EMFT Book

Draft version released 24th June 2012 at 12:20 CETDownloaded from http://www.plasma.uu.se/CED/Book Sheet: 1 of 284 .
Bo
id
FT
R A
FT
ELECTROMAGNETIC FIELD THEORY

S E
R A
FT
R A
FT
ELECTROMAGNETIC FIELD THEORY

S E
Bo Thid
Swedish Institute of Space Physics Uppsala, Sweden and
Department of Physics and Astronomy Uppsala University, Sweden and Galilean School of Higher Education University of Padua Padua, Italy
R A
FT
Also available
ELECTROMAGNETIC FIELD THEORY EXERCISES

by Tobia Carozzi, Anders Eriksson, Bengt Lundborg, Bo Thid and Mattias Waldenvik
A This book was typeset in L TEX 2" based on TEX 3.1415926 and Web2C 7.5.6
Copyright 19972009 by Bo Thid Uppsala, Sweden All rights reserved.
Electromagnetic Field Theory ISBN 978-0-486-4773-2
D
The cover graphics illustrates the linear momentum radiation pattern of a radio beam endowed with orbital angular momentum, generated by an array of tri-axial antennas. This graphics illustration was prepared by J O H A N S J H O L M and K R I S T O F F E R PA L M E R as part of their undergraduate Diploma Thesis work in Engineering Physics at Uppsala University 20062007.
R A
FT
Freely downloadable from www.plasma.uu.se/CED
To the memory of professor L E V M I K H A I L OV I C H E RU K H I M OV (19361997) dear friend, great physicist, poet and a truly remarkable man.
R A
FT
R A
FT
CONTENTS
Contents List of Figures Preface to the second edition Preface to the rst edition 1 Foundations of Classical Electrodynamics
1.1 Electrostatics . . . . . . . . . . . . . . . . . . 1.1.1 Coulombs law . . . . . . . . . . . . . 1.1.2 The electrostatic eld . . . . . . . . . . 1.2 Magnetostatics . . . . . . . . . . . . . . . . . 1.2.1 Ampres law . . . . . . . . . . . . . . 1.2.2 The magnetostatic eld . . . . . . . . . 1.3 Electrodynamics . . . . . . . . . . . . . . . . . 1.3.1 The indestructibility of electric charge . 1.3.2 Maxwells displacement current . . . . 1.3.3 Electromotive force . . . . . . . . . . . 1.3.4 Faradays law of induction . . . . . . . 1.3.5 The microscopic Maxwell equations . . 1.3.6 Diracs symmetrised Maxwell equations 1.4 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix xv xvii xix 1 2 2 3 6 6 7 9 10 10 11 12 15 16 18 21 21 22 23 24 27 30 34 35 35 36 39 40 41
2 Electromagnetic Fields and Waves
D
3.1 3.2 3.3 3.4
2.1 Axiomatic classical electrodynamics . . . . . . . . . . . . 2.2 Complex notation and physical observables . . . . . . . . 2.2.1 Physical observables and averages . . . . . . . . . 2.2.2 Maxwell equations in Majorana representation . . . 2.3 The wave equations for E and B . . . . . . . . . . . . . . 2.3.1 The time-independent wave equations for E and B 2.4 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . .
R A
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
3 Electromagnetic Potentials and Gauges

The electrostatic scalar potential . The magnetostatic vector potential The electrodynamic potentials . . Gauge conditions . . . . . . . . . 3.4.1 Lorenz-Lorentz gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
FT
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CONTENTS
3.4.2 Coulomb gauge 3.4.3 Velocity gauge 3.5 Gauge transformations 3.5.1 Other gauges . 3.6 Bibliography . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
44 46 46 51 53 55 55 57 60 61 61 62 66 71 79 79 80 81 82 84 87 90 94 95 96 96 97 98 99 99 101 106 109 110 111 111 118 119
4 Fundamental Properties of the Electromagnetic Field

4.1 Charge, space, and time inversion symmetries 4.2 Electromagnetic duality . . . . . . . . . . . . 4.3 Conservation laws . . . . . . . . . . . . . . . 4.3.1 Conservation of electric charge . . . . 4.3.2 Conservation of total electric current . 4.3.3 Conservation of energy . . . . . . . . 4.3.4 Conservation of linear momentum . . 4.3.5 Conservation of angular momentum . 4.3.6 Conservation of centre of energy . . . 4.3.7 Electromagnetic virial theorem . . . . 4.4 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 Fields from Arbitrary Charge and Current Distributions
6 Radiation and Radiating Systems
6.1 Radiation of linear momentum and energy . . . . . . . . . . 6.1.1 Monochromatic signals . . . . . . . . . . . . . . . . 6.1.2 Finite bandwidth signals . . . . . . . . . . . . . . . 6.2 Radiation of angular momentum . . . . . . . . . . . . . . . 6.3 Radiation from a localised source volume at rest . . . . . . . 6.3.1 Electric multipole moments . . . . . . . . . . . . . . 6.3.2 The Hertz potential . . . . . . . . . . . . . . . . . . 6.3.3 Electric dipole radiation . . . . . . . . . . . . . . . . 6.3.4 Magnetic dipole radiation . . . . . . . . . . . . . . . 6.3.5 Electric quadrupole radiation . . . . . . . . . . . . . 6.4 Radiation from an extended source volume at rest . . . . . . 6.4.1 Radiation from a one-dimensional current distribution 6.5 Radiation from a localised charge in arbitrary motion . . . . 6.5.1 The Linard-Wiechert potentials . . . . . . . . . . .
R A
5.1 5.2 5.3 5.4
Fourier component method . . . . . . . . . . . . The retarded electric eld . . . . . . . . . . . . . The retarded magnetic eld . . . . . . . . . . . . The total elds at large distances from the sources 5.4.1 The far elds . . . . . . . . . . . . . . .
FT
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . .
. . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
CONTENTS
xi
6.5.2 Radiation from an accelerated point charge 6.5.3 Bremsstrahlung . . . . . . . . . . . . . . . 6.5.4 Cyclotron and synchrotron radiation . . . . 6.6 Bibliography . . . . . . . . . . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
121 131 136 144 145 145 146 147 152 155 156 157 158 160 163 165 165 165 171 172 180 181 182 182 183 185 186 187 187 192 193 195 203 205 205 205 206 206
7 Relativistic Electrodynamics
7.1 The special theory of relativity . . . . . 7.1.1 The Lorentz transformation . . . 7.1.2 Lorentz space . . . . . . . . . . 7.1.3 Minkowski space . . . . . . . . 7.2 Covariant classical mechanics . . . . . . 7.3 Covariant classical electrodynamics . . 7.3.1 The four-potential . . . . . . . . 7.3.2 The Linard-Wiechert potentials 7.3.3 The electromagnetic eld tensor 7.4 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8 Electromagnetic Fields and Particles
9 Electromagnetic Fields and Matter
D
F
9.1 Maxwells macroscopic theory . . . . . . . . . . . . . 9.1.1 Polarisation and electric displacement . . . . . 9.1.2 Magnetisation and the magnetising eld . . . . 9.1.3 Macroscopic Maxwell equations . . . . . . . . 9.2 Phase velocity, group velocity and dispersion . . . . . 9.3 Radiation from charges in a material medium . . . . . L 9.3.1 Vavilov-Cerenkov radiation . . . . . . . . . . . 9.4 Electromagnetic waves in a medium . . . . . . . . . . 9.4.1 Constitutive relations . . . . . . . . . . . . . . 9.4.2 Electromagnetic waves in a conducting medium 9.5 Bibliography . . . . . . . . . . . . . . . . . . . . . . .
R A
8.1 Charged particles in an electromagnetic eld . . . . . . . . . . . 8.1.1 Covariant equations of motion . . . . . . . . . . . . . . 8.2 Covariant eld theory . . . . . . . . . . . . . . . . . . . . . . . 8.2.1 Lagrange-Hamilton formalism for elds and interactions 8.3 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Formul
F .1
The electromagnetic eld . . . . . . . . . . . . . . . . . . . . . . . . . . . . F .1.1 Microscopic Maxwell-Lorentz equations in Diracs symmetrised form F .1.2 Fields and potentials . . . . . . . . . . . . . . . . . . . . . . . . . . F .1.3 Energy and momentum . . . . . . . . . . . . . . . . . . . . . . . . .
FT
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xii
CONTENTS
F .2
Mathematical Methods
M .1
R A
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Electromagnetic radiation . . . . . . . . . . . . . . . . . . . . F .2.1 The far elds from an extended source distribution . . F .2.2 The far elds from an electric dipole . . . . . . . . . . F .2.3 The far elds from a magnetic dipole . . . . . . . . . . F .2.4 The far elds from an electric quadrupole . . . . . . . F .2.5 Relationship between the eld vectors in a plane wave F .2.6 The elds from a point charge in arbitrary motion . . . F .3 Special relativity . . . . . . . . . . . . . . . . . . . . . . . . . F .3.1 Metric tensor for at 4D space . . . . . . . . . . . . . F .3.2 Covariant and contravariant four-vectors . . . . . . . . F .3.3 Lorentz transformation of a four-vector . . . . . . . . F .3.4 Invariant line element . . . . . . . . . . . . . . . . . . F .3.5 Four-velocity . . . . . . . . . . . . . . . . . . . . . . F .3.6 Four-momentum . . . . . . . . . . . . . . . . . . . . F .3.7 Four-current density . . . . . . . . . . . . . . . . . . . F .3.8 Four-potential . . . . . . . . . . . . . . . . . . . . . . F .3.9 Field tensor . . . . . . . . . . . . . . . . . . . . . . . F .4 Vectors and tensor elds in 3D . . . . . . . . . . . . . . . . . F .4.1 Cylindrical circular coordinates . . . . . . . . . . . . . F .4.2 Spherical polar coordinates . . . . . . . . . . . . . . . F .4.3 Vector and tensor eld formul . . . . . . . . . . . . . F .5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
207 207 207 207 208 208 208 209 209 209 209 209 209 209 210 210 210 210 211 212 214 220 221 222 223 224 233 233 235 236 239 239 240 241 241 241 245 248 252 252
Scalars, vectors and tensors . . . . . . . M .1.1 Vectors . . . . . . . . . . . . . M .1.2 Fields . . . . . . . . . . . . . . M .2 Vector algebra . . . . . . . . . . . . . . M .2.1 Scalar product . . . . . . . . . . M .2.2 Vector product . . . . . . . . . . M .2.3 Dyadic product . . . . . . . . . M .3 Vector calculus . . . . . . . . . . . . . M .3.1 The del operator . . . . . . . . . M .3.2 The gradient of a scalar eld . . M .3.3 The divergence of a vector eld M .3.4 The curl of a vector eld . . . . M .3.5 The Laplacian . . . . . . . . . . M .3.6 Vector and tensor integrals . . . M .3.7 Helmholtzs theorem . . . . . . M .4 Analytical mechanics . . . . . . . . . . M .4.1 Lagranges equations . . . . . .
FT
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
CONTENTS
xiii
Hamiltons equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 M .5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
M .4.2
Index
255
R A
FT
R A
FT
LIST OF FIGURES
1.1 1.2 1.3 1.4
Coulomb interaction between two electric charges . . . . Coulomb interaction for a distribution of electric charges Ampre interaction . . . . . . . . . . . . . . . . . . . . Moving loop in a varying B eld . . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
3 5 7 13 90 103 105 112 113 116 119 121 131 133 137 139 141 143
5.1 Fields in the far zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 6.11 6.12 6.13 Multipole radiation geometry . . . . . . . . . . . . . . Electric dipole geometry . . . . . . . . . . . . . . . . Linear antenna . . . . . . . . . . . . . . . . . . . . . . Electric dipole antenna geometry . . . . . . . . . . . . Loop antenna . . . . . . . . . . . . . . . . . . . . . . Radiation from a moving charge in vacuum . . . . . . An accelerated charge in vacuum . . . . . . . . . . . . Angular distribution of radiation during bremsstrahlung Location of radiation during bremsstrahlung . . . . . . Radiation from a charge in circular motion . . . . . . . Synchrotron radiation lobe width . . . . . . . . . . . . The perpendicular electric eld of a moving charge . . Electron-electron scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1 Relative motion of two inertial systems . . . . . . . . . . . . . . . . . . . . . . . . . 146 7.2 Rotation in a 2D Euclidean space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 7.3 Minkowski diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154 8.1 Linear one-dimensional mass chain . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 L 9.1 Vavilov-Cerenkov cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
M .1
Tetrahedron-like volume element of matter . . . . . . . . . . . . . . . . . . . . . . . 229
D
xv
R A
FT
R A
FT
PREFACE TO THE SECOND EDITION

This second edition of the book E L E C T RO M AG N E T I C F I E L D T H E O RY is a major revision and expansion of the rst edition that was published on the Internet (www.plasma.uu.se/CED/ Book) in an organic growth process over the years 19972008. The main changes are an expansion of the material treated, an addition of a new chapter and several illustrative examples, and a slight reordering of the chapters. The main reason for attempting to improve the presentation and to add more material is that this new edition is now being made available in printed form by Dover Publications and is used in an extended Classical Electrodynamics course at Uppsala University, at the last-year undergraduate, Master, and beginning post-graduate/doctoral level. It has also been used by the author in a similar course at the Galilean School of Higher Education (Scuola Galileiana di Studi Superiori) at University of Padova. It is the authors hope that the second edition of his book will nd a wid use in Academia and elsewhere. The subject matter starts with a description of the properties of electromagnetism when the charges and currents are located in otherwise free space, i.e., a space that is free of matter and external elds (e.g., gravitation). A rigorous analysis of the fundamental properties of the electromagnetic elds and radiation phenomena follows. Only then the inuence of matter on the elds and the pertinent interaction processes is taken into account. In the authors opinion, this approach is preferable since it avoids the formal logical inconsistency of introducing, very early in the derivations, the effect on the electric and mangetic elds when conductors and dielectrics are present (and vice versa) in an ad hoc manner, before constitutive relations and physical models for the electromagnetic properties of matter, including conductors and dielectrics, have been derived from rst principles. Curved-space effects on electromagnetism are not treated at all. In addition to the Maxwell-Lorentz equations, which postulate the beaviour of electromagnetic elds due to electric charges and currents on a microscopic classical scale, chapter 1 on page 1 also introduces Diracs symmetrised equations that incorporate the effects of magnetic charges and currents. In chapter 2 on page 21, a stronger emphasis than before is put on the axiomatic foundation of electrodynamics as provided by the Maxwell-Lorentz equations that are taken as the postulates of the theory. Chapter 3 on page 35 on potentials and gauges now provides a more comprehensive picture and discusses gauge invariance in a more satisfactory manner than the rst edition did. Chapter 4 on page 55 is new and deals with symmetries and conserved quantities in a more rigourous, profound and detailed way than in the rst edition. For instance, the presentation of the theory of electromagnetic angular momentum and other observables (constants of motion) has been substantially expanded and put on a rm basis. Chapter 9 on page 181 is a complete rewrite and combines material that was scattered more or less all over the rst edition. It also contains new material on wave propagation in plasma and other media. When, in chapter 9 on page 181, the macroscopic Maxwell equations are introduced, the inherent approximations in the derived eld quantities are clearly pointed out. The collection of formul in appendix F on page 205 has been augmented quite substantially. In appendix M on page 221, the treatment of dyadic products and tensors has been expanded signicantly and numerous examples
R A
xvii
FT
xviii
P R E FAC E TO THE SECOND EDITION
R A
have been added throughout. I want to express my warm gratitude to professor C E S A R E B A R B I E R I and his entire group, particularly F A B R I Z I O TA M B U R I N I , at the Department of Astronomy, University of Padova, for stimulating discussions and the generous hospitality bestowed upon me during several shorter and longer visits in 2008, 2009, and 2010 that made it possible to prepare the current major revision of the book. In this breathtakingly beautiful northern Italy where the cradle of renaissance once stood, intellectual titan G A L I L E O G A L I L E I worked for eighteen years and gave birth to modern physics, astronomy and science as we know it today, by sweeping away Aristotelian dogmas, misconceptions and mere superstition, thus most profoundly changing our conception of the world and our place in it. In the process, Galileos new ideas transformed society and mankind irreversibly and changed our view of the Universe, including our own planet, forever. It is hoped that this book may contribute in some small, humble way to further these, once upon a time, mind-bogglingand dangerousideas of intellectual freedom and enlightment. Thanks are also due to J O H A N S J H O L M , K R I S T O F F E R PA L M E R , M A R C U S E R I K S S O N , and J O H A N L I N D B E R G who during their work on their Diploma theses suggested improvements and additions and to H O L G E R T H E N and S TA F F A N Y N G V E for carefully checking some lengthy calculations and to the numerous undergraduate students, who have been exposed to various draft versions of this second edtion. In particular, I would like to mention B RU N O S T R A N D B E R G. This book is dedicated to my son M AT T I A S , my daughter K A RO L I NA , my four grandsons M A X , A L B I N , F I L I P and O S K A R , my high-school physics teacher, S TA F F A N R S B Y , and my fellow members of the C A P E L L A P E DAG O G I C A U P S A L I E N S I S . Padova, Italy February, 2011
FT
BO THID www.physics.irfu.se/ bt
PREFACE TO THE FIRST EDITION

Of the four known fundamental interactions in naturegravitational, strong, weak, and electromagneticthe latter has a special standing in the physical sciences. Not only does it, together with gravitation, permanently make itself known to all of us in our everyday lives. Electrodynamics is also by far the most accurate physical theory known, tested on scales running from sub-nuclear to galactic, and electromagnetic eld theory is the prototype of all other eld theories. This book, E L E C T RO M AG N E T I C F I E L D T H E O RY , which tries to give a modern view of classical electrodynamics, is the result of a more than thirty-ve year long love affair. In the autumn of 1972, I took my rst advanced course in electrodynamics at the Department of Theoretical Physics, Uppsala University. Soon I joined the research group there and took on the task of helping the late professor P E R O L O F F R M A N , who was to become my Ph.D. thesis adviser, with the preparation of a new version of his lecture notes on the Theory of Electricity. This opened my eyes to the beauty and intricacy of electrodynamics and I simply became intrigued by it. The teaching of a course in Classical Electrodynamics at Uppsala University, some twenty odd years after I experienced the rst encounter with the subject, provided the incentive and impetus to write this book. Intended primarily as a textbook for physics and engineering students at the advanced undergraduate or beginning graduate level, it is hoped that the present book will be useful for research workers too. It aims at providing a thorough treatment of the theory of electrodynamics, mainly from a classical eld-theoretical point of view. The rst chapter is, by and large, a description of how Classical Electrodynamics was established by J A M E S C L E R K M A X W E L L as a fundamental theory of nature. It does so by introducing electrostatics and magnetostatics and demonstrating how they can be unied into one theory, classical electrodynamics, summarised in Lorentzs microscopic formulation of the Maxwell equations. These equations are used as an axiomatic foundation for the treatment in the remainder of the book, which includes modern formulation of the theory; electromagnetic waves and their propagation; electromagnetic potentials and gauge transformations; analysis of symmetries and conservation laws describing the electromagnetic counterparts of the classical concepts of force, momentum and energy, plus other fundamental properties of the electromagnetic eld; radiation phenomena; and covariant Lagrangian/Hamiltonian eld-theoretical methods for electromagnetic elds, particles and interactions. Emphasis has been put on modern electrodynamics concepts while the mathematical tools used, some of them presented in an Appendix, are essentially the same kind of vector and tensor analysis methods that are used in intermediate level textbooks on electromagnetics but perhaps a bit more advanced and far-reaching. The aim has been to write a book that can serve both as an advanced text in Classical Electrodynamics and as a preparation for studies in Quantum Electrodynamics and Field Theory, as well as more applied subjects such as Plasma Physics, Astrophysics, Condensed Matter Physics, Optics, Antenna Engineering, and Wireless Communications. The current version of the book is a major revision of an earlier version, which in turn was an
R A
xix
FT
xx
P R E FACE TO THE FIRST EDITION
D
Uppsala, Sweden December, 2008
outgrowth of the lecture notes that the author prepared for the four-credit course Electrodynamics that was introduced in the Uppsala University curriculum in 1992, to become the ve-credit course Classical Electrodynamics in 1997. To some extent, parts of those notes were based on lecture notes prepared, in Swedish, by my friend and Theoretical Physics colleague B E N G T L U N D B O R G , who created, developed and taught an earlier, two-credit course called Electromagnetic Radiation at our faculty. Thanks are due not only to Bengt Lundborg for providing the inspiration to write this book, but also to professor C H R I S T E R WA H L B E R G , and professor G R A N F L D T , both at the Department of Physics and Astronomy, Uppsala University, for insightful suggestions, to professor J O H N L E A R N E D , Department of Physics and Astronomy, University of Hawaii, for decisive encouragement at the early stage of this book project, to professor G E R A R D U S T H O O F T , for recommending this book on his web page How to become a good theoretical physicist, and professor C E C I L I A J A R L S KO G , Lund Unversity, for pointing out a couple of errors and ambiguities. I am particularly indebted to the late professor V I TA L I Y L A Z A R E V I C H G I N Z B U R G , for his many fascinating and very elucidating lectures, comments and historical notes on plasma physics, electromagnetic radiation and cosmic electrodynamics while cruising up and down the Volga and Oka rivers in Russia at the ship-borne Russian-Swedish summer schools that were organised jointly by late professor L E V M I K A H I L OV I C H E RU K H I M OV and the author during the 1990s, and for numerous deep discussions over the years. Helpful comments and suggestions for improvement from former PhD students T O B I A C A RO Z Z I , R O G E R K A R L S S O N , and M AT T I A S WA L D E N V I K , as well as A N D E R S E R I K S S O N at the Swedish Institute of Space Physics in Uppsala and who have all taught Uppsala students on the material covered in this book, are gratefully acknowledged. Thanks are also due to the late H E L M U T K O P K A , for more than twenty-ve years a close friend and space physics colleague working at the Max-Planck-Institut fr Aeronomie, Lindau, Germany, who not only taught me the practical aspects of the use of high-power electromagnetic radiation for studying space, but A also some of the delicate aspects of typesetting in TEX and LTEX. In an attempt to encourage the involvement of other scientists and students in the making of this book, thereby trying to ensure its quality and scope to make it useful in higher university education anywhere in the world, it was produced as a World-Wide Web (WWW) project. This turned out to be a rather successful move. By making an electronic version of the book freely downloadable on the Internet, comments have been received from fellow physicists around the world. To judge from WWW hit statistics, it seems that the book serves as a frequently used Internet resource. This way it is hoped that it will be particularly useful for students and researchers working under nancial or other circumstances that make it difcult to procure a printed copy of the book. I would like to thank all students and Internet users who have downloaded and commented on the book during its life on the World-Wide Web.
R A
FT
BO THID www.physics.irfu.se/ bt
FOUNDATIONS OF CLASSICAL ELECTRODYNAMICS

The classical theory of electromagnetism deals with electric and magnetic elds and their interactions with each other and with charges and currents. This theory is classical in the sense that it assumes the validity of certain mathematical limiting processes in which it is considered possible for the charge and current distributions to be localised in innitesimally small volumes of space.1 Clearly, this is in contradistinction to electromagnetism on an atomistic scale, where charges and currents have to be described in a nonlocal quantum formalism. However, the limiting processes used in the classical domain, which, crudely speaking, assume that an elementary charge has a continuous distribution of charge density, will yield results that agree perfectly with experiments on non-atomistic scales, however small or large these scales may be.2 It took the genius of J A M E S C L E R K M A X W E L L to consistently unify, in the mid-1800s, the theory of Electricity and the then distinctively different theory Magnetism into a single super-theory, Electromagnetism or Classical Electrodynamics (CED), and also to realise that optics is a sub-eld of this supertheory. Early in the 20th century, H E N D R I K A N T O O N L O R E N T Z took the electrodynamics theory further to the microscopic scale and also paved the way for the Special Theory of Relativity, formulated in its full extent by A L B E RT E I N S T E I N in 1905. In the 1930s PAU L A D R I E N M AU R I C E D I R AC expanded electrodynamics to a more symmetric form, including magnetic as well as electric charges. With his relativistic quantum mechanics and eld quantisation concepts, Dirac had already in the 1920s laid the foundation for Quantum Electrodynamics (QED ), the relativistic quantum theory for electromagnetic elds and their interaction with matter for which R I C H A R D P H I L L I P S F E Y N M A N , J U L I A N S E Y M O U R S C H W I N G E R , and S I N -I T I RO T O M O NAG A were awarded the Nobel Prize in Physics in 1965. Around the same time, physicists such as S H E L D O N G L A S H OW , A B D U S S A L A M , and S T E V E N W E I N B E R G were able to unify electrodynamics with the weak interaction theory, thus creating yet another successful super-theory, Electroweak Theory, an achievement which rendered them the Nobel Prize in Physics 1979. The modern theory of strong interactions, Quantum Chromodynamics (QCD ), is heavily inuenced by CED and QED. In this introductory chapter we start with the force interactions in classical
R A
1
FT
1 2
Accepting the mere existence of an electrically charged particle requires some careful thinking. In his excellent book Classical Charged Particles, F R I T Z R O H R L I C H writes To what extent does it make sense to talk about an electron, say, in classical terms? These and similar questions clearly indicate that ignoring philosophy in physics means not understanding physics. For there is no theoretical physics without some philosophy; not admitting this fact would be selfdeception.
Electrodynamics has been tested experimentally over a larger range of spatial scales than any other existing physical theory.
2j
1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS
1.1 Electrostatics
The physicist, mathematician and philosopher P I E R R E M AU R I C E M A R I E D U H E M (18611916) once wrote: The whole theory of electrostatics constitutes a group of abstract ideas and general propositions, formulated in the clear and concise language of geometry and algebra, and connected with one another by the rules of strict logic. This whole fully satises the reason of a French physicist and his taste for clarity, simplicity and order. . . .
4 3
C H A R L E S - AU G U S T I N D E C O U L O M B (17361806) was a French physicist who in 1775 published three reports on the forces between electrically charged bodies.
R A
1.1.1 Coulombs law
The theory which describes physical phenomena related to the interaction between stationary electric charges or charge distributions in a nite space with stationary boundaries is called electrostatics. For a long time, electrostatics, under the name electricity, was considered an independent physical theory of its own, alongside other physical theories such as Magnetism, Mechanics, Optics, and Thermodynamics.3
It has been found experimentally that the force interaction between stationary, electrically charged bodies can be described in terms of two-body mechanical forces. Based on experimental observations, C H A R L E S -A U G U S T I N E D E C O U L O M B postulated, in 1775, that in the simple case depicted in gure 1.1 on the facing page, the mechanical force on a static electric charge q located at the observation point x, due to the presence of another stationary electric charge q 0 at the source point x0 , is directed along the line connecting these two points, repulsive for charges of equal signs and attractive for charges of opposite signs. This postulate is called Coulombs law 4 and can be formulated mathematically as qq 0 1 qq 0 0 1 qq 0 x x0 F es .x/ D D r D r 4 "0 jx x0 j3 4 "0 4 "0 jx x0 j jx x0 j (1.1)
FT
electrostatics and classical magnetostatics and introduce the static electric and magnetic elds to postulate two uncoupled systems of differential equations for them. Then we see how the conservation of charge and its relation to current leads to the dynamic connection between electricity and magnetism and how the two can be unied into classical electrodynamics. This theory is described by a system of coupled dynamic eld equationsthe microscopic Maxwell differential equations introduced by Lorentzwhich, in chapter 2, we take as the axiomatic foundation for the theory of electromagnetic elds. At the end of this chapter 1 we present Diracs symmetrised form of the Maxwell-Lorentz equations that incorporate magnetic charges and magnetic currents into the theory in a most elegant way. In practical work, such as in antenna engineering, magnetic currents have proved to be a very useful concept. We shall make use of this symmetrised theory of electricity and magnetism, sometimes called Electromagnetodynamics, throughout the book.
1.1. Electrostatics
j3
q x x
x0
q0 x0
Figure 1.1: Coulombs law postulates that a static electric charge q, located at a point x relative to the origin O, will experience an electrostatic force F es .x/ from a static electric charge q 0 located at x0 . Note that this denition is independent of any particular choice of coordinate system since the mechanical force F es is a true (polar) vector.
Instead of describing the electrostatic interaction in terms of a force action at a distance, it turns out that for many purposes it is useful to introduce the concept of a eld. Thus we describe the electrostatic interaction in terms of a static vectorial electric eld Estat dened by the limiting process Estat .x/ lim
def
R A
F es .x/ q!0 q
1.1.2 The electrostatic eld
where F es is the electrostatic force, as dened in equation (1.1) on the facing page, from a net electric charge q 0 on the test particle with a small net electric charge q. In the SI system of units, electric elds are therefore measured in NC 1 or, equivalently, in Vm 1 . Since the purpose of the limiting process is to ascertain that the test charge q does not distort the eld set up by q 0 , the expression for Estat does not depend explicitly on q but only on the charge q 0 and the relative position vector x x0 . This means that we can say that any net electric charge produces an electric eld in the space that surrounds it, regardless of the existence of a second charge anywhere in this space.7 However, in order to experimentally detect a charge, a second (test) charge that senses the presence of the rst one, must be introduced.
FT
5 6
where, in the last step, formula (F.169) on page 218 was used. In SI units, which we shall use throughout, the electrostatic force5 F es is measured in Newton (N), the electric charges q and q 0 in Coulomb (C), i.e. Ampere-seconds (As), and the length jx x0 j in metres (m). The constant "0 D 107 =.4 c 2 / 8:8542 10 12 Farad per metre (Fm 1 ) is the vacuum permittivity and c 2:9979 108 ms 1 is the speed of light in vacuum.6 In CGS units, "0 D 1=.4 / and the force is measured in dyne, electric charge in statcoulomb, and length in centimetres (cm).
Massive particles also interact gravitationally but but with a much weaker force.
The notation c for speed stems from the Latin word celeritas which means swiftness. This notation seems to have been introduced by W I L H E L M E D UA R D W E B E R (18041891), and R U D O L F K O H L R AU S C H (1809 1858) and c is therefore sometimes referred to as Webers constant. In all his works from 1907 and onward, A L B E RT E I N S T E I N (18791955) used c to denote the speed of light in vacuum.
(1.2)
In the preface to the rst edition of the rst volume of his book A Treatise on Electricity and Magnetism, rst published in 1873, Maxwell describes this in the following almost poetic manner: For instance, Faraday, in his minds eye, saw lines of force traversing all space where the mathematicians saw centres of force attracting at a distance: Faraday saw a medium where they saw nothing but distance: Faraday sought the seat of the phenomena in real actions going on in the medium, they were satised that they had found it in a power
4j
Using equation (1.1) on page 2 and equation (1.2) on the previous page, and formula (F.168) on page 218, we nd that the electrostatic eld Estat at the observation point x (also known as the eld point), due to a eld-producing electric charge q 0 at the source point x0 , is given by q0 q 0 x x0 q0 1 1 r D r0 Estat .x/ D D 4 "0 jx x0 j3 4 "0 4 "0 jx x0 j jx x0 j (1.3) In the presence of several eld producing discrete electric charges qi0 , located at the points x0 , i D 1; 2; 3; : : : , respectively, in an otherwise empty space, the i assumption of linearity of vacuum8 allows us to superimpose their individual electrostatic elds into a total electrostatic eld Estat .x/ D
In fact, vacuum exhibits a quantum mechanical non-linearity due to vacuum polarisation effects manifesting themselves in the momentary creation and annihilation of electron-positron pairs, but classically this nonlinearity is negligible.
8
FT
1 X 0 x x0 i qi 4 "0 x x0 3 i i 1 x x0 dq 0 4 "0 jx x0 j3 1 x x0 d3x 0 .x0 / 4 "0 jx x0 j3
(1.4)
If the discrete electric charges are small and numerous enough, we can, in a continuum limit, assume that the total charge q 0 from an extended volume to be built up by local innitesimal elemental charges dq 0 , each producing an elemental electric eld dEstat .x/ D (1.5)
R A
dEstat .x/ D
By introducing the electric charge density , measured in Cm 3 in SI units, at 0 0 0 the point x0 within the volume element d3x 0 D dx1 dx2 dx3 (measured in m3 ), 0 3 0 0 the elemental charge can be expressed as dq D d x .x /, and the elemental electrostatic eld as (1.6)
Integrating this over the entire source volume V 0 , we obtain Z Z 1 x x0 stat stat E .x/ D dE .x/ D d3x 0 .x0 / 4 "0 V 0 jx x0 j3 Z 1 1 D d3x 0 .x0 /r 4 "0 V 0 jx x0 j Z 0 1 .x / D r d3x 0 4 "0 jx x0 j V0
(1.7)
where we used formula (F.168) on page 218 and the fact that .x0 / does not depend on the unprimed (eld point) coordinates on which r operates. We emphasise that under the assumption of linear superposition, equation (1.7) above is valid for an arbitrary distribution of electric charges, including
1.1. Electrostatics
j5
q x
x0 i
qi0 x0 i
Figure 1.2: Coulombs law for a continuous charge density .x0 / within a volume V 0 of limited extent. In particular, a charge denPN 0 0 sity .x0 / D x0 / i qi .x i would represent a source distribution consisting of N discrete 0 charges qi located at x0 , where i i D 1; 2; 3; : : : ; N .
V0
discrete charges, in which case tions:9
is expressed in terms of Dirac delta distribuX

i
.x0 / D
qi0 .x0
as illustrated in gure 1.2. Inserting this expression into expression (1.7) on the facing page we recover expression (1.4) on the preceding page, as we should. According to Helmholtzs theorem, discussed in subsection M.3.7, any wellbehaved vector eld is completely known once we know its divergence and curl at all points x in 3D space.10 Taking the divergence of the general Estat expression for an arbitrary electric charge distribution, equation (1.7) on the facing page, and using the representation of the Dirac delta distribution given by formula (F.171) on page 218, one nds that Z x x0 1 stat d3x 0 .x0 / r E .x/ D r 4 "0 V 0 jx x0 j3 Z 1 1 3 0 0 D d x .x /r r 4 "0 V 0 jx x0 j (1.9) Z 1 1 3 0 0 2 D d x .x / r 4 "0 V 0 jx x0 j Z 1 .x/ D d3x 0 .x0 / .x x0 / D "0 V 0 "0
FT
9
x0 / i
(1.8)
Since, by denition, the integral Z d3x 0 .x0 x0 / i V0 Z 0 d3x 0 .x 0 xi /

V0
.y 0
0 yi /.z 0
0 zi / D 1
which is the differential form of Gausss law of electrostatics. Since, according to formula (F.155) on page 217, r r .x/ 0 for any R3 scalar eld .x/, we immediately nd that in electrostatics Z 1 .x0 / stat 3 0 r E .x/ D r r dx D0 (1.10) 4 "0 jx x0 j V0 i.e. that Estat is an irrotational eld.
R A
is dimensionless, and x has the SI dimension m, the 3D Dirac delta distribution .x0 x0 / must have i 10 H E R M A N N L U DW . G F E R the SI dimension m 3 I D I NA N D VO N H E L M H O LT Z (18211894) was a physicist, physician and philosopher who contributed to wide areas of science, ranging from electrodynamics to ophthalmology.
6j
To summarise, electrostatics can be described in terms of two vector partial differential equations r Estat .x/ D r .x/ "0 stat E .x/ D 0 (1.11a) (1.11b)
representing four scalar partial differential equations.
1.2 Magnetostatics
1.2.1 Ampres law
11 A N D R - M A R I E A M P R E (17751836) was a French mathematician and physicist who, only a few days after he learned about the ndings by the Danish physicist and chemist H A N S C H R I S T I A N R S T E D (17771851) regarding the magnetic effects of electric currents, presented a paper to the Acadmie des Sciences in Paris, postulating the force law that now bears his name.
Experiments on the force interaction between two small loops that carry static electric currents I and I 0 (i.e. the currents I and I 0 do not vary in time) have shown that the loops interact via a mechanical force, much the same way that static electric charges interact. Let F ms .x/ denote the magnetostatic force on a loop C , with tangential line vector element dl, located at x and carrying a current I in the direction of dl, due to the presence of a loop C 0 , with tangential line element dl0 , located at x0 and carrying a current I 0 in the direction of dl0 in otherwise empty space. This spatial conguration is illustrated in graphical form in gure 1.3 on the next page. According to Ampres law the magnetostatic force in question is given by the expression11 I 0 I x x0 0 II ms 0 F .x/ D dl dl 4 jx x0 j3 C C0 (1.12) I I 0 1 0 II 0 D dl dl r 4 jx x0 j C C0
R A
"0
0
In SI units, 0 D 4 10 7 1:2566 10 6 Henry per metre (Hm 1 ) is the vacuum permeability. From the denition of "0 and 0 (in SI units) we observe that D 107 (Fm 4 c2
1
FT
) 4 10
7
Whereas electrostatics deals with static electric charges, i.e. electric charges that do not move, and the interaction between these charges, magnetostatics deals with static electric currents, i.e. electric charges moving with constant speeds, and the interaction between these currents. Here we shall discuss the theory of magnetostatics in some detail.
(Hm
) D
1 2 (s m c2
(1.13)
1.2. Magnetostatics
j7
C I dl x x x0 I 0 dl0 C0 x0
Figure 1.3: Ampres law postulates how a small loop C , carrying a static electric current I directed along the line element dl at x, experiences a magnetostatic force F ms .x/ from a small loop C 0 , carrying a static electric current I 0 directed along the line element dl0 located at x0 .
which is a most useful relation. At rst glance, equation (1.12) on the facing page may appear asymmetric in terms of the loops and therefore be a force law that does not obey Newtons third law. However, by applying the vector triple product bac-cab formula (F.108) on page 216, we can rewrite (1.12) as F ms .x/ D
0 II 0
I
C0
4
0 II 0
dl0 dl
I
C
Since the integrand in the rst integral is an exact differential, this integral vanishes and we can rewrite the force expression, formula (1.12) on the facing page, in the following symmetric way F ms .x/ D
0 II 0
R A
4 dl
0 C C0
dl
which clearly exhibits the expected interchange symmetry between loops C and C 0.
1.2.2 The magnetostatic eld
In analogy with the electrostatic case, we may attribute the magnetostatic interaction to a static vectorial magnetic eld Bstat . The small elemental dBstat .x/ of the static magnetic eld at the eld point x due to a small line current element I 0 dl0 D d3x 0 j 0 .x0 / of static current I 0 with electric current density j measured
FT
dl r 1 jx x0 j x x0 x0 j3 (1.14) jx dl0 x x0
C0
jx
x0 j3
(1.15)
8j
in Am
in SI units, directed along the local line element dl0 of the loop at x0 , is dBstat .x/ lim dF ms .x/ 0 0 0 D I dl I !0 I 4 x x0 0 3 0 0 0 D d x j .x / 4 jx x0 j3
def
x jx
x0 x0 j3 (1.16)
R A D
which is the differential form of the Biot-Savart law. The elemental eld vector dBstat .x/ is perpendicular to the plane spanned by the line element vector dl0 at x0 and the vector x x0 from x0 to the (elemental) eld point x. The corresponding local elemental force dF ms at x is directed perpendicular to the local plane spanned by dBstat .x/ and the eld sensing current element I dl.x/ D d3xj.x/. The SI unit for the magnetic eld, sometimes called the magnetic ux density or magnetic induction, is Tesla (T). If we integrate expression (1.16) above around the entire loop at x, we obtain Z stat B .x/ D dBstat .x/ Z x x0 0 D d3x 0 j.x0 / 4 jx x0 j3 V0 (1.17) Z 1 0 D d3x 0 j.x0 / r 4 jx x0 j V0 Z j.x0 / 0 r d3x 0 D 4 jx x0 j V0 where we used formula (F.168) on page 218, formula (F.141) on page 217, and the fact that j.x0 / does not depend on the unprimed coordinates on which r operates. Comparing equation (1.7) on page 4 with equation (1.17) above, we see that there exists an analogy between the expressions for Estat and Bstat but that they differ in their vectorial characteristics. With this denition of Bstat , equation (1.12) on page 6 may we written I F ms .x/ D I dl Bstat .x/ (1.18)
C
In order to assess the properties of Bstat , we determine its divergence and curl. Taking the divergence of both sides of equation (1.17) above and utilising formula (F.154) on page 217, we obtain Z 0 0 stat 3 0 j.x / r B .x/ D r r dx D0 (1.19) 4 jx x0 j V0 a/ vanishes for any
since, according to formula (F.154) on page 217, r .r vector eld a.x/.
FT
1.3. Electrodynamics
j9
With the use of the operator bac-cab formula (F.151) on page 217, the curl of equation (1.17) on the preceding page can be written Z j.x0 / 0 r Bstat .x/ D r r d3x 0 4 jx x0 j V0 Z 1 0 (1.20) D d3x 0 j.x0 / r 2 4 jx x0 j V0 Z 1 0 3 0 0 0 0 C d x j.x / r r 4 jx x0 j V0 In the rst of the two integrals on the right-hand side, we use the representation of the Dirac delta function given in formula (F.171) on page 218, and integrate the second integral by parts, utilising formula (F.135) on page 217, as follows: Z 1 d3x 0 j.x0 / r 0 r 0 jx x0 j V0 Z 1 @ O D xk d3x 0 r 0 j.x0 / 0 @xk jx x0 j V0 Z 1 3 0 0 0 0 (1.21) d x r j.x / r jx x0 j V0 Z @ 1 O O D xk d2x 0 n0 j.x0 / 0 0 @xk jx x0 j S Z 1 d3x 0 r 0 j.x0 / r 0 jx x0 j V0 We note that the rst integral in the result, obtained by applying Gausss theorem, vanishes when integrated over a large sphere far away from the localised source j.x0 /, and that the second integral vanishes because r j D 0 for stationary currents (no charge accumulation in space). The net result is simply Z r Bstat .x/ D 0 d3x 0 j.x0 /.x x0 / D 0 j.x/ (1.22)
V0
1.3 Electrodynamics
As we saw in the previous sections, the laws of electrostatics and magnetostatics can be summarised in two pairs of time-independent, uncoupled partial differential equations, namely the equations of classical electrostatics r Estat .x/ D r .x/ "0 Estat .x/ D 0 (1.23a) (1.23b)
R A
FT
10 j
and the equations of classical magnetostatics r Bstat .x/ D 0 r B

stat
(1.24a)
0 j.x/
.x/ D
(1.24b)
Since there is nothing a priori which connects Estat directly with Bstat , we must consider classical electrostatics and classical magnetostatics as two separate and mutually independent physical theories. However, when we include time-dependence, these theories are unied into Classical Electrodynamics. This unication of the theories of electricity and magnetism can be inferred from two empirically established facts:
2. A change in the magnetic ux through a loop will induce an electromotive force electric eld in the loop. This is the celebrated Faradays law of induction.
A more accurate model is to assume that the individual charge and current elements obey some distribution function that describes their local variation of velocity in space and time. For instance, j can be dened in statistical mechanical terms as j.t; x/ D R 3 P q d v v f .t; x; v/ where f .t; x; v / is the (normalised) distribution function for particle species carrying an electric charge q .
12
R A
1.3.1 The indestructibility of electric charge
Let j.t; x/ denote the time-dependent electric current density. In the simplest case it can be dened as j D v where v is the velocity of the electric charge density .12 The electric charge conservation law can be formulated in the equation of continuity for electric charge @ .t; x/ C r j.t; x/ D 0 @t (1.25)
which states that the time rate of change of electric charge .t; x/ is balanced by a divergence in the electric current density j.t; x/, i.e. the inux of charge.
1.3.2 Maxwells displacement current

We recall from the derivation of equation (1.22) on the preceding page that there we used the fact that in magnetostatics r j.x/ D 0. In the case of non-stationary sources and elds, we must, in accordance with the continuity equation (1.25) above, set r j.t; x/ D @ .t; x/=@t. Doing so, and formally repeating the steps in the derivation of equation (1.22) on the preceding page, we would obtain
FT
1. Electric charge is a conserved quantity and electric current is a transport of electric charge. As we shall see, this fact manifests itself in the equation of continuity and, as a consequence, in Maxwells displacement current.
j 11
the formal result r B.t; x/ D d3x 0 j.t; x0 /.x x0 / Z 1 0 @ 3 0 0 0 d x .t; x /r C 4 @t V 0 jx x0 j @ D 0 j.t; x/ C 0 "0 E.t; x/ @t
0 V0
(1.26)
The result is Maxwells source equation for the B eld @ r B.t; x/ D 0 j.t; x/ C "0 E.t; x/ D 0 j.t; x/ C @t
1.3.3 Electromotive force

If an electric eld E.t; x/ is applied to a conducting medium, a current density j.t; x/ will be set up in this medium. But also mechanical, hydrodynamical and chemical processes can give rise to electric currents. Under certain physical conditions, and for certain materials, one can assume that a linear relationship exists between the electric current density j and E. This approximation is called Ohms law:14
R A
where "0 @E.t; x/=@t is the famous displacement current. This, at the time, unobserved current was introduced by Maxwell, in a stroke of genius, in order to make also the right-hand side of this equation divergence-free when j.t; x/ is assumed to represent the density of the total electric current. This total current which can be split up into ordinary conduction currents, polarisation currents and magnetisation currents. This will be discussed in subsection 9.1 on page 182. The displacement current behaves like a current density owing in free space. As we shall see later, its existence has far-reaching physical consequences as it predicts that such physical observables as electromagnetic energy, linear momentum, and angular momentum can be transmitted over very long distances, even through empty space.
FT
0 "0
where, in the last step, we have assumed that a generalisation of equation (1.7) on page 4 to time-varying elds allows us to make the identication13 Z 1 1 @ d3x 0 .t; x0 /r 0 4 "0 @t V 0 jx x0 j Z 1 1 @ 3 0 0 d x .t; x /r D (1.27) @t 4 "0 V 0 jx x0 j Z 1 .t; x0 / @ @ r d3x 0 D E.t; x/ D 0 @t 4 "0 @t jx x0 j V
13
Later, we will need to consider this generalisation and formal identication further.
@ E.t; x/ @t (1.28)
In semiconductors this approximation is in general applicable only for a limited range of E. This property is used in semiconductor diodes for rectifying alternating currents.
14
12 j
j.t; x/ D E.t; x/
(1.29)
R A
C
where is the electric conductivity measured in Siemens per metre (Sm 1 ). We can view Ohms law equation (1.29) as the rst term in a Taylor expansion of a general law jE.t; x/. This general law incorporates non-linear effects such as frequency mixing and frequency conversion. Examples of media that are highly non-linear are semiconductors and plasma. We draw the attention to the fact that even in cases when the linear relation between E and j is a good approximation, we still have to use Ohms law with care. The conductivity is, in general, time-dependent (temporal dispersive media) but then it is often the case that equation (1.29) above is valid for each individual temporal Fourier (spectral) component of the eld. In some media, such as magnetised plasma and certain material, the conductivity is different in different directions. For such electromagnetically anisotropic media (spatial dispersive media) the scalar electric conductivity in Ohms law equation (1.29) has to be replaced by a conductivity tensor. If the response of the medium is not only anisotropic but also non-linear, higher-order tensorial terms have to be included. If the current is caused by an applied electric eld E.t; x/, this electric eld will exert work on the charges in the medium and, unless the medium is superconducting, there will be some energy loss. The time rate at which this energy is expended is j E per unit volume (Wm 3 ). If E is irrotational (conservative), j will decay away with time. Stationary currents therefore require that an electric eld due to an electromotive force (EMF ) is present. In the presence of such a eld Eemf , Ohms law, equation (1.29) above, takes the form
FT
j D .Estat C Eemf /
C
(1.30)
The electromotive force is dened as I E D dl .Estat C Eemf /
(1.31)
The term electromotive force is something of a misnomer since E represents a voltage, i.e. its SI dimension is V.
15
where dl is a tangential line element of the closed loop C .15
1.3.4 Faradays law of induction
In subsection on page 3 we derived the differential equations for the electrostatic eld. Specically, on page 5 we derived equation (1.10) stating that r Estat D 0 and hence that Estat is a conservative eld (it can be expressed as a gradient of a scalar eld). This implies that the closed line integral of Estat in equation (1.31) above vanishes and that this equation becomes I E D dl Eemf (1.32)
j 13
B.x/ O d2x n
B.x/
Figure 1.4: A loop C that moves with velocity v in a spatially varying magnetic eld B.x/ will sense a varying magnetic ux during the motion.
C dl
where m is the magnetic ux and S is the surface encircled by C , interpreted as a generic stationary loop and not necessarily as a conducting circuit. Application of Stokes theorem on this integral equation, transforms it into the differential equation r E.t; x/ D @ B.t; x/ @t (1.34)
that is valid for arbitrary variations in the elds and constitutes the Maxwell equation which explicitly connects electricity with magnetism. Any change of the magnetic ux m will induce an EMF. Let us therefore consider the case, illustrated in gure 1.4, when the loop is moved in such a way that it encircles a magnetic eld which varies during the movement. The total time derivative is evaluated according to the well-known operator formula d @ dx D C r dt @t dt (1.35)
This follows immediately from the multivariate chain rule for the differentiation of an arbitrary differentiable function f .t; x.t//. Here, dx=dt describes a chosen
R A
It has been established experimentally that a non-conservative EMF eld is produced in a closed circuit C at rest if the magnetic ux through this circuit varies with time. This is formulated in Faradays law which, in Maxwells generalised form, reads I d E.t / D dl E.t; x/ D m .t/ dt C Z Z (1.33) @ d O O d2x n B.t; x/ D d2x n B.t; x/ D dt S @t S
FT
14 j
path in space. We shall choose the ow path which means that dx=dt D v and @ d D Cv r dt @t (1.36)
where, in a continuum picture, v is the uid velocity. For this particular choice, the convective derivative dx=dt is usually referred to as the material derivative and is denoted Dx=Dt . Applying the rule (1.36) to Faradays law, equation (1.33) on the preceding page, we obtain Z Z Z d @B O O O E.t / D d2x n B D d2x n d2x n .v r /B (1.37) dt S @t S S
@ @ B.t; x/ D r B.t; x/ D @t @t
FT
r r E.t; x/ D 0 r B.t; x/ D 0 r .B v / D .v r /B Z v B/ D
S
Furthermore, taking the divergence of equation (1.34) on the previous page, we see that (1.38)
where in the last step formula (F.154) on page 217 was used. Since this is true for all times t , we conclude that (1.39)
R A D
or, rearranging the terms, I dl .Eemf
C
also for time-varying elds; this is in fact one of the Maxwell equations. Using this result and formula (F.143) on page 217, we nd that (1.40)
since, during spatial differentiation, v is to be considered as constant, This allows us to rewrite equation (1.37) above in the following way: I Z d emf O E.t / D dl E D d2x n B dt S C Z Z (1.41) @B 2 2 O O D d xn d x n r .B v / @t S S With Stokes theorem applied to the last integral, we nally get I I Z @B emf 2 O dl .B v / E.t / D dl E D d xn @t C C S
(1.42)
O d2x n
@B @t
(1.43)
j 15
where Eemf is the eld induced in the loop, i.e. in the moving system. The application of Stokes theorem in reverse on equation (1.43) on the preceding page yields r .Eemf v B/ D @B @t (1.44)
An observer in a xed frame of reference measures the electric eld E D Eemf v B (1.45)
and an observer in the moving frame of reference measures the following Lorentz force on a charge q
corresponding to an effective electric eld in the loop (moving observer) Eemf D E C v B
Hence, we conclude that for a stationary observer, the Maxwell equation r ED @B @t
FT
16
F D qEemf D qE C q.v
B/
(1.46)
In CGS units the microscopic Maxwell equations are r ED4 1 @B D0 c @t 4 1 @E D j c @t c
(1.47)
r BD0 EC B
r r
(1.48)
is indeed valid even if the loop is moving.
1.3.5 The microscopic Maxwell equations
R A
r ED "0 r BD0 r r EC B @B D0 @t 1 @E D c 2 @t
0j
in Heaviside-Lorentz units (one of several natural units) r ED r BD0 r r EC B 1 @B D0 c @t 1 1 @E D j c @t c
We are now able to collect the results from the above considerations and formulate the equations of classical electrodynamics valid for arbitrary variations in time and space of the coupled electric and magnetic elds E.t; x/ and B.t; x/. The equations are, in SI units,16 (1.49a) (1.49b) (1.49c) (1.49d)
and in Planck units (another set of natural units) r ED4 r BD0 r r EC B @B D0 @t @E D4 j @t
In these equations D .t; x/ represents the total, possibly both time and space dependent, electric charge density, with contributions from free as well as induced (polarisation) charges. Likewise, j D j.t; x/ represents the total, possibly both time and space dependent, electric current density, with contributions from conduction currents (motion of free charges) as well as all atomistic (polarisation
16 j
and magnetisation) currents. As they stand, the equations therefore incorporate the classical interaction between all electric charges and currents, free or bound, in the system and are called Maxwells microscopic equations. They were rst formulated by Lorentz and therefore another name frequently used for them is the Maxwell-Lorentz equations and the name we shall use. Together with the appropriate constitutive relations that relate and j to the elds, and the initial and boundary conditions pertinent to the physical situation at hand, they form a system of well-posed partial differential equations that completely determine E and B.
1.3.6 Diracs symmetrised Maxwell equations

If we look more closely at the microscopic Maxwell equations (1.49), we see that they exhibit a certain, albeit not complete, symmetry. Dirac therefore made the ad hoc assumption that there exist magnetic monopoles represented by a magnetic charge density, which we denote by m D m .t; x/, and a magnetic current density, which we denote by j m D j m .t; x/.17 With these new hypothetical physical entities included in the theory, and with the electric charge density denoted e and the electric current density denoted j e , the Maxwell equations will be symmetrised into the following two scalar and two coupled vectorial partial differential equations (SI units):
e
JULIAN SEYMOUR S C H W I N G E R (19181994) once put it: . . . there are strong theoretical reasons to believe that magnetic charge exists in nature, and may have played an important role in the development of the Universe. Searches for magnetic charge continue at the present time, emphasising that electromagnetism is very far from being a closed object.
17
A FT
r ED "0 r BD
0 m
(1.50a) (1.50b)
The magnetic monopole was rst postulated by P I E R R E C U R I E (18591906) and inferred from experiments in 2009.
@B m D (1.50c) 0j @t 1 @E r B D 0je (1.50d) c 2 @t We shall call these equations Diracs symmetrised Maxwell equations or the electromagnetodynamic equations. Taking the divergence of (1.50c), we nd that r EC @ m .r B/ (1.51) 0r j 0 @t where we used the fact that, according to formula (F.154) on page 217, the divergence of a curl always vanishes. Using (1.50b) to rewrite this relation, we obtain the equation of continuity for magnetic charge r .r E/ D
@ m C r jm D 0 (1.52) @t which has the same form as that for the electric charges (electric monopoles) and currents, equation (1.25) on page 10.
j 17
Faradays law derived from the assumed conservation of magnetic charge P O S T U L AT E 1.1 (I NDESTRUCTIBILITY OF MAGNETIC CHARGE ) Magnetic charge exists and is indestructible in the same way that electric charge exists and is indestructible. In other words, we postulate that there exists an equation of continuity for magnetic charges: @
m .t; x/
E X A M P L E 1.1
@t
C r j m .t; x/ D 0
(1.53)
Use this postulate and Diracs symmetrised form of Maxwells equations to derive Faradays law. The assumption of the existence of magnetic charges suggests a Coulomb-like law for magnetic elds: Z Z x x0 1 0 0 Bstat .x/ D D d3x 0 m .x0 / d3x 0 m .x0 /r 4 4 jx x0 j jx x0 j3 V0 V0 (1.54) Z m .x0 / 0 3 0 r dx D 4 jx x0 j V0
[cf. equation (1.7) on page 4 for Estat ] and, if magnetic currents exist, a Biot-Savart-like law for electric elds [cf. equation (1.17) on page 8 for Bstat ]: Z Z x x0 1 0 0 d3x 0 j m .x0 / d3x 0 j m .x0 / r D Estat .x/ D 4 4 jx x0 j jx x0 j3 V0 V0 Z m .x0 / j 0 D r d3x 0 4 jx x0 j V0 (1.55) Taking the curl of the latter and using the operator bac-cab rule, formula (F.151) on page 217, we nd that Z j m .x0 / 0 d3x 0 r Estat .x/ D r r D 4 jx x0 j V0 Z 1 0 (1.56) d3x 0 j m .x0 /r 2 D 0 4 jx x0 j V Z 1 0 d3x 0 j m .x0 / r 0 r 0 0 4 jx x0 j V
D
by analogy.
Comparing with equation (1.20) on page 9 for Estat and the evaluation of the integrals there, we realise that Z m r Estat .x/ D d3x 0 j m .x0 / .x x0 / D (1.57) 0 0 j .x/
V0
It is intriguing to note that if we assume that formula (1.55) is valid also for time-varying magnetic currents, then, with the use of the representation of the Dirac delta function, equation (F.171) on page 218, the equation of continuity for magnetic charge, equation (1.52) on the facing page, and the assumption of the generalisation of equation (1.54) above to time-dependent magnetic charge distributions, we obtain, at least formally,
A FT
18 j
E.t; x/ D
d3x 0 j m .t; x0 /.x x0 / Z 1 0 @ d3x 0 m .t; x0 /r 0 4 @t V 0 jx x0 j @ m B.t; x/ 0 j .t; x/ @t

0 V0
(1.58)
[cf. equation (1.26) on page 11] which we recognise as equation (1.50c) on page 16. A transformation of this electromagnetodynamic result by rotating into the electric realm of charge space, thereby letting j m tend to zero, yields the electrodynamic equation (1.49c) on page 15, i.e., the Faraday law in the ordinary Maxwell equations. This process would also provide an alternative interpretation of the term @B=@t as a magnetic displacement current, dual to the electric displacement current [cf. equation (1.28) on page 11]. By postulating the indestructibility of a hypothetical magnetic charge, and assuming a direct extension of results from statics to dynamics, we have been able to replace Faradays experimental results on electromotive forces and induction in loops as a foundation for the Maxwell equations by a more fundamental one. At rst sight, this result seems to be in conict with the concept of retardation. Therefore a more detailed analysis of it is required. This analysis is left to the reader.
1.4 Bibliography
[1]
T. W. BARRETT AND D. M. G RIMES, Advanced Electromagnetism. Foundations, Theory and Applications, World Scientic Publishing Co., Singapore, 1995, ISBN 981-02-2095-2. R. B ECKER, Electromagnetic Fields and Interactions, Dover Publications, Inc., New York, NY, 1982, ISBN 0-486-64290-9. W. G REINER, Classical Electrodynamics, Springer-Verlag, New York, Berlin, Heidelberg, 1996, ISBN 0-387-94799-X. E. H ALLN, Electromagnetic Theory, Chapman & Hall, Ltd., London, 1962.
[2] [3] [4] [5]
D
[6] [7] [8]
K. H UANG, Fundamental Forces of Nature. The Story of Gauge Fields, World Scientic Publishing Co. Pte. Ltd, New Jersey, London, Singapore, Beijing, Shanghai, Hong Kong, Taipei, and Chennai, 2007, ISBN 13-978-981-250-654-4 (pbk). J. D. JACKSON, Classical Electrodynamics, third ed., John Wiley & Sons, Inc., New York, NY . . . , 1999, ISBN 0-471-30932-X.
L. D. L ANDAU AND E. M. L IFSHITZ, The Classical Theory of Fields, fourth revised English ed., vol. 2 of Course of Theoretical Physics, Pergamon Press, Ltd., Oxford . . . , 1975, ISBN 0-08-025072-6. F. E. L OW, Classical Field Theory, John Wiley & Sons, Inc., New York, NY . . . , 1997, ISBN 0-471-59551-9.
A FT
End of example 1.1
1.4. Bibliography
j 19
[9]
J. C. M AXWELL, A Treatise on Electricity and Magnetism, third ed., vol. 1, Dover Publications, Inc., New York, NY, 1954, ISBN 0-486-60636-8.
[10] J. C. M AXWELL, A Treatise on Electricity and Magnetism, third ed., vol. 2, Dover Publications, Inc., New York, NY, 1954, ISBN 0-486-60637-8. [11] D. B. M ELROSE AND R. C. M C P HEDRAN, Electromagnetic Processes in Dispersive Media, Cambridge University Press, Cambridge . . . , 1991, ISBN 0-521-41025-8. [12] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0-201-05702-6. [13] F. ROHRLICH, Classical Charged Particles, third ed., World Scientic Publishing Co. Pte. Ltd., New Jersey, London, Singapore, . . . , 2007, ISBN 0-201-48300-9.
[15] J. VANDERLINDE, Classical Electromagnetic Theory, John Wiley & Sons, Inc., New York, Chichester, Brisbane, Toronto, and Singapore, 1993, ISBN 0-471-57269-1.
R A
FT
[14] J. A. S TRATTON, Electromagnetic Theory, McGraw-Hill Book Company, Inc., New York, NY and London, 1953, ISBN 07-062150-0.
R A
FT
ELECTROMAGNETIC FIELDS AND WAVES

