Special Theory of Relativity by Ugarov

SPECIAL
THEORY
OF
RELATIVITY
VA. UGAROV
CnEUV!AJlbHA51
TEOPVI51
OTHOCVITEJlbHOCTY1
B. A. YfAPOB
H3,nATEJibCTBO «HAYKA:.
MOCKBA
SPECIAL
THEORY
OF
RELATIVITY
V.A.UGAROV
TRA:-.ISLATED
FROM
THE IWSSIA"l'
BY
YURI
ATANOV
MIR PUBLISHERS
MOSCOW
Firs! published 1979
Revised from the 1977 Russian edition
Ha QN2AUIICKOM RSI>!Ke
@ f.IIIBH&Il peAIHUH11 $H3HKO•M&TeMITH'IeCKOII: JIBTepaT)'pW

H3AITellbCTBa cHayKu, 1977
© English 1ranslal ion, Mir Publishers, 1979
PREFACE
It gives me pleasure to thank B. M. Bolotovsky and S. N. Sto-

lyarov who have writtC'n §§ 6.14, 6.15 of this book. I wish to ex-
press my special gratitude to V. L. Ginzburg. This book quotes
many things which I learned at the seminar led by him. A few
questions were discussed with him directly; in particular. the
problem of an energy-momentum-tension tensor should be men-
tioned. Finally, V. L. Ginzburg has written the article "Who
Developed the Special Theory or Relativity, and How?" to he
published in this book (Supplement 1). In my opinion, this article
gh·es very precise answers to questions which would be met by
anyone interested in the history of the STR evolution. I feel myselr
honoured to have this article included in the book.
The author
CONTENTS
!Preface
Chapter I. CLASSICAL MECHANICS AND THE PRINCIPLE OF RELA-

TIVITY • • • II
§ 1.1. A coordinate system and a reference frame in classical mechanics II

§ 1.2. The choice nf a reference lr3me. 14
§ I 3. Tile Galilean transformation . 15
§ 1.4. The Galilean principlt:> of relativity. Newlon's second law. 19
§ I 5. Newton's laws and inertial frames of reference. 24
§ 1.6. Absolute lime and absolule space 29
§ 1.7. How physics was approaching the theory of relativity • • 30
§ 1.8. The generalization of lhe Galilean principle of relativity • 33
§ 1.9 The velocity of light in vacuo . . 36
Chapter 2 THE EINSTEIN POSTULATES THE INTERVAL BETWEEN

EVENTS. THE LORENTZ TRANSFORMATION . 38
§ 2.1. Einstein's postulates 38

§ 2 2. The relalivislic frame of reference 41
§ 2.3. The direct consequences of Einstein's postulates (a few imaginary
experiments) 45
§ 2 4 The relalivily of synchronization of clocks belonging to two iner-
tial fran1es of reference. The direct derivation of the Lorentz
transfornlation 52
§ 2.5. The Lorentz transformation as a consequence of Einstein's pos-
~~ M
§ 2.6. The propagation of the light wave profile. An interval between
-§ 2.7. -
The-Lorentz transformalion as a consequence of the invariance M
of the interval between events . • • . • , . • • 63
-§ 2.8. Complex values in the STR. Symmetric designations . . • . . 65
'§ 2.9. A geometric illustration of the Lorentz transformation • • , , 69
Contents
Chapter 3. CONSEQUENCES OF THE LORENTZ TRANSFORMATION.

THE CLASSIFICATION OF INTERVALS AND THE PRINCI·
PLE OF CAUSALITY. THE K CALCULUS . 71
§ 3.1. On the measurement of lengths and time inlen·als. The relativity
of simultaneity 71
§ 32. Relativity of length of moving rulers (scales). A vis1ble shape
of obj&ts moving at relativistic velocihes . 74
§ 3.3. Relativity of lime inte1 vals between events . 83
§ 3.4. The classification of intervals and the principle of causahty 90-
§ 3 5. The transformation of velocity components of a particle on Iran·
sition from one inertial frame of reference to another . . . . . 94
§ 3.6. The transformation of an absolute value and the direction of the
velocity of a particle. , !Cl
§ 3.7. The K calculus (the radar method) . • 105-
Chapter 4. THE FOUR-DIMENSIONAL SPACE-TIME • 117

§ 4.1. Three-dimensional and four-dimensional Euclidean spaces. , 117
§ 4..2. The 4-space-lime, or the four-dimensional pseudo-Euclidean space I 18
§ 4.3. 4-vectors and 4-tensors • 120
§ 4.4. A pseudo-Euclidean plane . . . • . . . • 123-
Chapter 5. RELATIVISTIC MECHANICS OF A PARTICLE , 133

§ 5.1. A 4-velocity and 4-acceleration. . . 134
§ 52. A 4-force and a four-dimensional equahon of motion . . 140
§ 53. A three-dimensional relativistic equation of motion of a particle
(the second law of Newton in a relativistic form) . . 143
§ 5.4. The relativistic express1on for a particle's energy . . 149
§ 55. A 4-vector of energy-momentum . • 153
§ 5.6. The rest mass of a system. The binding energy . • 157
§ 5.7. Some problems of relativistic DJechanics of a particle . 161
§ 5.8. The conservation laws of relativistic mechanics. , 175-
Chapter 6. THE MAXWELL THEORY IN A RELATIVISTIC FORM. , 180

6.1. The three-dimensional system of Maxwell's equations. A 4-poten-
tial and 4-current . . 181
6 2. The transformation of a 4-potential and 4-current . • • • • , 184
6 3. An electromagnetic !ield tensor • 188
6 4. The transformation of electric and magnetic !ield components , 192
6 5. The electromagnetic 6eld invariants . . , 193
66. The Lorentz force, , •• , , 199
6.7. Covariance of the S}'stem of the Maxwell equations • , • 205
6.8 The Minkowski equations for moving media (the transformation
. oJ material equations) , • , . , , • • • , • • • , • • 208-
Contents
§ 6.9 The transformation of electric and magnetic moments . . 214

§ 6 10 Some problems involving the transformation of an electromag-
netic field . . . . . . • 216
§ 611. An en.ergy-momentum-tension tensor ol an electromagnetic field
in vacuo . . . . . . . . . 222
§ 6 12. An energy-momentum-tension tensor of an electromagnetiC field
in a medium. The Minkowski tensor and Abraham tensor . . 233
§ 6 t3. An energy-momentum-tension tensor of a spherically symmetric
-~ .m
§ 6.14. The field potentials in a moving non-conducting medium . . . 210
§ 6.t5. The field potentials in a moving conducting medium . . . . 246
Chapter 7. OPTICAL PHENOMENA AND THE SPECIAL THEORY OF

R,ELATIVITY . :153
§ 7.1. Properties ol plane light waves . • • . 258
§ 7.2. A 4-wave vector. The Doppler effect. Aberration of light . 2fi1
§ 7.3. A plane wave limited in space. The transformation of the plane
wave energy and amplitude . . . • . . . 2 5
§ 7.4. The pressure exerted by an electromagnetic wave (light) on a
surface • • . . . • • . . 210
§ 7.5. The light frequency variation on renection from a moving sur-
-~~ .m
§ 7.6. Light quanta (photons) as relativistic particles . 276
§ 7.7. Light quanta in a medium. The Vavilov-Cherenkov effect. The
anomalous Doppler effect • . . • • , • • . • • • • • . 280
Chapler 8. ON CERTAIN PARADOXES OF THE SPECIAL THEORY OF

RELATIVITY . 286
§ 8 I. Faster-than-light veloC'ities • . 287
§ 82 The thread-and-lever paradox • • • • 292
§ 8 3. Tho! tachyons • . 297
§ 8-1. The clock paradox • . . . . 303
§ 8.5. The "equivalence" of mass and energy. The zero rest mass . . 3t0
SUPPLE.'o1ENT . 3t7
I. Who developed the special theory of relativity, and how,
(~ L. GmzbJr·)) . 3t7
II. The unsuccessful senrch for a medium for the propagation of light 328
III. Was Michelson's experiment "decisive" for the creation of the spe-
cial theory of relativity? . . . . . . . . . 345
IV. Why shouldn't the mass-velocity dependence, or the relativtslic
-- -
mass, be introduced? . . , . 350
V. Non-inertial fram~s of reference. The :;pecial theory ot relativity
and the advance to gravitational theury (the general theory of
to Contents
MAIN EVENTS RELATED TO THE HISTORY OF THE STR . • 361

Appendix 1. • • • • • 362
§ 1. The symmetric notation. The summation rules • 362
§ 2. The transformation of coordinates in the case of a rotation of the
Cartesian system of coordinates • • . 364-
§ 3. The tensors • 368.
§ 4. The invariance of a 4-divergence and d'Alembert's operator. . 373-
§ 5. The convolution ("rejuvenation") of tensor indices . . 375-
§ 6. Some data on determinants. The dual tensors . . 377
§ 7. The stress tensor . • • • • • • . • 383-
§ 8. The rectilinear oblique-angted systems of coordinates . • 38&
§ 9. The definition of the hyperbolic functions and some relationships
between them • • , , • 392"
Bibliography to Appendix I • • • . • • . • • 393-
Appendix II. The basic formulae of electrodynamics in the Gaussian system 394
Bibliography • 399t
Index • . • . • . • • . . . • • . • . • . • . • • . • • • 40&
CHAPTER r
CLASSICAL MECHANICS
AND THE PRINCIPLE
OF RELATIVITY
§ t.l. A coordinate system and a reference frame in classical

mechanics. All natural phenomena happen in space and in the
.course of time, and an element of any phenomenon is something
Qccurring at a given moment of time and at a given point in space.
In the special theory of relativity* it is customary to refer to that
"something" taking place at a given point and at a given moment
.of time (in fact, something concentrated in a sufficiently small
volume of space and limited by a small time interval) as an event.
This definition shows that concrete features of an event may be
very different. That is why it is usual to indicate that "the event
.consists in ... ". The examples of events can be the emission of a
light signal from a certain point in space at a certain moment of
time, or the presence of a moving particle (a material point) at
.a given point in space and at a given moment of time.
When an event is realized, one says that it "happened" (or is
happening, or will happen). Any physical phenomenon represents
a sequence of events. A description of a separate event serves as
a basis for the description of any phenomenon and therefore we
begin with the description of a separate event.
To characterize a point in space where an event occurred, every
point in space has to be labelled before specific physical pheno-
mena arc analysed. But space is uniform and isotropic and this
implies that all points in space and all directions in it are equal.
It should be pointed out at once that we deal here with the free
space, or vacuum. The investigation of physical phenomena in
uacuo is of prime importance for the special theory of relativit}'.
Ewn though ,·acuum is a complex physical system, it is sufficient
for our purpose to assume that in the space domain which we
take for vacuum, no substance possessing a finite rest mass is
practically present and gravitational and electric fields are not too
strong.
But even when all points in space are equal, one can still
smgle out a certain point by placing a material object, i.e. an
~fter the complete term uspedal theory of relativity" will be some·
LLmes abbreviated as STR.
12 Spedal Theory of Relativity
object having a finite rest mass, in it. Points in space are usually
labelled by means of a coordinate system. With the help of the
material object we distinguish a point which is the origin of co·
ordinates The simplest coordinate system is the Cartesian syo;.
tern. Its construction begins with the tracing of three mutu.t!ly
perpendicular straight lines, i.e. the coordmnte X, Y, Z axes. !11
terms of physics, however, these are not just abstract strarg~rt
lines. Theoretically, the coordinate axes are rigid non-deformable •
solids. By the way, instruments, standards and other objects of
a given reference frame \\"ill be always fixed to them and there-
fore it should be borne in mind that a physical coordinate system
is always a material object.
In the Cartesian coordinate system points are quite easy to
label. From any point M in space one can construct the perpen·
diculars to the X, Y, Z axes or, in other words, project this point
on the coordinate axes. Having measured the distances of the
point projections from the origin along the X, Y, Z axes by means
of the chosen scale, 've obtain the numbers x, y, z, which are called
the Cartesian coordinates of the point. The distances can be mea':>·
ured via the step-by-step transposition of a unit scale along the
axis from the origin to the point projection on the axis. In fad,
such a procedure used for length measurements in everyday life
can also be used for determining the length of a stretch or an
object if it is at rest in a given coordinate system. As we shall
see later, the special theory of relativity furnishes a very con-
venient method of measuring distances without recourse to rigid
scales and their step-by-step transposition (see Chapter 2). Both
methods are equivalent, of course.
Thus through the introduction of the Cartesian coordinate S\ s-
tem every point in space acquires three numbers, that is the thi-ee
Cartesian coordinates x, y, z. The principal objective of physics,
however, is to study motion. Although mechanical motion is th•.!
simplest type of motion, its description requires time measure-
ments and therefore the coordinate system has to be of necesstly
supplemented by a clock. This clock is needed to register tile
occurrence of events at various points in space. How many clocl-s
are needed?
In classical mechanics they do not usually hesitate over the
answer to this question and l<'lcitly assllme that one clock restinJ;
in a given coordinate system is enough. It is useful to find o.1t
what this assumption implies. Let the clock be located at the
origin of the coordinate system. Events may happen at any points
. • The STR negates the existence of absolute solids (see Chapter 8) but lor
the coordinate axes it is just sufficient not to be very elastic.
Classical Mechanics and the Principle of Relativity 13
in space including those removed lar enough from the origin.

Then how can the clock, removed from the place where an event
happens, register that event? Obviously, just at the moment the
event occurs a certain signal has to be sent from the place of
occurrence of the event to the clock located at the origin. If the
veloc1ty of the signal is finite, it will reach the clock some time
after the onset of the event, and the time lag will depend on the
distance between the point where the event occurred and the clock.
In c\gssical mechanics, however, it is assumed that basically there
may be signals propagating infinitely fast. It is obvious that in
this case one clock ti.xed ngidly to any point in the coordinate
system will be enough.
ll JS implied that the onset of an event is registered as follows:
at the moment of an event occurring at any point in space a
signal is sent from that point to the clock, and the time of its
arrival is thus the time of the onset of the event (the velocity of
the signal is infinite!). The assumption concerning the infinitely
fast signals applies, of course, not only to the registration of
events. In Newtonian m!O'chanics it is incorporated intrinsically:
interactions between bodies are transmitted infinitely fast (see
§ 1.4).
Modern physics, however, claims that all signals (interactions)
<Jre transmitted at a finite speed; in other \Vords, there is a finit~
velocity of interaction transmission. How can this fact be recon·
cilcd with the evidence that Newtonian mechanics based on the
assumption about the infinitely fast signals copes excellently with
many problems (for example, calculates superbly the motion of
planets in the solar system)? The answer to this que!>tion is very
simple. The ultimate speed at which a signal, or an interaction, is
transmitted is very great. According to the contemporary ideas it
is the speed of electromagnetic waves in vacuo, which is equal to
approximately 3-108 m/s. It follows that as far as velocities of
objects to be considered are essentially less than that of light in
vacuo and characteristic distances are such that the time of light
propagation along them is negligibly small, Newtonian mechanic3
is correct and one clock is enough to register the time of events.
Yet it is at once clear that in the case of a fast motion (v ::::: c)
and extended systems a time of an event has to be registered
othenvise and the whole science of mechanics has to be based on
different premises. In fact, this is just what the special theory of
relativity does when it explicitly takes into account the finite
velocity of the interaction transmission.
Now let us get back to the classical pattern. A reference frame
is formed by a reference object with a coordinate system, a set
of length standards and a clock fixed rigidly to a reference obje1.·t.
ln physics a reference frame is always implied since any
14 Special Theory of RelaUvity
measurement tahen by an instrument produces a result that is

related to the reference frame in which this instrument is at rest.
§ 1.2. The choice of a reference frame. To tackle concrete prob-
lems, we choose a convenient reference frame and a convenient
coordinate system. How does this opportunity of choice come
about? As to a clock, in classical mechanics every reference frame
needs only one ideal clock. But a reference object, an origin and
directions of coordinate axes can be chosen at will. It is well
Known how this circumstance is utilized in geometry. For example,
+
the equation of an ellipse has the simple form x 2fa 2 y2Jb2 = 1
only if the origin is placed in the centre of the ellipse and the
coordinate axes coincide with its principal axes. No doubt, all
typical features of the ellipse remain for any other choice of the
coordinate system, but all formulae become immeasurably more
complex. It is important to point out here that in analytical geo-
metry the transition from one coordinate system to another varies
only the algebraic form of equations of geometric object~ while
the objects themselves naturally remain invariable.
Considering physical phenomena, one may also set up a coor·
dinate system rather arbitrarily. However, the two most signi·
ficant properties of vacuum space are implicitly meant in thi5
case: uniformity and isotropy. Uniformity is identity of all points
in space. This property is very essential. Actually, it enables us
to use physics. Laws of physics prove to be the same at various
points of the Earth, and everywhere within the solar system, for
that matter. But this is just what permits the origin to be placed
at any convenient point. When we turn a coordinate system
around the origin, we do not expect anything to change. This
implies that all directions running from a given point are identi-
cal in their properties. And this is exactly how isotropy of space is
defined. In classical mechanics, or, more precisely, in reference
frames where the Newtonian laws are valid (see§ 1.5), uniformity
.and isotropy of free space are assumed.
In contrast to geometry, in physics there is another choice of
reference frames: one may consider those moving relative to one
another. This is quite superfluous in geometry. But in physics the
reference frames moving relative to one another are the inevitable
occurrence. For example, physical experiments can be carried out
abord a spaceship and on the Earth. These are the two reference
frames, each of which may have instruments motionless relative
to the frame. As soon as we accept the reference frames moving
relative to one another, the two intrinsically different but funda-
mental questions crop up.
I. How does the motion of a reference frame affect physical
phenomena observed in it, i.e. do the physical laws change on
transition from one such frame to another?
m;~n~u~f~~~tr~~e~~=e~::u~g c~nc;e~~rt~~~s~~~~r!nhc~n1~~~.n a~a

obtain some values as a result of measurements of physical
quantities characterizing this phenomenon. The same phenome11on
can be observed in another reference frame moving relative to the
first one. The measurements conducted in the second coordinate
system will give us certain numbers defining the same physical
quantities. How do these quantities correlate?
It is important to note here that the same phenomenon is ob-
served in both systems. We must know how to correlate these
quantities. After all, a reference frame is an artificial constructiou
created for measurement purposes. The phenomenon itseli, just
as the laws of nature, cannot be affected by the choice of a ref-
erence frame. The natural phenomenon is the objective reality
existing outside our senses and measurements.
Of course, the results of measurements may prove to be different
in different reference frames but in any case we must know how
to convert the results of observations obtained in one frame into
those that are obtained, or can be obtained, in another. In short,
we need a method to transform results of measurements. How can
such a method be found?
The answer to the first question leads us to the principle of rel-
ativity and via Newton's laws helps to distinguish a special class
of reference frames, that is inertial frames (§ 1.5). The answer
to the second question is given by the rules for the transformation
of coordinates of an event, i.e. the Galilean transformation in
classical mechanics (§ 1.3) or the Lorentz transformation in rel-
ativistic mechanics(§§ 2.4, 2.5, 2.7).
§ 1.3. The Galilean transformation. The transition from one
reference frame to another one moving relative to the first one
was performed long before the advent of the theory of relativity.
Apparently the first to use the technique was Huygens who treated
in this way the problem of the collision of spheres. For the cake
of brevity we shall designate the reference frame by the letter K,
and provided there are several frames we shall introduce super-
scripts (K 0, K', K", ... ). We have mentioned that an event can
be considered as an "element" of some physical phenomenon. It
is natural to begin with the conversion of the quantities charac-
terizing an event when a transition from one reierence frame to
another takes place. From now on "the transition from one ref·
crence frame to another" will be everywhere understood as the
consideration of those reference irames which move relative to
one another. A shift of the origin as well as a rotation of coor-
dinate axes will not be taken for a "transition".
In an arbitrary reference frame K an event is described by the
four numbers: x, f:l, z, t. The first three of these are the coordinates
16 Special Theory of Relatlolty
of the point at which the event happened, while the last one
specifies the moment of time at which it happened. We want to
know how the same four numbers x', y', z', t' look in another rei-
erence frame K' moving relative to the frame K.
From the very beginning we are compelled to restrict our prob-
lem and consider only those reference frames which move uni-
formly and rectilinearly relative to one another, and do not ro-
tate around the origin. In other words, none of the considered
M reference frames moves with
K
K'( . v !~oth~~elef~!~~- r~~~~~h~~
\ r! __...-g•later it will become clear
\ ~ ;1 that we deal here with th::!
0 collection of so-called iner-
tial frames of reference.
However, since such refer-
ence frames can be discrimi-
nated only on the basis of
Newton's laws, we shall post-
pone the definition and iden-
tification of such frames till
§ 1.5. For the present, we
'' shall consider the two frames
~~g~·:i~· t~~ea~bi~r~~f~edi~~cft~dm:~e: /n:. K and K' moving uniformly
z and x', y', z'. The frame K' moves rela· and rectilinearly (translu·
th·e to K at the velocity V. The radius tionwise) relative to each
vector of the point M is equal to the other, in terms of geometry.
vector ,. m th~ frame K and to r' in the Let us assume that the re-
~:~~n ~~~~c:~d~'=.. ~. \~-he;:c~rissut~~ ference frame K' moves re-
radtus vector of the origin 0'. lative to K at a velocity V.
Suppose that at a given mo·
ment of time t the radius vector of the point M in the frame K' is
equal tor'. Then it can be seen from Fig. 1.1 that r' = r-R,
where ,. is the radius vector of the same point in the frame K, and
R is the radius vector of the origin of the coordinate system K'
taken from the origin of K. This relation is valid for any moment
of time and R varies according to the familiar law R = Vt + R,J,
where Ro is the radius vector specifying the location of the origin
0' at the moment of time t = 0. Taking into account that at the
moment t = 0 both origins coincide, R = Vt, and we obtain the
coordinate transformation law in the vector form:
r'=r-Vf, (I. I)
where the components of the vector V are defined in the frame K.
Now we can resort to isotropy of space and rotate each of the
systems K and K' around its respective origin. It is convenient
Class/tal Methanlts and the Principle of Relativity 17
to perform this in the following way. First, rotating the reference

frames, we orient the x and x' axes along the direction of the
relative velocity of these frames K and K'. Then rotating the
frames around the common axis x, x', we orient axes y, y' and
z, z' in parallel to each other. In such a way, having lost none of
the generality in terms of physics, we come to the relative posi-
tion of the coordinate systems shown in Fig. 1.2. In this case the
velocity V has the components (V, 0, 0). The origin of the sy,:;-
tem K' slips along the common axis at the velocity V, while at
Fig. 1.2. The two reference frames K and K' with parallel axes move relative
to each other at the velocity V (Vis the velocity of motion of K' rei alive to K).
In classical physics coordinates of an Mevent" are transformed from the rcfer-
el:ce fr11me K Ia K' according to the formulae of the "Galilean transformation":
x'=x-Vf, Y'=y, z'=z, 1'=1.
the initial moment of time both origins coincided. It is seen from

the vector formula (1.1) or directly from Fig. 1.2 that the rela-
tionship between the coordinates of the point M, where we believe
the event happened, in the systems K and K' is determined by
the following equations:
x'=x- VI, y'=y, z'=z.
Now, in order to establish fully what are the coordinates of the
event in the system K', one has to know the time of the event
by the clock of the system K' (now we have two clocks: one in
the system K and another in the system K'). Since in both systems
we employ infinitely fast signals, the finite relative velocity of
the systems is inessential to such signals. Indeed, the infinite w-
locity remains infinite in both systems. Consequently, the time ol
18 Special Theory of Relativity
the event registered by the clocks of both systems will be the

same, i.e. t = t'. This conclusion is confirmed by our own "com-
mon sense", because we do not detect any influence of the motion
on the clock rate in everyday life. But we should bear in mind.
however, that infinitely fast signals were only assumed, and al-
though the common sense does not deceive us in everyday life,
we must be prepared to the fact that in the case of a finite ve-
locity it may turn out that t =F t'.
But within the scope of classical mechanics, as we have
established by now, the formulae of transformation from the
'·coordinates" of an event determined in the system K (x, y, z, t)
to those of the system K' (x', y', ?!, t') may be written as follows:
x'=x-Vt,
(1.2)
z'=z,
t'=t.
Naturally, these equations are valid only for the relative position
of the reference frames shown in Fig. 1.2. The transformation of
the event "coordinates" from the frame K to the frame K', as
given by Eqs. ( 1.2), is called the Galilean transformation. We
would like at once to draw readers' attention to the fact that time
turns out to be the fourth coordinate of an event so that when
speaking of coordinates of an event, we imply four numbers (x, y,
z, t). This is done not only for the sake of speech brevity. In the
special theory of relativity such a terminology gains a complete
justification (see Chapter 4).
We have already pointed out the equivalence of frames moving
uniformly and rectilinearly relative to one another. The frames K
and K' that will be referred to from now on differ only in the
velocity of K' relative to K being equal to V, whereas the velocity
of the frame K relative to K' is equal to -V. Hence, in order to
get the reverse transformation formulae, it is sufficient to make
primed and unprimed quantities change place, having changed the
sign of V in the process. We get
x=x'+Vt',
y=y',
(1.3)
z=z',
t=t'.
Surely, these same equations can be derived in a direct algebraic
manner.
Note one of the consequences of the Galilean transformatior..
Suppose two events occurred in the frame K on the x axis: one
Classical Mechanics and the Principle of Refafiuily 19
at the point x 1 at the moment t~o and the other at the point x2 at
the moment t 2 (1 1 =f= 12). 15 it possible to select the frame K' in
v.hich both events would happen at the same point in space? Let
us find the x coordinates of these events in the frame K': xl =
= x1 - Vt 1, xl = x2 - Vt 2; and compose the difference x~-xl=x2 -
.__ x 1 - V (1 2 - t 1). Having bidden x2- xl = 0, we obtain the
equation from which the velocity of the frame K' relative to K is
determined: V = (x2 - x 1)/(t2- tt). The meaning of the result
is \'Cry simple: during the time period t2 - t 1 the frame K' suc-
ceeds in bringing the point xl to the place where the second event
occurred by a requisite moment. We see that it is always possible
to select a frame K' satisfying the required condition. It is pos-
silJle, however, only because classical mechanics permits the ve-
locity V to have any magnitude. In the theory of relativity, where
the velocity of a reference frame, just as any other material ob-
ject, is limited, the required frame is far from being always found.
Before proceeding to the Galilean principle of relativity, let us
agree on one term. For the ease of speech "various observers"
or "observers in different reference frames" are often mentione<:l.
In the past such a terminology provoked blustering arguments,
because there were many who imagined that it implied a ~objec
tive approach to physical measurements. But the presence of an
obser\·er is not at all mandatory as far as measurements are con·
cerned: they can be taken by means of instruments and without
man's assistance. It is indeed the case, for example, with space-
ships, even when there arc people aboard. "An observer from a
frame K" is, in fact, taken to mean a set of instruments resting
in this frame. One should not be surprised by the fact that instru-
ments placed in different reference frames will give different re-
sults for measured quantities associated with one and the same
phenomenon inasmuch as relative motion is a fundamental phy~i
cal quality. Objecti\'ity of la\\'s of nature manifests itself when
from results of observation in one reference frame one can find
results of observation of the same phenomenon in any other
frame. One may hope that after these remarks the appearance of
an "observer" on the pages of this book will not give rise to any
objections.
§ 1.4. The Galilean principle of relativity. Newton's second law.
The Galilean principle of relativity pertains to mechanical pheno-
mena exclusively; it was the first step toward the establishment
of the principle of relativity that later embraced all physics. Ga-
lilee noticed that uniform and rectilinear motion does not affect
mechanical phenomena. It is necessary to formulate precisely
what it means. As we already know, a reference frame is needed
to describe any physical phenomena, including mechanical ones.
Let us consider two reference frames moving uniformly and recti-
,. Special Tlteory of Relalfulfy
linearly relative to each other and Jet us conduct a "mechanical

experiment" in one of these frames. For example, we shall study
the motion of a mathematical pendulum or the free fall of bodie.,,
The principle of relativity states that identical experiments con·
dueled in the two frames mentioned yield identical results. Hence,
it is impossible to detect relative mot_ion of the frames by mean;
of such experiments. Of course, relative motion is easy to detect
in many ways provided the experiments of another kind are
undertaken. The first formulation of the principle of relativity can
be found in Galileo's book Dialogue Concerning the Two Chief
World Systems- Ptolemaic and Copernican (1632). This formu·
lation is of purely qualitative nature. We shall quote a short ex·
tract from this boo!< illustrating the essence of the problem:
"Shut yourself up with some friend of yours in the main cabin
below decks on some large ship, and have with you there some
flies, butterflies, and other small flying animals. Have a large
bo,vl of water with fish in it: hang up a bottle that empties drop
by drop into a wide vessel beneath it. With the ship standing
still, observe carefully how the little animals fly with equal spel!d
to all sides of the cabin. The fish swim indifferently in all direc·
tions; the drops fall into the vessel beneath; and. in throwing
something to your friend, you need throw it no more strongly in
one direction than another, the distances being equal; jumping
with your feet together, you pass equal spaces in every direction.
When you have observed all this carefully (though there is no
doubt that when the ship is standing still everything must happen
in this way), have the ship proceed with any speed you like, !.a-
long as the motion is uniform and not fluctuating this way and
that. You will discover not the least change in all the effects
named, nor could you tell from any of them whether the ship was.
moving or standing still."
The significant consequences follow from the qualitative for·
mulation of the Galilean principle of relativity, which is the iden-
tity of results of identical mechanical experiments conducted by
h\o observers moving uniformly and rectilinearly relative to
each other. Indeed, if the laws governing mechanical phenomena
are known and all identical mechanical experiments produce the
same result regardless of a reference frame chosen, the Jav..s of
mechanics must also be identical in such frames. In other words,
the equations of mechanics must be the same in all reference
frames moving uniformly and rectilinearly relative to each other.
Thus, the principal equations of mechanics written via coordi-
nates and readings of a clock of its own reference frame must
have the same form. At the same time it is clear that many quan-
tities vary on transition from one reference frame to another. In-
deed, let us examine the motion of a particle* in the frame 1{.

Usually it is defined as a time dependence of a radius veclor
r = r(t). According to Eq. (1.1) the motion of the s::~me partie!!!
in K' is defined by the variable radius vector r'(t) = r(t)- Vt.
Di!Terentiattng both sides of the last equation with respect to t
and taking into account that dr'/dt = v', and dr/dt = v, we ob-
tain
v'=v-V. (1.4)
Hence the velocity of the particle in the frames /( and K' i:;
.different. The quantities that vary on transition from one coordi-
nate system to another are called relative. Thus x coordinates and
the velocity of a particle arc relative quantities. Its acceleration,
however, is the same in both frames, K and K'. This bccom""
evident immediately after differentiating Eq. (1.4).
(~~· =4lf- (V=const).
The fact that the acceleration of objects is the same for all ob-
servers in frames moving uniformly and rectilinearly relative to
each other, is immediately evident. But this result makes it pos-
sible for us to understand the statement that "the equations have
the same form in all reference frames". The fundamental equation
of classical mechanics is that expressing the second law of Ne\V-
ton. This equation relates a force F acting on a body and the ac·
celeration acquired by it due to the action of this force:
m-M-=F; (1.5)
the factor m is called the mass of a body.

II the la\VS of mechanics in all reference frames moving uni·
formly and rectilinearly relative to one another are really the same,
Eq. ( 1.5) has to retain its form in all reference frames of this
kind. It is not difficult to see that this is really the case. We have
already shown that acceleration is the same in all reference
frames being investigated. But what happens to forces on tran-
sition from one reference frame to another? Suppose we investi-
gate two objects: I and II. Let the force of their interaction
depend on the distance between them, their relative velocity and
time. But the Galilean transformation does not change any of
these quantities. Indeed, let us write out the coordinates and vc·
• What is meant is a small objed possessing a mass, but still so minute that
there is no need to take into account its rotation. In mechanics, in thts case,
they speak of a m11ss point, but since we shall have to deal with points in space
far too much, the term "particle" is preferable.
Special Theory of Relaliuily
locities of objects I and II in the frames K and W. using the

<Jalilean transformation:
I
I___'_'"_'_"_'_"·--~--------------~----------
Coordloal0s aod
velocities of obje~ls Coordloales lo K 1
f
Tran;~k,'~r~~!~oo o
--;-1 x,, y,, ,,; v, Ix;~x,-Vt, u;~y,, ,;~,, I v; -••- V

