observable we shall mean a familiar mechanical quantity, such as
the position, momentum, or energy of a particle, which we are accustomed
to think of as subject to experimental measurement. For each observable,
say the momentum p, we shall introduce an operator p0 and a symbol
pm. Then we shall relate these by means of an equation. The operator
indicates that a measurement is made on a quantity, and the symbol
represents the measured value of the quantity; the equation tells us
what possible values we may obtain for pm.
Postulate I defines the operators to be used. As in advanced dynamics,
we may start with the generalized coordinate q and its conjugate mo¬
mentum p. For these the operators are chosen as follows:

Observable Operator Measured value

Coordinate Qm
II
O

h d
Momentum Pm
V° ~ 2n dq

Thus, for the observable x the operator is x itself, while for linear mo¬
mentum px it is (h/2-iri) (d/dx).
Postulate II states that the only possible values which measurement of
an observable may yield are those for which the equation

(Operator) \[/ = (Measured value) \p (15-44)

leads to a solution for \p subject to the same restrictions (\f/ must be single¬
valued, etc.) as those listed for Schroedinger’s equation.
(a) Position. Equation (15-44) for the coordinate x is

xi = XrJ, (15-45)
or
(x — xm)\p = 0. (15-46)
15-10] THE OPERATOR APPROACH 343

First we should note that this equation is satisfied by x = xm, which

agrees with the previous statement that xm is the measured value of x.
Second, we see that postulate II imposes no restrictions on the possible
values xm may have. Third, observe that for a given xm we must have a
particular ^-function \f/m which, according to Eq. (15-46), must vanish for
all values of x except x = xm; in fact, this ipm behaves like the Dirac
delta function, 8(x — xm).*
(b) Linear momentum. Here the equation for \[/ is

2= *-*• ^

where pxm is a possible measured value of the linear momentum px = mx.

We integrate Eq. (15-47), considering pxm as constant for a given state,
and get
^ (JgZ* iPxmx I h

As for xm, all finite real values of pxm are possible.

(c) Angular momentum. We now write (15-44) as

2s ah(1M8)
We then integrate as in (b) to get

* =

We now find that pgm must be restricted to a discrete set of values. Adding
2t to 6, which means a complete revolution in space, must bring ^ back
to its original value if \f/ is to be single-valued at a point. Thus we must
limit pem to

n = 0, 1, 2, 3 • • • ,

which is the Bohr quantum condition, Eq. (15-13). To each value of n

corresponds a state of the system, the state being associated with the
angular momentum rih/2t and the ^-function

= Cmeine.

*Dirac’s delta function 8(q) is defined by 8(q) = 8(—q), 8(q) - 0 except for
q = 0, 8{q)dq = 1.
344 WAVE MECHANICS [CHAP. 15

(d) Energy. We must express the total energy, T + V, as a function

of the conjugate coordinates and momenta. This function is called the
Hamiltonian * To obtain the Hamiltonian operator, here designated by
H, we replace each momentum with its corresponding operator. Let E
be a possible, or measurable, value of the total energy. Equation (15-44)
now becomes
H$ = E\p, (15-49)

an equation which turns out to be equivalent to Schroedinger’s equation.

For example, in rectangular coordinates, T = mix2 + y2 + z2)/2, V =
V(x, y, z), px = mx, etc., so that the Hamiltonian function is

T V = jr— (pi + pi + pl) + E(x, y, z).

We substitute for the p’s their operators, and obtain

H =
V
G
8ir2m \dx2
—
3
+—) + F
dy2 " dz2/ '

Equation (15-49) then becomes

d2j
- = E\p,
8ir2m \dx2 dy2 ^ 522
or
8ir 2m
v• 4- (E - F)* = 0,

which is Schroedinger’s equation. From here on we proceed as before to

integrate the equation and find the allowed values of the energy E.
We have seen that Eq. (15-44) is more general than Schroedinger’s
equation, but includes it as a special case. We now turn to the more general
interpretation of \p.
Postulate III states that the expected mean of a series of measurements
on an observable such as p is

- = ///Pl^l2 dx dy dz
(15-50)
fSjWdxdydz

By properly choosing the constant C in the ^-function, dx dy dz

is usually made equal to unity; this is called normalizing the wave function,
and we shall assume it to be done.

*See Yol. I, Section 10-4.

15-11] THE UNCERTAINTY PRINCIPLE 345

Let the observable in Eq. (15-50) be x rather than p, and consider a

one-dimensional problem. Then for a particle in the state m, we have

Xm — dx.

According to Born, dx is the probability that, for the state m, the

particle will be found in the element dx, or that the coordinate of the
particle will lie between x and x + dx, so that fx\\pm\2 dx should give the
average value of x for the state m. Hence postulate III is in accord with
Born’s interpretation of \p and x corresponds to a statistical “center of mass. ”
While the position of a particle in an atomic system can be treated only
statistically, its energy Em for a given state may be found experimentally.
If we apply postulate III to Em, we have

E = §j$Em\\pm\2 dx dy dz

= Emfff\ipm\2 dx dy dz

== Em,

since ij/m is assumed normalized. The result seems obvious, but it lends
support to postulate III and points up the difference between an observable
whose discrete values may be found experimentally and one which must
(for atomic systems) be treated statistically.
For further development of the general theory of quantum mechanics
the reader is referred to one of the many specialized texts available. The
above treatment will show the broad abstract nature of the theory, which
has successfully handled such matters as the intensity and polarization of
spectral lines, electron spin, etc.

15-11 The uncertainty principle. We have repeatedly seen that in

wave mechanics all predictions are statistical, whereas in classical mechanics
it was assumed that, given sufficient data, future events could be forecast
with certainty, as in celestial mechanics. Of course, in atomic physics we
do not have the “sufficient data. ” Attempts to simultaneously measure
the position and the velocity of a particle always meet with frustration.
Suppose for example, that we decide to observe the position of an elec¬
tron. To do so, we must let at least one photon of light be scattered by the
particle. We know that an electron is too small to ‘see in a microscope
using visible light because no matter how large we may be able to make
the magnification, we are limited by the resolving power, which in turn
depends on the diffraction of the light. Since the resolving power is in¬
versely proportional to the wavelength of the light, we may increase the
practicable magnification by using radiation of shorter wavelengths, such as
346 WAVE MECHANICS [CHAP. 15

x-rays, but then the trouble is that one photon of such radiation strikes our
electron with greater momentum and energy than does one photon of radia¬
tion of longer wavelength. We have seen in the Compton effect that when a
photon is scattered by an electron, the latter recoils in some forward direc¬
tion. The energy imparted to the electron increases with the energy of the
incident photon. When we use radiation of shorter wavelength we improve
our resolving power and thus may locate an object more accurately in
position, but at the same time we increase the disturbance of the particle
and make its subsequent momentum more uncertain. Let Aq be the smallest
resolvable separation for the wavelength used, and let Ap be the corre¬
sponding uncertainty in the momentum of the scattered quantum, then the
product Aq Ap is of the order of magnitude of Planck’s constant h, or