As a rst step in the study of the dynamical properties of the classical electromagnetic eld, we shall in this chapter, as alternatives to the rst-order MaxwellLorentz equations, derive a set of second-order differential equations for the elds E and B. It turns out that these second-order equations are wave equations for the eld vectors E and B, indicating that electromagnetic vector wave modes are very natural and common manifestations of classical electrodynamics.1 But before deriving these alternatives to the Maxwell-Lorentz equations, we shall discuss the mathematical techniques of making use of complex variables to represent real physical observables in order to simplify the mathematical treatment. In this chapter we will also describe how to make use of the single spectral component (Fourier component) technique, which simplies the algebra, at the same time as it claries the physical content.
2.1 Axiomatic classical electrodynamics

In chapter 1 we described the historical route which led to the formulation of the microscopic Maxwell equations. From now on we shall consider these equations as postulates, i.e. as the axiomatic foundation of classical electrodynamics.2 As such, these equations postulate, in scalar and vector differential equation form, the behaviour in time t 2 R and in space x 2 R3 of the relation between the electric and magnetic elds E.t; x/ 2 R3 and B.t; x/ 2 R3 , respectively, and the charge density .t; x/ 2 R and current density j.t; x/ 2 R3 [cf. equations (1.49) on page 15] r ED "0 r BD0 r r EC B (Gausss law) (No magnetic charges) (Faradays law)
0j
R A
@B D0 @t 1 @E D c 2 @t 21
FT
2
1 In 1864, in a lecture at the Royal Society of London, J A M E S C L E R K M A X W E L L (18311879) himself said:
We have strong reason to conclude that light itselfincluding radiant heat and other radiation, if anyis an electromagnetic disturbance in the form of waves propagated through the electro-magnetic eld according to electromagnetic laws.
F R I T Z R O H R L I C H writes in Classical Charged Particles that A physical theory, in the narrow sense of the word, is a logical structure based on assumptions and denitions which permits one to predict the outcome of a maximum number of different experiments on the basis of a minimum number of postulates. One usually desires the postulates (or axioms) to be as self-evident as possible; one demands simplicity, beauty, and even elegance.
(2.1a) (2.1b) (2.1c) (2.1d)
(Maxwells law)
22 j
2 . E L E CTROMAGNETIC FIELDS AND WAVES
FT
e
and are not to be viewed as equations that only describe how the elds E and B are generated by and j, but also the other way around. We reiterate that in these equations .t; x/ and j.t; x/ are the total charge and current densities, respectively. Hence, these equations are considered microscopic in the sense that all charges and currents, including the intrinsic ones in matter, such as bound charges in atoms and molecules as well as magnetisation currents in magnetic material, are included, but macroscopic in the sense that quantum effects are neglected. Despite the fact that the charge and current densities may not only be considered as the sources of the elds, but may equally well be considered being generated by the elds, we shall follow the convention and refer to them as the source terms of the microscopic Maxwell equations. Analogously, the two equations where they appear will be referred to as the Maxwell-Lorentz source equations. I we allow for magnetic charge and current densities m and j m , respectively, in addition to electric charge and current densities e and j e j, we will have to replace the Maxwell-Lorentz equations by Diracs symmetrised equations r ED "0 (2.2a) (2.2b) (2.2c) (2.2d)
r BD r r
c 2 "0
0j
R A
EC B
3
@B D @t 1 @E D c 2 @t
0j
and consider them to be the postulates of electromagnetodynamics.
2.2 Complex notation and physical observables
A physical observable is something that can, one way or another, be ultimately reduced to an input to the human sensory system. In other words, physical observables quantify (our perception of) the physical reality and as such they must, of course, be described by real-valued quantities.
In order to simplify the mathematical treatment, we shall frequently allow the mathematical variables that represent the elds, the charge and current densities, and other physical quantities be analytically continued into the complex domain. However, when we use such a complex notation we must be very careful how to interpret the results derived within this notation. This is because every physical observable is, by denition, real-valued.3 Consequently, the mathematical expression for the observable under consideration must also be real-valued to be physically meaningful. If a physical scalar variable, or a component of a physical vector or tensor, is represented mathematically by the complex-valued number , i.e. if 2 C,
2.2. Complex notation and physical observables
j 23
2.2.1 Physical observables and averages
R A
i!t
As just mentioned, it is important to be aware of the limitations of the mathematical technique of using a complex representation for physical observables, particularly when evaluating products of complex mathematical variables that are to represent physical observables. Let us, for example, consider two physical vector elds a.t; x/ and b.t; x/ that are represented by their Fourier components, a0 .x/ exp. i!t/ and b0 .x/ exp. i!t /, i.e. by vectors in (a domain of) 3D complex space C3 . Furthermore, let be a binary operator for these vectors, representing either the scalar product operator ( ), the vector product operator ( ), the dyadic product operator (juxtaposition), or the outer tensor operator (). Then we make the interpretation a.t; x/ b.t; x/ D Re fag Re fbg D Re a0 .x/ e i!t Re b0 .x/ e
D
and
We can express the real part of the complex vectors a and b as Re fag D Re a0 .x/ e D 1 a0 .x/ e 2
i!t
Re fbg D Re b0 .x/ e
i!t
1 b0 .x/ e 2
FT
i!t
then in classical electrodynamics (in fact, in classical physics as a whole), one makes the identication observable D Re mathematical . Therefore, it is always understood that one must take the real part of a complex mathematical variable in order for it to represent a classical physical observable, i.e. something that is observed in Nature or measured in an experiment.4 For mathematical convenience and ease of calculation, we shall in the following regularly useand tacitly assumecomplex notation, stating explicitly when we do not. One convenient property of the complex notation is that differentiations often become trivial to perform. However, care must always be exercised. A typical situation being when products of two or more quantities are calculated since, for instance, for two complex-valued variables 1 and 2 we know that Re f 1 2 g Re f 1 g Re f 2 g. On the other hand, . 1 2/ D 1 2 .
This is at variance with quantum physics, where observable D mathematical . Letting denote complex conjugation, the real part can be written /, i.e. as Re f g D 1. C 2 the arithmetic mean of and its complex conjugate . Similarly, the magnitude can be written j jD. /1=2 , i.e. as the geometric mean of and . Under certain conditions, also the imaginary part corresponds to a physical observable.
(2.3)
C a0 .x/ ei!t
(2.4a)
i!t
C b0 .x/ ei!t
(2.4b)
24 j
ha bi t hRe fag Re fbgi t D
R A D
r G In regions where r
(2.6) 1 1 D Re fa0 b0 g D Re fa0 b0 g 2 2 This is because the oscillating function exp. 2i!t/ in equation (2.5) above vanishes when averaged in time over a complete cycle 2 =! (or over innitely many cycles), and, therefore, ha.t; x/ b.t; x/i t gives no contribution.
2.2.2 Maxwell equations in Majorana representation
It is often convenient to express electrodynamics in terms of the complex-eld six-vector, also known as the Riemann-Silberstein vector, G.t; x/ D E.t; x/ C icB.t; x/ (2.7)
where G 2 C3 even if E; B 2 R3 . If we use this vector, the Maxwell equations (2.1) on page 21 can be written r GD "0 (2.8a) (2.8b)
i @G D ic 0 j c @t D 0 and j D 0 these equations reduce to r GD0 GD i @G c @t
FT
1 1 Re fa b g D Re fa bg 2 2
respectively. Hence, the physically acceptable interpretation of the scalar product of two complex vectors, representing classical physical observables, is a.t; x/ b.t; x/ D Re a0 .x/ e i!t Re b0 .x/ e i!t 1 1 D a0 .x/ e i!t C a0 .x/ ei!t b0 .x/ e i!t C b0 .x/ ei!t 2 2 1 D a0 b0 C a0 b0 C a0 b0 e 2i!t C a0 b0 e2i!t 4 (2.5) 1 1 D Re fa0 b0 g C Re a0 b0 e 2i!t 2 2 1 1 D Re a0 e i!t b0 ei!t C Re a0 e i!t b0 e i!t 2 2 1 1 D Re fa.t; x/ b .t; x/g C Re fa.t; x/ b.t; x/g 2 2 In physics, we are often forced to measure the temporal average (cycle average) of a physical observable. We use the notation h i t for such an average and nd that the average of the product of the two physical quantities represented by a and b can be expressed as
(2.9a) (2.9b)
2.2. Complex notation and physical observables
j 25
which, with the help of the linear momentum operator y pD can be rewritten y p GD0 i @G D ciy p @t G (2.11a) (2.11b) i r (2.10)
where 0 0 0 B S D @0 0 0 i
1 0 1 0 0 0 0 i 0 C B C B O O iA x1 C @ 0 0 0A x2 C @ i 0 i 0 0 0
Then by introducing the Hamiltonian-like operator y y H D cS p D
we can write the Maxwell-Lorentz curl equations as
R
i
i.e. as a Schrdinger/Pauli/Dirac-like equation. This formulation of the freespace electromagnetic eld equations is known as the Majorana representation of the Maxwell-Lorentz equations or the Majorana formalism.6
Products of Riemann-Silberstein vectors with themselves for E; B 2 R3 C3 ,
One fundamental property of the 3D complex vector space to which the RiemannSilberstein vector G D E C icB belongs, is that inner (scalar) products in this space are invariant under rotations just as they are in R3 . However, as discussed in example M.4 on page 233, products in C3 can be dened in two different ways. Considering the special case of the scalar product of G with itself, assuming that E 2 R3 and B 2 R3 , we have the following two possibilities: 1. The inner (scalar) product dened as G scalar multiplied with itself
A FT
i @G y D c S pG @t (2.12) 1 i 0 C O 0 0A x3 0 0 (2.13) ic S r (2.14) @G y D HG @t (2.15)
6
The rst equation is the transversality condition p D k ? G where we anticipate the usual quantal relation between the eld momentum p and the wave vector k,5 whereas the second equation describes the dynamics. Using formula (F.160) on page 217, we can rewrite equation (2.11b) above as
5 The scalar quantity D h=.2 / is the reduced Planck constant where the Planck constant proper h 6:62606957 10 34 Js.
It so happens that E T T O R E M A J O R A NA (1906-1938) used the denition G D E icB, but this is no essential difference from the denition (2.7). One may say that Majorana used the other branch of p 1 as the imaginary unit.
E X A M P L E 2.1
26 j
G G D .E C icB/ .E C icB/ D E E
c 2 B B C 2icE B
(2.16)
Since length is a scalar quantity that is invariant under rotations, we nd that E2 c 2 B 2 D Const E B D Const (2.17a) (2.17b)
2. The inner (scalar) product dened as G scalar multiplied with the complex conjugate of itself G G D .E C icB/ .E icB/ D E E C c 2 B B E 2 C c 2 B 2 (2.18)
is also an invariant scalar quantity. As we shall see in chapter 4 on page 55, this quantity is proportional to the electromagnetic eld energy density. 3. As with any 3D vector, the cross product of G with itself vanishes: G G D .E C icB/ DE E
FT
.E C icB/ B/ c B
2
B C ic.E
B/ C ic.B B/ D 0
E/
(2.19)
D 0 C 0 C ic.E
ic.E
4. The cross product of G with the complex conjugate of itself does, however, not vanish. Instead it is G G D .E C icB/ DE D0C0 .E icB/
ECc B ic.E
ic.E
B/ C ic.B B/ D
E/
(2.20)
R A
5. The dyadic product of G with itself is GG D EE or, in component form, .GG/ij D Ei Ej 6. The dyadic product of G with itself is G G D EE
B/
ic.E
2ic.E
B/
which is proportional to the electromagnetic energy ux (the so called Poynting vector or the electromagnetic linear momentum density), to be introduced in chapter 4 on page 55.
c 2 BB C ic.EB C BE/
(2.21)
c 2 Bi Bj C ic.Ei Bj C Bi Ej /
(2.22)
c 2 BB
ic.EB C BE/ D .GG/
(2.23)
7. The dyadic product of G with its own complex conjuagte from the right is GG D EE C c 2 BB ic.EB BE/ (2.24)
and from the left it is G G D EE C c 2 BB C ic.EB BE/ D .GG / (2.25)
End of example 2.1
2.3. The wave equations for E and B
j 27
2.3 The wave equations for E and B

The Maxwell-Lorentz equations (2.1) on page 21 are four rst-order coupled differential equations (both E and B appear in the same equations). Two of the equations are scalar [equation (2.1a) and equation (2.1b)], and two are in 3D Euclidean vector form [equation (2.1c) and equation (2.1d)], representing three scalar equations each. Hence, the Maxwell equations represent eight (1 C 1 C 3 C 3 D 8) scalar coupled rst-order partial differential equations. However, it is well known from the theory of differential equations that a set of rst-order, coupled partial differential equations can be transformed into a smaller set of second-order partial differential equations that sometimes become decoupled in the process. It turns out that in our case we will obtain one second-order differential equation for E and one second-order differential equation for B. These second-order partial differential equations are, as we shall see, wave equations, and we shall discuss their properties and implications. In certain propagation media, the B wave eld can be easily obtained from the solution of the E wave equation but in general this is not the case. To bring the rst-order differential equations (2.1) on page 21 into second order one needs, of course, to operate on them with rst-order differential operators. If we apply the curl vector operator (r ) to both sides of the two Maxwell vector equations, equation (2.1c) and equation (2.1d), assuming that our physical quantities vary in such a regular way that temporal and spatial differentiation commute, we obtain the second-order differential equations
R A
.r E/ C r r @B @ D r .r E/ C .r @t @t 1 @ .r B/ .r E/ D 0 r c 2 @t 1 @2 E Cr c 2 @t 2 1 @2 B Cr c 2 @t 2 .r E/ D
0
As they stand, these second-order partial differential equations still appear to be coupled. However, by using the Maxwell equations once again we can formally decouple them into
.r
B/ D
If we use to the operator triple product bac-cab formula (F.151) on page 217, which gives r .r E/ D r .r E/ r 2E (2.28)
when applied to E and similarly to B, Gausss law equation (2.1a) on page 21, and then rearrange the terms, we obtain the two inhomogeneous vector wave
FT
B/ D 0 j (2.26a) (2.26b) @j @t (2.27a) (2.27b)
0r
28 j
equations
2
r "0 2 B D 0r ED
@j @t
(2.29a) (2.29b)
A FT
2 2
Clearly, if the current density in the RHS of equation (2.29b) is a function of E, as is the case if, for instance, Ohms law j D E is applicable, the coupling is not removed.
where 2 is the dAlembert operator, dened according to formula (M.116) on page 242. These are the general wave equations for the electromagnetic elds in regions where there exist sources .t; x/ and j.t; x/ of any kind. Simple everyday examples of such regions are electric conductors (e.g. radio and TV transmitter antennas) or plasma (e.g. the Sun and its surrounding corona). In principle, the sources and j can still cause the wave equations to be coupled, but in many important situations this is not the case.7 We notice that outside the source region, i.e. in free space where D 0 and j D 0, the inhomogeneous wave equations (2.29) above simplify to the well-known uncoupled, homogeneous wave equations ED0 BD0 (2.30a) (2.30b)
These equations describe how the elds that were generated in the source region, e.g. a transmitting antenna, propagate as vector waves through free space. Once these waves impinge upon another region which can sustain charges and/or currents for a long enough time, e.g. a receiving antenna or other electromagnetic sensors, the elds interact with the charges and the currents in this second region in accordance with equations (2.29) above.
A single plane wave is a mathematical idealisation. In reality, plane waves appear as building blocks of wave packets, i.e., superpositions of a (possibly innite) number of individual plane waves with different properties (frequencies, directions,. . . ). E.g., a radio beam from a transmitting antenna is a superposition (Fourier sum or integral) of many plane waves with slightly different angles relative to a xed, given axis or a plane.
R
Wave polarisation
E X A M P L E 2.2
Since electromagnetic waves are vector waves they exhibit wave polarisation. Let us consider a single plane wave that propagates in free space,8 i.e., a wave where the electric and magnetic eld vectors are restricted to a two-dimensional plane that is perpendicular the propagation direction. Let us choose this plane to be the x1 x2 plane and the propagation O vector (wave vector) k to be along the x3 axis: k D k x3 . A generic temporal Fourier mode of the electric eld vector E with (angular) frequency ! is therefore described by the real-valued expression E.t; x/ D E1 cos.!t O kx3 C 1 / x1 C E2 cos.!t O kx3 C 2 / x2 (2.31)
where the amplitudes Ei and phases i , can take any value. In complex notation we can write this as E.t; x/ D E1 ei1 ei.kx3
!t /
O x1 C E2 ei2 ei.kx3
!t / !tC1 /
!t /
O x2 (2.32)
O O D E1 ei1 x1 C E2 ei2 x2 ei.kx3 O O D E1 x1 C E2 ei0 x2 ei.kx3
j 29
where 0 D 2 1 . When this phase difference 0 vanishes, the electric eld oscillates along a line directed at an angle arctan .E2 =E1 / relative to the x1 axis. This is called linear wave polarisation. When 0 0 the wave is in general in a state of elliptical wave polarisation. For the special cases 0 D =2 and E1 D E2 D E0 the wave can, in complex notation, be described as E.t; x/ D E0 ei.kx3
!t C1 /
O x1 iO 2 x
(2.33)
As discussed in example M.1 on page 225, this shows that the eld vector E rotates around the x3 axis as it propatates along this axis. This rotation is called circular wave polarisation. The helical base vectors 1 O O h D p x1 iO 2 / x 2
which are xed unit vectors, allow us to write equation (2.33) above E.t; x/ D p 2E0 ei.kx3
!tC1 / O
R A
2
O O We use the convention that hC represents left-hand circular polarisation and h right-hand circular polarisation.9 Left-hand (right-hand) circular polarised waves are said to have positive helicity (negative helicity) with respect to the direction of propagation (along k). End of example 2.2
Wave equations expressed in terms of Riemann-Silberstein vectors If we use the Maxwell-Lorentz equations expressed in the Riemann-Silberstein vector G D E C icB, i.e., equations (2.8) on page 24, we can perform similar steps as we did when deriving equations (2.29) on the facing page. We then nd that r "0 @j Ci @t
GD
Taking the real and imaginary parts of this equation, assuming that E; B 2 R3 , we recover the wave equations (2.27) on page 27, as expected. End of example 2.3
FT
(2.34) h (2.35)
9
In physics, two different conventions are used. The handedness refers to the rotation in space of the electric eld vector, either when viewed as the wave propagates away from the observer or toward the observer.
E X A M P L E 2.3
0 cr
(2.36)
30 j
2.3.1 The time-independent wave equations for E and B

Often one can assume that the temporal dependency of E and B and of the sources and j is well-behaved enough that it can be represented by the sum of a nite number N of temporal spectral components (temporal Fourier components), or, in other words, in the form of a temporal Fourier series. In such situations it is sufcient to study the properties of one arbitrary member of this set of spectral components f!n W n D 0; 1; 2; : : : ; N 1g, i.e. E.t; x/ D En .x/ cos.!n t/ D En .x/Re e B.t; x/ D Bn .x/ cos.!n t/ D Bn .x/Re e
i!n t
En .x/e Bn .x/e
i!n t i!n t
(2.37a) (2.37b)
i!n t
10
R A
r 2 En e
i!n t
When subtle classical and quantum radiation effects are taken into account, one nds that all emissions suffer an unavoidable, intrinsic line broadening. Also, simply because the Universe has existed for only about 13.5 billion years, which is a nite time, no signals in the Universe can be observed to have a spectral width that is smaller than the inverse of this age.
where En ; Bn 2 R3 and !n ; t 2 R is assumed. This is because the MaxwellLorentz equations are linear, implying that the general solution is obtained by a weighted linear superposition (summation) of the result for each such spectral component, where the weight of the spectral component in question is given by its Fourier amplitude, En .x/, and Bn .x/, respectively. In a physical system, a temporal spectral component is identied uniquely by its angular frequency !n . A wave containing only a nite number of temporal spectral components is called a time-harmonic wave. In the limit when only one single frequency is present, we talk about a monochromatic wave. Strictly speaking, purely monochromatic waves do not exist.10 By inserting the temporal spectral component equation (2.37a) above into equation (2.29a) on page 28 one nds that for an arbitrary component the following equation is obtained: C
2 !n En e c2
FT
i!n t
C i!n
0 jn e
i!n t
r n e "0
i!n t
(2.38)
After dividing out the common factor exp. i!n t/, we obtain the time-independent wave equation r 2 En C
2 !n En C i!n c2 0 jn
r n "0
(2.39)
and similarly for B. Solving this equation, multiplying the solution obtained by exp. i!n t / and summing over all N such Fourier (spectral) components with frequencies !n ; n D 0; 1; 2; : : : ; N 1 present in the sources, and hence in the elds, the complete solution of the original wave equation is obtained. This is a consequence of the superposition principle which is valid as long as nonlinear effects can be neglected. In the limit of very many frequency components, the Fourier sum goes over into a Fourier integral . To illustrate this generic case, let us introduce the Fourier
j 31
transform of E.t; x/ Z 1 1 E! .x/ D dt E.t; x/ ei!t 2 1 and the corresponding inverse Fourier transform Z 1 E.t; x/ D d! E! .x/ e i!t
1
(2.40a)
(2.40b)
D0 since E!0; t!1
i! E! .x/
and that, consequently, Z 1 2
dt
@2 E.t; x/ i!t e D @t 2
R A
r 2 E! C !2 E! C i! c2
0 j!
Fourier transforming equation (2.29a) on page 28 and using (2.41) and (2.42), we thus obtain D r ! "0 (2.43)
which is mathematically identical to equation (2.39) on the preceding page. A subsequent inverse Fourier transformation of the solution E! of this equation leads to the same result as is obtained from the solution of equation (2.39) on page 30. Hence, by considering just one temporal Fourier component we obtain results which are identical to those that we would have obtained by employing the machinery of Fourier transforms and Fourier integrals. Hence, under the assumption of linearity (superposition principle) there is usually no need for the formal forward and inverse Fourier transform technique. What was said above in general terms about temporal spectral components is true also for spatial spectral components (spatial Fourier components) only that we must use a three-dimensional Fourier representation Z 1 1 Ek .t / D d3x E.t; x/ e ik x (2.44a) .2 /3 1 Z 1 E.t; x/ D d3k Ek .t / eik x (2.44b)
1
FT
! 2 E! .x/ (2.42)
where the amplitude E! .x/ 2 C3 is a continuous function of (angular) frequency ! 2 R and of x 2 R3 . We see that the Fourier transform of @E.t; x/=@t becomes Z 1 1 @E.t; x/ i!t dt e 2 @t 1 Z 1 1 1 i!t 1 i! E.t; x/ e dt E.t; x/ ei!t D (2.41) 1 2 2 1
32 j
Since we always assume that the real part shall be taken (if necessary), we can pick an arbitrary pair of the spatial amplitudes in equation (2.37a) and equation (2.37b) on page 30, denote the members of this pair by E0 and B0 , respectively, and then represent them as the Fourier modes E0 .x/ D e0 Re eik0 x D E!;k eik x (2.45a) ik x B0 .x/ D b0 Re e 0 D B!;k eik x (2.45b) respectively, where k0 is the wave vector (measured in m 1 ) of mode 0; in the last step we introduced a complex notation and also dropped the mode number since the formul are valid for any mode n. Now, since
and O r eik x D xi
@ ikj xj O e D xi ikj ij eikj xj D iki xi eikj xj D ikeik x (2.46b) @xi we see that for each spectral component in equations (2.45) above, temporal and spatial differential operators turn into algebraic operations according to the following scheme: @ ! i! @t r ! ik ! ik ! ik (2.47a) (2.47b) (2.47c) (2.47d)
R A
r r We note that r E D ik r B D ik
FT
r E D ik E D ik Ek E D ik B D ik E? B? r B D ik B D ik Bk r D ik j D ik j? r j D ik j D ik jk r j D ik r E? D 0 r Ek D 0
@ e @t
i!t
i!e
i!t
(2.46a)
(2.48a) (2.48b) (2.48c) (2.48d) (2.48e) (2.48f) (2.48g)
Hence, with respect to the wave vector k, the r operator projects out the spatially longitudinal component, and the r operator projects out the spatially transverse component of the eld vector in question. Put in another way, (2.49a) (2.49b)
j 33
and so on for the other observables. This can be viewed as a special instance of the Helmholtzs theorem, discussed in subsection M.3.7 on page 248, saying that if E falls off suitably rapidly at innity, it can be written as the sum of a rotational part Erotat and an irrotational part Eirrot : E D Erotat C Eirrot where, according to equations (M.152) on page 249, r Erotat D 0 r by making the formal identication11 E? D Erotat Ek D E For the magnetic eld
irrot
(2.50)
(2.51a) (2.51b)
11
irrot
D0
B D Brotat C Birrot where
R A
r Brotat D 0 B
irrot
D0
we make the analogous identication
B? D Brotat Bk D B
irrot
As seen from equations (2.47) on the facing page, the Fourier transform of a function of time t is a function of angular frequency !, and the Fourier transform of a function of the position vector x is a function of the wave vector k. One says that ! is a reciprocal space to t and that k spans a space that is reciprocal to x. In the reciprocal ! and k space the Maxwell-Lorentz equations are ik Ek D "0 ik Bk D 0 ik ik E? i!B D 0 ! B? C i 2 E D c (2.57a) (2.57b) (2.57c)
0j
FT
(2.52a) (2.52b) (2.53) (2.54a) (2.54b) (2.55) (2.56) (2.57d)
Strictly speaking, the spatial Fourier component is a plane wave and plane waves do not fall off at all at innity as required for the Helmholtz decomposition to be applicable. However, only wave packets made up of a sum of plane waves are physically acceptable and such packages can be localised well enough.
34 j
Applying the Helmholtz decomposition, the Maxwell-Lorentz equations become r Eirrot D r Birrot r r "0 D0 @B D0 @t 1 @E D c 2 @t (2.58a) (2.58b) (2.58c)
0j
Erotat C Brotat
(2.58d)
2.4 Bibliography
[16] J. D. JACKSON, Classical Electrodynamics, third ed., John Wiley & Sons, Inc., New York, NY . . . , 1999, ISBN 0-471-30932-X. [17] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0201-05702-6. [18] C. H. PAPAS, Theory of Electromagnetic Wave Propagation, Dover Publications, Inc., New York, 1988, ISBN 0-486-65678-0.
R A
FT
ELECTROMAGNETIC POTENTIALS AND GAUGES

As described in chapter 1 on page 1, the concepts of electric and magnetic elds were introduced such that they are intimately related to the mechanical forces between charges and currents given by Coulombs law and Ampres law, respectively. Just as in mechanics, it turns out that in electrodynamics it is often more convenient to express the theory in terms of potentials rather then in terms of the electric and magnetic elds (Coulomb and Ampre forces) themselves. This is particularly true for problems related to radiation and relativity. As we shall see in chapter 7 on page 145, the potentials pay a central rle in the formulation of relativistically covariant electromagnetism. And at the quantum level, electrodynamics is almost exclusively formulated in terms of potentials rather than electric and magnetic elds. In this chapter we introduce and study the properties of such potentials and nd that they exhibit some remarkable properties that elucidate the fundamental aspects of electromagnetism, lead naturally to the special theory of relativity, and pave the way for gauge eld theories.
3.1 The electrostatic scalar potential
As we saw in equation (1.10) on page 5, the time-independent electric (electrostatic) eld Estat .x/ is irrotational. According to formula (F.155) on page 217 we may therefore express it in terms of the gradient of a scalar eld. If we denote this scalar eld by stat .x/, we get Estat .x/ D r stat .x/ (3.1)
Taking the divergence of this and using equation (1.9) on page 5, we obtain Poissons equation r 2 stat .x/ D r Estat .x/ D .x/ "0 (3.2)
R A
35
FT
36 j
3 . E L E C T ROMAGNETIC POTENTIALS AND GAUGES
If we compare with the denition of Estat , namely equation (1.7) on page 4, we see that this equation has the solution Z 1 .x0 / stat .x/ D d3x 0 (3.3) 4 "0 V 0 jx x0 j where the integration is taken over all source points x0 at which the charge density .x0 / is non-zero. The scalar function stat .x/ in equation (3.3) above is called the electrostatic scalar potential .
3.2 The magnetostatic vector potential

Let us consider the equations of magnetostatics, equations (1.24) on page 10. According to formula (F.154) on page 217 any vector eld a has the property that r .r a/ 0 and in the derivation of equation (1.19) on page 8 in magnetostatics we found that r Bstat .x/ D 0. We therefore realise that we can always write Bstat .x/ D r Astat .x/ (3.4)
R A
0
where Astat .x/ is called the magnetostatic vector potential . In the magnetostatic case, we may start from Biot-Savarts law as expressed by equation (1.17) on page 8. Identifying this expression with equation (3.4) above allows us to dene the static vector potential as Z j.x0 / 0 (3.5) Astat .x/ D d3x 0 4 jx x0 j V0
From equations (3.1) and (3.4) we conclude that if we transform equations (3.3) and (3.5) in the following way stat .x/ 7! stat .x/ D stat .x/ C .x/ A
stat
FT
stat 0
(3.6a) (3.6b)
.x/ 7! A
.x/ D A
stat
.x/ C a.x/
the elds Estat and Bstat will be unaffected provided r .x/ D 0 r a.x/ D 0 (3.7a) (3.7b)
i.e. if is an arbitrary scalar function that is not dependent on x, e.g. a constant, and a.x/ is an arbitrary vector eld whose curl vanishes. According to formula (F.155) on page 217 such a vector eld can always be written as the gradient of a
3.2. The magnetostatic vector potential
j 37
scalar eld. In other words, the elds are unaffected by the transformation (3.6) if .x/ D Const a.x/ D r .x/ where is an arbitrary, differentiable function of x.
Multipole expansion of the electrostatic potential The integral in the electrostatic potential formula Z 1 .x0 / stat .x/ D d3x 0 4 "0 V 0 jx x0 j E X A M P L E 3.1
(3.8a) (3.8b)
(3.9)
1 jx x0 j
j.x
x0 /
1 3 X 1 X 1 D C n jx x0 j nD1
@n
1 jx x0 j
@xi1 D 1 jx x0 j
@xin C
1 X
nD1 n1 Cn2 Cn3 Dn ni 0 0 .x1
Inserting this into the integral in formula (3.9), we obtain the expansion " Z 1 X X 1 1 . 1/n d3x 0 .x0 / C stat .x/ D 4 "0 jx x0 j V 0 n1n2 n3
R
Z
V0 0 d3x 0 .x1 V0 V0
1 x0 j n n n @x1 1 @x2 2 @x3 3
@n jx
D
1 x0 j n n n @x1 1 @x2 2 @x3 3
@n jx
Clearly, the rst integral in this expansion is nothing but the static net charge Z qD d3x 0 .x0 /
If we introduce the electrostatic dipole moment vector Z d.x0 / D d3x 0 .x0 x0 / .x0 /
A FT
1 .x0 x0 /j
3 X i1 D1 in D1 0 .xi1
where is the charge density introduced in equation (1.9) on page 5, is not always possible to evaluate analytically. However, for a charge distribution source .x0 / that is well localised in a small volume V 0 around x0 , a series expansion of the integrand in such a way that the dominant contributions are contained the rst few terms can be made. E.g., if we in Taylor expand the inverse distance 1= x x0 with respect to the point x0 D x0 we obtain
x0i /
1
0 .xin
x0in /
(3.10)
. 1/n n1 n2 n3
0 x01 /n1 .x2
0 x02 /n2 .x3
x03 /n3
nD1 n1 Cn2 Cn3 Dn ni 0 0 x02 /n2 .x3
(3.11) # x03 /n3 .x0 /
0 x01 /n1 .x2
(3.12)
(3.13)
38 j
with components pi , i D 1; 2; 3, and the electrostatic quadrupole moment tensor Z Q.x0 / D d3x 0 .x0 x0 /.x0 x0 / .x0 /
V0
(3.14)
with components Qij ; i; j D 1; 2; 3, and use the fact that @ jx and that @2 jx
1 x0 j 1 x0 j
@xi
xi jx
x0i x0 j3 x0 j2 ij
(3.15)
@xi @xj
3.xi
x0i /.xj jx
x0j / x0 j
5
jx
(3.16)
where Einsteins summation convention over i and j is implied. We see that at large distances from a localised charge distribution, the electrostatic potential can, to the lowest order, be approximated by the (Coulomb) potential from a single point charge q located at the moment point x0 . We also see that Z Z Z d3x 0 .x0 / d3x 0 x0 .x0 / x0 d3x 0 .x0 x0 / .x0 / D d.x0 / D V0 V0 V0 Z d3x 0 x0 .x0 / x0 q (3.18) D
V0
R D
or
Q0 D Q
0 Qij D Qij
from which we draw the conclusion that if q 0, it is always possible to choose the moment point x0 such that d D 0, and if q D 0, then d is independent of the choice of moment point x0 . Furthermore, one can show that 1 3.xi ij 2 x0i /.xj jx x0j / x0 j 5 jx x0 j2 ij D0 (3.19)
where is an arbitrary constant. Choosing it to be Z 2 1 D d3x 0 x0 x0 .x0 / 3 V0
A FT
Z ij D
V0
then we can write the rst three terms of the expansion of equation (3.9) on the previous page as " q 1 1 x x0 stat .x/ D C d 4 "0 jx x0 j jx x0 j x0 j2 jx # (3.17) 1 3 .xi x0i / .xj x0j / 1 C Qij ij C : : : 2 jx x0 j jx x0 j 2 jx x0 j3
(3.20)
we can transform Qij into d3x 0

0 .xi 0 x0i /.xj
x0j /
1 0 x 3
2 x0 ij
.x0 /
(3.21) Z
13 D
V0
d3x 0
.x0
x0 /.x0
x0 /
13
1 0 x 3
2 x0 .x0 /
(3.22)
0 O O where 13 D xi xi is the unit tensor. It follows immediately that Qi i D 0 (Einstein summa-
3.3. The electrodynamic potentials
j 39
tion), i.e., that Q0 is traceless. Rotating the coordinate system, it is possible to diagonalise the tensors Q and Q0 . For any spherical symmetric distribution of charge, all components of Q0 vanish if the moment point x0 is chosen as the symmetry centre of the distribution. If the charge distribution .x/ is made up of discrete point charges qn with coordinates xn , the denitions above of q; d; Q and Q0 become X qD qn (3.23a)
n
dD
QD Q0 D
X
n
qn .xn qn .xn qn .xn
x0 / x0 /.xn x0 /.xn x0 / x0 /
13
(3.23b) (3.23c) 1 jxn 3 x0 j2 (3.23d)
X
n
X
n
3.3 The electrodynamic potentials
R A
B.t; x/ D r A.t; x/ r E.t; x/ D @ r @t A.t; x/ D r E.t; x/ C r @ A.t; x/ D 0 @t
Let us now generalise the static analysis above to the electrodynamic case, i.e. the case with temporal and spatial dependent sources .t; x/ and j.t; x/, and the pertinent elds E.t; x/ and B.t; x/, as described the Maxwell-Lorentz equations (2.1) on page 21. In other words, let us study the electrodynamic potentials .t; x/ and A.t; x/. According to the non-source Maxwell-Lorentz equation (2.1b), the magnetic eld B.t; x/ is divergence-free also in electrodynamics (if magnetic charges are not included). Because of this divergence-free nature of the time- and spacedependent magnetic eld, we can express it as the curl of an electromagnetic vector potential : (3.24)
Inserting this expression into the other non-source Maxwell-Lorentz equation, viz. equation (2.1c) on page 21, we obtain @ A.t; x/ @t (3.25)
or, rearranging the terms, (3.26)
As before we utilise the vanishing curl of a vector expression to write this vector expression as the gradient of a scalar function. If, in analogy with the
FT
End of example 3.1
40 j
electrostatic case, we introduce the electromagnetic scalar potential function .t; x/, equation (3.26) on the preceding page becomes equivalent to E.t; x/ C @ A.t; x/ D @t r .t; x/ (3.27)
This means that in electrodynamics, E.t; x/ is calculated from the potentials according to the formula E.t; x/ D r .t; x/ @ A.t; x/ @t (3.28)
R A
3.4 Gauge conditions
r 2A 1 @2 A c 2 @t 2
Inserting (3.28) and (3.24) on the preceding page into Maxwells equations (2.1) on page 21, we obtain, after some simple algebra and the use of equation (1.13) on page 6, the equations r 2 D .t; x/ "0 @ .r A/ @t 1 @ 0 j.t; x/ C 2 r c @t (3.29a) (3.29b)
Subtracting .1=c 2 /@2 =@t 2 from both sides of the rst equation and rearranging, these equations turn into the following general inhomogeneous wave equations .t; x/ @ 1 @ 2 D C r AC 2 (3.30a) "0 @t c @t 1 @ 2 A D 0 j.t; x/ r r A C 2 (3.30b) c @t These two second-order, coupled, partial differential equations, representing in all four scalar equations (one for and one each for the three components
FT
r .r A/ D
and B.t; x/ from formula (3.24) on the previous page. Hence, it is a matter of convention (or taste) whether we want to express the laws of electrodynamics in terms of the potentials .t; x/ and A.t; x/, or in terms of the elds E.t; x/ and B.t; x/. However, there is an important difference between the two approaches: in classical electrodynamics the only directly observable quantities are the elds themselves (and quantities derived from them) and not the potentials. On the other hand, the treatment becomes signicantly simpler if we use the potentials in our calculations and then, at the nal stage, use equation (3.24) on the preceding page and equation (3.28) above to calculate the elds or physical quantities expressed in the elds. This is the strategy we shall follow.
3.4. Gauge conditions
j 41
Ai ; i D 1; 2; 3 of A) are completely equivalent to the formulation of electrodynamics in terms of Maxwells equations, which represent eight scalar rst-order, coupled, partial differential equations. As they stand, equations (3.29) and equations (3.30) on the preceding page look complicated and may seem to be of limited use. However, if we write equation (3.24) on page 39 in the form r A.t; x/ D B.t; x/ we can consider this as a specication of r A. But we know from Helmholtzs theorem [see subsection M.3.7 on page 248] that in order to determine the (spatial) behaviour of A completely, we must also specify r A. Since this divergence does not enter the derivation above, we are free to choose r A in whatever way we like and still obtain the same physical results! This illustrates the power of formulating electrodynamics in terms of potentials.
3.4.1 Lorenz-Lorentz gauge
If we choose r A to full the so called Lorenz-Lorentz gauge condition 1 r AC 1 @ D0 c 2 @t
the coupled inhomogeneous wave equations (3.30) on page 40 simplify to the following set of uncoupled inhomogeneous wave equations:
R A
2
.t; x/ "0 2 A D 0 j.t; x/ D
where 2 is the dAlembert operator, discussed in example M.9 on page 242. Each of these four scalar equations is an inhomogeneous wave equation of the following form:
2
.t; x/ D s.t; x/
where denotes for either or one of the components Ai ; i D 1; 2; 3 of the vector potential A, and s is a shorthand for the pertinent source component, .t; x/="0 or 0 ji .t; x/; i D 1; 2; 3, respectively. We assume that our sources are well-behaved enough in time t so that the Fourier transform pair for the generic source function s Z 1 s.t; x/ D d! s! .x/ e i!t (3.34a) 1 Z 1 1 s! .x/ D dt s.t; x/ ei!t (3.34b) 2 1
FT
1
(3.31)
(3.32a) (3.32b)
In fact, the Dutch physicist HENDRIK ANTOON LORENTZ (1853-1928), who in 1903 demonstrated the covariance of Maxwells equations, was not the original discoverer of the gauge condition (3.31). It had been discovered by the Danish physicist L U DV I G VA L E N T I N L O R E N Z (1829-1891) already in 1867. This fact has sometimes been overlooked in the literature and the condition was earlier referred to as the Lorentz gauge condition.
(3.33)
42 j
exists, and that the same is true for the generic potential component : Z 1 .t; x/ D d! ! .x/ e i!t (3.35a) 1 Z 1 1 dt .t; x/ ei!t (3.35b) ! .x/ D 2 1 Inserting the Fourier representations (3.34) and (3.35) into equation (3.33) on the previous page, and using the vacuum dispersion relation for electromagnetic waves relating the angular frequency !, the speed of light c, and the wave number k D .2 /= where is the vacuum wavelength , ! D ck (3.36)
r 2 ! .x/ C k 2 ! .x/ D
FT
s! .x/ .x x0 /
V0
the generic 3D inhomogeneous wave equation (3.33) on the preceding page, turns into (3.37)
R A D
which is the 3D inhomogeneous time-independent wave equation, often called the 3D inhomogeneous Helmholtz equation. As postulated by Huygenss principle, each point on a wave front acts as a point source for spherical wavelets of varying amplitude (weight). A new wave front is formed by a linear superposition of the individual weighted wavelets from each of the point sources on the old wave front. The solution of (3.37) can therefore be expressed as a weighted sum of solutions of an equation where the source term has been replaced by a single point source r 2 G.x; x0 / C k 2 G.x; x0 / D (3.38)
and the solution of equation (3.37) above which corresponds to the frequency ! is given by the weighted superposition Z ! .x/ D d3x 0 s! .x0 /G.x; x0 / (3.39) (plus boundary conditions) where s! .x0 / is the wavelet amplitude at the source point x0 . The function G.x; x0 / is called the Green function or the propagator. Due to translational invariance in space, G.x; x0 / D G.x x0 /. Furthermore, in equation (3.38) above, the Dirac generalised function .x x0 /, which represents the point source, depends only on x x0 and there is no angular dependence in the equation. Hence, the solution can only be dependent on r D jx x0 j and not on the direction of x x0 . If we interpret r as the radial coordinate in a spherically polar coordinate system, and recall the expression for the Laplace operator in such a coordinate system, equation (3.38) above becomes d2 .rG/ C k 2 .rG/ D dr 2 r.r/ (3.40)
j 43
Away from r D jx takes the form
x0 j D 0, i.e. away from the source point x0 , this equation (3.41)

0
d2 .rG/ C k 2 .rG/ D 0 dr 2 with the well-known general solution

0
the volume integrated equation (3.38) on the facing page can be approximated by Z 1 3 0 2 .CC C C / dx r jx x0 j V0 Z Z (3.44) 1 3 0 2 3 0 0 C k .CC C C / dx D d x . x x / jx x0 j V0 V0
In virtue of the fact that the volume element d3x 0 in spherical polar coordinates is proportional to r 2 D jx x0 j2 [see formula (F.74) on page 213], the second integral vanishes when jx x0 j ! 0. Furthermore, from equation (F.171) on page 218, we nd that the integrand in the rst integral can be written as 4 .jx x0 j/ and, hence, that the two constants C must full the condition 1 (3.45) CC C C D 4 Now that we have determined the relation between CC and C , we insert the general solution equation (3.42) above into equation (3.39) on the preceding page and obtain the general solution in the ! domain: Z Z ikjx x0 j ikjx x0 j 3 0 0 e 3 0 0 e ! .x/ D CC d x s! .x / CC d x s! .x / (3.46) jx x0 j jx x0 j V0 V0 In order to nd the solution in the t domain, we take the inverse Fourier transform of this by inserting the above expression for ! .x/ into equation (3.35) on the facing page: i h 0 Z Z 1 exp i! t kjx! x j .t; x/ D CC d3x 0 d! s! .x0 / jx x0 j V0 1 h i (3.47) 0 Z Z 1 exp i! t C kjx! x j CC d3x 0 d! s! .x0 / jx x0 j V0 1
R A
FT
e ikr eikjx x j e ikjx x j eikr CC CC CC (3.42) 0j r r jx x jx x0 j where C are constants. In order to determine the constants C , we insert the general solution, equation (3.42) above, into equation (3.38) on the preceding page and integrate over a small volume around r D jx x0 j D 0. Since 1 1 G.x x0 / CC (3.43) CC ; x x 0 ! 0 0j 0j jx x jx x G D CC
44 j
R A
3.4.2 Coulomb gauge
In fact, inspired by ideas put forward by PAU L A D R I E N M AU R I C E D I R AC (1902 1984), J O H N A R C H I BA L D W H E E L E R (19112008) and RICHARD PHILLIPS FEYNMAN (19181988) derived, in 1945, a consistent electrodynamics based on both the retarded and the advanced potentials.
0 0 If we introduce the retarded time tret and the advanced time tadv in the following way [using the fact that k=! D 1=c in free space, according to formula (3.36) on page 42]: 0 x x0 .t 0 / k x x0 .tret / ret 0 0 x x 0 / D t Dt (3.48a) tret D tret .t; c 0 0 ! 0 0 x x .t / k x x .tadv / adv 0 0 DtC (3.48b) tadv D tadv .t; x x0 / D t C ! c and use equation (3.34) on page 41, we obtain Z Z 0 0 f .tadv ; x0 / f .tret ; x0 / .t; x/ D CC d3x 0 CC d3x 0 (3.49) jx x0 j jx x0 j V0 V0 This is a solution to the generic inhomogeneous wave equation for the potential components equation (3.33) on page 41. We note that the solution at time t at the eld point x is dependent on the behaviour at other times t 0 of the source at x0 and that both retarded and advanced t 0 are mathematically acceptable solutions. However, if we assume that causality requires that the potential at .t; x/ is set 0 up by the source at an earlier time, i.e. at .tret ; x0 /, we must in equation (3.49) above set C D 0 and therefore, according to equation (3.45) on the previous page, CC D 1=.4 /.2 From the above discussion about the solution of the inhomogeneous wave equations in the Lorenz-Lorentz gauge we conclude that, if we discard the advanced potentials, the electrodynamic potentials in free space can be written Z 0 0 tret ; x0 .tret / 1 .t; x/ D d3x 0 (3.50a) 0 4 "0 V 0 jx.t/ x0 .tret /j Z 0 0 j tret ; x0 .tret / 1 A.t; x/ D (3.50b) d3x 0 0 4 "0 c 2 V 0 jx.t/ x0 .tret /j These retarded potentials were obtained as solutions to the Lorenz-Lorentz inhomogeneous wave equations (3.32). The expressions (3.50) are therefore valid only in the Lorenz-Lorentz gauge. In other gauges the expressions for the potentials are analytically different but will, of course, yield the very same physical elds E and B as the expressions (3.50) do. As they stand, we shall use expressions (3.50) quite frequently in the sequel.
In Coulomb gauge, often employed in quantum electrodynamics, one chooses r A D 0 so that equations (3.29) or equations (3.30) on page 40 become .t; x/ r 2 D (3.51a) "0 1 @2 A 1 @ r 2A D (3.51b) 0 j.t; x/ C 2 r 2 @t 2 c c @t
FT
j 45
The rst of these two is the time-dependent Poissons equation which, in analogy with equation (3.3) on page 36, has the solution Z 1 .t; x0 / .t; x/ D d3x 0 (3.52) 4 "0 V 0 jx x0 j We see that in the scalar potential expression, the charge density source is evaluated at time t . I.e., the scalar potential does not exhibit any retardation, which means that the effect of the charge shows up in the potential instantaneously, as if the propagation speed were innite. This is not in disagreement with the laws of nature since in classical physics potentials are not physical observables. Since in the Coulomb gauge the scalar potential does not suffer any retardation (or advancement) but the elds E and B themselves must be physical and exhibit retardation effects, all retardation must occur in the vector potential A, i.e. the solution of the inhomogeneous wave equation (3.51b) on the facing page. As we see, the last term in the RHS of this equation contains the scalar potential , which, according to equation (3.52) above, in turn depends on the charge density . The continuity equation (1.25) on page 10 provides a relation between and j on which we can apply Helmholtz decomposition and nd that j D j rotat C j irrot where (3.53)
R A
r j rotat D 0 r j irrot D 0 "0 r @ D j irrot @t
0j
[cf. equations (2.52) on page 33]. Then the equation of continuity becomes @ @ C r j irrot D "0 r 2 C r j irrot @t @t @ Dr "0 r C j irrot D 0 @t r D 0 and r j irrot D 0, one nds that @ r "0 r C j irrot D 0 @t
Furthermore, since r
D
r 2A 1 @2 A D c 2 @t 2
According to Helmholtzs theorem, this implies that (3.57)
The inhomogeneous wave equation (3.51b) on the facing page thus becomes C 1 @ r D c 2 @t
0j
FT
(3.54a) (3.54b) (3.55) (3.56) C
0j irrot
0j
rotat
(3.58)
46 j
which shows that in Coulomb gauge the source of the vector potential A is the rotational (transverse) component of the current, j rotat . The irrotational (longitudinal) component of the current j irrot does not contribute to the vector potential. The retarded (particular) solution in Coulomb gauge of equations (3.29) on page 40 is therefore [cf. equation (3.50a) on page 44]: Z t; x0 .t/ 1 d3x 0 .t; x/ D 4 "0 V 0 jx.t/ x0 .t/j Z rotat 0 0 j tret ; x0 .tret / 0 A.t; x/ D d3x 0 0 4 jx.t/ x0 .tret /j V0 (3.59) (3.60)
The Coulomb gauge is also called the transverse gauge or the radiation gauge.
3.4.3 Velocity gauge
If r A fulls the velocity gauge condition, sometimes referred to as the complete -Lorenz gauge, r AC @ D 0; c 2 @t D c2 v2 (3.61)
The value D 1, corresponding to an imaginary speed v D ic, gives the Kirchhoff gauge, introduced already in 1857 by G U S TAV R O B E RT K I R C H H O F F (18241884).
R A
2
we obtain the Lorenz-Lorentz gauge condition in the limit v D c, i.e. D 1, and the Coulomb gauge condition in the limit v ! 1, i.e. D 0, respectively, where v is the propagation speed of the scalar potential. Hence, the velocity gauge is a generalisation of both these gauges.3 Inserting equation (3.61) above into the coupled inhomogeneous wave equations (3.30) on page 40 they become .t; x/ "0 @ @ c 2 @t @t 1 @ r 0 j.t; x/ C c2 @t 1
FT
D AD
(3.62a) (3.62b)
3.5 Gauge transformations

We saw in section 3.1 on page 35 and in section 3.2 on page 36 that in electrostatics and magnetostatics we have a certain mathematical degree of freedom, up to terms of vanishing gradients and curls, to pick suitable forms for the potentials and still get the same physical result. In fact, the way the electromagnetic scalar potential .t; x/ and the vector potential A.t; x/ are related to the physical observables gives leeway for similar manipulation of them also in electrodynamics.
3.5. Gauge transformations
j 47
In analogy with equations (3.6) on page 36 we introduce .t; x/ 7! 0 .t; x/ D .t; x/ C .t; x/ A.t; x/ 7! A .t; x/ D A.t; x/ C a.t; x/
0
(3.63a) (3.63b)
By inserting these transformed potentials into equation (3.28) on page 40 for the electric eld E.t; x/ and into equation (3.24) on page 39 for the magnetic eld B.t; x/, we see that the elds will be unaffected provided r .t; x/ C r @a.t; x/ D0 @t a.t; x/ D 0 (3.64a) (3.64b)
a.t; x/ D r .t; x/
which, when inserted into the rst condition in turn requires that @.t; x/ .t; x/ D @t
Hence, if we simultaneously transform both .t; x/ and A.t; x/ into new ones 0 .t; x/ and A0 .t; x/ according to the scheme
R A
.t; x/ 7! 0 .t; x/ D .t; x/
@.t; x/ @t A.t; x/ 7! A0 .t; x/ D A.t; x/ C r .t; x/
and insert the transformed potentials into equation (3.28) on page 40 for the electric eld and into equation (3.24) on page 39 for the magnetic eld, we obtain the transformed elds @ @A @.r / @A0 E0 D r 0 D r C r @t @t @t @t (3.67a) @A @.r / @.r / @A D r C D r @t @t @t @t 0 0 B D r A D r A C r .r / D r A (3.67b) where, once again equation (F.155) on page 217 was used. This explicit calculation clearly demonstrates that the elds E and B are unaffected by the gauge transformation (3.66). The function .t; x/ is called the gauge function. A transformation of the potentials and A which leaves the elds, and hence Maxwells equations, invariant is called a gauge transformation. Any physical law that does not change under a gauge transformation is said to be gauge invariant. It is only those quantities (expressions) that are gauge invariant that are
FT
(3.65a) (3.65b) (3.66a) (3.66b)
If we introduce an arbitrary, sufciently differentiable scalar function .t; x/, the second conditioned is fullled if we require that
48 j
A very important extension is the Yang-Mills theory, introduced in 1954. This theory has had a profound impact on modern physics.
observable and therefore have experimental signicance. Trivially, the electromagnetic elds and the Maxwell-Lorentz equations themselves are gauge invariant and electrodynamics is therefore a gauge theory and as such the prototype for all gauge theories.4 As just shown, the potentials .t; x/ and A.t; x/ calculated from equations (3.29) on page 40, with an arbitrary choice of r A, can be gauge transformed according to (3.66) on the previous page. If, in particular, we choose r A according to the Lorenz-Lorentz condition, equation (3.31) on page 41, and apply the gauge transformation (3.66) on the resulting Lorenz-Lorentz potential equations (3.32) on page 41, these equations will be transformed into 1 @2 c 2 @t 2 1 @2 A c 2 @t 2 @ .t; x/ 1 @2 2 r C r D 2 @t 2 @t c "0 1 @2 r 2A r r 2 D 0 j.t; x/ c 2 @t 2
2
(3.68a) (3.68b)
We notice that if we require that the gauge function .t; x/ itself be restricted to full the wave equation 1 @2 c 2 @t 2 r 2 D 0 (3.69)
these transformed Lorenz-Lorentz equations will keep their original form. The set of potentials which have been gauge transformed according to equation (3.66) on the preceding page with a gauge function .t; x/ restricted to full equation (3.69) above, or, in other words, those gauge transformed potentials for which the equations (3.32) on page 41 are invariant, comprise the Lorenz-Lorentz gauge.
E X A M P L E 3.2
R
shows that
Longitudinal and transverse components in gauge transformations
If we represent the vector potential A.t; x/ in the reciprocal k space as described at the end of subsection on page 30, the gauge transformation equation (3.66b) on the previous page becomes Ak .t/ 7! A0 .t/ D Ak .t/ k ikk .t/ (3.70)
we can separate it into its longitudinal and transverse components A0 D Ak k A0 ? r (3.71a) (3.71b)
A Helmholtz decomposition [see formula (M.150) on page 248] A D Arotat C Airrot (3.72)
A FT
D A?
j 49
A0 A
rotat
D Arotat
irrot
(3.73a) (3.73b)
0 irrot
DA
Hence, a law (expression) that depends on A only through its transverse/rotational component A? D Arotat is gauge invariant, whereas a law that depends on the longitudinal/irrotational component Ak D Airrot is in general not gauge invariant and therefore does not represent a physical observable. With the caveat about plane waves mentioned near formul (2.52) on page 33, the following then applies for the electric eld and the magnetic eld E D E? C Ek and B D B? C Bk , respectively: E? D Ek D @A? @t r @Ak @t (3.74a) (3.74b) (3.74c)
In terms of rotational and irrotational parts, the electric and magnetic elds are given by the sums E D Erotat C Eirrot and B D Brotat C Birrot , respectively, where the individual terms are Erotat D Eirrot D @Arotat @t (3.75a)
R
i
Gauge transformations and quantum mechanics
As discussed in section 2.2 on page 22, quantum theory requires that we take the magnitude rather than the real part of our mathematical variables in order to turn them into quantities that represent physical observables. In non-relativistic quantum mechanics, the physical observable probability density is D j j2 , where the wave function 2 C solves the Schrdinger equation @ y DH @t (3.76)
y and H is the Hamilton operator. The non-relativistic Hamiltonian for a classical particle with charge q in an electromagnetic
A FT
B? D r Bk D 0 A? (3.74d) r @Airrot @t (3.75b) (3.75c) Brotat D r
irrot
Arotat
D0
(3.75d)
End of example 3.2
E X A M P L E 3.3
50 j
eld, described by the scalar potential and vector potential A, is H D 1 .p 2m qA/2 C q (3.77)
where p is the linear momentum. The corresponding quantal Hamilton operator is obtained y from the corrsepondence principle, viz., by replacing p by the operator p D i r , referred to as minimal coupling. This gives the Schrdinger equation i @ 1 . i r D @t 2m qA/2 C q (3.78)
The idea is to perform a gauge transformation from the potentials .t; x/ and A.t; x/ to new potentials
A FT
A ! A0 .t; x/ D A.t; x/ i @.ei / 1 . i r D @t 2m qA/2 ei C qei i @ 0 1 i r D @t 2m qA C .qr /2
0
! 0 .t; x/ D .t; x/ C
@.t; x/ @t r .t; x/
(3.79a) (3.79b)
and then nd a .t; x/, expressed in the gauge function .t; x/, so that the transformed Schrdinger equation can be written (3.80)
Under the gauge transformation equation (3.79) above, the Schrdinger equation (3.78) transforms into C q
0
Cq
@ @t
(3.81)
Now, setting
.t; x/ D ei.t;x/ .t; x/
(3.82)
R
we see that i r D i r
qA C .qr /
qA C .qr / i r qA C .qr / i ei .r i/
qA C .qr / ei qAei C .qr /ei i (3.83)
D i r h
i ei .r /
D i r qA C .qr / ei i r i .r i/ qA C .qr / h i D i .r i/ei i ei r qAei C .qr /ei i r De i r D ei i r

i
i .r i/ i .r i/
qA C .qr / qA C .qr /2
qA C .r / C .qr /2 q.r /, or,
Clearly, the gauge transformed Hamilton operator is unchanged iff .r / D
j 51
equivalently, iff .t; x/ D i @ 0 @t D i2 q @ @t @ i e @t
q.t; x/= . This has as a consequence that

0
Di
@ 0 @ 0 @ i @ i C Di .e /C e @t @t @t @t @ @ i @ C i ei C e D ei i @t @t @t
(3.84)
Inserting this into the transformed Schrdinger equation (3.81) on the facing page, we recover the untransformed Schrdinger equation (3.78). We conclude that for a gauge transformation of the potentials and A and using the minimal coupling as in equation (3.78) on the preceding page, the Schrdinger equation is invariant, but that the wave function changes from to e i where the phase angle is realvalued. This means that the non-relativistic quantum physical observable j j2 is unaffected by a gauge transformation of the classical potentials and A that appear in the Hamilton operator. The fact that .t; x/ is coordinate dependent, means that we are dealing with a local gauge transformation.
The fact that the probability density D j j2 and the probability current i . r r /=.2m/ do not change under a gauge transformation of and A means that the R charge density D q j j2 and therefore also the charge V 0 d3x 0 is a conserved quantity. In other words, the electromagnetic gauge symmetry corresponds to electric charge conservation and vice versa. For the gauge transformation given by formul (3.79) on the facing page, W O L F G A N G PAU L I introduced the notation gauge transformation of the second kind whereas he called a wavefunction phase change, expression (3.82) on the preceding page, a gauge transformation of the rst kind . End of example 3.3
3.5.1 Other gauges

Other useful gauges are
The Poincar gauge (multipolar gauge, radial gauge) .t; x/ D

0
R
Z A0 .t; x/ D Z
0 1 0
where is a scalar parameter. We note that in Poincar gauge A0 and x are orthogonal.
The Weyl gauge, also known as the temporal gauge or Hamilton gauge, dened by 0 D 0. The axial gauge, dened by A0 D 0. 3
The process of choosing a particular gauge condition is known as gauge xing.
A FT
1
d x E.t; x/ B.t; x/
(3.85a) (3.85b)
d x
52 j
E X A M P L E 3.4
Electromagnetodynamic potentials In Diracs symmetrised form of electrodynamics (electromagnetodynamics), Maxwells equations are replaced by [see also equations (2.2) on page 22]:
e
r ED r BD r r
"0
0 m 0j m e
(3.86a) (3.86b) (3.86c) (3.86d)
@B EC D @t 1 @E D B c 2 @t
0j
A FT
ED r e .t; x/ BD 1 r m .t; x/ c2 @ e A .t; x/ r Am @t 1 @ m A .t; x/ C r Ae c 2 @t
e .t; x/ @ r Ae D @t "0 m .t; x/ @ 2 m m r C r A D @t "0 1 @2 Ae 1 @e r 2 Ae C r r Ae C 2 D 0 j e .t; x/ c 2 @t 2 c @t 1 @2 Am 1 @m 2 m m D 0 j m .t; x/ r A Cr r A C 2 c 2 @t 2 c @t
In this theory, one derives the inhomogeneous wave equations for the usual electric scalar and vector potentials .e ; Ae / and their magnetic counterparts .m ; Am / by assuming that the potentials are related to the elds in the following symmetrised form: (3.87a) (3.87b)
In the absence of magnetic charges, or, equivalently for m 0 and Am 0, these formul reduce to the usual Maxwell theory, formula (3.28) on page 40 and formula (3.24) on page 39 respectively, as they should. Inserting the symmetrised expressions (3.87) above into equations (3.86) above, one obtains [cf., equations (3.29) on page 40] r 2 e C (3.88a) (3.88b) (3.88c) (3.88d)
R D
By choosing the conditions on the divergence of the vector potentials as the generalised Lorenz-Lorentz condition [cf. equation (3.31) on page 41] 1 c2 1 r Am C 2 c these coupled wave equations simplify to r Ae C
2 2
@ e D0 @t @ m D0 @t
e .t; x/
(3.89a) (3.89b)
e D
"0
e 0 j .t; x/ m .t; x/
(3.90a) (3.90b) (3.90c) (3.90d)
Ae D m D
2 2
"0
0j m
Am D
.t; x/
3.6. Bibliography
j 53
exhibiting, once again, the striking properties of Diracs symmetrised Maxwell theory. End of example 3.4
3.6 Bibliography
[19] L. D. FADEEV AND A. A. S LAVNOV, Gauge Fields: Introduction to Quantum Theory, No. 50 in Frontiers in Physics: A Lecture Note and Reprint Series. Benjamin/Cummings Publishing Company, Inc., Reading, MA . . . , 1980, ISBN 0-80539016-2.
[21] J. D. JACKSON, Classical Electrodynamics, third ed., John Wiley & Sons, Inc., New York, NY . . . , 1999, ISBN 0-471-30932-X.
[22] H.-D. NATURE, The Physical Basis of The Direction of Time, fourth ed., SpringerVerlag, Cambridge . . . , 1920, ISBN 3-540-42081-9.
[23] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0201-05702-6.
[24] J. A. W HEELER AND R. P. F EYNMAN, Interaction with the absorber as a mechanism for radiation, Reviews of Modern Physics, 17 (1945), pp. 157.
R A
FT
[20] M. G UIDRY, Gauge Field Theories: An Introduction with Applications, John Wiley & Sons, Inc., New York, NY . . . , 1991, ISBN 0-471-63117-5.
R A
FT
FUNDAMENTAL PROPERTIES OF THE ELECTROMAGNETIC FIELD

In this chapter we will explore a number of fundamental properties of the electromagnetic elds and their sources, as well as of the physical observables constructed from them. Of particular interest are symmetries since they have a striking predictive power and are essential ingredients of the physics. This includes both discrete and continuous geometric symmetries (reection, translation, rotation) and intrinsic symmetries (duality, reciprocity). Intimately related to symmetries are conserved quantities (constants of motion) of which our primary interest will be the electromagnetic energy, linear momentum, angular momentum, and centre of energy. These conserved quantities can carry information over large distances and are all more or less straightforwardly related to their counterparts in classical mechanics (indeed in all eld theories). But we will also consider other conserved quantities, where this relation is perhaps less straightforward. To derive useful mathematical expressions for the physical observables that we want to study, we once again take the microscopic Maxwell-Lorentz equations (2.1) on page 21 as our axiomatic starting point.
4.1 Charge, space, and time inversion symmetries

Let us rst investigate how the charge density , the current density j, and the elds E and B behave under charge conjugation, i.e., a change of sign of charge q ! q 0 D q, called C symmetry; under space inversion, i.e., a change of sign of the space coordinates x ! x0 D x, called parity transformation and P symmetry; and under time reversal , i.e., a change of sign of the time coordinate t ! t 0 D t , called T symmetry. Recalling that /q (4.1a)
D
and
R A
jD vD dx dt 55
FT
(4.1b)
56 j
4 . F U N DA M E N TA L PROPERTIES OF THE ELECTROMAGNETIC FIELD
the transformation properties follow directly from the Maxwell-Lorentz equations (2.1) on page 21. Let us study them one by one. C SYMMETRY, CHARGE CONJUGATION q ! q 0 D !
0
q: (4.2a)
D
0 dx 0
j ! j0 D
dt
dx D dt
dx D dt
(4.2b) (4.2c)
r ! r0 D r
r E0 .t; x/ D and r
E0 .t; x/ D r
respectively, implying that
FT
0
@ @ @ ! 0 D (4.2d) @t @t @t When inserted into the two rst of the Maxwell-Lorentz equations (2.1) on page 21 we nd that they become "0 D "0 D r E.t; x/
E.t; x/ D
@ @ B0 .t; x/ D B.t; x/ @t @t
E0 .t; x/ D B .t; x/ D
0
E.t; x/
(4.3a) (4.3b)
R A
!
0
B.t; x/
Consequently, the Maxwell-Lorentz equations are invariant under charge conjugation. No violation of this invariance has yet been found experimentally. x: (4.4a) d. x / D dt
0
P SYMMETRY, SPACE INVERSION x ! x0 D D

0
j ! j0 D
0 dx
r ! r0 D
dt r
dx D dt
(4.4b) (4.4c) (4.4d)
@ @ @ ! 0 D @t @t @t
implying that E0 .t; x/ E.t; x/ D

0
E.t; x/
(4.5a) (4.5b)
B .t; x/ B.t; x/ D B.t; x/
Since, by assumption, t , and "0 are ordinary scalars and that, consequently, ="0 is an ordinary scalar, and that the position vector x is the prototype of all ordinary (polar) vectors, we note that j and E are ordinary vectors, whereas B is a pseudovector (axial vector) as described in subsection M.2.2 on page 235.
4.2. Electromagnetic duality
j 57
T SYMMETRY, TIME REVERSAL t ! t 0 D !

