Rast

ais

. oe
VECTOR DEFINITIONS,
IDENTITIES, AND THEOREMS

DEFINITIONS
Rectangular Coordinates
Erol ag OF ee, Oley.
I. a! oss
2 Vid = ey
Ox
S M
_ (94, _ 9A, 04, 94). (dA, OA \k
2 VAS Oy ar Oz ) Ox oy

vy? ot + Of3
: Vi Bi oe Oz?
5. V4 = VAG + aoe + W?A.k
Cylindrical Coordinates
of 1 of of|
iL: Vf= PAL ge de nn

1 0A,
2A yp Ae) + 2
p 0g

AVX
A= (OSS
10A, OA,
oe )ot oa
0A, Lyra
82a
Ten
4)
0A,

a vy = US: By nes Loy ee

p Op p Op pag? az?
2) V’A = V(V-A) —V XV X A(See Section 1.9.6.)

Spherical Coordinates

el ly
Mie ar"! 7 roo + rsin 8 dy i

pe I Peers
-A=——(r? 1 1 0A,
Meer so euerty iatSe an ede
0
ae E (A, sin 0) — oan

so He
rsin 6 (eX)

+3le OA, _ C40) 6, 42| 0A,

sin 8 d¢ or or 06

Vee igie. r
I /, of
(ao
Or
ree es, r?Ses
or r? sin 000 06 sin? § dg?
5); V?A ll= V(V-A) —V XV &X A (See Section 1.9.6.)
 “TOM Cartson)

IDENTITIES
1. VUfg) =fVe+eVvf
Dy V(4-B) = (B-V)A+ (4:V)B + BX(V XA +A4AX(V X B),

2f V-(fA) = (Vf)-A4+f(V-A),
4. V-(A X B) =B-(V X A) —A-(V X B),
=F Vi xt) = (Vx 4/0 x A),
6. V X (A X B) = (B-V)A — (4-V)B + (V-B)A — (V- ADB.
7 VXVXA= ee — V?A (See Section 1.9.6.)
oe OB,).
oe (A-V)B = |eae
ae F 2: i

aB,
+A. ax te “6gore oi
aB,
OB, OB, OB,
+| Ag+ Ars Blk

Devs Gty =!—, where the gradient is calculated at the source point (x’, y’,z’) and

r, is the iia vector from the source point (x’, y’, 2’) to the field point (x, y, z).

LOD V. (;) = a where the gradient is calculated at the field point, with the

same unit vector.

1. [ vrar = [faa
12. Je pee f AX das
where S is the surface which bounds the volume rT.

THEOREMS

1. Divergence theorem: [iA-da = [vA dr,

where S is the surface which bounds the volume r.

2. Stokes’s theorem: fpAnal = fe x A)-da,

S.
where C is the closed curve which bounds the surface

Mo= 4Tt X/o-*

> 3.95 x/07*
 Digitized by the Internet Archive
in 2021 with funding from
Kahle/Austin Foundation

https://archive.org/details/electromagneticfOO0Olorr_x6d3
ELECTROMAGNETIC FIELDS AND WAVES
 Second Edition

ELECTROMAGNETIC
FIELDS AND WAVES
Paul Lorrain
University of Montreal

Dale R. Corson
Cornell University

W. H. FREEMAN AND COMPANY

San Francisco
Copyright © 1962, 1970 by W. H. Freeman and Company

No part of this book may be reproduced by any mechanical,

photographic, or electronic process, or in the form of a
phonographic recording, nor may it be stored in a retrieval
system, transmitted, or otherwise copied for public or
private use without written permission from the publisher.

Printed in the United States of America

International Standard Book Number: 0-7167-0331-9

Library of Congress Catalog Card Number: 72-94872

LOROSS7,
Contents

Preface XIX
List of Symbols XXII

VECTORS 1
eat! Vector Algebra 2
Examples 83)
EZ Invariance 5
1.3 The Time Derivative 8
Example 9
1.4 The Gradient 10
Example 1]
Flux and Divergence. The Divergence Theorem 13
Examples 13, 16
Line Integral and Curl 16
Examples 16, 19

Lev Stokes’s Theorem 21

Example 22
1.8 The Laplacian 22
Example 22
 CONTENTS

1 Curvilinear Coordinates 23
1.9.1 Cylindrical Coordinates 24
1.9.2 Spherical Coordinates 26
1.9.3 The Gradient 27
1.9.4 The Divergence 28
1.9.5 The Curl 30
1.9.6 The Laplacian 31
1.10 Summary 3
Problems 36

2
PEEGLROSTATIG AFIELD Ss! 40
Electrostatic Fields in a Vacuum

Dal Coulomb’s Law 40

Example 4]
op) The Electric Field Intensity 42
Des} The Electric Potential 43
2.4 The Electric Field Inside and Outside Macroscopic Bodies 46
Pe) Gauss’s Law 47
Example: The Average Potential over a Spherical Surface 49
2.6 The Equations of Poisson and of Laplace SIL
Review 51
Za. Conductors a2
Example 38
2.8 Calculation of the Electric Field Produced by a Simple Charge
Distribution 53
Example: Field of a Uniform Spherical Charge Distribution wn=

DE) The Electric Dipole 61

2.10 The Linear Electric Quadrupole 64
Pall Electric Multipoles 65
AM The Electric Field Outside an Arbitrary Charge Distribution 66
Zs The Average Electric Field Intensity Inside a Sphere Containing
an Arbitrary Charge Distribution 70
2.14 Potential Energy of a Charge Distribution 72
PENS Energy Density in an Electric Field 76
Example 78
2.16 Forces on Conductors 78
Example 80
Example: Forces on a Parallel-Plate Capacitor 80
PAH Summary 81
Problems 84
 Vii

3
ELECTROSTATIC FIELDS II wl
Dielectric Materials

| The Electric Polarization P 92

Se Electric Field at an Exterior Point 92
3.2.1 The Bound Charge Densities p, and o, 94
3.2.2 The Polarization Current Density 96
33 Electric Field at an Interior Point 97
3.3.1 Electric Field Intensity E” due to the Distant Dipoles 58
3.3.2 Electric Field Intensity E’’’ due to the Near Dipoles 100
Review 101
3.4 The Local Field 102
3.5 The Electric Susceptibility x, 104
3.6 The Divergence of E. The Electric Displacement D 105
3.6.1 The Relative Permittivity ¢,. Poisson’s Equation
for Dielectrics 108
3.6.2 The Free Charge Density p; and the Bound
Charge Density p, 108
357 Calculation of Electric Fields Involving Dielectrics 109
Example: The Dielectric-Insulated Parallel-Plate Capacitor 110
Example: The Free Charge Density o;, the Bound Charge Density
o,, and the Electric Displacement D at a
Dielectric-Conductor Boundary al
Example: Dielectric Sphere with a Point Charge
at its Center 111
Example: The Bar Electret 113
3.8 The Clausius-Mossotti Equation 115
59 Polar Dielectrics 116
3.9.1 The Langevin Equation Wil7/
3.9.2 The Debye Equation 120
3.10 Frequency Dependence, Anisotropy, and Nonhomogeneity 121
Examples 122
Ba Potential Energy of a Charge Distribution in the Presence
of Dielectrics 123
Example: Energy Stored in a Parallel-Plate Capacitor 124

Forces on Dielectrics 125

Example: Force per Unit Volume on the Insulating Material
in a Coaxial Cable 127
Sy i18) Forces on Conductors in the Presence of Dielectrics 128

Example: Forces on a Parallel-Plate Capacitor Immersed

in a Liquid Dielectric 128

3.14 Summary 129

Problems 131
Vili CONTENTS

4
ELECTROSTATIC FIELDS III 138
General Methods for Solving Laplace’s and Poisson’s Equations
4.1 Continuity of V, D,, E, at the Interface Between Two Different
Media 139
4.1.1 Potential 139
4.1.2 Normal Component of the Electric Displacement 139
4.1.3 Tangential Component of the Electric Field Intensity 140
4.1.4 Bending of Lines of Force 141
4.2 The Uniqueness Theorem 142
4.3 Images 144
Example: Point Charge Near an Infinite Grounded
Conducting Plane 144
Example: Point Charge Near a Grounded Conducting
Sphere 146
Example: Point Charge Near a Charged Conducting
Sphere 149
Example: Charged Sphere Near a Grounded Conducting
Plane 150
Example: Charge Near a Semi-infinite Dielectric 153
4.4 Solution of Laplace’s Equation in Rectangular Coordinates 156
Example: Field Between Two Grounded Semzi-infinite Parallel
Electrodes Terminated by a Plane Electrode
at Potential V, 158
Example: Field Between Two Grounded Parallel Electrodes
Terminated on Two Opposite Sides by Plates
at Potentials V; and Vy 161
4.5 Solution of Laplace’s Equation in Spherical Coordinates. Legendre’s
Equation. Legendre Polynomials 163
Example: Conducting Sphere in a Uniform Electric Field 168
Example; Dielectric Sphere in a Uniform Electric Field 173
Solution of Poisson’s Equation for V 176
Example: p-n Junction Diode in Silicon 176
Solution of Poisson’s Equation for E 180
Examples 180
Summary 181
Problems 183

5
RBEEATIVIT Yel LoS
The Basic Concepts
yall The Galilean Transformation 194
ee Breakdown of the Galilean Transformation and of Classical
Mechanics at High Velocities 195
CONTENTS
1x

Example: Particle Velocities Never Exceed c 195

Example: The Addition of Velocities 195
Example: Time Dilation 195
Des} Inadequacy of the Galilean Transformation for Electromagnetic
Phenomena 196
Example: The Trouton and Noble Experiment 196
Example: Magnetic Fields 197
Example: The Jaseva-Javan-Murray-Townes Experiment 198
5.4 The Fundamental Postulate of Relativity 201
55 Invariance of a Physical Law as Illustrated by Classical
Mechanics 201
Example: Classical Mechanics Inside a Moving Train 202
Example: The Law F = ma 202
The Lorentz Transformation 203
Transformation of a Length 205
Example: The Apparent Shape of a Rapidly Moving Object 207
Transformation of a Time Interval 209
Example: The Time Read on a Rapidly Moving Clock 210
Example: The Relativistic Doppler Effect for Electromagnetic
Waves 212
Simultaneity 213
Causality and Maximum Signal Velocity 214
Transformation of a Velocity Bilis
Examples PANG, 2G
Transformation of an Acceleration 217
Relativistic Mass AAT
Transformation of a Mass 220
Relativistic Energy & 220
Example 221
The Four-Vector r el
Transformation of a Momentum and of a Relativistic Energy.
The Four-Momentum 223
Example: The Relation &? = moc* + p’c? 224
Transformation of a Force 225
Examples 226
Transformation of an Element of Volume 226
Invariance of Electric Charge 228
Example 229
Transformation of an Electric Charge Density and of an Electric
Current. The Four-current Density 229
Examples 231
The Four-Dimensional Operator (1 232
The Conservation of Charge 233
Summary 234
Problems 239
 CONTENTS

6
RELATIVITY II 247
The Electric and Magnetic Fields of Moving Electric Charges
6.1 Force Exerted on a Moving Charge by Another Charge Moving
at the Same Constant Velocity 0 248
6.2 Field of a Charge Moving at a Constant Velocity 251
6.2.1 The Electric Field 256
6.2.2 The Magnetic Field 259
Example: The Field of a 10 GeV Electron
ati — 90. 260
6.3 Transformation of Electric and Magnetic Fields 261
Example: The Parallel-Plate Capacitor 263
6.4 The Vector Potential 4 264
Example: The Vector Potential for a 10-GeV Electron
at @ = 90° 266
6.5 The Scalar Potential V. The Electric Field Intensity E Expressed
in Terms of V and 4 266
Example: V,VV, and dA/dt for a 10-GeV Electron 268
6.6 Transformation of the Electromagnetic Potentials V and 4. The Four-
Potential A 269
Example: The Parallel-Plate Capacitor 270
6.7 The Lorentz Condition 271
6.8 Gauss’s Law 271
Example 272
6.9 The Divergence of B Pe
6.10 The Curl of E 273
6.11 The Curl of B 274
Onl Maxwell’s Equations 276
Example: The Magnetic Field Near a Straight Wire Carrying
a Steady Electric Current 277
Example: The Magnetic Field of a Short Element of Wire
Carrying an Electric Current 280
Example: Force on a Second Element J dl 282
Example: force on a Moving Charged Particle 283
6.3 Summary 283
Problems 286

i
MAGNETIC FIELDS I 292
Steady Currents and Nonmagnetic Materials
ae Magnetic Forces 293
7.2 The Magnetic Induction B. The Biot-Savart Law 295
CONTENTS
xi

Example: The Magnetic Induction Due to a Current Flowing

in a Long Straight Wire 296
Example: Force Between Two Long Parallel Wires 297
Example: The Circular Loop 299
eS The Force on a Point Charge Moving in a Magnetic Field 299
Example: Hall Effect in Semiconductors 300
Example: The Hodoscope 301
7.4 The Divergence of the Magnetic Induction B 302
i) The Vector Potential 4 303
Example: The Long Straight Wire 305
Example: Pair of Long Parallel Wires 306
7.5.1 The Line Integral of the Vector Potential 4 Over a
Closed Curve 308
Lh The Curl of the Magnetic Induction B 308
ee Ampere’s Circuital Law 310
Example: Long Cylindrical Conductor 311
Example: The Toroidal Coil ils
Example: The Long Solenoid BIS
Example: Refraction of the Lines of B at a Current Sheet 316
Example: The Short Solenoid Bil
7.8 The Magnetic Dipole 319
Example: The Long Solenoid 3p
7.9 Summary B22)
Problems 323

8
MAGNETIC FIELDS II a2
Induced Electromotance and Magnetic Energy
8.1 The Faraday Induction Law B32
Example: The Expanding Loop 335

8.1.1 The Faraday Induction Law in Differential Form 336

8.2 The Induced Electric Field Intensity EF in Terms of the Vector
Potential A 337
Example: The Electromotance Induced in a Loop by a Pair of Long
Parallel Wires Carrying a Variable Current 338

8.3 Induced Electromotance in a Moving System 339

Example: The Electromotance Induced in a Fixed Loop
in a Time-dependent Magnetic Field 342
Example: The Electromotance Induced in a Loop Rotating
in a Fixed Magnetic Field 342

8.4 Inductance and Induced Electromotance 343

8.4.1 Mutual Inductance 343
8.4.2 Self-inductance 345
xii CONTENTS

Example: Self-inductance of a Long Solenoid 346

Example: Self-inductance of a Toroidal Coil 347
Example: Mutual Inductance Between Two Coaxial
Solenoids 348
8.4.3 Coefficient of Coupling 350
Example 351
135) Energy Stored in a Magnetic Field Bil
Example 352
8.5.1 Magnetic Energy in Terms of the Magnetic Induction B 353
Example 354
8.5.2 Magnetic Energy in Terms of the Current Density J;
and of the Vector Potential A 354
Examples B55
8.5.3 Magnetic Energy in Terms of the Current /
and of the Magnetic Flux ® 355
Examples 355
8.5.4 Magnetic Energy in Terms of the Currents
and of the Inductances 355
Examples 356
8.6 Self-inductance for a Volume Distribution of Current 356
Example: Self-inductance of a Coaxial Line 356
8.7 Magnetic Force Between Two Circuits 358
8.7.1 The Magnetic Force When the Currents are Kept Constant 359
Example: Force Between Two Coaxial Solenoids 361
8.7.2 Magnetic Force When the Fluxes Are Kept Constant 363
Example: Force Between Two Coaxial Solenoids 363
8.8 Magnetic Torque 364
Example: Magnetic Torque on a Current Loop 365
8.9 Magnetic Forces Within an Isolated Circuit 366
Example: The Pinch Effect 366
8.10 Magnetic Pressure 367
Example: Magnetic Pressure Inside a Long Solenoid 369
8.11 Summary 370
Problems 373

y
MAGNETIC FIELDS III 8 wo
(Fe)

Magnetic Materials

W),I The Magnetization M 384

Dee Magnetic Induction B at an Exterior Point 384
9.3 Magnetic Induction B at an Interior Point. The Divergence of B 387
Example: The B Field of a Uniformly Magnetized Cylinder 393
CONTENTS
xiii

9.4 The Magnetic Field Intensity H. Ampére’s Circuital Law 393

9.5 Magnetic Susceptibility Xm and Relative Permeability p, 395
9.5.1 The Equivalent Current Density V < M and the Free
Current Density J; 396
9.6 Hysteresis 396
9.6.1 Energy Dissipated in a Hysteresis Cycle 398
Example 400
9.7 Boundary Conditions 400
9.8 Magnetic Field Calculations 402
Example: The B and H Fields of a Bar Magnet 402
Example: The Bar Magnet and the Bar Electret 404
9.8.1 Magnetic Circuits 405
9.8.1.1 Magnetic Circuit With an Air Gap 407
Example 409
9.8.1.2 Magnetic Circuit Energized by a Permanent
Magnet 409
Example 411
9.8.2 Solution of Poisson’s Equation for B 413
Example: The Magnetic Induction at the Center
of a Rotating Disk of Charge 413
9.9 Summary 414
Problems 417

10
MAXWELL’S EQUATIONS 422
10.1 The Conservation of Electric Charge 422
Example: Charge Density in a Conductor 424

10.2 The Potentials V and A 424

10.2.1 The Retarded Potentials 427
Example: The Retarded Potentials for an Oscillating
Electric Dipole 427
Example: The Retarded Potential 4 for an Oscillating
Magnetic Dipole 428

10.3. The Lorentz Condition 432

Example: The Leaky Spherical Capacitor 434
10.4 The Divergence of EF and the Nonhomogeneous Wave
Equation for V 436
10.5 The Nonhomogeneous Wave Equation for A 437
10.6 The Curl of B 437
Example: The Displacement Current Density
in a Conductor 439

10.7. Maxwell’s Equations 439

10.7.1 Maxwell’s Equations in Integral Form 442
xiv CONTENTS

10.8 Duality 444

Examples 446
10.9 Lorentz’s Lemma 446
Examples 445, 446
10.10 The Nonhomogeneous Wave Equations for E and B 448
10.11 Summary 450
Problems 453

Ii
PROPAGATION OF
ELECTROMAGNETIC WAVES I 459
Plane Waves in Infinite Media

iil Plane Electromagnetic Waves in Free Space 460

11.1.1 The Poynting Vector 464
Examples 465
Example: Energy Flow Through an Imaginary Cylinder 465

The E and Hf Vectors in Homogeneous, Isotropic, Linear,

and Stationary Media 467
Propagation of Plane Electromagnetic Waves
in Nonconductors 470
Example 471
11.4 Propagation of Plane Electromagnetic Waves in Conducting
Media 471
11.4.1 The Poynting Vector in Conducting Media 473
Propagation of Plane Electromagnetic Waves in Good
Conductors 475
Example: Propagation in Copper at 1 Megahertz 479
Example: Joule Losses in Good Conductors 480

Propagation of Plane Electromagnetic Waves in Low-pressure Ionized

Gases 481
Example 483

11.6.1 The Conductivity o of an Ionized Gas 483

Example 485

11.6.2 The Plasma Angular Frequency w, 486

11.6.3 Wave Propagation at High Frequencies where w > w, 489
11.6.4 Wave Propagation at Low Frequencies where w < w, 491
Examples 492
Example: The lonosphere 492
Summary 493
Problems 495
 XV

12
PROPAGATION OF
ELECTROMAGNETIC WAVES II 504
Reflection and Refraction
12.1 The Laws of Reflection and Snell’s Law of Refraction 505
12.2 Fresnel’s Equations 508
: 12.2.1 Incident Wave Polarized with its E Vector Normal
to the Plane of Incidence 508
12.2.2 Incident Wave Polarized with its E Vector Parallel
to the Plane of Incidence 510
12.3 Reflection and Refraction at the Interface Between Two Nonmagnetic
Nonconductors pill
12.3.1 The Brewster Angle 515
Example: Measuring the Relative Permittivity of the Moon’s
Surface at Radio Frequencies SF
12.3.2 The Coefficients of Reflection and of Transmission at an Interface
Between Two Nonconductors Sly
12.4 Total Reflection at an Interface Between Two Nonconductors 520
Example: Light Emission from a Cathode Ray Tube 525
Example: The Critical Angle and Brewster’s Angle 526
12.4.1 Demonstration of the Validity of Snell’s Law, of the Laws
of Reflection and Refraction, and of Fresnel’s Equations
in the Case of Total Reflection Si
12.4.1.1 The Wave Numbers ki, and ky, for the Reflected
Wave 528
12.4.1.2 The Wave Numbers ky, and ky, for the
Transmitted Wave 528
12.4.1.3 The Amplitudes of E and H in the Reflected and
Transmitted Waves 529
12.4.1.4 The Poynting Vector for the Transmitted Wave 531
12.5 Reflection and Refraction at the Surface of aGood Conductor SB
Example: Communicating with Submarines at Sea 535
Example: Standing Waves at Normal Incidence 535
Example: Transmission of an Electromagnetic Wave Through a
Thin Sheet of Copper at Normal Incidence 536
12.5.1 Demonstration of the Validity of Snell’s Law, of the Laws of Reflection
and Refraction, and of Fresnel’s Equations at the Interface Between
a Nonconductor and a Good Conductor 540
12.5.1.1 The Wave Numbers ke; and ke, for
Refraction in a Good Conductor 541
12.5.1.2 The Amplitudes of E and H in the Reflected and
Transmitted Waves 542
12.6 Radiation Pressure at Normal Incidence on a Good Conductor 543
Examples 546
12.7. Reflection of an Electromagnetic Wave by an Ionized Gas 547
12.8 Summary 549
Problems all
Xvi CONTENTS

Ils)
PROPAGATION OF
ELECTROMAGNETIC WAVES III Spy
Guided Waves
13.1 Propagation in a Straight Line 557
13.1.1 TE and TM Waves 560
13.1.2 TEM Waves 561
13.1.3 Boundary Conditions at the Surface of Metallic Wave Guides 565
13.2 The Coaxial Line 566
13.3. The Hollow Rectangular Wave Guide 568
13.3.1 The TE Wave 569
13.3.2 Internal Reflections 575
13.3.3 Energy Transmission Si
13.3.4 Attenuation 578
13.4 Summary 582
Problems 585

14
RADIATION OF ELECTROMAGNETIC WAVES ae3s.
14.1 Electric Dipole Radiation 595
14.1.1 The Scalar Potential V 597
14.1.2 The Vector Potential 4 and the Magnetic Field
Intensity H 598
14.1.3 The Electric Field Intensity E 601
14.1.4 The Average Poynting Vector and the Radiated Power 603
14.1.5 The Electric and Magnetic Lines of Force 605
14.1.6 The KX Surface 607
14.2 Radiation from a Half-wave Antenna 611
14.2.1 The Electric Field Intensity E 611
14.2.2 The Magnetic Field Intensity H 614
14.2.3 The Average Poynting Vector and the Radiated Power 614
14.3. Antenna Arrays 616
14.4 Electric Quadrupole Radiation 620
14.5. Magnetic Dipole Radiation 423
14.5.1 The Potentials V and A 424
14.5.2 The E and H Vectors 624
14.5.3 The Average Poynting Vector and the Radiated Power
14.6 Magnetic Quadrupole Radiation 426
14.7. The Electric and Magnetic Dipoles as Receiving Antennas
14.8 The Reciprocity Theorem 629
14.9. Summary 633
Problems 635
CONTENTS
XVli

APPENDIXES 641
A. Conversion Table 643

B. The Complex Potential 644

B.1 Functions of the Complex Variable 644
B.2. Conformal Transformations 646
B.3 The Function W (z) as a Complex Potential 647
- Example: 648
B.4 The Stream Function 649
Example: The Parallel-plate Capacitor 651
Example: The Cylindrical Capacitor 651
Example: Field of Two Parallel Line Charges of Opposite
Polarities 653
Example: Field of Two Parallel Conducting Circular Cylinders
of Opposite Polarities 654
Problems 656

C. Induced Electromotance in Moving Systems 657

Example: Experiment 1 657
Example: Experiment 2 660
Example: Experiment 3 662

D. The Exponential Notation 665

D.1 The jw Operator 665
Example: Solving a Linear Differential Equation with Constant
Coefficients, Using the Exponential Notation 668

E. Waves 670
E.1 Plane Sinusoidal Waves 670
E.2. Waves on a Stretched String. The Differential
Equation for an Unattenuated Wave 674
E.3 Solution of the Differential Equation for an Unattenuated Wave
by the Separation of Variables 676
E.4_ Reflection of a Wave on a Stretched String at a Point
Where the Density Changes from p; to p» 677
E.5 Waves on a Stretched String with Damping. The Differential
Equation for an Attenuated Wave 679
E.6 Solution of the Differential Equation for an Attenuated Wave
by the Separation of Variables 681
E.7 Wave Propagation in Three Dimensions 682
E.8 Wave Propagation of a Vector Quantity 683
E.9 The Nonhomogeneous Wave Equation 683

ANSWERS 685
INDEX 699
~~”
7
ro‘ee Gi Vv
wi 6
Preface

This is a revised version of /ntroduction to Electromagnetic Fields and Waves

by the same authors. The general level is unchanged, despite the fact that
the phrase ‘Introduction to” has been omitted in the title.
Like the first edition, this book is intended primarily for students of
Physics or Electrical Engineering at the junior or senior levels, although
some schools will prefer to use it with first-year graduate students. The book
should also be useful for scientists and engineers who wish to review the
subject.
The major change has been the addition of two chapters on relativity.
Those who are pressed for time may omit these without losing continuity.
The other chapters have been largely rewritten, with many additions and
deletions. There are about 100 new figures; as in the first edition, three-
dimensional objects and phenomena are represented as such, and the field
maps (such as those in Chapter 4) have been plotted on a computer. Most of
the 140 examples and most of the 413 problems are new.
The aim of this book is to give the reader a working knowledge of the
basic concepts of electromagnetism. Indeed, as Alfred North Whitehead
stated, half a century ago, “Education is the acquisition of the art of the
utilization of knowledge.” This explains the relatively large number of ex-
amples and problems. It also explains why we have covered fewer subjects
more thoroughly. For instance, Laplace’s equation is solved in rectangular
and in spherical coordinates, but not in cylindrical coordinates.

CONTENTS
So as to reduce the mathematical requirements, we have included a chapter
on vectors (Chapter 1), a discussion of Legendre’s differential equation
xXx PREFACE

(Section 4.5), an appendix on the technique that involves replacing cos wf

by exp jwf, and an appendix on wave propagation.
After the introductory chapter on vectors, Chapters 2, 3, and 4 describe
electrostatic fields, both in a vacuum and in dielectrics. All of Chapter 4
is devoted to the solution of Laplace’s and of Poisson’s equations.
Chapters 5 and 6 discuss relativity. They may be omitted without losing
continuity. Indeed, this book presents both the conventional and the relativistic
approaches to electromagnetism, which are, in a sense, complementary. Chap-
ter 5 is a short exposition of the basic concepts of special relativity, with
little reference to electric charges. It requires nothing more, in the way of
mathematics, than elementary differential calculus and the vector analysis
of Chapter 1. Chapter 6 contains a demonstration of Maxwell’s equations that
is based on Coulomb’s law and on the Lorentz transformation and which is
valid only for the case where the charges move at constant velocities.
Chapters 7 and 8 deal with the conventional approach to the magnetic
fields associated with constant and with variable currents. Here, as elsewhere,
references to Chapter 6 may be disregarded.
Chapter 9 contains a discussion of magnetic materials that parallels, to a
certain extent, that of Chapter 3 on dielectrics.
In Chapter 10, the Maxwell equation for the curl of B is rediscovered,
without using relativity. This is followed by a discussion of the four Maxwell
equations, as well as of some of their more general implications. The point
of view is different from that of Chapter 6, and there is essentially no repe-
tition.
The last four chapters, 11 to 14, concern various applications of Max-
well’s equations: plane waves in infinite media in Chapter 11, reflection and
refraction in Chapter 12, guided waves in Chapter 13, and radiation in
Chapter 14. The only three media considered in Chapters 11 and 12 are
perfect dielectrics, good conductors, and low-pressure ionized gases. Simi-
larly, Chapter 13 is limited to the two simplest types of guided wave, namely
the TEM mode in coaxial lines and the TE;,) mode in rectangular guides.
Chapter 14 discusses electric and magnetic dipoles and quadrupoles, as well
as the essential ideas concerning the half-wave antenna, antenna arrays, and
the reciprocity theorem.
For a basic and relatively simple course on electromagnetism, one could
study only Chapters 2, 3 (less Sections 3.3, 3.4, 3.8, 3.9, and 3.10), 4 (less
Sections 4.4 and 4.5), 7, 8, 9 (less Section 9.3 but conserving the equation
V-B = 0), and 10. For a rather advanced course, on the other hand, Chapters
2, 3, 4, 5, 7, 8, and 9 could be reviewed briefly using the summaries at the
end of each chapter. One would then start with Chapter 6, and then go on to
Chapter 10 and the following chapters. There are, of course, many other
possibilities.
In Chapter (2, Sections 12.3 and 12.7 could be dispensed with. They
 PREFACE Xxi

involve the application of Fresnel’s equations to particular cases and

are not
essential for the remaining chapters. Chapter 13 is instructive, both because
of the insight it provides into the propagation of electromagnetic waves
and
because of its engineering applications, but it is not required for understand-
ing Chapter 14. Finally, Chapter 14 is based on Chapter 10 and on the first
two sections of Chapter 11.

PROBLEMS
This second edition contains an extensive collection of problems, many of
which are drawn from the current literature of Physics and Engineering. Such
problems should serve as a clear indication that electromagnetism has a
wealth of applications in a wide variety of fields. They should favor the
transferability of what has been learned, and they should develop creativity.
Few persons will spend the time and energy required for solving all of the
longer problems. Those left unsolved should provide some profitable reading.
Double daggers (1) indicate those problems for which hints are given.
It is suggested that the problems be solved on a computer when this is
feasible.

UNITS AND NOTATION

The units used are those of the Systéme International d’Unités (designated
Sl in all languages), adopted in Paris at the 1960 Eleventh General Conference
on Weights and Measures. This system, which is based on the meter, kilo-
gram, second, ampere, kelvin, and candela, is essentially the same as the
Giorgi, or MKSA, system in the field of electromagnetism.
The exponential function for periodic phenomena can be either exp (jwf)
or exp (—jwf), since the real part of both of these functions is cos wt. We
have used the positive exponent, which is preferable at this level; otherwise,
an impedance becomes R — /X instead of the conventional R + jX.

ACKNOWLEDGMENTS
The authors owe much to Gilles Cliche, to Gaétan Marchand, and to Paul
Carriere who performed the numerical calculations and assisted in the design
of the figures for the first edition. Several other persons assisted in the
preparation of this second edition. Francois Lorrain, Luce Gauthier-Labonté,
and Gilles Labonté all checked the complete manuscript and proposed in-
numerable improvements. Francois Lorrain and Guy Basque performed the
numerical calculations required for the new figures. Jean and Michel Barrette
checked the answers of many of the problems, Roland Savage drew the
one-hundred-odd new figures, and Angele Elias typed over a thousand pages
of text with admirable care. Despite the invaluable assistance of all these
xxii PREFACE

persons, the undersigned is clearly responsible for any errors or inaccuracies

that may remain.
The authors are grateful to the many students who proposed corrections
and modifications to the first edition. They are especially grateful to the many
persons who took the trouble to send their comments and lists of errata.
Among these it is a pleasure to mention Professors C. Rutherford Fischer of
Adelphi College; Gaston Fischer of the University of Montreal; R. H. Good
of California State College, Hayward; W. M. Nunn, Jr., of the Florida Insti-
tute of Technology; Frank E. Rose of Flint College; Paul A. Smith of Coe
College; Herschel Weil of the University of Michigan and R. E. Worley of
the University of Nevada.
The number of bits of information in a book such as this is truly enor:
mous, and errors are bound to occur. Further feedback from readers will
therefore be appreciated.
Finally, I wish to thank the Universidad de Madrid, where I was on
sabbatical leave during the final stages of the preparation of the manuscript,
and to thank Maximino Rodriguez Vidal and his group for many fruitful
discussions.
I note with regret that Dale Corson’s duties as Provost, and later as
President of Cornell University prevented him from participating in the
preparation of this second edition.

May 1969 PAUL LORRAIN

List of Symbols

SPACE, TIME, MECHANICS

Element of length Frequency a= Vl
or distance dl, ds, dr Angular frequency Oui
Total length or dis- Velocity u, v, U,
tance PS ar 1
Element of area da 1 (= W/o}?
Total area Acceleration a = 0u/0t
Element of volume Angular velocity (63)
Total volume Angular accelera-
Solid angle tion a = 0w/dt
Normal toa surface Mass m
Wave length Density p
Wave length in free Curvilinear
space coordinate q
Wave length of a Four-vector r
guided wave Ng Momentum P
Radian length A = K/2n = 1k, Four-momentum p
Wave number k =k, — jki Moment of inertia I
F
Attenuation Force
constant k; Torque r
Attenuation Pressure p
distance 5 = 1/k; Energy W
Time Tt
Period T= 1p

ELECTRICITY AND MAGNETISM

Quantity of Volume charge
electricity Q density p
Velocity of light C—O 8 aE Surface charge
meters/second density o
XXiv LIST OF SYMBOLS

Linear charge Avogadro’s number Na = 6.023

density v, x 10?8/mole
Electric potential, Molecular weight MM
scalar potential V Boltzmann constant A = 1.381 x 10773
Induced Electro- joule/kelvin
motance, Voltage Electronic charge B= IOP 3
Electric field coulomb
intensity E Planck’s constant
Electric displace- divided by 27 # = 1.055 & 107-*4
ment D Vector potential A
Permittivity of Four-potential A
vacuum (cy == weasels < Oe Magnetic induction B
farad/meter Magnetic field
Relative intensity H
permittivity €; Magnetic flux p
Permittivity @ a IDE Permeability of
Q of a medium Q vacuum fii he KbHt
Electric dipole henry /meter
moment Pp Relative perme-
Electric polariza- ability Ly
tion, electric Permeability
dipole moment Magnetic dipole
per unit volume P moment per
Molecular unit volume M
polarizability a Magnetic suscepti-
Electric bility Xm
susceptibility Xe Magnetic dipole
Electric quadrupole moment m
moment q Resistance R
Electric current I Capacitance G
Volume current Self-inductance IE
density J Mutual inductance M
Four-current Resistivity p
density J Conductivity o
Surface current Poynting vector S
density r

MATHEMATICAL SYMBOLS
Approximately Real part of z Re z
equal to awe Imaginary part of z Im z
Proportional to ac Modulus of z |z|
Factorial n n! Decadic log of x log x
Exponential of x e7, exp (x) Natural log of x In x
LIST OF SYMBOLS XXV

Arc tangent x arc tan x Unit vectors in

Complex conjugate cylindrical
of z zs coordinates Pi, Pi, 21
Vector E Unit vectors in
Gradient v spherical
Divergence YD coordinates ri, 91, Yi
Curl Vx Unit vector alongr rm
Quad Oo Field point 35 We
Laplacian We Source point eas
Unit pce in Average E
Cartesian
coordinates Alp te
ELECTROMAGNETIC FIELDS AND WAVES
CHAPTER |

VECTORS

We shall discuss electric and magnetic pneu etin terms of the yields of

It is therefore essential that you acquire at the very outset a thorough

understanding of the mathematical methods required to deal with fields. This
is the purpose of the present chapter on vectors. Note that the concept of field
and the mathematics of vectors are essential not only to electromagnetic
theory but also to most of present-day physics.
We shall assume that you are not familiar with vectors and that a
thorough discussion is required.
If, on the contrary, you are already quite familiar with vector analysis,
you may wish to omit Sections 1.1 to 1.8 and concentrate your attention on
Section 1.9, which deals with curvilinear coordinates.
Mathematically, a field is a function that describes a physical quantity
at all points in space. In scalar fields this physical quantity is completely
specified by a single number for each point. Temperature, density, and elec-
tric potential are examples of scalar quantities that can vary from one point
to another in space. For vector fields both a number and a direction are
required. Wind velocity, gravitational force, and electric field intensity are
examples of such vector quantities.
Vector quantities will be indicated by boldface type; lightface type will
indicate either a scalar quantity or the magnitude of a vector quantity.
We shall follow the usual custom of using right-hand Cartesian coordinate
 Figure 1-1. A vector A and the
three vectors A,i, A,j, A.k,
which, when placed end-to-end,
are equivalent to A.

systems as in Figure 1-1: the positive z direction is the direction of advance

of a right-hand screw rotated in the sense that turns the positive x-axis into
the positive y-axis through the 90° angle.

1.1 VECTOR ALGEBRA

A vector can be specified by its components along any three mutually perpen-
dicular axes. In the Cartesian coordinate system of Figure 1-1, for example,
the vector A has components 4,, A,, Az.
The vector A can be uniquely expressed in terms of its components
through the use of unit vectors i, j, k, which are defined as vectors of unit
magnitude in the positive x, y, z directions, respectively:
A= Ajit Ajj + Ak. (1-1)
The vector A is the sum of three vectors of magnitude A,, A,, A., parallel
to the x-, y-, z-axes, respectively.
The magnitude of A is

A = (4; + Ay + AZ)”. (1-2)

The sum of two vectors is obtained by adding their components:

Ach B= y+ BMY Be Ga (1-3)

Subtraction is simply addition with one of the vectors changed in sign:

A= B= AS (8) =o 2,)i + Ay — By Gt — Boke (4)

We shall use two types of multiplication: the scalar, or dot product; and
the vector, or cross product. The scalar, or dot product, is the scalar quantity
obtained on multiplying the magnitude of the first vector by the magnitude
 Figure 1-2. Two vectors A and B in
the xy-plane. The vector C is their
vector product 4 X B.

of the second and by the cosine of the angle between the two vectors. In
Figure 1-2, for example,
A-B = ABcos(¢ — 6) (1-5)
It follows from this definition that the usual commutative and distribu-
tive rules of ordinary arithmetic multiplication apply to the scalar product:
A-B = B.-A, (1-6)
and
A-(B+ C)=A-B+A4-C. (1-7)
From the definition of the scalar product it follows that
fi=1, jj=Hl, kek =1, (1-8)
g-k=0, kt=0, ij = 0. (1-9)
Then
A-B = (Axi + A,j + Azk)-(B.i + B,j + Bk), (1-10)
ards ated Bs. (111)
It is easy to check that this result is correct for two vectors in a plane, as in
Figure 1-2:
A-B = ABcos(¢ — 6) = ABcos¢cosé+ ABsingsin#é, (1-12)
= A,B, + A,B. (1-13)

Example A simple physical example of the scalar product is the work W done
by a force F acting through a displacement s:W = Fs.

The vector product, or cross product, of two vectors is a vector whose

direction is perpendicular to the plane containing the two initial vectors and
 Figure 1-3. An example of vector
multiplication. The torque T of the
force F about the point Ois r X F.
This vector has a magnitude of rF sin 6
and is oriented as shown.

whose magnitude is the product of the magnitudes of those vectors and the
sine of the angle between them. We indicate the vector product thus:
AXB=C, (1-14)
where the magnitude of C is
C = |ABsin(¢ — 6)|, (1-15)
with » and 6 defined as in Figure 1-2. The direction of C is given by the right-
hand screw rule: it is the direction of advance of a right-hand screw whose
axis, held perpendicular to the plane of A and B, is rotated in the sense that
rotates the first-named vector (A) into the second-named (B) through the
smaller angle.
The commutative rule is not followed for the vector product, since
inverting the order of A and B inverts the direction of C:
AX B= —(B X A). (1-16)
The distributive rule, however, is followed:

AX(B+0=(4XB)+(AX ©). (1-17)

This will be shown in Problem 1-7.
From the definition of the vector product it follows that

ixi=0, jxj=0, kX k=0, (1-18)

and, for the usual right-handed coordinate systems, such as that of Figure 1-1,

ixpek: jXk=i, kXti=j, j Xi —==—k, andso on: (1-19)

Writing out the vector product of A and B in terms of the components,
we have

A xX B= (A. “te A,j =. A,k) x (B.i =n Byj a B.k), (1-20)

 Figure 1-4. Another example of vector
multiplication. The area of the paral-
lelogram is A X B = S. The vector S
is normal to the parallelogram.

ADGB(A2B, Ath (A,B. A.B); (A,B, = A,B) 1-21)

ieiaek
ea ba bor ie (1-22)
Beep een,
We can check this result for the two vectors of Figure 1-2 by expanding
sin (g — 6) and noting that the vector product is in the positive z direction.

Examples A good physical example of the vector product is the torque T pro-
duced by a force F acting with a moment arm r about a point O, as
in Figure 1-3, where T = r X F.
A second example is the area of a parallelogram, as in Figure
1-4, where the area S = A X B. The area is thus represented by a
vector perpendicular to the surface.

Many other multiplication operations could be defined. The scalar and

the vector product operations as defined here are unique, however, in that
they are useful in describing real physical quantities.

1.2 INVARIANCE

You have probably noticed that we have two definitions for the scalar prod-
uct: Equations 1-5 and 1-11.
The first definition clearly does not refer to a particular coordinate sys-
tem and depends solely on the magnitudes of the two vectors and on the
angle between then y that is independent of the coordinate system

The second definition, Eq. 1-11, involves the components of A and of B

 Figure 1-5. Two coordinate
systems. The x’, y’, z’ system is
obtained from x, y, z by a
translation. Corresponding
coordinate axes, such as Ox and
O’x’ are parallel, but the origins
O and O’ do not coincide.

in some particular coordinate system. Since this second definition is deduced

from the first one, it should also be invariant. For example, given a force F
and a displacement s, this second definition should always give the correct
numerical result for any right-handed Cartesian coordinate system. If it does
not, then it is of no interest to us.
We shall show that the second definition, Eq. 1-11, really is invariant.
First, let us consider two right-handed coordinate systems x, y, z and x’, y’,
z’, as in Figure 1.5, one system being displaced, but not rotated, with respect
to the other. It is quite obvious that the components of F and of s are the
same in the two systems and that Eq. 1-11 therefore leads to identical results
in the two systems.
Figure 1-6 shows a more general type of relation between the x, y, z and
x’, y’, z’ systems, involving both a translation and a rotation.
Since we have already disposed of translation, we need only consider
rotation. Now any rotation of a coordinate system can be decomposed into

Figure 1-6. The x’, y’, z’ system

is obtained from x, y, z by a
translation and a rotation. The
two origins O and O’ do not
coincide and corresponding axes
are not parallel.
 Figure 1-7. The x’, y’, z’ system is
obtained from x, y, z by a rotation
about the z-axis.

three rotations performed successively about three different axes in space. It

will therefore suffice to investigate whether Eq. 1-11 is invariant under only
one such rotation, namely the rotation illustrated in Figure 1-7, where the
x’- and y’-axes are tilted at an angle @ with respect to the x- and y-axes. For
this condition the coordinates x, y, z and x’, y’, 2’ of a point P are related as
follows:
x’ = xcosé+ ysin8é, (1-23)
y’ = —xsinéd + ycos 8, (1-24)
ZR aZ, (1-25)
x = x’ cosé — y’sin8@, (1-26)
y= x sine yy cos ?, (1-27)
Zire, (1-28)
and the components of the vector
A=Ai+A,j+ Ak = Api’ + Ayj’ + Avk’ (1-29)
obey similar equations of transformation:
Az = A, cos 6+ A, sin 6, (1-30)
Ay = —A; sin 6 + A, cos 6, (1-31)
Ay= Ay (1-32)
A, = A, cos @ — A, sin 4, (1-33)

Ay = Aw sin 6 -+ A,’ COS, (1-34)

A, = Ay. (1-35)

Similar equations apply to the vector B. If we use the equations of trans-

formation in
A,B, ain A,B, = A.B., (1-36)
A-B —
8 VECTORS

we should obtain a similar equation in terms of the primed components, if

the expression on the right is invariant. We find indeed that
doB = AB od BAB = AB A Be Bee oT)
We have therefore shown that the scalar product, as defined in Eq. 1-11,
is an invariant under both a translation and a rotation of the coordinate axes.
Now what about the vector product? It has also been defined in two
different ways, first in Eq. 1-14 and then in Eq. 1-22.
From the first definition, the vector product of two vectors is an invariant
since it depends only on the magnitude of the two vectors and on the angle
between them. Is the vector product as defined in Eq. 1-22 also an invariant?
It should be, since Eq. 1-22 follows directly from Eq. 1-14. We have again
the same equations of transformation 1-30 to 1-35, and the unit vectors i, j,
k must also be transformed to 7’, j’, k’ according to the following rule:
i = cos #i’ — sin 67’, (1-38)
j = sin i’ + cos 6’, (1-39)
k=k’. (1-40)
We find that

Dia ae ’ jj’ #K
SSN ei in Re el (1-41)
Roath Bis (ie beeen
The cross product as defined in Eq. 1-22 is therefore also invariant.

1.3 THE TIME DERIVATIVE

We shall often be concerned with the rates of change of scalar and vector
quantities with both time and space coordinates, and thus with the time and
space derivatives.
The time derivative of a vector quantity is straightforward. In a time At,
a vector A, as in Figure 1-8, may change by AA, which in general represents
a change both in magnitude and in direction. Since AA has components A4A,,
AA,, and AA,,
AA = AA,i + AA,j + AA. (1-42)
On dividing AA by Ar and taking the limit in the usual way, we arrive at
the
definition of dA/dt:
 Figure 1-8. A vector A, its incre-
ment AA, and their components.

dA me CAGE
AN) (2)
a At : (1-43)
24 A Ad u PAA Je ’
eo At Oa
Lehn
Sidaes ik athe
ae © (1-45)
The time derivative of a vector is thus equal to the vector sum of the time
derivatives of its components.

Example The time derivative of the position r of a point is its velocity v, and
the time derivative of v is the acceleration a.

We again have two definitions, Eq. 1-43, which is necessarily invariant,

and Eq. 1-45, which refers to a given coordinate system. Is the right-hand
side of Eq. 1-45 invariant? It is not affected by a translation of the coordinate
system, since this changes neither the unit vectors i, j, k nor the components
A,, A,, A,» We need only check whether or not Eq. 1-45 is affected by a
rotation of the coordinate axes. If we express Eq. 1-45 in terms of the primed
coordinates, using the equations of transformation 1-33 to 1-35 for A,, A,,
A,, and 1-38 to 1-40 for i, j, k, we find that

The derivative of a vector with respect to time, as defined in Eq. 1-45,

is therefore an invariant.
10 VECTORS

1.4 THE GRADIENT

We shall be interested in one particular function of the space derivatives of a

scalar quantity (the gradient) and in two particular functions of the space
derivatives of a vector quantity (the divergence and the curl). Again, many
other such functions could be defined, but these particular ones are unique
in that they are useful to describe certain physical quantities.
Let us consider a scalar quantity that is a continuous and differentiable
function of the coordinates and has the value f at a certain point, as in Figure
1-9. We wish to know how f changes over the distance dl measured from that
point. We know that

OY =aor
ae Ox oh:
ay dy. :
(1-47)

Now df is the scalar product of two vectors A and dl with

SeTies
ay ,
(1-48)

dl = dxi-+ dyj. (1-49)

The vector A, whose components are the rates of change of fwith distance
along the coordinate axes, is called the gradient of the scalar quantity f. Gradi-
ent is commonly abbreviated as “grad,” and the operation on the scalar Pia
defined by the term gradient is indicated by the symbol V, called “‘del.’’ Thus

A = gradf=Vf. (1-50)
In three dimensions the operator V is defined as

(1-51)

Figure 1-9. The quantity f, a function

of position, changes fromf to f+ df
Seeds over the distance dl.
1.4 THE GRADIENT
11

The partial differentiations indicated are to be carried out on

whatever scalar
quantity stands to the right of the V symbol. Thus

va ue a si a ah, (1-52)
and

IVf| = |(2) ae (2) + zy)". (1-53)

The magnitude of df can also be written as

df = Vf-dl = |\Vf|\dl| cos 6, (1-54)

where @ is the angle between the vectors Vf and dl.
We can now ask what direction we should choose for dl in order that
df shall be maximum. The answer is: the direction in which cos@ = 1 or
6 = 0, that is, the direction of Vf. The gradient of f is thus a vector whose
magnitude and direction are those of the maximum space rate of change of f,
To summarize, the gradient of a scalar function is a vector having the
following properties.

(1) Its components at any point are the rates of change of the function
along the directions of the coordinate axes at that point.
(2) Its magnitude at the point is the maximum rate of change of the
function with distance.
(3) Its direction is that of the maximum rate of change of the function.
(4) It points toward Jarger values of the function.

The gradient is thus a vector point-function derived from a scalar point-

function.

Example As an example of the gradient, consider Figure 1-10, in which E,

the elevation above sea level, is a function of the x- and y- coordi-
nates measured on a horizontal plane. Points of constant elevation
may be joined together by contour lines. The gradient of the eleva-
tion E at a given point then has the following properties: (a) it is
perpendicular to the contour line at that point, (b) its magnitude is
equal to the maximum rate of change of elevation with displacement
measured in a horizontal plane at that point, and (c) it points toward
an increase in elevation.

Again we have two definitions: the gradient is a vector whose magnitude

and direction are those of the maximum space rate of change of f, and it is
also the vector of Eq. 1-52.
12

Figure 1-10. Topographic map of

a hill. The numbers shown give
the elevation E in meters. The
gradient of E is the slope of the
nill at the point considered, and
it points toward an increase in
elevation: VE = (0E/dx)i +
(OE/dy)j. The arrow shows VE
at one point where the elevation
is 400 meters.

The latter definition will be invariant if the operator V is invariant, that

is, if

ae OM aOt ae Om on Gees :
Vomias! Ws ay Jigs oma at ate ag eer ee SoEE):
The scalar quantity f is of course invariant.
Let us check whether V is invariant under a rotation of axes. We have

of _ of ax, af ay’ | af az!

dxam 0 0x dp One dz dn 20}
or, from Eqs. 1-23 to 1-25,

oe
ay = COS 6 ee
a7 cn
Sin 6 ay? 7
(1-57)
Thus
0 0 ; ta]
ae cos 6 =; — Sa (1-58)

Similarly,
0 : 0 0
ay = Sin ox? + cos @ ay” (1-59)

CA
oz ue oe
Using these last three equations, together with the equations of transforma-
tion for the unit vectors i, j, k, we do find Eq. 1-55.
We have shown that the V operator is invariant and therefore that the
gradient Vf is also invariant.
 13

1.5 FLUX AND DIVERGENCE.

THE DIVERGENCE THEOREM
¢

It is often necessary to calculate the flux of a vector through a surface.

The
flux d® of a vector A through an infinitesimal surface da is defined as

caamenas me: (1-61)

he flux atis the componen O Re vector normal to the nirics
aLaSIee by da. For a finite surface we find the total flux by integrating
A-da over the entire surface:

(1-62)
It should be obvious that this integral is invariant since the scalar product
A-da is invariant.
For a closed surface bounding a finite volume, the vector da is taken to
point outward.

Example Let us consider fluid flow. We define a vector pv, p being the fluid
density and v the fluid velocity at a point. The flux of pv through
any closed surface is the net rate at which mass leaves the volume
bounded by the surface. In an incompressible fluid this flux is always
equal to zero.

The outward flux of a vector through a closed surface can be calculated

either from the above equation or as follows. Let us consider an infinitesimal
volume dx, dy, dz and a vector A, as in Figure 1-11, whose components 4,,
A,, A, are functions of the coordinates x, y, z. We consider an infinitesimal
volume and first-order variations of the vector A.
The value of A, at the center of the right-hand face can be taken to be the
average value over the entire face. Through the right-hand face of the volume
element, the outgoing flux is

Fie (4.4 as
ue) ay dz, (1-63)

since the normal component of A at the right-hand face is the x-component

of A at that face.
14

Figure 1-11. Element of volume

dx, dy, dz around a point P,
where the vector 4 has the value
illustrated by the arrow.

At the left-hand face,

0A, dx ‘

The minus sign before the parenthesis is necessary here because, 4,i being
inward at this face and da being outward, the cosine of the angle between the
two vectors is —1. The net outward flux through the two faces is then

Fist: G25, = as ax vine 24sdr, (1-65)

where dr is the volume of the infinitesimal element.
If we calculate the net flux through the other pairs of faces in the same
manner, we find the total outward flux for the element of volume dr to be

APDio, = ( (1-66)
OA 0A, 0A,
u )dr.
Ox 1 oy nk OZ

Suppose now that we have two adjoining infinitesimal volume elements

and that we add the flux through the bounding surface of the first volume to
the flux through the bounding surface of the second. At the common face the
fluxes are equal in magnitude but opposite in sign, and they cancel. The sum
then, of the flux from the first volume plus that from the second is the flux
through the bounding surface of the combined volumes. The total outward
flux is
dA, , 0A, y , OA;:
aon =3(-— dA, Boia
, 0A, . OA H\ Me tearie
a (= ! ay us t) dn + (= 7 dy 4 mn), de ae
To extend this calculation to a finite volume, we sum the individual fluxes
1.5 FLUX AND DIVERGENCE. THE DIVERGENCE THEOREM 15

for each of the infinitesimal volume elements in the finite volume and the total
outward flux is
0A, , OA 0A
Biottot = | Co+ !
eS): a, :
(1-68)

At any given point in the volume, the quantity

0A, 0A, 0A,
(Ox Pay e)
is thus the outgoing flux per unit volume. We call this the divergence of the
vector A at the point.

According to the rules for the scalar product, we write the

where the operator V is as defined in Eq. 1-51.

The divergence of a vector as defined above must be invariant since, as
we have shown, both the operator V and the scalar product are invariant.
We can also check the invariance of V-A by substituting primed quantities
for A,, A,, A., 0/0x, 0/dy, 0/0z, using the transformation equations 1-33 to
1-35 and 1-58 to 1-60. This demonstration is formally identical to that of the
invariance of the scalar product.
The operator V- has physical meaning, not by itself, but only as it
operates on a function standing to the right of it. The symbol V is not a
vector symbol, of course, but it is convenient to use the notation of the
scalar product to indicate the operation that is to be carried out.
In Eq. 1-68 the total outward flux is also equal to the surface integral of
the normal outward component of A, thus

iz da- . a
AW te fe
[(C43 Browns
4 EVE|
f y - zs —
/ V-Z
A dr ‘
(1-70)
-

This is the divergence theorem. Note that the left-hand side involves only the
values of A on the surface S, whereas the right-hand side involves the values
of A throughout the volume 7 enclosed by S. This is also called Green's
theorem.
If the volume + is allowed to shrink sufficiently, so that V-A does not
vary appreciably over it, then

| Aaa = (V:A)r, (1-71)

S
16 VECTORS

and the divergence can therefore be defined as

V-A= fame A-da. (1-72)

707 JS

As we have seen, the divergence is the outward flux per unit volume as the
volume 7 approaches zero.

Examples In an incompressible fluid, V -(pv) is everywhere equal to zero since

the outward mass flux per unit volume is zero.
Within an explosion, V -(pv) is positive.

1.6 LINE INTEGRAL AND CURL

The integrals
b b b
[ +a [ axar and fra.

evaluated from the point a to the point b on some specified curve, are ex-
amples of line integrals.
In the first one, which is especially important, each element of length dl
on the curve is multiplied by the local value of A according to the rule for the
scalar product. These products are then summed to obtain the value of the
integral.

Example The work W done by a force F acting from a to 6 along some speci-
fied path is
W = [? Feat,
a
(1-73)
where both F and dl must of course be known functions of the
coordinates if the integral is to be evaluated analytically. Let us
calculate the work done by a force F, which is in the y direction and
has a magnitude proportional to y, as it moves around the circular
path from a to 6 in Figure 1-12. Since
Feaipj and dla ueieew 7,
a, [Fa » ikkydy = sia (1-74)

A vector field A is said to be conservative if the line integral of A-dl

around any closed curve is zero:

fara =0 (1-75)
 17

Figure 1-12. The force F is propor-

tional to y, and its point of application
moves from a to b. The work done is
given by the line integral of F-dl over
x the circular path.

The circle on the integral sign indicates that the path of integration is
closed.
Let us calculate the value of the above integral in the more general case
where it is not equal to zero.
For an infinitesimal element of path dl in the xy-plane, and from the
definition of the scalar product,
A-dl = A, dx + A, dy. (1-76)
Thus, for any closed path in the xy-plane and for any A,

fpa-at = fa. dx + fa, dy. (1-77)

Now consider the infinitesimal path in Figure 1-13. There are two con-
tributions to the first integral on the right-hand side of the above equation,
one at y — (dy/2) and one at y + (dy/2):

1 dA, dy a ( 0A, 2) ; 1-78

fpAadx = (4.-%4 y)dx Aaya 5 ORR (1-78)

There is a minus sign before the second term because the path element at
y + (dy/2) is in the negative x direction. Therefore

fA Aes = aye (1-79)

Similarly,
aA
fpAy dy _ a 9Ay dx dy, 1-80
(1-80)
and
0A OA
Pal
fA-dl = (a y a = )dx\Waib'e dy 1-81
(1-81)
18

Figure 1-13. Closed, rectangular

path in the xy-plane, centered
on the point P(x, y, 0) where
the vector A has the value illus-
trated by the heavy arrow. The
integration around the path is
performed in the direction of the
light arrows.

for the infinitesimal path of Figure 1-13.

If we set
dA, OA;
83 ae ay (1-82)

then

f a-at = g; da, (1-83)

where da = dx dy is the area enclosed on the xy-plane by the infinitesimal

path. Note here that this is correct only if the line integral is evaluated in the
positive direction in the xy-plane, that is, in the direction in which one would
have to turn a right-hand screw to make it advance in the positive direction
along the z-axis.
Let us now consider g; and the other two symmetrical quantities as the
components of a vector

GA, P0A,\ & dna TUAy\ 0A, O0Az\

(= whit (S a )i+(Z a) ke ey

which may be written as

ieee
0 0 0
vx43 | ee (1-85)
A eden
he quantity g; is then its z component.
1.6 LINE INTEGRAL AND CURL 19

If we now consider the element of area as a vector da pointing in the

direction of advance of a right-hand screw turned in the direction chosen for
the line integral, then
da = dak (1-86)
and

fa-dl = (7X Ada. (1-87)

This means that the line integral of 4-dl around the edge of an element
of area da is equal to the scalar product of the curl of A by this element of
area, as long as we observe the above sign convention.
We have arrived at this result for an element of area dx dy in the xy plane.
Is this result general? Does it apply to any small area, whatever the orientation
with respect to the coordinate axes? It does if it is invariant. We have already
seen that the scalar product is invariant. Thus the above line integral is also
invariant. We have also seen that the V operator and the vector product are
invariant. If you wish, you can check to see whether V X A really is invariant.
Equation 1-87 does apply to any element of area da and

© X A,= tim Lh Anat:

S—0 S
(1-88)
the component of the curl of a vector normal to a surface S is equal to the
line integral of the vector around the boundary of the surface divided by the
area of the surface when this area approaches zero.

Example Let us consider a fluid stream in which the velocity v is proportional

to the distance from the bottom. We set the z-axis parallel to the
direction of flow, and the x-axis perpendicular to the stream bottom,
as in Figure 1-14. Then
i, = 0) (Ds, => (0), 0, = CX. (1-89)

We shall calculate the curl from Eq. 1-88. For (V X v): we choose a
path parallel to the yz-plane. In evaluating

f v-dl

around such a path we note that the contributions are equal and
opposite on the parts parallel to the z-axis, hence (V X v); = 0.
Likewise, (V X v). = 0.
For the y-component we choose a path parallel to the xz-plane
and evaluate the integral around it in the sense that would advance
a right-hand screw in the positive y direction. On the parts of the
path parallel to the x-axis, v-dl = 0 since v and dl are perpendicular.
20

Figure 1-14. The velocity v ina

viscous fluid is assumed to be in
the direction of the z-axis and
proportional to the distance x
from the bottom. Then
Vee — — Gy:

On the bottom part of the path, at a distance x from the yz-plane,

(pee Dsl exN ze (1-90)

whereas at (x + Ax)

fae v-dl = —c(x + Ax) Az. (1-91)

For the whole path,

f v-dl = —c Ax Az, (1-92)

and the y-component of the curl is

f vedl —c Ax Az
(1-93)
ea ee Sead Aca ia
Calculating V X v directly from Eq. 1-85,
toy ok
Gy ye) te)
V <i) = ax ay az aa al 1 (1-94)
OD Omecs
which is the same result as above.

We have used the operator V for the gradient of a scalar point-function

and for the divergence and curl of a vector point-function. In all three cases,
V is defined by a single expression, and we obtain the gradient, the divergence,
or the curl by performing the appropriate multiplication. This relatively simple
situation is peculiar to the Cartesian coordinate system. As we shall see later
on, other coordinate systems do not permit a single definition for the operator
1.7 STOKES’S THEOREM 21

V but lead to more complicated expressions for the gradient, the divergence,
and the curl.

1.7 STOKES’S THEOREM

Equation 1-87 is true only for a path so small that V X A can be considered
constant over the surface da bounded by the path. What if the path is so large
that this condition is not met? The equation can be extended readily to arbi-
trary paths. We divide the surface—any surface bounded by the path of
integration in question—into elements of area day, daz, and so forth, as in
Figure 1-15. For any one of these small areas,

f Aral = (V X A)-da,. (1-95)

We add the left-hand sides of these equations for all the da’s, and then we
add all the right-hand sides. The sum of the left-hand sides is the line integral
around the external boundary, since there are always two equal and opposite
contributions to the sum along every common side between adjacent da’s.
The sum of the right-hand sides is merely the integral of (V X A)-da over the
finite surface. Thus, for an arbitrary path,

fpaval= [e x A)-da. (1-96)

This is Stokes’s theorem. It relates a line integral to a surface integral over
any surface bounded by the line integral path.
Figure 1-15 illustrates the sign convention.

Figure 1-15. An arbitrary surface

bounded by the curve C. The sum
of the line integrals around the
curvilinear squares shown is equal
to the line integral around C.
22 VECTORS

Example Under what condition is a vector field conservative? In other words,

under what condition is the line integral of 4-dl around an arbi-
trary closed path zero? From Stokes’s theorem, the line integral of
A-dl around an arbitrary closed path is zero if V K A = O every-
where. This condition can be met if
VAG—OVfe (1-97)
For then

A,
_ of
as ax’ A,
=a of
oy
_ of
A; rei az
:
(1 98)

and
2 2

(V XA), = dara ede oy OT a 0, (1-99)

dy oz Oydz Ozdy
and so on for the other components of the curl.
The field of A is therefore conservative if A can be expressed
as the gradient of some scalar function f.

1.8 THE LAPLACIAN

The divergence of the gradient is of great importance in electromagnetic

theory, as well as in many other parts of physics. Since

vp oeLi i+ ee
Le, (1-100)
then
2 2 EY)
ViVi = Vi = Sat ot (1-101)
The divergence of the gradient is the sum of the second derivatives with
respect to the rectangular coordinates. The quantity V-Vf is abbreviated to
Vf, and is called the Laplacian of f. The operator V? is called the Laplace
operator.
The Laplacian is invariant because it is the result of two successive in-
variant operations.

Example We shall show in the next chapter that the Laplacian of the electro-
static potential is equal to zero in regions where there is zero space
charge.

We have defined the Laplacian of a scalar point-function f. It is also

useful to define the Laplacian of a vector point-function A:
V4 = V2A,i + V2A,j + V°A, k, (1-102)
The Laplacian of a vector is also invariant.
 23

1.9 CURVILINEAR COORDINATES

It is frequently convenient, because of the symmetries that exist in certain

fields, to use coordinate systems other than the rectangular one. Of all the
other possible coordinate systems, we shall restrict our discussion to cylin-
drical and spherical polar coordinates, the two most commonly used.
We could calculate the gradient, the divergence, and so on, directly in
both cylindrical and spherical coordinates, but it is less tedious and more
useful to introduce first the idea of orthogonal curvilinear coordinates.
Consider the equation

S(% y, Z) = 4, (1-103)
in which q is a constant. This equation determines a family of surfaces in
space, each member of the family being characterized by a particular value
of the parameter g. An obvious example is x = qg, which determines the sur-
faces parallel to the yz-plane in Cartesian coordinates.
Consider now the three equations

Ail, ys Z) — Cal

FAX, Y, Z) = 42s (1-104)

SX; Vs Z) = 43,

which are chosen so that the three families of surfaces are mutually perpen-
dicular, or orthogonal. A point in space can then be defined as the intersection
of three of these surfaces, one of each family; the point is completely defined
if we state the values of gi, qo, gs corresponding to these three surfaces. The
variables qi, go, gs are called the curvilinear coordinates of the point, as in Fig-
ure 1-16.
Let us call di an element of length perpendicular to the surface q. This

Figure 1-16. Element of volume in

curvilinear coordinates. The unit
vectors uj, V1, Ww are respectively
normal to the qi, q2, gz surfaces at the
Kaa : point P(q, g2, q3). They are mutually
oe a, perpendicular and are oriented in such
Goa a way that wi X v1 = UW.
24 VECTORS

element of length is the distance between the surfaces gq: and q, + dq in the
infinitesimal region considered. The element of length d/, is related to dq, by
the equation
dh, = h dg, (1-105)
in which /, is, in general, a function of the coordinates q, q2, qs. Similarly,
dl, = h, dq, (1-106)
dl; = hg dq. (1-107)
In the Cartesian system of coordinates, /,, 4), hz are all unity.
The unit vectors u;, 11, Ww, are of unit length, normal respectively to the
Gis 925 3 Surfaces, and oriented toward increasing values of these coordinates.
They are chosen so that uw. X m = wy.
The orientation of the three unit vectors depends, in general, on the point
of space considered. Only in rectangular coordinates do they all have fixed di-
rections.
The volume element is
dr = dl, dl; dl; = hy hy h3(dqy dq, dq;). (1-108)
We can now find the q’s, the h’s, the elements of length, and the elements
of volume for cylindrical and spherical coordinates.

1.9.1 Cylindrical Coordinates

In cylindrical coordinates, as in Figure 1-17, the position of any point P in
space is specified by p, y, z. The coordinate p is the perpendicular distance
from the z-axis; ¢ is the azimuth angle of the plane containing P and the

Figure 1-17. Cylindrical coordi-

nate system.
 25

Figure 1-18. Element of volume

in cylindrical coordinates.

z-axis, measured from the xz-plane in the right-hand screw sense; and z is
the distance from the xy-plane. The q’s are thus p, ¢, z in this case.
At the point P there are three mutually orthogonal directions specified
by three unit vectors: p; in the direction of the perpendicular from the z-axis
extended through P; yg; perpendicular to the plane containing the z-axis and
P in the direction corresponding to increasing y; and z in the positive z
direction. These unit vectors do not maintain the same directions in space as
the point P moves about, but they always remain mutually orthogonal.
The vector describing the position of P is

rs pmb ea | (1-109)
Note that the angle ¢ does not appear explicitly on the right-hand side; it is
given by the orientation of the unit vector py.
Elements of length corresponding to infinitesimal changes in the coordi-
nates of a point are important. If the coordinates ¢ and z of the point P are
kept constant while p is allowed to increase by dp, P is displaced by dr =
dp pi. On the other hand, if p and z are held constant while ¢ is allowed to
increase by dy, then P is displaced by dr = pdg ¢1. Finally, if p and ¢ are
held constant while z is allowed to increase by dz, then dr = dz x. For arbi-
trary increments dp, dg, dz,

dr =dppit pdegi+ dzm — (1-110)

and
(1-111)
. Figure 1-18 shows the volume element whose edges are the elements of
26 VECTORS

length corresponding to infinitesimal increments in the coordinates at the

point P of Figure 1-17

dr =pdpde dz. (1-112)

1.9.2 Spherical Coordinates

In spherical coordinates the position of a point P is specified by r, 6, ¢, where
r is the distance from the origin, 6 the angle between the z-axis and the radius
vector, and ¢ the azimuthal angle. At the point P the unit vectors are shown
in Figure 1-19: r; is in the direction of the radius vector extended through P,
6, is perpendicular to the radius vector in the plane containing the z-axis and
the radius vector, and gy is perpendicular to that plane. Again, these unit
vectors do not maintain the same directions in space as the point P moves
about.
The vector r describing the position of P is now simply r = rr,, the
coordinates @ and ¢ being determined by the orientation of the unit vector ry.
The distance element dl corresponding to arbitrary increments of the
coordinates is

— dl= drry +rdo6,+ rsin bdo os, (1-113)

dl = [arp + Pao? + sin? dey}?. (1-114)
The volume element at point P is shown in Figure 1-20, and

dr = r*sin 6drdéde. (1-115)

Figure 1-19. Spherical coordi-

nate system.
 27

Figure 1-20. Element of volume

in spherical coordinates.

The table below shows the correspondence of curvilinear coordinates to

Cartesian, cylindrical, and spherical coordinates.

Curvilinear Cartesian Cylindrical Spherical

1 x p if
q2 J 2 0
43 Zz Z 2)
hy 1 1 1
hy 1 p r
h, 1 1 r sin 6
ul i P1 ry

V1 i 71 A;
wi k zy 91

It should be noted that the angle ¢ in both cylindrical and spherical

coordinates is undefined for points on the z-axis.
We are now in a position to find the gradient, divergence, and curl
operators in curvilinear coordinates. Once these are found, it will be a simple
matter to find the operators in cylindrical and spherical coordinates, for then
we need only substitute the appropriate values of qi, q2, qs, Mm, M2, fs, Us, U1,
Wy.

1.9.3 The Gradient

To find the form of the gradient, we require the rate of change of a scalar
functionf in each of the three coordinate directions:
28 VECTORS

vp= hint 0Ft 0Foo (1-116)

1

IE
a hy aq ct Uk 3s eee
uj, Se hy qo V1 ae hy 0q3 U1, (1-117 )

where wy, V1, Ww, are the unit vectors as in Figure 1-16.
In cylindrical coordinates,

ogy ae ON ea (1-118)

while, in spherical coordinates,

On the z-axis, p and sin @ are both zero and neither of these expressions
is valid.
We must ask ourselves again whether or not these expressions are in-
variant. Let us just check the one for cylindrical coordinates.
Let us first check whether it is invariant to a change from cylindrical
coordinates to Cartesian coordinates. You can do this yourself by using the
transformation equations
(1-120)
(1-121)
You will find that Vf in cylindrical coordinates does transform into the proper
expression for Vf in Cartesian coordinates, so that Vf is invariant to a change
from cylindrical to Cartesian coordinates. The inverse is equally true.
Now what about invariance to a change from one system of cylindrical
coordinates S; to another system of cylindrical coordinates S,? The gradient
is also invariant under this transformation for the following reason. Call C,
and C, the corresponding Cartesian systems. We have just shown that we have
invariance in going from S, to C;. Now we also have invariance in going from
C; to Cy, as we showed earlier in Section 1.4, and invariance in going from
C, to Sp. It follows that the gradient is invariant if one changes from S; to S».
The gradient expressed in spherical coordinates is similarly invariant to
a change from a spherical system of reference to either a Cartesian system, or
to a cylindrical system, or to a different spherical system.

1.9.4 The Divergence

To find the divergence, we consider the volume element of Figure 1-21. The
quantity Aj is the gq; component of the vector A at the center of the volume
 29

Figure 1-21. Element of volume

P(41,92.43) bere A
in curvilinear coordinates cen-
eae tered on the point P(q:, q2, 43),
IA where the vector A has the value
ee ee Ha peal ; shown by the arrow.

element; /,, /», hz are the h values at the center. Then the outward flux
through the left-hand face is

d®; = — Ayr hor hgr dqo dqa, (1-122)

== — (4.
(4.- 24d)
ee] (p,
lee shed)
tee] ie (, _ an
dai
ae ke a
It must be remembered that /. and h3 may be functions of gq, as well as Aj.
If we neglect differentials of order higher than the third, then

d®, = — Axhyhy dqz dq; atPars

(Aah 2) 5 dq3. (1-124)

By a similar argument,
0 da
d®p = Ahh; dgz dqz + aa:(Ayhyh3) ne dqo dqs (1-125)

for the right-hand face. The net flux through this pair of faces is then

0
d®rp = aa (Ajhohs) dqi dgz dqs. (1-126)
1 y

The same calculation can be repeated for the other pairs of faces to find
the net outward flux through the bounding surface. Dividing by the volume
of the element then gives the divergence. Since we have considered only
differentials up to the third order, we have already passed to the limit r— 0,
and

arpainhale E (Aiahs) + 3a.(Asli) + 5.(As huh) | (1-127)

30 VECTORS

Substituing the h’s and q’s gives the divergence in cylindrical coordi-_

3p PAP) Tes oa eae (1-128)

(1-129)

aL aA
aa . male (7? sin 0A,) + 25
STA — rae], (1-130)
a¢ sin 0A,) + 2

These expressions are again not valid on the z-axis where p and sin @ are
zero.

1.9.5 The Curl

From the fundamental definition given in Eq. 1-88,

(V X Ah = oyth sat (1-132)

where the path of integration must lie in the surface defined by qi = constant,
and where the direction of integration must be related to the direction of the
unit vector uw; by the right-hand screw rule. For the paths labeled a, b, c, d in
Figure 1-22, we have the following contributions to the line integral:

(a) — Ash; dqs,

(b) (4+
get
5% ae 2)(is- coas) dq,
va

Figure 1-22. Path of integration for

component 1 of the curl in curvilinear
coordinates.
1.9 CURVILINEAR COORDINATES 31

(c) Aghy dqp,

(@)d (42+ etdA24 da)(ie+ phe”

ad iy. 2 dap)dp
oh

Then, adding up these four terms and neglecting the higher order differentials
because we are interested in the limit S — 0,
1 0 as
F)
V =
Mars) hohs dqz dgs fee(Ashs) digo ds 0q3 (Ashe) de den
| NESS
ot? ito 0
= lols E (Ashs) = Ags (Ash) | (1-134)

Corresponding expressions for the other two components of the curl can
be found either by proceeding again as above or by proper rotation of the
indices. Finally,
hyuy hov, haw

1 0 0 0 : 2
Va eA
Iyhzhs3|0g. 9q2 9s (1-135)

hy A, hy A» hg As

Foroyiical eoordinates, Pi pP1 1

ila 0 .
V x A = Al ee ee b (1-136)

Ay pAw -A,

whereas, forsphericalcoordinates, ry 10, r sin 091

1 OmmnO) 0
V XA andlor 36 ay CD)
Aw TA Frsil 0A.

These definitions are valid everywhere except on the z-axis.

1.9.6 The Laplacian

We calculate the Laplacian of a scalar function fin curvilinear coordinates by
combining the expressions for the divergence and for the gradient:
Vf =V-Vf. (1-138)
1 0 (hohs | x ee a 9 (= v) |. 1-139
a Iyhiohs Ee (hy Og a O02 hy 0q2 0g3 hs 0q3 ( = )
32 VECTORS

For cylindrical coordinates,

except on the z-axis and, for spherical coordinates,

again except on the z-axis.

The divergence, the curl, and the Laplacian are all invariants.
We have already seen in Section 1.8 that the Laplacian of a vector A in
Cartesian coordinates is itself a vector whose components are the Laplacians
of A,, A,, A;. It will be shown in Problem 1-28 that
Vex V XA] Wied) — Vd (1-143)
is then an identity in Cartesian coordinates. In the more general case of
curvilinear coordinates, V?A is defined as the vector whose components are
those of V(V-A) — V X V X A, and not as a vector whose components are
the Laplacians of 41, A», A3.

1.10 SUMMARY

In Cartesian coordinates a vector quantity is written in the form

A=A,i+4,j+ 4.k, (1-1)
where i, j, k are unit vectors directed along the x, y, z axes respectively.
The magnitude of the vector A is the scalar

A= (Ai + Ay + A). (1-2)

Vectors can be added and subtracted:

A+ B=(A,+ Bi + (4, + B)j + (4. + Bk, (1-3)

A — B=(A, — B,)i+ (A, — B,j + (A. — BR. (1-4)
The scalar product (Figure 1-2)

A-B = AB cos (¢ — 9), (1-5)

= A,B, + A,B, + A.B., (1-11)
obeys the commutative and distributive rules:
1.10 SUMMARY
33

A-B = B.-A, (1-6)

A(B+C)=A4-B+4-C. (1-7)
The vector product (Figure 1-2)
Ax BEG, (1-14)
C = ABsin (¢ — 8), (1-15)
follows the distributive rule
AX(B+C)=AXB+H+AXC, (1-17)
but not the commutative rule:
AX B=-BX A. (1-16)
The time derivative of A is
dd dA Pee Amar GA:
(1-45)
at "at 7a aegis:
The del operator is defined as follows:

V- 0 : cs) : ) ?
se Cape T 927°
(1-51)
and the gradient of a scalar function f is written
i naw (1-52)
eas ose
The gradient gives the maximum rate of change of f with distance at the
point considered, and it points toward larger values of f.
The flux ® of a vector A through a surface S is

Q = [Ada (1-62)
Ss

For a closed surface the vector da is chosen to point outward.

The divergence of A,

dA, , 0A, , 0A (1-69)

Ea Arn a ay Ee
is the outward flux of A per unit volume at the point considered.
The divergence theorem states that

[esa | saa, (1-70)

T S

where S is the surface bounding the volume r.

34 VECTORS

The Jine integral

b
i A-dl

over a specified curve is the sum of the terms 4-dl for each element dl of the
curve between the points a and b and, for a closed curve C which bounds a
surface S, we have Stokes’s theorem:

fpaval= [io < A)-da, (1-96)

C Ss
where
is
0 0 (a)
VxXA= oo ay ae (1-85)

A, A, A;

is the curl of the vector function A. The above surface integral is evaluated
over any surface bounded by the curve C.
The Laplacian is the divergence of the gradient:

ViaOf ae ue
,54+
oF, ,
& (1-101)
The Laplacian of a vector in Cartesian coordinates is defined as
V?A = V?A,i + V2A,j + V2A_ k. (1-102)
In cylindrical coordinates, Figures 1-17 and 1-18,
r = pp, + zm, (1-109)
dr =dppit+pdegitdzn, (1-110)
dr = p dp de az, (1-112)

vf= Foto at ea, (1-118)

V a p p ¢ 4
f = a oe AF Fi a5 az? (1-129)

Pl pi Fi

iL|)@) 0 0
V xX A = alas Sden de I E
(1-136)

Ay pas Ay
. NS) feSR RIES of
VFaf = one~ | ap: = ae at a7 : 3
(1-140)

These last four equations are meaningless on the z-axis, where ps0:
1.10 SUMMARY 35

In spherical coordinates, Figures 1-19 and 1-20,

[PEA Ms

dr = dr ry --- rade 0, + rsin 6 dy 71, (1-113)

dr = r’ sin
6 dr d6 do, (1-115)
moh 1 of lof
Ai easeat ae (-119)

no 0A,
V4 === =
oa ae eea
rain 0g" Ga)
ry 10, rsin 6g;

1 ay 8} 0
re eee | BED)
A, YAg rsin 6A,

i
ZeOl SO) COLO 0; mm Oy. S
1 of =
ep r or ay or? a alt) ver 06? ipr?sin? @ dg?” CRE)

These last four equations are also meaningless on the z-axis, where sin 6 = 0.
In other than Cartesian coordinates, VA is defined as follows:
VA=V(V-A)-—VXVXA.
We shall have occasion to use the following operations:*

AOL AE
ERETAME
V(4A-B) = (B-V)A+ (A:V)B+BX(V XA+AXW XB);
V-(fA) = (Vf)-A4+f(V-A);

V-(A X B) = B-(V X A) — A-(V X B);

Var ay Vi )o6 A (Vex A);
V X (AX B) = (B-V)A — (A-V)B + (V-B)A — (V- ADB;
VXVXA=V(V-A) — V°A (see Section 1.9.6);

(A-V)B = |4 oe ! Ayet Ae| +|45;PY AyetA al

aB, OB,
ait |4.52 sine. y * 4 k;

* See J. Van Bladel, Electromagnetic Fields (McGraw-Hill, New York, 1964), Appendixes
1 and 2, for an extensive collection of vector identities and theorems.
 VECTORS
36

the unit vec-

where V’ (*)is calculated at the source point (x’, y’, 2’) and ri is
r
tor from the source point (x’, y’, z’) to the field point (x, y, vag

when the gradient is calculated at the field point, with the same unit vector;

| wan I |faa:

[x nar=- [4x da
7 8
where S is the surface that bounds the volume r.

PROBLEMS

NOTE: Double daggers ({) indicate problems for which hints are given at the end of the
Problems section.

1-1. Show that the two vectors A = 9i + j — 6k and B = 4i — 6j + 5k are per-

pendicular to each other.
1-2. Show that the angle between the two vectors 4 = 2i + 3j + k and B =
i — 6j + k is 130.5°.
1-3. The vectors 4, B, C are coplanar. Show graphically that
A(B+ C)=A4-B+A4-C
1-4. If Aand B are adjacent sides ofa parallelogram, C = 4 + BandD = A—B
are the diagonals, and 0 is the angle between 4 and B, show that (C? + D*) =
2(42 + B?) and that (C? — D*) = 4AB cos 0.

1-5. Let a and b be two unit vectors lying in the xy-plane. Let a be the angle a
makes with the x-axis, and let 8 be the angle b makes with the x-axis, so that
a = cosai+sinaj and b = cos8i+ sinGj.
Show that the trigonometric relations for the sine and cosine of the sum
and difference of two angles follow from the interpretation of a-b anda &X b.

1-6. Show that the magnitude of (4 X B)-C is the volume of a parallelepiped

whose edges are 4, B, C, and show that (4 X B)-C = A-(B X C).

1-7. Show that the distributive rule applies to the vector product (Eq. 1-17).t
1-8. Show that a X (b X ce) = b(a-c) — c(a-b).

1-9. If r-(dr/dt) = 0, show that r = constant.

 PROBLEMS 34)

1-10. A gun fires a bullet at a velocity of 500 meters/second and at an angle of 30°
with the horizontal. Find the position vector r, the velocity vector v, and the
acceleration vector a of the bullet, ¢ seconds after the gun is fired. Draw a
sketch of the trajectory and show the three vectors for some time f.

1-11. Let r be the radius vector from the origin of coordinates to any point, and
let 4 be a constant vector. Show that V(4-r) = 4.

ee k=12. Phe vector r is directed from P’(x’, y’, z’) to P(x, y, z).
= (a) If the point P is fixed and the point P’ is allowed to move, show that the
gradient of (1/r) under these conditions is
A
1 ry

vs©) is
! — = —

where r; is the unit vector along r. Show that this is the maximum rate of
change of 1/r.

rQ--8
(b) Show similarly that, if P’ is fixed and P is allowed to move,

r r

1-13. (a) Show that V-r = 3.

(b) What is the flux of r through a spherical surface of radius r?

1-14. Show that

V-(f4) =f{VAT AVS,
where f is a scalar function and 4 is a vector function.

1-15. The vector A = 3xi + yj + 2zk, andf = x+y? + 2’.

f4 and
(a) Show that V-(fA4) at the point (2, 2, 2) is 120, by first calculating
then calculating its divergence.
(b) Make the same calculation by first finding Vfand V- 4, and then using the
identity of Problem 1-14 above.
(c) If x, y, z are measured in meters, what are the units of V-(fA)?

1-16. It is suggested that V-(A X r/r’) is identically equal to zero as long as the
vector A is a constant.
Is this correct?

1-17. The components of a vector A are

AS
af, of
yao aes ay’ A, Se
=—Z Ae
Dies
x ie A, ee
x ay y ae

wheref is a function of x, y, z. Show that

Are Vis Aer —)0; and AV j—10)

1-18. Show that V X (f4) = (Vf) X 4+ f(V X A), wheref is a scalar function
and 4 is a vector function.

1-19. Show that V-(A X D) = D-(V X A) — A(V X D), where A and D are any
two vectors.
 VECTORS
38

1-20. One of the four Maxwell equations states that V x E = —0B/dt, where E
and B are respectively the electric field intensity in volts/meter and the mag-
netic induction in teslas at a point.
Show that

|f Beal
is 20.0 microvolts over a square 10.0 centimeters on the side when B is
2.00 10-*¢ tesla and normal to the square.

1-21. Show that

V-(V2A) = V(V-A)
in Cartesian coordinates.

1-22. Show, by differentiating the appropriate expressions for r, that the velocity
r in cylindrical coordinates is
ppi + pegi + 2m,
while in spherical coordinates it is
rr; + 160, + rsin Ogi.
1-23. The vector 4 is everywhere perpendicular to, and directed away from, a
given straight line; that is, in cylindrical coordinates, A. = A, = 0. Calculate
the net outgoing flux for a volume element, and show that

eet
p dp
. A fluid rotates with an angular velocity w about the z-axis. The direction of
rotation is related to that of the z-axis by the right-hand screw rule.
(a) Find the velocity v of a point in the fluid, and show that
VX v= Zook:

(b) If now w is a function of the radius r, show that V X v = 0 if w =

Constant/r’.

1-25. The gravitational forces exerted by the Sun on the planets are always directed
toward the Sun and depend only on the distance r. This type of field is called
a central force field.
Find the potential energy at a distance r from a center of attraction when
the force varies as 1/r?. Set the potential energy equal to zero at infinity.

1-26. The azimuthal force exerted on an electron in a certain betatron is propor-

tional to r°-4.
Find the curl of this force in cylindrical coordinates, and show that the
force is nonconservative.

1-27. A vector field is defined by 4 = f(r)r.

(a) Show that f(r) = Constant/r3 if V-A = 0.
(b) Show that VY X 4 is always equal to zero.
PROBLEMS
39

1-28. Show that

V X(V X A) = V(V-A) — V?A
in Cartesian coordinates, but not in the general case of curvilinear coordi-
nates, if V*A is taken to be the vector whose components are the Laplacians
of A, Ao, A3.

1-29. Show that

[ vrar
= [faa,
wheref is a function of (x, y, z) and S is the surface bounding the volume 7.{

. Show that

[wxda=- faxda,
where 4 is an arbitrary vector and S is the surface bounding the volume r.t{

Hints
1-7. Write out the components in Cartesian coordinates.

1-28. You can solve this problem without having to write out the equations in full.

1-29. Multiply both sides by a constant vector (scalar product) and use the identity of
Problem 1-14.

1-30. Multiply both sides by a constant vector (scalar product) and use the identities
of Problems 1-6 and 1-19.
CHAPTER »,

ELECTROSTATIC FIELDS I
Electrostatic Fields in a Vacuum

We begin our study of electromagnetic fields by investigating those fields that

originate from stationary electric charges. This will occupy us during the next
three chapters.
We shall start with Coulomb’s law for electrostatic forces because it is
fundamental. We shall deduce from it the basic concepts and laws of electro-
statics: the electric field intensity and the electric potential, Gauss’s law and
Poisson’s equation, and, finally, the electrostatic energy density, which is
related to the electrostatic forces.

2.1 COULOMB’S LAW

It has been found experimentally that the force between two stationary elec-
tric point charges Q, and Q, (a) acts along the line joining the two charges,
(b) is proportional to the product Q,Q,, and (c) is inversely proportional to
the square of the distance r separating the charges.
If the charges are extended, the situation is more complicated in that the
“distance between the charges” has no definite meaning. Moreover, the pres-
ence of Q, can modify the charge distribution within Q,, and vice versa,
leading to a complicated variation of force with distance.
We thus have Coulomb’s law for stationary point charges:

ae (2-1)
 41

Figure 2-1. Charges Q, and Q,

separated by a distance r. The
force exerted on Q, by Q, is Fiz
and is in the direction r; along
the line joining the two charges.

_ Where Fy is the force exerted by Q, on Q», K is a constant of proportionality,

and r; is a unit vector pointing in the direction from Q, to Q,, as in Figure 2-1.
The force is repulsive if Q, and Q, are of the same sign; it is attractive if they
are of different signs.
The magnitude of the constant of proportionality K depends on the units
that are used for the measurement of force, charge, and distance. In the
rationalized MKSA (Meter, Kilogram, Second, Ampere) system, which we
shall use throughout this book,

where the force F,, is measured in newtons; the charges Q, and Q), in
coulombs (the magnitude of which will be defined in terms of magnetic
interactions in Chapters 6 and 7); and the distance r, in meters. The quantity
47 appears in explicit form so as to simplify other equations that are used
much more extensively than Coulomb’s law. The constant ¢ is called the
permittivity of free space:

€o = 8.8542 & 10—!? farad/meter.

Coulomb’s law applies to a pair of point charges situated in a vacuum.

It also applies in dielectrics and conductors if F.» is taken to be the direct
force between Q, and Q,, irrespective of the forces arising from other
charges within the medium.
Substituting the value of « into Coulomb’s law,

where the factor of 9 is accurate to about one part in 1000; it should really
be 8.988.

Example The Coulomb forces in nature are enormous when compared with
the gravitational forces, which are given by
42 ELECTROSTATIC FIELDS I

MIN
F = 6.67 X 10" ~~ 1. (2-4)

One good example is the following. The gravitational force on a

proton at the surface of the Sun (Mass = 2.0 x 10* kilograms,
Radius = 7.0 X 108 meters) is equal to the electric force between
a proton and one microgram of electrons, separated by a distance
equal to the Sun’s radius.

It is remarkable indeed that we should not be conscious of these enor-

mous forces in everyday life. The positive and negative charges carried
respectively by the proton and the electron are very nearly, if not exactly, the
same. Recent experiments have shown them to be equal within about one
part in 102%.* Ordinary matter is thus neutral, and the enormous Coulomb
forces prevent the accumulation of any appreciable quantity of charge of
either sign. You will be able to show in Problem 2-35 that a sphere of protons
about one foot in diameter and weighing about 20 kilograms would have
enough energy to pulverize the Moon!

2 oes vUEC TRIG BIBLEDs INTENSELY

We think of the interaction between the point charges Q, and Q, in Cou-

lomb’s law as being an interaction between Q, and the field of Qs, or vice
versa. We define the e/ectric field intensity E to be the force per unit charge
exerted on a test charge in the field.-Thus the electric field intensity due to
the point charge Q, is

The electric field intensit iS measured in volts/meter

the held or
tne (

When the electric field is produced by a charge distribution that is dis-

turbed by the introduction of a finite test charge Q’, we can define E to be
the limiting force per unit charge as the magnitude of the test charge Q/
tends to zero:

2 at
E = lim —- (2-6)
Q’>0

* John G. King: private communication.

2.3 THE ELECTRIC POTENTIAL 43

ee) PYL opt (2-7)

Arey J F ;

This is the principle of superposition. In this integral, p is the electric charge

density at the source point (x’, y’, z’), r; is a unit vector pointing from the
source point to the field point (x, y, z) where E is calculated, r is the distance
between these.two points, and dr’ is the element of volume dx’ dy’ dz’. If
there exist surface distributions of charge, then we must add a similar integral
with p replaced by o and 7’ by A’.

Poetoee EPUB CTRIG. POTENTIAL

Consider a test point charge Q’ that can be moved about in an electric field.
The work W required to move it at a constant speed from a point P; to a
point P, along a given path is

The negative sign is required to obtain the work done against the field. Here
again, we assume that Q’ is so small that the charge distributions are not
appreciably disturbed by its presence. ‘
If the path is closed, the total work done is

Wa -p EQ’ -dl. (2-9)

Let us evaluate this integral. To simplify matters, we first consider the electric
field produced by a single point charge Q. Then

fproral rey Mash (2-10)

Are ie

Now the term under the integral on the right is simply dr/r? or —d(l/r).
r has the
The sum of the increments of (1/r) over a closed path is zero, since
line integral
same value at the beginning and at the end of the path. Then the
in moving a point charge Q’ around any
is zero, and the net work done
is fixed, is zero.
closed path in the field of a point charge Q, which
QO but by
If the electric field is produced not by a single point charge
onding to each in-
some fixed charge distribution, the line integrals corresp
44 ELECTROSTATIC FIELDS I

dividual charge of the distribution are all zero. Thus, for any distribution of
fixed charges,

el =, (2-11)

ne Acer follows raat ee the fact that the Coulomb force is a

central force.
Then, from Stokes’s theorem (Section 1.6), at ali points in space,
VXE=0, (2-12)
and we can write that

_E=-vv,! (2-13)
where V is a scalar point function, since V X VV = 0

en car int towat decrease | ytential, score to the usual

convention. It is aoa to note hae Vjis not uniquely defined; we can
add to it any quantity that is independent of the coordinates without affect-
ing EF in any way.
We can now show that the work done in moving a test charge at a
constant speed from a point P; to a point P, is independent of the path.
Let a and b be any two different paths leading from P; to P). Then these
two paths together form a closed curve and the work done in going from
P, to P; along a, and then from P» back to P; along b is zero. Then the work
done a aus from P, to P; is the same along a as it is along b.
ember th ith If there
were moving ATR present, V X E would not necessarily be zero, and VV
would then describe only part of the electric field intensity FE. We shall
investigate these more complicated phenomena later on.
According to Eq. 2-13,

Brdl =—VVedl = ar, (2-14)

Then

as in Figure 2-2.
Note that the electric field intensity E determines only differences between
the potentials at two different points. When we wish to speak of the electric
 45

Figure 2-2. The potential differ-

ence V; — V2 between two points
is given by the line integral of
E-dl from 1 to 2, where E is the
electric field intensity and dl is
an element of the path along
which the integral is to be calcu-
lated. The light lines represent
lines of force.

potential at a given point, we must therefore arbitrarily define the potential

in a given region of space to be zero. It is usually convenient to choose the
potential at infinity to be zero. Then the potential V at the point 2 is

The work W required to bring a charge Q’ from a point at which the

potential is defined to be zero to the point considered is VQ’. Thus V is
W/Q’ and can be defined to be the work per unit charge. The potential V
is expressed in joules/coulomb, or in volts.
When the field is produced by a single point charge Q, —

(2-17)
It will be observed that the sign of the potential V is the sam e as that of Q.
4 7 T 1]
electri
| tential V aS well

The set of points in space that are at a given potential defines an equi-
potential surface. For example, the equipotential surfaces about a point
charge are concentric spheres. We can see from Eq. 2-13 that the electric
field intensity FE is everywhere normal to the equipotential surfaces (Sec-
tion 1.4).
46 ELECTROSTATIC FIELDS I

If we join end-to-end infinitesimal vectors representing FE, we get a curve

in space—called a /ine of force—that is everywhere normal to the equipo-
tential surfaces. The vector E is everywhere tangent to a line of force.

24, THB ELECTRIG* BIEED OUTSIDE

AND INSIDE MACROSCOPIC BODIES

We have so far been thinking in terms of point charges. Now electric charges
are usually distributed over macroscopic bodies, which are composed of
positively charged nuclei and negative electrons. This brings up two questions.

(a) First of all, can we calculate the field owtside an electrically charged
body by assuming that the charge distribution inside the body is continuous?
If our assumption is valid, we can calculate the field by integrating over the
charge distribution. If our assumption is not valid, then we must find some
other form of calculation.
It is, in fact, usually appropriate to treat the discrete charges carried by
nuclei and electrons within macroscopic bodies as though they were con-
tinuously distributed within these bodies. Even the largest nuclei have
diameters that are only of the order of 10-4 meter. Nuclei and electrons
are therefore so small and so closely packed compared to the dimensions of
ordinary macroscopic objects that we may define an average electric charge
density p, measured in coulombs/meter®, as AQ/Ar’, where AQ is the net
charge within Ar’.
The volume Ar’ may be assumed large enough to render the fluctuations
in AQ with time, or from one Ar’ to a neighboring one, negligible. At the
same time the volume Ar’ may be assumed small enough to permit the use of
integral calculus.
Thus, instead of summing the potentials of a large number of individual
charges, we may integrate over a continuous distribution p(x’, y’, z’). An
element of charge p dr’ contributes at a point P(x, y, z) outside dr’ a potential

Wiehrmaie (2-18)
tos lope ee

where r is the distance from dr’ to P, and

idee p dr’
igs Are Jo Tr aes,
The integral is evaluated over all regions where the net charge density p is
not zero.
2.5 GAUSS’S LAW 47

The above integral is not always applicable, however, because in some

cases it diverges. In such cases you can calculate E and then deduce a po-
tential V by integration.
(b) Now what about the electric field inside a charged body? It is
obvious that the electric field intensity in the immediate neighborhood of a
nucleus or of an electron must be enormous. It is also obvious that this
electric field must change rapidly with time, since the charges are never
perfectly stationary. It is not useful for our purposes to look at the electric
field as closely as that. We shall be satisfied if we can calculate average values
of E and V inside a charged body by assuming a continuous distribution of
charge. This should be meaningful, since nuclei and electrons are so exceed-
ingly small.
But is it then really possible to define the electric field at a point P inside
a continuously distributed charge? It appears at first sight that the dV con-
tributed by the charge element p dr’ at P is infinite, since r is zero. In fact,
it is not infinite.
Consider a volume element dr’ that is a spherical shell of thickness dr
and radius r centered on P. The charge in this shell contributes at P a dV of
(p/4:e)(4r7r? dr/r). Another shell of smaller radius contributes a smaller dV
because, as 7 decreases, the 47r? dr in the numerator decreases more rapidly
than does the r in the denominator. The electric potential V therefore con-
verges, and the integral is finite. A similar argument shows that the electric
field intensity E also converges.

The electric fields of real charge distributions can therefore be calculated

with the usual techniques of the integral calculus, both inside and outside
the distributions.

2.5 GAUSS’S LAW

Gauss’s law relates the flux of E through a closed surface to the total charge
enclosed within the surface. Consider Figure 2-3, in which a point charge Q
is located inside a closed surface S at a point P’. We can calculate the flux
of the electric field intensity E through the closed surface as follows. he
flux of E through the element of area da is

oh) Q nda
=.= (2-20)
UN 2
 Figure 2-3. A point charge Q
located at P’ inside a closed
surface S. The total flux of the
electric field intensity E through
the surface S is Q/€.

where r-da is the projection of da on a plane normal to r;. Then

Ee aay
Are
(2-21)
where dQ is the element of solid angle subtended by da at the point P’.
To find the total flux of E, we integrate over the whole surface S:

If the point charge Q were outside the surface at some point P”, the
solid angle subtended by the surface S at P’’ would be zero.
The situation remains unchanged if the surface is convoluted in such
a way that a line drawn outward from P’ cuts the surface at more than one
point. The total solid angle subtended by the closed surface is still 47 at an
inside point P’, and is still zero at an outside point P”’.
If more than one point charge resides within S, the fluxes add alge-
braically, and the total flux of E leaving the volume is equal to the total
enclosed charge divided by e. This is Gauss’s law.
Gauss’s law provides us with a powerful method for calculating the
electric field intensity FE of simple charge distributions.
If the charge enclosed by the surface S’ is distributed over a finite
volume, the total enclosed charge is

Q= |pdr’, (2-23)
where p is the charge density and 7’ is the volume enclosed by the surface S’.
Then
2.5' GAUSS’S LAW 49

Applying the divergence theorem to the left-hand side,

[ovnar a2 foan (2-25)

Since this equation is valid for any closed surface in the field, the integrands
must be equal; at every point in the field,

Pe aes (2-26)

This i It concerns the derivatives of E

with respect to the coordinates, and not E itself.

Example THE AVERAGE POTENTIAL

OVER A SPHERICAL SURFACE
As an illustration of SDSS: Sslaw, we shall now show thai

Let us first think of a spherical shell of radius a carrying a uni-

form surface charge density o and a total charge Q, as in Figure
2-4a. Then, from Gauss’s law, the electric field intensity at some
point P situated outside at a distance R from the center of the shell
is O/47e)R? and the potential is

- 25 \Lead } (2-27)

J
Also, according to Coulomb’s law,

y= [ 4reor
2%, (2-28)

where the integral is evaluated over the surface of the shell. Equating
the two values of V and remembering that o is O/47a’, we obtain a
purely geometrical relation concerning a sphere of radius a and a
point P at a distance R from its center, as can be seen by canceling
the Q’s:
Os (ag okele (2-29)
4iregR * 4ra’ An eor

We now shift our attention to Figure 2-4b. The left-hand side

of this equation is the potential at the center of the sphere due to
the charge Q, and the right-hand side is the average potential on the
spherical surface.
We have tberetore eS acliae that
point char
50

Figure 2-4. (a) Spherical surface carrying a uniform charge distribution

a. (b) Point charge Q outside an imaginary spherical surface of radius a.
(c) Imaginary spherical surface in a charge-free region. It is shown that
the average potential over the surface is equal to the potential at the
center. (d) Spherical surface carrying a uniform charge distribution co.
The point P is situated inside. (e) Point charge Q inside an imaginary
spherical surface of radius a. (f) Imaginary spherical surface enclosing a
charge Q. It is shown that the average potential over the surface is equal
to Q/4reca.

his result also applies to any charge

distribution situated outside the sphere, as in Figure 2-4c, because
of the principle of superposition.
Problem 2-16 illustrates one application of this property of the
electric potential to the calculation of electric fields.
Now imagine for a moment that there is a potential maximum
at some point O in a region where p = 0. Then the average potential
over some sphere centered on O must be lower than the potentia
at O, which is contrary to the above resul
potential maxim

We can proceed in a similar fashion to find the average poten-

tial over a spherical surface when the charges are situated inside.
We start again with a charge Q distributed uniformly over the sur-
face as in Figure 2-4d. At any point P inside, the electric field
intensity must be zero for the following reason. The symmetry of
the field requires that Ey and E, be zero. To find E, we apply Gauss’s
law to a concentric spherical surface having some radius smaller
than a and we find that E, is also zero. Then E is zero inside, and
 2.6 THE EQUATIONS OF POISSON AND OF LAPLACE 51

the potential V at the point P inside must be equal to that at

the
surface, namely Q/4z€9a. Using now Coulomb’s law as previousl
y,
Pk”) Ds o da drm Q da
Areca 4rer = 4ra? J Areor
(2-30) ©
But the last term is just the average potential over the spherical sur-
face of Figure 2-4e.

2.6 THE EQUATIONS OF POISSON

AND OF LAPLACE

If we replace E by —VV in Eq. 2-26,

This isPoisson’s equation. Ina region of the field where the charge density p
is Zero,

which isLaplace’s equation.

oo 039
The general problem of finding the electric potential V corresponding to
a given charge distribution amounts to finding a solution of either Laplace’s
or Poisson’s equation that will satisfy the given boundary conditions.

REVIEW. At this stage it is worthwhile reviewing briefly the material that

we have studied since the beginning of the chapter.
Coulomb’s law states that the force between two stationary electric
charges (a) is proportional to the magnitudes of each one of the two charges,
(b) is directed along the line joining the two charges, and (c) varies inversely
as the square of the distance between them. We interpreted this force as the
product of the magnitude of either one of the charges by the electric field
intensity of the other at the point occupied by the first one.
The principle of superposition follows from the observation that each
charge produces its own field independently of all the other charges and that
the resultant electric field intensity is therefore the vector sum of all the
individual fields.
It follows from (b) and from the principle of superposition that the
52 ELECTROSTATIC FIELDS I

electrostatic field is conservative, and that we can describe the electric field
intensity in terms of a potential V: E = —VYV.
Finally, (a), (b), (c), and the principle of superposition led us to Gauss’s
law V-E = p/e. When we expressed this in terms of the electric potential V,
we found Poisson’s equation V?V = —p/«, and Laplace’s equation V*°V = 0
for the special case where p = 0.

2.7 CONDUCTORS

A conductor can be defined as a material inside which electric charges can

flow freely. Since we are dealing with electrostatics, we assume a priori that
the charges have reached their eau aniy positions and are fixed in space.

Thae se a conductor is charged, the charges arrange themselves so

that the net electric field due to all the charges can be zero inside the con-

ict

sible to ree a charge pamEreiRon one cnc ar an isolated nine conducting

plate without having an identical charge density on the other side so as to
produce within the las two fields of equal intensity but of app direc-

Coulomb’s law applies within conductors, even though the net field is
zero. Then Gauss’s law V-E = p/e, which is a consequence of Coulomb’s
law, must also be valid within conductors. Now, under static conditions,
FE = O inside a conductor, and hence the charge density p must be zero.
As 2 corollary gigeiaeeiehic bit0S Cosiclesec ea
At the surface of a conductor, the electric field intensity E must be
normal, for if there were a tangential component of E, charges would flow
along the surface, which would be contrary to our hypothesis. Then, accord-
ing to Gauss’s law, just outside the surface, E = o/«, where o is the surface
charge density.
2.8 CALCULATION OF ELECTRIC FIELDS
53

It is paradoxical that we should be able to express the

electric field
intensity at the surface of a conductor in terms of the local surface
charge
density o alone, despite the fact that the field is of course due to
all the
charges, whether they are on the conductor or elsewhere.

Example A hollow conductor has a charge on its inner surface that is equal
in magnitude and opposite in sign to any charge that may be en-
closed within the hollow. This is readily demonstrated by considering
a Gaussian surface that lies within the conductor and that encloses
the hollow. Since E is everywhere zero on this surface, the total-
enclosed charge must be zero.

2.8 CALCULATION OF THE ELECTRIC FIELD

PRODUCED BY A SIMPLE
CHARGE DISTRIBUTION

A problem often encountered in electrostatics is that of finding the electric

field intensity E produced by a given charge distribution. The vector E can
be found in several different ways.

(a) We can use Eq. 2-7 and integrate over the whole charge distribution.
We must of course keep in mind that the electric field intensity is a vector
quantity.
(b) Another way consists in writing down the potential produced by an
element of charge, integrating over the whole charge to obtain V, and then
calculating the electric field intensity from E = —VV. The potential calcula-
tion is generally simpler because the potential is a scalar quantity that can be
integrated more easily than the vector E.
(c) In some cases we can integrate either Laplace’s or Poisson’s equa-
tion, depending on whether the charge density p is zero or not, if the geom-
etry is simple enough, and if we have sufficient information about the field
to determine the constants of integration.
(d) We can use Gauss’s law if the charge distribution possesses a sym-
metry that ensures constancy of the electric field intensity over certain
imaginary surfaces in the field. When this method is applicable, it is usually
the easiest one.
(e) We shall find still another method in Section 4.7.

These five types of calculation are often inadequate, and we must then
use other methods, described in Chapter 4 and in Appendix B. Even then,
54 ELECTROSTATIC FIELDS I

only quite simple charge distributions can be treated by analytical means.

The fields of more complex charge distributions are calculated on electronic
computers.

Example FIELD OF A UNIFORM SPHERICAL

CHARGE DISTRIBUTION
As an illustration, we shall calculate the electric field intensity E,
both inside and outside a uniform spherical charge distribution,
using the methods (a), (b), (c), and (d). Our calculation will be quite
detailed and will serve as a review of all the material we have
discussed so far.
The distribution has a radius R and a uniform electric charge
density p, as in Figure 2-5. Our problem is to find the electric field
intensity E as a function of the distance r from the center O of the
sphere to the point P. It should be obvious, by symmetry, that E is
independent of the two other spherical coordinates 6 and ¢.
We shall use the index o to indicate that we are dealing with the
field outside the charge distribution, and the index i inside.

Field FE, at an External Point

Calculation of E, Using Coulomb’s Law. Let us first calculate E,

directly from Eq. 2-7. We can find the contribution to E, due to the
charge pdr’ in the element of volume dr’ and then integrate the
resultant expression over the whole sphere. It is convenient to use
spherical coordinates, since the charge has spherical symmetry. Then
the volume element is dr’ r’ d6’ r’ sin 6’ dy’. The charge in this volume
produces a field at the point P which is directed away from the vol-
ume element if p is positive, and toward the volume element if p is
negative. Its magnitude is

1 - pr’? sin 0? ar’ do” dg

de, (2-33)
Arr eo Ww

where s is the distance from the volume element to the point P. The
axis along which @ = 0 can be taken to be the line OP. The element
of electric field intensity is written d’e,, since it is a third-order
differential.
It should be obvious, from the symmetry of the charge distribu-
tion, that E, must be radial. For example, while the charge element
shown in Figure 2-5 produces an electric field intensity d’e, that is
not along the radius OP, there is another symmetrically placed ele-
ment that produces a symmetrically oriented field of the same mag-
nitude, and the net result is a field along OP.
 SB)

Figure 2-5. An element of charge at the point (r’, @) inside a uniform,

spherical charge distribution produces an element of electrostatic field
intensity d*Kj at a point P outside the sphere. The projection of d3E/ on
the axis joining P and the center of the sphere is d°E,.

We therefore consider only the radial component of e, and

1 pr’? Toessin&, dr’ , dd’ , de , pe
OND, = a. (2-34)
4ireéo gs?

The integration over the azimuth angle ¢’ around OP is straight-

forward, and the angle ¢’ varies from 0 to 27. We can carry out the
other two integrations by using r’ and s as independent variables.
To do this, we eliminate a with the aid of the cosine law:
get
pr? — 7?
PS ape (2-35)
Similarly,
12 Oi ee 2

cos 6’ = 7 as (2-36)
Now we wish to eliminate sin 6’ d0’ from the expression for d°E,. We
can find sin 6’ d6’ in terms of 7’, r, s, by differentiating the above
equation. Here we must remember that r is a constant and that r’ is
taken as a constant when we integrate Eq. 2-34 with respect to 0’.
Thus we must differentiate Eg. 2-36, taking both r and r’ as con-
stants, and thus

one rr
(2-37)
56 ELECTROSTATIC FIELDS I

If we substitute Eqs. 2-35 and 2-37 into Eq. 2-34, and integrate,

)ds ar’,(2-38)
s=r+r’ , rP—r? r :
15;
reg ho tAbie r (1$55
BeArey ee ee (2-39)
3

ie 4ire, r”
where Q is the total charge (4/3)7R’p, and where nr; is the radial unit
vector directed outward.
The vector E, is directed outward along OP if Q is positive,
and inward along OP if Q is negative.
This result is the same as if the total charge Q were concentrated
at the center of the sphere.

Calculation of E, Using the Potential. To compute the electric field

intensity EH, from the potential V, we use the same element of charge
as above. Then, from the definition of V,,
1 pr? sin 6’ dr’ dé’ dy’
(2-40)
4rreo Ky
There is no cos a term now, since V, is a scalar. To carry out the
integration, we eliminate 0’ and integrate over ¢’, s, and r’ as we did
before. The result is

eek
V,= tren (2-41)
r

The electric potential V,, like E,, is the same as if the total charge Q
were concentrated at the center of the sphere.
To find E,, we now calculate VV,. By symmetry, E, must be
radial, hence
emery 6 EC
Le eit Gla ae cage 1
Airey r?
ri, (2-42)
as previously.

Calculation of E, Using Laplace’s Equation. By hypothesis, p = 0

outside the sphere, and
VV, = 0. (2-43)
Now, by symmetry, V, is independent of both @ and y. Therefore

(ee aay
E,= -4, (2-46)
where A is a constant of integration. We shall be able to determine
its value later on, after we have found E;.
2.8 CALCULATION OF ELECTRIC FIELDS Sy

Calculation of E, Using Gauss’s Law. The simplest way to compute

the electric field intensity in this case is to use Gauss’s law. Consider
an imaginary sphere of radius r > R concentric with the charged
sphere. We know that E, must be radial. Then, according to Gauss’s
law,

477r°E, = 2 (2-47)
€
and again i
caAA
Ares r1. (2-48)

If the charge were not distributed with spherical symmetry, E,

would be a function of @ and of ¢, and it would not be constant over
the imaginary sphere. Gauss’s law would then only give the average
value of the normal component of E, over the imaginary sphere.

Field E; at an Internal Point

Let us now calculate the electric field intensity at a point P within
the charge distribution, as shown in Figure 2-6. We can proceed as
in the case of the external point, first writing down the contribution
from an element of charge either to E; or to V;, and then integrating
over the whole charge distribution. However, since the integration
is difficult to carry out, we shall simplify the problem by dividing
it into two distinct parts.
We draw an imaginary sphere of radius r through the point P,
as in Figure 2-6, to divide the charge distribution into two parts.
We then calculate the electric field intensity due to the charge con-
tained within the sphere of radius r and that due to the charge
contained within the hollow outer sphere of inner and outer radii r
and R, respectively. Then, by the principle of superposition, the
resultant field intensity for the two charge systems must be the vector
sum of the two component field intensities.
The separation of the charge into two systems is especially ad-
vantageous in this case because, as we shall see, the field produced
by the hollow outer sphere at a point on its inner surface, or at any
point within the hollow, is zero. This can be demonstrated as fol-
lows, without integrating.
We draw a small cone of solid angle dQ, having its vertex at the
point P and extending in both directions as in Figure 2-6, and con-
sider the volumes that these small cones intercept within a spherical
shell of radius r’ and thickness dr’ concentric with the sphere. The
distances between these volumes and P are s; and 5».
On the left, the volume element is
. s? dQ
Ci dr’, (2-49)
COs a
whereas on the right it is
Pree ar
COS @
(2-50)
58

Figure 2-6. To find the electrostatic field intensity E; at a point P inside

a uniform spherical charge distribution, we divide the sphere into a shell
and a core by means of an imaginary sphere of radius r. Then any pair
of volume elements such as those shown in the shell produces equal and
opposite fields at P. The field at P is thus due solely to the charges in the
core. The exploded view shows one of the volume elements in detail.

The charge in the left-hand volume element contributes at P a field

of magnitude
PE. = —? sj dQ
dr’, (2-51)
“ — Gareos? cos a
which is directed toward the right if p is positive. Similarly, the
charge on the right contributes an identical field, opposite in direc-
tion, with the result that the two fields ce Since this result is
valid for any dQ and any dr’, the fi u
point on its inner surface, or at any point within the HoUGWesis zero.
A much simpler way of showing that the field due to the hollow
sphere is zero anywhere inside the hollow is to use Gauss’s law.
Imagine a concentric spherical surface within the hollow. According
to Gauss’s law, the average radial electric field intensity over this
surface must be zero, since there is no enclosed charge. Now the
symmetry of the problem requires that the electric field, if it exists,
be radial and be the same over all the surface of the sphere. Hence
the electric field intensity must be zero at every point on any sphe-
rical surface within the hollow.

Calculation of E; Using Coulomb’s Law. With the electric field of

the hollow sphere thus disposed of, we can calculate the contribution
to EF; that is due to the inner sphere of radius r, just as we did in the
case of the external point:
1 (4/3)mr'p . pr
E; = 1 r\ (2-52)
4treo rr 3€0
2.8 CALCULATION OF ELECTRIC FIELDS 59

The electric field intensity thus increases linearly with r inside the
spherical charge distribution.

Calculation of E; Using the Potential. We can arrive at the same

result by first calculating the potential V; as a function of r within
the charge distribution. To do this, we could proceed by direct inte-
gration. However, it will again be easier and more instructive to
divide the charge distribution into two pele as above.

rmed. We choose our icvicntity

volume to be a thin shell ofane ip‘and of thickness dr’. Thus the
part of V; that is due to the hollow sphere is
1 R p4arr? dr’ p
= R? — r’), 2-53
Aire / ip 2€ ( He) ( )
Next we compute the potential due to the interior sphere of
radius r. The calculation is the same as for an external point, and we
can use Eq. 2-41. This term is
1 p(4/3)rr? p
Pr, (2-54)
At €0 r 3€

Adding these two contributions, we obtain the potential V; at

a radius r inside the spherical charge distribution:

Yoian= (Og
2(4 eis a
_) :
(2-55)

The potential V; can also be written as

Gaia) o Ps
Ae 4renR® 2 AreoR ( )

where the second term is the potential at the surface of the sphere,
and where the first term is the increase above the surface value for
the interior points.
Then

Ey= VV = a = (2-57)

Calculation of E; Using Poisson’s Equation. We now have, inside

the charge distribution,
AL | (2-58)

10 («mo) = 2h) (2-59)

60 ELECTROSTATIC FIELDS I

O53, OViN 2 ee bs
x(r a) = en aa
Ee ese: (2-61)
i 3

or 3€

pe ae (2-62)
where B is a constant of integration.
It is intuitively obvious that E; cannot become infinite at the
center of a uniform spherical charge distribution; B must therefore
be zero, and

= = ri. (2-63)
We are now in a position to find the value of the constant of
integration A, which we found when calculating E, with the equation
of Laplace. Should not the two values we have found for the electric
field intensity E, one valid inside and one valid outside (Eqs. 2-39
and 2-52), be equal at the surface of the charge distribution atr = R?
How could they be different? According to Gauss’s law, they could
be different if we had a surface charge density as on the surface of a
charged conductor. But, by assumption, the charged sphere has a
uniform volume density p out to the radius R and there can be no

pR?[
2€

pR?|
3€ |

Figure 2-7. In a uniform spherical charge distribution, the electric tield

intensity E rises linearly from the center to the surface of the sphere and
falls off as the inverse square of the distance outside the sphere. On the
other hand, the electric potential falls from a maximum at the center in
parabolic fashion inside the sphere, and as the inverse first power of the
distance outside.
2.9 THE ELECTRIC DIPOLE 61

discontinuity in E = —0V/dr at the surface. Thus our two values

of E must be equal at the surface:

al is
ER = R = Ben (2-64)

so that
moh Gee ees
7 Ta eae eb)
and Eq. 2-46 does give the correct value for E,.

Calculation of E; Using Gauss’s Law. To calculate the electric field

intensity at the internal point from Gauss’s law, we draw an imag-
inary sphere of radius r through the point P. Symmetry requires that
the electric field intensity be radial; thus

derg, — O/B )Rrt

t ey (2-66)

E, = s ry (2-67)
as previously.
Figure 2-7 shows E and V for our spherical charge distribution
of radius R as a function of the radial distance r.

Zee bee eC PRIC] DIPOLE

The electric dipole shown in Figure 2-8 is one type of charge distribution that
is encountered frequently. We shall return to it later in this chapter and also
in Chapters 3 and 14.
The electric dipole consists of two charges, a positive and a negative
charge of the same magnitude, separated by a distance s, which is small
compared to the distance r to the point P at which we require the electric
potential V and the electric field intensity E.
At P,
y_ 2 (Z ") (2-68)
Ameo \71p =a
where

m= pet (5) + rs cos 8, (2-69)

r
= Vas
[1+(5) + £cos |bes (2-70)

=ay fae2 f+
\4p £0088)
r +38 (s+
\4r?2
2eose) — ---
i/
(2-71)
62

Figure 2-8. The two charges +Q

and — Q form a dipole. The
electric potential at P is the sum
of the potentials due to the
individual charges.

F : s
or, if we neglect terms of order higher than —>
r
ie = 3c0s? Gk
ie 1-— = cos 6 eT 5 (2-72)

Similarly,

F214
To
Xcose + 5-8 y)F—, (2-73)

V= Qs = COs 0 CaP) 8 (2-74)

Aregr?

It is interesting to note that potential

the due to a dipole falls’ off as
1/r?, whereas the potential from a single point charge varies only as 1/r.—
This comes from the fact that the charges of a dipole appear close together_
for an observer some distance away, and that their fields cancel more and
more as the distance r increases.
We define the dipole moment p = Qs as a vector whose magnitude is Qs |
and which is directed from the negative to the positive charge. Then
 63

Figure 2-9. A dipole p produces

an electric field intensity E with
components E, and E, at the
point (r, 6). The element dl of the
line of force is parallel to E at
the point.

The components of E in spherical coordinates can be computed from

the gradient of V:

k=me apaele 23
cert cos 6, (2-76)

een OV sors le Dye

ae r 00 4reg r8 Ie (2-77)

fe oy
SS sre 0g a
‘Id intensity of a dipole thus falls off as the cube of the distance.
The equation for the lines of force of a dipole can be found by con-
sidering Figure 2-9, which shows an element dl of a line of force. Since the
two vectors are parallel, the components ofE and of dl are proportional and

rdo_E& _ sing
dr E, 2cosé (2-79)

dr _ 2d(sin@)
6)
r sind en)
r= Asin? 6. (2-81)
This equation determines the family of lines of force for an electric
dipole. The constant A is a parameter that varies from one line of force to
another.
Figure 2-10 shows lines of force and equipotential lines for an electric
dipole. Equipotential surfaces are generated by rotating equipotential lines
around the vertical axis.
Figure 2-10. Lines of force (arrows) and equipotential lines for the dipole of Figure
2-8. The dipole is in the vertical position at the center of the figure with the
positive charge very close to and above the negative charge. In the central region
the lines come too close together to be shown.

2.10 THE LINEAR ELECTRIC QUADRUPOLE

The linear electric quadrupole is an arrangement of three charges as in

Figure 2-11. The separation s of the charges is again assumed to be small
compared to the distance r to the point P. At P,

V = i (2 — 22 + °), (2-82)

i 8)
The ratios r/ra and r/r, can be expanded as previously, except that s re-
 65

P(r,0,¢)

Figure 2-11. Charges + Q, —20Q,

and + Q arranged along a line
to form an axial quadrupole.

places s/2. Thus, neglecting terms of order higher than s?/7?,

r _s s? (3 cos@
? — 1)
l 7 COS 6+ 2 9) ,
Va a)
r s s? (3 cos? 9— 1)
1+ COS Dae 2 9) ,
lp )
and
2Qs 2 (3 cos? 2 6 =
1) (r3 >
V 5°),
(2-86)
Aregr® D)

The electric potential V due to a linear electric quadrupole thus varies

as 1/r’, whereas the electric field intensity E, calculated as for the dipole,
varies as 1/r*+. The fields of the three charges + Q, —2Q, + Q cancel almost
completely for r > s.

ZLIB ELECTRIC MULTIPOLES

It is possible to extend the dipole and quadrupole concepts to larger num-

bers of positive and negative charges located at small distances from each
other. Such charge arrangements are known as multipoles and are defined as
follows. A single point charge is called a monopole. A dipole is obtained by
66 ELECTROSTATIC FIELDS I

displacing a monopole through a small distance s, and replacing the original

monopole by another of the same magnitude but of opposite sign. Likewise,
a quadrupole is obtained by displacing a dipole by a small distance s; and
then replacing the original dipole by one of equal magnitude but of opposite
sign. For the linear quadrupole, s2 = s1.
The multipole concept can be continued indefinitely. For example, the
quadrupole may be displaced by a small distance s;, and the original quadru-
pole replaced by one in which the signs of all the charges have been changed.
This produces an octupole. In general, we can produce 2!-poles, where / is the
number of independent displacements s;, s2, --- required to specify the
arrangement.
We have seen that the dipole potential varies as 1/r? and that the quad-
rupole potential varies as 1/r*. For the general multipole, characterized by
the letter /, the potential varies as 1/r’*! and the field intensity as 1/r’*?.

2.12 THEVEVEC TRIGsHIELDF OUSIDE

AN ARBITRARY CHARGE DISTRIBUTION

Let us consider an arbitrary charge distribution of density p(x’, y’, z’) occupy-
ing a volume 7’ and extending to a maximum distance rfax from the origin of
coordinates O. We select O either within the volume or close to it. Such a
distribution is illustrated in Figure 2-12.

Figure 2-12. An arbitrary charge distribution of density p within

a
volume 7’ produces an electric potential V at a point P (x,
y, z) outside
7’. It is assumed here for the calculation of V that r > Page
2.12 THE ELECTRIC FIELD OUTSIDE AN ARBITRARY CHARGE DISTRIBUTION 67

We wish to find the electric potential V at some point P(xay)z)isuch

that r > ee This is

pdr’
V= ibAap: (2-87)

where 7’ is the distance between the point of observation P and the position
P'(x', y’, z') of the element of charge p dr’:

r= [xe — xP + = PF 2). (2-88)

The point of observation P(x, y, z) is taken to be fixed; thus 7” is a function
Olax 5Vi, 2
Since 7’’ is a function of x’, y’, z’, let us expand 1/r’’ as a Taylor series
near the origin:

al ih | EP a CRRA
tl
eatYayttazLF)
IPS Va ill oeER
15 PI | ax’ ae 3 ay’ 32 w\ (=) ale ’ (2-89)

where the subscripts 0 indicate that the derivatives are evaluated at the origin.
In squaring the second bracket, we consider the factors x’, y’, z’ to be con-
stants. The same applies to all the terms that follow.
Now
3 (5) - - m5 meee
” =
(2-90)
and
Oval x if
E (=) |,7a =) a r. (2-91)

where / = x/r is the cosine of the angle between the vector r and the x-axis.
The other first derivatives with respect to y’ and z’ are given by similar
expressions.
The second derivatives of the third term on the right-hand side of Eq.
2-89 are calculated similarly, the substitution of x’ = 0, y’ = 0, 2’ = 0,
r’’ = r being made only at the end.
The result of the calculation is that at the point P

1 p dr’ Le, ; np dr
y= |)
1 r 4rre9
+ [Ao Bey ee a

+- / 4 [3mny’z’ + 3nlz’x’ + 3lmx'y’

pdr’
4+ 4 GP — 1x? + § Gm? — 1y* + 4 Gn? — 12?) T+ +++, (2-92)
68 ELECTROSTATIC FIELDS I

where /, m,n are the direction cosines of the line joining the origin to the point
of observation P.
Let us now examine the various terms in succession. The first term is
merely the electric potential that we would have at P if the whole charge were
concentrated at the origin. It is called the monopole term and is zero only if
the total net charge is zero. If the charges are all of the same sign, then it is
the most important term of the series, since it decreases only as 1/r.
The second term varies as 1/r?, as does the electric potential of a dipole.
Let us find the value of this term when the charge distribution is simply a
dipole, as in Figure 2-8. The integral must then be evaluated over the two
charges Q and —Q situated at z’ = s/2 and z’ = —s/2 respectively. The
result is the value of V for the dipole, as in Eq. 2-74. The second term V2 of
Eq. 2-92 can thus be taken to be a dipole term. In fact,

mee Lil i
ee 4rreor” Ces
if we set

p= / (xi + yj + 27k) pdr’ = / r’p dr’, (2-94)

r, being the unit vector along r in the direction of P:

rn = i+ mj +nk. (2-95)
The quantity p is the dipole moment of the charge distribution. It is a
vector whose components are

Pe = i x'p dr’, Py = [ y'p dr’, Pz = iez’p dr’. (2-96)

Therefore V2, the second term of Eq. 2-92, is the same as if we had at the
origin a small dipole with dipole moment p, as in Eq. 2-94.
The dipole moment of an extended charge distribution can also be
defined as
p = Or, (2-97)
Px = Qz', Py = QY’, pe =07; (2-98)
where Q is the total net charge in the distribution, and where 7’ (the vector
distance from the origin to the center of charge) is defined in a manner
analogous to the center of mass in mechanics:

[oro dr’ is: dr’

7’

= face iar
= T wi

(2-99)

If Q is zero, then F’ > oo from the above equation, and p is indeterminate

2.12 THE ELECTRIC FIELD OUTSIDE AN ARBITRARY CHARGE DISTRIBUTION 69

in Eq. 2-97. Equation 2-94, however, always determines p unambiguously.

When Q = 0, the dipole moment is independent of the choice of origin
(see Problem 2-23).
If the net charge is not zero, the dipole moment of the distribution can
always be made zero by choosing the origin at the center of charge, for then
rFo= 0:
Let us now consider the third term V3 of Eq. 2-92. It involves a 1/7?
factor as the Vof the linear quadrupole of Section 2-10. If we calculate V; for
the linear quadrupole with charges Q, — 20, O atz=s,z=0,z=-—s
respectively, we find that it gives the electric potential V of Eq. 2-86.
Then the third term V3; of Eq. 2-92 is the same as if we had a small
quadrupole at the origin. Let us write out its quadrupole moment. This third
term can be rewritten as follows:

V3 = athe |smn y'z'p dr’ + 3nl | z'x’pdr'

4regr® 2 7

PAB in /_x'y'p dr! + 5Gr i) /x" dr!

+ 5Gre =31) / yp dr’ +5 Gre ae) iDy ar’ | (2-100)

Note that these integrals, just like the integrals of Eq. 2-94, depend solely on
the distribution of charge within the volume 7’, and not on the coordinates
(x, y, Z) of the point of observation P.
These six integrals specify a quantity called the quadrupole moment of the
charge distribution:

Paz = /x" dr! = Ox”, (2-101)

Pw = /yp dr’ = Oy”, (2-102)

Pes = /229 dr! = Q2”, (2-103)

Pye = Pa = /Y'2'p dr! = Qy’Z’, (2-104)

Dn = pe = fez'x'p dr! = Qz'x’, (2-105)

Poy = Pyz = /xy'p dr’ = Oxy’, (2-106)

where the bars indicate average values. Note that the quadrupole moment has
six components. Thus
70 ELECTROSTATIC FIELDS I

V3 = =
Aas [3mn Pyz + 3nl pez + 31m Pry
0)

+4.GP = 1) per +4 3m! =) Pw +5 Gr* — 1) pa] @-107)

If the charge distribution displays cylindrical symmetry about the z-axis,
the elements of charge at the points (x’, y’, z’), (—x’,y’, 2’), ’, —y’, 2’),
(—x’, —y’, 2’) are all equal; their contributions to the integrals for Pz,
Pye» Dex cancel; and Pry = Pye = Pex = 0. Also, by symmetry, Pzze = Py-
In such cases it is convenient to define a single quantity g, often called
the quadrupole moment of the charge distribution:
gq = 2 (Dez — Pra); (2-108)

= [a2 -rpar, (2-109)

= Q(z?
— FP). (2-110)
For a charge distribution that has cylindrical symmetry about the z-axis,
the quadrupole potential at the point (7, @) is
GeiGcos'6.—11)
V3 =
4reo 4r8
(2-111)
Note that, in general, the various multipole terms depend on the position
of the origin, except for the monopole term that is proportional to the net
charge Q of the system. The dipole term is independent of the position of the
origin in the special case where Q = 0.
In summary, the potential due to an arbitrary charge distribution at a
point outside the distribution such that r > rmax is the same as: (a) the
potential of a point charge equal to the net charge of the distribution, plus
(b) that of a point dipole with a dipole moment equal to the dipole moment of
the distribution, plus (c) that of a point quadrupole with a quadrupole
moment equal to that of the distribution, and so on, all located at the origin.
Similarly, the electric field intensity KE= —VV outside an arbitrary
charge distribution is the sum of the electric field intensities of the above
point charge, dipole, quadrupole, ete.

2.13 THE AVERAGE ELECTRIC FIELD: INTENSITY

INSIDE A SPHERE CONTAINING
AN ARBITRARY CHARGE DISTRIBUTION

In discussing dielectrics in the next chapter, we shall require the average

electric field intensity within a spherical surface enclosing an arbitrary charge
 71

Figure 2-13. A point charge OQ

located at a distance r’ from the
center of a sphere of radius R.
The average field intensity which
Q produces within the sphere is
proportional to the dipole
moment of Q, is inversely pro-
portional to the volume of the
sphere, and is in the direction
opposite to the dipole moment of
Q. Figure 2-6 shows the elements
of volume in detail.

distribution. Let us calculate the average electric field intensity at the interior
point directly. To do this we shall first calculate the average E for a single
point charge.
Figure 2-13 shows a point charge Q located at a distance r’ from the
center of a sphere of radius R. The z-axis is taken to be along the line joining
the center of the sphere and the charge Q. By symmetry, the average field
over the volume of the sphere must be along the z-axis.
We wish then to calculate

jgpeeSok = 1 | Bede (2-112)

where 7 is the volume of the sphere. This integral can be separated into two
parts, one over the spherical shell between the radii 7’ and R, and one over
the sphere of radius 7’.
We can show that the integral over the outer volume is zero by
using an argument somewhat analogous to that used on pages 57 to 61 to
find the electric field inside a uniform spherical charge distribution. In Figure
2-13 the solid-angle element dQ intercepts the volume elements dr; and dry
in the shell. Since the value of E, decreases as the square of the distance from
Q, whereas dr increases as the square of the distance (see Eq. 2-49), their
product remains constant. However, at dr, E, is positive, whereas at dry E; is
negative; thus the two contributions to the integral cancel. The same is true
‘app. ELECTROSTATIC FIELDS I

for the whole outer shell, the contributions to the integral canceling in pairs.
To calculate the integral of E, dr over the inner volume, which is equal
to the same integral over the complete volume 7, we shall use spherical polar
coordinates and set the origin at the position of the charge Q. Then, at the
point P,

= Apepie oe 6, (2-113)

and
g=2r 6=7 r!! = — 2r’cos6
/py / i qm €08 8 (r'” sin 6 dr’”’ do dy). (2-114)
e e g=0 @= a/2 r/=0 Areor

In this integral, 7’’ varies from zero to —2r’ cos 6, which is the distance from
Q to the surface of the inner sphere in the direction 6. The negative sign comes
from the fact that cos 6 is negative. Then 6 sweeps from 7/2 to z and ¢ from
0 to 27. Thus

ih
Ear = — oes cos’é sin 6 dé, (2-115)
T €0 a/2

Ore Real _2..Gr dl

© Se) ee Sere:
and
Pr eet oe OR 4
isa 369 47R2 —— ArregR? (2-117)
or, since Qr’ is the dipole moment of the charge Q (Equation 2-94),

Eaves!
AregR® (118)
This average electric field intensity has been calculated for a single charge
Q on the z-axis. For an arbitrary charge distribution the field is the super-
position of the fields due to the different charges, according to the theorem of
superposition. Then E is again as above, with Pp equal to the dipole moment
of the arbitrary charge distribution within the sphere of radius R.

2.14 POTENTIAL ENERGY OF

A CHARGE DISTRIBUTION

Let us first consider a set of small charges distributed in space as in Figure

2-14. We shall assume that the charge configuration is of finite extent and that
no other charges are present. Each individual charge is situated in the electric
field produced by the other charges, at a point where the potential V due to
 73

Figure 2-14. Set of point charges

Qi, Qo, ..., Qs separated by
distances 1D), o.0-0 G Mla

the other charges has some definite value. Each charge thus has associated
with it a definite potential energy, either positive or negative, and the system
as a whole has a potential energy that we shall calculate.
We assume that the charges remain in equilibrium under the action of
the electric forces and of restraining mechanic forces.
Let Q;, Qo, Q3, ---, On be the N charges; let 712 be the distance between
Q, and Qs», r13 the distance between Q, and Q;, and so on, as in Figure 2-14.
Now let Q; recede to infinity slowly, so that the electric and the mechanical
forces are always in equilibrium. In this way there is no acceleration and,
therefore, no kinetic energy involved. The other charges remain fixed.
The decrease W, in the electric potential energy of the charge is equal
to the work done by the electric forces. This is the product of the charge Q,
and the potential V; produced by the other charges at the original position of
Q,;. Then
w= 2 (24 24 2), (2-119)
Arey \ Nie ri3 Yin

Note that all the charges except Q; are represented in the series of terms
within the parentheses.
Now that Q,; has been removed, let Q») recede to some point infinitely
distant from Q;. The decrease W, in the potential energy of Q» is

Wy = 2 (24 4... + OY),

Arey \ ros V4 lon
(2-120)
This series has only N — 2 terms within the parentheses, the Q; and Q» terms
being absent. We continue the process for all the remaining charges, there
being progressively fewer terms in the series, until finally the Nth charge can
74 ELECTROSTATIC FIELDS I

be removed without any change in energy, since it is left in a zero field once all
the other charges have been removed.
The total potential energy of the original charge distribution is then
W= Wit Wit Wse+---+ Wy, (2-121)

a eae =)
lad ~ae ee)
+2 ( 0 ): (2-122)
Let us now rewrite the array of terms within the parentheses, adding in,
to the left of and below the diagonal line of zeros, terms that are equal to their
counterparts on the other side of the diagonal. Then every term of the series
appears twice, and

Pi
Tr€0
(o +2424
Vio ri3
24... 4 Ov)
V4 MN

pi (S+0 +2424
Té9 \ Foi 93 V4
+ 2)
ron

QO; /Qi , Qs QO; On

Arey @ st 39, | r 134 zy
©! Wicentie herTeyiei iene 6: o/b) 'e te: (eli@: 'e) «wus 0.0.6 © el 6 66) 6 © 4 ‘bie. oleae ae

_ Ow (24242 +O peee sp 0). (2-123)

N1 Ine I'n3 l'n4

Thus the first line is Q,V1, the second line is Q.V2, and so forth, so that

2W = QW; + QV. + Q3V3 + --- + OnVy, (2-124)

and the potential energy of the original charge distribution is

(2-125)

“We en not igken into account the self energy of the individual charges,
that is, the energy that would be liberated if each one were allowed to expand
to an infinite volume.
2.14 POTENTIAL ENERGY OF A CHARGE DISTRIBUTION 715

The reason the factor } appears in the above equation should be clear
from the reasoning we have used to arrive at it. It is that the potential at the
position of a given charge at the time it is removed to infinity is less, in general,
than the potential at the same point in the original charge distribution. On
the average, the potential at the time of removal is just one-half the potential
in the original charge distribution.
You will note that this energy W, which does not include the energy
required to assemble the individual charges themselves, can be either positive
or negative. It is the energy required to assemble already existing point
charges. For example, if we have two positive point charges Q; and Q», then
the positive charge Q, is situated at a position where the potential Vi, due to
the other charge Q», is positive, so that Q,V, is positive. The product QoV,
is also positive and the potential energy of the charge distribution
QO1Q>/4eor? is positive. However, if we have a positive charge Q, and a nega-
tive charge Q», then V; is negative and Q,V; is negative. The other product
Q2V. is similarly negative, and W is negative. Note also that, for a single
point charge, the above W is zero.
For a continuous electric charge distribution of density p(x’, y’, 2’), we
simply replace Q; by p dr’ and the summation by an integration:

W= tf Vp dr’. (2-126)
This operation is misleading, however, because we have now included the
energies required to assemble the individual macroscopic charges. In fact, as
we shall see in the next section, this integral is always positive. It is evaluated
over any arbitrary volume 7’ that contains all the charges in the system.
If we have conducting bodies carrying surface charge densities o, then
the stored energy is

W= at aV da’, (2-127)

where S’ is the sum of the areas of all the conductors.

