Sharipov Classical Electrodynamics and T

arXiv:physics/0311011 v1 4 Nov 2003
RUSSIAN FEDERAL COMMITTEE

FOR HIGHER EDUCATION
BASHKIR STATE UNIVERSITY
SHARIPOV R. A.
CLASSICAL ELECTRODYNAMICS
AND THEORY OF RELATIVITY
the manual
Ufa 1997
2
UDC 517.9
Sharipov R. A. Classical Electrodynamics and Theory of
Relativity: the manual / Publ. of Bashkir State University —
Ufa, 1997. — pp. 163. — ISBN 5-7477-0180-0.
This book is a manual for the course of electrodynamics and

theory of relativity. It is recommended primarily for students
of mathematical departments. This defines its style: I use el-
ements of vectorial and tensorial analysis, differential geometry,
and theory of distributions in it.
In preparing Russian edition of this book I used computer
typesetting on the base of AMS-TEX package and I used cyrillic
fonts of Lh-family distributed by CyrTUG association of Cyrillic
TEX users. English edition is also typeset by AMS-TEX.
This book is published under the approval by Methodic Com-
mission of Mathematical Department of Bashkir State University.
Referees: Chair of Algebra and Geometry of Bashkir State

Pedagogical University (BGPI),
Prof. V. A. Baikov, Ufa State University for
Aviation and Technology (UGATU).
ISBN 5-7477-0180-0 c Sharipov R.A., 1997

English Translation c Sharipov R.A., 2003
3
CONTENTS.
CONTENTS. ....................................................................... 3.
PREFACE. .......................................................................... 5.
CHAPTER I. ELECTROSTATICS AND MAGNETO-
STATICS. ..................................................................... 7.
§ 1. Basic experimental facts and unit systems. ...................... 7.
§ 2. Concept of near action. ................................................ 13.
§ 3. Superposition principle. ................................................ 15.
§ 4. Lorentz force and Biot-Savart-Laplace law. .................. 18.
§ 5. Current density and the law of charge conservation. ..... 21.
§ 6. Electric dipole moment. ............................................... 24.
§ 7. Magnetic moment. ....................................................... 26.
§ 8. Integral equations of static electromagnetic field. ........... 31.
§ 9. Differential equations of static electromagnetic field. ...... 41.
CHAPTER II. CLASSICAL ELECTRODYNAMICS. ......... 43.

§ 1. Maxwell equations. ...................................................... 43.
§ 2. Density of energy and energy flow for electromagnetic
field. ........................................................................... 46.
§ 3. Vectorial and scalar potentials of electromagnetic
field. ........................................................................... 54.
§ 4. Gauge transformations and Lorentzian gauge. ............... 56.
§ 5. Electromagnetic waves. ................................................ 59.
§ 6. Emission of electromagnetic waves. ............................... 60.
CHAPTER III. SPECIAL THEORY OF RELATIVITY. ..... 68.

§ 1. Galileo transformations. ............................................... 68.
4
§ 2. Lorentz transformations. .............................................. 73.

§ 3. Minkowsky space. ........................................................ 77.
§ 4. Kinematics of relative motion. ...................................... 82.
§ 5. Relativistic law of velocity addition. .............................. 90.
§ 6. World lines and private time. ........................................ 91.
§ 7. Dynamics of material point. ......................................... 95.
§ 8. Four-dimensional form of Maxwell equations. ............. 100.
§ 9. Four-dimensional vector-potential. .............................. 107.
§ 10. The law of charge conservation. ................................ 112.
§ 11. Note on skew-angular and curvilinear
coordinates. ............................................................. 115.
CHAPTER IV. LAGRANGIAN FORMALISM

IN THEORY OF RELATIVITY. .............................. 119.
§ 1. Principle of minimal action for particles and fields. ...... 119.
§ 2. Motion of particle in electromagnetic field. .................. 124.
§ 3. Dynamics of dust matter. ........................................... 128.
§ 4. Action functional for dust matter. ............................... 133.
§ 5. Equations for electromagnetic field. ............................. 141.
CHAPTER V. GENERAL THEORY OF RELATIVITY. ... 145.

§ 1. Transition to non-flat metrics and curved Minkowsky
space. ....................................................................... 145.
§ 2. Action for gravitational field. Einstein equation. .......... 147.
§ 3. Four-dimensional momentum conservation law
for fields. .................................................................. 153.
§ 4. Energy-momentum tensor for electromagnetic field. ..... 155.
§ 5. Energy-momentum tensor for dust matter. .................. 157.
§ 6. Concluding remarks. .................................................. 160.
REFERENCES. ............................................................... 162.
CONTACTS. ................................................................... 163.

PREFACE.
Theory of relativity is a physical discipline which arose in

the beginning of XX-th century. It has dramatically changed
traditional notion about the structure of the Universe. Effects
predicted by this theory becomes essential only when we describe
processes at high velocities close to light velocity
c = 2.998 · 105 km/sec.
In XIX-th century there was the only theory dealing with such
processes, this was theory of electromagnetism. Development of
theory of electromagnetism in XIX-th century became a premise
for arising theory of relativity.
In this book I follow historical sequence of events. In Chapter I
electrostatics and magnetostatics are explained starting with first
experiments on interaction of charges and currents. Chapter II is
devoted to classical electrodynamics based on Maxwell equations.
In the beginning of Chapter III Lorentz transformations are
derived as transformations keeping form of Maxwell equations.
Physical interpretation of such transformation requires uniting
space and time into one four-dimensional continuum (Minkowsky
space) where there is no fixed direction for time axis. Upon
introducing four-dimensional space-time in Chapter III classical
electrodynamics is rederived in the form invariant with respect to
Lorentz transformations.
In Chapter IV variational approach to describing electromag-
netic field and other material fields in special relativity is con-
sidered. Use of curvilinear coordinates in Minkowsky space and
appropriate differential-geometric methods prepares background
for passing to general relativity.
6
In Chapter V Einstein’s theory of gravitation (general rela-

tivity) is considered, this theory interprets gravitational field as
curvature of space-time itself.
This book is addressed to Math. students. Therefore I paid
much attention to logical consistence of given material. References
to physical intuition are minimized: in those places, where I
need additional assumptions which do not follow from previous
material, detailed comment is given.
I hope that assiduous and interested reader with sufficient
preliminary background could follow all mathematical calculations
and, upon reading this book, would get pleasure of understanding
how harmonic is the nature of things.
I am grateful to N. T. Ahtyamov, D. I. Borisov, Yu. P. Ma-
shentseva, and A. I. Utarbaev for reading and correcting Russian
version of book.
November, 1997;
November, 2003. R. A. Sharipov.
CHAPTER I
ELECTROSTATICS AND MAGNETOSTATICS
§ 1. Basic experimental facts and unit systems.

Quantitative description of any physical phenomenon requires
measurements. In mechanics we have three basic quantities and
three basic units of measure: for mass, for length, and for time.
Quantity Unit Unit Relation

in SI in SGS of units
mass kg g 1 kg = 103 g
length m cm 1 m = 102 cm
time sec sec 1 sec = 1 sec
Units of measure for other quantities are derived from the

above basic units. Thus, for instance, for measure unit of force
due to Newton’s second law we get:
(1) N = kg · m · sec−2 in SI,
(2) dyn = g · cm · sec−2 in SGS.
Unit systems SI and SGS are two most popular unit systems
in physics. Units for measuring mechanical quantities (velocity,
acceleration, force, energy, power) in both systems are defined in
quite similar way. Proportions relating units for these quantities
CopyRight c Sharipov R.A., 1997, 2003.

8 CHAPTER I. ELECTROSTATICS AND MAGNETOSTATICS.
can be derived from proportions for basic quantities (see table

above). However, in choosing units for electric and magnetic
quantities these systems differ essentially.
Choice of measure unit for electric charge in SGS is based on
Coulomb law describing interaction of two charged point.
Coulomb law. Two charged points with charges of the same
sign are repulsing, while points with charges of opposite signs are
attracting with force proportional to
quantities of their charges and in-
verse proportional to square of dis-
tance between them:
Q1 Q2
(1.1) F ∼ .
r2
Measure unit of charge in SGS is

chosen such that coefficient in for-
mula (1.1) is equal to unity. Hence
Fig. 1.1 we have the following relation:
unit of charge in SGS = dyn 1/2 · cm= g 1/2 · cm3/2 · sec−1 .
Coulomb law itself then is written in form of the equality
Q1 Q2
(1.2) F = .
r2
Force F defined by the relationship (1.2) is very strong. How-
ever, in everyday life it does not reveal itself. This is due to the
screening. The numbers of positive and negative charges in nature
are exactly balanced. Atoms and molecules, which constitute all
observable matter around us, have the same amount of positive
and negative charges. Therefore they are electrically neutral in
§ 1. BASIC FACTS AND UNIT SYSTEMS. 9
whole. Force (1.2) reveals itself in form of chemical links only

when atoms are pulled together.
Electric current arises as a result of motion of charged points.
This occurs in metallic conductor, which usually have lengthy
form (form of wire). Current in such conductor is determined by
the amount of charge passing through it within the unit of time.
Therefore for unit of current we have:
unit of current in SGS = unit of charge in SGS · sec−1 =

= g 1/2 · cm 3/2 · sec−2 .
Let’s consider straight conducting rod of the length l. Current

in it leads to misbalance of charges in its ends. Charges of definite
sign move to one end of the rod, while lack of these charges in
the other end of the rod is detected as the charge of opposite
sign. Then Coulomb force (1.2) arises that tends to recover
balance of charges in electrically neutral rod. This means that
in such rod current could not flow in constant direction during
long time. Another situation we have with conductor of the
form of ring or circuit. Here current does not break the balance
of charges. Direct current can flow in it during unlimitedly
long time. Circular conductor itself thereby remains electrically
neutral and no Coulomb forces arise.
In spite of absence of Coulomb forces, in experiments the inter-
action of two circular conductors with currents was detected. This
interaction has other nature, it is not due to electrical, but due to
magnetic forces. The magnitude of magnetic forces depends es-
sentially on the shape and mutual arrangement of circular conduc-
tors. In order to reveal quantitative characteristics for magnetic
forces one should maximally simplify the geometry of conduc-
tors. For this purpose they are deformed so that each possesses
straight rod-shaped part of sufficiently big length l. These rod-
shaped parts are arranged par-

allel to each other with the dis-
tance r between them. In the
limit, when l is much larger than
r, this configuration of conduc-
tors can be treated as a pair
of infinitely long parallel conduc-
tors. In experiments it was found
that such conductors do interact
according to the following law.
Ampere law. Force of inter-
Fig. 1.2 action of two infinite parallel
conductors with currents per unit
length of them is proportional ti the values of currents in them
and inverse proportional to the distance between them:
F I1 I2
(1.3) ∼ .
l r
Two co-directed currents attract each other, while opposite di-
rected currents repulse each other.
The unit of current in SGS was already introduced above.
Therefore coefficient of proportionality in formula (1.3) is unique
quantity that should be determined in experiment. Here is the
measure unit for this coefficient: sec 2 · cm−2 . It coincides with
inverse square of velocity. Therefore formula (1.3) in SGS is
written as
F 2 I1 I2
(1.4) = 2 .
l c r
Constant c in (1.4) is a velocity constant. The value of this
constant is determined experimentally:
(1.5) c ≈ 2.998 · 1010 cm/sec.
§ 1. BASIC FACTS AND UNIT SYSTEMS. 11
As we shall see below, constant c in (1.5) coincides with velocity

of light in vacuum. Numeric coefficient 2 in (1.4) is introduced
intentionally for to provide such coincidence.
In SI measure unit of current 1 A (one ampere) is a basic unit.
It is determined such that formula (1.3) is written as
F 2 µ0 I1 I2
(1.6) = .
l 4π r
Here π = 3.14 . . . is exact (though it is irrational) mathematical
constant with no measure unit. Constant µ0 is called magnetic
susceptibility of vacuum. It has the measure unit:
(1.7) µ0 = 4π · 10−7 N · A−2 .
But, in contrast to constant c in (1.5), it is exact constant. Its
value should not be determined experimentally. One could choose
it to be equal to unity, but the above value (1.7) for this constant
was chosen by convention when SI system was established. Due
to this value of constant (1.7) current of 1 ampere appears to
be in that range of currents, that really appear in industrial and
household devices. Coefficient 4π in denominator (1.6) is used in
order to simplify some other formulas, which are more often used
for engineering calculations in electric technology.
Being basic unit in SI, unit of current ampere is used for
defining unit of charge of 1 coulomb: 1C = 1A · 1sec. Then
coefficient of proportionality in Coulomb law (1.1) appears to be
not equal to unity. In SI Coulomb law is written as
1 Q1 Q2
(1.8) F = .
4πǫ0 r2
Constant ǫ0 is called dielectric permittivity of vacuum. In con-
trast to constant µ0 in (1.7) this is physical constant determined
experimentally:
(1.9) ǫ0 ≈ 8.85 · 10−12 C 2 · N−1 · m−2 .
Constants (1.5), (1.7), and (1.9) are related to each other by the
following equality:
1
(1.10) c= p ≈ 2.998 · 108 m/sec.
ǫ0 µ0
From the above consideration we see that SGS and SI systems

differ from each other not only in the scale of units, but in for-
mulas for two fundamental laws: Coulomb law and Ampere law.
SI better suits for engineering calculations. However, derivation
of many formulas in this system appears more huge than in SGS.
Therefore below in this book we use SGS system.
Comparing Coulomb law and Ampere law we see that electrical
and magnetic forces reveal themselves in quite different way.
However, they have common origin: they both are due to electric
charges. Below we shall see that their relation is much more
close. Therefore theories of electricity and magnetism are usually
united into one theory of electromagnetic phenomena. Theory of
electromagnetism is a theory with one measurable constant: this
is light velocity c. Classical mechanics (without Newton’s theory
of gravitation) has no measurable constants. Newton’s theory of
gravitation has one constant:
(1.11) γ ≈ 6.67 · 10−8 cm3 · g−1 · sec−2 .
This theory is based on Newton’s fourth law formulated as follows.

Universal law of gravitation. Two point masses attract
each other with the force proportional to their masses and in-
verse proportional to the square of distance between them.
Universal law of gravitation is given by the same formula
M1 M2
(1.12) F =γ
r2
§ 2. CONCEPT OF NEAR ACTION. 13
in both systems: in SGS and in SI.

According to modern notion of nature classical mechanics and
Newton’s theory of gravitation are approximate theories. Cur-
rently they are replaced by special theory of relativity and general
theory of relativity. Historically they appeared as a result of
development of the theory of electromagnetism. Below we keep
this historical sequence in explaining all three theories.
Exercise 1.1. On the base of above facts find quantitative re-
lation of measure units for charge and current in SGS and SI.
§ 2. Concept of near action.

Let’s consider pair of charged bodies, which are initially fixed,
and let’s do the following mental experiment with them. When
we start moving second body apart from first one, the distance r
begins increasing and consequently force of Coulomb interaction
(1.2) will decrease. In this situation we have natural question:
how soon after second body starts moving second body will feel
change of Coulomb force of interaction ? There are two possible
answers to this question:
(1) immediately;
(2) with some delay depending on the distance between bodies.
First answer is known as concept of distant action. Taking this
concept we should take formula (1.2) as absolutely exact formula
applicable for charges at rest and for moving charges as well.
Second answer is based on the concept of near action. Ac-
cording to this concept, each interaction (and electric interaction
among others) can be transmitted immediately only to the point
of space infinitesimally close to initial one. Transmission of any
action to finite distance should be considered as a process of
successive transmission from point to point. This process always
leads to some finite velocity of transmission for any action. In
the framework of the concept of near action Coulomb law (1.2) is

treated as approximate law, which is exact only for the charges
at rest that stayed at rest during sufficiently long time so that
process of transmission of electric interaction has been terminated.
Theory of electromagnetism has measurable constant c (light
velocity (1.5)), which is first pretender for the role of transmission
velocity of electric and magnetic interactions. For this reason
electromagnetic theory is much more favorable as compared to
Newton’s theory of gravitation.
The value of light velocity is a very large quantity. If we settle
an experiment of measuring Coulomb force at the distances of the
order of r ≈ 10 cm, for the time of transmission of interaction we
would get times of the order of t ≈ 3 · 10−10 sec. Experimental
technique of XIX-th century was unable to detect such a short
interval of time. Therefore the problem of choosing concept could
not be solved experimentally. In XIX-th century it was subject for
contests. The only argument against the concept of distant action
that time, quite likely, was its straightness, its self-completeness,
and hence its scarcity.
In present time concept of near action is commonly accepted.
Now we have the opportunity for testing it experimentally in the
scope of electromagnetic phenomena. Let’s study this concept
more attentively. According to the concept of near action, process
of transmitting interaction to far distance exhibits an inertia.
Starting at one point, where moving charge is placed, for some
time this process exist in hidden form with no influence to both
charges. In order to describe this stage of process we need to
introduce new concept. This concept is a field.
Field is a material entity able to fill the whole space and able
to act upon other material bodies transmitting mutual interaction
of them.
The number of fields definitely known to scientists is not big.
There are only four fundamental fields: strong field, weak field,

§ 3. SUPERPOSITION PRINCIPLE. 15
electromagnetic field, and gravitational field. Strong and weak

fields are very short distance fields, they reveal themselves only
in atomic nuclei, in collisions and decay of elementary particles,
and in stellar objects of extremely high density, which are called
neutron stars. Strong and weak interactions and fields are not
considered in this book.
There are various terms using the word field: vector field, tensor
field, spinor field, gauge field, and others. These are mathematical
terms reflecting some definite properties of real physical fields.
§ 3. Superposition principle.
Let’s apply concept of near action to Coulomb law for two
charged points. Coulomb force in the framework of this concept
can be interpreted as follows: first charge produces electric field
around itself, and this field acts upon other charge. Result of such
action is detected as a force F applied to second charge. Force
is vectorial quantity. Let’s denote by F vector of force and take
into account the direction of this vector determined by verbal
statement of Coulomb law above. This yields
r2 − r1
(3.1) F = Q1 Q2 .
|r2 − r1 |3
Here r1 and r2 are radius-vectors of points, where charges Q1 and

Q2 are placed. Let’s consider vector E determined as the ratio
E = F/Q2 . For this vector from formula (3.1) we derive
r2 − r1
(3.2) E = Q1 .
|r2 − r1 |3
Vector E depends upon the position of first charge and upon

its value. It depends also on the position of second charge, but
it doesn’t depend on the value of second charge. One can take
vector E for quantitative measure of electric field produced by first

charge Q1 at the point r2 , where second charge is placed. Vector
E can be determined by formula (3.2) or it can be measured
experimentally. For this purpose one should place test charge q
to the point r2 and one should measure Coulomb force F acting
upon this test charge. Then vector E is determined by division of
F by the value of test charge q:
(3.3) E = F/q.
Now consider more complicated situation. Suppose that

charges Q1 , . . . , Qn are placed at the points r1 , . . . , rn . They
produce electric field around them, and this field acts upon test
charge q placed at the point r. This action reveals as a force F
applied to the charge q. Again we can define vector E of the form
(3.3) and take it for the quantitative measure of electric field at
the point r. This vector is called vector of intensity of electric
field or simply vector of electric field at that point.
Generally speaking, in this case one cannot be a priori sure
that vector E does not depend on the quantity of test charge q.
However, there is the following experimental fact.
Superposition principle. Electric field E at the point r
produced by a system of point charges Q1 , . . . , Qn is a vectorial
sum of electric fields that would be produced at this point by each
charge Q1 , . . . , Qn separately.
Superposition principle combined with Coulomb law leads to
the following formula for the intensity of electric field produced
by a system of point charges at the point r:
n
X r − ri
(3.4) E(r) = Qi .
i=1
|r − ri |3
§ 3. SUPERPOSITION PRINCIPLE. 17
Using superposition principle, one can pass from point charges

to continuously distributed charges. Suppose that the number of
point charges tends to infinity: n → ∞. In such limit sum in
formula (3.4) is replaced by integral over 3-dimensional space:
Z
r − r̃ 3
(3.5) E(r) = ρ(r̃) d r̃.
|r − r̃|3
Here ρ(r̃) is spatial density of charge at the point r̃. This value
designates the amount of charge per unit volume.
In order to find force acting on test charge q we should invert
formula (3.3). As a result we obtain
(3.6) F = q E(r).
Force acting on a charge q in electric field is equal to the product

of the quantity of this charge by the vector of intensity of field at
the point, where charge is placed. However, charge q also produces
electric field. Does it experience the action of its own field ? For
point charges the answer to this question is negative. This fact
should be treated as a supplement to principle of superposition.
Total force acting on a system of distributed charges in electric
field is determined by the following integral:
Z
(3.7) F = ρ(r) E(r) d3 r.
Field E(r) in (3.7) is external field produced by external charges.

Field of charges with density ρ(r) is not included into E(r).
Concluding this section, note that formulas (3.4) and (3.5) hold
only for charges at rest, which stayed at rest for sufficiently long
time so that process of interaction transmitting reached the point
of observation r. Fields produced by such systems of charges are
called static fields, while branch of theory of electromagnetism
studying such fields is called electrostatics.
§ 4. Lorentz force and Biot-Savart-Laplace law.

Ampere law of interaction of parallel conductors with currents
is an analog of Coulomb law for magnetic interactions. Accord-
ing to near action principle, force F
arises as a result of action of mag-
netic field produced by a current in
first conductor upon second conduc-
vt tor. However, parallel conductors
cannot be treated as point objects:
formula (1.4) holds only for l ≫ r.
In order to get quantitative measure
of magnetic field at some point r
let’s consider current I2 in (1.4) as
Fig. 4.1 a flow of charged particles of charge
q each, and each moving along con-
ductor with constant velocity v. If we denote by ν the number
of such particles per unit length of conductor, then in the whole
length l we would have N = ν l particles. Then during time in-
terval t we would have n = ν v t particles passing through a fixed
cross-section of the conductor. They carry charge amounting to
Q = q ν v t. Therefore for current I2 in second conductor we get
I2 = Q/t = q ν v.
Upon calculating force acting on a segment of conductor of the

length l by formula (1.4) we should divide it by the number of
particles N contained in this segment. Then for the force per each
particle we derive
2 I1 I2 l 2 I1 q v
(4.1) F = = 2 .
c2 r N c r
Formula determines (4.1) qualitative dependence of F on q and
on v: each charged particle moving in magnetic field experiences
§ 4. LORENTZ FORCE AND . . . 19
a force proportional to its charge q and to the magnitude of its

velocity vector v = |v|, i. e. we have proportionality
(4.2) F ∼ q v.
Force and velocity both are vectorial quantities. Simplest way to
relate two vectors F and v is to use vector product of v with
some third vectorial quantity H:
q
(4.3) F= [v, H(r)].
c
Here c is scalar constant equal to light velocity. Vectorial quantity
H(r) is a quantitative measure of magnetic field at the point r.
It is called intensity of magnetic field at that point. Scalar factor
1/c in (4.3) is used for to make H to be measured by the same
units as intensity of electric field E in (3.6). Force F acting on
a point charge in magnetic field is called Lorentz force. Total
Lorentz force acting on a charge in electromagnetic field is a sum
of two components: electric component and magnetic component:
q
(4.4) F = qE+ [v, H].
c
Formula (4.4) extends formula (3.6) for the case of general electro-
magnetic fields. It holds not only for static but for time-dependent
(non-static) fields. Surely the above derivation of formula (4.4) is
empiric. Actually, one should treat formula (4.4) as experimental
fact that do not contradict to another experimental fact (1.4)
within theory being developed.
Let’s turn back to our conductors. Formula (4.3) can be
interpreted in terms of currents. Each segment of unit length of
a conductor with current I in magnetic field H experiences the
force
F I
(4.5) = [τ , H]
l c
acting on it. Here τ is unit vector tangent to conductor and di-

rected along current in it. Total force acting on circular conductor
with current I is determined by contour integral
I
I
(4.6) F= [τ (s), H(r(s))] ds,
c
where s is natural parameter on contour (length) and r(s) is

vector-function determining shape of contour in parametric form.
Let’s consider the case of two parallel conductors. Force F now
can be calculated by formula (4.5) assuming that first conductor
produces magnetic field H(r) that acts upon second conductor.
Auxiliary experiment shows that vector H is perpendicular to
the plane of these two parallel conductors. The magnitude of
magnetic field H = |H| can be determined by formula (4.1):
2 I1
(4.7) H= .
c r
Here r is the distance from observation point to the conductor

producing field at that point.
Magnetic field produced by conductor with current satisfies
superposition principle. In particular, field of infinite straight
line conductor (4.7) is composed by fields produced by separate
segments of this conductor. One cannot measure magnetic field of
separate segment experimentally since one cannot keep constant
current is such separate segment for sufficiently long time. But
theoretically one can consider infinitesimally small segment of
conductor with current of the length ds. And one can write
formula for magnetic field produced by such segment of conductor:
1 [I τ , r − r̃]
(4.8) dH(r) = ds.
c |r − r̃|3
§ 5. CURRENT DENSITY AND . . . 21
Here τ is unit vector determining spatial orientation of infinitesi-

mal conductor. It is always taken to be directed along current I.
In practice, when calculating magnetic fields produced by circular
conductors, formula (4.8) is taken in integral form:
I
1 [I τ (s), r − r̃(s)]
(4.9) H(r) = ds.
c |r − r̃(s)|3
Like in (4.6), here s is natural parameter on the contour and r̃(s)

is vectorial function determining shape of this contour. Therefore
τ (s) = dr̃(s)/ds. The relationship (4.8) and its integral form (4.9)
constitute Biot-Savart-Laplace law for circular conductors with
current.
Biot-Savart-Laplace law in form (4.8) cannot be tested exper-
imentally. However, in integral form (4.9) for each particular
conductor it yields some particular expression for H(r). This
expression then can be verified in experiment.
Exercise 4.1. Using relationships (4.6) and (4.9), derive the
law of interaction of parallel conductors with current in form (1.4).
Exercise 4.2. Find magnetic field of the conductor with cur-
rent having the shape of circle of the radius a.
§ 5. Current density
and the law of charge conservation.
Conductors that we have considered above are kind of ideal-
ization. They are linear, we assume them having no thickness.
Real conductor always has some thickness. This fact is ignored
when we consider long conductors like wire. However, in some
cases thickness of a conductor cannot be ignored. For example, if
we consider current in electrolytic bath or if we study current in
plasma in upper layers of atmosphere. Current in bulk conductors
can be distributed non-uniformly within volume of conductor.

