Web: http: // www. pearl - hifi . com 86008, 2106 33 Ave.

SW, Calgary, AB; CAN T2T 1Z6

E-mail: custserv @ pearl - hifi . com Ph: + .1.403.244.4434 Fx: + .1.403.245.4456
Inc.
Perkins Electro-Acoustic Research Lab, Inc. ❦ Engineering and Intuition Serving the Soul of Music

Please note that the links in the PEARL logotype above are “live”
and can be used to direct your web browser to our site or to
open an e-mail message window addressed to ourselves.

To view our item listings on eBay, click here.

To see the feedback we have left for our customers, click here.

This document has been prepared as a public service . Any and all trademarks
and logotypes used herein are the property of their owners.

It is our intent to provide this document in accordance with the stipulations with
respect to “fair use” as delineated in Copyrights - Chapter 1: Subject Matter and
Scope of Copyright; Sec. 107. Limitations on exclusive rights: Fair Use.

Public access to copy of this document is provided on the website of Cornell Law School
at http://www4.law.cornell.edu/uscode/17/107.html and is here reproduced below:

Sec. 107. - Limitations on exclusive rights: Fair Use

Notwithstanding the provisions of sections 106 and 106A, the fair use of a copyrighted work, includ-
ing such use by reproduction in copies or phono records or by any other means specified by that section,
for purposes such as criticism, comment, news reporting, teaching (including multiple copies for class-
room use), scholarship, or research, is not an infringement of copyright. In determining whether the use
made of a work in any particular case is a fair use the factors to be considered shall include:

1 - the purpose and character of the use, including whether such use is of a
commercial nature or is for nonprofit educational purposes;

2 - the nature of the copyrighted work;

3 - the amount and substantiality of the portion used in relation to the copy-
righted work as a whole; and

4 - the effect of the use upon the potential market for or value of the copy-
righted work.

The fact that a work is unpublished shall not itself bar a finding of fair use if such finding is made
upon consideration of all the above factors

♦ PDF Cover Page ♦

♦ Verso Filler Page ♦
MAGNETIC FIELDS
A Comprehensive Theoretical Treatise
for Practical Use

Heinz E. Knoepfel

A Wiley-Interscience Publication

JOHN WILEY & SONS, INC.

New York. Chichester. Weinheim. Brisbane. Singapore. Toronto
♦ Verso Filler Page ♦
This book is printed on acid-free paper.

Copyright 0 2000 by John Wiley & Sons, Inc. All rights reserved

Published simultaneously in Canada.

No part of this publication may be reproduced. stored in a retrieval system or transmitted in any
form or by any means, electronic. mechanical, photocopying, recording, scanning or otherwise,
except as pennined under Section 107 or 108 of the 1976 United State Copyright Act, without
either the prior written pt:rmission of the Publisller. or autllorization through payment of the
appropriate pt:r-copy fee to the Copyright Clearance Center, 222 Rosewood Drive. Danver5, MA
01923. (978) 7S0-8400, fax (978) 7S0-4744. Requests to the' Publisher for permission should be
addressed to the Permissions Department, John WHey & Sons, Inc., 605 Third Avenue. New York,
NY 10158·0012. (212) 850·6011. fax (212) 850·6008. E·M.;I, PERMREQ@ W1LEY.COM

For ordering and customer service, call l-800-CALL WILEY.

Library of Congress Cataloging-in-Puhlication Dala:

Knocpfcl. Heinz, 193 1-

Magnetic fields: A comprehensive theoretical treatise for
practical use I Heinz E. Knoepfel.
p. cm.
Includes bibliographical references and index.
ISBN 0-47\·32205·9 (cloth: alk. paper)
I. Magnetic fields. I. Title.

QC754.2MJ K6J 1999 99·058796

537-·dc21

Printed in the United States of America

10 9 8 7 6 5 4 3 2
To the Memory of my Parents,
Eisa and Arthur Knoepfel-Bossart
 CONTENTS

1 FOUNDATION OF MAGNETIC FIELD THEORY 11

1.1 Maxwellts Field Equations /2

ntroduclion /2
Dlferenlal equalons /4
Material-related electromagnetic quanmies /5
Magnetic quantities
Magnetic charges and nOflsolenoidal fields
E.xample: Magnetic rod
E.lectric quantities

Ohm's law /16

1.2 Electromagnetic Field Approximations /19

Magnetostalcs and electrostatics /20
Ouasistalionary magnelc fields /21
Quasistalonary electric fields /24

1.3 Electrodynamics of Moving Media /25

Coordl1ate lameS of reference /25
MOv11g coordl1ate system /26
Maxwe//-Lorf'ntz transformation
Maxwel/-Ca/Uean trtmsjormarion
MOv11g media /31

1.4 Alternative Electromagnetic Equations /32

Integral laws /32
Ampere's law
Faraday's law
Miscellaneous laws

VII
viii CONTENTS

mfferentiai laws 137

Gurrenl densroes 13B
Magnetic crclils 140

1.5 Boundary Conditions 143

Field condilOhS 144
General conditions
Nonsupercondflding media
Example: Ca/culation of the conditions
Current densrty conditions 149
Moving boundanes 151

2 MAGNETIC POTENTlALS 155

2.1 Magnetic Scalar Potential 155

POlenlial n emply space /56
Definitions and solutions
Boundary conditions

Al<symmeUiC SYSlemS /60

General eqllations
Expansion near the axis
Examplr: Diamagnetic sphere

Two-dmensiona plane SYSlemS /64

Gr"rral I!qllations
Example: TransversI! field 0" diamagnetic rod
The conformal transformation method
Example: Parallel lines
POlential in and Ou1Slde magnetized malerial /70
Example: Permanent splJ"ica/ mag"et
Example: Magnetized sphrrr
Method of images

2.2 Magnetic Vector Potential 176

Base eqJations and soUtions /76
Equations
Solutions
Boundary conditions
Retarded potentia/s
AxiSymmeUiC syslems IB4
Plane IWo·dmensional syslems IB6
Bbl-Sava1law IB6
Fj/amentary conductor
Mll/lipo/e expansion
Magnetic dipole
 CONTENTS ix

2.3 Fields in Cylindrical and Straight Conductors 191

Cicuar loop 191
MuI1loop system 194
Coit pair
Field shaping

Thn solenoid 197

Thick solenod 11 00
Bitter solenod 1104
Torodal mag1ets 1105
Straight conductors I, 07
Example: Rectangular bar conductor
Example: Strip condllctor
Example: Two-dimensional nW/lipol! configurations
Saddle-shaped cols 1114
Short filamentary coil
Long coils

2.4 Inductance of Conductors 1117

Serr-nductance 1117
filamentary approximation
faraday law
Exampif': Wire transmission lille

Mutual nductance 112'

Filamentary approximation
General upressions

2.5 Structure of Magnetic Fields I, 25

Magnebc field lines 1126
Magnetic surfaces 1128
Toroidal slIrfaces

3 PERIODIC FIELDS AND WAVE PHENOMENA 1133

3.1 Complex Functions /133

StlusoKJal functions 1134
Complex variables and functions 1135
Example: Conformal transformations
Comptex vectors I, 38

3.2 Sinusoidal and Rotating Electromagnetic Fields /140

Electromagnetic equations 1140
GeneralZed wave equabons 1'42
Magnetic potential I' 42
Rotatng magnetic fields I' 44
x CONTENTS

Two-phased coil arrangement

Three-phased coil ammgement

3.3 Wave Phenomena 1146

Wave equalions 1147
Example: Radiation from Q Her/zion dipole
Plane wa-leS n a urdform medium 1149
Plane waves n a conduc1i1g mediJn 11 so

4 MAGNETIC FIELD DIFFUSION AND EDDY CURRENTS /153

4.1 Magnetic Diffusion Theory 11 53

General equalions for a soOCJ conductor /154
OuasistatiOnary dlfusion equatiOns 1156
SolutiOns to the dlfusion problem 11 58
Solutions and analogies
Response to a step-junction field
Gene'IJ/ization
DiffUSion time

4.2 Field Diffusion into Plane Conductors 1165

Half-space conductor 1165
Step-junction field
General solution
Characteristic diffusion parameters
Spec� boundary flelds on haHpace 11 7 1
Transient polynomial fields
Transient sinllsoida/ field
Steady sinusoidal field
Exponential field
Slab conductor 1179
Step-function field
Steady sinusoidal field

4.3 Fletd Diffusion In Cylindrical Geometry 1183

EquatiOns n smpie geometry 1184
One-dimensional equations
Axisymmttric equations
Two-dimensional equations
Dlfusion of <OOaIlields 1187
Step-function field on R rod
Steady sinusoidal field on R rod
DiffUSion through a hollow condllctor
 CONTENTS XI

Diffusion of transverse fields /194

Transverse step-function field on a rod
Steady and rotating transverse jiffds
Transacrs! diffusion Ihrollgh a hollow condfIClor

4.4 Eddy Currents /201

The eddy cllTent �oach /201
Two-dmensional currents n sheets /203
The thn-sheet approximation /207
Field equations
Vector potential equations
Example: Thil1 disc
Example: Moving thin disc
Erample: Moving sheet

Eddy ClXfents n cyindrk:al shels /214

Limiting CIISt'S
The thin-walled shell
The magnetically thin shell
EXiHnple: Transverse field on thin shelf with gaps

MagnetIC sIleldng /220

Example: Shielding Dj Q steady sinu5{JideJl field
EXilmplc: Shielding by Q thick hollow conductor
EXQmple: Shielding by Q thin hollow conductor

4.5 Diffusion in Electric Circuits 1225

Lumped circuits /225
Skin-effect impedance
Example: Step-function WHent on pl(HlC conductor
Ernmple: Step-junction current on a rod

GeneraHzation of the resuns and equiva�nt crcu!s /230

Alternating currents /232

5 ELECTROMAGNETIC AND THERMAL ENERGIES 1235

5.1 Electromagnetic Energy Equations 1235

Electromagletic energy /236

Field equations
Potential equations
Sinllsoidal fields
Inductive energy
Example: Coaxial transmission line

Thermal energy /244

f.xrlmple: Thermal diffUSion ;n slab
xii CONTENTS

S.2 Magnetic Heating in Constant Conductivity Conductors /248

Diffusion healilg /249

fault's law
Energy equations
Healilg illhe condJclilg half-space /252
Polynomial field
Step·function field
SinusoidQ/ field
Energy ski" depth
Healilg of thiCk sheets and cyndncal conductors 1260
Slab
Rod
Hollow cylindrical conductor

5.3 Heating in Thin Conductors 1264

p� anc cy1i1d!ical geomelries /264
Thin slab
Thin rod
Thil! disc

Electric conductMty law /268

Temperafllu-dependeP11 condllctivify
Cllrrent integral
Diffusion anc healilg il hoIow conductors /270
Example: Step-flmction field diffusion
Example: Transient sinusoidal field diffllsion
Example: Steady sinllsoidal heating

5.4 Nonlinear Magnetic Diffusion 1277

Nonlinear diffusion equaTIons 1278
One-dmenslonal diffuskln /279
An QPproximated 50h,tion
Energy skin depth
Dffusion and healilg il magnetic materials /285
StetJdy sinusoidal magnetic field heating
Ferromagnetic induction heating
Ferromllgnefic hystnesis heating

6 MAGNETIC FORCES AND THEIR EFFECTS /295

6.1 Electromagnetic Forces 1295

ElementaJ forces /296
Lorenlz force
 CONTENTS XI"

Biol-Savart force
Electromagnetic force density
Forces on li1ite tJo<jes /300
Magnetic jOTU density
Example: Arisymml!tric force density
Marwelrs magnetic stress tensor
Magnetic mu/tipoJe expansion
More on tiectromagnetic forces and torques

6.2 Magnetic Forces on Rigid Conductors /311

tlducwe forces /311
Virtual work formalism
Loops and solenoids
MagnetiC suspension and braki1g /315
Magnetic dipole suspension
Magnetic braking
Example: Moving disc
bllmp/e: Moving sheet
Example: Rotating sheet
Homopolar generator
MagnellC levitation /326
Fcrromagnetic track
Conducting track
Magnetomechanica macht1es /333
Rotating machines
Synchronous machines
Induction machines
More on magnetomeciltmica/ machines
Forces wilhn a conductor /339

6.3 Magnetic Propulsion Effects in Conducting Sheets 1342

Plane Sheet propUslO/1 /343
Equation of motion
Thermal velocity limitation for thin condllctors
Shock waves
Ablation effect
Velocity limitation for thick conductors
tnptosion of a C}Yldrical sheet /352
GenerQI equations
Constant current driver

6.4 Acceleration and Trajectories of Particles /356

Equation of motion /357

PeDet acceleration methods /359

XIv CONTENTS

Magnetic acceleration /361

Magnetic dipole Qcceleration
Diamagnetic sphere
Example: Toroidol coil
Magnelic raif gun
Particle trajectories n electrornagletic lields /370
Cartesian equations
Motion in static. uniform fields
Motion in variable fields
Motion in oXisymmetric fields
Example: Trajectories in toroidol fields
Electromagnetic optics

7 MAGNETOMECHANICAL STRESSES /387

7.1 Stress-Strain Relations and their Limits 1388

The stress tensor /388

Elastic stress-stran relaliOns /392

The displacement vector /395

Effects of thermal expansion 1397

ThermoelastiCity n dynamic system 1398

Falue a�eria 1399

7.2 Slresses In Solenoldal Magnels 1404

Thick cy1inder and disk 1404

Solenoidal cois 1407

Yielding i1 cy1ndrica contaners /412

Thin-waled cytindrical magnet 1414

Dynamc contailment 1415

7.3 Slresses In Toroldol Magnel. 1417

Contnr sheI of genera shape 1417

Toroidal magnet 1420

Toroidal maglet with constant stress /423

7.4 Force·Free Fields and Current Distributions /427

General formula�on 1428

Force-fi"ee configlKaliOns 1428

Plane geometry
 CONTENTS xv

Cylindrical geometry
Spherical geometry
Variable-pitch solution

8 MAGNETOHYDRODYNAMICS AND PROPERTIES OF

MATIER /435

8.1 Magnetohydrodynamic Equations /436

General formulation /436
Elllerian frame
Lagrangian frame
Magnetic diffusion /442
Field diffusion
Vector potential diffllsjon
Limiting cases
Cylindrical one·dimensional geometry

8.2 Magnetic Flux Compression /450

CLJrent generators /451
Flux compression c;Twit
Flux compression by incompressible plane conductors
mplosion of an idealllJx compressng shel /45B
Field diffusion
Ideal fluX compression
Ulrahigh fieid generators /463
Incompressible shell WitJl FMile conductivity
Compressible shell with infinite conductivity
Cylindrical flux compression

8.3 Some Properties of Matter 1471

Ek'>ctric conductMty /472
Conductivity in the solid state
Conductivity in liquid, vapor, and plasma states
5 uperconduct iuitlj

MagnetiC maner /4B4

Diamagnetism
Paramagnetism
ferromagnetism
Permanent magnets
Quantum parameters
Critical magnetic fields and monopoles
Nuclear magnetic resonance
H�h-energy densities /503
XVI CONTENTS

Equation of state
Specific heat and phase transitions
Shock waves

9 NUMERICAL AND ANALOG SOLUTION METHODS /51 3

9.1 Numerical Computation Methods 1514

Fintte-dife
f rence method /514
Poisson's difference equation
Difference solution method
Fintte-elemenl method /51 8
Shape fllnctions
example; Leakage field in an electromagnet
Fntte-elemenl ne1Wolk melhods /523
Network definition
Network soilltion
Magnetic reluctance network

9.2 Approximation by Filamentary or Simple Elemental Conductors /530

Ctcu� bops and so�ods /530

Approximatlon by Namenlary straight sectlons /532

tlductance and magnetic energy /533

Force and neld tines /534

9.3 Analog Methods /535

Electromagnetlc model scaiing /535
General formulation
Qllasistationary approximation
Magnetomechank::ai model scaing /538

APPENDICES /541

A.I Electromagnetic Units and Equations 1542

A.2 Mathematical Functions /545

The error tunclon /545

Bessel tunctlons of integer order /547

Functlons b1 sertes formuJalon /553

A.3 Coordinate Systems, Vector and Tensor Relations /555

Vector formulae /555
 CONTENTS xvii

Definitions
Vector operators
Differential relations
Integral relations
Vectors n Canesian, C)Ii1c1:icaJ, and spherical coordnates /558
Various coordinates and transformations
Differential expressions
Lap/acian of a vector

Vectors n or1hogonal cUlYlnear coordnates /562

Definitions
Differential expressions
Direction of the curl vector
Matrices and tensors /564
Definilions
Transformations between coordinate systems

A.4 Some Solutions of Second-Order Differential Equations 1567

Per1i'1ent differen\ial equaUons /568

Lap/aCe equatiOn /569

Separation in Cartesian coordinates
Separation in cylindrical coordinates
Example: Solution in It periodic geometry
Separation in spherical coordinates
Poisson equa!on /576
Simple and double series solution
Example: Infinitely permeable. rectangular boundary
Example: Axisymmetric, supercondllCting bOl4l1dary
Helmhoitz equation /581

Diffuson equ aUon /583

Separation in Cartesian coordinates
Separation in cylindrical coordinates
Separation in spherical coordinates

BIBLIOGRAPHY 1587

INDEX 1605
♦ Verso Filler Page ♦
 PREFACE

he aim of this book is to present a comprehensive treaUTIent of the problems

T connected with the generation and application of magnetic fields. It is basically a

treatise on the magnetic aspects of classical electromagnetic theory and is intended
for physicists and engineers working-however occasionally-with magnetic fields. It
should also be of interest to university students as a complementary text in courses on
electromagnetics. As such it contains both elementary concepts and advanced topics. The
attentive reader will probably realize that the book is nOl just a simple review of the vast
amount of literature on classical electrodynamics. On the contrary. the theoretical
presentations and the discussions of examples often follow unconventional paths, chosen
so alii to be clear and significant.

The book lies about midway between a theoretica1 monograph and a practical handbook. It
covers a broad spectrum of practica1 problems, ranging from the generation of magnetic
fields by a variety of coils, magnetomechanical and thenna! effects in system components
(magnets, coils, levitated vehicles, induction heaters, etc.), the magnetohydrodynamics of
ulltahigh field generalion, up 10 magnelic acceleration mechanisms and the trajeclories of
charged palticles and projectiles. Each topic is clearly related to the framework of
fundamenta1 electromagnetic theory, and the theoretica1 discussions are detailed enough to
allow the interested reader to follow the derivation of the results and to extend the solution
methods and applications to similar or more complicated problems. The extensive
Bibliography provides additional infonnation and a foothold from which to start a more
extensive literature search.

I have used a deductive approach. starting with Maxwell's four field equations, together
with two medium-dependent equations (chapter I), and then adding the Lorentz force
equation (chapter 6) or, alternatively, the correlated energy equations (chapter 5). From this
concise yet extremely powerful set of equations the theory of magnetic phenomena and its
applications unfolds in a natural and orderly way. In some aspects the presentation is
xx PREFACE

similar to my previous 1970 book. Pulsed High Magnetic Fields. which remains a useful
complementary source for solving some specific problems.

It is a'5umed that the reader has the mathematical background required by most textbooks
on electrodynamics; in particular, ordinary differential calculus and equations, vector
algebra, and differential relations. In any case, helpful reminders are given in the
Appendices, which contribute to making [he book largely self-sufficient. The International
System of Units (SI) is used in text and fannulae; but in deference to still widespread
laboratory practice, the ghost of the practical egs units (cm. g. s, dyn, erg, oersted. gauss,
together with ampere, volt, henry. coulomb. etc.) lingers on through some duplicated basic
equations, clearly marked by an asterisk. In addition, two comprehensive tables provide the
basic equations written in different systems (SI, Gaussian. emu, esu, practical cgs).

This book has evolved gradually through the years and is largely based on notes taken
during my long involvement in activities specifically related to magnetic fields. In the
foHowing are listed some of the projects I have worked on and just a few of the teachers,
colleagues and scholars to whom I express my gratitude for their direct or indirect
contributions and scientific enlightenment: In the 1950s, a ganuna-ray spectrometer based
on a highly unifonn NMR-controlled magnet (with Peter Stall and WilIy W51fli), this
thesis work was carried out at the Swiss Federal Institute of Technology (ETH) in Zurich
under Paul Scherrer, whilst I was also writing up and editing (with Fritz Herlach)
Wolfgang Pauli's lecture on Wave Mechanics for publication; in the 1960s magnetic flux
compression experiments and theory at Frascati with Jirka Linhart, Fritz Herlach and
Riccardo Luppi; in the 1970s, electron runaway studies in the Onnak magnetic tokamak at
Oak Ridge National Laboratory with John Clarke, Don Spong and Stewart Zweben; still in
the 1970s, design of an eJ:perimenral magnetic tokamak machine for relativistic electron­
beam studies at the Massachusetts Institute of Technology (Mln with Bruno Coppi, and
later. in the 1990s involvement in his superhigh-magnetic-field Ignitor project; back in the
1980s, I was engaged. as chainnan of the European Advisory Group on Fusion
Technology, in establishing the technology base for the European Fusion Program-in
particular, wirh regard to superconducting magnet technology; since the end of this period 1
have been following these problems also as director of the School of Fusion Reactor
Technology founded by Bruno Brunelli at the "Enore Majomna" Center for Scientific
Culture in Erice. Sicily.

I am grateful ro Robeno Andreani. the present and long-time director of the Eumtom-ENEA
Fusion Program, for the hospitality extended to me at the Frascati Research Center of the
Italian Agency for New Technologies, Energy and tht" Environment (ENEA) well beyond
my employment there by the European Commission. Last but not least, it is a pleasure to
acknowledge the substantial technical help received from many persons at Frascati. above
a1l from Lucilla Crescentini for the dedicated and professional management of the many
draflS of the manuscript, and from Peter Riske for the artwork. I would also like to thank
Carolyn Kent for the English editing, Nadia Gariazzo for typing the final camera-ready
copy, and Maria Polidoro for the secretarial help over all these years.
 PREFACE XXI

Finl.llly. I apologize for any errors in the text. equations, or figures thl.lt have been
overlooked despite careful proofreadi ng, and would greatly appreciate their being called to
my allenrion.

HEINZ E. KNOEPFEL

March 20lKI

Auocialilm EURA TOM·ENEA

£N£A Rt'xearrh Ce!ltu,
Fr'mHlli, fluly
♦ Verso Filler Page ♦
MAGNETIC FIELDS
♦ Verso Filler Page ♦
 MAGNETIC FIELDS; A Comprehensivc Theorctical Treatise for Practical Use
Ileinz E. Knoepfel
Copyright Cl 2000 by John Wiley & Sons. Inc.

Chapter 1

FOUNDATION OF MAGNETIC
FIELD THEORY

War e.s ein GOll der diese Gleichungen schrieb? ("Was it a god who wrote these
equations?") Thus wrote L.E. Boltzmann. one of the great scientists of the 19th century. at
the beginning of the introduction to his Lectures Oil Maxwell's Theory of Electricity and
Light (Munich. 1893). This motto well reflects the powerful beauty and conciseness of the
so-called Maxwell equations. particularly when presented (as below) in vectorial form.
These equations. which had been published in final fonn about ten years earlier, represent
the concluding highlight of centuries of discoveries and studies in electromagnetism and set
the comprehensive foundation of classical electromagnetic theory.

In contrast to the historical approach used in many textbooks (where the fundamentaJ
effect'; are gradually developed into the final set of the electromagnetic equations), in this
boo k. MaxweU's equations are used as the starting point for presenting and discussing the
mathematical and physical aspects of electromagnetism. with particular reference to
magnetic phenomena. The presentation of various (simplified) fonns of MaxweU's
equations and some related mathematical constraints is the main aim of this chapter.

Note. Equations are referred to by their designation, for example, (1.5-23) mean s the
equalion labeled (1.5-23) in seclion 1.5 The Inlemalional SySlem of UrnlS (SI) applies
throughout the book; however. an asterisk added to the equation designation-for example,
(1.1-17)*-indicales that equalion (1.1-17) is written in practical cgs units. which are

defined in table A.I-U of appendix A.1. Superscripts refer la the Bibliography, which is
subdivided per chapter and given at the end of the book.
2 CHAPTER 1 FOUNOATION OF MAGNETIC FIELD THEORY

1.1 MAXWELL'S FIELD EQUATIONS

Introduction

The history of the study of magnetic and electric effects is as old as that of physics.
which originated in the Ionian Greek culture· as an offspring of philosophy about 600 BC.
The philosopher Thales of Miletus, who is credited as the founder of science and thus
of physics ("the study of nature"), knew about the peculiar properties of lodestone as
attracting iron or assuming a north-south orientation. A large deposit of lodestone is known
to have existed near the ancient town of Magnes (today's Manissa. near Izmir, Turkey).
from which. in fact, the word "magnetism" is derived. These simple magnetic, as well as
some electrostatic, phenomena remained a curiosity for centuries.
The year 1600 AD saw the publication of De magnere, the first elementary treatise (in
Latin) on magnetism, by W. Gilbert, who was an influential medical doctor at the English
Court. The birth in the 17th century of the inductivHeductive science of J. Kepler, G .
Galilei, and I . Newton led in the 18th century to numerous and ordered observations of
magnetic and electric phenomena by D. Bemoulli. H. Cavendish, Ch. A. Coulomb, B.
Franklin, A. Galvani, and A. Volta. just t o name a few of the many scientists involved. A
fundamental contribution to the progress of science, and to the study of electromagnetic
phenomena in particular. was the refinement of mathematical analysis through the
differential and variational calculus introduced by I. Newton and G.W. Leibniz in
1670-75, and extended by L. Euler and J.L. Lagrange in 1744-55.
In the pioneering ftrst half of the 19th century, it became possible to study magnetic
and electric phenomena more systematically. Electricity and magnetism, which previously
were entirely separate subjects (the former dealt with such things as eat's fur, glass rods,
batteries, frog's legs, lightning; the latter with bar magnets. compass needles, the Earth's
poles) rapidly merged into electromagnetism. The beginning of this period coincides with
H.Ch. Oersted's discovery in 1819 that a compass needle is deflected if placed near a
current-carrying conductor. Mathematician A. M. Ampere's interest in physics was
stimulated by this discovery. and within a few months (1820) he extended both
experimentally and theoretically the understanding of magnetic effects related to electric
currents. For this work, he can be considered the Hfather" of electromagnetism.
In the same year, 1.B. Biot and F. Savart formulated the law that gives magnetic
fields as generated by filamentary currents, and in 1826 O.S. Olun established the relation
between electric fteld and current. In 1831 M. Faraday described the law of induction and
introduced the concept of magnetic lines of force. (He also revived the concept of Ether,
"the vacuum�filling medium"; see remarks in the introduction to section 2.5.) In the
following two decades, electromagnetic phenomena were gradually formulated in more
exact mathematical-theoretical tenns by the contributions of C.F. Gauss, W.E. Weber, W.
Thomson, R. Kohlrausch. H. Helmholtz. and others.
 1.1 MAXWELL'S FIELD EQUATIONS 3

In 1855 l.C. Maxwell further extended the ideas abom field lines; seven years later he
introduced the concept of displacement currents, which self-consistently completed the
electromagnetic equations. His Treatise on Electricity and Magnetism, U first published in
1873, and Ihe conlribulions in 1884-85 by H. Hertz and O. Heaviside gave Maxwell's
equations their final form. After lP. Joule stated the equivalence of heat and mechanical
energy in 1845, H.A. Lorentz (who formulated the electromagnetic force on an electric
charge in 1879), J,H. Poynting, and P.N. Lebedev gradually introduced the concepts of
electromagnetic force and energy into the theoretical framework in the second half of the
19th century. Clearly, a large number o f other important physicists and mathematicians
contributed substantially to establishing electrodynamics. Many of them are mentioned in
the text in connection with their fundamental work (full names and dates are listed in the
Index).
In the first half of the 20th century. classical electromagnetic theory was coupled with
quantum mechanics into quantum electrodynamics. As one of the four fundamental fields of
forces-thus of energy-in Nature [gravitation, electromagnetism, weak. interaction
(nuclear beta-decay), strong interaction (nuclear reactions»). electrodynamics was blended
in the second half of the century with the field of weak interactions into the unified electro­
weak. theory, one important step toward including all four fields in the so-called Grand
Unification.
The great applications of electricity started while classical electromagnetic theory was
being completed. In 1879 the first railway vehicle was driven by electric power in Berlin,
and three years later Th. A. Edison built the first electric power station to partly supply
lighting to New York City. This marks the beginning of a new evolution, which
. characterizes the 20th century, since the availability of electricity introduced increasingly
sophisticated and efficient applications of it in industry. transport. and communications.
In concluding these historical notes LlS it is instruclive to give a glance at table 1. 1-1,
which shows that the magnetic fields of interest today extend over about 102J orders of
magnitude. This extremely large range reflects the numerous aspects of the subject we shall
cover in this and the following chapters.