Aq Ap ~ h. (15-51)

This is Heisenberg’s uncertainty principle. The uncertainties referred to are

inherent in the nature of light and matter, according to quantum theory;
the usual errors due to imperfect equipment and measurement are addi¬
tional to these. Since errors must be treated statistically, the numerical
factor in Eq. (15-51) depends on whether we are referring to possible errors,
probable errors, standard deviations, etc.
Let us return to the concept that the motion of a particle may be repre¬
sented by that of a group of waves with a range of frequencies and with
a different phase velocity u corresponding to each frequency. We identified
the velocity v of the particle with the group velocity of the waves, which
means that the particle moves in the same manner as a region of phase agree¬
ment of the waves. If the range in frequency, Av, is increased, the sharpness
of the wave packet will also increase. Thus increasing Av makes it possible
to decrease Aq, the uncertainty in the position of the particle; but an increase
in Av means an increase in the uncertainty AE in the energy of the particle,
since we have
E = hv, E AE = h(v -j- Ar).

An increase in AE, in turn, involves an increase in the uncertainty Ap in

the momentum.
If the waves are to build up a packet of constructive interference in a
region of length Aq (this pattern will not be repeated outside of Aq),
we must have a range AX in wavelength such that

Aq
= n, > n + 1,
T

where n is an integer. If we eliminate n between these equations, we get

15-11] THE UNCERTAINTY PRINCIPLE 347

Aq (> ~ — X. (15-52)

The momentum of the particle is related to the wavelength by the de

Broglie equation
h
V = -,
so that

|Ap| = p AX.
We then have

|Ag| \Ap\ > h (l - = h (i — y) •

For accuracy, we must have Ap/p <SC 1, or

|A$| |Ap| ~ h,

as stated in Eq. (15-51).

As a result of this inherent uncertainty, an event considered “certain”
in classical physics becomes only “highly probable” in wave mechanics,
and an event termed “impossible” in Newtonian theory is now classed as
one having a “very low probability.” This explains the “leak” effect of
Section 15-9, where the particle was assumed to have a definite energy E
and hence a definite momentum p. If we take Ap = 0, then Aq becomes
infinite, which means that the particle might be found anywhere, for ex¬
ample on the far side of the barrier. If, on the other hand, we definitely
limit the position to an uncertainty less than the width of the barrier,
then E must be regarded as uncertain, with a finite probability that E > V0,
so that the particle may escape across the barrier. It is in the nature of
things that, since Planck’s constant has a finite value, infinitely detailed
experience and sharp predictions are physically impossible. This is the
underlying principle of quantum mechanics.
348 WAVE MECHANICS [chap. 15

Problems

1. Derive the expression X = (1.24 X oscillator, plot {'f'nl2 vs. q for n = 0,

104/PM> where X *is the wavelength, 1, and 2.
in angstroms, of light whose photons 9. A particle is confined to linear
each have an energy of V electron motion in a “box,” that is, V = 0
volts. between x = 0 and x = L, while
2. Explain how the photoelectric V = °° for all other values of x. Solve
effect and Eq. (15-5) may be used to Schroedinger’s equation for this system
determine Planck’s constant experi¬ and show that the allowed energies are
mentally.
3. From the three preceding equa¬
En = nh2/8mL2, n = 0, 1, 2, • • - .
tions, derive Eq. (15-9) for the Comp¬ Find the corresponding wave functions
ton shift. 'Pn•
4. Suppose that a photon, whose 10. The rotator. A rigid system, with
“associated mass” equals that of an moment of inertia I about an axis, is
electron at rest, strikes such an electron free to rotate about this axis. Taking
in a head-on Compton collision. Find V — 0, solve Schroedinger’s equation
what fraction of the photon’s energy for this system and show that the al¬
will be transferred to the electron. lowed energies are
5. Show that when V is sufficiently
large (over a few thousand volts), so Em = n2h2/8iv2I, n — 0, 1, 2, • • •.
that one must use the relativistic ex¬
11. Taking Ap = m Ay, and Av =
pression for the kinetic energy eV of
1 m/sec (about the accuracy of an auto¬
an electron, the de Broglie wavelength
mobile speedometer), find the uncer¬
is given by
tainty Aq for (a) a 1000-kg car, (b) a
X = h/[2moeV(l + b/2)]l/2, 20-gm bullet, (c) a proton, (d) an elec¬
tron.
where b = eV/moc2. Expand this ex¬
12. If the energy of an electron is
pression as a power series (a) in b, when
known to within 0.01 ev, find the
b <<C 2, (b) in 1/5, when b )$> 2.
uncertainty Aq for an electron whose
6. From Eq. (15-22), show that
energy is (a) 1 ev, (b) 100 ev.
passing from a region where V = 0
13. Show that if At is the time taken
(no force field) to one where the po¬
by a particle in traversing the distance
tential V exists, corresponds to enter¬
Aq, and T = \mv2, then Eq. (15-51)
ing a region whose index of refraction
becomes
for the i/'-waves is
AT At ~ h.
n = V(E - V)/E.
14. Show that if = YlnCn'&n,
7. Derive Schroedinger’s equation where \P„ = ypne2wiyn* and vn = En/h,
containing the time from the equation then I'Pl2 will contain terms involving
HT? = ESP, where 4 is a function of the frequencies postulated by Bohr in
the coordinates and time, and E is now Eq. (15-1), these frequencies appearing
the operator —(h/2m)(d/dt). as “beat frequencies.”
8. (a) Find the Hermite polynomials 15. Suppose that the wave packet
Ho, Hi, and H2. (b) For the harmonic discussed in Section 15-11 is a pulse of
 PROBLEMS 349

radiation that falls on an absorbing 16. For the rectangular potential

target in the time At — Aq/c. If AE barrier of Section 15-9, take L =
is the uncertainty in the amount of 10-15 m, Vo — E = 5X 10-14 joule
energy transferred to the target, show (about 300,000 ev), V0 = 1.6 X 10~13
that joule (106 ev), and m = 6.6 X 10-27 kg
AE At ~ h. (as for alpha particles), (a) Find k' and
2k'L. (b) Evaluate l/|i4i|2.
 References

The companion volume to this book, referred to in the text as Vol. I, is

Constant, F. W., Theoretical Physics: Mechanics of Particles, Rigid and
Elastic Bodies, Fluids, and Heat Flow. Reading, Mass.: Addison-Wesley, 1954.