~--x'_·_"_'·__''_;_•_·~x_;_~_x,_-_v_'·_•_~_~_"'_·_'_~=_'_'~_·_;_-_._,-__v__
At once it becomes evident that x2- x, = x2- xi, Y2- Yt =
= y2-yi, z2- z1 -z2-z\. And in so far as the distance
between the bodies is equal to
"tj(x2 x1F + (y2 Y1) 2 + (z2 z1)~ in the frame K
and
~(x; x;) 2 + (Y; y~) 2 + (z~ z;) 2 in the frame K',
it is clear that it remains constant on transition from K to K'. As
to the relative velocity,
i.e. it remains permanent. In accordance with the Galrlcan trans·

formation time is invariant: t = t'. Consequently, the forces de-
pendent on the variables cited do not at all vary on transition
from K to K'. But the forces considered in mechanics depend
either on a distance (gravitational forces, forces of electric inter-
action, elastic forces) or on a relative velocity (friction forces).
Hence, forces occurring in mechanics stay permanent under th~
Galilean transformation. Inasmuch as all the quantities appear-
ing in Eq. ( 1.5). accelerations and forces, do not vary under the
Galilean transformation, the fundamental equation of classical
mechanics, the second law of Newton, relating forces and accele-
rations, has the same form in the frames K and K' and differ<>
only in the designations or variables. (Surely, it is assumed that
a mass is a constant quantity; the mass invariance is one of the
basic postulates of classical mechanics.") The equation describ-
• Note that in Newtonian mechanics motion of bodies of a variable mass
can be exammed, e. g. when jet propulsion or motion o£ a drop accompanied by
conllensation is studred. But m all these cnses a body either donates substance
to the environment or acquires from it. When mass variability is mentioned rn
the STR (see Supplement IV), it is implrcd that a mass of a body stays con-
stant in the reshng frame, i.e. there is no mass exchange bet\\een a body and
its envrronment.
Ctassicat Mechanics and the Principle of Relativity 23
ing the second law of Newton in the frame K has the following
form, provided the force depends on a distance and time:
m ~;~ = F (r 12, f).
Accordingly, in the frame K'

m' ~;~ = F (r;2, {).
An equation which does not change in case of a transforma-

tion of variables appearing in it, i.e. an equation with its terms
invariant, is called invariant with respect to a given transforma-
tion. Thus we have shown that the equation describing the second
law of Newton is invariant with respect to the Galilean trans-
formation.
Now we can formulate more precisely the terms on which
"identical experiments produce identical results". Newton's Eq.
( 1.5) is an ordinary differential equation of the second order. Its
solutions describe motion of the system. To make the solutions
of Eq. ( 1.5) coincident in the frames K and K'. i.e. to ensure the
"identity" of motion, it is necessary for the initial conditions to
coincide. The invariance of the basic equation of mechanics en-
sures that mechanical phenomena proceed alike in all reference
frames moving uniformly relative to one another, only when the
initial conditions coincide in these frames.
When the initial conditions for the same phenomenon differ in
diverse reference frames, the phenomenon itself will look different.
For example, while a raindrop falls down vertically from the
viewpoint of an observer standing on a platform, the same rain·
drop will move along a parabola from the viewpoint of an ob-
server in a train. (We suppose that a raindrop falls down with
an acceleration.) However, the initial data in these frames were
different. From the viewpoint of an observer in a train the raindrop
had initially a horizontal component of a velocity. An observer
standing on a platform had to assume that at the initial moment
a raindrop had no horizontal component of a velocity whatsoever.
We have already mentioned above that in Newtonian mechanics
an interaction between bodies is assumed to be transmitted in-
finitely fast. Now we are able to explain this in detail. "Interac-
tion" between bodies is specified by forces. In classical mechanic~
forces are regarded as being dependent on distances between bo-
dies. The same is assumed to be correct for bodies moving relative
to one another. But a distance between two moving bodies has.
to be put down as
r 12 ~ .,f,.-[x.,-2 "(t),--,x-,d"'t)"'J''+'l.,..y2'("1)-y._,"(t")]2'+.--c:lz',("l)-z=-,"VJ"'l'·
Special Tlleory of Relativity
Assuming that an interaction, that is a force, is transmitted at

a finite velocity, one cannot presume that the equation defining
this force still incorporates r 12. If we want to find the force that
is exerted on body I by body II, the position of body II should he
registered not at the moment t but earlier by the time interval
needed for the interaction to be transmitted from body II to body
/. When this time lag is ignored, it means that the velocity 3t
,,·hich the interactions are transmitted is assumed infinitely fa~t.
It b precisely how problems are tackled in Newtonian mechanics.
The same thing happens when a potential energy is introduced.
Having written down a central force for the interaction of two
particles as usual in the form
F~- VU(I r, (I)- r,(t) 0.
we explicitly ignore the time lag in the interaction transmission.
The instantaneous transmission of interactions, formerly re-
ferred to as a long-range action, appears amazing and obscure
to us. The transmission of any signal, i.e. an impulse or energy
capable of accomplishing some action, e.g. switching on a certain
device, requires some time. As our experience teaches us, it i->
impossible to transmit a signal from "here" to another place
("there") instantaneously. Yet in Newton's time no other idea
except the long-range action could emerge, as far as a transmis·
sion of interactions is concerned. The finite velocity of the trans-
mission of interactions appeared together with the concept of a
field which was introduced into the theory of electromagnetism
by Maxwell. fn Maxwell's theory the interaction of charges or
currents is realized through a field to which an independent ex·
istence is attributed. rt follows from the theory that a field propa-
gates at a finite velocity. This means that the velocity of propaga-
tion of interaction is the same as that of the field. The propaga-
tion velocity of an electromagnetic field in vacuo plays a funda-
mental role in the theory of relativit~·· It is designated by the
letter c and is approximately equal to 3-108 m/s. Since field vari-
ations are transmitted from point to point, field theories are
referred to as short-range ones. As we shall see later, the theory
of relativity rejects the long-range action as a matter of principle.
§ 1.5. Newton's laws and inertial frames of reference. The basic
laws of mechanics, Newton's laws, make it possible for us to
distinguish among all conceivable reference frames, the speci<~l
class of frames in which not only laws of mechanics but also all
other physical laws look particularly simple. These are the so-
called inertial frames of reference. _An inertial frame of reference-
is a irame (or rather frames, since it will turn out later that there
are an infinite number of them) in which all three laws of New-
ton arc valid.
Classical Mechanics and the Principle of Relaliuify 25
We begin by showing how important the first law of Newton

is for the discrimination of inertial frames of reference among all
others. The first law of Newton, the law of inertia, claims that a
body subjected to no forces moves due to inertia, i.e., uniformly
and rectilinearly. Frequently one could hear, or even read in a
textbook, that the first law is not an independent statement, but
only a consequence of the second law.
Formally it is the case. The resultant of all forces acting on
a body appears in the right·hand side of Eq. (1.5). The second
law just claims that an acceleration acquired by a body is direclly
proportional to this resultant and inversely proportional to its
mass. It follows from Eq. ( 1.5) that if the resultant of all forct•s
is equal to zero, or there are no forces whatsoever, the body gains
no acceleration. And if a body gains no acceleration, it either
moves uniformly and rectilinearly or is at rest. It used to be con·
eluded from this that the law of inertia could be obtained from
the law of dynamics.
Then why was it necessary for Newton to formulate the law
of inertia separately? It is doubtful that Newton d1d not realize
that the law of inertia is a consequence of the law of dynamic->.
The problem is more complicated than it may seem at first sight
Newton understood very well that neither Eq. (1.5) nor the la~v
of inertia can be equally valid in all reference frames. It is not
accidental that the definition of an inertial frame involves all
three laws of Newton. Let us recollect the third law: to every
action there is an equal and opposite reaction. This Jaw empha·
sizes that all forces in Newtonian mechanics are intrinsically a'>·
sociated with an interaction between bodies.
Let us examine one useful, even though very plain example.
Let a body be at rest in an inertial frame of reference K. Th~:n
according to the second law of Newton no forces act on this body.
Without touching it let us consider it from the viewpoint of an
observer moving relative to the frame K with an acceleration a.
This observer will note that the body in question moves relative
to him with an acceleration -a. If the second law of Newton were
valid in his frame, he could say that the body experiences th~
force -ma. But we know from the observer in an inertial frame
of reference that there is no force acting on the body. Therefore,
the second law of Newton is merely not valid in the reference
frame moving relative to the inertial frame with an acceleration.
Many readers have already realized, of course, that passing into
the reference frame moving at an accelerating velocity, we detect
"a force of inertia" which is not adually a force in Newtonian me·
chanics (see Supplement V). Since the Jaws of Newton are not
valid in all reference frames, Newton had to point out that a cer·
lain reference frame was available in which all these laws were
valid. And the first law of Newton is, in fact, equivalent to thi<;
statement. This law postulates that an inertial frame of reference,
i.e. a reference frame in which the law of inertia is valid, is
available. In other words, one can find a reference frame in which
a body that interacts with no other bodies, moves due to inertia,
i. e. uniformly and rectilinearly.
The law of inertia represents a special case of the law of con-
servation of momf"ntum. On the one hand, it is a consequence of
the second and the third laws of
Newton, and on the other hand, a
The Earth consequence of the second law anrl
rotation axis
the assumption about uniformity of
~
space (equivalence of all its points),
i.e. Newtonian mechanics assumes
the uniformity of space in any iner·
tial frame of reference.
~
Now suppose we have found one
inertial frame of reference. Then ac-
cording to the Galilean principle of
Fig. 1.3. The Foucault experi· relativity all reference frames mov-
menl rlcsign.:!d to dele<:! one of ing uniformly and rectilinearly rela-
IFRs. For the sake of simplicity tive to it will be inertial as well.
the drawing illustrates the Fou· Therefore, it is clear that there is an
~~~t aetf~;~~~~~ 1~eir~~t. f~~r~;.' infinite number of inertial frames of
pcrimenl was conducted in Pa· reference.
ris. but this circumstance does How does one find at least one
not change the matter. inertial frame of reference? Of course,
the discovery of such a frame is a
matter of experience. The famous pendulum experiment first con·
dueled by Foucault is suitable for the purpose. For the sake of
simplicity we shall describe the experiment the way it could be
conducted at one of the Earth's poles (Fig. 1.3). A heavy ball is
suspended on a thread which is attached to a frame constructed
at the Pole. The point of the pendulum suspension is located on
the Earth's axis. The attachment of the thread is free and so th~
frame does not carry the thread along in the process of rotation
around the Earth's axis. The equilibrium position of the pendulum
thread coincides with the Earth's axis. If one deflects the pendu-
lum from the equilibrium position and then lets it go without im-
parting an initial velocity, it will start oscillating in a certairt
plane. The two forces acting on the pendulum are the gravita-
tional force mg and that of the tensile stress T of the thread. Bolh
forces act in the plane P of the pendulum oscillations and cannot
remove the pendulum from that plane. If the second law of Newton
were strictly valid on the Earth, the plane of the pendulum oscil-
lations would maintain its orientation relative to the Earth. But
C/(J.ssic(J./ Mech(J.nics (J.nd the Principle of Re/(J.fivillf 2i
in the experiment the Earth withdraws from under the pendulum

<Jnd thereby "registers" the fac! that in the coordinate system a<>-
~ociated with the Earth the E<:cor.d law is not, strictly ~peaking,
valid.
One should not be too m11ch annoyed hecause of this, since
Newton's laws still can be usee on the E~rth to great advantage.
It is evident, for example, frorr: the fact that the whole engineer-
ing and theoretical mechanics rely on the second Jaw of Newton
without any corrections. Surely, this is because corrections are
small: they are caused by the Earth rotation which is not very
fast. Therefore, the Earth can be treated as an i-:Ntial frame even
in a school textbook.
But fundamentally the Earth is not an inertial frame. An iner-
tial frame involves such a coordinate system relative to which
the plane of pendulum oscillations remains constant. This plane
can be found from the same Foucault experiment. The system
turns out to be rather "exotic". Its centre is located in the Sun
and the three coordinate axes arc directed to the "stationary"
stars, i.e. the stars moving rigidly together with the so-called
celestial sphere. Due to the singular role of the Sun the inertial
frame based on this system is referred to as heliocentric. In the
choice of an inertial frame most important is the choice of direc-
tions for coordinate axes. The choice of the origin in the centre
of inertia of the Sun is convenient because the Sun possesses the
largest mass in the whole solar system. The motion of planets
appears particularly plain in this frame. Note that the axes of the
heliocentric reference frame do not participate in the rotation of
the Sun. By the way, the reference frame with the coordinate axes
fixed rigidly to the Earth, i.e. rotating with it, is referred to as
geocentric. As Foucault's experiment showed, this frame is non-
inertial.
Thus, the Newtonian laws of dynamics are applicable in the
heliocentric frame. In accordance with the Galilean principle of
relativity the laws of Newton are equally valid in all reference
frames which move uniformly and rectilinearly relative to the
heliocentric one. We shall refer to all these reference frames as
inertial frames of reference*. Although the number of inertial
frames of reference is infinite, they still get lost among all feasible
kinds of frames. If it were possible to gather all kinds of frames
into a sack and then to draw out of it one frame at random, we
would get most likely a non-inertial frame.
Foucault's experiment is far from being the only one permitting
of detecting a deviation of the geocentric reference frame from
• Hereinafter we shall often abbreviate the term '"inertial Frame oF reference"

to the initial letters, i. e IFR.
Spec/at Theory of RetallfJ/Iy
an inertial one. We shall indicate another experiment of lhts kinO.

When a heavy object is dropped from some height, it does not
fall vertically down as it should due to the gravitational force,
but deviates slightly to the east. The deviation of the motion of
free falling objects from the vertical makes it possible to detcd
the non-inertial nature of the geocentric frame and to find a:t
inertial frame of reference.
In mechanics there is one more conServation law for closed-typ~
systems which is the conservation of moment of momentum. It is,
just as the law of conservation of momentum in a closed-type
system, the consequence of the second and third laws of Newton.
Moreover, it can be obtained as a consequence of the second law
and the assumption about the isotropy of space. This implies that
Newtonian mechanics presupposes the uniformity of space.
Tile law of conservation of energy for closed-type systems turns
out to be a consequence of the second law of Newton and the as-
sumption about a potential character of forces acting betweE-n
particles constituting the system. On the other hand, it stems
from the motion equations of the system and the assumption
about the uniformity of time. It follows that in Newtonian me-
chanics the uniformity of time is presupposed.
That is why an inertial frame of reference can be determined
as one relative to which space is uniform and isotropic, and time
uniform.
Sometimes an inertial frame of reference is defined as a frame
fixed rigidly to a free-moving object. Although this definition is
basically true, it cannot be practically used for the purpose of an
experimental identification of an IFR. There is no "free-moving"
object at our disposal, since the gravitational force cannot be
cancelled. Consequently, it is more correct to define an IFR as a
frame in which all three Jaws of Newton are valid.
Inertial frames are distinguished among other, non-inertia!,
reference frames not only in mechanics. An electric charge does
not radiate electromagnetic waves when at rest in an inertial
frame of reference, whereas in a non-inertial frame it does.
Inertial frames of reference play a tremendous role in physics.
It is for these frames that familiar laws of physics are recorded.
The transition to non-inertial frames is associated with consid-
erable difficulties. The special theory of relatiYity instructs us
how to describe all kinds of physical phenomena in any inertial
frame of reference"'. But what does it mean and how is it practi-
• It must be strE-ssed that the STR can also be formulated for non-inertial
framt>s of reft>rence In fact, the STR can be employed in any reference frame.
as long as there are no gravitational forces. i. e. in a plane lour-dimensional
space-lime. However, the form of the STR suggested by Einstein and to- be
developed in this book is applicable only to inertial frames of reference. ,
Classical Mechanics and the Principle of Relativity 2)
cally carried out? We have much to discuss before we can get

answers to these questions.
§ 1.6. Absolute time and absolute space. Although in deriving
the Galilean transformation we have, in fact, already spoken or
everything that was meant in that transformation by space and
time, we shall repeat the pertinent statements. Usually, when
"classical" physics is mentioned, Newtonian mechanics is implied
for the views of Newton of space and time reflect precisely the
classical approach to these concepts. Newton's ideas are worth
dwelling on more carefully because they correspond to our every-
day experience. and are customary and comprehensive, while the
transition to the concept of space and time inherent in the speci~l
theory of relativity presupposes renouncing these ideas. In addi-
tion. a still more decisive step further from thesP concepts wa<>
made by Einstein in his theory of gravitation which is sometimes
refNred to as the general theory of relativity. This is what one
can read in Newton's Philosophiae Natura/is Principia Mathema-
tica (1687) · "Absolute space, in its own nature, without relation
to anything external, remains always similar and immovable."
So. according to Newton, space represents a giant empty box
which contains material objects and where physical phenomena
take place. At the same time Newton was aware that the Galilean
principle is valid in mechanics. And this indicates that the states
of immobility and uniform rectilinear motion are equivalent. Then
how should one single out .. motionless absolute" space?
Of course, it is impossible to single out "motionless absolute''
space just by observing mechanical phenomena. Detection of ab-
solute space and absolute motion involves studies outside the
scope of mechanics. Such a detection is assumed to be possible
in the process of interpreting optical phenomena. Consequently,
in the historic essay dedicated to the interpretation of some ex-
perimental facts {see Supplement II) we shall presume that New-
ton's privileged, selected reference frame, that is motionless abs'l-
lute space, is the heliocentric frame. Finally, it will be clear, though,
that there is no such thing as a privileged frame at all, but ther<>
is a whole privileged class of reference frames in which laws of
physics appear particularly simple. Such is the class of inertial
frames of reference.
Now let us see what Newton wrote abm1t time:
"Absolute, true, and mathematical time, of itself, and from its
o-u n nature, flows equably without relation to anything external,
and is otherwise called duration."
Again we come across the statement that time is something
external relative to nature. Thus, in accordance with Newton':'
ideas, time and space exist by themselves and do not depend on
material bodies located in space. Surely, Newton's concepts of
Special Tlleory of Relativity
space and time seem very scholastic lo us. l·lov,.evN, they ::.houtd
not be underestimated. Here is a short excerpt from the book [II J:
"In conversations with one of lhe authors of this book at ,·ar-
ious limes over the years, Einstein emphasized his great respert
for Newton and, in particular, his admiration for Newton's cour-
age. He stressed that Newton was even better aware than his
17th century critics of the difficulties with the ideas of absolute
space and time. However, to postulale those ideas was the only
practical way at that time to get on with the task of describing
motion".
Of course, the natural question arises: why does classical me-
chanics based on such concepts of space and time that can hardly
be explained, function so efficiently? It turns out, however, that
these concepts are approximately correct and the departures from
them in everyday life are quite insignificant. The departures from
the classical ideas become clearly visible only when micropartic-
les are investigated and also in outer space conditions which
modern physics has already begun studying. Such investigations,
however, require special conditions and sufficiently complex equip·
men!.
To end this brief section it is necessary to give a concise pre-
sentation of the up-to-date approach to the problem. From the
modern point of view there is no absolute space and, consequently,
no absolute motion. All inertial frames of reference are equivalent.
The special theory of relativity shows that time readings for
events prove to be different in different inertial frames of reference.
Thus, time reading is found to depend on the state of motion. The
gravitational theory of Einstein goes still further. In terms of this
theory properties of space and time are not prescribed for ever
but are specified by objects located in space. Since in accordance
with dialectic materialism space and time are forms of existence
of matter, the conclusions of Einstein's theory of gravitation ap-
pear far more satisfactory than the Newtonian concepts of space
and time.
§ 1.7. How physics was approaching the theory of relativity.
From the point of view of modern physics it is useful to tra::e
how relativistic effects were showing well before the crea-
tion of the special theory of relativity. This section does not
claim to be an historic essay (Supplement II is closer to that). It
is intended only for promoting the understanding of the next two
sections, where, in fact, the first principles of the theory are pre-
sented.
No doubt, the first step to the development of the special theory
of relativity was the discovery by Galileo of the principle of rela·
tivity for mechanical phenomena.
The natural question arises: why did Galileo confine his prin·
Classical Mechanics and llle Principle of Relativity 31
ciple within the framework of mechanics? The answer is very

straightforward: in Galileo's times there were just no "other
branches of physics" as we call them now. In fact, mechanics re.
presented the whole physics. If one also takes into consideration
~~=t b!~\t~fs~:~h~~f~s0 ~t~:;~stt~~f t~~te~~t~1 :~e b~9t~x~~a~~~:y. 0~
becomes clear that the principle of relativity formulated by Ga·
lileo encompassed the "whole physics" at that time.
The next important step along the road to the theory of rela·
tivity was the establishment of the finiteness of the velocity of
light. The conclusion was made by Roemer on the basis of his
astronomical observations ( 1676). Before Roemer the velocity of
lig;l~epr~~fifea:~onp~~~i;l~sugre~e~~t~~if;fi~~~el.d be expressed in a
mathematical form only after equations of mechanics had been
written down (Newton, Philosophiae Naturalis Principia Mathe·
matica, 1687). Since coordinates and time are the basic variables
in\'olved in equations of mechanics, their transformation on tran·
sition from one reference frame to another, moving relative to the
former, requires appropriate equations for transformation of co·
ordinates and time under such a transition. It followed from the
~al~~~~~iJ~~~~i~l~d o!i~~la!~~~1d t~~~ ~~~e~e~~~sif~r~a~tf~~':~~~.~~
laws (§ 1.4). This transformation is that of Galileo.
th~nP!~~~e~~e i:0~~:i~~t J~i~~u~~fini~:fyerj~~~~s~~~~eSe[~~r~!~th~!
rotation and indicated the inertial frame of reference (§ 1.5). In
fact, one could conclude the description of mechanical phenomena,
linked directly with the theory of relativity, with this experiment.
The Galilean principle of relativity, Newton's laws and the Ga·
lilean transformation are all closely interrelated. The direct con·
sequence of the Galilean transformation is the classical formula
for the velocity transformation (Eq. (1.4)): v' = v- V. In 1851
Fizeau performed an experiment to siiOw explicitly that this for·
mula is not always correct. The Fizeau experiment with flowing
water was schematically conducted as follows. In the reference
frame K water was flowing along a tube at a velocity V, and the
velocity of light in water was being measured. Now we can reason
strictly in terms of kinematics. Let us fix the inertial frame K' to
mo\'ing water. In this frame the vrlociiy of light v' is determined
by the familiar relationship v' = cin. where n is the refraction
index of water. To find the velocity of light in tl1e frame K. one
can use Eq. (1.4), and then v = c/tl +
V. But Fizeau's result,
confirmed also by modern mea,urements, turned out to be
v=f+V(l-~).
32 Special Theory of RelaUvily
As it is seen from here, the classical Eq. ( 1.4) is not correct in

this case. And this is exactly what we wanted to emphasize. As
to the details of the experiment and its contemporary interpreta-
tion. all that can be found in § 3.6.
The theory of relativity owes very much to the Maxwell theory
~h~m~,~-~t-~~~~~! ir,::~~~~at(:SS~~1 ;;J~~~e~h:~d ~!!e\tfeu~:~:~e~eld
theory in which an interaction was supposed to be transmitted at
a finite velocity, that is the field propagation velocity. The theory
provided a quite definite value for this velocity which was speci·
fically equal to I/-1/eoJJ.o in vacuo, where e0 and Jlo are the electric
and magnetic constants. Naturally, the question arose at once as
to whether the principle of relativity was satisfied; in other words,
whether the Maxwell equations retained their form under the
Galilean transformation. One can easily check that the Galilean
transformation changes the appearance of the Maxwell equation;;.
Owing to this fact it was suspected that the principle of relativity
did not extend to dynamics. Several decades were needed to r~
alize what wonder of a theory Maxwell de,·eloped. Neither knowing,
nor even suspecting anything about the theory of relativity, Max-
well nevertheless developed his theory in a complete agreement
with the requirements of the theory of relativity.
Now, when we know full well where the influence of the theory
of relativity can be "detected", it is easy to come back to essen-
tial facts. The theory of relativity reveals itself when velocities
of objects get closer to that of light in vacuo, such velocities being
referred to as relativistic. However, there are no macroscopic
objects possessing relativistic velocities. Only microscopic par-
ticles can travel at velocities close to that of light. The first micro-
particle to be discovered was an electron (Thomson, 1894-1896).
Thomson determined the ratio of the charge of an electron to its
mass experimentally. His experiments were carried out in dis-
charge tubes where electron VP.Iocities were far below that of light.
But in 1896 the natural radioactivity was discovered. The stream
of electrons found among radiations emitted by radioactive sub-
stances was very soon identified with electrons in a discharge
tube. Velocities of these electrons turned out to be close to that
of light. When in 1902 Kaufmann investigated the motion of such
electrons in electric and magnetic fields, the classical equation of
motion, i.e. the second law of Newton, was found to describe their
behaviour incorrectly. Thus, the departure from the Newtonian
laws was observed for the first time.
Summing up, it can be said that by the beginning of the 20th
century it became obvious that Newtonian mechanics and the Ga-
lilean transformation are not always true, and the fastest signal->
Classical Mechanics and the Principle of R.elaliuify 33
of all known, that is light signals, are transmitted at a finite

velocity.
Although there are no macroscopic objects moving at a relati-
vistic velocity, one relativistic object, light, was always at men's
disposal. Naturally, optical experiments played a significant role
in the history of the STR: the interpretation of optical experi-
ments is associated with the emergence of a hypothesis concern-
ing a "luminoferous medium". Rejection of this hypothesis took
much effort, but now it is worth mentioning only as a page in the
history of physics (see Supplement II).
§ 1.8. The generalization of the Galilean principle of relativity.
The Galilean principle of relativity covered only mechanical ph-e-
nomena. We found that the second law of Newton expressed in
a differential form in combination with the Galilean transforma-
tion satisfied the principle of relativity. From the formal point of
view it implied that Eq. (1.5) remained invariant and only desig-
nations of variables changed. Naturally, the question arises: wily
must the principle of relativity cover only mechanical phenomena?
Why is it impossible to believe that all physical phenomena hap-
pen in the identical manner in all inertial frames, provided th~
initial conditions of these phenomena are identically specified? In
other words, why is it impossible to assume all inertial frames of
reference to be completely equal with respect to all physical phe-
nomena?
These questions did not worry physicists too much till the
middle of the 19th century since they reduced all physics to me-
chanics. But by the middle of the 19th century it became evident
that physics cannot be reduced to mechanics. By the same time
the conviction had grown as to the universal relationship between
phenomena, and between physical phenomena in particular. The
subdivision of physics into "mechanics", "electricity", "heat" etc.
is justified by the fact that each group of phenomena possesses
its own set of basic equations and so is caused by rather educa.-
tional requirements and is not intrinsically imperative. Looking
more carefully into even "purely mechanical" phenomena, one can
discern a manifestation of regularities of another kind. The colli-
sion of billiard balls is always cited as a classical example from
mechanics. But at the moment of collision, when the balls are
slightly flattened, the elastic forces defined by electromagnetic
forces come into play. Hence, no "purely mechanical" phenomena
can exist in nature. It follows that the principle of relativity must
either cover "all physics" or be wholly incorrect.
Thus, the extension of the principle of relativity to all physical
phenomena wa.s quite natural from the viewpoint of physics at the
end of the 19th century. But such generalization of the Galilean
2-97
principle of relativity is exactly what is called the first postulate

of Einstein, or the Einstein principle of relativity.
However, the equations of electrodynamics were at once found
to contradict the equivalence of inertial frames of reference.
First of all, so far as the basic system of electrodynamic equa·
ttons, that is the Maxwell equations, is concerned, they alter their
appearance unde~ the Galilean transformation, i.e. do not retain
their form, and it follows from here that electromagnetic pheno·
mcna are descrilled differently in different IFRs. In other words,
electromagnetic phenomena do not obey the principle of relativity.
In particular, this means that in the reference frame in which th('
Maxwell equations are written down in the conventional form
(see Chapter 6), the propagation velocity of electromagnetic
waves c is equal to J/,Ye 0JJ. 0, while in all other reference frames
moving relati\e to the first one, the velocity is different. But
vacuum occupies a special place relative to reference frames. In·
deed, it is remarkable because it has no "medium" possessing
I he rest mass. One can always fix a reference frame to a materi<1l
medium, i.e. single out such a frame in which the medium is :1t
rest as a whole or in a limited region. But this particular reference
frame is the chosen one. Another equivalent frame of reference
moving relative to the first one must possess the same property
of the motionless medium. But this creates a different physical
situation. Thus, the presence of a medium always distinguishes
one reference frame from all others. But it is impossible to single
out such a system in vacuo because there is no reference frame
in which vacuum is al rest. Consequently, all reference frame<>
are equivalent relative to vacuum. It follows logically from here
that prodded all inertial observers are equal, the velocity of
electromagnetic waves must be the same, 1/,Ye0 J.Lo, in all IFRs.
As to the classical formula for the transformation of velocities,
Eq. (1.4) shows that this is not the case. Let in an inertial frame
of reference K the velocity of light in vacuo be equal to c. Then
in another inertial frame of reference K' the velocity of light in
t•awo c' is equal to c- V. Hence, the velocity of light in vacuo c
is equal to 1/.YeoJ.Loonly in one privileged reference frame. Thus,
the principle of relativity seemed to be incorrect for electromag·
netic phenomena.
The above reasoning was based on the fundamental assumption
which was absolutely unacceptable for the 19th century physic~:
electromagnetic waves, i.e. light, can propagate in vacuo or, ex·
pressed otherwise, no matter is needed for their propagation. This
is a very difficult point to comprehend, when a transition from
classical physics to relativistic is undertaken.
Classical Mechanics and /he Principle of Relativity 35
But what could be done in such a situation? Logically three

possibilities were opening.
(I) The principle of relativity could be assumed to cover only
mechanics and have nothing to do with electrodynamics in which
there is an "absolute" frame of reference. But, as it was men-
tioned before, such a possibility is rejected when the general re-
lationship of physical phenomena is taken into consideration.
(2) The principle of relativity could be regarded to be univer-
sally applied, and inasmuch as the system of Maxwell's equations
does not satisfy this principle, that is it changes its appearancE'
under the Galilean transformation, it should be discarded. But
the system of Maxwell's equations showed itself as a reliable and
comprehensive theory within one inertial frame of reference, a la-
boratory frame. On the other hand, Newtonian mechanics and the
Galilean transformation associated with it did not prove to be
always correct. Because of this it would be reasonable to keep th~
system of Maxwell's equations.
(3) If the principle of relativity is assumed to be applicable
to all phenomena of nature, and the system of Maxwell's equations
correct, the transition from one inertial frame of reference to
another cannot be described by the Galilean transformation which
changes the form of Maxwell's equations. On the other hand, a
new transformation cannot leave the form of equations of me-
chanics intact. Consequently, the equations of mechanics have to
be changed so that the new transformation leaves them intact.
The last possibility formulates concisely the programme which
is realized by the special theory of relativity: (I) the principle of
relativity covers all phenomena of nature, (2) the velocity of
electromagnetic waves in vacuo is the same in all IFRs (thic;
follows from the invariance of Maxwell's equations).
But how must a transformation of coordinates and timE" look
like in order to meet both requirements set above? Such a trans-
formation will turn out to be the Lorentz transformation and we
shall examine it closely in the next chapter. In conclusion, we
shall point out the following.
As soon as the Galilean principle of relativity was extended to
cover all physical phenomena, it turned into a genuine principle
of physics. Evidently it is advisable to differentiate laws and prin-
ciples of physics. When laws of physics are spoken of, their vali-
dity for a limited scope of physical phenomena is implied. For
example, Newton's laws describe phenomena of mechanics. Max-
well's equations pertain to electrodynamics: and so they are the
laws of electrodynamics. The three laws of thermodynamics deal
with thermal phenomena. As to the principles of physics, they
are universally important, for they cover all physical pheno-
mena.
,.
36 Special Theory of RelatitJity
The most widely known principle of physics is that of conser·

~~~~0:rv~{i:~e~fY~n~:;Y f~sm~~fie~o~kheb~,t;;;,i~};"~P ~~~~~~;~dn~~
of Energy (1931). We believe that the law of conservation of
energy is true for all physical phenomena, just as we are sure
that the law of conservation of momentum is true for all physical
phenomena. The principle of relativity occupies its place in physics
along with the principles of consetvation of energy and momen-
tum.
§ 1.9. The velocity of light in vacuo. The velocity of light in
vacuo occupies a special place in nature because in accordance
with present-day conceptions it is the greatest possible velocity
at which an interaction between objects can be transmitted. Trans-
mission of an interaction, i.e. transmission of a certain action
produced by one object onto another, is often referred to as trans-
mission of a signal; it is this term that is especially popular in
the theory of relativity. To transmit a signal means to transmit
a momentum and energy (taken to be inseparable in the theory
of relativity (see§ 5.5)) which are capable of "switching on" a
certain device, e.g. a trigger mechanism.
It does not follow from anywhere that there exists an upper
limit for the velocity at which signals can be transmitted in na-
ture. However, both theory and experiment show that all known
interactions propagate at a finite velocity; and the fastest velocity
at which a signal is transmitted is that of light in vacuo. We
shall recall that this is also the propagation velocity of electro-
magnetic waves of any frequency in vacuo. As was already men-
tioned, the classical theory assumed tacitly that a signal can pro-
pagate infinitely fast.
If one admits that there is an ultimate velocity of signal pro-
pagation in nature, its absolute value must be the same in all
inertial frames of reference. In fact, all these frames are equiva-
lent according to the principle of relativity, and it is impossible
to suggest a physical experiment to detect the difference between
them. Had the velocity of interaction transmission been different
in different inertial frames of reference, it would have been pos-
sible to distinguish one inertial frame from another. This is im-
possible, however, provided the principle of relativity is assumed
to be universal. It follows immediately from this that the velocity
of light in vacuo must be the same in all inertial frames of re-
ference.
And what if a source moves toward an observer or an observ~r
moves toward a source? Such a motion cannot change the magni·
tude of the ultimate velocity at which a signal is transmitted.
Consequently, the velocity of light in vacuo cannot depend on the
motion of either a source or an observer.
Classical Mechanics and the Principle of Relativlfy 37
Obviously, the velocity of light in vacuo has unique properties.

All velocities are relative, i.e. they change on transition from one
inertial frame to another. But the absolute value of the velocity c
remains the same. Although there is no privileged frame among
all inertial frames, there is one privileged velocity in all of them.
Both these circumstances are intrinsically associated with the fact
that electromagnetic waves can propagate in vacuo. In other
words, no material medium is needed for their propagation. Na-
turally, the assumption concerning the privileged, i.e. invariant,
velocity upsets drastically the classical arrangement, that is Eq.
(1.4) and, hence, Eqs. (1.2) and (1.3).
Nevertheless, no matter how strict or beautiful the logical
reasoning is, an experiment was and will for ever be a supreme
judge in physics. An experiment supports quite unambiguously the
iollowing two statements: (I) in a given IFR the velocity of light
in vacuo is equal at all points and in all directions; (2) in all
IFRs this velocity has the same value. Here we refer to the Mi-
chelson-Morley and Kennedy-Thorndike experiments described in
Supplement II.
CHAPTER 2
THE EINSTEIN POSTULATES.
THE INTERVAL BETWEEN EVENTS.
THE LORENTZ TRANSFORMATION
§ 2.1. Einstein's postulates. The final part of the foregoing

chapter was devoted to the explanation of two basic assumptions
of the STR which are called the Einstein postulates. On account
of their significance we shall repeat them once more here and
supplement some comments.
Postulate I. All identical physical phenomena proceed alike
in inertial frames of reference in the case of equal initial condi·
lions. In other words, there is no privileged frame among IFRs,
and the state of absolute motion is impossible to find.
Thi5 postulate extends the Galilean principle of relativity to all
phenomena of nature. It puts an end to absolute space once and
for all: since all inertial frames of reference are equivalent, they
cannot have any privileged frame among them. It was just abso-
lute space that served as such a privileged frame. The conception
of "absolute" motion in vacuo which was meant as the motion
relative to the absolute frame of reference is rejected exactly in
the same way (sec§ 1.6).
Postulate II. The velocity of light in vacuo is equal in all di-
:~Ji~;:a~~~ ~7t f:!rt~:~n~:::!saof~~~;,~~:;~ial frame of reference,
th~~=~~c~r~s oflr~h\a~~ ~a:,~t~~e~~t"~e~ec7~~h b~h~h!t~~~~c~r; ~~a~
source. This, however, follows immediately from Postulate II for-
~~~~it;ld ;~a~: ~{~ef~~ee~c:bfi~!d ~~i~~y a~~ rt~u~h~:n h~~~r~~
I a
moves nou-uniformly and/or along a curved line, an instantaneous
co-moving inerltal frame can be found. In such a frame a source
is at rest, while all other inertial frames move relative to it (and
a source moves relative to them). Since in accordance with Postu-
late II the velocity of light is the same in all frames, it does not
depend on the velocity of a source. As to the motion of the ob-
server, it is the relative velocity of a source and an observer that
is essentia I, so that the preceding reasoning disposes of the ques-
tion.
It should be clearly realized what Postulate II implies. For this
purpose let us imagine that the velocity of light is measured in the
Einstein Postulates. Lorentz Transformation 39
~~~ffr~int~;;~f~~o~i~To::~h~~ t~!is~~~~~!c~e: f~;h~o~~1n:; !~

the moment t2. Then c = (x2- x1)/(t2- t1). Now the same two
events, that is the sending and reception of a signal, are viewed
from the frame K'. The sending of a signal occurs for an observer
from the frame K' at the point xl at the moment tl and the re-
ception at the point x2 at the moment t2. In spite of the fact that
the frames K and K' move relative to each other along the com-
mon axis x, x', we have to get the ratio (xl- xl)/(f~- 11) equal
to c. From the viewpoint of the "common sense" this must not be
the case. (This becomes clear if one draws the diagram of the
experiment.) However, this is exactly what Postulate II prescribes.
We have formulated Postulate II, in fact, the way it was done
by Einstein himself in his article of 1905. However, in our ti':l~
it is advisable to formulate it otherwise, n8riiely, 1o proceeOTrom
the assumption that there is the ultimate velocity of signal trans-
mission in nature. This i::. the principal assumption. Then this
ultimate velocity is identified with the velocity of electromagnetic
waves, i.e. light, in vacuo. The last assumption is not obligatory:
basically, the STR would not have lost its meaning if the ultimate
velocity had turned out to be different. However, the STR makes
use of just this assumption. If one assumes that the velocity uf
light in vacuo is the ultimate velocity at which an interaction can
be transmitted, it follows directly that it must have the same
magnitude in aJIIFRs (see§ 1.9).
Having formulated the first principles of the theory of relativity,
that is two Einstein's postulates, one can formulate the general
objective of the special theory of relativity. Its basis is the prin-
ciple of relativity, i.e. the equivalence of all inertial frames of re-
ference with respect to all physical phenomena. The theory of re-
lativity has to give such a description of physical phenomena
which will be the same in all inertial frames of reference. Thus,
if we have some equations at our disposal describing one or
another group of phenomena, these equations must appear alike in
all inertial frames of reference, each frame using its own vari-
ables. Recall that equations of mechanics and electrodynamics
intrinsically contain coordinates of an event and a moment of its
occurrence. These coordinates and the moment of time registered
for an event are transformed on transition from one inertial frame
of reference to another. The Galilean transformation changes the
appearance of Maxwell's equations, but since we want to preserve
them as equations of an electromagnetic field correct in all iner-
tial frames, we ought to find such a transformation of coordinates
and time that will keep the appearance of Maxwell's equations
invariable. Such a transformation will turn out to be the Lorentz
transformation.
The Lorentz transformation, however, follows directly from the

Einstein postulates. The point is that the Maxwell theory wa,;
developed from the very beginning as the relativistic one. The in-
herent cause for this consists in the fact that it described cor-
rectly the properties of light, the most relativistic object of all.
Thus, having found the transformation of coordinates and time
satisfying the Einstein postulates, we have to be sure that the
basic equations of physics are the~same in all inertial frames, i.e.
covariant relative to this transformation. The meaning of the
term "covariant" will be explained in § 4.3. Now we have to dwell
on the "basic laws" of physics.
The laws of Newton are referred to as the basic laws in me-
chanics, the Jaws of Maxwell as the basic laws in electrodynamics,
and the equations expressing the first and the second principles
as the basic laws in thermodynamics.
Relative quantities were known in classical physics, e.g. velo-
cities, coordinates, velocity directions, but the special theory of
relativity adds to them, rather unexpectedly for our intuition, rt:!-
Jative time intervals between events and relative scale lengths,
Le. distances. li_owever, this is the "pri-ce" that we must pay in
order to realize the principle of relativity with respect to all
physical phenomena.
And still the predominant feature of the theory of relativity,
in spite of its title, is not at all the relativity of various quantities,
i.e. their dependence on the choice of an inertial frame of re-
ference. The essence of the theory of relativity consists in just the
opposite. The theory of relativity shows that the laws of natur\!
in inertial frames of reference do not depend on the choice of a
reference frame and on a position and motion of an observer, but
measurement results in different reference frames can be cor-
related. Speaking in terms of philosophy, the theory of relativity
underscores the objective character of the Jaws of nature and not
the relativity of knowledge.
Of course, trying to alter an historically established name,
79~h i~y-ath:Q;e~~si'~sil~~RP~~ri~~n~O~e!~;~:~~~eb~-~~t~ao~~ {~
pay attention to. The controversy over the correct name for tile
theory, "special" or "partial", is not essential. In a sense, the
problem is how to restrict the theory to inertial frames of rei-
erence. Essentially this restriction results in the theory \Vhich 1_,
correct in the absence of gravitational fields or, practically, in
weak gravitational fields. That is why the most correct name
would be the "restricted" theory of relativity, which is adopted
Jn the French literature.
AlthOUgh the Einstein postulates are the first principles of the
theory of relativity, !.hey___are __nqi SJJfficient _f~ i~ 9~~~~~~n_!.:._
Einstein Postulafes. Lorenfz Transformation 41
The construction of a relativistic frame of reference is funda-