Table 1.1-1 Typical (upper limit) magnetic fields iD tesla

associated with utrophyslcal, technical, nuclear. and human
sources

Intergaloctic space < 10- 13 Superconducting coils 20

Brain jWI2 Fe-nucleus. hyperfine field 30

Heorth 10-10 Hybrid coils 40
Galaxy < 3.10-10 Pulsed coils 60
Earth 10-4 Ho-nucleus. hypcrfine field 730
Sunspots < 10- 1 Laboratory flux compression 103
Permanent magnets Uhrahigh intensilY lasers 104
Black holes 105
P ulsars 108
Neutron slars, up to 1010
4 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

Differential equations

Maxwell'sfield equations in SI units (International System of Units) are here written

in the vectorial fonn

ao
V'xH=j+­ (1.1-1)
ar

�
V'xE=-­ (1.1-2)
ar

'1·8 =0, ( 1.1-3)

( 1.1 4)
-

where the five vectors. one scalar function. and one operator arc:

H, magnetic field strength (or magnetic intensity) [dimension: ampere-rn-I; for the
electromagnetic units see table A.l-ll in appendix A.I];
B. magnetic flux density (or magnetic induction) [tesla];
D, electric flux density (or electric induction or displacement) [coulomb·m·l1;
E, electric field strength (or electric intensity) [volt-m·I];
j, free current density (that is, the current density related to the transport of free electric
charges) [ampere'm-l];
Pe. volume density of free electric charges [coulomb·m-)];
V. nabla operator (defining the operation curl, divergence, and so on; see table A.3-1f in
appendix A.3) [m·').

The equalions are also known as !he laws of Am�re-Maxwell (1.1-1), Faraday (1.1-2),
Gauss-Faraday (1.1-3), and Gauss (1.1-4), but these denominations are used more
appropriately in connection with the corresponding integral fonns presented in section 1.4.
In the previous differential equations, j and Pe
can be considered as the sources that
determine the electromagnetic fields H. D, E. D. They are related by the charge or
current conservation equation.

, ilPe = o
ov·J+- ( l.l-5)
at

which is oblained by laking Ihe divergence of ( l.l-I) and using (1.1-4) and Ihe facl lhal
divergence of a curl is zero, thereby conunuting the V and atat operators (because we
assume a space where, at each point, the field vector and all its derivatives are continuous).
To make a general solution possible, three more equations are required, which are
known as constitutive equations, that is, Ohm's law
 1.1 MAXWELL'S FIELD EQUATIONS 5

j=oE , (1.1-6)

and the relations

B=J1H , (1.1-7)
D=eE . ( 1.1 8)
-

where a is the eiectric conductivity [dimension: ohm-1.m-I].

(1.1-9)

is the magnetic permeability with 110 = 4JrlO- [henry-rn-II. and

7

(1.1-10)

is the dielectric constant (or permjllivity) with eO = 8.854 x 10-12 (farad'rn-I] (the
dimensionless parameters J1R and£R are discussed below). These quantities, which
characterize the medium, can themselves be functions of various parameters (for example.
the temperature. or even H itself), in addition to space and lime. For the more general case,
when the medium has nonisotropic propenies with respect to electromagnetic phenomena,
these parameters actually become tensors (see the end of appendix A.3)3.2. The whole
problem then becomes fonnally quite cumbersome, but nowadays such cases can be treated
by numerical computation (see chapter 9). In this book. however. we shall limit our
aUenlion to isotropic media and nearly always assume the electric conductivity (J, relative
magnetic penncability 1lR. and dielectric pennittivity ER to be constants.

Material-related electromagnetic quantities

We have seen that the magnetic field H is related to the free current density j
through Ampere's law ( l.l-I), whereas the magnetic induction or flux density B is related
to the electric field E through Faraday's Jaw (1.1-2). We shall see throughout the book that
B is the dominant magnetic vector quantity because it appears explicitly in all magnetically
induced effects: electric fields, magnetic forces, moments. and so on. For this reason, and
for simplicity, B is often also called the magnetic field, which well matches the electric field
E, the dominant electric quantity (rather than the electric flux density D) since it appears in
all electrically induced effects: currents, forces, moments, and so on.
The magnetic and electric properties of a medium can be described with the help of
two vectors, the magnetic (M) and electric (P) polarization vectors:

B
M=--H (1.1-11)
/10
6 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

P=D-EOE. (1.1-12)

which exist only in a medium since they actually vanish In free space where
B=I'oH. D=EOE. Introducing them in to (1.1-1. 4) yields

B
Vx-=j+ -+VxM
((Jp ) +£0- OIl
(1.1-13)
1'0 dt dt

EOV·E=p,+(-V·P) . (1.1-14)

These equations formal ly show that when material is present in an electromagnetic field,
internal sources of currents and charges appear. In fact, the terms

. (Jp
irn =VxM. J =­
P dt
(1.1-15)

n
may be i t erpreted as material-related equivalent magnetilJltion or electric-polarization
current densities, whereas

Pp=-V.p (1.1-16)

may be interpreted as an equivalent electric charge density introduced by the electric

polarization.
In the following we shall give further relations and information on these macroscopic
magnetic and electric properties of material. Then. in chapter 8 these quantities are put in
relation with some microscopic elements of material.

Magnetic quantities

It is an expetimentaJ fact that in most materials. when subjected (0 a (curre nt

generated) magnetic field H, an additional magnetic field component M is generated locally
(defined as magnetic polarilJJtion or magnetization), and the two add together to give the
total local induction
(1.1-17)

or. in the practical egs system [B in gauss; H in oersted; M in erg G-1. cm-3],
B=H+41tM . (1.1-17)*

Sometimes. the magnetic polarization is also defined as

( 1.1-18)
 1.1 MAXWELL'S FIELD EQUATIONS 7

By introducing the magnetic susceptibility

( 1 . 1 - 19)

we can write

( 1 . 1-20)

Note that if a given outer magnetic field Ho is applied to a finite body of magnetic
material, the local magnetic behavior is not, in general, described by the above relations
containing the substitution HO -+ H; for example, the magnetic induction in the body is not
B =I'HO but B=I'H;, where the internal local field H; (which determines the local
magnetization M=XmHi) is itself the result of the addition of the outer field and the
magnetization component. For the same reason, the induction outside the body is
B = JioHe. where the external local field ne is co-determined by the magnetic effect of the
body. To detennine Hi at any point within the body and If outside it requires solving a
magnetic (potential) problem, as outlined in sections 2.1 and 2.2. The example of a
magnetic rod is dealt with later in this section and will help to get a better understanding of
these magnetic field components. A simple case is given by a closed magnetic circuit or a
ring of magnetic material (as discussed in section 1.4 in connection with figure 1 .4-3a, with
x = 0) with an evenly wound coil around it, which generates a magnetic field Ho within

the structure. In this idealized situation it is simply He = 0 and Hi = HO, thus B =tlHo
everywhere. The magnetic properties of materials. expressed by the magnetization M,
depend on two main atorrllstic effects, which can give rise to large local magnetic fields, as
we will discuss in more detail in section 8.3 [see, in particular (8.3-26,36)]: (I) the orbital
motion of electrons around the nucleus, which can be seen as current loops of atomistic
dimensions or as small magnetic dipole moments; (2) the intrinsic spin of the electrons (or
nuclei) with the related magnetic dipole moment. The relative magnetic penneability PR or
the magnetic susceptibility Km ' which define M through ( 1 . 1 - 1 9, 20), vary widely, as
shown in table 1 . 1 -11. In the so-called diamagnelic malerials the susceptibility is negative; in
paramagnetic, positive; and in jerromagnetic. very large (Xm => 104 ). Moreover, it can
depend in a complicated way on H (see in section 8.3).
We have already seen in ( 1 . 1 - 15) that from a macroscopic poinl of view magnetization
may be expressed in tenns of an equivalent magnetization current density by

(1.1-21)

Alternatively we will see in ( 1 . 1 -23) that it may also be described as a volume density of
magnetic dipole moments, to which we will relate in ( 1 . 1 -27,29) a magnetic charge density.
These three macroscopic models are Obviously equivalent and justified by the atomistic
explanation of the magnetization given above.
When M is uniform there are no such currents in the medium; instead, a magnetiwtion
surface current density as defined in ( 1.4-35),
( 1 . 1 -22a)
8 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

Table 1.1·11 Magnetic s usu:ptibilitiesa

Material Magnetic susceptibility Xm

Diamagnetic:
Bismuth - 1 7.6 x 10-S
S ilver -2.4 x 10-5
Copper -0.88 )( 10-5
Water -0.90 x 10-5
Carbon diollide - 1 . 2 x IO-s
Paramagnetic.'
Oxygen 0. 1 9 x l O- 5
Sodium 0.85 x 10-s
Aluminum 2.3 x IQ" 5
Tungslen 7.8 x l O so

Gadolinium 41:1"000 x 10-5

Ferromagnetic (mn,values):
iro, 30 x 10]
Iron-nickel 80-300 x 103
£ISee section 8.3 (tables S.3- VI. VII) for more detailed information.

appears, which. on the boundary of the medium (e.g., surrounded by free space), has the
value [see in (1 .5-6)]

Im = -n x M . (1 1 -22b)
.

where n is the unit nonnal vector to the boundary, pointing outwards. Ampere already
suggested that magnetic properties might be described by such formal currents. They are

considered to be made up of elementary cells that include circulating, whirling currents, as

depicted in figures 1 . 1 - 1 and 1 .1 �2, which is in qualitative agreement with the above
expressions. In fact, when M is constant and the whirling current cells equal, all the
currents, except those on the surface of the medium. cancel each other out at the common
boundaries of the cells, thus giving rise to the surface current density. These equivalent
Amperian currents, pictured as flowing without dissipation in a magnetized medium. are an
artifact that can be used to describe magnetization effects in simple physical terms. In
section 8.3 we show that qualitative support of this picture is provided by the atomistic
explanation of magnetization, i.e., by microscopic. inaccessible curreOlS of atomic origin.
The magnetization M can be seen as the magnetic moment per unit volume in the
fann

pm d I
= Z'PQ x Jm.Q
•

M PQ = dl': (1 .1 -23)
Q

where rPQ is the coordinate vector with respect to a point of origin P of the current
density jm Q . The total magnetic moment Pm of a given volume of material is thusLS
.
 1.1 MAXWELL'S FIELD EQUATIONS 9

I __� jm

Figure 1.1-1 Magnetic material can be thought of as consisting of elementary

cells containing Amperian current densities Jm; addition (subtraction) of these
currents provides a bound surface currenl density im flowing on the boundary. which
exists even if the elementary amperian currents are identical and thus cancel each
other out in the interior (when the magneti7.alion M is unifonn). Alternatively, a
magnetic dipole moment Pm can be associated 10 each whirling current cell. (For
simplicity. hen: all moments are aligned. but in reality they could also be at
ntndom.)

( 1 . 1 -24)

which is a purely kinematic definition (nol involving any magnetic interaction) that is
independent of the choice of the point P. Note that this expression is formally identical to
the magnetic moment (2.2-49) for a free current density. In fact, it can be calculated
similarly and with the same assumptions made for the multipole expansion that will lead to
(2.2-49). For example, from this integral it follows immediately fas we shall see in (2.2-
45)1 that for a filamentary loop. carrying the current / and enclosing the area S, the
magnetic moment is

Pm = IS , ( 1 . 1 -25)

where the vector S is perpendicular to che plane of the loop.

Magnetic charges and nonsolenoidal fields

Another formulation of magnetic properties is ba.."ed on the [oonal concept of

magnelic charges or poles. We rewrite V·B = 0
with B ::::: PoH + .uOM as

where
10 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

Figure 1.1-2 Magnetization M and related current density Jm "" V x M.

Pm = -V · M
,
( 1 . 1-27)

in analogy [0 the electric polarization charge density ( 1 . 1 - 16), can be considered as a

magnetic cho.rge density, which is the source of the nonsolenoidal field component Hn.
We use the magnetic field decomposition ( 1 .1-35), where for the solenoidal component it is
V · U' = 0.
When M is uniform, there are no such charges in the bulk of the material under
consideration; however, a magnetic surface charge density defined as in ( 1.4-36),

fm = lim Pmtlh , ( 1 . 1 -28)

&1--+0
Slill exisls al lhe boundary, which is given in (1.5-11),

Fm = -n · (M2 - M 1 ), ( 1 . 1 -29)

and the relative nonsolenoidal field component (for simplicity we drop the superscript n) is
defined in ( 1.5-10),

( 1 . 1 -30)

where n is again the outward IX'inting unit vector, from medium 1 to 2, nonnal to the
boundary. These relations show that a magnetized material has at its surface formal
magnetic charge densities 'm. which arise whenever the nonnal component of M goes
through a discontinuity. The magnetic charges generate in the interior of the material a so­
called demagnetizing field � that by definition points from the nonh (+) to the south (-)
 1.1 MAXWELL'S FIELD EQUATIONS 11

pole [similarly to electric fields pointing from (+) to H electric charge] and is thus opposed
to M.

( 1 . 1 -3 1 )

Here km. defined as the demagnetizing/actor, is in general the result of an exact calculation
[see table 1 . 1 -111, and the example of a magnetized sphere. in connection with (2. I 7 1 80))
- ,

and depends on the geometry and other parameters of the arrangement (in special cases, it
could be a tensor). If an external field Ho is applied to the material, the magnetization ( 1 . 1-
201.

originates from the action of the local internal field

thus

H = Ho /( I + kmXml ( 1 . I-32a)

which is smaller than the applied field. If there is no applied field, H o = 0, but a
pennanenl magnetization, M = Mo. the local field is simply

The charges also generate an external field Ht, pointing again from the north to the south
pole, which roughly speaking has dipole shape and opposite sign to the internal
demagetizing field Hd [qualitatively, this ensures that the circuitation of H around any
closed contour line is equal to zero, as required by Ampere s law ( 1 .4-4b) because there are
'

no free currents; for more details see the examples illustrated by figures 1 . 1-4, 2 . 1 -8, 2. 1 -
9).
The properties of magnetized material fonnulated by magnetic charge densities have to
be equivalent to those obtained with the magnetic moment or current densities introduced in
( 1 . 1 2 1, 23), For the magnetic moment description ( 1 . 1-23) it suffices to consider each tiny
-

magnetic dipole made up of a pair of magnetic charges ±g [with dimension A·m] at distance
I , such that its moment is

( t . t -33)

When the dipoles are aligned (figure 1 . 1 -3) there is, locally, addition or subtraction of
these charges in such a way that even if the di�le density were constant (thUS the charges
fully neutralize in the bulk) there would remain a surface charge density ± r m = ± g S, as
given by ( 1 . 1 -29), where S is the surface area. As we have seen in ( 1 . 1-23), the magnetic
dipole moment can alternatively be ascribed to the Am�rian current loops. In conclusion,
which of the three magnetization models (magnetic charge, dipole moments, Am�rian
currenl� ) to use is dictated by con venience in each problem. as we will see with some
12 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

<:=-E> E -E> <:�-E>

<:=-E> <:=-E> <:=-E>
<: -E> <:�-E> <: -E>
Hm

<:=-E> E -D <:=-E>
M �

Figure 1.1·3 Magnetized malerial can be thought of as consisting of

elementary magnetic dipoles; the addition (subtraction) uf the magnetic charges
provides a bound surface charge density ± r.... which exists even if the dipoles art:
identical.

examples. In any case, the equivalence of these models often provides an interesting
consistency check for a COlTect understanding of magnetic problems.
We have thus shown that, as a consequence of 0 . 1-3), V · B "" 0, the formal
magnetic charges appear intrinsically in pairs. Contrary to free electric charges. they cannot
be separated in MaxweU' s theory. (With tbe artifact of a long magnetic rod we can formally
distinguish between nonh and south charges, but they always appear together: If the bar is
cut open in two hairs, each one will show north and south charges. )
However, there i s n o fonnal theoretical reason in electrooynamics that prohibits the
existence of isolated charges. Their existence would introduce a magnetic charge density
Pmp; and for symmetry reasons it is to be expected that the associated magnetic current

"Pi'
density j mp' defined by

V · Jrnp
' =- , ( \ . I ·34a)

would produce an electric field. simi larly 10 Ihe electric cUITenl densily j. ( 1 . 1·5). lhal
produces a magnetic field H , ( 1 .2-1). The existence of free magnetic charges would thus
enhance the formal symmetry of the electromagnetic equations. lA For example. the
magnetostatic and electrostatic equations ( 1 .2-1 to 4) with no magnetic and electric
polarizations would be respectively

V x B = /loi . V . H = Pmp ( 1 . 1 ·34b)

and

vx E = /loimp V ' D = p, . ( \ . I -34c)

to which are added the relations ( 1 . 1-5. 34a) between the charge and current densities. and
 1.1 MAXWELL'S FIELD EQUATIONS 13

we would also add the force densities to be introduced in (6. 1-12) and derivable from (8.3-
65).
Despite experimental investigations lasting decades, the existence of free magnetic
charges. sometimes called maglletic monopu/es, has nO( been established. But extensive
theoretical work in quantum electrodynamics and general relativity. more in general in the
domain of elementary particle physics, have outlined some of the characteristics of magnetic
monopolcs. if they exist at all, which will be presented in connection with (8.3-64).
It is useful to recall at this point Helmholtz 's theorem. according 10 which any static
vector field at any point in space (which. together with its derivative. is finite, continuous.
and vanishes at infinity) may be decomposed intol.' [see also ( 1 . 1-54)]

H= Hn +Hs , ( 1 . 1 -35)

where H O = -V4> is the irrotational component and H!> = V x A is the rotational or

solcnoidal component. The first component can be directly related 10 the magnetization M
through ( 1 . 1 -27.29); the second, to free and displacement currents through the
Ampere-Maxwell equation ( 1 . 1- 1 ).
Because from CA.3-7. 1 1 ) we have for these scalar and vector fields

V x V4> = O . V · (V x A) =O .
wc obtain

V · H = V · H" . V x H = V x H' . ( 1 . 1-36)

Example: Magnetic rod

Consider a cylindrical magnetic rod placed in a homog.eneous field Ho with direccion parallel 10 the:
rod axis (figure1 . 1 -4), For simplicity. we assume a very large aspect ratio. 2a « I. Thu , at the rod ends
s
there will be magnetic charge densities rm given by ( 1 . 1 -29), which art sources (+rm ) and sinks (·rm )
for magnetic field line�. which. by definilion, poinl from ( + rm ) (nonh) lowards ( - rm ) (south). TIle:
internal demagnetizing Held lid given by ( 1. 1 ·3 1 ) has the opposite direction to M. In general. the vector
lield Hd will depend on the distribution or M and on the geometry or the magnetic medium and is the
�olution or a magnetic potential problem (see section 2.1). Here. ror the cylinder with unirorm
magnetization along its axis. we can write the result in the simple approximated rorm

( 1 . I·37)

where the demagnetizing ractor km is given as a runction or the aspect ratio in table 1 . 1·111.
The lotal internal induction. directed along Ihe axis and along HO. is

( 1 . 1·38)
14 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

11'

T
2n
� 4!
n P

Figure 1.1-4 Very long magnetic rod to which a uniform outer field Ho
parallel 10 the axis is applied. inducing a magnetization M = (PR- I )H;; the related
m
magnetic swface charge densities ± r sets up a demagnetizing internal field
Hd -kmM. where k m is a geometric factor detennined from the general solution
;;;

of the corresponding magnetic potential problem.

whereas the (total) external field ne results from Ho. with the superposilion of a dipole-type field generated
by the magnetic rod (which is found through the solution of a magnetic potential problem). Here, we are
interested only in the internal field Hi and the external field just al the end surfaces �. both dim:ted along
the axis. for which we find
(1.1-39a)
that is. with
M=(J..l R-l)Hi

also
(1.I·39b)

and, from the boundary condition (1.5-3),

(1.1·40)

also
(1.1·41)

For a paramagnetic material with IlR > 1. and since 0 S km S I. the following will always hold:
HI S HO' H; � Ho. In particular. for a soft ferromagnetic material (section 8.3) with B ='llollRH and very
large permeability IlR. the internal magnetic field (1.1-39b) tends to vanish: We can say it will be "eltpel1ed"
because the magnetic charges are "free to move" towards the ends where they neutralize the given field.
 1.1 MAXWELL'S FIELD EQUATIONS 15

Table 1.1-111 Demagnetizing factor ror cylindrical macnet i.22 a

1123 0 I 2 5 10 lOO
km I 0,27 0,14 0,040 0.0172 0.00036
a For a sphere from (2.1-71) it is km = 113.

For a permanently magnelized rod with M = Mo independent of H and with no applied OUler field,
HO = 0 (figure 1 . 1 .4), from ( 1 . 1-39, 40) follows

( 1 . 1·42)

Only when the rod is very long, 2a « I, and thus km « 1 . will the outer field at the magnet poles be
H: == Mo. We can fonnally express these propenies also with the equivalent descriJXion involving the
(bound) surface current density i m . Since JUSt outside the cylindrical surface of the (very long) magnetic rod
the field is zero. from the boundary condition ( 1 .5-5) we derive

( 1 . 1.43)

The: magnetic field in the external free space produced by a pennanently magnetized rod is equivalent 10 the
dipole-type field generated by a solenoidal coil with the same cylindrical geometry and the current density
i m . The internal rlClds are quite different, as requirtd to satisfy the integral conditions ( 1 .4-4); In the tOO
Hi Hd -kmMo; in the equivalent cylindrical coil H I = H= = ( I -km )MO.
=: =

Electric quantities
For completeness' sake, we also present the analogous quantities relative to dielectric
material. Similarly to ( 1 . 1-17), for the elecLric displacement vector, from experience and in
relation to ( l . 1 - 12), we write

( 1 . 1 -44)

where the electric polariwtio1l is

( 1 . 1 -45)

and the electric susceptibility is

( 1 . 1 -46)

The electric polarization P is directly related to the electric propenies of materials

(which can thus be expressed through the parameters Xe and ER) and, according to
16 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

( 1 . 1 - 16), can be interpreted as deriving from an equivalent poiariwtion charge density

through

pp = -v . P . ( 1 . 1·47)

Somewhat analogously to magnetic material [see in connection with ( 1 . 1 -29)}, when P is

unifonn there are no such charges in the medium. However, a bound polarization surface
charge density ( 1 .4·36) is slill present, p = Iim(.:1h --> O)pp.:1h , and on the boundary of
r
the medium (e.g.. surrounded by free space) it has the value (1 .5·14),

( 1 . 1 -48)

and the electric fields on the boundary are defined in ( 1 .5-13),

n · (E2 - Et )EO = rp . ( 1 . 1 -49)

The model to interpret this polarization effect considers that the meclium contains electric
dipole charges that align under the action of an electric field. When the dipole density is
constant, the ± charges neutralize, but there is still a surface charge density, qualitatively
like the magnetic dipoles shown in figure 1 . 1-3. We recall that molecules or ions can have
pennanent dipoles, or dipoles generated through an applied electric field that displaces the
center of positive charges with respect to that of negative charges.
The eleclric polarization P can be seen as the electric moment per unit volume
(somewhat analogously to the magnetization (1.1·23)],

PQ = d V. = Pp,Q"PQ
d

Q
Pe
P ( 1 . 1-50)

where
Pp,Q is the net, locally bound polarization charge density and

vector with respect to a point of origin

rp,Q is the coordinate
P. The total electric moment of a given volume of
material is

( 1 . 1 -5 1 )

which is independent of the choice of the origin P.

Ohm's law

Under the action of a force, an electric charge moves in a conductor. thereby

establishing a current. We write

j = of , ( 1 . 1·52)
 1.1 MAXWELL'S FIELD EQUATIONS 17

where r is the mean force exened per unit charge, and the empirical material constant (J
(the electric conductivity) varies according 10 the materials (rable l . t -IV) and parameters. In
principle, r could include a variety of forces, such as gravitational. centrifugal. and so on,
but here we are interested in electromagnetic fields, thus (as discussed in section 8.3)

j = aE . ( 1 . 1 -53 )

that is. the electric field E drives the current density j in a conductor at rest. This
phenomenological linear relation, Ohm's law, is not always valid; fortunately, it is so in
metallic conductors within a large parameter range [see in connection with (8.3-3)]. In
some anisotropic situations (e.g .. in some crystals. or when there are significant magnetic
fields), the electric conductivity mllst be expressed by a tensor1.2 [see also in (A.3-33)].
What kind of special electric field can drive a current in a conducting loop located in
space? It is useful here to recall once again Helmholtz's theorem ( 1 . 1 -35), according to
which any vector can be decomposed into

( 1 . 1 ·54)

where E n "" -VU is derived from a scalar potential U and is thus the irrotational or non­
solenoidal component. and ES = V x C i s the rotational or solenoidal component derived
from a vector potential C. The gradient theorem (A.3-13) applies straightforwardly to the
potential u:

b b
U(b) - U(a) = J(VU) . dl = -J E n dl , ( 1 . 1-55)
a a

that is, the potential difference along an open path is due solely to the line integra1 of the
irrotational electric field component. We define the electromotive jfJrce (emt) e along a
closed contour C by

e- f E . d ' - f(E" + E' ) · d l = fE' · d l ( 1 . 1 -56)

c c c

where the line integral over the irrotational component En vanishes according to ( 1 . 1 -55)
when b � a .
With the local current density

j = a(E" + E') ,
( 1 . 1 -57)

driven by either En or ES according to ( 1 . 1 -55, or 56), the voltage drop along a conducting
circuit (e.g.. wire) can be written as

( 1 . 1-58)
18 CHAPTER 1 FOUNOATION OF MAGNETIC FIELD THEORY

Table l.t·IV Rnistlvities at 20°C

Material Resistivity 11 = lID' (ohm-meter)

Conductors;
Silver 1.59 X 10-8
Copper 1.67 x t o-8
Gold 2.35 x \0-8
Aluminum 2.6.5 x 1O-!
Nichrome 100 x 10-8
Semiconductors:
Sail water (�turated) 0.044
Gennanium 0.46
Silicon(dependingonpurity) 300-400
Insulators:
Waler (pure) 2.5 x IOS
Wood 1()8-I011
Glass 1010_1014
Qwutz 1013
Sui fur 2 x 1015
Rubber 1013_1016

where

R= f! ( 1 . 1-59)

is the resistance of the whole conducting circuit, 1 = jI is the total current, and we have
assumed that j and a are constant across any cross-sectional area .r (filamentary
conductor and/or stationary current approximations).
For a stationary as well as a quasistationary current, from ( 1 . 1-5, 57) and because
V · E" = 0 follows

V . j = V . (oE') = O ( 1 . 1 -60)

[see also in (1.2-2, 21)], which means that the current paths close on themselves or extend
to infinity and that a stationary or quasistationary current in a closed circuit is driven by an
emf due to a rotational electric field extending in space [such a field can be provided, e.g.,
by a time varying magnetic flux, as shown by ( 1 . 1 -2), or more explicitly by ( 1 .4-9)]. Often
the emf in a circuit is generated by a local source, such as an electronic voltage generator, a
battery, a charge separating device. or a thennocouple.
The previous expressions applied to ftlamentary conductors give rise to Kirchhoffs
two laws for an electric circuit. Consider, for example, the triple junction of figure 1 . 1-5.
By inlegrating ( 1 . 1-60) with Gauss' theorem (A.3-l6) over a space delimited by a closed
surface S surrounding the junction. we get
 1.2 ELECTROMAGNETIC FIELO APPROXIMATIONS 19

,
,

Figure t.I-5 CUITenl� flowing into a circuit junction (Kirchhoffs first law).