0
t: (4.6a) j @ @t (4.6b) (4.6c) (4.6d)
D
0
j ! j0 D @ @ ! 0 D @t @t implying that
r !r Dr
E0 .t; x/ E. t; x/ D E.t; x/ B .t; x/ B. t; x/ D

0
(4.7a) (4.7b)
We see that E is even and B odd under time reversal. The Universe as a whole is asymmetric under time reversal. On quantum scales this is exhibited by the uncertainty principle and on classical scales by the arrow of time which is related to the increase of thermodynamic entropy in a closed system. The CPT theorem states that the combined CPT symmetry must hold for all physical phenomena. No violation of this law has been observed so far.
4.2 Electromagnetic duality
We notice that Diracs symmetrised version of the Maxwell equations (1.50) on page 16 exhibit the following symmetry (recall that "0 0 D 1=c 2 ): E ! cB
e m e
R A
cB ! c ! ! E
m
c
m
cj ! j jm !
cj e
which is a particular case ( D =2) of the general duality transformation, also known as the Heaviside-Larmor-Rainich transformation (indicated by the Hodge
FT
(4.8a) (4.8b) (4.8c) (4.8d) (4.8e) (4.8f)
B.t; x/
58 j
star operator
in the upper left-hand corner of the symbol in question)

?
E D E cos C cB sin E sin C cB cos

e
(4.9a) (4.9b) (4.9c) (4.9d) (4.9e) (4.9f)
c BD c
? e
Dc D
cos C
e
sin
m
? m ? e
c
e
sin C
m
cos
c j D cj cos C j sin
? m
j D
cj sin C j cos
Recall that the Taylor expansion for cos contains only even powers of whereas the expansion for sin contains only odd.
E X A M P L E 4.1
R
r
?
Duality of the electromagnetodynamic equations
Show that the symmetric, electromagnetodynamic Maxwell-Lorentz equations (2.2) on page 22 (Diracs symmetrised Maxwell equations) are invariant under the duality transformation (4.9). Explicit application of the transformation yields
e
A FT
E D r .E cos C cB sin / D cos C c 0 "0 ? e 1 1 m e D cos C sin D "0 c "0 e 1 r ?B D r . E sin C B cos / D sin C c c"0 D 0 c e sin C m cos D 0 ? m
m
This transformation leaves the symmetrised Maxwell equations, and hence the physics they describe (often referred to as electromagnetodynamics), invariant. Since E and j e are true (polar) vectors, B a pseudovector (axial vector), e a (true) scalar, we conclude that the magnetic charge density m and , which behaves as a mixing angle in a two-dimensional charge space, must be pseudoscalars1 and j m a pseudovector. The invariance of Diracs symmetrised Maxwell equations under the duality transformation (4.9) means that the amount of magnetic monopole density m is irrelevant for the physics as long as the ratio m = e D tan is kept constant. So whether we assume that the particles are only electrically charged or have also a magnetic charge with a given, xed ratio between the two types of charges is a matter of convention, as long as we assume that this fraction is the same for all particles. Such particles are referred to as dyons. By varying the mixing angle we can change the fraction of magnetic monopoles at will without changing the laws of electrodynamics. For D 0 we recover the usual Maxwell electrodynamics.
sin (4.10a)
cos
(4.10b)
4.2. Electromagnetic duality
j 59
EC
@?B Dr @t D
D r
?
1 E sin C B cos c 1 @E @B m cos C c 0 j e sin C sin 0 j cos (4.10c) @t c @t 1 @E @B m e sin C cos D 0 j cos C c 0 j sin c @t @t e m ? m 0 . cj sin C j cos / D 0 j .E cos C cB sin / C @ @t .
1 @?E Dr c 2 @t 1 D c
Duality expressed in Riemman-Silberstein formalism
Expressed in the Riemann-Silberstein complex eld vector, introduced in equation (2.7) on page 24, the duality transformation equations (4.9) on the facing page become
?
G D ?E C ic ?B D E cos C cB sin D E.cos i sin / C icB.cos
R A
i sin / D .E C icB/e
?
iE sin C icB cos

i
from which it is easy to see that G

?
2 G D ?G D Ge
whereas
G D G Ge
D
@
?
Furthermore, assuming that D .t; x/, we see that spatial and temporal differentiation of ?G leads to @?G D i.@ t /e i G C e i @ t G @t @ ?G r ?G D ie i r G C e i r G @ t ?G Gr
?
GD
ie
which means that @ t ?G transforms as ?G itself only if is time-independent, and that r ?G and r ?G transform as ?G itself only if is space-independent. End of example 4.2
FT
QED End of example 4.1 D Ge
i
1 1 @ .E cos C cB sin / E sin C B cos / c c 2 @t 1 @B 1 @E m cos C 0 j e cos C 2 cos 0 j sin C c @t c @t (4.10d) 1 @E 1 @B cos sin 2 c @t c @t 1 m D 0 j sin C j e cos D 0 ?j e c
E X A M P L E 4.2
(4.11)
G ei D jGj2
(4.12)
2i
(4.13)
(4.14a) (4.14b) G (4.14c)
GCe
60 j
4.3 Conservation laws

It is well established that a deeper understanding of a physical system can be obtained by studying the systems conserved quantities (constants of motion), i.e., those observables that do not change with time. In other words, the task is to nd the conservation laws for the physical system under study, which, according to Noethers theorem,2 is closely related to nding the symmetries of the system. Consider a certain physical property or substance, e.g. electric charge, that ows, in a time-dependent way, in 3D space. Let us denote the volume density of this substance by %.t; x/ and its ow velocity by v .t; x/. Let us also introduce a closed volume V , xed in space and enclosed by a perfectly permeable surface O O S with a innitesimally small directed area element d2x n D dxi dxj n, where dxi and dxj are two coordinates that span the local tangent plane of the surface O and n is a unit vector, orthogonal to the tangent plane and pointing outward. In the special case that the volume V is spherical, it is usually referred to as the control sphere. The property or substance ows into the volume V at the rate %v per unit area. Clearly, this inward ow will increase the amount of substance in V . We also allow for an increase due to the production of the substance inside V . This increase is quantied by the source density s. Recalling that the normal vector O n points outward, the following balance equation must hold: Z d d3x % dt V
R A
I D
Total change within V
FT
Z O d2x n %v C d3x s V S
Flow into V Production inside V
AMALIE EMMY NOETHER (18821935) made important contributions to mathematics and theoretical physics. Her (rst) theorem, which states that any differentiable symmetry of the action of a physical system has a corresponding conservation law, is considered to be a fundamental tool of theoretical physics.
(4.15)
With the help of Gausss theorem, formula (F.177) on page 218, this balance equation can be rewritten Z d3x @% C r .%v / @t s D0 (4.16)
Since this equation must hold for an arbitrary volume V , the integrand must vanish and we obtain the continuity equation @% C r .%v / D s @t (4.17)
which expresses, in differential form, the balance between the explicit temporal change of the density of the substance, the ow of it across the surface of a closed volume, and the production of the substance in the volume enclosed by the surface. If there is no production of the substance in V , i.e., if s D 0, the equation describes the constancy of the amount of substance.
4.3. Conservation laws
j 61
In fact, since r .r a/ D 0 for any arbitrary, differentiable vector eld a, we can generalise equation (4.17) on the facing page to @% C r .%v C r a/ D s (4.18) @t Furthermore, since the time derivative d=dt operating on a scalar, vector or tensor eld, dependent on t and x.t /, is d @ D Cv r dt @t according to the chain rule [cf. equation (1.36) on page 14], and since r .%v / D %r v C v r % we can rewrite the equation of continuity as (4.19)
(4.20)
d% C %r .v C r a/ D s (4.21) dt where, again, a is an arbitrary, differentiable pseudovector eld with dimension m2 s 1 , e.g., the moment of velocity with respect to x0 , i.e., .x x0 / v .
4.3.1 Conservation of electric charge
@ .t; x/ C r j.t; x/ D 0 (4.22) @t This is our rst conservation law derived from the microscopic Maxwell equations (2.1) on page 21. This result is, of course, consistent a posteriori with the fact that these axiomatic laws were formulated with the a priori assumption that the continuity equation (1.25) on page 10 was valid. The continuity equation (4.22) above describes the fact that electric charge is conserved, i.e., indestructible.3
R A
1
0
When the substance density % in equation (4.17) on the facing page is the electric charge density , we exclude any charge-generating mechanisms, i.e., we assume that s D 0, and interpret v as the electric current density j. Then the continuity equation takes the form
4.3.2 Conservation of total electric current
If we use the notation j c for the electric current associated with electric charge transport and j d for the displacement current @."0 E/=@t, equation (2.1d) on page 21 can be rewritten r B D jc C jd (4.23)
FT
3
This was postulated in 1747 by B E N JA M I N F R A N K L I N (1706 1770), printer, scientist, inventor, philosopher, statesman, and one of the founding fathers of the United States of America.
62 j
Differentiating this with respect to time t and using the Maxwell equation (2.1c) on page 21 and formula (F.159) on page 217, we obtain the following conservation law for the total current @j c @j d C Cr @t @t
13
1
0
.r
E/ D 0
(4.24)
This is a rather unusual way of writing the wave equation for the electric eld vector Er. Usually it is written as in equation (2.29a) on page 28.
4.3.3 Conservation of energy
r .E
B/ D B .r D B
FT
E/ E .r B/ D
0
The continuity equation (4.22) on the preceding page contains the divergence of the rst-order vector quantity j. As one example of the divergence of a second order vector quantity, let us study the divergence of E B. Using the Maxwell equations (2.1) on page 21 we obtain the balance equation
@B @E "0 0 E 0E j @t @t @ "0 E E C c2B B C j E @t 2
(4.25)
R A
ueld D SD 1 E
0
where formula (F.137) on page 217 was used. Let us introduce the electromagnetic eld energy density "0 .E E C c 2 B B/ 2 (Jm
3
(4.26)
and the electromagnetic energy ux or the Poynting vector B D "0 c 2 E B (Wm

2
(4.27)
that can also be viewed as the electromagnetic energy current density. Employing formula (3.24) on page 39, the Poynting vector S can be expressed in terms of the vector potential A vector as SD D D D 1
0
.r
A/ E .r A/ r .EA/ C .r E/A r .EA/
(4.28a) (4.28b) (4.28c) (4.28d)
1
0
.r A/ E .r A/ E
1
0
1
0
A C .r A/ E
j 63
where we used formula (F.147) on page 217 in the rst step, formula (F.140) on page 217 in the second step and the Maxwell-Lorentz equation (2.1a) on page 21. For each spatial Fourier component, we can use the results in subsection on page 30 to resolve A into its longitudinal and transverse components Ak and A? , respectively, with the result that SD D D D 1
0
.r
A? / E .r A/? r .EA? / C .r E/A? r .EA/?
(4.29a) (4.29b) (4.29c) (4.29d)
1
0
.r A? / E .r A? / E
1
0
1
0
With the help of these denitions we can write the balance equation (4.25) on the preceding page as @ueld Cr S D j E @t i.e., as a continuity equation with a source density sD j ED E v (4.30)
@umech Dj E @t Equation (4.30) above can therefore be written
R A
From the denition of E, we conclude that E represents force per unit volume (Nm 3 ) and that E v therefore represents work per unit volume or, in other words, power density (Wm 3 ). It is known as the Lorentz power density and is equivalent to the time rate of change of the mechanical energy density (Jm 3 ) of the current carrying particles
@umech @ueld C Cr S D0 (4.33) @t @t which is the energy density balance equation in differential form. R Expressing the Lorentz power, V d3x j E, as the time rate of change of the mechanical energy: Z dU mech D d3x j E (4.34) dt V and introducing the electromagnetic eld energy U eld D U e C U m , where Z "0 Ue D d3x E E (4.35) 2 V
FT
(4.31) (4.32)
A? C .r A? / E
64 j
is the electric eld energy and Z Z 1 "0 Um D d3x B B D d3x c 2 B B 2 0 V 2 V
(4.36)
is the magnetic eld energy, we can write the integral version of the balance equation (4.33) on the previous page as dU mech dU eld C C dt dt I
S
O d2x n S D 0
(4.37)
A FT
j D .E C Eemf / ED j Eemf Z d3x j Eemf
V V V
where This is the energy theorem in Maxwells theory, also known as Poyntings theorem. Allowing for an EMF and assuming that Ohms law (4.38)
is a valid approximation, or, equivalently, that
(4.39)
we obtain the relation Z Z j2 d3x j E D d3x
(4.40)
which, when inserted into equation (4.37) above and use is made of equation (4.34) on the previous page, yields I Z dU eld O d3x j Eemf D C d2x n S d3x j Eemf dt V S V Z
Field power
R D
E X A M P L E 4.3
(4.41)
Supplied electric power
Radiated power Supplied electric power
This shows how the supplied power (left-hand side, LHS) is expelled in the form of a time rate change of electric and magnetic eld energy, radiated electromagnetic power, and Joule heat power, i.e., Ohmic losses in the system (right-hand side, RHS).
Conservation of the total energy in a closed system
Show, by explicit calculation, that the total energy U D U mech C U eld of a closed system comprising N non-relativistic particles of mass mi , speed vi and charge qi , and pertinent electromagnetic elds E and B, U D Z N 1X "0 2 mi vi C d3x .E E C c 2 B B/ 2 2 V
i D1
(4.42)
j 65
is conserved. To show that the total energy U of the closed system is conserved, i.e. is constant in time, is to show that the derivative of U with respect to time t vanishes. Direct differentiation yields Z N X dU dvi @B @E D mi vi C "0 d3x E C c2B dt dt @t @t V
i D1
(4.43)
From the Lorentz force equation (4.61) on page 68 and Newtons second law we nd that mi which means that
N X iD1
dvi D qi E.t; xi / C vi dt
B.t; xi /
i D1
Furthermore, from the Maxwell-Lorentz equations (2.1) on page 21 we nd that @E 1 j D c2r B @t "0 @B D r E @t
Substitution into equation (4.43) above yields, after some rearrangement of terms, Z N X dU D qi vi E.t; xi / d3x j E.t; x/ dt V iD1 Z 1 C d3x E .r B/ B .r E/
0 V
Since the current is carried by discrete charged particles i D 1; 2; : : : ; N , the current density can be represented as
R
jD
N X i D1
and therefore the two rst terms in equation (4.45) above cancel. If we use also formula (F.137) on page 217 we see that equation (4.45) can be written Z Z I dU 1 O D d3x r .E B/ D d3x r S D d2x n S (4.47) dt 0 V V S
where in the last step the divergence theorem, formula (F.177) on page 218, was used.
This result shows that the rate at which the total energy is lost in a volume equals the amount of energy ux that ows outward across a closed surface enclosing this volume. Of course, this is nothing but what is stated in the energy theorem (Poyntings theorem), formula (4.37) on the facing page. Now, if the surface lies entirely outside the boundaries of the system under study, and this system is closed, no energy ux passes through the surface and hence dU D0 dt
A FT
i D1 0
mi vi
N N X X dvi D qi vi E.t; xi / C qi v i dt
vi
N X B.t; xi / D qi vi E.t; xi / i D1
(4.44a)
(4.44b)
(4.45)
qi vi x
xi .t/
(4.46)
66 j
showing that the total energy U D U mech C U eld of a closed system is indeed a conserved quantity. QED Put in another way: If we observe a change in the radiated power from a given system, we can deduce that there has been a similar change, but with opposite sign, of the total energy of the system itself. This can be useful to know if we want to study the dynamics of a remote system, e.g. in the Universe. End of example 4.3
4.3.4 Conservation of linear momentum

The derivation of the energy conservation formula (4.30) on page 63 introduced the Poynting (energy ux) vector S.t; x/. We now seek a balance equation involving the time derivative of S and nd, using the Maxwell equations (2.1) on page 21, that @S @E D "0 c 2 @t @t 2 D "0 c E
FT
BCE .r @B @t E/ C c 2 B .r B/ c2j B E rE D 1 r .E E/ 2 E rE 1 r .EE/ C r .E E/ C .r E/E 2 1 13 .E E/ C E r EE 2 f D ECj B gmech .t; x/ D %m .t; x/v
(4.48)
A combination of formul (F.147) on page 217, (F.133), and (F.140) on page 217, and formula (F.159) on page 217, using the Maxwell-Lorentz equations (2.1) on page 21 once more, allows us to write
R A
E .r E/ D r E E D D
(4.49)
and similarly for B. Dividing out c 2 , and rearranging terms, we can write equation (4.48) above as h1 i 1 @S C "0 r 13 E E C c 2 B B EE C c 2 BB C f D 0 (4.50) c 2 @t 2 where (4.51)
The polar vector f has the dimension Nm 3 and we therefore identify it as a force density and call it the Lorentz force density. According to classical mechanics (Newtons second law, Eulers rst law), this mechanical force density must equal the time rate change of the mechanical linear momentum density (4.52)
j 67
where %m is the (volumetric) mass density (kgm 3 ). In the same vein, we introduce the electromagnetic linear momentum density geld by making the identication geld .t; x/ D "0 E or, alternatively, S D geld c 2 (4.54) BD S.t; x/ c2 (4.53)
@gmech @geld C Cr TD0 @t @t where

T D 13
"0 E E C c2B B 2
R A
Tij D ij ueld "0 Ei Ej 1 Bi Bj
0
is the electromagnetic linear momentum ux tensor (the negative of the Maxwell stress tensor), measured in Pa, i.e., Nm 2 . The tensor T is the electromagnetic linear momentum current density of the electromagnetic eld. In tensor component form, it can be written (4.57)
D
where
where the electromagnetic energy density ueld is dened in formula (4.26) on page 62. The component Tij is the electromagnetic linear momentum in the ith direction that passes a surface element in the j th direction per unit time, per unit area. Integration of equation (4.55) above over the entire nite volume V , enclosed by the surface S , yields the conservation law for linear momentum dpmech dpeld C C dt dt I
S
O d2x n T D 0
pmech D
Z
V
d3x gmech D
Z
V
FT
(4.55) "0 EE C c 2 BB (4.56) (4.58) d3x %m v (4.59a)
Since the LHS of this equation is (energy density) (velocity) and the RHS is (mass density) (velocity) c 2 , we see that this relation forebodes the relativistic relation E D mc 2 , where E is energy and m is mass. We are now able to write down the linear momentum density balance equation (4.48) on the preceding page in differential form. It is
68 j
and p or Z
V eld
Z D
V
d xg
eld
Z D
V
d3x "0 .E
B/
(4.59b)
Force on the matter
d xf
d C dt
Z
V
d x "0 .E
I B/ C
EM eld momentum
Linear momentum ow
O d2x n T
S
D0
(4.60)
R A
V V V V
which is the Lorentz force; see also equation (1.46) on page 15. Note that equation (4.61) above follows directly from a conservation law, and therefore is a consequence of a symmetry of the microscopic Maxwell equations (2.1) on page 21. Hence, the Lorentz force does not have to be presupposed. Using formula (4.28d) on page 62 in formula (4.59b), the Maxwell-Lorentz equation (2.1a) on page 21, and applying the divergence theorem, we can write the electromagnetic linear momentum of the eld as Z Z I O peld D d3x A C "0 d3x .r A/ E "0 d2x n EA (4.62)
S
For each spatial Fourier component we use equation (4.29d) on page 63 to nd the manifestly gauge-independent expression Z Z I O peld D d3x A? C "0 d3x .r A? / E "0 d2x n EA? (4.63)
S
If we restrict ourselves to consider a single temporal Fourier component of the transverse component and use the results in example 3.2 on page 48 we nd that ED @A D i!A @t (4.64)
which allows us to replace A by iE=!, yielding Z I i"0 i"0 O peld D Re d3x .r E/ E C Re d2x n EE trans 2! V ! S
FT
V
This is the linear momentum theorem in Maxwells theory which describes how the electromagnetic eld carries linear momentum. If we assume that we have a single localised charge q, i.e., that the charge density is given in terms of a Dirac distribution as in equation (1.8) on page 5, where the sum is over one particle only, the evaluation of the rst integral in equation (4.60) above shows that the force on this single charged particle is Z Z F D d3x f D d3x . E C j B/ D q.E C v B/ (4.61)
(4.65)
j 69
If EE falls off sufciently rapidly at large distances, we can discard the surface integral term and nd that the cycle averaged linear momentum of the transverse eld is Z i"0 eld ptrans D Re d3x .r E/ E (4.66) 2! V where we used formula (2.6) on page 24. In complex tensor notation with Einsteins summation convention applied, this can be written Z i"0 eld ptrans D d3x Ei r Ei (4.67) 2! V If we recall that the quantal linear momentum operator is
we see that we can write the expression for the linear momentum of the electromagnetic eld in terms of this operator as Z "0 y d3x Ei pEi (4.69) peld D trans 2 ! V i.e., as a quantum expectation value where the components of the electric eld vector behave as wavefunctions.
Conservation of the total linear momentum in a closed system
Show, by explicit calculation, that the total linear momentum p D of a closed electromechanical system comprising N non-relativistic particles of mass mi , speed vi and charge qi , and pertinent electromagnetic elds E and B,
R
pD
N X i D1 iD1
is conserved.
To show that the total linear momentum p of the closed system is conserved, i.e., is constant in time, is to show that the derivative of p with respect to time t vanishes. Direct differentiation yields Z N X dvi @E dp D mi C "0 d3x dt dt @t V BCE @B @t (4.71)
From the Lorentz force equation (4.61) on the preceding page and Newtons second law we nd that dvi qi D E.t; xi / C vi dt mi B.t; xi /
Substitution of this expression and equations (4.44) on page 65 into equation (4.71) yields,
A FT
y pD i r (4.68)
pmech C peld Z
V
E X A M P L E 4.4
mi vi C "0
d3x .E
B/
(4.70)
70 j
after some rearrangement of terms, Z N X dp D qi E.t; xi / C vi B.t; xi / d3x j dt V i D1 Z "0 d3x E .r E/ C c 2 B .r B/

V
B.t; x/ (4.72)
By rst applying formula (F.148) on page 217 and then formula (F.140) on page 217 on the last term, we nd that Z N X dp D qi E.t; xi / C vi B.t; xi / d3x j B.t; x/ dt V iD1 Z "0 d3x .Er E C c 2 Br B/ ZV "0 d3x 1 r .E E C c 2 B B/ r .EE C c 2 BB/ 2
V
(4.73)
We can now use the Maxwell-Lorentz equations to make the substitutions "0 r E D r B D 0 to obtain Z N X dp qi E.t; xi / C vi B.t; xi / d3x E C j B.t; x/ D dt V i D1 Z "0 d3x r 1 .E E C c 2 B B/13 .EE C c 2 BB/ 2
V
A FT
D
N X
and
(4.74)
Since the current is carried by discrete charged particles the current density j can be represented as formula (4.46) and the charge density as
qi .x
xi /
(4.75)
R
dp D dt I "0
S
iD1
the two rst terms in equation (4.74) above cancel and, after applying Gausss formula (F.177) on page 218, we are left with O d2x n
1 2
.E E C c 2 B B/13
.EE C c 2 BB/ D
I "0
S
O d2x n T (4.76)
where the tensor T is the electromagnetic linear momentum current density (the negative of the Maxwell stress tensor); cf. equation (4.56) on page 67. This result shows that the rate at which the total linear momentum is lost in a volume equals the amount of linear momentum ux that ows outward across a closed surface enclosing this volume. If the surface lies entirely outside the boundaries of the system under study, and this system is closed, no linear momentum ux passes through the surface and hence dp D0 dt
j 71
showing that indeed the total linear momentum p D pmech C peld of a closed system is a conserved quantity. QED If our system under study can be considered to be a closed electromechanical system and we observe a change in the electromagnetic linear momentum in this system, we can deduce that there has been a similar change, but with opposite sign, of the mechanical linear moment of the system. This allows us to determine mechanical properties of a system by analysing the radiation (radio, light, . . . ) from it. In the quantum picture the linear momentum of a photon is k where k is the wavenumber and therefore inversely proportional to the radiated wavelength. The wavelength shift caused by a change in the electromagnetic linear momentum is the translational Doppler shift. Since w D ck for a photon in free space, we experience the change of electromagnetic linear momentum as an associated frequency shift, a redshift if it is to the long wavelength (low frequency) side and a blueshift if it is to the short wavelength (high frequency) side. This vocabulary is used regardless of whether the radiation falls in the optical range or not. End of example 4.4
4.3.5 Conservation of angular momentum
R A
J mech .x0 / D mech C Lmech .x0 / .x0 / D dJ mech .x0 / dt
Already 1776 it was shown by L E O N H A R D E U L E R that the most general dynamical state of a mechanical system is the sum of its translational motion, described by the linear momentum mechanical linear momentum pmech , and its rotational motion, described by the systems mechanical moment of momentum, or mechanical angular momentum J mech .x0 / about the moment point x0 , and that these two momenta are in general independent of each other 4 . In the special case of a classical rigid body for which the contribution from internal angular momenta cancel the angular momentum around x0 is given by the pseudovector J mech .x0 / D .x x0 / pmech . For a closed system of rotating and orbiting bodies, e.g., a spinning planet orbiting a (non-rotating) star, the total mechanical angular momentum of the system is the vectorial sum of two contributions. (4.77)
D
mech
where mech is the intrinsic mechanical spin angular momentum, describing the spin of the planet around its own axis, and Lmech is the extrinsic mechanical orbital angular momentum, describing the motion of the planet in an orbit around the star. As is well known from mechanics, J mech of a closed mechanical system is conserved and to change it one has to apply a mechanical torque (4.78)
Starting from the denition of the mechanical linear momentum density g , formula (4.52) on page 66, and suppressing the obvious dependence of
FT
4
Rational mechanist C L I F F O R D A M B RO S E T RU E S D E L L I I I (19192000) wrote several excellent accounts on how this result was arrived at by Euler in the latter half of the 18th century.
72 j
x, we dene the mechanical angular momentum density about a xed moment point x0 as hmech .x0 / D .x x0 / gmech (4.79)
Analogously, we dene the electromagnetic moment of momentum density or electromagnetic angular momentum density about a moment point x0 as the pseudovector held .x0 / D .x x0 / geld (4.80)
@hmech .x0 / @ D .x @t @t C.x
FT
x0 / gmech D dx dt gmech f x0 / @gmech D .x @t x0 / x0 / geld D dx dt geld C .x x0 / x0 / x0 / @ .E @t B/ D 1 .x c2 @S @t
K.x0 / D .x
where, according to equation (4.54) on page 67, geld D "0 E B D S=c 2 . Now, if f is the Lorentz force density as dened in equation (4.51) on page 66, then
(4.81)
is the Lorentz torque density. Furthermore, @held .x0 / @ D .x @t @t D "0 .x
@geld @t
R A D
(4.82)
where we used the fact that dx=dt is the energy velocity S=ueld D geld c 2 =ueld . By introducing the electromagnetic angular momentum ux tensor x0 /
T
(4.83)
where T is the electromagnetic linear momentum ux tensor given by expression (4.56) on page 67, the differential form of the balance equation for angular momentum density can be written @hmech .x0 / @held .x0 / C C r K.x0 / D 0 @t @t (4.84)
where the pseudotensor K.x0 / represents the electromagnetic angular momentum current density around x0 . Integration over the entire volume V , enclosed by the surface S , yields the conservation law for angular momentum I dJ mech .x0 / dJ eld .x0 / O C C d2x n K.x0 / D 0 (4.85) dt dt S
j 73
where the mechanical and electromagnetic angular momentum pseudovectors are Z J mech .x0 / D d3x hmech .x0 / (4.86a)
V
and J
eld
.x0 / D
d x h .x0 / D d3x .x V Z D "0 d3x .x x0 / .E B/

eld V V
x0 /
geld (4.86b)
Torque on the matter
Field angular momentum
Angular momentum ow
D
V
We call it the Lorentz torque. If we study formula (4.86b) above a little closer, we see that it can be written Z J eld .x0 / D J eld .0/ x0 "0 d3x E B D J eld .0/ x0 peld (4.89)
V
where use was made of equation (4.59b) on page 68. With the help of equation (4.28) on page 62 we can write Z Z 1 J eld .0/ D "0 d3x x .E B/ D 2 d3x x S (4.90) c V V Z Z D "0 d3x x .r A/ E "0 d3x x E .r A/ (4.91)
V
R A
This angular momentum theorem is the angular analogue of the linear momentum theorem, equation (4.60) on page 68. It shows that the electromagnetic eld, like any physical eld, can carry angular momentum. For a single localised charge q, i.e., for a charge density given by equation (1.8) on page 5 with summation over one particle only, the evaluation of the rst integral in equation (4.87) above shows that the mechanical torque on this single charged particle is Z Z 3 .x0 / D d x .x x0 / f D d3x ..x x0 / . E C j B// V V (4.88) D .x x0 / q.E C v B/
FT
I (4.87) O d2x n K.x0 / D 0 S
respectively. We can formulateand interpretthis conservation law in the following way: Z Z d 3 d x .x x0 / f C "0 d3x .x x0 / .E B/ dt V V
74 j
Partial integration of this yields Z Z J eld .0/ D "0 d3x E A C "0 d3x x .r A/ E V V Z Z 3 "0 d x r .Ex A/ C "0 d3x .x A/.r E/
V V
(4.92)
By introducing and
eld
Z D "0
V
d3x E
(4.93a)
FT
.r A/ E C A
V S
Leld .0/ D "0
d3x x
d3x x
(4.93b)
applying Gausss theorem, formula (F.177) on page 218, and using the MaxwellLorentz equation (2.1a) on page 21, we nd that expression (4.92) above can be written I eld eld eld O J .0/ D C L .0/ "0 d2x n .Ex A/ (4.94) If there is no net electric charge density in the integration volume, the second integral in the RHS of equation (4.93b) above vanishes, and if Ex A E.x A/ falls off sufciently fast with jxj, the contribution from the surface integral in equation (4.94) above can be neglected. Furthermore, for a single Fourier component and in the complex notation, we obtain the expressions Z "0 eld D i d3x .E E/ (4.95a) 2! V Z "0 Leld .0/ D i d3x Ei .x r /Ei (4.95b) 2! V Recalling that in quantum mechanics the spin angular momentum operator is b j k D i O ij k xi (4.96)
R A
y LD i x
which, with the help of the matrix vector expression (M.33) on page 228 can be written b D
S
(4.97)
and the orbital angular momentum operator is r D i O ij k xj @k xi (4.98)
j 75
we can write Z "0 b D d3x Ej j k Ek 2 ! V Z "0 y d3x Ei LEi Leld .0/ D 2 ! V

eld
(4.99a) (4.99b)
Hence, under the assumptions made above, we can interpret eld as the electromagnetic spin angular momentum and Leld as the electromagnetic orbital angular momentum.
Conservation of the total angular momentum of a closed system Show, by explicit calculation, that the total angular momentum around a momemtum point x0 , J .x0 / D J mech .x0 / C J eld .x0 /, of a closed electromechanical system comprising N non-relativistic particles of mass mi , speed vi and charge qi , and pertinent electromagnetic elds E and B, Z J .x0 / D .xi x0 / mi vi C "0 d3x .x x0 / .E B/ (4.100)
V
E X A M P L E 4.5
is conserved.
To show that the total angular momentum J of the closed system is conserved, i.e., is constant in time, is to show that the derivative of J with respect to time t vanishes. For simplicity we put the origin at the moment point x0 . Direct differentiation then yields
N X dJ dxi D mi vi Cxi dt dt i D1 0
Z C "0 D
N X V
d3x
h dx dt
R
V
dvi C "0 dt i D1 Z h @E C "0 d3x x @t V mi xi
From the Lorentz force equation (4.61) on page 68 and Newtons second law we nd that qi dvi D E.t; xi / C vi dt mi B.t; xi /
Substitution of this expression and equations (4.44) on page 65 into equation (4.101) yields, after some rearrangement of terms,
N X dJ D qi xi E.t; xi / C xi vi B.t; xi / dt iD1 Z Z C "0 d3x v .E B/ d3x x .j B/ V V Z "0 d3x x E .r E/ C c 2 B .r B/
By rst applying formula (F.148) on page 217 and then formula (F.140) on page 217 on the
A FT
dvi dt .E B/ C x d3x v @E @t B Cx E @B i @t (4.101) Z .E B/
V
B C E
@B i @t
(4.102)
76 j
last term, we nd that

N X dJ D xi qi E.t; xi / C vi B.t; xi / dt iD1 Z Z C "0 d3x v .E B/ d3x x .j B/ V V Z "0 d3x x .Er E C c 2 Br B/ ZV 2 1 "0 d3x x r .EE C c 2 BB/ 2 r .E E C c B B/ V
(4.103)
We use the Maxwell-Lorentz equations to make the substitutions "0 r E D and obtain Z N X dJ D xi qi E.t; xi / C vi B.t; xi / d3x x E C .j dt V i D1 Z C "0 d3x v .E B/ ZV "0 d3x x r 1 .E E C c 2 B B/13 .EE C c 2 BB/ 2
V
and r B D 0
A FT
B/ dJ D dt I "0
S
(4.104)
Since the current is carried by discrete charged particles, the current density j can be represented as formula (4.46) on page 65 and the charge density as formula (4.75) on page 70. Hence, the two rst terms in equation (4.104) above cancel and, after applying Gausss formula (F.177) on page 218, we are left with O d2x n K (4.105)
where the tensor K is the electromagnetic angular momentum current density; cf. equation (4.83) on page 72. This result shows that the rate at which the total angular momentum is lost in a volume equals the amount of angular momentum ux that ows outward across a closed surface enclosing this volume. If the surface lies entirely outside the boundaries of the system under study, and this system is closed, no angular momentum ux passes through the surface and hence dJ D0 dt
showing that indeed the total angular momentum J D J mech C J eld of a closed system is a conserved quantity. QED If we observe a change in the electromagnetic angular momentum in a given system, we can deduce that there has been a similar change, but with opposite sign, of the mechanical angular moment of the system. End of example 4.5
j 77
Spin angular momentum and wave polarisation Consider a generic temporal Fourier mode of the electric eld vector E of a circularly polarised wave with (angular) frequency !. According to equation (2.35) on page 29 it can be written O E.t; x/ D E.t; x/ h where E.t; x/ D and p 2E0 ei.kx3
!tC1 /
E X A M P L E 4.6
(4.106)
(4.107)
O O As before, we use the convention that hC represents left-hand circular polarisation and h right-hand circular polarisation. Noting that O . h / O h D iO z
we see that E
R
eld D "0 2! Z
V
When this is inserted into equation (4.95a) on page 74 one nds that the spin angular momentum of a circularly polarised wave is U eld O z !
where U eld is the eld energy. Considering the wave consisting of N photons, then U eld D N ! which means that the spin of the wave is eld D N ! O O z D N z !
Hence, each photon of a right-hand or a left-hand circular polarised wave carries a spin angular momentum of or , respectively. End of example 4.6
A FT
1 O O h D p x1 iO 2 / x 2 (4.108) (4.109) O O E D i jEj2 z D iE 2 z (4.110) O d3x E 2 z D (4.111) (4.112)
78 j
E X A M P L E 4.7
Orbital angular momentum y The Cartesian components of the quantal orbital angular momentum operator (OAM) L as given by expression (4.98) on page 74 are @ @ y Lx D i y z (4.113a) @z @y @ @ y Ly D i z x (4.113b) @x @z @ @ y Lz D i x y (4.113c) @y @x In cylindrical coordinates . ; '; z/ they are @ yx D i sin ' z L @ @ y Ly D i cos ' z @
A FT
y Lz D y Lx D i sin ' y Ly D i @ @ C cos ' cot @ @' @ @ cos ' sin ' cot @ @' @ y Lz D i @' E D E0 .t; x/.'/ y Lz E.t; x/ D i @E0 .t; x/ .'/ @' E0 .t; x/ @.'/ @' .'/ D
1 X mD 1
@ z @ C cos ' @z @' @ z @ sin ' @z @' @ i @'
(4.114a) (4.114b) (4.114c)
and in spherical coordinates .r; '; / they are
(4.115a) (4.115b) (4.115c)
For an electric eld E that depends on the azimuthal angle ' in such a way that
(4.116)
R D
we see that
we nd that
(4.117)
If E0 .t; x/ is rotationally symmetric around the the z axis, so that E0 D E0 .t; ; z/ in cylindrical coordinates, E0 D E0 .t; r; / in spherical (polar) coordinates etc., and if the azimuthal part is expressed in a Fourier series cm eim' (4.118)
y Lz E.t; x/ D
1 X mD 1
cm m E0 eim' D
1 X mD 1
cm m Em
(4.119)
i.e., that E is a weighted superpostion of OAM eigenstates Em D E0 eim' . Furthermore, y Lz Em D m Em (4.120)
j 79
which means that for a rotationally symmetric beam with an azimuthal phase dependence given by exp.im'/ one deduces, according to formula (4.95b) on page 74, that the z component of the orbital angular momentum of each photon in this beam is m . End of example 4.7
4.3.6 Conservation of centre of energy

In analogy with the existence of a centre of mass in a classical mechanical system, there exists a centre of energy in electrodynamics. It can be shown that this vectoral quantity is also conserved.
Examples of other conservation laws In addition to the conservation laws for energy, linear momentum and angular momentum, a large number of other electromagnetic conservation laws can be derived. For instance, as can be shown by direct calculation based on the Maxwell-Lorentz the following conservation law holds in free space (vacuum): @ Cr XD0 @t E X A M P L E 4.8
where the pseudoscalar
def
E .r
is the chirality density, and the pseudovector XE is the chirality ow.

def
If instead of vector multiplying the linear momentum densities gmech and geld by the position vector relative to a x point x0 as we did in subsection 4.3.5 on page 71, we scalar multiply, the following differential balance equation is obtained:
D
@ .x
R
x0 / gmech @ .x C @t
4.3.7 Electromagnetic virial theorem
This is the electromagnetic virial theorem, analogous to the virial theorem of Clausius in mechanics. The quantity .x x0 / geld is the electromagnetic virial density. When integrated over space and time averaged, this theorem is a statement of the partitioning of energy in electrodynamics and nds use in, e.g., plasma physics.
A FT
(4.121) E/ C c 2 B .r B/ (4.122) @E C c2B @t @B @t (4.123) End of example 4.8
x0 / geld Cr @t
.x
x0 / T D ueld (4.124)
80 j
4.4 Bibliography
[25] J. D. JACKSON, Classical Electrodynamics, third ed., John Wiley & Sons, Inc., New York, NY . . . , 1999, ISBN 0-471-30932-X. [26] F. M ELIA, Electrodynamics, Chicago Lectures in Physics. Cambridge University Press, Cambridge and London, 1991, ISBN 0-226-51957-0.
R A
FT
FIELDS FROM ARBITRARY CHARGE AND CURRENT DISTRIBUTIONS
Whereas the electric and magnetic elds E and B generated by prescribed chrarge and current sources and j canat least in principlebe obtained by directly solving the Maxwell-Lorentz differential equations given at the beginning of chapter 2 on page 21, or the wave equations given later in the same chapter, it is often technically easier and physically more lucid to calculate the elds from the electromagnetic potentials and A. As shown in chapter 3 on page 35, these potentials can in Lorenz-Lorentz gauge be readily obtained in the form of an integral over the spatial distribution of the source elements, divided by the distance between the actual source element and the observer. In this chapter we will use the electromagnetic potentials to derive exact, closed-form, analytic expressions for the electric and magnetic elds generated by prescribed but completely arbitrary charge and current distributions located in otherwise free space. As we shall nd, these electric and magnetic elds are made up of several parts, each characterized by its particular direction and falloff with distance relative to the source such that in free space outside the source region, different eld components dominate in different zones. These zones are customarilyand self-explanatorilyreferred the near zone, the intermediate zone, and the far zone, respectively. Those parts of the eld that have the slowest fall-off as a function of distance from the source and therefore dominate in the far zone, are referred to as the far elds. Because of their importance, a special analysis of these far elds is given at the end of the chapter. Surprising as it may seem, it will be shown in chapter 6 on page 95 that certain physical observables receive their dominant far-zone contributions not from the dominant far elds but from a combination of far elds and sub-dominant near-zone and intermediate-zone elds. For completeness, we therefore include a derivation of approximate expressions for all eld components, dominant as well as sub-dominant, valid at large distances from an arbitrary source.
R A
81
FT
82 j
5 . F I E L D S F RO M A R BITRARY CHARGE AND CURRENT DISTRIBUTIONS
5.1 Fourier component method

As discussed earlier, the linearity of the Maxwell-Lorentz equations (2.1) on page 21 means that the complete solution of them can be found in terms of a superposition of Fourier components, each of which solves these equations; the resulting superposition is called a wavepacket. We recall that in order to use the Fourier component method to nd the solution (3.49) on page 44 for the generic inhomogeneous wave equation (3.33) on page 41, we presupposed the existence of the temporal Fourier transform pair equations (3.34) on page 41 for the generic source term: Z 1 f .t; x/ D d! f! .x/ e i!t (5.1a) 1 Z 1 1 f! .x/ D dt f .t; x/ ei!t (5.1b) 2 1 That such transform pairs exist is true for most non-erratic physical variables which are neither strictly monotonically increasing nor strictly monotonically decreasing with time. For charge and current densities that vary in time we can therefore, without loss of generality, work with individual temporal Fourier components ! .x/ and j! .x/, respectively. Strictly speaking, the existence of a signal represented by a single Fourier component assumes a monochromatic source (i.e., a source containing only one single frequency component), which requires that this source must have existed for an innitely long time. However, by taking the proper limits, we may still use this approach even for sources and elds of nite duration. This is the method we shall utilise in this chapter in order to derive the electric and magnetic elds in vacuum from arbitrary given charge densities .t; x/ and current densities j.t; x/, dened by the temporal Fourier transform pairs Z 1 .t; x/ D d! ! .x/ e i!t (5.2a) 1 Z 1 1 dt .t; x/ ei!t (5.2b) ! .x/ D 2 1 and j.t; x/ D d! j! .x/ e i!t 1 Z 1 1 j! .x/ D dt j.t; x/ ei!t 2 1 Z
1
R A
FT
(5.3a) (5.3b)
respectively. The derivation will be completely general except we make the reasonable assumption that only retarded potentials produce physically acceptable solutions and that the source region is at rest relative to the observer.
5.1. Fourier component method
j 83
The temporal Fourier transform pair for the retarded scalar potential can then be written Z 1 .t; x/ D d! ! .x/ e i!t (5.4a)
1
! .x/ D
1 2
1 1
dt .t; x/ ei!t D
1 4 "0
Z
V0
d3x 0
0 ! .x /
eikjx x j jx x0 j
(5.4b)
1 A! .x/ D 2
1 1
dt A.t; x/ ei!t D
1 4 "0 c 2
Analogous transform pairs exist for the elds themselves. In the limit that the sources can be considered monochromatic containing only one single frequency !0 , we have the much simpler expressions .t; x/ D
0 .x/e i!0 t
R A
.t; x/ D 0 .x/e A.t; x/ D A0 .x/e
i!0 t i!0 t
j.t; x/ D j0 .x/e
where again the real-valuedness of all these quantities is implied. As discussed above, we can safely assume that all formul derived for a general temporal Fourier representation of the source (general distribution of frequencies in the source) are valid for these simple limiting cases. We note that in this context, we can make the formal identication ! D 0 .! !0 /, j! D j0 .! !0 / etc., and that we therefore, without any essential loss of stringency, let 0 mean the same as the Fourier amplitude ! and so on. In order to simplify the computations, we will work in ! space and, at the nal stage, inverse Fourier transform back to ordinary t space. We shall be using the Lorenz-Lorentz gauge and note that, in ! space, the Lorenz-Lorentz condition, equation (3.31) on page 41, takes the form r A! k i ! D 0 c (5.7)
This provides a relation between (the temporal Fourier transforms of) the vector and scalar potentials A! and ! .
FT
Z
V0
where in the last step, we made use of the explicit expression for the temporal Fourier transform of the generic potential component ! .x/, equation (3.46) on page 43. Similarly, we must require that the following Fourier transform pair for the vector potential exist: Z 1 A.t; x/ D d! A! .x/ e i!t (5.5a) d3x 0 j! .x0 / eikjx x j (5.5b) jx x0 j
0
(5.6a)
i!0 t
(5.6b)
(5.6c) (5.6d)
84 j
5.2 The retarded electric eld

In order to calculate the retarded electric eld, we use the temporally Fourier transformed version of formula (3.28) on page 40, with equations (5.4) on the preceding page as the explicit expressions for the Fourier transforms of and A: E! .x/ D r ! .x/ C i!A! .x/ Z 0 1 eikjx x j D r d3x 0 ! .x0 / 4 "0 jx x0 j V0 Z 0 i! eikjx x j C d3x 0 j! .x0 / 4 "0 c 2 V 0 jx x0 j Z 0 ikjx x0 j 1 .x x0 / ! .x /e D d3x 0 4 "0 V 0 jx x0 j3 Z 0 ikjx x0 j ik .x x0 / 3 0 ! .x /e dx 4 "0 V 0 jx x0 j2 Z ikjx x0 j ik 3 0 0 e C d x j! .x / 4 "0 c V 0 jx x0 j
FT
r 0 j! .x0 / i!
! .x 0
(5.8)
R A
Taking the inverse Fourier transform of this expression in an analogous manner to what was done to obtain equation (5.24) on page 88, the following expression for the retarded electric eld is obtained: Z 0 1 .tret ; x/ x x0 E.t; x/ D d3x 0 4 "0 V 0 jx x0 j2 jx x0 j Z 0 1 P.tret ; x0 / x x0 C d3x 0 (5.9) 4 "0 c V 0 jx x0 j jx x0 j Z P 0 1 j.tret ; x0 / d3x 0 4 "0 c 2 V 0 jx x0 j
0 0 Letting dq 0 D d3x 0 .tret ; x0 / and dI 0 D d3x 0 j.tret ; x0 /, the corresponding formula in innitesimal differential form becomes 1 x x0 1 x x0 1 P0 1 C dq 0 dE.t; x/ D dq 0 P dI (5.10) 4 "0 c jx x0 j jx x0 j3 jx x0 j2 c 2
which in the static limit reduces to the innitesimal Coulomb law, formula (1.5) on page 4. We shall now further expand equation (5.9) above. To this end we rst note that the Fourier transform of the continuity equation (4.22) on page 61 /D0 (5.11)
5.2. The retarded electric eld
j 85
can be used to express
in terms of j! as follows
! .x 0
/D
i 0 r j! .x0 / !
(5.12)
Doing so in the last term of equation (5.8) on the preceding page, and also using the fact that k D !=c, we can rewrite this equation as E! .x/ D Z 0 ikjx x0 j 1 .x x0 / ! .x /e d3x 0 4 "0 V 0 jx x0 j3 0 ikjx x0 j Z 1 r j! .x0 /.x x0 / e d3x 0 ikj! .x0 / 0j c V0 jx x jx x0 j K! (5.13)
The last vector-valued integral can be further rewritten in the following way (where l and m are summation indices and Einsteins summation convention is assumed): Z K! D Z D
V0 V0
dx dx
3 0
r 0 j! .x0 /.x jx x0 j @j!;l xm 0 @xl jx

0 xm x0 j
x0 /
3 0
But, since 0 @ xm xm ikjx e j!;l 0 @xl jx x0 j2
R A
x0 j
eikjx x j ikj! .x / jx x0 j ikjx x0 j e O ikj!m .x0 / xm jx x0 j

0
@j!;l 0 @xl
C j!;l
@ 0 @xl
we can rewrite K! as " Z

3 0
D
V0
@ xm K! D d x j!;l 0 @xl jx V0 Z @ xm C d3x 0 0 j!;l @x jx V0 " l Z D d3x 0 j! .x0 / r 0 Z C

V0
0 xm
x0 j2
0 xm
O xm e
x0 j 2 x jx
O xm e x0 x0 j 2
ikjx x0 j
d3x 0 r 0
j! .x0 /
x jx
x0 x0 j 2
eikjx
FT
(5.14)
xm
0 xm
jx
x0 j2
eikjx
x0 j
xm
0 xm
jx
x0 j2
eikjx
x0 j
(5.15)
ikjx x0 j
eikjx x j C ikj! jx x0 j
eikjx x j C ikj! .x / jx x0 j
0
0
ikjx x0 j
x0 j
(5.16)
86 j
where, according to Gausss theorem, the last term vanishes if the dyadic inside the big parentheses is limited and tends to zero at large distances. Further evaluation of the derivative in the rst term makes it possible to write
Z K! D 2
V0
d3x 0 j! .x0 / .x d3x 0 j! .x0 / .x

0
x0 / .x x0 / .x
x0 /
eikjx jx jx eikjx
x0 j
x0 j4
x0 j
Z C ik Z C
V0 V0
x0 /
x0 j3
d x j! .x /
3 0
eikjx jx
x0 j
(5.17)
x0 j2 eikjx x j jx x0 j
0
Z ik
V0
d3x 0 j! .x0 /
Using the triple product bac-cab formula (F.108) on page 216 backwards, and inserting the resulting expression for K! into equation (5.13) on the previous page, we arrive at the following nal expression for the Fourier transform of the total E eld:
R A
E! .x/ D 1 r 4 "0 Z d3x 0
V0
i! eikjx x j C d3x 0 j! .x0 / 4 "0 c 2 V 0 jx x0 j Z 0 ikjx x0 j 1 .x x0 / ! .x /e D d3x 0 4 "0 V 0 jx x0 j3 Z 0 1 j! .x0 /eikjx x j .x x0 /.x x0 / C d3x 0 c V0 jx x0 j4 Z 0 1 j! .x0 /eikjx x j .x x0 / .x x0 / C d3x 0 c V0 jx x0 j4 Z 0 ikjx x0 j ik .x x0 / .x x0 / 3 0 j! .x /e dx c V0 jx x0 j3 Z
FT
0 ! .x /
eikjx x j jx x0 j
0
(5.18)
Taking the inverse Fourier transform of equation (5.18) above, once again using the vacuum relation ! D kc, we nd, at last, the expression in time domain
5.3. The retarded magnetic eld
j 87
for the total electric eld: Z 1 E.t; x/ D d! E! .x/ e i!t 1 Z 0 .tret ; x0 /.x x0 / 1 d3x 0 D 4 "0 V 0 jx x0 j3
Retarded Coulomb eld
1 4 "0 c 1 4 "0 c
Z
V0
d3x 0
0 j.tret ; x0 / .x
x0 /.x x0 j 4
x0 / (5.19) x0 /
jx
0 j.tret ; x0 /
Intermediate eld
Z
V0
d3x 0
.x jx
x0 / x0 j4 x0 /
.x
Intermediate eld
1 C 4 "0 c 2 where, as before
Z
V0
d3x 0
P 0 j.tret ; x0 /
Far eld
def P 0 j.tret ; x0 /
@j @t
Here, the rst term represents the retarded Coulomb eld and the last term represents the far eld which dominates at very large distances. The other two terms represent the intermediate eld which contributes signicantly only to the elds themselves in the near zone and must be properly taken into account there.
5.3 The retarded magnetic eld
D
B! .x/ D r
Let us now compute the magnetic eld from the vector potential, dened by equation (5.5a) and equation (5.5b) on page 83, and formula (3.24) on page 39: B.t; x/ D r A.t; x/ (5.20)
Using the Fourier transformed version of this equation and equation (5.5b) on page 83, we obtain A! .x/ D 1 r 4 "0 c 2 Z
V0
R A
d3x 0 j! .x0 /
FT
.x .x x0 / jx x0 j3
0 tDtret
eikjx x j jx x0 j
(5.21)
88 j
1 Equation (5.9) and equation (5.24) seem to have been rst introduced by Panofsky and Phillips. Later they were given by Jemenko and they are therefore sometimes called the Jemenko equations.
R A
1 C 4 "0 c 3 Z where
From this expression for the magnetic eld in the frequency (!) domain, we nally obtain the total magnetic eld in the temporal (t ) domain by taking the inverse Fourier transform (using the identity ik D i!=c): Z 1 B.t; x/ D d! B! .x/ e i!t 1 hR i 0 1 Z d! j! .x0 /e i.!t kjx x j/ .x x0 / 1 1 3 0 D dx 4 "0 c 2 V 0 jx x0 j3 i hR 0 1 Z .x x0 / d! . i!/j! .x0 /e i.!t kjx x j/ 1 1 3 0 C dx c V0 jx x0 j2 (5.23)
Comparing with equations (3.48) on page 44, we can identify the exponents 0 of the exponentials in the integrands as i!tret and nd that the total retarded magnetic eld can be written as the sum of two terms1 Z 0 1 j.tret ; x0 / .x x0 / B.t; x/ D d3x 0 2 4 "0 c V 0 jx x0 j3
Retarded induction eld
FT
dx
V0 Far eld 3 0
Utilising formula (F.141) on page 217 and recalling that j! .x0 / does not depend on x, we can rewrite this as " !# Z 0 eikjx x j 1 3 0 0 B! .x/ D d x j! .x / r 4 "0 c 2 V 0 jx x0 j Z x x0 1 0 d3x 0 j! .x0 / eikjx x j D 4 "0 c 2 V 0 jx x0 j3 Z x x0 ikjx x0 j 1 3 0 0 (5.22) C d x j! .x / ik e 0 jx x0 j jx x0 j V Z 0 j! .x0 /eikjx x j .x x0 / 1 d3x 0 D 4 "0 c 2 V 0 jx x0 j3 Z 0 . ik/j! .x0 /eikjx x j .x x0 / C d3x 0 jx x0 j2 V0
P 0 j.tret ; x0 / jx
.x x0 j 2
x0 /
(5.24)
def P 0 j.tret ; x0 /
@j @t
(5.25)
0 tDtret
5.3. The retarded magnetic eld
j 89
The rst term, the retarded induction eld that dominates near the current source but falls off rapidly with distance from it, is the electrodynamic version of the Biot-Savart law in electrostatics, formula (1.17) on page 8. The second term, the far eld , dominates at large distances. The spatial derivatives (r ) gave rise to a time derivative (P) so this term represents the part of the magnetic eld that is generated by the time rate of change of the current, i.e., accelerated charges, at the retarded time. In innitesimal differential form, formula (5.24) on the preceding page becomes 1 x x0 1 P0 x x0 0 dB.t; x/ D dI (5.26) C dI 4 "0 c 2 c jx x0 j3 jx x0 j2
Alternative expressions for E and B
The formul for the elds E and B expressed directly in integrals of generic source terms, can be rewritten in several ways. For instance, by introducing the vector .t 0 ; x0 ; x/ D
R
C 1 c jx c jx x0 j x0 j C 1
.t 0 ; x0 / x jx x0 j2 jx
D
where and the magnetic eld as
1 P j.t 0 ; x0 / c 2 jx x0 j x c
one can write the electric eld as E.t; x/ D 1 4 "0 Z