~~etl!aTIYiinportant for t!_lc t,l!_eory, so we tu~l! Ql.l_r atten_tiOrl_lo
this aspect now. - -
·f 2:2. The relativistic frame of reference. In the construction of
a relativistic frame of reference, just as in the construction of the
whole theory, the validity of both Einstein's postulates is assum- \
1
7sd.th~lft~ev:~~ i\/{f~:hic~~s~g~sv~~~c~7~-~:~f*~a:i4h~ara~~~ )
assump tOn IS not present in the Einstein postulates. However, as f
we shall see later on, it has to be inevitably incorporated in the \
theory if we want the principle QL~us.ality lQ. ~ome._inj_Q__ effe~t '
(see§ 3.4). In§ 1.1 we spOke fii detail on how a reference fraffie is
constructed in classical mechanics. There we indicated that it was
sufficient for each reference frame to have one clock, since it was
assumed that infinitely fast signals could be used. But in the
STR the existence of a finite velocity of a signal is explicitly
allowed for, .so that w~eo !~e__y_elocities.in question.geLclo_s~r io
that ultimate one, it becomes jpcoquenjent and eyen impassible
to use o~Jy one clock. But it is just these velocities that are of
interesnor £pTffie"§FY_9_f_relativity.
Therefore.~t_j)Lclock""ills to. .be. add~Q_ to a coordinate system
constructed exactly in the way described in § 1.1. Basically, the
STR implies that clocks are located at every point of space. This
is not needed in practice, but as a matter of principle a clock must
be at any point where the moment of an event is registered. All
the cleelts--of a--giveR referen~e_ frame are motionle~s_.relaU.~J9Jf:
It is assumed in the STR that-1t is pos·sible lo have at one's
disposal as many ideal identical clocks as one needs. This as-
sumption is easily realized in our time. According to quantum
mechanics all microparticles of the same kind are identical. In
particular, characteristic oscillation frequcnci.es of atoms of t~1e
same kind coincide precisely. Taking the atoms themselves for
the clocks and the periods of atomic characteristic oscillations
for the time standards, we obtain a sufficient number of required
clocks.
Length standards can be dealt with in just the same manner.
The wavelength of a chara!::_t_e_ri~~L~_r_a~li.Pn of a given atom can
be chosen quite aAeqqaj_e)y_~s a leng!h_ __u_Q_it. Even prior to the
~f~t'~~~~~a{;~~n~~ma ~~~~~~~n~ct~~ -~If ~!li~ri~~z!~a~sth;nwi~:~~i~t\~
length standard: to wit, that is how the length of the metre was
immortalized by Michelson at the beginning of the 20th century.
When we consider two IFRs moving relative to each other, the
length scales and the clocks of each frame are at rest only with
respect to "their own" reference frame. Is it possible to belieYe
that we have identical length scales and clocks in different IFRs,
., Special Theory of Relativity
if, say, there are such scales and clocks only in one frame? In
some literature one can come across a discourse to the point that
length scales and clocks can be transferred from one IFR into
another. There is no doubt that one should not do this. Transfer-
ring clocks and length scales from one IFR into another, we im-
part acceleration to them. Theoretically, acceleration varies the
length of scales and the clock rate. H~re are some straightforward
examples: drop a clock or a ruler on a stone ftoor. A clock may
just stop, and a ruler break. Even an atomic clock breal{s down
when atoms get destroyed. All that is the effect of acceleration.
But in order to obtain identical length and time standards in
different lFRs, one does not need to transfer anything from one
frame into another. lt is sufficient to take a pure substance in any
reference frame and its radiation will provide us with requir~d
standards. It should be emphasized how important it is to have
length and lime standards in each IFR which are truly identic..tl
with those in all other frames. Indeed the pri_n~iQ!e.....uLLel.aii\tity
iHl_d. t~J~!l!livalenc~_QLi!.IU.E.B.Sfn co_':~.Qi_natJon....w.ilhJhe _identity of
lengt_h and tune Standards maKe !fpossible to attain lli.e ·complete
identity of these reference frames.
So, every IFR has as many adequate clocks as needed. The
time of an event at a given point is the reading of the clock
located at the point where the event occurred at the moment of
the occurrence of the event. If two events occurred at different
points in space and the clocks at these points registered the same
time for the occurrence of these events, we have to regard these
events as simultaneous. But obviously the synchronism of events
occurring at different points in space depends on how the initial
time readings of these clocks were adjusted, the clock rates being
assumed absolutely identical. Thus the determination of the syn·
chronism of events and the adjustment of the initial lime readings
of all the clocks belonging to d given IFR, i.e. the clock syn-
chronization, are the same thing. It should be pointed out that the
clock synchronization, that is the determination of the synchronism
of events, can be accomplished in different ways. The advantages
of the synchronization suggested by Einstein will be explained
later on. All the same, it should be emphasized that the synchro-
nism of events is determined, and this determination can be ac·
complished not in a single way.
Here is the example showing how important it is to know how
to determine the synchronism of events. How is a velocity of a
particle found? Let a particle move along the x axis. To obtain
its velocity, one must know the position x 1 of the particle at the
moment t 1 and also its position x 2 at the moment 12• Provided
the motion is uniform the velocity is equal to (x2 -xl)/(t2 -tt).
But the arrival of the particle at the point x 1 is registered by a
Einstein Postulates. Lorentz Transformation
clod. located at that point and the arrival at the point x2 by a

clock located at the point x2. To determine the velocity, one has
to be sure that the clock located at the point x 2 was showing at
the moment t 1 the same time as the clock located at the point x1•
Only in that case the determination of the velocity would have
any sense. But this just implies that the clocks must be synchron-
ized.
Having explained that the determination of synchronism and
clock synchronization are the same thing, we pass over to th2
procedure of clock synchronization within one IFR. The first
~:ir~fy \0h;~maen~0 th~~·~erui;nd t~e~ t~01l~~~r t~=s~~~~~e a~o~~~- sp~f:
lowing Einstein we shall reject this procedure, because every clock
transfer is associated with an acceleration that the clocks gain.
Theoretically, every acceleration affects a clock rate. Consequently,
it is better first to set the clocks at their respective points and only
then to \"erify them *.
How can one verify, i.e. synchronize, the clocks located at va-
rious points in space? Let a clock which we shall call a reference
one be located at the origin of a given IFR. Of course, this par-
ticular clock does not differ in any detail from all others. One can
send a signal from the reference clock to any clock of a given
IFR. It is assumed in this case that the distances from each of
the doCks to the reference one are know-n~ w·fflj-- r:-_o Clocks_ b_eing
nece~s_ary to.. 00ter.mine_1!Jese d_istances. KnoWing the velocity at
which a signal is transmiTte<I, ""Oileean find the time it takes a
signal to travel from the reference clock to any clock of the frame.
If the signal from the reference clock is sent at the moment t=O,
a synchronized clock should display just exactly this time at the
moment when the signal reaches this clock. Although generally
speaking one can use any signal, it is most convenient to choo,;e
a light signal in vacuo for the purpose of clock synchronization
in all IFRs, since it propagates at the same velocity in all IFRs.
The utilization of a light signal in vacuo for clock synchroniza-
tion is one more factor ensuring the complete equivalence of all
IFRs.
Thus, a synchronization "agent" is a light signal. Let us de-
scribe now a synchronization procedure for a given lFR according
to Einstein.
I. Clocks are set at their respective points and actuated. Coor-
dinates of points at which the clocks are located are known, and
• There exists the voluminous literature uillustrating"' that the infinitely slow
tr~nsportalion of docks does not affect their rates. No doubt, it is plausible in
terms of physics. But inasmuch as the relativistic theory deals with relativistic
velocities and distances, a procedure of this kind is hardly of any interest to us.
Special Theory of Relativity
so the distances from each of the clocks to the reference one are
known.
2. At an arbitrarily chosen moment ! 1 a light signal is sent
from the reference clock to the clock to be synchronized. The liglit
signal travels in vacuo along the known path and its arrival is
registered by an observer or by means of a device.
3a. The reading of the clock at the inoment of the signal arrival
+
is to be set tot= t 1 r/c, where r is the distance to the reference
clock. The "initial" reading of the clock is thereby chosen, the
clock is "verified" against the reference one.
One may also use another equivalent method.
3b. Mirrors are set at all points where clocks are located in
crder to reflect light back to its source. If the reference clock
registers the return of the signal at the moment ! 2, the moment
registered by the clock at the mirror is to be set equal to t =
~t,+(t,+t,)/2.
The last procedure of clock synchronization has one delicate
point. Using procedure 3a we presume the velocity of light c
known. But we have already seen that two synchronized clocks
are necessary to determine the velocity of motion in one direc-
tion. On the other hand, the velocity of light is usually determined
from the motion of a beam along a closed path. In particular, the
velocity of light could be found by means of re~eclion from a
mirror, when only one clock is available. In this case one has to
make use of procedure 3b and to know the distance from the rei-
erence clock to the mirror. If this distance is equal to r, then
c = 2r/(t2 - tt). However, if the velocity of light propagating
"there" is not equal to that of light propagating "back", we are
not able to establish this fact. It is impossible to ascertain this
fact experimentally just because our clocks are synchronized to
give the value c for the velocity of light. The theory of relativity,
however, proceeds from the assumption that the velocity of light
in vacuo is the same in all directions. Besides, the totality of ex-
perimental data does not contradict either this statement or the
consequences of the theory of relativity.
Thus, we have come to a relativistic frame of reference compris·
ing a coordinate system of rigid axes and synchronized clocks
fixed rigidly to this system. Such clocks in a given IFR will be
referred to as a "set" of clocks. The synchronization procedure
according to Einstein is such that it can be performed in the same
manner in any IFR.
In accordance with the adopted rule for clock synchronization,
synchronism of events can also be determined as follows. Let two
events occur at the points of space equally removed from the
third point. If at the moment of the occurrence of the two events
the light signals are sent from the points of events to that third
Einstein Postu.lales. Lorentz Transformation
"
point, the events are assumed simultaneous if both signals reach
the third point at the same moment of time.
Of course, accelerated motion of bodies can be treated in the
framework of the STR, whereas accelerated motion of reference
frames, relative to inertial ones, cannot be considered. Since
length standards and clocks are rigidly fixed to their IFR, it is
clear that these standards and clocks should not be accelerated.
Otherwise, the study of an influence of acceleration on length
standards and clocks would make us examine their specific struc-
ture and de~rive the theory of its universal character.
§ 2.3. The direct consequences of Einstein's postulates (a few
imaginary experiments). The two dtrect consequences of Ein-
stein's postulates, "the relativity of .l~!!~lh_ scales:: and ".there~a
tivity of time intervals bet~veen _e-Ve®', can be obtained directly
from the postulates themselves. M.ost often they are obtained
from a transformation of coordinates and time of an event. This
transformation is compatible with the Einstein postulates and is
called the Lorentz transformation. However, this convenient me-
thod to be discussed in § 3.2 is not at all obligatory. Now we
shall describe a few "imaginary experiments" by means of whici-J
\\e shall draw necessary conclusion~aginary experiments play
a conspicuous role in conclusions of the STR".--::llley.::r"eJireSenrso·nfe-
hypothetical experiments nOt TO be nece~aTily con~uctea in p--rac:
lice. In fa_~t these are_on1y generalizat1ons permi~!i"i:tg 4efinile coil-
sequences ·to l:ieObfained ffom the given· premiSes.* Now We pass
over to a description of several imaginary exper1inents whose re-
sults we shall obtain once more when consequences of the LorentL
transformation are discussed.
We shall begin with a very simple imaginary experiment which
illustrates____ib_e _relativiry of synchronism, provided the second
Einstein postulate rs·satisfied. Later·on We shall obtain the same
result by different methods. The experiment is performed in the
Einstejn t~a_i~- _This term i"S1.-ppl1ed to a·ny train moViri"g uhi-
formly and rectilinearly at, preferably, a relativistic velocity. In
an imaginary experiment one can assume even such a thing. Tlte
middle of the train is easy to find precisely. _This is done in the
traj~); reference frame and does not present anya-tfl!£y_TlLes~UO
server r rs "lb~ated in th-e l'nT<fdTeClf flie·-rriin ana -observer 2 at
the station. Light signals are sent to observer I from the ends
of the train whiclt are equally removed from him. The imaginary
experiment is so performed that the signals travelling from the
• One should not thmk that ~imaginary experiments'" (Gedankenexperimente)

are characlerist•c only of the theory of relativity. The series oF ~imaginary
experiments'" devoted to quantum me::haroics can be found in the diScussiOn ol
N. Bohr and A. Einstein, UFN 66, 571 (1958).
46 Special Theory of Relatiuily
ends of the train reach observer I just at the moment when he

rides up to observer 2. In the imaginary experiments one do~s
not usually take interest in how to accomplish this practically.
It is essential to us what conclusions will be drawn by the two
observers from the fact of a simultaneous arrival of the signals
at the middle of the tr11in.
Observer 1. The light signals must travel equal distances until
they reach me. Consequently, they are sent simultaneously.
Observer 2. The light signals reached me when the middle of
the train was moving past me. Consequently, they were sent
somewhat earlier. But "earlier"
y f( the head __of til"etralnwa"s
~--=ta::"_me -than the--tail.
T'Fl""erefore, the- Slgiial from thC
tail had to be sent slightly
c in advance in order to reach
);;.-----7,0i--~.z;x' me simultaneously with the
signal travelling from the head.
/ Consequently, the signal from
the tail was sent earlier than
Flg. 2.1. The "Imaginary experiment" tharrrorn llle he3d. ,
permitting one to establish that the It is evident from this plain
~~ng1~~ f~ t1;: J~~~t~On°r~fnt~: a~1~\f\~! reasoning that two simulta-
motion of coordinate systems, does not neous events in one reference
\ ary when mensured in any I FR. frame, that is the train's frame
in our example, are far from
being simultaneous in another, which in our example is the frame
fixed to th(! Earth.
All subsequent imaginary experiments will be of quantitative
nature. In all of them we shall be considering two IFRs designat-
ed by K and K' with their relative velocity directed along the
common axis x, x' (see Fig. 1.2). It is assumed that the Cartesian
axes of the two frames coincide at the initial moment t = t' = 0.
(a) A comparison of the lengths of parallel rulers oriented in
the direction perpendicular to that of the relative motion of two
JFRs. Let us take rulers of the same length in each reference
frame K and K' and place them along the corresponding axes 11
andy'. The equal rulers BC and B'C' are illustrated in Fig. 2.1. A
and A' are the middle points of the rulers in each reference
frame. Let the rulers move so that when the axes y and y' coin-
cide, the middle points A and A' also coincide. The frames K
and K' are geometrically identical at the moment t = t' = 0. The
question is: what quantity will be obtained for the ruler's length
B'C' as a result of measurements by the observer from the frame
K. and what for the ruler BC, when measured by the observer
from the frame K'? The observers have to mark the positions of
the two ends of the rulers moving past them simultaneously in

their respective reference frames. For the case considered here
synchronism is conveniently established as follows. When the
points C' and 8' find themselves on the y axis, light signals are
sent to the point A'. In the frame K' the sections A'C' and 8' A'
are equal, the velocity of light c is the same, so that both signals
will reach the point A' simultaneously. Consequently, the points
C' and 8' will cross the y axis simultaneously in the frame K'.
Exactly in the same manner the points C and 8 will cross the y'
axis simultaneously in the frame K' as seen from the frame K.
Now let us measure the length of the ruler 8'C' in terms of the
frame K and the length of the ruler BC in terms of the frame K'
at the moment t = t' = 0, when the y and y' axes coincide. In
this case all four points C, C', 8, 8' find themselves on the com-
mon y, y' axis, and the observers in the two frames can compare
their results. If it turned out that C8 > C' 8', or vicE" versa
C' 8' > C8, it would be possible to detect the difference between
the reference frames K and K' . .Ibis js __ inadmissibl~___ due_ io the
initial assum~~i_Q.r:t_ a~g_ut the equivalenc~ of ai_~ 11'ier~i_!!Lli.~.!D£S_Of
reference.-That is why -lhe OOservers- Tiom tlie ·rrimes K and K'
Can- Only- Certify that CB = C'8'.
Consequently, the lengths (and the length units) oriented in
a direction perpendicular to that of the relative motion remain
constant when measured in any IFR. But this means that t~
coordinates of points along the axes perpendicular to the motion
direction also remain invariable. Thus, exactly like in the Gali-
lean tranformation
y'=y, z'=z. (2.1)
(b) The comparison of clock rates in the frames K and K'. Ob·
serving clock rates in the two frames K and K' moving relati\e
to each other, one can only compare readings of one clock from
one frame with readings of several clocks from another frame,
because two clocks from different reference frames get together
at the same point in space only once. In one of the frames there
must be at least two clocks which are supposed to be synchronized
in the way described in § 2.2 of this chapter. For the sake of d~
finiteness we shall be comparing one clock from the frame K'
with two clocks from the frame K.
Let a clock and a light source be located at the origin 0' of the
frame K' (Fig. 2.2a). A mirror is set on the z' axis at the distance
z~ from the light source (and the clock) in the direction perpen-
dicular to that of the relative motion. A light signal is transmitted
from the source to the mirror from which it is reflected back and
returns to the point 0' in the time interval !J.t' = 2z0/c. Both the
"light source and the mirror are at rest in the frame K' and the
signal travels "there" and "back" along the same straight line,
i.e. the z' axis.
Now let us consider the propagation of the same signal in the
frame K relative to which the source and the mirror move to the
right together with the frame K' at the velocity V. Although the
signal was sent fro_m the two coinci.dent origins 0 and 0', the
reflection from the mirror will occur at some other point x 1 of
the frame K and the reception of the retlected signal at the point
x2 of the x <Jxis. In this way the path of the signal in the frame K
!(' /( J('
I v
r+
Mtrror
(a) (b)
x,z'
Fig. 2.2. The "imaginary experiment" showing that the interval between two
ennis measured in terms o[ proper lime IS always less thdn the time interval
bCh\·een the same events registered by means of two clocks o[ any other refer-
ence frame. (The ~ltght clock" expenment.) (a) Tile calcu[ahon of the proper-
lime interval between the sendmg and reception of a ltght signal at the
oligin 0' o[ the coordinate system (b) The calculation of the l1me interval
bet\\'een the same events m the reference frame K relative to which a light
source and a mirror move.
traces out the two sides of an equilateral triangle. As the path

tra\'elled by light in the frame K is greater than that in the frame
K', one can expect that the time interval 1t between the sending
and reception of the signal. when measured in the frame K. will
be greater than !1i'. Indeed, the observer from the frame K will
certify that the two events, i.e. the emitting of light from the
point 0' and its return to the point 0', occur at the two different
points of space 0 and B (Fig. 2.2b). The time interval M betweeil
these two e\'ents in the frame K will be measured in this case
by the two clocl\s removed from each other by the distance V1t
along the motion direction. The velocity of light is equal to c in
all reference frames. Therefore, having divided the length of the
lateral sides of the triangle OAB by the velocity of light c, we ob-
tain the time interval !it expressed implicitly:
11t=2~z~+ ( v:t Y/c.

Finding !1t from the last equation, we get

!11=~ 1 =r~
c . I V2 c •
'VI-er
where r=(t--¥:-)-''•.
Considering that z0 =z~. it follows that
ill ~ AI' ~
. I I - y:_
_._t'-
~I' ill'
.Yr=Bl •
(2.2)
'V ,,
where the designation 8 = V/c is adopted. Since both eventc;
occurred at the same point in the frame K', they were registered
bv means of the same clock. ,A._time interval between eve~ts re-
gistered __!>..¥ means of the sam: -CloCk- JW1iiC~-im[fieSlhaTlllc
-evmts_~9CCL!rred at the same pomt of spac&__ts re erred f() as_ a
P!_<?E_fr:-f[f!i§_ inter_£Cifl?~een~e$e_- eVen(S. Of COUrSe, a- time· in-
-lerval the initial anO the final moments of which are registered
at different points of the reference frame and, consequently, by
means of different clocks will not be a proper-time interval be-
tween events. In the example just examined the proper-time in-
terval is equal to !1t'. It is seen from Eq. (2.2)_ that a time in-
terval between ev_wiils.._lM.J~sLwhen it is deferrriTiied ·rn--Sikh
~~ wher~__!_h~~~- _ev~ntSli'aj)p~if1l!1h~~e-pOTIIt
Trf space. As we-Shall see in § "3.4\I IS- possible to inaicale tlte
·conditions providing the existence of a reference frame in which
two given events occur at one point.
int~~~=l \b~t~~~~ i~':ev~smi~st~ i~~~:~~~t qc~~~~~~;0ni~~~
~di----oo---the-
cti6iCe-aLa··reference frame. Nothing oftnC-SOiT'
was ever l<nown in classicalpliysiCs;' \Vliere· tlnie-r~TV-ats par
~e_ssed_~-~tie..s_._. -
This example effectively illustrates that time readings them-
selves must be different in different frames. When the origins 0
and 0' coincided, the clocks from the frames K and K' located
at this point registered, according to our condition, the moments
i 1 = 0 and !'1 = 0. When the light signal returned to 0', the
clock from the frame K' registered the time 12=1] +M'. But at
the same moment and at the same point there is the clock from
the frame K. This clock is not the one located at 0, but another
+
one synchronized with it. Its reading will be i 2 = i 1 !1i. As we
hHe already established !1t 4= 1t', i.e. the clock readings are
different. This just means that times of events are registered dif-
ferently in different reference frames. Note that this calculation
of clock readings in the frame K corresponds completely to the
50 Special Theory of Relaliulfy
rule of clock synchronization proposed by Einstein and described

in§ 2.2.
(c) A comparison of the lengths of rulers arranged parallel to
the relatiue uelocity direction. A proper frame of reference with
respect to a given object is such a frame in which this object is
at rE'st. It is customary to designate such a frame by J<.O. Let us
suppose now that a ruler positioned alqng the x 0 axis is at rest
in th1s frame. Let us designate the length of the ruler in this
frame, i.e. the proper length of the ruler, by 10. To find the length
of the ruler in any reference frame, one has to determine the coor-
dinates of the ends of the ruler simultaneously in this frame. Qne-
,, K'
(a) (b)
fig. 2.3. The "nnagmary experiment" which perm1ts dctectmg the "contraction"
of the ruler's length when measured in d reference frame in which the ruler
nlO\·es uniformly and rectilinearly The ruler is ori!nted in parallel w1th its mo-
}:~~evek~ci:lhe (~~o:e~e le~~~~u~rm:n~u~r)~h{b:e~~~h n~!a:ur~~:~t a~f r~h~ ~~~;~:
length in the reference frame relative to which the ruler moves at the velocity V.
does not _c~~~ultanci~...cl~m~nJs.on!J i~ the

\1/b-Jc..h._ lb.e.J.!!!er rests. Since in everyday life we illea-
1i'"ii'i'leJ{D in
su:FeThe ·proper length or o15jeCf5, the procedure of length mea-
syr~me!lt is si~..illl..~..£!19r!lled by means of a dire.rt.scale.
transi)OSIIIOrl.__ ·--- - -·
A ruler is at rest only in one unique reference frame. In all
other inertial frames of reference moving relative to one another
the ruler moves, and the direct tran~pos_ition of a unit scale be-
comes impossible. Let us resort to the methol:l·ot-length measure-
men~atso suitable for measuring the length of a ruler
moving relative to a reference frame.
Let us place the ruler's left end at the origin 0 0 where the'
light source I is also located. At the ruler's right end the mirror S
is fixed perpendicular to the x 0 and x axes (Fig. 2.3a). Now let
us consider the following two events. The first event: a light
signal is sent from the source I toward the mirror along the X:>
axis at the moment t = t0 = 0. The second event: having reflected
from the mirror S, the light signal gets back to the ruler's leit
Einstein Postulates Lorentz Transformation 51
end at the point 0 0• Both events are registered at the point o~

bv means of one clock. So the time interval between the events
iS the proper-time interval !11 0 which obviously be written
down as
(2.3)
The same two events look somewhat different to the observer

[rom the frame K (Fig. 2.3b). At the moment when the signal is
emitted the source I in the frame K is positioned at the point 0
and the mirror S at the point St. By the moment of reflection the
mirror will be shifted to the point S 2 and the source I to tht~
point I 1• By the moment of the arrival of the signal reflected from
the mirror at the ruler's left end the source will already be shHted
to the point 12. The moments of time corresponding to the first
and second events are registered in the frame K at different
points and, consequently, by means of different clocks. This means
that the time interval Ill between these events can be expressed
in terms of !lt 0 according to Eq. (2.2). When light propagates t;">
the right. the velocity at which it overtakes the mirror S is equal
to c- V, according to the classical velocity summation in tile
given I FR. When light propagates to the left, it moves toward the
nmror at the velocity c +
V. Designating the ruler's length, un-
known so far, by I in the frame K. we obtain the time in which
light gets from the source to the mirror, t 1 = 1/(c- V), and the
time in \Vhich light gets from the mirror to the source, t2 =
= 1/(c + V). Therefore, the time interval between the sending
and reception of the light signal in the frame K is
fll=tt+t2=cb-+c:Fv=-¥- 1 -.! 8a, B=f
Recalling that l'!lo=.Yt-B2 /'!!, and taking into account Eq.
(2.3), we obtain from the last equation
I= c (I ; D1 ) Ill = c · ~lo ,Yf="W = lo ,Yf"='B2 = ¥-. (2.4)
Eq. (2.4) gives the length of the ruler when measured in an)'
inertial frame of reference. In the frame in which the ruler is at
rest (B = 0) its length is equal to 10 . It is just the fact from
which we started our reasoning. Eq. (2.4) is asymmetric with
respect to the lengths I and !0, since it relates the proper length / 0
of the ruler in the frame K0 to the improper length I in any other
reference frame K.
Thus we have ascertained the relativity of time intervals be-
tween events and the relativity of the ruler's lengths or scal~s
directly from Einstein's postulates, the quantities, which, in clas-
sical mechanics, were equal in all inertial frames of reference.
52 ~pedal Theory of Relativity
These results are inherent in the theory of relativity and require

a comprehensive discussion. However, we shall postpone the dis·
cussion of the results obtained till §§ 3.2, 3.3, since these results,
because of their significance, will be again derived by several
methods to reveal some new circumstances essential to the inter-
pretation of Eqs. (2.2) and (2.4).
§ 2.4. The relativity of synchronizatiOn of clocks belonging to
two inertial frames of reference. The direct derivation of the Lo-
rentz transformation. Up to now we have been considering syn-
chronization of a set ot clocks belonging to a given inertial frame.
But all inertial frames are equivalent and any event can be reg-
istered by an observer located in any inertial frame. An ob-
server marks the coordinates of the event in his own coordinate
grid. The set of clocks in every inertial frame of reference registers
the time of the event by means of the clock located at the moment
of the event at the point in space where this event occurs.
Speaking figuratively, all space is tilled up with moving clocks
belonging to different reference frames and a momentary light
flash at a given point in space, illuminating the dials of all clocks
located at it, makes it possible to determine the time of the
event, i.e. the light flash, in all reference frames whose clocks
were at that point at the moment of the light flash. In order to
make out what happens here, it is sufficient to consider the two
frames K and K'.
The question is what the clocks of the two frames K and K' will
show when they find themselves at one point. Of course, if we
want to compare the readings of the clod;s from different frame~.
a certain relationship should be established between the readings
of corresponding sets of clocks. The comparison is meaningless
without such a relationship. It should be recalled here that all th~
clocks belonging to each of the frames are synchronized.
It turns out that all one should do is to synchronize only the
two clocks, one of the frame K and the other of K', which get to-
gether at a given moment. Having synchronized this pair of clocks,
we thereby reset the readings of the remaining clocks in eacr.
frame. Then it turns out that the clocks of the frames K and K'
show different time at all other points in space. This is a very
significant result: the. clock synchronized in one reference fr_ame. is
dis-synchronize!l_~terms of any other inertial frame of reference.
In other- WOTOS, if the readlrigs ol all the clocks belonging to the
frame K are simultaneously fixed in the frame K', it will be seen
that all these clocks show different time in the frame K. Now we
shall derive the requisite equations.
Usually a relationship between the sets of synchronized clock~
in the frames K and K' is established as follows. When the origins
of the frames K and K' coincide, the clocks of K and K' located
Einstein Postulates. Lorentz Transfarmalion 53
at the common origin are set to the marks t = 0 and t' = 0. As

we shall soon see, iL_@~~--!19..l.lQUo..w _f{QID lhis _tb.aLthe.-.clock..s_
_!rom the frames K and K' show th~ same tlme at all other points
jn_~~-
We shall need a formula for coordinate transformation for
points in space on transition from th~ frame K to the frame K'.
When the origins of coordinates coincide, the coordinate grid of
the frame K' is contracted 1/f times in terms of the frame K.
The proper unit scales are assumed to be the same in the frames
K and K'. Consequently, at the initial moment the coordinates x
and x' are related by the ratio (see Eq. (2.4))
x=-f (r= .Yt~u~).

By the moment t the coordinate grid of the frame K' will
shift as a whole by the distance Vt, so that we obtain x = x'/1' +
+ Vt at that moment. Therefore, if the coordinate of the point
in the frame K is equal to x at the moment t, its coordinate x' in
the frame K' will bee~--- . _ . --. \
/x'(x,l)~r(x-Vt). J (2.5)
The coordinate grid does not vary along they and z axes (§ 2.3),
and because of this
y'=y, z'=z.
Now we are interested in the reading of the clock of the frame K'
located at the point x at the moment t. Let us designate this read-
ing by t'(x, t). This quantity can be defined by many methods,
but now we shall obtain it using a clock synchronization proce-
dure.
We shall do the following: when the origins of coordinates 0
and 0' coincide and the readings of the two clocks, one from K
and the other from K', are equal to zero, a light signal is sent
along the common x, x' axis in the direction of the growing
values of x, x'. Next, we consider the moment i by the clock of the
frame K. At this moment the signal arrives at the point x2 = ct
of the frame K. The arrival of the signal at the point x 2 at the
moment t represents the event with the coordinates (x2, t) in the
frame K. In the frame K' the same event will have the coordinates
(xl, t~). with x2 = ct2 according to the second postulate. But Eq.
(2.5) fs valid for all events, so substituting x 2 and x2 into its
left-hand and right-hanc!..J;:Ldes....a.nd....c.anc~lling by c, we get
lt;~rt(I-B). \ (2.6)
This means that a cloc~fOrlii~t;-the set of clocks of the
frame K' shows at the point x2 the time /~ which does not at all
Special Tfleory o/ Relalit~ity
coincide with the time t showed at the same point by a clock of

the frame K. Thus we have different time readings; this fact has
already been discussed in § 2.3.
Now we can find the reading of still another clock of the
frame K' at the moment t. The origin 0' will get at the point
x 1 = Vt at the moment t. The rerercnce clock of the frame K' will
also shift to this pOint together wifh the origin. During its
shifting it will register the proper-time interval 1!!.1~ = t]- 0 = tl.
K'
t IHumentt I
t;~tjr t;-rtt-.B)t
W'~~-~
Fig. 2.4. The dts·synchronization of the clocks of the frame K' in terms of the
frame K When the origins 0 and 0' coincide, two docks of the frames K and
K', which happen to be at this point, are set so that their readings are t = 0
and I'= 0 AI the moment t (by the frame K dock) the readings of the frame
K' dock can be found at the points x 1 = Vt and x2 =ct.
Meanwhile the time interval between the moment when the origins
0 and 0' coincide, and the moment when the origin 0' shifts to
the point x 1, is equal to 11 = t - 0 = t, when measured by the
clock of the frame K. According to Eq. (2.2)
,; ~1/l'. (2.7)
Thus we have come to the conclusion (Fig. 2.4) that at the
moment t (by the clock of the frame K. i.e., simultaneously in the
frame K) the clock of the frame K' shows different time when
located at different points in the frame K:
at the point x 2 =ct t;=r(l-B)t,
at the point x 1 = Vt t~ = 1/f.
And all this in spite of the fact that all the clocks from the set
of the frame K' have been synchronized within their own frame.
But the calculation shows that all these clocks are dis-synchron-
ized in the frame K. We have also found that a dis-synchroniza-
tion depends on what point of the frame K is selected for clock
-comparison. Let us find the difference of readings of clocks of
the frame K' at the points x 2 and x 1:
!it' =t;-t~ =fBt(B-1).
Einstein Postulates. Lorentz Tram/ormation 55
This difference of readings accumulates at the distance 1x =

= x 2 - x 1 = ct (I - B). Having assumed that a dis-synchroni-
zation depends continuously on the distance along the x axis, one
can determine a dis-s nchroniz i er unit length:
~~-r~
6x ·
j (2.8)
It is seen from Eq. (2.8) that a dis-synchronization per unit

· ~~~~ti~e~0~~~~ 0~/f6:na~s~~:;eb~t~~~~~ i~ea c~~~e~} fhebrrta~e dJ(;
as measured in the frame K.
Now one can get for an ar- t'(r, 0}
bitrary pair of points
t•O
t;-t~=-f~(x 2 -x 1 ).
We have already mention-
ed that sets of synchronized
clocks of the frames K and x
K' are adjusted to each other logs behind
by setting clocks to the zero
reading at the point X1 = 0
at the moment when the co- fig. 2.5. The frame K' clock readings at
ordinate systems of the the moment t = 0 (by the frame K clock)
frames coincide. In othrr at the points of a coordinate x.
words. the values t = 0 and
t' = 0 are ascribed to the readings of the clocks located at th.it
point. Substituting x2 = x, we obtain from the last equation
~, (x. i;,;o)-==- r£ (2.9)
It is seen from Eq. (2.9) what the clock of the frame K' located
at the point x will show at the moment t = 0 (by the clock of
the frame K). Its readings are illustrated graphically in Fig. 2.5.
A _do~k of t~e frame K' _oute._aces a__c)ock of the __frame K. when
-i0c3ied to tile left -of the· Orlgill,- and -Jags--behin-d -il _to the- righT
onne··orlgiil~ - ---- - - - - - - - - - - · - - --- ------
Now it is not difficult to ascertain what a clock of the frame
K' will show at the moment t, when located at the point x. We
shall take advantage of the fact that the difference of the readings
of two clocks of the frame K' does not depend on the choice of
the moment t. The clock of the frame K' which was at the point
x- Vt at the moment t = 0, and according to Eq. (2.9) lagged
behind the reference clock by the period r-f(x- VI), will get at
the point x at the moment t. This clock will always lag behind
the reference one by this period. But at the moment t the re-
S6 Speclaf Theory of Relalivity
ference clock will show the time t~=l/1' (see Eq. (2.4)), while
the clock located at the point x will show the time
ft'(x, r~-(~~-vt)=:r(t -~x).)
/)=+-- (2.10)
---------- - -·- _J
Eqs. (2.5), (2.6) ~nd _(_2._10} constitut~ the Lorentz transforma-
tion. Of course, the derivation of Eq. (2.10) may seem clumsy and
eVen superfluous. Indeed, applying the· reasoning that has led us
to Eq. (2.5) to the transition from the frame K' to K, we obtain
x(x',t')~f(x'+Vt'). (2.11)
Solving Eq. (2.11) with respect to t' and substituting x' ac-
cording to Eq. (2.5), we immediately get Eq. (2.10):
1'~+-{-T- <'}~+-{+- r (x- Vt)}~r(t--¥-x).