Imagine now a conductor that is situated a long distance away from any
other body. Because of the repulsion between similar charges, energy must
be expended to charge the conductor. Accordingly, its potential rises as
charge is added, the magnitude of the charge in potential being proportional
to the amount of charge added and depending on the geometrical configu-
ration of the conductor. According to the above equation, the stored energy 1S

w=ov,! (2-128)
where Q is the net charge carried by the conductor and V is its potential.
16 ELECTROSTATIC FIELDS I

The ratio Q/V is called the capacitance of the isolated conductor:

(2-129)
ron Bel5
and thus its stored energy can also be written as

The capacitance is the electric charge that must be added per unit increase in
potential.
One might well ask why the capacitance C should be a constant for a
given conductor, as we have implied above. This can be shown as follows. The
isolated conductor carries a charge Q, and the electric potential in the region
surrounding it is V(x, y, z). At the surface of the conductor the charge density
iS é£ (Section 2-7), or — & times the rate of change of V in the direction normal
to the surface. Since the potential V satisfies Laplace’s equation V*V = 0
(Eq. 2-32), and since this equation is linear, any multiple of V is also a
solution. Let us therefore increase V everywhere by some factor a. Then,
since E = — VV, the surface charge density will also increase by a, and the
total charge Q on the conductor will increase likewise. The charge on an
isolated conductor is therefore proportional to the potential, and C is a
constant.
Imagine now that we have two uncharged isolated conductors. If a
charge Q is transferred from one to the other, a potential difference AV is
established between them and, by definition, the capacitance C between the
conductors is O/AV.
The unit of capacitance is the coulomb/volt, or farad.

2.15 ENERGY DENSITY IN AN ELECTRIC FIELD

The potential energy W can be related to the electric field of the charge
distribution as follows.
According to Poisson’s equation,

p= — aV?V (2-131)
at every point in the field. If we substitute this into Eq. 2-126, then

W=—- ah VWV dr’,

a
(2-132)
where 7’ is again any volume containing all the charges in the system.
2.15 ENERGY DENSITY IN AN ELECTRIC FIELD 77

We may simplify this equation through the vector identity

V-fA = f(V-A) + A-Vf, (2-133)

where f and A are scalar and vector functions respectively. Let f = V, and
A = V V. Then, rearranging,
VWV =V-VVV) — (VV), (2-134)
and

ine = /V+ VV) de! — /vy a} (2-135)

We can transform the first term on the right into an integral over the surface
S bounding the volume 7’ by using the divergence theorem. Thus

ae S| [ wey-ae asi (vy a} (2-136)

Now 7’ can be any volume that includes all the charges in the system. We
are thus free to choose the bounding surface S’ at a large distance from the
charge distribution. Then, in the first integral, V falls off at least as fast as Wie
since the monopole term in the potential of a charge distribution falls off as
1/r, whereas the dipole, quadrupole, and higher terms fall off as 1/72, 1/r? and
so forth, as we saw in Section 2.12. Then VV falls off at least as fast as 1/r?.
Since the surface area increases as r’, the whole integral decreases at least as
fast as 1/r and can be made arbitrarily small by choosing the surface S’
sufficiently far off.
We are then left with

Ue = 2 /EF? dr’, (2-137)

since E = |V V|. The volume 7’ need not be infinite now, but only large
enough to include all regions where E differs from zero. This integral takes
into account the self-energies of the charges; it is always positive, and it is
obviously not zero for a single charge.
The potential energy W of Eq. 2-126 is equal to the W we have deduced
from it above and is therefore also positive for all charge configurations.
Equation 2-137 shows that the energy associated with a charge distribu-
tion, that is, the energy required to assemble it, starting with a configuration
in which the charge is spread over an infinite volume, may be calculated by
associating with each point of the field an energy density

OME
Zz = 7 E 24joules/meter’.3 (2-138)
A
718 ELECTROSTATIC FIELDS I

In the next section we shall arrive at this same result by another method.
We can therefore calculate the energy stored in an electric charge distri-
bution, either in terms of the charge density p and the potential V, or in terms
of the electric field intensity E.

Example An isolated spherical conductor of radius R carries a surface charge

. density o. Then the potential energy is

Woe 2 iIs sda 2

OV (2-139)
where
ale 2-140
pid 4repR. ( )

Thus
Oz
= . 2-141
87regR ( )

We can also write that

= © f ar, (2-142)
Pan
aa e 4rr?
(25) dr = SreR ce (2-143)eS
which is the same result as above. We have integrated from R to
infinity, since there is zero E inside the sphere.

2.16 FORCES ON CONDUCTORS

dhe (=
APJEE EN
+ ore
1ana is thererore Suvdyje owe,
ee ly —

To calculate the magnitude of this force, let us consider a conductor with

a surface charge density o and a field E at the surface. From Gauss’s law, the
electric field intensity just outside the conductor is ¢/¢) and is perpendicular to
the conducting surface. Now the force on the element of charge adais not
Ec da, since the field that acts on a da is the field due only to the other charges _
in the system.
Let us first calculate the electric intensity produced by o da itself. We can
do this by using Gauss’s law. The total flux emerging from o da must be
o da/e, half of it inwards and half outwards, as in Figure 2-15. Then the
 79

Figure 2-15. The local electric charge density at the surface of a con-
ductor gives rise to oppositely directed electric field intensities ¢/2¢€ as
shown by the two arrows on the left; the other charges on the conductor
give rise to the field ¢/2¢) shown by the arrow on the right. The net
result is ¢/€) outside, and zero inside. The vector n is a unit vector
normal to the conductor surface, and it points outward.

electric field intensity due to the local surface chage density must be o/2¢
inwards and o/2¢) outward.
The element of charge o da itself therefore produces exactly half the total
field at a point outside, arbitrarily close to the surface. This is reasonable, for
the nearby charge is more effective than the rest.
Now if o da produces half the field, then all the other charges must
produce the other half, and the electric field intensity acting on o da must be
o/ 2€.
The force dF on the element of area da of the conductor is therefore
given by the product of its charge o da multiplied by the field of all the other
charges:
dF = 5-0da, (2-144)
and the force per unit area is
dF “ ig
da 5s eg 2h)

Now this force is just equal to the energy density that we found in the
preceding section. We should really have expected this for the following
reason. Let us imagine that the conducting bodies in the field are disconnected
from their power supplies. Then, if the electric forces are allowed to perform
mechanical work, they must do so at the expense of the electric energy stored
in the field. Now imagine that a small area a of a conductor is allowed to be
80 ELECTROSTATIC FIELDS I

pulled into the field by a small distance x. The mechanical work performed is
ax times the force per unit area. It is also equal to the energy lost by the field,
which is ax times the energy density. Then the force per unit area must be
equal to the energy density.
If we use the same type of argument for a compressed gas, we find that
its energy density is equal to its pressure.
‘We have therefore found another and a more simple demonstration of
the fact that we can calculate the energy stored in an electric field by inte-
grating eE?/2.
This second demonstration is in some ways preferable to that of Section
2.15. First, itis purely local; it does not require that the field be integrated
over an infinite volume. As you will remember, we had to perform such an
operation to eliminate the first integral in Eq. 2-136. Second, it relies solely on
Gauss’s law and on the fact that the force on a charge Q situated in an electric
field Eis QE. Both of these postulates are valid even if the field is not constant.
The proof of Section 2.15, on the other hand, is valid only for electrostatic
fields.
It is important to note that

Example , Electric forces are never very large. For example, even for E = 108
volts/meter, the electric force is only 4.4 newtons/meter?,
or 0.45
kilogram/meter?, or about 2 ounces/foot?. See Problem 3-23.

Example FORCES ON A PARALLEL-PLATE CAPACITOR

As a simple example, let us calculate the forces on the plates of a
parallel-plate capacitor of area S and plate separation s carrying a
charge density + o on one plate and — o on the other, as in Figure
2-16. We shall assume that the separation of the plates is small com-
pared to their linear extent, in order that the parallel-plate capacitor
approximation be valid.
The force per unit area on the plates is equal to the energy
density in the field eE?/2, and the force of attraction between the
plates is therefore 6,28 /2.
Let us now calculate this force from the change in energy corre-
sponding to a slight decrease in the spacing s. The energy associated
with the electric field between the plates is

W = 5€0 B'Ss. (2-146)

If the plates are insulated, so that the charges remain constant, and
 81

Figure 2-16. A charged parallel-plate

capacitor with its plates insulated.
The plates are held in equilibrium
by mechanical forces F, which act
on each plate in a direction tending
to increase the separation s, and by
electric forces F, tending to decrease
the separation.

if the plates are moved closer together by a small distance ds, the
electric field intensity E remains unchanged, according to Gauss’s
law. The volume between the plates decreases, however, and hence
the energy in the field also decreases. Since we must have conserva-
tion of energy, the energy lost by the field has gone into the work
F,,, ds against the mechanical forces F,,, holding the plates apart. Then

A ES dSa—S ass (2-147)

and
€ o? Q?
Nis NIN
2 = he, S= 2€,S ' 2-14
as)
Since the electric force F, is in equilibrium with the mechanical
force, it must be of equal magnitude and oriented so as to pull the
plates together.

2.17 SUMMARY

It is found empirically that the force exerted by a point charge Q, on a point

—, is

Fn = al Q.0» Y1, (2-2)

where r is the distance between the charges and r; is a unit vector pointing
from Q, to Q». This is Coulomb’s law. We consider the force Fy as being the
product of Q» by the electric field intensity due to Qa,

E, — 2 Let (2-5)

or vice versa.
82 ELECTROSTATIC FIELDS I

According to the principle of superposition, two or more electric fields

acting at a given point add vectorially.
The electrostatic field is conservative,

Vaxeh =, (2-12)
hence
Rien v4 (2-13)
where

ralAbay /Ii dae

WF
(2-19)
p dr’ is the element of charge contained within the element of volume dr’, and
r is the distance between this element and the point where V is calculated.
Gauss’s law follows from Coulomb’s law. It can be stated either in
integral or in differential form:

i E-da’ = u i p dr’, (2-24)

is! €0 Jr!
or
V-E= *. (2-26)
€0
Poisson’s equation

Vyasa (2-31)
€0
then follows from the differential statement of Gauss’s law and from the
definition of the potential.
When the charge density is zero we have Laplace’s equation

Vv = 0. (2-32)
The above is summarized schematically in Figure 2-17.
The electric field produced by simple charge distributions can be calcu-
lated in four different ways: (a) by evaluating dE for an element of charge
and integrating over the complete charge distribution, (b) by calculating V
in the same manner and then calculating its gradient, (c) by integrating
Poisson’s or Laplace’s equation, or (d) by using Gauss’s law. The fields due
to more complex charge distributions can be calculated by other methods
described in Chapter 4 and in Appendix B.
The potential produced by an arbitrary charge distribution at a distance
r > Tmax is the same as that produced by a point charge, a point dipole, a
point quadrupole, and so forth, all situated at the origin, where the monopole
 83

Coulomb’s Law

F is Fee Be Principle of Superposition|

central ie

Potential V

Poisson’s Equation

\
|Laplace’s Equation
Gauss’s Law

Figure 2-17.

carries the total charge of the distribution, the dipole has the dipole moment
of the charge distribution,

p= /op dr’, (2-94)

the quadrupole has the quadrupole moment of the charge distribution, and
so forth.
If Q is the net charge carried by an isolated conductor, and if V is its
potential, the ratio Q/V is called its capacitance.
The capacitance between two isolated conductors is the ratio Q/AV, where
Q is the charge transferred from one to the other, and AV is the resulting
potential difference.
The potential energy associated with a charge distribution can be written
either as

W = al Vp dr’ (2-126)
OF as

W = 3/ E? dr’. (2-137)

In the first integral, r’ must be chosen to include all the charge distribution, and
in the second it must include all regions of space where £ is nonvanishing. The
assignment of an energy density eE’/2 to every point in space therefore leads
to the correct potential energy for the whole charge distribution.
Finally, electric forces on conductors can be calculated by either of two
methods. We can either (a) utilize the fact that the force per unit area is
o?/2e = e&?/2, where E is the electric field intensity near the surface of the
conductor, or (b) equate the work performed by the electric forces to the work
 84 ELECTROSTATIC FIELDS I

performed against the restraining mechanical forces during an infinitesimal

virtual displacement.

PROBLEMS
2-1. Show that e, is expressed in farads/meter, that E is expressed in volts/meter,
and that the volt is a joule/coulomb.

(3-2Xa) Calculate the electric field intensity E that would be just sufficient to bal-
~~ ance the gravitational force on an electron.
(b) If this electric field were produced by a second electron located below
the first one, what would be the distance between the two electrons? (The
charge on an electron is —1.6 X 107-19 coulomb and its mass is 9.1 & 107*!
kilogram.)

3./It is known that the maximum electric field intensity which can be maintained
in air at NTP is about 3 X 10° volts/meter. This determines the maximum
surface charge density.
If a spherical body is charged to half this maximum surface charge
density and placed in a vertical electric field of 1.5 & 108 volts/meter, what
is the largest solid sphere that can be supported by the electric field ?Choose
any material you wish for the sphere.
Assume that it is nonconducting and that its charge is uniformly dis-
tributed over its surface.

2-4. A charge + Q is situated at (—a, 0, 0) and a charge —2Q is situated at (a, 0, 0).
Is there a point in space where E = 0?

(25. A thin infinite conducting plate carries a surface charge density c. Show that
a one-half of the electric field intensity E at a point situated z meters from the
surface of the plate is due to the charge located on the plate within a circle
of radius V3 z.

2-6. Show that the potential due to two parallel line charges of opposite polarity
is (A/27€) In (71/re), where r; is the distance from the point considered to the
negative line, and rz is the corresponding distance to the positive line.
If your answer turns out to be 0/0, try another method.

2-7. The electron beam in an oscilloscope is deflected vertically and horizontally

by two pairs of parallel deflecting plates that are maintained at appropriate
voltages. As the electrons pass between one pair they are accelerated by the
electric field, and their kinetic energy increases. Thus, under steady-state con-
ditions, we achieve an increase in the kinetic energy of the electrons without
any expenditure of power in the deflecting plates, as long as the beam does
not touch the plates.
PROBLEMS 85

Could this phenomenon not be used as the basis of a perpetual motion

machine?
2-8. In 1906, in the course of a historic experiment which demonstrated the small
size of the atomic nucleus, Rutherford observed that an alpha particle (Q =
2 X 1.6 X 10~ coulomb) with a kinetic energy of 7.68 10° electron-volts
(7.68 X 10° & 1.6 X 107! joule) making a head-on collision with a gold
nucleus (Q = 79 X 1.6 X 107!9 coulomb) is repelled.
(a) What is the distance of closest approach at which the electrostatic poten-
tial energy is equal to the initial kinetic energy? Express your result in fem-
tometers (10~!° meter).
(b) What is the maximum force of repulsion?
(c) What is the maximum acceleration in g’s? The mass of the alpha particle
is about 4 times that of a proton, or 4 X 1.7 X 10-2’ kilogram.
2-9, Electric charge is distributed along an infinite straight line with a density
\ = do/(\x| + /), where k is a positive constant.
(a) Is the total charge infinite?
(b) Can you calculate V at a distance p from the wire at x = 0? If not, can
you show that it is not infinite?

2-10. A flat sheet of copper carries a surface charge density of ¢ coulombs/meter?

on each face.
Use Gauss’s law to show that the electric field intensity just outside the
sheet is ¢/€) and that it is zero inside.

2-11. (a) Show that the electric field intensity near a flat sheet of charge having a
uniform surface density o is ¢/2¢.
(b) Deduce the value of the electric potential.

2-12. An electric field points everywhere in the z direction.

(a) What can you conclude about the values of the partial derivatives of E,
with respect to x, to y, and to z, (i) if the charge density p is zero, (ii) if the
charge density p is not zero?
(b) Sketch lines of force for some fields that are possible and for some that
are not.

2-13. A 1.00-microampere beam of protons is accelerated through a difference of

potential of 10,000 volts.
(a) Calculate the charge density once the protons have been accelerated,
assuming that the current density is uniform within a diameter of 2.00 milli-
meters and is zero outside this diameter.
(b) Calculate the radial electric field intensity both inside and outside the
beam.
(c) Draw a graph of the radial electric field intensity for values of r ranging
from zero to 1.00 centimeter.
(d) Now let the beam be situated along the axis of a grounded cylindrical
conducting tube with an inside radius of 1.00 centimeter. Draw a graph of V
inside the tube. :
86 ELECTROSTATIC FIELDS I

2-14. Two infinite parallel plates separated by a distance s are at potentials 0 and Vo.
(a) Use Poisson’s equation to find the potential V in the region between the
plates where the space charge density is p = po(x/s). The distance x is meas-
ured from the plate at zero potential.
(b) What are the charge densities on the plates?

_ A two-dimensional electrostatic field varies with the coordinates x and y but

is independent of z.
Show that the average value of the potential V on any circle parallel to
the xy-plane equals the potential at the center of the circle, provided the
charge density in the region is zero.

. You can use the property demonstrated in Problem 2-15 to calculate the
potential distribution within a square, two of whose adjacent edges are main-
tained at 100 volts while the other two are maintained at 0 and at 50 volts.
(a) First draw a square grid of 36 points. Of these, 20 will be on the edges
of the square, and 16 will be in the interior.
(b) Guess carefully the potentials at the interior points and write them on your
grid. This is your first approximation.
(c) Now you can correct your first approximation as follows. Start at an
interior point, near one corner of the square. Let us call this point P. Calculate
the sum ofthe potentials at the four nearest points and subtract four times the
potential at P. This number is called the residual at P. Show this number on
your grid, preferably in a different color, and repeat the calculation for the
other interior points.
To obtain a second approximation, select the points where the residuals
are the largest and add to your original estimates one quarter of the residuals.
Round off the potentials to the nearest volt. This makes the potential distribu-
tion satisfy more closely the condition found in Problem 2-15.
(d) Now calculate the new residuals and repeat the operation until the re-
siduals are everywhere less than 2, corresponding to corrections of less than
one-half volt.
You might wish to solve this problem on a computer.
(e) Sketch in some equipotentials.
This is a good illustration of the relaxation method of calculation, which
we largely owe to R. V. Southwell, and which is described in many books
on applied mathematics. The method is applicable to a wide variety of prob-
lems, and not only to the solution of Laplace’s equation.

2-17, A charged conductor has a small-diameter hole in a region where the surface
charge density is oc.
Show that the electric field intensity in the hole close to the surface of the
conductor is o/2¢.

2-18. The electric field intensity at the surface of a real conductor falls from its ex-
ternal value FE, to zero within the conductor in a finite distance 56,
1 Fi 6
Show that the surface charge density ¢ = if p(x) dx = e£,.
 PROBLEMS
87

2-19. t an early stage in the development of the atomic theory,

J. J. Thomson had
proposed an atom consisting of a positive charge Ze,
where Z is an integer
and e is the fundamental unit of charge (1.6 X 10~® coulomb
), uniformly
distributed throughout a sphere of radius a. The electron
s, of charge —e,
were considered to be point charges embedded in the positive
charge.
(a) Find the force acting on one electron as a function of its
distance r from
the center of the sphere. Assume that the other charges are uniforml
y dis-
tributed throughout the sphere.
(b) What type of motion does the electron execute?
(c) What is the frequency for a typical atomic radius of 1 angstrom?
(d) How does this frequency compare with that of the visible light radiated
by atoms?

2-20. (a) Calculate the electric field intensity in volts/meter at the surface of an
iodine nucleus (53 protons and 74 neutrons).
(b) Find the electric potential at the center of the nucleus.
Assume that the charge density is uniform and that the radius of the
nucleus is 1.25 X 107!°41/8 meter, where 4 is the total number of particles
in the nucleus.
(2A) A spherical charge distribution is given by

p=m(1-2), <a)
as) (Ca)
(a) Calculate the total charge QO.
(b) Find the electric field intensity E and the potential V outside the charge
distribution.
(c) Find £ and Vinside.
(d) Show that the maximum value of E is at (r/a) = 0.745.
(e) The above charge distribution applies roughly to light nuclei. Draw
graphs showing p, E, and V as functions of r/a for calcium (atomic number
20), assuming that pp = 5.0 & 102° coulombs/meter* and a = 4.5 femtometers
(1 femtometer = 10~' meter).

2-22. We shall see in Chapter 12 that an electromagnetic wave incident on a con-

ductor exerts a pressure on the conduction electrons. Imagine that you have
an insulated copper sphere one micron in diameter and that you suddenly
illuminate it with a powerful laser pulse that blasts essentially all the conduc-
tion electrons out of the sphere. There is one conduction electron per atom
in copper.
(a) What would be the voltage at the surface of the sphere immediately after
the irradiation ?
(b) What energy would be required in the laser pulse?
(c) Do you think the experiment is feasible?

2-23. Show that the dipole moment of an arbitrary charge distribution is independ-
ent of the choice of origin, provided the net charge in the distribution is zero.
88 ELECTROSTATIC FIELDS I

2-24. Compute the dipole moment of a spherical shell with a surface charge density
co = 0, cos 6, where @ is the polar angle.

2-25. Calculate the potential for a dipole exactly and indentify the quadrupole and
octupole terms.

2-26. A charge Q is uniformly distributed inside a cube of side a. Show that the
dipole and quadrupole moments of the charge distribution are zero.

2-27. A charge Q is uniformly distributed along a straight line of length a.

(a) Calculate V by integration, choosing the origin of coordinates at the
center of the charge.
(b) Calculate the monopole, dipole, and quadrupole terms in the multipole
expansion for V.
(c) For what values of the distance r to the center of the charge is the quad-
rupole term less than one per cent of the monopole term?
(d) Expand the value of V found in (a) up to terms in 1/r* and compare
with (b).

2-28. A parallel-plate capacitor is formed of two parallel conducting plates of area

S separated by a distance s in air. Use Gauss’s law to show that its capacitance
is e9S/s farads. Neglect edge effects.
Can you suggest reasonable dimensions for a 1 picofarad (10~” farad)
air-insulated parallel-plate capacitor which is to operate at a potential differ-
ence of 500 volts?

2-29. Find the capacitance per unit length of a capacitor consisting of a pair of
infinite coaxial cylinders having inner and outer radii a and 4, respectively.

a. The capacitance of a parallel-plate capacitor is e9S/s, where S is the area of

the plates and s is their separation, as long as edge effects are negligible.
(a) How is the capacitance affected if an insulated sheet of metal of thickness
s’ is introduced between the plates, but without touching either plate? (The
sheet need not be parallel to the plates.)
(b) What do you think of the suggestion of moving this sheet in order to vary
the capacitance without moving the capacitor plates?

2-31. Two capacitors of capacitance C, and C, have charges Q; and Qs, respectively.
(a) Calculate the amount of energy dissipated when they are connected in
parallel.
(b) How is this energy dissipated?

2-32. A capacitor consisting of two concentric spheres is arranged so that the outer
sphere can be separated and removed without disturbing the charges on either.
The radius of the inner sphere is a, that of the outer sphere is 6, and the charges
are O and — Q respectively.
(a) If the outer sphere is removed and restored to its original form, find the
increase in energy when the two spheres are separated by a large distance.
(b) Where does this extra energy come from?
 PROBLEMS 89

2-33. Consider an imaginary sphere of radius a centered on a dipole of moment p.

Show that the electric energy associated with the region of space outside the
sphere is W = (p?/127reoa*) by integrating the energy density from r = a to
r=,

2-34. According to the theory of relativity, a particle at rest has an energy myc®,
where mp is the particle’s mass and c is the velocity of light (Section 5.15).
Imagine that this is the electrostatic energy of the electron.
Find the radius of the electron (a) if the charge e is distributed uniformly
throughout its spherical volume, and (b) if the charge is distributed uniformly
r its spherical surface.

a) Calculate the electric potential energy of a sphere of radius R carrying a

total charge Q uniformly distributed throughout its volume.
(b) Then calculate the gravitational potential energy of a sphere of radius
R’ of total mass M.
(c) The Moon has a mass of 7.33 X 10? kilograms and a radius of 1.74 « 108
meters. Calculate its gravitational potential energy.
(d) Imagine that you can assemble a sphere of protons with a density equal
to that of water. What would be the radius of this sphere if its electric poten-
_tial energy were sufficient to blow up the Moon?

2-36. A light spherical balloon is made of conducting material. It is suggested that

it could be kept spherical simply by connecting it to a high-voltage source.
The balloon has a diameter of 10.0 centimeters, and the maximum breakdown
voltage in air is 3 X 10° volts/meter.
(a) What must be the voltage of the source if the electric force is to be as large
as possible?
(b) What gas pressure inside the balloon would produce the same effect?

Gaye have shown in Section 2.16 that the force of attraction between the plates
of a parallel-plate capacitor is (1/2)e9E2S when the plates are disconnected
from the voltage source. Check this result by repeating the calculation for the
case where the capacitor is connected to the source. The force should be the
same.
You will find that one half of the energy supplied by the source performs
mechanical work and that the other half increases the energy in the field.
This is a general result.

2-38. Find the time required for the plates (mass mp per unit area) of a parallel-
plate capacitor to come together when released from a separation Xo, (a) when
the plates are charged with a charge density o and then insulated, and (b) when
the plates are maintained at a constant potential difference V.

2-39. A variable capacitor consists of two thin coaxial cylinders of radii a and 5,
with (6 — a) <a, free to move with respect to each other in the axial direc-
tion.
 90 ELECTROSTATIC FIELDS I

(a) Using energy methods, compute the magnitude and direction of the force
on the inner cylinder when it is displaced with respect to the outer one.
(b) Explain how this force arises.

|
2 ia)Imagine the following simple-minded high-voltage generator. A parallel-plate
capacitor has one fixed plate that is permanently connected to ground, and
N
one plate that is movable. When the plates are close together at the distance
_s the capacitor is charged by a battery to a voltage V. Then the movable plate
is disconnected from the battery and moved out to a distance ns. The voltage
on this plate then increases to nV, if we neglect edge effects. Once the voltage
has been raised to nV, the plate is discharged through a load resistance.
(a) Verify that there is conservation of energy.
(b) Can you suggest a rough design for such a high-voltage generator with a
more convenient geometry?

2-41./Can you suggest a rough design for an electrostatic motor?

Draw a sketch, explain its operation, specify voltages and currents, and
make a rough estimate of what its power would be.
Why is it that electric forces should be so much weaker than magnetic
forces in such cases?

2-42. Normal seeds can be separated from discolored ones and from foreign objects
by means of an electrostatic seed-sorting apparatus that operates as follows.
The seeds are observed by a pair of photocells as they fall one by one inside
a tube. If the color is not right, voltage is applied to a needle that deposits a
charge on the seed. The seeds then fall between a pair of electrically charged
plates that deflect the undesired ones into a separate bin. One such machine
can sort peas at the rate of 100/second, or about 2 tons per 24-hour day.
(a) If the seeds are dropped at the rate of 100/second, over what distance
must they fall if they must be separated by 2 centimeters when they pass be-
tween the photocells? Neglect air resistance.
(b) Assuming that the seeds acquire a charge of 1.5 X 10-° coulomb, that
the deflecting plates are parallel and 5 centimeters apart, and that the poten-
tial difference between them is 25,000 volts, how long should the plates extend
below the charging needle if the charged seeds must be deflected by 4 centi-
meters on leaving the plates? Assume that the charging needle and the top of
the deflecting plates are close to the photocell.

Hint

2-37. Do not forget to take into account the energy supplied either by or to the battery.
 CHAPTER 3

ELECTROSTATIC FIELDS I
Dielectric Materials

Dielectrics differ from conductors in that they have no free charges that can
move through the material under the influence of an electric field. In dielec-
trics, all the electrons are bound; the only motion possible in the presence
of an electric field is a minute displacement of positive and negative charges
in opposite directions. The displacement is usually small compared to atomic
dimensions.
A dielectric in which this charge displacement has taken place is said to
be polarized, and its molecules are said to possess induced dipole moments.
These dipoles produce their own field, which adds to that of the external
charges. The dipole field and the externally applied electric field can be compa-
rable in magnitude.
In addition to displacing the positive and negative charges, an applied
electric field can also polarize a dielectric by orienting molecules that possess
a permanent dipole moment. Such molecules experience a torque which tends
to align them with the field, but collisions arising from the thermal agitation
of the molecules tend to destroy the alignment. An equilibrium polarization
is thus established in which there is, on the average, a net alignment. >
This is, in fact, a simplified view of dielectric behavior because many
solids are not made up of molecules but of individual ions which interact with
a large number of other ions. Sodium chloride is an example of such a solid.
We shall limit most of our discussion to dielectrics that are made up of
molecules that are either distorted or oriented, or both distorted and oriented
in an electric field.
In a polarized dielectric each molecule acts as an electric dipole of moment
92 ELECTROSTATIC FIELDS II

p. We shall treat these dipoles both from a molecular and from a macro-
scopic point of view: we shall consider the fields of individual dipoles, as well
as those of large numbers of dipoles.
In the end we shall deal with dielectrics entirely from the macroscopic
point of view, but we must justify this procedure by first examining carefully
the molecular aspect.

3 STHES PLECTRICSPOLARIZATIONS?

Let us first consider nonpolar dielectrics, that is, dielectrics whose molecules
have zero permanent electric dipole moment. We shall not have to consider
polar dielectrics until Section 3.9.
The electric polarization P is the dipole moment per unit volume at a
given point and is a macroscopic quantity. If p is the average electric dipole
moment per molecule in a small volume r, and if NV is the number of molecules
per unit volume, then
abr (3-1)
It will be recalled from Problem 2-23 that the dipole moment of a charge
distribution is independent of the origin of coordinates, provided the net
charge in the distribution is zero, as it is in molecules.
The volume + over which we define P must be chosen large enough to
render the fluctuations in P, from one instant to the next or from one rt toa
neighboring one, negligible. Although on the average the volume 7 contains
Nr molecules, we may expect, from statistical considerations, that there can
be fluctuations of +(Nr)!/ in the number Nr. The actual number of molecules
per unit volume thus fluctuates between N — (N/r)!2 and N + (N/r)!2, or
by +100 (1/Nr)!”? %. This restriction on the size of 7 is ordinarily unim-
portant. Take the case of water, for example, where N Y 3 X 1028 molecules/
meter’. Even in a submicroscopic cube measuring 100 angstroms or 10-8
meter on the side, the statistical fluctuation in P is only about one half of one
percent.

3.2 ELECTRIC FIELD AT AN EXTERIOR POINT

Figure 3-1 shows a block of dielectric material with a dipole moment P per
unit volume, P being a function of position within the dielectric. Let us
calculate the electric potential V which the dipoles in the dielectric produce at
a point P outside. This potential can then be added to that produced by all
 93

Figure 3-1. Block of dielectric

with a dipole moment P per unit
volume. The dipoles within the
element of volume shown inside
the block give rise to an electric
potential dV at the point P.

the other charges in the system to give the total potential. The negative
gradient of this total potential at the point P is the electric field intensity at
that point.
At the point P(x, y, z) the potential dV due to the dipole P dr’ situated at
(en ez.).15
1 P-r,
dV
Aney Fr?
dr’, (3-2)

xsi [P-v" (| He (3-3)

where V’ is evaluated at (x’, y’, z’) (Problem 1-12). We have assumed that the
distance r is large compared to the dimensions of a molecule so that the
dipole approximation can be valid. Integrating over the block of dielectric
gives

pracd i!|P-v" =)
Arey |7 r
a. (3-4)
This integration can be carried out analytically only for simple geome-
tries and only if P is a simple function of position within the dielectric.
Furthermore, P is never known apriori, although we can often deduce it from
measurable quantities. We shall nevertheless assume, for the time being,
that P is known. This integral can be put into a more interesting form by
transforming it through the vector identity

V' fA =fV'A+LAV'S, (3-5)

where
f and A are scalar and vector functions respectively. Let 1/r be the
scalar and P the vector function. Then

I P I V oP
V== Det
—— (vae . )dr ’ Aicet i -— dr dr, 3-6
(3-6)
 ELECTROSTATIC FIELDS II
94

We can use the divergence theorem to express the first term as a surface
integral; then

ge (ae ale [ME

Aireg Ip Are le
ee
where S’ is the surface that bounds the volume 7’ of the dielectric and the
vector da’ points outward.
Remember that this is the potential due to the dipoles in the dielectric.
To obtain the total potential we have to add to this V the V due to the charges
that produce the applied electric field.

3.2.1 The Bound Charge Densities p, and o,

Now both integrals in Eq. 3-7 involve a 1/r dependence and are multiplied by
1/4ré. We thus have the following remarkable result: these two terms are
exactly the potentials that would result from surface and volume charge
distributions having densitites

(3-8)
and
(3-9)
respectively, where n is the unit outward normal vector at the surface of the
dielectric. Thus
en til oda’ 1 pdr’
ie af. rp T Are, i" x ale

We have omitted the prime on V since it is obvious that the divergence must
be rmence at thea where the volume charge density is py.
» diele herefore be replaced by the bound chargedistributions —
OD — — re re ee field outside the dielectric. We shall see
later that we can use the same procedure to find a macroscopic potential at
points inside the dielectric.
We can demonstrate with the aid of Figure 3-2 that the bound electric
charge densities o, and p, represent actual accumulations of charge. Let us
first consider the surface density o,. We imagine a small element of surface da’
inside the dielectric. Under the action of the field, an average charge separa-
tion s is produced in the molecules. Positive charge crosses the surface by
moving in the direction of the field; negative charge crosses it by moving in
the opposite direction. For the purpose of our calculation, we may consider
the positive charge to be in the form of point charges Q and the negative
charge to be in the form of point charges — Q. Furthermore, we may consider
 95

Figure 3-2. Under the action of an electric field E, which is the resultant of an
external field and of the field of the dipoles within the dielectric, positive and
negative charges in the molecules are separated by an average distance s. In the
process a net charge dO = NQs-da’ crosses the surface da’, N being the number
of molecules per unit volume and Q the positive charge in a molecule. The vector
da is perpendicular to the shaded surface. The circles indicate the centers of
charge for the positive and for the negative charges in one molecule.

the negative charges to be fixed and the positive charges to move a distance s.
The amount of charge dQ that crosses da’ is then just the total amount of
positive charge within the imaginary parallelepiped shown in Figure 3-2. The
volume of this parallelepiped is
dr’ = s-da’, (3-11)
and
dQ = NQs-da’, (3-12)

where JN is the number of molecules per unit volume and Qs is the dipole
moment p of a molecule. Then
dQ = P-da’. (3-13)
If da’ is on the surface of the dielectric material, dQ accumulates there in a
layer of thickness s-n. Since the thickness of the layer is of the order of the
dimensions of a molecule, we may treat the charge as a surface distribution
with a density
dQ
= = P-n. (3-14)

The bound surface charge density o, is thus equal to the normal component
of the polarization vector at the surface.
We can show similarly that —V-P represents a volume density of
charge. The net charge that flows out of a volume r’ across an element da’ of
 96 ELECTROSTATIC FIELDS II

its surface is P-da’, as we found above. The net charge that flows out of the
surface S’ bounding 7’ is thus

O= | Pda’,is
(3-15)
and the net charge that remains within the volume r’ must be — Q. If p» is the
volume density of the charge remaining within this volume, then

[ova = Sor -| P-da’, (3-16)

==} (VeP) dr. (3-17)

Since this equation must be true for all 7’, the integrands must be equal at
every point, and the bound electric charge density is

eal ig (3-18)
We refer to o, and p, as either bound, polarization, or induced charge
densities, as distinguished from the free or conductible charge densities o; and
py. Bound charges are those that accumulate through the displacements that
occur on a molecular scale in the polarization process. The other charges are
“called free charges. The conduction electrons in a conductor and the electrons
injected into a dielectric with a high-energy electron beam are examples of —
free charges.*
oulomb’s law applies to any net accumulation of charge, regardless of—
a
other matter that may be present.

3.2.2 The Polarization Current Density

When a dielectric is placed in an electric field that is a function of the time, the
motion of the bound charges gives rise to a polarization current. We can find
the polarization current density J, by using again the fact that charge is
conserved.
For any volume r bounded by a surface S, the rate at which bound charge
flows out through S must be equal to the rate of decrease of the bound charge
within S:

*The concepts of free and bound charges, and even the concept of polarization, are
based on the assumption that the dielectric is composed of molecules. If there are no well-
defined molecules, the values of os», oy, and P become rather arbitrary. See, for example,
Edward Purcell, Electricity and Magnetism, Berkeley Physics Course Volume 2 (McGraw-
Hill, New York, 1965) p. 344,
3.3. ELECTRIC FIELD AT AN INTERIOR POINT 97

0
| san = - 2 f var (3-19)

Using the divergence theorem on the left and substituting the value of p, on
the right from Eq. 3-18,

0 oP
[ven dr = 2 [ve (eye |v : ap OT (3-20)

We have put the d/d¢ under the integral sign because it is immaterial whether
the derivative with respect to the time is calculated first or last. Since 7 is any
volume, the integrands must be equal and the polarization current density 1s

P
j= -.amperes/meter?. (3-21)

3:3 ELECTRIC FIELD AT AN INTERIOR POINT

It is often necessary to know the electric potential V and the electric field
intensity FEat points within a dielectric. We may wish to know, for example,
the potential difference between two charged conductors separated by a
dielectric.
The electric field intensity at a particular time and at a particular point
within the dielectric is a rapidly fluctuating quantity, both with regard to
position and to time. For example, it reaches enormous magnitudes at the
surfaces of nuclei. There are also large variations in its direction: on one side
of a nucleus it is in one direction, and on the other side it is in the opposite
direction. Thermal agitation of the molecules also produces large time
fluctuations at any given point.
The macroscopic electric field intensity FE, which is what we really wish
to know, is the space-and-time average of this electric field intensity.
The macroscopic electric field intensity E is a slowly varying function of
the coordinates and is independent of the time in the static case. It is this
electric field intensity that we shall integrate in order to calculate potential
differences.
We shall show that the part of this macroscopic field that originates from
the dipoles within the dielectric can be calculated from the bound charge
distributions o, and p, discussed above, exactly as in Eq. 3-10. This is not
intuitively obvious.
The reason the type of calculation used for the exterior point might not
be valid at an interior point is that, in the immediate neighborhood of the
point considered, there may be large numbers of molecules for which our
98

Figure 3-3. Imaginary sphere of

radius R, volume r’” and surface
S” inside a block of dielectric.
The remaining volume is 7’, and
its outside surface is S’.

expressions for the dipole field are not valid, since Eq. 3-2 requires that the
distance between the dipole and the point considered be much larger than
the dimensions of the dipole.
To calculate the macroscopic electric field intensity E inside a dielectric,
let us consider Figure 3-3, which shows a small imaginary sphere of radius R
centered on P. The surface S” of this sphere divides the dielectric into two
volumes, 7’’ outside and 7’” inside. The potential at P may be calculated for
all the polarized material outside S’” by treating P as an exterior point, as in
Section 3.2; however, the molecules within S” are too close to P to be treated
in this fashion. The volume r’”’ is macroscopically small, and the macroscopic
quantities E, P, and p, = —V-P do not vary significantly from one side to the
other.
Thus the macroscopic electric field intensity that the dipoles of the
dielectric produce at the point P is the sum of two terms, which we shall
calculate separately:
E=E" +E", (3-22)

3.3.1 Electric Field Intensity E”

Due to the Distant Dipoles
Since the point P is external to the volume 7’’, we can calculate E” from the
bound charge (a) on the outer surface S’, (b) in the volume 7”, and (c) on
the inner surface S’’. This latter surface is not real, but the discussion in
Section 3.2 applies even when only a portion of the dielectric is under con-
sideration, as it is here.
Each element of polarization charge p, dr” in 7’’ contributes at P an
element of field
lL py ar”
ri,
4rreo r
 99

Figure 3-4. Spherical cavity of radius R within a polarized dielectric. The

charges shown are due to the polarization P.

where r; is the unit vector in the direction from the element of volume dr’’ to
the point P, and r is the distance from dr”’ to P. Proceeding similarly for the
surface charges,
}
Vie L ov da’ I py dr”
f I
ely oy da”
oy da’’

Airey i ae | dire I po 7 Are Jgr? ry. |(3-23)

ake
Let us just calculate the third term for the moment. Figure 3-4 shows the
spherical surface S” with an axis drawn parallel to the polarization vector P
The polarization charge density on the surface S’’ is
o, = P-n = —P cos 0, (3-24)

and the charge on the annulus of width R dé contributes an element of field

intensity at P of
ae pepe 27 R? sin 6 dé cos 0.
Ate R

The first cos @ factor comes from o», and the second comes from taking the
100 ELECTROSTATIC FIELDS II

component of the electric field intensity in the axial direction, since symmetry
requires the resultant field to be in that direction. On performing the integra-
tion over the whole surface S’’, from 6 = 0 to 6 = zm, we find that the third
integral of Eq. 3-23 is equal to P/3¢e0.

Sos clectrics Ficidaintensity

Due to the Near Dipoles
The electric polarization P may be taken to be uniform inside the small sphere
S’”. Then we may assume that the electric field intensity E’”” at the center is
equal to the average electric field intensity inside S’’ due to this polarization;
this is the quantity that we calculated in Section 2-13, and
yee = N(4/3)nR?p | P (3-25)

ier ae
where WN is the number of molecules per unit volume and p is the average
dipole moment per molecule.
Thus the contribution of the molecules within S” to the macroscopic
electric field intensity E at P is equal in magnitude and opposite in direction
to the field of the bound charges on the surface S” (the third integral of Eq.
3-23), and
E = B+ Ms 4rreq
/ HE
r? Arey
| pe
pe

The second integral is calculated only over that portion of the dielectric
that is outside the spherical surface S”’. Of course, it would be simpler to
calculate the second integral if it extended over the complete volume 7’ of the
dielectric, but this would add the term
1 Db dr!"

Tite
Are) Jan?
Since the bound charge density does not vary significantly over the small
sphere 7’’’, this integral is merely the electric field intensity at the center of a
uniform spherical charge distribution. For every element of charge giving an
element of electric field intensity in one direction, there is a symmetrical
element of charge giving a field in the opposite direction, and this integral is
zero. The second integral of Eq. 3-26 can therefore be extended to the
com-
plete volume 7’ of the dielectric, and

_ 1 Ob da’ 2 l Pb dr’

‘ al ie piss Arey i ry - (3-27)

a l P-da’ : 1 i V-P d;’
Are if r Lin Arey [ a te AG (3-28)
 3.3. ELECTRIC FIELD AT AN INTERIOR POINT 101

That part of the electric field intensity that is due to the dielectric itself
can therefore be calculated at points inside the dielectric just as it is calculated
outside, namely by replacing the dielectric by a space charge distribution
—V-P and a surface charge distribution P-n.

_ REVIEW. —In the presence of an electric field, a dielectric becomes polarized

and acquires a dipole moment P per unit volume. We wish to calculate the
electric field produced by the polarized dielectric.
In Section 3.2 we calculated the field at a point situated outside the
dielectric. We started with the value of the electric potential V for a small
dipole P dr and integrated over the volume of the dielectric. The result (Eq.
3-28) showed that the field is just what one would expect from a volume charge
density p, = —V-P and a surface charge density o, = P-n. We then showed
that these two quantities do indeed represent real accumulations of charge.
Next we calculated the macroscopic electric field intensity EF at a point
inside the dielectric. This is the space-and-time average of the electric field
intensity over a small region of the dielectric.
We divided the dielectric into two regions, a small sphere of volume 7’”’
centered on the point P, and the remaining volume 7’’. The electric field
intensity at P due to the dipoles in r’’ comprises three terms arising from (a)
the surface charge density on the outside surface S’ of the dielectric, (b) the
volume charge density in 7’’, and (c) the surface charge density on the surface
S’’. It turns out that this last term just cancels the field at P due to the dipoles
within the sphere 7’’”’. We also found that the second term can be extended
without error to the complete volume of the dielectric.
The macroscopic electric field intensity E at a point inside a dielectric can
therefore be found by replacing the dielectric with the bound charge densities
p» and o» in exactly the same manner as for an external point. In other words,
we have found that the charge densities p, and o, represent actual accumula-
tions of charge, even though bound, and that they produce electric field
intensities according to Coulomb’s law. This is true inside as well as outside
the dielectric.
In calculating the macroscopic field E, we have said nothing about the
field of the external charge distribution that is responsible for the polarization
in the first place. It simply adds at every point to the field of the polarized
dielectric, to give the total macroscopic field.
The electric potential V and the electric field intensity E are therefore
given correctly by the following integrals, whether or not there are dielectrics,
or even conductors, in the field:
102 ELECTROSTATIC FIELDS I

1 Pf Se Pb ! 1 IOf ais ob da’ (3-29)

ee
ae r aa Ares J s/ r j

2 ye dr’ at = (a; 2,ar da’, (3-30)

=— Arey Js’ r

where p; is the free charge density,p, is the bound charge density, r is the
distance between the source point where the total charge density is py + p, and
the field point where V and E are calculated, r; is a unit vector pointing from
the source point fo the field point, and r’ is any volume enclosing all the
charges.

3.4 THE LOCAL FIELD

Although the above discussion clarifies our understanding of polarization

effects in dielectrics, it cannot be used to calculate the macroscopic E, since
we do not know as yet how to relate the electric polarization P to E. We must
eventually be able to calculate the electric potential and field intensity at any
point, either inside or outside a dielectric, without knowing P. We must be
able to calculate V and E knowing only the potentials of, or the charges on,
the conductors, plus the geometry and characteristics of the conductors and
dielectrics in the system.
In order to proceed we must know the average electric field intensity
acting on a molecule in the dielectric. We shall call this the /ocal field and shall
designate it Ejoc.
In the next section we shall see that the dipole moment p induced in a
single molecule is usually proportional to the local field.
We define F;,. in the following way. Imagine that we stop the thermal
motion of all the molecules in the dielectric at some particular instant. We
remove the molecule in question, keeping all the other molecules frozen in
position, and calculate the space-averaged electric field intensity in the cavity
previously occupied by the molecule. We replace the molecule, return the
system to normal temperature, and repeat the process many times. The
resultant space-and-time average of the electric field intensity in the cavity is
Evo.
We expect the local field to be larger than the macroscopic E, since Ejoc
does not include the field of the molecule under consideration, whereas the
macroscopic field does. In Section 2.13 we found that the average electric field
intensity within a sphere containing an arbitrary charge distribution is in the
direction opposite to the electric dipole moment of the distribution. Thus
 103

Gr)
Jas)
Ses
Reiners,
cai oes Figure 3-5. The field of an indi-
p=Qs vidual dipole in an electric field.

the molecule in question contributes an average field in the direction opposite

that of the macroscopic electric field intensity E, as indicated in Figure 3-5, at
least if the material is isotropic so that the molecules are free to orient their
dipole moments according to the direction of E.
A numerical calculation of the average electric field intensity due to the
molecule would require detailed information about its shape and charge
distribution, but we can perform the calculation in general terms, assuming
only the shape of the molecule. If we select the simplest shape, namely the
sphere, then this average field is (again from Section 2-13)
Pp P P
~ AreoR? ~ 4reNR? 3€eNTn (1)

where p is the dipole moment of the molecule and r,, is its volume.
If we further set 7, ~ 1/N, then

Evo E+ es (3-32)
3

This expression for the local field is at best approximate, owing to the
approximations that have gone into it. Although we assumed a spherical
shape for the molecule, we set tm = 1/N. This means that all of space is filled
with molecules, which contradicts the assumption of spherical shape. None-
theless, dielectric behavior predicted from the above expression for the local
field agrees surprisingly well with the experimental data for many substances.
Fortunately, our dielectric theory does not require that the expression
for the local field be quantitatively accurate; our conclusions will all be valid
if only
104 ELECTROSTATIC FIELDS II

eter eb ‘, (3-33)
0

where b is a constant that depends on the nature of the dielectric.

So far we have discussed dielectrics consisting of only one type of
molecule. More generally,

Pa Nipie Nopra (3-34)

where JN, is the number of molecules of the first type per unit volume, with the
average dipole moment pi, N, is the number of the second type per unit
volume, and so on. The theory is not otherwise modified.

Selb re CURIGsSUSGEPTIBIV
ERY Sy,

We can now find a relationship between the electric polarization P and the
macroscopic electric field intensity E. To do this, we consider the electric
dipole moment p induced in a single molecule. The magnitude of p depends
on the local field E,,.: the positive and negative charges are separated under
the action of the local field until the restoring force, which is an internal
electrostatic force arising from their separation, just balances the local field
LOrce:
In most dielectrics the molecular charge separation is directly propor-
tional to, and in the same direction as, the local field. Thus

P = AB, = a (B+b =) 0
(3-35)
where the constant a is known as the molecular polarizability.
Dielectrics which show this simple dependence of polarization on local
field are said to be linear and isotropic. Many commercially important
dielectrics are homogeneous, as well as linear and isotropic. We shall desig-
nate such materials as Class A dielectrics and shall confine our attention to
them for the time being.
The dipole moment per unit volume in Class A dielectries is

P = Np = Na Ej, = Na (B+b =) (3-36)

€0

where N is again the number of molecules per unit volume, and

P = No :
t= (Nob/e) al!
= ex, (3-38)
3.6 THE DIVERGENCE OF E. THE ELECTRIC DISPLACEMENT D 105

where x, is a dimensionless constant known as the electric susceptibility of the

dielectric. The constant ¢ is included explicitly to make x, dimensionless.
For a Class A dielectric, the electric polarization P is thus proportional
to the macroscopic electric field intensity E as well as to the local field Ejcc.

3.6 THE DIVERGENCE OF E.

THE ELECTRIC DISPLACEMENT D

We have found that the electric field due to the atomic and molecular dipoles
of a polarized dielectric can be calculated from the bound charge densities and
from Coulomb’s law, just as for any other charge distribution. Let us investi-
gate the implications of this fact on Gauss’s law, which is a direct consequence
of Coulomb’s law.
Gauss’s law relates the flux of the electric field intensity E through a
closed surface to the total net charge Q; enclosed within that surface:

| B40 = iV-Edr= Qe. (3-39)

€0

For dielectrics, Q; must include bound as well as free charges:

Ot i(oy + pr) dr, (3-40)

where the integration is intended to cover both the surface and volume
distributions of free and bound charges.
If we substitute this value of Q; into Gauss’s law and equate the inte-
grands of the volume integrals, then

VB =e (3-41)

at every point, where py = py + p» is the total net charge density. This is

Gauss’s law in its more general form. It is one of Maxwell's four fundamental
equations of electromagnetism. It follows from (a) Coulomb’s law, (b) the
concept of electric field intensity, (c) the principle of superposition, and (d)
the fact that the field of the dipoles in a polarized medium can be calculated
everywhere from the bound charge densities p, and op.
Since E = —VV, it follows that

a ees (3-42)
106 ELECTROSTATIC FIELDS II

This is Poisson’s equation for V. This equation is valid in dielectrics and is

more general than Eq. 2-31. (In Chapter 2 there was no need to distinguish
between free and bound charges because all the charges considered were of
the former type; the p’s of Chapter 2 are therefore p;’s.)
Now p, = —V-P, and, for any dielectric,

V-E = +(/—V-P),
0
(3-43)
or
V -(ek + P) = py. (3-44)

The vector esf + P is therefore such that its divergence depends only on the
free charge density p;. This vector is called the electric displacement and is
designated by D:
D = ok + P. (3-45)
Thus
V-D = py. (3-46)
In integral form Gauss’s law for D becomes

[Deda = |evar
S T
(3-47)
and the flux of the electric displacement D through a closed surface is equal
to the free charge enclosed by the surface.
From the definition of the electric displacement,

ee (3-48)
and the electric field intensity E inside a dielectric is the sum of two fields:
D/e associated with the free charges, since

Ve (7)
D(a eeePy"
(3-49)
%

and —P/e9 associated with the bound charges, since

Win ((tas)P-) eeePb z

(3-50)

Lines of D begin or end only on free charges, whereas lines of

E begin or
end on either free or bound charges. This does not mean,
however, that the
magnitude and direction of D at a particular point depend
only on the free
charges. The only quantity that depends on the free charges
alone is V-D. To
find D, we must integrate Eq. 3-46, Subject to whatever boundar
y conditions
 107

Np On = 12
P=

: Pa
ait field intensity
Electric =oe =eg
= = = |]
Electric displacement ——>—__—__

Figure 3-6. A block of dielectric located between the plates of a parallel-plate

capacitor. Each molecule of the dielectric has a dipole moment p, leading to a
polarization P. The electric displacement D depends only on the free charges +o;
and is the same inside and outside the dielectric. The electric field intensity E, on
the other hand, is reduced inside the dielectric because the polarization charges
+o, produce a field in the opposite direction.

obtain in the particular field. Figure 3-6 shows the vectors p, FE, D and the
electric charge densities o; and o, for a parallel-plate capacitor.
In writing down the expressions for V-E and V-P, we have implicity
assumed the existence of the space derivatives of E and P, These derivatives
do not exist, of course, either at a point charge or at the interface between two
media. In such cases, we must revert to the integral form of Gauss’s law:

«| E-da = {we + p,) dt. (3-51)

Ss Tt
108 ELECTROSTATIC FIELDS II

3.6.1 The Relative Permittivity «,.

Poisson’s Equation for Dielectrics
For a Class A dielectric we have, from Eqs. 3-45 and 3-38,
D = &(1+~x.) E = oe E = ch, (3-52)
where
(3-53)

is a dimensionless constant known as the relative permittivity, or the dielectric

constant of the material. In a vacuum, x, = 0, e, = 1, and in all Class A
dielectrics «, > 1.
The relative permittivity of dielectrics lies typically between 2 and 5.
However, some substances which are not Class A dielectrics can have
relative permittivities as high as 10° or even higher.
In Class A dielectrics D is proportional to E, and, from Eqs. 3-46 and
3-52, if ¢, is not a function of the coordinates,

Veh (3-54)
€,€0 €

Then
4 pe eal (3-55)
Er€Q €

This is Poisson’s equation for homogeneous Class A dielectrics. If the dielectric

is not Class A, we must write

Vive _ (or + po), (3-56)

€0

3.6.2 The Free Charge Density p,

and the Bound Charge Density p,
The bound charge density p, is related to the free charge density p; in a Class
A dielectric as follows. Since

e,
—l
P=D—6E=
€;
D, (3-57)
and since the dielectric is homogeneous (c, is independent of the coordinat
es)
by hypothesis,

V-eP= € — 1 1
( )v-p = (1- =|Pf, (3-58)
€; ,
3.7 CALCULATION OF ELECTRIC FIELDS INVOLVING DIELECTRICS 109

or

re (*és‘) (3-59)
The total charge density

pe = pp + po = pp —V-P = & (3-60)

E>
is smaller than the free charge density p;, since pp cancels part of py.
At any point in a homogeneous, isotropic, linear dielectric, p, is zero if
py is zero, from Eq. 3-60. Usually, p; = 0, V-P = 0, and the polarization
charges are located only on the surfaces of the dielectric.

3.7 CALCULATION OF ELECTRIC FIELDS

INVOLVING DIELECTRICS

We are now in a position to calculate electric fields involving Class A die-

lectrics.
If the geometry of the field is simple and if the free electric charge density
py 1S Zero, as it often is, we can usually integrate V-D = 0 to find the electric
displacement D. The constants of integration can be determined from the
boundary conditions, which can be, for example, the charge densities on
the conductors. The integral equation

[oda = [ oer, (3-61)

S T

is also often useful for determining D.

Once D is determined, E can be found at once from E = D/e,¢), provided
that the relative permittivity «, is known. It is then a simple step from E to
potential difference or to capacitance, or to other physical quantities depend-
ing on E.
Combining Eqs. 3-29 and 3-60, we also have

eae Ee a Bae | Case ip (3-62)

Ag Jy er Arey S’ ‘

for linear and isotropic dielectrics. Note that p;/e is equal to p;/¢, but that no
such relation exists for the surface charge densities a; and a; + oa». Indeed, oy
and o, are unrelated. For example, a dielectric carrying zero of acquires a oy
when it is placed in an electric field. (However, see the second example that
follows.)
110 ELECTROSTATIC FIELDS II

Example THE DIELECTRIC-INSULATED

PARALLEL-PLATE CAPACITOR
Figure 3-7 shows such a capacitor with the spacing s exaggerated
for clarity. We assume that s is small compared to the linear extent
of the plates in order that we may neglect fringing effects at the edges.
The surface charge densities +-c,; on the lower plate and —o; on the
upper plate produce a uniform electric field directed upward. The
polarization in the dielectric produces a bound surface charge den-
sity —o» on the lower surface of the dielectric and +c, on the upper
surface as in Figure 3-6. These bound charges produce a uniform
electric field directed downward which cancels part of the field of
the charges situated on the plates. Since the net field between the
plates must remain equal to V/s, where V is the battery voltage, the
charge densities +-c; and —oa; on the plates must be larger than when
the dielectric is absent. The presence of the dielectric thus has the
effect of increasing the charges on the plates at a given value of V,
and hence of increasing the capacitance.
To calculate the capacitance, we apply Gauss’s law for the dis-
placement D to a cylinder as in Figure 3-7. Then the only flux of D
through the Gaussian surface is through the top, and D is numeri-
cally equal to a;. Also E = a;/e,€o, the potential difference between
the plates is o;s/e,¢), and the capacitance

= =. = Ge 5, (3-63)
where S is the area of one plate. The capacitance is therefore in-
creased by a factor e, through the presence of the dielectric.
The measurement of the capacitance of a suitable capacitor
with and without a dielectric provides a convenient method for
measuring a relative permittivity e,.

Figure 3-7. Dielectric insulated

plane-parallel capacitor.
3.7 CALCULATION OF ELECTRIC FIELDS INVOLVING DIELECTRICS 111

Example THE FREE CHARGE DENSITY o;,

THE BOUND CHARGE DENSITY «,
AND THE ELECTRIC DISPLACEMENT D
AT A DIELECTRIC-CONDUCTOR BOUNDARY

At the interface between a dielectric and a conductor there is a

bound surface charge density on the dielectric and a free surface
charge density on the conductor. For example, at the lower plate
of the capacitor considered above,

oo = P-n = —P, (3-64)

since n is the outward normal to the dielectric surface and points
downward, while the electric polarization P points upward like E.
Thus
—o, =P = D— oE, (3-65)

SeWere 1 \
eeeTee
(3-66)
€r €;

and

iS = (3-67)
as in Eq. 3-60.
Since e, > 1, the bound surface charge density o, and the free
surface charge density o; have opposite signs. The bound charge
density 1s also smaller in magnitude than the free charge density by
the factor (e, — 1)/e,. This relationship is always true whenever a
Class A dielectric is in contact with a conductor, or carries a free
surface charge density.
The free charge density a; on a conductor in contact with any
dielectric is always equal to the electric displacement D just inside
the dielectric. This follows from Gauss’s law, as in the case of the
parallel-plate capacitor. Furthermore, if a conducting surface is
introduced into a dielectric so as to coincide with an equipotential
surface, then the free surface charge density a; on the conductor at
a particular point is equal in magnitude to the electric displacement
D in the dielectric at that point.

Example DIELECTRIC SPHERE WITH A POINT CHARGE AT

ITS CENTER
Let us consider a Class A dielectric sphere of radius R with a point
charge Q embedded at the center, as in Figure 3-8. (It is possible
to trap negative charges within dielectrics by bombarding them with
high-energy electrons. This example is of little practical interest,
but it proves to be an excellent illustration of the behavior of di-
electrics.)
112

Figure 3-8. Dielectric sphere with a

point charge Q at the center.

Again, we can find D from Gauss’s law. We draw an imaginary

sphere of radius r < R and use it as a Gaussian surface. Then

= mre (3-68)

ee~ Aarecer?
3-69
Ce
Fey€; jee
én 4nr
(3-70)
At the outer surface of the sphere,
¢-—1 Q
op = Pen =P= aa ne (3-71)

The total amount of bound charge on the outer surface is thus

,—1
One €,
(3-72)
Since there is no volume density of free charge, there should be
no volume density of bound charge, as we saw in the last section.
This is correct:
pp = —V°P = —5 5 P) == (). (3-73)

There is also a bound charge on the surface of the cavity con-

taining the free charge Q at the center. This can be calculated as
follows. Let the radius of the cavity be 6. Then, if —o», is the density
of bound charge on the cavity surface, the total bound charge on the
cavity is

Or = 04408 = =Pa ee — (== _)Ds4n8, (3-74)

E a) 0. | (3-75)
The dielectric as a whole therefore remains neutral, as must be ex-
expected. The net charge at the center is then
3.7 CALCULATION OF ELECTRIC FIELDS INVOLVING DIELECTRICS 113

Qret = Q — (a= )Q= g (3-76)

and is smaller than the free charge by the factor e,. This accounts
for the reduction of the electric field intensity within the dielectric
by the factor ¢,, as in Eq. 3-69.
It is instructive to compare the electric field intensity outside
the sphere and that inside. Outside, from Gauss’s law,

Fas (3-77)

=Pa
eee (3-78)

which is to be compared to Eqs. 3-68 and 3-69 for the field inside the
sphere. At the surface of the sphere the electric field intensity is dis-
continuous, the magnitude just outside the surface being e, times as
large as the magnitude just inside the surface. The difference is due
to the bound charges, which produce an opposing field within the
dielectric.

Example THE BAR ELECTRET

An electret is the electric equivalent of a bar magnet. In most di-
electrics the polarization disappears immediately when the electric
field is removed, but some dielectrics retain their polarization for a
very long time. Indeed, several polymers have extrapolated lifetimes
of several thousand years at room temperature.
One way of charging a dielectric is to place it in a strong electric
field at some high temperature. The bound charge density on the
surfaces then builds up slowly as the molecules orient themselves.
Free charges are also deposited on the surfaces when sparking occurs

(a) (b)

Figure 3-9. (a) Bar electret polarized uniformly parallel to its axis.
(b) The E field of the bar electret is the same as that of a pair of circular
plates carrying uniform surface charge densities of opposite polarities.
114

Figure 3-10. Lines of E (solid) for the bar electret of Figure 3-9. Lines of
D are shown broken inside the electret; outside, they follow the lines
of E.

between the electrodes and the dielectic. The free and the bound
charges have opposite signs. The sample is then cooled to room tem-
perature without removing the electric field,
In the presence of an external electric field the polarization P
inside an electret is related to the total E as follows:

P= x.0E + Po (3-79)
where Py is the permanent polarization. Then
D = @E + P = e(1 + x,)E + Po, (3-80)
= «E+ Pp. (3-81)
Since there are no free charges inside the material, p, = 0 (from
 3.8 THE CLAUSIUS-MOSSOTTI EQUATION 115

Eq. 3-60) and the electric field of an electret can be calculated solely
from the surface charges.
As an exercise, let us see how one can calculate the field of the
bar electret illustrated in Figure 3-9a. We assume that the polariza-
tion P is uniform. We also assume that the surface density of free
charges oy is zero.
The electric field intensity E both inside and outside is the same
as that of a pair of parallel circular plates carrying charge densities
+P as in Figure 3-9b. This field is shown as solid lines in Figure 3-10.
The lines of D are identical to the lines of E outside the electret
since D = ¢,E there. Inside the electret, D is «JE + P with E and P
pointing in approximately opposite directions. If the electret were a
thin sheet polarized in the direction normal to its surface, the E inside
would be P/e, and D would be zero. However, inside a bar electret,
€,E is smaller than P, the result that D points in the direction of P,
and E and D point in approximately opposite directions, as in Figure
3-10.

3.8 THE CLAUSIUS-MOSSOTTI EQUATION *

Let us return now to the basic polarization mechanism and compute the rela-
tive permittivity e, in terms of the molecular properties. We shall assume the
approximate expression of Eq. 3-32 for the local field:

ie
Evo = E + (3-82)
3€

As we saw, this equation is not necessarily correct, in that the factor multi-
plying P/e) may differ from 3, but 3 is of the right order of magnitude.
If we set b = 3, then from Eq. 3-36,

ji (E x) (3-83)
3€

or
ee ,
e(e, — IE = Na ( te —— ) (3-84)

and
Glee Ne. (3-85)
—> €; +. 2} 3€

the molecular
where N is the number of molecules per unit volume and a 1s
polarizability.

y.
* Sections 3-8 to 3-10 can be omitted without losing continuit
116 ELECTROSTATIC FIELDS II

If the dielectric is a compound consisting of a number of different types

of molecules or atoms,
nee
é +2 36 Ds Wee
0; (
3-86 )
a

where N; and a; are the appropriate quantities for the ith type of molecule or
atom.
Equation 3-86 is known as the Clausius-Mossotti equation. It relates the
relative permittivity «, to the mass density of a material, since

N= ArNa, (3-87)
where p is the mass density, V4 is Avogadro’s number, and M is the molecular
weight. In MKSA units, p is expressed in kilograms/meter*®, NV, is the number
of molecules in one kilogram molecular weight, or 6.02 X 10%, and M is the
kilogram molecular weight (for example, the kilogram molecular weight of
oxygen is 32 kilograms). Then
Me— Na
a= ay. (3-88)
p&+2 3¢€

The quantity az is called the molar polarization.

Now in this equation Avogadro’s number N, and the permittivity of
free space e are of course the same for all substances, while both the molec-
ular weight M and the molecular polarizability a depend only on the type
of molecule of which the dielectric is made. Then Eq. 3-88 shows that the
quantity (e, — 1)/p(e, + 2) is independent of density if our model for the
polarization is valid. This prediction of the Clausius-Mossotti equation is
borne out in a wide variety of gases and is approximated in nonpolar liquids.
Examination of Eq. 3-88 shows that ¢, must increase rapidly and tend to
infinity as the number of molecules per unit volume approaches the critical
value 3¢)/a. Physically, the expression ¢, > 2% corresponds to infinitely polar-
izable molecules. In gases and in liquids the polarizabilities are relatively
small, with the result that this condition is never approached.
In crystalline solids the interactions between the dipoles produce polar-
izations that are more complicated than those of our simple model, and the
Clausius-Mossotti equation is not valid.

SO ePOCARSDIELECTRICS

Let us now investigate the behavior of dielectrics in which the

molecules
possess permanent dipole moments. Molecules consisting of
two or more
 117

Figure 3-11. Dipole with a per-

manent dipole moment p = Qs
=O) ea oriented at an angle @ with
7 E 1. respect to the local field.

dissimilar atoms exhibit such permanent moments. For example, a molecule

composed of two atoms, one of which carries a positive charge and the other
a negative charge, is polar. Other mechanisms exist by which other molecular
types can also have permanent moments. Diatomic molecules consisting of
two similar atoms never have permanent moments because there is no asym-
metrical way for them to arrange themselves.
In addition to their permanent dipole moments, polar molecules also
exhibit induced moments of the type we have been discussing, but for the
time being we shall neglect these induced moments.
Permanent dipole moments are usually of the order of 10-* to 10-*
coulomb-meter.
Equation 3-1 obviously does not apply to polar dielectrics, since the
individual p’s are not necessarily aligned with the field. Our problem is to
calculate their net orientation.
A permanent dipole oriented so that its dipole moment p makes an angle
6 with the local field, as in Figure 3-11, is subjected to a torque that tends to
align it with the field. An unaligned dipole therefore has a greater potential
energy than an aligned one.
It is a simple matter to calculate the potential energy of such a dipole.
We select arbitrarily the position 9 = 90° to correspond to zero potential
energy. The axis ofthe dipole is then perpendicular to Ejoc. This is permissible,
since only variations in potential energy have physical significance. Then the
energy lost by the dipole when its axis rotates from 90° to @ is Ejo. Qs cos 6
and the potential energy
— — p* joc. (3-89)

The potential energy is minimum when p is parallel to Ejoc.

3.9.1 The Langevin Equation

In gaseous and liquid polar dielectrics, thermal agitation brings about colli-
sions between molecules that tend to destroy any alignment with the local
118 ELECTROSTATIC FIELDS II

field. However, the local field exerts a restoring force between collisions and,
on the average, effects a net alignment, and consequently a net dipole moment
P per unit volume.
To compute this net polarization, we consider a unit volume containing
N dipoles. In the absence of an external electric field, the dipoles are oriented
at random, and, at any instant, dN are oriented at angles lying between @ and
6 + dé with respect to a given direction. The fraction dN/N is merely the ratio
of the solid angle corresponding to the angular interval dé to the total solid
angle 47:
dN 2xrsin@dé _ sinédé
(3-90)
Nite ain ee? ae
If the dipoles are subjected to a local field, the solid-angle elements are no
longer equally probable, since a dipole has a potential energy W, which de-
pends on its orientation relative to the field. It is shown in statistical me-
chanics that if a large number of molecules are in statistical equilibrium, the
number possessing a particular energy W is proportional to exp (—W/kT),
where k is the Boltzmann constant 1.381 X 10-* joule/kelvin and T is the
absolute temperature in kelvins.
In the present case the dipoles whose axes lie in the range between @ and
6 + dé, measured from the direction of the local field, all possess an energy
W = —p- Ej. = —PpE to. cos 8. Then the number of dipoles per unit volume
lying in the interval 6 to 6 + dé is

dN = C exp (p>Ejoc/kT) sin 6 d8 = C exp (u cos 6) sin 6 dé, (3-91)

where we have set

(3-92)
2! PEt.
u= kT ’

and where the constant C must be chosen so that the total number of mole-
cules in a unit volume is NV:

N= all exp (u cos 6) sin 6 dé. (3-93)

0

Now the molecules whose moments lie in the angular interval 6 to 6 +

dé
possess a total dipole moment in the direction of the local field
of
dP ip dN cos bie ne (u cos 6) sin
6 cos 0 do
; (3-94)
> eXP (u cos @) sin 6 dé

To obtain the dipole moment per unit volume P we now

integrate the numer-
ator from @ = 0 to 6 = x:
3.9 POLAR DIELECTRICS 119

i exp (u cos 6) sin 6 cos 6 dé

PEND = (3-95)
if exp (u cos 6) sin 6 dé

So as to simplify the integration we set

t= Pees cos 6 = ucos 0. (3-96)

Then
Np +u
ao exp ft) t dt
pa wim SEDI) _ Np [t (expt) — (exp )\r% (3-97)
ie (exp ft) dt a [exp ¢]=%

na ~~ 1 rs. DENce % il
= Np (coth u ‘)= Np (cotn kT =) (3-98)

This is known as the Langevin equation.

Figure 3-12 shows P/Np as a function of u. For large values of u =
PE\o./kT, that is, for large fields and low temperatures, P approaches Np.
The dipoles are then all aligned with the field, and the polarization is maxi-
mum.
The region of practical interest is where pE,../kT is small compared to
unity. At room temperature kT ~ 4 X 10°?! joule, whereas a typical dipole
moment is of the order of 10-*° coulomb-meter. Thus, even with a local field
of 10’ volts/meter, pE\../kTis only of the order of 2 X 10-*. We can therefore
expand the exponentials in Eq. 3-98 and retain only the terms up to u*. Then

SS Pe geen :
Pa Nola eae ~ul Coy
Np
ANE
ee
ogy
OD)
ie
(1= =)
6
= u|
Al (3-100)

~py Npu
OE _ Np
TE Boe :
(3-101)

Thus, when pE\o. < kT, the polarization P in a polar dielectric is proportional
to the local field, and the dielectric is linear. We found earlier in Eq. 3-36 that
the same applies to nonpolar dielectrics. When pE\oc < kT the susceptibility
of polar dielectric is inversely proportional to the temperature.
From a practical point of view, the only feature that distinguishes a polar
from a nonpolar dielectric is the temperature dependence of the susceptibility
and, therefore, of the relative permittivity.
120

ea em a a (a Pe a Ee ee ol
0 1 2 3 4 5 6 7 8 9 10 11 1:
yx PE toc

kT
Figure 3-12. The Langevin function.

3.9.2 The Debye Equation

Let us now consider real polar dielectrics in which there are both induced and
permanent dipoles. Combining Eqs. 3-36 and 3-101,

=
Pea («a LB
ia) Ejiocs :
(3-102)
where
ieee JED
(3-103)
as in Eq. 3-32 or, for a linear dielectric (Eqs. 3-38 and 3-53),

Eee ot? E. (3-104)

Then
P=2 ee Sn a
— DE=N PO \ fe 72
(«+ )( 3 )E, (3-105)

at he
peas oe (0T Fz) Golly
Multiplying by the molecular weight M and dividing by the mass density
p, aS we did in discussing the Clausius-Mossotti equation, we find a new
expression for the molar polarization that is valid for polar dielectrics:

_ iM geal NG P
OM a hace (0+ i) (3-107)

This is known as the Debye equation. It is similar to Eq. 3-88, except for
the term p?/3kT, which comes from the alignment of the polar molecules.
In principle, the Debye equation can be used to determine both the
 121
NANG 1
(P| [Ge ae 2

Figure 3-13. The molecular

polarizability a and the dipole
moment per molecule p can be
determined from the intercept on
the vertical axis and from the
1/T slope of this curve.

molecular polarizability a and the permanent dipole moment p of a molecule.

Plotting (M/p)(e, — 1)/(e, + 2) versus 1/T gives a straight line, as in Figure
3-13, whose intercept on the vertical axis is N4a/3e) and whose slope is
Nap’/9ek, where N4 is Avogadro’s number and k is the Boltzmann constant.
In practice, this equation provides a reliable way of measuring a and p only
for gases, where the dielectric coefficient differs only slightly from unity, and
for dilute solutions of polar molecules in a nonpolar solvent.
We have made one important assumption in deriving the Debye equation.
We have assumed that the field F,,.. = E + (P/3e0) is responsible both for the
induced and the oriented polarization. We saw that the factor 4} is only
approximate. Furthermore, there is another more subtle effect that further
invalidates the Debye equation. As a permanent dipole turns, part of the
local field turns with it and plays no part in orienting the dipole. This comes
about because the field of the dipole polarizes the surrounding molecules, and
these in turn produce a field at the dipole that is in the same direction as its
moment. Furthermore, molecular associations in liquids and in solids com-
plicate the problem. The result is that dielectric behavior in liquids, and
especially in crystalline solids, is more complicated than the simple picture
we have developed here. The molar polarization of water, for example, shows
little temperature dependence, even though water has a large permanent
dipole moment of 6.2 X 10-* coulomb-meter. The basic features of the po-
larization process are as we have discussed them, but a precise determination
of molecular polarizability and of permanent dipole moment must rest on a
more sophisticated treatment.

310, FREQUENCY: DEPENDENCE,

ANISOTROPY, AND NONHOMOGENEITY

We have discussed so far two basic polarization processes. We first discussed

induced polarization, in which the center of negative charge in a molecule is
122 ELECTROSTATIC FIELDS II

displaced relative to the center of positive charge when an external field is

applied. This type of polarization is also called electronic. We then considered
orientational polarization, in which molecules with a permanent dipole mo-
ment tend to be aligned by an external field, the magnitude of the suscepti-
bility being inversely proportional to the temperature. There is also a third
basic polarization process, which we may call ionic. This process occurs in
ionic crystals, in which ions of one sign may move with respect to ions of the
other sign when an external field is applied.

Examples Water has a relative permittivity of 81 in an electrostatic field, and

of about 1.8 at optical frequencies. The large static value is attrib-
utable to the orientation of the permanent dipole moments, but the
rotational inertia of the molecules is much too large for any signifi-
cant response at optical frequencies.
Similarly, the relative permittivity of sodium chloride is 5.6
in an electrostatic field and 2.3 at optical frequencies. The larger
static value is attributed to ionic motion, which again is impossible
at high frequencies.

For a given magnitude of E, both the magnitude and the phase of p are
functions of the frequency, first because of the various polarization processes
that come into play as the frequency changes, and also because of the exist-
ence of resonances. Thus, since e, is a function of the frequency, e, is strictly
definable only for a pure sine wave.
In many substances the relative permittivity decreases by a large factor
as the temperature is lowered through the freezing point.

Example In nitrobenzene, e, falls from about 35 to about 3 in passing through

the freezing point at 279 kelvins. In the solid state, the permanent
dipoles of the nitrobenzene molecules are fixed rigidly in the crystal
lattice and cannot rotate under the influence of an external field.

Anisotropy is a common departure from the ideal Class A dielectric

behavior. Crystalline solids commonly have different dielectric properties in
different crystal directions because the charges which constitute the atoms
of
the crystal are able to move more easily in some directions than in others.
The
result is that the susceptibility depends on direction. Thus, in general,
P is not
in the same direction as E. For example,

Pe — €0(XerrLx a Nowy Lig = Xanalra) (3-108)

or, more generally,

3.11 POTENTIAL ENERGY OF A CHARGE DISTRIBUTION 123

Pee egEy J
(3-109)
where the subscripts i and j represent the three coordinate directions Kaa es
All three components of P depend on all three components of E, with different
susceptibilities for each. The susceptibility x, thus has nine components and is
a tensor. Actually, there are only six independent components, and, if the
coordinate axes are properly chosen, these six components reduce to three.
The relationship between P and Eis still linear but is more complicated than
for isotropic dielectrics.
For anisotropic dielectrics, Eq. 3-44 still applies:

V-(ek + P) = py, (3-110)

where p; is the free charge density. It is again useful to define the displacement
vector as in Eq. 3-45:
D = oE+ P, (3-111)
except that now D, E, and P are generally not all in the same direction.
The relation between the displacement D and the electric field intensity
E is then
Di = © >, &risBi, (3-112)
Ht)

where e¢,;; is a tensor which, again, has six independent components. The re-
lationship between D and E is still linear, and the general features of the
discussion of Class A dielectrics are valid.
If the relative permittivity is a function of position in the dielectric, then
a volume density of bound charge may exist when there is no corresponding
volume density of free charge (see Eq. 3-60). Demonstration of this fact will
be left as an exercise.

3.11 POTENTIAL ENERGY

OF A CHARGE DISTRIBUTION
IN THE PRESENCE OF DIELECTRICS

If dielectrics are present in the vicinity of a charge distribution, we may still

calculate the electric potential energy of the system in the manner of Section
2.14. We allow each free charge to recede to infinity as before, except that now
the potential at a given point in the field depends not only on the charges and
on the geometry, but also on the characteristics of the dielectric. Then

W= 5| Var (3-113)
124 ELECTROSTATIC FIELDS II

where the volume r is any volume which includes all the free charges in the
system. Now p; = V-D, and, if we use the vector identity of Problem 1-14,

w= :i V(V-D) dr, (3-114)

Z S| w-vpya jh [ove (3-115)

Transforming the first term to a surface integral,

WV || VD-da + i [on dr. (3-116)

Djs 25
We now proceed as in Section 2.14 and let r and S tend to infinity. In
the first integral, V falls off at least as fast as 1/r, D falls off at least as fast as
1/r?, and da increases as r*. The first integral falls off at least as fast as 1/r
and goes to zero as the surface S recedes to infinity. Thus

W= 5 ome, (3-117)
where 7 is any volume that includes all the points where D and E differ from
zero. This expression is independent of the type of dielectric present in the
system and is completely general.
As for free space, we may define an energy density
dw 1
aaa 7 D-E, (3-118)

which is useful in calculating the work required to assemble a charge distri-

bution.
For a Class A dielectric, D = ¢,6F,

Wea | a dr, (3-119)

and the energy density is «E£?/2.

Example ENERGY STORED IN A

PARALLEL-PLATE CAPACITOR.
Imagine a dielectric-insulated plane parallel capacitor with one plate
connected to ground and the other plate at a potential V. According
to Eq. 3-113, the stored energy is QV/2, or CV?/2. Now the energy
density is «E?/2, and, if each plate has an area S and if the dielectric
has a thickness s, the stored energy is
1 1
5 eE’Ss = € =(Es)? = : AY (3-120)
No!
as before.
 125

Joes hORCES FON DIELECTRICS

A dielectric material placed in an electric field is subjected to forces and

torques that arise from the interaction of the electric field with the dipoles
in the dielectric. Although these forces and torques are ordinarily small, they
can also be quite large. See for example Problem 3-19.
A dipole in a uniform electric field experiences a torque
Te Oe (3-121)
which tends to align it with the field, but the net force is zero. There can be a
net force only if the field is nonuniform, so that one end of the dipole is sub-
jected to a greater force than the other.
Let us calculate the force that a dipole experiences in a nonuniform elec-
tric field. We select axes as in Figure 3-14 and set the field intensity at the
origin to be E. Then the x-component of the force on the dipole is

Fee On 40 (E.pa Esk) (3-122)

ag ee
OE OE OE.
=a ae a
ys oz Ds: ( 3-123 )

Similarly,

pe
= OE, dE, dE,
(De )

dz
Then

B= (o02 +5, + Pej) Eat Bi + EW) (3-126)

Figure 3-14. Dipole with negative

charge at the origin.
 ELECTROSTATIC FIELDS II
126

F= (p:V)E. (3-127)

This is the force exerted on a single dipole.

or (P-V JE.
The force per unit volume of the dielectric is N times larger,
Since
P = («— @)E, (3-128)
the.force per unit volume is

GIrGKE VEG ie) (ao 4+E, 2te =) (Bir Ep Een

(3-129)
and its x-component is
2
dE,
a7 ‘|
(e = —
«)| BE + Eo

Now, in an electrostatic field, V X E = 0 and

Ea ee) (3-130)
oy Ox

a (3-131)
OZ Ox

so that the x-component is

«- 6) [BB et EE)=56- ORE G13

and, finally, the force per unit volume is

aC =) VP=2 : ly ( E) (3-133)

inside a homogeneous dielectric. Note that the quantity between parentheses

on the right is the electric energy density.
This is the force per unit volume exerted on a dielectric inside which the
electric field intensity is E. It is clear that the force is in the direction of an
increase in the magnitude of E and that it is unaffected by a change in the
polarity of the field.
This expression is usually valid not only in electrostatic fields but also
in alternating electric fields for the following reason. We shall see in Section
8.1.1 that, in the general case, the curl of E is equal to 0B/dt where B is the
magnetic induction. For example, the x-component of the curl gives
3.12 FORCES ON DIELECTRICS
127

OF, OF, _ OB,

Ons Oye roy CARD
We may therefore set V X E = 0 whenever

OE, OF,
e ~ |e >|OB,
ar (3-135)

which is usually the case when we have to deal with dielectrics.

The x-component of the force on the dipole is zero if all three terms on
the right of Eq. 3-123 are zero, for example, if p is parallel to the x-axis
(Py = pz = 0) and if E, does not vary with x (aE,/dx = 0). This is intuitively
quite obvious, and, of course, the same applies to the two other components
F, and F,. It is a simple matter to discover other conditions where F, is zero.
Forces on dielectrics can also be calculated using the method of virtual
work (Section 2.16).

Example FORCE PER UNIT VOLUME ON THE INSULATING

MATERIAL IN A COAXIAL CABLE

A wire of radius R; is insulated with a dielectric of outer radius R»

that is itself enclosed in a grounded conducting sheath. We wish to
calculate the force per unit volume exerted on the dielectric.
We require the electric field intensity E inside the dielectric in
terms of the applied voltage V and of the radius p. Let the charge
per unit length on the wire be \. From Gauss’s law, E is \/27rep, the
voltage between the inner and outer conductors is

a Ere Ew dp = a eee
In R. (3-136)

and
r V
(3-137)
i= 27rep 2 p In (R2/Ri)

Thus, according to Eq. 3-133, the force per unit volume is

1 Ve 1 2 ale — DY? (3-138)

5 col€r a 1) In? (R2/Ri) Vv p
~ In? (R2/ Ri) p?
The force is radial, and it is directed inward. The reason for this is
as follows. Let us assume that the voltage V on the wire is positive.
The dipoles then align themselves radically with their negative ends
facing the wire, and the inward force on the negative charges is
slightly larger than the outward force on the positive charges.
128 ELECTROSTATIC FIELDS II

You will be able to show in Problem 3-19 that this force can be
larger than the gravitational force by a few orders of magnitude.

3.13 FORCES ON CONDUCTORS

IN| THE, PRESENCE OF DIELECTRICS:

The calculation of the forces between conductors in the presence of dielectrics

is best done by the method of virtual work illustrated by the example in Sec-
tion 2.16.
When the conductors are immersed in a Jiquid dielectric, the forces are
always found to be smaller than those in air by the factor e, if the charges are
the same in both cases. They are larger than in air by the factor e, if the electric
fields (and hence the voltages) are the same.
It follows (see Section 2.16) that the energy density in a dielectric sub-
jected to an electric field E is e£?/2, as in Section 3.11.
The case of solid dielectrics will be illustrated in Problem 3-22.

Example FORCES ON A PARALLEL-PLATE CAPACITOR

IMMERSED IN A LIQUID DIELECTRIC

We found in the example in Section 2.16 that the force between the
plates of an air-insulated parallel-plate capacitor is (o?/2¢€)S, where
a is the surface charge density and S is the area of one plate. We can
perform a similar calculation for a pair of plates immersed in a liquid
dielectric. In this case, if the spacing s between the plates decreases
by ds, the work done against the mechanical forces F,, holding the
plates apart is again F,, ds and

F,, ds = ;E°S ds. (3-139)

Then

Fn Soy
= 5 BS =
= 6,©0
5B'S, (3-140)
is

and, for a given electric field strength E, or for a given voltage differ-
ence between the plates, the electric force is €, times /arger than in air.
We also have that
D: ae QO? 1 Q?
lal = eli — = -
2e 2 2e - 26S &, Zen (3-141)
and, for given charges +-Q and —Q on the plates, the electric force
1s ¢, times smaller than in air.
 129

3.14 SUMMARY

When a nonpolar dielectric material is placed in an electric field, the positive

and negative charges in the molecules are displaced, one with respect to the
other, and the molecules become polarized. The induced dipole moment per
unit volume P is called the electric polarization.
This produces real accumulations of charge that we can use to calculate
V and E both inside and outside the dielectric:
oO, = Pen, (3-14)
py = —V-P, (3-18)
where a is the surface density and p, is the volume density of bound charge.
The unit vector n is normal to the surface of the dielectric and points outward.
The polarization current density 1s

Jo = oeamperes/meter’?. (3-21)

The Jocal electric field intensity Eo. is the space-and-time average of the
electric field intensity acting on a particular molecule:

Eyo & E+ hos (3-32)

3€0

This result is approximate and, in general,

P
Ew. = E+ 6 = (3-33)
0

where b is a constant that depends on the nature of the dielectric.

In linear and isotropic dielectrics the dipole moment per molecule p
is proportional to the local field, the factor of proportionality a being the
molecular polarizability,
P = Evo; (3-35)
and the electric polarization P is then proportional to E:
P = ox. #, (3-38)

where x, is a dimensionless constant called the electric susceptibility.

cs:
Gauss’s law can be expressed in a form which is valid for dielectri

Vika (3-41)
130 ELECTROSTATIC FIELDS II

where the total charge density p, is the free charge density p, plus the bound
s of
charge density py. This is one of Maxwell’s four fundamental equation
electromagnetism.
The electric displacement

D = wE + P, (3-45)
= e(1+ x.)E = eek = cb, (3-52)

where ¢, is the relative permittivity, or the dielectric constant, and is the

permittivity of the medium. The vector D is related to the free charge density

a al LP (3-46)
or

[ o-da= |oar (3-47)

Poisson’s equation for Class A (linear, isotropic, homogeneous) dielec-
trics is
VV (3-55)
In a Class A dielectric, the bound charge density is related to the free
charge density as follows:
—po= Py. (3-59)

At a dielectric-conductor boundary the free surface charge density o; on

the conductor is equal to the electric displacement just inside the dielectric,
and the bound surface charge density o, is smaller in magnitude by the factor
(ce, — 1)/e,:
e, — l é- — l
== €,
D= €;
oh (3-66)
The relative permittivity ¢, is related to the mass density p and to the
molecular polarizability a through the Clausius-Mossotti equation:
Ceara a No
Eee nae (3-85)
and
-_ Me, a i -_ Na
OM hee) = Se Q, (3-88)

where ay is the molar polarization.

For polar molecules we have the Langevin equation:

P=N; P E kT
(cothoes
kT ‘ 7) : 3
(3-98)
PROBLEMS
131

where p is the permanent dipole moment of the molecule; and for real
polar
dielectrics, in which there are both induced and permanent dipoles, we have
the Debye equation:

mn M eo 1 = Na Pp

oe (ea) = 3 (2+): CR
In the presence of dielectrics the potential energy of a charge distribution
is given by an integral similar to that for free space:

1
W= 5 | War (3-113)

l
= [on dr. (3-117)

And we can define an energy density

dw FG1
7 (D-E), (3-118)

which is equal to (1/2)eE? for Class A dielectrics.

The torque on a dipole is

T=pXE, (3-121)
while the force is
F =(p-V)E. 127)
The force per unit volume is (P-V)E, or (4)(€ — €)VE°.
Charged conductors are subjected to electric forces that depend on the
nature of the dielectric in which they are immersed. In a liquid dielectric the
dielectric decreases the forces by a factor of ¢, if the charges are kept constant,
and it increases the forces by a factor of e, if the voltages are kept constant.

PROBLEMS
3-1. A sample of diamond has a density of 3.5 grams/centimeter® and a polariza-
tion of 10-7 coulomb/meter?.
(a) Compute the average dipole moment per atom.
(b) Find the average separation between centers of positive and negative
charge. Carbon has a nucleus with a charge +6e, surrounded by 6 electrons.

3-2. Show that the bound charge density at the interface between two dielectrics
1 and 2 that is crossed by an electric field is (P; — P2)-n. The polarization in 1
is P, and is directed into the interface, while the polarization in 2 is P, and
 ELECTROSTATIC FIELDS II
132

points away from the interface. The unit vector n is normal to the interface
and points in the direction from 1 to 2.

3-3. A large block of dielectric contains small cavities of various shapes that may
be assumed not to disturb appreciably the polarization.
(a) Show that, inside a needle-like cavity parallel to P, E is the same as in the
dielectric. ;
(b) Show that, inside a thin crack perpendicular to P, E is e, times larger than
in the dielectric. .
(c) Show that, at the center of a small spherical cavity, E is P/3¢0.

. The space between the plates of a parallel-plate capacitor with plate separation
s and surface area 5S is partially filled with a dielectric plate of area S and of
thickness t < s.
Show that the capacitance is
€oS
s — [(e, — 1)t/e,]
3-5. A dielectric sphere of radius R contains a uniform density of free charge py.
Show that the potential at the center is
26.+1 p,R*
2€, 3€,
3-6. Draw graphs of D, E, V as functions of r for a point charge at the center of a
dielectric sphere as in the second example on page 111. Set Q = 1.00 X 1079
coulomb, R = 2.00 centimeters, «¢, = 3.00.
Show also the curves of D, E, V in the absence of the dielectric sphere.

. A dielectric sphere of radius R is polarized so that P = (K/r)n, ri being the

unit radial vector.

(a) Calculate the volume and the surface density of bound charge.
(b) Calculate the volume density of free charge.
(c) Calculate the potential inside and outside the sphere.
(d) Sketch a curve of potential versus distance from r = 0 tor =~.
3-8. A conducting wire carrying a charge \ per unit length is embedded along the
axis of a circular cylinder of dielectric. The radius of the wire is a; the radius
of the cylinder is b.
(a) Show that the bound charge on the outer surface of the dielectric is equal
to the bound charge on the inner surface, except for sign.
(b) Show that the net charge along the axis is \/e, per unit length.
(c) Show that the volume density of bound charge is zero in the dielectric.

3-9, The relative permittivity of the dielectric between the plates of a parallel-plate
capacitor varies linearly from one plate to the other. If €,, and €,) are the val-
ues at the two plates, where €,) > €,1, and if the plate separation is s, show that
the capacitance per unit area is

€0 (€2 — €,1)
Sain (€,2/ €r1)
 133

Back plate
Diaphragm
Metallized surface

—. Figure 3-15,

3-10. (a) If the space between two long, coaxial cylindrical conductors were filled
with a dielectric, how would the relative permittivity have to depend on the
distance p from the axis in order that the electric field intensity be independent
of p?
(b) What would be the volume density of bound charge?

. A capacitor is formed of two concentric spherical conducting shells of radii

r,; and r3. The space between the shells is filled from r, to r2 with a dielectric
of relative permittivity ¢,1, and from rz to rz; with a dielectric of relative per-
mittivity €,.
Show that the capacitance is

“Eeie-d
Are
1 1 ib ff i ry
énl1 €r73 r2 \Ero €r1

3-12. An electret has the form of a thin circular sheet of radius R and thickness f¢,
permanently polarized in the direction parallel to its axis. The polarization P
is uniform throughout the volume of the disk.
Calculate E and D on the axis, both inside and outside the disk.

. Figure 3-15 shows a microphone in which an electret film is used as dia-

phragm. The right-hand side of the electret is metallized and serves as one
electrode, the other electrode being the back plate.
In actual practice the surface of the back plate is slightly roughened and
the diaphragm is placed in physical contact with it. The diaphragm then
touches the back plate at a large number of small isolated spots. The back
plate is perforated in order to reduce the stiffness of the air cushion and thus
increase the sensitivity.
134 ELECTROSTATIC FIELDS II

The surfaces of the film carry both free and bound charges, the free
charges being predominant. The net charge density has a time constant that
can be as large as a few thousand years at room temperature.

(a) Calculate the induced charge densities on the electrodes, under steady-
state conditions, for the case of a 0.006 millimeter thick Mylar (e, = 3.0) film
and a 0.013 millimeter air gap. Assume constant free charge densities of
10-4 coulomb/meter? on the film surfaces and neglect permanent bound
charges. Neglect edge effects.
(b) Find a differential equation for c; when the diaphragm oscillates at an
angular frequency w.

. When a block of insulating material such as Lucite is bombarded with high-

energy electrons, the electrons penetrate into the material and remain trapped
inside. In one particular instance a 0.1 microampere beam bombarded an
area of 25 centimeters? of Lucite (e, = 3.2) for 1 second, and essentially all
the electrons were trapped about 6 millimeters below the surface in a region
about 2 millimeters thick. The block was 12 millimeters thick.
In the following calculations neglect edge effects and assume a uniform
density for the trapped electrons. Assume also that both faces of the Lucite
are in contact with grounded conducting plates.
(a) What is the bound charge density in the charged region?
(b) What is the bound charge density at the surface of the Lucite?
(c) Sketch graphs of D, E, V as functions of position inside the dielectric.
(d) Show that the potential at the center of the sheet of charge is about 4
kilovolts.
(e) What is the electric field intensity in the charge-free region?
(f) What is the energy stored in the block? What is the danger that the block
will explode?
(g) How would the curve of V be affected if the sheet of electrons were closer
to one face of the block than to the other?

. The dipole moment of the H,O molecule is 6.2 X 10-%° coulomb-meter. Find
the maximum electric polarization of water vapor at a temperature of 100°C
and at a pressure of 760 millimeters of mercury.

3-16. (a) Show that a nonhomogeneous dielectric can have a volume density of
bound charge in the absence of a free charge density.
(b) Calculate py in this case.
3-17, (a) Show that the maximum energy stored per cubic meter in a parallel-p
late
capacitor is ea?/2, where a is the dielectric strength of the insulator (maximu
m
electric field intensity before breakdown),
(b) It is suggested that a small vehicle could be propelled by an
electric motor
fed by charged capacitors. Comment on this suggestion, assuming
that the
only problem is one of energy storage. A good dielectric to
use would be
Mylar which has a dielectric strength of 4000 volts per mil
(0.001 inch) and
a relative permittivity of 3.2.
PROBLEMS 135

3-18. A dipole of moment p is lined up with the z-axis at the origin of coordinates.
A second dipole of moment p is centered at the point (a, 0, a) and is pointed
toward the origin. Calculate the force on the second dipole.
3-19. (a) Calculate the force per cubic meter on the dielectric of a coaxial cable
whose inner conductor has a radius of 1 millimeter and whose outer conductor
has an inner radius of 5 millimeters. The dielectric has a relative permittivity
of 2.5. The outer conductor is gounded, and the inner conductor is maintained
at 25 kilovolts.
(b) Show that the electric force near the inner conductor is about 300 times
larger than the gravitational force if the dielectric has the density of water,
namely 10% kilograms/meter?.

3-20. In the presence of a nonuniform electric field, a particle of dielectric suspended

in a fluid having a lower relative permittivity is submitted to a force that pulls
it toward the region where the field is highest. This phenomenon is known as
dielectrophoresis.
Unlike electrophoresis, dielectrophoresis does not require charged par-
ticles, but requires a nonuniform field that can be either direct or alternating.
Dielectrophoresis is in general a weak effect that can be used only with
strong fields. It has been used to separate living cells from dead ones, carbon
particles from a polymer, various minerals and chemicals, etc.
Let us calculate the drift velocity of spherical particles of radius a and
relative permittivity p in a fluid of relative permittivity f < p, when the field
is cylindrical with inner and outer radii p; and pp» respectively.
First, the polarizing field inside the particles is smaller than the field E
in the fluid by the factor
3f
+p
as we shall see in the example on page 173. This factor is strictly valid only
for a uniform field, but it is applicable here as long as the particles subtend
a small angle at the axis of symmetry. Then the net force on the particle is the
force on the particle minus the force on the droplet of fluid that it replaces.
(a) You should be able to show that
Dae; 3 p= AT ew) Vipi
ieee cae 3ios Tae AVE 3 €oa A pin? (ps/p.) .

where Vy is the voltage difference between the inner and outer electrodes, and

A =
(Gig De PU = a
fay):
Chaar)
(b) Under what conditions is A positive? Sketch a curve of A as a function
of p, for a given value of f.Can you explain qualitatively the shape of this
curve? '
(c) The drift velocity » is given by Stokes’s equation:
Fret = Omnar,
where 7 is the coefficient of viscosity of the fluid.
136

— D

| a4 ‘ Figure 3-16.

The drift velocity v is therefore proportional to the square of the radius

a of the particles and inversely proportional to the cube of the distance p
from the axis.
Show that v © 3 X 10-4 meter/second for particles of a polymer with
p = 4.6 in benzene (f = 2.3, 7 = 6.5 X 10-4 newton second/meter’),
Vo = 4 X 10° volts, pp = 1 centimeter, p: = 0.5 millimeter, a = 0.1 milli-
meter, and p = 0.5 centimeter.

3-21. It is possible to accelerate a neutral molecule by means of an electric field in

the following way. Figure 3-16 shows a pair of spheres carrying charges +Q
and —Q, respectively, and a molecule whose dipole moment is p. It is clear
that the molecule is subject to a net force F. At the moment the molecule
reaches the midpoint between the spheres, the spheres are both grounded,
and the molecule continues at a constant velocity.
The purpose of such an accelerator is to study in detail the inelastic
processes which occur in molecular collisions.
(a) Show that the kinetic energy of a molecule accelerated in this way is
2pQ =
m€,D
if x > D initially.
Of course the dipole moment p increases as the molecule moves into the
electric field between the two spheres, and the kinetic energy is in fact larger.t
(b) In one particular accelerator the electrodes have a radius of 0.25 milli-
meter, D = 1.00 millimeter, and they are maintained at +40 kilovolts. Show
that the kinetic energy acquired by a molecule with p = 2 X 10-29 coulomb-
meter is approximately 0.01 electron-volt.{
(c) Since this energy is too low, the accelerator has 700 stages with an average
distance of 1.4 centimeters/stage, for a total length of 10 meters. The elec-
trodes are fed at a frequency of 500 kilohertz. Assuming that the total energy
is then 700 times larger, show that the equivalent temperature of the accel-
erated molecules is about 6 x 10 kelvins.

. A capacitor composed of two electrodes of area A separated by a sheet of di-

electric of thickness ¢ is connected to a source of voltage V.{
PROBLEMS 137

(a) Calculate the force of attraction between the plates.

(b) In one particular case S = 68 square inches, t = 0.030 inch, ¢, = 3.0,
V = 60 kilovolts. Calculate the force of attraction in metric tons (1000 kilo-
grams).
(c) The solid dielectric is removed and the plates are submerged in a liquid
of the same relative permittivity, so that the space between the plates is always
filled with dielectric. Calculate the force of attraction again in metric tons.

3-23. One author states that he can attain fields of 4 < 10° volts/centimeter over a
0.1 inch gap in purified nitrobenzene (¢, = 3.5) and that the resulting electric
force on the electrodes is then 36 pounds per square inch, or more than two
atmospheres. Is the force really that large?

3-24. Electrostatic clamps are used for holding work pieces while they are being
machined. They utilize an insulated conducting plate charged to several
thousand volts and covered with a thin insulating sheet. The work piece is
placed on the sheet and grounded.
One particular type operates at 3000 volts and is advertised as having a
holding power of 30 pounds per square inch. If the insulator is Mylar
(ce, = 3.2), what is its thickness?

. It is stated in Section 3.13 that the electric forces on conductors immersed in

liquid dielectrics are larger than in air by the factor e, if the voltages are the
same, with and without the dielectric, and that they are smaller than in air
by the same factor e, if the charges are the same. Can you justify this general
statement?

Hints

3-21. (a) Assume that +Q and —Q are point charges, and calculate the force exerted
on them by the molecule.
(b) You will have to calculate an approximate value for Q.

3-22. (a) Remember that, if the plate separation is increased by ds, the extra field is in air.
3-24. See hint for 3-22(a).
CHAPTER |

ELECTROSTATIC FIELDS Ill

General Methods for Solving Laplace's
and Poisson's Equations

Up to this point our discussion of electrostatic fields has been limited to rather
simple charge distributions; we shall now develop methods for calculating
more complex fields.
Except for Section 4.1, this chapter deals with the solution of the differ-
ential equations V?V= 0, V7V= — p,/e,, VE = Vp,/e,. The first is of course
Laplace’s equation, while the other two are Poisson’s equations for V and E.
Now these equations are valid not only for electrostatic fields but also for
several other classes of phenomena. For example, the flow of heat in a
medium of thermal conductivity K obeys Poisson’s equation V°T = —q/K,
where T is the temperature and q is the thermal energy generated per unit
volume and per unit time. This chapter therefore has broad applications.*
We shall first discuss the continuity of various quantities at the interface
between two different media, and then we shall prove the uniqueness theorem,
according to which there is only one physically possible electric field that can
satisfy both Poisson’s equation and a given set of boundary conditions. Then
we shall illustrate the method of images, and finally we shall discuss at con-
siderable length the solution of Laplace’s and Poisson’s equations in rec-
tangular and in spherical coordinates.
If you are interested in this type of calculation, you can work through
Appendix B, which describes a method that is widely used for calculating
two-dimensional electrostatic fields when the volume charge density p is zero.

ceSections 4.1 and 4.2 are fundamental, but the rest of this chapter can be omitted if
time is lacking; it is not required for what follows.
 139

Figure 4-1. Gaussian cylinder on the interface between two different

media 1 and 2. The difference D,, — Dj,» between the normal components
of D is equal to the surface charge density o;.

It should be realized that, once the electrostatic field is known, it be-

comes possible to deduce many other quantities—such as charge densities and
total charges on conductors, potential differences between conductors, capaci-
tances, focusing properties of electrostatic lenses, and so forth.

aie CONTINUITY OF VD. Ey AT THE INTERFACE

BETWEEN TWO DIFFERENT MEDIA

The solution of Poisson’s or Laplace’s equation must be consistent with

certain boundary conditions.

4.1.1. Potential

The potential must also be zero at infinity if the charge distributions are
of finite extent, and it must be constant throughout any conductor as long
as the electric charges are at rest.

4.1.2 Normal Component of the Electric Displacement

Consider a short Gaussian cylinder drawn about a boundary surface, as in
Figure 4-1. The end faces of the cylinder are parallel to the boundary and
140

Figure 4-2. Closed path of inte-

gration crossing the interface
between two different media 1
and 2. Whatever be the surface
charge density a; the tangential
components of E on either side
of the interface are equal:
Eu = Ev.

arbitrarily close to it. The boundary carries a free surface charge density oy.
If the area Sis small, D and o; do not vary significantly over it, and according
to Gauss’s law the flux of D emerging from the flat cylinder is equal to the
charge enclosed:

(Dr — Dro)S = oS. (4-1)

The only flux of D is through the end faces, since the area of the other surface
is arbitrarily small. Thus
Dns == Dye =) Op. (4-2)

At the boundary between two dielectric media the free surface charge
density o; is generally zero and then D, is continuous across the boundary.
On the other hand, if the boundary is between a conductor and a dielectric,
and if the electric field is constant, D = 0 in the conductor and D, = o; in
the dielectric, o; being the free charge density on the surface of the conductor.

4.1.3 Tangential Component

of the Electric Field Intensity
Consider the path shown in Figure 4-2, with two sides parallel to the
boundary and arbitrarily close to it. The other two sides are infinitesmal.
If the path is short enough, E, does not vary significantly over it and the line
integral of E-dl is E,,L— E,,L. Now, according to Stokes’s Theorem, this
line integral is equal to the integral of V x E over the surface enclosed by
the path. By definition, the enclosed area is zero. So, even when V x E is
not zero, its surface integral is zero and

Eyl = Ex = 0, (4-3)
or

Eu a Ey. (4-4)
4.1. CONTINUITY oF V, D,, E, AT INTERFACES
141

The tangential component of E is therefore continuous

across the boundary.
If the boundary lies between a dielectric and a conduct
or, then EK =
in the conductor and E; = 0 in both medialis ERNE Ane
eT

4.1.4 Bending of Lines of Force

It follows from the boundary conditions that the D and E vectors change
direction at the boundary between two dielectrics. In Figure 4-3,

D, cos 6; = D2 cos 6, (4-5)

or
€r1€0E1 COS 6, = €,2€9)Ey COS 62, (4-6)
and
Ey sin 01 -- EF, sin A. (4-7)

Then, dividing the second equation by the third,

€,1 COL 6, = €;2 COt Bo, (4-8)

or
tan A, _ &1

tan; é5 cc

The larger angle from the normal is in the medium with the larger relative
permittivity.
For the case of an electret, see Problem 4-2.
The electric field is therefore subjected to several boundary conditions
at an interface, but these conditions do not all have the same degree of gener-
ality. The condition that V be continuous is perfectly general. Equation 4-2

Figure 4-3. Lines of D or of E

crossing the interface between
two different media 1 and 2. The
lines change direction in such a
Way that €,, tan A, = €,) tan 61.
142 ELECTROSTATIC FIELDS III

is also general. However, the condition D, = a, at the interface between a

dielectric and a conductor is valid only for static fields, because only then is
D = Oin conductors. Similarly, the condition E, = Ey at any interface, and
the bending of lines of force, as above, are also valid only for static fields
because they are based on the hypothesis that V X E = 0, which is not
general, as we shall see in Chapter 8.

4.2 THE UNIQUENESS THEOREM

We shall demonstrate that a potential V that satisfies both Poisson’s equation

and the boundary conditions pertinent to a particular field is the only possible
potential.
This is an important theorem because it leaves us free to use any method,
even intuition, to determine an electric field; if we can somehow find a field
that satisfies both of the above conditions, then it is the only possible one.
Let us consider a finite region of space that may contain charged con-
ductors at specified potentials, Class A dielectric materials of specified prop-
erties, and volume distributions of free charges with specified densities. To
demonstrate the uniqueness theorem we shall assume that at each point there
are two possible solutions, V; and V2, both of which satisfy Poisson’s equation
and both of which reduce to the specified potentials on the surfaces of the
conductors. This does not imply that a given point can be at two different
potentials at the same time. Our assumption is that either one or the other of
two different fields can exist in the region for which boundary conditions are
specified. We shall find that V; = V2. This is the uniqueness theorem.
Corresponding to V; and V2, there are two possible electric field inten-
sities
Espen Wy, BE, = —VV; (4-10)
at every point in the field.
We have assumed that Poisson’s equation is satisfied by both V; and V,
everywhere. Then

V-Di = pj, V+ Dy = py (4-11)

where py; is the free charge density.

We now focus our attention on the difference between the two solutions
and call it V;:
V3 = Ve — Vy. (4-12)

Then the corresponding D,, D2, D3 are such that

4.2 THE UNIQUENESS THEOREM 143

D; = D, — D,, (4-13)
and
V-D; = V-D, —V-D, = 0 (4-14)
at every point. On the surfaces of the conductors V3 = 0 since both V; and
V, reduce to the specified boundary values.
We now use the vector identity

V-V;D; = VV-D3) + D3-VV3. (4-15)

Integrating over a volume + and using the divergence theorem,

| veda = / VV - Ds) dr + |DeVV) ar, (4-16)

where the surface integral on the left is evaluated over all the surfaces that
bound the volume r. Let us take this volume to be the volume external to the
conductors, extending to infinity in all directions.
The surface integral is then to be evaluated over the surfaces of the
conductors and over an imaginary sphere of infinite radius. Since the quantity
V; is zero on all these surfaces, this portion of the integral is zero. To evaluate
the integral over the sphere of infinite radius, we consider the integral over a
finite sphere and let its surface recede to infinity. Both V, and V; must fall off
as 1/r at sufficiently large distances, since all the charge in the system will
appear as a point charge from a distance large in comparison to the dimen-
sions of the charge system. Then V3, the difference between V, and V;, must
also fall as 1/r. Now D; must fall off as VV3, or as 1/r?. Since the area S over
which the integration is performed increases as 7’, the whole integral falls off
as 1/r and approaches zero at infinity. The left side of the equation is thus
zero.
The first term on the right is also zero, since V-D; = 0 at every point. We
are thus left with the second term on the right, which must be identically equal
to zero. Thus

i(D,-E;) dr = 0. (4-17)

In homogeneous, isotropic, linear dielectrics the quantity D- FE= ¢E? is posi-

tive, and the only way in which the integral can be zero is to have D,; and E;
equal to zero at every point.
It therefore follows that

VV, = VN, (4-18)

or that V, can differ from V, at most by a constant. Since V; and V2 must be
144 ELECTROSTATIC FIELDS III

the same on the surfaces of the conductors, they must be the same everywhere.
Therefore V, = Vi, and there is only one possible potential V.
We have therefore shown that the solution of the Poisson equation for
given boundary conditions is unique, as long as D-E is positive throughout
the dielectric material in the system.

4.3 IMAGES

The method of images involves the conversion of an electric field into another
equivalent field that is simpler to calculate. It is particularly useful for point
charges near conductors: it is possible in certain cases to replace the con-
ductors by one or more point charges in such a way that the conductor
surfaces are replaced by equipotential surfaces at the same potentials. Since
the boundary conditions are then conserved, the electric field thus found is
the correct one for the region outside the conductors.

Example POINT CHARGE NEAR AN INFINITE

GROUNDED CONDUCTING PLANE
As a first example, consider a point charge Q at a distance D from
an infinite conducting plane connected to ground, as in Figure 4-4a.
This plane may be taken to be at zero potential. It is clear that if we
remove the grounded conductor and replace it by a charge —Q ata
distance D behind the plane, then every point of the plane will be
equidistant from Q and from — Q and will thus be at zero potential.

10

(a) . (b)
Figure 4-4. (a) Point charge Q near a grounded conducting plane.
(b) The conducting plane has been replaced by the image charge —Q
to
calculate the field at P.
4.3 IMAGES
145

In the region to the left of the plane the two point charges
must there-
fore give the proper solution for the point and the plane.
The charge
— Q is said to be the image of the charge Q in the plane.
The potential V at a point P, whose coordinates are r and
0, as
in Figure 4-4b, is given by

4reV = 2 — £, (4-19)
where
r! = Vr? + 42? — 4rD cos 6.
(4-20)
The components of the electric field intensity at P are given by the
components of VV:
OV Q Q(r— 2D cos 6)
4reE, = —4reo SiC ya ? (4-21)

10V 20D sin 6

4recE, = — Aire = crimes g iat (4-22)

The lines of force and the equipotentials are shown in Figure 4-5.
The induced charge density o’ on the surface of the conducting
plane is readily found from the normal component of the electric
field intensity at the conductor, since

E, = 0'/€0. (4-23)

4
<r.XN
,
Faex

Figure 4-5. Lines of force (identified by arrows) and equipotentials for a

point charge near a grounded conducting plane. Equipotentials and lines
of force near the charge cannot be shown because they get too close
together. Equipotential surfaces are generated by rotating the figure
about the axis designated by the curved arrow. The image field to the
right of the conducting plane is indicated by broken lines.
146

Figure 4-6. The electric field intensity

E,, at the surface of the grounded
conducting plane is calculated from
the fields of Q and of its image
—Q. It is the vector sum of E, and
E;, and is normal to the surface.

In this particular case the surface charge density is negative, and the
electric field intensity points to the right at the surface of the con-
ducting plate.
From Figure 4-6, r = r’ at all points on the plane and

LE ACOs ck aun ees: (4-24)

4ne,r® es

,_ QD
CS a (4-25)

A charge Q induces on the conductor a charge

QD 2sds 2
Oe iSige ene 2 Jo (s? + D3? — :
(4-26)
as expected.
Itwill be observed thé
‘i In this
case we requireOne field in the region to the left of the Bites whereas
the image is to the right.
Now what is the electrostatic force of attraction between the
charge Q and the grounded plate? It is obviously the same as be-
tween two charges Q and —Q situated at a distance 2D one from
the other, since Q cannot tell whether it is in the presence of a point
charge —Q or of a grounded plate. The force on Q is
Sabet This is i te true. T)

Example POINT CHARGE NEAR A GROUNDED

CONDUCTING SPHERE
Another case where the image method is applicable is that of a point
charge near a grounded conducting sphere, as in Figure 4-7a. We
 (a) (b)
Figure 4-7. (a) Point charge Q at a distance D from the center of a
grounded conducting sphere of radius a. When Q is positive, the induced
surface charge density o’ is negative. (b) The boundary condition V = 0
on the spherical surface is satisfied by the original charge Q and its
image —Q’.

remove the conductor and try to find the position and magnitude of
an “image” charge Q’, as in Figure 4-7b, that will make the potential
zero on the spherical surface.
It is clear from the symmetry of the problem that if such a
charge exists it must lie on the line connecting Q and the center of
the sphere. We begin by making the potential zero at the points P;
and P:. Then
Q jaan.
(Dieaye bynes sy
Q
WET wi
oo
CED
:
(4-28)

Solving these two equations gives

g=-50,
peed, +4 = —-. (4-29)
-29

We still have to find whether this charge arrangement will make

the potential zero at a general point P; on the surface of the sphere.
At P3;

repr amiciery sega

ro
(4-30)
where
r = V(D? + a? + 2Da cos 6), (4-31)
ry = V(b? + a + 2ba cos 6). (4-32)
Taking Q/ and b as in Eq. 4-29 does in fact make V = 0 at P;, as
required.
Since the original point charge and the image charge Q’ satisfy
the boundary condition, they must give the correct field at every
point in the space outside the conducting sphere.
148

Figure 4-8. The electric potential V at the

point P(r, @) is calculated from Q and its
image —Q’.

We can now write down the potential at an arbitrary point

P(r, 6) as in Figure 4-8:

bee 16
Es
V4
(4-33)
where ave, Pees 2 £
= VD? + r+ 2Drcos 8, (4-34)

weel perecean (4-35)

Then, to find E, we merely have to calculate —VV.
As in the case of the point charge and the plane, we calculate
the induced charge density o’ from the value of the electric field
intensity on the surface of the sphere. We calculate E, and evaluate
itatr=a:

Ora D text) —a0(r+5 5088)

4regE, = a Dri (4-36)

and, at r = a,

pone pee +PA a? + 2Da

seed. cos (4-37)
\ as 2

4a (D? 6)?"
If we integrate this density over the surface of the sphere, we
get the total induced charge:

(sae ih" o!2na' sin 6.9 = —* Q. (4-38)

The total induced charge on the real conducting sphere is thus
the same as the image charge that replaced the sphere. This must be
true because of Gauss’s law: if we draw a Gaussian surface just
outside the sphere, then the flux of E through this surface, and hence
the enclosed charge, must be the same, no matter whether the con-
ducting sphere is present or whether it is replaced by the image
 149

Figure 4-9. Lines of force (identified by arrows) and equipotentials for

a point charge near a conducting sphere. Equipotential surfaces are
generated by rotating the figure about the axis identified by the curved
arrow. Equipotentials and lines of force in the vicinity of the point charge
are again not shown because they get too close together.

charge, since the fields outside the sphere are identical in either case.
Figure 4-9 shows lines of force and equipotentials for a point
charge in the vicinity of a grounded conducting sphere.

Example POINT CHARGE NEAR A

CHARGED CONDUCTING SPHERE
If the sphere is at a potential other than zero, we may still determine
the field by the method of images. We first replace the conducting
sphere with an image charge Q’, as we did for a grounded sphere.
This makes the surface occupied by the sphere an equipotential.
We next add a second image charge at the center to raise the spher-
ical surface to the required potential.
If we are given a sphere of radius a with a charge Q, on it, and
if its center is at a distance D > a from a point charge Q, we can
replace the sphere by an image charge Q’ = —(a/D)Q at a distance
b = a*/D from the center, plus a charge (Q, — Q’) at the center.
The surface density of charge is then a’ + o’’, where o’ is the non-
150

uniform distribution calculated from QO and Q’, and where o”’ is the
uniform distribution calculated from (Q,; — Q’).

Example CHARGED SPHERE NEAR A

GROUNDED CONDUCTING PLANE
Some fields may be calculated by the method of images through suc-
cessive approximations. As an illustration we shall calculate the
capacitance of a charged sphere near a grounded conducting plane.
We shall replace both the sphere and the plane by a set of point
charges that will maintain these surfaces as equipotentials.
First, we put a charge Q,; at the center of the sphere, as in
Figure 4-10b. This makes the sphere, but not the plane, an equi-
potential. Next we put the image — Q; of Q; to the right of the plane.
This makes the plane an equipotential but destroys the spherical
equipotential, so we put the image Q» of —Q, inside the sphere.
This makes the sphere again an equipotential but upsets the plane.
We continue the process, which converges rapidly, until we have the

(a) = (b)
Figure 4-10. (a) Conducting sphere carrying a charge Q near a grounded
conducting plane. When Q is positive, the induced surface charge
density ao’ is negative. The plane is assumed to be infinite. (b) The field
outside the sphere and to the left of the conducting plane is calculated
by successive approximations by using the image charges Q1, Qo,...,
etc., and —Qi, —Qy,..., etc.
4.3. IMAGES 151

required precision. The charges and their locations are shown

in
Table 4-1.
If we set (a/2D) = r, then
Or = QO,

OQ» = rQ,,
r2

QO; == (hae Or,

r3
Os = aS aye, ere an (4-39)
(1 =a ae i=)
a

Qs= —____—_—__" r3

ig,
Ce) (1- —) E Bis rT
~ al
and so on. The equipotentials for D = 3a are shown in Figure 4-11.
The total charge on the sphere is

Q-O(1+r+ 254+...) (4-40)

but only Q; contributes to its potential: the charges —Q, and Q,
make the potential of the sphere zero, and the same is true of all the
following pairs of charges. The potential of the sphere is therefore

= a
4treoa
’
(4-41)
and the capacitance between the sphere and the plane is
Ct DRONES Toit)
4rea(1 +r-+---). (4-42)
~V O1/4reoa

Table 4-1. Images for the Case of a Charged Sphere Near

a Grounded Conducting Plane

Left of Plane Right of Plane

Distance Distance
from Center from Center
Charge of Sphere Charge of Sphere

Or 0

OQ» = (a/2D)Q1 a*/2D —Q; | 2D

a a a’ ~Q, |2D — (a?/2D)

Cre @\ 2D 2" |2D — (a2/2D)
2D — —
2D

a a/2Dy OQ aA ep

we
Sa) 1 — (a/2D)*
a NEE EEE EEE
152

Figure 4-11. Equipotentials for a charged sphere near a

grounded conducting plane.

The presence of the plane increases the capacitance of the sphere,

as must be expected.
Figure 4-12 shows the ratio C/a as a function of the ratio r.
For r = 0 the sphere is infinitely distant from the grounded plate,
and C is simply the capacitance of an isolated sphere, 47re9a farad.
For r— 1/2, the sphere comes infinitely close to the conducting
plane, and the capacitance tends to infinity.

2.0

LES

1.0
0 0.10 0.20 0.30 0.40 0.50
-

Figure 4-12. The ratio C/a as a function of r.

4.3. IMAGES 153

Example CHARGE NEAR A SEMI-INFINITE

DIELECTRIC
The method of images can also be used to calculate fields that involve
dielectrics. As an example, consider a point charge Q at a distance
D from a semi-infinite block of Class A dielectric, as in Figure 4-13.
The electric field of the charge Q polarizes the dielectric, and a sur-
face charge density o, appears on the surface. The resultant electric
field at any point can then be computed by replacing the dielectric
with this surface charge and treating the problem as if these charges,
together with Q, were in free space. The value of o, is, of course, a
function of the distance s.
We have two different fields, that due to Q and that-due to oa».
Let us consider only the normal components. That part of the nor-
mal component of the electric field intensity at the dielectric surface
arising from Q is % “
E = Ecos? Arey (8? + D232 (7 i

This field is the same, both in magnitude and in direction, just inside
and just outside the dielectric surface, as in Figure 4-13.
From Gauss’s law, the normal component of the field of o, is
a,/2€. This field is directed away from the boundary if a, is positive
(QO negative), and into the boundary if a» is negative (Q positive).
Therefore, just inside the dielectric surface, the normal com-
ponent of the electric field intensity is

Eqs = QO D

~ Arey (s? + D8?

aa Ob

eo (4-43)
This quantity is positive if E,,; points outward.

Figure 4-13. Point charge Q near the

plane surface of a large block of
dielectric. The quantity E,,, is the
normal component of E due to the
induced surface charge density op;
Eon is the normal component of E
due to the charge Q. So as to
simplify our calculation, the arrows
for Eo, and for E,,, are oriented on
the assumption that Q and o» are
positive. We find, as expected, that
if Q is positive o, is negative and
that the fields H,,, point into the
boundary.
154 ELECTROSTATIC FIELDS III

Now, from Eqs. 3-8 and 3-51,

a =Pn=P, = éo(ér — 1)Ens, (4-44)

where the unit vector n is normal to the surface and points outward,
and
Q D ob
Oh = = Eq Grae.) Fe (s? + D3? a 5° | (4-45)

ie = han (4-46)
Oe 2(er Ge
The induced charge density o, and Q have opposite signs, as
expected.
At this stage we could calculate the electric potential and field
intensity at any point, either within the dielectric or in free space,
by using Coulomb’s law, integrating over the o, distribution, and
adding the contribution from the point charge Q. But this is not the
simplest way to deal with this field. We shall find instead a set of
image charges that will satisfy the boundary conditions.
To find these charges, we confine our attention to the boundary
and write down the normal components of the resultant electric
field intensity for a point just inside the dielectric, E,;, and just out-
side, En.. From Eqs. 4-43 and 4-46,

Ee =|i - (2=:ti == + Pye sett

ey ee QD
(4-48)
SS G =f » Arreo(s? + D®)3/2
For the outside point,

Eno

=
= es a ee
2€, OD
- € + , 4reo(s? + D3? (4-50)
It will be observed that the normal component of D is con-
tinuous across the boundary:

€rEing = Eno. (4-51)

This is because there is zero free charge density at the surface of
the dielectric.
Can we now find a set of image charges that will give these nor-
mal field components? We recall that an image charge is always
outside the region in which the field is to be determined, and that
both E,; and E,. point outward from the dielectric. We consider
first a point just outside the dielectric. We can see from Eq. 4-49
that E,. is the same as if we replaced the dielectric by the image
charge

= ee
E>,
_(e mz 1)
(4-52)
 Figure 4-14. (a) If the dielectric is replaced by the image charge
—(e, — 1)Q/(e, + 1), the field is unaffected outside the dielectric. (b) If
the dielectric is extended to both sides and 2e,Q/(e, + 1) is substituted for
Q, the field is unaffected inside the dielectric.

located at a distance D behind the boundary, as in Figure 4-14a.

For a point inside the dielectric, there are two sets of point
charges that will give the proper normal component of E: (a) the
point charge Q together with an image charge

(= 1)
OS =e Q (4-53)
(e, + 1)
at the image position inside the dielectric and (b) a single charge

D\c Eng ~ Oe
Oo, Oa— —- é-+1 Q, (4-54)

that replaces Q, the dielectric extending in this case on both sides

of the boundary. The first of these must be ruled out because we al-
ways want the image charge to be outside the region in which the
field is required. The image charge must therefore be located as in
Figure 4-14b.
What we have done, then, is the following. In order to find the
field outside the dielectric we replaced the dielectric by an image
charge Q’ located at a distance D behind the boundary, as in Figure
4-14a. In order to find the field inside the dielectric, we replaced Q
by a single charge Q” at the position of Q, and we extended the
dielectric to both sides of the boundary, as in Figure 4-14b. Since
these combinations of charges satisfy the boundary conditions, we
know from the uniqueness theorem that they provide the correct
field. The shape of the field is shown in Figure 4-15.
In general, for two media having relative permittivities €,, and
€,», With the point charge Q in medium 1, the point charges Q’ and
Q” are the following. The field in medium 1 is the same as if we had
156

Figure 4-15. Lines of D (identified by arrows) and equipotentials for a

point charge near a dielectric. As previously, equipotential surfaces are
generated by rotating the figure about the axis indicated by the curved
arrow. Equipotentials and lines of D near the point charge are not
shown.

the same medium 1 on both sides of the interface, the original charge
Q, and a charge
;
eae en
en a Erg
4-55
2 Sa eas g (

at the image position. The field in medium 2 is the same as if medium

2 extended on both sides of the interface and the original charge Q
were replaced by
2€,.
Oe et (4-56)
En + € ro

4.4 SOLUTION OF LAPLACE’S EQUATION

IN RECTANGULAR COORDINATES

The methods that we have considered until now for the calculation of electric
fields are useful only in special cases. We must find more general methods
for solving Poisson’s equation,
4.4 SOLUTION OF LAPLACE’S EQUATION IN RECTANGULAR COORDINATES 157

V2 Vacate: Ge (4-57)
To begin with, we shall confine our attention to electric fields with zero
space charge density, and we shall deal with Laplace’s equation,
(4-58)
functions have a number of general properties, of which we shall use the
following one. If the functions V;, V2, V3, --- are solutions, then any linear
combination 4,V; + A,.V.+ A3;V3; + --- of these functions, where the A’s
are arbitrary constants, is also a solution. This can be readily demonstrated
by substitution into the original equation.
It is usually possible to find solutions of Laplace’s equation that will
satisfy required boundary conditions by separating the variables. For ex-
ample, in Cartesian coordinates, we can usually find a solution of the form
V = X@YO)ZO), (4-59)
where X(x), Y(), Z(z) are respectively functions only of x, y, z. We can then
fit boundary conditions by adding a series of such solutions multiplied by
suitable coefficients. The uniqueness theorem assures us that the solution
found in this way is correct.
We can find the form of the functions X(x), Y(yv), Z(z) by substituting
V = X(x)Y(y)Z(z) into Laplace’s equation. Then

(4-60)

where we have written total instead of partial derivatives, since each one of
the X, Y, Z functions is a function of a single variable. Dividing through by
XeZe

Now since the second and third terms are independent of x, and since the
three terms must add to zero at all points, the first term must also be independ-
ent of x. It is therefore constant in value and

Similarly,

ye ieee een: t
158 ELECTROSTATIC FIELDS III

The problem then is to solve three ordinary differential equations, subject to

this condition and to the boundary conditions.

Example FIELD BETWEEN TWO GROUNDED SEMI-INFINITE

PARALLEL ELECTRODES TERMINATED
BY A PLANE ELECTRODE, AT A POTENTIAL’ #,
Figure 4-16 shows two grounded, semi-infinite, parallel electrodes
separated by a distance b. At x = 0 an electrode is maintained at a
potential Vo. The problem is to find the potential V at any point
between the plates.
By hypothesis, the potential is independent of z, and the con-
stant C3 is zero. We must therefore solve two ordinary differential
equations:
ax
ae = ReX = 0, (4-66)

ay,
qn (4-67)
We have substituted k2 for C; and —k? for GC. so as to eliminate
square roots in the solution. The choice between C; and C, as the
negative constant is immaterial; the boundary conditions will force
us to the same final solution in either case.
We solve Eq. 4-67 by setting
Y = Asinky + Bcos ky, (4-68)
where 4 and B are arbitrary constants. This can be easily verified
by substitution. Our value of V must satisfy the boundary conditions
vV=0 (y = 0,» = d), (4-69)

Figure 4-16. Grounded, plane-

parallel electrodes terminated by
a plane electrode at potential Vy.
The electrodes are assumed to be
infinite in the direction perpendicu-
lar to the paper and are assumed
to extend infinitely on the right.
4.4 SOLUTION OF LAPLACE’S EQUATION IN RECTANGULAR COORDINATES 159

ag (r= 50); (4-70)

V-0 (x > 00). (4-71)
In order to have V = 0 at y = 0 we must have B = 0; and in order
to have V = 0 at y = 5 we must have

kb = nt (n= 1,2, ---). (4-72)

Thus

Y = Asin ae (n = 1,2, +++). (4-73)

The value n = 0 must be omitted, for it corresponds to zero field.
The X equation is now

ax ni \?

and
X = Genrz/b + He-nrz/b, (4-75)
where G and H are arbitrary constants. We can again verify this
solution by substitution. The condition that V->0 as x > re-
quires that G = 0.
Altogether then,

Cay) i. Gin = e-nrz/b, (4-76)

where C is another arbitrary constant.

The solution as it is will obviously satisfy the boundary condi-
tions stated in Eqs. 4-69 and 4-71. It will not, however, satisfy Eq.
4-70. We therefore use an infinite sum of such solutions and set

Vix, y) = >> Casi a en naz/b, (4-77)

To evaluate the coefficients C,, we use the boundary condition

at x = 0, namely

V(0, y) = ay Coin
a (4-78)
n=1

The expression on the right is called a Fourier series. It can be shown

that, provided an infinite series of cosine terms is also included, it
constitutes a complete set of functions. This means that an arbitrary
boundary condition can be satisfied with such an infinite series.
Using a technique devised by Fourier, we multiply both sides
by sin [(pry)/b], where p is an integer, and integrate from y = 0
Ona

y
iE Vy sin ans dy -[ 2BC,, sin sin oe dy. (4-79)
160

V/Vo

0 0.1 0.2 OS 04 0.5 0.6 0.7 0.8 0.9 1.0

y/b

Figure 4-17. The condition V = V, as satisfied by a Fourier series taking

(a) only the first term, (b) the first 3 terms, (c) the first 10 terms, and (d)
the first 100 terms. The Fourier series provides an increasingly better
approximation as the number of terms is increased.

On the left-hand side,

2bVo Peer
: — f
ie Vo sin Se dy = pr ab aoc. (4-80)
: 0 if p is even,
whereas, on the right-hand side,

nt tg . sas
ieCG. ob 5 dy = jo. oe ae (4-81)

Thus the only term of the infinite series on the right-hand side
of Eq. 4-79 that differs from zero is the one for which n = p. A se-
quence of functions possessing this property is said to be orthogonal.
Combining Eqs. 4-80 and 4-81,

=>
[see
Bt AT
it wis odd, (4-82)
if n is even.
We can now write down the potential V at any point (x, y, z):

Vo
AV, ~. I ine eh
V(x, y, Zz) = (4-83)
T noe>Sess b
The successive terms in the series become progressively less impor-
tant, both because of the factor (1/n) and because of the exponential
function. The degree of approximation that is achieved at x = 0
with one, three, ten, and one hundred terms of the series is indicated
in Figure 4-17. At y = 6 the first term alone gives a good approxi-
mation. The equipotentials are shown in Figure 4-18.
 161

Figure 4-18. Three-dimensional plot of the potential V for the configu-

ration of Figure 4-17. The U-shaped curves are equipotentials; the
others show the intersections of the potential surface with planes parallel
to the xV-plane.

Example FIELD BETWEEN TWO GROUNDED PARALLEL

ELECTRODES TERMINATED ON TWO OPPOSITE
SIDES BY PLATES AT POTENTIALS V,; AND J,
In Figure 4-19 two grounded plane parallel electrodes of width a
are separated by a distance b and extend to infinity in the other
direction. At x = 0a conducting surface is maintained at a potential
V,, and the plane at x = ais occupied by a conductor maintained at
a potential V2. The problem is again to find the electric potential V
at any point between the plates.
The field is again independent of z, and the constant C; is zero.
Since the Y part of the solution is identical with that of the previous
example, Eq. 4-73 is valid, as are Eqs. 4-74 and 4-75.
From this point on, the solution differs from that of the previous
example, since the boundary conditions are different. Here we have
ana at 58 ==! O), (4-84)
V= V2 at = a (4-85)
The most general solution, and the one required to satisfy the bound-
ary conditions, is

V(x, ¥, 2) = >, (Ane! + Byer?!) sin ae (4-86)

n=1
162

Figure 4-19. Grounded plane-

parallel electrodes terminated on
two sides with plane electrodes at
potentials V; and V2. The electrodes
are assumed to be infinite in the
direction perpendicular to the paper.

where 4, and B, are again constants that must be determined from

the boundary conditions.
A trce—a) 3
“ . mTy
i 1 ya +
An + Ba) sin —- B,)sin =
(4-87)

The coefficients are evaluated by the same Fourier method used in

the previous example. On multiplying by sin 2 and integrating

from y = 0 to y = 5b, we have again, out of the whole infinite series,
only one term corresponding to p = n:

Vr [sin
0
™b dy = (Ay + Br) e,
2
(4-88)
and
2 Vi
Piaey | if is odd,
(4-89)
0 if n is even.
We can find another relationship between A, and B, from the
boundary condition at x = a: from Eq. 4-86,

V. ——= »»
.
(deel?
—ntu
++ Byeot=l)
na
sin+. ANY
=. (4-90)

Multiplying by sin a and integrating from y = 0 to y = 4, as

before,
4V, pie
Ane nralb -f- Berto = \ if nis odd, (4-91)

0 if n is even,
and, from Egs. 4-89 and 4-91,
4 (Vy — Voeral
A, = nw :Sad a ) (4-92)

g, = 2 (fake ) (4-93)
4e-nta/b V2 — Vie nralb
 163

Figure 4-20. A three-dimensional plot of the potential V for the configur-

ation of Figure 4-18 with V; = V2 = V,

where n = 1, 3,5, ---. The potential V at any point (x, y, z) is given

by Eq. 4-86 with A, and B,, as above.
Figure 4-20 shows the equipotentials for V; = V2 = Vo.

