I V Savelyev Fundametals of Theoretical Physics Vol 1 PDF

I. V.
Savelyev
FUNDAMENTALS
OF THEORETICAL
PHYSICS
Volume 1
Mechanics
Electrodynamics
Mir Publishers Moscow

1;1. B. CaeenbeB
OCHOBbl TEOPETI;14ECKO~ ~1;13HKH
ToM 1
MEXAHHKA
811EKTPOAHHAMHKA
f.13P,A TEllbCTBO «HAYKA•

FUNDAMENTALS
OF THEORETICAL
PHYSICS
Volume1
Mechanics
Electrodynamics
Translated from the Russian
by
G. Leib
MIR PUBLISHERS MOSCOW

First published 1982
Revised from the 1975 Russian edition
TO THE READER
Mir Publishers would be grateful for your comments on the

content, translation and design of this book. We would also be
pleased to receive any other suggestions you may wish to make.
Our address is:
Mir Publishers
2 Pervy Rizhsky Pereulok
I-110, GSP, Moscow, 129820
USSR
..- ...... _--~-~~-: :;:,·r\_ ,.E~

~
qH ..... T. ' • '
HAY ' ' · . catet
1\am•w· ' •: , 1-\ H A
u~. B \:. · ....-
© ll3AaTeJibCTBo <<HayRa>>, :;;;--·~{{9 {3 ~H

@ English translation, Mir Publishers, 1982
PREFACE
The bonk being offered to the reader is a logical continuation of

the author's three-volume general course of physics. Everything
possible has been done to avoid repenting what has been sel out
in the three-volume course. Particularly. the experiments underlying
the advancing of physical ideas are not treated, and some of the re-
sults obtained are not discussed.
In the part devoted to mechanics, unlike the established tradi-
tions, Lagrange's equations are derived directly from Newton's equa-
tions instead of from d 'Alembert' s principle. Among the books I have
acquainted myself with, such a derivation is given in A. S. Kompa-
neits's book Theoretical Physics (in Russian) for the particular case
of a conservative system. In the present book, I have extended this
method of exposition to systems in which not only conservative,
but also non-conservative forces act.
The treatment of electrodynamics is restricted to a consideration
of media with a permittivity c and a permeability ~t not depending
on the fields E and B.
Sections 40 and 69 devoted to the energy-momentum tensor are
appreciably more complicated. They have been inelndccl in the book
becnuse they con tnin an excellent illustra timt of ho\v LngTnngian
formalism is generalized for non-mechanical systems. A reader to
whom these sections will seem too difficult may omit them without
any harm to his understanding the remaining sections of the book.
I have devoted much attention to the vnriational principl<>. with
the consistent use of the following procedure-first the required resnlt
is obtnined with tlte aid of mrthods which the render is alt·endr nc-
quuintPtl with, and then the ~ame n'sult is obtained u~ing the varia-
tional principle. The object here was to ensure the reader treating
the variational prlneiple as a quite reliable and powerful means of
research.
An appreciable difflculty appearing in studying theoretical physics
is the circumstance that quite often many malhematical topics have
eitl1er never been studied by the reader or have been forgotten by
him fundamentally. To eliminate this difficulty, I have provided
the book with detailed mathematical appendices. The latter are
sufficiently complete to relieve the reader of having to tum to
6 PREFACE
mathematical aids and find the required information in them. This

information is often set out in these aids too complicated for the
readers which the present book is intended for. Hence, the informa-
tion on mathematical analysis contained in a college course of
higher mathematics is sufficient for mastering this book.
The book has been conceived as a training aid for students of non-
theoretical specialities of higher educational institutions. I had
in mind readers who would like to grasp the main ideas and methods
of theoretical physics without delving into the details that are of
interest only for a specialist. This book will be helpful for physics
instructors at higher schools, and also for everyone interested in
the subject but having no time to become acquainted with it (or re-
store it in his memory) according to fundamental manuals.
Igor Sauelyev
CONTENTS
Preface 5
Part One. Mechanics 11
Chapter I. The Variational Principle in Mechanics 11

1. Introduction 11
2. Constraints 13
3. Equations of Motion in Cartesian Coordinates 16
4. Lagrange's Equations in Generalized Coordinates 19
5. The' Lagrangian and Energy 24
6. Examples of Compiling Lagrange's Equations 28
7. Principle of Least Action 33
Chapter II. Conservation Laws 36

8. Energy Conservation 36
9. Momentum Conservation 37
10. Angular Momentum Conservation 39
Chapter III. Selected Problems in Mechanics 41

11. Motion of a Particle in a Central Force Field 41
12. Two-Body Problem 45
13. Elastic Collisions of Particles 49
14. Particle Scattering 53
15. Motion in Non-Inertial Reference Frames 57
Chapter IV. Small-Amplitude Oscillations 64

16. Free Oscillations of a System Without Friction 64
17. Damped Oscillations 66
18. Forced Oscillations 70
8 CONTENTS
l 9. Oscillations of a System with Many Deg1·ees of Freedom 72

20. Coupled Pendulums 77
Chapter V. Mechanics of a Rigid Body 82

21. Kinematics of a Rigid Body 82
22. The Euler Angles 85
23. The lncrlia Tensor 88
24. Angular Momentum of u Higid Body 95
25. Free Axes of Rotation 99
26. Equation of Motion of u Rigid Body 101
27. Euler's Equations 105
28. Free Symmetric Top 107
29. Symmetric Top in a Homogeneous Gravitational Field 111
Chapter VI. Canonical Equations 115

30. Hamilton's Equations 115
31. Poisson Brackets 119,
32. The Hamilton-Jacobi Equation 121
Chapter VII. The Special Theory of Relativity 125

33. The Principle of Relativity 125
34. Interval 127
35. Lorentz Transformations 130
36. Four-Dimensional Velocity and Acceleration 134
37. Relativistic Dynamics 136
38. Momentum and Energy of a Particle 139
39. Action for a Relativistic Particle 143
40. Energy-Momentum Tensor 147
Pari Two. Electrodynamics 157
Chapter VIII. Electrostatics 157

41. Electrostatic Field in a Vacuum 157
42. Poisson's Equation 159
43. Expansion of a Field in Multipoles 161
44. Field in Dielectrics 166
45. Desq:iption of the Field in Dielectrics 170
46. Field in Anisotropic Dielectrics 175
Chapter IX. Magnetostatics 177

47. Stationary Magnetic Field in a Vacuum 177
48. Poisson's Equation for the Vector Potential 179
49. Field of Solenoid 182
50. The Biot-Savart Law 186
51. Magnetic Moment 188
52. Field in Magnetics 194
Chapter X. Time-Varying Electromagnetic Field 199

53. Law of Electromagnetic Induction 199
CONTENTS
54. Displacement Current 200

55. Maxwell's Equations 201 '
56. Potentials of Electromagnetic Field 203
57. D'Alembert's Equation 207
58. Density and Flux of Electromag-netic Field Energy 208
59. Momentum of Electromagnetic Field 211
Chapter X I. Equations ol" Eleetrodynarnit's in thP For11·-llinrcn"ional

Form 216
60. Four-Potential 2'16
61. Electromagnetic Field Tensor 219
fi2.Field Transformation Formulas 222
63. Field Invarinnt~ 225
u4. Maxwell's Equations in the Four- Dimen~ional Form 228
65. Equation of Motion of a Particle in a Field 230
Chapter XII. The Variational Principle in Electrodynamics 232

66. Action for a Charged Particle in an Electromagnetic Field 232
u7. Action for an Electromagnetic Field 234
G8. Derivation of Maxwell's Equations from the Principle of Least
Action 237
Gil. Energy-l\lornentnm Trn~ot· of an Electromagnrtic Firld 239
70. A Ch<trged Particle in an Electromagnetic Field 21,1,
Chapter XIII. Elrctromagnetic Waves 248

71. The Wave Equation 248
72. A Plane Electromagnetic Wave in a Homogeneous and Isotropic
Medium 250
7~. A Monochromatic Plane Wave 255
74. A Plane Monochromatic Wave in a Conducting 1\Iedium 260
75. Non-Monochromatic Waves 265
Chapter XIV. Hadiation of Electromagnetic Waves 269

7u. Hetnnled Potentials 269
77. Field of a Uniformly Moving Charge 272
78. Field of an Arbitrarily Moving Charge 276
79. Field Produced by a Sy~tem of Charges at Great Distances 283:
80. Dipole Hadiation 288
81. Magnetic Dipole and Quadrupole Hadiations 291
Appendices 297
I. Lagrange's Equations for a Holonomic System with Ideal Non-
Stationary Constraints 297
II. Euler's Theorem for Homogeneous Functions 299
III. Some Information from the Calculus of Variations 300
IV. Conics 309
V. Linear Differential Equations with Constant Coefficients 313
VI. Vectors 316
VII. Matrices 330
VI II. Determinants 338
IX. Quadratic Forms 347
10 CONTENTS
X. Tensors 355
XI. Basic Concepts of Vector Analysis 370
XII. Four-Dimensional Vectors and Tensors in Pseudo-Euclidean
Space 393
XIII. The Dirac Delta Function 412
XIV. The Fourier Series and Integral 413
Index 419
Part One
MECHANICS
Chapter I
THE VARIATIONAL PRINCIPLE

IN MECHANICS
1. Introduction
Depending on the nature of the objects being studied, mechanics

can be divided into particle mechanics, mechanics of a rigid body 1 ,
and continuum mechanics. The latter, in turn, is divided into hydro-
dynamics, gas dynamics, the theory of elasticity, the theory of plas-
ticity, and so on.
A continuum (continuous medium) is the most difficult object
for studying in mechanics because it is a system with an infinitely
great number of degrees of freedom. Besides, methods and equations
of thermodynamics, electrodynamics, etc. are used in solving a num-
her of problems treated by continuum mechanics in addition to
those of theoretical mechanics. These circumstances are why contin-
uum mechanics is the most complicated branch of mechanics. We
shall not deal with topics on continuum mechanics in this book.
In the general course of physics, problems of mechanics are solved
with the aid of Newton's equations. In this ehapter, we shnll acqunint
ourselves with a different approach to a description and the studying
of the motion of mechanical systems. By a mechanical system, we
shall understand a collection of point particles whose motion may
either be free or restricted by constraints. Particularly, a collection
of point particles joined by rigid constraints forms a rigid body. In
the following, for brevity's sake, we shall call point particles simply
particles.
In accordance with the approach mentioned above, a function of
generalized coordinates and generalized velocities of a system, and
abo of time, namely,
L = L (coordinates, velocities, time)
called a Lagrangian function or a Lagrangian is associated with each
mechanical system. Generalized coordinates qk are defined to be any
1 What is meant is a perfectly rigid body.

12 MECHANICS
quantities by means of which the position of a system in space can

be set. Generalized velocities qk are defined to be the time derivatives
of the generalized coordinates.
The Lagrangian can be used to characterize not only systems with
a finite number of degrees of freedom, but also systems with an
infinite number of them-continuous media and electromagnetic and
other physical fields. Thus, the significance of the Lagrangian extends
beyond the scope of classical mechanics.
Having established the form of the Lagrangian for the mechanical
system being considered, we can describe the motion of the system
with the aid of equations relating the partial derivatives of the func-
tion L with respect to the coordinates and velocities. These equations.
known as Lagrange's equations, replace Newton's equations.
The use of Lagrange's equations instead of Newton's equations
has the advantage that the number of the former equals the number
6£ degrees of freedom of the system, which, when constraints restrict-
ing the motion of the system are present, will be less than the
triple number of particles in the system. The number of Newton's
equations needed to describe a system of N particles, on the other
hand, is 3N. In addition, Lagrange's equations do not include the
reactions of the constraints\ which are unknown beforehand. Con-
sequently, when using Lagrange's equations, the reactions of the
c'onstraints are automatically excluded from consideration, and this
noticeably simplifies the solution of the relevant problem. True, the
solution in this case gives information only on the motion of the
system, the values of the reactions remaining unestablished. But in
the majority of physical problems, the values of the reactions are
of no interest, so that the data obtained by the method of Lagrange's
·equations are:quite sufficient. We can indicate as an example the prob-
lem of the oscillations of a simple pendulum (Fig. 1.1). The equation
of Newton's second law for the particle m has the form
'{ f\C
···.···.·.,,
~· ' .. . . . .... .,,... mr = mgA-'+1 R) 'f''
1 ~A.J.-
tv\
where 'A is the reaction of .the thread. Projecting all the vectors onto
ax~ x. y and:: (the :-axi~ i~ d~ted beyond the drawing). we obtain
~~ax equa.~:
m""x. = R:u my= mg + R,,, m~- = 0
If we characterize the position of the system by the generalized

coordinate
,_ .
cp, one equation will be sufficient instead of three, namely.
ml2;p = - mgl sin cp (1.1)
1 This is true only for ideal constraints, i.e. constraints without friction.
THE VARIATIONAL PRINCIPLE IN MECHANICS 1.3
It is Lagrange's equation for the given case1 • It does not contain the
reaction R. Solution of the equation yields <pas a function of t. We
shall meanwhile not deal with the form of the function L for the
-system being considered (see Example 1
in Sec. 6). 0
Hence, the motion of a mechanical gystem x·
·can be described with tho aid of either
Newton's or Lagrange's equations. Natu-
rally, the latter can be arrived at proceed-
ing from Newton's equations (this will be
.done in Sec. 4). A very significant circum- y
.stance, however, is that Lagrange's equa-
tions can be obtained with the aid of a
quite general variational principle-the
principle of least action. It can be used as m
the foundation of classical mechanics instead
of Newton's laws. A merit of the principle y
Df least action is that it can· readily be mg
extended to systems that are not mechanical
·Or purely mechanical, for example, elastic
~. electromagnetic fields, an<j. fields of .. I.· Fig. 1.1.
elementary particles. LYhl .k~1 Yl'\1)»"'-:
Summarizing, the approach to the studying of mechanical systems
-set out in the present chapter can be said to be much more general
than the method based on Newton's laws.
2. Constraints
For a system of particleg with the masses m< 1>, m< 2 >, ••• , the equa-,
tions of Newton's second law can be written as follows:
, m;xi = Fi (i = 1, 2, ... , n) (2.1)
where m 1 = m 2 = m 3 = m< 1 > is the mass of the 1st particle, m4 =
= m& = m 6 =m< 2 > is that of the 2nd particle, ... , x 1 is the coordinate
:c of the 1st particle, x 2 is the coordinate y of the 1st particle, Xa
is the coordinate z of the 1st particle, x~, x 5 , x 6 are the Cartesian coor-
dinates x, y, z of the 2nd particle, ... , F 1 is the projection onto the
x-axis of the resultant force F< 1 > acting on tho 1st particle. F 2 is the
projection of the}orce F< 1 >onto the y-axis, F 3 is the projection of the •
same force onto the z-axis, F 4 , F 5 , F 6 are the x-, y-, z-components of
1 The generalized velocity enters this equation in the form E._~. We remind~
dt
QUr reader that dots over a symbol stand for the time derivative: ; = dx/dt,:
and ;· = tf.2x/dt 2 •
MECHANICS ' . .
the· force F< 2> acting on the second particle. etc. The number n of
equations contained in (2.1) equals the triple number of particles
in the system. ·
· Restrictions of a geometric or kinematic nature may be imposed on
the positions and ':elocities o) a_ sys~em' s pa:t;ticlesh~jles,e restricmt · ons
are called constramts. ~~ , "'1-lAL.:J•~ tk ~~" , ~~ t.;,'J
· Examples of systems with geometric constraints are: 'it~t.··· -~
(1) a particle which in its motion cannot leave a given surface or
a given curve. The surface or curve may be stationary (a stationary
constraint) or move in a preset way (a 'non-stationary constraint);
· (2) two particles A and B joined by a rigid weightless rod of
length l. In this case, the restriction imposed by the constraint can
be written in the form of the equation · · ·
I. L ' .
(.xA-xB )2 + (YA-YB
,' . )2 + (zA-zB
,.. . )2 = l2 (2.2)
'' .....~":". t~--. ·.
(3) two particles joined by a weightless thread of length l. The

~nalyticalexpression of such a constraint is~
j l,. 1_,
(xA-
... ,
xB) 2 + (YA- :
Ys) 2 +(zA- is) 2 ~ l 2
; . ' ~ I' i ~ ,' ' I
.
'
(2.3)
(4) a perfectly rigid body which can be considered as a system of

particles with unchanging distances between them, i.e. experiencing
constraints like (2.2).
An example of a system with a constraint of both a geometric and
a kinematic nature is a ball rolling without slipping over a rough
surface. The kinematic restriction is that the velocity of the point
of contact must be zero .
.. .In the general case, a geometric constraint can. be represent~d by
the 'equation. ;,
f (x1, X2, ••. , Xn, t) = ~ · (2.4)
,
'
~1.
........ ,
(n = 3N, where N is the number of particles in the system) .
.. Whe_p... restrictions are imposed on the velocities of particles in
addition to their coordinates, the equation of a constraint i&
. '/J'l : ~·-~: • • •
.:~; '•ill" ,., q> (xl, X2•' ~ •• , Xn, X1, X2, • ,.• , Xn, t) = 0 (2.5)
..
If Eq. (2.5) can be integrated over time, it is evidently equivalent
to.an.equation in the form of (2.4). ·
· Constraints of the kind given by (2.4) and integrable constraints
given by (2.5)that can be reduced to them are known as holonomic
(t>r!'integrable) constraints. Systems with such constraints are also
called holonomic. The systems in Examples 1, 2, and 4 treated above
belong to the holonomic type.
Hence, with holonomic constraints, the restrictions they impose
are expressed in the form of equations relating the coordinates of
the particles and time [see Eq. (2.4)1.
Non-integrable constraints in the form of (2.5), and also constraints
expressed as inequalities (see Example 3), are called non-holonomic.
Constraints that do not change with time are known as stationary
(or scleronomous). Constraints that change with time are known as
non-stationary (or rheonomous).
The equation of a holonomic stationary constraint is
\.
(2.6}
It differs from Eq. (2.4) for a holonomic non-stationary constraint

in that it does not include the time t explicitly.
There are no general methods of solutions for problems with non-
holonomic constraints. An individual approach to each problem is
required. We shall not consider non-holonomic systems.
Every holonomic constraint, i.e. every constraint expressed by
Eq. (2.4) or (2.6), allows us to represent one of the coordinates as
a function of the others. Consequently, every such constraint
diminishes the number of independent coordinates by one.
Recall that the number of independent quantities needed to
determine the position of a system in space is called the number of
degrees of freedom of the system. We can therefore say that every
holonomic constraint diminishes the number of degrees of freedom
of the system by one.
If constraints are absent, a system of N particles has n = 3N
degrees of freedom. When there are r constraints, the number of de-
grees of freedom will be s = n - r = 3N - r.
Constraints act on the particles of a system with the forces R< 01 >
called reactions. A constraint with no friction is said to be ideal.
If an ideal constraint is also stationary, its reaction R is always per-
pendicular to the direction of the possible elementary displacement
of the particle which the forceR is applied to (for instance, in Fig. 1.1
the reaction R is directed along the thread, while the velocity of
the particle is perpendicular to it). This is why reactions of ideal
stationary constraints do not work on a system. If a constraint (even
an ideal one) depends on the time, the work done by its reaction will
be non-zero, as a rule.
The expression for the elementary work done on the particles of
a system by the reactions of the constraints is dA = ~ R 1 dx 1•
We have seen that this work is zero for a system with ideal stationary
constraints. Consequently, for such systems
(2.7)
MECHANICS
ivhere dx 1 are the projections onto the coordinate axes of the dis-
placements of the particles allowed by the constraints imposed on
·them1•
3. Equations of Motion in Cartesian Coordinates

Consider a system formed by interacting particles. Assume that
-external forces also act on the particles. We shall use the same sym-
bols as we did in the preceding section (see the first paragraph of
~ec. 2) for the coordinates of the particles and the components of
J.
I ''~
:die rorces. . · ·
In some cases, the forces acting on particles (or at least a part of
-these forces) can be represented in the form
,.
F ,=- -
au (3.1)
. iJ:Ct
. '' U = U (x1 , x 2 , •• :, Xn, t) • (3.2)

·H(a function of the coordinates of the particles and the time known
.tis the potential of tHe system. If the f1inction U dods not include
the time t explicitly, it is the potential energy of the system.
A force determined by form'ula (3.1) is called a potential force.
:Stationary and non-stationary potential forces should be distinguished.
Forces depending only on the coordinates of a particle and not
·depending on the time explicitly are called stationary. The func-
tion U not containing the time explicitly corresponds to these forces.
When U contain~) t explicitly, the force depends not only on the
·c'oordinates, but also on the time and, consequently, will be non-
:stationary. · Stationary potential forces are called conservative.
, Systems in which only conservative forces act are also called' con-
:SerVative.
· 1'We shall note that in accordance with(3.1), the force acting on the
particle numbered a can be. represented as a gradient of the func-
tion (3.2): · ·
- ..... • : 'i· F~ = -'V~U (xh t) (3.3)
ile~;e. '\!~ is a~operator whose components equal the partial deriva-
·tives with respect to the coordinates of the a-th particle. By Xt
here and below in the symbols of functions we denote the set of all
·the'coordinates: x1 , x 2 , •• • , Xn·
. 1 -ln the sums encountered in theoretical physics, the index over which sum-
mation is performed (the dummy index) is repeated twice, as a rule. In this
-~?.?.~~~tion,_it is customary practice to omit the symbol L} and understand sum-
mation· over the twice repeated indices. For instance, simply atbt is written
-instead of 2jatbt. Although this way of writing is distinguished for its brevity,
however, we shall not use it. Summation will be understood only where the
:Symbol 2j is written.
THE VARIATIONAL Pll.lNCIPLE IN MECHANICS 17
Formula (3.3) can also be written as follows:
F a; =; - au
ora;
(3.4)
where ra. is the position vector of the particle1•

Assume that part of the forces exerted on the particles of a systent
are potential and part of them are non-potential. Equation (2.1)
can therefore be written as
••
mtXt= - -
au .
-+Ff (£=·1, 2, ... , .n) (3.5)
8 Xj
where F1 are the components of the non-potential forces, and n is

the triple number of particles. .
Equation (3.5) can be written in a form very convenient for gener~
alizations. For this purpose, the Lagrangian mentioned in Sec. 1 is
introduced. It is determined as follows for the system of particles
we are considering:
• mi;~
L(xtt x 1 , t)= LiT-U(x 1 , t) (3.6)
i
.
Time differentiation of the partial derivative of L with respect to
x 1 yields the left-hand side of Eq. (3.5):
d aL d • ••
dt -.- = dt (mtxt) = mtXt
axi
The partial derivative of L with respect to Xt gives the i-th compo-
nent of the potential force:
aL au
OXj =- OXj
Consequently, we arrive at the relations

..!:..._
dt
iJL -
•
iJL
ox·
= F'!'t (.L = 11 21 • • • ' n
) (3.7)
OXi t
named Lagrange's equations.

For systems in which only potential forces act, Lagrange's equa-
tions are
d 8L iJL .
71t -.---a-=0
OXf Xj
(£=1,2, ... ,n) (3.8)
1 By the derivative of the scalar cp with respect to the vector a is understood

a vector having the components 8cp/8ax, iJcp/iJay, 8cp/8az. Consequently, the sym-
bol iJcp/iJr stands for a vector with the components 8cp/iJx, ocpioy, 8cp/8z (i.e. grad cp
or Vcp; see Appendix XI).
2-018
f8 MECHANICS
Lagrange's equations can sometimes be written in the form of (3.8)

even when the forces acting on particles depend on the velocities
(the Lorentz force is an example of such forces). This can be done
provided that such forces can be obtained from a certain function
U*(x 1, ; 1, t) with the aid of the relation
.·1k1.. · , F au• d au• .
\J.~) .Jif>i;v~·; .!1 il' 't= -8Xf+df a; .(3.9)
. t
(we leave it to the reader to verify this statement).
:'.The function U* .(xt>. ; 1 , t) is known as the generalized potential.
We shall apply the adjective generalized-potential to forces corre-
sponding to this potential. When these forces are present, the Lagran-
gian i~ written as follows:
• :·:JJ·:t-~ ;· -. ;: •
.
·mixr , •...
·,.; t ... ,: L(x, x 1; t)= ~-2 --U(x 1 , t)-U*(x" x 1 , t) (3.10)
.:. ; .
.' ~
i
The forces depending on the velocities of particles also inclu<le
dissipative forces. This adjective is used to designate forces always
~ii,re~ted oppositely to the velocities of particles and, consequently,
causing their deceleration. Dissipative forces include, for example,
forces of friction. When dissipative forces are present, the total
fn:echanical energy of a system diminishes (dissipates), transforming
into other non-mechanical .kinds of energy (for instance,
into the internal energy of bodies or, as we customarily say, into
heat). ,.
· The dissipative forces F<d> are often proportional to the velocities
of particles so that their components along the coordinate axes are
determined by the equation
.,,.. F~d'=-k 1 x'1 (i=L2, ... , n) (3.11)
In this case, the dissipative forces can be expressed in terms of Rayleigh's
dissipative function 1 equal to
1 •
. .
.,, .. ,) ..
D 2 ~ k 1 x~ =-"" .
(3.12)
i
Indeed, comparing expressions (3.11) and (3.12), we can easily see
that . ... ,,. . p~d> = ·~ a?
(3.13)
ax,
Substituting this expression into formula (3. 7) for Fi and assuming
tha~ there are no otMt non-potential forces, we obtain the equation
~ aL _ aL -j- fJD =--=O (3 . 14)
dt <I.
vXj
ax; ".
UXj
"I, 1 -This function is customarily designated by the letter F orR. To avoid con-
f~sion, however. we have preferred the symbol D.
THE VARIATIONAL PRINCIPLE IN MECHANICS 19
The function D has a simple physical meaning. The work done by

dissipative forces during the time dt is . ~,
dA = Lj F~4 > dx 1 = 2j Fid>;, dt = - 2j 8 ? ;, dt = -2D dt

" az 1 ,,. ,
(we have used Euler's theorem on homogeneous functions, see Appen-

dix II). This work is done at the expense of the system's store of
energy. Hence, the quantity -dA/dt = 2D gives the rate of energy
dissipation.
· Lagrange's equations are superior to Newton's equations (3.5) in
that, as will be shown in the following section, they retain their form
when we transform from Cartesian to any generalized coordinates.
When we pass over to independent generalized coordinates, the
reactions of the ideal holonomic constraints vanish from the equa-
tions, which greatly facilitates the solution of problems.
4. Lagrange's Equations in Generalized Coordinates
Generalized coordinates qlt are defined to be any quantities (lengths,

angles, areas, etc). that determine the position of a mechanical
system in space. As an example, we can indicate the spherical coor-
dinates of a particle: r, '1'7, cp. Cartesian coordinates are obviously
a particular case of generalized coordinates.
The time derivatives of the generalized coordinates, i.e. the quan-
tities qk, are known as the generalized velocities of a system.
The number of independent generalized coordinates needed to set
the position of a system equals the number of its degrees of freedom.
In the following, we shall always choose the generalized coordinates
so that their number coincides with that of the degrees of freedom
of a system (i.e. so that they are all independent of one another).
We must note that generalized coordinates are often helpful in
systems without constraints too (in this case their number coincides
with that of the Cartesian coordinates). For instance, in solving
a problem involving the motion of a particle in a central field of
forces, the spherical coordinates r, '1'1, cp are more convenient than
the Cartesian coordinates x, y, z. ..
The following representation is very helpful. Let us introduce into
consideration a system of coordinates in an imaginary s-dimensional
space (it is called a configuration space or a q-space). We plot the
values of the coordinates qk (t) along the axes of this system. Hence,
for each instant t, a point in the configuration space corresponds to
the position of the system in conventional space. The motion of
a point in our imaginary s-dimensional space corresponds to the
motion of the system in the real three-dimensional space.
2•.
.{.
20 MECHANICS.
We shall prove that Eqs. (3. 7) remain true in a transition from the
Cartesian coordinates xi to the generalized coordinates q1" and also
that Lagrange's equations written in independent generalized coor-
dinates contain no constraint reactions.
· Consider a holonomic system with ideal stationary constraints.
We shall divide the non-potential forces acting on the system's parti-
cles into two categories: the constraint reactions R i and the other
non~ potential forces Ff. Equations (3. 7) will therefore become
8~- !~ =R 1 +Ft (i=1,2, ... ,n)

~ ox vx,
(4.1)
W "li.H : !· i '' .
. L.et r constraints expressed by the conditions
' 111.~ ·. 1
L fz (x1 , x 2 , ••• , Xn) = 0 (l · 1, 2, ... , r) (4.2)
be imposed on the system (since the constraints are stationary, the
function f does not contain the time explicitly).
The Cartesian coordinates xi determining the position of a sys-
tem's particles can be represented as functions of the generalized
coordinates qk. If the equations of the constraints do not contain t
explicitly, we can always choose qk so that the functions expressing x 1
in terms of qk do not contain t, i.e. so that these functions are in the
~?.rm.
Xt . Xt (q1 , q2 , • • • , q8 ) (i = 1, 2, ... , n) (4.3)
.(s i~ the number of degrees of freedom equal ton -r). In the following,
for brevity's sake, we shall write expressions of the kind
Xt (q 1, q2, ••• , q8 ) in the form Xt (qk)·
· In accordance with (4.3), the time derivatives of the functions
x 1 (qk) have the form
(4.4)
Siimma"'tion is performed over all the qk's, i.e. the subscript k takes
oti; all the values from 1 to s. Expression (4.4) can be written for
anY' i from 1 to n. In the following, when this will not lead to mis-
understandings, we shall not indicate the values taken on by the sub-
script for which summation is being performed. This subscript is
called a dummy index. We must note that in summation formulas
a dummy index may be designated by any letter-the use of one
subscript instead of another does not change the sum. Particularly,
expression (4.4) can be written just as successfully, for instance, in
the form· ·," 't ,;1 ··.. ;:
• iJXi •
.'.)'1.!;;;.-. ! l ,,;
x,= ~ -"'-qz
vqz
(4.5)
l
The quantities 8xJaq~~. do not contain the generalized velocities ,j~~..

We can therefore state on the basis of (4.4) that
Oxi Ox;
dq'h. = Oqlt
(4.6)
We must also note that since functions (4.3) do not contain tho quan-
tities q~~., the partial derivatives of x 1 with respect to these quanti-
ties are zero:
0~ 1 = 0 (4. 7)
Oqk
Finally we shall obtain another relation that we shall need later
on. The quantities 8x 1 /aq~~. are functions only of the coordinates q~~.
(we have already noted that these quantities do not contain the
velocitiesqlt)· Renee,
..!!.__ OXi = ~ -~ ( Ox; ) • = " iJ 2x1 • (4.8)
dt aq~~. ..t....~ oq 1 aq~~. J q, ~ aq, aq~~. q1
l l
Now, we shall differentiate expression (4.5) with respect to qlt. Since
the derivatives a;l! /aq~~. equal zero, we obtain
OXi " 0 ( uxi ) • "" i) 2 xi

iJq" = LJ aq" uqz qz = ~ aq" aq 1 q1
l l
Comparing this expression with (4.8), we arrive at the relation

ux1 d OXi ( 4. 9)
uqk = dt uqk
Having obtained all the required relations, let us now turn to

the proof. Multiplication of both sides of each of Eqs. (4.1) by
8x 1/8qk and summation of all the equations yield
""'~ (.!:_
dt
OL ) ax 1
8; 1 aqh.
_ aL ax 1
""'
~ ox 1 aqh.
= ""\.l R ax,
~ 1 aqll.
+ ", F~
~ 1
ax,
aqk
( 4.10)
1 1. • 1 1
The first of the sums on the right-hand side of this expression equals
zero. To prove this, we multiply it by dqk:
( ~ R 1 ;;~) dq~~.= ~ R 1 ( ;;~ dq~~.) = ~ R 1 dx 1

i i i
Here dx 1 are the increments of the Cartesian coordinates appearing

when qk receives the increment dqh., the other-generalized coordinates
remaining unchanged. According to (2. 7}, however, the sum ~ R; dx 1
1
... . -
22 MECHANICS
fo'r any dx/s allowed by their constraints is zero (we remind our
reader that we assume the constraints to be stationary and ideal).
Hence,
( ~ R1 ::: ) dq,_ = 0
{
and, since dqk =1= 0, we arrive at the conclusion that

~;:II 'li.! ,'! ,.-;'-·•)j~'
(4.11)
The quantity
(4.12)
•~rl t 1' ' .••.
. j. : . .
[see the second term on the right-hand side of Eq. (4.10)1 is referred
to as the generalized force. This name is based on the grounds that
. .~'he expressiOn
. ·,•.c •
0: dq', = ~ Ft ( ;;~ dqk) == ~ Ft dx 1

i i
gives the work done by the forces F! in the displacements of parti-

cles due to an increase in the generalized coordinate qk by dq,_.
Consequently, the right-hand side of Eq. (4.10) is simply Qh,.
Let, us now consider. the left-hand side of this equation. We add
the sum~ 0 ~
. .. { OX!
:xi to the minuend and the subtrahend of the left-
q,.
hand side. Hence, the left-hand side of (4.10) will contain the
difference of the expressions
(4.13)
and
•}l
(4.14)
Expression (4.14) is the derivative of the Lagrangian with respect

to the generalized coordinate q,., i.e. aL!aq,..
":;With a view to (4.9), we can write expression (4.13) in the form
~ (.!:..__ aL ) ~
L..l dt "· oqk
+ LJ
~ .3.!i. ( .!:..__ ax
•. . dt aqk
1 ) = ~ "" oL oxt
dt L..l 8 ; aqk
; uXt i 8 X1 i I
THE VARIATIONAL PRINCIPLE IN MECHANICS
Finally, substituting o;/oq 11 for Ox/oq 11 in accordance with (4.6),

we arrive at the expression
.2_ ~ aL iJx;
dt L.J • •
i ox 1 8qr..
w1uc . exac tl y dt
. h 1s 8L . Indeed, according to the rules for the
d -.-
aq"
differentiation of a composite function
ilL
•
=~
£..J
aL ax 1
ax •
+~
LJ
aL
•
a;,•
aq" i t aq" i ax, iJq"
but according to (4.7), all the axi/oq 11 's equal zero so that the
first of the sums vanishes. We have thus shown that expression
d fJL
(4.13) equals Tt -.-.
8q"
Therefore, the difference between expressions (4.13) and (4.14)
is simply : .!..~ -
t aqR.
:L .
Introducing this value into (4.10), we
q"
arrive at the equations
a 8L
Cit-.-- Bq11. = Q"
8L •
(k= 1, 2, ..• , s) (4.15)
aq~~.
that differ from Eqs. (4.1) in containing the generalized coordi-

nates q11 instead of the Cartesian coordinates x 1 and the generalized
forces Q: instead of the forces Ff. Equations (4.15) do not contain
the reactions R 1 of the constraints.
·If all the forces acting on the particles of a system (except for the
reactions of the constraints) are potential ones, Eqs. (4.15) become
d. 8L 8L
Tt-.-- 8 =0 (k=1,2, ... ,s) (4.16)
8q" q"
Equations (4.15) and (4.16) are Lagrange's equations in general-
ized coordinates. The function L = L (q1 , q2 , • • • , q8 , 1 , 2 , • •• , qq qs)
in them is a function obtained from (3.6) ?Y substituting functions
(4.3) and (4.4) for the quantities x; and x 1•
It can be shown (see Appendix I) that Eqs. (4.15) and (4.16) also
remain true for holonomic systems with ideal non-stationary con-
straints.
We have thus proved that Lagrange's equations have the same
form in both Cartesian and generalized coordinates.
Expression (3.6) for the Lagrangian in Cartesian coordinates shows
that the derivative of L with respect to x 1 equals p 1-the projection
.
24 MECHANICS
of the momentum of the relevant particle onto one of the coordinate

axes; '
aL
- . - = m1x1 = Pt (4.17)
i:&t
and the derivative of L with respect to x 1 equals F 1-the projection
of the potential force acting on the particle:
fHiJ -'i ; 1 ' '" ' ·' I aL _ _ au ~F
OX! - axi - f
(4.18)
By analogy with (4.17) and (4.18), the quantity
.,,
I ·, ~
(4.1!)
is· called the generalized momentum, and the !fUantity

iJL
Qlt= aqk (4.20)
~he .·generalized force.
· With the use of tllese quantities, Eqs. (4.16) can be written as
lj'i-/ : \'-' \
?,It =Qit (4.21)

similar to Newton's equations
dpt -F
dt- f
We must note that definition (4.20) is more general than defini-

tion (4.12) which we have introduced for non-potential forces. If we
were to extend definition (4.12) to potential forces, i.e. to forcesr that
can be written in the form F1 = - aulax1, we would arrive at
the expression
((J l. i
Q _ "" F ax1 _ _ "" au ax1 _ _ ou
11 - L.J 1 aqlt - (4.22)
L.J ax1 aqk - oqk
i i
According to definition (4.20), however,
~· ... : Qh= :q~ =a(~;:U)= :~- :~ (4.23)
Expression (4.23) differs from (4.22) in the term aT/aqk which, as
,will be shown in the following section, generally speaking, is non-
zero [see formula (5.9)1.
·5. The.Lagrangian and Energy
'· The Lagrangian L (qk, q 1t~ t) is a characteristic function of a mech-

anical system. It is quite natural that not only momenta and forces
[see formulas (4.19) and (4.20)], but also the energy of a system can
be expressed in terms of this function.
The expression for the total energy of a system in terms of the
Lagrangian is
8L • •
E;= lj-.-q~~.-L(q~~., qh., t) (5.1)
h. 8qll.
The grounds for determining E in exactly this way will be revealed
below. Expression (5.1) is more general than the one known from
the general course of physics, i.e.
(5.2)
(Here T is the kinetic energy, and U, the potential energy.) Defmi-
tion (5.1) also holds when the total energy cannot be represented as
the sum of the kinetic and potential energies.
Let us calculate the total time derivative of the quantity (5.1)
iE 8L • • • d 8L 8L •
dt = ~ -.- qh. + ~ ql>. dt -.-- ~ 8qlt ql>.
It 8qP. It iJqh, It
8L '· 8L ( d 8L 8L) • 8L
- ~ -.- qlt- at= ~ dt -.- - 8q7>. qlt- at
It 8qlt It 8qlt
In accordance with Lagrange's equation (4.15), the expression in
parentheses equals the non-potential generalized force Q7:. Hence,
dE= 'Q Qkq.lt- 8L
dt ~ at
= W*-~
at
(5.3)
II.
where W* is the power of all the non-potential forces acting on the

system's particles.
When the Lagrangian does not contain the time explicitly, 8L/8t=
= 0, and formula (5.3) becomes
!:!_= W*
dt
which is known from the general course of physics.
A glance at (5.3) shows that for conservation of the total energy
of a system, the absence of non-potential forces is not sufficient.
It is also essential that the time be absent in the Lagrangian explic-
itly, i.e. that the system be conservatiYe. We have thus established
that for a closed system in which only conservative forces act, the
quantity determined by formula (5.1) remains constant. This is why
quantity (5.1) is known as the total energy of a system.
The functions of the quantities qlt and qlt that in the motion of
a system retain a constant value determined by the initial conditions
are called integrals of motion in mechanics. Consequently, the energy
26 ':MECHANICS';' ·•;;' :: ...
of ··a conservative closed system can be said to be an integral of

motion.
· · Let us find the conditions in which the definition (5.1) transforms
into (5.2). To do this, we shall investigate the expression for T in
generalized coordinates. We assume that the constraints are holono-
mic. We express the Cartesian coordina,tes x 1 of the system's parti-
cles in terms of the generalized coordinates qk:
;lbi,- '•; .. , . :'i.i: 1°
. (•'Jl Xi= Xi (ql, q'l., ••. , q8 , t) (i = 1, 2, ... , n) (5.4)

We must note that the time is contained explicitly in these relations
only for non-stationary constraints.
' We find the total time derivatives of the functions x 1• Since the
quantities qk are functions of t, the required derivatives have the
form
i L; ·~' ' ' ' '• j
.we introduce these expressions into the formula for the kinetic
energy:
'
T 1 "" •2 1 "" '( OXf
:.:= 2 LJ m,xi = 2 LJ m, Tt + LJ" OXi •
aqk qA
) 2
i i k
'i' ~ ~~ ( : 1)2 + ~ m 1~~ ~ ;;~ qk+ ~ ~~ ( ~ ::~ qk) 2 (5.5)

i i k i A
(i.I'n accordance with (5.4), the expres~ions 8x 1/fJt and oxiloqkare

functions of qA and t; they do not contain the quantities q·k· Conse-
quently, the first of the sums in formula (5.5) is also a function of
.q" and t. Designating this function by the letter a, we have
. '
a (qA, t) = 2: ~~ ( ~~~ )2 (5.6)
. Chan~ing the sequence of summation over i and k in the second

M the sums in formula (5.5), we give it the form
·:.' :r1',', _: '· Lj qA {Lj mt ~: 1 ;;~} = Lj ~A (qJ, t) q"
A · i A
where.;,, i ·
. A ( ) "'Q OXt ox; (5.7)
H ;-, "''tl'l .•. ,,,, __ .. PA qJ, t = L.Jm'Tt aqk
finally, representing one of the factors in the third term of (5.5)

•
~x 1 1" and the other factor in the form
in the form "LJ vqk q "" 0: 1
. LJ vqz
qz,
k l
and then, changing the sequence of summation over i, k, and ·z,

we write the third term as follows:
~
L..J
mi ~
2
ax, • ~
L.J aq~~. q~~. L..J
ax, • "" • •
aqz q1 = L.J q~~,qz
~
L..J 2
m, axi ax,
aq~~.
aqz
t k I 11., I i
= ~ '\'11.1 (q,, t) ;]ll.qz

11., l
where
(5.8)
The expression for the kinetic energy can thus be written as
T =a (q,, t) + 2J ~II. (q1 , t) q~~, + ~ '\'11. 1 (q" t) q~~,q 1 (5.9)

II. II., I
With stationary constraints, t is not contained explicitly in func-

tion (5.4). Consequently, ox 1/ot = 0, so that the coefficients a (q 1, t)
and ~h (q 1, t) vanish. In this case, thecoefficients '\'11.z do not contain
the time explicitly. The result is
(5.10)
Therefore, the kinetic energy of a system with stationary con-

straints is a homogeneous quadratic function of the generalized ve-
locities qh.
We established .in Sec. 3 that for a system of particles, the Lagran-
gian in Cartesian coordinates is
• m.;x~
L (x 1, x 1 , t) = Lj---;;!-- U (x 1, t)
i
In the transition from Cartesian to generalized coordinates, the

first term on the right-hand side of the above equation transforms,
generally speaking, into expression (5.9). If we limit ourselves to
a consideration of stationary constraints, however, the first term
m nst. be replaced with expression (5.10). The second term becomes
U (q 1., t). With station'l.ry constraints, the time is contained explicit-
ly in U provided that the potential forces are non-stationary.
We have thus established that for stationary constraints the
Lagrangian of a particle system written in generalized coordinates is
L (q~~,, q 1p t)=T (q~~,, q~~,)-U (q~~,, t)
28 MECHANICS, .
where T is determined by formula (5.10). Using this expression in

formula (5.1}, we get
oT • •
E= ~ -.- qk-T (qk, qk) +U (qk, t) (5.11)
k 8qk. ,
In accordance with (5.10), Tis a homogeneous quadratic function of

the quantities qk. Consequently, the following relation holds for it
on the basis of Euler's theorem (see Appendix II):
8T •
~ -.-qk=2T
1r, 8q~r, .
Making such a substitution in (5.11), we arrive at the formula

• ! ;
~ = 2T - T +-U = T (qk, qk) + U (qk, t)

We obtained this fo~mula with only a single assumption, namely,
that the constraints are stationary. If, in addition, the potential
forces are also stationary, the potential U (qk, t) transforms into the
potential energy U (qk), and formula (5.1) transforms into (5.2).
Definition (5.1) thus coincides with (5.2) when both the constraints
and the potential forces are stationary.
(6. Examples of Compiling Lagrange's Equations

Having selected the independent generalized coordinates conveni-
ent for describing the system being considered, we must establish
the .form . of the functions
Xi = Xi (qk, t) and Xi
. =
. .
Xi (qk, qk, t)
: .t ..
and introduce them instead of the quantities Xi and xi into the ex-
pression for the Lagrangian
L=T(;i)-U(xi, t)= Lj ~~ ;~-U(xi, t)
The, result is the Lagrangian in generalized coordinates:
If' the expression obtained for L contains terms that do not depend
on qk and qk, they may be discarded because they do not contribute
to the .quantities aL!aqk and aL!aqk and, consequently, do not affect
tlie form of Lagrange's equations. ·
' 1 We sometimes succeed in greatly simplifying the operation of
l. •
finding the function T (qk, fJk, t). This is possible when, it is easy to
establish the relation between a particle's elementary displacement

ds and the increments of the generalized coordinates q11 • For instance,
in polar coordinates on a plane (Fig. 6.1), the displacement ds is
a diagonal of a rectangle constructed on tho sides dr and r dcp (we must
0
0
Fig. 6.1. Fig. 6.2.
remember the smallness of dcp). Consequently, ds2 = dr + r2 dcp 2 •

Dividing this quantity by dt 2 , we obtain the square of the velocity
of a particle: v2 = ; 2 +
r2 ~ 2 • Finally,
1 1 • .
T = 2 mv2 = 2 m (r2 + r2cp2) (6.1)
For cylindrical ·coordinates, a third coordinate z is added to the
previous two (r and cp ). The displacement ds is a diagonal of a rectang-
ular parallelepiped with the sides dr, r dcp, and dz. Hence,
f • • •
T= 2:. m (r2 + r 2 cp2 + z2) (6.2)
Three mutually perpendicular segments (Fig. G.2) of lengths
dr, r dq,, r sin {} dcp (the last segment is directed beyond the drawing1 )
correspond to the increments of the polar coordinates r, {}, cp. The
displacement ds coincides with a diagonal of a rectangular paral-
lelepiped constructed on these segments. Therefore,
1 • • •
T = 2 m (r 2 + r 2{}2 + r 2 sin2 {}cp2) (6.3)
1 We shall depict segments perpendicular to the plane of the drawing by a

circle with a dot if the segment is directed towards us, and by a circle with a
cross if the segment is directed behind the drawing.
30 MECHANICS
1 ,~et. us consider several examples. . .. . .:

.A~. A Simple Pendulum (see Fig. 1.1) .. The displacement of the
p~~t.icle ,m is ds = l d<p. Hence, T = ~- ml2~ 2 • The potential energy
is 'fr = ·- mgl cos <p. Consequently, the Lagrangian is
1 •
L =2 ml 2 !p 2 + mgl cos <p
We invite our reader to write Lagrange's equation and convince
himself that it coincides with Eq. (1.1) •
. We find the generalized momentum and gep.e:r:alized force. By
formula (4.19) · · ·
lJL •
prp = -. = ml 2 !p = mvl
{)cp
Hence, in the given case, the generalized momentum coincides with

the moment of the conventional momentum mv (the angular momen-
tum) relative to the point of suspension of the pendulum.
By formula (4.20),,
oL -mg·z SID
Qrp=-= . !p
lJcp
and it is exactly the moment N of the force mg relative to the point
Of suspension. We must note that the elementary work is
:·>) ' dA-:- Qrp d<p = N d!p
The last expression coincides with that for the work in rotation
known from elementary mechanics.
, ' Finally, we find the energy of the pendulum by formula (5.1) (we
b'an do this in the given case because the constraint is stationary):
'' lJL· • ' • 1 • ·:' 1 •'
E =-. <p-:7'L = ml2!p2- 2 ml 2!p2- mgl cos.!p . 2 m (l!p) 2 -mgl cos IP
{)cp ' .• ' '
2. · A"'Pendulum with a Uniformly Moving Point of Suspension.
Assume that the point of suspension of a simple pendulum moves in
a horizontal direction at the constant velocity v in the plane of
oscillations of the pendulum (Fig. 6.3). The equation of the con-
straint is . , :
. f (x, y, t) = (x- vt)Z + y 2 --:- l 2 = 0
(the constraint is non-stationary).
From the expressions for the Cartesian coordinates
x = l sin !p +
vt, y = l cos <p
it follows that
;I ti'/
; = (l cos <p) ~ + v, y = - (l sin <p)~
THE VARIATIONAL PRINCIPLE IN MECHANICS 3f
whence
1 • .
T = 2 m [l2cp 2+ (2vl cos cp) cp + v2]
The potential energy U = - mgy = - mgl cos cp.
In compiling the Lagrangian, the constant term. ; mv2 may be-
discarded. Consequently
1 • •
L = 2 m [l2cp 2 + (2vl cos cp) cp] + mgl cos <p
Notwithstanding the non-stationary nature of the constraint, the
time does not enter the function L explicitly.
vt
0 ,.--A---.., v
X
I
I
I
I
I
I
. I
Y _ _.:. _ _ _ _L_ m
m X I
I
y
y
m·g
mg
Fig. 6.3. Fig. 6.4.
We invite our reader to find Lagrange's equations in this and the-

following examples.
3. A Pendulum with a Poirit of Suspension Moving with Constant
Acceleration. Consider a simple pendulum whose point of suspen-
sion S moves with the constant acceleration a along a straight line
making the angle a with a horizontal plane (Fig. 6.4). The coordi-
nates of the point of suspension are ·
Xs=(; acosa) t 2 ,,.-Ys=(; asina)t 2

and the coordinates:ofJpoint m are
x= (; a cos a) t 2 + l sin cp, y= ( ; a~sin a) t2 + l cos cp

' < ,..__ ! ~- - MECHANICS ..
Differentiation of x and y with respect to t yields

.
x =(a cos ct) t + (l cos cp) cp,. y.=(a sin ct) t - (l sin cp) cp.
Introducing these values of ; and y into the expression for the ki-
netic energy, we obtain·
1
T = 2 ma2t2 + mal (cos ct cos cp- sin ct sin cp) cpt• + 21 ml2cp• 2
The potential e~ergy is U = - mgy = - mg [ ( ! a sin ct) t2 +
4'1 cos cp J. In writing the expression for the Lagrangian, we may
omit the terms fma 2 t 2 in T and - (} mga sin ct) t 2 in U because
.
they do not contain cp and <p and therefore cannot affect the form
of Lagrange's equations. As a result,
y ..;·- we· find that
L =mal (cos ct cos cp- sin ct sin <p) cpt +
1 •
+2 ml2cp2 +mgl cos <p
The Lagrangian depends explicitly
wt
cp = {
2
IX I ,
on t (this is due to the non-stationary
nature of the constraint), the time
having entered L through the agency
X
of the kinetic energy T.
4. A Particle Moving along a Uni-
Fig. 6.5. formly Rotating Straight Line. ,Assume
that a particle of mass m expe-
riences a non-stationary constraint consisting in that the particle
can move only along a straight line rotating at a constant angular
velocity (cp = rot) in a vertical plane (Fig. 6.5). In addition to the
reaction of the constraint; the particle is ac.ted upon by the poten-
tial force mg. The Cartesian coordinates of the particle are
I ' !~
· x = r cos rot, y = r sin rot

Cpnsequently, ..
x
. = r.cos rot - rro sin rot, y
. = r. sin rot + rro cos rot
The kinetic energy is
The potential energy is

U = mgy = mgr sin rot
THE VARIATIONAL PRINCIPLE IN MECHANICS
(when we passed from the Cartesian coordinates to the generalized

coordinate r, the time t entered the expression for U explicitly).
Finally, the Lagrangian is
L= ~ m (~2 + r2ro2)- mgr sin rot
It depends explicitly on t; now the time has entered L in terms
of the potential energy U.
The generalized momentum is
Pr=-. =mr=mv
aL
or
where v is the particle's velocity along the straight line.
The generalized force
aL = mrro 2 - mg sm
Qr = ar •
rot
eonsists of two terms. The first of them, mrro 2, is the centrifugal force
of inertia, and the second, -mg sin rot, is the projection of the force
mg onto the direction r.
5. A Particle Moving along a Straight Line Rotating with Accelera-
tion. Assume that the straight line giving rise to a constraint (see
Fig. 6.5) rotates not uniformly, but with acceleration (q> = at2 ).
Hence,
x = r cos at2, y = r sin at2
. .
x = r cos at2- r 2at sin at 2
:"l
. .
y = r sin at + r 2at cos at 2
2 • ' 1 ~
1 •
T ='2m (r 2+r24a 2 t 2)
U = mgy = mgr sin atz
Accordingly,
1 •
L = -2 m (r2 + r2 4a2t2)- mgr sin at2
.
In this example, the time entered L explicitly in terms of both T
and U.
1. Principle of Least Action I

I
Instead of Newton's laws, mechanics can be based on the principle
of least action or the Hamilton principle2 • The action S during the
, 1 Before beginning to study this section, the reader should acquaint him-
self with Appendix III.
· 2 This principle was established by the Irish mathematician William Hamil-
ton in 1834.
3-018
34. MECHANICS ·~·
thne interval from t 1 to t 2 is defined to be the integral

,,., II'· . . t. . i . ;.:
S = JL (q'll., qlu t). dt t· ' ·

t,
(7.1)
where L (qlt, qk,t) is the Lagrangian of the system being considered.
Integration is performed from the instant t 1 at which the position of
the system is characterized by the values of the coordinates qk (t1)
to the instant t 2 at which the position of the system is determined
by the values of the coordinates qk (t 2).
According to the principle of least action, a system moves between
the positions qlt (t1) and qk (t 2) in such a way [i.e. the functions
qlt (t) have such a form] that the action (7 .1) has the smallest possible
value. Using the notion of configuration space, we can say that a
point depicting the position of a system moves in this space along the
curve for which the action S is minimum. ,
The Hamilton principle is the most general formulation of the
law of motion of mechanical systems. A merit of this formulation is
t'liaf it"can easily be, extended to systems that are noi' purely me-
chanical, for instance to elastic media and electromagnetic fields.
_.Inspection. ,of (7.1) shows that the dimension of action equals that
p¥ 'the 'product of energy and time (or momentum and displacement).
Pla'nck'S,:constant, also called a quantum of action,. has the same
<iimension.
It is a simple matter to obtain Lagrange's equations from the prin-
ciple of least action. The action S is a functional (see Appendix III).
According to the calculus of variations, ·a functional reaches an
extreme value provided that its variation is zero. Consequently.
the principle of least action can be expressed by the condition
t. .
88 = 8 \ L (qk, qk, t)'dt = 0 (7.2)
,,t~.
For thi£ reason, the principle of least action is known as the varia-
tional principle of mechanics. .
Jt is known from the calculus of variations (see Appendix III)
that the. variation of a functional of the type (7 .1) 1 vanishes if we
take functions satisfying Euler's equations (III.25) as qk(t). In the
given case, Euler's equations have the form
aL - ..!:.._ aL =0 (k 1 2 ) (7.3)
. aqk
91ftb··.•·,:· · r: ;:·· . .
dt •
aq"
= • • · · ·' 8
i.e; coincide with Lagrange's equations.
-11il' Coi:npare with (111:20). In the given ease, the role of f(x, Yk• y:) is played
.hy.L (t, q
qk, 11). The role of the independent variable x is played by t, that of
;,, (x) is played by q11 (t), and that of y/. (x), by the function 11 (t). q
TJ!E VARIATIONAL PRINCIPLE IN MECHANICS 35
We have thus convinced ourselves that the Hamilton principle

leads to Lagrange's equations.
Appendix III shows [see the text following formula (III.19)] that
the addition to the integrand in (7 .2) of the total derivative with
respect to t of any function of the generalized coordinates and time
does not change the conditions of the extremum, i.e. Eqs. (7.3).
Consequently, the Lagrangian must be determined to within the
additive summands that are the total time derivative of an arbitrary
function of the generalized coordinates and time (a constant is
a particular case of such a function). We have already noted this
circumstance in Sec. 4 and used it in Sec. 6.
Chapter II
p:;] CONSERVATION LAWS
Jn1!: l ! .. i
tW
"rn; : ., ··
.
"t; .''
-~· ~ ~--: ~ . . i! .
,·
8~ Energy Conservation
The conservation laws considered in mechanics are based on the

properties of space and time. Energy conservation is associated
with the homogeneity of time, momentum conservation-with the
homogeneity of space, and, finally, angular momentum conservatioD
is associated with the isotropy of space 1 .
We shall begin with the law of energy conservation. Assume that
a system of particles is in unchanging external conditions (this occurs
if the system is closed or experiences the action of a constant
external force field); the constraints (if present) are ideal and station-
ary. In this case, the time, owing to its homogeneity, cannot enter
the Lagrangian explicitly. Indeed, homogeneity signifies the equi-
valence of all instants. Therefore, the replacement of one instant
with another without changing the values of the coordinates and
velocities of the particles should not alter the mechanical properties
of the system. This is naturally true only if the replacement of one
instant with another does not change the conditions in which the
system is, i.e. if the external field is time independent (particular-
ly, this field may be absent) .
. Hence, for a closed system or one in a stationary force field, we
have fJL/fJt = 0. Therefore,
dL 8L • 8L • •
dt= ~ ~ql!. ql!. + ~ -.- q, (8.1)
11. 11. aqk
If a system is conservative, the motion of the particles obeys

Eqs; {4.16). In accordance with these equations, we shall substi-
t~te ;t
0.L for :L . Expression (8.1) can therefore be written as
aq~~. ql>.
'·'
1 . Homogeneity signifies identical properties at all points. Isotropy signifies
identical properties at each point in all directions. Homogeneity and isotropy
are independent of each other. A medium may have different properties at
different points, the properties at one point differing from those at others, but
being the same in all directions. Such a medium will be non-homogeneous, but
isotropic. A medium is possible whose properties are the same at all points, but
differ (in the same way at all points) in different directions. Such a medium will
be homogeneous, but anisotropic. It is quite obvious that there may be homoge-
neous and isotropic media, and also non-homogeneous and anisotropic ones
CONSERVATION LAWS 37
follows:
~; = ~
k
! (a~ ) q" + ~
{}qlt k
a_L ~
{}qk
q" = :t .~
It {}qk
a.L qlt
The latter expression can ·be given the form
d "' oL • dL d ("' rJL •
Tt ~ -.- q11- dt" = lit ~ -.- q"- L ) = 0
11 {}qlt 11 {}qk
According to formula (5.1), the quantity in parentheses is the ener-

gy E of the system. We have thus arrived at the statement that
dE/dt = 0, whence
E = const (8.2)
Hence, the homogeneity of time leads to the following law: the

energy of a closed conservative system of particles (or a system' in a station-
ary external force field) remains constant. -
Inspection of (5.1) shows that E is a function of the generalized
coordinates qlt and the generalized veloc.ities qlt. Functions of the
quantities q" and qh that retain in motion a constant value deter-
mined by the initial conditions are called integrals of motion. Couse..:
quently, the energy of a closed system is an integral of motion.
9. Momentum Conservation
Consider a closed system of particles. Closure signifies that the
action of external bodies on the system's particles is negligibly
small. Owing to the homogeneity of space, the displacement of all
the particles of the system through an identical length l\r must not
change the mechanical properties of the system-the Lagrangian
must retain its previous value. For an unclosed system, such a trans-
lation would cause a change in the arrangement of the particles
relative to the bodies interacting with them, which would affect the
system's mechanical properties. Therefore, we may state only for
a closed system of particles that the parallel translation of the system
as n whole is not nttended by a change in the function L (i.e. l\L = 0).
Assuming the displacement l\r to be very small, we can write
(9.1)
1 See the footnote on p. 17. In accordance with what is said in this footnote,
ora. of the particles are the same and
equal 8r.
According to our assumption, 8r =fo 0. It thus follows from (9.1)
that
(9.2)
Lagrange's equations (3.8) allow us to write

iJL
lJL d lJL d lJL
Multiplying the first, second, and third of these equations by the

1mit vectors ex, e~., and ez, respectively 1, and summating them, we
<>btain the expression
lJL d lJL
ora =dt ava
(9.3)
Equation (9.2) can thus be written as
.!!:.._~~-0 (9.4)
flt LJ iJv~ -
a
iJL lJL 8L
The quantity-8- is a vector with the components-,-=-:--,
Va. vllax iJ:ca
8L
-,-=-.-,
va., oya
8L
and -0-=-.-.
8L
Vaz8za.
iJL .
Accordmg to (4.17), these products
are the projections of the conventional (not generalized) momentum
Pa of the a-th particle onto the coordinate axes. Hence,
iJL
--=
iJVa.
r~ (9.5)
With account taken of this circumstance, Eq. (9.4) will be written
as follows:
1 Instead of the symbols i, j, k, we shall designate the unit vectors of th&

coordinate axes by the symbol e with the relevant subscript.
CONSERVATION LAWS 39
Hence it follows that
p = ~ Pa. = const (9.6)

a.
where pis the total momentum of the system.
Proceeding from the homogeneity of space, \Ve have thus arrived
at the following law: the total.momentum of a closed system of parti-
cles remains constant. Consequently, the momentum of a closed sys-
tem is also an integral of motion.
10. Angular Momentum Conservation

Owing to the isotropy of space, the mechanical properties of
a closed system of particles should not change upon the arbitrary
rotation of the system asa whole in space. Accordingly, the Lagran-
gian should also remain unchanged (6L = 0). Let us find the incre-
ment of the Lagrangian fJL in an arbitrary very small rotation of
a system through the angle fJq>. All the vectors characterizing the
system will rotate together with it. As a result, they will receive
certain increments that will be of the same order as 6q>. According
to formula (VI.46)
6ra. = [fJq>, ra.], 6va. = [6q>, v,J (10.1)
Owing te the smallness of the quanti ties 6r a. and 6v a.
2j -_,-
aL 6ra. + ~ aL 6va.
fJL = ura.
-,-
uV a.
a. a.
{we remind O\lr reader that L = L (ra., ra.) = L (ra, va.), and L does
not contain the time explicitly]. With a view to (10.1),
.~: " 8L ""\:I 8L (10.2)
vL= £..J ora [6q>, ra.J+ LJ ava [6q>, Va.]
a. a.
It is known from' vector algebra that a cyclic transposition of the
multipliers may be performed in a scalar triple product [see formu-
la (VI.3)l. Such a transposition in (10.2) yields
6L = ~
a.
r
6q> r a., :r~ ]+ 2j a, :v~ ]
a.
oq> [ v
Let us put 8q> outside the sum sign, simultaneously substituting

:t ::: for :r~ in accordance with Lagrange's equations (9.3):
~L = 6cp { 2j [ r a., ~
a.
::: J+ ~a. [ v a., ::: J} = 8q> :t ~ [
a.
r a. • :v~ ]
"OV MECHANICS
According to our assumption, <S<p =F 0, therefore the condition

6L = 0 is equivalent to the condition
~t ~ [ r ,., ::: J= 0
IX
Ol
2j [ r~,· aL
ava. J=canst
IX
According to (9.5), 8L/8va. is the conventional momentum Pa.·

The quantity M = [r, p] is the angular momentum (moment of mo-
mentum) of the relevant particle with respect to the origin of coor-
dinates. We have thus arrived at the statement that
M = ~ Mcx =~[rex, Po:J =canst (10.3)

<X <X
In this relation, Mcx is the angular momentum of the a-th particle,
and M is the resultant angular momentum of the system.
With a view to the isotropy of space, we have thus arrived at the
following law: the resultant angular momentum of a closed system of
particles remains constant. Consequently, the angular momentum of
a closed system, like its energy and momentum, is an integral of
motion.
Assume that a system of particles is in an external central force
field, i.e. a field in which the force action on any of the particles
has a direction passing through the same fixed point 0 (the centre
of the field), while the magnitude of the force depends only on the
distance r to the particle from this point. The potential energy of
a particle in such a field is
U = U (r) (10.4)
Arbitrary rotation of the system in space about the point 0 does
not change the mechanical properties of the system (the arrange-
ment of the particles relative to the force centre 0 remains constant
in such rotation). Hence, although in the given case the system of
particles. is not closed, its angular momentum will be constant.
True, this holds only for the angular momentum with respect to
the point 0. For a closed system, however, the angular momentum
with respect to any point is conserved.
If an external field has axial symmetry (this signifies that the
potential ~mergy of a particle depends only on the distance R to the
particle from the given axis), the mechanical properties of the sys-
tem will not change upon rotation about the axis of the field. There-
fore, the angular momentum (moment of momentum) of the system
relative to this axis will be constant (we remind our reader that the
moment relative to an axis is defined as the projection onto this
axis of a moment taken relative to any of the points on the axis).
Chapter Ill
SELECTED PROBLEMS
IH MECHANICS
U. Motion of a Particle in a Central Force Field
Consider the motion of a particle in a central field of the kind

U(r)=.!.
r
(11.1).
where a is a constant that may be either positive or negative. A posi-
tive constant corresponds to repulsion of the particle from the force
centre (for instance, the Coulomb force of repulsion), and a negative
one to attraction of the particle to the centre (the Coulomb force of
attraction or the force of gravitational interaction of the particle
with a stationary particle at the centre of the field).
We established in Sec. 10 that the angular momentum of a particle
remains constant in a central field, i.e.
M = [rp] = canst
A vector product is perpendicular to the plane containing the vectors
being multiplied. It follows that with a constant direction of the
vector M, the vector r is always in one plane perpendicular to M,
and the trajectory of the particle is a pJane curve. We shall deter-
mine the position of the particle with the aid of the polar coordi-
nates rand .-p, making the origin of coordinates coincide with the centre·
of the field. In these coordinates, the Lagrangian is [see formu-
la (6.1)1
1 · •
+ a
L = 2 m (r2 r2rp 2) ---;-
The function L does not contain the coordinate cp explicilty. Gener-
alized coordinates not contained in a Lagrangian explicitly are·
called cyclic. In the abscnc.e of non-potential fot'CC's, Lagrange's
equations corrcsroncting to c.yclie coordinates are as follows:
~ oL =O
dt •
{}qk
Hence,
aL
Pk = -.- = const (11.2)·
f)qh
Therefore, the generalized momenta corresponding to the cyclic·

coordinates are constant, i.e. are integrals of motion.
-42 MECHANICS
In the problem we are considering, Eq. (11.2) has the form

fJL •
pf(l = -. = mr2cp = M = const (11.3)
{JCll
We could have written this equation immediately, taking into

.account that mr 2 cp is the angular momentum of the particle relative
·to the origin of coordinates, which must be conserved in a central
·force field.
The energy of a particle in a central field is also an integral of
motion. Consequently, calculations by formula (5.11) allow us to
-write
1 • · a
E = 2- m (r 2+r2cp 2) + -r = const (11.4)
To find the trajectory of a particle, it is better to proceed from

Eqs. (11.3) and (11.4) than from Lagrange's equations. This way is
-simpler because Lagrange's equations contain second derivatives
·-of the coordinates, whereas Eqs. (11.3) and (11.4) contain the first
.derivatives of the coordinates with respect to time.
Excluding q> from Eqs. (11.3) and (11.4), we obtain
1 • M" a
E=-mr 2+--+-
2 2mr2 r
~whence
• dr
r=-=-
dt
1
m
V 2mE----- 2am
r
Ma
r2
:From Eq. (11.3)
• dcp 1 M
<p = Tt =--;- ---;:2
Excluding dt from the last two equations, we find that
Introducing the notatioA
2mE+ ( ' ; r =bz, ( - ~ dr=du)

-we can write that
cp-
-- J
r du
Vb2-u2
=cos-t ~ +\Po
"
-where cp 0 is an integration constant.
SELECTED PROBLEMS IN MECHANICS 43
Returning to our previous notation, we get an equation of a parti-

cle's trajectory in polar coordinates:
IP- <Jlo = cos-t -:-;~a=m=:/M;::=:;=+==M=/;=:r~

V2mE+(am/M)2
i+(M2Jam) (1/r)
('1 Ui)
+
V 1 ·u-.·M~/mat
Inspection of Eq. (11.5) shows that at a preset value of r, the
difference cp - cp 0 can have two values differing in their sign
{cos(-a)=cosal. Itisthusa simple matter to conclude that the
curve described by Eq. (11.5) is symmetrical relative to the straight
line making the angle q> 0 with the axis from which <:p is measured.
To reveal the nature of the curve described by Eq. (11.5), let us
introduce the notation
M2
liiTTit= p ( 11.6)
V 1 + 2EM 2 /ma 2 = e (11. 7)

The equation of the trajectory thus becomes
1 ± p/r
<:p- cp0 = cos-t __::::....:..__
e
<>r, after simple transformations,
r==F P (11.8)
1-e cos (cp-cpo)
(the upper sign corresponds to repulsion, and the lower to attraction
<>f the particle to the centre of force).
The equation we have obtained is one of a conic section (or conic-
see Appendix IV) with the focal parameter p and the eccentricity e.
Let us first consider repulsion (a > 0). In this case, U> 0 so that
the total energy E cannot be negative. Hence, by (11.7), we have
e > 1. Consequently, in repulsion, the particle's trajectory can only
be n hyperbola. Taking the upper (minus) sign in (11.8), we get the
.eqHation of the trajectory
r= -p
1-e cos (<p-<po)
The value of cp 0 is determined by the choice of the reference axis
for cp. If the angle cp is measured from the axis of symmetry of the
eurve (from the straight line passing through its foci), r should not
change when the sign of cp changes. This occurs only when qJ 0 = 0 or

44 MECHANICS
coinciding with Eq. (IV .14) that describes the right-hand branch of
a hyperbola [provided that the origin of coordinates, i.e. the centre
of force, is placed at the outer (left-hand) focus of the hyperbola].
Assuming that CJlo = nand taking into account that cos (cp- n)=
= -cos qJ, we get tho equation
r- -p
- 1+e cos cp
coinciding with Eq. (IV.13) that describes the left-hand branch of
a hyperbola [provided that the origin of coordinates is placed at the
outer (right-hand) focw; of the
hyperbola; Fig. 11.1].
U=~
Ia I I
Now let us consider attrac-
r I
I tion (a< 0). The lower (plus)
(repulsion) I
I sign in formula (11.8) corre-
I sponds to it. l-Ienee, the equa-
I
I tion of the trajectory is
I
I r= P (11.9)
I 1-e cos cp
I
for cro = 0, and
\ Force
\ centre r = 1+epcos cp ( 11. 10)
\
\ for cro = n.
\
As shown in Appendix IV,
\
\ both equations describe either
\
\ an ellipse, or one of the
branches of a hyperbola,or a pa-
\
rabola [see Eqs. (IV.11)
and (IV.12)l. The valile of e
Fig. 11.1. determines which of these
curves we have to do with.
In attraction, U < 0, consequently the total energy E may be
either positive or negative, particularly, it may be zero. It follows
from fnrmula (11. 7) that when E > 0, the eccentricity is greater
than one, and the trajectory will be a hyperbola. Equation (11.9)
gives the right-hand branch of a hyperbola, and Eq. (11.10), the
left-hand one. Unlike repulsion, the origin of coordinates, i.e. the
centre of force, is at the inner focus for the given branch (Fig. 11.2) 1 •
AtE = 0, the eccentricity e=1, and the trajectory will be a parab-
ola. This case occurs if a particle begins its motion from a state of
rest at infinity.
1 The solid curves in Fig. 11.2 are depicted for the same valuP of the angular
momentum M and, consequently, for the same value of the focal parameter p.
The dashed ellipse corresponds to a smaller value of M. For a smaller M, the
vertex of a parabola may be to the right of that of a hyperbola corresponding to
a greater M.
Finally, at E < 0, the eccentricity is less than one, and the tra-
jectory will be an ellipse. In this case, the curves described by Eqs.
(11.9) and (11.10) differ in the position of the centre of force. Curve
(11.9) is obtained if the centre of force (the origin of coordinates)
-tit~
\ U= -"F
\ (attraction)
\
\
\
\ E > 0, hyperbola
\
\ E = 0. parabola
\ E < 0, cllip~e
\
\
\ ---... ......
I
I
"'
I
'II
I I
I
I
I
I --
Force
centre
/
/
I
I
I
I
I
I
I
I
I
Fig. 11.2.
is at the left-hand focus of the ellipse. Curve (11.10) corresponds ts
the centre of force being at the right-hand focus.
If 1+ (2EM 2 /ma 2 ) = 0, i.e. E = -ma2 /2!Vl 2 , a glance at for-
mula (11.7) shows that the eccentricity vanishes-the trajectory is
a circle. At a given M, such an energy value in the given force
field is the minimum possible one (an imaginary value of e corre-
sponds to smaller E's).
When a particle moves in a restricted region of space (the particle
does not travel to infinity), the motion is called fmite. In infinite
motion, a particle travels to infinity. Motion in an ellipse is finite
(recall that in this case E < 0), and motion in a hyperbola or pa-
rabola is infinite (E 0). >
12. Two-Body Problem
Consider a closed system formed of two interacting particles of
masses m1 and m 2 • The potential energy of the system depends on
4fi MECHANICS
the distance· r between the particles. This distance can be treated

as the magnitude of the vector r drawn from one particle (say m 2 ) to
the other one (m 1): U = U (r) (Fig. 12.1).
The system has six degrees of freedom. We shall take as the gener-
alized coordinates the three Cartesian coordinates of the centre of
mass C of the system (which the position vector R of the point C
Fig. 12.1.
is equivalent to) and the three projections of the vector r onto the
coordinate axes (the vector r is equivalent to them). The vector r
can be represented as
r (12.1)
where r 1 and r 2 are the position vectors of the particles relative to the
centre of mass. According to the definition of the centre of mass
(12.2)
S~lving Eqs. (12.1) and (12.2) simultaneously, we find that

(12.3)
The position of the particles relative to the origin of coordinates

is determined by the vectors R +r 1 and R +
r2.
Let us write the Lagrangian of the system:
m • • m • •
L=z!- (R +rJ +T- (R+r
2 2) 2 -U (r)
SELECTED PROBLEMS IN MECHANICS 4T
Squaring and taking into account that by (12.2) we ·have m 1r 1

.+
+ m 2 r 2 = 0, we obtain
L= mxtm2 R2+ ~i ;i+ ~2 ;;-u (r)

Finally, substituting ; for ; 1 and ; 2 in accordance with (12.3), we·
arrive at an expression for L in the "coordinates" R and r which we
have adopted 1 :
(12.4)>
where
is a quantity known as the reduced mass of the system.

Function (12.4) breaks up into two independent terms:
L =L
. + L (r, r).
(R)
The first of them describes the behaviour of the centre of mass of the
!lystem, and the second, the motion of the particles relative to the·
centre of mass.
A glance at (12.4) shows that the coordinate R is cyclic (it is not
contained explicitly in L; see Sec. 11). Consequently,
PR = (m1 +m 2) R = const
(the momentum of the system is conserved; this could have beeu fore-
seen because the system is closed). The centre of mass of the system
moves rectilinearly· and uniformly (or is at rest).
The motion of particles relative to the centre of mass is described
by the function
(12.6}
It can be considered as the Lagrangian of a particle of mass j..l moving
in a central field with a stationary centre. The position of a particle
relative to the force centre is determined by the position vector r.
Therefore, the problem on the motion of a system consisting of two
interacting particles (the two-body problem) has been reduced to-
the problem on the motionofone particle in a central force field. We
have treated this problem in Sec. 11. We established that when U =
=air, the trajectory of a particle will be a conic. Consequently,
the tip of the position vector r = r 1 - r 2 slides in motion of the
I The projections of the vectors R and r onto the coordinate axes, and not the-
Yectors themselves, are the actual coordinates.
-48 MECHANICS
particles along a curve that is a conic. By (12.3), the position vec-

tors r 1 and r 2 are proportional to the position vector1 r. Therefore,
·each of these vectors also circumscribes a conic. Depending on the
nature of the interaction (attraction or repulsion) and the magnitude
-of the total energy of tho system, the trajectory of each of the parti-
des will be either an ellipse, a parabola, or a hyperbola.
/
I
/'
/'
........... ........ -- ............. .._. ______ ,../
Fig. 12.2.
Assume that an auxiliary imaginary particle ~ moves in an el-

lipse having the equation
r= P
1-ecos cp
According to (12.3), the vector r 1 at any instant has the :3alll.e direc-
tion as r (i. e. cp 1 = cp), while its magnitude is m~l(m 1 + m 2 )
times greater. Hence, the particle m1 moves in the ellipi3e
r _ P1 (12; 7)
1- 1-ecos<p 1
where p 1 = pm 21(m 1 + m 2 ).
The vector r 2 at each instant is directed oppositely to the vector r
'lsee (i2.3)1. Consequently, when the vector r is oriented at the an-
-gle cp, the vector r 2 is oriented at the angle cp 2 = cp + :n:. The magni-
tude of the vector r 2 is m 1/(m 1 + m 2 ) times greater. Hence, having
in view that cos (cp 2 - :n:) = - cos cp 2 , the equation of the ellipse
.along which the particle m 2 travels must be written as
rz= 1+ecos<p2 (12.8)
where Pz = pm/(m 1 + mz).

1 For this reason, we must assume that the imaginary centre of force under
whose action the particle !J. moves is situated at the point from which the vectors
T 1 and r 2 emerge, i.e. at the centre of mass of the system C.
SELECTED PROBLEIIIS IN MECHANICS 49
In the case corresponding to Eq. (12. 7), the origin of coordinates

(i.e. the centre of mass C) is at the left-hand focus of the elliuse [see
formula (IV.11)l. In the case corresponding to Eq. (12.8), the origin
of coorrlinntes (the point C) is at the right-hanrl focus of the ellipse
[soo formula (IV.t2)l. Consequently, the trajectories of the pnrt.iclos
are as shown in Fig. 12.2 1•
!-'- \
\
(a) (b)
Fig. 12.3.
We invite our reader to see for himself that in motion along hyper-
bolas the trajectories of particles will appear as shown in Fig. 12.3a
(for mutual attraction of the particles) and in Fig. 12.3b (for repul-
sion).
13. Elastic Collisions of Particles

A collision is dul"ttwd to be a process consisting in that particles
interacting with each othet· and al'fiving fmm inl'tnil.y (i.e. from
a distance such that their interaction may be disregarded) approach
each other. and then either recede again to infinity, or remain at
a finite distance from each other. In the first case, the collision is
called the scattering of the particles, and in the second, their capture.
The latter can obviously be observed only if the interaction of the
particles has the nature of attraction.
When we speak about collisions of particles, we do not at all
assume that the particles come into contact, as is the case. for in-
stance, in the collision of two spheres. What we have in mind is only
1 Fij:t'ures 12.2 and 12.3 are for mtlmi = 2/3; in Fig. 12.2, e = 0.8, ani in
Fig. 12 ..'1, e = 1.5.
50 MECHANICS
the fact that owing to their interaction, the particles change the
direction or nature of their motion.
Consider the elastic collision of two particles repelling each other.
A collision is defined as elastic if it is not attended by a change in the
internal energy of the particles. Consequently, in elastic collisions,
the mechanical energy of the system of colliding particles remains
constant.
It is the simplest to consider a collision process in a reference frame
associated with the centre of mass of the particles (it is known as
a c-frame). In practice, however, collisions are observed in a refer-
ence frame relative to which the centre of mass of the particles
moves with the velocity vc. This reference frame is called a labora-
tory one or, more briefly, an Z-frame. In a laboratory frame, one of
the particles is usually at rest before a collision.
The following relation obviously holds between the velocity of
the i-th particle in an Z-frame (we shall designate it by the symbol vi)
and the velocity of the same particle in a c-frame [we shall dt>signate
this velocity by v~c>]:
Vt = Vc + Vt (C)
(13.1)
It follows from the definition of the centre of mass that
m 1 v 1 +m 2v 2
vc = --''--"--'--=--=-
ml+m2
In the following, we shall treat only the case when in an Z-frame the
second particle is at rest before a collision. Therefore, designating
the velocity of the first partide before the collision by the sym-
bol v 10 , we have
(13.2}
Using this value o{ Vc in formula (13.1), we get the following expres-

sions for the velocities of the particles in a c-frame before the colli-
sion:
vl~) = ml~m2 Vio }

(13.3)
y(C) =- mi V1o
20 m1 +m 2
Multiplying the first of these velocities by m1 and the second by m~,
we find the momenta of the particles before the collision in the
c-frame:
Pl~l = 1-tV1o
P~~) = -1-tVlo
Here 1.1. = m 1m 2 /(m1 +m 2} is the reduced mass of the particles

As expected, the total momentum of the particles in the c-frame

before they collide is zero. It follows from the law of momentum
conservation that after the collision too, the momenta of the particles
in lhe c-frnmo can differ only in their sign: p\Cl '-== -p~Cl,
The total kinetic energy of the particles ns n resnll of an clastic
collision] cannot change (we assume that both before and after
colliding the particles are so far from each other that their mutual
potential energy is negligibly small). We can therefore write the
relation
2
+ 2m 2
2
2
2m 1
+ 2m
2
..fu.... ...l!.1fL = _P_t_ ..ll_
2m 1 2
[we have omitted the superscript "(C)" on the symbols of tho momon-
tal. In combination with the condition that I p<,~l I = I p~~l I and
1p\C>I = I p~C>I, the relation we have written indicates that the
momenta (and, consequently, the velocities) of the particles as a re-
sult of colliding only turn in the c-frame through a certain angle "f.,
remaining constant in magnitude. Let e 1 stand for the unit vector of
the first particle's velocity in the c-frame after the collision. In
accordance with formulas (13.3), we can therefore write the following ·
expressions for the velocities of the particles after the co1lision: ·
To obtain the velocities of the particles after the collision in the