The concept of current density j is best one for describing such

situation.
Current density is vectorial quantity depending on a point of
conducting medium: j = j(r). Vector of current density j(r)
indicate the direction of charge transport at the point r. Its
magnitude j = | j | is determined by the amount of charge passing
through unit area perpendicular to vector j per unit time. Let’s
mark mentally some restricted domain Ω within bulk conducting
medium. Its boundary is smooth closed surface. Due to the
above definition of current density the amount of charge flowing
out from marked domain per unit time is determined by surface
integral over the boundary of this domain, while charge enclosed
within this domain is given by spatial integral:
Z Z
3
(5.1) Q= ρ d r, J= j, n dS.
Ω ∂Ω
Here n is unit vector of external normal to the surface ∂Ω

restricting domain Ω.
Charge conservation law is one more fundamental experimental
fact reflecting the nature of electromagnetism. In its classical
form it states that charges cannot appear from nowhere and
cannot disappear as well, they can only move from one point to
another. Modern physics insert some correction to this statement:
charges appear and can disappear in processes of creation and
annihilation of pairs of elementary particles consisting of particle
and corresponding antiparticle. However, even in such creation-
annihilation processes total balance of charge is preserved since
total charge of a pair consisting of particle and antiparticle is
always equal to zero. When applied to integrals (5.1) charge
conservation law yields: Q̇ = −J. This relationship means that
decrease of charge enclosed within domain Ω is always due to
§ 5. CURRENT DENSITY AND . . . 23
charge leakage through the boundary and conversely increase of

charge is due to incoming flow through the boundary of this
domain. Let’s write charge conservation law in the following
form:
Z Z
d 3
(5.2) ρd r + j, n dS = 0.
dt
Ω ∂Ω
Current density j is a vector depending on a point of conducting

medium. Such objects in differential geometry are called vector
fields. Electric field E and magnetic field H are other examples
of vector fields. Surface integral J in (5.1) is called flow of vector
field j through the surface ∂Ω. For smooth vector field any surface
integral like J can be transformed to spatial integral by means of
Ostrogradsky-Gauss formula. When applied to (5.2), this yields
Z
∂ρ
(5.3) + div j d3 r = 0.
∂t
Ω
Note that Ω in (5.3) is an arbitrary domain that we marked men-

tally within conducting medium. This means that the expression
being integrated in (5.3) should be identically zero:
∂ρ
(5.4) + div j = 0.
∂t
The relationships (5.2) and (5.4) are integral and differential
forms of charge conservation law respectively. The relationship
(5.4) also is known as continuity equation for electric charge.
When applied to bulk conductors with distributed current j
within them, formula (4.6) is rewritten as follows:
Z
1
(5.5) F= [ j(r), H(r)] d3 r.
c
Biot-Savart-Laplace law for such conductors also is written in

terms of spatial integral in the following form:
Z
1 [ j(r̃), r − r̃] 3
(5.6) H(r) = d r̃.
c |r − r̃|3
In order to derive formulas (5.5) and (5.6) from formulas (4.6)

and (4.8) one should represent bulk conductor as a union of linear
conductors, then use superposition principle and pass to the limit
by the number of linear conductors n → ∞.
§ 6. Electric dipole moment.

Let’s consider some configuration of distributed charge with
density ρ(r) which is concentrated within some restricted domain
Ω. Let R be maximal linear size of the domain Ω. Let’s choose
coordinates with origin within this domain Ω and let’s choose
observation point r which is far enough from the domain of
charge concentration: |r| ≫ R. In order to find electric field E(r)
produced by charges in Ω we use formula (3.5):
Z
r − r̃ 3
(6.1) E(r) = ρ(r̃) d r̃.
|r − r̃|3
Ω
Since domain Ω in (6.1) is restricted, we have inequality |r̃| ≤ R.

Using this inequality along with |r| ≫ R, we can write Taylor
expansion for the fraction in the expression under integration in
(6.1). As a result we get power series in powers of ratio r̃/|r|:

r − r̃ r 1 r r r̃ r̃
(6.2) = + · 3 · , − + ... .
|r − r̃|3 |r|3 |r|2 |r| |r| |r| |r|
Substituting (6.2) into (6.1), we get the following expression for

§ 6. ELECTRIC DIPOLE MOMENT. 25
the vector of electric field E(r) produced by charges in Ω:
r 3 r, D r − |r|2 D
(6.3) E(r) = Q 3
+ + ... .
|r| |r|5
First summand in (6.3) is Coulomb field of point charge Q placed

at the origin, where Q is total charge enclosed in the domain Ω.
It is given by integral (5.1).
Second summand in (6.3) is known as field of point dipole
placed at the origin. Vector D there is called dipole moment. For
charges enclosed within domain Ω it is given by integral
Z
(6.4) D= ρ(r̃) r̃ d3 r̃.
Ω
For point charges dipole moment is determined by sum

n
X
(6.5) D= Qi r̃i .
i=1
For the system of charges concentrated near origin, which is

electrically neutral in whole, the field of point dipole
3 r, D r − |r|2 D
(6.6) E(r) =
|r|5
is leading term in asymptotics for electrostatic field (3.4) or

(3.5) as r → ∞. Note that for the system of charges with
Q = 0 dipole moment D calculated by formulas (6.4) and (6.5)
is invariant quantity. This quantity remains unchanged when
we move all charges to the same distance at the same direction
without changing their mutual orientation: r̃ → r̃ + r0 .
Exercise 6.1. Concept of charge density is applicable to point

charges as well. However, in this case ρ(r) is not ordinary function.
It is distribution. For example point charge Q placed at the point
r = 0 is represented by density ρ(r) = Q δ(r), where δ(r) is Dirac’s
delta-function. Consider the density
3
X ∂δ(r)
(6.7) ρ(r) = D, grad δ(r) = Di .
i=1
∂ri
Applying formula (5.1), calculate total charge Q corresponding to

this density (6.7). Using formula (6.4) calculate dipole moment
for distributed charge (6.7) and find electrostatic field produced
by this charge. Compare the expression obtained with (6.6) and
explain why system of charges described by the above density (6.7)
is called point dipole.
Exercise 6.2. Using formula (3.7) find the force acting on

point dipole in external electric field E(r).
§ 7. Magnetic moment.
Let’s consider situation similar to that of previous section.
Suppose some distributed system of currents is concentrated in
some restricted domain near origin. Let R be maximal linear
size of this domain Ω. Current density j(r) is smooth vector-
function, which is nonzero only within Ω and which vanishes at
the boundary ∂Ω and in outer space. Current density j(r) is
assumed to be stationary, i. e. it doesn’t depend on time, and it
doesn’t break charge balance, i. e. ρ(r) = 0. Charge conservation
law applied to this situation yields
(7.1) div j = 0.
§ 7. MAGNETIC MOMENT. 27
In order to calculate magnetic field H(r) we use Biot-Savart-

Laplace law written in integral form (5.6):
Z
1 [ j(r̃), r − r̃] 3
(7.2) H(r) = d r̃.
c |r − r̃|3
Ω
Assuming that |r| ≫ R, we take Taylor expansion (6.2) and

substitute it into (7.2). As a result we get
Z
[ j(r̃), r] 3
H(r) = d r̃ +
c |r|3
Ω
(7.3) Z
3 r, r̃ [ j(r̃), r] − |r|2 [ j(r̃), r̃] 3
+ d r̃ + . . . .
c |r|5
Ω
Lemma 7.1. First integral in (7.3) is identically equal to zero.
Proof. Denote this integral by H1 (r). Let’s choose some

arbitrary constant vector e and consider scalar product
Z Z
e, [ j(r̃), r] 3 j(r̃), [ r, e] 3
(7.4) H1 , e = d r̃ = d r̃.
c |r|3 c |r|3
Ω Ω
Then define vector a and function f (r̃) as follows:
[r, e]
a= , f (r̃) = a, r̃ .
c |r|3
Vector a does not depend on r̃, therefore in calculating integral

(7.4) we can take it for constant vector. For this vector we derive
a = grad f . Substituting this formula into the (7.4), we get

Z
H1 , e = j, grad f d3 r̃ =
Ω
(7.5) Z Z
= div(f j) d3 r̃ − f div j d3 r̃.
Ω Ω
Last integral in (7.5) is equal to zero due to (7.1). Previous inte-

gral is transformed to surface integral by means of Ostrogradsky-
Gauss formula. It is also equal to zero since j(r̃) vanishes at the
boundary of domain Ω. Therefore
Z
(7.6) H1 , e = f j, n dS = 0.
∂Ω
Now vanishing of vector H1 (r) follows from formula (7.6) since e

is arbitrary constant vector. Lemma 7.1 is proved.
Let’s transform second integral in (7.3). First of all we denote
it by H2 (r). Then, taking an arbitrary constant vector e, we form
scalar product H2 , e . This scalar product can be brought to
Z
1
(7.7) H2 , e = j(r̃), b(r̃) d3 r̃,
c |r|5
Ω
where b(r̃) = 3 r, r̃ [r, e] − |r|2 [r̃, e]. If one adds gradient of

arbitrary function f (r̃) to b(r̃), this wouldn’t change integral
in (7.7). Formulas (7.5) and (7.6) form an example of such
invariance. Let’s specify function f (r̃), choosing it as follows:
3
(7.8) f (r̃) = − r, r̃ r̃, [r, e] .
2

§ 7. MAGNETIC MOMENT. 29
For gradient of function (7.8) by direct calculations we find
3 3
grad f (r̃) = − r̃, [r, e] r − r, r̃ [r, e] =
2 2
3
= −3 r, r̃ [r, e] − r r̃, [r, e] − [r, e] r, r̃ .
2
Now let’s use well-known identity [a, [b, c]] = b a, c − c a, b .

Assuming that a = r̃, b = r, and c = [r, e], we transform the
above expression for grad f to the following form:
3
(7.9) grad f (r̃) = −3 r, r̃ [r, e] − [r̃, [r, [r, e]]].
2
Right hand side of (7.9) contains triple vectorial product. In order

to transform it we use the identity [a, [b, c]] = b a, c − c a, b
again, now assuming that a = r, b = r, and c = e:
3 3
grad f (r̃) = −3 r, r̃ [r, e] − r, e [r̃, r] + |r|2 [r̃, e].
2 2
Let’s add this expression for grad f to vector b(r̃). Here is

resulting new expression for this vector:
3 1
(7.10) b(r̃) = − r, e [r̃, r] + |r|2 [r̃, e].
2 2
Let’s substitute (7.10) into formula (7.7). This yields

Z
−3 r, e r, [ j(r̃), r̃] + |r|2 e, [ j(r̃), r̃] 3
H2 , e = d r̃.
2 c |r|5
Ω
Note that quantities j(r̃) and r̃ enter into this formula in form of
vector product [ j(r̃), r̃] only. Denote by M the following integral:

Z
[r̃, j(r̃)] 3
(7.11) M= d r̃.
2c
Ω
Vector M given by integral (7.11) is called magnetic moment for

currents with density j(r̃). In terms of M the above relationship
for scalar product H2 , e is written as follows:
3 r, e r, M − |r|2 e, M
(7.12) H2 , e = .
|r|5
If we remember that e in formula (7.12) is an arbitrary constant

vector, then from (7.3) and lemma 7.1 we can conclude that the
field of point magnetic dipole
3 r, M r − |r|2 M
(7.13) H(r) =
|r|5
is leading term in asymptotical expansion of static magnetic field

(4.9) and (5.6) as r → ∞.
Like electric dipole moment D of the system with zero total
charge Q = 0, magnetic moment M is invariant with respect
to displacements r → r + r0 that don’t change configuration of
currents within system. Indeed, under such displacement integral
(7.11) is incremented by
Z
[r0 , j(r̃)] 3
(7.14) △M = d r̃ = 0.
2c
Ω
Integral in formula (7.14) is equal to zero by the same reasons as

in proof of lemma 7.1.
§ 8. INTEGRAL EQUATIONS . . . 31
Exercise 7.1. Consider localized system of currents j(r) with

current density given by the following distribution:
(7.15) j(r) = −c [M, grad δ(r)].
Verify the relationship (7.1) for the system of currents (7.15) and
find its magnetic moment M. Applying formula (5.6), calculate
magnetic field of this system of currents and explain why this sys-
tem of currents is called point magnetic dipole.
Exercise 7.2. Using formula (5.5), find the force acting upon
point magnetic dipole in external magnetic field H(r).
Exercise 7.3. By means of the following formula for the torque
Z
1
M= [r, [ j(r), H]] d3 r
c
find torque M acting upon point magnetic dipole (7.15) in homo-

geneous magnetic field H = const.
§ 8. Integral equations
for static electromagnetic field.
Remember that we introduced the concept of flow of vector
field through a surface in considering charge conservation law (see
integral J in (5.1)). Now we consider flows of vector fields E(r)
and H(r), i. e. for electric field and magnetic field:
Z Z
(8.1) E= E, n dS, H= H, n dS.
S S
Let S be closed surface enveloping some domain Ω, i. e. S =

∂Ω. Electrostatic field E is determined by formula (3.5). Let’s
substitute (3.5) into first integral (8.1) and then let’s change order
of integration in resulting double integral:
Z Z
r − r̃, n(r)
(8.2) E= ρ(r̃) dS d3 r̃.
|r − r̃|3
∂Ω
Inner surface integral in (8.2) is an integral of explicit function.

This integral can be calculated explicitly:
Z (
r − r̃, n(r) 0, if r̃ 6∈ Ω,
(8.3) dS =
|r − r̃|3 4π, if r̃ ∈ Ω.
∂Ω
Here by Ω = Ω ∪ ∂Ω we denote closure of the domain Ω.

In order to prove the relationship (8.3) let’s consider vector
field m(r) given by the following formula:
r − r̃
(8.4) m(r) = .
|r − r̃|3
Vector field m(r) is smooth everywhere except for one special

point r = r̃. In all regular points of this vector field by direct
calculations we find div m = 0. If r̃ 6∈ Ω special point of the field
m is out of the domain Ω. Therefore in this case we can apply
Ostrogradsky-Gauss formula to (8.3):
Z Z
m, n dS = div m d3 r = 0.
∂Ω Ω
This proves first case in formula (8.3). In order to prove second

case, when r̃ ∈ Ω, we use tactical maneuver. Let’s consider spher-
ical ǫ-neighborhood O = Oǫ of special point r = r̃. For sufficiently
small ǫ this neighborhood O = Oǫ is completely enclosed into the
domain Ω. Then from zero divergency condition div m = 0 for

the field given by formula (8.4) we derive
Z Z
(8.5) m, n dS = m, n dS = 4π.
∂Ω ∂O
The value of last integral over sphere ∂O in (8.5) is found by direct

calculation, which is not difficult. Thus, formula (8.3) is proved.
Substituting (8.3) into (8.2) we get the following relationship:
Z Z
(8.6) E, n dS = 4π ρ(r) d 3 r.
∂Ω Ω
This relationship (8.6) can be formulated as a theorem.

Theorem (on the flow of electric field). Flow of electric field
through the boundary of restricted domain is equal to total charge
enclosed within this domain multiplied by 4π.
Now let’s consider flow of magnetic field H in (8.1). Static
magnetic field is determined by formula (5.6). Let’s substitute
H(r) given by (5.6) into second integral (8.1), then change the
order of integration in resulting double integral:
Z Z
1 [ j(r̃), r − r̃], n(r)
(8.7) H= dS d 3 r̃.
c |r − r̃|3
∂Ω
It’s clear that in calculating inner integral over the surface ∂Ω

vector j can be taken for constant. Now consider the field
[ j, r − r̃]
(8.8) m(r) = .
c |r − r̃|3
Like (8.4), this vector field (8.8) has only one singular point r = r̃.
Divergency of this field is equal to zero, this fact can be verified
by direct calculations. As appears in this case, singular point
makes no effect to the value of surface integral in (8.7). Instead
of (8.3) in this case we have the following formula:
Z
1 [ j, r − r̃], n(r)
(8.9) dS = 0.
c |r − r̃|3
∂Ω
For r̃ 6∈ Ω the relationship (8.9) follows from div m = 0 by

applying Ostrogradsky-Gauss formula. For r̃ ∈ Ω we have the
relationship similar to the above relationship (8.5):
Z Z
(8.10) m, n dS = m, n dS = 0.
∂Ω ∂O
However, the value of surface integral over sphere ∂O in this case

is equal to zero since vector m(r) is perpendicular to normal
vector n at all points of sphere ∂O. As a result of substituting
(8.9) into (8.7) we get the relationship
Z
(8.11) H, n dS = 0,
∂Ω
which is formulated as the following theorem.

Theorem (on the flow of magnetic field). Total flow of mag-
netic field through the boundary of any restricted domain is equal
to zero.
Let r(s) be vectorial parametric equation of some closed spatial
curve Γ being the rim for some open surface S, i. e. Γ = ∂S. Open
surface S means that S and Γ have empty intersection. By S we
denote the closure of the surface S. Then S = S ∪ Γ. Taking

s for natural parameter on Γ, we define circulation for electric
and magnetic fields in form of the
n
following contour integrals:
n
S I
e= E, τ ds,
Γ
(8.12) I
n
h= H, τ ds.
Γ
Γ
Substituting (3.5) into (8.12) and
Fig. 8.1 changing the order of integration in
resulting double integral, we get the
following equality for circulation of electric field:
Z I
r(s) − r̃, τ (s)
(8.13) e= ρ(r̃) ds d 3 r̃.
|r(s) − r̃|3
Γ
Due to (8.13) we need to consider vector field (8.4) again. For

r̃ 6∈ Γ, taking into account Γ = ∂S and applying Stokes formula,
we can transform contour integral in (8.13) to surface integral:
I Z
r(s) − r̃, τ (s)
(8.14) ds = rot m, n dS = 0.
|r(s) − r̃|3
Γ S
Values of integral (8.14) at those points r̃ ∈ Γ are of no matter

since when substituting (8.14) into integral (8.13) such points
constitute a set of zero measure.
Vanishing of integral (8.14) for r̃ 6∈ Γ follows from rot m = 0,
this equality can be verified by direct calculations. Singular point
r = r̃ of vector field (8.4) is unessential since surface S, for which

Γ is a boundary, can be deformed so that r̃ 6∈ S. The result of

substituting (8.14) into (8.13) can be written as an equation:
I
(8.15) E, τ ds = 0.
∂S
Theorem (on the circulation of electric field). Total circula-

tion of static electric field along the boundary of any restricted
open surface is equal to zero.
Formula like (8.15) is available for magnetic field as well. Here
is this formula that determines circulation of magnetic field:
I Z
4π
(8.16) H, τ ds = j, n dS.
c
∂S S
Corresponding theorem is stated as follows.

Theorem (on the circulation of magnetic field). Circulation of
static magnetic field along boundary of restricted open surface is
equal to total electric current penetrating this surface multiplied
by fraction 4 π/c.
Integral over the surface S now is in right hand side of formula
(8.16) explicitly. Therefore surface spanned over the contour Γ
now is fixed. We cannot deform this surface as we did above
in proving theorem on circulation of electric field. This leads to
some technical complication of the proof. Let’s consider ε-blow-
up of surface S. This is domain Ω(ε) being union of all ε-balls
surrounding all point r ∈ S. This domain encloses surface S and
contour Γ = ∂S. If ε → 0, domain Ω(ε) contracts to S.
Denote by D(ε) = R3 \ Ω(ε) exterior of the domain Ω(ε) and
then consider the following modification of formula (5.6) that
expresses Biot-Savart-Laplace law for magnetic field:

Z
1 [ j(r̃), r − r̃] 3
(8.17) H(r) = lim d r̃.
ε→0 c |r − r̃|3
D(ε)
Let’s substitute (8.17) into integral (8.12) and change the order of
integration in resulting double integral. As a result we get
Z I
1 [ j(r̃), r(s) − r̃], τ (s)
(8.18) h = lim ds d 3 r̃.
ε→0 c |r(s) − r̃|3
D(ε) Γ
In inner integral in (8.18) we see vector field (8.8). Unlike vector

filed (8.4), rotor of this field is nonzero:
3 r − r̃, j (r − r̃) − |r − r̃|2 j

(8.19) rot m = .
c |r − r̃|5
Using Stokes formula and taking into account (8.19), we can

transform contour integral (8.18) to surface integral:
I
1 [ j(r̃), r(s) − r̃], τ (s)
ds =
c |r(s) − r̃|3
Γ
Z
3 r − r̃, j(r̃) r − r̃, n(r) − |r − r̃|2 j(r̃), n(r)
= dS.
c |r − r̃|5
S
Denote by m(r̃)
e vector field of the following form:
3 r̃ − r, n(r) (r̃ − r) − |r̃ − r|2 n(r)

m(r̃)
e = .
c |r̃ − r|5
In terms of the field m(r̃)

e formula for h is written as
Z Z
h = lim m(r̃),
e j(r̃) dS d 3 r̃.
ε→0
D(ε) S
Vector field m(r̃)

e in this formula has cubic singularity |r̃ − r|−3
at the point r̃ = r. Such singularity is not integrable in R3 (if
we integrate with respect to d 3 r̃). This is why we use auxiliary
domain D(ε) and limit as ε → 0.
Let’s change the order of integration in resulting double integral
for circulation h. This leads to formula
Z Z Z Z
3
m(r̃),
e j(r̃) d r̃ dS = grad f (r̃), j(r̃) d 3 r̃ dS,
S D(ε) S D(ε)
since vector field m(r̃)

e apparently is gradient of the function f (r̃):
r̃ − r, n(r)
(8.20) f (r̃) = − .
c |r̃ − r|3
Function f (r̃) vanishes as r̃ → ∞. Assume that current density

also vanishes as r̃ → ∞. Then due to the same considerations as
in proof of lemma 7.1 and due to formula (7.1) spatial integral in
the above formula can be transformed to surface integral:
Z Z
(8.21) h = lim f (r̃) j(r̃), ñ(r̃) dSe dS.
ε→0
S ∂D(ε)
Let’s change the order of integration in (8.21) then take into

account common boundary ∂D(ε) = ∂Ω(ε). Outer normal to the
surface ∂D(ε) coincides with inner normal to ∂Ω(ε). This coin-

cidence and explicit form of function (8.20) lead to the following
expression for circulation of magnetic field h:
Z Z
j(r̃), ñ(r̃) r̃ − r, n(r) e
(8.22) h = lim dS dS.
ε→0 c |r̃ − r|3
∂Ω(ε) S
Let’s denote by V (r̃) inner integral in formula (8.22):

Z
r̃ − r, n(r)
(8.23) V (r̃) = dS.
|r̃ − r|3
S
Integral (8.23) is well-known in mathematical physics. It is called

potential of double layer. There is the following lemma, proof of
which can be found in [1].
Lemma 8.1. Double layer potential (8.23) is restricted func-
tion in R3 \ S. At each inner point r̃ ∈ S there are side limits
V± (r̃) = lim V (r),

r→±S
inner limit V− (r̃) as r tends to r̃ ∈ S from inside along normal

vector n, and outer limit V+ (r̃) as r tends to r̃ ∈ S from outside
against the direction of normal vector n. Thereby V+ − V− = 4π
for all points r̃ ∈ S.
In order to calculate limit in formula (8.22) we need to study
the geometry of ε-blow-up of the surface S. On Fig. 8.2 below
we see cross-section of the domain Ω(ε) obtained from the surface
S shown on Fig. 8.1. For sufficiently small ε boundary of the
domain Ω(ε) is composed of three parts:
(8.24) ∂Ω(ε) = S0 ∪ S+ ∪ S− .
Surface S0 is a part of ε-blow-up of the contour Γ Area of this

surface S0 satisfies the relationship
(8.25) S0 ∼ επL as ε → 0,
n S+ where L is length of contour

Γ. Surfaces S+ and S− are ob-
tained as a result of normal shift
of surface S to the distance ε
n along normal vector n, and to
the same distance against nor-
S mal vector n.
S− Substituting (8.24) into (8.22)
we break surface integral over
S0 ∂Ω(ε) into three parts. Since
double layer potential and func-
tion | j(r̃)| are restricted, we get
Fig. 8.2 the relationship
Z
j(r̃), ñ(r̃) e
(8.26) lim V (r̃) dS = 0.
ε→0 c
S0
For other two summand we also can calculate limits as ε → 0:

Z Z
j(r̃), ñ(r̃) e j(r), n(r)
(8.27) V (r̃) dS −→ ± V± (r) dS.
c c
S± S
We shall not load reader with the proof of formulas (8.24), (8.25)
and (8.27), which are sufficiently obvious. Summarizing (8.26)
and (8.27) and taking into account lemma 8.1, we obtain
Z
4π
(8.28) h= j(r), n(r) dS.
c
S
§ 9. DIFFERENTIAL EQUATIONS . . . 41
This relationship (8.28) completes derivation of formula (8.16)

and proof of theorem on circulation of magnetic field in whole.
Exercise 8.1. Verify the relationship div m = 0 for vector
fields (8.4) and (8.8).
Exercise 8.2. Verify the relationship (8.19) for vector field
given by formula (8.8).
Exercise 8.3. Calculate grad f for the function (8.20).
§ 9. Differential equations
for static electromagnetic field.
In § 8 we have derived four integral equations for electric and
magnetic fields. They are used to be grouped into two pairs.
Equations in first pair have zero right hand sides:
Z I
(9.1) H, n dS = 0, E, τ ds = 0.
∂Ω ∂S
Right hand sides of equations in second pair are non-zero. They

are determined by charges and currents:
Z Z
E, n dS = 4π ρ d 3 r,
∂Ω Ω
(9.2) I Z
4π
H, τ ds = j, n dS.
c
∂S S
Applying Ostrogradsky-Gauss formula and Stokes formula, one

can transform surface integrals to spatial ones, and contour inte-
grals to surface integrals. Then, since Ω is arbitrary domain and
S is arbitrary open surface, integral equations (9.1) and (9.2) can

be transformed to differential equations:
(9.3) div H = 0, rot E = 0,

4π
(9.4) div E = 4πρ, rot H = j.
c
When considering differential equations (9.3) and (9.4), we should

add conditions for charges and currents being stationary:
∂ρ ∂j
(9.5) = 0, = 0.
∂t ∂t
The relationship (7.1) then is a consequence of (9.5) and charge

conservation law.
Differential equations (9.3) and (9.4) form complete system
of differential equations for describing stationary electromagnetic
fields. When solving them functions ρ(r) and j(r) are assumed to
be known. If they are not known, one should have some additional
equations relating ρ and j with E and H. These additional
equations describe properties of medium (for instance, continuous
conducting medium is described by the equation j = σ E, where σ
is conductivity of medium).