4.5 SOLUTION OF LAPLACE’S EQUATION

IN SPHERICAL COORDINATES. LEGENDRE’S
EQUATION. LEGENDRE POLYNOMIALS

Although electrostatic fields can usually be calculated in Cartesian coordi-

nates, certain cases are best treated in spherical polar coordinate
_ equation then takes the form

equation al KNOWN a pherical harmonic JUuncLIONS.

e shall restrict ourselves here to fields with axial symmetry, that is, to
fields where V is independent of the angle ¢. Then
164 ELECTROSTATIC FIELDS III

anol 1 ai vy 2 ri
ar (" ean Bg et cae)
As in Cartesian coordinates, we seek solutions in which the variables are
separated and set
Vir, 0) = R(r) 0), (4-96)
where R is a function of r only and © is a function of @ only. Substituting
V = RO into Eq. 4-95,
a (aR Ra zy 7 :
oe (" =) "sie a8 (sina) = oot)
and dividing through by RO,
1d (dR 1 d dO
Rae a) Pea ee Cy
We have written total instead of partial derivatives, since R and © are each
functions of a single variable.
Since the second term is independent of r, the first term must also be
independent of r. The first term must therefore be constant:
ied (ooh as
ale ele oe
and then
Le ep ade ;
© sin 6 dé (sing a) = = arn
since the sum of the two constants must equal zero.
Let us examine the R equation first. Multiplying both sides by R and
differentiating the term between parentheses, we obtain

? os tty kR = 0. (4-101)
The solution of this equation is of the form

R = Ar” +
B
(4-102)
prt

On substituting we find that

nin + 1) =k. (4-103)

Let us now examine Eq. 4-100 for 0:

d
WB (sino
6 a)
—4 + nin + 1)sin@®= 0 (4-104)
4.5 |SOLUTION OF LAPLACE’S EQUATION IN SPHERICAL COORDINATES 165

It is convenient to change variables in this equation and to let

b= COS 0. (4-105)
Now, for any function f(y),

df_ fie
dear ik ae _ae
5 = -(1—- eae off (4-106)

and our equation for © becomes

5d ja =) d=| +n(n + 10 = 0. (4-107)

This is Legendre’s equation. When nis an integer, its solutions are pelynomials
in cos @ and are known as Legendre polynomials. They are designated by P,,(u)
or P,(cos @):
@ = P(u) = P-(cos 6), (4-108)
where n is called the degree of the polynomial. There is a different polynomial
for each value of n.
Before proceeding to find solutions of Eq. 4-107, we must point out an
important property of Legendre’s equation. The number 1 must satisfy Eq.
4-103, but
n= —(n+ 1) (4-109)
will equally satisfy this equation because
n(n’ + 1)=na+ 1) =k. (4-110)

That is, Eq. 4-107 remains unchanged when the number n’ is substituted
for n. Hence
P_+1(cos 6) = P,{cos 6), (4-111)

and, for every solution of Laplace’s equation of the form

Vi= AreP (eos), (4-112)

there is another solution of the form

B B (4-1 13)
[= yori P_(n4y(COs 6) = prt P,(cos 6).

This also follows from the R function of Eq. 4-102.

Let us now proceed to find the Legendre polynomials P,,(cos 4) that are
solutions of Legendre’s equation. We know from our experience with point
charges that
Vee c (4-114)
is a solution of Laplace’s equation, C being a constant. This is readily veri-
166 ELECTROSTATIC FIELDS III

fied by substitution in Eq. 4-94. Since we are looking for solutions of the form
indicated in Eqs. 4-112 or 4-113, it follows from the latter equation that
Po(cos 6) = 1. (4-115)
We use a prime on the symbol P because the polynomials that we derive here
differ from the Legendre polynomials by constant factors, as we shall see
later. They are nevertheless solutions of Eq. 4-107. Substituting © =
Pi(cos 0) = 1 and n = 0 into Eq. 4-107 does in fact solve it.
Having found P(cos 6), how can we find Pi(cos 6) and all the other
polynomials corresponding to integral values of the index n in Eq. 4-107?
We shall do this starting with V;, but first we must know that any partial
derivative of a solution of Laplace’s equation with respect to any of the
Cartesian coordinate variables is also a solution. This is easily demonstrated
by substituting 0V/dx in Laplace’s equation and remembering that the order
of differentiation in partial derivatives is immaterial.
Let us therefore find the negative partial derivative of V; with respect
LOW
aC
eis Gor
waa (4-116)

where
See (2 + y+ 22 = Sarcaee
OZ
(4-117)
OZ iP ;
We therefore have a new solution of Laplace’s equation:

V, = CF (4-118)
Comparing once again with Eq. 4-113, we see that

Pi(cos 0) = cos @. (4-119)

Substitution of Pi(cos @) = cos @ for © into Eq. 4-107 shows that it really is
a solution when n = 1.
Equation 4-112 provides another possible solution for a given P,(cos 6).
In addition to V,, we have another solution:

V2 = DrPi(cos 6) = Dr cos 80. (4-120)

To find P2(cos 6), we differentiate V2 with respect to z:

ee Cos 0 a) z (3 cos* #— 1)
V3 as (c ) - (c4) C A > (4-121)
Z jee Oz

Comparing this with Eq. 4-113,

P3(cos 6) = (3 cos? 6 — 1).

(4-122)
 167

20 40 60 80 100 120 140 160 180°

Figure 4-21. The first three Legendre polynomials.

Again there is another solution:

V3 = Fr°(3 cos? 6 — 1), (4-123)
which corresponds to Eq. 4-112. We shall stop here, but we could continue to
find further polynomials in this way by repeated partial differentiations with
respect to Z.
It is convenient to multiply the above polynomials by normalizing factors
to make them all equal to unity at cos @ = 1. For example, P3(cos @) must be
multiplied by 3 to make P,(cos @) = 1 when cos 6 = 1. The general form of the
normalized Legendre polynomial is
1 Om
(cosa? a1)”, (4-124)
NOSES ye (cos 6)"
The first six are shown in Table 4-2; those for n = 1, 2, 3 are plotted as
functions of @ in Figure 4-21.

Table 4-2. Legendre Polynomials

n P,,(cos 0)

0 1
1 cos 8
2 3 cos? 6 — 4
3 5 cos’ @— 3 cos
4 85 cost @ — 1 cos? 0 + §
a 63 cos’ @ — 8° cos* 6 + 4,° cos 8
168 ELECTROSTATIC FIELDS III

Table 4-3. Solutions of Laplace’s Equation in Spherical Polar Coordinates

in the Case of Axial Symmetry

n r°P,, (COS 6) r_ YP. cos 6

i ae
1 rcosé f= COSiU,
2 4$r%(3 cos? @ — 1) 3r_*(3 cos? 6 — 1)
3 =4r%(5 cos? 8 — 3 cos @) 3r_*(5 cos* 6 — 3 cos A)
4 4r'(35 cos! 6 — 30 cos? 6 + 3) tr_°(35 cos! 6 — 30 cos? @ + 3)
5 4r*(63 cos> 6 — 70 cos? @ + 15 cos @) 4r (63 cos® @ — 70 cos? 6 + 15 cos 6)

A general solution of Laplace’s equation in spherical polar coordinates,

assuming axial symmetry, is therefore

V= 2d,int
CE COS 0) \ 22d,Brr-°*» P,(cos 8). (4-125)

The lower order terms are given in Table 4-3.

It can be shown that the above functions form a complete set offunctions:
an arbitrary boundary condition with axial symmetry can be satisfied with
such an infinite series. Moreover, any function of the polar angle @ can be
represented as a series of Legendre polynomials, provided the function is
continuous within the range of 6 considered. Finally,

Hey 0 if m # n,
/ : P,,(cos 0) P,(cos 6) d(cos 6) = ieeea
ifm=n. (4-126)
(2n + 1
This property of orthogonality of the Legendre polynomials is important in
evaluating the coefficients 4, and B,, of Eq. 4-125.

Example CONDUCTING SPHERE IN

A UNIFORM ELECTRIC FIELD
Figure 4-22 shows an insulated conducting sphere situated in a uni-
form electric field E,. At any point, either inside or outside the
sphere, the electric field intensity is E, plus that due to the induced
charges. We assume that the charges that produce E, are so far away
that they are unaffected by the presence of the sphere. This is rea-
sonable because, as we shall see, the field is distorted only in
the im-
mediate neighborhood of the sphere. The induced charges
arrange
themselves on the conducting sphere so that the total field is
zero
inside. We shall calculate the field outside the sphere by
solving
Laplace’s equation in three different ways.
 169

a
Figure 4-22. Lines of force (indicated by arrows) and equipotentials for
a conducting sphere in a uniform electric field. The lines of force are
normal at the surface of the sphere, and there is zero electric field
intensity inside. Observe that the field is hardly disturbed at distances
larger than one radius from the surface of the sphere. The origin is at the
center of the sphere and the polar axis used in the calculation points to
the right.

(a) The field is best described in terms of spherical polar coor-

dinates, with the origin at the center of the sphere and the polar
axis along E,. Our boundary conditions are then
V=0 (r = a), (4-127)
= —E£,z = —E,rcosé@ (r 00), (4-128)
At r = a, from Egs. 4-125 and 4-127,

0 = >> A,a"P,(cos 6) + >) Bra~*PP,(cos 6). (4-129)

n=0 n=0

The method of evaluating the coefficients A, and B, is similar to

that which we used for evaluating the C,’s of Eq. 4-78. We multiply
both sides of the equation by P,,(cos 6) and integrate from cos 6 =
—1 to cos @ = +1;

= 1
)Pm(cos 9) d(cos 8)
io 2,/af A,,a"P (COS

+ DD [7 Bra Py(C08AP n(cos 6)d(cos6).(4-130)

n=0°

According to Eq. 4-126, the only nonvanishing terms are those for
for which n = m and
170 ELECTROSTATIC FIELDS III

+1
()) = Algae fi
eeP?(cos 8) d(cos 0) + B,a~ “TY /= P?(cos @) d(cos 6),
(4-131)

= Aaa (525)
2n+1
+ Barss (5 5)
2n+1
(4-132)
Thus
Ba A ae (4-133)
For r — © the potential V is — E,r cos @. All the terms involving
inverse powers of r go to zero, and

—E,rP,(cos 6) = >) Anr"P,(cos 8). (4-134)

n=0

Inspection of this equation shows that the only term that is not zero
on the right-hand side is that for which n = 1. We can show this in a
formal manner by multiplying both sides by P,,(cos #) and integrat-
ing from cos 6 = —1 to cos @ = +1. By either method
Ae ee (4-135)
and all the other A,,’s are zero. Then all the B’s are also zero except
153
By = —A,a* = E,a’. (4-136)
Finally, at any point (r, 6),
3 3
VGx0) = —E,rcos6 + E, 4% sok
r
= —E, (1— 5) ros 6

(4-137)
and
OV 2a
E, = Gane E, (1= =) cos 6, (4-138)

10V i \e
Ee = Soe = —E, (1= “;)sin é. (4-139)

The surface density of induced charge on the sphere is simply equal

to €) times E, at r = a:

o = 3e6£E, cos 0. (4-140)

Returning now to Eq. 4-137 we observe that the first term is the
potential corresponding to the uniform field E,. The second term
has the form of the potential due to a dipole (Section 2.9). In fact,
if we replace the sphere by a dipole of moment

Dp = 4reE,a’ (4-141)
located at the center, the field outside the surface previously occu-
pied by the sphere remains unchanged. We shall examine the image
aspect of this field in a problem at the end of this chapter.
(b) We could also have determined the field quickly from Eq.
4-125 by a less formal method. We must have the term — Fr cos @ to
fit the condition at infinity. No other function with positive powers
 171

P(r,0)

Figure 4-23. The electric potential V at P(r, 6) can be calculated from

the applied electric field intensity E, and the induced surface charge
density o. We first calculate V on the axis of symmetry at P(r, 0) by a
simple integration. This value is then used as a boundary condition to
determine the coefficients of the Legendre polynomials in the series
for V at any point P(r, 8).

of r can be included. This one term, however, is inadequate to fit the

condition at r = a, where V must be independent of 6. We must
therefore add another function which also includes the cos @ factor
in order that the coefficient of cos 6 can be zero at r = a. Then

B cos6
V = —£E,rcos@ +
poe
(4-142)
We finally set B = E,a* to make V = Oatr = a. Our solution satis-
fies both Laplace’s equation and the boundary conditions; thus, it
is the correct solution, according to the uniqueness theorem.
(c) There is still another method of calculating this same field
that will add to our understanding and that will further illustrate
the use of Legendre polynomials. Consider Figure 4-23. As indicated
previously, the potential at any point (r, @) arises from two charge
distributions: (1) that which produces the electric field intensity
E, and which resides on electrodes situated far away, and (2) that
which is induced on the surface ofthe sphere. This latter distribution
is unknown, and we denote it by o(6’). We use a prime on @ to dis-
tinguish it from the polar angle for a point (r, @) outside the sphere.
At the general point (r, @) the total potential from these two sources
must be of the form shown in Eq. 4-125.
Now it is possible to compute from Coulomb’s law the potential
at a point P on the axis @ = 0 at a distance r = z from the center
of the sphere:
teo(0 ee 0! do" (4.143)
V = —£&,z+ ;
Atréo
172 ELECTROSTATIC FIELDS III

where a is the radius of the sphere and s is the distance from a point
on the sphere to P as in Figure 4-23:
SS ee 9
s=V24
@ — 2azcos 6’ = Zot + 5 cos 6. (4-144)

Expanding 1/s and grouping terms involving the same power of

(a/z),
ae = [1 + OnZeose We l + (5costa
== enh eeea) 2

+ GSA cos? 6 mets3 C08 7)s,

a + ot
, Z
(4-145)

='+4
+=© P,(cos 6’) ae~Ps(cos Oy) + -+., (4-146)

as long as z > a.
We have already seen that any continuous function of the polar
angle @ can be expanded as a series of Legendre polynomials. Thus
a(6’) = by + b:Pi(cos 6’) + boP2(cos 6’) + -+-, (4-147)
where do, b;, --- are constants. Equation 4-143 then becomes
2

Yai Baris 0
itie [bo + biP\(cos 6’) + beP2(cos 6’) + +++]
—

x [E+ 5Piccos 6") + SPacos 6") + + + |atcos 9). (4-148)

The orthogonality property of the Legendre polynomials makes this
an easy integral to evaluate, and

V >= —Ear F(t

a (2b) ,2ha , 2a
St St) 4149)
Thus, when 6 = 0 and r = z, the general solution for V shown
in Eq. 4-125 reduces to the above form, and we can match coefficients
term by term to find V at any point (r, 6) outside the sphere. On do-
ing this we find that all the 4,’s are zero, except for

A, = —E,, (4-150)
and that
b,, a” +2
(4-151)
~In+1 ee

To evaluate the 5,’s, we use the fact that the potential is zero at
~ |= a. Substituting the above coefficients into Eq. 4-125 and setting
r="a:

= —E,aP,\(cos @) +78
wae—« Pileos 6) + 36,7(cos @) + -

(4-152)
which must be true for all @. Thus both the term boa/eo, which is
4.5 |SOLUTION OF LAPLACE’S EQUATION IN SPHERICAL COORDINATES 173

independent of 6, and the coefficients of all the P,,’s must be equal

to zero:
bo = 0, (4-153)
and

hace Rat (4-154)

3€ z
or
by = 3e0E,. (4-155)
All other 3,,’s are zero.
The potential V at any point (r, 6) is thus given by substituting
into Eq. 4-125 A; = —E,, as in Eq. 4-150, and
Be Eas (4-156)
as in Eqs. 4-151 and 4-155. The field is the same as that found pre-
viously.
The surface charge density o(6’) on the conducting sphere can
be obtained from Eq. 4-147 now that the 5,’s are known: we find
the value previously found in Eq. 4-140.

Example DIELECTRIC SPHERE IN

A UNIFORM ELECTRIC FIELD
We can calculate this field by either of the formal methods discussed
above if we write a general solution like Eq. 4-125 for points outside
the sphere, and write another solution with different coefficients for
points inside the sphere. The coefficients must be chosen in such a
way that the boundary conditions are satisfied:
V — —E,r cos 6 (r
> 0);

V is continuous across the boundary (aa)

the normal component of D is continuous (=a)

Instead of following such a formal procedure, however, we

shall write down a combination of spherical harmonics that will
satisfy all the boundary conditions.
Outside the sphere, we must have — E,r cos @ as one ofthe terms
in the solution to satisfy the condition at r > «©. Furthermore, this
is the only harmonic with a positive power of r that we can permit,
for otherwise the condition at r > » would be violated. As regards
this condition, all the terms with inverse powers of r are acceptable.
Consider now the solution for points inside the dielectric sphere.
No inverse powers at all are permissible here, since such terms
would make the potential infinite at the center. This is clearly im-
possible, since the only charges in the system are those which pro-
duce the field E, and those induced on the surface of the sphere, if
we assume a Class A dielectric, with the result that no volume dis-
tribution of induced charge exists.
174 ELECTROSTATIC FIELDS III

Writing V, for the potential outside the sphere and V; for that
inside,

V, = —E,r cos + >) Bur"*)P,(cos 8), (4-157)

n=0

Vie CTP (Cuat6). (4-158)

We require that V, be equal to V; at r = a and that
eae 0) --(« ore)
0) (4-159)
r= rT=a

where e, is the relative permittivity of the sphere. These are the sec-
ond and third boundary conditions discussed above. Therefore
B,P,(cos @) seB»P(cos
= 6) rae
—E,aP\(cos 0) + on =

= Cy) + CiaP,(cos 0) + Cra*P(cos #) + ---, (4-160)

and
PE
Pcosey a | Zea 8) sees @) i"
a a a

= —e,C\P\(cos 6) — 2€,CraP2(cos 8) -- ---. (4-161)

In order that these two equations be true for all values of 6, the
coefficient of each Legendre polynomial on the left must be equal
to the coefficient of the same Legendre polynomial on the right.
Thus, from Eq. 4-160,

By. (4-162)
a

—E,a + 2 See (4-163)

By P
Be Coa?,+ +, (4-164)

and, from Eq. 4-161,

Bo |
oe 0, (4-165)

2B
Ei ae = = —€,Or, (4-166)

3B,
4 = —2e,Cra. (4-167)

These sets of equations lead to the following values for the coeffi-
cients:
Bo —— C=SS 0, (4-168)

iy 1 |a(S 7— 5) seake3 :
(4-169)
 175

Figure 4-24. The field near a dielectric sphere (e, = 3) ina uniform electric
field. The lines of D (indicated by arrows) crowd into the sphere as
shown, with the result that D is larger inside than outside. Since there is
no free charge at the surface of the sphere, the lines of D neither origi-
nate nor terminate there, and they are continuous across the boundary.
The equipotentials spread out inside, corresponding to a Jower electric
field intensity E. The electric field intensity E is discontinuous at the
surface, and the density of lines of force is lower inside than outside. As
in the conducting sphere, the field is hardly disturbed at distances larger
than one radius from the surface. The field inside is uniform. The origin
is chosen at the center of the sphere and the polar axis used in the
calculation points to the right.

BE
Cragg (4-170)
B,= C, =10 (n > 1). (4-171)
Thus
WAC -|1 a (2 =5 7 E,rcos0,
e- -1\a’
(4-172)
and
Vir, 0) = s 5)Eencoss—=
-(- mg = 5)E,z.
-(< 7 oz. (4-173
(4-173)

We may calculate the electric field intensity inside and outside

the sphere by calculating —VV. It will be observed that the electric
field inside the sphere is uniform, is along z, and that
3
E; = (- = 5)E,. (4-174 )

Lines of D and equipotentials are shown in Figure 4-24.

 ELECTROSTATIC FIELDS II
176

4.6 SOLUTION OF POISSON’S EQUATION FOR V

We have as yet dealt only with solutions of Laplace’s equation, since we have
concerned ourselves only with regions where the electric charge density p, = 0.
When p, ~ 0 we must find a solution of Poisson’s equation VV = —p;/é
(Eq. 3-42) which is consistent with the given boundary conditions. We have
already seen in Section 4.2 that such a solution is unique for a given p.

Example p-n JUNCTION DIODE IN SILICON*

We shall use Poisson’s equation to study the electrical properties
of a silicon p-n junction diode in open circuit.
Semiconductors contain two types of fixed charges and two
types of mobile charges. There are (a) fixed positive donor atoms,
which lack an electron; (b) fixed negative acceptor atoms, which
have an extra electron; (c) mobile conduction electrons; (d) mobile
positive holes.
A hole is a vacancy left by an electron liberated from the valence
bond structure in the material. A hole behaves as a free particle
carrying a positive charge numerically equal to that of the electron,
and it moves through the semiconductor much as an air bubble
rises through water.
Donor and acceptor atoms are atoms of foreign substances that
are added in minute amounts (approximately one part per million)
to the basic semiconductor material, which is usually germanium
or silicon. Since they are fixed in the crystal lattice, they constitute
fixed charges. The addition of donor atoms adds conduction elec-
trons without adding corresponding holes and yields n-type material,
which conducts mostly by negatively charged electrons. Similarly,
the addition of acceptor atoms gives p-type material, which conducts
mostly by positively charged holes.
Let us call P and N the numbers of positive and negative fixed
charges per cubic meter, and p and n the numbers of positive and
negative mobile charges per cubic meter. Then the electric charge
density is
pi = e(P —N+p-—n), (4-175)
and

V-E= —V°V ="(P—N+p-—n), (4-176)

* See R. B. Adler, A. C. Smith, and R. L. Longini, /ntroduction to Semiconductor Physics,

and Paul E. Gray, David DeWitt, A. R. Boothroyd, and James F. Gibbons, Physical
Elec-
tronics and Circuit Models of Transistors, SEEC Books, Volumes
I and 2 (John Wiley, New
York, 1964) for a more detailed discussion of semiconductors.
4.6 SOLUTION OF POISSON’S EQUATION FOR V 177

where € = €,€9. The relative permittivity €, of germanium is

16, while
that of silicon is 12.
Let us assume that the numbers P and N are graded uniformly
inside a plane infinite junction as in Figures 4-25a, b, c. This
one-
dimensional approximation is valid for most solid-state diodes. Then
the semiconductor is n-type to the left of the origin and p-type to the
right.
The conduction electrons and the holes diffuse inside the mate-
rial like the molecules of a gas; under thermal equilibrium, p and n
satisfy the Boltzmann equation of statistical mechanics:
D = No exp (—eV/KT), (4-177)
n ll No exp (eV/kT), (4-178)
where 1 is the value of p or of n at the point where the potential V
is chosen to be zero—let us say at the origin. It turns out that No IS
the density of either conduction electrons or holes in the basic semi-
conductor material. The density p decreases exponentially with in-
creasing V, while nm increases exponentially with V (or decreases
exponentially with —V).
Combining the last three equations and utilizing the fact that
all quantities are independent of y and z by hypothesis, we find the
differential equation for V:

Oe
dx €
tsa (ce TO)

= 2m ©sinh (eV/kT) — 2 :)x (4-180)

where ¢ is the thickness of the junction and Py is as in Figure 4-25a.
This equation does not appear to have an analytical solution;
however, it has been solved numerically by using the relaxation
method (see Problem 2-16).* Figure 4-25 shows the curves obtained
for a linearly graded p-n junction in silicon at equilibrium and at
room temperature.
Such p-n junctions are used as diodes: relatively large currents
can flow when the p-type material is made positive with respect to
the n-type, and negligible currents flow in the opposite direction.
When the applied voltage makes the p-type material positive, posi-
tive holes from the p side and negative electrons from the n side are
driven into the junction, where they combine. On the other hand,
when the p-type material is made negative, the holes and the elec-
trons are removed from thejunction, and the current is smaller by a
few orders of magnitude.

*D. P. Kennedy and R. R. O’Brien, On the Mathematical Theory of the Linearly-

Graded P-N Junction, IBM Journal of Research and Development 11, no. 3 (May 1967),
22s
178 LO, 0a

OO Sen 00 O20 ON 200 OOS Oe LOO 120

(a) Microns

—120—100 —80 —60 —40 20 O 20" +40" (60 80 100 120

(b) Microns

120

(c) p-type n-type

Figure 4-25. Curves for a plane p-n junction in silicon at 300 Kelvins. Density of donor and
acceptor atoms on either side No = Py = 107 meter~, thickness f = 200 microns, or
2 X 10-4 meter. Curves (a) and (b) show respectively the densities of donor and acceptor
atoms as functions of x. The x-axis is perpendicular to the junction and the origin is at
the center of the junction. Curve (c) shows the net density of fixed charges. Remember
that donor atoms give fixed positive charges and mobile electrons, while acceptor atoms
give fixed negative charges and mobile positive holes. Curve (d) shows the potential V as
a function of x. There exists a permanent contact potential between the two sides of the
junction. Note that V varies roughly like P — N. The potential is positive on the side
that has a positive net fixed charge (n-type material) and negative on the other side. This
is because thermal agitation forces the free electrons from the ” side to diffuse into the
p side, while it forces holes from the p side to diffuse in the other direction. The electric
field intensity E = —dV/dx is shown in the next curve, (e). Since it is negative, it points
away from the n-type material, which contains a majority of fixed positive charges. This
electric field limits the diffusion of electrons and holes across the junction. The net
charge density p, curve (f), is proportional to minus the second derivative of V. It has
the same sign as V and as the fixed charge density P — N. Finally, curves (g) and (h)
show the numbers ” and p of mobile negative and positive charges. Curve (f) for the
charge density p is related to curves (c), (g), (h):

p=e(P—N+p-—Rn).

The product np is everywhere equal to n2.

 179

—120 —100 —80 —60 —40 —20 60 80 100 120

Microns

Mads

(d) — 0.064

E ete

ee eS Se
—120 —100 —80 —60 —40 —20 D0 OO eS LOO ml e20)
Sol Microns
L

(e) =80

—80 —60 —40 —20 0 20) 40) 160) 50 100 120

120 —100
(g) Microns

ae ! | |
ae 80 100 120
—60 —40 —20 0 20 40 60
—120 —100 —80
Microns
(h)
180 ELECTROSTATIC FIELDS III

4.7 SOLUTION OF POISSON'S EQUATION FOR E

We can find the Poisson equation for E as follows. In Section 2.3 we found
that the curl of E is zero. Now, from Eq. 1-143,
VXVXE= —-WE+ V(V-E) = 0, (4-181)
and. thus
VE = V(V -EB), (4-182)

= Voy + po), (4-183)

€0
from Eq. 3-41.
The solution of this equation (Appendix E) is

(eeAre /
J7
Aad)
if
(4-184)
We have added a prime to the del to stress the fact that it contains derivatives
with respect to the source point (x’, y’, 2’).
This equation relates the electric field intensity to the gradient of the
total charge density. Note the negative sign and the first power of r in the
denominator. It is valid whatever the nature of the media that are present in
the field, as long as the gradient is definable.
The more usual expression for the electrostatic field intensity of a volume
distribution of charge is a consequence of Coulomb’s law and was found in
Eq. 3-30:
poe ea ae (4-185)
Although the two integrals for E are equal, the integrands are obviously un-
equal, since there exists no general relationship between p, and V’p, at a point
in space. Indeed, the two integrals are equal only if they extend over ail the
charge distribution.
It will be shown in Problem 4-30 that, as a result of the equality of these
two integrals,

|Vv ("")a: = 0 (4-186)

for any volume distribution of charge, again, if the gradients exist.

Example Imagine a sphere of charge that has a uniform density throughout

its volume, except near the periphery, where the density gradually
decreases to zero. Then, if E is calculated by means of the integral
of Eq. 4-184, only the region near the surface contributes to the
integral, since V’p, is zero everywhere else (see Problem 4-29).
 181

4.8 SUMMARY

In this chapter we have dealt with electric fields that cannot be calculated
easily by the methods of Chapter 2.
We first established the conditions that must be satisfied by the potential
V, by the normal component D, of the electric displacement, and by the
tangential component £, of the electric field intensity at the boundary between
two media: all three must be continuous across any boundary that does not
carry a surface charge. If there is a surface charge, then D,, is discontinuous
but V and E£, are still continuous.
We then demonstrated the uniqueness theorem according to which, for a
given set of boundary conditions, there is only one possible electric field. This
theorem is of great practical importance: if we can somehow find a potential
V (x, y, Zz) that satisfies both the boundary conditions and Poisson’s equation,
then we know that we have found the correct potential.
The method of images can sometimes simplify the calculation of electric
fields that involve conducting surfaces and dielectrics. For example, the field
of a point charge Q near an infinite conducting plane is the same as if the
plane were replaced by a charge —Q at the position of the image of Q. The
force on Q is of course the same in both cases. The method of images gives
the correct field only outside the region where the image is situated. In some
cases there is an infinite set of images.
We then solved Laplace’s equation for a number of different boundary
conditions requiring either Cartesian or spherical coordinates. Solutions of
Laplace’s equation are called harmonic functions. We found such solutions by
separating the variables: in Cartesian coordinates we set

V = X(x) YZ), (4-59)

where X is a function of x only, Y is a function of y only, Z is a function of z
only, while in spherical coordinates we had
V = R(r)O(0). (4-96)
There was no ® (¢) function, since we considered only fields that were inde-
pendent of the azimuthal angle ¢.
For both of the fields that we calculated in Cartesian coordinates we
were able to fit the boundary conditions only by adding an infinite number
of solutions involving sine functions. Such infinite series of the form
- . aw
S, Ge Sl n Ale,

n=1
b
182 ELECTROSTATIC FIELDS Ili

are called Fourier series; an arbitrary boundary condition V(y) can be sat-
isfied by such a series. The value of the C,, coefficients is found by multiplying
both the boundary condition V(y) and the Fourier series by sin (pry/b),
integrating from 0 to 6, and utilizing the orthogonality property of the se-
quence of sin (nry/b) functions:
ie ii p # n,
(B C, sin nny
|= sin =7 dy of
ATTY b
= ]On5 if 4-8]
pep ar
The separation of variables in spherical coordinates transforms Laplace’s
equation into two ordinary differential equations:

Hed
Pe
Months
dr n(n + 1)R = 0, (4-101, 4-103)

and Legendre’s equation,

: ja “S| sie il YOY (4-107)

where u = cos 6.
The first equation is readily solved by functions of the type

R(r)= Arm + —
ae (4-102)

The solutions of Legendre’s equation are called Legendre polynomials:

P, (cos 0) = a
Jnl a(cos (cost) —
6)" (cos? 6 — 1) I -12
(4-124)

The general solution of Laplace’s equation in spherical coordinates, for

axial symmetry, is then

V = >* AnrP,(cos 6) + >, Bar" P,(cos 6). (4-125)

n=0 n=0

The individual terms of this equation constitute a complete set of functions;

any arbitrary boundary value of the potential having axial symmetry can be
satisfied with such a series. The coefficients in the series can be determined by
using the specified potentials on the boundaries and by using the orthogonality
property of the Legendre functions:

+1 2 Li 201. 9%. 0,
ie P,,(cos 0) P,,(cos 6) d(cos 0) = | (4-126)
if m =n,
 183

Figure 4-26.

Poisson’s equation for V is

Vy pr Pb how abt fee (3-42)
€0 €0

Poisson’s equation for E is

WE
_ Vos + ps) _ Vow (4-183)
€0 €0

and its solution leads to an alternate integral for E:

eee / EIN Seaat Dy als / V'pe (4-184)
4irep } Are ie

It follows that, for any volume distribution of charge,

(4-186)

as long, of course, as the gradients exist.

The reasoning that led us to Eqs. 4-183 and 4-184 is shown schematically
in Figure 4-26.

PROBLEMS
in the atmosphere
4-1. It is found experimentally that the electric field intensity
ter and that it points
near the surface of the Earth is about 100 volts/me
184 ELECTROSTATIC FIELDS III

downward: the potential increases as we go up. This field is maintained by

thunderstorms, which, on the average, deposit negative charges on the earth.
Calculate the electric charge carried by the Earth.

4-2. Figure 3-10 shows the E and D fields of a bar electret.

(a) Show that the lines of E are not bent at the cylindrical surface and that
they are bent at the end faces.
(b) Show that the inverse is true for the lines of D.

. It is interesting to speculate about the electrostatic potential at the surface

of the Sun (M = 2.0 X 10*° kilograms, R = 7.0 108 meters).
The electrons in the hot plasma at the surface easily attain the escape
velocity (2GM/R)! for an uncharged star, where G = 6.67 X 1071! is the
gravitational constant. Their mean kinetic energy is (3/2) kT, where k =
1.38 < 10-8 joule/kelvin is Boltzmann’s constant and the temperature T is
about 6 X 10’ kelvins. The protons have much lower thermal velocities be-
cause of their larger mass.
Electrons cannot escape indefinitely, however, because the Sun acquires
a positive charge that increases their escape velocity.
(a) Calculate the fraction of the electrons that possess enough energy to
escape when the charge is zero. This fraction is
exp [—(escape kinetic energy) /kT].
(b) Calculate the charge that is required to neutralize the gravitational force
on a proton. Note that this charge is independent of the radius.
(c) Show that it is only necessary to remove 160 electrons/meter? of the sur-
face of the Sun to achieve this charge.
(d) Assuming that equilibrium is achieved when the net attractive force on a
proton is zero, show that the average voltage at the surface of the Sun is about
2 X 10° volts.

4-4, Various devices, such as Van de Graaff particle accelerators for example,
have high voltage electrodes maintained under pressure in a metal tank. Let
us assume that the electrode is spherical and that it has a radius r;. The elec-
trode must operate at a voltage V with respect to the tank, which has a radius
r, and which is grounded.
We focus our attention on the electric field intensity at the surface of the
electrode, for that is where it is highest, and we try to optimize r; and fro.
(a) If we disregard the cost of the tank, then its optimum radius will be that
for which the electric field intensity in the insulating gas will be minimum, for
this will permit using a minimum gas pressure. What is this optimum value
of ro?
(b) In actual practice, cost, weight, and space requirements limit ry. One must
therefore optimize the ratio r./r by varying ri. Show that the E at the surface
of the high-voltage electrode (r = ri) has a minimum value of 2V/r; when
i), = Dp,
(c) Can you explain qualitatively why there should be an optimum conditio
n?
(d) To determine whether the optimum condition is critical or
not, plot the
PROBLEMS 185

value of E at r = r; for an actual case where rz was 0.483 meter and for values
of r; ranging from 0.1 to 0.4 meter.
(e) What range of values of 7; can be tolerated if one can tolerate an electric
field intensity 10% higher than 2V/r,?
(f) Calculate 2V/r, for V = 5 X 10° volts?

4-5. Perform the same calculations as above for a cylindrical geometry. You
should find that, when the radius r. of the outer cylinder is fixed, the electric
field intensity at the surface of the inner cylinder has a minimum value of
V/r, when r; = r2/e, where e is the base of the natural logarithms.

4-6. According to the uniqueness theorem, the Poisson equation V?V = —p,/€o
can have only one solution if the potential V is determined at the boundaries
of the field.
Show that two solutions can differ at most by a constant if the normal
component of VV is determined everywhere at the boundaries.

4-7. When the space between the plates of a parallel-plate capacitor is filled with a
dielectric e, it has a capacitance of C farads. If the dielectric is replaced by a
material whose resistivity p is much smaller than that of the electrodes, the
resistance between the electrodes is R ohms.
(a) Show that RC = pe, neglecting edge effects.{
(b) Show that this result also applies to cylindrical and spherical capacitors.
(c) Show that it applies to any pair of electrodes submerged in a medium
whose resistivity p is much smaller than that of the electrodes.
You should be able to show that the field is unaffected by the conduc-
tivity, with the above restriction.
One important application of this fact is the electrolytic plotting tank,
which is used for plotting electric fields in two, and in some cases three,
dimensions.

4-8. Ekctrified dust particles are ejected into the atmosphere and form an elon-
Fr ~ gated cloud of approximately cylindrical form at an altitude of 40 meters in a
3 meter/second wind. The current feeding the cloud is 80 microamperes.
Calculate the resulting electric field intensity at the surface of the Earth
directly under the cloud. Assume the Earth to be flat and conducting.

4-9. A point charge Q is situated between two horizontal parallel conducting

plates separated by a distance s and at a distance x above the lower plate.
(a) Calculate the force due to the image charges in the form of an infinite
series.
(b) Find an approximate value for the force when Q is situated (i) near one
of the plates and (ii) near the position x = s/2.
(c) In the Millikan oil-drop experiment a small oil droplet carrying a few
excess electrons is situated in the electric field between two charged parallel
plates separated by a distance s. The force on the droplet is calculated from the
electric field V/s, where V is the difference in potential between the plates.
The image force is neglected. Is this serious?
186 ELECTROSTATIC FIELDS III

4-10. (a) Show that the force of attraction or repulsion between a point charge g
and a conducting sphere of radius R carrying a charge Q 1s
q {Q+(R/D)q _- Rq .
Arr, D? DED ae
where D is the distance from g to the center of the sphere.
(b) Show that the force is attractive when q and Q are of the same sign, for
Opes Bie
7 ORY
4-11. In the last example in Section 4.3 we calculated the surface charge density
induced on the surface of a block of dielectric by a point charge Q at a dis-
tance D in front of it.
Show that this surface charge density and Q give the correct electric
potential V (a) at the foot of the perpendicular drawn from Q to the dielectric
surface, and (b) at a distance D behind the boundary.

4-12. Show that there exist solutions of Laplace’s equation that are of the form
X(x) + Y(y) + Z(@).
4-13. Calculate the potential at x = O for the electrodes of Figure 4-16, using the
first five terms of the Fourier series. Perform the calculation for y = 0.1 5,
0.2 6b, and so on.

4-14, Use the fact that 1/r is a solution of Laplace’s equation to show that

earl al ate A) lati tacr ol

are also solutions when a, 6, c are constants.

4-15, A dielectric sphere of radius R has a uniform polarization P.

(a) Show that the electric field intensity inside is uniform and equal to
— P/3¢9.
(b) Show that the electric field intensity outside is that of a dipole situated
at the center of the sphere and having a dipole moment equal to PV, where V
is the volume of the sphere.{
4-16. A grounded, infinite, circular, cylindrical conductor is introduced into a pre-
viously uniform electric field with its axis perpendicular to Ep.
Show that
V = E{1 — (a’/p*)|p cos ¢.t

4-17. Two point charges +@Q and —Q are situated on a diameter of a conducting
sphere of radius a at distances D > a to the right and to the left, respectively.
(a) Show that the image charges constitute a dipole of moment 2a*Q/D?
at the center of the sphere.
(b) Now let D and Q approach infinity in such a way that Q/D® remains
constant. Superpose the fields of + Q and of the dipole to find the field outside
the sphere,
PROBLEMS
187

This is another way of calculating the field around an uncharged conduct-

ing sphere that is introduced into a previously uniform field,
(c) Show that the surface charge density is 3¢.£, cos 6, where E, is the field
due to the charges +Q.

4-18. Show that a sum of terms of the form r” sin n@, where n is any positive or nega-
tive integer, is a solution of Laplace’s equation in two dimensions.

4-19. A linear distribution of charge of \ coulombs/meter extends along the z-axis

from z = —a to z = +a. Show that, at any point for which r > a,

4reV = —*4 Py(cos 0) +ea" Px(cos 0) + oe

a”P,(cos 0)+-

4-20. The axial electric field intensity E, on the axis of the accelerating tube in a
particular type of ion accelerator is given approximately by

FE, = En + kz’,

where z is measured from the center of the tube along its axis. The azimuthal
component £E, is zero.
(a) Show that the radial electric field intensity in the neighborhood of the
axis is —kzp, assuming that the charge density is zero.
(b) Draw a rough sketch of the lines of force in a plane that contains the axis.
(c) What is the maximum charge density that can be tolerated if the value
calculated above for the radial field is to be accurate within 5 % at the ends of
the tube? The accelerating tube is 1.00 meter long, E,9 is 7.5 X 10° volts/meter,
and k is 1.00 * 10° volts/meter’.

4-21. It is in some cases necessary to ionize a low pressure gas by bombarding it

with electrons. One example is ion pumping. In ion pumps the gas is removed
from the system to be evacuated by driving the positive ions into a suitable
cathode where they are absorbed.
Now such pumps are especially desirable at very low pressures, at which
the mean free path of electrons between collisions with molecules of the re-
sidual gas can be of the order of thousands of kilometers. Methods have there-
fore been devised for increasing the electron path length within the pump.
The orbitron is one of several types of ion pumps. It utilizes a cylindrical
cathode and an axial rod as anode. Electrons from one or more filaments are
injected near one end of the cylinder with sufficient angular momentum to
make them orbit around the anode. They are also given a slight axial velocity,
and they describe helices along the length of the pump. End plates connected
to the cathode reflect electrons at both ends. One end of the cylinder is pro-
vided with a flange for connection to the vessel to be evacuated.
One particular orbitron has a diameter of 10 centimeters and a length of
4O centimeters. The anode has a diameter of 5 millimeters and is maintained
at +5 kilovolts with respect to the cylinder, which is grounded.
Assume first that there is zero space charge, and neglect both the axial
velocity and end effects. It is shown in mechanics that a particle in a central
force field will describe a stable circular orbit if
188 ELECTROSTATIC FIELDS III

where f(r) is the magnitude of the radial force of attraction exerted on the
particle.
(a) Are the orbits stable when there is zero space charge?
(b) Now assume that the gas pressure is zero and that the space charge arises
solely from the orbiting electrons. Qualitatively, how will this space charge
affect the electric field intensity? Will it eventually lead to unstable orbits?
(c) Assume that, in the presence of gas, the radial electric field intensity is
constant. Are the electron orbits again stable?
It is possible to use essentially the same device for transporting charged
particles over considerable distances. In this case the particles are captured
into spiral orbits at one end of the tube and travel to the other end where they
are detected. The distance of travel is limited only by the quality of the vac-
uum. One particular guide that was used for extracting alpha particles from a
target inside a reactor had a length of 6.3 meters and a diameter of 7.5 centi-
meters. The central wire was maintained at —30 kilovolts.
4-22. In 1959 Lyttleton and Bondi* suggested that the expansion of the Universe
could be explained on the basis of Newtonian mechanics if matter carried a
net electric charge.
Imagine a spherical volume of astronomical size containing un-ionized
atomic hydrogen of uniform density », and assume that the proton charge
€» = —(i + y)e, where e is the electron charge.
(a) Show that, for y > 10718, the electrostatic repulsion becomes larger than
the gravitational attraction and the gas expands.
(b) Show that the force of repulsion on an atom is then proportional to its
distance R from the center and that, as a consequence, the radial velocity of
an atom at R is proportional to R. Assume that the density is maintained
constant by the continuous creation of matter in space.
(c) Show that the velocity » = R/T, where T is the time required for the radial
distance R of a given atom to increase by a factor of e. This time T can be
taken to be the age of the Universe.
(d) In the Millikan oil-drop experiment an electrically charged droplet of oil
is suspended in the electric field between two plane horizontal electrodes. It is
observed that the charge carried by the droplet changes by integral amounts
within an accuracy of about 1 part in 10°,
Imagine that the proton charge e, is equal to —(1 + y)e, where e is the
charge of the electron. Can you show that the Millikan oil-drop experiment
leads us to believe that y is less than about 10-17?
. Electrostatic precipitation is used extensively for the elimination of dust par-
ticles from industrial gases, for example, for eliminating fly-ash from the
smoke of coalfired electric power plants. This method was developed by
Cottrell at the beginning of the century. In this process the dust particles are
charged by the ions formed in a corona discharge, and they then drift in the
electric field to the electrodes, where they are deposited.

* Proc. Roy. Soc. (London) A ide,

252, p. 313.
PROBLEMS 189

In one type of precipitator the anode is a grounded cylinder having a

radius R of 15 centimeters, and the cathode is a wire supported along the axis
of the cylinder and maintained at a potential V of —50 kilovolts. The gas is
ionized and ions of both signs are formed in the corona discharge near the
wire. The positive ions quickly reach the center wire, while the negative ions
move out radially to the cylinder. The space charge is thus negative over most
of the volume of the cylinder.
Under these conditions it is found experimentally that E is approximately
equal to V/R for all values of r. If the dust particles are at least slightly con-
ducting, they acquire a negative charge QO of 127) Ea”, where a is their radius.
It turns out that if they are nonconducting Q is somewhat smaller.
(a) Set i to be the electric current per meter and vy to be the mobility
(velocity/E) of the negative ions. Show that, for any r, the current due to the
ions is
i= 27nrpyk;
and that the space charge density
p = ©E/r.
(b) The drift velocity of the dust particles is given by Stokes’s law: it is the
force EQ divided by 6zna, where 7 is the viscosity of the gas. Show that their
drift velocity » is 2e)E?a/n.
(c) Calculate i, p, v, and the time required for a dust particle to drift from the
cathode to the anode, when y = 2 X 10-4 meter?/volt second, a = 5 microns,
and 7 = 2 X 10~° kilogram/meter second.
This simplified theory neglects turbulence, which turns out to be impor-
tant in practice.

4-24. Would it not be possible to build a perpetual-motion machine that would

utilize the contact potential in a junction diode?
4-25. If the number of electron-hole pairs generated per unit volume and per unit
time in a semiconductor is G, and if the number of electron-hole recombina-
tions is R, show that
on
VeIn — ear = —e(G — R),

Op
V-J> + anes = e(G = R),

of
where n and p are, respectively, the number of electrons and the number
holes per unit volume, and where J, and J, are, respective ly, the electron
current density and the hole current density.
is a
The number of recombinations R per unit time and per unit volume
concentra tions n and p, while the number G depends
function of the carrier
on the temperature, on the illumination, etc.
and collected
4-26. In the vacuum diode, electrons are emitted from a hot filament
Show that, for an idealized diode having two plane parallel
at the anode.
electrodes, the electron current density is
J = 2.34 X 10-*V3/2/s? amperes/meter’,
190 ELECTROSTATIC FIELDS III

where Vy is the potential difference and s is the distance between cathode and
anode.
This is the Child-Langmuir law. It is valid only for the plane parallel
diode and for electrons emitted with zero velocity. More generally,
J = kV3”,
where K is a constant, for both ions and electrons and for any geometry, as
long as the current is space-charge limited, and as long as uw is negligible at the
source. ft

4-27. Let us see how current flows through an ionized gas. Two large parallel elec-
trodes of area S are separated by a distance s which is small enough to render
edge effects negligible. Pairs of ions are created throughout the gas between
the electrodes at the constant rate of m) pairs/meter? second. Let us set
nt and n~ to be the numbers of positive and negative ions per cubic
meter, respectively ;
vt and v-, the velocities of the positive and negative ions;
J+ and J- the currents due to the positive and negative ions;
e, the absolute value of the electric charge of an ion;
the x-axis to be perpendicular to the plates, in the direction of the electric
field, with the origin at the positive plate. With these conventions v*
is positive and v~ is negative.
Show that the total current density
J = entv* — en-v~ = enox + en(D — x) = emD,
and that the charge density
as x — D
Pa eno yt y-

The velocities are both proportional to the electric field intensity E:

Or SS pis. Vn — ie
where ut and uw are the mobilities and where E is itself a function of x. If
the negative particles are electrons, u~ >> ut.
We have neglected both thermal diffusions and recombination.

4-28. The thrust produced by a rocket motor is equal to m’v, where m’ is the mass
of propellant ejected per unit time and v is the exhaust velocity. Rocket engi-
neers therefore strive to make m’ as small as possible by increasing v.
One way of achieving large values of » is to eject a beam of charged par-
ticles, as in Figure 4-27, which shows a schematic diagram of an ion motor.
The propellant is ionized in the ion source and is ejected as a positive ion
beam at a velocity corresponding to the accelerating voltage V. Electrons are
injected into the beam to prevent the rocket from charging up.
(a) The current / of positive ions in the beam is carried by particles of
mass m
and charge ne, where e is the electronic charge.
Show that the thrust is

F = [(2Vm/ne)}??.
 191

Figure 4-27. Schematic diagram of an ion motor. The propellant is admitted at P

and ionized in S; A is a beam-shaping electrode and B is the accelerating
electrode, maintained at a voltage V with respect to S; the filament F injects
electrons into the ion beam to make it neutral.

(b) What is the value of F for a 0.1 ampere beam of protons accelerated to
50 kilovolts? Ion motors are used in outer space where small thrusts are re-
quired to correct either the attitude or the trajectory of satellites.
(c) Show qualitatively that the thrust is independent of the shape of the field
or of the presence of space charge in the acceleration region and that it is
always given by the above formula.
(d) If we call W the power JV spent in accelerating the beam, show that

» A ied ola Leo) Dah ks

F = (2Wm’')"/ ee W =)

Thus, for given values of W and m’, the thrust is independent of the
charge-to-mass ratio of the ions. Or, for a given W, F is inversely proportional
to v. The last expression shows that, for a given power expenditure W, it is
preferable to use heavy ions carrying a single charge (n = 1) and to use as low
an accelerating voltage V as possible.
(e) Sketch a graph of the voltage along the axis, inside and outside the motor,
(i) when the electron source is on, and (ii) some time after it has been turned
off. Assume that the body of the rocket is connected to electrode B.
(f) If the electron source is turned off, and if the beam current J is one ampere,
how long will it take the body of the rocket to attain a voltage equal to the
accelerating voltage if V is 50 kilovolts? Assume that the rocket is spherical
and that it has a radius of 1 meter. At that point the motor ceases to operate
because the ions follow the rocket.

4-29. A sphere of electric charge has a density p which is a function of the radius
as in Figure 4-28.
(a) Use Gauss’s law to show that, at a distance r > B from the center of the
sphere,

a
Po_ (8? + a°\(6 + a).
12e0r
192

(b) Show that one arrives at the same result by using Eq. 4-184.
Note that the result is unaffected as (8 — a) — 0.
(c) Does this value of E make sense when a = 6?

4-30. We have shown in Section 4-7 that, for static fields,

1 Put ot = = : [ee
Aire, Je 9 4rr€ i}

r;

Figure 4-28,

Show that, as a consequence,

i7 Vv’ (£)
ir
dr' = 0.t
4-31. Show that
i)v’ x (2) ar’ =0
for any finite charge distribution.

Hints

4-7. The charge density is zero in a uniform conducting medium.

4-15. Choose the z-axis along P and use Legendre polynomials to calculate V. You can
determine the values of all the coefficients from the properties of the field without
performing any integration.

. Show that this V obeys Laplace’s equation in cylindrical coordinates and that it
gives the correct boundary conditions.

Solve Poisson’s equation, setting the charge density p equal to J/u, where u is
the velocity of the electrons and (1/2)mu? = eV.
If your calculation involves (—J)'/2, disregard the minus sign and consider
simply the magnitude of the current.
Assume that the electrons are emitted with u = 0.
Set dV/dx = 0 at the cathode. This means that there is an unlimited supply of
electrons available at the cathode and that the current is limited by the electron
space charge, and not by the nature of the cathode.

4-30. Expand this last integral.

4-31. Use the identity of Problem 1-30.

CHAPTER 5

RELATIVITY I
The Basic Concepts

We have now studied at quite some length the fields of stationary electric
charges. At this point we can either go on directly to the fields of moving
charges, or we can leave the subject of electromagnetism aside, for the mo-
ment, and study the basic concepts of relativity. The reason for this digression
is that magnetic fields result from relativistic transformations of electric fields.
The longer path is more interesting, as always, but it may not be the
better one. There is no doubt that relativity throws much light on the nature
of magnetic fields. It also leads directly to the fundamental equations of
electromagnetism. But Chapters 5 and 6 do not replace the more conventional
approach that begins with Chapter 7. Selecting one path or the other is a
matter of time and personal taste.
Therefore, if you wish to go on with electromagnetism without delay, you
can omit Chapters 5 and 6, and go directly to Chapter 7 without losing conti-
nuity.
The present chapter is devoted to the basic concepts of relativity, with
little reference to electrical phenomena except near the end. In the next
chapter we shall utilize these concepts, first to reveal the origin of magnetic
fields, and then to establish several fundamental relations that we shall later
rediscover without using relativity.
We shall keep our discussion of relativity as simple as possible. In fact,
the only mathematical requirement for this chapter is elementary differential
calculus and the vector analysis of Chapter 1.*
* For a more detailed introduction to the basic concepts of relativity, see Edwin F.
A. Wheeler, Spacetime Physics (W. H. Freeman and Company, San
Taylor and John
Francisco, 1966).
194

Figure 5-1. Two Cartesian coordinate systems, one moving

at a velocity Ui with respect to the other in the positive
direction of the common x-axis. The two systems overlap
when the origins O, and QO» coincide. We shall always refer
to these two coordinate systems whenever we discuss rela-
tivistic effects.

We could start directly with the Lorentz transformation, which forms the
basis of special relativity, but this transformation is so contrary to everyday
experience that we shall first demonstrate the inadequacy of the more obvious
Galilean transformation.

5.1 THE GALILEAN TRANSFORMATION

The special theory of relativity is concerned with the observations made by

two different observers, one of whom has a constant velocity with respect to
the other. The general theory of relativity has to do with gravitation, and we
shali have no occasion to use it.
We therefore consider two Cartesian coordinate systems, as depicted in
Figure 5-1, where system 2 has a constant velocity Vi with respect to system 1
in the direction of the common x-axis. Corresponding axes are parallel and
in the same direction. Neither system is accelerated.
We shall constantly refer to these two particular coordinate systems
throughout this book whenever we discuss relativistic effects.
According to prerelativistic physics, the two systems are related by the
following intuitively obvious equations:
5.2 BREAKDOWN OF THE GALILEAN TRANSFORMATION 195

Xj = Xo -+ OF Yi = ya; Z1 = Zp. (5-1)

We have assumed that the two coordinate systems coincide at the time t = 0.
This set of equations constitutes a Galilean transformation.
Now, although the Galilean transformation is self-evident, and although
it is completely compatible with classical mechanics, it is not generally valid,
as we shall see, either for mechanical or for electromagnetic phenomena. We
shall have to use a more complex transformation that will reduce to Eqs. 5-1
for everyday mechanical phenomena.

5.2 BREAKDOWN OF THE GALILEAN

TRANSFORMATION AND OF CLASSICAL
MECHANICS AT HIGH VELOCITIES

Both classical mechanics and the Galilean transformation break down when
velocities approach the velocity of light. Many examples can be given.

Example PARTICLE VELOCITIES NEVER EXCEED c

It is found experimentally that the velocity of a particle does
not increase indefinitely as its energy increases.* Instead, the ve-
locity approaches asymptotically the velocity of light c = 3 x 108
meters/second. The usual expression (1/2)mv? of classical mechanics
for the kinetic energy of a mass m moving with a velocity v cannot
therefore be correct.

Example THE ADDITION OF VELOCITIES

According to the Galilean transformation, vector addition applies
to velocities. For example, if a passenger walks at a velocity v toward
the front of a train that itself moves forward at a velocity U, then his
velocity with respect to the ground is simply the sum of the two
velocities » + U. But this simple addition of velocities is found to be
incorrect in nuclear reactions where the velocities approach the ve-
locity of light. The resulting velocity is in fact always smaller than
v + UV, and never exceeds c.

Example TIME DILATION

The Galilean transformation assumes that the time rf, as measured

on the train, is the same as that measured on the ground, if the two

* The film ‘‘The Ultimate Speed; an Exploration with High Energy Electrons” by
W. Bertozzi demonstrates this phenomenon clearly. It was produced by Educational
Development Center, Newton, Mass.
196

Figure 5-2. The electric charge Q, is situated at the origin O2 and Q, is

situated at (x2, ye, O), also in 2.

observers have identical synchronized clocks. This also proves to

be incorrect at high velocities. In particular, it has been observed
that, for the high-energy mesons of the cosmic radiation, time flows
about nine times more slowly than the laboratory time. In other
words, a period of one microsecond as measured by the meson is
equivalent to about nine microseconds measured in the laboratory.
This phenomenon is called time dilation.*

3.3 INADEQUACY OF THE GALILEAN

TRANSFORMATION FOR
ELECTROMAGNETIC PHENOMENA

The Galilean transformation is totally inadequate for electromagnetic phe-

nomena,

Example THE TROUTON AND NOBLE EXPERIMENT

Consider two electric charges, as in Figure 5-2, Q, being situated
at the origin of reference frame 2, and Q, at (x2, ys, 0), also on frame
2. What is the force exerted by Q, on Q, as observed in frame 2.
Trouton and Noble attempted to observe this force in 1903.

* See American Journal of Physics, 37, 342 (1963). See also the film “Time
Dilation;
an Experiment on Mu-mesons” by F. Friedman, D. Frisch, and J.
Smith, produced by
Educational Development Center, Newton, Mass.
 197

Figure 5-3. In the Trouton and Noble experiment a mica capacitor C

(0.0037 microfarad) was suspended from a light phosphor bronze strip S.
One side of the capacitor was grounded by a platinum wire that dipped
in dilute sulfuric acid, while the other side was connected through the
suspension to a Wimshurst machine that operated at 2000 volts. The
light celluloid bulb B was covered with gilt paint and grounded to
eliminate electrostatic forces on the capacitor C. The mirror M served
to reflect a ray of light through the windows W onto a scale situated
One meter from C. The deflection of C was predicted taking into account
both the Earth’s orbital motion and the Sun’s proper motion. While the
predicted deflection varied from 0.0 to 6.8 centimeters, according to the
time at which the reading was taken, the observed deflection did not
exceed 0.36 centimeter and was unrelated to the calculated value.

They suspended a charged parallel-plate capacitor in such a way

that its plates were vertical and it could rotate around a vertical axis,
as in Figure 5-3. Since the Earth rotates around the Sun at an orbital
velocity of 3 & 104 meters/second, the force measured in the labo-
ratory is Fy.
The Galilean transformation assumes implicitly the existence
of an absolute, fixed frame of reference. If this assumption is correct,
we expect from elementary electricity that the charge Q, will be sub-
mitted not only to an electrostatic force but also to a magnetic force,
because the moving charges should produce magnetic fields. This
magnetic force should tend to turn the capacitor, bringing its plates
parallel to its velocity. Trouton and Noble were unable to observe
such a torque.

Example MAGNETIC FIELDS

Another example of the inadequacy of the Galilean transformation
for electromagnetic phenomena is simply the existence of a magnetic
field in the neighborhood of a conducting wire.
198 RELATIVITY I

We shall be able to show in the next chapter that the magnetic

fields of conduction currents are due to an exceedingly small change,
approximately one part in 10”, in the field of the moving electrons.
Magnetic fields can in no way be explained solely by Coulomb’s law
and the Galilean transformation.

Example THE JASEVA-JAVAN-MURRAY-TOWNES

EXPERIMENT*
Although the full potential of this experiment has yet to be achieved,
it already provides a highly sensitive method for detecting the effects,
if any, of the earth’s orbital velocity on the velocity of propagation
of light waves.
If there does exist an absolute fixed reference frame, then light
waves can be expected to propagate at the fixed velocity c with re-
spect to it, and it should be possible to measure the orbital velocity
of the earth. The hypothetical fixed medium of propagation is called
the ether.
The Jaseva-Javan-Murray-Townes experiment is a modern ver-
sion of the Michelson-Morley experiment, which was repeated many
times between the 1880’s and the 1930’s and which is described in
most books on relativity.
The modern version utilizes two lasers (Figure 5-4) set at right
angles on a rotating platform as in Figure 5-5. The two light beams
are mixed by means of a half-silvered mirror, and the beat frequency

Figure 5-4. Schematic diagram of a gas laser. The glass tube is

terminated by the two mirrors M; and Ms, and contains an appropriate
gas mixture, for example, a mixture of helium and neon. Energy
is
supplied to the gas by a de discharge between electrodes E. The
space
between the two mirrors constitutes a resonant cavity for the light
emitted by the gas, and a light beam comes out through mirror
M,,
which is half-silvered.

te S. Jaseva, A. Javan, J. Murray, and C. H. Townes, Physical Review

133A) 1221
 199

Figure 5-5. The Jaseva-Javan-Murray-Townes experiment. The half-

silvered mirror M mixes the light beams of lasers A and B, and the
photomultiplier and associated electronic equipment PM measure the
beat frequency. The support rotates so that A and B can be alternately
parallel and perpendicular to the orbital velocity of the Earth. The
object is to observe effects of this velocity on the velocity of propagation
of light with respect to the Earth. No effect has been observed so far.

is measured as in Figure 5-5.

Since a laser is a resonant cavity, its frequency
(Sat 'G oat nce
v=
NT play OMe: (5-2)
where L = n(\/2), is the distance between the two mirrors; n is the
number of half-wavelengths in the distance L; and c is the velocity
of light, as usual.
Let us assume that laser A is parallel to its velocity © with re-
spect to the ether. Then, in laser A, the time required for light to go
from one mirror to the other and back is
1 Thypi ae IEA Ybyi 1 " i= 2y7LA
» (5-3)
a Cc
0 co. “e (l—«ve)|
where
1
(5-4)
Y* = o/h”,
and laser A should oscillate as if its length were y?La, instead of La.
With © equal to the orbital velocity of the earth, 3 X 104
meters/second, (U/c)? is equal to 10-8, and v4 is smaller than if U
were zero by one part in 10°.
Laser B is perpendicular to the direction of motion and, from
Figure 5-6,
= Die = 2L23 Cc mel 2yLa (5-5)

eee ce OC
200

Figure 5-6. When laser B is perpendicular to the orbital velocity of the

Earth, light must travel through a distance 2L’z to make a round trip
through the tube. It is assumed that light goes from P to R at the
velocity c while the laser moves from Q to R at the velocity VD.
Then L’s/Lg = c/(c? — 0*)!/?.

Then laser B should oscillate as if its length were yLz, and the beat
frequency should be
Tom fn | 1
Vy = Yep—
Pa =
2 Ley e Ty :
(5-6)
If the complete set-up is now rotated through 90°, the beat
frequency should change by
NOG (elas —3)-(& go ES) 7
Avy = 2 UN ey Lay’ Lay’? Lay J’ Sot
Bee 7 al 1 1 1K peg1 2)
> EtE JG Jee : =o
where L is the average length of the lasers, and where the term (U/c)?
is obtained by expanding 1/7 using the binomial theorem.

(2) 9
Thus
Avy, ; 0 2

With v = 3 X 10" hertz, 2v, ~ 3 X 10® hertz. Since the fre-

quency stability of lasers can be of the order of 20 hertz, it should
be possible to observe a frequency shift even 10° times smaller than
the above 2y,.
It turns out to be exceedingly difficult to achieve the theoretical
accuracy, but it has been demonstrated that the frequency shift is
less than one thousandth of the effect predicted on the assumption
that light has a fixed velocity with respect to the ether.
5.5 INVARIANCE OF A PHYSICAL LAW 201

5.4 THE FUNDAMENTAL POSTULATE

OF RELATIVITY

The experiments of Trouton and Noble, and of Townes et al. had two points
in common. Both were designed to detect effects of the velocity of the Earth
through the ether, and both gave negative results. Many other experiments
that had the same objective also gave negative results.*
It is on the basis of such evidence that Einstein proposed in 1905 the
fundamental postulate of relativity, which can be stated as follows: it is
physically impossible to detect the uniform motion of a frame of reference from
observations made entirely within that frame.
This postulate is quite clear in itself, but it is so fundamental that we
shall state it in another way to emphasize its meaning. It means that any
experiment gives precisely the same result, whether it is performed in refer-
ence frame | or in reference frame 2, or whether it is performed in a standing
or in a moving vehicle, as long as it is not accelerated.
Then, if we are given some physical law like Fy = mq which is valid in 1,
there exists an identical law F, = mod, which is valid in frame 2. In fact, there
must exist a transformation, different from the Galilean transformation, that
renders all the laws of nature identical in frames 1 and 2.
We shall develop this concept of the invariance of a physical law in the
next section, using classical mechanics and the Galilean transformation as an
example.

5.5 INVARIANCE OF A PHYSICAL LAW AS

ILLUSTRATED BY CLASSICAL MECHANICS

The Galilean transformation leaves the laws, and hence the phenomena, of
classical mechanics unaltered in going from frame | to frame 2, and inversely.
In other words, the laws of classical mechanics are invariant under a Galilean
transformation. Classical, or prerelativistic, mechanics disregards the effects
that we shall discuss later, which become prominent when velocities become
comparable to the velocity of light.

* For a summary of these experiments see W. K. H. Panofsky and M. Phillips, Classical

Electricity and Magnetism (Addison-Wesley, Reading, Mass., 1955).
202 RELATIVITY I

Example CLASSICAL MECHANICS

INSIDE A MOVING TRAIN
If reference frame 1 is fixed with respect to the ground, and 2 is fixed
with respect to a railway train moving at a uniform velocity U, a
ball thrown vertically upward by a passenger sitting in the train
behaves, for that observer inside the train, precisely as it would if the
experiment were performed on the ground. The reason, of course, is
that the ball thrown vertically upward inside the train maintains its
horizontal velocity U throughout its trajectory. In fact, an observer
on the ground would see the ball in the moving train describe a
parabola.
Many other such experiments can be devised. For example, the
passenger can observe the period of a pendulum, or the acceleration
of a mass pushed by a spring, or the collision of billiard balls, etc.
In all cases the phenomenon as observed by the passenger in the train
is precisely the same whether the train is stopped or moving at a
constant velocity.
It is therefore impossible to design a mechanical device that
would permit the passenger to measure, or even to detect, the uni-
form motion of the train from observations made entirely inside the
train.
Of course, if the train accelerates, either by changing its speed
or its direction, the passenger observes inertial forces, which permit
him to measure his acceleration.

Example THE LAW F = ma

Let us write down the equations for a simple experiment that the
passenger performs, first when the train is stopped, and then when
the train is moving at a constant velocity U. We shall see that the
equations of motion in reference frame 1 are of the same form as
those in frame 2.
The passenger is given a mass mm on which he exerts a known
force F in the direction of the track by means of a calibrated spring.
As previously, reference frame 1 is fixed with respect to the ground
and 2 is fixed with respect to the train. We assume that the law
F = ma applies in frame 1, and we deduce the corresponding law
in frame 2.
When the train is stopped we can assume that the two frames
overlap; then x; = x. and

@ d2
F=m aia ae a5 (5-10)

Then F = ma applies in both systems.

5.5. INVARIANCE OF A PHYSICAL LAW 203

When the train has a uniform velocity U, x; = x2 + Uf and

end ck
(5-11)
hac i,5 On <P) 1a
and the law F = ma again applies in both systems.
What if the force F is exerted in a direction perpendicular to
the direction of motion? We again have two similar equations:

Famay
2

(5-12)
with respect to the ground, and
2
ea ape? (5-13)

with respect to the train, since y; = ye according to the Galilean

transformation.
We therefore say that the law F = ma of classical mechanics
applies in both reference frames, or that it is invariant under a
Galilean transformation.

5.6 THE LORENTZ TRANSFORMATION

Toward the turn of the century, several persons discovered independently

that invariance of all physical laws could be achieved by using the following
equations of transformation:

fe kee UE Sy salts aa Vh
1 Tt = (erp T= (O/ory”
Yi = yo, yo =), (5-14)
41 = 2, 22 = Z1,

er oT= a acy
ae to +o (U/C?) Xe ee (Kh (0/ ce") ’

This is the Lorentz transformation. \t applies to the coordinate systems of

Figure 5-1, which are repeated in Figure 5-7.
Since we shall be using these equations frequently, we have rewritten
them in a more concise form in Table 5-1.
The Lorentz transformation forms the basis of special relativity, and it
has been confirmed in innumerable experiments. We shall spend the rest of
this chapter discussing some ofits strange consequences. For the moment, we
can immediately note a few of its more or less obvious features.
204

Vi

Figure 5-7. The Lorentz transformation refers to these two

coordinate systems. System 2 moves at a velocity Ui with
respect to 1 in the positive direction of the common x-axis.
The two systems overlap when O; and O, coincide. This
figure is identical to Figure 5-1.

(a) The Lorentz transformation reduces to the Galilean transformation if

we set the velocity of light c equal to infinity.
(b) The relative velocity © of the two systems cannot be larger than c,
for otherwise x, y, z, ¢ become imaginary in one system or the other.
(c) There are really only four independent equations, since the right-hand
column can be deduced from the left-hand one, and vice versa. You should
check this immediately.
This is a general rule: the relation between quantities in one frame and the
corresponding quantities in the other frame can always be expressed by either
one of two equations which are equivalent.

Table 5-1. The Lorentz Transformation

x1 = y[x. + Ute| x. = y(x1 — Uh]

yi = ya va
oie 22> 21
th = [te + (0/c?)xo] t = y[ti — (0/c*)x1]

eeee SS
= Teen
 205

y2

Xo 0 x5 = lo X4
t—=10) t,=0
(a)

Vi

SP paps |

x, =0 Xie lov |1
t; =0 t, =0

(b)

Figure 5-8. (a) A ruler is fixed parallel to the x-axis in reference frame 2 and, for
an observer in that frame, it extends from x2. = 0 to x. = 4, and hence it has
a length /. (b) At the time 4; = 0 observer 0, in reference frame 1 notes the
positions of the two ends of the same ruler. He finds that it extends from x; = 0
to x1 = //y, and hence that it is shorter than J). This is the Lorentz contraction.
Lengths parallel to the y and z axes are unaffected.

(d) The right-hand column is identical to the left-hand column, except

that the subscripts | and 2 are interchanged and that — is substituted for
=
This is also a general rule: if a quantity in one frame is known in terms of
quantities in the other frame, the inverse relation is obtained by interchanging
the subscripts 1 and 2, and changing the sign of VU.
The reason for this rule is simply that we can either consider that frame
2 moves at a velocity + with respect to 1, or that frame 1 moves at a veloc-
ity — with respect to 2.
The Lorentz transformation provides us with transformation equations
for the space coordinates and for the time. We shall now deduce from them
transformation equations for many other quantities.

5.7 TRANSFORMATION OF A LENGTH

Let us imagine that an observer 0, in reference frame 2 fixes a ruler of length

l on his x-axis, as in Figure 5-8a, so that its extremities are at x. = 0 and at
X. = I. The length h is the length ofthe ruler as measured in its own reference
frame and is called its proper length. What will be the length of this same
ruler for an observer 0; on 1?
206 RELATIVITY I

By observer we mean either a human being equipped with proper instru-

ments, or some device that can take readings either automatically or by
remote control. Observer 0; is stationary in reference frame | and observer
0» is stationary in 2.
Observer 0, determines the length of the ruler by noting the positions of
both ends at the same time t,. Imagine that he makes his observations at the
time 4; = 0 when the two origins coincide. Then 0; notes that the left-hand
end is at x; = 0, since x. = 0, t; = ft. = O at the left-hand end. He also notes
that the right-hand end is at

h = yo + Vtr), (5-15)
according to the Lorentz transformation. We must find fy.
Now we know that 4 = 0 and x2 = / for o2’s reading at the right-hand
end. Then
b= = (0/e yh, (5-16)
and
h = ybl1 — (0/eP] = hl — (0/c)]!? = h/y. (5-17)
Thus the ruler that is fixed in reference frame 2 and that has a proper
length /) according to observer 0, appears shorter by a factor of 1/y to observer
01, as in Figure 5-8b. In other words, a ruler moving in the direction of its
length at a velocity U relative to the observer appears to be shortened by the
factor 1/y = [1 — (U/c)*]!”. This is the Lorentz contraction. It is independent
of the sign of 0. Both observers agree on the correctness of this figure; their
disagreement bears on the validity of the measurements. Observer 0; maintains
that his measurement is valid because he has observed the positions of the two
ends simultaneously, namely, at ¢; = 0. But observer 0, maintains that, on the
contrary, 0;’s measurements were not simultaneous—that 0, first noted the
position of the right-hand end at the time 4, = — (U/c*)l and Jater noted the
position of the left-hand end.
Of course the Lorentz contraction would be precisely the same if the
ruler were fixed anywhere else on the x» axis.
If the ruler is fixed along the x-axis in reference frame 1, observer 0, finds
it shortened by the same factor 1/y, as in Figure 5-9,
Thus 0; tells 0, that o’s meters are too short, and Q tells 0; that o,’s
meters are also too short! This is not really contradictory,
because the two
comparisons are not really the same. They involve two different
pairs of
measurements,
This is a general rule: the proper length of an object, measure
d in a certain
direction, is always LONGER than the same length measur
ed in a frame of
reference moving in that particular direction.
 207

i Figure 5-9. The ruler is now fixed in

(b) reference frame 1.

What if the ruler moves relative to the observer in the direction perpen-
dicular to its length? It then appears to have its proper length hh.

Example THE APPARENT SHAPE

OF A RAPIDLY MOVING OBJECT
We have just seen that the Lorentz contraction has the effect of con-
tracting an object in the direction of its motion by the factor
[1 — (V/c)?]!/? when it moves at a velocity © with respect to the
observer. The other dimensions remain unchanged.
One would therefore expect that a photograph of an object
moving rapidly in a direction perpendicular to the line of sight would
simply show it to be contracted in the direction of its motion. In
fact, it is quite easy to show that the object appears rather to be
rotated by an angle @ = arc sin (U/c), as long as the angle subtended
by the object at the camera is small. If the object moves in another
direction, or if the angle it subtends at the camera is not small, the
apparent distortion becomes quite complex.
Figure 5-10 shows a camera taking a photograph of a cube of
side / moving at a velocity U. The photograph registers photons that
arrive at a given instant. If the cube is quite far away, photons from
the face CDFE all arrive at approximately the same time. Edges
CE and DF appear shortened by the factor 1/y, while CD and EF
appear to be of the same length as if the cube were not moving with
respect to the camera. Thus the square face CDFE appears on the
photograph as a rectangle, as in Figure 5-11.
Photons from the edge AB in Figure 5-10 leave the cube some
time previously at A’B’, where A’A = B’B is the distance traveled
by the cube during the time that the light takes to travel the distance
1, or (l/c). The picture therefore has the appearance shown in
Figure 5-11.
Figure 5-12 shows that this is the photograph one would obtain
208

Figure 5-10. A camera photographs a cube

moving at a velocity © perpendicular to
the line of sight. Photons from the edge
AB leave the cube at A’B’, so that the
camera sees the left-hand face as in
Figure 5-11.

Figure 5-11. The photograph shows

the edge AB of the cube displaced
to the left and the edges CE and DF
contracted by the factor 1/y.

Figure 5-12. If the cube of Figure

5-10 were rotated through an angle
Es 6 = arc sin (U/c), it would appear
on the photograph approximately as
in Figure 5-11. The solid angle sub-
tended by the cube at the camera is
to camera
assumed to be small.
5.8 TRANSFORMATION OF A TIME INTERVAL 209

if the cube were rotated through the angle 6 = arc sin (U/c). (If the
velocity U of the cube were equal to the velocity of light c, the photo-
graph would only show the left-hand face!)
Of course, the object would have the same appearance to an ob-
server as it does on the photograph, because the eye, like the camera,
registers photons that arrive together at a given instant.
It will be noticed that the appearance of the object is not exactly
the same as if it were rotated, because the face CDFE remains
normal to the line of sight.

5.8 TRANSFORMATION OF A TIME INTERVAL

Now imagine that observer 0, measures the duration of a certain phenomenon
that is fixed with respect to his frame and that starts at tf = 0 and ends at
ty = To. The time 7) is called the proper time. What will be the duration of
this same phenomenon for observer 0?
To perform his measurement, 0; has a single clock situated at the origin
O».. Observer 0; uses two identical and synchronized clocks, one at O, and
one at x; as in Figure 5-13. He chooses x so that his second clock is next to
that of 02 at the end of the time interval.

0; x; 0, x}

Figure 5-13. Observer 02 measures the duration of a phenomenon which occurs at

the origin Oy. He uses a single clock at O2 and finds 7) seconds. To measure the
duration of the same phenomenon, observer 0; uses two synchronized clocks, one
at x; = 0 where the time interval begins, and one at x1 = U(y7)) where it ends.
Upon comparing the readings of his two clocks, observer 0; concludes that the
time interval is y7) seconds, which is /onger than 7). Figure (a) shows the clocks
at the beginning of the time interval: all three show the same time. Figure (b)
shows the clocks at the end of the time interval. Observer 0; says that 02’s clock
runs slow.
210

Figure 5-14. The situation here is the inverse of that illustrated in Figure 5-13. The
phenomenon now occurs at the origin O, instead of O., observer 0; measures its
duration with a single clock, and 0 uses two clocks. The time interval measured
by the moving observer is again longer than 7) by the factor y.

Thus, for oi, the time interval begins at 4, = 0 and ends at

T, = y[To + (O/c?)x2]. (5-18)

But x. = 0; thus
To
(5-19)
Py i coer
Observer 0; therefore measures a time interval that is Jonger than o»’s and
concludes that 0's clock runs slow.
What if it is 0: who measures a time interval with a single clock as in
Figure 5-14? Then 0, finds that 0,’s clock runs slow.
In other words, a moving clock appears to run slow by a factor of Y
whenever its time is checked as above. This is the phenomenon called time
dilation (see footnote, Section 5.2).
This is also a general rule: the proper time interval between two phenomena
is always SHORTER than the same time interval measured by a moving
observer.

Example THE TIME READ ON

A RAPIDLY MOVING CLOCK
We have just seen that, if 0 uses two identical clocks at two different
points on 1 to measure a time interval that 02 Measures at a fixed
5.8 TRANSFORMATION OF A TIME INTERVAL 211

point on 2 with a single clock, then o, finds that his own time intervals
are longer than those of 0 by the factor y = 1/[1 — (0/e)?]1/2,
Now, would 0, arrive at the same conclusion if he used a single
clock and /ooked at o,’s moving clock, as in Figure 5-15? We shall
assume that 02’s clock, which is situated at the origin O, of reference
frame 2, is moving away from 0, who stays at O;. This means that Oz
is to the right of O,, or that tf, is positive.
Let us first imagine that observer o, has a set of identical syn-
chronized clocks along his x-axis. As 0»’s clock goes by each one of
these, the relation 4; = yfy holds.
But 0; is at the origin O; of his system, and the light from o,’s
clock takes some time to reach 0. Thus 0, takes his readings later
than 4.
Suppose 0; reads a time fy on o»’s clock. What time is it on o,’s
clock? Let us call this time t{. Then ¢{ is the above t; plus the time
required for light to travel the distance Uf; between o»’s clock and
ors:

f=nt—=1+@/olr, (5-20)
BGO) :
“ft — @/on” ey)
—
1+ (v/e) 2
Fest to > fo. (5-22)

Thus, if observer 0; /ooks at 0o’s clock when it is moving away

from him, he arrives at the above result and concludes that o.’s clock
is even slower than with the previous method of measurement. If 0,

Figure 5-15. Observer 0; uses a double mirror M to photograph simul-

taneously his own clock and the moving clock. The relation ft; = yf) holds
again, but the light from the moving clock does not reach the camera
until a later time t’.
212 RELATIVITY I

takes into account the time taken by the signal to arrive to him, then
he finds the previous result 4) = yh.
What if 02’s clock is moving toward 0,? Then the origin O» is to
the left of O,, and both 4 and fy are negative. Therefore the time f{
at O, when 0, reads f2 on 0»’s clock is
|ur| Uh
ti =e ty +- c th = é ’ (5-23)

and fr is also negative, since U < c. Thus

{=| Ve cee 5-24

‘lel * ey
Now o»’s clock appears to run fast (tj and f) are both negative) by
the above factor, but, again, if 0, takes into account the time that
the light takes to reach him, he finds that 4 = yfo.

Example JHE RELATIVISTIC DOPPLER EFFECT

FOR ELECTROMAGNETIC WAVES
The shift in frequency that is observed at a detector when it moves
with respect to a source of waves is called the Doppler effect. This
phenomenon is well known in the field of acoustics.
Imagine that we have a source of periodic electromagnetic
waves of frequency f, at O. and a detector at O;. What will be the
frequencyfz measured at O,?
This problem is really identical to the clock problem we have
Just discussed because the source can be considered as a clock that
beats periods 7, = 1/f., instead of seconds. Therefore, when the
source recedes from the observer at a velocity U, as in Figure 5-16a,
the apparent period at the detector is

T,
a(eueiays
= E a (v/c) ie > Ties (5-25)

and the apparent frequency at the detector is

fa rye
= 1/T, = ELee
+(0/e)a)" Torah (5-26)

If the source moves toward the detector as in Figure 5-16b, the

apparent frequency is

l +- CU/c)
fa= Fess = is (5-27)
An interesting feature of these equations is that they are always
valid, whether it is the source or the detector that moves. This is
because there is no way ofknowing which one it is that moves, accord-
ing to the fundamental postulate of relativity. The relative velocity
0 is a positive quantity in all three equations.
 213

Figure 5-16. The Doppler effect for electromagnetic waves. S is a source

and A, B, C are three detectors. At A the source recedes from the
detector and fg < f,. At B the source moves toward the detector and
fa > fs. At C the source moves at right angles to the line joining it to
the detector and f; < f.. The frequency shifts are the same in (a) and (b):
it is only the re/ative velocity that matters.

If the relative velocity forms a right angle with the line joining
the source to the detector, we have either the first or the second
equation below, depending on the reference frame in which the
angle is measured:

fa = Yfs (source) or hi= ft(detector). (5-28)

5.9 SIMULTANEITY

Imagine that observer 0; sees two events A and B that, to him, occur at the
same x-coordinate x4; = Xg1, and, at the same time, t4; = fs:. He maintains
that A and B are simultaneous. Are they also simultaneous for 02?
From the Lorentz transformation,

tas = tas — (0/7

C8)X45) = yl tey — (0/C) x1] = tee. (5-29)

Then two events that occur at the same value of x and that are simultaneous
for one observer are also simultaneous for another observer. Note that the two
events need not occur at the same place, because the y’s and the z’s can be
different.
What if the events do not occur at the same value of x? If they are
simultaneous for 01, then t4; = fz, but since x4; # Xm, they are not simul-
taneous for the other observer 0».
214 RELATIVITY I

In fact, the time interval between the two events, as seen from reference
frame 2 is
tre — tar = YW(U/c? X41 — Xz), (5-30)

and it can be either positive or negative, depending on the sign of x41 — Xz.

5.10 CAUSALITY AND

MAXIMUM SIGNAL VELOCITY

We have just seen that two events that are simultaneous in frame | are not
simultaneous in frame 2, unless they occur at the same x. We can go even
further: the order in which two events occur can be different in different
frames because
tee — tao = y[(tar — tar) — (0/C?xn1 — Xa1)], 6-31)
and the signs of tz. — t42 and tp; — tai can be different.
This is a disturbing result indeed: it appears to violate the principle of
causality, according to which a cause necessarily occurs before its effect.
For example, imagine that observer 0; throws a ball in the direction of
the x-axis, and the ball, after a flight of a few seconds, breaks a windowpane.
The Lorentz transformation surely cannot imply that the series of events,
starting with the throwing ofthe ball and ending with the strewn broken glass,
would occur backwards in time for certain observers.
Let us imagine two events A and B. Event A occurs at the origins O, and
O, at the moment they coincide. Event A is the cause of B, which occurs at
Xp at a Jater time fs), and in frame 2, at xp. Event B cannot occur before A
in frame 2. Then fs: must not be negative. Event A could be the throwing of
the ball, and event B the breaking of the glass. Event A causes event B
through some device, in this case it is the ball, which propagates in some way
the signal from A at a velocity 2, in frame 1. Then

tre = [tm — (0/c)xm] = y[tm — (0/c rts), (5-32)

= 7 ta[l — (1 0/c)). (5-33)
The factor y is positive; by definition, tg: is also positive. If »; and U
have different signs, then fp» is always positive and event B always comes after
A. However, if 2, and © have the same sign, we must always have

nv < c. (5-34)
Let us assume that », and U are both positive. We have already seen in Section
5-6 that ‘0 can be as large as, but never larger than c. So the above inequality
must be satisfied even when U = c. Then
5.11 TRANSFORMATION OF A VELOCITY 215

m1 SC. (5-35)
If 2, and © are both negative, we arrive at the same result, by symmetry.
In other words, a signal can never be propagated at a velocity larger
than the velocity of light. Otherwise, the principle of causality would be
violated, and certain observers would perceive some effects before their causes.

5.11 TRANSFORMATION OF A VELOCITY

If 02 observes that an object has some velocity 2», in the direction of the x-axis,
what is the velocity of this same object according to observer 0;? (The velocity
V2; need not be constant.)
The velocity 14,, by definition, is dx,/dt, and
dx dx; dt
V2 = dh = diy dt; (5-36)

where
dx;
pra= VS
dts
Pres + Ute)= ¥ (22 + V). (5-37)

To calculate dt./dt, we remember that f, is a function of both 4 and m,

and that x, is itself a function of 4. In more familiar terms, we wish to calculate
dz/dx on a curve z = ax + by, where y is a specified function of x:

if z= ax + by(x), then az=a+b a. (5-38)

Then
Sige (0/2)x] = ¥ [1 — (0/c?) m2],
2 = 177 (5-39)

and
Ue = V(vx + V)[1L — (0/c?)rr2]. (5-40)
Putting both 2, terms on the left-hand side and simplifying,
Vor ave : (5-41)
U2 =
1 + (v2 0/c’)

The velocity 1, is measured at the time h, and %z is measured at the corre-

sponding time fz.
by
Therefore the velocity in reference frame 1 is smaller than ?2 +
the factor 1 + (vz U/c’).
The inverse transformation is

Voz as eee,
i Lan (diz U/c*) (5-42)
216 RELATIVITY I

Example This result makes the velocity of light equal to c in both systems.
For example, if the moving object is a photon in frame 2, ve, is equal
to c, and 2, is also equal to c, for any value of U! Or, if 0 = c, then
Viz = c for any value of v2;.

What if the velocity is not along the x-axis? Let us find a relation between

_ nd i
dy»
(5-43)
hy ris a dt,
Paade
The relation between 2, and v2, will be of the same form, by symmetry.
We proceed as above:

dy, dy, dtp

Bee epee -44
OY edie wede dh aa)
and, since y; = yo,

=
dy dts
———— =
dt.
4, — = Vo == Ziel -4
Vly dty dt, Voy dt, Voy y[1 (V/c )r1 i (5 5)

Then
Viy

*y == ST — (0/0) .
aie
5-46

The inverse transformation is

dy = oe (5-47)
afl + (@x20/2)]
Similarly,
Vo
UT Ne =
0/2
Ee
Bey
5-48

Vos Mis (5-49)

~ a1 — (n0/2))
Table 5-2 shows all six equations.

Table 5-2. Transformation of a V.elocity

ee
ee ee eee
hoe Vox ap V Pabgee M2 — U

i‘ 1 af (W220 /c?) fr 1 — (v1.0 /c?)

Vy = Voy : = Viy

VfL + (020 /c%)] vy = @0/c?)]

Voz Viz
uz = Fy Va. =
v1 + (v2,0/c%)] y[1 — (v10/c?)]
ee
5.13 RELATIVISTIC MASS
217

Example An object whose velocity in reference frame 2 is purely along the y

axis has a velocity Vi + (v2,/y)j in frame 1:0, is simply the relative
velocity Ui of the two frames, and diy is smaller than 2», by the factor
1/y.

5.12 TRANSFORMATION OF AN ACCELERATION

We can transform an acceleration by proceeding again in the same manner.
For example,

_ dvi, _ Ariz ate

cach me dtidy
a
= {4 TES ORV} y[1 — (0/c2)vx], (5-51)

= TF On0/OP
Qoz

pas)
Table 5-3 gives the six transformation equations.

Table 5-3. Transformation of an Acceleration

Qrz
ar:=
y*[1 + (02,0/c*)]?
peerage Sey = ans
Ay ALL +E(sd/e))? + od
il A Voz0 }

ie =ALL + Cn0/e9 2
Zz =. Sa
a ES ‘ze a

cod
eS aaa a Zz

> Air
Qo
~ [1 — (@1e0/c?)]*
1 if 01,0 }

ey Al ae O/e) oe Sn
1 if 0120
ag, =
v= owen oe |

5.13 RELATIVISTIC MASS

We have seen that relativity requires that lengths, times, velocities, and
accelerations be transformed according to rules that are different from those
218

—VW Vv =

Figure 5-17. This idealized experiment permits us to find a relation between mass
and velocity. Observer 0, slides two equal masses A and B with equal and opposite
velocities along the x-axis. They collide and rebound elastically. We conclude
from observer 02’s measurements that m = m/[1 — (v/c)?]*/*, where » is the
velocity of the mass with respect to the observer.

of classical mechanics. We can therefore expect that many other quantities

will be affected. Let us find out how mass should be transformed.
Imagine the following experiment, which is illustrated in Figure 5-17.
Observer 0, has two identical masses A and B which slide in opposite direc-
tions along the x-axis with zero friction. He observes that their velocities are
equal and opposite. The two masses collide and rebound with velocities that
are again equal and opposite. The fact that they do rebound with equal and
opposite velocities demonstrates that the two masses are equal. We assume
that the collision is elastic.
Observer 02 sees 0; perform his experiment and measures the velocities
of the masses A and B. The velocities with respect to the two frames of refer-
ence are shown in Table 5-4.
Let us call m4 and mg the masses of A and of B in reference frame 2
before the collision. Both are unknown for the moment. Let us set M to be
the total mass in frame 2 at the time of the collision.
We assume first that the total mass of A and B remains constant and
equal to M in frame 2:
ma, + mp = M. (5-53)
During the collision, part of the kinetic energy is transformed into potential
energy because the masses are elastically distorted, but the total energy, and
hence the total mass, remains constant.
We also assume that there is conservation of momentum during the
collision, as seen from 2:

Mav, + mpdp = —MD. (5-54)

On the left-hand side we have written the sum of the momenta of A and of B
before the collision, 14 and vz having the values shown in Table 5-4. The
term on the right is the momentum during the collision; it is the total mass
M,
times the common velocity —O with respect to frame 2. Remember that
during the collision the masses have zero velocity with respect to reference
frame 1, and thus a velocity —v with respect to 2.
5.13 RELATIVISTIC MASS 219

Table 5-4. Velocities of Masses A and B

If we multiply the first of these two equations by VU and then add it to the
second, we find that
mg Ute
a OL: (5-55)

This ratio is always positive. Also,

ma = ae

Mp =[5 — (va/cP eo)

You can show that these two values of m4/mp are equal by squaring and
equating the right-hand sides. (Hint: do not substitute the values of v4 and
vp until you have simplified the equality.) Note that v4 and vp cannot both
be equal to zero simultaneously.
Then we can write that
leg Le
es :
ma = TE (oslo eth
ie a : (5-58)
[1 — (ve/eP)”
where 1m is called the rest mass. These are the masses of A and B as measured
by observer 02, with reference to whom they have velocities v4 and vz, re-
spectively.
More generally, the mass of an object moving with a velocity » with
respect to the observer is
=
oe
Mo iran BS 5-59
™ =1 @/ol®
The quantity m is the relativistic mass.
Note that, if the observer follows the object, » = Oandm= mo, whatever
be the velocity of the object with respect to some other observer.
220 RELATIVITY I

5.14 TRANSFORMATION OF A MASS

Observer 0; on reference frame 1 observes a moving object and finds that its
mass is 7. Observer 0, in reference frame 2 observes the same object, and he
finds a mass 7». What are the two relations between m, and my»?
‘For observer 0, the mass is
Mo
m
SS
re
SS
(v/cpp?
Se
(5-60 )

where 7m is the mass of the object when it is at rest with respect to the observer,
and 2 1s the velocity of the object with respect to reference frame 1. Thus

Mo
™ = To +b + yey Se
Now, from Eqs. 5-41, 5-47, and 5-48,

Vie + Uy 0 a tA + OP + Uy + vee (5-62)

c cy? 1 + (02,0/c?)]?
Upon simplifying we find that

m = y[1 + (v20/c?)]me. (5-63)

The inverse relation is
mM, = y[1 — (120/c?)|m. (5-64)
As usual, these two equations are equivalent: they both give the same value
for the ratio m,/mp.
Note that the y and z components of velocity do not affect the ratio
m/mMo. If ve. = 0, then m, = ym».

5.15 RELATIVISTIC ENERGY €

What is the physical interpretation for the product mc2? Dimensionally, it is
an energy. From our definition of m,

mc = moc? [1 — (v/c)*]?, (5-65)

= moc! 1 bs a+e5t ob
Lae Se
(5-66)
ys
I moc? + 5mon" + ;Mo a ~- SURES (5-67)

The term (1/2)mpv? is the kinetic energy of classical mechanics. The term
 5.16 THE FOUR-VECTOR r 221

mic” is an energy that is associated with the rest mass and that
is called the
rest energy.
The quantity mec? is the relativistic energy &, and the difference mc: —
MC?
is called the relativistic kinetic energy. The relativistic kinetic energy is equal
to (1/2)mv? for 0? « ce.

Example An excellent demonstration of the existence of this rest energy is the

- annihilation of electrons. This phenomenon is well known to nuclear
physicists: positive electrons combine with negative electrons as in
Figure 5-18 to give two gamma rays, each of which has an energy
equal to the rest energy of one electron, or 0.511 MeV (1. million
electron volts = 1.6 X 107} joule).

Figure 5-18. A positive electron e+

and a negative electron e~ are anni-
hilated to form two gamma rays hy.
If the kinetic energy of the electrons
is small, the energy 2h/v of the two
gamma rays is equal to the rest en-
ergy 2moc? of the two electrons.

Se OmUP erhOUR-VEBOLOR Tr

An event such as the emission of a flash of light is characterized by its position

x, y, z and by the time ¢ at which it occurs. Let us examine these four variables
more closely. We shall find they can be grouped into a single expression that
is invariant under a Lorentz transformation and that proves to be useful for
discovering other invariant quantities in mechanical and electromagnetic
phenomena.
Under a Galilean transformation the distance between two points
(Xas Vay Za) and (Xs, Ys, Zp) iS an invariant:

Yap = (Xa1 ae on a (Va — yo)? + (Zar a oi): (5-68)

= (Xan — Xo)” (Var — Yoo)” + (Za2 — 202). (5-69)

Note that the word invariant is not synonymous with the word constant.
In fact, if the point a were fixed in frame 1 and b were fixed in 2, r,, would be
a function of the time. The distance r,,, is said to be invariant because it has
the same numerical value in both frames.
222 RELATIVITY I

Now you can easily show that, with the Lorentz transformation, ra, is
not an invariant. However, there does exist a corresponding quantity that is
invariant under a Lorentz transformation. Imagine that a flash of light is
emitted at O, at the moment when the two origins coincide. The light prop-
agates in all directions at the velocity c in both systems, and
2 2 2 2 2
aa = ams == Bla aN a ( (5-70)
1

The light arrives at (%4, yi, 21) at the time 4 and at the point (x, yo, Zz) at the
time f,. Thus, in this case,

M+yitzai-— et =<6+y+a— eh = 0. (5-71)

More generally, the coordinates x4, 1, Z1, 4, and x2, ye, Ze, fe for any single
event are related by the equation

am+yitz—ce# = 6@+y¥%4+2- es, (5-72)

or
ni— ef = 3 - eh. (5-73)
You can check this quite easily. The quantity r? — cf is therefore invariant
under a Lorentz transformation.
Now this property of x, y, z, t follows directly from the Lorentz trans-
formation. Then, for any set offour quantities that transform like x, y, z, t, we
have a corresponding invariant quantity. We shall use this result on several
occasions.
Such sets of four quantities are called four-vectors.
By analogy with three-dimensional geometry, we specify the coordinates
of the event as being
Ceey ect):
where j = (—1)'”, and the magnitude of the four-dimensional distance
between the event and the origin is the Square root of the sum of the squares
of the components. Thus
r = (r,jet). (5-74)
The quantity (x? + y? + 2? — @#)'” is called the magnitude of the four-vect
or
r and can be considered to be the square root of the scalar product
of r by
itself;
PS
»
PPS x ye ee pe ae (5-75)
This quantity is invariant, as we have shown above.
 223

5.17 TRANSFORMATION OF A MOMENTUM

AND OF A RELATIVISTIC ENERGY.
THE FOUR-MOMENTUM
The momentum of a mass m moving at a velocity v is

p=mv (5-76)
1 hEaG oar
It is to be understood that v is the velocity of m with respect to the observer.
In reference frame 1,
Pi = m1, (5-77)
and it has the three components

Piz = M12, Py = My, Piz = Md. (5-78)

It is a simple matter to transform py,:

Pic = Miz = y[1 + (V22°0 /c?) |g Fescmuatt (5-79)

= YMA + V), (5-80)

= y[Pox + V(&2/c)], (5-81)
where & = m,c’ is the relativistic energy of the mass m in reference frame 2.
The inverse transformation is

Pz = Y[ Piz a U(E:/c’)], (5-82)

and
Pry = Pr (5-83)
Piz = Poz. (5-84)
It is even simpler to transform a relativistic energy:

& = me? = [1 + @2x0/c?)|mc?, (5-85)

=[& + Uprl, (5-86)
and
&2 = y[&1 — Up). (5-87)
It will be observed in Table 5-5 that p,, p,, pz, &/c? transform like x, y, z, t.

Table 5-5. Transformation of a Four-Momentum

Pu = Y [Paz =i U(E2/c)] Pz = Y[Piz = U(Ei/c’)]

Ply = Pry Py = Ply
Piz = P2z Po = Piz

Ei/c? = y[(&2/c?) + (0/c")p2z] E2/c? = y[(E1/c?) — (U/c")piz]

224 RELATIVITY I

Table 5-6. The Four-Momentum Pp

Coordinates x y z t
Corresponding variable Dz Py Dz &/c?

Components of r Ey, y i jet

Components of p Dx Py Dz J&/c

Squared magnitude ofr Me ye eR22 — Of? = pr? — cp?

Squared magnitude of p p. + pe + p2 — (82/c2) = p* — (82/c?)

These four quantities are therefore the components of a four-vector

P = (p, jé/c), (5-88)
which is called the four-momentum. See Table 5-6.
The squared magnitude of the four-momentum,

Ve nt Ol ey erie
9

must be an invariant.

Example THE RELATION &? = mic4 + p*c?

If one reference frame moves with an object whose rest mass is Mo,
and if in some other frame the object has a momentum p and a mass
m, then
Dp? — mc? = —mrc* (5-89)
and
& = m’ct = mict + pre?. (5-90)
This general relation is illustrated in Figure 5-19.
The second term p’c? is negligible when 262 = 82/(1 — Bb) <ls
or 26° < 1, where B is the ratio v/c. As expected, at low velocities,
& & me?. (6? <« 1) (5-91)
On the other hand, if y?62>> 1, or 62 ~ 1,
& & pe = mvc & mc?. (6? = 1) (5-92)
For a photon, mp = 0 and
& = pe. (Photon) (5-93)
The momentum of a photon is therefore

& hy tw fh
p , (5-94)
aes chee leet,
where / is Planck’s constant 6.626 X 10-34 joule second, v is the fre-
quency, c is as usual the velocity of light in a vacuum 3 X 108
meters/second, #4 = h/2r = 1.05 X 10-84 joule-second, w = 27»,
and X is the wavelength divided by 27.
 225

Figure 5-19. The relation between

the relativistic energy &, the rest
energy mc”, and the momentum p of
a mass m is illustrated by this right-
angled triangle: 2 = mo’c* + p?c?.
If p = 0, & = moc?; if mo = 0 asfor
Moc? a photon, & = pe.

5.18 TRANSFORMATION OF A FORCE

At the very beginning of the next chapter we shall need to transform a force
from one reference frame to another. This is fairly simple now that we know
how to transform a momentum. We use the relation

_ dp
Lire (5-95)

Let the force be F; in frame 1 and F, in frame 2:

F, a Agi a Nij _ Lik and F, = Xt + Yoj + Zok. (5-96)

Then
Ppl
xX, a
as tye
dt, Piz =
|
dtp [vl Pes To U(E2/¢ »)
2 \e
dt,
'
(5 97)

AD 2 vd 1
= —— 5-98
o dto He‘ Cc dto fs yl + (vor /c?)]’ ( )

d&
X + (0/c) oF
= S, (5-99)
1 + (02.0/c?)

Now, from Problem 5-30,

o = Xe, + Yovoy + ZyVo, (5-100)

2

since d&/dt, is the rate at which the energy & = mc” builds up under the
action of the force F,. Finally,

Xy = Xo - etre (Voy Y> + VozL2). (5-101)

c? + v2,0
226 RELATIVITY I

Table 5-7. Transformation of a Force

X= Xo+ ae (VoyYo + 0222)

Cc + Vor 0

ae Reta ae
+ y[1 + W20/e?)]
vA 41
* ¥f1 + Q2x0/c?)]

Sy2 po
1 C2 7 M10 (lyAh,
ly4l lzH1

Y;
Woy ’
> yf = (@u0/c?)]
wz,
vA :
~ y[l — @n0/c?))

See Table 5-7 for the complete set of transformation equations. It is

understood that the velocity of the point of application of the force is v in
frame 1 and x in frame 2.
Note that the transformations do not involve the coordinates of the point
of application. They involve only the velocities 1, v2, Vi.
We shall use these relations repeatedly in the next chapter.

Examples (a) Surprisingly enough, a force directed along the x-axis (Y =

Z = ()), in the direction of the relative velocity Ui, is the same in both
systems.
(b) Forces that are equal and opposite in one frame are not neces-
sarily so in another frame. They remain equal and opposite only
if their points of application have equal velocities.
(c) Whenever the force has a component perpendicular to the
x = axis, the transformation changes both its magnitude and its
direction. Then two forces that are collinear in one frame are not
necessarily collinear in another frame.

5.19 TRANSFORMATION OF
AN ELEMENT OF VOLUME

We now wish to transform an electric charge density and an electric current.

As a first step we shall transform an element of volume, after which we shall
 227

Figure 5-20. An element of volume having the form of a

small cube of volume /*) has a velocity v2, with respect to
frame 2. It appears to have a smaller volume dr2 < /% for
observer 02 and a different volume dr, for observer 0}.

discuss charge invariance. After that, it will be a simple matter to deduce the
required transformations.
An element of volume at rest in reference frame 2 has a volume dro
for observer 02. What is its volume for 0,? Let us assume that the element of
volume is a small cube of volume /j in 2.
The element of volume is affected by the Lorentz contraction only in the
direction of the x-axis. Then, for 0;,

dr, = Ii[1 — (0/c}”, (5-102)

ears (5-103)
of

Remember that we have assumed that the element of volume is at rest

with respect to reference frame 2. With this assumption the element of volume
transforms like a length parallel to the x-axis.
We shall have to deal with elements of volume that have a velocity with
respect to frame 2. Let us first set this velocity to be along the x-axis, and let
us call it v2, as in Figure 5-20.
Then, for 0»,
dro =
i) [1 — (vm/c)"]” (5-104)
and similarly, for 0,
dr = BE — (n/c),
- (5-105)
where 2, is the velocity of the element of volume with respect to frame 1.
228 RELATIVITY I

Now we have already calculated 2;, in Section 5-11:

= Von a 0) ,
(5-106)
Vie
~ 1+ (»,0/c)

Thus substituting this value for 1, and simplifying,

dr ==
aL — (Oye
Ln)
Gas ee
1 + (v2,0 /c?) 2
:
(5 107)

- dro .
(5-108)
Al + @x0/c?))]
This transformation is really the same as for a length parallel to the x-axis and
moving at a velocity v2, with respect to reference frame 2.
The inverse transformation is
dr
dr. =
¥(1 = @20/e)]
; (5-109)
where the element of volume now has a velocity 2, with respect to frame 1.
We have assumed that the velocity in reference frame 2 is along the
x-axis. What happens if it has components along the y- and z-axes? According
to the Lorentz transformation, y; = y2 and z, = Ze. Therefore distances
perpendicular to the x-axis are the same in both frames, and the y and z
components of velocity have no effect on the ratio dr;/dr2. Equations 5-108
and 5-109 are thus valid for any value of x, and of ws.

5.20 INVARIANCE OF ELECTRIC CHARGE

From the experience that we have gained to date with relativistic transfor-
mations, we might venture to guess that an electric charge of Qy) coulombs
that is stationary with respect to reference frame 2 would appear to carry
either Qyy or Qo/y coulombs for an observer in frame 1.
This is in fact wrong. Electric charge is invariant, and a charged body
carries the same electrical charge for all observers.
Possibly the most direct demonstration of the invariance of charge is the
fact that the charge-to-mass ratio e/m for a charged particle moving at a
velocity v is found experimentally to agree with the law

a ee (5-110)
e e

No

The elementary charge e therefore remains constant and equal to 1.6 X

10—® coulomb, irrespective of the velocity of the particle, while the mass m
 5.21 TRANSFORMATION OF AN ELECTRIC CHARGE DENSITY 229

varies with the velocity, as in Section 5-13. This relation is found to apply in
particle accelerators up to the highest energies attained to date.
Another convincing demonstration of the invariance of charge is the
fact that a metal does not acquire an electric charge when it is heated or
cooled, despite the fact that the average kinetic energy of its conduction
electrons is much /ess affected than that of its atoms.* One might expect, at
first sight, that the extra charge would be negligible, since the change in
velocity is small. However, as we saw in Chapter 2, the electric charges in
matter are enormous, and it is only because their fields cancel perfectly at all
temperatures that ordinary matter remains macroscopically neutral.

Example Ten kilograms of copper contain about 102° atoms and 1.5 x 107
coulombs of conduction electrons. If the charges on the atoms could
be modified by only one part in 10! by heating or cooling, the total
charge in ten kilograms of copper would be 1.5 < 10-8 coulomb.
If the copper were spherical, it would have a radius of about 6.5
centimeters and, assuming that the extra charge migrated to the
surface, the copper would have a potential of about 2 kilovolts.
Such electrostatic effects have never been observed.

5.21 TRANSFORMATION OF AN ELECTRIC CHARGE

DENSITY AND OF AN ELECTRIC CURRENT.
THE FOUR-CURRENT DENSITY

Consider an element of volume that contains dn electric charges of Q coulombs

each and that has a velocity ». with respect to reference frame 2.
We can show that the total charge Qdn in the element of volume is the
same for both observers 0; and 0». First, the individual charges Q are the same
for both observers, since Q is an invariant, as we have just seen. Second, dn
must also be an invariant, because it is simply a number of objects that both
observers can count. Then Qdn is invariant.
Then the two densities are

a=dn = Qdn
(5-111)
where dr; is the volume of the element of volume as seen by 0, and dr» is the
volume of the same element as seen by 0. Thus

* See for example, Charles Kittel, Introduction to Solid State Physics, Third Edition
(John Wiley, New York, 1966) p. 209 ff.
230 RELATIVITY I

pi = eee
T1
= pxy[1 + (220/c’)]. G-112)
The inverse relation is

p= ae
drs
= pry[1 — (uz0/c?)). (G-113)

The quantity pode, is the product of the electric charge density and of the
velocity of these charges in the direction parallel to the x-axis, as seen by
observer 05. This is analogous to the product pv, which we used in Section 1.5
for the mass flux in water. The product p., is the electric charge flowing per
second and per square meter at the point considered, in the direction of the
x-axis, or the electric current density /2, in amperes/meter? as observed by 0»:
hoe = poz. (5-114)
Similarly,
Siz = pie. (5-115)
Then our two equations for p; and p, can be rewritten as follows:

pi = Y[p2 + (0/c*)Joz], (5-116)

p2 = ylpr — (V/c*) Ae]. (5-117)
These are the equations of transformation for the electric charge density.
We can now deduce the equations of transformation for electric current
densities:
Jizg = pies (5-118)

= pry[1 + (22,0/c?)]
/ 2
It +Vor@,0/2)’
+ 0) me
(5-119)

= (oz + Up»), (5-120)

and
Jox = Wiz Pax Up). (5-121)

If the charges were fixed with respect to frame 2, Jo, would be zero and dis
would be Up.
We have therefore found transformation equations for the electric charge
density and for the electric current density parallel to the x-axis. To be com-
plete, we require corresponding equations for Jiy, Joys Jz, Jo. We need only
find a relation between Jj, and Jy,:

Jiy = prdtys (5-122)

Vo
=
poyll + (r2.0/c*)] Wl + n/a)
6 om if 2 “y
= 7!
—_ oo
(5-123)
:

and, by symmetry,
Jig = Joa (5-124)
See Table 5-8.
5.21 TRANSFORMATION OF AN ELECTRIC CHARGE DENSITY 231

Table 5-8. Transformation of a Four-Current Density

ee es ee ees as SL ie.
Siz = (Jor ae Up») ta Viz — Up)
Jy =F Voy Jy = dh
Ji ad Jo2 Jo — Jie
pi = Y[p2 + (O/c*)Joz] p2 = y[p1 — (U/e»)iz]
a
eeeEeeeeeSSSSSSSSSSSSSFSSSSSSFMFFFssee

Examples If we have fixed charges with a density p» in frame 2, an observer

in frame 1 observes a charge density yp: and a corresponding current
density Upi, or Uypo.
If we have a conducting wire in frame 2, p2 is zero, and the
current density Jo, is associated with zero net charge density. If the
wire is parallel to the x-axis, and if the current flows in the positive
direction of the x-axis,

pi = (V/c)\(yJoz), Jie = Yor. (5-125)

Thus, contrary to what one would expect, Ji; is not equal to Up.
You will be able to solve this paradox if you solve Problem 5-35.

It will be observed that the equations of transformation for J and p are

of the same form as those ofthe Lorentz transformation, with the components
of the current playing the role of the coordinates x, y, z, and with the charge
density playing the role of the time.
The corresponding four-vector

J = VJ, jcp) (5-126)

is called the four-current density, and its squared magnitude
J? — Gps

is an invariant (see Table 5-9).

We shall use this four-vector shortly to state one of the fundamental
postulates of electromagnetism, but we must first be able to calculate a di-
vergence in four dimensions.

Table 5-9. The Four-Current Density J

Coordinates ie y Zz t
Corresponding variable Af As ie p

Components of r 36 y Z jet
Components of J Ak: If, Ap jcp

Squared magnitude of r x? + + z — f? = rp? — cf?

Squared magnitude ofJ Jz + Jy + Jz — c?p? = J? — cp?
232

Table 5-10. Transformation of the Partial Derivatives

cae loa Cale orga oe qre eal

0 )

0 0 ) 0
ay : OY2 Oye 7 ay

paoO22
OZ;
eee
OZ, OZ

0 0 0 0 0 0
2a Man Deeikpo tral

5.22 THE FOUR-DIMENSIONAL OPERATOR []

By analogy with the four-vector (x, y, Zz, jet), we expect to find that the com-
ponents of the four-dimensional operator (.) (occasionally called “‘quad’’),
which is the equivalent of the V operator, will be 0/dx, 0/dy, 0/dz, (1/jc)(0/dt).
This is in fact correct. The transformation equations for the partial derivatives
will indicate to us which are the proper components:

Ao ¥ - < — = | — (0/e*) Pal (5-127)

— = oe (5-128)

:~ - (5-129)

fs_ fei:" a = a | —_ = | (5-130)

These equations are tabulated in Table 5-10.

These transformation equations are of the same form as the Lorentz
equations of Table 5-1 if the operators corresponding to the four components
x, y, Z, t are selected just as we expected (Table 5-11).
Then
ie (v, 3). (5-131)
Formally, the operator [] is used much like the V of Chapter 1. It can
be used for calculating a gradient

co) Ae ete) oes: 1h aa) ; :

(JF = (5-132)
(2 dy OZ jc =| F,
5.23 THE CONSERVATION OF CHARGE 233

Table 5-11. The Four-Dimensional Operator [_]

Coordinates Ss y Zz t
Corresponding operators 0/0x 0/dy 0/dz —(1/c?)(0/02)

Components of r x, y zz Jct
Components of [] 0/0x 0/dy 0/dz (1/jc)(0/0t)

Squared magnitude ofr x? + y2 + 22 — cf? = r? — cp?

fal? (0/dx)? + (0/dy)? + (0/0z)? — (1/c?)(0/at)?
= V? — (1/c’)(0/0t)?

or a divergence
_ 0a; e da, , | da;
(UratS sy 7 =U az oe aie (5-133)

The scalar product of (_] by itself is called the d’Alembertian and is written (_}?:

2—
0? 0? 0 =
16? =
ego ay Ton” Gar eae
Sy 1 2
Aas (5-135)

Also
faye eeClotVie (5-136)
Setting the d’Alembertian equal to zero gives the wave equation corre-
sponding to a phase velocity equal to the velocity of light c.

5.23 THE CONSERVATION OF CHARGE

Let us calculate the divergence of the four-current ieoae:

O-d Je asau lasSJ, +o (5-137)

I V+ n (5-138)

This divergence is important. Let us see what it means, physically.

Consider a small volume +. The rate at which the charge rp enclosed in the
volume 7 increases with time is r(dp/01). The rate at which the enclosed charge
decreases with time is 7(V-J), since the current density J is the charge flowing
out per unit time and per unit area as in Section 1.5.
234 RELATIVITY I

Now, in all experiments so far, charge has always been found to be

conserved. Then
AN pee e 25) (5-139)
or
V-J=——:dp 5-140

This is the Jaw of conservation of charge. It says that, in any given frame of
reference, electric charge is neither created nor destroyed. The law of conser-
vation of charge can also be written as
()-J = 0. (5-141)
This law must not be confused with charge invariance. Charge invariance
means that the electric charge carried by an object is independent of the
velocity of the object with respect to the observer. In other words, that the
charge is the same in all frames of reference.

5.24 SUMMARY

We constantly refer to the two reference frames of Figure 5-1 (or 5-7), where
U is the velocity of reference frame 2 with respect to reference frame 1—in the
positive direction of the common x-axis. It is also the velocity of reference
frame | with respect to 2, in the negative direction of the common x-axis.
According to classical mechanics, it is possible to transform one set of
coordinates into the other by means of the obvious relations

xX = X. + Uf, Vi = yo, Z; = Zo. (5-1)

This set of equations constitutes a Galilean transformation. The Galilean

transformation is valid for mechanical phenomena when the velocities are
much smaller than the velocity oflight. It is not valid for electrical phenomena,
even at low velocities.
On the assumption that the Galilean transformation was correct, many
experiments were performed with the object of measuring the velocity of the
Earth with respect to the ‘‘ether.” The failure of all these experiments led to
the fundamental postulate of relativity, which can be stated as follows: Jt is
physically impossible to detect the uniform velocity of a frame of reference from
observations made entirely within that frame,
This fundamental postulate means that all the laws of nature must be
invariant when transformed from reference frame 1 to reference frame 2.
5.24 SUMMARY 235

This can be achieved if the Galilean transformation is replaced by the

Lorentz transformation (Table 5-1);
X= y[%2 + Vt], Xx, = y[% — Vi],
i aa 2s Ue Bar
41 — 29, Zo = 24,

t= [te + (0/c?)x2], te = y[h — (0/c?)x),

&, l| AB ie 00709 line c 3 X 108 meters/second.

TRANSFORMATION OF A LENGTH PARALLEL TO THE X-AXIS. THE LORENTZ CON-

TRACTION.

Object of proper length Object of proper length

at rest in 2: at rest in I:

Te oop ae
af af

TRANSFORMATION OF A LENGTH PERPENDICULAR TO THE X-AXIS. No change.

TRANSFORMATION OF A TIME INTERVAL. TIME DILATION.

Proper time interval 7) measured Proper time interval 7) measured

at position of a single clock at at position of single clock at
rest in 2, and rest in 1, and
T; measured with two clocks at T2 measured with two clocks at
rest in |: restin 2:
Ty = T) ST. Tz, = yIy > To.

RELATIVISTIC DOPPLER EFFECT FOR ELECTROMAGNETIC WAVES. The velocity U

is the velocity of the source relative to the detector.

(a) The velocity is along the line joining the source to the detector

ae Reson (5-25)
Nile (O/o)| 4" (5-26)
where fy is the frequency measured at the detector, and f, is the frequency
measured at the source. The velocity U is taken to be positive when the source
recedes from the observer.
(b) The velocity is perpendicular to the line joining the source to the
detector
Sa = rf, (source) leu Pe: tyf. (detector). (5-28)
236 RELATIVITY I

SIMULTANEITY. Two events that are simultaneous in one frame of reference

are also simultaneous in other frames of reference if and only if their x
coordinates are the same.

MAXIMUM SIGNAL VELOCITY. It is impossible to transmit a signal at a velocity

that is larger than the velocity of light.

TRANSFORMATION OF A VELOCITY (SECTION 5.11).

Y Vor == © » pe OES wa D

: 1 + (vo:0/c?) aia) OS (v1,0/2)

Vy = Poy ’ Oi = os
© YES @e0/c)] ~~ yl (eG /e*))
Voz Viz
Viz
ey =,
+ On0/A) , 3
=™ ~ afl — (z0/2)]
\2z ae f 9 .

TRANSFORMATION OF AN ACCELERATION (SECTION 5.12).

a _ 2x

ALT On0/2)]*
l ==; fi
\do — = Vo,2y U 7 1
OS La Gate 1 2 2 ea
a, =

1 V9, 0 |
Q, = in aarp ae =. PS
{ iG -- Vor VU ay J
: [1 == (dex /c?)]?

Qo. = Air
22 fl a (v420/c)]*”

cae 4 Ao
Vy O
rg |
uN

¥ =~Vlae — @nb/O? {ay+a

I
Oy =

l a fa : aV120 .
Q2, = or
(20 / ec) \ s 5 Cc? — 4,0 ay ,|
y [1 =

RELATIVISTIC MASS. The mass m of an object moving with a velocity v with

respect to the observer is

Pa ee ee
l — w/o” 5-59
ee
where mp is the rest mass.

TRANSFORMATION OF A MASS.

m, = [1 + (v220/c*) |e, (5-63)

m:, = y[1 — @20/c?)]m. (5-64)
5.24 SUMMARY 237

RELATIVISTIC ENERGY.

&=me = fl — @/o]* Cy (5-65)

= moe? + 3 my + § mout/c? + +>, (5-67)

where mc? is the rest energy and mc? — myc, or (1/2) my? + ---, is the
relativistic kinetic energy.

THE FOUR-VECTOR Ir = (r, jcf).

Components: x, y, Z, jct
Invariant squared magnitude: x? + y? + 2 — ef, or Pr — er

RELATIVISTIC MOMENTUM.

p=mv= Mo
(5-76)
[I — @o/cp}?”
TRANSFORMATION OF A MOMENTUM AND OF A RELATIVISTIC ENERGY (SECTION

5.17).
Piz = Y[P2x + U(E2/c’)] Pr = [Pre — U(Ei/c*)],
Py = Pr Py = Py
Piz = Px P22 = Pu
1/2 = y[(S2/2) + (0/C)por] &o/c? = y[(Ei/c?) — (0/c)pis]

THE FOUR-MOMENTUM P= = (p, /&/c).

Components: De Dus Deo) C
Invariant squared magnitude: p? — &/c? = p? — mc? = — moc? (5-90)
For a mass m,
& = moc! + pre’. (5-90)
For a photon,
pe t= =—=2 (5-94)

where h is Planck’s constant 6.626 X 10~*4 joule second, » is the frequency,

c = 3 X 108 meters/second, 4 = h/2x = 1.05 X 10-* joule second, w =
2rv, and X is the wavelength divided by 2r.

TRANSFORMATION OF A FORCE (SECTION 5.18).

<0)
Xy i Xe ap
eee!
C2 + Vo2°0
(Voy Yo + V2,L2), Xo = Xi
am go et + Zn),
238 RELATIVITY I

Moyes 2 yj oa
yl + (ox0/c’)) > yl = (@20/e)]
Zi a Zi Ai
~ fl + Cac0/2))

TRANSFORMATION OF AN ELEMENT OF VOLUME.

dr, =
dr»
(5-108)
y[1 + (e20/c*))"
dry =
dr
(5-109)
vL1 = (20/2)

INVARIANCE OF ELECTRIC It is found experimentally that the electric

CHARGE.
charge carried by an object is the same for all observers.

TRANSFORMATION OF AN ELECTRIC CHARGE DENSITY AND OF AN ELECTRIC

CURRENT. (SECTION 5.21).

Diz = [Jor = Upel, Joz = [Siz — Voi],

Sy = Joy, Joy = ee

Jz = Jo, Jo, ay Digs

pi = [2 + (0/C’\oe], D i)
ll yer — (0/c*) Az].

THE FOUR-CURRENT DENSITYJ = (J, jcp) (SECTION 5.21).

Components: Tiss, Jius Sigs Jcep

Invariant squared magnitude: J? — cp?

TRANSFORMATION OF THE PARTIAL DERIVATIVES (SECTION 5.22).

oO a 0 2 =9 0 0 Re
Ox, SW E (w/e) } cs v[& + (o/c

od Bi ie
dV1 dy.’ OVe oy,

GEE AG 6aaa
0%, OZ OZ» 7 Oz,
PROBLEMS
239

THE FOUR-DIMENSIONAL OPERATOR C] = (v = ai)

J

Components:
oleic
dx dy dz jc dt
Four-dimensional gradient:

ole. Tom lio

Fe > ’ ’ -
0 (2 Oy OZ jc a) (5-132)

Four-dimensional divergence:

anh = da x da y da z 1 da t x 25
Ox ap oy a Oz ie or O19)
Four-dimensional Laplacian, or d’Alerzbertian:

fe (v:Ea
COL
(5-136)
LAW OF CONSERVATION OF CHARGE. It is found experimentally that, in any
given frame of reference, electric charge is always conserved:

V-J=——%0ot (5-140)
or
O-J =0. (5-141)

PROBLEMS
Several of the following problems are adapted, with permission, from a book by
Edwin F. Taylor and John A. Wheeler.*

5-1. For what value of U does the value of y differ from unity by 1%?

5-2. Draw a graph of y as a function of U. Try to find “‘the’’ best way of conveying
the manner in which y depends on V.{

5-3. Two events occur at the same place in the laboratory and are separated in
time by 3 seconds.
(a) What is the spatial distance between these two events in a moving frame
with respect to which the events are separated in time by 5 seconds?
(b)What is the relative speed of the moving and laboratory frames?

5-4. Three men, A, O, and B, ride on a train moving at a velocity VU. A is in front,
O is in the middle, and Bis at the rear. A fourth man, O’, stands beside the

* Spacetime Physics (W. H. Freeman and Company, San Francisco, 1966).

240 RELATIVITY I

rails. At the moment O passes O’, light signals from A and B reach O and O’.
Both O and O’ are asked who emitted his light signal first. What do they
answer ?

. Observers 0; and 02 repeat the measurement of the length of the ruler (Section
5.7), but in a different way. Observer 0» fixes to the ends of his ruler a pair of
flash bulbs that can project the shadow of the edge of the ruler on a photo-
graphic plate in reference frame 1. Midway along the ruler, he sets up an elec-
tronic circuit that can send pulses in both directions to flash the bulbs simul-
taneously (according to him).
(a) What will be the distance between the edges of the ruler on the photo-
graphic plate?
(b) How can 0; account for this result?

5-6. A straight line passing through the origin O, of reference frame 2 forms an
angle a» with the x-axis.
(a) Calculate the value of a, as measured by an observer 0; on reference
frame 1.
(b) What is the value of a; when the velocity U of reference frame 2 with re-
spect to 1 approaches c?

5-7. A physicist is arrested for going through a red light. In court he pleads that
he approached at such a speed that the red light appeared green to him. The
judge, a graduate of a physics class, changes the charge to speeding and
fines the defendant one dollar for every kilometer per hour he exceeded the
speed limit of SO kilometers per hour. What is the fine? (Agreen & 5.3 X 1077
meter, Area & 6.5 X 1077 meter).

5-8. The radio galaxy 3C295 has a red shift of 46°%. The astronomers mean by
this that the observed wavelength is 1.46 times the wavelength of the same
radiation produced in the laboratory.
(a) Calculate the radial velocity of this galaxy.
(b) Some quasars have red shifts of 200°%. What is their radial velocity?

5-9. The twin, or clock, paradox can be illustrated as follows. On their twenty-first
birthday, Peter leaves his twin Paul behind on the Earth and goes off in a
straight line for seven years of his time at a speed of 0.96c. Peter then reverses
direction and returns at the same speed.
(a) What are the ages of Peter and of Paul at the moment of reunion?
(b) Peter and Paul, expecting this strange result, performed the following
experiment during Peter’s trip. They both observed a distant variable star
whose light alternates from dim to bright at a frequency f when observed from
the Earth. The variable star is ina direction perpendicular to Peter’s trajectory.
They of course both counted the same number of pulsations during the trip.
Use the expression for the Doppler shift to verify the difference in age between
Peter and Paul at the end of the trip.

5-10. It has been observed that some quasars exhibit light fluctuations with a period
of about one day.
Can you infer an upper limit for their size?
PROBLEMS 241

5-11. The Lorentz transformation implies that the relative velocity U of two frames
of reference cannot exceed the velocity of light c. We have also shown that a
mass and a signal cannot exceed the velocity of light. Discuss the following
cases.
(a) A very long straight rod, which is inclined at a small angle @ in relation to
a horizontal axis, moves downward at a velocity v.
What is the speed of the point of intersection of the lower edge of the
rod with the axis?
Can this speed be greater than c?
Can it be used to transmit a signal?
(b) The same rod is initially at rest with the point of intersection at the origin.
The rod is struck a downward blow at the origin with a hammer,
Can the motion of the point of intersection be used to transmit a signal
at a velocity greater than the velocity of light?
(c) A powerful laser is rotated rapidly about a vertical axis.
Can the azimuthal velocity of the beam exceed the velocity of light?
Can the beam be used to transmit a signal between two points at a veloc-
ity greater than c?
(d) The manufacturers of some oscilloscopes claim writing speeds in excess
of the speed of light.
Is this possible?

. We have found six formulas for calculating the velocity components in one
frame when the velocity components in the other frame are known. Show that
they can be written in the following vector form:
v= [v2 ==) == (vw, /y))/01 ak (vo: V/c?)].

. Light moves more slowly through a material medium than through a vacuum,
its phase velocity » being c/n, where v is the index of refraction of the medium.
If now the medium itself moves at a velocity U « c with respect to the labo-
ratory, show that the phase velocity of the light with respect to the laboratory
is approximately
1
¢ ++ 1b) (1ee +)
n n

. Aring of electrons rotates about its axis of symmetry and also moves parallel
to itself in the direction of this axis.
If v, is the axial velocity of the ring with respect to the laboratory, and
if vp is the azimuthal velocity of the electrons in the reference frame of the
ring, show that, with respect to the laboratory,

(a) 6? = 62 + 63 — 6:63,
(b) y = Ye.

5-15. A flash of light is emitted at an angle a» with respect to the X-axis.

(a) Show that
rein euaeeeenics ya!
an O41 = [cos a2 + (U/c)]’
242 RELATIVITY I

(b) Show that, as U approaches c, the angle a is small, except near a, = 7.

In frame 1 the light is then concentrated in a narrow forward cone. This is
called the headlight effect.
(c) Show that, for small angles and for 0 <c,

a= ov
a2 ak Cc

5-16. Show that the ratio m14/m1pz of Eq. 5-55 is always positive.
Sell7/, Show that the two values of the ratio m4/m, given in Eqs. 5-55 and 5-56 are
equal.

5-18. A nuclear bomb is exploded in an underground cavity and the products of

the explosion are allowed to cool.
Sketch graphs of » mec? and SS, moc? as functions of the time.

sril®), An atom of hydrogen is accelerated up to a very high velocity.

Show that the ratio of the kinetic energies of the proton and of the elec-
tron is equal to the ratio of their rest masses.

5-20. A proton has a kinetic energy of 500 million electron-volts.

(a) Calculate its mass.
(b) Calculate its velocity.

5-21. Draw a table of relativistic kinetic energies, expressed in electron-volts, for

electrons and protons that have velocities of 0.1c, 0.3c, 0.9c.
Since both the electron and the proton carry one electronic charge, the
energy * electron-volts is equal to the accelerating voltage.

5-22. A cosmic-ray particle, which may be assumed to be a proton, has been ob-
served to have an energy of 16 joules.
(a) Calculate its mass in micrograms.
(b) How long would this particle take to cross our galaxy (diameter 10° light
years), as measured by a clock moving with the proton? Express your answer
in seconds (1 year + 7 X 10’ seconds).
5-23. The Stanford Linear Accelerator is used to accelerate electrons up to energies
of 40 GeV (40 10° electron-volts).
(a) Calculate the mass of an electron that has the full energy. How does this
mass compare with that of a proton at rest (mass 1.7 X 107-27 kilogram). An
electron at rest has a mass of 9 X 10°54! kilogram.
(b) What is the length of the accelerator in the reference frame of an electron
that has the full energy? The length of the accelerator, as measured on the
ground, is 3000 meters.
(c) What time would be required for such an electron to go from one end
ofthe accelerator to the other, (i) in the laboratory frame, (ii) in the electron’s
frame of reference?
(d) What is the velocity ofthis electron, according to its own measurements?
5-24. It is shown by the example in Section 8.9 that the net outward force on an ion
PROBLEMS 243

that is at the periphery of an ion beam is smaller than the electrostatic force
of repulsion by a factor of 1/72, U being the velocity of the ion.
Show that, if the kinetic energy of the ion is equal to its rest energy, the
electrostatic force of repulsion is reduced by a factor of 4.
5-25. Visible light can be transformed into high-energy gamma radiation in the
following way.
Head-on electron-photon collisions are produced by reflecting a ray of
light from a laser backwards on a high-energy electron beam. Let us assume
that the photons have an energy hy of 2 electron-volts and that the electrons
have a kinetic energy of 6 X 10° electron-volts.
Let us disregard the electron recoil, for the moment. In the reference frame
of the electrons, the incident photons are Doppler-shifted to hv’ >> hv, and then
reflected forward at the same energy hv’. With respect to the laboratory, there is
a further shift to hv” >> hy’.
Calculate the final photon energy, using conservation of both momentum
and energy in the laboratory frame.

5-26. Since the thrust produced by a rocket motor is equal to the product m’v, where
m’ is the mass of propellant ejected per second and 2 is the exhaust velocity,
the ultimate rocket would transform its propellant into radiation and eject
photons backward at the velocity of light. The weight of the propellant would
then be minimum.
(a) Show that the power-to-thrust ratio W/F for a photon motor is c.
(b) Show that the thrust F is c(dM/dt), where M is the mass of the propellant
remaining at the time f.
Note that these two relations are independent of the frequency: the
source of radiation need not be monochromatic.
According to this last equation a photon rocket burning one gram of
matter per second would have a thrust of 30 tons. The difficulty is to trans-
form an appreciable fraction of the propellant mass into radiation, as the
following example will show.
(c) An ordinary flashlight has a capacity of about 2 ampere hours at about 2
volts. If a flashlight is switched on in outer space, show that its terminal veloc-
ity will be of the order of 10~4 meter/second.
(d) Show that, in the process, the flashlight loses about one part in 10' of its
original mass.

5-27. Let us investigate some of the conditions under which interstellar travel would
be possible.
(a) First, time should be contracted by, say, a factor of 10. Then y = 10.
Show that v/c must then be 0.995.
(b) Imagine a space ship equipped with a photon motor. The motor annihi-
lates the fuel and produces a beam of light directed backwards. This type of
motor consumes the least amount of fuel; see the preceding problem.
You can find the fractionf of the initial mass that remains after the ship
has attained a velocity of 99.5% c by writing out the equations for the conser-
vation of mass-energy and for the conservation of momentum, Take into
 RELATIVITY I

account the mass-energy and the momentum of both the rocket and the
radiation.
You should find that f = 0.05.
The space ship must later be braked to a stop, and this requires 99.5%
of the remaining mass. At the end of the return trip we are left with a frac-
tion f4, or 6.25 & 10~°, of the initial mass.
(c) It has been suggested that, in principle, the rocket could collect and burn
‘interstellar matter which has a density of about 1 atom of hydrogen per cubic
centimeter.
Calculate the mass of gas collected during one year if the rocket sweeps
out a volume 1000 square meters in cross-section at the velocity of light.
Is the interstellar gas a useful source of fuel?

5-28. A photon of energy hy is emitted at the surface of a star of mass M and

radius R.
(a) Show that the fractional frequency change Av/y is GM/Rc? after it has
escaped to infinity, where G is the gravitational constant.
This is the gravitational red shift.
(b) Calculate Av/y for the Sun (G = 6.67 X 10711, R = 7.0 X 108 meters,
M = 2.0 X 10*° kilograms) and for the Earth (R = 6.4 X 10° meters, M =
6.0 * 1024 kilograms).
(c) The star Sirius and a smaller star revolve about one another. By analyzing
this rotation using Newtonian mechanics, astronomers have been able to
establish that the mass of the smaller star is about equal to that of the Sun.
However, light from this star has a Ay/y of 7 X 1074.
What is its average density?
(d) The Sun rotates once in about 24.7 days.
What Doppler shift should we observe for light of \ = 5 X 10-7 meter
from the edge of the Sun’s disk at is equator?
(e) Compare this Doppler shift with the gravitational red shift.
5-29. An excited nucleus of Fe*’ formed by the radioactive decay of Co*? emits a
gamma ray of energy 14.4 10? electron-volts.
(a) By what fraction is the energy of the emitted ray shifted because of the
recoil of the nucleus, if the iron atom is completely free?
Assume that the mass of the nucleus is equal to that of 57 protons
(S7 X 1.7 X 107?” kilogram).t
It was discovered by R. L. Méssbauer in 1958 that, when the iron is in
solid form, a significant fraction of the atoms recoil as if they were locked
rigidly to the rest of the solid. This is Mossbauer effect.
(b) If the solid has a mass of 1 gram, by what fraction is the frequency of the
emitted ray shifted in the “‘recoilless” process?
(c) The natural line width Av/v of the Fe*? gamma ray is 3 X 107-8, What is
the value of Av?
(d) How does this compare with the Av for the decay of a free iron atom and
for a recoilless process ?
A sample of normal Fe®’ absorbs gamma rays of 14.4 KeV by the inverse
PROBLEMS 245

recoilless process much more strongly than it absorbs gamma rays of any
nearby energy. The excited nuclei thus formed re-emit the 14.4 KeV radiation
in random directions some time later. This process is called resonant scat-
tering.
(e) If now the source is moved toward the absorber at a velocity v, what must
be the value of » if the absorber is to see gamma rays shifted in frequency by
3 parts in 10%, corresponding to one resonance line width?
(f) What happens to the counting rate of a counter placed behind the absorber
when the source is moved (i) toward the absorber, (ii) away from the absorber
at the same speed?
(g) If a 14.4 KeV gamma ray, emitted without recoil by an Fe®’ nucleus,
travels 22.5 meters vertically upward, by what fraction will its energy be
reduced ?
(h) An Fe*’ absorber located at this height must move with what speed and
in what direction in order to scatter these gamma rays by recoilless processes ?
R. V. Pound and G. A. Rebka found that Av/y = (2.56 + 0.26) 10735
in this case.

5-30. Show that, if a force F is exerted on a mass m,

dé
eon
5-31. Use the equation
d
F= dt (mv)

to show that the three vectors F, v, a are coplanar, where a is the acceleration.

5-32. Show that a mass density transforms as follows:

2a E 1 ue]
Pi I ey if oF Cc? P25

Viz V0 2
p2 ll ve [1aia ae i

5-33. Imagine that electric charge is not invariant and that Q = Qo[1 — (0/e)7i"2)
(Remember that charge is in fact invariant, according to all experiments per-
formed to date.) The charge Qp is that measured by an observer moving with
the charge, and Q is the charge for an observer moving at a velocity U with
respect to it.
(a) If the electrons in a given sample have an average energy of 100 electron-
volts, what percentage increase in their charge must we expect if their velocity
increases by 1%?
(b) Would this be enough to produce an observable effect?