Having shown the relativity of clock synchronization, we cleared
up the physical meaning of different readings of clocks in different
inertial frames. Besides, the understanding of "clock dis-syn-
chronization" permits many baffling questions to be avoided. In
conclusion, note that the point at which the readings of the clocks
{)f the frames K and K' coincide, moves continuously along the
positive x axis at the velocity to be found from Eq. (2.10) by
substituting t = t' in it:
(2.12)
§ 2.5. The Lorentz transformation as a consequence of Ein-

stein's postulates. Let us derive the equations defining the coor-
dinates (x', y', z', t') of an event in the frame !(' from the coor-
dinates (x, y, z, t) of the same event in the frame K. These equa-
tions performing the transformation of the coordinates of the
~nt must comply with Einstein's postulates.
e~:c~\ tp~1o~s~ct~:t"~dfl~~~nJ~~~~~~~~
a es o e even · sl__b~_JinJ:ar- In-
]lfea,Te e origin of the coordinates and the time be changed, i.e.
the transformation x=X + xo, y=jj +Yo.z=Z +
zo, t=l +
t0 be
performed. If the relationship between the coordinates of the event
is linear in the frames K and K', we obtain. for x', for example
x' =a 1x-ta2 y + a;JZ-1- a~t=
= a1X + a2fi + a3 Z + a~t + (a1xo + a2Yo + a3Zo +- a4/0),
where a~. a2, a3 and a4 are constants. From the last equation it
is seen that the origin has shifted in the frame K' as well, since
the expression in parentheses is the same for all points of the
Einstein Postulates. Lorentz Transformation 5T
frame K'. However, such a shifting is immaterial due to the uni-

formity of space and time in all IFRs.
Now let us introduce into the transformation equation at least
one second-order term:
x' = btX2 + ...
= bti 2 +
2btixo + ...
Then the second term in the third link of this equaton will depend
on X and give rise to a distort~n .Qf...liefor.maliun.. ..aLsp....a~e. This
cannot be tQlerated. Conseg_ue_Qt_l.y,_th.e._ translormaiian_to_ be_.d.e,..
rived is linear. - - -
-we -s"hall-make use of the arrangement of reference frames il-
lustrated in Fig. 1.2. The relative velocity of the frames is directed
along the common x, x' axis with the y and z axes parallel to
the y' and z' axes respectively. At the moment t = t' = 0 the
coordinate systems coincide. The velocity of the frame K' relative
to the frame K is equal to V. The x axis results from the inter-
section of the planes y = 0 and z = 0. So if the x and x' axes
coincide, then y' = 0 and z' = 0 due to the condition y = 0 and
z = 0. Thus, the transformation equations for the variables y and
z must have the following form:
y' = Az+ By, z' =Cy+ Dz.
Here A, B, C and D are constants. Since spatial rotations of co-
ordinate systems are inessential for a description of physical phe-
nomena, one can ensure, through rotation of the y' and z' axes
around the x' axis, that the plane y = 0 is transformed into the
plane y' = 0 and the plane z = 0 into the plane z' = 0. Thus,
one can put y' =By and z' = Dz. However, since the directions
y and z are equivalent, i.e. space is isotropic, and the relative
velocity of the reference frames is directed along the x, x' axi.:;,
it should be true that B =D. Hence,
_u'=Dy, z'=Dz.
It remains to determine the coefficient B. Let us consider a unit
ruler located in the frame K along the y axis with the coordi-
nates of its ends being y 1 = 0, y 2 = I. These coordinates would
be y~=O, y;=D in the frame K', and its length would be
l' = y;- y; =D. If we took a unit ruler located in the frame K'
along they' axis (y;=o, u;= 1), the coordinates of its ends in
the frame K would be y 1 = 0, y 2 = 1/D and the length would be
1=y2-YJ = 1/D.
Thus, when measuring a unit ruler of the frame K, an observer
from the frame K' will find its length equal to D, while an ob·
server from the frame K measuring a unit ruler in the frame K'
will find its length equal to 1/D. Since all inertial frames are
58 Special Theory of Refativlty
equivalent, such a result cannot be tolerated unless D = I.

Therefore,
y=y', z=z',
just as it was directly obtained from Einstein's postulates (sec
§ 2.3).
~'!ow let us derive the transformation equations for the variables
x and t. Since the transformation is linear,
x'=A 11 x+A 12t+A 10, (•)
and vice versa

x = A21x' + A22t' + A20,
where all coefficients A are constants. From the initial condi·
tions t = 0 and t' = 0 when the origins 0 and 0' coincide. Hencr,
A 10 = 0 and A2o = 0. When observing the point 0' we can say
that its coordinate x is equal to Vt at the moment t. So from Eq.
l*) we get
0=A 11 Vt+Ad;
consequently, A 1JA 11 = -V. Designating* A12 by r', we can
rewrite Eq. (*) in the form (see§ 2.4):
x' ~I'' (x- VI), (2.13)
and from the analogous reasonings
x~r(x' + V(). (2.14)
Thus, the problem has reduced to the determination of the coef·
ficients r and r'. Due..J..a__ihLYniformity of time and sp_a_~.lli.Jo
the isotrop)r of space bath the"SfCOelficients can o~!y _depend on
~th~~~s~~~~-tV:~~: ~~!f~Y~~n-deed, let the sca·l: located along
the x axis in the frame K have the proper length /0• If one of its
ends is placed at the origin of the frame K, the coordinates of its
ends will be x 1 = 0 and x 2 = l0 respectively. According to Eq.
(2.14) xl = 0, x2= 1r1r at the moment t = t' = 0. (Recall that
the two frames coincide geometrically at that moment.) Conse-
quently, the length of the scale in terms of the frame K' is equal
to l' = xJ- xl =loll''. Let us take the scale of the same length
fixed to the frame K' and also located along the x axis. Then the
coordinates of its ends will be xl = 0 and xJ = 10 • But in terms
of the frame K these coordinates will be x 1 = 0 and x2 = 1011'' at
the moment t = 0 according to Eq. (2.13). Consequently, its
• It w1ll be shown very soon that the quantities f and f' introduced here are
equal and coincide with the quantities f used in Eqs. (2.2) and (2.4),
Einstein Postulates. Lorentz Trans/ormation
••
length is equal to i = x2- x1 = lo/f', i.e. the scale is contracted
I'' times. Since the frames K and K' are equivalent and their rel-
ative velocity is the same, the contraction must be identical and,
consequently, r = r'.
Now let us define the quantity r. The difference between thi!
frames K and K' lies only in their relative motion, and r can
depend on the absolute value of lhe velocity V alone. Let us take
advantage of the post_!I_I~JLQn_lhe_ln.v.ar.ianLvelocily of lighLin
-racUOJir_!!1___~ppose, at the moment t = t' = 0, when the
origins 0 and 0' of the two frames coincide, a light signal is sent
from the common origin. Let the event consist in the arrival of
the signal at some moment (t in the frame K or t' in the frame-
K') at some point (x in the frame K or x' in the frame K')
located at the x axis. In the frame K this point has the coordinate
x = ct, whereas in the frame K' the same point has the coordi·
nate x' = ct'. These times and coordinates are interrelated IJy
Eqs. (2.13) and (2.14). so that substituting these expressions
for x and x' in Eqs. (2.13) and (2.14), we obtain
ct' = rt (c - '!), ct = rt' (c +V).
Multiplying term-by-term the left-hand and right-hand sides
of these two expressions and cancelling them by tt', we get
r-- - -1- - -
- ;--v-, - - -1-
-v'l=D' . B-~
- ' . (215)
.
~1--;:;-
We have thus seen that this quantity r coincides with th~

quantity r which first appeared in Eqs. (2.2) and (2.4).
In order to find a transformation equation for the variable t
we shall define t' from Eq. (2.14), taking into account Eq. (2.13):
t'= r~ -f= ;v- rcx;vn =r{t +V(~- l)}=

~r(t-f,-x).
Thus, we finally obtain the transformation equations in tile
following form:
x'=r(x-Vt), y'=y, z'=z, t'=r(t-JJ-x). (2.16)
Eq. (2.16) is referred to as the Lorentz transformation. It is
easy to rewrite it for an arbitrary direction of the relative ve-
locity V of the frames K and K'. Indeed, we know that coordinates
change in a motion direction and remain invariable in a direction
perpendicular to motion. Let us resolve the radius vector r of the
60 Special Theory of Relafivil11
point into components: one parallel to the motion direction ru

(rull V) and another perpendicular to it r .L (r .L .l V):
r=r.L+'u·
Then
but
Therefore
r~ ~r;+r~ ~f(r 1 -Vt)+r.._ ~r[(r 1 +r.._)-Vi]-
r .L =r - r 1 =r -V~=/V~~)J
o.nd, consequently,
r'~ r(r- Vi)+ (I'- I) lfr~~ VJ,
t'~r(t-4,}
This is the Lorentz transformation in a vector form for an
arbitrary direction of the relative velocity. The equation for r'
corresponds to the classical Eq. (1.1) and transforms into it
when r =I.
Once again we shall postpone discussing the meaning of the
Lorentz transformation 1 (Eq. (2.16)) till we derive it by still
another method. That method will lead us to a realization that
the real physical world in which all phenomena of nature occur
is a four-dimensional manifold, the so-called ypace-tim~_· The
special theory _o_r rdaliY_ity wlli ~p__pear before us as th~ thepry of
-rour--diffienslQm!.l_~ace-time, ~s_well ~-~ thellleofy poSsessing an
015yious=_ge_ome!ri~!_c::Di~(IT!iLDUe tq __ ITS-jilf)'slca:l sco-pe-~nd a
~Possibility or a lurt~er _g~~{iiT£~~~on_ ~uch an afproach proved
to be of_ _eAr~m~ i[n~6fl_anc~Uo_ Ql,l[whole-viSion o the world and
the first step towarif the creation of the theory of gravitation.
§ 2.6. The propagation of the light wave profile. An interval
~~~~~:;ri~;e~:sin Lf!r~~ ~fnt~t I~R0;,h~ ~~~g~~r;h:xrr:r~~e%;
moving along the common x, x' axis at the velocity V in vacuo.
At the initial moment t = t' = 0, when the origins 0 and 0'
coincide, a light flash is triggered. According to the second Ein-
stein postulate light propagates in all directions in the frames K
and K' at the same velocity c. Consequently, the wave profile, i.e.
Einstein Postulates Lorentz Transformation 61
the surface of equal phases, will look like a sphere in each of the
frames K and K'. The equation of this sphere can be easily written
down:
In the frame K
x2 + y2 + z2= c2t2.
I
In the frame K'
x'2 + yf2+ z'2= c2t'2,
Even if we forget everything that was spoken about different time
readings t and t' in the frames K and K', we still can explain
now why we wrote t' instead of t for the frame K'. Let us sup·
pose the time in the frames be equal, i.e. t = t'. Then the radii
of the spheres turn out to be equal at a given moment t. Thus,
the same physical object, the wave profile, is equally described by
the two spheres of equal radii with their centres located at th~
two points 0 and 0'. This is an absurdity. Hence, one cannot
assume t = t'. Let us put down the equations in the form
c'l'- (x'+ y' +z'l -o,
c2(2- (x'2 + y'2 + z'~) = 0.
In this imaginary experiment we deal, in fact, with two events.
The first one consists in sending a signal from the origin x 0 = 0,
y 0 = 0, z0 = 0 at the moment t0 = 0, and the second in the arrival
of the signal at an arbitrary point of the sphere having the coor-
dinates x, y, z at the moment t. If one makes up the expression
~~u ~ ~ ~ ~ ~ ~ ~-
- -.,l"c'"'P~'x'~'y',-'z'',
that is referred to as the interval between these two events and
designated by s. the result obtained can be formulated as follow<;·
the square of the interval between the two events, consisting in
sending a signal from one point and its arrival at another, mu<;t
be equal to zero in any reference frame:
s'-0, s~-o. (2.17)
Of course, the interval between the events can be found not
only for the sending and arrival of a light beam. If the coordi·
nates of Event I are defined by the numbers X1, y1. Z1, 11 and the
coordinates of Event 2 by the numbers x 2, y 2, z2, t2, the interval
s 12 between these events is equal to
s,- -./ c' (1,- 1,)' '<c;-x,--:-;x,)',---,(y:-,--:y"',)'-c=(z,-:-::-::-z,")'.
The interval s 12 for arbitrary events, however, is not equal to zero.
Frequently it is convenient to consider events occurring at
infinitely near points and at infinitely near moments. Assuming
., Special Theory of Relativity
in this case 12 - t 1=dl, x 2 - x•=dx, Y2- Y•=dy, Z2- z1=dz,

we obtain the interval squared in the form
ds 2 = c2dt 2 - dx2- dy2 - dz2.
As we shall illustrate now, the basic property of the interval
between events is its invariance on transition from one inertial
frame to another.
According to Eq. (2.17) it follows fr6m this imaginary experi-
ment, that is the sending and reception of a light signal, that if
ds 2 = 0 in one IFR, ds' 2 = 0 in any other. Both ds and ds' are
infinitesimal quantities of the same order and consequently must
be proportional to each other. Therefore, one may put down
ds 2 =ads'2,
where a is a proportionality factor. This relationship must be
\'alid for the interval between any pairs of events. Indeed, there
are no conditions imposed on the relationship between the in-
tervals ds and ds' for a pair of arbitrary events. As to the speci<~l
events, that is the sending and reception of light signals, the re-
lationship has to be just like it is shown above.
The coefficient a cannot depend on the coordinates x, y, z and
the time t. Otherwise it would mean that different points in space
and different moments of time are not equivalent. Since we regard
space and time as uniform, a must be a constant depending only
on the absolute value of the relative velocity of the two IFRs
under consideration. Indeed, the coefficient a cannot depend on the
direction of the relative \'elocity of the two IFRs, because other-
wise it would imply inequivalence of different directions in space.
Due to the isotropy of space we havt> to presume that a can only
depend on the absolute value of the relative velocity of the iner-
tial frames of reference in question.
Let us consider three IFRs, having designated them K. K' and
K'' respectively, with V 1 being the velocity of the frame K' re-
lative to K and V2 the velocity of K" relative to K. We can write
that
ds 2 =a(Vd dsi. (•)
ds 2=a (V2) ds~. (oo)
Considering directly the frames K' <md K", one can write that
ds~=a(V 1 2)ds~·
where V12 is the absolute value of the velocity of the frame K'
relative to the frame K". Substituting the last expression in Eq.
(*) and comparing it with Eq. ( .. ),we find that
::~:~=a (Vd. (•u}

Einstein Posfulales Lorentz Transformation 63
Since V12 depends not only on the absolute values of the vectors
V1 and V2 but also on the angle between them (which does not
enter explicitly in the last equation), this relationship can !Je
evidently satisfied only when the coefficient a is reduced to a con·
stant value. From the last equation it is clear that the constant a
can be equal only to unity. Hence,
ds 2 =ds'2;
from the equality of infinitesimal intervals it follows that
s=s',
i.e. the interval a is invariant with respect to the transformation
of coordinates and time, complying with the Einstein postulates.
(Note that the intervals s and s' cannot differ by an arbitrary
constant, since from s = 0 it follows that s' = 0.) We have al·
ready seen and shall make sure again that such a transformation
is the Lorentz transformation.
Thus, the expression c2t 2 - x 2 - y 2 - z2 must remain invariable
on transition from the frame K to K'. When the frames K and K'
are arranged the way it is shown in Fig. 1.2, then y = y', z = z'
and the sum y 2 + z2 becomes an invariant. In this case the ex·
pression
(2.18)
will be, in fact, the transformation invariant.
§ 2.7. The Lorentz transformation as a consequence of the in-
variance of the Interval between events. In the previous section
it was shown that the coordinates of two events must satisfy the
equation
c2t'2- x'2 = c2P- x2, or x'2- c2t'2 = x2- c2t2 (2.19)
on transition from one inertial frame of reference to another. Here

we suppose for the sake of simplicity that one of the events has
the coordinates (0, 0, 0, 0), and due to our agreement al>out n~
ference frames it means that this event also has the coordinates
(0, 0, 0, 0) in another frame.
Let us take one more step to simplify the notation. The reader
has evidently already noticed how common is the product of the
velocity of light by the time ct. Let us introduce a new time unit,
a light metre, which is the time interval that light requires to prop-
agate over the distance I m. Obviously, I m of time= I m/s, i.e.
I m is covered in 1/c seconds.
Light travels T metres in the time •·1/c = t s. Hence, it is
clear that
'(light· m) = ct (s). (2.20)
This unit will not appear exceptional if one recalls that in as-
tronomy distances are measured in terms of time (and the velocity
of light), that is in light years.
So, if time is measured in light metres, the expression for the
invariant interval between events becomes quite simple:
(2.2!)
The easiest way to find a transformation satisfying Eq. (2.21)
is as follows. We know from § 2.5 that a transformation of coor-
dinates and time must be linear. Let us write down such a trans-
formation using indefinite constant coerficients in the form
X1 = OtX + b(r:, (2.22)
't'=a2 x+b2't.
Substituting Eq. (2.22) into the lert-hand side of Eq. (2.21)
and grouping the coefficients at x2, -r 2 and 2xr:, we obtain
x' 2 - -r 2 =x2(a~ -aD- 't' 2 (b~- bD + 2x-r(a 1b1 - a2b2 ) ===x2 --r2•
(2.23)
The last link of this equation is written down according to Eq.
(2.21); it must be identically satisfied for any x and -r. This re·
quires, however, the following equations to be complied with:
a~-a~=!, b~-b~=l, a1b1 -a2b2 =0. (2.24)
These equations are very easy to satisfy, having assumed the
coefficients a and b equal to hyperbolic functions (defined and
described in Appendix I, § 9):
a 1=cosh a.. a2 =sinh a~o b2 =cosh a2, b1=sinh a2.
In this case the first two equations of (2.24) are satisfied auto·
matica\ly. It follows from the third equation, rewritten as ada 2 =
= b2lb •. that tanh a. = tanh a2. and in order to satisfy this equa-
tion it is sufficient to assume a. = a2= a. Thus, the transforma-
tion (2.22) takes the form
X 1 =X COSh a+ 't sinh a,
(2.25)
't 1 =x sinh a+ 't cosh a.
The parameter a can depend only on the relative velocity V. It is
referred to as a velocity parameter and plays an important role
in the STR (see § 3.5). To define it, one should use the first
equation (2.25). Assuming x' = 0 for the origin 0', we obtain
~~-tanh a. (2.26)
However, the left-hand side of the equation, when written i:J.

conventional time units, appears as x/ct = B. since x/t at the
origin 0' is just equal to the velocity of the frame K' relative to l(.
So, we have found the relationship between the parameter a and
th<' velocity V:
tanh a=- B. (2.27)
One easily finds from this (see Appendix I, § 9)
cosh9= ,Y1-:anh26 = h =f,

(2.28)
sinh a= tanh a cosh a=- fB.
Finally, we obtain the sought for transformations

x' ~xr+ T ( - rB)~ r(x -B<),
(2.29)
<' ~x(- rB) + <f ~r(< -Bx).
For the reverse transformations we obtain

x~ r (x' + B<'), (2.30)
'l'=I'('l''+Bx').
(The easiest way is to replace the primed quantities by the un·
primed ones and B with -B.)
These perfectly symmetric equations are easily identified with
the same Lorentz transformation of Eq. (2.16); it is enough to
make the substitution of Eq. (2.20).
Thus, we have again obtained the Lorentz transformation, pro-
ceeding from the invariance of the interval and the uniformity of
space and time. There is no \\'Onder in this, because the interval
invariance is a direct consequence of the Einstein postulates.
§ 2.8. Complex values in the STR. Symmetric designations.
Sometimes for the sake of the formal convenience an imaginary
time coordinat~ i= ict = i'l' is introduc~d. This practice is efficient
when used in the frame\vork of the special theory of relativity,
because it frees us of necessity to introduce and distinguish co-
and contra\·ariant coordinates (see Appendix I, § 8). An intro-
duction of such coordinates is inevitable in relativistic electro-
dynamics, unless the imaginary lime is used. It should be pointed
out that an introduction of an imaginary time is only a matter
of convenience and one can do without it. Therefore, there ts
nothing mysterious about the appearance of the number i. In the
final form all formulae for coordinates and time do not contain
the number i. This confirms once more that it plays only an aul\-
iliary role.
,_,
66 Special Theory of R.elatlvlty
So, for the sake of the formal convenience we introduce an

imaginary coordinate i = ict. Then
s2=c2t2-x2=- (.il+i2) (J2=-I).
Here is the derivation of the Lorentz transformation using the

imaginary variable i. Consider the plane of variables (x, "t"). In
this plane the expression r + 't 2 represents a distance from the
origin of coordinates to the point (x, •). This distance does not
vary on rotation of the coordinate system through the angle cp in
the plane (x, 't).
Rotation in a conventional (Euclidean) plane through the
angle cp is described by the following equations (see Appendix I,
§ 2)'
x' = xcos<p+ y sin tp, y' = -x sin cp+ ycos<p, (2.31)
where all the quantities are real.

Let us consider rotation in the plane (x, 't) for the case when
one of the coordinates is purely imaginary. We shall assume that
Eq. (2.31) retains its appearance in this case as well. As we shall
see later, the geometric meaning of the equations with an imagi-
nary variable differs essentially from that of Eqs. (2.31). So let us
put down the sought for transformation in the form
X1 -= XCOS!p +
i Sin cp, (2.32a)
i' =icoscp- x sin 'P· (2.32b)
Let us clear up the meaning of the parameter IP· It can be as-

sociated only with the velocity V of a relative motion of the frames
K and K', because this is the only parameter by which they differ.
Take any point in the frame K' (x' = const). It moves relative
to K just as the whole frame does, i.e. at the velocity V. For any
point rigidly fixed to the frame K' one can write V = dx/dt. As-
suming x to be a function of i, differentiate Eq. (2.32a) with re-
spect to i. We shall obtain dx cos cp +
di sin IP = 0, whence it fol·
lows that
*=:;~%=-tan !Jl,
and, consequently,
tanip=iB. {2.33)
The tangent proved to be an imaginary quantity. This reminds

us once again th:lt there is an imaginary quantity among the
variables.
Using conventional formulas of trigonometry one can find from

Eq. (2.33)
sin<p= 1/i~nt:n~qJ = ,Yi,~R~ =iBf,
(2.34)
COS<p= ,Yt+ltan~qJ =,Yt~Bl =f,
where the designation used already in Eq. (2.15), 1'=(1- B2)-'''•
is introduced. Substituting the values of cos <p and sin <p in Eq.
(2.32), we obtain the sought for transformation for the vari·
abies x, 'l:
+
x' ~ r (x iBi), i' ~ r (i- iRx). (2.35)
The equations for the transformation of the coordinates of the
event from the frame K to K' must differ from those for the
transformation of the coordinates of the same event from the
frame K' to K only by the substitution of the primed values for
the unprimed ones and vice versa. Besides, the sign of the ve-
locity V should be changed to the opposite. In this way we obtain
from Eq. (2.35)
X= r (x' - iBi'), i' = r (i' + iBx'). (2.36)
Of course, the same result would be obtained if Eqs. (2.35) were
solved directly with respect to x and i.
In Eqs. (2.35) and (2.36) one can easily change over to the
real variables x and t by substituting i = ict and T' = ict'. Then
we directly get the transformation formulae (2.16). The direct
and inverse transformations of the variables x and t have the
form
x'~r(x- VI) (a), X~ r (x' VI') (c),+
I'- r (1 --'i- x) (b), 1~ r(1' +; x') (d).
(2.37)
Subsequently a comparison of the Lorentz transformation given

in the form of Eqs. (2.25) and (2.32) will be of use to us. Recall
that in Eq. (2.25) all the quantities are real, while in Eq. (2.32)
a time variable is imaginary. Having made use of the relation-
ships (see Appendix I.§ 9)
cosia=cosha, sinia=isinha, tania=itanha,
we see that it is sufficient to substitute <p = - ie in Eq. (2.25)
in order to convert it into Eq. (2.32). Thus, in a plane of red!
variables x, t we deal in formal terms with the rotation of a Car-
tesian system through an imaginary angle. Such a rotation resem-
bles very little a true rotation of a Cartesian system, and Eqs.
(2.25) defining. it are only a "parody" of Eqs. (2.31) describing a
true rotation. We shall explain somewhat later in this section the
3'
. Special Theory of Refaliuify
geometrical meaning of the "'rotation" of the x, ' axes according

to Eqs. (2.32). and now we shall derive the Lorentz transforma-
tion in a symmetric form to be used hereinafter.
Let us introduce the symmetric designations of the basic va-
riables as follows:
x 1 =x, Xz=y, X3=Z, X~=icf=i't (2.38)
for imaginary time and
x0 =cl=-r, x 1 =x, x'l=y, x1 =z (2.39)
for real time.
The set of variables (Eq. (2.38)) is convenient to use when
relativistic electrodynamics is described. As to the set or va-
riables of Eq. (2.39), it is adopted in the book (91 which contains
a description oi the general theory or relativity. A transition from
the STR to the general theory of relativity is more expedient to
make without resorting to the number i. Let us rewrite the corres-
ponding transformation of variables (Eq. (2.30)) in the real form:
,;>~rx" + l'Bx'' +O · x'' + 0 · x".
x1 =rB.xO' + rx 1' +O· x2' +O . .x;J',
(2.30')
x' ~ 0 · x"' + 0 · x'' + I · x'' + 0 · x'',
il=O· xO' +0 · x1' +O·x2' + l·x3',
and using imaginary time (Eq. (2.36)):
Xt rx] + 0 • x2+ 0 • X.~ -i[Jrx.\,
=
x2=0 · xl+I· x2+ 0 • xJ + 0 • x~. (2.36')
X1=0· x~ +O·x~+ l•xi+O · x4.
x~=iBr · x( +O· xZ+O· x3+ r · xt.
The transformations (2.30') and (2.36') can be written in the
abbreviated form:
x1 = a, 11 x"' (a>. 1 x, = a,,.x; (b). (2Ao>
In Eqs. (2.40a) and (2.40b) summation is performed for an
index k running from 0 to 3 in (a) and from I to 4 in (b). The
inde:< i is "free" taking on all values from 0 to 3 in (a) and
from I to 4 in (b), the coefficients et,,. and a,,. making up the
matrices
a,.~(~B ~B
0001
H) (a) I a,.~( ~! ~ -~BI')
iBroor
(b),
(2.41)
Einstein Poslulales. Lorentz Transformation 69
respectively which 1.1re rderred to as the Lorentz transformation

matrices. Matrices of this form are always used for transforma-
tion of coordinates and time on transition from one inertial frame
of reference to another. These matrices differ only in the value of
the relative velocity ~·. i.e. in different values of B = V/c.
'j<' 'y
~/
'
'<
,-; ..
/, 1"'B~
/
(a)
Fig. 2.6. A geometric illustration of the Lorentz transformation The Lorentz
transformation reduces to the rotation of the ~ and T axes through the angle
To !h!rrctfi~aT ~~~~rodn!h;.,o~i~i~hi~ ;t~~i~k~U~~s ~ c~~;tdi:r~1 en~~g~a~~f:t~~
o;,
the O·t' axis and the straight lines 1' = const are parallel to the Ox axis (we
have passed over to the rectilinear oblique-angled syst!!m of coordinates). The
transition from the frame K to K' corresponds to the convergence of the r' and
Jf! axes (a); the inverse transition to the divergence of the T and x ax!!s (b).
The equations of an inverse transition, i.e. a transition from

the frame K to K', are obtained by substituting -8 for D. Let us
designate the matrix of the transition from the frame K to K' by
alt. so that
(2.42)
For the matrix with the real elements the indicated substitution
brings about a completely new matrix a/ 11 • But the matrix a/t
turns into ih, (with lines and columns transposed), when -B is
substituted for B, so that
(2.43)
§ 2.9. A geometric illustration of the Lorentz transformation.
Since in our choice of a relative position of inertial frames of
~~~:tdne~e t~~e t~~~~1~~~=tic!n ~~dt~e dr~f~~:n~:rfr·a~e~s i~u~~~ie;l~~~
(x, <).
Let the x and"( axes of the frame K be depicted by two mutually
perpendicular straight lines (Fig. 2.6). To draw the axes of the
70 Special Theory of RelaUvlty
!~!~~i:;n~0o~h!~ed~!f::e~~e1 ef~a11~e~es;r~~~ ~,·a;;·;~!nfrc~Tnc7Sii~~

(when x = 0 and t' = 0, x' = 0 and '(' = 0 as well). The pomt
x' = 0, i.e. the origin of the frame K' moves at the velocity V
relative to the frame K. Hence, its motion is illustrated in this
diagram by a straight line making an angle c:p with the axis t',
the angle 1p being defined by the relation rp = arctan B. But the
straight line x' = 0 is the time axis in the frame K'. Conse·
quently, the Lorentz transformation of the -r axis reduces to an
inclination of the T' axis by an angle c:p to the T axis.
The x' axis is defined by the condition t'' = 0. But from Eq.
(2.29) it is seen that this condition is satisfied in the frame K on
the line -r = Bx. Of course, the -r' axis could also be found from
the condition x' = 0, but then we would obtain the same straigllt
line x =:= Ht from (2.29). Thus the equations for the new axes will
be written as follows:
axis T': T=fx: axis x': T=Bx. (2.44)
The angle between the x and x' axes is defined from the relation
1p = arctan B. Thus, the Lorentz transformation reduces to a con-
version of the rectangular reference frame x, T to the oblique-
angled one x', T'; the x and T axes rotate around the origin in
the direction of the bisector of the coordinate angle through the
same angle cp =arctan B (see Fig. 2.6a). This is what rotation
through an imaginary angle means! In formal terms rotation of
a rectangular system just considered does not at all resemble ro-
tation of a Cartesian system of coordinates.
Our result shows that we cannot remain within the framework
of orthogonal axes x, T when considering inertial frames of ref·
erence and resorting to a geometric illustration of this transfor-
mation. Even if the axes of the initial frame are orthogonal, a
transition to any frame K' makes it oblique-angled. Fig. 2.6b illu-
strates the transition from the orthogonal frame K' to K accord·
ing to Eq. (2.30). But an emergence of oblique-angled coordinate<;
makes it necessary to distinguish between co- and contravariant
coordinates (see Appendix I. § 8). That is why it is so difficuit
to bypass these notions in the STR without hiding beyond the
number i (see§ 2.8).
CHAPTER 3
CONSEQUENCES
OF THE LORENTZ TRANSFORMATION.
THE CLASSIFICATION
OF INTERVALS AND THE PRINCIPLE
OF CAUSALITY. THE K CALCULUS
One should not expect to obtain any new consequences of the

Lorentz transformation which are not obtainable directly from
the Einstein postulates. In the final analysis the Lorentz trans-
formation itself is a consequence of the Einstein postulates. The
beginning of this chapter is indeed dedicated to the analysis of
the results that have been already obtained in Chapter 2. Surely,
they are obtained in a much simpler way from the Lorentz trans-
formation and we shall take advantage of this fact. We shall not
even discard an opportunity to show how all these consequences,
including the law of velocity transformation, can be obtained even
without resorting to a construction of a coordinate system (the
K calculus). Naturally, the question may arise as to why the
Lorentz transformation is significant, since all the results ob-
tained until now can be derived by other means. The point is that
despite the significance of the results obtained, they are not every-
thing we need yet. In order to get convinced in the validity of
the principle of relativity, one has to know how the basic equations
of physics are transformed on transition from one IFR to another.
It is the Lorentz transformation that makes the basis for such
transformations. '
§ 3.1. On the measurement of lengths and time intervals. The
relativity ol simultaneity. The Lorentz transformation makes it
possible to compute the coordinates of an event (including a tim<!
"coordinate") on transition from one IFR to another. But an
event is only an element of a physical phenomenon, and in th~
final analysis the major task of the STR is the corresponding
computation of the physical quantities observed. However, prior
~~o~~iJ~ar~nforth~a~~~k pi~Y~~c~re~~=~1it\~s.dw;~1e onm~~~ ~~~~~~~~1
physical measurements are those of distances (lengths) and time
intervals. We are accustomed to measuring lengths of objects or
distances between points, when these objects or points are motion-
less relative to us. If an object is at rest, it is sufficient to trans-
pose a unit scale along the length to be measured the necessary
number of times. This is how we do it in our everyday life, when

measuring, say, cloth or the length of a room.
It is also rather easy to measure a time interval between two
events occurring at the same point where a clock is located. One
has only to register the moment when the first event occurred and
the moment of time when the second one did. The difference of
the clock's readings will give the time· interval between the events.
In just this way, for example, a duration of a lecture or a football
match is ascertained.
But how to measure the length of an object moving relative to
us? Let a train move past us at a great velocity and we want to
determine its length. It is far from being simple for one man to
do this. He has to note simultaneously the positions of the head
of the train (the locomotive) and the tail of the train relative to
some certain motionless points on the ground. But as soon as he
notes the position of the locomotin and begins to turn his head,
the tail of the train will have gone ahead. Consequently, one
should take special care to mark simultaneously the positions of
the head and the tail of the train.
Having marked the simultaneous positions of the head and the
tail of the train on the ground, we can readily measure the
distance by conventional means used for measuring motionless
objects.
And how to measure a time interval between events occurring
at different points in space? Recall how they measure a time which
a sprinter takes to run a hundred metre race. The events in this
case are represented by the sportsman's start and finish. And there
is only one clock! A starter's shot serves as a signal to start the
race and to actuate a timer located at the finish. Sound propa·
gates in air at the velocity of 330 m/s, so that the sportsman will
start running before an umpire located at the finish actuates his
timer. This is not very essential, though, because the velocity
of the runner is very small (at best about 36 km/h = 10 m/s,
which is small even relative to the velocity of sound). But from
this example one may perceive that the determination of a time
intcn·al between two events happening at different points in
space requires attention.
In § 2.1 we discussed how such a problem is dealt with in the
theory of relativity: each inertial frame has its own coordinate
system and motionless clocks are located at all its points,
\\ herever needed. These clocks are synchronized within that re-
ference frame, so that equal readings of the clocks correspond
to the same moment of time in the frame.
When we turn to comparing events in the two inertial frames K
and K', we link the readings of the synchronized set of clocks of
the frame K with the readings of the analogous set of the frame
Consequences of Lorentz Transformation. KCalculus 73
K' by assuming t = 0 and t' = 0 at the coinciding origins 0

and 0' (see§ 2.2). Recall that the Lorentz transformation is just
a conversion of the coordinates of the event in the frame K, i.~.
the coordinates (x, y, z, !), into the coordinates of the same event
in the frame K'. From Eqs. (2.16) or (2.29), or (2.40), we obtai;,
(x', y', z', t') expressed via (x, y, z, t). Of course, all these sur-
prising, from the viewpoint of the "common sense", consequences
of the Einstein postulates that we discussed in Chapter 2, can be
obtained from the Lorentz transformation. Now we shall be oc•
cupied with this.
First, let us derive two convenient equations that we shall need
later. Consider the two arbitrary events I (x,, Y~o Z~o t 1) and
II (x2, y 2, z2, t2 ). The conversion of coordinates and times of these
events into the frame K' yields, according to Eq. (2.37a, b),
t;~r(t,--fx,).l x;~r(x,-Vt,),
t;=r(t,-fx 1) . x;=r(x 1 -Vt 1).
Making up the differences 12- t~ and x2- xi, i.e. subtracting
the lower equations from the upper ones, and designating 6.x =
=x2 -X~o !J.x'=x2-xl. M'=t2-t;, M=t 2 - t 1, we obtain
the necessary equations (the inverse transition equations are also
\\ritten out):
6x' ~ f(6x- M),V (3.1) 16x~ r (6x' +V
M'), (3.1')
u~r(M--f6x). (3.2) M~r(M+-fu). (3.2')
In fact. Eqs. (3.1) and (3.2) as well as Eqs. (3.1') and (3.2') are
the Lorentz transformations for differences of spatial coordinates
and times of the two events. These equations should be suppl~
mented with the relations !1y' = 1.y and !J.z' = !1z.
It immediately follows from Eq. (3.2) that two simultaneous
events in the frame K are not simultaneous in the frame K'. In-
deed, assuming !11 = 0 in Eq. (3.2). we obtain
llt'=-f*b.x. (3.3)
It is seen that 1.1' "4= 0, if 6.x -4= 0. But if !1x = 0 at !1t = 0 as
\Veil, the events either coincide or happen in the plane x =canst.
For such events 1t' = 0.
Two conditions follow from Eq. (3.2). Provided they are satis-
fied, one may ignore the relativity of time intervals between
events. First, one should suppose B « I; then r ;: : : I and the fol-
lowing expression can be written: !1t' = M- (VIc) (.1.x/c). The
second term in this expression can be ignored if the ratio IJ.x'r
is rated small. This second condition is positively satisfied if

events occur in a limited region of space along the x axis, i.e. at
small ~x. No limitations are imposed, however, on the region of
space along the directions y and z, since all events in the plane
x = const happen at the same moment of time according to the
clock of the frame K'.
Certainly, the relativity of simultaneity reveals itself in the
dis-synchronization of clocks that we examined in detail in § 2.4.
U is sufficient to put t' = 0 in Eq. (2.37b), and we immediately
obtain Eq. (2.9) which did not come easily before. As soon as we
assume 11t = 0 in Eq. (3.2), Eq. (2.8) is obtained. This example
shows how much can be hidden in plain, by appearance, "trans-
formations".
In everyday life a violation of simultaneity is not perceptible:
the time difference !lt is proportional to B/c, as it is seen from
Eq. (3.2). provided that M = 0 is assumed (simultaneity in the
frame K). From the same Eq. (3.2) it is seen that if ~ !lx has
a substantial value, !lt' can also assume a substantial value pro-
vided that 11x is great.
It is very important to point out that the relativity of simul-
taneity is dictated by the finiteness of the velocity of light. If d
formal passing to the limit c-oo is performed (in fact. it means
that B-0), simultaneity becomes absolute. This result corres-
ponds to the case of small relative velocities of reference frames.
It is seen from Eqs. (3.1) and (3.1') that two events occurring
at points of space with the same x coordinate in the frame K. i.e.
at the same point of the frame K, will have different x' coordinates
in the frame K'. Indeed, from Eq. (3.1) we get 11x' =
= f(!lx- V !lt) = - rV 1.t. But 11t is a proper-time interval, and
so r tit =!it'. Hence, !lx' = - V !lt'. The meaning of the last
result is evident; it defines the displacement of the point x rela-
tive to the frame K' registered in the frame K'.
§ 3.2. Relativity of length of moving rulers (scales). A visible
shape of objects moving at relativistic velocities. Let us consider
now a measurement of the length of a moving ruler. Let the
clocks in the frame K be synchronized and spatial marks made.
Suppose that a ruler oriented along the x axis moves relative :o
the frame K at the velocity V. One can fix the frame K' to this
ruler. How to measure the length of the same ruler in the frame
K? Obviously, the coordinates of the ruler's ends have to be di!·
termined simultaneously in the frame K. This requirement of si·
multaneity leads us to the strange result which we have already
discussed in § 2.3: the ruler's length, when measured in a rci-
erence frame relative to which the ruler moves, turns out to be
less than the length of the same ruler in the reference frame
CollSequenccs of Lorentz Transformation K Calculus 75
''here it is at rest. So, let the ruler be at rest in the frame K' and
the coordinates of its ends be xl and x~. By definition, its length
in the frame K' called, as we have indicated, the proper length
is equal to x2- x;. The proper length of the ruler is designated
by l 0, i.e. ! 0 = xl.- x]. Since the ruler is motionless in the fram~
K', one may not worry about the simultaneity of measurements of
the coordinates of its ends: its length can be measured by any con·
ventional means.
In the frame K the coordinates of the ruler's ends will be de-
termined according to the Lorentz transformation (see Eq.
(2.37a)):
x~=r (x2 - Vt 2), x~ =f(x 1 - Vt 1).
Having formed the difference x2 - x~, we obtain
x;- x;- r ((x2 - x,)- V (12 - t,)). (3.4)
The proper length of the ruler constitutes the left-hand side of