CHAPTER II
CLASSICAL ELECTRODYNAMICS
§ 1. Maxwell equations.
Differential equations (9.3) and (9.4), which we have derived
in the end of Chapter I, describe fields generated by stationary
charges and currents. They are absolutely unsuitable if we are
going to describe the process of haw electromagnetic interaction is
transmitted in space. Note that the notion of field was introduced
within framework of the concept of near action for describing
the object that transmit interaction of charges and currents. For
static fields this property is revealed in a very restrictive form,
i. e. we use fields only to divide interaction of charges and currents
into two processes: creation of a field by charges and currents is
first process, action of this field upon other currents and charges
is second process. Dynamic properties of the field itself appears
beyond our consideration.
More exact equations describing process of transmitting elec-
tromagnetic interaction in its time evolution were suggested by
Maxwell. They are the following ones:
1 ∂H
(1.1) div H = 0, rot E = − ,
c ∂t
4π 1 ∂E
(1.2) div E = 4πρ, rot H = j+ .
c c ∂t
44 CHAPTER II. CLASSICAL ELECTRODYNAMICS.
It is easy to see that equations (1.1) and (1.2) are generalizations

for the (9.3) and (9.4) from Chapter I. They are obtained from
latter ones by modifying right hand sides. Like equations (9.3)
and (9.4) in Chapter I, Maxwell equations (1.1) and (1.2) can be
written in form of integral equations:
Z
H, n dS = 0,
∂Ω
(1.3) I Z
1 d
E, τ ds = − H, n dS,
c dt
∂S S
Z Z
E, n dS = 4π ρ d3 r,
∂Ω Ω
(1.4) I Z Z
4π 1 d
H, τ ds = j, n dS + E, n dS.
c c dt
∂S S S
Consider contour integral in second equation (1.3). Similar

contour integral is present in second equation (1.4). However,
unlike circulation of magnetic field, circulation of electric field
I
(1.5) e= E, τ ds
∂S
possess its own physical interpretation. If imaginary contour

Γ = ∂S in space is replaced by real circular conductor, then
electric field with nonzero circulation induces electric current in
conductor. The quantity e from (1.5) in this case is called electro-
motive force of the field E in contour. Electromotive force e 6= 0
in contour produce the same effect as linking electric cell with
§ 1. MAXWELL EQUATIONS. 45
voltage e into this contour. Experimentally it reveals as follow:

alternating magnetic field produces electric field with nonzero cir-
culation, this induces electric current in circular conductor. This
phenomenon is known as electromagnetic induction. It was first
discovered by Faraday. Faraday gave qualitative description of
this phenomenon in form of the following induction law.
Faraday’s law of electromagnetic induction. Electromo-

tive force of induction in circular conductor is proportional to the
rate of changing of magnetic flow embraced by this conductor.
Faraday’s induction law was a hint for Maxwell when choosing

right hand side in second equation (1.1). As for similar term in
right hand side of second equation (1.2), Maxwell had written it
by analogy. Experiments and further development of technology
proved correctness of Maxwell equations.
Note that charge conservation law in form of relationship (5.4)
from Chapter I is a consequence of Maxwell equations. One
should calculate divergency of both sides of second equation (1.2):
4π 1 ∂ div E
div rot H = div j + ,
c c ∂t
then one should apply the identity div rot H = 0. When combined
with the first equation (1.2) this yields exactly the relationship
(5.4) from Chapter I.
Equations (1.1) and (1.2) form complete system for describ-
ing arbitrary electromagnetic fields. In solving them functions
ρ(r, t) and j(r, t) should be given, or they should be determined
from medium equations. Then each problem of electrodynam-
ics mathematically reduces to some boundary-value problem or
mixed initial-value/boundary-value problem for Maxwell equa-
tions optionally completed by medium equations. In this section
we consider only some very special ones among such problems.

Our main goal is to derive important mathematical consequences
from Maxwell equations and to interpret their physical nature.
§ 2. Density of energy and energy flow

for electromagnetic field.
Suppose that in bulk conductor we have a current with density
j, and suppose that this current is produced by the flow of charged
particles with charge q. If ν is the number of such particles per
unit volume and if v is their velocity, then j = q ν v. Recall that
current density is a charge passing through unit area per unit
time (see § 5 in Chapter I).
In electromagnetic field each particle experiences Lorentz force
determined by formula (4.4) from Chapter I. Work of this force
per unit time is equal to F, v = q E, v . Total work produced
by electromagnetic field per unit volume is obtained if one multi-
plies this quantity by ν, then w = q ν E, v = E, j . This work
increases kinetic energy of particles (particles are accelerated by
field). Otherwise this work is used for to compensate forces of
viscous friction that resist motion of particles. In either case
total power spent by electromagnetic field within domain Ω is
determined by the following integral:
Z
(2.1) W = E, j d3 r.
Ω
Let’s transform integral (2.1). Let’s express current density j

through E and H using second equation (1.2) for this purpose:
c 1 ∂E
(2.2) j= rot H − .
4π 4 π ∂t
§ 2. DENSITY OF ENERGY AND ENERGY FLOW . . . 47
Substituting this expression (2.2) into formula (2.1), we get

Z Z
c 3 1 ∂
(2.3) W = E, rot H d r − E, E d3 r.
4π 8π ∂t
Ω Ω
In order to implement further transformations in formula (2.3)

we use well-known identity div [a, b] = b, rot a − a, rot b .
Assuming a = H and b = E, for W we get
Z Z Z
c 3 c 3 d |E|2 3
W = div[H, E] d r + H, rot E d r − d r.
4π 4π dt 8π
Ω Ω Ω
First integral in this expression can be transformed by means of

Ostrogradsky-Gauss formula, while for transforming rot E one can
use Maxwell equations (1.1):
Z Z
c d |E|2 + |H|2 3
(2.4) W+ [E, H], n dS + d r = 0.
4π dt 8π
∂Ω Ω
Let’s denote by S and ε vectorial field and scalar field of the form
c |E|2 + |H|2
(2.5) S= [E, H], ε= .
4π 8π
The quantity ε in (2.5) is called density of energy of electromag-
netic field. Vector S is known as density of energy flow. It also
called Umov-Pointing vector. Under such interpretation of quan-
tities (2.5) the relationship (2.4) can be treated as the equation
of energy balance. First summand in (2.4) is called dissipation
power, this is the amount of energy dissipated per unit time at
the expense of transmitting it to moving charges. Second sum-
mand is the amount of energy that flows from within domain Ω to
outer space per unit time. These two forms of energy losses lead
to diminishing the energy stored by electromagnetic field itself
within domain Ω (see third summand in (2.4)).
Energy balance equation (2.4) can be rewritten in differential
form, analogous to formula (5.4) from Chapter I:
∂ε
(2.6) + div S + w = 0.
∂t
Here w = E, j is a density of energy dissipation. Note that in

some cases w and integral (2.1) in whole can be negative. In such
a case we have energy pumping into electromagnetic field. This
energy then flows to outer space through boundary of the domain
Ω. This is the process of radiation of electromagnetic waves from
the domain Ω. It is realized in antennas (aerials) of radio and
TV transmitters. If we eliminate or restrict substantially the
energy leakage from the domain Ω to outer space, then we would
have the device like microwave oven, where electromagnetic field
is used for transmitting energy from radiator to beefsteak.
Electromagnetic field can store and transmit not only the
energy, but the momentum as well. In order to derive momentum
balance equations let’s consider again the current with density j
due to the particles with charge q which move with velocity v.
Let ν be concentration of these particles, i. e. number of particles
per unit volume. Then j = q ν v and ρ = q ν. Total force acting
on all particles within domain Ω is given by integral
Z Z
1
(2.7) F = ρ E d3 r + [ j, H] d3 r.
c
Ω Ω
In order to derive formula (2.7) one should multiply Lorentz force

acting on each separate particle by the number of particles per
unit volume ν and then integrate over the domain Ω.
Force F determines the amount of momentum transmitted

from electromagnetic field to particles enclosed within domain Ω.
Integral (2.7) is vectorial quantity. For further transformations of
this integral let’s choose some constant unit vector e and consider
scalar product of this vector e and vector F:
Z Z
3 1
(2.8) F, e = ρ E, e d r + e, [ j, H] d3 r.
c
Ω
Substituting (2.2) into (2.8), we get

Z Z
1
F, e = ρ E, e d3 r + e, [rot H, H] d3 r −
4π
Ω Ω
(2.9) Z
1
− e, [∂E/∂t, H] d3 r.
4πc
Ω
Recalling well-known property of mixed product, we do cyclic

transposition of multiplicands in second integral (2.9). Moreover,
we use obvious identity [∂E/∂t, H] = ∂ [E, H] /∂t − [E, ∂H/∂t].
This yields the following expression for F, e :
Z Z
1
F, e = ρ E, e d3 r + rot H, [H, e] d3 r −
4π
Ω Ω
Z Z
1 d 1
− e, [E, H] d3 r + e, [E, ∂H/∂t] d3 r.
4 π c dt 4πc
Ω Ω
Now we apply second equation of the system (1.1) written as

∂H/∂t = −c rot E. Then we get formula

Z Z
d e, [E, H] 3
F, e + d r= ρ E, e d3 r +
dt 4πc
Ω Ω
(2.10) Z
rot H, [H, e] + rot E, [E, e] 3
+ d r.
4π
Ω
In order to transform last two integrals in (2.10) we use the

following three identities, two of which we already used earlier:
[a, [b, c]] = b a, c − c a, b ,

(2.11) div [a, b] = b, rot a − a, rot b ,
rot [a, b] = a div b − b div a − {a, b}.
Here by curly brackets we denote commutator of two vector

fields a and b (see [2]). Traditionally square brackets are used
for commutator, but here by square brackets we denote vector
product of two vectors. From second identity (2.11) we derive
rot H, [H, e] = div [H, [H, e]] + H, rot [H, e] .
In order to transform rot[H, e] we use third identity (2.11):

rot[H, e] = −e div H − {H, e}. Then
3
X 3
X ∂H i
H, rot [H, e] = − H, e div H + Hi ej =
i=1 j=1
∂rj
1
= − H, e div H + e, grad |H|2 .
2
Let’s combine two above relationships and apply first identity

(2.11) for to transform double vectorial product [H, [H, e]] in
first of them. As a result we obtain

rot H, [H, e] = div H H, e − div e |H|2 −
1
− H, e div H + e, grad |H|2 .
2

But div e |H|2 = e, grad |H|2 . Hence as a final result we get
rot H, [H, e] = − H, e div H +

(2.12) 1 2
+ div H H, e − e |H| .
2
Quite similar identity can be derived for electric field E:
rot E, [E, e] = − E, e div E +

(2.13) 1 2
+ div E E, e − e |E| .
2
The only difference is that due to Maxwell equations div H = 0,

while divergency of electric field E is nonzero: div E = 4πρ.
Now, if we take into account (2.12) and (2.13), formula (2.10)
can be transformed to the following one:
Z
E, e n, E + H, e n, H
F, e − dS +
4π
∂Ω
Z Z
(|E|2 + |H|2 ) e, n d e, [E, H] 3
+ dS + d r = 0.
8π dt 4πc
∂Ω Ω
Denote by σ linear operator such that the result of applying this

operator to some arbitrary vector e is given by formula
E E, e + H H, e |E|2 + |H|2
(2.14) σe = − + e.
4π 8π
Formula (2.14) defines tensorial field σ of type (1, 1) with the
following components:
|E|2 + |H|2 i E i Ej + H i Hj
(2.15) σji = δj − .
8π 4π
Tensor σ with components (2.15) is called tensor of the density
of momentum flow. It is also known as Maxwell tensor. Now let’s
define vector of momentum density p by formula
[E, H]
(2.16) p= .
4πc
In terms of the notations (2.15) and (2.16) the above relationship
for F, e is rewritten as follows:
Z Z
d
(2.17) F, e + σ e, n dS + p, e d3 r = 0.
dt
∂Ω Ω
Operator of the density of momentum flow σ is symmetric, i. e.

σ e, n = e, σ n . Due to this property and because e is
arbitrary vector we can rewrite (2.17) in vectorial form:
Z Z
d
(2.18) F+ σ n dS + p d3 r = 0.
dt
∂Ω Ω
This equation (2.18) is the equation of momentum balance for

electromagnetic field. Force F, given by formula (2.7) determines
loss of momentum stored in electromagnetic field due to transmit-

ting it to moving particles. Second term in (2.18) determines loss
of momentum due to its flow through the boundary of the domain
Ω. These two losses lead to diminishing the momentum stored
by electromagnetic field within domain Ω (see third summand in
(2.18)).
The relationship (2.18) can be rewritten in differential form.
For this purpose we should define vectorial divergency for tensorial
field σ of the type (1, 1). Let
3
X ∂σji
(2.19) µ = div σ, where µj = .
i=1
∂ri
Then differential form of (2.18) is written as

∂p
(2.20) + div σ + f = 0,
∂t
where f = ρ E + [ j, H]/c is a density of Lorentz force, while
vectorial divergency is determined according to (2.19).
Thus, electromagnetic field is capable to accumulate within
itself the energy and momentum:
Z Z
|E|2 + |H|2 3 [E, H] 3
(2.21) E= d r, P= d r.
8π 4πc
Ω Ω
It is also capable to transmit energy and momentum to material

bodies. This confirms once more our assertion that electromag-
netic field itself is a material entity. It is not pure mathematical
abstraction convenient for describing interaction of charges and
currents, but real physical object.
Exercise 2.1. Verify that relationships (2.11) hold. Check on
the derivation of (2.12) and (2.13).
§ 3. Vectorial and scalar potentials

of electromagnetic field.
In section 2 we have found that electromagnetic field possess
energy and momentum (2.21). This is very important conse-
quence of Maxwell equations (1.1) and (1.2). However we have
not studied Maxwell equations themselves. This is system of four
equations, two of them are scalar equations, other two are vecto-
rial equations. So they are equivalent to eight scalar equations.
However we have only six undetermined functions in them: three
components of vector E and three components of vector H. So
observe somewhat like excessiveness in Maxwell equations.
One of the most popular ways for solving systems of algebraic
equations is to express some variable through other ones by
solving one of the equations in a system (usually most simple
equation) and then substituting the expression obtained into other
equations. Thus we exclude one variable and diminish the number
of equations in a system also by one. Sometimes this trick
is applicable to differential equations as well. Let’s consider
Maxwell equation div H = 0. Vector field with zero divergency is
called vortex field. For vortex fields the following theorem holds
(see proof in book [3]).
Theorem on vortex field. Each vortex field is a rotor of some

other vector field.
Let’s write the statement of this theorem as applied to mag-

netic field. It is given by the following relationship:
(3.1) H = rot A.
Vector field A, whose existence is granted by the above theorem,

is called vector-potential of electromagnetic field.
§ 3. VECTORIAL AND SCALAR POTENTIALS . . . 55
Let’s substitute vector H as given by (3.1) into second Maxwell

equation (1.1). This yields the equality

1 ∂ 1 ∂A
(3.2) rot E + rot A = rot E + = 0.
c ∂t c ∂t
Vector field with zero rotor is called potential field. It is vector

field E + (∂A/∂t)/c in formula (3.2) which is obviously potential
field. Potential fields are described by the following theorem (see
proof in book [3]).
Theorem on potential field. Each potential field is a gradi-
ent of some scalar field.
Applying this theorem to vector field (3.2), we get the relation-
ship determining scalar potential of electromagnetic field ϕ:
1 ∂A
(3.3) E+ = − grad ϕ.
c ∂t
Combining (3.1) and (3.3), we can express electric and magnetic
fields E and H through newly introduced fields A and ϕ:
1 ∂A
E = − grad ϕ − ,
(3.4) c ∂t
H = rot A.
Upon substituting (3.4) into first pair of Maxwell equations

(1.1) we find them to be identically fulfilled. As for second pair of
Maxwell equations, substituting (3.4) into these equations, we get
1 ∂
− △ϕ − div A = 4 π ρ,
c ∂t
(3.5)
1 ∂ 1 ∂2A 4πj
grad div A − △A + grad ϕ + 2 2 = .
c ∂t c ∂t c
In deriving (3.5) we used relationships
div grad ϕ = △ϕ,

(3.6)
rot rot A = grad div A − △A.
Second order differential operator △ is called Laplace operator. In

rectangular Cartesian coordinates it is defined by formula
X3 2
∂ ∂2 ∂2 ∂2
(3.7) △= = + + .
i=1
∂ri ∂x2 ∂y 2 ∂z 2
In order to simplify the equations (3.5) we rearrange terms in

them. As a result we get

1 ∂2ϕ 1 ∂ 1 ∂ϕ
− △ϕ = 4 π ρ + + div A ,
c2 ∂t2 c ∂t c ∂t
(3.8)
1 ∂2A 4πj 1 ∂ϕ
− △A = − grad + div A .
c2 ∂t2 c c ∂t
Differential equations (3.8) are Maxwell equations written in

terms of A and ϕ. This is system of two equations one of
which is scalar equation, while another is vectorial equation. As
we can see, number of equations now is equal to the number of
undetermined functions in them.
§ 4. Gauge transformations and Lorentzian gauge.

Vectorial and scalar potentials A and ϕ were introduced in
§ 3 as a replacement for electric and magnetic fields E and H.
However, fields A and ϕ are not physical fields. Physical fields
E and H are expressed through A and ϕ according to formulas
(3.4), but backward correspondence is not unique, i. e. fields A

§ 4. GAUGE TRANSFORMATIONS . . . 57
and ϕ are not uniquely determined by physical fields E and H.

Indeed, let’s consider transformation
Ã = A + grad ψ,
(4.1) 1 ∂ψ
ϕ̃ = ϕ − ,
c ∂t
where ψ(r, t) is an arbitrary function. Substituting (4.1) into
formula (3.4), we immediately get
Ẽ = E, H̃ = H.
This means that physical fields E, H determined by fields Ã, ϕ̃

and by fields A, ϕ do coincide. Transformation (4.1) that do not
change physical fields E and H is called gauge transformation.
We use gauge transformations (4.1) for further simplification of
Maxwell equations (3.8). Let’s consider the quantity enclosed in
brackets in right hand sides of the equations (3.8):

1 ∂ϕ 1 ∂ ϕ̃ 1 ∂2ψ
(4.2) + div A = + div Ã + − △ψ .
c ∂t c ∂t c2 ∂t2
Denote by the following differential operator:
1 ∂2
(4.3) = − △.
c2 ∂t2
Operator (4.3) is called d’Alambert operator or wave operator.
Differential equation u = v is called d’Alambert equation.
Using gauge freedom provided by gauge transformation (4.1),
we can fulfill the following condition:
1 ∂ϕ
(4.4) + div A = 0.
c ∂t
For this purpose we should choose ψ solving d’Alambert equation

1 ∂ ϕ̃
ψ = − + div Ã .
c ∂t
It is known that d’Alambert equation is solvable under rather

weak restrictions for its right hand side (see book [1]). Hence
practically always we can fulfill the condition (4.4). This condition
is called Lorentzian gauge.
If Lorentzian gauge condition (4.4) is fulfilled, then Maxwell
equations (3.8) simplify substantially:
4πj
(4.5) ϕ = 4 π ρ, A = .
c
They look like pair of independent d’Alambert equations. How-

ever, one shouldn’t think that variables A and ϕ are completely
separated. Lorentzian gauge condition (4.4) itself is an additional
equation requiring concordant choice of solutions for d’Alambert
equations (4.5).
D’Alambert operator (4.3) is a scalar operator, in (4.5) it acts
upon each component of vector A separately. Therefore operator
commutates with rotor operator and with time derivative as
well. Therefore on the base of (3.4) we derive
4π ∂j 4π
(4.6) E = −4π grad ρ − , H = rot j.
c2 ∂t c
These equations (4.6) have no entries of potentials A and ϕ.

They are written in terms of real physical fields E and H, and
are consequences of Maxwell equations (1.1) and (1.2). However,
backward Maxwell equations do not follow from (4.6).
§ 5. ELECTROMAGNETIC WAVES. 59
§ 5. Electromagnetic waves.
In previous Chapter we considered static electromagnetic fields.
Such fields are uniquely determined by static configuration of
charges and currents (see formu-
y las (3.5) and (5.6) in Chapter I ).
They cannot exist in the absence
E0 of charges and currents. How-
A0 ever, as we shall see just below,
k x Maxwell equations have nonzero
solutions even in the case of iden-
H0
tically zero currents and charges
in the space. Let’s study one of
z such solutions. We choose some
Fig. 5.1 right-oriented rectangular Carte-
sian system of coordinates and
take some constant vector k directed along x-axis (see Fig. 5.1).
Then we choose another constant vector A0 directed along y-axis
and consider the following two functions:
(5.1) A = A0 sin(k x − ω t), ϕ = 0.
Here k = |k|. Suppose ρ = 0 and j = 0. Then, substituting (5.1)

into (4.4) and into Maxwell equations (4.5), we get
ω
(5.2) k 2 = |k|2 = .
c
It is not difficult to satisfy this condition (5.2). If it is fulfilled,
then corresponding potentials (5.1) describe plane electromag-
netic wave, ω being its frequency and k being its wave-vector,
which determines the direction of propagation of that plane wave.
Rewriting (5.1) in a little bit different form
(5.3) A = A0 sin(k(x − c t)),

we see that the velocity of propagating of plane electromagnetic

wave is equal to constant c (see (1.5) in Chapter I).
Now let’s substitute (5.1) into (3.4) and calculate electric and
magnetic fields in electromagnetic wave:
E = E0 cos(k x − ω t), E0 = |k| A0 ,

(5.4) H = H0 cos(k x − ω t), H0 = [k, A0 ],
|E0 | = |H0 | = |k| |A0 |.
Vectors k, E0 , and H0 are perpendicular to each other, they form

right triple. Wave (5.4) with such vectors is called plane linear
polarized electromagnetic wave. Vector E0 is taken for polarization
vector of this wave. Wave
E = E0 cos(k x − ω t) + H0 sin(k x − ω t),

H = H0 cos(k x − ω t) − E0 sin(k x − ω t)
is called circular polarized wave. It is superposition of two linear

polarized waves. Natural light is also electromagnetic wave. It
has no fixed polarization, however it is not circular polarized as
well. Natural light is a superposition of numerous plane linear
polarized waves with chaotically distributed polarization vectors.
§ 6. Emission of electromagnetic waves.