Theoretical Physics in General

1. Bergmann, P. G., Basic Theories of Physics. 2 volumes. Englewood Cliffs,

N.J.: Prentice-Hall, 1949.
2. Joos, Georg, and Freeman, I., Theoretical Physics. 2nd ed. New York:
Stechert-Hafner, 1951.
3. Lindsay, R. B., Concepts and Methods of Theoretical Physics. Princeton,
N. J.: Van Nostrand, 1951.
4. Menzel, D. H., Mathematical Physics. Englewood Cliffs, N. J.: Prentice-
Hall, 1953.
5. Morse, P. M., and Feshbach, H., Methods of Theoretical Physics. New
York: McGraw-Hill, 1953.
6. Page, Leigh, Introduction to Theoretical Physics. 3rd ed. Princeton, N. J.:
Van Nostrand, 1952.
7. Planck, Max, Introduction to Theoretical Physics (translated by Henry
Brose). 5 volumes. New York: Macmillan, 1949.
8. Slater, J. C., and Frank, N. H., Introduction to Theoretical Physics. New
York: McGraw-Hill, 1933.
9. Sommerfeld, Arnold, Lectures on Theoretical Physics. (English transla¬
tion.) 6 volumes. New York: Academic Press, 1949.

Thermodynamics and Statistical Mechanics

1. Allis, W. P. and Herlin, M. A., Thermodynamics and Statistical Me¬

chanics. New York: McGraw-Hill, 1952.
2. Gurney, R. W., Introduction to Statistical Mechanics. New York: McGraw-
Hill, 1949.
3. Kennard, E. H., Kinetic Theory of Gases. New York: McGraw-Hill, 1938.
4. Lindsay, R. B., Introduction to Physical Statistics. New York: Wiley, 1941.
5. Rossini, F. D. (editor), Thermodynamics and Physics of Matter. Princeton,
N. J.: Princeton University Press, 1955.
6. Sears, F. W., An Introduction to Thermodynamics, the Kinetic Theory of
Gases, and Statistical Mechanics. Reading, Mass.: Addison-Wesley, 1950.
7. Zemansky, M. W., Heat and Thermodynamics. 4th ed. New York: McGraw-
Hill, 1957.
Electromagnetism

1. Bitter, F., Currents, Fields and Particles. New York: Wiley, 1956.
2. Di Francia, Toraldo, Electromagnetic Waves. New York: Interscience,
1955.
350
 REFERENCES 351

3. Fowler, R. G., Introduction to Electric Theory. Reading, Mass.: Addison-

Wesley, 1953. j
4. Harnwell, G. P., Principles of Electricity and Electromagnetism. 2nd ed.
New York: McGraw-Hill, 1949.
5. Jeans, J. H., The Mathematical Theory of Electricity and Magnetism. 5th
ed. England: Cambridge University Press, 1925.
6. Kraus, J. D., Electromagnetics. New York: McGraw-Hill, 1953.
7. Panofsky, W. K. H. and Phillips, M., Classical Electricity and Mag¬
netism. Reading, Mass.: Addison-Wesley, 1955.
8. Peck, E. D., Electricity and Magnetism. New York: McGraw-Hill, 1953.
9. Slater, J. C. and Frank, N. H., Electromagnetism. New York: McGraw-
Hill, 1947.
10. Smythe, W. R., Static and Dynamic Electricity. 2nd ed. New York:
McGraw-Hill, 1950.
11. Whitmer, R. M., Electromagnetics. Englewood Cliffs, N. J.: Prentice-Hall,
i954.
12. Whittaker, E., History of the Theories of Aether and Electricity. London:
Nelson and Sons, 1951.

Optics

1. Ditchburn, R. W., Light. New York: Interscience, 1953.

2. Frank, N. H., Introduction to Electricity and Optics. 2nd ed. New York:
McGraw-Hill, 1950.
3. Jenkins, F. A. and White, H. E., Fundamentals of Optics. 2nd ed. New
York: McGraw-Hill, 1950.

Quantum Theory

1. Bohm, David, Quantum Theory. Englewood Cliffs, N. J.: Prentice-Hall,

1951. • AT V 1
2. Lindsay, R. B. and Margenau, H., Foundations of Physics. New York:
Wiley, 1936. .
3. Pauling, L. and Wilson, E. B., Introduction to Quantum Mechanics. New
York: McGraw-Hill, 1935.
4. Richtmyer, F. K, Kennard, E. H. and Lauritsen, T., Introduction to
Modern Physics. 5th ed. New York: McGraw-Hill, 1955.
5. Schiff, L. I., Quantum Mechanics. 2nd ed. New York: McGraw-Hill, 1956.
 ANSWERS TO PROBLEMS

CHAPTER 1
1. 250°, 350°
2. (a) 491.8°, 671.8°, (b) —459.8°F
3. 115.5°, 215.5°
4. (a) 1.183, 1.40, (b) 161.7°, 46.2°
5. 101°
6. Extrapolate p vs. T to zero pressure.
7. (a) p(Vb - Va), (b) RT In (Vb/Va)
8. (a) 1.013 X 105 n/m2, (b) 1.69 X 105 joules

CHAPTER 2

1. (a) —40 joules, (b) 10 joules, (c) —40 joules

2. 3.62 X 104 joules/kg-mole-deg
3. (a) 4.207 joules/gm-deg, 4.186 joules/gm-deg, 4.193 joules/gm-deg
(b) No, C varies.
4. (a) W > Q2- Qi, (b) W/Q2 > (T2 - Tx)/T2
5. (a) 5.45%, (b) 6%, (c) 10%
6. (a) 1250 joules, (b) 1500 joules, (c) 5
7. 625 joules, 875 joules
8. (a) 0, (b) 0.42 joule/deg

9. (a) 37.5% vs. 50%, (b) 750 joules, 500 joules, (d) —5 joules/deg,
2.5 joules/deg, 2.5 joules/deg, 0
10. (a) 0, (b) 3.33 joules/deg, (c) 3.33 joules/deg
11. (a) 0, (b) 0, (c) 0; 1000 joules vs. 0
12. —24,800 joules/kg-mole-deg
13. (a) —R In 2, (b) R In 2, (c) 0
14. (a) 0, (b) R In 2, (c) R In 2