V0
A FT
x0 x0 j j.t 0 ; x0 / 2 j.t 0 ; x0 / 2 x jx x jx x jx x0 x0 x x0 j jx x0 x0 j x0 x0 j x0 x0 j x jx x jx x0 x0 j x0 x0 j (5.27)
0 t 0 D tret D t
0 where, as before, dI 0 D d3x 0 j.tret ; x0 /. This is the dynamic generalisation of the static innitesimal Biot-Savart law, equation (1.16) on page 8 to which it reduces in the static limit. With this we have achieved our goal of nding closed-form analytic expressions for the electric and magnetic elds when the sources of the elds are completely arbitrary, prescribed distributions of charges and currents. The only assumption made is that the advanced potentials have been discarded [recall the discussion following equation (3.49) on page 44 in chapter 3 on page 35].
E X A M P L E 5.1
(5.28)
d3x 0 .t 0 ; x0 ; x/
(5.29)
90 j
Figure 5.1: Relation between the O unit vector n, anchored at the observation point x and directed along x x0 , and hence normal to the surface S.x0 / which has its centre at x0 and passes through x, and the wave vector k.x0 /, directed along x x0 , of elds generated at the source point x0 near the point x0 in the source volume V 0 . At distances jx x0 j much larger than the exO tent of V 0 , the unit vector n and the O unit vector k.x0 / are nearly coincident.
O n O k.x0 /
S.x0 /
x x
x0 x0 x0 x0 x0 x0
B.t; x/ D
FT
V0
1 4 "0 c 2 Z
V0
d3x 0
x jx
x0 x0 j
.t 0 ; x0 ; x/
(5.30)
End of example 5.1
R A
5.4 The total elds at large distances from the sources

For many purposes it is convenient to have access to approximate expressions for the electric and magnetic elds that are valid in the far zone, i.e., very far away from the source region. Let us therefore derive such expressions. As illustrated in gure 5.1 we assume that the sources and j are located at xed points x0 near a xed point x0 inside a volume V 0 that is not moving relative to the observer. Hence the distance from each source point x0 to the observation point (eld point) x is assumend to be constant in time. The nonmoving source volume V 0 is located in free space and has such a limited spatial extent that sup jx0 x0 j inf jx x0 j, and the integration surface S.x0 /, cenO tred on x0 and with an outward pointing normal unit vector n D x x0 , has a 0 large enough radius jx x0 j sup jx x0 j. Noting that the (exact) wave vector for elds generated at x0 can be written x O k.x0 / D k k.x0 / k x x0 D k jx x0 x0 j (5.31)
expressing the fact that its magnitude k D !=c is constant but its direction is along x x0 and thus is dependent on the location of the source element at x0 in
5.4. The total elds at large distances from the sources
j 91
V 0 , we can, for the geometry just described, make the magnitude approximation k x x0 k.x0 / .x k jx x0 j x0 / k.x0 / .x k.x / .x
0 0
x0 /
k.x0 / .x0
x0 /
x0 /
(5.32)
in the phases (exponentials) and the cruder approximation x x0 jx x0 j (5.33)
E! .x/
1 eikjx x0 j 4 "0 jx x0 j2
Z
V0
d3x 0 Z
V0
! .x
/e
ik.x0 / .x0 x0 /
1 eikjx x0 j C 4 "0 c jx x0 j2 C 1 eikjx x0 j 4 "0 c jx x0 j2
d3x 0 j! .x0 /e d3x 0 j! .x0 /e
Z
V0
R A
ik eikjx x0 j 4 "0 c jx x0 j Z
V0
d3x 0 j! .x0 /e
B! .x/
1 eikjx x0 j 2 4 "0 c jx x0 j2
V0
d3x 0 j! .x0 /e
3 0 0
ik eikjx x0 j 4 "0 c 2 jx x0 j
V0
d x j! .x /e
D
x O k.x0 / D jx
At a eld (observation) point x located sufciently far away from the source volume V 0 such that sup jx0 x0 j sup jx0 x0 j, jx x0 j, and inf jx x0 j we can assume that the direction of all wave vectors from the sources in V 0 are parallel to each other. I.e., we can make the paraxial approximation x0 D x x0 x0 j x O k.x0 / D jx x0 Dx x0 j
O where n is the constant unit vector normal to the surface S.x0 / of a large sphere centred on x0 and passing through the (xed) eld point x (see gure 5.1 on
FT
O k.x0 /
ik.x0 / .x0 x0 /
in the amplitudes (denominators) in equation (5.18) on page 86, and equation (5.22) on page 88, and get the following approximate expressions for the Fourier amplitudes of the electric and magnetic elds, valid at sufcently large distances from the bounded source volume:
O O k.x0 / k.x0 / O k.x0 /
ik.x0 / .x0 x0 /
O k.x0 /
ik.x0 / .x0 x0 /
O k.x0 /
O k.x0 / (5.34a)
ik.x0 / .x0 x0 /
O k.x0 / (5.34b) O k.x /

0
ik.x0 / .x0 x0 /
O x0 n
(5.35)
92 j
O which, after reordering of scalar and vector products, using the fact that n is a constant unit vector, can be written Z 1 eikjx x0 j 0 O O E! .x/ d3x 0 ! .x0 /e ik n .x x0 / n 4 "0 jx x0 j2 V0 Z eikjx x0 j 1 O 3 0 0 ik n .x0 x0 / OO d x j! .x /e nn C 4 "0 c jx x0 j2 V0 Z eikjx x0 j 1 O 3 0 0 ik n .x0 x0 / O O C d x j! .x /e n n 4 "0 c jx x0 j2 V0 Z ik eikjx x0 j O 3 0 0 ik n .x0 x0 / O O d x j! .x /e n n 4 "0 c jx x0 j V0 (5.37a) ikjx x0 j Z e 1 0 O O d3x 0 j! .x0 /e ik n .x x0 / n B! .x/ 2 2 4 "0 c jx x0 j V0 (5.37b) Z ik eikjx x0 j O .x0 x0 / O d3x 0 j! .x0 /e ik n n 4 "0 c 2 jx x0 j V0 or E! .x/ 1 eikjx x0 j O Q! .x0 / n 4 "0 jx x0 j2 C 1 eikjx x0 j OO I ! .x0 / nn 4 "0 c jx x0 j2 O n/ O n/ O n O n
R A
FT
1 eikjx x0 j C .I ! .x0 / 4 "0 c jx x0 j2 ik eikjx x0 j .I ! .x0 / 4 "0 c jx x0 j
page 90). Then formul (5.34) on page 91 can be further approximated as Z 1 eikjx x0 j 0 O O d3x 0 ! .x0 /e ik n .x x0 / n E! .x/ 4 "0 jx x0 j2 V 0 Z 1 eikjx x0 j 0 O O O d3x 0 j! .x0 /e ik n .x x0 / n n C 4 "0 c jx x0 j2 V 0 (5.36a) Z 1 eikjx x0 j O 3 0 0 ik n .x0 x0 / O O C d x j! .x /e n n 4 "0 c jx x0 j2 V 0 Z ik eikjx x0 j 0 O O O d3x 0 j! .x0 /e ik n .x x0 / n n 4 "0 c jx x0 j V 0 Z 1 eikjx x0 j 0 O O B! .x/ d3x 0 j! .x0 /e ik n .x x0 / n 4 "0 c 2 jx x0 j2 V 0 (5.36b) Z ik eikjx x0 j O 3 0 0 ik n .x0 x0 / O d x j! .x /e n 4 "0 c 2 jx x0 j V 0
(5.38a)
5.4. The total elds at large distances from the sources
j 93
B! .x/
1 eikjx x0 j I ! .x0 / 4 "0 c 2 jx x0 j2 ik eikjx x0 j I ! .x0 / 4 "0 c 2 jx x0 j
O n (5.38b) O n
where Q! .x0 /
def
Z
V0
d3x 0
! .x
/e dx
O ik n .x0 x0 / O ik n x0
De and
V0
O ik n x0
Z
V0
(5.39a)
3 0
! .x
/e
V0
Inverse Fourier transforming these expressions we nd that the total retarded E and B elds very far away from a source to a good approximation are given by E.t; x/ 1 eikjx x0 j 4 "0 jx x0 j2 Z
V0
O d3x 0 .t 0 ; x0 /n Z Z
R A
1 eikjx x0 j C 4 "0 c jx x0 j2 eikjx x0 j 1 C 4 "0 c jx x0 j2 C
V0
O O d3x 0 j.t 0 ; x0 / nn (5.40) O n O n O n O n d x j.t ; x /

3 0 0 0
V0
eikjx x0 j 1 2 jx 4 "0 c x0 j
P d3x 0 j.t 0 ; x0 /
V0
and
D
respectively, where t 0 .x0 / D t C t
B.t; x/
eikjx x0 j 1 2 4 "0 c jx x0 j2
V0
1 eikjx x0 j 4 "0 c 3 jx x0 j
Z
V0
k .x0 x0 / ! k jx x0 j k jx x0 j C Dt ! !
FT
(5.39b) d3x 0 j.t 0 ; x0 / P d x j.t ; x /
3 0 0 0
I ! .x0 /
def
O d3x 0 j! .x0 / e ik n .x x0 / Z O O 0 D eik n x0 d3x 0 j! .x0 / e ik n x
O n (5.41) O n
jx
x0 j c
jx
x0 j
(5.42)
94 j
5.4.1 The far elds

When jx x0 j ! 1 and the source region is of nite extent, the only surviving components in expression (5.19) on page 87 and expression (5.24) on page 88 are the far elds Z P 0 1 j.tret ; x0 / .x x0 / .x x0 / far E .t; x/ D d3x 0 (5.43a) 4 "0 c 2 V 0 jx x0 j3 Z P 0 j.tret ; x0 / .x x0 / 1 d3x 0 (5.43b) Bfar .t; x/ D 3 4 "0 c V 0 jx x0 j2 In the frequency (temporal Fourier) domain, these far elds are represented exactly by Z 0 1 j! .x0 /eikjx x j .x x0 / .x x0 / Efar .x/ D ik d3x 0 ! 4 "0 c V 0 jx x0 j3 (5.44a) Z ikjx x0 j 1 3 0 e 0 O 0 / k.x0 / O dx j! .x / k.x D ik 4 "0 c V 0 jx x0 j Z 0 ikjx x0 j 1 .x x0 / 3 0 j! .x /e far dx B! .x/ D ik 4 "0 c 2 V 0 jx x0 j2 (5.44b) Z 0j ikjx x 1 3 0 e 0 0 O D ik dx j! .x / k.x / 4 "0 c 2 V 0 jx x0 j
R A
respectively. Within the approximation (5.32) on page 91, the expressions (5.44b) and (5.44a) for the far elds can be simplied to Z 1 eikjx x0 j 0 far O (5.45a) ik B! .x/ d3x 0 j! .x0 /e ik .x x0 / k 4 "0 c 2 jx x0 j V 0 Z 1 eikjx x0 j 0 far O O ik E! .x/ d3x 0 j! .x0 /e ik .x x0 / k k (5.45b) 4 "0 c jx x0 j V 0 Inverse Fourier transforming the above expressions into t domain, we nd that Z eikjx x0 j 1 far P O (5.46a) B .t; x/ d3x 0 j.t 0 .x0 /; x0 / k 4 "0 c 3 jx x0 j V 0 Z 1 eikjx x0 j P O O (5.46b) Efar .t; x/ d3x 0 j.t 0 .x0 /; x0 / k k 4 "0 c 2 jx x0 j V 0
where t 0 is given by the approximate equation (5.42) on the preceding page. In conclusion, if sup jx0 x0 j jx x0 j, then the far elds behave as spherical waves multiplied by dimensional and angular factor integrals, where the integrands contain only points in the source region. The magnitudes of the elds are directly proportional to the magnitude of the time derivative of the current.
FT
RADIATION AND RADIATING SYSTEMS
In chapter 3 on page 35 we were able to derive general expressions for the scalar and vector potentials from which we later (in chapter 5 on page 81) derived exact analytic expressions for the total electric and magnetic elds generated by completely arbitrary distributions of charge and current sources that are located in certain regions. The only limitation in the calculation of the elds was that the advanced potentials were discarded onadmittedly not totally convincing physical grounds. In chapter 4 on page 55 we showed that the electromagnetic energy, linear momentum, and angular momentum are all conserved quantities and in this chapter we will show that these quantities can be radiated all the way to innity and therefore be used in wireless communications over long distances and for observing very remote objects in Nature, including electromagnetic radiation sources in the Universe. Radiation processes are irreversible in that the radiation does not return to the radiator but is lost from it forever.1 However, the radiation can, of course, be sensed by other charges and currents that are located in free space, possibly very far away from the sources. This is precisely what makes it possible for our eyes to observe light, and even more so radio signals, from extremely distant stars and galaxies. This consequence of Maxwells equations, with the displacement current included, was veried experimentally by H E I N R I C H R U D O L F H E RT Z about twenty years after Maxwell had made his theoretical predictions. Hertzs experimental and theoretical studies paved the way for radio and TV broadcasting, radar, wireless communications, radio astronomy and a host of other applications and technologies. Thus, one can, at least in principle, calculate the radiated elds, ux of energy, linear momentum and angular momentum, as well as other electromagnetic observables at any time at any point in space for an arbitrary charge and current density of the source. However, in practice, it is often difcult to evaluate the source integrals unless the charge and current densities have a simple distribution in space. In the general case, one has to resort to approximations. We shall consider both these situations in this chapter.
R A
95
FT
1 This is referred to as time arrow of radiation.
96 j
6 . R ADIATION AND RADIATING SYSTEMS
6.1 Radiation of linear momentum and energy

We saw in chapter 5 on page 81 that at large distances r from the source, the leading order contributions to the E and B elds are purely transverse and fall off as 1=r. Hence, at large distances r, the dominating component of the Poynting vector, Sfar , falls off as 1=r 2 and is purely radial there. Consequently, when Sfar is integrated over a large spherical shell (centred on the source) for which the diO O O rected area element is d2x n D r 2 d r D r 2 sin d d' r [cf. formula (F.73) on page 213], the total integrated power, as given by the surface integral in equation (4.37) on page 64, tends to a constant at innity, showing that energy U eld and electromagnetic linear momentum peld is carried all the way to innity and is irreversibly lost there. This is the physical foundation of the well-known fact that peld and U eld can be transmitted over extremely long distances. The force action (time rate of change of peld ) on charges in one region in space can therefore cause a force action on charges in a another region in space. Todays wireless communication technology, be it classical or quantal, is based almost exclusively on the utilisation of this translational degree of freedom of the charges (currents) and the elds. Let us consider the linear momentum and energy that is carried in the far elds Bfar , equation (5.44b), and Efar , equation (5.44a) on page 94. Signals with limited lifetime and hence nite frequency bandwidth have to be analysed differently from monochromatic signals.
R A
hSi D 1
0
6.1.1 Monochromatic signals
If the source is strictly monochromatic, we can obtain the temporal average of the radiated power P directly, simply by averaging over one period so that hE Bi D 1 2
0
FT
Re fE
i!t
B gD D 1 2
0
1 2
0
Re E! e
i!t
.B! e
i!t
Re E!
B! e
i!t
Re fE!
B! g (6.1)
From formula (F.72) on page 213 and formula (F.73) on page 213 we see that d2x D jx x0 j2 d D jx x0 j2 sin d d' O O We also note from gure 5.1 on page 90 that k and n are nearly parallel. Hence, we can approximate O O k d2x n jx x0 j 2 d2x jx x0 j2 O O O O k nd k n d (6.2)
6.1. Radiation of linear momentum and energy
j 97
Using the far-eld approximations (5.34a) and (5.34b) on page 91 for the p elds and the fact that 1=c D "0 0 , and also introducing the characteristic impedance of vacuum r def 0 R0 376:7 (6.3) "0 for the vacuum resistance R0 , we obtain Z far 1 1 d3x 0 .j! R0 S D 2 2 32 jx x0 j V0
k/e
ik .x0 x0 /
2 x jx
x0 x0 j
(6.4)
6.1.2 Finite bandwidth signals
D
0
A signal with nite pulse width in time (t ) domain has a certain spread in frequency (!) domain. To calculate the total radiated energy we need to integrate over the whole bandwidth. The total energy transmitted through a unit area is the time integral of the Poynting vector: Z 1 Z 1 1 dt S.t / D dt .E B/ 0 1 1 (6.6) Z 1 Z 1 Z 1 1 0 i.!C! 0 /t 0/ e D d! d! dt .E! B!
1 1 1
If we carry out the temporal integration rst and use the fact that Z 1 0 dt e i.!C! /t D 2 .! C ! 0 /
1
R A
This formula is valid far away from the sources and shows that the radiated power is given by an expression which is resistance (R0 ) times the square of the supplied current (the integrated current density j! ), as expected. We note that the emitted power is independent of distance and is therefore carried all the way to innity. The possibility to transmit electromagnetic power over large distances, even in vacuum, is the physical foundation for the extremely important wireless communications technology. Besides determining the strength of the radiated power, the integral in formula (6.5) also determines the angular distribution.
FT
(6.7)
Consequently, the amount of power per unit solid angle d that ows across an innitesimal surface element r 2 d D jx x0 j2 d of a large spherical shell with its origin at x0 and enclosing all sources, is 2 Z dP 1 d3x 0 .j! k/e ik .x0 x0 / (6.5) D R 2 0 d 32 V0
98 j
where the last step follows from physical requirement of real-valuedness of E! and B! . We insert the Fourier transforms of the eld components which dominate at large distances, i.e., the far elds (5.44b) and (5.44a). The result, after integration over the area S of a large sphere which encloses the source volume V 0 , is 2 Z r I Z 1 1 0 2 d3x 0 j! k eikjx x0 j k O O (6.9) d xn d! U D 0j 4 "0 S jx x 0 V0
R A D
Inserting the approximations (5.32) and (6.2) into equation (6.9) above, introducing the spectral energy density U! via the denition Z 1 def U d!U! (6.10)
0
and recalling the denition (6.3) on the previous page, we obtain Z 2 dU! 1 0 d! R0 d3x 0 .j! k/e ik .x x0 / d! 0 d 4 V
FT
equation (6.6) on the previous page can be written Z 1 Z 1 2 d! .E! B ! / dt S.t / D 0 1 1 Z 1 Z 0 2 D d! .E! B ! / C d! .E! B ! / 0 1 Z0 1 Z 1 2 D d! .E! B ! / d! .E! B ! / 0 0 Z0 1 Z 1 2 D d! .E! B ! / C d! .E ! B! / 0 0 0 Z 1 2 d! .E! B ! C E ! B! / D 0 0 Z 1 2 D d! .E! B! C E! B! /
(6.8)
(6.11)
which, at large distances, is a good approximation to the energy that is radiated per unit solid angle d in a frequency band d!. It is important to notice that Formula (6.11) includes only source coordinates. This means that the amount of energy that is being radiated is independent on the distance to the source (as long as it is large).
6.2 Radiation of angular momentum

Not only electromagnetic linear momentum (Poynting vector) can be radiated from a source and transmitted over very long distances, but the same is also true
6.3. Radiation from a localised source volume at rest
j 99
where, in complex notation, I.t 0 / and P I.t 0 / Z

V0
Z
V0
d3x 0 j.t 0 ; x0 /
P d3x 0 j.t 0 ; x0 /
6.3 Radiation from a localised source volume at rest
In the general case, and when we are interested in evaluating the radiation far from a source at rest and which is localised in a small volume, we can introduce an approximation which leads to a multipole expansion where individual terms can be evaluated analytically. Here we use Hertzs method , which focuses on the physics rather than the mathematics, to obtain this expansion.
6.3.1 Electric multipole moments

Let us assume that the charge distribution determining the potential in equation (3.50a) on page 44 has such a small extent that all the source points x0 can
R A
We see that at very large distances r, the angular momentum density held falls off as 1=r 2 , i.e., it has precisely the same behaviour in the far zone as the linear momentum density and can therefore also transfer information wirelessly over large distances. The only difference is that while the direction of the linear momentum (Poynting vector) becomes purely radial at innity, the angular momentum becomes perpendicular to the linear momentum, i.e. purely transverse, there.
FT
(6.13) (6.14)
for electromagnetic angular momentum J eld . Then torque action (the time rate of change of J eld ) in one region causes torque action on charges. The use of this rotational degree of freedom of the elds has only recently been put to practical use even if it has been known for more than a century. After lengthy calculations, based on the results obtained in chapter 5 on page 81, one nds that the complete cycle averaged far-zone expression for a frequency component ! of the electromagnetic angular momentum density generated by arbitrary charge and current sources can be approximated by P eld O n Re .cq C In /I 1 h .x0 / D 32 2 "0 c 3 c jx x0 j2 O n Re f.cq C In /I g (6.12) C jx x0 j3
100 j
be assumed to be located very near a point x0 . At a large distance jx x0 j, one can then, to a good approximation, approximate the potential by the Taylor expansion [cf. example 3.1 on page 37] " 1 q.t/ 1 .x x0 /i .t; x/ D C d .t; x0 / 2 i 4 "0 jx x0 j jx x0 j jx x0 j # 1 1 3 .x x0 /i .x x0 /j ij C : : : C Qij .t; x0 / 2 jx x0 j jx x0 j 2 jx x0 j3 (6.15) where Z q.t / D
V0
is the total charge or electric monopole moment, Z d.t; x0 / D d3x 0 .x0
FT
V0 0 x0 / .tret ; x0 /
0 d3x 0 .tret ; x0 /
(6.16a)
(6.16b)
with components di , i D 1; 2; 3 is the electric dipole moment vector, and Z

Q.t; x0 / D
d3x 0 .x0
R A
0 t D t.tret /
V0
x0 /.x0
0 x0 / .tret ; x0 /
(6.16c)
with components Qij ; i; j D 1; 2; 3 (Einsteins summation convention over i and j is implied), is the electric quadrupole moment tensor. The source volume is so small that internal retardation effects can be ne0 glected, i.e., that we can set tret t jx x0 j =c. Then
0 tret C Const
(6.17)
where
Const D
jx c
x0 j
(6.18)
0 Hence the transformation between t and tret is a trivial. In the subsequent analy0 sis in this Subsection we shall use t 0 to denote this approximate tret . For a normal medium, the major contributions to the electrostatic interactions come from the net charge and the lowest order electric multipole moments induced by the polarisation due to an applied electric eld. Particularly important is the dipole moment. Let P denote the electric dipole moment density (electric dipole moment per unit volume; unit: C m 2 ), also known as the electric polarisation, in some medium. In analogy with the second term in the expansion
j 101
equation (6.15) on the preceding page, the electric potential from this volume distribution P .t; x0 / of electric dipole moments d at the source point x0 can be written Z x x0 1 d3x 0 P .t 0 ; x0 / d .t; x/ D 4 "0 V 0 jx x0 j3 Z 1 1 3 0 0 0 D d x P .t ; x / r (6.19) 4 "0 V 0 jx x0 j Z 1 1 D d3x 0 P .t 0 ; x0 / r 0 0 4 "0 V jx x0 j Using expression (M.120a) on page 243 and applying the divergence theorem, we can rewrite this expression for the potential as follows: Z Z r 0 P .t 0 ; x0 / 1 P .t 0 ; x0 / d3x 0 r 0 d3x 0 d .t; x/ D 0j 4 "0 V 0 jx x jx x0 j V0 I (6.20) Z 1 P .t 0 ; x0 / r 0 P .t 0 ; x0 / O D d2x 0 n0 d3x 0 4 "0 S 0 jx x0 j jx x0 j V0
Comparing this expression with expression equation (3.50a) on page 44for the potential from a charge distribution .t; x/, we see that r P .t; x/ has the characteristics of a charge density and that, to the lowest order, the effective charge density becomes .t; x/ r P .t; x/, in which the second term is a polarisation term that we call pol .t; x/.
6.3.2 The Hertz potential
In section 6.3.1 on page 99 we introduced the electric polarisation P .t; x/ such that the polarisation charge density
pol
R A
D r P
true def
where the rst term, which describes the effects of the induced, non-cancelling dipole moment on the surface of the volume, can be neglected, unless there is a O discontinuity in n P at the surface. Doing so, we nd that the contribution from the electric dipole moments to the potential is given by Z r 0 P .t 0 ; x0 / 1 d3x 0 (6.21) d .t; x/ D 4 "0 V 0 jx x0 j
If we adopt the same idea for the true charge density due to free charges and introduce a vector eld .t; x/, analogous to P .t; x/, but such that r (6.23a)
FT
(6.22)
102 j
which means that the associated polarisation current now is the true current: @ D j true (6.23b) @t As a consequence, the equation of continuity for true charges and currents [cf. expression (1.25) on page 10] is satised: .t; x/ @ @ C r j true .t; x/ D r Cr D0 (6.24) @t @t @t The vector is called the polarisation vector because, formally, it treats also the true (free) charges as polarisation charges. Since in the microscopic MaxwellLorentz equation (2.1a) on page 21, the charge density must include all charges, we can write this equation r ED @
true
FT
true
"0
C "0
pol
r P
"0
(6.25)
i.e., in a form where all the charges are considered to be polarisation charges. We now introduce a further potential e with the following property r e D
e
(6.26a)
R A
2
1 @ DA (6.26b) c 2 @t where and A are the electromagnetic scalar and vector potentials, respectively. As we see, e acts as a super-potential in the sense that it is a potential from which we can obtain other potentials. It is called the Hertz vector or polarisation potential . Requiring that the scalar and vector potentials and A, respectively, satisfy their inhomogeneous wave equations, equations (3.30) on page 40, one nds, using (6.23) and (6.26), that the Hertz vector must satisfy the inhomogeneous wave equation e D 1 @2 e c 2 @t 2 r 2 e D "0 (6.27)
Assume that the source region is a limited volume around some central point x0 far away from the eld (observation) point x illustrated in gure 6.1 on the facing page. Under these assumptions, we can expand the Hertz vector, expres0 sion (6.32) on the next page, due to the presence of non-vanishing .tret ; x0 / in the vicinity of x0 , in a formal series. For this purpose we recall from potential theory that eikjx x j eikj.x x0 / .x x0 /j jx x0 j j.x x0 / .x0 x0 /j 1 X D ik .2n C 1/Pn .cos /jn .k x0
nD0
0 0
(6.28) x0 /h.1/ .k jx n x0 j/
j 103
x x x
x0 x0 x0 x0 V0 x0 x0
Figure 6.1: Geometry of a typical multipole radiation problem where the eld point x is located some distance away from the nite source volume V 0 centred around x0 . If k jx0 x0 j 1 k jx x0 j, then the radiation at x is well approximated by a few terms in the multipole expansion.
where (see gure 6.1)
is the angle between x0
x0 and x
x0
Pn .cos / is the Legendre polynomial of order n jn .k x0 x0 / is the spherical Bessel function of the rst kind of order n h.1/ .k jx n
x0 j/ is the spherical Hankel function of the rst kind of order n
According to the addition theorem for Legendre polynomials Pn .cos / D
R A
mD n
n X
m . 1/m Pn .cos /Pn m .cos 0 /eim.'
m where Pn is an associated Legendre polynomial of the rst kind , related to the m spherical harmonic Yn as s 2n C 1 .n m/ m m P .cos / eim' Yn .; '/ D 4 .n C m/ n
and, in spherical polar coordinates, x0
x 0 D . x 0 x0 D .jx
x0 ; 0 ; ' 0 /
x0 j ; ; '/
This equation is of the same type as equation (3.33) on page 41, and has therefore the retarded solution Z 0 1 .tret ; x0 / (6.31) e .t; x/ D d3x 0 4 "0 V 0 jx x0 j
0 ikjx x0 j ! .x /e jx x0 j
with Fourier components e .x/ D ! 1 4 "0 Z

V0
d3x 0
FT
'0/
eikjx x j is a Green function or propagator jx x0 j
(6.29)
(6.30a) (6.30b)
(6.32)
104 j
If we introduce the help vector C such that CDr e (6.33)
we see that we can calculate the magnetic and electric elds, respectively, as follows BD 1 @C c 2 @t C (6.34a) (6.34b)
EDr
e !
1 n ik X X m D .2n C 1/. 1/m h.1/ .k jx x0 j/ Pn .cos / eim' n 4 "0 nD0 mD n Z 0 d3x 0 ! .x0 / jn .k x0 x0 / Pn m .cos 0 / e im' V0
FT
x0 j/ . i/nC1 eikjx x0 j k jx x0 j x 0 / 2n n k x0 .2n C 1/ x0
n
Clearly, the last equation is valid only if r E D 0 i.e., if we are outside the source volume. Since we are mainly interested in the elds in the so called far zone, a long distance from the source region, this is no essential limitation. Inserting equation (6.28) on page 102, together with formula (6.29) on the previous page, into equation (6.32) on the preceding page, we can in a formally exact way expand the Fourier component of the Hertz vector as
(6.35)
R A
h.1/ .k jx n
We notice that there is no dependence on x x0 inside the integral; the integrand is only dependent on the relative source vector x0 x0 . We are interested in the case where the eld point is many wavelengths away from the well-localised sources, i.e., when the following inequalities (6.36) k x0 x0 1 k jx x0 j hold. Then we may to a good approximation replace h.1/ with the rst term in n its asymptotic expansion: (6.37)
and replace jn with the rst term in its power series expansion: jn .k x0 (6.38)
Inserting these expansions into equation (6.35) above, we obtain the multipole expansion of the Fourier component of the Hertz vector e !
1 X nD0
e .n/ !
(6.39a)
j 105
O k x3 x Efar Bfar
Figure 6.2: If a spherical polar coordinate system .r; ; ') is chosen such that the electric dipole moment d (and thus its Fourier transform d! ) is located at the origin and directed along the polar axis, the calculations are simplied.
d O r x2
' x1
where
This expression is approximately correct only if certain care is exercised; if many e .n/ terms are needed for an accurate result, the expansions of the spherical ! Hankel and Bessel functions used above may not be consistent and must be replaced by more accurate expressions. Furthermore, asymptotic expansions as the one used in equation (6.37) on page 104 are not unique. Taking the inverse Fourier transform of e will yield the Hertz vector in ! time domain, which inserted into equation (6.33) on the facing page will yield C. The resulting expression can then in turn be inserted into equations (6.34) on the preceding page in order to obtain the radiation elds. For a linear source distribution along the polar axis, D in expression (6.39b) above, and Pn .cos / gives the angular distribution of the radiation. In the general case, however, the angular distribution must be computed with the help of formula (6.29) on page 103. Let us now study the lowest order contributions to the expansion of the Hertz vector.
R A
V0
e .n/ D . i/n !
1 eikjx x0 j 2n n 4 "0 jx x0 j .2n/ Z d3x 0 ! .x0 / .k x0
FT
x0 /n Pn .cos / (6.39b)
106 j
6.3.3 Electric dipole radiation

Choosing n D 0 in expression (6.39b) on the preceding page, we obtain e .0/ ! eikjx x0 j D 4 "0 jx x0 j Z
V0
d3x 0
! .x
/D
1 eikjx x0 j d! 4 "0 jx x0 j
(6.40)
Since represents a dipole moment density for the true charges (in the same R vein as P does so for the polarised charges), d! D V 0 d3x 0 ! .x0 / is, by denition, the Fourier component of the electric dipole moment Z Z d.t; x0 / D d3x 0 .t 0 ; x0 / D d3x 0 .x0 x0 / .t 0 ; x0 / (6.41)
V0 V0
[cf. equation (6.16b) on page 100]. If a spherical coordinate system is chosen with its polar axis along d! as in gure 6.2 on the preceding page, the components of e .0/ are ! O e e .0/ r D r !
def
FT
1 eikjx x0 j d! cos 4 "0 jx x0 j 1 eikjx x0 j d! sin 4 "0 jx x0 j 1 ik jx x0 j eikjx x0 j O d! sin jx x0 j ik
(6.42a) (6.42b) (6.42c)
def O e e .0/ D ! def
O e e .0/ D 0 ' !
R A
.0/ O C! D C!;' D
Evaluating formula (6.33) on page 104 for the help vector C, with the spherically polar components (6.42) of e .0/ inserted, we obtain ! 1 4 "0 (6.43)
Applying this to equations (6.34) on page 104, we obtain directly the Fourier components of the elds B! D ! i 4
0
jx x0 j " 1 1 E! D 2 4 "0 jx x0 j2 1 C jx x0 j2
eikjx x0 j O d! sin (6.44a) jx x0 j ik x x0 cos jx x0 j jx x0 j # (6.44b) ik eikjx x0 j 2 O k sin d! jx x0 j jx x0 j
Keeping only those parts of the elds which dominate at large distances (the radiation elds) and recalling that the wave vector k D k.x x0 /= jx x0 j where k D !=c, we can now write down the Fourier components of the radiation
j 107
parts of the magnetic and electric elds from the dipole:

far B! D
! 4
eikjx x0 j O d! k sin D jx x0 j
! 4
eikjx x0 j .d! jx x0 j
k/ k/
(6.45a) k (6.45b)
Efar D !
1 eikjx x0 j O d! k 2 sin D 4 "0 jx x0 j
1 eikjx x0 j .d! 4 "0 jx x0 j
These elds constitute the electric dipole radiation, also known as E1 radiation.
Linear and angular momenta radiated from an electric dipole in vacuum The Fourier amplitudes of the elds generated by an electric dipole, d! , oscillating at the angular frequency !, are given by formul (6.44) on the preceding page. Inverse Fourier transforming to the time domain, and using a spherical coordinate system .r; ; '/, the physically observable elds are found to be 1 ! 0 k O d! sin cos.kr !t 0 / B.t; x/ D sin.kr !t 0 / (6.46a) 4 r r2 ! 1 k k2 1 0 0 0 O d! sin cos.kr !t / C 2 sin.kr !t / cos.kr !t / E.t; x/ D 4 "0 r r3 r 1 k 1 O C cos.kr !t 0 / C 2 sin.kr !t 0 / r (6.46b) d! cos 2 "0 r3 r where t 0 D t r=c is the retarded time. E X A M P L E 6.1
Applying formula (4.53) on page 67 for the electromagnetic linear momentum density to the elds from a pure electric dipole, equations (6.46) above, one obtains geld D "0 E C C ! 16 C BD ! 8
0 2 d 2 !
k sin2 .kr r4
0 2 d 2 !
!t 0 /
R
sin2 1 sin.kr r5 !t 0 / cos2 .kr O !t 0 / r k2 sin2.kr r3
0 2 d 2 !
D
C ! 32 2
k sin2 .kr r4 k3 C 2 cos2 .kr r
Using well-known trigonometric relations, this can be put in the form 1 ! 0 2 k geld D d! sin cos 5 sin2.kr !t 0 / 2 4 cos2.kr 16 2 r r O !t 0 / 1 sin2.kr r5 !t 0 / C !t 0 / 2 k cos2.kr r4 O !t 0 / r
sin2
k2 sin2.kr r3
A FT
sin cos cos2 .kr 1 sin.kr r5 !t 0 / !t 0 / cos.kr !t 0 / k2 sin.kr r3 !t 0 / 2 k2 sin.kr r3 !t 0 / cos.kr !t 0 / (6.47) !t 0 / cos.kr O !t 0 / !t 0 / cos.kr !t 0 / !t 0 / !t 0 / (6.48) k3 1 C cos2.kr r2
108 j
which shows that the linear momentum density, and hence the Poynting vector, is strictly radial only at innity. Applying formula (4.80) on page 72 for the electromagnetic angular momentum density with the momentum point chosen as x0 D 0, i.e., held D x geld (6.49)
and using equation (6.48) on the previous page, we nd that for a pure electric dipole ! 0 2 1 eld h D d sin cos 4 sin.kr !t 0 / cos.kr !t 0 / 8 2 ! r C k sin2 .kr r3 !t 0 / cos2 .kr !t 0 / k2 sin.kr r2 !t 0 / cos.kr O !t 0 / (6.50) or
A FT
held D ! 16
0 2 d 2 !
sin cos
1 sin2.kr r4
!t 0 /
k cos2.kr r3
!t 0 /
k2 sin2.kr r2
O !t 0 /
(6.51)
The total electromagnetic linear momentum is (cf. formula (4.59b) on page 68) Z d3x 0 geld .t 0 ; x0 / peld D
V0
(6.52)
and the total electromagnetic angular momentum is (cf. formula (4.86b) on page 73) Z J eld D d3x 0 held .t 0 ; x0 / (6.53)
V0
R D
and held,far .t; x/
In order to get a total net J eld , it is convenient to superimpose several individual dipoles of (possibly) different strengths and relative phases. Perhaps the most common conguration yielding a total net J eld is two orthogonal co-located dipoles with =2 phase shift between them. We note that in the far zone the linear and angular momentum densities tend to geld,far .t; x/ ! 16 ! D 32
3 0 2k d 2 ! r2 3 0 2k d! 2 2 r
sin2 cos2 .kr
!t 0 /O r O !t 0 / r (6.54)
sin2 1 C cos2.kr
! 0 2 k2 d sin cos sin.kr !t 0 / cos.kr 8 2 ! r2 ! 0 2 k2 O D d sin cos sin2.kr !t 0 / 16 2 ! r 2
O !t 0 /
(6.55)
j 109
respectively. I.e., to leading order, both the linear momentum density geld and the angular momentum density held fall off as 1=r 2 far away from the source region. This means that when they are integrated over a spherical surface / r 2 located at a large distance from the source [cf. the last term in the LHS of formula (4.37) on page 64], there is a net ux so that the integrated momenta do not fall off with distance and can therefore be transported all the way to innity. End of example 6.1
6.3.4 Magnetic dipole radiation

The next term in the expression (6.39b) on page 105 for the expansion of the Fourier transform of the Hertz vector is for n D 1: Z eikjx x0 j d3x 0 k x0 x0 ! .x0 / cos e .1/ D i ! 4 "0 jx x0 j V 0 (6.56) Z 1 eikjx x0 j 3 0 0 0 D ik d x .x x0 / .x x0 / ! .x / 4 "0 jx x0 j2 V 0 Here, the term .x .x x0 / .x0 x0 / .x0 x0 / x0 /
0 ! .x / ! .x 0
/ can be rewritten x0;i /.xi0 x0;i /
D .xi
and introducing i D xi 0 i D
R A
xi0 x0;i x0 / .x0 x0 /
! .x 0
x0;i
the j th component of the integrand in e .1/ can be broken up into ! 1 0 i !;j 0 C !;i j i 2 (6.59) 1 0 0 C i !;j i !;i j 2 i.e., as the sum of two parts, the rst being symmetric and the second antisymmetric in the indices i; j . We note that the antisymmetric part can be written as 1 1 0 0 0 i !;j 0 j .i !;i / !;i j D !;j .i i / i 2 2 1 (6.60) D ! . 0 / 0 . ! /j 2 1 D .x x0 / ! .x0 x0 / j 2 The utilisation of equations (6.23) on page 101, and the fact that we are considering a single Fourier component, f.x /gj D .t; x/ D
!e i!t
FT
0 ! .x /
(6.57)
(6.58a)
(6.58b)
(6.61)
110 j
allow us to express
in j! as
!
Di
j! !
(6.62)
Hence, we can write the antisymmetric part of the integral in formula (6.56) on the previous page as Z 1 .x x0 / d3x 0 ! .x0 / .x0 x0 / 2 V0 Z 1 (6.63) d3x 0 j! .x0 / .x0 x0 / D i .x x0 / 2! V0 1 D i .x x0 / m! !
The nal result is that the antisymmetric, magnetic dipole, part of e .1/ can ! be written e, antisym !
.1/
FT
D eikjx x0 j k .x 4 "0 ! jx x0 j2 x0 / m! eikjx x0 j .m! 4 jx x0 j
0
where we introduced the Fourier transform of the magnetic dipole moment Z 1 m! D d3x 0 .x0 x0 / j! .x0 / (6.64) 2 V0
(6.65)
R A
far B! .x/ D
In analogy with the electric dipole case, we insert this expression into equation (6.33) on page 104 to evaluate C, with which equations (6.34) on page 104 then gives the B and E elds. Discarding, as before, all terms belonging to the near elds and transition elds and keeping only the terms that dominate at large distances, we obtain k/ k k (6.66a) (6.66b)
Efar .x/ D !
k eikjx x0 j m! 4 "0 c jx x0 j
which are the elds of the magnetic dipole radiation (M1 radiation).
6.3.5 Electric quadrupole radiation

e, sym .1/
The symmetric part ! of the n D 1 contribution in the equation (6.39b) on page 105 for the expansion of the Hertz vector can be expressed in terms of the electric quadrupole tensor, which is dened in accordance with equation (6.16c) on page 100: Z 0 Q.t; x0 / D d3x 0 .x0 x0 /.x0 x0 / .tret ; x0 / (6.67)
V0
6.4. Radiation from an extended source volume at rest
j 111
Again we use this expression in equation (6.33) on page 104 to calculate the elds via equations (6.34) on page 104. Tedious, but fairly straightforward algebra (which we will not present here), yields the resulting elds. The components of the elds that dominate in the far eld zone (wave zone) are given by
far B! .x/ D
i 8
0!
eikjx x0 j .k Q! / jx x0 j
k k k
(6.68a) (6.68b)
Efar .x/ D !
eikjx x0 j .k Q! / 8 "0 jx x0 j i
This type of radiation is called electric quadrupole radiation or E2 radiation.
6.4 Radiation from an extended source volume at rest
Certain radiating systems have a symmetric geometry or are in any other way simple enough that a direct (semi-)analytic calculation of the radiated elds and energy is possible. This is for instance the case when the radiating current ows in a nite, conducting medium of simple geometry at rest such as in a stationary antenna.
Let us apply equation (6.5) on page 97 to calculate the radiated EM power from a one-dimensional, time-varying current. Such a current can be set up by feeding the EMF of a generator (e.g., a transmitter) onto a stationary, linear, straight, thin, conducting wire across a very short gap at its centre. Due to the applied EMF, the charges in this thin wire of nite length L are set into linear motion to produce a time-varying antenna current which is the source of the EM radiation. Linear antennas of this type are called dipole antennas. For simplicity, we assume that the conductor resistance and the energy loss due to the electromagnetic radiation are negligible. Choosing our coordinate system such that the x3 axis is along the antenna axis, the antenna current density can be represented, in complex notation, by 0 0 0 0 O j.t 0 ; x0 / D .x1 /.x2 /J.t 0 ; x3 / x3 (measured in Am 2) where J.t 0 ; x3 / is the current (measured in A) along the antenna wire. Since we can assume that the antenna wire is innitely thin, the antenna current must vanish at the endpoints L=2 and L=2. At the midpoint, where the antenna is fed across a very short gap in the conducting wire, the antenna current is, of course, equal to the supplied current.
R A
6.4.1 Radiation from a one-dimensional current distribution
FT
112 j
Figure 6.3: A linear antenna used for transmission. The current in the feeder and the antenna wire is set up by the EMF of the generator (the transmitter). At the ends of the wire, the current is reected back with a 180 phase shift to produce a antenna current in the form of a standing wave.
sink.L=2
0 x /
3
L 2
j.t 0 ; x0 /
L 2
0 For each Fourier frequency component !0 , the antenna current J.t 0 ; x3 / can 0 be written as I.x3 / exp. i!0 t 0 / so that the antenna current density can be repre0 0 sented as j.t ; x / D j0 .x0 / exp. i!0 t 0 / [cf. equations (5.6) on page 83] where 0 0 0 O j0 .x0 / D .x1 /.x2 /I.x3 / x3
FT
I. L=2/ D I.L=2/ D 0 ; I.0/ D I0
(6.69)
0 and where the spatially varying Fourier amplitude I.x3 / of the antenna current fulls the time-independent wave equation (Helmholtz equation)
R A
d2 I 0 C k 2 I.x3 / D 0 ; 02 dx3
(6.70)
This second-order ordinary differential equation with constant coefcients has the well-known solution 0 sink.L=2 x3 / 0 I.x3 / D I0 (6.71) sin.kL=2/
where I0 is the amplitude of the antenna current (measured in A), assumed to be constant and supplied by the generator/transmitter at the antenna feed point (in our case the midpoint of the antenna wire) and 1= sin.kL=2/ is a normalisation factor. The antenna current forms a standing wave as indicated in gure 6.3. When the antenna is short we can approximate the current distribution formula (6.71) above by the rst term in its Taylor expansion, i.e., by I0 .1 0 2jx3 j=L/. For a half-wave antenna (L D =2 , kL D ) formula (6.71) sim0 plies to I0 cos.kx3 /. Hence, in the most general case of a straight, innitely 0 thin antenna of nite, arbitrary length L directed along the x3 axis, the Fourier amplitude of the antenna current density is 0 x / 3 0 0 0 sink.L=2 O j0 .x / D I0 .x1 /.x2 / x3 (6.72) sin.kL=2/
j 113
O r x3 D z
L 2
O x O
Figure 6.4: We choose a spherical polar coordinate system .r D jxj ; ; ') and arrange it so that the linear electric dipole antenna axis (and thus the antenna current density j! ) is along the polar axis with the feed point at the origin.
j! .x0 /
O k x2
' x1
L 2
For a half-wave dipole antenna (L D =2), the antenna current density is simply
R A
0 0 0 j0 .x0 / D I0 .x1 /.x2 / cos.kx3 / 0 0 j0 .x0 / D I0 .x1 /.x2 /.1
while for a short antenna (L
) it can be approximated by
In the case of a travelling wave antenna, in which one end of the antenna is connected to ground via a resistance so that the current at this end does not vanish, the Fourier amplitude of the antenna current density is
0 0 0 j0 .x0 / D I0 .x1 /.x2 / exp.kx3 /
In order to evaluate formula (6.5) on page 97 with the explicit monochromatic current (6.72) inserted, we use a spherical polar coordinate system as in
FT
(6.73) 0 2 x3 =L/ (6.74) (6.75)
114 j
gure 6.4 on the preceding page to evaluate the source integral 2 Z 0 I D d3x 0 j0 k e ik .x x0 / 0 V Z L 2 0 2 sink.L=2 x3 / 0 cos ikx cos 0 0 D dx3 I0 k sin e ikx3 e 0 L sin.kL=2/ 2 2 2 Z L 2 2 2 2 k sin ikx0 cos 0 0 0 0 D I0 2 L e dx3 sin.kL=2 kx3 / cos.kx3 cos / 2 0 sin .k 2 / 2 cos.kL=2/ 2 cos.kL=2/ cos D 4I0 sin sin.kL=2/ (6.76) Inserting this expression and d D 2 sin d into formula (6.5) on page 97 and integrating over , we nd that the total radiated power from the antenna is 1 4 Z
0
P .L/ D
2 R0 I0
FT
d cos.kL=2/ cos cos.kL=2/ sin sin.kL=2/ 2 sin lim P .L/ D 2 L 12
2 R0 I0
(6.77)
R A
kL!0
One can show that
(6.78)
where is the vacuum wavelength. The quantity
Rrad .L/ D
P .L/ P .L/ D 1 2 D R0 2 6 Ieff I 2 0
2 L
197
2 L
(6.79)
is called the radiation resistance. For the technologically important case of a half-wave antenna, i.e., for L D =2 or kL D , formula (6.77) reduces to 1 4 Z d
0
2 P . =2/ D R0 I0
cos2
cos sin
2
(6.80)
The integral in (6.80) can always be evaluated numerically. But, it can in fact
j 115
where in the last step the Euler-Mascheroni constant D 0:5772 : : : and the cosine integral Ci.x/ were introduced. Inserting this into the expression equation (6.80) on the preceding page we obtain the value Rrad . =2/ 73 .
Radiation from a two-dimensional current distribution
As an example of a two-dimensional current distribution we consider a circular loop antenna of radius a and calculate the far-zone Efar and Bfar elds from such an antenna. We choose the Cartesian coordinate system x1 x2 x3 with its origin at the centre of the loop as in gure 6.5 on the following page.
In our case the generator produces a single frequency ! and we feed the antenna across a small gap where the loop crosses the positive x1 axis. The circumference of the loop is chosen to be exactly one wavelength D 2 c=!. This means that the antenna current oscillates in the form of a sinusoidal standing current wave around the circular loop with a Fourier amplitude j! D I0 cos ' 0 .
0
R
O D
According to equation (5.34b) on page 91 the Fourier component of the radiation part of the magnetic eld generated by an extended, monochromatic current source is Z i 0 eikjxj 0 far B! D d3x 0 e ik x j! k (6.82) 0 4 jxj V
For the spherical coordinate system of the eld point, we recall from subsection F.4.2 on page 212 that the following relations between the base vectors hold: O O O O r D sin cos ' x1 C sin sin ' x2 C cos x3 O O O D cos cos ' x1 C cos sin ' x2 O O sin ' x1 C cos ' x2 O sin x3
A FT
1:22 (6.81)
O a/.z 0 /0 (6.83)
also be evaluated analytically as follows: Z Z 1 cos2 2 cos cos2 2 u d D cos ! u D du D sin u2 0 1 1 1 C cos. u/ cos2 u D 2 2 Z 1 1 C cos. u/ 1 du D 2 1 .1 C u/.1 u/ Z 1 Z 1 1 C cos. u/ 1 1 1 C cos. u/ D du C du 4 1 .1 C u/ 4 1 .1 u/ Z h 1 1 1 C cos. u/ vi D du D 1 C u ! 2 1 .1 C u/ Z 1 2 1 cos v 1 D dv D C ln 2 Ci.2 / 2 0 v 2
E X A M P L E 6.2
116 j
Figure 6.5: For the loop antenna the spherical coordinate system .r; ; '/ describes the eld point x (the radiation eld) and the cylindrical coordinate system . 0 ; ' 0 ; z 0 / describes the source point x0 (the antenna current).
O r x3 D z D z 0 O x O O k x2 O z0
and
With the use of the above transformations and trigonometric identities, we obtain for the cylindrical coordinate system which describes the source:
R
O 0 D D
A FT
x0 x1 '0 O 0 O 0
O O O x1 D sin cos ' r C cos cos ' O sin O sin ' O O O O x2 D sin sin ' r C cos sin ' C cos ' O O x3 D cos r O O O 0 D cos ' 0 x1 C sin ' 0 x2 D sin cos.' 0 sin sin.' 0 '/O C cos cos.' 0 r '/O r O sin cos sin.' 0 O '/ C sin.' 0 O '/ C cos.' 0 O '/ O '/ O O sin ' 0 x1 C cos ' 0 x2 O O O z0 D x3 D cos r k x0 D ka sin cos.' 0 '/ O 0 k D kcos.' 0 O '/ C cos sin.' 0 O '/
j! .x0 /
'
(6.84)
(6.85) (6.86)
O O This choice of coordinate systems means that k D k r and x0 D a0 so that (6.87)
and (6.88) D
With these expressions inserted, recalling that in cylindrical coordinates d3x 0
j 117
0 d 0 d' 0 dz 0 ,
Z
V0
the source integral becomes Z 2 0 d3x 0 e ik x j! k D a d' 0 e

0
ika sin cos.' 0 '/
O I0 cos ' 0 0
k (6.89)
Z D I0 ak
0
ika sin cos.' 0 '/
cos.' 0
O '/ cos ' 0 d' 0 sin.' 0 O '/ cos ' 0 d' 0
Z C I0 ak cos
0
ika sin cos.' 0 '/
Utilising the periodicity of the integrands over the integration interval 0; 2 , introducing the auxiliary integration variable ' 00 D ' 0 ', and utilising standard trigonometric identities, the rst integral in the RHS of (6.89) can be rewritten Z 2 00 e ika sin cos ' cos ' 00 cos.' 00 C '/ d' 00
0
D cos ' D cos ' D
e
0
Z
0
e Z
0 2
ika sin cos ' 00
1 cos ' 2 C
e Z
0 2
ika sin cos ' 00
1 cos ' 2
Analogously, the second integral in the RHS of (6.89) can be rewritten Z 2 00 e ika sin cos ' sin ' 00 cos.' 00 C '/ d' 00
0
1 sin ' 2
1 sin ' 2 / D . 1/n Jn . / Z i n /D e i

0
Jn . Jn .
R
cos '
As is well-known from the theory of Bessel functions, i n 2
which means that
D
Z e
0
ika sin cos ' 00
ika sin cos ' 00
Putting everything together, we nd that Z 0 O O d3x 0 e ik x j! k D I C I'

V0
D I0 ak cos ' J0 .ka sin /
O C I0 ak cos sin ' J0 .ka sin / C J2 .ka sin /
A FT
cos2 ' 00 d' 00 C a vanishing integral 1 1 C cos 2' 00 2 2 d' 00 d' 00 (6.90)
ika sin cos ' 00
ika sin cos ' 00
cos.2' 00 / d' 00
ika sin cos ' 00
d' 00
(6.91)
ika sin cos ' 00
cos 2' 00 d' 00
Z
0
cos n' d' D
i cos '
(6.92) cos n' d'
d' 00 D 2 J0 .ka sin / (6.93) cos 2' d' D

00 00
2 J2 .ka sin /
O J2 .ka sin /
(6.94)
118 j
so that, in spherical coordinates where jxj D r,

far B! .x/ D
ikr 0e
4 r
O O I C I'
(6.95)
From this expression for the radiated B eld, we can obtain the radiated E eld with the help of Maxwells equations. End of example 6.2
R A
.t; x/ D 1 4 "0 Z
0
6.5 Radiation from a localised charge in arbitrary motion

The derivation of the radiation elds for the case of the source moving relative to the observer is considerably more complicated than the stationary cases that we have studied so far. In order to handle this non-stationary situation, we start from the retarded potentials (3.50) on page 44 in chapter 3 on page 35
0 0 tret ; x0 .tret / 0 jx.t/ x0 .tret /j V0 0 0 j tret ; x.tret / d3x 0 0 jx.t/ x0 .tret /j V0
FT
Z d3x 0 4
To obtain the desired physical magnetic eld in the radiation (far) zone we must Fourier transform back to t space and take the real part: ) ( 0 i 0 e.ikr !t / O O I C I' Bfar .t; x/ D Re 4 r 0 O O D sin.kr !t 0 / I C I' 4 r (6.96) I0 ak 0 0 O D sin.kr !t / cos ' J0 .ka sin / J2 .ka sin / 4r O C cos sin ' J0 .ka sin / C J2 .ka sin /
(6.97a) (6.97b)
A.t; x/ D
and consider a source region with such a limited spatial extent that the charges and currents are well localised. Specically, we consider a charge q 0 , for instance an electron, which, classically, can be thought of as a localised, unstructured and rigid charge distribution with a small, nite radius. The part of this charge distribution dq 0 which we are considering is located in dV 0 D d3x 0 in the sphere in gure 6.6 on the next page. Since we assume that the electron (or any other similar electric charge) moves with a velocity v 0 whose direction is arbitrary and whose magnitude can even be comparable to the speed of light, we R 0 cannot say that the charge and current to be used in (6.97) is V 0 d3x 0 .tret ; x0 /
6.5. Radiation from a localised charge in arbitrary motion
j 119
x.t / x x0
v 0 .t 0 / dS0 x .t /
0 0
dr 0 dV 0
Figure 6.6: Signals that are observed at time t at the eld point x were generated at time t 0 at source points x0 on a sphere, centred on x and expanding, as time increases, with velocity c D c.x x0 /= jx x0 j outward from the centre. The source charge element moves with an arbitrary velocity v 0 and gives rise to a source leakage out of the volume dV 0 D d3x 0 .
dq 0 c
R 0 and V 0 d3x 0 v .tret ; x0 /, respectively, because in the nite time interval during which the observed signal is generated, part of the charge distribution will leak out of the volume element d3x 0 .
6.5.1 The Linard-Wiechert potentials
The charge distribution in gure 6.6 on page 119 which contributes to the eld at x.t/ is located at x0 .t 0 / on a sphere with radius r D jx x0 j D c.t t 0 /. The radius interval of this sphere from which radiation is received at the eld point x during the time interval .t 0 ; t 0 C dt 0 / is .r 0 ; r 0 C dr 0 / and the net amount of charge in this radial interval is
0 dq 0 D .tret ; x0 / dS 0 dr 0 0 .tret ; x0 /
R A
.x
where the last term represents the amount of source leakage due to the fact that the charge distribution moves with velocity v 0 .t 0 / D dx0 =dt 0 . Since dt 0 D dr 0 =c and dS 0 dr 0 D d3x 0 we can rewrite the expression for the net charge as .x x0 / v 0 3 0 0 0 .tret ; x0 / dq 0 D .tret ; x0 / d3x 0 dx c jx x0 j .x x0 / v 0 0 D .tret ; x0 / 1 d3x 0 c jx x0 j
FT
x0 / v 0 .t 0 / 0 0 dS dt jx x0 j (6.98) (6.99)
120 j
or
0 .tret ; x0 / d3x 0 D
dq 0 1
.x x0 / v 0 cjx x0 j
(6.100)
which leads to the expression

0 .tret ; x0 / 3 0 dx D jx x0 j jx
dq 0 x0 j
.x x0 / v 0 c
(6.101)
R A
where s D s.t 0 ; x/ D x D x x0 .t 0 /
2
For a sufciently small and well localised charge distribution we can, assuming that the integrands do not change sign in the integration volume, use the mean value theorem to evaluate these expressions to become Z 1 1 q0 1 .t; x/ D (6.103a) dq 0 D 4 "0 jx x0 j .x x0 / v 0 4 "0 s c Z 1 q0 v0 v0 A.t; x/ D dq 0 D 4 "0 c 2 jx x0 j .x x0 / v 0 4 "0 c 2 s c (6.103b) v0 D 2 c x0 .t 0 / v 0 .t 0 / c x x0 .t 0 / v 0 .t 0 / 0 0 x .t / 1 c jx x0 .t 0 /j 0 0 0 0 v .t / x x .t / x0 .t 0 / c jx x0 .t 0 /j x
FT
c
This is the expression to be used in the formul (6.97) on page 118 for the retarded potentials. The result is (recall that j D v ) Z dq 0 1 (6.102a) .t; x/ D 0 0 4 "0 jx x0 j .x x / v c Z v 0 dq 0 0 A.t; x/ D (6.102b) 0 0 4 jx x0 j .x x / v
(6.104a) (6.104b) (6.104c)
These results were derived independently by A L F R E D M A R I E L I NA R D (18691958) in 1898 and E M I L J O H A N N W I E C H E RT (18611928) in 1900. When v 0 k .x x0 / and v ! c, the potentials become singular. This was rst pointed out by A R N O L D J O H A N N E S WILHELM SOMMERFELD (18681951) in 1904.
D x
is the retarded relative distance. The potentials (6.103) are the Linard-Wiechert potentials.2 In section 7.3.2 on page 158 we shall derive them in a more elegant and general way by using a relativistically covariant formalism. It should be noted that in the complicated derivation presented above, the observer is in a coordinate system which has an absolute meaning and the velocity v 0 is that of the localised charge q 0 , whereas, as we shall see later in the covariant derivation, two reference frames of equal standing are moving relative to each other with v 0 .
j 121
q0 x0 .t 0 / 0
jx x0 j 0 v c
x0 .t / 0
v 0 .t 0 /
Figure 6.7: Signals that are observed at time t at the eld point x were generated at time t 0 at the source point x0 .t 0 /. After this time t 0 the particle, which moves with nonuniform velocity, has followed a so far (i.e., at t ) unknown trajectory. Extrapolating tangentially the trajectory from x0 .t 0 /, based on the velocity v 0 .t 0 /, denes the virtual simultaneous coordinate x0 .t /.
x x x0
x0 x.t /
The Linard-Wiechert potentials are applicable to all problems where a spatially localised charge in arbitrary motion emits electromagnetic radiation, and we shall now study such emission problems. The electric and magnetic elds are calculated from the potentials in the usual way: B.t; x/ D r E.t; x/ D A.t; x/ r .t; x/ (6.105a)
6.5.2 Radiation from an accelerated point charge