Plane wave (5.4) is an endless wave filling the whole space. It
is certainly kind of idealization. Real electromagnetic waves fill
only some restricted part of the space. Moreover, they are not
eternal in time: there are sources (radiators) and absorbers of
electromagnetic fields. Formula (5.4) is an approximate descrip-
tion of real electromagnetic field in that part of space which is far
apart from radiators and absorbers.
§ 6. EMISSION OF ELECTROMAGNETIC WAVES. 61
In this section we consider process of generation and radiation

of electromagnetic waves. Usually radiator is a system of charges
and currents, which is not static. We describe it by means of
functions ρ(r, t) and j(r, t). Let’s consider Maxwell equations
transformed to the form (4.5). These are non-homogeneous differ-
ential equations. Their solutions are not unique: to each solution
already found one can add arbitrary solution of corresponding
homogeneous equations. However, if we assume ρ(r, t) and j(r, t)
to be fast decreasing as r → ∞ and apply similar condition to
ϕ(r, t) and A(r, t), then we restrict substantially the freedom in
choosing solutions of the equations (4.5). In order to find one
of such solutions we need fundamental solution of d’Alambert
operator. This is distribution of the form:
c
(6.1) u(r, t) = θ(t) δ(c2 t2 − |r|2 ),
2π
where θ and δ are Heaviside theta-function and Dirac delta-
function respectively. Function (6.1) satisfies d’Alambert equation
with distribution in right hand side:
u = δ(t)δ(r).
In physics such objects are called Green functions. Knowing

fundamental solution (6.1) of d’Alambert operator, now we can
write solution for the equations (4.5) in form of contractions:
4π
(6.2) ϕ = 4π u ∗ ρ, A= u ∗ j.
c
Here ∗ denotes contraction of two distributions, see [1]. Due to
the properties of this operation from charge conservation law (see
formula (5.4) in Chapter I) we derive Lorentzian gauge condition
(4.4) for scalar and vectorial potentials (6.2). For smooth and
sufficiently fast decreasing functions ρ(r, t) and j(r, t) potentials

(6.2) are reduced to the following two integrals:
Z
ρ(r̃, t − τ ) 3
ϕ(r, t) = d r̃,
|r − r̃|
(6.3) Z
j(r̃, t − τ ) 3
A(r, t) = d r̃.
c |r − r̃|
Here the quantity τ = τ (r, r̃) is called time delay. It is determined

by the ratio τ = |r − r̃|/c. Potentials (6.3) are called retarded
potentials.
Retarded potentials have transparent physical interpretation.
Scalar potential ϕ at the point r at time instant t is a superposi-
tion of contributions from charges at various points of the space,
the contribution from the point r̃ being determined not by charge
density at present time instant t, but at previous time instant
t − τ . Time delay τ is exactly equal to the time required for the
signal spreading with light velocity c from the source point r̃ to
get to the observation point r. Similar time delay is present in
formula for vector potential A.
Note that fundamental solution of d’Alambert equation is not
unique. For example there is a solution obtained from (6.1)
by changing τ for −τ . Such solution corresponds to advanced
potentials. However, in physics advanced potentials have no
meaning, since they would break causality principle.
Let’s consider system of charges located in some small domain
Ω surrounding the origin. Let R be maximal linear size of
this domain Ω. Using formulas (6.3), we calculate we calculate
electromagnetic field of the system of charges at the point r which
is far distant from the domain Ω, i. e. |r̃| ≤ R ≪ |r|. Due to these
inequalities the ratio r̃/|r| is small vectorial quantity. Therefore
we have the following asymptotic expansions for |r − r̃| and t − τ :
r, r̃
|r − r̃| = |r| − + ... ,
|r|
(6.4)
|r| r, r̃
t−τ =t− + + ... .
c |r| c
The ratio |r|/c in (6.4) determines the time required for electro-
magnetic signal to get from the domain Ω to the observation point
|r|. Posterior terms in the series for t − τ are estimated by small
quantity R/c. This is the time of propagation of electromagnetic
signal within domain Ω.
Denote t′ = t − |r|/c and let t − τ = t′ + θ. For the quantity θ
we have the estimate |θ| ≤ R/c. Then let’s consider the following
Taylor expansions for ρ and j:
∂ρ(r̃, t′ )
ρ(r̃, t − τ ) = ρ(r̃, t′ ) + θ + ... ,
∂t
(6.5)
∂j(r̃, t′ )
j(r̃, t − τ ) = j(r̃, t′ ) + θ + ... .
∂t
The condition R ≪ |r| is not sufficient for the expansions (6.5) to
be consistent. Use of expansions (6.5) for approximating ρ(r̃, t−τ )
and j(r̃, t − τ ) is possible only under some additional assumptions
concerning these functions. Denote by T some specific time for
which functions ρ and j within domain Ω change substantially. In
case when one can specify such time T , the following quantities
are of the same order, i. e. equally large or equally small:
∂ρ ∂nρ
ρ≈T ≈ ... ≈ Tn n ,
∂t ∂t
(6.6)
∂j ∂nj
j≈T ≈ ... ≈ Tn n.
∂t ∂t

Now (6.5) can be rewritten as follows:

∂ρ(r̃, t′ ) θ
ρ(r̃, t − τ ) = ρ(r̃, t′ ) + T + ... ,
∂t T
(6.7)
∂j(r̃, t′ ) θ
j(r̃, t − τ ) = j(r̃, t′ ) + T + ... .
∂t T
Correctness of use of expansions (6.7) and (6.5) is provided by
additional condition R/c ≪ T . This yields θ/T ≪ 1.
The condition R/c ≪ T has simple meaning: the quantity
ω = 2π/T is a frequency of radiated electromagnetic waves, while
λ = 2πc/ω = c T is a wavelength. Hence condition R/c ≪ T
means that wavelength is mach greater than the size of radiator.
Suppose that both conditions R ≪ c T and R ≪ |r| are fulfilled.
Let’s calculate retarded vector potential A in (6.3) keeping only
first term in the expansion (6.5):
Z
j(r̃, t′ ) 3
(6.8) A= d r̃ + . . . .
|r| c
Ω
In order to transform integral in (6.8) let’s choose some arbitrary

constant vector e and consider scalar product A, e . Having
defined vector a and function f (r̃) by the relationships
e
a= = grad f, f (r̃) = a, r̃ ,
c|r|
we make calculations analogous to that of (7.5) in Chapter I:
Z Z Z
j, grad f d3 r̃ = div(f j) d3 r̃ − f div j d3 r̃ =
Ω Ω Ω
(6.9) Z Z Z
∂ρ 3 ∂ρ(r̃, t′ ) e, r̃ 3
= f j, n dS + f d r̃ = d r̃.
∂t ∂t |r| c
∂Ω Ω Ω
Since e is arbitrary vector, for vector potential A from (6.9) we

derive the following formula:
Z
∂ρ(r̃, t′ ) r̃ 3 Ḋ
(6.10) A= d r̃ + . . . = + ... .
∂t |r| c |r| c
Ω
Here Ḋ = Ḋ(t′ ) is time derivative of dipole moment D of the

system of charges at time instant t′ .
In a similar way, keeping only initial terms in the expansions
(6.4) and (6.5), for scalar potential ϕ in (6.3) we find
Z
ρ(r̃, t′ ) 3 Q
(6.11) ϕ= d r̃ + . . . = + ... ,
|r| |r|
Ω
where Q is total charge enclosed within domain Ω. This charge

does not depend on time since domain Ω is isolated and we have
no electric current in outer space.
Let’s compare the expressions under integration in (6.10) and
(6.11) taking into account (6.6). This comparison yields
R
|A| ≈ ϕ.
cT
The estimate R/(c T ) ≪ 1 following from R ≪ c T means that
vectorial potential is calculated with higher accuracy than scalar
potential. Hence in calculating ϕ one should take into account
higher order terms in expansions (6.4) and (6.5). Then
Z
Q ∂ρ(r̃, t′ ) r, r̃ 3
ϕ= + d r̃+
|r| ∂t |r|2 c
Ω
(6.12) Z
ρ(r̃, t′ ) r, r̃ 3
+ d r̃ + . . . .
|r| |r|2
Ω
Calculating integrals in formula (6.12), we transform it to
Q Ḋ, r D, r
(6.13) ϕ= + 2
+ + ... .
|r| |r| c |r|3
Potentials (6.10) and (6.13) are retarded potentials of the
system of charges in dipole approximation. Dependence of ρ and j
on time variable t lead to the dependence of D on t′ in them. Let’s
consider asymptotics of of these potentials as r → ∞. Thereby
we can omit last term in (6.13). Then
Q Ḋ, r Ḋ
(6.14) ϕ= + + ... , A= + ... .
|r| |r|2 c |r| c
Now on the base of formulas (3.4) and (6.14) we find
asymptotics of electric and magnetic fields at far distance from
the system of charges. In calculating rot A and grad ϕ we take
into account that t′ = t − |r|/c in argument of Ḋ(t′ ) is a quantity
depending on r. This dependence determines leading terms in
asymptotics of E and H:
[r, [r, D̈]] [r, D̈]

(6.15) E= + ... , H=− + ... .
|r|3 c2 |r|2 c2
Vectors E and H (more precisely, leading terms in their asymp-
totics) are perpendicular to each other and both are perpendicular
to vector r. This situation is similar to that of plane wave. How-
ever, in present case we deal with spherical wave being radiated
from the origin. The magnitude of fields |E| ≃ |H| decreases as
1/|r|, which is slower than for static Coulomb field. Using formula
(2.5), one can find the density of energy flow for waves (6.15):
|[r, D̈]|2
(6.16) S= r + ... .
4π |r|5 c3
For modulus of vector S we have |S| ∼ 1/|r|2 . This means that

total flow of energy through the sphere of arbitrarily large sphere
is nonzero. So we have real radiation of electromagnetic energy.
The amount of radiated energy is determined by second time
derivative of dipole moment. Therefore this case is called dipole
approximation in theory of radiation.
Exercise 6.1. Applying formula (6.16), find angular distribu-
tion of the intensity for dipole radiation. Also find total intensity
of dipole radiation.
Exercise 6.2. Particle with charge q is moving along circular
path of radius R with constant velocity v = ωR for infinitely long
time (ω is angular velocity). Calculate retarding potentials and
find angular distribution for intensity of electromagnetic radiation
of this particle. Also find total intensity of such cyclotronic radia-
tion.
Exercise 6.3. Assume that charge density ρ is zero, while cur-
rent density j is given by the following distribution:
(6.17) j(r, t) = −c [M(t), grad δ(r)]
(compare with (7.16) in Chapter I). Find retarding potentials (6.2)

for (6.17). Also find angular distribution and total intensity for
magnetic-dipole radiation induced by current (6.17).
CHAPTER III
SPECIAL THEORY OF RELATIVITY
§ 1. Galileo transformations.
Classical electrodynamics based on Maxwell equations histor-
ically was first field theory. It explained all electromagnetic
phenomena and predicted the existence of electromagnetic waves.
Later on electromagnetic waves were detected experimentally and
nowadays they have broad scope of applications in our everyday
life. However, along with successful development of this theory,
some difficulties there appeared. It was found that classical elec-
trodynamics contradicts to relativity principle. This principle in
its classical form suggested by Galileo and Newton states that
two Cartesian inertial coordinate systems moving with constant
velocity with respect to each other are equivalent. All physi-
cal phenomena in these two systems happen identically and are
described by the same laws.
Let’s consider two such Cartesian inertial coordinate systems
(r, t) and (r̃, t̃). Suppose that second system moves with velocity
u relative to first one so that coordinate axes in motion remain
parallel to their initial positions. The relation of radius-vectors of
points then can be written in form of the following transforma-
tions known as Galileo transformations:
(1.1) t = t̃, r = r̃ + ut̃.

§ 1. GALILEO TRANSFORMATIONS. 69
First relationship (1.1) means that watches in two systems are

synchronized and tick synchronously. Let r̃(t̃) be trajectory of
some material point in coordinate system (r̃, t̃). In first coordi-
nate system this trajectory is given by vector r(t) = r̃(t̃) + ut̃.
Differentiating this relationship, due to t̃ = t in (1.1) we get
∂r ∂r̃
(1.2) = + u, v = ṽ + u.
∂t ∂ t̃
Last relationship in (1.2) is known as classical law of velocity

addition. Differentiating (1.2) once more, we find the relation for
accelerations of material point in these two coordinate systems:
∂2r ∂ 2 r̃
(1.3) = , a = ã.
∂t2 ∂ t̃2
According to Newton’s second law, acceleration of material point

is determined by force F acting on it and by its mass m a = F.
From (1.3) due to relativity principle we conclude that force F
is invariant quantity. It doesn’t depend on the choice of inertial
coordinate system. This fact is represented by the relationship
(1.4) F(r̃ + ut̃, ṽ + u) = F̃(r̃, ṽ).
Now let’s consider charged particle with charge q being at

rest in coordinate system (r̃, t̃). In this coordinate system it
produces Coulomb electrostatic field. In coordinate system (r, t)
this particle is moving. Hence it should produce electric field and
magnetic field as well. This indicate that vectors E and H are
not invariant under Galileo transformations (1.1). Even if in one
coordinate system we have pure electric field, in second system we
should expect the presence of both electric and magnetic fields.
70 CHAPTER III. SPECIAL RELATIVITY.
Therefore transformation rules for E and H analogous to (1.4) for

F should be written in the following form:
E(r̃ + ut̃, t̃) = α(Ẽ(r̃, t̃), H̃(r̃, t̃), u),

(1.5)
H(r̃ + ut̃, t̃) = β(Ẽ(r̃, t̃), H̃(r̃, t̃), u).
Due to superposition principle, which is fulfilled in both coordi-

nate systems, functions α and β are linear and homogeneous with
respect to Ẽ and H̃. Therefore (1.5) is rewritten as
E(r, t) = α1 Ẽ(r̃, t̃) + α2 H̃(r̃, t̃),

(1.6)
H(r, t) = β1 Ẽ(r̃, t̃) + β2 H̃(r̃, t̃),
where α1 , α2 , β1 , β2 are some linear operators which depend on

u only. Vectors E and H determine the action of electromagnetic
field upon charges in form of Lorentz force (see formula (4.4) in
Chapter I). Substituting (1.6) into that formula and taking into
account (1.2) and (1.4), we get
q
qα1 Ẽ + qα2 H̃ + [ṽ + u, β1 Ẽ]+
(1.7) c
q q
+ [ṽ + u, β2 H̃] = q Ẽ + [ṽ, H̃].
c c
The relationship (1.7) is an identity with three arbitrary parame-
ters: ṽ, Ẽ, H̃. Therefore we can equate separately terms bilinear
with respect to ṽ and Ẽ. This yields [ṽ, β1 Ẽ] = 0, hence β1 = 0.
Now let’s equate terms bilinear with respect to ṽ and H̃. This
yields [ṽ, β2 H̃] = [ṽ, H̃]. Hence β2 = 1. And finally we should
equate terms linear with respect to H̃ and Ẽ. This yields the
following formulas for operators α1 and α2 :
1
α2 H̃ = − [u, H̃], α1 = 1.
c

§ 1. GALILEO TRANSFORMATIONS. 71
Now, if we substitute the above expressions for operators α1 , α2 ,

β1 , β2 into formula (1.6), we get the relationships
1
(1.8) E = Ẽ − [u, H̃], H = H̃.
c
The relationships (1.8) should complete Galileo transforma-

tions (1.1) in electrodynamics. However, as we shall see just
below, they cannot do this mission in non-contradictory form.
For this purpose, let’s transform Maxwell equations written as
(1.1) and (1.2) in Chapter II to coordinate system (r̃, t̃). For
partial derivatives due to transformations (1.1) we have
3
X
∂ ∂ ∂ ∂ ∂
(1.9) = i, = − uk .
∂ri ∂r̃ ∂t ∂ t̃ k=1 ∂r̃k
Now, combining (1.8) and (1.9), we derive
div H = div H̃,

1
div E = div Ẽ + u, rot H̃ ,
c
rot H = rot H̃
1 1
rot E = rot Ẽ + {u, H̃} − u div H̃,
c c
∂H ∂ H̃
= − {u, H̃},
∂t ∂ t̃
∂E ∂ Ẽ 1 1
= − {u, Ẽ} + [u, {u, H̃}] − [u, ∂ H̃/∂ t̃ ].
∂t ∂ t̃ c c
Here by curly brackets we denote commutator of vector fields (see

[2]). Thereby vector u is treated as constant vector field.
When substituting the above expressions into Maxwell equa-

tions we consider the case of zero charges and currents: ρ = 0,
j = 0. This yields the following equations:
div H̃ = 0,
1
div Ẽ = − u, rot H̃ ,
c
1 ∂ Ẽ 1
rot H̃ = − {u, Ẽ}+
c ∂ t̃ c
1 1
+ 2 [u, {u, H̃}] − 2 [u, ∂ H̃/∂ t̃ ],
c c
1 ∂ H̃
rot Ẽ = − .
c ∂ t̃
Only two of the above four equations coincide with original

Maxwell equations. Other two equations contain the entries
of vector u that cannot be eliminated.
This circumstance that we have found is very important. In the
end of XIX-th century it made a dilemma for physicists. The way
how this dilemma was resolved had determined in most further
development of physics in XX-th century. Indeed, one had to
make the following crucial choice:
(1) to admit that Maxwell equations are not invariant with
respect to Galileo transformations, hence they require the
existence of some marked inertial coordinate system where
they have standard form given in the very beginning of
Chapter II;
(2) or to assume that formulas (1.1) are not correct, hence
relativity principle claiming equivalence of all inertial co-
ordinate systems is realized in some different way.
§ 2. LORENTZ TRANSFORMATIONS. 73
First Choice had lead to ether theory. According to this theory,

marked inertial coordinate system is bound to some hypothetical
matter, which was called ether. This matter has no mass, no
color, and no smell. It fills the whole space and does not reveal
itself otherwise, but as a carrier of electromagnetic interaction.
Specified properties of ether look quite unusual, this makes ether
theory too artificial (not natural). As a compromise this theory
was admitted for a while, but later was refuted by experiments of
Michaelson and Morley, who tried to measure the Earth velocity
relative to ether (ether wind).
Second choice is more crucial. Indeed, refusing formulas (1.1),
we refuse classical mechanics of Newton in whole. Nevertheless
the development of science went through this second choice.
§ 2. Lorentz transformations.
Having refused formulas (1.1), one should replace them by
something else. This was done by Lorentz. Following Lorentz,
now we replace Galileo transformations (1.1) by general linear
transformations relating (r, t) and (r̃, t̃):
3
X 3
X
(2.1) c t = S00 c t̃ + Sk0 r̃k , ri = S0i c t̃ + Ski r̃k .
k=1 k=1
In (2.1) we introduced c as a factor for time variables t and t̃ in

order to equalize measure units. Upon introducing this factor all
components of matrix S appear to be purely numeric quantities
that do not require measure units at all. It is convenient to denote
c t by r0 and treat this quantity as additional (fourth) component
of radius-vector:
(2.2) r0 = ct.
Then two relationships (2.1) can be united into one relationship:
3
X
(2.3) ri = Ski r̃k .
k=0
In order to have invertible transformation (2.3) one should assume

that det S 6= 0. Let T = S −1 . Then inverse transformation for
(2.3) is written as follows:
3
X
(2.4) r̃i = Tki rk .
k=0
By their structure transformation (2.3) and (2.4) coincide with

transformations of coordinates of four-dimensional vector under
the change of base. Soon we shall see that such interpretation
appears to be very fruitful.
Now the problem of deriving Lorentz transformations can be
formulated as problem of finding components of matrix S in (2.3).
The only condition we should satisfy thereby is the invariance
of Maxwell equations with respect to transformations (2.3) upon
completing them with transformations for ρ, j, E and H.
For the beginning let’s consider the case with no currents and
charges, i. e. the case ρ = 0, j = 0. Instead of Maxwell equations
let’s study their differential consequences written in form of the
equations (4.6) in Chapter II:
(2.5) E = 0, H = 0.
Invariance of (2.5) under the transformations (2.3) and (2.4) is

necessary (but possibly not sufficient) condition for invariance of
Maxwell equations from which the equations (2.5) were derived.
§ 2. LORENTZ TRANSFORMATIONS. 75
Further we need the following formula for d’Alambert operator

used in the above equations (2.5):
3 X
X 3
∂ ∂
(2.6) = g ij .
i=0 j=0
∂ri ∂rj
Here by g ij we denote components of matrix

 
1 0 0 0
0 −1 0 0 
(2.7) g ij = gij =  .
0 0 −1 0
0 0 0 −1
It is easy to see that inverse matrix gij for (2.7) has the same
components, i. e. gij = g ij .
From (2.3) and (2.4) we derive the following transformation
rules for first order differential operators:
X 3 X 3
∂ ∂ ∂ ∂
(2.8) = Tik k , = Sik k .
∂ri ∂r̃ ∂r̃i ∂r
k=0 k=0
Substituting (2.8) into formula (2.6), we get
3 X
X 3
∂ ∂
= g̃ pq ,
p=0 q=0
∂r̃p ∂r̃q
where matrices g ij and g̃ pq are related by formula
3 X
X 3
(2.9) g̃ pq = Tip Tjq g ij .
i=0 j=0
In terms of inverse matrices gpq and g̃pq this relationship (2.9) can
be rewritten as follows:
3 X
X 3
(2.10) gij = Tip Tjq g̃pq .
p=0 q=0
Theorem 2.1. For any choice of operator coefficients α1 , α2 ,

β1 , and β2 in formulas (1.6) the invariance of the form of equations
(2.5) under the transformations (2.3) and (2.4) is equivalent to
proportionality of matrices g and g̃, i. e.
(2.11) g̃ ij = λ g ij .
Numeric factor λ in formula (2.11) is usually chosen to be equal

to unity: λ = 1. In this case from (2.10) and (2.11) we derive
3 X
X 3
(2.12) gij = Tip Tjq gpq .
p=0 q=0
In matrix form this relationship (2.12) looks like
(2.13) T t g T = g.
Here g is a matrix of the form (2.7), while by T t in (2.13) we

denote transposed matrix T .
Definition 2.1. Matrix T satisfying the relationship (2.13) is
called Lorentzian matrix.
It is easy to check up that the set of Lorentzian matrices form
a group. This group is usually denoted by O(1, 3). It is called
matrix Lorentz group.
§ 3. MINKOWSKY SPACE. 77
From the relationship (2.13) for Lorentzian matrix we derive

the equality (det T )2 = 1. Hence det T = ±1. Lorentzian matrices
with unit determinant form the group SO(1, 3), it is called special
If i = j = 0, from (2.12) we obtain the following formula
relating components of Lorentzian matrix T :
(2.14) (T00 )2 − (T01 )2 − (T02 )2 − (T03 )2 = 1.
Inequality |T00 | > 1 is immediate consequence of the relationship

(2.14). Hence T00 > 1 or T00 6 −1. Lorentzian matrix with T00 > 1
is called orthochronous. The set of orthochronous Lorentzian ma-
trices form orthochronous matrix Lorentz group O+ (1, 3). Intersec-
tion SO+ (1, 3) = SO(1, 3)∩O+ (1, 3) is called special orthochronous
Exercise 2.1. Prove theorem 2.1 under the assumption that

transformation (1.6) given by operator coefficients α1 , α2 , β1 , and
β2 is invertible.
§ 3. Minkowsky space.
In previous section we have found that each Lorentzian matrix
from group O(1, 3) determines some transformation (2.1) pre-
serving the form the equations (2.5). In deriving this fact we
introduced notations (2.2) and united space and time into one
four-dimensional “space-time”. Let’s denote it by M . Four-
dimensional space M is basic object in special theory of relativity.
Its points are called events. The space of events is equipped with
quadratic form g with signature (1, 3). This quadratic form is
called Minkowsky metric. Thereby inertial coordinate systems are
interpreted as Cartesian coordinates for which Minkowsky metric
has canonical form (2.7).

Equivalence principle. All physical laws in any two inertial

coordinate systems are written in the same form.
Let’s choose some inertial coordinate system. This choice
determines separation of event space M into geometric space V
(space of points) and time axis T :
(3.1) M = T ⊕ V.
Matrix of Minkowsky metric in chosen coordinate system has

canonic form (2.7). Therefore geometric space V is orthogonal to
time axis T with respect to Minkowsky metric g. Restriction of
this metric to V is negative quadratic form. Changing its sign,
we get positive quadratic form. This is standard Euclidean scalar
product in V .
Now let’s consider another inertial coordinate system. Like
(3.1), it determines second expansion of M into space and time:
(3.2) M = T̃ ⊕ Ṽ .
In general time axes T and T̃ in expansions (3.1) and (3.2) do

not coincide. Indeed, bases of these two coordinate systems are
related to each other by formula
3
X
(3.3) ẽi = Sij ej ,
j=0
where S is Lorentzian matrix from (2.3). For base vector ẽ0

directed along time axis T̃ from (3.3) we derive
(3.4) ẽ0 = S00 e0 + S01 e1 + S02 e2 + S03 e3 .
In general components S01 , S02 , and S03 in Lorentz matrix S are

nonzero. Therefore vectors ẽ0 and e0 are non-collinear. Hence
T 6= T̃ .
Non-coincidence of time axes T 6= T̃ for two inertial coordinate

systems leads to non-coincidence of geometric spaces: V 6= Ṽ .
This fact lead to quite radical conclusion when we interpret
it physically: observers in two such inertial systems observe two
different three-dimensional geometric spaces and have two different
time ticks. However, in our everyday life this difference is very
small and never reveals.
Let’s calculate how big is the difference in the rate of time ticks
for two inertial coordinate systems. From (2.4) we get
3
X T0
(3.5) t̃ = T00 t + k
rk .
c
k=1
Let t → +∞. If Lorentzian matrix T is orthochronous, then

T00 > 0 and t̃ → +∞. If matrix T is not orthochronous, then
t → +∞ we get t̃ → −∞. Transformations (2.4) with non-
orthochronous matrices T invert the direction of time exchanging
the future and the past. It would be very intriguing to have such
a feature in theory. However, presently in constructing theory of
relativity one uses more realistic approach. So we shall assume
that two physically real inertial coordinate systems can be related
only by orthochronous Lorentz matrices from O+ (1, 3).
Restriction of the set of admissible Lorentz matrices from
O(1, 3) to O+ (1, 3) is due to the presence of additional structure
in the space of events. It is called polarization. Let’s choose some
physical inertial coordinate system. Minkowsky metric in such
system is given by matrix of canonical form (2.7). Let’s calculate
scalar square of four-dimensional vector x in Minkowsky metric:
(3.6) g(x, x) = (x0 )2 − (x1 )2 − (x2 )2 − (x3 )2 .
By value of their scalar square g(x, x) in Minkowsky metric g

vectors of Minkowsky space M are subdivided into three parts:

(1) tome-like vectors, for which g(x, x) is positive;
(2) light vectors, for which g(x, x) = 0;
(3) space-like vectors, for which g(x, x) is negative.
Coordinates of light vectors satisfy the following equation:
(3.7) (x0 )2 − (x1 )2 − (x2 )2 − (x3 )2 = 0.
It is easy to see that (3.7) is the equation of cone in four-

dimensional space (see classification of quadrics in [4]). This cone
(3.7) is called light cone.
Time-like vectors fill interior of
light cone, while space-like vectors
future
fill outer space outside this cone.
Interior of light cone is a union of
two parts: time-like vectors with
x0 > 0 are directed to the future,
others with x0 < 0 are directed to
the past. Vector directed to the
past future can be continuously trans-
formed to any other vector directed
to the future. However, it cannot
Fig. 3.1 be continuously transformed to a
vector directed to the past without
making it space-like vector or zero vector at least once during
transformation. This means that the set of time-like vectors is
disjoint union of two connected components.
Definition 3.1. Geometric structure in Minkowsky space M

that marks one of two connected components in the set of time-
like vectors is called polarization. It is used to say that marked
component points to the future.
Let e0 , e1 , e2 , e3 be orthonormal base in Minkowsky metric*.

In the space M with polarization one can consider only those
such bases for which unit vector of time axis e0 is directed to the
future. Then transition from one of such bases to another would
be given by orthochronous matrix from group O+ (1, 3).
Definition 3.2. Four-dimensional affine space M equipped

with metric g of signature (1, 3) and equipped with orientation**
and polarization is called Minkowsky space.
According to special theory of relativity Minkowsky space,

which is equipped with orientation and polarization, is proper
mathematical model for the space of real physical events. Now
we can give strict mathematical definition of inertial coordinate
system.
Definition 3.3. Orthonormal right inertial coordinate system

is orthonormal right coordinate system in Minkowsky space with
time base vector directed to the future.
It is easy to verify that any two inertial coordinate systems as

defined above are related to each other by Lorentz transforma-
tion with matrix S from orthochronous Lorentz group SO+ (1, 3).
Let’s choose one of such coordinate systems and consider related
expansion (3.1). It is clear that e0 ∈ T , while linear span of
spatial vectors e1 , e2 , e3 defines subspace V . Taking orthonormal
base e1 , e2 , e3 for the standard of right bases in V , we equip
this three-dimensional space with orientation. This is concor-
dance with the fact that geometric space that we observe in our
everyday life possesses orientation distinguishing left and right.
* i. e. base for which Minkowsky metric has the form (2.7).