352
 ANSWERS TO PROBLEMS 353

CHAPTER 3

1. V = [RT/(V - b)] - a/V2

2. (a) 3/2/2, (b) (3/2/2) + aR/(pV2 - a)
6. 271.6°C
7. —92,350 joules
8. (a) 24.6 m3, 600°K
(b) 12.3 m3, 300°K
(c) 17.4 m3, 424°K
(d) 49.2 m3, 1200° K
(e) 6.15 m3, 150°K
9. Wi = Qi = RT2 In (Vb/Va)
W2 = R(T2 - Ti)/(7 - 1), = 0
W3 = Qs = /m In (Fd/Ec)
tf4 = /2(ri - r2)/(7 - i), Qi — o
14. (a) V — k = T/(p), (b) Same
15. (2a/RT) - b
16. RT, RTb — a
18. (a) 3.45 X 103 joules-m3/kg-mole, 0.0235 m3/kg-mole
(b) 3.64 X 105 joules-m3/kg-mole, 0.0428 m3/kg-mole

19. 3, 3, 2
20. 3, 2, 1

CHAPTER 4

1. (a) 3.34 X 10-27 kg, (b) 5.31 X 10-26 kg

2. (a) 2.69 X 1019/cm3, (b) 2.69 X 1013/cm3
4. (a) 5 m/sec, 5 m/sec
(b) 5 m/sec, 5.39 m/sec
(c) 5.50 m/sec, 6.21 m/sec
5. (a) 1.25 X 1011, (b) 333 m/sec, (c) 354 m/sec
6. (a) 100 n, (b) 140 n
7. (a) 482 m/sec, (b) 681 m/sec, (c) 681 m/sec
8. 7457°C
11. (b) 4/vA
12. 1.33 X 10 -29 m3, 1.47 X 10~10 m
13. (a) 5.26 km, (b) 0.312
 354 ANSWERS TO PROBLEMS

14. (a) In (p/po) = (mgNo/Rk) In [(To — ky)/T0]

(b) 5.36 km
16. 1, —2, 1; proportional to T1/2

CHAPTER 5

1. (a) 1, 10, 45, 120, 210, 252, 210, 120, 45, 10, 1, (b) 34.4%
2. (a) 1, (b) 5, (c) 10, (d) 20, (e) 30
3. (a) 51.4%, (b) 18.6%, (c) 7.5%
4. —1, —1, 1260
5. (a) 221, (b) 0
6. 0.42
7. 0.818
8. 0.922
9. 377 m/sec, 425 m/sec, 461 m/sec
10. (a) 2.36 X 1023, (b) 1.21 X 1024, (c) 1.09 X 1024, (d) 3.94 X 1023
11. 2
13. kT/2
14. (a) 366 m/sec, 259 m/sec, (b) 0.52 cm, 0.73 cm
15. 3R, 4/2, 1.33
16. 772/2, 9/2/2, 1.29

CHAPTER 6

1. (a) 7.4 X 106 m/sec, (b) 3.3 m/sec, (c) 2.2 X 10~9 m/sec
2. (a) 15, (b) 20, (c) 18, (d) 12, (e) 5
3. (a) 0, (b) 2, (c) 3, (d) 0, (e) 0
6. 1.8°K
10. (8?rhc d\/\5)/(eh^kT - 1)
12. 6.71 X 10~^4 joule-sec, 1.40 X 10~23 joule/deg
13. 2.4 X 103
14. (a) 0.98, (b) 0.50, (c) 0.02

CHAPTER 7

1. (a) ML/T2Q (b) M/TQ

(c) ML/Q2 (d) MLZ/T2Q2 (e) L/T
2. 1000 n/coul in direction of B X v.
 ANSWERS TO PROBLEMS
355

3. 4 X 1033 *
4. 8.85 X 10-12 coul2/n-m2, 12.56 X 10-7 n/amp2
5. ri is dimensionless.
6. 9 X 1011 n-m2/franklin2, 10-5 n-sec2/franklin2, 9 X 1016 m2/sec2
7. (a) 10 amps (b) 9 X 1020 dyne-cm2/abcoul2
(c) 3 X 1010 cm/sec (d) 10-4 w/m2
8. (a) 3 X 109 statcoul
(b) (1/9 X 1020) dyne-sec2/statcoul2
(c) 3 X 1010 cm/sec
9. (a) 3.14 X 10~9 volt, (b) 3.14 X 10-7 volt
10. 3.5 X 10-15 w/m

CHAPTER 8
2. qx/4ireo(x2 -f- a2)3/2
3. cr/eo
4. (a) 0, (b) pr/3e0
5. — q', q + q’
8. 1.3 X 10~15 kg
9. (a) — axy, (b) 4 X 10-6 joule, (c) 0
10. —a/y, ax/y2, —2eoax/y3
11. qd cos 6/^ireor2
12. (a) (gd2/167reor3)[(3x2/r2) — 1]
(b) 3qd2xy/47reor5
13. V = (V0/L)x + k(L2x - x3)/6e0
14. V = 2n2mCnm sin (nirx/Li) sin sin (hrz/Ls), (n/Li)2 +
(m/L2)2 + (l/L3)2 = 0
15. (a) 4:ireoVoab/(b — a), —47reoEoab/(b — a)
(b) —47T€oFoab/(b — a), 47reoEo52/(^ ~ a)
16. — q
17. (a) —q at 90°, 210°, and 330°; q at 150° and 270°; (b) 0.013 q2/t0 d2
18. Behind plane: —47reoaEo[l + (o/d) + (a2/d2)]
On sphere: 47reoaFo[l + (a/d) + («2/d2)]
19. (a) 47reo«Eo, ■—^rreoaVob/d
(b) 4-]reoaVo(l + ab/d2), —i-ireoaVob/d
20. Infinite charged plane
21. Semi-infinite charged plane
22. (a) q2/8ire0a, (b) q2/8Tre0a
 356 ANSWERS TO PROBLEMS