Consider a localised charge q 0 and assume that its trajectory is known experimentally as a function of retarded time x0 D x0 .t 0 / (6.106)
(in the interest of simplifying our notation, we drop the subscript ret on t 0 from now on). This means that we know the trajectory of the charge q 0 , i.e., x0 , for all times up to the time t 0 at which a signal was emitted in order to precisely arrive at the eld point x at time t. Because of the nite speed of propagation of the elds, the trajectory at times later than t 0 cannot be known at time t . The retarded velocity and acceleration at time t 0 are given by v 0 .t 0 / D dx0 dt 0 dv 0 d2 x 0 D 02 dt 0 dt (6.107a) (6.107b)
As for the charge coordinate x0 itself, we have in general no knowledge of the velocity and acceleration at times later than t 0 , and denitely not at the time of
R A
P a0 .t 0 / D v 0 .t 0 / D
FT
@A.t; x/ @t (6.105b)
122 j
observation t ! If we choose the eld point x as xed, the application of (6.107) to the relative vector x x0 yields d x dt 0 d2 x dt 0 2 x0 .t 0 / D x0 .t 0 / D v 0 .t 0 / P v 0 .t 0 / (6.108a) (6.108b)
The retarded time t 0 can, at least in principle, be calculated from the implicit relation t 0 D t 0 .t; x/ D t jx x0 .t 0 /j c (6.109)
R A
@ @t 0
x
and we shall see later how this relation can be taken into account in the calculations. According to formul (6.105) on the preceding page, the electric and magnetic elds are determined via differentiation of the retarded potentials at the observation time t and at the observation point x. In these formul the unprimed O r , i.e., the spatial derivative differentiation operator r D xi @=@xi means that we differentiate with respect to the coordinates x D .x1 ; x2 ; x3 / while keeping t xed, and the unprimed time derivative operator @=@t means that we differentiate with respect to t while keeping x xed. But the Linard-Wiechert potentials and A, equations (6.103) on page 120, are expressed in the charge velocity v 0 .t 0 / given by equation (6.107a) on page 121 and the retarded relative distance s.t 0 ; x/ given by equation (6.104) on page 120. This means that the expressions for the potentials and A contain terms that are expressed explicitly in t 0 , which in turn is expressed implicitly in t via equation (6.109) above. Despite this complication it is possible, as we shall see below, to determine the electric and magnetic elds and associated quantities at the time of observation t . To this end, we need to investigate carefully the action of differentiation on the potentials.
6.5.2.1 The differential operator method

We introduce the convention that a differential operator embraced by parentheses with an index x or t means that the operator in question is applied at constant x and t , respectively. With this convention, we nd that x p x0 .t 0 / Dp @ 0 j2 p @t 0 jx x .x x0 / v 0 .t 0 / D p jx x0 j2 p x x0 x
x
FT
x0 .t 0 /
(6.110)
j 123
Furthermore, by applying the operator .@=@t/x to equation (6.109) on the facing page we nd that 0 @t @ jx x0 .t 0 .t; x//j D1 @t x @t c x 0 @ @t jx x0 j D1 (6.111) @t 0 x c @t 0 x .x x0 / v 0 .t 0 / @t D1C c jx x0 j @t x This is an algebraic equation in .@t 0 =@t /x that we can solve to obtain 0 @t jx x0 j jx x0 j D D @t x s jx x0 j .x x0 / v 0 .t 0 /=c
(6.112)
where s D s.t 0 ; x/ is the retarded relative distance given by equation (6.104) on page 120. Making use of equation (6.112) above, we obtain the following useful operator identity 0 @t @ @ jx x0 j @ D D (6.113) @t x @t x @t 0 x s @t 0 x Likewise, by applying .r / t to equation (6.109) on the preceding page we obtain x x0 jx x0 .t 0 .t; x//j 0 .r t / t D r D r .x x0 / t c c jx x0 j t (6.114) x x0 .x x0 / v 0 .t 0 / 0 D C .r t / t c jx x0 j c jx x0 j This is an algebraic equation in .r t 0 / t with the solution .r t 0 / t D x x0 cs
R A
which gives the following operator relation when .r / t is acting on an arbitrary function of t 0 and x: x x0 @ @ 0 0 D C .r / t C .r / t 0 (6.116) .r / t D .r t / t @t 0 x cs @t 0 x
With the help of the rules (6.116) and (6.113) we are now able to replace t by t 0 in the operations that we need to perform. We nd, for instance, that 1 q0 r .r / t D r 4 "0 s t (6.117a) q0 x x0 v 0 .t 0 / x x0 @s D 4 "0 s 2 jx x0 j c cs @t 0 x
FT
(6.115)
124 j
and, from equation (6.103b) on page 120 and equation (6.113) on the previous page,) 0 0 0 @A @A @ 0 q v .t / D @t @t x @t 4 s x (6.117b) 0 0 q0 x x0 s v .t / x x0 v 0 .t 0 / @s P D 4 "0 c 2 s 3 @t 0 x Utilising these relations in the calculation of the E eld from the Linard-Wiechert potentials, equations (6.103) on page 120, we obtain E.t; x/ D @ A.t; x/ @t " q0 x x0 .t 0 / jx x0 .t 0 /j v 0 .t 0 /=c D 4 "0 s 2 .t 0 ; x/ jx x0 .t 0 /j x x0 .t 0 / jx x0 .t 0 /j v 0 .t 0 /=c @s.t 0 ; x/ cs.t 0 ; x/ @t 0 x # 0 0 0 0 P jx x .t /j v .t / 2 c r .t; x/
FT
x x0 .t 0 / jx x0 .t 0 /j v 0 .t 0 / c
Acceleration (radiation) eld
(6.118)
R A
q0 x x0 .t 0 / C 4 "0 s 3 .t 0 ; x/ c2 (
Starting from expression (6.104a) on page 120 for the retarded relative distance s.t 0 ; x/, we see that we can evaluate .@s=@t 0 /x in the following way 0 0 0 @ @s x x0 .x x / v .t / D @t 0 x @t 0 c x @ D x x0 .t 0 / @t 0 x 0 0 @v .t / 1 @x x0 .t 0 / 0 0 0 0 v .t / C x x .t / 0 c @t @t 0 x x 0 0 0 0 02 0 0 0 0 0 .x x .t // v .t / v .t / .x x .t // a .t / D C c c jx x0 .t 0 /j (6.119) where equation (6.110) on page 122 and equations (6.107) on page 121, respectively, were used. Hence, the electric eld generated by an arbitrarily moving localised charge at x0 .t 0 / is given by the expression q0 v 02 .t 0 / jx x0 .t 0 /j v 0 .t 0 / E.t; x/ D x x0 .t 0 / 1 4 "0 s 3 .t 0 ; x/ c c2
Velocity eld (tends to the Coulomb eld when v ! 0)
) a0 .t 0 / (6.120)
j 125
The rst part of the eld, the velocity eld , tends to the ordinary Coulomb eld when v 0 ! 0 and does not contribute to the radiation. The second part of the eld, the acceleration eld , is radiated into the far zone and is therefore also called the radiation eld . From gure 6.7 on page 121 we see that the position the charged particle would have had if at t 0 all external forces would have been switched off so that the trajectory from then on would have been a straight line in the direction of the tangent at x0 .t 0 / is x0 .t /, the virtual simultaneous coordinate. During the arbitrary motion, we interpret x x0 .t/ as the coordinate of the eld point x relative to the virtual simultaneous coordinate x0 .t/. Since the time it takes for a signal to propagate (in the assumed free space) from x0 .t 0 / to x is jx x0 j =c, this relative vector is given by x x0 .t / D x x0 .t 0 / jx
The magnetic eld can be computed in a similar manner: B.t; x/ D r D A.t; x/ r

0
R A
A
t
This allows us to rewrite equation (6.120) on the preceding page in the following way " v 02 .t 0 / q0 x x0 .t / 1 E.t; x/ D 4 "0 s 3 c2 # (6.122) x x0 .t/ a0 .t 0 / 0 0 C x x .t / c2
D r
q x 4 "0 c 2 s 2 jx
x x0 j
v0
x x c jx x0 j
D
so that
where we made use of equation (6.103) on page 120 and formula (6.113) on page 123. But, according to (6.117a), x x0 c jx x0 j r
t
q0 x 2 s 2 jx 4 "0 c
@A x x0 B.t; x/ D r t 0j c jx x @t x x x0 .t 0 / D E.t; x/ c jx x0 .t 0 /j
FT
x0 .t 0 /j v 0 .t 0 / c (6.121) A x x0
t0 0
cs @A @t x
@A @t 0
(6.123)
x0 x0 j
v0
(6.124)
(6.125)
126 j
The electric far eld is obtained from the acceleration eld in formula (6.120) on page 124 as q0 jx x0 j v 0 far 0 0 a0 E .t; x/ D .x x / .x x / 4 "0 c 2 s 3 c (6.126) q0 0 0 0 0 x x .t / fx x0 .t/ a .t /g D 4 "0 c 2 s 3 where in the last step we again used formula (6.121) on the previous page. Combining this formula and formula (6.125) on the preceding page, the radiation part of the magnetic eld can be written Bfar .t; x/ D x x0 .t 0 / c jx x0 .t 0 /j Efar .t; x/ (6.127)
6.5.2.2 The direct method
An alternative to the differential operator transformation technique just described is to try to express all quantities in the potentials directly in t and x. An example of such a quantity is the retarded relative distance s.t 0 ; x/. According to equation (6.104) on page 120, the square of this retarded relative distance can be written s 2 .t 0 ; x/ D x C 2 x0 .t 0 / x x x0 .t 0 / x0 .t 0 / v 0 .t 0 / 2 c 2 x x0 .t 0 / v 0 .t 0 / c
R A
If we use the following handy identity .x x0 / v 0 c 2 C .x D jx
FT
x0 / c v0 2 D jx x0 j2 v 02 c2 .x x0 / c v0 2 x0 .t 0 / v 0 D x x0 .t/ v0
(6.128)
jx
x0 j2 v 02 jx x0 j2 v 02 2 0 cos2 0 C sin c2 c2 x0 j2 v 02 jx x0 j2 v 02 .cos2 0 C sin2 0 / D c2 c2
(6.129)
we nd that .x
x0 / v 0 c
(6.130)
Furthermore, from equation (6.121) on the preceding page, we obtain the identity x (6.131)
j 127
which, when inserted into equation (6.130) on the preceding page, yields the relation .x x0 / v 0 2 .x x0 / v 0 2 jx x0 j2 v 02 D (6.132) c c2 c Inserting the above into expression (6.128) on the facing page for s 2 , this expression becomes 0 0 0 2 02 .x x0 / v 0 2 2 x x0 2 2 x x0 .x x / v C jx x j v s D c c2 c 0 0 2 0 2 .x x0 / v jx x j v D .x x0 / c c 0 2 .x x0 / v D .x x0 /2 c x x0 .t / v 0 .t 0 / 2 jx x0 .t /j2 c (6.133)
where in the penultimate step we used equation (6.121) on page 125. What we have just demonstrated is that if the particle velocity at time t can be calculated or projected from its value at the retarded time t 0 , the retarded distance s in the Linard-Wiechert potentials (6.103) can be expressed in terms of the virtual simultaneous coordinate x0 .t /, viz., the point at which the particle will have arrived at time t , i.e., when we obtain the rst knowledge of its existence at the source point x0 at the retarded time t 0 , and in the eld coordinate x D x.t/, where we make our observations. We have, in other words, shown that all quantities in the denition of s, and hence s itself, can, when the motion of the charge is somehow known, be expressed in terms of the time t alone. I.e., in this special case we are able to express the retarded relative distance as s D s.t; x/ and we do not have to involve the retarded time t 0 or any transformed differential operators in our calculations. Taking the square root of both sides of equation (6.133) above, we obtain the following alternative nal expressions for the retarded relative distance s in terms of the charges virtual simultaneous coordinate x0 .t/ and velocity v 0 .t 0 /: s x x0 .t/ v 0 .t 0 / 2 s.t 0 ; x/ D jx x0 .t /j2 (6.134a) c s v 02 .t 0 / 2 D jx x0 .t /j 1 sin 0 .t/ (6.134b) c2 s v 02 .t 0 / x x0 .t/ v 0 .t 0 / 2 2 D jx x0 .t /j 1 (6.134c) C c2 c
R A
FT
128 j
If we know what velocity the particle will have at time t , expression (6.134) on the previous page for s will not be dependent on t 0 . Using equation (6.134c) on the preceding page and standard vector analytic formulae, we obtain " # .x x0 / v 0 2 v 02 2 2 C r s D r jx x0 j 1 c2 c v 0v 0 v 02 (6.135) C 2 .x x0 / D 2 .x x0 / 1 c2 c 0 v0 v D 2 .x x0 / C .x x0 / c c
E X A M P L E 6.3
3
The elds from a uniformly moving charge
This problem was rst solved by O L I V E R H E AV I S I D E in 1888.
In the special case of uniform motion,3 the localised charge moves in a eld-free, isolated space and we know that it will not be affected by any external forces. It will therefore move uniformly in a straight line with the constant velocity v 0 . This gives us the possibility to extrapolate its position at the observation time, x0 .t/, from its position at the retarded time, P x0 .t 0 /. Since the particle is not accelerated, v 0 0, the virtual simultaneous coordinate x0 will be identical to the actual simultaneous coordinate of the particle at time t , i.e., x0 .t/ D x0 .t/. As depicted in gure 6.7 on page 121, the angle between x x0 and v 0 is 0 while then angle between x x0 and v 0 is 0 . In the case of uniform velocity v 0 , i.e., a velocity that does not change with time, any physical observable f .t; x/ has the same value at time t and position x as it has at time t C dt and position x C v 0 dt . Hence, f .t; x/ D f .t C dt; x C v 0 dt/ (6.136)
R D
Taylor expanding f .t C dt; x C v 0 dt ), keeping only linear terms in the innitesimally small dt , we obtain f .t C dt; x C v 0 dt/ D f .t; x/ C @f dt C v 0 rf dt C O .dt/2 @t (6.137)
From this we conclude that for uniform motion @f D @t v 0 rf (6.138)
Since f is an arbitrary physical observable, the time and space derivatives must be related in the following way when they operate on any physical observable dependent on x.t/ [cf. equation (1.36) on page 14]: @ D @t v0 r (6.139)
Hence, the E and B elds can be obtained from formul (6.105) on page 121, with the
A FT
We shall use this result in example 6.3 above for a uniform, unaccelerated motion of the charge.
j 129
potentials given by equations (6.103) on page 120 as follows: ED 1 @v 0 v0 @ @A D r D r @t c 2 @t c 2 @t v0 v0 v 0v 0 D r C r r D 1 c c c2 0 0 vv D r 13 c2 0 v0 v0 v Dr D r BDr ADr c2 c2 c2 0 0 0 0 v0 v v v0 vv D 2 r r r D 2 13 c c c c c2 v0 D 2 E c r
(6.140a)
(6.140b)
O O Here 13 D xi xi is the unit dyad and we used the fact that v 0 v 0 0. What remains is just to express r in quantities evaluated at t and x. From equation (6.103a) on page 120 and equation (6.135) on the preceding page we nd that 1 q0 q0 r D r D r s2 4 "0 s 8 "0 s 3 (6.141) 0 v q0 v0 .x x0 / D .x x0 / C c c 4 "0 s 3
When this expression for r is inserted into equation (6.140a) above, the following result 0 0 0 0 vv q0 vv E.t; x/ D 13 r D 13 r s2 c2 8 "0 s 3 c2 ( 0 q0 v v0 D .x x0 / .x x0 / C c c 4 "0 s 3 0 0 0 ) v0 v v 0v 0 v v .x x0 / .x x0 / c c c c c2 " (6.142) 0 q0 v0 v v 02 D .x x0 / C .x x0 / .x x0 / 2 c c 4 "0 s 3 c 0 # v0 v .x x0 / c c ! q0 v 02 D .x x0 / 1 4 "0 s 3 c2
obtains. Of course, the same result also follows from equation (6.122) on page 125 with P v 0 0 inserted. From equation (6.142) above we conclude that E is directed along the vector from the simultaneous coordinate x0 .t/ to the eld (observation) coordinate x.t/. In a similar way, the magnetic eld can be calculated and one nds that ! 0 v 02 1 0q B.t; x/ D 1 v 0 .x x0 / D 2 v 0 E (6.143) 4 s3 c2 c
A FT
130 j
From these explicit formulae for the E and B elds and formula (6.134b) on page 127 for s, we can discern the following cases: 1. v 0 ! 0 ) E goes over into the Coulomb eld ECoulomb 2. v 0 ! 0 ) B goes over into the Biot-Savart eld 3. v 0 ! c ) E becomes dependent on 0 4. v 0 ! c; sin 0 5. v 0 ! c; sin 0 0 ) E ! .1 1 ) E ! .1 v 02 =c 2 /ECoulomb v 02 =c 2 /
1=2 ECoulomb
End of example 6.3
6.5.2.3 Small velocities
s D x
x0
FT
.x x0 / v 0 c x x0 ; v0 c x0 / jx x0 j v 0 c x x0 ; v0 c x0 j 3 .x x0 / .x x0 / P v 0 ; v0 q0 c3 jx x0 j2 P v 0 .x x0 /; v0 c d D q 0 x0 .t 0 / R P q0v 0 D d ! x x Dx
0
If the charge moves at such low speeds that v 0 =c page 120 simplies to
1, formula (6.104) on
(6.144)
and formula (6.121) on page 125 x x0 D .x
(6.145)
R A
Efar .t; x/ D q0 4 "0 c 2 jx Bfar .t; x/ D 4 "0
so that the radiation eld equation (6.126) on page 126 can be approximated by c (6.146)
from which we obtain, with the use of formula (6.125) on page 125, the magnetic eld (6.147)
It is interesting to note the close correspondence that exists between the nonrelativistic elds (6.146) and (6.147) and the electric dipole eld equations (6.45) on page 107 if we introduce the electric dipole moment for a localised charge [cf. formula (6.41) on page 106] (6.148)
and at the same time make the transitions ! 2 d! (6.149a) (6.149b)
x0
j 131
v D 0:5c v D 0:25c v
vD0
Figure 6.8: Polar diagram of the energy loss angular distribution factor sin2 =.1 v cos =c/5 during bremsstrahlung for particle speeds v 0 D 0, v 0 D 0:25c, and v 0 D 0:5c.
The energy ux in the far zone is described by the Poynting vector as a function of Efar and Bfar . We use the close correspondence with the dipole case to nd that it becomes Sfar D
0q 02
P .v 0 /2 x0 j 2
16
2c
jx
sin2
P where is the angle between v 0 and x x0 . The total radiated power (integrated over a closed spherical surface) becomes
D
Efar .t; x/ D
6.5.3 Bremsstrahlung
An important special case of radiation is when the velocity v 0 and the acceleraP P tion v 0 are collinear (parallel or anti-parallel) so that v 0 v 0 D 0. This condition (for an arbitrary magnitude of v 0 ) inserted into expression (6.126) on page 126 for the radiation eld, yields q0 .x 4 "0 c 2 s 3 x0 / .x x0 / P v 0 ; P v0 k v0 (6.152)
R A
P q 0 2 v 02 P .v 0 /2 D (6.151) 6 c 6 "0 c 3 which is the Larmor formula for radiated power from an accelerated charge. Note that here we are treating a charge with v 0 c but otherwise totally unspecied motion while we compare with formul derived for a stationary oscillating dipole. The electric and magnetic elds, equation (6.146) on the facing page and equation (6.147) on the preceding page, respectively, and the expressions for the Poynting ux and power derived from them, are here instantaneous values, dependent on the instantaneous position of the charge at x0 .t 0 /. The angular distribution is that which is frozen to the point from which the energy is radiated. P D
0q 02
FT
x jx x0 x0 j (6.150)
132 j
from which we obtain, with the use of formula (6.125) on page 125, the magnetic eld Bfar .t; x/ D q 0 jx x0 j 0 P v 4 "0 c 3 s 3 .x x0 /; P v0 k v0 (6.153)
The difference between this case and the previous case of v 0 c is that the approximate expression (6.144) on page 130 for s is no longer valid; instead we must use the correct expression (6.104) on page 120. The angular distribution of the energy ux (Poynting vector) far away from the source therefore becomes Sfar D
0q 0 2 02
v P
sin2
v0 c
16
2c
jx
x0 j2 1
cos
x jx
x0 x0 j
(6.154)
dU rad . / d dt
D S .x
FT
x 0 / x x0 d D v P 2c 16
0q 0 2 02
It is interesting to note that the magnitudes of the electric and magnetic elds are P the same whether v 0 and v 0 are parallel or anti-parallel. We must be careful when we compute the energy (S integrated over time). The Poynting vector is related to the time t when it is measured and to a xed surface in space. The radiated power into a solid angle element d , measured relative to the particles retarded position, is given by the formula sin2
v0 c 6
cos
R A
dU rad d dt 0 D
(6.155)
On the other hand, the radiation loss due to radiation from the charge at retarded time t 0 : dU rad @t dU rad d D d (6.156) dt 0 dt @t 0 x Using formula (6.112) on page 123, we obtain dU rad s d dt jx x0 j D S .x x0 /s d (6.157)
Inserting equation (6.154) above for S into (6.157), we obtain the explicit expression for the energy loss due to radiation evaluated at the retarded time dU rad ./ d dt 0 D v P 2c 16
0q 0 2 02
sin2 1
v0 c
cos
(6.158)
The angular factors of this expression, for three different particle speeds, are plotted in gure 6.8 on the preceding page. Comparing expression (6.155) above with expression (6.158) above, we see that they differ by a factor 1 v 0 cos =c that comes from the extra factor s= jx x0 j introduced in (6.157). Let us explain this in geometrical terms.
j 133
dS dr x d q0 0 x0 v dt x0
2
Figure 6.9: Location of radiation between two spheres as the charge moves with velocity v 0 from x0 to 1 x0 during the time interval .t 0 ; t 0 C 2 0 /. The observation point (eld dt point) is at the xed location x.
x0 C c dt 0 2
R A
d D x d2x 2 x0 2 d3x D d2x dr D x 2 x0 d dr 2 x0 C c dt 0 2 x x0 2 D c x x x0 2 x0 2 ! v0 dt 0 D x cs dt 0 x0 2
During the interval .t 0 ; t 0 C dt 0 / and within the solid angle element d the particle radiates an energy dU rad . /=dt 0 dt 0 d . As shown in gure 6.9 this energy is at time t located between two spheres, one outer with its origin at x0 .t 0 / 1 and radius c.t t 0 /, and one inner with its origin at x0 .t 0 Cdt 0 / D x0 .t 0 /Cv 0 dt 0 2 1 and radius ct .t 0 C dt 0 / D c.t t 0 dt 0 /. From Figure 6.9 we see that the volume element subtending the solid angle element (6.159)
is
Here, dr denotes the differential distance between the two spheres and can be evaluated in the following way dr D x x x0 2 v 0 dt 0 x x0 2 0 v cos
where formula (6.104) on page 120 was used in the last step. Hence, the volume element under consideration is s d2x cdt 0 d3x D d2x dr D (6.162) x x 0 2
FT
(6.160) (6.161)
134 j
We see that the energy that is radiated per unit solid angle during the time interval .t 0 ; t 0 C dt 0 / is located in a volume element whose size is dependent. This explains the difference between expression (6.155) on page 132 and expression (6.158) on page 132. Q Let the radiated energy, integrated over , be denoted U rad . After tedious, but relatively straightforward integration of formula (6.158) on page 132, one obtains 3 0 2 02 Q P 2 q 0 2 v 02 P v 02 1 dU rad 0q v D 1 D (6.163) 3 02 dt 0 6 c 3 4 "0 c 3 c2 1 v2 c If we know v 0 .t 0 /, we can integrate this expression over t 0 and obtain the total energy radiated during the acceleration or deceleration of the particle. This way we obtain a classical picture of bremsstrahlung (braking radiation, free-free radiation). Often, an atomistic treatment is required for obtaining an acceptable result.
E X A M P L E 6.4 Bremsstrahlung for low speeds and short acceleration times
Calculate the bremsstrahlung when a charged particle, moving at a non-relativistic speed, is accelerated or decelerated during an innitely short time interval. We approximate the velocity change at time t 0 D t0 by a delta function: P v 0 .t 0 / D v 0 .t 0
1
A FT
t0 / v 0 .t0 / D Z
1
(6.164)
which means that
P dt 0 v 0
(6.165)
Also, we assume v=c
R D
1 so that, according to formula (6.104) on page 120, s x x0
(6.166)
and, according to formula (6.121) on page 125, x x0 x x0 (6.167)
From the general expression (6.125) on page 125 we conclude that E ? B and that it sufces to consider E Efar . According to the bremsstrahlung expression for Efar , equation (6.152) on page 131, ED q 0 sin 0 v 0 .t 0 4 "0 c 2 jx x0 j t0 / (6.168)
In this simple case B Bfar is given by BD E c (6.169)
Beacuse of the Dirac behaviour in time, Fourier transforming expression (6.168) above
j 135
for E is trivial, yielding E! D Z 1 q 0 sin 0 1 v 0 .t 0 t0 / i!t 0 dt 0 e jx x0 .t 0 /j 4 "0 c 2 2 1 q 0 sin 0 v 0 .t0 / e i!t0 D 2 " c 2 jx 8 x0 .t0 /j 0
(6.170)
We note that the magnitude of this Fourier component is independent of !. This is a consequence of the innitely short impulsive step .t 0 t0 / in the time domain which produces an innite spectrum in the frequency domain. The total radiation energy is given by the expression Z 1 I Z 1 Q dU rad Q O D dt 0 d2x 0 n0 E U rad D dt 0 dt 0 1 S0 1 I I Z 1 Z 1 1 D d2x 0 dt 0 EB D d2x 0 0 S0 0c S 0 1 I Z 1 D "0 c d2x 0 dt 0 E 2
S0 1
B
0 1 1
dt 0 E 2
(6.171)
According to Parsevals identity [cf. equation (6.8) on page 98] the following equality holds: Z 1 Z 1 dt 0 E 2 D 4 d! jE! j2 (6.172)
1 0
which means that the radiated energy in the frequency interval .!; ! C d!/ is I Q rad U! d! D 4 "0 c d2x 0 jE! j2 d!
S0
Q rad We see that the energy spectrum U! is independent of frequency !. This means that if we would integrate it over all frequencies ! 2 0; 1/, a divergent integral would result.
In reality, all spectra have nite widths, with an upper cutoff limit set by the quantum condition !max D 1 m.v 0 C v 0 /2 2 1 02 mv 2 (6.175)
which expresses that the highest possible frequency !max in the spectrum is that for which all kinetic energy difference has gone into one single eld quantum (photon) with energy !max . If we adopt the picture that the total energy is quantised in terms of N! photons radiated during the process, we nd that Q rad U! d! D dN! ! (6.176)
For our innite spectrum, equation (6.170) above, we obtain I sin2 0 q 0 2 .v 0 /2 Q rad d2x 0 d! U! d! D 3" c3 16 jx x0 j2 0 S0 Z Z q 0 2 .v 0 /2 2 D d' 0 d 0 sin 0 sin2 0 d! 16 3 "0 c 3 0 0 0 2 q02 v d! D 3 "0 c c 2
A FT
(6.173) (6.174)
136 j
or, for an electron where q 0 D dN! D e2 2 4 "0 c 3
jej, where e is the elementary charge, v 0 c 2 d! ! 1 2 137 3 v 0 c 2 d! ! 1=137. (6.177)
where we used the value of the ne structure constant D e 2 =.4 "0 c/
Even if the number of photons becomes innite when ! ! 0, these photons have negligible energies so that the total radiated energy is still nite. End of example 6.4
6.5.4 Cyclotron and synchrotron radiation (magnetic bremsstrahlung)

Formula (6.125) and formula (6.126) on page 126 for the magnetic eld and the radiation part of the electric eld are general, valid for any kind of motion of the localised charge. A very important special case is circular motion, i.e., the case P v 0 ? v 0. With the charged particle orbiting in the x1 x2 plane as in gure 6.10 on the facing page, an orbit radius a, and an angular frequency !0 , we obtain '.t 0 / D !0 t 0
0 0 0 0 0 0
FT
0 0 0 0
(6.178a) (6.178c)
R A
P v .t / D v D P
0 0 0 2 R x .t / D a!0 O 1 x 0 2 v D a! P 0 0 0
O x .t / D aO 1 cos '.t / C x2 sin '.t / x
(6.178b)
P O O v .t / D x .t / D a!0 x1 sin '.t / C x2 cos '.t / 0 0 v D a!0 v D O cos '.t / C x2 sin '.t /
0 0
(6.178d) (6.178e) (6.178f)
Because of the rotational symmetry we can, without loss of generality, rotate our coordinate system around the x3 axis so the relative vector x x0 from the source point to an arbitrary eld point always lies in the x2 x3 plane, i.e., x x0 D x O x0 .O 2 sin C x3 cos / x (6.179)
where is the angle between x x0 and the normal to the plane of the particle orbit (see Figure 6.10). From the above expressions we obtain .x .x x0 / v 0 D x x0 v 0 sin cos ' P x0 / v 0 D x x0 v 0 sin sin ' D x P (6.180a) P x0 v 0 cos (6.180b)
where in the last step we simply used the denition of a scalar product and the P fact that the angle between v 0 and x x0 is .
j 137
x2 x x v0 b 0 a0 '.t 0 / x1 q0 0 0 x .t / x0
Figure 6.10: Coordinate system for the radiation from a charged particle at x0 .t 0 / in circular motion with velocity v 0 .t 0 / along the tangent P and constant acceleration v 0 .t 0 / toward the origin. The x1 x2 axes are chosen so that the relative eld point vector x x0 makes an angle with the x3 axis, which is normal to the plane of the orbital motion. The radius of the orbit is a.
x3
The energy ux is given by the Poynting vector, which, with the help of formula (6.125) on page 125, can be written SD 1
0
.E
B/ D
1 c
0
Inserting this into equation (6.157) on page 132, we obtain dU rad .; '/ jx x0 j s D jEj2 dt 0 c 0
R A
c
where the retarded distance s is given by expression (6.104) on page 120. With the radiation part of the electric eld, expression (6.126) on page 126, inserted, and using (6.180a) and (6.180b) on the preceding page, one nds, after some algebra, that 2 02 v0 rad 0 2 02 1 1 v 2 sin2 sin2 ' sin cos ' c dU .; '/ P c 0q v D 5 0 0 2c dt 16 1 v sin cos '
The angles and ' vary in time during the rotation, so that refers to a moving coordinate system. But we can parametrise the solid angle d in the angle ' and the angle so that d D sin d d'. Integration of equation (6.183) above over this d gives, after some cumbersome algebra, the angular integrated expression Q dU rad D dt 0
0q
v P 6 c
0 2 02
In equation (6.183) above, two limits are particularly interesting:
FT
jEj2 x jx x0 x0 j (6.181) (6.182) (6.183) 1
v 02 c2
(6.184)
138 j
1. v 0 =c 2. v 0 =c
1 which corresponds to cyclotron radiation. 1 which corresponds to synchrotron radiation.
6.5.4.1 Cyclotron radiation

For a non-relativistic speed v 0 reduces to c, equation (6.183) on the preceding page
0 2 02 dU rad .; '/ P 0q v D .1 dt 0 16 2 c
sin2 sin2 '/
(6.185)
But, according to equation (6.180b) on page 136
where is dened in gure 6.10 on the preceding page. This means that we can write
0 2 02 dU rad . / P 0q v D .1 0 2c dt 16
FT
cos2 / D
0q 0 2 02
sin2 sin2 ' D cos2
(6.186)
v P
16
2c
sin2
(6.187)
Consequently, a xed observer near the orbit plane ( =2) will observe cyclotron radiation twice per revolution in the form of two equally broad pulses of radiation with alternating polarisation.
R A
6.5.4.2 Synchrotron radiation
When the particle is relativistic, v 0 c, the denominator in equation (6.183) on the previous page becomes very small if sin cos ' 1, which denes the forward direction of the particle motion ( =2; ' 0). The equation (6.183) on the preceding page becomes
0 2 02 P 1 dU rad . =2; 0/ 0q v D dt 0 16 2 c 1 v0 c 3
(6.188)
which means that an observer near the orbit plane sees a very strong pulse followed, half an orbit period later, by a much weaker pulse. The two cases represented by equation (6.187) above and equation (6.188) above are very important results since they can be used to determine the characteristics of the particle motion both in particle accelerators and in astrophysical objects where a direct measurement of particle velocities are impossible. In the orbit plane ( D =2), equation (6.183) on the previous page gives 2 02 v0 rad 0 2 02 1 1 v 2 sin2 ' cos ' c dU . =2; '/ P c 0q v D (6.189) 5 0 0 2c dt 16 1 v cos '
c
j 139
x x x0
x2
Figure 6.11: If the observation point x is in the plane of the particle orbit, i.e., if D =2, the lobe width is given by .
v0 b 0 q0 0 0 x .t / a0 '.t 0 / x1
x3
This vanishes for angles '0 such thati cos '0 D sin '0 D v0 c s 1
Hence, the angle '0 is a measure of the synchrotron radiation lobe width ; see gure 6.11. For ultra-relativistic particles, dened by s 1 v 02 Ds 1; 1 1; (6.191) c2 v 02 1 c2 one can approximate
R A
s '0 sin '0 D 1 v 02 1 D c2 1 t 0 D !0
Hence, synchrotron radiation from ultra-relativistic charges is characterized by a radiation lobe width which is approximately (6.193)
This angular interval is swept by the charge during the time interval (6.194)
FT
(6.190a) v 02 c2 (6.190b) (6.192)
140 j
during which the particle moves a length interval l 0 D v 0 t 0 D v 0 !0 (6.195)
in the direction toward the observer who therefore measures a compressed pulse width of length v 0 t 0 v0 l 0 D t 0 D 1 t 0 t D t 0 c c c 0 0 0 0 1 v 1C v c c v 1 v 1 D 1 1 D v0 c !0 c !0 !0 1C (6.196) c 1 1 1 v 02 D 3 1 2 c 2 !0 2 !0 1= 2
Typically, the spectral width of a pulse of length t is ! 1=t . In the ultrarelativistic synchrotron case one can therefore expect frequency components up to
FT
!max 1 D2 t
3
R A
!0
(6.197)
A spectral analysis of the radiation pulse will therefore exhibit a (broadened) line spectrum of Fourier components n!0 from n D 1 up to n 2 3 . When many charged particles, N say, contribute to the radiation, we can have three different situations depending on the relative phases of the radiation elds from the individual particles: 1. All N radiating particles are spatially much closer to each other than a typical wavelength. Then the relative phase differences of the individual electric and magnetic elds radiated are negligible and the total radiated elds from all individual particles will add up to become N times that from one particle. This means that the power radiated from the N particles will be N 2 higher than for a single charged particle. This is called coherent radiation. 2. The charged particles are perfectly evenly distributed in the orbit. In this case the phases of the radiation elds cause a complete cancellation of the elds themselves. No radiation escapes. 3. The charged particles are somewhat unevenly distributed in the orbit. This happens for an open ring current, carried initially by evenly distributed charged particles, which is subject to thermal uctuations. From statistical mechanics
j 141
vt q0 b jx B O E? x3 0 x0 j O v D v x1
Figure 6.12: The perpendicular electric eld of a charge q 0 moving O O with velocity v 0 D v 0 x is E? z.
we know that this happens forp open systems and that the particle densities all exhibit uctuations of order N . This means that out of the N particles, p N will exhibit deviation from perfect randomnessand thereby perfect radiation eld cancellationand give rise to net radiation elds which are p proportional to N . As a result, the radiated power will be proportional to N , and we speak about incoherent radiation. Examples of this can be found both in earthly laboratories and under cosmic conditions.
We recall that the general expression for the radiation E eld from a moving charge concentration is given by expression (6.126) on page 126. This expression in equation (6.182) on page 137 yields the general formula dU rad .; '/ D dt 0
0q
R A
jx x0 j .x 16 2 cs 5
02
6.5.4.3 Radiation in the general case
x/
.x
Integration over the solid angle
gives the totally radiated power as v P 6 c

0 2 02
Q dU rad D dt 0
0q
1 1
P where is the angle between v 0 and v 0 . 0 0 P If v is collinear with v , so that sin D 0, we get bremsstrahlung. For P v 0 ? v 0 , sin D 1, which corresponds to cyclotron radiation or synchrotron radiation.
FT
x/
0
jx
x0 j v 0 c
P v0 (6.198)
v 02 c2
v c2
sin2 3 02
(6.199)
142 j
6.5.4.4 Virtual photons

Let us consider a charge q 0 moving with constant, high velocity v 0 .t 0 / along the x1 axis. According to formula (6.142) on page 129 and gure 6.12 on the preceding page, the perpendicular component along the x3 axis of the electric eld from this moving charge is v 02 q0 O 1 .x x0 / x3 (6.200) E? D E3 D 4 "0 s 3 c2 Utilising expression (6.134) on page 127 and simple geometrical relations, we can rewrite this as E? D q0 4 "0 b
2
This represents a contracted Coulomb eld, approaching the eld of a plane wave. The passage of this eld pulse corresponds to a frequency distribution of the eld energy. Fourier transforming, we obtain Z 1 1 b! q0 b! E!;? D K1 (6.202) dt E? .t / ei!t D 2 4 2 "0 bv 0 v0 v0 1 Here, K1 is the Kelvin function (Bessel function of the second kind with imaginary argument) which behaves in such a way for small and large arguments that
R A
E!;? E!;? q0 4
2" 0 0 bv
FT
; b! v0 , b ! v0 b ! v0 1 1 v0 ,
bmin 1
.v 0 t 0 /2
C b2=
2 3=2
(6.201)
(6.203a) (6.203b)
0;
b!
showing that the pulse length is of the order b=.v 0 /. Due to the equipartitioning of the eld energy into the electric and magnetic elds, the total eld energy can be written Z Z bmax Z 1 2 2 Q U D "0 d3x E? D "0 db 2 b dt v 0 E? (6.204)
V
where the volume integration is over the plane perpendicular to v 0 . With the use of Parsevals identity for Fourier transforms, formula (6.8) on page 98, we can rewrite this as Z 1 Z bmax Z 1 2 Q Q U D d! U! D 4 "0 v 0 db 2 b d! E!;? 0 0 bmin (6.205) Z v0 =! Z 1 db q 02 d! 2 2 "0 v 0 1 b bmin
p2
p1
from which we conclude that Q U!
where an explicit value of bmin can be calculated in quantum theory only. As in the case of bremsstrahlung, it is intriguing to quantise the energy into photons [cf. equation (6.176) on page 135]. Then we nd that N! d! 2 ln c d! !
where D e2 =.4 "0 c/ 1=137 is the ne structure constant. Let us consider the interaction of two (classical) electrons, 1 and 2. The result of this interaction is that they change their linear momenta from p1 to p0 and p2 1 = p1 p0 to p0 , respectively. Heisenbergs uncertainty principle gives bmin 2 1 so that the number of photons exchanged in the process is of the order
R
N! d! 2 ln c N! d! 2
Since this change in momentum corresponds to a change in energy ! D E1 and E1 D m0 c 2 , we see that
a formula which gives a reasonable semi-classical account of a photon-induced electron-electron interaction process. In quantum theory, including only the lowest order contributions, this process is known as Mller scattering. A diagrammatic representation of (a semi-classical approximation of) this process is given in gure 6.13.
A FT
0 q 02 v ln 2" v0 2 bmin ! 0 (6.206) bmin ! (6.207) ! p1 d! p0 1 ! (6.208)
0 E1
E1 cp1 ln m0 c 2 E1
j 143
p0 2
Figure 6.13: Diagrammatic representation of the semi-classical electron-electron interaction (Mller scattering).
p0 1
! cp0 d! 1 0 E1 !
(6.209)
144 j
6.6 Bibliography
[27] H. A LFVN AND N. H ERLOFSON, Cosmic radiation and radio stars, Physical Review, 78 (1950), p. 616. [28] R. B ECKER, Electromagnetic Fields and Interactions, Dover Publications, Inc., New York, NY, 1982, ISBN 0-486-64290-9. [29] M. B ORN AND E. W OLF, Principles of Optics. Electromagnetic Theory of Propagation, Interference and Diffraction of Light, sixth ed., Pergamon Press, Oxford,. . . , 1980, ISBN 0-08-026481-6. [30] V. L. G INZBURG, Applications of Electrodynamics in Theoretical Physics and Astrophysics, Revised third ed., Gordon and Breach Science Publishers, New York, London, Paris, Montreux, Tokyo and Melbourne, 1989, ISBN 2-88124-719-9. [31] J. D. JACKSON, Classical Electrodynamics, third ed., John Wiley & Sons, Inc., New York, NY . . . , 1999, ISBN 0-471-30932-X. [32] H.-D. NATURE, The Physical Basis of The Direction of Time, fourth ed., SpringerVerlag, Cambridge . . . , 1920, ISBN 3-540-42081-9. [33] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0201-05702-6. [34] J. A. S TRATTON, Electromagnetic Theory, McGraw-Hill Book Company, Inc., New York, NY and London, 1953, ISBN 07-062150-0. [35] J. VANDERLINDE, Classical Electromagnetic Theory, John Wiley & Sons, Inc., New York, Chichester, Brisbane, Toronto, and Singapore, 1993, ISBN 0-471-57269-1.
R A
FT
RELATIVISTIC ELECTRODYNAMICS
We saw in chapter 3 on page 35 how the introduction of electrodynamic potentials led, in a most natural way, to the existence of a characteristic, nite speed of propagation of electromagnetic elds (and related quantities) in free space (vacp uum) that equals the speed of light c D 1= "0 0 and which can be considered a constant of Nature. To take this nite speed of propagation of information into account, and to ensure that our laws of physics be independent of any specic coordinate frame, requires a treatment of electrodynamics in a relativistically covariant (coordinate independent) form. This is the objective of this chapter.1 The technique we shall use to study relativity is the mathematical apparatus developed for non-Euclidean spaces of arbitrary dimensions, here specialised to four dimensions. It turns out that this theory of Riemannian spaces, derived for more or less purely mathematical reasons only, is ideal for a formal description of relativistic physics. For the simple case of the special theory of relativity, the mathematics is quite simple, whereas for the general theory of relativity it becomes more complicated.
The Special Theory of Relativity, by the physicist and philosopher D AV I D J O S E P H B O H M (1917 1992), opens with the following paragraph: The theory of relativity is not merely a scientic development of great importance in its own right. It is even more signicant as the rst stage of a radical change in our basic concepts, which began in physics, and which is spreading into other elds of science, and indeed, even into a great deal of thinking outside of science. For as is well known, the modern trend is away from the notion of sure absolute truth, (i.e., one which holds independently of all conditions, contexts, degrees, and types of approximation etc.) and toward the idea that a given concept has signicance only in relation to suitable broader forms of reference, within which that concept can be given its full meaning.
7.1 The special theory of relativity
An inertial system, or inertial reference frame, is a system of reference, or rigid coordinate system, in which the law of inertia (Galileos law, Newtons rst law) holds. In other words, an inertial system is a system in which free bodies move uniformly and do not experience any acceleration. The special theory of relativity describes how physical processes are interrelated when observed in different inertial systems in uniform, rectilinear motion relative to each other and is based on two postulates:
P O S T U L AT E 7.1 (Relativity principle; P OINCAR , 1905) All laws of physics (except the laws of gravitation) are independent of the uniform translational motion of the system on which they operate.
P O S T U L AT E 7.2 (E INSTEIN , 1905) The velocity of light in empty space is independent of the motion of the source that emits the light. 145
R A
FT
146 j
7 . RELATIVISTIC ELECTRODYNAMICS
Figure 7.1: Two inertial systems and 0 in relative motion with velocity v along the x D x 0 axis. At time t D t 0 D 0 the origin O 0 of 0 coincided with the origin O of . At time t , the inertial system 0 has been translated a distance vt along the x axis in . An event represented by P .t; x; y; z/ in is represented by P .t 0 ; x 0 ; y 0 ; z 0 / in 0 .
vt y
0
y0 v P .t; x; y; z/ P .t 0 ; x 0 ; y 0 ; z 0 /
O z
x z0
O0
x0
7.1.1 The Lorentz transformation
R A
D
Let us consider two three-dimensional inertial systems and 0 in free space. They are in rectilinear motion relative to each other in such a way that 0 moves with constant velocity v along the x axis of the system. The times and the spatial coordinates as measured in the two systems are t and .x; y; z/, and t 0 and .x 0 ; y 0 ; z 0 /, respectively. At time t D t 0 D 0 the origins O and O 0 and the x and x 0 axes of the two inertial systems coincide and at a later time t they have the relative location as depicted in gure 7.1, referred to as the standard conguration. For convenience, let us introduce the two quantities v c 1 1
2
FT
Dq ct 0 D .ct x D .x y Dy
0 0
A consequence of the rst postulate is that all geometrical objects (vectors, tensors) in an equation describing a physical process must transform in a covariant manner, i.e., in the same way.
(7.1) (7.2)
where v D jv j. In the following, we shall make frequent use of these shorthand notations. As shown by Einstein, the two postulates of special relativity require that the spatial coordinates and times as measured by an observer in and 0 , respectively, are connected by the following transformation: x/ vt/ (7.3a) (7.3b) (7.3c)
7.1. The special theory of relativity
j 147
z0 D z
(7.3d)
Taking the difference between the square of (7.3a) and the square of (7.3b) we nd that c 2 t 02 x 02 D c 2 t 2 2xct C x 2 2 x 2 C 2xvt v 2 t 2 v2 v2 1 2 2 2 c t 1 x 1 D c2 c2 v2 1 c2 2 2 Dc t x2
2
(7.4)
c2t 2
x2
y2
z 2 D c 2 t 02
7.1.2 Lorentz space
Let us introduce an ordered quadruple of real numbers, enumerated with the help of upper indices D 0; 1; 2; 3, where the zeroth component is ct (c is the speed of light and t is time), and the remaining components are the components of the ordinary R3 position vector x dened in equation (M.1) on page 223: x D .x 0 ; x 1 ; x 2 ; x 3 / D .ct; x; y; z/ .ct; x/ (7.6)
R A
which means that if a light wave is transmitted from the coinciding origins O and O 0 at time t D t 0 D 0 it will arrive at an observer at .x; y; z/ at time t in and an observer at .x 0 ; y 0 ; z 0 / at time t 0 in 0 in such a way that both observers conclude that the speed (spatial distance divided by time) of light in vacuum is c. Hence, the speed of light in and 0 is the same. A linear coordinate transformation which has this property is called a (homogeneous) Lorentz transformation.
In order that this quadruple x represent a physical observable, it must transform as (the component form of) a position four-vector (radius four-vector) in a real, linear, four-dimensional vector space.2 We require that this four-dimensional space be a Riemannian space, i.e., a metric space where a distance and a scalar product are dened. In this space we therefore dene a metric tensor, also known as the fundamental tensor, which we denote by g .
FT
x 02 y 02 z 02 (7.5)
2
From equations (7.3) on the preceding page we see that the y and z coordinates are unaffected by the translational motion of the inertial system 0 along the x axis of system . Using this fact, we nd that we can generalise the result in equation (7.4) above to
The British mathematician and philosopher A L F R E D N O RT H W H I T E H E A D (18611947) writes in his book The Concept of Nature: I regret that it has been necessary for me in this lecture to administer a large dose of fourdimensional geometry. I do not apologise, because I am really not responsible for the fact that Nature in its most fundamental aspect is four-dimensional. Things are what they are. . . .
148 j
7.1.2.1 Radius four-vector in contravariant and covariant form

The position four-vector x D .x 0 ; x 1 ; x 2 ; x 3 / D .ct; x/, as dened in equation (7.6) on the preceding page, is, by denition, the prototype of a contravariant vector (or, more accurately, a vector in contravariant component form). To every such vector there exists a dual vector. The vector dual to x is the covariant vector x , obtained as x Dg x (7.7)
where the upper index in x is summed over and is therefore a dummy index and may be replaced by another dummy index , say. This summation process is an example of index contraction and is often referred to as index lowering.
7.1.2.2 Scalar product and norm
The scalar product of x with itself in a Riemannian space is dened as g x x Dx x (7.8)
R A
7.1.2.3 Metric tensor
This scalar product acts as an invariant distance, or norm, in this space. In order to put the physical property of Lorentz transformation invariance, described by equation (7.5) on the previous page, into a convenient mathematical framework, we perceive this invariance as the manifestation of the conservation of the norm in a 4D Riemannian space.
In L4 one can choose the metric tensor g to take the simple form 8 1 if D D 0 < g D D 1 if D D i D j D 1; 2; 3 :0 if or, in matrix representation, 0 1 B0 B /DB @0 0 0 1 0 0 0 0 1 0 1 0 0C C C 0A 1
FT
(7.9)
(7.10)
i.e., a matrix with a main diagonal that has the sign sequence, or signature, fC; ; ; g, the index lowering operation in our at 4D space L4 becomes nearly trivial:
j 149
In matrix representation the lowering of the indices of x becomes 1 10 1 0 0 1 0 x0 x0 x0 1 0 0 0 Bx C B0 1 0 0 C Bx 1 C B x 1 C C CB C B B 1C B C CB C D B B CDB @x2 A @0 0 1 0 A @x 2 A @ x 2 A x3 x3 0 0 0 1 x3 I four-tensor notation, this can be written x D x D .ct; x/
(7.11)
(7.12)
x x D .ct; x/ .ct; x/ D c 2 t 2
which indeed is the desired Lorentz transformation invariance as required by equation (7.13) above. Without changing the physics, one can alternatively choose a signature f ; C; C; Cg. The latter has the advantage that the transition from 3D to 4D becomes smooth, while it will introduce some annoying minus signs in the theory. In current physics literature, the signature fC; ; ; g seems to be the most commonly used one. Note that our space, regardless of signature chosen, will have an indenite norm, i.e., a norm which can be positive denite, negative denite or even zero. This means that we deal with a nonEuclidean space and we call our four-dimensional space (or space-time) with this property Lorentz space and denote it L4 . A corresponding real, linear 4D space with a positive denite norm which is conserved during ordinary rotations is a Euclidean vector space. We denote such a space R4 . The L4 metric tensor equation (7.9) on the facing page has a number of interesting properties: rstly, we see that this tensor has a trace Tr D 2 whereas in R4 , as in any vector space with denite norm, the trace equals the space dimensionality. Secondly, we nd, after trivial algebra, that the following relations between the contravariant, covariant and mixed forms of the metric tensor hold:
R A

D D D D
D D
FT
x2 y2 z2 (7.13) (7.14a) (7.14b) (7.14c) (7.14d)
Hence, if the metric tensor is dened according to expression (7.9) on the facing page, the covariant position four-vector x is obtained from the contravariant position four-vector x simply by changing the sign of the last three components. These components are referred to as the space components; the zeroth component is referred to as the time component. As we see, for this particular choice of metric, the scalar product of x with itself becomes
150 j
Here we have introduced the 4D version of the Kronecker delta , a mixed four-tensor of rank 2 that fulls ( 1 if D D D (7.15) 0 if Clearly, the matrix representation of this tensor is 1 0 1 0 0 0 B0 1 0 0C C B . / D B C @0 0 1 0A 0 0 0 1
(7.16)
7.1.2.4 Invariant line element and proper time
The differential distance ds between the two points x and x C dx in L4 can be calculated from the Riemannian metric, given by the quadratic differential form ds 2 D dx dx D dx dx D .dx 0 /2 .dx 1 /2 .dx 2 /2 .dx 3 /2 (7.17)
R A
q
ds D c dt 1 r D c dt 1 D c dt q 1
where the metric tensor is as in equation (7.9) on page 148. As we see, this form is indenite as expected for a non-Euclidean space. The square root of this expression is the invariant line element 2 !2 1 4 dx 1 C c2 dt dx 2 dt !2 C dx 3 dt !2 3 5 v2 c2 (7.18)
1 .vx /2 C .vy /2 C .vz /2 D c dt c2 dt 2 D c D c d
where we introduced d D dt= (7.19)
Since d measures the time when no spatial changes are present, i.e., by a clock that is xed relative the given frame of reference, it is called the proper time. As equation (7.19) above shows, the proper time of a moving object is always less than the corresponding interval in the rest system. One may say that moving clocks go slower than those at rest.
FT
s 1
i.e., the 4
4 unit matrix.
j 151
Expressing the property of the Lorentz transformation described by equations (7.5) on page 147 in terms of the differential interval ds and comparing with equation (7.17) on the facing page, we nd that ds 2 D c 2 dt 2 dx 2 dy 2 dz 2 (7.20)
is invariant, i.e., remains unchanged, during a Lorentz transformation. Conversely, we may say that every coordinate transformation which preserves this differential interval is a Lorentz transformation. If in some inertial system dx 2 C dy 2 C dz 2 < c 2 dt 2 ds is a time-like interval , but if (7.21)
dx 2 C dy 2 C dz 2 > c 2 dt 2 ds is a space-like interval , whereas
dx 2 C dy 2 C dz 2 D c 2 dt 2
7.1.2.5 Four-vector elds
Any quantity which relative to any coordinate system has a quadruple of real numbers and transforms in the same way as the position four-vector x does, is called a four-vector. In analogy with the notation for the position four-vector we introduce the notation a D .a0 ; a/ for a general contravariant four-vector eld in L4 and nd that the lowering of index rule, formula (7.7) on page 148, for such an arbitrary four-vector yields the dual covariant four-vector eld a .x / D a .x / D .a0 .x /; a.x // (7.24)
The scalar product between this four-vector eld and another one b .x / is a .x /b .x / D .a0 ; a/ .b 0 ; b/ D a0 b 0 a b (7.25)
which is a scalar eld , i.e., an invariant scalar quantity .x / which depends on time and space, as described by x D .ct; x; y; z/.
R A
is a light-like interval ; we may also say that in this case we are on the light cone. A vector which has a light-like interval is called a null vector. The time-like, space-like or light-like aspects of an interval ds are invariant under a Lorentz transformation. I.e., it is not possible to change a time-like interval into a spacelike one or vice versa via a Lorentz transformation.
FT
(7.22) (7.23)
152 j
7.1.2.6 The Lorentz transformation matrix

Introducing the transformation matrix 0 B B DB @ 0 0 0 0 0 0 1 0 1 0 0C C C 0A 1
(7.26)
the linear Lorentz transformation (7.3) on page 146, i.e., the coordinate transformation x ! x 0 D x 0 .x 0 ; x 1 ; x 2 ; x 3 /, from one inertial system to another inertial system 0 in the standard conguration, can be written
FT
x0 D x D 1 C 2 1 C 1 2
(7.27)
7.1.2.7 The Lorentz group
It is easy to show, by means of direct algebra, that two successive Lorentz transformations of the type in equation (7.27) above, and dened by the speed parameters 1 and 2 , respectively, correspond to a single transformation with speed parameter
R A D
7.1.3 Minkowski space
(7.28)
This means that the nonempty set of Lorentz transformations constitutes a closed algebraic structure with a binary operation (multiplication) that is associative. Furthermore, one can show that this set possesses at least one identity element and at least one inverse element. In other words, this set of Lorentz transformations constitutes a mathematical group. However tempting, we shall not make any further use of group theory.
Specifying a point x D .x 0 ; x 1 ; x 2 ; x 3 / in 4D space-time is a way of saying that something takes place at a certain time t D x 0 =c and at a certain place .x; y; z/ D .x 1 ; x 2 ; x 3 /. Such a point is therefore called an event. The trajectory for an event as a function of time and space is called a world line. For instance, the world line for a light ray that propagates in vacuum (free space) is the trajectory x 0 D x 1 .
j 153
X0 X
00
x 01 x1
Figure 7.2: Minkowski space can be considered an ordinary Euclidean space where a Lorentz transformation from .x 1 ; X 0 D ict / to .x 01 ; X 00 D ict 0 / corresponds to an ordinary rotation through an angle . This rotation leaves the Euclidean distance 2 2 x 1 C X 0 D x 2 c 2 t 2 invariant.
X 0 D ix 0 D ict X 1 D x1 X Dx X Dx p
3 2 2 3
dS D ids where i D
1, we see that equation (7.17) on page 150 transforms into dS 2 D .dX 0 /2 C .dX 1 /2 C .dX 2 /2 C .dX 3 /2
R A
X 00 D
01
i.e., into a 4D differential form that is positive denite just as is ordinary 3D Euclidean space R3 . We shall call the 4D Euclidean space constructed in this way the Minkowski space M4 .3 As before, it sufces to consider the simplied case where the relative motion between and 0 is along the x axes. Then dS 2 D .dX 0 /2 C .dX 1 /2 D .dX 0 /2 C .dx 1 /2 (7.31)
and we consider the X 0 and X 1 D x 1 axes as orthogonal axes in a Euclidean space. As in all Euclidean spaces, every interval is invariant under a rotation of the X 0 x 1 plane through an angle into X 00 x 01 : x 1 sin C X 0 cos
1 0
x D x cos C X sin
See gure 7.2. If we introduce the angle ' D i , often called the rapidity or the Lorentz boost parameter, and transform back to the original space and time variables by
FT
(7.29a) (7.29b) (7.29d) (7.29c) (7.29e) (7.30)
3
Introducing
The fact that our Riemannian space can be transformed in this way into a Euclidean one means that it is, strictly speaking, a pseudo-Riemannian space.
(7.32a) (7.32b)
154 j
using equation (7.29) on the preceding page backwards, we obtain ct 0 D

0
FT
x sinh ' C ct cosh ' ct sinh ' x D x cosh ' sinh ' D cosh ' D tanh ' D tanh '1 C tanh '2 1 C tanh '1 tanh '2
Figure 7.3: Minkowski diagram depicting geometrically the transformation (7.33) from the unprimed system to the primed system. Here w denotes the world line for an event and the line x 0 D x 1 , x D ct the world line for a light ray in vacuum. Note that the event P is simultaneous with all points on the x 1 axis (t D 0), including the origin O. The event P 0 , which is simultaneous with all points on the x 0 axis, including O 0 D O, to an observer at rest in the primed system, is not simultaneous with O in the unprimed system but occurs there at time jP P 0 j =c.
x D ct
0
w x 00 x0 D x1
' P0 ' O D O0 P ct x1 D x
x 01
(7.33a) (7.33b)
R A D
tanh.'1 C '2 / D
which are identical to the original transformation equations (7.3) on page 146 if we let (7.34a) (7.34b) (7.34c)
It is therefore possible to envisage the Lorentz transformation as an ordinary rotation in the 4D Euclidean space M4 . Such a rotation in M4 corresponds to a coordinate change in L4 as depicted in gure 7.3. equation (7.28) on page 152 for successive Lorentz transformation then corresponds to the tanh addition formula (7.35)
The use of ict and M4 , which leads to the interpretation of the Lorentz transformation as an ordinary rotation, may, at best, be illustrative, but is not very physical. Besides, if we leave the at L4 space and enter the curved space of general relativity, the ict trick will turn out to be an impasse. Let us therefore immediately return to L4 where all components are real valued.
7.2. Covariant classical mechanics
j 155
7.2 Covariant classical mechanics

The invariance of the differential distance ds in L4 , and the associated differential proper time d [see equation (7.18) on page 150] allows us to dene the four-velocity 0 1 dx u D d B C B C v Bs c C D .u0 ; u/ ;s D .c; v / D B C 2 2 A @ v v 1 1 c2 c2 (7.36)
p D m0
dx d
B C B m0 c m0 v C C D .p 0 ; p/ D m0 .c; v / D B s ;s B C @ v2 v2 A 1 1 c2 c2
From this we see that we can write p D mv where m D m0 D s
R A
1 v2 c2 m0 c 2 cp 0 D m0 c 2 D s D mc 2 v2 1 c2 cp D .cp 0 ; cp/ D .E; cp/
m0
We can interpret this such that the Lorentz covariance implies that the mass-like term in the ordinary 3D linear momentum is not invariant. A better way to look at this is that p D mv D m0 v is the covariantly correct expression for the kinetic three-momentum. Multiplying the zeroth (time) component of the four-momentum p by the scalar invariant c, we obtain
Since this component has the dimension of energy and is the result of a Lorentzcovariant description of the motion of a particle with its kinetic momentum described by the spatial components of the four-momentum, equation (7.37) above, we interpret cp 0 as the total energy E. Hence, (7.41)
FT
(7.37) (7.38) (7.39) (7.40)
which, when multiplied with the scalar invariant m0 yields the four-momentum 0 1
156 j
Scalar multiplying this four-vector with itself, we obtain cp cp D c 2 p p D c 2 .p 0 /2 .p 1 /2 .p 2 /2 .p 3 /2 (7.42)
D .E; cp/ .E; cp/ D E 2 c 2 p2 v2 .m0 c 2 /2 1 D .m0 c 2 /2 D 2 c2 1 v2

c
Since this is an invariant, this equation holds in any inertial frame, particularly in the frame where p D 0 and there we have E D m0 c 2 This is probably the most famous formula in physics history. (7.43)
7.3 Covariant classical electrodynamics
Let us consider a charge density which in its rest inertial system is denoted by 0 . The four-vector (in contravariant component form)
R A
j D
0
FT
dx d D
0u
.c; v / D . c; v /
(7.44)
with
(7.45)
is the four-current. The contravariant form of the four-del operator @ D @=@x is dened in equation (M.100) on page 239 and its covariant counterpart @ D @=@x in equation (M.101) on page 239, respectively. As is shown in example M.9 on page 242, the dAlembert operator is the scalar product of the four-del with itself:
2
D@ @ D@ @ D
1 @2 c 2 @t 2
r2
(7.46)
Since it has the characteristics of a four-scalar, the dAlembert operator is invariant and, hence, the homogeneous wave equation 2 f .t; x/ D 0 is Lorentz covariant.
7.3. Covariant classical electrodynamics
j 157
7.3.1 The four-potential