** remember that orientation is geometric structure distinguishing left and
right bases (see [4]).
Exercise 3.1. By analogy with definition 3.3 formulate the

definition of skew-angular inertial coordinate system.
§ 4. Kinematics of relative motion.

Galileo transformations are used in mechanics for describing
physical processes as they are seen by two observers representing
two inertial coordinate systems. Lorentz transformations, which
we have derived from the condition of invariance of electrodynam-
ical equations (2.5), are designed for the same purpose. However,
this is not immediately clear when looking at formulas (2.3) and
(2.4). Therefore we shall bring these formulas to the form more
convenient for studying their physical nature.
Let’s fix two inertial coordinate systems related by Lorentz
transformation (2.1). First one is related with orthonormal base
e0 , e1 , e2 , e3 in Minkowsky space and with the expansion (3.1).
Second is related with the base ẽ0 , ẽ1 , ẽ2 , ẽ3 and with the
expansion (3.2). If time axes are parallel e0 = ẽ0 , then Lorentz
matrix S in (2.3) is reduced to orthogonal matrix O ∈ SO(3)
relating two right orthonormal bases e1 , e2 , e3 and ẽ1 , ẽ2 , ẽ3 . It
has the following blockwise-diagonal shape:
1 0 0 0 
 0 O11 O21 O31 
(4.1) S=  0 O2 O2 O2  .

1 2 3
0 O13 O23 O33
Thus, in case if T k T̃ two inertial coordinate systems differ only
in directions of spatial axes. They do not move with respect to
each other.
Now let’s consider the case T ∦ T̃ . Hence e0 6= ẽ0 . Let H be
linear span of vectors e0 and ẽ0 . Denote by W intersection of
subspaces V and Ṽ from (3.1) and (3.2):
(4.2) H = Span(e0 , ẽ0 ), W = V ∩ Ṽ .
§ 4. KINEMATICS OF RELATIVE MOTION. 83
Lemma 4.1. Two-dimensional subspaces H and W in (4.2) are

perpendicular to each other in Minkowsky metric g. Their inter-
section is zero: H ∩ W = {0}, while direct sum of these subspaces
coincides with the whole Minkowsky space: H ⊕ W = M .
Proof. Subspace H is two-dimensional since it is linear span
of two non-collinear vectors. Subspaces V and Ṽ are three-
dimensional and V 6= Ṽ . Hence their sum V + Ṽ coincides with
M , i. e. dim(V + Ṽ ) = 4. Applying theorem on the dimension of
sum and intersection of two subspaces (see [4]), we get
dim(W ) = dim V + dim Ṽ − dim(V + Ṽ ) = 3 + 3 − 4 = 2.

In order to prove orthogonality of H and W we use orthogonal-
ity of T and V in the expansion (3.1) and orthogonality of T̃ and
Ṽ in (3.2). Let y be an arbitrary vector in subspace W . Then
y ∈ V . From V ⊥ T we get y ⊥ e0 . Analogously from y ∈ Ṽ we
get y ⊥ ẽ0 . Now from orthogonality of y to both vectors e0 and
ẽ0 we derive orthogonality of y to their linear span: y ⊥ H. Since
y is arbitrary vector in W , we have W ⊥ H.
Now let’s prove that H ∩ W = {0}. Let’s consider an arbitrary
vector x ∈ H ∩ W . From x ∈ H and x ∈ W due to orthogonality
of H and W , which is already proved, we get g(x, x) = 0. But
x ∈ W ⊂ V , while restriction of Minkowsky metric to subspace
V is negative quadratic form of signature (0, 3). Therefore from
g(x, x) = 0 we derive x = 0. Proposition H ∩ W = {0} is proved.
From H ∩ W = {0} we conclude that sum of subspaces H and
W is direct sum and dim(H +W ) = 2+2 = 4. Hence H ⊕W = M .
Lemma is proved.
Now let’s return back to considering pair of inertial coordinate
systems with bases e0 , e1 , e2 , e3 and ẽ0 , ẽ1 , ẽ2 , ẽ3 . There is the
expansion (3.4) for vector ẽ0 . Let’s write it as follows:
(4.3) ẽ0 = S00 e0 + v.
Here v = S01 e1 + S02 e2 + S03 e3 ∈ V . Since matrix S is or-

thochronous and since ẽ0 6= e0 , we have
(4.4) S00 > 1, v 6= 0.
For any real number a > 1 there exists a number α > 0 such that
a = cosh(α). Let’s apply this observation to S00 in (4.3):
(4.5) S00 = cosh(α).
From (4.3), from (4.5), and from orthogonality of vectors e0 and

v in Minkowsky metric we obtain
1 = g(ẽ0 , ẽ0 ) = (S00 )2 g(e0 , e0 ) + g(v, v) = cosh2 (α) − |v|2 .
Using this equality we can find Euclidean length of vector v in

three-dimensional subspace V :
(4.6) |v| = sinh(α), where α > 0.
Let’s replace vector v by vector of unit length h1 = v/|v| and

rewrite the relationship (4.3) as follows:
(4.7) ẽ0 = cosh(α) e0 + sinh(α) h1 .
Due to (4.7) vector h1 is linear combination of vectors e0 and ẽ0 ,

hence h1 ∈ H. But h1 ∈ V as well. Therefore h1 ∈ V ∩H. Vectors
e0 and h1 are perpendicular to each other, they form orthonormal
base in two-dimensional subspace H:
(4.8) g(e0 , e0 ) = 1, g(h1 , h1 ) = −1.
From (4.8) we conclude that restriction of Minkowsky metric to

subspace H is metric with signature (1, 1).

Now we need another vector from subspace H. Let’s determine

it by the following relationship:
(4.9) h̃1 = sinh(α) e0 + cosh(α) h1 .
It is easy to check that vectors ẽ0 and h̃1 form another orthonor-
mal base in subspace H. Transition matrix relating these two
bases has the following form:
!
cosh(α) sinh(α)
(4.10) SL = .
sinh(α) cosh(α)
Matrix (4.10) is called the matrix of Lorentzian rotation.

There is four-dimensional version of matrix (4.10). Indeed,
vector h1 ∈ V is perpendicular to subspace W ⊂ V . Therefore we
have the expansion of subspace V as a direct sum:
V = Span(h1 ) ⊕ W.
Let’s choose two vectors h2 and h3 forming orthonormal base

in subspace W and complementing h1 up to an orthonormal
right base in V . Then four vectors e0 , h1 , h2 h3 constitute
orthonormal right base in M with time vector e0 directed to
the future. Transition matrix relating this base with the base
ẽ0 , h̃1 , h2 h3 has the following blockwise-diagonal form:
 
cosh(α) sinh(α) 0 0
 
 sinh(α) cosh(α) 0 0 
 
(4.11) SL = 

.

 0 0 1 0 
 
0 0 0 1
Transition from base e0 , e1 , e2 e3 to base e0 , h1 , h2 h3 is given

by a matrix of the form (4.1). This is because their time vectors
do coincide. In a similar way transition from base ẽ0 , h̃1 , h2 h3
to base ẽ0 , ẽ1 , ẽ2 ẽ3 is given by a matrix of the same form
(4.1). Ultimate change of base e0 , e1 , e2 e3 for another base
ẽ0 , ẽ1 , ẽ2 ẽ3 then can be done in three steps.
Theorem 4.1. Each Lorentzian matrix S ∈ SO+ (1, 3) is a
product of three matrices S = S1 SL S2 , one of which SL is a
matrix of Lorentzian rotation (4.11), while two others S1 and S2
are matrices of the form (4.1).
In order to clarify physical meaning of Lorentz transformations
let’s first consider transformations with matrix S of the form
(4.11). Let ct = r0 , r1 , r2 , r3 be coordinates of some vector
r ∈ M in the base e0 , h1 , h2 , h3 . By ct̃ = r̃0 , r̃1 , r̃2 , r̃3 we denote
coordinates of the same vector in the base ẽ0 , h̃1 , h2 h3 . For
matrix S of the form (4.11) formula (2.3) leads to relationships
sinh(α) 1
t = cosh(α) t̃ + r̃ ,
c
(4.12) r1 = sinh(α) c t̃ + cosh(α) r̃1 ,
r2 = r̃2 ,
r3 = r̃3 .
Let r̃1 , r̃2 , r̃3 be coordinates of radius-vector of some point A

which is at rest in inertial coordinate system with base ẽ1 , ẽ2 , ẽ3 .
Then r̃1 , r̃2 , r̃3 are constants, they do not depend on time t̃ in
this coordinate system. Upon calculating coordinates of this point
A in other inertial coordinate system by means of formulas (4.12)
its first coordinate r1 appears to be a function of parameter t̃.
We use first relationship (4.12) in order to express parameter t̃
through time variable t in second coordinate system:
t tanh(α) 1
(4.13) t̃ = − r̃ .
cosh(α) c
Substituting (4.13) into other three formulas (4.12), we get
r1 = r1 (t) = c tanh(α) t + const,

(4.14) r2 = r2 (t) = const,
r3 = r3 (t) = const .
From (4.14) we see that in second coordinate system our point A

is moving with constant velocity u = c tanh(α) in the direction of
first coordinate axis.
In contrast to parameter α in matrix (4.11), parameter u
has transparent physical interpretation as magnitude of relative
velocity of one coordinate system with respect to another. Let’s
express components of matrix (4.11) through u:
1 u 1
cosh(α) = r , sinh(α) = r .
u2 c u2
1− 1− 2
c2 c
Let’s substitute these formulas into (4.12). As a result we get
u
t̃ + 2 r̃1
c ut̃ + r̃1
t= r , r1 = r ,
u2 u2
1− 2 1− 2
c c
(4.15)
r2 = r̃2 , r3 = r̃3 .
Denote for a while by r and r̃ the following two three-

dimensional vectors in subspaces V and Ṽ :
r = r1 h1 + r2 h2 + r3 h3 ,
(4.16)
r̃ = r̃1 h̃1 + r̃2 h2 + r̃3 h3 .
Then we define linear map θ : V → Ṽ determined by its action

upon base vectors h1 , h2 , and h3 :
θ(h1 ) = h̃1 , θ(h2 ) = h2 , θ(h3 ) = h3 .
This map θ is orientation preserving isometry, since it maps right

orthonormal base of subspace V to right orthonormal base in
subspace Ṽ . Using above notations (4.16) and the map θ, we can
write formulas (4.15) in vectorial form:
θu, r̃
t̃ +
c2
t= r ,
|u|2
1− 2
c
(4.17)
θu, r̃
θu t̃ + θu
|u|2 θu, r̃
θr = r + r̃ − θu.
|u|2 |u|2
1−
c2
Here u = u h1 is vector of relative velocity of second coordinate
system with respect to first one. Formulas (4.17) are irrespective
to the choice of bases in subspaces V and Ṽ . Therefore they are
applicable to Lorentz transformations with special matrix of the
form (4.11) and to arbitrary Lorentz transformations with matrix
S = S1 SL S2 (see theorem 4.1).
Very often the sign of map θ realizing isomorphism of subspaces

V and Ṽ in formulas (4.17) is omitted:
u, r̃
t̃ +
c2
t= r ,
|u|2
1− 2
c
(4.18)
u, r̃
u t̃ + u
|u|2 u, r̃
r= r + r̃ − u.
|u|2 |u|2
1− 2
c
Formulas (4.18) represent “conditionally three-dimensional” un-
derstanding of Lorentz transformations when vectors r and r̃
treated as vectors of the same three-dimensional Euclidean space,
while t and t̃ are treated as scalar parameters. However, accord-
ing to modern paradigm four-dimensional Minkowsky space is real
physical entity, not purely mathematical abstraction convenient
for shortening formulas (compare (2.3) and (4.17)).
When writing formulas (4.17) and (4.18) in components we
should expand vectors r and u in the base of one coordinate sys-
tem, while vector r̃ is expanded in the base of another coordinate
system. Thereby the difference in the shape of these two formulas
completely disappears.
Exercise 4.1. Using expansions (4.16) for vectors r and r̃, de-
rive the following formulas:
θu, r̃ θu, r̃
r̃1 = , r̃2 h2 + r̃3 h3 = r̃ − θu.
|u| |u|2
Combining these formulas with (4.15), derive formulas (4.17).

§ 5. Relativistic law of velocity addition.

Classical law of velocity addition was first consequence that we
obtained from Galileo transformations:
(5.1) v = ṽ + u,
see formulas (1.2). Replacing Galileo transformations by Lorentz

transformations, now we should derive new relativistic law of
velocity addition.
Suppose that vector-function r̃(t̃) describes the motion of a
point A in inertial coordinate system (r̃, t̃) and suppose that this
coordinate system moves with velocity u with respect to other
inertial coordinate system (r, t). For passing to coordinate system
(r, t) we use Lorentz transformation given by formulas (4.18). As
a result we get two functions
u, r̃(t̃)
t̃ +
c2
t(t̃) = r ,
|u|2
1− 2
c
(5.2)
u, r̃(t̃)
u t̃ + u
|u|2 u, r̃(t̃)
r(t̃) = r + r̃(t̃) − u.
|u|2 |u|2
1−
c2
Let’s calculate first derivatives of functions (5.2):
u, ṽ
1+
dt c2
(5.3a) = r ,
dt̃ |u|2
1− 2
c
§ 6. WORLD LINES AND PRIVATE TIME. 91
u, ṽ
u+ u
dr |u|2 u, ṽ
(5.3b) = r + ṽ − u.
dt̃ |u| 2 |u|2
1− 2
c
By ṽ we denote the velocity of the point A in coordinates (r̃, t̃):
˙ t̃) = dr̃ .
ṽ = r̃(
dt̃
In a similar way by v we denote the velocity of this point in other
coordinates (r, t). To calculate v we divide derivatives:

dr dr dt
(5.4) v= = ṙ(t) = .
dt dt̃ dt̃
Substituting (5.3a) and (5.3b) into (5.4), we get formula
u, ṽ u, ṽ
u+ u ṽ − ur
|u|2 |u|2 |u|2
(5.5) v= + 1− 2 .
u, ṽ u, ṽ c
1+ 2
1+ 2
c c
Formula (5.5) is relativistic law of velocity addition. It is mach
more complicated than classical law given by formula (5.1). How-
ever, in the limit of small velocities |u| ≪ c formula (5.5) reduces
to formula (5.1).
Exercise 5.1. Derive relativistic law of velocity addition from
formula (4.17). Explain why resulting formula differs from (5.5).
§ 6. World lines and private time.

Motion of point-size material object in arbitrary inertial coor-
dinate system (r, t) is described by vector-function r(t), where t

is time variable and r is radius-vector of point. Four-dimensional

radius-vector of this material point has the following components:
(6.1) r0 (t) = ct, r1 (t), r2 (t), r3 (t).
Vector-function with components (6.1) determines parametric line

in Minkowsky space M , this line is called world line of material
point. Once world line is given, motion of material point is
described completely. Let’s differentiate four-dimensional radius-
vector (6.1) with respect to parameter t. As a result we get
four-dimensional vector tangent to world line:
(6.2) K = (c, ṙ1 , ṙ2 , ṙ3 ) = (c, v 1 , v 2 , v 3 ).
Last three components of this vector form velocity vector of

material point. Velocity of most material objects is not greater
than light velocity: |v| < c. When applied to vector K in (6.2)
this means that tangent-vector of world line is time-like vector:
(6.3) g(K, K) = c2 − |v|2 > 0.
Definition 6.1. Smooth curve in Minkowsky space is called

time-like curve if tangent-vector of this curve is time-like vector
at each its point.
World lines for most material objects are time-like curves.

Exception are world lines of photons (light particles) and world
lines of other elementary particles with zero mass. For them
|v| = c, hence we get g(K, K) = 0.
World line have no singular points. Indeed, even if g(K, K) =
0, tangent vector K in (6.2) is nonzero since K 0 = c 6= 0.
Let’s consider world line of material point of nonzero mass.
§ 6. WORLD LINES AND PRIVATE TIME. 93
For this line we have the condition (6.3) fulfilled, hence we can
introduce natural parameter on this line:
Zt p
(6.4) s(t) = g(K, K) dt.
t0
Integral (6.4) yields invariant parameter for world lines. For any
two points A and B on a given world line the quantity s(B)−s(A)
does not depend on inertial coordinate system used for calculating
integral (6.4). This quantity is called interval length of the arc
AB on world line.
Theorem 6.1. Straight line segment connecting end points of
an arc on smooth time-like curve is a segment of time-like straight
line. Its interval length is greater than interval length of corre-
sponding arc.
Let A and B be two successive events in the “life” of material
point of nonzero mass. The answer to the question what time in-
terval separates these two events depend on the choice of inertial
coordinate system from which we observe the “life” of this ma-
terial point. So this answer is relative (not invariant). However,
there is invariant quantity characterizing time distance between
two events on world line:
s(B) − s(A)
(6.5) τ= .
c
This quantity τ in formula (6.5) is called interval of private time
on world line.
Concept of private time determine microlocal concept of time
in theory of relativity. According to this concept each material
point lives according to its own watch, and watches of different
material points are synchronized only in very rough way: they
count time from the past to the future. This rough synchroniza-
tion is determined by polarization in Minkowsky space. Exact
synchronization of watches is possible only when material points
come to immediate touch with each other, i. e. when their world
lines intersect. However, even after such exact synchronization
in the point of next meeting watches of tho material points will
show different times. This difference is due to different ”life
paths” between two meetings.
Concept of private time is illustrated by so-called twins prob-
lem, well-known from science fiction. Suppose that one of twins
goes to far-away travel in interstellar spacecraft, while his brother
stays on the Earth. Which of them will be older when they meet
each other on the Earth in the end of space voyage.
The answer is: that one who stayed on the Earth will be older.
World lines of twins intersect twice. Both intersections occur on
the Earth, one before travel and other after travel. Its known that
Coordinate system associated with the Earth can be taken for
inertial coordinate system with high degree of accuracy (indeed,
acceleration due to rotation of the Earth around its axis and due
to orbital rotation around the Sun is not sensible in our everyday
life). Therefore world line of twin stayed on the Earth is straight
line. World line of twin in spacecraft is curved. In the beginning
of travel he accelerates reaching substantial velocity comparable
with light velocity in the middle of the path. Then he experiences
backward acceleration in order to brake before reaching target
point of his travel. Then he accelerates and brakes again in his
back way to the Earth. According to theorem 6.1 interval length
of curved world line connecting two events is shorter than interval
length of straight world line connecting the same two events.
Hence twin stayed on the Earth will be older.
Exercise 6.1. Remember proof of the fact that the length of

curved line connecting two points A and B in Euclidean space is
§ 7. DYNAMICS OF MATERIAL POINT. 95
greater than the length of straight line segment AB. By analogy

to this proof find the proof for theorem 6.1.
§ 7. Dynamics of material point.

Motion of material point in theory of relativity is described by
its world line in Minkowsky space. Let’s choose natural parameter
on world line and consider four-dimensional tangent vector
dr(s)
(7.1) u(s) = ,
ds
where r(s) is four-dimensional radius vector of events on world
line. Vector u is (7.1) is called vector of 4-velocity. It is time-like
vector and it is unit vector in Minkowsky metric: g(u, u) = 1.
Upon choosing some inertial coordinate system we can write
components of 4-velocity vector explicitly:
c
1 v1
(7.2) u= p .
c2 − |v|2 v2
v3
Here v 1 , v 2 , v 3 are components of three-dimensional velocity
vector v. Note that components u0 , u1 , u2 , u3 of 4-velocity vector
are absolute numbers (without measure unit). It is easy to see
from (7.2). Upon multiplying u by scalar mc with the measure
unit of momentum we get vector of 4-momentum
c
m v1
(7.3) p= r
|v|2 v2
1− v3
c2
for material point with mass m. Vector p plays important role
in physics since there is fundamental law of nature: the law of
conservation of 4-momentum.
Momentum conservation law. Vector of 4-momentum of

material point which do not experience external action remains
unchanged.
According to the law just stated, for particle that do not
experience external action we have p = const. Hence u = const.
Integrating the equation (7.1), for r(s) we derive
r(s) = r0 + u s.
Conclusion: in the absence of external action material point moves

uniformly along straight line.
External actions causing change of 4-momentum of material
point are subdivided into two categories:
(1) continuous;
(2) discrete.
Continuous actions are applied to material particle by external
fields. They cause world line to bend making it curved line. In
this case p 6= const. Derivative of 4-momentum with respect to
natural parameter s is called vector of 4-force:
dp
(7.4) = F(s).
ds
Vector of 4-force in (7.4) is quantitative characteristic of the

action of external fields upon material particle. It is determined
by parameters of particle itself and by parameters of external
fields at current position of particle as well. We know that
vector of 4-velocity u is unit vector, therefore g(p, p) = m2 c2 .
Differentiating this relationship with respect to s and taking into
account that components of matrix (2.7) are constant, we find
(7.5) g(u, F) = 0.
The relationship (7.5) means that vector of 4-force is perpendicu-

lar to vector of 4-velocity in Minkowsky metric, i. e. force vector
is perpendicular to world line of particle.
Choosing some inertial coordinate system, we can replace nat-
ural parameter s in (7.5) by time variable t of this coordinate
system. Then, taking into account (7.3), from (7.4) we derive
dpi p
(7.6) = c2 − |v|2 F i , where i = 1, 2, 3.
dt
Now, if we p
denote by f three-dimensional vector with compo-
nents f i = c2 − |v|2 F i , then for three-dimensional vector of
momentum from (7.6) we obtain differential equation
dp
(7.7) = f.
dt
The equation (7.7) is treated as relativistic analog of Newton’s
second law. Instead of classical formula p = mv relating momen-
tum and velocity vectors here we have the following relationship:
mv
(7.8) p= r .
|v|2
1− 2
c
In order to write (7.8) in classical form we introduce the quantity
m
(7.9) mv = r .
|v|2
1− 2
c
Constant m is called mass at rest, while mv in (7.8) is called
dynamic mass of moving particle. Now p = mv v, and Newton’s
second law is written as follows:
(7.10) (mv v)′t = f .

Formulas (7.9) and (7.10) are the very ones which are in mind
when one says that mass in theory of relativity depends on
velocity. It seems to me that such terminology is not so good.
In what follows we shall mostly use four-dimensional invariant
equation (7.4) and, saying mass, we shall imply mass at rest.
Discrete external actions appear
p̃2 in those situations when 4-momen-
p̃1 tum of material particle changes
abruptly in jump-like manner. Such
p̃n
situation arise in particle collisions,
particle confluence, and particle de-
p1 cay. Collision of particles corre-
p k spond to that point in Minkowsky
p2
space where world lines of two or
several particles come together. Af-
Fig. 7.1 ter collision particles can simply fly
out from that point. But if these
are molecules of ingredients in chemical reaction, then after col-
lision we would have new molecules of reaction products. In a
similar way in collisions of atomic nuclei nuclear reactions occur.
Let’s consider simultaneous collision of k particles. Denote by
p1 , . . . , pk their 4-momenta just before the collision. Suppose that
as a result of collision he have n new particles created from initial
ones. Denote by p̃1 , . . . , p̃n 4-momenta of outgoing particles just
after the collision. If k = 1 this is particle decay process, while if
n = 1 we have particle confluence into one composite particle.
Momentum conservation law. Total 4-momentum of ingo-
ing particles before collision is equal to total 4-momentum of out-
going particles after collision:
k
X n
X
(7.11) pi = p̃i .
i=1 i=1

As an example we consider process of frontal collision of two

identical particles of mass m leading to creation of one particle of
mass M . Suppose that velocities of initial particles are equal by
magnitude but opposite to each other:
c c
m v1 m −v 1
p1 = r , p2 = r .
|v|2 v2 |v|2 −v 2
1− 2 v3 1− −v 3
c c2
For 4-momentum of resulting particle we have
c
M w1
p̃1 = r .
|w|2 w2
1− w3
c2
Applying momentum conservation law (7.11) to this situation, we
get w = 0 and additionally we obtain
2m
(7.12) M=r .
|v|2
1− 2
c
From (7.12) we see that mass at rest of resulting composite
particle is greater than sum of rest masses of its components:
M > m + m. Conclusion: the low of mass conservation is fulfilled
approximately only in the limit of small velocities |v| ≪ c.
Let’s multiply zeroth component of 4-momentum of material
particle by c. Resulting quantity has the measure unit of energy.
Let’s denote this quantity by E:
mc2
(7.13) E=r .
|v|2
1− 2
c
The quantity (7.13) is called kinetic energy of moving particle.

Writing relationship (7.11) for zeroth components of colliding
particles, we get energy conservation law:
k
X n
X
(7.14) Ei = Ẽi .
i=1 i=1
Thus, 4-momentum conservation law for collision includes both

energy conservation law (7.14) and the law of conservation for
three-dimensional momentum.
Note that for zero velocity v = 0 the above quantity (7.13)
does not vanish, but takes nonzero value
(7.15) E = mc2 .
This quantity is known as rest energy of material particle. For-

mula (7.15) is well-known. It reflects very important fact absent
in classical physics: the ability of energy to mass and mass to
energy conversion. In practice conversion of energy to mass is
realized in particle confluence (see M > m + m in formula (7.12)).
Converse phenomenon of particle decay yields mass defect (mass
decrease). Lost mass is realized in additional amount of kinetic
energy of outgoing particles. Total conversion of mass to en-
ergy is also possible. This happens in process of annihilation,
when elementary particle meets corresponding antiparticle. Large
amount of energy released in annihilation is scattered in form of
short-wave electromagnetic radiation.
§ 8. Four-dimensional form of Maxwell equations.