CHAPTER 9

1. (a) 10-7 n attraction

(b) 0
(c) 0.353 X 10-7 n in+F- and — F-directions

(d) 0
(e) 0.192 X 10-7 n in — F- and -f-X-directions
2. (a) 10~° w/m2 in Z-direction
(c) 3.53 X 10~7 w/m2 in Z-direction
(d) 0
(e) 2.7 X 10~7 w/m2 at 45° with —Y- and +Z-directions
3. (a) 2 X 10-5 w/m2 in Z-direction
(c) 2 X 10~5 w/m2 in Z-direction
(d) 0
(e) 1.41 X 10-5 w/m2 at 45° with —F- and -(-^-directions
5. (a) 7.8 X 10~n m/sec, (b) 1.6 X 10-6 m
6. (a) 0
(b) 1.6 X 10 _14 n in Z-direction
(c) 1.6 X 10-14 n in — F-direction
(d) 1.13 X 10-14 n in Z-direction
(e) 1.31 X 10_14 n in FZ-plane

7. y0!R2/2{x2 + R2)3/2

8. (noIN/2l){[(l - x)/VR2 + (l - x)2] + (x/VR2 + x2)}

9. \/'2noNI /it a
10. (a) 3.5 X 107 m/sec, (b) 7 X 104 volts/m in —Z-direction

12. (a) 3.54 X 105 m/sec, (b) 7.3 X 10~3 w/m2

14. 5 X 10-9 amp
16. (a) fjLoIr/2ira2
(b) noI/2-irr
(c) mo I(c2 — r2)/2irr(c2 — b2)
(d) 0

17. k(x2 — y2)/nox2y2 in Z-direction

18. NIA

19. (mo//2)(1 - x/VR2 + x2)

21. (mo^/27t) In (R2/R1) in Z-direction
22. (yola2/Tr)[i(—y/r3) + j(x/r3)]; {2>y0Ia2/^){xz/rb),
(3yoIa2/Tr)(yz/r5), {yola2/-w){2z2 — x2 — y2)/r5
 ANSWERS TO PROBLEMS 357

CHAPTER 10

1. (a) NBAw sin 6, (b) NBAw/y/2

2. (a) l2Bo:/2, (b) 0
3. 0.157 volt
4. 26 volts clockwise; 28 volts counterclockwise
5. 14 volts clockwise; 8 volts counterclockwise
6. (a) (moH/27t) In (b/a), (b) (moZ/2tt) In (b/a)
7. 7ra2/xoA^iAr2/25
8. (a) iM)I/2ivr, (b) (nol/2ir) In (b/a)
9. (moH2/4tt) In (b/a)
10. (b) —e2BR2/Am
11. (hqI/2-k) In (r2/ri)
15. (L1L2 — M2)/(Li + L2 4= 2M)

CHAPTER 11

L (a) V• pv = -p, (b) V-VT = (pc/k)T

3. (a) V»VE = eopoE + cr/xoE
(b) V-VE* = —w2£omoE* — iWpoE*
4. V*VA = comoA — moJ
5. 377 volts/amp
7. (a) coul2-sec/kg-m3, (b) 1.04 X 1012
9. (a) 4.75 volts/m, 1.26 X 10-2 w/m2, (b) 2 X lO'10 joule/m3
10. (a) 300 m, (b) 6 X 10~3 joule, (c) 6000 watts
13. (a) (l/r2)(d/dr)[r2(du/dr)] = (l2/a2)(d2u/dt2)
14. Poynting vector directed around magnet

CHAPTER 12

1. RlR2R3/(RlR2 + R2R3 + R3R1)

2. (a) RI+.L(dI/dt) + q/C = 60^

(b) L(d2I/dt2) + R(dl/dt) + I/C = iu&0eM

3. (a) I « E3/2, (b) I x V3

4. -q(Ke - 1)/Ke
5. K2 tan 9i = K1 tan 02
 ANSWERS TO PROBLEMS
358

6. (a) 10.6 X 10 9 farad

(b) Ex = 240,000 volts/m, E2 = 120,000 volts/m; Di = D2 =
6.36 X 10-6 coul/m2
(c) Pi = 4.24 X 10~6 coul/m2, P2 = 5.30 X 10-6 coul/m2
(d) —1.06 X 10-6 coul/m2
7. (a) Ei = E2 = 200,000 volts/m; Pi = 1.77 X 10“6 coul/m2, D2
3.54 X 10~6 coul/m2
(b) 1.77 X 10-6 coul/m2 and 3.54 X 10~6 coul/m2
(c) 4.4 X 10-9 farad
9. (a) C/A = (e0/d)(l + 5e~VIE»d)
(b) W = (e072/2d) + (5e0V2/d)e-v/E°d + 5e0EoVe~v/E°d
+ 5e0Po de-v/E°d — 5e0E% d
10. (a) 0.159 n
(b) 7.97 X 10“5 joule, 15.94 X 10~5 joule
(c) -7.97 X 10-5 joule
12. (a) 4 amp-turns/m, 11, 5.53 X 10~5 w/m2, 40 amp/m
(b) 103 amp-turns/m, 1.1, 1.38 X 10~3 w/m2, 100 amp/m
(c) 104 amp-turns/m, 1.01, 1.27 X 10~2 w/m2, 100 amp/m
(d) 10, 0.1, 0.01

CHAPTER 13

1. (a) r is imaginary, cos r = ±jV(n\/n2)2 sin2 i — 1

(b) Transmitted wave represented by
g ja[<—(!/ sin r/®2)] g— al-y/(nilni)2 sin2 i—l](a:/»2)

Second term represents damping as x increases.

(c) Poynting vector has no component into medium 2.
8. 3v/2
9. Velocity must be proportional to the frequency.
12. (a) Vl\, V2P, V3P, • • • VnlX
(b) Width = Vl\/4n ; area = tIX
14. Up oscillates between limits a\ and 0 and finally becomes ai/2.

CHAPTER 14

2. (a) v = c2B/E and B' = 0.