J \1 . j d V f l d
= s = 11 + 12 + 13 = 0 ( 1 . l -6 1 )
,

More in general. Kirchhoffs first law states that the sum of the currents out of a junction
(or node) is zero. that is,

( 1 . 1 -62)

where j ' d s , hence In . are mken as positive when the currents are flowing OUl of the
junction. In addition. by extending ( 1 . 1 -SS) around any closed loop. we write, with
Ohm's law and en = Un .

( 1 . l -63)

which represents Kirchhoffs second (restricted) law. It can be extended to quasistationary

currents and then includes the potential difference across inductances ( 1 .4-10),
capacitances, and pure resistors [see (4.5-1)]. These two laws represent the basis of the
electric circuit theory. In section 1.4 we shall see that for the so-called magnetic circui[ two
fonnally similar laws apply.

1.2 ELECTROMA G N ETIC FIELD

A PPROXIMA TI O N S

The complexity of Maxwell's equations makes it mandatory to look for simplifying

conditions. or approximations. In this section. simplifications are introduced by
considering the time rate of changes that appear in the Am�re-Maxwell and Faraday
20 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

equations (1.2-1, 2): if there is no time dependence, the static equations with the related
phenomena apply; if the rate of changes is sufficiently slow with respect to the dynamic
phenomena of interest, the quasistationary approximation is valid. In fact, this book IS
mostly devoted to the study of static and quasistationary magnetic field approximations.

Magnetostatics and electrostatics

In the cases where there is no time dependence, the displacement current term
aDI i!t in ( 1 . 1- 1 ) drops out, and Ampere's law is simply
VxH=j ( 1 .2- 1 )

or, in practical egs units (oersted, ampere, cm),

V x H = O.4nj ( 1 .2-1)*

As a consequence (since the divergence of any curl vanishes), the cu"enl conservation
equation ( 1 .1-5) reduces to
V'j=O , ( 1 .2-2)

which expresses the fact that all current lines either close on themselves or extend to
il il
infinity, with no charge accumulation ( Pe l t :::: 0). In addition, from ( 1 . 1 -3) we know
that

V·B=O , with B = I'H , ( 1 .2-3)

meaning the magnetic flux: lines are solenoidal (they either close on themselves or go to
infinity; see also section 2.5).
In magnetostatics, basically three groups of problems are solved (see also the
schematic in figure 2.2-3, chapter 2).
(a) Given Ht find j [a relatively simple problem, particularly in the integral form for the
total currenl I, as in Ampere'S law ( 1.4-3)].
(b) Given the currents, find H (this problem is the main subject of the first two chapters
and basically requires the solution of a second-order differential equation).
(c) Given the magnetostatic fields. determine the inductances, forces, and energies
related to it (treated in the following chapters).
Note that the analogous electrostatic equations,

V xE=O ( 1 .2-4a)

V · D =Pe' with D = EE , ( 1 .2-4b)

are obtained by requiring no magnetic flux density variation, aB I t :::: 0 and

d ilPe I ill :::: 0,
 1 .2 ELECTROMAGNETIC FIELD APPROXIMATIONS 21

that is, Ihe charge density Pe to remain constant at each point in space. As a consequence of
these equations, there is no interdependence between magnetostatic and electrostatic fields
and they can be treated separately.

Quasistationary magnetic fields

In the quasistationary (or quasistatic) approximation, some time dependences of the

fields are explicitly allowed. This obviously requires the conditions for toHerablc time
variations 10 be defined with respect to the dynamic processes of interest. For this purpose
we rewrite the time-dependent MaxweU equations ( 1 . 1 - 1.2) in a dimension less form, with
quantities marked by an asterisk (* ), by introducing a characteristic system dimension 10
and a characlen'stic dynamic time r (e.g., for a harmonic time variation it would be
<=1/ ",),

· ,
r=Jor· (i.e., x = /ox , etc.), t = Tt , ( 1 .2-5)

which also yields

a I a
- -> - ­ ( 1.2-6)
(Jr < (Jr'

and a characteristic magneticfield Ho.

,
H = HOH . ( 1 .2-7)

We want to preserve Faraday's equation in the form

n '
y x E"' - - --
_ aB'· ( 1 .2-8)
ai
which implies

E = 1l1o
Ho E '
( 1 .2-9a)
<

With this and Ohm's law ( 1 . 1 -6),

j=oE , ( 1.2-9b)

Ampere's equation ( 1 . 1- 1 ) is

2 � ,
n'
V X H* - -
_
1'0 --
fm E· + - UI!. 0.2-10 )
, ,2 Ut"

where
22 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

ro = �
c ,
( 1 . 2- 1 1)

is the time required by an electromagnetic (plane) wave to propagate down the characteristic
system dimension to at the light velocity c;

( 1 .2-12)

is the magnetic or current density diffusion time (that is. the characteristic time an
electromagnetic field tied to a current density requires to diffuse into a conductor. as will be
discussed in chapter 4); and

£
fC = ­
(j ( 1.2-13a)

whereby

is a jorrt1lJi eleclricfield or charge relaxation time in the conductor-that is, a characlcristic

time after which a variation of the electric charge density. or its related electrostatic field,
senles to stationary conditions. In fact. from the continuity equation ( 1 . 1 -5), Gauss' law
( 1 . 1 -4), and Ohm's law ( 1 .2-9b) we obtain the equation for the free electric charge density
Pc'
�+ �Pc = O ( 1 .2-13b)

which has the exponential solution

(1.2- 1 3c)

The inconsistency introduced by the charge relaxation time Te will be discussed below,
Within a conductor, the displacement current [that is. the last term in (1.2· 10)] can be
neglected with respect to the conduction current (first term on right-hand side), provided

� ro = re « l
2
( 1 .2-14)
tm f
2 r
Neglecting the displacement term in free space, where there is no conduction current
«j = 0, rm = 0), requires

ro / r « 1 , ( 1.2- 15a)

which, by introducing the wavelength of the imposed (harmonic) electromagnetic

field, A. = 2m: r , and with (1.2-11), yields

( 1 .2- 1 5b)
 1.2 ELECTROMAGNETIC FIELD APPROXIMATIONS 23

This means that the whole conductor system is subject to the same quasistationary field;
there is no tag effect in the field propagation, whose velocity may be regarded as infinite.
This conclusion is well known from antenna theory, where currents, whose wavelength is
long compared to the largest dimension of the antenna conducting circuit, do not give rise to
appreciable radialion [see in (3.3-16)].
In electric conductors, where the effects of magnetic fields predominate over those of
electrostatic fields, conditions ( 1 .2-14, 15) are generally well satisfied. For example, for a
copper conductor (0' 6 107 ohm- I m- I , l = 0.3 m) the times are typically 1'0
= = 10-9 s, fm
= 7 s. re 1.5 10- 19 s.
=
However, Ohm's law ( 1.2-9b), which has been used here to establish both the
dimensionless Ampere equation ( 1 .2-10) and the charge equation (1 .2-13), is valid only for
time scales much larger (han (he average time between collisions of the free electron (with
electric charge e and mass me) in the conductor

( 1.2-16)

as will be shown in connection with (8.3-4); for copper it is typically 'tf = 5 . 10- 14 s (table
8.3-1). For the above considerations to remain valid in a good conductor, it is at least
necessary that T be long compared to Tp in which case the current will be in phase with
the electric field. It is, therefore, physically meaningful to take, instead of Te, the much
longer relaxation time r.. as a limit in most situations. The physical phenomena, which
define the effective charge relaxation time in a good electrical conductor, are more subtle.
but the rough estimate of Te--+r.. remains fairly valid. loB In any case, for typical
quasistationary magnetic-dominated phenomena, where 't � 10-6 s, it follows that

(1.2-17)

whereas T can be of the same order as 'rm' We conclude that the quasistationary magnetic
field approx.imation holds at least for aU frequencies up to those corresponding to the
infrared spectrum.
Consequent to these conditions, the displacement tenn in Ampere's equation ( 1 . 1 - 1 )
[as well as the charge density variation tenn in ( 1 . 1-5), which directly derives from it]
drops out, and the magneloquasistalionary equntions are

VxH= j . ( 1.2-18)

V X E = _ilI ( 1.2-19)
ar
V·B=O . ( 1 .2-20)

where B "" J.lH and j "" oE. The equation

V·j=O ( 1 .2-21)
24 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

is a resuh of (1 .2-18), as was pointed out in connection with ( 1 .2-2). The dynamics in this
magnetoquasistationary approximation enters through the Faraday law ( 1.2-19). There is
also the equation

for the unlikely circumstance of wanting to know a posteriori the very small free charge
density Pe in a magnetic problem.
In conclusion. the magnetoquasistationary approximation is a powerful and very
useful theoretical formalism because it holds for the majority of fields used in
electromechanical engineering and for some in radioengineering.

Quasistationary electric fields

It is instructive to present the electroquasistationary equations pertinent to an electric­

field-dominated problem, which is typically found in a dielectric or very poorly conducting
12 l I
medium. For example, in a glass insuJator «(1 = 10 ohm- m- , I = O.3 m) we find
1'0 = 10- s, 'I'm = 1 . 1 10- s, 1'e = 9 s; that is,
9 19

( 1 .2-23)

whereas r can be of the same order as t'e ' Using similar considerations and procedure to
that for the magnetoquasistationary case, we get the etectroquasjstationary equations:

( 1 .2-24)

( 1 .2-25)

VxE=O . ( 1 .2-26)

where D = rE and j = oK There are also the equations

aD
V X H=j+ ( 1 .2-27)
dt

[Ihis one being consislenl with (1 .2-24. 25)J and

( 1 .2-28)

which can be used in the unlikely circumstance that the extremely weak magnetic field must
be known a posteriori in the electric quasistationary approximation. Note that here the
dynamics enters through the charge conservation equation (1.2-25).
 1.3 ELECTRODYNAMICS O F MOVING MEDIA 25

1.3 ELECTRODYN AMICS O F M O V I N G

MEDIA

In this section we introduce two generalizations regarding the study o f electrodynamic

phenomena: First, we consider two coordinate frames moving with velocity v with respect
to each other, and we study how the electromagnetic quantities transfonn in the two
systems; second, we allow the medium (to which the electrodynamic phenomena refer) to
flow with velocity u ::: u(x,y,z) and study its effects. This study provides the basic
equations of magnetohydrodynamic theory, a subject that is expanded in chapter 8.

Coordinate frames of reference

We consider two (Cartesian) coordinate systems (one primed and the other unprimed)
moving with relative, constant velocity " (figure 1.3-1). Relativity theory describes how
physical variables change by switching from one coordinate system to the other. The

r=r-L,
LorenlZ transfonnation prescribes how time and the space coordinate change: With

{J = " . ( 1 .3-1 )
.JI-(�r
c ·

we haveJ.l 2

( 1 .3-20)

Figure 1.3·1 FiKed or laboralOry-frame coordinate system (x.Y.z). and re.�I­

frame system (x·.y·,z') moving witll constant velocity v relative to tile fonner. in the
Galilean coordinate transfonnation approximation (here. tile rest frame is defined in
relalion 10 a rigid body moving with velocity v within which it is at rest).
26 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

where the subscripts I and .1 indicate components of the position vector parallel and
perpendicular la V. For example, for Cartesian coordinates with the primed system moving
with velocity v along the z-direction:

' et' = r(ct - /k) .

" = r(' - (3ct) . x =x, y' = y , (1 .3-2b)

or
'
, = re,' + (3ct') , x=x , y = y' , et = r(ct' + P,') . ( 1.3-2e)

When {3 « I-that is, r == 1 (or, formally. c --+ -}-these relations reduce to the Gali/ean
trans/omIQtion:

r' = r - VI . I' -.. / . ( 1.3-3a)

or, as above,

, ' = , - vl , x' = x . y' = y . ,' = 1 . ( 1.3-3b)

The Galilean transfonnation applies to the quasistationary electromagnetic equations, since

in both fonnalisms the same assumption applies ( c � - ) Both Lorentz and Galilean
.

transformations are based on lhe approximation of a nonaccelerated movement- that is,

approximately constant v. They are applied within the special theory of relativity and
express the relation between the quantities given in nonaccelerated. or so-called inertial.
frames of reference.'·"" Otherwise. Einstein's general theory of relativity applies. However.
objects moving with velocity u within each frame can be accelerated.

Moving coordinate system

We are concerned in general about the relationship between two enlltles,

electromagnetic fields and media wilh electromagnetic properties (conductors, magnetic
materials, etc.), so we must understand how the properties expressed by them change when
viewed within two frames of reference in relative motion.
With respect to the second entity, we consider first a medium that presents itself as a
rigid body. We define as the rest-!r"",,, (or moving) coordinate system, characteriud by
primed parameters (figure 1.3-1), the one where the body and any observer P' placed on it
are at rest, but moving with relative constant velocity v with respect to the laboratory or
fued/rame (labeled with unprimed parameters).
We want to detennine the differences in the electromagnetic quantities and properties
perceived by an observer P' at rest in the moving frame versus those perceived by an
observer P in the laboratory frame, with respect to the same phenomena.
 1.3 ELECTROOYNAMICS OF MOVING MEDIA 27

Maxwell-Lorentz transformation

Fortunately. Maxwell's equations ( 1 . 1 - 1 to 4) remain the same in all frames,

regardless of their relative (nonaccelerated) motion. In fact, it is a postulate of special
relativity that physical laws must be the same in any inertial coordinate system. In parallel to
the unprimed equations ( 1 . 1 - 1 to 4) valid in the laboratory frame. we have the identical
primed equations valid in the moving frame

V' X H' = j' + �' ( 1 .3-4)

V' x E' := - c:: ( 1 .3-5)

V' · B' = 0 ( 1 .3-6)

( 1 .3-7)

where the primed nabla operator is V' == (a Iax',a I ay',a I dz'), The constitutive equations
( 1 . 1-6 to 8) are strictly valid for the medium only in its rest frame. so

j' = G'E' • ( 1 .3-8)

R' = ,uH' , D'= EE' . ( 1 .3-9)

The parameters (J.j1.e of the medium should also be primed, since they could be
dependent of the relative velocity; but this effect is neglected here.
The problem then is to determine the transformation rules that must be applied to
switch from the one set of equations to the other. For example, an observer P in the
laboratory frame wants to know the primed electromagnetic quantities in the moving frame.
but expressed as a function of the unprimed quantities of his frame (which he can measure).
In the general formulation these transformation rules can be found by applying the
Maxwel/-Lorentz transformations [i.e., the Lorentz transfonnation ( 1 .3-2) for the coordi­

�
nates, plus the condition that Maxwell's equations remain invariant to it]. For the operators

- -( l -
from (1 .3-Zb) follows [for frames moving along the z-axis (v,. 0, v,. 0)] �

a _ z:. a ilL L _ a p a _
a a a _ a
a; iIz iJI.' + iIz ar' r Ji' - c7i" ' a. - ;v, dY ay' , ( 1.3-10)

(1.3-11)

We are now in the position to show under which conditions the unprimed Maxwell
equations ( 1 . 1 - 1 to 4) transform into the primed ones ( 1 .3-4 to 7). For this purpose we
28 CHAPTER 1 FOUNOATION OF MAGNETIC FIELD THEORY

write the former in their components and apply the above operators.S. 1 We find that the
resulting primed equations are identical to (1.3-4 to 7). provided that we set [the subscripts
.1 and 11 refer to the nonnal (x.y). and parallel (l) components relative to the constant (z­
component) velocily1

( 13 - 12)

Bi =r(B - -'" x E)l. ' ( 13- 13)

c

( 1 .3-14)

( 13 - 1 5)

)i =h ' ( 1.3- 16a)

I . )' )
Pe, = r( Pe -" ( 1 . 3-1 6b)
c

From these relations we note the relative nature of magnetic and electric fields: Their
magnitude can be different in different reference systems. For example, even if it were
H = 0 in the unprimed system. we can have a finite H ', ( 1 .3-12), in the primed system.
This is in contrast to the constancy of Maxwell's equations-thar is. (he laws governing
these fields.

Maxwel/-Galilean transformation

Fortunately, the operations are easier for nonrelativistic velocities since they remain
invariant to the simpler Gali1ean transformation ( 1 . 3-3). In addition to the space
transformation, the following rules [which derive from (1.3- 10, 1 1)] apply to the
derivatives of a scalar function cP, or vector A:

V' .... V ( 1.3- 1 7.)

J..
(Jr'
.... (!!... + ,. V)=�DI
(Jr
( 1 .3- 17b)

where the last operator is called the convective or Lagragan

i time derivalive (which can also
stand for a total time derivative). The first rule states that the spatial derivatives are the
 1.3 ELECTRODYNAMICS OF MOVING MEDIA 29

same, irrespective of the frame in which they are taken. This follows logically from the fact
that an instant physical picture of the phenomenon must show up identical in the two
frames. Clearly, the situation is different with the lime derivative taken in the two frames
(second rule). For an observer P' in the primed rest frame. by definition, the time derivative
is taken with the primed special coordinates held fixed; but in the laboratory frame. on the
contrary, any function cP, or vector A, which is in the position r ;;: (x,y,z) al time I moves
into position r+ v.1t during the following time interval .1 / . This implies the application of

(ox oy it..)
the chain rule differentiation, which. when applied to a scalar function cP = 4>(x, )', z. t) with
,.
V - ill ' ", ,,, ' s
�

( 1 .3- 1 8 )

and which, analogously, when applied t o a vector A ;::::: (Ax,A ,Az), is

y

( 1 .3-19)

I n particular, using (A3-5, 9) we get

04> d<P
- � - + V · (<Pv) ( 1 .3-20)
01 ar

if V·v = 0 [i.e., for a constant velocity or an incompressible body, see (8. 1 -4)]; and

O A JA
- � - + v(V · A) - V x (v x A) , ( 1 .3-2 1 )
01 ar

if i n addition ( A · V)v = 0 li.c., for a rigid body translating with constant velocity, see,
e.g., (8. 1 -43»).
With respect to the Maxwell-GaJilcan transformation (where C� 00, Y == I), the
relations ( 1.3-12 to 16) can be rewritten in the form

E' = E+vxB ( 1 .3-22)

H' = H - v x D ( 1 .3-23)

j' = j - Pe v , ( 1 .3-24)

and

H' = H , D' = D , P� = Pe . ( 1 .3-25)

plus the constitutive equations ( 1 .3-8,9). For example, to determine H', E' from Maxwell's
30 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

equations (1.34 to 7) and using (1.3-17,21) we obtain

V x H' = a(E + v x B) + ao - V x Cv x 0)+ Pev ( 1 .3-26)

al

aB
V x E' = - + V x (v x B) ( 1 . 3-27)
al

( 1 .3-28)

The magnetic quasistalionary approximation ( 1 .2·18 to 20) is compatible with the

Galilean transfonnalion (both fonnally assume c----+ 00), and in this case the set of relations
(1.3-22 to 25) reduce to

E' = E + v x B , ( 1.3-29)

whereas

H' = H, B' = B. J'" = J ( 1 . 3-30)

remain untouched.
Equation ( 1 . 3-29) is also intuitively deduced from the Lorentz force (6. 1 - 1 ). In fact.
an observer moving through a magnetic field and carrying a charge q will experience a force
q(E + v x B), which is fonnaJly equivalent to saying that in his frame of reference he sees
the electric field (1 .3-29). A direct consequence of (1 .3-29. 30) is that [by substituting into
( 1.3-6)) Ohm's law must be written as

j = a(E + v x B) ( 1 .3-3 1 )

In practice we can say that this law applies in any frame of reference that moves across
magneticflux lines (or in which the body carrying the current j moves with respect to the
magnetic field source).
For completeness' sake we point out that the Galilean transfonnation rules ( 1 .3-3.17)
are also consistent with the transfonnation of the primed into the unprimed electric
quasistalionary equations (1 .2-24. 25.26). provided that

( 1 .3-32)

H' = H - v x O . ( 1 .3-33 )

whereas
'
E' = E . 0' = 0 . Pc Pc=
( 1 .3-34)

remain untouched. The current density is here affected by a convection current that in the
magnetic approximation is implicitly negligible with regard to the conduction current.
 1.3 ELECTRODYNAMICS OF MOVING MEDIA 31

Moving media

An important generalization of the electromagnetic system so far considered is

introduced by substituting the rigid body with a conducting medium that flows with
approximately constant velocity u = u(x,y.z) with respect to a laboratory frame of
reference. A rest frame can here be defLned only in relation to a single fluid element
L1 V'centered on point P', since all the other elements would generally have different
velocities (figure 1 .3-2), On this basis, however. we can apply all the previous results
established in relation to the laboratory and rest frames by substituting v � u.
Before moving on, it is useful to revisit the concepts we have just introduced.
According to our definitions, (1.3-29) means, for example, that the observer pi in the rest
fmme (x'.y', z'), which moves with the (relative) velocity u of the fluid element 11 V'with
respect to the laboratory frame, detects an elecUic field E' = E + U X B, with the important
addition of the electric field component u x B that results from the crossing of the magnetic
flux lines. Here, the fields E and B are supposed to be generated (and measured by the
observer P) in the laboratory frame. Clearly, it is the field E' measured by the observer P'
that makes the current flow in the conductor placed in the rest frame (x', y', z '). E' drives
the current density l' = DE' = a(E + u x B) in the fluid element 11 V ' . Alternatively, the

the same current j

As in
=
observer P in the laboratory frame, who measures the fields
1 a(E + u x B) in the medium that flows past him.
=

( 1.3-27) we rewrite Faraday's law (1 .3-5) in terms of E' and B = B':

E. B . and the velocity u. sees

V x E' = - ': + V x(u x B) ( 1 .3-35)

v
"
v B'

E' l' (x.Y.z)

y'
rest frame
"

fluid v

laboratory frame

Figure 1.3·2 Fixed- or laboratory-frame system (x,y.z) (where: the fields are
generated). and the rest-frame system moving with the fluid element .dV' centered
on P'.
32 CHAPTER 1 FOUNOATION OF MAGNETIC FIELD THEORY

It is customary to introduce the Lagrangian derivative (1.3-21). so that

DB ( 1 .3-36)
V x E' = -
Or
These equations. together with

VxH=j ,
V·B=O ,
E' = E + u x B ( 1 .3-37)
j = aE' ,
B = IIH ,

provide the set of electromagnetic equations used by magnetohydrodynamic theory (in the
magnetic quasistationary approximation),
By considering d'd'= 0/01 -+ d I dl as a total time derivative and having introduced
E'. the above equations are written in wgrangitut form, i.e., with respect to rest frames
(x',y',z') moving with the fluid (a fonn often convenient in computationa1 evaluations). If.
on the contrary. we reintroduce the electric field E and take into account the fann ( 1 .3-
2 1 l, then Ihe equations return to the Eulerian form thal is, wilh respecl to the laboratory
-

frame fixed in the (x. y. z) space.

1.4 ALTERNATIVE ELECTROMAGNETIC

EQUATIONS

It may sometimes be convenient to use Maxwell's relations in different forms to those

presented at the beginning of this chapter. This is particularly true for the integral form,
which introduces the overall, lumped parameters of voltage, current, flux. and electric
chorge, with which an understanding of electromagnetic phenomena is often more
straightforward.

Integral laws

Ampere's law

By integrating (1.1-1) over any (open) cap surface S bounded by a closed contour C
(figure 1.4-1) and applying Stokes' theorem (A.3-14l,
 1 04 ALTERNATIVE ELECTROMAGNETIC EQUATIONS 33

"

<1>.1_- ___

surface S

dl

Figure 1.4·1 Surface S limiled by a contour line C.

f(VXH)ods = fH o dl ( 1 .4-1)
s c

we get

( 1 .4-2.)

H
(hat is. the circuitation (line integral) of around any closed contour C is equal to the (ora)
normal current flowing across the corresponding cap surface S bounded by C,

IS =
f jods • ( 1 .4-2b)
s

plus the time rate change of the normal D-flux across S. The positive direction around C
and of the nonnal unit vector n on S follow the right-hand-screw convention. Note that if
Ihe surface and contour move wilh velocily v. only Ihe lasl leom in ( 1 .4-2a) is affected.
similarly to the magnetic flux, as will be indicated in ( l .4-9), In the quasistationary
magnetic approximation (when the last term can be neglected), this reduces to Ampere's
law
fHodl = /S ( 1 .4-3)
C

or, in practical cgs units (oersted. ampere. cm), to

fH d = 0 4
o
l 0 1tls 0 ( 1 .4-3) *
c

If we separate

H= H" + H'
34 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

as in ( 1 .1-35) into the irrotational component Hn and rotational component H S , we find

from Ampere's law ( 1 .4-3)

f Hs . dl = /s ( 1.4-4a)

t H" . dl = O
because

, ( 1 .4-4b)

according to the gradient theorem (A.3-l3), The solenoidal component H S is completely

independent of whether magnetic materials are present or not along the path (there is
simply no IlR factor in these equations).

Faraday 's law

Similarly, by applying Stokes' theorem ( 1 .4-1), the Faraday law follows from

t E dl = -f �s = - ! f B ds ,
( 1 . 1 -2):

e= ( 1 .4-5)
c s S

or, in practical cgs units (volt/cm. gauss, cm),

( l .4-5)*

(Note that the pania/ differentiation can be exchanged with me integral over the surface.)
The induced electric field along C produces in the circuit the (inductive) electromotive force
e ( 1 . 1 -56), which here is due to the time variation of the magnetic flux

f
IfI = B ds ( 1 .4-6)
s

within the fixed contour C; so, we write

e__ [iJlfI]
at \1 =0
( 1 .4-7)

More in general, when the circuit C moves with velocity v, the emf along C from

t t (E + v X B) . dl .
(1.3-29) is

e = E' . dl = ( 1 .4-80)
C C
 1.4 ALTERNATIVE ELECTROMAGNETIC EQUATIONS 35

For the last integral term, with the definitions v = dr I dr, drx dl = ds, and the b'ivector

J = _ a;
relation (A.3-Id), we write

Cf(v X B) . dl � -f B . (v X dl) � -
C Slat
B . ds / .1t [ ] ( 1 .4-gb)
8=con�t.

To obtain this integral. the following remarks are useful (figure 1.4-2):

(a) A lateral surface Slat is defined as the contour C moves to C' in the time interval Al.
(b) The differential area element of this lateral surface is I v x dl l .1r . and consequently

( l .4-Bb) represents the increment L1'1' of the magnetic flux through the surface S lat
during time L1t.
(c) Since B is a solenoidal field [see ( 1 .4-13)J. the lotal flux Ihrough Ihe closed surface
S +S lat+S ' is zero; Iherefore ( 1 .4-8b) is equal to the difference Ll'l' of the fluxes
through the open surfaces S and S' (the outward pointing ds is opposite on S and S'),
(d) The resulting .1¥'h1t teon is expressed by the last equality, since the partial time
derivative stands for the rate of change of flux due to the motion of the contour C in a
constant field B.