Eq. (3.4). In the braces of the right-hand side there are 6.x and
fj.t for the two events, the position of the ruler's left end being x- 2
at the moment t 2 and the position of the ruler's right end being x1
at the moment t 1 (in the frame K).
The quantity 6.x will be the ruler's length in the frame K only
if the positions of the ruler's ends are registered simultaneously
in this frame. Otherwise we can get any value for 6.x. Recall th~
example cited at the beginning of § 3.1 concerning the determina·
tion of the length of a moving train: you have just marked th?
position of the tail of the train and are slowly turning your head
toward the locomotive. If you mark the position of the locomotive
and then measure the distance between the marks, the distance
measured will be greater than the proper length of the train. Now
proceed in the opposite direction: first mark the position of the
locomotive and then turn your h~ad slowly to the tail of the
train. It is easy to contrive that if one turns his head slowly
enough, the train's length may even get equal to zero. This is
just what Eq. (3.4) expresses. Thus, in order to determine the
ruler's length unambiguously in the frame K, one has to consider
the two simultaneous events in K: the coincidence of the ruler's
left end with a certain spatial mark, say x 11 and that of the ruler's
right end with another spatial mark, say x2• This means that
/).x = l only if ~~ = 0. But then it follows from Eq. (3.4) that
h = rl, where l is the ruler's length determined in the frame K.
According to custom the last equation is written as follows:
./-V' ==! -v'I-82.

l= ~~~ 1 - 7 (3.5)
0
Of course, under the same conditions of the problem one may

make use of the inverse transformation equations:
x,~qx;+ vr;). x, ~qx; + vr;).
Subtracting the right-hand equation from the leU-hand one, we
obtain
t.x~rt,+ rv (t; -r;) ~n,+ rv M'.
But l:!x becomes the length l only under the condition M = 0.
Having expressed l:!t' in terms of 11t and 6.x from Eq. (2.37b).
AI'= r ( M- f 6.x), and having assumed M = 0 in this equa·
tion, we obtain Eq. (3.5) again.
o' o'
e:t----------------,:)
B
Fig. 3.1. Measurement of the length of a moving ruler.
Here are two more methods of determining the length of a

moving ruler which, naturally, bring about the same result. Let
the ruler AB be at rest in the frame K, so that its left end tt
coincides with the origin 0 (Fig. 3.1). At the moment t = 0 the
origin 0' of the frame K' coincides with 0, the clock's reading
in 0' being t' = 0. Then the observer located at 8 registers the
moment when 0' passes the point B. Let it be the moment t. Th~
velocity of the frame K' is known. Therefore, the proper length of
the ruler 10 = V.1.!. This is the proper length of the ruler becau<>e
it is measured by the scale and clocks of the frame K in which
the ruler is at rest. The velocity is also determined in the frame K..
On the other hand, the observer from the frame K' located at the
point 0' will register his passing by the points A (at the moment
t' = 0) and 8 (at the moment t') using his clock. But the ruler
moves past this observer also at the velocity V (in the opposite
direction). and he will find the length of the moving ruler to be
equal to l = V ·..'11 1 • But ~!' = (!' ~ 0) is the proper-time interval
between the two events, the coincidence of 0' with A and then
with JJ. As to ..lt = (t- 0) = t, it is the time interval between
the same C'vents in the frame X. According to Eq. (2.2) ~t=r !1t',
Consequences of Lorentz Transformation. K Calculus 77
and inasmuch as 7;;= ~; ===+·

l=-f-l0 =lo-v'i-H2•
The relativity of lengths is a direct consequence of the relativity
of simultaneity. Let a ruler be at rest in the frame K', the coor·
din~tes ~f its ends being xl and x\. The proper length !0 = l!x' =
=X!-XJ.
Let two light bulbs fixed at the ruler's ends in the frame K'
flash simultaneously (1.t' = 0), and let these two events be reg·
istered in the frame K. Let us find the distance between the
Fig. 3.2, Relativity of lhe rulers' lenglhs as a consequence of the relativity of

simullaneily of two evenls.
points at which these events occur in the frame K: l!x =

= l'(l!x' +
V !!t') = f!!x' = rt
0 (see Eq. (3.1')). This means that
the distance between the points at which these two events occur
is greater than the proper length of the ruler. But this distance is
not the ruler's length which will be measured by the observer
from K. In order to find the ruler's length in the frame K. one
must find the coordinates of the ruler's ends simultaneously in
this frame. The events which arc simultaneous in the frame K'
are delayed relative to one another by lit= I' ( !!t'
= 1'*/ 0 in the frame K (Fig. 3.2). But during this time the
+
l!x') = *
ruler's end x] will shift in the direction of motion by the
distance V ~t = 1'132 / 0. Hence the measured length of the ruler
will be less than 1'/0 by V ·M, i.e.
I~ 1'/0 - I'll'/,~ I, -v'T=!l'
We shall also recall that Eq. (3.5) was obtained directly from
the Einstein postulates (§ 2.3). Now it is time to dwell on the
physical meaning of Eq. (3.5). We have found that the length
of a physical object, for example a ruler, is relative, i.e. different
in different reference frames. The ruler possesses a maximum
length in the frame in which it is motionless, i.e. the proper length
is the greatest. If the ruler's length is determined in an inertial

frame, relative to which the ruler moves, its length will prove
to be less than the proper length. It follows from Eq. (3.5) th~t
if the ruler were able to move at the velocity c, its length would
be equal to zero. But this just cannot happen: any object pos-
scsstng a finite rest mass, including any feasible frame of re-
ference, cannot reach the velocity c. ·
What does a contraction of a ruler mean? Frequently one may
hear the question as to whether the ruler "actually" gets shorter.
To begin with, it is clear that no contraction of the ruler can
tal>e place. This follows from the basic principle of the STR,
the principle of equivalence of all IFRs. The physical state of the
ruler is the same in all IFRs. So there is no question of an
emergence of any stresses leading to the ruler's deformation. The
ruler's "contraction" comes about solely due to differing methods
of length measurements in two reference frames. On the oth.:::r
hand, the observed relativity of the ruler's length is not due to
the observer's illusion. This result can be obtained by any reaso·
nablc method of measuring the length of a moving object. Mor~·
over, analysing physical phenomena in a given reference fram~.
the quantity I should be adopted as the object's length according
to Eq. (3.5), and not 10•
It is extremely unfair to speak of the "Lorentz contraction"
when alluding to Eq. (3.5), although indeed G. A. Lorentz was
the first to suggest this equation in 1892. However, it was inter·
preted quite differently (see Supplement II) from what we have
just discussed.
It was Einstein who clearly said about the reality of the Lo·
rentz contraction: "There is no point to question whether the
Lorentz contraction is real or not. The contraction is not real,
since it does not exist for an observer moving with the object.
However, it is real, since it can be fundamentally proved by
physical means for an observer not moving with the object."
Another question is often raised: what is the "actual" length
of ihe ruler? This question has no meaning, if asked "in a broad
serose". The question about the ruler's length reg::udless of a
frame does not have any meaning. The ruler has its length in
each reference frame; this is just its "actual" length. All inertic~l
frames of reference arc equivalent and so are the ruler's length
Yalues determined in these frames. In any reference frame the
ruler will behave as if it has the length determined in this fram.~.
Although all IFRs are equivalent among themselves, there is still
one "selected" coordinate system to which we are accustomed.
This is the system in which the ruler is at rest. From the view·
point of our customary concepts this is just the "actual" length of
the ruler. We are prone to adopt it as a true length, but this
length defines the behaviour of the ruler only in this "inherent"

reference frame.
Finally, the last remark. The ruler exists objectively, i.e. outside
our consciousness and ourselves. But is there any length before
measurements are made? A length as a certain number emerges
as a result of measurements and a choice of units of length. Of
course, the ruler possesses extent, or length, H you want, as a
quality before a measurement, but the numerical value of length
originates only after measurement. Thus, the numerical value of
length of existing objects emerges after measurement, and the
result of measurement, as we have established, depends on what
kind of instruments are used.
Let us consider rulers having the same proper lengths I~ and L'O
in the two reference frames K' and K". Measurements carried out
in the frame K' will give
l~>l~. (3.6)
Measurements carried out in the frame K" will give
~~ >l~. (3.7)
The inequalities (3.6) and (3.7) are far from being contradictory
because (3.6) was obtained for the scales and clocks of the frame
K' and (3.7) for the scales and clocks of the frame K". Th~
difference in the values of the ruler's length is dictated by the
fact that simultaneity in the frames K' and K" is defined different-
ly. The difficulty in the interpretation of the conclusions of the
special theory of relativity lies not in the existence of relative
quantities, but in the detection of the equivalence of all inertial
frames of reference. The inequalities (3.6) and (3.7) indicate just
this equivalence.
The conclusion of the special theory of relativity concerning
the relativity of the length of a moving object is unusual partially
because in our everyday life we do not perceive such an effect.
Let us consider the fastest motion within reach, that is the orbital
motion of the Earth. In this case V = 30 km/s. The ratio VIc ~
~ IQ-4 and I= t0.,Y I - 10- 8 ~ /0 (I - -4- · 10-H) ~ l0. It should
~=q~~~~=e~f at~!ifi~i~!~ets~e ~f0 ~~~a~t~?:ci~; ~er"fii~t.islfa tg!r~~locc~~;
of light were infinite, the ruler's length would be the same in all
reference frames in accordance with Eq. (3.5). This can also b~
seen from the fact that in the case of c-+ oo the simultaneity of
events becomes absolute.
Although until now we always discussed the relativity of
lengths of objects (rulers), it should be borne in mind that actually
we dealt with the relativity of distances between two motionless
Spec/41 Theory of RelatirJitg
points in one reference frame when measured by instruments from

another frame.
Let us consider a cube at rest in the frame K with the side<;
.1.x, 1.y, 1.z and the proper volume 1.1"0 = 1.x 1.y ~z. According to
the Lorentz transformation in an arbitrary IFR K' we have
l!J.x' = (1/f)l!J.x, lly' = ~y. llz' = ~z and, consequently
~r' ~ 6x' 6y' ~z' ~ (ltr) ~x ~Y ~z ~(Iff) 6r0 •
Hence, the change of the cube's volume on transition from t ... e
frame K to the frame K' is determined as
r' ~ r, .y 1 - B'. (3.8)
It follows from the result obtained that the proper volume of an
object is the invariant of the Lorentz transformation:
r' d'P'' ~ r" d'P'" ~ r,.
Is it possible to observe directly the Lorentz contraction by, say,
taking photographs of a rapidly moving object? In the first paper
by Einstein dedicated to the theory of relativity one mc:y read the
following: a moving body which at rest has a spherical shape is
observed from a stationary frame as an ellipsoid with the semi·
axes R (I - ~ 2 ) ''•. R. R. The word "observed" here can be inter-
preted as a visual observation or photographing. F9r about fifty
years -~fter the advent of the STR everybody wa~- sure that a
visible shape of a retativistic ~phere was an ellUJsoid. However,
it t1,1rned out that_ the problem of the_ viSL_bfe ~_h_ape __ Qf objecls -
moving a[ a relalivistk velo"city requires __i1IED-Y-~irc\.unstances __to
be takeri 1nto account, and tnat a -raprary- moving sphere remains
sph~lf one assumes that an eye and a photographic plate
fiX an instantaneous image produced by light, this will mean th::~t
the image is produced by rays coming from different sections of
the observed object simultaneously on a retina or a photographic
plate. But if the optical paths of light going from various points
~!g~~t:r 0~~=r~~~it~b~:c~f athee ~~~:~~~t,p~in~~o!~gJifiPe~~~tp~~~~ill~
of time prior to the moment of photographing. The whole effect
~~ac1~~~:~ ;~ !~:~~~~~~ve~h~f !h~i~~ll~c~fi'a~~ 1~fh:. ~~~?ngg as&~:~·~
coincides with that of a stationary sphere.
Let us imagine a luminous cube moving along a straight line
parallel to one of its faces past a photographic camera or an
observer. The photographing or observation takes place at the
moment when the centre of the cube crosses the line perpen-
dicular to the motion direction and drawr. through the point at
which the observer is located (Fig. 3.3a). Naturally, we mut>t
Consequences of Lorentz Transformatlnn K Calculus 81
know beforehand that the moving object has the form of a cube
in its own reference frame.
At a definite moment all photons emitted simultaneously (in
the frame fixed to the plate) at the points of the line AD will
reach the plate together with those emitted at the point 8 earlier
by the time interval lfc, I being the edge length of the cube. But
at that moment the point 8 was in the position 8'. The simul-
taneous determination of the
positions of the points A and B A B
D in the frame fixed to th~ f-ht.B c lj] tl/Fjil (b}
plate leads, in accordance Ltgkt tJ
with the conventional rule oi
length measurement, to the
Lorentz contraction: l' =
A ~
ni,,J
= l(l-p 2 )'''· On the other
hand, 88' ~(If c) u ~ ~1.
From Fig. 3.3b and c one
Lrght ! L1ght
tW
lsinp tcos~p
~
can realize that the picture
of a moving cube that would
be seen by a motionless id~
alized observer coincides with
the picture of a motionless {d) Ohserver
cube turned through a cer·
tain angle cp. This angle is fig. 3.3. Visual observation of a cube mov-
determined from the relation ing past an observer: (a) the mutual d1s·
sin cp = ~- This is a particu· posilion of the observer and the cube at
Jar case of a more general 0=0; (b) the visible p1clure oF the mov·
ing cube, (c) the possible inlerpretal10n
result: any three-dim~l of the visible picture by one obs!!rver. the
moving-objeCiiSSeen __turned rotation of the cube through the angle
jit -a --given fuOrn~~T. Tf- tfie f41 = arcsin j3: (d) the observation of the
cube is so positioned relative moving cube at the angle 0
to the observer that it is seen
at the angle a' relative to the x' axis when at rest, it will he
observed turned through another angle. If the cube is removed
far enough from the observer, light travelling from it can lle
taken for a parallel pencil of rays. When this pencil is observed
in the frame K. it propagates at the angle a to the x axis, as
seen by the observer from K. the angles a and a' being related
by the equation (see Eq. (7.11)):
cosB~(cosB' +~)/(! +~cosB').
A variation of a plane wave front direction on transition from one

reference frame to another, moving relative to the former one,
is called a~_~berr<!tiqn of light The image registered on the plate
in this Case corresponds to the cube observed in the frame K at
the angle a and turned through the angle a- a'.
82 Special Theory of Relatluily
From this it is clear that a sphere turns as well, but its

outline does not vary, of course (see [28] for details).
And still- can an object be photographed, so that a plate will
register a relativistic contraction? To avoid difficulties associated
with a turn, one can consider a one-dimensional object which is
easy to compare with its own image on the plane. For this pur-
pose the observer in the frame K must know beforehand that a
rod moves along a given direction. The rod is at rest in the
frame K', and its proper length is also known. In this case the
/(1 Kl
L....u oi lo
/
l ~ .r,.r'
,.//"
:p'
1/
~
1-'llotqgrophJ.C plate
~~~- t!' 4be1 ~ho~~~i~a=~~~t~~:~t ~~:r:;;/~~?eg o1r~h; 0 ~~~~0~o;~[:c~~~~ f~eaii~:'~r:J.

a special device opens the shutter at P for an instant to pass through the rays
emtlled by the rod's points at the moment when the point 0' crosses the line PO.
observer constructs a counterpart of the moving rod in his

frame K and takes a photograph of the moving rod against the
background of its own length.
The simplest arrangement intended for photographing a rocl
experiencing the Lorentz contraction could be of the type shown
in Fig. 3.4. The rod is parallel to the x axis and moves along it.
An observer is located on a perpendicular to the x axis, the per·
pendicular going through the middle of the rod's counterpart at
rest in the frame K'. As soon as the middle of the moving lumi·
nous rod finds itself on the perpendicular, a mechanism trigger!>
a camera's shutter, and the simultaneous positions of the rod's
ends get registered on a photographic plate. As to the stationary
counterpart of the rod, it can be photographed at any time, of
course. Most likely, this is also an "imaginary experiment".
It should also be mentioned, however, that a "turn of a cube"
moving at a relativistic velocity was qualitatively photographed.
We refer the reader interested in details to (27].
In general, a "visible" picture may differ quite essentially for
different observers. Here is a simple example. Let an observer
Consequenres of Loren/? Tran.s/ormafion K Calculus 83
be at some distance from a plane at which electric bulbs flash

simultaneously in the frame where the plane and the observer are
at rest, or the light flash is produced. Then due to the finiteness
of the velocity of light the observer will see that the plane starts
glowing gradually with an "illumination wave" running from the
centre to the periphery. In particular, if an infinite thread is illu-
minated instantaneously, the remote observer will see two lumi-
nous points running apart.
Owing to the sam~ reason a visible (observable) velocity---.ll!Y
~a! one: if may prove to be even hi_g_her th_an
the velocity' of ligm. --- ...
- § 3.3. Relativity o hme mtervals between events. Suppose that
two events occurred at some point of the frame K' at time mo-
ments 1' 1 and 1'2• The time interval separating these events can
be registered by the clock located at that point. According to the
defmition a time interval between e.Y.eni.s____th.ai....Q.ccurred at the
1 1
:::: c ~~nt 0ol a sce;!~~ rsef:;e~ce f~~;er~L~~i*~I!Jf'IY-!~~·
the events. bestgnatlng the proper-time interval by t1! 0 , we obtain
Tn-ouTcase
!J.i 0 =llf'=t~-t~,
Let us define now a time interval between the considered events

in the frame K. According to Eq. (3.2) tJ.t = rAt'= r M 0• This
result is already familiar to us (Eq. (2.2)).
However, if the events in the frame K' occurred at one point
in space, this is not the case in any other frame K. Indeed, let
two events occur in the frame K' at different moments but at one
point, t.e. t1x' = 0 but !lt' =F 0. According to Eq. (3.1) 1x = rv.
t1t' =F 0. The meaning of the last result is obvious: all points of
the frame K' move relative to K at a velocity V, and 1x is just
a displacement of any point of the frame K' during the interval
considered. Since ftl.t' = ft1t 0 = At, then t1x = V tl.t.
Therefore, time intervals between the same pair of events turn
out to be different in different IFRs. We shall observe the least
time interval between events in the reference frame in which these
events occur at the same point and, consequently, are registered
~{~same clock. In other words, E!,.e proper-time interval is the
A time interval between events in terms of the frame K can

also be measured as follows. The point of the frame K' at which
the two considered events occur, moves at the velocity V re\ativ~
to the frame K. If the events occurred in the frame K at the
points x 1 and x 2, the time interval tl.t between the events is,
obviously, equal to t1t = tl.x/V. But according to Eq. (.3.1) ax=
= I'V .lt' (.lx' = 0), whence we obtain again
llt=rllt 0 =~ (3.9)

~
The direct experimental confirmation of the conclusion of the

STR aiJout the relativity of time interv.als is widely known. Light
elementary particles '(muons) were discovered, on the one hand.
in a laboratory as a result of nuclei splitting and, on the other
hand, in cosmic rays. The lifetime (the half-life) of muons
mea~ured in laboratory conditions proves to be equal to about
2- I 0- 6 s. This lifetime can be regarded as a proper lifetime, since
the velocities of laboratory muons are non-relativistic (see E4.
(3.9), the velocity of the coordinate system K' is the velocity of
the frame fixed to a muon). In the intervalllt 0 = 2-J0-6 sa muon
breaks up into other particles.
It is known that the muons observed in cosmic rays at the sur-
face of the Earth originate in the upper layers of the atmosphere
at the height of from five to six kilometres due to the primary
cosmic radiation. The velocity of the generated muons moving
toward the Earth is comparable with that of light. According to
Eq. (3.9) the half-life 6.t of a muon in the laboratory frame is
equal to 6.t = r .1.t0 . In the case of muons r ;: : : : 10 and in th~
laboratory frame of reference 6.1 = 2-I0-5 s. During this time a
muon travels the distance eM= 3·10 10 X2·I0- 5 ;::::::: 6 km. But
for the relativity of time intervals, muons would have travelled
only about 600 m and we would not have observed them at th~
sea level. Thus, only the relativistic transformation of time in-
tervals makes it possible to explain muon showers observed or~
the Earth.
An excellent illustration of the relativity of time intervals is
p1ovided by the Doppler effect.\ This effect consists in the fact
that if a light sOJJrr"e---arrd--arr"1iDserver (a receiver) move relative
to each other*, the light frequency determined by the observ:::r
differs from the frequency that would be observed by him if he
\\as stationary relative to the source. It is natural to refer to a
ligltt frequency determined by the observer stationary relative to
the source as a proper frrquency of light. Let us designate it by
w0• All our conclusions pertain to vacuum.
Let us first consider the case \vhen the direction of light propa-
gation coincides with the direction of the relative velocity of lhe
sourcf' and the observer, the so-called radial Doppler effect.
alv.~ay~o~etilti~:t amty0p~~arl~bsr~f~~~\st1~s~ussss~~np5tL~~ ~~!~s~~al~ar;li:Jcsm~~~u~hai~

needed for light propagation. That is why only the relative velocity of a light
source and an observer is essential for us
Consequences of Lorentz Transformation K Calculus 8S
In Fig. 3.5a this corresponds to light propagation along the x, x'

axis. To consider the Doppler effect, one can imitate a light wave
by sending short pulses from the source at the interval (period) T.
Let us suppose now that such pulses are sent from the origin of
the frame K. i.e. from the point 0. Any observer at rest in this
frame will discover that these pulses come to him at the same
intervals T. Now let us fix the observer to the frame K'. Let the
first pulse be sent from the point 0 at the moment when the
origins 0 and 0' coincide (t = 0, t' = 0). Naturally, the same
pulse will be registered by the ollserver at the point 0'. The next
~
L
n"l v,' ----- 0
Source
t!rz.x"
L
Observer
(b)
f1g. 3.5. The deri\·alion of equations of the radial Doppler elfect: (a) an
observer moves away from a source; (b) an observer approaches a source.
pulse leaves the point 0 after the time interval T (by the clock
of the frame K). But at that moment the origin 0' is already at
the distance VT from 0. The velocity of light relative to 0' in the
frame K is equal to c- V, so that a light ray will need the
additional time VT/(c- V) to reach the origin 0'. Therefore, the
observer at 0 will receive the pulse after the time interval
T'=T+ c~TV =(1 + 1 ~1~/c)T=~T (3.10)
when registered by the clock of the frame K. We have oUtained

the time interval between the first and the second signals received
by the observer at 0' and registered by the clock of the frame K.
The reception of the signal in the frame K' takes place at one
point 0' and the transition to the proper-time interval can be
accomplished in accordance with Eq. (3.9): T~ = (1/1') T', so that
the period TO will be equal for the observer at 0' to
' I
1-U
,--------;-;;
To=--T-vl-lY= ,YT+lf
-I --lT
l
. (3.11)
Passing over to frequencies (Wo=2n/T, w'=2n/T~). we obtain

JT="B
w ~'VT+lf"""'""(l-
I • B
). (3.12)
86 Special iheory of Relatlvify
The last operation is performed in the easier way as follows:

the numerator and denominator under the radical sign are multi-
plied by (I -B). The right-hand side of Eq. (3.12) is immediately
obtained when the term B2 is ignored in the denominator ex-
pression ( I - B 2 ). This equation is valid for an observer moving
away from the source: the observed frequency is less than the pro-
per one. When an observer approaches the source (in Fig. 3.5b the
point 0" is to the left of 0) and the signals are sent from the
frame K", the analogous reasoning (the source and the observer
converge and the relative velocity is c + V) brings about the
equations T" = 1 ~ 8 T, T'O = ~: ~: T and finally
(3.13)
Eqs. (3.12) and (3.13) (without approximations) are the exact

relativistic equations describing the radial Doppler effect. Th.:!y
will be obtained later (§ 7.2) on the basis of strict rela~
tivistic equations. But the derivation cited here is faultless in
terms of physics. At the same time it clearly shows that the
Doppler effect is formed of two independent parts: (I) it is con·
nected with a continuously changing distance between the observer
and the source; (2) it is also connected with the transformation
of time intervals between events on transition from one reference
frame to another. The first factor does not pertain to the theory
of relativity in the slightest degree. The radial Doppler effect
follows qualitatively from the classical theory, with the corres·
ponding equation being obtained from Eq. (3.10). There is noth·
ing to change in Eq. (3.10) in terms of the classical theory, since
time intervals in all reference frames are the same. The differen~e
between the classical equation and the relativistic one is essential
only to the order of magnitude of B2• The last approximate equa-
tion in (3.12) just gives the classical expression obtained from
Eq. (3.10). Inasmuch as the ratio B is determined by the relative
velocity of the source and the observer, it is very small at least
for macroscopic sources, and the Doppler effect is defined pri~
marily by a variation of the distance between a source and an
observer. However, there is a case of a zero relative velocity of a
source and an observer, although the frames, in which the source
nnd the observer are at rest, move relative to each other. This hap-
pens when the moving source is obser\'ed at the moment when its
\'elocity is perpendicular to the observation directton (the line of
\'ision) (Fig. 3.6a). At the moment of observation illustrated in
Fig. 3.6a the distance between the source and the observer doe~
not vary. Consequently, no Doppler effect is possible from a clas-
Consequences of Lorenlz Transformation K Calculus 87
sical viewpoint. But in terms of the relativistic theory the period

TO between signals in the frame K' is the proper-time interval
and roo-= 2n/f0. Having converted TO into the observer's time ac·
cording to Eq. (3.9), i.e. T=T~1/l-B 2 , we obtain the equation
of the transverse Doppler effect:
w'~vl-B'""· (3.14)
This is the equation of the second order with respect to B. The
transverse effect is more difficult to observe than the radial one,
/fj
o'If~----~x' -~r,
Source :
-----c+-
~xvrcose:
Obsef'l!er
{a) (h)
Fig. 3.6. The derivation of the Doppler effect equations: (o) the transverse
effect: (b) the general case.
but still it was observed in 1938. Its discovery, as it is seen from

the foregoing reasoning, is a direct evidence of the relatiYl!y__gf
~-inte.rv_aJ.s____Qetween_events. It-s1tollltl.l:ie emplla.Sized once more
~v~e~~ee~gr~~~~e~4~~nf2j!~~:-
be derived if the radiation at the angle to the direction of the
source motion is considered (Fig. 3.6b). If the first pulse is
emitted at the point A and the second at the point B, the path
difference of parallel rays travelling at the angle e to the velocity
direction is equal to VT cos e. It is clear from this that
T' = T- VT ~os 8 in terms of the observer, and, consequently,
w' = ---v-.
1--;:cose
Note that w' and w0 in this equation are mea·
sured in the same reference frame. Thus we make sure once

again that there is no transverse Doppler effect (8 = n/2) in
classical physics: w' = wo.
We have made sure that when two events in a certain reference
frame occur at the same point, i.e. are single-positioned, the time
.. Special Theory of Relativity
interval between them is defined as the proper-time interval, i.e.

is measured by one clock, whereas the time interval between the
same events in any other frame can be calculated according to
Eq. (3.9). The following question arises: is it always possible to
convert a time interval found in an arbitrary reference frame into
a proper-time interval? It turns out that this cannot be alway'>
done, and the stipulation, under which this becomes possible, will
be found in § 3.4.
Now let us introduce the concept of the object's proper time.
Let an object move uniformly and rectilinearly relative to th:!
frame K. The frame K' can be fixed to the moving object. The
object is at rest in this frame, so that events happening with this
object or at it are registered by one clock. This clock counts th.!
proper time at the point where the object is located; it can be
said that this clock counts the object's proper time. Eq. (3.9)
shows in this case that the interval between events that happened
with the object or at it, is always less in terms of the objed'os
proper time than the time interval between the same events reg-
istered by the clocks of any IFR relative to which this object
moves. It should not be forgotten here that the proper-time in·
terval is registered by one clock, while the time interval in the
frame relative to which the object moves is registered by at least
-two clocks. This is very important because in interpreting Eq.
_;f~~~v~~~ ~~s ~~d~ ~ii~x~re~!7~" a h~~:~:d~~~~~
confuse th~uation. ln]a!=t.. the clock rates are the same in all
IFRs. What turns ou~to be-j)[~ent ·rs1]1~- rea_~]~gs_qiJl!!i~-~n~_
TerVals betwee~ _e~'enls.¥_it __ ~s_ onfy naturaT,Decause the clocKs
synchronize"d i1f orie-TFR _are dis-synchrQ.niz~_in__anather.
~ The proper rime can also be introduced for a particle movint:::
with acceleration. To do this, let us consider the motion of a par-
ticle during an infinitesimal time interval. Let the particle velocity
at a given moment be equal to V. Consider now the inertial frame
of reference K* moving at the velocity V. In this reference frame
the equation d"t"=--Jf'="'Wdt is valid. This equation is also approx-
imately valid for the instantaneously co-moving particle of the
frame K'. The frame !(• differs from the instantaneously co-mov·
ing frame K' in that the latter, K', moves with acceleration, while
the former, K*, does not, although both of them move at the same
velocity at a given moment. The less the time interval dt is, the
more applicable becomes the equation d-r = v--r=-wdt in thl:'
frame K'. Having integrated this equation, we obtain the precise
expression in terms of -r, which is in essence the overall "proper''
time of any coordinate system K*.
Proceeding from these considerations we shall suppose that if
the interval between events that happened to the object turns out
Consequences of Lorentz Transformalion. K Calculus 89
to be equal to d-r when measured by the clock fixed to the ref-

erence frame co-moving with the object, the time interval between
the same events dt, measured by the clock of another IFR rela-
tive to which the object moves, will be, according to Eq. (3.9),
dt=yd"t,
where
(3.15)
now v = v(t) is the velocity of the object (and not of the ref-
erence frame). Owing to this circumstance the designations jl
and y have been introduced. When the velocity of the object varie.;
according to the equation v = v(t) the relationship between th?
final time interval "t and the time interval registered by the clock
of the frame relative to which the object moves is obtained by in-
tegration
' ----
./ ('('))'
"t--r0 =
..1 V I - --;- dt. (3.16)
What is the quantity that is seen on the left-hand side of Eq.

(3.16)? Of course, it can be called the object's proper time. But
how to measure it? Strictly speaking, Eq. (3.9) is valid for the
clocks in inertial frames of reference. But if the clock is fixed
to an arbitrarily moving object, it will undergo an acceleration.
No doubt, an acceleration affects the clock rate in a varying
degree depending on the design of the clock. (If you do not believe
in this, drop your clock on the floor.) Consequently, one can hardly
speak of time readings made by means of such a clock. The rea-
sonable interpretation of Eq. (3.16) lies in the fact that -r- To is
the overall time measured in many inertial frames co-moving
with the object or, which is the s~me, the time registered by the
clock fixed rigidly to the object and not affected aL.gJL12.!LM-.
acceleration_gj.Jb.e...JJ.bj.ecL- ------
~-rr-sliOiiRI be stressed that the difference in readings of clocks
from different inertial frames of reference which we obtained has
no relation whatever to any irregularity of the clock rate in on!:'
or another frame. As in the case of a measurement of a ruler's
length, we deal here with different methods of time measurement.
~ rates of all ~!ocks in all reference frames are absolutely the
same~ ~urementS or-lTmeTiifefvaTsbetween t\VO evefitS
-perfornled. by fhe two sets of clocks ·from different referenc-e~
~mes,-~~q@e(f"' ~llfi'1__n Uleir ..!!Sj)eCfl~e_1r_ar1ie~]eaQ ~~9~~
res~iT~QIJWnecf: a proper.-tJiil]- i11terval--oetween two events t!:l!_l}.§
oUtfobealwaystheleast. ------ --- ------
90 Special Theory of Relcrflollg
Let us consider an example, which is quite analogous to that

discussed in connection with the variation of the scale's length,
showing that deceleration of time is caused by different methods
of its comparison. Let us take two quite identical clocks: A in the
frame K and A' in the frame K'. These can be atoms of the same
kind. Suppose we observe the clock A' of the frame K', i.e. com~
pare the clock A with the set of clocks synchronized with the
dock A'. Then observers from K' will discover that the clock A
goes slower than the set of clocks from K'. On the contrary, ob-
servers from the frame K viewing one clock A' of the frame K'
will discover that it goes slower than the set of clocks from K.
Are these results contradictory? No. We clearly see that the meth-
-ods or clock comparison in the first and the second cases are
different. The clock wl"!!ch is _f!)~~cLwiih different clocksJ'=Q!!J
another reference f This amaZ1ng-srruation
proves o be i11_e~ The equivalence of ~!1_1£__~-~~
_theory plL_e_l~ti_vjj:y, ~o ~f1Wre]aJ1~vaTU.eS eroe_~g_e, th~ em~e
in the same manner in all IFRs. _ - - -
_ § 3.4. The classification of intervals and the principle of causal-
ity. It is seen from Eqs. (3.1) and (3.2) that when two arbitrary
events are considered, both the distance and the time interval
between them prove to be relative values: (1-x' =1= l!u, l!J.t' =1= M).
Unbl now we dealt with the events of the special type: in length
measurements the coordinates or a ruler's ends x2 and x 1 were
.considered simultaneously (/ 1 = t 2); when a time interval was
determined, the moments of time tl and t2 were considered at the
same point x! = xl. But even in those cases the spatial and tem-
poral "distances" between events turned out to be relative. No
wonder they do also in a general case. In addition to that, it
iollows directly from the Einstein postulates that the interval
between events is, as we know, the invariant of the Lorentz trans-
formation:
~-#~ ~ ~ ~ ~ ~ ~ ~-
- vc' (MJ' (~x)' (~y)' (~z) 1 . (3.17)
The designations are the same as were used in the derivation of

Eqs. (3.1) and (3.2). It is convenient to introduce also the special
designations for the spatial and temporal distances between
events:
1f,-(6x)'+(6y)'+(6z)', M-t". (3.18)
Having written out the squared interval between two events

in the frame K as si2 = cii2 -li2 and in the frame K' as s:~=
= c2t~~ -t;~. we obtain the condition for the interval inv~riance

sa=si2 as
Considering the events in an arbitrary reference frame K, we

shall most likely discover that they happened at different points
in space and at different moments of time.
Is it possible through the choice of the reference frame K' to
ensure that (a) ~nts.J..AnJi lLb.§_p_p_g_n_at t)le same pQl~! !n_ spa~e.
i.e. be single-positioned; (b) events I and II happen at tfle same
moment of time; and, finally, (c) events I and II happen at one
point in space and at the same moment of time? Let us begin
f1om the beginning.
(a) Is it possible to choose such a system K', in which these
events will happen at the same point in space, i.e. will be single-
positioned? This means that the following condition should be
met: li2=0. But then it follows from Eq. (3.19) that
(3.20)
i. e. si2~ 0, and the interval s12 must be real. In the frame K'
the events considered happen at one point in space and the time
interval between them is equal (with an accuracy to within the
factor c) to
(3.21)
That is why_ the real intervals between events are referred to as

time-like interva15."Tbe cond,fion for a time-like i~~ also
be wrillen-m::tfWJorm l,? < ct,?.
1 e t us consider the motion of a particle possessing a rest mass.
Conventional mechanics deals with objects of only this kind.
Suppose for the sal<e of simplicity that this particle moves uni·
formly along the x axis, covering the distance 6x in the time .lt.
In the frame K' this particle will travel the distance ~x' in the
time 6t' which is determined in accordance with Eqs. (3.1) and
(3.2). The ratio ¥,
= v is the velocity of the particle in the
frame K. Taking this into account, we can rewrite Eqs. (3.1) and
(3.2):
t.x' ~ r(t.x- V '')~I' ( ¥,- V)M ~ rtv- V)At (3.22)
llt' =I' ( llt- fa-6x )=r (1-{r 1f}~t =I' ( 1-7) 6t. (3.23)
Special Theory of Relaliulty
Assuming !lx' = 0 in Eq. (3.22), one can easily find the velocity
of the frame K' in which the two events in question are single-
positioned. From the right-hand side of the equation we immedi-
ate!) obtain the evident answer V = u. So, this is just the frame
co-moving with the particle. Another important consequence fol-
lows from Eq. (3.23). Let M = /2- t1 > OJhis.___impfu_Ltl!!_t
event_II happ_ened_af_ter event I. Is there such a fr_ame K' in which
-Slr< .0. i.e. th~ tirn_e ~iqUeilce of the events IS"""ii1VerSed i.SCo-nl-
pared to the Trame K? Eq. (3.23) shows the sign of !lt' to coincide
\\tith that of llt when (I - ~) > 0. But this condition is always
satisfied, since the velocity of an object v is always less than c.
(The reference frame is also a material body.) The same condi·
1~~~r~sal~ 1 ~~dse~t~~f~~~ ~'h';r~n~n\a~f 0Jq~v(~-~3) 0~~~~~i~[ th:i:~~~~:
sian (1 -f cA:t ). According to Eq. (3.18) l 12 ~ llx, and if
ct 12 > 112, eM> llx a fortiori. This implies that the ratio llx/c·~t
is less than unity; V/c is always less than unity, so that
{I -f ~) > 0 and, consequently, tl.l' > 0.
Hence, for these two events, considered in terms of the frames K
and K', the conCf_pis-"later_':_l!n~ "earlier" have an i_dentical,_ tQat
is absolute, character. In general, H {he intervaT-Oetween--events
is time-like (recall that the interval is the invariant quantity), the
time sequen~e_ of eyents r_emains the :Safll~_ih_r_ough.auJ a_ll IFR:>._
Tater on we sfia n _see llla1lfi_rs:~Do1Jb.e c;tse for_ intervals WhiCh
differ fronLiimedi.ke. ~_r_v~~-
What is the significance of the invariance of time sequence in
all inrrtial frames of reference? We have already ascertained that
two events separated by a time-like interval will be single-posi·
tioned in some reference frame. If one of them happened "earlier"
and the second "later", the first event may be the cause for the
origination of the second, i.eJ.he,L_!!!O:)_y_Jle_connecte_Q_frr__the c_au~e_
and-effect relationship. But in ITi1s case their time sequenceCai"lnot
depend on [he· choice of a reference frame. It is from our results
that the criterion for the possibility of the cause-and-effect rela·
tionship follows (the interval is time-like). As to the time sequence
of events, it automatically remains the same in all reference
frames.
of ~h~a~i~~~~~~ek~~~r~~l at~~~~~~Pn b~~!~~sn i;\?~~f~e~o!h~nfy b~icb~~~~
1 0
it provides the identical time sequence in all IFRs. It indicates
the physical opportunity of one event affecting another. It follows
from the inequality l'f2 < c2t~2 , determining a time-like interval,
that during the time passing between the two events light can
Consequen.:es of Lorentz Tran#ormafion K Calculus 93
a Jortwri cover the distance from the point where event I occurred
to the point where event II did, the product c(t 2 - t!) being the
path travelled by light during the lime (1 2 - tt). This means that
basically a certain interaction (signal) could propagate from the
point where event I occurred to the point where event II did
during the time interval between the events. Without claiming th~
generality in formulating the problem we shall assume that one
event can affect another only through a physical (force) inter-
action. Then, if event I happened, a "signal" about this fact can
reach the point where event II will happen prior to the moment
of occurrence of event II. This means that event I can be the cause
of event II, and event II can be the effect of event I. In this case
the events can have the cause-and-effect relationship. Thus, the
eve-nts separated by a time-like interval can have the cause-and-
effect relationship in terms of physics as well. It is understood
that they may not be in such a relation. We only point to thl\_
theoretical possibility. What is essential, the _time sequence c-an--)
~~~e~ea ft~~t~t!nc_:~~-~Se:_ol- sUCli_ irilemlls:-ihe--CamequeAce ~
(b) Now let us PaSS over to the consideration of intervals of
the opposite sign. Let us examine again the condition of the in-
terval invariance (Eq. (3.19)) and determine whether we can find
such a coordinate system K' in which the two given events I and
II happen simultaneously. This means that in this system tl2=0.
si
Hence 2 = -t;i < 0. The squared interval between the events
must be negative, and the interval proves to be imaginary. In the
frame K' the events in question happen at the same moment of
time, and the interval between them is reduced (with an accuracy
to within the number i) to the spatial intervall] 2 = is 12. That is
why imaginary intervals are referre_!)_t_g_as_ spa~.e-l.i_kln!etva-!5.
The condition -ror a-Spac-e-ltke ·intefval can also be written in the
form lt 2 > cl12·
Can one find the reference frame in which 6.t' = 0 for two given
events? Assuming M' = 0, we obtain from Eq. (3.2):
(3.24)
Since one can always choose the events so that l 12= 11x, it fol-
lows from the condition for a space-like interval that in thi~ case
~x > c-~t. Eq. (3.24) testifies that we can get V < c, that is,
basically, such a frame can be chosen. The ratio l!!x/c-M appears
in the third link of Eq. (3.23); as we have mentioned. it can ex-
ceed unity. But this means that the factor (I - f ,I:J.;t) can be
made negative by the appropriate choice of V.
It follows from here that the time sequence of two events