(b) v = E/B and E' = 0.
3. (a) 50 sec, 62.5 sec, (b) 80 ft
 ANSWERS TO PROBLEMS 359

4. (b) 0.44 fringe

5. (a) 0.867c, (b) 400 lb, (c) 33' years
7. (a) p = rav = mov/vA — v2/c2
8. (a) 5.12 X 104 ev, (b) 5.12 X 105 ev, (c) 4.61 X 106 ev, (d) 5.07 X 107 ev

9. (a) 9.42 X 107 ev, (b) 9.42 X 108 ev, (c) 8.48 X 109 ev, (d) 9.33 X 1010 ev
10. (a) 2.57 X 103 ev, (b) 7.86 X 104 ev, (c) 6.60 X 105 ev, (d) 3.62 X 106 ev
11. (a) e2/87re0a, (b) a = e2/87re0m0c2, (c) 1.41 X 10~15 m, (d) 7.66 X 10~19 m

12. 6.1 X 10-9 gm

13. (a) 1.5 X 10-10 joule = 9.3 X 108ev
(b) 0.37%
(c) 3.35 X 1014 joules = 93 X 106 kwh
14. 4.67 X 10-6 n/m2

CHAPTER 15

4. S _
5. (a) X = (h/V2m0eV)( 1 — 56 + )
(b) X = (hc/eV)[ 1 - (1/6) + f(l/62)-]

8. (a)«o, aiq, ao(2q2 — 1)

11. (a) 6.6 X 10-37 m (b) 3.3 X 10~32 m
(c) 4.0 X 10-7 m (d) 7.3 X 10~4 m
12. (a) 2.5 X 10-7 m, (b) 2.5 X 10-6 m
16. (a) 2.44 X 1014/m, 0.244, (b) 0:935
 INDEX

Abampere, 132, 136 Boltzmann’s distribution law, 81 ff.

Absolute temperature scale, 19 Bose distribution function, 106
Absolute zero, 21 Bose-Einstein statistics, 99, 103 ff.
Accelerated charge, 239 Bose’s derivation of Planck’s law, 110,
Action, 295 111
principle of least, 329 Boundary conditions, for electric field,
Adiabatic process, 11 252
equation for, 45 for electric potential, 156, 159, 160
Ampere, the unit of current, 132, 172, for electromagnetic waves, 274
178 for magnetic field, 264
Ampere’s law, 132, 171, 219, 260 for waves of matter, 340
circuital or integral form of, 181 Boyle’s law, 42
point statement of, 183
Amperian currents, 125, 257 ff. Calorie, 14
Angle, of incidence, 272 Capacitance, 254
of reflection, 273 Capacitor, energy of charged, 256
Angular frequency, 230 Carnot cycle, 16
Angular momentum quantization, 343 Carnot refrigerator, 18
Annihilation of electron and proton, Carnot’s theorems, 18
317 Cavendish experiment, 170
Apparent mass, 314 Cavity radiation, 110, 111
A priori probability, 85 Cell, in phase space, 77, 102, 112
Associated mass of a photon, 323 reversible, 41
Atmosphere, variation of pressure with Charge, conservation of, 220
height, 70, 71 point, 129, 235
Available energy, 32 Charge density, 139
Average kinetic energy per molecule, Charges, group of, 148 ff.
66 ff. polarization of, 248
Average speed of molecule, 60 Circuital form of Ampere’s law, 179 ff.,
Avogadro’s number, 60 259
determination of, 71 Class A media, 244
Clausius-Clapeyron equation, 39, 40
Beattie-Bridgman equation, 51 Clausius statement of second law, 16
Bernoulli’s molecular hypothesis, 59 Coefficient, of coupling, 214
Betatron, theory of, 206 ff. of expansion, 38
Biot’s formula, 174 of mutual inductance, 210
Biot-Savart law, 176 of reflection, 275
Black-body radiation (see Cavity radi¬ of self-inductance, 211
ation) of viscosity, 73
Bohr frequency relation, 321 Coefficients, thermodynamic, 37
Bohr-Sommerfeld theory, 102, 321 Collisions, molecular, 68
Boiling point, variation with pressure Colloidal suspensions, 71
of, 40 Components, number of, 56
Boltzmann’s constant, 67, 91 Compressibility, 38
360
 INDEX 361

Compton effect, 324 Diffusion, 74 f

Conducting sphere in uniform fieldj 156 Dimensions, 126 ff.
Conductivity, electrical, 244 Dipole, electric, 147
thermal, 74 magnetic, 187
Conductor, surface charge density of, Dipole moment, 150
145 Dipoles, induced and permanent, 247
Conjugate functions in electrostatics, Dispersion of material waves, 331
164 ff. Displacement current, 225
Conservation, of charge, 220 Displacement, electric, 222, 250
of energy, 79, 311 Distribution of particles, among cells
of mass-energy, 317 in phase space, 83, 89, 106, 114
of matter, 79 with respect to energy, 87, 116
of momentum, 311, 313 with respect to speed, 85 ff., 107
Conservative field, 142 Divergence, of electric field, 141
Constitutive equations, 243, 265 of magnetic field, 188
Continuity, equation of, 220 ff. Divergence theorem, 141
Correspondence principle, 323 Domain, magnetic, 263
Coulomb, the unit of charge, 133 Double layer, 187
Coulomb’s law, 133, 137, 218, 300 Dual nature, of atomic particles, 326 ff.
integral statement of (see Gauss’ of light, 322 ff.
law) Dulong-Petit law, 97
point statement of, 141
Coupling, coefficient of, 214 Efficiency, thermodynamic, 17
Critical point, 52 Eigenwerte (proper values), 334
Curl, of electric field, 142, 209 Einstein’s mass-energy relationship,
of magnetic field, 182, 222 317
Current density, 181 Einstein’s quantum statistics of a gas,
Current element, 129, 173 106 ff.
Cycle, 9 Electric charge (see Charge)
Carnot, 16 Electric current (see Current)
Electric dipole, 147
D’Alembert’s equation, 234 Electric displacement D, 222, 250
DeBroglie’s wave mechanics, 326 ff. boundary conditions for, 252
Decay constant, radioactive, 341 Electric field, 130
Degenerate gas, 108, 115 nonelectrostatic, 197
Degradation of energy principle, 31 Electric intensity E, 130, 138
Degrees of freedom, 55, 93, 94 boundary conditions for, 252
V (del) operator defined, 144 curl of, 142, 209
Density, charge, 139 divergence of, 141
current, 181 in terms of potential functions, 144,
Diamagnetism, 208, 262 209
Dielectric constant, 251 Electric moment, 150
Dielectric media, 246 ff. Electric permittivity, 138, 251
Differential, exact, 14, 23 Electric polarization P, 247
inexact, 8 Electric potential, 142
Differentiation in thermodynamics, 34 due to continuous charge distribu¬
Diffraction, 283 tion, 152
 INDEX
362