Z-frame, we substitute expression (13.2) for "c and the above expres-
sions for v~c) in formula (13.1). The result is
Vt= mt
mt +m2 Vto
+ ml m2
+m2 v!Oet
Vz = mt
m 1 +m 2
Vto-
m1
~ m2 v!Oet
We obtain the following expressions for the momenta of the parti-

cles in the Z-frame after the collision:
Pt= m ~m 2
1 '
Pzo+ m
1
n+~2 m 2 Ptoel}
(13.4)
m. m2
P2 = mt +m2 Pto- mt +mz Pwel
The toliuwmg geometrical construction will be an excellent illus-
tration of the obtained relations. Let us depict the vector p1 o by
the segment AD (Fig. 13.1) and let the point 0 on it divide the length
of the vector in the ratio m 1 : m 2 • We draw a circle passing through
the tip of the vector p10 with the point 0 as its centre. The radius of
52 MECHANICS
this circle is PH,m 2 /(m 1 + m 2 ). If m 1 < m 2 , the point A will be

inside the circle (Fig. 13.1a); if m 1 > m 2 , the point A will be outside
the circle (Fig. 13.1b); if m 1 =
= m 2 , the point A will be on the
circle (Fig. 13.1c). We lay of[
from the point 0 at the angle
X relative to p10 the unit vector
~~~~::::=:::==~ D e 1 of the direction in which we
assume the first particle to be
flying relative to the c-frame.
Hence, the segment OB will de-
pict the vector e1p 10m2 /(m 1 +m 2 ),
and in accordance with formulas
(a) (13.4) the segment AB will de-
pict the vector p1 , and the seg-
ment BD, the vector p 2 •
The angle 81 between the vec-
tors p1 and p10 is called the scatter-
ing angle. It characterizes the
A ~=:t==;::====~~==:::~~ D deviation of the first particle
observed in the l-frame. The
angle 8 2 between the vectors p 2
and p 10 is called the recoil angle.
The sum 81 + 82 is called the
divergence angle of the particles
(b)
after the collision. The angles 81
and 8 2 can be expressed in terms of
x-the deflection angle of the
first particle in the c-frame. With
a view to the length of the seg-
ment OB being p 10 m/ (m 1 + m 2),
A \F----'.....;..-~...._-~-l!>D we can write
tan a,=
x
[p 1 0 m,/(m 1 +'ll 1 )] sin
p, ,m 1 /(m +m 2 )+[p 10 m 2 /(m, +m,)) cos X
(c) or
m 2 sin X
Fig. 13.1. tan 81 = m + m 2 cos X (13.5)
1
From the isosceles triangle OBD, we get the relation
82 = n-;-x (13.6)
Inspection of Fig. 13.1a shows that the.ligh~er particl~ can di-
verge on the heavier one (m 1<m 2) in any duect10n (the pomt B can
SELECTED PROBLEJIIS IN MECHANICS 53
be at any place on the circle). The angle of divergence of the parti-

cles in this case is always greater than n/2.
When m1 > m 2 (Fig. 13.1b), the scattering angle CRnnot exceed
fl certain extreme VRlue e\,111HX (the point n' in the figure corre-
sponds to it). The sine of this angle equals the ratio of the segments
OB' and OA, i.e.
. e1• max=-~-,;-
SID
m2
1
When m1 > m 2 , the divergence angle of the pRrticles is always less

than n/2.
If the masses of the particles are the same (m 1 = m 2 ), the parti-
cles after colliding fly apRrt at right Rngles to each other (8 1 +
+ 8 2 = n/2; Fig. 13.1c).
In a head-on collision, the particles fly apart at the angle el e'.l
-I_
equal either to n (when m 1 < m 2 ; Fig. 13.1a) or to zero (when m 1 >

> m 2 ; Fig. 13.1b). The angle x in a head-on collision is n. When the
masses of the particles are the same (m 1 = m 2 ), the momentum p1 is
zero, and p 2 = p10 (see Fig. 13.1c; in this case, the point B coincides
with the point A). Consequently, particles of the same mass e:xchange
momenta in a head-on collision. This result can also be obtained
quite easily from formulas (13.4).
The results we have obtained are a corollary of the laws of energy
and momentum conservation and do not depend on the nature of
particle interaction. To determine at what angle X a particle di-
verges, we must lmow the law ofinteraction of the particles and their
mutual arrangement in colliding. The following section is devoted
to a treatment of this matter.
14. Particle Scattering

We showed in Sec.12that the problem on the motion of two inter-
acting particles reduces to the problem on the motion of a particle
of mass f.l (f.l is the reduced mass) in a central force fiCld, the distance
from this partielo to lh(' n'lltrc of forcos lwing ('(]!lfll to the distanec
between the pal'l.ielcs in question. Afler l'1nding the t.rnjer.lory of llw
imaginary particle of mass ~t, it is a simple matter to lind the trajec-
tories of both particles.
We shall use this procedure for studying the process of the diver-
gence of the particle m 1 by the particle m 2 which is initially station-
ary in an Z-frame. Let us pass over to a c-frame and consider the
particle ~t moving in a force field whose centre coincides with the
centre of mass C of the system. \Ve shall consider the fteld to be so
weak at large distances from the centre that the motion of a particle
at these distances may be considered rectilinear.
54 MECHANICS
Let us introduce the impact parameter b equal to the distance

from the centre of force at which a particle would fly past it if the
field did not act on it (Fig. 14.1). It is evident that the angle of
deflection of a particle is a function of the impact parameter: X =
= X (b), and, generally speaking, x should grow with diminishing b.
Inversion of this function yields
b = b (X) (14.1)
Assume that a beam of identical particles flying far from the
centre of force C in the same direction and with the same velocity v0
',
''" "\.
"\.
"\.
"\
'
''fo
~
x·v~
~-----~~~Y-:~=-;/r~r----
------~~J~------
c ----
Fig. 14.1.
is incident on the centre. The beam can be characterized by the den-

sity of the particle flux j, by which is meant the number of particles
flying a second through unit area at right angles to the bea'm. We
shall assume that the beam of particles is homogeneous, i.e. that
far from the scattering centre the flux density is the same at all points
of the beam's cross section.
The beam particles deflect through different angles x depending
on the ·impact parameter of a particle approaching the centre. Particles
whose impact parameter ranges from b to b db will be scattered +
within the limits of angles from x to x dx. Let us denote the flux+
of such particles (i.e. the number of particles scattered at angles from
x to x + dx in unit time) by dN'I.. The ratio
da = dNx. (14.2)
j
is known as the differential effective cross section of scattering. One
reason why this name was introduced was the circumstance that da,
as follo\vs from (14.2), has the dimension of area. It is simple to see
that da determines the relative number (fraction) of partieles scat-
tered within a given range of angles.
SELECTED PROBLEMS IN lltECHANICS 55
vVhen a beam has a homogeneous cross section, the flux of particles

whose impact parameter ranges from b to b +db is j2nb db (the
flux equals the flux density times the area). This flux is scattered
at angles from X to X + dx. I-I once, dN x = j2nb db. Using this
value in formula (14.2), we obtain
r:kJ = jZnb db 2nb db

j
or, passing over from the variable b to x [see (14.1)],
da = 2nb (x) I ~~ Idx (14.3)

(we have taken the absolute value of dbldx because dbldx is less
than zero).
The quantity dNr. in formula (14.2) can be interpreted as the flux
of particles flying within the solid angle dQ = 2n sin x dx (this is
the value of the solid angle confmed between cones with apex angles
of x and x + dx). Substituting dQ/sin x for 2n dx in (14.3), we can
reduce the formula for the differential effective cross section of
scattering to the form
dcr= b.(x) l!!:!!_ldQ (14.4)
smx dx
Formula (14.3), like (14.4), is the most general-it determines the
differential effective scattering cross section for any central scatter-
ing field. The quantities b (x) and dbldx are determined by the
nature of the force field, i.e. by that of particle interaction. Con-
sequently, dcr is determined by the kind of scattering field and is the
most important characteristic of a scattering process. We can pro-
cure information on the nature of a force field by experimentally
studying dcr.
Up to now, we have dealt with the scattering of a beam of parti-
cles on one scattering centre. In practice, however, scattering occurs
on a collection of identical scattering centres. In this connection,
we shall note the following circnmstanco. Tho deflection angles X
are appreciable only for particles that approach the scattering
centre sufficiently closely (for which the impact parameter is small).
Therefore, if there are n identical, not overlapping, and sufficiently
rarefied scattering centres in the path of particles, they will scatter
the particles independently of one another, and the flux of particles
deflected 'vithin the range of angles from x to x +
dx will ben times
greater than when there is only one centre. Hence, with n centres,
we have
dNx = nj dcr (14.5}
Let us now go over from a particle of mass fl. deflected by a station-
ary force centre· at the point C to real particles m 1 and m 2 • The
56 MECHANICS
trajectories of these particles are geometrically similar to that of

the particle 1-t· Indeed, according to (12.3), the position vector of
the first particle emerging from the point C is m 2 /(m 1 +
m 2) times
longer than the position vector of the. particle 1-t emerging from the
same point. A similar relation also holds for the second particle.
This signifies that the particle m1 deflects through the same angle x
as the particle 1-t in a c-frame. The impact parameter must he taken
equal to tho distance at which tl1e first particle would fly past tho
second one if the particles did not interact (the position vector r of
the particle 1-t emerging from the force centre C coincides with r 1 - r 2 ).
Hence, formula (14.3) also holds (in a c-system) for a beam of
particles m1 scattered by the particle m 2 • To go over from the c-frame
to the laboratory reference frame in which scattering is being
observed, we must transfer from the variable X to the variable 91 in
formula (14.3). We use formula (13.5) for this transition. There-
sulting formula is very cumbersome in the general case.
In the particular case when m1 « m 2 (the particles being scat-
tered are much lighter than the ones scattering them), 91 ~ x [see for-
mula (13.5)] so that formulas (14.3) and (14.4) can be written in
the Z-frame:
da = 2nb (8 1) I :: Id9
1 1}
(14.6)
I I
da = b. (81) _j}!_ dQ
sm 8 1 ae 1
We must note that in this case 1-t ~ m1 and r ~ r 1 (the trajectory of
the particle m 1 virtually coincides with that of the particle ~t).
Consider a Coulomb scattering field, i.e. a field of the kind U =
= air, assuming that m1 « m 2 • The energy of the particle being
deflected can be represented by the expression E = ; m 1 v~, where v0
is the initial (and final) velocity of the particle m 1 • The angular
momentum of the particle m1 relative to the scattering centre (coin-
ciding with the particle m 2 ) is 111 = m 1 v 0 b (see Fig. 14.1). Using
these values of E and M in formula (11.5), we arrive at the relation
_1 1+(m 1 v3bl/a) (1/r) ( 14 7)
!p- !po =cos V 1+(m 1 v~b/a) 2 '
A glance at Fig. 14.1 shows that when r = oo there are two values
of <p-zero and 2!p 0 • In the first case, the left-hand side of formu-
la (14. 7) becomes -<p 0 , and in the second, +~Po· Therefore, assuming
in (14.7) that r = oo, we can write
I
<Jlo = cos-1
V 1+(m1 1 vijb/a) 2 I
whence
(14.8)
Further, it follows from Fig. 14.1 that X = :rt - 2cp 0 , i.e. {Po =
= :rt/2 - x/2 = :rt/2 - 81 /2 (in the case being considered it is possi-
ble to assume that 81 = X)· Introducing this value of cp 0 into formu-
la (14.8), we obtain
sin 2 ~= '1
2 1+(m 1 vfibia) 2
Solving this relation forb, we arrive after i'imple transformaliuns at
the expression
(14.9)
Differentiation with respect to 81 yields

db a '1
d 81 =- m1 vij '""2:-s-:-i1-:1~::-,(""8-
1 /""2:-)
('14.10)
Finally, using expressions (14.9) and (14.10) in (14.6), we get for-

mulas for the differential effective cross section of scattering of the
particles of.mass m 1 in the Coulomb field se(up by the particle of mass
m 2 (here m~ »m 1 ):
d a=:rt ( -a- ) 2 cos (8 1 /2) d8 1 (14.1'1)
m vg 1 sin 3 ( 8 /2) 1
da = (_a_)
2m !J5
2 ---;--;d--,Q~
sin 4 (8 1 /2)
(14.12)
1
We have obtained Rutherford's formula for the scattering of alpha-

particles on heavy nuclei known from the general course of physics.
This can be verified by substituting 2Ze 2 for a and multiplying for-
mulas (14.11) and (14.12) by the flux density i of alpha-particles
and the number of atoms n of the scattering substance per unit cross-
sectional area of the alpha particle beam. The expression ni da will
be obtained on the left-hand side of the formula, which gives dN e,-
the flux of alpha particles scattered in the range of angles from 8 1 to
81 + d8 1 , or dNa- the flux of alpha particles sea ttered in the solid
angle dQ [see formula (14.5)1.
We must note that the expressions we have found for da do not de-
pend on the sign of a so that the result obtained holds not only for
the Coulombian repulsion of tho particles m 1 and m:!, but also for
their Coulombian attraction.
t 5. Motion in Non-Inertial Reference Frames

The Lagrangian of one particle has the form
1
L= 2 mv 2 -U(r) (15.1)
only in inertial reference frames. Let us fmd the form of L in an
arbitrary non-inertial referenr,e frame. Figure 15.1 depicts the iner-
~8 MECHANICS
tial reference frame K and the frame K' whose origin (the point 0')
moves in the frame K at the velocity v 0 (t). The frame K', in addition,
rotates relative to the frame K at the angular velocity ro (t). Let
IUS express the function (15.1) in terms of the vector r' determining
the position of a particle in the
y' K' frame K', and in terms of the
t..l(t) velocity v' of the particle obser-
ved in the same frame.
We assume first that v 0 (t) =
y K, ==
0 and that the origins of both
reference frames coincide. Hence,
X'
the following relation would hold
between the velocities of the par-
ticle in both frames:
0 X
v = v' + [ror'l (15.2)
Fig. 15.1. (a particle that is stationary in
the frame K' would have a veloc-
ity equal to [ror'] 1 in the frame K). If v 0 (t) is non-zero, on the other
band, relation (15.2) becomes
v = v 0 (t) + v' +- [ror') (15.3)
Let us introduce the expression we have obtained for v into
formula (15.1). This yields
m mv'2 m
L =2 (v 0 (t)J2+-
2- +z [ror']2
+ mv 0 (t) v' + mv 0 (t) [ror'] + mv' [ror'] (15.4)
The first term in this formula is the preset function of time,' which
can be represented as the total derivative with respect to t of another
function. We established in Sec. 7 that the Lagrangian must be
-determined to within the additive terms that are the total time
derivative of an arbitrnry function of the generalized coordinates
and time. For this reason, the term (m/2) [v 0 (t)P should be omitted.
Consider the fourth and fifth terms in formula (15.4). Factoring
<>fit mv 0 (t), these terms can be written in the form
mv 0 (t) {v' + [ror'J} = mv 0 (t) ---cit+ [dai
{ d'r' ,r' ]}
_ (t) {d'r' + [dcp, r')}
- mvo dt (15.5)
Here d'r' is the increment of r' observed during the time dt in the
frame K' (we remind our reader that v' is the velocity of the particle
1 This expression is obtained from formula (VI .46) if we assume that a = r'
in it and divide the relation obtained by dt.
observed in the frame K'), and dcp is the angle through which the
frame K' turns during the time dt.
If a reference frame rotates relative to another one, the increment
of a Vl'dot· n obsorYed in both frames will he different. This is easy
to understand by assuming that the vec.tor docs not change with
respeet to the rotating frame, i.e. the increment of the vector in
this frame (we designate it by K') is zero: d'a = 0. Consequently,
the increment of the vector in the stationary frame (the frame K)
can be written as
da = [dcp, a)
[see formula (VI.46)1. If the increment of the vector d'a observed in
the rotating frame is non-zero, the increment observed in the sta-
tionary frame will be
da = d'a + [dcp, a] (15.6)
[assuming that a =r' and dividing by dt, we arrive at formula (15.2)].
By comparing (15.0) with the exprPssion in braces on the right-
hand ~ide of formula (15.5), we arriYe at the conclusion that this
expression is thP increment of the vector r' observed in the frame K,
i.e. dr'. l-Ienee, the sum of the fourth and fifth terms of formula (15.4)
can be written as
dl''
mv0 (t) lit
Let us transform this expression as follows:

dr' d { mv ( t ) r '} -mr , dt
mv 0 ( t ) dt=dt dv0
0
We may discard the first term as the total time derivative of the
function of the coordinates and time. In the second term, dv 0 /dt is
w0 (t)-the acceleration of the origin of coordinates of the frame K'
observed in the frame K.
Wn haYe l.hns nJTived at the following expression for the Lagran-
gian in the variables r' and v':
L' = ! mv'2+! m [ror']2-mr'w0 (t) +mv' [ror'] -U (r') (15.7)
W c have obtained the general form for the Lagrangian of a parti-
cle in an arbitrary non-inertial reference frame. We must now consid-
er that the function U is set in the variables r' [in formula (15.1) it
was set in the variables rl. The transition from one set of variables
to another is accomplished by the formula
r = r 0 (t) + r' (15.8)
where r 0 (t) is the position vector of the origin of coOl'dinates of the
frame K' (see Fig. 15.1).
60 MECHANICS
We must note that even if the time were not contained explicitly
in the function (15.1) (it could have been contained in the force
function U), the function (15.7) contains the time because w 0 and ro
are, generally speaking, functions oft. The time also enters explicitly
the term U (r') as a result of the transition from r to r' accomplished
by formula (15.8).
Before beginning to compile Lagrange's equation, let us replace
the second term of expression (15. 7) in accordance with formula (VI.6).
The result is
L l = 21 mv 12 , 21 mw 2r /2 - 21 m ( ror ')2
I
+
- mr'w 0 (t) mv' [ror']- U (r') (15.9)
Taldng advantage of a cyclic transposition of the multipliers [see
formula (VI.3)], the nc:xt.,.to-last term equal to mv' [ror'l could he
written in the form
m lv'w]r' (15.10)
Lagrange's equation in the frame K' is as follows:
d 8L' 8L
dt 8v' = 8r'
(15.11)
[see formula (9.3) and the footnote on page 171. A glance at expres-
sion (15.9) shows that
8L'
av' = mv' + m [ror']
whence
d 8L'
dt av'
•
= mv' •
+ m [wr'J + m [ror']•
We remind our reader that from the very instant when we expressed
L in the variables r' and v', we have been "living" in the reference
frame K'. Consequently, by ~' we must understand the accelera-
tion 'W.' of a particle observed in the frame K', and by r', the veloci-
ty v' of the particle in the same frame. Hence,
d aL'
dt av' = mw '+ m r·ror 'J+ m rrov ,1 ('15.12)
As regards ro, it is the time derivative of the function ro (t) that is

set in the frame K.
In calculating DL'Ior', we shall assume that the next-to-last term
in formula (15.9) is represented in the form (15.10). We thus obtain
U' W
8T = mro 2 r'- m (ror') ro- mw0 (t) + m [v' ro]- Br' (15.13)
The first two terms in this expression are the triple vector product
m [w, [r'(J)ll written according to formula (VI.5). Expression (15.13)
can therefore be written as
~;: =m[(!l, [r'w]J-mw0 (t)+m[v'w]- :~ (15.14)
Introducing expressions (15.12) and (15. H) into formula ('15. H)

and performing transformations, we arrive at an equation of motion
of a particle in the frame K':
mw' = - 7au -mw0 (t) + m [r'w]
· + m [w, [r' w]] +2m [v'w] (15.15)
Vve see that the acceleration of a particle in the frame K' is deter-
mined, in addition to the force -r7U/8r' due to the force fleld, by
a number of additional forces called, as is well known, forces of
inertia. The term m[m, [r' mll gives the centrifugal force of inertia,
and the term 2m [ v' ml, the Coriolis force. The force m[r' ~] is associat-
ed with the non-uniformity of rotation; it has no special name.
If the frame K' has only translational mo lion relative to th8 frame
K (in this case w = 0 and, consequently, ~ also equals 0), the
equation of motion contains only one force of inertia equal to
f1n = - mw 0 (t) (15.16)
It is remarkable that this force, like the force of gravity mg, is
proportional to the mass of the particle. This circumstance underlies
the general theory of relativity.
For a uniformly rotating system of coordinates having no trans-
lational acceleration [w 0 (t) = 0, ~ = 0], the Lagrangian has the
form [see (15.7)]
L' =
· mv'
--2
2 + m (ror')
2
+ mv , [wr '] -
2 U (15.17)
Let us find the momentum, the angular momentum, and the energy
of a particle for this case. By formula (9.5)
, aL'
p = av'
Taking the derivative of function (15.17) with respect to v', we get
p' = mv' + m [wr'] = m {v' + [w'l} (15.18)
If the frame K' has neither a translational acceleration nor a trans-
lational velocity (v 0 = 0), inspection of (15.3) shows that the ex-
pression in braces in (15.18) is the velocity v of a particle relative to
the inertial frame K. Hence, p' equals mv, i.e. it coincides with the
momentum p of a particle in the inertial frame:
p' = p (15.19)
Further, if the origins of the frames K and K' coincide [see
Fig. 15.1)], the position vectors r and r' also coincide. Hence, with
a view to (15.19), it follows that the angular momentum M' =
= [r'p'] in the frame K.' coincides with the angular momentum
M :- [rp] in the frame K:
M' =M (15.20)
By formula (5.1), the energy of a particle in the frame [(' is deter-
mined by the expression
E ' = ""'
L.J -aL' ·, L'
. - Xt-
i axi
where xi are the Cartesian coordinates of the particle in the frame K'.
According to (4.17), 8L'ID;i is pi-the projection of the momentum
p' of the particle onto the i-th coordinate axis, and ;i is the projec-
tion of the velocity v' of the particle onto the same axis. Consequent-
ly, the expression for the energy can be written as
E' = p'v'- L' (15.21)
Substituting for p' its value from (15.18) and for L' expression (15.17),
we get the following formula:
1 mv' 11 m ,
E = - 2- + U - -
2
[ror ]2 (15.22)
Rotation of the reference frame manifested itself in the appearance

of the term
(15.23)
not depending on the particle's velocity u' in the expressiori for the
energy. This additional "potential" energy is called centrifugal.
Let us substitute v -[ror'] for v' in formula (15.22) [see (15.3);
we assume that v 0 ( t) = 0]. The result is
E' = mt +U -mv [ror'J (15.24)
The first two terms give the energy E of a particle in the frnme [(.
If the origins of the frames K and K' coincide, r' may be replaced
by r. The last term in (15.24) by means of cyclic transposition can
now be givev. the form
mv [ror] = ro [r, mv) = rol\:1
Therefore, the following relation holds between the energies E and
E' of the particle in the frame~ K and K' respectively:
E' = E - roM (15.25)
SELECTED PROBLEMS IN MECHANICS 63-
We remind our reader that this formula has been obtained assuming·
that the origins of both reference frames coincide. Consequently,
instead of M in formula (15.25), we can write M' [see (15.20)1.
Summarizing, if the reference frame K' rotates uniformly relative
to the reference frame K, and the origins of both frames coincide,
the momentum and the angular momentum of a particle in both
frames coincide, while the energy of the partiele in the frame K' is
less than that in the frame K by the magnitude of the scalar product
of the vectors w and M.
Chapter IV
SMALL·AMPLITUDE
OSCILLATIONS
16. Free Oscillations of a System Without Fridion
Consider a system with one degree of freedom in which friction

forces are absent. The potential energy of such a system has the
form U = U(q), where q is a generalized coordinate. The potential
energy is known to be minimum in the position of stable equilibri-
um. \Ve shall measure q from this position. Let us expand the func-
tion U (q) in powers of q in close proximity to the point q = 0.
Owing to the smallness of q, we shall limit ourselves to the ftrst terms
of the expansion:
u (q) = u (O> + U' (O) q+ +u" (O) q 2
The condition of equilibrium yields U' (0) = 0. Let us measure

the potential energy from the equilibrium position, i.e. assume
that U (0) = 0. Finally, let us introduce the symbol U" (0) = x
(remember that the second derivative is positive at a point of a min-
imum, hence, x > 0). As a result, we arrive at the expression
xq2
U(q)=-2- (16.1)
We shall consider the constraints to be stationary. Therefore,
by (5. 7)
T = y (q) q2
.
In P<lssing through the equilibrium position, T does not vanish.
Consequently, y (0) is non-zero. Expanding y (q) into a series and
retaining only the zero term of the expansion owing to the smallness
Of q, we can write
.
!J.q2 (16.2)
T=-2-
where ll = 2y (0) (do not confuse it with the reduced mass!).

Let us compile the Lagrangian:
(16.3)
SMALL-AMPLITUDE OSCILLATIONS 65
Lagrange's equation
11q + xq = 0 or q + w~q = 0 (16.4)
[here w! = (x/11) >0] is a linear homogeneous second-order differen-
tia I eq 11<1 lion with conslnnl codliden ts (seo Appendix V). Using
the substitution q = eA-t, we arrive at the characteristic equation 1
')._2 +
(J)~ = 0
The roots of this equation are 'A1 = +iw 0 and A. 2 = -iw 0 • Con-
sequently, the general solution has the form
q = C lelooot + C 2e-iooot (16.5)
where C1 and C 2 are complex constants.
The values of q must be real; this signifies that the condition q* =
= q (q* is the complex conjugate of q) must be observed. Introducing
expression (16.5) for q into this condition, we obtain
Cfe-iooot + C2eiooot = C1eioo0t +
C 2 e-ioo 0 t
The above relation is observed if C1 = C2 (correspondingly Ct = C2 ).
Having this in view, we shall write the coefficients C1 and C2 as
C1 = Teta,
a . a .
C 2=-e-ta (16.6)
2
(a and a are arbitrary real constants). The use of these values is
formula (16.5) yields
(16. 7)
Therefore, the free motion of the system near the position of

stable equilibrium has the nature of a harmonic oscillation (natural-
ly, provided that q remains small in the process of motion).
It is known from the general course of physics that a is called the
amplitude, a-the' initial phase of the oscillation, and w0 -the natu-
ral frequency of the system 2 •
Let us transform expression (16.7) according to the formula for
the cosine of a sum:
rj = a (cos a cos w0 t - sin a sin w0 t)
and introduce the notation
c1 = a cos a, c2 = -·a sin a
The solution of Eq. (16.4) can thus be written as
q = c1 cos w0 t + c2 sin w0 t (16.8)
1 See formulas (V.7) and (V.9). .

2 As a rule, we shall not give the information on a question being con~ider~d
that can be found in textbooks of general physics.
l\lECHANICS
where c1 and c2 are real constants whose values are determined from
the initial conditions [from q0 and (q) 0 l.
Finally, we shall give another form of writing a harmonic oscil-
lation1: ·
(16.9)
where
A= aeia (16.10)
is a complex amplitude; its magnitude equals the ordinary amplitude,
and its argument equals the initial phase of the oscillation. Introduc-
ing the value of A from (16.10) into (16.9) and taking the real part
of the expression obtained, we arrive at formula (16. 7).
Consequently, a harmonic oscillation can be represented in the
form of any of the three formulas (16.7), (16.8), or (16.9).
17. Damped Oscillations

- In a real oscillatory system, forc'es act that retard the motion of
the system and lead to a gradual attenuation of the amplitudes
(damping) of the oscillations. The mechanical energy of the system
transforms into the internal energy of the system ap.d the surroundings
(for brevity, the energy is usually said to transform into heat, but
this is not quite strict). Such a process is called the dissipation of
energy .
. We shall limit ourselves to a treatment of cases when the gener-
alized force of friction retarding a system is proportional to the
generalized velocity of the system:
Q* = -rq
.
This is a non-potential force, therefore Lagrange's equation will
have the form of (4.15), and the function (16.3) must be taken as L.
Hence, damped oscillations are described by the equation
1-lq + xq = -rq
.
(the symbols ~l and x have the same meaning as in the preceding
.section). Let us write this equation in the form
(17.1)
where
--~ - - -r
.' ro~ =->
f-1
0 and 2~ =-
fl
The substitution q = e"t leads to the characteristic equation
/.. 2 + 2~/.. + ro8 = 0 (17.2)
1 WP shall use a cap over a symbol to designate complex quantities.
Provided that ~2 < ro3, the roots of the characteristic equation are
complex:
At= -~+ill ro~-~~. /..2= -~-q/ro~-~ 2
The general sol uti on of Eq. (17.1) is
q = Cle'-lt +
Cze).2t = e-f:lt (Clefrot + Cze-frof)
where <I) = V
ro~ - ~ 2 • The solution we have found differs from the
function (16.5) in the factor e-Bt and in the substitution of ro for ro 0 •
Fig. 17.1.
The requirement that q be real leads to the condition C1 = C~;

Introducing the notation (16.6) and performing elementary transfor-
mations, we arrive at an expression for damped oscillations:
q = ae-Bt cos (rot a) + (17.3)
Wlien ~ 2 > ro~, the roots of characteristic equation (17.2) are real:
'l..j = ~~+ V ~ 2 -ro;_=·~a~, 1..2 = -{3- V~~-ro~ = ,-a2
(since V ~ 2 - ro3 < ~. the quantity a 1 is positive; the quantity a 1
is also positive, and a 2 > a 1 ). The solution in this case is
(17.4)
where cl and c2 are real cons'tants. ; .
, Hence,· with strong friction (when ~~ > w~}, no oscillations
.appear-the system . brought out. from its equilibrium position
returns to it asymptotically. The motion of the system may have
the nature described either by curve 1 or c.~·rv~ 2 ·(Fig. 17.1). In :t~e
6~-- MECHANICS
latter case, the system first passes through the equilibrium position,
deviates to the other side of it, and only then approaches the equilib-
rium position asymptotically. Such motion of a system is called
aperiodic damping (or an aperiodic process).
How a system will return to its equilibrium position (following
CUrve 1 Or CQrVe 2) depends on the ratio of the coefficients C1 and C2
which, in turn, is determined by the initial conditions [i.e. by the
values of the generalized coordinate q 0 and the generalized velocity
v0 = (q) 0 at the initial instant).
Let us establish the conditions in which aperiodic motion has a
specific nature. We express the coefficients C1 and C 2 in terms of
q0 and v0 • Assuming in (17.4) that t = 0, we get
(17.5)
Differentiating (17 .4) with respect to time and assuming that t = 0
in the expression obtained, we find that
• (17.6)
Vo = (q)o = -a1C1- a2C2
It follows from Eqs. (17.5) and (17.6) that
(17. 7)
We equate expression (17.4) to zero:

C1e-c:ttt + C 2e-cx2t = 0 (17.8)
When aperiodic damping occurs according to curve 2 (see Fig. 17.1),
Eq. (17.8) must have a finite positive solution. Solving this equation
for. t, we obtain
t = 1 In ( _.£..!_) = 1 In atqo+vo
~-~. C1 ~-~ a~+~
•[we have introduced the values of C 1 and C 2 from (17.7)1. The differ-
ence a~- a 1 is greater than zero (see above). Therefore, twill be
positive when the expression inside the logarithm symbol is greater
than +L The latt.er condition is observed if the expressions
(a~q 0 + v 0 ) and (a 2 q0 + v 0 ) have the same signs and, in addition,
the magnitude of the first expression is greater than that o£ the
second one:
sgn (a 1q0 + v0) = sgn (a2q0 v0) } + (17.9)
!atqo+ Vol>la2qo+vol
.,.. The coefficients a 1 and a 2 are positive, and a 2 > a 1 • Hence, for
l3atisfying the second of conditions (17.9), q0 and v 0 inust have differ·
'ent signs. This occurs if the initial velocity is directed towards the
equilibrium position [when the system is deflected to the right
(q 0 > 0), the velocity is directed to the left (v 0 < 0), and vice versa].
Figure 17.2 shows graphs of the functions y = a 1q0 + v 0 and y =
= a 2 q0 + v 0 • The graphs have been plotted for q0 > 0, therefore
v0 < 0. The values taken on by q 0 are divided into three regions.
It is easy to see that both condi-
tions (17.9) are satisfied only in y
region I, i.e. at q0 's not exceeding
-v 01a 2 • In region I I, the first of
the conditions is not observed, and
in region III, the second one.
Hence, aperiodic damping occurs
in accordance with curve 2 (see
Fig. 17.1) when v0 and q0 have I
different signs and, in addition, I
I
Jq0 J< ~
I
IX2
or Jv 0 J> a 2 Jq 0 J (17 .10) I
I I
(we remind our reader that a 2 = Vo Region I I Region /II Region l!J
=~ + V~ 2 - w8).
Special attention must be given Fig. 17.2.
to the case when characteristic equa-
tion (17.2) has multiple roots. This occurs provided that ~ 2 .
w6. Consequently, /.1 = 1.. 2 = -~. According to formula (V.11),,
in this case the general solution of Eq. (17.1) is
q = C1e-llt + C te-llt
2 = (C 1 + C t) e-llt
2
After the relevant calculations, we fmd that

C1 = qo and C2 = ~qo + Vo
From the condition q = 0, we get (except for t = oo) the value
t=-.fl_=- qo
c2 ~qo+vo
It will be positive if
qo < 0 or ~qo <0
~qo +L'o ~qo + Vo
(multiplication by ~ does not change the sign of a quantity because
~ > 0). The last condition is observed when the sign of v0 is opposite
to that of q0 and, in addition,
I Vo I>~ I qo I (17.11)
Hence, with multiple roots, aperiodic damping can also occur
either monotonously (see the curve 1 in Fig. 17.1) or with passing
through the equilibrium position (see the curve 2 in Fig. 17.1). The
latter case occurs if a system brought out of its equilibrium position
70 MECHANICS
by q0 receives an impetus towards the equilibrium position, impart-

ing a sufficiently high initial velocity to it [a velocity satisfying
condition (17 .11); when the roots are different, the velocity must
comply with condition (17.10)1.
18. Forced Oscillations
Assume that the system tr.eated in the preceding section experiences
the periodically changing external generalized force
Q* = Q 0 cos (rot+ a.) (18.1)
which we shall call the driving force for brevity's sake. Lagrange's
..
equation (4.15) therefore becomes
fA.q + xq =
.
-rq + Q0 cos (rot +a.)
Let us transform it to the form
.q.+ . +
2~q ro~q = /cos (rot+ a.)
0 (18.2)
where / 0 · Q 0 1~-t; the other quantities are explained in the preceding
sections.
We have arrived at a linear non-homogeneous differential equation
with constant coefficients. According to theorem (V.6), we can obtain
its general solution by adding a particular solution of Eq. (18 ..2)
to the general solution of the corresponding homogeneous equation,
i.e. to the function (17.3). To find the particular solution, let us
proceed in accordance with what is said at the end of Appendix V,
namely, let us add the imaginary function if 0 sin (rot + a.) to the
right-hand side of (18.2) and seek the complex solution of the q
q,
equation obtained; after finding we shall take its real part, and
the latter will be the solution of Eq. (18.2). Hence, we shall' solve
the equation
'q + 2~q + ro:q = fo [cos (rot+ a.) + i sin (rot + a.)]
Its right-hand side can be written as
f oe'~<oot+r7.) = f oeioot
where
(18.3)
is the complex amplitude 1 of the driving force (more exactly, the
force divided by fl, but for brevity's sake we shall call Q*l~-t simply
a force). The differential equation written in the new notation will be
'q' + 2~q + ro~q
= J0 eioot (18.4)
(we have omitted the cap over q to avoid complicated symbols).
1 .1 Compare with {16.10).
We shall seek the solution of Eq. (18.4) in the form

q = aeiCilt (18.5)
where a is the complex amplitude of the oscillation. Differentiation
with respect to t yields
.
q = iwaeiCilt,
.q.= (iw)~ aeiCilt = -w 2 aeiCilt (18.6)
We see that in the complex representation of harmonically varying
quantities time differentiation consists in multiplying the quantities
by iw (in integration-in dividing them by iw).
Introducing expressions (18.5) and (18.6) into Eq. (18.4) and
cancelling the common factor eiCilt, we obtain the equation
-w 2 a + 2i~w~ + w~a = fn
from which we find
a= fo
(wG- w2) + 2i~w
We represent the complex number in the denominator as
(w~- w2 ) + 2i~w = peicp (18.7)
where p is the modulus and cp is the argument of this number. There-
fore,
- fo 1 · ...
a=-. = -e-t'J!fo (18.8)
pe''fi p
It follows from (18. 7) 1 that
p= V (wo2 - w2 ) 2 + 4B 2 w2 ' 2 ~w
tan cp = w5-w2 (18.9)
Using the values of p and ] 0 [see (18.3)1 in (18.8), we get the fol-
lowing expression for the complex amplitude:
a= I ro: et<a:-cp> = aet<a-cp>

l (wfi-w2)2+4~2w2
Finally, introducing the value of ainto formula (18.5), we find the

complex expression for q:
q= aei(Cilt+a-cp)
1 Recall that a complex number can be depicted by a point P on a plane.

The abscissa x of this point equals the real part of the number, and the ordinate
y equals its imaginary part. The modulus of a complex number equals the modu-
lus p of the position vector of the point P, while the argument <p is the angle.
made by the position vector and the axis of abscissas. It thus follows that p =
y +
= x~ y~ and tan <p = y/x.
72 MECHANICS
Its real part coincides with the expression for steady-state forced
oscillations known from the general course of physics:
q= fo cos ((i)t +a-:- m) (18.10)
V(w~-w2)2+4~2w2 . '~'
(in textbooks of general physics, it is usually assumed that a = 0).
We obtain tho general solution of Eq. (18.2) by summation of tho
functions (17.3) and (18.10). We shall not stop off to analyse this
solution and consider the phenomenon of resonance because this is
done in sufficient detail in general courses of physics.
19. Oscillations of a System with Many Degrees

of Freedom I
Consider a conservative system with sdegrees offreedom and having
a position of stable equilibrium. In this position, the potential
energy of the system U = U (q1 , q 2 , • • • , q.) has a minimum. We
shall measure the generalized coordinates q1 from the equilibrium
position. Bearing in mind that we shall limit ourselves to small-
amplitude oscillations, let us expand the potential energy in pow-
ers of q1, disregarding the terms of the higher orders of smallness:
U_
-
U0 + "L.J (3!!._)
aq; o q; + .!._ " ( aq;a2U
2 L.J aqk ) o q;qk
i i. k
In the equilibrium position, all the generalized forces Q1 =

= -- (oU/oq 1) 0 vanish. We also assume the energy U 0 to vanish.
The expression for the potential energy can thus be written as
U=; ~xo.q 1 qk
i. k
,(19.1)
',Vhere
X -X - ( 82U )
ik- ki- 8q; 8qk 0
are positive quantities (at a minimum, the second derivatives are

positive). Since U is measured from its minimum value taken as
zero, quadratic form (19.1) is positive defmite.
With stationary constraints, the kinetic energy is determined by
a positive definite quadratic form of the variables 1 [see (5.10)1: q
1 • •
T = 2 2.;!-ltkqtqk (19.2)
i. k
whtlre
!lik = 'Yik (0)
Before beginning to read this section, acquaint yourself with Appendices
1
VII, VIII, and IX.
SMALL-AMPLITUDE OSCILLATIONS
are the zero terms of the expansion of the coefficients 'Vik (q). By
formula (5.8), 'Vik = '\'ki• therefore f.tik = f.tki·
Subtraction of expression (19.1) from (19.2) yields the Lagrangian:
1 . . 1
L=2 ~f.ttkqtq"-2 ~"'tllqtqll (19.3)
i. h i. h
To lind the derivatives of L with respect to q; and we write the q;,
expression of the total differential of the function (19.3):
1, ••
dL =2 2::flt~r.qt aq~~. + 21.2:flt~r.q~r.
, .• 1, 1
dqt- 2 1xtkqt aq~~.- 2 .2:x;~~.q~~. dq,
i,ll i.k i.k i.h
The subscripts i and k are dummy ones, therefore any letter may be·
used for either of them. Taking advantage of this, let us exchange the
places of the subscripts i and k in the first and third sums:
1 • • 1 • •
aL = 2 .2:fl"'q"dqt+ 2 .2:fltkqk dq1
i. k i. k
1 1 . •
·-2 .2: X~r,;qk dqi -2 ~ X;kqk dql = .2: f.ttkqk dqi- ~ Xtkqk dql
i.ll i,k i,k i,k
(recall that f.tlli = f.ttk• and "'"; = x;k)· The expression we have
obtained can be written as
(19.4)
In an expression for the total differential of a function of several

variables, the factor of a differential of a variable equals the partial
derivative of the function with respect to this variable. It thus
follows from (19.4) that
d aL
Since the quantities f.ttk are constants, the derivative - -
dt ()ql
= "51, ~t 1 kq~. Hence, Lagrange's equations have the form

k
(19.5)
[compare with Eq. (16.4) for one variable].

We have arrived at a system of linear homogeneous differential
equations with constant coefficients. Let us attempt to seek the
unknown functions qk (t) in the form [compare with (16.5)1
(19.6)
14 MECHANICS
where c~~. are complex constants that have to be determined. The

:functions (19.6) are complex, whereas the generalized coordinates
are real. Consequently, upon completing our calculations, we shall
have to take the real parts .of functions (19.6) (see Appendix V).
The introduction of expressions (19.6) into Eq. (19.5) yields
LJk f.lt'll. ( - ro2) C~~.eiwt + ~ x 1hC11.eiwt = 0
11.
(i = 1, 2, ... , s)
Cancelling eiwt in all the equations, we obtain
LJh. (x 1 ~~.-ro 2 f!t11.) C11.=0 (19. 7)
We have arrived at a system of s linear homogeneous algebraic

-equations with the unknowns C1 , C2 , • • • , C 8 • For this system to have
.a non-zero solution, it is necessary and sufficient that its determinant
be zero:
Xu- ro 2J.Lu Xtz- ro 2fltz • • • Xts- ro 2f.lts
X21- ro 2flzt X22- ro 2f!zz • • • Xzs- ro 2f.l2S
=0 (19.8)
ro 2f.lst Xsz- ro 2fls2 • • • Xss- ro 2J.Lss

Xs1-
,[see the text following formula (VIII.26) in Appendix VIII].
Equation (19.8) is known as a characteristic equation. It is an equa-
tion of degree s relative to ro 2 • In the general case, this equation
has s different 1 real positive roots: roi, ro~, ..•• , ro;. The quantities
roet(a = 1, 2, ... , s) found in this way are called the natural frequen-
-eies of the system.
Let us prove that the roots of Eq. (19.8) are real and positive.
For this purpose, we multiply each of Eqs. (19. 7) by Ct (i.e. by a
quantity that is the complex conjugate of the coefficient C ;) and
then summate all the equations. The result is
LJ
i, k
(xtk- ro 2f!i11.) ere k = 0
-or
·whence
~ Xihcrck
·,·. (1)2 = -=i.:.,:k.::....__ __ (19.9)
~ flikqck
i, k
The numerator and denominator of Eq. (19.9) contain quadratic

iorms like (IX.21). It is shown in Appendix IX that such a form
1 Multiple roots may he obtained in particular cases.
equals the sum of the quadratic forms, 2Ji,k xil,a 1a 11 + 2Ji,k x 1 ~~.b 1 b 11 ,
and correspondingly ~ r.taa 1ak
i,k
+ 2Ji,k
f.ttkbtbk (a 1 is the real part,
and b 1 is the imaginary part of C 1). The
latter forms, in turn, are,
first, evidently real and, second, positive defmite [see (19.1) and
(19.2)]. We have thus proved that the numerator and denominator of
Eq. (19.9) and, consequently, w~ are real and positive.
Thus, having solved the characteristic equation (19.8), we find s
natural frequencies of the system: <u 1 , w 2 , • • • , W 8 • Introducing in
turn the values w& into the system of equations (19.7) and sohing
the system, we find C ~~.'s corresponding to different wa. 's. If the
matrix of the system (19.7) has the rank s - 1 (which is usually the
case), by (VIII.28) the solutions of the system are
C k(~) -- ca A(a)
mk
where ca is an arbitrary complex eonstant, and A~l is the signed

minor of the element Xmk - w&~tm 11 in the determinant of the system
(m is chosen arbitrarily but with at least one A~~ being non-zero).
Since all the elements of this determinant are real, the quantities
A~~ are also real.
Hence, for each generalized coordinate qk, we obtain s different
solutions of the form
qk = c~ A mke
(~) {(l)~tl (ct = 1 . 2 ' . . . ' s) (19 . 10)
where A~k are real constants determined by the values of the coef-
ficients Xtk and !-ttk• and also of the frequencies Wa·
We obtain the general solution by summation of all expressions
(19.10):
Passing over to the real part of this expression, we obtain

qk = Re {lJ CaA~~ei(l)at} = ~ A~~Re {caeiwett}
~ a
Finally, representing ca as aa.e• 0a

(here aCT. is the modulus of ca,
i.e. a real positive quantity), we arrive at the expression
•
qk = 2J A~~aa cos (Wat +6a.) (19.11)
Cl=1
Consequently, the change in each generalized coordinate qk with

time is the superposition of s harmonic oscillations whose frequencies
equal the natural frequencies of the system. The quantities aa and
6rt are determined from the initial conditions.
76 MECHANICS
Expressions (19.11) can be greatly simplifted with a special selec-

tion of the generalized coordinates. It is shown in Appendix IX that
when we have two quadratic forms--one from the variables qk and
.
the other from the variables qh, the first of them being positive
definite, a linear transformation of the variables qh. exists such that
reduces both forms to a diagonal type [see diagram (IX .37)1. Let us
pass over with the aid of such a transformation from the variables qk
to the variables ~It· The quadratic forms (19.1) and (19.2) will now
becom~ diagonal:
1 •
T=z- ~~k, (19.12)
k
The Lagrangian will be as follows in the new variables:
L= ; 2:~it- ; 2:'-~t~k
k k
and Lagrange's equations will be
'ik + '-~t~k = 0 (k = 1, 2, ... , s) (19.13)

The equations of motion in the coordinates ~It thus split up into s
independent equations each of which is identical to Eq. (16.4). We
must note that owing to the positive definiteness of the quadratic
form for the potential energy U, all the coefficients '-~t are positive.
They can therefore be written as
'),k = (J)~
where ffi~t are real quantities.

Let us write the solutions of Eqs. (19.13):
Sk = ak cos (rokt + 611.) (k = 1, 2, ... , s) (19.14)
[see formula (16. 7)].
We have found that the generalized coordinates £h perform a simple
harmollic oscillation independently of one another, and each with
its own frequency ffik· The generalized coordinates determined in
t~is way are called normal (or principal), and the simple harmonic
oscillations they perform-normal oscillations of the system.
We must note that the normal coordinates 6~t are related to the
arbitrary generalized coordinates qk by means of linear transforma-
tions,· i.e. transformations of the form
(19.15)
Hence, ~It can be obtained as a linear combination of the coordin-

ates q;.
20. Coupled Pendulums

Consider the small-amplitude oscillations of a system consisting
of two identical simple pendulums connected by a weightless spring
{Fig. 20.1). Assnmo that the peiHlnlums can oscillnto only in the
plane of the drawing so that the sys-
tem has two degrees of freedom. Wo
choose <p 1 and <p 2 -the angles of deflec-
tion of the pendulums from the
vertical direction-as the generalized
coordinates. The length of each pen-
dulum is l and its mass is m. The
ends of the spring are fastened on the
pendulum rods at a distance b from
the point of suspension. The spring
is chosen so that when <p 1 = <p 2 , its
tension is zero. m m
Let us write an expression for the
potential energy U of the system,
assuming that U is zero in the equi- Fig. 20.1.
librium position:
U = mgl (1- cos <p 1) + mgl (1- cos <p 2) + ~ k (b sin <p 2 - b sin <p 1) 2
For small-amplitude oscillations, we can assume that sin <p = <p
and cos <p = V
1 - sin <p = 2 V1 - <p = 1 -
2 ~ cp The expression
2•
for U thus becomes

U = 21 mgl<p~ + 21 mgl<p; +:r1 kb2 (<p 2 - <p 1)2
1
= 2 [(mgl + kb 2) <p~- kb 2 <p 1<p 2 - kb 2 <p 2 <p 1 + (mgl + kb 2) <p~] (20.1)
The kinetic energy in the same approximation is
1 • •
T = 2 [ml 2 cri + m.ZZ<p~] (20.2)
A comparison of expressions (20.1) and (20.2) with expressions
(19.1) and (19.2) yields the following values for the coefficients
Xtk and !l11t:
X 11 = X 22 = mgl +
kb 2 , x 12 = x 21 = -·kb 2 (20.3)
flu = !l22 = ml2, ~l12 = !lzt = 0
The introduction of these values of the coefficients into Eq. (19.5)
leads to the differential. equations
ml 2~~~ + (mgl + kb2) 1 = C1eiwt and q:> 2 = C2eiwt (20.5}
Let us substitute these expressions into Eq. (20.4). After cancelling
and combining similar terms, we get a system of equations for
etoot
determining the constants C1 and C 2 :
(mgl + kb 2 - ml 2 w2 ) C 1 - kb 2C2 = 0 }
- kb 2C1+ (mgl + kb 2 - ml2ro2) C2 = 0 (20 ·6)
For this system to have a non-zero solution, its determinant must
equal zero:
I
(mgl + kb 2 - mZ 2ro 2)
(- kb2)
(- kb 2)
(mgl + kb 2 - ml2ro2)
0 I=
i.e. the following condition must be satisfied:
(mgl + kb 2 - mZZro 2 ) 2 - ( -kb2 ) 2 = 0.
The latter equation can be written as follows after simple transfor-
mations:
(ro2)2-2( f +! ~: )ro2+[~: +2 ( f ! ~: )]=0
We have arrived at a quadratic equation relative to ro 2 • The roots
of this equation are
2 g . 2 . g . . k b2 •
rot = T and ro2 = T + 2 In 12
Consequently, the natural frequencies of the system will be
, /g / g 2 k b2
ffit = v T and Wz = 1 T + 12 m.
(20. 7)
· Let us introduce the square of the first natural frequency, i.e.
ro~ into Eq. (20.6) instead. of ro 2 • After simplification, system (20.6)
becomes
kb 2 C1 - kb 2 C ~ = 0
-kb2C1 + kb 2C 2 = 0
= ==
'The solutions of this system are obvious:
cl Cz cl alefOt
where c1 is an arbitrary complex constant, a 1 is its modulus, and
.. (20.8)
IS 1 is its argument. ·
The introduction of (20.8) into (20:5) yields complex solutions of
differential equations (20.4) corresponding to the frequency w1 :
q:>jtl · · . c1eiw,t =·. a1ei(w,t+Otl.
\C ,.
<p~~~ = c1eiw 1 t =:' ~ 1-ei(w,t+_o,~·
SMALL-AMPLITUDE OSCILLATIONS 79·
Taking the real part of the functions we have found, we obtain

q>~ 1 ' = a 1 cos (w 1 t + <\), q>~ 0 = a 1 cos (w 1 t + 61 ) (20.9}-
Now let us introduce the square of the second natural frequency ..
i.e. w~, into Eq. (20.u) inslcnd of w2 • Tho result is
-kb 2 C1 - kb 2C 2 = 0
-kb 2C1 - kb 2C 2 = 0
The system is satisfied by the values
cl = -C2 = Cz = a2ei 0•
The functions (20.5) will accordingly be

q>~~, = c2 eilil 2 t = a 2 ei(lil2t+6 2) }
q>~z, = -c 2eilil2t = - a 2ei(Cil2t+llt>
and their real parts will be
<p~ 2 ' = a 2 cos (w 2t + 62) . } ( 20.10)
<p~2 ' = - a2 cos (w2 t + 62 )
We obtain a general solution of system (20.4) by the summation of
solutions (20. 9) and (20.10). Consequently,
Cf>t = <Jl~ 1 ' + <p~ 2 ' = ai cos (wit+ 61) + a 2 cos (w 2 t + 62) l (20 _11}
 2 to th&
new variables ~ 1 and 62 , which· we shall determine as follows:
St _:_ ~ (<Jlt + <Jlz) and Sz = ~ ( <Jl1- <Jl2)

With a view to (20.11), we obtain
Si=alcos(wit+6t)} (20.12)
S2 = a 2 cos (w 2 t + 62)
The variables £1 and £2 are thus normal coordinates of the system of
coupled pendulums being considered. The generalized coordinates
s
cp 1 and q> 2 are expressed in terms of £1 and 2 with the aid of the
linear equations
<Jl1 = 61 + sz and cr2 = £1 - 62 • (20.131
Assume that only the first normal oscillation is performed in th&
system. This signifies that 62 ==
0. Inspection of (20.13) shows that
in this case
80 MECHANICS
i.e. both pendulums oscillate like a single whole with the frequency
ro 1 , being at each instant deflected to the same side through the
same angle (Fig. 20.2a). The spring is not deformed so that each
pendulum oscillates as if the other one were absent (w 1 = V gil).
Now assume that only the second
normal oscillation is being per-
. formed in the system. A glance at
(20.13) shows that in this case
(jl1 = -ep2 = £2 = a 2 cos(w 2t+6 2)
At each instant, the pendulums are
deflected through an identical angle,
(a) (b) but in opposite directions (Fig.
20.2b).
Fig. 20.2. The connection between the pen-
dulums can be characterized with
the aid of the spring constant k. Let us call the latter the coupling
coefficient. Consider the case of a weak connection, i.e. a small k.
If (kim) ~ (gil), the difference between the natural frequencies will
be much smaller than the frequencies themselves:
(20.14)
Let us deflect the first pendulum through the angle cp 10 = a, keeping

the second one at its zero position. Now let the system oscillate.
The initial conditions in this case will obviously be
(jl1o = a, <Jl2o = 0, (~l)o = 0, (~2)o = 0
\Ye ::.hall fmd the values of the constants a 1 , a 2 , 61 , and 6 2 • For
this purpose, we assume that t = 0 in (20.11). The result is
ep 10 =a= +
a 1 cos 61 a 2 cos 62 }
(20.15)
ep 20 = 0 = a 1 cos 61 - a 2 cos 52
Now let us differentiate expressions (20.11) with respect to time

and assume that t = 0 in the formulas obtained. This leads to the
expressions
(~ 1 ) 0 = 0= -a 1w 1 sin B1-a 2wlsin 52}. (20.16)
( ~ 2 ) 0 ·= 0 = - a 1w1 sin B1 + a 2 w2 sin 52
Solving Eqs .. (20.15) and (20.16) simultaneously, we find that

Hence, in the case being considered, the oscillations have the

form
a. (
qJ 1 = 2 cos co 1t +cos w~t ) =a: cos--~-
w2 -w 1 w.+w 1
t. cos~ t
qJ 2 = za. (COS w t - COS w t ) =

1 2
· W2- WI
2-
0: S l l l - -
•
t · Slil W2
--
+2-
Wl
t
With a weak coupling, we have (w 2 - w1 ) « (w 2 + w1 ) [see

(20.14)1. We may thus consider that each of the pendulums performs
harmonic oscillation at the frequency (w 2 + w1 )/2 ;::::::; C•l 1 with a
slowly varying amplitude. Hence, the motion of each of the pendulums
has the nature of beats. The amplitudes change with a phase shift of
n/2. When the amplitude of one of the pendulums reaches its maxi·
mum value, the amplitude of tho second one vanishes, and vice versa.
In the process of oscillations, energy is pumped, as it were, from one
pendulum to the other.
When only one normal oscillation ~ 1 or ; 2 is produced, no transition
of energy from one pendulum to the other occurs.
Chapter V
MECHANICS OF A RIGID
BODY
21. Kinematics of a Rigid Body
By dividing a continuous rigid body into elementary volumes of

mass p dV (here tJ is the density of the body), it can be represented
as a system of particles with rigid constraints.
A rigid body is known to have six degrees of freedom-three trans-
lational and three rotational ones. To describe the motion of a rigid
body, let us choose the inertial reference frame K (with the axes
X 1 , X2, X a) which we shall consider to be stationary. We shall
rigidly connect to the body another frame K' (with the axes x 1 , x 2 , x 3 )
and place its origin at the point A of the body. It is convenient to
take the three coordinates of the origin of the frame K' (the position
vector RA corresponds to them) and the three angles characterizing
the orientation of the axes x 1 , x 2 , x 3 relative to the axes X 1 , X 2 , X 3
as the generalized coordinates determining the position of the body.
These axes make nine angles with one another, but only three of
them are independent; the other six can be expressed through the
values of the first three 1 • It is customary practice to use the Euler
angles cp, 'it, '\jJ (see Sec. 22) as the three angles determining the
mutual orientation of the axes of the frames K and K'.
Any elementary displacement of a rigid body can be represented
as the sum of a translational displacement when all its points are
displaced over the same distance dRA and rotation through the angle
d<D about nn nxis passing through the point A.
Since the velocities v of the points of the body in the frame K'
are zero, formula (15.3) for the velocity of a point whose position
in the frame K' is determined by the position vector r<Al 2 acquires
, 1 There are six relations between the cosines of these angles [see formu-
la (VI.39)):
2J<Xfm<X!tm = 6tk (i, k = 1, 2, 3; i ~ k)
m
[<Xt~t =
cos (xt, Xk)]•
Let us agree on notation. In this chapter, we shall use indices of two kinds:
2
(1) without parentheses, and (2) in parentheses. Indices without parentheses
will indicate a particle or point which the given quantity relates to. For in-
stance, mu. is the mass of a particle whose number is a, r91 is the position vector
of the same particle, and RA is the position vector of the point A.
Indices in parentheses will indicate the point from which a position vector
emerges, or the point relative to which a moment or angular momentum is cal-
MECHANICS OF A RIGID BODY 83
the form
(21.1)
where VA is the translational velocity of the body (the veloeity of
the point A observed in the frame K), and W(A) = d<l>ldt is the
angular velocity of rotation of the body about an axis passing through
the point A. The first term in this formula is the same for all the
points of a body, the second is a position function.
If we had placed the origin of the frame K' at another point of
the body, say at a point B, formula (21.1) would be as follows:
V = VB + [w, rl (21.2)
where VB is the velocity of the point B observed in the frame K,
and WeB> is the angular velocity of rotation of the body about an
axis passing through the point B.
The position of an arbitrary point of the body in the frame K
is determined by the same position vector in both cases:
R = RA + f(A) = RB + f(B)
It thus follows that the position vector reB) can be represented as
reB> = a + rcA> (21.3)
where a = RA - RB is a position vector from the point B to the
point A, i.e. a quantity not depending on which point of the body
we write formula (21.3) for.
Using the value given by (21.3) in formula (21.2), we obtain
V = VB +
[w<Bl• a] +
[ro<Bl• r<All (21.4)
The first two terms on the right-hand side of (21.4) are identical
for all the points of the body, while the third term is a position
function.
Formulas (21.1) and (21.4) determine the same quantity-the
velocity of the point of the body being considered in the frame K.
culated, etc. Depending on the circumstances, we shall use these indices either
as subscripts or superscripts on the relevant symbol. For example, rcA) or r<A)
will stand for a position vector emerging from the point A; M(A) or M(A), for
the angular momentum relative to the foint A. The absence of an index in
parentheses at the symbol r or M wil signify that the relevant quantity is
taken relative to the centre of mass C of the body. Hence, we shall designate a
position vector emerging from the point C either by the symbol r<c> or simply
by r, the angular momentum relative to the point C, by the symbols M<c> and
M, and so on.
We shall use the symbol R only in one quite definite case-to denote a posi-
tion vector emerging from the origin of the stationary reference frame K (with
the axe! X, Y, Z). Therefore, there is no need to use an index in parentheses in
the given case, and we shall not write it.
We shall use lower case letters (r, x, y, z, etc.) to denote position vectors
drawn from the origin of the reference frame K' rigidly connected to a body,
the coordinates in the frame K', etc.
6*
84 MECHANICS
Consequently, at any r(Al• the right-hand sides of these formulas

must coincide. This is possible provided that
VA = VB + [ro(BJ• a) (21.5)
[ro(Al• r(A)] = [ro• r<Al) (21.6)
[the identity sign stresses that equality must hold at any values of
r(A)].
A glance at identity (21.6) shows that
(J)(A) = (J)(B)
i.e. that the angular velocity of rotation about any axis is the same,
and we can speak simply of the angular velocity ro of the body regard-
less of our choice of the reference frame K'. The translational veloc-
ity, as can be seen from relation (21.5), does not have an absolute
nature, however; it depends on the position of the origin of the
frame K' (that is, VA =I= VB)·
Suppressing the superfluous subscript on ro, let us write relation
(21.5) as follows:
(21.7)
Two cases are possible: (1) the vectors VA and w are mutually per-
pendicular, and (2) the vectors VA and w make an angle differing
from n/2. It is easy to see that in the first case the vectors VA and
[roa] are coplanar. Consequently, the vectors VB and VA are also
coplanar. Hence, the vector VB, like the vector VA• will be per-
pendicular to the vector w. This allows us to make the following
conclusion: if the vectors VA and w are mutually perpendicular with
our choice of the origin of the frame K', these vectors will also be
mutually perpendicular at any other choice of the origin of the frame
K' (with any other choice of the point A).
Let us now turn to formula (21.1) and write it as
V = VA -t- [ro, r(A)] (21.8)
This formula shows that when the vectors VA and ro are mutually
perpendicular (which, if this occurs, is observed with any choice of
the point A), the vectors V and VA will be coplanar, and the veloci-
t~es V of all the points of a body are in planes perpendicular to the
vector ro. By varying our choice of the point A, we can fmd a position
of it for which
VA = V - [ro, r(A)) (21.9)
vanishes 1 (here the point A may be outside the body). As a result,
1 Both terms on the right-hand side of (21.9) are position functions of points
of the body (V is the velocity of a point of the body in the frame K, and
f(A) is the position vector of this point in the frame K'). The difference of these
terms for all the points of the body is the same and equals VA.
the motion of a solid will be represented as only rotation about an

axis called the instantaneous axis of rotation of the body [see (21.8)].
When the vectors VA and ro are not perpendicular to each other,
we can choose the point A so that these vectors will be collinear.
Consequently, the motion of the body at each instant is the super-
position of two motions: rotation about an axis at the angular veloc-
ity ro and translational motion at the velocity VA along thP same
axis. We shall not stop to prove this statement.
We must note that the formulas of the dynamics of a rigid body
become especially simple if we make the origin of the frame K'
coincide with the centre of mass C of a body. In the following, we
shall usually proceed in exactly this way. Formula (21.8) will
therefore acquire the form
V = Vc +
[ror] (21.10)
22. The Euler Angles

The Euler angles are determined as follows. Assume that the axes
of the frame K' associated with a bodv ftrst coincided with the axes
of tho frame K. Next the body turn~d, as a result of which the
z z z
(a) (b) (c)
Fig. 22.1.
orientation of the axes of K' in space changed. Any suelt rotation

can be performed with the aid of the three rotations :,;lwwn in
Fig. 22.1.
1. Rotation about the Z-axis through the angle cp (Fig. 22. Ja).
The direction n followed by the x-axis is called the nodal line.
2. Rotation about the nodal line through the angle {} (Fig. 22.1b).
3. Rotation about the z-axis through the angle 1p (Fig. 22.1c).
The direction of each of these rotations is related to the direction
of the axis about which it occurs by the right-hand screw rule.
Examination of Fig. 22.2 shows that the nodal line is the line of
intersection of the coordinate planes XY and xy. The angle cp is formed
86 MECHANICS
by the X-axis and the nodal line, the angle 1.p by the nodal line and
the x-axis,· and, fmally, the angle {} is the angle between the axes Z
and z. The angles {} and cp are
z· the polar coordinates of the point
of intersection of the z-axis and
a sphere of unit radius. This
point is known as the apex.
For the set of angles cp, {}, and
1.p determining each real rotation
Y to be unique, it is assumed that
-+-+----J~:&~77:'>1--+-- the angles <p and 1.p can have
values from zero to 2:rt, while the
values of the angle {t are limited
to the interval from zero to :rt.
If the angle {t were also allowed
to have values from 0 to 2:rt, the
rotation depicted in Fig. 22.3,
for instance, could be charac-
terized either by the set of an-
Fig. 22.2. gles cp = n/2, 'l't = n/2, 1.p = 0 (the
upper sequence of rotations; the
axes X, Y, Z are not shown in the figure, their orientation
coincides with the initial orientation of the axes x, y, z) or by the
set cp=3n/2, {} = 3n/2, 1.p = :rt (the lower sequence of rotations).
Fig. 22.3.
Assume that the Z-axis is directed vertically and the frame K'
is rigidly associated with a top (gyroscope), the z-axis coinciding
with the top's axis of proper rotation. It is now a simple matter to
see that a change in the angle 1.p corresponds to rotation of the top
itself, a change in the angle <p corresponds to rotation of the vertical

plane containing the z-axis, i.e. to precession of the top, and, ftnally,
a change in the angle{} to motion of the top's axis called nutation 1 •
Accordingly, the angle <p is called the precession angle, the angle
1J-the nutation angle, and the angle '¢-the angle of proper rotation
(or the angle of pure rotation) 2 •
The rato of the change in the angle <p can be characterized by
the angular velocity vector ro'P directed along the Z-axis (see
Fig. 22.2); the magnitude of this vector is cp. Let us resolve the vector
roll' into two components, one of which is directed along the z-axis
(its magnitude is <p cos-&), and the second is perpendicular to the
z-axis, i.e. is in the plane xy (its magnitude is ~sin{}). Tho second
component is obviously perpendicular to the nodal line n and,
consequently, makes the angles n/2 - tjJ and tjJ with the axes x and
y, respectively. We can conclude from the above that the projections
of the vector ro~ onto the axes of the frame K' are
(w..,) 1 =~sin{} cos (n/2- "') =~sin{} sin "' )

(w,.):: = <p sin {} cos tjJ (22.1)
(ro'P) 3 = ~ cos {}
The rate of the change in·· the angle {} is characterized by the
vector ro~ directed along the nodal line: its magnitude is .fr. The
nodal line is perpendicular to the z-axis, and makes the angles '~
and tjJ + n/2 with the axes x and y, respectively. Consequently, the
projections of the vector ro" onto the axes of the frame K' are
. . cos"'
(w~) 1 = 1J
1 (22.2)
(w"h = .fr. cos("'+ n/2) = - .fr sin tjJ
(w,)a = 0
Finally, the rate of the change in the angle 'I' is characterized
by tho vector ro"' directed along tho z-axis (its magnitude is~). The
projections of this vector onto the axes of the frame K' are
(w'l>h = 0, (w¢) 2 = 0, (w¢) 3 = tjJ (22.3)
1 General courses of physics usually deal only with regular precession char-
acterized by the angle between the top's axis and a vertical line remaining un-
changed. Actually, as a rule, the top's axis oscillates in the plane Zz about a
certain middle position. This oscillation is what we call nutation.
2 The letter '1\J is sometimes used to denote the angle of precession, and the
letter <p-the angle of proper rotation.
88 MECHANICS
The vector of the angular velocity (J) at which the body rotates
relative to the frame K (the body is always stationary relative to the
frame K') can be represented as the sum of the angular velocities of
each of the three rotations corresponding to the changes in the Euler
angles:
(J) = (J) ;p + (J)~ + (J)Ij)
Therefore, the following values are obtained for the projections of
I
the angular velocity (J) onto the axes of the frame K' with account
taken of formulas (22.1), (22.2), and (22.3):
w1 =~sin it sin ¢+~cos cp

w2 = <p sin tt cos¢- it sin '\jl (22.4)
(1)3 = ~cos tt + ~
We shall need these formulas in the following.
23. The Inertia Tensor 1
Assume that we are observing the motion of a solid in the stationary

reference frame K (whose axes will be designated XI. x2. Xa, or
X, Y. Z). In accordance with what was stated in Sec. 21, let us
connect the reference frame K' having the axes xi, x 2 , x 3 (or x, y, z)
to the body rigidly. We divide the body mentally into particles of
mass ma 2 • According to formula (21.10), the velocity of the a-th
particle will be written as follows:
Va = Vc + [ro, ral (23.1)
Let us calculate the kinetic energy of the body. It is
T= ~ i.:maV&={ .l:ma{Vc+f(J), ra1} 2
Ct
= ~ .l:ma V~+ 2:>~a Vc (ro, ra] +; 2:;ma (ro, raJ2

In the first term on the right-hand side, we can put the factor V~
ou,tside the sum sign. This term therefore becomes ; m V~, where
m = ~ ma. is the mass of the body. In the second term, we shall
perform a cyclic transposition of the factors [see (VI.3)], after which
we shall put the constant factor outside the sum sign. The result
is the expression [Vc, (J)] ~ ma.r a. = l VC• ro] mrc, where rc is the
1 Before beginning to study this section, acquaint yourself with Ap-
pendix X.
2 The subscript a indicates a particle's number. We use the Latin subscripts
;,, k, 1, ... to number coordinate axes, vector components, etc.
MECHANICS OF A RIGID BODY 89'
position vector of the centre of mass C. If, as we have agreed upon,

we place the origin of the frame K' at the point C, the second term
vanishes.
Hence, the kinetic energy of a rigid body breaks llp into two terms ..
The first of them,
(23.2).
is the kinetic energy of translational motion. The second term
Trot=~ ~ma [ro, raJ 2 (23.3}

Ct.
is the kinetic energy of rotation.

We must stress the fact that bot.h these energies are absolutely
independent-one depends only on Vc, and the other only on ro.
Since the origin of the frame K' coincides with the point C, the term
containing both V c and ro vanishes.
Let us transform expression (23.3). First, we shall replace the
square of the vector product in accordance with (VI.6):
Trot= ; ~met. {ro 2 r~- (ror a) 2} (23.4}

Ct.
We shall now write this expression using the projections of the-

vectors ro and ret. onto the axes of the frame K'. The projections of
ret. onto the axes of the frame K' equal simply the coordinates of the
x x x
particle 1 a, 2 a, 3 a. Let co 1 , co 2 , co~ stand for the projections of the
vector ro onto the axes of the frame K'. Expression (23.4) therefore
acquires the following form in the components of the wctors:
Trot=~ ~ma [(~coi) ( ~xra)-( ~COtXta) ( ~cohxha)]

Ct. i l i h
= ~ 2>~a [ (~co~ )( 2:xfa)- 2:wtCOhXtaXha]

et. i I l, h
(we remind our render that a dummy index can be designated by any
letter).
In the expression we have obtained, the quantities CO; and coh
do not depend on the subscript a. and they could be put outside the-
sign of the sum over a.. This is prevented by the circumstance. how-
ever, that the first term of the expression includes the sum of the-
squares coi and the second term, the sum of the products co;coh. This
obstacle can be eliminated by replacing the sum of the quantities
coi with the expression ? 1, co 1coh6ih which is obviously equivalent
"i:h
90 MECHANI<!S
to 2j cui- The formula for the rotational energy therefore becomes

i
Trot={ ~ma [( ~ffitffi~t6t~t) ( ~x1a)- ~ffitffiJtXtaX~ta] (23.5)

a: i, k l i. k
(we must note that ~ Xfa is simply a scalar 1 depending on the sub-
t
script a; each of the addends ro;roh6 11, is multiplied by this scalar).
In expression (23.5), summation is first performed over the sub-
scripts i and k, and then over the subscript ct. Let us change the
sequence of summation so that summation over the subscripts i
and k will be performed the last, i.e. rewrite (23.5) as follows:
Trot= ~ ~(J)i(J)lt2:ma:
i, k a
r
61/t ( ~xra)- Xta.X~ta J
I
If we introduce the symbol2
lilt= ~ma:
a.
[ 611t ( ~x1a) -x 1 a.X~ta.
I
J (23.6)
the expression for the kinetic energy of rotation can be written as
Trot=~ ~ft~tffit(l)k (23.8)

i, k
•· The quantity determined by formula (23.6) is a number (but not

an invariant!) depending on the subscripts i and k. There are al-
together nine such numbers. It can be seen that the set of quantities
1 1" forms a second-rank tensor 3 • Indeed, the product of the scalar
2j .x~a=r~=inv.
l
When calculating the quantities lth for a continuous body, p dV must be
2
taken instead of ma and summation replaced by integration. Hence
ltl! = J[llt~t ( ~.xf ) -.xt.xl!] P dV

l
(23.7)
' 8 This conclusion can also be arrived at by the following reasoning. Let us
write expression (23.8) in the form
Trot=~ ~Wt ~I,,.w,.,

i k
This expression can be invariant only if the quantities lJitltwh are the i-th com-
k
ponents of a vector [see Appendix VI, the text following formula (VI.28)]. The
latter, in turn, is possible only when the quantities Jilt are the components of a
tensor [see Appendix X, the text between formulas (X.22) and (X.23)].
hl xfa. and the unit tensor 6ik is a tensor [see formulas (X.17) and
(X.19)l, and the products x 1a.xka are products of the components
of the vector ra., i.e. also a tensor [see (X.16)l. Finally, the difference
of the relevant components of two tensors also gives the components
of a tensor [see (X.18)l.
l-Ienee, quantities (23.G) are the components of a tensor. Tho latter
is called an inertia tensor. Quantities (23.6) do not change when
the subscripts i and k are transposed. Consequently, the inertia
tensor is symmetrical (1 111 = I 11 i)·
Let us write the components of the inertia tensor using the con-
ventional notation of Cartesian coordinates:
-~ m (y2+ z2) -~mxy - ')1, mxz
(Itl!) = - Lj myx ~ m (x 2+z2) - h myz (23.9)
-,Lmzx -~. mzy 2J m (x2 + yz)_
(to avoid making the formulas more c.omplicated, we have suppressed
the subscript a on m and on the coordinates x, y, z; all the sums are
taken over this subscript).
The diagonal components of the tensor are known as axial mo-
ments of inertia. They coincide with the moments of inertia of
a body relative to the corresponding coordinate axes known from
the general course of physics. The non-diagonal components are
called centrifugal moments of inertia.
The geometric shape of a symmetric tensor is an ellipsoid. In the
case being considered, it is an ellipsoid of inertia. The directions
in a body coinciding with the semiaxes of an ellipsoid of inertia
are called its principal axes of inertia. They intersect at the centre
of mass of the body. If we direct the axes of the frame K' (i.e. the
axes x, y, z; we remind our reader that these axes are rigidly con-
nected to the body) along the principal axes of inertia of a body, the
inertia tensor will be reduced to a diagonal form
/1 0 0)
(I,k) = ( 0 /2 0 (23.10)
0 0 13
The values of / 1 , / 2 , ! 3 of tho diagonal components of a tensor
(in the case when itj has been reduced to a diagonal form) are
called the1 principal moments of inertia of a body (they could be
designated by the symbols I:x, IY, I,).
If the principal axes of inertia have been chosen as the axes x, y, z
associated with a body, expression (23.8) for the kinetic energy of
rotation is simplified as follows:
Trot=; (! 1 w~+l 2 w:+I 3 wD (23.11)
92 MECHANICS
or
(23.12)
(do not forget that (J) 1 , (J) 2 , (J) 3 are the projections onto the axes
x. y, z of the vector (J)-the angular velocity of rotation of the body
observed in the frame with the axes X, Y, Z).
When the vector (J) coincides with one of the principal axes of
inertia along which we direct, say, the z-axis, the expression for the
energy becomes even more simple:
(23.13}
An expression similar to (23.13) is also obtained when a body
rotates about an axis fixed relative to the body and passing through
its centre of mass 1 • By directing, say, the z-axis along this axis, we
find that rox = roy = 0, and roz = ro. Consequently, of the nine
addends of formula (23.8), only the one in which i = k = z will be
non-zero, so that
T rot= 21Izzffiz2 = 21Izzffi 2 (23.14)
where I zz in the general case is not one of the principal moments of

inertia.
We must note that, for example, for a body such as a homogeneous
sphere, the ellipsoid of inertia degenerates into a sphere. Therefore,
the principal axes of inertia are not f1Xed relative to the body. This
signifies that any three mutually perpendicular axes passing through
the centre of the sphere can be taken as the principal axes. In this
case, all the principal moments of inertia are the same: I 1 = I 2 =
= I 3 = I, and the tensor of inertia can be written as '
(23.15)
where (cSuJ is a unit tensor [see (X.17)l, and I is a scalar.
Everxthing that we have said about a sphere also holds for a
homogeneous cube. Indeed, relation (23.15) evidently holds for it.
Consequently, the ellipsoid of inertia for a cube degenerates into
a s'phere. For this reason, any axis (and not only an axis of symmetry)
passing through the centre of a cube may be considered as a principal
axis of inertia. This is why a cube. in addition to a sphere and other
bodies for which I 1 = I 2 = I 3 , is called a spherical top 2 •
1 If this axis does not coincide with any of the principal axes of inertia, it
must be retained in place with the aid o[ bearings.
2 For a spherical top, the energy is always expressed by formula (23.13),
where by I, we must understand the scalar factor I in expression (23.15) for
the inertia tensor.
A body for which two principal moments of inertia are equal (for
instance, I 1 = I 2 =I= I 3 ) is called a symmetrical top. Finally, a body
for which all three principal moments of inertia are different is called
an asymmetrical top.
Up to now, in considering the inertia tensor, we a~sumed that the
·origin of the frame K' associated \vith a body is at tho centre of
mass C of the bod v. Tho iHPrtia
tensor can also lJe ~letermined by
formula (23.6), however, with re-
spect to the frame K(Al associated
with a body and having its origin K(cJ
at an arbitrary point A. The tensor
components in this case will be
I (ikA)
= )-,, ma {6tk [~ (x~~>) 2 ] -x~~lxft~l}
a l
(23.16)
The values of I~~> are related by
simple expressions to the values
I ih of the tensor components deter-
mined with respect to the frame K(c> Fig. 23.1.
with its origin at the point C and
with axes parallel to those of the
frame K(Al (Fig. 23.1). To find these relations, let us use the symbol a
to designate the position vector of the point A in the frame K(c>·
Hence, for any point of the body, r(A) = r - a and, consequently,
x~A> = x 1- a 1 (23.17)
where a1 is the i-th coordinate of the point A in the frame K(c>•
Let us introdu'ce the values (23.17) into formula (23.16):
I~t> = ~ ma {6;k [~~ (Xza- az)2]- (Xta- a;) (xka- an)}
a l
= 2Ja. ma {6tk (~l Xfa.) -Xta.Xka} + 2Ja ma. (6tk (2J1 aF)- a;ak}
- ~ ma.Oik 2J 2xzaa 1 + ~ ma.X 1aak + Y ma.Xkaa 1
a. l a. a.
The first of the five sums on the right-hand side is I 111 • In the second
sum, none of the quantities in the braces has the subscript a. In
addition, '', a[ is the square of the vector a, i.e. a 2 • Therefore, the
second sum can be written as m(a 2o111 - a 1a 11 ). The third sum can be
written as ~ 2o 111 a 1 )1, ma.Xla.· But ~ maxla = mx1c = 0, so that
I a a.
the third sum vanishes. Similarly, factoring out the multiplier in
94 MECHANICS
the fourth and fifth sums that does not depend on a and taking into
account that xiC = x 11 c = 0, we find that both these sums also vanish.
We thus arrive at the relation
(23.18)
which is the tensor form of writing the parallel axis (Steiner) theorem.
To verify this, let us fmd the component I~~> = IW. According to
(23.18)
I~~>= lzz + m(a 2 - a~)= Izz + m(a~ +a~)= Izz + ma}_
(23.19)
where a.L is the distance between the axes z and z<A>.
We calculate the kinetic energy of a body rotating with the angular
velocity ro about a stationary (in the frame K) axis fixed relative
to the body and not passing through its centre of mass. If the centre
of mass is at a distance ·a .L from the axis of rotation, its velocity is
V c = roa .L· Consequently,
T trans= 21 mro 2aJ.
2
where m is the mass of the body. We find the rotational energy by

formula (23.14), directing the z-axis parallel to the stationary axis of
rotation. Summating both energies, we obtain
T = 21 mro 2a.L2 + 21 I zzffi 2 = 21 ( ma.L2 + I zz ) ro 2 = 21 J<A>
zz ro 2 (23.20)
[see formula (23.19)1.

Consequently, the formula for the kinetic energy of a rotating
body considered in the general course of physics is true not always,
but only in two cases: (1) when the body rotates about one of its
principal axes of inertia [see formula (23.13)], and (2) when the rbody
rotates about a stationary axis fixed in it [see formulas (23.14) and
(23.20)1.
In concluding, we shall find the form of the inertia tensor when
only on~ of the coordinate axes, say tho z-axis, coincides with one
of the principal axes of inertia of a body. The transition from a fr<Jme
all of whose axes coincide with the principal ones to the frame we
are interested in is achieved by rotation about the z-axis through the
angle cp. It is easy to see that the table of the transformation coeffi-
cients is as follows in this case:
a 11 a 1z. 0 cos cp sin cp 0
a 21 a 22 0 = - sin cp cos cp 0
0 0 1 0 0 1
The components of the inertia tensor in the reference frame we
are interested in are obtained from the components of the tensor
reduced to the principal axes according to the transformation for-

mula (X.10). Let us represent the components of the tensor (23.10)
in the form lzm = 61ml m· Therefore, denoting the components of
the tensor in the new "primed" frame simply by I ik (without a
prime), we can write
Ilk= ' : CXttCXI!m6lml m = ~ CXtmCXkm[ m
I, m m
According to this formula

Iu = ~ aim! m = a~ 1 1 1
m
+ ai 212 = cos2 cp/1 + sin 2 q>/2
Similar calculations show that I a= :o.in 2 r:p/ 1 + cos 2 crl 2 • 1 33 =
=la, II2=l-z.1=sinq>coscp(/2-l1). /13=l31=0, /23=
= I 32 = 0. Hence, the inertia tensor in the new frame has the form
/11 /12 0 )
(I,k) = ( /21 /22 0 (23.21)
0 0 13
We must note that when / 1 = 1 2 =I, it follows from the for-
mulas we have obtained that / 11 = I 22 =I, and 1 12 = 1"!. 1 = 0,
i.e. that the new tensor does not differ from the initial one. This is
exactly how matters should stand because when the moments I 1
and I 2 are equal, the principal axes x and y are not fixed.
24. Angular Momentum of a Rigid Body

As in the preceding sections, we shall consider the motion of a
rigid body in the reference frame K with the axes X 1, X 2 , X 3 • We
shall rigidly associate with the body the frame K' whose origin will
first be placed at an arbitrary point A. We shall designate the axes
of this frame by x1 , x 2 , x 3 (or x, y, z). \Ve shall mentally divide the
body into particles of mass ma..
By formula (21.8), the veloeity of the a-th particle in the frame
K is
(24.1)
where VA is the velocity of the origin of the frame[(', (tl is the angular
velocity of rotation of the body in the frame K, and r~A) is the
position vector of the particle emerging from the point A.
Let us find the angular momentum M(A) of the body relative to
the origin of the frame K' (relative to the point A). The position
vector leading from the point A to the a-th particle is r~A>. Hence,
M<A> = ~ [r~A), ma. V a.1

a
'96 MECHANICS
:Substitution into this expression of the value (24.1) for Va yields

(24.2)
'The second term in (24.2) is the value M(AJ which the angular momen-
tum would have provided that the point A were stationary. Con-
i'!equently, M(AJ is the angular momentum due only to rotation of
a body.
Let us transform the first term in (24.2) using the distributivity of
.a vector product:
2J·a [r~l, ma VA]= ]2a [mar~A), VA]=[(~a mar~l), VA]
= [mr<AJ,
c VA]= [r<AJ,
c mVAJ
'Here m is the mass of the body, and rtAJ is the position vector of the
·centre of mass in the frame K' (the position vector from A to C).
Expression (24.2) can thus be given the form
(24.::!)
We have already noted that the term M(A) is due to rotation of the
body. It can be called the proper angular momentum of the body.
The second term is due to the translational motion of the body.
If we place the origin of the frame K' (i.e. the point A) at the
·centre of mass of the body, rtA) vanishes, and as a result (24.3) be-
·comes M(c) = M(CJ· It thus follows that the angular momentum of
.a body relative to its centre of mass observed in the stationary frame
K coincides with its proper angular momentum, i.e. it is determined
·only by rotation and does not depend on whether the centre of
mass of the body is moving or is at rest. .
Let us find an expression for the angular momentum 1 of a body
Telative to its centre of mass. If the point A coincides with C, the
;first term in (24.2) vanishes. Consequently,
M(cl = ~ ma Ira., [ro, ra.JJ (24.4)
a.
·(recall that ra. is the position vector of a particle emerging from
point C). Let us transform this expression according to formula
·(VI.S):
M<c> = ~ ~
L ma {ror~- r a. (ror a)}=..:...:. ~ ~
mo: {ro ( /-1 xra.)- r a (.2.; rokxka)}
a a l k
{we have expressed the scalar products in terms of projections of the

relevant vectors onto the axes of the frame K' associated with the
body).
1 Observed in the frame K. In the frame K', the body is at rest, so that the
.angular momentum in this frame is always zero.
Let us calculate the components of the vector 1\I(c) along the axes
>f the frame K' (in the following formulas we shall omit the sub-
;cript "C" in parentheses). For the component along the i-th axis,
we obtain
L,et us write w1 in the form

Wt = _>j c511t(l)ll
k
H:ence
f"inally, let us change the sequence of summation over the subscripts

:x. and k:
The expression in brackets is a component of the inertia tensor JUt

[see formula (23.6)]. Consequently, for the projection of the vector M
onto the i-th axis of the reference frame associated with the body,
we get the following expression
M 1 = )J l "w"
1 (i = 1, 2, 3) (24.5)
"
(do not forget that wh is the projection of the vector w onto the
k-th axis of the frame K').
Examination of formula (24.5) shows that the vectors M and w
are, in general, not collinear to each other. If the axes of the frame
K' (i.e. the axes x,. y, z) arc directed along the principal axes of
inertia of a body, formula (24.5) becomes simplified as follows:
Mt = ItWt (i = x, y, z) (24.6)
Here I 1 is the i-th principal moment of inertia.
Assume that a body rotates, for example, about the third prin-
cipal axis of inertia. Therefore, Wx = Wy = 0, and Wz = w, so that
M = 11-fz = IzWz = lzW
The last relation can be written in the vector form:

M = I zW (24. 7)
A glance at formula (24.6) shows that for a spherical top (i.e. a
body for which / 1 = / 2 = / 3 = /), the vectors M and w will also
be collinear, and M = I w.
If a body rotates about the z-axis fixed in it that does not coincide
with any of the principal axes of inertia, Wx = Wy = 0, and Wz = w.
7-018
98 MECHANICS
Formula (24.5) therefore yields

M x = I,.zW, My = I yzW, lvf z = I zzW
The projection of the vector M onto an axis passing through the
point relative to which M has boon determined is called the angular
momentum of the body relative to
... this axis. Consequently, the angular
momentum relative, for example, to
the z-axis is
M z = IzzW (24.8)
We must stress that the vector M itself
in the last case is not collinear to the
vector w and rotates about the direc-
tion of ro together with the axes x
and y.
In conclusion, let us consider the
case when a body rotates about a sta-
tionary (in the frame K) axis fixed in
it that does not pass through its centre
of mass C (Fig. 24.1). Let us take an
arbitrary point A on the axis of rota-
Fig. 24.1. tion. We draw a vector which we
1 •. shall designate by the symbol a from
this point to the point C. Therefore, for each of the body's particles
r~A) = a + rez (24.9)
where r~A) is the position vector of the a-th particle emerging from
the point A, and rez, is the position vector of the same particle from
the point C.
According to our assumption, the axis of rotation is stationary.
Consequently, the velocity VA equals zero. Let us find the angular
momentum of the body relative to the point A, i.e. MeA>· By for-
mula (24.2)
MeA>=~ [r~A>, mez [ro, r~A>]]
ez
We introduce into this expression the value of r~A) from (24.9):
M<A>=2: [(a+rez), ma.[ro, (a+ra.)]] (24.10)
a.
Taking advantage of the distributivity of a vector product, after
simple transformations we can write expression (24.10) as follows:
MeA>· [a, [wa]J (2: mez) +[a, [ro, (2: mezrez)J]
a a
+ [(2.j mezrez), [roa]]
a.
+ ,22a [r 01 , mez [roraJ1 · (24.11)
Since ~ mara = mrc = 0, the second and third terms in the expres-
sion we have obtained vanish. Using formula (VI.5) and substituting
the mass m of the body for ~ ma, we transform the first term into
m lwa 2 - a (roa)l = m l(afl -1- al) (') - a 11 (t)a],
where an is the component of the vector a parallel to ro, and a.L
is the component of a perpendicular to ro (a J. is the distance between
the axis of rotation and the z-axis parallel to it and passing through
the point C).
Finally, a comparison with formula (24.4) shows that the last
term in (24.11) is M<c>-the angular momentum of the body relative
to the point C. Hence, in the case shown in Fig. 24.·1
l\'I<A> = m (an+ al) ro - ma 11 wa + .M<c> (24.12)
Let us find the component of the angular momentum (24.12) along
the axis of rotation. We shall designate it M\f>. The component of
the first term equals this term itself. Since wa 11 = a 11 w, the com-
ponent of the second term can be written as maf1w. By formula (24.8),
the component of the third term is I zzW, where I zz is the moment
of inertia of the body relative to the z-axis (passing through the
point C). Consequently,
+ +
Mff> = ma}. ro I zzW = (ma}. I zz) ro = I~~>w (24.13)
Here I~1> is the moment of inertia of the body relative to a fixed
axis of rotation passing through the point A determined by the
parallel axis theorem (see formula (23.19)].
We must note that the vector M<A> itself, in
general, does not coincide in direction with
It
the vector ro. T
25. Free Axes of Rotation ..... .... __
Free axes of rotation of a body are defined
to be axes that retain their position in space
without the action of external forces on them.
We shall prove in the present section that only o
the principal axes of inertia can be free axes
of rotation.
Assume that a body rotates at the angular Fig. 25.1.
velocity w about a fixed axis associated
with it (Fig. 25.1). The following acceleration must be imparted
to each particle of the body:
Wa. = -w2pa
where Pa. is the component of the position vector ra.
of the given particle perpendicular to the axis of rotation
,.
100 MECHANICS
(rr.z emerges from the point 0 on the axis ~of rotation 1 ).

To impart such an acceleration to a particle, the force
(25.1)
must be applied to it whose moment relative to the point 0 is
No: = [rex, Fo:l = -mo:(J) 2 [ra., Po:l (25.2)
By summating all the forces (25.1), we get the resultant external
force that has to be applied to a body to ensure its rotation about the
axis being considered 2 :
(25.3)
The resultant moment of the external forces must equal the sum
of the moments (25.2):
(25.4)
Let us associate with a body a system of coordinates having its

origin at the point 0 and its z-axis directed along the vector ro. The
components of the vector Pa. along the axes of such a system are
Xa., Yo:, 0. Consequently,
ex e, Cz
[r a., Pa.l = X a. y ex zo:
Xa. Ya. 0
so that the components of the vector [ra., Pal are
[ra, Palpr.x = Ya ·0 - Za.Ya = -YaZa.
[ra., Palpr.y = ZaXa. - Xa ·0 = Xo:Za (25.5)
[ra, Palpr.z = XaYa. -· Ya.Xa = 0
Now let us write tho components of the resultant foree F anrl the
resultant moment N. By formula (25.3), we have
Fx = - (1) 2 :LJ ma.Xa. = - (J) 2mx0
a
Fy= -(1)2 :LJ ma.Ya.= -(J)2myc
a
1 In accordance with the notation we have adopted, this vector ought to be

designated by the symbol r~>. But since in this section we shall not encounter
position vectors emerging from the point C, we shall suppress the index "(0)"
on the symbols of vectors and coordinates to simplify our notation.
2 The forces Fa include both external and internal forces, but it is general
knowledge that the resultant of the internal forces is zero.
where xc and Yc are the coordinates of the body's centre of mass.

If the axis of rotation passes through the centre of mass, these coor-
dinates will be zero, so that all the components of the force and,
consequently, the force F itself, will vanish.
By formulas (25.4) and (25.5), we obtain
N 11 =- ro 2 ~ mo: (xo:zo:) = ro 2I xz
0:
Nz= - ro2 ~ m o: · 0 = 0
0:
where 1 11 z aud l.~z are the centrifugal moments of inertia of the body
[see formula (23.9)]. If the z-axis about which rotation occurs is one
of the principal axes of inertia, the centrifugal moments J.,z and I yz
are zero [see (23.21)] so that all the components of the resultant
moment of the forces and, consequently, the moment N itself vanish.
We have thus proved that when a body rotates about one of its
principal axes of inertia, the resultant of the external forces and the
resultant moment of these forces equal zero. Hence, for such an
axis of rotation to retain its position in space, no external forces
are needed.
26. Equation of Motion of a Rigid Body

Let us take as the generalized coordinates determining the posi-
tion of a body in the stationary frame K the Cartesian coordinates
X 1 c, X 2 c, X 3 c of the centre of mass (the position vector Rc corre-
sponds to them) and the Euler angles cp, {)·, ']~, and let us direct the
axes of the frame J{' associated with the body along its principal
axes of inertia.
We established in Sec. 23 that the kinetic energy of a rigid body
consists of the energy of translational motion (23.2) and that of
rotation, which with our choice of the axes of the frame K' is deter-