related by a space-like interval can be reversed on transition from
one IFR to another. This does not apply to the events which
could have the cause-and-effect relationship. But in terms of
physics they just cannot be so related. Indeed, the conditioll
tr
l'f2 > c2 2 signifies that no "signal" can be transmitted from the
point where event I happens to the Point where event I I docs
during the time interval between these events. Consequently,
events separated by a space-like j~r_v.;j_l cannot be in a ca_!!s~
a-riO-elleg~--- ------------ -
ThUS, the s ecial theor of relativity ~ke.!.JLQ.C??_~ible to in-
~t~. .L Wllditions under w ....caus~~t ieli!IOJ:ishtp:
~~~~~b~~in~~sii~1 eth~~=n~~af f~~Z f~~~o~~~~~
premises. It should be emphasized once more, of course, that all
of our reasonings are based....Q!1__t!Je premise of the finite ~city
of sjgn.al t&ansmission..___ --
(c) If we are interested in the reference frame in which the
events would be both simultaneous and single-positioned, i.e. the
two conditions 1]2 =0 and ll 2 =0 would be satisfied, the two in-
equalities s 12 ;a= 0 and 112 ~ 0 would have to be simultaneously
complied with. This is possible only when s 12 = 0 and s]2 = 0. If
the events in question do not represent the sending and reception
of light signals, the intervals can be equal to zero only in the
case when the two events coincide in each of the frames K and K'.
Of course, the coincidence of events does not depend on the choice
of a reference frame.
The interval between two even~s that happened in a given frame
at different points in space and at different moments of time and
whose absolute value is equal to zero, is relevantly referred to as
a light-like interval. A light-like jp_k!:val links to&!!!!er events
:=rons1stmg m ~-.!lg_~!-~ng_£o~UJi_yeJY-=-v_a!iQUS_ RQillf~
it"!_ space. We made sure of this at the beginning of § 2.6.
§ 3.5. The transformation of velocity components of a particle
on transition from one inertial frame of reference to another. From
the Galilean transformation for coordinates and time (Eq. ( 1.2))
we obtained Eq. (1.4) showing how the particle velocity is trans-
formed on transition from one IFR to another: v' = v- V. This
transformation rule does not satisfy the second postulate of Ein-
stein, since the velocity of light in vacuo turns out to be different
in different reference frames. The velocity transformation equa-
tions following from the Lorentz transformation satisfy the Ein-
stein postulates. Now we proceed to the derivation of these equa-
tions.
Consequences of Lorentz Tran.~formallon K Calculus 95
Let us consider the particle motion in terms of the two IFRs:

K and K'. The velocities are determined as usual:
In the frame K In the frame K'
If If
x=x(l), y=y(l), z=z(l), x'=x'(t'), y'=y'(t'), z'=z'(t'),
then then
V~=%, V11 =!{/t, v:= ~~:, V~= :r, V~= :;: .
d•
Vz=(fl·
The relationship between x, t and x', t' is meant to be estab-

lished via the Lorentz transformation, so that t, for example, can
be adopted as an independent variable. When t varies by dt, all
variables get increments; differentiating the Lorentz transforma-
tion (see Eq. (2.16)), we obtain these increments in terms of
differentials:
dx=f(dx'+Vdt'), dy=dy', dz=dz',
dt= r ( dl' + fdx'). (3.25)
Having divided termwise the first three equations of Eq. (3.25)

by the last one, we get
dx dx' + V dt' dy dy'
dt= dt'+fdx'' dt- r(dt'+fdx')'
d•
dt
r (dr +fdx') ·
Dividing the numerator and denominator of the right-hand sides
of these equations by dt', we finally obtain
' . 1---v>
v11 'V 1--;r
'./-V'
Vz'Vl--;r
V11 = V , Vz (3.26)
!+fv~
1
l+creo.o:
These are the formulae of the relativistic transformation of velo-
cities. Using them, the components v.c. v11 , Vz in the frame K can
be found from the velocity components v~. v~. v~ in 'the frame K'
and the velocity of the frame K relative to K' which is equal
to - V. In order to get the formulae for the inverse transforma-
tion, it is necessary to change the sign of the velocity V, to prime
the unprimed quantities and to withdraw primes from the primed

ones:
, v_.-v
(3.27)
v..,= 1-B-¥'
Of course, the same result will be obtained from Eq. (3.26)

directly. It is seen from Eqs. (3.26) and (3.27) that a uniform
motion in one IFR will be Wliform in all other JFRs. Hence, the
-JJA-i-iomCan-d rectilinear motion is distinguished from all othi!r
kinds of motion. On the contrary, according to relativistic kine-
mati~s_ t~~ unifor~ly accelerated motion in a certain IFR may
-n-ot tie that in oth-er)F_Rs_($ee § 5~1).
In Eqs. (3.26) and (3.27) the x axis is distinguished from the
y and z ax~s. This is only because the relative velocity of the
reference frames K and K' is directed along the x axis. Passing
to the limit B-+ 0, i.e. assuming formally c-oo in these equa·
tions, we get back to the Galilean transformation (Eq. (1.4)).
)'his_ me~_!IS that the Galilean transformation is sufficiently ac·
-%~~~r!~-£1t3~Fftifi{~ri~d_g~e~~:-~~~n~ot~~~ :~ ~~7~~-
tivistic concepts in our everyday life. Here we have learned again
that the difference between relativistic and classical concepts is
~~~ \~.~~) ~~~t~~~~~l~~ ~heeri~:~o~!In~f alifo~t~-~~:n~~~~a~~pp(?~!~/1
to the theory of relativity.
Let us consider the motion along the x' axis. In this case the
velocity components in the reference frame K' will be v~ = v',
L~ = 0, v~ = 0. From Eq. (3.26) we see that in the frame K the
components Vy and Vz are equal to zero. Consequently, the motiou
in the frame K also takes place along the x axis and Vx = ~.
That is why according to Eq. (3.26)
(3.28)
Setting v' = c, we get v = c from Eq. (3.28). This corresponds

to the second postulate of Einstein: the velocity of light in vacuo
is the same in all IFRs.
Note here that the substitution v' = c in Eq. (3.28) is not
quite consistent, since material particles representing a "signal"
cannot_ move aJ_ t~~-'::~~ocity_.f, the_equatioR bei_~g _4e.r:iyed I_!?_r ma-
terial parl.JcleS (m =F OJ. However, one may assume v' = c in
Eq. (3.28) considering light quanta (photons) as relativistic par·
tides (§ 7.6). Besides, there are ultra-relativistic particles whose
velocity is close to that of light. For example, the velocity of elec-
trans in the electron accelerator in Erevan differs from c in the

ei~~h s~1lud~~~a~i~~ir~~ explanation of the results of the Fizeau
experiment as another example of utilizing Eq. (3.28) for tho.!
velocity summation. In this experiment the velocity of light pro-
pagating in water motionless relative to an observer (laboratory)
was compared with that in water moving at the velocity V. The
fig. 3.7. The schematic drawing of the Fizeau experiment. A ~mi-transparent

plate PP splits a light beam from a source into two light 11encils, one going
:i~~,n a~~v~~~t~~d~!,a~~~lc:;:t~a:h~ ~~~dci~;e~'fe~i·g;r~ !u~je~ 1i6hlo~rtlfea~b:
fi~~ri:e~~:r:nl~~w;~ct~r b~l~~ ~~~=r t~t v~~eh;r~~~~y si~'n V:.i!t., t~:rr::~~~:rs r~
light propagation along the water How, and the sign "-" to light propagation
against the water How: the coefficient k is called the drag coefficient. This coef-
1kient was sought by Fizeau who conducted the described t:xperiment. It lol-
lowt'd from the Fizeau experiment that k = I - i/n~ The theory of rela1Lv1ly
explains quite naturally Fizeau's resu!l (see the text). The details of the expert-
men! can be found in books (13] and (15].
velocity of light in motionless water is equal to c/n, where n i-;

the refraction index of water. The light pencil travelled through
l~: l~b~:~for~af~e;m~F~f·r:i:J~n~~~r~~ :~~~~itt!rf:r~~c~e~r~:~el1g~~
pencils going one along and another against the water How. It
followed from the results obtained in the Fizeau experiment that
the phase velocity of light in motionless water should be in-
creased by the velocity of water V multiplied by ( I - 1/n 2). Thu.,;,
if the phase velocity of light in motionless water is u' = c/n, the
phase velocity, found in the laboratory frame, turns out to be
(3.29)
98 Special Theory of RelalirJily
Using Eq. (3.28), we conclude that in the laboratory frame (due to

the velocity summation law)
v=~= -;-+v (f+v)(!--f,;-)

l+*tf l+fn. 1-Cfn.Y
Neglecting the quantity (V/cn)2 in the denominator whose small-
ness is due to the non-relativistic velocity of water ((V/c) 2 «:I),
we obtain
v~ f+ V (1 -~) -£=+[! +f(l -~ )n -;]. (3.30)

In Eq. (3.30) we again neglect the term V2 /c 2 and get the Fizeau
result (Eq. (3.29)). Thus the Einstein equation for the velocity
transformation provides a natural interpretation of the results
of the Fizeau experiment (see§ 1.7).
We have already pointed out that the most important assump-
tion of contemporary physics is the statement concerning im-
possibility of transmitting signals (interactions) at the velocity
exceeding that of light. No doubt, a moving object can be used !o
transmit a signal (energy, momentum), and, consequently, t~~
velocity of an object cannot exceed c. Relativistic mechanics in-
fers that the velocity of a material object, i.e. an ol>ject possessing
a_ rest •'T:t~s.-ls __@l~v_itys less than c and never reaches this value.
BUflhts is valid in a· "definite IFR:_ls_____il:__ p_a_s.s1bJ~ ,t_q_ choos-e suc11
an lERln_:w_b_i_cllJ!luelocity of an object will exce~d_t;F- - - -
1-lad it followed fro-m cl~fsSicat mecharTICSfllit in a given IFR
the velocity of an object never exceeds c, one would have ob-
tained the velocity of an object exceeding c via a choice of a suital>le
reference frame. Indeed, according to Eq. (1.4) v = r/ + V, where
v' is the velocity of the object relative to the frame K' and V is
the relative velocity of the frames K and K'. If the velocities L''
and V exceed 0.5 c, the velocity v of the object in the frame K
will be greater than c.
But in the STR the velocity transformation is carried out
differently. It is seen from Eqs. (3.26) and (3.27) that the \'e-
locities of a particle and of a frame do not add up as vectors do.
MoreO\·er, the velocity summation in the STR obeys the incredible
+
rule c c =c.
It follows from Eqs. (3.26) and (3.27) that if the velocity of a
p<~rliclc is less than c in the frame K (vfc < 1), and the velocity
of the frame K' relative to K is also less than c (VIc< 1), the
velocity of this particle determined in the frame K' is always less
than c. The simplest demonstration of this statement can be car-
ried out for the case of a unidimensional motion by means of Eq.
Consequences of Lorentz Transformation K Calculus 99
(3.28). Having composed the expression (vjc- I), we write out