due to group of charges, 148 ff. for ideal gas, 44

Electrical conductivity, 244 van der Waals, 51, 69
Electromagnetic energy equation, 236 virial form of, 51
Electromagnetic equations, for air or Equilibrium state, 78
vacuum, 226 Equilibrium, thermodynamic, 3
for any medium, 265 Equipartition of energy, 93 ff.
invariance of, 297, 300 Equipotential surface, 144
Electromagnetic system of units, 127, Ether hypothesis, 302
136 Exact differential, 14, 23
Electromagnetic waves, 226 ff. Exclusion principle, 99, 112
Electromotive force, defined, 134, 197, Expansion, coefficient of, 38
206 free, 25
induced, 202 Extensive variable, 5
motional, 197 ff.
Electron gas, 98, 114 ff.
Electron volt, 68, 145 Farad, the unit of capacitance, 255
Electrons, free and bound, 247 Faraday constant, 42, 72
Electrostatic problems and potential Faraday’s induction law, 134, 203, 218
theory, 145 ff. integral statement of, 206
Electrostatic system of units, 127, 136 point statement of, 209
Energy, available for work, 32 Fermat’s principle, 278 ff., 292, 329
of charged capacitor, 256 Fermi-Dirac distribution function, 114
conservation of, 12, 311 Fermi-Dirac energy distribution, 116
of current-carrying coil, 215 Fermi-Dirac statistics, 99, 112 ff.
distribution functions, 87, 116 Ferromagnetism, 263
electromagnetic, 237 Field (see Electric field and Magnetic
equation, 100, 236 field)
equipartition of, 93 ff. Fields, general remarks, 129 ff.
internal, 12 First law of thermodynamics, 12
of quantum, 323 First and second laws of thermody¬
Energy density, of electric field, 169, namics combined, 34
256 Fitzgerald (see Lorentz)
of electromagnetic field, 237, 266 Fixed points in thermometry, 7
of magnetic field, 216 Flux, magnetic, 134
Enthalpy, 38 measurement of, 203 ff.
Entropy, definition of, 23 ff. Force, between current elements, 132,
function for ideal gas, 46 171
and irreversibility, 27 of magnetic field on moving charge,
and probability, 90 ff. 175
of steam, 26 between static charges, 133, 137
and unavailability of energy, 31 Forces, types of, 129
of universe, 30 Free expansion, 25
Equation of continuity, 220 ff. Frequency associated with a material
Equation of state, 3, 50 ff. particle, 326
based on results of porous-plug ex¬ Fresnel zones, 290 ff.
periment, 50 Fresnel’s equation for reflected and
Beattie-Bridgman, 51 refracted amplitudes, 277, 278
 INDEX 363

Galilean transformation, 297, 298 Invariance of physical laws, 296

Galvanometer, moving-coil type, 179 Inversion temperature, 55
Gas, electron, 114 Irreversible process, 4
ideal, 42 ff. entropy change for, 28
photon, 108 Irrotational field of force, 142
quantum statistics of a, 106 Isobaric process, 5
van der Waals’, 68 Isochoric process, 5
Gauss’ law, 139, 250 Isothermal process, 8
Generator, electric, 199
g-factor, 85 Joule heating effect, 197, 237
Gibbs-Helmholtz equation, 41 ff. Joule, the unit of energy, 8
Gibbs’ functions, 38 Joule-Kelvin experiment, 46 ff.
Gibbs’ phase rule, 55 ff. Joule’s law for ideal gases, 42
Green’s theorem, 285
Group velocity, 281 ff. Kelvin absolute temperature scale, 19,
44
Hamiltonian operator, 344 Kelvin-Planck statement of second
Heat, 10 law, 16
of reaction, 42 Kilogram-mole, 5
Heisenberg’s uncertainty principle, Kinetic energy, of gas molecules, 66 ff.,
102, 345 ff. 93 ff.
Helmholtz equation, 231 relativistic expression for, 316
Helmholtz free-energy function, 38 Kinetic-theory concept, of pressure,
Henry, the unit of inductance, 211 61 ff.
Hermite polynomials, 336 of temperature, 66 ff.
Hertzian waves, 228 Kirchhoff diffraction formula, 288 ff.
Historical outline of electromagnetism, Kirchhoff laws for a network, 245
124 ff.
Huygens’ principle, 283 Lagrange’s method of undetermined
Hydrogen atom, theory of, 327, 337 ff. multipliers, 82
Laplace’s equation, 152 ff.
Ideal gas, defined, 42, 66 Latent heat, 37
entropy of, 46 Law of atmospheres, 71
equation of state for, 44 Laws of reflection and refraction, 274
Image, electrical, 160 Legendre polynomials, 155
optical, 279 Lenz’s law, 208
Images, method of, 160 ff. Light, speed of, 135, 228, 296, 311
Index of refraction, 266 Line integral, 134, 179
Induced currents, 196 Linear circuit, field due to, 181, 183,
Induction, electromagnetic, 196 191
Inertial systems, 295 Linear medium, 243, 251
Inexact differential, 8 Lines of force, 144
Intensity of magnetization M, 257 Liquefaction of a gas, 55
Intensive variable, 5 Lorentz-Fitzgerald contraction, 303,
Interferometer, Michelson, 302 304, 306
Intermolecular forces, 69 Lorentz force, 175
Internal energy, 12 Lorentz transformations, 298, 304 ff.
364 INDEX

Macrostate, 77, 103 Moving clocks, 308

Magnetic, deflection, 176 Multipoles, 148 ff.
dipole, 187 Mutual inductance, 210
field, 130, 173 unit of, 211
flux, 134, 201
force, 175 Newton’s laws of motion, 59, 311
induction B, 131, 173 invariance of, 296, 311
intensity H, 222, 260 Normalization, 344
moment, 257
permeability, 261 Observable quantity, 342
permittivity, 260 Ohm, the unit of resistance, 244
poles, 187 Ohm’s law, 244
scalar potential, 185 ff. Operators in quantum mechanics,
Magnetization, intensity of, 257 342 ff.
Magnetomotive force, 223 Optical length, 278
Mass, apparent or momental, 314 Orthogonality of lines of force and
longitudinal, 315 equipotential surfaces, 144, 164
variation of with speed, 314 Oscillator, simple harmonic, 100, 101,
Mass-energy equivalence, 317 334 ff.
Maxwell-Boltzmann energy distribu¬
tion function, 87 Parallel-plate capacitor, 255
Maxwell’s electromagnetic equations, Paramagnetism, 263
226, 265 Pauli’s exclusion principle, 99, 112
invariance of, 297 ff. Penetration of potential barrier, 339 ff.
Maxwell’s postulate (displacement cur¬ Permeability, magnetic, 261
rent), 135, 221 ff. Permittivity, electric, 138, 251
Maxwell’s relations in thermodynam¬ magnetic, 172, 260
ics, 38 ff. Perpetual motion machine, 15, 16
Maxwell’s speed-distribution function, Perrin’s sedimentation experiment,
86, 89 71 ff.
Maxwell-Boltzmann statistics, 76 ff. Phase change in reflection, 275
Measured value of an observable, 392 Phase rule, 55 ff.
Mechanical equivalent of heat, 14 Phase space, 77, 101
Michelson-Morley experiment, 301 ff. cell in, 77, 102, 112
Microstate, 77, 103 Phase velocity, 281, 327
Mksc system, 128 Photoelectric effect, 324
Models, use of, 187, 320 Photon, 321
Molecular beam, 90 energy of, 323
hypothesis, 59 mass of, 323
Molecule, size of, 68 momentum of, 323
Moment, electric, 150 Photon gas, 108
magnetic, 257 Planck’s constant h, 100, 102
Momentum, conservation of, 311, 313 Planck’s hypothesis, 100
of a photon, 323 Planck’s radiation law, 110 ff.
Motor, electric, 200 Plane waves, 228 ff.
Moving charge, in a magnetic field, 176 Point charge, 129, 235
magnetic field due to, 192 Poisson’s equation, 152
 INDEX 365