5-34, Calculate v, 6, y for electrons in matter, assuming that they have kinetic
energies of 10 electon-volts.
in refer-
5-35. We have seen in Section 5-21 that the charge and current densities
246 RELATIVITY I

ence frame 1 that correspond to a conduction current parallel to the x-axis

in frame 2 are
[one == (0/c2)(yJox) and ie = YI 225

so that Ji; # Up.

How can this apply to a conducting wire?t¢

Hints

5-2. You might think of using either a semilog or a log-log plot. You might also think
of plotting several graphs, or of plotting y — 1, etc.

5-29. (a) Call the initial rest mass of the excited nucleus mz and its final rest mass 1.
Then, immediately after the emission of the gamma ray, the nucleus has a mass m7’
and a velocity uw. Eliminate both m’ and u. If there were no recoil, the gamma-ray
energy would be AE = (1) — mj)c2. Show that

hy = AE{1 — (AE/2myc?)}.

5-35. Remember that, in a conductor, we have fixed positive charges and mobile negative
electrons. Treat both sets of charges separately, and then calculate the net charge
density and the net current density in 1.
CHAPTER 6

RELATIVITY II
The Electric and Magnetic Fields
of Moving Electric Charges*

Our knowledge of relativity is now amply sufficient to permit us to calculate

the electric and magnetic fields of moving electric charges, as long as we can
neglect accelerations. The fields of ccelerated charges are more complicated,
and we shall not discuss them from the point of view of relativity.
We shall first calculate EF and B for an individual charge moving at a
constant velocity 0 along the x-axis. Then we shall make the usual assump-
tion, which was first made by Lorentz, that all electric and magnetic fields are
due to elementary electric charges. Under this assumption, the basic proper-
ties of our E and B fields for a single charge must apply to the E and B fields
of any set of charges moving at arbitrary constant velocities.
We shall be able to deduce Maxwell’s four fundamental equations of
electromagnetism, in their general form, but our reasoning will really be
valid only for charges moving at constant velocities.
We shall use Coulomb’s law and the concepts of relativity that we
developed in Chapter 5. You will remember that the latter are all direct
consequences of the Lorentz transformation (Section 5.6).
We have already found one of Maxwell’s equations (Eq. 3-41), and, in
Chapters 7 to 10, we shall rediscover the other three without using relativity.
We therefore intend to deduce Maxwell’s equations twice, with and without
relativity. The two approaches are complementary, and there will be a
negligible amount of duplication.

* This chapter is based on Chapter 5. If you have not studied Chapter 5 or its equivalent,
proceed to Chapter 7.
248 RELATIVITY II

6.1 FORCE EXERTED ON A MOVING CHARGE

BY ANOTHER CHARGE MOVING
AT THE SAME CONSTANT VELOCITY

Let us return to the experiment of Trouton and Noble, which we discussed

briefly in Section 5.3. You will recall that they attempted to detect a torque on
a charged parallel-plate capacitor that was suspended so that its plates were
vertical. A torque was expected to arise from a magnetic interaction between
the two charges moving along with the Earth around the Sun. No such torque
was observed. This result is consistent with the fundamental postulate of rela-
tivity, according to which the force between two electric charges must be pre-
cisely the same, for an observer at rest with respect to the charges, whether the
charges move at a constant velocity, or whether they remain fixed with respect
to some reference frame.
We shall calculate the force on one of the charges in a reference frame
with respect to which the charges move along two parallel paths at the same
constant velocity 0. For example, if the charges are two electrons emerging
from an electron accelerator, the force that we shall calculate is the one that
changes the momentum of one of the electrons with respect to the laboratory.
The magnitude of © can have any value less than or equal to c.
After we have studied this relatively simple case we shall go on to more
general concepts.
Let us call the two charges Q, and Q, as in Figure 6-1, Q, being situated
at Os, and Q» at (Xe, ye, 0). Both charges are fixed in reference frame 2 and
move at the constant velocity Ui with respect to reference frame 1.
In frame 2 the force on Q, is Fy, which has the components
Vene O.Q0sX2 Y= QaQrvyr2
> ’ Zo er 0. (6-1)
Areor3 4reor3
Let us calculate the force on Qs, as measured by an observer 0; in
reference frame 1, at the time t; = f = 0 when the two frames overlap.
Observer 0, sees the two charges moving at the same velocity Ui. We simply
use the transformation equations of Table 5-7 to find Fy, remembering that
Vox = Voy = 22 = O in this case and that Q, and Q, are invariant. This gives

Xvi
&
a Xve2 —
= QaQoxe
Arreo(x3 ic yaya? (6-2)

Va = Yo2 as Q.Qoype (6-3)

y — 4mevy(x3 + p25?”
Zu = 0. (6-4)
 249

Figure 6-1. Two charges Q, and Q, are fixed in reference

frame 2 and hence have the same velocity Ui with respect to
frame 1. The charge Q, is at O» and Q, is at (x2, yo, O). We
calculate the force exerted by Q, on QO).

Substituting the values of x. and y. at 4h = fg = 0 from the Lorentz trans-

formation, and omitting the subscript 1 for simplicity,

iG, me
= rekv¥QaQox
BE ae ype (6 a >)

yan OQ.Qvy _ Amen(yy¥QaQvy

2x? + y wu [1 —(/c)"], (6-6)
Amevy(y2x2 + p82 +e
Z, = 0.

Reference frame | can be taken to be fixed with respect to the laboratory ;

then, the above force and the above coordinates are those measured by an
observer in the laboratory.
This is the force exerted by Q, on Q, at the time ¢ = 0 when the charge
Q, 18 at x = 0.
We can rewrite F, as follows:

4 yOaQor — Y¥DaQs Oyj .

tes Areg(y2x2 + y?)8/2 — Aaregc®(y2x? + yl? (6-8)

x¥QaQs (ST eaee| ;

~ Aen y2x? a jaye c2 Wf (2)

ON
|) 6-10

= 20 dradorixt +p + LO ™dreccr(x? + ype Ls

| eo yQa0 yk 5
as if 7 Q.r Y
250

Figure 6-2. The electric field intensity

E, and the magnetic induction B, due
to the charge Q, moving at the uniform
velocity U.

The product 4:ec?, which appears in the last term, occurs in many
calculations and is defined to be 10’:
Aregc? = 10’. (6-11)
This, in effect, defines the coulomb as the unit of charge, because the units for
the other variables F, U, x, y are already defined from elementary mechanics.
In fact, the constant ep which appears in Coulomb’s law is best determined by
first measuring the velocity of light c and then using the above equation.
It is customary to write that

“ = po = 4r X 10-7 henry/meter, (6-12)

where jo is the permeability of free space.

The first term between the braces in Eq. 6-10 is the electric field intensity
E, of the charge Q, at the point x, y, z and at the time t = 0, when the
origins of the two systems coincide, as you can easily verify for the limiting
case where U = O andy = 1.
In fact,
F, = QE, + U X B,), (6-13)
where
a yQar =
E, = Aren(y2x? + p32 (6-14)
and
Pie Koy QaUyk Q (6-15)
4n(y?2x? + yp? sl?

is the magnetic induction due to Q, at r. The magnetic induction B is expressed

in volt-seconds/meter’, or in webers/meter2, or in teslas. The vectors E, and
B,, and the corresponding forces are shown in Figures 6-2 and 6-3.
The force F; is called the Lorentz force.
Note that the magnetic field of reference frame 1 has appeared as a result
 251

Figure 6-3. The electric force Q,, and

the magnetic force 0,0 X B, exerted
on Q,. The charges Q, and Q, move
at the same uniform velocity U.

of the application of a relativistic transformation to the electric force in reference

frame 2.
The electric field intensity E, is radial, as well as the electric force Q,Ea.
Figure 6-4 shows two limiting cases: in position (a), E, is y times larger in
frame | than if © were zero, but in position (b) it is smaller by the factor
Ly? = 1 — (0/c)*.
The magnetic induction B, is oriented in the positive direction of the
z-axis and is therefore related to U as in Figure 6-2. It exerts a force that is
perpendicular to the trajectory of Q, and that tends to bring Q, and Q, closer
together if Q, and Q, are of the same sign, as in Figure 6-4. In position (a) the
magnetic force cancels part of the electric force, and the total force tends to
zero when the velocity U of the charges approaches c. In position (b) the
magnetic force is zero.

6.2 FIELD OF A CHARGE

MOVING AT A CONSTANT VELOCITY

We now wish to obtain a more general expression for the field of a charge
moving at a constant velocity. To do this we shall use two charges as in
Figure 6-5: the force exerted on Q, will give us the field of Q, at the position
of Qs. The charge Q, is situated at O. and thus moves at the constant velocity
Vi with respect to the laboratory (reference frame 1). The charge Q, has some
unspecified velocity v,, also with respect to the laboratory, The velocity vp
need not be constant.
We calculate again the force exerted by Q. on Qz, first in reference frame
2, and then in reference frame 1.
252

(a) (b)

Figure 6-4. The two charges Q, and Q, again move at the same uniform
velocity U. In (a) the electric and magnetic forces are in opposite di-
rections. In (b) there is only an electric force.

In reference frame 2 we assume that the only field acting on Q, is that

of Q,.
In frame 2, Q, is stationary and creates a field E,, as in Section 2.2. We
also assume that the force acting on Q, in frame 2 is independent of the
velocity of Q, and is given simply by Coulomb’s law, or by the product
E4.0s-
Then the force on Q, in frame 2 has the three components

X= Q.QsX2
Areor3
vu Q.Qrvy2
Arers
) Zug = 2226-16)
4rers

We again use the transformation of Table 5-7, except that now Uns = 0:

Q.Q> | VpoyO Upoz O

z } = 9 |9 -
Xo =
i Areor3 *a COE tiseO Yer C2 -E tO 2 | ory

:
|f
Q.0s | Vo
Je
Yunbl = Arerey |\ Toe are (6-18)

Q.Q0> | Z2
ZnS F =
(6 19)
Areoray 1 f OTe |

Now we must transform the coordinates and the velocities on the right-hand
side. Let us start with the brackets:
 253

Figure 6-5. Charges Q, and Q, have velocities Ui and v,

respectively with respect to reference frame 1.

Xe+
PGiVo2yVSah y DS Vpoz U0 A,
C2 + Vpe2V C2 + Vp00V
= ¥(% — Vb) + yUoyV0/c?)y1 + y(0m120/c?)z1, (6-20)

AGG = [1 — Quzd/c)|y1, (6-21)

Z9
1 + (0922°0/c?)
= [1 — (v120/c*)]21. (6-22)
The vector drawn from Q, to Q; is
r= (x — UDi+ yj + Zk. (6-23)
Then the force exerted by Q, on Q, in reference frame 1, with respect to which
Q, has a velocity Ui and Q, has a velocity vs, is

F, =1222) 6 + (/e2)[ (omy + 0012)i — tog — rmzh]}, (6-24)

Areors

= YOO {r + (1/c?)v, X (UO X r)}. (6-25)

Ar eors

We have again omitted the subscript 1 for simplicity. We have made only one
assumption: we have assumed that the velocity © of Q, is constant. Now
re = yx — Of)? + y? + 2, (6-26)
ay ren Ul) aye 8), (6-27)
= (1 — B? sin? 6), (6-28)
254

Figure 6-6. The E and B vectors at a

point P due to a charge Q moving at a
velocity U.

where r and @ are as in Figure 6-5, and 8 = U/c. Then

F, —= trey rl Q.Q> oN,

— B sin? 62 [ri + (1/c?)us =
X (OX m1)] = (6-29)

where r; = r/r is the unit vector pointing from Q, to Q.

The quantity between the braces shows that the force has again two
components, the electric force represented by the first term r:, and the mag-
netic force represented by the double cross-product term:

F, = QE. + vs X Bi), (6-30)

where
es Q (6-31)
ri,
~ Greyy?r(1 — B? sin? 6)8/?
oQWD sin 6
B
~ Geel — BP sin? He PP eae)
as in Figure 6-6. We have now omitted the subscript a since we shall not need
it from now on. You can verify that these equations reduce to Eqs. 6-9 and
6-10 for the special case where t = 0, z = 0.
In Cartesian coordinates,

yOl(x — VA)i + yj + 2k] (6-33)

Ameo y?(x = Ur)? -f y? -f- sol ae

paAn [y?(x vOU[—2j + yk)

— Ut)? + y? + 22]8/2
(6-34)

This is the electric field intensity E and the magnetic induction B at the point
(x, y, z) and at the time ¢ due toa charge Q moving at a constant velocity Vi
and passing through the origin at the time t = 0.
We have arrived at this result by calculating the force on a second moving
 255

Figure 6-7. Figure (a) shows the force F, exerted by Q, on Q,. Figure
(b) shows the new r; and the new @ which we use to calculate F,. It
is found that F, # —F,, contrary to what one would expect from ele-
mentary mechanics.

charge, first in a reference frame moving with Q, and then in the reference
frame of the laboratory.
What is the force exerted by Q; on Q,? According to elementary me-
chanics, F, should be equal to — F,. However, our experience warns us to be
cautious, so let us calculate Fy.
We can find F, from the value of F; given in Eq. 6-29. In this equation U
is the velocity v, of Q, and

0.0% 9 ,
(6-35)
y= Aregy?r2(1 — 6? sin? 6)3/2 {r: + (1/c?)uv X (va X rij.
=

This is the force exerted by Q. on Q, if the various terms are defined as in

Figure 6-7a, and if

B = 4/C, y= pe (6-36)

To find F, we interchange the roles of Q, and Q, in the above formula.

This means
(a) replacing r; by —r, as in Figure 6-7b,
(b) interchanging v, and vs,
(c) replacing 6 by 6’,
(d) replacing 8 = v,/c by B’ = v/c.
256

(a) (b)
Figure 6-8. (a) Typical line of E, and (b) typical lines of B for a charge
QO moving at a constant velocity U, as seen by a stationary observer.
The electric field is radial. The lines of B are circles centered on the
trajectory and perpendicular to it.

Thus

ee pee sin? 6’)3/2 {ri + (1/c?)va X (vs X M1}. (6-37)

The first and second terms within the braces give, respectively, the
electric and magnetic components of Fy.
It is obvious that F, ~ —F,. The electric components are exactly in
opposite directions, but their magnitudes are not the same because of the 6’,
6’, y’ terms, and the magnetic components are completely different.
There is a difference between F, and F, that is worth noting. The
expression for F, is valid if, or v,, is constant. That for F,, on the contrary,
is valid if v, is constant.
The fact that F, ~ —F, is not peculiar to electric phenomena. It is a
purely relativistic effect that would be observed with any type of force.

6.2.1 The Electric Field

Note that the electric field of Q is radial: the lines of force of E are straight
lines that converge on Q as in Figure 6-8. The fact that the electric field is
radial comes as a surprise. Imagine that you have fixed a light source to Q,
and that you move with Q,. You will not see Q, where it is at the moment
when you look at it, but where it was when the light that reaches your eye left
Qu. At the time ¢ you will see Q, where it was at the previous time t — (r/c), r
being the distance traveled by the light that arrives at the time ¢. You must
 B=0.85 ae B=0.90
Figure 6-9. Lines of force for a charge Q moving along the diameter of an
imaginary sphere. The dots show where lines emerge from the sphere at the
instant when the charge is at its center. The density of the dots is a measure of
the electric field intensity. The total number of dots is the same in all six figures,
so as to satisfy Gauss’s law (Section 6.8). Note how the field shifts to the region
of 0 = 90° as the velocity increases. For U0 = c the field is all concentrated at
() = OOF,
 2.0

0.4

! ll | | fae re ml
cara
WOE ee Ee
0° 30° 60° 90' 120° 150° 180°

Figure 6-10. The electric field intensity E of a moving point charge as a function
of the polar angle @ of Figure 6-6, for seven values of 8 = U/c. The observer is
stationary and sees the charge moving at the uniform velocity 0. For 8 = 0 the
field is isotropic. It is hardly disturbed at 8 = 0.25. As the velocity increases the
field increases near 6 = 90° and decreases both ahead of the charge (near @ = 0)
and behind it (near 6 = 180°). At extremely high velocities most of the electric
field is concentrated near 9 = 90°. These curves explain qualitatively the validity
of Gauss’s law for moving charges: as the velocity increases, the flux of E shifts
from the regions where 06 ~ 0 and ~ 180° to 6 = 90° and the total flux of E
remains constant. (Then why are the areas under the curves not equal?) Note that
the electric field is always symmetrical about 90°. This means that there is no
way of telling, from the shape of the field, whether the charge is moving to the
right or to the left. The vertical scale gives E divided by Q/4ze0r?.

therefore look, not in the direction of the charge, but some distance back; at a
given moment, the light rays originating from Q, are not radial lines converg-
ing on Quz.
But the electric field is radial—as if the information concerning the
position of the charge traveled at an infinite velocity! Actually, it is only when
the velocity of the charge is constant that its electric field is radial; if the
charge is accelerated there is a retardation effect, and the field is not radial.
We have assumed implicitly that the velocity of the charge has been equal to
6.2 FIELD OF A CHARGE MOVING AT A CONSTANT VELOCITY 259

i for an infinite time; otherwise, any disturbance in the velocity would

appear as a distortion of the field.
Figure 6-9 illustrates how the lines of E migrate toward the region 6 =
90° as the velocity increases, and Figure 6-10 shows the magnitude of E asa
function of @ for several values of B = U/c.

6.2.2 The Magnetic Field

The lines of B are circles centered on the trajectory and perpendicular to it,
as in Figure 6-8. At any moment, the magnetic field is symmetrical about the
instantaneous position of the charge, as is shown in Figure 6-11. Here again,
the field is the same as if the information concerning the position of the charge
traveled at an infinite velocity.

30° 60' 90° 120° 150° 180°

0
Figure 6-11. The magnetic induction B of a moving point charge as a function of
the polar angle 6 for seven values of 8. For 6 = 0 there is no magnetic field. As
6 increases, B first increases at all angles. Then B continues to increase near
8 = 90°, while decreasing both ahead of the charge and behind it. At extremely
high velocities, most of the magnetic field is concentrated near the plane @ = 90°.
Note that the magnetic field is symmetrical with respect to 6 = 90°. The vertical
scale gives B divided by nQyc/4nr?, and thus the maximum ordinate on any
curve is f.
260 RELATIVITY iI

The second term on the right in Eq. 6-30 gives the magnetic force on
Q,. The fact that this force is of the form Q,v, x B, has three important
consequences.
(a) The magnetic force can exist, in a given frame of reference, only if
the charge Q, has a velocity with respect to that particular frame.
(b) The magnetic force on Q, is independent of the component of v,
that is parallel to the B of Q,.

(c) The magnetic force is always perpendicular to the velocity v,, and it
does no work.
Then, if & is the relativistic energy mc? of a particle of mass m which
carries the charge Q,

re
dt
any | Peg) See (6-38)
It is only with an electric field that one can increase or decrease the kinetic
energy of a charged particle; a magnetic field can deflect a charged particle,
but cannot change its speed.

Example THE FIELD OF A 10-GeY ELECTRON AT @ = 90°

Imagine that we have an E-meter and a B-meter with infinitely short
response times. What will be the maximum values of E and of B
at a distance of 1 centimeter from the path of a single 10 GeV elec-
tron? One GeV is 10° electron-volts.
From Eqs. 6-31 and 6-32, as well as from Figures 6-10 and 6-11,
both E and B are maximum at @ = 90°, or when the line joining the
particle to the point of observation is perpendicular to the path of
the particle. Then sin @ = 1 and

yQ
ER ace = , (6-39)
: 4rreor?
in the radial direction,
_ MoyYQV
Bee I Arr? (6-40)

in the azimuthal direction. The relativistic kinetic energy gives us y:

(m — mo)c? = my — 1c? = 10" X 1.6 X 10-9 joule, (6-41)

ry 1.6 X10"
9.1 X 10" xX 9 x 108 +1=20X104% (6-42)
Then
Re peisz ip eee Liles
= 0.29 volt/meter,
Ome
(6-43)
6.3 TRANSFORMATION OF ELECTRIC AND MAGNETIC FIELDS 261

pe LUTON PLONE C10 19 x 398108

max ~~ 10-4

= 9.6 < 107! tesla. (6-44)

We have set U = c because, at such a large value of y, » xc.
Although E is 2 X 104 times larger than if the electron were
stationary in the laboratory frame, it is still quite small. As for B,
it is about 10~® times that of the Earth.

6.3 TRANSFORMATION OF ELECTRIC

AND MAGNETIC FIELDS

We have just seen that an electric field in one reference frame becomes both
an electric and a magnetic field in another reference frame. There exists a
general rule for transforming electric and magnetic fields, and we shall be able
to deduce it by again transforming the force on a moving charge.
A charge Q has a velocity v; in a region where there is an electric field
intensity E, and a magnetic induction By, all with respect to reference frame 1.
Then
F, = O(F, + vu X B)). (6-45)
To find FE; and B; in a reference frame moving at the velocity Ui with respect
to the first, we shall transform F, into F,. We should find that
F, —= O(E, + (2D) x B,), (6-46)

and this should give us E, and B, in terms of Ki, Bi, V.

We first write out the components of F:

X = O[Biz + (MB — 12Bry)], (6-47)

V12Biz) |, (6-48)
1G = Ol Fay aie (01, Biz —

ae V1, Biz) |. (6-49)

Zi = O[ Fi, si (U12Biy

Now, from Table 5-7,

x0)
X, = X Ce — m0 (dy ¥1 + 01221), (6-50)

ae Y; — (6-51)
Ye = TT — (uix0/e)]
ie yl — (0/c?)|
AN i (6-52)
We must substitute into these equations the above values of X;, Yi, Zi, and
then set
262 RELATIVITY II

Table 6-1. Transformation of Electric and Magnetic Fields

Fiz = Eo, Foz — Ez,

Evy = y[Ex + UBz]), Ey, = y[Ey — UBx),
1D, = y [Ex = UBy], E», = y[Eiz + UBy |,
Bi, = Boz, Boz = Biz,

By = y(By — (O/c*) Ex], By = y[By + (0/c?)Fiz],

By, = [Boz + (O/c?)Esy], Bo, = y[Bu — (0/c’)
Ei,).

a
i Le
SY CE
a,
GyEy
Voz a V
6-53
hues)

ty apne
=T+eee(sd)
yee, 6-54
oe
eee ee (6-55)
y[1 + (v2°0/c?)]
as in Table 5-2.
We start with Y, because it is relatively simple to calculate:

Y, = O{y[Ey — UB] + Bis — to2y[Bu — (U/c*)Ey)}. (6-56)

This is of the form
Y, = Q[ Eo, + 022Bo.2 — V2rBoz|, (6-57)
with
Ey = y[Ey — UB), (6-58)
Boz —— Bios (6-59)

Bo, = [Bis — (0/c*) Ey). (6-60)

We have now found three of the six components of E and B in frame 2.
Repeating the calculation with Z, gives us two more components:

Ep, = [Ez at UBy |, (6-61)

By = y[By + (0/c?)E,]. (6-62)

To find E,, we calculate X, for the special case where Vy = 1, = 0. This
immediately gives us
Bo, = Ejz. (6-63)
Grouping our results and adding the inverse transformations, we have
the equations of Table 6-1.
Now we can demonstrate quite easily that

Bi-S = BE =, (6-64)
and that
E,-B, —— E,- By. (6-65)
6.3 TRANSFORMATION OF ELECTRIC AND MAGNETIC FIELDS 263

Therefore both

B- — and E-B
are invariants.

Example THE PARALLEL-PLATE CAPACITOR

As an example, we shall transform the electric field of a moving
air-insulated parallel-plate capacitor.
We choose the y-axis perpendicular to the plates and in the
direction of the electric field, as in Figure 6-12a. Then
Ep, = 0, En, = Fp, Ey, => 0, (6-66)

Bo, = Boy, — Bo, = 0), (6-67)

If now the capacitor moves at a velocity U in the direction of the

x-axis, a stationary observer sees the field of Figure 6-12b:
Ei, = 0, By ae yE, Ey a 0, (6-68)

Bea! Bi, — 0} By, = y(0/c2)E>. (6-69)

This is not too difficult to explain. Let us first consider the
electric field. In the reference frame of the capacitor, we have charge
densities --co, on the plates, where o2 = ef. Now in frame 1 the
plates are shorter by the factor 1/y, but they carry the same charge.
Then o; is larger than a» by the factor y, and then FE, must also be
larger than E, by the factor y. (We shall see in Section 6.8 that
Gauss’s law applies to moving charges.)
What about the potentials on the plates? Since the distance s
between the plates is measured in a direction perpendicular to the
velocity, it is the same in both frames. But
Eos = Vo, Fs = Vo. (6-70)

Figure 6-12. Figure (a) shows a parallel-plate capacitor as seen in its own
reference frame. The electric field intensity inside it is E2. For a sta-
tionary observer who sees the capacitor moving to the right at a velocity
U, the field is that of Figure (b): the electric field intensity is y times
stronger and there is a magnetic induction y(‘O/c?)E2 in the direction of
the z-axis.
264 RELATIVITY II

Then the voltage difference between the plates is y times larger in

frame 1 than in frame 2. We shall return to this subject in the ex-
ample on page 270.
But why should there be a magnetic field? The answer is that
observer 0; in reference frame 1 sees a current flowing to the right
in the lower plate and a current flowing to the left in the upper plate.
You will be able to show in Problem 7-27 that
By, = poo V = mo(yeok2)V = y(V/c?)E». (6-71)

Finally, let us check whether E? — c?B? and E-B are invariant.

In frame 1,
E? — 0(°B? = 7°EB — c*¥y(0/c)"ES, (6-72)
= [1 — (07/c)
|B = E, (6-73)
which is the value of E3 — c2B3. Similarly,
E,-B, = E,- B, = 0. (6-74)

6.4 THE VECTOR POTENTIAL A

In Section 6-2 we found that the field at x, y, z, due to a charge Q traveling

at a constant velocity U along the x-axis and passing through the origin at
= 0 is given by

_ _Qi — Vii + yf + zk] (6-75)

+ 2?7]8?
Areo|y?(x — Ut)? + y?

eee OO —ei or sR) (6-76)

4n [yx — OA? + yy? + 2)?
We are now ready to deduce from these two equations several funda-
mental relations involving FE and B.
We start with the vector potential A. You will remember that the
electrostatic field intensity EK can be expressed as minus the gradient of a
certain quantity V, which we called the electric potential: E = —VV. The
magnetic induction B can be expressed in a similar fashion as

B=VXA, (6-77)
where the quantity A is the vector potential at x, y, z, t. The vector potential
is expressed in webers/meter. We shall show that, when the field is due to a
single charge Q moving at a constant velocity U along the x-axis,

_ Ho yOUVI
(6-78)
4a [ye UD sy? oe
 265

Figure 6-13. The vector potential A is

parallel to the velocity of the moving
charge when the velocity is constant. It
is inversely proportional to r and
inversely proportional to
(1 — B? sin? 6)1/2, where B = U/c.

It is in Some cases more convenient to write A in the following form:

_ Ho OVi z
A= 4r r(1 — B? sin? py? (6-79)
where
P(e 00) 2 pie izit? (6-80)
is the distance between the charge and the point of observation (x, y, z). At
the time ¢, the charge is situated at (Uf, 0, 0).
Note that the vector potential is in the direction of the velocity of the
charge as in Figure 6-13.
We first write down the four partial derivatives of the term between
brackets in Eq. 6-78, as we shall require them quite often:

2 [yx — v0 + y? + 24] = 2x — 00), (6-81)

5pLyx — 00)? + y+ 24] = 2y, (6-82)
Se — ON + y? + 24] = 2, (6-83)
4 Ive — UN? + y+ 22] = —2ytv(x— 04). (6-84)
Then the curl of the above A is

La wed!
Yo
- Ox
0 0
oy
0
dz
eye
OAM
az
0A
oy e
6-85
ee
enone
= Gn [yx 7x Ut)? =. y? -f- z2|3/2

as required.
266 RELATIVITY II

Example THE VECTOR POTENTIAL

FOR A 10-GeV ELECTRON
Let us calculate the maximum value of A for the 10-GeV electron
of the example on page 260. From Eg. 6-79, A is maximum at 6 =
90° and

Arie — poe (6-87)

4rr

We have already found that y is 2.0 « 104 and that U we.

Thus, at 1 centimeter from the path of the particle,
Abe DS Ort 3 ASS I D< Il Xe I 3 I
Amax = A x 10-2 > (6 88)

= 9.6 < 10-!2 weber/meter. (6-89)

The vector potential is everywhere parallel to the trajectory, and
it points in the same direction as the velocity of the particle.

6:5 THE SCALAR POTENTIAL FV.

THE EVECITRIC PIELD INTENSITY £E
EXPRESSEDSIN: TERMS: OF CAND A

We have seen in Chapter 2 that accumulations of electric charge produce

electric fields for which E = —VV. Now, if some of the charges are in
motion, there is also a magnetic field and, if the magnetic field is not constant,
there appears a second electric field that adds to the first one. We shall show
that the electric field intensity associated with a changing magnetic field is
—0A/dt, and that the general expression for E is
0A
E = —\rs a — — (6-90)
-

as in Figure 6-14.
We shall demonstrate this general result for the special case of a single
charge moving at a constant velocity.

Figure 6-14, The electric field intensity

is the vector sum of —VV and
—0A/dt.
 R075 8 = 0.80

B = 0.85
B = 0.90
Figure 6-15. Equipotential surfaces V = Constant for a point charge Q moving
either to the right or to the left. The equipotentials near Q are not shown because
they are too close together.
268 RELATIVITY II

The electric potential at the point (x, y, z) and at the time ¢ due to our
charge Q is
Oe ;
uot th (6-91)
y? + 27]?
ou 4rreg [yx — Ut)? +
It is again sometimes more convenient to write V in the following form:

= ! Q: . (
6-92 )
Are r(1 = Be sin2 6) 1/2

Note the analogy between the formulas for V and A. Figure 6-15 shows
equipotential surfaces for six values of 8.
We can prove the validity of Eq. 6-90 by substituting our expressions for
V and A:
eV. OV ee ey, 0A
Gi. Say ede ar Sak
E=

».
[y?(x =
7yQ
Ut)? + y? — Bene |ge1 Ge — UNE ; + yj‘ + 2
=OAr yx — Opi) (6-94)
_ _yQl(x — vii + yj + 2k) (6-95)
Srey — Oi? + y? + 2]
which is the electric field intensity E as given in Equation 6-75.

Example V, VV, AND 0A/dt FOR A 10-GeV ELECTRON

We return again to the 10-GeV electron and calculate V, VV, and
0A/dt. The sum of the last two should equal the E we calculated in
the example on page 260.
The scalar potential is maximum at 6 = 7/2 and

Vena mie
= Goes (6-96)
‘

At | centimeter from the path of the particle,

2 xX 104 & 1.6 x 10719
[a = (6-97)
4m X 8.85 X 10-2 x 10-?
= 2.88 X 107? volt. (6-98)
To calculate VV at this point, let us use Eq. 6-92 for V, and
spherical coordinates. Then
_ oF lov
VV an r\ + 5 06 0,, (6-99)

Qf —1 B2 sin 6 cos 6 \
~ 4areg LL — B? sin? 2"! * 721 — B? sin? 972 91| ’

(6-100)
6.6 TRANSFORMATION OF THE ELECTROMAGNETIC POTENTIALS V AND A 269

‘YQ
VWV=- aL at 0 = 90°.é (6-101)

We can find the value of 04/dt at 6 = 90° from Eq. 6-79 with-
out performing any calculation. As the particle sweeps by the point
1 centimeter away from its trajectory, r is minimum at 6 = 90°. The
angle @ changes from 0 to z, but sin? 6 has the same value at corre-
sponding angles on either side of 6 = 7/2. Then, as a function of
the time, A is maximum at the instant considered when 6 = 7/2,
and our 04/0dt is zero. Then, for this particular case, E is simply
equal to VV, and this agrees with the result we obtained in the fore-
going example (page 266).

6.6 TRANSFORMATION OF
THE ELECTROMAGNETIC POTENTIALS
V AND A. THE FOUR-POTENTIAL A

The electric potential V and the components of the vector potential A

transform somewhat like x, y, z, t, as is apparent from Table 6-2. We shall
show that these transformation equations are correct by showing that they
transform E and B correctly.
For the moment, we can see that Aj,, A1,, Ar, V/c? transform like x, y, z,
t, and are therefore the components of a four-vector

= (A, jV/c). (6-102)

Then the squared magnitude of A,

V2
A? — =

is an invariant.
Let us first check whether the transformation equations of Table 6-2 for
A and V are consistent with the fact that Ai, = F2, as in Eq. 6-63. To trans-
form

Pig Seed (6-103)

we substitute the values of 0/dx, and 0/dt, from Eqs. 5-127 and 5-130, and we
substitute the values of V; and A ;, from Table 6-2:

Se, a~(v je) =|cy[(Va/c2) + (0/c2)Ave], (6-104)

ll s w | x
Q> ue
aay (U/c’)
0 ‘
= + (02/c2) ae (6-105)
270 RELATIVITY II

Table 6-2. Transformation of the Four-Potential

Aix = y[Aox aF U(V2/c?)], Aox a y[Aiz = UMi/c?)],

Ay = Ay, Ay = Aly,
Ai, — Aoz, Ao, = Az,

Vi/c? = y[V2/c?) + (0/c*) Azz], Voie? = y[Vi/e?) — (0/c*)Aiz).

Similarly,

— 9s = fo + core)
Oh 2 — SE — (w/e El, Ot» J
6-106)
Upon simplifying, we do find that

OV ene ‘
Eis = OXs Oty 7s Foz, (6 107)

as required.
We shall not show that the transformation equations for the other
components of E and B are satisfied, but you might wish to try one or two
more.

Example THE PARALLEL-PLATE CAPACITOR

Let us return once more to the parallel-plate capacitor. We have
already transformed its field in the example on page 263; we should
arrive at the same results by transforming the potentials.
In reference frame 2, with respect to which the capacitor is
stationary,
Apis — Ase — Os (6-108)
while the bottom plate is at +-Vo/2 and the top plate is at —V%/2,
as in Figure 6-1 2a.
Then, from Section 6.6, the bottom plate in frame 1 is at yVo/2
while the top plate is at —yV/2, so that
Ey = vy, (6-109)
as previously.
To calculate the vector potential A at any point inside the capac-
itor, we need the value of V2 as a function of pe

V2 = —* Vow (6-110)
Then

Anes —yuseye Vis, Ay, = 0, Ay, = 0. (6-111)

To find B, we compute the curl of A:

6.8 GAUSS’S LAW 271

ub if Ik
1 dj@
B =— y0/c%)2) — Vosew esl
ans =
(6-112)

» OO @
= 7(0/c) Esk, (6-113)
again as in the example on page 263.
You can easily check that A? — (V2/c?) is the same in both
frames of reference, and that it is therefore invariant.

6.7 THE LORENTZ CONDITION

The integrals for the vector potential A and for the electric potential V are so
similar that one suspects the existence of some general relation linking them
together. Such a relation does exist and it is called the Lorentz condition:

Vio ad=, (6-114)

or, in four-dimensional notation,

[j-A =0, (6-115)

where [_] is the four-dimensional quad operator defined in Eq. 5-131, and A is
the four-potential of Eq. 6-102.
You can easily verify the Lorentz condition by substituting the values of
A and of V from Eqs. 6-78 and 6-91.
We shall return to the Lorentz condition on several occasions, particu-
larly in Chapters 10 and 14.
We now turn to the four fundamental equations of electromagnetism,
which are known as Maxwell’s equations.

6.8 GAUSS’S LAW

In discussing electrostatic fields in Chapters 2 and 3 we discovered a simple

relation between the divergence of EF and the total charge density p;:

V-E= %. (6-116)
€0
This was Eq. 3-41 and is called Gauss’s law. It is one of Maxwell’s equations.
We shall now show that it applies to the E of Eq. 6-31. Gauss’s law applies in
fact to all electric fields.
272

Figure 6-16. Gauss’s law applies to a

moving charge: the surface integral of
E evaluated over the fixed sphere gives
Q/e, as if the charge were stationary.
/ The excess E near 6 = 7/2 is just
ue compensated by the weak field near
we 6 = O and @ = 7. See Figures 6-9 and
Se ad 6-10.

Integrating over a volume 7 bounded by the surface S, and using the

divergence theorem, we obtain the integral form of Gauss’s law:

| eda mee (6-117)

where Q, is the total charge enclosed within the surface S.
We shall check the validity of Gauss’s law by integrating the flux of E
over a spherical surface, of radius R, which encloses the charge Q at a given
moment and which is fixed in reference frame 1. The charge passes through
the sphere at the velocity U as in Figure 6-16.
We can simplify the calculation by choosing ¢ equal to zero. Then

8 ar sina
Jo (1 — B? sin? 63/2 (6-118)
iE:am Arey?

oe OR le sin 6 dé
= Dey? Jo [2 00870 + Uy] (6-119)
Bae a dx
J—1 [82x? + (1/72) ]8 (6-120)
2evy?

==€0 (6-121)
Gauss’s law therefore applies to the E of point charges moving at
constant velocities.

Example In a previous example (page 263), when we checked our transforma-

tion of the field of a parallel-plate capacitor, we used Gauss’s law
in both frames to find the ratio of E, to E,. We were therefore assum-
ing that Gauss’s law applies equally well to stationary and mobile
charges.
6.10 THE CURL OF E 273

6.9 THE DIVERGENCE OF B

The divergence of B is quite easy to calculate from Eq. 6-34:

te ( a
Vip une

Ode lay \EA@ = UP PF

a8 dzE ze — or Y+ y+ aan) (6-122)

= 1— 3yz 3yz
{r yap pas % (6-123)
where we have set
i= la Ut) oye? (6-124)
This is the second of Maxwell’s equations:

|V-B=0 | (6-125)

or, using the divergence theorem,

/ Beda 10. (6-126)

Ss

Intuitively, it is quite obvious that the divergence of B should be zero:

since the lines of B are circles centered on the path of the moving charge,
which is the x-axis in Figures 6-9 and 6-10, the total flux of B flowing out of
any imaginary volume must be zero. This is a general result: the divergence of
B is always zero.

CLUS THESCURTZ ORE

There is a third Maxwell equation, which is stated as follows:

oB
VXE=—-5, ‘
(6-127)

or, using Stokes’s theorem,

0 Od
pip heEe eo2
f real | Bde atrp (6-128)
274 RELATIVITY II

where Sis any surface bounded by the curve C, and where ® is the magnetic
flux linking the curve C.
Let us verify Eq. 6-127 by substituting the values of E and B from
Eqs. 6-33 and 6-34:

VXE=|— — =} (6-129)

yO [dO Zz 0 yah =o
(VX Be = tre Lay 1 (6-130)
az 1
TON Oo ULC a 2 i
(V X E), = rele 4 3? ax [ 3/2) (6-131)
10 (es Ue x ar =e,
(6-132)
Arey | ai
wo NGL) oe Ve
<r Ate [ 52 (y tay (6-133)

LON ore ye ern

(6-134)
WEE i 4ireg \Ox[ 82 ay [ ees
— Ut)y os 3(x — vNy |,
ey” Toy
(6-135)
area ci J
_ 370 [(x — v4) need (6-136)
4trep [ 15? (1 VD |

while
OB wovQ 3y?0(x — V2)
(—zj + yk). (6-137)
Ot = 4r [ ee
We can see that Eq. 6-127 is satisfied since

| — 92= — By = —(0/c)7 (6-138)

This is again a general result: the curl of E is always equal to minus the
time derivative of B. You should be able to show that this is in agreement
with the expression for E in terms of V and A, which we found in Section 6.5.

6.11 THE CURL OF B

The fourth and last of Maxwell’s equations is the following:

6.11 THE CURL OF B 275

VX B= wp @ ae a) (6-139)
where J, is the current density at the point P, where the electric field intensity
is E and the magnetic induction is B, or

OE :
f Bat = w | @ + € =) -da = Molt. c iy
C Ss ot

The meaning of the index m, for matter, will become apparent later in this
section. Let us disregard it for the moment. The current /; is the total current
linking the curve C.
We shall be able to deduce this equation from the fact that Gauss’s law
is invariant under a Lorentz transformation.
Charges that are stationary at P in reference frame 1, in which E and B
are measured, contribute nothing to V X B. Let us disregard them. Imagine
that the density of moving charges at P is p and that these charges move at a
velocity Ui. Then
J = pvi. (6-141)

We have chosen the x-axis in the direction of J at P.

In reference frame 2, which follows these charges, Jo, = 0, and we have a
charge density
p= - (6-142)
from Eq. 5-116. In frame 2, B = 0, and the only information we have about
E, is Gauss’s law:
V-E, =— pe = udad (6-143)

€0 Y€0
or
F
aa F)
ee F) i aep 6-144
OX2 |oe a OV, Ey Ar OZ» E, Ep ( )

We can deduce an equation for the field in frame 1 by using the equations
of transformation of Table 5-10 for the partial derivatives, and those of
Table 6-1 for the components of E:

0 Vd 0 0
(2+35 ) E; bY 5, Bs Ub,
UB) ~(E. ++ UB,) UBy) == p/ven, , ((6-145 )
+ ¥ —(£,

e OE} dB. OB, wath imei) (6-146)

A ey o(% | €0
276 RELATIVITY Il

Table 6-3. Maxwell’s Equations

Differential Form | Integral Form

V-E== €0
i)EdeAS €0

V-B=0 [x da = 0
S
OB 0 fees)
Vy, fra --3 | pda = ar

VxXxB Usk Hod fp-a =v |(t+ ) da

= pol
2 Br Ss 0

We have again omitted the subscripts 1. But Gauss’s law also applies in
frame 1, and V-E = p/e. Then, dividing by V,

Lois (0B, @B,\ _ pO) = Sails (6-147)

Xt oy Oz €C?
9
€9C~

We have arrived at this equation by postulating that the current density

vector J was directed along the x-axis. In the more general case where J has
three components, we have two other equations that can be deduced from
this one by rotating the indices. Combining the three,

Vx B-— o ee Lod m. (6-148)

C0

Since the exact nature of the current is immaterial, we have added a

subscript m on J to indicate that it represents any type of current in matter.
For example, in a lossy dielectric, J,, is the sum of the conduction current
density J; plus the polarization current density dP/dt.

6.12 MAXWELL’S EQUATIONS

Maxwell’s equations are grouped together in Table 6-3. You will be able to
show (Problem 6-18) that they are invariant under a Lorentz transformation.
These are the four fundamental equations of electromagnetism. We shall
have many occasions to discuss them, and especially to use them, throughout
the remaining chapters. We shall not therefore say more about them for the
moment.
6.12 MAXWELL’S EQUATIONS 277

Example THE MAGNETIC FIELD NEAR A STRAIGHT WIRE

CARRYING A STEADY ELECTRIC CURRENT
Until now we have studied the electric and magnetic fields of a single
point charge. In practice, magnetic fields are produced by electric
currents flowing in wires. Such currents are more complex than the
simple cases we have studied so far: (a) the conduction electrons are
distributed over a finite volume, (b) they move in all directions,
(c) they have a broad spectrum of velocities, (d) they drift in the
direction of the electric field through a lattice of fixed positive
charges, and (e) their drift velocity changes in direction as they go
around bends in the wire.
The thermal agitation of the conduction electrons gives a zero
average magnetic field; thus we can concentrate our attention on
the drift velocity.
We first dispose of the acceleration of the drifting cloud of elec-
trons as it goes around bends in the wire. We have not discussed the
fields of accelerated charges, and all that we can say about these
accelerations is that they have no appreciable effect on the magnetic
field. The accelerations produce fields that are of the order of U/c,
or 3 X 10~!* times the one we shall calculate below.
Let us calculate the magnetic field of a current-carrying wire
in two simple cases. The first case will be that of a long straight wire;
we shall rediscover the well-known formula for B. The second case
will be that of a short length of wire and will be the subject of the
next example (page 280). This will give us an expression that can be
integrated over any circuit.
A long straight wire is stationary in the laboratory and carries
an electric current / as in Figure 6-17. Inside the wire, the net charge
density is zero, as we shall see later on in the example on page 424.
There are surface charges, but we can forget about them because
they simply superpose an electric field over the magnetic field we are
interested in. The electric field depends on the resistance of the wire
as well as on the current J flowing through it.
We shall use the same procedure as previously. We shall calcu-
late the force on a charge QO moving at some velocity v in the vicinity
ofthe wire. This force will be of the form Qv X B and will give us B.
We must consider separately the conduction electrons and the
positive charges in the wire. Let us call the fixed positive linear
charge density inside the wire \, and the mobile negative linear
charge density \,, both measured with respect to the laboratory.
Linear charge densities are expressed in coulombs/meter. Then, in
the frame of the laboratory,
Ae Ny 0) (6-149)
, being positive and i, negative. As usual, reference frame 1 is the
laboratory frame, and we omit the subscript 1.
278

Figure 6-17. A charge Q moves at

a velocity v parallel to a straight
wire carrying a current J. The
magnetic induction B of the wire
exerts a force Ov * B on the
moving charge because the wire
appears, to the charge Q, to be
negatively charged.

Frame 2 moves in the positive direction of the x-axis at the

drift velocity Ui of the electrons. The charge Q, on the other hand,
has some velocity v with respect to the laboratory as in Figure 6-17.
So as to simplify the calculation we choose our y-axes so that Q lies
on them at ft; = f2 = 0, and we calculate the force on Q at that
moment in the frame of the laboratory.
The positive charges are fixed with respect to the laboratory
and exert on Q a force

Pp
ers (6-150)
27 e0p
where the coefficient of Q is the electric field intensity at a distance p,
from an infinite line charge of density \, coulombs/meter. Note that
p is the radial distance to Q and not a charge density.
The negative charges are stationary in frame 2 and, similarly,

Y= os
AnQ | (6-151)
Both the charge Q and the distance p, which is perpendicular to the
x-axis, are the same in frames 1 and 2. Then, from Table 5-7,
\n2O

¥, = 2rrenp
y[1 — @,0/c’)], (6-152)
where 0, is the x component of the velocity of Q.
Now the linear charge density inside the wire is simply the vol-
ume charge density multiplied by the cross-sectional area of the
wire, the latter quantity being the same in both frames. Thus ),,»
must transform like a volume charge density, as in Eq. 5-117:
\n2 — [An ry (0/c?)In], (6-153)

where J, is the electron current \,,U. Thus

Ane = YAn[1 na) Cie) c)?} = An, (6-154)

ah
and then

Y, ms 2reop Pan = (n/enial, (6-155)

The first term is an electric force, while the second is a magnetic
6.12 MAXWELL’S EQUATIONS 279

force since it depends on the velocity of Q. Then the total force

exerted on the charge Q is
ie fi
Y, + Yn ey Srl
oe Qv., :
(6-156)
and the electric fields of the positive and negative charges cancel per-
fectly. Remember that \, = —), and that ec? = 1/yo from Eq.
6-15.
Finally, since J, = \,0 = —I, where J is the electric current
flowing in the positive direction of the x-axis,

Y== Hol
oe Qd;. (6-157)

If we express this as a magnetic force Qv X B, we find that

=,ee
a ?1. (6-158)

This is the magnetic induction expressed in teslas at a distance

of p meters from a long straight wire carrying a current of J amperes.
The force Qv X B and the vector B are illustrated in Figure 6-17.
This formula is in principle valid only for an infinitely long wire,
because it is based on Eq. 6-150; in practice it is valid within 1%
if (a) p is less than 7% of the length of the straight portion of the
wire and (b) the point considered is near the middle of the wire
(Problem 7-9).
Could we have arrived at this same result from Maxwell’s equa-
tions? Indeed we could have. In the frame of the laboratory there is
no electric field, except that which comes from the surface charges
and which is irrelevant. Then, from Eq. 6-139, V x B is equal to
oJ, and, integrating over a circle centered on the wire and perpen-
dicular to it,

ip(EAB) daar iLdeed: (6-159)

Or, from Stokes’s theorem,

f,Beale h (6-160)
Now, by symmetry, B must be azimuthal as in Figure 6-17, from
what we know about the magnetic field of a single moving charge
(Figure 6-8). Then the line integral must be simply 2rp B and we
have again Eq. 6-158.
This magnetic field results from a relativistic transformation
of the electric field of the moving electrons. This is indeed surprising
because the drift velocity 0 of the electrons in a conductor is only
of the order of 10-4 meter/second, or one foot per hour, or 3 X 1078
times the velocity of light. Then y is equal to unity within one part
in 1025! Relativistic effects could hardly be expected at such small
velocities.
280 RELATIVITY II

Let us therefore return to Eq. 6-155, and let us calculate the

numerical value of the two terms within the bracket for a typical
case. Let us say that the wire is made of copper, that it carries a cur-
rent of one ampere, and that it has a cross-sectional area of one
square millimeter. Copper contains 10° atoms/meter* and has one
conduction electron per atom. Then
Nn = —107° X 107° X 1.6 K 10-19 = —1.6 X 104 coulombs/meter
(6-161)
This is an enormous charge. The force of attraction or repulsion,
between two point charges of 1.6 « 104 coulombs each, 10 centi-
meters apart, would be about 3 X 1014 tons, which is about the force
of attraction between the Earth and the Moon.
Let us say that the charge Q is an electron, and that v, is its
drift velocity in a second wire that is parallel to the first one. Then

(*:)
Uz
i (5xLOSEam) x 1 2~ 10-*!mad coulomb/meter.
/
(6-162)
2

Therefore the bracket in Eq. 6-155 is the sum of two terms, the
first one of which is about 102° times larger than the second. But the
first one is perfectly cancelled by \,, and we are left with only the
second.
Of course the force on the second wire is appreciable for the
simple reason that it also contains an enormous number of conduc-
tion electrons.

Example THE FIELD OF A SHORT ELEMENT OF WIRE

CARRYING AN ELECTRIC CURRENT
It was interesting to rediscover the well-known formula for the mag-
netic field of a long straight wire because the calculation showed
quite vividly the origin of the magnetic fields of conduction currents.
We now wish to be able to calculate B for a wire of arbitrary shape,
and, for this purpose, we must know the field of a short element so
that we can calculate the total B by integration.
We therefore consider the short length of wire dl at the origin
O; as in Figure 6-18. It contains a positive charge \, d/ which is
stationary, and a negative charge \,, d/ which moves in the direction
of the x-axis at the velocity — ‘U. The total charge is zero:
Ap al + dz dl = 0. (6-163)
To find the field of this short length of wire we shall add the
electrostatic field of the stationary positive charges to the E and B
fields of the moving electrons, calculated as in Section 6.2.
As mentioned at the beginning of the last example (page 277),
we neglect the fact that the electrons are accelerated in drifting
around the curve, and we calculate E and B by considering only
their drift velocity Vi.
 281

Figure 6-18. Element of wire of

length dl carrying a current J. It
produces E and B fields at the
point r. The E field is negligible,
in practice.

The positive charges produce at r(x, y, z) an electric field in-

tensity _
=, Beaten
Pp
~ 4mreor?
(6-164)
and no magnetic field.
The negative charges in d/ can be treated as a single charge X, d/
passing through the origin at a velocity —. The charge X, d/ is
negative. Then, from Eqs. 6-31 and 6-32,
(1 — B)A, dl
En = Fray — B? sin? a (6-165)
po (1 — B?)0AX, di sin 6
rX1 — 6? sin? 92? (6-166)
him ie
Therefore, the fields EK and B at r must be FE, + E, and B,:

SG ee eas (6 167)
roe 61 = B sin2 ae ri,
i

Bp — bo A= BY), di sin 8
~ 4x PL — Bin? ye? P*
(6-168)

The electric field intensity is not zero! The electric fields of the
moving electrons and of the stationary electric charges do not can-
cel, although they are of different signs. This result is apparently in
contradiction with the previous section. As we shall see below, the
E for a short length of wire is positive for a certain range of 0, and
negative elsewhere; it turns out that, once integrated over an infinite
straight wire, the two contributions cancel perfectly. Integrating the
above B over an infinite straight wire also yields the same result as
in the previous section.
Let us deduce approximate values for E and B by setting 6? < 1.
The approximations will be extraordinarily good because, as we have
SCelly O72 LOn-o hus
282 RELATIVITY II

mePet aie eo (1Riz58?sin? @)|ri, (6-169)

TE

a Rea ie Hees5sin 6))ri.

da
:
(6-170)

Thus E is directed outward for sin? @ < 2/3 and inward for larger
angles. The explanation is simple: the E of the positive charges
points outward and is the same in all directions, while the E of the
negative charges points inward and is slightly larger near 6 = 7/2,
slightly smaller near 0 = 0 and r.
Since J = \,(— UV)= ApV, where J is the current in the positive
direction of the x-axis, and since (1 /e9c”)= po,
bol al ee +o ) S
ER Aap? o(1 3 sin 6) ry. (6-171)

Similarly,
Ree fed —89(1 +3 8 sin? 6)sin 6
(6-172)

wy esia sin 6 ¢,, (6-173)

Ce caer (6-174)
ae Ho Tdl x Yr)

where the vector dl points in the direction of the current as in Figure

6-17.
For a closed circuit,

BeHl ee
dh} (6-175)
This is the Biot-Savart law. We shall return to it in the next chapter.

Example FORCE ON A SECOND ELEMENT /dl

We can use these simplified expressions for E and B to calculate the

force on a second wire carrying an electric current.
For two elements of wire as in Figure 6-19, the electric field
intensity of dl, exerts equal and opposite forces on the positive and
negative charges of dl;, since An» = —Ap». Then the force d2F4,
exerted by dl, on dl, is
id: ab = (And dly)Vs x B,, (6-176)

= bo Jadla X 11
= (Anv dh)Vy X eee ge (6-177)

i 4a
BY 17, ES ae
(6-178)
We shall use this result at the very beginning of the next chapter.
To find the force d?Fy, exerted by I,dl, on I,dl., we exchange
 283

Figure 6-19. The magnetic force d2F,,

exerted by the element dl, on the element
dly.

subscripts a and b, and change the sign of r;. Contrary to what one
would expect from elementary mechanics, the result is not equal to
—d?F x, unless dl, and dl, are parallel. This purely relativistic effect
was mentioned earlier in Sections 5.18 and 6.2.

Example FORCE ON A MOVING CHARGED PARTICLE

If instead we have at the point r(x, y, z) a single charge Q moving
at the velocity v, then the electric force is
bol al See
Fea & dar? U0 (1_ 7 Sin? 0)On, (6-179)

while the magnetic force is

bo Idlsin 6
Prag Ov X Aes sts fi. (6-180)

So as to simplify matters, let us assume that v is along the x-axis.

Then
Fey Ss ofl — (3/2) say (6-181)
eae v sin 0
The trigonometric term is of the order of unity, except at 6 = 0,
where the magnetic induction B is zero. The velocity © of the con-
duction electrons is so small (-10~ meter/second) that F.) is com-
pletely negligible in practice. For example, if we imagine that the
particle is a Xe ion (Atomic weight 130) with an energy of only one
electron-volt (1.6 * 107!* joule), its velocity v is of the order of 108
meters/second, and even then the electric force is smaller than the
magnetic force by a factor of 10’.

6.13 SUMMARY

We found that the force on a particle carrying a charge Q and moving at a

velocity v is of the form
F= Q(E+vXB), (6-13)
284 RELATIVITY II

where E and B are respectively the electric field intensity and the magnetic
induction at the point occupied by the particle. This is called the Lorentz
force.
By investigating the force between two moving particles we were able to
show that the field of a charge Q moving at a constant velocity V is given by

Q
~ Aregy?r?(1 — B? sin? 6)3/2 oe
(6-31)
HoQVD sin 6
(6-32)
ae 4ry2r(1 — B? sin? 9)32 ©”
where r and @ are defined as in in Figure 6-2 and where 6 = V/c,
Teer
The permeability of free space is defined as

tio = 4m X 107 = henrys/meter. (6-12)

€oC 2

Using then the equations of transformation for the components of a

force, we found the equations of transformation for E and B (Section 6.3):
Ey, = Foz, Ey, = Ez,

Ey, = y[Exy + UBoz], Foy = y [Ey — VB),

Fy, = y| Ex: — UBzy|; Ey, = y[Exz + VBy],
Bi, = Boz, Bor = Biz,
By = y[Buy —(0/c?\Ex], Bry = YB + (0/c?)Ex],
we | y[ Bo, =F (O/J7VE., | By, y[ Bi = (0 pag rcee

It follows that
oc CB? and E-B
are both invariant.
The magnetic induction B is related to the vector potential A as follows:

B= Yas (6-77)
where
ee. yOUI
4m [yx — Ot)? + yp? + 22}? (6-78)

_ Ho Ovi
4x r(1 — B? sin? 6)?” Said.
for a single charge Q moving at the constant velocity Vi.
Similarly, the electric field intensity E can be related to both the vector
potential A and the scalar potential V:
0A
i —V ¥ here ar (6-90)
-
6.13 SUMMARY 285

Table 6-4. Four-Vectors Discussed in Chapters 5 and 6

Four-Vector Components Squared Magnitude

r (r, jct) Gey, Z, jct) r? — cf?

p (p, J&/c) Ga Pus Pzs J&/c) p? ans (8?/c?)

J Gi: cp) Cs Jus dks icp) I= (c?p)

A (A, iV/c) (A,, Ay, A,, iV/c) AS (Vc?)

il @) a @ @ I @ 1 0?
Vv ag Tae ’ ’ > 2 bean ee ere

O ( jc =) (2 dy Oz jc >. Me c? Or

where
Vow v2 :
Arely(x —U1)? yy? 271 Oe
Q
~ 4rer(1 — B? sin? 6)!”
again for a single charge Q having a constant velocity Vi.
The three components of A and V are the components of the four-
potential
A = (A, jV/c), (6-102)
and hence A;, A,, A,, V/c? transform like x, y, z, ¢ (Section 6.6):

Aiz = y[Aoz = U( V2/c?)], Ao = [Au i OVa/e));

Ay, = Ay, Ary = Ay

Ai, oa Aoz, Ao — Aiz,

Vi/c? = y[(V2/c?) + (0/c?) Are], V2/c? = y[(Vi/c?) — (0/c?) Are],

and the squared magnitude of A,
ve
A* — ae

is an invariant.
The Lorentz condition provides us with a relation between V and 4:

V-A+ cots = (0): (6-114)

or
(a2 An—=0, (6-115)

We then demonstrated the validity of Maxwell’s equations for the field

of a charge moving at a constant velocity. These four equations are the
following:
286 RELATIVITY II

SE aa 2 Vie Ea ee,or’
OE
V-B=0, VX B= wo(Jn + 050):

Finally, we used the transformation equations to find (a) the field of a

long straight wire carrying a current / and (b) that of a short element J dl. For
the long wire we found the familiar formula

B= ol $1. (6-158)
2p

This result constituted a striking demonstration of the fact that the magnetic
fields associated with conduction currents result from an exceedingly small
relativistic correction to the e/ectric field of the moving electrons.
For an element of wire dl carrying a current /,

_ wo ldl Xr
ior fr?
(6-174)
The magnetic force exerted by an element of wire dl,, carrying a current J,, on
an element of wire dl,, carrying a current J;, is

PF, ae Ko
ie fees dl, x
=
(dl, x ry)
(6-178)
The element / dl also has an electric field, but the resulting electric force
is either zero or negligibly small.

PROBLEMS

6-1. Show that the force exerted by Q, on Q;, when both charges move at the same
velocity U, as in Section 6.1, can also be written

aXG = 0.0, COS 0

° Arevy2r2(1 — B? sin? 6)3/?

fs Q.Q, sin 0
4revy4r2(1 — 8? sin? 93/2”
where r is the distance between the two particles and @ is the angle between
r and 0, in reference frame 2,
PROBLEMS 287

6-2. Repeat the calculation of Section 6.1 for a pair of equal masses m, and my,
pulled together by a spring under tension exerting a force Kr.
(a) Show that, in frame 1,

F, = —1K {r+ x (Snx)}

at t = 0 when the two systems coincide.

Note that the “‘magnetic’’ force cancels part of the y component of the
radial term and that the net force is not along the line joining m, to mp.
(b) Show that the force on m, is equal and opposite to that on m,. (You
should read Section 6.2 before attempting this.)
(c) Show that the torque exerted on the system, in frame 1, is

yB?K(xu1 — Xa1)Vo1

in the counterclockwise direction.

6-3. Calculate the force, as observed in the laboratory system, between two elec-
trons moving side by side along parallel paths one millimeter apart if they
each have a kinetic energy of (i) one electron-volt, and (ii) one million
electron-volts.

6-4. A proton having a kinetic energy of 50 million electron-volts is deflected by a

magnetic field of 2.5 teslas, which is perpendicular to the trajectory.
Show that the deflecting electric field “seen”? by the proton is about
2.3 X 108 volts/meter.

6-5. Draw a curve of the magnetic induction B as a function of the time at a point
1 centimeter away from the path of a single electron having an energy of 10
GeV (10° electron-volts).

6-6. Show that the equation of motion of a particle of rest mass my and charge Q
can be written as

ym &= o(F+oxB-2uk)

For motion in a static magnetic field,

a =vxXw

where
OB
Orn
is called the Larmor frequency, or cyclotron frequency.

6-7. Comment on the following statement. “There is no such thing as a magnetic

field; what is called a magnetic field is simply an electric field observed in the
wrong frame of reference.”

. (a) Show that, if there exists a field E,, B, with respect to the laboratory, then
the field E,, B, seen by an observer moving at a velocity VU is
288 RELATIVITY II

E,, = y (Fi, + 0 X Bi),

1
Ble (Bi, — 50x h:)
Ey, = Ey\,
By, = By),
where the symbols || and | identify the components that are, respectively,
parallel and perpendicular to U.
DLIISHe OlmOe <a
E,=F, +0 X By,

B, = B, —— 0 X E,.

(b) Show that, if E and B are orthogonal, the field in reference frame 2 is
purely magnetic if

(c) Show that, in that case,

B, —

(d) The motion of a charged particle in uniform and mutually perpendicular

electric and magnetic fields with E < cB consists of a spiraling motion about
the direction of the magnetic field, plus a drifting motion
_ Ex Be
a Bi ca
Note that vp is independent of the nature of the particle.
Sketch the trajectories of positive and negative particles in such a field.
Note that, in a plasma, the positive and negative charges all drift together
at the same velocity vp; no macroscopic current is generated in this way.
(e) Calculate the drift velocity of a proton at the equator under the combined
action of the gravitational attraction and of the magnetic field of the Earth.
In which direction does it drift?
Assume that the magnetic induction of the Earth is horizontal and equal
to 4 & 10~ tesla.
(f) In which direction would an electron drift under the same circumstances ?

6-9. Show that B-F is invariant under a Lorentz transformation.

6-10. Show that B? — (E?/c?) is invariant under a Lorentz transformation.

6-11. Given a field FE, B in the laboratory frame, under what conditions is it possible
to find (a) a reference frame in which either E or B is zero; (b) one in which
FE and B are parallel, say in the direction of the y-axis; (c) one in which E
and B are perpendicular?

6-12. (a) A dielectric-filled parallel-plate capacitor is situated in reference frame 2

with its plates parallel to the xy plane.
Show that €r1 = E92.
PROBLEMS 289

(b) We now turn the capacitor so that its plates are parallel to the yz plane.
The capacitor is again situated in frame 2.
Show that e, is again the same in both reference frames.

6-13. Calculate E and B as functions of the time at a point 1 centimeter away from
the path of a 10 million electron-volt proton.

6-14. The capacitor of the examples on pages 263 and 270 moves in a direction
perpendicular to its plates.
Calculate the potentials on the plates, and the A, E, B inside the capacitor
in the reference frame of the laboratory.

6-15. The magnetic induction inside a long solenoid of radius R that has N’
turns/meter and carries a current J is uo N’/. The magnetic induction outside
is zero (example on page 315).
(a) Calculate the values of E, and B,, both inside and outside such a solenoid,
as measured by a stationary observer, when the solenoid moves at a velocity
© in a direction perpendicular to its length.
Assume that the axis of the solenoid is the z axis and that U = Vi.
(b) Now calculate this same field by transforming the potentials.
First show that, in the frame of the solenoid, the vector potential
Bhy LB :
Ay = —— yoi + 5 xl

gives a uniform field B, oriented in the positive direction of the z-axis.

Note that there are an infinite number of possible expressions for A,. For
example, we could have set
A, = B,x,J.
(c) Now calculate 4;, V;, and then Ej, By, inside the solenoid.
You should arrive at the same result as under (a) above.
Both V, and A, depend on the expression chosen arbitrarily for A,, but we
always obtain the correct values for E, and B, from

E, 1 = —VV, -—

BV
1
eA

6-16. The solenoid of the preceding problem moves at a velocity VU in the direction
of its length.
Calculate E; and B, both inside and outside the solenoid, as measured
by a stationary observer.
290 RELATIVITY II

6-17. Verify that the Lorentz condition applies to the potentials 4 and V of Egs.
6-78 and 6-91.

6-18. Show that Maxwell’s equations are invariant under a Lorentz transformation.tf
6-19. Calculate the drift velocity of conduction electrons in copper from the follow-
ing data:
J = 1 ampere/millimeter?;

Atomic weight, 64;

Density, 9.

Express the velocity in meters/hour.

6-20. A charge QO moves at a velocity V <c parallel to a wire carrying a current J

and zero net charge density. The conduction electrons in the wire move at a
velocity 0 <ce.
Show that, for the charge Q, the wire appears to have a net charge density
vV/c? times the charge density of the conduction electrons in the wire and
that this explains the magnetic force on Q.

6-21. There presumably exists a gravitational equivalent of magnetic forces. Why

have such forces not been observed?

6-22. A beam of particles of mass m and velocity v has a radius r. The charge density
may be assumed to be constant throughout the cross-section of the beam.
Show that the outward force on an ion situated at the periphery of the
beam is
IQ I y2
>
9
21 eqrv Cc?

where J is the current in the beam and Q is the magnitude of the charge on an
ion.
6-23. A certain linear accelerator produces a beam of electrons having a kinetic
energy of 100 million electron-volts. The beam is pulsed, each pulse lasting
25 picoseconds (25 X 10-!? second) and containing 3.0 X 10° electrons. At
the exit of the accelerator the beam has a diameter of 5.0 millimeters.
You are required to investigate the divergence of the beam and, in par-
ticular, to calculate what distance a pulse could travel before blowing up toa
diameter of 2.0 centimeters. The beam travels inside a grounded conducting
tube that has an inside diameter of 10 centimeters.
One may assume that the charge density inside each pulse is uniform.t
(a) Calculate the length, diameter, and linear charge density of a pulse at the
exit of the accelerator in the reference frame of the electrons.
(b) To calculate the divergence of the beam, we shall first calculate the di-
vergence in the frame of the electrons, and then we shall transform to the
laboratory system under (c) below.
You should have found under (a) above that, in the moving frame, the
pulses are long cylinders of charge; so let us calculate the radius of a cylinder
as a function of the time.
PROBLEMS 291

Call the initial radius at the exit of the accelerator ry and the radius of the
tube r;, and calculate the beam radius r as a function of the proper time fo.
To do this you can calculate r as a function of f) for an ion at the periph-
ery of the beam. Radial velocities are small compared to c, and the mass can
be taken to be the rest mass #9.
Neglect end effects.
The integral for r is evaluated as follows:
Tr
iL dii 7
ip U U
= / dr = i ?r du — ar f exp ur du.
ie u
To (in") To 0 0
0

This last integral is tabulated in Tables of Functions by E. Jahnke and F. Emde,

fourth edition, page 32.
You should find that

ps exp (u2) du = — em tn
o Teen

where Xz is the linear charge density and f. is the time, as measured in the
moving frame.
(c) Now transform to the laboratory system and draw a curve of r as a func-
tion of the distance x; traveled by a pulse in the laboratory, for values of r
ranging up to 10 millimeters.
You should find that the pulse has to travel about 400 meters before its
radius becomes equal to 10 millimeters. (We have neglected scattering by gas
molecules.)
(d) We have assumed that the radial velocity dr/dt, in the frame of the elec-
trons is small compared to c.
Was this assumption justifiable?
(e) We have also assumed that the charge density was uniform inside each
pulse.
Was this assumption really necessary?
(f) We shall see in Problem 8-30 that a continuous beam diverges in the same
manner as above.
Why is this?

Hints

6-18. Use Tables 5-8, 5-10, 6-1.

6-22. Calculate the charge per unit length and the force in the reference frame of the ions,
and then transform to the reference frame of the laboratory.

6-23. Use the relation Potential Energy + Kinetic Energy = Constant, setting the po-
tential energy equal to zero at r;, and the radial kinetic energy equal to zero at ro.
CHAPTER 7

MAGNETIC FIELDS I
Steady Currents
and Nonmagnetic Materials

The next three chapters will deal with various aspects of magnetic fields. We
shall start by studying in some detail the properties of two vector quantities,
namely, the magnetic induction B and the vector potential A, which are used
to describe magnetic fields. We shall limit ourselves to steady currents and
nonmagnetic materials. We shall then discuss varying currents in Chapter 8,
and magnetic materials in Chapter 9.
We already know that there are several types of electric current.
(a) There are first the usual conduction currents through good conductors
such as copper, which are due to the drift of conduction electrons.
(b) There are also the conduction currents in semiconductors, which are
due to the drift of either or both conduction electrons and holes (see the
example on page 176).
(c) There are electrolytic currents.
(d) The motion of ions or electrons in a vacuum—for example, in an ion
beam or in a vacuum diode—gives convection currents.
(e) The motion of macroscopic charged bodies also produces electric cur-
rents.
All these currents are associated with the motion of free charges, and they
are the ones we shall be thinking of in this chapter. We shall use the symbol J;
for the current density associated with free charges, and J or J; for the corre-
sponding current.
(f) The polarization current density dP/dt in dielectrics (Section 3.2.2) is
associated with bound charges, and we shall assume for the moment that it is
zero.
7.1. MAGNETIC FORCES 293

(g) There are two other types of electric current, which we shall study
later, in Chapters 9 and 10. These are (a) equivalent currents in magnetized
substances and (b) displacement currents associated with changing electric
fields.
If you have not studied Chapter 6, simply disregard references to it in this
chapter and in all following chapters.

7.1 MAGNETIC FORCES

It is common experience that circuits carrying electric currents exert forces on
each other. For example, the force between two straight parallel wires carrying
currents J, and J; is proportional to J,/;,/p, where p is the distance between the
wires. The force is attractive if the currents flow in the same direction, and it
is repulsive if they flow in opposite directions.
For the more general case illustrated in Figure 7-1, the force that one
current exerts on the other when both are in free space is more complex, but
it is again proportional to the product J,J;. This force is

iB dl, X (dl, Xr)

Fu a. ae i, pp r2 2 (7-1)

where F,,, is the force exerted by current J, on current J,, and where the line
integrals are evaluated over the two circuits.
This is the magnetic force law. The vectors dl, and dl, point in the
direction of current flow, r; is a unit vector pointing from dl, to dl, and r is
the distance between the two elements dl, and dl,. The force is measured in
newtons, the currents in amperes, and the lengths in meters. This law is in
agreement with Eq. 6-178.
The meaning of the double integral is as follows. We choose a fixed
element dl, on circuit 6 and add the vectors dl, X (dl, X 1r)/r? corresponding
to each element dl, of circuit a. We then repeat the operation for all the other
elements dl, of circuit b and, finally, calculate the over-all sum. In general,
this integration cannot be performed analytically. We then divide the circuits
into small finite elements and evaluate the sum numerically.
As we saw in Section 6-1, the constant po is defined as follows:

uo = 4x X 107 newton/ampere? (7-2)

and is called the permeability of free space. Since po is defined in this way,
Eq. 7-1 can be considered to be a definition ofthe unit ofcurrent, the ampere
or coulomb/second.
294

Figure 7-1. Two currents J,

and Ih.

Coulomb’s law gave us the force of interaction between stationary elec-

tric charges. The magnetic force law now states the force between electric
currents. In both laws there are constants of proportionality, ¢ and uo, and it
iS yo that is defined arbitrarily. The coulomb is thus defined, not from Cou-
lomb’s law, but from the magnetic force law. It turns out, experimentally,
that the value of « in Coulomb’s law must be 8.85 X 10-” farad/meter, as
stated previously.
The force F,, is expressed above in such a fashion that dl, and dl, do not
play symmetrical roles. This is quite disturbing, since Newton’s third law
surely applies to a pair of circuits carrying steady currents, and F,, must equal
— Fry.
The force F,, can be expressed in a symmetrical form by expanding the
triple vector product under the integral sign:

dl, X (dla X ri) e: dl,(dl, +11) xn ri(dl,+dly)

r r r (7-3)

We now show that the double integral of the first term on the right is zero.
First
ffa(dly-r) ¥ f il fBets (7-4)
hi a b

Now, on the right-hand side, the second integral is zero for the following
reason. It is the integral of dr/r? around a closed curve, circuit b. It is therefore
the integral of (1/r?) dr with identical upper and lower limits of integration,
and it is zero.
7.2 THE MAGNETIC INDUCTION B. THE BIOT-SAVART LAW 295

We are thus left with the double integral of only the second term for the
triple vector product dl, X (dl, X 11), and

es Pes
re EE pp r:(dl,-dl,)
err ars (7-5)
Therefore, F., = —F%., since the unit vector r; is directed toward the circuit
on which the force is to be calculated, with the result that it is oriented in one
direction for F, and in the opposite direction for Fa.

7.2 THE MAGNETIC INDUCTION B.

THE BIOT-SAVART LAW

Despite the fact that the above integral for F,, is simpler and more sym-
metrical than that of Eq. 7-1, it is not as useful. The reason is that, with the
above integral, the force cannot be expressed as the interaction of current b
with the field of current a. We can perform such an operation on Eq. 7-1,
however, since

ora ng an x (en dl,

GX), a
andarxn, (7-7)
where

is the magnetic induction due to circuit a at the position of the element dl, of
circuit b. As usual, the unit vector r; points from the source, to the point of
observation: it points from the element dl,, to the point where B is calculated.
The magnetic induction is expressed in fes/as or in webers/meter?:

1 tesla = 1 weber/meter? = 104 gauss. (7-9)

The weber is a volt-second. The gauss is not an MKSA unit, but it is fre-
quently used because of its convenient order of magnitude.
The above equation for B is the Biot-Savart law, which we have already
found in the example on page 280. The integration can be performed analyti-
cally only for the simplest geometrical forms. It shows that the element of
force dF on an element of wire of length dl carrying a current / in a region
where the magnetic induction is B is

dF =TdlX BL (7-10)
296 MAGNETIC FIELDS I

If the current J is distributed in space with a current density J; amperes/

meter?, then J becomes J; da and must be put under the integral sign. Then
J, da dl can be written as J; dr, where dr is an element of volume, and

re a
Pie at) ee
Ee ae
(7-11)
The integration is carried out over any volume 7’ that includes all the currents.
The subscriptfon the current density J; refers to the fact that the currents we
are considering in this chapter are due to free charges. We assume, as we do
throughout this chapter, that J; is not a function of the time and that there are
no magnetic materials in the field.
Can this integral be used to calculate B at a point inside acurrent-—
carrying conductor? Since r is the distance between (a) the point of observa-
‘tion where B is measured and (b) the point where the current density is J;,
we are led to expect that the contribution of the local current density will be
infinite because of the 1/r? factor. The integral does not, in fact, diverge; it
does apply within current-carrying conductors. This can be seen by analogy
with electrostatics, where the same problem arises in calculating the electric
field intensity FEinside a charge distribution. The components of B and of E
both vary as 1/7’. Since those of FEdo remain finite within a charge distribu-
tion, those of B must also remain finite within a current distribution. We
assume that both the charge density and the current density are finite.
As in electrostatics, where we used lines of force to describe an electric
field, we can describe a magnetic field by drawing Jines of B that are every-
where tangent to the direction of B.
Similarly, it is convenient to use the concept of flux, the flux of the mag-
netic induction B through a surface S being defined as the normal component
of B integrated over S:

v= | Badu (7-12)
S
The flux @ is expressed in webers.

Example THE MAGNETIC INDUCTION DUE TO A CURRENT

FLOWING IN A LONG STRAIGHT WIRE
An element dl of a long straight wire carrying a current J, as in
Figure 7-2, produces a magnetic induction dB as shown in the figure,
with
bol dl sin 6
fi. (7-13)
Se Ae Ve
 297

Figure 7-2. The magnetic induction dB produced by an element / dl of

the current / in an infinitely long straight wire. The vector dB lies in a
plane that is perpendicular to the wire and which passes through P.

Expressing d/, sin 6, and r? in terms of a and p,

=
bol
—
4rp
+7/2
di
J—1/2 take dose ab
=
bol

2mp a
(7-14)
as in Eq. 6-158.
The magnitude of B thus falls off inversely as the first power of
the distance from an infinitely long wire and is in the direction per-
pendicular to a plane containing the wire. The lines of B are circles
lying in a plane perpendicular to the wire and are centered on it.

Example FORCE BETWEEN TWO LONG PARALLEL WIRES

The force between two infinitely long parallel wires carrying currents
I, and J,, separated by a distance p, as in Figure 7-3, follows imme-
diately from the above result. The current J, produces a magnetic
induction B,, as above, at the position of the current /,. The force
acting on an element /dl, of this current is
dF = I,(dl, X B), (7-15)
dp = [tlouole,
2p
(7-16)
and the force per unit length is
dF
dP _ polals 107
dl 2np ry
298

Figure 7-3. Two long parallel wires

carrying currents in the same direc-
tion. The element of force dF acting
on the element dl; is in the direction
shown.

The force is attractive if the currents are in the same direction, andit
_is repulsiveifthey are in opposite directions.
This equation is the basis of the international definition of the
ampere: two long parallel wires separated by a distance of one meter
exert on each other a force of 2 X 10-7 newton per meter of length
when the current in each is one ampere. It is assumed that the diam-
eters of the wires are negligible compared to their separation.

Figure 7-4. The magnetic induction

dB produced by an element J dl at
a point on the axis of circular
current loop of radius a. The
projection of dB on the axis is dB,.
 7.3. THE FORCE ON A POINT CHARGE MOVING IN A MAGNETIC FIELD 299

Example THE CIRCULAR LOOP e

Let us calculate the magnitude and direction of B on the axis of a
circular loop of radius a carrying a current I, as in Figure 7-4.
An element J dl of current produces a magnetic induction dB,
as indicated in the figure. By symmetry, the total magnetic induction
will be along the axis, and

bolal
Us, = dq p28 0; (7-18)
hence
bol 27a Lola?
Tee) oe Sermon Ce
The magnetic induction is maximum in the plane of the ring and
drops off as z3 for z? > a?.

7.3 THE FORCE ON A POINT CHARGE

MOVING IN A MAGNETIC FIELD

Equation 7-7 gave us the force on a closed circuit immersed in a magnetic

field; it can also give us the force on a single charge Q moving at a velocity
v in a magnetic field B. We shall see that Eq. 7-7 is consistent with the
Lorentz force we have already found in Eq. 6-8.
The force on a current element J d/is [dl X B. Now, if the cross-sectional
area of the wire is da,
I= n(dav)O-. (7-20)

where nv is the number of carriers per unit volume, v is their average drift
velocity, and Q is the charge on one carrier. The reason for this relation is
that the total charge flowing per second is the charge on the carriers that are
contained in a length » of the wire.
Then the force on the element dl is
ndadlQv X B,
and the force on a single charge Q moving at a velocity v in a field B is
Ov X B.

This force is perpendicular both to the velocity v and to the local magnetic
induction B.
More generally, if there is also an electric field E, the force is

_ This is the Lorentz force.

300

p-type material n-type material

Figure 7-5. Hall effect in semiconductors. In p-type material conduction

is due to the drift of positive charges (holes) and the Hall voltage is as
shown in (a). In n-type material conduction is due to the conduction elec-
trons and the Hall voltage has the opposite polarity as in (b). Ordinary
conductors such as copper behave as in (b). Figure (c) shows the two
opposing transverse forces QE; and Qu X B on a positive charge
drifting along the axis of the bar at a velocity v. Figure (d) shows how
these forces are reversed in n-type material.

Example HALL EFFECT IN SEMICONDUCTORS

We have already seen in the example on page 176, that semicon-
ductors contain either or both of two types of mobile charges,
namely conduction electrons and holes. When a current flows
through a bar of semiconductor in the presence of a transverse mag-
netic field, as in Figures 7-5a or 7-5b, the mobile charges drift, not
only in the direction of the applied electric field, but also in a direc-
tion perpendicular to both the applied electric and magnetic fields.
This gives rise to a voltage difference between the upper and lower
electrodes.
If the voltmeter connected to these electrodes draws a negligible
amount of current, the plates charge up until their field EZ, is suffi-
cient to stop the transverse drift. This transverse electric field is called
 7.3. THE FORCE ON A POINT CHARGE MOVING IN A MAGNETIC FIELD 301

the Hall field. The net transverse force on the mobile charges inside
the bar is then zero as in Figures 7-5c and 7-5d.
Let us assume that the conduction is due to holes carrying
charges +-e, e being the magnitude of the electronic charge, and that
there are p holes per cubic meter. When the transverse drift has
stopped,
Aa, = ef = evB. (7-21)
Now
I = Jab = (pev)ab, (7-22)
and
.-—. = pea
973 (7-23)
If the conduction is due rather to conduction electrons carrying
a charge —e, the Hall field has the opposite polarity, as in Figure
7-5b.
Note that in both cases the carriers are swept down by the mag-
netic field.
Although the Hall effect is in fact more complex than we have
assumed it to be, our value of V is nonetheless approximately correct.
The Hall effect is commonly used for measuring magnetic fields and
for studying conduction phenomena in semiconductors.

Example THE HODOSCOPE

The hodoscope is a device that simulates the trajectory of a charged
particle in a magnetic field. The principle involved is quite simple:
if the charged particle, of mass m, charge Q, and velocity »v, is re-
placed by a light wire fixed at the two ends of the trajectory and car-
rying a current J, the wire will follow the trajectory if
iy IE
oO a je (7-24)

where T is the tension in the wire. This statement is by no means

obvious, and we shall have to demonstrate its validity.
The advantage of the so-called floating wire lies in the fact that
it is much easier to experiment with a wire than with an ion beam;
the wire is, in effect, an analog computer.
Let us first consider a region where the ion beam is perpendic-
ular to B as in Figure 7-6a. Then
mov?
BQv = —,
R, 7-2
(7-25)

where R; is the radius of curvature of the trajectory:

mov
te : 7-26
302

Q »
ee
Regt! i
Fee
Ay
Lge
vi \

Figure 7-6. (a) Charge Q moving at a velocity v in a magnetic field B.

The radius of curvature of the trajectory is R;. (b) Light wire carrying a
current J in the opposite direction in the same magnetic field. The tension
in the wire is T and its radius of curvature is R». It is shown that the
wire has the same radius of curvature as the trajectory if mv/Q is equal
to T/T.

In Figure 7-6b we have replaced the charged particles by a light

wire carrying a current J flowing in the opposite direction. If the ele-
ment d/ is in equilibrium, the outward force B/ d/ is compensated
by the inward component of the tension T:
i
BI dl = 2Tsin d(6/2) = dl. (7-27)

The radius of curvature of the wire is thus

Rt (7-28)

and the two radii of curvature will be the same if

m T
fais, (7-29)

Note that the particle will be deflected downward if the mag-

netic force is downward, while the wire will curve downward if the
force is upward. The magnetic forces must therefore be directed
in opposite directions.
If the magnetic field is not uniform, and if the beam is not per-
pendicular to B, then the wire does not always follow the trajectory.
For example, magnetic fields are often used both to deflect and focus
an ion beam. In such cases the focusing forces on the beam can be-
come defocusing forces on the wire, which is then deflected away
from the trajectory.

7.4 THE DIVERGENCE

OF THE MAGNETIC INDUCTION B
In Section 6.9 we demonstrated that the magnetic fields of moving charges
were such that V-B = 0. It is also possible to arrive at this same result for
steady currents starting from the Biot-Savart law, Eq. 7-8. Since
7.5 THE VECTOR POTENTIAL A 303

wo | Ie Xn
“i 4r Jy rr
dr’, (7-30)

V. B= oth Vv. Heery dr’ = ie ic G, x< =) dr’. (7-31)

But, from Problem 1-19

Jv xtet-@xn-i x8), om
where the first term on the right is zero because J; is a function of the source
point x’, y’, z’, while the del operator involves derivatives with respect to
the field point x, y, z. The second term on the right is also zero because

i J k
Ae uae ©. 2 a a
WEDS eye) VE agers Ox oy dz ee
(x = xr — yYrt @ Zr
Then
a (7-34)
This equation follows from the definition of B given in Eq. 7-30. In
Chapter 6 we also showed that it is a consequence of Coulomb’s law and of
the Lorentz transformation. The fact that V-B is zero means that there can-
not exist sources of B.
The net flux of magnetic induction through any closed surface is equal
to zero since

(7-35)

7 mlb ey EC LTORSPOTBNTVALUTA

We have shown in Section 6.4 that the magnetic induction B of a single©

moving charge can be expressed as the curl of a certain quantity A, which |
we called the vectorr potential. The same relation applies, of course, if we use
the Biot-Savart law, as we shall see. The vector potential A is an important
quantity; we shall use it repeatedly.
According to the Biot-Savart law, Eq. 7-8, the magnetic induction at the
point P of Figure 7-7 is

-hf peta (7-36)

 304

Figure 7-7. A current element / dl at a source point P’ produces an

element of magnetic induction dB at a field point P.

where 7’ is the volume of the conductor, or

B=] 4r ae
v (2) x dea
lf
(7-37)
from Problem 1-12. Using the vector identity of Problem 1-18,

V()x tar ely1 x (7-38)

where the second term is zero, again because J; is a function of x’, y’, z’, while
V contains derivatives with respect to x, y, z. Then

B= ef Ve (7-39)
4r J r

We can change the order of differentiation and integration, and thus

R=a [e ihJian} (7-40)

4r Jur
=V XA, (7-41)
where

A=ryAeA beeiia dr :
(7-42)

is the vector potential of the current distribution.

This integral, like that for B, appears to diverge inside a current-carrying
conductor because of the 1/r factor, but it actually does not. This
can be seen
from the fact that its components vary as I/r, like the electric
potential V,
which does not diverge within a charge distribution.
The fact that B is equal to the curl of A follows from the equation
_V-B = 0, because the divergence of the curl of a vector
— is identically equal
to zero.
7.5. THE VECTOR POTENTIAL A 305

If the current is limited to a conducting wire,

aloes
nol
cfal (7-43)

Note that the vector potential A is not uniquely defined by the above
integrals. Indeed we can add to these integrals any quantity whose curl is
zero without affecting the value of B in any way. Since we have no reason
to add such a term for the moment, we shall use these integrals for defining
A, and we shall defer this question until later.
Note also that B depends on the space derivatives of A, and not on A
itself. The value of B at a given point can thus be calculated from A only if
A is known in the region around the point considered.

Example THE LONG STRAIGHT WIRE

We have already found the magnetic induction B for a long straight
wire twice, in Section 6.13 and in Section 7.2.1. Let us calculate this
B once more, using the vector potential A this time.
Elements J dl of the current in Figure 7-8 contribute to the
vector potential elements

AV ee (7-44)
all in the same direction. From the fundamental definition of the
curl in terms of a line integral (Eq. 1-88), and from the azimuthal
symmetry of the field, V X A = B is in the azimuthal direction
around the conductor.
For an infinitely long conductor, dA is proportional to d/// for
large values of r, since r = /, and A tends to infinity logarithmically.
However, the fact that a function is infinite does not necessarily
mean that its derivatives are also infinite; that is, B can be finite even
though A is infinite.
Let us calculate A and B for a current of finite length 2L, and
then we can let L go to infinity. Referring to Figure 7-8,

pol fL dl .

= elIn (1 + (p? + pica (7-46)

= oe
Lol f
(ne 1+ 1+ 4) — inp} wa in
_, boll nz,
(7-47)
ys

0) Oar << IE
 Figure 7-8. An element / dl of a current J
in a long straight wire produces an element
of vector potential dA at the point P.

To calculate B = V X A, we use cylindrical coordinates, keep-

ing in mind that 4 is parallel to the z-axis and independent of ¢:

Pi pPi 1
Ke) fe]
lt Palle Ge (7-48)
0 OQ. =A;
B, = 0, (7-49)
Bass (7-50)

OA,
By = — ap (7-51)

epel (p?< L). (7-52)

27p
as in Eqs. 6-158 and 7-14.

Example PAIR OF LONG PARALLEL WIRES

Figure 7-9 shows two long parallel wires that are separated by a dis-
tance R and that carry currents J of equal magnitude but in opposite
directions. To calculate 4 and B we begin with wires of finite length
2L, use Eq. 7-47 for a single wire, and add the two vector potentials
together:

iy peess 2L
Ave aes (in —In ) (7-53)
2n Pa Ps
 307

Figure 7-9. Pair of long parallel

wires carrying currents of equal
magnitude in opposite directions.
The element of current J dl pro-
duces the element of vector potential
dA shown. The vector potential A
due to both wires is directed in the
opposite direction, since the point
considered is closer to J, than to iy.

In this case we can let L tend to infinity before computing the curl,
and

Pe eee
4r pz
a ee
Qn Pa
(7-54)
With the Cartesian coordinates shown in Figure 7-9,

A,=0, 4,=0, 4,= 4%In | oa 7 y)| (7-55)

2 — 2

B, SL OA,
ay me Hol
oe
E my)
3 = 0ill,
| 7-56
(7-56)

= Gees OA, Pena ae

een):
*) 7
aa]
By, Ox 2iraN ps Pa
B, = 0. (7-58y
At the midpoint between the two conductors, x = 0, y = R/2,
hence

petiole tg AU0l, Bae O, Bu= ONn(759)

or (2) ie te
as expected.
308 MAGNETIC FIELDS I

7.5.1 The Line Integral of the Vector

Potential A Over a Closed Curve

The line integral of 4-dl around a closed curve C is equal to the magnetic
flux @ linking C:

where S is any surface bounded by the curve C.

7, OeRHE CUREROP SLE

MAGNETIC INDUCTION B

We have shown that the magnetic induction is always equal to the curl of
the vector potential -X A. We shall now show that

VX B= poly, (7-61)
again assuming a steady state and the absence of magnetic materials.
In terms of A,

VXB=VXVXA=V(V-4)—V2A (7-62)
from Section 1.9.6.
We have already shown in Section 6.7 that V-A is proportional to the
time derivative of the electric potential V, and we shall also show this again
in Section 10.3. Then, with the above assumptions,

V-4=0, (7-63)
For the second term, we have from the definition of A that

V24 = ne i vl! dr’, (7-64)

where we have interchanged the order of differentiation and integration.

The meaning of this equation can be understood by referring to Figure
7-10, At the field point P(x, y, z), where we wish to compute V2.4, we form
the vector J; dr’/r, where J; and dr’ are respectively the current density and
the volume element at the source point P’, and where r is the distance from
P’ to P. We compute the Laplacian of this vector at P by taking the appro-
priate derivatives with respect to the coordinates x, y, z of P. We then sum
 309
P(x, y,2Z)

Figure 7-10. Source point P’ and field

point P for the calculation of V2A.

the contributions from all such sources in the volume 7’, which includes all
points at which J; exists. The volume 7’ may include the field point P, where
r = 0, as will be shown below.
Since J; is not a function of the coordinates of P, we can write the integral
as
Df) =. eo | ay 2 ()
van 1 ar’ 2
(7-65)

Now, by differentiation of
1 1
(7-66)
r (@—xP+0—-y¥ +E- zy)”
we find that V7(1/r) = 0 if r # 0. There can thus be no contribution to the
integral from any element dr’, except possibly if P and P’ coincide and r is
zero.
To investigate the integral at r = 0 we consider a small volume enclosing
a point P, where we wish to calculate V*A, situated inside the current distribu-
tion.
We take the volume so small that J; does not change appreciably within
it; J; may then be removed from the integral:

v4 = bods i Vv? (;)dr’. (7-67)

1 7’ 0 7
The meaning of this integral is as follows. For each element of volume dr’
centered at the point P’ within 7’, we calculate

v (;)a (a os) , (7-68)

r) = \ae* a *at) [@— xP +0 — YP + 2
multiply by dr’, and sum the results. Since V(1/r) = V(1/r),
310 MAGNETIC FIELDS I

vd = th fon (2)ae (7-69)

ieae {a7 (;)dr’, (7-70)

2 [Vie (7) mi, (7-71)

from the divergence theorem. Then

van lel| ede, (7-12)

from Problem 1-12, where r; is the unit vector from the source point to the
field point. In this case r; points inward toward the point P. Thus

v4 = — bod / dQ, (7-73)

4r Js0

where dQ is the element of solid angle subtended at the point P by the element
of area da. Since the surface S’ completely surrounds P,

V2A = — wd, (7-74)

and
VX B= inJ;- (7-75)
This result is again valid only for static fields and in the absence of mag-
netic materials.

7.7 AMPERE’S CIRCUITAL LAW

We can express the above result in integral form by integrating the normal
component of V X B over an arbitrary surface S:

i (V X B)-da = pw /J;+da. (7-76)

S AS

Then, using Stokes’s theorem, we transform the left side of this equation into
a line integral around the closed path C, which bounds the surface S, and

This 1: the line integral of B around a closed

path is equal to po times the total current / crossing any surface bounded by
7.7 AMPERE’S CIRCUITAL LAW 311

the line integral path. Again, we are limited to nonmagnetic materials and
to static fields.
In many cases the same current crosses the surface bounded by the
integration path several times. With a solenoid, for example, the integration
path could follow the axis and return outside the solenoid. The total current
crossing the surface is then the current in each turn multiplied by the number
of turns, or the number of ampere-turns.
The circuital law can be used to calculate B when it is constant along
the path of integration. This law is thus somewhat similar to Gauss’s law,
which is used to compute E when E is constant over a surface.

Example LONG CYLINDRICAL CONDUCTOR

Let us investigate the magnitude and direction of the magnetic in-
duction B and of the vector potential A inside and outside a long,
straight cylindrical conductor, as in Figure 7-11, carrying a current J
uniformly distributed over its cross section with a density

J; w Ble (7-78)
wR?
Outside the conductor, B is azimuthal and independent of ¢,
so that, according to the circuital law,

ites Hol (7-79)

27p
Inside the conductor, for a circular path of radius p,
ay Mod yp? an Holp : 7-80
ae 2rp 2nR? ( )

The magnetic induction B therefore increases linearly with p inside

the conductor. Outside the conductor, B decreases as 1/p. The curve
of B as a function of p is shown in Figure 7-12.

Figure 7-11. Long cylindrical

conductor of circular cross section
carrying a current J. The circle of
radius p is a path of integration
for calculating B inside.
312

0 1.0 2.0 3.0 4.0 5.0 6.0

p (millimeters)

Figure 7-12. The magnetic induction B as a function of radius for a wire

of 1 millimeter radius carrying a current of 1 ampere.

To calculate the vector potential 4 at a point within the con-

ductor, we cannot use Eq. 7-43, which yields an infinite value for A,
as it did for a single infinite wire; instead, we apply Eq. 7-60 to a
path lying in a radial plane, as in Figure 7-13. The path follows the
surface of the conductor parallel to the axis and returns at a distance
p from the axis. Since the vector potential must be parallel to the
current density J;, it is everywhere parallel to the axis, and, for a
long conductor, it is independent of the axial coordinate z. Hence,
from Eq. 7-60,
— A(R)! + A(p)l = ©, (7-81)
where A(R) is the magnitude of A at the radius R, A(p) is its mag-
nitude at p, and ® is the flux through the radial plane surface en-
closed by the path. Now
_ R polp boll ee
oa pga epi ea| (7-82)
Also, the function A(p) for the region outside the conductor was
found in Eq. 7-47. Setting p = R in this equation,

Figure 7-13. Long cylindrical

conductor of circular cross section
carrying a current J. Part of the
wire is removed to show the
magnetic flux @ and the path of
integration for calculating 4 inside.
 313
20-x 1077

3
3
B
5
pe)
oO
= =

cs

SS SS SS ee ee eae Se eee eee

0 5 10 15 20 25
p (millimeters)

Figure 7-14. The vector potential A for a length of wire 2 meters long
and | millimeter in radius carrying a current of 1 ampere.
T or
ae iy (7-83)
eo Ter
where L. is the half-length of the conductor. Then

Hol ( 4P? a
A(p) = In +1 (op < R). (7-84)
4n R? R?
The logarithmic term is constant for a given wire and for a given
current, whereas the term p?/R* increases as the square of the radius
p inside the conductor. Figure 7-14 shows the value of A both inside
and outside a wire that is 1 millimeter in radius and 2 meters long
and that carries a current of 1 ampere.
We have assumed that the conductor is nonmagnetic.
The only term of importance inside the parenthesis is p?/R*, since
the others are independent of the coordinates. One may therefore set
A equal to

_ Hol p?
Aq R*

The curl of this A, we find the B of Eq. 7-80, as expected.

Example THE TOROIDAL COIL

As a second example, consider a close-wound toroidal coil of square
cross section, as in Figure 7-15, carrying a current J.
Along path a, the line integral of B is equal to zero, since there
is no current linking this path. Then the azimuthal B is zero in this
region. The same applies to c and to any similar path outside the
toroid. Then the azimuthal B is zero everywhere outside.
314

on ween a
Ne
posdbunnedds |

NINN
Hf
NE a
Figure 7-15. Toroidal coil of square cross section carrying a current J.
The figures shown by broken lines are paths of integration for B.

Inside, along path 5, B is fairly constant throughout the section

if the toroid is thin in the radial direction. Thus
2rpB = woNI, (7-85)
where JN is the total number of turns, and
N
B = Uo en EE (7-86)

There exist nonazimuthal components of the magnetic induction

outside the toroid. For a path such as d in Figure 7-15, the area
bounded by the path is crossed once by the current in the toroidal
winding and, at a distance large compared to the outer radius of the
toroid, the magnetic induction is that of a single turn along the mean
radius.
It is interesting to note that, although the magnetic induction B
outside the toroid is essentially zero, the vector potential A is not.
This will be evident if one remembers that A is a constant times the
integral of J dl/r, where r is the distance between the element dl and
the point where 4 is calculated. For example, at a point close to the
winding, A is due mostly to the nearby turns, it is parallel to the
current, and it has approximately the same value inside and outside.
The vector potential 4 can therefore exist in a region where
there is no B field. This simply means that we can have at the same
time A ~ 0 and V X A = 0, which is entirely plausible. For exam-
ple, 4 = ki, where k is a constant, satisfies this condition. We are
already familiar with a similar situation in electrostatics: the electric
potential can have any uniform value in a region where E =
—VV = 0.*

* See The Feynman Lectures on Physics (Addison-Wesley, Reading, Mass., 1964) Vol II,
Section 15-5, for a discussion of the quantum-mechanical aspects of this question.
7.7 AMPERE’S CIRCUITAL LAW 315

Example THE LONG SOLENOID

As another example, let us calculate B inside and outside the long
solenoid of Figure 7-16. We shall select a region remote from the
ends, so that end effects will be negligible; also, we shall assume that
the pitch of the winding is small. Let us choose cylindrical coordi-
nates with the z-axis coinciding with the axis of symmetry of the
solenoid.
1. We first note that B has the following characteristics both
inside and outside the solenoid.
(a) By symmetry, B is neither a function of z nor of ¢.
(b) Moreover, B, = 0 for the following reason. Consider an axial
cylinder of length / and radius p, either larger than the solenoid
radius, or smaller, as in Figure 7-16. The integral of B-da over its
surface is simply 27p/B, since the integrals over the two end faces
cancel. But, according to Eq. 7-35, the integral of B-da over any
closed surface is zero. Then B, = 0.
These three characteristics satisfy the condition V-B = 0.
(c) Finally, the curl of B must be zero everywhere except inside
the wire, and thus
een, (7-87)
dp

Figure 7-16. Long solenoid carrying

a current J. The figures shown by
broken lines are paths of integration
for B.
316

Figure 7-17. Conducting sheet

carrying a current density of
amperes per meter. Since
V-B = 0, the normal component
of B is the same on both sides
of the sheet. According to the
circuital law, however, the
tangential component is not
conserved and a line of B is
deflected in the direction shown.

2. Outside the solenoid.

(a) We can show that B, = 0 by considering path a in Figure
7-16. The net current linking this path is zero, and the line integral of
B-dl around it is therefore also zero. Now, since the line integrals
along sides 1 and 2 are zero (B, = 0), the line integrals along sides
3 and 4 must cancel. But sides 3 and 4 can each be situated at any
distance from the solenoid, and B, must therefore be either zero or
constant along them.
Since the lines of B have neither beginning nor end, and since
the total return flux in the infinite space outside must equal the finite
flux inside the solenoid, B, must be zero outside.
(b) A path such as 3 is linked once by the current, and, outside
the solenoid,

")
_ Hol
spree (7-88)
3. Inside the solenoid.
(a) B, = 0 inside because the line integral of B, d/ over a circle
of radius p, say the top edge of the small cylinder shown in Figure
7-16, is 27pBy, and this must be zero according to Eq. 7-75 because
there is no current enclosed by the path.
(b) Considering now path c in Figure 7-16, and remembering that
B, = 0 both inside and outside, and that B, = 0 outside, we see
that B,s = uoN’Is, and
Bosna (7-89)
The magnetic induction inside a long solenoid in the region
remote from the ends is therefore uniform and equal to po times the
number of ampere-turns/meter.

Example REFRACTION OF THE LINES OF B AT A

CURRENT SHEET

Imagine a thin conducting sheet carrying a current density of \

amperes/meter, as in Figure 7-17. We can find how the lines of B
 317

Figure 7-18. A short solenoid.

are refracted in passing through the sheet by proceeding as in Section

4.1.2. Since the divergence of B is zero, the normal component of B
is conserved:
Bu = Brno. (7-90)
Also, if we apply Ampére’s circuital law to a path of length L that
is perpendicular to the sheet as in Figure 4-2,
PANE (7-91)
Bry = By — bod. (7-92)
The line of force is therefore rotated in the clockwise direction for
an observer looking in the direction of the vector X.
We could have arrived at this result in another way. The mag-
netic induction B is due to the current sheet itself and the other cur-
rents flowing elsewhere in the system. According to Ampére’s cir-
cuital law, the current sheet produces just below itself in Figure
7-17a B that is directed to the left and whose magnitude is poA/2.
Similarly, the B just above the sheet is directed to the right and has
the same magnitude. If we add this field to that of the other currents,
we see that the tangential components of B must differ as above.

Example THE SHORT SOLENOID

We can calculate B on the axis of a short solenoid by summing the

contributions of the individual turns, using Eq. 7-19. If the length
of the solenoid is / and if its radius is a, the magnetic induction at
the center is

eB serene (7-93)
+1/2 any
Lo a*N'l dz
= ——__—___, 7-93

= oN’ sin ine (7-94)

as in Figure 7-18. We have assumed that the solenoid is close wound.

For a long solenoid 0, — 1/2, and B— N’/, as in the example
on page 315.
318

Figure 7-19. Lines of B for a solenoid whose length is equal to twice

its diameter.

At one end, again on the axis,

sin Be
B= wN'l 2,
(7-95)
The magnetic induction thus decreases at both ends of the
solenoid, and this is of course due to the fact that the lines of B flare
out as in Figure 7-19.
Upon crossing the current sheet, the radial component of B
remains unchanged, but the axial component changes both its mag-
nitude and its sign. For example, the axial component in the upper
left-hand side of the solenoid in Figure 7-19 changes from, say,
—0.9 woN’Z to +0.1 poN’/, since the surface current density \ is N’.
 319

7.8 THE MAGNETIC DIPOLE

We shall now show that a small loop of wire of area S, situated at the origin
in a plane perpendicular to the z-axis and carrying a current J, produces a
magnetic field

Bos Ae a3) cos 6, (7-96)

bom .,
Vi ie “a Sin 6, (7-97)

Bei) (7-98)

where m = IS is called the magnetic dipole moment of the loop. Note that
this B field is similar to the E field of an electric dipole (Section 2.9 and
Figure 2-10).
We shall start by calculating

pol A dl
HM a (7-99)

for a circular loop, as in Figure 7-20, and then we can generalize to any shape
of loop.
Since the dl vectors have no z component, 4 can only have x and y
components. A little thought will show that, for any given value of r’, we
have two symmetrical dl vectors whose y components add and whose x com-
ponents cancel. Then we need only calculate the y, or azimuthal, component
of A. If ais the radius of the loop, then

7 Te 0
La lie
(7-100)
where g; is the azimuthal unit vector in spherical polar coordinates.
We must now express 7’ in terms of r and of y. First

r2 = p? + @ — 2arcosy, (7-101)
" 2 A 1/2
= (1te ee tons v) (7-102)
le |
i lie

Eo | wet7a
a
T we7. COS Y;
a
(7-103 )

if we make the assumption that a<_r. Since r/r’ & 1, we have substituted
r for r’ in the two correction terms on the right-hand side.
320

x y

Figure 7-20. Circular loop of wire carrying a current J. The vector potential 4 at
the point P is directed in the azimuthal direction.

To replace the cos y term by a function of y, we use the fact that the
scalar product r-a can be expressed either as
(xi + zk)-(acosei+asing j) = xacos¢ (7-104)
or as racos y. Then
cosy = =COS 9, (7-105)

if ie ax
ae 1 — > + 72 608 ¢, (7-106)
and
fol ae a ax
As Ae . a COs ~ 1 — ip + Pp COS ¢ dp 71; (7-107)

_ AboInd OS aae Ind

Be ee (7-108)

= = sin 0 v1, (7-109)

bo m X ry
= Ave ro)aes (7-1 10)
 321

Figure 7-21. Current loop and its

Magnetic moment m.

if the dipole moment is written in the vector form

m=IS kei, (7-111)
We selected a circular loop so as to make the integral for A tractable.
If the loop is plane but not circular, it can be considered to be formed of
closely packed circles of various sizes, and the vector potential is then given
by the same expression as above, except that xa? must be replaced by the area
S of the loop.
To find B, we simply compute V X A, and we do find that

B, _= Mo
4° 2m
= cos 8, :
(7-112)

By _= 7Mo Mm Sin 8, '

(7-113)
Bees (7-114)
This result is valid for a <r. The analogy with the field of the electric dipole
(Section 2.9) is obvious.
It is possible to generalize the concept of magnetic dipole moment in the
following way. For any closed circuit, whether it is planar or not, we can
write that

m 1 pr x Il, (7-115)
where r is the vector from the origin to the current element / dl, and where
the integral is evaluated over the circuit. Figure 7-21 shows the vector m for
a simple loop.
The concept of magnetic dipole moment can be further generalized to a
current distribution J; in a volume r:

Ti — a xX Jy dr. (7-116)
322 MAGNETIC FIELDS I

Example THE LONG SOLENOID

To find the magnetic dipole moment of a long solenoid we must
calculate the line integral of Eq. 7-115 over the wire of the solenoid.
If the pitch of the winding is small, each turn has a dipole moment
aR*I, and, for the complete solenoid,
m = TR°NI k, (7-117)
where JN is the total number of turns.
At distances that are large compared to the length of the sole-
noid, the B field is given by Eqs. 7-112 to 7-114.

7.9 SUMMARY

We are concerned in this chapter with the magnetic fields that are due to the
motion of free charges. As a rule, we think in terms of conduction currents
in wires.
The magnetic force exerted by a circuit a carrying a current J, on a circuit
b carrying a current J; is
ee See fp dl, X ee x ri) (7-1)

where r; is the unit vector pointing from the element dl, to the element dl).
The constant yo is arbitrarily chosen to be exactly 44 X 10-7 newton/ampere?.
The magnetic interaction is best thought of as the interaction between the
current J, and the magnetic field of J,, or inversely. The magnetic induction of
current J, is
dl,
B, = 1, foun (7-8)
and is expressed in fes/as or in webers/meter®. For continuous distributions
of current,

sieas Pt de! (7-11)

The magnetic flux is defined as

=| B-da. (7-12)
S

The magnetic force acting on a charge Q moving with a velocity v in a

magnetic field B is
OQvxX B
If there is also an electric field E, then the total force is
PROBLEMS
323

Q(E + v X B).
This is the Lorentz force.
Starting from our definition of B, we then showed that

V-B= 0. (7-34)
Then

| B-da =a); (7-35)

where Sis any closed surface, and

Be Vek: (7-41)
where

A= _ ual
ie falz (7-42)

is the vector potential. The line integral is evaluated over the circuit carrying
the current J.
If there are only conduction currents,

VX B= wd), (7-61)
and then

f Beat= w |J;-da. (7-77)

C Ss
This is Ampére’s circuital law.
At distances large compared to the size of a current loop of area S and
carrying a current J, the magnetic induction B has the same form as the E of
an electric dipole if we call
HRS (7-111)
the magnetic dipole moment of the loop.
For a continuous current distribution, the magnetic dipole moment is
defined as
mn = sf rX Jy dr. (7-116)

PROBLEMS

Note: An asterisk indicates that the problem requires a knowledge of relativity.

7-1. We have seen in Chapter 2 that ¢ is expressed in farads/meter and, in this

chapter, that wo is expressed in newtons/ampere®. Now é uo = 1/c?, where c
is the velocity of light. Show that this equation is dimensionally correct.
324 MAGNETIC FIELDS I

*7-2. In Section 6.1 we calculated the force exerted by a charge Q, moving at a

velocity V on another charge Q, moving at the same velocity.
Let us set x» = 0. Then, from Eqs. 6-2 to 6-4, the force exerted on Q,; is

QQ» 2)1/2+
reper: See:

Now calculate this same force using Coulomb’s law and the magnetic
force law of Section 7.1, substituting Q,0 for J, and Q,V for J;.
Does the magnetic force law give the correct result in this case?
The magnetic force law should never be used to calculate forces between
single particles.

7-3. An electric arc is 5.00 centimeters long and carries a current of 400 amperes in
a direction perpendicular to a magnetic field of 5O gauss.
Calculate the magnetic force exerted on the arc.

7-4. Calculate the force due to the Earth’s magnetic field on a horizontal wire 100
meters long carrying a current of 50 amperes due north (magnetic). Take the
magnetic field to be 0.5 & 10~4 tesla at an angle of 70° with the horizontal.

7-5. (a) Find the current density necessary to float a copper wire in the Earth’s
magnetic field.
Assume the experiment to be done at the Earth’s magnetic equator in a
field of 10~4 tesla. The density of copper is 8.9 grams/centimeter?.
(b) Will the wire become hot?
The resistivity of copper is 1.7 * 10-8 ohm-meter.

7-6. Show that the total force is zero on a closed circuit carrying a current Jina
uniform magnetic field.

7-7. Show that the magnetic force on a coil is zero in regions where J; is parallel
to B.

. Show that the magnetic induction B can also be written in the form

p-%4r [Jr’ iSr

/

where 7’ is any volume that encloses all the currents J;, and where the operator
V’ is evaluated at the source point x’, ’, 2’, where the current density is oF

7-9. It was stated in the example on page 277 that the magnetic induction due to
the current flowing in a straight wire is given by ol/2mrp within 1 °%, if p is
less than 7% of the length of the wire, and for points near the middle of the
wire. Is this correct? Neglect the field of the rest of the circuit,
7-10. Figure 7-22 shows part of a coil whose cross section has the shape of a pair
of intersecting circles.
(a) In what general direction is the field oriented in the region between the
conductors?
(b) One author states that the magnetic induction is uniform in the
hollow.
Is this true?
 325

Figure 7-22.

Assume that the length of the coil is infinite.

You can solve this one by straight integration if you wish, but there is a
much easier way...

7-11. (a) A straight, flat conductor of width 2a carries a current J. Show that

a Mol
oe 4ra™

The Mol ro
Dire 4ra us r
in the first quadrant, when the coordinate axes are chosen so that the edges
of the conductor are situated at x = --a, and when the current flows in the
direction of the positive z-axis.
The distance from the point where B is measured to the edge at x = a is
r;. The angle @ is that between r; and r, and is positive in the direction from
r; tO ro.
(b) Calculate B,, B,, B at a distance of 26.0 centimeters from the axis of the
conductor at an angle of 72.0° from the x-axis in the first quadrant for a strip
10 centimeters wide carrying a current of 5.76 amperes.
(c) Find the magnetic induction B (i) at an external point in the plane of the
strip at a distance D from its axis and (ii) at an external point in the plane
perpendicular to the strip and at the same distance D from its axis.
(d) How do the results of (i) and (ii) compare when D > a? Explain.

ide Along straight conductor has a circular cross section of radius R and carries
a current J. Inside the conductor, there is a cylindrical hole of radius a whose
axis is parallel to the axis of the conductor and at a distance 5 from it.
Show that the magnetic induction inside the hole is uniform and is equal
to
Mobl +t ——-
27(R? — a’) ‘Lt

. Use the Biot-Savart law to compute the magnetic induction B at the center
of a square current loop carrying a current /.

A short solenoid carries a current J and has N’ turns/meter.

326 MAGNETIC FIELDS I

(a) Show that, at any point on the axis,

= SuolN'(cos Q1 +°COS a2),

where a and a» are the angles subtended at the point by a radius R at either
end of the solenoid.
For example, if the coil has a length 2Z and if the point is situated at a
distance x from the center,
L—x aoe es L-x ;
COS a, =
[R?+(L — x)” Co [R?+(L + x2

(b) Plot a curve of 2B/yoIN’ as a function of x/R for L = 10R and for values
of x/R ranging from 0 to 15.
Note how quickly the field drops off outside the solenoid.

7-15. A solenoid has an inner radius R;, an outer radius Ro, and a length L.
(a) Show that the magnetic induction at the center, when the solenoid carries
a current J, is
_ pont poe+ (a? + B12
where
Ba en {ae
_® _
a, PIR

and n is the number of turns per square meter of the cross-sections.

(b) Show that, if Vis the volume of the winding, the length of the wire is

l= Vn = 2rn(a? — 1)8R}.
(c) Fabry has shown that, at the center of any solenoid,
BNA
B=G EF
734 Pe

where G is a factor which depends on the geometry of the solenoid, P is the

power dissipated in the solenoid, A = nar? (where r is the radius of the conductor)
is the filling factor (the fraction of the coil cross section that is occupied by the
conductor), and p is the resistivity of the conductor.
Check the Fabry equation for the above type of solenoid.
7-16. The Zeeman effect observed in the spectra of sunspots reveals the existence
of magnetic fields as large as 0.4 tesla.
Let us assume that the magnetic field is due to a disk of electrons 107
meters in radius rotating at an angular velocity of 3 X 10-2 radian/second.
The thickness of the disk is small compared to its radius.
(a) Show that the density of electrons required to achieve a B of 0.4 tesla is
about 10!°/meter?.
(b) Show that the current is about 3 xX 10!* amperes.
(c) In view of the enormous size of the Coulomb forces, such charge densities
are clearly impossible. Then how could such currents exist?
 327

i Magnetic
et field

Figure 7-23.

7-17. (a) Compute the period and radius of curvature of the path of an electron
moving in a plane perpendicular to the Earth’s magnetic field. Assume
B = 107% tesla,
e = 1.6 X 10719 coulomb,
m= 9.1 X 10-* kilogram,
electron energy = 3000 electron-volts,
one electron volt = 1.6 < 107! joule.
*(b) Repeat the calculation for an electron having a kinetic energy of 3.0
million electron-volts.

7-18. In the parallel-plate magnetron, the cathode and the anode are flat parallel
plates, and a magnetic field is applied in a direction parallel to the plates.
Electrons are emitted from the cathode at essentially zero velocity.
If the anode is at a distance s from the cathode, and if it is held at a poten-
tial V with respect to the cathode, show that for
eB*s?
fo 2m
no current will flow to the anode. The magnetic induction is B, e is the
magnitude of the electronic charge, and m is the mass of an electron.{
The parallel-plate magnetron is apparently not a useful device, but certain
types of crossed-field amplifiers use a fairly similar geometry.*

7-19. A magnetron consists of an electron-emitting filament at the center of a cylin-

drical anode situated in a uniform axial magnetic field. Electrons of charge e
and mass m are emitted with negligible velocity from the filament.
Find the minimum potential difference between filament and anode for
electrons to reach the anode.}

7-20. Figure 7-23 shows one type of rocket motor that utilizes the magnetic force.
An arc A ionizes the gas, which enters from the left, and the ions are blown
into the crossed electric and magnetic fields. If the current flowing between

* See, for example, A. F. Harvey, Microwave Engineering (Academic Press, New York,
1963), Chapter 12.
328 MAGNETIC FIELDS I

the electrodes C and D is J, then the thrust F developed by the motor is BJs,
if the values of B and s are assumed to be constant throughout the thrust
chamber, and if end effects are neglected.
If m’ is the mass of gas flowing through the motor per unit time, and if v
is the exhaust velocity, then F is m’v, and the kinetic power communicated
to the gas is
Vo=21 0 = Sho BI
If the efficiency of the motor is defined as

1 Wet Wo
where Wp is the power dissipated as heat between C and D, show that

n = ==
1 Lea =>
1
—_———_ =>
1
———’

2m Dy 2E
italy Fe eatery May
where go is the electrical conductivity, t is the volume of the plasma in the
crossed fields, J is the current density, and E is the transverse electric field
produced by the electrodes C and D.
7-21. When a direct current flows in a long straight conductor, the current density
may. be assumed to be the same throughout the cross section, unless the cur-
rent is subjected to a magnetic field originating elsewhere. Let us see how this
comes about.
(a) Calculate the magnetic force on a conduction electron drifting at the sur-
face of a wire 1 millimeter in radius and carrying a current of 1 ampere.
Set the electron velocity equal to 10-4 meter/second.
(b) Calculate the radial electric field intensity required to cancel the mag-
netic force at a radius r inside the wire, and find the corresponding volume
charge density.
(c) How many extra electrons are required per meter to supply this field, and
how does this compare with the number of conduction electrons per meter?
(d) Is it reasonable to assume that the current density is uniform throughout
the cross section of the wire?
7-22. A solenoid has a uniform turn density, except near the ends, where extra turns
are added to obtain a higher magnetic induction than near the center.
Show qualitatively that, if the axial velocity is not too large, a charged
particle that spirals along the axis in a vacuum inside the solenoid will be
reflected back when it reaches the higher magnetic field.
The regions of higher magnetic field are called magnetic mirrors.
Viegs. In 1949 Fermi proposed the following mechanism, now called Fermi accel-
eration, to explain the existence of very high energy particles in space.
Imagine a clump of plasma traveling in space at some velocity v,i. The
plasma carries a current, and thus has a magnetic field. Imagine now a par-
ticle traveling in the opposite direction at a velocity —2,i, both », and v, being
positive quantities. The particle is deflected in the magnetic field and acquires
a velocity vi, where v, is being also a positive quantity.
PROBLEMS
329

Set ». = v2 = c/2. Show that, after reflection, the y of the particle has
increased from 1.15 to 2.69.
See also Problem 5-25.

7-24. A sheet of dielectric situated in the x-y plane has a velocity vi in a region where
the magnetic induction is Bk. Find the polarization P and the surface charge
densities o,.

7-25. A current J flows in a wire bent around a square form measuring 2a meters
on a side.
(a) Calculate the vector potential 4 and the magnetic induction B along an
axis passing through the center of the square and parallel to one of the sides.
(b) Draw curves of A and B as functions of distance from the center of the
square when J is 1.00 ampere and a is 10.0 centimeters.
7-26. (a) Calculate the axial component of the vector potential A at the center of a
helix of 2N turns, of radius R, and of length 2H, carrying a current J.
(b) Show that the result is the same as that for a single wire of length 2H
along the side of the helix that carries a current J.
Why is this so?
(c) Is it possible to use this result to calculate the axial component of B at the
center ? :
(d) Show that the axial component of B at the center of the helix is
LolN

RES es
7-27. /A conducting sheet carries a current density of X amperes/meter.
(a) Show that, very close to the sheet, the magnetic induction due to the cur-
rent in the sheet is woA/2 in a direction perpendicular to X.
*(b) Show that the B given in the example on page 263 in the case of the
moving capacitor agrees with the above result.

7-28. In a Van de Graaff high-voltage generator, a charged insulating belt is used

to transport electric charges to the high voltage electrode.
(a) Calculate the current that can be carried by a belt 50.0 centimeters wide
that is driven by a pulley 10.0 centimeters in diameter and that rotates at 60
revolutions per second, if the electric field intensity at the surface of the belt
is 20 kilovolts/centimeter.
(b) Calculate the magnetic induction close to the surface of the belt, neglect-
ing edge effects.

7-29. A circular loop whose axis is coincident with the z-axis carries a current J.
(a) Calculate 0B,/dp at a point on the axis, (b) B, in the neighborhood of the
axis, (c) 0B,/dp at a point on the axis, and (d) B, in the neighborhood of the
axis.

7-30. Show that the permeability of free space uo can be defined as follows: if an
infinitely long solenoid carries a current density of 1.0 ampere/meter, then
the magnetic induction in teslas inside the solenoid is numerically equal to po.
330 MAGNETIC FIELDS I

*7-31. According to Ampére’s circuital law, the magnetic induction at a distance p

from the axis of a beam of electrons carrying a current J is uol/2mp, and is
independent of the manner in which the current density J varies with p inside
the beam. (It is assumed that J is not a function of the azimuthal angle 6, and
that p is larger than the radius of the beam.)
Show that this is a consequence of Gauss’s law.

7-32. A magnetic bottle is a field which has the property of containing a plasma.
One type of magnetic bottle operates as follows. An evacuated tube sey-
eral meters long and about half a meter in diameter is enclosed in a long sole-
noid that produces a uniform magnetic field. An accelerator injects electrons,
which spiral inside the tube, forming in effect a second winding whose field
opposes that of the solenoid. Magnetic mirrors (regions of stronger magnetic
field; see Problem 7-22) at both ends of the tube reflect the electrons, causing
them to spiral back and forth along the length of the tube, thus increasing
the magnetic field of the electrons which becomes larger than that of the
solenoid.
(a) Sketch a graph of B as a function of radius.
(b) Sketch lines of B in a plane containing the axis of the solenoid.
(c) Show, qualitatively, that the magnetic field prevents the electrons from
straying away from the current sheet.
(d) The electrons ionize the residual gas in the tube. Sketch the path of a
low-energy proton formed by the collision of one of the high-energy electrons
with a molecule or an atom of hydrogen.

. A conducting sphere of radius R is charged to a potential V and spun about

a diameter at an angular velocity w.
(a) Show that the surface current density is
A = ewV sin @ = M sin8,
where M is ewV.
(b) Show that the magnetic induction at the center is

2Ver 2
Bo a oe M.
a (gh
(c) What is the numerical value of this magnetic induction for a sphere 10.0
centimeters in radius, charged to 10.0 kilovolts, and spinning at 1.00 x 104
turns/minute.
(d) Show that the dipole moment is

47R3M k,
where k is a unit vector along the axis and is related to the direction of rotation
by the right-hand screw rule.
(e) What is the dipole moment of a sphere as in (c) above?
(f) What current flowing through a loop 10.0 centimeters in diameter would
give the same dipole moment?
PROBLEMS 331

7-34. Anelectron revolves in a circular orbit with an angular momentum 2!/24 about
a fixed proton. The constant # = 1.05 < 10~*4 joule-second is Planck’s con-
stant divided by 27.
(a) Calculate the magnetic moment in terms of #.
(b) Calculate B at the position of the proton.

Hints
7-8. Use the results of Problems 1-12 and 1-30. The current density J; must be zero
everywhere on the surface of r’.

7-12. Imagine that the hole is filled with a conductor carrying current of the same density.
Use the circuital law to find B. Then imagine that another current of the same density
but of opposite direction is superposed in the space occupied by the hole. Superpose
the two magnetic fields.

7-18. Use the Lorentz force and remember that its magnetic component does no work.
Thus the kinetic energy of an electron is equal to the potential energy it has lost
in the electric field. If the y-axis is normal to the plates and if the origin is at the
surface of the cathode,

dia?
> =e ie
5 ys)

and, for the above critical value of v, u, = Oat y = s.

7-19. Use the conservation of energy and Newton’s second law in the form Torque = Rate
of Change of Angular Momentum. Use polar coordinates.

7-29. Use the divergence and curl properties of B.

CHAPTER 8

MAGNETIC FIELDS II
Induced Electromotance
and Magnetic Energy

In the last chapter we discussed the magnetic fields of steady currents; we

shall now study time-dependent magnetic fields and the nonconservative elec-
tric fields which accompany them. We shall then be able to calculate the
energy stored in a magnetic field, as well as magnetic forces, torques, and
pressures.
We shall continue to disregard magnetic materials. As usual, you may
neglect references to Chapter 6 if you have not worked through it.

8.1 THE FARADAY INDUCTION LAW

We have seen in Section 2.3 that an electrostatic field is conservative, or that

fpk-dl = 0, (8-1)
Thus the work performed by electrostatic forces is zero when a charge moves
around a closed path.
Equation 8-1 is not applicable if the path is linked by a changing mag-
netic flux. Let us consider a simple path having the shape of a loop. Then, if
® is the magnetic flux linking the loop,

if the direction in which the flux is taken to be positive is related to the direc-
 333

Figure 8-1. The positive direc-

tion around an integration path
in a magnetic field B. The posi-
tive direction around the path
and the positive direction for B
are related according to the
right-hand screw rule.

tion in which the line integral is evaluated according to the right-hand screw
rule, as in Figure 8-1. This is the Faraday induction law; the line integral of
E-dl around the path is called the induced electromotance.
If there are no sources in a circuit, the current is equal to the induced
electromotance divided by the resistance of the circuit, exactly as if the electro-
motance were replaced by a battery of the same voltage and polarity. The
induced electromotance is expressed in volts, and it adds algebraically to the
voltages of the other sources that may be present in the circuit.
The above equation can be rewritten as

where S is any surface bounded by the path of integration chosen for the line
integral. This path can be chosen at will, and it need not lie in conducting
material.
If the path ofintegration is not just a simple loop, the surface over which
the magnetic induction must be integrated to obtain the flux linkage can be
complicated. The procedure for a solenoid is illustrated in Figure 8-2.
If the lines of B are parallel to the axis of the solenoid, they cross the
spiral ramp once for each turn of the solenoid, and the total flux crossing
the spiral surface, of the flux linkage, is N times the magnetic flux crossing the
surface corresponding to a single turn. The lines of B are then said to link all
N turns. If the lines of B are not parallel, the flux linkage varies from turn to
334

Figure 8-2. Solenoid BCD connected to a source. To calculate the flux

linkage, the circuit is replaced by ABFGDE, in which the current flows
downward along the dotted part, plus BCDGFB, in which the current
flows upward in the dotted part. The flux linkage for the real circuit is
then the sum of the flux linkages through the spiral ramp and through
the surface AFGE.

turn and the electromotance induced in the solenoid is equal to minus the
rate of change of the total flux linkage.
In general, therefore, the quantity ® in Eq. 8-2 must be interpreted as
the flux linkage. Flux linkage is measured in weber-turns.

Induced electromotance due to the motion of all or part of a circuit in a

fixed magnetic field is known as motional-induced electromotance. It is a direct
consequence of the magnetic force (Sections 6.2 and 7.1), which itself follows
8.1 THE FARADAY INDUCTION LAW 335

from the magnetic force law discussed at the beginning of the last chapter.
Transformer-induced electromotance is observed in a rigid, fixed circuit
when it is linked by a variable magnetic flux. The distinction between motional
and transformer electromotance is artificial in some cases, in that different
observers would identify them differently. They would always agree, however,
on the electromotance. This will be shown in Appendix C.
If the flux linkage © increases, d®/dt is positive, and the electromotance is
negative, that is, the induced electromotance is in the negative direction.
On the other hand, if & decreases, d®/dt is negative, and the electromotance
is in the positive direction. >
irectionof the induced current is always such that it produces a
marnete field that opposes, to a greater or lesser extent, the change in flux,
depending on the resistance in the circuit. Thus, if ® increases, the induced
current produces an opposing flux. If @ decreases, the induced current pro-
duces an aiding flux. This is Lenz’s law.

Example THE EXPANDING LOOP

Figure 8-3 shows a closed rectangular circuit, one side of which can
slide parallel to itself with a velocity u in a region of uniform mag-

x x

x
B
x

x x

Figure 8-3. A conducting wire ab

slides with a velocity u along
conducting rails in a region of
uniform magnetic induction B.
The magnetic force on the elec-
trons in the wire produces a
current / in the circuit. The elec-
tronic charge is taken to be Q.
336 MAGNETIC FIELDS II

netic induction B. A free charge Q in the moving wire then expe-

riences a magnetic force Q(u X B) (Section 7.3).
The vector u X B is called the induced electric field intensity.
It is the magnetic force exerted on a unit charge.
The line integral of the induced electric field intensity around
the circuit is not zero. Taking the line integral in the clockwise direc-
tion, and B in the direction shown in the figure,

f E-dl = piu X B)-dl = ip(u X B)-dl (8-4)

= —wuB. (8-5)
We have assumed that the resistance of the loop is so large that the
current flowing through it has a negligible effect on the value of B.
The right-hand side is the area swept by the wire per unit time
multiplied by the magnetic induction B. It is thus the magnetic flux
swept per unit time as in Eq. 8-2.
If we had assumed that two or more sides of the circuit moved
in the magnetic field, the end result would have been the same; the
induced electromotance is always —d®/dt in such cases.

8.1.1 The Faraday Induction Law

in Differential Form
The Faraday law as stated in Eq. 8-2 says nothing about E at a given point;
it only gives the electromotance in a complete circuit.
Let us state this law in differential form. Using Stokes’s theorem to trans-
form the line integral into a surface integral,

i (V X E)-da = — aff B-da, (8-6)

Ss dt Js
where S is any surface bounded by the closed integration path. If this path
is fixed in space, we may interchange the order of differentiation and integra-
tion on the right-hand side and

Jw x Baa = -ae da. (8-7)

s s of
We have used the partial derivative of B because we now require the rate of
change of B with time at a fixed point. In a later section we shall examine the
more general case in which the path moves.
Since the above equation is valid for arbitrary surfaces, the integrands
must be equal at every point, and as in Eq. 6-127,

iepos S
Ot
(8-8)
8.2 THE INDUCED ELECTRIC FIELD INTENSITY 337.

This is one of the four Maxwell equations. The only other one that we
found without using relativity, was Eq. 3-41. All four equations were discussed
in Chapter 6.
Equation 8-8 is a differential equation that relates the space derivatives
of E at a particular point to the time rate of change of B at the same point.
The equation does not give the value of EF, unless it can be integrated.

8.2 THE INDUCED ELECTRIC FIELD

INTENSITY IN TERMS OF
VHESVEC TOR POTENTIAL-A

The electric field intensity E induced by a changing magnetic field is related

to the vector potential A of Sections 6.4 and 7.5. Since

(glvxas (8-9)
then, from Eq. 8-8,
VXE= -20 x 4), (8-10)
aA
or

vx (Eae=) =. (8-12)
The term between parentheses must equal a quantity whose curl is zero,
namely a gradient. Then
0A
Sle er VV. (8-13)

For steady currents, A is a constant and Eq. 8-13 reduces to Eq. 2-13.
The quantity V is therefore the electric potential of Section 2.3, which is also
known as the scalar potential.
Equation 8-13 is the general expression for EF, which we have already
found in Section 6.5. This equation states that an electric field intensity can
arise both from accumulations of charge, through the —VV term, or from
changing magnetic fields, through the —0A/dt term.
We could also have found Equation 8-12 in the following way. According
to the Faraday induction law, the induced electromotance for any given path
1S

db d
pe. dl SSS
i ap fpfi {-dl
* 5 (8-14 )
338 MAGNETIC FIELDS II

from Section 7.5.1. Interchanging the order of differentiation and of integra-

tion on the right, which we may do since we are considering fixed paths,

f ( as “) a0. (8-15)
We have used a partial derivative under the integral sign because we require
the time derivative of A at a given point on the path. Thus

f (e+). a= [vx (e+ %)-aa=0 (8-16)

over any fixed closed curve bounding the surface S. The integrand on the
right is therefore zero.

Example THE ELECTROMOTANCE INDUCED IN A LOOP

BY A PAIR OF LONG PARALLEL WIRES
CARRYING A VARIABLE CURRENT
A pair of parallel wires, as in Figure 8-4, carries equal currents J in
opposite directions, and J increases at the rate d//dt. We shall first
calculate the induced electromotance from Eq. 8-2 and then from
Eq. 8-13.
The current / in wire a produces a magnetic induction

(8-17)

Figure 8-4. Pair of parallel wires

carrying equal currents J in opposite
directions in the plane of a closed
rectangular loop of wire. When J
increases, the induced electro-
motance gives rise to a current J’
in the direction shown. The vector
potentials A and the induced electric
fields —dA/dt are shown on the
vertical wires. The induced current
rh >| I’ flows in the counterclockwise
=o y
direction because —0A4/dt is larger
on the left than on the right.
8.3. INDUCED ELECTROMOTANCE IN A MOVING SYSTEM 339

and a similar relation exists for wire b. The flux through the loop is
thus
ii Tot+w h dpa ro+w
bie = (p chan / ites) (8-18)
To Pa rb Pb

_ bol KS at:%) (8-19)

2r rales + w) |
and it points into the paper as in Figure 8-4. From Eg. 8-2, the
induced electromotance is —d®/0dr:

fan ae dl

—ttdn[Mee]
ro(Fa + w)

The fact that the line integral is negative for a positive d//dt indicates
aan
that the induced current /’ is in the negative direction with respect
to B. Since B is directed into the paper in Figure 8-4, J’ must flow
in the counterclockwise direction. This is in agreement with Lenz’s
law: the current J’ produces a magnetic field that opposes the in-
crease in ®,
Let us now use Eq. 8-13 to calculate this same electromotance
from the time derivative of the vector potential 4, V being equal to
zero. From Section 7.5, A is parallel to the wires. Choosing the up-
ward direction as positive,
_ tol
A, = es In ae (8-21)

_ Kol
wu(nts) iy a

along the left- and right-hand sides of the loop, respectively. Thus
oan
Mo dl Tp
Se ee -2
Ey, 2m dt % Ce)
Ho dl (0+ *) 8.24

~ Qa dt ue eo! ( )

oe poh dl (rx + w) .
PEdi= oF alBees | oP}
Ho Talo WwW
8-25

To find the electromotance induced in the loop by a changing

current in a single conductor, we set r, > ©, and then
- pooh aI n( Ta ) 8-26
f DL: Ae ta 1 W Ceo)

8.3 INDUCED ELECTROMOTANCE

IN A MOVING SYSTEM

Our differential form of the Faraday law, Eq. 8-8, was limited to systems at
rest. We shall now consider moving systems.
340

Figure 8-5. A path of integra-

tion moves from C, to C; in the
time dt. The displacement is
general and involves a transla-
tion, a rotation, and a distortion.
The point P is assumed to move
with a velocity u in a region
where the magnetic induction is
(dl x u) dt B.