If we introduce the four-potential A D c ;A (7.47)
where is the scalar potential and A the vector potential, dened in section 3.3 on page 39, we can write the uncoupled inhomogeneous wave equations, equations (3.32) on page 41, in the following compact (and covariant) way:
2
A D
0j
(7.48)
With the help of the above, we can formulate our electrodynamic equations covariantly. For instance, the covariant form of the equation of continuity, equation (1.25) on page 10 is @ j D0
and the Lorenz-Lorentz gauge condition, equation (3.31) on page 41, can be written
The Lorenz-Lorentz gauge is sometimes called the covariant gauge. The gauge transformations (3.66) on page 47 in covariant form are A 7! A0 D A C @ .x / (7.51)
If only one dimension Lorentz contracts (for instance, due to relative motion along the x direction), a 3D spatial volume element transforms according to 1 s 1 q v2 c2
R A
dV D d3x D dV0 D dV0 1 2 D dV0 dV D
0 dV0
@ A D0
where dV0 denotes the volume element as measured in the rest system, then from equation (7.45) on the facing page we see that (7.53)
i.e., the charge in a given volume is conserved. We can therefore conclude that the elementary electric charge is a universal constant.
FT
(7.49) (7.50) (7.52)
158 j
7.3.2 The Linard-Wiechert potentials

Let us now solve the the inhomogeneous wave equations (3.32) on page 41 in vacuum for the case of a well-localised charge q 0 at a source point dened by the position four-vector x 0 .x 00 D ct 0 ; x 01 ; x 02 ; x 03 /. The eld point (observation point) is denoted by the position four-vector x D .x 0 D ct; x 1 ; x 2 ; x 3 /. In the rest system we know that the solution is simply 0 1 q .A /0 D ;A D ;0 (7.54) c 4 "0 c jx x0 j0 vD0 where jx x0 j0 is the usual distance from the source point to the eld point, evaluated in the rest system (signied by the index 0). Let us introduce the relative position four-vector between the source point and the eld point: R Dx
FT
x 0 D .c.t t 0 /; x x0 / t 0 /; .x x0 // D c 2 .t t 0 /2 x x0 D c.t t 0/ R R D0 R D . x x0 ; x x0 /
(7.55)
Scalar multiplying this relative four-vector with itself, we obtain R R D .c.t t 0 /; x x0 / .c.t
2 x0 (7.56)
R A D
We know that in vacuum the signal (eld) from the charge q 0 at x 0 propagates to x with the speed of light c so that (7.57)
Inserting this into equation (7.56) above, we see that (7.58)
or that equation (7.55) above can be written (7.59)
Now we want to nd the correspondence to the rest system solution, equation (7.54) above, in an arbitrary inertial system. We note from equation (7.36) on page 155 that in the rest system 0 1 B C B C v Bs c C .u /0 D B ;s D .c; 0/ C 2 2 A @ v v 1 1 c2 c 2 vD0 (7.60)
j 159
and .R /0 D .x x0 ; x x 0 /0 D . x x0 0 ; .x x 0 /0 / (7.61)
As all scalar products, u R is invariant, which means that we can evaluate it in any inertial system and it will have the same value in all other inertial systems. If we evaluate it in the rest system the result is: u R D u R D .u /0 .R /0 0 D .c; 0/ .x x0 0 ; .x x0 /0 / D c x (7.62)
x0 0
We therefore see that the expression
Generalising expression (7.1) on page 146 to vector form: D v O

def
R A
v c
def s x
subject to the condition R R D 0 has the proper transformation properties (proper tensor form) and reduces, in the rest system, to the solution equation (7.54) on the facing page. It is therefore the correct solution, valid in any inertial system. According to equation (7.36) on page 155 and equation (7.59) on the facing page u R D .c; v / x x0 ; .x x0 / D c x x0 v .x x0 / (7.64)
and introducing
x0
v .x x0 / x c
we can write
D
and
u R D cs
u D cu R
1 v ; cs c 2 s
from which we see that the solution (7.63) can be written q0 1 v A .x / D ; 2 D ;A 4 "0 cs c s c
FT
(7.65) x0 .x x0 / (7.66) (7.67) (7.68) (7.69)
A D
q0 u 4 "0 cu R
(7.63)
160 j
where in the last step the denition of the four-potential, equation (7.47) on page 157, was used. Writing the solution in the ordinary 3D way, we conclude that for a very localised charge volume, moving relative an observer with a velocity v , the scalar and vector potentials are given by the expressions 1 q0 1 q0 D 0j .x x0 / 4 "0 s 4 "0 jx x 0 0 q v q v A.t; x/ D D 2 s 2 jx 0j 4 "0 c 4 "0 c x .x x0 / .t; x/ D (7.70a) (7.70b)
These potentials are the Linard-Wiechert potentials that we derived in a more complicated and restricted way in subsection 6.5.1 on page 119.
Consider a vectorial (cross) product c between two ordinary vectors a and b: cDa bD O ij k ai bj xk
D .a2 b3
We notice that the kth component of the vector c can be represented as ck D ai bj aj bi D cij D cj i ; i; j k
FT
a3 b2 /O 1 C .a3 b1 x a1 b3 /O 2 C .a1 b2 x r ED @B @t BDr ED r A @A @t @j Ai @ t Ai
7.3.3 The electromagnetic eld tensor
a2 b1 /O 3 x (7.71)
(7.72)
R A D
Bij D
In other words, the pseudovector c D a b can be considered as an antisymmetric tensor of rank two. The same is true for the curl operator r operating on a polar vector. For instance, the Maxwell equation (7.73)
can in this tensor notation be written @Ej @Ei @Bij D (7.74) i j @x @x @t We know from chapter 3 on page 35 that the elds can be derived from the electromagnetic potentials in the following way: (7.75a) (7.75b)
In component form, this can be written @Aj @Ai D @i Aj @x i @x j @ @Ai Ei D D @i i @x @t (7.76a) (7.76b)
j 161
From this, we notice the clear difference between the axial vector (pseudovector) B and the polar vector (ordinary vector) E. Our goal is to express the electric and magnetic elds in a tensor form where the components are functions of the covariant form of the four-potential, equation (7.47) on page 157: A D ;A (7.77) c Inspection of (7.77) and equation (7.76) on the facing page makes it natural to dene the four-tensor F D @A @x @A D@ A @x @ A (7.78)
This anti-symmetric (skew-symmetric), four-tensor of rank 2 is called the electromagnetic eld tensor or the Faraday tensor. In matrix representation, the contravariant eld tensor can be written 1 0 0 Ex =c Ey =c Ez =c BE =c 0 Bz By C C B x .F / D B (7.79) C @Ey =c Bz 0 Bx A Ez =c By Bx 0
R A
F D
We note that the eld tensor is a sort of four-dimensional curl of the fourpotential vector A . The covariant eld tensor is obtained from the contravariant eld tensor in the usual manner by index lowering F D @ A @ A (7.80)
which in matrix representation becomes 0 0 Ex =c B E =c 0 B x F DB @ Ey =c Bz Ez =c By
Comparing formula (7.81) above with formula (7.79) we see that the covariant eld tensor is obtained from the contravariant one by a transformation E ! E. That the two Maxwell source equations can be written @ F D
0j
is immediately observed by explicitly solving this covariant equation. Setting D 0, corresponding to the rst/leftmost column in the matrix representation
FT
Ey =c Bz 0 Bx 1 Ez =c By C C C Bx A 0 (7.81) (7.82)
162 j
of the covariant component form of the electromagnetic eld tensor, F , i.e., equation (7.79) on the previous page, we see that @F 10 @F 20 @F 30 1 @Ex @Ey @Ez @F 00 C C C D0C C C @x 0 @x 1 @x 2 @x 3 c @x @y @z (7.83) 1 0 D r E D 0j D 0c c or, equivalently (recalling that "0
0
D 1=c 2 ), "0 (7.84)
r ED
@F 01 @F 11 @F 21 @F 31 C C C D @x 0 @x 1 @x 2 @x 3 D
1 0j D 0
FT
1 @Ex @Bz C0C c 2 @t @y @By @z @By @z "0
0
which we recognise as the Maxwell source equation for the electric eld, equation (1.49a) on page 15. For D 1 [the second column in equation (7.79) on the previous page], equation (7.82) on the preceding page yields (7.85)
vx
This result can be rewritten as @Bz @y or, equivalently, as
@Ex D @t
0 jx
(7.86)
R A
.r B/x D and similarly for form as r BD
@Ex (7.87) @t D 2; 3. In summary, we can write the result in three-vector

0 jx
C "0
@E (7.88) @t which we recognise as the Maxwell source equation for the magnetic eld, equation (1.49d) on page 15. With the help of the fully antisymmetric pseudotensor of rank 4 8 1 if ; ; ; is an even permutation of 0,1,2,3 < D 0 (7.89) if at least two of ; ; ; are equal : 1 if ; ; ; is an odd permutation of 0,1,2,3
0 j.t; x/
C "0
which can be viewed as a generalisation of the Levi-Civita tensor, formula (M.28) on page 227, we can introduce the dual electromagnetic tensor
?
1 2
F D
(7.90)
7.4. Bibliography
j 163
with the further property

? ?
(7.91)
In matrix representation the dual eld tensor is 0 0 Bx By BB 0 Ez =c B x ? F DB @By Ez =c 0 Bz Ey =c Ex =c
1 Bz Ey =c C C C Ex =c A 0
(7.92)
can then be written
Explicit evaluation shows that this corresponds to (no summation!) @ F C@ F
R
@
?
sometimes referred to as the Jacobi identity. Hence, equation (7.82) on page 161 and equation (7.96) above constitute Maxwells equations in four-dimensional formalism. It is interesting to note that equation (7.82) on page 161 and F D
0 jm
where jm is the magnetic four-current, represent the covariant form of Diracs symmetrised Maxwell equations (1.50) on page 16.
7.4 Bibliography
[36] J. A HARONI, The Special Theory of Relativity, second, revised ed., Dover Publications, Inc., New York, 1985, ISBN 0-486-64870-2. [37] A. O. BARUT, Electrodynamics and Classical Theory of Fields and Particles, Dover Publications, Inc., New York, NY, 1980, ISBN 0-486-64038-8.
A FT
r ED @B @t (7.93) r BD0 (7.94) @
?
i.e., the dual eld tensor is obtained from the ordinary eld tensor by the duality transformation E ! cB and B ! E=c. The covariant form of the two Maxwell eld equations
D0
(7.95)
C @ F D 0
(7.96)
(7.97)
164 j
[38] R. B ECKER, Electromagnetic Fields and Interactions, Dover Publications, Inc., New York, NY, 1982, ISBN 0-486-64290-9. [39] W. T. G RANDY, Introduction to Electrodynamics and Radiation, Academic Press, New York and London, 1970, ISBN 0-12-295250-2. [40] L. D. L ANDAU AND E. M. L IFSHITZ, The Classical Theory of Fields, fourth revised English ed., vol. 2 of Course of Theoretical Physics, Pergamon Press, Ltd., Oxford . . . , 1975, ISBN 0-08-025072-6. [41] F. E. L OW, Classical Field Theory, John Wiley & Sons, Inc., New York, NY . . . , 1997, ISBN 0-471-59551-9. [42] H. M UIRHEAD, The Special Theory of Relativity, The Macmillan Press Ltd., London, Beccles and Colchester, 1973, ISBN 333-12845-1.
[44] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0-201-05702-6. [45] J. J. S AKURAI, Advanced Quantum Mechanics, Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1967, ISBN 0-201-06710-2. [46] B. S PAIN, Tensor Calculus, third ed., Oliver and Boyd, Ltd., Edinburgh and London, 1965, ISBN 05-001331-9.
R A
FT
[43] C. M LLER, The Theory of Relativity, second ed., Oxford University Press, Glasgow . . . , 1972.
ELECTROMAGNETIC FIELDS AND PARTICLES

In previous chapters, we calculated the electromagnetic elds and potentials from arbitrary, but prescribed distributions of charges and currents. In this chapter we rst study the opposite situation, viz., the dynamics of charged particles in arbitrary, but prescribed electromagnetic elds. Then we go on to consider the general problem of interaction between electric and magnetic elds and electrically charged particles. The analysis is based on Lagrangian and Hamiltonian methods, is fully covariant, and yields results which are relativistically correct.
8.1 Charged particles in an electromagnetic eld
8.1.1 Covariant equations of motion
We will show that for our problem we can derive the correct equations of motion by using in four-dimensional L4 a function with similar properties as a Lagrange function in 3D and then apply a variational principle. We will also show that we can nd a Hamiltonian-type function in 4D and solve the corresponding Hamilton-type equations to obtain the correct covariant formulation of classical electrodynamics.
8.1.1.1 Lagrangian formalism

In analogy with particle dynamics in 3D Euclidean space, we introduce a generalised 4D action Z S4 D L4 .x ; u / d (8.1) where d is the proper time dened via equation (7.18) on page 150, and L4 acts as a kind of generalisation to the common 3D Lagrangian. As a result, the 165
R A
We rst establish a relativistically correct theory describing the motion of charged particles in prescribed electric and magnetic elds. From these equations we may then calculate the charged particle dynamics in the most general case.
FT
166 j
8 . E L E C TROMAGNETIC FIELDS AND PARTICLES
variational principle Z S4 D
0 1
L4 .x ; u / d D 0
(8.2)
with xed endpoints 0 ; 1 must be fullled. We require that L4 is a scalar invariant which does not contain higher than the second power of the four-velocity u in order that the resulting equations of motion be linear. According to formula (M.174) on page 252 the ordinary 3D Lagrangian is the difference between the kinetic and potential energies. A free particle has only kinetic energy. If the particle mass is m0 then in 3D the kinetic energy is m0 v 2 =2. This suggests that in 4D the Lagrangian for a free particle should be
FT
1 m0 u u C qu A .x / 2 u D dx d
Lfree D 4
1 m0 u u 2
(8.3)
Again drawing inferences from analytical mechanics in 3D, we introduce a generalised interaction between the particles and the electromagnetic eld with the help of the four-potential given by equation (7.77) on page 161 in the following way L4 D (8.4)
R A
0
We call this the four-Lagrangian and shall now show how this function, together with the variation principle, formula (8.2), yields Lorentz covariant results which are physically correct. The variation principle (8.2) with the 4D Lagrangian (8.4) inserted, leads to Z 1 m0 S4 D u u C qu A .x / d 2 Z 10 m0 @.u u / @A D u C q A u C u x d (8.5) 2 @u @x 0 Z 1 D m0 u u C q A u C u @ A x d D 0
According to equation (7.36) on page 155, the four-velocity is (8.6)
which means that we can write the variation of u as a total derivative with respect to : dx d .x / u D D (8.7) d d
8.1. Charged particles in an electromagnetic eld
j 167
Inserting this into the rst two terms in the last integral in equation (8.5) on the facing page, we obtain Z 1 d d .x / C qA .x / C qu @ A x d S4 D m0 u (8.8) d d 0 Partial integration in the two rst terms in the right hand member of (8.8) gives Z 1 dA du x q x C qu @ A x d (8.9) S4 D m0 d d 0 where the integrated parts do not contribute since the variations at the endpoints vanish. A change of irrelevant summation index from to in the rst two terms of the right hand member of (8.9) yields, after moving the ensuing common factor x outside the parenthesis, the following expression: Z 1 du dA S4 D m0 q C qu @ A x d (8.10) d d 0 Applying well-known rules of differentiation and the expression (7.36) for the four-velocity, we can express dA =d as follows:
Since, according to the variational principle, this expression shall vanish and x is arbitrary between the xed end points 0 and 1 , the expression inside in the integrand in the right hand member of equation (8.12) above must vanish. In other words, we have found an equation of motion for a charged particle in a prescribed electromagnetic eld: du D qu @ A @ A (8.13) d With the help of formula (7.80) on page 161 for the covariant component form of the eld tensor, we can express this equation in terms of the electromagnetic eld tensor in the following way: m0 du D qu F (8.14) d This is the sought-for covariant equation of motion for a particle in an electromagnetic eld. It is often referred to as the Minkowski equation. As the reader may easily verify, the spatial part of this 4-vector equation is the covariant (relativistically correct) expression for the Newton-Lorentz force equation. m0
R A
dA @A dx D D@ A u (8.11) d @x d By inserting this expression (8.11) into the second term in right-hand member of equation (8.10) above, and noting the common factor qu of the resulting term and the last term, we obtain the nal variational principle expression Z 1 du S4 D m0 C qu @ A @ A x d (8.12) d 0
FT
168 j
8.1.1.2 Hamiltonian formalism

The usual Hamilton equations for a 3D space are given by equation (M.186) on page 253 in appendix M on page 221. These six rst-order partial differential equations are @H dqi D @pi dt dpi @H D @qi dt (8.15a) (8.15b)
FT
p D @L4 @u H4 D p u L4 @H4 dx D @p d @H4 dp D @x d
where H.pi ; qi ; t / D pi qi L.qi ; qi ; t/ is the ordinary 3D Hamiltonian, qi is a P P generalised coordinate and pi is its canonically conjugate momentum. We seek a similar set of equations in 4D space. To this end we introduce a canonically conjugate four-momentum p in an analogous way as the ordinary 3D conjugate momentum (8.16)
and utilise the four-velocity u , as given by equation (7.36) on page 155, to dene the four-Hamiltonian
R A
1
(8.17)
With the help of these, the position four-vector x , considered as the generalised four-coordinate, and the invariant line element ds, dened in equation (7.18) on page 150, we introduce the following eight partial differential equations: (8.18a) (8.18b)
Recall that in 3D, the Hamiltonian equals the total energy.
which form the four-dimensional Hamilton equations. Our strategy now is to use equation (8.16) above and equations (8.18) above to derive an explicit algebraic expression for the canonically conjugate momentum four-vector. According to equation (7.41) on page 155, c times a fourmomentum has a zeroth (time) component which we can identify with the total energy. Hence we require that the component p 0 of the conjugate fourmomentum vector dened according to equation (8.16) above be identical to the ordinary 3D Hamiltonian H divided by c and hence that this cp 0 solves the Hamilton equations, equations (8.15) above.1 This latter consistency check is left as an exercise to the reader.
8.1. Charged particles in an electromagnetic eld
j 169
Using the denition of H4 , equation (8.17) on the preceding page, and the expression for L4 , equation (8.4) on page 166, we obtain H4 D p u L4 D p u 1 m0 u u 2 qu A .x / (8.19)
Furthermore, from the denition (8.16) of the canonically conjugate four-momentum p , we see that @ @L4 1 D m0 u u C qu A .x / D m0 u C qA (8.20) p D @u @u 2 Inserting this into (8.19), we obtain H4 D m0 u u C qA u 1 m0 u u 2 qu A .x / D 1 m0 u u 2 (8.21)
Since the four-velocity scalar-multiplied by itself is u u D c 2 , we clearly see from equation (8.21) above that H4 is indeed a scalar invariant, whose value is simply H4 D m0 c 2 2 (8.22)
However, at the same time (8.20) provides the algebraic relationship u D 1 .p m0 qA /
and if this is used in (8.21) to eliminate u , one gets m0 1 1 .p H4 D p qA qA / 2 m0 m0 1 .p D qA qA / p 2m0 1 D p p 2qA p C q 2 A A 2m0
R A
@H4 D @x D D D q .p m0 q m0 u m0 @A qu @x dp D d qA / @A @x @A @x m0 du d q @A u @x
That this four-Hamiltonian yields the correct covariant equation of motion can be seen by inserting it into the four-dimensional Hamilton equations (8.18) and using the relation (8.23):
FT
(8.23) (8.24) (8.25)
170 j
where in the last step equation (8.20) on the preceding page was used. Rearranging terms, and using equation (7.80) on page 161, we obtain m0 du D qu d @ A @ A D qu F (8.26)
which is identical to the covariant equation of motion equation (8.14) on page 167. We can therefore safely conclude that the Hamiltonian in question yields correct results. Recalling expression (7.47) on page 157 for the four-potential, and representing the canonically conjugate four-momentum as p D .p 0 ; p/, we obtain the following scalar products:
FT
.p A/ .A/2 q2 2qp A C q 2 .A/2 C 2 c
.p qA/2 2
p p D .p 0 /2 1 0 A p D p c 1 A A D 2 2 c
.p/2
(8.27a) (8.27b) (8.27c)
R A
.p 0 /2 2q 0 p c .p/2 with two possible solutions p0 D
Inserting these explicit expressions into equation (8.24) on the preceding page, and using the fact that H4 is equal to the scalar value m0 c 2 =2, as derived in equation (8.22) on the previous page, we obtain the equation m0 c 2 1 2 q2 D .p 0 /2 .p/2 q p 0 C 2q.p A/ C 2 2 q 2 .A/2 2 2m0 c c (8.28) which is a second-order algebraic equation in p 0 : m2 c 2 D 0 0 (8.29)
q c
.p
qA/2 C m2 c 2 0
(8.30)
Since the zeroth component (time component) p 0 of a four-momentum vector p multiplied by c represents the energy [cf. equation (7.41) on page 155], the positive solution in equation (8.30) above must be identied with the ordinary Hamilton function H divided by c. Consequently, q H cp 0 D q C c .p qA/2 C m2 c 2 (8.31) 0 is the ordinary 3D Hamilton function for a charged particle moving in scalar and vector potentials associated with prescribed electric and magnetic elds.
8.2. Covariant eld theory
j 171
The ordinary Lagrange and Hamilton functions L and H are related to each other by the 3D transformation [cf. the 4D transformation (8.17) between L4 and H4 ] LDp v H (8.32)
Using the the explicit expressions given by equation (8.31) and equation (8.32), we obtain the explicit expression for the ordinary 3D Lagrange function LDp v q q c .p qA/2 C m2 c 2 0 (8.33)
and if we make the identication p qA D s
v2 c2
where the quantity mv is the usual kinetic momentum, we can rewrite this expression for the ordinary Lagrangian as follows: L D qA v C mv 2 D mv
2
q A v/
q c m2 v 2 C m2 c 2 0 mc D
2
R A
q. q C qA v
What we have obtained is the relativistically correct (covariant) expression for the Lagrangian describing the mechanical motion of a charged particle in scalar and vector potentials associated with prescribed electric and magnetic elds.
8.2 Covariant eld theory
So far, we have considered two classes of problems. Either we have calculated the elds from given, prescribed distributions of charges and currents, or we have derived the equations of motion for charged particles in given, prescribed elds. Let us now put the elds and the particles on an equal footing and present a theoretical description which treats the elds, the particles, and their interactions in a unied way. This involves transition to a eld picture with an innite number of degrees of freedom. We shall rst consider a simple mechanical problem whose solution is well known. Then, drawing inferences from this model problem, we apply a similar view on the electromagnetic problem.
FT
D mv (8.34) s m0 c
2
m0 v
v2 c2
(8.35)
172 j
Figure 8.1: A one-dimensional chain consisting of N discrete, identical mass points m, connected to their neighbours with identical, ideal springs with spring constants k. The equilibrium distance between the neighbouring mass points is a and i 1 .t /, i .t /, iC1 .t / are the instantaneous deviations, along the x axis, of positions of the .i 1/th, i th, and .i C 1/th mass point, respectively.
i 1
iC1
k a
k a
k a
k a
m x
8.2.1 Lagrange-Hamilton formalism for elds and interactions

Consider the situation, illustrated in gure 8.1, with N identical mass points, each with mass m and connected to its neighbour along a one-dimensional straight line, which we choose to be the x axis, by identical ideal springs with spring constants k (Hookes law). At equilibrium the mass points are at rest, distributed evenly with a distance a to their two nearest neighbours so that the equilibrium O coordinate for the i th particle is xi D iax . After perturbation, the motion of O mass point i will be a one-dimensional oscillatory motion along x. Let us denote the deviation for mass point i from its equilibrium position by i .t/O . x As is well known, the solution to this mechanical problem can be obtained if we can nd a Lagrangian (Lagrange function) L which satises the variational equation Z L. i ; Pi ; t/ dt D 0 (8.36) According to equation (M.174) on page 252, the Lagrangian is L D T V where T denotes the kinetic energy and V the potential energy of a classical mechanical system with conservative forces. In our case the Lagrangian is 1 X h P2 LD m i 2
i D1 N
R A D
LD where 1 Li D 2 " m P2 a i
FT
k.
i C1 i/ 2
(8.37)
Let us write the Lagrangian, as given by equation (8.37) above, in the following way:
N X i D1
a Li
(8.38)
ka
i C1
2 # (8.39)
j 173
is the so called linear Lagrange density, measured in J m 1 . If we now let N ! 1 and, at the same time, let the springs become innitesimally short according to the following scheme: a ! dx m dm ! D a dx ka ! Y
i C1 i
(8.40a) linear mass density Youngs modulus (8.40b) (8.40c) (8.40d)
a we obtain
@ @x
LD where
L
L dx
@ @ ; ; ;t @t @x
1 D 2
"
@ @t
Notice how we made a transition from a discrete description, in which the mass points were identied by a discrete integer variable i D 1; 2; : : : ; N , to a continuous description, where the innitesimal mass points were instead identied by O a continuous real parameter x, namely their position along x. A consequence of this transition is that the number of degrees of freedom for the system went from the nite number N to innity! Another consequence is that L has now become dependent also on the partial derivative with respect to x of the eld coordinate . But, as we shall see, the transition is well worth the cost because it allows us to treat all elds, be it classical scalar or vectorial elds, or wave functions, spinors and other elds that appear in quantum physics, on an equal footing. Under the assumption of time independence and xed endpoints, the variation principle (8.36) on the facing page yields: Z L dt ZZ @ @ D L ; ; dx dt @t @x 3 2 (8.43) ZZ @L @L @ @L @ 5 4 D C C dx dt @ @t @x @ @ @ @ D0
R A
@t @x
FT
(8.41) 2 Y @ @x 2 # (8.42)
174 j
As before, the last integral can be integrated by parts. This results in the expression 2 0 0 1 13 ZZ @L @ @ @L A @ @ @L A5 4 dx dt D 0 (8.44) @ @t @ @ @x @ @ @t @x where the variation is arbitrary (and the endpoints xed). This means that the integrand itself must vanish. If we introduce the functional derivative 0 1 L @L @ @ @L A D (8.45) @ @x @ @
@x
we can express this as
FT
L 1 0 @ @ @L A D0 @t @ @ @t @2 @t 2 Y @2 D0 @x 2 ! D0 1 @2 2 v' @t 2 @2 @x 2
@x @x
(8.46)
which is the one-dimensional Euler-Lagrange equation. Inserting the linear mass point chain Lagrangian density, equation (8.42) on the previous page, into equation (8.46) above, we obtain the equation of motion for our one-dimensional linear mechanical structure. It is: (8.47a)
R A
or
(8.47b)
i.e., the one-dimensional wave equation for compression waves which propagate p with phase speed v' D Y = along the linear structure. A generalisation of the above 1D results to a three-dimensional continuum is straightforward. For this 3D case we get the variational principle Z ZZ Z @ L dt D L d3x dt D L ; d4x @x 2 0 13 ZZ (8.48) @L @ @ @L A5 4 4 D dxD0 @ @x @ @ where the variation is arbitrary and the endpoints are xed. This means that the integrand itself must vanish: 1 0 @L @ @ @L A @L @L D @ D0 (8.49) @ @x @ @.@ / @ @
j 175
This constitutes the four-dimensional Euler-Lagrange equations. Introducing the three-dimensional functional derivative 0 1 L @L @ @ @L A D @ @x i @ @
@x i
(8.50)
we can express this as L @ @ @L A D0 @t @ @ @t 0 1 (8.51)
In analogy with particle mechanics (nite number of degrees of freedom), we may introduce the canonically conjugate momentum density .x / D .t; x/ D and dene the Hamilton density @ H ; ; i It D @x
@ @t
If, as usual, we differentiate this expression and identify terms, we obtain the following Hamilton density equations
R A
L tot D L mech C L interaction C L eld
@ @H D @ @t H @ D @t
The Hamilton density functions are in many ways similar to the ordinary Hamilton functions for a system of a nite number of particles and lead to similar results. However, they describe the dynamics of a continuous system of innitely many degrees of freedom.
8.2.1.1 The electromagnetic eld
Above, when we described the mechanical eld, we used a scalar eld .t; x/. If we want to describe the electromagnetic eld in terms of a Lagrange density L and Euler-Lagrange equations, it comes natural to express L in terms of the four-potential A .x /. The entire system of particles and elds consists of a mechanical part, a eld part and an interaction part. We therefore assume that the total Lagrange density L tot for this system can be expressed as (8.55)
FT
@L
@ @t
(8.52)
@ @ ; @t @x i
(8.53)
(8.54a) (8.54b)
176 j
where the mechanical part has to do with the particle motion (kinetic energy). It is given by L4 =V where L4 is given by equation (8.3) on page 166 and V is the volume. Expressed in the rest mass density %0 , the mechanical Lagrange density can be written
L mech D
1 %0 u u 2
(8.56)
The L interaction part describes the interaction between the charged particles and the external electromagnetic eld. A convenient expression for this interaction Lagrange density is
For the eld part L eld we choose the difference between magnetic and electric energy density (in analogy with the difference between kinetic and potential energy in a mechanical eld). With the help of the eld tensor, we express this eld Lagrange density as
L eld D
FT
1 4 F F
0
L interaction D j A
(8.57)
(8.58)
R A
L tot D
so that the total Lagrangian density can be written 1 1 %0 u u C j A C F 2 4 0 F (8.59)
From this we can calculate all physical quantities. Using L tot in the 3D Euler-Lagrange equations, equation (8.49) on page 174 (with replaced by A ), we can derive the dynamics for the whole system. For instance, the electromagnetic part of the Lagrangian density
L EM D L interaction C L eld D j A C
1 4
0
(8.60)
inserted into the Euler-Lagrange equations, expression (8.49) on page 174, yields two of Maxwells equations. To see this, we note from equation (8.60) above and the results in Example 8.1 that @L EM Dj @A (8.61)
j 177
Furthermore, @L EM 1 @ D @ @.@ A / 4 0 @ 1 @ D 4 0 @.@ A ( @ 1 @ D 4 0 @.@ A
/ /
@ F F @.@ A / h .@ A @ A /.@ A @ A @ A
@ A /
@ A @ A
(8.62) )
@ A @ A C @ A @ A D But @ @ A @ A @.@ A / @ D @ A @.@ A @ D @ A @.@ A @ D @ A @.@ A D 2@ A Similarly, 1 2

0
@ @ @ A C @ A @ A @.@ A / @.@ A / @ @ A C @ A @ A / @.@ A / @ @ A C @ A @ A / @.@ A / @ @ A C @ A @ A / @.@ A / D @ A
so that @
R
@L EM 1 D @ .@ A @.@ A / 0 @ F
@ @ A @ A D 2@ A @.@ A /
D
or
This means that the Euler-Lagrange equations, expression (8.49) on page 174, for the Lagrangian density L EM and with A as the eld quantity become @L EM @L EM 1 @ Dj @ F D0 (8.66) @A @.@ A / 0
which, according to equation (7.82) on page 161, is a Lorentz covariant formulation of Maxwells source equations.
A FT
(8.63) (8.64) @ A /D 1
0
@ @ A @ A @.@ A /
@ A @ A
@ F
(8.65)
0j
(8.67)
178 j
E X A M P L E 8.1
Field energy difference expressed in the eld tensor Show, by explicit calculation, that 1 4
0
1 D 2
B2
0
! "0 E
2
(8.68)
i.e., the difference between the magnetic and electric eld energy densities. From formula (7.79) on page 161 we recall that 0 BEx =c DB @Ey =c Ez =c 0 Ex =c 0 Bz By Ey =c Bz 0 Bx 1 Ez =c By C C Bx A 0
(8.69)
and from formula (7.81) on page 161 that 0 B Ex =c B D@ Ey =c Ez =c 0
FT
Ex =c 0 Bz By Ey =c Bz 0 Bx 1 Ez =c By C C Bx A 0
2 Ey =c 2 2 Ez =c 2 2 2Ey =c 2 2 2 2 2 2Ez =c 2 C 2Bx C 2By C 2Bz
(8.70)
where denotes the row number and direct substitution yields F F
the column number. Then, Einstein summation and
R A
D0
2 Ex =c 2
D F 00 F00 C F 01 F01 C F 02 F02 C F 03 F03
C F 10 F10 C F 11 F11 C F 12 F12 C F 13 F13 C F 20 F20 C F 21 F21 C F 22 F22 C F 23 F23 C F 30 F30 C F 31 F31 C F 32 F32 C F 33 F33
2 2 2 Ex =c 2 C 0 C Bz C By
(8.71)
2 2 2 Ey =c 2 C Bz C 0 C Bx 2 2 2 Ez =c 2 C By C Bx C 0
2 2Ex =c 2
2E 2 =c 2 C 2B 2 D 2.B 2
E 2 =c 2 /
or [cf. equation (2.17a) on page 26] 1 4

0
1 D 2
B2
0
1 c2
0
! E
2
1 D 2
B2
0
! "0 E
2
"0 2 .E 2
c2B 2/ (8.72)
where, in the last step, the identity "0
D 1=c 2 was used.
QED End of example 8.1
j 179
8.2.1.2 Other elds

In general, the dynamic equations for most any elds, and not only electromagnetic ones, can be derived from a Lagrangian density together with a variational principle (the Euler-Lagrange equations). Both linear and non-linear elds are studied with this technique. As a simple example, consider a real, scalar eld which has the following Lagrange density:
L D
1 @ 2
m2
(8.73)
Insertion into the 1D Euler-Lagrange equation, equation (8.46) on page 174, yields the dynamic equation
with the solution
which describes the Yukawa meson eld for a scalar meson with mass m. With D 1 @ c 2 @t
we obtain the Hamilton density 1h 2 H D c 2
R
@ F
which is positive denite. Another Lagrangian density which has attracted quite some interest is the Proca Lagrangian
L EM D L interaction C L eld D j A C
which leads to the dynamic equation
This equation describes an electromagnetic eld with a mass, or, in other words, massive photons. If massive photons do exist, large-scale magnetic elds, including those of the earth and galactic spiral arms, should be signicantly modied from what they ar to yield measurable discrepancies from their usual form. Space experiments of this kind on board satellites have led to stringent upper bounds on the photon mass. If the photon really has a mass, it will have an impact on electrodynamics as well as on cosmology and astrophysics.
A FT
D ei.k x
!t/ e mjxj
m2 / D 0
(8.74)
jxj
(8.75)
(8.76)
C .r /2 C m2
(8.77)
1 4
0
C m2 A A
(8.78)
m2 A D
0j
(8.79)
180 j
8.3 Bibliography
[47] A. O. BARUT, Electrodynamics and Classical Theory of Fields and Particles, Dover Publications, Inc., New York, NY, 1980, ISBN 0-486-64038-8. [48] V. L. G INZBURG, Applications of Electrodynamics in Theoretical Physics and Astrophysics, Revised third ed., Gordon and Breach Science Publishers, New York, London, Paris, Montreux, Tokyo and Melbourne, 1989, ISBN 2-88124-719-9. [49] H. G OLDSTEIN, Classical Mechanics, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1981, ISBN 0-201-02918-9. [50] W. T. G RANDY, Introduction to Electrodynamics and Radiation, Academic Press, New York and London, 1970, ISBN 0-12-295250-2.
[52] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0201-05702-6. [53] J. J. S AKURAI, Advanced Quantum Mechanics, Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1967, ISBN 0-201-06710-2. [54] D. E. S OPER, Classical Field Theory, John Wiley & Sons, Inc., New York, London, Sydney and Toronto, 1976, ISBN 0-471-81368-0.
R A
FT
[51] L. D. L ANDAU AND E. M. L IFSHITZ, The Classical Theory of Fields, fourth revised English ed., vol. 2 of Course of Theoretical Physics, Pergamon Press, Ltd., Oxford . . . , 1975, ISBN 0-08-025072-6.
ELECTROMAGNETIC FIELDS AND MAT TER
R A
D D D.t; xI E; B/ H D H.t; xI E; B/ 181
The microscopic Maxwell equations derived in chapter 1 on page 1, which in chapter 2 on page 21 were chosen as the axiomatic basis for the treatment in the remainder of the book, are valid on all scales where a classical description is good. They provide a correct physical picture for arbitrary eld and source distributions, on macroscopic and, under certain assumptions, microscopic scales. A more complete and accurate theory, valid also when quantum effects are signicant, is provided by quantum electrodynamics. QED gives a consistent description of how electromagnetic elds are quantised into photons and describes their intrinsic and extrinsic properties. However, this theory is beyond the scope of the current book. In a material medium, be it in a solid, uid or gaseous state or a combination thereof, it is sometimes convenient to replace the Maxwell-Lorentz equations (2.1) on page 21 by the corresponding macroscopic Maxwell equations in which auxiliary, derived elds are introduced. These auxiliary elds, viz., the electric displacement vector D (measured in C m 2 ) and the magnetising eld H (measured in A m 1 ), incorporate intrinsic electromagnetic properties of macroscopic matter, or properties that appear when the medium is immersed fully or partially in an electromagnetic eld. Consequently, they represent, respectively, electric and magnetic eld quantities in which, in an average sense, the material properties of the substances are already included. In the most general case, these derived elds are complicated, possibly non-local and nonlinear, functions of the primary elds E and B :
An example of this are chiral media. A general treatment of these elds will not be included here. Only simplied, but important and illuminating examples will be given.
FT
(9.1a) (9.1b)
182 j
9 . E L E CTROMAGNETIC FIELDS AND MATTER
9.1 Maxwells macroscopic theory

Under certain conditions, for instance for small magnitudes of the primary eld strengths E and B, we may assume that the response of a substance to the elds can be approximated by a linear one so that D H "E
1
(9.2) B (9.3)
9.1.1 Polarisation and electric displacement
R A
total
By writing the rst microscopic Maxwell-Lorentz equation (2.1a) on page 21 as in equation (6.25) on page 102, i.e., in a form where the total charge density .t; x/ is split into the charge density for free, true charges, true , and the charge density, pol , for bound polarisation charges induced by the applied eld E, as r ED .t; x/ D "0
true
FT
.t; x/ C "0
pol
i.e., that the electric displacement vector D.t; x/ is only linearly dependent on the electric eld E.t; x/, and the magnetising eld H.t; x/ is only linearly dependent on the magnetic eld B.t; x/. In this chapter we derive these linearised forms, and then consider a simple, explicit linear model for a medium from which we derive the expression for the dielectric permittivity ".t; x/, the magnetic susceptibility .t; x/, and the refractive index or index of refraction n.t; x/ of this medium. Using this simple model, we study certain interesting aspects of the propagation of electromagnetic particles and waves in the medium.
.t; x/
true
.t; x/
r P .t; x/ "0 (9.4)

2
and at the same time introducing the electric displacement vector (C m D.t; x/ D "0 E.t; x/ C P .t; x/
) (9.5)
one can reshufe expression (9.4) above to obtain r "0 E.t; x/ C P .t; x/ D r D.t; x/ D
true
.t; x/
(9.6)
This is one of the original macroscopic Maxwell equations. It is important to remember that only the induced electric dipole moment of matter, subject to the eld E, was included in the above separation into true and induced charge densities. Contributions to D from higher-order electric moments were neglected. This is one of the approximations assumed.
9.1. Maxwells macroscopic theory
j 183
Another approximation is the assumption that there exists a simple linear relationship between P and E in the material medium under consideration P .t; x/ D "0
e .t; x/E.t; x/
(9.7)
This approximation is often valid for regular media if the eld strength jEj is low enough. Here the variations in time and space of the the material dependent electric susceptibility, e , are usually on much slower and longer scales than for E itself.1 Inserting the approximation (9.7) into equation (9.5) on the facing page, we can write the latter D.t; x/ D ".t; x/E.t; x/ where, approximately, ".t; x/ D "0 1 C (9.8)
The fact that the relation between the dipole moment per unit volume P and the applied electric eld E is local in time and space is yet another approximation assumed in macroscopic Maxwell theory.
e .t; x/
For an electromagnetically anisotropic medium such as a magnetised plasma or a birefringent crystal , the susceptibility e or, equivalently the relative dielectric permittivity e .t; x/ D ".t; x/ D1C "0
e .t; x/
R A
will have to be replaced by a tensor. This would still describe a linear relationship between E and P but one where the linear proportionality factor, or, as we shall call it, the dispersive property of the medium, is dependent on the direction in space. In general, however, the relationship is not of a simple linear form as in equation (9.7) above but non-linear terms are important. In such a situation the principle of superposition is no longer valid and non-linear effects such as frequency conversion and mixing can be expected.2
9.1.2 Magnetisation and the magnetising eld
An analysis of the properties of magnetic media and the associated currents shows that three such types of currents exist: 1. In analogy with true charges for the electric case, we may have true currents j true , i.e., a physical transport of true (free) charges.
2. In analogy with the electric polarisation P there may be a form of charge transport associated with the changes of the polarisation with time. Such currents, induced by an external eld, are called polarisation currents and are identied with @P =@t.
FT
D "0 e .t; x/ (9.9) (9.10)
2
The nonlinearity of semiconductor diodes is used, e.g., in radio receivers to convert high radio frequencies into lower ones, or into the audible spectrum. These techniques are called heterodyning and demodulation, respectively. Another example of the nonlinear response of a medium is the Kerr effect.
184 j
3. There may also be intrinsic currents of a microscopic, often atomistic, nature that are inaccessible to direct observation, but which may produce net effects at discontinuities and boundaries. These magnetisation currents are denoted jM. Magnetic monopoles have not yet been unambiguously identied in experiments. So there is no correspondence in the magnetic case to the electric monopole moment, formula (6.16a) on page 100. The lowest order magnetic moment, corresponding to the electric dipole moment, formula (6.16b) on page 100, is the magnetic dipole moment [cf. the Fourier component expression (6.64) on page 110] Z 1 m.t/ D d3x 0 .x0 x0 / j.t 0 ; x0 / (9.11) 2 V0 Analogously to the electric case, one may, for a distribution of magnetic dipole moments in a volume, describe this volume in terms of its magnetisation, or magnetic dipole moment per unit volume, M. Via the denition of the vector potential A one can show that the magnetisation current and the magnetisation is simply related: jM D r M (9.12)
R A D
HD LHS D r H RHS D j true C
In a stationary medium we therefore have a total current which is (approximately) the sum of the three currents enumerated above: j total D j true C @P Cr @t M (9.13) B
One might then be led to think that the right-hand side (RHS) of the r Maxwell equation (2.1d) on page 21 should be @P true RHS D 0 j C Cr M @t
However, moving the term r M from the right hand side (RHS) to the left hand side (LHS) and introducing the magnetising eld (magnetic eld intensity, Ampre-turn density) as B
0
FT
M @P @D D j true C @t @t "0 @E @t
(9.14)
and using the denition for D, equation (9.5) on page 182, we nd that
9.1. Maxwells macroscopic theory
j 185
Hence, in this simplistic view, we would pick up a term "0 @E=@t which makes the equation inconsistent: the divergence of the left hand side vanishes while the divergence of the right hand side does not! Maxwell realised this and to overcome this inconsistency he was forced to add his famous displacement current term which precisely compensates for the last term the RHS expression.3 In chapter 1 on page 1, we discussed an alternative way, based on the postulate of conservation of electric charge, to introduce the displacement current. We may, in analogy with the electric case, introduce a magnetic susceptibility for the medium. Denoting it m , we can write H.t; x/ D where, approximately, .t; x/ D and m .t; x/ D .t; x/
0 0 1 1
.t; x/B.t; x/
(9.15)
This term, which ensures that electric charge is conserved also in non-stationary problems, is the one that makes it possible to turn the Maxwell equations into wave equations (see chapter 2 on page 21) and, hence, the term that, in a way, is the basis for radio communications and other engineering applications of the theory.
m .t; x/
D1C
is the relative permeability. In the case of anisotropy, m will be a tensor, but it is still only a linear approximation.4
9.1.3 Macroscopic Maxwell equations
Field equations, expressed in terms of the derived, and therefore in principle superuous, eld quantities D and H are obtained from the Maxwell-Lorentz microscopic equations (2.1) on page 21, by replacing the E and B in the two source equations by using the approximate relations formula (9.8) on page 183 and formula (9.15) above, respectively: r DD ED
true
R A
r @B @t r BD0 r H D j true C @D @t
This set of differential equations, originally derived by Maxwell himself, are called Maxwells macroscopic equations. Together with the boundary conditions and the constitutive relations, they describe uniquely (but only approximately) the properties of the electric and magnetic elds in matter and are convenient to use in certain simple cases, particularly in engineering applications.
FT
D
0 m .t; x/
(9.16)
m .t; x/
(9.17)
This is the case for the Hall effect which produces a potential difference across an electric conduction current channel, orthogonal to this current, in the presence of an external magnetic eld which is likewise perpendicular to the current. This effect was discovered 1879 by the US physicist E DW I N H E R B E RT H A L L (18551938).
(9.18a) (9.18b) (9.18c) (9.18d)
186 j
It should be recalled that Maxwell formulated these macroscopic equations before it was known that matter has an atomistic structure and that there exist electrically charged particles such as electrons and protons, which possess a quantum mechanical property called spin that gives rise to magnetism!
However, the structure of these equations rely on certain linear approximations and there are many situations where they are not useful or even applicable. Therefore, these equations, which are the original Maxwell equations (albeit expressed in their modern vector form as introduced by O L I V E R H E AV I S I D E ), should be used with some care.5
9.2 Phase velocity, group velocity and dispersion

If we introduce the phase velocity in the medium as c Dp (9.19) e m 0 p where, according to equation (1.13) on page 6, c D 1= "0 0 is the speed of light, i.e., the phase speed of electromagnetic waves, in vacuum. Associated with the phase speed of a medium for a wave of a given frequency ! we have a wave vector, dened as 1 v' D p " 1 Dp e "0 m
FT
O k k k D k v' D O
def
! v' v' v'
(9.20)
The ratio of the phase speed in vacuum and in the medium
R A
where the material dependent quantity n.t; x/
def
p c p def D e m D c " n v'
(9.21)
p c D e .t; x/m .t; x/ v'
(9.22)
In fact, there exist metamaterials where e and m are negative. For such materials, the refractive index becomes negative: nDi D
je j i
jm j
je m j1=2
Such negative refractive index materials, have quite remarkable electromagnetic properties.
is called the refractive index of the medium and describes its refractive and reective properties.6 In general n is a function of frequency. If the medium is anisotropic or birefringent, the refractive index is a rank-two tensor eld. Under our simplifying assumptions, in the material medium that we consider n D Const for each frequency component of the elds. In certain materials, the refractive index is larger than unity (e.g., glass and water at optical frequencies), in others, it can be smaller than unity (e.g., plasma and metals at radio and optical frequencies). It is important to notice that depending on the electric and magnetic properties of a medium, and, hence, on the value of the refractive index n, the phase speed in the medium can be smaller or larger than the speed of light: v' D c ! D n k (9.23)
9.3. Radiation from charges in a material medium
j 187
where, in the last step, we used equation (9.20) on the preceding page. If the medium has a refractive index which, as is usually the case, dependent on frequency !, we say that the medium is dispersive. Because in this case also k.!/ and !.k/, so that the group velocity @! (9.24) @k has a unique value for each frequency component, and is different from v' . Except in regions of anomalous dispersion, vg is always smaller than c. In a gas of free charges, such as a plasma, the refractive index is given by the expression vg D n2 .!/ D 1 where
2 !p D 2 !p
!2
(9.25)
X N q2 "0 m
9.3 Radiation from charges in a material medium

When electromagnetic radiation is propagating through matter, new phenomena may appear which are (at least classically) not present in vacuum. As mentioned earlier, one can under certain simplifying assumptions include, to some extent, the inuence from matter on the electromagnetic elds by introducing new, derived eld quantities D and H according to D D ".t; x/E D e .t; x/"0 E B D .t; x/H D m .t; x/
0H
R A
is the square of the plasma frequency !p . Here m and N denote the mass and number density, respectively, of charged particle species . In an inhomogeneous plasma, N D N .x/ so that the refractive index and also the phase and group velocities are space dependent. As can be easily seen, for each given frequency, the phase and group velocities in a plasma are different from each other. If the frequency ! is such that it coincides with !p at some point in the medium, then at that point v' ! 1 while vg ! 0 and the wave Fourier component at ! is reected there.
L 9.3.1 Vavilov-Cerenkov radiation

As we saw in subsection on page 128, a charge in uniform, rectilinear motion in vacuum does not give rise to any radiation; see in particular equation (6.140a)
FT
(9.26) (9.27) (9.28)
188 j
on page 129. Let us now consider a charge in uniform, rectilinear motion in a medium with electric properties which are different from those of a (classical) vacuum. Specically, consider a medium where " D Const > "0 D
0
(9.29a) (9.29b)
This implies that in this medium the phase speed is v' D c 1 Dp n " <c
0
(9.30)
R A
.t; x/ D 1 4 "0 jx q0 x0 j A.t; x/ D
Hence, in this particular medium, the speed of propagation of (the phase planes of) electromagnetic waves is less than the speed of light in vacuum, which we know is an absolute limit for the motion of anything, including particles. A medium of this kind has the interesting property that particles, entering into the medium at high speeds jv 0 j, which, of course, are below the phase speed in vacuum, can experience that the particle speeds are higher than the phase L speed in the medium. This is the basis for the Vavilov-Cerenkov radiation, more commonly known in the western literature as Cherenkov radiation, that we shall now study. If we recall the general derivation, in the vacuum case, of the retarded (and advanced) potentials in chapter 3 on page 35 and the Linard-Wiechert potentials, equations (6.103) on page 120, we realise that we obtain the latter in the medium by a simple formal replacement c ! c=n in the expression (6.104) on page 120 for s. Hence, the Linard-Wiechert potentials in a medium characterized by a refractive index n, are
FT
n .x 1 q0 D x0 / v 0 4 "0 s c 1 q0v 0 D x0 / v 0 4 "0 c 2 s c q0v 0 n .x x0 j x
0
(9.31a)
1 4 "0 c 2 jx
(9.31b)
where now s D x n .x x0 / v 0 c (9.32)
The need for the absolute value of the expression for s is obvious in the case when v 0 =c 1=n because then the second term can be larger than the rst term; if v 0 =c 1=n we recover the well-known vacuum case but with modied phase speed. We also note that the retarded and advanced times in the medium are [cf.
j 189
x.t /
c x .t /
0 0
q0
Figure 9.1: Instantaneous picture of the expanding eld spheres from a point charge moving with constant speed v 0 =c > 1=n in a medium where n > 1. This generates L a Vavilov-Cerenkov shock wave in the form of a cone.
equation (3.48) on page 44] 0 0 tret D tret .t; x 0 0 tadv D tadv .t; x x0 / D t
k jx x0 j jx x0 j n Dt ! c k jx x0 j jx x0 j n 0 x /DtC DtC ! c
so that the usual time interval t t 0 between the time measured at the point of observation and the retarded time in a medium becomes
For v 0 =c 1=n, the retarded distance s, and therefore the denominators in equations (9.31) on the preceding page, vanish when n.x x0 / v0 D x c nv 0 x0 cos c D x c x0 (9.35)
or, equivalently, when
c (9.36) nv 0 In the direction dened by this angle c , the potentials become singular. During the time interval t t 0 given by expression (9.34) above, the eld exists within a sphere of radius jx x0 j around the particle while the particle moves a distance cos c D l 0 D .t t 0 /v 0 (9.37)
along the direction of v 0 . In the direction c where the potentials are singular, all eld spheres are tangent to a straight cone with its apex at the instantaneous position of the particle and with the apex half angle c dened according to c (9.38) sin c D cos c D nv 0
R A
t0 D
jx
x0 j n c
FT
(9.33a) (9.33b) (9.34)
190 j
The rst systematic exploration of this radiation was made in 1934 by PAV E L A L E K S E E V I C H L C E R E N KOV (19041990), who was then a doctoral student in S E R G E Y I VA N OV I C H VAVI L OV s (18911951) research group at the Lebedev Physical Institute in Moscow. Vavilov wrote a manuscript with the experimental L ndings, put Cerenkov as the author, and submitted it to Nature. In the manuscript, Vavilov explained the results in terms of radioactive particles creating Compton electrons which gave rise to the radiation. This was indeed the correct interpretation, but the paper was rejected. The paper was then sent to Physical Review and was, after some controversy with the American editors, who claimed the results to be wrong, eventually published in 1937. In the same year, I G O R E V G E N E V I C H TA M M (18951975) and I LYA M I K H A I L OV I C H F R A N K (1908 1990) published the theory for the effect (the singing electron). In fact, predictions of a similar effect had been made as early as 1888 by O L I V E R H E AV I S I D E (18501925), and by A R N O L D JOHANNES WILHELM SOMM E R F E L D (18681951) in his 1904 paper Radiating body moving with velocity of light. On 8 May, 1937, Sommerfeld sent a letter to Tamm via Austria, saying that he was surprised that his old 1904 ideas were now becoming interesting. Tamm, Frank and L Cerenkov received the Nobel Prize in 1958 for the discovery and L the interpretation of the Cerenkov effect [V I TA L I Y L A Z A R E V I C H G I N Z B U R G (19162009), private communication]. The VavilovL Cerenkov cone is similar in nature to the Mach cone in acoustics.
This is illustrated in gure 9.1 on the previous page. The cone of potential singularities and eld sphere circumferences propagates with speed c=n in the form of a shock front. The rst observation of this type of radiation was reported by M A R I E S K L O D OW S K A C U R I E in 1910, but she never pursued the exploration of it. This radiation in question is therefore L called Vavilov-Cerenkov radiation.7 In order to make some quantitative estimates of this radiation, we note that we can describe the motion of each charged particle q 0 as a current density: j D q 0 v 0 .x0 v 0 t 0 / D q 0 v 0 .x 0 v 0 t 0 /.y 0 /.z 0 /O 1 x (9.39)
which has the trivial Fourier transform
This Fourier component can be used in the formul derived for a linear current in subsection on page 111 if only we make the replacements "0 ! " D n2 "0 n! k! c (9.41a) (9.41b)
R A
In this manner, using j! from equation (9.40) above, the resulting Fourier transL forms of the Vavilov-Cerenkov magnetic and electric radiation elds can be calculated from the expressions (5.22) on page 88) and (5.18) on page 86, respectively. The total energy content is then obtained from equation (6.8) on page 98 (integrated over a closed sphere at large distances). For a Fourier component one obtains [cf. equation (6.11) on page 98] Z 2 1 rad 3 0 ik x0 d x .j! k/e U! d 0 d 4 "0 nc V (9.42) Z 2 h i q 0 2 n! 2 1 0 ! 0 2 D exp ix k cos dx sin d 16 3 "0 c 3 1 v0
O1 where is the angle between the direction of motion, x0 , and the direction to the O The integral in (9.42) is singular of a Dirac delta type. If we limit observer, k. the spatial extent of the motion of the particle to the closed interval X; X on the x 0 axis we can evaluate the integral to obtain h i 0 2 02 2 sin2 1 nv cos X! 0 c v q n! sin rad U! d D d (9.43) nv 0 ! 2 4 3 "0 c 3 1 cos 0
c v
which has a maximum in the direction c as expected. The magnitude of this maximum grows and its width narrows as X ! 1. The integration of (9.43)
FT
j! D
q 0 i!x 0 =v0 e .y 0 /.z 0 /O 1 x 2
(9.40)
j 191
over therefore picks up the main contributions from c . Consequently, 2 2 we can set sin sin c and the result of the integration is Q rad U! D 2
rad U! . / sin d D dcos D cD2 0 i h Z 1 sin2 1 C nv0 X! 2 2 02 c v0 q n! sin c h i2 d 2" c3 0 2 ! 0 1 1 C nv c v0
1 1 rad U! . / d
(9.44)
sin c
1
sin2 h
1C
nv 0 c
! v0
X! v0
i d
1C
nv 0 c
i2
cX D !n
leading to the nal approximate result for the total energy loss in the frequency interval .!; ! C d!/ q02X Q rad U! d! D 2 "0 c 2 1 c2 n2 v 02 ! d! (9.46)
As mentioned earlier, the refractive index is usually frequency dependent. Realising this, we nd that the radiation energy per frequency unit and per unit length is
R A
Q rad U! d! q02! D 1 2X 4 "0 c 2 c2 n2 .!/v 02
This result was derived under the assumption that v 0 =c > 1=n.!/, i.e., under the condition that the expression inside the parentheses in the right hand side is positive. For all media it is true that n.!/ ! 1 when ! ! 1, so there exist L always a highest frequency for which we can obtain Vavilov-Cerenkov radiation from a fast charge in a medium. Our derivation above for a xed value of n is valid for each individual Fourier component.
FT
c2 n2 v 02 1 cX !n Z
1 1
The integrand in (9.44) is strongly peaked near D c=.nv 0 /, or, equivalently, near cos c D c=.nv 0 /. This means that the integrand function is practically zero outside the integration interval 2 1; 1. Consequently, one may extend the integration interval to . 1; 1/ without introducing too much an error. Via yet another variable substitution we can therefore approximate sin2 x dx x2
c2 n2 v 02
(9.45)
d!
(9.47)
192 j
9.4 Electromagnetic waves in a medium

In section 2.3 on page 27 in chapter 2 on page 21 we derived the wave equations for the electric and magnetic elds, E and B, respectively, 1 @2 E c 2 @t 2 1 @2 B c 2 @t 2 r 2E D r 2B D r "0
0r 0
@j @t
(9.48a) (9.48b)
R A D
r 2E .x/ .x/ @E @t ".x/ .x/
where the charge density and the current density j were viewed as the sources of the wave elds. As we recall, these wave equations were derived from the Maxwell-Lorentz equations (2.1) on page 21, taken as an axiomatic foundation, or postulates, of electromagnetic theory. As such, these equations just state what relations exist between (the second order derivatives of) the elds, i.e., essentially the Coulomb and Ampre forces, and the dynamics of the charges (charge and current densities) in the region under study. Even if the and j terms in the Maxwell-Lorentz equations are often referred to as the source terms, they can equally well be viewed as terms that describe the impact on matter in a particular region upon which an electromagnetic wave, produced in another region with its own charges and currents, impinges. In order to do so, one needs to nd the constitutive relations that describe how charge and current densities are induced by the impinging elds. Then one can solve the wave equations (9.48) above. In general, this is a formidable task, and one must often resort to numerical methods. Let us, for simplicity, assume that the linear relations, as given by formula (9.8) on page 183 and formula (9.15) on page 185, hold, and that there is also a linear relation between the electric eld E and the current density, known as Ohms law:
FT
j.t; x/ D .t; x/E.t; x/ @2 E D .r @t 2 E/ r ln .x/ r r ln ".x/ E C r
(9.49)
where is the conductivity of the medium. Let us make the further assumption that " D ".x/, D .x/, and .x/ are not explicitly dependent on time and are local in space. Then we obtain the coupled wave equations
(9.50a)
" @H @2 H r 2H .x/ .x/ ".x/ .x/ 2 D .r H/ r ln ".x/ @t @t r r ln .x/ H C r ln ".x/ . .x/E/ r .x/ E
(9.50b)
9.4. Electromagnetic waves in a medium
j 193
For the case D 0 (no magnetisation) and tions (9.50) on the facing page simplify to r 2E r 2B
0
D Const in the medium, equa-
@E @t @B .x/ @t C .x/
@2 E D r r ln ".x/ E @t 2 @2 B ".x/ 0 2 D .r B/ r ln ".x/ @t . E/ E 0 r ln ".x/ 0 r .x/ ".x/