Starting from electromagnetic equations E = 0 and H = 0
in previous sections we have constructed and described Lorentz
transformations preserving form of these equations. We also have
§ 8. FOUR-DIMENSIONAL FORM . . . 101
given geometric and physical interpretation of Lorentz transfor-

mations and even have described dynamics of material points on
the base of new relativistic notion of space and time. Now time
has come to remember that equations E = 0 and H = 0
are not primary equations of electrodynamics, they were derived
from Maxwell equations. To have complete picture we should we
should write Maxwell equations in four-dimensional form. Let’s
begin with second pair of these equations containing charges and
currents (see equations (1.2) in Chapter II. Let’s modify them:
1 ∂E 4π
− rot H = − j, − div E = −4πρ.
c ∂t c
Then rewrite these equations in components using Levi-Civita

symbol for to express rotor (see [3]):
3 3
∂E p X X ∂H k 4π
0
− εpqk q = − j p ,
∂r q=1
∂r c
k=1
(8.1)
3
X ∂E q
− = −4π ρ.
q=1
∂rq
Here we used notation r0 = ct associating time variable with

zeroth component of radius-vector in Minkowsky space.
Using Levi-Civita symbol and components of vector H, we can
construct skew-symmetric 3 × 3 matrix with elements
3
X
(8.2) F pq = − εpqk H k .
k=1
Due to (8.2) we can easily write explicit form of matrix F :

 
0 −H 3 H2
(8.3) F pq =  H3 0 −H 1  .
−H 2 H1 0
Let’s complement the above matrix (8.3) with one additional line
and one additional column:
 0 −E 1 −E 2 −E 3 
 E1 0 −H 3 H2 
 
(8.4) F pq =  2 .
E H3 0 −H 1 
E3 −H 2 H1 0
Additional line and additional column in (8.4) are indexed by

zero, i. e. indices p and q run over integer numbers from 0 to 3.
In addition, we complement three-dimensional vector of current
density with one more component
(8.5) j 0 = ρc.
By means of (8.4) and (8.5) we can rewrite Maxwell equations

(8.1) in very concise four-dimensional form:
3
X ∂F pq 4π p
(8.6) =− j .
q=0
∂rq c
Now let’s consider first pair of Maxwell equations (see equations

(1.1) in Chapter II). In coordinates they are written as
3 3 3
∂H p X X ∂E k X ∂H q
(8.7) + ε pqk = 0, = 0.
∂r0 q=1 ∂rq q=1
∂rq
k=1
The structure of the equations (8.7) is quite similar to that of

(8.1). However, their right hand sides are zero and we see slight
difference in signs. Main difference is that components of vectors
E and H have exchanged their places. To exchange components
of vectors E and H in matrix (8.4) we need four-dimensional
analog of Levi-Civita symbol:


 0, if among p, q, k, s there



 are at least two equal num-



 bers;
εpqks = εpqks = 1, if (p q k s) is even permuta-



 tion of numbers (0 1 2 3);



 −1, if (p q k s) is odd permuta-


tion of numbers (0 1 2 3).
Let’s define matrix G by the following formula for its components:
3 3 3 3
1 X X X X pqks
(8.8) Gpq = − ε gkm gsn F mn .
2 s=0 m=0 n=0
k=0
Here g is matrix (2.7) determining Minkowsky metric. Matrix G

with components (8.8) can be expressed in explicit form:
 0 −H 1 −H 2 −H 3 
 H1 0 E3 −E 2 
 
(8.9) Gpq = 2 .
H −E 3 0 E1 
3 2
H E −E 1 0
The structure of matrix (8.9) enable us to write remaining two

Maxwell equations (8.7) in concise four-dimensional form:
3
X ∂Gpq
(8.10) = 0.
q=0
∂rq
Usage of both matrices F and G in theory is assumed to be too

excessive. For this reason equations (8.10) are written as
3 X
X 3 X
3
∂Fks
(8.11) εpqks = 0.
q=0 k=0 s=0
∂rq
Matrix Fks is obtained from F mn by means of standard index

lowering procedure using matrix (2.7):
3 X
X 3
(8.12) Fks = gkm gsn F mn .
m=0 n=0
Four-dimensional form of Maxwell equations (8.6) and (8.11)

gives a hint for proper geometric interpretation of these equations.
Matrix (8.4) defines tensor of the type (2, 0) in Minkowsky space.
This tensor is called tensor of electromagnetic field. Tensorial
interpretation of matrix (8.4) immediately yields transformation
rules, which were lacking so far:
3 X
X 3
(8.13) F pq = p
Sm Snq F̃ mn .
m=0 n=0
These relationships (8.13) determine transformation rules for com-

ponents of vectors E and H. Before now we express these rules in
undetermined form by the relationships (1.6). For special Lorentz
matrices (4.11) vectors of electric and magnetic fields E and H in

two inertial coordinate systems are related as follows:
u u
Ẽ 2 + H̃ 3 Ẽ 3 − H̃ 2
c c
E 1 = Ẽ 1 , E2 = r , E3 = r ,
u2 u2
1− 2 1− 2
c c
u u
H̃ 2 − Ẽ 3 H̃ 3 + Ẽ 2
c c
H 1 = H̃ 1 , H2 = r , H3 = r .
u2 u2
1− 2 1− 2
c c
According to theorem 4.1, general Lorentz matrix is a product
of special Lorentz matrix of the form (4.11) and two matrices of
spatial rotation in tree-dimensional space. The latter ones can be
excluded if one writes Lorentz transformation in “conditionally
three-dimensional” vectorial form:
u, Ẽ 1
Ẽ − u − [u, H̃]
u, Ẽ |u|2 c
E= u+ r ,
|u|2 |u|2
1− 2
c
(8.14)
u, H̃ 1
H̃ − u + [u, Ẽ]
u, H̃ |u|2 c
H= u+ r .
|u|2 |u|2
1− 2
c
From (8.13) we derive the following rule for transforming co-
variant components of the tensor of electromagnetic field:
3 X
X 3
(8.15) Fpq = Tpm Tqn F̃mn .
m=0 n=0

This relationship (8.15) provides invariance of the form of Maxwell

equations (8.11) under Lorentz transformation (2.3). In order to
verify this fact it is sufficient to apply relationships (2.8) for
transforming derivatives and then remember well-known property
of four-dimensional Levi-Civita symbol εpqks :
3 X
X 3 X
3 X
3
(8.16) Tap Tbq Tck Tds εabcd = det T εpqks .
a=0 b=0 c=0 d=0
The condition of invariance of Maxwell equations (8.6) with re-

spect to Lorentz transformations leads to the following transfor-
mation rule for components of four-dimensional current density:
3
X
(8.17) jp = p m
Sm j̃ .
m=0
In (8.17) it is easy to recognize the transformation rule for com-

ponents of four-dimensional vector. In case of special Lorentz
matrix of the form (4.11), taking into account (8.5), one can write
the above relationship (8.17) as follows:
u
ρ̃ + 2 j̃ 1
c uρ̃ + j̃ 1
ρ= r , j1 = r ,
u2 u2
1− 2 1− 2
c c
(8.18)
j 2 = j̃ 2 , j 3 = j̃ 3 .
Remember that here u = c tanh(α) is a magnitude of relative

velocity of one inertial coordinate system with respect to another.
§ 9. FOUR-DIMENSIONAL VECTOR-POTENTIAL. 107
In vectorial form relationships (8.18) are written as
u, j̃
ρ̃ +
c2
ρ= r ,
|u|2
1− 2
c
(8.19)
u, j̃
u ρ̃ + u
|u|2 u, j̃
j= r + j̃ − 2
u.
|u| 2 |u|
1− 2
c
In such form they give transformation rule for charge density ρ

and three-dimensional current density j under Lorentz transfor-
mations with arbitrary Lorentz matrix.
Exercise 8.1. Prove the relationship (8.16), assuming T to be
an arbitrary 4 × 4 matrix.
Exercise 8.2. Using (2.12), derive the relationship (8.15) from
(8.12) and (8.13).
Exercise 8.3. Using (8.15), (8.16) and (2.8), transform Max-
well equations (8.11) from one inertial coordinate system to an-
other. Verify that the form of these equations is invariant.
Exercise 8.4. Using (8.13), (8.17) and (2.8), transform Max-
well equations (8.6) from one inertial coordinate system to another.
Verify that the form of these equations is invariant.
§ 9. Four-dimensional vector-potential.
Due to special structure of Maxwell equations one can intro-
duce vector-potential A and scalar potential ϕ. This was done in
§ 3 of Chapter II. Here are formulas for components of E and H:
∂ϕ 1 ∂Ap
Ep = − p
− ,
∂r c ∂t
(9.1) 3 X
X 3
∂Ak
Hp = εpqk ,
q=1 k=1
∂rq
(see formulas (3.4) in Chapter II). Denote A0 = ϕ and consider

four-dimensional vector A with components A0 , A1 , A2 , A3 . This
is four-dimensional vector-potential of electromagnetic field. By
lowering index procedure we get covector A:
3
X
(9.2) Ap = gpq Aq .
q=0
Taking into account relationships (2.7) for components of matrix

gpq , from formula (9.2) we derive
A0 = A0 , A1 = −A1 ,
(9.3)
A2 = −A2 , A3 = −A3 .
Moreover, let’s write explicitly covariant components for the ten-

sor of electromagnetic field:
 0 E1 E2 E3 
 −E 1 0 −H 3 H2 
 
(9.4) Fpq = .
 −E 2 H3 0 −H 1 
−E 3 −H 2 H1 0
Due to (9.3) and (9.4) first relationship (9.1) can be written as
∂Aq ∂A0
(9.5) F0q = − .
∂r0 ∂rq
In order to calculate other components of tensor Fpq let’s ap-
ply (8.2) and second relationship (9.1). Thereby let’s take into
account that Fpq = F pq and Ap = −Ap for p, q = 1, 2, 3:
3
X 3 X
X 3 X
3
∂An
(9.6) Fpq = − εpqk H k = εpqk εkmn .
∂rm
k=1 k=1 m=1 n=1
Further transformation of (9.6) is based on one of the well-known

contraction identities for Levi-Civita symbol:
3
X
(9.7) εpqk εkmn = δpm δqn − δqm δpn .
k=1
Applying (9.7) to (9.6), we get

3 X
X 3
∂An ∂Aq ∂Ap
(9.8) Fpq = (δpm δqn − δqm δpn ) m
= p
− .
m=1 n=1
∂r ∂r ∂rq
Combining (9.8) and (9.5), we obtain the following formula for all
covariant components of the tensor of electromagnetic field:
∂Aq ∂Ap
(9.9) Fpq = − .
∂rp ∂rq
In essential, formula (9.9) is four-dimensional form of the rela-
tionships (9.1). It unites these two relationships into one.
Remember that vectorial and scalar potentials of electromag-
netic field are not unique. They are determined up to a gauge
transformation (see formula (4.1) in Chapter II). This uncertainty

could be included into transformation rule for component of four-
dimensional potential A. However, if we assert that A0 , A1 , A2 ,
A3 are transformed as components of four-dimensional vector
3
X
(9.10) Ap = Sqp Ãq ,
q=0
and A0 , A1 , A2 , A3 are obtained from them by index lowering

procedure (9.2), then we find that quantities Fpq defined by
formula (9.9) are transformed exactly by formula (8.15), as they
actually should.
From (9.10) one can easily derive explicit transformation
formulas for scalar potential ϕ and for components of three-
dimensional vector-potential A. For special Lorentz transforma-
tions with matrix (4.11) they are written as follows:
u u
ϕ̃ + Ã1 ϕ̃ + Ã1
c c
ϕ= r , A1 = r ,
u2 u2
1− 2 1− 2
c c
(9.11)
A2 = Ã2 , A3 = Ã3 .
Note that one can rederive transformation rules for components
of electric and magnetic fields (see § 8 above). However, we shall
not do it now.
In case of Lorentz transformations with arbitrary Lorentz ma-
trix the relationships (9.11) should be written in vectorial form:
u, j̃
ϕ̃ +
c
(9.12a) ϕ= r ,
|u|2
1− 2
c
u u, Ã
ϕ̃ + u
c |u|2 u, Ã
(9.12b) A= r + Ã − u.
|u|2 |u|2
1− 2
c
Theorem 9.1. Each skew-symmetric tensor field F of type
(0, 2) in four-dimensional space satisfying differential equations
(8.11) is determined by some covector field A according to the
above formula (9.9).
Proof. Each skew-symmetric tensor field F of type (0, 2)
in four-dimensional space can be identified with pair of three-
dimensional vector fields E and H depending on additional pa-
rameter r0 = ct. In order to do this one should use (9.4). Then
equations (8.11) are written as Maxwell equations for E and H:
1 ∂H
div H = 0, rot E = − .
c ∂t
Further construction of covector field A is based on considerations
from § 3 of Chapter II, where three-dimensional vector-potential
and scalar potential were introduced. Then we denote A0 = ϕ
and thus convert three-dimensional vector-potential into four-
dimensional vector. And the last step is index lowering procedure
given by formula (9.2).
Choice of vector field A in formula (9.9), as we noted above,
has gauge uncertainty. In four-dimensional formalism this fact is
represented by gauge transformations
∂ψ
(9.13) Ak → Ak + ,
∂rk
where ψ — is some arbitrary scalar field. Formula (9.13) is four-
dimensional version of gauge transformations (4.1) considered
in Chapter II. It is easy to verify that gauge transformations

(9.13) do not break transformation rules (9.10) for contravariant
components of vector A.
Exercise 9.1. Prove theorem 9.1 immediately in four-dimen-
sional form without passing back to three-dimensional statements
and constructions.
§ 10. The law of charge conservation.

Earlier we have noted that charge conservation law can be
derived from Maxwell equations (see § 1 in Chapter II). To prove
this fact in four-dimensional formalism is even easier. Let’s
differentiate the relationship (8.6) with respect to rp and add one
more summation with respect to index p:
X3 X 3 3
∂ 2 F pq 4π X ∂j p
(10.1) = − .
p=0 q=0
∂rp ∂rq c p=0 ∂rp
Double differentiation in (10.1) is symmetric operation, while

tensor of electromagnetic field F pq , to which it is applied, is
skew-symmetric. Therefore the expression under summation in
left hand side of (10.1) is skew-symmetric with respect to indices
p and q. This leads to vanishing of double sum in left hand side
of formula (10.1). Hence we obtain
3
X ∂j p
(10.2) = 0.
p=0
∂rp
The equality (10.2) is four-dimensional form of charge conserva-

tion law. If we remember that j 0 = cρ and r0 = ct, we see that
this equality coincides with (5.4) in Chapter I.
Conservation laws for scalar quantities (those like electric char-
ge) in theory of relativity are expressed by equations analogous

§ 10. THE LAW OF CHARGE CONSERVATION. 113
to (10.2) in form of vanishing of four-dimensional divergencies for

corresponding four-dimensional currents. For vectorial quantities
corresponding current densities are tensors. Thus the law of con-
servation of 4-momentum for fields is represented by the equation
3
X ∂T qp
(10.3) = 0.
p=0
∂rp
Tensor T qp in (10.3) playing the role of current density of 4-mo-

mentum is called energy-momentum tensor.
Theorem 10.1. For any vector field j in n-dimensional space
(n > 2) if its divergency is zero
n
X ∂j p
(10.4) = 0,
p=1
∂rp
then there is skew-symmetric tensor field ψ of type (2, 0) such that
n
X ∂ψ pq
(10.5) jp = .
q=1
∂rq
Proof. Choosing some Cartesian coordinate system, we shall

construct matrix ψ pq of the following special form:
 0 ... 0 ψ 1n 
 .. .. .. .. 
 . . . . 
(10.6) ψ pq
=

.

 0 ... 0 ψ n−1 n 
−ψ 1n . . . −ψ n−1 n 0
Matrix (10.6) is skew-symmetric, it has (n − 1) independent com-

ponents. From (10.5) for these components we derive
∂ψ kn
= j k , where k = 1, . . . , n − 1,
∂rn
(10.7) n−1
X ∂ψ kn
= −j n .
∂rk
k=1
Let’s define functions ψ kn in (10.7) by the following integrals:
n
Zr
ψ kn = j k (r1 , . . . , rn−1 , y) dy+
0
(10.8)
Zrk
1
+ j n (r1 , . . . , y, . . . , rn−1 , 0) dy.
n−1
0
It is easy to verify that functions (10.8) satisfy first series of

differential equations (10.7). Under the condition (10.4) they
satisfy last equation (10.7) as well. Thus, theorem is proved.
Theorem 10.1 can be easily generalized for arbitrary tensorial

currents. Its prove thereby remains the same in most.
Theorem 10.2. For any tensorial field T of type (m, s) in the

space of dimension n > 2 if its divergency is zero
Xn
∂Tqp11...
... pm
qs
pm
= 0,
p =1
∂r
m
§ 11. NOTE ON COORDINATES. 115
then there is tensorial field ψ of type (m + 1, s) skew-symmetric

in last pair of upper indices and such that
n
X p ... p p
∂ψq11 ... qsm m+1
Tqp11...
... pm
qs = .
pm+1 =1
∂rpm+1
Exercise 10.1. Verify that the equation (10.4) provides last

equation (10.7) to be fulfilled for the functions (10.8).
Exercise 10.2. Clarify the relation of theorem 10.1 and theo-

rem on vortex field in case of dimension n = 3.
§ 11. Note on skew-angular and

curvilinear coordinates.
In previous three sections we have managed to write in four-
dimensional form all Maxwell equations, charge conservation law,
and the relation of E, H and their potentials. The relationships
(8.6), (8.11), (9.9), (9.13), (10.2), which were obtained there, pre-
serve their shape when we transfer from one rectangular Cartesian
coordinate system to another. Such transitions are interpreted as
Lorentz transformations, they are given by Lorentz matrices.
However, all these relationships (8.6), (8.11), (9.9), (9.13), (10.2)
possess transparent tensorial interpretation. Therefore they can
be transformed to any skew-angular Cartesian coordinate system
as well. Thereby we would have minor differences: the shape
of matrix g would be different and instead of εpqks in (8.11) we
would require volume tensor with components
p
(11.1) ω pqks = ± − det ĝ εpqks .
Matrix gpq in skew-angular coordinate system is not given by

formula (2.7), here it is arbitrary symmetric matrix determining
quadratic form of signature (1, 3). Therefore differential equations
E = 0 and H = 0, which we are started from, have not their
initial form. They are written as F pq = 0, where d’Alambert
operator is given by formula (2.6) with non-diagonal matrix g ij .
In arbitrary skew-angular coordinate system none of axes
should have time-like direction. Therefore none of them can
be interpreted as time axis. Three-dimensional form of electrody-
namics equations, even if we could write them, would not have
proper physical interpretation in such coordinate system. In par-
ticular, interpretation of components of tensor F pq as components
of electric and magnetic fields in formula (8.4) would not be
physically meaningful.
Tensorial form of four-dimensional electrodynamics equations
enables us to make one more step toward increasing arbitrariness
in the choice of coordinate system: we can use not only skew-
angular, but curvilinear coordinates as well. To make this step we
need to replace partial derivatives by covariant derivatives:
∂
(11.2) → ∇p
∂rp
(see [3] for more details). Connection components required for

passing to covariant derivatives (11.2) are determined by compo-
nents of metric tensor. The latter ones in curvilinear coordinate
system do actually depend on r0 , r1 , r2 , r3 :
3
1 X ks ∂gsj ∂gis ∂gij
(11.3) Γkij = g + − .
2 s=0 ∂ri ∂rj ∂rs
No we give list of all basic equations, which we derived above, in

§ 11. NOTE ON COORDINATES. 117
covariant form. Maxwell equations are written as follows:
3
X 4π p
∇q F pq = − j ,
q=0
c
(11.4)
3 X
X 3 X
3
ω pqks ∇q Fks = 0.
q=0 k=0 s=0
Here components of volume tensor ω pqks are given by formula

(11.1). Tensor of electromagnetic field is expressed through four-
dimensional vector-potential by formula
(11.5) Fpq = ∇p Aq − ∇q Ap ,
while gauge uncertainty in the choice of vector-potential itself is

described by the relationship
(11.6) Ak → Ak + ∇k ψ,
where ψ is arbitrary scalar field. Charge conservation law in

curvilinear coordinates is written as
3
X
(11.7) ∇p j p = 0.
p=0
Instead of formula (2.6) for D’Alambert operator here we have
3 X
X 3
(11.8) = g ij ∇i ∇j .
i=0 j=0
Dynamics of material point of nonzero mass m 6= 0 is described

by ordinary differential equations of Newtonian type:
F
(11.9) ṙ = u, ∇s u = .
mc
Here dot means standard differentiation with respect to natural
parameter s on world line, while ∇s is covariant derivative with
respect to the same parameter.
Exercise 11.1. Using symmetry of Christoffel symbols (11.3)
with respect to lower pair of indices i and j, show that the re-
lationship (11.5) can be brought to the form (9.9) in curvilinear
coordinate system as well.
CHAPTER IV
LAGRANGIAN FORMALISM
IN THEORY OF RELATIVITY
§ 1. Principle of minimal action

for particles and fields.
Dynamics of material points in theory of relativity is described
by their world lines. These are time-like lines in Minkowsky space.
Let’s consider some world line corre-
B sponding to real motion of some particle
under the action of external fields. Let’s
fix two points A and B on this world
line not too far from each other. Then
consider small deformation of world line
A in the range between these two points
A and B. Suppose that we have some
coordinate system in Minkowsky space
Fig. 1.1 (either Cartesian, or curvilinear, no mat-
ter). Then our world line is given in
parametric form by four functions
(1.1) r0 (s), r1 (s), r2 (s), r3 (s),
where s is natural parameter. Then deformed curve can be given

120 CHAPTER IV. LAGRANGIAN FORMALISM . . .
by the following four functions:
(1.2) r̂i (s) = ri (s) + hi (ε, s), i = 0, . . . , 3.
Here s is original natural parameter on initial non-deformed world

line (1.1), while hi (ε, s) are smooth functions which are nonzero
only within the range between points A and B. Note that
functions hi (ε, s) in (1.2) depend on additional parameter ε which
is assumed to be small. Moreover, we shall assume that
(1.3) hi (ε, s) → 0 as ε → 0.
Thus, in (1.2) we have whole family of deformed lines. This family

of lines is called variation of world line (1.1). Due to (1.3) we
have the following Taylor expansion for hi (ε, s):
(1.4) hi (ε, s) = ε hi (s) + . . . .
Under the change of one curvilinear coordinate system for an-

other quantities hi (s) are transformed as components of four-
dimensional vector. This vector is called vector of variation of
world line, while quantities
(1.5) δri (s) = εhi (s)
are called variations of point coordinates. It is clear that they

also are transformed as components of four-dimensional vector.
Due to formulas (1.4) and (1.5) parametric equations of deformed
curves (1.2) are written as follows:
(1.6) r̂i (s) = ri (s) + δri (s) + . . . .
By this formula we emphasize that terms other than linear with

respect to small parameter ε are of no importance.
§ 1. PRINCIPLE OF MINIMAL ACTION . . . 121
By varying functions hi (ε, s) in (1.2) and by varying parameter

ε in them we can surround segment of initial world line by a
swarm of its variations. Generally speaking, these variations do
not describe real dynamics of points. However, they are used
in statement of minimal action principle. Within framework of
Lagrangian formalism functional of action S is usually introduced,
this is a map that to each line connecting two points A and B put
into correspondence some real number S.
Principle of minimal action for particles. World line con-
necting two points A and B describes real dynamics of material
point if and only if action functional S reaches local minimum on
it among other lines being its small variations.
Action functional S producing number by each line should de-
pend only on that line (as geometric set of points in M ), but
it should not depend on coordinate system (r0 , r1 , r2 , r3 ) in M .
By tradition this condition is called Lorentz invariance, though
changes of one curvilinear coordinate system by another form
much broader class of transformations than Lorentz transforma-
tions relating two rectangular Cartesian coordinate systems in
Minkowsky space.
Action functional in most cases is integral. For single point of
mass m in electromagnetic field with potential A it is written as
Zs2 Zs2
q
(1.7) S = −mc ds − g(A, u) ds.
c
s1 s1
Here q is electric charge of particle, while u = u(s) is vector of

its 4-velocity (unit tangent vector of world line). First integral
in (1.7) yields action for free particle (in the absence of exter-
nal fields), second integral describes interaction of particle with
electromagnetic field.
If we consider system of N particles, then we should write

integral (1.7) for each of them and we should add all these
integrals. And finally, in order to get the action functional for
total system of field and particles we should add integral of action
for electromagnetic field itself:
 
XN Zs2 (i) s2 (i)
Z
 qi 
S= −mi c ds − g(A, u) ds −
i=1
c
s1 (i) s1 (i)
(1.8)
ZV2X
3 X
3 p
1
− Fpq F pq − det g d4 r.
16 π c
V1 p=0 q=0
Last integral in (1.8) deserves special consideration. This is

four-dimensional volume integral over the domain enclosed be-
future tween two three-dimensional hyper-
surfaces V1 and V2 . Hypersurfaces
V2 V1 and V2 are space-like, i. e. their
normal vectors are time-like vectors.
V1 These hypersurfaces determine the
fissure between the future and the
past past, and over this fissure we inte-
grate in (1.8). Thereby change of
Fig. 1.2 field functions (here these are com-
ponents of vector-potential A) when
passing from V1 to V2 reflects evolution of electromagnetic field
from the past to the future.
Electromagnetic field is described by field functions. Therefore
variation of field is defined in other way than that of particles.
Suppose that Ω is some restricted four-dimensional domain en-
closed between hypersurfaces V1 and V2 . Let’s consider four
smooth functions hi (ε, r) = hi (ε, r0 , r1 , r2 , r3 ) being identically
§ 1. PRINCIPLE OF MINIMAL ACTION . . . 123
zero outside the domain Ω and vanishing for ε = 0. Let’s define
(1.9) Âi (r) = Ai (r) + hi (ε, r)
and consider Taylor expansion of hi at the point ε = 0:
(1.10) hi (ε, r) = ε hi (r) + . . . .
The following functions determined by linear terms in the above

Taylor expansions (1.10)
(1.11) δAi (r) = ε hi (r)
are called variations of field functions for electromagnetic field.

Deformation of vector-potential (1.9) now can be written as
(1.12) Âi (r) = Ai (r) + δAi (r) + . . . .
Principle of minimal action for fields. Field functions de-

termine actual configuration of physical fields if and only if they
realize local minimum of action functional S in class of all varia-
tions with restricted support*.
The condition of minimum of action for actual field configura-
tion and for actual world lines of particles, as a rule, is not used.
In order to derive dynamical equations for fields and particles it is
sufficient to have extremum condition (no matter minimum, max-
imum, or saddle point). For this reason minimal action principle
often is stated as principle of extremal action.
Exercise 1.1. Verify that hi (s) in (1.4) are transformed as
components of vector under the change of coordinates.
* Variations with restricted support are those which are identically zero
outside some restricted domain Ω.
Exercise 1.2. Prove that under gauge transformations (11.6)

from Chapter III action functional (1.8) is transformed as follows:
N
X
qi qi
(1.13) S→S− ψ(r(s2 (i))) − ψ(r(s1 (i))) .
i=1
c c
Explain why terms added to action functional in (1.13) are not

sensitive to variation of word lines (1.2).
§ 2. Motion of particle in electromagnetic field.