In conclusion, by adding the contributions from ( 1 .4-7,8), for the electromotive force
e or the induced voltage U in a closed circuit in the general case, we obtain

e - u - _ [ a", ] _ [ a", ] = _ d l/f ( 1 .4-9)

- - d v=O dt B=const dt '

where the total time derivative now stands for the total rale of change of the magnetic flux
through the moving contour C. By introducing the inductance parameter L, which will be

'X;C(IOOP ...-- S
at tImet)
-
"

,
,
,

--,
'..:'"
,
,

dl s· C (loop al llme r+d rj

Figure 1.4-2 The COn!our C moving with velocity " = drJdl defines after a
lime.1t a cylindrical surface with lateral. top. and bottom surfaces.
36 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

defined in more detail in section 2.4. the Faraday law can also be written in the form

u = - diU) .
( 1 .4- 1 0)
dl

F.
This result could have been obtained directly by using the generalized Leibniz rule. A.I valid

�JF ds = J [ aF +(V'F)v] .ds- f(FXV).dl ,

for a vector

( 1 .4- 1 1 )

F
dl
S S '" C

and applied to the magnetic induction, 1:1 B . where V · B = O. In any case, the positive
direction around C and of the normal unit vector n on S follow the right-hand-screw
convention. The integral form of the Faraday law thus stales that the emf, or the circuitation
of E, around any closed contour C is equal to the negative time rate of change of the lota]
normal B-flux across any cap surface S bounded bye, whether this change is due to a time
variation of B or to motion of the contour C. The negative sign is determined by Lenz's
law: The current is (or would be) driven by the emf and will be in such a direction that the
magnetic field it produces tends to counteract the change in flux that induced the emf.
Faraday's law OA-8) is the basis for the generation of an inductive emf and electric
current in an associated circuit. Two fundamental methods follow from ( 1 .4-8a): either by
magnetic flux variation as applied in an alternator generator, or by movement within a
magnetic field (by "cutting field lines") as in a homopolar generator. Both applications will
be discussed in section 6.2, in connection with figures 6.2-9 and 1 3 ,

Miscellaneous laws

By applying Gauss' theorem (A.3-l6) over any volume V bounded by a closed

J (V B)dV f B ds
surface S,

= , ( 1 .4-12)
v s

we get with ( 1 . 1 -3)

fB.ds = O , ( 1 .4- 1 3 )
s
that is, the total outward normal B-flux across any closed surface S vanishes. (This is
evidently a general consequence of the rotational nature of D.) Similarly. the current
conservation (1 .2-2 or 1.2-21) yields

§j.ds = O , ( 1 .4-14)
S
 1.4 ALTERNATIVE ELECTROMAGNETIC EQUATIONS 37

see Kirchhoffs law ( 1 . 1 -62). Integration of ( 1 . 1 -4) gives, for the electric flux density,

fD.dS = fPedV . ( 1 .4-15)

S V

These integral laws remain unaffected when the surface moves across the electromagnetic
field lines.

Differential laws

Maxwell's differential laws in a fluid moving with quasiconstant velocity u are given
in (1.3-26, 27). As an example of the many possible variations, we introduce explicitly the
magnetic and electric polarizations ( 1 . 1- 1 1 , 12) in the fluid,

( 1.4-16)

D = EOE + P . ( 1.4-17)

and require that the current density j include only the terms that in the frame-transfonnation
are related to free electrical charges,

j = a(E + u x I'oH) - Pe u ( 1 .4-18)

thus excluding the additional term ux J.loM deriving from the atorrUstic currents, as

required by (1.3-22). This particular formulation of Maxwell's equation is sometimes called

I
the Chu-formulation. . B We deduce from ( 1 .4-18) that on this basis the Galilean
transformation relations, instead of ( 1.3-22, 23), are

E' = E + u x I'oH ( 1.4-19)

H' = H - u X EOE . ( 1 .4-20)

By inserting them into (1.3-26, 27), we obtain a fully consistent set of equations:

a
V x H = j + ar (eoE + P) + V x (P X U) ( 1 .4-2 1 )

( 1 .4-22) .

and from ( 1 . 1 -3 to 5)
38 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

V · H = -V · M , ( 1 .4-23)

( 1 .4-24)

V . j = _OPe ( 1 .4-25)
at

In addition, from ( 1.3-25) we have

8' = B . 0' = D . p; = p� ( 1.4-26)

c = <PoEa )
-112
hence with also we obtain

JIoM' = JIoM + -tu x E, P' = P - u X JIoH . ( 1 .4-27)

c

These equations easily can be reduced into the magnetic or electric quasistationary forms.
similarly to (1.3-29 to 3 1 ) or ( 1 .3-32 to 34). When u = 0 we obtain the original Maxwell
equations ( 1 . 1- 1 to 8).

Current densities

As mentioned in section 1 . 1 , it can be useful to formally express some terms in

Maxwell's equations as current densities. In analogy to the vacuum displacement CUffent
density.

( 1 .4-28)

(already recognized by MaxweU as a necessary term to complete the nonstationary

electromagnetic equations even in empty space), in ( 1 . 1- 15) we have introduced. by
analogy. the magnetic and electric polarization current densities defined as

( 1 .4-29)

The component

( 1 .4-30)

can be defined, by its very nature, as a capacitive current density. whereas

 1 . 4 ALTERNATIVE ELECTROMAGNETIC EQUATIONS 39

. j
JI = + Jm = 11;;
. [ VXB] ( 1 .4·31)

is an inductive current density, where j is the current density sustained by free charges. The
total current density is

( 1 .4·32)

where

( 1 .4·33)

represents the current densities that can be considered as being generated by the motion of
free and/or bound electric charges, since they are related 10 the respecti ve charge densities

V.jj + =a:-
by

Jp = o , i = nil,d,m,p, or a combination of them, ( 1 .4-34 )

[see also in ( 1 . 1 ·5); nOle that the current and charge densities corresponding to free charges
are designated throughout the book as j, Pe. the laner to be distinguished from the mass
density p I .
For later use we introduce the concept of a surface currenl density ij defined as

ii = l im 1.1h ; ( 1 .4- 35 )
6h-+O
that is, the current flowing as current density ji in a layer of depth i1h from the conductor
surface is compressed into an infinitesimal surface thickness .1h -+ 0 [see comments in
connection with (1.5·33) for the free surface current density]. Clearly. the same definition
can be applied to the charge densities,

'I = lim Pitih , ( 1 .4·36)

""...0

so that

V, · i·, +�
ar
=O ( 1 .4·37)

where the index. i is the same as in ( 1.4·34).

The surface vector i is defined by two surface coordinates (for ex.ample. for a plane
shut and Cartesian coordinates) by (x,y). Consequently, here the nabla operator Vs is a
surface operator with derivatives along its surface only (see also in (4.4- 12)]. For example,
in Cartesian coordinates with the unit vectors e .r ' ey along the x, y axes. we obtain

(1 .4-38)
40 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

Magnetic circuits

For the sludy of magnelic syslems composed of ferromagnelic components Ihat

almost entirely confine the magnetic flux. it is very convenient to resort to the magnetic­
circuit concept, which is analogous to the elecoic (resistive) circuit. The circuits shown in
figure 1.4·3 and the listing of the corresponding parameters in table 1.4-1 illustrate this
concept. In fact. the application of Am�re's law ( 1 .4-3) to an appropriately chosen mean
contour line C in the magnetic circuit of figure 1 .4-3a gives the so-called magnetomolive
force (mmt) or magnelomolance

( 1 .4-39a)

where

( 1.4-39b)

is induced here by the Ne windings carrying the current le. Since the nonnaJ magnetic flux
through any cross section S is constant [because of ( 1 .4-1 3)],

( 1 .4-40)

.)

I
/12 - - - - - -...
- - - - - - - - -,
,
8
S
I, , ,
, ,
Po ,
+
RIU
I
,
N� ,
,
x
,
, Po « #
,
, ,
, ,
/12, -------- ------- ,

RI2
/

RI2

Figure 1.4-3 Analogy between magnetic (a), and electric circuit (b).
 1.4 ALTERNATIVE ELECTROMAGNETIC EQUATIONS 41

we can also write

( 1 .4-4 1 )

where

'.
RHn = -- ( 1 .4 -42)
5P• •

is the reluctance of the nth magnetic circuit element; for example, RHx = x I(J1oS). Note that
here we have implicitly made some assumptions: (a) constant flux along me magnetic
branches In (no significant leakage. see remark below); (b) small gap, x « .JS (no fringe
fields, straight field lines); (c) small aspect ratio of the closed magnetic structure (definition
of a mean contour line of length In and field Hn); (d) constant magnetic

penneability f1 = 1l0IJR (no saturation effects). Relation ( 1 .4-41) is fonnally analogous to

Kirchhoffs law ( 1 . 1 -63) for the electric circuit in figure 1 .4-3b,

V = IR+ IRx • ( 1 .4-43)

with the equivalent parameters given in table 1 .4-1. In ( 1 . 5 - 18) we will show that the
tangential field and the normal induction,

( 1 .4-44)

Table 1.4·1 Equlvaleot parameters of magnetic and electric circuits

Magnetic circuit Electric circuit

Definilion Symbol and relation Symbol and relation Definition

Magnelic flu, density '" B � pH j� oil Current density (Ohm's

induction law)

Magnetic permeability J.I = J.lOJ.lR G Electric conduclivilY

Magnetic field H E Electric field

Magnetomotive force (mmf) '

VH = H · dJ = Nclc V � jE-dI Voltage
[see Ampere's law (1.4-4)]

Total magnelic flux

", = 85 = "* l = jS = 7i Total current

(= jB-d,) d
(= j j - s)

Magnetic reluctance RH = �s R= � Resistance

...L 1.
RH R
Pt:rmeance Conductance
42 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

remain steady when crossing the boundary of the magnetic material. The first boundary
condition implies that there is also some magnetic flux (leakage flux) external to the
magnetic branches (for example. as shown in figure 9.l-7� see section 9.1); however, the
assumption of constant flux ( 1.4�40) is in most cases sufficiently correct because the ratio
of internal (J1lII ) over ex(ernal (J1o Ht ) induction is very large for a ferromagnetic material,
since pI JJo » I. (Note the difference to the electric circuit, where no current is present

outside the conductor, because there (1= 0.) The second boundary condition, Bn constant.
is expressed by ( 1.4-40), which in combinalion wilh the flux delennined by ( 1 .4-41)
defines the magnetic field in the gap,
H = Nc/c
x H_ '
( 1 .4-45)
PR

This field can be relatively large for a narrow gap width x in a ferromagnetic core
(PR » I) . [In an electric circuit a gap would fully stop the current, here it only introduces a
large reluctance RHx = x 1(l1oS) in the magnetic circuit, giving rise to the large magnetic
field.]
By applying these laws to two closed loops and one junction of the circuit shown in
figure 1.4-4, and on the basis of the assumptions made before, we obtain

VHI + VH2 = 'I'IRHI + 'I'2RH2 .

VHI = 'I'IRHI + 'I'3(RHJ + RH4 + RH. ) ( 1 4-46)

.

'1'1 = '1'3 + '1'2 '

where

( 1 .4-47)

and for the tapered branch leading to the gap (from Sx to S3) we find

( 1 .4-48)

From these relations it is easy to find all the required quantities; in particular. the field in the
gap will have a form somewhat similar to ( 1.4-45) but with an additional amplifying factor
SISx: [For the magnetic forces acting in the gap, see in connection with (6.2-50).]

The rules eslablished before in (1 .4-40.41) can be eXlended to more complicated

circuits by applying the magnetic version of Kirchhoffs laws ( 1 . 1-62, 63) concerning the
magnetic flux 1jI and motive force VH in a junction and around a magnetic loop.
respectively:

( 1 .4-49)
n
 1.5 BOUNDARY CONDITIONS 43

11
IJn
RHI RH'
f - - - - - - - J. - - - �- - - , I,
, , RHJn
, , 1"1
RH412 '1',
I" N"
,
,
VHI RH, VH2
RH412
N" 1"
-

SI RHJf2
'I'I\}, I.y? I"J

Figure 1.4·4 Magnet with three branches and tapered gap in air, and related
circuit.

( 1 .4-50)
n n

Designs of electric transformers, where time varying currents are transfonned from primary
to secondary currents, largely follow the considerations made above.

1.5 B O U N D A RY C O N D I TI O N S

An electromagnetic problem is generally characterized by its boundary (and initial)

conditions, which detenrune the specific solution of the appropriate electromagnetic
equations. In facl, these equations are common la all problems, but the boundary
conditions along with the geometric configuration characterize in particular each specific
problem.
Generally, electromagnetic problems are referred to free space containing various
media. The latter have electromagnetic properties defined by quantities such as the
conductivity (1 and the magnetic and dielectric parameters J1 and £, which are related to
the main electromagnetic fields and currents through the constitutive equations presented in
section 1 . 1 . At the boundary separating two media. (1,11.£ might be discontinous, and
consequently also the related fields and current vectors. The conditions of how the latter
change when crossing the boundary are called boundary conditions. In this section these
conditions will be derived from Maxwell's equations in their integral forms. {The boundary
conditions regarding the magnetic potentials cP and A , which are introduced in the next
chapler, are given following (2.1-6) and (2.2-23).)
44 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

Field conditions

General conditions

The fundamental boundary condilions for the tangential and nonna) electromagnetic
field component5 between two media will be derived in ( 1 .5-24, 25. 26. 29. 32); they are

o x (H2 - HI ) = ; ( 1 .5- 1 )

o X (1':2 - 1':1 ) = 0 ( 1 .5-2)

0- (8, - 81 ) = 0 , ( 1.5- 3)

n ' (D� - Dl ) = ro: , ( 1 .5-4)

where n is the unit normal vector on the boundary pointing from medium t to medium 2
(see figure 1 .5 - 1 ); i and re are the surface current and charge den�itics related to free
electric charges [according to definitions ( 1 .4-35, 36»). which can exist in superconductors
where the electric conductivity is infinite. a -t 00, Note that if in medium 2. (12 -+ then D<> .

H2 = 0, and the boundary field H , is purely tangential.

B = JlO " + .uoM. and obtain

It is instructive to rewrite [he boundary condition ( 1 .5- 1 ) by introducing the
magnetization M through ( 1 . 1 -17),

where

( 1.5-6)

is the surface magnetization current density defined by ( 1 .4-35). By inserting

( 1.5-7)

into ( 1 .5-1), an alternative to ( 1.5-6) is

( 1.5-8)

since. in fact.

( 1 .5-9 )

Also, condilion ( 1 .5-3) cao be separaled wilh ( 1.5-26b) inlo Ihe form

0 - (H2 - H I ) = rm ' ( 1 .5 - 1 0)
 1.5 BOUNDARY CONDITIONS 45

medium 2

tu
ph
__ __ ium 1 __
rn«I _....__

Figure 1.5-1 Boundary surface between two media labeled I and 2. 'The
orthogonal coordinate system, defined at any surface point P with the unit vectors
(p.�.n), is such that p,s lie in the plane tangent on P at the boundary. with n
normal to it. R, is the projection of B onlO the tangent plane. and it can itself be
composed of the components B, and D,.

where

is the surface magnetic charge density defined in ( 1 . 1-28). Allematively, condition ( 1 .5-3)
can be written as

(Note that jf medium 2 is superconducting ( c � QC ) and does not admit any magnetic
field. it is ,u I H1, n = 0 ; that is, magnetic fields are tangential to a superconducting surface.)
Similarly. introducing the electric polarization P from ( 1 . 1 - 12), D = £oE + P. into
( 1 .5-4) gives

( 1 .5-13)

where

( 1 .5-14)

is the surface electric polarization charge density defined in ( 1 . 1 -48). [Note ( 1 .5-25) and
that the expression for the electric surface polarization is obtained similarly as in ( 1 .5-26)
for the magnetic polarization,] With
46 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

( 1 .5-15)

a modified fonn of (1 .5-4) is

( 1 .5 - 1 6)

and a modified fonn of ( 1 .5-2) is

( 1 .5-17)

Nonsuperconducting media

In the magnetic quasistationary approximation and for normally conducting media

(where I and re vanish), conditions ( 1 .5-1 to 4) mean that the tangential (t) and nannal
(n) components

( 1 .5-18)
and
( 1 .5-19)

remain steady when crossing the boundary. This also means that the other componems Ho.
B,. and En, D,. in conforming wilh ( 1 .5-8, 12) and ( 1 .5-16, 17), may jump according to
the magnetic and electric propenies of the adjoning media 1,2 according to

Hn.2 J.l l
-- = - ( 1 .5-20)
Hn.1 112

a) b)

m<:dium 1 medium 2

n .H

Figure 1.5-2 Refraction of a magnetic field line when crossing the boundary
between twO magnetic media (here. qualitatively 1-/2 < Ill). and (b) decomposition of
the related field vectors.
 1.5 BOUNDARY CONDITIONS 47

and

( 1 .5-21)

These conditions detemtine a refraction of the field lines as they cross the boundary
between media t and 2, which is similar to Soell's refraction law in optics (figure 1.5-2),
For example, for the magnetic field, conditions (1.5-18) can be rewritten in scalar fonn as

from which, by eliminating HI, H2• follows

Ianq,l l!i
= ( 1 .5-22)
tan.p, 1'2
and analogously for the electric field from conditions (1.5- 19). This means that if a H(or E)
field line enters a medium with a smaller magnetic or dielectric coefficiem, it is refracted
towards the nonnal. We also see (figure 1.5-2b) that when crossing the boundary. the
tangential cOffiIXment Ht remains untouched [see ( 1 .5-18)], but the nonnal component
increases by the amount 'm given in ( 1 .5-lO). Also note that if for medium 2. J.l2 = 00, then
4'1 = 0, or Ji,H1,t = 0; that is, the field at the boundary is purely normal. Conversely, if
112 = O. the field at the boundary is purely tangential, which is formally analogous to the
superconducting case 0'2 = 00 [in both cases, medium 2 is fully diamagnetic; see also
comments made in relation to (1.5-1). J

Example: Calculation of the conditions

It is instructive to derive the boundary conditions (1.5-1.3) [and by similarity (1.5-2.4)] by using
the integral forms of Maxwell's equ8Iions near the boundary. Consider (1.4-13), which is the integral form
of V · B = O, that is.
fB d = O
. S .

applied over the surface of the sufficiently small cylinder shown in figure 1.5·3 with height ilh. and em
surfaces ils, and cui by the boundary surface separating the media 1 and 2, so as 10 obtain

(DI · B . + D2 · B2)4s'+ f B w · ds = O , ( 1.5-23)

w

where the first term represents the flUlt leaving the end surfaces ( BI, 82,Bw are taken to be the mean values
over .heir respective surfaces) and the integral gives the flux leaving the side wall W of the cylinder.
Wc let the cylinder height shrink towards Ihe boundary surface. that is. ilh ---? 0, in which case the
flux contribution from the side wall W vanishes because it depends on the wall surface, which with ilh ---? 0
48 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

�n tn, , ,
-r
,

, j

O,jl,£ medIUm I

Flllure 1.5·3 Boundary surface between media I and 2 with elemental cylinder
and loop used to derive the boundary conditions for 8, D and H. E, respectively.

also goes (0 zero. As the ends can remain finite. though small (& * 0), from the above equation with an

appropriate labeling (n 0; "2 = -" I ) follows

Similarly. from ( 1 . 1 -4), V · D = Pe . follows

(nl ,D1 + 02 , D 2 ) & + JDw · ds = J PcdV . ( 1.5-25a)

w

The last inlegral represents the free electric charges contained in the cylinder and is .. PctJItis. As .1h -+ O.
it vanishes, unless on the boundary there exists a surface charge density re as defined in ( 1 .4-36), and we

obtain

( I .5-25b)

When the cylindrical body has magncti:t:ation M or elcclric polarization P. we proceed in the same way. For
example. with B = IJoH + iJoM we obtain. instead of ( 1 .5-23),

(nl ' HI +n2 ·H2).ds+ JHw ' ds + f M ·ds = 0 (1 .5-260)

w

The last integral. by inlrOducing the magnetic charge density Pm "" -V · M , transforms with Gauss'
theorem into .. pmA.t.1h: and, as above, letting .1h � 0 we find

( 1 .5·26b)

where the surface magnetic charge density (l.l-28) can also be written in the form (1.5-1 1).
To demonstrate the boundary condition (1.5-1) we apply Ampere's inlegral law (1 .4-2),
 1.5 BOUNOARY CONDITIONS 49

f H dl � J(J+ !,)ds , ( 1 .5-27)

c s

over the small rectangular loop C with surface S shown in figure 1.5-3. We introduce the orthogunal
coordinate system with unit vectors (p. s, ft) such that p = s x n .dI = p.dl (parallel branches), .dh = n.1h
•

(venicaJ branche.o;). p = P2 =-p,. n = n2 = -R, and use, as before. the appropriate labeling
P = P2 = -PI' n = n2 = -RI_ The above integral relation can thus be rewritten approximatdy as

( 1 .5-28)

We let the loop height shrink towards the boundary, L1h -+ 0, whereas dl. though small, can remain finite
( til � 0). in which case the second term (derivil1g from the venical branches) vanishes and this relation
reduces 10

( 1.5-29)

To obtain this boundary condition we have: (a) used with me vecTor idemity (A.3-ld) the relation

( 1 .5-30)

Cb) introduced the surface current density i defined in ( 1 .4-35); and (c) assumed that D remains bounded; that
is. for tlh � O. Iim(Odh) --t O . More general. we could have introduced two
components.ll --+ i,1 +i,2' as both media could. in principle. carry surface currenl.. on their common
boundary.
By proceeding similarly. from Faraday's integral law ( 1 .4-5) we obtain

( 1 .5-31)

from which we get the boundary condition

( 1 .5-32)

provided [like in point (c)] that B remains bounded while ah --+ O.

Current density conditions

From the current conservation equation ( 1 . 1-5) and by the procedure used for (1 .5-24,
26), we gel

n ' (J'2 - J- ] )A_

al = (Vs ' ')A.
l al + d(p,&tJ
dt A•
al .

only that here the surface integral over the side wall W of the elemental cylinder [see (1.5-
23. 25)] gives a contribution. providing a surface current density
50 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

i = lim (j&.) ( 1 .5-33)

&.->0
exists. In fact, note that Gauss' theorem (A.3-16),

jj-ds = jV j dV .
W

n.1hdl, dV == �hL1s can be transfonned into the related surface relation

fj ds -&.-:-:---+-·f-'o;c->' (n . i)dl = j
with d s ==

(Vs · i)ds = (Vs · i)& • ( 1 .5-34)

W
where Vs is the surface operator ( 1 .4-42). By letting tlh � 0, the above main equation
reduces to

( 1 .5-35)

where 'eis the surface density of free charges defined in ( 1 .4-36). Note that the last term
derives directly from the displacement tenn in ( 1 . 1 - 1 ) and is thus canceled out in the
quasistationary magnetic approximation. In any case, in a situation where there are no
currents crossing the boundary. there must be

Vs · I + � =0 ( 1 .5-36)

as we know from ( 1 .4-37). Whenever there are no surface currents or charges, or in any
case if this relation holds, it follows that

( 1 .5-37)

that is, the nonnaJ component jn remains steady when crossing the boundary. The
condition for the tangential component h follows with j = oR directly from ( 1 .5-2),
(1 .5-38)

that is

jt2 = JtI ( 1.5-39)

CT2 £Tt
With regard to the surface current density i = is used in ( 1.5-29,33) (that is, we
chose the unit vector s along i), we note that its existence, because of j � 00, also implies
£T --t 00 . Nevertheless, this concept is also useful when the current-carrying layer has finite
thickness L1h. provided that it is much smaller than the extension of the conductor,
&. « t1J . Wilh ( 1.5-29) and (A.3-le) we oblain
 1.5 BOUNOARY CONOITIONS 51

i
I .11
.. - -- --- --- - - - - --- -- -- -- -.....

Figure 1.5·4 Change of the tangential component of the magnetic field

H, = H� when crossing a thin curre nt-carrying layer.

( 1 .5-40)

and the boundary condition can be written in the simple form

lil = j = H, 2 - HIl . ( 1 . 5-4 1 )

where IIi = Hp = p . H is the projection of H on the tangent plane. along the unit vector p
Ihal is normal to s and thus to the current density, such that n= p x s (figure 1 .5-4),

Moving boundaries

If the boundary surface moves with velocity u = u(x. y. z) (with respect to the
laboratory frame (figure 1 .5-5)), then the introduction of the e1crtrlc field ( 1 .3-22) into
condition ( 1 .5-2) will result in the resfJrame boundary condition

( 1 .5-42)

With Ihe vector relation (A.3-3) and condition ( 1 .5-3) this becomes

( 1.5-43)

For example, i f the media move parallel to their boundary (that is, n · u :::. 0), the velocity
tenn vanishes (e.g.. on the surface of a cylindrical conductor rotating about its ax.is).
In the magnetoquasistationary approx.imation this i s the only boundary condition
affected by the movement of the boundary. By using the vector component notations
defined in figure 1.5-5, the last equation can be rewritten as
52 CHAPTER 1 FOUNDATION OF MAGNETIC FIELD THEORY

_ _ B

Figure 1.5-5 Boundary moving with. velocity u; here B and E are decomposed
into 'lle or1hogonal �yslem with unit vectors (p.s,n) chosen so that EIs "" n x E.

or, in analogy to ( 1 .5-18, 19), we can also say that

( 1 .5-44)

(as well as unBn) are steady when crossing the moving boundary. ( Bfp is the tangential
component parallel to Et whereas Bu is nonnal (0 it.)
I

If we consider, for completeness' sake, a moving dielectric medium to which the

velocity-dependent quasistalionary electric equations ( 1 .3-17. 19) apply. then the boundary
conditions (1 5- 1 37. 38) will change into
. ,

n X [("2 - "1)+ UX (02 - °1 )] � n X ("2 - "1) + (n · u)(02 - 0]) i,.� ( 1 .5-45)

n ' (h jl ) � (n' U)(Pe2 -Pel)

- • ( 1 .5-46)

(1 5-47 )
.

from which follow, like before. the conditions for the tangential or Dannal components of
the parameters involved, in particular those concerning the convective current term PeD,

which is typical for the electric quasistationary approximation.

In the following chapter we shall detennine the boundary conditions for other
electromagnetic quantities. namely. for the magnetic scalar and vector potentials, all of
which are summarized in table 1 .5-1.
 1.5 BOUNOARY CONOITIONS 53

Table 1.5·J Boundary conditions (magnetic: quaslstationary

approximation)Q

h Reference
Parameter"
Across Reference Across
Parameter
boundmy equation boundmy equation

Pitlds and currenls c

Potenlials

11';:
Rn = IJHn 0 1 .5·3 1"1> 0 2. 1 - 1 1

0 2. 1 - 7

ii<P
i,
'¥ 2.1-9

HI =: 0
7 i 1.5-\ A 2.2-23

On '" cEn 1. iMj 0 d

'. 1.5-4
Il ai; 2.2_24

.l .M, i, 2.2-24
P '"

E, = ; 0 1 .5·2 U 0
i. 0 1 .5-37 au 0
dx;
iJ.
• 0 1 .5·39
·f '.
from 1.5-16

,
Moving boundary

£. -u n� , 0 1.5-44

Ps
-.s", 0 1 .5-44
•
Assumptions: possible existence ofis. and SJ.z T- J.l2' E2 T- E, ; no existence of
surface electric polarization and surface magnetic moment.
b
Simplified notation with following meanings: first condition,
8n2 - Bnl = 0, or n .(82 - 8,) = 0; second condition. Ht - HIl = 0, or

,
See in chlipter 2: cp, magnetic scalar patenlial (section 2.1); A , magnetic vector
potential (section 2.2): U. eloctric scalar pc>(ential (section 2.2),
sirnuhaoeously Xj = n and Ai = As
d
Xj = s, p,n: A; = AJ, Ap'AII; bUl nOl «('"
definitions s.ee figure 2.2-4).
,
All other conditions remain unaffected by the velocity D, including those for V, A.
provided that they are in the Lorentl. or Coulomb gauge (see in section 2.2).
S4 CHAPTER 1 FOUNDATION OF MAGNETIC FIELO THEORY

The artwork here and at the end of each

chapter was done by Sergio Costa
JIj,1" "
during a course on Magnetic Tokamak ,1)1
,
Problems held in July 1985 at the Erice I,l..
(Sicily) Center of Scientific Culture:
Simple but elegant impressions from the
Meditemurean ambience (within which
the study ofmagnetism began more than
2600 years ago).
 MAGNETIC FIELDS: A Comprehensive Theoretical Treatise for Practical Use
Hein:.t: E. Knoepfel
Copyright Cl 2000 by John Wiley & Sons. Inc.