mined by formula (23.11). Hence, for the Lagrangian of a rigid
body, we can write the following expression:
(26.1)
To obtain an expression for L in the generalized coordinates we

have adopted, let us substitute Rc for Vc and express the projection$
of the vector ro onto the axes of the frame K' in terms of the Euler
102 MECHANICS
angles [see (22.4)1. The result is

1 • 1 . .
L = 2 mRt + 2 {1 1 ( qJ sin -&sin \jJ + {} cos \jJ )2
. . . .
+I2 (q1 sin 'l'l' cos \jJ- {}sin ~')2 +I 3 (<p cos{)'+ \jl) 2} - U (Rc, qJ, {)', \jl)
(26.2)
(remember that I 1 , I 2 , I~ are the principal moments of inertia of the
body).
Knowing the form of the function U (Rc, qJ, 'fr, ~1), we can compile
Lagrange's equation and solve the relevant problem on the motion
of a rigid body. Lagrange's equation corresponding to the coordinates
of the centre of mass has the form
d •
-mRc=
au
---=-VU=F
dt aRc
whence we get the equation of motion of the centre of mass of the
body:
mRc = F (26.3)
where F is the resultant of the external forces acting on the body.
To obtain an equation determining the variation of the angular
momentum M of a body with time, let us remember that for an indi-
vidual particle
'la. = lra., Fa.l = Na.
i.e. the time derivative of the angular momentum equals the moment
of the force acting on the particle. Summation over all the particles
of a body yields
• d
M=dt 2j Ma.= 2j [ra., Fa.]=N (26.4)
where N is the sum of the moments of all the external forces acting
on tlre body relative to the point C (the sum of the moments of the
internal forces is zero).
Let us write Eq. (26.4) in projections onto the axes of the frame K'
(onto the principal axes of inertia of a body):
(26.5)
The projection N 1 can be written as
w au oL
N, = - 6<D, = - ami = ami (26 · 6)
where 8<Di is the angle of rotation of the body about its i-th principal
axis ((J.); = d<Dddt). Indeed, when the body rotates through the angle
6<!>;, the forces applied to it do the work

8A = )-,. F6R = ~ F [(6CD;e 1), r]
Here summation is performed not over the subscript i. hut oYer all
the external forefls acting on the body. 6R is thP displncPnwn t of tht>
point of application of the relevant force. e; is the unit nctor uf the
i-th principal axis, and r is the position \"ector of the point of appli-
cation of the force emerging from the point C (R extends from the
origin of the frame K). After a cyclic transposition of the factors
and putting the common factor 6<D;e 1 outside the sum sign, we
arrive at the expression
8A = 8<!>;e 1 "\; [rFl ,= BcD;e,N = N;8Cl>;
where N; is the projection of the resultant moment of the forces
onto the axis about which rotation through the angle OCD; has oc-
curred. The work oA we have calculated is done at the expense of
the decrement of the potential energy U, i.e.
6A = Nt 6<!>1 = -6U
whence follows formula (26.6).
We have obtained formula (26.6) in considering rotation about
one of the principal axes of rotation of a body. This formula also
holds in the most general case--for rotation about an arbitrary
axis (naturally, provided that the force whose moment is being
considered is a potential one).
A glance at formula (26.6) shows that the quantities N 1 are gener-
alized forces corresponding to the generalized coordinates cD i [see
(4.20)].
Let us now differentiate function (26.1) \Vith respect to w;. The
result is
aL
--=l·w·
OW; I I
=M1
Isee formula (24.6)]. The projection of the angular velocity w; can
be represented as <il; [see the text following formula (2o.6)1. We
can therefore write that
M 1 = ..!!:__ = 0 ~ (26. 7)
8w; B<Dt
[compare with formula (11.3)1. It follows from formula (26.7) that

the quantities Jl![ i are generalized momenta corresponding to the
generalized coordinates <!>; [see (4.19)1.
Using relations (26.6) and (26.7), we can represent formula (26.5)
as follows:
d aL aL
(26.8)
Tt acb, = a<Dt
104 MECHANICS
i.e. as Lagrange's equation corresponding to the generalized coor-

dinate <Di. .
It should be borne in mind that with our selection of the axes of
the frame K' only the Euler angle 'ljJ corresponds to rotation about a
principal axis. The other angles correspond to rotations about the
fixed axis X 3 and about the nodal line.
By differentiating function (26.2) with respect to ~. we shall find
an expression for the projection of the angular momentum onto the
axis x 3 (onto the z-axis) in terms of the Euler angles:
.
M '1/> = M 3 = I 3 ( !Jl cos -6- + 'ljJ)
. (26.9)
Assume that the Euler angle -6- equals zero. This signifies that the
Z- and z-axes coincide all the time-the body rotates about an axis
associated with it that is fixed in the frame K. In this case, the sum
of the angles !Jl + 'ljJ determines the total angle of rotation of the
body about the z-axis. The position of the nodal line in this case is
indefinite-it may be located at any place between the X- and
x-axes. Partieularly, the nodal line can be made to coincide \Vith the
X-axis. Now !Jl = 0, and rotation of the body about the z-axis will
be characterized by the angle of proper rotation 'ljJ. If we make the
nodal line coincide with the x-axis, the angle 'ljJ will vanish, and
rotation about the z-axis will be described by the precession angle cp.
When -6- = 0, formula (26.9) becomes
M3
. .
= 1 3 (!p + 'tjJ) = l 3 ro 3
where ro 3 is the angular velocity of rotation of the body about the

z-axis.
When -6- = .rt/2, formula (26.9) is simplified as follows: r
M$ = M 3 =
.= l
l:~'l)J 3 ro'1/>
where ro'l/> is the angular velocity of proper rotation of the body.

In concluding, we shall compile Lagrange's equation for a body
rotating about an axis rigidly associated with it that is fixed in the
frame K. To be more general, we shall consider that the axis of
rotation does not pass through the centre of mass C of the body and
is not parallel to any of its principal axes of inertia.
We direct the axes Z and z along the axis of rotation of the body.
Hence, the Euler angle -6- will be zero. Since the nodal line is not
fixed in this case, we shall make it coincide with the x-axis, and the
angle 'ljJ will also vanish. With our choice of the z-axis (along the
•
direction of the vector ro), we have rox = ro!J = 0, and roz = u) = c+'·
We have calculated the kinetic energy for such a case on an earlier
page. It was equal to the value given by (23.20). Consequently, the
MECHANICS OF A RIGID BODY 105-
Lagrangian has the form
L = ; 1~~\\l)z- U (cp) = ; 1~~l~z- U (cp) (26.10}

(A is a point not coinciding with C at which the origin of llte fram&
K' is).
We compile Lagrange's equation:
d iJL iJL
Tt a~ = oq>
It follows from (26.10) that the left-hand side equals 1~1' ~- By

(26.6), the right-hand side, equal to -8U/8cp, is the projection of the-
resultant moment of the forces onto the axis of rotation, i.e. N z·
We thus arrive at the equation
1~1)
. = Nz
Cil
27. Euler's Equations

Lagrange's equations corresponding to the Euler angles (i.e.
describing the rotation of a body), as can readily be concluded from
the form of the function (26.2), are very complicated. It is sometimes-
more convenient to use other equations that were obtained by L. Euler-
and bear his name. To arrive at these equations, we shall proceed from
relation (26.4):
dM =N (27.1}
dt
Equation (27 .1) holds in an inertial reference frame (i.e. in the·
stationary frame K)~ We shall try to find an equation that holds in
the frame K' rotating together with the body whose axes coincide·
with the principal axes of inertia of the body.
In Eq. (27.1), dM is the increment of the vector M during the tim&
dt, observed in the frame K. By(15.6), this increment can be written
as
dM = d'M [dcp, M] +
where d'M is the increment of the vector M during the time dt, ob-
served in the frame K', and dcp is the angle through which the frame
K' rotates during the time dt. Dividing the last equation by dt, we
obtain
dM d'M
Tt=Cit+ at' M
[ dq> J (27.2)·
where dM!dt is the rate of the change in the vector M observed in the-
frame K, d'M/dt is the rate of change in the same vector observed in
106 MECHANICS
the frame K', and ro is the angular velocity of rotation of the frame
K' (i.e. the angular velocity of rotation of the body).
Formula (27.2) holds for any vector, particularly, it also holds
for the vector ro. In the latter cnse, we have
dro d'ro
dt =at+ [roroj
Since [<oro] = 0, we arrive at the relation
dro d' ro
lft=d't (27 .3)
from which it follows that the rates of the change in the vector ro
·observed in the frames K and K' are identical.
Let us substitute expression (27 .2) for the left-hand side of for-
mula (27.1). The result is the equation
d'M
at+ [roM]= N
Let us project all the vectors onto the i-th axis of the frame K',
taking into account that the projection of the vector M onto this
.axis is M 1 = / 1rol 1 [see (24.6)1:
M N
d' (/jWf)
dt + [ro ]pr.:.:l = i
Knowing that I 1 = const and d' roldt = dro/dt [(see (27 .3)], we can
write the equation obtained as follows:
(27.4)
Let us represent the projection of the vector product onto the

.axis x 1 by formula (VI.33):
[rol\l]pr.x 1= ">';
k,l
8tJtlrolt.MI = L Bt~tlroltiirol
k, l
.and introduce this expression into (27.4). As a result, we arrive at

:the equations
• dwt ~
It---a;-+ ~et~tzltrohroz=N 1 {i=1, 2, 3) (27.5)
k, l
which are Euler's equations. Assuming consecutively that i = 1,

i = 2, i = 3, and summating over k and l, we obtain three equations:
It d:: + (Ia- I 2) ro2roa = N t 1

12 dd~2 +(I 1 - I3)w 3 Wt=N2 } (27.6)
I3 d:t +(J2-I1) ro1w2=Na j

We must note that each following equation is obtained from the

preceding one by means of a cyclic transposition of the subscripts
1. 2, 3.
It is easy to see that for a spherical top, Eqs. (27.6) transform
into the equation I ro = N.
28. Free Symmetric Top

Let us use Euler's equations for studying the motion of a symmetric
top (i.e. a body for which I 1 = I 2 =/= I 3 ) not experiencing the action
of external forces. In this case, the centre of mass of the body moves
at a constant velocity or is at rest [see formula (26.3)1. Therefore,
from among all the inertial frames, we can choose the frame K
whose origin coincides with the centre of mass C of the body. In
this frame, translational motion of the body is absent, and it remains
for us to establish only the nature of the rotation.
In the absence of external forces, the angular momentum M of
the body remains constant in magnitude and direction relative to
the stationary frame K. Let us choose the direction of the vector M
as the Z-axis. We shall see that the vector l\1, generally speaking,
constrmtly changes its direction relative to the frame K' associated
with the body.
Since I 1 = I 2 and the moment of the external forces is zero. Eqs.
(27.6) have the following form 1 :
af+ (I a- /1) W2Wa =
I1 dw· 0 1
dw ·
2 -(Ia-I 1)WaWt==O ~
1 1-dt- (28.1)
dwa 0
dt= J
We directly obtain from the third equation that w3 = w11 = canst.
This signiftes that the projection of the angular velocity vector onto
the z-axis associated with the body remains constant.
Introducing the symbol
n Ia-II (28.2)
11 ~'WJJ=
we can write the first two of Eqs. (28.1) as follows:
.
wl = -Qw2,
.
(1)2 = Qwl
It is a simple matter to see that the system we have obtained is
satisf1ed by the functions
w1 = ro1_ cos (Qt +a), ro 2 = w1_ sin (Qt +a) (28.3)
-----
We remind
1 our reader that Euler's equations are written in a coordinate
system whose axes coincide with the principal axes of inertia of the body.
108 MECHANICS
where (J) .L and a are constants, (J) .L = Y (J)i + (J): being the magnitude
of the projection of the vector (J) onto the plane xy perpendicular to
the axis of proper rotation of the body-the z-axis. Inspection of
(28.3) shows that the component (J).L perpendicular to the z-axis,
remaining constant in magnitude, rotates uniformly in the plane xy
at the angular velocity Q determined by formula (28.2). We have
seen that the component rou parallel to the z-axis also remains con-
stant in magnitude. We thus conclude that the vector ro rotates
relative to the body at the angular velocity Q, describing a cone
about the z-axis, the magnitude of the
z vector ro remaining unchanged (Fig.28.1 1 ).
M According to (24.6), the projections of
the vector M onto the axes x, y, z are
lVfx = /l(J)lt My= /1(1)2, 111z = /3(1)3
Hence,
M = /l(J)lex + /l(J)2ey + Ia(J)aez
=l1 ((J)Iex + W2ey) + la(J)~ez
where ex, ey, ez are the unit vectors of
the relevant axes (these axes rotate to-
Fig. 28.1. gether with the body). The sum (J) 1e._, +
+ (J) 2ey gives the component (J) .L perpen-
dicular to the z-axis; (J) 3e 2 is the component ron parallel to the z-axis.
Hence,
M = I 1 roJ. + I 3 ro 11 (28.4)
The directions of the vectors M and ro 11 pass through a common
origin for the frames K and K', i.e. the point C. Consequehtly,
these vectors determine a certain plane (the plane Zz). For equality
(28.4) to be observed, the vector ro .L must be in the same plane.
Hence, the vector ro = ro.l -+ ro 11 is in the plane Zz. We thus conclude
that the vectors M, (J), and the axis z of proper rotation of the body
are in the same plane at each instant (this plane is hatched in
Fig. 28.1). The plane rotates about the direction of M. The z-axis
rotates together with it and describes about the Z-axis a circular
cone. Such rotation of the body's z-axis is called regular precession.
The latter is characterized by a constant nutation angle it.
We have established that the vector ro rotates relative to the
body about the z-axis at the angular velocity Q. At the same time,
this vector remains in the plane Zz. It thus follows that relative to
1 In practice, a body of revolution is taken as a symmetric top. We have de-
picted a body of an irregular shape in the figure, however, to underline the fact
that the only condition for a body to be a symmetric top is the equality of two
of its principal moments of inertia.
the plane Zz, the body rotates about the z-axis in the opposite direc-
tion at the same velocity Q.
Figure 28.2 shows the components I 1 ro J. and I 3 ro 11 whose sum
gives the angular momentum M [see (28.4)]. A glance at the figure
shows that
z
(28.5)
M
We established earlier that wJ. and w11

are constants. Therefore, tan -&, and,
consequently, the nutation angle{} itself
remain constant. The angle between the
vector w and the z-axis is also constant
(its tangent is w J./ roll)·
The quantities WJ. and ro 11 are deter-
mined by the body's kinetic energy T Fig. 28.2.
and the angular momentum M. Irtdeed,
T = { (I1roi+ Itwi+ law~)= {(Itroi + Iarofl)} (28.6)

M 2 = (I1ro1)2+ (I1"!2) 2 + (/aroa)2 = Iiro1 + Iirofl
Solving this system of equations for roJ. and ro 11 , we frnd the expres-
sions of these quantities in terms of T and M:
(28. 7)
Assume that the body is flattened along the z-axis. Hence, I 3 > I 1 ,
and the denominator in the radicand of the expression for ron will
be positive and for WJ., negative. Consequently, for ro, 1 and wJ. to be
real, it is necessary that the conditions 1lf2- 2TI 1 ~ 0 and
M 2 - 2TI 3 ~ 0 be satisfied; they can be combined into a single
formula:
M2 M~
u~r~ 2!·1 (28.8)
3
The energy of a free symmetric top cannot have values beyond the
indicated limits (at a given M). If the energy has its lowest possible
value, i.e. at llf2=2TI 3 , the quantity wJ. vanishes [see (28.7)1.
Examination of formula (28.5) shows that in this case {} = 0-the
axes Z and z coincide, the directions of the vectors M and w also
coincide, and the vector w does not move relative to the body or
relative to the frame K. We must note that in this case the following
relation holds:
(28.9)
110 MECHANICS
Assume that the energy has its highest possible value, i.e. 11!f2 =
= 2T/ 1 • Now w11 vanishes [see (28.7)1. By formula (28.5), we have
{} = n/2 in this case- the axes Z and z are mutually perpendicular,
there is no rotation about the z-axis, the vector ro = ro _1_ coincides
in direction with the vector M, and the energy is related to the angular
momentum by the expression
M2
T = 211 ( 28.1 0)
The relations we have obtained can be given a fine geometric

interpretation. Let us rewrite formulas (28.6) as follows:
rof ro~ro§
2T/!1 + 2T/! + 2T!I1= 1 8 (28.11)
~ {I)~ ~ 1
M2/Jf + M2ff§ =
'
M2fJf -r (28.12)
Each of these equations describes an ellipsoid of revolution. If,
as we have assumed, / 3 > / 1 , both ellipsoids are flattened along the
z w 3 Energy
----lf--~ellipsoid
Fig. 28.3.
axis u) 37 which coincides with the z-axis (an ellipsoid of inertia in

this case, conversely, is extended along the z-axis). It is a simple
matter to comprehend 1 that the ftrSt ellipsoid (let us call it an
energy ellipsoid) is flattened less than the second one (which we shall
call the angular momentum ellipsoid). Figure 28.3 shows both
ellipsoids. The values of w1 , w2 , w3 satisfying Eqs. (28.11) and
(28.12) are determined by the lines of intersection of both ellipsoids.
It thus follows that the tip of the vector ffi must slide along this
line of intersection. Consequently, the vector ffi rotates relative to
the axes w1 , w2 , w3 , describing a cone about the axis w3 • It must be
remembered that ffit is the projection of ro onto the i-th principal
1 For this purpose, we must take into account that /§/If > / 3//1 •
axis of inertia of the body. Hence, the axes wi coincide with the
axes x, y. z. Rotation relative to the axes lu; thus signifies rotation
rclativ£> to the body itself.
We can increase !t1 with T remaining constant until the energy and
angular momentum ellipsoids have only two common points. This-
occurs provided that Y2T// 3 = M/1 3 [compare with (28.9)!. In
this case, the vector w coincides in dit·cctioll with the z-axis. If
we continue to increase 111, i.e. assume that Jll~/2/ 3 > T. the ellip-
soids stop touching each other, and the system of equations (28.11)
and (28.12) has no common solutions. Such a case cannot be realized.
Consequently, we have obtained the lower boundary for T [see·
(28.8)1.
By diminishing M with T remaining constant, we arrive at a
situation when both lines of intersection of the ellipsoids merge
into a single line of contact in the equatorial plane. This occurs
provided that Y2T// 1 = M/1 1 [compare with (28.10)]. In this case,
the vector ro rotates about the z-axis, constantly remaining per-
pendicular to it. If we continue to decrease 111, i.e. assume that
T > Jvl2/2/ 11 the ellipsoids stop touching each other, and the system
of equations (28.11) and (28.12) has no common solutions. We have
thus arrived at the upper boundary for T [see (28.8)].
29. Symmetric Top in a Homogeneous Gravitational Field

Consider the behaviour of a symmetric top with one fixed point
in a homogeneous gravitational field. We must note that a general
solution1 of the problem on the motion of a body with one fixed
point in a homogeneous gravitational field can be obtained only in
three cases:
(1) for an asymmetric balanced top (a top is called balanced if the·
fixed point coincides with the centre of mass of the top). This case
is known as Euler's problem;
(2) for a symmetric unbalanced top (the fixed point does not coin-
cide with the centre of mass) in which the centre of mass is on the
z-axis-Lagrange's problem;
(3) for a symmetric unbalanced top for which / 1 = I 2 = 21 3 , and
the centre of mass is in the plane xy-the problem of S. Kovalevskaya.
We shall consider Lagrange's problem. The equations of motion
in this case are integrated in a very complicated way. \Ve shall
therefore limit ourselves to writing the initial equations and dis-
cussing their solutions.
We place the origins of both coordinate systems K and K' at the
fixed point A of the top (at the point of support). We direct th&
1 That is, the solution obtained with the aid of quadratures with arbitrary
initial conditions.
112 MECHANICS
Z-axis of the stationary frame K along a vertical line, and the z-axis
-of the frame K' associated with the top along its third principal
axis of inertia (/ 1 = I 2 =f= I 3 ). With such a choice of the coordinate
.axes, the top's potential energy has the form U = mgl cos 'l't, where
J, is the distance from the supporting point to the centre of mass C
(we assume that the coordinate z of the centre of mass, i.e. zc, is
-greater than zero).
Let us find the expression for the kinetic energy in the given case.
With a view to the point A being fixed, we can write
T =-} ~ ma. [ro, r~A>]Z

a.
:and perform the same transformations that we made for expression
{23.3). As a result, we arrive at a formula which will differ from
(23.5) only in containing the coordinates x~~>, x~!>, etc. instead of
the coordinates x 1a., xha.• etc. Consequently, we obtain an expression
similar to (23.8) for the kinetic energy:
T = 21 "Q
LJ
/(A)
ih (J)i(J)k
f1 •
where I~t> are the tensor components determined by formula (23.16).

According to (23.18), these components are
I~t> = I 111. + m(a 2f>th - aiah)
In our case, a 1 = a 2 = 0, and a 3 = -a = - l [here a 1 is the i-th

coordinate of the point A in the frame Kf.c>• see (23.17)1. In addition,
-since the z-axis coincides with the third principal axis of inertia,
while the axes x and y are parallel to the two other principal ax~s1 , we
have I 1h = I 1f>th· We can thus conclude that the tensor I~t> is
diagonal, its non-zero components being
/~A> = 11 + mP, J~A> = / 2 + mZZ, J~.A> = /3
Hence, taking into account that / 1 = 1 2 , we obtain the following
-expression for the kinetic energy:
1
T = 2 [(I 1 + ml 2 ) ( ro~ + ro:) +I3ro:1
Introducing values (22.4) for the components of ro into this expres-
-sion, we obtain
1 In a symmetric top, any two mutually perpendicular axes perpendicular

to the axis of symmetry may be principal axes of inertia.
MECHANICS OF A RIGID BODY
Let us write the Lagrangian1

1 • •
L =2 (I1 + ml2) (rp2 sin2 it+ it2)
1 . •
+2 I 3 (q> cos{}+ ljl) 2 -mgl cos it (29.1)
The coordinates rp and 'i' are cyclic [see the text in Sec. 11 related
to formula (11.2)]. Therefore, the generalized momenta prp and Po;
(a) (b) (c)

Fig. 29.1.
are integrals of motion. The energy E is a third integral of motion.

We thus have three equations:
Prp =oLio~= [(I1 +ml 2 ) sin 2 it+ I 3 cos 2 it]~+ I 3 cos~= Mz = const
P$ =oLio~= I 3 (~ cos it+~)= Mz = const2
E=T +U = const
An analysis of the solutions of these equations leads to the follow-
ing results. The angle it varies periodically within the limits from
-&1 to it 2 determined by the initial conditions (particularly, by the
relation between the energy and the angular momentum of the top).
The oscillations of the top's axis corresponding to the variations of
the angle it are called nutation. Simultaneously, the axis of the top
precesses, i.e. turns about the Z-axis. As a result, the apex. i.e. the
point of intersection of the z-axis (the axis of the top) and a sphere
of unit radius, describes one of the curves shown in Fig. 29.1. The
sign of the derivative~ either remains unchanged (Fig. 29.1a and b)
. .
or changes (Fig. 29.1c). Case (b) occurs when cp and it simultaneously
vanish.
1 The force applied to a top at its point of support is a reaction of the con-
straint, which, as we know, is not included in Lagrange's equations.
2 Compare with (26. 9).
8-018
114 MECHANICS
The nature of the behaviour of ~' like the values of '1'1 1 and '1'1 2 ,
depends on the initial conditions. Motion of the top's axis such as
that in case (b) corresponds to natural initial conditions when the
top is first brought into rotation about its axis, after which the axis
. .
is released and begins its motion. At the moment of axis release,
both cp and '1'1 equal zero. In these initial conditions, the top fust
inclines, and then upon reaching the boundary angle '1'1 2 it begins to
rise (see Fig. 29.1b).
In absolutely specific initial conditions, both boundary values of
'1'11 and '1'12 coincide so that the top's axis precesses without nutation.
As we have already noted, such precession is called regular. To ob-
tain regular precession, a top must be given an initial impetus of
a quite definite magnitude and direction.
For a "rapid" top (i.e. a top whose kinetic energy of proper rotation
is high compared \vith its energy in a gravitational field), the action
of gravitational forces may be disregarded in a f1rst approximation.
Consequently, the motion of the top can be represented as the free
precession of the top's axis about the direction of the angular momen-
tum M, considered in Sec. 28 (this precession corresponds to nutation
of a heavy top), onto which small disturbances due to the action of
the gravitational force are superposed. These disturbances cause
slow precession of the angular momentum M about a vertical line.
Calculations show that the more rapidly a top rotates, the smaller
is the amplitude of nutation. In addition, in a real rapid top, nutation
is damped by friction at the support. Therefore, in practice the
nutation of a sufficiently rapid top is sometimes unnoticeable, and
the top seems to uniformly precess about a vertical axis. Since such
precession is regular only approximately, it is known as pseudo-
regular precession.
Chapter VI
CANONICAL EQUATIONS
30. Hamilton's Equations
In solving problems on the motion of a system with s degrees of

freedom using Lagrange's equations, we have to solve a system of
s second-order differential equa~ions. The generalized coordinates qk
and the generalized velocities qk are the independent variables in
these equations.
W. Hamilton obtained equations of motion in which the general-
ized coordinates qk and the generalized momenta Pk are the indepen-
dent variables. Hamilton's equations or, as they are also called,
canonicaP equations (qk and Pk are accordingly called canonical
variables), unlike Lagrange's equations, are first-order differential
ones. But on the other hand, their number needed to describe a
system with s degrees of freedom is 2.~.
Hamilton's equations can be derived either from Lagrange's
equations or directly from the principle of least action (we shall
give both derivations below). It is natural that they give nothing
novel in essence. But canonical equations are more symmetric than
Lagrange's equations and, in addition, being invariant with respect
to canonical transformations, they unveil great possibilities for
generalizations playing an important role in electrodynamics, sta-
tistical physics, and quantum mechanics.
Hamilton took the energy (5.1) expressed in terms of the variables
qk and Ph as the funetion characterizing a mechanical system. Hav-
ing in view that
(30.1)
1 Hamilton's equations are called canonical because they remain invariant

upon quite general transformations of the variables. With the aid of such ca-
nonical transformations, we can pass over from the variables qk and Pk to other
canonical variables: Qt (q 11 , Pk· t) and P, (q 11 , Pk• t). Here Hamilton's equations
retain their form, true, with the new Hamiltonian H' (Qt. Pt. t) that replaces
the function Jl (q1., Pk• t). The variables Qt and P1 may have a different physical
meaning than the variables qk and Pk·
8*
U6 MECHANICS
[see (4.19)1, let us write this function as follows:
H (q", P"' t) =' ~

k
p"q"- L (q", q", t) (30.2)
(qk is assumed to be expressed in terms of q" and Pk)· The character-

istic function H is called the Hamiltonian function or simply the
Hamiltonian.
We shall give as an example the Hamiltonian of a particle moving
in the potential field U = U (x, y, z, t):
H= 2~ (pi+ P~+ pi) +U (x, y, z, t) (30.3)
For a particle moving in a stationary field, H has the same form, but
U does not depend explicitly on t.
Let us derive Hamilton's equations proceeding from those of
Lagrange. For this purpose, we shall find the total differentials of
the left-hand and right-hand sides of formula (30.2) and equate
them to each other. The total differential of the left-hand side is
(30.4)
The differential of the right-hand side is

• • {)L lJL • {)L
dH = ~ Pk dq 1t
k
+ ~ q" dp"- ~
k k
aq" dq"- ~ - . - dq" -Tt dt
k {)qk
(30.5)
In view of relation (30.1), the first and fourth sums cancel each other.
It follows from Lagrange's equation (4.16) 1 that
{)L d {)L
aqk = Tt-.-= Ph. (30.6)
{)qh.
p
.Let us substitute 11 for fJLifJq 11 in the third sum of formula (30.5).
As a result, expression (30.5) becomes
~ • ~ • lJL
dH = £.J qk dpk- L.J P11. dqh. ---atdt (30.7)
k k
For expressions (30.4) and (30. 7) to be equal at arbitrary values
<lf dq 11 , dpk, and dt, the following conditions must be ob-
1 We assume that all the forces acting in the system are potential ones;
this is why we use Eq. (4.16) instead of (4.15).
CANONICAL EQUATIONS H7
served:
an
P~t= -- - (k= 1, 2, ... , s) (30.8)
iJqk
ran aL
Tt=-Tt (30.9}
Equations (30.8) are the required Hamilton or canonical equations.,

As we have already noted, they are first-order d fferential equations.-
Their total number is 2s.
For a particle described by the Hamiltonian (30.3), Hamilton's
equations have the form
au ;_.!!£ t
P:&= --a;-· m ' e c.
... -
Now let us deri':e Hamilton's equations from the principle of

least action. We remind our reader that according to this principle,
a system moves so that the action S [see (7 .1)1 has the smallest poss-
ible value. This statement is written in the form
t,
.,
fJS = 6 ) L (q 1 qk, t) dt = 0
t,
Let us introduce into this equation the value of L obtained from

relation (30.2):
t,
8 J (~ P1tq1t-H) dt=O
t, It
(30.10)
The variation on the left-hand side of (30.10) can be written as
Let us integrate the first term by parts:
Here we have taken advantage of the fact that 8qk = (8qk) :t

[see (111.4)1. The variations t3qk vanish when the integration limits
1.f8 MECHANICS
are introduced 1 • Therefore, the first term of the expression obtained

must be discarded. Consequently, condition (30.10) will. be written
.a.s. follows: ·
Owing to the arbitrary nature of the variations 6qk and 6pk, this
eondition can be satisfied only if the expressions in parentheses will
be zeros. Hence we directly obtain Eqs. (30.8).
Let us investigate the Hamiltonian H. We find the total derivative
of this function with respect to time:
Taking into account values (30.8) for qk and Pl<t we find that
dH an (30.11)
Tt=---at
Thus, if the function H does not depend explicitly on the time, it
retains its value. This was to be expected because H is the total
energy of a system that is retained provided that
oL!ot = 0 [see (30.9)1.
Let us replace Pk in the second of equations (30.8) in accordance

with (30.6). As a result, we find that
(30.12)
lt thus follows that the generalized coordinates which are cyclic,

i.e. which are not contained explicitly in the function L, are also
not contained explicitly in the function H. On an earlier page [see
.(11.2)], we established that the generalized momenta corresponding
to cyclic coordinates are integrals of motion. We can conclude from
what has been said that the generalized momenta corresponding
to the coordinates qk not included explicitly in the Hamiltonian
{i.e. cyclic relative to the function H) remain constant:
Pk = const provided that ~H = 0

vqk
(30.13)
. 1 We remind our reader that in the variation of trajectories in a configura-

tion space (see Sec. 4), the initial and final points of the trajectories are assu-
med to be fixed.
CANONICAL EQUATIONS 119
31. Poisson Brackets
Let us take a function of the canonical variables q,. and Ph· and
also of the time t, that is, f (qh, Ph• t) and establish in what condi-
tions this function will be an integral of motion 1 . To do this, we
calculate the total derivative of this function with respect to time:
. .
We substitute the values for the derivatives qk and Ph from (30.8):
(31.1)
If we have two functions <p (qk, p 1.., t) and tj.1 (qh, Ph• t), the
expression
(31.2)
is called the Poisson bracket for the functions <p and 'lj.l 2 • When nec-
essary, the independent variables with respect to which the partial
derivatives are taken are indicated as subscripts of the symbol of
the Poisson bracket, i.e., for instance, bracket (31.2) is written as
{cp, tj.1} •. It is easy to see that
q,p
(31.3)
Using the Poisson bracket, expression (31.1) can be represented as

df at
"dt=-at+U. H}q. p (31.4)
or as
(31.5)
A glance at (31.4) shows that the condition in which t.he function f
Do not forget that the functions of dynamic variables (qh and ~h or qh and
1
Pk) that remain constant in the motion of a system are known as integrals of
motion.
2 Sometimes square brackets and even round ones (parentheses) are used to
designate the Poisson bracket instead of curly brackets (braces).
120 MECHANICS
is an integral of motion is
at
at+{/, H}=O (31.6)
We thus conclude that when the integral of motion f does not depend
explicitly on the time, its Poisson bracket with the Hamiltonian
equals zero.
Below are given some obvious properties of the Poisson bracket:
{cp, lp}= -{¢, cp} (31. 7)

{cp, cp}=O (31.8)
{(cp1+cp2), 'i'}={cpt, 'i'}+{(P2, 'i'} (31.9)
{(cpicp2)' 'I'}= cpd cp2, "'} + cp2 {cpi, 'I'} (31.10)
:t {cp, 'I'}={~~' 'll}+{cp, ~;} (31.11)
Particularly, we may take canonical variables as cp or tj), or as both

of them. This yields the following relations:
{q,, 'I'}= :~ (31.12)

iJ'I'
{p,, "' } = - iJqi (31.13)
{q,, ql!} = 0 (31.14)

'
{Ph P~t}=O (31.15)
{q,, P~t} = lltit (31.16)
'We invite our reader to obtain formulas (31.12)-(31.16) as an exercise.

In deriving them, take into account that op 1/oq 11 = 0, and oq;fop 11 =
, 0 (because q11 and p 11 are independent variables).
A very important property of the Poisson bracket is its invariance
relative to canonical transformations. This signifies that
(31.17)
where Q and P are variables obtained from q and p with the aid
of canonical transformations.
In quantum mechanics, we shall acquaint ourselves with the quan-
tum Poisson bracket that is the' quantum mechanical analogue of
the classical Poisson bracket treated in this section.
CANONICAL EQUATIONS 12t
32. The Hamilton-Jacobi Equation

Variation of the ac.tion
s= r
tt
Ldt (32.1}
in finding the true trajectory of motion of a system (we have in

mind the trajectory in configuration space, i.e. in a space with s
dimensions; s is the number of degrees of freedom) consists in com-
paring the values of S for close trajectories with fixed ends, i.e. with-
Fig: 32.1. Fig. 32.2.
identical values of q,(t 1) = q},11 and qk(t 2) = q'l..2 ,. This can be-
illustrated with the aid of Fig. 32.1. Only the trajectory for which S'
is minimum corresponds to actual motion (it is depicted by a solid
line in the figure).
In the present section, we shall consider the action S as a quantity
characterizing motion along true trajectories, and study the behav-
iour of this quantity upon changes in the point q< 2 > (with t 2 =
= canst), and also upon changes in t 2 [the symbol q< 2 > signifies the·
set of all the ql,2 ,s]. We shall thus treat the action as the function
S = S(q,, t) (32.2}
where q, are the coordinates of the final position of the system, and t
is the instant when this position is reached.
Let us take near the point q< 2 > a point with the coordinate q< 2 > +
+ Oq which the system reaches at the same instant t 2 in which it.
'122 MECHANICS
.arrives at the point q< 2 > (Fig. 32.2). The action for the trajectory bring-
ing the system to the point q< 2 ) -~ oq differs from the action for the
trajectory bringing the system to the point q< 2 > by the quantity
!2
oqk + -.- 6q" dt

-.
6S = I
J
:2:: ( -,-
fJL
vqk
fJL
fJqL
' )
(32.3)
!t k ..
Here {)qk is the difference between the values of qh taken for both
trajectories at the same instant t; similarly, oq~ is the difference be-
tween the values of qh at the instant t.
Let us integrate the second term in (32.3) by parts:
t2 !2 !2
~ {)~ 6qh dt = {)~ 6qk

tt oq,. aq k tt
- ~ (
tl
:t {)~ )6q,. dt
aqk
(32.4)
For the true trajectory, aL!aqh is the generalized momentum Ph·

The origins of both trajectories coincide, hence oqk (t 1) = 0. The
-quantity 6qk (t 2 ) can be designated simply by 6qh. Consequently,
the first term in (32.4) can be represented in the form phoq,..
Let us introduce (32.4) into expression (32.3):
6S = ~ Ph6qk + ~ :~- ~
k
r(
t1 k
a.L ) 6q,. dt
oqk
True trajectories satisfy Lagrange's equations. Therefore the inte-

;grand and, consequently, the integral itself, vanishes. We have thus
Qbtained the following value for the increment of the action S due
to the change in the coordinates of the final position of the' system
.by 8q,. (at a constant time of motion):
oS = ~ P~tflqk (32.5)
k
Here .Ph is the value of the momentum at the instant t 2 •

It follows from expression (32.5) that
fJS
fJq11. =Ph (32.6)
•Consequently, the partial derivatives of the action with respect

to the generalized coordinates equal the corresponding generalized
momenta.
Now assume that the upper limit of integration in (32.1) is not
flxed. To underline this fact, we shall write the action in the form
t
S= ~ Ldt (32. 7)
!J
CANONICAL EQUATIONS 123
The action represented in this way is a function of the upper integra-

tion limit, i.e. S = S (t). It can be seen from (32. 7) that
~~ = L (32.8)
At the same time, in accordance with (32.2), we can write that
as as "" as • as "" • (32.9)
dl=-at+ .LJ aqk qk = Tt+ .LJ PR.q,.
k k
[we have taken relation (32.6) into account]. Equating the right-
hand sides of expressions (32.8) and (32. 9). we get the following
value for the partial derivative of S with respect to t:
as
Tt =- ( ~ Pkqk-
• L)
I!
The expression in parentheses is the Hamiltonian H. Hence,
as
Tt = -H (qk, P~<• t) (32.10)
In accordance with formulas (32.6) and (32.10), the differential
of function (32.2) can be written as
dS = .2J· Pk dqk- H dt (32.11)
k
Let us substitute for the Prt'S in Eq. (32.10) their values from
(32.6) and write this equation as follo,vs:
as ( as as as )
-at+H qt, q2 , ••• , q.; aq 1 , aq 2 , ••• , aq. ; t =0 (32.12)
We have obtained a differential equation that must be satisfied
by the function S (q 1 , q2 , • • • , q.; t). It is called the Hamilton-
Jacobi equation. It is an equation in partial derivatives of the first
order.
Equation (32.12) is tho cornerstone of a general method of integrat-
ing equations of motion. But a treatment of this method is beyond
the scope of our course.
For a conservative system with stationary constraints, the time
is not contained explicitly in the function 1!, and H = E = const
[see (30.9)1. Consequently, according to (32.10), the dependence of S
on t is expressed by the term -Et. Therefore, the action breaks
up into two terms, one of which depends only on the generalized
coordinates, and the other only on the time:
(32.13)
The function S 0 (q 11 ) is called the contracted action. Introducing S
in the form of (32.13) into Eq. (32.12), we arrive at the Hamilton-
124 MECHANICS
Jacobi equation for contracted action:

H ( qh q21 ' ' ' I
aS 0
qS; -,-!
aS0 I
-a- aso ) = E
'"I -,- (32.14)
uql q2 uqs
The Hamilton-Jacobi equation plays an important role in optics
and quantum mechanics. It underlies optical-mechanical analogy 1
which led E. Schri:idinger to the formulation of wave mechanics.
Let us write the Hamilton-Jacobi equation for a particle moving
in a non-stationary potential field. Taking into account formulas
(30.3) and (32.12) 1 we obtain
1 {( as
2m Tx )2 + (as
ay )2 + (az
as )2} +U(x~ Y1 Z1 t)= -a;-
as (32.15)
If the field in which the particle is moving is stationary instead of

1
(32.15), an equation for the contracted action S 0 is considered:
2~ {( ~:0 ) 2+( a;Yo ) 2+( ~: 0 ) 2}+U(x, y, z)=E (32.16)

Chapter VII
THE SPECIAL THEORY

OF RELATIVITY
33. The Principle of Relativity
The special theory of relativity is based on two postulates formu-

lated by Albert Einstein:
1. All laws of nature are the same in all inertia reference frames.
In other words, we can say that the equations expressinr; the laws of
nature are invariant 1 with respect to transformations of coordinates and
time from one inertial reference frame to another.
2. Light always propagates in a vacuum at a definite constant speed c
not depending on the state of motion of the emitting body.
The first postulate is called Einstein's principle of relativity, and
the second is called the principle of constancy of the speed of light.
N e'vtonian mechanics proceeds from the assumption that in tar-
actions are transmitted instantaneously from body to body. This,
particularly, manifests itself in that the interaction of particles is
described with the aid of the potential energy U (r 1 , r 2 , • • • • ),
which depends only on the coordinates of the particles. It is thus
assumed that a change in the position of one of the particles affects
the other particles at the very same instant.
Actually, as shown by experiments, there are no instantaneous
interactions in nature. If the position of a particle changes, this
change begins to tell on another particle interacting with it after
a finite interval of time elapses, needed for the interaction propagat-
ing at a finite veloCity to cover a path equal to the distance between
the particles. Consequently, we must acknowledge the existence of
a maximum velocity of propagation of interactions. Experiments
show that this velocity equals c-the speed of light in a vacuum.
It also follows from the second pbstulate that the velocity of
propagation of interactions is the same in all inertial reference
frames, i.e. is a universal constant.
In accordance with Galileo's mechanical principle of relativity,
the laws of mechanics are invariant relative to the Galilean transfor-
mations:
z = z',. t = t' (33.1)
1 The invariance of an equation signifies that the form of the equation

does not change when the coordinates and time of one reference frame are
replaced in it with the coordinates and time of another frame.
126 MECHANICS
(vo is the velocity of the frame K' relative to the frame K). These
transformations lead to the classical law of velocity summation:
v = v' + v 0 (33.2)
The latter equation and, consequently, Eqs. (33.1) from which
it follows, does not ngree with Einstein's second postulate accord-
ing to which c = c' for a light signal.
Consider two inertial reference frames K and K'. We shall select
the coordinate axes of these frames so that the axes x and x' are
directed along the velocity v 0 of the frame K', and the axes y and z
y K y'
X x.'
Fig. 33.1.
are parallel to the axes y' and z' (Fig. 33.1 ). We shall begin to
measure the time in both frames from the instant when the origins
of the systems coincide. Assume that a light signal propagating
in all directions was sent at the instant t = t' = 0 from the coincid-
ing origins of coordinates. By the instant t, the signal in the frame K
will reach points at the distance l = ct from 0. The coordinates of
these points satisfy the equation
c2 t2 - x2 - y 2 - z2 = 0 (33.3)
Similarly, by the instant t', the signal in the frame K' will reach
points of a sphere of the radius ct'. The coordinates of these points
satisfy the equation
c2t'2 - x'2 - y'2 - z'2 = 0 (33.4)
Equations (33.3) and (33.4) have the same form, which manifests
the invariance of the law of propagation of light with respect to
a transformation of coordinates and time from one frame to another.
If we introduce into (33.3) the values of the unprimed coordinates
and time determined by formulas (33.1), we get the relation
c2t•2 - x'2 - y'2 - z'2 - 2vox't' - v!t'2 = 0
not coinciding with (33.4). Hence, we have again arrived at the
conclusion that the Galilean transformations are not compatible
with the principle of the constancy of the speed of light.
THE SPECIAL THEORY OF RELATIVITY f27
According to Einstein's principle of relativity, all the laws of

nature, including the laws of mechanics and electrodynamics. must
be invariant relative to the same transformations of the coordinates
and time performed in passing from one reference frame to another.
But Newton's equations and Maxwell's equations do not meet this
requirement. Whereas Newton's equations are invariant with respect
to the Galilean transformations, Maxwell's equations, as can readily
be seen by direct verification, are not invariant relative to these
transformations. This circumstance led Einstein to the conclusion
that Newton's equations need refinement, as a result of which the
laws of mechanics and electrodynamics would be invariant with
respect to the same transformations. The required modification of
the laws of mechanics •vas performed by Einstein. The result was
the appearance of mechanics agreeing with Einstein's principle of
relativity, which was given the name of relativistic mechanics. The
present chapter is devoted to a treatment of the fundamentals of
this mechanics.
34. Interval
An event occurring with a particle is characterized by the place-
where it occurred (i.e. a set of values of x, y, z) and the time t
when it occurred. If we introduce an imaginary four-dimensional
space along whose axes we lay off the space coordinates x, y, z
and the time t (or a quantity proportional to t), an event will be
characterized in this space by a point. A point depicting an event
in a four-dimensional space is called a world point. YVith time.
a world point corresponding to a given particle moves in four-di-
men:<>ional space, describing a line known as a world line.
Let us consider two events, the first of which consists in the emis-
sion of a light signa"! from a point with the coordinates x 1 , y 1 , z1
at the instant t 1 , and the second in the arrival of this signal at a point
with the coordinates x 2 , y 2 , z2 at the instant t 2 • The following
relation holds between the coordinates and time of these two events:
c 2 (t 2 - t 1 )2 - (x 2 - x 1 ) 2 - (y 2 - !h? - (z 2 - z1 ) 2 = 0 (34.1)
The quantity
(34.2)
is the square of the distance (or of the interval) between two points
in conventional space. vVe can speak in a similar way about the
distance (interval) between two points in a four-dimensional space.
The interval between two events is defined to be the quantity s12
whose square is determined by the formula
s~2 = C2 (t2 - tt) 2 - (x2 - Xt) 2 - CY2 - Yt) 2 - (z2 - Zt) 2
=c 2 (t 2 - t 1) 2 - l~ 2 (34.3)
'128 !MECHANICS
For two infinitely close events, the square of the interval is

ds 2 = c2 dt 2 - dl 2 (34.4)
For two events consisting in the emission of a light signal at one
point and its arrival at another point, the interval is zero:
~s2 = c2 M2 - ~F =0 (34.5)
lsee formula (34.1)). Owing to the constancy of the speed of light,
Eq. (34.1) must hold in any inertial reference frame. Consequently,
if an interval equals zero in the frame K, it will be zero in any other
frame K'.
Hence, an interval must become equal to zero in all reference
frames simultaneously. It thus follows that the interval ~s between
.events expressed in the frame K must be related to the interval ~s'
between the same events expressed in the frame K' by the equation
~s = a~s' (34.6)
But owing to the complete equivalence of the frames K and K', we
can write on the same grounds that
~s' = a~s (34. 7)
where a has the same value as in formula (34.6).
Multiplying relations (34.6) and (34. 7), we find that
a2 = 1
whence a = ± 1. It is natural to assume that the sign of the interval
in all the reference frames must be the same. Therefore the value of a
equal to -1 must be discarded. We thus arrive at the conclusion that
the interval between two events is an invariant:
~s = ~s' (34.8)
The result we have obtained indicates the expediency of the defmi-
tion of the interval between two points of four-dimensional space
that we have adopted. The interval determined by formula (34.3)
is invariant with respect to transformations of the coordinates and
time from one reference frame to another, i.e. behaves like the
distance (34.2) between two points in conventional space.
We must underline the fact that the conclusion on the invariance
of quantity (34.3) is a logical corollary of Einstein's postulates.
Proceeding from the invariance of the interval, we can write that
(34.9)
Assume that ~s 2 > 0, i.e. that the interval is real. We can therefore
find such a reference frame K' in which ~l' will equal zero. In this
frame, events separated by the interval ~s will occur at one point.
THE SPECIAL THEORY OF RELATIVITY 129
The time interval between events in the frame K' is

ilt' =~
c (34.10)
Real intervals are called time-like.

Now assume that !::.s2 < 0, i.e. that the interval is imaginary.
We can therefore find such a reference frame K' in which !::.t' = 0,
i.e. events occur simultaneously. The distance between the points
at which the events occurred in the frame K' is
t::.l' = i ils (34.11)
Imaginary intervals are called space-like.

Events occurring with the same particle can be separated only by
a time-like interval. Indeed, since a particle cannot move at a veloc-
ity greater than c, the distance ill which it travels during the time
M cannot exceed c M, i.e. ill ~ c M. Hence, ils2 ~ 0.
A space-like interval can separate only events having no causal
relation. Indeed, if ils2 < 0, then ill> c !::.t. Consequently, no
action emerging from one point of space can reach another point
during the time M and affect an event occurring at this point.
Consider a particle moving uniformly at the velocity v relative
to a frame K (a laboratory frame). Assume that two events occur
with this particle and are separated by a time interval equal to dt
in the frame K. Let us introduce a frame K' relative to which the
particle is at rest. In this frame, the time interval between the events
being considered is
dt' =.!!!
c
[see (34.1 0) ].
It is a simple matter to see that the time interval dt' is measured
by a clock moving together with the particle relative to K. The time
measured by means of a clock moving together with a body is called
the proper time of the body. Denoting the proper time by the sym-
bol -r, we can write
ds
d-r=-
c
(34.12)
Since ds is an invariant, and c is a constant, the proper time d-r is

an invariant.
Let us find the relation between the proper time d-r and the time
dt measured by means of a clock belonging to the frame I< relative
to which the particle and the (proper) dock associated with it move
at the velocity v. For this purpose"' we shall introduce into (34.12)
the expression for ds in terms of tM coordinates and the time in the
9-018
130 MECHANICS
frame K:
d-e = !!!_ = V c2 dt2- dl2
c c
[see formula (34.4)1. Let us transform the expression obtained as

follows:
d-c=dt v 1
(dl/dt) 2
But dl/dt is the velocity v of the particle. Hence,

d-e . dt V1 ~-
v2
c2 (34.13)
We conclude from (34.13) that the proper time of the particle is
always less than the relevant time interval in the stationary (labora-
tory) frame.
We have obtained formula (34.13) for the uniform motion of a
particle. It is also valid for non-uniform motion. We can therefore
write the following equation for finite time intervals:
rv
t2
£1-r= 1- :: dt (34.14)
ft
where v = v (t) is the velocity of the body for which the proper time
is being calculated.
35. Lorentz Transformations

We established in the preceding section that the interval M be-
tween two points in four-dimensional space is an invariant, i.e.
behaves like the magnitude of a vector in Euclidean space. This
gives us the grounds to consider L\s as the magnitude ("length") of
a four-dimensional vector (a four-vector) conducted from one world
point to another.
If we introduce the notation
. x0 = ct, x 1 = x, x 2 = y, x 3 = z (35.1)
the square of the interval becomes
£1s2 = (£\x0)2- (£1x1)2- (£1x2)2- (£1x3)2
The following relation holds for the distance L\l between two points
in Euclidean space:
£1Z 2 = I r 2 - r 1 12 = L\x~ + L1x! + L\x~
i.e. L\l equals the magnitude of the difference of the points' position
vectors. Similarly, the interval L\s can be represented as the magni-
tude of the difference of four-position vectors of the relevant world
THE SPECIAL ,:'HEORY OF RELATIVITY 131
points. Consequently, the coordinates x0 , xl, x2 , x 3 are the compo.,..

nents of the four-position vector of a world point. The square of the
magnitude of this position vector is
(xo)2 _ (xl)2 _ (x2)2 _ (xs)2
A comparison of the last expression with formula (XII.5) shows that

the space in which an event is depicted by a world point with the
coordinates (35.1) has a pseudo-Euclidean metric determined by the
tensor (XII.4).
Consequently, the square of t:he four-position vector can be repre-
sented as
3
x 0x 0 + xtxt + x 2x 2+ x3x 3 = ~ x~'x11 (35.2)
J.I.=O
[see formula (XII.31)}.
The components of a four-position vector are transformed by the
formula
3
x'J.I.= LJ aJ.I.x"
v=o· v
(35.3)
Let us take two inertial reference frames as the coordinate systems

K and K' in a pseudo-Euclidean space. We shall direct the axes of
these frames in accordance with Fig. 33.1. Hence, as is established.
in Appendix XII, the matrix of the transformation coefficients is
as follows [see (XII.21)}:
[a") = [ at
a 0 at 00]
ao 0 0 (35.4)
v 0 0 1 0
0 0 0 1
while
a: - a; = 1 (35.5)
[see formula (XII.22)).
The coefficients a 0 and a 1 can depend only on the relative velocity
v 0 of the frames. To find the form of this relation, let us write formu-
la (35.3) for x' 1 • With a view to (35.4), we obtain
x' 1 = a 1x0 + a 0xl
We replace x'1 with x', x0 with ct, and x1 with x [see (35.1)):
x' = a 1ct + aoX (35.6)
Let us write the expression_ obtained for point 0'- the origin of
coordinates of the frame K' ,(see Fig. 33.1). For this point, x' = 0,
9*
132 MECHANICS
and x = v 0 t. Substitution into (35.6) yields

0 = a 1ct a 0 v0 t +
whence
a. 1 = - a. 0 ~0 (37 .5)
Introducing this value of a 1 into relation (35.5), we find that
or
(/.,~ (1 - =~ ) = 1
(35.8)
We obtain from (35.7) that
-v le
0
(1.1 = ---:7'"7===:::r=;;
Vi-vile'
(35.9)
Hence, the matrix (35.4) in the case we are interested in will be
1 -~ 0 0
vt-~' vt-~ 2
-~ 1 0 0 (35.10)
[a.~J = vt-~ 2 vt-~'
0 0 1 0
0 0 0 1_
We have introduced the symbol
~ = ~
e
(35.11)
The matrix [a~] of the reverse transformation differs from (35.10)
only in that ~ in the numerator is preceded by a plus sign [see the
matrix (XII.23)l. ·
Substitution of the values of a.~ we have found into (35.3) leads
to formulas for the transformation of the components of a four-
position vector:
zD-~zl -~xD+xl
x'2=x2 x'3=x3
x'O=·-
v1- ~~ x't=
, -v 1- ~2
(35.12)
' ,
. The formulas for the inverse transformation are

x'O+~x'l ~x'"+x'l
xO = x2 = x' 2, x3 = x' 3 (35.13)
x1 =
v1-~· '
v1-~ 2 '
Going over in formulas (35.12) and (35.13) to the conventional
symbols t, x, y, z, we obtain
t( = tv(Vole2) X X'= .v-Vot ' y' = y, z' = z (35.14)
1-~2 .. 1-~2
t = t'+(v0le 2) x' X= z' +vot' Y = y', z = z'

V"t-~~ v"t-~'· .' (35.15)
Formulas (35.14) and (35.15) are known as the Lorentz transforma-

tions. We invite our reader to convince himself that these transfor-
mations leave the interval s12 between two events invariant.
The formulas for the inverse transformation (35.15) differ from those
for the direct transformation (35.14) only in the sign of v 0 • This
should be expected with a view to the total equivalence of both
reference frames, and also to the circumstance that for the given K
and K' the projections of the velocity of relative motion onto the
axes x and x' differ in their signs. Indeed, if the velocity of the
frame K' relative to the frame K is directed to the right (and its
projection onto the x-axis is positive), the velocity of the frame K
relative to K' is directed to the left (and its projection onto the
x'-axis is negative).
At velocities v0 so small that the ratio v 0/c may be disregarded in
comparison with unity, it is easy to see that the Lorentz transforma-
tions change into the Galilean ones.
The Lorentz transformations allow us to obtain formulas for the
transformation of the lengths and time intervals in passing from
one inertial reference frame to another (this is done in any general
course of physics). We shall limit ourselves to recalling the formula
for the Lorentz contraction of a body's length (in the direction of
its motion):
(35.16)
Here l 0 is the proper length of the body, for instance a rod (i.e. the
length of the body in the reference frame in which it is at rest), and l
is the length of the body in the reference frame relative to which
it is moving at the velocity v.
Since the lateral. dimensions of a body do not change during its
motion, the volume of the body contracts in accordance with the
formula
(35.17)
where V 0 is the proper volume of the body, and Vis its volume in
the frame relative to which it is moving at the velocity v.
Let us find formulas for the transformation of the components of
a particle's velocity. From formulas (35.14), we have
dt'= dt-(u 0 /c2)dx, dx'= dx-u 0 dt d '-d dz'=dz
l"t-~2 Vi-~2 , y - y,
Consequently,
, dx' dx-u 0 dt dx!dt-uo Ux-Vo
V:c =Cit'= dt- (u0 /c2) dx = 1- (u 0/c 2) (dx/dt) = 1- uouxlc 3
134 MECHANICS
Therefore,
V~=1~v~:ic2 (35.18)
Similar calculations lead to formulas for transforming the two other
·.velocity components:
, VII Vf-~l , Vz Vi-~'
VII= 1-vov:rfc2 ' Vz = 1-vov:rfc2 (35.19)
The formulas for the inverse transformation are
v' Vi-~ 2
v - v~+vo
: - 1 + v v~/c 2
V = II
1+v 0 v~/c2
,
0 II
)
'
(35.20)
v; Vi-~2
Vz= 1+v 0 v~jc 2
. Assume that the velocity .v of a particle makes the angle e with

the x-axis, arid the angle 8' with the x'-axis (the axes x and x',
by parallel translation, can always be brought into a position in
which they will be· in the same plane as .v). Let us find the relation
petween the angles 8 and 8'. We arrange the axes y and y' in the
plane defined by the x-axis and the direction of the vector v. The
latter will therefore be in the plane xy, and we can write
Vx = V COS S, c V 11 = V sin 8
v~ =
v' cos 8', :;~ = v' sin 8'
where vis the magnitude of the velocity in the frame K; and v' is
its magnitude in the frame K'. · ·
With the aid of formulas (35: 18) and (35.19)~ we find that
· V1 V ' 1 1 - A2 · ' 8 , / - fA2
.
· : tan·e• = ~= · V: r · ,., ' = v sm r -,.,
Vz Vx- Vo v cos 8- Vo
This formula allows us to ftnd the angle 8' made by the vector .v'
with the x'-axis from the known values of v and e. Similarly, we
can find a formula allowing us to determine the angle e between
the vector v and the x-axis when we know v' arid 9'. Using formulas
(35.20), we find that ·
· v 11 v~ Vi-~ 2 : v• sinS' Vi-~2
tan9=-=
Vx
,
Vx +v
= , ,
v cos~8 +v0
(35.22)
0
36. Four-Dimensional Velocity and Acceleration
In Appendix XII, we defined a four-vector as a set of the quanti-

ties a0 , a 1 , a\ a 3 which in passing over from one system of coordi-
nates to another are transformed according to the same rules as the
components of a four-position vector. Consequently, the formulas for

transforming the quantities all are similar to formulas (35.12):
'O aO-~al '1 -~ao+al
a = V 1-~2 ' a = -Y~1===~2:-' a's =as (36.1)
The formulas for the inverse transformation differ from (36.1) in

the sign of ~:
a'O+~a'l ~a'O+a'l
ao = a 1= a 2= a' 2, as= a'S (36.2)
V 1- ~z ' V 1- ~ 2 '
Let us consider the four-vectors of velocity and acceleration. In

non-relativistic mechanics, both the space intervals dl and the time
intervals dt are assumed to be invariant. Therefore, the set of quanti-
ties obtained after dividing the components of the three-dimensional
vector dr by the invariant dt forms the three-dimensional vector v-
the velocity vector of a particle. Similarly, the set of quantities
obtained after dividing the components of the vector dv by the in-
variant dt is the acceleration vector w.
We have seen that actually neither dl nor dt is invariant. What
is invariant is the interval ds related to dl and dt by the expres-
sion ds 2 = c2 dt 2 - d[2. The invariance of the interval made it
possible to introduce a four-position vector with the components
x 0 , x1 , x 2 , x3 that is an analogue of a three-dimensional position
vector with the components x 1 , x 2 , Xs· Let us attempt to find the
four-dimensional analogues of the three-vectors v and w.
It is evident that the set of the four quantities dxll/dt does not
have the properties of a four-vector because dt is not an invariant
and 2] (dxll/dt) (dxu.ldt) does not retain its value in Lorentz transfor-
mations. But we know an invariant that is a "relative" of dt. It is
the proper time d-r; = ds/c [see (34.12)). Since d-r; is an invariant
(i.e. a scalar), the quantities
dxll dxf.l
uiJ.=-=c-
d't ds
(36.3)
have the properties of components of a four-vector. It is called the

four-dimensional velocity (four-velocity) of a particle 1 •
Similarly, a four-vector with the components
d 2xl1 dul1 du 14
w11 = -d-r2- = --
d't = c:z. -ds
- (36.4)
is called the four-dimensional acceleration of a particle.

1 Einstein defined the four-velocity as a vector with the components ut-4 =
= dxl-4/ds. It is obvious that the velocity defined in this way is a dimensionless
quantity similar to v/c.
136 MECHANICS
Taking into account the values of dxJ.L, and also the circumstance
that
d-r:=dt V 1-- 11 2
c2 (36.5)
[see (34.13)], it is a simple matter to obtain the following values for

the components of the four-velocity:
uo- c (36.6)
- V 1-v2jc2 '
which can be written as
(36.7)
[see formulas (XII.34) and (XII.35)1. Here v is the conventional

three-dimensional velocity of a particle, and vk are its projections
onto the axes x, y, z. Of importance for our further discussion is
the fact that when v ~ c, the spatial part of the four-velocity trans-
forms into the conventional velocity v.
We easily find from formulas (36.6) that
3
2J uJLu11 = c2 (36.8)
J.I=O
(if we determine uJ.L in the same way as Einstein did, 2J uJ.Lu'l = 1).
Differentiating formula (36.8) with respect to 't', we obtain
~ ~u; u 11 + ~ uJ.L ~u_; = ~ wJ.Lu + ~ u~'w11 = 0

11
J.L Jl J.L Jl
According to (XII.33), both sums are equivalent, so that
2J w~'uJ.L=O
J.L
(36.9)
It follows from Eq. (36.9) that the vectors of the four-dimensional

velocity and acceleration are mutually perpendicular.
37. Relativistic Dynamics

Newton's equations are invariant with respect to the Galilean
transformations, but are not invariant with respect to the Lorentz
ones. Consequently, to satisfy Einstein's principle of relativity,
Newton's second law must be replaced with a more general one.
Having in view that when ~ -+ 0 (i.e. when v 0 /c -+ 0), the Lorentz
transformations convert to the Galilean transformations, the relativ-
istic-invariant equations of motion at v ~ c must convert to the
Newtonian equations:
;t (mv 1) = F1 (i = 1, 2, 3) (37.1)
The following relations are a natural four-dimensional generaliza-
tion of these equations:
:r: (mull)= K 11 (~ = 0, 1, 2, 3) (37 .2)
where 't' is the proper time, m is an invariant quantity characterizing
the inert properties of a particle (the mass of the particle), u1-1 is
a component of the particle's four-velocity. and, finally, Kl• is
a four-vector known as the Minkowski force. The values of KI-L must
be determined so that when v « c, the spatial components of
Eqs. (37.2) transform into Eqs. (37.1) like the spatial components
of the four-velocity in this case transform into the conventional
velocity v.
Taking into consideration expressions (36.5) and (36.6) for 't'
and u11, we shall write Eqs. (37.2) as
1 d ( me ) _ KO
Y1-v2fc2 dt Y1-v2jc2 - '
1 .!!__( mvi )=K 1 (i=1, 2, 3)
Y 1-v2fc2 dt Y 1-v2fc2
Multiplying these equations by Yt - v /c 2 2, we get

.!!__ ( me )- KO Y 1- v2/c 2 (37 .3)
~dt Y1-v2fc2 -
- d (
; dt
mv·
Y 1-v2fc2
) ' =K 1 vr 1-v2 /c 2
1
(i = 1, 2, 3) (37.4)
If we determine the spatial components Ki of the Minkowski force

so that they are related to the components of the conventional three-
dimensional force F 1 by the expressions
F 1 =K,V1-v 2 /c 2 (i=1, 2, 3) (37.5)
Eqs. (37.4) become
.!!__(
dt
mv,
Y 1-v2fc2
)=F 1 (i=1, 2, 3) (37.6)
It can be seen that when v « c, Eq. (37.6), as is required, trans-

forms into Newton's equations (37.1).
To determine the time component K 0 of the Minkowski force, let
us multiply Eq. (37.2) by the four-velocity u11. The result is
138 MECHANICS
(we have taken into account that m is an invariant and, consequently,

it can be put outside the derivative sign). Summation of the obtain-
ed equations over f..l. yields
3 3
2J K
11-0
11 u 11 = m 2J w11u11 = 0
11-0
[see (36.9)]. Substitution for utL of the values (36.6) and for K 1 of
the values obtained from (37. 5) yields
3
KO c ~ Ft Vi
0
Vf-v2fc2- V 1-v2fcZ V 1-v2 fc 1
i=i
whence
3
1 Fv
KO= ~ F 1v 1 = (37. 7)
cV1-v 2 /c 2 c V 1-v2 /c 2
i=1
Now we can write all the components of the Minkowski force.

With a view to formulas (37.5) and (37. 7), we have
(i=1, 2, 3) (37.8)
Hence,
K 11 - ( Fv F ) (37.9)
- c V 1-v2Jc2 ' V 1-,-v2jc2
The scalar product of the three-vectors F and v gives the work done
by the force F on a particle in unit time. This work equals the rate
of change in the particle's energy, i.e. dE/dt. Consequently, expres-
sion (37. 7) for K 0 can be given the form .
(37.10)
where E is the energy of a particle .

. We have thus established that the relativistic-invariant 1 equation
of the dynamics of a particle has the form of (37.2), where utL is the
four-velocity with the components (36.6), and KtL is the four-force
(the Minkowski force) with the components (37.8). The spatial com-
ponents of Eq. (37.2) can be represented in the form of (37.4) or
(37.6). At the limit when v ~ c, these equations transform into
Newton's equations [see (37.1)]. ·
1 Since ul1 and KIL are four-vectors, the form of Eq. (37 .2) in Lorentz trans-
formations remains unchanged (m is an invariant by definition).
The time component of Eq. (37.2) [see (37.3)] after substituting

for K 0 its value from (37.10) becomes
d ( me ) 1 dE (37.11)
dt V1-v2fc2 =edt
We thus conclude that the relativistic expression for the energy of
a particle is
mc
E= V 1-
2
v /c2
+ const
2
(37 .12)
38. Momentum and Energy of a Particle

In classical mechanics, the momentum of a particle is defined as
a three-vector having the components
p~cl) =
mv 1 (i = 1, 2, 3) (38.1)
The four-dimensional analogue of this momentum is the four-vector
having the components
pJJ. = muJJ. (~-t = 0, 1, 2, 3) (38.2)
where uJJ. are the components of the four-velocity 1 • Introducing the
values (36.6) for uJJ., we obtain
Po= V 1-v2jc2
me ' Pi= V 1-v
mvi .. (i = 1. , 2, 3) (38.3)
2 /c 2
which can be written in the form

' ''
JJ. 'me
(··. mv ·) ( 38 4)
p. = V1-v2jc2 '· V 1-v2fc2 . •
It is easy to see that when.v « c, the formula for the spatial com-
ponents of the relativistic ID:omentum transforms into the Newtonian
formula (38.1). This gives us the grounds to adopt the following
formula as the relativistic expression for the conventional three-
dimensional momentum:
mv
(38.5)
Let us now turn to the time component of the four-momentum.

At the end of Sec. 37, we obtained formula (37.12) for the energy E
of a particle without dealing with the value of the integration con-
stant. Comparing expression (38.3) for p 0 with formula (37.12), it is
a simple matter to see that by assuming the constant to be zero, we
1 If the four-velocity was defined as u!J. = d:xll/ds (see the footnote on p. 135),
the four-momentum is defined as piJ. = mcuiJ..
140 MECHANICS
can obtain the relation

E
pO=- (38.6)
C
In this case, the expression for the four-momentum becomes

pll = (~ I p) (38. 7)
where p is the quantity determined by formula (38.5).
The energy and (conventional) momentum are thus the components
of a single four-vector-the four-momentum of a particle (this four-
vector is sometimes called the momentum-energy four-vector). This
circumstance allows us to use formulas (36.1) to find the rules for
the transformation of E and p in passing from one inertial reference
frame to another. Introducing the relevant values of p~' into (36.1),
we easily obtain
E-E'+v 0p; P~+vo(E'/c 2 ) ' ' (38 8)
- yt-~ 2 • Px= Yi-~ 2 ' Py=py, Pz=Pz ·
The inverse transformations differ in the sign of v 0 •
Let us find the square of the four-momentum. From Eq. (38. 7),
we get
3
~ E2
L.J pllpll = C2- p2
Jl=O
[see formula (XII.38)l. At the same time
3 3 3
Lj pl1p11 = ~ (mu~') (mull)= m2 Lj u~-'u"' = m2c2 (38.9)
Jl=O Jl=O Jl=O
[see (36.8)1. We thus arrive at the relation
E2
-2 -p2=m2cZ (38.10)
c
We must note that the square of the four-momentum, like that of
any four-vector, is an invariant.
Assuming the constant in formula (37.12) to be zero, we obtain
t~e following expression for the energy of a particle:
E= mc2 (38.11)
y 1-v /c 2 2
The quantity E determined by expression (38.11) is called the total

energy of a particle. It must be borne in mind that E does not in-
clude the potential energy of the particle in an external force fi.eld 1 •
1 It must be remembered that by (37.7) and (37.10) dE/dt = Fv. In Newto-
nian mechanics, on the other hand, the work of the resultant of all the forces
acting on a particle equals the increment of its kinetic energy T, and not of its
summary energy T +
U.
For a particle at rest (i.e. when v = 0), expression (38.11) becomes

E 0 = mc2 (38.12)
where E 0 stands for the value of E at v = 0. This value is known as
the rest energy of a particle. \Ve remind our reader that for brevity's
sake, we use the term particle to denote a point particle, i.e. a body
whoso dimensions we may disregard. Tho rest onorgy of such a body
consists of the rest energies of the particles which the body consists
o£1 , of the kinetic energies of these particles, and of the energy of
their interaction with one another. It thus follows that
where m is the mass of the body, and ma are the rest masses of the
particles forming the body. Hence, the mass of a body does not equal
the sum of the masses of its parts.
Einstein at one time spent a lot of effort to substantiate the cor-
rectness of the assumption that the constant in (37.12) equals zero or,
in other words, to substantiate the statement that the energy mc 2
is stored in the mass m [see (38.12)]. For this purpose, he considered
several specific phenomena and showed that in each of them the
change in a body's energy by 1'1E leads to a change in its mass by
!::.m = 1'1E/c2 • Matters are much simpler at present. To substantiate
the relation 1'1E = c2 !::.m, it is sufficient, for example, to consider
the process of the transformation of an electron and a positron at
rest into two gamma-quanta. The corresponding measurements
show that the total energy of these gamma-quanta exactly equals
the sum of the rest energies of the electron and the positron.
The difference between the total energy (38.11) and the rest energy
(38.12) gives the kinetic energy of a particle
(38.13)
At small v's, this formula transforms into the Newtonian expression

for the kinetic energy
(38.14)
Let us now consider the momentum of a particle. Expression (38.5)

which we have obtained can be interpreted to signify that the depen-
dence of the momentum of a particle on the velocity is actually more
complicated than is assumed in Newtonian mechanics. A different
interpretation is also possible-we may consider that the relativistic
momentum of a particle equals, as in Newtonian mechanics, the
product of the mass and the velocity, but the mass is not constant
iL We now give a different meaning to the term particle than we did up to now.
142 MECHANICS
and depends on the velocity according to the law

m
m (v)- V 1-v /c"
2
(38.15)
Hence, for the relativistic momentum, we can write the expression
p = m (v) v
that is similar to the Newtonian expression p = mv.
Assuming in formula (38.15) that v equals zero, we find that
m (0) = m. Consequently, the invariant quantity m can be consid-
ered as the value of the mass of a particle at rest. In general courses of
physics, this quantity is called the rest mass and is usually designat-
ed by m 0 • The mass m (v) determined by formula (38.15), on the
other hand, is called the relativistic mass and is designated by m.
It is customary practice in theoretical physics, however, to dealonly
with the invariant mass. There is therefore no need to use the sub-
script "0" on the symbol m and the word "rest" in the name of the
mass of a particle.
The equation of motion (37.6) established in Sec. 37 can be written
as follows:
..'!!_ _ .!!___ { mv ) _ F (38.16)
dt - dt V1-v2fc2 -
It is not difficult to see that

d { mv ) m dv mvv dv
(38.17)
dt V 1-v2jc2 =V 1-v'fc2 dt+ c' (1-v2/c2)312 dt
Assume that the velocity v and the force F acting on a particle are
collinear. Now the velocity changes only in magnit,ude, and
v (dv/dt) = v (dv/dt). After making this substitution in the second
term of formula (38.17), putting dv/dt outside the parentheses, and
performing simple transformations, we get
m dv = F (38.18)
(1-v2fc2)3/2 dt
Now assume that the force F is perpendicular to the velocity v.

Hence, the velocity changes only in direction, and dv/dt = 0. Conse-
quently, formula (38.16) becomes
m dv =F (38.19)
V 1-vZfcZ dt
A comparison of expressions (38.18) and (38.19) shows that the

coefficient of proportionality between the force and the acceleration
in these two cases is different (in Newtonian mechanics it has a_single
value equal to the mass of a particle).
In concluding, we shall note that the relation

E2 = p 2 c2 + m c'2 (38.20)
follows from formula (38.10). The energy expressed in terms of the
momentum is called the Hamiltonian (see Sec. 30.). Consequently,
the relativistic expression for the Hamiltonian of a particle is
H=cV p2+m2c2 (38.21)
if the particle is free, and
H =C V p2+m2c2+U (38.22)
if the particle is in an external force field [ U is the potential energy
of the particle in this field; see the paragraph following formula
(38.11)1.
39. Action for a Relativistic Particle

Let us find the expression of the action for a free (i.e. not experienc-
ing the action of any forces) particle. The integral expressing the
action must be invariant relative to the Lorentz transformations.
Consequently, it must be taken over a scalar, and the latter must
have the form of a differentialto the first power. The only scalar of
this kind that can be associated with a free particle is a quantity
proportional to the interval ds. Denoting the coefficient of propor-
tionality by a:, we get the following expression for the action:
i2 t, /--;c-
s= Ja: ds = Ja:c];
1 t,
1- ~: dt (39.1)
[we have used formula (34.4) for ds 2 and taken into account that
d[/dt equals the velocity v of the particle].
Comparing (39.1) with expression (7.1), we arrive at the conclu-
sion that the Lagrangian for a free relativistic particle must be
/ --v-2
L=acJ 1-C2 (39.2)
At the limit when v ~ c, this function must transform into the

Newtonian expression
L = 21 mv2 (39.3)
Let us expand function (39.2) in power( of vic. Ignoring the
terms of the higher orders, we_ obtain
L=a.c V 1--~
v2
c2 a:c--
2c
av 2
144 MECHANICS
We may discard the constant term ae (see Sec. 7). Consequently,

in the Newtonian approximation, L = -av 2/2e. A comparison with
(39.3) shows that a must be assumed equal to -me. We have thus
established the form of the Lagrangian for a free particle1 :
,/-v2
L= -mc 2 V 1- 7 (39.4)
Knowing the form of the Lagrangian, we can easily find the mo-
mentum and energy of a particle. Using formulas (9.5) and (5.1),
we get
aL mv
p=-= (39.5)
av "Vt-v2fc2
mv 2 2
mc
E=pv-L=
V1-v2fc2
+me 2 V1-v 2 /e 2 = V1-v2fc2 (39.6)
We have thus arrived at the same formulas for the momentum

and energy of a particle that were obtained in Sec. 38. It must be
borne in mind, however, that we obtained formulas (39.5) and (39.6)
for a free particle, whereas in Sec. 38 similar formulas were obtained
with the assumption that a particle experiences the action of forces.
Let us again revert to expression (39.1). With a view to the value
of a we have found, we can write the action as
2
8=-me jas (39.7)
1
The true trajectory of a particle is determined by the condition
6S = 0 (39.8)
For the variation of the action, we have the expression
2 2
6S = -me6 Jds= -me J6 (ds)
1 1
The interval is
ds = y~ dx!L dx11
According to (XII.40), the variation of the radicand can be written as

6 ~ dxll dx11 = 2 .LJ dxiJ.t) dxJJ.
11 ll
J For a particle in an extemal potential force field, we have

L= -mc 2 V 1-v2/c 2 -U
where U is the potential energy of the particle. We must note that in relativistic
mechanics, L does not equal T - U.
THE SPECIAL THEORY OF RELATIYITY 145
Consequent! y,
2 2 '\l dx 6 dxl1
6S = -me _f
1
61/ ~ dx'-' dx
l.l
1, = -me J1/
1 '\; dx
11
-'-;;"':-::::;====-
1' dx 11
JJ.
2 ~ dx 11 6dx'-'
= - me l -'I-'-'---:,..---- (39.9)
.1 ds
I
The derivative dx 11 /ds is u 1Je [see (36.3)1. In addition, 6 (dx11) -

= d (6xJJ.) [see (III.4)l. We therefore arrive at the expression
2
6S= -m J ~ uJJ.d(8xll)
1 l.l
Let us integrate this expression by parts:
2
6S= -m ~ u11 6xJJ.[ 2 +m
1
dT r ~ fJxJJ. d;'
(39.10)
J1
11
JJ. JJ.
[we have represented du 11 in the form (du/dT) dT, where 't" is the
proper time of a particle). At the ends of the trajectory, 6x'-' = 0.
Therefore, condition (39.8) becomes
6S=m j
r ~ 6xJJ.Citd-c=0
L.J
dull
(39.11)
JJ.1
For this condition to be observed with arbitrary values of 6xll,

it is essential that the quantities duJJ.!dT: vanish, i.e. the four-velocity
of a particle be constant, which obviously holds for a free particle.
Let us find the action as a function of the coordinates of a particle
(i.e. as a function of the upper limit of integration; see Sec. 32).
For real motion, du 11 /d, = 0, hence the second term in (39.10)
vanishes. We consider that the lower limit of integration is fixed,
therefore (6xJJ.) 1 = 0. Consequently, the action as a function of the
coordinates of a particle satisfies the relation
3
6S= -m ~ uJ.L6xJJ.
J.L=O
(we have omitted the subscript "2" on Oxll}. The quantities mull
give covariant components of the four-momentum of a particle
[see (38.2)1 so that the increment of the action can be written as
3
6S = - 2J pll6xJJ. (39.12)
JJ.=O
146 MECHANICS
In Sec. 32, we obtained the following expression for the increment

of the action due to a change in the finite position of a particle in
conventional (three-dimensional) space:
3
6S = _L; Pi6x 1 (39.13)
i=i
lsee formula (32.5)1. It is a simple matter to note that the terms of

formula (39.12) corresponding to the spatial coordinates yield the
sum (39.13) after the index on 6x~ has been lowered.
Examination of (39.12) shows that the covariant components of
the four-momentum can be determined as follows:
as
P~1=- ax~ (39.14)
[compare with (32.6)1.
It was established in Sec. 38 that ~ pllp~ = m 2c2. Lowering the
index of the first multiplier, i.e. replacing the contravariant mul-
tipliers with the ,corresponding covariant ones, we find that
p~ - p~ - p: - p~ = m2c2
Substituting for p~ 1 in this equation their values from (39.14), we
arrive at the Hamilton-Jacobi relativistic equation
_1
c2
(.E_)2
at
_ (E-.)2
ax
_ (E--)2
ay
_ (E-.)2
az
= mzcz (39.15)
(we have introduced ct instead of x 0 , x instead of x 1 , etc.).

The action S in (39.15) differs from the non-relativistic action S'.
This can readily be understood if we have in mind that the action is
related to the energy by the expression E = -(8S/8t) [see formu-
la (32.10)1. The non~relativistic energy E', on the other hand, differs
from the relativistic energy E in the term mc 2 (i.e. E = E' + mc 2 ).
HenCiJ
as as·
--= at--+mc 2 at
'
or
S = S'- mc 2 t
Using this relation in (39.15), 'we get an equation for S':
·(!£)2
_f_·
2mc'
_!£_ _ _
at ' . IJt
1 [(!£)2- (~)2
2m ax
(!.£.)21= 0
t- ay + az J .
which at the limit when c .- oo transforms into the classical Hamil-
ton-Jacobi equation for a free particJ.e [see formula (32.15) in which
we must a~~ume that U = 0].
40. Energy-Momentum Tensor

In this section, we shall carry out n very important generalization
of the expression for the action. In such a generalized form, the
expression for the action can be applied not only to purely mechani-
cal systems, but also to an electromagnetic field and other physical
systems.
Up to now, we have written the expression for the action
2
S= ~ L (qk, qk, t) dt (40.1)

1
where Lis the Lagrangian, qk = qk (t) are the generalized coordinates

determining the position of the particles of the system, and qllr.
are the generalized velocities equal to dq,../dt. The quantities q11
and qk are assumed to depend only on the time.
When we write equations in the four-dimensional form, we have
to .. do with the four formally equivalent variables x 0 , x 1 , x 2 , x 3
that must be present in equations in a similar way. To reflect this
circumstance, we shall write the expression for the action as
S
1 r .
= c J !-* (qa, qav• x 0 , x 1 , x 2 , x 3) dx 0 dx 1 dx 2 dx3 (40.2)
where by qa we understand the set of the quantities q1 , q2 , • • •
determining the state of a system (the parameters of the system) ..
There may be any number of these parameters, particularly, au
infinitely large one. By qav
is understood the set of the partial deriv-
atives of the parameters q" with respect to the coordinates xv:
• 8qa
qav = axv (a==i, 2, ... ; v=O, 1, 2, 3) (40.3)
The factor 1/c has been introduced for convenience.

The quantities qa and qav are considered as functions of the coordi·
nates x 0 , x 1 , x 2 , x 3 • Particularly, the parameters qa may be found
to depend only on x 0 • Here, we arrive at the case \Ve already know
when qa = qa(t). .
We must note that since o2qalaxvoxll .,.,-- o2qaloxtJ. axv, the follow-
ing relation holds:
. .
8qav = 8qatJ.
(40.4)
8xll axv
To establish the conformity between expressions (40.2) and (40.1),
let us take into account that an element of the four-volume dV*
is related to an element of the volume dV in conventional space and
148 MECHANICS
the time interval dt by the following expression:

dV* = dx 0 dx 1 dx 2 dx 3 = c dt dV (40.5)
Introducing this value of dV* into (40.2), we get the expression
2
8= JL*dtdV= Jdt JL*dV 1
(40.6)
Integration with respect to dt is performed within the given interval

of time, with respect to dV -over the entire three-dimensional vol-
ume.
A comparison of expressions (40.1) and (40.6) shows that
L = JL* dV (40.7)
Thus, the function L* is the "density" of the Lagrangian of the sys-
tem being considered.
For a closed mechanical system, the Lagrangian does not depend
explicitly on t (see Sec. 8). Similarly, the absence of an explicit
dependence of L* on the coordinates x0 , xi, x2, x3 should be a math-
ematical expression of the fact that the system is closed. Hence,
for a closed system, the action has the form
8= c1 Jr L* (qa, ·
qav) dV* (40.8)
Let us find the equations of motion for a closed system. For this
purpose, we shall calculate the variation of the action (40.8) and
equate it to zero.
The variation of expression (40.8) is
B8=-;
1
Jf ( ~ oL* oL* " )
oqa 6qa+ ~ -,-Bqav dV*
a a, v oqav
By an.alogy with the relation By' = (By)', we have

• - oqa
,·
Bqa'Y- {j - - v
OX
= vOX0 Bqa
After performing this replacement, we obtain

1
88 = -
c
J(~ -,-
oL* Bqa + ~ -.-
uqa
aL* ·a
(}q ax
--v Bqa ) dV* (40.9)
a a, v av
Xc-cording to the rules for the differentiation of a product, we have

~ o ( oL* .) "'V oL* a "'V i} oL*
4J oxv. - .- Bqa
a. v : oqav.
= Ll - .- axv Bqa
a. v oqav
+·a,L:
'Y
Bqa oxv -.-·-
oqav
The first sum on the right-hand side is identical to the second term
of the integrand in formula (40.9). We can therefore write
The second of the sums
is a four-divergence of a vector whose v-th component is

~ 8L*
L._, - . - 6qa
a 8qav
Therefore, using the four-dimensional analogue of the Ostrogradsky-
Gauss theorem [see (XII. 72]), the second of the three terms in formu-
la (40.10) can be replaced with an integral over the closed hypersur-
face confining the four-volume over which integration is being per-
formed in (40.10):
(40.11)
On the boundary of the four-volume being considered, however,

the variations 6qa = 0. (Similarly in mechanics, the variations 6q,
at boundary points are zero.) Consequently, integral (40.11) vanishes,
so that only the first and third sums remain in formula (40.10).
Let us combine them, factoring out the common factor 6qa:
68 = _1
c
l )'
J ~
rL 8L"
aqa
- ~ _!__ 8L*
Li azv •
J6qa dV*
a v 8qav
Owing to the arbitrary nature of the variations 8qa, the expres-

sion we have obtained can equal zero only if all the expressions in
brackets vanish. We thus arrive at the following equations of mo-
tion:
aL* "" a aL* (40.12)
-a-= Li - - - (a= 1, 2, ... )
qa azv aq"
v av
These equations are a generalization of Lagrange's equations
aL a aL
(40.13)
aq;=di aq;
i50 MECHANICS
[see formula (7.3)1. The right-hand side of (40.12) is the sum of four
derivatives because the role played by the single variable t is now
played by four variables xv. It is not difficult to see that if qa 's depend
only on x 0 , i.e. on t, Eq. (40.12) transforms into Eq. (40.13).
Let us multiply Eq. (40.12) by gal! and summate over a:
aP • • a aL*
2j aqa qall = ~ qall a:xv - . -
a a,V aqay
The right-hand side can be transformed by the formula
av a (uv) au
·u~=---v--
azv axv axv
The result is
(40.14)
Let us perform the substitutions ga~-1 = oqaloxJJ. in the sum on the

left-hand side and oqaJJ.Iaxv = oqavfoxJJ. [see (40.4)1 in the second sum
on the right-hand side, and also group the terms of expression (40.14)
in a different way. The result is the relation
"" aL* aqa + "" aL* aqav = "'V _a_ ( • aL* )
LJ aqa azll LJ • ax11 .L! axv qaJJ. •
a a, v aqav a, v aqav
The left-hand side of the expression obtained is fJL*IoxJJ.. Conse-
quently, we have arrived at the formula
aL* _ ..;;• _a_ ( • aL• )
lJ Jl - ~ a v qaj). •
X a, V X aqav
The left-hand side of this formula can be written as 1
aL* = "'V &v aL* = )' _a_ (&v L*)
axJJ. LJ 11 axv ~ axv Jl
v v
which yields the relation
LJ a v (&v11L*)
"" _a = LJ
"" _a "1 ·
a v ( LJ qall aL•
• )
V X V X a oqav
The latter can be transformed as follows:
~ a:v [~a qaJJ. aqav
a.L* - &~L* J= 0 (40.15)
v
Equation (40.15) is a collection of four equations corresponding

to different values of the index f.L (f.L = 0, 1, 2, 3). The expressions
1 It must be remembered that arp/axY is a covariant component of a four-veotor.
in brackets have the properties of mixed components of a four-tensor

of rank two. Denoting this tensor by the symbol T~ v, we obtain:
(40.16)
Using this symbol, we can write Eq. (40.15) in the form

"'0
£..--0
ar IJ.v _
(Jl=O, 1. 2, 3) (40.17)
axv
v
We remind our reader that we have obtained Eq. (40.15) by equating

the variation of the action (40.8) to zero.
The tensor TJ.I. v satisfying Eq. (40.17) is determined non-uniquely.
Any tensor of the form
8Qvp
TJ.I.v =TJ.I.v+ "'0_J.I. (40.18)
£.. i)xP
p
where Q'(! is a tensor antisymmetrical with respect to the in-

dices v and p (Q~P = -Q~v) also satisfies Eq. (40.17). Indeed,
owing to the antisymmetry of the tensor Q~P,· we have
()2Q~P ()2Q~v
-
and, consequently,
The following condition will therefore follow from (40.17):
(40.19)
By properly choosing the tensor Q~P, we can always make the

·tensor (40.18) be symmetric. We shall assume in the following that
this condition is observed and
TWtl = TVIJ. (40.20)
We shall note that since Q~v = 0,
TJ.I. IJ. =Til
- IJ.
(40.21)
i.e., TIJ.IJ. is determined by formula (40.16).
152 MECHANICS
·: The transition of the index 1-L upward in (40.19) will either leave
all the T ll"'' s unchanged (if !l = 0), or will reverse the sign of all the
TJL"''s (if f.t = 1, 2, 3). Consequently, it follows from (40.19) that
fJTIL"'
""\.1
L-1
-ax=
'V
0 (!J.=O, 1, 2, 3) (40.22)
v
The tensor TJLv is known as the energy-momentum tensor of a
system. The grounds for this name will come to light below.
It is shown in Appendix XII that when the tensor TJLv satisfies
the condition (40.22) and all the TJL"~'s vanish at infinity, the vector
having the components
pJJ. =a J~ Tp.v d/v (40.23)
v
[see formula (XII.86)l, remains constant in time (is conserved). In
Eq. (40.23), a is an arbitrary constant, and dfv is a component of
the four-vector of an element of the hypersurface. The integral
in Eq. (40.23) is taken over an arbitrary hypersurface including
the entire three-dimensional space. If we take the hyperplane x 0 =
= canst as the hypersurface over which integration is being per-
formed, all the dfv's except df 0 = dV vanish, and expression (40.23)
will be simplified as follows:
pll=a JTJJ. 0 df 0 =a JTJL 0 dV (40.24)
We know that for a closed system the total (i.e. taken over the
entire volume) energy and the total (i.e. taken over the entire vol-
ume) momentum are conserved. The energy and the momentum are
the components of the four-momentum. Consequently, the four-
momentum of a closed system must also be conserved. This gives us
the grounds to identify the vector determined by formulas (40.23)
and (40.24) with the four-momentum of a system. The constant a
must be chosen so that formulas (40.23) and (40.24) agree with the
previons definition of the four-momentum [see (38.7)], according
to which, for example, p 0 = Elc. Assuming in formula (40.24) that
1-L· = 0, we get
Po= a Jroo dV = ~ (40.25)

Using formulas (40.21) and (40.16), we can write
- • fJL* • fJL*
roo= Too= Too=~ qau-. --6~L*= ~ qa -.--L*
a fJqao a oqa
where qa = iJq /iJt [we have taken into
4 consideration that q.ao =
= iJqafiJx 0 = oqufiJ (ct) = (1/c) (oqafiJt) = q
(1/c) 4 ).
The integral J L* dV gives the Lagrangian L. We therefore arrive

at the relation
i 00
Jr dV= ~ qa. -.--L
fJL
a flqa
According to formula (5.1), the expression on the right-hand side

determines the energy E of a system. Hence
J roo dV = E (40.26)
and roo is the energy density w:
roo= w (40.27)
Substitution of the value (40.26) into formula (40.25) leads us to
the conclusion that a = 1/c. Consequently,
pll=+ J2: rllv dfv

v
(40.28)
if integration is performed over an arbitrary hypersurface including

all the three-dimensional space, and
pll = +J rll 0 df 0 = +J rll 0 dV (40.29)
if integration is performed over the hyperplane x 0 = const.
To reveal the meaning of the components rov, let us write expres-
sion (40.22) for F = 0:
3 3
"" fJTOv = aroo "" fJTOk =0
L.J
V=O
{)z v {)zO + k=1
L.J f}zk
or
3
1 fJT0° _ "" fJTOk
- cat - L.J ---;;x;;:-
k=1
Multiplication of the relation obtained by c and integration over
a certain volume V yield
3
- .!_ \ roo dV = \ "" c fJTOk dV (40.30)
ot J J L.J .{)zk
v v k=l .
The integral on the left-hand side equals the energy confined in the
volume V. The integrand on the right-hand side is the conventional
(three-dimensional) divergence of a certain vectorS having the com-
ponents
(40.31)
154 MECHANICS
Formula (40.30) can thus be written as
8E=
--8t
1VSdV
v
Applying the Ostrogradsky-Gauss theorem to the right-hand side,
we obtain
_.M_= f.. Sdf
8t ';(
I
where the integral is taken over the closed surface confining the
volume V.
The decrement of the energy in the volume V per unit time must
~qual the energy flux through the surface. Consequently, S is the
vector of the energy flux density. By (40.31),
rot_..!..
- c
s %'
ro3 = _!_s
c z (40.32)
Owing to symmetry of the tensor TJJ.v, the relation TM = T 0"'
is observed, so that
T 10 = .!_
c S :1:> r 20 = -c1 Sy, T 30 = _!_
c Sz (40.33)
Hence, the components roll. and TM to within the factor 1/c equal
the relevant components of the vector of the energy flux density.
The spatial components of the vector (40.28) are
pi!.=~) r~~.odV
We thus conclude that the vector g with the components ri!.Ofc
.determines the density of the momentum of a system:
1 1 1
g:~:=-T•o, gy=-T2o, gz=-T3o (40.34)
c c c
A comparison of (40.34) and (40.33) shows that

1
g"=-2-S"
c
(40.35)
·.Or in the vector form
(40.36)
We have thus arrived at the conclusion that there is a relation

between the energy flux and the momentum-the density of the
momentum equals the density of the energy flux divided by c2 •
To establish the meaning of the components Tilt, let us write

relation (40.22) for fl. = i:
Hence, having in view that x 0 = ct and fio = cg;, we find that

3
ag 1 ~ aTilt
-at = k-1 --a;k
lt=1
Let us integrate the expression obtained over a certain volume V.

Taking into account that J
g dV = p (here p is the momentum of
the part of the system confined in the volume V), we obtain
a
--p1= 1. ')13 aTilt
--dV
at . ~ axk
l' lt=1