0
(9.51a)
(9.51b)
Making the further assumption that the medium is not conductive, i.e., that D 0, the uncoupled wave equations r 2E r 2B are obtained. ".x/ @2 E D r fr ln ".x/ Eg "0 c 2 @t 2 ".x/ @2 B D .r B/ r ln ".x/ "0 c 2 @t 2 (9.52a) (9.52b)
9.4.1 Constitutive relations
In a solid, uid or gaseous medium the source terms in the microscopic Maxwell equations (2.1) on page 21 must include all charges and currents in the medium, i.e., also the intrinsic ones (e.g., the polarisation charges in electrets, and atomistic magnetisation currents in magnets) and the self-consistently imposed ones (e.g., polarisation currents). This is of course also true for the inhomogeneous wave equations derived from Maxwells equations. From now one we assume that and j represent only the charge and current densities (i.e., polarisation and conduction charges and currents, respectively) that are induces by the E and B elds of the waves impinging upon the medium of interest.8 Let us for simplicity consider a medium containing free electrons only and which is not penetrated by a magnetic eld, i.e., the medium is assumed to be isotropic with no preferred direction(s) in space.9 Each of these electrons are assumed to be accelerated by the Lorentz force, formula (4.61) on page 68. However, if the elds are those of an electromagnetic wave one can, for reasonably high oscillation frequencies, neglect the force from the magnetic eld. Of course, this is also true if there is no magnetic eld present. So the equation of motion for each electron in the medium can be written
R A
m d2 x dx Cm D qE 2 dt dt
FT
8
If one includes also the effect of the charges on E and B, i.e., treat and j as sources for elds, singularities will appear in the theory. Such so called self-force effects will not be treated here.
9
To this category belongs unmagnetised plasma. So do also, to a good approximation, uid or solid metals.
(9.53)
194 j
where m and q are the mass and charge of the electron, respectively, the effective collision frequency representing the frictional dissipative force from the surrounding medium, and E the effective applied electric eld sensed by the electron. For a Fourier component of the electric eld E D E0 exp . i!t/, the equation of motion becomes ! 2 qx.t/ i! qx.t/ D q2 E m (9.54)
If the electron is at equilibrium x D 0 when E D 0, then its dipole moment is d.t / D qx.t /. Inserting this in equation (9.54) above, we obtain dD m.! 2 q2 E C i! / (9.55)
This is the the lowest order contribution to the dipole moment of the medium from each electron under the inuence of the assumed electric eld. If Nd .x/ electrons per unit volume can be assumed to give rise to the electric polarisation P , this becomes P D Nd d D Nd q 2 E m.! 2 C i! / (9.56)
Using this in formula (9.5) on page 182, one nds that
FT
D.t; x/ D ".x/E.t; x/ Nd .x/q 2 1 2 C i! "0 m ! s !p .x/ D Nd .x/q 2 "0 m
2 !p
R A
where ".x/ D "0 The quantity
(9.57)
Nd .x/q 2 D "0 1 m.! 2 C i! /
(9.58)
(9.59)
is called the plasma frequency and s s ".x/ n.x/ D D 1 "0
! 2 C i!
(9.60)
is called the refractive index. At points in the medium where the wave frequency ! equals this plasma frequency and the collision frequency vanishes, the refractive index n D 0, and the wave is totally reected. In the ionised outer part of the atmosphere called the ionosphere this happens for radio waves of frequencies up to about 10 MHz. This is the basis for over-the-horizon radio communications.
j 195
9.4.2 Electromagnetic waves in a conducting medium

We shall now restrict ourselves to the wave equations for the electric eld vector E and the magnetic eld vector B in a electrically conductive and neutral medium, i.e., a volume where there exist no net electric charge, D 0, no dielectric effects, " D "0 , and no electromotive force, Eemf D 0. A highly conductive metal is a good example of such a medium.
9.4.2.1 The wave equations for E and B

To a good approximation, metals and other conductors in free space have a conductivity that is not dependent on t or x. The wave equations (9.51) on page 193 are then simplied to r 2E r 2B
0
@E @t @B @t
1 @2 E D0 c 2 @t 2 1 @2 B D0 c 2 @t 2
D
In the limit of long
Multiplying by e i!t and introducing the relaxation time D "0 = of the medium in question, we see that the differential equation for each spectral component can be written !2 i 2 r E.t; x/ C 2 1 C E.t; x/ D 0 (9.64) c ! (low conductivity ), (9.64) tends to r 2E C !2 ED0 c2 (9.65)
R A
which are the homogeneous vector wave equations for E and B in a conducting medium without EMF. We notice that for the simple propagation media considered here, the wave equation for the magnetic eld B has exactly the same mathematical form as the wave equation for the electric eld E, equation (9.61) above. Therefore, in this case it sufces to consider only the E eld, since the results for the B eld follow trivially. For EM waves propagating in more complicated media, containing, e.g., inhomogeneities, the wave equations for E and for B do not have the same mathematical form. Following the spectral component prescription leading to equation (2.39) on page 30, we obtain, in the special case under consideration, the following timeindependent wave equation !2 2 r E0 C 2 1 C i E0 D 0 (9.63) c "0 !
FT
(9.61) (9.62)
196 j
which is a time-independent wave equation for E, representing undamped propagating waves. In the short (high conductivity ) limit we have instead r 2 E C i!
0
ED0
(9.66)
where in the last step we used the characteristic impedance of vacuum dened according to formula (6.3) on page 97.
R A
O r Dn
9.4.2.2 Plane waves
Consider now the case where all elds depend only on the distance O plane with unit normal n. Then the del operator becomes @ O D nr @
FT
C "0
0
which is a time-independent diffusion equation for E. For most metals 10 14 s, which means that the diffusion picture is good for all frequencies lower than optical frequencies. Hence, in metallic conductors, the propagation term @2 E=c 2 @t 2 is negligible even for VHF, UHF, and SHF signals. Alternatively, we may say that the displacement current "0 @E=@t is negligible relative to the conduction current j D E. If we introduce the vacuum wave number ! kD (9.67) c p we can write, using the fact that c D 1= "0 0 according to equation (1.13) on page 6, r 1 1 0 D D D D R0 (9.68) ! "0 ! "0 ck k "0 k
to a given
(9.69)
and the microscopic Maxwell equations attain the form @E @ @E O n @ @B O n @ @B O n @ O n D0 D (9.70a) (9.70b) (9.70c)
0 j.t; x/
@B @t
D0 D @E D @t
0
E C "0
@E @t
(9.70d)
O Scalar multiplying (9.70d) by n, we nd that @B O O O 0Dn n Dn 0 C "0 @
@ 0 @t
E (9.71)
j 197
which simplies to the rst-order ordinary differential equation for the normal component En of the electric eld dEn C En D 0 dt "0 with the solution En D En0 e
t ="0
(9.72)
D En0 e
t=
(9.73)
or O n @B D0 @t
9.4.2.3 Telegraphers equation
In analogy with equation (9.61) on page 195, we can easily derive a wave equation @2 E @ 2
0
R A
@E @t 1 @2 E D0 c 2 @t 2 @2 E @ 2 1 @2 E D0 c 2 @t 2 Ei D f . ct / C g. C ct/;
From this, and (9.70c), we conclude that the only longitudinal component of B must be constant in both time and space. In other words, the only non-static solution must consist of transverse components.
describing the propagation of plane waves along in a conducting medium. This equation is is called the telegraphers equation. If the medium is an insulator so that D 0, then the equation takes the form of the one-dimensional wave equation (9.77)
As is well known, each component of this equation has a solution which can be written i D 1; 2; 3 (9.78)
FT
(9.75) (9.76)
This, together with (9.70a), shows that the longitudinal component of E, i.e., the component which is perpendicular to the plane surface is independent of and has a time dependence which exhibits an exponential decay, with a decrement given by the relaxation time in the medium. O Scalar multiplying (9.70b) by n, we similarly nd that @E @B O O O 0Dn n D n (9.74) @ @t
198 j
where f and g are arbitrary (non-pathological) functions of their respective arguments. This general solution represents perturbations which propagate along , where the f perturbation propagates in the positive direction and the g perturbation propagates in the negative direction. In a medium, the general solution to each component of equation (9.99) on page 200 is given by Ei D f . v' t/ C g. C v' t/; i D 1; 2; 3 (9.79)
If we assume that our electromagnetic elds E and B are represented by a Fourier component proportional to exp. i!t/, the solution of equation (9.77) on the preceding page becomes E D E0 e By introducing the wave vector
i.!tk /
D E0 ei.
!t/
(9.80)
FT
O k D kn D ! !O O nD k c c E D E0 ei.k x
!t/
(9.81)
this solution can be written as
(9.82)
R A
O n or, solving for B, BD k O n ! ED 1 k !
Let us consider the lower sign in front of k in the exponent in (9.80). This corresponds to a wave which propagates in the direction of increasing . Inserting this solution into equation (9.70b) on page 196, gives @E O D i!B D ik n @ E (9.83)
ED
1 O k c
ED
p "0
O n
(9.84)
Hence, to each transverse component of E, there exists an associated magnetic eld given by equation (9.84) above. If E and/or B has a direction in space which is constant in time, we have a plane wave. Allowing now for a nite conductivity in our medium, and making the spectral component Ansatz in equation (9.76) on the previous page, we nd that the time-independent telegraphers equation can be written @2 E C "0 @ 2
2 0! E C i 0
!E D
@2 E C K 2E D 0 @ 2
(9.85)
where K 2 D "0 !2 1 C i 0 "0 ! D !2 c2 1Ci "0 ! D k2 1 C i "0 ! (9.86)
j 199
where, in the last step, equation (9.67) on page 196 was used to introduce the wave number k. Taking the square root of this expression, we obtain r D C i (9.87) K Dk 1Ci "0 ! Squaring, one nds that k2 1 C i or 2 D 2 D k 2"0 !
2
"0 !
D . 2
2 / C 2i
(9.88)
k2
(9.89) (9.90)
Squaring the latter and combining with the former, one obtains the second order algebraic equation (in 2 ) 2 . 2 k2/ D k4 2 4"2 ! 2 0 (9.91)
R A
E D E0 e
which can be easily solved and one nds that vr u 2 u u 1C C1 t "0 ! Dk 2 vr u 2 u u 1C 1 t "0 ! Dk 2
As a consequence, the solution of the time-independent telegraphers equation, equation (9.85) on the preceding page, can be written
D
BD 1 O Kk !
ei.
With the aid of equation (9.84) on the facing page we can calculate the associated magnetic eld, and nd that it is given by ED 1 O .k ! E/. C i/ D 1 O .k ! E/ jAj ei (9.94)
where we have, in the last step, rewritten C i in the amplitude-phase form jAj exp.i /. From the above, we immediately see that E, and consequently also B, is damped, and that E and B in the wave are out of phase.
FT
(9.92a) (9.92b)
!t/
(9.93)
200 j
In the limit "0 ! K Dk 1Ci p D "0

1 2
, we can approximate K as follows: r k.1 C i/ 2"0 ! (9.95)
1 "0 ! 2 1 i Dk i "0 ! "0 ! r r 0 ! D .1 C i/ 0 !.1 C i/ 2"0 ! 2
Hence, both elds fall off by a factor 1=e at a distance s 2 D 0 !
FT
@2 E @ 2 @E @t " @2 E D0 @t 2 @2 E @ 2 " @2 E D0 @t 2 E D E0 ei.k x
!t/
In this limit we nd that when the wave impinges perpendicularly upon the medium, the elds are given, inside the medium, by r r 0 ! 0 ! 0 E D E0 exp exp i !t (9.96a) 2 2 r 0 O .n E 0 / (9.96b) B0 D .1 C i/ 2!
(9.97)
R A D
This distance is called the skin depth . Assuming for simplicity that the electric permittivity " and the magnetic permeability , and hence the relative permittivity e and the relative permeability m all have xed values, independent on time and space, for each type of material we consider, we can derive the general telegraphers equation [cf. equation (9.76) on page 197] (9.98)
describing (1D) wave propagation in a material medium. In chapter 2 on page 21 we concluded that the existence of a nite conductivity, manifesting itself in a collisional interaction between the charge carriers, causes the waves to decay exponentially with time and space. Let us therefore assume that in our medium D 0 so that the wave equation simplies to (9.99)
As in the vacuum case discussed in chapter 2 on page 21, assuming that E is time-harmonic, i.e., can be represented by a Fourier component proportional to exp. i!t /, the solution of equation (9.99) can be written (9.100)
j 201
where now k is the wave vector in the medium given by equation (9.20) on page 186. With these denitions, the vacuum formula for the associated magnetic eld, equation (9.84) on page 198, BD p O " k ED 1 O k v' ED 1 k ! E (9.101)
is valid also in a material medium (assuming, as mentioned, that n has a xed constant scalar value). A consequence of a e 1 is that the electric eld will, in general, have a longitudinal component.
Electromagnetic waves in an electrically and magnetically conducting medium Derive the wave equation for the E eld described by the electromagnetodynamic equations (Diracs symmetrised Maxwell equations) [cf. equations (2.2) on page 22]
e
E X A M P L E 9.1
"0 @B r ED @t r BD 0 m r B D "0
0
under the assumption of vanishing net electric and magnetic charge densities and in the absence of electromotive and magnetomotive forces. Interpret this equation physically.
Assume, for symmetry reasons, that there exists a linear relation between the magnetic current density j m and the magnetic eld B (the magnetic dual of Ohms law for electric currents, j e D e E) jm D
m
R
.r E/ D D
0r
Taking the curl of (2.2c) and using (2.2d), one nds, noting that "0 @ .r B/ @t @ 1 @E m e r B 0j C 2 @t c @t 1 @E m e @E e 0 EC 2 0 @t c @t jm
Using the vector operator identity r .r E/ D r .r E/ r 2 E, and the fact that r E D 0 for a vanishing net electric charge, we can rewrite the wave equation as m @E 1 @2 E e 2 m e ED0 (9.105) r2E C 2 0 0 @t c c 2 @t 2 This is the homogeneous electromagnetodynamic wave equation for E that we were after. Compared to the ordinary electrodynamic wave equation for E, equation (9.61) on page 195, we see that we pick up extra terms. In order to understand what these extra terms mean physically, we analyse the time-independent wave equation for a single Fourier
A FT
r ED (9.102a)
0j m
(9.102b) (9.102c)
@E C @t
0j
(9.102d)
(9.103)
D 1=c 2 , that
(9.104) 1 @2 E c 2 @t 2
202 j
component. Then our wave equation becomes r E C i!

2 2
!2 Dr EC 2 c
C 2 c " 1
EC 1 0 ! 2 "0
!2 E c2
m e
2 m e E 0
Ci
# C m =c 2 ED0 "0 !
(9.106)
Realising that, according to formula (6.3) on page 97, 0 ="0 is the square of the vacuum radiation resistance R0 , and rearranging a bit, we obtain the time-independent wave equation in Diracs symmetrised electrodynamics 0 r2E C !2 c2 1
2 R0 2 !
1
eC "0 ! 1 m =c 2
2 R0 !2
!
m e
B B1 C i @
From this equation we conclude that the existence of magnetic charges (magnetic monopoles), and non-vanishing electric and magnetic conductivities would lead to a shift in the effective wave number of the wave. Furthermore, even if the electric conductivity e vanishes, the imaginary term does not necessarily vanish and the wave therefore experiences damping or growth according as m is positive or negative, respectively. This would happen in a hypothetical medium which is a perfect insulator for electric currents but which can carry magnetic currents. p m e def ! , the time-independent Finally, we note that in the particular case ! D R0 m wave equation equation (9.106) above becomes a time-independent diffusion equation r 2 E C i!
e m
A FT
! R0 p
m e 0
m e
C CE D 0; A
(9.107)
c2
ED0
(9.108)
R D
which in time domain corresponds to the time-dependent diffusion equation @E @t Dr 2 E D0 (9.109)
with a diffusion coefcient given by DD

0
1
e
m c2
(9.110)
Hence, electromagnetic waves with this particular frequency do not propagate. This means that if magnetic charges (monopoles) exist in a region in the Universe, electromagnetic waves propagating through this region would, in this simplistic model, exhibit a lower cutoff at ! D !m . This would in fact impose a lower limit on the mass of the photon, the quantum of the electromagnetic eld that we shall come across later. End of example 9.1
9.5. Bibliography
j 203
9.5 Bibliography
[55] E. H ALLN, Electromagnetic Theory, Chapman & Hall, Ltd., London, 1962. [56] J. D. JACKSON, Classical Electrodynamics, third ed., John Wiley & Sons, Inc., New York, NY . . . , 1999, ISBN 0-471-30932-X. [57] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0201-05702-6. [58] J. A. S TRATTON, Electromagnetic Theory, McGraw-Hill Book Company, Inc., New York, NY and London, 1953, ISBN 07-062150-0.
R A
FT
R A
FT
FORMUL
This appendix is a compilation of the most important formul derived in the various chapters. It also contains an extensive collection of vector and tensor algebra and calculus formul, including a few that are perhaps not included in every electrodynamics textbook.
F .1
The electromagnetic eld
F .1.1
Microscopic Maxwell-Lorentz equations in Diracs symmetrised form

e
R A
r ED "0 r BD EC
0 m
@B D @t 1 @E D c 2 @t
F .1.1.1
Constitutive relations
1 cDp "0
D
r
0
.D 299 792 458 ms

0
"0
D R0 .D 119:9169832 D H j P "E
1
E "0 E 205
FT
(F.1) (F.2) (F.3) (F.4)
0j m 0j e 1
(F.5) (F.6) (F.7) (F.8) (F.9) (F.10)
376:7 /
206 j
6. FORMUL
F .1.2
F .1.2.1
Fields and potentials

Vector and scalar potentials
BDr ED r A @A @t (F.11) (F.12)
F .1.2.2
The velocity gauge condition in free space

c2 D 2; v ( D 1 ) Lorenz-Lorentz gauge D 0 ) Coulomb gauge (F.13)
F .1.3
F .1.3.1
Energy and momentum
Electromagnetic eld energy density in free space

ueld D 1 "0 .E E C c 2 B B/ 2
FT
SD 1
0
@ r AC 2 D 0; c @t
(F.14)
R A
F .1.3.2 F .1.3.3
Poynting vector in free space

E B (F.15)
Linear momentum density in free space

g D "0 E B (F.16)
F .1.3.4
Linear momentum ux tensor in free space

T D ueld 13
"0 EE C c 2 BB "0 Ei Ej "0 c Bi Bj

2
(F.17) (F.18)
Tij D u
eld
ij
F .1.3.5
Angular momentum density around x0 in free space

h D .x x0 / g D .x x0 / "0 E B (F.19)
F .2.
Electromagnetic radiation
j 207
F .1.3.6
Angular momentum ux tensor around x0 in free space

K.x0 / D .x
x0 /
(F.20)
F .2
Electromagnetic radiation
The far elds from an extended source distribution
i k eikjx x0 j 4 "0 c jx x0 j i
far B! .x/
F .2.1
Efar .x/ !
d3x 0
k e 4 "0 c jx
k eikjx x0 j 4 "0 c 2 jx x0 j i
x0 j Z
.I ! .x0 /
k e O I ! .x0 / n 4 "0 c 2 jx x0 j Z 0 O I ! .x0 / D d3x 0 j! .x0 / e ik n .x x0 /

V0
V0 ikjx x0 j
d3x 0 j! e
F .2.2
The far elds from an electric dipole

Efar .x/ D ! k 2 eikjx x0 j .d! .x0 / 4 "0 jx x0 j O k/
R A
far B! .x/ D V0
k 2 eikjx x0 j O d! .x0 / k 4 "0 c jx x0 j Z 0 d.t; x0 / D d3x 0 .x0 x0 / .tret ; x0 /
D
F .2.3
The far elds from a magnetic dipole

Efar .x/ D ! k 2 eikjx x0 j m! .x0 / 4 "0 c jx x0 j O k (F.29) (F.30) (F.31)
far B! .x/ D
k 2 eikjx x0 j O O m! .x0 / k k 4 "0 c 2 jx x0 j Z 1 0 m.t; x0 / D d3x 0 .x0 x0 / j.tret ; x0 / 2 V0
FT
O n/ O n (F.22)
ik .x0 x0 /
V0 ikjx x0 j
j! e
ik .x0 x0 /
O k
O k
(F.21)
O k
(F.23)
(F.24)
(F.25)
O k
(F.26) (F.27) (F.28)
208 j
6. FORMUL
F .2.4
The far elds from an electric quadrupole
Efar .x/ D i !
k 3 eikjx x0 j 8 "0 jx x0 j
O k Q! .x0 /
O k
O k
(F.32) (F.33) (F.34)
k 3 eikjx x0 j O far O B! .x/ D i k Q! .x0 / k 8 "0 c jx x0 j Z 0 Q.t; x0 / D d3x 0 .x0 x0 /.x0 x0 / .tret ; x0 /
V0
F .2.5
Relationship between the eld vectors in a plane wave
FT
BD 1O k c E x0 / 1 v 02 c2 C .x
0
(F.35)
R A
F .2.6
E.t; x/ D q .x 4 "0 s 3 B.t; x/ D
The elds from a point charge in arbitrary motion
x0 /
.x
x0 / c2
a0 (F.36) (F.37) (F.38) (F.39) (F.40) (F.41) (F.42)
x E.t; x/ x0 j dx0 .t 0 / v 0 .t 0 / D dt 0 dv 0 .t 0 / a0 .t 0 / D dt 0 v0 s D x x0 .x x0 / c v0 x x0 D .x x0 / x x0 c 0 @t jx x0 j D @t x s
1 x c jx
F .3.
Special relativity
j 209
F .3
Special relativity
Metric tensor for at 4D space
0 1 B0 B DB @0 0 0 1 0 0 0 0 1 0 1 0 0C C C 0A 1
F .3.1
(F.43)
F .3.2
Covariant and contravariant four-vectors

a D a
F .3.3
Lorentz transformation of a four-vector

x0 D x 0 B B DB @ 0 0 0 0 1 Dp 1 2 v D c
R A
ds D c dt Dcd u D dx d D .c; v / p D m0 u D E ;p c
F .3.4
Invariant line element
D
F .3.5 F .3.6
Four-velocity
(F.50)
Four-momentum
(F.51)
FT
(F.44) (F.45) 0 0 1 0 0 0C C C 0A 1 1 (F.46) (F.47) (F.48) (F.49)
210 j
6. FORMUL
F .3.7
Four-current density
j D
0u
(F.52)
F .3.8
Four-potential
A D ;A c (F.53)
F .3.9
Field tensor
FT
F D@ A @ A 0 Ex =c 0 Bz By Ey =c Bz 0 Bx 1 Ez =c By C C C Bx A 0
3 X i D1
(F.54)
0 BE =c B x DB @Ey =c Ez =c
(F.55)
R A
F .4
r O xi
Vectors and tensor elds in 3D
Let x be the position vector (radius vector, coordinate vector) from the origin of the Euclidean space R3 coordinate system to the coordinate point .x1 ; x2 ; x3 / in the same system and let jxj denote the magnitude (length) of x. Let further .x/; .x/; : : :, be arbitrary scalar elds, a.x/; b.x/; : : : , arbitrary vector elds, and A.x/; B.x/; : : : , arbitrary rank two tensor elds in this space. Let denote complex conjugate and denote Hermitian conjugate (transposition and, where applicable, complex conjugation). The differential vector operator r is in Cartesian coordinates given by @ def @ def O xi @ @xi @xi (F.56)
O O O O O O O where xi , i D 1; 2; 3 is the ith unit vector and x1 x, x2 y, and x3 z. In component (tensor) notation r can be written @ @ @ @ @ @ ri D @i D ; ; D ; ; (F.57) @x1 @x2 @x3 @x @y @z
F .4.
j 211
The differential vector operator r 0 is dened as r0 or ri0 D @0 i D @ @ @ 0 ; 0 ; 0 @x1 @x2 @x3 D @ @ @ ; ; 0 @y 0 @z 0 @x (F.59)
3 X iD1
O xi
@ def @ def O xi 0 @ 0 @xi0 @xi
(F.58)
F .4.1
F .4.1.1
Cylindrical circular coordinates

Base vectors
O O O D cos x1 C sin x2 O D O O z D x3 C YLINDRICAL CIRCULAR TO C ARTESIAN O O x1 D cos O O x3 D z
O O sin x1 C cos x2
O sin
R A
O O O x2 D sin C cos O O O O dl D dx x D d C d' C dz z dV D d3x D d d' dz
F .4.1.2
Directed line element
F .4.1.3
Directed area element
D
F .4.1.4 F .4.1.5
O O O dS D d' dz C d dz C d d' z
Volume element
(F.64)
Spatial differential operators
In the following we assume that the scalar eld D . ; '; z/ and that the vector eld a D a. ; '; z/.
FT
(F.60) (F.61) (F.62) (F.63)
C ARTESIAN TO CYLINDRICAL CIRCULAR
212 j
6. FORMUL
T HE GRADIENT O r D T HE DIVERGENCE r aD T HE CURL 1 @. a / 1 @a' @az C C @ @' @z @ O 1 @ C z @ O C @ @' @z
(F.65)
(F.66)
FT
O aD 1 @ @ @ @ C 1 @2 @2 C 2 2 @' 2 @z O sin x3 O sin ' O sin
1 @az @a' @' @z @a @az O C @z @ 1 @. a' / @a O Cz @ @'
(F.67)
T HE L APLACIAN
R A
r 2 D
(F.68)
F .4.2
Spherical polar coordinates

Base vectors
F .4.2.1
C ARTESIAN TO SPHERICAL POLAR
O O O O r D sin cos ' x1 C sin sin ' x2 C cos x3 (F.69) O O sin ' x1 C cos ' x2
O O O D cos cos ' x1 C cos sin ' x2 O D
S PHERICAL POLAR TO C ARTESIAN O O O x1 D sin cos ' r C cos cos ' O O x3 D cos r
O O O O x2 D sin sin ' r C cos sin ' C cos '
(F.70)
F .4.
j 213
F .4.2.2
Directed line element

O O O O dl D dx x D dr r C r d C r sin d' (F.71)
F .4.2.3
Solid angle element

d D sin d d' (F.72)
F .4.2.4
Directed area element
F .4.2.5
Volume element
dV D d3x D dr r 2 d
F .4.2.6
Spatial differential operators
T HE GRADIENT
R A
O r D r @ O 1 @ C 1 @ O C @r r @ r sin @' r O aDr
In the following we assume that the scalar eld D .r; ; '/ and that the vector eld a D a.r; ; '/.
T HE DIVERGENCE
r aD T HE CURL
1 @.r 2 ar / 1 @.a sin / 1 @a' C C 2 r @r r sin @ r sin @'
1 @.a' sin / @a r sin @ @' 1 @ar @.ra' / O1 C r sin @' @r 1 @.ra / @ar O C r @r @
FT
(F.74) (F.75) (F.76) (F.77)
O O O dS D r 2 d r C r sin dr d' C r dr d
(F.73)
214 j
6. FORMUL
T HE L APLACIAN r 2 D 1 @ @ 1 @ @ 1 @2 r2 C 2 sin C 2 @' 2 2 @r r @r r sin @ @ r 2 sin
(F.78)
F .4.3
F .4.3.1
Vector and tensor eld formul

The three-dimensional unit tensor of rank two
O O 13 D xi xi (F.79)
with matrix representation, denoted by .: : : /, 0
FT
1 1 0 0 B C .13 / D @0 1 0A 0 0 1 ( ij D 0 if i j if i D j 1 .ij / D .13 / if i; j; k is an even permutation of 1,2,3 if at least two of i; j; k are equal if i; j; k is an odd permutation of 1,2,3
ij k ij k
(F.80)
F .4.3.2
The 3D Kronecker delta tensor
R A
F .4.3.3
(F.81)
has the same matrix representation as 13 : (F.82)
The fully antisymmetric Levi-Civita tensor

8 1 < D 0 : 1
ij k
(F.83)
has the properties D

j ki j ik
D D
kij ikj
(F.84) (F.85) (F.86)
ij k ilm
D j l km
j m kl
F .4.
j 215
F .4.3.4
Rotational matrices
The rotational matrix vector O S D Si xi D i. x ij k /O i (F.87)
F .4.3.5
General vector and tensor algebra identities

a b D b a D ij ai bj D ab cos a aDa D
2
R A
bD b aD O ij k xi aj bk b/ D
2
O j ki xj ak bi
2
.a
ia Sb D
.a b/ C .a
b/ D a b
O O A D xi Aij xj
O O A D xj Aij xi Tr A D Ai i
O O Aij D xi A xj O O ab D xi ai bj xj Tr ab D a b
a.b C c/ D ab C ac .a C b/c D ac C bc c ab D .c a/b ab c D a.b c/ c ab ab D .c c D a.b a/b c/
ab cd D .b c/ad
FT
3j k /
has the matrices Si as components, where 0 1 0 0 0 B C S1 D @0 0 iA D i. 0 i 0 0 1 0 0 i B C S2 D @ 0 0 0A D i. i 0 0 0 1 0 i 0 B C S3 D @ i 0 0A D i. 0 0 0
1j k /
(F.88a)
2j k /
(F.88b)
(F.88c)
(F.89)
(F.90) (F.91)
O kij xk ai bj
iaS b
2 2
(F.92) (F.93) (F.94) (F.95) (F.96) (F.97) (F.98) (F.99) (F.100) (F.101) (F.102) (F.103) (F.104) (F.105) (F.106)
216 j
6. FORMUL
a .b a .a .a b/ .c .b b/ a .a
F .4.3.6
c/ D .a
b/ c c.a b/ a.b c/ b/ D 0 b c/d
(F.107) (F.108) (F.109) (F.110) (F.112)
c/ D ba c c D ba c .b b/ c/ C b .c .c
ca b D b.a c/ ab c D b.a c/ .c a/ C c b d/c .a
d/ D a b
d/ D .a c/.b d/ .a
.a d/.b c/ (F.111)
d/ D .a
Special vector and tensor algebra identities

13 a D a 13 D a
(F.113) (F.114) (F.115) (F.116) (F.117) (F.118) 1
.1 3
a/ D .a
R A
r a aD r D
F .4.3.7
General vector and tensor calculus identities

O r D xi @i r a D @i ai a r D ai @i O ij k xi @j ak O ij k xi aj @k
FT
0 B 13 / D @ a3 a2 a .13 .a 0 a3 0 a1 a2 C a1 A D 0 iS a a .13 b/ D a b b/ D a b/ D ba b
13
13
aDa
13
ab
(F.119) (F.120) (F.121) (F.122) (F.123) (F.124) (F.125) (F.126) (F.127) (F.128) (F.129) (F.130) (F.131) (F.132) (F.133)
O O r a D xi xj @i aj Tr r a D r a
O O ar D xi xj ai @j
D
r .a b/ D a
O r A D xj @i Aij ar b D a@i bi r r D r 2 O r r a D r .r a/ D xi @i @j aj O a r r D .a r /r D xi aj @j @i r ./ D r C r .r b/ C b .r a/ C a r b C b r a
F .4.
j 217
r .a
b/ D .r a/
.r b/
a
2
(F.134) (F.135) (F.136) (F.137) (F.138) a/ b (F.139) (F.140) (F.141) (F.142) (F.143) (F.144) b/
r .a/ D a r C r a .r / .r / D r .r / r .a .r a/ .r b/ D b .r a/ D .r a/ a/ a a .r / .r b/ D b r r r r .a r r .a r r / a .r r .r .r / .r / a rb rb
r .ab/ D .r a/b C a r b .a/ D r .r / D .r / .ab/ D .r a .r a a .a .a r .r a r .b .r .r r/ r/ a/b
b/ D ar b
br a C b r a a
r .A/ D .r /A C r a b/ D .a r/ b b/ D .r b/ a
1 a/ D 2 r .a2 /
b D .r b/ a
r D a rr
R A
c/ D .a r b/ r .r r .r cCb a b .r c/ D b .a r c/ r D 0 a/ D 0 r/ a D 0 r .13 / D r r .13 a/ D r a r .13 a/ D r a/ D r ir Sa D r xD3 xD0 a ir S a .r
a/ D r r a
r r a D r .r a/
F .4.3.8
Special vector and tensor calculus identities
In the following S is the matrix vector dened in formula (F.87) and k is an arbitrary constant vector. (F.157) (F.158) (F.159) (F.160) (F.161) (F.162)
FT
(F.145) (F.146) (F.148) (F.149) (F.150) a rb (F.147) a ra ar
2
.r b/a
r a
(F.151)
.a r c/
(F.152) (F.153) (F.154) (F.155) (F.156)
a .b r c/
218 j
6. FORMUL
r x D 13 r .x k/ D k r .x a/ D a C x.r a/ C .x r / a x r jxj D jxj x x0 r jx x0 j D D r 0 jx x0 j jx x0 j x 1 D r jxj jxj3 1 x x0 1 0 r D D r jx x0 j3 jx x0 j jx x0 j x 1 r D r2 D 4 .x/ 3 jxj jxj x x0 1 r D r2 D 4 .x x0 / 0 j3 jx x jx x0 j k 1 k x r Dk r D jxj jxj jxj3 x k x r k D r if jxj 0 3 jxj jxj3 k 1 r2 D kr 2 D 4 k.x/ jxj jxj r .k a/ D k.r a/ C k .r a/ r .k a/
(F.163) (F.164) (F.165) (F.166) (F.167) (F.168) (F.169) (F.170) (F.171) (F.172) (F.173) (F.174) (F.175)
R A
F .4.3.9
Integral identities
I NTEGRAL THEOREMS
Let V .S / be the volume bounded by the closed surface S.V /. Denote the 3D volume element by d3x. dV / and the surface element, directed along the outO O ward pointing surface normal unit vector n, by dS. d2x n/. Then Z
V
FT
d x r D Z
V 3
I dS IS dS a IS aD
S
(F.176) (F.177) (F.178) (F.179)
d3x r a D d3x r
Z
V
dS I
Z
V
d3x r A D
dS A
S
F .4.
j 219
If S.C / is an open surface bounded by the contour C.S/, whose line element is dl, then I Z dl D dS r (F.180) C S I Z dl a D dS r a (F.181)
C S
G ENERIC INTEGRAL IDENTITIES Z r

V0
d3x 0
A jx x0 j
Z D
V0
d3x 0
r0 A jx x0 j
I
S0
O d2x 0 n0
A jx x0 j
(F.182) and
Z r r
V0
d3x 0
A jx x0 j
Where A D .x0 / or A D a.x0 /. S PECIFIC INTEGRAL IDENTITIES Z rr a.x0 / dx D jx x0 j V0

3 0
Z
V0
R A
I C
S0
dx r O dx n Z
2 0
3 0
d3x 0
V0
a.x0 / jx x0 j
D rr
a.x0 / d3x 0 jx x0 j V0 Z a.x0 / r r d3x 0 jx x0 j V0 Z 1 3 0 0 0 0 D 4 a.x/ d x r a.x /r jx x0 j V0 I 0 0 a.x /.x x / O C d2x 0 n0 0 jx x0 j3 S (F.185)
H ELMHOLTZ DECOMPOSITION
Any regular, differentiable vector eld u that falls off sufciently fast asymptotically can be decomposed into two components, one irrotational and one rotational, such that u.x/ D uirrot .x/ C urotat .x/ (F.186a)
FT
4 A.x/ (F.183) a.x /r jx x0 j 0 a.x /.x x0 / jx x0 j3
0 0
where is either nothing (juxtaposition) and A D .x0 /, or D or D A D a.x0 /.
(F.184)
220 j
6. FORMUL
where u
irrot
Z .x/ D r
urotat .x/ D r
r 0 u.x0 / 4 jx x0 j V0 Z r 0 u.x0 / d3x 0 4 jx x0 j V0 d3x 0
(F.186b) (F.186c)
R A
F .5
Bibliography
[59] G. B. A RFKEN AND H. J. W EBER, Mathematical Methods for Physicists, fourth, international ed., Academic Press, Inc., San Diego, CA . . . , 1995, ISBN 0-12-0598167.
[60] P. M. M ORSE AND H. F ESHBACH, Methods of Theoretical Physics, Part I. McGrawHill Book Company, Inc., New York, NY . . . , 1953, ISBN 07-043316-8. [61] W. K. H. PANOFSKY AND M. P HILLIPS, Classical Electricity and Magnetism, second ed., Addison-Wesley Publishing Company, Inc., Reading, MA . . . , 1962, ISBN 0201-05702-6.
FT
If v is a vector eld with the same properties as u, or u itself, then Z Z Z r 0 u.x0 / 3 irrot irrot 3 (F.187) d x u .x/ v .x/ D d x r v .x/ d3x 0 4 jx x0 j V V V0 Z Z Z r 0 v .x0 / (F.188) d3x urotat .x/ v rotat .x/ D d3x r u.x/ d3x 0 4 jx x0 j V V0 ZV d3x uirrot .x/ v rotat .x/ D 0 (F.189) V Z d3x uirrot .x/ v irrot .x/ D 0 (F.190) V Z Z Z r 0 v .x0 / d3x urotat .x/ v rotat .x/ D d3x r u.x/ d3x 0 4 jx x0 j V V V0 (F.191)
MATHEMATICAL METHODS
Physics is the academic discipline that systematically studies and describes the physical world, nds new fundamental laws of Nature or generalises existing ones that govern Natures behaviour under various conditions, andperhaps most important of allmakes new predictions about Nature, based on these new physical laws. Then these predictions are put to systematic tests in independent, carefully designed and performed, repeatable experiments that produce objective data. Merely describing Nature and explaining physical experiments in terms of already existing laws is not physics in the true sense of the word.1 Had this non-creative, static view been adopted by all physicists since the days of Newton, we would still be doing essentially Newtonian physics. Even if such a scientic giant as M I C H A E L F A R A DAY , who had very little mathematical training, was able to make truly remarkable contributions to physics (and chemistry) using practically no formal mathematics whatsoever, it is for us mere mortals most convenient to use the shorthand language of mathematics, together with the formal methods of logic (inter alia propositional calculus), in physics. After all, mathematics was once introduced by us human beings to make it easier to quantitatively and systematically describe, understand and predict the physical world around us. Examples of this from ancient times are arithmetics and geometry. A less archaic example is differential calculus, needed by S I R I S A AC N E W T O N to formulate, in a compact and unambiguous manner, the physical laws that bear his name. Another more modern example is the delta function introduced by PAU L A D R I E N M AU R I C E D I R AC . But the opposite is also very common: the expansion and generalisation of mathematics has more than once provided excellent tools for creating new physical ideas and to better analyse observational data. Examples of the latter include non-Euclidean geometry and group theory. Unlike mathematics per se, where the criterion of logical consistency is both necessary and sufcient, a physical theory that is supposed to describe the physical reality has to full the additional criterion that its predictions be empirically testable. Ultimately, as G A L I L E O G A L I L E I taught us, physical reality is dened by the outcome of experiments and observations, and not by mere Aristotelean logical reasoning, however mathematically correct and logically consistent this may be. Common sense is not enough and logic and reasoning can never outsmart Nature. Should the outcome of repeated, carefully performed,
R A
221
FT
1
As S T E V E N W E I N B E R G puts it in the Preface To Volume I of The Quantum Theory of Fields: . . . after all, our purpose in theoretical physics is not just to describe the world as we nd it, but to explain in terms of a few fundamental principles why the world is the way it is.
222 j
13. MATHEMATICAL METHODS
The theoretical physicist S I R R U D O L F P E I E R L S (19071995) described an ideal physical theory in the following way: It must rstly leave undisturbed the successes of earlier work and not upset the explanations of observations that had been used in support of earlier ideas. Secondly it must explain in a reasonable manner the new evidence which brought the previous ideas into doubt and which suggested the new hypothesis. And thirdly it must predict new phenomena or new relationships between different phenomena, which were not known or not clearly understood at the time when it was invented.
independent experiments produce results that systematically contradict predictions of a theory, the only conclusion one can draw is that the theory in question, however logically stringent and mathematically correct it may be, is wrong.2 On the other hand, extending existing physical theories by mathematical and logical generalisations is a very powerful way of making hypotheses and predictions of new physical phenomena. History shows that if one has several alternative ways of generalising a theory, the theory with most generality, simplicity, elegance and beauty often is the best one. But it is not until these hypotheses and predictions have withstood empirical tests in physical experiments that they can be said to extend physics and our knowledge about Nature. This appendix describes briey some of the more common mathematical methods and tools that are used in Classical Electrodynamics.
M .1
Scalars, vectors and tensors
The Latin word vector means carrier.
Every physical observable can be represented by a mathematical object. We have chosen to describe the observables in classical electrodynamics in terms of scalars, pseudoscalars, vectors, pseudovectors, tensors and pseudotensors, all of which obey certain canonical rules of transformation under a change of coordinate systems and are completely dened by these rules. Despite certain advantages (and some shortcomings), differential forms will not be exploited to any signicant degree in our mathematical description of physical observables. A scalar, which may or may not be constant in time and/or space, describes the scaling of a physical quantity. A vector describes some kind of physical motion along a curve in space due to vection.3 A tensor describes the local motion or deformation of a surface or a volume due to some form of tension and is therefore a relation between a set vectors. However, generalisations to more abstract notions of these quantities have proved useful and are therefore commonplace. The difference between a scalar, vector and tensor and a pseudoscalar, pseudovector and a pseudotensor is that the latter behave differently under those coordinate transformations that cannot be reduced to pure rotations. For computational convenience, it is often useful to allow mathematical objects representing physical observables to be complex-valued, i.e., to let them be analytically continued into (a domain of) the complex plane. However, since by denition our physical world is real, care must be exercised when comparing mathematical results with physical observables, i.e., real-valued numbers obtained from physical measurements. Throughout we adopt the convention that Latin indices i; j; k; l; : : : run over the range f1; 2; 3g to denote vector or tensor components in the real Euclidean
R A
FT
M .1.
j 223
three-dimensional (3D) conguration space, and Greek indices ; ; ; ; : : : , which are used in four-dimensional (4D) spacetime, run over the range f0; 1; 2; 3g.
M .1.1
Vectors
M .1.1.1
Position vector
R A
xD
3 X i D1
The most basic vector, and the prototype against which all other vectors are benchmarked, is the position vector (radius vector, coordinate vector) which is the vector from the origin of the chosen coordinate system to the actual point of interest. Its N -tuple representation simply enumerates the coordinates of the position of this point. In this sense, the vector from the origin to a point is synonymous with the coordinates of the point itself. In the 3D Euclidean space R3 , we have N D 3 and the position vector x can be represented by the triplet .x1 ; x2 ; x3 / of its coordinates xi 2 R, i D 1; 2; 3. The coordinates xi are scalar quantities which describe the position along the O unit base vectors xi which span R3 . Therefore one convenient representation of the position vector in R3 is5 O O xi xi xi xi
def
FT
(M.1) (M.2)
6
Mathematically, a vector can be represented in a number of different ways. One suitable representation in a vector space of dimension N is in terms of an ordered N -tuple of real or complex numbers4 .a1 ; a2 ; : : : ; aN / of the components along N orthogonal coordinate axes that span the vector space under consideration. Note, however, that there are many ordered N -tuples of numbers that do not comprise a vector, i.e., do not have the necessary vector transformation properties!
It is often very convenient to use complex notation in physics. This notation can simplify the mathematical treatment considerably. But since all physical observables are real, we must in the nal step of our mathematical analysis of a physical problem always ensure that the results to be compared with experimental values are realvalued. In classical physics this is achieved by taking the real (or imaginary) part of the mathematical result, whereas in quantum physics one takes the absolute value.
5 We introduce the symbol which may be read is, by denition, to equal in meaning, or equals by denition, or, formally,
def
deniendum deniens. Another symbol sometimes used is .
def
where we have introduced Einsteins summation convention (E) that states that a repeated index in a term implies summation over the range of the index in question. Whenever possible and convenient we shall in the following always assume E and suppress explicit summation in our formul. Typographically, we represent a vector or vector operator in 3D Euclidean space by a boldface letter or symbol in a Roman font, for instance a, r , , and . Alternatively, we can describe the position vector x in component notation as xi where xi .x1 ; x2 ; x3 / In Cartesian coordinates6
def
In spherical polar coordinates .x1 ; x2 ; x3 / D .r; ; '/ and in cylindrical coordinates .x1 ; x2 ; x3 / D . ; '; z/.
224 j
.x1 ; x2 ; x3 / D .x; y; z/
(M.3)
This component notation is particularly useful in 4D space where we can represent the (one and the same) position vector either in its contravariant component form, (superscript index form) as the quartet x .x 0 ; x 1 ; x 2 ; x 3 / or its covariant component form (subscript index form) x .x0 ; x1 ; x2 ; x3 /
def def
(M.4)
(M.5)
M .1.2
Fields
R A
M .1.2.1
A eld is a physical entity that depends on one or more continuous parameters. Such a parameter can be viewed as a continuous index that enumerates the innitely many coordinates of the eld. In particular, in a eld that depends on the usual position vector x of R3 , each point in this space can be considered as one degree of freedom so that a eld is a representation of a physical entity with an innite number of degrees of freedom.
Scalar elds
We denote an arbitrary scalar eld in R3 by .x/ D .x1 ; x2 ; x3 / .xi /

def
FT
.x 0 ; x 1 ; x 2 ; x 3 / .x /
def
The relation between the covariant and contravariant forms is determined by the metric tensor (also known as the fundamental tensor) whose actual form is dictated by the properties of the vector space in question. The dual representation of vectors in contravariant and covariant forms is most convenient when we work in a vector space with an indenite metric. An example is Lorentz space L4 which is a 4D Riemannian space frequently employed to formulate the special theory of relativity.
(M.6)
This eld describes how the scalar quantity varies continuously in 3D R3 space. In 4D, a four-scalar eld is denoted (M.7)
which indicates that the four-scalar depends on all four coordinates spanning this space. Since a four-scalar has the same value at a given point regardless of coordinate system, it is also called an invariant.
M .1.
j 225
M .1.2.2
Vector elds
We can represent an arbitrary 3D real vector eld a.x/ as follows: a.x/ D ai .x/O i 2 R3 x In component notation this same vector can be represented as ai .x/ D .a1 .x/; a2 .x/; a3 .x// D ai .xj / (M.9) (M.8)
A 3D complex vector eld c.x/ is a vector in C3 (or, if we like, in R6 ), expressed in terms of two real vectors cR and cI in R3 in the following way c.x/ cR .x/ C icI .x/ D cR .x/O R C icI .x/O I c.x/O 2 C3 c a c which means that
def def
(M.10)
The use of complex vectors is in many cases a very convenient and powerful technique but requires extra care since physical observables must be represented by real vectors (2 R3 ).
The physical interpretation of a complex vector
A FT
Re fcg D cR Im fcg D cI (M.11a) (M.11b)
C i sin / ; C 2 R (M.12) 1 O x x h D p .O 1 iO 2 / 2 O h D1 O hC D 0 (M.13) O O O hC hC D h (M.14) (M.15) O O O hC h D h C i sin /.O 1 iO 2 / x x O O O sin x2 / C iC.sin x1 cos x2 / (M.16) O sin x2 / (M.17) measures the angle of rotation with angluar frequency
EXAMPLE
M .1
Study the physical meaning of vectors in complex notation where both the magnitudes and the base vectors are complex. Let us choose the complex magnitude as p c D jcjei D 2C.cos
that full the conditions
D
If
With these choices, our two vectors can be written O c D jcjh D C.cos O D C.cos x1
In order to interpret this expression correctly in physical terms, we must take the real part O Re fc g D C.cos x1 D !t, as is be the case when
and, without lack of generality, choose as base vectors the two complex unit base vectors, expressed in the two orthogonal base vectors x1 ; x2 2 R3
226 j
!, then O Re fc g D C cos.!t/ x1 O sin.!t/ x2 (M.18)
We see that the physically meaningful real part describes a rotation in positive or negative O sense, depending on the choice of sign. and therefore h are called helical base vectors. End of example M.1
In 4D, an arbitrary four-vector eld in contravariant component form can be represented as a .x / D .a0 .x /; a1 .x /; a2 .x /; a3 .x // or, in covariant component form, as (M.19)
where x is the position four-vector (radius four-vector, coordinate four-vector). Again, the relation between a and a is determined by the metric of the physical 4D system under consideration.
M .1.2.3
Coordinate transformations
R A D
We note that for a change of coordinates x ! x 0 D x 0 .x 0 ; x 1 ; x 2 ; x 3 /, due to a transformation from one coordinate system to another coordinate system 0 , the differential position vector dx transforms as dx 0 D @x 0 dx @x (M.21)
This follows trivially from the rules of differentiation of x 0 considered as a function of four variables x , i.e., x 0 D x 0 .x /. Analogous to the transformation rule for the differential dx , equation (M.21) above, the transformation rule for the differential operator @=@x under a transformation x ! x 0 becomes @ @x @ D 0 0 @x @x @x (M.22)
which, again, follows trivially from the rules of differentiation. Whether an arbitrary N -tuple fulls the requirement of being an (N -dimensional) contravariant vector or not, depends on its transformation properties during a change of coordinates. For instance, in 4D an assemblage y D .y 0 ; y 1 ; y 2 ; y 3 / constitutes a contravariant four-vector (or the contravariant components of a four-vector) if and only if, during a transformation from a system with coordinates x to a system 0 with coordinates x 0 , it transforms to the new system according to the rule y0 D @x 0 y @x (M.23)
FT
a .x / D .a0 .x /; a1 .x /; a2 .x /; a3 .x //
(M.20)
M .1.
j 227
i.e., in the same way as the differential coordinate element dx transforms according to equation (M.21) on the preceding page. The analogous requirement for a covariant four-vector is that it transforms, during the change from to 0 , according to the rule y0 D @x y @x 0 (M.24)
i.e., in the same way as the differential operator @=@x transforms according to equation (M.22) on the facing page.
M .1.2.4
Tensor elds
The 3D Kronecker delta tensor has the following matrix representation 1 0 1 0 0 C B (M.27) .ij / D @0 1 0A 0 0 1 Another common and useful tensor is the fully antisymmetric tensor of rank three, also known as the Levi-Civita tensor 8 1 if i; j; k is an even permutation of 1,2,3 < (M.28) 0 if at least two of i; j; k are equal ij k D : 1 if i; j; k is an odd permutation of 1,2,3 Clearly, this tensor fulls the relations
ij k
D
and
R A
D
j ki
Strictly speaking, the tensor eld described here is a tensor of rank two. A particularly simple rank two tensor in R3 is the 3D Kronecker delta tensor ij , with the following properties: ( 0 if i j ij D (M.26) 1 if i D j
ij k
j ik
FT
7
We denote an arbitrary tensor eld in R3 by A.x/. This tensor eld can be represented in a number of ways, for instance in the following matrix representation:7 0 1 A11 .x/ A12 .x/ A13 .x/ def def B C A.x/ Aij .xk / @A21 .x/ A22 .x/ A23 .x/A (M.25) A31 .x/ A32 .x/ A33 .x/
When a mathematical object representing a physical observable is given in matrix representation, we indicate this by enclosing the mathematical object in question in parentheses, i.e., .: : :/.
kij
(M.29)
ikj
(M.30)
228 j
and has the following further property

ij k ilm
D j l km
j m kl
(M.31)
A FT
0
111 121 131 211 221 231 112 122 132 212 222 232 ij k /iD1
In fact, tensors may have any rank n. In this picture, a scalar is considered to be a tensor of rank n D 0 and a vector to be a tensor of rank n D 1. Consequently, the notation where a vector (tensor) is represented in its component form is called the tensor notation. A tensor of rank n D 2 may be represented by a two-dimensional array or matrix, and a tensor of rank n D 3 may be represented as a vector of tensors of rank n D 2. Assuming that one of the indices of the Levi-Civita tensor ij k , e.g., the rst index i D 1; 2; 3, denotes the component of such a vector of tensors, these components have the matrix representations (the second and third indices, j; k D 1; 2; 3, are the matrix indices) 0 0 0 C B 0 123 A D @0 0 1 133 1 0 0 0 213 C B D @0 0 223 A 1 0 233 1 0 0 1 313 C B AD@ 1 0 323 0 0 333
113
B D@ 0
B ij k /iD2 D @
311 321 331
312 322 332
B ij k /iD3 D @
1 0 C 1A D iS1 0 1 1 C 0 A D iS2 0 1 0 C 0A D iS3 0
(M.32a)
(M.32b)
(M.32c)
Here we have introduced the matrix vector
R D
0 0 B S1 D @0 0 0 i 0
O S D Si xi
(M.33)
where the vector components Si are the matrices 1 0 C iA 0 1 0 0 i B C S2 D @ 0 0 0A i 0 0 0 0 0 B S3 D @ i 0 1 i 0 C 0 0A 0 0
(M.34)
which satisfy the angular momentum commutation rule Si ; Sj D i

ij k Sk
(M.35)
Tensors of rank higher than 3 are best represented in their tensor notation (component form). It is important to remember that a tensor of any rank is fully and totally characterized by its transformation properties under the change of coordinates. This is a very strict constraint.
M .1.
j 229
x3
Figure M.1: Tetrahedron-like volume element V containing matter.
O n
d2x x2 V
x1
Tensors in 3D space
A FT
T
O n
EXAMPLE
M .2
R
D cos 1 Tx1 d2x O cos 2 Tx2 d2x O
Consider a tetrahedron-like volume element V of a solid, uid, or gaseous body, whose atomistic structure is irrelevant for the present analysis; gure M.1 indicates how this volO ume may look like. Let dS D d2x n be the directed surface element of this volume element and let the vector Tn d2x be the force that matter, lying on the side of d2x toward which the O O unit normal vector n points, acts on matter which lies on the opposite side of d2x. This force concept is meaningful only if the forces are short-range enough that they can be assumed to act only in the surface proper. According to Newtons third law, this surface force fulls
Tn O
(M.36)
D
Tn d2x O
Using (M.36) and Newtons second law, we nd that the matter of mass m, which at a given instant is located in V obeys the equation of motion cos 3 Tx3 d2x C Fext D ma O (M.37)
where Fext is the external force and a is the acceleration of the volume element. In other words m Fext Tn D n1 Tx1 C n2 Tx2 C n3 Tx3 C 2 a (M.38) O O O O m dx Since both a and Fext =m remain nite whereas m=d2x ! 0 as V ! 0, one nds that in
230 j
this limit Tn D O
3 X i D1
ni Txi ni Txi O O
(M.39)
From the above derivation it is clear that equation (M.39) above is valid not only in equilibrium but also when the matter in V is in motion. Introducing the notation Tij D Txi O
j
(M.40)
for the j th component of the vector Txi , we can write equation (M.39) above in component O form as follows Tnj D .Tn /j D O O
3 X i D1
ni Tij ni Tij
(M.41)
Using equation (M.41) above, we nd that the component of the vector Tn in the direction O O of an arbitrary unit vector m is O Tnm D Tn m O O O D
3 X 3 3 X X
A FT
! Tnj mj D O ni Tij O O mj ni Tij mj D n T m
j D1 j D1 i D1
(M.42)
Hence, the j th component of the vector Txi , here denoted Tij , can be interpreted as the O ij th component of a tensor T. Note that Tnm is independent of the particular coordinate O O system used in the derivation. We shall now show how one can use the momentum law (force equation) to derive the equation of motion for an arbitrary element of mass in the body. To this end we consider a part V of the body. If the external force density (force per unit volume) is denoted by f and the velocity for a mass element dm is denoted by v, we obtain Z Z Z d v dm D f d3x C Tn d2x (M.43) O dt V V S The j th component of this equation can be written Z Z Z Z Z d vj dm D fj d3x C Tnj d2x D fj d3x C ni Tij d2x O V dt V S V S where, in the last step, equation (M.41) above was used. Setting dm D divergence theorem on the last term, we can rewrite the result as Z Z Z @Tij 3 d vj d3x D fj d3x C dx V dt V V @xi @Tij D0 @xi @Tij D0 @xi
R D
or, equivalently
(M.44)
d3x and using the
(M.45)
Since this formula is valid for any arbitrary volume, we must require that d vj dt fj (M.46)
@vj C v r vj @t
fj
(M.47)
M .1.
j 231
Note that @vj =@t is the rate of change with time of the velocity component vj at a xed point x D .x1 ; x1 ; x3 /. End of example M.2
In 4D, we have three forms of four-tensor elds of rank n. We speak of a contravariant four-tensor eld , denoted A a covariant four-tensor eld , denoted A a mixed four-tensor eld , denoted A
1 1 2 ::: n
.x /,
2 ::: n
.x /,
1 2 ::: k kC1 ::: n
.x /.
The 4D metric tensor (fundamental tensor) mentioned above is a particularly important four-tensor of rank two. In covariant component form we shall denote it g . This metric tensor determines the relation between an arbitrary contravariant four-vector a and its covariant counterpart a according to the following rule: a .x / g
def
This rule is often called lowering of index. The raising of index analogue of the index lowering rule is: a .x / g
def
R
g
k k
More generally, the following lowering and raising rules hold for arbitrary rank n mixed tensor elds: A
1 2 ::: k 1 k kC1 kC2 ::: n
D
g
k k
1 2 ::: k 1 k kC1 ::: n
Successive lowering and raising of more than one index is achieved by a repeated application of this rule. For example, a dual application of the lowering operation on a rank two tensor in its contravariant form yields A Dg
g
i.e., the same rank two tensor in its covariant form.
A FT
a .x / (M.48) a .x / (M.49) .x / D A
1 2 ::: k 1 k kC1 ::: n
.x /
(M.50)
.x / D A
1 2 ::: k 1 k kC1 kC2 ::: n
.x /
(M.51)
(M.52)
232 j
EXAMPLE
M .3
Contravariant and covariant vectors in at Lorentz space The 4D Lorentz space L4 has a simple metric which can be described by the metric tensor 8 if D D 0 1 < g D D (M.53) 1 if D D i D j D 1; 2; 3 : 0 if which in matrix notation becomes 0 1 B0 B /D@ 0 0 0 1 0 0 0 0 1 0 1 0 0C C 0A 1
(M.54)
which in matrix representation becomes
A FT
. 1 B 0 /DB @ 0 0 0 0 1 0 0 0 0 1 0 1 0 0C C 0A 1 a .a0 ; a1 ; a2 ; a3 / D .a0 ; a/
def
i.e., a matrix with a main diagonal that has the sign sequence, or signature, fC; ; ; g. Alternatively, one can dene the metric tensor in L4 as 8 1 if D D 0 < D 1 (M.55) if D D i D j D 1; 2; 3 : 0 if
(M.56)
i.e., a matrix with signature f ; C; C; Cg. Of course, the physics is unaffected by the choice of metric tensor. Consider an arbitrary contravariant four-vector a in this space. In component form it can be written:
R D
or D0W D1W D2W D3W
(M.57)
According to the index lowering rule, equation (M.48) on the preceding page, we obtain the covariant version of this vector as a
def
.a0 ; a1 ; a2 ; a3 / D
(M.58)
In the fC; ; ; g metric we obtain a0 D 1 a 0 C 0 a 1 C 0 a 2 C 0 a 3 D a 0 a1 D 0 a