In order to find world line of relativistic particle in external
electromagnetic field we shall apply particle version of extremal
action principle to functional (1.8). Let’s choose one of N par-
ticles in (1.8) and consider deformation (1.6) of its world line.
When we substitute deformed world line into (1.8) in place of
initial non-deformed one the value of last integral remains un-
changed. Thereby in first term containing sum of integrals only
one summand changes its value, that one which represent the par-
ticle we have chosen among others. Therefore writing extremity
condition for (1.8) we can use action functional in form of (1.7).
The value of (1.7) for deformed world line is calculated as follows:
Zs2p Zs2
q
(2.1) Sdef = −mc g(K, K) ds − g(A, K) ds.
c
s1 s1
Formula (2.1) visually differs from formula (1.7) because s is

natural parameter on initial world line, but it is not natural
parameter on deformed line. Here tangent vector
dr̂(s) dĥ(s)
(2.2) K(s) = = u(s) + ε + ...
ds ds
§ 2. MOTION OF PARTICLE . . . 125
is not unit vector. Therefore first integral (1.7) is rewritten as

length integral (see (6.4) in Chapter III). In second integral (1.7)
unit tangent vector is replaced by vector K.
Let’s write in coordinate form both expressions which are under
integration in (2.1) taking into account that we deal with general
curvilinear coordinate system in Minkowsky space:
v
u 3 3
p uX X
g(K, K) = t gij (r̂(s)) K i (s) K j (s),
i=0 j=0
(2.3)
3
X
g(A, K) = Ai (r̂(s)) K i (s).
i=0
Let’s substitute (2.2) into (2.3) and take into account (1.2) and
the expansion (1.4). As a result for the expressions (2.3) we get
the following power expansions with respect to small parameter ε:
p p 3
X
ε dhi (s)
g(K, K) = g(u, u) + p ui (s) +
g(u, u) i=0
ds
3 3 3
!
1 X X X ∂gij
+ ui (s) uj (s) hk (s) + ... ,
2 i=0 j=0 k=0
∂rk
3
X dhi (s)
g(A, K) = g(A, u) + ε Ai (r(s)) +
i=0
ds
X3 X3
∂Ai i
+ε k
u (s) hk (s) + . . . .
i=0
∂r
k=0
When substituting these expansions into (2.1) we should remem-

ber that u is unit vector. Then for Sdef we get
Zs2X
3 dhk (s)
q
Sdef = S − ε m c uk (s) + Ak (r(s)) ds −
c ds
s1 k=0
Zs2X
3 3 3 3
!
q X ∂Ai i mc X X ∂gij i j k
−ε u + u u h (s) ds + . . . .
c i=0 ∂rk 2 i=0 j=0 ∂rk
s1 k=0
Let’s apply integration by parts to first integral above. As a result

we get the expression without derivatives of functions hk (s):
s2
3
X
q
Sdef = S − ε m c uk (s) + Ak (r(s)) hk (s) +
c
k=0 s1
Zs2X
d
3
q
+ε m c uk (s) + Ak (r(s)) hk (s) ds −
ds c
s1 k=0
Zs2X
3 3 3 3
!
q X ∂Ai i mc X X ∂gij i j k
−ε u + u u h (s) ds + . . . .
c i=0 ∂rk 2 i=0 j=0 ∂rk
s1 k=0
Remember that function hk (s) vanish at the ends of integration

path hk (s1 ) = hk (s2 ) = 0 (see § 1). This provides vanishing of
non-integral terms in the above formula for Sdef .
Now in order to derive differential equations for world line of
particle we apply extremity condition for S. It means that term
linear with respect to ε in power expansion for Sdef should be

§ 2. MOTION OF PARTICLE . . . 127
identically zero irrespective to the choice of functions hk (s):
d q
m c uk (s) + Ak (r(s)) =
ds c
(2.4) 3 3 3
q X ∂Ai i mc X X ∂gij i j
= u + u u .
c i=0 ∂rk 2 i=0 j=0 ∂rk
Let’s calculate derivative in left hand side of (2.4). Then let’s

rearrange terms so that those with m c factor are in left hand
side, while others with q/c factor are in right hand side:
3 3
! 3
duk 1 X X ∂gij i j q X ∂Ai ∂Ak
mc − u u = − ui .
ds 2 i=0 j=0 ∂rk c i=0 ∂rk ∂ri
Now in right hand side of this equation we find tensor of electro-

magnetic field (see formula (9.9) in Chapter III). For transforming
left hand side of this equation we use formula (11.3) from Chap-
ter III. As a result we get the following equation for world line:
3 3
! 3
duk X X i qX
(2.5) mc − Γkj ui uj = Fki ui .
ds i=0 j=0
c i=0
In left hand side of the equation (2.5) we find covariant derivative

with respect to parameter s along world line:
3
qX
(2.6) m c ∇s uk = Fki ui .
c i=0
Comparing (2.6) with the equations (11.9) from Chapter III, we

get formula for the vector of four-dimensional force acting on a

particle with charge q in electromagnetic field:
3
qX
(2.7) Fk = Fki ui .
c i=0
Suppose that we have rectangular Cartesian coordinate system

in Minkowsky space. Then we can subdivide F into spatial and
temporal parts and canp calculate components of three-dimensional
force vector: f i = c2 − |v|2 F i (see formula (7.6) in Chap-
ter III). Upon easy calculations with the use of formulas (7.2) and
(9.4) from Chapter III for force vector f we get
q
(2.8) f = qE+ [v, H].
c
This formula (2.8) is exactly the same as formula for Lorentz force
(see (4.4) in Chapter I). Thus formula (2.7) is four-dimensional
generalization of formula for Lorentz force. Orthogonality condi-
tion for 4-force and 4-velocity (see (7.5) in Chapter III) for (2.7) is
fulfilled due to skew symmetry of tensor of electromagnetic field.
Exercise 2.1. Prove that gauge transformation of action func-
tional (1.13) does not change dynamic equations of material point
in electromagnetic field (2.6).
Exercise 2.2. Verify that the relationship (7.5) from Chap-
ter III holds for Lorentz force.
§ 3. Dynamics of dust matter.

Differential equation (2.6) describes motion of charged particles
in electromagnetic field. If the number of particles is not large,
then we can follow after the motion of each of them. When
describing extremely large number of particles continual limit is
§ 3. DYNAMICS OF DUST MATTER. 129
used, particles are replaced by continuous medium modeling their

collective behavior. Simplest model describing large number of
non-colliding particles is a model of dust cloud. In this model
Fig. 3.1 Fig. 3.2
particles of cloud move regularly (not chaotically). Their world

lines can be modeled by regular family of lines filling the whole
space (see Fig. 3.1).
Another model is a model of ideal gas. Here particles also
do not collide each other, i. e. their world lines do not intersect.
However, their motion is chaotic (see Fig. 3.2). Therefore if we
fill the whole space with their world lines, they would intersect.
Besides two models considered mentioned above, there are
models describing liquids and solid materials. Points of liquid
and solid media move regularly (as on Fig. 3.1). However, in
these media interaction of particles is essential. Therefore when
describing such media one should either use detailed microscopic
analysis and get macroscopic parameters by statistical averaging,
or should use some heuristic assumptions based on experiment.
In this book we consider only most simple model of dust cloud.
In this case one should assume Minkowsky space to be filled by
regular family of world lines. Some of them are world lines of real
particles, others are imaginary ones obtained by extrapolation in

continual limit. Therefore at each point of M we have unit vector
u, this is tangent vector to world line passing through this point.
This means that dynamics of dust cloud can be described by
vector field u(r).
Apart from vector field u, below we need scalar parameter ν(r)
which means the density of dust cloud. We define it as follows.
Let’s choose some small fragment of three-dimensional hypersur-
face in M orthogonal to vector u(r) at the point r. The number
of dust particles whose world lines cross this fragment is propor-
tional to its three-dimensional volume: N = ν(r) V , parameter
ν(r) is coefficient of proportionality. Parameter ν(r) has measure
unit of concentration, it can be treated as concentration of par-
ticles in small fragment of dust cloud near the point r measured
in that inertial coordinate system for which particles of this small
fragment are at rest. By means of ν(r) and u(r) we compose new
four-dimensional vector
(3.1) η(r) = c ν(r) u(r).
Vector (3.1) is called four-dimensional flow density for particles

in cloud. If we choose some inertial coordinate system, then
η 0 /c is interpreted as concentration of particles in dust cloud,
while other three components of four-dimensional vector η form
three-dimensional vector of flow density.
Suppose that dust cloud is formed by identical particles with
mass m and charge q. Then four-dimensional current density
vector can be represented as follows:
(3.2) j(r) = q η(r).
By analogy with (3.2) one can define mass flow density vector:
(3.3) µ(r) = m η(r).

§ 3. DYNAMICS OF DUST MATTER. 131
Total number of particles in cloud is fixed. This conservation law

is written as the following equality for η:
3
X
(3.4) ∇p η p = 0.
p=0
From (3.4) and (3.2) one can derive charge conservation law in
form of the relationship (11.7) from Chapter III. Taking into
account (3.3), we get rest mass conservation law:
3
X
(3.5) ∇p µp = 0.
p=0
Rest mass conservation law here is fulfilled due to the absence of

collisions when heavy particles can be produced from light ones
(see § 7 in Chapter III).
Let’s consider dynamics of particles composing dust cloud.
Vector field u is constituted by tangent vectors to world lines
of dust particles. Therefore these world lines can be determined
as integral curves of vector field u, i. e. by solving the following
system of ordinary differential equations:
dri
(3.6) = ui (r(s)), i = 0, . . . , 3.
ds
Having determined world line of particle from differential equa-

tions, we know vector of its 4-velocity u(s). Now let’s calculate
covariant derivative of vector u(s) with respect to parameter s:
3 3
dup (s) X X p k
(3.7) ∇s up = + Γnk u (s) un (s).
ds n=0
k=0
Calculating derivative dup /ds in (3.7) we take into account (3.6)

and the equality u(s) = u(r(s)). As a result we get
3
dup (s) X k ∂up
(3.8) = u .
ds ∂rk
k=0
Substituting (3.8) into (3.7), we derive the following formula:
3
X
(3.9) ∇s up = uk ∇k up .
k=0
Right hand side of (3.9) is covariant derivative of vector field u(r)

along itself (see more details in [3]). Substituting (3.9) into the
equations of the dynamics of material point, we get:
F
(3.10) ∇u u = .
mc
Here F = F(r, u) is some external force field acting on particles

of dust matter. For example in the case of charged dust in
electromagnetic field the equation (3.10) looks like
3
X 3
q X
(3.11) uk ∇k up = Fpk uk .
mc2
k=0 k=0
In contrast to the equations (11.9) from Chapter III, which

describe dynamics of separate particle, here (3.10) are partial
differential equations with respect to components of vector field
u(r). They describe dynamics of dust cloud in continual limit.
The equation for scalar field ν(r) is derived from conservation law
§ 4. ACTION FUNCTIONAL FOR DUST MATTER. 133
(3.4) for the number of particles. Combining these two equations,

we obtain a system of differential equations:
3
X Fp
uk ∇k up = ,
mc
k=0
(3.12)
3
X 3
X
uk ∇k ν = −ν ∇k uk .
k=0 k=0
System of partial differential equations (3.12) yields complete

description for the dynamics of dust cloud.
Model of dust matter cam be generalized a little bit. We can
consider mixture of particles of different sorts. For each sort of
particles we define its own vector field u(i, r) and its own scalar
field of concentration ν(i, r). Then formulas (3.2) and (3.3) for j
and µ are generalized as follows:
n
X n
X
j(r) = q(i) η(i, r), µ(r) = m(i) η(i, r).
i=1 i=1
Here η(i, r) = c ν(i, r) u(i, r). Each pair of fields u(i, r) and ν(i, r)
satisfies differential equations (3.12). We can derive mass and
charge conservation laws from these equations.
§ 4. Action functional for dust matter.

Let’s study the dynamics of dust matter in electromagnetic
field within framework of Lagrangian formalism. Fort this purpose
we need to pass to continual limit in action functional (1.8).
For the sake of simplicity we consider dust cloud with identical
particles. Omitting details of how it was derived, now we write

ultimate formula for action functional (1.8) in continual limit:
ZV2p p
S = −m g(η, η) − det g d4 r−
V1
ZV2 p
q
(4.1) − 2 g(η, A) − det g d4 r−
c
V1
ZV2X
3 X
3 p
1
− Fpk F pk − det g d4 r.
16π c
V1 p=0 k=0
Instead of deriving formula (4.1) from (1.8) we shall verify this

formula indirectly. For this purpose we shall derive dynamical
equation (3.11) from principle of extremal action applied to action
functional (4.1).
For describing dust matter in (4.1) we have chosen vector field
η(r) defined in (3.1). Other two fields u(r) and ν(r) can be
expressed though vector field η(r):
p η
(4.2) c ν = |η| = g(η, η) , u= .
cν
Dealing with variation of vector field η(r) we should always

remember that components of this field are not independent
functions. They satisfy differential equation (3.4). In order
to resolve tis equation (3.4) we use slightly modified version of
theorem 10.1 from Chapter III.
Theorem 4.1. Let M be some n-dimensional manifold, where

n ≥ 2, equipped with metric gij . For each vector field η with zero
divergency with respect to metric connection

n
X
(4.3) ∇p η p = 0
p=1
there is skew-symmetric tensor field ϕ of type (2, 0) such that the

following relationships are fulfilled
n
X
(4.4) ηp = ∇q ϕpq .
q=1
Proof. Writing relationships (4.3), we use well-known for-

mula for components of metric connection, see formula (11.3) in
Chapter III. As a result we get
n
X n
X n X
X n n
X
∂η p ∂η p
∇p η p = + Γpps η s = +
p=1 p=1
∂rp p=1 s=1 p=1
∂rp
n n n
1 X X X pk ∂gpk ∂gks ∂gps
+ g + − ηs .
2 p=1 s=1 ∂rs ∂rp ∂rk
k=1
Note that last two derivatives of metric tensor in round brackets

are canceled when we sum over indices p and k. This is because
g pk is symmetric. Hence
n
X n
X n n n
∂η p 1 X X X sk ∂gks p
∇p η p = + g η =
p=1 p=1
∂rp 2 p=1 s=1 ∂rp
k=1
(4.5)
Xn n
∂η p 1X −1 ∂g
= p
+ tr g p
ηp .
p=1
∂r 2 p=1
∂r
For further transforming of this expression (4.5) we use well

known formula for logarithmic derivative of determinant:

∂ ln | det g| −1 ∂g
(4.6) = tr g .
∂rp ∂rp
Substituting (4.6) into (4.5), we transform (4.5) so that

n n p
X 1 X ∂(η p | det g|)
p
(4.7) ∇p η = p .
p=1 | det g| p=1 ∂rp
Let’s carry out analogous calculations for right hand side of (4.4)
taking into account skew symmetry of the field ϕpq and symmetry
of connection components Γkpq . These calculations yield
n n p
X 1 X ∂(ϕpq | det g|)
pq
(4.8) ∇q ϕ = p .
q=1 | det g| q=1 ∂rq
p p
Denote j p = | det g| η p and ψ pq = | det g| ϕpq . Now on the
base of (4.7) and (4.8) it is easy to understand that proof of
theorem 4.1 is reduced to theorem 10.1 from Chapter III.
Remark. Generally speaking, theorem 10.2 has no direct
generalization for the case of spaces with metric. It is generalized
s
only for metric spaces with zero curvature tensor Rkpq = 0.
Let’s define deformation of the field η in a way similar to that
we used for vector-potential A in § 1:
(4.9) η̂ p (r) = η p (r) + ε ζ p (r) + . . . .
Both fields η̂ and η satisfy differential equation (3.4). Hence

vector field ζ defined in (4.9) also satisfy this equation. Let’s
apply theorem 4.1 to vector field ζ:

3
X
(4.10) ζp = ∇k ϕpk .
k=0
Theorem 4.1 does not specify tensor field ϕpk in (4.10), this can
be any skew-symmetric tensor field. However, we choose it in very
special form as follows:
(4.11) ϕpk = η p hk − hp η k .
This choice can be motivated by the following theorem.

Theorem 4.2. For any two vector fields ζ and η, where η 6= 0,
both satisfying differential equation (3.4) there is vector field h
such that vector field ζ is given by formula
3
X
ζp = ∇k (η p hk − hp η k ).
k=0
Our choice (4.11) leads to the following expression for the field η̂:
3
X
(4.12) η̂ p (r) = η p (r) + ε ∇k (η p hk − hp η k ) + . . . .
k=0
Quantities hi (r) in (4.12) are chosen to be smooth functions being

nonzero only within some restricted domain Ω in Minkowsky
space.
When substituting (4.12)pinto action functional (4.1) we use
the following expansion for g(η̂, η̂) :
p p X3 X 3
ε
g(η̂, η̂) = g(η, η) + p ηp ∇k ϕpk + . . . .
g(η, η) p=0 q=0
We have analogous power expansion for the expression under

second integral in formula (4.1):
3 X
X 3
g(η̂, A) = g(η, A) + ε Ap ∇k ϕpk + . . . .
p=0 k=0
Substituting these two expansions into (4.1), we take into account

(4.2). For the action Sdef this yields
Z X
3 X
3 p
Sdef = S − ε m up ∇k ϕpk − det g d4 r−
Ω p=0 k=0
(4.13) Z 3 3
εq XX p
− 2 Ap ∇k ϕpk − det g d4 r + . . . .
c p=0
Ω k=0
Further in order to transform the above expression (4.13) we

use Ostrogradsky-Gauss formula. In the space equipped with
metric this formula is written as follows:
Z X
3 p Z
k 4
(4.14) ∇k z − det g d r = g(z, n) dV.
Ω k=0 ∂Ω
Here z 0 , z 1 , z 2 , z 3 are components of smooth vector field z,

while n is unit normal vector for the boundary of the domain
Ω. In order to transform first integral in formula (4.13) we take
P
z k = 3p=0 up ϕpk . Then in right hand side of (4.14) we obtain
3
X 3 X
X 3 3 X
X 3
∇k z k = up ∇k ϕpk + ∇k up ϕpk .
k=0 p=0 k=0 p=0 k=0
Right hand side of (4.14) vanishes since ϕpk do vanish on the

boundary of Ω. Hence we have the equality
Z X
3 X
3 p
up ∇k ϕpk − det g d4 r =
Ω p=0 k=0
Z X 3 X
3 p
− ∇k up ϕpk − det g d4 r.
Ω p=0 k=0
In a similar way we transform second integral in (4.13). In whole

for the action Sdef we get the following expression
Z X 3
3 X p
Sdef = S + ε m ∇k up ϕpk − det g d4 r +
Ω p=0 k=0
(4.15) Z 3 3
εq XX p
+ 2 ∇k Ap ϕpk − det g d4 r + . . . .
c p=0 Ω k=0
Extremity of action S means that linear part with respect to ε

in formula (4.15) should vanish:
Z X 3
3 X p
q
m∇k up + 2
∇k Ap ϕpk − det g d4 r = 0.
c
Ω p=0 k=0
Let’s substitute formula (4.11) for ϕpk into the above equality.
Then it is transformed to the following one
Z X 3
3 X p
q p k
m∇k up + 2
∇k p η h
A − det g d4 r =
c
Ω p=0 k=0
Z X 3
3 X p
q k p
= m∇k up + ∇k Ap η h − det g d4 r.
c2
Ω p=0 k=0
Let’s exchange indices k and p in second integral. Thereafter

integrals can be united into one integral:
Z X 3
3 X
q
m∇k up − m∇p uk + ∇k Ap −
c2
(4.16) Ω k=0 p=0
q p
− 2
∇p Ak η p hk − det g d4 r = 0.
c
Now let’s take into account that in resulting equality hk = hk (r)

are arbitrary smooth functions vanishing outside the domain Ω.
Therefore vanishing of integral (4.16) means vanishing of each
summand in sum over index k in the expression under integration:
X3
q
(4.17) m ∇k up − m ∇p uk + 2 Fkp η p = 0.
p=0
c
Here we used the relationship (11.5) from Chapter III. It re-

lates tensor of electromagnetic field and four-dimensional vector-
potential.
In order to bring the equation (4.17) just derived to its ultimate
form we use the relationships (4.2), which relate vector field η
and vector field u: η p = c ν up . Since u is unit vector, we have
3
X
(4.18) up ∇k up = 0.
p=0

§ 5. EQUATIONS FOR ELECTROMAGNETIC FIELD. 141
Taking into account (4.18), we bring (4.17) to the following form:
3
X 3
q X
(4.19) up ∇p uk = Fkp up .
p=0
mc2 p=0
Now it is easy to see that (4.19) exactly coincides with the equa-
tion (3.11), which we have derived earlier. This result approves
the use of the action (4.1) for describing charged dust matter in
electromagnetic field.
Exercise 4.1. Prove that for any skew-symmetric tensor field
ϕpq vector field η determined by formula (4.4) has zero divergency,
i. e. is satisfies differential equation (3.4).
Exercise 4.2. Prove theorem 4.2. For this purpose use the
following fact known as theorem on rectification of vector field.
Theorem 4.3. For any vector field η 6= 0 there exists some
curvilinear coordinate system r0 , r1 , r2 , r3 such that η 0 = 1,
η 1 = 0, η 2 = 0, η 3 = 0 in this coordinate system.
Exercise 4.3. Prove theorem 4.3 on rectification of vector field.
Exercise 4.4. Derive Ostrogradsky-Gauss formula (4.14) for
the space equipped with metric on the base of the following integral
relationship in standard space Rn :
Z Z
∂f (r) n
d r= f (r) dr1 . . . dri−1 dri+1 . . . drn .
∂ri
Ω ∂Ω
§ 5. Equations for electromagnetic field.

In this section we continue studying action functional (4.1).
This functional describes dust cloud composed of particles with
mass m and charge q in electromagnetic field. In previous section

we have found that applying extremal action principle to S with
respect to the field η one can derive dynamical equations for
velocity field in dust cloud. Now we shall apply extremal action
principle to S with respect to vector-potential A. Deformation of
vector-potential is defined according to (1.9), (1.10), (1.11), (1.12):
(5.1) Âi (r) = Ai (r) + εhi (r) + . . . .
For components of tensor of electromagnetic field we derive
(5.2) F̂ij = Fij + ε (∇i hj − ∇j hi ) + . . . .
When substituting (5.2) into action functional (4.1) we carry out

the following calculations:
3 X
X 3 3 X
X 3 X
3 X
3
F̂pk F̂ pk = F̂pk F̂ij g pi g kj =
p=0 k=0 i=0 j=0 p=0 k=0
3 X
X 3 3 X
X 3
= Fpk F pk + 2 ε F pk (∇p hk − ∇k hp ) + . . . .
p=0 k=0 p=0 k=0
Taking into account skew symmetry of tensor F pk , this expansion

can be simplified more and can be brought to the form
3 X
X 3 3 X
X 3 3 X
X 3
F̂pk F̂ pk = Fpk F pk + 4 ε F pk ∇p hk + . . . .
p=0 k=0 p=0 k=0 p=0 k=0
Analogous calculations in substituting (5.1) into (4.1) yield

3
X
g(η, Â) = g(η, A) + ε η k hk + . . . .
k=0
§ 5. EQUATIONS FOR ELECTROMAGNETIC FIELD. 143
As a result for deformation of action functional (4.1) we get
Z X
3 p
εq
Sdef = S − η k hk − det g d4 r −
c2
Ω k=0
Z X
3 X3 p
ε
− F pk ∇p hk − det g d4 r + . . . .
4π c p=0
Ω k=0
Let’s transform second integral in the above expansion for Sdef

by means of Ostrogradsky-Gauss formula (4.14). For this purpose
P3
let’s choose z p = k=0 F pk hk and take into account vanishing of
hk on the boundary of the domain Ω. Then for Sdef we get
Z X
3 3
!
qη k X ∇p F pk p
Sdef = S + ε − 2 + hk − det g d4 r + . . . .
c p=0
4π c
Ω k=0
Extremal action principle means that linear in ε part of the above

expansion for Sdef should vanish. Note also that Ω is an arbitrary
domain and hk (r) are arbitrary functions within Ω. This yield
the following equations for the tensor of electromagnetic field:
3
X 4π q k
(5.3) ∇p F pk = η .
p=0
c
Remember that η(r) is related to current density by means of

(3.2). Then (5.3) can be written as
3
X 4π k
(5.4) ∇p F kp = − j .
p=0
c
It is easy to see that (5.4) are exactly Maxwell equations written

in four-dimensional form (see (11.4) in Chapter III). Another pair
of Maxwell equations written in four-dimensional form
3 X
X 3 X
3
ω pqks ∇q Fks = 0
q=0 k=0 s=0
is a consequence of the relationship Fpq = ∇p Aq − ∇q Ap (see

formula (11.5) in Chapter III).
Exercise 5.1. Which form will have differential equations (5.3)
if we consider dust cloud composed by particles of several sorts
with masses m(1), . . . , m(N ) and charges q(1), . . . , q(N ) ? Will
this change differential equations (5.4) ?
CHAPTER V
GENERAL THEORY OF RELATIVITY
§ 1. Transition to non-flat metrics

and curved Minkowsky space.
Passing from classical electrodynamics to special theory of re-
lativity, in previous two chapters we have successively geometrized
many basic physical concepts. Having denoted r0 = ct and com-
bining r0 with components of three-dimensional radius-vector in
inertial coordinate system, we have constructed four-dimensional
space of events (Minkowsky space). This space appears to be
equipped with metric of signature (1, 3), which is called Min-
kowsky metric. Thereby inertial coordinate systems are inter-
preted as orthonormal bases in Minkowsky metric.
In four-dimensional formalism dynamics of material point is
described by vectorial differential equations, while Maxwell equa-
tions for electromagnetic field are written in tensorial form. Due
to this circumstance in previous two chapters we managed to
include into consideration skew-angular and even curvilinear co-
ordinate systems in Minkowsky space. Thereby we got explicit
entries of metric tensor components gij , metric connection com-
ponents Γkij , and covariant derivatives ∇i in all our equations.
Next step in this direction is quite natural. One should keep
the shape of all equations and pass from flat Minkowsky metric
146 CHAPTER V. GENERAL RELATIVITY.
to metric of signature (1, 3) with nonzero curvature tensor:

3 3
k
∂Γkjq ∂Γkiq X k s
X
(1.1) Rqij = − + Γ Γ
is jq − Γkjs Γsiq .
∂ri ∂rj s=0 s=0
This crucial step was first made by Einstein. Theory he had dis-
covered in this way later was called Einstein’s theory of gravitation
or general theory of relativity.
Definition 1.1. Four-dimensional affine space equipped with
orientation, polarization, and with metric of signature (1, 3) and
nonzero curvature (1.1) is called curved Minkowsky space.
In non-flat Minkowsky space we loose some structures available
in flat case. In such space there are no coordinates for which
Minkowsky metric is given by matrix (2.7) from Chapter III, i. e.
here we have no inertial coordinate systems. This is substantial
loss, but it is not catastrophic since dynamic equation for material
points and Maxwell equations rewritten in vectorial and tensorial
form are not bound to inertial coordinate systems.
Geodesic lines in curved Minkowsky space do not coincide
with affine straight lines. Therefore affine structure becomes
excessive restriction in general relativity. As appears, one can
give up topologic structure of flat space R4 as well. Even in
two-dimensional case, as we know, apart from deformed (curved)
plain, there are surfaces with more complicated topology: sphere,
torus and sphere with several handles glued to it (see [5]). In
multidimensional case these objects are generalized in concept of
smooth manifold (see details in [2], [5], and [6]).
Smooth manifold M of dimension n is a topologic space each
point of which has a neighborhood (a chart) identical to some
neighborhood of a point in Rn . In other words M is covered by a
family of charts Uα , each of which is diffeomorphic to some open
set Vα in Rn . Such chart maps (chart diffeomorphisms) define
§ 2. ACTION FOR GRAVITATIONAL FIELD. 147
local curvilinear coordinate systems within their chart domains

Uα . At those points of manifold M , where two chart domains are
overlapping, transition functions arise. They relate one curvilinear
coordinate system with another:
r̃i = r̃i (r1 , . . . , rn ), where i = 1, . . . , n,

(1.2)
ri = ri (r̃1 , . . . , r̃n ), where i = 1, . . . , n.
According to definition of smooth manifold, transition functions

(1.2) are smooth functions (of class C ∞ ). Transition functions
determine transition matrices S and T :
∂r̃i ∂ri
(1.3) Tji = , Sji = .
∂rj ∂r̃j
Presence of transition matrices (1.3) lead to full-scale theory of
tensors, which is almost literally the same as theory of tensors
for curvilinear coordinates in Rn (see [3]). The only difference
is that here we cannot choose Cartesian coordinates at all. This
is because in general there is no smooth diffeomorphic map from
manifold M to Rn .
Definition 1.1. Four-dimensional smooth manifold equipped
with orientation, polarization, and with metric of signature (1, 3)
is called generalized Minkowsky space or Minkowsky manifold.
§ 2. Action for gravitational field.