We transform the right-hand side according to the Ostrogradsky-

Gauss theorem:
3
_.!.._ p 1 =tf, ~ T 1 kdf~t= _rha 1 df (40.37)
at J J
I k=1 I
where a 1 is a vect~r with the components

(40.38)
The left-hand side of formula (40.37) contains the rate of diminish-

ing of the i-th component of the momentum confined in the vol-
ume V. Consequently, the right-hand side has the meaning of the flux
of the component p 1 through the surface f confining V, and a; is
the density of the flux of p 1• The quantity (40.38) is the k-th compo-
nent of the flux density of the component Pt· Hence, the three-dimen-
sional tensor (40.38) determines the density of the momentum flux.
We must note that the flux density of a scalar quantity (for instance,
the energy) is a vector; the flux density of a vector quantity (for
instance, the momentum) is a tensor.
The momentum carried through unit area in unit time equals the
force acting on this area, i.e. the stress at the location of the area.
This is why the tensor a 1k is called the stress tensor.
We have thus established the meaning of all the components of
the tensor T11v. Combining (40.27), (40.32), (40.33), and (40.38),
156 MECHANICS
we obtain
(40.39)
The components of the tensor TJJ.v characterize the density of the

energy and the momentum, and also the densities of the fluxes of
these quantities, which explains the name of TJJ.V-the energy-
momentum tensor.
Part Two
ELECTRODYNAMICS
Chapter VIII
ELECTROSTATICS
41. Electrostatic Field in a Vacuum
In general, an electric and a magnetic field are closely associated

with each other, forming a single electromagnetic field. Stationary
(i.e. not varying with time) electric and magnetic fields, however,
may be treated separately. We shall begin with considering an elec-
trostatic field in a vacuum.
The basic (force) characteristic of an electric field is the field
strength E related to the force acting on a point charge e at a given
point in the field by the expression
F = eE (41.1)
This relation may be considered as the definition of the quantity E.
An electrostatic field is a potential one. This signifies that the
work done by the forces of this field on a charge over any closed
path is zero:
~ F dl = e ~ E dl = 0 (41.2)
A glance at formula (41.2) shows that the circulation of the electro-

static field strength vector over any contour (closed path) r is zero.
Using Stokes's theorem [sec (XI.23)l, we can write
~ E dl = J. [V'E] df =0 (41.3)
r t
where f is an arbitrary surface confined by the contour r, and df
is the vector of an elementary area taken on this surface.
Condition (41.3) must be satisfied for any arbitrarily chosen sur-
face f. This is possible only if the integrand function at each point is
zero. We thus arrive at the conclusion that the curl of an electro-
static field strength vector equals zero at every point of the field:
[VEl = 0 (41.4)
158 ELECTRODYNAMICS
The equality to zero of the strength curl is a feature of an electro-

static Held expressing its potential nature.
It is known from vector analysis that the curl of the gradient of
a scalar function always equals zero [see (XI.43)l. Therefore, the
strength of an electrostatic field can be represented as the gradient or
a scalar function <p:
E= -'V<p (41.5)
(the minus sign does not change matters, it has been taken from
physical considerations). The function <p is known as the potential
of an electrostatic field. It is obvious that c:p
has been determined to within an arbitrary
constant addend. The potential can there-
fore be measured from any point of a field
for which its value is taken equal to zero.
In electrodynamics, the potential is usually
assumed to be zero at infinity.
0 The potential of the field of a point charge
e equals, as is known,
<p (r) =.!.. (41.6)
r
where r is a position vector drawn from
Fig. 41.1. the point where e is to the point for which c:p
is being determined.
Taking the gradient of expression (41.6) and reversing its sign,
we find the field strength of a point charge:
E= -'\7 ( -e) e -r= -ee

r =r2
- r r2 r ' (41.7)
where er is the unit vector of the position vector r [see (XI.51)}.

Assume that a field is set up by a system of point charges c0 placed
at points determined by the position vectors r~ (Fig. 41.1). Hence,
according to the superposition principle, the potential of the field
set up by this system at the point determined by the position vec-
tor r is
<p (r) = ~ lr~r~l (41.8)
a
and the field strength is
(41.9)
If the charge setting up a field is distributed in space with the

density p = p (r'}, the potential and the strength of the field can be
ELECTROSTATICS 15~
calculated by formulas similar to (41.8) and (41.9):

I p(r')dV'
<p(r)= J \r-r'l (41.10)
v
p (r') (r-r') d'V'
E (r) = \
!r-r' \ 3
(41.11)
v
where dV' = dx' dy' dz' is a volume element at the point r' (here-
x', y', z' are components of the varying vector r').
The transition from the system of point charges ea to a charge
distributed i.n space with the density p (r) is accomplished with the
aid of the Dirac delta function (see Appendix XIII). This function
allows us to represent the system of point charges ea at points with
the position vectors r~ by means of the charge density:
p (r) = ~ e0 6 (r-r~) (41.12)

a
It is obvious that by introducing the function (41.12) into expres-

sions (41.10) and (41.11) and performing integration, we shall arrive
at formulas (41.8) and (41.9).
42. Poisson's Equation

The general course of physics acquaints us with Gauss'~ theorem.
which for a field in a vacuum can be worded as follows: the flux of the
vector E through a closed surface is proportional to the algehraic sum
of the charges confined within the surface. The proportionality con-
stant depends on the choice of the system of units. In the Gaussian
system usually employed in theoretical physics, it is 4rr. Hence,
r$Endf=4n~e
v
(42.1)
t
If the distribution of charges inside the surface f is characterized.
with the aid of the charge density p = p (r), Gauss's theorem can be
written in the form
~En df = 4n Jp dV (42.2)
t v
where V is the volume confmed by the surface f.
Applying the Ostrogradsky-Gauss theorem [se<1 (Xl.13)1 to the
left-hand side of formula (42.2), we arrive at the relation
) VE dV = 4n ) ~dV
v v
i60 ELECTRODYNAMICS
The latter expression must be observed for any arbitrarily selected

volume V. This is possible only if at every point of the field
VE = 4np (42.3)
Expressing the field strength in formula (42.3) through the potential
according to (41.5), we obtain
V (Vcp) = -4np
It is shown in Appendix XI [see (XI.38)1 that V (Vcp) = Llcp, where Ll
is the Laplacian (operator). We thus arrive at the relation
Llcp = -4np (42.4)
known as Poisson's equation. In the expanded form, this equation is
as follows:
azq>
axz + azq> o2q>
ayz + 022 =- 4np (42.5)
Poisson's equation allows us to find the potential at every point
of a field according to a given distribution of the charges in space
(according to the given function p). Knowing cp, we can determine E
by formula (41.5).
For points of a field at which p = 0, Eq. (42.4) is
Llcp = 0 (42.6)
Relation (42.6) is known as Laplace's equation.
In accordance with formula (41.10), the solution of Eq. (42.4) is
_ \ p (r') dV'
q>- J lr-r'l (42. 7)
where integration covers the entire region in which the charges set-
ting up the field are distributed. This statement can be proved strict-
ly mathematically by applying the Laplacian to the integral (42. 7).
We must note that expression (42. 7) also satisfies Poisson's equation
if an arbitrary constant (Ll const = 0) is added to it. Consequently,
generally speaking, for the solution of Poisson's equation to be fully
uniq1!e, the boundary conditions must be set. Solution (42. 7) (with
const = 0) is obtained when the potential at infinity is assumed to
be zero.
The solution of Poisson's equation can be shown to be the only
one with given boundary conditions. This proof, however, is beyond
the scope of the present text.
We shall show that the function cp (r) = 1/r satisfies Laplace's
equation [Eq. (42.6)1 at all points except r = 0. By (XI.51),
1 1r r
we have V -r = - - = ..,.....-.
r2 r r3
Applying the operator V once
more to this expression, we obtain
11 +
= V (V f) = - V ( ;3 ) = - /a (Vr) - rv ( ;a )
ELECTROSTATICS 161
The divergence of r is 3 [see (XI.49)]. Consequently
Ll_.!_= _.!__r (
r r3
_.!....!...)
r' r
= _2_+2..=0
r3 r3
Q.E.D.
We can find tlte values of ~ (1/r) for all points including r = 0.
To do this, we shall take advantage of the fact that for a point charge
e at the origin of coordinates p = e8 (r), and the potential cp =
= e/r. Introducing these values into Eq. (42.4) and cancelling e,
we arrive at the following mathematical relation:
1
Ll-=
r
-4n8(r) (42.8)
43. Expansion of a Field In Multlpoles

Let us consider a system of charges in a restricted volume and
investigate the field set up by this system at distances that are great
in comparison with the system's dimensions. We shall place the
origin of coordinates inside the volume occupied by the system.
This gives the following expression for the potential:
cp (r) = ~ I ea I (43.1)
· L.. r-ra
a
where according to our assumption r » ra [see (41.8); we have omit-
ted the prime on ra ].
\Ve expand expression (43.1) into a power series in r air. For this
purpose, we express cp (r) in terms of the components of the vectors r
and r 0 :
(43.2)
(the snbscript i takes on the Yah1es 1. 2, 3, and the subscript a,

the values 1, 2, ... , n, where n is llte number of charges in the
system). Considering the quantities ( -Xa;) as small increments of
the coordinates xb we can write the function (43.2) as
+ _21 ~ ~ 8 o; ( V ea
:X )I Xm "V :x~
) (-X all) (-X am)+ • • •
a k, m ~ t
= cro + cp1 + cp2 + . . . (43.3)

[each addend of the sum (43.2) is expanded into a series).
11-018
i62 ELECTRODYNAMICS
Expression (43.3) is known as the expansion of the potential in

multipoles. The first term of the expansion
")1, ea
fPo=-=;:-- (43.4)
(we have substituted r for v~ x1) has tho form of the potenlinl of
a point charge. The total charge~ ea is a zero-order multipole (it is
also called a monopole). When this multipole is non-zero, the term
'Po makes the main contribution to potential (43.2).
To establish the form of a first-order multipole, let us transform
the second term of the expansion (43.3) as follows:
<i't = ~~
a k
a:k ( era ) (-X all) = - ~
k
a!k ( f ) ~ eaXak
a
The sum ~ eaxa 11 is the projection onto the k-th coordinate axis
a
of the vector
(43.5)
This vector is the dipole moment of the system of charges 1• And it is

a first-order multipole.
The expression
a!k ( +)
gives the k-th component of the gradient of 1/r. Hence,
q>,=-~(v_!_)
k r pr. xk
·Pk=-p·V_!_
r
(43. 7)
The formula for qJ 1 can be obtained directly in the vector form,

taking advantage of the circumstance that by (XI.5) to within
first-order terms, we have
I (r + Br) = I (r) + VI (r) 8r
1 If a charge is distributed over the volume of a system having the density
p, the dipole moment is determined by the integral
p = J p (r) r dV (43.6)
It must be remembered that for a system whose total charge ~ea equals zero,
the dipole moment does not depend on the choice of the origin of coordinates.
Indeed, transferring the origin of coordinates to a point for which r = b, we
get the values r~ = ra - b for the position vectors from the new origin of coor-
dinates. The dipole moment in the new system is p' = ~ear~ = ~>ara -
- ~eab = p- b 2Jea. Since 2Jea = 0, we obtain p' = p.
ELECTROSTATICS 163
Let us apply this formula to each addend in (43.1), considering -ra

as or. The result is
cp (r) = ~
n
e: -~ear a
a
v+ (43.8)
The f1r-st ter-m roineidc~ with (43.4). and the ~ceond with (43.7).
We must note that V (1/r) is proportional to 1/r~. Consequently,
the addends of the second sum in (43.8) are quantities of the order
of ralr relative to the addends of the first sum.
+t a -e
y-------,
I I
I I
al 1a
I I
I
6-------~
I
-e a +t
Fig. 43.1.
After calculating V (1/r) and introducing it into (43. 7), we]arrive

at an expression for the potential of the field of a dipole:
IP1 = -pV _.!_=

r
-p ( - -r1-2 !..._)
r
=~
<·r•
(43.9)
Now let us find the field strength of a dipole:
E= - Y'Cflt - V ( ~~ ) = - (pr) V-;\-- r13 V (pr)

[see (XI.25)]. Using formulas (XI.51) and (XI.37), we obtain
3 (pr) r 1 { ,
E=-r4 --,--,:r- [p, [Vr]J+(r, (Vp]]+(rV)P+(PV)r}
The curl of r equals zero [sC'C' (XT..'"lO)I, [Vp] ,., 0 hC'cnuse p does
not depend on r, and (rV) p = 0 for the same rea:-on. Hence. of the
four terms in the braces, only the last one is non-zero, and by (XJ.34)
it equals p. Therefore,
(43.10)
where er is the unit Yector of the position yector r.

We must note that the field of a dipole has axial symmetry relative
to the direction of p.
Not only the total charge ~ ea, but also the dipole moment p =
= ~ eara may equal zero. This occurs, for example, for the system
of charges depicted in Fig. 43.1 and known as a quadrupole. Here,
11*
164 ELECTRODYNAMICS
the field is determined by the next term of the expansion of the func-
tion (43.1), quadratic relative to the quantities r 0 /r.
Let us write qJ 2 as follows [see (43.3)1:
q>2 = +~ ~ {Jxk~~m ( era ) XakXam
+~
a k. m
=
It, m
axka;xm; ( +) {~ o
eaXakXam} (43.11)
The quantity in braces is the (k, m)-th component of a symmetrical

tensor of rank two (see Appendix X). This tensor could be adopted
as the corresponding multipole. But as vre shall show below, of the
nine components of this tensor not six are independent (as in a sym-
metrical tensor), but only five. To underline this. the tensor char-
acterizing the property of the system th3t determines q> 2 is written
in a different form. \Ve saw in Sec. 42 that the function 1/r satisfies
Laplace's equation. i.e. ·
~2
L1 r1 = "0
L..J
k
iJxi r1 =0
(according to our assumption r» r 0 , so that r = 0 is not considered).
It can easily be seen that this formula can be written as
k,m
Multiplying the expression we have obtained by ear~/6 and then
summating over a, we obtain
_61 "0 ear~ "0 6km 1
iJXJtOXm -=0
{J2
LJ LJ r
a k,m
which can be written as follows:
+~ k,m
OXJt~:m +{~ ! ear~6km} =
a
0
Subtracting this expression from (43.11). we give the formula

for q> 2 the form
IJ>z=; ~ ax:;xm +{~ea(xakXam-; r~<\m)} (43.12)

It, m a
The set of quantities
Qkm = ~ ea (3xakXam- r~6km) (43.13)
a
is called the tensor of the quadrupole moment of a system. Let us

evaluate the trace of this tensor, i.e. the sum of its diagonal compo-
ELECTROSTATICS 165
nents:
T r (Q 11m) = ,-,
L.J
Qk k-- ~ ~
.:..J ..:....; ea (3 Xak-
2 r2) ~ ~ (3 2
a ="""'" ea-
2)
Xak- r a
k k a a k
,...., ea [3'1
= .:..J - 32} 0
Xak- r a =
2
(43.14)
a k
The equality to zero of Tr (Qhm) signifies that of the three diagonal

components of the tensor Qkm only two are independent, and, con-
sequently, there are altogether five independent components.
If we transform the tensor Q11 m to the principal axes, owing to
condition (43.14) only two of the three principal values are independ-
ent. If a system of charges has an axis of symmetry of an order
higher than two, this axis (we shall designate it by the letter z)
is one of the principal axes of the tensor Q11 m; the position of the
other two principal axes is arbitrary. In this case, it is evident
that Qxx = QYY' and owing to (43.14), we have
(43.15)
The principal value Qzz is called simply the quadrupole moment of

the system in this case. It can be shown that when the total charge
and the dipole moment of a system equal zero, the quadrupole mo-
ment does not depend on the ·choice of the origin of coordinates.
Using the symbol (43.13), we can write the potential of a field
due to a quadrupole as follows:
f "'
<1'2 = 6 L.J Qkm OXJIOXm
iJZ
rf (43.16)
k,m
Let us calculate the second derivatives in this expression:
(axmlax 11 = 8hm)· Introducing this value of the derivatives into

(43.16), we have
(43.17)
Examination of (43.17) shows that the potential of a quadrupole

diminishes with the distance as 1/r3 • vVe remind our reader that
the potential of a monopole diminishes according to the law 1/r
[see (43.4)], and that of a dipole according to the law 1/r2 [see (43.9)].
In general, the pntential of an n-th order multipole diminishes with
the distance 11ccording to the law 1/rn+ 1
16£ ELECTRODYNAMICS
Hence, the field of a system of charges can be represented as the

superposition of the fields set up by multi poles of different orders:
~ ea i
{j).(r) = CJlo+ CJl1 + CJl2+ · ·· = - r - - PV'-,:-
+
+ ! Lj Qkm -:a:-x-k8...,.~x-m- + '''
vVe shall not deal with multipoles of higher orders. We shall only
note that a multipole of the third order is called an octupole. We
canindicate as an example of an octupole a system of eight unlike
chargesidentical in magnitude and arranged at the corners of a cube
so that the closest neighbours are charges of opposite signs .
.44. Field in Dielectrics

Up to now, we have been dealing with an electrostatic field produced
by a given system of charges in a vacuum. It was assumed that
the charges setting up the field can move over macroscopic distances
{for instance, within the confines of the entire conducting body).
We shall call such charges free.
Matter!' become much more complicated in a field produced by free
charges in dielectrics. Here the field set up by the charges in the
atoms and molecules of the dielectric is superposed onto the field
of the free charges. Since these charges cannot leave the confines of
the atoms and molecules they belong to, they are called bound.
If Erree is the field of the free charges, and Ebound is that of the
bound ones, the strength of the resultant field Eres can be written as
Eres = Erree + Ebound (44.1)
Even if the free charges are stationary, the field (44.1) is not sta-
tionary (i.e. time independent) because the bound charges are in
mot-ion inside the molecules and, in addition, participate together
with the molecules in thermal motion. It is thus evident that the
field Ebound is a fast-varying function of time. In addition, Ebouod
changes greatly in the space between two adjacent molecules. Both
kinds of dependence (on the time and on the point in the space be-
tween molecules) vanish if we deal with a value of Ebound that has
been averaged, first, over a time interval much longer than the pe-
riod of intramolecular motion and thermal oscillations, and. second,
over a volume considerably exceeding that of a molecule. Conse-
quently. the field (Ebounct) is stationary. In addition, it changes
smoothly within the limits of a volume including many molecules.
The field (Ebound) is called macroscopic, unlike the microscopic
field Ebound·
ELECTROSTATICS 167
We shall call the macroscopic quantity

E = Etree + {Ebound} (44.2)
the fteld strength in a dielectric.
In the absence of an external field (i.e. a field of free charges), the
field (Ebound) usually equals zero. Under the action of an external
field. the mean positions of bound charges are displaced the more,
the stronger is the field acting on them. As a result, the field
(Ebound) becomes other than zero. In calculating the field (44.2),
·matters become complicated by the fact that the average displace-
ment of the bound charges is determined not by the field Etree,
but by the resultant field E that includes (Ebound) in it.
It is customary practice to characterize the state of a dielectric
by the dipole moment of a unit of volume of the dielectric, which is
called the polarization and is designated by the symbol P. It is
evident that P can be determined as
~ p·
P=~ (44.3)
t.V
where ~Vis an infinitely small volume 1 , and Pi is the dipole moment
of an individual molecule, summation is performed over all the
molecules confined in the volume ~ V. · ·
Having determined Pin this way. we have in essence performed
the averaging mentioned above when discussing Ebound (Pis a macro-
scopic quantity, and Pi is a microscopic one).
vVhen an electric field acts on a dielectric, the bound charges
become displaced (each remaining within the confines of its "own"
molecule), the positive ones along the field. and the negative ones
in the opposite direction. The result is the formation of bound
charges on the surface of the dielectric. In addition. space bound
charges may also appear.
Let us find an expression for Pbound- the volume density of the
bound charges. We mentally separate in a dielectric an infinitely
small volume ~ V. Assume that in the absence of a field the bound
charges ea (where ex = 1, 2, ... ) confined in this volume are at
points determined by the position vector·s rao· Since the diC'lectric
is not polarized in the absence of a field, the expression': ear-xo is
zero (the sum is taken over ~ V). Indeed, with an accuracy up to
1/ ~ V. this sum coincides with the polarization P, which vanishes
in an unpolarized dielectric.
1 By an infinitely small volume in physics is meant the volume il V co.n-

taining a sufficiently large number of molecules to allow us to ignore the dis-
creteness of the substance, and at the same time small enough for us to consider
that macroscopic quantities such as E or P are constant within the confines
of t.V.
168 ELECTRODYNAMICS
Assume that when a field is switched on, the bound charges become
displaced by the segments Llra. (we must note that these segments
are much smaller than the linear dimensions of the volume Ll V).
The result is the appearance of a polarization characterized by the
vector
P= 6.~ ~ea.(ra.o+Llra.)= 6.~ ~ea.Llra. (44.4)

a.
The bound charges in a dielectric can be divided into several
groups, in each of which the magnitude of the charge ea. and the dis-
placement Llra. are identical. Let us number such groups with the sub-
script~- Let nil stand for the number of charges of group ~ per unit
I
volume of the dielectric. Hence, the polariza-
I tion can be written as follows:
____Rid,
Ar~
p = 6.1V ~ ell .1rantl .1V = ~ ntletl .:lrll (44.5)
I( . 13 II
I
I D (nil .1 V is the number of charges of group ~
I contained in the volume Ll V).
~
\M~ \ I:
We calculate the algebraic sum of the bound
charges intersecting the boundary of the Yol-
ume .1 V when the field is switched on. Fig-
Fig. 44.1. ure 44.1 shows an element df of the surface
confining Ll V. The charges of the group bearing
the number ~ contained in an elementary volume of the size Llr 13 df
cross df and emerge from or enter the elementary -volume. They carry
along with them the charge
(44.6)
This expression is algebraic. Its sign depends on that of e11 and on
the sign of the scalar product LlrJl df, i.e. on the direction of Llr 13
relative to the outward normal to df (in Fig. 44.1 it is denoted by n).
Sumrnating expression (44.6) over ~. we obtain the total charg-e
crossing df: -
~ntletl Llr 11 df = Pdf (44. 7)
[see (44.5)].
Having integrated expression (44. 7) over the surface j, let us find
the total bound charge emerging from the volume .1 V when a field
is switched on. When this happens, the volume .1 V, which was pre-
viously neutral, acquires a bound charge:
.:lqbound = - 5Pdf
f
ELECTROSTATICS 169·
(the acquired charge equals the emerged one taken with the opposite·
sign). Using the Ostrogradsky-Gauss theorem, we obtain
.1qbound =- JV'P dV = - V'P

~v
.1V
Hence, the following expression is obtained for the density of bonnet

charges:
Pbound = - V'P (44.8)·
Formula (44.8) will also help us to find the surface density of the·
bound charges abound· For this purpose, let us consider a cylindrical
volume confined between two infinitely close sur-
faces of the magnitude .1/ at different sides of the
dielectric's surface. (Fig. 44.2). The bound charge
qbound confined in this volume can be represented
either in the form J

Pbound dV, or in the form
abound 11f. Using formula (44.8) and the Ostro-
gradsky-Gauss theorem, we can write
qbound = abgund .1f := - JV'P dV = - JP n' df
where n' is an outward normal to the surface of

the cylinder. The integral over the surface can be
divided into three parts: an integral oYer the outer
base of the cylinder that equals zero because P = 0
outside the dielectric; an integral over the side
surface of the cylinder that may be disregarded Fig. 44.2.
owing to this surface being infinitely small, and,
finally, an integral over the inner base of the cylinder. The latter·
equals -Pn' t:.j, where n' is an inward normal to the surface of
the dielectric. If instead of n' we take the outward normal n, we
must substitute +Pn for -Pn'· We have thus arrived at the
ex pression abound 6./ = P n tlf, whence
(44.9}
Here P n is the projection of the polarization Ponto an outward nor-

mal to the surface of a dielectric.
We must note that charges of the density determined by formu-
las (44.8) and (44.9) are not imaginary, but quite real charges.
Hence. in the presence of dielectrics, the field of bound surface
[see (44.9)] and volume [see (44.8)] charges is superposed onto the-
field set up by the free charges (let their density be p). Consequently •.
170 ELECTRODYNAMICS
the potential at a point determined by the position vector r is

_ \ p (r') dV' \ Pn (r') df' \ V'P (r') dV'
cp (r)- j R -j- j R - j R (44.10)
vP td · vd
where for brevity we have used the notation R = I r - r' I; dV' is

an elementary volume taken in the vicinity of the point r', and df'
is an element of the surface of the dielectric taken in the vicinity of
the point r'; the divergence of P (r') is taken over the primed coor-
dinates, therefore the operator V is primed [see (XI.52)1. The first
integrnl is taken over the volume where pis non-zero, the second in-
tegral is taken over all the surfaces confining the dielectric, and,
'finally, the third integral is taken over the entire volume of the
·dielectric .
.45. Description of the Field in Dielectrics

If VP differs from zero, every elementary volume of a dielectric
is equivalent to a point charge of the magnitude -VP dV and makes
the corresponding contribution to the macroscopic field E. There-
fore, when dielectrics are present, Eq. (42.3) must be written as
VE = 4n(p rree + Pbound) = 4n(p - VP) (45.1)
(by p is meant the density of the free charges). The equation in this
'form is hardly suitable for finding E because it determines the latter
not only in terms of the density of the free charges, but also through
the nature of polarization of the dielectric. The polarization, in
turn, is a fui;tction of E.
It is easy to note that if we introduce the auxiliary quantity
D = E 4nP + (45.2)
the following equation holds for it:
vn = 4np (45.3)
that is, vn is determined only by the density of the free charges.
· The quantity D is called the electric displacement (or the electric
induction).
A comparison of Eqs. (45.1) and (45.3) shows that the operation
of finding Dis much simpler than that of finding E. There would be
1ittle good from the quantity D, however, if it were not for the
·circumstance that in the majority of practically important cases D
is proportional to E. It is therefore possible to use a "roundabout
manoeuvre": instead of the main characteristic of a field E, first the
.auxiliary quantity D is calculated, and then the transition is per-
formed from D to E.
ELECTROSTATICS 171
The introduction of the quantity Dis also expedient because many

formulas written with the n~e of Dare much simpler than they would
be when expre~~erl in terms of E nnd P.
Experiments show that in mnny cases having a pmclical inll'rest,
the polarization is proportional to the field strength:
P = xE (45.4)
~we nre meanwhile limiting ourselves to a treatment of isotropic
dielectrics). The quantity X is known as the electric susceptibility
-of a substance. It is always positive. Introducing (45.4) into (45.2),
·.we find that
D = E __:__ 4.:-rxE = eE
where
e = 1 +
4rrx (45.6)
Js the relative permittivity or simply the permittivity of a substance
1
(also sometimes called the dielectric constant).

Hence, D and E are often proportional to each other. This is ex-
:actly why it is expedient to introduce D.
Proceeding from Eq. (45.3). it would seem pnssible fnr us to con-
·clude that D is determined only by the density of the free charges.
But this is not true. Equation (4.'>.3) alone is not sufficient for deter-
mining D. To understand this more clearly, let us recall how E is
found in the absence of a dielectric. In addition to the equation VE =
= 4rrp, we took advantage of the fact that [VEl = 0 and assumed
'that E = -V<p. Introducing this expression into the equation for
the divergence, we arrive at Poisson's equation: V 2 <p = -4rrp.
Solution of this equation allows us to find <p. and then E.
If we follow the same path in finding D. in addition to (45.3) we
must consider the equation
[VD] = f"V. eEl = [Ve, El +
e [VEl = [Ve, E)
lwe have used formula (XI.27) and taken into account that [VEl -
= OJ. This equation transforms into [VDI = 0 only when Vc = 0,
i.e. when the dielectric is homogeneons. In the general case, [VD]
<lepends on E. i.e. in the long rnn on the bonnd charges.
Although D, however, depends in general on the bound charges,
the collection of eqnntions
[VEl = 0, VD = 4rrp. D = eE
:allow5 u~ to calculate E and D according to the known distribution
-of the free charges in space.
Using relation (45.4). we can determine the conditions in which
Phound differs from zero. Let us introduce the value given by (45.4)
1 The absolute permittivity of a medium Ea = e 0 e is introduced in electri-

<:al engineering. It is deprived of a physical meaning, however, and we shall
not use it.
172 ELECTRODYNAMICS
into formula (44.8):

Pbound = -'VP = -V (XE) = -xVE- EVX (45. 7)
[see (XI.26)]. By (45.1), we have VE = 4n (Ptree +
Pbound)· Sub-
stituting this value for VE in formula (45. 7), we get
Pbound = -x4n(Ptree +
Pbound) - E VX
Now let us solve the equation obtained relative to Pbound:
1
Pbound = - -E (4nXPtree + EVX)
Inspection of the relation we have found shows that Pbounci is
non-zero, first, at the points where Ptree is non-zero, and, second, at
the points where VX =1= 0, i.e. at places where the dielectric is non-
homogeneous. We must note that space hound charges do not appear
in a homogeneous polarized dielectric (P = const).
Let us consider the field in a homogeneous dielectric. Assume that
in the absence of a dielectric with a given distribution of the free
charges, a field is produced that is characterized by the strength E 0
and the potential rp 0 • We know E 0 and rp 0 to he the solutions of the
equations
(45.8}
(45.9)
[see (42.3) and (42.4)1.
Now, without changing the arrangement of the free charges (i.e.
p), let us fill the entire space in which the field is non-zero with a
homogeneous ( e = const) isotropic dielectric. The field strength
will therefore become equal toE, and the potential will be rp. Let us
write equations for E and rp. According to (45.3)
VD = V(eE) = 4np (45.10)
Substituting -Vrp for E in this equation and taking into account
that e = const, we can write
-V (eVrp) = -V 2 (erp) = 4np
or
(4.5.11)
A comparison of (45.10) with (45.8) and (45.11) with (45.9) shows
that the equation for eE coincides with that for E 0 , and the equation
for erp with that for rp 0 • It follows that the filling of a space in which
a field is present with a homogeneous isotropic dielectric leads to
both the field strength an cl the potential becoming equal to 1/ e-th
of their initial values. Particularly, for the field of a point charge
placed in a homogeneous dielectric, we have
(45.12)
ELECTROSTATICS 173
It must be noted that by (45.11), Poisson's equation (42.4) for a

field in a homogeneous isotropic dielectric is as follows:
4n
LlrtJ= - £ P (45.13)
lVe shall indicate without a derh·ation (which can be found in
many textbooks of general physics) the conditions which E and D
must satisfy at the interface hetwrrn two diC'lC"ctr·ies:
En = E't2• E1En1 = E2En2 } (45.14)

Dnl = Dn2• D -nh1 = D n/e2
(the subscripts T and n denote the tangential and normal components

<>f the relevant vector, respectively).
It is a very difficult task to find the field by soh·ing Eqs. (42.4),
~45.3), etc. in the general case.
In cases involving symmetry. q'
it is possible to establish the
form of the field without solv-
ing any equations. \Ve shall
show this using the following
example. Assume that we have
the plane boundary of two se-
miinfinite homogeneous and·· _:£3,_2- - - A - - L - - - -1 ; r - -
isotropic dielectrics with the E1
permittivities e1 and e 2 • There
is a point charge q in the first
dielectric at the distance a
from the boundary. We are to
find the field in both dielec-
trics. .
We shall form the field in
the first dielectric from those Fig. 45.1.
<>f the point charge q and its
mirror image-the imaginary charge q'. This assumption satisfies
the main condition that the first dielectric contains only one source
of D-the chargC' q (q' is outsidC' the first die]C'etric). \\ie shall thus
seek the potential in the first dielectric in the form
q q'
cpt = er-
1
+ --:-;;-
"1
(45.15)
(what r and r' are is clear from Fig. 45.1).

We shall represent the field in the second dielectric as that of the
imaginary charge q" placed 'vhere q is. This assumption agrees with
the fact that there are no sources of D in the second dielectric (q" is
outside the second dielectric). \Ve shall thus seek the potential in
174 ELECTRODYNAMICS
the second dielectric in the form

q"
(j)z =-;-;- (45.16)
2
Let us try to choose the values of q' and q" so as to satisfy th&
boundary conditions (45.14) for D. According to our assumptions, D·
has the following values in the first and second dielectrics:
qr q'r' q"r
Dt = 7 +--;:;a , D2=
r3- (45.17}
At the boundary of the dielectrics, I r I = I r' I· Consequently,

q"r
D2b= -a-
r
(45.18)
Let us find the normal components of the vectors in (45.18). We-

direct the vector n from the first dielectric to the second one. Taking-
into account that the projection of r onto n equals a, and the pro-
jection of r' equals -a, we get
q-q' q"
Dtn=--a-a,
r
D2n=-3
r
a
Since the tangential components of the vectors r and r' are th~
same, i.e. r"' = r~, we can write
(45.19)
It follows from the equality D 1 n = D 211 that
q - q' = q" (45.20~
After introducing the values from (45.19) into the relation Dn/e 1 =
= D2-r/e 2 [see (45.14)1, we obtain
q + q' = ~
e2
q" (45.21}
The" simultaneous solution of Eqs. (4.5.20) and (45.21) yields th&
values of q' and q" satisfying the boundary conditions:
q
'
=q
81- 8 2
e1 +e 2 ' q
"
=q
28 2
e1 +e 2
(45 ' 22}
We have succeeded in "constructing" the functions D1 and D 2 [see·
(45.17)1, each of which satisfies the equation VD = 4np in its region.
In addition, these functions satisfy the boundary conditions. There-
fore, the functions (45.17) and, accordingly, (45.15) and (45.16)·
are the solutions of the problem [after introducing the values (45.22)
for q' and q" into theml. According to the uniqueness theorem (see·
Sec. 42), there are no other solutions.
·ELECTROSTATICS 17~
46. Field In Anisotropic Dielectrics

In anisotropic dielectrics, the directions of the vectors P and E~
generally speaking, do not coincide (see Appendix X). The relation
between the components of these vectors is given by the expressions
P1=
,,...:Xt1tEh
.
3
(i=1,2,3) (46.1)·
k=l
where Xil! is a symmetric tensor of rank two known as the electrical

susceptibility tensor.
By (4.5.2), we have D = E + 4nP or. in components.
Di = Ei + 4nPi
Introducing Pi from (46.1 ), we get
D·1 = £.l + 4n ~. x·~~.E~
- z .... K.
"
If we represent E 1 in the form "i'.l5ii,Ek, we can write
k
D 1 = ~ l5 1 ~tE1t + 4n ~ X 1tE1t = ~ (l5tl! + 4.rtXt~t) E11.

1
" k It
The quantities
(46. 2)'
are clearly the components of a symmetric tensor of rank two. It is-
called the permittivity tensor. With its aid, the relation between the
vectors D and E can be written as
The symmetric tensor has six independent components. If it is

reduced to the principal axes, it appears as follows:
We must note that since the principal values of the tensor Xil! are·
positive, those of the tensor eik are always greater than unity.
In crystals of the triclinic. monoclinic, and rhombic systems.
all three principal values of the tensor eik and. consequently. the-
semiaxes of the tensor ellipsoid are different. Such crystals are-
known as biaxial ones.
176 ELECTRODYNAMICS
In crystals of the tetragonal, rhombohedral, and hexagonal sys-

tems, two principal values coincide: e 1 = e 2 =I= e 3 • The tensor
-ellipsoid in this case is an ellipsoid of revolution. Such crystals are
:known as uniaxial ones.
In crystals of the cubic system, all three principal values of the
tensor eik are the same so that the tensor has the form e6ik· The
tensor ellipsoid in this case degenerates into a sphere. Such crystals
·do not differ from isotropic bodies in their dielectric (and optical)
properties.
Chapter IX
MAGHETOSTATICS
47. Stationary Magnetic Field in a Vacuum
A point chargee moYing at the velocity v expcrienet'~ Ill a mag-

netic fie! d the force 1
F=_:_c (vB] (47 .1)
(c is the speed of light in a vacuum). The vector quantity B called

the magnetic induction is the basic (strength) characteristic of
a magnetic fiCld. Relation (47.1) may be considered as the definition
of the qnantity B.
Owing to the absence in nature of magnetic charges similar to the
electric eharges 2 e. the lines of the vector B have neither a beginning
nor an end. This is why the flux of the vector B through any closed
surface always equals zero:
~Bndf=O (47.2)
f
Formnla (47.2) is an analytical expression of Gauss's theorem for

the magnetic induction vector.
Using the Ostrogradsky-Gauss theorem, expression (47.2) can be
written as
J\i'BdV= 0 (47.3)
v
Conditon (tl7.3) mw.;t he observed for any arbitrarily eho~en volume
V. This is po~~iblo only if the integrand fnnet ion is ZCI'o at each
poinl. We thus arrive at the conclusion that the divergence of the
magnetic induction vector is zero at every point of a field:
\i'B = 0 (47.4)
1 The sum of the forces (41.1) and (47.1) is called the Lorentz force.
2 Proceeding from the fact that the equations of physics in general and those
of electrodynamics in particular must be symmetric. Paul Dirac advanced the
hypothesis that ma~netic charges (Dirac's monopoles} ought to exist in nature.
Searches for these charges have meanwhile given no results, so that the question
of the existence of Dirac's monopoles remains open.
178 ELECTRODYNAMICS"
It is proved in vector analysis [see (XI.44)l that the divergence of

the curl of a vee tor function always equals zero. Therefore, the mag-
netic induction can be represented as the curl of a certain vector
function A:
B =[VAl (47.5)
The function A is called the vector potential of a magnetic field.
The vector potential, like its scalar counterpart <p, is determined
non-uniquely. Indeed, since the curl of the gradient of any function
is r,ero [see (XI.43)l, the addition to the vector potential of the quan-
tity V1p (here 'Pis an ;uhitrary fundion) does not chango tho value of
[VAl, i.e. B. Hence, if A is the vector potential corresponding to a
given magnetic field, the function
N = A + Vlj> (47.6)
is also a vector potential of this field.
The property (47.6) allows us to choose the potential in the most
convenient way, for instance, to impose definite conditions on the
divergence of A. Indeed, it can be seen from (47.6) that
VA' =VA+ V (V1p) =VA+ il1p
so that the proper choice of1p can give VA' any preset value. Within
the scope of magnetostatics, we shall choose 1p so that
VA= 0 (47.7)
To illustrate what has been said above, let us consider the vector
potential of the homogeneous magnetic field B = canst = B 0 • Let
us direct the z-axis along B 0 • Hence Bx = B 11 = 0, B. = B 0 , and
Eq. (47.5) written in components becomes
B _ OAx _ iJAz _ O
y- az ax - ,
(47.8)
iJAy iJAx
B z = -iJx- - - -
iJy
= B0
It can be seen that these equations are satisfied, say, by the follow-
ing value of the potential:
Ax= -B 0 y, A 11 = 0, A.= 0 (47.9)
Figure 47.1a depicts the lines of the vector A having the compo-
nents (47.9).
Solution (47.9) is not the only one. Equations (47.8) are also
satisfied by the following potential:
(47.10}
The lines of A for this case are shown in Fig. 47.1b. It is evident that
Ax = -r:xB 0 y, A 11 = (1 - r:x.) B 0x, Az = 0
MAGNETOSTATICS 179
where a is any number, will also be a solution. Particularly, Eqs.

(4 7.8) are satisfied by
(47.11)
Solution (47.11) can be represented as
1 1 1
Ax=2(B71 z-Bzy), A 11 =2 (B:x-Bxz), Az=T (BxY-Byx)
(remember that Bx =By = 0), whence
1 1
A= 2 [B 0r] = 2 [B 0R] (47.12)
where R is the component of the position vector r perpendicular to
the z-axis. The lines of the vector A corresponding to (47.12) are
shown in Fig. 47.1c.
y y y
X
I f
X X
(a) (b) (c)
Fig. 47.1.
All the values of the potential we have given satisfy condition

(47.7). We thus conclude that Eqs. (47.5) and (47.7) do not complete-
ly determine A. For its determination to be unique, we must set
the boundary conditions for it.
48. Poisson's Equation for the Vector Potential

It is known from the general course of physics that the cireula tion
of the vector B around any closed contour r taken in a stationary
magnetic field (in a field of steady currents) is proportional to the
algebraic sum of the currents enclosed by the contour:
~~ B1 dl = 4c1t 2:; i (48.1)

r
We shall treat formula (48.1) as a relation established experimen-
tally.
12*
180 ELECTRODYNAMICS
Introducing the current density j, we can represent the sum of the

currents as the flux of the vector j through the surface confined by
the contour r. Formula (48.1) thus becomes
~ Bl dl =
r
4; Jj df
t
Lel us transform the left-hand side of the relation we have obtained
according to Stokes' theorem. The result is
J[V'B] df
I
= 4; J df
t
j (48.2)
Assume that integration on the left-hand and right-hand sides is

performed over the same surface [although Eq. (48.2) is also ob-
served for different surfaces if only they rest on the same contour rJ.
Relation (48.2) must be observed for any arbitrarily taken surface f.
This is only possible if at every point of the field
[Y'B] = 4rt j ( 48.3)

c
(we remind our reader that we are considering the field of steady
currents).
Equation (48.3) plays the same basic role in magnetostatics as
Eq. (42.3) does in electrostatics. Together with Eq. (47.4), it allows
us to calculate the field of preset stationary currents.
Let us introduce the curl of A [see (47.5)1 instead of B into for-
mula (48.3):
[Y', [V'AJ]=~j
c
According to (XI.45), we have [Y', [V'All = Y' ('VA)- ~A. As-
suming, as we agreed upon [see (47. 7)1, that Y' A= 0, we get the fol-
lowing differential equation for A:
L1A = - 4rt j (48.4)
c
This vector equation is equivalent to three scalar equations:
L1Ak = - 4 n h (k = x, y, z) (48.5)
c
'each of which is similar to Poisson's equation for <p [see (42.4)].
Equations (42.4) and (48.5) are equivalent from the mathematical
vfewJ?oint. Consequently, substituting A k. for rp and hie for p i~ the
~olutwn of Eq. (42.4), we get the solutwn of Eq. (48.5). W1th a
view to formula (42. 7), we obtain
Ak = _ci Jf ik (r')
lr-r'l
dV' (k
=x,y,z
) (48.6)
MAGNETOSTATICS 181
where the integral is evaluated over the entire region in which the
currents producing the field flow.
The three expressions (48.6) can be combined into a single yector
one:
A=_!. [ j (r') dV' (48. 7)
c .l lr-r' I
FMm11la (It~. 7) allows 11s to calcillnlt' the \'Prior polC'nl ial of the
field set up by currents according to their known distribution in
space. Let us consider the field of a line (straight) current as an
example. It is general knowledge that the potential of thr electric
field produced by n thin infinitely long uniformly chargL'd l'lianH'nl.
can be written ns 1 (we arc considering the field outside the lilnment)
R R
cp(R)= -2Aln-= -2puln- (48.8)
Ro R0
where R is the distance from the filament, R 0 is the distance to
points whose potential is taken as zero (in the given case we cannot
assume that cp = 0 at infmity because with such normalization the
potential is infinitely great at finite vnlnes of R), and f... is the linear
density of the charge which, assuming the latter to be uniformly dis-
tributed over the entire cross section of the filament, can be written
as pu (here u is the cross-sectional area of the filament).
Now assume that we have a thin infinitely long straight wire
through which a current of density j flows that is uniformly distri-
buted over the cross section u. Directing the z-axis along the \\'ire,
we have jx = j!l = 0, jz = j. On the grounds that Eqs. (42.4) and
(48.5) are identical, we can obtain an expression for Az b~· :"ubsti-
tuting Uc = j/c for p in (48.8):
2ja R 21 R
A z (R)= - -
c J nR- = - -c l nRo
- (48.9)
0
where i = ju is the current flowing through the wire. The introduc-

tion into (48.8) of jx = 0 and j,1 = 0 instead of p yields zero nlues
for the components Ax and Arr Hence, the vector.\ can be written as
2i R
A= - -c l n - e
Ro z
(48.10)
where ez is the unit vector of the z-axis.

Taking the curl of expression (48.10), we find B:
B=['VAJ=- ~i [v, {In ;o ez)]
= --f2' Lr ( 'V ln R )
Ro ,
J 2i
ez ---;- R
ln Ro ['Vez]
1 It is a simple matter to find that r: (R} = 2i ..'R \\'ilh the aid of Gauss's
tl:eorem. Consequently, dcp/dR = -2)./ R. Integration lead~ to formula (48.8).
182 ELECTRODYNAMICS
[we have used formula (XI.27)l. Since ez = const, the second term
vanishes. The gradient of the function ln (R/ R 0 ) equals R/R 2 • There-
fore,
21 [ R
B= --c- 2i J
R2 'Cz =CJ[2(Cz, R] (48.11)
According to the result we have obtained, the vector B at every

point is in a plane perpendicular to the wire and directed along a
tangent to the circle encompassing the wire. The magnitude of
B is
B=~
cR
49. Field of Solenoid

Let us calculate the field of an infinite solenoid, representing it
as an infinite cylinder of radius a in whose surface layer of thickness
b (where b «a) a current of density j flows (here bj is equivalent to
ni, where i is the current in the sole-
Y noid, and n is the number of its turns
per unit length).
We choose rectangular coordinate axes
and make the z-axis coincide with
the geometrical axis of the solenoid
(Fig. 49.1). The projections of the
Y vector j onto the coordinate axes are
• • • • y
Jr.= -Jsma= -J-a ,
j y = j cos a =j : , jz = 0 (49.1)
Fig. 49.1.
It immediately follows from j 1 = 0
that Az = 0. In accordance with what
was said in Sec. 48, the component Ax coincides with the potential cp
produced by the charge distributed in the surface layer of the cylin-
der with the density Px = jxlc = -(jlc) (yla) = -p 0 (yla). We have
introduced the symbol
j nt
Po =-;=--;;c (49.2)
Similarly, the component Ay coincides with the potential of the

charge distributed in the surface layer with the density py =
= j 1)c = p0 (xla).
The density, which changes according to the law -y/a, can be
obtained as follows. Let us insert a cylinder charged uniformly with
the volume density +Po into another cylinder charged uniformly
MAGNETOSTATICS 183
Jx=-J. srn
. .y
ex =-Ja
•
Jy=J
• • X
cos ex =Ja
y y
. y X
Px=-Posmcx=-Poa Py= Po cos a= Po a-
(a) (b)
Fig. 49.2.
with the volume density -p 0 • As a whole, the system of two cylin-

ders is neutral at every point. Now let us move the negatively charged
cylinder over the length b/2 in the direction of the y-axis. and
the positively charged cylinder over the same distance in the oppo-
site direction (Fig. 49.2a). Since b -«:a (here a is the radius of the
cylinders). the formed system may be considered as a eylinrlf'r in
whose surface layer of thickness b a charge having the Yarying den-
sity Px is distributed. Actually, the density of thf' charge in the sur-
face layer i::- constant and equals p0 (m -p 11 ). whilf' what do£':' chang£'
is the thickness of the uncompensated layer. A glance at Fig. 49.2a
shows that this thickness Yaries according to thP law b sin tX so that
the charge per unit of cylinder surface is- p0 b sin tX= - p0 by/a.
Our error will not be noticeable, however, if we consider the thick-
ness of the charged layer to be the same eYerywhere and equal to b,
and the charge density in this layer to vary according to the law
__ PoY
P:x- (49.3)
a
184 ELECTRODYNAMICS
To obtain a density varying according to the law xla, let us shift

the cylinders as shown in Fig. 49.2b. Now the density of the charge
in an imaginary surface layer of thickness b will vary according to
the law
PoX
Py=-a- (49A)
In the cases being considered, the fields· are the superposition of

the two fields +c:p 0 and -<p 0 , identical in magnitude but opposite in
sign, displaced relative to each other over the very small distance b
(the latter is directed oppositely to the y-axis for Px• and along the
x-axis for py)· Assume that initially the fields were accurately super-
posed, as a result of which the potential at the point determined by
the position vector r was f+c:p 0 (r)l + [-<p 0 (r)), i.e. zero. Now let us
shift the field +c:p 0 by b/2, and the field -cp 0 by -h/2. Hence, the
potential +c:p 0 that was at the point (r - h/2) and the potential -<p 0
that was at the point (r + h/2) will be at the point with the position
vector r. Consequently,
<p(r)=<p0 (r- ~ )-c:p0 (r+ ~)
Owing to the smallness of h, both terms of the expression we have
obtained can be transformed by formula (X1.5), i.e. can be written as
<p (r) = [ cp 0 (r)- \7<p0 • b/2]- [<p 0 (r) + \7<p 0 • b/2]
= -\?<po· b = - ( 7iX
8<po b
x+ 8<po b
ay 8<po
y+~
bz )
For Px• the vector b has the components b:~e = bz = 0, and by =

= -b, so that
<p (r) = 88: 0 b (for A:~e) (49.5)
for py, we have by = bz = 0, and b:~e = b, so that
cp (r) = - 8{)~0 b (for Ay) (4!).6)
It now remains to find cp 0 and introduce its derivatives into (49.5)

and (49.6). Recall that <p 0 is the potential of the field produced by a
cylinder of radius a charged with the constant volume density +Po·
The potential inside and outside such a cylinder is determined by
different formulas.
Field Inside a Solenoid. We can easily find with the aid of Gauss's
theorem that the field inside a charged cylinder is E = 2np 0 R,
where R is the distance from the axis of the cylinder (R < a). The
potential <p 0 = -np 0 R 2 + const = -np 0 (x 2 + y2 ) + const corre-
l'vlAGNETOSTATICS 18S
sponds to this field. Its derivatives are

8!p 0 2npox' 8!Jlo 2
----az- = - ay =- 1tPoY
Let us introduce the found values of the derivatives into (49.5}
and (49.6):
(p (r) = -2np 0 by (for Ax)
(p (r) = 2np 0 bx (for Ay)
Substituting nil be for p 0 in these expressions [see (49. 2)], we obtai11:
expressions for Ax and A y (it was established above that A z = 0):·
A. = _ 2nni Y = _ ...!_ ( 4rtni ) Y )
X C 2 C
A y_- -2nni 1 ( .:lrtni )

c-X=2 -c- X
(49.7)·
Az=O
It was shown in Sec. 47 [see (47.11)1 that the potential (49.7) de-
termines a homogeneous magnetic field with the magnetic induction
B = 4c:1 ni (49.8)
parallel to the z-axis, i.e. to the axis of the solenoid.
Field Outside a Solenoid. The potential outside a cylinder charged
with the density Po is
'Po= - 2p 0na 2 In....!!.._
Ro
where R is the distance from the cylinder axis (R >a), and R 0 is.
a constant [compare with (48.8)1. The derivatives of 'Po can be writ-
ten as
8cpo = - 2p na2 _Y_
8cpo -
----az- -. - 2Pona 2 7f2
X
' ay o R2
Substitution of these values into (49.5) and (49.6) yields
(p (r) = -- 2p 0 na2b : 2 (for Ax)
cp (r) = 2p 0 na 2 b ; 2 (for Ay)

Substituting ni/bc for p 0 , we obtain Ax and Ay. Let us write the·
values of all three components:
A X = _ 2nnia 2
_!!_ = _ K .2._ )
c R_2 R2
2rtnia 2 x 1· (49.9}
Ay = - c - }-{2 =K R2
Az=O
186'
The quantity R 2 = x2 +y 2 does not contain z. Consequently,

.0Axl8z = 8Ayl8z = 0. Since Az = 0, it follows that oA)ax =
= 8A 2 !8y = 0. Hence, Bx and By vanish. Let us find B 2 :
i!Ay aAx [ 1 2x 1 2y y ) -,
Bz = {)x-- -ay = [{ ~ R2 -- Ra R
I X )
-
(
- R2 + R3 R J
=K 2
[ R2 -
2
R2
x2 + y2 J=
R2 0
We have thus found that the field outside an infinitely long sole-
noid is zero. The vector potential outside the solenoid, however, is
non-zero. The collection of formulas (49. 9) can be written as a single
vector formula:
A= 2nnia2 '[ R] (49.10)
cR~ ' ez,
A comparison with formula (48.11) shows that the field ot the vector
A outside the solenoid has the same nature as the field of the vector
B around a straight long current-carrying conductor.
50. The Biot-Savart Law

The expression
A{r) =-c
1 j' j (r') dV'
(50.1)
lr-r'l
makes it possible, knowing the distribution of currents in space, to
·determine the vector potential of a magnetic field [see (48.7)]. Let
us attempt to find a formula that would allow us to find B directly
according to the preset currents. For this purpose, let us calculate
the curl of the function (50.1). It must be borne in mind that inte-
gration in (50.1) is performed over the primed coordinates x', y z', 1
,
whereas differentiation in taking the curl is performed over the un-

primed coordinates x, y, z. Therefore, we may exchange the places
·of the operations of integration and taking the curl. With this in
view, we obtain
B (r) = [V, A (r)J =+ j' [ V, ;\:~;,I JdV'
Considering j (r')/\ r - r' I as the product of the vector function
j (r') and the scalar function 1/\ r - r' \,we shall use formula (XI.27):
[ V, j'(r')
lr-r'l
J= [ (V, lr-r'l
1 J ) • ( ')
'J r
J+ lr-r'l
1 (
V, J r
• ( ')]
"The second term vanishes because j (r') contains no unprimed coor-

·dinates. The gradient in the first term is
r-r'
V -,-r_1_r-:-,1:-- = V . 1
V ~ (xt-xj)2
l\IAGNETOSTATICS 187
We have thus arrived at the formula
B(r) = c1 Jr [- - r-r'
lr-r'l" , j (r') JdV'
=+) (j (r'), (r-r')] dV'
lr-r' 13
(50.2)
(we have put the scalar factor outside the sign of the vector product
and, in addition, have changed the places of the factors, which re-
sulted in vanishing of the minus sign).
The formula obtained is a solution of the problem we posed-it
allows us to calculate B from the preset currents. Formula (50.2) is
simplified if the currents flow only through thin wires. Figure 50.1
Fig. 50.1. Fig. 50.2.
shows a portion of such a wire. Inspection of the figure shows that

the expression j dV' can be written as
j dV' = jcr dl = jcr dl = i dl (50.3)

where a is the cross-sectional area of the wire, i is the current, and
dl is a vector coinciding with the wire element dl and having the
same direction as the current. Substituting i dl for j dV' in (50.2),
we obtain
B = _!_ f ij[dl, (r-r')]
(50.4)
c J lr-r' 13
(integration is performed over the length of all the wires).

Formula (50.4) is an analytical expression of the Biot-Savart law.
Figure 50.2 explains that r - r' is a vector drawn from the
point where the current element dl is to the point for which B is
being calculated.
188 ELECTRODYNAMICS
51. Magnetic Moment

Before turning to the topic of the present section, let us obtain the-
continuity equation, which is a corollary of the law of charge con-
servation.
Assume that currents characterized by the density j = j (r) flow in
a certain region. Let us separate an imaginary volume V confined by
the surface f in this region. The charge flowing outward through this:
surface in unit time can be written as
~ jn df = J'Vj dV
f v
The above expression equals the rate of diminishing of the charge-
confined in the volume V, which is determined by the expression
-- ~ I p dV =
dt J
-
JI ~
at
dV
v v
(we have written the partial derivative with respect to t because P·
is a function not only of time, hut also of the coordinates).
Equating the two expressions, we obtain
Jv 'Vj dV = - vJ ~~ dV
The above equation must be observed for any arbitrarily chosen
volume. This is possible only upon equality of the integrands at
every point of space. We thus arrive at the relation
• ap
'VJ= · - - (51.1}
at
known as the continuity equation in the differential form. In the-
integral form, the continuity equation is
~ jn df = -
I
:t J P dV (51.2}
Now let us consider a system of stationary currents circulating

within a restricted volume V and calculate the magnetic field set up.
by this system at distances that are great in comparison with the
dimensions of the system. This problem is similar to that treated in
Sec. 43.
First of all, we must note that owing to the stationary nature of
the currents the accumulation or dissipation of charges cannot occur
at any of the system's points, i.e. the condition op/ot = 0 must hold
everywhere. It thus follows by (51.1) that within the system
'Vj = 0 (51.3)
MAGNETOSTATICS 189
Further, there are no currents outside the system. Hence, every-

where on the surface confining the system, we have
jn = 0 (51.4)
{the currents do not intersect the surface).
Finally, we shall prove that
J jdV=O
I'
(51.5)
where the integral is evaluated over the entire volume of the system.
Owing to the stationary and restricted nature of the system, all the
-current tubes 1 that can be separated inside the system are closed.
Consequently, the entire volume of the system can be divided into
dosed current tubes. For each of the tubes, the integral over its
volume can be transformed as follows:
J
ovPr t uhc
j dV = J jcr dl = ~i dl = i~ dl = 0 ((51.6)
{a is the cross-sectional area of a tube, i is the current, and dl is an

·element of length of a tube, see formula (50.3)1. Summation of ex-
pression (51.6) over all the current tubes yields formula (51.5).
Let us choose the origin of coordinates inside the system and write
an expression for the vector potential:
A (r) = _!_ l' j (r') dV' (51. 7)
c J ! r-r' I
v
Here r is the position vector of the point for which A is being calcu-
lated, and r' is the position vector of the point in whose vicinity the
elementary volume dV' is; integration is performed over the primed
coordinates within the confines of the system's volume.
Taking advantage of the fact that according to our assumption
r' ~r, let us expand expression (51.7) into a series 2 in r' /r. To
within first-order terms, we have
A (r) = _!_ f j (r') dV' _ _!_ .r jJ(r') {v _!_r r'} dV' (51.8)
c J
v
r c
vJ
1 By a tube of an electric current is meant the same as by a tube of a flow
in a fluid, i.e. the volume confmed by lines tangents to which at each point
coincide with the direction of the vector j.
2 Considering -r' as a small increment of the position vector r, we write the
function f (r) = 1/r for lite value of the argument r - r' as
f (r - r') =I (r) + Vf (r) ( -r')
190 ELECTRODYNAMICS
The first term of (51.8), owing to (51.6), vanishes. Consequently.

substituting for V (1/r) its value. we obtain
A(r)=! .\ j(r')(~~')dV'
v
Let us write the k-th component of the potential:
Ak = c~s ) h (r') ( ~ xixi) dV' = c~ 3 ~ x 1 ~ hxi dV' (51.9)
i i
(we have expressed the scalar product rr' through the components of
the position vectors). The products hxi are the components of a ten-
sor of rank two 1 • Let us write this tensor as the sum of a symmetric
and antisymmetric tensors:
. , jkxi+hxk ikxi-hxk
lkXi= 2 2+ Skt+Akt (51.10)
[see (X.27)).
Let us prove that the integral of the symmetric component of the
tensor (51.10) is zero. For this purpose, we shall use the identity
V' (xix/.j) =- jV' (x(xi)+xixiV'j (51.11)
(see (XI.26)). In view of (51.3), the second term on the right-hand
side vanishes. Let us write the expression V'(xixi,) using the for-
mula for the gradient of the product of scalar functions [see (XI.25)1:
V' (xixk) = xiV' xk + x/,.v' xi
, "'0 8xk , "'0 8xj "'0 , ,
=Xi k.J em-;;-;-+
vXm
Xk k.J em -8Xm' = LJ em (xi<'>km +r Xk<'>tm)
m m m ·
(we have taken into account that ox//ox'm = Okm)· The expression
in parentheses is the m-th component of the gradient of the function
xixk·
Now let us calculate the second term of expression(51.11):
jV' (xi xi)=~ im {V' (xixk) }m = ~ im (xi<'>km
m m
+ XkOtm)
In the first sum on the right-hand side. only the addend with m =
= k is non-zero, and in the second-the addend with m = i. Hence
jV' (xi xi) = j kxi + j ,xi
The integral ht = ~ ikxi dV' is also a component of a tensor. Consequently,

1
the vector A to within the factor 1/cr' is the product of the tensor Ikt and the
vector x 1 [see (X.22)].
MAGNETOSTATICS 19t
Therefore, by (51.11)
S -
hi-
j,x£ +2 j;x/. 1 ' ( . , ')
=7 \l X;XkJ
Let us take an integral of this expression:
.\ S,<i dV' = ~ .\ 'V' (xix/.j) dV' = ~ ~ xix/Jn df
}' y ;
(we have employed the Ostrogradsky-Gauss theorem). In vie\v of
(51.4), the last integral vanishes. We have thus proved that the
integral of the symmetric component of the integrand on the right-
hand side in (51. 9) vanishes. This allows us to write expression
(51.9) as
1 'V
A k = Cr3 f ihxi-iixkdV'
L.J Xi J 2 = 2cr 3
1
Jf {"'-" (' ' . ')}
L.J Xt ]kXi- JiXk
dV'
i v v ;
(51.12}
We shall show that the integrand equals the k-th component of
the triple vector product [r [jr']]. Introducing the auxiliary notation
b = [jr'] and using formula (VI.33), we can write
[r,[jr'llk=[rb]k=~ehi!xibt=L:ehi!Xt ~ elmnjmx~
i.! i, l m, n
(we remind our reader that in a cyclic rearrangement of the subscripts

the value of e 1mn does not change). We summate over the subscript l
using relation (VI.16):
[r, [jr'lh = ~ 6km6inX;jmX~-
i. m.tn
':E
i. m. n
6hn6imXijmX~
Now let us summate over the subscripts m and n. In the first sum,
the addends with m = k and n = i are non-zero, and in the second-
the addends with n = k and m = i. Hence,
[r, [jr' Jl 11 = ~ X;hxi-
i
Li X;i;Xk =:_:X;
i
(hxi- itxl.)
A comparison with (51.12) allows us to ·write

Ah = ;r
2 3
v
~ [r, [jr']]k dV'
or in the vector form

A= 2c1rs ~ [(r'j], r] dV' = r~ [ { 2~ .\ (r'j) dV'}, r (51.13) J
v v
(we have exchanged the places of the factors in both vector products).
192 ELECTRODYNAMICS
The quantity
m= ;c ) [r' j] dV (51.14)
v
is known as the magnetic moment of a system 1 •
For a system of discrete charges, the expression of the magnetic
moment is
1 )1
m= 2c-;/ l , I
ea faVa
(51.15)
'The latter is obtained from (51.14) if we assume that

j (r') = p (r') v (r')= ~ ea v (r') 6 (r'- r~)
a
J
:[see (41.12)1 andj have in mind that eav (r') 6 (r' -r~) dV' taken
in the close proximity of the point r;, transforms into eava.
The magnetic moment depends only on the properties of a system
and, as can readily be seen, does not depend on the choice of the
·coordinate system. Indeed, let us displace the origin of the coordi-
nate system over the distance b. The position vectors in the new sy-
stem will now be r" = r' - b. The magnetic moment in the new
;system is
m' = ;c -~
y
[r"j] dV = 2~ r
y
[(r'- b), j) dV = 2~ J[r'j] dV- ;c .\
y v
[bj] dV
'The first term equals m, and the second one can be written as
- 2~ [ b, J dV J
j
But according to (51.5), ~· j dV vanishes. Consequently, we have

.arrived at equality of the moments m' and m, Q.E.D.
With a view to (51.14), expression (51.13) can be written as fol-
lows:
[mr)
A (r) =--;:a= - [_m, "Y-;1 H5t.16) J
[compare with formula (43. 7)).
To find B, we must calculate the curl of expression (51.16). As-
suming in formula (IX.29) that a = m and b = r/r 3 , we obtain
B =["VA]= [ \7, [ m, ; 3 J] = ( rra 'V) m- (m'V) r~ + m ( 'V ; 3 )
r
--;:a ("Vm)
1 We have omitted the prime on dV because the integrand contains no un·
primed coordinates. In the expressions containing both r' and r, the prime
on dV indicated that integration is performed over the primed coordinates.
In the following we shall also omit the prime on r in this expression.
MAGNETOSTATICS 1Q3
The vector m does not depend on r, therefore the first and last terms
on the right-hand side vanish. According to (XI.26)
"V -.r 1
,r8 =r"V3
T
+ 31 T
"Vr= r ( --,--
3
TT
r) +-1 3= 0
T3
Therefore, the third term also vanishes. Hence,

B = - (m"V) ; 8 (51.17)
By formula (XI.33), we have
(m"V) ~
T
= r ( ID· "V~)
T
+ -4- (m"V) r =
T
- r ( m.~..!.)
r~ T ·
+_!
r3 m
m-3er (mer)
ra
[see (XI.34), er is the unit vector of the vector r]. Consequently,
B = 3er (m;;)-m (51.18)
A comparison with formula (43.10) shows that a magnetic field is
expressed in terms of the magnetic moment by a formula like the
one expressing an electric field in terms
of the electrical dipole moment.
In concluding, let us calculate the
magnetic moment of a current flowing
through a thin wire that forms a plane
loop. We choose the origin of coordinates
in the plane of the loop (Fig. 51.1 ).
According to the definition (51.14)
m =
v
;c J[rj] dV
In the case being. considered, we can
make the substitution j dV = i dl [see Fig. 51.1.
formulas (50.3) and (51.6) Consequently,
the expression for m can be written as
rn=i- ;V (r, dl]
_c 't_
Designating the unit vector of a normal to the plane of the loop by
the symbol n (in Fig. 51.1 this unit vector is directed beyond the
drawing), the integrand can be written as nr sin a dl so that
m = ~ rh r sin a dl
c J 2
A glance at Fig. 51.1 shows that the integrand equals the area of
the hatched triangle. The integral therefore equals the area I of the
loop. We have thus arrived at an expression for the magnetic moment:
1 '/ n
m=-L
c
f94 ELECTRODYNAMICS
52. Field in Magnetics

In the presence of magnetics, the field produced by the molecules,
Bmoh is superposed on the field of the conduction currents Bi so
that the resultant. microscopic magnetic field can be written as
Bree = BJ + Bmol (52.1)
The field Bres. like the field Eres determined by formula (44.1), is a
fast-varying function of time. In addition, it varies greatly in the
space between two adjacent molecules. This is why the macroscopic
quantity
B=BJ+(Bmot> (52.2)
·is taken as a characteristic of the magnetic field in magnetics. It is
called the magnetic induction in the magnetic. The microscopic
.field Bmol is averaged in the same way as the .field Ebound (see
Sec. 44).
It is customary practice to characterize the state of a magnetic
by the magnetic moment of a unit volume of the magnetic, known as
the magnetization. We shall denote the magnetization (which is
a vector) by the symbol M. It is obvious that M can be determined as
M=
lJmi
---;w- (52.3)
where dV is an infinitely small volume, and mt is the magnetic mo-
ment of an individual molecule; summation is performed over all
the molecules contained in the volume dV.
The contribution made by a magnetic to a macroscopic magnetic
field can be calculated by formula (51.16). According to (52.3), the
volume element dV' of a magnetic has the magnetic moment dm =
= M dV'. Consequently, at the point determined by the position
vector r, this volume element produces the magnetic potential
dA (r) = !M (r'), (r- r')J dV' (52.4)
I r-r' 18
[r' is the position vector of the point where the volume element dV'
is, and M (r') is the magnetization at this point].
The integral of expression (52.4) evaluated over the entire volume
of a magnetic gives the contribution made by the magnetic to the
macroscopic magnetic potential. It must be added to the magnetic
potential produced by the conduction currents [see formula (48. 7)].
Consequently, in the presence of a magnetic, the field is [character·
ized by the potential
A(r)=_!_ i j(r')dV' +i [M(r'),(r-r')) dV'=l +l (52.5)
c J lr-r'l J lr-r'l 3 1 2
MAGNETOSTATICS f95
The second integral in this expression can he transformed as fol~

lows:.
IM (r'), (r-:-r')J
lr-r' Ia
dV' ~ f
J
[M (r') '. r-r' .
lr-r' Ia
J.dV'
, . ,
=HM(r'), .. \7' ( lr_!r'l )]av•·
(the prime ou V signifws that when taking the gradient, differenti-
ation is performed over the primed coordinates).
· Assuming in formula (X1.27) that cp = 1/i r - r' I and a =
== M (r'), we obtain
[ V, ~~(;:~ l [V' ( lr~r' I ) ' M (rf) J+{ lr~r' I ) IV'' M (r')l

whence
[ M (r'), V' ( lr-r'
1
I
)]=· [V', M(r')J
[r-r'] [ v', M (r') ]•
lr-r'l
Consequently,.the second term in formula (52.5,) can be written as
T = f [V'' M (r')] dV'- f [v' M (r') JdV' =I'+["

2 J Ir - r' I J . ' Ir-'- r' I 2 · 2
Let us transform the integra,! r; by formula (XI.60):

l"=f["Y'
2
M(r') ]dV'=I[df'
J . 'lr-r'l
M(r')
J _ 'lr-r'l
J
v t
If a magnetic occupies a finite volume or M(r') diminishes sufficien-
tly rapidly with an increasing distance from the origin of coordi-
nates, the last integral vanishes [when a magnetic is localized ina·
finite volume, the· integration surface can be chosen outside the
magnetic, and in this case M(r') = 0 everywhere on the surface].
Hence, in the presence of magnetics, the vector potential is deter-
A(r)=/ +I'=_!_ f j(r')dV' f IV', M(r')ldV'
1 2 c J lr-r' I. +J lr-r' I
=_!, [ j (r')+c [\'', M (r')] dV' (52 6\.
c J lr-r' 1 · r
The result we have obtained may be interpreted to signify that
the magnetization ma].;es the same contribution to the vector poten-
tial as the current of density
(52. 7)
(we have omitted the prime on '\7, and accordingly M is considered as.
a function of r instead of r'). It follows that in the presence of magnet..
ta•
196 ELECTRODYNAMICS
ics, Eq. (48.3) must be written as

[VB] = ~
c
j + 4n [VMJ (52.8)
Combining the terms containing a curl, we obtain

[V', (B-4nM)] =~
c
j
The quantity
H = B- 4nM (52.9)
is called the strength of a magnetic field. It is an auxiliary macro-
scopic characteristic of a magnetic field similar to the electric dis-
placement D [see (45.2)1. The following equation holds for H:
[VH] = :n j (52.10)
Experiments show that for diamagnetics and paramagnetics the

magnetization is proportional to the field strength:
M = XmH (52.11)
(The magnetic is assumed to be isotropic. In addition, we have in
mind fields for which the magnetization is far from saturation.)
The quantity Xm is known as the magnetic susceptibility of a sub-
stance. It is positive for paramagnetics and negative for diamagnet-
ics. Substitution of expression (52.11) into formula· (52.9) yields
B = H + 4nxmH = 11H (52.12)
where
11 = 1 + 4nxm '(52.13)
is the permeability of a substance.
Relations (52.11)-(52.13) are also used for ferromagnetics, treating
Xm and 11 as functions of the field strength H.
The-expediency of introducing H is due to the same considerations
that were set out in Sec. 45 to substantiate the expediency of intro-
ducing D.
Consider a field in a homogeneous isotropic magnetic.Assume that
in the absence of the magnetic the given conduction currents produce
.a field characterized by the induction B 0 and the potential A 0 • It is
..known that B 0 and A 0 are solutions of the equations
[VR 0 ] =~
c
i (52.14)
4n.,
D. Ao = - -
cJ
(52.15)
[see (48.3) and (48.4)1.

MAGNETOSTATICS 197
Now, without changing the conduction currents, let us fill the

entire space in which the fteld differs from zero with a homogeneous
(f.t = const) isotropic magnetic. The field induction will therefore
become equal to B, and the potential will become A. Let us write
equations for B and A. By Eq. (52.10), we have
[VHJ = [v
-
..!.!__] = i::_ j
fl c
(52.16)
Substituting [VAl forB in this equation [see (47.5)1 and taking into
account that fl = const, we can write
~ [V, [VA]] = ~ j
Let us develop the left-hand side of the expression obtained by for-

mula (XI.45). The result is
~{V (VA)-~A}=i::_ j
~ c
But we have agreed to choose A so that VA vanishes [see (47. 7)].
Hence, the first term in braces vanishes, and we arrive tat the
equation
(52.17)
A comparison of (52.16) with (52.14) and (52.17) with (52.15)

shows that the equation for B/11 coincides with that for B 0 , and the
equation for A/11 with that for A 0 • This shows that the filling of a
space containing a field with a homogeneous magnetic leads to an
increase in both the magnetic induction and the magnetic potential
of 11 times.
By Eq. (52.17), Poisson's equation (48.4) for a field in a homoge-
neous isotropic magnetic has the following form:
~A=- 4 n~j (52.18)
c
When a homogeneous isotropic magnetic with ~L = const (i.e.

with ll not depending on H) fills the entire space in which the field
differs from zero, the following relation holds:
H=_!_ [VAJ=[VA'I (52.19)

~
where A' = Ai11.
We shall underline the fact that the field strength H can be repre-
sented in the form of the curl of the function A/11 only when the
magnetic is homogeneous and fl = const. The magnetic induction
B, however, can always be written as B =[VAl because VB= 0 in
any conditions.
198 ELECTRODYNAMICS
Recall that at the boundary between two magnetics, the vectors

B and H must satisfy the following conditions:
Bnl = Bn2t B nl'flo 1 = B 1:2/!l£\.
(52.20)
Hn = H1:21 f.L1Hn1 = f.L~Hn2 J
A derivation of these boundary conditions can be found in textbooks
of general physics.
In anisotropic magnetics, the relation between M and H is given
by the expressions
3
Mt= ~ Xm,thHh (i=1, 2, 3) (52.21)
k=i
where Xm, u, is a symmetric tensor of rank two called the magnetic

susceptibility tensor.
The equations relating the vectors B and H can accordingly be
written as follows:
3
B, = 2J !-1-thH
k=i
11. (i = 1, 2, 3) (52.22)
where the quantities

1-'-tk = Bik + 4nxm,lll. (52.23)
are the components of the magnetic susceptibility tensor (compare
with the formulas in Sec. 46).
Chapter X
TIME-V ARYIHG
ELECTROMAGNETIC FIELD
53. Law of Electromagnetic Induction
A change in the magnetic induction flux through a closed contour

r is attended by the appearance of an induced e.m.f. in the circuit
equal to
'f, 1 = - ..!_ d<D = - ..!_ ...:!..._ \ B df
c dt c dt J
f
(<!> is the magnetic flux penetrating the contour). If the surface over
which the integral is evaluated does not vary with time, the opera-
tions of differentiation with respect to t and integration over the
coordinates may exchange tb.eir places. Henct>, the e.xpn?$5iun for
~~ can be written as
~~=
1
--
c j• -aB
atd f (53.'1)
(we have written the partial derivative with respect to t because B,
generally speaking, is a function of both time and the coordinates).
The e. m.f. by definition is the circulation around a given contour
of the field strength vector Eext of extraneous forces. In the present
case, the strength E of a vortical electric field produced by the vary-
ing magnetic field B is the strength Eext· Consequently,
~i = ~
r
E I dl = J[VE] df
t
(53.2)
(we have used the Ostrogradsky-Gauss theorem).

Equating the right-hand sides of expressions (53.1) and (53.2),
we arrive at the relation
J[VE] df= - - J-an df

i
c i}t
(53.3)
Assume that both integrals are taken over the same surface [Eq.
(~3.3) is also observed for different surfaces if only they rest on the
same contour rJ. Relation (53.3) must be satisfied for any surface f.
This is possible only if the integrand functions have the same value
at every point of space. We thus arrive at the equation
'[VE] = -~1. oB (53.4)
c i}t
200 ELECTRODYNAMICS
The equation we have obtained does not include the parameters of

the contour with whose consideration we began this section. It is
natural to assume that this equation must be observed for any point
of a field regardless of the presence in this field of a physically sepa-
rated (particularly a conducting) contour. This assumption is known
to be justified experimentally.
Examination of Eq. (53.4) shows that a time-varying magnetic
field B must set up an electric field E. Indeed, for [VEl to be non-
zero, the presence of a non-homogeneous (i.e. varying from point to
point) field E is needed.
54. Displacement Current
In analysing the equations describing electromagnetic phenomena,

J. Maxwell gave attention to the fact that in the non-stationary case,
the equation
[VH] =~j (54.1)
c
[see (52.10)1 is incompatible at ap/at =fo 0 with the continuity
equation
Vj=-~
i}t
(54.2)
[see (51.1)1. To convince ourselves that this is true, let us take the
divergence of both sides of Eq. (54.1). Since the divergence of a curl
is always zero [see (XI.44)l, we arrive at the conclusion that the
divergence of j and, consequently, ap/at too, cannot be non-zero. But
the conclusion that ap/at always equals zero does not agree with ex-
periments: in non-stationary processes, the density of the charges
quite often varies with time.
Equations (54.1) and (54.2) can be brought into agreement by
adding in (54.1) to j a quantity (we shall denote it jd) having the di-
mension of current density. This quantity must be determined so
that tiro condition
v (j + jd) = 0
will always be satisfied. It follows from this condition with a view to
(54.2) that the addend j d must satisfy the relation
• . ap
V'Jd= -'V]=-
i}t
(54.3)
Time differentiation of Eq. (45.3) yields the equation
! (VD) =4n :~
TIME-VARYING ELECTROMAGNETIC FIELD 20t
or, changing the sequence of differentiating D with respect to time

and the coordinates,
V (aD)
at =4n ~
at
Substituting the expression obtained from this equation for oplot
into formula (54.3), we arrive at the equation
. ,- 1 ( aD )
'YJct= 4nv ~...
It follows from this equation that the expression
jd = 4~ ~~ (54.4)
must be taken as jd [in the general case the arbitrary function of
time f (t) must be added to the right-hand side, but we have as-
sumed that it equals zero].
Maxwell thus arrived at the conclusion that for a time-varying-
field Eq. (54.1) must be written as
[VH] = ~ ( .
c J
+ _1
4n
aD )
at
or
(54.5).
He called the quantity jd the density of the displacement current

and the sum j + jd the density of the total current.
A glance at Eq. (54.5) shows that a magnetic field may be pro-
duced not only by conduction currents, but also by a time-varying-
electric field. Hence, the introduction of the displacement current
made the fields E. and ll equivalent with respect to the ability to·
produce each other.
SS. Maxwell's Equations

The collection of Eqs. (4.5.3), (47.4), (53.4) and (54.5) forms a sys-
tem of Maxwell's equations. It is customary practice to group these
equations in pairs. The equations
[VE]=--~
1 an (55.1)·
c at
and
VB= 0 (55.2)
are known as the first pair of Maxwell's equations, and the equations.
[VH[ = ~ j
c
+ _!__c aD
at (55.3).
202 ELECTRODYNAMICS
and
VD = 4np (55.4)
as the second pair of equations. We must note that the f1rst pair of
equations includes only the basic characteristics of a field, namely,
E and B. The second pair contains the auxiliary quantities D and H.
Maxwell's equations are the foundation of all electrodynamics.
They play the same role in classical electrodynamic.A that Newton's
equations do in classical mechanics.
The system of Maxwell's equations includes eight scalar equations
(each of the vector equations (55.1) and (55.3) is equivalent to three
scalar ones] containing twelve unknown scalar functions (the com-
ponents of the vectors E, B, D and H). Therefore, Eqs. (55.1)-(55.4)
by themselves are insufficient for determining the electromagnetic
fields in a substance. In this sense, the system of Maxwell's equations
is incomplete. To be able to analyse fields, Maxwell's equations must
be supplemented with equations relating the quantities D, j and E,
and also H and B, with one another. These equations (sometimes
called equations of state) have the form [see (45.5) and (52.12)]
D = eE. (55.5)
B = J.LH (55.6)
j = aE (55.7)
(here a is the electrical conductivity of the medium). To solve prob-
lems, we must know the characteristics of the medium, e, f.!, and cr,
which are constants in the simplest case.
We must note that Maxwell's equations can be written so as not
to contain the auxiliary quantities D and H. For this purpose, we
must replace these quantities with their values from (45.2) and (52.9)
in the second pair of equations. We invite our reader to convince
himself that after such a replacement Maxwell's equations become
[VE] = -+ , ~~ VB= 0 (55.8)
[VB] = : 10 ( j + c [VM] + ~~ ) + ~~ +, VE ~ 4n (p -- VP) (5.5. 9)

,To solve the system of equations (55.8), (55.9), we must know the
form of the functions P = P (E), M = M (B), and j = j (E).
In !he following, we shall have to do with electromagnetic fields
in homogeneous and isotropic media whose permittivity e and per-
meability 1.1. are constant. Here e and 1-1 can be put outside the signs
of the derivatives or, conversely, can be put inside these signs. Hence,
Eqs. (55.1)-(55.4) can be written as
CVE] = _ _!_ aB: VB=O (55.10)
. c iJt '
[VB}= 4nf.l.
c
J' +!t- ~
;::c iJt '
VE ~' 4n p
e
(55.11)
TIME-VARYING ELECTROMAGNETIC FIE:L.D 203
In the absence of media that can be polarized and magnetized, 8

and 11 equal unity, so that Maxwell's equations for a field in a
vacuum are as follows:
1 as
[VE] = - 7 at, VB= 0 (55.12)
[ VB]=~.+_!__~
,; J c at ' VE=4np (55.13)
Since in the Gaussian system of units H coincides with B for a field

in a vacuum, in this case we may write H instead of Bin Eqs. (55.12)
and (55.13) .
. S6. Potentials of Electromagnetic Field

In Sec. 47, taking advantage of the fact that VB = 0, we wrote
the magnetic induction as
B =[VA] (56.1)
where A is an auxiliary function known as the vector potential.
Expression (56.1) also holds for a time-varying field. In this case,
however, A must be considered as a function of not only the coordi-
nates, but also the time t, namely, A = A (r, t).
Let us substitute (56.1). into Eq. (55.1). The result is
]VE] a ['VA]= -c
1 at
= -7 1 [ V, aA
at J
This relation can be written as
[ \7, ( E + +~~)] = 0
Since the curl of the vector E +
(1/c) aA!at vanishes, this vector
can be written as the gradient of a function cp:
E+_!_ iJA = - \lcp (56.2)
t: iJt
The function cp is called the scalar potential of an electromagnetic
field. In the non-stationary case, it depends on r and on t. It can be
seen from ~(56.2) that the potentials cp and A have the same dimen-
sion.
By (56.2), we have
1 aA
E= -Vcp--- c at (5fL3)
The electric field strength in the general case is thus determined

not only by the scalar potential cp, but also by the vector potential
A. The second term in (56.3) is evidently due to electromagnetic in-
duction. For a stationary field, aA/at = 0, and formula (56.3)
transforms into (41.5).
204 ELECTRODYNAMICS
Relations (56.1) and (56.3) express the magnetic and electric

fields in terms of the vector and scalar potentials.
Let us find equations that will allow us to calculate the potentials
A and cp for a field in a homogeneous and isotropic medium with
constant e and f.!· For this purpose, we shall substitute in Eqs. (55.11)
expression (56.3) for E and the curl of A for B:
[n ["'A]]= 4np, • -~.!..... ("' )- £f-t2 ~
v' v c J c at v c:p c ot 2
v (Vc:p)+f-v ( ~~) = -'4; p

Taking into account that [V, !vAll = v (VA)- 8A [see (XI.45)1,
and V (Vcp) = 8cp, we can write the equations we have obtained as
follows:
8A-~ o2A = - 4n~-t '+V ( A+ ef-t~) (56.4)
2 2c ot c J v c at
4np 1 8
l1c:p= - - -e - - (c V
at A ) (56.5)
(in some terms of the equations we have changed the sequence of

differentiation with respect to the coordinates and time).
Equations (56.4) and (56.5) are exactly the ones we are looking
for to find A and c:p. These equations are quite complicated. The fact
that they are mutually related-each of them contains both A and
c:p-is especially unpleasant. We shall see below, however, that the
potentials can be selected so that these equations become greatly
simplified.
The potentials A and c:p are determined non-uniquely. There is
consequently a certain freedom in choosing them. Particularly, an
arbitrary constant vector may be added to A, and an arbitrary con-
stant to c:p without any change in the values of B and E. Naturally,
the potentials should be chosen in the most convenient way for the
given case. This most expedient choice of the potentials is known as
their gauging. We must note that we have already taken advantage
of the possibility of gauging in magnetostatics: in Sec. 47 we chose
A so that its divergence would be zero [see formula (47.7)1.
Let us determine the most general form of the gauge transforma-
tions, i.e. such transformations of the potentials A and c:p at which
the fields E and B remain constant. The field B = [VA] will not
change if we add to A the gradient of an arbitrary scalar function f
(the curl of the gradient is zero), i.e. pass over from A to A' equal
to A+ vt:
A~ A'= A+vt (56.6)
For the electric field E = -Vc:p- (1/c) fJA/fJt to remain unchanged
here, simultaneously with the transition (56.6) we must perform
TIME-VARYING ELECTROMAGNETIC FIELD 205
the transition
(56. 7)
where f is the same function as in (56.6). Indeed, the field E' deter-
mined by the potentials A' and cp' will in this case be
E' = - Vcp' _ _!_ iJA' = - Vcp+ _!_ \1 !1_
c iJt c iJt
1 iJA 1 iJ 1 iJA
-7Tt--c-a('Vt= -vcp-781=-E
(recall that V (8f/8t) = a (\lf)/8t]. Hence, the transformation of the
potentials by formulas (56.6) and (56. 7) does not change the value
()f B and E.
The transformations (56.6) and (56. 7) are gauge transformations
()f the most general kind. Since the fields B and E remain unchanged
in these transformations, all the equations describing these fields
must be invariant with respect to the gauge transformations. This
inv ariance is called a gauge in variance.
In practice, as we have already mentioned, the gauge that is the
most expedient in each specific case is used. Particularly, we can
choose the potentials so that the following condition will be observed:
VA+ ef.t ~=0 (56.8)

c iJt
It is called the Lorentz condition.

For a fwld in a vacuum, the Lorentz condition is
VA+_! ~=0 (56.9)

c iJt
We shall show that condition (56.8) can be satisfied by the proper

choice of the function f in formulas (56.6) and (56. 7). For this pur-
pose, we shall introduce the values of A' and cp' determined by these
formulas into Eq. (56.8):
"A+
v
11! + ~ ~-!..!::
c iJt c2
iJ 2 /
iJf!
=0
[\1 (V'/) = L\f]. Hence, we obtain an equation for finding the func-
tion f:
(56.10)
where F (r, t) = -VA- (ef.L/c) 8cpl8t is the preset function of r

and t. Introducing the function f obtained from the solution of this
equation into formulas (56.6) and (56. 7), we shall find the values of
the potentials A' and cp' satisfying condition (56.8).
206 ELECTRODYNAMICS
The gauging of potentials satisfying condition (56.8) is called

the Lorentz gauge. This gauge is in the greatest favour.
The Lorentz condition greatly restricts the set of potential values
suitable for describing a given field, but it nevertheless makes the
choice of tho potentials quite unique. Indeed, without violating
condition (56.8), we can perform the transformations
A-+ A' =A. +V¢ }
I 1 0'\jl (56.11)
<J>-+ ~c-Tt
(both sets of potentials-A, <p and A', <p' -·are assumed to satisfy
the Lorentz condition), where the function 'ljl is the solution of the
equation
(56.12)
Indeed, introducing into the left-hand side of (56.8} the primed po-
tentials from (56.11) instead of A and <p, we obtain the expression
VA + .1,,,
'I'
+ .!!::.
c
~
8t
- .!!::.
c2
82 '~'
8t2
= (VA+~
c
~
8t
)+ ( .1•••-
'I'
~
c2
82 ~) )
8t 2 .
which owing to (56.8) and (56.12) equals zero. Therefore, if the po-
tentials A and <p belong to the Lorentz gauge, the potentials A' and <p'
determined by transformations (56.11) [with 'ljl obeying (56.12)1 belong
to the same gauge. This allows us to impose on the potentials an addi-
tional condition besides condition (56.8). For instance, we can re-
quire that the potential <p vanish. For this purpose, according, to the
second of Eqs. (56.11), it is sufficient to choose the function 'ljl so that
its time derivative will be c<p.
We can also adopt the requirement that
VA= 0 (56.13)
aR an additional condition. It follows from (56.11) that VA' ~
-:- VA + .1'\jl. Therefore, for the requirement that VA' = 0 to be
satisfied, the equality
.1'\jl =-VA
must be observed. At the same time, from (56.12), we have
.1 '"
'1'
= .!!::...
c2 8t 2
82'\l
Therefore,· if we take- as 'ljl the solution of the equation

~ 82¢ =-VA
c2 8t 2
and introduce this solution into (56.11), we obtain values of A'

and cp' satisfying both the Lorentz condition~and the requirement
(56.13).
The Lorentz gauge satisfying the additional condition (56.13)
is called the Coulomb (or transverse) gauge. This is the gauge we
used in Chap. IX [see (47. 7)]. Examination of (56.5) shows that in
the Coulomb gauge the scalar potential sati~;fies Poisson's equation
.1cp=- 41tp
e
[see Eq. (45.13)], i.e. is a Coulomb potential. This explains the ori•
gin of the name "Coulomb gauge".
57. D'Aiembert's Equation

When condition (56.8) is satisfied, the last term in Eq. (56.4) van-
r
ishes. In addition, the time derivative of VA has the value
-(e~-tlc) fJ2ip/ot 2 • Consequently, Eqs. (56.4) andf (56.5)~ become
(57.1)
(57.2)
Hence, instead of two mutually related equations, we have ob-

tained two independent equations, the equations for A and cp having
acquired a similar form.
A differential equation of the kind
.1!-~
c2 ~=
iJt2 F (r, t) (57 .3)
is known as d'Alembert's equation [compare with (56.10)]. It can
be written in a very compact form if we introduce the d'Alembertian
operator
e!J. iJ2 iJ2 iJZ iJZ Bf.L iJS
0 = t:,. - cz ifi2 = iJx2 +
iJy2 +
iJz2 - "'C2 ifi2 (57 .4)
Now Eq. (57.3) becomes
Of = F (r, t) (57.5)
In the stationary case, the time derivatives vanish, and D' Alem-
bert's equation transforms into Poisson's one [see Eqs. (45.13) and
(52.18)].
When we use the symbol (57.4), Eqs. (57.1) and (57.2) become
4l'tf.L.
DA = - - c -) (57 .6)
41tp
Ocp= - - e - (57. 7)
'208 ELECTRODYNAMICS
From a mathematical viewpoint, Eqs. (57.6) and (57.7) are simpl-

er than Maxwell's equations. It is therefore simpler to calculate the
potentials A and <p than the fields B and E directly. When the poten-
tials are known, there is no great difficulty in finding the fields by
formulas (56.1) and (56.3). This circumstance justifies the introduc-
tion of the auxiliary quantities A and <p. In addition, as we shall
see in the following chapter, the introduction of the potentials allows
us to give the equations of electrodynamics a very compact and re-
fined form.
58. Density and Flux of Electromagnetic

Field Energy
Experiments show that an electromagnetic field has an energy

that propagates in space with a certain density w and can flow from
one place to another. This energy can also transform into other kinds
of energy, for example, it can go to do work on the particles of a
substance. ·
Let us separate in a substance in which there is a macroscopic
electromagnetic field the volume V confined by the surface f. This
volume contains energy of the field equal to
W= \ wdV (58.1)
v
The energy can flow outward from the volume V through the sur-
face f confining it. If we introduce the vector S of the energy flux
density, the flux of energy flowing outward through f can be writ-
ten as
<D=) Sndf=) V'SdV ,(58.2)
I V
(we have used the Ostrogradsky-Gauss theorem).
Let us find the work done in unit time bv the field forces on the
particles of a substance. The forces of a field do the following work
on a paTticle having the charge ea and travelling at the velocity Va
in unit time:
Na = ea {E +f fvaBJ} Va = eavaE
(the scalar triple product Va [ vaBl vanishes). Summating this expres-
sion over all the particles confined in unit volume, we obtain the
density of the power developed by the forces of the electromagnetic
field in doing work on the particles of the substance. Designating
the power density by N, we can write
N =-~ ~~ N a = E .~· ea V a
V=1
The sum ~eava taken over unit volume is the density j ofthe elec-
tric current (if all the particles are identical, travel at the same ve-
locity, and their number in unit volume is n, the sum ~eava trans-
forms into the expression j = env known from the general course
of physics). Hence,
N = Ej (58.3)
The energy W contained in the volume V may diminish because
of energy flowing outward through the surface f and because of work
being done on the particles of the substance. Consequently, the fol-
lowing relation must be observed:
dW= -<1>-
-
dt
i
v
NdV
(remember that N is the power density, i.e. power developed per unit
volume). Substitution into the above equation of expressions (58.1)-
(58.3) for W, <l>, and N yields
~ I w dV = -
v
) vs dV- )
v v
Ej dV
Let us exchange the places of time differentiation and integration

over the coordinates in the first integral on the right-hand side, and
also combine all three integrals into one. The result is
) ( ~~ +VS+Ej)~dV=O (58.4)
v
(we have used the symbol of the partial derivative because w in the
general case is a function not only of time, but also of the coordi-