0 0 0
(M.59)
1 2 3
1 a C0 a C0 a D
1 1
2 2
3 3 3
a a a
(M.60) (M.61) (M.62)
a2 D 0 a C 0 a
1 a C0 a D
2
a3 D 0 a C 0 a C 0 a
1 a D
a D .a0 ; a1 ; a2 ; a3 / D .a0 ; a1 ; a2 ; a3 / D .a0 ; a/ The radius 4-vector itself in L4 and in this metric is given by
(M.63)
M .2.
Vector algebra
j 233
x D .x 0 ; x 1 ; x 2 ; x 3 / D .ct; x; y; z/ D .ct; x/ x D .x0 ; x1 ; x2 ; x3 / D .ct; x 1 ; x 2 ; x 3 / D .ct; x/ Analogously, using the f ; C; C; Cg metric we obtain a D .a0 ; a1 ; a2 ; a3 / D . a0 ; a1 ; a2 ; a3 / D . a0 ; a/
(M.64)
(M.65)
End of example M.3
M .2
Vector algebra
The scalar product (dot product, inner product) of two arbitrary 3D vectors a and b in Euclidean R3 space is the scalar number O O O O a b D ai xi bj xj D xi xj ai bj D ij ai bj D ai bi
O O where we used the fact that the scalar product xi xj is a representation of the Kronecker delta ij dened in equation (M.26) on page 227.8 The scalar product of a vector a in R3 with itself is a a .a/2 D jaj2 D .ai /2 D a2 and similarly for b. This allows us to write
def
where is the angle between a and b.
Scalar products in complex vector space
R
c2 c c D c2 R
def
A FT
(M.66)
8
M .2.1
Scalar product
In the Russian literature, the 3D scalar product is often denoted .ab/.
(M.67)
a b D ab cos
(M.68)
EXAMPLE
M .4
D
from which we nd that
Let c be a complex vector dened as in expression (M.10) on page 225. The inner product of c with itself may be dened as c2 C 2icR cI c 2 2 C I
def
(M.69)
cD
c2 R
c2 C 2icR cI 2 C I
(M.70)
Using this in equation (M.10) on page 225, we see that we can dene the complex unit vector as being
234 j
O cD
c D q c
cR c2 R c2 I C 2icR cI c2 I 1
O cR C i q
cI c2 R c2 I C 2icR cI
O cI
q
D cR
c2 R
q
2icR cI 4c2 c2 sin2 R I .c2 C c2 /2 R I O cR C i cI
c2 R
c2 I 1
2icR cI 4c2 c2 sin2 R I .c2 C c2 /2 R I
s
.c2 C c2 / R I
s
.c2 C c2 / R I
O cI 2 C3
(M.71) On the other hand, the denition of the scalar product in terms of the inner product of a complex vector with its own complex conjugate yields jcj2 c c D c2 C c2 D jcj2 R I with the help of which we can dene the unit vector as cR cI c O O D q cR C i q cI jcj 2 2 2 2 cR C cI cR C cI
def
(M.72)
A FT
O cD
cR
c2 C c2 R I
(M.73)
c2 C c2 R I
O cR C i
cI
c2 C c2 R I
c2 C c2 R I
O cI 2 C3
End of example M.4
In 4D space we dene the scalar product of two arbitrary four-vectors a and b in the following way a b Dg a b Da b Dg a b (M.74)
R D
EXAMPLE
M .5
where we made use of the index lowering and raising rules (M.48) and (M.49). The result is a four-scalar, i.e., an invariant which is independent of in which 4D coordinate system it is measured. The quadratic differential form ds 2 D g dx dx D dx dx (M.75)
i.e., the scalar product of the differential position four-vector with itself, is an invariant called the metric. It is also the square of the line element ds which is the distance between neighbouring points with coordinates x and x C dx .
Scalar product, norm and metric in Lorentz space
In L4 the metric tensor attains a simple form [see example M.3 on page 232] and, hence, the scalar product in equation (M.74) above can be evaluated almost trivially. For the fC; ; ; g signature it becomes a b D .a0 ; a/ .b 0 ; b/ D a0 b 0 a b (M.76)
In L4 the important scalar product of the postition four-vector with itself therefore takes the
M .2.
Vector algebra
j 235
simple form x x D .x0 ; x/ .x 0 ; x/ D .ct; x/ .ct; x/ D .ct/2 .x 1 /2 .x 2 /2 .x 3 /2 D s 2 (M.77)
which is the indenite, real norm of L4 . The L4 metric is the quadratic differential form ds 2 D dx dx D c 2 .dt/2 .dx 1 /2 .dx 2 /2 .dx 3 /2 (M.78)
End of example M.5
M .2.2
Vector product
The vector product or cross product of two arbitrary 3D vectors a and b in ordinary R3 space is the vector9 cDa bD O ij k xi aj bk D O ij k aj bk xi (M.79)
Here ij k is the Levi-Civita tensor dened in equation (M.28) on page 227. Alternatively, a O b D ab sin e
O where is the angle between a and b and e is a unit vector perpendicular to the plane spanned by a and b. 0 0 0 A spatial reversal of the coordinate system, .x1 ; x2 ; x3 / D . x1 ; x2 ; x3 /, known as a parity transformation, changes sign of the components of the vectors a and b so that in the new coordinate system a0 D a and b0 D b, which is to say that the direction of an ordinary vector is not dependent on the choice of the directions of the coordinate axes. On the other hand, as is seen from equation (M.79) above, the cross product vector c does not change sign. Therefore a (or b) is an example of a true vector, or polar vector, whereas c is an example of an pseudovector or axial vector. A prototype for a pseudovector is the angular momentum vector L D x p and hence the attribute axial. Pseudovectors transform as ordinary vectors under translations and proper rotations, but reverse their sign relative to ordinary vectors for any coordinate change involving reection. Tensors (of any rank) that transform analogously to pseudovectors are called pseudotensors. Scalars are tensors of rank zero, and zero-rank pseudotensors are therefore also called O x O O O pseudoscalars, an example being the pseudoscalar xi .O j xk / D xi . ij k xi /. This triple product is a representation of the ij k component of the rank three Levi-Civita pseudotensor ij k .
A FT
9
Sometimes the 3D vector product of a and b is denoted a ^ b or, particularly in the Russian literature, ab.
(M.80)
236 j
EXAMPLE
M .6
The vector triple product The vector triple product is the vector product of a vector a with a vector product b can, with the help of formula (M.31) on page 228, be evaluated as a .b c/ D D D O ij k xi aj .b
kij
c and
c/k D
O O ij k xi aj . lmn xl bm cn /k D O ij k kmn xi aj bm cn i n j m /O i aj bm cn x O aj bj ci xi (M.81)
O ij k xi aj kmn bm cn
O kmn xi aj bm cn D .i m j n .a b/c D b.a c/
O D i m j n xi aj bm cn D .a c/b
O O i n j m xi aj bm cn D aj cj bi xi c.a b/
which is formula (F.108) on page 216. This is sometimes called Lagranges formula, but is more often referred to as the bac-cab rule. End of example M.6
M .2.3
Dyadic product
R A
In matrix representation
The dyadic product eld A.x/ a.x/b.x/ with two juxtaposed vector elds a.x/ and b.x/ is the outer product of a and b known as a dyad . Here a is called the antecedent and b the consequent. The dyadic product between vectors (rank one tensors) is a special instance of the direct product, usually denoted , between arbitrary rank tensors. This associative but non-commuting operation is also called the tensor product or, when applied to matrices (matrix representations of tensors), the Kronecker product. Written out in explicit form, the dyadic product A.x/ D a.x/b.x/ becomes O O O O A D ab a b D a1 x1 b1 x1 C a1 x1 b2 x2 C O O C a3 x3 b3 x3 (M.82a) (M.82b) (M.82c) O O O O O O D ai xi bj xj D ai bj xi xj D xi ai bj xj 10 1 0 O x1 a1 b1 a1 b2 a1 b3 CB C B D x1 x2 x3 @a2 b1 a2 b2 a2 b3 A @x2 A O O O O O a3 b1 a3 b2 a3 b3 x3 0
0 1 a1 B C .A/ D .ab/ .a b/ D @a2 A b1 a3
FT
b2 b3 a1 b1 B D @a2 b1 a3 b1 a1 b2 a2 b2 a3 b2
def def
1 a1 b3 C a2 b3 A a3 b3 (M.83)
which we identify with expression (M.25) on page 227, viz. a tensor in matrix representation. Hence, a dyadic of two vectors is intimately related to a rank two tensor, emphasising its vectorial characteristics. Scalar multiplication from the right or from the left of the dyad A.x/ D a.x/b.x/ by a vector c, produces other vectors according to the scheme O A c ab c a.b c/ D aj bi ci xj (M.84a)
M .2.
Vector algebra
j 237
O c A c ab .c a/b D ai bj ci xj
def
def
(M.84b)
respectively. These two vectors, proportional to a and b, respectively, are in general not identical to each other. In the rst case, c is known as the postfactor, in the second case as the prefactor. O Specically, if c D xj , then O O A xj D abj D ai bj xi O O xj A D aj b D aj bi xi which means that O O xi A xj D ai bj D Aij The vector product can be represented in matrix form as follows: 0 1 0 1 c1 a2 b3 a3 b2 B C B C .c/ D @c2 A D .a b/ D @a3 b1 a1 b3 A D ia Sb D iaS b c3 a1 b2 a2 b1 (M.85c) (M.85a) (M.85b)
where Sb is dyadic product of the matrix vector S,given by formula (M.33) on page 228, and the vector b, and aS is the dyadic product of the vector a and the matrix vector S.
Matrix representation of the vector product in R3
A FT
(M.86)
b is given by formula a S D ai Si (M.87) (M.88) a3 0 a1 10 1 0 a2 b1 a2 b3 a1 A @b2 A D i @a3 b1 0 b3 a1 b 2 1 a3 b2 a1 b3 A a2 b1 (M.89)
EXAMPLE
M .7
Prove that the matrix representation of the vector product c D a (M.86) on page 237.
According to formula (M.84) on the facing page, the scalar multiplication of a dyadic product of two vectors (in our case S and b) from the left by a vector (in our case a) is interpreted as a Sb D .a S/b where
D
from which we nd that
and the components Si are given by formula (M.34) on page 228. Hence 0 1 0 1 0 1 0 0 0 0 0 i 0 i 0 a S D a1 @0 0 iA C a2 @ 0 0 0A C a3 @ i 0 0A 0 i 0 i 0 0 0 0 0 20 1 0 1 0 13 0 0 0 0 0 a2 0 a3 0 D i 4@0 0 a1 A C @ 0 0 0 A C @a3 0 0A5 0 a1 0 a2 0 0 0 0 0 0 1 0 a3 a2 D i @ a3 0 a1 A a2 a1 0 0
R
0 .a S/b D i @ a3 a2
238 j
In other words, i.a S/b D a2 b3 ia Sb D @a3 b1 a1 b2 0 1 a3 b2 a1 b3 A D .a a2 b1 b/ (M.90)
Likewise, aS b D a.S b/ where 0 1 0 0 0 0 0 S b D Si bi D @0 0 iA b1 C @ 0 0 i 0 i 0 1 0 b3 b2 D i @ b3 0 b1 A b2 b1 0 and a.S b/ D i a1 from which follows a2
0 0 0
1 0 i 0 0A b2 C @ i 0 0
i 0 0
1 0 0A b3 0
(M.91)
FT
0 a3 0 @ b3 b2 b3 0 b1 1 0 b2 a2 b3 b1 A D i @a3 b1 0 a1 b2 1 a3 b2 a1 b3 A a2 b1 0 a2 b3 iaS b D @a3 b1 a1 b2 1 a3 b2 a1 b3 A D .a a2 b1 b/ iaS b
(M.92)
ia.S b/ D Hence, .a b/ D ia Sb D
(M.93) QED
R A
A
End of example M.7
Vector multiplication from the right and from the left of the dyad A by a vector is another dyad according to the scheme c D ab c a.b
def def
c/ D
O O j kl ai bk cl xi xj O O j kl al bi ck xj xi D O O ikl al bj ck xi xj
(M.94a) (M.94b)
ADc
ab .c
a/b D
respectively. In general, the two new dyads thus created are not identical to each other. O O Specically, if A D 13 D xi xi , i.e., the unit dyad or the second-rank unit tensor, then
13
cD
O O O O c3 x1 x2 C c2 x1 x3 O O C c3 x2 x1 O O c1 x2 x3
13
(M.95)
O O O O c2 x1 x3 C c1 x3 x2 D c
or, in matrix representation, 0 B c/ D @ c3 c2 0 c3 0 c1 1 c2 C c1 A D .c 0
.1 3
13 /
(M.96)
M .3.
Vector calculus
j 239
Using the matrix vector formula (M.33) on page 228, we can write this as .13 c/ D .c
13 / D
iS c
(M.97)
One can extend the dyadic scheme and introduce abc, called a tryad , and so on. In this vein, a vector a is sometimes called a monad .
M .3
Vector calculus
The del operator
M .3.1
O O r D xi ri xi
def
@ def @ def @ @xi @x
In 4D, the contravariant component representation of the four-del operator is dened by @ @ @ @ @ D ; ; ; (M.100) @x0 @x1 @x2 @x3 whereas the covariant four-del operator is @ @ @ @ ; ; ; @ D @x 0 @x 1 @x 2 @x 3
R A
y0 D @ x0 y y 0 D @0 x y
O where xi is the ith unit vector in a Cartesian coordinate system. Since the operator in itself has vectorial properties, we denote it with a boldface nabla (r ). In component (tensor) notation the del operator can be written @ @ @ @i D ; ; (M.99) @x1 @x2 @x3
D
and respectively.
We can use this four-del operator to express the transformation properties (M.23) and (M.24) on page 227 as (M.102)
FT
10
In R3 the del operator is a differential vector operator, denoted in Gibbs notation by the boldface nabla symbol r and dened as10 (M.98)
This operator was introduced by W I L L I A M R OW E N H A M I LT O N (18051865) who, however, used the symbol for it. It is therefore sometimes called the Hamilton operator.
(M.101)
(M.103)
240 j
M .3.2
The gradient of a scalar eld
The gradient of an R scalar eld .x/, is an R3 vector eld O grad .x/ r .x/ D @.x/ D xi @i .x/ If the scalar eld depends only on one coordinate, say, then @. / O D r. / @
def
(M.104)
O r .x/ D r . / D
(M.105)
O and, therefore, r D r. From this we see that the boldface notation for the gradient (r ) is very handy as it elucidates its 3D vectorial property and separates it from the nabla operator (r) which has a scalar property. In 4D, the four-gradient of a four-scalar is a covariant vector, formed as a derivative of a four-scalar eld .x /, with the following component form: @.x / @x
R A
M .8
with the contravariant form
EXAMPLE
Gradient of a scalar product of two vector elds
The gradient of the scalar product of two vector elds a and b can be calculated in the following way: O O r .a b/ D .O i @i /.aj xj bk xk / x O O O O D .O i @i /.aj xj / .bk xk / C .aj xj / .O i @i /.bk xk / x x O O D .O i @i aj xj / b C .O i @i bk xk / a D .r a/ b C .r b/ a x x (M.107)
This is the rst version of formula (F.133) on page 216. End of example M.8
FT
@ .x / D @ .x / D @.x / @x
(M.106a)
(M.106b)
M .3.
Vector calculus
j 241
M .3.3
The divergence of a vector eld

def
We dene the 3D divergence of a vector eld a in R3 as div a.x/ r a.x/ D @ai .x/ D @i ai .x/ @xi (M.108)
which, as indicated by the notation .x/, is a scalar eld in R3 . The four-divergence of a four-vector a is the four-scalar @ a .x / D @a .x / @x (M.109)
M .3.4
The curl of a vector eld
In R3 the curl of a vector eld a.x/ is another R3 vector eld dened in the following way: curl a.x/ r
def
a.x/ D
O ij k xi @j ak .x/
where S is the matrix vector given by formula (M.33) on page 228. The covariant 4D generalisation of the curl of a four-vector eld a .x / is the antisymmetric four-tensor eld A .x / D @ a .x / @ a .x / D A .x / (M.112)
D
M .3.5
A vector with vanishing curl is said to be irrotational .
The Laplacian
The 3D Laplace operator or Laplacian can be described as the divergence of the del operator: r r D r 2 D @2 D i
def
R A
3
where use was made of the Levi-Civita tensor, introduced in equation (M.28) on page 227. If a is an ordinary vector (polar vector), then r a is a pseudovector (axial vector) and vice versa. Similarly to formula (M.86) on page 237, we can write the matrix representation of the curl in R3 as 1 0 @2 a3 @3 a2 C B (M.111) .r a/ D @@3 a1 @1 a3 A D ir Sa D ir S a @1 a2 @2 a1
X @2 @2 2 @xi @xi2 i D1
FT
D O ij k xi @ak .x/ @xj (M.110) (M.113)
242 j
The symbol r 2 is sometimes read del squared . If, for a scalar eld .x/, r 2 < 0 at some point in 3D space, has a concentration at that point. Numerous vector algebra and vector calculus formul are given in appendix F on page 205. Those which are not found there can often be easily derived by using the component forms of the vectors and tensors, together with the Kronecker and Levi-Civita tensors and their generalisations to higher ranks and higher dimensions.
EXAMPLE
M .9
The four-del operator in Lorentz space In L4 the contravariant form of the four-del operator can be represented as 1 @ 1 @ @ D ; @ D ; r c @t c @t and the covariant form as
(M.114)
A FT
@ D 1 @ 1 @ ;@ D ;r c @t c @t @ @ D 1 @2 c 2 @t 2 r2 D
2
(M.115)
Taking the scalar product of these two, one obtains
(M.116)
which is the dAlembert operator, sometimes denoted opposite sign convention.
, and sometimes dened with an End of example M.9
EXAMPLE
M .10
Gradients of scalar functions of relative distances in 3D
R D
r x Likewise r
Very often electrodynamic quantities are dependent on the relative distance in R3 between two vectors x and x0 , i.e., on x x0 . In analogy with equation (M.98) on page 239, we can dene the primed del operator in the following way: O r 0 D xi @ 0 0 D@ @xi (M.117)
Using this, the corresponding unprimed version, viz., equation (M.98) on page 239, and elementary rules of differentiation, we obtain the following very useful result: 0 0 0 @ .x1 x1 /2 C .x2 x2 /2 C .x3 x3 /2 @x x0 0 O O x D xi D xi @xi @xi (M.118) 0 .xi xi / @x x0 x x0 O O D D xi D xi D r 0 x x0 0 @xi jx x0 j jx x0 j jx 1 x0 j D x jx x0 x0 j3 jx 1 x0 j (M.119)
r0
End of example M.10
M .3.
Vector calculus
j 243
Divergence and curl of a vector eld divided by the relative distance For an arbitrary vector eld the following relations hold: 0/ 0 a.x0 / a.x r 1 0 0 0 r D C a.x / r jx x0 j jx x0 j jx x0 j a.x0 / r 0 a.x0 / 1 r0 D C a.x0 / r 0 jx x0 j jx x0 j jx x0 j R3 a.x0 /,
EXAMPLE
M .11
(M.120a) (M.120b)
This demonstrates how the primed divergence and curl, dened in terms of the primed del operator in equation (M.117) on the preceding page, work. End of example M.11
The Laplacian and the Dirac delta function A very useful formula in r r R3 jx is 1
EXAMPLE
M .12
x0 j
where .x x0 / is the 3D Dirac delta function. This formula follows directly from the fact that Z I Z 1 x x0 x x0 O d3x 0 r r d3x 0 r d2x 0 n0 D D jx x0 j jx x0 j3 jx x0 j3 V0 V0 S0 (M.122) equals 4 if the integration volume V 0 .S 0 /, enclosed by the surface S 0 .V 0 /, includes x0 D x, and equals 0 otherwise. End of example M.12
The curl of a gradient Using the denition of the R3 r
A FT
D r2 1 jx x0 j D 4 .x x0 / (M.121) O ij k xi @j r .x/k D O ij k xi @j @k .x/ (M.123) D
ij k
EXAMPLE
M .13
curl, equation (M.110) on page 241, and the gradient, equation (M.104) on page 240, we see that r .x/ D
which, due to the assumed well-behavedness of .x/, vanishes:
R
O ij k xi @j @k .x/ D C C 0
@ @ .x/O i x @xj @xk @2 @x3 @x2 ! .x/O 1 x ! .x/O 2 x ! .x/O 3 x (M.124)
@2 @x2 @x3 @2 @x3 @x1 @2 @x1 @x2
@2 @x1 @x3 @2 @x2 @x1
244 j
We thus nd that r r .x/ 0 (M.125)
for any arbitrary, well-behaved R3 scalar eld .x/. In 4D we note that for any well-behaved four-scalar eld .x / .@ @ @ @ /.x / 0 (M.126)
so that the four-curl of a four-gradient vanishes just as does a curl of a gradient in R3 . Hence, a gradient is always irrotational . End of example M.13
EXAMPLE
M .14
The divergence of a curl
With the use of the denitions of the divergence (M.108) and the curl, equation (M.110) on page 241, we nd that r r a.x/ D @i r a.x/i D
ij k @i @j ak .x/
A FT
@i
ij k @j ak .x/
(M.127)
Using the denition for the Levi-Civita symbol, dened by equation (M.28) on page 227, we nd that, due to the assumed well-behavedness of a.x/, D @ @xi
ij k
@ a @xj k
@2 @x2 @x3
@2 @x3 @x2
a1 .x/ !
C C
@2 @x3 @x1 @2 @x1 @x2
@2 @x1 @x3 @2 @x2 @x1
a2 .x/ a3 .x/
(M.128)
R
i.e., that
r r
a.x/ 0
(M.129)
for any arbitrary, well-behaved R3 vector eld a.x/. The 3D curl is therefore solenoidal (has a vanishing divergence). In 4D, the four-divergence of the four-curl is not zero, for @ A D @ @ a .x /
2
a .x / 0
(M.130)
End of example M.14
M .3.
Vector calculus
j 245
The curl of the curl of a vector eldthe bac-cab rule for the del operator The curl of the curl of a vector eld can be viewed as the del operator version of the bac-cab that was derived in example M.6 on page 236. By using formula (M.31) on page 228 it can be evaluated as r .r a/ D D D O ij k xi @j .r O ij k xi @j a/k D O O ij k xi @j . lmn xl @m an /k O ij k kmn xi @j @m an i n j m /O i @j @m an x O 2 xi @j ai (M.131)
kmn @m an D
EXAMPLE
M .15
O kij kmn xi @j @m an
D .i m j n
O D i m j n xi @j @m an O D xi @j @i aj D r .r a/
O i n j m xi @j @m an r2a
O O xi @j @j ai D xi @i @j aj r2a D r r a
which is formula (F.151) on page 217.
M .3.6
Vector and tensor integrals
In this subsection we derive some (Riemann) integral identities involving vectors and/or tensors, Of particular interest, importance and usefulness are identities where the integrand contains the scalar 1= jx x0 j.11
M .3.6.1
First order derivatives
Let us start with the gradient of a volume integral of a scalar eld divided by jx x0 j. It can be written Z Z 0 1 3 0 .x / 3 0 0 r dx D (M.132) d x .x /r jx x0 j jx x0 j V0 V0 where we used formula (F.132) on page 216 with .x/ D 1= jx x0 j, noticing that is a function of x0 only and therefore behaves as a constant under differentiation with respect to x. The results obtained in example M.10 on page 242 allow us to make the replacement r ! r 0 in the integrand, leading to Z Z 0 1 3 0 .x / 3 0 0 0 r dx D d x .x /r (M.133) jx x0 j jx x0 j V0 V0 We can now integrate the RHS by part by invoking formula (F.132) once more, but this time with r 0 instead of r . The result is Z Z Z 0 0 0 .x0 / 3 0 r .x / 3 0 0 3 0 .x / D dx dx r (M.134) r dx jx x0 j jx x0 j jx x0 j V0 V0 V0 Formula (F.176) on page 218 enables us to replace the last volume integral with a surface integral, yielding the nal result Z Z I .x0 / .x0 / r 0 .x0 / O r d3x 0 D d3x 0 d2x 0 n (M.135) 0j 0j jx x jx x jx x0 j V0 V0 S0
R A
FT
11
End of example M.15
Such integrands are closely related to Green functions (propagators).
246 j
An analogous approach for the divergence of a volume integral of a regular vector eld a.x0 / divided by jx x0 j yields, with the use of formula (F.135) on page 217 and the results in example M.10 on page 242, Z Z 1 a.x0 / D d3x 0 a.x0 / r r d3x 0 jx x0 j jx x0 j V0 V0 (M.136) Z 1 D d3x 0 a.x0 / r 0 jx x0 j V0 We integrate this by part, again employing formula (F.135) on page 217 and formula (F.177) on page 218, to obtain Z I Z r 0 a.x0 / a.x0 / a.x0 / O (M.137) D d3x 0 d2x 0 n0 r d3x 0 0j 0j jx x jx x jx x0 j V0 S0 V0 For the curl of the same vector integral, we do not repeat the steps but only quote the nal result Z Z I a.x0 / r 0 a.x0 / a.x0 / O r d3x 0 D (M.138) d3x 0 d2x 0 n0 0j 0j jx x jx x jx x0 j V0 V0 S0 The above results can be summarised in the general partial integration formula Z Z I 0 A A 3 0 3 0 r A O r dx D (M.139) dx d2x 0 n0 0j 0j 0 0 0 jx x jx x jx x0 j V V S where is either (i) nothing (juxtaposition) and A D .x0 /, or (ii) D or D and A D a.x0 /. In the surface integrals in the formul above, the surface O element d2x 0 D dS0 n0 is proportional to r 2 D jx x0 j2 [cf. formula (F.73) on page 213]. Hence, if A falls off faster than 1=r, this surface integral vanishes when we let the radius r of the sphere S 0 , over which the surface integral is to be evaluated, tend to innity. If A falls off exactly as 1=r for large r, the surface integral tends to a constant, which, in the special cases D ; A D O O a ? n0 and D ; A D a k n0 is zero. If A falls off slower than 1=r at innity, the surface integral is singular and, consequently, the above formul are inapplicable. When the surface integral vanishes the following simple and very useful formula obtains: Z r
V0
R A
d3x 0 A jx
M .3.6.2
FT
Z x0 j D
V0
d3x 0
r0 A jx x0 j
(M.140)
Second order derivatives
Let us also derive two identities involving second derivatives of an integral over V 0 where the integrand is either a regular, differentiable scalar .x0 / or a regular vector eld a.x0 / divided by jx x0 j.
M .3.
Vector calculus
j 247
The divergence of the gradient of an integral where the integrand is of the rst kind can be written Z Z 1 .x0 / D d3x 0 .x0 /r r r r d3x 0 jx x0 j jx x0 j V0 V0 (M.141) Z 1 3 0 0 2 D d x .x /r jx x0 j V0 The representation of the Dirac delta function given by expression (M.121) on page 243 immediately gives the simple result that Z .x0 / r r d3x 0 D 4 .x/ (M.142) jx x0 j V0
where we used the identity (F.151) on page 217. The second integral in the RHS is the vector form of formula (M.142) above : Z a.x0 / r r d3x 0 D 4 a.x/ (M.144) jx x0 j V0
Using the fact that r operates on unprimed coordinates whereas a depends only on x0 , we can rewrite the rst integral in the RHS as Z Z 1 1 d3x 0 a.x0 / r r D (M.145) d3x 0 a.x0 / r 0 r 0 jx x0 j jx x0 j V0 V0 where we twice used the fact that r D r 0 when they operate on 1= jx x0 j and its gradient. Now we can utilise identity (F.140) on page 217 with b D r 0 .1= jx x0 j/ to integrate the RHS by parts as follows: Z 1 3 0 0 0 0 d x a.x / r r jx x0 j V0 Z Z 1 1 3 0 0 0 0 3 0 0 0 0 C dx r a.x /r D d x r a.x /r 0 jx x0 j jx x0 j V0 IV Z 1 1 3 0 0 0 0 2 0 0 0 0 O D d x r a.x /r C d x n a.x /r jx x0 j jx x0 j V0 S0 (M.146) where, in the last step, we used the divergence theorem for tensors/dyadics, formula (F.179) on page 218. Putting it all together, we nally obtain the integral identity
R A
FT
The curl of the curl of an integral with an integrand of the second kind is Z Z 0 a.x0 / 3 0 a.x / D rr d3x 0 r r dx jx x0 j jx x0 j V0 V0 (M.143) Z 0 3 0 a.x / r r dx jx x0 j V0
248 j
r r
Z
V0
d3x 0
a.x0 / jx x0 j
Z 1 D 4 a.x/ d3x 0 r 0 a.x0 /r 0 jx x0 j V0 I a.x0 /.x x0 / O C d2x 0 n0 (M.147) jx x0 j3 S0
where the surface integral vanishes for any well-behaved a.
M .3.7
Helmholtzs theorem
If u.x/ is regular and falls off rapidly enough with distance r D jx x0 j (typically faster than 1=r when r ! 1), we can use formula (M.140) on page 246 to rewrite this as u.x/ D where .x/ D r .x/ C r d3x 0 a.x/ (M.149a)
R A
Z
r 0 u.x0 / 4 jx x0 j V0 Z r 0 u.x0 / a.x/ D d3x 0 4 jx x0 j V0
Since, according to example M.13 on page 243, r .r / D 0, i.e., r is irrotational , and, according to example M.14 on page 244, r .r a/ D 0, i.e., r a is rotational (also called divergence-less or solenoidal ), u can always be decomposed into an irrotational and a rotational component u.x/ D uirrot .x/ C urotat .x/ Z r .x/ D r r 0 u.x0 / 4 jx x0 j V0 Z r 0 u.x0 / d3x 0 4 jx x0 j V0 d3x 0 (M.150a)
where uirrot .x/ D urotat .x/ D r (M.150b) (M.150c)
This is called Helmholtz decomposition. Furthermore, since r uirrot D 0 (M.151a)
FT
a.x/ D r
Let us consider an unspecied but well-behaved vector eld u.x/ that is continuously differentiable. From equation (M.143) and equation (M.144) on the preceding page we see that we can always represent such a vector eld in the following way: Z Z u.x0 / u.x0 / 3 0 3 0 u.x/ D r r dx Cr r dx 4 jx x0 j 4 jx x0 j V0 V0 (M.148)
(M.149b) (M.149c)
M .3.
Vector calculus
j 249
r urotat D 0 we notice, by invoking also equations (M.150) on the facing page, that r u D r uirrot D r . r / D r uDr u
rotat
(M.151b)
r 2 a/
(M.152a) (M.152b)
Dr
.r
from which we see that a vector eld that is well-behaved at large distances is completely and uniquely determined if we know its divergence and curl at all points x in 3D space (and at any given xed time t if the vector eld is time dependent). This is the (rst) Helmholtzs theorem, also called the fundamental theorem of vector calculus.
Products of rotational and irrotational components of vectors
A FT
(M.153a) (M.153b) r .x/ (M.154) .x/r .x/ .x/r 2 .x/ (M.155)
EXAMPLE
M .16
Derive the scalar and vector products of the rotational and irrotational components of two vector elds u.x/ and v.x) that have been Helmholtz decomposed into u D urotat C uirrot and v D v rotat C v irrot , respectively. Let us, in addition to expressions (M.149), introduce the denitions Z 1 r 0 v.x0 / .x/ D d3x 0 4 jx x0 j V0 Z 1 r 0 v.x0 / d3x 0 b.x/ D 4 jx x0 j V0 in order to simplify our notations and calculations. S CALAR PRODUCT
We can replace the RHS of this expression by using formula (F.136) on page 217 to obtain r .x/ r .x/ D r
If we insert the expression (M.149b) on the facing page for .x/ and expression (M.153a) for .x/ and then integrate over V , the rst term in the RHS can, with help of the divergence theorem, be written as a surface integral. Noticing that at large distances r D x x0 the integrand in this surface integral (the expression within square brackets) tends to zero as 1=r 3 , we conclude that this integral vanishes and that, consequently, Z d3x uirrot .x/ v irrot .x/ V Z Z Z 1 r 0 u.x0 / 1 D d3x d3x 0 d3x 0 r 0 v.x0 /r 2 jx x0 j jx x0 j 16 2 V V0 V0 (M.156)
The scalar product between the two irrotational components uirrot and v irrot can now be written uirrot .x/ v irrot .x/ D r .x/
250 j
We apply the Dirac delta function expression (M.121) on page 243 to the RHS of this equation to obtain the following non-local scalar product expression: Z Z Z r 0 u.x0 / r v.x/ 1 d3x uirrot .x/ v irrot .x/ D d3x d3x 0 4 jx x0 j V V V0 (M.157a) Z D d3x r v.x/ '.x/
V
where '.x/ D is a scalar potential . The evaluation of the scalar product of the two rotational (solenoidal) components urotat and v rotat is a little bit trickier. We start from 1 4 Z
V0
d3x 0
r 0 u.x0 / jx x0 j
(M.157b)
A FT
urotat .x/ v rotat .x/ D r a.x/ r b.x/ urotat .x/ v rotat .x/ D b r .r a/ r .r a/ b b r .r a/ D b r .r a/ r 2 a urotat .x/ v rotat .x/ D b r2a r .r a/ b Z C.x/ D
V0
(M.158)
and use formula (F.139) on page 217 to obtain
(M.159)
As the last term in the RHS will vanish when we integrate over V (divergence theorem), we focus our attention on the rst term. According to formula (F.151) on page 217 (M.160)
and, according to equation (M.149c) on page 248 and formula (F.135) on page 217, we obtain Z Z 0 1 1 v .x0 / 1 3 0 r 3 0 0 0 r aD D dx dx r v.x / r r 4 4 jx x0 j jx x0 j V0 V0 (M.161) Using formula (F.169) on page 218, formula (M.120) on page 243, and formula (F.154) on page 217, we can rewrite this as 0 0 Z I 1 r v.x0 / 1 r v .x0 / O r aD d3x 0 r D d2x 0 n0 (M.162) 4 4 jx x0 j jx x0 j V0 S0
R D
where
which tends to zero for a surface at large distances. Hence, we have found that (M.163)
Integration over V and the use of the divergence theorem, where the resulting surface integral vanishes, gives the result Z Z Z 1 r u.x/ r 0 v.x0 / d3x urotat .x/ v rotat .x/ D d3x d3x 0 4 jx x0 j V V0 (M.164a) ZV 1 D d3x r u.x/ C.x/ 4 V r 0 v .x0 / jx x0 j
d3x 0
(M.164b)
M .3.
Vector calculus
j 251
is a vector potential . The scalar product of uirrot and v rotat becomes uirrot .x/ v rotat .x/ D r .x/ r b.x/ (M.165)
With the use of a standard vector analytic identity formula (F.138) on page 217, we can rewrite this as uirrot .x/ v rotat .x/ D r b.x/ r .x/ (M.166)
If we integrate this over V and use the divergence theorem, we obtain Z

V
d xu
irrot
.x/ v
rotat
I .x/ D
S
O d2x n
b.x/
r .x/
(M.167)
Again, the surface integral vanishes and we nd that the non-local orthogonality condition Z
V
d3x uirrot .x/ v rotat .x/ D 0
holds between the irrotational and rotational components of any two arbitrary, continuously differentiable vector elds u.x/ and v .x/. If, in particlar, v D u, then Z
V
d3x uirrot .x/ urotat .x/ D 0
V ECTOR PRODUCT
The vector product between the two irrotational components uirrot and v irrot is uirrot .x/ v irrot .x/ D r .x/ r .x/ (M.170)
which can be evaluated in a similar manner as the scalar product. Using formula (F.142) on page 217 we immediately see that the result is uirrot .x/ v irrot .x/ D r .x/r .x/ (M.171)
D
Z
V
Integration over V yields, with the use of formula (F.178) on page 218, the result d3x uirrot .x/ v irrot .x/ D I
S
R A
O d2x n Z
V
where the RHS goes to zero at innity, resulting in the non-local parallelity condition d3x uirrot .x/ v irrot .x/ D 0 (M.173)
FT
(M.168) (M.169) .x/r .x/ (M.172) End of example M.16
252 j
M .4
Analytical mechanics
Lagranges equations
M .4.1
As is well known from elementary analytical mechanics, the Lagrange function or Lagrangian L is given by L.qi ; qi ; t/ D L .qi ; qi ; t/ D T P P where qi is the generalised coordinate, qi P
def
(M.174)
dqi dt
(M.175)
and the variational principle with xed endpoints t1 and t2 , S D 0
FT
t1
the generalised velocity, T the kinetic energy, and V the potential energy of a mechanical system, If we use the action Z t2 SD dt L.qi ; qi ; t/ P (M.176)
(M.177)
R A D
L0 D L C then
we nd that the Lagrangian L satises the Euler-Lagrange equations d @L @L D0 dt @qi P @qi
(M.178)
To the generalised coordinate qi one denes a canonically conjugate momentum pi according to pi D @L @qi P (M.179)
and note from equation (M.178) above that @L D pi P @qi (M.180)
If we introduce an arbitrary, continuously differentiable function D .qi ; t/ and a new Lagrangian L0 related to L in the following way @ @ d D L C qi P C dt @qi @t (M.181)
@L0 @L @ D C @qi P @qi P @q
(M.182a)
M .4.
Analytical mechanics
j 253
@L d @ @L0 D C @qi @qi dt @q Or, in other words, d @L0 dt @qi P where pi0 D and qi0 D @L0 @L @ @ D C D pi C @qi P @qi P @qi @qi @L0 @L D D qi @pi P @p P @L0 d @L D @qi dt @qi P @L @qi
(M.182b)
(M.183)
(M.184a)
M .4.2
Hamiltons equations
From L, the Hamiltonian (Hamilton function) H can be dened via the Legendre transformation H.pi ; qi ; t / D pi qi P L.qi ; qi ; t/ P
After differentiating the left and right hand sides of this denition and setting them equal we obtain @H @H @H dpi C dqi C dt D qi dpi C pi dqi P P @pi @qi @t @L dqi @qi @L dqi P @qi P @L dt @t (M.186)
According to the denition of pi , equation (M.179) on the preceding page, the second and fourth terms on the right hand side cancel. Furthermore, noting that according to equation (M.180) on the facing page the third term on the right hand side of equation (M.186) above is equal to pi dqi and identifying terms, P we obtain the Hamilton equations: @H dqi D qi D P @pi dt @H D pi D P @qi (M.187a) dpi dt (M.187b)
A FT
(M.184b) (M.185)
254 j
M .5
Bibliography
[62] M. A BRAMOWITZ AND I. A. S TEGUN, Handbook of Mathematical Functions, Dover Publications, Inc., New York, 1972, Tenth Printing, with corrections. [63] G. B. A RFKEN AND H. J. W EBER, Mathematical Methods for Physicists, fourth, international ed., Academic Press, Inc., San Diego, CA . . . , 1995, ISBN 0-12-0598167. [64] R. A. D EAN, Elements of Abstract Algebra, John Wiley & Sons, Inc., New York, NY . . . , 1967, ISBN 0-471-20452-8. [65] A. A. E VETT, Permutation symbol approach to elementary vector analysis, American Journal of Physics, 34 (1965), pp. 503507.
[67] P. M. M ORSE AND H. F ESHBACH, Methods of Theoretical Physics, Part I. McGrawHill Book Company, Inc., New York, NY . . . , 1953, ISBN 07-043316-8. [68] B. S PAIN, Tensor Calculus, third ed., Oliver and Boyd, Ltd., Edinburgh and London, 1965, ISBN 05-001331-9. [69] W. E. T HIRRING, Classical Mathematical Physics, Springer-Verlag, New York, Vienna, 1997, ISBN 0-387-94843-0.
R A
FT
[66] A. M ESSIAH, Quantum Mechanics, vol. II, North-Holland Publishing Co., Amsterdam, 1970, Sixth printing.
INDEX
L Cerenkov, Pavel Alekseevich, 190 acceleration eld, 125 advanced time, 44 Ampres law, 6 Ampre, Andr-Marie, 6 Ampre-turn density, 184 angular frequency, 30, 42 angular momentum commutation rule, 228 angular momentum theorem, 73 anisotropic, 186 anisotropic medium, 183 anomalous dispersion, 187 antecedent, 236 antenna, 111 antenna current, 111 antenna feed point, 112 antisymmetric tensor, 160 arrow of time, 57 associated Legendre polynomial of the rst kind, 103 associative, 152 axial gauge, 51 axial vector, 161, 235 axiomatic foundation of classical electrodynamics, 21 azimuthal phase, 79 bac-cab rule, 236 Barbieri, Cesare, xviii Bessel functions, 117 birefringent, 186 birefringent crystal, 183 blueshift, 71 Bohm, David Joseph, 145 braking radiation, 134 bremsstrahlung, 134, 141 chirality density, 79 chirality ow, 79 circular wave polarisation, 29 Classical Electrodynamics, 1, 10 closed algebraic structure, 152 coherent radiation, 140 collision frequency, 194 collisional interaction, 200 complete -Lorenz gauge, 46 complex conjugate, 210 complex notation, 22, 223 complex vector eld, 225 complex-eld six-vector, 24 component notation, 223 concentration, 242 conductivity, 192 conductivity tensor, 12 consequent, 236 conservation law, 61 conservation law for angular momentum, 72 conservation law for linear momentum, 67 conservation law for the total current, 62 conservation laws, 60 conservative eld, 12 conservative forces, 172 conserved quantities, 60 constants of motion, 60 constitutive relations, 16, 192 continuity equation, 60 contravariant component form, 148, 224 contravariant eld tensor, 161 contravariant four-tensor eld, 231 contravariant four-vector, 226 contravariant four-vector eld, 151 contravariant vector, 148 control sphere, 60 convective derivative, 14 coordinate four-vector, 226 coordinate vector, 223 corrsepondence principle, 50 cosine integral, 115 Coulomb gauge, 44 Coulombs law, 2 coupled differential equations, 27 covariant, 146 covariant component form, 224 covariant eld tensor, 161
Carozzi, Tobia, xx canonically conjugate four-momentum, 168 canonically conjugate momentum, 168, 252 canonically conjugate momentum density, 175 centre of energy, 79 CGS units, 3, 15 characteristic impedance of vacuum, 97, 196 charge conjugation, 55 Cherenkov radiation, 188 chiral media, 181
R
255
A FT
256
INDEX
covariant four-tensor eld, 231 covariant four-vector, 227 covariant four-vector eld, 151 covariant gauge, 157 covariant vector, 148 CPT theorem, 57 cross product, 235 Curie, Marie Sklodowska, 190 Curie, Pierre, 16 curl, 241 cutoff, 135 cycle average, 24 cyclotron radiation, 138, 141 dAlembert operator, 28, 41, 156, 242 de Coulomb, Charles-Augustin, 2 de Coulomb, Charles-Augustine, 2 deniendum, 223 deniens, 223 del operator, 239 del squared, 242 demodulation, 183 dielectric permittivity, 182 differential distance, 150 differential vector operator, 239 diffusion coefcient, 202 dipole antennas, 111 Dirac delta, 243 Diracs symmetrised Maxwell equations, 16 Dirac, Paul Adrien Maurice, 1, 44, 221 direct product, 236 dispersive, 187 dispersive property, 183 displacement current, 11 divergence, 241 divergence-less, 248 dot product, 233 dual electromagnetic tensor, 162 dual vector, 148 duality transformation, 57, 163 Duhem, Pierre Maurice Marie, 2 dummy index, 148 dyad, 236 dyadic product, 236 dyons, 58 E1 radiation, 107 E2 radiation, 111 Einsteins summation convention, 223 Einstein, Albert, 1, 3
electric and magnetic eld energy, 64 electric charge conservation law, 10 electric charge density, 4 electric conductivity, 12 electric current density, 7 electric dipole moment, 106 electric dipole moment vector, 100 electric dipole radiation, 107 electric displacement current, 18 electric displacement vector, 181, 182 electric eld, 3, 182 electric eld energy, 64 electric monopole moment, 100 electric permittivity, 200 electric polarisation, 100 electric quadrupole moment tensor, 100 electric quadrupole radiation, 111 electric quadrupole tensor, 110 electric susceptibility, 183 Electricity, 1 electricity, 2 electrodynamic potentials, 39 electromagnetic angular momentum current density, 72, 76 electromagnetic angular momentum density, 72, 108 electromagnetic angular momentum ux tensor, 72 electromagnetic energy current density, 62 electromagnetic energy ux, 62 electromagnetic eld energy, 63 electromagnetic eld energy density, 26, 62 electromagnetic eld tensor, 161 electromagnetic linear momentum current density, 67, 70 electromagnetic linear momentum density, 26, 67, 107 electromagnetic linear momentum ux tensor, 67 electromagnetic moment of momentum density, 72 electromagnetic orbital angular momentum, 75 electromagnetic scalar potential, 40 electromagnetic spin angular momentum, 75 electromagnetic vector potential, 39 electromagnetic virial density, 79 electromagnetic virial theorem, 79 electromagnetically anisotropic media, 12 Electromagnetism, 1 electromagnetodynamic equations, 16 Electromagnetodynamics, 2 electromagnetodynamics, 58
A FT
INDEX
257
far eld, 87, 89 far elds, 81, 94 far zone, 81, 90, 104 Faraday tensor, 161 Faradays law, 13 Faradays law of induction, 10 Faraday, Michael, 221 Feynman, Richard Phillips, 1, 44 eld, 224 eld Lagrange density, 176 eld momentum, 25 eld point, 4 eld quantum, 135 ne structure constant, 136, 143 four-current, 156 four-del operator, 239 four-dimensional Hamilton equations, 168 four-dimensional vector space, 147 four-divergence, 241 four-gradient, 240 four-Hamiltonian, 168 four-Lagrangian, 166
A FT
Galilei, Galileo, 221 Galileos law, 145 gauge xing, 51 gauge function, 47 gauge invariant, 47 gauge theory, 48 gauge transformation, 47 gauge transformation of the rst kind, 51 gauge transformation of the second kind, 51 Gausss law of electrostatics, 5 general inhomogeneous wave equations, 40 general theory of relativity, 145 generalised coordinate, 168, 252 generalised four-coordinate, 168 generalised velocity, 252 Gibbs notation, 239 Ginzburg, Vitaliy Lazarevich, xx, 190 Glashow, Sheldon, 1 gradient, 240 Green function, 42, 103 Green functions, 245 group theory, 152 group velocity, 187 Hall effect, 185 Hall, Edwin Herbert, 185 Hamilton density, 175 Hamilton density equations, 175 Hamilton equations, 168, 253
electromotive force, 12 electrostatic dipole moment vector, 37 electrostatic quadrupole moment tensor, 38 electrostatic scalar potential, 36 electrostatics, 2 Electroweak Theory, 1 elliptical wave polarisation, 29 EMF, 12 energy density balance equation, 63 energy theorem in Maxwells theory, 64 equation of continuity, 157 equation of continuity for electric charge, 10 equation of continuity for magnetic charge, 16 equations of classical electrostatics, 9 equations of classical magnetostatics, 10 Eriksson, Anders, xx Eriksson, Marcus, xviii Erukhimov, Lev Mikahilovich, xx Euclidean space, 153 Euclidean vector space, 149 Eulers rst law, 66 Euler, Leonhard, 71 Euler-Lagrange equation, 174 Euler-Lagrange equations, 175, 252 Euler-Mascheroni constant, 115 event, 152
four-momentum, 155 four-potential, 157 four-scalar, 224 four-tensor elds, 231 four-vector, 151, 226 four-velocity, 155 Fourier amplitude, 30 Fourier integral, 30 Fourier transform, 31, 41 Frank, Ilya Mikhailovich, 190 Franklin, Benjamin, 61 free-free radiation, 134 frequency conversion, 12 frequency mixing, 12 Frman, Per Olof, xix fully antisymmetric tensor, 227 functional derivative, 174 fundamental tensor, 147, 224, 231 fundamental theorem of vector calculus, 249 Fldt, Gran, xx
258
INDEX
identity element, 152 in a medium, 189 incoherent radiation, 141 indenite norm, 149 index contraction, 148 index lowering, 148 index of refraction, 182 inertial reference frame, 145 inertial system, 145 inhomogeneous Helmholtz equation, 42 inhomogeneous time-independent wave equation, 42 inhomogeneous wave equation, 41 inner product, 233 instantaneous, 131 interaction Lagrange density, 176 intermediate eld, 87 intermediate zone, 81 invariant, 224 invariant line element, 150 inverse element, 152 inverse Fourier transform, 31 ionosphere, 194 irrotational, 5, 241, 244, 248 Jacobi identity, 163 Jarlskog, Cecilia, xx
A FT
Hamilton function, 253 Hamilton gauge, 51 Hamilton operator, 49, 239 Hamilton, William Rowen, 239 Hamiltonian, 49, 253 Heaviside, Oliver, 128, 186, 190 Heaviside-Larmor-Rainich transformation, 57 Heaviside-Lorentz units, 15 helical base vectors, 29, 226 Helmholtz decomposition, 248 Helmholtz equation, 112 Helmholtzs theorem, 5, 33, 41, 249 help vector, 104 Hermitian conjugate, 210 Hertz vector, 102 Hertzs method, 99 Hertz, Heinrich Rudolf, 95 heterodyning, 183 Hodge star operator, 58 homogeneous vector wave equations, 195 Hookes law, 172 Huygenss principle, 42
Jemenko equations, 88 Joule heat power, 64 Karlsson, Roger, xx Kelvin function, 142 Kerr effect, 183 kinetic energy, 172, 252 kinetic momentum, 171 Kirchhoff gauge, 46 Kirchhoff, Gustav Robert, 46 Kohlrausch, Rudolf, 3 Kopka, Helmut, xx Kronecker delta tensor, 227 Kronecker product, 236 Lagrange density, 173 Lagrange function, 172, 252 Lagranges formula, 236 Lagrangian, 172, 252 Laplace operator, 241 Laplacian, 241 Larmor formula for radiated power, 131 law of inertia, 145 Learned, John, xx left-hand circular polarisation, 29, 77 Legendre polynomial, 103 Legendre transformation, 253 Levi-Civita tensor, 227 light cone, 151 light-like interval, 151 Lindberg, Johan, xviii line broadening, 30 line element, 234 linear mass density, 173 linear momentum density, 67 linear momentum operator, 25, 69 linear momentum theorem in Maxwells theory, 68 linear wave polarisation, 29 Linard, Alfred-Marie, 120 Linard-Wiechert potentials, 120, 160 local gauge transformation, 51 longitudinal component, 197 loop antenna, 115 Lorentz boost parameter, 153 Lorentz force, 15, 68 Lorentz force density, 66 Lorentz power, 63 Lorentz power density, 63 Lorentz space, 149, 224 Lorentz torque, 73
INDEX
259
Lorentz torque density, 72 Lorentz transformation, 147 Lorentz, Hendrik Antoon, 1, 41 Lorenz, Ludvig Valentin, 41 Lorenz-Lorentz gauge, 48 Lorenz-Lorentz gauge condition, 41, 157 lowering of index, 231 Lundborg, Bengt, xx M1 radiation, 110 Mach cone, 190 magnetic charge density, 16 magnetic current density, 16 magnetic dipole moment, 110, 184 magnetic dipole moment per unit volume, 184 magnetic dipole radiation, 110 magnetic displacement current, 18 magnetic eld, 7, 182 magnetic eld energy, 64 magnetic eld intensity, 184 magnetic ux, 13 magnetic ux density, 8 magnetic four-current, 163 magnetic induction, 8 magnetic monopoles, 16 magnetic permeability, 200 magnetic susceptibility, 182, 185 magnetisation, 184 magnetisation currents, 184 magnetised plasma, 12, 183 magnetising eld, 181, 182, 184 Magnetism, 1 magnetostatic vector potential, 36 magnetostatics, 6 Majorana formalism, 25 Majorana representation, 25 Majorana, Ettore, 25 mass density, 67 massive photons, 179 material derivative, 14 mathematical group, 152 matrix representation, 227 Maxwell stress tensor, 67, 70 Maxwells displacement current, 10 Maxwells macroscopic equations, 185 Maxwells microscopic equations, 16 Maxwell, James Clerk, xix, 1, 21 Maxwell-Lorentz equations, 16 Maxwell-Lorentz source equations, 22
A FT
natural units, 15 near zone, 81, 87 negative helicity, 29 negative refractive index, 186 Newtons rst law, 145 Newtons second law, 66 Newton, Sir Isaac, 221 Newton-Lorentz force equation, 167 Noethers theorem, 60 Noether, Amalie Emmy, 60 non-Euclidean space, 149 non-linear effects, 12 norm, 148, 235 null vector, 151 observation point, 4 Ohms law, 11, 192 Ohmic losses, 64 one-dimensional wave equation, 197 orbital angular momentum operator, 74, 78 orthogonality condition, 251 outer product, 236
mechanical angular momentum, 71 mechanical angular momentum density, 72 mechanical energy, 63 mechanical energy density, 63 mechanical Lagrange density, 176 mechanical linear momentum, 71 mechanical linear momentum density, 66 mechanical moment of momentum, 71 mechanical orbital angular momentum, 71 mechanical spin angular momentum, 71 mechanical torque, 71 metamaterials, 186 metric, 224, 234 metric tensor, 147, 224, 231 minimal coupling, 50 Minkowski equation, 167 Minkowski space, 153 mixed four-tensor eld, 231 mixing angle, 58 moment of velocity, 61 monad, 239 monochromatic, 82 monochromatic wave, 30 multipolar gauge, 51 multipole expansion, 99 Mller scattering, 143
260
INDEX
Palmer, Kristoffer, vi, xviii parallelity condition, 251 paraxial approximation, 91 parity transformation, 55, 235 Parsevals identity, 135, 142 Pauli, Wolfgang, 51 Peierls, Sir Rudolf, 222 phase velocity, 186 photon, 135, 202 photons, 181 physical observable, 22, 147 Planck constant, 25 Planck units, 15 plane wave, 28, 198 plasma, 187 plasma frequency, 187, 194 plasma physics, 79 Poincar gauge, 51 Poissons equation, 35 polar vector, 161, 235 polarisation charges, 182 polarisation currents, 183 polarisation potential, 102 polarisation vector, 102 position four-vector, 147, 226 position vector, 223 positive denite, 153 positive denite norm, 149 positive helicity, 29 postfactor, 237 postulates, 21 potential energy, 172, 252 potential theory, 102 Poynting vector, 26, 62 Poyntings theorem, 64 prefactor, 237 probability density, 49 Proca Lagrangian, 179 propagation vector, 28 propagator, 42, 103 propagators, 245 proper time, 150 propositional calculus, 221 pseudo-Riemannian space, 153 pseudoscalar, 222 pseudoscalars, 235 pseudotensor, 222 pseudotensors, 235 pseudovector, 56, 160, 222, 235
QCD, 1 QED, 1, 181 quadratic differential form, 150, 234 Quantum Chromodynamics, 1 Quantum Electrodynamics, 1 quantum electrodynamics, 44, 181 quantum mechanical non-linearity, 4 radial gauge, 51 radiated electromagnetic power, 64 radiation eld, 125 radiation gauge, 46 radiation resistance, 114 radius four-vector, 147, 226 radius vector, 210, 223 raising of index, 231 rank, 227 rapidity, 153 real vector eld, 225 reciprocal space, 33 redshift, 71 reduced Planck constant, 25 refractive index, 182, 186, 194 relative dielectric permittivity, 183 relative permeability, 185, 200 relative permittivity, 200 Relativity principle, 145 relaxation time, 195 rest mass density, 176 retarded Coulomb eld, 87 retarded induction eld, 89 retarded relative distance, 120 retarded time, 44 Riemann-Silberstein vector, 24 Riemannian metric, 150 Riemannian space, 147, 224 right-hand circular polarisation, 29, 77 Rohrlich, Fritz, 1, 21 rotational, 248 rotational degree of freedom, 99 Salam, Abdus, 1 scalar, 222, 241 scalar eld, 151, 224 scalar potential, 250 scalar product, 233 Schrdinger equation, 49 Schwinger, Julian Seymour, 1, 16 self-force effects, 193 shock front, 190
A FT
INDEX
261
SI units, 3, 15 signature, 148, 232 simultaneous coordinate, 128 Sjholm, Johan, vi, xviii skew-symmetric, 161 skin depth, 200 solenoidal, 244, 248 Sommerfeld, Arnold Johannes Wilhelm, 120, 190 source equations, 185 source point, 4 source terms, 22 space components, 149 space inversion, 55 space-like interval, 151 space-time, 149 spatial dispersive media, 12 spatial Fourier components, 31 spatial spectral components, 31 special theory of relativity, 145 spectral energy density, 98 spherical Bessel function of the rst kind, 103 spherical Hankel function of the rst kind, 103 spherical harmonic, 103 spherical waves, 94 spin angular momentum operator, 74 standard conguration, 146 standing wave, 112 static net charge, 37 Strandberg, Bruno, xviii super-potential, 102 superposition principle, 30 synchrotron radiation, 138, 141 synchrotron radiation lobe width, 139 tHooft, Gerardus, xx Tamburini, Fabrizio, xviii Tamm, Igor Evgenevich, 190 telegraphers equation, 197, 200 temporal average, 24 temporal dispersive media, 12 temporal Fourier components, 30 temporal Fourier series, 30 temporal gauge, 51 temporal spectral components, 30 tensor, 222 tensor eld, 227 tensor notation, 228 tensor product, 236 Then, Holger, xviii
A FT
vacuum permeability, 6 vacuum permittivity, 3 vacuum polarisation effects, 4 vacuum wave number, 196 vacuum wavelength, 42 variational principle, 252 Vavilov, Sergey Ivanovich, 190 L Vavilov-Cerenkov cone, 190 L Vavilov-Cerenkov radiation, 188, 190 vector, 222 vector potential, 251 vector product, 235 vector wave equations, 28 vector waves, 28 velocity eld, 125 velocity gauge condition, 46 virtual simultaneous coordinate, 125
thermodynamic entropy, 57 three-dimensional functional derivative, 175 time arrow of radiation, 95 time component, 149 time reversal, 55 time-dependent diffusion equation, 202 time-dependent Poissons equation, 45 time-harmonic wave, 30 time-independent diffusion equation, 196 time-independent telegraphers equation, 198 time-independent wave equation, 30, 112, 196 time-like interval, 151 Tomonaga, Sin-Itiro, 1 total charge, 100 total electromagnetic angular momentum, 108 total electromagnetic linear momentum, 108 total mechanical angular momentum, 71 traceless, 39 translational degree of freedom, 96 translational Doppler shift, 71 transversality condition, 25 transverse components, 197 transverse gauge, 46 Truesdell III, Clifford Ambrose, 71 tryad, 239 uncertainty principle, 57 uncoupled inhomogeneous wave equations, 41 unit dyad, 238 unit tensor, 238 universal constant, 157 unmagnetised plasma, 193
262
INDEX
von Helmholtz, Hermann Ludwig Ferdinand, 5 Wahlberg, Christer, xx Waldenvik, Mattias, xx wave equation, 62 wave equations, 21, 27 wave number, 42 wave packets, 28 wave polarisation, 28 wave vector, 25, 32, 186, 198 wavepacket, 82 Webers constant, 3 Weber, Wilhelm Eduard, 3 Weinberg, Steven, 1, 221 Weyl gauge, 51 Wheeler, John Archibald, 44 Whitehead, Alfred North, 147 Wiechert, Emil Johann, 120 world line, 152 Yang-Mills theory, 48 Yngve, Staffan, xviii Youngs modulus, 173 Yukawa meson eld, 179 rsted, Hans Christian, 6
A FT
A FT
ELECTROMAGNETIC FIELD THEORY ISBN 978-0-486-4773-2
A FT

EMFT Book

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

EMFT Book

Caricato da

Copyright:

Formati disponibili

Draft version released 24th June 2012 at 12:20 CETDownloaded from http://www.plasma.uu.se/CED/Book Sheet: 1 of 284 .

ELECTROMAGNETIC FIELD THEORY

ELECTROMAGNETIC FIELD THEORY

ELECTROMAGNETIC FIELD THEORY EXERCISES

Copyright 19972009 by Bo Thid Uppsala, Sweden All rights reserved.

Electromagnetic Field Theory ISBN 978-0-486-4773-2

Freely downloadable from www.plasma.uu.se/CED

2 Electromagnetic Fields and Waves

3 Electromagnetic Potentials and Gauges

44 46 46 51 53 55 55 57 60 61 61 62 66 71 79 79 80 81 82 84 87 90 94 95 96 96 97 98 99 99 101 106 109 110 111 111 118 119

4 Fundamental Properties of the Electromagnetic Field

5 Fields from Arbitrary Charge and Current Distributions

6 Radiation and Radiating Systems

5.1 5.2 5.3 5.4

8 Electromagnetic Fields and Particles

9 Electromagnetic Fields and Matter

Hamiltons equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 M .5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254

1.1 1.2 1.3 1.4

Tetrahedron-like volume element of matter . . . . . . . . . . . . . . . . . . . . . . . 229

PREFACE TO THE SECOND EDITION

P R E FAC E TO THE SECOND EDITION

PREFACE TO THE FIRST EDITION

P R E FACE TO THE FIRST EDITION

FOUNDATIONS OF CLASSICAL ELECTRODYNAMICS

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

1.1.2 The electrostatic eld

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

discrete charges, in which case tions:9

is expressed in terms of Dirac delta distribuX

Since, by denition, the integral Z d3x 0 .x0 x0 / i V0 Z 0 d3x 0 .x 0 xi /

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

representing four scalar partial differential equations.

1.2.1 Ampres law

1.2.2 The magnetostatic eld

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

since, according to formula (F.154) on page 217, r .r vector eld a.x/.

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

and the equations of classical magnetostatics r Bstat .x/ D 0 r B

1.3.1 The indestructibility of electric charge

1.3.2 Maxwells displacement current

1.3.3 Electromotive force

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

The electromotive force is dened as I E D dl .Estat C Eemf /

where dl is a tangential line element of the closed loop C .15

1.3.4 Faradays law of induction

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

corresponding to an effective electric eld in the loop (moving observer) Eemf D E C v B

Hence, we conclude that for a stationary observer, the Maxwell equation r ED @B @t

In CGS units the microscopic Maxwell equations are r ED4 1 @B D0 c @t 4 1 @E D j c @t c

is indeed valid even if the loop is moving.

1.3.5 The microscopic Maxwell equations

in Heaviside-Lorentz units (one of several natural units) r ED r BD0 r r EC B 1 @B D0 c @t 1 1 @E D j c @t c

and in Planck units (another set of natural units) r ED4 r BD0 r r EC B @B D0 @t @E D4 j @t

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

1.3.6 Diracs symmetrised Maxwell equations

1 . F OU N DATIONS OF CLASSICAL ELECTRODYNAMICS

d3x 0 j m .t; x0 /.x x0 / Z 1 0 @ d3x 0 m .t; x0 /r 0 4 @t V 0 jx x0 j @ m B.t; x/ 0 j .t; x/ @t

[2] [3] [4] [5]

End of example 1.1

ELECTROMAGNETIC FIELDS AND WAVES

2.1 Axiomatic classical electrodynamics

1 In 1864, in a lecture at the Royal Society of London, J A M E S C L E R K M A X W E L L (18311879) himself said:

(2.1a) (2.1b) (2.1c) (2.1d)