We return to the integral form of the Faraday law, Eq. 8-2, and consider
the rate of change of magnetic flux through a surface bounded by a path C
that moves in some arbitrary manner from C, to C; in the time df, as in
Figure 8-5. A given point P on the path C moves with a velocity u through
a region where the magnetic induction depends both on the coordinates and
on time.
The rate of change of flux through the circuit is

io J, But + do)-day — [ BA)-da,

dt dt att
where B,(t + dt) is the magnetic induction on some surface S;, bounded by
C, at the time ¢ + dt. Similarly, B,(¢) is the magnetic induction on S, bounded
by C, at the time ¢.
The magnetic flux emerging at the time t + dt from the volume + swept
out during the interval ¢ to t + dt is

ilBit + dt)-da, — / BA(t + dt):dag + at |“Bt + dt)-(dl X u)

b Sa Ca

= i V-Bit + dt) dr = 0, (8-28)

since the divergence of B is always zero. We have used B without a subscript

for the magnetic induction along the path of integration C,.
Now, on the surface S,,

BG} ai).da. = Bie © (Be-da,) dt, (8-29)

where the time derivative is evaluated at the time f¢.
8.3 INDUCED ELECTROMOTANCE IN A MOVING SYSTEM 341

Also, from Stokes’s theorem,

f me + dt)-(dl X u) = f [u X Bt + dt)]-dl, (8-30)

E ifly Seiad (8-31)

Substituting 8-29 and 8-31 into 8-28,

[ Bit + dt)-da, — ibB.(t)-daa — ar | oB aw, da,

m a Sa

+ ar | V X (u X B)-da, = 0, (8-32)
and, finally,

db
posi)wryil
big xX (u X B) ; aa + OB
a peed
oF da. (8-33)

The subscript on S is now unnecessary, since we have really calculated A®/At

in the limit At — 0. The two terms represent respectively: (a) the flux gained
through the sides of the volume traced out by the moving path, and (b) the
increase of flux by virtue of the change of B with time.
Thus, from the Faraday induction law,

fpBoal= [Ir X (u X B) — oF aa, (8-34)

or, using Stokes’s theorem,

5B da.
[vy ad er Foe [Ir Yatinein = oe (8-35)
Since this equation must be valid for any surface S bounded by any curve C,
the integrands must be equal at every point, and

This E is the induced electric field intensity as measured in a system

moving with a velocity u relative to that in which the magnetic induction
is measured as B. For example, if E is induced in a moving conductor, u
is the velocity of the conductor relative to the laboratory, and B is measured
with an instrument that is fixed with respect to the laboratory.*

* See Appendix C for a detailed discussion of Eq. 8-36.

342 MAGNETIC FIELDS II

Example THE ELECTROMOTANCE INDUCED

IN A FIXED LOOP IN A TIME-DEPENDENT
MAGNETIC FIELD
We consider the square loop of Figure 8-6. Let us first suppose that
the loop is at rest. There is a uniform magnetic field B, which is
parallel to the z-axis and which varies with time:
B= B)sin ot. (8-37)
Since the loop is at rest, wu= 0,
OB
VxkE= a7 (8-38)

f Edt = i,
S
(FGh)-da = |s OB
sda.
Ot
men (ea)
= —BwS cos 6 cos wt, (8-40)
where S = Wh is the area of the loop.
For a square loop 10 X 10 centimeters? normal to a magnetic
field varying at a frequency of 60 hertz with a maximum value of
0.01 tesla (100 gauss), the induced electromotance has a frequency of
60 hertz maximum value of about 38 millivolts/turn. At any
mome

Example THE ELECTROMOTANCE INDUCED

IN A LOOP ROTATING IN FIXED
MAGNETIC FIELD
Consider now the same loop of Figure 8-6 rotating with an angular
velocity w about the x-axis in a uniform, time-independent magnetic
field B parallel to the z-axis. Then, from Eq. 8-36,

Figure 8-6. Fixed loop in a_ time-

dependent magnetic field B. The
vector m is normal to the loop.
8.4 INDUCTANCE AND INDUCED ELECTROMOTANCE 343

VXE=VxX(u XB), (8-41)

and
13, = a S< RY (8-42)
as in Section 8.1. Thus, setting 6 = wt,

g Bed fu < B)-dl, (8-43)

==) > AB sin wt. (8-44)

The only contributions to the integral are on the vertical sides, since
(uw X B) is perpendicular to dl along the top and bottom parts of
the integration path. Hence

g E-dl = BoS sin wt, (8-45)

where S is again the area of the loop. The electromotance is zero

when the normal to the loop is parallel to B, since the free charges
inside the wire are then moving parallel to B and the induced electric
field intensity is zero.
The case of a loop rotating in a time-dependent magnetic field
is the subject of Problem 8-6.

8.4 INDUCTANCE AND

INDUCED ELECTROMOTANCE

To calculate the electromotance induced in one circuit when the current

changes in another circuit, it is convenient to express the magnetic flux linking
the first one in terms of the current in the second and of a geometrical factor
involving both circuits. This factor is known as the mutual inductance between
the two circuits.
The same procedure can be used to relate the linking flux and the current
for a single circuit, in which case the geometrical factor is known as the se/f-
inductance of the circuit.

8.4.1 Mutual Inductance

Let us seek an expression for the magnetic flux linking one circuit, but due
to the current in another circuit. The current /, in circuit a produces a mag-
netic flux ®,, linking circuit b as in Figure 8-7:

Pap = / B,- dap, (8-46)

So

where da, is an element of area of an arbitrary surface S, bounded by circuit

b, and where B, is the magnetic induction due to current J, at a point on S;.
344

Figure 8-7. Two circuits a and b. The flux ®,, shown linking 4 and
originating in a is positive. This is because its direction is related by the
right-hand screw rule to the direction chosen to be positive around b.

We can calculate ©,, from the vector potential A, produced by J,:

o,, = Hl(V X A,)-dap, (8-47)

Sp

= Pb Aural, (8-48)
as Hola dl, ,
= fp(a f = ) dl, (8-49)

a Mola dl,- dl,

ay es fp = (8-50)

a ee ee (8-51)
where
Bas Ho dl,: dl,
Ma = ae pp rs (8-52)

is the mutual inductance between the two circuits. This is the Neumann equa-
tion.
Similarly, the flux @,, linking circuit a is

Dia = Moral. (8-53)

Since the Neumann equation is symmetrical with respect to the subscripts

a and b,
Me Mis, (8-54)
8.4 INDUCTANCE AND INDUCED ELECTROMOTANCE 345

It will be noticed that mutual inductance is a quantity that depends

solely on the geometry of the two circuits and that, when multiplied by the
current in one circuit, gives the magnetic flux linking the other. A similar
situation exists in the case of a capacitance: the charge induced on a grounded
conductor is the product of the potential of a neighboring conductor, multi-
plied by a geometrical factor called the capacitance between the two
conductors.
The sign of the mutual inductance is arbitrary. It is positive if a current
in the direction arbitrarily chosen as positive in circuit a produces a flux in
the direction arbitrarily chosen as positive through b.
Since the mutual inductance is the magnetic flux linking one circuit per
unit in the othe

The mutual inductance between two circuits is one henry when a current
of one ampere in one of the circuits produces a flux linkage of one weber-turn
in the other.
The electromotance induced in circuit b by a change in the current J, is

fr-at tee
= —8 = Carpe
— m4 (8-55)
Similarly, the electromotance induced in circuit a by a change in the current
I, is

f beatee _ den
oe ay,abyAl (8-56)

This equation is convenient for computing the induced electromotance

since it involves only the mutual inductance and d//dt, both of which can be
measured.
This equation gives us a second definition of the henry: the mutual
inductance between two circuits is one henry if a current changing at the
rate of one ampere/second in one circuit induces an electromotance of one
volt in the other.
We shall have an example of mutual inductance after we have studied
self-inductance.

8.4.2 Self-inductance

A single circuit is of course linked by its own flux as in Figure 8-8, and

wp nl @57
where L is called the self-inductance of the circuit and depends solely on the
 346

Figure 8-8. Single circuit carrying a

current J.

geometry of the circui ike mutual inductance, is measured

in henrys. I
A circuit has a self-inductance of one henry if a current of one ampere
produces a flux linkage of one weber-turn.
A change in the current flowing through the circuit produces within the
circuit itself an induced electromotance

db
f ral Sage dl (8-58)

Dinter cereadds to whatever other voltages are present.

A circuit therefore has a self-inductance of one henry if a current chang-
ing at the rate of one ampere/second induces in it an electromotance of one
volt.
If the circuit is truly filamentary, that is, if the cross-sectional area of
the conductor is infinitely small, the flux @ and the inductance L become
infinite. This can be explained as follows. As the radius of the conductor tends
to zero, B tends to infinity in the immediate neighborhood of the wire. The
region where B is infinitely large is itself infinitely small, but the flux tends to
infinity logarithmically. This flux clings infinitely close to the wire. In prac-
tice, currents are distributed over a finite cross section, and the flux linkage,
and therefore the inductance, does not in fact diverge.
We shall calculate self-inductance from the ratio of flux linkage to cur-
rent for two cases in which the currents are effectively distributed over a
surface. Then both the flux and the inductance are finite.

Example SELF-INDUCTANCE OF A LONG SOLENOID

It was shown in the example on page 315 that the magnetic induction
inside a long solenoid, neglecting end effects, is constant, and that
B = woN'l, (8-59)
8.4 INDUCTANCE AND INDUCED ELECTROMOTANCE 347

Table 8-1. Representative Values of K.

Radius/Length K

0 1.00
0.2 0.85
0.4 0.74
0.6 0.65
0.8 0.58
1.0 0.53
ik 0.43
2.0 0.37
4.0 0.24
10.0 0.12

where N’ is the number of turns/meter. Thus

NI
che ane TR?, (8-60)

where N is the total number of turns, / is the length of the solenoid,

and R is its radius. Then
2
a N® = HoN? rR?, (8-61)
I 1

= poN”? | rR. (8-62)

The inductance of a short solenoid is smaller by a factor K,

which is a function of R//. Representative values of K are shown in
Table 8-1.

Example SELF-INDUCTANCE OF A TOROIDAL COIL

A toroidal coil of N turns is wound on a form of nonmagnetic mate-

rial having a square cross section as in Figure 8-9.
According to the circuital law, or from Eq. 7-86, the magnetic
induction in the azimuthal direction, at a radius p inside the toroid,
is
p= HONE, (8-63)
21p
Thus

gp = WNT : Lakh (8-64)

R+w/2

2m Jr—w/2 ?P

ae
aR,
= ww In biees (8-65)
348

ins
~ 3

Figure 8-9. Toroidal coil of square cross section and mean radius R.

and
a polV2w 2R a “| rs

ee 6 ees . —
For R>w,
ZR RS (ew w ’
2R—w (174 wa )(4 + am) eolh

wy l+RaWw 8-
(8-68)

2R ww :
In [et] a (8-69)
and
Bolas
Poop (R> w). :
(8-70)

The self-inductance of a toroidal coil is proportional to the

square of the number of turns and to the cross-sectional area en-
closed by the winding, and inversely proportional to the length of
the winding. The reason we obtained a similar expression for the
long solenoid is that we neglected end effects.

Example MUTUAL INDUCTANCE

BETWEEN TWO COAXIAL SOLENOIDS

Let us now add a second winding over the solenoid, as in Figure

8-10. We assume for simplicity that both windings are long with re-
spect to the common diameter, in order that end effects may be
neglected.
To calculate the mutual inductance between the two coils we
shall avoid evaluating the double integrals of the Neumann equa-
tion, 8-52, by using again the ratio of flux linkage to current. Let us
 349

Figure 8-10. Coaxial solenoids. The two radii are taken to be approx-
imately equal.

assume a current J, in coil a. Then the flux of coil a linking coil b is

RN,
ae tae (8-71)
and the mutual inductance is

N,®, R2N,
Vp Me)
Ih
UsHe
(8-72)
We can also calculate the mutual inductance by assuming a
current J; in coil 6. Then

_= Saye
Kom R°*NoI
sab
Pra (8-73)

This flux links only (/,//,)N. turns of coil a since B falls rapidly to
zero beyond the end of a long solenoid, as we saw in Problem 7-14.
Then the mutual inductance is

i [,NaPoa
a
a bot R2NLNs
(8-74)
as previously.
It is paradoxical that a varying current in the inner solenoid
should induce an electromotance in the outer one, since we have
shown (example, page 315) that the magnetic induction outside a
long solenoid is zero. The explanation is that the induced electric-
field intensity at any given point is equal to the negative time deriv-
ative of the vector potential at that point and that the vector poten-
tial A does not vanish outside an infinite solenoid, despite the fact
that B = V X A does.
We can actually calculate the vector potential just outside the
inner solenoid from the mutual inductance. If we consider the direc-
tion of the current /, in the primary as positive, then the electro-
motance induced in the secondary is — MdlI,/dt, or
pot R?N GN, dl,
Taare dt’
and the induced electric field intensity —d0A/dt is 2rRN, times
smaller:
350 MAGNETIC FIELDS II

= Ot = = PH.
as,
aki
8-75
CMB):
Integrating, we have the vector potential at any point close to the
solenoid:

yh me ip (8-76)
in the azimuthal direction. This can be shown to be correct by calcu-
lating the line integral of A-dl as in Section 7.5.1.

8.4.3 Coefficient of Coupling

Let us consider a single-turn coil a, through which a current J, produces a
flux ,,, and another single-turn coil b, arranged so that only a fraction k,
of ®,. passes through it:

Pap = KePaas (8-77)

The self-inductance of coil a is

= ® Ls Safe:

La i (8-78)

and the mutual inductance between the two coils is

ive ee Sere (8-79)

Likewise,
Mia kyL>. (8-80)

Then, since My. = Ma = M,

M2? = k,kp Ll, (8-81)

M = +k(L,Ly)'”, (8-82)
where
k — +(k,k»)”, (8-83)

called the coefficient of coupling between the two coils, can have values
ranging from —1 to +1. If the two single-turn coils coincide, the absolute
value of k is unity and its sign is the same as that of M.
Although this simple reasoning, which is based on single-turn coils, does
not apply to the coils which one uses in practice, Eq. 8-82 remains true and
the maximum possible value of k is unity.
We have seen in Section 8.4.2 that the self-inductance of a circuit tends
to infinity as the wire radius r tends to zero. The mutual inductance, however,
does not tend to infinity if one or both of the circuits is filamentary. The
reason is that the intense flux very close to a thin wire does not link another
8.5 ENERGY STORED IN A MAGNETIC FIELD 351

circuit some distance away. That is, as the radius of the wire tends to zero,
L— «,k-—0, and the mutual inductance remains about the same.

Example The coefficient of coupling for the two coaxial solenoids is

k a aoe (uomR?NN»)/la
(8-84)
(LaLs)¥? (oN 2a R2/1a)"! (ugN 3 R2/1,) 2
“sGy". (8-85)
It was assumed in calculating M that /, was shorter than, or equal
LOM

8.5 ENERGY STORED IN A MAGNETIC FIELD

To find the energy stored in a magnetic field, we shall calculate the energy
supplied by a source to an isolated circuit, as the current density increases
from zero to some value J;.
We shall assume that Ohm’s law applies to the conducting medium.
Then, at any given point outside a source,

ip = ae. (8-86)
where a is the conductivity of the medium, expressed in mhos/meter, at that
point. This is Ohm’s Law.
This electric field intensity is the sum of two terms: —VV, produced by
accumulations of charge on the terminals of the source and on the surfaces
of the conductor, and —dA/dt, due to the changing magnetic field. In a wire,
the surface charges adjust themselves so that the total EF is along the axis as
in Figure 8-11.
Inside a source there is also a third electric field E,, which comes from
the local generation of energy and

(ese: o(-vv-“+4,). (8-87)

The work done per unit time and per unit volume on the moving charges
at any point outside the source can be calculated as follows. Consider an
element of volume having the form of a rectangular parallelepiped oriented
so that one set of sides is parallel to the current density J;, as in Figure 8-11.
In one second, a charge J; da goes in through the left-hand face and an equal
charge comes out at the other end. The source maintains a difference in
potential of —VV-dl across these faces, and the integral of VV-dl around
 Figure 8-11. Rectangular par-
allelepiped parallel to the current
density J; in a conductor. The
aA electric field intensity E is the
Shap sum of —VV and —0A/dt.

the circuit is equal to the voltage supplied by the source. The source supplies
to the element of volume dt a power

—VV-dl J,da = —VV-J;dr = ( + =) - J; dr, (8-88)

and the total power spent by the source is

dw 0A
“dt = . (E-J + Ot y J) dr, (8-89)

2
= [at [4 > J; dr. (8-90)
= 0 Ok

The first term on the right gives the power spent in Ohmic or Joule
losses (Problem 8-19), while the second gives the rate at which work is done
by the source against the induced electromotance. This latter work is that
which must be done to establish the magnetic field, and it is the one that
concerns us here.
Writing W,, for the energy stored in the magnetic field,

dW,
fe _ edie
[ad (8-91)
where 7 is again the volume occupied by the currents.

Example For a long solenoid, 4 is given by Eq. 8-76 and, replacing J; dr by

T dl,
AW mar |dl, (8-92)
dt 21 dt
where the integral is evaluated over the length of the wire, 27RN.
Then
dWm _ THoR®N® , dl
dt l dt
(8-93)

Wn = :uol2N’21R2, (8-94)
8.5 ENERGY STORED IN A MAGNETIC FIELD 353

8.5.1 Magnetic Energy in Terms

of the Magnetic Induction B
It is possible to express the magnetic energy W,, in terms of B, just as the
electric energy was expressed in terms of E in Section 2.15.
We assume again that there are no magnetic materials in the field. We
also assume that conduction takes place in good conductors so that Eq. 7-75
can be applicable. Then
l
J; =—V XB, (8-95)
Ho
and
dW, 1 [0A
a = [% -(V X B) dr, (8-96)

“1 [a(x t)-e-(fxa)]a, aon

— i fa.3art 1
> fax. da, (8-98)
the volume 7 being any volume which includes all points where the current
density J; is not zero, and S being the corresponding surface.
As in the corresponding electrostatic case, the above equation becomes
simple if we choose 7 to include all space, in which case the surface S is at
infinity. The magnetic induction B falls off as 1/r* at large distances, This
was shown for the case of a current loop in Eqs. 7-112 and 7-113. The in-
duced electric field intensity —dA/dt falls off as 1/r?, because the vector
potential A itself falls off as 1/r? at large distances from a current loop (Sec-
tion 7.8). Since the surface area S increases only as r?, the surface integral
decreases as 1/r* and vanishes as the surface of integration becomes infinite
Thus
Bie
dt
eeeHo | Bia (8-99)
On the right we have a partial derivative because B is a function of both the
coordinates and of the time. If we remove this derivative from under the
integral sign it becomes a total derivative that applies to all the field:

dW,
Sade pst ld 8-100
dé Qu at (|RE a!
Setting W,, = 0 when B = 0,
Wu = | B? dr. (8-101)
354 MAGNETIC FIELDS II

The quantity W,,, is the work that must be done to establish a magnetic field
in terms of the magnetic induction B, either in free space or in nonmagnetic
matter.

Just as in electrostatics, Section 2.15, we may define an energy density

Hee
dr
Ve
210
7 (8-102)
associated with each point in space. We shall arrive at this same result in
Section 8.10 from the value of the magnetic pressure.

Example For a long solenoid of length / and radius R, B is uoN’7 throughout

the interior, and zero outside. Then

We2uo /c-) B? dr = 2
pol?NUR?. (8-103)

8.5.2 Magnetic Energy in Terms of the Current Density J,

and of the Vector Potential A
It will be recalled from Section 2.15 that we expressed the energy density in
an electric field both as eE?/2 and as pV/2. We have already expressed the
magnetic energy density as B?/2u9; we shall now express it in terms of the
free current density J; and of the vector potential 4.
We rewrite Eq. 8-101 as follows:

We =f (B-V x A) dr. (8-104)

Ko Jo
Using the vector identity of Problem 1-19 and the divergence theorem,

Wn = xf A:(V X B) dr + | (4 X B)-da. (8-105)

2 Lo co 210 Ee}

The surface integral vanishes again as in Eq. 8-98, and when vy X B = Mods
(Section 7.6),
l
Le = 3 fs dr, (8-106)

where r is any volume that includes all regions where J; is not zero.
It is convenient to assign an energy density
dW, _ 1
Fi ey 5 Wy A) (8-107)

to conductors carrying a free current density J;.

8.5 ENERGY STORED IN A MAGNETIC FIELD 355

Example In the case of the long solenoid, the current is distributed, not over
a volume, but over the surface of the solenoid with a density \ =
N’'T. The vector potential A is parallel to J and is given by Eq. 8-76.
Then

Wn = ;/ NA da = 5 pol?NIR?. (8-108)

8.5.3 Magnetic Energy in Terms of the

Current I and of the Magnetic Flux ®
With filamentary circuits we may express the energy in terms of the current J
and of the flux ® linking the circuit. Replacing J; dr in Eq. 8-106 by J dl,
dl being an element of the circuit that carries the current J,

1 1
Va sip A-al = 5/8, (8-109)
from Eq. 7-60. The positive directions for J and ® are related according to
the right-hand screw rule.

Example For the long solenoid,

Wn = 5ie= 5INB(rR?) = Bwol2NIR2, (8-110)

8.5.4 Magnetic Energy in Terms of the Currents

and of the Inductances
It is possible to express the energy stored in a magnetic field in still another
way, in terms of the currents and of the inductances. Since the magnetic
flux © can be replaced by the product of the self-inductance L and the cur-
rent J, from Section 8.42,

Wn = 5Lr (8-111)
For two circuits carrying currents /, and /h,

] l
Wn = 2 [Pa “- wy I,®,, (8-1 12)

where ®, and , are the total fluxes linking circuits a and b respectively and
consist of contributions from both circuits:

P, = Baa oe Poa, (8-1 13)

Py == Dap ar Py, (8-1 14)

356 MAGNETIC FIELDS II

Thus
1 1 1
Wr = 5Ube oe 5)TPy_ + 5)IPreg 0)TyPap. (8-115)

But
Di a Mi,, Pay —_ MI,, (8-116)

Daa = Lele Pp, = Iblbe (8-117)

and therefore

a Salt + eLalt + MI,S,. (8-118)

The first two terms on the right are the self-energies due to the interaction
of each current with its own field, whereas the third term is the interaction
energy due to the mutual inductance.

Example Returning once more to the long solenoid, its stored energy is

Wn = LP, (8-119)

and, using the value of L given in Eq. 8-62, we again find that

Ve 5 Hol*N"IR?. (8-120)

8.6 SELF-INDUCTANCE FOR A VOLUME

DISTRIBUTION OF CURRENT

The self-inductance of a circuit comprising currents distributed over a finite

volume is defined from the energy stored in the system:

We x | ey een (8-121)
2140 * 2

Then the self-inductance is

L= ae / B? dr, (8-122)
Mol Jj

where the integral is evaluated over all space.

Example SELF-INDUCTANCE OF A COAXIAL LINE

We assume that the frequency is low enough to ensure that the cur-
rents are distributed uniformly throughout the cross sections of the
conductors, and we neglect end effects for simplicity.
We shall calculate successively the magnetic energies per unit
 357

Figure 8-12. Coaxial line of radii

a, 6, ¢ carrying currents J in
opposite directions in the mner
and outer conductors.

length of the line in regions 1, 2, 3, 4 as in Figure 8-12, and then set

the sum of these energies equal to (1/2)L’/*. This will give us the
inductance L’ per unit length.
(a) In region 1, from the circuital law (Section 7.7) applied to a
path of radius p,
27pB = pol ™p
2
ae (8-123)
and

= OF bolo Hol?
=. -124
Wm all (ue MOE (G12)
(b) In region 2,
Ba Bol, (8-125)
27p
2
Ve ae (8-126)
4r a
(c) In region 3, the current within a circular path of radius p is
that in the center conductor, namely J, less that part of the current
that lies in the outer conductor between the radii b and p. Thus
LEN [7Se (2, es= )}
a (8-127 )

_ bol (S = a (8-128)
orp Xce —ob4
and 3¢2 — b
Wa =
wolf
Ar
cf
Ee — 5)? In b
. c¢ 3c2— B i 8-129
4(c? — BD?
(d) In region 4, B = 0 and W,,4 = 0.
Finally, the inductance per unit length is
(8-130)
Li =
nt 2(Wima a0 Wa
7=e Wns + Wma)

fel fl b ih | ed Cm. amen oS

Te oe net ol (@o bia Ae — BDA
(8-131)
358 MAGNETIC FIELDS II

The second term within the braces is normally the most important;
it gives the inductance associated with the magnetic energy in the
annular region between the conductors.

8.7 MAGNETIC FORCE BETWEEN TWO CIRCUITS

Although the magnetic force law stated in Eqs. 7-1 and 7-5 was the starting
point for Chapter 7, we have not really used it to calculate the magnetic
force between two circuits, because of the difficulty of evaluating the line
integrals.
We are now in a position to find other expressions that are more con-
venient to use because they are based on mutual inductance. Although mutual
inductance is just as difficult to calculate, because it too involves a line
integral (Eq. 8-52), it is easily measurable.
We shall use the same method as in the example on page 78, where we
calculated forces on conductors. We assume a small virtual translation of
one coil, without any rotation, and then apply the principle of the conserva-
tion of energy:
Work done by the sources = Increase in magnetic energy
+ Mechanical work done.

For simplicity we disregard Joule losses. We also assume that the displace-
ments are made infinitely slowly so as to avoid taking kinetic energy into
account.
We consider two loops carrying currents J, and J, in the same direction
as in Figure 8-13. Since the forces shown in the figure indicate that the loops
tend to move toward each other, the coils can be kept fixed in space only if
the magnetic forces are balanced by equal and opposite mechanical forces.
The virtual displacement can be made in any convenient way: we may,
for example, assume that either the currents or the flux linkages are kept
constant. Whatever the assumption, the result must be the same, since the
force acting between two fixed circuits obviously has some single definite
value. We had a similar situation in the example on page 78 and in Problem
2-37, where we found the force on a capacitor plate from the conservation of
energy, assuming first that the plates were insulated, and then assuming that
they were connected to a battery. The forces were found to be the same.
 359

Figure 8-13. The two loops carry currents J, and J,. Typical lines of B originating
in a are shown linking b. Observe that the elementary forces dF have a compo-
nent in the direction of coil a, with the result that the total force F is attractive.
To calculate F, coil b is assumed to be displaced parallel to itself by a distance
dr into the position shown by the broken line, and the mechanical work done,
F-dr, is found from the principle of conservation of energy.

8.7.1 The Magnetic Force When the Currents

Are Kept Constant
Let us calculate the magnetic force F,, on loop b with the above method,
assuming that the currents are kept constant. We imagine that loop b is
allowed to move a distance dr toward loop a. Since the currents are kept
constant, only the interaction energy changes; the self-energies remain con-
stant. Hence, from Eq. 8-118,
dW, = I,l, dM, (8-132)
— He dPig = I, dP», (8-133)

where ®,, is the flux originating in b and linking a, and similarly for Pas.
The mutual inductance M and the fluxes ,, and ®,, all increase in this case,
so that dW,,, is a positive quantity, and the stored magnetic energy increases.
The mutual inductance is positive here.
Consider next the work done by the sources producing the currents in
the loops. In loop 6, ®,» increases, and the induced electromotance is in
such a direction that it would, by itself, produce a magnetic field opposing Ba.
360 MAGNETIC FIELDS II

Therefore the induced electromotance in loop 6 must tend to oppose the

current J;. If J, is to be kept constant, the source voltage must be increased
at each instant by d®,,/dt. The source in loop 6 therefore supplies an amount
of work
as oe He oa laden (8-134)
ip ne (8-135)
By symmetry, the work supplied by the source in loop a is the same, and the
total work supplied by the sources is
dW, = 2I.t, dM, (8-136)
which is exactly twice the increase in magnetic energy. The remainder has
gone into mechanical work.
In other words, the mechanical work is accompanied by an equal in-
crease in magnetic energy, both being supplied by the sources. See Problem
2-37.
The mechanical work done is
F.y-dr = In, dM, (8-137)
= (dWr)r, (8-138)
where F,,, is the force exerted by circuit a on circuit b. The subscript J indi-
cates that the currents are kept constant.
Since the quantity on the right is positive in the present case, the scalar
product F,,-dr must be positive, and the force F,, must point toward coil a,
like dr. This is correct, since we found at the beginning that the coils tend
to move closer to each other.
Note that we would have obtained the wrong sign if we had neglected
the work done by the sources.
In the general case, the x-component of F,,, is

aM
Fave = Inte ae (8-139)

where the increment of x is the x-component of dr. Also,

fan, (28) = (Me)

IPD ga OW,
ceo
The partial derivatives are evaluated for constant currents.
Although we have considered a system comprising only two circuits,
the procedure remains the same for any number. The only difference is that
the expression for the force F,,, must then be replaced by a sum of such
8.7 MAGNETIC FORCE BETWEEN TWO CIRCUITS 361

terms. Stated in another way, the component of force in a given direction

on the 7th circuit is the product of its current by the space rate of change of
flux linking it when it is displaced in the specified direction, all the currents
in the system being held constant.
We now have two expressions for the force between two current-
carrying circuits: Eq. 7-1, or 7-5, and Eq. 8-137. Let us see whether we can
transform the latter equation into the former. We use the Neumann formula,
Eq. 8-52. Then
Leno dl,-dl,\
Fay-dr = 7 dole App pane ) (8-141)

It is to be understood that the differential of the term within the parentheses

on the right-hand side must correspond to the displacement dr of the coil b,
as in Figure 8-13. Interchanging the order of the d operator and of the line
integrals,

Fy-dr = Ar
tale D
aJb
(Pe
lk
(8-142)
In the process of moving coil 6 by a distance dr, both dl, and dl, remain
unaffected, and the first d under the integral signs therefore operates only
on the 1/r factor. If we now define r; as a unit vector pointing from a to b
in the direction of dr,

a(*) =-9=-"%,
l ~ har 2 _nidr
(8-143)
q

and

Ear TT, pp ais r,-dr. (8-144)

Aor ffi. © aR

Since the term dr on the right is independent of both dl, and dl,, it can be
removed from under the integral sign, and, finally,

oete
Nias = ri(dl, + dly) 8-145
Idk fp 2 : ( )

This is exactly Eq. 7-5, which is an alternative form of the magnetic force
law, Eq. 7-1. We have therefore rediscovered the law that was the starting
point for our discussion of magnetic fields in Chapter 7.

Example FORCE BETWEEN TWO COAXIAL SOLENOIDS

The case of two long solenoids, one of which extends a distance /

within the other as in Figure 8-14, will serve as an example. We must
again neglect end effects, for otherwise the calculation would be
362

Figure 8-14. Two coaxial solenoids of different lengths and approximately

equal diameters, the smaller one penetrating a distance / inside the other.
There is an axial attractive force F when the two currents J, and J; are
in the same direction. The solenoids have N/ and N; turns/meter,
respectively.

much more complex. We therefore assume that the solenoids are

long and thin. We also assume that their diameters are approxi-
mately equal and that the currents flow in the same direction, as
in the figure.
The mutual inductance is

M=
(ND®oo _ (NiDPov (8-146)
ie | ae
where N; and Nj are the numbers of turns per meter on the two sole-
noids. Or, setting S to be their cross-sectional area,
BMoNENIDS1 = HoNNelSI
M : : (8-147)
= woNiN, SI, (8-148)
OM Ir
are 2 KoNLN;S, (8-149)

F = poNiN SIs. (8-150)

The force is axial and it is attractive, since M increases with /.

Let us repeat the calculation, starting from Eq. 8-101:

1 z
(8 151)
Wan = —
20 i B 2 dr,

1 : :
(8-152)
== Pita [Bi(ta — SI) + Biro — SD) + (Ba + Bs)2S/],

“95,
ce © Bir» + 2B.BpS)), (8-153)
1 2 y,

where B, = uoNZ/, is the magnetic induction originating in solenoid

a and, similarly, B, is the magnetic induction originating in solenoid
b. These are constants, since the currents J, and J, are again assumed
to be constant. Then
8.7 MAGNETIC FORCE BETWEEN TWO CIRCUITS 363

dW,, = Fadl = aoa (8-154)

0

F = poNINESIalp. (8-155)
Note that the magnetic energy W,, is a function of / because
the magnetic energy density depends on the square of B. If it de-
pended on the first power of B, W,, would not be a functiom of J/,
and the force would be zero.

8.7.2 Magnetic Force When the Fluxes

Are Kept Constant
Imagine that during the virtual displacement we adjust the currents So as to
keep the flux linkages constant. There is then zero induced electromotance,
and the sources supply zero energy to the system, except for the Joule losses,
which we are neglecting. Then

F.:dr = (Aan) es (8-156)

and
ow
Fags ihe (=). (8-157)

In this case the mechanical work is accompanied by a decrease in magnetic

energy.

Example FORCE BETWEEN TWO COAXIAL SOLENOIDS

To calculate the force with the above method we can start with
Eq. 8-151 for W,,. Thus

[2B.(dBa)ta + 2B,(dBs)T» + 2(dB.) BS! + 2B.(dBy)SI + 2BaByS dl).

(8-158)
Since the total flux linkage for solenoid a is constant,

BN iS + BiNa .S = Constant, (8-159)

or
Bat Bt = Constant, (8-160)

and
dB, + dB, ;+ By ;dl = 0. (8-161)

The symmetrical equation for solenoid b is obtained by permuting

the indices. Solving these two equations yields
364 MAGNETIC FIELDS II

i a, dl, (8-162)
CB ee pare
and a symmetrical expression for dB. Substituting then these two
equations into Eq. 8-158, and remembering that 7, is /,S while 7,
is /,S, we find again an axial force

F= usBABS. (8-163)
Ho

8.8 MAGNETIC TORQUE

To calculate a magnetic torque we again use the method of virtual work,

and, by analogy with Eqs. 8-139, 8-140, and 8-157,

r= 1h, (8-164)
a (32). (8-165)

~ (FP). (8-166)
= - (2) (8-167)
where I’ is the torque and © is the flux linking the circuit.

Figure 8-15. Loop carrying a

current / in a uniform magnetic
field B. With the current in the
direction shown, the loop is subject
to a torque tending to increase the
angle @.
8.9 MAGNETIC FORCES WITHIN AN ISOLATED CIRCUIT 365

Example MAGNETIC TORQUE ON A CURRENT LOOP

Figure 8-15 shows a rectangular loop carrying a current J and set at
an angle @ in a uniform magnetic field B. The torque is
te)
a 30 (®) — BS cos 8), (8-168)
where ®) is the flux produced by the current J in the loop, and S is
the area of the loop. The flux BS cos 6 is negative because it links
the loop in the negative direction with respect to the current J. Thus
is Sisinigs (8-169)
The torque is positive (in the direction of increasing 6), for
0 < 6 < m. This result can be easily verified from the direction of
the elementary forces J(dl X B).

8.9 MAGNETIC FORCES

WITHIN AN ISOLATED CIRCUIT

In an isolated circuit the current flows in its own magnetic field and is there-
fore subjected to a force. This is illustrated in Figure 8-16: the interaction
of the current J with its own magnetic induction B produces a force that has
the effect of extending the circuit to the right.
Magnetic forces within an isolated circuit are calculated by using either
the magnetic force law or the principle of conservation of energy. In the case
of Figure 8-16, it is not practical to use the former method because it would
be too difficult to calculate the values of B and J inside the moving wire. The
energy method is therefore indicated. See Problem 8-28.

Figure 8-16. Schematic diagram of a rail gun. The battery A charges, through a
resistance R, the capacitor C, which can be made to discharge through the line by
closing switch S. The role of the capacitor is to store electric charge and to supply
a very large current to the loop for a very short time. If side D is allowed to
move, it moves to the right under the action of the magnetic force F’, Compare
this figure with Figure 8-3.
366 MAGNETIC FIELDS II

Example THE PINCH EFFECT

An individual ion moving at the periphery of an ion beam, as in
Figure 8-17, is subjected to two forces: the outward electrostatic
repulsion F, and the inward magnetic force F,,,. The magnetic force
thus tends to “‘pinch,’”’ or to concentrate the beam along its axis.
At lJow velocities the magnetic force is negligible and the beam
diverges. If the axial velocity is sufficiently small, the radial velocity
can be of the same order of magnitude or even larger, in which case
the beam disappears.
We shall calculate F, — F,, on the assumption that the charge
and current densities have cylindrical symmetry.
If the total current is / and if the radius of the beam is R, then

m= = Mol
See OF, (8-170)
8-170

since the magnetic induction pol/27R is perpendicular to the ve-

locity v.
Also, if Q’ is the electric charge per meter in the ion beam,

a es
IreaR ‘
(8-171)

according to Gauss’s law. Since J = Q’v,

i=meet!
ye (8-172)

Figure 8-17. The pinch effect. In an ion beam the individual ions are
subject to an outward electrostatic force F, and to an inward magnetic
force F,,. At low velocities F,, << F, but, as the velocity v approaches
the velocity of light c, F,, approaches F,. If the electrostatic repulsion is
cancelled by mixing ions of the opposite charge in the ion beam, Fn
acts alone and the beam contracts. This phenomenon is often observed
in positive ion accelerators. Residual gas in the path of the ion beam is
ionized by impact, and the resulting low-energy electrons are trapped
in the positive beam while the low-energy positive ions drift away. Thus,
if the vacuum is not too good, the positive space charge in the beam
can be reduced and the focusing improved. This phenomenon is known
as the pinch effect, or as gas focusing.
8.10 MAGNETIC PRESSURE 367

and the net outward force is

ee a
IQ
185, = QmeoRv a w €oMov”), (8-173)

- fe v?
~ QarepRv ey te)
The net force tends to zero for v > c. This result is valid even if
Vv SC.
If the kinetic energy of the particle is equal to its rest energy
(5 X 10° electron-volts for an electron and 10° electron-volts for a
proton), then the net force F, — F,, is smaller than F, by a factor
of four (Problem 5-24).
In practice, a vacuum is never perfect. Let us say the ions are
positive. If their energy is of the order of tens of electron-volts or
more, they ionize the residual gas, and the resulting positive ions drift
away from the positive beam. The low-energy electrons, however,
remain trapped in the beam and neutralize part of its space charge,
thereby reducing F,. This type of pinch effect is often called gas
focusing.

8.10 MAGNETIC PRESSURE

Except for the ion beam of the last example, we have considered as yet only
the magnetic forces acting on conducting wires. If, instead, we have a current
Sheet, then it is appropriate to think in terms of magnetic pressure.
Let us imagine a flat current sheet carrying \ amperes/meter and
situated in a uniform tangential magnetic field B/2 due to currents flowing
elsewhere, as in Figure 8-18a, with A normal to B. If \ is very small, the
magnetic field is not disturbed appreciably by the current sheet and the force
per unit area is AB/2.
We now increase \ until it produces an aiding field B/2 on one side and
an opposing field B/2 on the other side, as in Figure 8-18b. Then, from Prob-
lem 7-27, \ = B/o, and the force per unit area, or the pressure, is B?/2,0.
We have arrived at this result for a flat current sheet, but the same applies
to any current sheet when B = 0 on one side.
This value of the magnetic pressure permits us to calculate the energy
density (Sections 8.5 to 8.5.4) in still another way. Imagine that the current
sheet moves back infinitely slowly by a small distance x. Then the work
performed by the magnetic pressure on a small area a of the current sheet
is ax(B?/2u9). This work is supplied by the sources that maintain B constant,
and it is equal to the increase in the energy stored in the field (Section 8.7.1),
368

(a) (b)

Figure 8-18. (a) A current sheet carries a very weak current density of \ ampere
per meter of its width and is situated in a uniform magnetic field B/2. (b) The
total magnetic field is that of figure (a), plus that of the current sheet. Selecting
so that it just cancels the magnetic field on the left-hand side of the sheet, we have
a total field B on the right.

10

log P 8
(newtons/ meter
or ’
joules/meter) 6

Ll ee | 1 =
2 3
log B (teslas)

Figure 8-19. Magnetic pressure B?/2uo as a function of B. The magnetic

pressure is equal to the magnetic energy density.

or ax times the energy density. Then the energy density is B®/2uo, as j

Section 8.5.1.
We arrived at a similar result in Section 2.16 when we deduced the
8.10 MAGNETIC PRESSURE 369

energy density in an electric field from the electric force exerted on the
surface of a conductor. The remarks we made at that time concerning this
approach to the energy density in a field are equally valid here.
Figure 8-19 shows the magnetic pressure, or the magnetic energy density
B?/2yuo as a function of B.

Example MAGNETIC PRESSURE

INSIDE A LONG SOLENOID
Figure 8-20 shows an end view of a solenoid. We shall use the
method of virtual work to show that the magnetic pressure is B?/2,o.
Imagine that the current is maintained constant while the mag-
netic pressure increases the radius of the solenoid from R to R + dR.
Then the magnetic energy W,,, which we calculated in Section 8.5,
increases by
dW,, = tuol?NIR dR, (8-175)
and the work performed by the magnetic pressure P is 27R/(dR)P.
During the expansion, a voltage Nd®/dt is induced in the sole-
noid; according to Lenz’s law, this induced voltage opposes the in-
crease in flux, and hence opposes the applied voltage. So as to main-
tain the current J constant, the source must supply an extra voltage
Nd®/dt and an extra power NI d®/dt. Then the extra energy sup-
plied by the source during the expansion is
6
NI d® = N'UId(rR0N'T) = 2rpol2N”IR AR. (8-176)
One half of this energy is transformed into the magnetic energy
dW,,, and the other half into the mechanical work performed by P.
Thus
2mRI(dR)P = Tyol*N”IR AR, (8-177)
a bol2N”? _ B :
P= ore he OH (8-178)

since B is polN’.

Figure 8-20. End view of a solenoid. The

dots represent the lines of B, which are
normal to the paper, and the arrows show
the direction in which the magnetic
pressure is exerted.
370 MAGNETIC FIELDS II

8.11 SUMMARY

This chapter is concerned with (a) the nonconservative electric fields asso-
ciated with time-dependent magnetic fields, (b) the energy stored in magnetic
fields, and (c) the forces exerted on current-carrying circuits.
For a closed circuit,

db
fpral ear (8-2)

d
= -4 | Baa, (8-3)

where E is the total electric field intensity, including the induced part. The
directions in which the line integral is evaluated and the flux ® is taken to be
positive are related according to the right-hand screw rule.
This is the Faraday induction law, and the line integral is the induced
electromotance.
The negative sign indicates that the induced electromotance tends to
oppose the change in flux. This is Lenz’s law.
Induced electromotance 1s observed whenever the magnetic flux linking
a circuit changes with time, whether the circuit moves relative to the mag-
netic field, or whether a fixed circuit is linked by a variable flux, as in a trans-
former.
In differential form, the Faraday induction law becomes

OB
V SE eee eat :
oF (8-8)

This is one of Maxwell’s four fundamental equations of electromagnetism.

It leads to the general expression for the electric field intensity

ee a (8-13)

where the first term is that part of E that is induced by changing magnetic
fields, and the second is the part that is produced by accumulations of charge.
For a conductor moving with a velocity u in a changing magnetic field,
Eq. 8-8 becomes
OB
VXE=-9 +7 x (xB) (8-36)
8.11 SUMMARY 371

when E is measured in a coordinate system moving at a velocity u relative

to that in which the magnetic induction is measured as B.
The mutual inductance M between two circuits is equal to the magnetic
flux linking one circuit per unit current flowing in the other:

Pap a Marla, (8-51)

where ®,, is the flux originating in circuit a and linking circuit b. It is found
that
bo dl,-dl,
Ma = A pp r 2 (8-52)

This is the Neumann equation. It shows that M depends solely on the geometry
of the two circuits, if there are no magnetic materials present. By symmetry,

May = Mya (8-54)

Self-inductance L is a similar quantity, which applies to a single circuit:

oe (8-57)

In terms of self-inductance, the induced electromotance is

fpeal = — die (8-58)

a similar relation applies to mutual inductance.

The coefficient of coupling k between two circuits is defined by

Moms (Leg? (8-82)

and can have values ranging from —1 to +1. It is zero if none of the flux
of one circuit links the other.
The energy stored in a magnetic field is equal to that required to establish
the field, and the energy supplied to an element of volume dr of conductor
per unit time is

dW,
SIMI as
OA
aeew, ; 8-9]
dt [% Jp ar Ce
If there are no magnetic materials present, and if V X B= oJ; is
applicable,

Wan = = | B? dr, (8-101)

HO Fo)

a 5| Aer (8-106)
372 MAGNETIC FIELDS II

where 7 is any volume which includes all regions in which J; is not zero.
We also find that

WV = 51®, (8-109)

where ® is the flux linking the current /. As usual, the positive directions for J
and © are related according to the right-hand screw rule. In terms of the
inductances L,, Ly, M for two circuits a and b,

ae 5 Lil + 5alt + MI,Is. (8-118)

The self-inductance of a volume distribution of current is defined from

its stored magnetic energy:
|
L= = |B? dr. (8-122)
Bol? ce)

The magnetic force between two current-carrying circuits can be cal-

culated from the principle of conservation of energy applied to a virtual
displacement of one of the circuits. The x-component of the force F,., exerted
by a on bis found to be

Faa= Llp sey (8-139)

ay (2s). (8-140)

= Ce *). (8-140)

= -(% ") (8-157)

the subscripts indicating which parameter is kept constant.

Equation 8-145 is an alternative of the magnetic force law stated in Ed,
7-1, which was the starting point for our entire discussion of magnetic fields.
Magnetic torque is given by

i= (dW,, ), (8-166)
as well as by other expressions similar to the ones above for Fy», with x
replaced by 90.
Magnetic forces also occur within an isolated circuit because a current
flows in its own magnetic field.
 PROBLEMS 373

Magnetic pressure on a conductor is observed whenever the current

density is not parallel to the local magnetic induction.
If the magnetic induction is B on one side of a current sheet and zero
on the other, then the magnetic pressure is equal to the magnetic energy
density B?/2p.

PROBLEMS

Not ex An asterisk indicates that the problem requires a knowledge of relativity.

/

CoE carriage runs on two rails on either side of a long tank of water equipped
or testing boat models. The rails are 3.0 meters apart and the carriage has a
maximum speed of 20 meters/second.
Calculate the maximum voltage between the two rails if the vertical com-
__ ponent of the earth’s magnetic field is 2.0 & 10-° tesla.
Fed conducting bar slides at a constant velocity u along conducting rails in a
/
{
~
region of uniform magnetic induction, as in Figure 8-3. The resistance in the
circuit is R and the inductance is negligible.
(a) Calculate the current J flowing in the circuit.
(b) How much power is required to move the bar?
(c) How does this power compare with the power loss in the resistance R?
8-3/ If a conductor is given an acceleration a, the conduction electrons are sub-
\ jected to inertia forces —ma.
Show that, if E’ is the equivalent total electric field intensity in the con-
ductor, then
ed
Ot
ee
e
es
This effect was predicted by Maxwell, and was observed for the first time
in 1916 by Tolman and collaborators. c
The inverse effect, namely the acceleration of a body carrying a variable
current was also predicted by Maxwell, and was first observed by Barnett and
others in 1930.
/In the beratron, electrons are held in a circular orbit in a vacuum chamber
by a magnetic field B. The electrons are accelerated by increasing the magnetic
flux linking the orbit.
Show that the average magnetic induction over the plane of the orbit
must be twice the magnetic induction at the orbit if the orbit radius is to re-
main fixed as the electron’s energy is increased.

8-5. A toroidal coil is fed a voltage V(t) by a variable power supply. The length
of the wire is /.
374 MAGNETIC FIELDS II

Discuss the electric field intensity E inside the wire and show that E is
not equal to V/1.
In Problem 8-18, we shall see that the magnitude of the extra term is
0A/ot.

8-6. Show that the electromotance induced in a loop rotating at an angular velocity
w in a magnetic field By sin wr, as in Figure 8-6, is
— BySw cos 2wt.

. A thin conducting disk of thickness h, diameter D, and conductivity o is

placed in a uniform alternating magnetic field B = By sin wt parallel to the
axis of the disk.
(a) Find the induced current density as a function of distance from the axis
of the disk.
(b) What is the direction of this current?

8-8. Electromagnetic flowmeters operate as follows. A fluid, which must be at least

slightiy conducting (the blood of an animal, for example), flows in a noncon-
ducting tube between the poles of a magnet. The ions in the fluid are then
subjected to magnetic forces Qu X B in the direction perpendicular to the
velocity u of the fluid and to the magnetic induction B. Electrodes placed on
either side of the tube and in contact with the fluid thus acquire charges of
opposite signs and the resulting voltage difference is a measure of u.
For example, if the tube has a rectangular cross section ab with B
parallel to side b, and if the velocity were the same throughout the cross
section, then the voltage between the electrodes would be uwBa.
Note that, in the electromagnetic flowmeter, ions of both signs have the
same velocity u and that the polarity of the electrodes is always the same for a
given direction of flow and for a given direction of B. Compare with the Hall
effect discussed in the example on page 300.
Faraday attempted to measure the velocity of the Thames river in this
way in 1832. The magnetic field was of course that of the Earth.
(a) In the absence of turbulence, the fluid velocity uw in a tube of radius a is of
the form

if = w(1 _ a@
es

where p is the radial coordinate.

Show that the difference in potential between two points situated at the
ends of a diameter perpendicular to B is then

where Q is the volume of fluid flowing through the tube in one second.
This result is in fact incorrect, as we shall see below.
(b) Sketch a cross section of the tube, showing by means of arrows of various
lengths the magnitude of u X B at various points inside the tube.
(c) The fact that w X B is not the same everywhere in the fluid will cause
currents to flow.
PROBLEMS 375

Sketch another cross section of the tube, showing the lines of current
flow.
The current drawn by the electrodes is negligible.
It turns out that this current causes an Ohmic drop that reduces the
voltage calculated under (a) above by one quarter, and that the correct volt-
age difference is
2Q0B
Ta
(d) Show that this is the voltage one would expect if the velocity u were uni-
form and equal to the average velocity, and if there were zero Ohmic drop as
above.
(e) We have neglected edge effects in the regions where the conducting fluid
enters into and emerges from the magnetic field.
Sketch lines of current flow in these two regions.
These currents further increase the Ohmic drop.

8-9. A particle of charge Q, mass m, and kinetic energy T describes a circle of

radius r in a uniform magnetic field B.
The magnetic field increases at such a rate that dr/r «<1 during one
revolution.
(a) Show that, during one revolution,
AB
INE = 71?°O & = 49h b)
B
and that, therefore,
PF
— = Constan
B tant.

Neglect relativistic effects.{

(b) If the charge of the particle is imagined to be spread out around the cir-
cumference of the circle, it corresponds to a current Qw/2r7.
Show that the corresponding magnetic moment is 7/B and is therefore
approximately constant.
(c) Show that the flux linked by the trajectory is constant.

*8-10. Repeat part (a) of the preceding problem without restricting your calcula-
tions to nonrelativistic velocities.
You should find that
AT = ;mo? a.

*8-11. In one type of ion drag accelerator a ring of high-energy electrons rotating in
a magnetic field and containing a small percentage of protons is accelerated
as a unit along its axis of symmetry.
A useful analogy for explaining the accelerating force on the protons is
that of amarble in a saucer in a railway train. If the acceleration is sufficiently
gradual, the marble stays in the saucer and acquires the velocity of the train.
the force on the protons is of course due to the electric attraction of the
electrons.
376 MAGNETIC FIELDS II

The main advantage of this type of accelerator will be demonstrated

under (g) below.
Let us assume that the ring is initially at rest and that the electrons have
a kinetic energy 7, of 1.50 million electron-volts. Protons are injected into the
ring by admitting a small amount of hydrogen gas, which becomes partly
ionized. The kinetic energies of the protons are negligible, initially.
The magnetic induction B, is 2.00 « 107? tesla.
(a) Show that the initial radius r; of the ring is 32.4 centimeters.
(b) The magnetic field is then increased by a factor of 50 to B, = 1.00 tesla.
Show that the radius ofthe ring shrinks to r2 = 4.6 centimeters, and that
the kinetic energy of the electrons increases to 13.4 million electron-volts.t
(c) Let us call the axis of symmetry of the ring the z-axis.
Show qualitatively that, if B is independent of the time, but decreases
in the positive direction of the z-axis, the ring is accelerated in that direction.{
(d) The ring therefore acquires an axial velocity. But the static magnetic field
cannot increase the kinetic energy of the particles, because the magnetic force
Qu X B is perpendicular to the velocity u.
Then how can the ring acquire an axial velocity?
(e) Show that, in a plane where the axial component of B is B;., the ring has a,
radius
Gam

[sy =I fb) = :
B3,

Thus, at a plane where B3, is 0.2 tesla, r3 is 4.6 X 5!/2 = 10.3 centimeters.
This process is called expansion acceleration.t
(f) Show that y. corresponding to the axial velocity obeys the following
equation:
Yor (3)
VY 22 at B3:

(g) The protons have the same y. as the electrons, and their transverse velocity
is negligible.
(h) Show that the protons have a kinetic energy of 1.2 < 10° electron-volts
in the plane where Bz, is 0.2 tesla.
The protons therefore acquire energies of the order of 40 times that of
the electrons. This is the major advantage of this type of accelerator.
The ring could also be accelerated in an electric field (electric accelera-
tion). For example, if, the ring were accelerated through a difference of poten-
tial of 10 million volts, the protons would acquire a kinetic energy of 400
million electron-volts. Heavy ions would gain even higher energies.

. A wire bent in the form of a circle of radius R is placed so that its center is
at a distance D = 2R from a long straight wire, the two being in the same
plane.
Show that the mutual inductance is 0.268 yoR.
PROBLEMS
377

8-13. Two long parallel rectangular loops lying in the same plane have lengths /,
and /, and widths w; and we, respectively. The loops do not overlap, and the
distance between the near sides is s.
Show that the mutual inductance between the loops is

Mol» | S++ We
M=

20 a Ww
(1 sae
if 4 < h, and if the loops have but a single turn. Neglect end effects.

8-14. The coefficient of coupling k between two single-turn coils was defined in
Section 8.4.3. In general, ka ¥ ky.
Show that
ka _ Ly
ky ibe,

8-15. Show that the self-inductance of a close-wound toroidal coil of m turns, and
of major radius R and minor radius r is

L = pon? {R — (R? — r?)/2,

8-16. It is shown in Section 8.7.1 that, if one active circuit is allowed to move with
respect to another, the mechanical work performed by the sources is equal to
the increase in magnetic energy. The currents are assumed to be maintained
constant. Hence the force between two active circuits is given by the rate of
change of magnetic energy.
Show that, similarly, if the geometry of an isolated active circuit is
altered, the mechanical work performed by the source is equal to the increase
in magnetic energy. Assume again that the current is kept constant. It follows
that, on this assumption, the force on an element of an active circuit is given
by the rate of change of magnetic energy. t
See also Problem 8-28.

8-17. It is known that high-frequency currents do not penetrate into a conductor

as do low-frequency currents. This is called the skin effect.
Would you expect the self-inductance of a coaxial line to increase or to
decrease with increasing frequency?

8-18. (a) Show that the magnitude of the vector potential at the surface of a close-
wound toroidal coil of circular cross section is /27r, where ® is the magnetic
flux inside the coil and r is the minor radius of the toroid.
(b) Show that the magnitude ofthe extra term found in Problem 8-5 is 0.4/0t.

8-19. Show that the power lost by Joule heating per cubic meter in a conductor is
EJ;, or J?/c, where J; is the current density and a is the conductivity.

£20, compare the energies per unit volume in (a) a magnetic field of 1.0 tesla and
(b) an electrostatic field of 10° volts/meter.

8-21. Calculate the magnetic energy in a toroidal coil in at least two different ways.

8-22. Consider n rigid fixed circuits, the ith circuit carrying a current J/;.
378 MAGNETIC FIELDS II

Show that the magnetic energy associated with the system can be written
as a sum of self-energy terms of the form (3) L;J?, plus a sum of mutual energy
terms of the form M,,/;J;, each pair of currents appearing once.
*8-23. In the examples on pages 263 and 270 we transformed the field inside a
parallel-plate capacitor.
(a) Show that the field energy in frame 1 is y(1 + 6”) times that in frame 2.
(b) How do the energies & and & compare in the case of the solenoid of
Problem 6-15?
8-24. A coil of resistance R and inductance L is fed by a de power supply.
(a) Show that the power expended by the source when the voltage changes is
db
PR+ é G LI?) weit Tier
(b) Show that this is in agreement with Eq. 8-89.
8-25. Electromagnetic levitation forces are sometimes used to support objects in
space. For example, if a metal object is situated near a coil carrying an alter-
nating current J, eddy currents will flow in the object and there will result a
repulsive force.
Show that the force in the direction of x is

if the object is allowed to move in the x direction, and if the effective induc-
tance of the coil is L.t

8-26. (a) Show that a current-carrying coil tends to orient itself in a magnetic field
in such a way that the total magnetic flux linking the coil is maximum.
(b) Show that the torque exerted on the coil is m X B, where m is the mag-
netic moment of the coil and B is the magnetic induction when the current
in the coil is zero.

8-27. Many satellites require attitude control to keep their instrumentation properly
oriented. For example, communication satellites must keep their antenna
systems on target. Attitude control requires a mechanism for exerting appro-
priate torques as they are required.
Attitude control can be achieved by means of coils whose magnetic fields
interact with that of the Earth.
(a) Show that the torque exerted by such a coil is
NIBA sin 0,
where WN is the number of turns, / is the current, B is the magnetic induction
due to the earth, A is the area of the coil, and @ is the angle between the earth’s
magnetic field and the normal to the coil.
(b) Calculate the number of ampere-turns required for a coil wound around
the outside surface of a satellite whose diameter is 1.14 meters. The torque
required at 6 = 5 degrees is 10~* newton-meter, and the magnetic induction
at an altitude of 700 kilometers over the equator, where the orbit of the satel-
lite will be situated, is 4.0 X 10-5 tesla.
PROBLEMS 379

8-28. Let us calculate the force on the movable link D in the circuit of Figure 8-16.
Instead of being a metallic rod, the link can also be an electric arc. The
device then accelerates blobs of plasma and is called a plasma gun. Such
plasma guns can be used as rocket motors.
We cannot solve this problem by integrating the magnetic force J * B
over the volume of the link, since neither J nor B are known, or even easily
calculable. Instead, we shall find the force by investigating the magnetic and
mechanical energies involved.
If we set x = 0 at the initial position of the link, then the inductance L
in the circuit is Lyp + L’/x, where L’x is the inductance associated with the
magnetic flux linking the rails.
During the discharge of the capacitor C we can neglect the battery and
its resistance R, since the current flowing in that part of the circuit is negligible.
(a) Show that the force F on the link D is (i2/2)L’, where /is the instantaneous
current flowing in the right-hand side of the circuit.
Call the resistance on that side R’.t
(b) Show that, during the discharge,
COmmUOdE
__— ee
dQ
ptt 2D =—a()s

Vere Thea Wee

where Q is the charge on the capacitor at the time rf.
8-29. A superconducting power transmission line has been proposed that would
have the following characteristics. It would carry 10! watts at 200 kilovolts
dc over a distance of 10’ kilometers. The conductors would have a circular
cross section of 5 centimeters? and would be held 5 centimeters apart, center
to center.
(a) Find an approximate value for the magnetic force per meter.{
(b) Calculate B?/2u at the surface of a conductor, on the inner side.
(c) Calculate the stored energy and its cost at 0.5 cent per kilowatt-hour.
The inductance of such a line is 0.66 microhenry/meter.

. We have seen in the example on page 366 that an ion situated at the periphery
of an ion beam of radius r is subjected to a force

LO “)
27 €or CG

(a) Show that the divergence of the beam is given by the following equation:
1/2

dt ro

where ry is the radius of the beam at a point where the radial velocity is zero,
and where

en I
TEUIN
(8) vy
(
Sa

You can solve this differential equation by referring to Problem 6-23(b).

*(b) Show that this differential equation is equivalent to the equation
380 MAGNETIC FIELDS II

dr [ edo r ]1/2
— = In
dty TEM Fo
which we found in Problem 6-23 for a pulsed beam of electrons of charge e.

*8-31. Show that F,, is equal to F, (Figure 8-17) within 10% when the beam energy
(the kinetic energy of a particle) is at least equal to twice the rest energy of a
particle, approximately.
For electrons, this beam energy is 1 MeV, and for protons it is 2 GeV.

8-32. Check the graph of Figure 8-19 and show that magnetic pressure is equal to
one atmosphere when B is only about 0.5 tesla, or 5000 gauss.

8-33. (a) Show that magnetic pressure is about 4B? atmospheres, if B is expressed
in teslas.
(b) Draw a log-log plot of the electric force per unit area }¢9E? as a function
of E.
(c) Discuss the similarities and differences between the magnetic pressure and
the electric force per unit area.

8-34. A thin tube of wall thickness ¢ and inside radius R carries a current /.
(a) Show that the magnetic pressure is directed inward and is (uo/87?)(//R)?.
(b) Calculate the pressure in atmospheres for a current of 1000 amperes and a
tube of 1 centimeter radius.

8-35. Magnetic fields are used for performing various mechanical tasks that require
a high power level for a very short time.
For example, the magnetic pressure inside a coil can be used to crush a
light aluminum tube that acts as a shutter to turn off a beam of light or of
soft X-rays. When the coil is suddenly connected to a charged capacitor, the
induced electromotance d®/dt produces a large current in the aluminum tube,
which collapses under the magnetic pressure.
Let us calculate the pressure exerted on a conducting tube placed inside
a long solenoid. If the current 7 in the solenoid is increased gradually from
zero to some arbitrarily large value, the current induced in the tube is small
and the magnetic pressure is negligible. Let us assume that d//dr is so large
that the induced current in the tube maintains zero magnetic field inside it.
Then there is a magnetic field B only in the annular region between the sole-
noid and the conducting tube.
(a) Calculate the pressure on the tube for B = 1 tesla.
(b) What would be the pressure if the conducting tube were parallel = the
axis, but off center?
8-36. Extremely high magnetic fields can be obtained by discharging a capacitor
through a low-inductance coil. The capacitor itself must of course have a low
inductance. Such fast capacitors cost approximately one dollar per joule of
stored-energy capacity.
(a) Estimate the cost of a capacitor that could store an energy equal to that
of a 100 tesla (one megagauss) magnetic field occupying a volume of one
liter.
 PROBLEMS 381

(b) Estimate the cost of the electricity required to charge the capacitors.
(c) Calculate the magnetic pressure in atmospheres, at 100 teslas.
ee eg ah
(. 8-37 Flux compression is one method of obtaining large values of B. For example,
ee cs
a magnetic field Bo is established inside a conducting tube of radius Ro, and
the tube is then imploded by means of an explosive placed around the tube.
Currents flow in the tube, and the internal magnetic pressure builds up until
it is equal to the external gas pressure.
(a) Show that, if the radius of the tube shrinks very rapidly,
; B = Bo(Ro/R)?,
where B is the magnetic induction when the radius has been reduced to R.
(b) Show that the surface current density in the tube must be 10° am-
peres/meter to achieve a field of 10° teslas.
(c) If the initial B is 20 teslas, and if the tube has initially a diameter of 20
centimeters, what should be the value of B when the tube is compressed to a
diameter of about 5 millimeters?
(d) ‘Calculate the resulting increase in magnetic energy, and hence the explo-
sive energy, required to compress the field. Assume that the cylinder is 20
centimeters long, and neglect end effects. 2,
(e)) Flux compression can also be achieved by means of a conducting piston
sKot axially into a solenoid.
If the radius of the solenoid is Ro, and if the radius of the piston is R,
show that the magnetic induction in the annular region between the piston
and the solenoid reaches a value of
Bo

- (Ri)
Hints

8-4. Relate the centripetal acceleration to the magnetic force acting on the electron, and
then find the condition that lets the linear momentum of the electron increase with
fixed orbit radius. Use Newton’s second law with the tangential force on the electron
given by the Faraday induction law. |

8-9. (a) If you find a AT that is too large by a factor of 2, you have probably made an
approximation that is not legitimate. The increase in kinetic energy AT is the work
done by the force Q(0A/d1).

8-10. When the magnetic field is constant, the mass of an electron remains constant and
the magnetic force BeV is equal to the centripetal force mv’/r, as for the nonrela-
tivistic case, except that m is the relativistic mass mo.

. (a) See the hint for the preceding problem.

(b) Use the result of the preceding problem.
(c) Draw a set of lines of B in a plane containing the axis. See also Problem 7-22.
382 MAGNETIC FIELDS II

(e) Remember that the particles are submitted to a central force. What can you say
about their angular momentum?
(f) Use the fact that yo2y29 = y3zv3e = y, aS demonstrated in Problem 5-14, and
remember that v9 v39 & c in the reference frame of the ring.

8-16. Consider the general case of a single-turn circuit.

8-25. Consider the object as a secondary winding whose impedance Is jw». (The resistance
_ R, is made negligible by selecting a high enough frequency.) The effective inductance
L of the primary is then L,(1 — k?), and displacing the object changes k. Then calcu-
late the work done by the source in the primary for a small Ak which takes place
during the time t = 0 to t = T, where T is the period 27/w.*
8-28. (a) Equate the power supplied by the capacitor to the rate at which the energy
increases in the right-hand side of the circuit. The voltage across a variable induct-
ance is d(Li)/dt.

8-29. Assume that the current is concentrated along the axes of the conductors. Can you show
that this gives the correct result?

* For a more general proof of this result, see Sir James Jeans, The Mathematical Theory
of Electricity and Magnetism, Fifth Edition, Cambridge University Press (1951), Paragraph
564.
CHAPTER Y)

MAGNETIC FIELDS III

Magnetic Materials

Thus far we have studied only those magnetic fields that are due to the motion
of free charges. Now, on the atomic scale, all bodies contain spinning elec-
trons that move around in orbits, and these electrons also produce magnetic
fields.
Our purpose in this chapter is to express the magnetic fields of these
atomic currents in macroscopic terms.
Magnetic materials are similar to dielectrics in that individual charges
or systems of charges can possess magnetic moments, and these moments,
when properly oriented, produce a resultant magnetic moment in a macro-
scopic body. Such a body is then said to be magnetized.

The so-called transition elements, like man-

ganese, are examples of paramagnetic substances. When such a substance
is placed in a magnetic field, its atoms are subjected to a torque that tends
to align them with the field, but thermal agitation tends to destroy the align-
ment. This phenomenon is analogous to the alignment of polar molecules in
dielectrics (Section 3.9).
In diamagnetic materials, the elementary moments are not permanent
but are induced according to the Faraday induction law (Section 8.1). All
materials are diamagnetic, but orientational polarization may predominate.
Most magnetic devices use ferromagnetic materials, such as iron, in which
the magnetization can be orders of magnitude larger than that of either para-
or diamagnetic substances. This large magnetization is attributed to electron
spin and is associated with group phenomena in which all the elementary
 384 MAGNETIC FIELDS II

moments in a small region, known as a domain, are aligned. The magnetiza-

tion of one domain may be oriented at random with respect to that of a
neighboring domain izati

Although the behavior of magnetic materials seems at first glance to be

analogous to that of dielectric materials, there is one important difference.

e calculation of the fields associated with magnetic materials is there-

for largely empirical.

Oo) THE MAGNETIZATION M

If m is the average magnetic dipole moment per atom, and if N is the number
of atoms per unit volume, the magnetization is defined as

M=Nmg (9-1)
if the individual dipole moments in the element of volume dz considered are
all aligned in the same direction. If the moments are not all aligned, then the
magnetization is the net magnetic moment per unit volume. The volume dr
is subject to the limitations discussed in Section 3.1.

soitpolieatonPinia NNNSnSnon
9.2 MAGNETIC INDUCTION B
Ad ANCEXTERIORSPOINT

Let us calculate the magnetic induction at a point P outside a magnetized

body, as in Figure 9-1. Our discussion will parallel that of Section 3.2.
From Section 7-8, the vector potential of a current loop is

aes Mo m Pa al
: 4r
(9-2)
where m is the magnetic dipole moment of the loop, and where the distance
ris large compared to the size of the loop. The unit vector r; points, as usual,
from the source, which is the loop, to the point P where the field is calculated,
 385

Figure 9-1. Element of volume inside a

magnetized body, and an external
© P(x,y,2) point P.

Then the element of dipole moment dr’

M situated at P’(x’, y’, z’) pro-
duces at P(x, y, z) a vector potential

dA = (9-3)
and

eh ae ae
_ Mo [Mx
=e le \ciet
(2)ar, :
(9-5)

where 7’ is the volume of the magnetized material, or

Ae= del
aia | Ae / Ho
4 M= dr’ ! + a i)Veh
: ’
dr’, (9-6)

from Problem 1-18. Then, from Problem 1-30,

A=_4real |a MXn,,,
ie da eae VXM
u |Pair ,
dr’, :
(9-7)

where S’ is the surface bounding the volume 7’ of the magnetized body. We

may omit the prime on the y that operates on M since V clearly involves
derivatives with respect to the coordinates x’, y’, z’ of the point where the
magnetization is M.
Although these expressions for A are all equivalent, the last one lends
itself to a simple physical interpretation: both integrands have the form of
the vector potential calculations already familiar to us from Section 7.5,
386

.
Figure 9-2. Ampere

provided the numerators represent a surface current density in the first term
and a volume current density in the second term. We therefore define an
equivalent surface current density

qweMixend 0-8
and an equivalent volume current density

=v xXM. (9-9)
Then the vector potential at P is

The equivalent surface current density », for a uniformly polarized rod

is shown in Figure 9-2.
The physical nature of the equivalent or Ampérian currents can be under-
stood from a model due to Ampére. Consider Figure 9-2 and imagine cur-
rents \, flowing around cells in a piece of material whose length is one meter.
The current in one cell is nullified by the currents in the adjoining cells, except
at the periphery of the material. The net current density on the surface of the
rod is thus \,. Equating the magnetic moment of the current \, around the
periphery to the sum of the magnetic moments of the currents in the cells,
we have
Ao = MS, (9-11)
where S is the cross-sectional area of the rod and M is the magnetization, or
the magnetic moment per unit volume of the material, and

eM (9-12)
The origin of the equivalent surface current density A, can thus be understood
in terms of currents on the atomic scale.
9.3. MAGNETIC INDUCTION B AT AN INTERIOR POINT 387

ince J. is V X M, V-J, is always zero. Therefore charge cannot

accumulate at a point by virtue of J.. Furthermore, the equivalent currents
do not produce heating, since they do not involve electron drift and scattering
processes like those associated with conduction currents.
Since A can be ascribed to the current distributions A, and J., we can
immediately write down the value of B using the Biot-Savart law (Section
TEAS

(9-13)
Thus, if we know the magnetization M, we can find the equivalent
current densities. Then we can use the equivalent currents to calculate B
as if they were in a vacuum.
We have demonstrated the validity of this procedure only for points
situated outside the material, but the procedure is equally valid inside, as
we shall see in the next section.

9.3 MAGNETIC INDUCTION B AT AN

INTERIOR’ POINT) THE DIVERGENCE OF B

We now calculate the macroscopic magnetic induction, or the space and time
average of the magnetic induction on the atomic scale, at an interior point in
magnetized material.
We proceed as we did with dielectrics and divide the material into two
regions, one near the point P and one farther away. The near region is a small
sphere of radius R as in Figure 9-3, and the far region is the remaining volume
of magnetized material.

Figure 9-3. Small imaginary

sphere of radius R and volume
7’” centered at P within a
magnetized body. The remaining
volume is 7”’.
388 MAGNETIC FIELDS III

If the radius R is chosen large enough, the magnetic induction B” at the

center of the sphere due to the magnetized material in the outer volume 7”’ is
adequately given by

ib to fe tae
- mf KO Ge +B Me (9-14)
Led is i nr Ss”

_ We now calculate the macroscopic magnetic induction at P due to the

dipoles within the small sphere of volume 7’. We had a similar task in Section
3.3.2, and the average E turned out to be proportional to the polarization P
and in the opposite direction. In magnetic materials the same type of calcu-
lation gives a different result because the B field in the vicinity of a current
loop is different from the E field near an electric dipole, as shown in Figure
9-4, although at large distances the two fields have the same form. In fact, we

a”
 389

(b)

Figure 9-4. The fields (a) of a magnetic dipole and (b) of an electric dipole.

Let us first compute the average magnetic induction produced over the
volume of a sphere by a small current loop inside it. For simplicity let us
place the loop of magnetic moment m at the center of the sphere, as in
Figure 9-5. By definition, the average field in the sphere is

ne If aa! [ox dar (9-15)

ay Reo Tlie

where 7 is the volume of the sphere. Since we do not know the value of 4 in
the immediate vicinity of an atom, we transform the volume integral into a
surface integral, using the result of Problem 1-30:

B= af (n X A) da, (9-16)
eels!
 390

Figure 9-5. Current loop of

magnetic moment m at the
center of a sphere of radius R
large enough for the dipole
approximation to be valid on the
surface.

where S is the surface bounding 7, and n is the outward drawn normal to S.

At the surface the dipole approximation is valid and

A = (Re (9-17)
4r Pr

We can now evaluate the volume integral of Eq. 9-15 without knowing the
value of A inside, provided we know its value at the surface:

.
Be 4x Js
nee
qe
ag (9-18)
or, expanding the double vector product as in Problem 1-8,

_ Mo m(n-r;) — ri(m-m)
| Ba = wf R da. (9-19)

The first term in the numerator on the right is simply m. The second term is
rym Cos 6 and, by virtue of the symmetry, we take its component along m,
which makes it m cos? 6. Therefore

| Rap an a i m cos? 6(27 sin 6 dé), (9-20)

ps 4a 4r Jo
2
= 3 Hom, (9-21)

and the average magnetic induction inside a sphere of radius R due to the
single loop of magnetic moment m at its center is

(2/3)uom
——y
wom
=

‘ (A/3ycR® ow Deke
 9.3. MAGNETIC INDUCTION B AT AN INTERIOR POINT 391

If the current loop is not situated at the center of the sphere, the average B
is unaffected, as you will be able to show in Problem 9-3.
Since there are many atomic current loops inside the sphere, each one
produces an average field as above, and the total average field due to the near
dipoles is
47R$
Wy
eas 3 NR!
K
= 3 HM (9-22)
if they are all aligned in the same direction.
This result is open to question, because some of the loops will be too
close to the spherical surface for the dipole approximation of Eq. 9-17 to be
valid. However, the error can be made arbitrarily small by increasing the size
of the sphere. Molecules are so small that it is easy to choose a microscopic
sphere that is large enough to make this error negligible. This type of problem
was encountered previously in Section 3.3.
The magnetization M may be taken to be uniform inside the small sphere
S’’, and we may assume that the B at the center is this average B we have just
calculated, or (2/3)u.M.

(9-23)
Let us evaluate the third term. This term gives the magnetic induction at
the center of the sphere due to the currents flowing on the surface S’’. The
equivalent current density \, on S” is M X n and is in the direction shown
in Figure 9-6. Thus, at the center of the sphere,
Ae
Sra
X11 (MXn)Xn
R2 3
(9-24)
' These currents give a magnetic induction that is in the direction opposite to
M, and we must therefore take the component of (M X n) X n in the direc-
tion of —M:

mM Neore Tay uo |” M sin? 6 :

on iQue Oh = i,ser oR sind Rdo (9-25)

= — FM. (9-26)
392

Mxn

Figure 9-6. Equivalent current

density M X n on the imaginary
surface S’’.

hus, at an inside point; we are

left with only the first two integrals of Eq. 9-23:

pow { xn dr +H f Xn da’. (9-27)

pa Ho aie x ry wt Ho Ae x Let / a

Now, if we extended the first integral to the complete volume +’, which
is 7’’ + 7’’’, we would be adding the integral of (J. X ri)/r? over the volume
7’ of the small sphere. This latter integral is zero, by symmetry, because J,
is constant throughout the small volume 7’”’. Finally,

Mo
pam f Xm
[ 2 dr 7, etwoos f
an eX
p da’,Y, (9-28)

whereJ, = V X Mand vy, = M X n.

The equivalent currents therefore yield the correct value for the magnetic
induction B at points inside magnetized material as well as outside. In other
words,
We obtained a corresponding result for dielectrics in Sec-
(non 3),3).22,
Therefore the relation

| v-B=01| (9-29)
of Sections 6.9 and 7.4
This is one of Maxwell’s equations.
It also follows that the vector potential due to magnetized material is
9.4 THE MAGNETIC FIELD INTENSITY H. AMPERE’S CIRCUITAL LAW 393

More generally, if we include the free current densities J; and X,,

a= pf
T Jr
dt vxrBeal M 4
is Js
Sees
i
(OD)

Example THE B FIELD OF A UNIFORMLY

MAGNETIZED CYLINDER

We return to the cylinder that is uniformly magnetized in the direc-

tion of its axis, as in Figure 9-2. This is an idealized case, as we shall
see later on, but it is a good example to use at this point.
We have shown in Section 9.2 that the cylindrical surface carries
an equivalent current density \., which is numerically equal to M.
Inside the magnet, V X M = 0, and the volume current density J,
is zero. Over the end faces, M is parallel to the normal unit vector
n, and X, is zero.
The B field of a cylinder uniformly magnetized in the direction
of its axis of symmetry is therefore identical to that of a solenoid
of the same dimensions that carries a current density N’J = M,
where WN’ is the number of turns/meter. Figure 7-19 shows an exam-
ple of such a field.
Note that B and M are in the same general direction inside the
magnet. Near the center of a long uniformly magnetized cylinder,
B would be equal to uM.
vit

9.4 THE MAGNETIC FIELD INTENSITY H.

AMPERE’S CIRCUITAL LAW
We have found (in Section 7.6) that, for steady currents and nonmagnetic
materials,
VX B= wy. (9-32)

Now we have just seen that magnetized material can always be replaced by
its equivalent currents for calculating B. Consequently, if we have magnetized
material as well as steady currents,

UX B= wd; +JD. (9-33)

where J, is the equivalent volume current ree
 394 MAGNETIC FIELDS III

and we can substitute V * M for J.. Thus

VX B=.p15,+V7 X M), (9-34)

vx (7_ u) = J (9-35)
The vecto: between parentheses, whose curl is equal to the free current density
at the point, is the magnetic field intensity

and is expressed in amperes/meter. Note that H and M are expressed in the

same units.
Thus
B= w+ M) (9-37)
This equation is to be compared with Eq. 3-45, which applies to dielectrics:
py So (9-38)
Rewriting now Eq. 9-35, we have that, for steady currents, either inside
or outside magnetic material,

nex 029
Integrating over a surface S,

i (V X H)-da = i J;-da (9-40)

S S

or, using Stokes’s theorem on the left-hand side,

where C is the curve bounding the surface S, and J; is the current of free
charges linking the curve C.
currents, The term on the left is called the magnetomotance.
This is a more general form of Ampére’s circuital law (Section 7.7) in
that it can be used to calculate H even in the presence of magnetic materials.
‘tisrigorouslyvalidhoweser,onlyforsendyeurens we shall dea with
variable currents in the next chapter.
Sections 9.8.1.1 and 9.8.1.2 will illustrate the use of H and of Ampére’s
circuital law.
 395

9.5 MAGNETIC SUSCEPTIBILITY y,,

AND RELATIVE PERMEABILITY uy,
As for dielectrics, it is convenient to define a magnetic susceptibility Xm and a
relative permeability u, such that

Now, since
MM = xl) (9-42)
B= p(H + M), (9-43)
then

where
© B= wl + Xm)H = pouHl= wH, (9-44)
T= Ltd (9-45)
The quantity yu is called the permeability. Both x,, and yu, are dimensionless
quantities.
Equation 9-43 is general, but Eqs. 9-44 and 9-45 are based on the
assumption that the material is both isotropic and linear, in other words,
that M is proportional to H and in the same direction. This assumption is
unfortunately never completely true in ferromagnetic materials. In a perma-
nent magnet, for example, B and H point in roughly opposite directions
(example, page 402). Even in so-called soft magnetic materials, such as soft
iron, B and H do not always point in the same direction, and when they do,
u, 1s by no means constant. It can vary by orders of magnitude, depending
on the value of H and on the previous history of the material (Section 9.6).
The quantities x,, and yu, are therefore not as meaningful as the corresponding
quantities x, and e, for dielectrics.
Bommonsandstonep 0 Salesian ihn aL.
In paramagnetic materials M is given by the Langevin equation 3-98
with p replaced by m and £,,, by B. As for dielectrics, the magnetic energy
mB is much less than kT at room temperature. For example, m is of the order
of 10-3 ampere-meter?, and, even in a magnetic field of one tesla, mB/kT ~
2.5 X 10-*. Expanding the exponentials as in Section 3.9.1,
Nm
Ms Te
Pe ae (9-46)
ue

M M — wNm
Fi a) Lh,ee) ce 9-4
> OS Tg ey 3kT CD
The magnetic susceptibility of paramagnetic substances is smaller than unity
by several orders of magnitude and is proportional to the inverse of the
absolute temperature.
396 MAGNETIC FIELDS III

In diamagnetic materials the magnetization is in the direction opposite

to the external field; the relative permeability is Jess than unity and is inde-
pendent of the temperature.
If orientational magnetization predoininates, the resultant permeability
is greater than unity.

9.5.1 The Equivalent Current Density vx M

and the Free Current Density J;
In a linear and isotropic magnetic material, B = uH. If, moreover, u is not a
function of the coordinates, we have from Eq. 9-39 that

pV XH=V XpH=VXB= gp; (9-48)

or, using Eq. 9-37,

MoV X (H+ M) = pd = poutrds, (9-49)

VXM=u2J;—V X H, (9-50)
(9-51)
Thus, in linear, isotropic, and homogeneous magnetic materials, the
equivalent volume current density V X M is a multiple of the free current
density J;, the factor of proportionality being positive.
This relation reminds one of Eq. 3-58, which is the corresponding relation
for dielectrics.
Also,
ud; = wo(Js + V X M), (9-52)
which corresponds to Eq. 3-60. The above equation is valid only for steady
fields, as we shall see in Chapter 10.
It follows from Eq. 9-31 that, in a homogeneous medium up,

Dad Jr ay oe Ho vy + M xn
da’. (9-53)
An Jy 4a} sr r
A =

The surface integral cannot be transformed like the volume integral, because
A; and M X n are unrelated; for example, a piece of soft iron placed in a
magnetic field acquires a magnetization M even though its A; is zero.

9.6 HYSTERESIS

One can measure B as a function of H with a ring-shaped sample as in

Figure 9-7. The function of winding a, which has N turns and carries a
 397

Figure 9-7. Rowland ring for

the determination of Basa
function of H in a ferromag-
netic substance.

current J, is to apply an azimuthal magnetic field intensity

NI
ale eee (9-54)

The magnetic induction is

, (9-55)
where S is the cross-sectional area of the core and @ is the magnetic flux. One
can measure changes in ®, and hence changes in B, by changing J and inte-
grating the electromotance induced in winding 6 as in Problem 9-11. Both
B and H are therefore easily measurable.
If we start with an unmagnetized sample ofiron and increase the current
in coil a of Figure 9-7 to some specified magnitude, the magnetic induction
increases along a curve such as ab in Figure 9-8. This curve is known as the
magnetization curve. If the current in the winding is reduced to zero, B de-
creases along bc. The magnitude of the magnetic induction at c is the rema-
nence or the retentivity for the particular sample of material. If the current is
then reversed in direction and increased, B reaches a point d where it is
reduced to zero. The magnitude of H at this point is known as the coercive
force. On further increasing the current in the same direction a point e,
symmetrical to point b, is reached. If the current is now reduced, reversed,
and increased, the point b is again reached. The closed curve bcdefgb is known
as a hysteresis loop. If, at any point, the current is varied in a smaller cycle, a
small hysteresis loop is described.
The differential permeability is defined as the ratio
1 dB
po dH
It is not a constant.
398

& 1 J

100 200

(ae
meter

Figure 9-8. Magnetization curve

ab and hysteresis loop bcdefg. It
is shown in the example on page
402 that B and H inside a per-
manent magnet give a point such
as p in the second quadrant.
lx}

A characteristic feature of the hysteresis cycle is the saturation induction

beyond which M increases no further. Maximum alignment of the domains
is then achieved, and a further increase in H increases B only as the contribu-
tion from the conduction current increases. The saturation induction is
characteristically in the range from 1 to 2 teslas.
In a permanent magnet, the point p representing B and H on the hys-
teresis loop of Figure 9-8 is in the second quadrant.

9.6.1 Energy Dissipated in a Hysteresis Cycle

Energy is required to describe a hysteresis cycle. This can be shown by
considering Figure 9-7. When the current is increasing, the electromotance
induced in the winding opposes the increase in current, according to Lenz’s
law (Section 8.1), and the extra power spent by the source is
dw d® dB
= (va) = INS an (9-56)
 399

(teslas)
B

100 200

(222ce
meter

e =|

Figure 9-9. The shaded area gives the energy required, per unit volume,
in going from g to 4 on the hysteresis loop. The total energy per unit
volume required to describe a complete hysteresis loop is equal to the
area enclosed by the loop.

where S is the cross-sectional area of the ring, N is the number of turns, and
B is the average magnetic induction in the core. Also,
dw INSIdB dB
pee am ar aD oe
where / is the mean circumference of the ring, 7 = S/ is its volume, and
H = NI/I, as in Eq. 9-54. Thus
b

We = | HdB (9-58)
g

is the energy supplied by the source in going from the point g to the point b
in Figure 9-9. This integral corresponds to the crosshatched area in the figure
and is equal to the energy supplied per unit volume of the magnetic core.
400 MAGNETIC FIELDS III

When the current is in the same direction but is decreasing, the polarity
of the induced electromotance is reversed, according to Lenz’s law, with the
result that the energy

W,=+ i HaB b
(9-59)
is returned to the source.
Finally, the energy supplied by the source during one cycle is

We=r f H dB, (9-60)

where the integral is evaluated around the hysteresis loop. The area of the
hysteresis loop in tesla-ampere turns/meter or in weber-ampere turns/meter®
is therefore the number ofjoules dissipated per cubic meter and per cycle in
thecore:
Note that this energy loss is unrelated to the losses associated with the
eddy currents. Eddy-current losses, which are caused by changes in the mag-
netic field, can be minimized by using laminated materials, as in transformers,
by using powders dispersed in an insulator, or by using nonconducting ma-
terials. Hysteresis losses, however, can be minimized only by selecting a
material with a narrow hysteresis loop.

Example For the case illustrated in Figure 9-9, the area enclosed by the hyster-
esis loop is roughly 150 weber-ampere turns/meter®, or 150 joules
per cubic meter and per cycle, or 600 calories per cubic meter and
per cycle. This energy loss could be important if the material were
subjected to an alternating field.

9.7 BOUNDARY CONDITIONS

Let us examine the continuity conditions that B and H must obey at the
interface between two media. We shall proceed as in Section 4.1.
Figure 9-10a shows a short cylindrical volume whose top and bottom
faces are parallel and infinitely close to the interface. Since there is zero flux
through the cylindrical surface, the flux through the top face must equal that
through the bottom, and

Bas Bay o-60)

Consider now Figure 9-10b. The closed path has two sides parallel to
 401

Figure 9-10. (a) Gaussian surface

at the interface between two
media. (b) Closed path crossing
the interface.

the interface and close to it. From the circuital law, Eq. 9-41,

fpnat = 7, (9-62)

where J is the conduction current linking the path.

If the two sides parallel to the interface are infinitely close to it, J is equal
to zero since there are no true surface currents, except in superconductors.
Thus
(iA (9-63)
sn thetangent emponent ofiscontinousaos nineties.
If we can set B equal to wH, the relative permeabilities being those which
correspond to the actual values of H, then

|at £050,=ple 05 0-69

402 MAGNETIC FIELDS III

from the continuity of the normal component of B, and

ee | 9-65
from the continuity of the tangential component of H. Then

9.8 MAGNETIC FIELD CALCULATIONS

It should be obvious that magnetic fields involving ferromagnetic materials

do not lend themselves to rigorous mathematical analyses. Approximate
methods of calculation have therefore been devised, and these are usually
satisfactory. If more accuracy is required, one must resort to making meas-
urements on models.
As a first example of magnetic field calculations we shall return to the
field of a bar magnet and apply the concepts that we have developed so far.
Then we shall discuss a different approach that utilizes the concept of mag-
netic circuit. ff

Example THE B AND H FIELDS OF A BAR MAGNET

We first return to the idealized case of the uniformly magnetized
cylinder of Figure 9-2. This will help us understand the field of a
real bar magnet.
As we have seen previously in the example on page 393, the B
field, both inside and outside the bar magnet, is that of a solenoid
of the same size and carrying a current density N’7 = M, where N’
is the number of turns/meter. This field is shown in Figure 9-11.
Note that the lines of B are refracted at the cylindrical surface as
explained in the example on page 316.
Outside the magnet, H is simply B/o.
Inside the magnet,

fi) = be M, (9-67)
Ko

from Eq. 9-36. (Remember that the relation H = B/y is valid only
 403

Figure 9-11. Lines of H (solid) for a uniformly magnetized cylinder.

Lines of M (not shown) are parallel to the axis and point to the left. Lines
of B are shown broken inside the magnet; outside, they follow the lines
of H.
This figure is identical to Figure 3-10, and the above B field is the
same as that of Figure 7-19.

in soft magnetic materials that are both linear and isotropic.) Near
the center of a long bar magnet, B ~ uoM and H ~ 0. For a short
bar magnet we can use the values of B that we calculated for a short
solenoid in the example on page 317, substituting again the mag-
netization M for the current density /N’. Thus, at the center of the
magnet,
B= woM sin On, = —M(1 — sin @,,), (9-68)
while at the center of one face
404 MAGNETIC FIELDS III

B = 1M = Be, = —m(1 = ne), (9-69)

The H field is also shown in Figure 9-11.
It will be noticed that, inside the magnet, B and H point in
approximately opposite directions. This can be seen in Figure 9-11,
and also in Figure 9-8. Also, taking the curl of Eq. 9-67,
VX Ba WY KV x’) =—O0: (9-70)
V X H being zero because J; is zero (Eq. 9-39), and V X M being
also zero since the magnetization M is uniform, by hypothesis.
Note that the lines of H are nor refracted at the cylindrical sur-
face of the magnet.
The field of a real bar magnet is not as simple to analyze as the
one we have just described. Since the magnetic moments of the indi-
vidual atoms tend to align themselves with the B field, the magneti-
zation M, and hence also the equivalent current density on the cylin-
drical surface, are weaker near the ends. Moreover, since M is not
uniform, there can be Ampérian currents inside the magnet. The end
faces also carry Ampérian currents since M X n is zero only on the
axis. The net result is that there are “poles” near the ends of the
magnet from which lines of B appear to radiate in all directions.
The poles are most conspicuous if the bar magnet is long and thin.

Example THE BAR MAGNET AND THE BAR ELECTRET

It is instructive to compare the field of our uniformly magnetized
cylinder of magnetic material, Figure 9-12a, with that of its electric
equivalent, the uniformly polarized cylinder of dielectric of Figure
9-12b. We discussed the field of the bar electret in the example on
page 113.
In the case of the magnet, the B field is that of the equivalent
solenoid, Figure 9-12c which has a current density M. Then, to find
the H field, we use the relation
B = poH + woM. (9-71)
Thus H is B/o outside and B/u — M inside.
In the case of the electret, the E field is that of the bound charges
on the end faces as in Figure 9-12d, or ofa pair of parallel and oppo-
sitely charged disks carrying charge densities +P. The D field is
then given by
D=eE+P (9-72)
and is therefore €)/ outside and e+ P inside.
Mathematically, the fields of the magnet and of the electret obey
similar equations:
Electret Magnet
D=ek+P B = wH + wM
VsD=0 V-B=0 (9-73)
VXE=0 VxXH=0
 405

5 (a) . (b)

2
en 2

L (a)
Figure 9-12. (a) Bar magnet. (b) Bar electret. (c) At any point in space
the magnetic induction B due to the magnet is the same as that of a
solenoid of the same size carrying a current density N’J equal to M. The
number of turns per meter on the solenoid is N’. (d) Similarly, the
electric field intensity E due to the electret is the same as that of a pair
of disks occupying the positions of the end faces and carrying charge
densities +P. :

Indeed, the H field of the magnet can be calculated like the E field
of the electret if one assumes the existence of a magnetic pole density
equal to +M on the end faces.

9.8.1 Magnetic Circuits

Figure 9-13 shows a ferromagnetic core around which is wound a coil of N
turns carrying a current J. We wish to calculate the magnetic flux ® through
some cross section of the core.
In the absence of ferromagnetic material, the lines of B are as shown in
the figure. At first sight, one expects the B inside the core to be much larger
close to the winding than on the opposite side. This is not the case, h

This can be understood as follows. The magnetic induction due to the

current J magnetizes the core in the region near the coil, and this magneti-
zation produces Ampérian currents that both increase B and extend it along
the core. This further increases and extends the magnetization, and hence B,
until the lines of B extend all around the core.
Of course some of the lines of B escape into the air and then return to
406

Figure 9-13. Ferromagnetic toroid with concentrated winding. The lines

of force shown are in the plane of the toroid and are similar to those of
Figure 9-4a. They apply only when there is no iron present.

the core to pass again through the coil. This constitutes the leakage flux that
may, or may not, be negligible. For example, if the toroid is made up of a
long thin iron wire, the flux at P is negligible compared to that near the coil.
Let us assume that the cross section of the toroid is large enough to
render the leakage flux negligible. Then, applying Ampére’s circuital law,
Eq. 9-41, to a circular path of radius r going all around inside the toroid,

fu-a = NI, (9-74)

2nr == NI, (9-75)

_ BNI
== :
(9-76)
Then, taking R, to be the radius corresponding to the average value of B, and
R, to be the minor radius of the toroid,
= pmR3
® ig NI. (9-77)

The flux through the core is therefore the same as if we had a toroidal coil
(example, page 313) of the same size and if the number of ampere-turns were
increased by a factor of u,. In other words, for each ampere-turn in the coil
there are u, — 1 ampere-turns in the core. The amplification can be as high as
10° or more.
This equation shows that the magnetic flux is given by the magnetomo-
tance N/ multiplied by the factor

pars
2rR,y
9.8 MAGNETIC FIELD CALCULATIONS 407

which is called the permeance of the magnetic circuit. The inverse of the
permeance is called the reluctance. Thus
_ Magnetic Flux = Permeance X Magnetomotance (9-78)

The analogy with Ohm’s law is obvious: if an electromotance were in-

duced in the core, the current would be

a orR>
u io 27k, A;
(9-80)
Thus the corresponding quantities in electric and magnetic circuits are as
follows:
Current J Magnetic flux ®
Current density J Magnetic induction B
Conductivity Permeability u
Electromotance Magnetomotance
Electric field intensity E Magnetic field intensity H
Conductance Permeance
Resistance Reluctance
There is one important difference between electric and magnetic circuits:

Indeed, a magnetic circuit behaves much

as an electric circuit would if it were submerged in salt water: part of the
current would flow through the components, and the rest would flow through
the water.
If a magnetic circuit is not properly designed, the leakage flux can easily
be an order of magnitude Jarger than that flowing around the circuit.

9.8.1.1 Magnetic Circuit With an Air Gap. Figure 9-14 shows a circuit with
an air gap whose cross section is different from that of the soft-iron yoke.
Each winding provides N/J/2 ampere-turn.
We shall see that the magnetic flux in the air gap is equal to the mag-
netomotance NJ divided by the sum of the reluctances of the iron yoke and
of the air gap.
This is a general la

We assume that the leakage flux is negligible. As we shall see, this will
result in quite a large error.
408

Figure 9-14. Electromagnet. The

coils have been cut out to expose
the iron yoke.

Applying again Ampére’s circuital law to the circuit, we see that

NI =H Las AL (9-81)

where the subscript i refers to the iron yoke and g to the air gap. As we shall
see below, the requirement that the reluctance of the yoke be much smaller
than that of the gap makes H,;L; < H,L,. The path length L; in the iron can
be taken to be the length measured along the center of the cross section of the
yoke.
Now, according to Eq. 9-29, or 6-125, the net outward flux of B through
any closed surface must be equal to zero. This results from the fact that V-B
is always zero; indeed, this is one of Maxwell’s equations. Thus, if we neglect
leakage flux, the flux of B must be the same over any cross section of the
magnetic circuit and
B;A; = B,A,; (9-82)
where A; and A, are respectively the cross sections of the iron yoke and of
the air gap.
Combining these two equations,

L; L, =
BAG |+ = —+ |= NI, (9-83)

and the magnetic flux is

 9.8 MAGNETIC FIELD CALCULATIONS 409

b= ByAy = =NI ~ mde,

Ni woA
NI
(9-84)
v g 0

LA; Lo A g
Now we have two reluctances in series, L;/A, in the yoke, and L,/uoA, in
the gap.
You can now show that H,L; << H,L,, as in Eq. 9-81, if the reluctance of
the yoke is much smaller than that of the gap.
Since we have neglected leakage flux, this equation can only serve to
provide an upper limit for &.

Example If there is a total of 10,000 turns in the two windings, and if J= 1.00
ampere, 4; = 100 centimeters”, 4, = 50.0 centimeters?, 4, = 1,000,
L; = 90.0 centimeters, L, = 1.00 centimeter, then
104
p = pena! 0.9we ERE a [Oa a ee (9-85)

105 <4 On O2 10m al Ome

= 6.3 < 107* weber, (9-86)
‘G3 XX O- e :
a 510 1.3 teslas = 1.3 X 104 gauss, (9-87)

CS Omeel02
is 1.00 = 63 henrys ie
(9-88)

and the stored energy is

+LP = $ X 63 X 1.00 = 32 joules. (9-89)

In this particular case the leakage flux is 70% of the flux in the gap.
In other words, the magnetic induction in the gap is not 1.3 teslas,
but only 1.3/1.7 = 0.77 tesla.*

9.8.1.2 Magnetic Circuit Energized by a Permanent Magnet. Figure 9-15

shows a magnetic circuit that is similar to that of Figure 9-14, except that
the two windings and part of the yoke have been replaced by permanent
magnets. Let us see how we can calculate the magnetic flux. We neglect the
leakage flux, as previously.
The B and H vectors are oriented as shown in the figure. In the soft-iron
yoke and in the pole pieces, B and H are in the same direction and the oper-
ating point on the hysteresis curve (Section 9.6) is situated somewhere in the
first quadrant. In the air gap H = B/, and the two vectors are again oriented

* There exist empirical formulas for calculating leakage fluxes. See, for example, R. K.
Tenzer, Estimating Leakage Factors for Magnetic Circuits by a Simple Method, Electrical
Manufacturing, February 1957.
410

| Permanent pea aes ‘ igure 9-15.

Figure 9 Magnetic
gne uit energize
circuit energized
magnets oft iron yoke by a pair of permanent magnets.

in the same direction. As we saw in the example on

This is a good illustration of the fact that the flux of H over a closed
surface is not necessarily equal to zero (Problem 9.19). In fact, the ends of the
permanent magnets act as either sources or sinks of H. For example, the
lower end of the top magnet acts as a source of H whereas the top end acts
as a sink of H.
To calculate the magnetic flux we proceed as follows. Since there are
neither free currents nor displacement currents, the line integral of H-dl
around the magnetic circuit must be zero. Then

A,L; + Js i er ear tsPe = 0, (9-90)

Aylin = gh PF iiglag GS gles, (9-91)

where the subscripts i, g, m refer, respectively, to the soft iron, the air gap,
and the permanent magnets. The magnetomotance H;L; is again negligible
if the reluctance of the iron is much smaller than that of the air gap.
9.8 MAGNETIC FIELD CALCULATIONS 411

Also, if the leakage flux is negligible, & is the same over any cross section,
and
Dies Ze Ee (9-92)
These two equations are similar to Eqs. 9-81 and 9-82, except that NJ is
replaced by HinLm. Thus

= Tia tary? (9-93)

LA; bog

Remember that L; is the length of the path in the soft-iron yoke only,
and that the total distance around the circuit is L; + Lm + Ly.
The design of a magnetic circuit energized by a permanent magnet is
somewhat complicated by the fact that the position of the operating point
on the hysteresis curve (Section 9.6), and hence the value of H,,, depends on
the permeance of the magnetic circuit.
As a rule, one tries to use an operating point that makes the energy
product H,,B, maximum. The reason for this is the following. The purpose
of the magnetic circuit is to produce a magnetic field in an air gap, and the

Energy product - B,H,

Bie One 2 1.5 1 0.5

1.0

0.8

3 x 107 0.6 @
5 3
isa)
2 0.4

iS
1 0.2

0.5
= (he Sy eas ee a
5 x 10° 4 3 2 1 0
H (ampere-turns/meter)

Figure 9-16. Demagnetization curve for Alnico 5. The energy product is maximum
for B = 0.8 tesla.
412 MAGNETIC FIELDS III

magnetic energy required is the volume of the air gap, A,Z,, multiplied by
the energy density $40B?. Now

1 1
Ane A, LB; =e (H,L,)(A,B,), (9-94)
2 Lo 2

= 5(HnlnXAmBn) = 5(Hn Buh Emm)» (9-95)

from Eqs. 9-91 and 9-92.
The magnetic energy in the air gap is therefore equal to one-half the
energy product multiplied by the volume of the permanent magnet. As a rule,
one requires that the volume L,,A,, be as small as possible, for reasons of
economy, size, and weight. Then the operating point on the hysteresis curve
is chosen so that the energy product H,,B,, is maximum.
Figure 9-16 shows the demagnetization curve for the alloy Alnico 5,
which is commonly used for permanent magnets.

Example Let us design a magnetic circuit that will produce a magnetic field
of 1.3 teslas in an air gap 1.00 centimeter long and 50.0 centimeters?
in area as in the previous example. We again neglect the leakage flux
and the percentage error will be the same as in the previous example.
With the alloy Alnico 5, the energy product H,,,B,, is maximum

a B,, = 8000 gauss = 0.80 tesla, (9-96)

Hf, = 450 oersteds = 3.6 X 104 ampere-turns/meter. (9-97)
Then

A, = ae = an < 5.0 X 10-* = 8.1 X 107 meter? (9-98)

and the permanent magnets should be about 10 centimeters in

diameter.
To find the value of L,,, we use Eq. 9-91. If we use the same
geometry as in the previous example, the reluctance of the yoke is
even more negligible because L, is now shorter since it does not in-
clude the lengths of the permanent magnets and we can set

lobey eh = loti: (9-99)

Then

Iees
— Hole _ Bola
= Tis a bol m (9-100)

Le x Ome
= 4r X 1077 X 3.6 xX 10! = (0.29 meter. (9-101)

Each magnet should therefore be about 15 centimeters long. If per-

manent magnets of the proper shape are not available, smaller mag-
nets can be assembled in an appropriate series-parallel combination,
9.8 MAGNETIC FIELD CALCULATIONS 413

We have again neglected the leakage flux and B, is in fact only

0.77 tesla.

9.8.2 Solution of Poisson’s Equation for B

You will recall that there exists a Poisson equation for E (Section 4.7). There
exists a similar equation for B: taking the curl of Eq. 9-34,
VXV X B= —-V’B+ VV -B) = wV X (Jy +V X M) (9-102)
and, since V-B is zero (Hq: 9-29),
VB = ud X I+ XM). (9-103)
This is Poisson’s equation for B.
The solution of this equation will be found in Section E.9 of Appendix E.
It is

Re
Ar
i ESE I
SLD (9-104)
We have used primes on the dels on the right-hand side because they operate
at the source point (x’, y’, z’). Note the analogy with the corresponding
equation 4-184 for E. The above equation relates B, not to the current density
J; + V X M itself, but to its cur]. We again have ¢ to the first power in the
denominator, but a positive sign. Of course the integral can be calculated only
if the curl exists.
You will remember from Section 9.3 that, for a volume distribution of
charge,

B= #™ i Jee Vado A, r
(9-105)
Since the two integrals for B are equal, it follows (Problem 9-22) that

ifvx here =0 (9-106)

for any current distribution, as long as the curls exist.

Example THE MAGNETIC INDUCTION AT THE CENTER

OF A ROTATING DISK OF CHARGE

A thin disk of charge of radius R and thickness 2¢ < R has a density

iS Rpt =z)"
in cylindrical coordinates, p and z being the radial and axial coordi-
nates, respectively. When the disk rotates at an angular velocity w
the current density is
414 MAGNETIC FIELDS III

J; = K(R — p(t — 2)*ap ¢1. (9-107)

We shall calculate B, first with Eq. 9-104, and then with Eq.
9-105.
The curl of J; is
r Od, Il eG;
WADE cee pet 5 ap Pi) #1 (9-108)
and V’ < M is zero in this case. By symmetry, we can neglect the
first term on the right. Then B is axial and
14a
ey (pJ;)
Pes ( td ey 3 (9-109)
Grr a r

Since, by hypothesis, the disk is very thin, we may substitute p for r,

and then
palo [*" [?-* oKQR — 3p)¢ = 2) -2np dp dz, (9-110)
qr sme =0) p

2
Te boKwR?2?. (9-111)

We can arrive at the same result if we use Eq. 9-105. So as to

simplify the calculation, we shall use the formula which we found in
the example on page 299 for a current ring, setting z = 0. The ring
of inner radius p and outer radius p + dp carries a current
+t
I = K(R — p)wp dp ts (t — z)? dz, (9-112)

= 83 K(R — p)wpt? 3 dp, (9-113)

and
R =
8/3)K(R — 8
0 p

= 5: moKoR*t273 (9-115)
as before.

9.9 SUMMARY

The magnetization M corresponds to the polarization P in dielectrics and is

the net magnetic moment per unit volume.
According to the Ampérian model, the magnetization produces an equiy-
alent surface current density X and an equivalent volume current density Je,
where
re M Xn, (9-8)
J. Vx M. (9-9)
9.9 SUMMARY
415

We have shown that B can be calculated correctly, both inside and outside
magnetic material, by treating the equivalent currents as if they were real
conduction currents flowing in a vacuum. For example, the B field of a cyl-
inder uniformly magnetized in the direction of its axis of symmetry is identical
to that of a solenoid carrying a current density IN’ equal to M.
The magnetic field intensity H is related to B and to M as follows:

B
H=——M
=, " -
(9-36)
or
B= u(H + M). (9-37)
This equation is to be compared with
D=aE-+P, (9-38)
which applies to dielectrics.
In terms of H, Ampére’s circuital law is stated as follows:
VxXH=J (9-39)
or

fa = I, (9-41)
C;

where / is the current of free charges linking the curve C. This is a more
general form of Ampére’s law (Section 7-7), in that it is valid even in the
presence of magnetic materials, but it is still rigorously valid only for steady
currents.
It is the custom to write that
M = xnH, (9-42)
where Xm is the magnetic susceptibility. Then
B= pll + xn) = vowH = vw, (9-44)
where uy, is the relative permeability and p is the permeability.
In linear, isotropic, and homogeneous materials,

Je = (ue — IDSs, (9-51)

and
wd; = wd; + V XM). (9-52)
The hysteresis loop of Figure 9-8 is a plot of Bas a function of H for one
complete cycle. It shows clearly that », = B/uoH is definitely not a constant
for a given ferromagnetic material. The area of the hysteresis loop is equal to
the energy dissipated per cubic meter and per cycle in the material.
416 MAGNETIC FIELDS III

At the interface between two media | and 2, the boundary conditions for
B and H are as follows:
Bu a Bro, (9-61)

Jake = Jaley (9-63)

If the media are linear and isotropic, then the B and H vectors are parallel,
bothin medium | and in medium 2, and

tan, bn
(9-66)
tan OP) Mn,

where the angles are those between the B vectors and the normal as in
Figure 9-10a.
The concept of magnetic circuit is widely used whenever the magnetic
flux is guided mostly through magnetic material. We then have a magnetic
equivalent of Ohm’s law:

Magnetic Flux = Permeance X Magnetomotance (9-78)

__ Magnetomotance
~ Reluctance O44)
The correspondence between electric and magnetic circuits is as follows:

Current J Magnetic Flux &

Current density J Magnetic induction B
Conductivity o Permeability u
Electromotance Magnetomotance
Electric field intensity FE Magnetic field intensity H
Conductance Permeance
Resistance Reluctance

The magnetomotance of a coil is NJ, where N is the total number of

turns, while that of a cylindrical permanent magnet is H,,L,,, where H,, is
the value of H inside the magnet and L,, is the length of the magnet.
Poisson’s equation for B is

VB = —nwV X (J; +V X M), (9-103)

and thus
pa ie XU +V XM),
zal : (9-104)

This is an alternate expression for B, which should be compared with Eq.

4-184.
 417

PROBLEMS

1A long tube is uniformly magnetized in the direction parallel to its axis.

What is the value of B inside the tube?
9-2. Show that the torque exerted on a small permanent magnet of dipole moment
m situated in a magnetic field Bis m X B.
9-3. A small magnetic dipole of moment m is arbitrarily oriented at an off-center
point within a spherical surface of radius R.
Show that the average magnetic induction inside the sphere is wpm /27R?
as in Section 9.3.
9-4. Draw a parallel between the bound charges of Chapter 3 and the equivalent
currents.
9-5. /The magnetization of iron can contribute as much as 2 teslas to the magnetic
induction in iron. If each electron can contribute a magnetic moment of
9.27 < 10-7? ampere-meter” (one Bohr magneton), how many electrons per
atom, on the average, contribute to the polarization?
9-6. Show that the equivalent surface current density on the interface between two
different magnetic materials with magnetizations M, and Mz is
(M2 — M;) x n,

where n is a unit vector that is normal to the interface and that points from
medium 2 to medium 1.
9-7. Draw a parallel between the relative permeability 4, and the relative permit-
tivity €,.
(9-8. Imagine a parallel-plate capacitor whose plates are charged and insulated.
~ (a) How are Eand D affected by the introduction of a dielectric between the
plates?
(b) Show a polar molecule of the dielectric oriented in the field.
(c) Is its energy maximum or minimum?
(d) Now imagine a long solenoid carrying a fixed current.
How are B and H affected by the introduction of a cylinder of ferro-
magnetic material inside the solenoid?
(e) Show a small current loop representing the magnetic dipole moment of an
atom, and show the direction of the current on the loop.
(f) Is its energy maximum or minimum?

9-9. A conducting wire carrying a current / is embedded in a nonconducting mag-

netic material.
Find H, B, M and the equivalent currents.

9-10. A long wire of radius a carries a current / and is surrounded by a long hollow
iron cylinder. The inner radius of the cylinder is b and the outer radius c.
418 MAGNETIC FIELDS III

(a) Compute the flux of B inside a section of the cylinder / meters long.
(b) Find the equivalent current density on the inner and outer iron surfaces,
and find the direction of the equivalent currents relative to the current in the
wire.
(c) Find the equivalent current density inside the iron.
(d) Find B at distances r > c from the wire.
How would this value be affected if the iron cylinder were removed?

9-11. It was mentioned in Section 9.6 that the hysteresis loop for a given sample of
material can be obtained by using a ring-shaped sample, as in Figure 9-7.
Show that, if the ratio R/L (resistance/inductance) for the circuit of
winding d is large, and if the current in winding a is not changed too rapidly,
A® = RQ/N,
where A® is the change in flux, Q is the charge that flows through the circuit,
and N is the number of turns in winding b.

9-12. The peaking strip is a device that is used for measuring B. It consists of a fine
wire of magnetic material, oriented in the direction of B, and centered on the
axis of a solenoid as in Figure 9-17. A small pick-up coil of a few thousand
turns is fixed to the center of the wire.
The peaking strip is operated as follows. If the solenoid carries a direct
current that just cancels the field one wishes to measure, plus a small alter-
nating current, then the H on the axis of the solenoid is that due to the alter-
nating current alone, and the peaking strip goes through a hysteresis loop at
every cycle. The electromotance induced in the pick-up coil is N d®/dr, where
N is the number of turns in the coil and ® is the magnetic flux in the strip.
Now the right- and left-hand sides of the hysteresis loop (Figure 9-8) are
nearly vertical in certain materials, such a molybdenum permalloy, and the
electromotance then has two sharp peaks, one positive and the other negative,
which can be observed on an oscilloscope. If the oscilloscope sweep is syn-
chronized with the alternating current in the solenoid, the two peaks are sym-
metrical when the time average of H on the axis of the solenoid is zero, or
when the dec field of the solenoid exactly cancels the measured field.
The peaking strip has a rather limited range of applications, first because
of the large size of the solenoid: for maximum sensitivity the length of the
wire should be much larger than its diameter, because end effects reduce the
value of B inside it, and the length of the solenoid must be several times that
of the wire, again to reduce end effects. In practice, the solenoid is 10 to 20
centimeters long. Also, because, of limitations in the power that can be dissi-
pated in the solenoid, the peaking strip can be used only up to fields of a few
hundred gauss. Finally, the field of the solenoid can alter the B in the pole
pieces and thus change their permeability locally.
Calculate the peak value of the electromotance induced in the pick-up
coil under the following conditions.
Peaking strip diameter, 2.5 < 10~5 meter;
Number of turns in pick-up coil, 1000;
Maximum value of dB/dH, 0.8 tesla-meter/ampere-turn;
 419

ee
Solenoid

Peaking strip
Pick-up coil

Figure 9-17.

Frequency of applied alternating current, 60 hertz;

Amplitude of the alternating H, 0.25 ampere-turn/meter.
So as to simplify the calculation, replace the sine wave for H by straight
lines, one for each half cycle.

. If m is the unit vector normal to an interface, show that n-V x (A, — A,) = 0
and that n x (V x A,/p, — V x A,/p,) = 9.

. Show that both the normal and tangential components of the vector potential
A must be continuous across the interface between two media if the currents
are constant.

eee thin disk of iron of radius a and thickness ¢ is magnetized in the direction
parallel to its axis.
Calculate H and B on the axis, both inside and outside the iron.

9-16. A coil of 300 turns is wound on an iron ring (u, = 500) of 40 centimeters mean
diameter and 10 centimeters? in cross section.
(a) Calculate the magnetic flux in the ring when the current in the coil is one
ampere.
(b) Calculate the flux when there is a gap of 1.0 millimeter in the ring.

9-17. Show that the energy stored in a magnetic circuit is

= 12°R,
where ® is the magnetic flux and R is the reluctance.

9-18. A magnetic circuit links a one-turn coil.

(a) Show that the inductance L of the coil is equal to the inverse of the reluc-
tance R of the magnetic circuit.
(b) Show that, if the coil has N turns,
N?2
ae
420 MAGNETIC FIELDS III

9-19, We have seen in Sections 6.9 and 7.4 that V- B is zero. This equation is valid
even in nonlineai, nonhomogeneous and nonisotropic media.
Show that it does not follow that V-H is always zero.

9-20. Calculate A,, and L,, for a magnetic circuit, similar to that of Figure 9-15,
which will produce a field of one tesla in an air gap 1.00 centimeter? in area
and 2.00 millimeters long. Assume a leakage factor of 50%.

9-21. In the process of designing magnetic devices it is often necessary to predict

in detail the configuration of amagnetic field. Now magnetic fields can seldom
be calculated analytically. One solution is to perform a numerical calculation
on a computer. Another solution, which can be more convenient in some
cases, is to use an electrolytic tank in which the magnetic field is simulated by
an electric field as follows.
Consider a two-dimensional static magnetic field in the xy plane. In
regions where J; = 0 and where there is no permanent magnetism,
V-B=0; and Vas
Note that these equations are analogous to
Vie be—10) and VXE=0,
which apply to electrostatic fields.
(a) Show that
0A OA
= oy B, =.
B Ox
where 4 is the z-component of the vector potential.
(b) Show that a line of constant 4 is a line of B.
(c) Show that B is derivable from a potential u such that

and that lines of constant u are orthogonal to lines of B and thus to lines of
constant A.
(d) Hence show how a magnetic field can be simulated in an electrolytic tank,
using conducting electrodes to simulate the pole pieces. The pole pieces are
assumed to have infinite permeability.
Note that this method is not applicable if there are current-carrying con-
ductors or permanent magnets in the field.
(e) Show that
Vie — 08

The relaxation method described in Problem 2-16 can therefore also be used
for plotting magnetic fields.

9-22. Show that

[vx (2 +yx *) dr’ = 04

 421

Hints

9-3. The dipole m has two components, one that is radial and another that is perpen-
dicular to the diameter passing through m.
You should be able to show that, for a radial dipole mz,,
a cy sin’ @ de
Be Se eacra Sera
where S is the distance between the dipole and the center of the sphere, and @ = 0
at the position of the dipole. The integral on the right is equal to 4/3 for any value
of S>
Similarly, you can show that, for the other component my,
4r e. Melee COSa On—ma 057 -K) COS :
= [ Bar=am. f,Tl + (S/R? = 2(S/R) cos 8) 82 S18 8 48.

The integral on the right is equal to 8/3 and is also independent of S.

9-22. See Section 9.8.2.

CHAPTER 10

MAXWELL’S EQUATIONS

At this stage we have found all four of Maxwell’s equations in Chapter 6, but
we have found only three of them without using relativity. These were Eqs.
3-41, 8-8, and 9-29. Our first objective in this chapter is therefore to deduce
the fourth one, namely the equation for the curl of B, without using the
methods of relativity. Then we shall reexamine all four equations as a group
and deduce from them two general, and rather surprising, properties of elec-
tromagnetic fields.
The reasoning that will lead us to Maxwell’s equation for the curl of B is
unfortunately rather involved (see Figure 10-11), but we shall learn many
things on the way.
We assume, as uSual, that e, u, « are not functions of the time.
The following definitions will be useful

10.1 THE CONSERVATION

OF ELECTRIC CHARGE

According to all experiments performed to date, it appears that electric

charge is never created nor destroyed (Section 5.23). Therefore, if we consider
10.1 THE CONSERVATION OF ELECTRIC CHARGE 423

a surface S’ enclosing a volume 7’, the net free charge escaping through S$’
during a time interval Ar must be equal to the net free charge lost by 7’ during
this same time interval. Thus

o ypdae--{%oP dr. (10-2)

This is the Jaw of conservation of free charge stated in integral form.
We can also state this law in differential form if we replace the surface
integral by a volume integral. Then

[ va =— |Mar, (10-3)
and, since this equation is valid for any 7’,
0
V-J; = a (10-4)

There is also conservation of bound charges, since

cle 0 Opp
Monewtor
ViaP)
~
=
ot
: (10-5)
where 0P/dt is the polarization current density (Section 3.2.2), and —V-P is
the bound charge density p; (Section 3.2.1).
Finally, the charges involved in magnetization currents (Section 9.2) are
conserved because V-V X M = 0. The net outward flux of these charges
through any closed surface is therefore zero.
Thus, if we set J,, to be the current density in matter,
oP

and
Bh PP OP igre ae Dt, :

There is also a fourth type of current that can exist even in a vacuum and
that we shall use in Section 10.6. The corresponding current density is «0H/dt
and
Pe re Oi ee, (10-8)
since the divergence of E is equal to p,/é, from Section 3.6.
Then the total volume current density is

des yt euxM+ as = In te: (10-9)

424 MAXWELL’S EQUATIONS

and
V-J,= 0. (10-10)
The second and fourth terms in J, are often grouped together, and
Oke OP Bye aD
eo + cra 3 (eo + P)= an (10-11)

is called the displacement current density, since D is the electric displacement

(Section 3.6).
Thus
Pape: a 4VxXM. (10-12)
Example CHARGE DENSITY IN A CONDUCTOR

The current density in a linear and isotropic conductor obeys Ohm’s

law
J; = cE, (10-13)
where @ is the electric conductivity. Then, from Eqs. 10-4 and 3-54,
and if o is not a function of the coordinates,

V-J; = — oe = 0V-E= ee (10-14)

and
Pr = ape. (10-15)
Therefore, as long as our assumptions are satisfied, the free
charge density can only decrease! To achieve a nonzero p; in a homo-
geneous, isotropic, and linear conductor, one must inject charges in
some manner that does not satisfy Eq. 10-13. For example, a con-
ductor can be bombarded with a pulse of high-energy electrons that
come to rest inside the material.
“The inverse of the coefficient of + in the exponent is the
relaxation time. Our calculation is not entirely valid because we
have disregarded the fact that the conductivity o is frequency
dependent, and thus a function of the relaxation time. The calcula-
tion is valid for poor conductors, but it leads to values which are
too small by orders of magnitude in the case of good conductors.
Relaxation times in good conductors are nevertheless short and
py = 0 in practice. For copper at room temperature, for example,
the correct value is ~4 X 107'4 second.” *

10.2 THE POTENTIALS V AND A

We have seen in Sections 3.3.2 and 9.3 that, for a finite charge and current
distribution as in Figure 10-1, the scalar potential V and the vector potential

* W. M. Saslow and G. Wilkinson, American Journal of Physics, 39, 1244 (1971).

 425

PiCx, y, Zi Figure 10-1. The charges and

currents within the volume 7’
produce a scalar potential V and
a vector potential A at the field
point P(x, y, z). The current
density J, at the source point
P'(x', y’, z) is J; + V X M and
the charge density p, is py + po.

A are given respectively by the following integrals:

Va : 1 / Ps avs P iy! dt l |i op
+ 4 P-n yp. (10-16)
Tey | I Atrey | r

Ami | AM
V
ge, SEM aa’, (10-17)
where

py is the volume density of free charge,

P is the electric polarization (Section 3.1),
—V-P is the volume density of bound charge (Section 3.2.1), ps,
a; iS the surface density of free charge,
P-n is the surface density of bound charge (Section 3.2.1), a3,
J; is the volume density of current due to the motion of free charges,
M is the magnetization (Section 9.1),
V X Mis the volume density of equivalent currents (Section 9.2),
A, is the surface density of current due to free charges,
M X nis the surface density of equivalent currents (Section 9.2),
r is the distance between the points P and P’,
P(x, y, Z) is the field point where V and A are calculated,
P’(x’, y’, z’) is the source point where the charge and current densities are
Pys Pbs Of, CLC:

The integrals are evaluated over volumes 7’ and surfaces S$’, which
enclose all charges and currents. If the charge or current distributions extend
to infinity, these integrals do not converge, as a rule.
426 MAXWELL’S EQUATIONS

We have also seen in Sections 3.7 and 9.5.1 that, in a linear, isotropic, and
homogeneous medium e, p,*

rob fmare
dt [esha | Pe
co
yh -Lfds re Ho
Bf ye M X La r (10-19)
i a

If you have studied Chapter 6, you will recall that we found in Sections
6.4 and 6.5 that, for an isolated charge passing through the origin at a constant
velocity © along the x-axis,

yQ :
12" (10-20)
~ Are fyX(x — U1)? + yp? + zy

—+ yp? + 27} 1? (10-21)

aeAn {yx — Ut)?

The potentials V and A are related to the field vectors E and B by the
equations
hea Vig ““ (10-22)
=v ocd. (10-23)
These are Eqs. 6-90 and 6-77, or 8-13 and 7-41 respectively. In the first equa-
tion, —VV is that part of E that is due to accumulations of electric charge,
whereas —0A/0dt is the part that is due to changing magnetic fields. Once V
and A are known, it is merely a matter of differentiation to find E and B.
Therefore the six components of E and B can be deduced from only four
quantities, V and the three components of A.

* Our reasoning with respect to A will not be quite consistent. In Section 9.5.1 we used
the relation

Jy = wlJy+ VX M)
to transform Eq. 10-17 into 10-19. Now this relation is valid only for steady fiekds. The more
general expression for uJ; contains a time derivative on the right, and it can be deduced
from Eqs. 10-94 and 10-99. It is as follows:

dy = why +0 XM) — = no) 2.

The integrand uwJ;/r in Eq. 10-19 is valid for steady fields; is it also valid for variable
fields? It is, but we have not shown that the above time derivative is involved.
Note that 0D/dt is the displacement current. This quantity appears in one of Maxwell’s
equations (See Sections 6.11 and 10.6).
10.2 THE POTENTIALS V AND A 427

10.2.1 The Retarded Potentials

The integrals of Eqs. 10-18 and 10-19 have one serious fault: they do not take
into account the finite velocity of propagation of electric and magnetic fields.
For example, if the charge density changes in one region, the integrals imply
that V changes simultaneously throughout all space. This is of course in-
admissible (Section 5.10). The potentials V and A at the time ¢ cannot corre-
spond exactly to the charge and current distributions at that particular
moment, unless the charges are fixed in space. The analogy with astronomy
is obvious: one cannot see a Star as it is now, but only as it was thousands or
millions of years ago.
If we have charge and current distributions p; and J; in a vacuum, then

Sea | eile
vege. ace (10-24)

— bo [ LWrle
ankle fo de,y :
(10-25)
where the brackets indicate that the charge and current distributions are
those at the previous time t — (r/c). These are the retarded potentials.
From now on, square brackets will be used exclusively to identify retarded
values of charge density, current density, position, velocity, etc.
The expressions for V and A in the presence of matter are simple in one
particular case, namely that of an infinite linear, homogeneous, and isotropic
medium e, u. Then the surface integrals are zero, and

y=_ 1df
{| bbe
ee,gy, :
(10-26)

A= a EEL dr’, (10-27)

where p, and J; are now calculated at the retarded time ¢ — (r/v), v being the
velocity of electromagnetic waves in the medium.
The integrals are evaluated over any volume 7’ that encloses all the
charges and currents.

The velocity » is there-

fore strictly definable only if py and J; are pure sinusoidal functions of the
time, although, in practice, dispersion can often be neglected.
Fortunately, it is not always necessary to take retardation into account,
Retardation can be either important or neglibible, depending on the magni-
428 MAXWELL’S EQUATIONS

tude of the time delay, on the nature of the phenomena involved, and on the
time resolution required. The following two examples illustrate the impor-
tance of retardation for extended pulsating sources.

Example THE RETARDED POTENTIALS FOR

AN OSCILLATING ELECTRIC DIPOLE
One can imagine an electric dipole whose moment is a sinusoidal
function of the time, as in Figure 10-2, with

QO = Qe. (10-28)
Then
p= OQve™'s = poet, (10-29)

where
Po = Qos. (10-30)
For example, the charges could be situated on a pair of spheres
joined by a thin wire of negligible resistance and capacitance.
The upward current J flowing through the connecting wire is

[= “a = jwQoe! = le’, (10-31)

+Q

(b)
Figure 10-2. (a) An oscillating electric dipole. The total charge is zero and
the vector s is oriented from —Q to +Q as shown. (b) The charges and
the current are functions of the time.
 429
P(r,0,@)

-9@*> Figure 10-3. Oscillating electric di-

ia pole and point P where the scalar
potential is V.

and
Ih = jwQo, (10-32)
hs = jwpo. (10-33)
Is = jwp. (10-34)

If the wave propagates in a vacuum, the retarded scalar poten-

tial is

Qo exp jw (:— 7) Qo exp jw (:= ‘s)

Y= (10-35)
Atreéors At e(ra

where the numerators are the charges, as they appear at the point
P(r, 6, ¢) of Figure 10-3 at the time f.
Note that the charges — Q and + Q give scalar potentials which
differ not only in amplitude but also in phase. The amplitudes differ
because of the terms r, and r, in the denominators, and the phases
differ because of the r,/c and r;,/c terms in the exponentials.
Similarly,

A el ie
ell I, exp jw
; (:engl)‘)S.Ves :
(10-36)

We can find a more convenient expression for V that is valid

for r > s as follows. Since
430 MAXWELL’S EQUATIONS

AG +55 cos 0, (10-37)

S
ry Rr — 5 COS 0, (10-38)

then
rp F yescos¢
wo(t-2)wo(s-24+ ae (10-39)

fe w(: = \+s
+ —5 00S 6, (10-40)

and
vale Qo exp jw (:— *) exp G5
3K cos a) exp (1%
x cos 0)

ae 1— 2r
= cosé te > cos 6
(10-41)
where A = )/27m is called the radian length. Expanding the expo-
nential functions and the denominators of the two terms within the
braces as power series, and neglecting terms of the third order and
higher in s/r and s/X, we obtain

Po &Xp Jw (- *) x

é = i)cos @, (10-42)
AtreorA

where we have substituted py for Qos. We have assumed the length

of the dipole s to be small compared to both r and \: s <r and
S < X. This is our only approximation. We have made no assumption
as to the relative orders of magnitude of r and A.
It is important to check this result with that for the static dipole
(Eq. 2-74). For zero frequency, w = 0, A + ~, and the two expres-
sions agree as expected.

Example THE RETARDED POTENTIAL A FOR

AN OSCILLATING MAGNETIC DIPOLE

Figure 10-4 shows a magnetic dipole that is similar to that of Figure

7-20, but which is fed by a source of alternating current. It carries
a current Jp exp jwr.
Instead of Eq. 7-100, we now have that
Qn
rr Ih expjw( — >
Ale= ads acosydpg. (10-43)

This is the sum of the dA’s due to us elements J dl around the loop,
all retarded by the ae time ss et be
— Rewriting,

Moaly exp AG
jw _ df "exp j
A= —— COS ae oie (10- 44)
 431

Figure 10-4. Oscillating magnetic dipole.

If the loop is small with respect to X = \/2z, then

lr—r'| «KX, (10-45)
fr—r r—r
exp /( x )w1+j x (10-46)

and

poaly expjw (1 3) 2
c aed. ayes il
0 are Oia us

A —, sll =, = de :
d 4rr i lie Td X ¢ 1)) arg es
(10-47)
Also, from Eq. 7-106,

ie a ax
\ ~l- Ppt =| a COS Y, (10-48)

for a<~r. If we now integrate and then substitute r sin 6 for x, we

find that

Tyra? exp jw (:= ’) x

Jue aA BE (; +/) sin@ yg, (10-49)
4a rx r
432 MAXWELL’S EQUATIONS

or, setting 7a?) = mo,

Mo EXP Jwa = “) x
A= iaSEES (3as i)sin 6 g, (10-50)

ped AE “
SDHn (+i) exp jo (1= *), (10-51)
~ Ar
where
m = 1S (10-52)
is the magnetic moment of the loop. The magnitude of the vector S
is the area za? of the loop, and its direction is related to the direction
of current flow by the right-hand screw rule.
For zero frequency, w = 0, X >, and

A= He menZS (10-53)
as in Section 7.8.

L033 THES LORENTZ. CONDITION

In view of the fact that p; and J;, which appear in the integrals for V and A in
Eqs. 10-26 and 10-27, are related together by Equation 10-4 for the conserva-
tion of free charge, we may suspect the existence of an equation linking
together the potentials V and A.
We have already found such a relation for the potentials of a single
charge moving at a constant velocity, in Section 6.7.
We shall now show that a similar relation exists between the potentials
of Eqs. 10-24 and 10-25. We shall use the wnretarded potentials V and A in
order to simplify the calculation. This means that we shall assume that the
time delay r/v is negligible.
We calculate the divergence of A:
ep
Usd 4 checm oh i eee
: dr’. :
(10-54)

Since the divergence of A is calculated at the field point P(x, y, z), while the
integral is evaluated over the source points P’(x’, py’, z’) as in Figure 10-1, the
order of the del and of the integral sign can be inverted and

V-A=htFi .
a : Ji
: dr’., :
(10-55)

Now
]
WW u = avs eV . (10-56)
10.3. THE LORENTZ CONDITION 433

The first term on the right is zero since J; is a function of ey ee and not of
the coordinates x, y, z of the field point that are involved in the operator V.
Also, from Problem 1-12, V(1/r) is equal to —V’(1/r), and thus

Ve Ft
of = JVpg! (10-57)
Similarly,
ies
Wi Ot Ie.1 (10-58)
Then, combining these last two equations, we find that
1
ae ut= hoy ce (10-59)
and
: — 1 Ye Je i? ML / J; !
V A = a [ ; dr A ae “ ae ° (10-60)

The second integral is similar to the one we started from (Eq. 10-55),
except that now the divergence of J;/r is computed at P’ instead of at P. The
second integral is zero for the following reason. According to the divergence
theorem, it is equal to the integral of (J;/r)-da over the surface S’ bounding
the volume 7’. Now because, by definition, all currents are contained within
the surface S’, everywhere on S’ the vector J; is either zero or tangential, and
(J;/r):da is zero. Then
lid Ve ‘Ji
V-A dr’.! (10-61)
An jer 7

Now we have had to use a prime on the del operator that appears under
the integral sign because it was essential for us to discriminate between
derivatives with respect to x, y, z and derivatives with respect to x’, y’, z’. But
the V’-J; term is really the same as the V- J; which appears in Eq. 10-4. In this
latter equation we were only concerned with the source point, and a prime on
the del would have been superfluous. Then, from Eq. 10-4, V’-J; is —0p;/dt
and
aera SideAr /J, CLICr i (10-62)
Now + is not a function of the time, since it is the distance between the two
fixed points P and P’. We can therefore remove the 0/d¢ operator from under
the integral sign and
V-A
. oy € ome
fae
SS pr=a f (10-63)
-63

OV
———— aay —— 9, (10-64 )
from Eq. 10-26.
434 MAXWELL’S EQUATIONS

The relation

is cae a is an identity if V and A are defined as in

Eqs. 10-26 and 10-27, despite the fact that the above proof is valid only for
slowly varying fields.* It is based on the well-founded assumption that there
is conservation of charge.
For charges and currents in a vacuum,

V-A + coo = 0 (10-66)

and, for steady fields,

V-A = 0. (10-67)
We have already noted in Section 10.2 that the six components of E and
B can be computed from V and the three components of 4A; with the Lorentz
condition it suffices to know the three components of A to find V, if Visa
function of the time, and therefore to calculate the six components of the
field vectors E and B.

Example THE LEAKY SPHERICAL CAPACITOR

A spherical capacitor as in Figure 10-5 contains a slightly conducting
dielectric. We wish to calculate B and A when the capacitor dis-
charges through its own dielectric.
a) By symmetry, A» and A, are both zero. Then
B=VxXA=0. (10-68)
There are other ways of showing that B = 0. First, we recall
from Section 7.4 that

[Baa si 1) (10-69)
over any closed surface S. If we choose S to be a spherical surface
concentric with the capacitor, then B-da is B,da. By symmetry, B,
must be independent of 6 and ¢, so that 4m7r?B, is zero. Therefore B,
is also zero. But, again by symmetry, By and B, must also be zero,
and thus B = 0.
Also, from Section 8.1.1,
OB
VXE= ark (10-70)

In the present case E is radial and independent of 6 and y, so that

its curl is zero, from Eq. 1-137. Then B must be constant. But B is
zero for tf> ©. Then B is always zero. See also Problem 10-17.
 435

Figure 10-5. Spherical capacitor.

b) To calculate 4, we use the Lorentz condition. We first obtain an

approximate value for V from the following equation:
OV J; I
or BE co. 4nra
(10-71)
where J; is the conduction current density and g is the conductivity.
We have neglected the term —0A4/0dt of Equation 10-22; as we shall
see later, this is well justified. Then
1 & he-#®C, (10-72)
where R and C are, respectively, the resistance and capacitance of
the spherical capacitor.
Let us assume that the outer conductor is grounded and that
the inner conductor is at the potential V;. Then
lity2Tyet/ RC Them a t/ RE G 1
View SS ate a fae ee pet \
; [ 4rr’o a 4iro R, Ro (10-73)

and, at a radius r,
Inet Re G 1 )
mE — (- — — }: 10-74
“é 4iro ee Wo) ( )
According to the Lorentz condition,
OV The“Re (1 1 )
V rey =_ — J ———— -———_— (O)9/
: “Hor *P 4raRC \r Ro MeUr ie)
or, setting the coefficient before the parenthesis on the right-hand
side equal to K,
Kote ie (; 1
2 Dp (r°A,) ~ K -= ——)>
R, (10-76 )

OMe r2
—ar (r°A,) — K(r ——)>
R. 0-77
(10-77)

reA, & K(5 === | (10-78)

436 MAXWELL’S EQUATIONS

We have disregarded the constant of integration because it is

independent of all four variables r, 6, ¢, t for the following reasons.
It is of course independent of r, and it is independent of @ and ¢ by
symmetry. It is also independent of t because A decreases exponen-
tially with time, like V and J, and the exponential function is already
contained in the coefficient K. Thus

a (5-3):
eh eae I TEN GE VE Ati)
Finally we must justify our assumption that
OA, chee ay eee
Ot RC? 40 \2 = are
is negligible compared to dV/dr in Equation 10-71. First, we note
that the above parenthesis is of the order of unity. Now, since
(eu)—'/2 & velocity of light in the dielectric (Section 11.4), RC/(eyu)!/?
is the distance traveled by light in one time constant RC. This,
squared, is vastly larger than the r? of Equation 10-71, under our
assumption that the dielectric is only slightly conducting.