nates). ·
Condition (58.4) must be observed for any arbitrarily chosen vol-
ume V. We thus conclude that the integrand function must vanish
at every point. Consequently, we arrive at a differential equation
that can be written as
Ei= -~-vs
IJt
(58.5)
The energy density w and the energy flux density S are functions
of quantities characterizing a field. To find the form of these func-
tions, let us; attempt to transform the expression for Ej so that it
would become the sum of two addends, one of which would be the
time derivative of a scalar quantity, which we would be able to
identify with w, and the second-the divergence of a vector quantity,
which we would be able to identify with S. · ·
14-018
210 ELECTRODYNAMICS
Using Maxwell's equation (55.3), let us express j in terms of the

field characteristics H and D:
• c 1 ao
J = 4n [VH] -"'4nat
Now let us find the scalar product of this expression and E:
Ej =-c-E [VH]--1-E ~ (58.6)

4n: 4n: at
By formula (XI.28), we have V [EH] = H [VEl- E [VHl, whence
E [VH] = H [VE] - V [EH]. Introducing this value into (58.6),
we obtain
Ej = __c V [EH] + {-c-H
4n: 4n:
1 E
[VE] - -
4rt
aD}
iJt
(58.7)
Using Maxwell's equation (55.1), let us substitute -(1/c) a Blat for

[VEl. As a result, the expression in braces becomes
{... }= - 41n: { H ~~ + E ~~ }
Finally, let us use relations (55.5) and (55.6):
{• • •}= _ __!_{H
4n:
8(~-tH)
at
+E
_,
B(eE) }=-~( eE2
at at Bn
+ 11snn2)
(we have assumed that e and f..t are time-independent).
Formula (58. 7) can thus be written as follows:
E' = _.!..._ ( eE2+!1H2) _ V

J at sn
{-c
4n
[EHJ}
Acomparison of the obtained relation with formula (58.5) gives for

w and S the expressions
eE2+!1H2
W= 8:rt (58.8)
S= :n [EH] (58.9)
The vector S determined by formula (58.9) is called Poynting's

vector.
We must note that expression (58.8) includes both the proper ener-
gy of the field equal to
and the energy spent on polarization and magnetization of the medi•

urn in producing the field.
TIME-VARYING ELECTROMAGNETIC FIELD 211.
59. Momentum of Electromagnetic Field

It follows from the existence of the pressure of light that an electro·
magnetic field has not only energy, but also momentum. The mo·
mentum, like the energy, can ·'flow" from one place to another.
This process can be characterized by introducing the concepts of the
flux and flux density of the momentum.
The momentum flux, unlike the energy flux, is a vector, and not a
scalar. Consequently, the density of the momentum flux must be a
quantity such that when multiplied by the vector df (here dj is an
element of the surface) it yields a vector. It is shown in Appendix X
that the scalar product of a second-rank tensor and a vector is a
vector. We thus conclude that the density of the momentum flux is
a tensor. Let us designate the components of this tensor by the sym-
bol uik· Therefore, the flux of the i-th component of the momentum
through the area df is determined by the expression
2juik dfk
h
and the flux of the momentum vector by

~ e1 ~ u,k dfk
i k
In a system consisting of free charged particles and an electromag-
netic field, the total momentum, which is the sum of that of the
particles and that of the field, must be conserved. Hence, denoting
the total momentum of the particles by the symbol P, and the den·
sity of the momentum (i.e. the momentum of unit volume) of an
electromagnetic field by the symbol g, we can write
~ (P + Jg dV ) = ~ e 1 ~ ~ U1 k df k
V i I k
The left-hand side gives the rate of growth of the total momentum
contained in the volume V, and the right-hand side-the flux of the
momentum of the field flowing into the volume V throughhhe sur-
face f confining it. We assume that no particles intersect this sur-
face so that no momentum is carried across it by the particles.
Let us write the above relation as follows:
~P=-
dt .Ii ~dV
iJt
+~e. A!.~ a,k dfk
~ ' 'j' ~
(59.1)
v i f k
The rate of the change in the momentum of a particle is deter·

mined by the force acting on a particle:
dp e
-=eE+-[vB]
dt c
212 ELECTRODYNAMICS
Summation of this expression over the particles contained in a unit

volume yields
~o = pE + ! [jB]
where p 0 is the density of the momentum of the particles. Finally,

integrating this expression over the entire volume of the system,
we find the rate of the change in the total momentum of the par-
ticles:
:t P Jv pE dV + +- Jv [jB] dV
= (59.2)
Let us use Maxwell's equations to exclude p and j from this expres-

sion. Assume that the medium containing the particles and the field
are homogeneous and isotropic with constant e and 1-t· It can be seen
from Eqs. (55.11) that
e "E . c ["B] e aE
p = 4it y , J = 4:rtf.t y - 4i1 at
Substitution of these values into (59.2) yields
d P=
dt :n J EVE dV +
\ 1
4:rtf.t
\
J [[VB], B] dV- 4 :c Jf r· at, BJdV
BE
(59.3)
We shall transform Eq. (59.3) using the relation
! .. J
aE , B J + [ E, Tt
a [EB] = [ Tt
Tt aB J
whence
l~
at ' B] = _!__
ot [EB]- [E ' ~]
at
We substitute -c [VE] for a Blat here according to the first of the
equations (55.10). The result is
aE B
[ Tt, l =ata [EB] + c [E, [VE]J
Using the value obtained in (59.3), we arrive at the expression
4t'P= ~ JEVEdV+ 4!f.t J[[VB], B]dV

- 4 ~c J:e [EB] dV- 4: J[E, [VE]J dV
Let us make the substitution B = f.l.H in the second and third inte-
grals and, in addition, exchange the places of the multipliers in the
second integral (which will cause the sign to be reYersed):

:t P= J 4: EVEdV- 2~ J[H, [VH]]dV
-~ (' _!_[EH] dV __e_
4llc J at 4ll Jf [E
' [vEJ] dV
To make this expression symmetric with respect to E and H, let
J
us add (f.l/4n) H VH dV to its right-hand side. This will )not change
the expression because this term is zero [f.lVH = V (f.lH) = VB is
zero everywhere]. Finally, grouping the terms properly, we obtain
:t P= - Z~c J ~ [EH] dV
+ 4~ J{eEVE+f.lHVH-e[E, [VEJl-f.l[H, [VHJJ}dV (59.4)
The second integral, as we shall see below, can be transformed into
an integral over the surface f confining the volume V. Hence,
:t P= - .\ ~ { :~c
v
[EH]} dV +integral over surface f (59.5)
A comparison of the found relation with (59.1) allows us to draw

the conclusion that the density of the momentum of an electromag-
netic field is determined by the expression
g = Z~c [EH] (59.6)
Taking (58. 9) into account, we can write that

ef-t
g =-S
• c2
(59. 7)
where S is the Poynting vector. For a vacuum. this relation is

1
g=-2
c
s (59.8)
We must note that expression (59. 7), in addition to the proper mo-
mentum of a field, includes the momentum of the bound charges en-
tering the composition of the medium in which the field has been
produced. To obtain the momentum of the field alone, we must un-
derstand P in formula (59.1) to be the mechanical momentum not
only of the free charges, but also of the bound ones. Now we would
have to take not the averaged macroscopic field, but the microscopic
field as E and B and correspondingly use Maxwell's equations for
a field in a vacuum in the transformations 1• This is equivalent to
1 We did not do this from the very beginning to obtain a more general expres-
sion for the Maxwell stress tensor (see below) that is also suitable for the field
in a medium.
214 ELECTRODYNAMICS
assuming in all the formulas of the present section that e = f.t = 1.

As a result, we would arrive at formula (59.8). Consequently, the
density of the momentum of only a field in all cases (both in a vac-
uum and in a substance) is determined by formula (59.8). A com-
parison of this formula with formula (40.36) shows that there is
exactly the same relation between the densities of the energy and
momentum fluxes for an electromagnetic field as we obtained in
Sec. 40 for an arbitrary system.
Now let us consider the second integral in formula (59.4), i.e. the
integral
in ~ {eEV'E+f.tHV'H-e[E, [Y'Ell-f.t[H, [V'H]l}dV (59.9)
We shall attempt to transform it into a surface integral.

The integrand contains two similar expressions of the kind
aV'a - [a, [V'a])
One of them contains E instead of a, and the other H.
Assuming in the formula
V' (ab) = [a, [V'bll + [b, [V'a]] + (aV') b + (bY') a
[see (XI.37)1 that b = a, we obtain
V'a 2 = 2 [a, [V'all + 2 (aV') a (59.10)
Let us find the value of the expression (V'a) b in which it is as-
sumed that V' acts on both factors following it. According to the general
rule for calculating such expressions, we have
(V'a) b = (Y' a a) b + (Y'ba) b = bV'a (aV') b
or, assuming that b = a,
(V'a) a = aV'a + (aV') a
After finding (aV') a from the last expression and introducing its
value into formula (59.10), we obtain
V'a2 = 2 [a, [V'all + 2 (V'a) a - 2aV'a
whence
1
aV'a- [a, (V'a]] = (Y'a) a- 2 V'az
Applying such a transformation to the integral (59.9), we can
write it as follows:
4~ ) { e (V'E) _E + f.t (V'H) H + ~ V' (eEz + f.tHz)} dV

In this integral, the operator V' acts on all the functions following it.
Consequently, with the aid of the transformation
dV·V'-. df
TDIE-YARYING ELECTROMAGNETIC FIELD 215
[see (X1.65)l, we can transform this integral into a surface one:
4~ ~ { eE (E df) + ~tH (H df)- + (e£2 + ~H 2 ) df} (59.11)

{since df does not have the properties of a differential operator, we
can exchange the places of the factors in expression!' such as (df a) al.
Let us develop expression (50.11) using the components of the
vectors in it. The result is
· 4~ ~e~~{eE 1 ~Ekdf~t+~H 1 ~H~tdf~t-~ (eE 2 +~tH 2 )d/ 1 }

i k k
Let us write dfi in the last term as ~<Silt dfk and factor_out df-A. Let
h.
us also substitute D 1 for expressions of the kind eE 1 , and Bt for
~H 1 • As a result, we arrive at the expression
4~ ~e 1 ~ ~ {E 1 D~t+H 1 B~t--}(ED+HB)8 1 ~t}df~t

t k
that coincides with the last term of formula (59.1) if we assume that
O'ih. 1 { EPh. +
=-:rrt HiBh. IW (ED+HB) 8ik }
-~2. (59.12)
We established at the beginning of this section that the tensor a 111.

characterizes the density of the momentum flux (see also Sec. 40).
The tensor an, who!::'e components are determined by formula (59.12)
is called the Maxwell stress tensor.
To underline the symmetrical nature of the tensor a 171., its compo-
nents are sometimes written as
1
a 11~= Bn {E 1 D~t+E~tD 1 +H 1 B~t+H~tB 1 -(ED+HB)8 1 ~t}
For a field in a vacuum, formula (59.12) is simplified as follows:

atk= 41n{EtEk+BtBk- ~ (E 2 +B 2)6th} . (59.13)
The tensor O'th. allows us to reduce the problem of finding the force
acting on a certain volume of a substance in an electromagnetic
field to the calculation of the surface integral
~ ei~ ~aih.dfk
i I h.
Chapter XI
EQUATIONS OF ELECTRODYNAMICS
IN THE FOUR-DIMENSIONAL
FORM
60. Four-Potential•
According to the principle of relativity, the equations of.electro-

dynamics, like all other equations expressing nature's laws, must be
relativistically invariant, i.e. retain their form in Lorentz trans-
formations (when passing from one inertial reference frame to an-
other). Direct verification shows that Maxwell's equations meet this
requirement. We shall choose a different way, however-we shall
show that the equations of electrodynamics can be written in the
four-dimensional form as relations between four-vectors and four-
tensors, whence their relativistic invariance will follow.
Our starting point will be the thesis adopted in electrodynamics
(in accordance with experiments) that the electric charge is invariant,
i.e. that the magnitude of the charge of a particle is the same in all
inertial reference frames. It thus follows that the quantity p dV is
also invariant:
p dV = p dx1 dx2 d:J:3 = inv (60.1)
We know that the three-dimensional volume dV is not invariant
[see formula (35.17)]. We thus conclude from (60.1) that the charge
density p is also not invariant, but changes in a transition from one
reference frame to another according to a definite law. To establish
this law, we shall take into consideration that a four-dimensional
volume is invariant:
dV* = dx0 d,x1 dx2 dx3 = c dt dV = inv (60.2)
Indeed, in passing over to another reference frame, dt and dV
transform by the formulas
dt' = V dt 1 , dV' = dV Y 1- v2/c2
1-v2 /c 21
so that dt' dV' = dt dV.
1 In this chapter, we treat fields in a vacuum, i.e. we assume that e = 1 and

fl = 1.
We advise our reader to look through Appendix XII and Chap. VII before
beginning to read this chapter.
EQUATIONS OF ELECTRODYNAMICS IN FOUR-DIMENSIONAL FORM 217
A comparison of (60.1) and (60.2) shows that p transforms accord-

ing to the same law as dx 0 , i.e. like the time component of a four-
vector.
It is shown in Appendix XII that the contravariant components
of the Hamiltonian four-operator are
V* 11 = (i..~
c at '
_n)
v
[see formula (XII.45)l.

By (XII.38), the symbolic square of the vector v* is
1
V*2----V2-
a2 1 a?
- c2 at2 - - --11
c2 !Jat2
(60.3)
where 11 is the three-dimensional Laplacian operator.

A comparison of (60.3) and (57.4) shows that the ,d' Alembertian
operator for a field in a vacuum (e = 1, f.L = 1) differs from vu
only in its sign:
(60.4)
It follows from everything said above that Eq. (57. 7) can be writ-
ten as
(60.5)
We have ascertained above that p has the properties of the time
component of a four-vector, and V* 2 , like the square of any four-
vector, behaves in Lorentz transformations like an invariant. Tak-
ing this into account, we conclude on the basis of (60.5) that the
potential c:p must transform according to the same law as p, i.e. as a
four-vector time component.
Consider the current density vector j = pv. Its components are
(60.6)
In transformations, p behaves like ct or c dt. Consequently, i11. will

behave like dx" (where k = 1, 2, 3), i.e. like space components of
a four-vector.
Therefore, in transformations of the coordinates, p behaves like
a four-vector time component, and the quantities h like four-vector
space components. Hence, multiplying p by the scalar c (to obtain a
quantity of the same dimension as h has), we can combine p and j
into a single four-vector called the charge-current or simply the cur•
rent four-vector. Its components are
•3 •
l = l:r.
which can be written briefly as follows:
jll = (cp, j) (60. 7)
218 ELECTRODYNAMICS
We must note that the component j 0 = cp can~be written like the

components (60.6) as
·o dxO
] =pdt
(x0 = ct). Therefore, the components of the current four-vector can

be determined as follows:
(60.8)
The continuity equation following from charge conservation is

[see (51.1)]
Vj+ :~ =0
With a view to (60.7), this equation can be written in the four-di-
mensional form:
(60.9)
The left-hand side of relation (60. 9) is the four-dimensional diver-

gence of the current four-vector. Indeed, by analogy with the three-
dimensional divergence, the four-divergence of the vector al-l must
be determined as the scalar product of the vectors V* and all, i.e. as
[3 3 3 I
:2: V*~-taJ.l = :2: V~a~-t = :2: aa:l
J.l=O; J.l=OI J.l=O ax,'),
{see expression (XI1.42) for the covariant components of the vec-

tor V*].
That the four-divergence of the vector jll equals zero is an analyt-
ical expression of the law of charge conservation. The fact that the
vector .(60. 7) which we have introduced satisfies such a simple con-
dition is another argument in favour of combining p and j into a
single four-vector.
·Having in view that 0 = -V* 2 , let us write Eqs. (57.6) and
(57. 7) as follows:
n*2A-
v 4n •
- -c-J (60.10)
4n
'V*2cp = - (cp) (60.11)
c
Examination of Eq. (60.10) shows that the quantities Ak behave in

the same way as the quantities jh, i.e. like space components of a
-four-vector. This circumstance allows us to combine cp and A into a
single four-vector:
AJJ. = (cp, A) (60.12)
known as the four-potential of an electromagnetic field. Equations
(60.10) and (60.11) can therefore be written as a single equation:
(60.13)
(60.14)
The covariant components of the four-potential are as follows:

AJ.l. = (cp, -A) (60.15)
We remind our reader that the four-potential is determined non-
uniquely. By (47.6), the values of the space components A11. can be
replaced with the quantities
A =A + 8\jl(xl, x2, x3) (60.16)
1!. 1!. oxl!.
and the value of the component A 0 with the quantity

A0 = A0 +C (60.17)
{C is a constant; see the text following formula (41.5)1 without

changing the field characteristics B and E.
The Lorentz gauge condition [see (56.8)1 in the four-dimensional
form is
(60.18)
This signifies that the four-potential is chosen so that its four-diver-

gence vanishes [compare with (47. 7)1.
61. Electromagnetic Field Tensor

Let us go over from potentials to the force characteristics E and B
of a field. This transition is performed by formulas (56.1) and (56.3).
For convenience, we shall repeat these formulas:
B =~[\7 A) (61.1)
1 aA r
E= -vm---.
T C fJt!
(61.2)
220 ELECTRODYNAMICS
Let us write the expressions for the components of the vector E:

E -· Bcp 1 OAx
x - · - ox -cat
1 OAy
EY = -Bcp
By- -c -Bt-
E = _ Bcp _ _!_ Mz
z az c Bt
In four-dimensional symbols, these formulas can be written as
Ex= ~~; - ::~ = V~At- VfAo )
E y= oA2 BAo *A *A
Bxo -~=Vo z-V2 o
l
1r
(61.3)
E% = BA 3
BxO -
BA 0
Bx8 = v*A *A
0 3 - v3 0 J
[see formulas (XII.42) for the covariant components of the gradient

four-operator and formulas (60.15) for the covariant components of
the four-potential. By the last formulas, for instance, Ax = -A 1 ,
etc.].
Now let us write the expressions for the components of the vector
B. According to (61.1)
oAz
B =- oAy
----
BA 2 oA 8 *A *A )
X {)y 0'1.= - - ---=
{)z3 qz2
V a z-V2 3
B = BAx- oA: = oAa -BAr =V*A -"V*A

I
y i)z ox oxl (Jx8 1 3 3 1 (61.4)
B
z
=BAy_ BAx
ax {}y
= BA 1
fJ.7;2
_ BA2
{}xl
=V*A -V*A
2 1 1 2
Jf
It is known from tensor algebra that expressions of the kind al!-b" -
- a"b~'- are covariant components of an antisymmetric second-rank
tensor (a~' and bv are covariant components of arbitrary vectors).
It follow&, from formulas (61.3) and (61.4) that the components of
the vectors E and B can be interpreted as covariant components of
the antisymmetric four-tensor
. F JJ.v=VIl*A v-Vv*A IJ.= OAv OA11
Bxll- Bx"
(61 . 5)
The latter is called the e]ectromagnetic fieJd tensor.

Raising the indices f..l. and v on both sides of (61.5) and lowering
them on the right side, we get an expression for the contravariant
components of the electromagnetic field tensor:
(61.6)
A comparison of formulas (61.3) and (61.4) with expression (61.5)

shows that
E,; = F 01 = -F10 , Ey = F 02 = -F 20 , Ez = F 03 = -F 30 }
(61. 7)
Bx = F 32 = -F 23, By = F 13 = -F 31 , Bz = F 21 = -F12
The tensor F 11 v can thus be written as follows:
(61.8)
Taking into account formulas (XII.58), let us write the values of

the contravariant components of the electromagnetic field tensor:
(61.9)
Hence, in four-dimensional space, an electromagnetic field is

described with the aid of one antisymmetric second-rank tensor in-
stead of two vectors (E and B).
We must note that whereas the four-potential All is determined
non-uniquely, the components of the tensor F 11 v are unique. Indeed,
let us replace the components A 11 in (61.5) with the quantities A11
according to formulas (60.16) and (60.17). If fl and v equal 1, 2,
and 3, we obtain
If one of the indices (for instance fl) is zero, we get the relation
81\J (x 1, x2, x3) )
ax"
_ 8A 0 _ _!!!__ _ 8Av _ 8Ao F
ax'~~ axv - axo axv 0'11
We have thus proved that the quantities F 11 v are determined

uniquely.
222 ELECTRODYNAMICS
62. Field Transformation Formulas

Formulas for transforming the components of an antisymmetric
four-tensor are established in Appendix XII [see formulas (XII.62)1.
Substitution of the values (35.8) and (35. 9) for a 0 and a 1 results in
these formulas becoming
A'01=A01 A'OZ= A02_~Al2 A'03= A03-~Al3 }
' V1-~ 2 ' ¥1-~2
A t2_~A02 Al3 ~A 03 (62.1}
A'12 = .... A'13 = -.... A'23 = AZ3
v1- ~2 ' v1- ~2 '
(here ~ = v0 /c).
Writing formulas (62.1) for the tensor (61. 9), we obtain
p•ot =pot, i.e.
F'OZ = po2_ ~pt2
V 1-~2 ' i.e.
etc. Writing out relations (62.1) for all the components of the tensor
p11v and replacing p11v with the relevant values of Ek' and Bk, w&
arrive at formulas for transforming the components of the vectors
E and B in a transition from one inertial reference frame to another:
Eu-~Bz Ez+~By
E~=E:#, E~= E;=
V1-~2 ' v1-~ 2
B~ = B,;a B~=
B 11 +~Ez
v 1-~2 ,
B'- Bz-~Ey
~- 'Vt-~2
) (62.2)
r
The formulas for the inverse transformation differ from • these for-
mulas only in the sign of the terms containing the factor ~ (i.e. in
the sign of v0).
Resolving the vectors E and B into components parallel to th&
x-axis (And, consequently, to the vector v 0 ) and components perpen-
dicular to this axis (i.e. writing, for example, E as E 11 + E.L), we
can write formulas (62.2) in the vector form 1 :
, E .L +(1/c) [v 0 B .L]
Ell = E n, E .L = --=-:-,r1r.;====:=:::::=-
1-vBfc2
Bl1
, B.L -(1/c)[v0E.L]
1 ===;~=
= B n, B.L = ---==-:,r=:=
r 1-vij/c2
) (62.3)
We must note that since B 1 and v0 are collinear, f[v 0 B] = [v0 B11 ]
1 +
+ [v0 B.L]=[v0 B.L].Similarly, [v 0E] = [v0 E.d. For this reason in the vector
products of formulas (62.3), the subscript ".i" at B.L and E.L may be discarded~
(recall that Vx = v 0, and Vy = v, = 0). A glance at these formulas

shows that the longitudinaP components of the fields in a transition
from one reference frame to another do not change, and only the
lateral components are transformed.
If ~ « 1 (i.e. v0 «c), the expression 11V 1 - ~ 2 approximately
equals 1+{- ~ 2 • Consequently, within terms of the order of ~ =
= v 0 le, formulas (62.2) can be written as fol1ows:
E~ = Ex, E~ = Ey - (v 0 le) B, E~ = E: + (v 01e) By} (62.4)
B~ = Bx, B~ = By + (vofc) Ez, B; = Bz - (v 0 1e) Ey
It is not difficult to see that these formulas can be written in the
vector form:
(62.5)
[compare with formulas (62.3'].
If there is only an electric field in the frame K (i.e. E =I= 0 and
B = 0), both fields exist in the frame K'. By formulas (62.5), these
fields are
E'=E, B' = - _!_ [v 0 E]
c
Having in view that E = E', we can write that
B' = -'-_!_
c
[v 0 E'] (62.6)
Relation (62.6) 2 indicates that the fields B' and E' are mutually
perpendicular. It can also be seen from this relation that the field
B' is perpendicular to the vector v 0 , i.e. to the x-axis.
It can be shown similarly that when there is only a magnetic
field in the frame K, the vectors B' and E' are related by the expres-
sion
E' = _!_ [v 0B'] (62.7)
c
Consequently, in this case too the fields B' and E' are mutually
perpendicular. In addition, E' is perpendicular to v 0 , i.e. to the
x-axis.
Hence, if only one of the fields (E or B) exists in the frame K.
in any other frame K' the fields B' and E' are mutually perpendic-
ular. The opposite conclusion also holds: if the fields B and E
are mutually perpendicular in a frame K (and the magnitudes of
1 For simplicity, we shall use this name for the components parallel to the
vector of the relative velocity of the frames K and K' (to the vector v0 ). We
shall call the perpendicular components lateral ones.
2 We invite our reader to convince himself that the same relation between
the vector B' and E' can also be obtained in the case being considered from the
accurate formulas (62.2).
224 ELECTRODYNAMICS
the fields satisfy the conditions indicated below), frames K' exist
in which the field is purely electric, and also frames in which the
field is purely magnetic. Let us find the velocities of the relevant
systems beginning with a treatment of the case when v0 «c.
Assume that the vectors B and E are mutually perpendicular.
It follows from formulas (62.5) that for the field in the frame K'
to be purely electric (i.e. forB' to vanish), the vector v 0 must satisfy
the condition
1
B=-
c
[v 0E]
This condition will be observed if the vector v 0 is perpendicular

to B (the vector E is perpendicular to B according to our assump-
tion) and, in addition, v 0 E sin a = cB, where a is the angle be-
tween the vectors v 0 and E. Hence, the field is purely electric in all
frames travelling in directions perpendicular to B provided that
the velocity v0 of the given frame is cBI E sin a. Since the velocity
v 0 of the system cannot exceed c, the reference frames being consid-
ered exist only provided that cB < E sin a. If E ~ cB, this
condition is not observed at any angle a. Therefore, in this case,
notwithstanding the mutual perpendicularity of B and E, no frames
exist in which the field is purely electric.
It is not difficult to see that the result we have obtained is true
without stipulating that v 0 « c. For this purpose, let us turn to
formulas (62.3). Assume that B and E are mutually perpendicular.
Let us take the frame K' whose velocity v 0 is perpendicular to B
and equal in magnitude to eEl E sin a (here a is the angle between
the vectors v 0 and E). Since the vectors v 0 and B are mutually
perpendicular, the component B11 vanishes. By formulas (62.3),
Bj1 also vanishes. Let us consider the numerator of formula (62.3)
for B~. With the direction of v 0 we have chosen, the vector B.h
equals B. The vector product of the vectors v 0 and E can be written
as
[v 0 E] = [v 0Eul + [v 0 E.l] = [v 0 E.ll
(the first term vanishes because the vectors v 0 and E 11 are collinear).
Consequently, the formula for B~ for the case being considered can
be written as follows:
B ,_ _
l-
B- (1/c) [voE)
(62.8)
V1-v~/c2
According to our condition, B is perpendicular to both E and v 0 •

Therefore, the vector [v 0 El is collinear to the vector B. By properly
choosing the direction of the vector v 0 (to the right or the left),
we can make the vectors [v 0 E] and B have the same direction.
The numerator of formula (62.8) thus contains the difference of
two identically directed vectors whose magnitudes are B and
EQUATIONS OF ELECTRODYNAMICS-IN FOUR-DIMENSIO~AL FORM 225
(1/c) v 0 E sin a. If v 0 = cBIE sin a, the magnitudes of these vectors

will be the same, and the numerator in the formula for B~ will
vanish. Hence, both Bj 1 and B~ are absent in this case. We can show
in a similar way that the field is purely magnetic with mutual
perpendicularity of the fields B and E in the frames [(' travelling
in directions pt>rpcndicnlnr to the vceto1· E at a speed v 0 equal to
cEIB sin a (here a is the angle between the veetors v 0 and B). This
statement holds provided that cE < B sin a. If B ~ cE, no systems
exist in which the field is purely magnetic.
63. Field Invariants
Let us form the expression

B '2 - E'2 -..::....;
_ '\1 B'2 ,-.., E'2
k-.L! _ "\~ (B'2
k-.:....; k - E'2)
k
and substitute for ihe quantities Bit and Elt in it their expressions
in terms of the unprimed components [see (62.2)1:
B' 2 -E' 2 = ~ (Bit2 -K,n=Bi-E'i
(By+ ~Ez) 2 -(Ey-BBz) 2 + (Bz-~Ey) 2 -(Ez + BBy) 2
+ 1-B2 -
It is a simple matter to see that the right-hand side is reduced to

the form
(Bi- Ei:) + (B~- E;) + (B~- ED
We thus arrive at the conclusion that the difference between the
squares of the vectors B and E has the same value in all inertial
reference frames, i.e. is an invariant:
B2 - E 2 = inv (63.1)
Now let 11s form the scalar product of the vectors E' and B', i.e. the
sum ~ E/,15],. In trod ncing into this snm instead of Eli and Rlt their
values from (62.2), we obtain
"
.L..J
E'B' _ E B
h k- X X
+ (Ey-BBz) (Bu+BEzl+(Ez+BBv)
1- ~2
(Bz-BBu) _ " E
.L..J - h
B
k
Consequently, the scalar product of the vectors E and B is also

an invariant:
EB = inv (63.2)
Inspection of (63.2) shows that when the fields Band E are mutual-
ly perpendicular (i.e: EB = 0) in a certain reference frame, they
are mutuallyl perpendicular in any other inertial reference frame.
15-018
226 ELECTRODYNAMICS
Inspection of (63.1) shows that when the magnitudes of the vec-

tors B and E are the same (i.e. B 2 - E 2 = 0) in a certain reference
frame, they are the same in any other inertial reference frame.
In addition, the following conclusions can be drawn from the
invariance of expressions (63.1) and (63.2). When the vectors B
and E form an acute (or obtuse) angle (i.e. EB is greater or less
than zero) in a certain reference frame, they form an acute (or obtuse)
angle in any other frame. If B > E (or B < E) (i.e. B 2 - E 2 is
greater or less than zero) in a certain frame, the same relation will
be retained between the magnitudes of the vectors B and E in any
other frame.
When both invariants eqnal zero, the vectors B and E in all
inertial reference frames are mutually perpendicular and equal in
magnitude.
If only the invariant (63.2) equals zero, i.e. EB = 0, it is pos-
sible to fmd a reference frame in which either the field B or E is
zero depending on the value of the expression B2 - E2 • The opposite
is also true: if one of the fields BorE is zero in a certain frame, the
fields in any other frame will be mutually perpendicular (we already
arrived at this conclusion earlier when analysing the formulas for
field transformation).
One must have in view that the fields Band E, generally speaking,
vary from point to point. Therefore, the invariants (63.1) and (63.2)
may have different values at different points. The above statements
on the properties of fields relate to their points for which the assump-
tions we have adopted are observed (for instance, the equality to
zero of a given invariant, etc.). If these assumptions are observed
at all the points of a field, the statements on the properties of fields
will also naturally relate to all the points.
With a view to the above remarks. let us assume that at a 'Point
of a field in the frame K, the product EB is non-zero, i.e. that the
fields at the given point are not perpendicular to each other. We can
now find such a reference frame K' in which the fields at a given
point are parallel to each other. In this frame, E'B' = E' ll', so
that we obtain two equations:
B' 2 - E' 2 -'"" B2 - E2, E' B' = EB
Solving these equations simultaneously, we find the values of the

quantities E' and B' in the reference frame in which the fields E'
and B' are parallel (the vectors E and B are set).
The invariants of a field can be found proceeding from the general
properties of tensors. It is shown in Appendix X that the multipli-
cation of tensors of ranks m and n yields a tensor of rank (m n), +
and also that the contraction of a tensor over any pair of indices
lowers t.he rank of the tensor by two. Particularly, the contraction
EQUATIONS OF ELECTRODYNAMICS IN FOUR-Dii\IENSIONAL FORM 227
of a second-rank tensor, equal to the sum of its diagonal components,

is called the trace of the tensor and is an invariant [see formulas
(X.21) and (XII.63)l.
For an antisymmetric tensor like the electromagnetic field tensor,
the trace is zero so that this invariant is of no interest.
Let us form the produel of the tensors (o1.8) nncl (61.!l). i.e. a ten-
sor with tho components F 11 vFP 0 • It is a tensor of tho fourth rank.
We shall perform a double contraction of this tensor. assuming the
indices f.! and p, and also v and cr. to be equal and summating over
these indices. As a result, the rank of the tensor will lower by four
(each of the two contractions lowers the rank by two). and we obtain
a zcr·o-rank tensor, i.e. <Ill invariant:
'\:, F I!V p1LV = InV
.LJ 0
J.L, v
Introduction of the values of F J.LV and FJ.Lv [see (61.8) and (61.9)1
yields
3 3 3
2j FJ.LvpJ.Lv=-2 )', E~+2 ~ B~=2(B 2 -E 2 )=inv (63.3)
J.L, v=O k=1 k=1
which agrees with formula (63.1).
Now let us form a four-tensor of the eighth rank
eJ.LVpcr F r:r.flF .., 0 (63.4)
where cJ.Lvpcr is an absolutely antisymmetric unit four-pseudotensor
of the fourth rank [see Appendix XII, the text following formu-
la (XII.68)l. The non-zero components of this tensor equal +1
or -1 depending on whether an even or odd number of permutations
is needed to obtain the given sequence of indices f-l· v, p. a from
the sequence 0, 1, 2, 3. Table 63.1 gives all the permutations of
the indices with an 'indication of the sign at 1 corresponding to them.
TABLE 63.1
Sequence I Sign II Sequence I Sign I Sequence I Sign I Sequence I Sign
0123
0132
+
-
1320
1302
+
-
2013
2031
+
-
3210
3201
+
-
0231 + 1203 + 2130 + 3102 +
-
0213 - 1230 - 2103 - 3120
0312
0321
+
-
1032
1023
+
-
2301
2310
+
-
3021
3012
+
-
The four-fold contraction of the tensor (63.4) is the invariant

2: ellvpcrpJJ.vFpa=inv (63.5)
J.l,V, p, (]
l';Ll';Gl'J:tODYNAMICS
Having written 24 non-zero components of the sum (63.5) (the

signs of the components must be taken from Table 63.1), it is not
difficult to see that they can be divided into six groups each includ-
ing four coinciding expressions:
~ = 4FotF2a- 4FotF32 + 4Fo2Fat- 4FozFta
+ 4F03 F 12 -4F03 F 21 (63.6)
Owing to the antisymmetric nature of the tensor F 11 ,, the products
F 01 F~ 2 =
-F 01 F 23 , etc. Expression (63.6) is therefore simplified
as follows:
~ =8 (F 0 tF 23 +F02 Fat +F 0 aFt 2) = inv
The expression in parentheses is obviously also invariant. Substitu-
tion of the values for F 11 v from (61.8) yields
Ex (-Bx) + Ey (-B + Ez (-Bz) = 11 ) inv
which coincides with (63.2).
64. Maxwell's Equations in the Four-Dimensional Form

The first pair of Maxwell's equations [i.e. Eqs. (55.12)1 can be
··written as a single equation for the components of the tensor (61.8):
(64.1)
We must note that the indices in each of the addends form a cyclic
,transposition of the sequence IJ., v, p.
·· Expression (64.1) is a set of four equations, the first of which is
obtained at ,._.., v, p equal to 0, 1, 2, respectively, the second at
!J., v, p equal to 1, 2, 3, the third at !J., v, p equal to 2, 3, 0,
and, finally, the fourth at ~l, v, p equal to 3, 0, 1. Owing to anti-
symmetry of the tensor F 11 v, the equation obtained at any other
'combination of the three non-coinciding indices reduces to one of
the four indicated equations.
Let us-write Eq. (64.1) assuming that 1--l = 0, v = 1, and p = 2:
aF01
ax2 I
, aF 12
i)xO
+ aF2o
i)xl
= 0
Introducing the values of F 11 v and the coordinates x~t, we obtain

aEx _ ..!:._ iJBz _ aEy = 0
ay c at ax
whence
aEy
_ BE
__ _ _x_. ==· ____
1 as z
ax i}y c at
The found relation is the z-th component of the vector equation

(55.12). Similarly, the equations for ll = 2, v = 3, p = 0, and
ll = 3, v = 0, p = 1 give the x-th and y-th components of the
same equation. (An equation is obtained for the component whose
index is absent in tho ~et of values of fl. ''· p.)
Assuming that ll ""~ 1, v 2, p = 3, we arrh·o at lhr fnl'mnla
.!I..u.
axa
+ aF2s
axt
+ aFst
ax2
=0
which after introduction of the values of F J.tV and x~t becomes
- fJB, - 8Bx - aBy "-""' 0
az ax ay
This is equivalent to the second of equations (55.10).
We have thus seen that Eq. (64.1) is equivalent to the first pair
of l\Jaxwell' s equations.
The second pair of l\Iaxwell's equations, i.e. equations (55.13),
can be written as
3
~ fJpJ.lV
--=--]ll
4n .
(~.t=O, 1, 2, 3) (64.2)
fJxv c
V=O
Indeed, let us assume, for instance, that ~t = 1. Equation (Gif.2)
can therefore he written as follows:
apto aFt2 , apta 4rt .
a;o +-a;z I --a;a = - -c- J1
= 0). Substituting their values for FJ.tv and xll, and also group-
(F 11
ing the terms as required, we get the equation
fJBz _ aBy = 4n 'x+J... iJEx
au az c 1. c at
which is the x-th component of the vector equation (55.'13). Similarly,
the equations for ll = 2 and ~t = 3 give tho y-th and z-th component~
of the same equation.
When ll = 0, Eq. (64.2) becomes
opot aFo2 apoa
--;)Xl + ax2 + --;};il = - 4np
Introduction of the values of Fllv and xll yields
- aEx - aErl - aEz · = - 4np
ax iJy i)z
which coincides with the second of equations (55.13).

We have thns shown that Eq. (64.2) is eqtlivalenl to the second
pair of Maxwell's equations.
230 ELECTRODYNAMICS
6s. Equation of Motion of a Particle in a Field

According to (38.16), the equation of motion of a charged particle
in nn electromagnetic field is
d
dt
(
V 1mv
- v2 I c2
)
=e
E + ce [ v B] (65.1)
Dividing both sides of (65.1) by V 1 - v2/c 2 and taking into account

that dt V 1 - v2 /C 2 is d1: [see formula (34.13}], we obtain
d ( mv ) e cE+[vB)
d. V1-v2fc2 =c V1-v2fc2
or, in components,
d ( mvx = .!!..._ cEx+vyB 2 -v2 By
)
d. lff-v2fc 2 c lf 1-v2jc2 (65.2)
d ( mvy _ .!!..._ cEy+vzBx-VxBz
)
dT V 1-v2fc2 - c Vf-v2fc2
)
But uh!V 1 - v 2 /c 2 = u 11 , c!V 1 - v2 1c2 = u 0 [see (36.6)1. Further,
E X = p10 ' E y = pzo ' E Z = pao ' B X = pa2 =· _pza ' B .ll = pta=
= -F31, B z = F 21 = -Fl 2 • Therefore, Eqs. (65.2) can be written
as follows:
(65.3)
On the right-hand side, we have used the covariant components

of. the four-velocity so that all the terms will have a plus sign.
Now let us take advantage of the fact that the rate of change
in the energy of a particle equals the work done in unit time by
the forces acting on the particle:
d (
-dt mc2
V 1-v2fc2 ) = ( eE +-
e [vB] ) v=eEv (65.4)
c
Dividing both sides of this equation by c V 1- v2/c 2 , we obtain

duo
ma:t = ce (F 01 u1 + F 02 u2 + F 03 u3 ) (65.5)
EQUATIONS OF ELECTRODYNAMICS IN FOUR-DUIENSIONAL FORM 231
The set of equation~ (65.3) and (65.5) can be written as a single

equation
3
dull _ e ""i0 pll"
m ([t- cL..J . Uv (f.t = 0, 11 2, 3) (65.6)
V=O
This is the equation of m()tion of a particle in a field written in th~

four-dimensional form. Its space components are equival~ri.t to
Eq. (65.1), and its time component-to Eq. (65.4). ·
Equation (65.6) can be written in a somewhat different way.
Let us lower the free index f.t on both sides of formula (65.6).
In addition, let us simultaneously lower the dummy index v in Fllv, ·
and raise it in Uv. The result is
3
m--=- Lj
dull e
F ll'IJ u" (!-1=0, 1, 2, 3) (65. 7)
dT C
V=O
Equation (65. 7) can be obtained directly from Eqs. (65.1) and
(65.4) if we replace the left-hand side in them with the covariant
components of the four-velocity, and write the quantities E::. and B"'
as the components of the tensor (61.8).
Chapter XII
THE V.ARI.ATIONAL PRINCIPLE

IN ELECTRODYNAMICS
66. Action for a Charged Particle in an Electromagnetic

Field
Agreement with experiments is obtained if we take the following

~xpression as the action for a particle in a field: ·
2 3
S= ~ ( -mcds-: 2j A~dx~) (66.1)
;· 1 1'=0
where m is the mass of the particle, e is its charge, and A~ is the
four-potential of the field. We draw attention to the fact that there
is an invariant in the integrand, as there should be. At A~ = 0,
expression (66.1) transforms into the action for a free particle [see
formula (39. 7)1.
We must note that the ambiguity of the potential does not affect
the equations of motion. Indeed, substituting for A~ in (66.1) the
quantities A~ [see formulas (60.16) and (60.17)], we get additional
terms in the integrand:
3
- !!._
c
rL 2j o'iJ d:rfl
axk
+C dxOJ = - !._c (d\j) (X 1 X 2X 3 )
• •
+C dx J
0
k=i
Integration of these terms yields a constant quantity:
2
-.!..c J\ (dljl + C dx 0 ) = - !!._

c
[\j) (2) + Cx~2 ,] + _:_ [\j) (1) + Cx~ 0 ]
c
1
which in variation of the action vanishes.
The right-hand side of (66.1) consists of two terms. The first
depends only on the properties of a particle (on m). The second
describes the interaction of the particle with the fteld; it contains
accordingly both a quantity characterizing the particle (the charge e)
and a quantity characterizing the field (the potential A~). In general,
we ought to include another term describing the field itself. But in
considering the motion of a particle in a given field, this term may
be left out of consideration because owing to the determinacy of
the field it should not vary. True, this statement holds only provided
that the charge of the particle is so small that we may disregard
its influence on the field. In considering the motion of a particle
in a given field, we shall assume that this condition is satisfied.
THE VARIATIONAL PRINCIPLE IN ELECTRODYNAMICS 23~
We shall obtain equations of motion of a particle proceeding

from the principle of least action. According to this principle. the·
following condition is observed for the true trajectory of a particle:
2
68 = /) ~ ( -me
1
V 2j dxJJ. dxJJ.- : ~ A 1, dx") = 0
(we have made the snbstitnLion ds~ o-= 2J dx 1, dx~'). Perfonuing-
variation, we obtain
2
~8 = .Ir ( -me "L.J dxJJ.dsd6xiJ. -

rce L.J
"
A"d 6x"- rc
e " )
L.J 6A 11 dx'' = 0
1
(66.2}
[see formula (39.9); recall that 6 dxJJ. = d /)xll].
Let us integrate the first two terms in the integrand by parts.
As a result, the first term becomes
2
,~ " 6xll--
m L.J duJJ. d-r
dT
1
where • is the proper time of the particle, and u 11 is the four-Yelocity

of the particle [see formula (39.11 )]. Integration of the second term
by parts yields
2 2
- J ; 2j AJJ. d6xl' = -.: 2j A" 6xJJ.I: + J : ~ 6xll dAil
1 1
The first expression on the right vanishes because at the ends of

the trajectory 6xll =·o. Consequently, after integration by parts:
of the first two terms, expression (GG.2) becomes
2
68 = J (m 2jll /)xl' ~; d-r + : 2j,, 6x" dAil- :

1
~
\1
6Av dx'') = 0
(it wi.ll come to light in t]IC followiug LcxL LhaL it is expedient to-
denote the dummy index in the third term by the symbol '\').
Now let us perform the substitutions
dAJJ. = "L.J -"-

BAll dx'' = L.J
" -v-u"
BA 11 d-r in the second term, and
\1
ax \1
ax
6A \1 = " BAv 6xll and dx" = u" d-r in the third term
LJ axil
JJ.
234 ELECTRODYNAMICS
"The result is
2
fJS = Jr ( m ~ fJx~-t dull dT + !_c ~ fJx~-t ~ BAll uv dT
~ dT ~ ~ 8zv
1 It ll v
which can be written as

2
8S = l ~ (m
,I ~ .
dull
dT
+ .!_ ~
C ~
8AIJ. uv- .!_ "" 8Av uv)
8x V C ~ 8xll
6x~-t d-r: = Q
1 ll v v
For the condition indicated above to be satisfied at arbitrary values
·of 6xll, it is essential that all the terms in parentheses be zero. Hence,
we obtain the relations
m dull = !_ ~ ( 8Av - 8Ail ) uv
dT c ~ 8xll 8xv
v
·or
(!J. :::::;0, 1, 2, 3) (66.3)
v
'[see formula (61.5)1, which coincides with (65. 7).

We haYe thus obtained an equation of motion of a particle in
a field proceeding from the principle of least action.
67. Action for an Electromagnetic Field

In the preceding section, we considered the field in which a particle
inoves to be preset, and in this connection we took no account of
the term in the action describing the properties of the field itself.
Now let us consider a system made up of particles in an electromag-
neti~ field and attempt to find equations determining the field
proceeding from the principle of least action. Here we can give
up our assumption on tho smaHness of the particles' charges and
·obtain equations for the true fwld, i.e. the one obtained upon super-
position of the external field and that produced by the charges
themselves. Consequently, the values of A 11 will have to depend
·on the positions and velocities of the particles.
The action for the system field +
particle must consist of three
terms:
(67 .1)
Here S 1 is the part of the action that depends only on the properties
·of the field itself, i.e. the action for the field in the absence of char-
:ges; S m is the part of the action that depends only on the properties
THE VARIATIONAL PRINCIPLE IN ELECTRODYNAMICS 235.
of the particles, i.e. the action for the free charges. And, finally.
Smr is the part of the action that is due to the interaction between
the particles and the field.
As regards the last two terms, their values can be obtained by
summating expressions (66.1) for all the particles. Hence,
N
Sm=- ~mac
a=1
Jdsa. (67 .2)
N 3
Smr=- ~ :a ~ ~ A,1 adx~ (67.3)
a=1 v 1,1=0
Here a is the number of the particle, N is the number of p&rticles
considered, and A'-'a is the potential of the field at the point of four-
space where the a-th particle is. The subscript a must be distinguished
from the indices ~t, v, p, ... used to designate the components
of four-vectors and four-tensors. For the latter indiees, we must
distinguish between upper and lower positions. This distinction
II as no meaning for the index a.
Expression (67.3) can be written in a different way. For this
purpose, let us substitute (dx~!dt) dt for dx~ in it, i.e. write
Smr= __, ~
a
:a
· \
J ~ A~-ta
II
dx 11
---;d-dt
Next let us write the eollection of point charges ea as a charge

distributed in space with the density
p = ~ ea.6 (r- ra)
IJ
[see (41.12)1. Therefore, the eharge confined in the Yolume element

dV can he written as de = p dV, and a sum of the form ~eaf (xa.. Ya.
za) can be replaced with the integral
Jp (x, y, z) f (x, y, z) dV
As n l'esult, we obtain
S mr = - +J p dV J ~ A'-' d;;•
II
dt
Now let us take into account that p (dxJJ./dt) = jll. where jll is
a component of the four-current [see formula (60.8)]. Hence. we
-+ j
can write
Smr= ~ A 11 jlldV dt=- c~ j ~ A11 j!LdV* (67.4)
JJ. II
where dV* = dx 0 dx 1 dx 2 dx 3 = c dV dt.
236 ELECTRODYNAMICS
To obtain an expression for S 1 , let us take into account that, as

shown experimentally, an electromagnetic field obeys the super-
position principle. Therefore, the equations for a field must be
linear differential equations. The equations of a field are obtained
by variation of the action. But in variation, the power of the inte-
grand lowers by one. Consequently, the equations will be linear if
the action contains in the integrand an expression quadratic with
respect to the field. In addition, this expression must be invariant.
There are two very simple quadratic invariants that can be formed
from the characteristics of a field: ,lJA 11Ail and ljF IJ.VFwv. The
first of them is not suitable for our purposes because the four-poten-
tial is determined ambiguously. We thus arrive at the conclusion
that the integrand in the action must consist of the invariant
LjF 11 ,.F11v. To obtain the action for an entire field, we must integrate
over the entire four-space where the field is non-zero. Hence, we
obtain the expression
S t =a J~ F 11 vF11" dV*
Jl,V
where a is a constant, and dV* = c dV dt. Integration over the

coordinates is performed over the entire three-dimensional space,
and over the time between two preset instants t 1 and t 2 •
We obtain correct equations of a field from the expression we have
found if we assume (in the Gaussian system of units) that a =
= -1/16:rrc. Hence,
Sr = - 16~c J ~ FJJ.vF v
f.',V
11 dV* (67.5)
Summation of expressions (67.2), (67.4), and (67.5) yields the

aetion for the system field + particles:
S= :.__ ~ mac ~
a
dsa- c12 J~ AJ.Ijll dV*- iB~c )
J.1
~
Jl, v
F '"'''F 11 v dV*
(67.6)
By formula (63.3), we have lJFr1 vF~'" = 2 (B 2 - E 2 ). Introclnc-
ing this value into (67.5), and also substituting c dV dt for dV*,
we can transform the expression for S 1 as follows:
t.
s t = I dt j 8~ (E 2 - B2) dV (67. 7)
tt
A glance at this formula shows that the Lagrangian for the field
is determined by the expression
L 1= 8~ J(E 2- B 2) dV (67 .8)

THE VARIATIONAL PRINCIPLE IN ELECTRODYNAMICS 237
We have thus established the form of the action for an electro-

magnetic field. We shall find the equation!' of a field proceeding
from the principle of least action in the following section.
68. Derivation of Maxwell's Equations from

the Principle of Least .Adion
Let us fiud the equations of a field, considering the motion of the
particles to be preset. In this case, the term Sm in the action may
be left out of consideration because in view of the determinacy
in the motion of the charges it should not vary. We shall thus proceed
from the expression
S = St +Smt = - 16~c J~ FJJ.vpll"~ dV*- J~ AJJ.jll dV*

ci2 (68.1)
IJ.,V 1J.
[see (67.5) and (67.4)1.

Let us calculate the variation of expression (68.1) and equate
it to zero. We shall take into account that owing to the determinacy
of the motion of the charges the current jll should not vary. Hence,
68 = - 16~c ) 6 ( ~ F IJ.Vpll"~) dV*- c12 ) ~ j'-'6A'"' dV* (68.2)
We determine the variation in the first integrand:

o 2J F 11 vFilv = ~ F'"'vf>Fil"~ + 2J pllvoFilv
We shall raise the indices of the first factor in the first of the sums
and simultaneously lower them in the second factor. The result is
o >; FflvFIJ.V = 2 >; pllVoF'"'v
With a view to formula (61.5) for F 11 v, we shall write this expression
as follows:
0 "" F pllv = 2 "" pllv 0 ( aAv _ aAIJ. )

.LJ ).l.V .LJ axil 8XV
Taking advantage of the antisymmetric nature of the tensor Fll"~,

we shall substitute _pv'"' for FJ.l.v in the first smn on the right-haud
side. and then exchange the places of the indices ~l and ''· As a result,
the first sum will become identical to the second one, and we obtain
(68.3)
238 ELECTRODYNAMICS
By changing the sequence of differentiation and variation of the

quantities A~" we reduce (68.3) to the form
''
o L.J F )J.V pJJ.v-
- - 4 "" a oA ~~
L.J,pJJ.v azV
Now let us transform the expression obtained by the formulR uv'

= (uv)' - u'v:
0 ""F pJJ.V = -4""' _a_ (F)I.VoA) +4 ""'oA _a_ pJJ.V
L.J J.LV L.J azv I! L.J I! azV
Substituting this expression into (68.2), we arrive at the following

formula for the variation of the action:
oS = _'1_ l "\"I _a_ (FJ.LvoA ) dV* - -1- l "" oA _a_ p!!v dV*
4:rcc J L.J axv )J. 4:rcc J L.J ~· axv
J.L,V )J.,V
- c~ ~ 2j jJJ.oAJ.L dV* (68.4)

)J.
The first of the three integrals can be transformed Recording

to the Ostrogradsky-Gauss formula into a surface integral:
J~ J.L,V
a:v (F 11 vOAJJ.) dV* =) 2j pJJ.VoAJ.L dfv
)J.,V
At the boundary of the four-volume being considered, oA J.L = 0.

Hence, the integral we have written vanishes. Consequently, only
the second and third terms must be retained in formula (68.4).
Combining them and factoring out the common factor oA I!' we
obtain
os = _..!.c J""' (-4:n:1
L.J
""
L.J
aFJ.LV
axv +..!.c J·)1.) oA)J. av•
)1. v
Owing to the arbitrary nature of the variations oA !!' the value

of oS we have found may be zero only if all the expressions in paren-
theses vanish. We thus arrive at the equations
(~=0, 1, 2, 3) (68.5)
that are the second pair of Maxwell's equations [see Eq. (64.2)1.
We must note that the relation between the frelds B and E, on
the one hand, and the charges and currents, on the other, is deter-
mined exactly by the second pair of Maxwell's equations [see
Eqs. (55.13)]. The first pair of Maxwell's equations expresses the
properties of the fields B and E and their relation to each other
[see Eqs. (55.12)].
THE VARIATIONAL PRINCIPLE IN ELECTRODYNAMICS 239·
69. Energy-Momentum Tensor of an Electromagnetic Field 1
We established in Sec. 67 that the action for an electromagnetic

field is determined by the expression
St.= -- 161nc J~ Fw<JFJJ.V dV*
~·. 'I'
[see formula (67.5)].
A comparison of formulas (69.1) and (40.2) shows that we must
take the following expression as the density of the Lagrangian for
a fteld:
L • = - 16n
·! ""' F 11 F'.w
LJ " (ti9.2).
11,'1'
We must take the components of the potential A 11 as the general-
ized coordinates q4 for the field, and the derivatives of these com-
ponents with respect to the coordinates xv as the generalized veloc-
ities. To simplify the writing of the formulas, let us introduce the
designation
(69.3)-
Equating the variation of the action to zero and performing the

same calculations that led us to the equations of motion (40.'12)
in Sec. 40, we shall arrive at the relations
aL• )1 a aL*
(f.l=O, 1, 2, 3) (69.4}
aAI1 = ..:.J ax" aa11v
'I'
Equations (69.4) are equations of a field. To determine the nlues

of the derivatives in them, let us write the variation of the func-
tion L*. According to the general rules for calculating a variation,
we have·
8L* = ""' A 8A 11 + LJ
L.J -8aL* ""' - aL•
8-a,,v 6a 1,.., (69.5)
1•
11 11·"
On the other hand, the variation of the function (69. 2) is
6L* ==- -16n
1 -6 ""' F F 11 v =
LJ I1V
- 1-""'
4lt LJ
F'L\'8 iJA,,
azV
= -4lt1- "
L.J
F 11 '.8a J.1V
J..l., v J,L, v IJ.,"

A comparison of the expression obtained with formula (G9.5)
allows us to conclude that
aL* a£• = _i_pJJ.V (G9.6)
aA = 0 • 11 iJa11v 4n
1 Before beginning to read this section, recall the contents of Sec. 40.
240 ELECTRODYNAMICS
Substituting -F'VIl for Fll" in the second expression and then exchang-
ing the places_ of the indices f.t and v, we find that
aL* = __1_ pll" (69. 7)
oavJl 4:rt
Introduction of tho values (69.6) into formula (69.4) leads to the
following "equations of motion" for an electromagnetic fwld:
~ aFJJ."
L.J -v-=0 (f.t=O, 1, 2, 3) (69.8)
ax
v
~we have omitted the factor 1/4n). Equation (69.8) is Maxwell's

-equation (64.2) written for the case when jJJ. = 0. This result is
what should have been expected because we proceeded from the
action for only the field without any charges.
Now let us establish the form of the energy-momentum tensor
for an electromagnet~c field. Substituting the quantities aPJJ. [see
{40.3) and (G9.3)l for qaJJ. in formula (40.1G), we obtain the following
-expression for the components of this tensor:
-v "1\1 ~"L*
uL* - uJJ.
T ll = LJ apJJ.-,-
vapJl
p
Let us replace~iJL*/fJapv with -(1/4n)F"P in accordance with (69. 7).

In addition, let us substitute for aPIIo its value from (69.3), and
for L* expression (69.2). The result is
T"
ll
= __1_ "1\1 oAp pVP
4n LJ axil
+ _1_ ()v "1\1 F pP>v
16n ll LJ f3V
(G9. 9)
P B.v
[in the second sum, we may not designate the dummy indices by
the letters f.t and v as was done in formula (69.2) because in formula
(69.9) f.t and v were already chosen as free indices].
The tensor (69.9) is not symmetric. To make it symmetric, let
us add to it the~tensor
G" = _1_ "1\1 aA~' p"P (69.10)
ll 4:rt L.J uxP
p
which, as we shall show, can be written as

\'P
GvJJ. --- "'1

-·· .i..J
iJQJJ.
iJxP '
where Q'(i' = - Q~"
,.
p
[see formula (40.18)1. Indeed, let us apply to expression (69.10)

the transformation uv' = (uv)' - u'v:
G" = _1_ ""· OAJ.t F"P = _1_"" _iJ_ (A F"P) __1_ ""A aF"P
J' 4n LJ iJxP 4:rt L.J iJxP J.t 4:rt LJ J.t iJxP
p p p
Owing to (69.8) the second sum on the right-hand side vanishes

(A~ can be factored out of the sum sign). Consequently, we have
brought the tensor (69.10) to the form
Gv = """ _a ( _1 A pvp)
1.1 LJ axr> 4n ~
p
Since the expression in parentheses is antisymmetric with respect

to the indices v and p, expression (69.10) may be added to the ten-
sor (69.9) [see formula (40.18) and the text associated with it).
Summation of expressions (69. 9) and (69.10) gives the tensor
aA
1 """ _P 1 aA 1
Tv= _ _ pvp+- """-~~ pVP+-8v """F p~v
1.1 · 4rt £..j ox~l 4n LJ axP 16n ~ LJ /3i'
n P /3,v
=- _1_ """ ( aAp - aA~ ) pYP + _1_ 8" """ F . p~v
4i1 LJ OX~ ()xP 16:-c 1.1 LJ ll~
P B. v
The expression in parentheses is F ~P [see (61.5)). Consequently,

the formula for the mixed components of the energy-momentum
tensor becomes
T~ = - 4~ ~ F~pFvp+ 1 ~n 8~ ~ F13vF~v (69.11)
p /3. 'V
To go over to contravariant components, we shall raise the index t-t

in all tho terms of formula (69.H). Here 8~v will transform into
gwv [see (XII.66) and (XII.67)1. Hence,
T""-
-
_._1_ """ pll pVP
4n LJ P
+ _1_
16n
gJ.IV ""' F p~V
LJ llv (69.12)
P ll.v
Raising of an index in one factor with simultaneous lowering of
the samo index in tho second factor does not change the product.
We can therefore write with equal grounds that
T"v = __4rc1_ ""'

LJ
pi'P pv
P
+ _1_
16n
g~'v """ F Fllv
L.J flv
(69.13)
P ~.v
A comparison of expressions (G2.12) and (62.13) allows us to

conclude that the tensor r~v is indeed symmetric.
Let us calculate the trace of the tensor Tllv. According to for-
mula (69.11),
~ T~ = ~ ( ·- 4~ ~ F~.~pF~P) + ~ ( 1
u ~ p ~
in 8it ~ Ft3vF
fl.')
13")
242 ELECTRODYNAMICS
In the second term, all 1he factors P.xcept c5~ can be put outside the
sign of summation over f.t· The sum ~c5tt equals four. Hence
~Ttl=- 4~ ~ F"PF"P+4 1in ~ FBvpf3v

'" j.L, p (3, '\'
It is not difficult to see that the last expression vanishes. We have
therefore estab:5shed that the trace of the tensor Tllv is zero:
LT~=O (69.14)
j.L
Let us find the expressions for the components of the tensor fW'J
in terms of the components of the vectors E and B. For this purpose,
we shall introduce the values of the components FJtv into (69:13).
We shall take these values from (61.9), having in mind that lower-
ing of the time index does not change the components of the tensor,
while lowering of the space index reverses the sign of a component.
We shall also take into account that the sum ~F JJ.vFilv is an in-
variant that we calculated in Sec. 63. It eq<1als
.22 FIJ.VFIJ.V=2(B2 -E 2) (69.15)
).L,V
[see formula (63.3)].

Let us begin with calculation of T 00 • According to (69.13), (61. 9),
and (69.15), we have
~ 0 ~. __1_ ~ pOP pO + _1_ goo2 (82 _ £2)
;('
4n L
, P 16n
1
= - 4n (-£2) + 16rt1 2 (B2-E2)
1
=an (E2+B2) = w
[g00 = 1; see (XII.67)]. We have obtained an already known result:

the comJlonent T 00 equals the energy density w [see formula (40.27)].
Now let us find the component 1°1 • Since g01 = 0,
T01 • . - _1_ " pOP pip·
· 4n LJ
p
1
= -4n [O·Ex+ (-Ex) ·0+ ( -Ev) Bz+ ( -Ez) ( -Bv)l
--.: 4~ (EvBz-EzBv>={ 4~ [EBJL=! Sx (69.16)

where S x is the x-th component of the Poynting vector (in a vacuum
H = B). Similar calculations show that T 02 = Syfc, and T03 =
= Szlc [compare with formula (40.32)1.
THE VARIATIONAL PRINCIPLE IN ELECTRODYNAliiiCS 243
Finally, let us calculate, for example, T 12 • According to (XII.67)~

g12 = 0. Therefore
TI2=- 4~ ~ p1Pp2p
p
1
= -Tn'" [ExEv +0·( -Bz) + ( -B:)·O +BvBx]
1
= - 4it (ExE v + BxB y) = f1xy
where f1xy is a component of the Maxwell stress tensor {see formu-

la (59.12)1. It is a simple matter to verify that all the remaining
components having the form Tih coincide with the relevant compo-
nents of the tensor (59.12):
Till.= a 1k = 4~ {; (E 2 +B 2) 6 1 ~t -E 1 E~t -B 1 B~t} (69.17)
[in verifying, it must be taken into account that gii = -1 =

= -6u; see (XII.67)].
Let us consider the matter of diagonalization of the tensor T!!v.
In Euclidean space, such a transformation of a symmetric tensor
is always possible. In pseudo-Euclidean space, however, matters are
different, as we shall now see. According to (69.16), components
of the form T 0 i = Tio vanish provided that
[EB] = 0 (69.18)
According to (69.17), components of the form Ti" (i =1= k) vanish
provided that
(i =I= k) (69.19)
Therefore, for dhtgonalization of the tensor T~"', we must transfel"
to a reference frame in which the vectors B and E are collinear,.
or one of them vanishes [in which case condition (69.18) is satisfied].
We established in Sec. 62 that such a frame always exists except.
when Band E are mutually perpendicular and identical in magnitude.
In this reference frame, one of the coordinate axes must be directed
along the field. This will result in condition (69.19) being satisfied.
The tensor will therefore acquire a diagonal form. Let us find the
components Tii assuming that the x-axis has been chosen in the
direction of the fields and, consequently,
Ex= E, Ey = Ez = 0, Bx = +B, By = Bz = 0
Hence, by (69.17),
Tu= 4~ {~ (E2+B2)-E2-B2}=- 8~ (E2+B2)= -w
rzz = rsa = _1 (E2+B2) = w
8rt .
244 ELECTRODYNAMICS
We know that the component T 00 also equals w. Being reduced

to the diagonal form, the energy-momentum tensor of an electro-
magnetic field is as follows:
(T"') ~ u-~ ~ ~)
/w
0 0 w
0 w 0
(69.20)
If the vectors B and E are mutually perpendicular and identical

in magnitude, the tensor Til" cannot be given a diagonal form (the
mutual perpendicularity does not allow B and E to be transformed
so that they become collinear, and the equality of their magnitudes
does not allow them to be transformed so that one of the fields
vanishes).
We must note that in mixed components, the tensor (69.20) is
(69.21)
It directly follows from (69.21) that the trace of the tensor Til"
is zero. We obtained this result earlier in the general case [see
(69.14)1.
70. A Charged Particle in an Electromagnetic Field

We know that the action for a charged particle in an electro-
magnetic field is determined by the expression
2 3
S= j ( -mcds-; ~ A 11 dxll) (70.1)
1 !l=O
The covariant components of the four-potential can be written as
A 14 = (<p, -A)
[see (60.15)1. The components of the four-position vector are
xll = (ct, r)
Consequently,
3
2J A14 dxll = cpc dt- A dr = (cq>- Av) dt
!l=O
Let us substitute this value of the sum into formula (70.1). In

addition, let us substitute c V 1 - v2/c 2 dt for ds in accordance
with (34.4). The result is

2
S= J(-mc V 1-z· /c'l....:...; AY-eq-) dt

2 2
Hence we conclude that the Lagrangian for a charged particle in

a field is
L = - mc 2 V1-v 2/c2 +..:_ Av- e<p (70.2)
c
The first term is the Lagrangian for a free particle [sec (39.4)1.
The other two terms describe the interaction of the particle with
the field.
Knowing the Lagrangian, we can calculate the energy and momen-
tum of a particle. According to formulas (4.19) and (5.1), the gen-
eralized momentum is determined by the expression
P=~
av
and the energy, by the expression
oL
W=-v-L (70.3)
ov
Consequently, having differentiated the function (70.2) with respect
to v, we obtain the generalized momentum of a particle:
P= mv
y1-v2fc2
+ !.A=p+_:.
c c
A (70.4)
Here p is the conventional momentum of the particle [see formu-

la (38.5)]. A glance at formula (70.4) shows that the generalized
momentum differs from the conventional one in the term (e/c) A.
In the absence of a fwld, the generalized momentum coincides with
the conventional one.
Now let us determine the energy of a particle. In accordance
with (70.3) and (70.2), we have
W= OL v-L= Pv-L
av
=( mv
Y1-v2/c2
+.!..c A) v- ( - mc2 V 1- v2/c 2 + ..:_ Av- e<p)
c
- Y 1-v /c + e<p
mc 2
2 2
(70.5)
The first term in (70.5) is the energy of a free particle [see formu-
la (38.11)], the second term is the additional energy that a particle
in the field has.
ELECTRODYNAMICS
Replacing the velocity v in the expression for the energy in terms of

the generalized momentum P, let us find the Hamiltonian of a par-
ticle. To delete v from formulas (70.5) and (70.4), let us write these
equations as · ·
W-e<p _ me
c Y t-v2jc2
If. we square these equations and subtract the lower one from the
upper one, on the right we obtain m2c2 • Consequently, substituting
Q/8 for W, we have
(70.6)
whence
(70. 7)
This is the Hamiltonian for a particle in a field.

, .It was established in Sec. 32 that the components of the general-
ized momentum equal the derivatives of the action with respect
to the relevant generalized coordinates [see formula (32.6)}. In our
case, the role of the generalized coordinates is played by the Carte-
sian coordinates x 1• Consequently,
as
• a::c i or P=VS
P.=-
Further, according to (32.10), the derivative of the action with

respect to time gives the Hamiltonian taken with the opposite sign:
as
at= -d/8
-
Substituting -as/at for $8 and 7 S for P in formula (70.6), we

arrive at the Ramilton-J acobi equation for a particle in an electro-
magnetic field:
as
-c12 (.-+ecp
at
)2 - (VS--ce A )2 -m c 2 2 =0 (70.8)
In the classical approximation, i.e. when v ~ c, the function

(70.2). becomes
mv2 e
L=-2-+-c Av-ecp (70.9)
lwe have expanded (70.2) by the powers of v2 /c 2 and discarded the

constant -mc2 ].
Differentiation of (70.9) with respect to v yields the generalized

momentum
(70.10)
\Vhere p is the conventional momentum.

For the energy in this approximation, we obtain
e ) mv 2 e mv2
W=Pv-L= ( mv +-A
c v - -2 - - -
c Av-'-eqJ=--+e(fl
2 l
{70.11)
From (70.10), we have
Introducing this value of v into (70.11), we arrive at an expression

for the Hamiltonian:
618 =2m e A )2 _J_ eq::
1- ( P-- (70.12)
c 1
The Hamilton-Jacobi equation in the classical approximation is
-as
/}t
1 {vs--ce A )2 +e(fl=O
+2m
- (70. '13)
Chapter XIII
ELECTROMAGNETIC WAVES
71. The Wave Equation
It was shown in Sec. 57 that upon application to the potentials A

and !p of the Lorentz condition
VA+~~=O
c at
(71.1)
[see formula (56.8)1 they satisfy d' Alembert's equation:
0 A=- 4 :rt~ j (71.2)
c
4:rtp
0 !p = - - 8- (71.3)
[see Eqs. 1 (57.6) and (57.7)]. Here 0 is d'Alembert's operator equal
to
e~ 2
a _ a2 , a2 a2 e~ a2
0 = !l. - cr 7fi2 - fu2 -r ay2 + 7fii - C2 7ft2 (71.4)
[see (57.4)].
In the absence of charges and currents (i.e. at p = 0 and j = 0),
the equations for the potentials acquire the form
OA=O (71.5)
0!p=0 (71.6)
or, with account of (71.4)
!l.A-~ azA =0 (71. 7)
c2 at 2
Bfl a2<ll
!l.(p---=0 (71.8)
2 c2 at
Similar equations are also obtained for the vectors E and B:
!l.E- ~ azE =0 (71.9)
c2 at 2
(71.10)
(see Sec. 74).
1 Recall that Eqs. (57.6) and (57.7) were obtained on the assumption that
the medium in which the field is being considered is homogeneous and isotropic,
and, in addition, that e and fl. do not depend on E and H.
ELECTROMAGNETIC WAVES 249
Equations (71. 7)-(71.10) have non-zero solutions. Consequently,

electromagnetic fields can also exist in the absence of charges. Such
electromagnetic fields arc known as electromagnetic waves.
An equation of the kind
1 8 2/
t:.t- 7;2
=0 8t2 (71.11)
where v is a constant, is called a wave equation. It is known from
the general course of physics that v is the phase velocity of a wave.
Consequently, the velocity of electromagnetic waves is
c c
V=--=- (71.12}
Ye!J. n
where
(71.13)
is the refractive index of the medium in which a wave propagates.
It is a simple matter to obtain the wave equation in the four-
dimensional form. Let us do this for the field in a vacuum. By (64.2),
Maxwell's equations in the absence of charges and currents are-
(~=0, 1, 2, 3) (71.14)
The substitution for FIJ.v of their values from (61.6) yields

a ( aAv BAi-l )
~v axv ax"' - OXv I= 0
or
~ azAv o2Ai-l
.£-l---~ =0 (71.15)
.v
ox axv
1.1
L.J
v
ax ax"
v
The first sum can be written as follows:
~ o2 A" a aAv
v
axi-l axv = axi-l Lv axv
If the four-potential satisfies the Lorentz condition (60.18), then

~ (oAv/ox") = 0, and the first term in Eq. (71.15) will vanish.
We thus arrive at the equations
~
L.J
82AIL =0 (~=0,1,2,3) (7'1.16)
OXv axv
v
Raising of the index on oxv is equivalent to multiplying each
of the addends in (71.16) by gvv [see (XI1.30)]. Therefore, Eq. (71.16)
can be given the form
250 ELECTRODYNAMICS
Finally, taking into account that the non-diagonal components

of the tensor gvr1 are zeros, we can write
(~ = 0, 1' 2, 3) (71.17)
Equations (71.16) and (71.17) are wave equations in the four-

dimensional form. Substitution of values for g'~~P and x'" in them
results, as can readily be seen, in Eqs. (71. 7) and (71.8).
72. A Plane Electromagnetic Wave in a Homogeneous

and Isotropic Medium
:,. ·The solution of the wave equation is considerably facilitated
if the field depends on only one coordinate, say x. The wave in this
case is called plane. Using a plane wave as an example, we can
determine all the features of electromagnetic waves. For these
reasons, we shall treat only plane waves in this chapter.
Let f stand for any of the components of the vector potential A
or the scalar potential cp (with equal reason, we can understand f
to signify any component of the vector E or the vector B). For a plane
wave, the function f depends only on x and t and, consequently;
is a solution of the equation
o2f e~t o2 /
ox2 - C2 ot2 =0 (72.1)
This equation can be written as
D! = 0 r (72.2)
where D is a differential operator determined by the· formula
We shall represent this operator as
(72.3)
where n = V e~ [see (71.13)1.

Let us introduce the new variables
t X X
<::. = t - C77i ' T) = t+ c/n (72.4)
i.e. replace x and t by the formulas
T)-~ c
X=-2--;'
t-
-
11+;
2 (72.5)
By (72.5), we have
ara = axa arax +7fta arat = c

- 2n a;
() 1 ()
+ 2 8t = - 2n
c ( a
ax- c at
n () )
Consequently, in going over to the variables and s T), the first factor
in (72.3) must be replaced with -(2n/c) alfJs.
Similarly,
a a ax a at c a 1 a c ( a n a)
ail = ax ail + Tt ail = zn ax + :r at = 2ri""" ax + -c a~
so that the second factor in (72.3) must be replaced with (2n/c) fJ/oT).
The result is
D --
4n 2 a
- - - - - -c2- a~
-iJT]
-
f) 4n2 a2
- c2 a~ all -
Let us introduce the found value of D into Eq. (72.2), discarding
the factor -4n2 /c 2 • The result is the differential equation
a2j .
a~ all =0 (72.6)
An obvious solution of this equation is a function depending

on only one variable~ or T), i.e. the function / 1 (£) or / 2 (TJ). Summat-
ing the functions / 1 and / 2 , weget a general solution of Eq. (72.6)_:
t <£, TJ) = t1 <s) + t 2 (TJ) (72. 7)
We did not multiply / 1 by C1 and / 2 by C 2 because / 1 and / 2 are
arbitrary functions of the relevant variables.
Introducing into (72.7) expressions (72.4) for and TJ, we arrive s
at a solution of Eq. (72.1):
f(x, t)=f.{t- ,;n )+tdt+ c~n) (72.8)
The first term in this expression is a wave running at the velocity
c/n in the direction of the x-axis. Indeed, the values of / 1 are the
same for all the values of t and x related by the expression
t--=-c/n == const or x = ~
n t-~n const
whence it follows that any preset value of the function / 1 travels
along the x-axis at the velocity c!n.
Similarly, the second term in expression (72.8) is a wave travelling
at the velocity c/n in a direction opposite to the x-axis.
The shape of the wave (72.8), i.e. the form of the functions / 1
and / 2 , is absolutely arbitrary.
Consider a plane wave propagating in the direction of the x-axis.
Let us choose the potentials so that
q> = 0, VA= 0 (72.9)
252 ELECTRODYNAMICS
Such a gauging of the potentials satisfies the Lorentz condition

(71.1) and, consequently, does not affect the values of the vectors B
and E (see Sec. 56). When we apply condition (72. 9) to the poten-
tial <p, we obtain
1 aA
E = -c -at- (72 .10)
[see (56.3)1.
For the wave being considered, all the quantities characterizing
a field including A have the form
A= A (~) = A ( t - _c;nx ) (72.11}
so that
aAi _ d aAi _
-ay- - 0 an -{jz- - 0 (i=x, y, z)
Particularly, BAy!fJy and fJAzlfJz vanish. Hence, the second of condi-

tions (72. 9) in the case being considered becomes
&Ax =0
ox
Introduction of the last relation into Eq. (72.1) written for A:c
yields
According to the last equation

&Ax
-ae=cons t (72.12)
The derivative aA xlat determines the field component EX [see

(72.10)]. Relation (72.12) therefore signifies that Ex= const. We
have thus arrived at the conclusion that a non-zero component Ex
can be dl:ie only to a constant and homogeneous electric field. Such
a field has no relation to an electromagnetic wave. Consequently,
we can consider that Ax and, therefore, Ex too, vanishes.
Examination of (72.11) shows that
aA aA aA aA a~ n &A
ae=ar · a:x=arax= -c-or (72.13)
According to the first of these formulas
E= _.!_&A=_.!_~ (72 .14)
c at c a£
Since A depends only on the single coordinate x, in the expres-
sion B = [VA] we must retain only the component V equal to
ex (8/ax). We can therefore write that
B= [ex a~ , A J= [ex, ~~ J
or, according to the second of formulas (72.13),
B =[ex, (- ~ ~~) J
Finally, taking into account (72.14), we obtain
B = n [ex, E] = V e~-t [ex, E] (72 .15)
whence it follows that the vector B is perpendicular both to the
vector E and to the x-axis. We showed above that for a wave Ex = 0
and, consequently, the vector E is perpendicular to the x-axis.
We thus arrive at the conclusion that in a plane wave the vectors E
and B are perpendicular to the direction of propagation of the wave.
Consequently, electromagnetic waves are transverse.
In formula (72.15),ex is the unit vector of the direction in which
the wave propagates. Designating this unit vector by the symbol
k 0 , we arrive at the formula
B = V e~-t [k 0, E) (72 .16)
that does not depend on our choice of the directions of the coordi-
nate axes.
Substitution of ~-tH for B yields
V~H = [ko, VeEJ (72. 17)
It can be seen from (72.17) that the vectors H and E are mutually
perpendicular, and their magnitudes are related by the expression
v~H =VeE ts) <n.
In addition, it is easy to conclude from (72.17) that the vectors E,
H, and k 0 form a right-handed sequence.
An electromagnetic wave carries an energy whose flux is deter-
mined by the Poynting vector
S= _c_ [EH]
J.4:t
Let us express H in terms of E in accordance with (72.17) and use
formula (VI.5). Hence
S= :n V ; [E, [koEJJ = 4crt V~ E 2ko- ln V ~ (Eko) E
Since the vectors E and k 0 are mutually perpendicular, Ek 0 = 0
so that
254 ELECTRODYNAMICS
Using relation (72.18), we can write

S- c E2k - c lf2k - c (eE2+flH2) k
- 4n V Bfl 8 0- 4n V Sfl f.L 0- V Bfl Sn: 0
Finally, with a view to (eE 2 + f.LH 2 )/8n being the wave energy
density w, we obtain
S= ,; wk 0 =vwk0 (72.19)
J' Efl
In a vacuum
S = cwk 0 (72.20)
Expression (72.19) agrees with the relation known from the gen-
eral course of physics, according to which the density of the energy
flux carried by a wave equals the density of the energy multiplied
by the velocity of propagation of the wave. The direction of the
energy flux density vector coincides with that of wave propagation.
The momentum flux is determined by Maxwell's stress tensor whose
components are calculated by the formula
atk = 4~ { E 1Dk +H 1B,- +(ED+ HB) 61k}
[see (59.12)1. Let us direct the x-axis along k 0 , the y-axis along E,
and the z-axis along H (the vectors E, H, and k 0 will form a right-
handed sequence). Consequently, Ey = +E, Ex = Ez = 0, Dy =
= ±D, Dx = Dz = 0, Hz= +H, Hx = Hy = 0, Bz = +B, Bx =
=By = 0. It can be seen that with such a choice of the axes all
the non-diagonal components of the tensor crik vanish. Since E xD x
and HxBx equal zero, we obtain that
ED+HB
O"xx=- Sn: = -w
Since EyDy =ED, and H yBy == 0, we have
cryy= 4~ { ED-+(ED+HB)}=O
[by (72.18), ED = HBJ. Similarly
cr zz = 4~ { 11 B - + (ED + IfB)} =·: 0

Hence, with our choice of the coordinate axes, only one component
of Maxwell's stress tensor is non-zero-the component crxx• and
it equals the wave energy density w taken with the opposite sign.
We remind our reader that in accordance with the customary prac-
tice in the overwhelming majority of manuals on electrodynamics,
we defined the tensor aik to characterize the flux of the momentum
not flowing out of a given volume, but flowing into it (see Sec. 59).
This is equivalent to reversing the direction of a normal to an area.
Hence, the momentum flux carried in the direction of the x-axis

through the area df perpendicular to this axis is determined by
a positive quantity, namely, axx (-df) = w df.
13. A Monochromatic Plane Wave

A wave in which the field nt ('ach point vnries with time according
to a harmonic law (i.e. according to a cosine law) is called mono·
chromatic.
For a monochromatic wave, the solution of Eq. (71. U) is
f = a cos (wt + '1Jlt) (73.1)
where a is a quantity not depending on t, and 'ljJ 1 is a function of r.
In this case, 82 j!8t 2 = -w 2f. Substitution of this value into (71.11)
yields the equation
oo2
!!.!+ 7 !=0 (73. 2)
that determines how f depends on r.

Let us introduce the quantity
2n
k =-;;-- ko =
"'
-wv ko = V -e~-t -wc ko (73. 3)
known as the wave vector (k 0 is the unit vector of the direction in

which the wave is propagating). Equation (73.2) can now be written
as follows:
(73.4)
In the following, we shall limit ourselves to a treatment of a plane
monochromatic wave. Let us choose the x-axis in the direction of
wave propagation. Equation (73.4) therefore becomes
(73. 5)
The following function will be a solution of this equation:

f = a cos (±kx + 'ljJ 2 ) (73.6)
where a is a quantity not depending on x, and 'Pz is a function of t.
Expressions (73.1) and (73.6) can be brought into agreement,
assuming that;
f = a cos ( wt ± kx a) + (73. 7)
where a and a are quantities depending neither on t nor on x. The
different signs of kx correspond to different directions of wave pro-
pagation. We shall consider waves running in the direction of increas-
ing x, and in this connection we shall write a minus sign in front
of kx.
256 ELECTRODYNAMICS
With an arbitrary choice of the coordinate axes, formula (73. 7)

becomes
f = a cos (rot - kr ~) + (73.8)
Any quantity characterizing a monochromatic plane wave changes
according to this law, particularly the vector potential A. We thus
have
A = A 0 cos (rot- kr +a.) (73.9)
Here A 0 is the amplitude, ro is the frequency, k is the wave vector,
and ~ is the initial phase of the wave.
The expression (rot - kr + ~), called the phase of a wave, is an
invariant. Indeed, the fact that the field at a given point of space
at a given instant acquired, for example, a zero value [cos (rot -
- kr + ~) = 0] cannot depend on the choice of the reference frame.
It thus follows that for two arbitrarily chosen reference frames, the
conditinn
rot- kr + ~ = ro't' - k'r' +a.'
must be satisfied.
If we assume that t = 0 and r = 0, it can be seen from the Lorentz
transformations that t' and r' also equal zero. Consequently, we
must have ~ = ~' = inv, from which we conclude that the quantity
<1> = rot - kr (73.10)
also called the phase, is invariant.
For an electromagnetic wave propagating in a vacuum, the phase
(73.10) can be written as
w (J) 0
<D=-(et)-kr=-x
c c
-kr (73 .11)
Since.<!> is an invariant, it follows from (73.11) that role and k form

a four-dimensional wave vector
(73 .12)
This circumstance makes it possible to find the law of transfor-

mation of the wave frequency ro in passing from one inertial refer-
ence frame to another. According to the first of formulas (36.1),
kO-~kl
k iO =--::====
Vi-~2
whence after introducing the values k 0 = role, k1 = kx =
= (role) cos -fr (-fr is the angle between the direction of wave propa-
gation and the x-axis), and J3 = vic (v is the velocity of the frame
K' relative to the frame K), we find that

ro' = w [1- (L•/c) cos 1'7]
Vi-v2fc2
This formula for t} = 0 yields a formula for the longitudinal
Doppler effect:
ro' = w (1- vic) = w, /1 vjc
Y1-v2Jc~ V i+v!c
When t} = n/2, we arrive at a formula for the transverse Doppler
effect:
w' = w
Y 1-v2jc2
We must note that since the magnitude of the wave vector is
role, the square of the four-vector kiJ. vanishes:
3
~ kl1k)l.=(k0) 2 -k2 =0
Jl-=0
Let us again turn to expression (73. 9). It can be written as

A= Re {A 0·ei (oot-krJ} = Re{A} (73 .13)
where 10 is a constant complex vector:
Ao = Aaeia (73.14)
and 1 is the complex vector in the braces [compare with (16. 9) and
(16.10)].
A change in the sign of the exponent in an expression such as eN
does not change the real part of this expression. Formula (73.13)
can therefore be written as follows:
A= Re {A 0ei (kr-ooiJ} = Re {A} (73 .15)
In this case by the phase of the wave we must understand the expres-
sion differing from (73.10) in its sign. In addition, the sign ought
to be changed in the exponent of the quantity (73.14). But owing
to the arbitrary nature of a, this may not be done.
Expression (73.15) is in some respects more convenient than
expression (73.13) and, in addition, is more similar to the wave
function describing the motion of a free particle in quantum me-
chanics than (73.13).
It is a simple matter to see that in performing linear operations
on the quantity A determined by formula (73.15), for instance
differentiation, we can perform these operations on the complex
vector A and then take the real part of the obtained quantity. For
I 7-018
258 ELECTRODYNAMICS
example,
aA
ar=Re { a.A} ~
at =Re{ -HuA} . (73.16)
Let us calculate the curl of the vector A. This is a linear operation,

hence
[V' A] = Re {[V' AI}= Re {[V', A0 ei (kr-wt)l}
· Re {[V'ei (kr-wt), Ao]} (73 .17)
Since kr = ~ k1xJt we have
_8_ ei (kr-wt) = ik ·ei (kr-rot)
tJXj )
Hence,
vei (kr-wt) = iei (kr-wt) ~ ejkj = ikei (kr-rot)
j
Introducing this value of the gradient into (73.17), we obtain

[V' A]= Re {[ikei (kr-wt), A 0l}
= Re {[ik, A 0ei (kr-wt>l} = Re {i [kAJ} (73 .18)
Now we can write expressions for the fields E and B:
1 aA
E=---=Re (J) ~}
{.~-A (73 .19)
c at c
B =['VA]= Re{i [kAJ} (73.20)
[if we had proceeded from expression (73.13), a minus sign would
have appeared in the last two formulas]. A comparison of them shows
that the vectors B and E oscillate in the same phase.
Having in view that k = ( w V ef.L/c) k 0 , formula (73.20) can be
written as
B-=Re{i[Vef.L ~ k 0 , A.J}=Re{Vef.L[k 0 , i ~ .AJ}
The' second factor in the last vector product is E [see (73.19)1. Hence,
for the vector B whose real part gives B, we obtain the relation
(73. 21)
Introducing the notation i (wlc) A0 = E0 , we can write expres-
sion (73.19) as follows:
(73. 22)
If we direct the x-axis along the vector k, the vector E and, conse-
quently, E0 will be in the plane yz. It can therefore be written as
a linear combination of the unit vectors ey and ez:

E0 = ~ey + fJez
where ~ and fJ are complex numbers.
We shall show that upon the proper choice of a, the last expres-
sion can be transformed as follows:
(73.23)
where Eyo and Ezo are real quantities. To do this, we shall write
the following expression for the square of the vector E0 \vhich,
generally speaking, is a complex number:
E02 =real number·e-2 fa
Let us represent the real number as the square of a vector quan-
tity. Not only a real vector, but also the complex vector a+ ib
may be this quantity provided that a and b are mutually perpendi-
cular. Indeed, when ab = 0,
(a + ib) 2 = a 2 + 2iab + b2 = a2 + b2 = real number
We have thus obtained the expression
E~ = (a +
ib) 2 e· 2 ia,
whence
+
E0 =(a ib) e-ta (a _l_ b)
Directing the y-axis along the vector a, and the z-axis along the
vector b, we arrive at formula (73.23).
Let us introduce expression (73.23) into formula (73.22):
E =He {(EyoCy + iEzoCz) e-i (<ot-hs+a>}
(with the directions of the axes we have chosen, kr = kx). Hence
Ey = Eyo cos (ffit- kx +a), Ez = Ezo sin (ffit- kx +a) (73.24)
Inspection of these formulas shows that
E2 2
_Y +~=1 (73.25)
E~o E;o
All our reasoning and all formulas beginning with (73.22) also
hold for the vector B.
Our result signifies that the vector E rotates in a plane perpendic-
ular to the direction of propagation of the wave, its tip describing
an ellipse (the vector B behaves similarly). The direction of rotation
depends on what signs (identical or opposite) Eyo and E 20 have in
formulas (73.24). Such a wave is called elliptically polarized. At
Eyo = +Ezo• the ellipse (73.25) transforms into a circle. In this
case, the wave is circularly polarized. Finally, one of the quantities
Eyo and E 20 may be zero. Here, the vector E (and also B) is con-
17•
260 ELECTRODYNAMICS
stantly directed along the same straight line. The wave in this
ease is called linearly polarized or plane polarized.
A monochromatic wave is thus polarized (elliptically, circularly,
or linearly) in all cases. This signifies that the oscillations in the
wave are ordered in some way or other.
74. A Plane Monochromatic Wave in a Conducting Medium

Equations (71.7) and (71.8) were obtained assuming that the
medium contains no surplus free charges (p = 0) and no conduction
currents (j = 0). We established in the preceding section that for
a plane monochromatic wave the function (73.9) with a constant
amplitude is a solution of Eq. (71. 7). The failure of the amplitude
to depend on the coordinates signifies that the propagation of a plane
wave in a dielectric is not attended by a change in its intensity.
Now let us assume that a medium has the conductivity a so that
conduction currents j equal to aE can be produced in it. We shall
consider as previously that surplus free charges are absent (p = 0).
With these assumptions, Maxwell's equations (55.10) and (55.11)
will be written as follows:
1 8B
[VE]= -cat' VB=O (74.1)
(74. 2)
(we have substituted aE for j).
Let us take a curl of the first of Eqs. (7 4.1):
1 a
[V, ['VE]] =-; Tt [VB] (74.3)
According to (XI.45), [V, fvE)] = V (VE) - ~E. But VE = 0 so

that only the second term equal to -~E remains. Let us substitute
it for the left-hand side of Eq. (74.3). In addition, let us introduce
into the l'ight-hand side of this equation the value of [VB] from
(74.2). The result is
Let us write this equation as follows:

~E-~ iJ2E _ 4.rq.t.cr aE =O (74.4)
c2 ata c2 i}t
Equation (74.4) is called a generalized wave equation. It differs

trom Eq. (71. 9) in an additional term containing the first time
derivative of the required function. When a = 0, this term vanishes,
and Eq. (74.4) transforms into (71.9).
Considering the wave to be monochromatic, we shall seek the

solution of the equation in the form
E = Re {E (r) e-iwt} = Re {E} (74 .5)
where E (r) is a complex vector function of r, and E is the function
in braces. Differentiation of the function (74.5) with respeet to
t yields
iii""=
8E Re {- iffiE (r) e-iwt} = Re {- iffiE}
A A }
iJ2E (74.6)
ai2 = Re { - (!) 2E (r) e-icut} = Re { - ffi 2E}
A A
Substitution into Eq. (74.4) of the values (74.5) and (74.6), after
cancelling the common factor e-icut, results in the following dif-
ferential equation for E (r):
LlE (r) + B~~ 2 E(r) + i 4 rr~crro E (r) = 0 (74. 7)
Let us multiply the numerator and denominator in the third term

by Effi and substitute k 2 for Ef!ffi 2/c 2 [see (73.3)1. As a result, combin-
ing the second and third terms, we get the equation
LlE (r) + k 2 1+ i e;;;-
A ( 4ncr )
E (r)A
= 0 (74 .8)
When a = 0, this equation transforms into an equation like (73.5).

Let us represent the coefficient of E (r) in Eq. (74.8) as the square
of the complex wave number k = k 1 + ik 2 :
k 2 ( 1 + i ::cr) = (k 1 + ik 2) 2 = k:-k; + 2ik 1k2
Equating the real and imaginary parts, we arrive at a system of two
equations for k 1 and k 2 :
k:-k: = k2 }
2k k = 4ncrk2 (74. 9)
l
1 2 f.(J)
The solution of this system has the form
k 1 = k } / 1+ V 1+2(4ncr/ero)2
(74.10)
k2 = k V -1+ V 1i(4ncr/ero) 2
When (4na/effi) 2 ~ 1 (i.e. when a is small and ffi is large), for-

mulas (74.10) become greatly simplified: the quantity k 1 can be
taken equal to k, and in the expression for k~ we can assume that
262 ELECTRODYNAMICS
V1 + (4ncr/ero) 2 ::::::; 1 + ; (4ncr/ero) 2 , which yields the value 2ncrk/ero

for k 2 • Hence, at a low conductivity and a high frequency
(74.11)
Equation (74. 7) can be written as
,:lE (r) + k2E (r) = 0 (74.12)
where k = k 1 + ik 2 is a complex quantity whose real and imaginary
parts are determined by formulas (74.10).
Consider a plane wave. We choose the x-axis in the direction of
wave propagation. Therefore E (r) depends only on x, and Eq. (74.12)
can be simplified as follows:
82!S) + fczF; (r) = 0

The following function is the solution of this equation:
E (r) = E0e±ikx (74.13)
where E0 is a constant complex vector. The plus and minus signs
correspond to different directions of wave propagation. We shall
be interested in waves running in the direction of the x-axis. We
shall therefore take the sign "+"
in the exponent (we wrote the
factor depending on t as e-iwt).
Substituting the solution (74.13) which we have found into (74.5),
we obtain the following expression for E:
E = Re {Eoeil.xe-iwt}
Let us represent the complex vector E0 in this formula as E 0e-ia,
where E 0 is a real constant vector. In addition, let us express the
complex number lc through its real and imaginary parts (k = k 1 +
+ ik 2 ). The result is
E = Re {E 0 e-ittei(kl +ill 2 lxe-iwt}
Let 'us write this expression as follows:
E = Re {Eoe-h 2 xe-i(wt-ktx+al} (74.14)
The real part of the expression in braces yields
i E = E 0e-h,x cos (rot - k 1x + a.) (74.15)
Expression (74.15) describes a wave whose amplitude diminishes
according to the law e-k,x. The phase velocity of this wave is
(74.16)
Hence, the real part of the complex wave number k determines

the phase velocity of the wave, and the imaginary part-the at-
tenuation of the wave.
According to (74.10), with slight attenuation k 1 ~ k = (w/c) V ell,
and the attenuation factor is
k
2
2:rtcr k _ 2:rtcr
~ ----ec;) - -c- l /"7.:"
"f (74.17)
Let us obtain a solution of Eq. (74. 7) in another way. We shall

write this equation as
Ll:E (r) + (e+i 4:rt;/oo) 1100t E(r) = 0
If we introduce the complex permittivity
~ • 4:rtcr
e=e+t-- (74.18)
oo
we can write
~
LlE (r) + :~ w2E (r) = 0

Let us multiply this equation by e-frot and take into account that
E (r) e-irot = -8 2 Eiot 2 , where E = E (r) e-irot [see (74.6)]. As
(t) 2
a result, we arrive at the equation
LlE- Bfl
c2
aats
2
t =o (74.19)
differing from Eq. (71.9) only in that the electric field strength and
the permittivity are not real, but complex. The field E is the real
part of the solution of Eq. (74.19).
We found the solution of Eq. (71.9) in Sec. 73. It is
E = Re {E 0e-i(rot-kx+a>} = Re {E}
where k = V~w/c (the wave is assumed to run in the direction of
the x-axis). Consequently, when e is real, the function
(74.20)
is the solution of Eq. (74.19).
Substitution of the complex value of the permittivity for the
real one, e, does not change the form of the solution. But the real
wave number k must be replaced in it with the complex number
determined by the formula
~
k = k 1 + ik2 = v-=- 7 ell
(!)
(74.21)
264 ELECTRODYNAMICS
where e is the quantity from (74.18). Substitution into formula

(74.20) of the complex number k = k1 + ik 2 fork leads to an expres-
sion coinciding with (74.15).
With a view to (74.18), the square of the complex number (74.21) is
k2 = E)tro 2/c2= { e + i 4 ~0 ) ~tro 2 /c2
= 8Jl(l)2
ct
+i 4l'tOEJl ~ =
eoo c2
p ( 1 + i 4na )
eoo
This value coincides with k in formula (74.8). 2
Hence, with a conducting medium, both the wave number and
the permittivity become complex. The refractive index = i~-t n V
is also complex. Let us write it as
n= n+ i)(=
A v-;:- 1/ (e+
e~t= i (4:rto
; ) ) fJ. (74.22)
Let us square this relation:
(n+ ix) 2 = n2 - )( 2 + 2inx= (e + i 4:rto ) ~ = (I)
e~ + i 4l'tcreJ.L
(1)8
We equate the real and imaginary parts:

nz- xz = en 2n)( = 4mJefl
r• SCll
The system of equations obtained is identical with the system

(74.9). Therefore, its solutions can be found by substituting Efl
for k 2 in formulas (74.10). The result is
n= V-e~V
X=
.. I" 1+ V 1+(4l'to/soo)
V -8'/AV
2
2
.. I" -1+V1+(4:rtot8oo)2
2
l (74.23)
A comp-arison of formulas (74.10) and (74.23) shows that

k (J) k (J) 4
k 1 = ,;- n=-n, k 2 = ,;- X=-X (74.2)
y 8).1. c y 8)1 c
(we have determined k as ,ro e~tlc). V

Substituting for k 1 its value from (74.24) into formula (74.16),
we arrive at the relation
c
v=-
n
from which it follows that the real part of the complex refractive
index ii is the conventional refractive index n of the medium. The
second of formulas (74.24) shows that the imaginary part of

proportional to the attenuation factor of the wave (x CX: k 2 ).
is n
Up to now we dealt with the electric field of a wave. The magnetic
field can be obtained from relation (73.21) by introducing the com-
plex value of the permittivity and the complex function E into it.
By (74.14), we have
Consequently,
B= v.-el-l- [ko, Eoe-1t2xe-i(oot-lllx+o:>]
The complex number V e1-1- can be written as

vel-l-= n + ix = vn2 + x2ei1ll
where '\)> = tan- 1 (x/n). Taking advantage of this, we can writ&
the formula
+
B= V n2 xz. [ko, Eoe-llzxe-i(oot-ltix+o:-1!l)]
from which it can be seen that the vectors B and E oscillate not
in the same phase (as in a dielectric), but with the phase difference'\)>
determined by the expression
tan '\)>= n
X
(74.25)
The relation between the amplitudes of the electric and magnetic
fields is determined by the formula
Bo = V n2 + x 2 E 0
which after introduction of the values (74.23) for n and x become~
,,r- / 4ncr )z
B0 = E0 v e~-t 1 1 + ( ec;;- (74.26}
[compare with (73.21)].
75. Non-Monochromatic Waves t
Any non-monochromatic wave can be represented as the superpo-

sition of monochromatic waves of different frequencies. This opera-
tion is known as the spectral decomposition of the wave.
If the field of a wave is described by a strictly periodic function,
it can be expanded into a Fourier series. In this case, the spectral
expansion contains frequencies forming a discrete series of values:
ro 1 , ro 2 , ro 3 , • • • • The frequencies ron (n =I= 0) are integral multiples
1 Before beginning to read this section, acquaint yourself with Appendix X IV.
266 ELECTRODYNAMICS
of the fundamental frequency ro 0 , i.e. Wn = nro 0 • The fundamental

frequency is determined by the period T of the function describing
the field: ro 0 = 2n/T.
Consider a plane non-monochromatic wave propagating in a vac-
uum in the positive direction of the x-axis. Any quantity charac-
terizing such a wave (E, B, etc.) is described by the function
f=f(t-;) (75.1)
Isee (72.8)1. Upon fixing x, we obtain a function of t describing

the oscillations of the field at a given point. We have presumed that
this function j~is strictly periodic. Assume that its period is T:
f (t + T) = f (t)