CfUlpter 2

MAGNETIC POTENTIALS

The basic problem tackled in this chapter is that of establishing an explicit relation between
static or quasistationary magnetic fields and the related current distributions, including the
effects of magnetized materials. The study of this relation'·h logically based on the solution
of Maxwell's equations written in magnetic and electric fields. which is often a fonnidable
mathematical task. A powerful alternative approach consists in introducing the magnetic
scalar potential cP or vector potential A , together with the electric potential U where
necessary. and in establishing their related differential equations. The introduction and
handling of these potentials is the main subject of this chapter, and it will be shown that
their application allows a more convenient description of magnetic phenomena in specific
circumstances. Several analytical solution methods are presented; then useful solutions are
provided that are related to magnetic fields produced by some geometrically simple,
currcnt-carrying conductors or magnetized materials; finally. the structure of magnetic
fields is discussed in general tenns.

Applications. The theory developed in this chapter gives the relations between magnet
structures and the related magnetic field configurations. Solenoidal, toroidal,
homogeneous, gradient. multipole. maximum. and variously shaped magnetic fields are
calculated from solenoidal,2.1l toroidal, variable-winding, race-track, Bitter-type,l.lJ and
other current-carrying coil structures and also from magnetized materials.s.u These results
are Ihen used in other chapters to analyze the connected energies, forces. and mechanical
stresses.

2.1 MAG N ETIC SCALAR POTENTI A L

A relation between magnetic fields and current distribution can be conveniently

established by introducing the magnetic scalar potential CP. The advantage of using this
56 CHAPTER 2 MAGNETIC POTENTIALS

scalar potential in the free space external to the current-carrying conductor, instead of
Ampere's vectorial equations, is that the magnetic problem can be solved by one differential
equation instead of a system of three simultaneous differential equations regarding the three
components of the magnetic field vector. In addition. the solution methods developed for
potential problems in other fields of physics (i.e., in electrostatics) can be conveniently
applied. Another use of the magnetic scalar potential. discussed at the end of this section.
is its application to describe the magnetic field pattern in and around a magnetized material.

Potential in empty space

Definitions and solutions

Just as in electrostatics, where the electric field in empty space E can be expressed

-VU
by a potential U through

E= , (2. 1 - 1 )

it follows. from the magnetostatic or quasistationary equations (1.2-1, 18) in empty space,
that the magnetic vector field H can be expressed as the gradient of a scalar potential

•
, ,
•

,
, \....... field lines:
I current : H=·Vtb
: loop ,
,
,
,
,

fl. _._. ._._ _. .....

,
\."

,
.'-T-"
\
'- eqUlpotentla
. . t s:
, , , ,
, , , I <P=consl.
, ..... ....
, ,
I
, ,
, ,
, ,
, ,
, ,
, •

Figure 1.1-1 Current circular loop with its dipole-type magnetic field lines
(which will be calculated in section 2.3 with the help of the magnetic vector
potenlial). The rotation surfaces (lines) of constant magnetic potential tP are nonnal
to the H lines.
 2.1 MAGNETIC SCALAR POTENTIAL 57

H = -V4> . (2.1-2)
In facl, taking the curl of this equation leads to V x H = 0 because of (A.3- 1 1 ), as also
obtained from ( 1 .2-1,18) in empty space wherej = O . Since also JlV , H = 0 from ( 1 . 1-3),
the magnetic field problem reduces to a potential problem described by lAp/ace's equation

64> = 0 , (2.1-3)

together with the appropriate boundary conditions (2.1-7. 9, 1 1). In general this gives rise
to a Dirichlel problem [(see in connection with (A.4-6»), where the boundary conditions are
defined by a given surface potential or field.
The advantages in using the magnetic potential cP and its related LapJace equation
instead of the amperian equation ( 1 .2-18) written in the magnetic field H have already
been mentioned in the introduction to this section. But what are the merits of using cP
instead of the magnetic vector potential A (which is introduced in section 2.2)? The answer
depends on many aspects of the specific problem, as we shall see in this chapter. 1be
vector potential is of more general usefulness because it applies to both current-canying
and empty space, whereas the scalar (X>tential is limited only to the latter. But if there are no
symmetries that eliminate one or two of the three vector components in A.
the one­
valued scalar potential can be more practical; it is also easier to visualize. because the
surfaces of constant C/J are nonnat to H with their spacing inversely proportional to IHI.
according to the wen-known properties of the gradient of a sca1ar potential (figure 2.1-1).
The scalar potential can be used only in simply connected, current-free space regions;
the properties of the currents that now beyond these regions can, in general, be taken care
of by appropriate boundary conditions. However, one must be careful to avoid a multiple­
valued potential function, and it often is tricky to relate 4> to the current density j. which
is present in most cases.

p(.)

Figure 2.1-2 Integralion palhs for delermining Ihe scalar polential lb.
58 CHAPTER 2 MAGNETtC POTENTtALS

For the potential defined by (2.1-3), !he gradient theorem

4>(r) - 4>(8) = jV4> dl

· (2.1-4)

applies, where the integration is from a reference point Pea) along a path Cj up to the
variable point P(r) (figure 2.1-2). With (2.1-2) this is

4>(r) = - jH · d l + D (2.1-5)
c;

since the potentials defining H can differ by a constant D. Some care must be taken in
choosing the correct path. In figure 2.1-2, Cl and Cl are acceptable paths, even if they
differ by a constant; in fact, with ( 1 .4-2) we get

which shows that the change in tP around the closed path C 1 - Cl is proportional to the
encircled constant current, even when j = 0 along the path. Path C2 should be avoided; it
can be used only if one is sure that the current flowing between it and any outside path
remains constant with changing position of Per).
Most often, however. the magnetic potential problem is conveniently solved through
the Laplace equation (2.1-3), which can be tackled with various methods, as we shall see in
this chapter (reference is also made to appendix A.4 and to the extensive literature on
mathematical physics).

Boundary conditions

Before dealing with some solutions of the magnetic potential problem, we shall
introduce !he boundary conditions on !he magnetic scalar potential 4>. These conditions
follow direcdy from those given in section 1.5. For example. noting that

il4>
n · V4> = ­ (2.1-6)
iln

from ( 1.5-3) we obtain

(2.1-7)

The orthogonal coordinate system defined by the unit vectors p,S,n is oriented so that (p,
s) are tangential to the boundary and, in particular, that p is in the direction of the vertical
projection of H = -V 4> (figure 2.1-3). Hence
 2.1 MAGNETIC SCALAR POTENTIAL 59

a<p
- a;;-

<P, •

p
s · '··Pt

Figure 2.1-3 Decomposition of the gradienl tb vector and definition of the

onhogonal coordinate system with unit vectors p,s,n, such that the plane (p,s) is
tangential to the boundary al point P with s along the surface current density is'
Also indicated are the potentials �.� at poi nts PI,P2 across the boundary.

d<l>
n x V<I> = s (2.1-8)
dp

---=ls
and from (1.5-29) we have

d<l>2 d<l>, .
--
(2.1-9)
dp dp

More directly. the scalar potential across a boundary can be expressed with (2.1-5) by

1'2<1>2 (P2 ) - 1',<1>, (P, ) = - J !Ill ' dl

P,

(2.1-10)
P,

Clearly, however. the integral vanishes as the two extremes of the path (figure 2 . 1 -3)
2
approach 'he boundary (P -> P, ); hence.

(2. 1 - 1 1)

These boundary conditions are somewhat similar to those concerning the electrostatic
potential U (see in table 1 .5-1), so the magnetic potential problem is similar to the
electrustatic one. (Note, however, that the derivatives along the nonnat and tangential
direc,ions are inverted: d<l>ldn = O. d<1>ldp = i, ; whereas dUldn = p,. dUldX; = 0).
Therefore. the many known solutions of the latter can be transcribed with appropriate care
into the magnetic case.
60 CHAPTER 2 MAGNETIC POTENTIALS

Axisymmetric systems

General equations

In axisymmetric systems there is rotational symmetry abom the z-axis. the current'"
are purely azimuthal.

(2.1-12)

and consequently the related magnetic field i s purely poloidal; that is, it has n o component
along e, [see also the discussion in connection with (2.2-27)]. In this case it is convenient
to resort to cylindrical coordinates (X,r): or in special cases, to spherical coordinates (P.O).
Many magnetic field configurations of practical importance have this symmetry, such as
those generated by currents flowing in loops or cylindrical coils (figure 2.1-4).
The Laplace equation (2,1-3) for the potential <P(z,r} in cylindrical coordinates is,
according 10 (A.4- 15),

(2. 1 - 1 3)

,
•

lip

Fllure 2.1-4 Poloidal magnetic field H(z. r);- H(p. O) generated by an

axisymmetric current distribution J = j_e_, with field components in cylindrical

(l,r,�) coordinates; the on-axis component i!> "o(z = p) = "l(Z,O) = Hp(p,O).
 2.1 MAGNETIC SCALAR POTENTIAL 61

and the field "(Z,r) (2.1-2) has the componenls (table A.3-U)

H _ _ alP H, = _ alP (2.1-14)

, - az ' ar

In spherical coordinates the corresponding expressions for tP(p.8) and R(p.8) are

ap
( )
l.. 2 alP + __l..
p
ap
I
sine ae
(
Si n e
alP =
ae
o
'
) (2.1-15)

P-
H _ _ alP Ho = -­
I alP
(2.1-16)
ap ' - p ae

The general solutions of these equations are given by (AA-2 t, 22) in appendix A.4.
A simple example in spherical coordinates will be discussed with the solution (2. 1-26), The
series expansion technique is applied below to present some properties of the magnetic field
and its potentia1 near the axis of symmetry . On the contrary, the fields far from the
generating current sources can be conveniently described by the multipole expansion,
which will be described starting from the vector potential form (2.2-42).

Expansion near the aXIs

For the scalar magnetic potential we set the polynomial series in r

-
lP(z,r) = "O(z) + a2 (z)r2 + .. . = �>2nr2n (2.1-17)
n=O
(uneven tenns in T vanish because of symmetry). and by introducing it into the two­

[ao(2) +4a 1 + [a(2) + 1 6

dimensional Laplace equation (2.1-13) obtain

2 2
..
Q4 1T2 + . = 0 •

where for the differentiation we use the notation a&2) = d2[ao<z)]/dz2. Because this
equation holds for any value of T, the expressions contained in each pair of brackets must
be zero, and by defining

(2. 1-18)

we find
62 CHAPTER 2 MAGNETIC POTENTIALS

"0 = t1b '

n
a (_ I)
_ (2n)
2n - n , 2 <It)
(2 . )

thus the series representation of the solution is

4>(Z,r) =
�
( ')
L- n! n!
n

4>(2n)
0
('-)2n
2
(2. 1-19)
n=O
By normalizing with the axial field on axis and in analogy to (2.1-18), we obtain

n
u(n )(
"0 l.) -
_d HO(z) _ ".,!n+l)
- -"'1)
dzn
(2.1-20)

(
The axisymmetric magnetic field components follow from (2. 1-14):

-,I ), H�2nlcz ' -

Hz(z,r) =
�

L
n { )2n
, (2.1-21)
2
n=O n.n.

( l)n+1 {'-)2n+1
�n!(n + l )! 0
�

H (z.r)
, = '" - H (2n+I)( z . (2. 1 -22)
2
n o

[Note that the substitution n - 1 � n in (2.1 -22) will have the series starting again at
n ;:;:
nature of
O.} These solutions, which are a direct and general consequence of the rotational
H. (Le. V . H = 0), provide the magnetic field around the symmetry axis, in a
.
volume not containing any current, by knowing just the component Ho(z) = Hz(z,O),
which can be found independently [see, e.g., in (2.3-7, 12, 19, 35)].
In first-order approximation (n = 0), from (2.1-22) we get

H,(,-, r )
=
_ '- dHz(z,O) (2. 1 -23)
2 dZ •

and differentiating (2.1-21) in the approximation n = I with respect to r yields

[d or ] = £H�2)(zO) = [dHzdZ(Z, O) ]
_2 HZ(zo , r) _ [dH/Z,O) ]
az z=.zo
, (2. 1-24)
r=E Z=Zo+E

where we have used the definition of a differential (substitute �

H H�l»
 2.1 MAGNETIC SCALAR POTENTIAL 63

oH(z) = Iim H(z + e) - H(l)

.
dz £-+0 £

The lauer approximation shows that around a point on axis z = la in axial symmetry
(where aH? ' dr vanishes), the increment of the gradient of the axial field component along
the axis is twice as large in absolute value as the one perpendicular to it.

Example: Diamagnetic sphere

An �xample conveniently treated in spherical coordinates (figure 2 . 1 -5) is that of a fully diamagnetic
sphere with radius a (e.g., a superconducting sphere that fully excludes magnetic field,>) immersed in a
unifomt outer field Ho (const.), whose spherical components are

Hop = Ho cos8, Hoe = -Hosin8 (2.1·25)

The general solution of the differential equation (2.1-15) is given in (A.4-2S),

<P(p,B) = I,[Allpll + BIIP-II-I ]PII (8) •

(2.1 -26)
11=0

where PII(B) are [he Legendre polynomials (AA-25l. For the radial field component (2.1-16) we calculate

(2.1·27)

The boundary conditions require the following: (a) At infinity. Hp(p -+ ....; any 9 ) - + Hocos6. which
implies that only the term AI :::: -Ho remains: and (b) on the surface of the diamagnetic sphere.
H pep = a; any 8) = 0; thus B J = -Hoa3f2 (the boundary conditions for He are then also automatically
fullfilled). Consequently. the potential is
l
� cos8
4'(p.6) = -Hopcos6 - HO
2p (2.1-28)

where the first term represents the potential

4'u = -Ho P
. (2. 1-29)

:
9
:
-'-
,
'

:
:H
- ,

Figure 2.1-5 Magnetic field around a diamagnetic sphere.

64 CHAPTER 2 MAGNETIC POTENTIALS

of the uniform magnetic field (2.1 ·25), and the second term represents the potential (2.2·52)

P
I p. p
<Po = -
4K -:3
(2.1 -301

of the magnetic dipole with moment

(2.1-311

The spherical componenlS of the poloidal field R(p.Blare calculated from (2.1.16) to be

(2.1-32)

On the equator, (p . Q,8 _ 1r 12), the magnetic field increases independently of the r-oidius a by a factor of
1 .5 : we can say that the diamagnetic sphere acts there as a flux concentrator. On the surface of the
superconducting (diamagnetic) sphere. there flows a surface cu�nt (1.5-1, 4 1 )

(2.1-33)

Ihat prevents the magnetic field from penetrating, and thus turns lefl-hand with respecl lO the vector "0'
When the elecuic conductivity is finite. the induced current penetrates into the sphere, and the
potential is modified. In chapter 4 we show that it is then convenient 10 reson 10 other solution methods; in
particular. in appendix AA we di!:cuss Ihi.� problem by providing the: appropriate equation (A.4-74) for i[�
solution.

Two-dimensional plane systems

General equa tions

When the currenlS are purely cuial, j = jl.el.' the scalar potential 4> and the related
magnetic field H = -V lP became two-dimensional functions that can be described, e . g . t

by Cartesian (x,y) or polar (r.�) coordinates (figure 2 . 1-6). The Laplace equalion and the
field components take the fann

(2.1-34.)

Hx = _aq,
ax
H
Y
d</>
= --

ay
(2. 1-34b)

or, respectively.

( 2 . 1 -35a)
 2.1 MAGNETIC SCALAR POTENTIAL 65

\
,

, ',
,
,/
,(
, ,
/ '
"11 :
- . ,
,
,
,
,
,

Figure 2.1-6 Two-dimensional magnetic field H(x. y) = H(r. �) generated by

the orthogonal current density j jzcz. with field components in Cartesian (x.y)
=

and cylindrical (polar) (r, Ij» -coordinates.

M I M)
H -__ ) H� = --­ (2 . 1 -35b)
r ar , a�
For the general solution of these equations we refer again to appendix 4 and. in
particular. to (A.4-14. 21). For example. the two-dimensional magnetic p:>lemial and field
on the outside (r :? a) of a very long cylindrical conductor carrying the Iota] current I within
[he radius a is calculated in Cartesian coordinates from (2.1 -34),

4J(X' Y) = 2/rc arctan � • r2 ; x2 + y2 , (2.1-36,)

(2. 1-36b)

or in cylindrical coordinates from (2.1-35),

cP(r,�) =-f.� (2.1-37')

( 2. 1 -37b)

(This is also obtained directly from Amphe's law ( 1 .4-3).) When the current flows
uniformingly in the cross-section, (r � a), the results inside the conductor are obtained by
the substitution I � Ir2/a2 in the above expressions.
66 CHAPTER 2 MAGNETIC POTENTIALS

Many solution methods developed for other fields (in particular electrostatic) can
conveniently be taken over for solving magnetic problems-for example, the methods of
images (see at the end of this section and in connection with (6.2-5 1)]2- 1 and confonnal
tran,fonnation (introduced below).

Example: Transverse field on diamagnetic rod

We briefly discuss a cylindrical problem that is directly solved by applying the solution (A.4·21): A
uniform magnetic field Ho is applied transversally to a perfectly conducting (i.e., fully diamagnetic)
cylinder. This arrangement, discussed in more detail in connection with figure 4.3-4. requires the following
boundary conditions: (a) al infinity, � -; any tP) -+ Ho costP; and (b) on the surface of the
H,(r
diamagnetk:: cylinder, H,(r = a; any tP) ;::; O . Using a procedure similar to that of the spherical solution
(2.1-28), from (A.4-21) we get

(2.1-38)

with

(2.1-39)

The conformal transformation method

For the analytical solution of the p()[ential problems defined by the two-dimensional
potential or Laplace equations (2. 1-34a,35a), it is onen convenient to apply the method
2 41
based on confonnal transformation with complex functions. . 1 . . (Complex variables and
functions are introduced and defined in section 3.1.)
The basic theorems here are that, firstly, any analytical function (3.1-20),

�l:) = u(x. y) + i v(x.y) • (2. 1-40)

of the complex variable

l: = x + i y (2. 1 - 4 1 )

and, secondly. the real functions u, v in the variables x, y satisfy the potential equation
(2.1-34a). In fact. for the complex function !!:. from
 2.1 MAGNETIC SCALAR POTENTIAL 67

a)!' a)!, a.
- = -- = -
dw a)!, a)!' a.
-= --=,-
. dw
(2. 1-42)
ax a. ax d. ay a. ay d;;

we calculate

and summation shows the first theorem,

(2. 1 -43)

whereas introducing (2.1 -40) into this equation gives

(2. 1-44)

and. since each bracket term must vanish, this shows the second theorem. Similarly,
introducing (2.-40) into (2.1-42) results in the so-called Cauchy-Riemann differential
equations
ou dv (2.1-45)
ax - oy
_

which represent a necessary and sufficient condition for .11:,.(,) to be an analytical function.
As later discussed in section 3 . 1 [see in connection with (3.1-20)), the complex
function w(z) transforms the (x,y)-plane into the (u,v)-plane. The interest here is that this
fonnalism makes it possible in many cases to transform a potential problem from the real
(x,y)-plane into the (u,v)-piane, where the analytical solution may be much easier for
geometrical reasons.u The final solution is lhen obtained by transfonning back inlo the
original (x.y)-plane. A typical example is when one wants to detennine the magnetic field
generated by currents flowing in a group of circular conductors, or conductors with
polynomial boundaries. with or without the presence of ferrornagnetic materials2.21• 1he
application of numerical techniques to confonnal transfonnation has added interesting
possibilities to this solution method.2- 1

Example: Parallel lines

An example where the confonnal transformation method can be conveniently applied is represented
by the parallel transmission line problem illustrated in figure 2 . 1 -7a. where two identical (superconducting)
wires of radius a, al distance d, each carry on their surface the opposite total current ± I. Due to
fieJd---currenl interaction {proximity effect}, the magnetic pattern is not obtained by simply adding the
undisturbed vector field deriving from each wire taken separately. The azimuthal magnetic field around one
68 CHAPTER 2 MAGNETIC POTENTIALS

'1 bl

i)'
./.--1-....
,<I+p
"

x "

d
�p� �-plnne

Figu� 2.1-7 (a) Parallel wire transmission line and (b) its conformal lransformation.

cylindrical conductor follows directly from Ampere's law (1 .4-3) and was given together with the magnetic
potential in (2.1 -36. 37).
The complex transformation function

(2.1 -461

provides aconfomal transformation of the space outside the two superconducting conductor.; (with radii a
and at distance d) in the ,-plane into the space between two concentric conductors in the lr.-plane with radii
as indicated in the figure. where

p=Jd42 _02 . (2.1-471

Such transformation functions. after some experience and familiarity with the confonnal method. can be
constructed with ease, or simply found tabulated in the literature for the most signifieanl geometrics.u
Here the b:-plane was chosen so as to make the form of the magnetic potential very simple,

lP(w)
- - = i �lnw .
21t -
(2.1·481

In fact, if we consider the polar form (3. 1-12) �= peit • we can write in the 11:: -plane

(2.1.491

where in the real term we find the potential (2.1-37a). Transfonnalion into th.e z.-plane with (2.1-46) yields

For the magnetic field, which is the gradient of the potential, (2.1-2), in complex writing we get, by
applying the Cauchy-Riemann equations (2.1-45),
 2.1 MAGNETIC SCALAR POTENTIAL 69

(2.1·51)

that is. we do nOI need 10 delennine the real part of !1! and calculate the gradient, but we can differentiate
directly the complex potential with respect (0 Z. and obtain [laking again the conjugate (3. )· 14)1

(2. 1-52)

The magnetic field in the ,-plane outside the (superconducting) conductors thus has the componems

(2.1-53)

(2. 1-54)

The magnitude of the field becomes

(2.1-55)

On the conductor surface. described by the equation

(..r± �dr _a2 _y2 (2.1-56)

we have
1
IH 1 = - pi .
_6
(2.1-57)
2n a.x

which also means that the linear current density on the conductor surface flowing parallel 10 the axis. where

!!.. - aSx -.; !!.. + a. has the value ( 1 .5-4 1); that is.
2 2
j =. 1 pi
=_
l H (2.1·58)
J 2n ax

The magnetic field between the two wires is perpendicular to the x-axis, and its magnitude becomes. from
(2.1-54) (or y ::: O,

I P
H =--1
1t p 2 . x2
-- (2.1 -59)
J

whereas from (2.1·37) and ignoring th� proximity effect. w� obuain the smaller value

(2.) -60)

For Ihe increase of th� magnetic field berween (he IWO conductors there is a corresponding cumnl density
concentration on the conductor sides facing each other. Consequently, the apparent resiSlance and inductance
vary. as will be shown by (2.4·1 S) for the inductance.
70 CHAPTER 2 MAGNETIC POTENTIAlS

Potential in and outside magnetized materials

'The magnetic
scalar potential concept can be useful in describing the magnetic field in
and around magnetized malerial. To illustrate this problem, let us consider the case of a
permanenl (bul space dependent) magnetization M(r), such as represented by hard
ferromagnets where the magnetization is pennanent and independent of any applied
magnetic field (see in chapter 8). Taking the divergence of (2.1-2),
H = -V4> , (2. 1-61)

and, because of the definition ( 1 . 1-17), V · 8/).1 '" V , ( H + M ) '" a.results In the Poisson
0
equation
64> = V · M , (2.1-62)

and from ( 1 . 1-27) we can introduce the magnetic charge density

Pm = -V · M. (2.1-63)

With this fannulation the general solution for tP at point P of the inhomogeneous
differential equation can be found by various methods, for example it can be obtained
similarly 10 the electrostatic potential (2.2-20):

(2. 1-64 )

The last expression is obtained by applying Gauss' theorem (A.3-16); it is important when
M is uniform, because then in the whole volume V · M = 0, and only the surface gives a
contribution through the magnetic surface charge density introduced in ( 1. 1 -29):
rm = n · M . (2.1-65)

Example: Permanent spherical magnet

We iIIuslralt Ihe magnelic polenlial method with a permanent splltrical magnet with fixed unifonn
magnelizalion M = Mo (const.) and radius a (figure 2.I-S). We now have

Fm = Mcos 8 , (2.1-66)

and the solulion integral (2. 1-64) is

2
«P(p,S)
= Ma j"'d 8'j211 cosS'sin8' d4l' (2.1-67)
4x 0 0 PPQ
 2.1 MAGNETIC SCALAR POTENTIAL 71

,) b) c)

t
p

,
,
,
,
,

-�-_;--,,_
<� .. �� l ' I
'. '. , o·
..

Flllure 2 . 1 · 8 The permanently magnetized sphere: (a) geometry, (b)

approximate paltem of magnetic induction B (left), and (c) magnetic field H (right;
note opposite internal field Hi ).

PPQ = (p,) 2 + p2 _ 2pp'[cos(IP' _ ;)si09'5i09 +C050'cas 8] .

Integration yields" I)

11",<0,'
".

<P(p. 8) =

..
1
(2.)·68)
-"":1"('01
I
6. p'
.
1 ,

The Jailer solution represents the potential of a magnetic dipole (2.1·30) with moment

4
K 1
p = -a M (2.) ·69)
3

whereas the focmer corresponds 10 � potential of a uniform field H = -M/3, (2.1·29), where the

spherical components of M are

Mp = MCQs9, M6 =-Msin9 (2.) ·70)

TIle internal and extemal magnetic field components (2.1·16) arc thus

.
) i '
Hp
' - -- M cos9. H8 = '3 M Sin
' 11. pSa
3
(2.).7)

p�a .
72 CHAPTER 2 MAGNETIC POTENTIALS

Note that according 10 the definitions ( 1 . 1 ·37), the demagnetizing field here is Hd = H = - 1/3M
i

and the demagnetiz.ing faclor km = 1f3: the internal field is antiparallel 10 M and 10 the z-axis. The

induction

(2. 1 -72 )

in the sphere has the components

.
Bp "'3 J10M cos 8
2
= (2.1-73)

it is uniform and parallel to the z-axis (and opposite to the magnetic field), whereas outside it is
Bt = JloUt•

Example: Magnetized sphere

In the second. more general example described in figure 2.1·9, a unifonn outer field Ho. derivable

from a potential

4tu = -HopeD! 9 (2.1-74)

and with components

HOp = Ho cos8 . (2.1-75)

induces in the sphere the magnetization

(2. 1-16)

where Hi is the internal. as yet unknown. field. although it is plausible 10 assume (ha! it is parallel 10 "u'

We can proceed formally. as before. because M remains uniform and constant, and the field thus results
from lhe superposilion of solution (2.1-71) on Ihe oUler field (2.1-75), but we need an equatiun Ihat relates
M or H 10 Ho. This equalion is provided by the boundary condition ( 1 .5-18) requiring the normal (i.e.
i

radial) component of the magnetic induction to remai n steady at the boundary p "" a , that is,

(2.1-77)

JJlIo cosB - '3I JJM cos8 = J1oHo cos6+ ')2 JioM cos6

from which follows

M = 3aHO with (2. 1 -78)

or, from (2.1 -76),

 2.1 MAGNETIC SCALAR POTENTIAL 73

a) b)

: " "':
I
. .

JIi " M

.,
" ,
I'

Figure 1.1-9 Paramagnetic sphere ( PR > I) in uniform magnetic field flu

with approximate pattern of (a) magnetic induction B (Ier!) and (b) magnetic field H
(right; note reduced internal field Hi ).