10.4 THE DIVERGENCE OF E

AND THE NONHOMOGENEOUS
WAVE EQUATION: FOR V

It was in Section 3.6 that we found the first of our four Maxwell’s equations,
namely

(10-81)

where p, is the total charge density p; + p,. Although this result was arrived
at in discussing electrostatic fields, it is in fact general, as one can see from the
discussion in Section 6.8.
For a linear, homogeneous, and isotropic dielectric, we have Eq. 3-54,

VE="%, (10-82)
and
0A py
Vilarvpe
( V Sle
=) : g
(10-83)

Rewriting,

é 5 = 4 ts |
Se
Is (10-84)
10.6 THE CURL OF B 437

and, using the Lorentz condition,

AI cat ee (10-85)
This is the nonhomogeneous wave equation for V. At points where
py = 0 we have the usual wave equation (Appendix E); the phase velocity is
1/(eu)"?.
If py is not a function of the time, neither is V and, in a homogeneous
linear and isotropic dielectric,

Vy ee Pt (10-86)
€ €
which is Eq. 3-55.
The solution of the nonhomogeneous wave equation 10-85 for an in-
finite medium e, u (not functions of x, y, z, f) is the retarded scalar potential
of Eq. 10-24 (Appendix E).

10.5 THE NONHOMOGENEOUS

WAVE EQUATION FOR A

We can find the wave equation for A by analogy with that for V.
The vector equation 10-25 for A is equivalent to three scalar equations
for the three Cartesian components of A, each one of these equations being
similar to the single scalar equation for V. Then the three components of A
must satisfy three scalar differential equations similar to Eq. 10-85, and,
combining them together,
0°A
V2A — eu OP = —pJy. (10-87)

If J; is constant, A is also constant and

V2A = —p Jy. (10-88)
We have therefore found three functions of 4: V X A, which is the
magnetic induction B; V-A, which is related to 0V/dt by the Lorentz con-
dition; and V2A as above.

(Cie DHbeCUR OF Bb

We have already found the curl of B in Section 9.4, but that equation is
strictly valid only for steady currents. We have also found the curl of B in
438 MAXWELL’S EQUATIONS

Section 6.11, but we were not then familiar with the equivalent volume
current density V X M.
It will not be difficult to find the correct value of V X B, now that we
have the Lorentz condition and the wave equation for A.
In terms of the vector potential A,
VX B=VXVXA= —-V*%A+ V(V-A). (10-89)
Substituting now the value of V*A from the wave equation 10-87 and the
value of V-A from the Lorentz condition 10-65,

VX B= why a SS+¥ (—a *) (10-90)

= wy — a SS - a ZV (10-91)
= wh — wT a of—E- Sh (10-92)
= wdy + eu a (10-93)
If we remove the e and the uv which restrict the equation to linear, homogene-
ous, and isotropic media, we obtain the following equation:

aD
Ve ere (10-94)

The term dD/dt is called the displacement current density (Section 10.1).
This is another of Maxwell’s equations. Let us rewrite it in terms of E and
B, and without the e and the p:

VX B= wv X(H+M)=n.(V XH+VXM), (10-95)

Zoighh (Ja _ 4VX M), (10-96)
dE oP
= Ho (4+ € OL — Or are. Vi M), (10-97)

Or

[vx B=wal 00-98)

where the total current density J; is defined in Eq. 10-9.
One convenient form for this equation is the following:

1 OE
7) x B = af
me oe is Lod ms (10-99)
10.7 MAXWELL’S EQUATIONS 439

where c = 1/(€ojio)!/? is the velocity of light in a vacuum, as we shall see in the

next chapter, and where J,, is the current density in matter, as defined in
Eq. 10-6.
This result is in agreement with the value of V X B which we found in
Section 9.4 when we were dealing with steady currents. We can see now that
VX Be uly +V XM) (10-100)
if the displacement current (0/dt) (ef + P) is negligible compared to
J; + VX M. When this condition is satisfied the current is said to be quasi-
Stationary.

Example THE DISPLACEMENT CURRENT DENSITY

IN A CONDUCTOR

It is instructive to estimate the ratio of displacement current density

0D/odt to conduction current density J; in a conductor. For an alter-
nating electric field Eo cos wt within a material of conductivity o,
0D
ee —weky sin wt, (10-101)

Jy = gk) cos ot, (10-102)

ab
or We
ai mara (10-103)

The relative permittivity «, of a conductor is not readily meas-

ured, since polarization effects are usually overshadowed by conduc-
tion. However, for purposes of estimation, we may set ¢, ~ 1 and
ao = 10’ mhos/meter for a good conductor. Then
oD
ar
a ~ 10-7, (10-104)
wheref = w/2m is the frequency. The displacement current in a
good conductor is therefore negligible compared to the conduction
current at any frequency lower than optical frequencies, where
DMO” ava.
It is interesting to note that the conduction current is in phase
with the electric field intensity, while the displacement current leads
the electric field by 7/2 radians.

10.7 MAXWELL’S EQUATIONS

Let us group together the four Maxwell equations, which we found suc-
cessively as Eqs. 10-81, 7-34, 8-8, and 10-99:
440 MAXWELL’S EQUATIONS

Wie Fee, (10-105)

€0
V-B=0, (10-106)
VE - = 0, (10-107)
1 0E
Y OB ae le: (10-108)

These equations are also given in Table 6.3.

As usual,
E is the electric field intensity in volts/meter;
pt = py + po is the total electric charge density in coulombs/meter?;
py; iS the free charge density;
p» is the bound charge density —V-P, P being the electric polarization in
coulombs/meter?;
B is the magnetic induction in teslas;
Jm = Jy + OP/dt + V X M is the current density due to the flow of
charges in matter, in amperes/meter?;
J; is the current density of free charges;
OP /dt is the polarization current density;
V X Mis the equivalent current density in magnetized matter;
M is the magnetization in amperes/meter;
c is the velocity of light, 3 X 10° meters/second, and c? = 1/(€ouo);
€) is the permittivity of free space, 8.85 X 10~!? farad/meter;
Mo is the permeability of free space, 47 X 10-7 henry/meter.
In conductors obeying Ohm’s law,

J; = cE, (10-109)

where oa is the conductivity in mho/meter. Inside a source such as a battery,

there is also another electric field intensity E,, due to the local generation of
energy, and
J; = o(E + E,). (10-110)

Maxwell’s equations are partial differential equations involving space

and time derivatives of the field vectors E and B, the total charge density p,,
and the current density J,,. They do not yield the values of E and B directly,
but only after integration and after taking into account the proper boundary
conditions.
These are the four fundamental equations of electromagnetism. They
10.7 MAXWELL’S EQUATIONS 441

apply to all electromagnetic phenomena in media which are at rest with

respect to the coordinate system used. They are valid for nonhomogeneous,
nonlinear, and even nonisotropic media. The remaining chapters will be based
on them.
Maxwell’s equations are not all independent of each other. If we take the
divergence of Eq. 10-107, recalling that the divergence of the curl of any
vector is zero, then
OB
os 0, (10-111)

or, inverting the order of the operations,

te)
ai YB) = (0, (10-112)

The quantity V-B is therefore independent of the time at any point in space.
We can set the divergence of B equal to zero everywhere if we assume that, for
each point of space, it is equal to zero at some time, either in the past or in the
future. Under this assumption, Eq. 10-106 can be deduced from Eq. 10-107.
Equations 10-106 and 10-107 are sometimes called the first pair of Maxwell’s
equations.
Similarly, taking the divergence of Eq. 10-108,

oe = ae’a
v. OL Pode (10-113)
If we assume that there is conservation of charge, as we did in Section 10-1,

per
OK 1 Op,
hs, -114
Or €9 Ot al )

ay
0
apes~ (6) Pt
10-115
Or hy ) Ot €0 ( )

SE eng os (10-116)

where C is some quantity that can be a function of the coordinates, but that
is independent of the time. If we further assume that, at every point of space
at some time either in the past or in the future, both V- EFand p; are simultane-
ously equal to zero, then the constant of integration C must be zero, and we
are left with Eq. 10-105. Under these two assumptions, Eqs. 10-105 and 10-108
are therefore not independent. They form the second pair of Maxwell's
equations.
If we write out p, and J,, in full, Maxwell’s equations become
442 MAXWELL’S EQUATIONS

V-E= aay yp, (10-117)

V-B=0, (10-118)
VxXE+
53 =0, (10-119)
Vx Ba = (E+ +0 XM), (10-120)
CanOn Ot

and, if we use all four field vectors E, D, B, H, remembering that D is ek + P,

while B is uo(H + M),
Viel) ates (10-121)
WATT a) (10-122)
von . = 0, (10-123)

VX H- ee 27h, (10-124)
All three sets of equations (10-105 to 10-108, 10-117 to 10-120, and 10-121
to 10-124) are valid in nonhomogeneous, nonlinear, and nonisotropic media.
In the remaining chapters we shall be concerned mostly with fields where
E and H are sinusoidal functions of the time. In such cases we can replace the
time derivatives by factors of jw (Appendix D). Setting also D = eE and
B= wl,
a ete (10-125)
Via 20), (10-126)
TSC l= 0. (10-127)
UxCH jack =], (10-128)

10.7.1 Maxwell’s Equations in Integral Form

We have stated Maxwell’s equations in differential form; let us now state

them in integral form, as we did in Chapter 6, in order that we may arrive at a
better understanding of their physical meaning. We shall use Eqs. 10-105 to
10-108.
In Section 3.6 we deduced Eq. 10-105 from

|S Reda 4€0 |Jr bones€0 (10-129)

where S is the surface bounding the volume 7 and where Q, is the net charge
contained within 7. The outward flux of E through any closed surface S is
 443

Figure 10-6. Lines of E emerging from

a volume 7 containing a net charge Q,.
The outward flux of E is equal to O;/¢.

therefore equal to 1/e times the net charge inside, as illustrated in Figure 10-6.
This is Gauss’s law (Sections 3.6 and 6.8).
Integrating Eq. 10-106 in a similar manner, we find that the outward flux
of B through any closed surface S is equal to zero (Sections 6.9 and 9.3):

i B-da = 0. (10-130)
S

This is shown in Figure 10-7.

Equation 10-107 can be integrated over a surface S bounded by a curve

/ VX E-da = — | —-:da, (10-131)

S

or, if we use Stokes’s theorem on the left and invert the operations on the
right, the surface S being fixed in space,

Figure 10-7. Lines of B through

a closed surface S. The net out-
ward flux of B is equal to zero.
 Figure 10-8. The direction of the
electromotance induced around
C is indicated by an arrow for
the case where the magnetic
induction B is in the direction
\ shown and increases. The electro-
motance is in the same direction
if B is upward and decreases.

fpeed
. =Clc 2a |Beda
. =
dd
= (10-132)
-132

Then the electromotance induced around curve C is equal to minus the rate
of change of the magnetic flux @ linking C, as in Figure 10-8. The positive
directions for B and around C are related according to the right-hand screw
rule (Section 8.1).
Finally, if we also integrate Eq. 10-108 over an area S bounded by a
Cunver@.
dE
f eat = wf (J.+ € ) \ da = jel. (10-133)
G s ot
If we perform the same operation on Eq. 10-124,

fpn-at = | (a4) -aa, (10-134)

and the magnetomotance around C is equal to the sum of the free current plus
the displacement current linking C. This is illustrated in Figure 10-9. The
positive directions are again related by the right-hand screw rule.

10.8 DUALITY

If py and J; are both zero, Maxwell’s equations 10-121 to 10-124 reduce to

VD =, (10-135)
vin (10-136)
VXE+ 53=0, (10-137)
)
ae ana
ot
(10-138)
 445

Figure 10-9. The direction of the

magnetomotance around C is
indicated by an arrow for the
case where the current J; +
(0D/0t) is in the direction shown.
The displacement current is
downward if D is downward and
increases, or if it is upward and
decreases.

Let us consider one field where the electric field intensity is E and the
magnetic induction is B. Then the above equations are necessarily satisfied.
Let us now consider a different field E’, B’ such that
E’
—KB = —KpH, (10-139)
H’ I +KD = +KeE. (10-140)
Substituting E’ and B’ into Eqs . 10-121 to 10-124, and assuming that e and u
are independent of x, y, z, t, we find that Maxwell’s equations also apply to
the primed field:
V-D' =0, (10-141)
V-B’ =0, (10-142)
/

Vie Bae ~ = 0, (10-143)

Vie Hi - = 0, (10-144)

Figure 10-10. Pair of symmetrical fields. Lines of E are solid and lines of
B are hatched.
446 MAXWELL’S EQUATIONS

This duality property of electromagnetic fields is illustrated in Figure 10-10.

Example If we have a purely electric field {E(x, y, z), B = 0}, then it is possi-
ble to produce a purely magnetic field {B’(x, y, z), E’ = 0}, where
B’ is the same function of x, y, z as E, within a constant factor. The
inverse is also true. Experimentally, of course, one of the two fields
can be more difficult to obtain than the other.

Example The simplest electric field is

E = Constant X i. (10-145)

It is the field inside a parallel-plate capacitor when edge effects are

negligible.
The corresponding magnetic field
B’ = Constant X i (10-146)
is found inside a long solenoid.

Example We have found in Section 2.9 that the E field of an electric dipole is
given by

je. __l
= awe 2p
i cos 6, (10-147)

Ey =Ju PRLS
lee sin¢ 6, (10-148)

E, = 0, (10-149)
as long as r is much larger than the length of the dipole. Then, in
Section 7.8, we showed that the B field of a magnetic dipole is given
by
B== dq
Be 2M
pe 08 6. 0, (10-150)

By Bs
= tq Ra
ps SD 6, %
(10-151)

By = 0, (10-152)
if r is again much larger than the dipole.
The analogy is obvious: these two fields are mathematically
identical, within a constant factor.

10.9 LORENTZ’S LEMMA

We again consider two distinct fields, E,, B,, and E,, By, which are not
necessarily related as in Section 10.8. According to the principle of super-
10.9 LORENTZ’S LEMMA 447

position, these two fields can either exist separately or be superimposed

without disturbing each other, giving a third field EF, + Ey, Ba + Bp.
For these two fields we have the vector identity

V-(E, x B, = E, x B.) = By: V x E, a E,:V x B,

—BuV XE,+ Ey-V X Ba, (10-153)

and, using the Maxwell equations 10-107 and 10-108,

V-(E. X B, — E, X B.) = —B,- oe — E,- (Jus+ &% ay

+By + OB9+ Esto (Ine+ oo R)-

Eu
(10-154)
If the two fields are harmonic functions of the time and if they are of the
same frequency, the 0/dt operators can be replaced by jw and the time de-
rivatives cancel. Then

WC HB 3c BY gk (Jn+ Fey
P
x M,)
Ore
+ yok, - (J a5 air WS M,): (10-155)
ot

If, moreover, the medium is linear and isotropic, P is e9x.£, and the
above time derivatives also cancel. Finally, if the point considered is not
inside a source and if Ohm’s law applies, J; = cE and the J; terms cancel.
We are then left with the V X M terms. If the medium is nonmagnetic, then
M is zero and
Wik «Bhs XB) = 0 (10-156)
If the medium is magnetic, we can perform a similar calculation using
H’s instead of B’s and Eq. 10-124. This gives

V-(E, X H, — Ey X H.) = 0. (10-157)

In summary, therefore, if two fields a and 6 are sinusoidal and of the

same frequency, if also the medium is linear and isotropic, and if the point
considered is not inside a source, then Eq. 10-157 applies. If, moreover, the
medium is nonmagnetic, then we have Eq. 10-156.
This result is known as Lorentz’s lemma. It is remarkable in that it
establishes a relation between two independent electromagnetic fields.
A more general discussion of this type will lead us to an important
theorem in Section 14-8.
448 MAXWELL’S EQUATIONS

Example If the a field is purely electric and the 4 field is purely magnetic, then
V-(E, X HM) = 0, (10-158)
subject only to the above limitations.
This surprising statement is easily shown to be correct by ex-
panding the divergence:
V(X Hy) = AW XE.) — EW X ii), (10-159)
OB, ODs\ _
= —H,: “ar E.- (Jn45 oe) =(0, (10-160)

since, by hypothesis, B, and E, are both zero.

Example The field of a small permanent magnet is that of a magnetic dipole

and is given by Eqs. 7-112 to 7-114. If the magnet carries an electric
charge Q, it also has an electric field
Q
iS etapa (10-161)
and
r, 8, Pi
Vie (Bie OT) — Vee Os (10-162)
Be Be 10

= V-(E,Bs 91), (10-163)

ys (a es sing.) = 0. (10-164)
4reor? 4rr

10.10 THE NONHOMOGENEOUS WAVE

EQUATIONS FORTE AND B

These wave equations follow immediately from Maxwell’s equations.

Taking the curl of Eq. 10-107 and remembering from Section 1.9.6 that
VXVXE= —VE+ V(V-B), (10-165)
we find that
VE — V(V-E) = as x B), (10-166)
and, if « and w are independent of the coordinates and of the time,
IE
V2E — V(V-E) = kd(oJ;+ eu =), (10-167)
Ot Ot
. CE Pf OJ;
V-E
2 —— ew i
ap v(”)
sx + a7
pain (10-168)
=

This is the nonhomogeneous wave equation for E in a homogeneous, linear,

and isotropic medium e, u.
10.10 THE NONHOMOGENEOUS WAVE EQUATIONS FOR E AND B 449

In regions where p; and J; are both zero, we have the homogeneous wave
equation
VeE — ew * = 0. (10-169)

For steady fields we have again Eq. 4-183:

VE=V", (10-170)

DAAC Bay 3) (10-171)

€0
This latter equation is valid for any medium, since we could also have
deduced it from Eqs. 10-107 and 10-105, as we did in Section 4.7.
For an infinite medium «, y, the solution of the complete nonhomogene-
ous equation 10-168 is

E=Ss tee [ Mier

: le
dr’ — al | OSs ifetle.dr’.5, (10-172)

Note that the velocity v of the E wave is (eu)!/? (see Appendix E), so that the
terms between brackets are taken at the time ¢ — (r/v), if E is calculated at
the time f. It will be shown in Problem 10-19 that the first term on the right is
—VYV, while the second is —dA/dt, so that Eis —VV — 0A/dt, as usual.
We can find the nonhomogeneous wave equation for B which corre-
sponds to Eq. 10-168 by multiplying Eq. 10-124 by uw and then taking its curl,
assuming that u is independent of the coordinates. Then

2h (7 A) sane (J+e =) (10-173)

or, using Eqs. 10-106 and 10-107, and assuming that « is also independent of
the coordinates,
VB — ey a PV OTA (10-174)
In regions where J; is zero we have the homogeneous equation
0B
V°B — ex ape 0. (10-175)

For steady fields,

VB = -uV X J; (10-176)
or, since p is, by hypothesis, independent of the coordinates, and because of
Eq. 9-52, we find again Eq. 9-103:
V2B —- — UV x Mrd fs (10-177)

The second equation is valid for any medium.

450 MAXWELL’S EQUATIONS

The solution of the non-homogeneous wave equation 10-174 for B is

p= +T |
J}qf
Ly Elis (10-179)
for an infinite medium e, u. This is the curl of A, as will be shown in Problem
10-18.
See also Problem 7-8.

10.11 SUMMARY

There is conservation of free charge, and therefore

Vey ee (10-4)
There is also conservation of bound charge, and, setting
oP
Jn tT Vv RM (10-6)

to be the current flowing in matter,

V-In = — = (10-7)

where p; = ps + pp iS the total charge density.

The total volume current density is

OE
he=Utont+2auxM, (10-9)
t Ot
)
= J; + - +VXM, (10-12)
and
VJ, = 0. (10-10)

The term 0D/0dt is called the displacement current density.

In an infinite medium e, u, the retarded potentials are

Ey [i bee
V=7- a dr’, (10-26)

ge k /
A=H |esWikar, (10-27)
where [p;], and [J;], are the free charge and current densities at the source
point P(x’, y’, z’) at the time t — (r/v), r is the distance between P’ and
 451

Retarded V Retarded A
Eq. 10-26 Eq. 10-27

Lorentz condition Conservation of charge

Eq. 10-65 Eq. 10-4

Nonhomogeneous wave equation for V

Eq. 10-85

oATE
Nonhomogeneous wave equation for A
Eq. 10-87

Figure 10-11.

the point P(x, y, z) where V and A are calculated at the time ft, and » =
1/(eu)"/? is the phase velocity of electromagnetic waves in the medium.
The Lorentz condition relates together the electromagnetic potentials V
and A: ~
VeAt eu - = 0. (10-65)

When Vis a function of the time, the six components of the field vectors E and
B can be obtained from the three components of A.
The nonhomogeneous wave equations for V and A in a homogeneous
medium e, » are as follows:

2V OV fe, pr
VV — ew WEES
ap ; ly
(0-85)

ree ee os seeuy, (10-87)

2

The reasoning that led us to the equation for the curl of B is shown
schematically in Figure 10-11. This is one of Maxwell’s equations:

1 OE
V x B eee C2
aegetend
at Lod ms (10-99 )

or
452 MAXWELL’S EQUATIONS

oD
Vier, Hise dla (10-94)

Maxwell’s equations can be written in various forms:

V-E == we (10-105)
€0

(10-106)
(10-107)
1 OE
VXB ria aae Hodm5 (10-108)

or

Vikan €0
(aE), (10-117)
V-B=0, (10-118)
OB
V xX E+ ae
—_
= (10-119)
1 OE oP
V x B Se
C2 Ot wo
(+o a +0 x u)-
M); (10-120)
Or

V-D= Py, (10-121)

Fun o (10-122)
VXE+%3 =0, (10-123)

oD
V xX H cams
F FsPS
di (10-124)
or, for fields which are sinusoidal functions of the time in a medium e, Ml,

Vek = py, (10-125)

Vu = 0, (10-126)
VX E+ jouH = 0, (10-127)
VX jock = fy (10-128)
In integral form, Eqs. 10-105 to 10-108 are written as follows:
 PROBLEMS
453

0:
ikE-d us
=»
(10-129)
E

ikB-da = 0, (10-130)

) Hdl d®
tees: (10-132)

f B-dl = pols, (10-133)

where S is any closed surface, Q; = Q; + Q, is the total charge enclosed by

S, C is any closed curve, ® is the magnetic flux linking C, and /; is the total
current flowing through C:

i= f (J.+ € ) - da, (10-133)

Ss ot
this S being any surface bounded by the curve C.
It follows from Maxwell’s equations that, to any field E, B, with p; = 0,
J; = O, there corresponds another field
E’ = — KB, (10-139)
H’ = +KD. (10-140)
If we have two fields E,, H., and E,, Hy, which are sinusoidal functions
of the time and of the same frequency, then, in a region outside the sources
where the medium is linear and isotropic,
V-(E. X H, — Eb X H,) = 0. (10-157)
This is Lorentz’s lemma.
Finally, we found the nonhomogeneous wave equations for EF and Bina
homogeneous medium ¢, p:
VEOy Ae: use
OE Ree pr
=V () Od;
Ps + mail (10-168)
2

Alp ip - eh AeA (10-174)

2H)

PROBLEMS
\

| 10-. What is the relaxation time for a semiconductor having a rather low conduc-
tivity of about 100 mhos/meter and a rather large relative permittivity of 15?
454 MAXWELL’S EQUATIONS

10-2. ‘Two plane parallel electrodes are separated by a plate of thickness s whose
conductivity o varies linearly from ao near the positive plate to a) + a near
the negative plate.
(a) Calculate the space charge density p; when the current density is J;.
(b) Calculate p; near both plates for a = 1.00 * 10° mhos/meter, a) + a =
2.00 10’ mhos/meter, J; = 1.00 ampere/meter?, e, = 3, s = 1.00 centimeter.

10-3. Show that the nonhomogeneous wave equations for V and A are compatible
with the conservation of charge.

10-4. Calculate the magnetic induction inside and outside a large flat conducting
plate if the current density J decreases linearly with depth z inside the plate,
J being equal to Jo(1 — az). The plate thickness is 1/a.

10-5. Deduce Eq. 2-5 for the electric field intensity due to a point charge Q, start-
ing from Maxwell’s equations.

10-6. Write out Maxwell’s equations in terms of E and H only for a nonhomo-
geneous medium in which e, and yu, are functions of the coordinates.

10-7. We have seen in Section 10.2 that

OA
E=—VV — ay.

BS VA:
where V and A are the scalar and vector potentials, respectively.
Rewrite Maxwell’s equations in terms of these potentials, for linear
homogeneous media.

10-8. At very high frequencies currents are limited to the region close to the surface
of a conductor and there is essentially zero electric and magnetic field in the
interior.
(a) Setting E, and B, to be the tangential electric field intensity and magnetic
induction in the conductor, show that

OE,
@z
_ _ 0B.Or
where the z-axis is normal to the surface of the conductor and points outward,
E, is taken to be positive in the direction of the x-axis, and B, in the direction
of the y-axis.
(b) Show that, outside the conductor, B is tangential to the surface, or very
nearly so.
(c) Show that B is related to the surface current density A by the equations

B= pA X n) and B= pod,
where nm is a unit vector normal to the surface of the conductor and pointing
outward.
(d) Does this result depend on the way in which the current varies with depth
inside the conductor?
(e) Can there be a normal component of E outside?
 PROBLEMS 455

10-9. Inside a superconductor

E—0 and 1 = Q,
under ideal conditions. This is true (except for a thin region, of the order of a
tenth of a micron thick, near the surface), even if the material is cooled in a
magnetic field.
(a) Show that, just outside, B is tangential to the surface.
(b) Show that J; = 0 inside. Therefore currents must flow very close to the
surface.
~ Assume steady-state conditions and set the magnetization M equal to
Zero.
(c) Does there exist a relation between the tangential B and the surface current
density?
(d) What can you say about the magnetic forces?
(e) Show that the magnetic flux linking a curve C which is entirely situated
inside a pure superconducting material must be constant.
(f) What happens if a superconducting ring is moved from a region where the
magnetic induction is B, to another where it is B,?
10-10. The conduction current density in a liquid metal is
\
J; = o(E + u X B),
where o is the conductivity, E is the electric field intensity, u is the velocity
of the fluid, and B is the magnetic induction, all quantities being measured
in the frame of reference of the laboratory.
The displacement current density is negligible.
(a) Show that the Lorentz force per unit volume of fluid is
o(E + u X B) X B.
For example, a conducting fluid is pushed in the direction of FE< B in
crossed electric and magnetic fields. This is the principle of operation of
electromagnetic pumps. See also Problem 8-8. The field u X B then opposes E.
The Lorentz force is equal to
o(E, + wu, X B) X B,

where the subscript | indicates a component perpendicular to B.

(b) Show that the Lorentz force can also be written as
cB*wv — u,),

where
EX
v= ile

is a velocity perpendicular to B.
This means that the Lorentz force tends to make the component of the
velocity that is perpendicular to the magnetic lines of force equal to the ve-
locity v.
(c) Show that
OB 1
—
Ey = M6 Ee SE ap ——V-B:
456 MAXWELL’S EQUATIONS

In liquid metals 1/opo is of the order of unity. For example, mercury has
a conductivity of about 10° mhos/meter, and 1/cpo is about 10/47.

10-11. Imagine an expanding spherically symmetrical universe in which there is

continuous creation of charge. Then, by symmetry, the vector potential A
must be purely radial.
(a) Show that, under steady-state conditions, the current density J,, must be
everywhere zero, according to Maxwell’s equations.
Since this did not appear reasonable to them, Lyttleton and Bondi (see
Problem 4-22) suggested that, if continuous charge creation does exist, then
Maxwell’s equations must be modified.
They suggested that one should write
1 0E 1
VxXB= poln + 55" - (G4)

yea"tt —(4y)
(3 ey.
where the new terms are set between parentheses. The quantities V and A are
the usual scalar and vector potentials, which are related to E and B as in
Section 10.2:

E=-VV-—- ae
Ot
B=V XA.
The other two equations of Maxwell for V X E and V-B would remain un-
changed.
Lyttleton and Bondi suggested that the constant /, which has the dimen-
sions of a length, would be of the order of the radius of the Universe. The
new terms would therefore be negligible in all but cosmological problems.
(b) If these modified Maxwell equations were correct, would V and A be
measurable, in principle?
Remember that, with the above equations for E and B, it is only the rates
of change of V and A that affect E and B.
(c) Write out the equation for the conservation of the total charge (Section
10.1) in its modified form, assuming that charge is everywhere created at the
constant rate of g coulombs/meter*-second.
(d) Would the Lorentz condition (Section 10.3) still be valid?
(e) Now set A = A’r, where 4’ is a constant, and assume V to be a constant.
Show that

B= 0, E= 0, Jm = (q/3)r, pe = OV /P.

Assuming that the velocity of the outward flow of matter is the same as
that of the charge, namely J,,/p:, it follows that the radial velocity is propor-
tional to r, which is consistent with the linear velocity-distance relation ob-
served by astronomers: v = r/T, where T +3 X 10!” seconds is the Hubble
constant.

(f) Show that p, = g7/3.

PROBLEMS 457

Now the space-charge density

—_— ——— e
Pt oe?

where » + 10~*6 kilogram/meter® is the mass density, m is the proton mass

(the Universe is mostly hydrogen), and ye is the excess charge of a hydrogen
atom (See Problem 4-22). Similarly, the rate of charge creation g is related to
the rate of mass creation Q as follows:

Ee
oa (2 oi
(g) Show that, if this theory is correct, then Q ~1 hydrogen atom per
2 X 10!* meters?/second.
10-12. (a) Starting from an electromagnetic field characterized by the field vectors
E and H, utilize the duality property of electromagnetic fields four times to
find successively the fields FE’, H'; Et, H¥; Eiii, Hii: and Ev, Hiv,
(b) Assuming that E and H are respectively parallel to the x- and y-axes, show
the 10 vectors on 5 separate diagrams.
(c) Compare the fields when the constant of proportionality K is equal to the
velocity of light in the medium, that is, 1/(eu)!/*.
10-13. (a) Show that Maxwell’s equations for free space are invariant under the
transformation
E’ = Ecos6é+ cBsin 8,
B’ = —(E/c) sin@ + Bcos@.
The transformation E’ = —KB, H’ = KD of Section 10.8 and the trans-
formation E’ = —E, B’ = —B are special cases corresponding to 6 = 7/2
and to 6 = a respectively.
(b) Show that the energy density

=
1 1
2 e9H 2 + ea
oun B 2

and the Poynting vector FE < H for a plane electromagnetic wave (Section
11.1.1) are also invariant under this transformation.t

10-14. A rectangular plate ABCD of resistive material has a uniform thickness s

and a conductivity oc. If conducting electrodes are fixed to the two edges AB
and CD, the resistance is R;. When conducting electrodes are fixed to the
edges BC and DA, the resistance is R».
Show that

This equation also applies to any region ABCD bounded by lines of

current flow and equipotentials (which are interchanged when the connections
are modified as above). We shall use this theorem* in Problem 13-7.

* Proc. [EEE 55, 1238 (1967). See also D. J. Epstein, Proc, IEEE, 56, 198 (1968).
458 MAXWELL’S EQUATIONS

10-15. Verify Lorentz’s lemma graphically by considering two pairs of orthogonal

vectors E,, H,, and E,, H, lying in the same plane.

10-16. Imagine various pairs of electromagnetic fields in free space, and verify
whether Lorentz’s lemma applies.{

10-17. In the example on page 434 we showed that the leakage current in a spherical
capacitor does not produce a magnetic field. There are two other ways of
demonstrating this.
(a) Show that V x A = 0.
(b) Use the fact that
pe | Cs dr’.
Gr }_,
SF
r

(c) Use also Problem 4-7 to show that V X Hi is zero.

10-18. We have shown in Section 10.10 that

jeerArones|, Aer dr’

7i

for an infinite medium yp and if retardation effects are negligible.

Show that the term on the right is V X 4.t
10-19. Show that the two terms on the right-hand side of Eq. 10-172 are respectively
—VV and —04/ot.
Assume that the retardation is negligible and omit the brackets.{

Hints

10-13(b). Use Eq. 11-17.

10-16. You can try a capacitor in a magnetic field, the fields of the electric and magnetic
dipoles, the electric and magnetic fields in a cylindrical magnetron, and so forth.

10-18. Use Eq. 9-106.

10-19. Use Eq. 4-186.

CHAPTER ll

PROPAGATION OF
ELECTROMAGNETIC WAVES I
Plane Waves in Infinite Media

In this chapter we shall study the basic aspects of the propagation of electro-
magnetic waves in infinite media; reflection and refraction phenomena will
be the subject of the next chapter, after which we shall be prepared to study
guided waves in Chapter 13. We shall not deal with sources of electromag-
netic waves until Chapter 14.
Let us start the present discussion with propagation in free space, and
then take up three types of media: dielectrics, conductors, and ionized gases.
Dielectrics, which are defined as nonconductors, can be either magnetic or
nonmagnetic.
At this stage you would be well advised to work through Appendixes D
and E, unless you are familiar with the exponential notation for representing
the cosine function, and with wave propagation.
Maxwell’s equations impose no limit on the frequency of electromagnetic
waves. To date, the spectrum that has been investigated experimentally is
shown in Figure 11-1. It extends continuously from the long radio waves to
the very high energy gamma rays observed in cosmic radiation. In the former
the frequencies are about 104 hertz and the wavelengths about 3 X 10 meters;
in the latter the frequencies are of the order of 10% hertz (and higher), and
the wavelengths of the order of 3 & 10~' meter (and shorter). The known
spectrum thus covers a range of 20 or more orders of magnitude. Radio,
light, and heat waves, X-rays and gamma rays—all are electromagnetic,
although the sources and the detectors, as well as the modes of interaction
with matter, vary widely as the frequency changes by orders of magnitude.
The fundamental identity of all these types of waves is demonstrated by
460

Logo hf (electron volts)

—11 —10 —9-8 —7 -—6 -5 —-4 -3-2-10 1 2 3 4 5 6 7 8 9 10
steer etaeentr 6rakes chenrerverer iebeserandorent abrreesch ce leerete rie “eer caleen Ata ean Rt OA ETE ES

!
Log, / (meters)
5 4 3 2 1 O —1 —2 -—3 —4 —5 —6 —7 —8 —9—10 -11—12-—13-14
-15-16
(theo edeee teeter heehee th eben See es ae Sp eS oe eee

|
|
Logi, f (hertz)
| eNO) AM A Sle aie ils UG IG, ile iS), BIN PAL Bek pie Os!

—— Gamma rays

~ Visible X-rays
light
Ultra
violet
ae
HESTON
mies
ty
as
BS
2 Infra
= red ee

Figure 11-1. The spectrum of electromagnetic waves. The abbreviations VLF,

LF, MF, ... mean, respectively, very low frequency, low frequency, medium
frequency, high frequency, very high frequency, ultrahigh frequency, super high
frequency, and extremely high frequency. The limits indicated by the shaded re-
gions are approximate. The energy /f, where / is Planck’s constant (6.63 107*4
joule-second) andf is the frequency, is that of a photon or quantum of radiation.

many experiments covering overlapping parts of the spectrum. It is also

demonstrated by the fact that in free space they are all transverse waves with
a common velocity of propagation c given by Eq. 11-7. For example, simul-
taneous radio and optical observations on flare stars has shown that the
velocity of propagation is the same within experimental error for wavelengths
differing by more than six orders of magnitude.
We shall use H rather than B in discussing electromagnetic waves, in
spite of the fact that, until now, we have used H only for magnetic materials.
There are two reasons for using H instead of B in dealing with electromagnetic
waves: one is that FEX H is an energy flux density; the other is that E/H has
the dimensions of an impedance. These two concepts prove to be of great
practical value.

ltd PLANE, BLECTROMAGNE IG

WAVES IN FREE SPACE

Let us start with the relatively simple case of plane sinusoidal waves propa-
gating in a vacuum in a region infinitely remote from matter. Then Maxwell’s
11.1 PLANE ELECTROMAGNETIC WAVES IN FREE SPACE 461

equations 10-125 to 10-128 reduce to

V-E=0, (11-1)
T= 0) (11-2)
Wah a 10) (11-3)
WA SUE lcd) a) (11-4)
We obtain wave equations for E and H, as in Section 10-10 by taking the
curls of Eqs. 11-3 and 11-4:
VE + equ wk = 0, (11-5)
V7 + equ HH = 0. (11-6)
These differential equations are those of an unattenuated wave traveling
at the velocity 1/(¢uo)”. It follows that the field vectors can be propagated as
waves in free space at the velocity
1
(11-7)
es (€ U0)!”

These are two remarkable results. We have deduced from our investi-
gation of the basic electromagnetic phenomena: (a) the possibility of the
existence of electromagnetic waves and (b) the velocity of such waves in free
space.
The above expression for c is in itself remarkable. It links three basic
constants of electromagnetism: the velocity of an electromagnetic wave c,
the permittivity of free space ¢, which we first met in Section 2.1 while dis-
cussing Coulomb’s law, and the permeability of free space po, which enters
into the magnetic force law of Section 7.1.
It will be remembered from Sections 6.1 and 7.1 that the constant uo was
defined arbitrarily to be exactly 4r X 10~7 henry/meter. The constant e can
thus be deduced from the measured value for the velocity of electromagnetic
waves,
c = 2.9979 X 108 meters/second: (11-8)

é = = =.6.5)42 < 10," farad/ meter. (11-9)

The permittivity of free space ¢) can also be determined directly from measure-
ments involving electrostatic phenomena. The measurements lead to the above
value within experimental error, thereby confirming the theory.
For a plane electromagnetic wave propagating in the positive direction
of the z-axis, E is independent of x and y,

ae De OZ
(11-10)
462

Figure 11-2. Decomposition of

the vector E into two vectors F;
and Fk).

and the z component of E cannot be a function of z. We shall set

E, = 0, (11-11)
since we are interested in waves and not in uniform fields.
The same argument applies to the H vector, and
HH, = 0. (11-12)
A plane electromagnetic wave propagating in free space is therefore
transverse, since it has no longitudinal components.
We now assume that the wave is plane-polarized with its E vector point-
ing in the direction of the x-axis.* This does not involve any loss of generality
since any plane-polarized wave can be considered to be the sum of two waves
that are plane-polarized in perpendicular directions and in phase. For ex-
ample, in Figure 11-2, the vector E can be resolved into two mutually per-
pendicular vectors FE, and E).+
For a plane-polarized wave having its E vector in the direction of the
x-axis (Figure 11-3),

E = Eyexp jw (:= 4 i, (11-13)

* The plane of polarization was originally taken to be the plane containing the direction
of propagation and the H vector. It is now common either to avoid the use of the term
plane of polarization entirely or to use the term plane of vibration, which is defined to be the
plane containing the direction of propagation and the E vector. Radio engineers say that
a wave is polarized in the direction of its E vector. We shall always refer to the orientations
of the E vector as above.
+ One can also add two plane-polarized waves that differ in phase. Then, at a given point,
the maxima of E; and of FE, do not occur at the same time, and their sum E describes an
ellipse about the z-axis. We then have an elliptically polarized wave.
If E; and E, have equal amplitudes but are +/2 out of phase, the ellipse becomes a
circle and the wave is said to be circularly polarized. The polarization is said to be right-
or left-handed according to whether the vectors E and H rotate clockwise or counterclock-
wise for an observer looking toward the source.
 463

MAAK A Y YYWyyYY ¥ K KAAKAK A A

jejeje} @) @ x | x |x|x\xx\x!x| x | x © | @ jeleleesele| e |
e xXx ee,
lee
B|° x ee
xx/x}x
°{cf8si°
elelee
iste e x xx x e jee C)
e X RRR |x e\*sgit\e
Re e x rete x e lssie e

**) 6 x | PRREXPT e |*leigge\*)«

°. x PRX] see.
e
e@ele)e@
x x]
a
x | x |x |xkx\x/x|x | x
ee
ejelee
@ | @ jeleseee| e |
RRA OY YYWWYY Y 4 AAMM AD4
(b)

Figure 11-3. The E and H vectors for a plane electromagnetic wave traveling in
the positive direction along the z-axis. (a) The variation of E and H with z ata
particular moment. The two vectors are in phase, but perpendicular to each other.
(b) The corresponding lines of force as seen when looking down on the xz-plane.
The lines represent the electric field. The dots represent magnetic lines of force
coming out of the paper, and the crosses represent magnetic lines of force going
into the paper. The vector E X H gives the direction of propagation.

and, from either Eqs. 11-3 or 11-4,

H, = 0, (11-14)
H = ac Eyexp jw (1= “)j. (11-15)
Therefore H is perpendicular to EF, and
1/2
See) i (“) = 377 ohms, (11-16)
H €0C €0

3 = c = 3.00 X 10% meters/second. (11-17)

 Figure 11-4. The energy density —£? =
oH? as a function of z for a plane
electromagnetic wave traveling along
z the z-axis in free space.

The E and H vectors are perpendicular and oriented in such a way that
their vector product E X H points in the direction of propagation, as in
Figure 11-3. The E and H vectors are in phase, since E,/H, is real, and they
have the same relative magnitudes at all points at all times.
The electric and magnetic energy densities are in phase and are equal,
since
teok?
+ ie (11-18)
2

At any instant, the total energy density is therefore distributed as in Figure

11-4.

11.1.1 The Poynting Vector

We have found that a plane electromagnetic wave in free space propagates in
the direction of the vector E X H. Let us calculate the divergence of this
vector for any electromagnetic field in free space:

V-(EX H)= —E(WV XH)+

H-V X £), (11-19)
=e (F-0)
dE es (1-10 oH
=) (11-20)
4 0 feok oH?
oY al hes J (11-21)

Integrating over a volume r bounded by a surface S, and using the divergence

theorem,

i
[ex
Lee -da =
de
0 f[ (cE? | mol?
2{ (2 + 5 )ar, -22
(11-22)

The integral on the right is the sum of the electric and magnetic energies,
according to Sections 2.15 and 8.5.1. The right-hand side is thus the energy
lost per unit time by the volume 7 and the left-hand side must be the total
outward flux of energy through the surface S bounding r.
The quantity
“S=EXH (11-23)
11.1 PLANE ELECTROMAGNETIC WAVES IN FREE SPACE 465

is cal

The instantaneous value of the Poynting vector at a given point in space

is E X H or, according to Eq. 11-16,

For a plane sinusoidal wave, the average value of § is

Sav = CéepEims k = 4 ceokG k. (11-25)
= 2.66 X 10°? E?ms k watts/meter’. (11-26)
The energy can therefore be considered to travel with an average density

4 egEtms + 4 poHims = «oErms = 4 €oEo (11-27)

at the velocity of propagation ck.*

Example A laser beam has a power of 20 gigawatts and a diameter of 2 milli-

meters. Let us calculate the peak valuesof EandofB. , _ Q
From Eq. 11-26, iN
LOS 20 exe? V2
Eo eas PEI Breve == 21/2 5) (11-28)
200 ete al Oe?
2 2a)? Volts/meter (11-29)
This is an enormous electric field; it corresponds to a voltage differ-
ence of about a quarter of a volt over a distance of one angstrom
(10~'° meter), which is about the diameter of an atom.
We can now find By from Eq. 11-17:
Ee ee (2
B , — - = 3 x< 108 = 7.3
z teslas
¢ > (11-30)
-

This is about ten times larger than the magnetic induction between
the pole pieces of a powerful permanent magnet.

Example ENERGY FLOW THROUGH AN IMAGINARY

CYLINDER
Figure 11-5 shows wave fronts of a plane electromagnetic wave flow-
ing through an imaginary cylinder. If the amplitude of the wave is
constant, the electric and magnetic energy densities averaged over
one period are constant, and the net outward energy flow is zero.

* See The Feynman Lectures on Physics (Addison Wesley, Reading, Mass., 1964), Volume
II, Section 27-4 for an interesting discussion on energy density and energy flow in electro-
magnetic fields.
466

/ === —— /7

j\3- —/ gs
NGS eet wz ‘

ee
———— \ 7
owes SSS \ i
ee————
ae

Figure 11-5. The electromagnetic wave emanating from the horn sweeps
through the imaginary cylinder. The amplitude gradually decreases at
the source, and the intensity, represented by the Poynting vectors, is
larger on the right than on the left, as in Figure 11-6.

However, if the electromagnetic wave decreases in amplitude, the

average energy density decreases with time, and there must be a net
outward flux of energy.
The instantaneous value of E for a plane wave traveling in the
positive direction along the z-axis is the same as that near the source
at a previous time t — (z/c), where z is the distance from the source.
If the amplitude decreases linearly with time, the electric field in-
tensity at the source is
E = E01 — at) exp jot, (11-31)

whereas, at a distance z,

E = E{1- a(t= =)}exp joo(:= =), (11-32)

At any given time f, the wave then has the general shape shown in
Figure 11-6.
We now calculate the rate at which the electromagnetic energy
enclosed within the cylindrical volume decreases with time. We
assume that E changes only slightly during a period 1/f, so that
averages can be calculated as above. This requires that a/f be much
smaller than unity.
The net average outward energy flow is obtained by evaluating
the surface integral of E X H over the two ends of the cylinder, the
left-hand end being at z) and the right-hand end at z) + L. Since
H = eck,

ie(E X H)-da

= Dseas {(1 — ar ¢ SFO)

Cc
a + @CoM"1, (11-33)
S52 {2a(1 — at) + ©(22 + DL};
(11-34)
 467

Figure 11-6. Sine wave from a source whose amplitude decreases linearly
with time. The wave travels to the right.

whereas the average energy enclosed at the time ¢ is

z+L €
W= ii 75Eo (1— ) S dz, (11-35)

= Sf — ab + SL! + 3b? + 3281)

+2 = anQQa+L)L}- (11-36)
Then
_ dWdt _ SEB {2a(1 ive © (2% ae DL}. (11-37)
The net average outward energy flow per unit time is thus equal
to the rate at which the enclosed energy decreases with time, as
expected.

11.2 THE E AND H VECTORS IN HOMOGENEOUS,

ISOTROPIC, LINEAR, AND STATIONARY MEDIA

We now go on to the propagation of plane electromagnetic waves in homo-

geneous, isotropic, linear, and stationary media. We repeat the definitions of
these terms that we gave in the introduction to Chapter 10.

ee a)
468 PROPAGATION OF ELECTROMAGNETIC WAVES I

where e, », and the conductivity o are constants independent of FE and H, and

independent of direction. Crystalline media are usually anisotropic. We
shall call a medium stationary if it is at rest with respect to the coordinate
system used.
We assume that there are no sources in the region under consideration.
We also assume that the medium of propagation is infinite in extent. This
will avoid reflection and refraction; these phenomena will be the subject of the
next chapter.
The wave equations for E and H in such media follow from Eqs.10-168
and 10-174 if we simply substitute cE for J; and wH for B:
:2 ee CE OE ce Ny py
V7E — ew ap 7 OM pele
oe V ltl) 2
(11-39)

CH oH
WH = a TCP i le EP cae 0. (11-40)

In both equations the second term on the left comes from the displacement
current, while the third comes from the conduction current. The two equa-
tions are identical, except for the fact that there is no magnetic equivalent of
the electric charge density p;. These differential equations describe an atten-
uated wave (see Eq. E-55 in Appendix E).
Let us first dispose of the charge density that appears in Eq. 11-39. We
consider again a plane wave propagating in the positive direction of the
z-axis. Then all derivatives with respect to x and to y are zero, so that

aE we et”
O Kate
toes On :
(11-41)

Then
Vv (Me 0 nt) ae 0° i -49
e Oz (2 I gz" Hak tt
and the wave equation for E becomes

a” ‘ ‘ 0° 0 : :
a7 (£.i + B,J) — ple ait +o a (44 + Ej + Bk) = 0. (11-43)

The longitudinal component E£, must therefore satisfy the equation

@E, dE,
€ pate ere Bt (11-44)
with the result that E,, if it exists, must be of the form

E, = a+ bexp (—at/e), (11-45)

where a and 6 are constants of integration independent of t. Thus E, must
11.2 THE E AND HT veEcrors 469

decrease exponentially with time and there is no E, wave. If ¢ = 0, E, is of the

form a + bt and again there is no wave. Hence we may set
E. = 0, (11-46)
since we are concerned solely with wave propagation. Also, from Eq. 11-41,
we may set
pr = O (11-47)
for a plane wave in a conducting medium. This is in agreement with our
discussion of the example on page 424.
ee ee ae the wave again has no longitudinal
component of E.
We have therefore found that, for plane electromagnetic waves, (a) we
may set py = 0 in Eq. 11-39, and (b) the E vector is transverse.
It is easy to show that the H vector is also transverse: the divergence of
H being equal to zero,
eae) (11-48)
because the derivatives with respect to x and to y are both zero, by hypothesis.
Then H. is not a function of z, and, if we are concerned only with waves, we
may set
teh 210), (11-49)

Then the H vector is also transverse.

Plane electromagnetic waves are therefore transverse in any homogeneous,
isotropic, linear, and stationary medium.
Now that we have shown that FE and H are transverse, let us investigate
their relative orientations. We assume again a plane-polarized wave with
the E vector parallel to the x-axis:
E = Eyexp j(wt — kz)i (11-50)

where the wave number k (Appendix E) is in general complex. Then, from

the Maxwell equation 11-3,
fej
0 0 —jk| = — jw u(H.t
+ H,)). (11-51)
E 0 0
Then the x component of H is zero and

Heer eye). (11-52)

wp
470 PROPAGATION OF ELECTROMAGNETIC WAVES I

The E and H vectors in a plane-polarized wave are therefore (a) mutually

perpendicular; (b) oriented so that their vector product E X H points in the
direction of propagation; (c) not necessarily in phase, because the wave
number & can be complex.

11.3 PROPAGATION OF PLANE ELECTROMAGNETIC

WAVES IN NONCONDUCTORS

In nonconductors, « = 0, as a rule py = 0, and the wave equations 11-39

and 11-40 reduce to
RE = 0,
VE — on 2S (11-54)
VH — eu a a (11-55)
2

Substituting the values of E and H from Eqs. 11-50 and 11-52, we find that
—kh+ ww = 0. (11-56)

k= oe) (11-57)
is real and there i attenuation.

The phase velocity in nonconductors is therefore Jess than in free space and
th is

(11-59)

nee (11-60)
Note however that n and e, are both functions of the frequency. We have
discussed briefly the variation of ¢, with frequency in Sections 3.10 and 10.2.1.
Since tables of n are usually compiled at optical frequencies, whereas «, is
usually measured at much lower frequencies, such pairs of values cannot be
expected to correspond.
Also, from Eq. 11-53,

H, = ()"z (11-61)
11.4 PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 471

and
1/2
H = Ayexp j(wt — kz)j = () Ey exp j(wt — kz)j. (11-62)

——_—_
eB = bat (11-63)
The total instantaneous energy density is thus eE? or »H?, and the average
total energy density is eEtms, Or uHims. The average value of the Poynting
vector is
] ] €\1/2 5
Sav = g) EoHy k = zu Eo k, (11-64)

€\12 1 A
— Mu Bens k —- (cu)! Chae k, a 1-65)

= ucE2,; k, (11-66)
1/2
= 2.66 X 10-3 (=) E>, k watts/meter?. (11-67)

The average value of the Poynting vector is equal to the phase velocity wu
multiplied by the average energy density.

Example If the 20 gigawatt laser beam of the example given at the end of Sec-
tion 11.1.1 passes through a glass whose index of refraction is 1.6,
then «!”? is 1.6, u, is unity, and E) is smaller than in air by a factor
of 1.61, or by 1.26. Then & in the glass is 1.7 < 10° volts/meter.
To calculate By we can use the fact that the Poynting vector
(3)E, Hp is the same in the glass as in the air; since E is smaller in
the glass by a factor of 1.26, By must be /arger by the same factor.
You can check that this makes the electric energy density equal
to the magnetic energy density (Eq. 11-63).

11.4 PROPAGATION OF PLANE ELECTROMAGNETIC

WAVES IN CONDUCTING MEDIA

We must now solve the two wave equations 11-39 and 11-40 for a wave
traveling along the z-axis, with p, = 0. Substituting again Eqs. 11-50 and
11-52, we find that

—k? + wew — jwop = 0, (11-68)

er etAo (1e 2),

2
Wwe
(11-69)
472 PROPAGATION OF ELECTROMAGNETIC WAVES I

where Xp = Ao/2m7 = C/w is the radian length for a wave of the same angular
frequency w propagating in free space.
The quantity jwe/o is the ratio of the displacement current density dD/dt
to the conduction current density J; = cE. We shall call the magnitude of
this ratio the © of the medium:

aD)
Ot
\aeE
Ot we Ge
SO A ae er (11-70)
For nonconductors, Q—~ «©. For common types of conductors, o is of the
order of 107 mhos/meter (5.8 X 107 for copper) and we can set e, & | (ex-
ample, page 439). The ratio Q is thus very small for the usual conductors. At
optical frequencies o is a function of frequency, as explained in the Example
on page 424.
Then
o _ Str f, J\T ;
= re (1 Z) (11-71)

and the wave number k is complex:

k= k, — jk; (11-72)
where k, and k; are both positive:
S af ent oN ms. bles
ke = 5 () haw ee eae , (11-73)

(2) rg) —Y (74

1 sey" ( l )2 eS
k; a haie ; 1 —- | -

1/2 1/4
ie = (ertir) 1+ - exp {—/j are tan (k;/k,)\. (11-75)
ie @
Our expression for the arc tan function is correct, since k, is a positive quantity
(Appendix E). In a vacuum, k, = 1/%o, ki = 0.
The real part k, of the wave number is 1/A = 27/), where ) is the wave-
length in the medium; the imaginary part k; is the reciprocal of the distance 6
over which the amplitude is attenuated by a factor of e. The quantity 6 = 1/k;,
is called the attenuation distance.
Again from Appendix E, the phase velocity is

u= , (11-76)

+ Note that k? is used here for the square of the complex number k, and not for the
square of its modulus.
 11.4 PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 473

and the index of refraction is complex

nN = Kok = K(k, — Jk;).

(11-77)

The ratio E/H is again given by Eq. 11-53:

Ee One ee Be 1 .
om amn\te (eer: (11-78)
2

where - :
k;
6 = arc tan k. (11-79)
is the phase of E with respect to H.
Also,
E= Eyexp {jt — k,z) — k,z} i, (11-80)
H = Hyexp {jot — k,z — 6) — kz} j, (11-81)
with

(11-82)

The electric and magnetic energy densities are in the ratio

pe I
1 94

(11-83)
a ereee
5 HH (1+)

and the average total energy density is

55(;
oieHee He )\exp
0 + 5 HH6 exp (a kzle
iZ) = See
4 Eo (Tee)
+ & Poe 2 k iZ).
(11-84)

11.4.1 The Poynting Vector in Conducting Media

The Poynting vector
S=EXH (11-85)
was discussed in Section 11.1.1 for a wave propagating in free space. We
shall now reexamine this vector for the more general case of conducting
media, and then we shall find a general expression for calculating the time
average of the energy flux in a field.
We again have the vector identity
V-(EX H) = HV X E)— E-(V X BH). (11-86)
474 PROPAGATION OF ELECTROMAGNETIC WAVES I

Now, from Eqs. 10-123 and 10-124 for linear and isotropic media,

V-(E X H) = — Hu TE B.(e22 + J) (11-87)

0 [1 1
i at (5pH? + 2 ce) a E- Jy. (11-88)

Integrating over a volume 7 and using the divergence theorem on the left-hand
side, and then changing signs,

— | ex mda = of (Sam + jum) art | Beda, (11-89)

S . :
where S is the closed surface bounding the volume r.
The first term on the right gives the increase in the electric and magnetic
energies in the volume 7 per unit time; the second integral gives the electro-
magnetic energy removed from the same volume and transformed into heat
energy, also per unit time. Then the term on the left, with its negative sign,
must represent the rate at which electromagnetic energy flows into the vol-
ume 7.
Then

fe x H)-da

again gives the total outward flow of electromagnetic energy per unit time
through the surface S, as in Section 11.1.1.
As a rule, both E and H are expressed as exponential functions of the
time and of the space coordinates. However, since we require the product of
two such quantities, we must revert to the cosine functions (Appendix D).
For example, if E is of the form
Ey exp j(wt — kz)
one must write instead
Ey exp (—k,z) cos (wt — k,z).

In practice, it is more convenient to use the following formula, which

gives directly the time average of 8, with E and H expressed as exponential
functions:

Re (i x H*), (11-90)

where H* is the complex conjugate of Ht, and where the operator Re means
“Real part of.”

} The complex conjugate is obtained by substituting —/ for /.

11.5 PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 475

We can show this equation to be correct in the following way. Let us

write E and H in the form

E = (Eo, + j Enije**, (11-91)

H = (Mo, + j Moide**, (11-92)
where Eo,, Eo:;, Hor, Ho; are real functions of x, y, z.
Then
Sav = (Re E X Re A). (11-93)
= {(Eo, cos wt — Ey; sin wt) X (Hor cos wt — Hp; sin wt) }av. (11-94)
Now the average value of cos? wf is 3, the average of sin? wt is also 4, and the
average of sin wf cos wf is zero. Thus

Sav = 5 (Eo, x A, te Eo; x H;). (11-95)

But
ane (Ex H*) Zz Re |(Eor + j Foi) X (Hor — jHy)}, (11-96)

= 5 (Ey X Mo, + Eo: X Hoi), (11-97)

as above. We have therefore demonstrated the validity of Eq. 11-90 for
electromagnetic fields that are sinusoidal functions of the time. We have
made no assumption as to the relative orientations and phases of the vectors
E and H.
It is not difficult to show that

|Sv=(Average energy density) X(Phase velocity). (11-98)

We can thus consider again, in this more general case, that the average energy
density is propagated at the phase velocity.

11.5 PROPAGATION OF PLANE ELECTROMAGNETIC

WAVES IN GOOD CONDUCTORS

Let us return to the values of k, and k; for conducting media as given in Eqs.
11-73 and 11-74. In good conductors Q is much smaller than unity and

Abe
{ 1 1/2 ie | | ic)2 1/2 11-99

I talaee) ee HG eft, apse

A (<) (1+5),
Pe one ;
(11-100)
oe (<)" (11-101)
Q
within 1% when
476 PROPAGATION OF ELECTROMAGNETIC WAVES I

a oe (11-102)
or when the conduction current density cE is at least 50 times larger than
the displacement current density 0D/dt.
Good conductors will be defined as those for which the above condition is
satisfied. One must remember here that o is a function of w at optical fre-
quencies. According to this definition, copper is a ‘“‘good’’ conductor up to
frequencies of about 2 X 10'* hertz, or to the ultraviolet.
For good conductors, Eq. 11-68 simplifies to

k2 = — joy, (11-103)
and
1/2

k = (—jaon)'? = ($) a= 7 (11-104)

Then k,; = k, and
2 1/2

From Eq. 11-53, we have tha,Ab nee

1sdifference comes from the fact that the current

that is associated with H in conductors is the conduction current, and
not the displacement current as in nonconductors.
From Eqs. 11-50 and 11-52,

E = Eyexp 1 (or= ;)2 i}i, (11-107)

ee (2\" ere i («1oo 5)2 shi (11-108)

or, in terms of cosine functions,

E = Eye” cos (or_ 5)i, (11-109)

H = (=)" Eye-?/5 cos (or- ;- i):F (11-110)

= Hye-*!5 cos («1= := :)i. (11-111)

Figure 11-7 shows the curves of E/E) and of HyHy fort =O,
 Figure 11-7. The ratios
E/E, = e-#!® cos {wt — (z/6)}
and
H/H, = e*# cos {wt — (z/6) — (@/4)}
at ¢ = O as functions of z/X = z/6 for an electromagnetic
wave propagating in the positive direction of the z-axis in a
good conductor.

The amplitude of the wave is attenuated by a factor of 1/e = 0.368 in

one radian length i, and by a factor of (1/e)’" + 2 X 10-* in one wavelength
A, whereas the Poynting vector is attenuated by (1/e)? = 0.135 in X% and by
(1/e)** + 4 X 10° in X. The attenuation is so rapid that the wave is barely
discernible (Figure 11-7).
The attenuation distance 6 in conductors is called the skin depth, or the
depth of penetration. The skin depth decreases if either the conductivity o, the
relative permeability yu,, or the frequencyf increases. Good conductors are
therefore always opaque to light, except in the form of extremely thin films.
It does not follow, however, that substances that are nonconducting at low
frequencies are necessarily transparent at optical frequencies.
Table 11-1 shows the skin depth 6 for various conductors at four typical
frequencies. Note that the attenuation in iron is much larger than in silver,
despite the fact that iron is a relatively poor conductor.
 9
4of

*T-TT

afqey
syJdaq
ee

eee
dANLIOY

ee
urys

u1ySeee
ee

sdojonpuoy)
AyIANONpuOD
© Ayyiqeewlag zi1t9

@yideq
09 Z1I9H ZOYOITY
[| ZoYyRsIPy
€ ZHOYRSID
Jojonpuod /oyw) (1939 all /s9}0UL1)
PUODaS
—(¢,, (sI9}OUNTJUIO)
© (SIsJOUNTTFTW)
$= (SIo}oUNTTTIUN)
= (suosorur
cIMnUTnyy PSE LOLS 00'T £800 Et Lic £800 yl
sseig g°¢9) ‘ND Z'PE (UZ 69'T X LOT 00'T 9710 €9'T 86 9710 0€C
unruo1y) =Beé= X LOI 00'T 180°0 OT OG 180°0 ot
teddo5 08'S X 101 00'T 990°0 S30 eg 990'0 Cal
PIOD OS'P X LOT 00'T ¢L0'0 L60 8E°C SL0°0 vl
amydein OT lx 00°T I 6S SOC €0S 6ST 0°67
SHeuse
uor OT TX 7 X 20T T10°0 vo se0 T10°0 020
Tejouny
SL) ‘INZ ‘ID
¢ ‘ND81 (94 910 X LOI Z X vOl 6200°0 Le0'0 c60°0 6c00°0 €S0°0
[93{9IN a a Me, rb T X 2Ol vt0'0 810 vv vlo0 97°0
BIS J3}8M OS~ 00'T 7 X 2Ol € X 201 L X s0OI 76 AUS 4
J9ATIS ST9 X LOL 00'T 790°0 €8°0 £0'C 790'0 cl
& an OL8°0
X LOT 00'T ILO ECE Tp’s ILT0 cle
OUT, 98'T X 101 00°T LITO IST OL'E LLCO VT

:dounos pojdepy
wo. IY} uvdiaaup ajnjijsuy
fo saisdyg YOoqpuny ‘TITH-MBIDO])
MON OA ‘(£961
« IV @ = 7000 "E[S2}
4} 1Vv stui Aouanbay
O sI ynoGe
‘Zz puk vas Ja}eM
SI JOUB ,,poos,, 1OJONPUOD
*) & ‘(OL
re
11.5 PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 479

The phase velocity

ORAL
w= =Wh= (=) (11-112)
ou
is proportional to the square root of the frequency.
The ratio of the electric to the magnetic energy density is

5eB Wwe 1
=—— = 0S = (11-113)
Lelie o 50
D) 0

and the energy is essentially all in the magnetic form. This results from the
large conductivity o, which causes E/J; to be small. The electric field intensity
is weak, but the current density, and hence H, are relatively large.
From Eggs. 11-90 and 11-106, the average value of the Poynting vector is
1
sa 5 Re (ESCH): (11-114)

bs:(6) e-22/5 Fk. 1/2

(11-115)

Example PROPAGATION IN COPPER AT 1 MEGAHERTZ

Copper has a conductivity o of 5.80 107 mhos/meter. Then, at 1

megahertz.
pec 108 XS Boe
NO (11-116)
oe 5.80 & 107
The requirement that Q be less than ;'5 is indeed well met!
The skin depth, which is also the radian length, is

2 1/2
§=1 = Goon xis)
ll 6.6 < 10~ meter, (11-117)

= 66 microns. (11-118)

The wavelength 27x at 1 megahertz is about 0.4 millimeter in copper,

while it is 300 meters in air. In fact, the ratio of the wavelengths
in copper and in air is
r w 1/2 ;
— = 27 (=) Feo)? = 1.4 610-82) (11-119)
do WO Lo

The phase velocity is correspondingly low:

u = wh = 2r X 10° X 6.6 X 10> = 4.1 X 10? meters/second.

(11-120)
480 PROPAGATION OF ELECTROMAGNETIC WAVES I

This is about ten times smaller than the velocity of sound in copper,
which is 3.6 X 103 meters/second. Also,
| (* x 10° & 49 X 107-7\12
= 3.7 X 10-*ohm, (11-121)
H\ 58 10%
as against 1207, or 377 in free space (Eq. 11-16).

Example JOULE LOSSES IN GOOD CONDUCTORS

Let us compare the energy lost by the wave and that gained by the
medium through Joule losses. We consider a thin sheet of conductor,
perpendicular to the direction of propagation, ab meters? in area,
and of thickness Az, as in Figure 11-8. If the amplitude of E on the
left-hand face is E) then, on the right-hand face, it is
E) e428

and §,, decreases from

LG oka LS A ar ee
“lean Fon) st a ts
within the sheet. Then the average power lost by the wave in the
sheet is

ab (_o \? roy -aAslt aF (se) 2yAZ (11-122

2 (5) a aS 2 \2wp fae 5° )

~) ;abo Az E% watts (11-123)

if Az is small compared to 6. We could also have arrived at this result

by calculating (dS,,/dz) Az.
Now the space average of the peak voltage across the sheet is

Figure 11-8. Element of volume aé

meter? in area and Az meters thick,
normal to the direction of propa-
al leAz gation in a good conductor.
11.6 PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 481

1
5 Fell + e42/*)\q ~ Ey (1— 5) aw Eva (11-124)

and the resistance in the direction of current flow is a/(cb Az). Then
the Joule losses must dissipate

12a/obHie Az 1 2 abo Az Ej watts. (11-125)

The factor of 4 is required because Eya is the peak voltage across

the sheet, and not the rms voltage.
The energy lost by the wave is therefore equal to the heat energy
generated in the medium, as we could have expected.

11.6 PROPAGATION OF PLANE ELECTROMAGNETIC

WAVES IN LOW-PRESSURE IONIZED GASES

We shall use the general results of Section 11.2, but first we must find the
conductivity o of an ionized gas.
We shall assume that the gas pressure is low, so that collisions and energy
losses can be neglected. Under those conditions the conduction process in
ionized gases is very different from that in metals, because conduction elec-
trons in metals suffer large numbers of collisions with the crystal lattice.
We also neglect thermal agitation. This is equivalent to setting the
temperature T equal to zero.
Let us consider a plane electromagnetic wave traveling in the positive
direction along the z-axis with the E and H vectors respectively parallel to the
x- and the y-axes. An ion of charge Q, mass m, velocity u, situated at x, y, z,
is subjected to a Lorentz force as in Sections 6.1 and 7.3:

f=Q2 {E+ XB), (11-126)

—= Yai dz
Gas —. Bat
IPCs dx
BRS
ORV) aeFay 12
(11-127)

Then
iuieg 4
8)
= i
dz
a nd
79
a 28
dt? m (E z ji) ele

E 1 dz
10 tie) ae
where we have used the free-space relationship E/B = c (Eq. 11-17). As we
shall see later on in Section 11.6.3, the ratio E/B is even larger in ionized
gases so that we are overestimating the magnetic force. Similarly,
482 PROPAGATION OF ELECTROMAGNETIC WAVES I

CV .
=p = 0, (11-130)

az E dx
Ae Oar (11-131)

We neglect thermal agitation because random velocities produce zero net

current.
Let us assume that the velocity of the ion along the z-axis is much smaller
than the velocity of light:
1 dz
é dt
cat (11-132)
We shall justify this assumption later on. Then, in the direction of the x-axis,
the magnetic force is negligible compared to the electric force and

Ghee
[ie Otaies m
<9}
E8 £08 wt. (11-133)

We must not use the exponential notation here, because we shall have to
solve Eq. 11-131, which involves the product of the two variables E and
dx/dt (see Appendix D).
Thus
x= ve E) cos wt. (11-134)
wm
We have neglected the constants of integration, which are related to the mean
position and velocity of the charge during one cycle. Also,

= Fe F sinut. (11-135)
Substituting in Eq. 11-131,

GF = SE in Dut (11-136)
Fe8 PB 5 2, (11-137)
noes eS sin wt. (11-138)
Before proceeding further, let us find the conditions under which the
inequality expressed in Eq. 11-132 is satisfied. We expect that the ion velocity
will remain low, either if the field is not intense, or if the frequency is so high
that the time available for acceleration is small. Let us calculate
eae 2 ke
c
(\dt ) aT
~ Ae22c?
(11-139)
11.6, PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 483

For an electron, Q = 1.60 X 10-’ coulomb, m = 9.11 X 10-3! kilogram,

and
yee 2.17 X 103.
= (Gen 7 FR. (11-140)
Let us assume that the average Poynting vector is calculated as in free space
(Eq. 11-26):
eee colon 2. (11-141)

We shall see later on in Section 11.6.3 that this is correct only at high fre-
quencies. Then
1 fdz
=Al = i)
:) S R630 Speedie
?? =f 4!
(11-142)

The velocity dz/dt is proportional to the average flux of energy Say, and
inversely proportional to the square of the frequency. This seems to indicate
that it is possible for (dz/dt)max to exceed the velocity of light c. This is
because we have neglected relativistic effects in our calculation. The ratio
(dz/dt)max/¢ is usually much smalller than unity in practice.

Examples At one kilometer from an antenna radiating 50 kilowatts of power

isotropically at a frequency of one megahertz the ratio is about 10-8,
Much higher power densities with S,, ~ 108 watts/meter? are
common in wave guides, but the frequencies are then of the order
of several gigahertz and (dz/df)max/c 107°.
Laser beams can carry power densities of the order of 101°
watts/meter? at f +10! hertz, and then the ratio is about 10-8.
Our assumption of Eq. 11-132 is therefore well justified for
electrons; it is even better justified for ions, since (dz/dt)/c is propor-
tional to 1/m?.
In the case of the antenna, the ratio Zmax/Xmax, Which is equal to
QE, /8wmce, is about 2 K 107°, Xmax is about 8 millimeters, Zmax about
1.6 * 107 meter, and dz/dt is only 2 meters/second. This velocity
corresponds to an energy of about 10-" electron-volt, while Ep is
1.7 volts/meter.
The ion motion is thus due almost exclusively to the electric
field, and the space charge density in a plane wave remains unaffected
by the ion motion, since ions of a given type move as a group ina
plane perpendicular to the direction of propagation.

11.6.1 The Conductivity of an Ionized Gas

Since the ion drift velocity can be ascribed almost exclusively to the electric
field intensity E, we can consider the ionized gas as having a conductivity o
such that
 PROPAGATION OF ELECTROMAGNETIC WAVES I
484

oh = Jy. (11-143)
Now

y= ONO. alle (11-144)

where N, is the number of ions or electrons of a given type per cubic meter,
Q; is the charge per ion or electron in coulombs, and (dx/dt); is the drift
velocity along the x-axis, which we have found above. The quantity J; is the
convection current density. Then
n2
o Ey exp jwt = > meet Ey expj (ot— 5) (11-145)

and
FS ; 2

g=—-t> (11-146)
w i
Since the masses ofthe ions are larger than the electron mass by several orders
of magnitude, whereas their charges are at most a few times larger than that
of the electron, we can retain only the term corresponding to the electrons and

g=—j NQ?_= — j447 X {0-9 Sie peter, (11-147)

WM, :*

where JN, is the number of free electrons per cubic meter.

Let us now try to understand the meaning of this imaginary conductivity.
Imagine a cube of some resistive material such as carbon, with copper
electrodes covering two opposite faces. If it has a volume of one cubic meter,
the resistance R’ between the electrodes is
ee R ee=p5=p=-
= en (11-148)
where © is the applied voltage, / is the current, p is the resistivity, / is the
distance between the electrodes (1 meter), A is the cross section (1 meter’),
and a is the conductivity of the medium. The quantity R’ is numerically equal
to p, but is expressed in ohms.
Thus the conductivity « of a medium is the conductance G’ = 1/R’
between two opposite faces of an imaginary cube measuring One meter on the
side. More generally, o is the admittance Y’ = 1/Z’ between the same elec-
trodes.
In the present case o is imaginary and negative, so that

l . OMe
Z! =" =i xpi nel (11-149)
11.6 PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 485

where

le GE (11-150)

is the equivalent inductance of an imaginary cubic meter of plasma. The

quantity L’ is expressed in henry-meters.
The electron current Jags the electric field intensity by 7/2 radians and
the electron current is inductive. Since E and J; are 7/2 radians out of phase,
the scalar product E- J; is also purely imaginary, and there is no energy loss
in the medium (Problem 11-22). This means that, on the average, the oscil-
lating electrons do not gain energy from the field once it is established. It will
be remembered that we assumed at the beginning that the electrons do not
lose energy by collision with the gas molecules.
Since the displacement current 0D/dt = jwesE leads the electric field
intensity E by 7/2 radians, whereas the electron current Jags by the same
angle, the displacement and electron currents are therefore m radians out
phase.

Example If there is an ionized gas between the plates of a parallel-plate capac-

itor, the total current density is
aD ay
Sarvs + Jy = jweE — j ane E.
.N.Q? (11-151)

The two components of J; are shown as functions of the frequency

Figure 11-9. The displacement current density 0D/dt and the convection
current density J; in an ionized gas are plotted here as functions of the
circular frequency w. The total current density is inductive below wp,
zero at w,, and capacitive above wp.
486

Figure 11-10. If there is an ionized gas

between the plates of a parallel-plate
capacitor of capacitance C, the impedance
is the same as that of a capacitance C in
parallel with an inductance (s/S)L’, where
L’ is given by Eq. 11-150.

in Figure 11-9. If S is the area of the plates and s their separation,

the total current is

I = jweS(0/s) cxR22
NeGs (uy)
is) (11-152)

1
= (jaeit.Jol’s
v, (11-153)
where V is the applied voltage. The capacitor therefore behaves as
if were connected in parallel with an inductance L’s/S as in Fig-
ure 11-10.
If an alternating voltage is applied to the circuit, the current
through C and that through sL’/S both remain finite but become
equal and opposite in sign at the resonant frequency.

11.6.2 The Plasma Angular Frequency o,

We have just calculated the total current density in an ionized gas. Let us
rewrite J, in a different form:

Jes (1-5N,2), (11-154)

= fwenlt (1= “2, (11-155)

where the second term between the parentheses is the ratio of the convection
to the displacement current density and where
2\ 1/2
Wp — (==) (11-156)
€—/Me

is called the plasma angular frequency. This quantity depends solely on the
properties of the gas considered. It corresponds to a frequency

Fo = 8.98N,"~ hertz. (11-157)

 487

y
' € (Zp)
Zo agi Pee — — =. ae i)
|

Figure 11-11. The vertical line on

aa
|
the left at t = 0 shows sche-
matically the initial uniform
ee ei distribution of the electrons
along the z-axis at positions
Se labeled z,. The vertical line
| on the right shows the new
|
| distribution for a positive dis-
——————
| placement ¢ increasing gradually
|
|
with z,. The new positions are
een! labeled z. We have assumed for
|
simplicity that ¢ increases uni-
1 formly with z; however, ¢ can
O i be any continuous function of z.

In the plasma of a gas discharge, N, is typically of the order of 10'8

electrons/meter*’, and f, is about 10‘ megahertz, whereas in the ionosphere
N. is of the order of 10" electrons/meter* andf, is about 3 megahertz.
Let us investigate further this frequency, which is characteristic of an
ionized gas. The phenomena involved are complex, but it will nevertheless
be instructive to use the following simple model. We consider a neutral ion-
ized gas. The ions being much heavier than the electrons, we shall investigate
group motions of the electrons in the absence of an electromagnetic wave,
assuming that the ions remain essentially fixed in position. We neglect thermal
agitation. Under these conditions, the charge density due to the ions is uni-
form throughout the volume considered, and the electron density NV, can vary
from point to point.
Let us assume that the electrons move in the direction of the z-axis
distances ¢(Zo) as in Figure 11-11, (zo) being a function only of the original
value Z) of the coordinate z. We make no assumption as to the function ¢(2o).
Then an electron originally at z) moves out to

Zo ae (Zo),

while an electron at Zz) + dz) moves out to

Zo + dz + (20 + dz),
or to
488 PROPAGATION OF ELECTROMAGNETIC WAVES I

Zp + dz + §(Z) + Edi

where the derivative d{/dz is evaluated at Zo.

In the process, the charge originally occupying an element of volume of
thickness dz) and area S comes to occupy a volume
d
(ae 4 a dz.)AY

and the electron density at z changes from N, to

NaS he (11-158)
d¢
Nene
dz

Assuming now that the displacement ¢ is everywhere small and that it varies
smoothly with z, we can set
d¢
eeeae (11-159)
and then
Nefie= N.( ay
=) (11-160)

Then the net charge density is that due to the ions, minus that due to the
electrons, or

> NG; a N.Q.( =) 2) = N.Q. 2 (11-161)

pfs =

where Q., the magnitude of the electronic charge, is positive, and where N;
is the number of ions, either positive or negative, of a given type. The net
charge density p; is positive if d¢/dz is positive, because the electrons are
then spread out.
From the Maxwell equation 10-105, this net charge density gives rise to
an electric field in the direction of the z-axis and
dE, N.Q.dt
reer (11-162)

Integrating and neglecting uniform fields,

E, = NQe gs (11-163)
€0

Then E, is proportional to £, and the equation of motion for an electron of

mass ™, iS
az. N,O;
iaierr We
-How = €. (11-164)
11.6 PROPAGATION OF PLANE ELECTROMAGNETIC WAVES 489

Now Zz = Z + §, where Zp is the initial position and is a constant, so that

ad Nod NO:peg
lear (11-165)
This is the equation for an undamped simple harmonic vibration of
angular frequency

ie: (22°) (11-166)

2\ 1/2

which is the value of w, given in Eq. 11-156. Each electron can therefore
execute a simple harmonic vibration about its initial position.
We have therefore calculated the plasma angular frequency w, in two
different ways. We first considered the motion of the electrons under the
action of an applied alternating electric field and found that for one particular
value of w, namely w,, the relation between the applied electric field E and the
resulting electron motion was such that the total current was equal to zero.
We then considered the motion of the electrons under the action of the elec-
tric field created by their own nonuniform displacement and found that they
could oscillate about their initial position with the same angular frequency w».
We can also calculate w, in the following manner. If we assume that the
electron density can oscillate at some angular frequency w,, the resulting elec-
tric field intensity E will oscillate at the same frequency. Then we can use
Eq. 11-147 for the conductivity o of the ionized gas. Now Eq. 10-15 for the
free charge density py assumes only the conservation of charge and Ohm’s law
J; = cE. Substituting the value of o and setting e, = 1,

p= psespys
NQ* i (11-167)
WyINe€o

The imaginary exponent shows that p oscillates, as expected, at an angular

frequency given by the coefficient of jt. This angular frequency must again
be w,, and
2\ 1/2
Wp = (22) (11-168)
e&0

as previously.

11.6.3 Wave Propagation at High Frequencies

Where o >
We use the wave number k for a medium of conductivity o as given by Eq.
11-69, set «, = 1, u, = 1, and substitute the value of o from Eq. 11-147. Then
490 PROPAGATION OF ELECTROMAGNETIC WAVES I

(1 xe\"

k=+ ie ; (11-169)

= rt
=)"
é (11-170)
For a wave propagating in the positive direction of the z-axis and with its E
vector parallel to the x-axis,
E = E,exp jot — kz)i (11-171)

where k is positive. From the general relationship for E/H, Eq. 11-53,
/2 2 Wh
(eee (2) i (“*) Wen at ie ete
wu wo) | ap
When w > w,, the wave number &k is real and the E and H vectors
are in phase, the ratio E/H being larger than in free space by a factor of
1/{1 — (wp/w)?} 1/2. As w approaches w», H tends to zero as expected, since
the total current density J; + (0D/dt) tends to zero. The Poynting vector
S also tends to zero.
The index of refraction is

n= {1a (2) " (11-173)

= {1— 80.5 ate , (11-174)

and the phase velocity is

ee ; 212
Hie (2yNhs (11-175)
mers
At high frequencies where w > w,, the phase velocity u is therefore greater
than the velocity of light, the wave number k is real, and there is no attenu-
ation. Figure 11-12 shows both n and 1/n as functions of w/w,. Since the phase
velocity increases with increasing ion density, waves tend to bend away from
regions of high ion density.
For high frequencies where w >> w;, the transmission is hardly affected
by the presence of ionized gas.
Note that the phase velocity is the velocity at which a given phase is
propagated. This is not the velocity at which a signal can be transmitted.
The reason for this is that a signal can be transmitted only if the wave is
modulated in some way, for example by varying its amplitude at an acoustic
 491

Figure 11-12. The index of re-

fraction n for an ionized gas and
its inverse 1/n as functions of the
«o/, ratio w/w».

frequency, as in amplitude-modulation radio broadcasting. Since any modu-

lation involves frequencies other than the carrier frequency, a signal neces-
sarily involves more than one frequency.
A single frequency corresponds to a pure sine wave extending from
t= —o to t= +o. Clearly, such a wave can transmit no information.
Since the phase velocity in an ionized gas is frequency dependent, the various
frequency components of a signal travel at different velocities. The result is
that a signal travels at a velocity that is different from those ofits component
waves and that is always less than the velocity of light in a vacuum.* It was
shown in Section 5.10 that this general law is required to preserve the principle
of causality.
Since the component waves travel at different velocities, the shape of the
signal, that is, the envelope of the wave, changes with time as the waves
progress through the dispersive medium.

11.6.4 Wave Propagation at Low Frequencies

Where o < a,

For w < w,, the wave number k of Eq. 11-170 is imaginary. The vectors E
and H are then out of phase by 7/2 radian.

* Léon Brillouin, Wave Propagation and Group Velocity (Academic Press, New York,
1960).
492 PROPAGATION OF ELECTROMAGNETIC WAVES I

The ratio |E/H|is smaller than in free space for w < (w,/21/”) and larger at
higher frequencies, from Eq. 11-172.
The average Poynting vector 8,, = (1/2)Re(E X H™*) is zero, and there
is no energy transmission.
The index of refraction and the phase velocity are also imaginary, and
E = Eyexp (jot — k’z), (11-176)
H = Hy exp (jot — k’2), (11-177)
where k’ = jk is a real number. There is no wave, since the phases of E and
of H are independent of z, and the amplitude decreases exponentially with z.
An electromagnetic wave is therefore either transmitted without attenu-
ation, or not transmitted at all through an ionized gas, depending on the ratio
w/w,. High frequencies are transmitted, whereas low frequencies are not. Our
theory, however, is much simplified. In particular, we have neglected energy
losses, which is justifiable only at low gas pressures.

Example Telemetry measurements during a Saturn I-b launching showed that

there was complete radio blackout at a certain moment at all fre-
quencies up to 10 gigahertz (10!° hertz). At that moment, the mini-
mum electron density in the plasma surrounding the rocket was such
that
N. N.
80.5 fz = tO) 1020 = le (11-178)

and the electron density was = 10!8 electrons/meter?.

Example THE IONOSPHERE

In the region of the upper atmosphere ranging in altitude from ap-

proximately 50 to 1000 kilometers the ionization is sufficient to inter-
fere with the propagation of electromagnetic waves. The ionization
is attributed mostly to the ultraviolet radiation of the Sun. On the
whole, the ion density first increases with altitude and then decreases,
but it shows “ledges” where the ion density varies more slowly with
altitude. The ledges are commonly called the D, E, F,, and F, layers.
These layers are due to the fact that both the nature of the solar
radiation and the composition of the atmosphere change with alti-
tude. Both the heights and the intensities of ionization ofthese layers
change with the hour of the day, with the season, with the sunspot
cycle, and so on.
It appears that the conductivity o is due almost exclusively to
the electron density, except possibly near the lower limit of the
ionosphere.
The free electron density is typically 10'!/meter*, varying from
11.7 SUMMARY 493

about 10!° to 10!?/meter® from the lowest to the highest layer. Over
this range of altitudes the number of molecules per cubic meter varies
from about 102 to 10!°. The percent ionization therefore increases
very rapidly with altitude, but it always remains low. For N, =
10!!/meter’, the plasma frequency f, ~ 3 megahertz.
Frequencies lower than f, are not transmitted. At radio fre-
quencies, waves are bent back toward the Earth, since the electron
density increases with increasing height, and since the phase velocity
increases with increasing electron density.
The assumption that there are no collisions between the elec-
trons and the gas atoms or molecules is not satisfactory in the
lowest regions of the ionosphere, where the pressure is highest, at
frequencies of the order of 1 megahertz or lower.
In the presence of the Earth’s magnetic field the ionized gas be-
comes doubly refracting, with the result that there are two distinct
phase velocities, depending on whether the E vector of the wave is
parallel or perpendicular to the B of the Earth.

11.7 SUMMARY

We began our discussion of electromagnetic waves by considering the sim-

plest case, namely that of plane waves propagating in a vacuum. Then e, = 1,
ur = 1, o = 0, p, = O and, from Maxwell’s equations,

WE + equwk = 0, (T-3)
V7 + euo’H = 0. (11-6)

These are the wave equations for E and H in free space. They show that the
phase velocity of such waves is
l
oS ae (11-7)
(opto)?
Plane electromagnetic waves in free space are transverse. Their E and H
vectors are orthogonal and oriented so that E X H points in the direction of
propagation. The magnitudes of E and H are related to give equal electric
and magnetic energy densities:

fis (“") = 377 ohms. (11-16)

1/2

H €0

The quantity
s=-EXH (11-23)
is called the Poynting vector. It is expressed in watts/meter’ and, when its
normal component is integrated over a closed surface, it gives the electro-
494 PROPAGATION OF ELECTROMAGNETIC WAVES I

magnetic power flowing out of the surface. In the case of a plane wave, §$ is
equal to the energy density multiplied by the phase velocity.
In any homogeneous, isotropic, linear, and stationary medium, a plane
electromagnetic wave has the following characteristics:

(1) pr = 0 (Eq. 11-47);

(2) E and H are transverse (Eqs. 11-46 and 11-49);
(3) E and H are mutually perpendicular;
(4) E/H = wu/k (Eq. 11-53);
(5) E X H points in the direction of propagation.
(6) E X H equals the energy density multiplied by the phase velocity.
In nonconductors, the phase velocity is
C= G
(€fy)? (11-57)
and, in nonmagnetic media (u, = 1), the index of refraction n is related to
the relative permittivity by the relation
n= el, (11-59)
The vectors E and H are in phase, and the electric and magnetic energy den-
sities are equal (Eq. 11-63).
In conducting media the ratio

(11-70)
is the magnitude of the ratio of the displacement current density to the con-
duction current density. The wave number k is complex (Eq. 11-75), and E
leads H (Eq. 11-78). The average value of the Poynting vector is then given by

Sav = 9Re(E X H*). (11-90)

Good conductors are defined as media for which

1
Ss 50° (11-102)
Then

k=
we
5? (11-104)

where

beth
= - (=) s)\ (11-105)

is the attenuation distance. The attenuation is so rapid that the wave is hardly
discernible (Figure 11-7). The E vector leads the H vector by 7/4 radian, and
PROBLEMS 495

E 2
a os (““) esn/4h (11-106)
o

Most of the energy is in the magnetic form.

In a low-pressure ionized gas the motion of the ions is not appreciably
affected by the magnetic field and the conductivity o, defined by

ees ok = Jj, (11-143)

1S Imaginary:

N02
Ces om (11-147)

where WN, is the number of electrons per cubic meter, Q, is the charge of the
electron, and m, is its mass. The total current density is less than in free space:
oD ‘ ON
Ji = op t= jon {1— (2) be, (11-155)
where
2\ 1/2
Wp = (72) (11-156)

is the plasma angular frequency.

At high frequencies where w? >> w%,, the wave is unaffected by the presence
of the ionized gas. As the frequency decreases, the attenuation remains equal
to zero, and the phase velocity increases to values greater than c until it be-
comes equal to infinity at w = w,. At angular frequencies smaller than w,, the
wave number is imaginary, the field is attenuated exponentially with z, and
there is no wave.

PROBLEMS

Note: An asterisk indicates that the problem requires a knowledge of relativity.

*11-1. Show that the wave equations for E and B is free space are invariant under a
Lorentz transformation, but not under a Galilean transformation.

11-2. A plane electromagnetic wave of circular frequency w propagates in free space

in the direction of the unit vector m.
Setting the wave number k = km, show that
k-E = 0, k X E — wll = 0,
k-H = 0, k X H+ wek = 0.
11-3. A 3.0-megahertz plane electromagnetic wave propagates in free space, and its
peak electric field intensity is 100 millivolts/meter.
496 PROPAGATION OF ELECTROMAGNETIC WAVES I

Calculate the peak voltage induced in a 1.00 meter?, 10-turn receiving

loop oriented so that its plane contains the normal to a wave front and forms
an angle of 30 degrees with the electric vector.
. (a) Calculate the electric field intensity due to the radiation at the surface of
the Sun from the following data: power radiated by the Sun, 3.8 < 107° watts;
radius of the Sun, 7.0 * 10% meters.
(b) What is the electric field intensity due to solar radiation at the surface of
the Earth? The average distance between the Sun and the Earth is 1.5 & 10!
meters.
(c) Show that the average solar energy incident on the Earth is 2 calo-
ries/centimeter?-minute.
. Acircularly polarized wave results from the superposition of two waves that
are (a) of the same frequency and amplitude, (b) plane-polarized in perpen-
dicular directions, and (c) 90° out of phase.
Show that the average value of the Poynting vector for such a wave is
the sum of the average values of the Poynting vectors for the two plane-
polarized waves.

11-6. (a) Show that, for an arbitrary electromagnetic field in a vacuum,

VEX E+ eB XB) = —S (EV x E+ ctB-Y x B),

where the dots indicate partial derivatives with respect to the time.
Note that this equation has the form of a conservation law and that the
spatial density of the conserved quantity is given by the parenthesis on the
right-hand side. This spatial density is expressed in volts?/meter®.
The flux of the conserved quantity is given by the parenthesis on the left.
(b) Show that this flux vanishes for a linearly polarized wave.
(c) Show that the flux is nonvanishing for a circularly polarized wave, that it
is proportional to the frequency, and that it is in the direction of propagation
if the EF and B vectors rotate in the positive direction, but contrary to the
direction of propagation if F and B rotate in the negative direction.t+

. Acharged particle travels at a velocity u in the field of a plane electromagnetic

wave in free space, the velocity of the particle being parallel to the direction of
propagation of the wave.
Show that the net transverse force is in the direction of E, but that it
tends to zero as the velocity of the particle approaches c.

. Several phenomena currently under investigation involve the motion of the

particles of a plasma in the presence of both an electromagnetic wave and a
constant magnetic field. Examples are the propagation of electromagnetic
waves through the ionosphere or through the Earth’s magnetosphere, and
the heating of plasmas.
In this problem we shall be concerned with the energy transfer from the
wave to a single particle of mass m and charge e.

} Daniel M. Lipkin, Jour. Math. Phys. 5, 696 (1964).

PROBLEMS 497

(a) The particle is illuminated with a plane-polarized wave in the presence

of a constant and uniform magnetic field ® parallel to the direction of
propagation.
Show that, if the angular frequency w of the wave is equal to the rest-
mass cyclotron frequency e@/m, and if we neglect both the change of mass
with energy and the magnetic field of the wave, then the kinetic energy aver-
aged over one cycle increases indefinitely as the square of the time f:

Dep = Cs te
2 8719

where £) is the maximum value of E. We assume that the particle is at rest

at ¢ = 0 and that wf? > 1.f
(b) We now consider the more complicated case where the wave is circularly
polarized and where the B of the wave is not negligible.
Use the following expressions for the E and B of the wave:
E = E,{sin (wt — kz)i — cos (wt — kz) jt,
B = Bo{cos (wt — kz)i + sin (wt — kz) j}.
We shall require these particular expressions under (e) below.
Show that
dé _ a:
ie ae
where & is the total energy of the particle, or moc* plus its kinetic energy.
This result is valid even at relativistic velocities. It shows that the wave
cannot increase the energy of the particle without also increasing the z
component of its momentum. This is because the wave has a transverse B.t{
*(c) Since the particle has a forward velocity, it feels a frequency which is
lower than if it were stationary.
Use the results of the example on page 212 and of Section 5.14 to show
that the condition for resonance is then

or

eB _ (ie a
m c

where the mass m is the mass measured in the laboratory system, and where
dz/dt is the longitudinal velocity of the particle.
*(d) You can now use the result you obtained under (b) above to show that,
if there is resonance at ¢ = 0, then resonance will be maintained indefinitely
afterward. The frequency of the wave, measured at the source, is assumed to
be constant.
This result is most interesting. As the velocity of the particle increases,
its mass increases and its cyclotron frequency e®/m decreases. However, the
axial velocity also increases, and the resulting Doppler effect gives a frequency
shift which exactly compensates for the increase in mass.
498 PROPAGATION OF ELECTROMAGNETIC WAVES I

(e) It is quite difficult to find the energy & as a function of z, but you can do
part of the calculation.
First, you can show—if you have not already done so under (b)—that

abs= ef{(Ey — v,Bo) sin (wt — kz) + v,B},

oy = e{—(E, — 2,Bo) cos (wt — kz) — 0,8},

Pe = eBo{v, sin (wt — kz) — v, cos (wt — kz)}.

C. S. Roberts and S. J. Buchsbaum have shown? that these three equa-

tions lead to the following equation for &:

(S) + V8) = 0,
where
wo Eo 36 — &
VG) a2 each
for a particle which is initially at rest (69 = mc”). Since & > &, V is always
negative, and & increases indefinitely.
It is useful to sketch a curve of V as a function of & to understand quali-
tatively the behavior of d&/dt.
(f) We have found d&/dt in (e) above, and dz/dt in (c); you can show,
finally, that
& — & 1 3 mez"
&o i 21/38 ce

The left-hand side is the kinetic energy (7 — mo)c? expressed as a multiple of

the rest energy nyc’.
The above theory, or rather a similar one that applies to standing waves,
has been checked (see second reference in footnote) with electrons at a
frequency of one gigahertz. The predicted electron kinetic energy, which was
5.80 10° electron-volts, agreed with the experimental value. The electron
energy was deduced from the X-ray spectrum and also from the radius of the
electron trajectory.

11-9. A plane elliptically polarized wave results from the superposition of two
plane-polarized waves whose E vectors are oriented in perpendicular direc-
tions and out of phase.
For example, E can have the two components

E, = E, exp j(wt — kz),

E, = Ey exp j(wt — kz + ¢),

where £,9 and Ey, are real.

+ C. S. Roberts and S. J. Buchsbaum, Phys. Rey. 135, A383 (1964). See also H. R. Jory
and A. W. Trivelpiece, Jour. Appl. Phys. 39, 3053 (1968).
PROBLEMS 499

Show that the E and H vectors for an elliptically polarized wave are
orthogonal if g = 0 or if the wave impedance E/H is real.
11-10. Show that S,, for a plane wave is equal to the average energy density
multiplied by the phase velocity, in any homogeneous, isotropic, linear, and
stationary medium.
11-11. In the general case of a uniform plane wave in a nonconductor,
E = (Eost + Eqyj + Evck) exp j(wt — krx — kyy — k.z),
where the coefficients Foz, Eo,, Eo:, kz, ky, kz can be complex, but are inde-
pendent of the coordinates x, y, z and of the time ¢.
The vector H is given by a similar expression with the same values of
Kaw Kayeukese
(a) Show that
B+ B+i=h
where k is the wave number corresponding to the medium and the frequency
under consideration.
(b) Discuss the relative orientations of E, H, and k.
11-12. Discuss the propagation of a plane electromagnetic wave in a medium
where (a) Q = 1, (b) Q? > 1.

11-13. The quantity

€— 1k
o
is called the complex permittivity of a medium. Justify the use of this term.

11-14. It was shown in the example on page 424 that the charge density p in a
conductor decreases exponentially with a relaxation time of €/c:
p = poexp (—at/e)
Show that © must be at most about 3 if p/po is to be less than 1% within
one fourth of a period.
11-15. Using the values of E and of H that we have found for plane electromagnetic
waves in good conductors, verify that all of Maxwell’s equations apply.
11-16. A plane electromagnetic wave propagates in a good conductor in the positive
direction of the z-axis.
Calculate the total power lost per square meter by Joule heating between
z = 0 and z— », and show that this is equal to S,, at z = 0.

11-17. (a) Show that the real part n, of the index of refraction n, the attenuation distance
6 and the Q ofa good conductor are related by the following equations:
03? = 2€ é
po®
Q _ &br
- 6? 2h?

Qn = €H,
Oho

(b) Show that the minimum value of n, is 5 for “good” conductors.

500 PROPAGATION OF ELECTROMAGNETIC WAVES I

11-18. Compare the ratios |E/H| for an electromagnetic wave in air and in sea
water at 20 kilohertz and at 20 megahertz.

11-19. Calculate the attenuation of an electromagnetic wave in sea water and

in copper at 20 kilohertz and at 20 megahertz. State your result in
decibels/meter. (See Problem 13-25 for a definition of the decibel.)

11-20. A conducting body can be heated by immersing it in an alternating magnetic

- field.
Consider the simple case of a plane conducting surface exposed to a
tangential magnetic field intensity Hp exp jwt.
Show that a plane electromagnetic wave penetrates perpendicularly into
the conductor and that the energy dissipated per square meter is
1 (op \!?
5($) i,

if the conductor is much thicker than the skin depth 6.

Note that this power dissipation is proportional to the square root of
the frequency and that it occurs mostly within one skin depth of the surface.

11-21. It is interesting to draw a parallel between the flow of heat in a thermally

conducting medium and the propagation of an electric or magnetic field in
an electrically conducting medium.
Letting ® be the heat flux density in watts/meter?,
® = —) VT,
where ) is the thermal conductivity in watts/meter-kelvin and T is the
temperature.
Then, for conservation of energy,
ca
V-P= ae ak ai QO}

where p is the density in kilograms/meter*, c is the specific heat in

joules/kilogram-kelvin, and Q is the heat produced within the medium in
watts/meter?.
FomnO = 0)
:2T ee
pc
SS
OT
Ms NX Or
This equation is identical in form to that for an electromagnetic wave in a
good conductor, pc/\ corresponding to wo. Its solution for heat flow in one
dimension is entirely similar to Eq. 11-107:

T = Ty exp Vv( - S 5 Al
where

mea wpe
1/2

Compare the velocities of propagation of T and of B in copper and in

iron.
 501

Property Copper Tron Units

o Sate) xe {OY 1.0 x 107 mbhos/meter

Cc 0.092 Ona) calories/gram-kelvin
p 8.9 7.9 grams/centimeter®
ON 1.0 0.15 calories/second-centimeter-kelvin
fr 1.0 200

11-22. Show that there is zero energy dissipation in a medium when E and J; are
out of phase by 7/2 radians.

11-23. (a) A mass m slides on a frictionless horizontal surface under the action of
a spring of stiffness (stretching force/elongation) & fixed to a rigid support,
and of a force F = Fy cos wt in line with the spring.
Show that this mechanical system is mathematically equivalent to a
series-resonant electric circuit.
Discuss the amplitudes of the displacement and of the velocity as func-
tions of w.
(b) A mass m slides on a horizontal frictionless surface under the action of a
force F = Fy cos wt, which is applied to it through a spring of stiffness k.
Show that this mechanical system is mathematically equivalent to a
parallel-resonant circuit.
Discuss the amplitudes and the velocities of the two ends of the spring
as functions of w.

11-24. Show that the conductivity of an ionized gas is

Eo
Jw

11-25. Two plane electromagnetic waves of equal amplitude propagate in the iono-
sphere where the electron density is N. electrons/meter*. One wave has a
circular frequency w; and a corresponding wavelength \,; the other has a
slightly different circular frequency w, and a wavelength d..
(a) At a given time ¢ there exist values of z for which the two waves are
in phase and values of z for which they are opposite in phase. What is the
distance between the maxima?
(b) What is their velocity?
This velocity is called the group velocity Uy.
(c) Show that, in the limit,
ee gq dk

dw»
(d) Show that
Uplt, = Cc’,

where u, is the phase velocity given in Eq. 11-175.

mega-
(e) Calculate the phase velocities and the group velocity for f; = 5.3
hertz, f2 = 5.4 megahertz , and NV, = 5 x 10” electrons/ meter’.
(f) Calculate the distance and the number of waves between two minima.
502 PROPAGATION OF ELECTROMAGNETIC WAVES I

11-26. Pulsars, or pulsating radio stars, emit sharp bursts of radio energy about 5
to 50 milliseconds wide at intervals of about one second. For any given pulsar
the repetition frequency is stable within about one part in 10°. The amplitude
and shape of the pulses vary widely, but each pulsar has its own characteristic
mean pulse profile.
Within a few months after the discovery of pulsars, distance estimates
were obtained in the following manner. It was observed that the arrival time
of a pulse depends on the frequency of observation, the arrival time being
later at lower frequencies. This delay is attributed to dispersion in the inter-
stellar medium which is ionized hydrogen with an electror. “ensity V, ~ 10°
per cubic meter.
(a) Show that, if w? > w,, a plot of the time delay Ar as a function of

LS peed) Tell
lO G Ay
is a straight line whose slope is a measure of the distance to the pulsar.
(b) In the case of pulsar CP 0328, arrival times measured at 151, 408, and 610
megahertz gave the results shown below.

i At
Megahertz Seconds
151
4.18
408
0.367
610
Show that, according to these measurements, the distance to CP 0328 is
268 parsec. (The parsec is 3.086 X 10!° meters. It is the distance from which
the radius of the Earth’s orbit, 1.495 < 10!! meters, would subtend an angle
of one second).
The fact that such plots give straight lines passing through the origin
indicates that the assumption w? > w? is correct. The delay therefore occurs
over large distances in a low-density plasma such as that of interstellar space,
and not inside the pulsar itself, which is quite small.
11-27. The solar wind is a high-conductivity plasma which is emitted radially from
the surface of the Sun. Let us calculate the flux of electromagnetic energy in
the solar wind at the orbit of the Earth.
In the plane of the Earth’s orbit, the magnetic field of the Sun is approxi-
mately radial, pointing outwards in certain regions and inwards in others.
This field is ‘‘frozen” in the high-conductivity plasma. Since the Sun rotates
(with a period of 27 days), and the plasma has a radial velocity, the lines of B
are in fact Archimedes spirals and, at the Earth, they form an angle of about
45° with the Sun-Earth direction. This is the so-called garden-hose effect.
At the orbit of the Earth the solar wind has a density of about 107 proton
masses/meter® and a velocity of about 4 X 10° meters/second, while the
magnetic field of the Sun is about 5 < 10-9 tesla.
PROBLEMS 503

(a) First show that, in an electrically neutral (p: = 0) and nonmagnetic fluid
of conductivity o and velocity v, Maxwell’s equations become

V-E=0, REO,
OB
Me ae VX B= wlolE +0 xB) +o Hh

the polarization currents being negligibly small compared to the conduction

currents.
In these equations, E, B, and v are measured with respect to a “‘fixed”’
frame of reference, and not with respect to the solar wind. Since p, is zero,
the current density J; is the same in both frames, except for a factor y, which
is very close to unity. See Section 5.21. Also, since the positive and negative
charges drift at the same velocity, J; is zero and
E = —v XB.
(b) Show that the component of v which is normal to B is
1
Un = pe BX (v X B)

(c) Show that the Poynting vector for the solar wind is
5
S= 2 Uv, & 6 microwatts/meter?.

This is about 4 < 107° times the average value of the Poynting vector of the
solar radiation, which is about 1.4 kilowatts/meter?.
The Poynting vector of the solar wind is normal to the local B and it
points at an angle 45° away from the Sun-Earth direction.
(d) You can show that the kinetic, magnetic, and electric energy densities are
related as follows:
Ux > Uu > Uz.

Hints

11-8. (a) Remember that

dp _ hae
in dv se
_ Oe PO)

and solve for V, and V,.

(b) Since a magnetic field can only deflect a particle without changing the magnitude
of its velocity, the gain in energy is due solely to the electric field and

ae ev: E.
dt
CHAPTER l2

PROPAGATION OF
ELECTROMAGNETIC WAVES II
Reflection and Refraction

In Chapter 11 we studied the propagation of electromagnetic waves in infinite,

continuous media. We shall now examine the effects of a discontinuity in the
medium of propagation, as in Figure 12-1.
We shall investigate again the same three types of media as in Chapter 11:
dielectrics, good conductors, and low-pressure ionized gases. It will be re-
called from the introduction to Chapter 11 tha
shall
limit ourselves to nonmagnetic dielectrics.
We assume an ideally thin, infinite, plane interface between two linear,
homogeneous, isotropic media. Then an incident wave along n; gives rise to
both a reflected wave along n, and a transmitted wave along n,;. The three
waves of course satisfy the conditions of continuity for the tangential compo-
nents of E and of H at the interface. For the time being, we exclude total
reflection; this will be discussed later, in Section 12.4. So as to avoid multiple
reflections, we assume that the media extend to infinity on both sides of the
interface.
There are many cases where a wave incident on a discontinuity is partly
reflected and partly transmitted. For example, a sound wave incident upon a
wall gives both a reflected wave, which comes back into the room, and a
transmitted wave, which proceeds into the wall. A similar phenomenon occurs
at a discontinuity in an electrical transmission line, at the junction between
two different types of coaxial line for example. Waves on strings show the
same type of behavior, as is shown in Section E-4 of Appendix E.
 505

Interface

Figure 12-1. An electromagnetic wave in medium 1 is incident on the interface

between media 1 and 2 and gives rise to both a reflected and a transmitted wave.
The vectors n;, n,, and n; are unit vectors normal to the respective wave fronts,
and point in the direction of propagation. The angles 0;, 0,, 6; are, respectively,
the angles of incidence, of reflection, and of refraction. (Radio engineers use f
different convention and call the complementary angles (7/2) — 0;, (7/2)—
and (17/2) — 6; the angles of incidence, reflection, and refraction, respectively)
Total reflection will be treated separately in Section 12.4.

Zeb AWS, OF (REBLECTION

AND SNELL’S LAW OF REFRACTION

If we assume that the electromagnetic wave incident on the interface as in

Figure 12-1 is both plane and plane-polarized, then its electric field intensity
E; is of the form
E; = Ey: exp jw; (:= nit), (12-1)

where 1 is the phase velocity of the wave in medium 1. (See Appendix E.)
The time ¢ = 0 and the origin r = 0 can be chosen arbitrarily. This equation
defines a plane wave for all values of ¢ and for all values of r, that is, a wave
that extends throughout all time and all space. However, it is used only in
medium 1.
Both the reflected and the refracted waves from a plane interface are then
also plane and plane-polarized, since the laws of reflection and of refraction
for any given incident ray must be the same at all points on the interface.
Then the reflected and the transmitted waves are of the form

E, = Ey, exp jo, (1= net), (12-2)

uy
506 PROPAGATION OF ELECTROMAGNETIC WAVES II

E, = Eo: exp Jw, (1= nt), (12-3)

UD)

where u is the phase velocity of the wave in medium 2. Note that we have
made no assumption whatever as to the amplitudes, phases, frequencies, or
directions of the reflected and transmitted waves. The amplitudes Eo, and Eo,
can be complex to take phase differences into account.
‘We shall choose the origin at some convenient poi

To obtain continuity of the tangential components of E and of H at the

interface, some valid relation must exist between E;, E,, E, for all time ¢ and
for all points on the interface. Such a relation will be possible if (a) all three
vectors E;, E,, E; are identical functions of the time f, (b) all three vectors are
identical functions of position r; on the interface, and (c) there exist certain
relations between Eo;, Eo,, Eo:.
From condition (a),
iy =O = Oh. (12-4)
All three waves must therefore be of the same frequency. This is intuitively
quite obvious, since they are all superpositions of the wave emitted by the
source and of those waves emitted by the electrons executing forced vibrations
in media | and 2. It will be recalled from mechanics that forced vibrations
have the same frequency as the applied force.
From condition (b), we must also have, at any point ry on the interface,
Tey Tr My Tz
(12-5)
ae aee
a ts mee (12-6)
Uy

Then from the first of these equations,

Gu: — n,):rr = 0. (12-7)
Since the vector ry; lies in the interface, the vector n; — n, must be normal to
the interface and the tangential components of these two vectors must be
equal and of the same sign as in Figure 12-2. Then
Bp. (12-8)
 507

Figure 12-2. The vectors n;, n,,

Mt, ml.

Considering now Eq. 12-6,

(= = =) -rp = 0. (12-9)
Uy U2

Hence the vector between parentheses must also be normal to the interface,
so that n;, m,, m are coplanar and all four vectors n;, n,, n,, nm are in the
plane of incidence. Moreover, the tangential components of n;/im and of n;/u
must be equal, and
sin@; _ sin 6,
7A a (12-10)

or, since the wave number k is w/u,

k, sin 0; = ky sin 4. (12-11)

The quantity k sin 6 is therefore conserved in crossing the interface. We can

also write that

since k = n/Xo, where n is the index of refraction. This 1

It is important to note that this law, as well as the laws of reflection, are
general. They apply to any two media; they even hold true for total reflection,
as will be shown later on.
508

nterface

wy
%%
%,.
v
© 3.
9,a)

Figure 12-3. Coordinate system

used for the study of reflection
ly and refraction.

Thus, choosing the origin in the interface and axes as in Figure 12-3,

E; = Ey; exp j{wt — k,(x sin 6; — z cos 6;)}, (12-14)

E, = Eo, exp j{wt — k(x sin 6; + z cos 6;)}, (12-15)
E, = Ey, exp j{wt — k(x sin @, — z cos 4;)}. (12-16)

Zar RESNEE S*EQUATIONS

We shall turn now to the third condition mentioned in the preceding section.
We must find the relations between the quantities Eo:, Eo, Eo. that will ensure
continuity of the tangential components of FE and of H at the interface.
We recall from Section 11-2 that the E and H vectors in a plane electro-
magnetic wave are always perpendicular to the direction of propagation and
to each other. The E vector of the incident wave can thus be oriented in any
direction perpendicular to the vector nj.
It will be convenient to divide the discussion into two parts. We shall
consider successively incident waves polarized with their EF vectors normal,
and then parallel to the plane of incidence. Any incident wave can be sepa-
rated into two such components.

12.2.1 Incident Wave Polarized with Its E Vector

Normal to the Plane of Incidence

The E and the H vectors of the incident wave are oriented as in Figure 12-4.
If the media are isotropic, as we assumed at the beginning of this chapter, the
 509

Figure 12-4. The incident, re-

flected, and transmitted waves
when the incident wave is po-
larized with its E vector normal
to the plane of incidence. The
arrows indicate the directions in
which the Eo;, Ho;, Eo,, etc.,
vectors are taken to be positive
at the interface. The vectors
E X H point everywhere in the
direction of propagation.

E vectors of both the reflected and the transmitted waves will also be perpen-
dicular to the plane of incidence, as in the figure.
Considering the electric and magnetic field intensities of the incident wave
to be known, we have four unknowns: Eo,, Eo:, Ho, Ho. We also have four
equations that are provided (a) by the continuity of the tangential components
of E and H at the interface and (b) by the relations between E and H for plane
waves in medium | and in medium 2 as given in Eq. 11-53. It will therefore
suffice to

The continuity of the tangential component of E at the interface requires

that
Foi + Eor = For (12-17)

at any given time and at any given point on the interface. Similarly, the
continuity of the tangential component of H requires that
Hy; cos 6; — Ho, cos 0; = Hy, cos 6; (12-18)
or, from Eq. 11-53,

ee (Eo; — Eor) cos 0; = Bl Et COS 4. (12-19)

WUy WH

Thus, recalling that k = n/Xo, where n is the index of refraction and Xp is the
free-space wavelength divided by 2r,
510 ROMAGNETIC WAVES II

ue cos 6; — we COs 4;
Part gene ee Rd al O88 (12-20)
n Nye
=~ 608.0; -—— cos 6;
Mri Lr2

2 ak cos 6;
= ——_*t _,
hy 1/0)
(12-21)
— cos 6; + — cos @,;

where the subscript N indicates o: 1s normal to the plane of incidence.

These are two of Fresnel’s equations; the other pair will be deduced in
the next section. Fresnel’s equations give the ratios of the amplitudes of the
incident, reflected, and transmitted waves. They apply to any two media. We
shall find later on that they are valid even for total reflection.

12.2.2 Incident Wave Polarized with Its E Vector

Parallel to the Plane of Incidence

In this case the E vectors of all three waves must be in the plane of incidence
as in Figure 12-5. We have chosen the orientations of Ey, and of Eo; so that
Figures 12-4 and 12-5 become identical at normal incidence, except for a
rotation of 90° around the normal to the interface.
We now have
Hoi — Hor = Hor (12-22)
or

ap — Eq) = Ay Eot, (12-23)

My Whe
and
(Eoi + Eor) cos 6; = Ep: Cos 6. (12-24)

Then, solving for Eo, and

Myo Ay
— — cos6; + — cos8;
E T2é
a
= — cos 6; + —' cos 6;
Mro Mrt

Ny
2 — cos 6;
(=) =—_ _. (12-26)
ae ® Cos 6, + we COS6;

Equations 12-20, 12-21, 12-25, and 12-26 are Fresnel’s equations.

 511

nterface

mel

ge Waki h y Figure 12-5. The incident, re-

; flected, and transmitted waves
:/ when the incident wave is po-
| larized with its E vector parallel
Be a) to the plane of incidence. The
arrows for E and for H have the
iy same meaning as in Figure 12-4.

At normal incidence 6; = 6, = 0, the plane of incidence becomes undeter-

mined, and the pair 12-20 and 12-21 is identical to the pair 12-25 and 12-26.

12.3 REFLECTION AND REFRACTION

AT THE INTERFACE BETWEEN
TWO NONMAGNETIC NONCONDUCTORS
Now that we have established the laws of reflection, Snell’s law of refraction,
and Fresnel’s equations, all of which apply to the interface between any two
media, we shall consider the relatively simple case of the interface between
two nonmagnetic nonconductors. The indexes of refraction nm and m are real
numbers equal to or larger than unity, and
sin6, ™m
EL ese (12-27)

as in Eq. 12-1

From Fresnel’s equations,

(“) cos 6; — cos 6;

(2) oe ae (12-28)
Eoi) (“) cos 6; + cos 0;
2

(#). S oe (12-29)
Eo; =)cos 6; + cos
2

for a wave polarized with its E vector normal to the plane of incidence.
512

Interface Interface
0) ]

Ce
xe
aeBe
\by ie
\
\

n,

ly (a)

Air
Glass

Glass

if | , Air

(b)
 513

Figure 12-6. (a) The relative phases, at the interface, of the electric field intensities
in the reflected and transmitted waves for n. > n, and for n. < m with Ey;
normal to the plane of incidence. In the first case the reflected wave is 7 radians
out of phase with the incident wave. The transmitted wave is in phase in both
cases. (b) “‘Crests”’ of E corresponding to (a) at some particular time. They are
spaced one wavelength apart and travel in the directions shown. Note the phase
shift of 7 on reflection from a glass surface. Note also the interference pattern
resulting from the superposition of the incident and reflected waves. Constructive
interference occurs wherever two crests or two troughs meet; destructive inter-
ference occurs where a crest meets a trough. (c) Graphs of E;, E,, E, at normal
incidence on an air-glass interface. Where the wave emerges from glass into air,
as on the right, EZ; is larger than E;. Conservation of energy still applies, however.
See Section 12.3.2.

It will be observed that (E,/Ep:)w is always real and positive. This means
that at the interface the transmitted wave is always in phase with the incident
wave. The ratio (£p,/Eo:)w can, however, be either positive or negative, depend-
ing on the value of m/m, for if m/no > 1, then 6, > 6; and cos 6; > cos 4;
whereas if n/n. < 1, then 6, < 6; and cos 6; < cos 6. The reflected wave is
thus either in phase with the incident wave at the interface if m, > ny or is
radians out of phase if m < ny. Figure 12-6 illustrates the E vectors for both
types of reflection; Figure 12-7 shows the ratios of Eqs. 12-28 and 12-29 for
n/n, = 1/1.5. This corresponds, for example, to a light wave incident in air
on a glass with an index of refraction of 1.5.
For an incident wave polarized with its E vector parallel to the plane of
incidence,
514

0.0 ea ae a | SS (I Jere Lent

imal le 10 20: 30° 40 +50 60) 70) Vs0
7]

= 110

Figure 12-7. The ratios (E,,/E,;)y and (E,,/E:)y as functions

of the angle of incidence 6; for n,/n, = 1/1.5. This corre-
sponds to light incident in air on a glass with m = 1.5. The
wave is polarized with its E vector normal to the plane of
incidence.

ny
—cos 6; + (2) cos 6;
(=) ee eee eee (12-30)
a cos 6; + (“) cos 6;

ny
2{—)cos@;
=) (“:) .
i
—)
" = ——o“—-—:
cos 6; + (“')COS 6;
Seal
12-31

Ng

The second ratio is always positive. This indicates that the relative phases
of Eo, and Eo; are as in Figure 12-5, which we used in arriving at this result;
that is, the incident and transmitted electric field intensities are in phase at
the interface.
On the other hand, the ratio for Ey, can be either positive or negative,
which indicates that Ey, can point either as in Figure 12-5 or in the opposite
direction. The tangential components of Eo; and of Eo, can thus be either in
phase or w radians out of phase. The Ey, component is in phase with Ep; at
the interface if

(*) cos 6; — cos 6; > 0 (12-32)

2
 515

Eo, é
—0.4 Eo; ; z||
Figure 12-8. The ratios (E,,/E,;)p and (E,,/E,;)p as functions
of the angle of incidence 0; for n,/n, = 1/1.5. This corre-
sponds to light incident in air on a glass with n = 1.5. The
wave is polarized with its E vector parallel to the plane of
incidence.

or if
sin 6, cos @; — sin 6; cos 6; > O,. (12-33)
sin 20, — sin 26; > 0, (12-34)
sin (0; — 6;) cos (@, + 6;) > 0. (12-35)
This inequality will be satisfied either if
6, > 0; and b,x 5 (12-36)
or if
Oe < 0; and oe p> * (12-37)
The phase of the reflected wave in this case does not therefore depend only on
the ratio n2/n,; it depends on both 6; and 6;. The ratio Ey,/E), can be either
positive or negative, both for np > m, and for np. < nm. Figure 12-8 shows the
ratios of Eqs. 12-30 and 12-31, again for m/n, = 1/1.5.

12.3.1 The Brewster Angle

We have seen in the foregoing that the electric field intensity Eo, of the reflected
wave is either in phase or z radians out of phase with the incident wave,
depending on whether sin (@, — 6;) cos (0; + 6;) is greater or less than zero.
It will be gathered from this that there is no reflected wave when this ex-
pression is equal to zero, that is, when 6; = 6, = O or when 6; + ® = 7/2.
The first condition is incorrect, however; it arises from the fact that we have
multiplied the inequality
516

Z nterface nterface

ly ly

Figure 12-9. When the incident wave is polarized with its E vector parallel to the
plane of incidence, there is no reflected wave at 6; + 6, = 1/2. The angle of
incidence 6; is then called the Brewster angle. The position of the missing reflected
ray is at 90° to the transmitted ray. For any pair of media, the sum of the two
angles 6;2 is 90°.

(2) cos 6; — cos 6; > 0 (12-38)

7

by sin 6;, which is equal to zero at 6; = 0.

Thus, for

6;+6, = a (12-39)

there is a reflected wave only when the incident wave is polarized with its E
vector normal to the plane of incidence. This is rather remarkable because
it involves the passage of a wave through a discontinuity in the medium of
propagation without the production of a reflected wave. The conditions of
continuity at the interface are then satisfied by two waves only—the incident
and the transmitted waves—instead of the usual three. This is illustrated in
Figure 12-9. This angle of incidence is called the Brewster angle. It is also
called the polarizing angle, since an unpolarized wave incident on an interface
at this angle is reflected as a polarized wave with its E vector normal to the
plane of incidence.*

* The Brewster angle is often explained incorrectly as follows. For this particular angle
of incidence, the missing reflected ray is at 90° to the transmitted ray. It is argued that the
electrons excited in medium 2 do not radiate in their direction of oscillation (Section
14.1.4) and hence cannot give rise to a reflected ray in medium 1 in this case. This explana-
tion is incorrect, since the Brewster angle is observed even when medium 2 is a vacuum.
12.3. THE INTERFACE BETWEEN TWO NONCONDUCTORS 517

At the Brewster angle,

4 Tv
: Sint =e—" 0:
ny sin 0, ( w) e
—_—-
NM,
= r5
SIN Oyz
= eee
Sin biz
- SC= co “PD,
Le (12-40)
Example MEASURING THE RELATIVE PERMITTIVITY OF
THE MOON’S SURFACE AT RADIO FREQUENCIES
The nature of the Moon’s surface can be inferred, to some extent,
from the value of its relative permittivity ¢, = n? (Section 11.3) at
radio frequencies. The Brewster angle has been used to measure this
quantity in the following way. When the Moon is illuminated with
a radio wave originating from a satellite in lunar orbit, the reflection
observed at the Earth can be compared to the reflection of sunlight
from the surface of a lake: most of the light comes from the regions
that happen to be correctly oriented for specular reflection. The sur-
face of the Moon thus glistens over an area of the order of 100
kilometers in diameter, the area depending on the height of the
satellite above the surface of the Moon, and also on the roughness
of the surface. (If the detector on the Earth receives both the
reflected wave and a direct wave from the satellite, it is possible to
discriminate between the two by making use of the fact that the
Doppler effect makes the two frequencies slightly different.) A plot
of the intensity of the reflected wave as a function of the angle of
incidence when the E vector lies in the plane of incidence shows zero
reflection at the Brewster angle.
In one such measurement, performed at a frequency of 0.14
gigahertz, the Brewster angle was found to be 60 + 1° in the mare
northwest of Hanstein. This gives an e¢, of 3.0 + 0.2.
It is possible to perform similar measurements at other points
on the surface of the Moon because of the relative motions of the
three bodies involved, namely the satellite, the Moon, and the Earth.

12.3.2 The Coefficients of Reflection and of Transmission

at an Interface Between Two Nonconductors

It is useful to define coefficients of reflection and of transmission that are re-

lated to the flow of energy across the interface. The average energy flux per
unit area in the incident wave is given by the average value of the Poynting
vector, Eq. 11-64. Setting u, = 1,
he fc nee
8; ave (2) Eo ni, (12-41)
2 \Ho

S.= 5 (2) Ein, (12-42)

/2

2 \uo
1 fe\” 2
Stay =5 (— Eo. m (12-43)
2 \Ho
518

“O) {Oo “26P sor 40r 50 GO Jonkos? S02

6
Figure 12-10. The coefficient of reflection Ry and the
coefficient of transmission Ty as functions of the angle of
incidence 6; for m/n. = 1/1.5. The E vector of the incident
wave is normal to the plane of incidence.

The coefficients of reflection R and of transmission T are defined as the

ratios of the average energy fluxes per unit time and per unit area at the inter-
face:
S, av° TL Ei,
R = (|) ’ -4
Savi E32; (12 4)

where ni is the unit vector normal to the interface;

1
8; av ba 7
ee
€r2 2 EX COS 6,
; 3
Siay'n - (©) E2, cos 6; (12-45)
__ Ny Ep: Cos 6; 3
~ ny, E2, cos 0; (12-46)

Then, from Fresnel’s equations for nonconductors,

(“) cos 6; — cos A, l

Ry = 2
(12-47)
(“) cos 6; + cos 6 ;

4 (2)cos 6; cos 0;
Ty = 4,
— (12-48)
Ny

{() cos 6; + cos a

Rp = i cos 6; + (2
*)e080)

|cos6; + (“) cos 0; | Vag

 519

EO PVi0® 20iers0 40 HTS0RGOs P70 LES 90°

Figure 12-11. The coefficient of reflection Rp and the

coefficient of transmission 7p as functions of the angle
incidence 6; for m/n, = 1/1.5. The E vector of the incident
wave is parallel to the plane of incidence.

4 (“) cos 0; Cos 6;

Tp = 2
SEER TEC EES (12-50)
{cos6,;+ (“) cos a}
Ny

In both cases, R+ T = 1, as expected, since there must be conservation of

energy. At the Brewster angle, defined by Eq. 12-40, Rp = 0 and Tp = 1,
again as expected.
Figures 12-10 and 12-11 show the coefficients of reflection R and of
transmission T as functions of the angle of incidence 6; for m/n2 = 1/1.5,
whereas Figure 12-12 shows R and T as functions of m/n: at normal incidence.

(Roc aifesile veils STE ea fer ee eo]

0.1 Or 0.4 0.6 1.0 2.0 40 6.0 10.0
n,/Nz

Figure 12-12. The coefficient of reflection R and the coeffi-

cient of transmission 7 at normal incidence as functions of
the ratio m/no.
520 PROPAGATION OF ELECTROMAGNETIC WAVES II

12.4 TOTAL REFLECTION AT AN INTERFACE

BETWEEN TWO NONMAGNETIC
NONCONDUCTORS
We shall now consider the phenomenon of total reflection, which we have
excluded in Sections 12.3 to 12.3.2.
If m, > m, and if 6; is sufficiently large, Snell’s law

sin 6, = sin 6; (12-51)

2

leads to the apparently absurd result that sin @, is greater than unity. The
critical angle of incidence, for which sin @, = 1 and @, = 90°, is

sin 0. = 2. (12-52)
ny
It is observed experimentally that, when 6; > 0,., the wave originating in
medium | and incident on the interface is totally reflected back into medium 1
as in Figure 12-13. This phenomenon, which is called total reflection, does not
depend on the orientation of the E vector in the incident wave. For light
propagating in glass with an index of refraction of 1.6, the critical angle of
incidence is 38.7°.
It turns out, as we shall see in Section 12.4.1, that Snell’s law (Eq. 12-13),
the laws of reflection and refraction (Section 12.1), and Fresnel’s equations
(12-20, 12-21, and 12-25, 12-26) are all applicable to total reflection if we
disregard the fact that sin @, > 1 and if we set
cos @,= — (1 — sin? 6)”, (12-53)
s {1 (“) sin? 6.) (12-54)
oe a ny 2 oe iby ;

=Sei
—2cs{sin®o
BP me (*)
6.) pl ae (12-55)

Figure 12-13. For angles of

incidence @; equal to or greater
than the critical angle 6,,, the
wave is totally reflected back into
medium 1.
12.4 TOTAL REFLECTION 521

Then, from Eqs. 12-15 and 12-16,

E, = Ep, exp j{wt — ky(x sin 6; + z cos @;)}, (12-56)
E, = Eq. exp j{wt — k(x sin 6, — zcos 6,)\, (12-57)

= Ey; exp {i(wt — 3 x sin 6;) + - ((ou/my sin’é; — Linz. (12-58)

0

INCIDENT WAVE POLARIZED WITH ITS E VECTOR NORMAL TO THE PLANE OF

INCIDENCE. From Fresnel’s equations 12-20 and 12-21, and for nonmagnetic
nonconductors,
2) 1/2
{sin 6; — (“*)}
(7) = exp 2j arctan A Soup ja, (12-59)
Ey

2 COS 6;
Eo
pace, (12-60)
(=),: Pea cos 6; — j {int6; — (2) }
ny

=i 2 cos 0; =a expjare tan(5 ) (12-61)

ol
We first notice that the amplitude of the reflected wave is equal to that
of the incident wave, so that the coefficient of reflection R is equal to unity.
The energy is totally reflected and there is zero net flux of energy through the
interface.
The phase jump upon reflection varies from 0° at the critical angle of
incidence (sin 6; = m2/m) to 180° at glancing incidence, through positive
angles. This is shown in Figure 12-14.

Figure 12-14. The phase a of the reflected wave with respect

to that of the incident wave in the case of total reflection.
The incident wave is polarized with its E vector normal to
the plane of incidence; the ratio m/nz is equal to 1.50.
522

I Oe

=
l. tn 2.0

Figure 12-15. The ratio (Eo:/Eo:)v of Eq. 12-60 is plotted

here in the complex plane for various angles of incidence 6;
larger than the critical angle and for m/n. = 1.50. The
amplitude of the transmitted wave is represented by the
distance from a point on the curve to the origin, and is
greatest at the critical angle. The transmitted wave leads the
incident wave by an angle equal to the argument of the
complex ratio—for example, the angle ¢ shown.

As regards the transmitted wave, it is obvious that Ep; is not zero, despite
the fact that the net flux of energy across the interface is zero. Medium 2 can
be considered to act like a pure inductance fed by a source of alternating
current. The average power input to the inductance is zero, the power flow
being alternately one way and then the other, but there is nevertheless a
current through the inductance.
Figure 12-15 shows how the electric field intensity of the transmitted
wave varies both in amplitude and in phase with the angle of incidence.
The transmitted wave is quite remarkable. According to Eq. 12-58,
it travels unattenuated (as in Figure 12-16) parallel to the interface, with a
wavelength
Mo | AM
msin6; sind;
(12-62)
A; being the wavelength in medium | above the interface. The wavelength \,
is exactly the distance along the x-axis between two neighboring equiphase
points in the incident wave. This was to be expected, since the continuity
conditions must be satisfied at all points on the interface.
This result, namely that the wave travels unattenuated parallel to the
interface, is most surprising if we think of an incident wave of finite cross-
 523

j : >
t: H§ { : Air

Figure 12-16. “‘Crests’’ of E for the incident, reflected, and transmitted waves are
represented here schematically for the case of total reflection. They are spaced one
wavelength apart. The transmitted wave travels unattenuated below the interface,
and its amplitude decreases exponentially with depth in medium 2. The data
used for the figure are the following: nm, = 3.0, m. = 1.0, 6; = 75°. (See Problem
12-12.)

section. Does the transmitted wave on the other side of the interface extend
beyond the illuminated region? Our discussion cannot provide us with an
answer, since it is based on the assumption that the incident wave is infinite
in extent. Physically, what happens is this: a given incident ray, instead of
being reflected abruptly at the interface, penetrates into medium 2, where it
is bent back into medium 1. It is this phenomenon which gives rise to the
“transmitted” wave.*
The transmitted wave is damped exponentially in the direction perpen-
dicular to the interface in such a way that its amplitude decreases by a factor
of e over a distance
Ao
= 2 1/2" (12-63)
{(2) sin’é; — 1}
2

This is illustrated qualitatively in Figure 12-16.

The ratio 6,/X, is shown in Figure 12-17 as a function of the angle of
incidence 6; for m/n, = 1.5.

* See A. von Hippel, Dielectrics and Waves (Wiley, New York, 1954), p. 54, for a brief
account of work by F. Goos and H. Hanchen on this subject.
524

lon T
Nn
— =

4 | i (eee i | 1
40 50 60 70 80 90°
i

Figure 12-17. The ratio of 6, (the depth of penetration) to A»

(the wavelength in medium 2 divided by 27) for the trans-
mitted wave when there is total reflection. The index of
refraction of the first medium is 1.50 times that of the sec-
ond medium.

We Shall find in Section 12.4.1.4 that there is an energy flow parallel to

the interface in the positive direction of the x-axis. The flux is a function of
the angle of incidence as in Figure 12-18, and it decreases exponentially with
z. There is zero average energy flux in the direction normal to the interface.
The transmitted wave, has been observed both with visible light and with
microwaves.*

INCIDENT WAVE POLARIZED WITH ITS E VECTOR PARALLEL TO THE PLANE OF

INCIDENCE. Equations 12-56 and 12-57 for E, and E, are again valid, the

0.85 4
0.7
0.6
S05
q
@ 0.4

Figure 12-18. The flux of energy S,, parallel to and just

below the interface, is proportional to cos @; sin 20; in the
case of total reflection.

*See, for example, J. Strong, Concepts of Classical Optics (W. H. Freeman and Com-
pany, San Francisco, 1958), Appendix J by G. F. Hull, Jr., p. 516.
12.4 TOTAL REFLECTION AT AN INTERFACE 525

coefficient of reflection R is again equal to unity, but the phase jump upon
reflection is different. Total reflection of a wave polarized in some arbitrary
direction therefore gives an elliptically polarized reflected wave.

Example LIGHT EMISSION FROM A CATHODE RAY TUBE

In a cathode ray tube, light is generated in a fluorescent screen, which
is in contact with the tube face. The light is of course emitted in all
directions but, as a rule, the part that is emitted backwards in the
direction of the electron gun is reflected by a thin aluminum coating
laid over the fluorescent screen as in Figure 12-19. This increases the
light output by a factor of two.
The figure shows that most of the light emitted by the screen is
trapped inside the tube by total reflection. We can calculate the
fraction of the light that comes out through the tube face as follows.
Let us assume that, inside the cone 26;,, all the light passes
through the interface. Then the fraction F is the solid angle corre-
sponding to the critical angle 0;,., divided by 27. We divide by 27
and not by 4m because, with the reflecting coating, all the light is
emitted within a solid angle of 27. Now the solid angle correspond-
ing to 6;, is the area of the surface represented by a dashed line in
the figure, divided by R®. Thus

F
1 / 2nR sin RAD _ y _ cos 6,,).
Bic 7
(12-64)
~ Oy 0 TR

For a glass with an index of refraction of 1.6, F is only 22%,

with our assumption that the coefficient of transmission is equal to
unity for @ < 6,,. The fraction F is, in fact, even smaller than 22%,
but not very much so because the coefficient of transmission is
approximately equal to unity, except near the critical angle.

pee

Fluorescent Glass tube

screen face

Reflecting
coating ~7}

Electron
beam

Figure 12-19. Section of the face of a

cathode ray tube.
526 PROPAGATION OF ELECTROMAGNETIC WAVES II

Example THE CRITICAL ANGLE AND BREWSTER’S ANGLE

The critical angle arc sin (m2/m) (Eq. 12-52) is somewhat larger than
the Brewster angle arc tan (m:/m,) (Eg. 12-40). For example, again
for light propagating inside a glass with an index of refraction of
1.6 (n,), the wave is totally transmitted into the air at a glass-air
interface when the angle of incidence is the Brewster angle, 32.0°,
and is totally reflected into the glass at the critical angle of 38.7°.
Figure 12-20 shows these two angles as functions of the ratio
n/n. For large values of ;/n:, that is, for light incident in a rela-
tively ‘‘dense’’ medium, 6;, is nearly equal to 6;z. For media with
more similar indexes of refraction, the Brewster angle approaches
45°, whereas the critical angle approaches 90°.
It is interesting to note that, for a wave polarized with its E
vector parallel to the plane of incidence, the amplitude of the
reflected wave changes rapidly when the angle of incidence lies
between the Brewster angle and the critical angle.
This peculiar behavior of the reflected wave could be used for
measuring small angular displacements.

20 ie =

Total reflection

Critical angle

Fa
(SE
ETS
Ve
LS
Ta
LESS
Fag
!
Brewster angle

2 3 4 SO8: SAU
n)/ny

Figure 12-20. The critical angle and the Brewster angle as functions of
the ratio m/n, of the indexes of refraction on either side of the interface.
The wave is incident in medium 1. It is polarized with its E vector
parallel to the plane of incidence for the Brewster angle curve.
 527

12.4.1 Demonstration of the Validity of Snell’s Law,

of the Laws of Reflection and Refraction,
and of Fresnel’s Equations in the Case
of Total Reflection*
We shall study only the case where the incident wave is polarized with its E
vector normal to the plane of incidence.
It turns out that it is impossible to satisfy the requirement of continuity
of the tangential components of E and H, and of the normal components of
D and B at the interface, with only the incident and reflected waves. We
therefore conclude that there must exist some sort of transmitted wave.
The transmitted wave must, however, be of a rather special nature, since
it is not observable under ordinary conditions. It must of course satisfy the
general wave equation for nonmagnetic nonconductors (Eq. 11-54),
2 2 2
“ = = €2Lo oe (12-65)

We have set the derivative with respect to y equal to zero because, by hypoth-
esis, the field does not vary with the y-coordinate (Figure 12-3).
We can use Eq. 12-14 for the incident wave:
E; = Ey; exp j{wt — k(x sin 6; — zcos6;)}. (12-66)

For the reflected wave, we can write

E, = Ep, exp f(wt — kizx — kz), (12-67)

where E,,, kiz, kiz are unknown constants. We have used the same value of w
as for the incident wave since, as in Section 12.1, all three waves have the
same frequency. We have also used only x and z terms in the exponent since
the derivative with respect to y must, again, be zero.
We set
E, = (Borat + EowyJ + Eouk) exp j(wt a KoxX = ky,2), (12-68)

where Eoiz, Eoiy, Eotzs Kor, koz are also unknown constants. Again we have the
same w, and we have no y term in the exponent. We let the unknowns be
complex in order that the various components of E, and of E, can be out of
phase with each other and in order that the dependence on x and on z can be
more general than with a simpler expression, such as that for the incident
wave.