According to (XIV.11), we can write f (t) as
+co
f.(t) = ~ Cne-inooot (75.2)
n=-oo
where Cn is a constant calculated by the formula
+T/2
Cn = ~ ) f (t) einwot dt
-T/2
lsee (XIV.14)1.
According to formula (XIV.15), the expansion of the function
j (t) can also be written as
+co
f (t) = ~ C~einooot (~5.3)
n=-oo
The average intensity of a wave is proportional to the average
value of the square of E or B, i.e. is proportional to the expression
([/ (t) )') ~i I (.~~ C,e-••oo•) (j~C~e<m•o<) dt {75.4)
The' integrand equals [f (t)J2. We have taken one of the factors

in the form of (75.2), and the other in the form of (75.3), and in
the second sum denoted the dummy index by the letter m instead of n.
Let us transform expression (75.4) as follows:
+oo +T/2
cn c•m __.!..T jr ef(m-n)!Jlot dt
n, m=-co -T/2
+oo +oo +oo oo
~ C,.Cin6nm= ~ CnC~= ~ ICnj2=C~+2 ~ ICni 2 (75.5)
n, m=-~ n=-~ n=-oo n=i
[we have taken advantage of the orthogonal nature of the system

of functions einwot; see (XIV.13)l.
The result we have obtained signi1les that the average intensity
of a non-monochromatic wave is composed of the intensities of the
monochromatic components.
Now let us consider the case when a wave and, consequently,
the oscillation f(t) too exists during a limited time interval (and
not from t = -oo to t = +oo). In this case, f (t) is not periodic
and can be expanded not into a series, but into a Fourier integral
containing a continuous series of different frequencies. By (XIV.22)
(75.6)
where c(jJ is a function of the frequency (!) determined by the expres-

sion1
(75.7)
{see (XIV.21)]. It is evident that

c_CJ) = c: (75.8)
The total intensity of a wave during the time from -oo to +oo
is determined by the expression
+oo +oo
h:.=) [f(t)]2dt= ~ f(t)f(t)dt
-oo -oo
Substitution of expression (75.6) for one of the factors yields
+oo +oo
lr.=)
-oo
{f(t)/2n J Cwe-iwtdffi}dt
-oo
In the formula we have obtained, it is assumed that integration is

first performed over ffi, and then over t. Let us change the sequence
of integration, i.e. write
+oo +oo
Ir. = J C(j) { y~n J f (t)
-oo -oo
e-iwt dt} d(J)
l The integration variable in formula (75. 7) is customarily designated by

the letter t. We have used a different letter for it to stress that C00 is a quantity
not depending on t.
268 ELECTRODYNAMICS
A comparison with (75. 7) shows that the integral inside the braces
is C_u, i.e. C~. Consequently,
+oo +oo oo
h=) CwC:dw=) 1Cwl 2 dw=2 ~ 1Cwl 2 dw (75.9)

-oo -oo 0
It follows from the formula we have obtained that the quantity

I Cw 12 characterizes the fraction of the total intensity per unit
interval of frequencies.
The radiation of a charge experiencing deceleration can be men-
tioned as an example of a field that can be expanded into a Fourier
integral.
Chapter XIV
RADIATION
OF ELECTROMAGNETIC WAVES
76. Retarded Potentials
We assumed in the preceding chapter that charges and currents

are absent, and in this connection we presumed that the right-hand
side in d'Alembert's equations (71.2) and (71.3) equal zero. Now let
us turn to studying time-varying fields in the presence of arbitrarily
moving charges. In this case, the potentials of the field satisfy the
equations
1 82 q>
,1
'~'
m - - - = -4np
c2 8t2
(76.1)
.1A- _!__ 82 A - - i::_ l' (7G.2)

c 2 8t 2 - c
It is known from the theory of linear differential equations that

the general solution of a non-homogeneous equation equals the
sum of the general solution of a homogeneous equation and a partic-
ular solution of a non-homogeneous one. General solutions of
homogeneous equations were studied in the preceding chapter.
Consequently, to obtain the general solutions of Eqs. (71.1) and
(71.2), it is sufficient to find their particular solutions.
Let us divide the entire space containing charges and currents
into elementary volumes dV' and determine the field set up by
each of the charges de contained in a given dV'. Owing to the linear
nature of the equations, the required field will be the superposition
of the fields set up by all the charges de.
The charge de contained in a given elementary volume dV' is,
generally speaking, a function of time: de = de (t). If we pay no
attention to the presence of other charges de, the charge density
due to the point charge being considered can be written as
p' (r, t) = de (t) 6 (r- r') (76.3)
where r' is the position vector determining the position of the charge
de (t). We have designated this density by the symbol p' to distin-
guish it from the density p(r, t) determining de(t) by the formula
de(t) = p (r, t) dV'.
Substitution of expression (76.3) into Eq. (71.3) yields
1 82 q>
.1cp- C2 at 2 =- 4n de (t) 8 (R) (76.4)
where R = r - r'.
270 ELECTRODYNAMICS .
At all points except the one for which R = 0, the density p' = o.
and Eq. (76.4) has the form
(76.5)
It is obvious that in this case the field has central symmetry relative
to the point R = 0 and, consequently, is a function only of R.
Let us therefore write Eq. (76.5) in spherical coordinates. Using
expression (XI.88) for the Laplacian, we obtain
~2 0~ ( R2 :~ ) - 1
c2 ~:~ = 0 (76.6)
We shall seek the solution in the form
\jJ (R, t)
(jl = R (76. 7)
When this condition is observed, we have

fJq> \jJ 1 fJ\jJ
fJR = - R 2 +R oR
a ( R2 arp ) -
7ilf aR
a (
-7iif -
,1,
'~'
+R iil/1 ) -
oR - -ali
iil/1 + oR
o¢ -l- R oz¢ -
' DR 2 -
R oZ¢
aR 2
ii 2 q> 82 ( \jJ ) 1 8 2¢
at 2 = 7ii2 If = 7f ---a£2·
The introduction of these values into Eq. (76.6) and the cancelling
of 1/R yield the equation 1
fJ2\jJ 1 fJ2\jJ
fJR2 - C2 --at2 = 0
As we established in Sec. 72, the general solution of such an equa-
tion has the form
'i'=!t(t-~)+t2(t+:) (76.8)
[see form.ula (72.8); in the case we are now considering, n = 1].
The fust term in (76.8) describes an expanding spherical wave,
and the second term describes a spherical wave converging at the
poin't R = 0. We are interested in a particular solution. Let us
take the first term of formula (76.8) as such a solution. Introducing
it into (76. 7), we find an expression for qJ:
qJ(r, t)=f(t;_Ric) (76.9)
The solution (76.9) satisfies Eq. (76.5) with an arbitrary choice

of the function f (t - Ric). Let us try to choose this function so that
1 Remember that the point f111' which R = 0 has meanwhile been omitted
from consideration.
RADIATION OF ELECTROMAGNETIC WAVES 271
expression (76.9) will also satisfy Eq. (76.4) at the point R = 0.

We must note that at R ._. 0 the function (76.9) tends to infinity.
Consequently, its derivative with respect to the coordinates be-
comes very large for small R's, so that the term (1lc 2 ) (8 2 cpl8t 2 ) may
be disregarded in comparison with ~rp and Eq. (76.4) can be writte.n
as
Llrp = -4rrde (t) 8 (R)
We have arrived at Poisson's equation (42.4) for the potential
of a point charge. Hence, near the point R = 0, the function (76.9)
should transform into an expression of the kind rp = el R. This
will occur if we assume that f (t - Ric) = de (t - Ric). Conse-
quently, the required solution is
de (t- Ric)
rp = -'---=:R-.....:. (76.10)
It is evident that in the direct proximity of the point R = 0, this

expression becomes
de (t)
rp;::::; ---yr-
To find the potential for the arbitrary distribution of charges

described by the function p (r, t), let us summate the solution
(76.10) over all de= p dV'. We thus obtain
p ( r', t Ir: r' I ) dV'
rp (r' t) = ) --'----;-1-r_---:r,-:-1__;___ (76.11)
Equation (76.2) differs from (76.1) only in the right-hand side

containing j/c instead of the function p. We can therefore directly
write an expression for A, by analogy with (76.11):
1 j ( r', t - Ir: r' I ) dV'
A (r, t) ~c-- I I ,I (7G.-t2)
c J r-r
Expressions (76.11) and (76.12) are called retarded potentials.
The name is due to the fact that the values of the potentials at the
instant t are determined by the values of p and j at earlier instants
in advance of t by the retarded time 't = I r - r' lie needed for the
electromagnetic disturbance to reach the point r from the point r'.
To obtain the general solution of Eqs. (76.1) and (76.2), it is
necessary to add to the retarded potentials the general solutions
of the homogeneous equations that were found in the preceding
chapter. These solutions are not related to the field produced by
the system. They describe the external field acting on the system
and superposed onto the field set up by the system.
272 ELECTRODYNAMICS
In the stationary case (i.e. when p and j do not change with time),
formulas (76.11) and (76.12) transform into expressions (41.10) and
(48. 7), respectively.
We must note that if we had chosen the function / 2 as the partic-
ular solution [see (76.8)1, we would have obtained advanced poten-
tials instead of retarded ones. We shall not stop to consider this
in greater detail.
77. Field of a Uniformly Moving Charge

Assume that the charge e is moving at the constant velocity v
relative to the reference frame K. Let us associate the reference
frame K' with the charge. This frame, like the charge, moves rela-
tive to the frame K at the velocity v.
The charge is at rest relative to the frame K'. Consequently,
the potentials of the field in this frame are
' t ') =-,,
q> ( r, e A (r', t') =0 (77 .1)
r
The most general solution would be A = const, but owing to gauge
invariance, this constant could be taken equal to zero.
To find the potential in the frame K, let us transform the four-
vector
AIL = (q>, A)
to the frame K. For this purpose, we must take the formulas (36.2)
for the inverse transformation, according to which
(r t) = t y 1-~2 y 1-~2
A (r t)= ~<p(r', t')+Ax(r', t') =~<p(r', t')
X t y 1 - ~2 y
1- ~2
Ay =A~= 0, Az =A~ =0
Introducing for q> in these formulas its value from (77.1), we obtain
q> (r, t) = r' Y:-~2~ (77.2)

Ax (r, t) = ~c) e
r' 1-~ 2
Since only the component Ax differs from zero, while the vector v
is directed along the x-axis, the expression for the vector potential
can be written as
A (r t) = ev (77 .3)
' cr' V1-~S
Now in formulas (77.2) and (77.3), we must pass over from the
primed coordinates (i.e. from r') to unprimed ones. According to
the Lorentz transformations (35.14)
'= x-vt ,
X
y1-~2
' y = y, z , = z
Con seq uen tl y,
r' = V x' 2 + y' 2 + z'2 = f/ [(x- vt)IY 1- ~212 + y2 + z2

= V 1 Y(x-vt)2+(1-~2) (yz+z2) (77.4)
1-~2
Substitution of (77.4) into (77.2) and (77.3) yields
<p (r, t) = V (x-ut)2+(1-~2)

e
(y2+z2)
(77.5)
A (r, t) =c y (x-ut)2+(1-~2)
ev
(y2+z2)
(77.6)
A comparison of (77.5) and (77.6) shows that the following rela-

tion exists between the poten-
I(
tials: Poin-t of
observatfon
(77. 7)
Formulas (77.5) and (77.6) can

be simplified by expressing them
in terms of the length of the vector
R drawn from the charge to the
point of observation, and of the X
angle {} between the direction of "" r
this vector and the x-axis. If we
begin to measure the time from Fig. 77.1
the instant when the charge was
at the origin of coordinates, we can see from Fig. 77.1 that
R2 = (x - vW + y2 + z2
Therefore, by (77.4), we have
r'= 1 YR2-~2(y2+z2)= R .. /1-~2 y2+z2
y 1-~2 y 1-~z V R2
The ratio (y 2 + z )/R
2 2 equals sin 2 {}(see Fig. 77.1). Consequently,
r'= R Y1-~ 2 sin 2 {}

V1-~2
18-018
274 ELECTRODYNAMICS
Introducing this value into (77.2) and (77.3), we obtain

(77.8)
(77.9)
Knowing the potentials, we can use the formulas
E= -Vcp-! ~~ (77.10)
B = [VA] (77.11)
[see (56.3} and (56.1)] to calculate the fields E and B. Writing

(77 .10) in components, we obtain
E _ acp 1 iJA1
~-- ax 1 --c-at
Hence,
E:t = - ~-
iJ:r:
.!_ aA:~: = - ~-
c iJt iJx
.!_ ~ ( ..::£)
c at c
= - acp
ax -.!!.... ~
c2 at
(77 .12)
E--~
z- iJz (77 .13)
According to (77.5) and (77.6), cp and A have the same form
'ljl (r, t) = V (x-vt)2+(f-~2)

a
(y2+z2)
where a is a constant (scalar or vector). Introducing the notation

x - vt = 6, we can write that
wherice we conclude that

a.p a.p
-Dt= - v8x
- (77.14)
Substituting for 8cp/{)t in (77.12) its value from (77.14}, we arrive

~t the formula
Es= -~-.!!....~=
iJx c2 iJt
_fiJcp
iJ:r:
+~~=-
c2 iJ:r:
iJcp
iJ:r:
(1-A 2)
P
(77.15)
Introduction of the derivatives with respect to the coordinates

of expression (77.5) into formulas (77.15) and (77.13) yields
e (x-vt)
Ex=(i-~ 2 )
f(x-vt)2+(1-~2) (y2+z2)]3/2 l
Eu = (1-~2)
ey I
((:e-vt)2+(1-~2) (yl+z2)];1/2 ~ (77.16)
Ez= (1-~2)
ez I
[(x-vt)2+(1-~2) (y2+z2))3/2 J
These formulas can be written in the vector form:
E- (1 ~2) eR (77.17)
- - R 3 (1-~2 sin2 {}) 3/2
where R is a position vector drawn from the point where the charge
is to the point of observation (see Fig. 77 .1).
For the magnitude of E, we have
E- e(i-~2) (7718)
- R2 (1- ~2 sin2 'fr) 3/2 •
A glance at this formula shows that on the axis along which the
charge is moving (i.e. at -6- = 0 and :rt),
(77.19)
and in directions perpendicular to the velocity of the charge (i.e. at
-6- = :rt/2),
E- e (77.20)
- R 2 Vi-~~
The field flattens, as it were, in the direction of motion of the charge,
and to a greater extent with increasing v.
Let us find the magnetic field. By (77.11) and (77.7)
B =[VA]= [ V, v:] = ! [Vcp, v] = - ~ [v, V<p] (77.21)

(recall that v is constant).
According to (77 .1 0), we have
(77 .22)
[we have taken into account relation (77. 7), and also the circum-
stance that the velocity is independent of tJ.
We introduce expression (77.22) for -Vcp into (77.21):
B=.!.c [vEJ +..!.[v

c
...!...~]
' c at 2
276 ELECTRODYNAMICS
The second term is the vector product of collinear vectors and there-
fore vanishes. Hence,
1
B = - [vE]
c
(77 .23)
We see that the vector B at each point is perpendicular both to the
vector v and to the vector E. The vectors v and E are collinear
on the x-axis. Consequently, on this axis, B = 0.
If we introduce expression (77 .17) for E into formula (77 .23),
we obtain
B- (1 R2) e [vR] (77.24)
- - t-' c R3 (1-~2 sin2 {})3/2
When v ~ c, Eq. (77.24) transforms into the formula.

B=...:_ [vR) (77.25)
c Ra
We must note that expressions (77.17) and (77.25) can be obtain-
ed with the aid of formulas (62.2) for the transformation of fields,
proceeding from the fact that E' = er' /r' 3 and B' = 0 in the
frame K'. We invite our reader to do this as an exercise.
78. Field of an Arbitrarily Moving Charge

Let us find the field of a charge moving with acceleration. Assume
that motion occurs along the trajectory r 0 = r 0 (t) (Fig. 78.1).
We shall calculate the field at the point
of observation P determined by the posi-
tion vector r. Owing to retardation, the
potentials at the point P at the instant
~-__;~::..:.....~.-pP t are determined by the position and
velocity of the charge not at the same
instant t, but at the earlier instant t 0 •
The latter must satisfy the condition
c't = c (t - t 0 ) = R (t 0 ) (78.1)
where R (t 0 ) is the distance from the
poirit at which the charge is at the
Fig. 78.1. instant t 0 to the point of observation
P. The retardation is determined by
the time 't needed for the perturbation to reach the point r from
the point r 0 (t 0 ).
It is a simple matter to see that c-r and the vector R (t 0) form the
four-vector
RiJ. = (n, R) (78.2)
Indeed, the four-vectors
xiJ.={ct, r} and xf:'={ct 0 , r 0 (t 0)}
are two four-position vectors. The vector (78.2) equals the difference
between these two four-vectors and, consequently, is a four-vector
itself.
It can be seen from (78.1) that the square of the four-vector (78.2)
is zero:
(78.3)
(this expression is in essence the square of the interval between the
events of the appearance of a signal at one point and its arrival at
another point).
Let us associate the reference frame K' with the charge. In this
frame, the charge is at rest and, consequently, the potentials are
determined by the expressions
m' (r' t') =-e-= _e_ A' (r' t') = 0 (78.4)

't' ' R' c't' ' '
where r' and t' are the position vector of the point and the time of
observation determined in the frame K', and R' is the distance be-
tween the point where the charge is and the point of observation.
Hence, the four-potential of the field in the frame K' is
A'~'=f-e
\ C't' '
o) (78.5)
Let us attempt to find an expression for the four-potential such

that would transform into expression (78.5) when v = 0. It is evi-
dent that in an arbitrary frame, the potentials should depend on the
velocity v of the charge. Since we are seeking a four-dimensional
expression for the potential, the velocity must be taken as the four-
velocity
u~'= ( V1~v2Jc2' V1_:v2fc2)

If we assume that the four-potential is proportional to the four-
velocity, at v = 0 the vector potential A will also be zero. Further,
the denominator of the time component of the four-potential (78.5)
contains the time component of the four-vector (78.2). Consequently,
the four-vector R!J. should he introduced into the denominator of
the required expression for A~', and for the correct dimension to be
obtained, it must be multiplied by the four-velocity (we have already
introduced the four-velocity into the numerator).
11 The four-potential must thus be determined as follows:
(78.6)
278 ELECTRODYNAMICS
Substitution of values for Rv and Uv yields

~ R\' Uv = 1 (c 2"C- Rv)
v
V 1-v2jcl
Consequently, expression (78.6) can be written as
AJJ. = V1- v2fc2 eu!J.
c 2 't-RV
whence
AJJ. _ ( ec ev ) (78. 7)
- c2't-RV ' c2't-RV
It is not difficult to note that when v = 0, expression (78. 7) does

indeed transform into (78.5).
Taking into account that c"C = R, let us write the expressions
for the potentials as follows:
~ (r, t) = (R-eR~) to (78.8)

e~t
A (r, t) = (R-R~>t. = ~~ (78.9)
(we have used the notation f3 = v/c).
Expressions (78.8) and (78.9) are known as the Lienard-Wiechert
potentials. To obtain the values of the potentials at the instant t,
the values of v and R on the right-hand sides of (78.8) and (78.9)
must he taken at the instant t 0 determined by the condition
to=t-"C=t __ R(to) = t - I r-ro<to) I
c c
Let us write this condition as
F (x, y, z, t, t 0) = t - t 0
_ ([x-x 0 Uo)P+fY-Yo (to) J2+[z-so(to)J 2} 112 =0 (78.10)
c
The right-hand sides of (78.8) and (78.9) are functions of t 0 , and
the latter, in turn, is a function of x, y, z, and t (here x, y, z are
the_ coordinates of the point of observation):
t 0 = f (x, y, z, t) (78.11)
Relation (78.10) is an implicit expression of the functional rela-
tion (78.11).
When finding the values of the fields by the formulas
E= -V'cp-: 0:, B= [V'A]
it is necessary to calculate expressions of the form {)r:p/oxi, oA/ot,

and oA ~tloxt. where r; is the coordinate of the point of observation.
Since the functions q> and A depend on x, y, z, _ in a complicated

way-in terms of t 0 = f (x, y, z, t), we have to perform the calcu-
lations according to the following scheme:
aq> acp at 0
ax, = ato ax, (78.12)
aA aA at 0
ar=at;"a~ (78.13)
aAk aA, at0
axi =at; axl (78.14)
It follows from relation (78.12) that
aq>
\lcp = ato 'Vto (78.15)
We shall therefore need the values of the derivatives ot 0/ox 1 and

a t 0 /ot, which can be found with the aid of relation (78.10). Let us
write this relation as
_ t ( 2J [xi -Xoi (to)J2 J112
F(x 1 , x 2 , x 3 , t, t 0 , ) - t - o - c
{R2)1/2
= t - t0 - c =0 (78.16)
According to the rules for th.e differentiation of an implicit func-
tion, we have
ato aF/axi
axi = - fJF/8t 0
Differentiation of (78.16) with respect to t 0 yields
iiF ~ (xt-Xot) (ox 0tfot 0)
9t 0 = - f + cR
Taking into account that x 1 - x 01 = R, and fJx 01 /ot 0 is the t-th
component of the velocity of the charge at the instant t 0 , we arrive
at the expression
!..!_= _ 1 + Rv(t 0) = _ R-R~
(78.17)
Ot 0 cR R
Similar calculations yield

aF Rt
ax, = - cR (78.18)
Dividing (78.18) by (78.17) and reversing the sign, we obtain
ot 0 Rt
axi = - c (R-R~)
280 BLECTRODYNAMICS
whence it follows that

R
"Vto=- c(R-Rj}) (78.19)
Now let us ftnd ot 0 /ot. According to the relevanfrule for the differ-
entiation of an implicit function, we have
~at 9 aF/ot
-at=- oF/ot0
The derivative of the function (78.16) with respect to t is unity:

oF/ot = 1. The derivative 8F/ot 0 is determined by expression (78.17).
Consequently,
(78.20)
Now we shall commence calculating the values of o<plot 0 and oA/ot 0 •

By (78.8), we have
:~ = - (R~Rjl)2 ( :~- :~ ~-R :~)

We can see from the relations R = c (t - t 0 ), R = r - r 0 (t 0) that
oR oR 8r
~= - c, --at;= -Tt;0 = -v (to)
In addition, o~lot 0 = ~ = ~/c, where ; is the acceleration of the

charge at the instant t 0 • Hence,
ec (1-P'+R~/c)
0
8q> _ e ( -c+v~J-Rj})
Tt;-- (R-RW (R-Rp) 2
(78.21)
By (78.9), we have
Substitution of expression (78.21) for Ofp/ot 0 and (78.8) for <p yields
bA _ ecp (1-P 2 +R~!c) + e~ _ p (1-P 2 +R~!c)+$/c) (R-Rjl)
ut,- (R-Rjl)' R-Rp -ce (R-RIJ)'
(78.22)
Finally, we can write an expression for E:
E= -V _,!_ aA = -~Vt _..!_ aA ~
fP c at oto 0 c oto at
RADIATION OF ELECTROMAGNETIC W A YES 28t
[see formulas (78.15) and (78.13)}. Introducing expressions (78.21)~

(78.19), (78.22), and (78.20) into this equation, we obtain
E = ec (1- ~ 2 +R~/c)
(R-R~)~
[ R
c (R-R~)
J
fl (1- ~~+ R~/c) +(~/c) (R -R~) R
-e (R-Rfl)~ • R-R~
R (i-~ 2 +R6/c)-R~ (i-~ 2 +RP/c)-(R~/c) (R-R~)

= e (R-RB)3 (78.23}
By grouping together the terms containing the acceleration v
of the charge (i.e. ~}, we can give the expression for E the form
E_ (R-R~)(i-~ 2 ) + (R~/c) (R-R~) -(R~/c) (R-R~)
- e (R-R~)s e (R-R~)a
The numerator of the second term can he written as the vector triple-
product [R, [(R - R~}, ~/ell. This can be verified by expanding
the product according to formula (VI.5). As a result, we obtain the-
final expression
E _ (R-R~)(i-~ 2 ) +
[R, [(R-R~), p;c)]
(78.24}
- e (R-RW e (R-1\~)a
•
Remember that the values of R, ~ (i.e. v), and fi (i.e. v) in this for-
•
mula must be taken for the instant t 0 = t - 't.
The field described by expression (78.24) consists of two parts~
The first depends only on the velocity of the charge and at large dis-
tances diminishes as 1/R 2 (i.e. as a Coulomb field). The second part~
besides the velocity of the charge, also depends on its acceleration
and at large distances diminishes as 1/R (i.e. as the field strength in
a spherical electromagnetic wave).
If the charge moves uniformly, the second term in formula (78.24)
vanishes, and the field is determined by the expression '
E (Ro-R 0 ~) (1-~ 2 )i
=e (Ro-Ro~)a
(78.25}
Here the subscript "0" on R stresses the circumstance that the value
of R is taken for the instant t 0 • We have not used the subscript "0'•
on fi because in uniform motion ~ does not depend on t.
In Sec. 77 we obtained formula (77.17) for the field of a uniformly
moving charge. It can be written as
E- e R(i-~2) i {(826
7 . )
::r;:-~..-. R3 (1-~2 sin 2 {l-)3f2
Formulas (78.25) and (78.26) greatly differ from each other in their
appearance. It is a simple matter to show, however, that they are-
282 ELECTRODYNAMICS
actually identical. The matter is that in formula (78.25) the field

is expressed in terms of the distance from the charge to the point of
<1bservation taken at the instant t 0 = t - 't. In formula (78.26),
however, the field is expressed in terms of the distance from the charge
to the point of observation taken at the instant of observation t.
To prove that formulas (78.25) and (78.26) are identical, let us
eonsider Fig. 78.2. The distance OP is R 0 , eP equals R, and the seg-
ment Oe is the path travelled by the charge during the retarded time •·
Taking into account that 't = R 0 /c, this path can be written as R 0 ~.
Fig. 78.2.
The vectors R 0 , R, and R 0p are related by the expression R =

= R 0 - R 0 p, whence the identity of the numerators of formulas
(78.25) and (78.26) follows. .
·.. The length of the segment OQ equals the projection of the vector
R oP onto the direction of the vector R0 :
OQ = R 0 ~ cos a = RoP
Consequently,
(78.27)
{compare with the denominator of formula (78.25)1.
Let us express the length of the segment QP in terms of R. A
glance a{ Fig. 78.2 shows that
(QP) 2 = R 2 - b2 = R 2 - (RoP!sin a) 2
Further, it also follows from the figure that R 0 sin a = R sin~.
so that
(QP) 2 = R 2 - (Rp sin '1'}) 2
whence
QP=RV1-~ 2 sin 2 '1't {78.28)
Finally, equating the right-hand sides of expressions (78.27) and
(78.28), we arrive at the relation
Ro-RoP=RV1-~ 2 sin 2 '1'}
from which the identity of the denominators of formulas (78.25)

and (78.26) follows. We have thus proved the identity of expres-
sions (78.25) and (78.26).
Let us pass over to finding the field B. By formnla (XI.56), we have
fA depends on the coordinates in terms of t 0 , like a in formula (XI.56)

depends on the coordinates in terms of ~].
Introducing (78.19) and (78.22) into the last expression, we obtain
R ~ (1-~ 2 +R~/c)+(6!c) (R-R~)J

B= [ - c (R_:_R~) ' ce (R-R~) 2
= [~ -Rjl (1-~ 2 +R~/c)-(R~/c) (R-Rjl)J
_R ' e (R-RJl)3
Let us add in the numerator of the second factor the term R (1 -

- ~ 2 + R61c). This will not change the expression because [RRJ =
= 0. But the second factor will now transform into E [see (78.23)1.
We .thus arrive at the relation
B=n~. ~]
I •~c:.
(78.29)
Here R = R (t 0 ). Examination of (78.29) shows that the vector B

at each point is perpendicular to the vector E and to the vector drawn
from the point where the charge was at the instant t 0 to the point of
observation.
79. Field Produced by a System of Charges

at Great Distances
IAssume that we have a system of moving charges that do not leave
the confines of a certain volume in their motion. We shall presume
that the system as a whole is neutral. Let us consider the field pro-
duced by such a system at distances that are great in comparison with
its dimensions. We shall place the origin of coordinates inside the
system and characterize the distribution of the charge with the aid
of the function p = p (r', t). Hence, the charge inside the volume
dV' at the point with the position vector r' will be de (t) =
= p (r', t) dV'. Let r stand for the position vector of the observation
point P.In addition, we shall introduce the notation R = r - r'. It
is obvious that R is the vector drawn from de to the point P.
ELECTRODYNAMICS
Let us write expressions (76.11) and (76.12) for the retarded poten-
tials of the field produced by the system:
cp(r, t)= J p(r', t~R/c)dV' ( 79 . 1)
A(r, t)=! f j(r', t-R/c)dV' (79.2)

.c J R
According to our assumption, r ~ r'. Therefore, . the quantity
R = I r - r' I can be considered as the value of the function f (r) =
= I r I = r at the point r + 6r, where 6r = -r'. Taking advan-
tage of the formula f (r + 6r) = f (r) + 'Vf (r) 6r, we can write
R = lr-r' I= r + \lr ( -r') = r-..!..r'
r
= r-nr' (79.3)
where n is the unit vector of the position vector r.

Substitution of (79.3) for R in the formulas for the potentials
yields
p ( r', t - : + n:' ) dV'
cp (r, t) = JI r-nr
, (79.4)
1 J. ( r,
, r
t -- + -or' ) dV'
A (r t) = - f c c (79.5)
' c J r-nr'
We see that the retarded timeT consists of two parts. One of them,
equal to T0 = ric, does not depend on r' and is called the retarded
time of the system. It determines the time needed for an electro-
magnetic perturbation to travel the path from the origin of coordi-
nates to the point of observation. The second part, equal to_:t = -nr' /c,
is called the proper retardation. It characterizes the time nee'cled
for propagation of the perturbation within the limits of the system.
Let us expand the integrand in (79.4) into a series in the ratio
r' /r. Considering the quantity -nr' as the small increment Br
"f the argument r, we obtain
r: ....
p(r', t - r~nr') p(r', t-7)
r-nr' - r
8 [ P ( r' • t- ; ) ] ,
+a,; r (-nr )+ ... (79.6)
To see whether we can limit ourselves to the written terms in the

expansioa, we must assess the following terms. The latter will con-
tain higher derivatives of p with respect to r. It is easy to see that
the derivatives of p with respect to r are proportional to the deriva-
tives of p with respect to t. Indeed, assuming that t - ric = ~. we
can write
ap ap
7it a;
whence
ap 1 ap
a;=--; 7it (79.7)
Similarly
a2p = ( _..!_)2 02p amp= ( _ _!_)no amp
or' c ot 2 , ••• , arm c atm
The consecutive differentiation of the function p/r with respect
to r with the following replacement of the derivatives with respect
to r with those with respect to t yields
_!_
or
(.2..)
r
= - ..£_+.!. ( _ _!_) ~
r r .
2 c at
:r: (~ )= • ' ' ' ' ' ' ' + ~ ( - ! )m~7~

(in the last line we have written only the last term).
At large values of r, the first terms in the derivatives we have writ-
ten are much smaller than the last ones. Therefore, our task con-
sists in assessing the relative magnitude of expressions of the form
1 )m -
-r1 ( - - - ( -nr ')m-
amp --1 ( 1 )m - - n( r ')m
amp (79.8)
c atm r - c atm
[in expansion (79:6), the m-th derivative is multiplied by (6r)m].
Assume that p varies with time according to the harmonic law
p oo cos wt. Therefore, the m-th derivative of p with respect to t
will be of the order of wmp. Substitution into (79.8) yields
~ (r
~ (nr')m"' ~ ( w; r
where l are the linear dimensions of the system of charges being con-
sidered.
The following term of the expansion will be of the order
~ (~~ r+i
Thus, the ratio of the consecutive terms of expansion (79.6) in the
order of its magnitude is wllc. Replacing the frequency with the
period of the variation of p (by the formula w = 2n/T), we obtain
wZ 21tl Z
-=---
c cT cT
286 ELECTRODYNAMICS
What has been said above shows that the subsequent terms in the
expansion (79.6) may be disregarded if the condition
clT ~1 (79.9)
is satisfted. The ratio lie determines the proper retardation time 't'.
Consequently, condition (79.9) can be written as
't ' « T (79.10)
It can be seen from (79.10) that we may limit ourselves to the ftrst
terms of expansion (79.6) when the time needed for the propagation of
an electromagnetic perturbation within the limits of the system is
much smaller than the time during which the distribution of the
charges in the system changes appreciably.
Condition (79.9) can be written in two other ways. The product cT
gives the wavelength ')..of the radiation produced by the system. There-
fore, inequality (79.9) can be written as 1 ~- • ~
f~ fh-- V\ V..IYJ
(79.11)
lthe dimensions of the system must be much smaller than the wave-
(ength).
Finally, having in view that liT in the order of magnitude equals
the velocity v of the charges in the system, instead of inequality
(79.9) we can write
v«c (79.12)
A glance at the last relation shows that by interrupting expansion
(79.6) at the second term, we have limited ourselves to considering
the radiation of a non-relativistic system of charges.
Let us again turn to the calculation of the potentials, assuming
conditions (79.10)-(79.12) to be satisfied. The substitution of (79.6)
into (79.4) yields
q> (r, t) = ~ Jp ( r', t- : ) dV'- :r {.,! Jp ( r', t- : ) or' dV'}

. (79.13)
(we remind our reader that integration is performed over the primed
coordinates, therefore r can be put outside the integral; in addition,
we have changed the sequence of differentiation with respect to r
and of integration).
The density of the charge at the instant t - (ric) is inside the
first integral. Consequently, this integral gives the total charge of
the system, which owing to the presumed electroneutrality of the
system is zero. We must therefore retain only the second term in
formula (79.13). Putting n in it outside the integral and the deriva-
tive, we arrive at the expression

cp (r, t) =- n :r {~ ~ p ( r', t- ~ ) r' dV'}
The integral in this expression is the dipole electric moment which
the system had at the instant t - ric: o~ r ~_. fh:)l
\ f\ \ p ( t- : ) =) p ( r', t- : ) r' dV'
[compare with (43.6)]. We can therefore write

cp ( r, t) {} [
=-nTr p (t-r/c)
r
J (79.14)
Finally, having performed differentiation and taken into considera-
tion that aplor = -(1/c) ap/ot = -(1/c) [compare with (79.7)1~ p
we obtain
t) = np (t-rfc)
.
+ np (t-r/c)
(r
cp ' r, cr
The first term in this formula coincides with the potential (43. 9)
of a static dipole (n = rlr). We must note that the field corresponding
to this term at the distance r and the instant t is determined by the
value of the dipole moment at the instant t- ric. The first term dimin-
ishes with an increasing distance r much more rapidly than the
second term. Therefore, considering the field at great distances, we
can assume that
.
t) _ np (t-r/c)
cp (r 1 - cr (79.15)
Let us go over to ·determination of the vector potential. Formu-
la (79.2) differs from (79.1) only in that j (r', t - Ric) is inside the
integral instead of p (r' t - Ric). Consequently by expanding the
1 1
integrand into a series, we obtain an expression similar to (79.13):

A (r 1 t) = c~ ~ j { r', t- : ) dV'- :r { cir ) j ( r', t - : ) (nr') dV'}
(79.16)
If the currents were stationary, i.e. did not depend on t, the first.
integral would vanish [see (51.5)]. For non-stationary currents, how-
ever, this integral differs from zero. We may therefore retain only the-
first term in expansion (79.16) 1 • We can thus assume that . \
AV", t)= c~ Jj (r', t-:) dV' (6 }

& In formula (79.13), we could not disregard the second term because the
first one vanished.
'288 ELECTRODYNAMICS
j
We shall prove that j (r', t - r/c) dV' equals the time deriva-
tive of the dipole moment of the system taken for the instant t - ric.
It will be the simplest to prove this by passing from the continuous
distribution of the charges to a discrete one. Let us perform the sub-
stitution
Jr j dV' = jr pvdV'-+ ~ e.. v ..
(the velocities of the charges, like the function pv, must be taken
for the instant t - r/c). However,
• d •
'lj e4 v 4 = ~ e.. r~ = Tt ~ e.. r~ = p (t- r/c)
Consequently,
A(r, t)= p(t-r/c) (79.17)
cr
A comparison with (79.15) allows us to write
cp =An (79.18)
The potentials (79.15) and (79.17) are determined by the value
of the time derivative of the dipole moment of the system. This is
why they are called potentials calculated in a dipole approximation.
The dipole approximation is allowable when the conditions (79.10)-
(79.12) are observed.
80. Dipole Radiation

The region of a field that is at a distance r from the radiating system
much greater not only t.han the dimensions of the system l, but also
than the radiated wavelength (r » 'A » l) is known as a wave zone.
In this zone. conditions are observed in which the dipole approx-
imation treated in Sec. 79 holds. In this approximation
np (t-r/e) A= p(t-r/c) (80.1)
cp = cr ' cr
{see (79.15) and (79.17)].
To calculate E, we must find 'Yep and 8A/8t. Using formula (XI.51),
we obtain
8!p 8!p
'V<p=-'Vr=-
fJr Br
n
(recall that n = er = r/r). Hence,

'V
cp
= ..!.._ (
or
np (t-r/c))
cr
0 = _ np
cr 2
0 _ np 0
c•r (80.2)
[we have taken advantage of the circumstance that 8p/8r =

= -(1/c) 8pl8t].
The first term in the expression we have obtained diminishes with
an increasing distance much more rapidly than the second one. It
may therefore be ignored for great distances, and we may consider
that
" np (1-r/c)
v cp = - c2r n
The derivative 8A/8t = p.lcr. Hence,

E = _ V' _..!. aA _ (np) n ·-
..
..!_ L _
.. ..
(np) n - p
cp c at - c2 r c cr - c2 r
The numerator of this expression can be written as [n [np"J]. This

can readily be verified by expanding the vector triple product using
formula (V 1.5) and taking into account that nn = 1. The electric
field is thus determined by the formula
1 •• .1 ••
E = -c2- r [n, [np]] =--'--2 [[pn], n]
cr
Let us go over to calculation of the magnetic field. The vector
potential is a function of r. Therefore, by (XI.56), -we have
B=[V'AJ=[Y'r, ~~J=[n, ~~] (80.3)
Differentiation of expression (80.1) for A yields
aA = .!_ ( p (t-r/c) )
. = _ _L _
. _L
ar ar ' cr cr 2 c2 r
[compare with (80.2)1. Discarding the term proportional to 1/r2 , we
find that aA!ar = _:_plc 2r. Hence,
1 •. 1 •.
B = ---
c2 r
[np] = -c2-r (pn]
We shall write the ftnal expressions for E and B:

1 .• 1 .•
E=l!llpn], n],
c r
Bc·=--[pn]
c2 r
(80.4)
(remember that the values of.p must be taken for the instant t - ric).
A comparison of these expressions leads to the conclusion that
E = [Bnl (80.5)
whence it follows that the vector E is perpendicular to the vector B.
Examination of expressions (80.4) sho\YS that the vectors E and B
are perpendicular to the vector n [the perpendicularity of E to n
19-01~
290 ELECTRODYNAMICS
can also be seen from (80.5)]. Hence, as in a plane wave, the vectors B,
E, and n are mutually perpendicular [see formula (72.16)1]. In addi-
tion, the vectors B and E, as in a plane wave, are identical in magni-
tude, and
(80.6)
..
where {} is the angle between the directions of the vectors p and n.
That the relations observed for a plane wave were found to hold
for the field we are studying is not surprising. At distances that are
great in comparison with the
dimensions of the radiating sys-
tem, a wave must be spherical.
At the same time, provided that
r » 'A., small portions of the
spherical wave virtually coincide
with a plane wave.
It can be seen from (80.4) that
the fields E and B are determined
by the second derivative of the
dipole moment of the system.
This is the reason whv the radia-
tion being considered is called
dipole radiation.
Fig. 80.1. The dipole moment is deter-
mined by the expression p= ~er'.
Consequently, ·.p = lje~i = lje~. It thus follows that charges- emit
electromagnetic waves only when they move with acceleration.
To comprehend the pattern of the field at great distances, let us
introduce a spherical system of coordinates, measuring the polar
p.
angle~ from the direction of the vector (t- r/c) (Fig. 80.1). By
(80.4), tile vector B is perpendicular to the plane determined by the
vectors p and n. Consequently, B is directed along a tangent to a
"parallel", the vectors p, n, and B forming a right-handed system.
Examination of (80.5) shows that the vectors B, n, and E form a
right-handed system. Hence, it follows that E is directed along a
tangent to a "meridian", the directions of ·p
and E on the equator
being opposite. We stress once more that the vectors depicted in
·p
Fig. 80.1 relate to different instants: to the instant t - ric, and B
1 We multiply relation (80.5) by n and use formula (VI.5):
[nE]=[n, (Bn]]=B(nn)-n(nB)=B
We have arrived at formula (72.16) (in a vacuum VEf.l = 1, H = B).
and E to the instant t. The magnitudes of the vectors B and E are

proportional to sin -fr [see (80.4)1. Therefore, the fields have the maxi-
mum value at the "equator" and vanish at the "poles".
To determine the intensity of radiation in different directions and
the total radiated power, let us calculate the Poynting vector. With
a view to (80.6), we obtain
c
S=-,- [EB] =4- EBn= 4 3 2
c p2 sin2 {} n (80.i)
'!ll n nc r
Hence, the intensity of dipole radiation is proportional to sin 2 ofr.
The diagram showing the intensity as a function of -fr has. a two-lob~d..
pattern. J · ·· · " · ·'
To determine the radiated power P, let us find th€ 1 energy flux
th.rough t~e enti~e .spherical surface. The area of a ~eri~_al band. ~f
w1dth d-fr lS 2nr sm -fr dofr. Consequently, _.p::::: _; 0 c 3 J c,r:_l
:n: •• ••
P =
i r
J S df = J
p2 sin2 {} . .
4ncar2 2nr2 sm -fr dfr =
2p2
3c3 (80.8)
0
Assume that of all the charges of the system only one has accele-
ration. Hence,·.;= ~ea;a = e~ and the radiated power is
(80.9)
.
This formula also holds when there is only one charge moving with
the acceleration v.
81. Magnetic Dipole and Quadrupole Radiations
If the properties of a system of charges are such that p = 0, no

dipole radiation is produced. This does not signify, however, that
there is no radiation at all. In this case, we must take into account
the terms of the expansion of the potentials which we disregarded in
the dipole approximation.
We established in the preceding section that in the wave zone a
wave in small regions is close to a plane wave, for which relation
(80.5) holds. This relation makes it quite simple to find E if we know
B. To find B, on the other hand, it is sufficient to know only A.
We shall therefore limit ourselves to finding the vector potential.
Consider the second term of formula (79.16) which we disregarded
in the dipole approximation. Since in the case we are interested in
the first term in (79.16) is zero, the vector potential is
A(r, t)=- :r {:r J (r·, j t-+)(nr')dV'}

292 ELECTRODYNAMICS
Differentiation with respect to r yields two terms. One of them is pro-

portional to 1/r2 , and the second is proportional to 1/r. The first term
diminishes with the increasing distance much more rapidly than the
second one. We therefore disregard it, as we have already done many
times. In addition, we take into account that ojlor = -(1/c) oj/at.
The result is
A(r, t)=++ Jl" j(r', t-rlc)(nr')dV'
cr ul
To simplify our further calculations, let us pass over from a con-

tinuous distribution of the charges to a discrete one. The expression
for A therefore becomes
A (r, t) = c;r :t ~ eava (nr~) (81.1)
l1
The values of v a and r~ are taken for the instant t - r/c.

The expression v a (nr~) can be written as follows:
Va (nr~) = :t {r~ (nr~)}-r~ (nva) (81.2)
Let us divide v a (nr~) into two equal parts and replace one of them
with half of expression (81.2):
Va (nr~) = ~ Va (nr~) + +:t {r~ (nr~)}-+ r~ (nva)

The first and third terms can be written as a vector triple pro-
duct: ~ [n, [v a' r~]]; this can be verified with the aid of formula
(VI.5). Hence,
Va(nr~)=+[n, fvar~ll+ ~ :t {r~(nr~)}

Substitution of this expression into formula (81.1) yields
• 1 {)
A (r, t) = 2c2r at~ ea [n, [var~]]
a
+ 2: 2 r :t: ~ear~ (nr~) =Am+ AQ

a.
(81.3)
The meaning of the subscripts m and Q will be revealed below.

Let us put n outside the sum sign in the first term and exchange
the places of v a and r~. As a result, this term becomes
Am=- 2c~r ~ [ n, ~ ea [r~va] J=

a
:r :t [ c ~
2
1
n
ea [r~va], n J
The first factor is the magnetic moment m of the system [see for-
mula (51.15)1. We can therefore write that
1 •
Am=-
cr
[mn) (81.4)
where ;n = omlot is taken for the instant t - r/c. The potential

Am is thus determined by the changes in the magnetic dipole moment.
Therefore, the relevant radiation is called magnetic-dipole radiation.
For magnetic-dipole radiation, the magnetic fwld is
Bm=[VAml=l_vr, a!m J=r_n, !; J 8
[compare with (80.3)). If we disregard in oAmlor the term propor-

..
tional to 1/r2 and replace D/Dr with (-1/c) D/Dt [recall that m =
= m (t- ric)), we obtain
1 •• 1 ••
Bm = - c-
2r
[n ' [mn]J = -.-.
c-r
[[mn), n)
Using relation (72.16) equivalent to (80.5), we find that

1 •• 1 .•
Em= --.-
c~r.
[mn) =c2-
r
[nm)
Hence, the fields in magnetic-dipole radiation are

1 •• .
c2- r [nm),
Em = - }
1 •• (81.5)
Bm = -c2- r [[mn), n)
..
A comparison of the results obtained with formulas (80.4) for dipole
..
radiation shows that Bm is expressed in terms of m by a formula simi-
lar to the one expressing E in terms of p. The formulas for E, and B
differ, apart from the substitution of ·~1 for·,;, in the sign. lt thus fol-
lows that a pattern of the field in magnetic-dipole radiation in the
wave zone can be obtained by substituting ;l~ for.p, Bm forE, and -Em
for B in Fig. 80.1.
Let us turn to the second term of formula (81.3), i.e. to the expres-
sion
(81.6)
a
The addition to AQ of the expression 1 (r) n [here I (r) is any function

of r) will not change BQ because (\7, f (r) n) = 0 (see (XI.54)1.
294 ELECTRODYNAMICS
Let us take as f (r) the function

f (r) = - 1
6c2r {)
""Ft2 ""'
L..
a
[r~ is a function of the argument (t - ric), which when acted upon

by the operator 'V behaves like a function of r]. Multiplication of
this function by n and the addition of it to (81.6) yields
AQ = 6: 2 r a~: 2j ea {3r~ (nr~)- r~2 n}
(!
Let us write an expression for the i-th component of the vector AQ:
AQt = Gc12 r : 122 ~ ea { 3x~i ( 2j n~~.x~lr.)- r~2n 1 }

a h
The last term in the braces can be written as 2jr~2 6 1 ~r.nk. Hence, A Qi
k
will appear as follows:
AQi ='= fic\r ::2 ~ ea { 3x~i ( ~ nkx~k)- ( 2j r~2 6u,nk)}

a k k
= fic2r
1 {)2 ""'
{Jt2 LJ LJ ea {3 XaiXak-
nk ""' ' ' r,'2 Utk
$1 }
k a
But the sum over a is Q1k - a component of the tensor of the quad-
rupole moment of the system [see formula (43.13)]. Consequently
1 i)2 "\' 1 {)2
AQI = 6c2r {)t2 LJ Qihnk = 6c2r 8t2 Qt
k
where Qt is a component of the vector obtained when the vector n

is multiplied by the tensor Q 1 ~~.. Denoting this vector by the symbol
Q, we can write
1 ••
Aq = 6c 2 r Q (81. 7)
The potential AQ is determined by the changes in the quadrupole

moment of the system. For this reason, the corresponding radiation
is called quadrupole radiati or•.
The magnetic f1eld of quadrupole radiation is
BQ=['VAQl=[ Vr, a:rQ]=[n, a~o]
In calculating aAQ!fJr, we disregard the term proportional to 1/r2
and replace the derivative with respect to r with one with respect to
t. The result is
1 •••
Bq=-6
c3 r [Qn]
(81.8)
(we have transposed the factors to avoid writing the minus sign in-
troduced by the factor -1/c).
Taking advantage of relation (80.5), we find that
1 .••
Eq = Bc 3 r [lQ n], n] (81.9)
In comparison with formulas (80.4), expressions (81.8) and (81. 9)
have the additional factor 1/6c. In addition, they contain ·o_· instead
of p. Otherwise, the formulas for the fields in quadrupole radiation
are identical to the relevant formulas for dipole radiation.
Calculations show that the total power of all three kinds of radia-
tion (including dipole radiation) is determined by the expression
2 •• 2 •• 1 ...
P= 3ca P2 + 3ca mz+ 180c 0 ~ Q ik (81.10)
i, k
Let us assess the relative intensity of various kinds of radiation.

We shall assume for simplicity that the charges of the system move
according to a harmonic law. Hence
r' = l cos rot, v = lro sin rot
where l is a quantity of the order of the system's dimensions. The
average values of the magnitudes of cos rot and sin rot are quantities
of the order of unity. Therefore in the final expressions characterizing
the order of magnitude of the expressions being considered, we shall
discard the factors cos rot and sin rot. For instance, if we do not have
in mind the necessity of time differentiation, we can write that
r' ,.., l, v ,.., lro (81.11)
It can be seen from definition (43.5) that p is a quantity of the
order of er', i.e.
p ,...... el cos rot, p,._. elro sin rot, p
.. elro cos rot
~ 2
Hence we find that with respect to tho order of magnitude, the power
of dipole radiation is determined by the expression
(81.12)
[see (81.11)].
It follows from definition (51.15) that the magnetic moment has
a magnitude of the order of er'v!c, i.e.
m "' .1..
c
el2ro cos rot sin rot~ .1..
c
el2ro sin 2rot
• 1 •• 1
m ,...... - el 2 ro 2 cos 2rot m ~ - el2ro3 sin 2rot
c ' c
296 ELECTRODYNAMICS
whence
(81.13)
A comparison of expressions (81.12) and (81.13) :>hows Lhat
Pm:Pp=(+r=1
Recall that we have calculated the fields of radiated waves for the
non-relativistic case (i.e. for v ~c). Consequently, in the case we
have studied, the intensity of magnetic-dipole radiation is much
lower than that of electric-dipole radiation.
According to definition (43.13), the quadrupole moment (and,
therefore, the vector Q) is a quantity of the order of er' 2 , i.e.
Q ,..., el 2 cos 2 ffit,
Q,. . . , eP(J) cos ffit sin ffit ,...., eZ2ffi sin 2ffit
'(! ,.., eZ2ffi 2 cos 2ffit, '(!',......, e2 l 2 ffi 3 sin 2ffit
Hence,
(81.14)
Quadrupole radiation thus has an intensity of the same order of

magnitude as magnetic-dipole radiation.
APPENDICES
I. Lagrange's Equations for a Holonomic System

with Ideal Non-Stationary Constraints
For non-stationary constraints, conditions (4.2) have the form

/1 (x1 , x 2, ... , Xn, t) = 0 (l = 1, 2, .. , r) (!.1)
Accordingly, the time also enters functions (4.3):
X; =X; (q1 , q 2 , • • • , q8 , t) (i = 1, 2, ... , n) (1.2)
Another term appears in formulas (4.4) and (4.5):
• ox; ""' ax; •
xi=Tt+ L..J oqz q, (1.3)
l
From (1.3), we obtain the relation

ax; ax; (1.4\
-oqz. = aql '
coinciding with (4.6). Formulas (4. 7) also remain unchanged:

a~; =O (1.5)
aq~~.
Since the quantities ox 1 /Dq~~. contain not only q~~., l>ut also t, ex-
pression (4.8) becomes somewhat more involved:
d
Tt
ox; a ( ox; )
oqn =7ft oq11 + "'V a ( ox; ) •
L..J ---aq;- aqk q,
I
i)2x; "'V iJ2xt •
= atoq 11 + L..J l
oq 1 aq 11 fJL (Ui)
Differentiation of expression (1.3) with respect to q11 yields

iJ2x·
ox·
oq~ =
a ( 8x· )
oqk
""'
Tt + Ll l
a ( ox· ) •
oqk aq; qz = uq11 ~~ + ""'
Ll
l
o 2 x·
oq11 a~ 1 qz
Comparing this expression with (1.6), we arrive at the relation
(I. 7)
coinciding with (4. 9).

"298 APPENDICES
Let us multiply the equations

d aL aL
-d - - - = R 1 +Fi
• (i=1, 2, ... ,n)
t a·x; ax;
tsee (4.1)1 by fJx;IDqh and summate them over i:

""0 ( ~ aL ) ax; _"' aL ~
~ dt a;,. aqk ~...J ax; aqk
= "' R a:r,
~ 1 aqk
+ ""0
~
F~ ax;
1aq11
(I.S)
1 '& 1 '
Since relations (4.9), (4.6), and (4.7), used in Sec. 4 in the transfor-
mation of the left-hand side of this equation, remained unchanged
fsee (1. 7), (1.4) and (1.5)], the result will also remain unchanged.
Consequently, the left-hand side of (I.8) can be written as
d aL aL
Tt-a· - aqk
qk
Jsee the paragraph preceding formula (4.15)].
Before starting to consider the right-hand side of formula (1.8),
let us discuss the following matter. With stationary constraints,
x 1 = x 1 (q 11 ), and the increment of the coordinate x 1 during the time
dt will be
2j -ax·
dX·= ' dqk (1.9)
' aq~<
k
where dqk are the increments of the generalized coordinates during

the time dt. With non-stationary constraints, x 1 = X; (qk, t), and
the increment of x 1 during the time dt is
ax; "' a:r;
dx 1= Tt dt + L.J aqk dqk (1.10)
k
where dq 11 are again the increments of the coordinates qlt during the
time dt. -
If the constraints were to suddenly stop changing, fJx 1/ot would
vanish, and formula (1.10) would coincide with (I.9). Consequently,
the second term in (1.10) is the imaginary displacement of the system
that would be obtained with "frozen" constraints. It is called the
virtual (or possible) displacement and is designated by 1• Hence, ox
the true displacement dx 1 can be written as the sum of the virtual
displacement 8x 1 and an addend equal to (fJx 1/ot) dt:
(1.11)
With stationary constraints, the virtual displacement coincides with

the true one.
APPENDICES 299
Now let us consider the first term on the right-hand side of for-
mula (I.S). Multiplying it by dqk, we obtain
( ~ R 1 :;,:) dqk = ~ R 1 ( ;;~ dqk) = ~ R 16.r1 (1.'12)

i i i
where fix; is the virtual increment of the coordinate x 1 appearing
when only one generalized coordinate qk changes.
The work of the reactions R 1 with f1xed non-stationary constraints
(as with stationary constraints) is zero. Therefore, YR/JX; = 0,
i
and since dqk =I= 0, it follows from (I.12) that
(1.13)
We must note that the true work of the reactions of non-stationary

constraints according to (1.11) and (1.13) is
R 1 dx 1 = Li R/,x 1 + L.J R 1 af- dt = ~ R 1 af- dt

"'Q "'Q "'Q 0X • "'Q 0X ·
L.J
i i i i
and, generally speaking, is non-zero.
Hence, according to (I.13), the first term on the right-hand side
of formula (I.8) is zero-the reactions of the constraints have again
vanished from the equations. As regards the second term, by defi-
nition it is the generalized force Qi. [see formula (4.12)]. Consequently,
for non-stationary constraints too, we have arrived at Lagrange's
equations
II. Euler's Theorem for Homogeneous Functions
A function of any number of variables is called a homogeneous

function of these variables of the degree m if upon multiplying all
the varinblcs by the nrbitrary quantity a, tlw function is multiplied
by am, i.e.
f (ax1 , ax 2, ••• , axn) =
o.mf (x 1, x 2 , . . . , Xn)
Differentiation of this identity with respect to a yields
11
"'Q iJf ___ m-1j {..
~ a (ax;) X;- ma :.It, :rn)
i=1
Assuming that a = 1, we obtain
n
2j :~i Xt = mf (x 1, x~, ... , X 11 )
t=1
300 APPENDICES
We have proved Euler's theorem, which states that the sum of

the products of partial derivatives of a homogeneous function and the
corresponding variables equals the product of the function itself and the
degree of its homogeneity.
Ill. Some Information from the Calculus

of Variations
1. Functional. If to each number x belongiJJg to a certain class

there corresponds another number y, it is general knowledge that
we have to do with the ftmction y = y (x).
If to each function y (x) belonging to a certain class of functions
there corresponds a certaiu nnmber cD, the functional c:I) [y (x)l is
said to be set.
For purposes of clarity, we shall some-
'! times speak of a curve instead of a func-

tion.
Hence, a function establishes the corres-
pondence:
number ~number
whereas a functional establishes the cor-
respondence:
0 X
function (or curve) ~number
Consequently, when dealing with a func-
Fig. 111.1.
tional, the role of the argument is
played by a function (or a curve).
Let us explain what has been said above by means of the following
example. Assume that we are given two fixed points 1 and 2 in the
plane x, y (Fig. III.1). The distance l 12 between the points measured
along the curve joining them is a functional. To find an analytical
expression relating the quantity l 12 to the function y = y (x) de-
scribing \he curve, we shall take into account that the element dl
of the curve is related to dx and dy by the expression dl 2 = dx 2 +
dy 2 •
Writing dy as y' (x) dx, we obtain dl = dx 2 V +
[y' (x)F dx 2 =
= V +
1 [y' (x)J2 dx. Finally, integrating and designating l 12 by
 [y (x)l, we arrive at the expression
2
<D (y (x)] = JV 1 +
1
[y' (x)] 2 dx (III.1)
Taking different curves, i.e. different functions y (x), we shall ob-

tain different numbers <D.
APPENDICES 301
In the same way as there are functions not of one, but of several
variables, there are functionals depending on several functions:
Y1 (x), Y2 (x), · .. , Ys (x) 1 •
If a functional satisfies the conditions
(!J [cy (x)] = c<IJ [y (x)l (c = constant) (III.2)
(D [y 1 (x) +y 2 (x)l = cD [y 1 (x)l + <I) [y 2 (x)l
it is called linear. We shall denote a linear functional by the sym-
bol <D 11 n, or Fun, etc.
Functionals depending on several functions may be linear with
respect to some of them and non-linear with respect to the others.
In this case, for example, the symbol
<Dun v2 [yl (x), Y2 (x), •.. ]
will stand for a functional that is linear relative to the function y 2 (x).
It is the task of the calculus of variations to work out methods for
finding the extremal (i.e. maximum, minimum, or stationary) val-
ues of functionals. This task is in many aspects similar to the task
of fmding the extrema of conventional functions.
2. Variation of a Functional. Let us select an arbitrary function
y (x) from a class of functions being considered (we have used the
tilde sign to distinguish the s.elected function from the remaining
functions of the given class). Now let us select another function y (x)
from the same class. The difference of these two functions is called
the variation of the function y
(x). The variation is designated by
the symbol 8y (x) or simply 8y. The variation of a function is thus
determined by the following expression:
8y = y (x) - y (x) (III.3)
The variation of a function is similar to the increment ~x (or dx)
of the argument of an ordinary function: ~x = x - x.
The variation 8y of the function y(x) is evidently a function of x.
Differentiating this function with respect to x, we find in accordance
with (III.3) that
(f>y)' = y' (x) - (x) !/
The right-hand side of this expression is the variation of the function
y' (x). Consequently, we arrive at the relation
(8y)' = 8y' (III.4)
(the derivative of a variation is the Yariation of the derivative).
1 There are also functionals depending on the functions of ~cveral variables

r 1 , x 2 , • • • , xn. We shall not need such functional,, however. and shall not
r.onsider them.
302 APPENDICES
Let us determine what quantity in the calculus of variations cor-

responds to the differential of a conventional function. Recall that
if a function is continuous\ its increment t:.y equal to y (x) - (x) y
can be written as the sum of a term that is linear with respect to t:.x
and an infinitesimal of an order higher than the first one relative
to !:lx:
t:.y = y' (x) t:.x e t:.x+
where e is a quantity that vanishes together with !:lx (in other words,
lim e = 0). The first term in this expression is called the differen-
.o.x-e
tial of the function. Hence, a differential of a function is defined to
be that part of the increment of the function that is linear relative
to !:lx.
For the function y · x,
the differential coincides with the incre-
ment. Consequently, t:.x = dx, so that the expression for the differ-
ential of a function can be written as
dy = y' (x) !:lx or dy = y' (x) dx (111.5)
If a small change in a functional corresponds to a small change
in a function (to a small 6y), the functional is called continuous.
A quantity similar to the differential of a conventional function can
be introduced for continuous functionals.
The increment of a functional MD = <D [y (x) + 6y] - <D [y (x)]
is a quantity depending on two functions: y (x) and 6y. Therefore,
t:.<D is also a functional. This functional, generally speaking, will
be non-linear. If !:l<D [y (x)l can be written as the sum of the func-
tional Fun. 6 y [y (x), 6y] that is linear relative to 6y and an infi-
nitesimal of an order higher than the first one with respect to I 6y I max
(the maximum value of the magnitude of the function 6y), the main
part of the increment of a functional linear relative to 6y is called
the variation of the functional. Hence,
MD [y (x)l = 6<l> [y (x)] + e I 6y I max (II 1.6)
where
6<D [y (x)l = Fun, 6y [y (x), 6y] (Ill. 7)
is the variation of the functional, and e is a quantity vanishing to-
gether with I 6y I max·
The variation of a functional is the analogue of the differential
of a function determined by expression (111.5).
1 More strictly, the function ought to be not only continuous, but also differ-
entiable. As a rule, however, the functions considered in physics are also differ-
entiable if they are continuous. Although, generally speaking, continuous func-
tions are known in mathematics that are not differentiable.
APPENDICES 303'
The definition (III. 7) can easily be generalized for functionals

depending on several functions:
6CD [y 1 (x), y 2 (x), . . . , Ys (x)l
= Fun ~Yr. 6y2, ... , ~Ys IYt (x), Y2 (x), • · ., Ys (x), 6y1, 6y~, · · ·
... , 6ysl (I II.S).
As an illustration, let us find the increment of the functional
b
<D (y (x)] =) [y (x)J2dx (111.9}

a
This increment is
b b
!leD= <D [y (x) + 6y]- <D [y (x)] = ) [y (x) + 6y] 2 dx- ) [y (x)F dx
a a
b b b
=) [2y(x)6y+(6y)2] dx=) 2y(:r) oydx+ J(6y)2d.r

a a a
b
The integral ) (6y) 2 dx does not exceed the quantity I 6y lkax (b -
a
- a) = {I 6y Imax (b - a)} I 6y Imax• where I 6y Imax is the maxi-
mum value of the magnitude of the function 6y within the interval
a :::;;; x :::;;; b. The expression in braces vanishes together with I 8y lm"x·
b
Consequently, ) (6y) 2 dx can be written as e I 6y lmax• where e _. 0>

a
when I 6y lmax _. 0. Hence,
b
~Cl>=) 2y(.T)6ydx-f-e I B!!lmnx
"
[compare with (111.6)]. The iirst term in tllis expression is the func-
tional depending on the functions y (x) and oy:
b
Fun, ov [y (x), 8y] = J2y (:1.:) 6y dx (II I. '10)

a
It is easy to verify that this functional is linear with respect to 6y 1

[see the conditions (III.2)l. Hence, expression (111.10) giYes the Yaria-
tion of the functional (III.9).
1 The given functional is also linear with respect to y (x), but this is of no·
significance-only the linearity with respect to oy is important.
304 APPENDICES
3. Necessary Condition for the Extremum of a Functional. Let us

again begin with similar concepts from calculus. The function I (x1 ,
x2, ... , Xn) is said to have an extremum at the point Xz, xl,
... , Xn if the increment of this function
M =I (xt, Xz, ••• , xn) -I (xl, ~•... , Xn)
has the same sign for all the points (x 1 , x 2 , • • • , Xn) belonging to
the vicinity of the point (xu ~ •... , Xn)· There is a maximum at
a given point when M ::( 0, and a minimum when M ~ 0.
It is proved in calculus that a necessary condition for the existence
of an extremum at a point is the equality to zero of the differential
of the function at this point:
n
at
dl= ~ -dxk=O (III.11)
axk
k=i
The functional <D [y (x)l is similarly said to reach an extremum at
y = y
(x) if the increment of the functional
~<D = <D [y (x)] - <D [y (x)]
has the same sign for all the curves y (x) sufficiently close to the curve
y (x). When ~<1) ::( 0, a maximum of the functional is observed,
and when ~<D ~ 0, a minimum.
Let us find the condition necessary for the functional <]) [y (x)]
to reach an extremum at y = y
(x). For clarity, we shall consider the
case of a maximum. If cD [y (x)] reaches a maximum at y = (x), y
this signifies that
<D [y (x) + By] - <D (y (x)] ~ 0 (III.12)
for all 8y = By (x) for which I By lmax is sufficiently small. By (III.7)
and (111.8)
~<D = <D (x)] [y + oy] - cD !Y (x)]
= Flln, /Jy [y (x), By] + B I By lmax (III.13)
Let us separate from all Llw possible variations By those that can
be represented in the form By = a6y 0 , where By 0 is a fixed sufficient-
ly small variation, and a is a varying algebraic quantity. Introduc-
ing this variation into (III.13) and taking into account that owing
to linearity F [.y(x), aBy 0 ] = aF [y (x), 8y 0 ], we can write
~<D = aF [y (x), Byol + W IBYo lmax

In the last expression, F [y (x), 8y is simply a number.
0] If this
number is non-zero, at suff1ciently small a's the sign of ~cD will
APPENDICES 305
be determined by that of the expression aF [y

(x), 8y 0 ) (the term
ea I 8y 0 lmax diminishes much more rapidly than the first term 1),
and this expression will change its sign together with a. (which may
be either positive or negative). Hence, for the condition (Ill.12)
to be satisfied, the quantity F [y (x), c5y 0 ) must equal zero. The con-
dition (111.12) must be satisfied for all sufficiently small variations
<'>y without any exception. We have found, however, that if 6<1> =I= 0
for even part of the variations having the form a.c5y 0 , the above con-
dition is not observed. We can therefore state the following: for the
functional [y (x)l to achieve a maximum at y = (x), its variation y
(if it exists) must vanish at y = (x): y
c5<1> = Flln,au [y(x), c5y) = 0 (III.14)
[The identity sign underlines the circumstance that the condition
(III.14) must be satisfied for all the c5y's.]
It can be seen that by repeating our reasoning for the minimum
of a functional, we shall arrive at the same conclusion. Consequently,
formula (II1.14) expresses the condition necessary not only for a max-
imum, but also for a minimum, i.e. for an extremum in general.
This formula is an analogue of formula (III.11).
4. A Simple Problem in the Calculus of Variations. Let us find the
extremum of a functional having the form
x,
 [y (x)] = ~ f [x, y (x), y' (x)] dx (III.15)
The boundary points of the allowable curves are assumed to be fixed-

for all the allowable curves
y (x 1 ) = y1 and y (x 2 ) = y 2 (III.16)
The increment of the functional is
%1 x2
L\<1> = Jf[x, y + c5y, y' + c5y'] dx- Jf [x, y, y'] dx
We expand the integrand of the f1rst integral in powers of the small

quantities c5y and c5y'. The result is
...
+ e (<'>y, <')y')} dx- j f(x, y, y') dx
1 The conditione- 0 for I 6y lmax -+- 0 in the given case acquires the form
lim e = 0.
a ..... o
20-018
306 APPENDICES
where e (fJy, 6y') combines the terms of an order higher than the
first one relative to the quantities {)y and fJy'. This expression is
simplified as follows:
x. ,:.:,
.l<D=) { :~ 6y+ :~, 6y'}dx+) e(l3y, {)y')dx (111.17)
Let us integrate the second term in the first integral by parts. For
this purpose, we write it as
!ICJ !lCs
J= J ::, l3y' dx= J aU:

:x:,
(6y)' dx
{remember that fJy' = (oy)', see formula (III.4)]. Designating ofloy'

by u, and (6y)' dx by dv and taking advantage of the formula u dv = J
= uv - Jv du, we obtain
%1 :Ca ::e.
J= j ;:, (oy)'dx= :;, oy j- j ;z ( :;, )oydx
:tt X1 Zt
Since the boundary points of the allowed curves are fixed, the varia-
tion 6y at these points must vanish: 6y {x1 ) = 0, and 6y {x 2 ) = 0.
Therefore, the first term on the right-hand side vanishes, and
:x:.
J=-
Jr _j_(_!L)&ydx
dx fJy'
Let us introduce the found value of J into formula (III.17), factoring

out oy: '
:.:. :a:.
A<Il= J{;;-:X 8;, }6ydx+ Je(&y, &y') dx
The-main part of .l<t> is formed by the first integral that is a func-

tional linear with respect to &y. By definition, this integral is the
variation of the functional <IJ. Hence, the variation of the functional
(III.15) is
x.
6<D= J{-
of - - -
dx
fJy
at } &ydx
d
fJy'
(III.18)
and the condition (111.14) for an extremum will be written as

:x:.
J{-at -dz-
d at } 6ydx==.O
-
fJy 8y'
APPENDICES 307
The identity obtained must be observed for any sufficiently small

functions {)y = {)y (x). This is possible only provided that the inte-
grand in braces vanishes:
.!}__~ _!)_=0 (III.19)
oy dx oy'
This equation is known as Euler's equation. It is the condition for
an extremum of functionals of the form given by (II !.15). The curves
y = y (x, C1 , C 2 ), which are solutions of this equation, are called
extrema Is (C 1 and C 2 are integration constants).
We must note that the addition to the integrand in (III.15) of
the total derivative with respect to x of any function 'ljJ (y, x) does
not change the conditions of the extremum (III.19). Indeed, this
term after integration yields the quantity
whose variation is zero [according to the condition (III.16), the

curves do not vary at their ends]. Hence, the addition of d'4J/dx changes
only the value of the extremum of a functional, but does not affect
the form of the function y (x) at which this extremum is reached.
Let us use formula (III.19) to find the extremum of the functional
(III.1). In this case, f (x, y, y') = V 1 +
[y' (x)] 2 • Hence,
oj -0 of y' (x)
oy - , oy' = - Vt+IY' <x>J2
and Eq. (III.19) becomes

d 11·" ( 1 y'2 ) Yn=O
Tx Vt+y'2 = Vt+y'2- (i+y'7>3'2
A function for which y" = 0 and y' = a will be the solution of

this differential equation. Consequently, the function itself is a
linear one, y = ax+ b, whose coefficients must be chosen so as to
satisfy the conditions (III.16). Therefore, functional (111.1) reaches
an extremum (in this case, evidently, a minimum) if we presume
that y (x) is a straight line joining the points 1 and 2 (see Fig. III.1).
5. Extremum of Functionals Depending on Several Functions.
Consider a functional of the form
x,
... , y,] = Jf [x, Yt• Y2• · · ·, Ys,
(111.20)
308 APPENDICES
where Y11. = Y11. (x) (k = 1, 2, ... , s) are curves fixed at boundary

points, i.e. satisfying the boundary eonditions
Y11. (xl) = Yh1• Y11. (x2) = Y11.2 (k = 1; 2, ... , s)
(III.21)
In accordance with what has been said in paragraph 4, the varia-
tion of the functional (III.20) is determined by the expression
(111.22)
where 6y11. are sufficiently small functions of x vanishing at the

boundary points
(111.23)
It is necessary to find such a set of functions Y11. (x) satisfying the
conditions (111.21) at which the functional (111.20) reaches an extre-
mum. A necessary condition for an extremum is the vanishing of
the variation of the functional (III.22).
· Integrating by parts each of the s addends of the form
x,
J ::k ~Yk dx
Xt .
in formula (111.22), we can bring it to the form
Performing such a replacement in (III.22) and equating to zero the

expression obtained for 6<1>, we arrive at the necessary condition for
an extremum:
x, '
Jr ""
LJ
(..!L
auk _ _E_ _!4.-) 8yk dx = 0
dx ayk
(II1.24)
Xt 1!.=1
This identity must be obeyed for any sufficiently small functione

{)y,_- = 6y,_ (x) selected independently of one another. This is possibls
only provided that all s expressions in parentheses inside the sum
~re zero. We have thus arrived at a system of Euler's equations:
_fl_ _ _
dd 81 =0 (k==1, 2, ••. , s) (111.25)
ay,.. x 8u~~.,
The set of functions y,.. == Yn (x) satisfying these equations and the
boundary conditions (III.21) when substituted into the functional
(III.20) will give its extremum. ·
APPENDICES 309
IV. Conics
Conics (conic sections) are defined to be the lines of intersection
of a circular cone with a plane. Depending on the orientation of the
plane relative to the axis of the cone, these lines are either an ellipse
(circle), a hyperbola, or a parabola.
/
Fig. IV.1. Fig. IV.2.
An ellipse (Fig. IV.1) is defined to be the locus of points, the sum

of whose distances from two fixed points F 1 and F 2 called the foci is
a constant quantity:
· r 1 -L r 2 = 2a (IV.1)
The canonical equation of an ellipse is
.z2 y2
-+-=1
a2 b2 (IV.2)
where a and b are the major and minor semiaxes of the ellipse. The
quantity
c
e=- a
(IV.3)
where c is half the distance between the foci, is called the eccentricity
of the ellipse. When e = 0, an ellipse degenerates into a circle.
The quantities a, b and c are related by the expression
(IV.4)
A hyperbola (Fig. IV.2) is defined to be the locus of points, the
magnitude of the difference of whose distances from two fixed points
F 1 and F 2 called the foci is a constant quantity:
I r1 -rd = 2a (IV.5)
A hyperbola has two symmetric branches. The canonical equation of
a hyperbola is as follows:
(VI.6)
310 APPENDICES
where a and b are the real and imaginary semiaxes of the hyperbola.
The quantities a and b are related to c (half the distance between the
foci) by the expression
(IV. 7)
The eccentricity of a hyperbola is determined by the same for·
mula (IV.3) as that of an ellipse. It can be seen that for an ellipse
e < 1, and for a hyperbola e > L
p'
A parabola (Fig. IV.3) is defined to be the locus of points whose

distance r from a fixed point F (the focus) equals the distance d
from the fixed straight line D called the directrix of the parabola:
r = d. The canonical equation of a parabola is
yz = 2px (IV.8)
where p is the parameter of the parabola, equal to the distance from
the focus to the directrix (the x-axis is directed along the axis of
symmetry of the parabola, and the origin of coordinates coincides
with the apex of the parabola). As we shall see below, the eccen-
tricity of a parabola should be taken equal to unity.
Any conic can be determined as the locus of points for which the
ratio between the distance r (Fig. IV.4) to the point F (called the
focus) and the distance d to the straight lineD (called the directrix)
is a constant quantity e (called the eccentricity of the curve):
r
rr=e (IV.9)
The relevant conic is obtained depending on the value of the ec-

centricity e (Table TV.1).
APPENDICES 311
Table IV.f
Value of e
I Kind of curve
<1 Ellipse
1 Parabola
>1 Hyperbola
An ellipse and a hyperbola each has two foci and two directrices.
The condition (IV. 9) is observed for each of the foci and the directrix
corresponding to it. Figure IV.S shows the foci and directrices of an
ellipse (a) and a hyperbola (b) (compare with Fig. IV.3 for a parabola).
p' 2(a/e) p'
(a)
Fig. IV.5.
The distance p' from a focus to a directrix is called the parameter

of the relevant curve. The quantity
p = p'e (IV.W)
is known as the focal parameter. It is a simple matter to see that it
equals half the chord passing through a focus and parallel to the
directrix (Fig. IV.4). For a parabola, p = p'.
The distance between the directrices of an ellipse (hyperbola)
equals 2 (a/e), where a is the major (real) sem iaxis of the curve.
Let us write the equation of a conic in polar coordinates, placing
the origin of coordinates at one of the foci of the curve (Fig. IV.4).
In accordance with (IV.9)
r r
-=
d p' +rcos cp
=e
312 APPENDICES
whence
r= P {IV .11)
1-e cos q>
[p = p'e; see (IV.10)].
Equation (IV.11) describes an ellipse (with the origin of coordi-
nates at the point F 1 ; Fig. IV.Sa), the right-hand branch of a hyper-
bola (with the origin of coordinates at the point F 2 ; Fig. IV.Sb),
and a parabola. We must note that the focus F 2 is the inner one for
the right-hand branch of the hyperbola.
If we place the origin of coordinates at the right-hand focus 6f an

ellipse (point F 2 in Fig. IV.5a), a glance at Fig. IV.6 shows that the
equation of the ellipse is
r
p'-rcos q>
e
or
r=1+:cosq> (IV.12)
(p = p' e). The same equation describes the left-hand branch of a
hyperbola (see Fig. IV.Sb) provided that the origin of coordinates
is placed at the point F 1 (at the inner focus with respect to this branch),
and also a parabola that is the mirror (relative to D) image of the
parabola depicted in Fig. IV.3. The focus F of such a parabola is to
the left of the directrix D.
Let us fi.nd the equation of one of the branches of a hyperbola
(say, the left one) provided that the origin of coordinates is at the
external focus relative to this branch (at the point F 2 for the left
APPENDICES 313
branch; Fig. IV.5b). In this case (Fig. IV.7),

d' = r cos (n - cp) - 2 !.
e
- p' =- r cos cp - 2 !. - p'
e
According to the definition (IV.5) of a hyperbola, r- r' = 2a~
whence
r' = r - 2a
Introducing the values d' and r' we have found into relation (IV.9).
we arrive at the formula
r' r-2a
y= - - - - - - - - = e
-rcos cp-2 .!!:..-p'
e
After simple transformations, we obtain the required equation
r= -p (IV .13)
1+e cos cp
(p = p'e). It must be borne in mind that for the left branch cp >
> n/2 (see Fig. IV. 7), i.e. cos cp < 0. In addition, for all points
I e cos cp I> 1 so that the values of r obtained by formula (IV.13)
will be positive.
We invite our reader to convince himself that the similar equation
for the right branch (the origin of coordinates is at the point F 1 ) is
r= 1 _-:-:C,sq> (IV.14)
in this case, cp < n/2 so that cos cp > 0. In addition, e cos cp > 1
so that the values of r are positive.
V. Linear Differential Equations with Constant

Coefficients ·
A linear differential equation of the n-th order with constant coef-

ficients is defined to be an equation of the kind
y<n> +an-tY<n-t) + ·.·+ aty' +
aoy = f (x) (V .1)
s.e. an equation linear in the unknown function y (x) and its deriYa-
tives (the a;'s arc constant quantities that may also be zero).
If the right-hand side of an equation identically equals zero [f (x) =
= 0], the linear equation is called homogeneous, othenvise it is
non-homogeneous. A homogeneous equation has the form
y<n> +an_ 1 y<n-t) + ... +a1 y' a0y = 0+ (V.2)
The general solution of a differential equation is defined to be th&
multitude of solutions including all the particular solutions with
no exceptions. The general solution of an n-th order differential
equation contains n arbitrary constants (integration constants) ..
314 APPENDICES
i.e. has the form

(V.3)
By giving the constants C1 , C 2 , • • • , Cn definite values, we obtain
a particular solution. The latter contains no arbitrary constants.
It is proved in the theory of linear differential equations that if
y 1 , y 2 , • • • , Yn are linearly independent 1 solutions of the homoge-
neous equation (V.2), the general solution of this equation can be
written as
n
y(x, C" C 2 , ••• , Cn)= 2J C yt(x)
i=t
1 (V.4)
where C1 , C 2 , • • • , Cn are arbitrary constants.

Let y (x) be one of the particular solutions of the non-homogeneous
equation (V .1) and y (x) be the general solution of the same equation.
If we introduce the notation u (x) = y (x) - y
(x), the general solu-
tion can be written as follows:
y (x, C11 C,., . . . , Cn) = u (x, C1 , C2 , • • • , Cn) +
(x) (V.5) +y
Let us substitute this function into Eq. (V.1) and group separately
the terms of the kind u<k> and the terms of the kind y<k>:
.u<nl +an -lu<n-1) + ... +
alu' aou+
+ [y<n) + an_1y<n-l) + . . . + a y' + a y] = f (x)
1 0
y
The function (x) is a particular solution of the equation. Conse-
'<luently' the expression in brackets equals the right-hand side of
the equation. It follows that the function u (x, C1 , C 2 , • • • , Cn)
.satisfies the condition
u<n> +
a(n-llu<n-l) + ... +
a 1u' a 0u = 0 +
Hence; u is the general solution of the homogeneous equation (V.2)
-corresponding to the non-homogeneous equation (V.1), i.e. having
the same coefficients ak as Eq. (V.i).
The result we have obtained can be formulated as follows: the
general solution of a linear non-homogeneous equation equals the sum
of the general solution of the corresponding homogeneous equation and
a particular solution of the non-homogeneous equation:
y (gen., non-hom.) = y (gen., hom.) + y (part., non-hom.) (V.6)
1 The set of functions fit f 2 , • • • ,fn is called linearly independent if an expres-
sion of the kind
a;l ft +a2 fz+ • · ·+a;n fn== 0
.is observed only provided that all the ai's vanish.
APPENDICES 315
Linear homogeneous differential equations with constant coefficients

are solved with the aid of the substitution
y (x) = e""' (V.7)
where A. is a constant. Differentiating this function m times (m =
= 1, 2, ... , n), we obtain
(V.8)
Introducing the values of the function (V.7) and its derivatives
(V.8) into Eq. (V.2) and cancelling the non-zero factot· e""', we arrive
at what is called a characteristic equation:
A.n + an_1A.n-l + ... + a 1A. + a 0 = 0 (V.9)
The roots of this equation are the values of A. at which the function
(V.7) satisfies Eq. (V.2).
If all n roots of the characteristic equation are different (multiple,
i.e. coinciding roots are absent), n particular solutions of the kind
e'J..t"' will be linearly independent. Consequently, in the absence of
multiple roots, the general solution of Eq. (V.2) is as follows:
(V.10)
(Cu C 2 , • • • , Cn are arbitrary constants).

It can be shown that when the characteristic equation (V.9) has
multiple roots, piJ. linear independent particular solutions corre-
sponding to the root A.~1 of multiplicity p 11 must be taken in the form
eAle Ax 2'J..x
iJ. , xe iJ. , x e iJ. , • • •,
so that the contribution to the general solution corresponding to

them equals the sum
p
'\~ C .ll-t A.,X
""x e ...
..:....J
k=l
Consequently, if the root A.1 will be of the multiplicity p 1 , the root

A. 2 of the multiplicity p 2, ••• , Am of the multiplicity Pm (here
p1 + p 2 + ... +
Pm = n), the general solution can be written as
(V.11)
Let the coefficients ak in Eq. (V.1) be real, and the function f (x)
be complex. Writing it in the form
f (x) = /1 (x) + if2 (x)
316 APPENDICES
we get the equation

y<n) + an-1Y(n- 1) + • · · + a1Y + aoy =
1
= /1 (x) + if 2 (x) (V.12)

We shall seek the solution of this equation in the form
y (x) = y1 (x) +
iy 2 (x)
Substitution into (V.12) yields
(yln>+ an-1Y(n-1) + · ·· · + a1Yi + aoYl)
+ i (y~nl + an-JY~n-u + · · · + aly~ + aoy2) =
= /1
(x) +
i/2 (x) (V.13)
In complex numbers equal to one another, the real and imaginary
parts are equal to one another independently. Hence, Eq. (V.13)
breaks up into two independent equations of the form of (V .1 ). The
right-hand side of one of them contains the function f1 (x), and the
function y1 (x) is its solution. The right-hand side of the other equa-
tion contains the function / 2 (x), and the function y 2 (x) is its so-
lution. This property of Eq. (V.13) is due to its linear nature. It
allows us to use the following procedure that sometimes considerably
facilitates calculations. Assume that the right-hand side of the
equation (V.1) we are solving is real. We add an arbitrary imaginary
function to it. After now finding the complex solution of the equa-
tion obtained, we take its real part. It will be a solution of the initial
differential equation.
The following statement is obvious: if a linear homogeneous equa-
tion (V.2) with real coefficients has the complex solution y (x) =
= y 1 (x) + iy 2 (x), each of the functions y1 (x) and y 2 (x) separately
is a solution of this equation.
VI. Vectors
1. Basic Definitions. Vectors are quantities defined by a numerical
value (magnitude) and a direction and, in addition, are added geo-
metrically (i.e. according to the triangle or parallelogram method).
On a later page, we shall give a more general definition allowing us
to extend the concept of a vector to an n-dimensional space~
The scalar product1 of two vectors a and b is defined as the scalar
quantity
ah = ab cos (a, b) (VI.1)
A scalar product is commutative (ab = ba) and distributive
{a (b 1 + + ... )
b2 +
= ab 1 ab 2 + ... },
but is not associative
{a (be) -=1= (ab) c}.
1 Another way of writing a scalar product in addition to the one we use is
a·b, which explains the name not product sometimes used for it.
APPENDICES 317
A vector product is defined as the vector t
[ab] = ab sin (a, b)· n (VI.2)

where n is the unit vector of a normal to the plane containing the
vectors a and b, the sequence a, b, n forming a right-handed system.
A vector product is not commutative (lab] =I= (hal}, is distribu-
ti\·e (Ia, (b 1 + b 2 + ... )] = [ab 1 ] -1- [ab 2 ] + ... ), and is not
associative (!lab], c] =I= (a, [bell).
Let us consider a scalar triple ...,."1--------;::::ry
product of three vectors: a [be]. ~-:..-)-----~/--- /
Applying formulas (VI.1) and I I I
(VI.2), we obtain n / / /
a [bel= a {be sin (b, c)} cos (a, n) (a,n) c ----f/-;:;_.,.1
A glance at Fig. Vl.1 shows that ..,.;;..(-lrb.~c-)_ _ _ _v-
the expression we have obtained
€quals the volume of the paral~ Fig. Vl.1.
lelepiped constructed on the vec-
tors being multiplied 2 • Indeed, be sin (b, c) gives the area of
the base of the parallelepiped, and a cos (a, n) gives its altitude.
We can also take a face whose sides form the vectors c and a or
the vectors a and b as the base of the parallelepiped. In this case,
the volume will be determined by the scalar triple products b [cal
and c [ab]. Since the volume in all three cases is the same, we can
write
a [bel = b [cal = c [ab] (VI.3)
A scalar triple product thus allows a cyclic transposition of the
factors, i.e. the substitution for each factor of the one following or
preceding it. The v~ctor c is presumed to be followed by a, and a
is assumed to be preceded by c, which can be illustrated by the fol-
lowing diagram:
a-+b
'/ (VI.4)
c
We must note that in all three expressions of formula (VI.3) the
vectors have the same sequence as in the diagram (VI.4). If we take
the vectors in a sequence that is the opposite of what is shown in
the diagram (VI.4), the scalar triple product will change its sign.
A vector triple product is defined to be the vector [a [bell. We
can prove (this will be done somewhat later) that
[a, [bell = b (ac) - c (ab) (VI.5)
1 Another way of writing a vector product is a X b, which explains the
name cross product sometimes used for it.
2 The angle (a, n) is assumed to be acute. If this angle is obtuse, the scalar
triple product equals the volume of the parallelepiped taken with a minus sign.
318 APPENDICES
In accordance with the definitions (VI.2) and (VI.1), the square of

the vector product of the vectors a and b can be transformed as
follows:
[ab] 2 = a2 b2 sin 2 (a, b) = a2 b2 - a2 b2 cos 2 (a, b) = a 2b2 - (ab)2
We have arrived at the formula
(VI.6)
2. Formulas of Vector Algebra Expressed in Terms of Projections
of Vectors onto the Coordinate Axes. All the above definitions and
formulas do not depend on the choice of the coordinate system used
for our consideration. If we set up a coordinate system (we shall
consider only rectangular, i.e. Cartesian systems), each vector can
be set by three numbers-its projections onto the coordinate axes1 •
Consequently, the vector a is equivalent to the three numbers ax,
ay, az, the vector b-to the numbers bx, by, bz, etc.
Knowing the projections of a vector onto the coordinate axes, we
can find the vector itself. Denoting the unit vectors of the coordinate
axes by the symbols ex, ey, e:z:, 2 we can represent the vector as
(VI. 7)
To obtain the possibility of writing the formulas in a compact
form using the sum sign L,
we shall use the symbol x 1 instead of the
coordinate x, x 2 instead of y, and x 3 instead of z in the following.
Similarly, we shall introduce the symbols e1 , e 2 , e 3 for the unit
vectors of the axes. The correspondence between the previous and
the new symbols is shown below:
X--+Xt}
y--+xz (VI.8)
Z--+Xg
(VI.9)
In the new notation, formula (VI. 7) can be written as
a=~ eltak (VI.10)

k
1 We are treating free vectors, for which the point of application of the
vector and the straight line along which it is directed are not fixed.
2 These unit vectors are also designated by the symbols i, j, k. The notation
we have adopted, however, as will be seen from the following treatment, has an
undoubted advantage.
APPENDICES 319
(unless otherwise indicated, we shall always assume that the subscript

over which summation is performed-the dummy index-runs
through the values 1, 2, 3).
The triplet of vectors e 1 , e 2 , e 3 forms the basis of a coordinate sys-
tem. The setting of these vectors completely determines the system.
Since e 1 , e 2 and e 3 aro mutually perpendi-
cular, and their magnitudes equal unity, eJ
it can be seen from formula (VI.1) that
e;e 11 = 6il, (VI.11)
where 8; 11 is the Kronecker symbol deter-
mined as follows:
1 when i = k
{
6111 = 0 when i =I= k (VI. 12) e1
It must be noted that 6t 11 = 6 11 ; (it will Fig. Vl.2.
be shown in Appendix X that the set of
the quantities 6; 11 forms a symmetric second-rank tensor).
Examination of formula (VI.2) reveals that (Fig. Vl.2)
[ele2] = ea (VI.13)
[e2esl = el
[e 3e 1 J = e 2
Each of the relations (VI.13) can be obtained from the preceding
(or following) one by a cyclic transposition of the subscripts ac-
cording to the diagram
(VI.14)
Let us introduce the symboP

CiJll (VI.15}
standing for a set of 27 numbers that are determined by the fol-
lowing rules:
(1) if the values of at least two subscripts coincide, we have
Etlll = 0 (for instance, e. 111 = E 2 ~t 2 = Eiss = e 222 = 0);
(2) if all the subscripts are different and form a cyclic transposi-
tion of the sequence 1, 2, 3, we have E;~tz = 1 ( e123 = e 231 = e 312 =
= 1);
(3) if all the subscripts are different and form a cyclic transposi-
tion of the sequence 3, 2, 1, we have Eil!l = -1 ( E 321 = c 213 =
= E132 = -1).
1 It is sometimes called the Kronecker skew-symmetric symbol. It will be
shown in Appendix X that the set of quantities ttl!l forms an absolutely anti-
symmetric third-rank tensor.
320 APPENDICES
Hence, of the 27 values of e1kz, 21 are zero, 3 are +1, and 3 are
-1.
We must note that any cyclic transposition of the numbers 1, 2, 3
~an be obtained from 1, 2, 3 by an even number of transpositions
of two subscripts, and any cyclic transposition of the numbers 3, 2, 1
can he obtained from 1, 2, 3 by an odd nnmber of transpositions of
two subscripts. Indeed, by changing, for instance, the places of 1
and 2 in the sequence 1, 2, 3 (which yields 2, 1, 3), i.e. a cyclic
transposition of the numbers 3, 2, 1), and then changing the places
-of 1 and 3, we obtain the transposition 2, 3, 1.
Consequently, the values of the symbol Etkl can be determined as
follows: (1) they are zero if the values of at least two subscripts coin-
cide, (2) they are +1 or -1 depending on whether the sequence i, k,
l can be obtained from the sequence 1, 2, 3 by an even or an odd
number of transpositions.
We can also use the following rule to determine the sign of eikl·
\Vhen a larger number is ahead of a smaller one in a transposition,
we shall call this a disorder. For example, in the transposition 2, 1, 3
there is one disorder-2 is ahead of 1, and in the transposition 3, 2,
1 there are three disorders-3 is ahead of 1, 3 is ahead of 2, and 2 is
ahead of 1. Let us assign the value of +1 to e1kz if the number of
disorders in the transposition i, k, l is even, and -1 if it is odd. It
is easy to see that all three rules for determining the sign of e 111.1
which we have considered give the same result.
Let us prove the following very useful relation between the sym-
bols e and &:
~ Eik!Emnl = 8im811n- 8tn811m (VI.16)
l
We shall expand the sum on the left-hand side:

e·il!.l Emnl + eii!.2Emn2 + Etl!.aEmn3 (VI.17)
.and determine the values of the subscripts i, k, m, and n at which
this sum is non-zero. It is evident that for at least one term to differ
irom zero; the conditions
i =I= k and m =I= n (VI.18)
must be satisfied simultaneously. In addition, it is essential that
i = m, k = n or i = n, k = m (VJ.19)
Indeed, if the observance of the condition (VI.18) will not be attend-
ed by observance of the condition (VI.19), all three numbers 1, 2
and 3 will be present among the values of the first two subscripts
Q{ both factors in each of the terms in (VI.17). Therefore, the dummy
index l in each of the terms will coincide with the value of one of
the subscripts i, k, m ~nd n, so that all the terms will vanish.
APPENDICES 321
Let us combine the conditions (VI.18) and (VI.19) into one ex-
pressed by the formulas
i = m =1= k = n (VI.20)
i = n =1= k = m (VI.21)
lu the case corresponding to rela lions (Vl.20), the sum (VI.17)
becomes
emn1enm1 + emn2Emn2 + emnaEmna
It is obvious that only one term (for which m, n, and l are different)
will he non-zero, and it equals +L
In the case corresponding to relations (VI.21), the sum (VI.t7)
becomes
In this sum too, only one term is non-zero, and it equals the product
of +1 and -1, i.e. -1 (upon the transposition of two subscripts
enml changes its sign).
Now let us turn to the right-hand side of formula (VI.16), i.e. to
the expression
(VI.22)
If i = k (or m = n), this expression becomes 011711 fl1<n - 81w8hm (or
Oin{)l<n- O;nflR. 11 ). Both these expressions are zero. ]t thus follo,vs
that for expression (VI.22) to be other than zero. the following con-
ditions must be observed simultaneously:
i =I= k and m =I= n (VI.23)
[compare with (VI.18)l. In addition, one of the following two condi-
tions must be observed:
i = m, k = n (VI.24)
i = n, k = m (VI.25)
'Vhen the condition (Vl.24) is sntisi'Jl'd, the lhst lern1 ot' expn•s-
sion (VI.22) is +L Since i = m and m =I= n [see (VI.23)}, we have
i =1= n, and Lhe second term of expression (VI.22) vanishes. Conse-
quently, when the condition (VI.24) is observed, expression (VI.22)
is +1. The combination of the conditions (VI.23) and (VI.24) is
equivalent to the condition (VI.20) in which, as we have established,
the left-hand side of formula (VI.16) also becomes equal to +1.
When the condition (VI.25) is observed, which in combination
with (VI.23) is equivalent to the condition (VI.21), expression
(VI.22) becomes equal to -1.
· We have thus proved relation (VI.16).
21-018
322 APPENDICES
Using the symbol Btkl• we can write the combination of relations

(VI.13) as a single expression:
[e 1e"] = 2JI e 1" 1ez (VI.26)
Indeed, when i = 1 and k = 2, only the addend e123 e 3 equal to e 3

will be non-zero, when i = 2 and k = 3, only the addend e231 e 1 = e 1
will he non-zero, and, finally, when i = 3 and k = 1, the addend
e 312e 2 = e 2 will he non-zero.
The expression (VI.26) gives even more than a combination of
three relations (VI.13). It contains nine relations. It follows from
it that the vector product of any unit vector by itself is zero-when
i = k, all the addends, on the right-hand side of formula (VI.26)
vanish. In addition, (VI.26) contains expressions obtained from
(VI.13) by transposition of the multipliers. For instance, when
i = 2 and k = 1, the addend e 213e 3 = -e 3 on the right in (VI.26)
is non-zero, etc.
Let us form the scalar product of the vectors a= 2J e 1a 1 and b =
= 2Jekbk:
Using the property of distributivity, we can write

ab = 2J
i. 1\
e 1e"a 1b" = 2J
i, 1\
l3,katbk
[see formula (Vl.11)l. In accordance with the defmition of l3 1k, in

the last sum only addends having identical values of the subscripts
t and k are non-zero. Therefore
(VI.27)
or, going over to conventional symbols, we have

ab = axbx + a11 b'll + azbz (VI.28)
For the given vectors a anrl b, their projections onto the coordi-
nate axes depend on the choice of the coordinate system, but the
product ab itself does not depend on this choice. We thus conclude
that the expression o.xbx +
ayby +
azbz is an invariant, i.e. a quan-
tity identical in all coordinate systems.
Assume that we have been given a set of three numbers u, v, w
about which we know that in combination with the projections of a