the following chain of equations:
+-1~ ~ 4~~
,.
-1
-f+~-1-~
I+ vc~,
f(•-f)-(•-f) _,__('_-_,__-¥-)'--'(,.'-___,-7'-'--'
)
- v' < 0. (3.31)
I+~~~ I++
Whence it is clear that v < c.
But is it possible to get the relative velocity of reference frames
exceeding c by means of consecutive transitions from one frame
to another? Strictly speaking, a reference frame is a system of
material objects, so that in order to answer the question we can
make use of the theorem just formulated. Certainly, in the STR
one cannot obtain the relative velocity of frames exceeding c in
any case. But now we shall derive this result once again by
another method, which is instructive by itself.
Let us introduce, aside from the frame K, two more frames, K'
and K". What is the relative velocity of the frames K and K" if,
on the one hand, the relative velocity of the frames K' and K".
and, on the other hand, that of the frames K and K' are known?
Let the relative velocity of K and K' be equal to V and that of K'
and K" be equal to W. Introducing the designations B1 = V/c
and B2 = Wjc and, correspondingly, l/f 1 = -y'I- a:.
l/f2 =
=-v't- Bt we get
x~rdx'+Vt'), t~r,(t'+~x'). (3.32)
x' ~r,(x" + Wt"), t' ~ r,(t" +-'l'-x"). (3.33)
Substituting Eq. (3.33) into Eq. (3.32), we find the explicit rela-
tionship between coordinates and time in the frames K and K"
x=f 1r2(x" + Wt"+ Vt"+B1B2x")=
~r,r,((1 +B,B,)x" +(V + W)t"}~
=f1f 2(1 +B1D2)(x"+ t++B~~, t"). (3.34)
In much the same way one can obtain
1 ~"t~, x").
..
t=f1f2(1+B1B2)(t"+f (3.35)
100 Special Theory of RelaUvitg
Designating
~~~~~ = t!;:;c' =U, (3.36)
we shall calculate the first multiplier in Eqs. (3.34) and (3.35):

r,r,o +D,IJ,J-
-V 1--
u•.
,.
Then from Eqs. (3.34) and (3.35) one can get
xN+UI" t"+.fxn
X=------=- f = ----.
-VI-*' ~
Consequently, the_JwQ.__~!J.~ce~§:ive .~QLeni:t... .transformations Q.f
the r~~~eS-lLand ~olthe reference f~ames are egui-
~eri_L1o.._onetransformatiQ!!_ of the f-~w~-Yeloc1J.£Rife.t~~riii.Qea
according to Fq (3 36}-:-In allier worOS. the relative velocities of
~~f:~e~J;e~~~ma~!~ 4th~f s~~~ aac;~~~~~Pat-}~~-tll ~~~8 ~~o~~~ew:
ye]ocity greater than that of light.
Eq. (3.36) can readily be obtained by means of a complex ro-
tation (see § 2.8). In geometrical terms the transition from l(
to K' and then from K' to K" constitutes a consecutive rotation
in the plane (x, 't) through the angles fll• and fll 2 with tan fll• = iB 1
and tan q:2 = iD2.
The tangent of the resulting angle can be found according to
the conventional formula for the tangent of the sum of two angles
(~-~·+op,):
tanqJ 1 +tan ql2
tan fll I tan ql1 ·tan ql1 '
or
iD = iln+"t:·~; .
which is just Eq. (3.36) with B, Bt and B2 being replaced by their
respective values. The two last expressions show that the set of
the Lorentz transformations possesses the basic property of the
gf,_OJJ.P- (from the standpoint of the ma[hemat.cal group theory).
_two Lo~enlz transforma_t_iC!ns _a_gain _pro~~~e ~ L_o!et!_!~ !rit~!Qiffia:
Consequenees of Lorentz Transformation. K Calculus 101
tion. It is essential here, however, that the relative velocity is

-alWaYs directed along the x axis.
Here is the useful interpretation of Eq. (3.28). In § 2.7 we in-
troduced the parameter a associated with the relative velocity ol
reference frames by the relationship B =-tanh a. We can also
introduce the velocity parameter for a particle p =-tanh e. Then
Eq. (3.28) will take the following form:
f}==tanh9= ~~~~~~=- 1 1~n:a:;1 !,~ 81~~~ 0 -tanh(6'+a);(3.37)
the last link of the equation is written in accordance with the

equations of Appendix I, § 9. This is an interesting result. In the,
cl..a.ssit.aL.!h~ it is velacifies that a~ 1.2..4))-:-wnrr-e
in the rel~tivistic.the.ocy._Yelocity oarameters do. Therasr CoiF
tinge11Cy will be put to use in§ 5.7.
§ 3.6. The transformation of an absolute value and the direction
of the velocity of a particle. From Eq. (3.26) for the velocity com-
ponent transformation one can obtain expressions determining
the absolute value and direction of the velocity in the frame K,
provided the velocity components in the frame K' are know11.
First of all, it is evident that if the vei.Qfi_ty~_component v~ = .0 in
the _frame__ K'~~the. component Vz ==: 0 in the: frame I( t9o. This
meanS tb3.t if the ffiOtion in the -frame ]('Takes plaCe in the plane
(x', y~...__thc motion in the frame K will also· Hikif place· in the
~a~-\~~~fii h:!- r~ _f~;-~~~t\~:. ap}*~ th!sr~~~~ak,~h~:~oi~iti~

clear that if 9' is the angle between the direction of the velocity
u' and the x' axis then v~ = v' cos a', v~ = v' sin a'. We shall de-
note the angle between the direction of the velocity v and the x
axis in the frame K by a. Consequently, V:c = v cos a. Vg = v sin a.
Let us find the equations relating v and a with v' and a'. Having
expressed the components v~ and v~ in terms of v' and a', we
can rewrite the first iwo formulae of Eq. (3.26) in the following
form:
vcosa= v'co~O'~,v , vsina v' sine: -v'f~ . (3.38)

I+~D l+vc;se 0
Having divided the second formula by the first one, we obtain

the expression for tan 9:
(3.39)
In order to find the expression for the absolute value of the ve-
locity, it is sufficient to square and sum term wise Eq. (3.38); we
shall obtain at once
It follows from Eqs. (3.39) and (3AO) that the angle between
the velocity direction and the corresponding x axis, as well as the-
absolute value of the velocity, change on transition from the frame
K' to the frame K (Fig. 3.8a). (Recall that the geometric axes x
and x' coincide.) Of course, the same occurs in classical me·
chanics as well, although it is described by other equations.
Now let us derive a useful formula resulting from Eq. (3.40).
We shall need it when studying Chapter 7. Using the first equa-
tion of (3.40). we compose the following expression:
fJ 2 c2 {I + fJ~~ cos 9') 2

- v' 2 - V2- 2v'V cos e' + v;~· 2 sin2 9'
1--;:r= v'V 2
c2(1+C2cos9')
c2-v'2+~v'2-V 2 (1-~)(1-;-)
c2 (1+ 0;
v cosa'r
2 (1+ v;~ cosf' )'
Consequently,
(3.41)
(3.42)
From Eq. (3.41) it is also easy to obtain a convenient equation

for the square of the absolute value of the velocity:
(!-.:;;:)(!-;)
(3.43)
from which it immediately follows that if v'/c and VIc are less
than unity, then v < c. For the special case, when the velocity
was determined according to Eq. (3.28), the same theorem was
demonstrated above. Eq. (3.43) can be obtained, of course, by
squaring and summing up the left-hand and right-hand sides of
Eq. (3.26).
Consequences of Lorentz Transformation. K. Calculus 103
From Eqs. (3.38) and (3.39) one can readily obtain the equa·
1ions determining the variation of the direction of light rays on
transition [rom the frame K' to the frame K. In this case, having
.assumed v' = c in Eq. (3.43), we obtain, as it should be expected,
K B
tana.•:;r,:=='jfz
x,:r'
Ia) (b)
Fig. 3.8. (a) A particle moves in the plane (.r', y') in the frame K.'. The incli·
nation angle of its velocity to the .r' axis is equal to 0', with tan 0' = 7J'vlu'~·
In the frame K. the components fJ~ and u~ vary according to Eq {326). whence it
is clear that the angle a is not longer equal to 0' (see also Eq (339)). For
1he case shown in the diagram, a' > a. (b) Light propagates along the y' axis
in the frame K', i. e. along the perpendicular to the frame motion direction. Ob-
viously, u'~ = 0, u'v = c, a• = tt/2. In the frame K according to Eq (3 26),
u~ = V, t>v = c..,f! -D 2, '~hence tan 0 = ,Yr=R2/11. The aberration angle is
formed by the visible direction of incoming light in the frame K and the dircc·
1ion of hght in the frame K.', i e. the y axis. The aberration angle cz; = tt/2- 0.
v = v' = c. Taking this into account, we obtain from Eqs. (3.39)

.and (3.38) respectively
tan e = n..;~ :s~2, sin 9', (3.44)
sin9= 1 ~a,sin9', cose=t:~c!s~·· (3.45)

Eqs. (3.45) describe the light aberration consisting in the wave
front of a light wave changing its direction on transition from
one IFR to another. Let light propagate in the frame K' along
the perpendicular to the frame motion direction, for example,
along they' axis (Fig. 3.8b); this means that 9' = n/2. Then, ac~
cording to Eq. (3.44) tan e = ~ .
The aberration angle is constituted by the visible directions of
light in the two IFRs. In the frame K' light propagates along the
104 Special Theory of Relaliuify
y' axis, while m the frame K at the angle a= n/2- 9 to the y

axis. Obviously, the angle a is the aberration angle, and
tana=tan(2}-9)=cot9= .y~·
The calculation of the aberration angle a according to the Ga-
~e~." i~~~s~~~~;;i~~a1 d1h~herecl~\~~:::ii~n r6'~~~~:1f~ilfeirvse 5fr1~~ (\~
non-relativistic one by the term of the order of 8 3•
One can also readily obtain the equations for the aberration
angle in the case when light falls at an arbitrary angle to the
motion direction. We shall assume that 8 <t: I. Then, according
to Eq. (3.45),
sin9=(1--}B2 + ... )(1-Bcose'+ ... )sinO'.

Rejecting all the terms starting from 8 2 and higher, we obtain
sinO- sin9'=-Bsinflcos9'.
The angle 9'- 9 = 1.9 is the aberration angle. Since the right
part, proportional to B, is small, ~0 is small as well:
sin 0- sin 0'= 2cos e~e' sin 9 -; a·~- cosO' ~0.
Consequently,
l\0= B sinO'. (3.46)
This elementary equation describes the aberration of light falling
at the angle 0' in the frame K'. Fig. 3.8 illustrates the change of
the direction of particle velocity on transition from the frame K'
to K. as well as the calculation of an aberration angle in the case
of a perpendicular (relative to the motion) incidence of light. Sec-
Supplement II about the role that the aberration played in the de-
velopment of the STR.
Finally, let us calculate the relative velocity of two particles.
It is natural to define the relative velocity of two particles as tlv~
velocity of one of them in the frame K in which another particle
is at rest. Let the velocities of the particles in the frame K' L•e
vi and vl. Choose the coordinate system K such that V = - vf.
The particle velocities are immediately determined from Eq. (3.40).
The absolute value of the velocity v 2 is equal to zero, while th3t
of the first particle is
(v~- v~)2- ~[v~v~J
vi= ( vlv2 , ~)' (3.47)
I+ ""C2
Consequences of Lore11lz Transformation. I( Calculus 105
Thts expression defines the square of the relative velocity of th~

two particles. Eq. (3.47) is symmetric relative to v 1 and v 2•
It was shown (see Eq. (3.43)) that v 2 is always less than c2
in Eq. (3.40). This. Js also the case ~QC.Eq~ 13.47): the- relative-
yelocity of r_ar1iC!eS cannot e.xceed the velocity of light in vacuo.
§~7. The K calculus (the radar method). We shall present
below an elegant method of deriving the basic consequences of
the Einstein postulates. This method could be briefly presented on
the basis of the results obtained earlier by other methods. How-
ever, we intend to reiterate some conclusions in order to make
this section more or less self-consistent. The method is remarkable
because it dispenses with the coordinate routine, even in the
derivation of the Lorentz transformation. Although graphical illu-
strations used below involve coordinates, they have an auxiliary
character: these coordinates are not indispensable for the pre-
sentation of the method but are useful for those who are familiar
with space-time diagrams.
Only one spatial coordinate is to be considered. Many charac-
teristics of the STR are revealed even in this case permitting of
<lescriptive illustrations. Thus, let all events occur on the x axis
(and respectively on the coincident x' axis of the frame K', see
Fig. 1.2). In the K calculus all conclusions are drawn from the
imaginary experiments consisting primarily in the exchange of
light signals in vacuo; their sending, reflection and reception aN
examined. In the final analysis, such a play with light spots makes
tl possible to obtain the basic consequences of the Einstein postu-
lates.
The principal assumption which is made in the K calculus is
based on the Doppler effect (see § 3.3); in a uni<limensional case
it is always the radial Doppler effect. Thus, if a stationary radar
located in the frame K emits short pulses periodically with time
intervals (perio<ls) T, an observer in the frame K' moving away
from this radar at constant velocity will <liscover that the in-
terval between the incoming light pulses is different, despite the
fact that the rates of the clock fixe<l to the radar and that of the
observer from the frame K' are identical.
For the sake of simplicity we shaH speak not of the radar and
receiver, but of the two observers A and A' at rest in the reference
frames K and K' respectively. Thus, if the observer A sends light
signals separated by the time interval T accor<ling to his clock,
the observer A' will receive these signals separated by a <lifferent
interval as measured by his own clock. Let us designate this in·
1crval by KT. That is how the coefficient K appears, the key quan·
ltiy of the considere<l method.
It shoul<l be pointe<l out that T and KT are the time intervals
between the sending of the first an<l of the secon<l signals by the
observer A and the reception of these signals by the observer A'.

measured in each case by the clocks at rest in the frames K and
K' respectively.
Proceeding from the principal properties of space and time.
their uniformity and isotropy, one can assume that the coefficient
K depends neither on the positions of the receiver and the source,
nor on the time of sending and receiving the signal, nor on the
direction in which the signal is sent' (in other words, the dire,:-
tion of the common x, x' axis may be chosen arbitrarily in space}.
Certainly, this coefficient does not depend on the time interval
between the sendings of the signals. It may depend only on the
relative velocity of the observers A and A'. Indeed, as the expe-
rience shows, the variation of the light frequency due to the
Doppler effect depends only on the velocity of the relative motion.
The reason for the appearance of the coefficient K is evidenL
Let the observer A located at the origin of the reference frame K
send light signals to the observer A' located at the origin of the
frame K'. The frame K' moves away from the frame K to the
right. Let the first signal be sent at the moment of time t. Then
it is easy to determine the moment 'tt by the clock of the observer
A, when the observer A' receives this signal. Indeed, the signal
propagating at the velocity c has to travel during the time "t"! the
distance Vt which separated the observers A and A' at the mo-
ment t and the distance V"t" 1 which will be covered by the observer
A' during the time '1" 1: et 1 = Vt + V"t" 1, whence it follows that
'~"I= c ~ V t. The second signal is emitted at the moment t T +

and it reaches A' in the time '1"2 determined from the equation
cr2 = V(t+T)+ V'l" 2• Consequently, -r2 = vdt_+yT). The differ·
ence T2 - -r1 = c ~ V T gives the time interval between the signals
rel·eived by the observer A'. However, we have not yet found the
expression for the coefficient K, although it may seem so. The
coefficient K will be obtained as soon as we find the time interval
bdween the incoming signals registered by the clock of the ob-
,<;erver A'. But we have not determined the relationship between
the readings of the clocks A and A' so far.
Until now we put to use only the uniformity and isotropy of
time and space. Now we shall make use of the constancy of the
velocity of light in vacuo in all IFRs. We shall have to use this
p1operty of light very often. This condition can be formulated as
follows: "light cannot O\·ertake light." Now we pass over to the
problem \Vhich employs explicitly the equivalence of all inertial
observers, i.e. the first postulate of Einstein.
We have agreed that the signals sent by the observer A at the
intervals Twill be received by the observer A' at the intervals KT
Consequences of Lorentz TransformaUon K Calculus 107
.as measured by his clock. Due to the equivalence of the observers

we have to suppose that the signals sent by the observer A' at the
intervals T will be received by the observer A at the intervals KT
.as well. (The principle of relativity for two inertial observers A
and A'.)
It is worth mentioning that this assumption is strongly bas<>c:l
-on the fact that vacuum contains no medium in which light pro-
pagates. Had such medium existed, the coefficient K would have
depended on the velocities of the observers A and A' relative to
this medium. It was just such a medium (ether) that agitated the
minds of the 19th century physicists most of all. It caused a series
-of dramatic situations preceding the advent of the STR (see Sup-
plement II). At present it is quite reasonable to adopt the contem-
porary point of view.
Now we shall find the explicit expression of the coefficient K in
terms of a relative motion. In this procedure we shall need noth-
ing except a few imaginary experiments pertaining to the send-
ing, reflection and reception of light signals. The reflection can
be treated, if necessary, as the sending of the signals by the "ob-
-server" in the reverse direction at the moment when he receives
the incoming signal.
Let the first signal from the observer A to the observer A' be
-sent at the moment when the frames K and K' coincide. The ob-
-servers A and A' located at the origins of their respective frames
are positioned at this moment at the same point in space. Na-
turally, the transmission of this signal from A to A' and of the
reverse signal from A' to A does not require any time. After the
time interval T by his clock the observer A sends a light signal
to the observer A' who will receive it in the time interval KT after
the reception of the first signal. Let the observer A' send a signal
back to A immediately on the reception of the second signal (th~
-same as the mirror reflection). The two signals are separated by
the time interval KT by the clock of the observer A'. Hence, the
return signal will be sent from A' to A after this time interval.
But the observer A will not receive it after the time interval KT.
This time the interval will be increased K times again and will
be equal to K2 T. Consequently, the return signal will be received
at the moment K2T by the clock of the observer A. Hence, in terms
-of the observer A the total travel of the second signal sent at the
moment T to the obsen•er A' and back takes the time K 2T- T =
= (K 2 - I) T. Since the velocity of light is the same whether it
propagates in the direct or the opposite direction, the propagation
time from A to A' (or back) is equal to 1f2(K2- I) T. From this
it iollo,vs that the determination of the distance between A and
A' at the moment of reflection by means of a radar will give the
value 1/ 2 (K2 - I ) Tc.
Thus, we have found the distance between the observers A

and A' at the moment when the signal is reflected. But at what
moment by the clock of A did the reflection occur? Note that we-
speak of the clock located at A, while the event that we consider.
i.e. the reflection of the signal at A'. is removed from A. In this
case we cannot measure the time of the event directly but have
to ascribe a definite moment of time to it.
The second light signal was sent' at the moment T and was
received back at the moment K2 T. Hence, the moment of reflection
is determined as 1/2(T +
K2 T) = 1/2(K 2 +I) r. Consequently, dur-
ing the time interval 1/2(K 2 +I) T the observer A' moves away
from lhe observer A by the distance 1/ 2 (K 2 - I) Tc. So, the rela-
tive velocity of the observer A' is
I
V- 2(/(i-l)Tc v 1(2 -1
c= K'+,. (3.48)
- ].<K 2 +l)T'
It therefore follows • that
(3.49)
Here we shall write the two equations which we shall need later:
K:; I = ~ ::: r, K:; I =rD. (3.50)
It is very convenient to make use of a graphical diagram in

order to present descriptively the results obtained. Let us intra·
duce the Cartesian coordinate system on the plane with coordinate
axes x and ct. Later we shall see that the choice of the spatial
and time coordinates of the same dimension is downright inevi·
table, but for the present we shall be marking • along the axis of
ordinates which is proportional to time: "l = ct. The x and t axes
arc drawn in Fig. 3.9. Every point of the plane represents the
event defined by the coordinates (x, "t). The motion of a body is
a sequence of events consisting in the arrival of this body at a
given point at a given moment of time; it is depicted as a curve
in the plane (x, <).
The uniform motion of an object is depicted in this plane by a
straight line. The propagation of a light beam at the velocity c is
depicted by a bisecting line (the equation x = •) running through
quadrants I and III when light propagates in the positive direc·
tion of the x axis and through quadrants II and IV when light
• Later we shall see that this equation determines the change of a light ire·
quency on reHection from a moving mirror (see§ 7.5),
Consequences of Lorentz Transformation K Calculus 109
travels in the opposite direction. Since the velocity of an ob~ect

is always less than that of light, the uniform motion of any obJ~Ct
is depicted by a straight line forming an angle less than n/4 wtth
the 't axis.
It is easy to find the points on the plane which depict the mo-
tion of the observers A and A'. In the frame K shown in Fig. 3.9
the observer A is at rest; we shall suppose that he is located at th.!
point x = 0. Then his "world
line", i.e. the succession of
points in the plane (x, 't) cor-
responding to the events con-
sisting in his being at a given
point at a given moment of
time, will be represented by the
axis of ordinates. Hence, the
axis of ordinates is the world
line of the observer A. The
world line of the observer A' in
the frame K is represented by
the straight line inclined to the
't axis at the angle a whose
tangent is determined by the
ratio tan a= x/'t = x/ct = v/c.
If at the moment t = 0 the ob-
servers A and A' were located
at one point, the world line of Fig. 3.9. The graphic illustration of the
the observer A' passed through determination of the relat1ve ,-elocity
the origin 0. The sending and of two observers.
reception of light signals by the
observer A is depicted in the plot (x, 't) as follows. The first
''exchange" of signals takes place at the point 0. Then, after
the time interval T (at the world point AI), the observer A
sends a light signal. Its propagation is described by the straight
line A1Al parallel to the bisecting line. The observer A' will re-
ceive the light signal at the world point Aj. The propagation of
the light signal sent by the observer A' in the reverse direction
is depicted by the straight line AfA~ parallel to the bisecting line
of quadrants II and IV not shown in Fig. 3.9. The observer A
will receive the return signal at the world point A 2• According tG
the condition OA 1 = T * and, from the definition of the coefficient
K, OA 2 = K 2 T. In the frame K the point AI is associated wit!t
the moment of time (by the clock located at the point x = 0, i.e.
at the observer A) A 3. Obviously, OA 3 = 1/2(0A 1 +
OA 2 ) =
• In the coordinates Tone should write 011 1 = cT, Uut for the sJkc of Sim-
plicity we shall not do thi~.
!10 Special Theory of Relatlvlty
= 1/ 2 (K2+ I) T. The propagation time of the second signal from

A to A' is, naturally, 1/ 2 (0A2- OAI) = 1/2(1(2- I) T.
Next, we shall note a useful theorem of the K calculus. It is
seen from Eq. (3.49) that the change of the sign of the relative
velocity, i.e. of the quantity B. transforms the quantity K into
1/ K. This means that the receding and the approaching at the
same absolute \'alue of the velocity co.rrespond to reciprocal values
of the coefficient K.
Let us now consider the case when there are three reference
frames K. K' and K" and three observers located at the corre:;-
ponding origins 0, 0' and 0". Let the coefficient K be equal to
K(A, A') for the observers A and A'; it depends only on the rel-
ati\'e \'elocity of the frames K and K' which we shall designate
by V as before. If the relative velocity of the observers A' and A"
is equal to W, the coefficient K for these observers, K(A', A").
depends only on W'. Is it possible to find K(A, A") when K(A, A')
and K (A', A") are known? Let us derive the requisite equation.
Let the observer A send two light signals separated by the time
inter\'al T registered by his clock. The observer A' receiving these
signals will find that they come in separated by the time interval
K(A, A') T as it follows from the definition of the coefficient K.
But this time is registered by the clock of the observer A'. The
observer A" is located further from A than the observer A', so
that the signals passing the observer A' go onward to A". At the
moment when A' receives the first signal from A, he sends a light
signal himself without delay to A" (do not worry: it is an
imaginary experiment!). Now two signals propagate toward the
observer A": one travelling from A and another sent by the ob-
sen·er A'. Since both of them are light signals, they propagate at
the same veloctty, having left A' at the same moment. In fact,
they propagate as one signal.
The same procedure is repeated by the observer A' at the mo·
ment when the second signal from A comes in. And again one
signal propagates from A' to A", consisting of two light puls~s
.sent from A and from A'. ,
The observer A" will receive the two signals. On the one hand,
according to the definition he wi\1 register by his clock that the
time inten·al between the signals is equal to K(A, A") T. On the
other hand, these signals were sent by the observer A' with th!!
time interval between them K(A, A') T. According to lhe defini-
tion the observer A" wi\1 find that the time interval between these
signals is equal to K(A', A") ·K(A, A') T. But the signals from A,
.and A' arriye at A" simultaneously, so that
K(A, A")-I((A, A')·K(A', A"), (3,51)
Consequences of Lorentz Transformation K Calculus Ill
The result is remarkably simple. Knowing the coefficients K

for two pairs of reference frames in which one common frame is
contained, one can obtain the unknown coefficient K pertaining to
the last pair of frames by multiplication of the known coeffi.
cients K.
Graphically this result is readily obtained from Fig. 3.10. Here
the world lines of the three observers A, A' and A.., are depictecl.
All the observers were at the same point 0 at the moment t = 0.
The time interval T later, the ob-
server A sends from the world
point A 1 a light signal whose
world line is depicted by a dotted
straight line A 1Ai'A]'. According to
the condition, OA 1=T and by the
definition OAl = K (A, A')T, OA'{ =
= K (A, A") T. On the other hand,
evidently GAo~ K (A', A")· K (A,
AD T. Note that the proposed
diagram is suitable only for a
graphical depiction of "imaginary
experiments" but cannot be used
for a geometric determination of
various quantities. The plane
(x, 't) is not just a conventional
Euclidean plane (see Chapter 4).
However, combining a graphical
geometric description with alge· Fig. 3.10. The derivation of Eq.
braic determinations, we shall not (3.51).
make a mistake.
It is easy to lind the equation for the transformation of velocities
of coordinate systems. Suppose we want to lind the relative ve-
locity U of the frames K and K.., if the relative velocity of the
frames K and K', designated by V, and that of the frames K' and
K..,, designated by W, are known.
Introducing the familiar designations V/c = B 1 and W/c = B2,
\\e obtain from Eqs. (3.48), (3.49) and (3.51)
U K2 -1 K 2 (A,A')·K 2 (A',A")-I Bt+fl2
c = K 2 +1 K 2 (A,A')·K 2 (A'.A")+I I+B1I32'
Going back to conventional designations, we obtain the equa~
tion for the velocity transformation (Eq. (3.28)):
U= 1 !t,:;c2 ·
It is seen from the reasoning quoted that the moments of
occurrence of events and the time intervals between them prova
'"
to be different for observers from different IFRs. To detect this,
let u~ return to the experiment analysed earlier and involving
an exchange of light signals between the observers A and A'. We
shall recall that the first "exchange" is performed at the moment
when the observers are located at one point. At that very moment
the clocks of the observers A and A' are set to the zero reading.
Then after the time interval T by his. clock the observer A sends
a signal directed to A'; according to the definition, the time in-
terval separating the reception of the first and the second signals
by the observer A' is equal to KT by his clock. However, the ob-
server A will ascribe the moment of time 1/ 2 (K 2 +I) T to the
reception of the signal at A' and will assume that the signals
sent by him at the intervals T will reach A' with the intervals
1/ 2 (K2 +I) T. As it was mentioned, the same interval in term-;
of the clock A' is equal to KT. Hence, the time interval between
the two identical events, the arrival of the first and the second
signal at A', proves to be different: in terms of A' it is equal i.o
J(T and in terms of A it is equal to 1/ 2 (1(2 +I) T. Thus, we dis·
covered that the time of the event. i.e. the arrival of the second
signal, is relative: it is equal to KT in terms of A' and
1/ 2 (K2 +I) T in terms of A. The time interval between the two
events proved to be different for A and A' too._All. this i!l_dicates
that t_tt~ time _oLan event as well as the time interval betWeen
-evem-s are relative values. · -- ---
Under what conditions will these values coincide? It happens
when KT-;;:::::; 1/2(K 2 +I) T. It can readily be inferred that it is
possible when K ::::: I or, as it is seen from Eq. (3.48), when
V{c--+0. Thus, the difference in time readings and the relativity
of time intervals between events can be neglected in those IFRs
whose relative velocitieS are small compared to that of light.
The proper time. The K calculus makes it possible to determine
readily a relationship of a time interval between two events that
occur in a certain IFR at one point in space and are, consequently,
registered by one clock (the proper-time interval), and a timr
int~rval betwzen the same events registered by two clocks of
another IFR in which the considered events occur at different
points.
Now let us go back to the exchange of light spots. If A sends
signals at the interval T by his clock, A' receives them at the in-
terval KT by his clock. Ho\\·ever, as we saw before (p. 108), this
+
interval is equal to 1/ 2 (K 2 I) T in terms of A. It is the ratio of
these quantities that gives the relationship between the proper-
time interval .l:t = KT and the time interval 1.t registered by two
docks of another I FR. This ratio is equal to
l!.t KT 2K . ~--V-'
Tt= l/2(K~+ I) T =~=~I --zr •
Consequences of Lorentz Transformation. K Cafc"ulus 113
where in the last link Eq. (3.48) is used. This result is o[ course
familiar to us.
Relativity of rulers' lengths (distances). Suppose we have two
motionless points in the reference frame where the observer t1'
is at rest. One may presume, although it is far from being oblig-
atory, that these pomts are a ruler·s ends. Let the ruler move
[rom the observer A and the observer A' be located at the end of
the ruler which is nearer to A. (Do not forget that the ruler is
oriented along the direction of the relative velocity.)
To determine the length o[ the ruler, the observer A sends a
signal at the moment t~. registered by his clock, and waits for it
to return after reflection from the far end of the ruler. Let the
moment of the signal return be t4 by the clock of A. Obviously,
the moment of the signal re[]ection is equal to 1/ 2 (!1 +
!4). Exactly
in the same manner, a signal can be sent to the near end of the
ruler (say, at the moment ! 2) and the moment of its return deter-
mined ([or example, t3 ). The moment of the signal reflection [rom
the near end is equal to 1/ 2 (1 2 +! 3). Both signals are reflected
simultaneously (by the clock of A) from both ends of the ruler,
provided the following condition is met:
~~+W-~~+~ ~~
In an imaginary experiment this condition can be satisfied !.:l~·
choosing the times of sending of the first and the second signals.
The first signal from A, however, will be received by the ob-
server A', located at the near end of the ruler, at the moment Kt 1
(recall that the initial readings of the clocks of A and A' coin-
cided when the observers were located at one point). The signal
reflected [rom the far end of the ruler and returning to A at lhr!
moment t 4 will pass A' at the moment !1/K. Indeed, the signal
recei\·ed by A' at the moment 14/K will get to the observer A at the
moment (t 4/K) ·K = h From the viewpoint of the observer A' tile
doubled length o[ the ruler 10 is determined as the time interval,
taken by light to reach the far end o[ the ruler and get back, multi-
plied by the velocity of light, i.e.
-4-(~- Kt 1)c=l0 (3.53)
As to the relationship between !2 and ! 3, it follows directly from
the definition of the coe[ficient K:
t,- J('t,. (3.54)
The imaginary experiments performed to measure length are
illustrated in Fig. 3.11, which does not require any special expla-
nations after the diagrams of Figs. 3.9 and 3.10 have been anal-
ysed.
The ruler's length determined by the observer A is equal to th>!

difference of the distances from him to the far and near ends of
the ruler under the necessary condition that these distances are
determined simultaneously. This condition is satisfied owing to
the validity of Eq. (3.52). The distance from A to the far end
is equal to 1/ 2 (1 4 - tt) c and to the near end to 1/ 2 (1 3 - t 2 ) c. Con-
sequently, A has to assume the ruler's ~ength l equal to
I~ -fr [(t,- 10) - (1 3 - 13)) c. (3.55)
Eqs. (3.53)-(3.55) make it possible to find the relationship

between l and l 0. It follows from Eq. (3.52) that !4 = t2 !3- t1. +
Substituting the expression obtained for t~ into the left-hand side
of Eq. (3.53) and resorting to Eq. (3.54), we get
lo=f (t2 +~-t 1 _ Ktl) = f [ 12 (K 2 + I) -;I1(K 2 + I)]=
= f K2K+ I (/2- tl). (3.56)
Since according to Eq. (3.52) t2 - 11 = 14 - t3, it follows from

Eq. (3.55) that
(t,-tL)-(13-12) .!..
2 '
Now Eq. (3.56) takes the form
lo=lK22;' = ~·
where in the last equation the formula (3.50) is taken into ac-
count. This is exactly what we obtained earlier as Eq. (3.5).
This derivation shows quite distinctly how essential it is to find
the ruler's ends simultaneously when its length is determined.
Incidentally, note that the derivation of Eq. (2.4) involves, in
essence, a radar approach as well.
The Lorentz transformation. We have made sure that the K cal-
culus can be employed to derive all basic principles of the STR,
the Einstein postulates. The advantage of this derivation lies in
the fact that there is no need for an explicit introduction of a co-
ordinate system.
But, of course, the application of STR methods in physics re-
quires an explicit introduction of a reference frame. If so, the in-
troduction of the Lorentz transformation is outright inevitable.
The Lorentz transformation can be derived by means of the K
calculus.
Consider the two reference frames K and K' with the respective
observers A and A' registering the same event. In both frames the
imhal time reading is chosen so that t = t' = 0 when both

origins coincide. Then at the moment t 1 the observer A sends a
light signal to A' which is received by him at the moment 1] by
his clock; the signal sent by A proceeds further accompanied by
the signal sent by A' at the moment when he receives the signal
from A. In fact, one signal consisting of two propagates along
~ the x axis. Let the event
~ §- P represent the arrival 1)f
.k ~ that signal at some point
Observer~. <;;:: ~ (or the arrival of thr.:
woA'I'td tine A~ ~ signal coincides with the
'§!
~
;$
~ l moment of occurrence of
a certain event). At that
'
Fig. 3.11. The determination of the length Fig. 3.12. The derivation of
of a moving ruler. the Lorentz transformation.
point the signal is reflected (or, otherwise, the return signal is sent
immediately on the arrival of the direct one). First, it gets to the
observer A' at the moment t2; at the same moment A' sends his
stgnal in the direction of A. Now the single signal consisting, in
fact, of two signals propagates from A' to A. It is received by the
observer A at the moment t2 (Fig. 3.12).
The observer A will ascribe the coordinates to the event P as
follows. The time t of the event is just the half-sum of the time~
of sending and reception of the signal since the velocity of light
on the way "there" and ··back" is equal:
I~~ (1, + t,). (3.57)
The distance to the point where the event occurred can be found
if the propagation velocity of the signal c is multiplied by the
time which the signal takes to travel "there"; this time is equal to
116 Special Theory of Relalioity
half the total time spent by the signal. Since the signal travelled
a closed path during the time t2 - t~, the event coordinate x will
be determined by the observer A as
x~-l-(1 2 -l,)c. (3.58)
From Eqs. (3.57) and (3.58) we obtain
tl=t-f. /2=t+f. (3.59)
nut the observer A' will lind in exactly the same manner that
(3.60)
According to the definition of the coefficient K. and comparing

the intervals between the exchanges of signals, we get
t;-o~K(t,-0). t,-o~K(t;-o). (3.61)
According to Eqs. (3.59) and (3.60) we obtain
t'-f=K(t-f). (3.62)
t+f=K(t'+f). (3.63)
Multiplying crosswise Eqs. (3.62) and (3.63), we immediately
obtain that the quantity
t 12 - ;:
2
=fl-*.
retains its value in all IFRs, i.e. is the invariant. Having written
(3.64)
Eqs. (3.62) and (3.63) in the form more convenient for solution
t'--f=K(t-7). (3.65)
t'+f=-R-(t +7)· (3.66)

we readily find that
t' = K~~ I t - K~;; I X, x' = K~i I X- K~~ I ct.
Taking into account Eq. (3.50), we discover that this is just the
Lorentz transformation:
x'=l'(x- Vt), t'=f(t -~ x).
CHAPTER 4
THE FOUR-DIMENSIONAL
SPACE-TIME
§ 4.1. Three-dimensional and four-dimensional Euclidean spaces.

When we introduce a coordinate system, the position of every
point is specified by three numbers which are referred to as the
coordinates of a point. A manifold of three dimen,.ions is under-
stood as a set of all points. If we want to pass over from a mani-
fold to space possessing definite geometrical properties, we have
to define the expression for a distance between two infinite close
points of the manifold. Having assigned the square of the distanc~
between such points, one can define basic geometric quantitie5,
such as a vector's length, an angle between vectors, areas of two-
dimensional figures formed by vectors. Geometry, whose princip~l
laws were formulated by Euclid, is valid to a high degree of ac-
curacy in the world that we live in. In accordance with Euclidean
geometry the square of the distance between two infinitely close
points can be put down in the Cart~sian coordinates in the fo:-
lowing form:
+
ds2 = dx? dy2 dz2• + (4.1)
This equation represents nothing other than the Pythagorean
theorem written out for the diagonal of a rectangular three-
dimensional parallelepiped with the sides dx, dy, dz.
A coordinate system can be selected at will (the Cartesian coor-
dinate system is distinguished only for its simplicity), and a
distance between points, owing to its geometric meaning, should
not depend on the choice of a system. This means that Eq. (4.1)
has to be an invariant of any transformation of coordinates.
A disl ~nee between any two points has also to be an invariant
of the transformation of coordfnates. Thus, in Euclidean geometry
the invariant is the distance between two points:
r 12 = ...j(x2 x 1) 2 + (y 2 y 1)'2 + (z
2 Zt) 2 , (4.2}
where (x 1, Yt. zt) and (x 2, y2, z2) are the coordinates of two points
in space. Eqs. (4.1) and (4.2) relate the coordinates of two points
of space. In particular, the transformation equations for transi-
tions from one Cartesian system to another are given in Ap-
118 Special Theory of RelafitJity
pendix I, § 2. Such a transition represents a rotation, provided we

ignore the system's translation which is of little interest to us.
It is seen from this equation that in a new coordinate system any
new coordinate is expressed through all old ones.
In a three-dimensional Euclidean space one can introduce vec-
tors specified by a triad of numbers, i.e. vector components. The
coordinates of a point comprise the ,components of a radius vee·
tor. Consequently, components of any vector are transformed ac-
cording to the coordinate transformation rule. Norms of vectors,
their dot products and an angle between them are found via
Yector components according to the known rules.
What would the appearance of one more dimension in a Euclid-
ean space imply? Certainly, it is difficult to visualize a four-
dimensional space with one's own eyes. But there is no such need.
Having available the principal relationships for a three-dimen-
sional space, we just carry them over to a four-dimensional space.
Let the coordinates of a point in the four-dimensional space be
x, y, z, w. For a four-dimensional Euclidean space the square of
the distance between two infinitely close points will be written
in the following form (the symmetrical designations are also
given):
+ + +
ds2 = dx2 dy2 dz2 dW = dxft + dx 12 + dx -t- dx'
22 2, (4.3)
and the distance between points
r12=.Y(x2 Xt) 2 -t-(y2 Ytf-t-(z2 Ztf-t-(W!! Wt) 2. (4.4)
Eqs. ( 4.3) and ( 4.4) will be the invariants of the coordinate
transformation, and basic geometrical relationships will be found
in much the same way as they are found in three-dimensional
space.
§ 4.2. The 4-space-time, or the four-dimensional pseudo-Euclid-
ean space. Let us consider a four-dimensional manifold made up
of "points" whose coordinates are constituted by four numbers
x, y, z, -r = ct defining a four-dimensional point. One or another
-event representing an instantaneous physical process can occur
at any point of this manifold. The four-dimensional space-time is
a purely geometric notion. Sometimes, following Minkowski, this
space is called the "world". Any event occurs at some point of
the Minkowski world.
Geometric properties of the Minkowski world can be established
after some invariant relationship between coordinates of poinlo;
is found, which can be interpreted as the distance between two
points of a manifold. When the distances between points are
defined, we may pass from manifold to space. But how can the
necessary invariant relation be found? It should not be forgotten
that the coordinates of the "world" points arl' <Jefmcd with physi-
Four-Dimensional Space-Time 119
cally different quantities, so that it is impossible to presume in

advance that the "distance" in this world can be defined by the
expression of the (4.3) type. But the theory of relativity answer~
this question unambiguously. Considering only inertial frames
of reference, the interval between events (Eq. (3.19)) remains
the invariant for any pair of events, or, in terms of geometry, for
any pair of points in the Minkowski world. The transition from
one IFR to another is described by the Lorentz transformation,
and no other transformation is needed in the framework of the
STR. Consequently, from physical considerations we can take the
expression for the square of the interval between events
ds 2 -d<2 -dx2 -dy'-dz'-d:t" -dx' 2 - d:t"- dx'2 (4.5)
as the basic invariant quadratic form defining the "distance" in
the Minkowski world. Here • = ct. It is Eq. (4.5) that defines the
square of the distance between two infinitely close points in the
Minkowski world. Thus, the Einstein postulates, from which the
invariance of the interval between events follows, signify that
geometry of the four-dimensional space-time, i.e. the Minkowsl<i
space, is determined by the basic fundamental form of the (4.5)
type. It is seen from the appearance of this form that coordinates
and time are not equivalent.
It will be shown in Supplement V that the transition from
inertial frames of reference to non-inertial ones alters the ap-
pearance of the interval between events. Although this expression
always remains invariant, its form becomes different, so that the
square of the interval takes the following form:
ds 2=g1kdx 1dxk, (4.6)
where the indices i and k denote summation from I to 4 and the
coefficients g,k, referred to as metric coefficients, may depend on
coordinates and time. We need this general equation now only in
order to write out g;k for Eqs. (4.3) and (4.5). Using the sym-
metric designations of Eqs. (4.3) and (4.5), we obtain respec-
tively
go)=g••=g22=g33=l, (4.3')
goo=l, g••=g22=g33=-l. (4.5')
It is evident that Eqs. (4.3) and (4.5) differ by signs of metric
coefficients. A totality of these signs is called a signature of cor-
responding quadratic forms. The signature of Eq. (4.3) has the
form (+ + + +), while the signature of Eq. (4.5) (+-- -).
If a four-dimensional space were formed by a simple increase of
the number of dimensions of our conventional space, the signature
would be (+ + + +). Such a space would not differ from our
space in anything except the number of dimensions and it would
be referred to as the Euclidean (four-dimensional) space. Tne

signature of space considered in the special theory of relativity
corresponds to that of Eq. (4.5), i.e.(+---).
A change of a signature implies a variation of the "distance''
between points in space, the variation of properties of this space
as compared to those of the customary Euclidean space. This
four-dimensional space, possessing unusual geometric properties,
is e:..tremely important for the STR. 1t is in this space that all
physical phenomena take place.
The geometry of the Minkowski world differs from Euclidean
geometry, but not too much, since the coefficients in Eq. (4.5) as
well as those in Eq. (4.3) are constant. Accordingly, the geometry
defined by the quadratic form of Eq. (4.5) is customarily called
pseudo-Euclidean and the corresponding space a pseudo-Euclidean
space.
Thus, the space of the four variables x, y, z, ct of the special
theory of relativity is the four-dimensional pseudo-Euclidean space.
It is not originated just by adding the fourth (time) coordinale
ct to the three spatial ones x, y. z, but through the peculiar de-
finition (Eq. (4.5)) of the invariant distance between the points
of this space.
A physical motive for the consideration of the pseudo-Euclidean
1 arhde
~~;c:o u:~ ~~·et~te ~~~t n~ta~~~~v!re~\iai n tk~~ r~~d~~IT: ~r~~~~;
close connection.
§ 4.3. 4-vectors and 4-tensors. Exactly as in a three-dimensional
space, coordinates of a point in a four-dimensional space can he
treated as components of a four-dimensional radius vector drawn
from the origin of a coordinate system to a given point. All four-
dimensional vectors will be designated by an arrow over a letter;
in p_;uticular, a four-dimensional radius vector will be designated
by R. For the convenience of our readers we shall be presenting
basic relationships both in complex notation and via the real
variables. Complex notation simplifies the presentation of electro-
dynamics, while the usage of real variables leads us to the for-
malism of the general theory of relativity, where the introduction
of a complex coordinate is of no use. Most of the equations will
be written in a symmetric notation, and the presentation of coor-
dinates of a four-dimensional vector in a two-line form will make
it possible to recall the meaning of the introduced designations.
Thus, we introduce a four-dimensional radius vector in one of the
following ways:
i\(x' yx,
X
x3
Z
~·
l. =
• ):
!CI
a
( )
I ii( •
x'
= ct
x'
X
x'y x')
Z
(b).
(4.7)
Four-DimensioMl Space-Timo 121
The usage of superscripts in Eq. (4.7b) is not accidental. When

real values of coordinates are used, the difference between co-
variant and contravariant vector components needs to be em-
phasized (see Appendix I, § 8), and, consequently, superscripts
are used with contravariant components. So the square of the in-
terval between events will be written in the form
ds 2 = dx~ + dx~ f- dx~ + dx! = ds 2 = dxrfl- dx 12 - dx? 2 -
= 1~1 dxi = g 11l dx 1 dx 11 ; (a) -dx' 2 ~g,.dx'dx'. (b)

(4.8)
Differing from zero: Differing from zero:
gu=l, g22=l, goo= I, gu=-1,
g33=l, g44 =1. g22=-l, g33=-l.
Note that the intervals given in Eqs. (4.8a) and (4.8b) have
opposite signs. Since ds 2 may be either negative or positive, the
choice of signs for ds 2 is of no practical importance.
The Lorentz transformation is a transformation of four-dimen-
sional radius vector components, that is coordinates of an event.
We shall write them out again;
X~= r (x, + iBx4), xY = r (x-0- Dx 1),
X~=X2 , (a) x'" ~ r (x' - Dx"), (b)
(4.9)
x;=x 3,
x~ = r (x4 - iBx 1); x'' ~x'.
A four-dimensional radius vector is one of four-dimension::~!
vectors, so that if in the reference frame K the following four-
dimensional vectors are specified
:4(A,A2A3A~) I :4(A0 A1A2A1),
the components of the same vectors will be determined as follows
in the frame K':
A;~ i'(A, +iDA,), A'~ I' (A'- DA'),
A;=Al, (a) A''~ r (A'- BA~, (b)
(4.10)
A;=A3, A2'=A2,
A;~l'(A,-iDA,); A3'=A1.
Eqs. (4.8a, b) represent the square of an infinitesimal vector

(di?) 2• Consequently, the square of the norm of a four-dimensional
122 Special Theory of Refatluilg
vector (which is an invariant quantity) must be determined in

this way:
A2 =g 11,A 1A..,= IA 2 =gt,.,A 1A11 =A,.A"'=
=Ai+A~+A~+A~; {a) =Ao2-A 12 -A22 -A32 , {b)
(4.11)
Of course, Eqs. {_!-lla) and {4.llb) give opposite signs for the
invariant quantity A 2• But this is of no significance just as in the
case when the sign of the interval is determined {sec the not~
after Eq. (4.8)). II should be borne in mind, though, that different
signs of the interval alt~r the conditions defining "time-like" and
"space-like" intervals and vectors. (There is no harmony in the
literature concerning this issue.)
In Eq. (4.llb) we introduced covariant coordinates according
to the formulae of Appendix I,§ 8: Ak = g,"A;_ It is easy to notice
that A0 =A 0 , A1=-A 1, A 2 =-A 2 and A3=-A 3 •
Just as in the case of a three-dimensional space, we shall have
to deal with tensors. Most easily the tensor component transfor 4
mation law is derived from the transformation law for a pro1-

uct of two four-dimensional vector components. The transforma-
tion ~quatio2.s for the components of the four-dimensional vec-
tors A and 8 can be written down using the symmetric notation
(see Eqs. (2.40a. b))'
A 1=Ci 11 A;,
(4.12)
Bk=Cik,.B~;
Multiplying the left-hand and right-hand sides of these equa 4
tions, we obtain at once the transformation rules for vector com-

ponent products:
A 1 B~ = Ci 11 Ci~,.A;B~,(a) J A 1 B~ =a 11 a~,.A' 1 B'"'. (b) (4.13)
Thus we obtain the general transformation law for the tensorq
T,~= A,811 and Ti11 = A'Bk:
1,11 = a 11 a~,.r:,.. (4.14)
T'k=auak,.T' 1"'. (4.15)
Eq. (4.15) in which the difference between the covariant and con-
tra\'3riant coordinates is essential, represents the transformation
law for a twice-contravariant tensor.
In the 4-space the m~asured physical quantities should be '>0
arranged as to possess quite definite transformation properties
with respect to a transition from one IFR to another, i.e. to the
Lorentz transformation. But in the coordinate transformation (in 4
eluding the fourth coordinate of the Minkowski world) only tensor

quantities possess the definite transformation properties, and
tensors of different rank transform according to different rules.
Hence, all physical quantities to which we ascribe a real meaniug
have to be tensors: either scalars, i.e. zero-rank tensors, or 4-vec·
tors, i.e. first-rank tensors, or, finally, tensors of a higher-than-on~
rank. We shall see later that an electromagnetic field forms a
second-rank tensor (see Chapter 6). The transition from customary
three-dimensional quantities to four-dimensional ones (which is.
no doubt, necessary in the case of the Lorentz transformation) is
not always straightforward and is realized differently in different
cases. It is often possible to represent, with some modification, a
customary three-dimensional vector as a spatial part of a 4-vector.
As to the fourth component, its expression seems to be rather sur-
prising at first, but in the final analysis proves to be natural.
There is nothing amazing in this since in a non-relativistic limit
we nearly always come back from relativistic relationships to
classical ones.
Chapters 5-7 provide numerous examples of constructing four-
dimensional vectors and tensors.
§ 4.4. A pseudo-Euclidean plane. Characteristic features of the
p~eudo-Euclidean space can be illustrated by means of the pseudo-
Euclidean plane. One of the two
coordinate axes must neces-
sarily represent the time axis, or
the axis of time-proportional
quantities, since in the STR
purely spatial geometry remains
Euclidean, and only space-time
is described by pseudo-Eucli-
dean geometry. In our choice of
reference frames it is most con-
venient to consider the plane ject fig. 4.1. (a) The world line of an ob-
at rest at the point x = x0• (b)
(x, <). The world line of an object moving
Recall that the four-dimen· uniformly along the x axis.
siOnar· sp-are--time --co:-nUriiillrn
-WhLse pOints represent events is sometimes called the Minkowski
~Orld. Every event in our real ph-ysical world occurs at a definite
world point of the Minkowski world. Considering a particle, one
can regard its staying at a given point at a given moment of time
as an event. No matter whether this particle moves or not, the
sequence of events happening with the particle in the Minkowsld
world yields a certain curve called the wof(4__liae_.oi the_partick_,_
Let us draw the x, '( axes--or lhe frame K at right angles to
each other and analyse the simplest cases. Let a particle be
located at the point x = x0 in the frame K; ils world line in the
plane (x, T) of the Minkowski world will be a straight !me par-

allel to the T axis (Fig. 4.la). Let another particle move uni-
formly along the x axis in the frame K at the v'!locity v. Itt
world line in this frame will be a straight line inclined at th~
angle 9 to the "C axis (Fig. 4.lb).
A bit later we shall see that
9 =>arctan (v/c).
Now we shall examine an ar-
bitrary motion of a particle in
this r~ference frame. The motion
of this particle is represented
by the world line x = x(l") in
the plane (x, T), as it is depicted
in Fig. 4.2.
The inclination of the world
line to the T axis at each given
point is determined by the de-
rivative dx/dT at that point. In-
World lfl?!' deed (see Fig. 4.2),
oftiJlhl
rays
tana=*=~%=f. (4.16)
Fig. 4.2. The system of real coordina·
t~s x, 1: = ct. The particle's position at Thus, the inclination angle is
a ~i\"en moment is specdi.ed by the
pomt in this plane Th~ particle"s mo·
determined from the following
equation:
~~~~e~s ,~~~~dt~~n~n o1r~ ~~~~~ ~het~~o~fd
lm~s of motionless pomts are straig-ht
hnes parallel to the 1: ax1s The \\oriel
a= arctan f =arctan~. (4.17)
line ol lig-ht rays is the coordinate an·
g-le bisector. In the case of the vana·
where ~ = v/c and v is the in~
ble velocity the angle Jormcrl by the
tangent line to the world line and the
stantaneous velocity of the point
or the object. Inasmuch as
1: axis is defined from the relation 0 =
= arctan (v/r), where v is the instan·~ < I always, the angle a can·
taneous velocity of a particle. not exceed 45° for any mo\·ing
object. The world line of ligl1t
rays will be represented by the bisecting line of the coordinate
angle.
We saw in § 2.9 that the ·t', x' axes are obtained from the t", x
axes as a result of the Lorentz transformation, provided thes~
axes are drawn together in a scissors·like manner to the world
line of light rays. The relativity of simultaneity is graphically seen
in Fig. 4.3a where the t"', x' axes are drawn together with the "t", ..\.
axes. In the frame K' all events lying on the x' axis, or on the
straight lines "t"' =canst, are simultaneous. In terms of geometry,
all these lines parallel to the x' axis represent the simultaneity
lines in the frame K'.
Let us consider the two events A1 and A2 lying on the r axis,

both these events occurring simultaneously in the frame K' at
the moment t' = 0. To find the moments of time at which these
two events occur in the frame K, one should .. project" these events
on the 't" axis by drawing straight lines parallel to the x axis,
'~!·
__: ___ ~~
I N - N---
0
.r
lll
Fig. 4.3. (a) The Lorentz transformation reduces to the rotation of the x
and T axes through the angle ~ = arctan 8 about the origin of coordinates to·
w_nrd the coordinate angle bi!oeetor and their new positions x', T'. The straight
lines x' =cons! are now parallel to the OT' axis, while the straight lines
~~:;' ~blf~u:.r:n:r:dall:~·st~mth~f ~~~r~~~~te~~e T~~v~ef:t~~~ty o;fer si~uW:n;~) 11 /~
dear!) sten: the events A. and A2 which are simultaneous in the frame K' (ly·
ing on the straight line T' = cons!) are not simultaneous in the frame I( To
fiild the respective moments in the frame 1(, we project them on the T axis by
means of straight hnes parallel to the x uis. (b) Here are two world lin~s
of objects (LL and MM). The relativity of the distance between moving objects
is seen very well To find the distance between them, one has to determine the
coordinates of these objects simultaneously Let one of the objects be located at
~1 t t~~i~!r! ~~c;;e~~- ~~T~n°fte~~s r~f~~/rr~h~es~?~~e 0~~~~~~s ob~e~~e .!o~f\h~
point P at the same moment. The sections NR and NP corresponding to the
distances between the objects have different lengths.
since in the frame K1 the events lying on the straight lines T' =
= const (Fig. 4.3a) are simultaneous. We see that in the frame K
these events occur at different moments of time t 1 and t2 . Of
course, this is only a geometric illustration of the relativity of
clock synchronization that we dealt with in § 2.4.
wo~ldvJrXe~~F~~~n~bj:~t~1 ~~oJf~;~~{f~~m~i\ut·;tdi1Je::n~w:el~~~
ilies. To determine the distance between them at a given moment
of time, the coordinates of these objects should be found simul·
126 Special Theory of Relufiuity
taneously m the frame in which this distance is bemg determin~;>d,