* You may wish to omit Sections 12.4.1 to 12.4.1.4 inclusively.

528 PROPAGATION OF ELECTROMAGNETIC WAVES II

All the unknowns are independent both of the coordinates and of the
time, the only dependence on x, z, ¢ being expressed by the exponential
functions.
We represent the H vectors by similar expressions, in which the letter E
is replaced by the letter H and, for simplicity, we choose the origin of coor-
dinates at the interface, as in Figure 12-3.
We shall determine all these unknowns by using the boundary conditions
at the interface, the wave equation, and Maxwell’s equations.

12.4.1.1 The Wave Numbers ki, and k,, for the Reflected Wave. To satisfy
the boundary conditions at the interface, we proceed just as we did in Section
12.1. We first require that the exponents be equal at the interface z = 0.
Then, from Eqs. 12-66 and 12-67,

jo Sines = = sin 6. (12-69)

0

To find ky, we turn to the wave equation in medium 1.

ee ee ()> (12-70)
and

ki, = (@)a — sin? 6,), (12-71)

0

ee
ae cos 6; 75
(12-72)

We choose the positive sign in order to have a wave propagating in the proper
direction.
The reflected wave is thus of the form
E, = Eo, exp j{wt — k,(x sin 6; + z cos 6,). (12-73)
This is a plane wave reflected from the interface at an angle equal to the angle
of incidence.

12.4.1.2 The Wave Numbers k», and ky, for the Transmitted Wave. Equating
similarly the exponents for E; and for E,, again at the interface z = 0, we find
the wave number k», of the transmitted wave:

ni
x,0 sin 8. (12-74)
e
Kos = ky sin d; =

To find k»,, we again use the wave equation. Using the expression for E,,
12.4 TOTAL REFLECTION 529

ker + kd. = euow, (12-75)

Koz = + (expuow® — k,)'2, (12-76)
Na wate 1/2
=o (: a sin? 6s) , (12-77)
= Ny ny 2 A Ae 1/2
= +52 { (2) sin? 6; — 1} : (12-78)

The exponential function for the transmitted wave is thus

2 1/2
exp j{ot— 2xsinoi} +2 te sin?
6; — lp z}-
Xo Xo Ng

We have replaced the + sign before the z term by a + sign since the electric
field intensity must not become infinite as z > — «©. Then ky, can be written
with a plus sign:
2 1/2
kan=+524 (BY sint
oy—1b (12-79)
If @, were real, we would expect to have

Ny
ko, = — — COS 4. (12-80)
Xo
Comparing these two equations, we set*
2 1/2
COS 0; a {1= (™) sin? 6 : (12-81)
Nog

= — (1 — sin’ 9,)!”. (12-82)

We have elected to place the negative sign before the radical so as to preserve
the formalism. This is the value of cos 6, that we used earlier (Eqs. 12-53
and 12-54).

12.4.1.3 The Amplitudes of E and H in the Reflected and Transmitted Waves.

Now that we have found the wave numbers for the reflected and transmitted
waves, we only have to find the amplitudes of their FEand H vectors.

* Tt is probably useful to recall here that, if A is some positive real number, then
(—A)¥2 = jA'2,

and
ZAP = —j(— A)¥2,

These equations do not apply if A is negative. If you have any doubt concerning such
operations you should plot A, — A, and their square roots in the complex plane. It is rather
easy to be misled; for example,

A = (Ait = {(—1(—A))}2 = PA = —A!

530

nterface_

Figure 12-21. The E and H

vectors for the incident, reflected,
and transmitted waves in the
case of total reflection when the
incident wave is polarized with
its E vector normal to the plane
of incidence. The continuity of
the tangential component of
H across the interface makes
Aoty = O:

We can dispose of Hp, immediately because the relation between Ho, and
E,, is that which applies to plane waves in dielectrics (Eq. 11-61).
We can assume, as in Section 12.2.1, that the vectors Eo, and Ep, for the
reflected and transmitted waves are normal to the plane of incidence as in
Figure 12-21. Then
E; = Ep; exp j(wt = kyn;r) j, (12-83)

E, = Eo exp j(wt — kin,-r) j, (12-84)

E, = Eo: exp f(wt — Korx — ky,Z)j. (12-85)

We have written the exponents in the first two equations in a more compact
form than previously. The wave numbers k2, and ks, are now known (Eqs.
12-74 and 12-79).
For the incident and reflected waves, H is in the plane of incidence, just
as in Figure 12-4, and it has both x- and z-components:
H; ll= Ho(cos 6,t + sin 0;k) exp j(wt — kin;-r), (12-86)
H, = Ho(— cos 6,i + sin 6;k) exp j(wt — kyn,-r). (12-87)

For the transmitted wave, however, we must use a more general expres-
sion, since we know nothing as yet concerning its magnetic field intensity.
We therefore set

Hi. = (Hot + Howj + Howk) exp jot — keex — kez), (12-88)

and we utilize the fact that the tangential component of E must be continuous
across the interface. At the origin, which we have chosen at some arbitrary
point in the interface, the exponential functions of E;, E,, E, all reduce to
exp jwt; thus
Boe Eor = Evu. (12-89)
 12.4 TOTAL REFLECTION 531

Similarly, the continuity of the tangential component of H requires that

0 = Hoty, (12-90)

(Hoi — Hor) cos 6; = Hin, (12-91)

or, from Eq. 11-61, which applies to plane waves in dielectrics,
1/2
(*) M( Eo: — Eor) cos 0; = Apis. (12-92)
0

From the continuity of the normal component of B (or of H, since the

media are nonmagnetic) at the interface,
1/2

(*) nmy(Eo: + Eo) Sin 6; = Hote (12-93)

0

We are now left with four unknowns, E),, Eo:, Hotz, Ho.z, and only three
equations. We therefore turn to Maxwell’s equations. We choose one of the
simpler ones and apply it to the transmitted wave. Since

V-H, = 0, (12-94)
then
KozHoiz + kezAoi. = 0. (12-95)

Solving and substituting the values of ky, and of k»,, we find Eqs. 12-59 and
12-60, and
2
H 1/9 2JM COS 8; (sin?6; — (2) le
( v2) Ltt (*) al (12-96)
Ey: Lo eta ( ney. vil Ny i
N
COSIU any) ‘arg 0; A

(#)
Aon
2 (2)
€9\1
any sin 26;
aon (12-97)
COSI7 aay, eat ts = (™) i

To obtain E,, the value of Ey, must be substituted into Eq. 12-84. Simi-
larly, to obtain E,, H,,, H.,, the values of Ey,, Hoiz, Ho.z must be multiplied by
the exponential function of Section 12.4.1.2 for the transmitted wave.
Since both denominators on the right are identical, whereas one numer-
ator is real and the other imaginary, the two components of I are 7/2 radians
out of phase, and, at any given point, it has neither a fixed amplitude nor a
fixed direction but rotates in the plane of incidence. The transmitted wave is
thus rather complex.

12.4.1.4 The Poynting Vector for the Transmitted Wave. It is instructive to

calculate the Poynting vector for the transmitted wave. From Eq. 11-90,
532 PROPAGATION OF ELECTROMAGNETIC WAVES II

i Pak
(Sav = 5Re (KE. X Ht) = 5Rel0 0 |, (12-98)
He OE;

= 5 Re (EH) i - 1Re (E,H*) k, (12-99)

where
E, i acos 4; ue Ey,

cos 6; — Jysin? 6; — (“) ;

ae (2) sin? 0; — i} :} (12-100)

2 1/2

X exp {ieSe 0
sil 2

and where the stars identify complex conjugates (—/ substituted for /).
Then the x-component of Say is

:Re (EH) = (2)1/2 ny cos 6;. sinQ] 26; mes Ny (2) 2 seo 1}1/2 _
Mo _ (mY Ro (\ia
ny
(12-101)
and the z-component is zero.
The energy flow in the transmitted wave was discussed near the end of
Section 12.4.

12.5 REFLECTION AND REFRACTION AT

THE SURBACE OF A-GOOD, CONDUCTOR

We have found the wave number of a good conductor in Section 11.5:

~My (ooppw\'? gn ae ere

ie ( 2 ) (ore Jl 9), (12-102)
where 6 is the skin depth. Since the index of refraction my is complex and very
large, the direct application of Snell’s law, Eq. 12-13, leads to a very small
complex value for sin 6;.
As we shall see in Section 12.5.1, the laws of reflection, Snell’s law, and
Fresnel’s equations are again valid if we disregard the fact that my: is complex
and if we set
sin 0; Il _ sin 6;, (12-103)
2

cos 0, a ft= (™) sin? abe (12-104)

12.5 INTERFACE BETWEEN A DIELECTRIC AND A GOOD CONDUCTOR 533

Note that we require a positive sign before the square root, whereas we had a
negative sign in the corresponding equation 12-54 for total reflection.
The second term between the braces is complex, because of 1, but neg-
ligible since n, >> m (Section 11.5). Then

cos 6; & 1, 6, = 0, sin 0; 0 (12-105)

to a high degree of approximation. The wave penetrates into the conductor
essentially along the normal to the surface, whatever the angle of incidence.
The exponential function for £; is

4 ke Oe Z Zz
exp {i(wt X,x sin 0; 4- :)te ‘I.

This wave travels in a direction that is nearly perpendicular to the interface,

and it is attenuated by a factor of e over one skin depth 6.
The concept of skin depth therefore applies to an electromagnetic wave
incident at any angle on a good conductor. Whatever the angle of incidence 4;,
the transmitted wave can be considered to be a plane wave propagating along
the normal to the surface, with the enormous damping which is character-
istic of electromagnetic waves in good conductors.

INCIDENT WAVE POLARIZED WITH ITS E VECTOR NORMAL TO THE PLANE OF

INCIDENCE. From Fresnel’s equations 12-20 and 12-21,
A hs
MyU72 COS 0; — a (heap)
(2) = : Pb (12-106)
ON Maptre COS 8; + = (1 —/)
=) 21 py2 COS 0;
= ZaN\ 12-107
(= N Xo ( )
Ny plr2 COS 0; + F (ii)

The fact that the first ratio is negative means that the E vector of the reflected
wave is in the direction opposite to that shown in Figure 12-4, which was used
for the calculation. As to the second ratio, it is necessarily quite small since
Eo. = Eo: — Eo; and Eo, & Ep;. Also, the E in the conductor can be expected
to be small. The E vectors are shown in Figure 12-22.
Reflection from the surface of a dielectric with mn, >> m would also give
Ey,/ Eo; — 1 and a weak transmitted wave.
It is interesting to note that there is a small loss of intensity on reflection
from a good conductor, Ey, being somewhat smaller than Eo;. You will re-
member that, with total reflection, there is no loss of intensity and R is equal
to unity (Eq. 12-59).
534

Interface

Figure 12-22. The incident, reflected, and transmitted waves at the interface
between a dielectric and a good conductor. The incident wave is in the dielectric
and is polarized with its E vector normal to the plane of incidence. The H vector
of the transmitted wave is parallel to the interface, as shown, but it lags the E
vector by 7/4 radian.
The coefficient of reflection R is approximately equal to unity; the electric
fields of the incident and reflected waves nearly cancel on the surface, and a weak,
highly attenuated wave penetrates perpendicularly into the conductor.

Interface

Figure 12-23. Reflection and refraction at the interface between a dielectric and a
good conductor. The incident wave is in the dielectric and is polarized with its E
vector parallel to the plane of incidence. The E vector of the transmitted wave is
as shown, but it leads the H vector by 7/4 radian. The coefficient of reflection R
is approximately unity, the tangential components of the E vectors of the incident
and reflected waves nearly cancel on the interface, and a weak attenuated wave
penetrates perpendicularly into the conductor.
12.5 INTERFACE BETWEEN A DIELECTRIC AND A GOOD CONDUCTOR 535

INCIDENT WAVE POLARIZED WITH ITS E VECTOR PARALLEL TO THE PLANE OF

INCIDENCE. The three waves are as in Figure 12-23, again with

=)La
(= tonal (2)
ais (12-108)
The tangential components of Ey; and of Ey, nearly cancel at the surface of the
conductor, as expected, and the transmitted wave is again a weak, highly
attenuated plane wave which penetrates perpendicularly into the conductor.

Example COMMUNICATING WITH SUBMARINES AT SEA

One interesting application of the above discussion is the problem
of communicating with submarines at sea. For shore-to-ship com-
munication with the submarine antenna submerged, the efficiency is
extremely low, first because most of the incident energy is reflected
upwards at the surface of the sea, and second because the weak
transmitted wave is highly attenuated. It was shown in Problem
11-19 that the attenuation in sea water is about 53 decibels/foot at
20 megahertz, and 1.7 decibels/foot at 20 kilohertz. Very low
frequencies (15 kilohertz) are used at very high power. The
response of a submerged receiving antenna will be calculated in
Problem 14-24.
Ship-to-shore communication is presently impossible at these
low frequencies since the power required at the transmitter is too
large, and since it is impossible to use a sufficiently large antenna on
the submarine. Two-way communication is achieved at frequencies
of a few megahertz with the submarine antenna projecting from the
water.

Example STANDING WAVES AT NORMAL INCIDENCE

The three waves are shown in Figure 12-24. Since the direction of
propagation of the reflected wave is opposite to that of the incident
wave, and since E < H is in the direction of propagation, the H of
the reflected wave must be in phase with that of the incident wave
as in the figure. At the reflecting surface the electric field intensities
nearly cancel, and we have a node of E; the magnetic field intensities
add, and we have a loop of H. This is shown in Figure 12-25. The
nodes of E and H are thus not coincident but are spaced a quarter
wavelength apart.
A little thought will show that a similar situation exists for
reflection from any surface. Either the E or the H vector must change
direction on reflection, in order that the Poynting vector KE H can
change direction.
536 PROPAGATION OF ELECTROMAGNETIC WAVES II

Example TRANSMISSION OF AN ELECTROMAGNETIC

WAVE THROUGH A THIN SHEET OF COPPER
AT NORMAL INCIDENCE

We assume that the sheet is in air as in Figure 12-26.

We use the subscript ¢ for the wave transmitted through the
first interface, the subscript tr for the wave reflected back at the
second interface, and the subscript ¢t for the wave that emerges on
the other side of the sheet, as in the figure. We neglect multiple
reflections for the moment.
From Section 12.5 Eq. 12-107, setting m = 1, yu,» = 1, and
remembering that the ¢ wave progresses in the negative direction of
the z-axis,

E, = Ey exp ,(er3 a)Li ‘I, (12-109)

with
26
Eq:w= = rq
—— yh—_~ Eu: (12-110)
-

and, again from Section 12.5,

H, = Hy: exp {i
(wt+ ;)+ 5p (12-111)

with
€0 1/2
iva (<) Ev: (12-112)
Ho

so as to ensure continuity of Ho, at the interface. At the second inter-

face, we again have three waves: the ¢ wave with E, and H, as above,
the tr wave with

BOT,
< “Eo Ho
Nn ng
Interface
vi
\,/ Figure 12-24. Reflection at normal
incidence from the surface of a good
conductor. The electric fields of the
incident and reflected waves cancel
at the interface, or nearly so. The
magnetic field intensities add,
however, with the result that the
amplitude of the Ho, vector at the
y interface is 2Ho;.
 537

Figure 12-25. The standing-wave pattern resulting from the reflection of

an electromagnetic wave at the surface of a good conductor. The curves
show the standing waves of E and of H at some particular time. The
nodes of E and of H are not coincident but are spaced \/4 apart as
shown.

Ey Eotr EXP iu
Li(wt = ) - 3} (12-113)

Hy, = Go=)"Tein exo iu

Rea ;)= iI, (12-114)
and the ft wave with
Eu = Eout exp j(ot + x) (12-115)
0

€) \!/2 ; Zz
Ay (2) Four EXP J (or+- Z) (12-116)
Ho Xo

The continuity equations require that

E, a6 Ev, a Eu, (12-117)
and
HA, —-— Ay, = Aun (12-118)

or
= Ex
a-)% ?(E,— Ev) (12-119)
538

Figure 12-26. An electromagnetic wave incident normally on a con-

ducting sheet of thickness a. The subscripts /, r, t, tr, tt refer, respectively,
to the incident wave, to the wave reflected at the first interface, to the
wave transmitted at the first interface, and so on. Multiple reflections
inside the sheet are neglected.

at z = —a. Keeping in mind that (c/eqw)!/2>>1 for a good con-

ductor, we find that

eae
ee exp | Eo
at aliek
+A) p> (12-120

Eu
ae _ 2eer
2 exp {i _ 444
el TD , (12-121)

E,, = Ey exp {i( == t 22 aoe _ 2a, (12-122)

En = 2E exp {i(ora aad = ‘)- “I. (12-123)

It will be noticed that at z = —a, with the above approximations,
the E vector of the tr wave has the same phase and amplitude as
the t wave, while the E vector of the ¢f wave is also in phase but has
twice the amplitude.
In the incident wave,

Stay = ;Re (E; X HD), (12-124)

1 €0 1/2 >
3; ay = Sl3 iin ESOi> (12-125)
=

whereas, in the ¢ wave,

12.5 INTERFACE BETWEEN A DIELECTRIC AND A GOOD CONDUCTOR 539

1
Stay = 5 Re (E; X Hh), (12-126)
1/2
Stay = 23/2 (<2) eS. 4, (12-127)
Just below the first interface, at z = OM ona aoe eSIIICE
(weo/o)!/? & QV? and © is small (Section 11.5 and the example on
page 479). Then, inside the sheet, the Poynting vector is further
reduced by a factor of e~2/°, On the far side of the sheet, from Eqs.
12-121 and 12-110,
1 1/2
Sit avy — 2 () Ee, (12-128)

16 78
o
e-raldg, (12-129)

ll {2 (<2) }(e-218) {20ar; Siav» (12-130)

/

where the quantities in enclosures show respectively the loss in

energy flux that occurs at the first interface, in traversing the sheet
of thickness a, and at the second interface.
For a sheet of copper (¢ = 5.80 X 107 mho/meter) 25.4 microns
thick and at 1 megahertz, 6 = 6.6 X 10-5 meter (Table 11-1), and
Sttay = (2.8 XK 107)(0.46)(5.5 X 10-5)8; ay, (12-131)
1) oe ee (12-132)
It will be observed that the enormous attenuation arises mostly from
reflection at the faces of the conducting sheet, for the thickness we
have chosen.
Let us now take into consideration multiple reflections within
the copper sheet. We have seen that reflection is nearly perfect at the
second interface. This reflected wave proceeds to the left through
the conductor and is similarly reflected to the right at the first inter-
face, and so on. To take into account the wave that is reflected back
at the second and then at the first interface, we must add to the
above E;,; the term
PD IT. 40)ji(e ee : 4 _ ae +h
0

For the wave which is reflected four times, we add the term

26, exp fi(

oF + ae — at +h
0

and so forth. Then, taking into accoumt multiple reflections,

El, = 2E,, exp si(e nee 5)s =i +i

0

x +: exp | F +i at exp | a rahar “+ (12-133)

540 PROPAGATION OF ELECTROMAGNETIC WAVES II

where the quantity between braces is of the form

feb? byte spe1 (12-134)

Then
Eq: &Xp ji(oRe - = )t
Eo See (12-135)
a
sin j);-
<—(1 + if
inh a

exp j——
= oe epee TE (12-136)
* sinh +a +a
6

and

Siitvay ss >(=)"
1 /
ae
i
ie
eee | em . (12-137)

Ls(<2)
2
aA
a + es

4 WE,

o
5,5, (12-139)
sinh? ricos? hel + cosh? 7 sin? R
5
1 29107 ys (12-140)
which is larger than when multiple reflections are neglected. This is
of course true only for the particular case we have chosen.

12.5.1 Demonstration of the Validity of Snell’s Law,

of the Laws of Reflection and Refraction,
and of Fresnel’s Equations at the Interface
Between a Dielectric and a Good Conductor*
We shall proceed in the same manner as for total reflection and write that

E; = Ej; exp j{wt — k,(x sin 6; — z cos 6;)}, (12-141)

E, = Eq exp j{wt — k(x sin 6; + z cos 6,)}, (12-142)

E, = Ey: exp j(wt — Korx — kp,2), (12-143)

with axes as in Figure 12-27. We have not taken the trouble to show that E, is
of the form shown above; the demonstration is identical to that which led to

* You may wish to omit Sections 12.5.1 and 12.5.1.1.

 541

Figure 12-27. An electromagnetic

wave incident on a good con-
ductor, with its E vector normal
to the plane of incidence. The
angle of reflection is equal to
the angle of incidence. The
continuity of the tangential
component of H across the
interface makes Hor, = 0.

Eq. 12-73 for total reflection. We have set all three circular frequencies equal,
as previously.

12.5.1.1 The Wave Numbers ko, and ko, for Refraction into a Good Conductor.
We can again equate the exponential functions for E; and E, on the interface,
just as we did in Sections 12.1 and 12.4.1.2. Then

ny
(es [ea in Oe = x, sin 6;, (12-144)
as for total reflection.
We can find k,, from the wave equation for good conductors,

PE:
Ox:
,aE,
02?
= joo. (12-145)

The derivative with respect to y is zero since the wave is assumed to be inde-
pendent of the y-coordinate. Then

—k3, — kee = jwor, = — ki, (12-146)

where, from Eqs. 11-104 and 11-105,

ky = =i = - (12-147)
0

Thus
kee 1/2
hap = oe he (1- iH) ; (12-148)
2

== + va
hi (2)
a are
sin a (12-1499)

Since the “‘index of refraction” n: of the conductor is much larger than

that of the dielectric m, the second term under the square root can be neg-
lected. Thus
542 PROPAGATION OF ELECTROMAGNETIC WAVES II

gee aay es +t J, (12-150)

and
Kor K Keo,. (12-151)

We have selected the negative sign before kz so as to make the imaginary

part of ky, positive. This is required to prevent the wave from building up to
infinite amplitude as z— — ~.
If both m, and nz were real, we would have

I = — ky cos. (12-152)
Comparing Eqs. 12-149 (with a negative sign before k2) and 12-152, we see
that we were justified in writing cos @, as in Eq. 12-104.

12.5.1.2 The Amplitudes of E and H in the Reflected and Transmitted Waves.

We assume that E; is normal to the plane of incidence. Then the E vectors are
all normal to the plane of incidence and parallel to the interface as in Figure
12-22.
At first sight one expects a discontinuity in the tangential component of
the magnetic field intensity at the interface, because of the existence of surface
currents. This is incorrect, however, and there is continuity for the following
reason: currents flow inside the conductor (unless it is superconducting), and
they are distributed over a finite thickness. There is zero surface current. Thus,
if we draw a closed path like the one in Figure 9-106, with its two long sides
infinitely close to the surface, there is zero current enclosed by the path and
the tangential component of H must be the same on both sides of the interface.
If we make the path a little wider, keeping one side some distance away from
the conductor and pushing the other one into the conductor, then the current
linking the path is not zero and the tangential H’s are not equal. But then we
are not comparing the tangential components of H at the interface. This
point was mentioned in Section 9.7.
A similar argument shows that the tangential component of E is the
same on both sides of the interface. Thus

Eyi + Ew = For (12-153)

Os Any (12-154)
1/2
€
(*) ny( Ep; — Er) COS 0; => Hotz, (12-155)
0
 12.6 RADIATION PRESSURE AT NORMAL INCIDENCE ON A GOOD CONDUCTOR 543

€ 1/2
(=) n,(Eo; + #,,) sin 0; = H,,., (12-156)
0

KeeHoi2 + kezHo = 0. (12-157)

From the last equation,
iz > Abus (12-158)
since k», is much smaller than k», (Eq. 12-151). The wave in the conductor
therefore has its E vector parallel to the y-axis and its H vector parallel to the
x-axis. It propagates in the negative direction of the z-axis.
Upon solving for Ey, and E): we find Eqs. 12-106 and 12-107. Solving
for Hoi, shows that E and H in the conductor are related as in Eq. 11-106 and
that E leads H by 7/4 radian.

12.6 RADIATION PRESSURE AT NORMAL

INCIDENCE ON A GOOD CONDUCTOR

We have now discussed at some length the reflection and the refraction of an
electromagnetic wave at the interface between two dielectrics, and then at the
interface between a dielectric and a good conductor. We also wish to investi-
gate the same phenomena for the case of a dielectric and an ionized gas.
Before going on with this, however, we shall study another phenomenon
that is related to reflection from conductors, namely radiation pressure.
Let us consider an electromagnetic wave incident on a good conductor.
We limit ourselves to normal incidence. We have seen above that the electric
field intensity transmitted into the metal is small compared to that of the
incident wave, but that it is not zero. We have also seen that the FE and H
vectors of the transmitted wave inside the conductor are orthogonal, just as
in the incident wave.
According to Ohm’s law, the electric field intensity E, in the conductor
gives a current density cE,, where o is the conductivity of the medium. This
current, which is oriented like E,, is perpendicular to H;,. It turns out, as we
shall see presently, that under these conditions the conduction electrons are
pushed by the magnetic force Qu X B in the direction of propagation of the
wave. The electrons in turn push on the conductor in the process of colliding
with the atoms in their path, giving rise to radiation pressure.
The situation would be very different if the conduction electrons were
entirely free to move through the metal: the conductivity would then be
544

A nterface
Figure 12-28. The conduction
current cE, and the magnetic
induction pol which give rise to
radiation pressure in an element
ae of volume of a reflector. The
ice sides of the element of volume
Ca are parallel to the E and H vec-
dz tors. The element has a thickness
ee dz and is ab square meters in
Z ly area.

imaginary, as in the low density gases that we studied in Section 11.6.1, and
the radiation pressure would be zero.
Let us study the radiation pressure of a wave incident normally on a
good conductor. We consider an element of volume that is parallel to the
interface, of thickness dz as in Figure 12-28, and ab meter? in area. It carries
a current cEb dz, and is submitted to a magnetic force cEy)Hab dz in the
negative direction of the z-axis.
Note that the relative permeability u, of the conductor does not enter
into the calculation. It has been shown that the magnetic force exerted on an
electron within a magnetic material is proportional to u)H and not to uH, for
slow electrons like the ones we are considering.*
Then the pressure exerted by the wave on the element of thickness dz is

dP = oE oH dz. (12-159)

With £, and H, chosen to be positive in the directions shown in Figure 12-28,

a positive result will show that the wave pushes on the conductor.
From Eq. 12-107 for 6; = 0 and %) > 4,

E, — 2Hnmd
“=p Eo;pn exp f
iu(wt + 5:)4 23)” (12-160)

We cannot continue using the exponential functions, however, since we must

evaluate their product (see Appendix D). We must therefore rewrite them in
the form

*G. H. Wannier, Phys. Rev. 72, 304 (1947).

12.6 RADIATION PRESSURE AT NORMAL INCIDENCE ON A GOOD CONDUCTOR 545

2 p95
E, = aK, Eye? cos (or+ ;+ i) (12-162)

H, = 2Hp,e?/* cos (oray ;): (12-163)

Then

Bie reardz5 (12-164)

= : 4(2!?)n,
9
—
E, A;
¢22/5 dz
o2/
cog (orlaZ \cos (wien x (12-165)
c ) 5 4
This is the instantaneous value of the pressure on the element of thickness
dz at z. To find the average pressure we must replace the term between the
the braces by its average value over one period T = 21/w:
1 1h

rf cos (01+ ‘)cos (ot+ -+ i)di = ae (12-166)

Then

Jide ob = eels ss (12-167)

The average pressure is in the same direction at all depths within the con-
ductor.
Finally, integrating over all z within the conductor, from — to 0,
0

Pepa ee | er2/sos (12-168)

ae Fouts, (12-169)
This is a positive quantity. As we saw at the beginning of this calculation,
the fact that P,, is positive means that it is in the direction of propagation
of the incident wave. The quantity P,, is the average pressure exerted on
the conductor by the incident radiation, or the radiation pressure.
If the incident wave propagates in a vacuum,

Si av
Py = 2 wri (12-170)

where 8S; ay is the average of the absolute value of the Poynting vector for the
incident wave (Section 11.3).
We can ascribe this pressure to a change in momentum of 2 8,,/c per
unit time and per unit area in the incident wave, the factor 2 being required
because the wave is reflected with a momentum equal to its initial momentum,
but of opposite sign.
546 PROPAGATION OF ELECTROMAGNETIC WAVES II

An electromagnetic wave propagating in a vacuum can therefore be

considered to involve a flux of momentum that is equal to 8,,/c, or to its
energy density, according to Section 11.3; that is,
(Flux of momentum) = (Momentum density) X (Phase velocity), (12-171)
= (Energy density). (12-172)

These results agree with those of atomic physics, where we consider an

electromagnetic wave to involve photons of energy fo(f = 1.05 * 10-*4
joule-second is Planck’s constant h divided by 27) and of momentum f/A
traveling with a velocity u:
u = hw. (12-173)
The radiation pressure of electromagnetic waves has been observed
experimentally, and it has been found to agree with the above theory. This is
a confirmation of our hypothesis that the effective magnetic induction acting
on a slow electron inside a magnetic material is uo, and not zH. It also is a
demonstration of the fact that an electromagnetic field can have a momentum.

Examples Radiation pressure is small for the usual radiation intensities. In

sunlight the Poynting vector is approximately 1.4 kilowatts/meter?
at the surface of the Earth, giving a radiation pressure of about
9.3 < 10-° newton/meter? or 0.1 microgram/centimeter? on a me-
tallic reflector. The radiation pressure P varies as the inverse square
of the distance from the source, as does the Poynting vector S.
To find P at the surface of the Sun we must multiply the above
figures by
mean distance of Earth to Sun 3 G xX 101!
radius of Sun 5 7 X 108
y = 46 X 104,
(12-174)
which still gives only 4.4 milligrams/centimeter?, or 4.3 X 10-6
atmosphere. Radiation pressure is unimportant even in the interior
of the Sun, but it possibly plays an important role in the more lu-
minous stars. As is well known, comet tails point predominantly
away from the Sun. This phenomenon is explained in part by radia-
tion pressure.
Much higher power densities and pressures are available in
wave guides, where Ss, can reach values of the order of 10°
watts/meter?. Then the radiation pressure is 0.1 gram/centimeter?,
assuming that the waves propagating within wave guides are plane.
This is not correct, as we shall see, but the error is unimportant here,
since we are concerned only with orders of magnitude.
In a laser beam with a power density of 10"! watts/meter? the
radiation pressure is 10 grams/centimeter’.
 547

12.7 REFLECTION OF AN ELECTROMAGNETIC

WAVE BY AN IONIZED GAS

Finally, we study the behavior of an electromagnetic wave that encounters

an ionized gas. We assume, as in Section 11.6 that the electrons do not collide
with the molecules of the gas; in other words, we assume that the pressure is
low. Under those conditions the phase velocity is larger in the ionized gas
than in free space, the index of refraction being

we ‘= {1_ (*) } .
2)1/

(12-175)
1/
{1— 80,5 x _ (12-176)
from Eqs. 11-173 and 11-174, where c is the velocity of light, u is the phase
velocity of the wave, w, is the plasma angular frequency of the wave, N, is the
number of free electrons per cubic meter, and f is the frequency of the wave.
If the ionized gas had a definite boundary and a uniform value of N,
throughout its volume, reflection and refraction at its surface would be simple
to describe: the ionized gas would simply act as a dielectric with n < 1. As
a rule, neither assumption is valid and the wave 1s reflected in much the same
way that a light wave is reflected in a mirage.
It is possible to calculate the path of a ray by performing a numerical
integration of the ray equation stated in Problem 12-2. However, one can
deduce the main features of the reflection by simply using Snell’s law.
We select coordinates as in Figure 12-29 and assume that the index of
refraction n varies slowly with z, but not with the other two coordinates
x and y. To be more specific, we assume that n varies by a negligible amount
over one wavelength. If N, gradually increases with z, a given ray gradually
bends down as in Figure 12-29 to an angle 6, at a point where the index of
refraction is 7.
We can calculate @ in the following way. When refraction occurs at the
interface between any two media m and mm, the quantity n sin @ is conserved
in going from one side of the interface to the other. This is Snell’s law, Eq.
12-13. If the index of refraction varies continuously, the medium can be
imagined to be stratified in infinitely thin layers, and n sin @ is similarly con-
served all along the ray. Thus

nsin@ = ny, sin 6;, (12-177)

548

Figure 12-29. An electromagnetic wave incident on an ion-

ized region at an angle 6; is deflected at an angle @ after
penetrating to a distance z.

where n; is the index of refraction at z = 0. If we set m, = 1, then

nsin 6 = sin 6;. (12-178)
It is interesting to differentiate this equation with respect to the distance
] measured along a ray. We find that
do _ l1dn
t= a qi ano (12-179)

If the ray penetrates into an ionized region where the ion density increases
with z, the index of refraction n decreases with / and the derivative dn/d] is
negative, so that the angle @ increases with distance as in Figure 12-29. After
some distance, if n becomes sufficiently large, @ becomes equal to 90°. At that
point the tangent of @ becomes infinite, but dn/d/ becomes zero. After this,
tan 6 becomes negative, whereas the derivative dn/d/ becomes positive, and @
keeps increasing until the ray escapes from the ionized region at an angle equal
to the angle of incidence 6;.
At the top of the trajectory,

sin 6 I 1, (12-180)
Ngoc = Sin 6;. (12-181)

This is the index of refraction required for reflection when the angle of inci-
dence is 6;.
12.8 SUMMARY 549

From Eq. 12-175,

sec 6; = —, (12-182)
Wp
where w is the angular frequency of the wave and w, is the plasma angular
frequency at the value of z where 6 = 90°. For normal incidence, 6; = 0,
sec 6; = 1, and reflection occurs at a z sec wp» = w, that is, where n is zero
and the phase velocity is infinite. At oblique incidence, however, sec 6; > 1,
and reflection occurs where wy) < w, or at a lower value of z, if we assume that
the electron density N., and hence w,, increase with z.
We have found above the value of dé/d/. This is the reciprocal of the
radius of curvature R, and thus

1 _ 1dn
R = - dl tan @. (12-183)

A positive value of R correponds to a positive value of d@/dl and to a tra-

jectory which is concave downwards in this case. Also,

1 1 dn
(ae eile (12-184)

where /’ is perpendicular to the direction of propagation, as in Figure 12-29.

The ray therefore bends most sharply where the index of refraction n varies
most rapidly in the direction perpendicular to the ray.

12.8 SUMMARY

The laws of reflection, Snell’s law, and Fresnel’s equations result from the
continuity of the tangential components of EF and of H at the interface between
two media.
The Jaws of reflection are the following: (a) the angle of reflection is
equal to the angle of incidence, (b) the normals to the wavefronts of the
incident and reflected waves lie in a plane that also contains the normal to the
interface and that is called the plane of incidence.
According to Snell’s law,

SE
sin@;
te
Ne
(12-13)
where 6, and 6; are the angles of incidence and of refraction respectively and
where 7, is the index of refraction of the first medium and np is that of the
second medium.
Fresnel’s equations relate the amplitudes and phases of the reflected and
550 PROPAGATION OF ELECTROMAGNETIC WAVES II

transmitted waves to those of the incident wave. The subscripts N and P

indicate that the E vector of the incident wave is either normal or parallel to
the plane of incidence.
"1 cos 6; _ ” cos 6,
(=) = eee niy yams (12-20)
EgiJy Ma cos 6; + 2 cos 8,
rl Mr2

E 2 eis cos 6;
(?) = Se (12-21)
OT = 60s Ga —— COsU:
Mri Lr2

ey eee » 603 8,
Ew = Br2 Mri 5 (12-25)

Evi] p fz Ay
— cos 6; + — cos #;
br2 bri

E 2Dee cos 6;
(*) Be ER
oea (12-26)
1 = 63 6; — cos @,
ae Mri

At the Brewster angle for dielectrics there is no reflected wave if the

incident wave is polarized with its E vector in the plane of incidence:

my
= COUUsze (12-40)
9

The coefficient of reflection R gives the fraction of the incident power

which is reflected. Similarly, the coefficient of transmission T is the fraction of
the incident power which is transmitted. Thus R + T = 1.
Total reflection occurs when Snell’s law gives sin @, = 1. Then R = 1.
The transmitted wave travels along the interface, and is attenuated exponen-
tially in the direction perpendicular to the interface. The average energy flow
across the interface is zero, but the instantaneous flow is alternately one way
and then the other. Snell’s law and Fresnel’s equations give the correct results
when applied to total reflection if we set

cos 6, = —(1 — sin? 6,2 = -{1 — (2) sin? abe

se (12-54)

Good conductors have a coefficient of reflection R close to unity. The

transmitted wave penetrates nearly perpendicularly into the conductor,
whatever the angle of incidence, m) being much larger than 7. It is attenuated
by a factor e in one skin depth
PROBLEMS 551

fe (WOoM2
Z iy (12-102)
The Fresnel equations again apply if we set
* deren Ani m\2 . , \ 12
cos #; = +(1 — sin? 6,2 = a = (egypt aa » (12-104)

At normal incidence, standing waves are formed in front of a good con-

ductor, the reflecting surface giving a node of E and a loop of H. Nodes of
E and ofH are spaced )/4 apart.
Radiation pressure is due to the magnetic force on the electrons oscillating
under the influence of the electric field of the wave. It is shown that
8: av
dees = 2 ——+ (12-170)
u

where P,y is the average value of the radiation pressure in newtons/meter?,

S; ay 18 the average value of the magnitude of the Poynting vector in watts/
meter’, and wu is the phase velocity of the wave in meters/second. This leads to
the concepts of flux of momentum and of momentum density for an electro-
magnetic wave. The forces due to radiation pressure are small and are usually
negligible.
An electromagnetic wave is reflected by an ionized gas much as light
waves are reflected in a mirage. If the ionization density does not vary in the
direction parallel to the interface, we find that
nsiné = n, sin 6, (12-177)
where 6; is the angle of incidence, m is the index of refraction of medium 1,
and @ gives the orientation of the ray at the point where the index of refraction
is n (Figure 12-29). Ifn, = 1, then @ = 90° at a point where n is equal to sin 6;.
If N, increases with z, the value of z where 6 = 90° increases as the angle of
incidence 6; decreases and, at normal incidence, reflection occurs at a point
where w is equal to the plasma frequency w, and where the phase velocity is
therefore infinite.

PROBLEMS

12-1. In a plane sinusoidal wave, the components of EF and H are of the form
W = Wyexp/(wt — k-n),
= W exp j(wt — krx — kyy — k.z),
where W, is a constant that can be complex. See Problem 11-11.
552 PROPAGATION OF ELECTROMAGNETIC WAVES II

The vector k, which is called the wave number (not to be confused with
the unit vector along the z-axis), can also be complex. Then we can write that
i an, — jbny,

where mn, and ny are unit vectors, with k, = am: — jbno,, and similarly for
k, and k,.
(a) Show that surfaces of constant phase are normal to n,; and that surfaces
of constant amplitude are normal to no.
(b) Show that a is related to the wavelength, and 4 to the attenuation.
(c) Show that, in all operations involving V, this operator can be replaced
by —jk.
(d) Rewrite Maxwell’s equations utilizing this fact.
(e) It does not follow that E, H, k are orthogonal when p, = 0 and ¢,, y, are con-
stants. To show this, use Maxwell’s equations, with W expressed as a cosine
function, and then find the conditions for orthogonality.

12-2. A wave travels in a stratified medium whose index of refraction is a function of

the coordinate y.

(a) Show that the angle @ between a ray and the y-axis is given by:
G0 me ee
ds
where the distance s is measured along the ray.
(b) You can now verify the ray equation

d
7 ot) = Vn,

where t is a unit vector tangent to the ray at a point where the index of refrac-
tion is n.{
. Show that
tis 2uriNs COS 6;
seed) = .
Ao; N Mrolly COS 0; i Mrill2 COs 6,

fy) _ __ Mra COS 0; — Myilly COS 6,

A; N Mrolly COS 0; oti Mrilto COs Gs

12-4. Electromagnetic radiation is incident normally on a dielectric whose index

of refraction is n.
Show that the reflected wave can be eliminated by covering the dielectric
with a layer of a second dielectric whose index of refraction is 7/2 and whose
thickness is one-quarter wavelength.

5. Calculate the Brewster angles for the following cases:

(a) light incident on a glass whose index of refraction is 1.6,
(b) light emerging from the same type of glass,
(c) radio frequency wave incident on water (n = 9 at radio frequencies).

12-6. Show that there is no reflection from a p/ate of material illuminated at the
Brewster angle when the E vector of the incident wave lies in the plane of
incidence.
PROBLEMS 553

12-7. (a) Show that, for a wave incident in air on a nonconducting magnetic
medium, (£,/Ep;)p is zero for

tan? 6; = slereH
Err — 1
and hence that the Brewster angle exists only if €, > p,.
(b) Show that, similarly, (£o,/Eo;)y is zero for

tan? 6; = Hr(or — Er)

Err — |

In this case there is a Brewster angle only if u, > «,.

12-8. Show that the coefficient of reflection R and the coefficient of transmission 7
are both equal to 0.5 at normal] incidence on the interface between two di-
electrics if the ratio of the indexes of refraction is 5.83.
12-9. A 60-watt light bulb is situated in air one meter away from a water surface.
(a) Calculate the root mean square (rms) values of E and H for the incident,
reflected, and refracted rays at the surface of the water directly under the bulb.
Assume that all the power is dissipated as electromagnetic radiation. The
index of refraction of water is 1.33.
(b) Calculate the coefficients of reflection and of transmission.
12-10. Light is transmitted at normal incidence from a medium with index of refrac-
tion n to another of index n + a.
(a) Express the coefficient of transmission T as a power series in a/n, and show
that the first two terms give an accuracy of 1 % for (a/n) S +, approximately.
(b) What is the value of 7 for light passing from air to glass (n = 1.52)?
(c) Discuss the reasons for using coated lenses in optical instruments.
12-11. A plane wave is reflected at the interface between two dielectrics. The wave
is incident in medium 1, and m,/n2 = 1 + a.
(a) Show that the coefficient of reflection for waves polarized with their E
vectors normal to the plane of incidence is

where
A* = 1 — a(a + 2) tan? 6,,

0; being the angle of incidence.

(b) Show that A = 0 at the critical angle.
12-12. An electromagnetic wave polarized with its E vector normal to the plane of
incidence is totally reflected at the interface between air and a dielectric whose
index of refraction is 3.0. The angle of incidence is 75°.
(a) Calculate 6,/A, and 6,/Ai.
(b) Calculate the phases of the reflected and transmitted waves with respect
to the incident wave at any point on the interface.
554 PROPAGATION OF ELECTROMAGNETIC WAVES II

(c) Draw wave fronts for all three waves in the neighborhood of the interface,
showing the phase shifts found above.
Draw parallel lines spaced 27 apart to represent the wave fronts.
(d) Check the continuity of E across the interface.
(e) Compare your results with Figure 12-16.

12-13. A plane-polarized electromagnetic wave is totally reflected at the interface

between two media and the E vector of the incident wave has components
parallel and normal to the plane of incidence.
Show that the reflected wave is elliptically polarized and that the compo-
nent whose EF vector is parallel to the plane of incidence leads the other com-
ponent by
ih Ny 2) UyiP7
cos 9,4sin 6; — (®) f .
2 arc tan -
sin? 6;

12-14. In the case of the total reflection of a wave polarized with its E vector normal
to the plane of incidence, we have found in Eqs. 12-96 and 12-97 that the x-
and z-components of the H vector in the transmitted wave are out of phase
by 7/2 radians.
In what direction does the H vector rotate with respect to the direction
of propagation of the transmitted wave: does it rotate like the wheels of a
vehicle moving on a road or does it rotate in the opposite direction?

12-15. A scintillator is a substance that emits light when traversed by an ionizing

particle, such as an electron. This light is detected by a photomultiplier, which
produces electric pulses that actuate counters and other electronic equipment.
The scintillator has an index of refraction m and is fixed to the face of the
photomultiplier with a cement of index my. Light is emitted in all directions
in the scintillator, a fraction F being transmitted out of the scintillator in the
direction of the photomultiplier.

(a) Calculate the fraction F as a function of = assuming that all the light is

transmitted for angles of incidence smaller than the critical angle and that
the scintillator is surrounded by a nonreflecting substance.
(b) Draw a graph of F for values of 1)/n, ranging from 0.1 to 1.0.

12-16. Construct sin (a + jb) and cos (a + jb), where a and b are both real, in the
complex plane.
Over what ranges of values can these two functions vary ?

12-17. Show that, at the surface of a good conductor in air,

6
Eor

Eo a
. ~ 1 — = cos 6;,
a COS

if the conductor is nonmagnetic.

12-18. (a) Calculate (Eo,/Eo:)p for reflection from the surface of a good conductor.
(b) Show that the reflected wave leads the incident wave by the angle
PROBLEMS -
555

Vor areian 2a cos 6;

PW Sa
SS ==
a? — 2 cos? 6;
where

2 6
Qe nye aly
Bri Xo

12-19. A wave is incident at an angle 6; on a plate of dielectric backed by a perfect

conductor. Under what conditions can multiple reflections be avoided?

12-20. Draw two figures similar to the left-hand one in Figure 12-6c, showing E and
H for an electromagnetic wave incident on a good conductor.
You will of course have to exaggerate the values of Ey, and of \ in the
conductor.
Be sure to show the phases correctly. Show x, y, z axes on both figures
so as to relate one to the other.
12-21. We have seen in the second example on page 535 that the standing wave
pattern obtained by reflecting an electromagnetic wave at normal incidence
on a good conductor gives nodes of E which are half way between the nodes
of H.
Show that the electromagnetic energy density is constant throughout the
standing wave pattern.
12-22. The surface impedance of a conductor is defined as the ratio E;/H;, at the
surface, where E; and H; are the tangential components of EF and H.
It was shown in the Problem 10-8 that H; is numerically equal to the
current per unit width in the conductor.
(a) Show that the surface impedance of a good conductor is
. (ou?
Ol vd) ($4) a OE
a+a/
od
The quantity 1/c6 is called the surface resistivity.
Both the surface impedance and the surface resistivity are expressed in
ohms/square. They give the impedance, or the resistance, between opposite
sides of asquare surface of any size. For example, the surface resistivity of copper
is 8.25 x 10°? ohm/square at 1 gigahertz.
(b) Show that the power dissipated per square meter in the conductor is

1
206
yp
to

where Hj is the amplitude of H at the surface.

12-23. Calculate the radiation and gravitational forces exerted by the Sun on the
Earth.
Assume that the radiation incident on the Earth is all absorbed. Then the
radiation pressure is S,,/c, and not 28,,/e as in Eq. 12-170.
The value of Syy is 1.4 kilowatts/meter’.
556 PROPAGATION OF ELECTROMAGNETIC WAVES II

12-24. (a) Compare the forces due to the gravitational attraction and to the radia-
tion pressure of the Sun on a spherical particle of radius r and of specific
gravity 5.
The Sun has a mass of 2.0 * 10% kilograms, and it radiates 3.8 « 10%
watts in the form of electromagnetic radiation. The gravitational constant is
6.7 X 10-" newton-meter?/kilogram?’.
In computing the radiation pressure, assume that the particle absorbs the
radiation. See Problem 12-23.
(b) Calculate the value of r for which the two forces are equal.

12-25. A radio wave is incident at an angle 0; on the ionosphere. One may assume
that the electron density depends only on the altitude.
(a) Show that, at any given point in the ionosphere, the radius of curvature
of a ray is given by
nm
R=
~ (dn/dz) sin 0;
(b) Assuming that n decreases linearly with altitude, where does the ray bend
most sharply, and what is the value of R at that point?
(c) Show that a ray would describe a circular arc of radius R if the index of
refraction were given by
Rsin 6;
n=
z+ Rsin 6;
(d) Draw a sketch of such a ray showing 6; and the center of curvature.
Is n equal to sin 6; at the top of the trajectory?

Hint
12-2. (b) Select your y axis along Wn, and your x axis in the plane of incidence.
CHAPTER l3

PROPAGATION OF
ELECTROMAGNETIC WAVES III
Guided Waves

In Chapter 11 we studied the propagation of plane electromagnetic waves in

an unbounded region, first in a vacuum, and then in various media. Then, in
Chapter 12, we investigated the reflection and the refraction of plane waves
at the boundary between two different media. We shall now study the manner
in which waves can be guided in prescribed directions by metallic wave guides.
Various types of such wave guides are illustrated in Figures 13-1, 13-2, and
13-3.
We first investigate the propagation of electromagnetic waves in a
straight line, along the z-axis, without making any assumption as to their
dependence on the x- and y-coordinates. This discussion will therefore be
more general than those in Chapters 11 and 12. After that, we shall study two
relatively simple types of guided waves, which will serve as examples.

13.1 PROPAGATION IN A STRAIGHT LINE

We assume that the medium of propagation is isotropic, linear, and homo-

geneous; then D = cE and H = B/y. We also set its conductivity equal to
zero, since electromagnetic waves are strongly attenuated in conductors. This
does not exclude metallic wave guides because the wave is propagated outside
the conductors. Thirdly, we set the charge density p, equal to zero in the
medium of propagation. This makes V-E equal to zero (Section 3.6).
Fourthly, we assume zero attenuation. The method used for dealing with
attenuation will be discussed in Section 13.3.4. Finally, we assume that
558 PROPAGATION OF ELECTROMAGNETIC WAVES III

propagation occurs in a straight line, namely in the positive direction along

the z-axis.
Then, for a sinusoidal wave,
E= (ELozt oe EuyJ +- E,.k) exp j(wt — k,z) = E, exp j(wt = ie): (13-1)

H = (Aut + Aoyj + Aok) exp j(wt — kz) = Hy exp j(wt — k,z), (13-2)

where the coefficients Ey,, Eo,, Eo:, Hor, «++ are as yet unspecified functions of
x and of y only. The dependence on ¢ and on z appears in the exponential
function, which is characteristic of a wave propagating in the positive direc-
tion of the z-axis.
The wave number for the guided wave,
1
i x,’ (13-3)

is not necessarily equal to the wave number for a plane wave (Section 11.3)
and is real if there is zero attenuation.
It is interesting to compare our present procedure with that of the
preceding chapter. In discussing reflection and refraction, we used exponential
functions to describe the dependence of the reflected and transmitted waves
on all three coordinates and on the time, as for example in Eqs. 12-14 to 12-16.
Here we use an exponential function only for z and rt, whereas the dependence
on x and on y is left unspecified. Let us write out Maxwell’s equations for this
field. Since p; = 0, it follows that V-E = 0. Then
Olea Oleg. = oh
= + By Jk Eve. (13-4)

Similarly, since V-B = 0, and since yu is independent of the coordinates,

V-H = 0 and

fo ap
OAs 0H, ee
em (13-5)
From V X E = —0B/0t,
OEte 4 ;
—jouAoes, (13-6)
i + jk, Eo, =

[ ORs 3
Tk, Eon Ia cae = Jou, (13-7)

OE oy, OE ox = A
Ox oy a Jop Hoe, (13-8)

and, from V X H = 0D/0dt,

Obloxs es g
- + jk Ho, = JweEo, (13-9)
Oy
13.1 PROPAGATION IN A STRAIGHT LINE 559

; OM: :
Jk, Hox ae ce cs —JweEoy, (13-10)

0H, OA: ;
Te — ah = JweE;. (13-11)

We can now show that the four transverse components Ep,, Eo,, Hox, Hoy
can be deduced from the two longitudinal components £p, and Ho,. From
Eqs. 13-7 and 13-9,

Foz = i
—jou
i ih (= Facey
k, OE oz wae)
(X, # A). (13-12)
MG Sait
We have made the additional assumption that the radian length 4 =
1/w(eu)'/? for a plane wave is different from the radian length of the guided
wave, or that A, ~ X. The case where A, = A will be discussed separately in
Section 13.1.2.
Similarly,
jou a) ae
_ (13-13)
( wu dy ar (A, # A),
ce Lee
2 x2

syaloee oe — ee) ey, Orr

(XA,# A), ( )
B= ie len we Ox
2 x?

x ‘ois
x
OE oz k, at)
— — Joe
(XA,# X). ( )
aera (epee oy
le
He
It 1s obvious, by inspection of the above four equations, that the wave is
completely determined once Ey, and Ho, are known.
The wave equation 11-54,
OE
Vi = oe = —k°E, (13-16)

provides a differential equation for Ep.:

O7Ey, , 07 Ez see He ’
Ax? + ay? — Ey = —k’En., (13-17)

or
O°Ey, , OE _ (5 x 3) E 13-18
ax? + dy? re x2 x2 Oz» ( = )

and an identical equation applies to Ho,:

0°Hy. ,OH: (i “i3) iy (13-19)
Ox? oy?
560 PROPAGATION OF ELECTROMAGNETIC WAVES III

The radian length 4, of the guided wave is as yet unspecified; this is a

constant which must be selected in such a way that Eo. and Hp. can satisfy the
above differential equations, and the boundary conditions defined by the wave
guide. It turns out, as we shall see in Section 13.3.1, that only certain discrete
values of A, are possible. These values are called the characteristic or eigen
values of the equation, and they depend (a) on the frequency, (b) on the
geometry and on the electrical characteristics e,, 4, of the medium of propaga-
tion, and (c) on the geometry and on the electrical properties o, u, of the
guiding structure.
The general procedure for calculating FEand H is therefore the following.
We first solve the above wave equations for Ep. and Ao., using the boundary
conditions for the wave guide under consideration. This gives not only Eo. and
A., but also X,. The other components of F and of H are then deduced from
Eqs. 13-13 to 13-16. It is of course also possible to use Maxwell’s equations
directly, in conjunction with the appropriate boundary conditions.

13.1.1 TE and TM Waves

It is convenient to consider separately two types of wave: Transverse Electric

(TE) waves, for which Ey. = 0, and Transverse Magnetic (TM) waves, for
which Ho, = 0. We shall study one particular type of TE wave later on in
Section 13.3.
Let us write out the transverse components of Ey) and of Hy in vector form:

Eo: = Eort + Eq ij, (13-20)

Hy: = Aozt a A,j. (13-21)

We can gain information on the relative orientations of these two vectors

from their scalar product. We find that

Eoit Moir = EnrHoz + Eo,Hoy = 0 (13-22)

for both TE and TM waves. The transverse components of E and of H are

everywhere mutually perpendicular. This applies to any TE or TM wave
propagating in a straight line. Our demonstration is valid only for X, ~ A, but
the result is correct for any value of X,, as we shall see in Section 13.1.2.
Let us calculate the ratio Eo,/ Ho. for TE waves (Ey. = 0). We can select
our axes in such a way that, at the point considered, Ey, = Hor = 0. Then

For Bow wm (“\" Xg

| , (13-23)
Ay. shige Ke, € X
A
= 377 ei ohms (er = 1, ae = 1). (13-24)
0
 13.1 PROPAGATION IN A STRAIGHT LINE 561

Similarly, for TM waves (Ho. = 0),

Ev: = k, uy ( 1/2 X
Hewes Veins (13-25)
x
= 377 . ohms Ge ea (13-26)
g

The ratio Ey:/Hp; is called the wave impedance and is a real positive
number.
Our demonstration is again valid only for X, ¥ X, but the result is really
correct for any value of X,, as will also be shown in Section 13.1.2.

13.1.2 TEM Waves

Let us consider waves for which 4, = X. Since the factor

as
Re
is Zero, the parentheses on the right in Eqs. 13-12 to 13-15 must also be zero:
k Ey , OHo. _
fy Bil ga 0, (13-27)

kena Ep oii
ay ae 0, (13-28)

Oby | wk Oly
Ove bueroxe a ¥ arse)
OFo. , K OH. _
Aiea tockaye a 0. (13-30)

The last two equations are equivalent to the first two, since, by hypothesis,
A, = Aand k/wu = we/k = (c/n)? in this case.
These equations can be satisfied by setting Ey, and Ho, both equal to zero,
in which case we have a Transverse Electric and Magnetic (TEM) wave.
The TEM wave has some interesting characteristics. To begin with,
X, = X; hence the phase velocity u = wA is the same as that of a plane wave in
the medium of propagation. This velocity is 1/(eu)!/2 and is independent of
the frequency, insofar as e and p are themselves independent of the frequency.
If the wave propagates in.a vacuum, its velocity is c, whatever the
geometry of the wave guide, and whatever the frequency. Such a line is said
to be distortionless because the various frequency components of a complex
waveform are all transmitted at the same velocity.*

* This is only approximately correct in practice. Because of the finite conductivity of

metal wave guides, there is both attenuation and dispersion.
562

Figure 13-1. A hollow, perfectly conducting tube.

The TEM wave also has the following remarkable property. We have
seen in Sections 6.5 and 8.2 that
aA vy,
E= ———— (13-31)

where the first term is associated with changes in the magnetic field, and the
second with accumulations of charge. Now the currents are longitudinal, as
we Shall see in Section 13.1.3, and A must therefore also be longitudinal, as
well as 0A/dt. Then A = Ak, and

-1i-Fj-(F a
ax" ays Oz + Ot :
(13-32)

But, since E is transverse, the longitudinal component of VV must cancel that

of 0A/dt exactly at all points,

(13-33)
and

E= ——i-—j. (13-34)

Since we are dealing with a wave, E and V must be of the form

E = Eyexpj (otS ‘): (13-35)

V = Voexpj («1= ‘) (13-36)

where Ey and Vo are functions of x and of y only, and thus
 563

(a) (b)
Figure 13-2. (a) Shielded-pair line. The signal is applied between the two
wires; the outer cylindrical shield is grounded. (b) Parallel-wire line.

E= al (Fe
Mt i+ eA eej (« ee‘).
ay )exp (13-37)

= Ey expj @ — ;) (13-38)
with
Ey = _ Vy,
rea _ Vo,
ay ap (13-39)

This quantity Ep is the electric field intensity in a plane perpendicular to

the direction of propagation (that is, in a plane parallel to the xy-plane) for
corresponding values of z and ¢ which make the phase angle wt — (z/A) equal
to zero. This E, is derivable from the potential V) in exactly the same manner
as an electrostatic field.
If the wave guide is a hollow, perfectly conducting tube as in Figure 13-1,
the tangential component of E at its surface is zero, Vy is a constant all around
the tube, and the only possible solution inside is V) = constant. Now, if Vo is
constant throughout the inside of the guide, E) is zero, EK = 0, and, since
V X Eis —0B/dt, there is no H wave either. We therefore conclude that
transverse electric and magnetic, or TEM, waves cannot be transmitted inside
hollow conducting tubes. This is not rigorously true because TEM waves are
transmitted if the wavelength is very much less than the cross-sectional
dimensions. For example, light can of course be transmitted through a
straight length of metal pipe. We shall see in Section 13.3.2 that the TEM
wave is then a limiting case of a TE wave.
In a coaxial line, as in Figure 13-4, the inner conductor need not be at the
same potential as the outer one, and E need not be zero. A field is therefore
possible, and a TEM wave can be transmitted. Other lines with more than one
conductor, such as the shielded pair or the parallel-wire lines illustrated in
Figure 13-2, can also transmit TEM waves.
564 PROPAGATION OF ELECTROMAGNETIC WAVES III

Let us rewrite Eqs. 13-4 to 13-11 for Eo, = 0, Ao, = 0. These eight
equations are simply Maxwell’s equations, as applied to a wave propagating
in the positive direction along the z-axis. We now have only six distinct
equations:
Oboe , Oy
ae -+ ay e—— 0, (13-40 )

Oz Oo,
—— = 13-41
Ox oy 0, ( )
m EZ

y= = (*) Aor, (13-42)

he (*) ee (13-43)
m 1/2

OE
a
OEys
ee
_ 0, (13-44)

ony — Ee = 0. (13-45)
Ox Oy

We first note that, for a plane wave propagating along the z-axis, as in
Chapter 11, the field vectors depend solely on z and on f¢, and the derivatives
with respect to x and to y are zero. Then Eqs. 13-40, 13-41, 13-44, 13-45
become identities, and Eqs. 13-42 and 13-43 are equivalent to Eq. 11-61. Such
a plane wave is the simplest form of TEM wave.
For any TEM wave, we can see, from the third and fourth of the above
equations, that E and H are mutually perpendicular, just as for the transverse
components of TE and TM waves. The wave impedance EF,’ Ho; can be
calculated as in Section 13.1.1, choosing coordinate axes so that &), = Ay. =
0. Thus
Eo: as Eo: ie 7
Ho, % A, id (‘) ose

= 377 ohms (ec, = 1, wu, = 1). (13-47)

In addition, the electric and magnetic energy densities are equal,
l l
5 eR? = 5H’, (13-48)

and the total energy density is «£°, or wH?.

From Eq. 13-43, the average Poynting vector is

Sw = 5ries
Ree X HY) = 5heres
(5) ER, (13-49)
—s ({\"
u : rms k /t, 13-50
( a= )

— vEbrask, (13-51)

Up Hens ke. (13-52)

13.1 PROPAGATION IN A STRAIGHT LINE 565

This vector is directed in the positive direction of the z-axis, which is the
direction of propagation of the wave, and its magnitude is equal to the phase
velocity u = 1/(eu)? multiplied by the average energy density, just as for a
plane wave (Section 11.2).
To find E and H, one first chooses the static field Ey that corresponds to
the mode of propagation desired. For example, with the coaxial line of
Figure 13-4, Eo is radial. Then Eq. 13-37 gives E and Eqs. 13-42 and 13-43
give H. We shall work out the field of the coaxial line in Section 13.2.

13.1.3 Boundary Conditions at the Surface

of Metallic Wave Guides
For any type of wave the tangential component of E must vanish at the
surface of an infinitely conducting guide. Then, close to the surface of the
guide, E must be normal and, in particular, Ey), must be equal to zero. Real
conductors, of course, do not have infinite conductivity unless they are
superconducting, and we shall deal with this matter in Section 13.3.4.
There is also another boundary condition that applies to any type of
wave guided by a perfect conductor and that is interesting from a qualitative
point of view. Since the current density in the guide is tangent to its surface,
we may use the results of Problem 10-8. Then, close to the guide, H must be
(a) tangent to the surface, (b) perpendicular to the current density, and (c)
equal in magnitude to the surface current density expressed in amperes/meter.
For example, with TM and TEM waves, H is everywhere transverse and the
currents in the guide are therefore longitudinal. We have already used this
result in the preceding section.
Finally, in the case of a TE wave and a perfectly conducting guide, there
is a third boundary condition that concerns Hy, and that we can find by
writing out the value of Ey:
Ey = Eqri + EvyJ, (13-53)
Ey, being equal to zero by hypothesis. Using Eqs. 13-12 and 13-13,
_ 0H: . OH. ) i!
Jou
B=prs sr ( are 2 ase (13-54)

le OG

= Hk x Vie (13-55)
Ao ae
where k is the unit vector in the direction of the z-axis. The three vectors Ko,
k, V Ho, are shown in Figure 13-3. It will be observed that V Hp, is necessarily
566

Figure 13-3. Portion of a

rectangular wave guide. The
electromagnetic wave propagates
inside the tube. If the guide is
perfectly conducting, the electric
field intensity E is zero in the
conductor, and E is either nor-
mal or zero at the surface. For a
TE wave, it is shown that V Ap-
is tangent to the wall.

tangent to the conducting wall and, therefore, that the rate of change of Ho.
in the direction normal to the surface must be zero.

132° THE? COAXIAL, LINE

In the coaxial line illustrated in Figure 13-4 the electromagnetic wave propa-
gates in the annular region between the two coaxial cylindrical conductors,
and there is zero field outside. This type of wave guide is often used in
electronic equipment. The medium of propagation is a low-loss dielectric.
We shall make several simplifying assumptions, namely, that (a) the
guide is straight and its cross section is constant throughout its length, (b) the
electrical conductivity of the walls is infinite, (c) the time dependence of the
wave is described by a cosine function, and (d) the wave travels in the positive
direction of the z-axis and there is no reflected wave traveling in the opposite
direction. Assumptions (c) and (d) do not limit the generality of our calcula-
tion, since, according to the principle of superposition, the net field of a
number of waves is simply given by the vector sum of their fields.
We have seen above that a TEM wave can be transmitted down such a
guide; this is the mode that we shall study in this section. TE and TM modes
are also possible, but they are much more complex and they are not used in
practice.
Thus
Eq, = 0, Hy, = 0, (13-56)
and, from Eq. 13-37,
eet oo BOLE UIR OV. . _2
E = (Soi+ jp) exp (« ‘): (13-57)
 567

Figure 13-4. The E and H

vectors inside a coaxial line.
They are both transverse, E
being radial and H azimuthal.
The vector product E x H
always points in the direction
y of propagation.

The dependence of E on x and y is the same as for a two-dimensional

electrostatic field. For a given position z along the guide, and for a given time
t, Eis radial and varies as 1/p, where p is the radial distance from the axis to
the point considered. Then
E= *expj (1 -;) P1 (13-58).

in cylindrical coordinates, p; being the unit vector in the radial direction, and
C being a constant.
Since we have a TEM wave, the radian length X, of the guided wave is the
same as that of an infinite plane wave, or 1/w(euo)!/*, and the velocity of
propagation wA = 1/(euo)!/? is the same at all frequencies.*
The line voltage U, which is the potential of the inner conductor with
respect to the outer conductor, is
Po p Zz
V= i Edp = Clin * exp j (at ~Z) (13-59)
pi Pi nN

From the previous section, H is orthogonal to E, and

1/2
H = (<) casts j («1-7)1. (13-60)
Ho p A

The vectors E and H are oriented as in Figure 13-4 and the Poynting vector
E X H points in the direction of propagation.

* Assuming that the line has infinite conductivity and that e- is independent of the
frequency. In practice, the velocity of propagation does depend somewhat on the frequency.
See Section 13.1.2.
568 PROPAGATION OF ELECTROMAGNETIC WAVES III

The line current J flowing along the surface of the inner conductor can be
calculated from the circuital law, Eq. 7-77:
1/2
I= i}
Hp; de = 2p; (<) <exp j (ot-3), (13-61)

Bee
See! | w eex ==
€r= 1, br = 1) 2) (13-62)
An equal current flows in the opposite direction along the inner surface of the
outer conductor. Note that C is equal to 60 times the peak value of Jif e, = 1,
Te
The average transmitted power can be calculated by integrating the
average Poynting vector over the annular area between the two conductors:
Po
Wr = HeSav 2mp dp, (13-63)

where

SH SRE (EOGH = (<) Gre (13-64)

1/2 2

Z Ho 2p?
Thus

Wr =
Ces
= In— watts © =i =i) (13-65)
120 Pi

The average transmitted power can also be calculated from JimsUrms =

(CUR ere
The characteristic impedance of a coaxial line, which is the ratio U/I
when there is no reflected wave traveling in the opposite direction, as we have
assumed at the beginning, is

60 In ohms G. = go 1),
a

The wave impedance E/H is 377 ohms, as we saw earlier in Section 13.1.2.

13.3 THE HOLLOW RECTANGULAR

WAVE GUIDE
Hollow wave guides are extensively used at microwave frequencies. These
wave guides are simply metallic tubes inside which electromagnetic waves
propagate by reflection on the inner surfaces, in much the same way that
sound waves propagate through a tube. We shall consider hollow wave guides
of rectangular cross-section because their guided waves are relatively simple
and because they are the most widely used.
 569

Figure 13-5. A hollow, rec-

tangular wave guide. The wave
propagates in the positive direc-
tion along the z-axis.

We make the same simplifying assumptions as for the coaxial line of the
previous section. We also assume that the dielectric is air.
We have shown in Section 13.1.2 that TEM waves cannot propagate
inside a hollow tube. Thus the radian length 4, of the guided TE or TM waves
is different from 4%, which is Xo, or 1/c(epu)'/? in this case.
We select axes as in Figure 13-5, with the wave propagating in the
positive direction along the z-axis.

13:31) The TE Wave

We now consider the transverse electric (TE) wave (E, = 0). This wave is
simply a plane electromagnetic wave whose E and H vectors are oriented as
in Figure 13-6 and which is reflected back and forth on the walls parallel to
the xz-plane. We shall use this fact to simplify the analysis somewhat, but we
could also find the TE wave by solving for Ho,, using the proper boundary
conditions, without referring to the reflection.
We must find Eo,, Eoy, Hox, Hoy, Hoe, Ag = 1/k,. From Figure 13-6 we can
see immediately that
fon 0, His = "0; (13-66)
To determine the remaining four quantities we shall proceed as indicated at
the end of Section 13-1: we shall solve the wave equation for Ho, for the
proper boundary conditions. This will give Ho, and 4,. The values of Eo, and
Ay, will then follow from Eqs. 13-12 and 13-15.
We have found in Eq. 13-19 that Ho, obeys the wave equation
PH,
aePH. .(1 1| re (13-67)
570

Figure 13-6. Typical plane wavefront

with the £, H, and E x H vectors
inside a rectangular wave guide. The
reflection occurs on the faces parallel to
the xz-plane. The E vector is trans-
verse, that is, perpendicular to the
direction of propagation Oz, and the
wave is TE.

In the simple TE wave illustrated in Figure 13-6, the wave is independent of

the x-coordinate, and
dton [il |
ie (5 = a) A:. (13-68)

Solving this equation, we find that

Hy, = F sin Cy + Gcos Cy, (13-69)
where

=—C? = cele 13-70

te oo
We now apply the boundary condition discussed in Section 13.1.2:
OH:
reid 0 atx = Oandx =a, (13-71)

ae =0 aty = Oandy = 5b. (13-72)

The first condition is already satisfied, since Ho, is not a function of x. From
the second condition,

Oz
= C(f cos Cy — Gsin Cy) = 0 aty=Oandy=b. (13-73)
oy
To satisfy the boundary condition at y = 0, F must be zero. The condition
C = 0 must be rejected because it implies that A, = Ao, which is not compati-
ble with a TE wave. At y = 3,
sin Cb. = 0, (13-74)
and thus

c=" (13-75)
13.3. THE HOLLOW RECTANGULAR WAVE GUIDE 571

where is an integer that cannot be zero, for then C would also be zero. Then

Hue Gos “ay. (13-76)

Let us find X,. This is easy now that we know H),. We can either substi-
tute Ho, into the wave equation or use Eqs. 13-70 and 13-75 to obtain

a? | 1
BB WO choad} (13-77)
It will be noticed that A, can have only certain discrete characteristic
values corresponding ton = 1, 2,3,---.
Since the left-hand side of the above equation is negative, A, > Ao. This
means that the wavelength in the guide is /onger than that of a plane wave of
the same frequency propagating in free space. Then the phase velocity wu is
larger than c. We shall return to this point later on.
Now that we have found Ap. and X,, we can calculate the two remaining
unknowns £p, and Ao, from Eqs. 13-12 and 13-15, setting \ = Xp since e, = 1,
Hr = 1, by hypothesis:

ge eee
AT b
y, (13-78)

a Dees
Ay = nae sin | :
(13-79)
or, setting

(Bigg ne
nr
(13-80)
then
: nn : Z
E, = Egor sin | — y })exp j @ -Z) (13-81)
b X,

te in ( y)exp j («1 ) (13-82)

= 00x =) a »)

Hes_ Enon sg (3
nt; cos (0 v)exp j(
{ wt -z)
Kyi (13-83 )
where the values for X, are given by Eq. 13-77. The coefficient Eo, is the
maximum value of E inside the guide. We have already found that Ey, = 0,
Hor = 0 (Eq. 13-66).
Let us consider the value of E,. It can be understood qualitatively as
follows. In any plane wave front such as that shown in Figure 13-6, E is
independent of the x-coordinate; since the wave along the z-axis results from
the superposition of such plane waves by multiple reflections, its E must also
be independent of the x-coordinate. The same cannot be said about the y
dependence, however. If we consider a single elementary plane wave progress-
572

(a) (b)

Figure 13-7. The amplitude of E for a TE wave in a hollow rectangular wave

guide. The value of n of Eq. 13-78 is 1 in (a) and 2 in (b).

ing at an angle along the guide, the amplitude and the phase of E are constant
over a wave front such as that shown in the figure. The superposition of such
waves by multiple reflections gives an interference pattern, with the result that
the amplitude varies with the y coordinate.
Let us now consider the meaning of the quantity n. For n = 1, Eo, varies
from 0 at y = Oto a maximum of Eo, at y = b/2 and to zero again at yp= b
as in Figure 13-7a. Forn = 2, Ey, is zero in the middle ofthe guide at yp= 6/2,
and is of opposite sign on either side, as in Figure 13-7b. The different values
of n thus correspond to different modes of propagation inside the guide.
Now let us return to Eq. 13-77 for A,. It can be rewritten as
A l
i = Mo 2 ve (13-84)
= (35) }
1
= EE, fora = 1. (13-85)
2b) J
This equation shows that 4, is real if \) < 26. If X, is real, the exponential
functions of Eqs. 13-81 to 13-83 describe an wnattenuated wave. Therefore, if
the above inequality is satisfied, that is, if the frequency is high enough, and
if the walls are perfectly conducting, a wave can propagate inside the hollow
wave guide without attenuation.
If the above inequality is not satisfied, that is, if the frequency is too low,
then A, is imaginary, and the exponential functions show that the field is
13.3 THE HOLLOW RECTANGULAR WAVE GUIDE 573

attenuated exponentially with z. Then the phase does not vary with z, there is
no wave, and there is zero energy flow into the guide once the field is es-
tablished.
The attenuation is rapid. For example, if the frequency is too low by a
factor of 2, \) = 4b. Then, from Eq. 13-84,
1 pone
hes 3'5/ a (13-86)

We must choose the negative sign, for otherwise the amplitude would
increase exponentially with z. Then

exp (-7=) = exp (-=—), (13-87)

9 Xo

and the wave is attenuated in amplitude by a factor of 4 X 107° in one free-

space wavelength ))!
The average Poynting vector $,, and the average transmitted power are
attenuated by the square of this, or by a factor of 16 X 107!" in one do! The
wave guide therefore acts like a high-pass filter, with the lower frequency limit
determined solely by the width b, and not by a.
In the special case where the free-space wavelength X» is 2b, the guide
wavelength \, becomes infinite, as does the phase velocity. This corresponds
to w = w, in the case of propagation in an ionized gas (Section 11.6.3). The
quantity \, = 2b is called the cut-off wavelength.
Wave guides can therefore be used only if the free-space wavelength Xo
is shorter than twice the distance between the reflecting walls for the n = 1
mode. They are inherently high-frequency devices. For example, if b = 10
centimeters, \) must be shorter than 20 centimeters, and the frequency must
be higher than 3 X 10°/0.2, or 1.5 gigahertz.
Let us now rewrite the components of EF and H for the mode n = 1,
which is normally used:

a EST (2) exp j @ = ). (13-88)

9g

Fine 0. (13-89)
E, = 0, (13-90)
H, = 0, (13-91)

A, =_ Foor. (tv
Sin sin (b )exp/i
| wt = a
wy (13-922)

H, _= id
TEwe cos (
ay
h )Expy ; (1 ba x,
fee LY
4 13-93
(13-93)

sehaS
sh cos (h )exp jeen)
Ccot= x, (z+ I,
4) (13-94 )
574

Figure 13-8. Lines of E and lines of H for a TE wave with n = 1 propagating in

a hollow rectangular wave guide. The ovals are lines of H. The lines of E are
vertical straight lines and are represented by dots and crosses.

where the guide wavelength is

= (iy shia
2) 1/2 > No, (13-95)

{1- Gi) }
and the phase velocity is

u P eet ae
fi2 é is ag (13-96)
2b

It will be observed that £, and H, are in phase, but that H, has the same
oN
phase at z + z as have the two other vectors at z.

Figure 13-8 shows lines of E and ofH for a TE wave with n = 1 propa-
gating in a hollow rectangular wave guide.
13.3. THE HOLLOW RECTANGULAR WAVE GUIDE 575

13.3.2 Internal Reflections

Let us return to Figure 13-6, which shows a typical wave front for a wave
zigzagging down the guide. It will be instructive to investigate the field by
considering the interference resulting from the multiple reflections.
Figure 13-9 shows the multiple reflections in more detail. Let us assume
that along the fixed line AB the electric field intensity of the wave which
propagates upward and to the right is Ey exp jwt. The line AB is thus parallel
to the wave fronts for this wave. The lines BC and DE are similarly fixed and
parallel to wave fronts for the wave propagating to the right and downward.
These two waves must interfere at B to give zero E for all values of t at
the perfectly conducting wall, since their electric field intensities are per-
pendicular to the paper and parallel to the wall. Then the electric field
intensity along BC must be E) exp j(wt + 7), or —Ey exp jwt. At C, inter-
ference must again give zero E for all ¢. Then, along CD, E is Ey exp j(wt +
2m), or Ep exp jut.
Thus, AB and CD are one free-space wavelength \» apart, and, from
the triangle CH/ in the figure,

(13-97)

Figure 13-9. A plane electromagnetic wave propagating in a rectangular

hollow wave guide along a zigzag path. The lines AB and CD are parallel
to wave fronts for the wave propagating to the right and upward. Simi-
larly, BC and DE are parallel to wave fronts traveling to the right and
and downward. The angle a is the angle of incidence; the broken line
FCG represents a ray reflected at C.
576 PROPAGATION OF ELECTROMAGNETIC WAVES III

If the wavelength were n times smaller, the same lines AB and CD would
be n wavelengths apart, and then we would have

COSa =
= 5, :
(13-98)

There are obviously only certain discrete values of the angle a that permit
destructive interference to occur at the walls of the guide.
We have therefore found a geometrical interpretation for the ratio
\o/(2b/n). From Eq. 13-84,
Nye
do ol sin a
(13-99)

where the guide wavelength X, is either AC, or CE, or BD.

At the critical wavelength, \) = 2b/n, cosa = 1, a = 0, and the wave
fronts are parallel to the axis of the guide. The traveling wave then degener-
ates into a standing wave between the wave guide walls.
For \y K 2b/n, cos a > 0, and a — 7/2. The TE wave then approaches
a TEM mode as a limit.
The phase velocity is

tp = te= = >. (13-100)

This is the velocity at which the phase propagates along the guide. It is larger
than c because the individual plane wave fronts are inclined at an angle with
respect to the axis of the guide. This can be seen from Figure 13-9 as follows.
Consider AB to be a wave front propagating parallel to itself and to the right
at a velocity c. Then the point A moves along the z-axis at a velocity that is
larger than c.
We can also consider the velocity at which a given signal progresses
along the length of the guide. The z component of the velocity of an indi-
vidual wave front is only c sin a and is smaller than c. Thus, if we call this
velocity u,, then
Pe (ei ay (13-101)
No Py ee

= {1 -(F)} <6, (13-102)

and
Ustly = C*. (13-103)

This is in agreement with our requirement that the signal velocity never
exceed c (Section 5.10).
13.3. THE HOLLOW RECTANGULAR WAVE GUIDE 577

13.3.3 Energy Transmission

Let us consider the energy transmitted by a TE wave in a low-loss rectangular
wave guide. We shall assume the usual » = 1 mode. The field is then com-
pletely described by Eqs. 13-78 to 13-83, with n = 1, or by Eqs. 13-88 to
13-94, The first set is slightly more convenient for our purpose.
From Eq. 11-90, the average value of the Poynting vector is
1
si = 5)Re(E X H%*), (13-104)

where H* is the complex conjugate of H. In the present case,

BA Sheed:
Say = 5Re 12a Tig 5Re(—E.HYj + E,Hyk). (13-105)
OGAs eH:

Substituting the values of E,, H,, H., we find that the first term in the paren-
theses is imaginary, whereas the second term is real.
The energy therefore flows only in the direction of the z-axis and

Say == ae
Evoz sin” 2 |
(7 y)k.is (13-106)

The value of S.y is independent of x, as expected, since the amplitude and

phase of the wave are independent of the x coordinate. It is zero at the walls
y = Oand y = 5, where E is zero, and it is maximum at y = 5/2, again as
expected.
The total average transmitted power is thus
y=b 2
Foor . @ )
Wr r = i Tope a (omaWeredy, (13-107 )
Egorab
= AGG
——3 (13-108)
=a 0

= Epozab i { raat ey
etl 1/2
13-109
Acto ic 2b J ee )

Let us compare this transmitted power with the average electromagnetic

energy per unit length within the guide. The instantaneous electric energy
density is (1/2)e)E?, and its average value is (1/4)e£5. The average electric
energy per unit length is thus
b

1 ‘ < WE €() 2
= eghdon
if4 €09£00 Sill” (;
b y JadLy = —8 ab Eppe.
0) (13-110)

To find the average magnetic energy content per unit length, we proceed
578 PROPAGATION OF ELECTROMAGNETIC WAVES III

similarly for both the y- and z-components of H and add the results, since
HoH, ane, (13-111)
Again the result is (€/8)ab Ejor, and the average electric and magnetic
energies per unit length are equal. This is reasonable, since the plane electro-
magnetic waves that produce the field configuration by reflection at the side
walls involve equal electric and magnetic energy densities. It is not obvious,
however, because the interference effects tend to confuse the picture.
The total average electromagnetic energy content per unit length in the
guide is therefore epab Ejo,/4. Upon dividing the total average transmitted
power by this quantity, we find that

> ab Ap sy hae ogo ey (ye 13-112

Foor Awpoh, Exoreab 4 i x i 2b) J (1k)
i (13-113)
The average transmitted power is thus equal to the product of the
average energy per unit length times the signal velocity u, as we could have
expected, intuitively.

13.3.4 Attenuation
We have assumed until now that the walls were perfectly conducting; let us
now consider real wave guides of finite conductivity.
In the process of guiding electromagnetic waves, conductors dissipate
part of the wave energy in the form of Joule losses. This is because the waves
induce electric currents in the guide. A rigorous calculation of the field for a
guide of finite conductivity is difficult, but fortunately unnecessary.
The procedure used for calculating the Joule losses is the following. We
have performed a calculation on the assumption that the guide is perfectly
conducting. This led to a field in which there is a tangential H at the surface of
the guide. Since the tangential H must be continuous across any interface, we
know the value of H inside the conductor. Then, using Maxwell’s equations,
we can find the corresponding tangential E inside, which is not zero unless
the guide material is a perfect conductor. This small tangential E is then
considered to be a perturbation of the ideal field obtained with perfect
conductors. The method is entirely satisfactory because this E is so small that
it hardly disturbs the wave. We thus have a tangential E, a tangential H, and
a Poynting vector that is normal to the conducting surface and directed into
the metal.
That both EFand H vectors must exist inside the conducting walls can also
13.3. THE HOLLOW RECTANGULAR WAVE GUIDE 579

be shown as follows. To begin with, we must have a tangential H just inside

the wall. On the other hand, at some distance within the wall, there must be
zero field, since the attenuation distance 6 (Section 11.5) is quite short at fre-
quencies which are high enough to propagate in wave guides. For example,
if b = 7.5 centimeters, the frequency can be, say, 3000 megahertz and 6 is then
only 1.2 microns in copper (Table 11-1). The tangential component of H thus
decreases rapidly with depth inside the conducting wall. Then V X H is not
zero, and there is a current density J; parallel to the surface and normal to H
since V X H = J;. We must therefore have both a tangential E to produce
the tangential current density J; and a tangential H.
The average Poynting vector directed into the guide wall gives the
average power W,; which is removed from the wave per meter of length. We
then wish to calculate the attenuation constant k,;. This constant must be
such that, when both the E and the H of the transmitted wave are multiplied
by exp (—k,,z), the average Poynting vector for the transmitted wave and the
average transmitted power Wr decrease by a factor of
exp (—2k,: Az) & 1 — 2k_i Az (13-114)
in a distance Az. The approximation is excellent for ordinary types of wave
guide. Then
W1 Az = (2k, Az)Wr, (13-115)
or

kes = ar
— Wi
(13-116)
i

The real part k,, of k, can be taken to be the k, obtained on the assumption
of perfectly conducting walls.
It might be expected at first sight that the attenuation could be calculated
from the reflection losses. It will be recalled from Section 12.5 that an electro-
magnetic wave reflected from a good conductor is slightly weaker than the
incident wave. This method of calculation is incorrect because, as we shall
see, there are also energy losses in the guide faces parallel to the yz- plane.
Let us calculate k,;. The tangential H produces an electromagnetic wave
that penetrates perpendicularly into the wall. Inside the conducting wall,

Ea (~)" girls, (13-117)

o

as in Eq. 11-106.
We assume that the dielectric inside the guide is dry air, and is therefore
lossless. We also assume that n = 1 in order that the field can be described,
as a first approximation, by Eqs. 13-88 to 13-94.
580 PROPAGATION OF ELECTROMAGNETIC WAVES III

Along the face that lies in the xz-plane,

ig ete
wiob
eae
Naat
5)
2.
(ise Olas CS)
Then E, is not equal to zero for y = 0, as in Eq. 13-88, but is rather
1/2
E, = (=)
o
ee
lod
ay (otba XA
2 1)
4
(y = 0). (13-119)
Note that, as the conductivity « approaches infinity, E, approaches zero. The
average Poynting vector (1/2) Re (E X H*) is directed into the guide wall and
is equal to
TE, Ox 2 1
( 5 ) og! Qerpi) 8/2 (= Uy

This is the average energy flowing into the wall at y = 0, per square meter
and per second. It is interesting to note that this energy flux is the same at all
points on the face y = 0. The power lost to the wall per meter of length is a
times larger and, for the two faces parallel to the xz-plane,
tT Epox \2 2a
iva Geta error eee
This is the average power lost by reflection.
At the face x = 0, H has y- and z-components as in Eqs. 13-92 and 13-94.
Associated with H,, we have an electric field intensity

a (ts!)1/2 Es on (7) ae («xar

Z
i) (13-121)
o WHOA, g

The magnitude of the corresponding average Poynting vector directed into

the walls is now
(5) : sin? (? .
AG o}/( 2a)3/2 b

Similarly, the magnitude of the average Poynting vector corresponding to H,

1S

(He) 1 cos? =
b | eeu) by
Integrating the sum of these two quantities from y = Oto y = b and multiply-
ing by 2, we obtain the average power lost per meter in the two walls parallel
to the yz-plane:

We
___Ebe
bo Quy
_[, , (2b)
|} aie (2) | (13-122)
or, using Eq. 13-85,
1 Eo e 2D\*
Wye = bot Quo 7 (~) : (13-123)
 581

kid”

law,
he
al
ae
0 N i= i= = \ a ! |
02 0.4 0.6 08 1.00
My Atenagiob
Figure 13-10. Dependence of k,;b*/? on \o/d. for a/b = 0.5
and for copper.

The total average power loss per meter is then

W = Wrz Sig Wie (13-124)

See Cae 2b
= painanar {5 + (5) | a)
To find the attenuation constant k,; for the wave, we now use Eqs.
13-116, 13-125, and 13-109:

ie ae (13-126)

1G)
2ai fren?
RET (13-127)
Oba
For an infinite guide height a, only the second term remains in the
numerator and

ee ae)
No \?

"BOA"? fy Gey fe
(a «). (13-128)
i 2b
This term comes from W,,, and the losses then occur only on the guide faces
paraliel to the xz-plane. It is shown in Problem 13-22 that this value of k,; can
be accounted for entirely by the reflection losses.
In practice, the ratio 2a/b is close to unity, whereas (\o/2)? is of the
order of 1/2. The losses on the faces parallel to xz-plane are thus of the same
order of magnitude as those on the other pair of faces, but smaller by a factor
of about 2.
Figure 13-10 shows k,;b?/? as a function of the ratio \o/A, = do/2b for
582 PROPAGATION OF ELECTROMAGNETIC WAVES III

Table 13-1. Characteristics of a Few Types of Rectangular Wave Guides

(TE mode with n = 1)

Inside Cut-off Operating Attenuation Power

Dimensions Frequency Range Rating

inches gigahertz db/100 feet megawatts

2.84 X 1.34 2.080 2.60- 3.95 1.10—0.75 2.2-3.2

1.872 < 6.872 J155 3.95— 5.85 2.08-1.44 1.4-2.0
13720622 4,285 5.85— 8.20 2.87—2.30 0.6—0.7
0.900 X 0.400 6.56 8.20-12.4 6.45—4.48 0.2-0.3
0.622 X 0.311 9.49 12.4 -18.0 9.51-8.31 0.1-0.2

2a/b equal to unity. According to this curve, the optimum value of \»/26 is
about 0.4, but the minimum is very broad; actual values of \o/26 are larger so
as to achieve strong attenuation for the n = 2 mode.
The attenuation is of the order of a tenth of a decibel per meter at
frequencies of a few gigahertz, increasing as f?/? when the ratios a/b and
do/2b are kept constant (Problem 13-21).
Table 13-1 shows the characteristics of a few types of hollow rectangular
wave guides.

13.4 SUMMARY

In the general case of a sinusoidal electromagnetic wave propagating in the

positive direction of the z-axis,
E = (Eost + Eq j + Eok) exp f(wt — k,z) = Ey exp j(wt — k,z) (13-1)
A = (Hot + Ho,j + Hok) exp (wt — kz) = Hy exp fwt — kyz) (13-2)
We have shown that, if the radian length X, of the guided wave is not
equal to the radian length X of a plane wave in the medium of propagation,
then
_ Jou K, OE: 0H):
Foz = ifn: all
li(=: ax ay ) (13-12)

x2 x?

vad Jou 7 Ke OF): OM):

Evy ia il 5 sal Ww oy + ax b) (13-13)

mi
Jwe OE). k, 0H.
Ay0 = ———
i i |
(—ay — ree
2 —),) 5
(13-14)
13.4 SUMMARY 583

figs —Jwe (= k, on
Ox we Oy Us2l>)

The vectors E and H can therefore be calculated once Ep, and Hy, are known.
If Ey. = Owe have a TE wave, and if Hy. = Owe haveaTM wave. If both
Eo, and Ho, are zero, then we have a TEM wave and Agar iNe
For TE and TM waves we calculate Ep. and Ho. using the wave equations
and the boundary conditions imposed by the guide. The wave equations
for Ey, and Hp, are similar; that for Ep, is the following:

O*Ens | O° Ey, [1 1
axe a ay? (5 = zs)Ev. (13-18)
In TEM waves,
OV 0A
oe oT. (13-31)
and
LS ae oe : ae
E e Poe ay i)exp j (ot ‘) (13-37)

= Ey exp j (ota ‘): (13-38)

The field Ey can be derived from the potential V, in exactly the same manner as
for an electrostatic field. Then, inside a hollow perfectly conducting tube, Eo
must be zero and TEM waves are impossible, except as a limiting case of TE
and TM waves at very short wavelengths.
The transverse components of E and H are mutually perpendicular,
whether we have TE, TM, or TEM waves. The wave impedance 1s

AB
ao (pape i" for TE waves,
(‘) (13-23)

: (2) ‘for TM waves,

mM 1/2 X

€ Na
(13-25)
1/2
= (*) for TEM waves. (13-46)
€

The boundary conditions at the surface of a perfectly conducting wave

guide are as follows. (a) For any type of wave, the tangential component of
E is zero. (b) For any type of wave, the H vector is tangent to the surface,
perpendicular to the current density in the guide, and numerically equal to the
surface current density expressed in amperes/meter. (c) For TE waves, V Ho:
is tangent to the wall, and the rate of change of Mp; in the direction normal to
the surface is zero.
584 PROPAGATION OF ELECTROMAGNETIC WAVES III

To illustrate how one can deal with guided waves, we studied two
common types, namely, the TEM wave in a coaxial line and the TE wave with
= | in a rectangular wave guide. In a coaxial line,

(nen p
ees (ot= 5)
A
re (13-58)
1/2
H = (<) ay (or= 5)01, (13-60)
Ho p x

and the average transmitted power is

Wr = 120
— in ©?
ero,
watts (cr = 1, ur = 1). (13-65)

For the TE wave in a rectangular wave guide, we assumed a plane wave

zigzagging down the guide, as in Figure 13-6. We then found H,. from the
wave equation and from the required boundary conditions. The other
components followed immediately. The six components are shown in Eqs.
13-88 to 13-94 for n = 1. This parameter n is the number of “‘half cycles of E”’
in the guide as in Figure 13-7.
Rectangular wave guides are high-frequency devices, the cut-off wave
length \, being equal to twice the distance between the reflecting sides. The
phase velocity w, 1s larger than c, whereas the signal velocity u, is smaller, and
Ups = C?.
The average transmitted power is
2 2) 1/2
Wie =e {1es(33) } (13-109)
where £oo; is the maximum value of £ in the guide and the lengths a and b are
as in Figure 13-5,
It is possible to take into account the finite conductivity of the guide by
deducing the tangential E in the guide from the value of the tangential H
calculated on the assumption of infinite conductivity as above. This leads to a
Poynting vector that points into the guide material and that gives the Joule
losses. The result is that the imagninary part of the wave number is

ett 3B
Mo = GI200n)" fay"
2a No 2

(13-127)
i} 2b J
It corresponds to losses in all four faces.
 585

PROBLEMS

13-1. The transverse part of the vector E in a guided wave is written as a vector E,,.
What can you say about the orientation of this vector in space, as a func-
tion of the four variables x, y, z, r, in the case of TE waves? What if E,/Eoz
is not real?

13-2. Sketch a rather large cross-sectional view of a coaxial line in a plane contain-
ing the axis.
(a) Show lines of E and of H at a given instant over at least one wavelength.
The lines should be most closely spaced where the field is strongest. Indicate
the directions of the fields by means of arrow heads. Show the direction of
propagation.
(b) Add arrows at various points to represent Poynting vectors, using longer
arrows where the power flow is larger. Assume that the length of the arrow
represents the magnitude of the Poynting vector at its midpoint.
(c) How does the pattern change with time?
(d) Sketch a cross-sectional view of the coaxial line in a plane perpendicular
to the axis and show lines of E and of H at a particular instant.
Relate this plane to the figure you drew under (a) above.
(e) Explain how this pattern changes with time.
(f) Add plus and minus signs to both figures to show the surface charges.
The spacing between the signs should indicate qualitatively the relative mag-
nitude of the surface charge density.
(g) How does the charge pattern change with time at a given z?
(h) Now add arrows of various lengths to your first figure to represent surface
current densities.
(i) How does the current pattern change with time at a given z?

13-3. Show that, in the case of an idealized coaxial line of infinite conductivity, the
current is given by the linear charge density multiplied by the velocity of
propagation.

13-4. It is known, from transmission-line theory, that the characteristic impedance

of a line is given by
Lt 1/2
Ls — (3) ’

where L’ and C’ are, respectively, the inductance and capacitance per unit
length.
Show that this is correct in the case of the coaxial line, by calculating L’
and C’, and then comparing with the value of the characteristic impedance
given in Section 13.2.

. (a) If the maximum allowed field strength in a coaxial line is E,,, show that
the maximum allowed voltage is
Va, = pug Bol Po/pi
586

(b)
Figure 13-11.

(b) Show that, for a given value of p., V,, is greatest when p,/p; is equal to e.
(c) Show that the characteristic impedance is then 60 ohms if the line is air-
insulated.
(d) Show that, under those conditions, the maximum allowable current is
PoEm/163 ampere.

13-6. (a) Setting E,, to be the maximum allowed field strength in a coaxial line,
use the result of Problem 13-5 to show that the maximum allowed power is

ES °° 120 ee
(p/p)?es
Ex, 1 (p0/pi)

if the line is air-insulated.

(b) Show that, for a given value of p,, this power is greatest when p,/p; =
RGSS
(c) Show that this ratio corresponds to a characteristic impedance of 30
ohms if the line is air-insulated.

13-7. One important parameter of a TEM transmission line is the characteristic

impedance.
ae 1/2

z.=(G)"
where L’ and C’ are, respectively, the inductance and the capacitance per unit
length of the line.
In designing such lines it is therefore essential to predict the values of L’
and C’. If the geometry is such that L’ and C’ cannot be calculated analytically,
as, for example, in Figure 13-11a, it is sometimes useful to perform the follow-
ing measurements on a resistance-sheet analog.*
(a) We can find the value of C’ by cutting out a sheet of resistive material
in the shape of the cross section of the dielectric as in Figure 13-11b, and
measuring the resistance between electrodes 4 and B.

*D. J. Epstein, Comment on a Theorem in the Field of Steady Current Flow, Proc.
IEEE, 56, 198 (1968).
 587

Figure 13-12.

Show that, if the material has a conductivity o and a thickness s, and if

the permittivity of the dielectric is «, then
€
Ric =—,
So

where R; is the resistance between A and B.t

(b) To measure L’ we use a similar sheet with a radial cut and with electrodes
C and D as in Figure 13-11c.
Show that, if uw is the permeability of the dielectric,

Riles
So

where R2 is the resistance between C and D.,

Thus
ie bi 1/2 Ry

ae @ Ry
See also Problem 10-14.t

. Figure 13-12 shows a cross section of a microstrip line. The lower electrode
is grounded and the wave is TEM.
The microstrip has the advantage of being much less costly than either
the coaxial line or the rectangular wave guide. It is also particularly convenient
for conveying high-frequency signals within printed and integrated circuits.
Its main disadvantage lies in the fact that its field is not limited to the region
immediately below the upper electrode. Microstrip lines can therefore interact
with other elements in a circuit, unless they are either spaced or shielded
properly. Shielding can be achieved by adding a second grounded plane above
the strip, but this, of course, increases the complexity and cost of the line.
(a) Sketch lines of E and of H.
Use arrows to show the directions of E and H at a given time. Show the
direction of propagation.
(b) In practice, the width 4 ofthe strip is much larger than its distance / to the
grounded plane, and edge effects are small.
Show that the instantaneous value / of the transmitted power is equal
to the Poynting vector integrated over the cross section bh.
(c) Show that the characteristic impedance U/J is
p\V2h
(Z) b
(d) Show that one arrives at the same result if one defines the characteristic
impedance as in Problems 13-4 or 13-7.
(e) Show that the addition of a second grounded plane placed symmetrically
with the first reduces the characteristic impedance by a factor of two.
588 PROPAGATION OF ELECTROMAGNETIC WAVES III

13-9. Figure 13-8 shows schematically the E and H fields inside a rectangular wave
guide carrying a TE (n = 1) wave.
Draw sketches as in Problem 13-2 showing Poynting vectors E < H, sur-
face charge densities, and surface current densities.
To represent properly these last two quantities you will have to sketch
two or three neighboring faces. Be sure to relate these faces to those shown
in Figure 13-8.

13-10. If a wave guide is not properly connected to its load, part of the incident
wave is reflected back towards the source and a standing wave is established
along the line. Under such conditions the power fed to the load can become
negligible.
It is therefore useful to be able to move a small probe along a longitu-
dinal slot to sample the field inside the guide. The quantity that is usually
measured is the Voltage Standing Wave Ratio (VSWR). This is the ratio of the
maximum to the minimum time-averaged voltage picked up by the probe as it
slides along the slot. Under ideal conditions of power transfer, there is no
reflected wave and the VSWR is equal to unity.
The probe can be either a small loop that is coupled to the magnetic field,
or a short length of wire that responds to the electric field. In both cases the
projection into the guide is approximately one millimeter or less.
(a) In the case of a rectangular hollow wave guide and a TE mode, as in
Figure 13-8, where should the slot be cut? The slot should, of course, disturb
the wave as little as possible. Sketch a perspective view of the guide, showing
both the EF vector and the slot.
(b) If the probe is a small loop, how should it be oriented with respect to the
guide?
(c) How would you use the movable probe to measure ),?

13-11. Ordinary wave guides are brass or copper tubes with wall thicknesses of the
order of one or two millimeters. However, some microwave structures are
quite complex and would be too expensive to fabricate either from tubing or
from solid stock. Several techniques have therefore been developed for such
cases.
One technique, called e/ectroforming, utilizes a former, in the shape if the
internal volume required, which is electroplated to a thickness of a few milli-
meters. The former is then removed, leaving the required structure.
In some cases the former is a plastic and is left in place. The resulting
structure is thus dielectric-filled and is smaller (Problem 13-13), for a given
operating frequency, than if it were air-filled.
In this latter case, what minimum thickness of copper would you recom-
mend if the operating frequency is to be 3 gigahertz?
13-12. It is found that an electromagnetic wave propagating in the TE mode with
n = 1 (Section 13.3.1) through the 2.84 inch X 1.34 inch wave guide of Table
13-1 has a X, of 13.8 centimeters.
Calculate its frequency.
13-13. (a) Show that, if a hollow rectangular wave guide is completely filled with a
PROBLEMS 589

dielectric of relative permeability ¢,, its cut-off frequency is /ower than if it

were empty by a factor of &”:
1
Soa = =73ev»
€,

where the subscripts d and v refer, respectively, to ‘“‘dielectric” and to “‘vac-

uum.”
Thus, for a given operating frequency, a dielectric-filled wave guide is
smaller than an air-filled one. For example, if the dielectric is Teflon (e, = 2.1),
the cross-sectional dimensions a and b of Figure 13-5 are both reduced by a
factor of 1.45.
(b) Consider two wave guides, one air-insulated and with a cross section a, b,
and another that is dielectric-insulated and that has a cross-section a/e!/?,
b/é”.
Show that, for a given frequency,

Nod = =75 Nov:

€,

13-14. In Section 13.3.1 we showed that

1 1 ane
Sheei old g

for a TE wave propagating through a rectangular wave guide, as in Figure

13-6.
Show that, in the more general case of any TE or TM wave in a rectan-
gular wave guide,

where m is the number of half cycles of E or of Hin the direction of the x-axis,
and similarly for n.

13-15. A plane electromagnetic wave is incident at an angle on a flat perfectly con-

ducting surface, and E is normal to the plane of incidence.
(a) Draw carefully a set of equally spaced parallel lines representing the
*“‘crests”’ of E in the incident wave at a given instant, and draw broken lines
for the “troughs.” Draw similar lines for the reflected wave.
(b) Where is E always equal to zero?
(c) Can you relate this pattern to the n = 1, 2, etc., modes in a rectangular
wave guide?

13-16. (a) Sketch a graph of \o/A, as a function of Ao/A, for a rectangular wave
guide.
In actual practice this curve is meaningful only for \o/A, ~ 0.7. At much
smaller values of \o, higher modes of propagation (n = 2,n = 3, etc.), are pos-
sible, and the field configuration inside the guide becomes uncertain. At much
larger values of \o the attenuation becomes excessive.
(b) Sketch a curve of w as a function of k,.
590 PROPAGATION OF ELECTROMAGNETIC WAVES III

For a given operating point, the phase velocity u, is given by the ratio
w/Kg.
(c) Verify that the signal velocity is given by

Us
a ae ee
~ dk,/do dk
and thus by the slope of this curve.

13-17. Check the power rating of the largest rectangular wave guide shown in
Table 13-1. Assume that the operating frequency is 1.5 times the cut-off
frequency.
The maximum permissible electric field intensity in air is 3 x 108
volts/meter. Allow a factor of safety of 2 to take into account the effects of
irregularities in the inner surface of the guide.

13-18. A small single-turn loop of area A is situated at x = 0, y = 4/2, just inside

the hollow rectangular wave guide of Figure 13-5. The plane of the loop is
parallel to the x-z plane, and the field inside the guide is as illustrated in
Figure 13-8.
Show that the power flowing through the guide is given by
gab 40k
Wr = 3.36 X 10%

where U is the rms electromotance induced in the loop.

13-19. It is suggested that one could measure the power transmitted by a rectangular
wave guide by observing either the deflection or the energy gain of a beam of
electrons crossing the guide.
Discuss the posibility of such measurements.
13-20. (a) Show that the average force per unit area exerted by the field on the face
x = 0 of the rectangular wave guide shown in Figure 13-8 is given by
9 °
_ €9E oor Xo \? s, 2ry
Koehn ge 1 cops 20s tan
(b) Calculate F, for the first guide listed in Table 13-1 at the maximum power
rating and at a frequency of 3.00 gigahertz.
(c) Is it important to take this force into account in the design of rectangular
wave guides?
(d) Could this force be used to measure the transmitted power?

13-21. Show that the attenuation constant k,,; fora TE(n = 1) wave ina rectangular
wave guide, as given in Eq. 13-127, varies as f*/? when the ratios a/b and \o/2b
are kept constant.

13-22. (a) Show that an electromagnetic wave is reduced in amplitude by a factor of

approximately
1/2
i (=) cos 6;
oO
PROBLEMS 591

upon reflection from a good conductor of conductivity «. The angle of inci-

dence is 6;, the dielectric is air, and the E vector is normal to the plane of
incidence.
(b) Show that this loss leads to the attenuation constant of Eq. 13-128 for
rectangular wave guides.
13-23. (a) What is the surface current density in the face y = 0 of the wave guide
illustrated in Figure 13-5 when it carries a TE(n = 1) wave?
(b) Show that the power loss in the wall is the same as if the current were
uniformly distributed in a thickness 6 near the surface, 6 being the attenuation
distance or skin depth (Section 11.5).
This result is valid for curved surfaces, as long as the radius of curvature
is much larger than the skin depth.
The quantity 1/o6 is called the surface resistivity and is expressed in
ohms/square. A square sheet of given surface resistivity and of any size has
a resistance equal to the surface resistivity between two parallel edges. See
also Problems 11-20 and 12-22.

13-24. Use the result of Problem 11-20 to verify Eq. 13-125.

13-25. The degree of attenuation in a line can be expressed in nepers/meter, the

number of nepers/meter being simply the numerical value of the attenuation
constant k,; in MKSA units.
The degree of attenuation is also expressed in decibels/100 feet. The num-
ber of decibels/100 feet is 20 times the logarithm to the base 10 of the ratio
of the voltages (or the currents) at the two ends of a line 100 feet long.
Show that one neper/meter is equivalent to 264 decibels/100 feet.

13-26. (a) Check the attenuation of the largest type of rectangular wave guide
listed in Table 13-1. Assume that the guide material is copper and that the
operating frequency is 1.5 times the cut-off frequency.
See Problem 13-25 for transforming nepers/meter to decibels/100 feet.
(b) Calculate the power dissipated in the guide per meter of length at its maxi-
mum power rating of 2800 kilowatts.
(c) Calculate the rate at which the temperature of the guide would increase,
at a maximum power rating, if it were thermally insulated. The walls have a
thickness of 0.080 inch.
(d) How should the guide walls be cooled?
Would it be practical to transmit large amounts of electric energy over
long distances in this way ?
Wave guides are often used to transmit large amounts of power at low
duty cycles. The average power dissipation is then much lower than in the
above example.

13-27. Tabulate the values of the following quantities for the five types of rectang-
ular wave guide of Table 13-1:

(a) b/a;
(b) Vea fants :
(c) 0/26 at both limits of the operating range;
592 PROPAGATION OF ELECTROMAGNETIC WAVES III

(d) the attenuation in nepers/meter (see Problem 13-25) for the n = 2 mode
at both limits of the operating frequency.

13-28. (a) Weare required to calculate the maximum power that can be transmitted
at a frequency of 3.00 gigahertz through a short length of (a) coaxial line, and
(b) rectangular wave guide.
The coaxial line has a diameter 2p, of 2.50 centimeters and the wave guide
has an inside cross section of 3.75 & 7.50 centimeters.
The coaxial line satisfies the condition for maximum power transfer,
namely p./p; = 1.65 (Problem 13-6). The radius py is chosen to ensure
attenuation of higher order modes.*
In both cases the dielectric is air and the current-carrying surfaces are
silver plated. The maximum allowed electric field intensity is 1.5 « 108
volts/meter as in Problem 13-17.
It is assumed that there is no reflected wave.
(b) We are also required to calculate the power lost per meter of length in
both cases. See Problem 13-29a.
(c) Finally, we wish to know the rms voltage and current at the input end of
the coaxial line.

13-29, In calculating the resistance of a coaxial line it is correct to assume that the
current is uniformly distributed throughout a thickness equal to the skin depth
6 on both conductors (see Problem 13-23).
(a) Show that the resistance of a coaxial line per unit length is
1 1 1
on 216 Do + *),

where o is the conductivity of the material.

(b) Use the method of Section 13.3.4 to show that, for a coaxial line,
V21 + (po pi)
koi = 2.64 X 10-8 (! )
oO Po In (po/ pi)
(c) Show that, for a given value of p,, ky; is minimum when p,/p; = 3.6.
(d) Show that the characteristic impedance is then 77 ohms.

13-30. (a) Calculate the power rating of an air-insulated copper coaxial line whose
outer conductor has an inside diameter 2, of 2.84 inches, when the ratio
Po/pi 1S Selected as in Problem 13-6 for maximum power transfer. The maxi-
mum electric field intensity in air is 3 X 10° volts/meter. Allow a factor of
safety of 2 to take into account irregularities in the surfaces of the conductors.
(b) Use the results of Problem 13-29 to calculate the power dissipation in both
the inner and outer conductors per meter of length at the power calculated
under (a) above, and at a frequency of one gigahertz.

13-31. Let us calculate the attenuation constant k,; for a coaxial line, taking into
account energy dissipation in the dielectric.

* See, for example, A. F. Harvey, Microwave Engineering, Academic Press, New York,
1963, page 20.
PROBLEMS
593

Since the dielectric losses are normally small, we shall simply add their
attenuation constant to that associated with resistive losses in the conductors
(Problem 13-29) to obtain the value of k,: for the line.
(a) A dielectric is said to be lossy when it dissipates energy in the course of
the polarization process, or when it is slightly conducting.
Consider first a parallel-plate capacitor containing a lossy dielectric. The
plates have an area A and are separated by a distance s. An alternating voltage
voltage U is applied across the plates.
You should be able to show that, if there is energy dissipation in the
dielectric, then one can either say that the dielectric has a conductivity o, or
that its relative permittivity is of the form

ae > Jess

where e; and ¢/’ are positive real quantities, with

T = Wee, .
Energy dissipation in dielectrics is used extensively for welding plastics,
gluing wood, heating food, etc.
(b) The calculation of k, with a lossy dielectric is formally the same as with a
loss-less medium, except that e, is complex.
Show that the part of k,; that is due to the dielectric is
1 el! ) ay

ken e or aA (2

Thus, taking into account the losses in both the conductors and the dielec-
tric, and using the result of Problem 13-29, the attenuation constant for a
coaxial line is
= zy il = (po/ pi) €.
ke a= 2.64
2.64 X 1 0-8 ((+ ee
Palio sim MMOs) 5 110) —8 Seyi
GE meters =

where a is the conductivity of the conductors, and where e¢} — je!’ is the rela-
tive permittivity of the dielectric.
13-32. A gas lens utilizes the fact that the index of refraction of a gas is dependent
on its density, and therefore on its temperature. Thus, cool gas blown gently
through a heated tube becomes hotter and lighter near the periphery than near
the axis and acts as a weak converging lens.
After a certain distance the radial temperature gradient becomes negli-
gible and, if further convergence is required, the hot gas must be evacuated
and fresh cool gas injected in its place. Another possibility is to use the same
gas flowing through a succession of alternately hot and cold tubes. Then the
hot tubes act as focusing elements, while the cold tubes act as defocusing
elements, and, under proper conditions, there is a net focusing effect. This
technique is known as alternate gradient focusing and was developed for par-
ticle accelerators.
Such beam wave guides can be used for transmitting modulated light
beams carrying, for example, telephone messages.
The index of refraction is related to the density D and to the temperature
T as follows:
594 PROPAGATION OF ELECTROMAGNETIC WAVES III

p= i DRS
no — 1 ID; tan:

Use the ray equation given in Problem 12-2 to show that

where p is the radial position of a ray at a distance z along the guide.

Assume that
oO mS oO

Os” Oz
and neglect the term
Op On
Oz Oz
Since 77 is very close to unity,

d’p _ an
dz*~ dp
Then
ap Ty) OT
det ad (1 1) T? dp

and, finally, if we choose the reference temperature Jy to be the average tem-

perature, we can write that
d’p x 11g) = 1 an

dz? mory To dp.

This equation can serve as a starting point for the study of such beam
wave guides. Since the right-hand side is a complicated function of both p
and z, the equation must be solved numerically.

Hints

13-7. (a) See Problem 4-7.

(b) See Problem 9-18.

CHAPTER l |

RADIATION OF
ELECTROMAGNETIC WAVES

We have studied the propagation of electromagnetic waves in considerable

detail. In Chapters 11 through 13 we have studied successively their propaga-
tion in free space and in various media, across an interface, and then along
various guiding structures. The present chapter will now be devoted to the
processes whereby these waves are produced. It may not seem logical to
proceed in this order, but the reason is one of convenience and will become
apparent after a while: the phenomenon of radiation is rather complex, and
its discussion was best delayed until now. Indeed this chapter consists of a
series of examples illustrating concepts we have developed in the preceding
chapters.
We shall start with the electric dipole of Figure 14-1, which is the simplest
type of source. Once we have mastered the electric dipole, we shall be able
to study the radiation fields of the half-wave antenna and of antenna arrays.
We shall then go on to magnetic dipole radiation, which is closely related to
electric dipole radiation. After that it will be relatively easy to deduce the
radiation fields of simple electric and magnetic quadrupoles. Finally, we
shall prove a reciprocity theorem that is equally valid for antennas and for
electric circuits.

(AnlerLEG RIC DIPOLE RADIATION

It will be recalled from Section 2.9 and from the example on page 428 that an
electric dipole is formed of a pair of charges of equal magnitude and of
opposite signs, as in Figure 14.1. Its dipole moment p is Qs.
596

Figure 14-1. An electric dipole. The

total charge is zero and the vector s is
oriented from —Q to +Q.

In the oscillating electric dipole,

= Oe?" (14-1)

p = Os = Qyse*' = pre" (14-2)

and an alternating current

= Ihe#'t = jw Qoe™ (14-3)

flows in the wire connecting the two charges as in Figure 14-2.

+Q

-0

Figure 14-2. The charges —Q and +Q and the current J as

functions of time in the oscillating dipole.
 597

Figure 14-3. Electric dipole

antenna fed by an oscillator.

We shall use this simple model, but electric dipole radiation is also produced
by any charge distribution whose dipole moment

p= | eras (14-4)

is a sinusoidal function of the time. For example, charge oscillations can be

imagined to occur on the surface of a conducting sphere in such a way that
the electrons are driven alternately from one pole to the other. Figure 14-3
shows schematically an electric dipole antenna fed by an oscillator.
We shall calculate successively the potentials V and A, then the field
vectors E and H, and finally the average Poynting vector $,, and the radiated
power W.

(7-171 she Scalar Potential, V

We have already found in the example on page 428 the retarded V of the
oscillating electric dipole in free space fors<«Xands <r:
_ po exp jolt] € 5 ‘
V= Sie ee iy + j)} cos 6, (14-5)

where r and 6 are as in Figure 14-4 and [f] is the retarded time t — (r/c).
For a nonzero frequency, the exponential term appears to show that V
propagates as a wave at a phase velocity c. This is not quite correct, because
of the complex factor (A/r) + j. Rewriting the complex factor in exponential
form,

= rene é + 1)" exp {ja — :+ sarc tan :) cos 6. (14-6)

For r>>X, arc tan (r/A) & 1/2 and is approximately independent of r. The
phase velocity is then c. However, closer in to the dipole, where r is not much
larger than 4, arc tan (r/A) is not constant, and the effect of this term is to
give a phase velocity that is Jarger than c.
The scalar potential V varies as cos @ and is zero in the equatorial plane,
where the fields of the two charges cancel exactly, just as in electrostatics.
It varies as 1/r? as in the static case, but only as 1/r when r > 4. Also, for
598

Figure 14-4. An oscillating electric

dipole situated at the origin gives rise
Zz to a scalar potential V at the point
as y P(r, 6, ¢).

r >> X, V varies as 1/X and is thus proportional to the frequency. Figure 14-5
shows a radial plot of V as a function of @ and ¢.

14.1.2 The Vector Potential A and

the Magnetic Field Intensity H
We shall calculate the vector potential A at the point P(r, @, ¢) and then the H
vector from the curl of A. We assume again that the dipole is in free space,
that s<vr, and that s<i. The last condition eliminates standing-wave
effects in the connecting wire, and the current is then the same throughout the
length s of the dipole. We have already used these approximations in calcu-
lating V. We again make no assumption as to the relative magnitudes of r
and A.
The vector potential A is related to the current J in the dipole as in Eq.
10-36 and Figure 14-5 and 6:

A=ey a
eel I, expiD jw[t]
ij s.: :
(14-7)

It is independent of both @ and ¢, and it depends only on the distance r to the

dipole. It is also everywhere parallel to s, and thus parallel to the polar axis.
Expressing s in polar coordinates,

A= a ys (ex
jw[ t})(cos
p 6 r; — sin 6 @;) (14-8)

= EEP* (ex
jel#]\(cos
p @r: — sin 6 61) (14-9)
where the polar unit vectors ri, 0, g: are as in Figure 1-19. The vector po-
tential A propagates at a velocity c, even for r< i.
 599

Figure 14-5. The scalar potential V and the magnitude of the vector potential A
are shown here as functions of 6 and ¢ about an oscillating electric dipole ori-
ented as shown. The radial distance from the center of the dipole to the spheres
marked V is proportional to the value of V in that particular direction. The scalar
potential V is maximum at the poles; it vanishes at the equator, where the indi-
vidual potentials of the charge — QO and +0 of the dipole cancel. It is positive
in the northern hemisphere, where the field of +@Q is predominant, and negative
in the southern hemisphere. The magnitude of A is similarly represented by the
sphere marked A. The vector potential is independent, both in magnitude and in
direction, of the coordinates @ and ¢.

We should really have integrated da over the length s of the dipole.

However, it will be shown in Problem 14-1 that the integration leads to the
above result when s « i. The situation here is different from what it was for
V, because V is the difference between the scalar potentials of the two charges
— Q and+ Q. Since the two scalar potentials,are very nearly equal in
magnitude, their difference must be evaluated with care. For A, however,
the da’s all add up, and the phase angle w{t — (r/c)}, or wt — r/X, changes
but slightly from one element to the next, since s < i.
600

P(r,0,~)

Figure 14-6. An oscillating elec-

tric dipole of moment p situated
y at the origin gives rise to a vector
potential 4 parallel to p at the
point P(r, 6, ¢).

Thus
l 1 fa 0A,

The two other components of the curl are zero because A has no ¢ component,
and because it is not a function of ¢. Thus

Re (RING
herf (exp jolt)(7 +i) sin 61. (14-11)
As usual, we compare this new result with a previously acquired one,
namely the Biot-Savart law for steady currents, as stated in Eq. 7-8. According
to this equation, the magnetic field intensity due to an element of current /s
at the position of the dipole is

dSOS Ti. 1S SIU

hear elas PPh (w = 0). (14-12)

Thus, at zero frequency (w = 0, \— ©), the two results agree.

For a nonzero frequency, H propagates at the velocity c when r> Xi.
Closer in, the phase velocity is larger than c, as it is for V.
It will be observed that, at zero frequency, the last parenthesis of Eq.
14-11 reduces to the first term, \/r, and H varies as 1/7”.
The j term in the last parenthesis of Eq. 14-11 decreases only as 1/r and
becomes predominant at large distances from the dipole where r >> X. This is
the radiation term.
It is interesting to note how the amplitude of H changes from a 1/?r*
dependence for a constant current to a 1/r dependence for a varying current.
We have already observed a similar behavior for V.
14.1 ELECTRIC DIPOLE RADIATION 601

Since both parts of H involve sin 6, they are both maximum in the
equatorial plane and zero along the axis of the dipole. The magnitude ofH as
a function of @ and ¢ is plotted in Figure 14-8.
Since [hs = jwpo,

== 1)sin 6 1.
ee=va(exp jol)( i> (14-13)
For r> i,
Pave me exp jw[t] sin 0 gi (PA). (14-14)

14.1.3 The Electric Field Intensity E

To obtain E, we require the time derivative of A,

a ge= (exp jw[t])(cos 6 r; — sin 6 6), (14-15)

and the gradient of the scalar potential of Eq. 14-6,

OV 1 oV OV
— VV = — {-—
(Son + = ae +35 ow), (14-16 )

—Po 3
GiregrX? exp jolt]

IN nA Kr? I\ <6
xX \(-25 — jot 1)cos 6r; — 5 +] *)sin 604) (14-17)

Then

= = ri exp jolt]

oe Ae \|
1)| cos 6
x/{(-0+ (-25,- yt
2
+44) + (-4
9 — )}sinoa (14-19)

At distances where r >> X, the terms in /r and %*/r? become negligible. In

the coefficient of cos @ the first and last terms are the largest, but they cancel.
Finally,

Po ; MEER Aa srw! ; NGS ) Ware

eer (exp iol) 2(°5sil eos Or; ae (* l SWIG (14-20)
602

Figure 14-7. The E, H, § vectors

for an oscillating dipole are ori-
ented as shown at r > A when
the phase angle w(t] is zero.

Contrary to the H vector, which is entirely azimuthal, and thus transverse or

perpendicular to the direction of propagation, EF has a longitudinal component
in the radial direction, at least close to the dipole where r > X.
We can compare this E with that of a static electric dipole, as given in
Section 2.9. For X > «, Eq. 14-20 becomes

E = —?°_ (2 cos6r + sin

4reor?
00;) (w = 0), (14-21)
as required. These are the so-called static terms.
The electric field intensity E propagates through space with a velocity c
for 7 >A. as do VA, A;
Close to the dipole, where r is not much larger than X, E involves five
terms. Two of these vary as 1/r*, two others lead them by 7/2 radians and vary
as 1/r?, and a fifth term varies as 1/r and leads the first pair by m radians.
Far from the source where r > X, we are left with the radiation term

es Ly Po A : :
is ara exp jw[t] sin 6 0, (r > 2X). (14-22)

The magnitude of E at r > X is plotted as a function of 6 and ¢ in Figure 14-

8.
In this case E changes from a 1/r* dependence for a static field to a 1/r
dependence for a varying field. We shall see later on that a 1/r dependence is
14.1 ELECTRIC DIPOLE RADIATION 603

required of both E and H for an oscillating dipole to ensure conservation of

energy.
If we consider both E and H far from the source where r > X, we notice
that E lies in a plane passing through the polar axis, whereas H is azimuthal.
The Poynting vector E X H is radial, as in Figure 14-7. We shall calculate the
Poynting vector in the next section.
The ratio

ahs (#)" = 377 ohms (r >), (14-23)

just as for a plane wave in free space (Section 11.1). The electric and magnet
energy densities eE?/2 and woH?/2 are equal.
We have therefore discovered tha

14.1.4 The Average Poynting Vector

and the Radiated Power
It is interesting to calculate the power radiated by an oscillating electric
dipole. This can be found by integrating the average Poynting vector S,, over
a spherical surface centered on the dipole. Let us first calculate $,,:

Say = ;Re(E X H*), (14-24)

me 5Re {(E, 1; + Ey 61) X Hé gi}. (14-25)

Recalling that
mnxX = — GW, aixXa=n, (14-26)

Sav = 5Re (— E,HZ 0, + EoHe ri), (14-27)

the components of E and of H being given by Eqs. 14-20 and 14-13. Thus
Ps ADD)
_ Po sin’0 14-28
ay 32 negrAt ( )

If we wish to express the average Poynting vector in terms of the current

Ip, we must substitute pop} for ps, and then
Be, sin? 0 r4.
ae Serr (14-29)

One striking feature of the Poynting vector is that it involves only the
604

Figure 14-8. Polar diagrams of sin @ (outer surface) and of sin? @ (inner surface)
showing, respectively, the angular distributions of E or H, and of 8,, at a distance
r ><X of an oscillating electric dipole situated at the origin. The radial distance
from the dipole to one of the surfaces is proportional to the magnitude of these
quantities in the corresponding direction. Most of the energy is radiated near the
equatorial plane; none is radiated along the axis.

radiation terms, despite the fact that our calculation is valid even near the
dipole, where r is not much larger than X. In fact, it could have been calculated
correctly by disregarding the terms that are unimportant far away from the
dipole. The energy flow is everywhere purely radial, at least as long as the
dipole length s is small compared to both r and X.
The Poynting vector varies as |/r?. This is required for conservation of
energy since, under steady state conditions, the energy flow through any
given solid angle must be the same for all 7. This 1/r? dependence results from
the fact that the radiation terms for E and for H both vary as 1/r.
Since the energy flow varies as sin? 6, it is zero along the axis of the dipole
and maximum in the equatorial plane, as in Figure 14-8. An electric dipole
does not radiate energy along its axis.
The total radiated power W is obtained by integrating $,,. over the
surface of a sphere of radius r:
14.1. ELECTRIC DIPOLE RADIATION 605

W = oPoxdiie "sin6ddde, (14-30)

Cc

pe Tie = (14-31)

SOND, HHO,Ls (14-32)

The radiated energy varies as the square of the dipole moment py) = Qos, and
inversely as the fourth power of the wavelength, or directly as the fourth power
of the frequency.
When the average Poynting vector is expressed in terms of /,

7 _ Hoe (3)
(S\* R,22 (14-33)
2

00 (:) 2 watts, (14-34)

7

00 (;) Pang watts. (14-35)

It will be observed that the energy radiated by the electric dipole is
proportional to the square of the current flowing through it. The coefficient of
Tims is called the radiation resistance:
2
Rowe 20.0 (;) ohms. (14-36)

This is the resistance that would dissipate in the form of heat the same power
that the dipole radiates in the form of an electromagnetic wave, if it carried the
same current. It will be recalled that we have assumed that s < X.

14.1.5 The Electric and Magnetic Lines of Force

We have already found E in Eq. 14-20. To find the electric lines of force, we
can proceed as in Section 2.9 and set
E Es
=, 14-37
dr rdé ( )

since an element of a line of force, having components dr and r dé, is parallel

to the local E, whose components are E£, and Ey. The calculation is consid-
erably simplified by using the vector
A :
=a +i) sin 0 gy.
tad.
(exp jolt ne -38
(14-38)
606

HOS Or=0 ot ==>

2

co)
Figure 14-9. The electric lines of force of an oscillating dipole for wt = 0, 7/2,
mw, and 32/2. The dipole is situated in the center and is oriented in the vertical
direction. The decrease in wavelength with distance can be observed on these
figures. The magnetic lines of torce are circles perpendicular to the paper and
centered on the axis of the dipole.

Then
E=VXC (14-39)
ee ae
= rsing do. (14-40)

B=_ —kaye) (14-41)

The differential equation for the lines of force is therefore

ic) : ie
Hee eae (14-42)
14.1. ELECTRIC DIPOLE RADIATION 607

Or
0 , 0 :
30 (Cr sin 6) do + oe (Cr sin 6) dr = 0, (14-43)

and the total differential of the quantity Cr sin 6 is zero. Then

Cr sin 6 = Constant, (14-44)

or, substituting the value of C and omitting the constant terms,

Sin /Ae ;
x G as i)exp ja( ai ‘)= Constant, (14-45)

‘ x2 1/2 r r
sin? @ é os 1) cos (otas + arc tan z)=i, (14-46)

where K is a real parameter that varies from one line of force to the next. The
calculation is exact as long as r>>s5 and X>>s. All three factors determine
the shape of the line of force as a function of r and 6, whereas the cosine term
gives the radial motion. Figure 14-9 shows four families of lines of force.
For r >A,

sin? 6 cos (wr= :+ ) = 1K (14-47)

and the lines of force travel outward at a velocity wA = c. Closer in, however,
the arc tan term varies with r, with the result that the velocity of the lines of
force is larger than c.
The magnetic lines of force are much simpler: they are circles perpen-
dicular to, and centered on, the axis of the dipole. This can be seen from Eq.
14-11, which shows that H is everywhere azimuthal.

14.1.6 The Kx Surface*

It is instructive to represent the above equation in the form of a three-dimen-
sional surface, as in Figure 14-10. This shows KX as a function of 7 and 6, as
in Eq. 14-46, for t = 0. The loops drawn on the surface correspond to
constant values of KX and are therefore lines of force. They are in fact the
same lines of force as those shown in Figure 14-9, for wt = 0.
As time goes on, the angle wt — (r/A) + arc tan (7/4) increases, and the
result is that the ripples move out, somewhat like a damped wave, carrying the
lines of force with them. Let us examine what happens to the lines of force.
Figure 14-11 shows the intersection of the KX surface, again for t = 0,

* You may wish to omit this rather detailed discussion of the electric lines of force of
the oscillating electric dipole.
608

Figure 14-10. The parameter KA of Eq. 14-46 is plotted here as a function of the
coordinates r and @. The surface shown is that corresponding to tf = 0; with
increasing time f, the central peaks oscillate together from —» to +, and the
ripples move out radially. The loops are electric lines of force corresponding to
constant values of KX.

with the plane 6 = 7/2. This curve is situated inside its envelope, defined by
the curves
Kia + 1(3) me iy" (sin @ = 1), (14-48)
which are shown by broken lines. These curves approach infinity as r/X > 0
and approach unity when (r/A)? > 1. All values of KX are possible, from —
to + «. The figure also shows a succession of curves for successive values of f.
As the ripples move out, their height decreases rapidly at first, and soon
approaches unity. It is clear that lines of force with KX < 1 can travel out to
infinity. They give the radiation field. It is also clear that if KX > 1, they cannot
go far. Let us consider the case where KX is slightly larger than unity. The
loop formed by the corresponding line of force shrinks until it reaches the top
of the ripple and then disappears. This explains the relatively rapid decrease in
the field intensity in the region where r is of the order of X or less.
Closer to the dipole, which is situated at the origin, some lines of force,
such as that for which KX = 5, do not even get into a ripple. These lines of
force simply pulsate in and out without ever escaping into space. This is the
static field.
 609

One
ON
PSS
5
THEY

Figure 14-11. This figure illustrates how the intersection of the KX surface (Figure
14-10) with @ = 7/2 changes with time. The curves are all situated within the
envelope shown; thus the amplitude is infinite at r = 0 and approaches unity for
oN

Let us analyze the motion of the lines of force quantitatively. We can set
sin 6 = | and consider their motion only in the 6 = 7/2 plane. Then
iy Ay ie r .
(5 + 1) cos (otzi + arc tan ‘)= Ki (sind = 1). (14-49)

This represents a family of curves such as those of Figure 14-11. To obtain

the velocity of a line of force, we keep KX constant and calculate u = dr/dt.
We find that
i 1)sin X
Dasa LR (ai) a (14-50)
mee sin X — 7 cos X
where we have set
A = wt — :+ arc tan (14-51)

If we eliminate the cosine term,

(14-52)
610 RADIATION OF ELECTROMAGNETIC WAVES

The velocity u of the line of force becomes zero for sin X = 0, or for cos
X = 1. This corresponds to a point where the ripple touches the envelope.
The curve then “‘rolls’’ on the envelope and the line of force momentarily has
zero velocity.
The velocity u also approaches plus or minus infinity near

anes. (14-53)
3

or
Kx
r p\ 12 (14-54)
a (1+ 4)
Now this is precisely the condition that defines the top of a crest, or the bot-
tom of a trough. The infinite velocity simply results from the fact that, as the
ripple decreases in amplitude with increasing r, the line of force closes in with
infinite velocity just before disappearing. This is interesting in that it tells us
where the lines of force disappear.
Note that the top of a crest does not occur at cos X = 1, or sin X¥ = QO,
where the curve touches its envelope, but somewhat closer to the dipole at a
value of X defined by the Eq. 14-112, since the cosine term in Eq. 14-49 is
multiplied by another term that decreases with increasing r, namely

(a/ry? + 1p”.
For large values of 7/X, KX is about unity, and the top of a crest, or the
bottom of a trough, occurs at X~ 0. For small values of r/X, and for large
values of KX, these occur at |sin X| > 0.
For r > i, KX is about unity or less, and u— c. The lines of force then
travel outward at the velocity of light, as expected.
For Ki > 1 and r pA,

(14-55)

and these lines pulsate in and out without ever escaping.

It is important to note that the KX surface and Figure 14-9 do not give the
magnitude of KE, but only its direction. For example, in drawing a figure such
as 14-9, one naturally selects equal intervals of KX; this leads to a constant
density of lines of force for r >>, which appears to indicate that the ampli-
tude of E does not decrease with r. In fact, E decreases as 1/r for r >> X, from
Eq. 14-22.
 611

Io cos wt

Figure 14-12. Half-wave antenna. The current distribution shown as a broken line
is I, cos (//X) cos wt. This is the standing wave pattern at some particular time
when cos wt = 1.

There is, of course, no such thing as a distinct line of force; the KA

surface provides only the direction of E as a function of the position r, 6, and
of the time f.

14.2 RADIATION FROM A HALF-WAVE ANTENNA

The half-wave antenna illustrated in Figure 14-12 is commonly used for

radiating electromagnetic waves into space. It is simply a straight conductor
whose length is half a free-space wavelength. When a current Jy cos wt is estab-
lished at the center by means of a suitable electronic circuit, a standing wave
is formed along the conductor and the current / at / is

t= 16 COs :COS wf. (14-56)

Each element Jd/ of the antenna then radiates an electromagnetic wave

similar to that of an electric dipole, and the field at any given point in space is
the sum of all these fields.
In many cases the half-wave antenna is a one-quarter wave length mast
set vertically on the ground, which then acts as a mirror (Problem 14-9). The
mast and its image in the ground together form a half-wave antenna. Radio
broadcast antennas are often of this type. To achieve good conductivity, the
612 RADIATION OF ELECTROMAGNETIC WAVES

ground in the neighborhood of the antenna can be covered with a conducting

screen.
This description of the half-wave antenna is really contradictory, because
the standing wave along the wire can be truly sinusoidal only if there is zero
energy loss, and hence no radiated wave. It turns out, however, that a rigorous
calculation leads to nearly the same result as the approximate one. The current
distribution is not quite sinusoidal, but the distortion has little effect on the
field. It will therefore be sufficient for our purposes to assume a pure sinusoidal
current distribution.
The standing wave can be expressed in exponential form as follows:

J = Re 3 exp (01= 4 RE J (or+ :)t (14-57)

where “‘Re” means, as usual, “Real part of.’ The right-hand side shows that
the standing wave is the sum of two traveling waves, one in the positive di-
rection, and one in the negative direction, with amplitudes /)/2.
Then, from Eqs. 14-2 and 14-3, using the usual complex notation, we can
express the electric dipole moment p of the element d/ as

jop = jape®! = af exp j (ur— ‘)| exp (er + -)}dl. (14-58)

14.2.1 The Electric Field Intensity E

In calculating the electric field intensity at the point (r, 6, »), we shall assume
that the distance r to the point of observation is much greater than A. Then
the electric field intensity dE from the element d/ is given by Eq. 14-22, in
which we must substitute the value of po obtained from the above equation.
Thus
uA ee ee Wide gl abe 3 ee! Etre \a a
dE = Sacer’ exp i( or x 4 + exp (of + “Gal zy} sin 6 dl 6,

(14-59)
where r’ is the distance between the element d/ and the point (r, 6, ¢), as in
Figure 14-12,
r' ~r — Icos 8, (14-60)
and

E=PE as SC
Teer X sin @ Gadd
. lag 28
ee (ix (cos @ — 1)

+- exp (7:(cos @ + »)} dl@,. (14-61)

We have removed the 1/7’ from under the integral sign and set it equal to Lyin
14.2. RADIATION FROM A HALF-WAVE ANTENNA 613

since it is assumed that r >> X. With this condition, the dE’s can all be taken to
be parallel for a given point of observation and can be assumed to have the
same amplitude but different phases, and the integration can be limited to the
phases. Integrating,
oe Ih exp jw[t]
me J8rceodr
+/4
(exp AR (cos 6 — »} X exp ee (cos 6 + | :
X sin é Cy aN + Fee ip es oe. 6,. (14-62)

It is not permissible to expand the exponential functions between the main

parentheses in series form, since / is not small with respect to %. We have

te pire ior it\2 (cos 0

aa (e < )} aoe
sin y “
2 (cos 6 + 3)
4nceor cos @ — | cos é@ + |
(14-63)
The expression between the braces can be simplified by setting
pel
sin a ee Say 5
hyaay
= SeCOs(5
7 008 6 (14-64)

init
sin \2 (cose lho
I) — i COS @5 cos 6 ) (14-65)

and adding the two terms. Then

; cos (5cos 6)
i = Dacre Ih exp jw[t] = Tein Gaten 0,, (14-66)

cos G COs a)
= 60.07 fooui \—; 0, volts/meter. (r > A). (14-67)

At 6 = 0 and @ = z, the above equation is indeterminate because the trig-

onometric term becomes 0/0. According to l’Hospital’s rule, the limiting
value of such a ratio is equal to the ratio of the derivatives of the two functions
at the limit and
eee 70) 0) aw COS ( COS a)
Le:
lim ————_——
Fae
= 5, ———_—__;
Ne | (14-68)
son sind ay: |
do 0=0.0
ff :
| sin(5 cos a) sin 0
= “ = | = 0. (14-69)
cos @ 6=O0,7

Thus E is zero along the axis of the antenna.