vector a they yield a scalar, i.e. an invariant:
ua:x + va11 + wa 1 = inv
APPENDICES 323
On the basis of what has been said above, we can assert that u, v,
and w are the components1 of a vector.
Let us form the vector product of the vectors a = ~ e 1a1 and b =
= ~e11b11:
[ab] = [(lj e 1a 1), (~ ekbk)l
i h.
Owing to distributivity, we can write

[ab] = 2j [ete~tl a1b11
i, II
Let us replace [eiek] in accordance with formula (VI.26):
[ab] = 2j a 1b" 2j e 1'k!ez = 2j eo. za 1b'kel
i, k l i. k, l
The vector product can thus be written as
[ab] = ~ e 1" 1a 1bket (VI.29)
i, k, l
Of the 27 addends of this sum, only six are non-zero. Writing them
out, we obtain
[ab] = a 1b 2e 3 4- a 2b3e 1 4- a 3b1e 2 -- a 3b 2e 1 -- a 2b1e 3 -- a 1b3e 2
Finally, combining terms with identical unit vectors, we arrive at
the expression
[ab] = e 1 (a 2 b3 - - a 3 b2 ) + e 2 (a 3 b1 --a 1 b3 ) +"a (a 1 b2 --a 2 b1 ) (VI.30)
that can be written in the form of a determinant (.::ee Appendix VIII)
Ct e2 e3
[ab] = at a2 aa (VI.31)
bt b2 ba
or in conventional notation
ex ey Cz
[ab] = ax ay az (VI.32)
bx by bz
We must note that according to (Vl.29), the l-th component of a
vector product is determined by the formula
[abj, = ~ e 1kza 1b11. = ~ Bz 1 ~tatbk
i. It i. h
(we have performed a cyclic transposition of the subscripts on e,
which, as is known, does not change the numerical value of this
1 For brevity's sake, we shall use this term for the projections of a vector
onto the coordinate axes.
21*
~24. APPENDICES
symbol). To deal with the more customary sequence of letter sub-

scripts, we shall write an expression for the i-th component of a
:vector product:
(VL33)
"Let u~ prove formula (VI.5) using relation (VI.16). For this pur-
}Jose, we shall write the vector product in accordance with formu-
:la (VI.29):
d =[a, [be]]= 2J
It, l, i
ewak [bcjpr, 1e 1
Now, we shall substitute for [bclpr,z its expression obtained from

~iormula(VI.33):
·Let us perform a cyclic transposition of the subscripts on the sym-

:bols e so that their common subscript l will be in the last plaee. In
'addition, let us group the factors so that summation over l is per-
Harmed first of all:
,d = 2j
i, It, m, n
etakbmCn ~ Btk!Bmnl
l
= 2J
i. k, m, n
etakbmCn ({)im{)kn- {)tn{)km)
lwe have employed relation (VI.16)]. Further transformations yield

.d = LJ
.i, k
~lak 2Jm l>,mbm 2Jn {)knCn- i,2Jk e,alt 2Jn B,ncn 2Jm Bltmbm
= 2J
i, k
e 1a11.b 1ck- );
i, k
e1akciblt
Q.E.D.
3. True Vectors and Pseudovectors. Two kinds of vectors are dis-
tinguished: polar (or true) and axial vectors, also known as pseudo-
'Vectors1;·Upon inversion of the coordinate axes, i.e. when the
:directions of the coordinate axes arc reversed (Fig. VI.3), the com-
~ponents of a true vector change their sign. This signifies that such a
vector upon inversion remains unchanged. The components of a pseu-
dovector upon inversion do not change their sign. This signifies that
:.a pseudovector upon inversion reverses its diiection (i.e. changes
its sign).
Inspection of Fig. VI.3 shows that upon inversion of the coordi-
: nate axes, a right-hand system of coordinates transforms into a left-
i hand one. The distinction between a true vector and a pseudovector
.ean therefore be defined as follows: a true (polar) vector does not
1 "Pseudo" i~ a prefix meaning false or sham.
APPENDICES 325
change upon a transition from a right-hand system of coordinates to

a left-hand one, whereas a pseudovector reverses its direction in
such a transition. It will be shown in Appendix X that a pseudovector
is an antisymmetric second-rank tensor.
If both vectors a and b are true, the components of the ver.tor pro-
duct (VI.30) upon inversion do not change their sign (a 1 and b1a
separately change their sign, but their
product remains unchanged). Consequent- z
ly, the vector product of true vectors is
a pseudovector.
Scalars must also be divided into two fl(x)
kinds: true scalars and pseudoscalars. /
/
True scalars do not change in a transition /
from a right-hand coordinate system to a (;)----- 1 o y

left-hand one (or upon inversion of the 1
coordinate axes). They include mass, I
electric charge, and temperature. Pseudo- x I
scalars change their sign in a transi- 1
tion from a right-hand coordinate system f(z)
to a left-hand one. They include the scalar
expressions obtained as a result of mathe- Fig. VI.3.
matical operations on vectors. For ex-
ample, the scalar product [see (VI.27) l of a true vector and a pseudo-
vector changes its sign upon inversion and, consequently, is not
a true scalar, but a pseudoscalar.
If the vectors a, b, care true, expressio:m. (VI.3) will be a pseudosca-
lar-it changes its sign upon inversion. Hence, the scalar triple pro-
duct of true vectors is a pseudoscalar. ·
4. Transformations of Vector Components. Let us find formulas:
for the transformation of the components of a vector in a transition
from one coordinate system to another. Let us take two systems of:
Cartesian coordinates K and J{', setting them by their unit vectors
e 1 , e 2 , e:~ and e~, c~, e;. The arbitrary vector a can be written as
a = ~ c"a 1, where a" arc the projections of a onto the axes of the·
systrm K, or as a = ~ e/,al,, where ah are the projccti0ns of a onto
the axes of the systPill K'. Hunce,
.::..; ' ' = .t-.J
'\' ckak '\' ekak (VI.34)?
k k
We multiply (Vl.34) by the unit vector ei:
(VI.35},
By (VI:11), we have eiel, = {)il,. Consequently, of the three addends.

on the left-hand side, only the one with k = i that is equal to 8 11 ai =-
= ai ''ill be non-zero.
S26 APPENDICES
The scalar product eie 11 equals the cosine of the angle between the
axis xi of the system K' and the axis x 11 of the system K. Designat-
ing this cosine by the symbol CtiJu we can write
a, 11 = eieh = cos (xi, xh) (i, k = 1, 2, 3) (VI.36)
Using this notation, we can write relation (VI.35) as
ai = 2JII. et 111 a11 (i = 1, 2, 3) (VI.37)
Formula (VI.37) allows us to calculate the projections of the vector

a onto the axes of the system K' according to the known projections
of a onto the axes of the system K. To obtain formulas for the reverse
transformation (from K' to K), let us multiply (Vl.34) by the unit
vector e;. Repeating the reasoning that led us to formula (VI.37),
we obtain
(VI.38)
Formulas (VI.37) and (VI.38) differ only in that in one case summa-
1lion is performed over the second subscript of a 1 ~~., and in the other
case over the first one.
The nine quantities ail!. are not independent. Let us form the sum
~ et 1mall.m• Taking (VI.36) into account, we obtain
m
~ CttmCtkm = ~ (eiem) (e.ltem)
m m
The quantity e!em can be considered as the projection of the vector
ei onto the axis Xm of the system K; similarly, eltem is the projection
of the vector cit onto the axis Xm· Hence, the sum on the right can be
written as
[see formula (VI.27)). Consequently,

~ CtunCtll.m = 011, (VI.39)
m
It can be proved in a similar way (we invite our reader to do this)
that
(VI.40)
Transformations (VI.37) and (VI.38) can be adopted as a defini-

tion of a vector: a vector is defined to be a set of quantittes a 1 , a 2 , a 3 ,
which upon a transition from one coordinate system to another
are transformed by formulas (VI.37) and (VI.38) where Cttk are
quantities determined by fnrmula (VI.36).
APPENDICES 327
The last definition of a vector can readily be extended to a space

with any number of dimensions. Assume that we have an n-dimen-
sional space. Let us take a system of coordinates whose axes are
mutually perpendicular in this space. This signifies that the unit
vectors e 1 , e 2 , • • • , en of the axes satisfy the condition (VI.11).
A vector in n-dimensional spaco (an n-vector) is thus defined to be a
set of n quantities a 11 a 2 , • • • , an that upon a transition from one
coordinate system to another are transformed by formulas (VI.37)
and (VI.38). In summation, the dummy index runs through n val-
ues instead of 3. The number of equations in (VI.37) and (VI.38)
will also be n instead of 3.
A scalar product of vectors can also he readily generalized for an
n-dimensional space. Similar to (VI .27), we shall call the expres-
sion
(VI.41)
which is an invariant, the scalar product of two vectors having the

components a 1 , a 2 , • • • , an and b1 , b2 , • • • , bn. Vectors whose sca-
lar product is zero are said to be mutually orthogonal (or mutually
perpendicular).
The concept of a vector product cannot be extended to spaces with
other than three dimensions.
Inversion of the coordinate axes (see Fig. Vl.3) can be treated as
a transformation from the system K to the system K' whose coeffi-
cients have the value!'!
-1 when i = k
{
ath.= 0 when i=fok
or
(VI.42)
By formula (VI.37), the components of a vector upon inversion
are transformed 'aCCording to the law
, ,.,
a,=- LJ o,ltak = -a,
h
i.e. reverse their sign (this was already mentioned on an earlier

page).
Let us find the law of transformation of the components of a vec-
tor product upon inversion of the coordinate axes. We write ex-
pression (VI.33) for the system of coordinates K' (obtained as a re-
sult of inversion of the axes of the system K):
(VI.43)
328 APPENDICES
We have taken advantage of the circumstance that the quantities

e 1 ~tz are determined in the same way for all coordinate systems,
owing to which upon any transformations of the coordinates, we
have
(VI.44)
Let us express alt and h; in formula (VI.43) in terms of the un-
primed components of the relevant vectors using relation (VI.37)
[ab]i = L
k, l
eikl ~ ( - 6km) am~ ( - Bzp) bp =
m p
~
k, I, m, p
eiklOkmOlpambp
Summation over the subscripts m and p yields

[ab]i = Lj eiklakbl
k, I
In accordance with (VI.33), the last expression is [ab]i, We have

thus established that
[abli = [ab] 1
i.e. that the components of a vector product do not change in inver-
sion. A vector product of true vectors is therefore a pseudovector.
Let us write a scalar triple product of three vectors. By formu...:
las (VI.27) and (VI.33), we have
a [be]=~ a 1 [bc] 1= ~ a 1 L eihlbkcz = ~ e1kza 1bkcz (VI.45)
i i k.l i,k,l
Let us see how this quantity behaves in inversion. In the system K',
we have
(a [be])' = lj e 111 zaibitcl
i, k, l
We have obtained a result we already know: the scalar triple product

of trlJ.e vectors upon inversion change1; its sign, i.e. is a p,;;euuu.s<:alar.
5. Increment of a Vector in Hotation. Let us fmd the increment
which the vector a obtains in rotation through an infinitely small
angle dfP. We introduce two coordinate systems K and K' which we
choose so that their axes z and z' coincide with the vector dfP
(Fig. Vl.4). Assume that the system K' turns together with the
vector a through the angle dcp relative to the system K. Relative to
the system K', the vector a remains constant, while relative to the
system K it receives the increment da.
We shall first assume that the tail of the vee tor a is on the z-ax is
(Fig. VI.4). If a is initially in the plane yz, the increment dais collin-
APPENDICES 329
ear to the x-axis. The magnitude of this increment, as can be seen

from the figure, is a sin a dq;. It follows from the above that the
increment of the vector a can be written as
da = (dq;, a] (V1.46)
We shall prove that the formula we have found also holds with an
arbitrary arrangement of the vector a relative to the coordinate sys-
tems K and [{'. Let us introduce the
:' K'
unit vectors ex, ey, ez of the system K and
the unit vectors e~, e~, e; of the system K
K'. The vector a can therefore be set by
the expression
a = exax + eyay + ezaz (VI./~7)
or the . expression
(VI.48)
where ax, ay, az are the projections of the e'y y'
vector a onto the axes of the system K, and
a~, a~, a~ are the projections of a onto the
axes of the system K'.
When the vector turns together with the y
system K' through the angle dq;, it receives
an increment relative to K, that can be
written as the increment of expression
(VI.47):
Fig. VI.4.
da = Cx dax + e11 day + ez daz
or as an increment of expression (VI.48):
da = a~ de~ + a;, de~ + a; de; (VI.49)
where de~, de 1;, de~ are the increments of the uuit vectors of the sys-
tem f(' observed iu the system K (remember that the projections
a~, a;,, a; remain unchanged upon rotation).
With the din·ction of thl' z'-axis we have chosPn, the increment of
tile unit vector e; vm1isltos (de; = 0). Figure Vl.5 shows the incre-
ments de~ and de~ contained by the unit vectors e; and e~ when the
coordinate system IC turns through the angle dq;. Examination of
the figure shows that the direction of de~ coincides with the direc-
tion of the unit vector e~. The magnitude of de~, on the other hand,.
equals dq; (the magnitude, i.e. the length of any unit vector is uni-
ty). Consequently, the increment of the unit vector e~ obsen·ed in
the system K can be written as
de~ =c e; dq;
'330 APPENDICES
Similar reasoning results in the formula

de~= -e~ d<p
the minus sign is due to the fact that the vectors de~ and e~ are
directed oppositely).
Introducing the values of the unit vector
increments we have found into formula
(VI.49), we obtain
da = (a~e;- a~~) dcp
We shall show that the expression we
have found is equivalent to the vector
product [d<p, a]. To do this, we shall express
de~ the product in terms of the projections of
the vectors being multiplied onto the axes
Fig. VI.5. of the frame K', taking into account that
dff! is directed along the z' -axis. In accord-
.ance with formula (VI.32), we have
e~ e~ e;
[d<p,a]= 0 0 d
a~ a~ a;
We have thus arrived at formula (VI.46).
VII. Matrices
Definition. In Appendix VI, we obtained formulas for the trans-
formation of vector components in a transition from the coordinate
system K to the system K':
'V
a 1kah (i = 1, 2, 3)
I
a•= LJ (VII.1)
II.
a,=~ a~~.,ah. (i=1, 2, 3) (VII.2)

k
{see formulas (VI.37) and (VI.38)].

The transition coefficients can be written in the form of a square
table
au cx12 cx13
A= CXzt CXzz CXz3 (VII.3)

cx31 CX32 cx33
known as a transformation matrix. The quantities a 1k are the matrix

-elements. The first subscript indicates the number of the row in
which the given element is, and the second subscript-the number
~f the column.
APPENDICES 331
Let us come to an agreement about our notation. We shall denote

the elements of a matrix by lower-case letters with two subscripts,
and a matrix by the corresponding capital letter (for instance, the
element a 1k of the matrix A 1 ). We shall denote the components of a
vector by lower-case italic letters with one subscript, and the vector
itself by the same lower-case Roman (upright) letter in bold-face
type (for instance, a 1 is the component of a vector, and a is the vec-
tor itself).
The operation (VII.1) of transforming the components of a vector
can be written symbolically as the multiplication of a vector and a
matrix:
a'= Aa (VII.4)
The coefficients of the inverse transformation (VI 1.2) form the
matrix
au a21 aat
A-t= rx12 a22 aaz (VII.5)
called an inverse matrix. Denoting the elements of an inverse matrix

by the symbol aik, we can write
(VII.6)
The matrix obtained from A by interchanging the rows with the
columns is called a transposed matrix and is designated by A. If we
denote the elements of a transposed matrix by the symbol 1k, we a
can write
(VII.7)
A glance at formulas (VII.6) and (VII. 7) shows that the matrix
(VII.5) of the inverse transformation coincides with the transposed
matrix (VII.3) of the direct transformation:
A- 1 =A (VII.8)
Relation (VII.8) does not hold for any matrices 2• Matrices satis-
fying the condition (VI 1.8) are called orthogonal.
The inverse transformation (VII.2) is written symbolically as
a = A-la' (VII.9)
Without altering the formal mathematical aspect of the matter, re-
lations (VII.4) and (VII.9) [in other words, relations (VII.i) and
1 This is how the capital Greek letter "alpha" is written.
2 In general, not any matrix has an inverse one. A matrix for which no in-
verse one exists is known as a singular or degenerate one. But even if a matrix
is non-singular. it:> inverse and transposed matrices may not coincide.
332 APPENDICES
(VII.2)l can be treated not as operations of transition from one coor-

dinate system to another, but as operations transforming one vector
into another, both vectors being considered in the same coordinate
system. Having in view such an interpretation, we can write the
formulas for transformation as follows:
b =An (VII.10)
a= A-1b (VII. H)
Hence, the matrix A can be considered as a linear operator that by
acting on the vector a transforms it into the vector b.
Let us write the transformations (VII.10) and (VII.11) in the
explicit form, and for greater generality we shall consider that the
vectors a and b are determined not in a three-dimensional space, but
in a space with n dimensions. By analogy with (VII.1) and (VII.2),
we obtain
n
b,= 2J a,kak
k=t
(i=1, 2, ... , n) (VII.12)
n
a 1= 2J aikbk
k=t
(i = 1, 2, ... , n) (VII.13)
where a{k are the elements of the inverse transformation matrix

(the matrix A - 1). For an orthogonal matrix, aik = all.;·
The matrices A and A -l will now have n rows and n columns, for
example
CGu a12 ... atn
A=
a21 a22 ... a2n
(VIJ.14)
..
ant an2 ... ann
The matrix (VII.14) is square-the number of rows in it equals
the number of columns. In addition to square matrices, rectangular
ones are ·also considered, in which the number of rows m does not
equal the number of columns n:
A= A(min) = (VII.15)
The first subscript on the matrix symbol indicates the number of

rows, and the second-the number of columns. We shall drop these
subscripts when this does not cause ambiguity.
Hence, in the general case, a matrix is defined to be a set of m·n
elements arranged in the form of a rectangular array. The elements
APPENDICES 333
of a matrix may be functions, numbers, or other quantities on which

algebraic operations can be performed. A matrix with m rows and n
{:Olumns is said to be an m by n matrix (written m X n). An m X 1
matrix, i.e. a matrix with one column, is called a column matrix,
and a 1 X n matrix, i.e. a matrix with one row, is called a row
matrix.
Two matrices A and B are said to be equal (A = B) if the relennt
·elements of these matrices equal one another (a.; 11 = ~i 11 ).
The matrices A and B are considered to differ only in their sign
{A = -B) if the relevant elements of these matrices are related by
the expression a.;k = -~ill·
The square matrix (VII.14) (i.e. ann X n matrix) is a particular
-case of the matrix (VII.15). A matrix transforming a vector in an
.n-dimensional space into another vector in the same space will evi-
·dently be square.
If the elements of a square matrix satisfy the condition
(VII.16)
the matrix is symmetric. A symmetric matrix obviously coincides
with its transposed one:
(VIJ.17)
A square matrix whose elements satisfy the condition
(VII.18)
is said to be asymmetric or skew-symmetric. An asymmetric matrix
.differs from its transposed one only in the sign:
(VII.19)
A square matrix in which only the elements a. 11t with identical
values of the subscripts i and /.,; are non-zero is called diagonal. Such
a matrix has the form
0
0
(VII .20)
0 0 ... An
The elements of this matrix can be written as follows:
A.;k = A.koik (VII.21)
where 0; 11 is the Kronecker symbol [see (VI.12)].
If we change the coordinate system (i.e. the basis e 1 , e 2 , • • • , en),
the components of the vectors a and b [see formula (VII.10)l become
different. The elements of the matrix-operator will also change.
334 APPENDICES
Sometimes (particularly, when the matrix A is symmetric), the ba-

sis can be chosen so that the matrix A becomes diagonal.
In a transition from one coordinate system to another, the ele-
ments of a matrix change, but the sum of the diagonal elements called
the trace of the matrix (its symbol is Tr A) remains unchanged.
Therefore, the trace of a matrix is identical in all coordinate systems,
i.e. is an invariant:
TrA=2Jcxii=inv (VI1.22)
i
The determinant of the matrix [see (VIII.3)] also remains unchanged:

det II aik II = inv (VII.23)
Let us define a unit (or identity) matrix I which when multiplied
by a vector according to the rule (VII.10) yields the same vector:
a= /a
It is a simple matter to see that the elements of a unit matrix must
equal 6 1h [the substitution into (VII.12) of CXtk = 6tk leads to the
relation bt = ad. Hence,
1 0 0
I= 116tk II= 0 1 ... 0 (VII.24)
0 0 ... 1
We must note that this matrix is diagonal.
Matrix Algebra. Matrices are algebraic objects lending themselves
to addition, subtraction, and multiplication (the operation of matrix
division does not exist). ,
The sum of two matrices A and B is the matrix r = A + B whose
elements are determined by the formula
(VII.25)
The difference of two matrices is the matrix r = A - B with the
elements
(VII.26)
It is evident that only matrices having the same number of rows
and the same number of columns can be added and subtracted.
The product of the matrix A and the scalar 11 is defined to be the
matrix B = 11A with the elements
(VII.27)
Let us now consider the multiplication of matrices. Assume that
the action of the matrix A on the vector a transforms it into the
vector b, and the action of the matrix B on the vector b transforms
APPENDICES 335-
it into the vector c. It is natural to define the product of the matrices

A and B as the matrix r which when acting on the vector a trans-
forms it into the vector c. Hence,
b = Aa, i.e. bm =~ c:tmltak
k
.
C= Bb = BAa, I.e. - '\., A b - '\1 A '\1
Ct-,....JI-'tm m-..:..:ptmLJC:tmhak-..::....Jalt~t-'tmC:tm~
- '\1 "\, A
m m k k m
On the other hand,
A comparison of the two formulas for c and c1 leads to the rule of

matrix multiplication:
r= BA signifies that Vt1t = ~ ~ 1 mc:tm1t (VII.28}
m
According to this rule, to obtain an element of the matrix r at the

intersection of the i-th row and the k-th column, we must multiply
each element of the i-th row of the matrix B by the corresponding-
element of the k-th column of the matrix A and summate all the pro-
ducts. Thi~ can be explained by the following diagram:
alk···
.a2k ...
(VII.29)
~:: • .~i.2 .... .. ~~n ":::::::.::
(
.............. "•Clnk " '
· i-th row, k-th column'j
We must note that matrix multiplication, in general, is not com-

mutative, i.e.
BA =I= AB
Matrices for which the condition
BA = AB (VI 1.30)
is satisfied are called commutative.
It is easy to show that the product of matrices is associative:
(fB) A = r (BA) (VII.31)
This signifies that by first multiplying B and r, and then A and
(fB), we obtain the same result as we would by first multiplying the
matrices A and B and then multiplying the matrix (BA) and r. In-
336 APPENDICES
deed, according to the rule of matrix multiplication

{(fB) A},k = ~ (fB)tmGtmk =
m
2:m (~l "r'tl~lm) Gtmk
= ~ "r'tl (2J ~zmGtmk) = 2J y 1z (BA)z~< = {f (BA)}tk
l m l
(in the course of transformations we have changed the sequence of

summation over the subscripts m and l). Consequently, the property
{VI I.31) has been proved.
Non-square (rectangular) matrices can also be multiplied by each
-other. Examination of the diagram (VII.29) shows that such matrices
<:an be multiplied only when the number of columns of the matrix B
(the second matrix 1) coincides with the number of rows of the matrix
A (the first matrix). The product matrix will have the same number
-of rows as the second matrix (B) does, and the same number of co-
lumns as the first matrix (A) does. We shall explain this by the fol-
lowing example:
Gtu Gttz Gtta
Gtzt Gtzz Gtza
(LJ ~tkGtkt) (~ ~tkakz) (LJ ~tlta,.a) II

= II (~ ~2kaki) (~ ~zkctkz) (2J ~2kaka)
If the second matrix is square, i.e. is an n X n matrix, and the
first matrix has only one column with n elements, the product ma-
trix also consists of one column with n elements: 1
~11 ~12 ~in Gtt (2.; ~tlta,.)

~21 ~22 ~2n Gtz ( '\1
.LJ ~zkak
)
(VII.32)
..... - .....
~ni ~n2 ... ~nn
an (~ ~nkak)
When a column malrix is rn11lliplicd hy n row mnlrix, the rcs11lt
is simply a number (or a function if the malrix elements are func-
tions):
II ~t ~2 •• • ~n II (VII.33)
1 In the matrix product BA, the matrix A at the right should be considered
as the first multiplier. It is multiplied by a vector fnst of all, and only then the
second matrix B acts on the result.
APPENDICES 337
Particularly, if we take the transposed matrix II ex II as the matrix

II B 11. Eq. (VII.33) becomes
tXt
CXz - "\1
II tXt CX2 • • • CXn II -
,..,2
LJ I.NJt
k
Consequently, for the column matrix A(n.l)• the following relation

holds:
(VII.34)
If we take the compone11ts of the vectot a as the elements of a col-

umn matrix, and the matrix operator A as a sqnare matrix, rela-
tion (VII.32) will become
cxu CXt2 CXtn at bt
a21 (X22 CX2n a2 b2

. .. (VI1.35)
CXnt CXn2 ... tXnn an bn
where b1 = ~ cx 1kak [compare with (VII.12)]. It is not at all difficult

k
to see that relation (VII.35) is equivalent to relation (VII.10). Con-
sequently, a vector can be represented as a column matrix.
Consider the product of the unit matrix I and an arbitrary matrix
A. By rule (VII.28)
(Al)lk = '\."1, CX.tm8mk
m
(8mk are the elements of the matrix J). In this sum, only one addend
in which m = k will be non-zero. Therefore, (Al) 1k = cx 0~,. Simi-
larly,
It follows hom tho above that m11ltiplication by a unit matrix

(with any sequence of the factors) does not change the matrix A:
!A= AI= A (VII.36)
Relation (VII.36) signifies that a unit matrix commutes with any
matrix A.
It is obvious that by fi.rst applying to a vector tho transformation
A, and then its inverse transformation A -1, we must return to the
initial vector:
a=A- 1Aa (VII.37)
338 APPENDICES
This shows that the product of a direct and an inverse matrices must
equal a unit matrix: A-lA = Jl. The product of a direct and an in-
verse matrices is obviously commutative, hence
A - 1A = AA -l = I (VII.38)
Having written the elements of the prodnct of the matrices A and
A- 1 by formula (Vll.28), we can iiml the relation between the ele-
ments of the direct and the inverse matrices:
(VII.39)
For an orthogonal matrix, i.e. one satisfying the condition (VII.8),

we have eti~t = aki [see (VII.6)]. Making this substitution in (VII.39),
we obtain
(VII.40)
m
(VII.41)
The elements of an orthogonal matrix thus satisfy relations (VII.39)

and (VII.40) [compare with formulas (VI.39) and (VI.40)l.
VIII. Determinants
Aisume that we have the square matrix
(VIII.1)
Let us form the following expression from the elements of this

matrix:
(VIII.2)
where i, k, ... , m is a permutation of the numbers 1, 2, .. , n, and
e1k ... m is a quantity equal to +1 if the number of disorders (in-
versions)2 in the permutation i, k, ... , m is even and to -1 if it is
odd [compare with (VI.15)]. The number of permutations of n num-
bers taken n at a time is known to be n! We can therefore compile nl
ilifferent expressions of the form given by (VII.2).
1 This relation clarifies the symbol A -1 used for an inverse matrix (/ is

"unity").
2 Recall that a disorder in a permutation is the fact that a larger num her
precedes a smaller one (see p. 320).
APPENDICES 339_:
The sum of all the expressions having the form of (VIII.2) is de:..
noted by the symbols
ant Un2 • • • ann

'\1
L..! eil1. .. ma 11 a 21l ••• anm (VIII.3)
(ik .. . m)
and is called the determinant corresponding to the matrix (VIII.1).
The sum (VIII.3) is taken over all the permutations of the numbers
i, k, ... , m. Consequently, it contains n! addends.
If we add to the definition of e 11, ••. m the condition that this quan-
tity vanishes when the values of at least two of the n subscripts
coincide, the determinant D (A) can be evaluated as the sum
n
i, k,
2J
.. , m=!
Bik ••. mauaall • • • Unm (VIII .4)
in which all the subscripts i, k, ... , m take on values from 1 to n.

A glance at (VIII.3) shows that a determinant can be written as
an array similar to (VIII.1) with the difference that single vertical
bars are used instead of double ones.
The number of rows (or columns) of a determinant is said to be its
order.
We must note that the determinant of the diagonal matrix (VII.20)
equals the product of its diagonal elements:
det II A."8o. II = A. 1A. 2 • • • A.n (VIII.5)
and the determinant of a unit matrix is unity:
D (I) = 1 (VIII.6)
Let us explain what has been said above using a third-order de-
terminant as an example:
au a12 a13
D3 = a21 a22 UzJ

a31 a32 a33
We can make 31 = 6 permutations from the second subscripts of the

elements of this determinant:
123 0 disorder +
231 2 disorders +
312 2 disorders +
321 3 disorders
213 1 disorder
132 1 disorder -
340 APPENDICES
By formula (VIII.4), the expression

Da = aua22aaa + aua23aal + auauaa2 - ataa22aat
- a12a21asa - aua2aas2
is a determinant of the third order.
If we delete the i-th row and the k-th column in an n-order deter-
minant, we obtain a determinant of the n - 1 order known as the
minor of the initial determinant corresponding to the element ail<·
This minor is customarily denoted by the symbol ~ilt· The quantity
A111. = (-1)i+1t~il!. (VIII.7)
is called the algebraic cofactor of the element a 111 •
Properties of Determinants. We shall list the basic properties of
determinants without proving them.
1. The value of a determinant does not change if corresponding
rows and columns are interchanged:
au at2 ... atn au a21 . .. ant
... a2n atz a22 ...
..... . . .an2. .
au a22
ant aM ... ann atn azn . .. ann

or, more compactly,
det II a,lt II = det II a~~., II (VIII.S)
Interchanging of the rows and columns is called transposition.
A determinant in which such a change has been made is said to be
transposed. We can thus say that a transposed determinant equals
the initial one. '
Property 1 shows that the determinant of a transposed matrix
equals the determinant of the initial matrix:
D (A)= D (A) (VIII.9)
Indeed, these determinants differ only in the rows and columns
having been interchanged, which does not change the value of the
determinant.
2. The sign of a determinant is changed if any two rows or any
two columns are interchanged.
3. If two columns or rows are identical, the determinant is zero
(this follows from the property 2).
4. A determinant is a linear form 1 of the elements of a row or a
1 The linear form of the variables x1 , x2 , • • • , xn is defined to be the linear

homogeneous function of these variables, i.e. the expression
I= a1 x 1 + a2x2 + ... + anxn
APPENDICES 341
column:
n
det lla 1AII = ~ A 111 a 1k (linear form of the elements of the i-th row)
k-=1
(VIII.10)
or
n
det lla 1kll = ~ A 1katk (linear form of the elements of the k-th column)
i=1
(VIII. H)
the algebraic cofactors (VIII. 7) of the relevant elements being the
coefficients A 111 of the linear forms (VIII.10) and (Vlll.11).
5. If the elements of one row (or column) are multiplied by the
algebraic cofactors of the elements of another row (or column) and
the products obtained are summated, the sum will be zero (this sum
is a determinant with two identical rows or columns; see the pro-
perty 3).
Properties 4 and 5 can be combined in the form of the relations
~ A,AamA = det !larA II· Btm }
~ A,Aalm
t
. det llatk!l· BAm
(VIII.12)
Two more properties (6 and 7) follow directly from the property 4:

6. If all the elements of a row (or column) contain a common
factor, it can be written before the determinant:
au · · · ~atA atn au · · · a1A . . . atn
a au . . . a~aik . ... aain =a~ au ... a,k . .. aln (VIII.13)
ant ...~ank ann ant . .. anh . .. ann

7. If the elements of a row (or column) are the sum of two (or
more) addends, the determinant eq1.1als the sum of determinants in
which the rclovnnt addC'nds arC' tho elcml'nts of tho givl'n row (or
column), for example,
a 11 aih + aik a 1n a 11 al.h ... a 1n
azt a2A + a2A azn azt a2k · · · azn
ann ant ... a~~~ . .. ann

au ... ah atn
a2f ... ah a2n
+ (VIII.14)
12 n l ... a~k ann
APPENDICES
8. The value of a determinant is not changed if to the elements of

any row (or column) we add the corresponding elements of another
.row (or column) after first multiplying them by the same constant
quantity.
This property follows from properties 7, 6, and 3.
9. Tho product of Lwo delerminanls of Lho sumo order, dot II a 1h II
and det II b1h II, is a determinant of the same order, del II c 11, II,
whose elements are expressed by the formulas
(VIII.15)
A comparison of this formula with formula (VII.28) shows that de-

terminants are multiplied in the same way as the corresponding ma-
trices. Consequently, the determinant of a product matrix coincides
with the product of the determinants of the multiplier matrices.
It can be seen from the property 9 that the determinant of an ortho-
gonal matrix is +1. Indeed, for an orthogonal matrix A = A - 1
(see (VII.8)], owing to which AA = I. According to what has been
said above,
D (A) D (A) = D (I) = 1
[see (VIII.6)]. But by (VIII.9), we have D (A) = D (A), so that we

can write
(D (A)] 2 = 1
whence
D (A)= ±1 (VIII.16)
Systems of Linear Non-Homogeneous Algebraic Equations. Con-
sider a system of n linear algebraic equations with n unknowns,
xl, x2, ... , Xn:
auxl + a12X2 + • • • + alnXn = bl
a21X1 + a22X2 + • · · + a2nXn = b2
This system can be written as a single expression:

n
2j a;kxk=bt (i=1, 21 ••• , n) (VIII.17)
h=1
The coefficients at the unknowns can be seen to form a square

matrix similar to the matrix (VIII.1 ). Assume that the determinant
of this matrix (we shall call it the determinant of the system) is
APPENDICES 343
non-zero:
au a12 ... atn
D= au a22 ... a2n
:FO (Vlii.18)
ant an2 ... ann
We multiply the lirst of Eqs. (Vlll.17) by 11 1 m-the co[netor of the
element a1 m, the second equation by A 2 m, . . . , the n-th equation
by Anm' and summate the expressions obtained. The result is
We change the sequence of summation on the left-hand side of the

equation:
n n n
~ Xk
k=l
LJ Atmaik = i=l
i=l
L Atmbt
n
By the second of formulas (VIII.12), ~ Atmaik = det II aik I! Bmk =
i=!
= DBm 11 , where D is the determinant of the system. Hence, the
obtained relation can be written as
n n
LJ xkDBmk = i=-1
k=1
LJ Atmbt
Summation on the left over k yields the product xmD· A comparison
of the sum on the right with formula (VIII. H) allows us to conclude
that this sum is the determinant obtained from the determinant
(VIII.'18) by substituting the free terms of the system (VIII.17)
for the elements of the m-th column. Denoting this determinant by
the symbol D<m>, we can write that
XmD = D<m>
whence
[m has been replaced with k for the subscript on x to be designated

by the same letter in the given formula and in formula (VIII.t7)].
'vVe have arrived at Cramer's rule, which states: if the determinant
of a system is non-zero, it has one definite solution, the value of the
unknown x~r, being equal to a fraction whose denominator is the deter-
minant D of the system, and whose numerator is the determinant D<k>
obtained from D by replacing the elements of the k-th column with the
344 APPENDICES
free terms of the system:

au atz ... bt ... atn
a21 azz · · .bz · · · a2n
(VIII.19)
au a12 · · · atlt · · · atn
au a22 • • • azlt. • • • azn
The solution of the system of equations (VIII.17) can be made

very clear by using the following representation. The unknowns
x1 , x 2 , • • • , Xn can be considered as the components of a vector x
in an n-dimensional space, and the free terms b1 , b2 , • • • , bn as the
components of a given vector b. Now the system (VIII.17) can be
written symbolically as the relation
Ax= b (VIII.20)
where A is a matrix compiled from the coefficients of Eqs. (VIII.17):
au a12 · · • atn
A= au a22 • • • azn
Indeed, expanding relation (VIII.20) by formula (VII.12), we obtain

n equations:
n
L]a,kxk=b 1 (i=1, 2, ... , n)
k=i
coincidi_ng with the system (VIII.17).

Consequently, the problem of finding the unknowns x 1 can be
formulated as follows: we are given the matrix A and the vector b
in 'an n-dimensional space; it is necessary to find a vector x such
that when multiplied by the matrix A transforms into the given
vector b.
We multiply Eq. (VIIL20) by the matrix A -l. On the left, we
obtain the required vector x [see formula (VII.37)l, and we arrive
at the relation
X= A- 1b (VIII.21)
Hence, to find the solutions of the system of equations (VIII.17),
we must proceed as follows: find the matrix that is the inverse of the
system's matrix and substitute the elements of this matrix into the
APPENDICES 345
formulas
xk = ~aittb 1 (VIII .22}
i
[see formulas (VII.11) and (VII.13)].
Comparing formulas (VII1.19) and (VIII.22), we arrive at the
conclusion that the elements of the inverse matrix are determined
by the expressions
aii = ~k (VIII.23}
We shall indicate another form of writing the relations we have
considered. Representing the vectors x and b as matriees with n
rows and only one column, we can write the system of equations
(VIII.17) as
au a12 ••• ain xi bi
a21 a22 ••• a2n x2 b2
- (VIII.24)
ani an2 • • • ann Xn bn
or, more briefly,
A(n,n) ·Xcn,t) = B(n,l) (VIII.25)
[compare with (VII.35)].
Systems of Linear Homogeneous Equations. The system of equa-
tions (VII 1.17) in which all the free terms b1 are zero is said to be
homogeneous. Hence, a homogeneous system has the form
+ a x + ... + a nXn = 0 )
a 11 x 1 12 2 1
. . . .. . +. • .. • +. a2nXn
a2lxl + a22x2 =0
. . . . (VII 1.26)
an 1X 1 + an 2x 2 + ... + annXn = 0
(we are considering only systems in which the number of equations
equals the number of unknowns).
If the determinant of this system is non-zero, according to Cramer's
rule the system has one definite solution, which in the given case is
zero:
tk=1, 2, ... , n)
[see (VIII.19)].
Consequently, for a homogeneous system of equations to have a
non-zero solution, its determinant must be zero. It can be proved
that this condition is not only necessary, but also sufficient.
To be able to discuss the nature of the solutions of the system
(VIII.26), we must acquaint ourselves with the concept of the rank
346 APPENDICES
of a matrix. If the number of rows m of a matrix differs from its

number of columns n, the determinant (VIII.4) cannot be compiled
for it. But by deleting certain rows and columns from the matrix,
we can form a determinant from the remaining rows and columns.
The determinants obtained in this way are said to be included in the
composition of the matrix. Tho highest possible order of those deter-
minants equals the minimum of the numbers m and n determining
the size of the matrix, while the smallest order of these determinants
is unity, the first-order determinants being elements of the matrix.
Assume that all the determinants of the order l .\n the composition
of a matrix are zero. Hence, all the determinants of the order (Z +1)
are also zero (this follows from the property 4 of determinants). In
other words, if all the determinants of the order lin the composition
of a matrix are zero, all the higher-order determinants are also zero.
The highest order of a non-zero determinant in a matrix is called
its rank. Hence, the fact that the rank of a matrix is r signifies that
among the r-order determinants in the matrix at least one is non-
zero; all the determinants of a higher order are zero here.
The concept of the rank, naturally, may also be applied to a square
matrix. For example, with observance of condition (VIII.18), the
rank of the matrix formed from the coefficients of the non-homogen-
eous system of equations (VI II .17) is n.
For the homogeneous system (VIII.26) to have a non-zero solution,
its determinant must be zero. In other words, the rank of the matrix
formed from the coefficients of the system must be less than n. Assume
that the rank of the system's matrix is r (where 1 ~ r < n). Here
there are n - r linearly independent solutions:
x~ 01 >, x~01 >, •.. , x~fl (cc = 1, 2, ... , n - r)
The values of the unknowns determined by the expressions
(VIII.27)
where Ca are arbitrary constants, will be the most general solution.

In the '[larticular case when the rank of a system's matrix is r =
= n - 1, there is only one linearly independent solution. The
fol1owing values of the unknowns can be proved to be this solution:
(VIII.28)
where A 1d is the algebraic cofactor of the element a1d in the deter-
minant D of the system, cis an arbitrary constant, and k is selected
so that at least one of the A 11 /s (where i = 1, 2, ... , n) is non-zero.
The values of A 11 t obtained with a different selection of k correspon-
ding to the condition indicated above differ from one another by
a common factor that can be included in the constant c. Hence, the
form of the solution (VIII.28) does not depend on how k is selected.
APPENDICES 347
Assume that a set of values x 1 = q 1 satisfies the system (VIII.26).

It is not difficult to see that the values x 1 = f..q 1 (here f.. is an arbi-
trary constant) also satisfy the system. This explains the presence of
the factor c in formulas (VIII.28). It follows from the above that the
system (VIII.26) uniquely determines only the ratios x)xk, while
the Yalucs of x 1 themselves are determined to within an arbitrary
factor.
The problem of solving a system of homogeneous equations can be
given the following geometrical interpretation. We shall consider
the set of quantities (x 1 , x 2 , • • • , xn) as an n-vector x; similarly we
shall consider the set of quantities (a 11 , a 12 , • . • , a 1n) as ann-vector
a 1 (there will be n such vectors). The system (VIII.26) can therefore
be written as
a 1x =0 (i = 1, 2, ... , n)
[see formula (VI.4t)], and the problem itself formulated as follows:

n vectors a 1 are set in an n-dimensional space. It is necessary to find
a vector x such that would be perpendicular to all the vectors a 1•
It is evident that the multiplication of the vector x by the scalar c
does not violate its orthogonality to the vectors a 1• Therefore, the
unknowns x 1 , x 2 , • • • , Xn are determined by the system (VIII.26)
to within the arbitrary factorc so that the value of one of the un-
knowns can be chosen arbitrarily (for instance, we can assume that
x 1 = 1); now the values of the remaining unknowns will be deter-
mined uniquely (they will be expressed in terms of x 1).
IX. Quadratic Forms

A quadratic forin f of the variables x1 , x 2, . . . , X 11 is defined to
be a homogeneous polynomial of the second degree in these variables.
Such a polynomial can be written as
n
f= ~ ao,xtxlt (IX.1)
i. k=1
where
(IX.2)
are constant quantities that may be either real or complex. If all

the coefficients a 1k are real, the quadratic form is said to be real.
If at least one of the coefficients is complex, the quadratic form is
said to be complex.
A real quadratic form is called positive definite (negative definite)
if this form has positive (negative) values for any real values of the
variables x 10 x 2 , • • • , Xn not equal to zero simultaneously.
348 APPENDICES
In the following, we shall consider only real quadratic forms.

The symmetric matrix
au a12 ••• atn
A= a2t a22 • • • a2n (IX.3)
compiled from the coefficients of the polynomial (IX.1) is called

a matrix of a quadratic form. It is obvious that a quadratic form is
completely determined by its matrix.
The determinant
D (A) = det II ail, II (IX.4}
compiled from the coefficients of a quadratic form is called its.
discriminant.
A quadratic form can be written as the product of three matrices:
X2 = XAX (IX.5)
ant an2 • • • ann Xn
where X is a column matrix, and X is its transposed matrix. Indeed,
the product of a column matrix and a square matrix is a column
matrix [see (VII.32)1 whose elements in the case being considered are
2Ja 1kxk (i=1,2, .•. ,n)
k
The product of a column matrix and a row matrix is simply a function
[see (VII.33)]. In the given case, this function is
,..., x, L a,kxk
i k
The last expression can be seen to be identical to expression (IX.1).
A quadratic form such as
(IX.6)
containing no terms with products of different variables is known

as a canonical one. Such a form has a diagonal matrix and is there-
fore also called a diagonal quadratic form. It is obvious that if all
the 'Ak' s are greater (smaller) than zero, the form (IX.6) will be
positive definite (negative definite).
Any quadratic form can be reduced to a diagonal form with the
aid of a non-singular linear transformation 1 • Let us go over from the
1 A linear transformation B (where B is a matrix) is called non-singular
when its determinant is non-zero: D (B) =I= 0.
APPENDICES 349
variables x 1 , x 2 , • • • , Xn to the new variables y1 , y 2 , • • • , Yn associated

with the previous variables by means of the linear relations
x, = ");k
btkYk (IX.7)
Let us consider the collection of quantities x 1 and the collection of

quantities Yk as column matrices. This allows us to write formula
(IX.7) as
X =BY (IX.8)
where B is the matrix of the linear transformation whose elements
are the coefficients btk·
Introducing the values of x 1 determined by relations (IX.7) into
formula (IX.1), we find an expression of the quadratic form in the
new variables:
f = )1, a,k lJ buYz 2J bkmYm = ~ YzYm 2J a,R.bilbkm = ~. CzmYIYm (IX.9)
i, h. l m !, m i. k !, m
where
(IX.10)
It is not difficult to see that the condition a 1k = aki results in

Czm = Cml·
Let us write expression (IX.10) as follows:
,., ~ '\1- ~
Czm = ~ bu ~ a,R.bkm = Ll b11 ~ a,kbkm
i k i k
[we have replaced the elements of the matrix B with the correspond-
ing elements of the transposed matrix ii, see formula (VIl.7)].
According to (Vll.28), lJ a;kbkm is (AB);m-an element of the
k
matrix obtained by multiplying the matrices B and A. Similarly,
~b,, (AB) tm is (i3AB) 1m-an element of the matrix obtained by
i
multiplying the matrices (AB) and B. Consequently, the matrix C
whose elements are determined by formula (IX.10) can be written as
BAB c= (IX.11)
By (VIII.15), the determinant D (C) of the matrix C is
D (C) = D (B) D (A) D (B)
Since the matrices B and B differ in the columns being interchanged
with the rows, while the determinant does not change in this case
[see (VIII.8)1, we have D (B) = D (B), and we can write
D (C) =D (A) [D (B)] 2 (IX.12)
350 APPENDICES
Hence, in a linear transformation of variables, the discriminant of

the quadratic form is multiplied by the square of the determinant of
the transformation from the new variables to the initial ones.
If the transformation B is orthogonal [this signifies that the coeffi-
cients bill satisfy the conditions (VII.40) and (VII.41)], the transposed
matrix coincides with the inverse one: B = B- 1 [see (VII.8)]. Con-
sequently, for an orthogonal transformation, formula (IX.11) is as
follows:
C = B- 1 AB (IX.13)
The matrix C determined by relation (IX.'13) is a matrix of a
quadratic form in the new variables YR.· Let us find an orthogonal
transformation B such that the matrix C will be diagonal, i.e. will
have the form
"-t 0 ... 0
C=A= 0 1.. 2 ••• 0 (IX.14)
0 0 .. • l..n
Now the quadratic form in the new variables Yk will be canonic

[see formula (IX.6)].
We multiply both sides of formula (IX.13) by B. Since BB- 1 = I,
and multiplication by a unit matrix does not change the second
factor [see (VII.36)], we arrive at the relation
BC =AB
or
(IX.'15)
In accordance with (IX.14), the elements Cmk can be written as

Cmk = l..m8mk· Substitution of this value of Cmk into formula (IX.15)
yields
In tne sum on the left, only the addend with m = k will be non-zero,
and it equals bikl..k. We thus arrive at the equation
bikl..k=~aimbmk (i, k=1, 2, ... , n) (IX.16)
m
Expression (IX.16) can be treated as a set of n 2 equations with n 2

unknowns btk· These equations can be divided into n groups (differing
in the values of the subscript k). Each group consists of n equations
differing in the values of the subscript i. Transferring all the terms
in (IX.16) to one side, we can write the k-th group of efluations
APPENDICES 351
in the form
(au- AI!) b111
• •a:1b.lh.
+ a12b211
+(a~2~.A:,).b2k:.-:·.~
+ ... + a1nbn1t
.a~n~n~.
=0
=0
I
(IX.i 7)
anlbtk + an2b211 + ••· +(ann- Ak) bnk = 0
For the system of equations (IX.17) to have non-zero solutions,
its determinant must be zero (see Appendix VIII):
a 11 - A a 12 a 1n
a21 azz- A . . . a2n =0 (IX.18)
(we have dropped the subscript on A because similar conditions are

obtained with any k). This expression is an algebraic equation of
the n-th degree in the unknown A.. It is called the characteristic
equation of the matrix A.
Equation (IX.18) has n roots: A. 11 A. 2 , • • . , "-n• which are elements
of the required diagonal matrix (IX.14). Introducing the values of
'A 11 in turn into the system (IX.17) and solving this system relative
to the unknowns bik• we find the elements of the matrix of the tran-
sition from the variables y 11 in which the quadratic form is diagonal
to the previous variables xi [see (IX.7)]. The transition from the
variables xi to the variablei< y 11 can be performed with the aid of
the inverse matrix B-1 • Since in accordance with our condition
D (B) =1= 0 (the transformation B is non-singular), an inverse matrix
exists; its elements can be evaluated by formula (VIII.23).
The roots of Eq. (IX.18) (i.e. the quantities 'Ah) are real. This
follows directly from relations (IX.16), if we take into account the
real nature of the quantities ai 11 and b 0,. The matrix (IX.14) will
thus be real.
Consider the following diagonal quadratic form:
(IX.19)
It is positive definite. Its matrix is the unit matrix

A= l=il6,k II (IX.20)
Let us apply the arbitrary orthogonal transformation B to the
variables x 1 and see what the quadratic form (IX.19) will be in the
new variables. By (IX.13), the matrix C of the quadratic form in
the new variables is determined by the expression
C = B-1IB = B-1B =I
.352 APPENDICES
{see formulas (VII.36) and (VII.38)], Consequently, the quadratic

form (IX.19) in the new variables is
/=~ Yl
i
Any orthogonal transformation of variables thus leaves a quadratic
form of the kind given by (IX.19) unchanged.
Let us consider the real quadratie form
(IX.21)
where z1 , z2 , • • • , Zn are complex quantities1 • We shall represent

Zm as Xm +
iym, and zh. correspondingly as Xn - iYn· Hence,
f = ~ ah.m (xh.- iYn) (xm + iYm) = ~ ah.m (XnXm + YkYm)
h.,m h.,m
(IX.22)
Let us interchange the dummy indices k and m in the sum deter-

mining the imaginary part f 2 in (IX.22):
/2 = ~ ah.m (XnYm -XmYh.) = ~ amh. (XmYh. -Xh.Ym)
h., m m, k
=- 2J
k, m
ah.m(XnYm-XmYn)=-ft
{we have taken advantage of the fact that amh. = ah.m; see (IX.2)).
The relation / 2 = -/ 2 is possible only when / 2 = 0. We have thus
proved that the quadratic form (IX.21) has real values at any com-
plex z's. By (IX.22), let us write it as
f = ~ akmXh.Xm + ~ akmYh.Ym = fo (xh.) + fo (Yk) (IX.23)
h.,m h.,m
where
(~;ecall that a quadratic form is determined completely by its matrix

and does not depend on the designation of the variables).
Simultaneous Reduction of Two Quadratic Forms to a Diagonal
Form. Assume that we have two quadratic forms:
(IX.24)
/ 2 = ~ b1nx 1xk (IX.25)

i, k
1 Recall that a quadratic form with the real coefficients akm is said to be
real. The form (IX.21) is a particular case of the Hermitian form in which the
coefficients, generally speaking, are complex and satisfy the condition aik =
*
= aM. .
APPENDICES 353
We shall show that if one of them, say / 1 , is positive definite, it is

possible to find a linear transformation of the variables that reduces
both forms to a diagonal one. We shall carry out the required trans-
formation in several steps. First, using the orthogonal transformation
F, we shall pass over to the variables v1 in which the form (IX.24)
acquires a diagonal form:
/1 = 2J ~,vr
i
(such a transformation was considered earlier in detail).

Now let us pass over from the variables v 1 to the variables
u, = v, v ~~
(it can readily be seen that with at least one ~~=I= 1, this transfor-
mation is not orthogonal). Since the form / 1 is positive definite, all
the coefficients f.tt are positive so that the variables u 1 will be real.
In these variables
t1=..:::..
''i Ut
I
hence, I will be the matrix of h·

Finally, let us go over with the aid of the orthogonal transfor-
mation G from the variables u 1 to the variables y 1 such that the form
f 2 will become diagonal. The form f 1 here remains diagonal because,
as was shown above, a quadratic form with the matrix I [see (IX.20)l
does not change in any orthogonal transformation. Consequently, in
the variables y 11 the quadratic forms (IX.24) and (IX.25) will be
diagonal:
(IX.26)
The entire sequence of transformations can be represented by the

diagram:
(transformation F) (XV f.t 1) (transformation G)
x1 - Vt - Ut - Yt
'i,:Bk
a 1kXtXk- 2J
i
f.ttVi -
'i0 2
L Ut
i
"\'
-..:....;
i
Yi2
2J b11,x 1x11.- L:,

i, k i, k
bi~~.v 1 v~~.- )' bikuiuk- ~. A.,yi (IX.27)
i, k i
To establish a method of finding the coefficients 1. 11 let us compile

the auxiliary quadratic form
f=/ 2 -A.ft=~ (b 1 ~~.-t.a 1 ~~.)x 1 x~~.=~ (t.~~.-t..)yi (IX.28)
i, k
23-018
354 APPENDICES
whose coefficients contain the parameter A.. The discriminant of this

form in the variables x 1 is
det II b;~~. - A.a;~~. II (IX.29)
and in the variables y 1
det II (A.~~. - A.) ~~" II = (A.l - A.) (A.2 - A.) • • • (A.n - A.)
(IX.30)
Denoting the transformation of the direct transition from the
variables x 1 to the variables y 1 by the letter B (this transformation,
generally speaking, will not be orthogonal), let us write the following
expression for the matrix of the quadratic form (IX.28) in the
variables y 1:
c= BAB
where A stands for the matrix of the form (IX.28) in the variables x 1
[see formula (IX.11)]. According to (IX.12)
D (C) = D (A) lD (B)] 2
where D (C) is the determinant (IX.30), D (A) is the determinant
(IX.29), and D (B) is the non-zero determinant of the matrix of the
transformation B in which the parameter A. is absent lD (B) =F 0
because the transformation B is non-singular)].
Hence,
(A. 1 - A.) (A.2- A.) •.• (A.n- A.) = det II btk- A.atk II lD (B)JZ
The substitution of any of the values of A.~~. for A. makes the left-hand
side of the equation and, consequently, the factor det II b 111. - 'A.a 1k II
vanish. Therefore, the quantities A.~~. are the roots of the equation
!
bu- 'A.au b12- 'A.au ... bin~ Aatn

b21- A.a21 b22- Aa22 · · · b2n- Aazn = 0 (IX.31)
bnl- 'A.ant bn2- A.an2 · · · bnn- Aann
The task of reducing the quadratic forms (IX.24) and (IX.25) to a
diagonal form thus consists in finding the roots of Eq. (IX .31 ).
Let us consider in addition to the quadratic form in the variables x 1
f= >:
i, k
a 1kxixk (IX.32)
the similar quadratic form in the variables x 1

I ~l 0 0
f = L...i aikXtXk (IX.33)

i, h
where x 1 is the derivative of the Yariable xi with respect to a para-

meter t. Let u's call ~~ the i-th velo~ity.
APPENDICES 355
In the linear transformation

(IX.34)
m
from the variables x 1 to the new variables Yi> the velocities experience
the same lransformation:
(IX.35)
m
We express the forms (IX.32) and (IX.33) in the new variables.

To do this, we introduce expressions (IX.34) and (IX.35) into
formulas (IX.32) and (IX.33):
f = 1j ail,xixk = ~ a;k lj C;mYm \ ; Ck!Yl
i, k i, k m l
= 1j YmYl ~ aikcimckl = ~ bmlYmYl

m, l i; k m, l
where
(IX.36)
where bml has the same values (IX.36) as in the preceding case.
Hence, in any linear transformation from the variables x 1 to the
.
new variables y 1, the coefficients of the quadratic form of the velo-
cities x 1 are transformed in exactly the same way as the coefficients of
the similar quadratic form of x;. On these grounds, the diagram
(IX.27) can be modified as follows:
( xV ~~)
(transformation F)
- (transfotmation G)
---;.. y!
,,' y.
-
Xt ---;.. vi U;
'\1 • • ,.., •2 '\1 0 2
..._] a 1kXtXI!.- ~ ~;Vt ...::...; Ut ---;.. ~
- 1
i, k i i i
'')"'
....._) bikXtXI!.- ._:
'\1 b'ikV;VI!.- )' '\1
b"ii!.U;Uk- ~ f. tYi2 (IX.37)
i, k i, R. r;-k i
The quadratic form 1j a 1 k~i~l!. is assumed to be positive definite.
X. Tensors
1. Definition of a Tensor. To arrive at the concept of a tensor,
let us consider the polarization of an anisotropic dielectric.
23*
356 APPENDICES
In an isotropic dielectric, the polarization P is proportional to the

electric field strength E:
P = xE (X.1)
where X is the dielectric susceptibility. According to (X.1), the
vectors P and E are collinear.
In an anisotropic dielectric, the polarizability differs in different
directions. As a result, the direction of the vector P, generally
y
speaking, does not coincide with
that of the vector E. Experiments
show that in any anisotropic di-
electric there are three mutually
perpendicular directions such that
when the direction of E coincides
with one of them, the vector P is
collinear with E. These directions
are called the principal ones. Let
us direct the coordinate axes along
the principal directions of a dielec-
tric (Fig. X.1). The arbitrarily
Fig. X.1. directed vector E can be resolved
into the components Ex, Ey, and Ez
(the last component is perpendicular to the plane of the drawing).
The component Ex will set up the polarization p X = xxEx collinear
'with it, where Xx is the susceptibility in the direction of the x-axis.
Similarly, the other two components will set up Py = XyEy and
Pz = XzEz. It is not difficult to note that with different values of
Xx• 'Xy• and Xz the resultant vector P = P x + P y + P z will not be
·collinear with E. '
; · Let us take an anisotropic dielectric which we shall consider to
he a homogeneous unbounded medium. We associate with it a Car-
tesian coordinate system whose axes are oriented absolutely arbi-
trarily and coincide with none of the principal directions of the
dielectric: With the field Ex directed along the x-axis, not only Px,
but also Py and Pz will be non-zero, and
Px = XxxEx; Py = 'XyxEx, Pz = 'XzxEx (X.2)
where Xxx• Xyx• and Xzx are coefficients of proportionality between
Ex and the relevant components of P.
Similarly, the fields Ey and E z will cause the polarizations
.· P x = 'XxyEy, P y = 'XyyEy, P z = 'XzyEy }
(X.3)
Px='XxzEz, Py=XyzE:, Pz='XzzEz
With a field E not coinciding with any of the coordinate axes,
Ex, Ey, and Ez will exist simultaneously so that all the P/s deter-
mined by formulas (X.2) and (X.3) will appear. Combining the
APPENDICES 357
relevant components of the vector P, we find that

P :x = X:x:xE:x+X:xyEu+ x~zEz }
Py = XyxEx+XyyE"+Xy:Ez (X.4)
Pz= X:xE:x+XzyEu+XzzEz
Using numerical subscripts instead ofletters, we can write equations
(X.4) in a compact form:
P,=~XikEk (i=1, 2, 3) (X.5)
k
It follows from the above that to characterize an anisotropic

dielectric, we must set nine quantities Xik (a single quantity X
was sufficient for an isotropic dielectric).
Let us now go over from the previous coordinate system x1 , x 2 , x8
(the system K) to a new system x~, x;, x~ (the system K 1 ) whose axes
also do not coincide with the principal directions of the dielectric.
We shall find out how the quantities Xik transform in such a tran-
sition. In the new system of coordinates, the equations relating P(
and Ei are similar to equations (X.5):
P i = LJ Xik E'k
I ,., I
(X.6)
k
Here Xtk are the nine quantities characterizing the dielectric in the
new coordinate system.
By formulas (VI.37) and (VI.38), the components of the vector P
in the transition from the system K to the system K 1 are transformed
by the formula
Pi=~ a. 11 P1 (X. 7)
I
and the components of the vector E in the transition from the system
K' to the system K are transformed by the formula
(X.8)
(Recall that a.ih = eieh is the cosine of the angle between the i-th
primed and the k-th unprimed coordinate axes.)
Let us replace P 1 with Em in (X.7) according to relation (X.5).
The result is
Pi=~ a.uPz = ~ O.tz
I I
lJm XlmEm = I,lJm a.i!XImEm
Now let us substitute into this equation Em from formula (X.8):
Pi= ~ a.uX1m ~ CX~tmE/t =
1. m k
lJk Ei I,~m O.tzCXkmXIm
358 APPENDICES
Comparing the expression obtained with expression (X.6), we

find that
(X.9)
The set of the nine quantilie'l T 111 that transform in the transition
from the coordinate system K to the system K' by the formula
(X.10)
is known as a tensor of the second rank (or a tensor of the second

order).
The reverse transformation (from the system K' to the system K)
is performed by the formula
(X.11)
A tensor is written in one of the following three ways:
Tu T12 Tta)
T=(T 0 ,)= ( T2t T22 T2a (X.12)
Tat Ta2 Taa
The quantities T 1, are called the components of a tensor. The com-
ponents T 11 , T 22 , and T 33 are said to be diagonal.
Hence, the properties of an anisotropic dielectric are described by
the dielectric susceptibility tensor
(Xtl<) = (~:: ~:: ~:: ') (X.13)

Xat Xa2 Xaa
Of special interest is the case when the coordinate axes coincide
with the principal directions of a dielectric. Now the field component
E 1 sets up only the i-th component of the polarization, and
Pt = xuEi (i = 1, 2, 3)
A comparison with (X.5) leads us to the conclusion that only the
diagonal tensor components are non-zero in the present case so that
the dielectric susceptibility tensor is
(Xi~<)= (0Xt X20 0)

0 (X.14)
0 0
Xa
(we have left only one subscript on the non-zero tensor components
because both subseripts are the same for diagonal component<;).
APPENDICES 359
A tensor in which only the diagonal components are non-zero 1

is said to be reduced to the principal axes. The values of the diagonal
components obtained in this case are called the principal values of
the tensor.
Vve must note that in an isotropic dielectric all three principal
values of the dielectric susceptibility tensor arc the same: x1 = x2 =
= x3 =X· For an isotropic dielectric, any three mutually perpendicu-
lar directions can be taken as its principal directions.
Tensors not only of the second rank, but also of other ranks are
considered. For instance, a tensor of the third rank is a collection of
27 quantities Tu<l that transform in a transition from one coordinate
system to another by the formula
(X.15)
Tensors of other ranks are determined similarly. A tensor of rank

r has 3r components. It is a simple matter to see that a vector is a
tensor of the first rank (it. has 3 1 = 3 components), and a scalar is
a tensor of the zeroth rank (it has 3° = 1 component).
The concept of a tensor can readily be extended to ann-dimensional
space. A tensor of rank r in such a space (an n-tensor of rank r) is
defmed to be a set of nr quantities T;R...p (altogether r subscripts)
that transform according to a formnla differing from formula (X.15)
only in that the dummy indices take on n values instead of three
in summation.
V/e shall consider some more examples of tensors of rank two.
Let us take two vectors a and b and form products of the kind
Ilik = aibk (X.16)
from their components. It is not difficult to see that these products
are transformed by formula (X.10), i.e. have properties of the com-
ponents of a tensor of the second rank.
The tensor
(6ilt) = (100)
0 '1 0 (X.17)
0 0 '1
is called a unit tensor. By transformation formula (VI.39), its com-
ponents in the new coordinate system are
6i11 = LJ
l, m
ailallmOzm = )' aila~tz
I
= 0,11
[we have taken advantage of the property of the coefficients a11t
expressed by formula (X.39)). The components of a unit tensor
1 We must note that this is possible only with specially chosen coordinate
axes.
360 APPENDICES
are thus identical in all coordinate systems. Tensors having this

property are said to be invariant.
2. Tensor Algebra. Let us consider the fundamental operations
with tensors.
The sum of the tensors T lit and Gilt is defined to be a tensor with
the components
~~It= T11t +Gtk (X.18)
In accordance with (X.18), any tensor can be written as the sum of
two (or more) tensors.
The product of the tensor Tilt and the scalar a is defined to be the
tensor Gilt with the components
Gt~t = aTt~t (X.19)
The product of the tensors T tit and G1m is defined to be the tensor
IT 1 ~t 1 m
of the fourth rank with the components
Tii~tlm = Tt~tGzm (X.20)
The product of tensors of other ranks is determined similarly. Par-
ticularly, we have considered [see (X.16)] the product of tensors of
rank one, i.e. vectors. Examination of definition (X.20) shows that
the rank of a product tensor equals the sum of the ranks of the mul-
tiplier tensors (tensors of different ranks can also be multiplied).
By the contraction of a tensor is meant the following operation:
two subscripts on the tensor components are assumed to be the same
and summation is performed over them. The quantity obtained as
a result of such an operation is called the contraction of the tensor.
Contraction obviously lowers the rank of a tensor by two units.
The contraction operation can be used for tensors of a rank not lower
than the second one 1 • For a tensor of the second rank, its contraction
is a tensor of the zero rank, i.e. a scalar. The latter is called the
trace of the tensor (compare with the trace of a matrix, p. 334). It
equals the sum of the diagonal components:
Tr (T til)= 2J T u (X.21)
i
A scalar does not change upon the transformation of coordinates.

Consequently, the trace of a tensor is an invariant. For example, the
trace of the tensor (X.16) is the scalar product of the vectors a and b
which, as we know, is invariant relative to the transformation of the
coordinates [see the text following formula (VI.28)].
In physical applications, the multiplication of tensors is custo-
marily used in combination with the subsequent contraction of the
1 For a tensor whose rank r is higher than the second, contraction can be
performed in several ways (over different pairs of subscripts). The result will
be different tensors of the rank r - 2.
APPENDICES 36t
expression obtained. The result of these operations is known as the-

scalar product of tensors1 • A typical example is the scalar product
of vectors (tensors of the first rank) which we have just mentioned:
the tensor a 1b11 is contracted over the pair of subscripts i and k,
and the result is the expression ~ a11 b11 .
In accordance with the above, by a scalar product of the tensor·
T 111 and the vector a 11 is meant the vector b 1 with the components-
b1=2JT111a11 (X.22)
It
Let us convince ourselves that the set of quantities b 1 determined

in this way does form a vector. For this end, let us find the law of
transformation of the quantities b1• It is evident that
bi = 2J Ti11ak
It
Let us introduce into the last expression the values of Ti 11 and a~,
in terms of the unprimed components:
b( = ~ T(ltak =
I&
2JIt l,2Jm auakmTlm 2J• aksas = 2Jl au m,2Js Tzmas 2JIt akmak,
According to the property (VI.40), ~ akmak, = llms• Consequently,.
. 11
bi =~au 2J Tzmasl3m = ~l au2Jm Tzmam =

8 ~ aubz
l m, s l
The result we have obtained signifies that the quantities b 1 are-

transformed according to the law of transformation of vector com-
ponents. Hence, the quantities b 1 determined by expression (X.22) do-
indeed form a vector.
In a similar way, we can convince ourselves in the correctness of
the following statement: if a set of nine quantities X 111 taken with
the components of a vector a 11 in the combination
~· xlkak
It
gives the components of another vector bh the quantities X 111 are-

the components of a tensor. We encountered such a situation at the-
beginning of this Appendix. The set of nine quantities X11t taken
with the components of the vector Ek in the combination (X.5) gave-
the components of the vector P 1• On these premises, we showed that
the quantities Xllt are transformed according to the law of transfor-
mation of tensor components, i.e. that the dielectric susceptibility-
is a tensor.
1 The latter product is sometimes called an inner one, whereas expressio!ll
(X.20) is called an outer product of tensors.
APPENDICES
As a n~sult of the scalar multiplication of the vector a and the

tensor T, we obtain the new vector b:
b = Ta (X.23)
The tensor T can therefore be considered as a linear operator trans-
forming one vector into another one.
Let us fmd the product of a unit tensor and the vector ,ah. By for-
mula (X.22), we have
b1 = 2j 81hah = a 1
h
Hence, a vector does not change when multiplied by a unit tensor.
3. Symmetric and Antisymmetric Tensors. The tensor S;h whose
·Components satisfy the condition
(X.24)
is said to be symmetric. We must note that the dielectric susceptibil-
ity tensor treated above is symmetric.
The tensor A ih whose components satisfy the condition
A ih = -A"; (X.25)
is said to be antisymmetric.
The property of symmetry or antisymmetry belongs to the tensor
itself. This follows from the fact that this property is retained in
any transformations of the coordinutes. We shall give the following
-calculations to prove this statement:
Sih = 2j ailahmSlm = 2j CX;maltlsml = .'B ahlaimslm =ski
l, m m.l l. m
{we ftrst interchanged the dummy indices l and m, and then replaced
Smz with Szm equal to it). We have proved that a tensor symmetric
in the system K will be symmetric in any other system K'.
Similarly
Aih =-~ ailahmAlm = ~ CX;mal!lAmz =- ~ ahlCX;mAlm = - Alti
l, m m, l l. m
·(aftBr interchanging the dummy indices, we replaced Am 1 with

-A 1m equal to it). We have proved that a tensor antisymmetric in
K will be antisymmetric in K' too.
Any tensor T ik can be written as the sum of a symmetric and an
.antisymmetric tensors. Indeed, let us write Til, as
f. = T;h+Thi + Tih-Thi
zh 2 2
The lawfulness of such an expression is obvious. At the same time,
the first term does not change when the subscripts i and k are inter-
<:.hanged, i.e. it has the properties of the components of a symmetric
tensor; the second term, on the other hand, changes its sign when
APPENDICES 363
the subscripts are interchanged, i.e. has the properties of the com-
ponents of an antisymmetric tensor.
We can therefore always consider that
Tu, = Su, .4 111 + (X.2G)
-where
S T;k+T,,; A _ T;11-T1t1
ih = 2 ill- 2 (X .27)
The concepts of symmetry and antisymmetry can also be applied
to tensors of a higher rank. For instance, Till! is said to be symmetric
{antisymmetric) relative to the subscripts i and k (or i and l, or
k and l) if upon transposition of these subscripts the components of
the tensor do not change (reverse their sign). If the components of
a tensor do not change (or reverse their sign) upon the transposition
of any pair of subscripts, the tensor is said to be absolutely symmetric
(absolutely antisymmetric).
The set of 27 quantities e 1" 1 introduced in Appendix VI forms
an absolutely antisymmetric tensor of the third rank. Recall that
the quantities e 1111 (1) are zero if any two subscripts have identical
values; (2) equal +1 if all the subscripts are different and form a
cyclic transposition of the sequence 1, 2, 3; (3) equal -1 if all the
subscripts are different and form a cyclic transposition of the sequence
3, 2, 1. Of the 27 components of this tensor, 21 are zero, three are +1,
and three are -1. It can be shown by simple but cumbersome cal-
·cnlations that
eiltz = 2J
aimakpazsf:mps = eilil
m. p, s
(X.28)
i:e. that the tensor e 1" 1 is invariant.
4. True Tensors and Pseudotensors. Let us see how the components
-of a tensor transform upon the inversion of the coordinate axes. The
transformation coefficients in this case have the values
au, = -(~ill
(see (VI.42)). Therefore, the formula for the transformation of the
·components of a tensor of rank r, Tilt ... ! (altogether r subscripts), in
inversion is
1'ik... z= ~ (-<51m)(-61!p) ··· (-6z.)1'mp ... a
m, p .•• s
= ( -1Y ~ 6;m61tp ... 6z.T mp .. .• = ( -1Y Tik ... z
m, p ... s
Hence, the components of a tensor of rank r (we have in mind
tensors in three-dimensional space) are transformed in inversion
by the formula
Ti11 ... 1 = (-1}T Tik ... z
By (X.28). the quantities e.ihl in any transformation of the coor-
dinates (and, consequently, in inversion) remain invariant, i.e.
364 APPENDICES
do not change their sign. The components of a true tensor of rank

three, on the other hand, must reverse their sign in inversion. This
is why the set of quantities BiltZ forms a pseudotensor instead of
a true one.
A pseudotensor of rank r is defined to be a set of 3r components
P 111 ,., 1 which upon rotations of the coordinate system behave like
the components of an ordinary tensor, and upon inversion are trans-
formed by the formula
PfR ... z = (-1)'+1 Pt~t ...l
Let us consider examples of pseudotensors of different ranks. We
acquainted ourselves with pseudoscalars and pseudovectors in
Appendix VI. We treated a pseudotensor of the third rank e 11tz
above. The set of quantities P ilt formed from the products of the
components a 1 of the true vector and Pit of the pseudovector, i.e.
P 1'11.- a1p'll.
is a pseudovector of the second rank. Indeed, upon inversion, the
quantities a 1 change their sign, while the quantities PR. remain un-
changed. Therefore, P i'll. in inversion will change its sign. At the
same time, we know that the components of a true tensor of the
second rank do not change their sign in inversion. Hence, P i'll. is
a pseudo tensor.
It is a simple matter to see that the contraction of a pseudotensor
also results in a pseudo tensor [the evenness or oddness of a tensor's
rank in contraction does not change, while the formulas for the
transformation of the components of the initial tensor and its contrac-
tion differ by the factor (-1) 11 }. For instance, in contracting the
pseudotensor P I'll.• we obtain the scalar product of a true vector and
a pseudovector, i.e. a pseudoscalar. ,
The product of a pseudotensor P 1R. .. ,l of any rank and a true tensor
T mp ... s of any rank is a pseudotensor:
n llt ... lmP ... a =
p iR. ... I T mp ... s
To verify this statement, let us compile the following table:
Table X.i
PiR. ••• l Tmp ••• s lliR. ••• Imp ••• s
rank I sign upon

Inversion rank I sign upon
inversion rank I sign upon
Inversion
even changes even does not change even changes

even changes odd changes odd does not change
odd does not change even does not change odd does not change
odd does not change odd changes even changes
APPENDICES 365
Similarly, we can see that the product of two pseudotensors of

any ranks is a true tensor.
5. Properties of a Symmetric Tensor of the Second Rank. Of the
nine components of a symmetric tensor of the second rank, only
six are independent (S 12 = S 2u
S1a = Sat• S2s = Sa2).
A symmetric tensor of the se-
cond rank allows an important
geometric interpretation. Before
considering it, we shall note that
the vector a can be represented t
not only by a directed segment
of a straight line, but also by ~
a plane whose equation is
ar = 1 (X.29)
where r is a position vector of a Fig. X.2.
point on a plane (Fig. X.2). Since
ar = ara = 1, we have ra = 1/a. Consequently, Eq. (X.29) defines
a plane perpendicular to the vector a and at a distance 1/a from the
origin of coordinates. Expression (X.29) can also be written as
arr = 1, whence r = 1/ar. Hence, on the straight line passing through
the origin of coordinates and having the direction n (see Fig. X.2),
the plane (X.29) intercepts a segment of the length
(X.30)
Now let us turn to the symmetric tensor S. We shall correlate

with it a surface determined by the equation
r (Sr) = 1 (X.31)
It is obvious that this surface does not depend on the choice of the
coordinate system in which the components of the tensor S and of
the position vector r are determined. The nature of the surface
depends only on the properties of the tensor S.
Let us choose an arbitrary coordinate system x1 , x 2 , x 3 and express
the left-hand side of Eq. (X.31) in terms of the components randS
in this system. By (X.22), we have
(Sr) 1 = ~ S lkxk
It
Sr.alar multiplication of r and the vector Sr yields
~ xt(Sr) 1= 2J x 1 '5: S t!txll = ~ S 1 ~~..x 1 x!t

i i It t.ll
366 APPENDICES
Consequently, Eq. (X.31) has the form

~ S;kXIXk =1 (X.32)
i. h
In the expanded form, this equation appears as follows (recall that
sik = sk,):
Sux~ + s22x:+ SaaX: + 2S12xlx2 + 2S2aX2Xa + 2Sa1XaX1 = 1
(X.33)
Equation (X .33) determines a second-degree surface whose centre
is at the origin of coordinates. In the applications of tensor calculus
to physical problems, the diagonal components S ii may be greater
than zero (this happens, for example, with the quantities Xii)· In
this case, the surface (X.33) is an ellip'-
soid. It is exactly this ellipsoid that is
the geometrical image of a second-
rank symmetric tensor (naturally, in
three-dimensional space), like a
directed line segment or the plane
(X.29) is a geometrical image of a
first-rank tensor, i.e. a vector.
Let us find the distance from the
centre of the ellipsoid to points on
Fig. X.3. its surface. To do this, we draw an
arbitrary straight line n (Fig. X.3)
from the centre of the ellipsoid and assume it to be the x 1 -axis.
Now the distance p from the centre of the ellipsoid to the point
P will equal the value of x1 when x 2 = x 3 = 0. Assuming in
Eq. (X.33) that x 2 = x 3 = 0, we find that S 11 xi = 1, whence
1
x 1 = P= VSu
[compare with (X.30)]. Hence, the distance p is the reciprocal of the
square -root of the tensor component S 11 evaluated for the condition
that the direction along which p is measured has been taken as the
x1.-axis.
In transformations of the coordinates, the coefficients at S ; 11 in
Eq. (X.33) change, but the ellipsoid itself does not depend on the
choice of the coordinate system. If we direct the coordinate axes
along the semi axes of the ellipsoid, the equation of the latter, i.e.
Eq. (X.33), as is known, becomes simpler and acquires the form
S 11x~ + S 22x: +
S 33xi = 1
This signifies that with this choice of the coordinate axes, the non-
diagonal tensor components vanish. Consequently, the principal axes
of the tensor coincide with the semiaxes of the tensor ellipsoid.
APPENDICES 367
If the coordinate axes are directed along the principal axes of the·
tensor (i.e. along the semiaxes of the tensor ellipsoid), the tensor
acquires a diagonal form:
(X.34)
We have introduced the notation S 11 = A- 1 , S 22 = 1.. 2 , S 33 = 1.. 3 •

The quantities A1 , A2 , and A3 are the principal values of the tensor.
When /.. 1 = A2 = A3 =" A, the tensor ellipsoid transforms into a
sphere. Particularly, a sphere of unit radius corresponds to the unit
tensor {) u,.
Let us multiply a vector a by the tensor (X.34). The result is the·
vector b whose components are determined by formula (X.22). The
components of the tensor (X.34) can be written as S ik = A. 18ik·
Consequently,
(X.35)
Let the vector a be directed along the first principal axis of the tensor.
Hence, a 1 = a, a 2 = a 3 = 0 [having taken the tensor in the form
of (X.:34), we assumed that the coordinate axes are directed along the
principal axes of the tensor). By formula (X.35), the components of
the vector b will be b1 = A1a 1 , b2 = b 3 = 0. This signifies that the
direction of the vector b coincides with that of the 'ector a, while
the magnitude of the vector b is A1 times that of the vector a. The
same also holds for the other two principal axes. Therefore, if the
vector a is directed along one of the principal axes of the tensor, the
following equation ·holds:
Sa = A.a (X.36)
where A. is the relevant principal value of the tensor.
We have arrived at the following result: when a vector having the
direction of one of the principal axes is multiplied by a tensor, the
direction of the vector does not change, but its magnitude grows by
a number of times equnl to the rolevnnt princ.ipnl value of tho tens,Jr.
Equation (X.3G) holds for uuy coordinate system if only the vector
a is directed along one of the principal axes of the tensor. This cir-
cumstance can be used to flnd the principal values and the principal
axes of a tensor. For this purpose, we must vary the direction of the
vector a until the vector Sa coincides in direction with a. The direc-
tion found gives a principal axis of the tensor, and the ratio of Sa
to a-the corresponding principal value. Analytically, this can be
expressed as follows. Let us write Eq. (X.36) in components (we have
not yet reduced the tensor to its principal axes so that, generally
::368 APPENDICES
-speaking, all the S 11/ s are non-zero):

Suat + St2a2 + Staaa =!..at
Sztat + S22a2 + Szaaa = f..az
Sa tat+ S szaz + S aaaa = A.aa
Combining like terms, we obtain
(Su-f..)at+S12a 2 +Staas =0}

S2tat+(S22-f..)a2+S2aaa =0 (X.37)
Satat + Saza2 + (Saa-A.) aa = 0
We have arrived at a homogeneous system of three equations with
the unknowns a 1 , a2 , as-the components of the required vector a.
For this system to have a non-zero solution, it is necessary that its
determinant be zero [see Appendix VIII, the text following formula
{VIII.26)1. Hence, the following condition must be satisfied:
Su-I.. 812 Sta
S21 s22-A. s2a = o (X.38)
Sst Sa2 Saa-1..
The roots of this cubic equation in A. are the principal values of the
tensor: A. 1 , A- 2 , A. 3 • Introducing one of these roots into the system
(X.37), we can find the ratios a 2 /a 1 and a 3 /a 1 that will determine the
direction of the vector a satisfying the equation (X.36), i.e. the rele-
vant principal axis of the tensor. By performing this operation for all
three 1../s, we find the directions of all the principal axes.
We have assumed that all the roots A. 1 , A. 2 , and A. 3 are different.
When Eq. (X.38) has multiple roots, clarification is needed. With
two multiple roots (A. 2 = As = A., 1..1 =I= A.), the tensor ellipsoid will
be an ellipsoid of revolution. Only the principal axis of the tensor
<:oincidtng with the axis of symmetry of the ellipsoid will be deter-
mined uniquely. Any two mutually perpendicular axes normal to
the axis of symmetry of the ellipsoid may be taken as the other two
principal axes. If all three roots of Eq. (X.38) are the same (A.1 =
= A. 2 = As = A.), the tensor ellipsoid degenerates into a sphere, and
any three mutually perpendicular axes may be taken as the principal
ones.
Let us consider the properties of a symmetric tensor as an operator.
By (X.36), the action of S on the vector a directed along one of the
principal axes of the tensor causes only the length of the vector to
change A. times. The action of S on an arbitrarily oriented vector
causes, generally speaking, a change in both the length and the
direction of the vector (see, for instance, Fig. X.1).
APPENDICES 369
6. Properties of an Antisymmetric Tensor of the Second Rank.

Let us now turn to an antisymmetric tensor of the second rank.
Condition (X.25) for the diagonal components has the form Au =
= -A 11 • This is possible only when A 11 = 0. Consequently, the
diagonal components of an antisymmetric tensor are zero. Of the
remaining six components, only three are independent. Hence, an
antisymmetric tensor of the second rank, like a vector, is determined
by three quantities:
A12 = - A21 = aa}
A:la= -A32=a2 (X.39)
A31 = -At3 =at
Using this notation, we can write an antisymmetric tensor as fol-

lows:
(X.40)
The components of the tensor (X.40) can be written as
Ath = LJl eu"zaz = eihtat + eih2a2 + eikaaa (X.41)
where e 11lz is an absolutely antisymmetric pseudotensor [see the text

preceding formula (X.28)]. Indeed, the components Au determined
by this formula are zero because eiil at any l equals zero. Further,
A 12 = £121 a 1 + e 122 a 2 + e123a 3 = O·a1 + O·a 2 + 1·a 3 = a 3
A13 = £131a1 + £1a2a2 + £13aaa = O·a1 + (-1)·a2 +
O·aa =-a2
etc.
It was shown above that the multiplication of a true tensor by
a pseudotensor results in a pseudotensor. The product of two pseudo-
tensors, on the other hand, is a true tensor. The fact that the quan-
tities a 1 when multiplied by the pseudotensor £ 11! 1 [see formula (X.41)]
yield the components of a true tensor of the second rank indicates that
these quantities have the properties of a pseudotensor of the f1rst
rank, i.e. a pseudovector.
Consequently, a pseudovector with the components a 1 determined
by formulas (X.39) can be correlated with any antisymmetric tensor
of the second rank. And, conversely, a true antisymmetric tensor
of the second rank whose components are expressed in terms of the
components of a pseudovector by formula (X.41) can be said to
conform to any pseudovector a 1• ,
Let us apply the operator A (here A is an antisymmetric tensor)
to a vector b. By (X.22) and (X.41), we obtain the vector c having
24-~18
APPENDICES
the components
c1 = LJ A,~~.b~~. = 2j BtJt.lbkal
II. "· l
Comparing the expression obtained with formula (VI.33), we arrive
at the conclusion that c can be written as the vector product1 of the
vectors b and a:
c = Ab =[hal
The vector c is normal both to the initial vector b and to the pseudo·
vec.tor a corresponding to the tensor A. Hence, the action of the
tensor A causes the vector to turn through the angle n/2. In addition~
generally speaking, the length of the vector changes.
XI.· Basic Concepts of Vector Analysis

Gradient. Let us consider a scalar field, i.e. a region of space to
each of whose points there corresponds a definite value of the scalar <p'
<p = <p (P) = <p (r) = <p (x1 , x 2 , x 3 )
whPre r is a position vector, and x1 , x 2 , x 3 are the Cartesian coordinates
of the point P.
An identical value of <p corresponds to all the points of the surface
determined by the equation
1 (x1 , x 2 , x 3 ) = const (XI.i)
Asurface having the form of (XI.1) is said to be a constant <p surface.
Such a surface can be drawn through any point of a field.
When displaced from the point P over the distance dr, the function
<p receives the increment
The last expression does not depend on the choice of the coordinates
x, i.e. is-an invariant. The set of quantities dx 1 forms the vector dr.
We can therefore state (seA the text following formula (VI.28)J
that, the quantities 8cp/8x 1 are the projections of a vector onto the
x 1-axis. This vector is called the gradient of the scalar <p and is
designated by the symbol grad <p. Hence,
grad<p= ~ :: e 1 (Xl.2}
j •
or., in conventional notation,
8~ 8~ 8~
grad <p= Tx e~+ay ey + a;:ez
1 Recall that a vector product is a true vector only if one of the multipliers
is a pseudovector. We could also conclude from this circumst..mce that a is not.
11. truo vector, but a pseudovector.
APPEN:PICES 371.
lt is a simple matter to extend the definition (Xl.2) to an n-di-

mensional space. In the latter case, the number of terms in formula
(XI.2) will be n instead of three.
We shall prove that the componEmts of a gradient are transformed
by formulas (VI.37). Having taken two coordinate systems, K and
K',1 we can write
dcp = ~ -o<p- dx 1 = ~ --,
o<p d ,
Xf (XI.3)
OXt OX;
i I
Let us express dx 1 in terms of dx;. by formula (VI.38) and introduce
these expressions into (XI.3):
"" O<p ""
LJ OXt LJ ak I d Xk I
= O<p d :z:iI
LJ ox~
""
i h I 1
We change the sequence of summation over the subscripts t and k

on the left-hand side:
~ dxk ~au 0°cp
Xt
= ~ 8<r,
8xi ax;
k i i
The subscripts i and k are dummy ones. We have already noted
that adummy index can be designated by any letter. Therefore, the
snm on the left will not change if we interchange the subscripts i
and k. The result is
~ dxl ~ tJ. tk 8°cp
Xk
= ~ 0°<p,xi dx;
i k i
It follows from the relation we have obtained that
fJ <p "" o<p
Ox'
i
= Li
k
all! axk
which coincides with (VI.37). \Ve have thus shown that the quantitie~
ar;p/ax 1 upon transformations of the coordinates behave like the
eomponents of a vector. .
Willin111 Hamilton introduced the vector differentinl operator \7:
(the del operator or the Hamiltonian operator) that is a vector with
the components a/ax, a/ay, and 8/oz:
(XI.4).
The vector \i' has no meaning by itself. It acquires a meaning when·

applied to a scalar or vector function. For instance, upon the sym:.
bolic multiplication of V by cp, we obtain the gradient of cp:
! •·.• j ··-
372 APPENDICES
Consequently, Vep = grad ep. By (X1.3), the increment of ep can be

written as the scalar product of the vector grad cp and dr:
· dcp =grad <p·dr = ('Yep) dr (XI.5)
In movement over a constant cp surface, dfp = 0. It thus follows in

accordance with (XI.5) that the vector 'Yep at each point of the fwld
is directed along a normal to the constant cp surface. Let us find
the rate of change in ep along a certain direction l, i.e. dep/dl. By
(XT.5), the increment of ep over the segment dl is ('Yep) dl = (Vcp) 1 dl,
where ('Vep)z is the projection of the gradient onto the direction l.
Therefore, depldl = [(Vep) 1 dl]ldl = ('Vep) 1• Hence, the projection of
the gradient onto a certain direction gives the rate of change in the
function in the given direction.
We must note that the vector vep exists at every point of the scalar
field ep. Consequently, a gradient forms a vector field, i.e. a region
of space to each point of which there corresponds a definite value
of the vector \lep.
~Divergence. Assume that we are given the field of the vector a.
Tne flux of the vector a through the surface f is defined to be
<Da = Jan df = Ja df (XI.6)

I I
where an is the projection of the vector a onto a positive normal

to the surface element df, and df is the vector of the element; its
magnitude equals that of the surface element df, while its direction
coincides with that of a positive normal to the element. The direccion
of the positive normal is determined depending on the circumstantes.
For example, the outward normal is considered to be positive in
calculating the flux through a closed surface.
The name "flux" is due to the fact that for the field of a liquid's
velocity vector, the integral (XI.6) gives the flux of the liquid through
the surface f, i.e. the volume of the liquid flowing through f in unt
time. •
Let us surround the point P of a fteld with the closed surface f
and evaluate the flux <Da through this surface. The ratio of <Da to V
will characterize the properties of the field in the vicinity of the
point P averaged over the volume V confined within f. The smaller
the linear dimensions of the volume, the closer will the average
characteristic be to the true characteristic of the fteld at the point P.
The scalar quantity
.:.
d tva= 1'1m -y=
a i A., an df'.
1'1m y'j' (XI.7)
v-P v-P
I .
is known as the divergence of the vector field at the point P.

APPENDICES 373
The definition (XI. 7) is a most general one not depending on the

choice of the coordinate system. Let us find an expression for the
divergence in terms of the projections of a onto the axes of a Cartesian
coordinate system. We take a
volume in the form of a rectan- z
gular parallelepiped with faces
perpendicular to the coordinate
axes in the vicinity of the point
P (Fig. XI.1). Let us find the
flux of the vector a through the
faces 1 and 2 perpendicular to
the x-axis. An outward normal
to the face 1 coincides in direc-
tion with the x-axis. Hence, for
this face, an = ax 1 (the sub-
script 1 indicates that the value
of ax is taken at a point on the Fig. XI.1.
face 1). An outward normal to
the face 2 is opposite in direction to the x-axis. Therefore, we have
for it an = -ax 2 (the subscript 2 indicates that the value of ax
is taken at a point on the face 2). The total flux through the faces
1 and 2 is
<l>x= ~
(1)
axtdft- J
(2)
ax2df2= ~
f
(axt-ax2) df (XI. B)
where df = dj 1 = d/ 2 (see Fig. XI.1), ax 1 and ax 2 are taken for

points on the faces 1 and 2 with identical y' s and z' s. The integral
on the right is taken over the surface f of any of the faces 1 and 2.
Let us expand ax into a series in the vicinity of the point P:
ax=axp+( ";; )P(x-xp)+( o;yx )P(y-yp)
+( ":~'" )P(Z-Zp)+ex (XI.9)
Here Xp, yp, and Zp are the coordinates of the point P, a:.·P is the
value of ax at the point P, (DaxiBx)p, etc. are the values of the deriv-
atives at the point P, and ex is a quantity of a higher order of small-
ness than the differences (x- xp), (y- yp), and (z- zp), i.e.
a quantity diminishing more rapidly than the linear dimensions of
the parallelepiped.
Assuming in expression (XI.9) that x = x 1 , let us find the values
of ax at points on the face 1, i.e. ax 1 ; assuming that x = x 2 , we
obtain the values of ax 2 • Subtracting these values from one another,
we obtain the following expression for the opposite areas d/1 and d/ 1