Polar molecules, 247 Ray optics, 271

Polarization, of a dielectric, 247 Reflection and refraction, of electro¬
charges, 248 ff. magnetic waves, 271 ff.
Polarizing angle, 278 at normal incidence, 274
Porous-plug experiment, 46 ff. at oblique incidence, 276
Potential, of a dipole, 150 Refraction (see also Reflection)
electrostatic, 142 index of, 266
scalar, 209, 233 Refrigerator, Carnot, 18
scalar magnetic, 185 Relativistic mechanics, 311 ff.
vector, 189, 209 Relativity, principle of, 296
Potential difference, 143 postulates of special theory of, 296
Power, 199 Reservoir, heat, 17
Poynting vector, 237, 266 Resistance, 244
Poynting’s theorem, 237 Retarded potentials, 235
Pressure, 5 Reversible cell, 41
critical, 52 Reversible process, 4, 28
kinetic-theory concept of, 61 ff. Root-mean-square speed, 64 ff.
of light, 319 Rotator, 348
Principle, of least action, 329
of least time (see Fermat's principle) Scalar potential, 209, 233 ff.
of relativity, 296 Schroedinger’s equation, 333, 344
of superposition, 281 for hydrogen atom, 337
Probability, 79, 90 ff. for oscillator, 334
a priori, 85 solution of, 334
interpretation of matter waves, 327, Search coil, 203
334 Second law of thermodynamics, 16
thermodynamic, 79, 103 and availability of energy, 32
Process, 4 and probability, 91
adiabatic, 11 in terms of entropy, 30
irreversible, 4, 28 Self-inductance, 211
isobaric, 5 Separation of variables, 158
isothermal, 8 Signal velocity, 282
reversible, 4, 28 Simple harmonic oscillator, 100, 334 ff.
Propagation constant, 231 Simultaneity, 307
Snell’s law, 274
Solenoid, 213, 214, 261
Quadrupole moment, 150
Solenoidal vector, 188
Quantum, of light, 321
Specific heat capacity, 14, 37
numbers, 338
of monatomic gas, 73
principle, 320
of various gases, 95 ff.
Speed (see also Velocity)
Radiation, absorption and emission of, distribution, 85 ff., 107
322 mean, 64
cavity, 100, 108 ff. root-mean-square, 64 ff.
from accelerated charge, 239 ff. Spherical waves, 286, 288
Radiation pressure, 319 Spin of electron, etc., 99, 112
Ratio of specific heats, 37, 73, 95 Statcoulomb, 136
366 INDEX

State, change of, 40, 41 Lorentz field, 298

equation of, 3, 50 ff. Lorentz space-time, 298
of a gas, 76 of velocities, 310
of an oscillator, 101 Transverse nature of electromagnetic
of a system, 3 waves, 228
Statistics, Bose-Einstein, 103 ff. Triple point, 7
classical, 76 ff.
compared, 117 ff. Unavailable energy, 32
Fermi-Dirac, 112 ff.
Uncertainty principle, 102, 345 ff.
Maxwell-Boltzmann, 76 ff. Units, systems of, 126 ff.
Steam, entropy of, 26
Universal gas constant R, 44
Stirling’s theorem, 81
Stokes’ theorem, 182
Superposition, principle of, 281 Van der Waals’ constants, 51 ff., 68 ff.
Surface charge density, 145 Van der Waals’ equation, 51, 69
Surface integral, 134 Vapor, 52
Surroundings of a system, 3, 27 Variables, extensive and intensive, 5
System, 2 thermodynamic, 3, 8, 13
closed, 3 Vector potential function, 189 ff., 209,
isolated, 28 233 ff.
Systems of units, 126 ff. Velocities, transformation of, 310
Velocity, distribution function, 85, 86,
TE and TM modes in waveguides, 233 89, 107
Temperature, 5 of electromagnetic waves, 227, 266
absolute, 6, 19 group, 281 ff., 331
centigrade, 8 of light, 135, 228, 296, 311
critical, 52 phase, 281, 327, 330
Kelvin, 19, 44 of sound, 65
kinetic-theory concept of, 66 Virial coefficients, 51
Thermodynamic probability, 79, 103 Viscosity, coefficient of, 73
Thermodynamic variables or coordi¬ Volt, the unit of potential difference,
nates, 3, 8, 13 144
Thermodynamics, defined, 1 Volume, 5
first law of, 12
second law of, 15 Wave equation, 227, 230, 272
statement of the combined laws of, time-independent, 231
34 Wave functions, 334, 337
Thermometer, gas, 6, 7, 49 Wave mechanics, 320 ff.
Throttling process, 47 Wave optics, 271 ff.
Time, relativity of, 308 ff. Waveguides, 231 ff.
Tolman’s hypothetical collision ex¬ Wavelength of atomic particle, 326,328
periment, 312 Waves, electromagnetic, 226 ff.
Toroidal solenoid, 182, 213, 261 of matter, 326 ff.
Torque on current circuit, 178 plane, 228 ff., 238
Total reflection, 293 spherical, 286, 288
Transformation, Galilean or Newton¬ Weber, the unit of magnetic flux, 131,
ian, 297, 298 174
 INDEX 367

Work, in thermodynamics, 8 ff. Zero, absolute, 21

to move a charge, 142 Zero-point energy, 115
Zonal symmetry, fields with, 155
Zartman and Ko experiment, 90 Zones, Fresnel, 290 ff.
I
 I •

>

• >

' -7

.
'
•fV

Jt
-■" t