.
HI = -- Ho = ( I -a)Ho
3
(2.1 -79)
'uR + 2

The second boundary condition, requiring the tangential 8-componenl of the magnetic field (2.1·7 1 ) to
remain steady. is aUlOmalically satisfied; also note that if the medium surrounding the sphere is not empty
space bur has a J.iK2 � 1. then the above expressions change according to the substitution iJR 4 JiR I,uR2 '
The solution. resulting from the superposition of the fields (2.1-71. 75). with a given in (2.1 -78), is

H� "" Ho( l - a)cos8. H� = -Ho( l - a)sin8 for p�a . (2.1-80)

[
H� = -HO I - a(aI P)3 }inO for p i::: a (2.1·81 )

According 10 the definitions ( 1 . 1 -37) the demagnetizing field here: is Hd "" -aIIO and the demagnetizing
factor iskm '" a. Similarly. the: total pOle:ntial he:re: is obtained by supc:rposition of the potentials (2.1-68.
74). For a fully diamagnetic sphere (i.e ,uR --+ 0 and a --+ - 1 / 2 ). we: find that the field H( p.B) and the
.•

dipole: moment p are as in (2.1-31. 32). In a pararnagnetic material with,uR, > I. the internal field Hi is
always smallerthan Ho . but
--�
Oi _
3,uD H0 (2.1-82)
,uo PR, + 2

is always larger. and for ,uR very large (ferromagnetic materials) it approaches 3Ho; because of the

boundary condition (2.1 -77) this Bi/jJg '" IJ.RHi '" Ht! is also the field just outside the poles.
74 CHAPTER 2 MAGNETIC POTENTlAlS

Wc know from (1.1-22) that the effect of constant magnetization can also be described by the
equivalent surface current density

(2.1-83)

and here the only component becomes

im,8 = -M sin6 = -3aHo sinB (2.1-84)

Alternatively, from (1.1-29) we know that the field pattern can also be related 10 the surface magnetic charge
density (2.1-66),

'm = Mcos 8 = 3aHo cos 0 . (2.1-85)

Method of images
The method of images is a powerful easy-ta-handle solution technique that is used to
detennine magnetic potentials (or fields) when plane or circular boundaries (or of other
simple geometries) are presen{7. Consider, as an example, the arrangement of a line
current I (the field source S) parallel to a plane boundary between (Wo magnetic media with
different penneabilities J1.P 112 as shown in figure 2. 1-lOa. The problem is to find the
magnetic field H or inductions B in the two media. The solution method is quite
straightforward and consists in finding the magnetic potential tP through the Laplacc
equation (2.1-34) and its boundary conditions (2.1-7,9), which in their field formulation
( 1.5-18) require the tangential field and the normal inductions (2.1-36b).

(2.1-86a)

B,(x.z) = J1lI, = -f. -'r . (2. 1-86b)

,

to remain steady across Ihe boundary. (Through lhis formulalion of the problem here we
are taking advantage of the already established general solution (2.1-36a); but note the
difference in the coordinates.]
The result can be expressed fonnally by the image method, : 1 which states that the
magnetic potential tP (or field) in the /J..-region is obtained by the summation of the
potential 4>. from the source with current I plus the potential tP from the image source S'
2
with current I (j,l.,- 11,)/(j,l.,+ 11,); whereas in the 11, -region the potential <1>, (or field) is
due solely to the potential from the original source S, but with current 21/J./(J.l2 + /J..). as
shown in figure 2. 1 - lOb,c. This statement can be logically extended by superposition to
any closed circuit, and Ihus has general validity. It is motivated by the fact that both
potentials (or fields) of the general fonn (2. 1-36) are each solutions of the Laplace equation
 2.1 MAGNETIC SCALAR POTENTIAL 75

b) , c)

2,u,
I I
source S Il'}. +JJ ,
•
n-.edium JJ
I ,

medium
P2 P,

Image S'

Figure 2.1-10 Method of images for determining the magnetic field of a line
current I (flowing in the positive y-dircclion) in magnetic media with two magnetic
permeabilities. (a) The geometry of the problem with qualitative magnetic field lines
for the case where Jil = 4Jl j; the image CUn-COLS are given as required for calculating
(b) the field in the Ill-region and (c) the field in the JJ? -region.

(2.1-34) satisfying the boundary conditions, as can be easily checked. In fact, in the J.LI­
region of the boundary (z = 0) the tangential field is

(2.1-870)

whereas the nOllTlal induction is

(2.1-87b)

and the same is obtained in the ,LLz-region. Note that with these values we get the refraction
of the magnetic field lines at the boundary as defined in ( 1 .5·22).
The analogous case of a Hne currem in free space and parallel to a superconducting
plane follows formally from the above results by setting 112 = 0, as we know from the
comments made in connection with (1 .5·22). The field outside the superconductor results
then from the vector summation of the fields from source and image with currents +1 and
-I, respectively, whereas there is no field within the superconductor (both the
superconductor and the 112 = 0 medium are fully diamagnetic; see also discussion in
connection with figure 6.2· 1 1 ). For the tangential field on the superconducting boundary.
from (2.1-870) follows

(2.1-88)

whereas the surface current density from ( 1.5-41) is

76 CHAPTER 2 MAGNETIC POTENTIALS

(2.1-89)

that is, it falls off rapidly away from the projection point x=O (the total current flowing on
the surface is obtained by summation and amounts to -I). In this ideal system of a line
conductor parallel to the plane superconducting return conductor we find the current density
concentration typical for the so-called proximity effect. We have calculated this effect
already for the parallel transmission lines shown in figure 2.1-7.
When the conductors have finite electrical conductivity (7, the problem becomes more
complicated. m relation to a line current parallel to a conducting plane, we find in (5.4-32b)
the approximate mean heating power deposited in the plane by an oscillating line cUlTent;
and we find in (6.2-62) the drag and suspension forces on the line current moving nonnally
over the plane.

2.2 MAGNETIC VECTOR POTENTIAL

To deal with quasistationary magnetic problems, it is often convenient to introduce a

magnetic vector potential A as done in this section. This vector potential is generally
useful because it applies both to current--carrying and to empty space (whereas the scalar
potential introduced in the previous section is limited only to the latter) and is particularly
convenient when the symmetry allows one to eliminate two of its three vector components.
as for two-dimensional plane or axisymmetric magnetic field problems.

Basic equations and solutions

Equations
For the solution of the quasistationary magnetic field equation
'\7 x H = j (2.2- 1 )

within a space where the current density is not everywhere zero, it is often convenient to
introduce the vector potential A. defined through

B = '\7 x A ; (2.2-2)
 2.2 MAGNETIC VECTOR POTENTIAL 77

this is in agreement with ( 1 . 1 -3), which requires B = pH to be the curl of some vector
field; that is, the divergence of B vanishes according to (A.3-7). One of the physical
meanings of A is related to the magnetic flux IjI across any arbitrary curved surface
S bounded by a contour line C (figure 2.2-1); in fact, applying Stokes' theorem, (A.3-l4),
yields
(2.2-3)

( )
From Faraday's law ( 1 . 1 -2) through insertion of (2.2-2) we get

V X E + dA = D '
dt

because the curl vanishes, the vector in parentheses can be derived from a scalar potential
U defined by (A.3- 1 1 ),
-VU = E + �� (2.2-4.)

This potential U can be idenlified as an electric scalar potential due to electrostatic charges
or to external yohage sources. The total current density can thus include two componems,

( 2. 2-4b)

where

j, = -a'I' U (2.2-4c)

is the onc given by external electric sources, whereas

(2.2-4d)

is the inductive component generated by a time-dependent magnetic vector potential or

field.

11

ds=nds

surface S

dl
contour C

Figure 2.2-1 Magnetic flul( and vector potential.

78 CHAPTER 2 MAGNETIC POTENTIALS

We shift now briefly our attention to the general fannulation. which takes care of the
displacement current term in (1.1-1). Taking the divergence of (2.2-4a) and using (1. 1-4)

p,
(thereby assuming /I, E to be constants) yields

__
/'J,U - E
_ <lW-A)
at '. (2.2-5)

M = -/lj+ '1('1· A)+ IlEq+ /lEil�U

then inserting (2.2-2,4a) into Maxwell's equation ( 1 . 1-1) gives

(2.2-6)

= V . V has different meanings, according to whether it

at ot

where the Laplace operator Il.

operates on a scalar or on a vector [see in connection with (A.3-1O. 23)]:

f!.U=V·(VU) , (2.2-7)

M= (V · V)A '1('1 ·A)- VxV xA

= (2.2-8)
According to the definition of A in (2.2-2) and of the vector relation (A.3-I I), it is possible
to add a vector A', A A + A', without changing the value of the magnetic induction B .
V (plus a constant vector), A' = VV. This freedom to introduce a so-called gauge
�

provided that V x A I =0 -that is, provided that A is the gradient of some scalar function
I

rransfarmarian can be used to simplify (2.2-5,6) by adjusting V· A : The Lorenr. gauge

V · A -1lEi!f,-
= (2.2-9)

-/lEil'U &.
leads to the symmetric d'AJemben equations

M-/lEU = -/lj
= _
at2 £
l1U
(2.2-10)
ill'

[For the solution of these equations see (2.2-26).J

In the quasistationary magnetic approximation, where the Ja'it two tenns in (2.2-6)
vanish because they derive from the displacement tenn aD!at, it is convenient to use the
Coulomb gauge
V·A=Q (2.2- 1 1 )
to obtain

M = -J1.i
(2.2- 12)

(2.2- 13)
It is useful to point ou[ the significance of this fonnalism: The system of equations (2.2-10)
[or of (2.2-12, 13) in the quasistationary approximation] can substitute Maxwell's
 2.2 MAGNETIC VECTOR POTENTIAL 79

equations ( 1 . 1-1 to 4). Once the potentials A, U are determined, the electric and magnetic
fields are found through (2.2-2. 4).

pH = V x A . (2.2-14)

E=- ¥,- - VU (2.2- 1 5 )

Which o f these two options (i.e., the potential or the Maxwell equations) is more
convenient for the solution depends on the specific electromagnetic problem to be tackled,
as we shall see from some examples later on.
In conclusion we note that in the stationary case, the potential A follows from the
given stationary current density j = js

M = -!l.i . (2 .2- 16)

[As we will see in the following problems of this section, J5 is taken as given, without
bothering about its origin (2.2-4c),] In the quasist81ionary case, the current density (2.2-
4b), j = js + jj, includes in addition the inductive component (2.2-4<1),

such that
aA .
U
AA - IL " - = -ILls (2.2-17)
al
[i.e., we obtain the diffusion equation (4. 1 - 1 3), which will be discussed in chapter 4].

Solu tions
The stationary or quasistationary solutions of the above differential equations in the
pOlentials U. A follow standard rules. In particular. Ihe general solution of (2.2-13) must
be looked for separately in the spaces with and without currents; in the latter case the
vectorial LAplace·type equation

6A = O (2.2-18)

has to be solved. Finally, the two subsolutions mUSI be matched at the boundaries.
For the solution of the voc:torial equations presented so far [e.g., (2.2-13, 18)], it is
necessary to decompose them into equations for the components of the unknown vector
2
field. This requires correctly interpreting the Laplacian operator, 6 = V . V = V • as
pointed out in (2.2-7.8). The result is straighlforward (A.3-23) in Canesian coordinales
with A == (Ax ' Ay , Az). For example, the vectorial equation (2.2-13) decomposes into three
nonnal Poisson equations
80 CHAPTER 2 MAGNETIC POTENTlALS

Mj :: -j.lji • 1 = X.},.l , (2.2-19)

where the Laplacian 6 is appl ied here to the scalar component-function Ai and has the
well-known form given, for example, in table A.3-II. In other coordinate systems, the
decomposition of the vectorial Laplace or Poisson equation into component equations is
more complicated, as shown, for example, by the Laplacian (A.3-23b) in cylindrical
coordinates.
On the other hand, as is well known from mathematical physics, and in particular
from electrostatic problems, the solutions of (2.2- 12, 13) in an unbounded space can also

fP'Q
be expressed directly by integral formulae AI7: The electric potential becomes

I
4.ra: FPQ •
Up = - -dV,Q (2.2-20)

and the vector potential yields

(2.2-2 1 )

from which the magnetic field

(2.2-22)

is obtained in the quasistationary approximation; here P denotes the point in which the
magnetic field is to be established, and Q the variable integration point within the conductor
(figure 2.2·2). The last expression is called the Biot-Savart law. The Iransformation from
(2.2�21) to (2.2-22) is obtained using the vector relation

v x (� )=�VXiQ + (V �)X iQ
V xjQ = O. since the V operator is related only to point P. From (2.2·2 1)
V· A
and noting that
we easily find that the condition = 0 requires VQ·jQ = 0: (2.2·22) is therefore

iQ +
strictly correct only for closed conductors or conductors extending to infinity. The general
time-dependent fannulation of (2.2-2 1 . 22) has the substitution iQ-> iJD/at in the
integrals. which requires the calculation of the displacement current term; therefore, the
formulation given in (2.2-26b) is to be preferred.

"JQ
Q

Figure 2.2-2 Spatial current distribution.

 2.2 MAGNETIC VECTOR POTENTIAL 81

In figure 2.2-3 the relations between the current density j. the magnetic flux B , and
the scalar and vector potentials CI>,A are represented schematically. The magnetic vector
potential is very attractive for two-dimensional plane problems or axisynunetric magnetic
field problems (where (he CUITem is purely axial j i,e, or azimuthal j i.e.) because it
� �

then has only one component (A�. or A rt» . which is orthogonal to the plane containing the
magnetic field vector, as we shaH see. For three-dimensional problems, an three
components of the vector potential must in general be computed [from a system of three
coupled, paniai differential equations derived from the vectorial Poisson equation
(2.2-13)], and thus the vector potential becomes less useful than the magnetic scalar
potential (whenever the latter exists)_ The vector potential has the merit, in any case, of
being applicable both in free space and within current--carrying conductors and avoiding the
multiple·valued problem of the scalar potential.

pj=- M A �
1
4.
fD" -;er dV
(2.2- 13)
(equiv. 2.2-22)

A=
.1!.....
4.
f1 ,
dY
B ='VxA
(2.2-2)

4� J j;:
(2.2-21 )

H= dY (2.2-22)
j ='VxH (1.2-1)
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, ,
, , H = ·'VtP
, ,
,
(2.1·21
,
, ,
,
,
,
,
,
,
,

Figure 2.2-3 The magnelostalic equations relating currenl density j. magnetic

induction B = p H , scalar and vector potentials cP, A , with the equations shown
(with reference to their position in
the chapter); the polenlials can be convenient
intermediate functions for simplifying the solution IcP is defined in me space, so
there is no general meaningful relation to j in the conductor].
82 CHAPTER 2 MAGNETIC POTENTIALS

In discussing the general solution of the vector potential problem, we note that the
integral form (2.2-21 ) is the most useful to start from because it is valid in spaces with or
without currents, and the integration extends only over the conductor. An imponant
advantage of solving magnetic field problems through the vector potential fonnalism is that
this integral has a simple structure and is relatively easy to handle. As with the magnetic
scalar potential. the integral can be solved with various approximations-for example, with
the multipole expansion technique, valid far from the current source, which we present in
(2.2-42, 50).

Boundary conditions
The boundary conditions for the magnetic vector potential A can be derived from the
conditions on H and B given in section 1.5 but also depend upon the gauge chosen. From
the Coulomb gauge (2.2-1 1 ), V · A = 0, it follows [as in connection with (\ .5-23)] that
the normal component An remains steady across the boundary [this is also true for the
Lorentz gauge (2.2-9)}. On the other hand, the same property follows from definition (2.2-
2) for the tangential component Ap; in facl, consider (2.2-3) applied to a loop as depicted in
figure 2.2-4, and then let the loop shrink. dh�O. Thus, the vector potential remains steady
across the boundary,

(2.2-23 )

Boundary condition ( 1.5-6) also sets Hmits on the derivatives of the vector components of
A . It is convenient to decompose the vector along the (Pt S , n)-directions (figure 2.2-4),
A ;: (A p .As.A.n l: and from condition (1 .5-5) with B = V x A, and after some vector
calculus, all partial derivatives along the (Pt S, n)-directions of all components
(Ap, A s. An) are steady across the boundary. with the possible exception of

(2.2-24)

where the surface current density i = sis defines the s-direclion.

In addition. the current condition ( 1.5-37) with expressions (2. 1 -6), (2.2- 15) can be
rewriuen in [he fonn

(2.2-25)

Note that in magnetostatics (time-independent magnetic field and vector potential) this
condition reduces to one for the electric pocential.

which is similar to condition (2.1-7) for the magnetic potential.

 2.2 MAGNETIC VECTOR POTENTIAL 83

A,

p S . '"
,
I ,'"
--- "
--

P
_ _
A

Figure 2.2-4 Decomposition of the vector potential A . (A p . A s . A n ) am

definition of the orthogonal cOO!'"dinate system with unit vectors p . s. n. such that
the plane (P. s) is langenlial lo the boundary at point P with A, along the surface
current density is. Also indicated is a rectangular loop cuning nonnally the
boundary.

Retarded potentia/s
We return to the d'Alembel1 equations (2.2-10) and recall that they can be used
instead of the general Maxwell equations to detemtine the electromagnetic polentials and
their propagation as generated by the source terms Pe. j; the electromagnetic fields then
follow with the help of (2.2-14, 15). In free space these equations reduce to the familiar
fonn of the wave equation (4.1 -6). On the other hand, we have seen that in the
magnetoquasistationary approximation (where any electromagnetic disturbance is
transmitted instantaneously) they reduce to the Poisson equations (2.2- 12, 13) with the
solutions given in (2.2·20, 21).
What are the solutions of the d'Alembert equations? We note that the effect of any
change of the source tenns Pe j here is not transmitted instantaneously to a remote point;
instead, the corresponding disturbance will propagate outwards with the light velocity
c =- 1/ "£J.i. ( 1 .2- 1 1). Consequently. it is plausible that the solutions of the dynamic
equations (2.2-lO) have the same form (2.2-20) as for the static case but are retarded in
time at distance r from the source point by t -+ t - rlc. so PeQ . jQ are functions not of (t)
but of (t -rPf/c); hence, in unbounded space, we obtain the solutions of (2.2-1 0) represent­
ed by the so-called retarded potentials:

(2.2·260)
84 CHAPTER 2 MAGNETIC POTENTIALS

By taking the curl of the last equation, Le. proceeding as for the transition �.2-21 to 22),
we obtain the generalized Biot-Savart law (valid for any time dependence):2.

(2.2-26b)

here for simplicity we indicate the retardation through a square bracket,

[jQ1 = j(t - 'I'Q I c), and have used (A.3-4) and Ihe relation

Axisymmetric systems

In a cylindrical vector-potential problem. the cylindrical field components of

JJH = V X A are deduced from table A.3-Il:

_ _ 1- aA.
H, - J1 al. '
(2.2-27)

When the given. stationary currents are azimuthal or toeoidal, j = j�etP [i.e., flowing in
circular paths perpendicular to the z-axis and centered on it (see figure 2.1-4 or 2.4-5)1. we
know from the rule (A.3-3 1) established in connection with the curl of a vector. j = V x H.
Ihat Ihe field is purely poloidal, meaning H� = O. Similarly, from p,R = A x A it follows
that the vector potential is azimuthal. This is also consistent with the expression for the H�­
component in (2.2-27), which shows for H9 = 0 that Ar can depend, if at all, only on r,
and Az. only on z: and since Ar• Az do not appear in the other field components of (2.2-
27), it can be safely assumed that Az = Ar = O. In conclusion. only the A4I -component is
relevant in axisymmetric geomeuy.
By taking the Ikomponent of (A.3-24b), or by expressing the equation
(V x H)� = j� with the help of table A.3-Jl, we obtain

' '
a A;
+ 1 aA, _ � + a A. = -JJj. (2.2-28)
ar2 r dr r2 az2 Y'

The general solution of this Poisson equation referred to a space containing axisymmetric
currents is obtained as usua1: First, the solution of the homogeneous (j41 = 0) equation
must be found (it is given in general form in (A.4-30)]; then, a particular solution of the full
equation (2.2-28) must be added; finally. the solutions must be matched at the boundaries
(including the condition � 0 for , O.
= ) and for the initia1 conditions. Other solution
= 00
 2.2 MAGNETIC VECTOR POTENTIAL 85

methods exist, such as the interesting double series methoo presented and discussed in
(A.4-4I). NOle Ihal (2.2-28) in free space differs fonnally by a tenn -AiJr2 from lhe
analogous (2. 1-13) referred to the scalar magnetic potential l/J. When solving by numerical
computation, however. both the � or <I> approaches require roughly the same effort.
The series expansion around the axis of symmetry can be appHed as a solution
method to give � in a space not containing any current. From (2.1-21 or 22) and (2.2-27)

(-1)" (2nl( , )2n+1

we gel

� (l,r) /1 �
':"' .!("+I)! HO
� "2 ' (2.2-29)
n;;;;;:D
where the differential of the axial magnetic field denoted with the symbol Hb2n), which is a
function of 2:, is defined in (2.1-20); near the axis, a valid approximation is obtained by
considering the n = 0 term only.

(2.2-30)

Finally, there is the integral solution (2.2-21), which is particularly useful in

axisymmetnc geometry. [t will be applied in the next section, starting with the solution
(2.3-1) for calculating the field produced by circular coils.
In concl�ding this discussion on axisymmetric systems, we reintroduce the poloidaI
magnetic nux function 1jI(l..r), here taken across any surface S bounded by a concentric
circle with radius r (see in figure 2.5-4). Definition (2.2-3) leads to

If/(l, r) � 2"r� , (2.2-3 1)

with which the differential equation (2.2-28) can be written in the form

2 0 2
a ", a
1� + = � -2"WJ'
•

Of'
_ _ (2.2-32)
0,2 r 0' Oz.2

as easily checked. The magnetic field components (2.2-27) become

(2.2-33)

,e + Hz.ez. = 2�r er; x V II'

they can be expressed by the poloidal field defined as

Hp = H , (2.2-34)

where ez.,e" el/l are the orthogonal unit vectors in cylindrical geometry [the last expression
can be checked by insening e�,e"e, and applying the triveclOraJ relation (A.3-3). then

taking the vector components of V If/ from table A.3-Il]. We shall take up the lasl
expression again in section 2.5 and show in (2.5-12) that the 1jI = consi. lines are the
Hp-field lines.
86 CHAPTER 2 MAGNETIC POTENTIALS

Plane two-dimensional systems

We consider here the two-dimensional geometry where the given. stationary current
density is parallel la the z-axis, j = j,ez (figure 2. 1-6), and hence the magnetic vector
potentia1 has the Aceomponent only. [As with the axisymmelric system. this is obtained by
applying the rule established in connection with (A.3-31) to j = V x H and H = VxA.I
This problem, where A is translationally invariant in z, can be handled, for example, in
Canesian coordinates with Az.(x,y), or in cylindricaJ polar coordinates with Az(r.t,lI}, the
choice being dictated by geometric convenience. Contrary to what we found in
axisymmetric geometry, the Az-component of the vector potential and the scalar potential
<1> both obey formally identical differential equations.
The Poisson equation (2.2- 13), which here includes the axial component of (A.3-

1r
24b) only. reduces in polar coordinates to

a'A' + l a ' + ...l.. a2A, = _ .

or2 ,. ,.2 a�2 J1h. (2.2-35)

and the magnetic field components are

H� = - -'- �
p a,
(2.2-36)
.

The homogeneous form of (2.2-35) is identical to (A.4-l5) with the last term neglected, so
the homogeneous solution Az(r,q') can be transcribed from (A.4-21). [f there is also axial
synunetry in jz(r,l/J) = ft.(r), Hr = 0, the result is a toroidal field
which implies
HT = H�e�. As usual, the complete solution of (2.2-35) requires a particular solution to
be added to the homogeneous solution.
It is often convenient to use Cartesian coordinates (x.y) in solving for Az(x,y) in
plane problems. This case is discussed in section A.4. A simple Fourier series solution of
the related Laplacian equation (A.4-27) is given in (A.4-28). Of particular interest,
however, is the double Fouriee series solution (AA-3l) because it directly gives the
solution of the Poisson equation (A.4-27). Finally. there is the integrnl solution (2.2-21 ) ,
which i s particularly interesting in two-dimensional geometry [see the example in
connection with (2.3-41) at the end of the next section].

Biot-Savart law

Filamentary conductor
In filamentary cond�ctors (wires), the linear dimension of the cross section is much
smaller than all the other dimensions, such that approximately
 2.2 MAGNETIC VECTOR POTENTIAL 87

'PQ p

Figure 2.2-5 Biot-Savart law.

jdV", jsdl '" Idl (2.2-37)

where I ;:.js is the total current flowing in the conductor, with s the cross-sectional area,

= ..1.. 1j
as shown in figure 2.2-5. Solution (2.2-22) thus transforms into Biol Savart's law:
-
dlx'PQ
Hp 41t 3
(2.2-38)
'PQ

�I 1..!!L
whereas for the vector potential (2,2-21) we have

=
AP 4ft j rl'Q (2.2-39)

In the next section we shall use this integral form to find directly the so1ution of different
filamentary CUlTent systems. For the general time-dependem formulation of these two
integrals, see comments made in connection with (2.2-22. 26a).

Multipole expansIOn
.
When the magnetic field or vector potential has to be known only at large distances
from the current source, then the so-called multipole expansion approximntion can be
useful.A.LJ With reference to figure 2.2-6 it is

.fo = p2 + a2 - 2pacosa = ( ;:
p2 1 + _ 2 )= p2 (1
�cosa + c) . (2.2-40)

and by expanding into a Taylor series with respect to £ = (ti2Ip2}-(2a cosaJp) we obtain

(2.2-41 a)
88 CHAPTER 2 MAGNETIC POTENTIALS

p
'PQ

Figure 2.2-6 Geometry for the multipole expansion of the magnetic vector
potential at point P generated by a cum:nt loop.

a) are fhe Legendre polynomials, A . I the first five of which are given in
where P" :::: Pn(cos
(A.4-2S). This could have been developed directly in vector form6_2 for IPQ = p - a :

--L= .l [ I+£:!+3<p.a>2 -e2a2 + ]

'Po P p
' 2p'
.. .
(2.2-41 b)

(nsertion into (2.2-39) yields

(2.2-42)

here we have neglected the n :::: 0 (magnetic monopole) term because the total vector

f
displacement around a closed loop is a1ways zero:

dl = 0 (2.2-43)

The remaining n = 1.2 •. ,. terms correspond to the magnetic dipole. quadrupole. . . . terms.

Magnetic dipole
The dominant first n = 1 tenn in (2.2-42), called the dipole tenn, can be transformed
into
 2.2 MAGNETIC VECTOR POTENTIAL 89

A(p) = iJI , lacosOdI = iJI , l(p.a)dl = 4(p x p) (2.2-44)

41tp '1 41tp '1 4np

where
(2.2-45)

is defined as the magnetic dipole moment of the loop [see in (2.3-6) for a circular loop].
To obtain the last expression we have used the vector relation

p x (a x da) = (p · da)a - (p · a)da • (2.2-46)

derived from the triple vector product fannula (A.3-le), and the differential relation
d[(p· a)al = (p. d,), + (p . a)da • (2.2-47)

�
where the vector p is not affected by the differential, and in the previous notation it is
da = dl; we also note that with (2.2-43) we have d[(p . a)a] = O. Because the vector

H a xdl = H(a+ b) x dl
(a x dl)/2 is proportional to the shadowed triangular area in figure 2.2-6, the integral

s= (2.2-48)

gives the (minimal) vector area of the closed loop, whose value S is independent of the

f
chosen momental eenter O. (In fact. shifting the eenter by b leaves S unchanged
because b x dl = 0.) In particular. for a flat loop. S is the area enclosed by it. and the

vector S :; Sn is perpendicular to the plane and poinring in the direction given by the usual
right-hand rule (figure 2.2-7).
Instead of a closed filamentary loop, we consider a conducting body containing a
divergence-free current distribution, V · j = o. Such a distribution may be thought of as
being composed of an ensemble of filaments closed on themselves or extending to infinity
(which may have a variable cross section sn but contain the same currentjn all around).
each of which idenlifiable with a filamentary currenl through (2.2-37). In this situation we
can transform the expressions (2.2-39, 42, 44, 45) related to a filamentary circuit to those
of a body circuit (and vice versa) by the formal substitution Id. f-+ jdV. For example, the
magnetic dipole moment [which for a filamentary conductor is given in (2.2-45)], for a
conducting body becomes

Figure 2.2-7 Magnetic dipole moment p of plane loop, where S is the aR:a

enclosed by the loop.