614

Figure 14-13. Polar diagrams the functions of cos {(7/2) cos 6} /sin 6 and cos?
{(a/2) cos 6\ /sin? 6 showing, respectively, the angular distributions of E or H and
of S.y for the half-wave antenna at r > 4. The angular distributions are similar
to those for the dipole, except that the half-wave antenna radiates a larger frac-
tion of its power in the region of the equatorial plane.

It is interesting to note that the electric field intensity for a half-wave

antenna is independent of the frequency for a given current J): the E for an
elementary dipole is proportional to 1/A for a given current amplitude, and
the integration over the length of the antenna has introduced a factor of X.
Figure 14-13 shows the radiation pattern for a half-wave antenna. It is
quite similar to that for the dipole, except that a somewhat larger part of the
field is radiated in the region of the equator. The reason for this similarity is
that the phase difference between dE’s from the elements of current along the
antenna is small, except when the point of observation is near the polar axis,
where the electric field intensity is zero in any case.

14.2.2 The Magnetic Field Intensity H

It is now a simple matter to find the H vector. For the electric dipole we found
that H is azimuthal, as in Figure 14-7, and that
1/2

z = (2) = 377 ohms, (14-70)

€0

as in Eq. 14-23. For the half-wave antenna, H is again azimuthal, and the
above equation also applies; thus

COs (5cos 0)
i ay Ih exp jw| ft] sin 8 Yi. (14-71)
Z

14.2.3 The Average Poynting Vector

and the Radiated Power
The average Poynting vector $,, gives the average flux of radiated energy:
14.2 RADIATION FROM A HALF-WAVE ANTENNA 615

l
Say =o) Re (E X H*), (14-72)

. \ eos ae
G cos#)

81° ce 1? sin’0 ny Gale)

P cos? 5 cos 8)
== 0.50 2 = See watts/meter?. (14-74)

It points radially outward and varies as 1/7’, which ensures conservation of

energy. The average Poynting vector is shown as a function of @ in Figure
14-13.
The radiated power is again obtained by integrating S,,, over a sphere of
radius r:
ie ifioca
cost
cos? (Fos)
5 608
F
= 995 [Laas
2
+
2 ; r? sin 6 dé de, (14-75)

mICOS: ¢ cos 0)
= 60.0 [ens —__—____ qf (14-76)
0 sin 6

To perform this integration, we set

a T
= cos 6 = AD (14-77)

and
20
1 — cosa
Wer¢ 60.0rhin2 | eon da. ~
(14-78)

Then if we write
i i fil 1
al4r — 2a) 4r (;a 27 — ) Sete)

the radiated power becomes

° Or ws

W = 15.0 Ing (":de +fo [a 8 a) (14-80)

0

Now the two integrals between the braces are equal, as can be seen from the
curves for 1 — cos a, for a, and for 2x — a, shown in Figure 14-14. Since the
curve for 1 — cos a is symmetrical about a = 7, the area under the curve
(1 — cos a)/a must be equal to that under (1 — cos a)/(27 — a) between the
limits 0 and 27. Thus

W = 30.0 rea eco (14-81)

a 0
616

Figure 14-14. The three functions

1 — cos a, a, and 27 — a@as
functions of a for a = 0 to 27.
The curves for (1 — cos a)/a
and for (1 — cos a)/(2m — a)
are symmetrical about a = 7;
thus the areas under them must
be equal for the interval shown.

This integration cannot be performed analytically, but tables are available*

and the integral equals 2.4377. Thus

W =a Te awats, (14-82)
and
Rea = 73.1 ohms. (14-83)

The radiation resistance of a half-wave antenna is therefore 73.1 ohms. This

assumes that the current distribution on the antenna is sinusoidal, which is
not quite correct, as we saw at the beginning of Section 14.2.

14.3 ANTENNA ARRAYS

It is often desirable to radiate energy predominantly in some given direction.

This is achieved with arrays of antennas that utilize appropriate interference
effects. Two simple examples are shown in Figures 14-15 and 14-16.
It is clear that by phasing and by spacing antennas properly a great
variety of radiation patterns can be achieved.
Let us calculate the radiation pattern of Figure 14-15. We choose coor-
dinates as in Figure 14-17 and assume that r>> X. Then the electric field
intensity of each antenna is given by Eq. 14-67 and, if the antennas are
excited in phase, the total field is

* See, for example E. Jahnke and F. Emde, Tables of Functions (Dover, New York,
1951), pp. 3 and 6.
 617

Figure 14-15. A simple antenna array, represented by the two vertical bars, and
its radiation pattern for r >> A. The two half-wave antennas are spaced \/2 apart
and are excited in phase. The surface shows the magnitude of E plotted radially
as a function of 6 and ». The curves shown are situated on the surface for con-
stant values of y chosen at 10° intervals. The surface is cut into two parts for
clarity. Along the y-axis the two waves add, and the resulting electric field in-
tensity is twice that produced by a single antenna. The same applies to all of the
yz-plane, at least for r >> A. Along the x-axis, however, the two waves arrive in
opposite phases and cancel. For other directions along the xz-plane the waves do
not cancel completely, since the difference in path is smaller than \/2. There is
zero field along the z-axis for each of the antennas, and hence for the array.

T
, cos (Jcos ‘)
ee
Va sin 6
J (- D cos (i?cos i)i :
x ieee Ix + exp marge’ i 61, (14-84)

where the quantity on the second line accounts for the fact that the two waves
arrive out of phase, one having traveled a distance r + (D/2) cos y, and the
other a distance r — (D/2) cos y. Thus
Tv
5 cos ( COS a)
E = 120.0; BIEN
r
ke
sin @
cos (5
2K
cos v)6, (14-85)
618

Figure 14-16. The two half-wave antennas shown as vertical bars are spaced \/2
apart, but the one at x = — D/2 leads the other by zm radians. See legend of Figure
14-15 for an explanation of the surface. The two waves now cancel everywhere
on the yz-plane. All along the x-axis, the two waves arrive in phase to give twice
the field of a single antenna. There is again no radiation in the direction of the
z-axis.

We can replace the angle y by 6 and 9, since

rcosy = rsin@cos ¢, (14-86)

and then

, cos (5cos a)
E = 120.0) pel + cos & sin 6 cos e)6; volts/meter.
Dr
(14-87)
When the antennas are one-half wavelength apart, D/(2X) = 2/2. Then,
in the xy-plane, where 6 = 7/2, E varies as

COS
oaks
Lees

At ¢ = Oor 7, this function is zero, while at ¢ = 7/2 itis maximum. There is

constructive interference along the y-axis and destructive interference along
 619

Figure 14-17. Pair of parallel

half-wave antennas D meters
apart. The point P is situated
approximately at distances. r —
(D/2) cos y and r + (D/2) cos p
from the centers of the antennas.

the x-axis, as must be expected. In the xz-plane, g = 0 and E varies as

COS ¢ cos )
Nea. cos ¢ sin @ )-
sin 6 2)

The first term is the angular distribution for a single half-wave antenna; it is
zero at 6 = 0 and maximum at 6 = 7/2. The second term comes from the
interference between the two antennas; it is maximum at 6 = O and zero at
6 = 1/2. The product of the two is zero both at 6 = 0 and 6 = w/2. Finally,
in the yz-plane, ¢ = 7/2, and E varies simply as

COs ¢ COs s)

sin 6

as does the E of a single half-wave antenna. This is to be expected, since the

two waves arrive there in phase, and the total field is exactly twice that of a
single antenna, at least for r > A.
If the antenna centered at x = — D/2 has a phase lead of 7, then the
total electric field intensity is

T
cos { = cos 6
je 60,0) 1h exp jw[t
xp sole) ic__f
2 )

x {exp (- ese “fe expi(—sy =| r) 04, (14-88)

which is the same as Eq. 14-84 except for the addition of jr in the last ex-
ponent. Recalling that e’" = —1,
620 RADIATION OF ELECTROMAGNETIC WAVES

T
: cos (5cos s)
E = 1200 ie
expel — 4 sin (3 sin 6 cos e)6, volts/meter.
r i 2K
(14-89)
The radiation pattern is shown in Figure 14-16.

14.4 ELECTRIC QUADRUPOLE RADIATION

We now go on to the more elaborate type of radiation produced by a linear
electric quadrupole whose moment is a sinusoidal function of the time.
The linear electric quadrupole (Section 2.10) is composed of two dipoles
of opposite polarity arranged in line to give three charges + QO, —2Q, + Q as
in Figure 14-18. The dipole moment of such a charge distribution is zero, but
the quadrupole moment is
Pu = Se O27 = 20S, (14-90)
if
QO = Qe", (14-91)
then
Piz = 2Qys?e7", (14-92)
Pio = 2Qos?. (14-93)
There is no dipole radiation, since the dipole moment p = a Qz is always
zero. Nevertheless, there must be some sort of radiation from the moving
charges. This is what we shall investigate.
We shall use the above model to calculate the radiation field of a linear
quadrupole, although any charge distribution will produce exactly the same
field if it oscillates in such a fashion that

Past | pz? dr (14-94)

is a sinusoidal function of time. This type of radiation would arise, for example,
if currents were excited at the surface of a sphere in such a way as to drive the
electrons alternately to both poles and then to the equator.
We could proceed exactly as for electric dipole radiation and calculate
successively V, 4, E, H. This will be done in Problem 14-11. It is easier, how-
ever, to add the fields of the two component dipoles as follows. For simplicity,
we consider only the radiation field and set r >> X.
We now have two dipoles, one with moment —pye’*! centered at —s/2,
and another with moment + poe’! centered at + 5/2, as in Figure 14-18. The
electric field intensities of the two dipoles add vectorially, according to the
 621

P(r,0,@)

Figure 14-18. Linear electric

quadrupole formed of two di-
poles of opposite polarity, one
above the other. The dipole
centered at —s/2 has a moment
of — Qs; the dipole centered at
+5/2 has a moment of +Qs.
The charges are assumed to
pulsate with O = Qyei'.

principle of superposition (Section 2.2). From Eq. 14-22, the electric field
intensity in the radiation field of an electric dipole situated at the origin is

(Fy ane exp jut] sin 6 (r > 2). (14-95)

Since the two dipoles forming the quadrupole are centered some distance
away from the origin, their electric field intensities at the point (7, 4, ¢) will
differ slightly in direction, in amplitude, and in phase. We may easily neglect
the difference in direction and in amplitude for r > s, but not the difference
in phase. Thus

ta iat i :
oe Are? PAINS

X sin @ {exp (ia cos a)— exp i imCOs 6) 0;. (14-96)

The last two exponential functions can be expanded as power series in s/A for
s << X, and their sum then reduces to (js/X) cos 6. Thus

E = —~P#_ exp jw[r] sin 6 cos 0 (r>>A> 5). (14-97)

8rreqr'X?

There can be no radiation along the axis 6 = O or m, where neither of the

622

Figure 14-19. Radiation pattern for a vertical oscillating electric quadrupole at the
origin. The amplitude of E or of H in any given direction is proportional to the
distance between the origin and the outer surface in that direction. The inner
surface is a similar plot of the magnitude of §,,. There is no field along the axis or
along the equator of the quadrupole. The maximum field intensity occurs along
the surface of a cone at 45° to the axis.

dipoles forming the quadrupole radiate energy; there is also zero radiation
along the equator @ = 7/2, where the two dipoles give equal and opposite
fields.
The magnetic field intensity H is found in the same manner from Eq.
14-14 and
€0 1/2
H=(=) Fe: (14-98)
The E and H vectors for electric quadrupole radiation are therefore
oriented in the same manner as those for electric dipole radiation. This was
to be expected, since we have simply superposed the fields of two electric
dipoles. The amplitudes of both E and H are inversely proportional to r and
therefore decrease at the same rate as for dipole radiation. This makes the
14.5 MAGNETIC DIPOLE RADIATION 623

average Poynting vector (1/2) Re (KE X H*) decrease again as 1/r?, which is
necessary for conservation of energy.
There are two main differences between electric dipole and electric
quadrupole radiation: the former field increases with the square of the
frequency, whereas the latter increases as the cube of the frequency; the dipole
field is zero along the polar axis, whereas the quadrupole field is zero both at
the poles and at the equator.
Figure 14-19 shows the radiation pattern for an oscillating electric
quadrupole at the origin.

14.5 MAGNETIC DIPOLE RADIATION

We studied the static magnetic dipole in Section 7.8, where a current loop of
area S and current / was defined to have a magnetic moment
m = IS. (14-99)
The vector S is perpendicular to the small area limited by the loop, and its
direction is related to the current by the right-hand screw rule.
We shall consider a magnetic dipole as in Figure 14-20 carrying a current

ee
xX
lea
ae

Figure 14-20. Magnetic dipole antenna fed by an oscillator. The E, H, § vectors

are oriented as shown when the phase angle w{t — (r/c)} is zero.
624 RADIATION OF ELECTROMAGNETIC WAVES

T= he, (14-100)
but magnetic dipole radiation can also be produced by any current distribution
whose magnetic dipole moment

m = 5LG x J) dr (14-101)

is a sinusoidal function of the time. This general definition of the magnetic

dipole moment was stated in Eq. 7-116.
The conducting sphere can again be used to illustrate the oscillating
magnetic dipole. In this case charge would flow parallel to the equator,
alternately in one direction and then in the other.
We first calculate V and A, and then E, H, S, and W. We assume that the
loop is small with respect to X; if it were not, there would be wave effects along
its circumference, and the current would not be the same all around it.

14.5.1 The Potentials V and A

If the impedance of the loop is small, all its points are at the same constant
potential, which we can set equal to zero, since we are not interested in adding
an electrostatic field to the radiation field. Then V is zero throughout space.
We have already found 4 in the example on page 430:

_ wo Mo Xr (A : . 5
A= ae ges (-+) exp jw[t]. (14-102)

You will remember from Sections 6.7 and 10.3 that the Lorentz condition
states that V-A is —euo times dV/dt. Since V = 0 for a magnetic dipole, the
divergence of the above A must be zero. This will be shown in Problem 14-19.

14.5.2 The E and H Vectors

We can now find E and H for a magnetic dipole:

1
H=-VXA, (14-103)
Mo

= gq PHL {2(Ts+7) cosors + (Z+j> — 1)singer

m : x? .A : <n :

(14-104)
=p
ay exp jw[t] sin @ 0, (r > Xv); (14-105)
14.5 |MAGNETIC DIPOLE RADIATION 625

0A 0A

= wo\ "2 mo K .
= =(2) Aan (exp jw t]) (/Pa 1)sin 6 y, :
(14-107)

2 (2) pas exp jolt] sindg:

Lo 2m 5 A
>). (14-108)
The scalar potential V is zero, and the electric field intensity arises solely from
the changing magnetic field. The vector H lies in a plane passing through the
z-axis, whereas E is azimuthal, as in Figure 14-20.
At zero frequency (w = 0, A> 0),

(fix = (2 cos 0 r, + sin 6 0) (w = 0), (14-109)

E30 (cor—=s0): (14-110)
This is the field of a static magnetic dipole at a distance r that is large com-
pared to the radius of the loop. The value for H agrees with that found in
Eqs. 7-112 to 7-114.
It is interesting to compare magnetic and electric dipole radiation. The
FE and H vectors for the latter are given by Eqs. 14-22 and 14-14 for r >A. It
will be observed that the two fields are quite similar, except that the expres-
sions for E and H are interchanged. Also, the sign of E for the magnetic dipole
is opposite that of H for the electric dipole. Such a change in sign is required
to keep the Poynting vector directed outwards. The similarity between the
fields of the electric and of the magnetic dipoles is a good illustration of the
duality principle that was discussed in Section 10.8. It will be found that
these two fields satisfy the symmetry conditions and that the constant of
proportionality K is here equal to m/po when the magnetic dipole field is
chosen as the primed field.
The lines of force are similar to those of electric dipole radiation, except
that, again, KF and H are interchanged.

14.5.3 The Average Poynting Vector

and the Radiated Power
The average Poynting vector and the radiated power for the magnetic dipole
are calculated as in Section 14.1.4:

Say
== Cuomo Sin? 9 ri,
327 2r2X4 (
14-111 )

=r1,19 one
p24 r; watts /meter’; (14-112)
626 RADIATION OF ELECTROMAGNETIC WAVES

We
ecules
127! (
14-113 )
2

1010 a watts. (14-114)

To find the radiation resistance of a circular loop of radius a, we set
My = ral (14-115)

for the magnitude of the dipole moment, and then

4
oe OO (;) F2 watts. (14-116)
The radiation resistance is the coefficient of /5/2:
4
Rraa = 197 (;) ohms (a<\). (14-117)
The radiation resistance of the magnetic dipole is proportional to the fourth
power of the frequency, whereas that of the electric dipole was found to be
proportional only to the second power of the frequency in Eq. 14-36.

14.6 MAGNETIC QUADRUPOLE RADIATION

In the previous section we studied an oscillating magnetic dipole consisting

of a circular loop centered on the origin and carrying an alternating current.

a0)

Figure 14-21. Simple magnetic quadrupole comprising two

parallel loops of radius a separated by a distance a and
excited in opposite phases, as indicated schematically by +0
and —WV.
 627

Tale
5 R UV U
ees > |

NE

(a) (b)
Figure 14-22. Electric and magnetic dipoles used as receiving antennas. The
incident electromagnetic radiation induces an electric field in the wires, and the
resulting voltage U across the resistance R is measured with a suitable electronic
circuit.

We can now form an oscillating magnetic quadrupole with two such loops
parallel to each other on either side of the origin and oscillating in opposite
phases. We shall consider the simplest case, in which the distance between the
loops is equal to the radius a of the loops, as in Figure 14-21, and we shall
consider only the radiation field (r >> 4). The lower dipole is centered at
z = —a/2 and has a moment —mpe’*‘, whereas the upper dipole is centered
at z = +a/2 and has a moment +/me'. This arrangement is purposely
made similar to that of the linear electric quadrupole of Section 14.4.
As for the electric quadrupole, we may add the fields of the two magnetic
dipoles, neglecting the differences in amplitude and in direction, but taking
into account the difference in retardation. This is done by multiplying the
dipole fields by plus and minus exp (ja/A) cos 4, as in Section 14.4. Thus,
1/2

LESH () wie exp jw[t] sin 6 cos 6 ¢1 (r >A), (14-118)

€0

(r >A). (14-119)
LEee) Anake exp jw[t] sin 6 cos 6 6;

14.7 THE ELECTRIC AND MAGNETIC DIPOLES

AS RECEIVING ANTENNAS

Figure 14-22 shows electric and magnetic dipoles used as receiving antennas.
The tangential component of the incident electric field induces currents in the
wires that (a) reradiate energy and (b) produce a voltage 0 across the load
628 RADIATION OF ELECTROMAGNETIC WAVES

resistance R. It can usually be assumed that these currents have a negligible

effect on those of the transmitting antenna. The voltage across R can be
measured with some appropriate electronic circuit.
For the electric dipole, it can be shown that

U = Es, (14-120)

where E; is the component of the incident electric field intensity that is

tangent to the antenna and s is the length of the antenna. This assumes that
(s/2)? « X2 and that R—«. A proper demonstration is quite elaborate and
will not be gone into here.*
For the magnetic dipole, the induced electromotance is relatively easy to
calculate:

oA
fpeal Il — f (F + vv) - dl, (14-121)
C c\ ot

ll -{ (Vx +r xe) Pe (14-122)

S

where S is any surface bounded by the circuit. The second term on the right
vanishes, since the curl of a gradient is identically zero. Also, the order of the
operations in the first term can be interchanged to give

sate te) UV - = o2te) ; 3 : 4

A298
pe dl at xX A-da |B da Cl 123)

The induced electromotance in the loop is therefore equal to the rate of change
of the flux linking the loop. It is maximum when the normal to the loop is
parallel to the local B.
The voltage © when R—~ is not necessarily equal to the induced
electromotance, because the circuit may also be excited in the electric dipole
mode. For example, with a symmetrical loop such as that shown in Figure
14-22, if the E vector is parallel to the wires leading to R, charge oscillates
from one end of the circuit to the other, U is not affected by the electric dipole
oscillation, and the above relation is correct. On the other hand, if the E
vector is in the plane of the loop but is perpendicular to the pair of wires, an
extra voltage appears on R that comes from the dipole excitation and adds to
the above induced electromotance.

*R. W. P. King, Handbuch der Physik (Springer-Verlag, Gottingen, 1958), Vol. XVI,
p. 267.
 629

Re R,

_ Transmitter Receiver
Figure 14-23. Pair of loop antennas. The one on the left is fed by an oscillator
supplying a voltage U; the other is connected to a load resistance R, and to a
zero impedance ammeter. The current / could also be measured with a high-
impedance voltmeter across Rs. When a current / is drawn from a source such as
an oscillator, the voltage drops at the output by an amount AV. The ratio AU/I
is called the output impedance of the source. It is shown here as the resistance Ra.

14.8 THE RECIPROCITY THEOREM

According to the reciprocity theorem, the current in a detector divided by the

voltage at the source remains constant when source and detector are inter-
changed, as long as the frequency and all the impedances are left unchanged.
This theorem is widely used for investigating both electric circuits and
antennas. We shall prove it in the general case by using Maxwell’s equations.
A pair of loop antennas, one of which is used as transmitter and the other
as receiver, will serve to illustrate the discussion without restricting its
generality. These are illustrated in Figure 14-23. The conductors and the
medium of propagation are assumed to be isotropic. The source is connected
to the left-hand antenna and supplies a voltage U, while the detector is
connected to the right-hand antenna and measures a current /. The reciprocity
theorem states that the ratio J/VU is not affected if source and detector are
interchanged as in Figure 14-24.
We have seen in Section 10.9 that, for any two fields a and b of the same
frequency,
V-(E. X H.) = V-(Ey X H,) (14-124)
at any point except within the sources. This is Lorentz’s lemma. The more
general case that does not exclude the sources is more complex, but it is also
more interesting and more useful.
Let us consider the field a that is obtained when the antennas are used as
in the top part of Figure 14-24, and the totally different field b that is obtained
when the source is inserted in b and the current detector is in a as in the lower
630

R,

Transmitter Receiver

Field vectors E,, H,

Receiver Transmitter
Field vectors E,,, H,,

Figure 14-24. Pair of loop antennas with the source in a (top), and then in 6
(bottom). The frequency and the impedances are the same in both cases.

part of the same figure. The frequency and the impedances are assumed to be
the same in the two cases.
At any point in space including the loops, and even including the source
in loop a, we first have the mathematical identity

V-(E.
X Hy — Ey X H,)
=H)VXE,— EV X A, — Hav X E,+ Ey-V X Hy. (14-125)

Then, using Maxwell’s equations,

V-(Ba X Hy— ByX Hy)= —Hy- SP— By-(Js+ 2")

OB, = OD, 5
+H, cre + E, = (Ji = a (14-126)

or, replacing the operator 0/dt by jw,

V-(E, x H, om E, x H.,) — Ey: Sia Se E+ Jj. (14-127)

For points outside the source, the current density J, is simply cE, if we
assume that Ohm’s law applies. However, within the source, there is a further
electric field intensity E, (Section 8.5) and, in general,
14.8 THE RECIPROCITY THEOREM 631

Jja = o(Ea + Exc), (14-128)

Jip I= o( Ey + E,p). (14-129)
The quantities F,, and E,, are the applied electric field intensities within the
source when it is in loop a and when it is in loop b, respectively. The current
density at any point in space is Jj, when the source is in a, and it is J;, when
the source is in b, Eliminating FE, and EF, on the right-hand side of Eq. 14-127,
V-(CE, X Hy — Ey X Ha) = Esae Spo — Esve Sya- (14-130)

In general, the right-hand side is not equal to zero. This relation applies to any
pair of electromagnetic fields at any point in space, even inside the sources.
Let us integrate over all space:

/ V-(E, X H, — Ey X H,) dr = | (Eya* Jpop — Esye Spa) dr, (14-131)

or, from the divergence theorem,

/ (E. X H, — Ey X H,)-da = | (Eyat

J7p — Esy+Jya) dr. (14-132)

If we now assume that the sources are limited to a finite volume, the surface
of integration on the left-hand side is infinitely remote from them, and we
have a plane wave with E and H orthogonal and transverse:
EXCH — EH 1, (14-133)

where r; is the unit radial vector. This is shown in Figure 14-25. Then
€ 1/2
H, = (2) m1 x E,, (14-134)
Ho

and similarly for H, and E,. Thus

E, XH, — EX

= (=) (By X (r1 X Es) — Ey X (r1 X E,)} = 0 (14-135)

1/2

Ho

at points infinitely remote from the sources. We can show that the quantity
within braces is equal to zero merely by expanding it, recalling that ry is
perpendicular to both E, and E,. Therefore the integral on the right-hand side
of Eq. 14-132 must also be zero:

| (Esa: Jpop — Esye Spa) dr = 9. (14-136)

The integration is extended to all space, but it can of course be limited to

632

Figure 14-25. The field vectors E and H at a point infinitely

remote from a source of radiation are orthogonal and
transverse, as in a plane wave.

the sources, since E,, and E,, are zero everywhere else. Thus

| Bodo dr, (14-137)

| Bodnar =
a b

where the integrals are evaluated over the regions where E,, and E,, are
nonzero.
The meaning of this equation can be illustrated by referring again to our
pair of loop antennas. These are shown in Figure 14-24 for the field a, which
obtains when the source is in a, and for the field b with the source in b.
For these antennas,

iE,a* dsp dr = [ Boat Jy, da = Veale in a; (14-138)

where U,a is the voltage supplied by the source in a, and J; j, « is the current in
the same loop a when b is energized. The other integral of Eq. 14-137 can be
expressed similarly, and
Usals IG aa Usola in bs (14-139)

or
ye Oe I, ina

oe > Oks Coat

Physically, this means that the current induced in 6 when a is energized,

divided by the voltage applied on a, is the same as the current induced in a
when b is energized, divided by the applied voltage on 6, as long as the
frequency and the impedances remain unchanged. This is the reciprocity theorem.
We have illustrated it by referring to a pair of magnetic dipoles, but it is
equally valid for any pair of antennas.
It must be kept in mind that this theorem is concerned solely with the
ratio 1/0; it says nothing about the power expended by the source. This
usually changes when the source is moved from one position to the other.
 633

Figure 14-26. Simple electric circuit comprising a source

of voltage © and an ammeter J. The reciprocity theorem
applies: the ratio 7/0 remains constant when source and
ammeter are interchanged, provided the frequency and
the impedances remain unchanged.

One important consequence of the reciprocity theorem is that the radia-

tion pattern of an antenna must have the same shape as the corresponding
plot of the response of the antenna as a function of angle when it is used as a
receiver. This fact is commonly used for the determination of radiation
patterns.
Our demonstration of the theorem is general, except that the media have
been assumed to be linear and isotropic. It therefore applies to ordinary
electric circuits. It is a simple matter to check the reciprocity theorem for the
circuit of Figure 14-26.

14.9 SUMMARY

Electric dipole radiation is produced by any charge distribution whose dipole

moment p is a function of the time.
To obtain the field vectors E and H, one first calculates the retarded
potentials V and A. One finds that, in free space, and for
P = pre’, (14-2)

E Sieasies
eae j
exp jw[t] sin 6 0, (r > nNX), (14-22 )

H = —-?
AarK2
exp jw[t] sind o (r > 2). (14-14)
A transverse spherical wave therefore radiates away from the dipole, and
the ratio E/H is the same as for a plane wave in free space. Closer in, the wave
is more complex: there is a total of seven terms, and FE has a radial component.
The average Poynting vector is everywhere radial:
634 RADIATION OF ELECTROMAGNETIC WAVES

cpio sin? 6
Sam os Biterne | (14-28)

There is no energy radiated along the axis of the dipole.

The half-wave antenna is a conductor \/2 long, which carries a standing
wave of electric current. Its field is calculated by adding the radiation fields
of the individual dipoles formed by the elements dl along its length. The field
of the half-wave antenna is qualitatively similar to that of a dipole and, for
r>A,
j cos 5 Eos a)
E= Inrce Ih exp jw[t] “> gine 6; (14-66)

; COs (5cos 0)
Hie 2nr
ST cexpyoltaesin 6ee (14-71)
i
The radiated power is
W = 73.) Tims watts, (14-82)

where J,,,; is the rms current at the center of the antenna at the point where
it is fed by a source of power. The coefficient 73.1 is the radiation resistance of
the antenna.
Antenna arrays are sets of antennas properly spaced and phased so as to
obtain, by interference, special radiation patterns.
Electric quadrupole radiation is produced by a charge distribution whose
quadrupole moment is a function of the time. The simplest quadrupole
source 1s the linear quadrupole. It is a simple matter to deduce its field from
that of the electric dipole:

ba exp jw[t] sin 6 cos 6 0, (r>A>s). (4-97)

The vectors H and E are related as with the electric dipole.

Magnetic dipole radiation is produced when the magnetic dipole moment
of a current distribution is a function of time. If we assume a circular current
loop of negligible impedance, then V = 0. From the retarded vector potential
A, we find values of FE and H that are related to those of electric dipole
radiation according to the duality principle of Section 10.8.
Magnetic quadrupole radiation is similarly related to electric quadrupole
radiation.
Electric and magnetic dipoles can be used as receiving antennas. For a
short electric dipole, the voltage on a high resistance load is equal to the
tangential electric field multiplied by the length of the dipole. In the case of
PROBLEMS 635

the magnetic dipole, this voltage is the electromotance induced by the chang-
ing magnetic flux linking the dipole, plus, in some cases, a voltage due to
electric dipole excitation of the antenna.
The reciprocity theorem states that the current in a detector divided by
the voltage applied at the source remains constant when source and detector
are interchanged, as long as the frequency and the impedances are left
unchanged. This applies to any electromagnetic field and, in particular, to
electric circuits and to antennas.

PROBLEMS

14-1. (a) Calculate the vector potential A at a distance r on the axis of an oscillating
electric dipole of length s.
(b) Show that the result is the same as it would be if the current were localized
at the center of the dipole.
14-2. (a) Show that the phase velocity of the H field of an electric dipole is

u= (1=r zs)@
r
where 7’ = r/X.
(b) Draw a qualitative graph of u as a function of 7’.
14-3. (a) Show that the phase velocities of the r and 6 components ofE for an oscil-
lating electric dipole are, respectively,
1
w= (+55 C,

uw = {he
r’4— 2r”?
where r’ = r/X.
(b) Draw qualitative graphs of u, and of ue as functions of r’.
density to the average mag-
14-4. Calculate the ratio of the average electric energy
netic energy density in the field of an electric dipole for (a) r< i, (b) r = 4,
(Cine:
dipole is radiated
14-5. What fraction of the total power radiated by an electric
between -£45° of the equatorial plane?
r > X is equal to the
14-6. Show that the Poynting vector for the electric dipole at
energy density multiplied by c.
14-7. (a) Show that, in the field of an electric dipole,
6
Exms = 6.71 win,
where W is the radiated power and r > A.
636 RADIATION OF ELECTROMAGNETIC WAVES

(b) Show also that, for the half-wave antenna,

cos {(7/2) cos OF
errs = 7.01 Wile =
r sin 6

14-8. Calculate the electric field intensity in millivolts/meter at a distance of one

kilometer in the equatorial plane of a half-wave antenna radiating one kilo-
watt of power. Set A < 1 kilometer.
14-9. An antenna is normally situated near a conductor (the Earth, an airborne
vehicle, a satellite, etc.) Energy radiated toward the conductor is reflected, and
the total field is thus the vector sum of the direct wave plus the reflected wave.
It is convenient to consider that the latter is generated, not by reflection,
but rather by an image of the antenna located below the surface of the con-
ductor.
(a) Discuss the image of a horizontal half wave antenna over a perfectly
conducting Earth, and show that the current in the image and the current in
the antenna flow in opposite directions. {
(b) Discuss the image of a vertical antenna over a perfectly conducting Earth,
and show that the current in the image and that in the antenna flow in the
same direction.
Both rules apply to oblique antennas.i
(c) We have shown that the radiation resistance of a half-wave antenna is
73.1 ohms.
Show that the radiation resistance of a quarter-wave antenna perpen-
dicular to a conducting plane is one half of this, or 36.6 ohms.
14-10. Identical parallel half-wave antennas are arranged in line with a uniform
spacing D and are excited in phase.
(a) Show that the angular dependence of the electric field intensity E in the
plane perpendicular to the antennas is given by
sin {((ND 2A) cos y}
sin {(D/2A) cos ¢} ’
where WN is the number of antennas and ¢ is the angle between the direction
of observation and the plane containing the antennas.t
(b) Determine the angular positions of the maxima and minima of E for this
radiation pattern. Differentiation of the above formula will yield only the
maxima.
(c) Show that, for a given spacing D, the main lobe at ¢ = 7/2 becomes
narrower as the number of antennas WN is increased.
(d) Show graphically, by considering the summation in the complex plane,
that the main lobe is sharp and that the side lobes are small when N is large
and D < X.
(e) Plot the radiation pattern for an array composed of four parallel half-wave
antennas spaced \/2 apart.

14-11. Calculate the electromagnetic potentials V and 4 and the field vectors E and
H for the oscillating linear electric quadrupole directly, without using the field
of the oscillating electric dipole.
PROBLEMS 637

14-12. Show that the Lorentz condition applies to the radiation field of a linear
electric quadrupole for r > X.

14-13. Calculate the Poynting vector for the radiation field of an oscillating linear
electric quadrupole, and show that the total radiated power is

ilefe Se MO
QOxo
294

watts.

14-14. A sealed plastic box contains an antenna that radiates electromagnetic waves.
How could you tell whether the antenna is an electric or a magnetic dipole?

14-15. Show that the electric dipole and magnetic dipole fields illustrate the duality
property of electromagnetic fields (Section 10.8).

14-16. The gain G of an antenna is defined as the ratio of the Poynting vector at the
maximum of the radiation pattern to the Poynting vector averaged over a
spherical surface:
G
Spare

|
a I [7ssinkee
Wie fe
6 aode
The gain of an antenna is a measure of its directivity.
(a) Show that the gain of an electric or magnetic dipole is 1.5.
(b) Show that the gain of a half-wave antenna is 1.64.

14-17. (a) Show that the ratio |E/H| for the field of an electric dipole is much larger
close to the dipole than it is far away.
(b) Show that the inverse is true for the magnetic dipole.
(c) Which type of probe would you use to detect (i) electric dipole radia-
tion, (ii) magnetic dipole radiation, near the source?
14-18. In the field of a magnetic dipole V = 0 and A # O. Is it possible to have a
radiation field such that V # 0 and A = 0?
14-19. (a) Show that V- 4 = 0 for the field of the magnetic dipole.
This is the Lorentz condition, since V is equal to zero.
(b) Show also that

Vase ee
Or
0
for the field of the magnetic dipole.
This is the wave equation for A.

14-20. Two identical magnetic dipoles are perpendicular to each other and have a
common diameter.
(a) Show that the radiation pattern (amplitude as a function of @) is a circle
in the plane perpendicular to the common diameter if one dipole leads the
other by 7/2 radian.
(b) Explain the nature of the resulting field.
(c) How would you connect these antennas to a common source he
Such a pair of crossed coils can be used as an omnidirectional trans-
mitting or receiving antenna.
638 RADIATION OF ELECTROMAGNETIC WAVES

Figure 14-27.

14-21. Mobile radio receivers equipped with the usual vertical whip antennas are
subject to signal fading when they move through the standing wave patterns
that exist near electrically conducting buildings.
Let us see how this effect can be eliminated by using a half-loop antenna
like the one shown in Figure 14-27. The E vector is vertical and the plane of
the loop is normal to the local H.
(a) Show that the sum of the signals at A and B is a measure of E, while their
difference is a measure of H.
(b) Then, from Problem 12-21,
WGKCAN Si Te) <i ICAP = 15)
will be a constant in the standing wave pattern if M/WN is chosen properly.
Show that we must have that

where R is the radius of the loop andf is the operating frequency.

This ratio is approximately equal to unity at 100 megahertz.
The single half-loop is not omnidirectional; to obtain an omnidirectional
antenna one must use a pair of crossed half-loops connected as in Problem
14-20.
14-22. A loop antenna of inductance L feeds a load resistance R. The resistance of
the loop is negligible compared to R.
Show that there will be maximum power transfer to the load when
RS or,

14-23. In the design of loop antennas for satellites, weight and size limitations are of
foremost importance.
Show that, for a given loop diameter and for a given mass of copper, the
ratio wL/R is independent of the number ofturns, L and R being, respectively,
the inductance and resistance of the loop.

14-24. Compare the responses of the electric and of the magnetic dipole antennas
when used as receivers (a) in air and (b) in sea water.
In (b), assume a frequency of 20 kilohertz. Assume that the loop antenna
has a single turn, that its diameter is equal to the length / of the electric dipole,
and that /< i.

14-25. (a) Calculate the ratio 7/0 for the circuit of Figure 14-26, and show that
the reciprocity theorem applies.
(b) Calculate the power expended by the source when the circuit is as shown
and when and / are interchanged.
PROBLEMS 639

14-26. Astronomers are interested in the following cosmological problem. Consider

a particular class of objects, say quasars, and assume that they are all identical
and distributed uniformly in a Euclidean universe. Then, if one plots log N as
a function of log S, where N is the number ofobjects giving a radio flux greater
than S at the Earth, one should obtain a straight line whose slope is —1.5.
The slope for quasars is in fact larger and possibly indicates the existence of
cosmological evolution. *
Show that the expected slope is —1.5.

Hints

14-9. Think of the images of the charges flowing along the antennas.

14-10. (a) Perform the summation graphically in the complex plane.

s, Novem-
* See, for example, A. Braccesi and L. Forniggini, Astronomy and Astrophysic
ber 1969, page 3064.
APPENDIXES
>>

naa

b "i| A
raped 2
ok

7 | | |
APPENDIX A

Conversion Table

Examples: One meter equals 100 centimeters. One volt equals 108 electromagnetic
units of potential.

SI CGS SYSTEMS

esu emu

Length meter 10° centimeters 10? centimeters

Mass kilogram 10° grams 10° grams
Time second 1 second 1 second
Force newton 10° dynes 10° dynes
Energy joule 10’ ergs 10’ ergs
Power watt 107 ergs/second 107 ergs/second
Charge coulomb 3 Ike 10+
Electric potential volt 1/300 108
Electric field intensity volt /meter WCE Os) 108
Electric displacement coulomb/meter? 127 a 10P 4r X 10°
Displacement flux coulomb Wie 9 KOE Bb < IO
Electric polarization coulomb/meter? 3X 10° 10m?
Electric current ampere Been)? 107!
Conductivity mho/meter Oye AO! 1071

Resistance ohm LO L.O22) 10°

Capacitance farad yO) Om
Magnetic flux weber 1/300 108 maxwells
Magnetic induction tesla AGH INOS) 104 gausses
Magnetic field intensity ampere/meter 127, < 10! 47 X 107? oersted
Magnetomotance ampere aie >< IO) (47/10) gilberts
Magnetic polarization ampere/meter VG e103") iQ ®
Inductance henry O02) 10°
Reluctance ampere /weber 36a X 101! areal Ome
a ee LEEEEEEEEEEEEEEEEEEEEEEEE

Note: We have set c = 3 X 108 meters/second.

APPENDIX B

The Complex Potential

The complex potential provides a powerful method for calculating electro-

static fields. It is restricted, however, to two-dimensions, that is, to fields
which are essentially constant in one direction. End-effects due to the finite
length of the conductors in this particular direction are assumed to be
negligible in the region considered. The method also assumes zero space
charge density.
Similar methods of calculation are used in other fields of physics in the
study of quantities which satisfy Laplace’s equation. One important example
is the field of hydrodynamics.

B.1 FUNCTIONS OF THE COMPLEX VARIABLE

Let us first consider the complex variable

z=x+jy, (B-1)

where x and y are real numbers and where j = (— 1)". This quantity can be
represented by a point in the complex plane with x as abscissa and jy as
ordinate, as in Figure B-1. This z must not be confused with the z-coordinate
of Cartesian or cylindrical coordinates.
We can have functions W(z) such as z?, or 1/z, or In z, and so on, and

W(z) = U(x, y) + jV(x, y) (B-2)

with a real part U and an imaginary part j/V, U and V being both real func-
tions of x and of y. For example,
 645

Figure B-1. Point z = x + jy in the

complex plane. The quantity r is called
the modulus, and @ the argument, of
the complex number z.

Zz
(OE yy = ? — y*) 4 2ixy, (B-3)
in which
U=x?— y? and V = 2xy. (B-4)
The function W(z) = U + jV can be represented as a point on another
complex plane with U as abscissa and jV as ordinate. We then speak of the
z-plane and of the W-plane.
We shall consider functions W(z) such that the derivative dW/dz exists
in the region considered. This condition leads to an important pair of equa-
tions. First, let us examine the meaning of the derivative dW/dz by consider-
ing Figure B-2. The point W in the W-plane corresponds to the point z in
the z-plane, according to some specified function W(z). If z changes to
z+ Az, W changes similarly to W + AW, where the increments Az and AW
are complex numbers. The derivative dW/dz is the ratio of these increments
at the limit Az — 0. The value of this derivative can take on different values
for different values of z, but we wish to have a single value of dW/dz for a
given value of z, that is, for a given point in the z-plane, no matter how dz is
chosen.
We consider two particular values of dz: dx and j dy. In the first case,
dz is parallel to the x-axis; in the second case, dz is parallel to the jy-axis.
For both of these particular values of dz, we must have the same value of

jV ; iy

(a) W plane (b)

z plane
Figure B-2. The function W(z) = U(x, y) + jV(x, y) shown in the W-plane. When
z changes to z + Az, W changes to W + AW, and the ratio AW/Az is equal to
dW /dz for Az —> 0.
646 APPENDIX B

dW /dz. For the first case, the value of dW/dz becomes dW/dx, and
OW SoU en
Ox Ox J Ox ee)
For the second case, dW/dz becomes 0W/j dy, and
dW lou | oV
na cs ce e B-6
LOA CVS OY. oe
These two expressions must be equal:
OUD OV OU ey
ax dx 9 7 oy uy oy ee
Then
oU aV aU OV
ES = ay and ay => mae (B-8)

These are the Cauchy-Riemann equations. The functions U and V are related
to each other through these equations and are called conjugate functions.
A function W(z) is said to be analytic if its four partial derivatives exist
and are continuous throughout the region considered and, moreover, if they
satisfy the Cauchy-Riemann equations.

B.2 CONFORMAL TRANSFORMATIONS

Consider now a point z and two neighboring points z’ and z” in the z-plane.
These three points correspond to three other points in the W-plane: the point
W and the neighboring points W’ and W’’, as in Figure B-3. Since dW/dz
is unique at the point z, then
W _ W'—Ww rabedl | ghee 2°
ssid=" lim a SaaS lim spills (B-9)

(a) W plane (b)

z plane

Figure B-3. The three points W, W’, W” correspond, respectively, to z, z’, z’’. The
angles a and § are equal when z’ — z and z’’ — z, as long as dW//dz exists and
is not zero.
THE COMPLEX POTENTIAL 647

If dW/dz ¥ 0, it can be written in the form Ae’. (The argument » has no

meaning if dW/dz = 0.) Then, at the limits z’ — z and z’’ — z, considering
only the arguments of the various complex quantities,

arg (W’ — W) = arg(z’ — z)+ ¢, (B-10)

arg (W"” — W) = arg(z” — z)+ ¢, (B-11)
arg (W’ — W) — arg (W” — W) = arg (z’ — z) — arg(z”’ — z), (B-12)
or
a= 8, (B-13)
where the angles a and 8 are as in Figure B-3. This result does not apply to
points where dW/dz = 0.
The angle between two infinitesimal line segments is therefore conserved
in passing from the z- to the W-plane, as long as the derivative dW//dz exists
and is not zero. For example, the two families of straight lines represented
by U = Constant and V = Constant in the W-plane are clearly orthogonal.
Then, the corresponding curves U = Constant and V = Constant in the
z-plane are also orthogonal.
Since there is a one-to-one correspondence between the points in the
z-plane and those in the W-plane, we can imagine the W-plane to be distorted
into the z-plane according to the function W(z). Then a geometrical figure
in the W-plane is “mapped” into a corresponding figure in the z-plane, and
inversely. This process is called a conformal transformation.

Boel Hes PUNCTION YY

(Zz)
AS A COMPLEX POTENTIAL

Let us calculate the second derivatives of W with respect to x and to y:

aW =
dWaz_
=
dW , -14
Ox dz ox ) dz ae
ew of aw
aes B-15
Ox? dz* ( )
ow dWoz_ .dW
— = — — = j— B-16
oy dz oy az ( )
aw ew
—_ = ———- B-17
ay? dz? ( )
and
a eae) (B-18)
648 APPENDIX B

The function W(z) is thus a solution of Laplace’s equation in two dimensions.

Separating real and imaginary quantities, we find that
VU PU _ av, av |

Thus both U and V independently satisfy Laplace’s equation, and either

one can be set to be the electric potential that also satisfies Laplace’s equa-
tion, since we have assumed zero space charge density.
If the imaginary part of W is taken to be the electric potential, then the
equipotentials are given by V = Constant. Since the U = Constant curves are
orthogonal to the equipotentials, as we have seen above, they define the lines
of force, from Section 2.3. The function V is then called the potential func-
tion, U is called the stream function, and W is called the complex potential
function. It is shown in Problem B-2 that the electric field intensity is
dw
nae (B-20)

at any point in the field.

A _two-dimensional electrostatic field with zero space charge density is
therefore completely determined once the complex potential function W(z) is
known.
It is essential to recall here the uniqueness theorem derived earlier in
Section 4.2, according to which there is only one field configuration that
satisfies given boundary conditions. Thus, if in one way or another we can
find a satisfactory function W(z), then that function is the proper one and the
only one.
The determination of W(z) is usually intuitive and empirical. However,
much work has been done in this connection, and it is usually possible to
determine the proper function for simple geometries.* One can usually arrive
at W(z) by using some function of the complex potential for some other
known field.

Example As an illustration, let us consider the field inside an infinite parallel-

plate capacitor as in Figure B-4. The equipotentials and the lines of
force are respectively » = Constant and x = Constant.
Setting V to be the electric potential,
V
V= osy. (B-21)

* See, for example, E. Durand, Electrostatique (Masson et Cie, Paris, 1964), Vol. 1,
and E. Kober, Dictionary of Conformal Representations (Dover, New York, 1952).
 649

Figure B-4. A parallel-plate

capacitor with its lower plate
grounded and its upper plate at
potential V;. End effects are
neglected. The horizontal lines are
equipotentials V = (V,/s)y =
constant; the vertical lines are lines
= of force U = (V;/s)x = constant.

This suggests that the function W(z) must be

We) =U+V= Ot p= Oz (B-22)

Thus,

ae (B-23)

We could also have determined the stream function U through the

Cauchy-Riemann equations.

B.4 THE STREAM FUNCTION

We have seen above that the stream function is a constant along a line of
force. We shall now see that it is quantitatively just as important as the
potential function. We shall assume, as in Section B.3, that V is the potential
function.
Figure B-5 shows three equipotentials and three lines of force in a por-
tion of an electrostatic field. The vector dn is an element of length along the
line of force normal to the equipotential at the point considered. The vector
ds is an element of length along an equipotential and is oriented with respect
to dn so that it points to the /eff when viewed along dn. For convenience,
we choose our x- and jy-axes as shown, so that the x-axis is parallel to dn
and the jy axis is parallel to ds at the point considered.
According to the Cauchy-Riemann equations,
abe nu OV _ an (B-24)
Os Oy Ox on

Since —0V/dn is the electric field intensity E at the point, the positive direc-
tion for E being along dn, then, along an equipotential,
dU = Eds. (B-25)
650

Figure B-5. A set of lines of force and

a set of equipotentials in a portion of
an electrostatic field. The elementary
vectors dn and ds are directed, re-
spectively, along the line of force and
along the equipotential at the point
considered. The vector dn points in the
direction of E, and the vector ds
points to the /eft when one looks
along dn. For convenience, the axes
are chosen to be parallel to these vec-
| tors as shown.

This relation is valid for any point in the field as long as V is chosen to be
the potential function, and E and ds are oriented as in Figure B-5.
The stream function is thus related to Eds. This quantity Eds is the
flux of E crossing the equipotential in the direction of dn through an element
of area on the equipotential surface that is ds wide and whose height, meas-
ured in the direction perpendicular to the paper, is the unit of length, namely
one meter. Integrating Eq. B-25 along an equipotential between the lines of
force U; and Up,
Us Us
U,— U, = / Ms = / Eds. (B-26)
Ui ds Ui

The line of force for which the stream function is zero is chosen arbitrarily,
just as for the equipotential on which the potential is zero. Thus the charge
density o at the surface of a conductor is

og = gE = — —> (B-27)

where the vector ds points toward the /eft when one looks toward the outside
of the conductor into the field.
The total charge Q per unit length on a cylindrical conductor whose axis
is perpendicular to the paper is obtained by integrating o« ds around the
periphery of the conductor in the direction of increasing ds:

Q= fo ds = fdu. (B-28)

The stream function is thus useful for determining surface charge densities
and total charges on conductors.
THE COMPLEX POTENTIAL 651

Example THE PARALLEL-PLATE CAPACITOR

Let us return to the parallel-plate capacitor of Figure B-4, for which
we found the complex potential function W(z) in Eq. B-22. The
charge density on the lower plate is obtained from Eq. B-22 with
the vector ds pointing toward the /eft when one looks toward the
outside of the conductor into the field:
dU aU LAS
Slower = OE = Sa =F C0 ‘ (B-29)

Similarly, the charge density on the upper plate is found to be

+€9(V1/s). Both of these charge densities can be verified to be correct
by using Gauss’s law.
The capacitance C’ per unit area is

: f= = =,
aoe 3
(B-30)

and

je Le ess (B-31)
dz 8
as expected.
We have thus verified our method of calculation by applying it
to a well-known field.

Example THE CYLINDRICAL CAPACITOR

In the case of the cylindrical capacitor, as in Figure B-6, the equi-
potentials are concentric circles, whereas the lines of force are radial
straight lines.
We find empirically the following complex potential function,
which is justified below:

WO) ine. (B-32)

In ee
ry

or, writing
Zee (B-33)

Wee (inE +0) (B-34)

ine £
ry

eee a (B-35)
l a In ee te
ry; ry

Thus
fic eg: (B-36)
652

Figure B-6. A section through a

cylindrical capacitor of internal radii
r, and rz. The inner cylinder is
grounded, and the outer one is
at a potential V,, the charges being,
respectively, —Q and +Q.

eG: (B-37)
r La!
In —
ry

The above expression for W(z) is justified as follows: (a) the

logarithm serves to give equipotentials and lines of force of the re-
quired form r = Constant and 6 = Constant when the real and
imaginary parts of W(z) are set equal to constants; (b) the factor /
serves to make V the potential; (c) the ratio z/r; makes V equal to
zero at r = r;; (d) finally, the factor V,/In (r2/r1) serves to make the
potential equal to V; at r = ro.
Let us find the charge on the inner cylinder. To do this we must
integrate the stream function U in the direction of ds, or of increasing
0, between —7z and +7, as in Figure B-6. Then

Q inner = — Pe (B-38)

The charge on the outer cylinder is found by integrating U in the

direction of decreasing 0, and Qouter = — Qinner, AS expected.
The capacitance C’ per unit length normal to the paper is

o=|2 = “Te, (B-39)

which is correct, and E is also given correctly by

_|aw) | wtlee
=i (B-40)
In —
ry
THE COMPLEX POTENTIAL 653

(B-41)

Example FIELD OF TWO PARALLEL LINE CHARGES

OF OPPOSITE POLARITIES
We consider the field due to two parallel line charges of Q cou-
lombs/meter and of opposite polarities, as in Figure B-7. The line
charges shown at —a and at +a are presumed to be infinitely long
in the direction perpendicular to the paper.
The potential V at any point due to this charge distribution can
be found by integration, as in Problem 2-6:
Q ry
V = on —
In ae (B-42
4 )

where Q is the charge per unit length perpendicular to the paper,

and r; and 2 are as in Figure B-7. This equation shows that V — +
atm = 0, V — —o atr = 0,and V = Oatn = re, which is correct.
Also, for a given value of m, V varies as the logarithm of r2 and,
similarly, for a given value of rz, V varies as the logarithm of rn,
which is also correct.
We can also rewrite V as follows:

meeQ
Dare
I i
22 (B-43)
This suggests that

Way i Oren n=
ea
(B-44)
, @ z+a|en
=] Dac In ( a) (B-45)
z—a

és =|
ae 27reo mn
zt+a
re
oe se a |, (B-46)

\ Figure B-7. Line charges of — Q and

: Z +Q coulombs/meter are situated,
| ‘te \ respectively, at x = —a and
x = +a.
654

Figure B-8. Lines of force (indicated by arrows) and equipotentials for

two infinite line charges perpendicular to the paper. All the curves are
circles or arcs of circles. The equipotential surfaces are generated by
sliding the figure along an axis perpendicular to the paper.

and

rae 21reo
(= 6); (B-47)

Se 2ré9
OS ana
|Z2—a
ee AgHae2
27€)
(B-48)
The lines of force are given by the equation 6. — 6, = constant.
They are thus arcs of circles passing through the line charges and
centered on the jy-axis, both above and below the x-axis.
The equipotential V is determined by the equation

“ = exp (276V/Q). (B-49)

It is shown in Problem B-3 that this equipotential is a cylinder whose
radius is |acsch (27e)V/Q)| and whose axis is situated at x =
la coth 2re)\V/Q|. For V—> +, the equipotential surface reduces
to a line perpendicular to the paper and situated at x = a, as ex-
pected. Similarly, for V — —o, we have a line at x = —a.
The equipotentials and lines of force are shown in Figure B-8.

Example FIELD OF TWO PARALLEL CONDUCTING

CIRCULAR CYLINDERS OF OPPOSITE POLARITIES

The field investigated above was shown to have equipotentials

in the form of circular cylinders whose axes are all parallel. We can
THE COMPLEX POTENTIAL 655

place an uncharged conducting foil on any of these equipotentials

without disturbing the field in any way. When we do this, charges
migrate inside the foil so as to cancel the electrostatic field within it,
and charges of opposite polarity appear on the two surfaces. In this
way, the field remains everywhere exactly as it was, except for the
conducting region inside the foil, where the field is zero. If the line
charge surrounded by the foil is +-Q per unit length, a charge — Q
per unit length is induced on the inside surface of the foil, and a
charge of +@Q per unit length is induced on the outside surface.
These induced charges are of course due to the electrostatic field
and are due to both of the line charges. We can cancel the + Q and
— Q charges surrounded by the foil by shorting the line charge +O
and the foil, leaving us with a net charge +Q on the outside of the
conducting foil and zero charge inside. The field in the internal re-
gion limited by the foil is then zero, but the field outside is un-
changed. Instead of canceling the charges inside the foil as above,
we could similarly have canceled the charges outside, making the
field zero outside the region bounded by the conducting foil but
leaving it intact inside.
We can now find the field due to a pair of parallel conducting
circular cylinders by replacing two of the equipotentials with con-
ducting cylinders carrying charges equal in magnitude but opposite
in sign.
We consider two cylinders of radius R as in Figure B-9, with
their axes separated by a distance D and carrying known charges
—Q and +Q, respectively. Their potentials —V and +V are un-
known. These are the potentials with respect to that along the jy-
axis, which is taken to be zero. The potential difference 2V is espe-
cially important, since it is required for the calculation of the capac-
itance per unit length of the system.
The function W(z) for the pair of cylinders is the same as in
Eq. B-44, except for the fact that now the quantity a is an unknown,
with
Ik = eiceyoln C="), (B-50)

Jy

Figure B-9. A section through two

parallel cylinders of radius R whose
axes are separated by a distance D
and which carry charges — Q and
+ QO coulombs/meter, respectively.
The potentials are —V and +V.
The origin of coordinates is chosen
midway between the axes.
656 APPENDIX B

3 = acoth ea (B-51)
Since
coth? x — csch?x = 1, (B-52)
then

= (a) nsRe] (B-53)

Substituting this value of a in Eq. B-50 for R, recalling that the
capacitance per meter C’ = Q/2V, we find that
TEO
C= ; D2 1/2 (B-54)
sinh“! & _ 1)

TT€o :
(B-55)
~ cosh! (D/2R)
It will be observed that C’ depends only on the ratio D/R, and
not on the actual dimensions, just as in the case of the cylindrical
capacitor. When D = 4R, C’ = 21.2 picofarads/meter.

PROBLEMS

B.1 Two cylinders of length / and radius a have their axes parallel and separated
by a distance D in a liquid. A potential difference V is applied between the
cylinders. Calculate the force of attraction for / = 1.00 meter, a = 0.500
centimeter, D = 2.50 centimeters, ¢, = 2.6, and V = 1.00 X 104 volts. Neg-
lect end effects.

B.2 Show that the electric field intensity for a two-dimensional field in the xy-plane
iS
dw|
dz}
where W is the complex potential and z = x + jy.

B.3 Show that the equipotentials defined by Eq. B-49 are the circular cylinders

E —acoth ‘Gil + y? = a? csch? (228").

B.4 Consider an unknown two-dimensional field due to two charged conductors

A’ and B’ in the complex z’-plane, and a known two-dimensional field due to
the charged conductors A and B in the z-plane. Points in the z’-plane are re-
lated to those in the z-plane according to the transformation z’ = z’(z), and
the corresponding potentials are equal: V4 = V4: and Vg = Vz.
Show that there is conservation of charge under this transformation and
that there is, in consequence, conservation of capacitance.
APPENDIX C

Induced Electromotance
in Moving Systems*

To illustrate the meaning of Eq. 8-36,

OB
YE ier cher MiteX BB), (C-1)

in which the induced electric field intensity E is measured in one coordinate

system and the magnetic induction B in another, let us consider the following
experiments.

Example EXPERIMENT 1

Figure C-1 shows a circular disk D1 rotating with an angular veloc-

ity w about an axis perpendicular to its plane and parallel to a uni-
form magnetic field B. The disk is assumed to be both nonconducting
and nonmagnetic.
We now station two observers, one in the laboratory and one
on the rotating disk, both equipped with a “‘curl-meter,”’ a “‘B-meter,”’
and a stop watch. The “‘curl-meter”’ consists of a small loop of wire
capable of orientation in any direction and connected in series with
a sensitive, infinite-impedance voltmeter. By definition, the compo-
nent of V * Enormal to the plane containing the path of integration
is

E-dl
(V X E), = lim —~—: (C-2)
S—0 S

* See Dale R. Corson, Am. J. Phys. 24, 126 (1956).

658

Figure C-1. A circular, nonconduct-

ing, nonmagnetic disk rotating with
angular velocity w about an axis
perpendicular to its plane and
parallel to a uniform magnetic
field B.

The voltmeter reading divided by the area of the loop is thus the
component of the curl in the direction of the normal to the loop,
if the loop is small enough. The “‘B-meter”’ can be a cathode ray
tube with the deflection of the electron beam on the tube face cali-
brated in teslas. The stop watch is used to measure the time rate
of change of B. The observers know nothing about the magnetic
field except what they measure with their own instruments.
The laboratory observer measures a uniform magnetic field:
B, = B. He determines its direction and magnitude by observing
the deflection of his electron beam for at least two mutually perpen-
dicular orientations. He also observes that, for each orientation, the
beam deflection is time-independent. Then 0B,/dt = 0. Further-
more, his ‘‘curl-meter” reads zero for all orientations, since there is
no changing flux through the loop, and thus

VY Xi = =e = th (C-3)

When the disk observer points his electron beam in the plane
of the disk, he always records the same deflection, no matter where
he is on the disk. When he points it parallel to the axis of the disk, he
always records zero deflection. He therefore concludes that the mag-
netic induction is uniform and perpendicular to his disk. His value
Bp, is the same as that of the laboratory observer: Bp; = Br = B.
The “‘curl-meter”’ on the disk sees only a constant flux, and
V X Ep. = 0 (C-4)

everywhere on the disk. Thus, the disk observer finds

: OB
V X En = —— = 0. (C-5)

Now let us consider a second similar nonconducting and non-

magnetic disk D2 rotating with an angular velocity w about an axis
in the plane of the disk and perpendicular to the direction of B, as
indicated in Figure C-2. An observer on this disk, if equipped with
 659

Figure C-2. A nonconducting,

nonmagnetic disk rotating with
an angular velocity w about an
A axis in the plane of the disk and
perpendicular to the direction of
a uniform magnetic field B.

the same instruments as the other observers, will ascribe entirely

different properties to the field.
Let the observer on disk D2 establish a coordinate system with
its z-axis perpendicular to the plane of the disk and its x-axis parallel
to the axis of the rotation. When he points his electron beam in the
z direction, the x-component of the beam deflection measures (Bp»),,,
and the y-component measures (Bp»),. He finds that

(Bor): = 0, (C-6)
(Boo), = Bo sin wt. (C7)
When he points his beam in the y direction, the x-deflection measures
(Bp»)., and the z-deflection measures (Bp»);:
(Bp»), = 9, (C-8)
(Bp), = Bo cos wt. (C-9)
He finds these same fields no matter where he measures on his disk.
He can describe the field he measures as a uniform field By rotating
with angular velocity w about his x-axis. If he were to compare notes
with the laboratory observer, he would find that By) = Br = B.
What about his ‘“‘curl-meter’’? When the disk observer points
it so that the axis of the loop is in the x-direction, there is no flux
through the loop, and the reading is zero. Then
— WBo»)x
(V X Eps): = Pea ee 0. (C-10)

When he points the loop axis in the z direction, there is a changing

660 APPENDIX C

flux through the loop, and

(F
dt /p2
SS BSo stn. (C-11)
where 5S is the area of the loop. Then, for this orientation,

f Ep,-dl = BSw sin wt, (C-12)

and, from the definition of the curl,
(V X Ep), = Bw sin wt, (C-13)

(Ve Gon: bet Bes siiat. (C-14)

If the ‘‘curl-meter” reading is calculated for the y-component,

(V X Ep), = "Eos SS BN eOs cat: (C-15)

From this experiment we can see that
OB
VXE= me (C-16)

whenever E and B are measured in the same coordinate system.

Example EXPERIMENT 2
Let us now suppose that each observer is given some conducting
wire and told to try to arrange it so as to induce an electromotance
in a closed circuit at rest in his own coordinate system. For the
laboratory observer,
OB,
V x E, a Tar = 0 (C-17)

everywhere and, from Stokes’s theorem, the induced electromotance

p E,-dl = 0 (C-18)
for any closed path.
The observer on disk D1 has the same experience, and if D1 is
made of conducting material, there are no eddy currents induced
in it.
Again the situation is different in D2. Let the conductor be a
single loop around the rim of the disk, with a voltmeter connected
in series. The voltmeter can either be on the disk or in a fixed position
in the laboratory and connected to the loop by means of slip rings.
The voltmeter is read both by the observer on the disk and by the
laboratory observer. The former can calculate what the meter will
read by invoking the Faraday law, since he knows dBp:/0r:

f Ep:-dl ll= S eo (C-19)

= By)Sw sin wtf. (C-20)

INDUCED ELECTROMOTANCE IN MOVING SYSTEMS 661

He can also calculate the electromotance from Stokes’s theorem, since

he has measured V X Ep» for every point on the disk:

f Endl 2 i (Vc Beaedas (C-21)

= BSw sin wrt. (C-22)
The two results are the same, of course, since

VX Ep: = — 7B (C-23)
Now let us calculate what the laboratory observer thinks the
meter on the disk will read according to the Faraday law. He says
that db/dt differs from zero because the circuit is rotating in a time-
independent field, whereas the observer on the disk says that d®/dt
differs from zero because his circuit is in a time-varying field. The
laboratory man calculates di/dt through the rotating loop and gets

g Ee dl BISe ier, (C-24)

where his B; is the same as the By measured by the rotating observer.
The laboratory observer can also calculate the electromotance
using Equation C-1:
OB;
V X Em = aa +V xX (u X Bz). (C-25)
He must use this complete expression, since the path around which
he calculates the electromotance is moving in his coordinate system.
Since

i)
OB,
a = (C-26)
C-26

the equation
Va br) Vee a BD) (C-27)

holds everywhere on the disk or, excluding terms of zero curl,

Ep: =u x By. (C-28)

The laboratory observer then says that

f Eoudl x f (uw eB dL (C-29)

From Figure C-3, at an arbitrary point on the rim of the disk,
u = wr sin 6, (C-30)
lu X B,| = Brrw sin 6 sin wt. (C-31)
Thus

g (u X B_,)-dl Re Brw sin @ sin wt cos (5 — 0)rd@, (C-32)

2
ll SBw sin wt, (C-33)

as with the Faraday law.

The laboratory observer and the D2 observer therefore agree
662

Figure C-3. The vector product

u X B on the rim of a disk rotating
about an axis lying on its plane
and perpendicular to a uniform
magnetic field B. The vector u is
perpendicular to the plane of
the disk.

as to the voltmeter reading, but they disagree as to the reason for the
induced electromotance. The laboratory observer says that the mag-
netic field is static but that the magnetic force (uw xX B) on the free
charges in the moving conductor produces the electromotance. The
disk D2 observer, on the other hand, says that the conductor is at
rest but that it is in a time-dependent magnetic field.

Example EXPERIMENT 3

As a final example, let us consider the Faraday disk or, as it is also

called, the homopolar generator. Let us return to the disk Dl. We
place a conducting ring around its circumference and use a conduct-
ing axle. The axle and the ring are connected by a radial conducting
wire attached rigidly to the disk, and the circuit is completed by
brushes and a stationary wire with a voltmeter in series, as indicated
in Figure C-4.
According to the laboratory observer, 0B,/0t is everywhere
zero, and

f E-al 2 fw x B,)-dl. (C-34)

The part of the circuit that is stationary in the laboratory has u = 0.
On the rim and on the axle, (u X Br) is everywhere perpendicular
 663

Figure C-4. Homopolar generator.

to the path of integration, thus there is no contribution to the inte-

gral. Along the radial conductor, the laboratory observer finds that

f E-dl = ifmroBedn (C-35)

2
ae 28 (C-36)
With the Faraday law, it is important to specify carefully the
surface through which the flux is to be calculated. It can be any sur-
face bounded by the path of integration in the electromotance calcu-
lation. For convenience we may choose a surface lying in two planes,
as in Figure C-4. The only part of this surface in which the flux
differs from zero is the part that lies on the disk. The rate of change
of flux through this part of the surface is readily calculated.
If the observer on the disk calculates the electromotance, he
finds that 0Bp;/dt is also everywhere zero and that the only place
where (u > Bp;) differs from zero is in the portion of the circuit
external to his disk. He sees this part of the circuit rotating with re-
spect to the disk with angular velocity w. Again, since the only con-
tribution to the electromotance is in the radial parts of the circuit,
his calculation also gives

f E-dl = — (C-37)
2

The laboratory observer says there is an electromotance induced

in the circuit because of the magnetic force on the moving
charges of the disk, and the disk observer says there is an electro-
motance because of the magnetic force on the moving charges of the
664 APPENDIX C

portion of the circuit external to the disk, but they always agree on
the voltmeter reading.
If the whole disk is made of conducting material, the electro-
motance is calculated in exactly the same way. It makes no difference
what integration path we choose from the axle to the rim, as long as
the path is at rest relative to the disk. It is essential that this part of
the path be at rest relative to the conductor, since the charges which
experience the force resulting in the electromotance are, on the
average, at rest with respect to the conductor. The electromotance
is independent of the path in the conductor, since only the radial
components of the path elements contribute to the integral of
(u X B)-dl.
The electromotance may also be calculated for the conducting
disk from the Faraday law. Again we may choose any path in the
moving conductor, as long as it is at rest relative to it. The result
is the same as with the wire discussed above.
Our discussion is strictly valid only for large magnetic fields and
small angular velocities, since centrifugal and Coriolis effects have
been neglected. For practical laboratory purposes, however, this is
not a limitation. The ratio of the magnetic force to the centripetal
force on an electron in a disk rotating with angular velocity w in a
magnetic field B is
Fy, eB
itn (C-38)

For an angular velocity of 1800 revolutions/minute and a mag-

netic field of one tesla, this ratio is about 10°. This complete separa-
tion of electrical and mechanical effects is a consequence of the large
value of the ratio e/m for an electron.
APPENDIX D

The Exponential Notation

The subject of this appendix is a mathematical technique for solving what is

probably, from the point of view of the physicist, the most important class
of differential equations.
The sine and cosine functions play a particularly important role in
physics, mostly because of the relative ease with which they can be generated
and measured by the ordinary types of instruments. They are also relatively
easy to manipulate mathematically. All other periodic functions, such as
square waves, for example, are vastly more complicated to use, both experi-
mentally and mathematically.
The mathematical technique which we shall develop here is therefore
widely used, despite the fact that it can apply only to functions of the form
= X) COS (wt + 8), (D-1)
where Xy is the amplitude of x, w is the angular frequency, and t is the time.
The quantity (wt + @) is the phase, or phase angle, # being the phase at ¢ = 0.
We shall limit our discussion to the cosine function, since the sine can be
transformed to a cosine by an appropriate choice of 6. We shall assume that
the origin of time is chosen so that 0 = 0.

Dil THE jo OPERATOR

The procedure for differentiating cos wf, although elementary, is rather in-
convenient: the cosine function is changed to a sine, the result is multiplied
by w, and the sign is changed. To find the second derivative, the sine is
666 APPENDIX D

changed back to a cosine, and the result is again multiplied by w, but this
time without changing sign, and so on.
Differentiation can be simplified if it is kept in mind that

e! = cos wt + / Sin of, (D-2)

where j= V—1. Then

x= xX) cos ar = Re ye, (D-3)

where Re is an operator which means ‘‘Real part of’ whatever follows.

Let us calculate the first two derivatives:
ax ¢d =i
ai = di Re (xe? ), (D-4)

= Re (jwxe*), (D-5)
= —wXp Sin wf, (D-6)

oa = Re ((ia)me4}, (D-1)
= —w?Xy COS wh. (D-8)
The results are, of course, the same as by the usual method.
We now adopt the following convention: we shall express the quantities x,
dx/dt, d’x/dt?, and so forth, as exponential functions without writing the Re
operator, but with the tacit understanding that only the real part must be used.
Then x = xX cos wt will be written as

X = Xeirt,
7)
and

a = Jwxoe'* = jwx, ae

m7 = (ja)*xoe?™* = (ja)?x, ioe

With this convention, the operator d/dt can be replaced by the factor
jw.
This simplification is so useful that the exponential notation is almost in-
variably used to represent sinusoidally varying quantities, whether they are
mechanical, acoustical, or electrical.
The coefficient before the exponential function can itself be complex.
For example, jxoe”*’ means

Re (jxoe**") = —Xp Sin wf (D-12)

if Xo is real. Since j = e’"/?, one can also write that
Re (jxce*") = Re (xpetottn/2)) (D-13)
THE EXPONENTIAL NOTATION 667

= Xp COS (or+ =), (D-14)

= —X, Sin wf. (D-15)

One common case is

A ea aT : b
(a + jb)e' = Va? + b? expj (or+ arc tan :) (D-16)

where we have written

a+
: ees
jb = Va? + b? exp arc tan ({)
b (D-17)

in the usual manner, on the assumption that a is positive, in order that the
angle chosen lies either in the first or fourth quadrant. Thus

Re (a + jb)e#*t = Va? + b? cos (or+ arc tan ’) (D-18)

Warning No. 1: If a is not positive, one must be careful to use the proper
angle in the exponent. For example, the argument of (—1 + 2/) is (w — arc
tan 2), not arc tan (—2).
Space-dependent functions can also be represented with the exponential
notation. For example, in an electric field, the vector E is oriented in some
direction in space. It can also be a function of the time; for example

E = E, cos (wt + 8), (D-19)

where E, is a vector whose magnitude is the maximum value of E. We then

write
E = Eeie), (D-20)
= Eyeteist, (D-21)
The coefficient of exp jwt can therefore be, at the same time, a vector and
a complex quantity, the former property having to do with the orientation
of E in space, whereas the second property is related to the phase E in the
time dependence.
The exponential notation is used as follows. The sine or cosine func-
tions are expressed in the form xe’, which is of course equal to
Xo{cos (wt + 0) + sin (wt + 6)}. We are concerned only with the real part,
which is the first term; the second term can be considered as parasitic. Then,
as long as the mathematical operations on the exponential functions are
restricted to additions, subtractions, differentiations, and integrations, their
real and imaginary parts do not mix. The technique is thus useful for solving
linear differential equations with constant coefficients. Once the calculations
668 APPENDIX D

are completed, the resulting expressions are often left in exponential form.
However, if amplitudes and phases are required, the imaginary part is re-
jected, and the result is expressed again as a cosine function.

Warning No. 2: The exponential technique is valid only for mathematical

operations that do not mix the real and the imaginary parts of the exponential
functions. For example, for any two complex numbers A and B,

Re{ Aei#t + Beiot+)} = Re Ae! + Re Beret (D-22)

but
Re{ Aei#*Bet@#+)) =~ {Re Ae™'} {Re Bet} (D-23)

Whenever multiplications of e’*' terms are involved in the calculations, one

must revert to the cosine functions and not use the exponential technique.

Example SOLVING A LINEAR DIFFERENTIAL EQUATION

WITH CONSTANT COEFFICIENTS, USING
THE EXPONENTIAL NOTATION
As an illustration, let us consider the following differential equation,
where all the terms are real:

a a +8 “ + yx = Xcos wt. (D-24)

This could be the differential equation describing the motion of a

mass a under the influence of a force X cos wf, a restoring force
—x proportional to the displacement x from equilibrium, and a
damping force —8 0x/dt proportional to the velocity 0x/dr, as in
Figure D-1a: the product of the mass a and the acceleration 02x/0r?
is equal to the sum of the applied forces.
It could also be the differential equation for the charge x on a
capacitor whose capacitance is 1/y in series with an inductance a
and a resistance 6, for an applied voltage XYcos wf, as in Figure

—— TUT ee

E = X cosat pe

X cos wt i |[2

(b)

Figure D-1. Examples of oscillating systems.

THE EXPONENTIAL NOTATION 669

D-1b: the applied voltage is equal to the sum of the voltages across
a, B, and 1/y.
This equation can be solved, without using the exponential nota-
tion, as follows. We consider only the steady-state solution and neg-
lect all transient effects obtained by setting the right-hand side equal
to zero. There is, on the left, a sum of three terms involving the
unknown function x, its first, and its second derivatives, with con-
stant coefficients. Since the derivatives of the sine and cosine func-
tions are themselves sine and cosine functions, it is plausible to try a
function of the form
x = Acoswt + Bsin wt. (D-25)

Substituting this expression for x in Eq. D-24, and setting the coeffi-
cient of cos wt on the left-hand side equal to X and that of sin wr
equal to zero, one can determine the coefficients 4 and B. The sum
(A cos wt + B sin wf) is then put into the form C cos (wt + 6), and
the result is as follows:

x= a cos | wt — arc tan —*). (D-26)

Vy = aw)? + Bo? Y= ate
We have assumed that y > aw? in writing out the arc tan term;
otherwise this expression would not be correct. See Warning No. 1.
The procedure is much simpler with the exponential notation.
Substituting jw for the operator d/dt in Eq. D-24, we can write
directly that
Xeivt
= . D-27
© Saw + joB +7 O28
The amplitude and phase of x can be found easily:

a a a
jwt
(D-28)
Vy = a”)?+ 8%? exp
exp (are
( arc tan
tan ~ 8”)

=
= Me = as
oo)? ieBa pie (
iL wt ee
arc tan _ oe5)

+ j sin (wr— arc tan 5 pe 7 (D-29)

If we reject the imaginary part, we obtain Eq. D-26.

The value of dx/dt corresponding to the velocity in Figure D-la
and to the current in Figure D-1b can be calculated easily:
abe.
ee en ee jwXei* D-30)
dt Haw?
+ joB +7
or, rationalizing and again rejecting the imaginary part,

ee 5 « fwt + arc tan Ae~ ip (D-31)

au")|

Gast :
sh ee
APPENDIX E

Waves

A wave involves the propagation of a disturbance of some sort in space. One

can think, for example, of the waves formed when a stretched string is fixed
at one end and moved rapidly in a vertical plane in some arbitrary way at
the free end.
If we call y(t) the vertical position of the moving end, then it turns out
that the vertical position y at a distance z along the string is y{t — (z/u)},
assuming no losses and a perfectly flexible string. The vertical position at z
is thus given by the position at the moving end at a previous time t — (z/u),
the quantity z/u being the time required for the disturbance to travel through
the distance z at the velocity uw.
More generally, we can consider waves propagating in an extended
region, such as acoustic waves in air or light waves in space. The quantity
propagated can be either a scalar or a vector quantity. For example, in an
acoustic wave, we can consider the propagation of pressure, which is a scalar.
In an electromagnetic wave, we can consider the propagation of the electric
field intensity vector E or of its components E,, E,, E:.

F.1 PLANE SINUSOIDAL WAVES

If a certain quantity a propagating with a velocity u is given at z = 0 by

a = ay COS at, (E-1)

then, for any position z in the direction of propagation of the plane wave,
WAVES 671

a = ay COS w (:= ‘). (E-2)

This expression describes an wnattenuated plane sinusoidal wave, since the

amplitude ap is constant, and since a depends on z but not on x or on y.
The wave fronts, which are surfaces of constant phase at a given time, are
thus perpendicular to the z-axis. The quantity ao is called the amplitude of
the wave. For a given position z, we have a sinusoidal variation of a with
time. For a given time ¢, we also have a sinusoidal variation with z, as in
Figure E-1.
The phase angle shown between brackets is a constant, that is

t— 2 = Constant, (E-3)

for a point traveling with the velocity

cl (E-4)
The quantity u is called the phase velocity of the wave, since it is the velocity

with which the phase w (:= :)is propagated in space.

u
We often write
a = ay COS (wt — kz), (E-5)
where
a (E-6)
u
is called the wave number. It is important to note that this wave number is 27
times that used in optics. For this reason, the quantity k is also called the
circular wave number.
The wave length ) is the distance over which kz changes by 27 radians,

(a) (b)

Figure E-1. The quantity a = ao cos (wf — kz) as a function of z

and as a function of f.
672 APPENDIX E

as shown in Figure E-la:

kx = 20. (E-7)
If we write
r
ee r, (E-8)

where X is read as “‘lambda bar,”

|
— ti (E-9)

The quantity X is called the radian length. It is the distance over which
the phase of the wave changes by one radian; this is about \/6, and is there-
fore considerably less than one-quarter wavelength. It turns out that the
quantity that appears in nearly all the calculations is A, and not \. We are
already familiar with the fact that it is preferable to use the circular frequency
w instead of the frequency f. In both cases the intuitively simple quantity,
namely, the wavelength or the frequency, is not the one that is “natural”
from a mathematical standpoint.
With the exponential notation, the wave traveling in the positive direc-
tion along the z-axis is written*
a = ayexp j(wt — kz). (E-10)
Similarly, a wave traveling in the negative direction along the z-axis is
described by
= ao exp j(wt + kz). (E-11)

We shall have occasion to use plane sinusoidal waves traveling in some

given direction specified by a unit vector n. The wave fronts are then normal
to m, and such a wave is given by
a = aexp j{wt — k(n-r)}, (E-12)

as can be seen in Figure E-2. When n coincides with the unit vector in the z
direction, n-r = z.
If a plane wave traveling along the z-axis is attenuated, its amplitude
decreases exponentially, and

a = ao exp {j(wt — k,z) — k,z}. (E-13)

* We shall use the notation of Eq. E-10. Notice, however, that one could equally well
write
a = a exp {—j(wt — kz)},

since the cosine is an even function. The latter notation is frequently used where it is con-
venient to omit the factor e~’*¢ for brevity. A wave traveling in the positive direction is then
simply written
a= a exp jkz.
 673

Figure E-2. The plane shown is

defined by n+r = constant, the
constant being the distance
between the plane and the origin.

It is now k, that is related to the wave length:

k, == (E-14)

whereas the quantity k;, which is called the attenuation constant, is such that
the wave is reduced in amplitude by a factor of e in a distance

Sans (E-15)

i= ie (E-16)

We can also rewrite Eq. E-13 in the form

a = ayexp j(wt — kz), (E-17)

where
k = k, — jk. (E-18)
It is important to note the negative sign in the above expression.
The quantity k is still called the wave number in this general case. How-
ever, it is complex, its imaginary part corresponding to absorption. It will be
observed that an attenuated wave traveling in the positive direction along
the z-axis requires that the real part of k be a positive quantity and that the
imaginary part be a negative quantity. Otherwise, the wave would grow
exponentially in amplitude with increasing z. The quantities k, and k, in
Eq. E-18 are thus both positive quantities. In transmission line theory the
quantityjk is written y, and is called the propagation constant.
674 APPENDIX E

An attenuated wave traveling in the negative direction along the z-axis

is given similarly by
R =| a exp {j(wt + k,z) + k,z}, (E-19)
ay exp j(wt + kz), (E-20)
where k, and k, are again positive quantities and where k is defined as above.
We shall meet with cases where k, = 0. Then Eq. E-13 becomes
a = ay exp (—k,z) exp jot. (E-21)
The phase angle wf is then independent of z, all points are in phase, there is
no traveling wave, and the amplitude decreases exponentially with z.

E.2 WAVES ON A STRETCHED STRING.

THES DIFFERENTIAL BOUALION
FOR AN UNATTENUATED WAVE

It is interesting to consider at this point the simple case of transverse waves

propagating along a stretched string. We assume small transverse displace-
ments on a flexible string of mass p per unit length stretched with a tension F,
as in Figure E-3.
Figure E-3 shows an element of the string at some given time f¢. Both its
displacement y and its angle @ are functions of both the position z along the
string and the time t. We assume that there is no motion along the z-axis, in
order that the stretching force F be constant all along the string.

F/

eo
= \

Figure E-3. Element of string

stretched with a force F. The
angles and the displacement from
| z the z-axis are grossly exaggerated
Zodz for clarity.
WAVES
675

2
The element of mass p dz takes on an acceleration a under the action

of the forces F at either end, and

p dz@ " = F{sin| (0 + dd) — sin 6, (E-22)

= F d(sin 6), (E-23)
= F cos 6 do. (E-24)
Then
Rigo)
cos 6 a5 we Fae (E-25)

We have written a partial derivative for 6, since the dé found above was for
a given time ¢. For waves of small amplitude, we can set cos@ = 1 and
6 = dy/dz. Then

Oz
ee
F or
(E-26)

We can verify that the above differential equation does correspond to a

wave motion by substituting for y any function of {t — (z/u)} or of
{t + (z/u)}. We assume, of course, that the second partial derivatives exist.
We find that the phase velocity wu is

“= (=)". (E-27)

This result is general, and any uwnattenuated wave traveling with a velocity
u along the z-axis is described by the following differential equation:

(E-28)

The differential equation for an attenuated wave will be discussed in Sections

E.5 and E.6.
It will be observed that the above differential equation (E-28) is much
more general than Eq. E-10. It does not involve the amplitude ap, the angular
frequency w, or the wave number k, but only the velocity u. The differential
equation therefore applies equally well to any wave form, periodic or not,
and to any amplitude. Equation E-28 is also independent ofthe sign of u, since
it involves only uv”. It can therefore represent waves traveling in either direc-
tion along the z-axis.
For a sinusoidal wave.
676 APPENDIX E

ge ees
Oa w?
(E29)
0a
——
. = Ss E-30

E.3 7SOLUTION OF THE, DIFFERENTIAL EQUATION

FOR AN UNATTENUATED WAVE
BY THE SEPARATION OF VARIABLES

Equation E-28 can be solved formally by the method of separation of vari-

ables (Section 4.4). We set
a = T(t)Z(2), (E-31)
where 7 and Z are respectively functions of ¢t and z only. Substituting this
value of a in Eq. E-28 and dividing by TZ,
paz. a Ire
(E-32)
Zdz* wT dt?
The term on the left is a function only of z, whereas that on the right is
a function only of ¢. Then both sides of the equation can be equated to a
separation constant —k?:
ieee :
Z dz eg ae!
Dal eee.
aie oa oe (E-34)
and
T = A exp (jukt) + B exp (—jukn), (E-35)
Z = Cexp (jkz) + D exp (—jkz). (E-36)
Then

a = ACexp {jk(ut + 2)} + BD exp {—jk(ur + 2)} + ADexp {jk(ur — z)}

+ BC exp {—jk(ut — z)}. (E-37)
Now a is some physical quantity that must not increase or decrease
exponentially with either z or t. This condition requires that k be a real
number, The exponentials then reduce to sine and cosine functions.
Since the coefficients AC, BD, and so on, are as yet undetermined con-
stants, we can set AC to be some complex number

AC = Gel®, (E-38)
where G and ¢ are real. Also, a must be a real quantity, and we must have
WAVES
677

BD = Ge-ie (E-39)
in order that the first two terms of Eq. E-37 can add up to give a cosine
function.
Similarly, we can set
AD = Fe’ (E-40)
and
BG whes. (E-41)
Finally, a can be written as follows:

a = 2F cos (kut — kz + 0) + 2G cos (kut + kz + 4g), (E-42)

where the angles @ and ¢ are constant. Comparing this with Eq. E-5, we find
that the above equation determines a pair of waves traveling in opposite
directions with a common velocity u. The separation constant k is the wave
number, and
ia — ae Dry, (E-43)

Since a is a function only of z and of t, the waves are plane, with wave
fronts parallel to the xy-plane.
It will be observed that the formal solution that we arrived at by sepa-
rating the variables z and ¢ has led us to a very special class of waves; namely,
sine waves. We did not find a general function of {t — (z/u)\. This is quite
disturbing at first sight, since our formal solution is presumably general.
There is no contradiction, however, for the following reason. Since our dif-
ferential equation is linear, that is, since its terms are all of the first degree
in a or its derivatives, the sum of any number of solutions is also a solution.
Any type of periodic wave form encountered in practice can be expressed
as a Fourier series of sines and cosines of the fundamental frequency and
of its harmonics (Section 4.4). Even individual nonperiodic pulses can be
analyzed in a somewhat similar manner by means of Fourier integrals. Any
wave form can thus be synthesized by combining terms of the form shown in
Eq. E-42 with appropriate amplitudes and wave numbers.

E.4 REFLECTION OF A WAVE ON A STRETCHED

STRING AT A POINT WHERE THE DENSITY
CHANGES FROM p, TO p:
If two strings of different densities p; and p, are tied together and stretched
with a force F, as in Figure E-4, a wave traveling along the first section will
678

rem PEIPIIPPIDIE
LOOTED DOPE
F P1 0 P2 F

Figure E-4. Two strings of densities p; and p: fixed together at O

and stretched with a force F.

be partly reflected and partly transmitted at the knot. The phase velocities are

“(9
respectively
F 1/2

Pl
1/2
(E-44)
Wu = () 5)
p2

and the corresponding wave numbers are

Py w 7 pi 1/2
ky = Uy =» (%) ’

(E-45)
a) |
ky tS: \>
Uy 55]
=

Let us assume that a wave travels to the right along string 1. We shall
call this the incident wave and set

Yi = Yo: exp j(wt — kxz), (E-46)

y being the lateral displacement of the string and z = 0 being chosen at the
knot. For the wave transmitted to string 2,

Ye = Yor exp (wt — kyz). (E-47)

Finally, for the wave reflected back at the knot,

Vr = Yor exp j(wt + kyz). (E-48)

We assume, for simplicity, that there are no reflected waves originating

at the supports.
In the above three expressions for );, yi, y-, the amplitude yo; of the
incident wave can be assumed to be known. Thus we have two quantities to
determine: yo, and yo,.
It is possible to calculate the values of yo, and yo, in terms of poi, p1, ps
by considering the conditions of continuity which must be satisfied at the
WAVES 679

knot. First, there must, of course, be continuity of the displacement y: the

value of y just to the left of the knot must be equal to its value just to the right.
Then
yityr=y at z = 0,
or

Voi + Yor = Vor- (E-49)

Second, there must be continuity of the slope of the string dy/dz. The
reason for this is as follows. We have assumed implicitly that the knot was
weightless. Thus the sum of the forces acting on it must be zero, and the two
opposing tension forces F at that point must be along the same line. Then

—kyyoi + kyyor = —kayor. (E-50)

Solving these equations,

Jor ki — k pie = p2’”

yi kx the pl? + oh? Coy
= = - =9 E=

Jot 2k, 2pi”

(E-52)
Joi i ky + ke je pi Sic ae

Since the ratio yo/yo: is always real and positive, the transmitted wave is
always in phase with the incident wave. On the other hand, the ratio yo,/yo;
can be either positive or negative. The reflected wave is in phase with the
incident wave if p; > p2, and m radians out of phase if p1 < po. If p, = py there
is no discontinuity, no reflected wave, and yo: = oi.

ES WAVES ON A STRETCHED STRING WITH

DAMPING” [HE DIFFERENTIAL” EQUATION
FOR AN ATTENUATED WAVE

Let us return to the case of the stretched string of uniform density p. We

assume now that the string is in a viscous medium that provides a damping
force that is proportional to the velocity. The damping force on the element
of string dz is then

Ah ee 2, (E-53)

and, from Eq. E-24, we now have

0*y Oy
pdz ——
ry 2 «= F cos
6 dé a— bd. ahAt (E-54 )
680 APPENDIX E

or

Oy poy bey '

dZ* FF or 1:F ot Ca

We can show that this is the differential equation for an attenuated wave
by trying a solution of the form

y = poexp j(wt — k’z). (E-56)

Substituting, we find that

2
kit = SE (1-j2), (E-57)
F wp

and the wave number k’ is now obviously complex. The above differential
equation is therefore that of an attenuated wave. This is to be expected, since
it is identical to Eq. E-26, except for the addition of the second term on the
right-hand side. This term corresponds to a damping force which dissipates
energy.
Equation E-55 can also be rewritten as

os > —“e (1-i2) Re har (E-58)

or
C2
(S 7 =)y=0. (E-59)

This equation is similar to Eq. E-30, except that the wave number k’ is now
complex. It represents a pair of attenuated waves traveling in opposite
directions along the z-axis.
It will be observed from Eq. E-57 that the differential equation is equally
well satisfied by +k’ or by —k’, that is, by waves traveling in either direction
along the z-axis.
We can find both k, and k; by recalling Eq. E-18 and substituting

kK! = k, — jk; (E-60)

in Eq. E-57, which gives

/' ® \ 1/25 1/2

Ke, T R=, : 1+ (Ite =i) iter (E-61)
aa €
<ly pete: wp”

ae Pp 1/2 s ioe V2) V2

ki = + (4) es (1+ =a) 1 (E-62)
WAVES 681

Since, by definition, k, and k; are both positive and real (Section E.1), the
+ signs must all be replaced by + signs, and

ke a bh? 1/2) 1/2

k,
ol) U+( +a) adi
p yay b? 1/20 2/2
ki
°(e) (1+ (tam) Coe
It will also be recalled from Section E.1 that the velocity of the wave is
w divided by the real part of the propagation constant. This reduces to
(F/p)? when b = 0, as expected.

E.6 SOLUTION OF THE DIFFERENTIAL

EQUATION FOR AN ATTENUATED WAVE
BY THE SEPARATION OF VARIABLES

The differential equation E-55 that we found above is general, and any
attenuated wave traveling along the z-axis is described by
07"
az?
07a"
oF
hoe (E-65)
We can solve this equation formally by separating the variables, as in Section
E.3. We set
gael ()Z.(2), (E-66)
substitute in Eq. E-65, and divide by T’Z’. Then
1 @Z' gda@T' , hdr’
(E-67)
Fwd a Te dpe Taf
where the left-hand side is a function only of z, whereas the right-hand side
is a function only of ¢. Then
Len F
Say = OK E-6
Li az ee Wea)

BOT) oh al si :
Tier T 7 dt we Cad

From the first of these equations,

Z' = C’ exp (jk’z) + D’ exp (—jk’z), (E-70)

and k’ is the wave number.

The second equation takes on a more familiar form when rewritten as
682 APPENDIX E

GES!
+h = 55 ip a) (E-71)
S dP
We can assume a Sinusoidal wave without losing generality, as was noted in
Section E.3, and set

T' = A’eist, (E-72)

The quantity » must be real in order that the amplitude of the wave can
decrease only with z. Substituting in Eq. E-71, we find that
h
k (i BO)wg (1a y «):
ial é
(E-73)

This result is identical to that which we found in Eq. E-57, the coefficients
g and / of the general differential equation for an attenuated wave (Eq. E-65)
being equal, respectively, to p/F and to b/F in the differential equation for
an attenuated wave on a stretched string (Eq. E-55).
If we again set
ki = ky — jki, (E-74)

“OGY ex
we find that

Ee

k= 0 (§) {1+ (1445) }" (E-76)

as in Eqs. E-61 and E-62, since g = p/F, andh = b/F.
Finally, from Eqs. E-66, E-70, and E-72,

a! = A’C’ exp j(wt + k’z) + A’D’ exp j(wt — k’z), (E-77)

where k’ = k, — jk;. The first term represents a plane wave traveling in the
negative direction along the z-axis, whereas the second term represents a
similar wave traveling in the positive direction.
In this general case, the wave velocity w is w/k,, which reduces to (1/g)!/?
when h = 0. The wave is attenuated by a factor of e in a distance 6 = 1/k,,
which approaches infinity as A approaches zero.
These results are the same as those of the preceding section.

E.7 WAVE PROPAGATION IN THREE DIMENSIONS

In the case of an attenuated wave propagating in space, the wave equation is

similar to Eq. E-65, except that the second derivative with respect to z is
WAVES 683

replaced by the Laplacian

. = 07a
Ven ter +h 0a
ae (E-78)

The coefficients g and / again determine the wave number as in Eqs. E-75
and E-76. For sinusoidal waves,

V°a = (— gw? + jwh)a, (E-79)

Or

a + ka = 0, (E-80)
where k is the wave number. We have omitted the primes that were used
previously to identify the attenuated wave.
If there is no attenuation, / = 0, g = 1/u?, and

1 0a
Ae are (E-81)
for any waveform, u being the phase velocity.

B.8. WAVE PROPAGATION OF A VECTOR QUANTITY

As yet, we have only considered waves in which the quantity which is propa-
gated is a scalar. Vector quantities, such as an electric field intensity KE, for
example, can also propagate as a wave, and then, for a plane wave propagat-
ing in the positive direction of the z-axis,
E = Ey,exp j(wt — kz). (E-82)

Since E, is a vector, we may write that

E= (Ect = Loy = Ev-k) exp (wt +? kz), (E-83)

where Ey,, Eo,, and Ey. are the components of Ey. These components may
conceivably depend on x and on y, but they do not depend on z, because a
plane sinusoidal wave is characterized by the dependence on ¢ and on z
which is shown in the exponential function.

E.9 THE NONHOMOGENEOUS WAVE EQUATIONS

In the absence of attenuation, the scalar nonhomogeneous wave equation is

of the form

a Boe = SOY, 2's

0° he / /
(E-84)
684 APPENDIX E

where the function f describes a disturbance that causes a wave to propagate

in space. Thus, points where fis not zero are inside the source. The variable a
is measured at a field point x, y, z, while a point inside the source is identified
by the coordinates x’, y’, 2’.
Outside the source we have the homogeneous wave equation
0’a
Va — err 0. (E-85)

Although a thorough mathematical discussion of the nonhomogeneous

wave equation can be rather elaborate, its solution becomes intuitively quite
obvious if we proceed as follows.
In the special case wheref is not a function of the time,

Vea = f(x’, y’, 2’), (E-86)

and a@ is some function of x, y, z only. Now this is Poisson’s equation, Eq.
2-31, which we use in electrostatics, if a is the potential V and f is minus
the charge density divided by eo. The solution of this equation is well known.
It is

a= —
a 1 ee,
z
y’, z’)
dr /
(E-87)
as in Eq. 2-19, where

Te (Nae) eC ye ke (E-88)
is the distance between the source point x’, y’, z’ and the field point x, y, z.
If now fis a function, not only of x’, y’, z’ but also of r, we must take
into account the time required for the disturbance to travel from x’, p’, z
to x, y, z at the velocity u, as explained in Section 10.2.1, and thus

a=-_ ey J 2~—
(~,Ley
—_lee“w
‘) (E-89)
This is the solution of Eq. E-84.
The solution of the vector nonhomogeneous wave equation
CE

VE & = fx’, DS Zi, t)

or"
is, similarly,
ea ae 2
1 f(xy 21 ") '

E=—-—— | —————‘dr’. (E-90)

ANSWERS

In many cases the answer is included in the statement of the problem; the list below
contains one half of the remaining answers.

CHAPTER |

1-10. r = 433ri + (250¢ — 4.907%),

we 4337 (250 — 9:800;!
a = —9,80j.
1-16. Correct

1-26, |\V < Fl—= 1.4 Ar.

CHAPTER 2
Dem) a5.O me VOL metens
(b) 5.1 meters.

2-4. There is zero field at (—5.83a, 0, 0).

2-8. (a) 29.6 femtometers (1 femtometer = 10~'> meter).

(b) 41.5 newtons.
(C6324 >G 1023 g,

Mili, We = aoe z+ Vo, where Vo is an arbitrary constant.

0

2-13. (a) 2.32 X 10-7 coulomb/meter’.

(b) Inside, E = 1.31 X 10*r volts/meter.
Outside, E = 1.31 X 10~?/r volt/meter.
 ANSWERS

V E
IK10,° 15

1OxX10 16

Sal 9 Si
E
V

0 0 5 10 Figure A-1. Solution to Problem

Millimeters 2-13c,d.

Inside, V = —6.6 X 10%(r? — 10®) + 3.02 X 107? volt.

Outside, V = —(6.04 + 1.31 Inr)10~ volt.
See above figure.
e
2-19. F = ————_
ele) 4ne,a®
(b) Simple harmonic motion.
(c) f = 101 hertz.
(d) Same order of magnitude. See Figure 11-1.

2-21. (a) Q = (8/15) mpoa’.

_ 2poa*
CD) ees 15er?

_ 2p9a'
ee 15eor

= Pe ery
(©) ae €0 3 =)

(e)

0 2 3 4 5
Figure A-2. Solution to Problem 2-21e.
ANSWERS 687

2-24. (4/3)ma%ao.

eae [tae +S] -

2G
Waite Arreor’

Vor 03

De
7 967reor®
ieay
(c) Since 3n? — 1 is of the order of unity,
V3
alge
1
V7, < MES:
i00 lopip >2a Det.

2Tr€9
* In (6/a)
(CQ. — C01)"
Ze CCAG + C)
(b) The energy is dissipated as heat in the wires connecting the capacitors.

3e? e
2-34. (a) Tine (b) Srregmc?
2-36. (a) 150 kilovolts. (b) About 4 « 10~4 atmosphere.

2-39. 70 2‘
(a) in G/a) (b) The force is due to the fringing field.

2-42. (a) 20 centimeters. (b) 44 centimeters.

CHAPTER 3
3-1. (a) p 5.7 & 107%? coulomb-meter.
(b) s = 35.9' X 10-2 meter.

3-6.
} E D

1500 15x10* 15107’ [

|
1000 ~=10x10° 110x107)
\.- £ without dielectric
VA
é VA, —V without dielectric
500. Seal Sx10"* ~~
NS

r (centimeters)

Figure A-3. Solution to Problem 3-6.

688 ANSWERS

3-10. (a) €, = K/p.

(b) p, = A/2nKp, A being the linear charge density on the inner conductor.

3-13. (a) 1.33 & 10-* coulomb/meter?.

(b) «AR a = [ta + (tao + aCOS wt)]o, — taoa, Where A is the area of the film,

t, its thickness, and o, its charge density; o is the charge density on the
electrodes; and t., + a cos wt is the thickness of the air gap. Note that
AR(dcz,/dt) is the output voltage.

(c) 1.53 & 10% coulomb/meter?. (d) 6.1 volts.

3-15. 1.24 & 10-4 coulomb/meter‘.

3-17. (b) Assuming a 1 HP motor operating for one hour at 100% efficiency, the
capacitor would have a volume of the order of 10 meters* and would
weigh about 10 tons.

3-19. (a) 3.20 X 10~%/p* newtons/meter®.

3-22. (a) - CV (b) 1.1 tons. —(c): 0.37 ton.

3-24. 44 micrometers.

CHAPTER 4

4-1. 4.6 x 10° coulombs.

4-4. (a) m0. 8 (c) Forn Ym, EV /(re — ri) is large. For mn Kr,
E is large at the surface of the small inner sphere.
(CRON <0) 323 (f) 4.15 X 105 volts/meter.
4-8. 2.4 104 volts/meter.

4-21. (a) Yes.

(b) |df/dr| becomes larger and the orbits can become unstable.
(c) Yes.

4-24, No.

4-28. (b) 3.3 & 107% newton, or one third of a gram-force.

Electron source

on

|
|
|
|
|
|
|
Me
Figure A-4. Solution to Problem 4-28e.
ANSWERS 689

(f) About 5.5 microseconds.

CHAPTER 5

Sail, ayes

5-4. O says that A and B emitted their light signals simultaneously.

O’ says that B emitted his signal before 4.

5-6. (a) tan a, = y tan a.

(b) 90°.

5-8. (a) 0.363 c.

(b) 0.80 c.
5-10. Diameter < 10!* meters.

5-18. Dme*
|
|
(a) I
|
|

= .

Explosion Time

=Imyc

(b)
|
|

We Time A
Explosion Figure A-5. Solution to Problem 5-18.

5-22. (a) 1.78 K 107 microgram.

(b) 30 seconds.

5-25. 9.34 x 108 electron-volts.

5-28. (b) For the Sun, Av/y = —2.1 X 10~°.

For the Earth, Av/y = —7.0 X 10°".
(c) About 10!! kilograms/meter®.
(d) +6.9 xX 107°.
(e) The two shifts are of the same order of magnitude.

5-33. (a) AO/O = —4 X 10.

(b) Yes. See the example in Section 5-20.

CHAPTER 6
6-3. (i) 2.31 < 10°” newton.
(ii) 7.80 < 10~*% newton.
690

6-8. (d)
155

; B
(a) (b)

Figure A-6. Solution to Problem 6-8d.

(e) 2.5 X 10-8 meter/second eastward.

(f) Westward.

6-13. E = 1.45 x 10-5

(—4.34 x 10% + J)
volt/meter.
(19.2 X 1082 + 132
B = 7.00 X 107%
k tesla.
(19.2 X 1082 + 1?
6-15. (a) Inside, EK, = yUB.j, B, = yBok, if Bo = Bok.
(b) Outside, E, = 0, B; = 0.
(e) Yes.
(f)

Figure A-7. Solution to Problem 6-15f.

6-19. v = 0.27 meter/hour.

6-23. (a) A = 1.46 meters.

D = 5.0 millimeters.
\ = 3.28 X 10-! coulomb/meter.

(d) Yes:
dr Cc
ae 10! at x = 400 meters.
G ©
ANSWERS
691

(e) No.
(f) We have neglected end effects on the cylinders of charge.

GHAPTER 7
7-4. 0.234 newton.

7-10. (a) Perpendicular to the plane passing through the two axes.
(b) Yes.
21245 i
7-13.
T a

7-16. (c) The electrons form part of a neutral plasma.

7-19. V = (e/m)B*(b? — a®)*/8b2, where e and m are the charge and mass of the
electron, a is the radius of the cathode, and 4 that of the anode.

7-24. (a) ex-Bor.

(b) px = —2ex-Bw
oh = €0xX-Bua,
where a is the radius of the cylinder.

7-26. (a) _ ual ( HR (=+ HN

A= is In LAF 75

(b) It is only the axial component of the current which matters.

(c) No; to calculate the derivatives of A we require 4 as a function of p, ¢, z.

OB awe 3pola’z
7-29. (a) “Op a 4a? + z2)5/2

3yola*zp
(0) By = Hat + Z25/2

OB, 3uUola*p

gee Gl aye
2 3(a2 — 42”)?
(@ B= xe
5 tle
+ ml t aa sxeh
(1 +
73a. (C) 1.15 AO tesla,
(e) 4.83 & 10-§ ampere/meter?.
(f) 6.14 * 10 ampere.

CHAPTER 8
8-1. 1.2 millivolts.

mroLdE , trr’aR V
is E=-—>
ee l at I /

where L is the inductance, o is the conductivity of the wire, r is its radius, and
/is its length.
692

8-8. (b)

Figure A-8. Solution to Problem 8-8b.

(c)

Figure A-9. Solution to Problem 8-8c.

ney ne Figure A-10. Solution

to Problem 8-8e.

8-11. (d) The transverse velocity decreases.

8-17. L decreases with increasing frequency until the skin depth becomes small com-
pared to 6 — a.

8-71. eT (Ge) p

8-27. (b) 280 ampere-turns.

8-33. (c) Both are equal to the energy density and proportional to the square of the
magnitude of the field.
The electric force is directed into the field; the magnetic force is directed
away from the field.
The electric force per unit area is normal to the surface of a conductor and
is equal to €)?/2 only if the field is static.
The magnetic pressure is normal to the surface of a conductor and equal
to B®/2uo only if there is zero field inside the conductor. This condition is
achieved exactly with superconductors, and approximately if the frequency
is high.
ANSWERS
693

8-35. (a) 4.0 X 10° newtons/meter? = 4.0 atmospheres.

(b) The pressure would be unchanged since B is uniform inside a long sole-
noid.

8-37. (d) 3.2 X 10 teslas.

(e) 1.6 < 10° joules.

CHAPTER 9
9-5. “2.0.
9-8. (a) E is reduced.
D is unaffected.
(c) Minimum.
(d) B is increased.
H is unaffected.
(f) Minimum.

(a) &Wl In £.e

9-10. ORE
(b) Ay l|= XmI/27b, parallel to J.
Ne = Xml/2mc, opposite to I.
(c) Zero.
(d) uol/2m7p. Removing the iron cylinder would have no effect on this field.

9-15. Inside, H = M/2 in the direction opposite to M.

B = woM/2 in the direction of M.
Outside, |
LoM ta?
B=
a? + D231?

where D is the distance to the center, in the direction of M.

9-20. Use a circuit similar to that of Figure 9-15 with A,, = 2.5 centimeters? and
L,, = 8.84 centimeters (each magnet is 4.42 centimeters long).

CHAPTER 10

10-1. No. The relaxation time is of the order of 10~!2 second.

10-498 = jigJa/4a G6 0),

l| Lodo iw) ASE aly Th tent
Sr al (0 ¢z<
~ ye ‘)

-yh/sa (2<4); a

Rime ate =2 = 0.293/a,

694 ANSWERS

2
Vie
Ae al
The other two equations (V-B = OandV X E = —0B/0f) lead to identities.
10-9. (c) B = pod.
(d) The magnetic pressure is poA?/2.
(f) Currents are induced which maintain ® constant.
10-12. (Cc) EV = E; WV = dH.

CHAPTER 11

11-3. 54.4 millivolts.

11-12. For © = 1, k, = 1.10(ep,)/2/Ao, ki = 0.455(e,-p,)1/2/Xo, 6 = 2X.
|E/H| is smaller than for ¢ = 0 by a factor of 1.19.
E leads H by 0.393 radian or 22.5 degrees.

For Q? > 1, k, & w(eu)!/? (1+

+ ai) a~ 5(u/ QE) RPL a%

|E/H| is smaller than for ¢ = 0 by a factor of G- =):

E leads H by 1/2Q radian.
11-19. Attenuation in db/meter:

20 kilohertz 20 megahertz
Sea Water 5.4 170
Copper 1.9 x 104 5.9 < 105
c
2 @ CE PE = Py
1/2
(b) u, =c {1- (2“
(e) um = 3.24 108 meters/second.
u. = 3.23 & 108 meters/second.
u, = 2.78 X 108 meters/second.
(f) 2.73 & 10% meters; 45 waves.

CHAPTER 12

12-1. (d) jk-E = p./e kB = 0;

keh — spe —jk X B = pom + jwoE/c?.
ANSWERS
695

12-9. (a) E(volts/meter) H(ampere-turns/meter)

Incident 42.4 0.115
Reflected 6.00 0.016
Refracted 36.4 0.131
(b) R = 2.01%, T = 98.0%.
12-12. (a) 6,/X» = 0.368, Die = Meili
(b) Eo, leads Eo; by 2.59 radians.
Eo: leads Eo; by 1.29 radians.
geal n\?] 2
1221 55(a) ee — at ~ [ _ (2) ] e

(b) FF
0.5

0.3

0.1
2 Ss ee ee ee
eee

0 0.2 04 06 0.8 1.0 Figure A-11. Solution to Problem

N2/Ny 12-15b.

‘218 (F) _ =C =f) cos 6; + mipro(5/%o)

Ey:/p lh (1 — f) cos 0; + myyo(5/Xo)

12-20.

Figure A-12. Solution to Problem

E H 12-20.

12-24. (a) few = 8.86 X 10°r.

rad

(b) The two forces are equal for r = 0.1 micron.

696 ANSWERS

CHAPTER 13
13-11. The skin depth is 1.5 microns; the coating should have a minimum thickness
of about 5 microns.

13-15. (a)

Reflecting surface

Figure A-13. Solution to Problem 13-1Sa.

(b) E is always zero in the planes marked x.

(c) If there were a reflecting surface at x1, we would have the n = 1 mode.
A reflecting surface at x. would give the n = 2 mode, etc.

13-19. Both methods can, in fact, be used. See, for example: Harvey, Microwave
Engineering (New York: Academic Press), p. 144.

13-23. } = A,-~o. This current flows inside the conductor, very close to the surface.

13-27.

Attenuation
No /2b (n = 2)
Jest a in Cee oe La a a

Type |e
a inten Uicain Wee iiemtin eae

2.84 xX 1.34 ZA2 52 0.800 0.525 10.8 4.4

1.872°X 0.872 PLANS) 1.48 0.800 0.530 16.4 19
L3722X O162296 2122 1.40 0.735 0.525 2a 8.9
0.900 X 0.400 = 2.25 isi! 0.800 0.528 34.2 14.5
0.622 X 0.311 2.00 1.45 0.762 0.525 48.0 20.2

13-30. (a) 1.8 megawatts.

(b) 4.24 kilowatts/meter in the outer conductor.
7.00 kilowatts/meter in the inner conductor.
 697

CHAPTER14

Mee ay

De
?

fee ose ee ee
|
|
Figure A-14. Solution to
Problem 14-2b.

1 + 3 cos? 6 A?
14-4. (a)
Sift Ue eer
1 + 7 cos? 6
(b) 2 sin? 6

(cae

14-8. 222 millivolts/meter.

14-14. Explore the field with a small electric or magnetic dipole.

14-18. No.

14-24. In air, the ratio of the output voltages is

In sea water,
Ve 4
Vinag 1 (opow)!!?
For a frequency of the order of 104 (see Section 12.5.1.

APPENDIX B
B-1. 6.44 * 1074 newton.
 Index

Accelerating tube, electric field inside Barnett, S. J., 373

an, 187 Beam wave guides, 593
Acceleration, expansion, 376 Bertozzi, W., 195
Accelerator Betatron, 373
beam divergence, 290, 379 Biot-Savart law, 295
for neutral molecules, 136 B-meter, 657
ion drag, 375 Boltzmann constant k, 118
Stanford linear, 242 Boltzmann equation, 177
Van de Graaff, 184, 329 Bondi, H. 188, 456
Acceptor atoms, 176 Boothroyd, A. R., 176
Adler, R. B., 176 Boundary conditions, 139, 400, 565
Alnico 5, 412 Braccesi, A., 639
Ampere, 293 Brackets, use of, 427
Ampeére’s circuital law, 310, 393 Brewster angle, 516, 519, 526, 552
Amperian currents, 386 ff. Brillouin, Léon, 491
Amplitude, 665, 671 Buchsbaum, S. J., 498
Analytic function, 646
Angle, solid, 48 Capacitance, 76
Angular frequency w, 665 Capacitor
Anisotropic medium, 422 cylindrical, 651
Annihilation of electrons, 221 leaky spherical, 434
Antenna parallel-plate, 80, 110, 124, 136, 263, 270,
arrays, 616 289, 485, 651
dipole, 428, 595, 627, 638 Cartesian coordinates, 1
gain, 637 Cathode ray tube, 525
halfwave, 611, 636 Cauchy-Riemann equations, 646
loop, 430, 623, 627, 629, 638 Causality, 214
quarter wave, 636 Characteristic values, 560
receiving, 627 Charge conservation, 233, 422
Atmosphere, electric field intensity Charge density, bound (or induced,
in the, 183 or polarization), 96, 101, 108, 109
Attenuation, 578, 591, 673 at dielectric-conductor boundary, 111
Attenuation constant k;, 673 Charge density, free (or conductible), 96,
Attenuation distance 6, 472, 673 108
Avogadro’s number Na, 116 at dielectric-conductor boundary, 111
700 INDEX

Charge density, total, 109 Current density

Charge invariance, 245 equivalent, 386, 396, 423
Child-Langmuir law, 190 in matter, 423, 439
Circuital law, 310 polarization, 96, 423
Class-A dielectrics, 104, 108 Curvilinear coordinates, 23 ff.
Clausius-Mossotti equation, 115 Cut-off wavelength, 573
Clock Cyclotron frequency, 287, 497
paradox, 240 Cylindrical coordinates, 24
time read on a rapidly moving, 211
Coaxial line, 127, 135, 356, 563, 566, 585, d’Alembertian, 233
586, 592 Debye equation, 120
Coercive force, 397 Decibel, 591
Comets, 546 Del operator V, 10, 15, 20
Complete set of functions, 159, 168 Depth of penetration 6, 477, 478, 499
Complex conjugate, 474 DeWitt, David, 176
Complex numbers, 529 Diamagnetic materials, 383, 396
Complex variables, 644 Dielectric constant, 108
Conductivity o, 351 Dielectric losses, 592
of an ionized gas, 483 Dielectrics, 91 ff., 459
Conductors, 52, 424, 439, 471 ff., 476, 480, class-A, 104, 108
499, 532, 540, 543 nonlinear, anisotropic, nonhomogeneous,
Conformal transformation, 646 /f. 121
Conjugate functions, 646 polar, 116, 120
Conservation of charge, 233, 422 Dielectric sphere with a point charge
Conservative fields, 16, 44 at its center, 111
Contact potential, 178 Dielectrophoresis, 135
Coordinates Diffusion, thermal, 177
Cartesian, 1 Diode, p-n junction, 176
correspondence between the four Diode, vacuum, 189
systems, 27 Dipole; electric.ol jf— Gi, Live teieie>
curvilinear, 23 oscillating, 428, 595, 627
cylindrical, 24 Dipole, magnetic, 319
spherical, 26 oscillating, 430, 623, 627, 629
Corson Ds Reo Dispersion, 427, 491
Cottrell, F. G., 188 Displacement D, electric, 105, 109
Coulomb, 250 Displacement current density dD/ar, 424,
Coulomb’s law, 40, 52, 54, 58, 96, 105, 247, 438, 439, 472
250, 294 Distortionless line, 561
Coupling, coefficient of, 350 Divergence
Creation of charge, continuous, 456 four-dimensional, 233
Critical angle, 520, 526 in Cartesian coordinates, 13, 16
Curl in curvilinear coordinates, 28
in Cartesian coordinates, 16, 19 in cylindrical coordinates, 30
in curvilinear coordinates, 30 in spherical coordinates, 30
in cylindrical coordinates, 31 of A, 271, 308, 432
in spherical coordinates, 31 of B, 273, 302, 392, 420, 440
of A, 264, 304, 337 of D, 106, 109, 436
of B, 274, 308, 437, 440 of E, 49, 105 f7., 271, 436, 440
of E, 273, 336 of H, 420
of H, 438 Divergence theorem, 13, 15
Curl-meter, 657 Domain, magnetic, 384
Current Donor atoms, 176
conduction, 292 Doppler effect, 212, 243, 244, 497
convection, 484 Duality, 444
displacement, 424, 438, 439, 472 Durand, E., 648
electrolytic, 292
polarization, 96, 423 Eigen values, 560
INDEX
701

Einstein, A., 201 of a charge moving at a constant velocity,

Electrets, 113, 133, 141, 184 Doilerife
Electric displacement D, 105, 109 of a half-wave antenna, 616, 636
Electric energy density. See Energy density, of an oscillating electric dipole, 428, 595,
electric 608, 627
Electric field of an oscillating electric quadrupole, 620
between two grounded parallel electrodes of an oscillating magnetic dipole, 430,
terminated on two opposite sides 623, 627, 629
by plates at potentials V; and V2, 161 of an oscillating magnetic quadrupole,
between two grounded semi-infinite par- 626
allel electrodes terminated by a plane of antenna arrays, 616
calculations, 53, 54, 66, 92 #7, 109 of a short element of wire carrying
electrode at a potential Vj, 158 an electric current, 280
disturbed by a conducting sphere, 168 of a 10 GeV electron, 260, 266, 267
disturbed by a dielectric sphere, 173 of a 10 MeV proton, 289
in a plane p-n junction in silicon at 300 Electromotance, induced, 333, 334, 335,
kelvins, 178, 179 B39 342 OomiT:
of a bar electret, 113, 404 Electron annihilation, 221
of a charged sphere near a grounded con- Electronic polarization, 122
ducting plane, 150 Electrons, drift velocity in a conductor, 279,
of a moving solenoid, 289 290
of an electric dipole, 61 ff. Electrophoresis, 135
of an individual dipole in an electric field, Electrostatic precipitation, 188
103 Emde, F., 291, 616
of a point charge at the center of a Energy density, 465, 471, 473, 577
dielectric sphere, 111 electric, 76, 124, 128, 134, 464, 471, 473,
of a point charge moving at a constant Hi
velocity, 256, 258 magnetic, 353, 354, 464, 471, 473, 577
of a point charge near a charged conduct- Energy, kinetic
ing sphere, 149 classical, 195
of a point charge near a grounded con- relativistic, 221, 224
ducting sphere, 146 Energy, magnetic, 351 ff.
of a point charge near an infinite, Energy product, 411
grounded, conducting plane, 144 Energy, relativistic &, 220
of a point charge near a semi-infinite Energy transmission, 577
dielectric, 153 Epstein, D. J., 586
of a uniform spherical charge distribution, Equipotential surfaces, 45
54 ff. Equivalent currents, 386/f-
outside an arbitrary charge distribution, Exponential notation, 665 ff:
66
Electric field intensity EF, 42, 337 Fabry equation, 326
average, inside a sphere containing an Farad, 76
arbitrary charge distribution, 70 Faraday, M., 374
induced, 337 Faraday
inside a dielectric, 97 ff. disk, 662
local, 102, 121 induction law, 332, 336
macroscopic, 97 Fermi acceleration, 328
Electric force, 78, 80, 89, 125, 128, 136, 137 Ferromagnetic materials, 383, 395
Electric polarization P, 91, 92, 329 Feynman, Richard P., 314, 465
Electric susceptibility x-, 104, 119, 122 Fields, 1
Electroforming, 588 conservative, 16, 44
Electrolytic plotting tank, 185 Floating wire method, 301
for magnetic fields, 420 Flowmeter, 374
Electromagnetic field Flux, 13
inside a coaxial line, 566 compression, 381
inside a hollow rectangular wave guide, electric. See Gauss’s law
568 ff. leakage, 406, 409, 412
702 INDEX

Flux (cont.) Homopolar generator, 662

linkage, 334 Hubble constant, 456
magnetic, 332. See also Divergence of B, Hull, G.F., 524
and Inductance Hysteresis, 397, 398
Focusing, alternate gradient, 593
Four-current density J, 229 Image forces, 185
Four-momentum p, 223 Image of a current, 636
Four-potential A, 269 Images, electric, 144 ff., 636
Four-vector r, 221 Impedance
Four-dimensional divergence, 233 characteristic, 568, 585, 586
Four-dimensional gradient, 232 wave, 568
Four-dimensional operator Quad (J, 232 Incidence, plane of, 507
Fourier series, 159 Inductance
Fresnel’s equations, 508, 527, 532, 540 mutual M, 343, 348
Friedman, F., 196 self L, 343, 345 ffi, 356, 377
Frisch, D., 196 Inertia effects on conduction electrons, 373
Functions Invariance, 5, 8, 9, 12, 13, 15, 19, 22, 28, 32
conjugate, 646 of a physical law, 201
harmonic, 163 of electric charge, 228
stream, 648ff. Ion drag accelerator, 375
Ionic polarization, 122
Gain of an antenna, 637 Ionized gases: conduction at low frequen-
Galilean transformation, 194 cies, 190
breakdown of, 195 Ionosphere, 492, 501, 556
Gamma_ radiation, production of high Ion rocket, 190, 379
energy, 243 Isotropic medium, 422, 467
Garden-hose effect, 502
Gases, ionized, 481 ff., 547 ff. Jahnke E., 291, 616
Gas lens, 593 Jaseva, T.S., 198
Gauss, 295 Jaseva-Javan-Murray-Townes experiment,
Gauss’s law, 47, 57, 61, 105 77, 271 198
Gibbons, James F., 176 Javan, A., 198
Goos, F., 523 jw operator, 665 ff.
Gradient Jory, H.R., 498
four-dimensional, 233 Joule losses, 352, 377, 474, 480, 578#7.
in Cartesian coordinates, 10
in curvilinear coordinates, 27 Kennedy, D.P., 177
in cylindrical coordinates, 28 King, John G., 42
in spherical coordinates, 28 King, R.W.P., 628
of V, 44, 266, 337, 426 KX surface, 607
Gravitational equivalent of magnetic Kober, E., 648
field, 251
Gray, Paul E., 176 Langevin equation, 117
Green’s theorem, 15 Laplace operator V?, 22
Guide for charged particles, 188 Laplace’s equation, 51, 56, 138 f#:, 156, 163,
168, 186, 187
Hall effect, 300 Laplacian
Hanchen, H., 523 of a scalar function, 22, 31, 32
Harmonic functions, 157 of a vector function, 22, 32
spherical, 163 of A, 308, 437
Harvey, A.F., 592 of V, 51, 56, 59, 105, 108, 163, 436
Headlight effect, 242 Larmor frequency, 287
Heat conduction, 138, 500 Laser, 198, 465, 483, 546
Henry, 345 Legendre’s equation, 163, 165
Hodoscope, 301 Legendre polynomials, 163, 165, 167, 168
Hole, 176 Length, proper, 205
Homogeneous medium, 422, 467 Lenz’s law, 335
INDEX 703

Levitation, electromagnetic, 378 Magnetic induction B, 250, 295, 324, 384

Light, velocity of, c, 195, 204, 233, 427, 439, Magnetic mirror, 328, 330
461 Magnetic pressure, 367, 380
Line integral, 16 Magnetic susceptibility, 395
Linear medium, 422, 467 Magnetic torque, 364
Lines of B, 296 Magnetization, 384
of refraction at a current sheet, 316 Magnetization curve, 397
Lines of force, electric, 46, 141 Magnetomotance, 394, 407
Longini, R.L., 176 Magnetosphere, 496
Lorentz condition, 271, 432, 456 Magnetron, 327
Lorentz Mass
contraction, 206 relativistic, 217
force, 250, 299 rest, 219
lemma, 446, 629 Maxwell’s equations, 105, 247, 273, 276, 290,
transformation, 203, 247 337, 439 ff.
Lyttleton, R.A., 188, 456 Michelson-Morley experiment, 198
Microstrip line, 587
Magnet, permanent, 393, 402, 404, 409 Millikan oil-drop experiment, 185, 188
Magnetic bottle, 330 MKSA system of units xvi, 41, 116
Magnetic circuits, 405 ff Modes of propagation, 572
Magnetic dipole, 319 Molar polarization ajz, 116
moment, 319, 321, 384 Molecular polarizability a, 104, 121
oscillating, 430, 623, 627, 629 Momentum, flux of, 546
Magnetic energy density. See Energy Momentum, relativistic, 223
density, magnetic Monopole, 65
Moon, 517
Magnetic field, 292 ff.
at the center of a rotating disk of charge, Mossbauer effect, 244
413 Multipoles, electric, 65
inside and outside a long cylindrical con- Murray, J., 198
ductor, 311
near a Straight wire carrying a steady elec- n-type material, 176
tric current, 277 Neumann equation, 344
of a bar magnet, 393, 402, 404 Nitrobenzene, 122
of a current flowing in a long straight
wire, 296
O’Brien, R.R., 177
of a long solenoid, 289, 315, 322
Observer, 206
of a point charge moving at a constant
Octupole, electric, 66
velocity, 259
Oersted, 412
of a short solenoid, 317, 326
Ohm’s law, 351
of a toroidal coil, 313
Orbitron vacuum pump, 187
on the axis of a circular loop, 299
Orbit stability, 187
resulting from a relativistic transforma-
Orientational polarization, 122
tion of an electric field, 250ff.
Magnetic field calculations, 402
Magnetic field intensity H , 393 Panofsky, W. K. H., 201
Magnetic flux &, 296 Parallel-wire line, 563, 654
linkage, 333 Paramagnetic materials, 383, 395
Magnetic force Parsec, 502
between two circuits, 293, 358 ff. Particle, charged, in E, B field, 288
between two coaxial solenoids, 361, 363 Peaking strip, 418
between two current elements, 282 Pea sorter, 90
between two long parallel wires Permeability
on a moving charge, 260, 544, 546 differentia], 397
on a point charge moving in a magnetic of free space po, 250, 284, 293, 323, 329,
field, 299 461
within an isolated circuit, 365 relative p,, 395
Magnetic force law, 293 ff., 324 Permeance, 407
704 INDEX

Permittivity Quasi-stationary field, 439

complex, 499
of free space eo, 41, 461 Radian length %, 672
relative «,, 108, 119, 395, 439, 470, 593 Radiation resistance, 605, 616
Phase, 665, 671 Radiofrequency heating, 500
Phillips, M., 201 Reciprocity theorem, 629
Photon, 224, 546 Red shift, 240
Photon rocket, 243 gravitational, 244
Pinch effect, 290, 366 Reflection, 504 ff.
Plasma coefficient of, 517
angular frequency w,, 486 laws of, 505, 527, 532, 540
gun, 379 total, 520 ff, 554
Plasmas, 481 ff., 496 Reflections, multiple, 539
Poisson’s equation Refraction, 504 ff.
for B, 413 index of, 470, 473, 499, 541
for E, 180 of lines of B at a current sheet, 316
for V, 51, 59, 106, 108, 138 ff, 142, 436 of lines of B at an interface, 400
Polar dielectrics, 116, 120 of lines of E at an interface, 141
Polarizability, molecular a, 104, 121 Relativity, 193 ff.
Polarization fundamental postulate of, 201, 248
circular and elliptical, 462, 496 Relaxation
current density, 96, 423 method, 86, 177, 420
electric, 91, 92, 329 time, 424
electronic, 122 Reluctance, 407
ionic, 122 Remanence, 397
molar amu, 116 Re operator, 474, 666
orientational, 122 Resonant circuits, 501
plane of, 462 Retarded potentials, 427 ff.
Polarizing angle, 516 Retentivity, 397
Pole, magnetic, 404 Right-hand screw rule, 4
Potential, complex, 644 Roberts, C. S., 498
electric V, 43, 49, 50, 56, 59, 105, 108, 163 Rocket motor, 190, 327
scalar V, 226, 267, 337, 424, 562, 597 Rutherford experiment on the size of the
vector A, 264, 266, 267, 303, 305, 306, 308, atomic nucleus, 85
337, 424, 562, 598
Potential energy Satellites, attitude control of, 378
of a dipole in an electric field, 117 Saturation induction, 398
of an electric charge distribution, 72, 123 Scalars, 1
Potentials Scattering, resonant, 245
electromagnetic, 424 Sea water, 478, 500, 534
retarded, 427 ff. Semiconductors, 176, 189
Poynting vector 8, 464, 471, 473, 492, 546, Separation of variables, method of, 157,
564, 568, 573, 577, 579, 585, 587, 603, 676, 681
614, 625 Shape, apparent, of a rapidly moving object,
Pressure, radiation, 543ff: 207
Propagation constant y, 673 Shielded pair line, 563
p-type material, 176 Signal velocity, 214, 241, 490, 576, 578
Pulsars, 502 SI system of units, xvi
Pumps, electromagnetic, 455 Simultaneity, 213
Purcell, Edward M., 96 Skin
depth 5, 477, 478, 499, 533
Q of a medium, 472, 475, 499 effect, 454
Quad operator 1, 232 Smith, A. C., 176
Quadrupole, electric, 64 Smith, J., 196
oscillating, 620 Snell’s law, 505, 527, 532, 540, 547
Quadrupole, magnetic, oscillating, 626 Solar wind, 502
Quasars, 240, 639 Southwell, R. V., 86
INDEX 705

Spectrum of electromagnetic waves, 459 of the field of a long solenoid, 289

Spherical coordinates, 26 of the partial derivatives 0/0x,0/0y 0/dz,
Standing wave ratio, 588 0/d0t, 232
Stanford linear accelerator, 242 of a time interval, 208
Static field of an oscillating electric dipole, of a velocity, 215, 216, 241
608 of electric and magnetic fields, 261, 287
Stationary medium, 422, 468 of the electric field inside a parallel-plate
Stokes’s capacitor, 263, 289
law, 135, 189 of the electromagnetic potentials, 269
theorem, 21 Transmission
Stratton, J. A., 434 coefficient of, 517
Stream function, 648 ff. line, superconducting, 379
String, stretched, waves on, 674 ff. Travel, interstellar, 243
Strong, J., 524 Trivelpiece, A. W., 498
Submarines, communicating with, 535 Trouton and Noble experiment, 196, 248
Sun, 184, 326, 496, 546, 555, 556 Tsukada, Masanobu, 457
Superconductors, 455 Twin paradox, 240
Superposition, principle of, 43
Surface Uniqueness theorem, 142, 185
impedance, 555 Units, xvi, 41, 116
resistivity, 555, 591 conversion table, 643
Systéme International d’Unités (SI), xvi Universe, expansion of, 188

Van Bladel, J., 35

Taylor, Edwin F., 193, 239 Van de Graaff accelerator, 184
TEM waves, 561, 564, 565, 566 Vectors, 1
Tenzer, R. K., 409 addition, 2
Tesla, 250, 295 components, 2
Test charge, 42 cross product, 3
TE waves, 560, 565, 569 dot product, 2
Thomson atom, 87 magnitude, 2
Time scalar product, 2
dilation, 195, 210 subtraction, 2
proper, 209, 210 time derivative, 8
T waves, 560, 565, 569 vector product, 3
Tolman, Richard C., 373 Vector potential A. See Potential, complex,
Torque vector A
electric, on a dipole in a uniform electric Vectors, unit, 2, 24, 25, 26
field, 125 Velocities, addition of, 195
magnetic, 365 Velocity, group, 491, 501
Townes, C. H., 198 limit, 195
Transformation maximum signal, 214, 241
conformal, 646 ff. of light c, 195, 204, 233, 427, 439, 461,
Lorentz, 204 491
of a force, 225 phase, 470, 471, 475, 490, 491, 567, 571,
of a four-current density, 231 576, 671, 673
of a four-momentum, 223 signal, 214, 241, 490, 576, 578
of a four-potential, 270 Vibration, plane of, 462
of a length, 205 Volt, 45
of a mass, 220 Voltage standing wave ratio (VSWR), 588
of a momentum, 223 von Hippel, A., 523
of an acceleration, 217
of an electric charge density, 229 Water, 122, 134
of an electric current, 229, 245 Wave equation, 675, 679
of an element of volume, 226 for A, 437
of a relative permittivity, 288 for B, 448
of a relativistic energy, 223 for E, 448
of Maxwell’s equations, 457 for V, 436
706 INDEX

Wave equation, nonhomogeneous, 683 Waves, 670 ff.

for A, 437 attenuated, 468, 679
for B, 448 in three dimensions, 682
for E, 448 Waves, electromagnetic
for V, 436 in conducting media, 471
Wave front, 671, 672 ion ionized gases, 481 ff.
Wave guides, 557ff. in free space, 460
beam, 593 in homogeneous, isotropic, linear,
coaxial line, 127, 135, 356, 563, 566, 585, and stationary media, 467
~ 586, 592 in good conductors, 475
hollow rectangular, 568 ff: in nonconductors, 470
microstrip line, 587 spectrum of, 459
parallel-wire lines, 563, 654 standing, 535
Wave impedance, 561 TE, 560, 565, 569
Wave length, 671 TEM, 561, 564, 565, 566
Wave number, 469, 470, 472, 475, 490, 499, TM 560, 565, 569
507, 509, 528, 541, 558, 671 Weber, 250, 295
Wave propagation in a straight line, 557 ff Wheeler, John A., 193, 239
 Ne bem CV Ones
; F = 4c = ae QiQz ec
\

D]

—_ rs
o ya ¢ .
eae

Y°Cc $9

r. 1 SiGe b
¢
i ile
Q
a.
Vv 7
+ |

B>VxA
—_
(n

O OPas= Uo.t

. F: YXA
 ,

os Py * Sin 4
vA
R=

.= sin? act ws? 7

Z°

x
Bienens R dk dodz

Z
dQ = dkR + RAPF + 42

dS * Rdo dz
 es Spheri cot

X= 15in 9 coo?
Mean
a rans sin?
ali : R°sin@ skaeds
dle dr Rrrde Gr rsinode $i
ASunfiee 2R™ sin@ doa’
OK
(4 > sm @ cosh sine Sgr ws

S,* ws cosh, cos sing, ~Sin © (c

% = - SinGir (0S b 4
‘
CONVERSION TABLE

Examples: One meter equals 100 centimeters. One volt equals 108 electromagnetic
units of potential.

SI CGS SYSTEMS

esu emu

Length meter 10” centimeters 10? centimeters

Mass kilogram 10° grams 10° grams
Time second 1 second 1 second
Force newton 10° dynes 10° dynes
Energy joule 10’ ergs 10’ ergs
Power watt 107 ergs/second 107 ergs/second
Charge coulomb se ele oO
Electric potential volt 1/300 108
Electric field intensity volt /meter iGae10:) 10°
Electric displacement coulomb/meter? 127 X 10° Ale 9 Ir"
Displacement flux coulomb 127 x 10° Ara ealOme
Electric polarization coulomb/meter? Sex LO? Ore
Electric current ampere 83 9 Ue 10=
Conductivity mho/meter 9 & 10° 1071!

Resistance ohm 1/(9 & 101) 10°

Capacitance farad 9 x 101 Ome
Magnetic flux weber 1/300 108 maxwells
Magnetic induction tesla 1/(3 X 10°) 104 gausses
Magnetic field intensity ampere/meter eel! 4r X 107* oersted
Magnetomotance ampere 127 x 10° (47/10) gilberts
Magnetic polarization ampere/meter W/(3re< 10%) 10s
Inductance henry 1/9 < 10") 10°
Reluctance ampere/weber 367 X 10}! Ane x10

Norte: We have set c = 3 & 108 meters/second.