~374 APPENDICES
·(the values of y and z are the same for them):

·,:i'
where again e~ is a quantity diminishing at a faster rate than the

linear dimensions of the volume.
UsiJ?.g the value we have found in formula (X 1.8), we obtain
<D:~:= l a;; )P(x1-x2) ~ df+ j e~ df= (a;;)

f f
P (xt-:-x2) !+ \
~
~ df
A glance at Fig. XI.1 reveals that the product (x 1 - x 2 ) f is the
vq!ume V of the parallelepiped. Hence,
'· Cl>x. '(a;; ) P V +e~
where e~ is & quantity of a higher order of smallness than V. Similar
expressions· are also obtained for the fluxes through a pair of faces
perJ?endicular to the axes y and z:
L',,-1': cD~=7-( ~: )PV+~, C!>z=( ~:z )PV+e;

Summating cDx, <f>y, and cDu we obtain the total flux of the vector
1J through the surface of the parallelepiped. By dividing this flux
1 oy V in accordance with the definition (XI. 7) and performing the
limiting process 1 V -+ 0, we arrive at the formula
·:· :•
I
d'
1V8=--
8a:x + -Bay- +-aa~-
i h ~ •
(we have discarded the superfluous subscript P on the derivatives).
Expression (XL 10) we have found for the divergence can be
written as
d'1V8= £.J
""' OXf
Bat (XI. H)
f
ln thi.'3 [orrn, lhe concept of lhe divergence can be extended to vcGtor
.f.ields in an n-dimensional space. The definition (XI. 7) can also be
eXtended to an n-dimensional space. In this case, by an element of
volume we should understand dV* = dx 1 dx 2 dx 8 • • • dxn The
integral must be evaluated over an (n - i)-dimensional hypersurface.
-~n element of the hypersurface perpendicular to the x1t-axis will
equal df"' = dx 1 dx 2 ••• dxk _1 dxk + 1 • • . dxn. In a four-dimensional
~pace, a conventional three-dimensional volume will be the hyper-
surface.
1 In the limiting process, the quantities e"/V shrink to zero.

APPENDICES 375
Expression (Xl.11) can be considered as the sum ol products of the

quantitiE>s Y' 1 = fJ/{)xt and a 1, i.e. as the scalar product of the ver.tors
Y' and a. The divergence can therefore be written as
div a= Y'a (X I.12)
The quantity Y'a exists at every point of the vector fleld of a. Con-
sequently, the divergence forms a scalar field determined in the
same part of space as the field of the vector a.
Let us take the finite volume V confined
within the surface f (Fig. XI.2) in the field
of the vector a. We divide this volume into
,elementary volumes d V. By (XI.7) for the.
flux del> a through the surface of such a volume,
we can write da ~ div a·dV = Y'a·dY.
Let us summate these expressions for all the
elementary volumes. In summating d<f>a, the
fluxes through the faces separating two ad-
jacent volumes mutually nullify each other
(for adjacent volumes, the fluxes differ in their Fig. XI.2 .
.signs because the outward normals n and n'
have opposite directions). Only the fluxes through the portions of the
outer surface .f remain uncompensated, so that the sum gives us the
Oux of the vector a through this surface. The sum on the right in
the limit (when d V _. 0) transforms into an integral over the entire
volume. The approximate equality in the limit will transform into
a strict equality. As a result, we obtain
J
~a df = va dV (XI.'l3)
:t v
The relation we have obtained is called the Ostrogradsky-Gauss theo-
rem. It states that the flux of a vector through a closed surface equals
the integral of the divergence over the volume confined by this surface.
Curl. The circulation of the vector a over the contour r is defined
to be the expression
Ca=~adl=~atdl (XJ.t4)
r! 1r
For example, in a potential field of forces, the circulation of the
vector F equals the work of the forces on the closed path r.
Let us take in the vicinity of the point P the contour r in a plane
passing through P. We shall find the circulation Ca around this
contour. The ratio of Ca to the surface f enclosed by the contour will
characterize the properties of the field in the vicinity of the point P,
averaged over the surface f. The smaller the linear dimensions of the
surface, the closer will the average characteristic be to the true
376 APPENDICES
characteristic of the f1eld at the point P. At the limit, when the

contour shrinks to the point P, the average characteristic will trans-
form into the true one. Therefore, the properties of the vector field
at the point P can be characterized by the quantity
lim ~a = lim
t-o f-0
Tr ~ a dl (XI.15)
The quantity (Xl.15) depends not only on the properties of the

field at the point P, but also on the orientation of the plane in
which the contour is. The orientation of this
plane in space can be set by a normal to the
plane associated with the direction of circum-
vention of the contour r in integration by the
right-hand screw rule (Fig. XI.3). For different
directions of n, the quantity (X 1.15) will have
different values at the same point P, and it is not
: difficult to see that values of the quantity
(X1.15) differing only in their sign will corres-
pond to opposite directions of n. Consequently,
Fig. XI.3. the quantity defined by formula (XI.15) behaves
like the projection of a ve~tor onto the direction
of a normal to the contour r. This vector is known as the curl of the
vector field at the point P and is designated by the symbol curl a.
Hence,
(curl a)n =lim Cfa =lim -1 1 ~a dl (XI.'16)
t-o t-o 'Y
Formula (X1.16) gives a most general definition of the curl that
does not depend on the choice of the coordinate system. Let us find
an expression for the curl in terms of y
the projections of the vector a onto the
axes of a Cartesian coordinate system. f
We shall begin with determination of
the projection of the vector curl a onto
the x-axi':l._ For this purpose, we take in
the vicinity of the point P the contour
r in a plane perpendicular to the x-axis .
(Fig.' XI.4). We choose the direction of
circumvention of the contour so that
it forms a right-handed system with x
the direction of the x-axis. Hence, the Fig. XI.4.
directions of n and the x-axis will coincide, and expression
(XI.16) will give (curl a)x· For the contour we have chosen, we have
~ az dl = ~ a dl = ~ (ay dy + az dz)
r r r
(for all the elements of the contour dx = 0).
APPENDICES 377
The values of ay for points of the contour can be written as1
p (y- yp) + ( az
aa!I ) aa y )
ay = ayp+ ( ay p (z-:-Zp) +ey
= const + ( a:: ) P y+ ( ~:u )P z + ey

where the constant term includes three addends not depending on
y and z, and ey is a quantity of a higher order of smallness than the
linear dimensions of the con tour. Consequently,
~ ay dy = const ~ dy + ( a:: ) P ~:Y dy + ( ~:u )P ~ zdy + ~ ey dy

It is not difficult to see that§ dy = 0. In exactly the same way,
the integral ~ y dy = ; ~ d (y 2 ) is zero. It is simple to see from
Fig. XI.4 that ~ z dy = - f, where f is the area of the contour.
Hence,
(XI.17)
where e~ is a quantity of a higher order of smallness than the area f

of the contour.
Similar calculations for az yield the expression
p az dz = const p dz + ( ~:z ) ~~P y dz + p ( ~:z )~

P z dz + ~) Bz dz
The integrals ~ dz and p z dz = f ~ d (z 2) are zero, and ~ y dz = f.

Therefore,
pazdz= ( ~:z )Pf+e~ (XI.18)
The sum of expressions (XI.17) and (XJ.18) yields ~a dl. Dividing

this sum by f in accordance with the defll)ition (XI.1G) and per-
forming a limiting process 2 , we obtain
oa: aay
(curl a)x = ay-o;-
Having considered the circulation for contours oriented with the
normal n along the y- and z-axes, we can obtain expressions for the
1 The meaning of the quantities in this expression is similar to the meaning-

of the ones in formula (XI.9): the term containing (x - xp) is absent because
x = xp for all the points of the contour.
2 The quantities s' It vanish in a limiting process.
\'378 APPENDICES
(lrojections of the curl onto these axes:

aax
(cur l a )11 =-a;-- aa. aay aax
--a;- , (curla) z =ax- - -
ay.
It is easy to remember the formulas for the projections of a curl
-onto the coordinate axes by giving attention to the fact that in
each of them the subscript on curl a and the letters on the righ l in
the denominators form a cyclic transposition following the scheme
x~y
'/ z
Knowing the projections, it is simple to find the vector itself:
. I· ( aa. aav·) (' aax aaz ) ( aay aax )
cur a= ex --ay -a;- +.e 11 ---;;;---a? +ez a?---ay
(Xl.19)
With a· view to the fact that, for example, aazlay can be written as
V 11az, etc. (V 11 = a/ay is the projection of the vector V onto the
y-axis), formula (XI.19) becomes
ex ey ez
curl a= Tx ay a;
a a
(Xl.20)
IJ
ax ay az
Finally, a comparison with formula (VI.31) allows us to write that
curl a== [Val (XJ.21)
Taking advantage of formulas (VI.29) and (Vl.33) for a vector
.product, we can write the following expressions for a curl and its
k-th .component:
aak
• [Va] = ~ Btkz-a-e"
Xf
(XI.22)
i, k, l m.n
The quantity [Val exists at every point of the vector field of u.
Consequently, the curl forms a vector field determined in the same
part of space in which the field of the vector a is set.
Let us take an arbitrary surface j enclosed by the contour r (the
latter does not necessarily have to be plane and can have any shape)
in the field of the vector a. We divide this surface into small elements
Aj (Fig. X1.5). By (XI.16) for the circulation A.Ca around the bound-
ary of an element A.j, we can write the expression A.Ca ~ [Valn M,
where [Valn is the projection of [Val onto a normal to the given
element A.f related to the direction of circumvention by the right-
hand screw rule.
APPENDICES 379
Let us summate these expressions for all the 6.f's. In the summation
(>f !:lCa, the integrals Ja dl taken along the boundaries between
adjacent areas cancel each other (for adjacent areas these integrals
differ in their sign because they are evaluated in opposite dirPetions)
Only the integrals) a dl for the sections coinciding with thn contour f
enclosing f remain uncompensated. The sum of these integrals gives
the circulation of a around the contour r.
The sum on the right in the limit (when
!:if - 0) transforms into an integral over the sur-
face. The approximate equality in the limit trans-
forms into a strict equality. The result is r
~a dl = J[\i'a) df (XI.23)
p f
The relation we have found is called Stokes's Fig. XI. 5.
theorem. It states that the circulation of the vec-
tor a around the closed contour r equals the flux of the vector [\i'a]
through the surface enclosed by the contour r.
The surface over which the integral on the right-hand side of
formula (XI.23) is evaluated may be any one, it is only important
that it bound on the contour r. The direction of the normal n must
be made to agree with the direction of circumventing the contour
r in integration.
Application of the Operator \7 to the Product of Functions. When
formulas including \7 are being compiled, it is necessary to adhere
to the rules of both vector algebra and differential calculus. Assume,
for instance, that cp and 'IJ> are scalar position functions. Therefore,
\7 (cp'IJ>} = \7"' (Cfl'IJ>) + \7 cp (cp\jl} (XI.24)
(the subscript on \7 indicates which of the functions it is acting on).
The factor on which \7 does not act in the given term can be put
outside the symbol \7 (the operator \7 acts only on quantities follow-
ing it). Formula (XI.24) now becomes \7 (cp,~) = <rY'w'ii + 'IJ>\i''Pcp.
lu this expression, the subscripts on \7 are superfluous, so that we
finally have
(XI.25}
(the right-hand side is read "phi gradient of psi plus psi gradient
of phi"). .
Let us apply \7 to the product cpa. Here there are two possibilities-
a scalar or a vector product of the vectors \7 and cpa. The corre-
sponding results are:
\7 (cpa)= \i''P (cpa)+ Y'a (cpa)= a\i'cp +
cp\i'a (XI.26)
380 APPENDICES
("a, gradient of phi plus phi divergence of a") and

[V, (cpa)] = [V Ill' (cpa)] + [Va, (cpa) I = [(Vcp), a] + cp [Val (XI.27)
Now let us apply V to the product [abl, first obtaining a scalar
product of the vectors: V [abl = Va [abl +
Vb [ab]. We perform the
cyclic transposition (VI.3) in each of the terms:
V [ab] = b [Vaal+ a [bVbl = b [Vaal- a [Vbb]
(we have exchanged the places of band Vb in the second term to have
the vector b following the operator Vb that acts on it; the sign in
the vector product has changed accordingly). Omitting the sub-
scripts needed no longer, we arrive at the formula
V [abl = b [Val - a [Vbl (XI.28)
("b curl of a minus a curl of b").
The vector product of [ab] by V yields [V, [abll = lVa, [abll +
+ [V b• [ab]]. Let us expand each of the terms using formula (VI.5):
[V, [ab)] =a (Vab)- b (Vaa) +a (Vbb)- b (Vba). Arranging the
factors so that we may suppress the subsc'ripts on V, we obtain
lv, [abll = (bV) a - (aV) b + a (Vb) - b (Va) (XI.29)
The expressions (aV) and (bV) are scalar differential operators. For
example,
(XI.30)
These operators can be applied to both scalar and vector functions.

When applied to the scalar cp, the operator (XI.30) gives ~
(aV) cp = 2j a 1 :~ =a (Vcp) (XI.31)

{
The action of the operator (aV) on the vector b yields the expression
(aV)·b = 2j a 1
i
:!, = 2j a
i
1 0 ~ 1 ( ~ e"b") = ~ eh ~ a1
h k i
::: (XI.32)
Let us apply the operator (XI .30) to the product of the scalar
function cp and the vector function b:
(aV) (cpb) = (aVq:) (cpb) + (aVb) (cpb) = b (aV) cp +
cp (aV) b
= b (a·Vcp) + cp (aV) b (XI.33)
It is useful to know the value of the expression (aV) r, where r
is a position vector, and a is an arbitrary vector. Substituting r
for b in (XI.32) and taking into account that ox11/{}x 1 = 6 11" we
APPENDICES 381
Qhtain
(XI.34)
We have obtained formulas (X1.25)-(X1.29) quite easily. The

finding of the gradient of the scalar product of the two vectors
V (ab) is more ilwolved beeausc it is not clear what shoHlrl be unrler-
stood, for example, by the expression Va (ab). It cannol be inter-
preted as (~aa) b because the operations of multiplying a and b
and of applying ~a cannot be interchanged. This difficulty can be
surmounted by employing auxiliary relations following from formula
(VI.5): [a, [~bll = ~b (ab) - b (~ba) = ~b (ab) - (a~) b, whence
'Yb (ab) = [a, [~bll + (a~) b (XI.35)
Having written [b, [~all in the same way, we arrive at the relation
~a (ab) = [b, [~all+ (b~) a (XI.36)
Substitution of relations (XI.35) and (XI.36) into the formula
~ (ab) = ~a (ab) + "'h (ab)
yields the following expression for the gradient of the scalar product
Qf the vectors a and b:
~ (ab) = fa, [~bll + [b, [~a]] + (a~) b + (b~) a (XI.37)
Repeated Application of the Operator ~. The action of the del
Qperator on scalar or ver.tor functions results in new vector or scalar
functions which the operator ~ can be applied to, in turn.
The gradient of the function cp is a vector. Consequently, both the
divergence and the curl operations can he applied to it. Let us
calculate the divergence of a gradient. By formulas (X 1.2) and
(XI.11), we have
~ (V'cp) = ~ 2 cp =""_a_ (~cp) 1 = ""_a_~= ~ a2~ = ~cp
L.J axi L.J axi a.ri oxr
i i i
Hence
V' (~cp) = ~cp (XI.38)
where d is the Laplacian opemlor:
i)2 i)2 a2
~= iJx2 + iJy2 + az2 (XJ .39)
Our calculations show that

'\72 = !!. (Xl.40)
It must be borne in mind, however, that such a relation between the
operators V' and !!. occurs only in Cartesian coordinCttes. In other
coordinate systE>ms, for instance in a cylindrical or a spherical one.
382 APPFNDTCES
relation (XI.40) is not observed. The most general definition of the

operator 11 that holds in any coordinate system is that following
from relation (XI.38), which can be written as
11<p = div grad <p (XJ.41}
Let us find curl' grad <p. By (XI.22), we have
"" etkl ax,
['\7, '\?<p]= L.J a ( ax
a<p \} ez (XI.42)
11
i, k, l
Since fJ 2 <plfJx 1 fJxk = fJ 2 <p/fJxk fJx 1 , expression (XI.42) is zero, so that

curl grad <p = 0 (XI.43~
This is what should be expected because ['\7, '\?<p] = [Wl <p, and the
vector product of a vector by itself is zero.
Let ns calculate the divergence of a curl. By formulas (X 1.11) and
(Xl.22), we obtain
'\7 ['\78] = ~ a!11 ( 2j ekmn ;;: ·) = ~ Bkmn ox~2 ~~m

k m,n h,m,n
Seeing that fJ 2a11 18:ck fJxm = 82 a,J8xm fJxk, the last expression equals~ ·
zero. Consequently,
div curl a = 0 ' .. (XI.44)
We could have arrived at this result direr,tly by taking into account
that the scalar triple product of vectors (which is what '\7 -['ya] is)
equals the volume of a parallelepiped constructed on the vectors
being multiplied. Therefore, when two of the three factors coincide.,
sueh a product is zero.
To calculate curl curl a, we shall proceed from formulas (XI.22):
['\7, [V'a Jl = 2j ethl iJ~t ( 2j C:kmn ::: ) ei

i. h., I '"· n
Let. us perform a cyclic transposition of the subscripts on e so that
th(~subscript k is the last on both c:'s, and let us use relat.ion (VI.16).
The result is
APPENDICES 38:5
We summate over the subscripts m and n. This eauses the expression.

we have obtained to become
""' a2a, ""' a2az
LJ e, ax 1ax1 - LJ ez ax~
i, l i, I 1
2j ez a~, { ~ :;~ )- 2j a: ! { 2j a ei) =

2
1 V' (Y'a)- fla
/ j i ' /
We have thus arrived at the formula
l.v, [Vall =v (Va) - ~a (XI.45}
or
curl curl a = grad div a - ~a (XI.46)
It i1" a simple matter to see that relation (XI.45) can be obtained
if we expand [V, [yoa]] by formula (VI.5), treating V' as a conventional
vee tor.
Examination of (XJ.46) shows that
grad div a =' curl curl a + ~a (XI.47)
The divergence is a scalar. Therefore, no operation except that of
finding the gradient can be applied to it.
~ome Formulas of Vector Analysis. Let us fmd the divergence and
curl of the position vector r, and also the gradient of the magnitude
of r Taking into account that r = (~ xD 112 , we obtain hy (XJ.2)
~e ar 2j e1 -x =
1 r (Xl.48)
Vr=
' .axi- = r
-=
r
e
r
'i i
where er is the unit vector of the position vector r.

liy (XI.11), we have
Vr = "
LJ
art = "LJ fJxt
iJxt
ax 1 = 3 (XI.49)
; I
(wP have taken advantage of the fact that r 1 = x 1). In an n-dimen-

sional spaee, vr = n.
1t is not difficult to see that the curl of the position vector is zero:
[yor] = 0 (XI.50)·
For this end, we can use formula (XI.22), substituting .Lk for a11
in it We obtain
[V'r] = ""'
LJ CiJd axt ez = "LJ Etkl6il,el
ax"
i, h. / i, "· /
Tbi~ expression is zero because at identical values of the subscripts
384 APPENDICES
i and k, the factor Ethl vanishes, and at different values of them,

the factor 6 1k. vanishes.
Relation (XI.50) also follows from the fact that the circulation
Qf the vector r around any contour is zero. Indeed, according to
(XI.1.4)
Cr=~rdl=~(xdx+ydy+zdz)= ~ ~d(x2+y2+z2)
There is a total differential in the integrand, therefore the latter
is zero.
Now let us find the gradient of a function of r, i.e. Y'<r (r). The
partial derivative of <p (r) with respect to xk has the form 1 o<ploxk =
= acp/i:Jr· or/ox, = o<plor· x,!r. Consequently,
Y' Vr= oq> .!..=~er (XI.51)
or or T OT
Assume that we have a function of the distance between two points:

<p (I r-r' I)= <r ( 1/~ (x 1 -xiF) = <r(R), where R = V~ (x 1 -xi)2
1 1
The operation of finding the gradient can be applied to this function

in two ways-we can perform differentiation either with respect to
the coordinates x 1 or to the coordinates xi. To distinguish between
these two gradients, in the first case we shall use the symbol V,
and in the second, v'. The components of the gradient in both cases
have the form
It is evident that the derivatives oR/axe and oR!oxi differ only in

their sign. We thus conclude that
Vrp = -V'cp (XI.52)
Let us calculate the curl of the unit vector er = rlr. Representing
r/r as the product of cp == 1/r and a = r, we shall apply formula
(XI.27). As a result, we find that [V, r/r] = [V (1/r), r) + (1/r) [yrr].
By (XI.51), we have V (1/r) = - (1/r2 ) r/r. Therefore, the first term
is zero. The second term is zero in accordance with (XI.50). Hence,
[V, r/rl = [V, erl = 0 (XI.53)
1 We have .written oq>/fJr instead of d<p/dr because 'it· may happen. that 1p,
in addition to r, also depends, say, on t.
APPENDICES 385
Now let us find the curl of the function qJ (r) en where qJ (r) is an
arbitrary function of r. We again apply formulas (XI.27) and (XI.51):
[V, qJ (r) erl = [V'qJ (r), erl + qJ (r) [V, erl
= [(oq:>/or) en erl + qJ (r) [V, erl = 0
[sec (Xl.53)]. Hence
[V, cp (r) erl = 0 (XI. 54)
Assume that we are given the vector function a (~) = ~ a 11 (s) e 11
It
of the scalar quantity £ that, in turn, is a function of the coordinates
x 1, that is, £ = £ (xi)· Let us find the curl and divergence of this
function. According to the rules for the differentiation of a composite
function,
(XI. 55)
By (XI.22)
The quantity oa 1Jo£ is the k-th component of the vector oa/86, and
o£/oxiis the i-th component of the gradient of the function £. There-
fore,
[Val= ~
i, k, l
eil.z (V's)r ( :~ ) 11 ez = [ V's. :~ J
[we have used formula (VI.29)]. Hence, if a = a(£), where £ =
= £ (x1 , x 2 , x 3 ), we have
[V'a] = [_ \76, :~ J (X 1.56)
To fmd the divergence of the function a(£), we proceed from
formula (XI.11). With a view to (XI.55), \Ve obtain
va = "" oak = ""
LJ axh LJ as
a~" ~
axk
= ")' ( a~ ) (\7£) = !!.. . \76 (XI.5 7)
k-J as . k II a~
h It It
Assuming in formulas (XI.5li) and (X1.57) that = r (where r s
is the magnitude of the position vector), we fmd that
[\7, a(r)J=[vr. :;J=[e, :~J (XI. 58)

. aa aa r
\78 (r) = Tr vr =a;: r (Xl.59)
[see (XI.48)1.
We shall prove the formula
Jv [V'a] dV = ~ {na] df = ~
f f
[df, a] (XI.60)
25-0!8
386 APPENDICES
where a is a vector function, Vis an arbitrary volume, f is the surface

confming it, and n is an outward normal to the surface element df.
For this purpose, we fmd the scalar product of the vector [Val and
an arbitrary constant vector b. Inspection of the formula V [abl =
= b [Val- a [Vbl [see (XI.28)1 reveals that b [Val = V [abl +
+ a [Vbl = V [abl ([Vbl = 0 because b = const). Now let us in-
tegrate the relation obtained over a certain volume V and apply the
Ostrogradsky-Gauss theorem to the right-hand side:
Jh [Va] dV = JV [ahl dV = ~ [abl n df

V V I
Performing a cyclic transposition of the vectors being multiplied
in the last integral [see (VI.3)], we obtain ~ b [Val dV = ~ b [na] df.
V I
vVe put the constant vector b outside the integral:
b Jv [Va] dV = h ~ [na) df
I
This relation must be observed with an arbitrary choice of the vec-
tor h. We can therefore cancel b. As a result, we arrive at formula
(XI.60).
Integral Determination of the Operator V. Let us consider the
integral
~ cp df (XI. 61)
over the closed surface f, where cp is an arbitrary scalar functi9n,
df = n df is an elementary vector of the area df. Let us take the
surface of an infinitely small rectangular parallelepiped with the
sides dx, dy, dz as the surface over which the integral is being eval-
uated. The integral (XI.61) can therefore he written as the sum of
six integrals each of which is evaluated over one of the faces of the
parallelepiped. Owing to the smallness of the faces, the value of (p
within the limits of each face may be considered as constant. Con-
sequently,
~cp df=ex[cp(x+dx, y, z)-cp(x, y, z)]dydz
+ey[cp(x, y+dy, z)-cp(x, y, z)ldxdz
+ez[cp(x, y, z+dz)-cp(x, y, z)]dxdy
= (ex;; +ev :~ +ez ~~) dxdydz=VcpdV

whence
V'cp= lim ~ ~ cp df (XI.62)
v-o
APPENDICES 387
In accordance with formula (XI.62), the operator '\7 can be deter-

mined as follows:
'\l = lim ...!.. f., df (XI.63)
v ~o V '-ji
Let us take a volume V and divide it into elementary volumes~ V.
By formula (XI.62), the approximate equality vcp·~V;::; ~cp df is
observed for each volume.
Now let us summate this equation over all the elementary volumes.
In summation, tho right-hand sides corresponding to the boundaries
between adjacent volumes mutually eliminate one another. Only
the terms corresponding to the outer surface remain uncompensated.
Therefore in passing to the limit at which all the ~ V' s shrink to
zero, we arrive at the relation
) '\lcp dV = ) cp df (XI. 64)
v f
According to (XI.64), the integral of the scalar function cp over the
closed surface f can be transformed into an integral over the volume
confined in it by replacing the surface element df with the operator
'\l dV. Here the components df experience the transformation
f)
df 1 -+ dV -0 (XI. G5)
Xi
Naturally, the reverse transformation is also possible.

Matters are similar with the integral of a vector function. Indeed,
according to the Ostrogradsky-Gauss theorem [see (XI.13)l, we
have .~ '\Ia dV = ~a df, which also agrees with (XI.65).
v f
In general, the transformation (XI.65) is always allowable regard-
less of the specific kind of expression in the integrands. An example
of such a transformation is relation (XI.60).
Curvilinear Coordinates. IL is sometimes convenient to determine
the position of a point not by means of the Cartesian coordinall~s
x 11 x 2 , x 3 , but by using three other numbers q1 , q2 , q3 corresponding
better to the nature of the problem being considered. These numbers
are known as the curvilinear coordinates of a point.
Imposing (if the need appears) restrictions on the region of chang-
ing of curvilinear coordinates, we can achieve a one-to-one corre-
spondence of the variables x 1 and q1• Hence, q1 = qi (x11 x 2 , x 3), and
xi = Xt (q1 , q 2 , q 3 ) (i = 1, 2, 3).
Surfaces described by the equation q1 (x 1 , x 2 , x 3) = const are
said to be coordinate surfaces. The lines of intersection of two coor-
dinate surfaces are called coordinate lines. Only one coordinate varies
along a coordinate line; the other two remain unchanged.
25*
388 APPENDICES
An example of curvilinear coordinates is the spherical system

r, {}, q>. In this case, (1) spheres, r = const, (2) half-angle cones,
oO- = const, and (3) half-planes, q> = const, are the coordinate sur-
faces. The coordinate lines are (1) radii-the lines r, (2) meridians-
the lines -6-, and (3) parallels-the lines q>.
If the coordinate lines at every point are mutually perpendicular,
the curvilinear coordinates are said to be orthogonal. We shall limit
ourselves only to a treatment of orthogonal coordinates. They in-
elude spherical and cylindrical coordinates.
Let us introduce for every point P the unit vectors e:, e~, e~
directed along tangents to the coordinate lines at the given point
towards an increase in the relevant variables q1• Owing to orthogo-
nality, the following relations hold for the unit vectors ef:
ete: = 6t~t (XI.66)
We determine the derivative of the position vector r = r (q 1 , q2 , q3}
with respect to the coordinate q 1• The remaining two coordinates do
not change in differentiation. Consequently, when the coordinate q1
is given the increment 6q~o the tip of the vector r travels along the
coordinate line q1• Therefore, the vector fJr/fJq 1 is directed along
a tangent to the coordinate line qtt i.e. is collinear with et. Denoting
the magnitude of the vector fJr/fJq 1 by the symbol Hi> we can write
that
(XI.67)
Representing r in the form of Ljxkek (here ek are the unit vectors

of the Cartesian coordinate axes), we can write that
or = ~ -;;--
-,- OXk
ek = H ,ei• (i=1, 2, 3)
vqi vqt
k
Squaring of this relation yields
IH=~(~~:t (i=1, 2, 3) (Xl.68)

h.
The quantities H1 are known as the Lame coefficients. Their values

can be found by formula (XI.68) if we know the form of the functions
xk = xk (q 1). For example, for a spherical coordinate system, x1 =
= r sin-& cos <p, x 2 = r sin\} sin <p, x 3 = r cos ofr. Hence
~:: = 8;: =sin{} cos <p, :;; = a;; = sin-& sin q>, :;1 = 8; : = cos-&
so that by (XI.68)
Hi = (sin-& cos <p) 2 + (sin\} sin <p} 2 + (cos -&) 2 = 1
APPENDICES 389
Similar calculations for H 2 and Il 3 yield

H 1 = 1, H 2 = r, H 3 = r sin{)- (XJ.69)
Let us find the square of the distance ds between two points. By
(XI.67), we have
dr= ~
i
:;i dq = 1 ~ H 1eT dq 1
i
(XI. 70)
Consequently,
ds 2 = I dr 12 = 2Ji Hf dql (XI.71)
For a spherical coordinate system, this formula yields

ds2 = dr 2 +
r 2 d-6-2 +
r 2 sin 2 {)- dq> 2 (XI. 72)
Let us draw coordinate surfaces through the tail and the tip of the
vector dr. The result is an infinitely small parallelepiped whose edges,
by (XI. 70), are
(XI.73)
The faces of this parallelepiped have the areas
d/1 = H2Hadq2dqa, d/2 = HaHldqadqt, dfa = H1H2dq1dq2
(X I. 74)
The volume of the parallelepiped is
dV = H 1H 2H 3 dq 1 dq 2 dq 8 (XI.75)
Instead of calculation by formulas (XI.68), the Lame coefficients
can be found with the aid of expressions (XI.73). For instance, for
spherical coordinates (Fig. XI.6), an elementary parallelepiped has
the edges dl 1 = dr, dl 2 = r dl't, and dl 3 = r sin{} dcr, whence we
directly obtain the values (XI.69) for the Lame coefficients.
For a cylindrical coordinate system (Fig. XI. 7), the edges of an
elementary parallelepiped are dl 1 = dp, dl 2 = p dq>, and dl 3 = dz,
whence
(XI.76)
Let us find the expression for the gradient in curvilinear coordi-
nates. By formula (XI.5), the projection of the gradient of the function
'IJ' onto the direction of the unit vector et
is &¢1ol 1• With a view to
(XI.73), we obtain
0'\jl 1 8\jJ
Tzi=lli 8q;
The gradient itself is determined by the formula

"Q &\jJ * "Q 1 8\jJ • (XI. 77)
V''i' = £.J azi e; = £.J ""lTj' aqj' ei
i i
390 APPENDICES
Consequently, in a spherical coordinate system

a¢ 1 a¢ 1 a¢
V1jl = ar er + 7 a{} e, + r sin {} alP e<P (XI. 78)
and in a cylindrical system

&¢ 1 ii¢ ii¢
:, V1Jl=apeP+paiPe'~~+Tze, (XI. 79)
To determine the divergence in curvilinear coordinates, we shall

proceed from the definition (XI. 7) according to which
va = lim
v-o
+~ 'j
a dl = lim
v-o
~ (XI. SO)
Let us take as V the volume of an infinitely small parallelepiped

including the point P for which the divergence is being calculated.
Fig. XI.6. Fig. XI.7.
Heasoning in the same way as when obtaining formula (XI.11), let

us write an expression for the flux through the faces perpendicular
to the unit vector e~:
Cf>1 = a~M~ - a~ M~ = (alH 2H a)' t!.q2t!.qa - (alH 2H a)" t!.q2t!.qa
[see (XI.74)]. Here a 1 is the projection of the vector a onto the direc-
tion of e~. One prime indicates the values of quantities relating to
one of two opposite faces, and two primes indicate the values relating
to the other face.
Expanding a 1H 2 H 3 into a series in the vicinity of the point P,
we obtain
APPENDICES 391
where V is the volume of the parallelepiped lsee (XI.75)], and e1

is a quantity of a higher order of smallness than V. The fluxes through
two other pairs of faces are evaluated in a similar way:
<l> _ 1 a (a 2 H 8 H 1) V+ <l> _ 1 a (a 3 H 1 H 2) V +
2 - IltHtHa oqt e2, 3- HtH2Ha oqs e3
Substituting <l> = <1> 1 + <1> 2 + <1> 3 into (XI.SO) and performing

a limiting process, we arrive at the formula
Va= 1 {o(a1 J/ 2Ha)+o(a 2 H 3 H 1 )+o(aaH 1 H 2)} (XI.Si)
H 1 H 2 H3 oq 1 oq 2 oq 3
Taking (XI.69) into account, we obtain an expression for the

divergence in spherical coordinates:
V _ 1 a (r2ar) + 1 a (sin -6-a~) + 1 oa~p (XI 82)
a - r2 OT T Sin -6- o-6- T sin -6- aql •
In cylindrical coordinates
Va =..!_a (pap)+..!_ oa~p + oaz (XI.83)
p ap p ac:p az
By (XI.16), the projection of a curl onto the direction of the unit
vector ef is determined by the expres-
sion
[Va]f = lim -11 ~a dl = lim c1. (XI.84)
1
11-o t • 11-o '
where I 1 is a surface element perpen-

dicular to et, and. C1 is the circulation
-of the vector a around the contour confin-
ing this element.
To evaluate the projection of [\7a)
onto the unit vector e;, let us take 11
as an infmilely small curvilinear
rectangle with the sides !11 2 = H 2 Aq 2 Fig. XI.S.
and Al 3 = H 3 Aq 3 • The circulation
of a around the contour of this rectangle can be represented in the
form of four integrals like \ a dl. Two of them are evaluated over
the opposite sides Al~ and Llt;, and two over the opposite sides Al;
and At; (Fig. Xl.8). As a result, with a view to (XI. 73), we obtain
(XI.85)
892 APPENDICES
Expanding a 3 H 3 and a 2 H 2 into a series in the vicinity of the point

P for which the curl is being calculated, we can write that
aa"H"a-aa'H'a= a (a3H
,· s) LJ.qz
A + Ea, ,
a~n~-a~
TT" 'H's= a (a2H
" 2) LJ.q3
A + Ez
vqa vqa
Consequently, expression (XI.85) can be written as
C _a (asH
aq
8 ) ,1 ,1
q'l. qa -
a(aaq2 H 2) ,1 qa ,1 q2 + s
1- 2 3
= _1_
H 2H 3
{a (asHs)-
aq
a(a2Ha)}
aq
It+ e
2 3
where / 1 is the area of the rectangle [see (XI. 74)1 and e is a quantity
of a higher order of smallness than / 1 •
Introducing the value of C1 we have found into (XI.84) and per-
forming a limiting process, we arrive at the formula
(V'a] 1 = _1_
HaHa
{a (aaqaH _a (aaqsH
3 3) 2 2)}
Similar formulas are obtained for the two other projections of the
vector [V'al. All three formulas can be combined into a single one:
1 { a (azHz) a (~R.qHz R.) } {&" = 1' 2, 3) XI • 86 )

(V'alt = HR.Hz aqk u (
(the subscripts i, k, l form a cyclic transposition of the sequence 1,

2, 3). Finally, let us find an expression for the Laplacian operator
in curvilinear coordinates. By (XI.41), we have .1'\jl = V' (V''Ijl) =
= div grad 'ljl.
Using formulas (XI.81) and (XI. 77), we obtain
.1'1J = V' (V''Ijl) = {_!_
aq
1
H 1 H 2H 3
( H2Ha
H
~)
aq 1 \ 1 1
+-a (
aq 2
H 8H 1
H2
~)+-a ( H H 2 ~)}
aq aq
2 H aq 3
1
3 3
(XI.87)
Introducing the values (XI.69), we obtain an expression for the

Laplacian operator in spherical coordinates
,1 'ljJ = :a !( r2 :; ) + r2 si1n2 ~ a~ (sin ~ :: )

1 a2 '(l
+r 2 sin 2 'fr at:pa
(XI.88)
In cylindrical coordinates
1 a ( a'~') 1 aa'l'
.1'\jl=-p-ap P Tp +p2 acpa + aa"¢
aza (XI.89)
APPENDICES
XII. Four-Dimensional Vectors and Tensors in Pseudo-Euclidean Space
In Appendices VI and X, we considered vectors and tensors in

Euclidean space in which the square of the position vector is deter-
n
X~+ x: + ... +X~= t=1
~ xf (XII.1)
If we introduce the tensor

... 0)
••• 0
(XII.2)
1
expression (XII.1) can be written as
(XI1.3)
The tensor g 1k is called a metric tensor.

The equations of the special theory of relativity and electro-
dynamics acquire an especially siniple and clear nature if they are·
written as relations between vectors and tensors in four-dimensional
space whose metric is determined by the tensor
1 0 0 0)
v 0 -1 0 0
(g~t ) ==(gil..,) = ( 0 0 -1 0 (XII.4}
0 0 0 -1
For reasons that will be revealed later on, we shall use the follow-
ing notation. We shall write the indices on the components of a four-
position vector not as subscripts, but as superscripts. \V e shall
write the Kronecker symbol as 6~, i.e. with one subscript and one·
superscript. We shall use superscripts or subscripts on the compo-
nents of the metric tensor depending on the circumstances.
We shall adhere to the following rules in writing formulas:
1. In each pair of dummy indices (i.e. indices over which sum-
mation is to be performed), one will be a superscript and the other·
a subscript.
2. Free indices (i.e. those over which summation is not performed).
will be placed in the same position-either as a superscript or a sub-
script-on both sides of an equation.
In the coefficients of the linear transformation of vector and tensor
components, we shall also make one index a superscript and the
;394 APPENDICES
other a subscript 1 • Consequently, unlike the transformation formulas

.a( = 2J
k
a;kak which we had to do with in Appendices VI and X,
we shall write similar formulas as follows: a'IJ. = 2Jv a~av.

Recall that in accordance with the condition we have adopted,
whenever the symbol 2J is absent, summation even over paired
indices is not performed.
A space whose properties are determined by the tensor (XII.4)
is called a pseudo-Euclidean one. We shall distinguish the coordina-
tes, and also vector and tensor components in this four-space with the
aid of the Greek indices ll• v, etc., that can take on the four values
0, 1, 2, and 3. We shall use the Roman indices i, k, ... (running
through the values 1, 2, and 3) on coordinates and vector compo-
nents in conventional three-dimensional space (in space with a
Euclidean metric). We shall sometimes use Roman indices on the
.components of four-vectors and four-tensors to underline that we
have in mind non-zero values of the indices.
Having determined the square of a position vector in the four-
space being considered by means of an expression similar to (XII.3)
.and taking into account (XII.4), we obtain
3
~ gJJ.vxJJ.xv = (.rO)z _ (x1)2 _ (x2)2 _ (x3)2 (XII.S)
IJ., \'=0
In the theory of relativity, the product of the timet and the speed
-of light in a vacuum is taken as the coordinate x 0 , and the coordinates
in conventional three-dimensional space are taken as the remaining
-coordinates xi. Hence,
x 0 = ct, x1 = x, x 2 = y, x 3 = z (XII'.6)
Having this in view, we can write expression (XII.S) as
3 3
~ gllvxl1xV=(x0)2- ~ (xi)2=c2t2-r2 (XII. 7)
• j.t, \'=0 i=!
Formulas for the Transformation of Coordinates. In a transition

to another coordinate system, the components of a four-position
vector are transformed according to the linear law:
3
x'!J.= ~ a~xv (XII.8)
v=O
1 We shall see in the following that one index is a superscript and the other
one is a subscript on mixed components of tensors. It must be remembered,
llowever, that linear transformation coefficients do not have the properties of
tensor components.
APPENDICES 395
The inverse transformation is performed by the formula

3
xll = )1 allx'" (XII.9)
v";;;o "
where a~ are the coefficients of the inverse transformation.
Owing to the invariance of the square of the four-position vector,
3 3
the condition must be observed that ~ g~-'"xllx" = ~ g~-'"x'llx'".
llo V=O f.t, v=O
Substituting for x'll and x'" in this expression their values from
(XII.8), we obtain
3 3 3 3
"\-.,
~
g vx llV
x =
~
L "\1 llP"\1 VO
gl!" Ll aPx Ll aax
11
ll• v=O ll· v=O p=O cr=O
Let us exchange on the right the places of the dummy indices 1.1.
3
and p, and also of v and a. The result is ~ gt.tvXIlX" =
ll. V=O
3 3
- ~ xllx" ~ g paa~a~.
1-1. v-o p, cr=O
This gives us the conditions which the coefficients of the linear
transformation (XII.8) must satisfy:
3
~ gpaO:~a~ =gil"
p, cr=O
Having in view that g per differ from zero only when p = a [see
(XII.4)], this relation can be simplified as follows:
(XII.10)
For example, for 1.1. = v = 0, we obtain by this formula

(a:) 2 - (a~)2- (o:~) 2 - (a:) 2 = 1 (XII.11)
and for 1.1. = 1 and v = 2:
(XII.12)
We must note that if we took the tensor (XII.2) as gpo• we could
reduce formula (XII.10) to the form ~ o:~o:~ = 61-1"' which coin-
P
cides with formula (VI.40).
It is obvious that relations similar to (XII.10) also hold for the
coefficients a~.
Now let us find the relation between the coefficients of the direct
and inverse transformations. Remember that in Euclidean space,
this relation is aik = a 11 i [compare formulas (VI.37) and (VI.38)l.
396 APPENDICES
Let us consecutively apply the transformations (XII.9) and

(XII.8). The result is
3 3 3 3 3
xll =
p=O p
2.: allp v=O
2J ctllx'P = p=O 2J aex" = \1=0
2J X" p=O
2J a~-tae
p
The component x~-t can be written as LJ <'l~x". It thus follows that

3 "
" " allctP=6v
.!...J '- ll (p., v=O, 1, 2, 3) (XII.13)
p=O p v
It is not difficult to see that the system of equations (XII.13)

will be satisfied if we assume that
0
ct·= _,.,,.--:-
,.,, _ _ ,.,,0 - -,.,, L( ; k= 1, 2, 3)
: " _,.,,..
• ""o• ""o- ""•• ""11.- ""i .,
(XII.14)
Indeed, let p. = v = 0. Equation (XII.13) therefore becomes
3 3
2J CX"ga~ = 1 or, with account of (XII.14), agag - 2J a~a~
i=O
= 1
P=O
which agrees with (XII.11).
Assume that p. = 1 and v = 2. Equation (XII.13) now becomes
3 3
2J a~af =
P=O
0 or, with account of (XII.14), -a~a~ + 2j a:a; =
i=1
0
which agrees with (XII.12). We can convince ourselves similarly
that the remaining 14 equations contained in (XII.13) are satisfied
by the solution (XII.14).
We must note that relations (XII.14) can be written as a single
relation
(XII.15)
Indeed, when p. = v = 0, we have g~-t~-t = g""' therefore ag = a~;
when p. = 0 and v = i =1= 0, as well as when p. = i =I= 0 and v = 0,
we have gllll = -gvv• owing to which a~ = -aA and a.~ = -a~;
finally, when p. = i =I= 0 and v = k =I= 0, both factors (gllll and gvv)
equal -1, so that a~ = a~. As a result, we have arrived at relations
(XII.14).
The theory of relativity usually deals with transformations in
which the coordinates x 2 = y and x 3 = z remain unchanged (x' 2 =
= x 2 and x' 3 = r). The matrix of the transformation coefficients
in this case can evidently be written as follows:
(XII.16)
APPENDICES 397
The matrix of the inverse transformation coefficients has a similar

appearance.
Let us designate by p a parameter that can be used to characterize
the difference between the systems [(and IC. It can be, for example,
the angle of rotation of one system relative to the other. In the theory
of relativity, the role of the parameter p is played by the relative
velocity of the systems K and K'. It is evident that as a result of a
limiting process in which p shrinks to zero, both systems coincide,
so that the matrix (XII.16) is
[cx~Jp ... Q = [H H]
0 0 0 1
Hence,
lim ex~ = lim ex~ = 1, lim ex~ = lim ex~ = 0 (XII.17)
p ... o p ... o p ... o p ... o
The four coefficients ex~ differing from zero and unity are not inde-
pendent. Having written relation (XII.10) for 1-1. = v = 0, we frnd
that
(XII.18)
A similar relation also holds for the coefficients of the inverse
transformation: (a~) 2 - (~~) 2 = 1, whence with a view to (XII.14)
we obtain the equation
(cx~) 21 - (cx~) 2 = (XII.19)
Now let us write condition (XII.10) for 1-1. = 0 and v = 1:
cx~cx~ - cx~cx~ = 0 (X I I. 20)
A comparison of Eqs. (XII.18) and (XII.19) shows that (cx~) 2 =
= (cx~) 2 , i.e. ex~ = ±ex~. A comparison of this condition with
Eq. (XII.20) gives two possible alternatives of the relations bet,veen
the coefficients:
( 1) a 00 = at1 if cx 01 = at0' and (2) cx 00 = -at1 if aot = -ext0
The first alternative agrees with the property of the coefficients
expressed by formula (XII.17). Consequently, we must assume that
a~ = ex~, and a~ = ex~. Introducing the notation ex~ = a~ = cx 0
and a~ = ex~ = a 1 , we can write the matrix of the transformation
coefficients as follows:
[all]= [ cxi CXo 0 0

cx 0 a1 00] (XII.21)
v 0 0 1 0
0 0 0 1
398 APPENDICES
This matrix contains only one independent coefficient (whose value

is determined by the specific form of the transformation) because
cx 0 = cxg and a 1 =a~ are related by Eq. (XII.19):
ct~ - ct~ = 1 (XII.22)
By (XII.14}, ~ = ct~ = ct 0 , and ;:~ = -ct~ = -ct1. Therefore,
the matrix of the inverse transformation is
(ctlL) = [ -
cx 0 -ct 1
cx1 CX00 0
00] (XII.23)
'V 0 0 1 0
0 0 0 1
Contravariant and Covariant Vectors. Substitution into (XII.8)
of the values of ex~ from (XII.21) leads to formulas for the trans-
formation of the components of a four-position vector:
x' 0 = a 0x 0 + a 1xl, x' 1 = a 1 x 0 + cx 0x\ x' 2 = x 2 , x' 3 = x 3 (XII.24)
The set of the four quantities a0 , a\ a 2 , and a 3 that transform in
a transition from one coordinate system to another according to
the same rules as the components of a four-position vector, i.e.
according to the formulas
3
a'IL = ~ a~av (XII.25)
v=O
is said to be a four-dimensional vector (a four-vector). The coef-
ficients ctlLv in (XII.25) have the same values as in (XII.8).
The component a 0 corresponds to the component x 0 of a four-
position vector that in the theory of relativity is taken equal to ct
[see (XII.6)]. In this connection, the component a 0 is called a time
one. The components a 1, a 2, a 3 correspond to the components x 1 = x,
x 2 = y, x 3 = z, and are therefore called spatial components.
Substitution into (XII.25) of the values of cxlLv from (XII.21) leads
to the foll~wing formulas:
a' 0 = a 0 a0 + a 1a 1, a' 1 = cx 1a 0 +
cx 0a 1, a' 2 = a2, a' 3 = a 3 (XII.26)
The square of a four-vector is determined by analogy with the
square of a four-position vector [see (XII.5)1:
3
~ gJ.LJ.LalLalL = (aD)Z- (a1) 2 - (a2) 2 - (a3)2 (XII.27)
J.L=O
We must note that the square of a four-vector can be either positive
or negative; particularly, it may be zero.
In writing vector formulas for a pseudo-Euclidean space, we become
confronted with a major inconvenience because the square of a four-
vector is determined by expression (XII.27) that cannot be written
APPENDICES 399
in a compact form as ~ (all) 2. This inconvenience is eliminated by

introducing two kinds of four-vectors. They are distinguished by
using a superscript on vectors of one kind and a subscript on vectors
of the other kind. Tho first kind of vectors are called contravariant,
and the second, covariant. Hence, by using a superscript on th~
components of a four-position vector, we have related this vector
to the category of contravariant ones.
A covariant vector all corresponds to each contravariant vector aJJ.·
(and vice versa), and it is assumed that
(XII.28}
[see (XII.4)l, i.e.
a0 = a0 , a1 = -a\ a2 = -a 2 , a3 = -a 3 (XI1.29)
It is not difficult to see that relation (XII.28) can be written as
all = gllllall (XII.30}
Consequently, raising or lowering of the index 1.1. on the component
of a four-vector is attended by multiplication of the component by
g llll or gllJJ..
When using contravariant and covariant components, the square
of a four-vector [see (XII.27)l can be written as follows:
3
~ a~'all = a 0 a 0
Jl=O
+a a + a a +a a
1 1 2 2 3 3 (XII.31)
By analogy with formula (XII.31), which gives the product of

two identical four-vectors, we can determine the scalar product of
two different four-vectors:
3
~ a~'b11 = a 0b0 + a 1b! + a b + a3b3 =
2 2 aobo- atbt- a2b2 - a3b3
JJ.=O
(XII.32}
The following expression is obvious:
3 3
~l
..::...J a.-"b11 = ~
.::...J a~,
bll (XII.33)
J.l=0 ~l=O
In general in any pair of dummy indices, we may exchange the places

of the superscript and subscript.
It must be noted that in purely spatial rotations (i.e. transforma-
tions not affecting the component a 0 ), tho three spatial components
of the four-vector a~' behave like the components of a vector in three-
dimensional Euclidean space (the component a 0 behaves here like
a three-dimensional scalar). In this connection, the components
of a four-vector can be written as
all= (a 0 , a;) (i = 1, 2, 3) (XII.34}
400 APPENDICES
()r, more briefly,

(XII.35)
The covariant components of the same vector are as follows:
a!! = (a 0 , -a;) (i = 1, 2, 3) (XII.36)
()r
al-l = (a 0 , -a) (XII.37)
The square of a four-vector can be expressed as
:Lj al!al! = (aO)II- ~ af = (a0)2- az (XII.38)
I! (
and the scalar product of four-vectors as

2J al-lb 14 = aOb0 - ab (XII.39)
"
It is a simple matter to see that the increment of the square of
a four-vector can be written in two ways:
&2J a~'a 14 = 2 :Lj a~t&a 14 = 2 ~ a14 &a" (XII.40)
J.t J.t "
Four-Dimensional Gradient. Assume that we are given a scalar
function of the quantities x" :
cp (x 0 , xl, x 2 , x 3)
According to the rules of differential calculus, the increment of
this function is given by the expression
dm
-r
= Bcp dxo
azo
+ 8zBcp 1
dxi + 8xBcp
2
dx2 + 8zBcp
3
dx3
The increment of a function cannot depend on the coordinate

system it is being calculated in, i.e. is an invariant. We thus con-
clude that the quantities
Bcp Bcp Bcp Bcp 4 )
axo • iJxl • iJz2 • /Jza (XII. 1
behave like covariant components of a vector in transformations
of the coordinates. This vector is called the four-gradient of the
function cp.
If we want to introduce the four-operator of the gradient (the oper-
ator V*), its covariant components must be determined as follows:
"'* a
·•o=8ijf• (XII.42)
or
V~ = ( a:o , V) (XII.43)
APPENDICES 401
where 'i1
is the operator of the gradient in three-dimensional Eucli-
dean space. Consequently, the contravariant components oJ the
operator 'il* will be
'il*o=~
axo , 'il*•= -~
oxl , 'i1*2= __
i]x2 - _ _!.._
a_ , 'i1*3- axa (XII.44)
'il·~ = ( a~o • - 'i1) (XII.45)

We are now able to explain the essence of the difference between
contravariant and covariant vectors. By (XII.8), the derivative of
x'~ with respect to x" is a~:
(XII.46)
Introducing these values of a~ into (XII.8), we can give the for-

mula for transforming the contravariant components of a four-·
vector to the form
3 ,
a'IJ.= ~ --a"
axl.t
(XII.47)
ax"
V=O
Considering a scalar function cp as a composite function of the
form
we can write that
By (XII.41), ocp/ox'IJ. is the 11-th covariant component of the gradient

of cp calculated for the system K' [let us designate it by (V'*cp) '),
and ocplox" is the v-th covariant component of V'*<p calculated for
the system K [let us designate it by (V'*<p)vl. We can therefore write
3 "
('il*<p)~ = "" ~ ('il*IP)v
.L.J ax IJ.
(XII.48)
V=O
A comparison of transformations (XII.47) and (XII.48) shows that
they are not identical. The role of the coefficients in one of them
is played by the quantities ox'~lox", and in the other by the quanti-
ties ox"/ox'IJ..
Hence, vectors are said to be contravariant whose components
transform according to the law (XII.47), i.e. like the components of
lf2 26-018
402 APPENDICES
the vectors x~t or dx~t (like the differentials of the coordinates).

Vectors are said to be covariant whose components transform accord-
ing to the law (XII.48), i.e. like the components of a gradient (like
the partial derivatives with respect to the coordinates).
We have seen above that if we are given a contravariant vector all,
we can always use the rule (XII.29) to introduce the covariant vector
corresponding to it (and vice versa). Since the quantities all and a,..
are close relatives, we shall treat them not as components of two
different vectors, but as the contravariant and covariant components
of the same vector.
. For the Euclidean metric, the inverse transformation coefficients
a~ are related to the direct transformation coefficients a~ by the
~xpression a~ = a~. Consequently, the direct and inverse trans-
formations of coordinates have the form
whence
ax' 11 ax v
---;;;v = ax' 11
Hence, when a space has a Euclidean metric, the coefficients of the
transformations (XII.47) and (XII.48) coincide, and the difference
between the contravariant and covariant vectors (and also the
tensors) vanishes.
Transformation of Covariant Components. In a transition from the
system K to the system K', the covariant components of vectors
obviously transform like their contravariant counterparts according
to a linear law. Designating the coefficients of covariant component
transformation by the symbol a~, we can write
(XII.49)
Let us find the relation between the coefficients of transformations

of contravaPiant and covariant components, i.e. between the quanti-
ties a~ and a~. Raising or lowering of the index f.t (or v) on the com-
ponent of a vector is attended by the multiplication of this compo-
nent by giJ.IJ. (or gvv)· Expression (XII.49) can therefore be transformed
as follows:
(XII.50)
-:At the same time for contravariant components, we have

,_,, ... ,,. 3
a'll= ~ a~av
i- V=O
APPENDICES 403
Multiplying the left-hand and right-hand sides by g 1111 , we obtain

3
'\1 (
a 'llg llll = ll
L! av g 1111) av (XII.51)
V=O
A comparison of formulas (XII.50) and (XII.5'1) shows that
-v ll
ai-Lg "" = avg~,ll (XII.52)
With a view to (XII.4), it is simple to obtain that
-o
a 0 --a 0o, -o
ai -- -a 0
i , a-i0 -- -a o -"- a,.i ("z, k -- 1 , 2 , 3) (XII . 53)
i• ai-
A comparison of relations (XII.15) and (XII.52) allows us to
write
aV = aV (XII.54)
ll ll
where a~ are the coefficients of inverse transformation of the contra-

variant components.
We saw above that a~ = ox'V/{}xll [see (XII.46)]. It is evident
that
-aV =--
ax"
ll ax'll
Introducing this expression into (XII.54), we obtain coefficients
for the direct transformation of the covariant components
aV =aV = - -
ax"
ll ll ax' ll
This agrees with (XII.48).
Four-Dimensional Tensors. By a four-tensor of the second rank
is meant a set of the 16 quantities TllV which in a transition from
one coordinate system to another are transformed by the formula
(XII.55)
where a~ are the coefficients from (XII.21) [compare with (X.JO)l.

In the inverse transformation
3
Til"= ~ aJ.LavT'pa (XII.56)
p, a=O P 0
A particular case of a four-tensor is the tensor with the com-

ponents
illl" = allb" (XJI.57)
where all and b" are components of four-vectors.
The components of a four-tensor can be represented in three
forms: as contravariant ones Til", covariant ones T 11 ..,, and mixed
lf 2 26•
:{04 APPENDICES
ones TJJ. v· By analogy with (XII.57), raising or lowering of the time

index does not change the sign of a component, while raising or
lowering of a space index reverses the sign of a component [compare
with (XII.29)l. Haising or lowering of an even number of space
indices obviously leaves a component unchanged. Hence,
Too= roo, Tal= -To\ ... , Tlo = -Tlo, Tn = ru, • ..
roo = roo, Tol = rol, ... , rol = -TOI, rlo = -TlO,
... , T / = T\ = - T 11 , ••• (XII. 58)
Formulas for the transformation of the covariant components of

the tensor (XII.57) can be obtained by taking (XII.49) into account:
Similarly
(XII.59)
where the coefficients a~ are determined by the rules (XII.53).

In the same way, we can arrive at the formula
(XII.60)
Examination of (XII.58) reveals that in general we must distin-

guish between the mixed components TJJ.v and TvJJ.• i.e. see which
index-the first or second-is a superscript and which is a subscript.
Indeed, for instance, in the general case. T 1 0 = T 10 =I= T 01 = T 0 1 •
For the symmetric tensors S!J.V (for which S!J.V = sv!J.), the mixed
components SJJ.v and SvJJ. evidently coincide so that the indices
may be placed one above the other. For antisymmetric tensors, the
.- A JJ. v = -AvJJ. holds.
relat10n
In an antisymmetric four-tensor, six components are independent
(four diagonal ones are zero, the others satisfy the condition AJJ.V =
= -AVJJ.). Consequently, the array of the components of an anti-
symmetric four-tensor is as follows: ·
Aot
0
-Ata (XII.61)
-Ata
APPENDICES 405
Let us find formulas for transforming the components of the tensor

(XII.61). By (XII. 55) for the component A 01, we have
3 3 3
A' 01 = 2J aga~AP 11 =
p, <1=0
~
p~O
ag L]
<r=O
a~Apa
3 3 3 3
=ao '\1 atAOa+ao '\1 a1Ata+ao ); a1A2<J+,.,.o )l, ,.,.1A3a
o - a 1 .LJ a 2 - a '-"a ·- '-"a
cr=O a=O a=O cr=O
The coefficients a~ and a~ are zeros [see (XII.21)], therefore the last
two sums will Yanish. Of the four coefficients a~, only two are non-
zero: a~ and a~. Consequently, in the first two sums, only the first
two addends are non-zero. Hence,
A' 01 = a~ (a~A 00 + a~ A 01) + a~ (a~A 10 + a~ A 11 )
Taking into account that A 00 = A 11 = 0, A 10 = -A 01, substituting

a 0 for a~ and a~, and also a 1 for a! and a~, we arrive at the formula
A 101 = (a: _ ai) A o1 = A o1
[see relation (XII.22)1. In a similar way, we can obtain transform-
ation formulas for the other components of the tensor Al-1".
The formulas for the transformation of all six independent com-
ponents are as follows:
A 101 = A or, A 102 = aoA o2 +
a 1A 12, A 'oa = aoA o3 a 1 A 13 +
A '12 = aoA 12 +
a 1A o2, A 'la = aoA 13 a 1 A oa, A '23 = A 23 +
(XII.62)
We must note that the formulas for the inverse transformation
differ from formulas (XII.62) only in the sign of the terms containing
the factor a 1 [see (XII.23)].
We need formulas (XII.62) in electrodynamics.
Let us form the following expression from the components of
the tensor IIl-1" = al-1b":
lJ II llll =--= lJ allb
).1 ).1
14
This expression is an invariant, i.e. a scalar. Similarly, for any ten-

sor T~'", the expression
~ T 11J.L = T 0 0 + T1 1 + T 2 I + T 3 s (XII.63)
11
is a scalar [compare with (X.21)]. It is called the trace of the tensor.
A glance at (XII.33) shows that lJ IIIl11 = ~ II/. Similarly
~ Tllll = ~ T llll (XII.64)
).1 ).1
406 APPENDICES
The scalar product of the four-vector av and ·the four-tensor Tllv

is defined to be the four-vector bll whose components are determined
by the formula
(XII.65)
Invariant Four-Tensors. A tensor which when multiplied by a

vector leaves the latter unchanged is naturally called a unit tensor.
The components of this tensor must obviously be taken equal to
1 if f.t=V
oil=
v
{
0 if 1-t =I= v
(XII.66)
Indeed, the introduction of these values into formula (XII.65)

yields
biJ. = 2j 0~ av = all
v
as matters should be for a unit tensor.
It follows from (XII.66) that &~ = &~, i.e. that the tensor &~
is symmetric. This is why we have arranged the indices one over
the other. The trace of the tensor&~ is
3
~ 0~=4
iJ.=O
The contravariant and covariant components of a unit tensor are
customarily designated by the symbols g11v and giJ.v• It can be seen
that r
(g!!V) = (gjJ.V) =
( 1 0 0 0)
0 -1
0 0 -1
0 0
0 (XII.67)
0 0 0 -1
A c9mparison with (XII.4) shows that the tensor gllv (like giJ.v)
is a metric tensor.
Let us form the expression
Taking into account (XII.67), we find that when 1-t = 0, this exp!es-
sion is a0 = a0 , while when f.t = i (i = 1, 2, 3), we obtain -at =
= ai. It thus follows that
APPENDICES 407
Similarly, we can see that
Consequently, the scalar product of the vectors aiL and b" can
be written in any of the following three ways:
LJ g11 ..,a1Lb"
2JIL a1Lb11 = IJ.,V · Lj giL"a"'b"
IJ.,V
(XII.68)
The tensors 6~, g"'"' and giL" are invariant-their components are
identical in all coordinate systems. The absolutely antisymmetric
unit four-pseudo tensor of the fourth rank, £IL"P 0 , is also invariant.
The components of this tensor are determined similarly to those
of the pseudo tensor e;kl in Euclidean space [see (VI.15)]. If at least
two indices coincide, eJ.tvpa is zero. Therefore, of its 4 4 = 256 com-
ponents, only 4! = 24 components are non-zero. It is assumed that
eot2a = +1 (XII.69)
The remaining 23 components are assumed to equal +1 or -1 de-
pending on what number of permutations of two indices-even or
odd - gives us the sequence f!, v, p, a from the sequence
0, 1, 2, 3. It is obvious that 12 components have the value +1
and 12 the value -1.
According to the rule we have adopted above, lowering of all
the non-zero indices in e"'vpa must change the sign of the relevant
component. Therefore, £ 0123 = -1. Similarly, e 11 vpa = -£IL"P0
(of the four indices, one must be 0, and the other three-1, 2, 3).
It thus follows that
LJei'VP<JeiJ.Vp:J= -24
J..L,'Y,p,a
(XII. 70)
The Ostrogradsky-Gauss Theorem. In Euclidean space, the Ostro-

gradsky-Gauss theorem is written as follows [see (XII.13)]:
3 3
J2j
V i=t
::: dV=~ ~
I i=l
a 1 df 1 (XII. 71)
(the integral of the divergence of the vector a over a certain volume V

equals the flux of this vector through the surface f enclosing V).
The quantities dft are components of the vector df = n df. They have
the values dfx = dy dz, d[y = dz dx, dfz = dx dy.
The following relation is a generalization of the Ostrogradsky-
Gauss theorem for pseudo-Euclidean four-space:
3 3
J 2j :;: dV*
1'=0
= ~ 2j ~" df"'
11=0
(XII. 72)
408 APPENDICES
where a~ is a four-vector, dV* = dx 0 dx 1 dx 2 dx 3 = c dt dV is an

element of the volume in four-space, df~ is a component of the four-
vector of an element of the hypersurface enclosing the four-volume
over which the integral on the left-hand side of formula (XII. 72)
is taken. The components df~ have the values d/ 0 = dx 1 dx 2 dx3 =
= dV, dfl = dx 0 dx 2 dx 3 = c dt dy dz, etc.
Let us use the Ostrogradsky-Gauss theorem to prove a relation
that we need in Sec. 40. To make our proof more obvious, we shall
x,(1) • (2)
xl
xl x,
f, !2 f, t;.
(a) (b)
Fig. XII.L
first give it for Euclidean three-space, and then perform similar

calculations for four-space.
Assume that we have the vector a 1 whose divergence is zero:
3
~ Bat =0
£.J 8Xj
i=1
Let the vector a 1 be non-zero in a restricted region of space. By

(XII. 71), we have
r~i
::~ dV = ~ 2j a! df I = 0
I i
(XII. 73)
Equation (XII. 73) holds for an arbitrarily taken volume V and

the surface f enclosing it. Let us take as V the volume confined be-
tween two infinite planes x1 = x~ 11 = const and x1 = x~ 2 > = const
(Fig. XII.1a). The integral over the side surface of this volume is
zero because, by assumption, at infinity at = 0. Consequently, the
right-hand side of formula (XII. 73) can be written as (for the plane
) a 1 d/ 1 -
(1) (2)
Jatd/ =0 1
APPENDICES 409
~1 = const, the components d/ 2 and d/ 3 are zero). Hence

Ja, d/
(1)
1 = ~ a1 dft = const
(l!)
The result we have obtained signifies that the integral Ja d/

1 1
;aken over any infinite plane x1 = const has an identical value.

We must note that the coordinates x 2 and x 3 take on all values from
-oo to +oo in integration.
Now let us take as V a volume enclosed by two surfaces of an arbi-
lrary shape for all of whose points the coordinate x 1 is finite, while
~he coordinates x 2 and x 3 vary from -oo to +oo. Consequently,
lhe edges of the surfaces are at infinity (Fig. XII.1b). In this case,
~he right-hand side of formula (XII. 73) can be written as
J~a,dfi- ~ ~
1
a1 dft =0
whertcc-
j~a 1 df 1 = const (XII.74)
t
i.e. has the same value for any surface including all the two-dimen-
sional space x 2 x 3 (the space yz).
Now let us assume that there is a tensor Tit• satisfying the con-
dition
8
~ 8Jx' 11 =0
11 (i=1, 2, 3)
(XII.75)
11-1
The components of T 111 differ from zero in a restricted region of space

Let us form the vector a having the components
(XII.76)
where bi are the components of an arbitrary constant vector (bi -

= const). The divergence of the vector a will be zero. Indeed,
·~ ~=: = ~ 8 ~ 11 2j
11 II t
Ttllb, = 2j b, ~
t k
J:: =0
8
Therefore, the vector (XII. 76) satisfies the conditions in which

relation (XII. 74) is observed. Substitution of the values (XII. 76)
into (XI1.74) yields
j ~ ak d/11= j 2j ( ~ Ttltbi) d/11 =
II II i
J
2j b, ~ T,ll d/11 =
h
const (XII.77)
410 APPENDICES
(integration is performed over the surface including the entire two-

dimensional space x 2x 3 ).
Introducing the symbol
Pi = J_Lj T i~tdf~t (XII. 78)
It
we can write relation (XII. 77) as

2j biPi = const
i
Because of the constancy and arbitrariness of the quantities bit
the last expression can be observed only provided that all the p; 's
are also constant.
We have thus arrived at the following statement: if the tensor.
T;lt satisfies condition (XII. 75) and its components are non-zero in
a restricted region of space, the values of the vector components Pi
do not depend on over which of the surfaces confining the entire two-
dimensional space x 2 x 3 integration in formula (XII. 78) is performed.
Now let us reason similarly for four-space. Assume that there is
the vector a" whose four-divergence is zero:
3 "
~~=0 (XII.79)
LJ ax"
v-o
(a" are non-zero in a restricted region of four-space). By (XII. 72),
we have
JLj ::: dV*=~ ~a" dfv=O
v v
(XII.80)
This equation holds for any closed hypersurface; the integral on the
left-hand side is evaluated over the four-volume confined by this
hypersurface. Let us take as this volume the part of four-space con-
fined between two infinite hyperplanes x 0 = x~u = const and
x 0 = xf21 = const. The coordinates xl, x 2 , x 3 for points of such
hyperplanes-vary from -oo to +oo. Consequently, each of the hyper-
planes consists of the whole three-dimensional space taken at the
instant t 1 = x~ 1 ,/c for the first plane and at the instant t 2 = x~2 ,/c
for tho second one.
For the chosen four-volume, the right-hand side of relation
(XII.80) can be written as
Ja d/ Ja d/ =0
( 1)
0 0-
(2)
0 0
(since dx 0 = 0, all the dfi's are zero). Hence
Ja 0 df 0 - const (XII.81)
APPENDICES 411
This means that the value of the integral (XII.81) does not depend
on which of the hyperplanes x 0 = const integration is performed
over.
Now let us take a four-volume enclosed by two hypersurfaces
of an arbitrary shape for all of whose points the coordinate x 0 is
finite, and the coordinates xl, x 2 , x 3 take on values from -oo to +oo.
Such hypersurfaces include the entire three-dimensional space.
Writing the right-hand side of relation (XII.80) for such a volume,
we have
j~avdfv- J~avdfv=O
(1) v (2) v
Hence, the integral
J~ avdfv=COnst
v
(XII.82)
i.e. does not depend on which of the hypersurfaces enclosing the

entire three-dimensional space it is taken over. In other words, this
integral is time-independent, and its value is conserved.
Let us take as av the four-vector with the components
av = Lj TJJ.VbJJ. (XII.83)
IJ.
where bJ.L are the components ·of an arbitrary constant four-vector
(b" = const), and T11v is a four-tensor satisfying the condition
3
oTIAV
~ - v =0 (f.t=O, 1, 2, 3) (XII.84)
V=O
ox
The components of this tensor are assumed to be non-zero in a restrict-
ed region of four-space. It is not difficult to see that the vector
(XII.83) satisfies condition (XII. 79). Consequently, relation (XII.82)
must be observed for it. Substitution of the values (XII.83) into
(XII.82) yields
J~ avdfv= J~ (~ TJJ.Vb... ) dfv= ~ bll J~ TllVdfv=Const
v v ll ll v
(intog•·ntion i:;~ performed over n11 nddtrnry hyppr·:-ndaec induding
the entire three-dimensional space).
If we introduce the symbol
pll = J~ Tllv df
v
v (Xll.85)
the relation we have obtained can be written as

Lj bJJ.pll = const
IJ.
27*
412 APPENDICES
Owing to the arbitrariness and constancy of the quantities b!J., the

latter condition can be observed only provided that all the pl!'s are
constant (i.e. time-independent). We thus arrive at the following
conclusion. If there is a tensor T~tv whose components are non-zero
only in a restricted region of four-space and satisfy condition ( XI!.84),
the components of the four-vector (X/1.85) are conserved, i.e. do not
change their value with time. It is obvious that the four-vector with
~he components
(XII.86)
v
where a is an arbitrary constant, will also be conserved.
XIII. The Dirac Delta Fundion
The Dirac delta function (6 function) is defined as the function

determined as follows: 6 (x) = 0 at all x's differing from zero; at
:c = 0, the function l3 (x) becomes infinite, and so that
+oo
Jl'J(x)dx=i (XIII.i)
-oo
The delta function is useful owing to its following property:
+oo
) f (x) 6 (x) dx = f (0) (XIII.2)
-oo
where f (x) is an arbitrary continuous function of x. This property
follows from the definition of the delta function. Indeed, since
6 (x) = 0 at all x =1= 0, only the vicinity of the point x = 0 makes
a non-zero contribution to the integral (XIII.2). In this vicinity,
f (x) can be assumed equal to f (0). Putting f (0) outside the integral
sign and takjng (XIII.i) into account, we arrive at (XIII.2).
It is evident that the function o(x - a) has the same properties
in the vicinity of the point x = a as the function l'J(x) in the vicinity
of the point x = 0. Particularly,
+oo
Jf (x) o(x- a) dx =I (a) [(XIII.3)
-oo
. The integration region in formulas (XII1.2) and (XIII.3) must

not necessarily extend from -oo to +oo, it is sufficient for this
region to include a singular point at which the delta function is
non-zero.
Like our introduction of 6 (x), a three-dimensional delta function
designated by l) (r) is defined. It is zero everywhere except for the
APPENDICES 413
onglll of coordinates. At the origin of coordinates, 6 (r) becomes

infinite so that
J
6(r) dV = 1 (XII1.4)
The integral is evaluated over the entire three-dimensional space.
The three-dimensional delta function can be written as the
product of three one-dimensional delta functions:
6(r) = 6(x) 6(y) 6(z) (XIII.5)
It follows from the definition of the delta function that
J/(r) 6(r- r 0) dV = /(r 0 ) (XIII.6)
The integration region in formula (XIII.6) must not necessarily
include the entire three-dimensional space, it is sufficient that this
region include the point determined by the vector r 0 .
XIV. The fourier Series and Integral
A function satisfying the condition

I (t + T) = f (t) (XIV.1)
where T is a constant, is said to be periodic 1 . The quantity T is
called the period of the function. A very simple example of a periodic
function is the harmonic function f (t) = a cos (rot + a), where
ro = 2n/T is the cyclic (angular) frequency of the function.
The overwhelming majority of periodic functions encountered
in physical problems can be written as the series
00
(XlV.2)
n=O
For brevity, we have used the notation
2rt
roo=-r (XIV.3)
where 1' is the period of the function.
The series (XIV.2) is called the Fourier series. The constants an
and bn are called Fourier coefficients. We shall not discuss the con-
ditions which a function must satisfy for its values to coincide with
those of the series (XIV.2), referring readers interested in this matter
to texts on calculus.
A non-periodic function can also be represented as a Fourier
series. But such a representation will be suitable for the non-periodic
function only on the segment from -T/2 to +T/2.
1 Having in mind the applications, we have designated the independent.
variable by t instead of by x.
4f4 APPENDICES
Expression (XIV.2) is an expansion of the function f (t) into a series

in the functions
1, cos ffi 0 t, sin ffi 0 t, ... , cos nffi 0 t, sin nffi 0 t, . . . (XIV.4)
The system of functions (XIV.4) is orthogonal over the segment
-T/2, +T/2. This signifies that an integral of the product of two
different functions of the system over this segment is zero, while
a similar integral of the square of any function is non-zero. The
property of orthogonality can be written briefly as
+T/2
J '!Jn '!Jm dt = {Jnmqn
-T/2
Here 'Pn and '!Jm are any functions belonging to the system (XIV.4).
It is a simple matter to see that q 0 = T and qn = T/2 (for n =F 0).
For this end, it is sufficient to recall that the average values of the
square of a cosine and the square of a sine are 1/2.
The orthogonality of the system of functions (XIV.4) allows us
to find the values of the Fourier coefficients. Let us multiply relation
(XIV.2) by the function cos mffi 0 t (here m =1= 0) and then integrate
over the segment -T/2, +T/2. Owing to orthogonality, of all the
integrals on the right-hand side, only one will be non-zero, namely,
+T/2
Jam cos 2 mffi 0 t dt=am ;
-T/2
+T/2
We therefore arrive at the formula .\ f (t) cos mffi 0 t dt = amT/2,
-T/2
whence
+T/2
am= ; J f (t) cos mffi t dt 0 (m =I= 0) (XIV .5)
-T/2
Similar calculations lead to expressions for bm and a 0 :
+T/2
bm = ; J f (t) sin mffi 0 t dt (m =I= 0) (XIV.G)
-T/2
+T/2
a0 = ~ J f (t) dt, b0 = 0 (XIV. 7)
-T/2
The Fourier series can be written in the complex form. For this
purpose, we replace the cosine and sine with exponentials:
• e inroot _ e- inroot
sm nffi0 t = 2i
APPENDICES 415
The expression in parentheses in formula (XIV.2) now becomes

e inwot + e- inwot e inwot _ e- inC!lot
an 2 + bn 2!
_ Un +ibn e- inwot 4-- Un- iblt einCilot

2 , 2
= C n e-inWot + C -n einwot (XIV.8)
We have introduced the following coefficients instead of a 0 , an,

and bn:
C = an+ibn (XIV .9)
n 2 '
·we must note that
(XIV.10)
(the asterisk stands for a complex conjugate quantity).
Making the substitution (XIV.S) in (XIV.2), we obtain
+oo
t (t) = '\1
--I
cn e-inwot (XIV.11)
n=-oo
Expression (XIV.11) is an expansion of the function f(t) into

a series in the functions
1, e±iwot, e±i2wot, ••• (XIV.t2)
The system of functions (XIV .12) is also orthogonal over the seg-
ment - T/2, +T/2. For complex functions, this signif1es that
+T/2
J '¢n'¢~ dt = 6nmqn
-T/2 .
(here qn =I= 0). Indeed, it is not difficult to
see that
+T/2
J ei<n-m)wot dt = l)nmT (XIV .13)
-T/2
The orthogonality of the system of functions (XIV.12) allows us
to lind tho values of the coefficients C n· Let us multiply relation
(XIV.11) by etmw.t and then integrate over the segment -T/2,
+T/2. Owing to orthogonality, of all the integrals on the right-hand
side, only one will be non-zero, namely
+T/2
J Cme-i(n-m)wot dt = CmT
-T/2
+T/2
We therefore arrive at the formula J f(t) eimw• 1 dt = CmT,
-T/2
·'418 APPENDICES
If two functions c.p and I are related by an expression of the form
c.p (x) = J" I (t) [( (x, t) dt (XIV.23)

a
the function cp (x) is said to be the integral transform of the function

J (t), and the function [( (x, t), the kernel of the integral transform.
A comparison of formulas (XIV.21) and (XIV.23) shows that the
·function c.p (w) determined by expression (XIV.21) is the integral
transform of the function I (~), the kernel of the transform being
.K ( w, ~) = (1/V 2n) efro;. This is why the function c.p ( w) is called
the Fourier transform of the function I (£). The function f (t) de-
termined by expression (XIV.22) is called the inverse Fourier trans-
form. We must note that the direct [formula (XIV.21)1 and the
inverse [formula (XIV.22)l Fourier transforms differ only in the
:sign of the exponent.
The function f (g) in expression (XIV.21) is also called the Fourier
lim age of the function c.p ( w).
INDEX
Acceleration, four-dimensional, 135, 136 Coord ina tcs,

Action, 121ff cylindrical, 389
charged particle In electromagnetic field, generalized, 11t, 19, 82
232ff cyclic, 41
for closed system, 14S normal, 76
contracted, 123 principal, 76
dimension, 3<i transformation formulas, 394ff
for dectromagnetic field, 234ff Curl, 375ff
field and particle system, 236 position vector, 383
generalized expression, 147 Curve,
quantum, 34 eccentricity, 31 or
for relativistic part!clP, 143ff parameter, 311
variation, 238
Angle(s),
EulPr, 85ff Damping, aperiodic, 68f
nutation, 87 Density,
precPssion, 87 charge, 159
proper rotation, 87 hound, 169, 172
Apex, 86, 113 displacement current, 201
Axes, electric current, 209
coordinate, Inversion, 327f energy, electromagnetic field, 209f
free, of rotation, 99ff momentum, 154
principal, or inertia, 91, 92 electroma(!netic field, 213!
momentum nux 155
electromagnetic field, 211
Body, rigid, see Rigid body power, electromagnetic field forces, 208!
Bracket, Poisson, 119f total current, 201
Determinant(s), 338ft
minor, 340
Charges, mul tipllcation, 342
bound, 166ff order, 339
free, 166 transposed, 340
·Coefftclent(s), Dipole,
coupling-, 80 field JJotentlal, 163, 165
Lame, ;lBBI field ,;tr<'ngth, 163
.Cofactor, algebraic, 340, 341 Dirac, P.A.l\1., 177
Collisions, clastic, 49ff DlrPctions, principal, 356
·Condition, Lorentz gauge, 205 Discriminant, quadratic form, 348
in four-dimensional form, 219 Disorder, in transposition, 320
·Conics, 309 ff Dtsplacl·ment, eiPctric, 170f, IH
.Constant, diei<•ctric, 171 true, 298
{:onstraint(s), 1!of virtual, 298
equations, 1/.d' DivPq::ellce. :n2ff
gcotlll'lric, 14 In curvilllll'nr coonlillutes, 390!
ilolonomic, 14t In n-dimcnsional spaet•, 3H
ideal. 15 position VC'Ctor, 38~
integralde, 14
kincma tic, 14
non-holonomic, 15 Effect, Doppler,
non-integrable, 15 longitudinal, 257
non-stationary, 15, 297ft transverse, 257
rheonomous, 15 Einstein, A., 125, 127, 141
scleronomous, 15 Electric field, 2891
stationary, 15 arbitrarily moving charge, 280ft
>Continuum, II dipole radiation, 293
Coordinates, uniformly moving charge, 275, 281
Cartesian, and generalized coordinates, 20 Electromagnetic field,
curvilinear, 387ff energy, 208
orthogonal, 388ft energy density, 209f
cyrlir, liB energy nux density, 209!
420 INDEX
Electromagnetic field, Equations,

four-potential, 219, 221 motion,
momentum density, 213! particle In field, 230!
momentum flux, 211 rigid body, 102ff
momentum flux density, 2! t Newton's, 127
potentlal(s), 203ft second law, 12, 13
gauge transformations, 204! parabola, 312
scalar, 203ft Poisson's, 160, 207
vector, 203ft field in homogeneous dielectric, 173
Ellipse, 309 l'teld In magnetic, 107
canonical equation, 309 tor vector potential, 179ft
eccentricity, 309 wave, 249
equation, 312 tour-dimensional, 249
Ellipsoid, generalized, 260
· angular momentum, 110! Euler, L., 105
energy, 11 or Expansion, potential, In multipoles, 162
inertia, 91 Extremals, 307
tensor, 366, 368 Field,
Energy, electric, see Electric field
centrifugal, 62 electromagnetic, see Electromagnetic Aeld
closed system, 37 electrostatic, in vacuum, 157ft
dissipation, 66 Infinite solenoid, 182ft
free symmetric top, 109! invariants, 225ff
kinetic,: line current, 181
particle, 1U magnetic, see Magnetic fleld
rigid body, 89, 94 In magnetics, 194ft
rotation, 89, 90! point charge in homogeneous dielectric.
particle, In uniformly rotating coordinate 172
system, 62 inside solt>noid, 184!
pendulum, 30 outside solenoid, 185!
potential system, 16, 64 strength, 157ft
relativistic expression, 139, 144 in dielectric, 167
rest, particle, 141 point charge, 158
total, Flux,
particle, 140 magnetic Induction, 177
system, 25 momentum, 211
translational motion, 89, 94 electromagnetic wave, 254f
Equation(s), see also Formula(s) Forcc(s),
canonical, 115ff cen trlfugal, of Inertia, 61
ellipse, 309 conservative, 16
hyperbola, 309f Coriolis, 61
parabola, 310 dissipative, 18
' characteristic, 74 driving, 70
matrix, 351 complex amplitude, 70
constraints, 14! electromotive, Induced, 199
continuity, 200 generalized, 22, 24
from charge conservation, 218 pendulum, 30
In differential form, 188 !(enerallzed-potential, 18
In integral form, 188 !flertia, 61
d' Alembert's, 207, 248 Lorentz, 177
ellipse, 312 Mlnkowski, 137!
Euler's, 105ft, 307, 308 time component, 137!
spherical top, 107 non-potential, 66
Hamilton-Jacobi, 123 potential, 16
classical approximation, 247 stationary, 16
for contracted action, 124 Forms, quadratic, eee Quadratic !orm(s~
relatlvlstle, 146 Formula(s), see also Equation(s)
Hamil ton's, 115ft field transformation, 222ff
hyperbola, 312! Rutherford's, 57
lnvarlance, 125 Four-gradient, 400ft
Lagrange's, 12t, 17!f, 23 Four-momentum, 152
corresponding to cyclic coordinates, U clo!ed system, 152
generalization, 149f covariant component, 146
in generalized coordinates, 19ft, 23 Four-operator, del, 400
for holonomic system, 297ft contravariant components, 401
from least action principle, 34 Four-potential, electromagnetic field, 219.
Laplace's, 160 221, 277
linear, Four-tensors, 40 3ff
differential, with constant coefficients; antlsymmetrlc, 404!
313ft invariant, 406f
homogeneous systems, 345ft Four-vector, 134f, 398!
non-homogeneous systems, 342ft charge-current, 217
Maxwell's, !27, 20t!f, 228ft, 238 components, 398f
derivation from !toast-action principle, contravariant, 399, 401f
237f covariant, 399, 402
motion, covariant components, transformation.
centre of mass, !02 '•02ff
INDEX 421
Four-vector, Lagrangian,
current, 217t in generalized coordinates, 27t
momentum-enerf(y, 140 for non-inertia I reference frames, 57ft
scalar product, 399, 400 pendulum, 30, 31, 32
spatial component, 398 in polar coordinates, 41
square, 398, 400 rigid body, 101t
time component, 398 for uniformly rotating coordinate system,
Four-velocity, 135f 61
Frame, reference, see Reference frame Law,
Frequencies, natural, 74 B!ot-Snvart, 187
Function, conservation,
Dirac delta, 412f angular momentum, 39f
Hamiltonian, 116 charge, 218
Lagrangian, 11, see also Lagrangian energy, 36!
non-periodic, 413!, 416 momentum, 37ft
periodic, 413, 416 electromagnetic induction, 199f
Rayleigh's dissipative, 18! Length, Lorentz contraction, 133
Functional, 300ft Llne(s)
continuous, 302 coordinate, 387
extremum, 304, 306ft nodal, Sa
linear, 301 world, 127
maximum, 305
minimum, 305
variation, 302! Magnetic field, 2891
arbitrarily moving charge, 283
dipole radiation, 293
Gauge, quadrupole radiation, 29ft
Coulomb, 207 uniformly moving charge, 275t
Lorentz, 206 Magnetization, 19ft
transverse, 207 diamagnet!cs, 196
Gauging, potentials, 20/o paramagnetics, 196
Gradient, 370ff Mass,
in curvilinear coordinates, 389f reduced, system, ft7
four-dimensional, 400ft relativistic, 142
In n-dimensional space, 371 rest, 142
Matrix(ces), 330ff
algebra, 334ff
Hamilton, w. R., 33, 115, 371 asymmetric, 333
Hamiltonian, 116, 118 characteristic equation, 3~ 1
relativistic expression, 143 column, 333
Homogeneity, 36 commutative, 335
Hyperbola, 309 degenerate, 331
canonical equation, 309! diagonal, 333f
eccentricity, 310 difference, 334
equation, 312f elements, 330
identity, 334
inverse, 331
Index( Ices), as linear Ol?erator, 332
dummy, 16, 20, 371, 393 multlplicatJOn, 334ff
free, 393 orthogonal, 331, 338
refractive, complex, 2641 quadratic form, 348, 350
Induction, rank, 346
electric, 170 row, 333
magnetic, 177, 194, 196! singular, 331
Integral(s), skew-symmetric, 333
Fourier, 417 squarr, 332, 346
motion, 25f, 37, 113, 119, 120 sum, 334
Interval, 127ft, 144 symmetric, 333
space-like, 129 trace, 33ft
time-like, 129 transformation, 330
Invarlance, gauge, 205 transposed, 331, 340
Invariants, fleld, 225ff unit, 334, 337
Isotropy, 36 Maxwell, J. c., 200, 201
Mechanics, relativlstic 1 12i
Minor, determinant, 3~t0
Kernel, Integral transform, 418 Moment(s),
Kovalevskaya, S. v., 111 dipole, 162
inertia,
axial, 91
Lagrange, J. L., 111 centrifugal, 91
Lagrangian, 11f, 17, 18, 23f principal, 91
charged particle in field, 245 magnetic, 192f, 295
density, 148 plane loop with current, 193
for field, 239 system or discrete charges, 192
and energy, 24ff quadrupole, 296
tor fleld, 236 Momentum,
for free relativistic particle, 1431 angular, 62
422 INDEX
Momentum, Particles,
four-dimensional, 139! scattering, 49, 53ff
generalized, 24 scattering angle, 52
pendulum, 30 total energy, 140
relativistic particle, 141t, 144 trajectory, 42ff
system, density, 154 equation In polar coordinates, 43ff
three-dimensional, relativistic formula, velocities after collision, 51
139 Pendulum(s),
for uniformly rotating coordinate system, with constantly accelerating suspensloa
61 point, 31f
Monopoles, coupled, 77ff
Dirac's, 177 kinetic energy, 77
tleld potential, 165 natural frequency, 78
Motion, normal coordinates, 79
aperiodic, 681 potential energy, 77
flhlte, 45 simple, 30
lntlnlte, 45 with uniformly moving suspension point,
Multlpole, 162 30!
field potential, 165f Permeability, 196
first-order, 16 2 Permittivity, 171
complex, 263
relative, 171
Number, Phase, wave, 256
complex wave, 261, 263 Poln t, world, 127
degrees of freedom, 15 Polarization, dielectric, 167!, 3561
Nutation, 87, 113f Postulates, Einstein's, 125
Potentlal(s),
calculated In dipole approximation, 288
Octupole, 166 field,
Operator, dipole, 163, 165
d'Alembertlan, 207, 217 electrostatic, 158f
del, 371 monopole, 165
application, 379ff multipole, 165f
Integral determination, 3861 point charge, 158
Hamiltonian, 371 quadrupole, 165
Laplacian, 381t generalized, 18
In curvilinear coordinates, 392 L!enard-Wiechart, 278
Oscillations, magnetic, 194
damped, 66ff retarded, 271f
forced, 70ff field of charge system, 2M
tree, 64ff system, 16
harmonic, 65f vector, 194ff, 291ft
complex amplitude, 66 magnetic field, 178, 189
normal, 76 Power, dipole radiation, 291, 295
small-amplitude, 64ff Precession,
pseudoregular, 114
regular, 87, 108
Parabola, 310 Principle,
canonical equation, 310 constancy of light speed, 1251
eccentricity, 310 Einstein's, relativity, 125, 127
equation, 312 Gallleo's, mechanical, relativity, 1251
Parameter, Hamilton, 33ft
focal, 311 least action, 13, 33ff, 117, 233
Impact, 54, 56 variational, mechanics, 34
Partlcle(s), Problem, Lagrange's, 111!
capture, 49 Process, aperiodic, 68
charged, in electromagnetic l'leld, Pseudoscalars, 325, 328
action, 232ff PReudotcnsor, 364,
mot1o11, 230 contraction, an:,
defiectlon angle, 52, 54 Pseudovector, 3241, 328, 369
and Impact parameter, 54
divergence angle, 52f
elastic collisions, 49ff Quadratic form(s), 3471r
energy, 245 canonical, 348
In central field, 42 diagonal, 348
In fteld, Hamiltonian, 246f discriminant, 348
generalized momentum, 245 matrix, 348 350
head-on collision, 53 negative definite, 34 7, 348
motion in central force field, 41ff positive definite, 347, 348
moving along stralgbt line, 321 simultaneous reduction to diagonal form,
generalized force, 33 3521
generalized momentum, 33 Quadrupole, 163f
Lagrangian, 33 moment, 1641
In non-inertial reference frame, 57ff
recoil angle, 52
relativistic, 143ft Radiation,
rest energy, 141 dipole, 290f
INDEX
Radiation, Tensors,
magnetic dipole, 293, 296 Maxw<>ll stn•ss, 215, 254
quadrupole, 294! m~tric, 393, 406
Reaction, constraint, 15 multiplication, 360
Reference frame, In n-dim~nslonal space, 359
c-, 50 permittivity, 175
centre-of-mass, 50 principal values, 359, 367f
1-, 50 rank, 359
I a bora tory, 50 scalar product, 361
Retardation, proper, 284, 286 second rank, 358, 365ft, 369ft
Rigid body, stress, 155
angular mom~ntum, 05ff sum, 360
proper, 96 symmetric, 362f
degrees of freedom, 82 as operator, 368
elementary displacement, 82 of second rank, 365ft
equations of motion, 102ff trace, 227, 360, 405, 406
kinematics, 82ll true, 3631
Rule, Cramer's, 343f unit, 359, 406
Th~orem,
Euler's, for homogeneous functions, 301}•
Scalars, 359 Gauss's, 159, 177
true, 325 Gauss's divergence, stt Theorem, Ostro-
Scattering, differential effective cross sec- gradsky-Gauss
tion, 54!, 57 Ostrogradsky-Gauss, 375, 407ft
SchrMinger, E., 124 for pseudo-Euclid<'an four-space, 407t
Series, Fourier, 413ft parallel axis, 9~
In complex form, 414f St<•iner, 94
Space, Stokes's, 157, 379
configuration, 19, 34 Theory, r<>latlvity, special, 125
pseudo-Euclidean, 394 Time,
q-, 19 proper, body, 129
Strength, magnetic field, 196, 197 retarded, system, 284
Surface(s), Top,
constant cp, 370 asymmetrical, \13
coordinate, 387 rapid, 114
Susceptibility, spherical, 92
electric, 171 Euler's equations, 107
magnetic, 196 symmetrical, 93, 114
Symbol, Kronecker, 319 free, 107ff
skew-symmetric, 319 in homogeneous gravitational field, 111
System(s), unbalanced, 111ft
closed, Trace,
four-momentum, 152 matrices, 334
energy, 37 tensor, 227, 360, 405, 406
conservative, 16, 36 Traj~ctorv, motion, 121
coordinate, see Reference frame Transform, Fourier, 418
energy-momentum tensor, 152ft inverse, 418
holonomlc, 14, 20 Transformations, Lorentz, 132f
with Ideal constraints, tsf .
mechanical, 11
number of degrees of freedom, 15 Variation, function, 30l
potential, 16 Vector(s), 316ff, 359
potential energy, 16 ttccelrration, 135
reduced mass, 47 circulation, 375
retarded time, 284 components, 323
total energy, 25 transformations, 325f
two interacting particles, 45ll cross product, 317
dot product, 316
nux, :l72
T<•nsor(s), :l~f>ff four-. srt· Four-v(•ctor
antisynunetrlc, 362C four-liilll<'llSiOIIDl, ;1\)!(, •tt ttlso Jo'our-V<'CtOr
of second rank, 369f four-position, 130ff
cofftponents, 358 free, 318
diagonal, 358 Increment In rotation, 328ft
contraction, 360 n-, 327
dielectric susceptibility, 358 In n-dimensional space, 34'
electrical susceptibility, 175 Poynting's 210, 253f, 291
electromagnetic field, 220f scalar product, 316, 322, 327
elllpsold, 366, 368 tor n-dimensional space, 327
energy-momentum, 152ft scalar triple product, 317, 325, 32&-
for electromagnetic field, 240ft square of vector product, 318
four-dimensional, 403ft true, 324!
inertia, 91ff vector product, 317, 323
Inner product, 361 vector triple product, 317
invariant, 360 velocity, 135
as linear operator, 362 wave, 255
magnetic susceptibility, 198 four-dimensional, 256
-424 INDEX
Veloclty(les), Waves,
centre ot mass, SO . non-monochromaUc, 265ft
tour-dimensional, 13St plane, 250
generalized, 12, 19 In conducting medium, 260U
Volume, contraction. 133 oolarlzed,
circularly, 259
elliptically, 259
Wave(s), linearly, 260
attenuation tactor, 263 spectral decomposition. 265
average Intensity, 266t Work, dissipative torces, HJ
electromagnetic, 249
magnetic l'leld, 263
monocbromatfc, 255 Zone, wave, 288
:Y!l!i 530.1
Hropb BnanuMnpoa!lq Caaenbea

OCHOBbiTEOPETHqECHOn ~H3HHH
TOM I
Mexamn~a u aJieKTponnuaMHKa
flayqubtlt pena~<Top H. B. ;r(eitHeKo. Pe~~:aKTOpbt JI. A. CyoooTHHa, T. B. Jiaaapesa.

Xy)I011<Hl!K B. K BHccenrannes. Xy~~:omecTsennbiit pen;aKTOP H. B. 3oTosa.
·TexunqeCKHit pe~~:aKTDP JI. II. BnpJOKosa. HoppeKTOPbi M. 3. <l>perep, JI. H. BacHJihesa.
liB No 3306
CnaHO B Haoop 25.12.81. llO)IUHCaHO K neqaTH 23.06.82. ~OpMaT 60X90 1/ 16,
ByMara THnorpa!j>CKaii No 1. rapliHTypa OObiKIIODeHH31I. lleqaTb BbiC01131I.
Oo'beM 13,25 oyM. JI. Yen. neq, a. 26,50. YcJI. Kp.-oTT. 26,89. Yq.-H3JI. n. 24,89. r
'Hal\. No 17/1604. THpam 6450 aKa. 3aK. 018. U:ena 2 p. 34 11.
JII3;r(ATEJibCTBO «MHP». Moctma, Hi PHli<CKHit nep., 2.
<lpneua Tpy)IOBOrO HpacHOro 3HaMeHH MOCKOBCK31I THnorpa!j>HII N\ 7 «HCJ<pa pesOJI!OnHH»
CowanoJIHrpa<f>npoMa rocyn;apcTneunoro RoMHTeTa CCCP no AeJiaM nan;aTeJihCTB,
l!OJIItrpa<jJUH ll KHHli<HOit TOprOBJIH. MOCKBa 103001, Tpexnpyi\Hblit nep., 9.
'C 1704020000-016 28-1982 2

056(01)-82 , 'i.
Printed in the Union of Soviet Socialist Republics

Igor V. Savelyev, D.Sc. (Physics
and Mathematics), has been head
of the Deparfment of General Physics
at the Moscow Institute
of Engineering Physics for about 2S years,
alter devoting quite a lew years
to experimental physics.
He is the author ol about 100 published
works in physics, including Physics.
A General Course, in three volumes,
translated into English
by Mir Publishers, and A Collection
of Questions and Problems
In General Physics (in Russian).
His principal interests include
instruction in physics at
higher technical schools
and its improvement.
Professor Savelyev holds the title
ol Honoured Scientist ol the RSFSR
and is a USSR State Prize Winner.

I V Savelyev Fundametals of Theoretical Physics Vol 1 PDF

Uploaded by

Copyright:

Available Formats

I V Savelyev Fundametals of Theoretical Physics Vol 1 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

I V Savelyev Fundametals of Theoretical Physics Vol 1 PDF

Uploaded by

Copyright:

Available Formats

What is the subject matter covered in the book?

What is the subject matter covered in the book?

What mathematical topics does the author provide appendices for?

What mathematical topics does the author provide appendices for?

I. V.

Mir Publishers Moscow

OCHOBbl TEOPETI;14ECKO~ ~1;13HKH

f.13P,A TEllbCTBO «HAYKA•

MIR PUBLISHERS MOSCOW

Mir Publishers would be grateful for your comments on the

..- ...... _--~-~~-: :;:,·r\_ ,.E~

© ll3AaTeJibCTBo <<HayRa>>, :;;;--·~{{9 {3 ~H

The bonk being offered to the reader is a logical continuation of

mathematical aids and find the required information in them. This

Part One. Mechanics 11

Chapter I. The Variational Principle in Mechanics 11

Chapter II. Conservation Laws 36

Chapter III. Selected Problems in Mechanics 41

Chapter IV. Small-Amplitude Oscillations 64

l 9. Oscillations of a System with Many Deg1·ees of Freedom 72

Chapter V. Mechanics of a Rigid Body 82

Chapter VI. Canonical Equations 115

Chapter VII. The Special Theory of Relativity 125

Pari Two. Electrodynamics 157

Chapter VIII. Electrostatics 157

Chapter IX. Magnetostatics 177

Chapter X. Time-Varying Electromagnetic Field 199

54. Displacement Current 200

Chapter X I. Equations ol" Eleetrodynarnit's in thP For11·-llinrcn"ional

Chapter XII. The Variational Principle in Electrodynamics 232

Chapter XIII. Elrctromagnetic Waves 248

Chapter XIV. Hadiation of Electromagnetic Waves 269

THE VARIATIONAL PRINCIPLE

Depending on the nature of the objects being studied, mechanics

1 What is meant is a perfectly rigid body.

quantities by means of which the position of a system in space can

m""x. = R:u my= mg + R,,, m~- = 0

If we characterize the position of the system by the generalized

(3) two particles joined by a weightless thread of length l. The

(4) a perfectly rigid body which can be considered as a system of

It differs from Eq. (2.4) for a holonomic non-stationary constraint

3. Equations of Motion in Cartesian Coordinates

. '' U = U (x1 , x 2 , •• :, Xn, t) • (3.2)

Formula (3.3) can also be written as follows:

where ra. is the position vector of the particle1•

where F1 are the components of the non-potential forces, and n is

Consequently, we arrive at the relations

named Lagrange's equations.

1 By the derivative of the scalar cp with respect to the vector a is understood

Lagrange's equations can sometimes be written in the form of (3.8)

The function D has a simple physical meaning. The work done by

dA = Lj F~4 > dx 1 = 2j Fid>;, dt = - 2j 8 ? ;, dt = -2D dt

(we have used Euler's theorem on homogeneous functions, see Appen-

4. Lagrange's Equations in Generalized Coordinates

Generalized coordinates qlt are defined to be any quantities (lengths,

8~- !~ =R 1 +Ft (i=1,2, ... ,n)

The quantities 8xJaq~~. do not contain the generalized velocities ,j~~..

OXi " 0 ( uxi ) • "" i) 2 xi

Comparing this expression with (4.8), we arrive at the relation

The quantities 8xJaq. do not contain the generalized velocities ,j..