90 CHAPTER 2 MAGNETIC POTENTIALS

(2.2-49)

Similarly. the vector potential (2.2-44) for a conducting body transfonns in this
approximation into

(2.2-50)

meaning that the vector potential at a distant point P can be expressed by the magnetic
dipole momenl (figure 2.2-8).
The magneticfield associated with a magnetic dipole is, using (2.2-50) and the vector
transformations (A.3-3.9),

(2.2-5 1 )

By comparing (2.1-2) with (2.2-5 1), it follows that the scaJar magnetic potential of .
magnetic dipole is

�
'Y0 - 4 .1l' ] - ...
- .l.!'£ - L
41r l'..
2 cos e • (2.2-52)
P P

and thus the spherical field components are obtained through the differentiations (2. 1 - J 6),

(2.2-53)

Figure 2.2.1 Magnetic dipole moment of a spatial current dislribution.

 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 91

here p is .ssoei.ted with the dipole moment ofthe specific problem. as given in (2.2-45) or
(2.2-49) and .Iso in (2.3-6).
We recall that a static electric dipole with moment
Pe = ql (2.2-54)

(that is, referre� to an idealized electric field source having two equal and opposite point
charges ±q at some vanishing distance I apart) formally possesses the same potential

(2.2-55)

and hence the same field pattern

E = -VUD (2.2-56)
os .h•• given by (2.2-53).

2.3 FIELDS IN CYLINDRICAL AND

STRAIGHT CONDUCTORS

The general quasistationary magnetic field theory developed so far is now applied to
the calculation of magnetic fields produced by some of the most simple current-carrying
conductors.

Circular loop

The filamentary circular current-loop is a fundamental element for the analytic and
numeric computation ofaxisymmetric magnetic field configurations, so we will look at the
various ways of calculating the magnetic field and its approximations.
From the comments made in connection with the cylindrical field components (2.2-
27), we know that the magnetic vector potentia] of the concentric circular current-loop with
radius a of figure 2.3-1 has only the azimuthal component �. In the spherical coordinates
(p.�. 8) and wi.h r':' = p' + a' 2pa sin 8cos�• •he in.egral (2.2-39) is

� �p2+la a2 J �lc"'� d�
2K
-

A.( P 8) - .1'... [!M -

.""f . 4n: j 'l'Q E COS 4'
(2.3- 1)
0
-
- 4n: •
92 CHAPTER 2 MAGNETIC POTENTIALS

'PQ
,
., .-.�
8 , z
p ,
,
p

Figure 2.3-1 Current loop and coordinate system.

2
e = 2apsin 81(p £ . requiring either p » a or p « a or sin 9
where +a
1 ), For small �( 1.
with ( I _ £ cos�)-
1/2
:: 1 + t ec05". we obtain
(2.3-2)

The spherical magnetic field components of .uH = V x A are deduced from tablc A.3-II.
(2.3-3)

and become in this approltimation

(2.3-4)

In particular, at large distances (p » a) they reduce to the dip:>le field [figure 2.3-2 and
relations (2.2-53)].

(2.3 -5)

where

(2.3-6)

is the magnetic dipole moment (2.2-45 or 49), The on-axis field

,
Hp ( p � l.8 � O) = Hz(7.,O)= I. f (2. 3-7)
2(:2+02)

is a precise (not approximated) expression and could llave been obtained directly by a
straight integration of (2.2-38). Knowing the on-axis axial field, it is possible (in a current
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 93

Hp

. Figure 2.3-2 Dipole-type magnetic field generated by a current loop.

free space around the axis) to use the series expansions for the cylindrical components
Hz(z,r), H,.(z.r) given in (2.1-2t, 22), or for the spherical components
Hp (p.O). Ho (p.O) deduced from (2.1-32).

Note that the contribution of only a circular arc conductor with radius a (extending
symmetrically over an angle -a to a with respect to the y-axis, see in figure 2.3-19) to the
vector potential component A � and, for example, the normal field Ho on axis (0 = 0) at

�' h
distance p =I can be recalculated, from (2.3- 1 , 3) with E = 0, to be

A� = sina , where A. = 1l. (2. 3- 80)

1tV 1 +.-1.2 P ,

(2.3-8b)

According (0 the comments made in connection with (2.2-22), this field contribution is only
correct if the arc is part of a closed circuits.
The exact (numerical) calculation of the off·axis magnetic field of a circular loop can
start from the vector potential in spherical coordinates given in (2 .3-1). An alternative is to
refer to eJliplic integrals: By introducing the cylindrical coordinates z = psin8,

k2(Z. r)
r = p case and the parameter

4D'
=
(a+r) 2 +z: 2
from (2.3-1), after simple transformations, we obtain
94 CHAPTER 2 MAGNETIC POTENTIALS

(2.3-9)

where

t f
K(k ): ( 1 - k2sin2 y)-1/2 d y
f E(k) : f ( 1 - esin2y)1/2 d y (2.3-10)
o o

are the complete elliptic integrals of the first and second kind. whose tabulated values and
functional properties are well known and published.A.3 After differentiation of (2.3-9) and

; E)
some rearrangement, the magnetic field components are obtained:

+- O(:ur� ) '/;; [(a + r)2 + ,2]-1/ K+ 0'_';-,

2

[
H,(" r) : r'
:
J:n
[ (a-r) +z
OA.
H" (z r) = -- " = ..LL (a + r)2 + z2
·-
paz x r
-t/2 ] [
- K+ Q +r +z E
(0_r)2
" +z2' ] (2.3- 1 1 )
2

For small values of k (that is. r » a or r « a) the expansions

K-I
_
I T
(
n + k' + 7>4k
� 4+
.. )
� _ n ( k' 9 4
E - I I - T - "64k + ... )
hold.A.' On axis (r-+O, k-+O), the result (2.3-7) is confirmed.
In conclusion. the off-axis magnetic field of a filamentary circular current loop can be
found through the exact solution (2.3-11) by using the tabulated values for the elliplic
integrals:"" ) In certain situations, it can be more convenient to take advantage of the
approximations (2.3-4, 5) or expansions (2,1 w21, 22), with the field onwaxis being given
by (2,3w7). Finally, there are tables that provide the field distribution as a function of
nonnalized loop parameters,H

Multiloop system

By adding one or more coaxial current loops to the original one, the axia1 magnetic
field can be modified according to special requirements, A powerful method of ana1ysis is
to expand the magnetic potentia1 that results from the array of loops in spherical harmonics,
similarly as with the multipole expansion in (2.2w42).2.25 In the following, instead. we will
take advantage of the single loop results obtained above.

Coil pair
In the case of a symmetric coil pair (figure 2.3-3), the sum of the axial fields (2.3 w7 )
gives
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 95

+/e;) H) ! \ W -I
a) , b)

+{� H , wf -!. '!... 2a-+-

� : �t
� .Qi- � 2a�
� i �1
L I l..
I t.
_._. ._._. . _ _._ . ..

I I •

� \!( �l
I. 21 --<.0-11 1�.-- 21 .1
Figure 2.3-3 Parallel (a) and amiparalleJ (b) current loop pair (the laner has a
magnetic quadrupoJe field panem).

(2.3-12)

where the ± sign applies for paralJeUanliparalleJ current f]ow in the loops. This will now be
developed in lenns of the small parameter (Z2 ± 2Iz)/(a + P)---that is, for a region near
2

the origin, where z2 1 R; « 1 with � = a2 + /2 . We consider here the parallel current coil
pair [for the antiparallei pair, see in connection with (2.3-15)]. Rearrangement of the (enns

Hi(z) = I "f;; ,[ " a ) (-t)2 il fsl4 +a�

! f4/� ( )4 ..
gives

1_ + I ) -t - .
4 1 2a "
] • .
(2 3 - 1 3 )

as the uneven terms cancel oul because of the system's symmetry.

When the separation of the coil pair is larger than the radius. 21 > a, there is a field

= a (figure 2.3-4), generates the central field

minimum al the origin: The parallel coil pair generales a magnetic bottle or mirror
.
configuration (figure 2 3- 3a). On the contrary, a maximum is obtained for 21 < Q. A

H,(O) = 5�' �
HelmholLZ. coil pair, defined by 21

� = O.7 1 6 • .
(2 3- 1 4)

which, according to (2.3-13), is homogeneous within order (lJRo) . Field homogeneity in a

4
direction perpendicular to the z-axis is even better (twice as much) than field homogeneity
along the axis. as we know from (2.1-24) for an axisymmetric system.

Field shaping
Addition of one or more loop pairs can further improve the homogeneity or allow
more conditions to be satisfied. Consider. for example, the simple case where the second
96 CHAPTER 2 MAGNETIC POTENTIALS

Figure 2.3-4 Helmhollz coil pair.

pair has shrunk 10 one single loop, I, = 0 and Ihus a, = 8" in figure 2.3-5. Adding the
l
field (2.3-13), H;l + Hzl, of the two pairs and then requiring the parameters of the Z and
4
Z terms to vanish simultaneously provide two equations from which we easily find
I, I n. = H and I, l l, 32/49 (the total current in the central loop being 21,); alternatively,
=
the current ratio can be approximated, by using the corresponding winding ratio of a wire
carrying the same current. N2 ' NI = 12 '/1, This particular coil system thus generates a
central field tha( is homogeneous within order (tJRO)6. By proceeding similarly, bur
allowing a finite coil separation, '2 ;t 0, we find a homogeneous field within order (z/RO)8
for the following parameters: I, I R" = 0.76506, I, I 8,,= 0.28523, and 1, I Iz = 0.682 1 1 .
Another convenient method is the use of current shims-that is, small, mainly coaxial loops
of circular and rectangular shape--to locally correct the given main magnetic field according
to the specific needs. 2.26
Besides homogeneity, other conditions can be satisfied by an array of loops, such as
minimum spatial ripple within a maximum volume, or maximum field gradient. For the
latter case, we consider the antiparallel coil pair (magnetic quadrupole) of figure 2.3-3; and
from (2.3-13) and by developing for small dR", we obtain the approximation

H,(Z) = I hlM
R.t I/o + ... (2.3-15a)

At the origin z = 0 the field vanishes. but its gradient

z
dH, /d z = 3/a /l R6 (2.3-15b)

is finite (the gradient coil pair). Differentiating the gradient further with respect to I shows
that it is maximized for 21 = a- that is, in an inverse Helmholtz or Maxwell coil pair. It can
be shown [by using (2.1-23), (2.3-7)] that the zero field point is also a minimum gradient
point; that is, there the gradient increases in both z- and r-directions.
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 97

,
,
,

,
,
, ,
, ,

,
,
, ,

,
, ,

,
, ,

,
, , ,
,
p : ,

, ,
,

�.
, ,
,

:
,

,
� -.-.�.
' ,
' ,

-,-, _ . _ . -+. . .J.. ...

' ,

: z
_

, 0 , ,
,0,
0,
, ,

,
,
, ,
, ,
, ,
,
, , ,
, , ,
, , ,
, , ,
, , ,
, ,
, ,
,
,
,
,
,
,

Figure 2.3-5 Double coil pair (magnetic oclupoie fidd pattem) inscribed on a
spherical surface of radius Ra; [he central field can be homogeneous wilhin the sillth,
or higher. order in zlRO.

Thin solenoid

A thin solenoid made of N total windings (figure 2.3-6) and carrying a total current
NI can be thought of as an infinite array of identical loops. Integration over the
contribution to the axial field Hz. (z)at point P from the elemental loop positioned at z' with
current id z' gives. with (2.3-7),

(2.3-16)

where j = NlI2b is the constant linear current density. By introducing the eenter-field
parameter

Hoo = Nll2b , (2.3-17)

which is the axial field at z = 0 for an infinitely long coil. or, in practical units,

Hoo = O.41tNII2b , (2.3-17)*

98 CHAPTER 2 MAGNETIC POTENTIALS

r ...
·_ . _
. _ · ·_O-
_ · ........
.. . --r

2b ------+-101
Figure 2.3-6 Thin-coil coordinatt: system.

and the dimensionless position parameters

� = zla and /3 = bl a , (2.3-18)

integration yields

(2.3-19)

or, alternatively,

where the angles (J are defined in figure 2.3-6, In particular. the eenler field at � = 0 is

H,(O) = If{, ( 1+;')'" (2.3-20)

Off-axis fields can be approximated by the series expansions (2.1-21, 22), or

calculated exactly by integmting expressions (2.3- 1 1) over the elemental loops. It is to be
expected that the largest axial field is obtained at the midplane near the windings and that the
largest radial component is obtained at the coil ends (z = ±b. r -+ a), as we shall now
continn.
For the vector potential of the thin solenoid at r = a . by integrating (2.3-9) similarly

� rHt !) f] d z'
to (2.3-16) we have

�( z, r = a) = - K- (2.3-21)
:
where now

(2.3-22)

in analogy (0 definition of k2 given in connection with (2.3·9).

Since from (2.2-27) we have /lff, = - dA./d zand d k' Id z = -dk' Id z', we directly
get the radial component
 99

[( ) K(k) - -
2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS

E( k ) k=k+{Z'=b)
H,(,. r = a) = {% t - ! -
k ]k=k_{z =-b)
, (2.3·23)

which can be detennined for any z value with the help of the K and E elliptic integral
tables'\.}. We explore the field behavior near the end of the coil. at r = a and z = b +

�
2a£ where e is a small number, and obtain from (2.3·22) with f1 = h/a

'
k+(V = a; ,' = b) = -- , k.(z,r = a;,' = -b) =
1+£ I +�

When £2 « I . and since A.I

for k =- 1 up to order ( 1 - k 2 ), the approximation

H,(z = b + 2ae, r = a) '" 2�\,[-lne - Cl
is obtained from (2.3-23), where e(L lis the contribution from k = k_ which vanishes for
a very long coil, fJ » 1. *_ == 0 [as obtained from the approximarions given in connection
with (2.3- 1 1 )) . In this extreme case, Hr diverges around the coil ends, as one would have
expected.

2.0 r-r-r-r-r-r-r-r-r-TTTl

il2b

1.5

0.5

0.5 0�L-L-L-'-:;
0':
.5-'--L-
z .L-"""'1.0
b

Figure 2.3-7 Current distribution in an idealized single-turn solenoid.

100 CHAPTER 2 MAGNETIC POTENTIALS

In a single-turn solenoid (made of one conducting sheet), the linear current

distribution i(z) will establish itself by interaction with the field and in accordance with the
geometry and conductivity of the conductor. Let us consider an ideal (superconducting)
thin solenoid, which again can be considered to be an array of filamentary loops, each
carrying a current i(z)dz, where i(z) is not constant but must be determined. This problem
can be solved by considering the vector potential with the appropriate boundary condition
or, more simply, by using the loop Cannulae (2.3-1 1). In fact. the contribution to the radial
field Hr at the point (Z,T = a). due to the infinitesimal loop located in the plane z'. can be
wriuen as

d H,(z - z'.a) = I(z - z'.a);(z')d z · • (2.3-24)

where I(z - z',a) is clearly detennined from (2.3- 1 1). The boundary condition requires
Hr ;;: 0 an along the solenoid; that is. the integral equation

f
+b , ,
I(z - i , a) ; (z ) d z = 0 (2.3-25)
-b

detennines the unknown 1(,). The numerically calculated result2.17 is shown in figure 2.3-7.
The current density diverges at the end point as (bLz2)-II2 . This case is an example of the
genercil problem of field behavior around sharp corners of (superconducting)
conductors.2_11

Thick solenoid

After the filamentary loop and the thin solenoid. the next logical extension is the thick
cylindrical solenoid (figure 2.3-8). We assume that the azimuthal current is characterized by
the mean conSlanl-currenl density g. where j is related to the cross section I of each of
the N total windings and f is the filling factor defined as

f - conductor val. _ NI
(2.3-26)
- coil volume - 2b(1I:2 a[)

We start by looking for the general case-that is. the off-axis magnetic field.
Consider the vector potential A == (Ax .Ay , A(). (2.2-21), in Cartesian coordinates. We
know that in this axisymmetric case, it will have only the azimuthal component �, which
in the system of figure 2.3-8 (where P lies in the y = 0 plane- that is, the plane including
the x,z-axes) is identical to its component A :
y

(2.3-27)

with
 2.3 FIELDS IN CVLlNDRICAL AND STRAIGHT CONDUCTORS 101

,
, I
,
,

.z :
I. '," " ,
"

X4S"��Y
I ' P' (,. ,1
o · /- .;.
'--"._.,,-_"'::
'PQ
�
�
.,�----'-
I
'1 y
i
..--
_ _ 1 __
-
..__- 1 - .....
P (z. r) I
7-� I
.--- --.....J
I

Figure 2.3-8 VecTor potential for a thick. axisymmetric coil at point P in the
y = 0 plane (for practical reasons, the decomposition of the current density j' �
flowing at point Q is made in The z = 0 plane of The drawing).

and

(2.3-28)

which is detennined through the triangle PP'Q. The solution is

fjn: f"'fb
p4-
(r.t.) - - -b 12"
r'cos 4l d r'dl'd4'
A �
'"f/J - (2.3-29)
a, 0 [(z'- Z)2+ r' 2+r2 -2r'r cost'lj1/2

with the magnetic field components (2.2-27)

aA,
H , (r. z) = _ .1p ill. and (2.3-30)
102 CHAPTER 2 MAGNETIC POTENTIALS

For a general discussion of these analytic results we refer to the literature.2.11,2.19

Nowadays, numeric integration of this result is relatively straightforward; care must
be taken near Ihe axis (r --> 0), where il might be better to use approximations of the type
shown in (2.1-21, 22). If a computer is unavailable 'or when the fields have to be calculated
only in a few points, it might be convenient to use tabulated fields given as a function of the
dimensionless parameters r I � and z I � . 2.16 In fact, figure 2.3-9 makes it clear that any
regular solenoid with constant current density can be considered the resultant superposition
(with appropriate signs) of four semi-infinite solenoids with zero inner radius
(a, --> 0, b --> �) . We see from (2.3-29, 30) that the normalized fields for each such coil,
H,
hr = /ja • - !!L (2.3-3 1 )
z ''4
- /jaz '
depend only on the dimensionless parameters rl a1 and z/ G-z. and can thus be tabulated
conveniently.
The on-axis field component. Hz(P},
can be obtained by extending, as before. the
loop or thin solenoid ex.pressions (2.3-7. or 16), or by calculation from the general result
(2.3-30) with r � O. The latter, however, needs some tedious transformations, using
partial integrations to avoid singularities. We prefer to return to the original integral (2.2-
P
22) and write the axial field component at point as (figure 2.3-10)

tl-
Ifr\

t
a,
___ _.
..
_. _ _p�._. Hz I
+-

- -0
,

, ,
:.. 2b ----+-l
, ,

Figure 2.3-9 Reclangular coil with constanl current density as SUperposilion of

four semi·infinite coils with zero inner radius.
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 103

-'--�
I I I p
.-.-.
1 0 --"<' -'-'
1 ' --"
� ---
H,(P l '"'

Figurt: 2.3-10 AKisymmetric thick solenoid.

Hz (p) = _1 J fJ r sinB
4n: 2 'f'
d�dr d z' . (2.3-32)
''0

lmegrating over 4J and introducing sine = r I rPQ.rPQ = [12 +(7. _ 4 ')' ]1/1 yields the general
fonn

Hz(p) = tff jj , I'" lI, d r d z' (2.3-33)

[I +(z-z ) J

which can be applied to various cylindrical coils.

j
Provided that the current density and the filling factor /. (2.3-26), are constant over
the rectangular cross section of the coil, it is convenient to introduce the total current

l = jI
flowing within one conductor winding with cross section £ and to eliminate f. With
parameters similar to those in (2.3-17, 18),

(2.3-34)

(the origin of � being at the eenler-point of the coil) and after a first integration over d 7.'
[similar to the integration in (2.3-16)) and then a second over d r, we get
104 CHAPTER 2 MAGNETIC POTENTIALS

In particular. the center field al � = 0 is

H,(O) = Hoo
L
a-I
ln( J
a+ a' +p'

J
I+ I+P'
) (2.3-36)

This expression can be simplified for various limiting cases. For example, for a thin disk
coil, where � « I , it reduces (for unifonn current density) 10

(2.3-37)

Bitter solenoid

Instead of the multitum arrangement considered so far. a thick solenoid can be made
from sheet windings or conducting disks. The disk magnets were introduced by Biuer-8
and are particularly suitable as high-field generators because they are mechanicaJly strong
and easy to cool (figure 2.3- 1 1).
The steady-state, resistively determined current distribution within a disk is described
by the nonuniform radial dependence

(2.3·38)

a) b)
j(r)

000
a
f\
�L--e _. .....
,
,)

I-
2b
-I
BD��
Figure 1.3- 1 1 (a) Bitter magnet made of a stack of conducting disks; in
practice, the stack can be obtained from machining along a helical pattern from a
solid thick solenoid [schematically shown in (b)]. or the single disks are cut open
along a radius and connected 10 adjacenl disks as indicated in (c).
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 105

where )1 ' the maximum vaJue at the inner bore, at r = al • is given in (23-39b). Current
density and field in the conductor are given in (7.2-29, 30).
The axial magnetic field is again calculated from (2.3-33), as for all coaxial solenoids.
In particular, for the eenter field in a rectangular Bitter coil, after inserting (2.3-38) and
integrating,we obtain

(2.3-39a)

where the lotal current to be used in Hoo with the filling factor J. (2.3-34), is

(23-39b)

Toroidal magnets

Toroidal magnets in various arrangements are widely used in thermonuclear fusion

research since they produce (nearly) closed, toroidal magnetic field configurations for
containing hot ionized gases or plasmas. Ideally, a toroidal magnet consists o f a current
(density) flowing around the minor axis of an axisymmetric toroidal surface, obtained
by rotating any cross section (circular. D-shaped, etc.) around the major axis. This can be
well approached by a densely packed winding of N turns around the torus (figure 2.3- 12a).
In practice and most often, however, a toroidal magnet is made from a number "c of
discrete coils (figure 2.3 - 1 2b; see also figure 2.5-2).

Figure 2.3- 12 Toroidal magnet: (a) Ideal magnet with uniform current density j
nowiog around the minor axis (in practice this can be obtained by a unifonn,
densely packed winding); (b) magnet made of a finite number of coils regularly
placed around the altis (only some of the coils are shown).
106 CHAPTER 2 MAGNETIC POTENTIALS

In the ideal axisymmetric coil, where the current density flows around the (J direction,
the magnetic field inside the torus reduces to only the azimuthal (toroidai) component Hq, =
HT. which is invariant along a concentric circular path with radius r. Thus, by applying
Ampere ' s law ( 1 .4·3) 10 such a path we gel

(2.3·40)

where N is the total number of turns. each carrying the current I. The field is as if it were
produced by a CU!Tenl NI flowing along Ibe major axis [see (2.3-43)J bUI limiled to Ibe
inside of the torus only.
When the toroidal magnet is made of a finite number of separate coils, the magnetic
field pattern is disturbed with respect to the ideal azimuthaJ configuration, as illustrated in
figure 2.3-13. The perturbation is often characterized for practical purposes by a so-called
magneticfield ripple parameter

magnetic
field lines

/�
/�T ·
equimagnitud
e lines

'.

Figure 2.3-13 Toroidal magnet made of an array of n, = 16 equally spaced

coils. This partial cross section in the z = 0 equatorial plane (with 4 coils) also
shows some magnetic field lines and, in the lower sector. some equimagnilOOe lines;
they clearly indicate the magnetic field ripple between each coil pair.
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 107

and Hmax and Hmin are the maximum and minimum magnetic field values occurring along a
circle of radius r, on the intersection of the equatorial plane z = 0 with the midplane of each
coil and with the midplane between two adjacent coils, respectively. The ripple pamrneter e

strongly depends on the total number ne of coils and on the radial position r; it also depends
somewhat on other quantities-for example. the azimuthal thickness of each coil and the
separation distance between each coil pair. The magnetic field configuration and the related
ripple parameter for a discretized toroidal magnet has to be calculated numerically.
However. it is also possible to establish analytical approximations by considering the dipole
field of each coil, or approximated coil pair, as given, for example, in (2.3-4, 1 1 , 1 3 , or
30). Such approximations can be useful in system studies of toroidal magnets when the
problem is, for instance, to determine the simplest (and least expensive) magnet systems
that provide an acceptable ripple within a given toroidal chamber. In fact, the size of the
toroidal magnet, the number of coils, and the coil dimensions are parameters that influence
the cost and the energy requirement of the system, as well as the maximum {olemble
stresses in it. (The stress distribution in toroidal magnets is discussed in section 7.3.)

In many applications of toroidal systems (in particular, for the confinement of

thermonuclear plasmas), in addition to the toroidal field HT. also poloidal fields Hp
(generated b y azimuthal or toroidal currenl..<; ) are used. The resulting helical field pattern (the
result of the vectorial combination H = HT + H p which generally maintains the
axisymmetry) is discussed in section 2.5.

Straight conductors

The very long straight conductors shown in figure 2.3-14 can be considered as a
bundle of filamentary currents, each with cross section ds and current jds. The vector
potential has only the A::-componcnt, and for cach filamcntary current of length 2J (figure
2.3-15) the corresponding contribution dA. from the integral (2.2-39) becomes

d Az ::: .fJr jdSr:�=�!r2 ::: tJr jd s ln( <: + �z2 + r2 �::::: - tJr j d s[lnr - ln(I + �J 2 + (2 )]

(2.3-4 1 )

The last term is independent of, when I ») r. so it does not influence the calculation of the
magnetic field ,uH ::: V x A and we neglect it in the Az-component:

Az = - in J jlnrds = - tkf Inrds = -!klnr (2.3-42)

S

here the second expression with the total current I holds when the current density j is
constant over the cross section S of the conductor, whereas the last expression is for a
filamentary conductor. In this case from (2.2-36) we obtain
108 CHAPTER 2 MAGNETIC POTENTIALS

P(X,y) a)
b)

T2b
I
2b --­
x

1
Figure 2 . 3 - 1 4 (a) Cross sections of a rectangular bar with current flowing
unifonningly along the positive z-axis towards (he reader, and (b) the related
isomagnetic lines.

(2.3-43)

which is the azimuthal magnetic field (2.1-37b) outside a very long cylindrical conductor.
When the limit I » r does not apply. we obtain from (2.3-41 ) for a filamentary
conductor extending symmelrically from -I to +1 (figure 2.3-15)

A, = � In[1+ �f,+ I ] = f.H -lnA + In(1 + �I ,f)].