Einstein equation.
Space of events in general relativity is some smooth Minkowsky
manifold M . This circumstance provides additional arbitrariness
consisting in choosing M and in choosing metric on M . Nonzero
curvature described by tensor (1.1) is interpreted as gravitational
field. Gravitational field acts upon material bodies and upon

electromagnetic field enclosed within M . This action is due to

the presence of covariant derivatives in dynamic equations. The
magnitude of gravitational field itself should be determined by
presence of matter in M in form of massive particles or in form of
electromagnetic radiation, i. e. we should have backward relation
between geometry of the space and its content.
In order to describe backward relation between gravitational
field and other physical fields we use Lagrangian formalism and
extremal action principle. Let’s start from action functional (4.1)
in Chapter IV. It is sum of three integrals:
(2.1) S = Smat + Sint + Sel .
First integral Smat is responsible for material particles in form of

dust cloud, second integral describes interaction of dust cloud and
electromagnetic field, third term in (2.1) describes electromagnetic
field itself. In order to describe gravitational field one more
summand in (2.1) is added:
(2.2) S = Sgr + Smat + Sint + Sel .
This additional term is chosen in the following form:

ZV2 p
c3
(2.3) Sgr =− R − det g d4 r.
16πγ
V1
Here γ gravitational constant same as in Newton’s universal law

of gravitation (see formula (1.11) in Chapter I). Scalar quantity R
in (2.3) is scalar curvature determined by curvature tensor (1.1)
according to the following formula:
3 X
X 3 X
3
(2.4) R= g qj Rqkj
k
.
q=0 k=0 j=0
Ricci tensor is an intermediate object relating curvature tensor

(1.1) and scalar quantity (2.4). Here are its components:
3
X
k
(2.5) Rqj = Rqkj .
k=0
Ricci tensor is symmetric (see [3]). Scalar curvature R is obtained

by contracting Ricci tensor and metric tensor g qj with respect to
both indices q and j. This fact is obvious due to (2.5) and (2.4).
Note that sometimes in the action for gravitational field (2.3)
one more constant parameter Λ is added:
ZV2 p
c3
Sgr =− (R + 2 Λ) − det g d4 r.
16πγ
V1
This parameter is called cosmological constant. However, accord-

ing to contemporary experimental data the value of this constant
is undetectably small or maybe is exactly equal to zero. Therefore
further we shall use action Sgr in form of (2.3).
Note also that metric tensor describing gravitational field enters
in implicit form into all summand in (2.2). Therefore we need
not add special terms describing interaction of gravitational field
with material particles and electromagnetic field. Moreover, such
additional terms could change the form of dynamical equations for
matter and form of Maxwell equations for electromagnetic field
thus contradicting our claim that these equations are the same in
general and in special relativity.
Now let’s begin with deriving dynamical equations for grav-
itational field. For this purpose we consider deformation of
components of metric tensor given by the following relationship:
(2.6) ĝ ij (r) = g ij (r) + ε hij (r) + . . . .

Functions hij (r) in (2.6) are assumed to be smooth functions

vanishing outside some restricted domain Ω ⊂ M . Deformation of
matrix g ij lead to deformation of inverse matrix gij :
ĝij = gij − ε hij + . . . =
(2.7) 3 X
X 3
= gij − ε gip hpq gqj + . . . .
p=0 q=0
Let’s differentiate the relationship (2.7) and let’s express partial

derivatives through covariant derivatives in resulting formula:
∂ĝij ∂gij ∂hij ∂gij
k
= k
−ε k
+ ... = − ε ∇k hij +
∂r ∂r ∂r ∂rk
(2.8) 3
X 3
X
+ε Γpki hpj + ε Γpkj hip + . . . .
p=0 p=0
In (2.8) we used covariant derivatives corresponding to non-

deformed metric gij . Now on the base of (2.8) we calculate
the following combination of derivatives:
∂ĝkj ∂ĝik ∂ĝij ∂gkj ∂gik ∂gij
+ − = + − −
∂ri ∂rj ∂rk ∂ri ∂rj ∂rk
(2.9) 3
!
X p
− ε ∇i hkj + ∇j hik − ∇k hij − 2 Γij hpk + . . . .
p=0
Let’s use the relationships (2.6) and (2.8) in calculating defor-

mation of connection components. For this purpose let’s apply
well-known formula to Γ̂pij (see formula (11.3) in Chapter III):
3
ε X pk
Γ̂pij = Γpij + g (∇i hkj + ∇j hik − ∇k hij ) + . . . .
2
k=0
This expansion for Γ̂pij can be written in symbolic concise form
(2.10) Γ̂pij = Γpij + ε Yijp + . . .
by introducing the following quite natural notation:

3
1 X pk
(2.11) Yijp = g (∇i hkj + ∇j hik − ∇k hij ) .
2
k=0
Now let’s substitute the expansion (2.10) into the formula (1.1)
for curvature tensor. This yields
k k

(2.12) R̂qij = Rqij + ε ∇i Yjqk − ∇j Yiqk + . . . .
Upon contracting (2.12) with respect to one pair of indices we get

similar expansion for deformation of Ricci tensor:
3
X
(2.13) R̂qj = Rqj + ε ∇k Yjqk − ∇j Ykq
k
+ ... .
k=0
We multiply (2.13) by gqj using formula (2.6). Then we carry out

complete contraction with respect to both indices q and j. This
yields deformation of scalar curvature:
3 X
3 3
!
X X
qj qj k k
R̂ = R + ε Rqj h + g (∇k Yjq − ∇j Ykq ) + ... .
j=0 q=0 k=0
Let’s introduce vector field with the following components:

3
3 X
X
Zk = Yjqk g qj − Yjqj g qk .
j=0 q=0
Then we can rewrite deformation of scalar curvature R̂ as

X 3
3 X 3
X
(2.14) R̂ = R + ε Rqj hqj + ε ∇k Z k + . . . .
j=0 q=0 k=0
When substituting (2.14) into action integral (2.3) we should note

that second sum in (2.14) is exactly covariant divergency of vector
field Z. Components of Z are smooth functions vanishing outside
the domain Ω. Therefore integral of such sum is equal to zero:
Z X
3 p Z
k 4
∇k Z − det g d r = g(Z, n) dV = 0.
Ω k=0 ∂Ω
This follows from Ostrogradsky-Gauss formula (see (4.14) in

Chapter IV). Hence for deformation of Sgr we get
Z X3 X 3
p
ε c3 R
Sdef = Sgr − Rqj − gqj hqj − det g d4 r + . . . .
16πγ j=0 q=0
2
Ω
In deriving this formula we also used the following expansion:

3 X 3
!
p p X gqj hqj
(2.15) − det ĝ = − det g 1 − ε + ... .
j=0 q=0
2
It follows from (2.6). Now we shall not calculate deformations of

other three terms in (2.2) in explicit form. This will be done in
§ 4 and § 5 below. However, we introduce notation
(2.16) Sm.f. = Smat + Sint + Sel .
Here Sm.f. denotes overall action for all material fields other than
gravitation. The number of terms in the sum (2.16) could be much
§ 3. CONSERVATION LAW . . . 153
more than three, if one consider more complicated models for

describing matter. But in any case action of gravitational field is
excluded from this sum since gravitational field plays exceptional
role in general relativity. Now we shall write deformation of the
action (2.16) in the following conditional form:
Z X
3 X
3 p
ε
(2.17) Sdef = Sm.f. + Tqj hqj − det g d4 r + . . . .
2c
Ω q=0 j=0
Then extremity condition for total action (2.2) is written as

R 8πγ
(2.18) Rqj − gqj = 4 Tqj .
2 c
This equation (2.18) is known as Einstein equation. It is ba-
sic equation describing dynamics of metric tensor gij in general
theory of relativity.
Exercise 2.1. Derive the relationships (2.7) and (2.15) from
the expansion (2.6) for deformation of tensor g ij .
§ 3. Four-dimensional momentum
conservation law for fields.
Tensor T in right hand side of Einstein equation (2.18) is called
energy-momentum tensor for material fields. It is determined
by the relationship (2.17) and comprises contributions from all
material fields and their interactions. In the model of dust matter
in electromagnetic field tensor T is composed of three parts (see
formula (2.16)).
Energy-momentum tensor is related with 4-momentum conser-
vation law for material fields. In order to derive this conservation
law we use well-known Bianchi identity:
p p p
(3.1) ∇k Rsij + ∇i Rsjk + ∇j Rski = 0.
More details concerning Bianchi identity (3.1) can be found in [2]

and [6]. Let’s contract this identity with respect to i and p:
3
X p
(3.2) ∇k Rsj + ∇p Rsjk − ∇j Rsk = 0.
p=0
Here we used skew symmetry of curvature tensor with respect

to last pair of indices (see [3]). Let’s multiply (3.2) by g sj and
contract it with respect to double indices s and j. Upon slight
ps sp
transformation based on skew symmetry Rij = −Rij we get
3
X 1
(3.3) ∇s Rks − ∇k R = 0.
s=0
2
Now let’s raise index j in the equation (2.18), then apply covariant
differentiation ∇j and contract with respect to double index j:
3
X 3
1 8πγ X
(3.4) ∇j Rqj − ∇q R = 4 ∇j Tqj .
j=0
2 c j=0
Comparing (3.3) and (3.4), we get the following equation for

energy-momentum tensor of material fields:
3
X
(3.5) ∇j Tqj = 0.
j=0
The equation (3.5) expresses 4-momentum conservation law for

the whole variety of material fields. It is usually written in the
following form with raised index q:
3
X
(3.6) ∇j T qj = 0.
j=0
Energy-momentum tensor is symmetric therefore the order of

indices q and j in (3.6) is unessential.

§ 4. ENERGY-MOMENTUM TENSOR . . . 155
§ 4. Energy-momentum tensor
for electromagnetic field.
Energy-momentum tensor for whole variety of material fields
is defined by the relationship (2.17). By analogy with (2.17) we
define energy-momentum tensor for electromagnetic field:
Z X
3 X
3 p
ε
(4.1) Sdef = Sel + Tqj hqj − det g d4 r + . . . .
2c
Ω q=0 j=0
Basic fields in the action Sel are covariant components of vector-

potential Ai (r). Covariant components of tensor of electromag-
netic field are defined by formula
∂Aj ∂Ai
(4.2) Fij = ∇i Aj − ∇j Ai = i
− j
∂r ∂r
(see also formula (11.5) in Chapter III). Ultimate expression in

right hand side of (4.2) has no entry of connection components
Γkij . Therefore covariant components Fij are not changed by
deformation of metric (2.6). Upon raising indices we get
3 X
X 3
pk
F̂ = ĝ pi ĝ kj Fij
i=0 j=0
and, using this formula, for contravariant components F pq of

tensor of electromagnetic field we derive the expansion
3 X
X 3
(4.3) F̂ pk = F pk + ε (hpi g kj + g pi hkj ) Fij + . . . .
i=0 j=0
Substituting F̂ pk and ĝ into action functional Sel , we get
ZV2X
3 X
3 p
1
Sdef =− Fpk F̂ pk − det ĝ d4 r
16π c
V1 p=0 k=0
Then, taking into account (4.3) and (2.15), we derive formula

Z X
3 X
3 3 X
X 3
ε
Sdef = Sel − 2 Fpq g pi Fij −
16π c
Ω q=0 j=0 p=0 i=0
3 3
!
1 XX p
− Fpi F gqj hqj − det g
pi
d4 r + . . . .
2 p=0 i=0
Comparing this actual expansion with expected expansion (4.1)

for Sdef , we find components of energy-momentum tensor for
electromagnetic field in explicit form:
3 3
1 XX pi 1 pi
(4.4) Tqj =− Fpq g Fij − Fpi F gqj .
4π p=0 i=0 4
Raising indices q and j in (4.4), for contravariant components of

energy-momentum tensor T we derive
3 3
1 XX 1
(4.5) T qj = − F pq gpi F ij − Fpi F pi g qj .
4π p=0 i=0 4
By means of formula (4.5) one can calculate covariant divergency

for energy-momentum tensor of electromagnetic field:
3
X 3
1 X ps
(4.6) ∇s T ps = − F js .
s=0
c s=0
Formula (4.6) shows that 4-momentum conservation law for sepa-

rate electromagnetic field is not fulfilled. This is due to momen-
tum exchange between electromagnetic field and other forms of
matter, e. g. dust matter.
Exercise 4.1. Verify the relationship (4.6). For this purpose
use well-known formula for commutator of covariant derivatives
3
X
s
(∇i ∇j − ∇i ∇j )Ak = − Rkij As
s=0
and properties of curvature tensor (see details in [3]).

Exercise 4.2. Calculate components of energy-momentum
tensor (4.5) in inertial coordinate system for flat Minkowsky met-
ric. Compare them with components of Maxwell tensor, with den-
sity of energy, and with vector of energy flow for electromagnetic
field (see formulas (2.5) and (2.15) in Chapter II).
§ 5. Energy-momentum tensor
for dust matter.
Let’s consider energy-momentum tensor related with last two
terms Smat and Sint in the action (2.16). They contain entries of
vector field η whose components satisfy differential equation
3
X
(5.1) ∇p η p = 0,
p=0
see (3.4) in Chapter IV. This circumstance differs them from com-
ponents of vector-potential A. Metric tensor gij enters differential
equation (5.1) through connection components Γkij of metric con-
nection. Therefore by deformation of metric gij → ĝij one cannot
treat η p as metric independent quantities.
In order to find truly metric independent variables for dust

matter we use formula (4.7) from Chapter IV and rewrite differ-
ential equation (5.1) as follows:
3 p
X ∂(η p − det ĝ )
= 0.
p=0
∂rp
p
Denote η̂ p = η p − det g . These quantities η̂ p can be treated as
metric independent ones since differential constraint for them is
written in form of the equation that does not contain metric:
3
X ∂ η̂ p
(5.2) = 0.
p=0
∂rp
Expressing η p through η̂ p , for action functional Sint describing

interaction of dust matter and electromagnetic field we get
ZV2X
3
q
(5.3) Sint =− 2 η̂ p Ap d4 r.
c p=0
V1
It is easy to see that integral (5.3) does not depend on metric

tensor. Therefore action functional Sint makes no contribution to
overall energy-momentum tensor.
Now let’s express η p through η̂ p in action functional Smat for
dust matter. As a result we get formula
v
ZV2u
uX 3 X 3
(5.4) Smat = −m t gpq η̂ p η̂ q d4 r.
V1 p=0 q=0
The dependence of this functional on metric tensor is completely

determined by explicit entry of gpq under square root sign in right
hand side of (5.4). Therefore power extension for Smat is easily

calculated on the base of the expansion (2.7):
Z X 3 X 3
!
ε m ηp ηq p
Sdef = Smat + p hpq − det g d4 r + . . . .
2 p=0 q=0 g(η, η)
Ω
Let’s compare this expansion with expected expansion for Sdef :

Z X
3 X
3 p
ε
Sdef = Smat + Tpq hpq − det g d4 r + . . . .
2c
Ω p=0 q=0
By this comparison we find explicit formula for components of

energy-momentum tensor for dust matter:
mc ηp ηq p
(5.5) Tpq = p = mc g(η, η) up uq .
g(η, η)
Contravariant components of energy-momentum tensor (5.5)

are obtained by raising indices p and q:
mc η p η q p
(5.6) T pq = p = mc g(η, η) up uq .
g(η, η)
Using collinearity of vectors u and η (see formula (3.1) in Chap-

ter IV) and recalling that u is unit vector, we can bring formula
(5.6) to the following simple form:
(5.7) T pk = mc up η k .
Formula (5.7) is convenient for calculating covariant divergency of

energy-momentum tensor for dust matter:
3
X 3
q X ps
(5.8) ∇s T ps = F ηs .
s=0
c s=0
Now, applying formula (3.2) from Chapter IV, we can transform

formula (5.8) and write it as follows:
3
X 3
ps 1 X ps
(5.9) ∇s T = F js .
s=0
c s=0
Let’s compare (5.9) with analogous formula (4.6) for energy-

momentum tensor of electromagnetic field. Right hand sides of
these two formulas differ only in sign. This fact has transparent
interpretation. It means that in our model the overall energy-
momentum tensor for matter
Tm.f. = Tmat + Tel
satisfies differential equation (3.6). This fact is in complete

agreement with 4-momentum conservation law.
Another important conclusion, which follows from of (4.6) and
(5.9), is that 4-momentum conservation law for the whole variety
of material fields can be derived from dynamical equations for
these fields. Therefore this law is valid also in special relativity,
where Einstein equation (2.18) is not considered and where in
general case for flat Minkowsky metric it is not fulfilled.
Exercise 5.1. Derive the relationship (5.8) on the base of equa-
tions (3.4) and (4.19) from Chapter IV.
§ 6. Concluding remarks.
Event space in general theory of relativity is some Minkowsky
manifold M with Minkowsky metric of signature (1, 3). This
metric is determined by material content of the space according
to Einstein equation (2.18). However, topology of the manifold
M has great deal of arbitrariness. This manifold can have local
singularities at the points with extremely high concentration of
§ 6. CONCLUDING REMARKS. 161
matter. Such objects are called black holes. Moreover, global

topology of M also can be nontrivial (other than topology of R4 ).
In contemporary physics most popular models of M include big
bang in the very beginning of times. According to these models
in far past times our Universe M was extremely small, while
density of matter in it was extremely high. In further evolution
our Universe was expanding up to its present size. Will this
expansion last infinitely long or it will change for contraction ?
This problem is not yet solved. The answer to this question
depends on estimates of total amount of matter in the Universe.
In this book we cannot consider all these fascinating problems
of modern astrophysics and cosmology. However, I think the
above theoretical material makes sufficient background for to
continue studying these problems e. g. in books [2], [7], and [8].
I would like also to recommend the book [9] of popular genre,
where these problems are discussed in commonly understandable
and intriguing manner.

REFERENCES
1. Vladimirov V. S. Equations of mathematical physics, Nauka

publishers, Moscow, 1981.
2. Dubrovin B. A., Novikov S. P., Fomenko A. T. Modern geome-
try, vol. I, Nauka publishers, Moscow, 1986.
3. Sharipov R. A. Course of differential geometry, Publication of
Bashkir State University, Ufa, 1996.
4. Sharipov R. A. Course of linear algebra and multidimensional
geometry, Publication of Bashkir State University, Ufa, 1996.
5. Borisovich Yu. G., Bliznyakov N. M., Izrailevich Ya. A.,
Fomenko T. N. Introduction to topology, Nauka publishers,
Moscow, 1995.
6. Kobayashi Sh., Nomizu K. Foundations of differential geometry,
Interscience publishers, New York, London, 1963.
7. Landau L. D., Lifshits E. M. Course of theoretical physics,
vol. II, Field theory, Nauka publishers, Moscow, 1988.
8. Bogoyavlensky O. I. Methods of qualitative theory of dynamical
systems in astrophysics and in gas dynamics, Nauka publishers,
Moscow, 1980.
9. Davis P. Superforce. The search for a grand unified theory of
nature, Symon and Schuster publishers, New York, 1984.
CONTACTS
Address: Home address:

Ruslan A. Sharipov, Ruslan A. Sharipov,
Math. Department, Rabochaya street 5,
Bashkir State University, 450003, Ufa, Bashkortostan,
Frunze street 32, Russia
450074, Ufa, Bashkortostan,
Russia
Phone: E-mails:
7-(3472)-23-67-18 R [email protected]
7-(3472)-23-67-74 (FAX) [email protected]
ra [email protected]

Sharipov Classical Electrodynamics and T

Uploaded by

Copyright:

Available Formats

Sharipov Classical Electrodynamics and T

Uploaded by

Document Information

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

Sharipov Classical Electrodynamics and T

Uploaded by

Copyright:

Available Formats

arXiv:physics/0311011 v1 4 Nov 2003

RUSSIAN FEDERAL COMMITTEE

BASHKIR STATE UNIVERSITY

This book is a manual for the course of electrodynamics and

Referees: Chair of Algebra and Geometry of Bashkir State

ISBN 5-7477-0180-0 c Sharipov R.A., 1997

CHAPTER II. CLASSICAL ELECTRODYNAMICS. ......... 43.

CHAPTER III. SPECIAL THEORY OF RELATIVITY. ..... 68.

§ 2. Lorentz transformations. .............................................. 73.

CHAPTER IV. LAGRANGIAN FORMALISM

CHAPTER V. GENERAL THEORY OF RELATIVITY. ... 145.

REFERENCES. ............................................................... 162.

CONTACTS. ................................................................... 163.

Theory of relativity is a physical discipline which arose in

c = 2.998 · 105 km/sec.

In Chapter V Einstein’s theory of gravitation (general rela-

ELECTROSTATICS AND MAGNETOSTATICS

§ 1. Basic experimental facts and unit systems.

Quantity Unit Unit Relation

Units of measure for other quantities are derived from the

CopyRight c Sharipov R.A., 1997, 2003.

can be derived from proportions for basic quantities (see table

Measure unit of charge in SGS is

unit of charge in SGS = dyn 1/2 · cm= g 1/2 · cm3/2 · sec−1 .

Coulomb law itself then is written in form of the equality

whole. Force (1.2) reveals itself in form of chemical links only

unit of current in SGS = unit of charge in SGS · sec−1 =

Let’s consider straight conducting rod of the length l. Current

shaped parts are arranged par-

As we shall see below, constant c in (1.5) coincides with velocity

From the above consideration we see that SGS and SI systems

(1.11) γ ≈ 6.67 · 10−8 cm3 · g−1 · sec−2 .

This theory is based on Newton’s fourth law formulated as follows.

in both systems: in SGS and in SI.

§ 2. Concept of near action.

the framework of the concept of near action Coulomb law (1.2) is

CopyRight c Sharipov R.A., 1997, 2003.

electromagnetic field, and gravitational field. Strong and weak

Here r1 and r2 are radius-vectors of points, where charges Q1 and

Vector E depends upon the position of first charge and upon

vector E for quantitative measure of electric field produced by first

Now consider more complicated situation. Suppose that

Using superposition principle, one can pass from point charges

Force acting on a charge q in electric field is equal to the product

Field E(r) in (3.7) is external field produced by external charges.

§ 4. Lorentz force and Biot-Savart-Laplace law.

Upon calculating force acting on a segment of conductor of the

a force proportional to its charge q and to the magnitude of its

acting on it. Here τ is unit vector tangent to conductor and di-

where s is natural parameter on contour (length) and r(s) is

Here r is the distance from observation point to the conductor

Here τ is unit vector determining spatial orientation of infinitesi-

Like in (4.6), here s is natural parameter on the contour and r̃(s)

CopyRight c Sharipov R.A., 1997, 2003.

The concept of current density j is best one for describing such

Here n is unit vector of external normal to the surface ∂Ω

charge leakage through the boundary and conversely increase of

Current density j is a vector depending on a point of conducting

Note that Ω in (5.3) is an arbitrary domain that we marked men-

Biot-Savart-Laplace law for such conductors also is written in