+ where A = f (2.3-44)

(r being the radial distance from the conductor), with the azimuthal field at point P of
(2.3-45)

More in general, we will show in (9.2-4) that the contribution to the azimuthal field
Hf,12 at point P of a straight section defined by the vectors r l,r1 (figure 2.3-15) amounts
to
(2.3-46,)

that is,

(2.3-46b)

In particular, the contribution of a conductor extending from points 0 to I is [compare with

(2.3-45))

H41,01 _ ..L 1 ' '-

"'Ui - 701 '
-
�I+�I
, (2.3-46c)
- 4Jtr
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 109

- - .- - 'I
--..:..
- O ...;...-- 2
--....
101 112 l
.

l
.;;.-
. - . -. �

• •

Figure 2.3-15 Straight. long filamentary current.

with which we can also write

H.,.12 = H.,.02 - H�.OI

Example: Rectangular bar conductor

- 14 is
The vector potential component (2.3-42) for the rectangular straight condoclOr shown In figure
23
. a

u b
11
Al ( x.y) = -ift fab J J1nIX'-X)2 +( Y' -y)2j 2d,x'dy' (2.3-47)
-u -b

- - 16�b { - x)(b
which can be integrated 10 give1.1

Al ca - y}ln[(a - .e)l + (b - y)2 J +(0 + x)(b - y)lol(o + x)2 + (b _ y)2 ]

J ( J
+(a-x)(b+ )')In((a -x)2 +(b+ y)2 )+{o +..r)(b+ y)ln[(a+ x)2 +(b+ y)2)
+ (a - X)2[arc tan !=: + arc tan ::; + (a + x) arctan !:! + arc lan ::;

u+'!] + (b + [ sd.!]l
2 (2.3-48)

+(b - y)2 [arc lan Q.::,!

b-y
+ arc tan b-y 2
y) arc tan )i-x
+,
+ arc tan
b+,

The magnetic field components are calculated from j.tH = V x A with table A.3-1J,

H = -'-�
.! JI ay .
HY = _ l.�
p ih
' (2.3-49)

After differentiation of the above result and then defining the boundary corner by tile polar coordinates
(figure 2.3-14a) suell tllat

2 +(b- y )2 1 _- 2Inr ,
In((a - x )
b-y _
are tan _ - 414,
4 ...•
o .r ...•
110 CHAPTER 2 MAGNETIC POTENTIAlS

�[(y + b)(�1 - �)-(y- b)(414 -�) + (x +a)ln � -(X - a)IO;}]

we obtain for the field componems

� [(X +a)(� - �)- (x -a}«(tl - ';4)+ (y+ b)Jn -: -<y b)ln 'fr] .
H,,(x. y) -'"

2
( 3-5. 0)
Hy (x,y) '" -

We note from (2.5-13) that the m!lgnelic (flux or) field lines in the (x,y) plane are defined
as At(x,y) = conSI. curves (figure 2.3-14b). More in general, expressions similar to (2.3-50) can be found
1_l1
for the magnetic field generated by long. straight conductors with polygon cross sections.

Example: Strip conductor

As an altemale solution method for finding the magnetic field H without going through the vector potential
A , we now calculate the field related to a suaighl, thin strip conductor whose shape in the (x,y) plane is
described by the curve y = y(x) (see figure 2.3-16a). From Biol-Savart's law (2.2-22), with
jdVcrjd)dxdz=idxrlz and integrating from (: = -DC> to +DC>, or simply from (2.3-43), we know that the
contribution dH to the field in P(x,y) from a snip filament with width dw' in Canesian coordinates is

dH "' ....L�
2. ,
where i is the constant surface current density, and

,2 = (x _ x,}2 + (y _ y,)2
(dw,)2 = (dx,)2 + (dy,}2

Figure 2.3·16a shows that dH� = sina dH. sina'" (y - y' )/r, and so on. Thus. we find

a) b)

p(x,y)
y

'2
L.:-:-
,-- ",* P
P(x.y)
y
, dH x ,
'

, ,
,
, �2 (i d..- l
.. • ..

a :
.J __ Q (x',y)
, h
___
dy '

d..-'
((:)+
-a «(:)+ a x
I' 2a '1

Figure 2 . 3 · 1 6 (a) Strip conductor of infinite length along the z-axis, whose
cross section in the (x y)-plane is described by the curve y = y{x) [the surface current
.

density i(x',y') flows along the positive (:-axis. towards the reader], with details of
the decomposition of the transverse line element dw' and related field contributions;
(b) plane srrip conductor.
 111

dH.( = _ ...L d dw'

211' ,2
(2.3.5 1 )
dH := ..L x-x dw'
Y 211' ,2

and the field components Hx ' Hy 31 P(x, y) are obtained by integrating these equations along Ihe cross­
seclional curve y' = y'(x") of the strip.
As a simple example we consider the flat strip of width 2a shown in figure 2.3-16b (thus.

[ j
y' = h = consl.. dw' = dx') for which by integration we obtain

46 y-h
arc lan ..!=.JI..
y-h
Hx =-..L arclan ..Lt.t -

(2.3-528)

By inlroducing the angles �I' � and dislances 'I' rl (figure 2.3-16b), similarly as in (2.3-50) with

and the total cUrTem l = 20; flowing in the strip. we rewrite these componentS in the simple form

H, = - f,;; (�I - �I )
(2.3-52b)
H = ...Lln !l.
Y 2.cu '1

(Through figure 2.3-16b we see that � -;t = rp, - rh·)

Example: Two-dimensional multipole configurations

The distribution of conlinuous or discrete line currents along the surface of a circular cylinder
produces two-dimensional field configurations, which can be: calculated in closed form and have significanl
propenies for many applications. Consider the el'illmple shown in figure 2.3-1711 with the continuous surface
current density

(2.3-53a)

where N(+) current sheets (the first eXlending over the angle -1CI2N to + 7fl2N) alternate with N(-)
sheets. each of the 2N sheets carrying the current

(2.3·53b)

The related magnetic vector potential component (2.3-42) is expressed as

2.
Az(r.41) = �f izlnRd a (2.3·54)
o
-
112

.) b)

y +1
./ " - -
,
,0'
,
,
(-) (+)
x , x
,

"'1::::1=
: ,:$''';(
,
,
0, a (-It' I
.. ..
here N = I ) (here N = 4)
- - , ,,n= 2N- 1

Figure 2.3-17 Current distribution along a very long cylindrical surface, producing
a two-dimensional. 2N-mullipole magnetic field: (a) continuous, (b) discrete 2N
filaments.

With

R _0 +, eastl/! - a) = cos,cosa + sinl/l sina (2.3-55)

2 2 2:
__ 2are05(, -a),

the calculation of (2.3-54) through integration by parts gives

Ij� p N cosH;
ria
Al(p. ;) = t. (2.3.56)

where [here and up to (2.3·6 1 b») p = and the upper sign applies for r S. a and the lower for , � a. The
cylindrical field components are defined by (2.2-36),

(2.3-57a)

whereas the Cartesian components are then obtained as indicated in table A.3-I.

(2.3-57b)

Aroond the symmelry z-axis (p-+()j we recognize the muhipole character of this field configuration:

H = const.(H-t = O . H = const.)
dipole(N = I)
y
quadrupole(N = 2) H =0; VH = const.
(2.3-5H)
l\exapole (N = 3) H = 0; VH =0; aH - cons!.
octupole(N = 4) H = O, VH = O, <lH - O

The dipole and quadrupole configurations will be discussed in (2.3-63 and 67).
When the current is bundled into 2N equally spaced filaments carrying allemat�ly the current ±J as
shown in figure 2.3-17b. the vector potential component is calculated similarly, but with (2.3....2) now as

2N-1
A, = _E!..
2.
r. (-I)"lnR" + const., (2.3·59a)
".0,1.....

where Rn is the distance from each singular current filament so tha! with (2.3-55)
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 113

[
2InRlI = lno + p2 ) + ln l - 2P2 COS(Ii"- nZ)
'+p
] . (2.3-59bl

For the calculation of the summation at , S a use repeatedly the expansIOn

In( I ± z) = ± z 13n ± 1.3/3 r.·'4± . . valid for z < 1; the final result is 1.21
- - .

(2.3-60)

Nole that this expression and (2.3-6Ia) are valid at r 2: a as well as r S a, since they do nOI change when
substituting p with lip.
The cylindrical field components are calculated again from (2.2-36)

(2.3-61 a)

In first-order approximation by neglecting ptlN == O. we obtain

(2.3·61 b)

where again the upper sign applies for r < a and the lower for ' > 0. Around the symmetry axis
(p rJa......O) we find the mullipole paUem (2.3-58), whereas al large distances (p = rla 1 ) the multipole
= ..

field decreases as 1/,.N�1.

Magnetic fields resulting from other current distributions along the circular cylinder surface have
been calculated-for example, the filamentary arrangement of figure 2.3-17b but with all currents nowing
in the same directionl.l1, or the 2N equally spaced sheet conductorsl.l1 that represent an azimuthal extension
of the filamentary conductors in figure 2.3-l7b. The magnetic fields of toroidal multipole configurations
have also been presented as closed analytic eJlpressions. 1.1t
In this type of problem, it is often convenient to use series expansion for the current distribution.
The sol utions for the vector potential and the field components are then also in the form of infinite series.l.ll
In addition, the calculation of two-dimenionsional magnetic fields with compleJl variables and functions (as
introduced in section 3.1) is often particularly uscful.llI

. _�tL
z

Figure 2.3-18 Uniform magnetiC fields in relatively large volumes can be produced
by a pair of long racetrnck coils. It is, however, convenient to bend the ends along
the surface of a cyl inder to have optimal access to the uniform field hence the -

name saddle-shaped.
Saddle-shaped coils

Saddle-shaped coils are used to produce uniform, transverse magnetic fields over a
relatively large and long volume (figure 2.3-18), as, for instance, required in particle
accelerators for bending charged particle beams. We first discuss a filamentary coil with
finite length and then analyze the two-dimensional field pattern produced by filamentary.
sheet, and bulk straight conductors.

Short filamentary coil

The saddle-shaped coil shown in figure 2.3-19 consists of two identical filamentary
conductors wound on the surface of a circular cylinder with radius a and length 2/. The
field anywhere in space may be calculated by applying the Biot-Savart law (2.2-38, 39).
At the center of the coil (and of the coordinate system) it consists (because of symmetry)

] .
only of the vertical HOy-component. which is

Hoy = u[
M -p ", l+il2
+ 2 3/2
(I+A: )
sma, '-
,, - !!
/ . (2.3-62)

The first term comes from the four straight sections [each contributes a component
H. sina; see (2.3-45)], and the second term comes from the four circular arcs extending
over the arc 2a [see (2.3-8)].

.
) b)
y

'(i
/
,
,
x

. -:-----...�L71
....-- -·"
2

, / 'a
2/

Figure 1.3·19 Filamentary saddle-shaped coil producing a dipole-type magnelic

field. [In the cross section. the (+) sign means. as always. that the quantily
points/flows out of the page towards the reader; the opposite applies for the (-)
sign.)
 2.3 FIELDS IN CYLINDRICAL AND STRAIGHT CONDUCTORS 115

The position and extension of all the arcs and straight sections contribute to the
homogeneity of the central field. This can be analyzed similarly to the loop pair
conductors discussed in connection with (2.3-15). For example. it can be shown2.14 that
good homogeneity (where all second derivatives of HOy with respect to the coordinates
x,y,z vanish) can be obtained with A = 0.5 and 2a = 120°,

Long coils
We now consider very long straight conductor arrangemems, where the
contribulions of the connecting ends becomes negligible. The question here is: Are there
any conductor arrangements (that is. current distributions) that provide unifonn, purely
transverse magnetic fields in such two-dimensional configurations? The answer is yes; in
fact we have already found them. For the dipole current sheet with N 1 in figure =

2.3·17a, from (2.3·57) we find a homogeneous field within the whole cylindrical space
r $. a:

(2.3-63)

where J is the total current carried in the sheet in the (+) direction, as well as in the (-)
direction. The same applies for the filamentary current with N = 1 in figure 2.3·17b, but
limited to a small space around the symmetry axis, where from (2.3-61b) we obtain
(2.3-64)

y
a) b)
y

2a

Figure 2.3·20 Two overlapping cylindrical conductors carrying the opposite,

uniform current density jl = -il for the production of a uniform magnetic field H,. In
the overlapping region the cum:nts cancel out, so we can consider lhis a conductor­
fru aperture.
116 CHAPTER2 MAGNETICPOTENTIALS

There are other such arrangements. where the conductors have finite thickness.
Consider two overlapping circular conductors with their centers at distance l' in which Ihe
uniform currents flows in the opposite direction (figure 2.3-20). Within one single
conductor, the azimuthal field is [see (2.3-43)]
2
H", = j��r = tjr (2.3-65)

Within the overlapping region. the field components are calculated to be (figure 2.3-20b)

Hx = tj( �siniflJ - "1sinlP2 ) = O ,

(2.3-66)
Hy = t j(-rlcos� + r2coslP2 ) = t,j
We thus find a uniform. pmely transverse field in the overlapping region. In (his region
the net current is zero, so we can consider it to be the aperture of the magnet surrounded
by two circular sectional conductors.
We know from sections 2.1 and 3 , 1 that two-dimensional magnetic field problems,
such as those resulting here from very long straight conductors, are conveniently solved
by application of the conformal transformation method with complex functions. The
method transforms complicated current distributions of the real coil (for example/·I� two
overlapping ellipses, as shown in figure 2.3-2Ic) into the overlapping circular conductors
(figure 2.3-20) where the problem is solved easily. as above. This is then transformed
back imo the solution of the real coil.
Besides uniform fields, which arc obtained as we have seen by dipole-type magnets.
important applications also depend on fields with constant gradients-for example. to
produce strong-focusing fields for guiding and slabilizing charged particle beams.6.8 The
quadrupole magnets shown in figure 2.3-21 produce zero magnetic field on the axis and
constant field gradients around it, as mentioned in (2.3-58). For example. for the current
sheet of figure 2.3-2Ia we obtain from (2.3-53. 57) with N = 2 and y = r sin t/I .
x = rcos t/l.

.) cl
)
b)
y (- Y

12b
-/
--- -.•
(- ) ," -
, • ,
, ,
, ,
(+1 (+1 +/ ' • +/
,
,
x •• 1+) x
, ,
, , 2.
, ,
, ,
-

(·1
1+1 '"
-/
-_ (.}
.....

2a
I• .1
I. 2b .1
Figure 2.3-21 Example of two-dimensional quadrupole magnets (cross sections).
The filamentary and sheet conductors arc as in a previous figure with N = 2; the bulk
conductors in (c) result from tWO overlapping ellipses. as mentioned in the text.
 2.4 INDUCTANCE OF CONDUCTORS 117

,
Hy � -lxI2a- (2.3-67)

which provide the properties just mentioned.

2.4 INDUCTANCE OF CONDUCTORS

The magnetic properties of a conductor system are conveniemly expressed by the

inductance parameter L. which (at least in steady conditions) is a purely georneuic quantity.
In this section, its relation with the magnetic field-and, in particular, with the magnetic
flu<-is established and its properties then illustrated with different examples.

Self-inductance

The magneticflux '" referred to a closed filamentary loop C subtending a surface S is

J B d s � t A . dl
defined in (2.2-3) as

"' � . (2.4- 1 )
s c

When the flux. refers instead to a fmite closed conductor with volume V (figure 2.4-I),
straightforward extension with the relation (2.2-37) yields

'" � HA. jd V (2.4-2)

v

__ dVp

\,
Figure 2.4-1 Inductance of a closed conductor.
118 CHAPTER 2 MAGNETIC POTENTIALS

The vector potential at the point of integration P in both cases is the result of all the currents

J1 f JQ
flowing in the conductor, and from (2.2-21) it is given by

Ap = 4x 'l'Q d VQ (2.4-3)
vQ

If we assume the current density J to remain everywhere proponional to the tota]

current 1(1) in the system [i.e. . j = Ig(x,y,z), with g being a purely geometric vector
function], by combining the last two equations it follows that the flux tbreaded by the
conductor system can be wrinen in the form

'I' = U , (2.4-4)

L = : -;, f f J��:�o d VQ,d VQ '

where we define the self-inductance, or simply inductance, by

. (2.4-5)
vQ, vQ

which is a purely geometric expression. By the same token, and because of the flux
integrals (2.4-1, 2), we can also express the inductance with the magnetic and vector
fields:

L = +fB . ds , (2.4-6)
S

1
L =*fA.jdV
v
(2.4-7)

Note that because the integral (2.4-2) is taken only over the conductor volume V (outside t
i

vanishes because j = 0), the consequent integral (2.4-5) also extends (twice) over volume;
to express this we have substituted P ---+ Q'. In (5.1-33) an alternative definition of the
inductance is proposed, which is linked to the magnetic energy thread by the conductor
system.

Filamentary approximation
When the cross section of the conductor is small compared to all other dimensions,
(figure 2.4-2), it is possible to apply the ftlamentary conductor approximation, expressed
by the transformation scbeme (2.2-37), and the vector potential (2.2-39) then formally
becomes

(2 4- 8)
.
Instead of (2.4-5, 7), we get
 2.4 INDUCTANCE OF CONDUCTORS 119

\
dl

Figure 2.4-2 Inductance of a closed conductor in the filamentary approximation.

(2.4-9)

L = + jA d l . (2.4- 10)

The expressions (2.4-8, 9) [and as a consequence also (2.4-10)) diverge because 'QQ' or
�Q vanish when Q' � Q or Q -+ P t and thus have no physical meaning; however. their
fonn is useful in calculating mutual inductances, as we shall see. The filamentary conductor
approximation, which is sometimes used to model complicated conductor systems
(section 9.2), is convenient for numerically computing magnetic fields, electromagnetic
forces. and mutual inductances between conductors. but is useless for calculating the (self)
inductances, so we have to go back to the real conductor, or al least to a physically
meaningful approximation.
A further comment on these formulae concerns multitum contour coils consisting of a
total number of N (densely packed) filamentary windings. Because of the double line

1ff I
integration (2.4-9), we now find

A . d l = � D . d. ,
2
L= (2.4-1 1 )

where the integration is over one-valued integrals (but again, these integrals diverge for the
filamentary approximation). Here I, as everywhere in this chaprer, represents the current
flowing in the coil-that is, in one of the N filamentary windings, with B and A being
genemled by 1 (in other words, NI would be the total current flowing in the winding pack
that generates the vector fields NB, NA), By definition, in the flux relation (2.4A) the
multitum faclor N 2 is always included through the inductance L.
120 CHAPTER 2 MAGNETIC POTENTIALS

In conductor systems where there are only surface currents, the formulae based on
the filamentary approximation can also be v ery useful. The integration paths Cp. CQare

now any lines lying on the conductor surface and appropriately encircling it (figure2.4-1),
and by avoiding Iheir intersection, one can eliminate any divergence in (2.4-9). [This
follows directly by physics arguments but it can also be formally established by rewriting
(2.4-5) into surface integrals with the transfonnation scheme used in connection with (2.3-
16).J

Faraday law
The inductance represenL" a Jumped parameler (in addition to the similar resistance
and capacitance parameters; see in connection with figure 4.5-1) that can be convenient for
describing (inductive) circuit behavior. The inductance of a conductor system is also
practical for ex:pressing the magnetic energy linked to it, as we shall see in section 5 . 1 .
Through the magnetic flux expression (2 .4-4), it also provides a simple expression for

U = ' E . d l = - d��)
Faraday's law (1 .4-9): The emf. or induced voltage. in a closed (filamentary) circuit is

(2.4-12)
C

Example: Wire transmission line

As a simple example, we calculate the inductance of the parallel wire transmission line of figure 2.1-
7 in section 2.1. In the high-frequency limit (surface currents only), the flux linked to the system is
represented by the flux lines crossing the x-axis between the two wires, and we find, with (2.1-59),

td-a td-a p+u-1d

=Po f J dx 2
kf / = Y.
I 1
a-2d
H d x = If
l.ttQpl 0 p2_x2
= -E.2.1n
If p_u+.l..
2d
(2.4-t J)

After a simple transformation with

(2.4-14)

we derive, for the inductance,

(2.4- 15)

or in second·order approximation (when ald « I)

(2.4-16)
 2.4 INDUCTANCE OF CONDUCTORS 121

The inductance based on (2.1-60), whict

.. ignores the proximity effect. is obtained analogously
(formally by the substitution p --+ 1f2d), and in the same apprmdmalion it becomes

L=
PO
,
[
I ln " - �
" d
l . (2.4-17)

The c",ample of the coaxial transmission line is given in section 5.1 .

Mutual inductance

In many practical arrangements, the conductor system consists of separated

subsystems, so the inductance can be structured appropriately in a simple and significant
form.

Filamen tary approxima tion

We consider filamentary subsystems and begin with two loops ( 1 and 2) carrying
currents I] and h (figure 2.4·3). The integration path in the integrals must be substituted
by

(2.4-18)

or, for simplicity,

(2.4-19)

Similarly. the integration volumes split into

VQ, -> V\+ V2 ' (2.4-20)

By combining (2.4-4, 8, 10) the total flux is now

(2.4-21)

which gives four double-line integrals, each having the current ( h or 12 ) pertaining to
circuit I or 2 on which the moving integration point Q lies (figure 2.4-3): With Q either on
circuit 1 or 2, we also introduce the corresponding notations d I( , d 12.1j 2,and so on, and so
we obtain
122 CHAPTER 2 MAGNETIC POTENTIALS

--- QJ

-----

Filun 2.4·3 MulUal inductance between two loops.

(2.4-22)

with

(2.4-23)

where �I = �2 = M is the so-called mutual inductance between the (WO circuits (often
labeled with M); the (+) sign holds for currenlS II .h turning in the same direction, and the
H sign holds for ihe opposiJe [ihe practical meaning of ihis will be explained in connection
with (2.4-31)]. The self-inductance

L1 1
P 1 1 d Ird I;
(2.4-24)
4ijc1 Ye;
-

'i I
-

and the analogous �2 fann, which also results from (2.4-21), diverge, as we know from
ihe comments on (2.4-10), and in this form ihey have no physical meaning. Self­
inductances can be calculated by referring to finite conductors or current distributions as

expressed by (2.4-5, 6, 7).

When ihe subsystems have NI and N2 windings, respectively, by applying ihe
reasoning relative to (2.4- 1 1 ), ihe mutual induc tance becomes

(2.4-25)
 2.4 INDUCTANCE OF CONDUCTORS 123

For example. if there are two coaxial, circular, multitum loops with radii al .a2 at distance
s(figure 2.4-4), the mutual inductance in this axisynunetric arrangement can be expressed
by (2.4·32) in which 1\;.1 is the azimuthal magnetic vector component on loop 2, generated
by the current 1I flowing in loop 1 ; it is given in (2.3-9) with a � ai , r --7 a2 , Z -4 s.
Consequently,

(2.4-26)

where

k2 = 4a]Q2
(al+a2 )2+s2
and K, E are the complete elliptic integrals defined in (2.3-10).
It is easy to generalize (2.4-4. 8, 9) for a system consisting of n linearized.
(fiIamentary) circuits labeled j = 1,2,.... n and to obtain

(2.4-27)

'Ilk = I,4k 1i. (2.4-28)

j=!

N' 1let A . . dlk = J!. ..N ,N 1 1 d " dl, l

I.
-,k. = I; I 4n- I k 1c,I; Ye; rit (2.4-29)

where the purely geometric expression Li/c = Lki is the mutual inductance between the ith
and kth circuits; as in (2.4-24), the self-inductances L./ck diverge and have to be calculated
for finite conductors through (2.4-5, 6, or 7). (Note that Aj stands for the vector potential
lk
component on the segment d of the loop Ck generated by the current I; flowing in loop
Cj with Nj windings.)

,
I "I

- '12 LI.;;-
dl2 --...
�
FiguN! 2.4-4
subsystem J
/ '\ subsystem 2
Mutual inductance between twO conc�ntric circular loops.
124 CHAPTER 2 MAGNETIC POTENTIALS

General expressIOns
The mutual inductance (2A-?) between two not necessarily fi1amentary circuits with

v, J{AI ' iz)d V2

notalion (2.4-23) is

LI2 = (2.4-30)
v,

When at least one (subsystem 2) is a filamentary conductor (figure 2.4-5) from (2.4-6, 10)

-i; J Al -i; JRI ·dsz

we get, instead,

Lr z = d lz = (2.4-31)
Cz S2

where 82 is a surface subtended by C2 . The last expression shows that the mutual
inductance between two conductor systems 1,2 (figure 2.4-3) is defined as the magnetic
flux lhrough circuil 2 produced by the unit current in system 1 (or vice versa). The (+) or
(-) sign applies in (2.4-22) according to whether the: flux integral is positive or negative.
that is, whether or not the inducing flux in circuit 2 has the same sign as the one generated
by its own current h.
For an axisymmelric system (figure 2.4-5), where an azimuthal current distribution in
subsystem I with total current 11 produces the azimuthaJ vector potential A4I. 1 around loop
2 with radius a2 , the simple form
"<.,
1.� 2 = 2Jra2 ' ,
--
(2.4-32)

s
-I

Figure 2.4-5 Mutual inductance of an axisymmetric conductor system. where

2
at least subsystem is a filamentary circular loop.
 2.5 STRUCTURE OF MAGNETIC FIELDS 125

is oblained from (2.4-31). The vector potential components A� generated by a loop, a thin

solenoid, or a thick one are given in (2.3-9. 2 1 , 29).

If there are NI (density packed) total windings in subsystem I and N2 in subsystem
2. then for the same reasons that led to (2.4-25) the above mutual-inductance formulae
(2.4-3 1 , 32) will include an additional factor N2 (the other factor NI is implicitly expressed
by the ratio Al l I1 or Bl ' I,). For later use we also note that with the magnetic field
expressions (2.2-27), which are valid in an axisymmetric conductor system, the
components of the magnetic induction on the loop ( z = S, r = a2 ) can, through (2.4-32),
be written also in the form
(2.4-33)

(2.4-34)

2.5 STRUCTURE OF MAGNETIC FIELDS

Studies on the magnetic confinement of hot plasmas of thermonuclear interest and,

more in general. on magnetohydrodynamics, as well as on modern electromechanical
engineering have extended our practical knowledge of the structure and properties of
magnetic fields. A magnetic field is a single quantity that can be related (as shown in this
chapter) to a scalar or vector potential extending generally to infinity. Study of its
morphology not only helps to understand the basic aspects of the field itself but also can
lead to important practical benefits.H• These considerations are conveniently expressed if
we base them on the concepts of magnetic lines and surfaces.
There is an interesting historical and philosophical-physical evolution, which can be
characterized by the idea of Ether (the medium filling the vacuum) and the related specific
concept of electromagnetic field lines or lines of force.l.u Not surprisingly, this was
already a subject of debate in the Greek culture: Aristotle thought that "Nature abhors a
vacuum." Much later, Newton believed in a continuous medium filling all space, but his
equations did not in fact require it. In the 19th century. Faraday and then Maxwell revived
the idea that space was filled with a medium having physical properties in itself, but again
their equations could stand by themselves. The advent of special relativity at the very
beginning of the 20th century started to change these ideas, but initially the
Faraday-Maxwell concept of elecuic and magnetic fields as media filling all space was still
retained. However, the concept of Ether in classical electromagnetic theory was then
gradually abandoned. or just forgotten, because of its uselessness. whereas it continued to
evolve and to motivate important laws of physics via general relativity and quantum field
theory.
126 CHAPTER 2 MAGNETIC POTENTIALS

Magnetic field lines

The lines offorce of a magnetic field (i.e., the field lines) provide an intuitive and
useful view of the magnetic vector field H . which can be defined in space, for example,
through its Cartesian components Hx ' Hy . Hl. ' Magnetic field lines are lines in space that
are everywhere tangenr to the magnetic field (figure 2.5- 1). They are therefore described, in

Cartesian coordinates, by the equations

coost. , (2.5- 1 )

or, in cylindrical coordinates, by

= const. (2.5-2)

Note that these relations can also be written in vector fonn as H = const.dr, or H X dr = 0,
where r is the position vector of point P from the origin. and dr is thus a differential vector
tangent to the field lines.
It is convenient to introduce the arc length s as an indepen(Jent variable, such that the
field lines in space are described by the s-dependent functions

x = x(s). y = y(s). z = z(s) (2.5-3)

!ft-
These functions are clearly defined by (2.5-1), where now

� = cosax = = h�(s)
dy
J
!#-
H
ill = cosu = = hy (s) (2.5-4)
y

* = cosaz = = hz(S)

(2.5-5)

and the initial conditions are

(2.5-6)

#-
Note that with (2.5-4) we can also write

ff-, = = ff- 1: = = const.

, ,

Fonnally. knowing the field lines (2.5-3) and the field magnitude H = IHI. (2.5-5). IS

uniquely equivalent to knowing the vectorfield H. In practice, there is quite a difference;