It is clearly seen that the distance between objects measured in
the frames I( and }(' proves to be different. Due to the equivalence
of reference frames none of the distances obtained can be regarded
true. But then all laws of mechanics, in which force depends on
distance, become ambiguous in the case of moving objects. Nat·
urally, this problem did not emerge
x--ct .x"" ct in Newtonian mechanics where
time was regarded absolute.
Let us consider the x, 't axes
of the frame K (Fig. 4.4). The
IV square of the interval between two
I
world points is defined by the ex·
Absolutely Ab#f#,f[jY pression si~=(• 2 -'t 1 ) 2 -(x 2 -x 1 ) 2 •
remote
events Absolute event.;
past For the sake of simplicity let :1s
//I suppose that event I occured at
the point x = 0 at the mo-
ment T = 0, i.e. at the point 0.
Fig. 4.4, The intersection of the-
space-time cone by the plane (x, t). Any events that occurred on the
The point 0 represents event I. Att x axis before and after event l
events located in quadrants Ill and are depicted by points in the
IV represent absolutely remote plane (x, T). Since the square of
events wilh respect to event 0. The
events located in quadrant I repre- the interval, that is the distance,
sent absolute future while the from event I to any other event
events located in quadrant II abso· is equal to s2 = 'T2- x2, this plane
lute past. is subdivided into four quadrants
I, II, Ill, IV by the straight lin~s
x = T, which correspond to the sequence of events consisting
in the emission of a signal from the point x = 0 at the mo-
ment T = 0 and its arrival at the point x at the moment T. The
interval between the events located on the straight lines
T2 - x2 = O_i_s_l_!gb!.·Hke, and the "distance" between such events
is equal to zero m the pseudo~~Jl_cHdec3-n_p!!n~. Now let us consid-
~~at~;a~~·y s~u~~~n~ x~x;r~~r c~n~~~u~i~~~.urhe si~f~;vh1! 1 ~~~~e!~
any event of quadrant I and event I is time-like. For all events
of this quadrant T > 0; consequently, all of them will occur after
event I, and no choice of a reference frame can alter this situa-
tion. This means that quadrant I is the region of absolute future
with respect to 0. In quadrant II s2 > 0 also, but here for all
events T < 0; hence, quadrant II is the region of absolute past
with respect to event I.
In quadrants Ill and IV s2 < 0, i.e. the interval between any
event located in this region and event I is space-like. All these
events occur at points which do not coincide with the point at
which event I occurred, and again it is impossible to alter this

by the choice of a reference frame. However, one can find such
reference frames where a given event from quadrant III or IV
can happen before or after, or, finally, simultaneously with event
I, since the concepts "simultaneously", "before" and "later" are
relative for the events located in this region.
If one examines two events located arbitrarily in the plane
(t, x), the character of the interval between them will be de·
termined from the slope of the straight line c:onnecting these two
points. If the straight line is inclined
~"
to the x axis at the angle exceeding
n/4, the interval between events I
and 2 is time-like; if the angle is
less than n/4, the interval is space-
like. Finally, if this line is parallel
to the bisecting line, the interval is
light-like.
In the four-dimensional space the
equation describing the propagation Pse(Jt/o-Pyll!ogorean tl!eorcm
of light has the form c'l['l- x'l- AB 2~JJC 2-AC 2
thit e~u!~i: ~e~~e~~n~: ~f ·~~~~·~t~~ ~~~~ t~~or!~e inp~h~d~~~~~~~~~:

0 4
the four-dimensional space. Usually. clidean space.
this cone is called a light __ cO@._The
internal caVIties of this cone correspond to the regions of "absolute
future" and "absolute past". The light cone surface on which the
light-like directions are located is remarkable owing to the fact that
its position in the four-dimensional space remains invariable for
every world point under all transitions from one IFR to another.
Let an event consist in the arrival of a light ray at a certain
world point where an observer is located. Thus, we deal with the
observation of light signals at a given point of space and at a
given moment of time. The light rays can get at a given world
point only along those directions of the four-dimensional space
which lie on the "light cone of past" down to infinity (practicaliy
far enough in terms of light units). Each generatrix of this cone
can be associated with the point on the spatial sphere of an in-
finitely great radius in whose centre the observer is located. Such
an assumed sphere is used for the observation of celestial bodies
and is called the sky sphere.
When depicting the pseudo-Euclidean plane on a sheet of pa-
per, it should be remembered that we are used to such relations be-
tween the lengths of rulers, which are customary in the Euclidean
plane. In Fig. 4.5 a right triangle is shown with the side AC equal
to x2 - x1 and BC to t 2 - t 1. But in this plane AB 2 = BC 2 - AC2,
according to the definition of the square of the interval and con·
trary to the Pythagorean theorem; so this is the pseudo-Pytha-

f~.r~rs~~~~~e~ ;e~f~~~~d' ~~~tf~~FY~rison of lengths in
the plane
In the Euclidean plane (x, y) the locus of points equidistant
from the origin of coordinates is defined by the equation of th~
+
circumference r2 = x2 y 2 =canst. In the pseudo-Euclidean
pla.ne (x, T) where the square of
the distance from the origin of
coordinates is defined by the rela-
tionship s2 = T 2 - x 2 , the locus
of points "equidistant" from the
x' ~i,,inpaffer~00f~~~a~e;pe~~~f~s 4 '~2
is not necessarily positive). If one
chooses the hyperbola for which
s2 = I and draws rays from the
origin of coordinates till they in-
tersect with this hyperbola, the
section of each of such rays will
determine the unitary "pseudo-
Euclidean" length in the corres-
ponding direction. It is possible
~~f~s\~·2h:2 ~u~.ei~~~;r~ ~yp:;~ to give the physical interpreta-
tion for plotting the hyperbola
~;~~~~~c~n ~~~eL~~~~~~n~~~~:r:rt~~ti~~ s2 = I. Let particles having vari-
leaves the expression -r2- x2 = ous velocities but the identical
= &/2- x2 invariant, we shall also lifetime -r 0 = I be generated at
obtain the hyperbolas 'f'2- x' 2 = I, the world point -r = 0, x = 0.
x'2- 'f'2 = -I in the new oblique· Then the locus of the world points
angled coordinate system. But this
means that these four equilateral at which these particles decay,
hyperbolas cross the axes x, T, x', will be the hyperbola s2 = I, and
T ill the distances from the origm the world lines of these particles
equal to unity The hyperbolas plot· will be represented by the rays
led are referred to as scale hyper·
bolas. outgoing from the world point (0,
0) and reaching this hyperbola.
Let us consider the two pairs of equilateral hyperbolas in the
plane (x, -r):
-r2-x2=l, (4.18)
(4.19)
One can readily subdivide the plane (x, -r) into four quadrants,
each containing one hyperbola. The dividing lines between the
quadrants prove to be the asymptotes of these hyperbolas. Indeed,
substituting the equation of the ray -r = kx, passing through the
origin of coordinates with the arbitrary slope k (k =tan a), into
the equations of hyperbolas (4.18) and (4.19), we discover that
the intersection coordinate is determined from the equation x 2 =

= ±(l/(l-k2 )). This equation has a real root only if k 2 <I.
When k 2 = I, the coordinate of the intersection point on the x
axis moves away into infinity. This means that the rays '= x
are asymptotes of these hyperbolas.
Thus, the world lines of the light rays r'
x = ct are the asymptotes of the hyper-
bolas defined by Eqs. (4.18) and (4.19).
Each of these hyperbolas intersects
only one of the axes: x or •· The inter-
section points of the hyperbolas (4.19) i
with the x axis arc determined from the
condition • = 0. We see that the hyper-
bolas (4.19) intersect the x axis at the 0
points x =±I. In a similar way one Fig. 4.7. The geometric illu-
can find that the hyperbolas (4.18) in- stration of the relativity of
tersect the, axis at the points 1" =±I. ruler lengths Quadrant I of
Inasmuch as the hyperbolas (4.18) and Fsi~. r~fe/satd~~~~t~~ t~!r£~ 11~~
(4.19) cut off the unitary sections on K. The world lines of its ends
the coordinate axes, it is natural to call are OT and AA". The h\rcr·
them the scale hyperbolas. bola x2 - T2 = I intersttt-;
Since the expression , 2 - x2 = :~: ~,a;~~sat ;~eth:i~~i~t a~~
= c t - x is the invariant of the Lo- Thus, OA = 1 and OA' = 1.
2 2 2
rentz transformation, the equations To find simultaneously the
't'2-x'2= I, 't'2-x'2=-l will be positionoltheruler'sendsin
va:tid in the frame K'. It follows directlv the framt> K', the world lines
thaf [h·e same fi}'perbolas cut off the uni"- r:r!~~t rl~~~~ e~od~esh~~~:i~~i
tary sections on the new oblique-angle.-;! line r' = cons!. lor example,
axes x' and 't' as well. with the .r:' axis (correspond·
It is directly seen from Fig. 4.6 that mg to the moment I'= 0).
the unitary sections of the X and x' axes Jhhee~ra~: ;,ult~;~ 5 1 ~~ft~o ~~
are far from being equal. It should be equal to OA". But OA" <
remembered though that the represent~- < OA' = 1.
tion of the pseudo-Euclidean plane in
the Euclidean one is conditional and the "proper" units of length
are identically chosen.
Now it becomes easy to explain in geometrical terms how the
contraction of a moving ruler comes about. Let us show the x, 't
axes and x', t' axes in one figure, and plot that part of the hyper-
bola that passes through quadraht I of the coordinate systems K
and K' (Fig. 4.7). The section OA represents a unitary rul-er
which is at rest in K. Its world lines in the frame K are straight
lines parallel to the Q, axis and passing through the points 0
and A. But in terms of the frame K' the simultaneous position
of the ends of the section OA at the moment t' = 0 corresponds
to the intersection of its world lines with the x' axis, i.e. to tht!
5-W
130 Speclal Theory of Relatiuify
points 0 and A". The unitary ruler in K' is equal to OA'; it i::.
seen from Fig. 4.7 that OA" < OA' = 1.
Suppose now that a unitary ruler is at rest in the frame K'
'ILl
(Fig. 4.8). Then its length is equal to OA' and its world lines ar~
parallel to the Or' axis, one of them being the or axis itself, and
another the straight line A'B. In order to determine simultane-
ously the coordina~es of the ruler's ~nds in terms of the frame K.
,~\
,l·
~ ~-~
'1 p
0
A'
BA
i
X
BL
0,0 :r:, ;r
Fig. 4.8. The case is illustra- Fig. 4.9. The geometric illustration of the relati-
ted when a ruler is at rest in vity of time intervals between two events Let a
the frarne K'. The world lines clock be at rest in the frame K' and be located
oi its ends are straight at the origin of coordinates 0. Its world line
. nes parallel to Qy' (the Ot' coincides with the Ot' axis. The readmg of this
~ ;is itself and the straight clock at the world point B' differs by a unit from
iine passing through B). The its readin~ at the point 0'. But in the frame K
ruler's length in K is deter-
mined by the intersection or ~e <r;i~~ o~ i~h~i~~~~n~~~!i;~\h f~ee w~~ ~~~~~
these world lines with the x wtlh the P.Oint B') at which the clock (located
axis (I = 0) and proves to be at this point and at rest in the frame K) will
equal to OB But OB<OA= indicate the lime determined by the section OB
=I, and we obtain the same relative to the readi(f. of another clock from K
:~~t~B a~IIQ,,g?~t I. ·,!~i!s i~~ije~r~hatththefifi~=
result: the ruler's length
ls the greatest in the frame
where Ute ruler LS o~t rest. interval, during which the clock from the frame
K' moves, is less in terms of K' than in terms
of K.
the world lines of the ruler's ends are to be intersected by any

straight line -r =canst. It is more convenient for us to draw the
straight line -r = 0. From Fig. 4.8 it is seen that OB < OA = I.
Let us dwell on a geometric illustration of the relativity of time
intervals (Fig. 4.9). Let a clock be at rest at the origin of the co-
ordinate system K'. Its world line will be the Or.' axis. At the
moment of time t = 0 a moving clock was at the origin of the
coordinate system K where we had its reading compared agains-t
one of the clocks of the system K' located at this point.
As before, we suppose that the clocks from both systems show
the time. t = 0 and t' = 0 at the moment when 0 and 0' coin-
FoUr-Dimensional Spate-Time 131
cide. Then the sections 08 and 0'8' correspond to the time read-
ings of the clocks of the systems K and K'.
Fig. 4.10. The same as in the preceding figure. only r~
now the clock is at rest at the origin oi the frame
K. The world line oi the clock is the 0-r axis. AI a"/_.
~~ee ~~i~~ts8 Jy\~~
c~~~h:i!~r~f:~~a\~n: ~~~~lelmt~ B / B'
~~e s~ui~!~~~~ ~~t~i~ffis t~g~~~t t~e th~i'£~a~e \\k~~ t

II is clear that 08" > 08' = l, i.e. a motionless
clock will register the lesser time interval as com-
pared to a moving clock. 0, 0 , .r
At the world point 8' the reading of the moving clock will in~
crease by unity compared to that at the point 0'. But the point 8'
in the frame K is simultaneous with all events located at the
straight line T = canst passing through the
point 8'. In particular. the world line of the Wortdllneof
ctockQ.'
clock located at the point x 1 and at rest in K .,
passes exactly through the point 8'. This
means that H the moving clock of K' regis-
ters the proper-time interval 0'8', the time
interval registered by the two clocks of K
(located at the points 0 and xl) is equal to
08. It is seen in the figure that the time in-
terval registered by the clock of K' is less,
because 0'8' = I and 08 > I.
And if the clock is at rest in the system K.
it will register a time unit at the world point
8 (Fig. 4.10) which is simultaneous with the
point 8" in the system K' (08" is the read-
ing of the clock of the system K' which an 0 Qt a2 QJ r
observer from the system K will get at thl' Fig. 4.11. The differ-
point 8"). The point 8" is obtained as a re- ence between the
suit of the intersection of the straight line :gj~ft~~/i1h! c~r:C
parallel to the x' axis and passing through nate time registered by
~e J~Jn~ ~; '~~~s:~~e~~~, at~:- :o~~in~8;;o~ ~:en~c: 1 f~~~~~ r~~=~~r:~
will again register the longer time interval to which the object
than two motionless clocks .• The length of the . moves
world line afc (in the pseudo-Euclidean plane!) is directly associat-
ed with the proper time of the object, being just proportional to it:
ds = c dt:. Hence, the length of the "·orld line arc enables us to
conjecture about the proper time that was registered by the clock
fixed to the particle. It should be remembered, however, 'that one
should be careful in the evaluation of the arc length in the pseudo-
..
Euclidean plane. The "risk" is clearly_ visible f.rom the fact that th~
132 Special Theory of Relatloltg
"arc length" lor two points located at the finite spatial distance
from each other may turn out to be equal to zero. Think for your-
self why in the foregoing reasoning we obtained the correct results
on the basis of geometry. Naturally, the peculiarities of the pseudo·
Euclidean plane interfere with the interpretation of the results.
As an example let us consider the difference between the proper
time and the coordinate one, i.e. the time registered by the clock
of the system relative to which an object moves. Let the clock Q'
be at rest at the origin of the system K' and its world line be OA 3
(Fig. 4.11). As usual, the coinciding clocks at 0 and 0' indicate
t=O,t'=O.
The world lines of all clocks Q at rest in K are represented by
straight lines parallel to the "t" axis. At the world points A~. A2,
A3, ••. one can check the clock Q' against the clocks Q~o Q2, Q3, •••
synchronized in K and indicating the common, unified for K. time
at any world point A 1, A2, A3, ••.• Its value at the world point A 1
is equal to the length of the world line Q 1A 1• For the clock Q',
however, the length of the world line connecting 0' and A1 is
equal to OA 1• But OA~=Q 1 A?-OQi, from where it is clear
that OA 1 < OtA 1• This implies that the clock Q' checked against
the clocks Q~o Q2, ••• at rC'st in the frame K is slow compared to
thf' docks Q~. Q2, ••• synchronized in the frame K.
B
fig. 4.12. The world lines of two "twins". The world
line of the "traveller" is the broken line DAB, that of
the "stay-at-home" the straight line DB The "travel·
ler" undergoes an acceleration when he reverses his mo·
tion direction at the point 8 and thereby gets into a
non-inertial reference frame for this time interval. The
leongth of the world line of an object determines its
proper-time interval. The proper-time interval is obvi·
ously less for the "traveller" than for the "stay-a!•
home" (see the pseudo-Pythagorean theorem in Fig. 4.5),
Finally, let two persons ("twins") be at the point 0 at first.

Then one of them ("a traveller") moves uniformly and rectili·
nearly except for a short time interval needed to reverse the veloc-
ity direction before returning to the initial point 0. The other
"twin" remains at the point 0 all the time. It is seen from Fig. 4.12
that the world line of the "traveller" OAB is longer than that of
the "stay-at-home". However, in accordance with the pseudo-Py-
thagorean theorem this means that the "traveller" spent less of
his local time than the "stay-at-home" did. We shall come bacl<
again to this problem in Chapter 8.
CHAPTER 5
RELATIVISTIC MECHANICS
OF A PARTICLE
The Einstein principle of relativity is valid provided that the

basic laws of physics are formulated similarly throughout all
inertial frames and differ only by the notation of variables asso-
ciated with the given reference frame. In terms of physics the last
statement implies that in every IFR measurements are carried out
by means of instruments which are at rest in that frame. But th!!
transformation of the coordinates of an event on transition from
one IFR to another is the Lorentz transformation. Consequently,
the equations of mechanics, for example, have to retain their ap-
pearance (in the above-mentioned sense) in any IFR This condi-
tion is automatically fulfilled if the equations of mechanics are
put down in a four-dimensional vector form. Indeed, in this case
the transformation law of the left-hand and right-hand sides of
such an equation is known, and it does not change the appearance
of the equation. When put down in the vector (or the more general
tensor) form, the equation is said to be written in the covariant
form.
The Newtonian equation relating forces and accelerations is
covariant relative to the Galilean transformation, although it is
not covariant relative to the Lorentz transformation. However,
the Lorentz transformation follows unambiguously from the Ein-
stein postulates which are for certain confirmed experimentally.
In order to satisfy the principal Einstein postulate on the equiv-
alence of inertial frames of reference, one has to ensure the co-
variance of the equations of mechanics under the reLativistic trans-
formation of coordinates and time, i.e. the Lorentz transformation.
The required equations of mechanics ai-e fairly easy to write u.5·
ing the STR's four-dimensional geometric concept. We shall pro-
ceed in just this manner.
Certainly, the development oi science does not cancel previously
known ("correct") laws, but only sets limits to their application.
There is always some conformity between various theories de-
scribing one and the same group of phenom~na in extreme cases.
The majority of equations of classical mechanics correspond to the
I'Xtreme cases of relativistic equations with ~-+ 0. In other words,
134 Special Theory of RetafirJily
classical mechanics is the extreme case of relativistic mechanics

corresponding to the velocities which are small in comparison
with that of light. Nevertheless, relativistic mechanics brings for·
ward such conclusions that could not even be alluded to in the
framework of classical mechanics (for example, the existence of
the rest energy of an object).
§ 5.1. A 4·velodty and 4-acceleration. To write down the rela·
tions between physical quantities in space-time, we must con·
struct the required 4-vectors. While doing this, we should re-
member that in the extreme case of small velocities the Lorentz
transformation turns into the Galilean one, the relativity of time
intervals and lengths does not manifest itself any more, and the
Newtonian equations correspond to the Galilean principle of rela-
ti\'ity provided it describes the transition from one IFR to an-
other. In this extreme case time and space are not related, and we
can utilize conventional three-dimensional quantities. Therefore,
while composing four-dimensional quantities, we shall always
try to make their three (spatial) components resemble the corres-
ponding three-dimensional quantities. In the extreme case of small
velocities (p- 0) the three components of four-dimensional
quantities must turn into the conventional mechanical quanti-
ties.
We shall compose a 4-velocity and a 4-acceleration in the same
way as we do the corresponding quantities in the three-dimension-
al space where a particle position is specified by the three-di-
mensional radius vector r and the 3-velocity is determined as th~
derivative of the radius vector with respect to time, dr/dt. Thl!
4-velocity cannot, however, be defined as a derivative of the 4-ra-
dius vector R with respect to time. To get the 4-vector velocity,
we have to divide the 4-vector of the increment dR by a scalar (an
invariant of the Lorentz transformation). But neither time nor
its differential is a scalar.
One can take the interval or the proper time of a particle (seP
§ 3.3) as an inYariant time-dependent quantity. We shall intro-
duce once more the proper-time concept, having associated it with
the interval between events. We make use of the fact that the
motion of a particle in the 3-space is a continuous sequence of
events consisting in a particle occupying a definite point in space
at a given moment of time. Let the coordinates of a particle :n
the frame K change by dx, dy, dz during the time dt, and its dio;-
placement be equal to d!=.,Yd:il+dy 2 +dz'l· Consider the in·
stantaneous inertial frame K' co-moving with the particle, i.~.
the frame moving at the constant velocity V equal to the in-
stantaneous velocity of the particle. In the frame K' the coordi-
nates of the particle do not change during the infinitesimal timl!
Relatluistic Mechanics of a Particle 136
interval dt': dx' = dy' = dz' = 0. The interval between events is

invariant, so that
ds2=c2dt2-dx2- dy2- dz2=c2dt'2,
In the frame K' the time interval dt' is the proper-time interval.
In this chapter we shall designate it by d• (we shall not use the
designation • = ct as we did in previous chapters). From the
foregoing equation we have
dT = 7 = ~ll=~dx~':i+,~~~::~,+[•!l;lz[a dt =
~ ,.,;-,----·;.-(,.-.""•:""')"2 dt =,.,; 1- ?- dt.
We have obtained the familiar result (§ 3.3) and demonstrated
the invariance of the proper time (d• = ds/c). Here are the equa·
tions to be needed later:
dT~dtfy, ds~CdT, y~(J -~2 )-'1', ~(t)~v(I)/C. (5.1)
We see that the proper time of a particle is registered by a clock
of an instantaneous co-moving IFR. But these instantaneous
co-moving IFRs change during a finite time interval in the case
of a particle movi1;1g with an acceleration. The final proper time
of such a particle iS defined as the overall time registered by many
IFRs. As a matter of principle, the clock should not be rigidly
linked with the particle, since any acceleration affects the clock
rate. The proper time can be registered by the clock fixed rigidly
to the particle only if the acceleration to which this particle is
subjected does not affect the clock rate. The "proper time" can,
however, be readily obtained from the time registered by the clocl<
of the frame K (relative to which the particle moves) provided
fhat the time dependence of the particle velocity, i.e. v = v (t), is
known:
It is seen from the last equation and equations (5.1) that the
coordinate time, that is the time registered by all clocks of K. is
a function of the proper time •· From the equation ds = c dT one
can see that in addition to the proper time dT one may equally
use the interval ds, with all equations differing by various powers
of the invariant factor c.
Now let. us introduce the 4-vector velocity
Since d't 1s an invariant and dR a vector, V is also a vector,

no doubt. Let us disclose a three-dimensional connotation of the
first three components of (5.2) in the notation of Eq. (4.7a):
uu = ~~u =v d~a = VV 11 (a= I, 2, 3), (5.3)
where vrz are the components of the Conventional 3-velocity. There-

fore, the first three components of the 4-velocity are those of the
conventional 3-velocity multiplied by the factor y depending on
the absolute value of the particle velocity. The fourth component
is to be found separately:
dx, d(icf) •
u~=--;rr=v-,,-=lCy. (5.4)
In accordance with the notation of Eq. (4.7) we have

uo= d;~o) =v d~~tl =vc.
Similarly to Eq. (4.7a, b) one can write
v( u,VV..:
u2
YVv
UJ
Ytlz
u4 ) ·
icy '
()
a
I v(uoyc VV~: VVv u3)
ul u2
VOz. •
(b) (5.5)
When ~-+ 0, i.e. when the velocity of an object v « c, the factor

y ~ I, and the first three components of the 4-velocity of (5.5a)
as well as the last three components of Eq. (5.5b) coincide with
the conventional velocity. Of special interest is the fourth com-
ponent of (5.5a) and the zeroth one of (5.5b) for the 4-velocity.
They are different from zero even when a particle is at rest (if
v = 0, y = I and u 4 = ic, but u0 =c). The last result has the
obvious meaning: time cannot be stopped, it always flows withoLJt
interruption. Accordingly, there is no quiescence in the four-di-
mensional world (in the sense that V
=F 0). As to the "velocity
of the time flow", it is defined by the choice of time units, of
course.
The components of the 4-velocity can be also put down as fol-
lows:
V(yv. icy); . (a) I V(cv, vv). (b) (5.6)
The Square of the 4-vector is an invariant. It can be found from
Eqs. (4.1Ia) and (4.1Ib) respectively:
V: = v2v2 _ c2-f::::::! _ c2; V2 = (c2v2 _ v2v2> = c2.
The computation is easiest when made in the inherent rei-
erence frame of a particle at rest (v = 0). Then in (5.5a) only
Relativfstlc Mechanics of a Particle 137
u 4 = tc will differ from zero, and in (5.5b) only u 0 = c. Conse-

quently,
Y 2 =u~+ui+u~+u~=-c 2 ; (a) I Vil=urfl- u 12 -
-u22_u32=c2; (b) (5.7)

the squares of the 4-velocity in Eqs. (5.7a) and (5.7b) are op-
posite in sign due to the different determination of the int!rval
(see Chapter 4). When the appropriate determination of V2 !s
chosen, however, this sign does not change, and it follows that
v < c in all cases.
As soon as the velocity in 4-space is written down in the form
of a 4-vector, the transformation equations for the velocity com-
ponents on transition from one inertial frame to another can b!
obtained at once. Let the components of the 4-velocity V
(u 1, U2, u3, u~) be specified in the frame K. In accordance with
Eq. (4.10a) we shall obtain in the frame K'
u;=r(u 1 +iBu4), ~=":!· u;=u3, u;=r(u~-iBu 1 ), (5.8)
but the 4-velocilies have the components ii(yv, icy), Vi(y'v', icy').
Having substituted them in Eq. (5.8), we get
y'v: = r (Yvx- yV), y'v~ = yv 11 , y'v~ = Yr.Jz,
icv' ~ r (icy- iBvo,). (5.9)
It follows from the last equation of (5.9) that
V I
(5.10)
yr= r(1-f,-vx)'
Substituting this expression in the first three equations (5.9),
v~=f,f(r.Jx-V). v~=fr.J 11 , r.J~=fvz•
we shall obtain the equations for the velocity components in K'
which were derived in Chapter 3 from the Lorentz transforma-
tion.
Noh•, incidentally, that if in place of Eq. (5.8) of transition
from K to K' one uses the equations for the reverse transition
from K' to K, the following equation is obtained
f~r(1 +.);. o;). (5.10')
instead of Eq. (5.10). This way we obtain the value of y/y' in
terms of the velocity components in the frame K'. From Eq. (5.10').
188 Special Theory of RelatirJitg
follows Eq. (3.41} derived otherwise here:

~0=B'
It follows from Eq. (5.10) that if a particle is at rest inK (tr=O},

then y' = f; this result is obvious, because a particle which is at
rest inK moves at the velocity -V relative to K'.
The same result is obtained if one makes use of Eq. (5.6b) and
the transformation equation (4.10b). We suggest that the reader
do it himself. Our result is obvious: the spatial components of the
4-ve\ocity determine the transformation of the conventional 3-ve-
\ocity.
Now we are to define the 4-acce\eration which we shall also
construct as a 4-vector:
(5.11)
or expressed via components:

dul d 2 :c 1
Wt="'"'(lf="'"'(ki""; (a)
I w'=---;tf=---;w:-·
du 1 d'x 1
(b) (5.12)
Below we shall write out a few formulae dealing with accelera-

tion, using the notation of Eq. (4.7a). They will be needed only
in special cases. The four-dimensional acceleration components
can be expressed by means of the three-dimensional components
of the vectors v and V. We get
w~~=ft<vv~~)-*=yv~~*+Y2 d:t =q+ cz~~~~2J2. (5.13)

because, as it is easy to verify,
(5.14)
and dt/d-r; = y The fourth component of the acceleration is
w4 =ft(icy)-*=icy~=f dd~2 =icy 4 ~~ =7 (I ~V~ 2 ) 2 • (5.15)
In the case of the uniform motion (ti = 0) all four acceleration

components turn into zero. In the reference frame in which a par-
ticle is at rest
wll=V~, wg=V", ~=ti 2 , uP.=O, (5.16)
i.e. the three spatial components of the 4-acce\eration coincide
with the conventional three-dimensional components of the acccl~
Relalio(siU: J11echan(cs of a ParUcte I~
r;_ti(5.i6)h;~~tthe lime component turns into zero. It is seen _from
~=wr=v2>0.
Due to the invariance of the square of the 4-vector norm (see Ap-
pendix I,§ I) one may regard the 4-vector acceleration as a space-
like vector (see the definition of the interval (Eq. (4.5)).
Let us write out the components of the 4-vector acceleration .;
in the notation of Eq. ( 4.7):
; ( Y ft<vv), f dl) ~( Y~ + y4pPv.-/;;; .!!J-). (a)
(5.17)
; (y ft (cy), y ft (yv)) ~ ( -;'c ~~1K , y ft (yo)). (b)
The particle energy 8 introduced here will be defined later on

(see Eq. (5.32) below). Using Eqs. (5.15) and (5.17), one can
easily obtain
~=«7=v 6 (V2 -*(vVF)>o. (5.18)
Now let us write out the transformation equation for the 3-ac-
celeralion (ti = dv/dt, il = dv'/dt') on transition from one IFR to
another. The Galilean transformation leaves the 3-acceleration of
a particle invariable. The Lorentz transformation changes the
3-acceleration components. The simplest way to derive the trans-
formation equation for the 3-acceleration components is as fol-
lows. Regarding v_.. and v~ as functions of t and t' respectively,
and taking into account the relationship between t and t' (E<J..
(2.16)), and, finally, designating (v_..jc)=P..,. (v:Jc)=P: etc., we
shall obtain from Eq. (3.26)
dtl~ do; - n (~; do: - ~~ du;)
dv..,= r2(!+B~~Y, dvu r(t+o,;)'
and in much the same way, dvz. Having divided the left-hand and
right-hand sides of these equations by the left-hand and right-hand
dt
sides of the equation = r ( dt' + -fdx'),
respectively, we get
"· ~ r"(t ~ .,;) v;,

ti: + B (~:v:- ~:v:)
r'{t +"';)'
Of course, the same result will +be obtained via the transforma-
tion of the 4-vector acceleration w. In the frames K and K' it is
140 Special Theory of Relatlrllty
.;,asies~ to put it down in the form fZ (y 2V+yo y, icy y),

w'(y' 2v' + y'v'y', icy'Y').
From the transformation equations for the 4-vector components
w1 = f (w:- iBw~). w4 = r (w~ iBw;) +
we get
yti"' + v:cV = r-f (v'o: +fv~ + VV'),
v~ rfw +B(v'P~+ v'P;)].
Using Eq. (5.10'), ti ... will be found from these relations. Sub-
stituting w2 = w2 and w3 = w:l in the transformation equations,
we obtain the equations for transformation of fly and Oz.
The transformation equation for the 3-acceleration components
involves the velocity of a particle. But the 3-acceleration appears
only when the velocity varies. Consequently, even when the 3-ac-
celeration 1s constant in one IFR. it varies with time in all other
frames: in relativistic mechanics a uniformly accelerated motion
in one IFR is not such in all others.
§ 5.2. A 4-force and a four-dimensional equation of motion. Here
are some three-dimensional classical relations to be referred to
later on quite often. In classical mechanics a mass of a particle
is treated as a constant. We shall designate it by m. The second
law of Newton is put down as follows in classical mechanics:
f,!mv)~F (5.19a)
l!ft-=P. (5.19b)
where P is a three-dimensional vector of a conventional force;

the quantity p = mv is called a classical momentum of a par-
ticle.
Multiplying the left-hand and right-hand sides of Eq. (5.19a)
by v dt, we derive by means of simple transformations the coral·
lary of the second law of Newton:
d (mv2/2) ~ Fv dt. (5.20)
The right-hand side of Eq. (5.20) represents the work accom-
plished by the force F; in accordance with the energy conserva·
tion law the left-hand side must incorporate the change of energy.
Hence, the energy of a particle can be defined as T = mv2f2 with
an accuracy of a constant addendum. Here the addendum is
adopted to be equal to zero, so that the particle at rest possesses
no energy. Consequently, the energy mv 2/2 is associated only with
Relatlvistic Mechanics of a Particle 141
the motion of the particle; accordingly, it is appropriately called

the kinetic energy (the "motion energy"). It should be pointed out
that if we assumed that a motionless object possesses the energy
go the "total" energy of a moving object would beg= T 8 0• +
The constant go can be interpreted as a permanent potential, or
internal, energy. But in classical mechanics there is no reason to
do this, and g 0 can be treated as an arbitrary constant. Conse-
quently, in classical mechanics the "total" energy of a free object
may have either sign in principle (depending on the sign of th.:!
constant g 0 ). Customarily g 0 = 0, and the total energy of a free
object coincides with the kinetic one.
Suppose now that a particle is in a potential field, i.e. the force
acting on a particle can be expressed as F =-grad U, where
U(x, y, z) is a potential energy. Since v dt = dr and grad U dr =
= dU, Eq. (5.20) will take the form
d (mo'/2) ~- dU.
from where follows the important law of classical ml;!chanics, thaL
is the law of conservation of the total energy:
f,IT+U)~O;
+
in other words, T U = canst.
Now we can pass over to the definition of a 4-momentum of a
particle P. As in the case of the 3-momentum (p = mv) we spec1fy
the 4-momentum as the product of the invariant (scalar) mass m
by the 4-velocity V, P
so that = m Therefore V.
P(mvv. imyc); (a) I P(mvc. myv). (b) (5.21)
As it will be clear later, the invariant mass m is expedientlv
called a rest mass. Analogously with Eq. (5.19) one may suppos'e
that the four-dimensional equation of motion has the form
iP ... IS.22)
-;rr=F
or in components
m ~:1 =6,. (5.23)
where the differentiation is naturally accomplished with respect

to the invariant proper time d"t (otherwise a vector relation will
not be obtained), while the right part of the equation contains the
4-vector force f('U~o 'iJ2, 'jJ3, 'i5 4), whose components are still to be
determined.
Special Theory of Relat(vitg
\Ve shall recall once more (see § 4.3) why it is important to

haVe the motion equation in a four-dimensional vector form (Eq.
l5.22)). The point is that in accordance with the first postulate
of' Einstein all basic laws of physics must have the same form in
3.11 IFRs. In mathematical terms this means that equations descrih-
frig physical laws have to be represented in the covariant form
with r·espect to the Lorentz transformation. Equations are said
to be written down in the covariant form if their left-hand and
right-hand sides change alike under the Lorentz transformation.
But this implies that the left-hand and right-hand sides must be
correspondingly either scalar (invariant) quantities or 4-vectors,
or tensors of the same rank (to be dealt with in Chapter 6). This
is enough to ensure the invariance of relations presented in this
form on transition from one IFR to another. Having put down
the motion equations in the vector form (Eq. (5.22)). we ensured
the covariance of this equation under the Lorentz transformation,
i.e. the universal character of the ~instein principle of relativity.
The components of the vector dP/d-c are familiar to us, for we
know the components of i\itd-c from Eq. (5.17a, b) and m is th~
invariant:
*(
4 ( ft.
Y
y
(myv), iy -ft. (myc)),
ft. (myc), y ft. (myv)).
(a)
(b)
We have denoted the components of the 4-vector F by the gothic

+
(5.24)
letter \J supplied by indices, i.e. FUJ 1, ij2, \}3, \}4) or 'F(ijo, \} 1, ij2, ij3).
Equating 4-vectors, we equate their components. The first three
components of Eq. (5.24a) and the last three components of Eq.
l5.24b) are obtained as follows (a= I, 2, 3):
v f, (myv,)- ~,; (a) I v f, (myv,)- )1•. (b) (5.25)
Now let us determine the first three components of the 4·force

~':l' Obviously, they are proportional to the 3-force components
~~~~o~na:heeq~i~W~~r~t"~~~~o~.~f0o~! ~~:~ntso tt~t :oanc;e~~i~~~lc~~:

finition of force and supposes, as before, that a "force determines
a change of momentum", one should write
~~~=vFa, ~~~=vFu,
\Vhere Fa are the components of a conventional three-dimensional

force. Having substituted the expressions for Sa into the right-
Relaliulstlc Mechanics of ti Particle 143
hand side of Eq. (5.25a), we obtain
f,-<mvva) =Fa (a= l, 2, 3),
~~~yh~~~~Ji:a~~t~~~f~r :~ha~~ t~:s~eia1~~~~o~~u~yobtaci~:Js[uo~~~J.

we get the motion equation in a vector form:
ft-<mvo)=P. (5.26)
Comparing Eq. (5.26) with the non-relativistic motion equatioa

(5.19), we notice that they differ only in the definition of a mo-
mentum. The relativistic (three-dimensional) momentum is rep-
resented by the quantity
p=mvv; (5.27)
in this case Eq. (5.26) resembles Eq. (5.19) in outer appearance
Thus, the three spatial components of Eq. (5 23) have ytelded
the second law of Newton in a relativistic form. However, the
meaning of the fourth (or zeroth) relation is yet to be cleared up.
In order to do this, ~ 4 (or ij0 ) should be known. But it turns out
that having the three components of the 4-force determined, we
thereby get the fourth component also determined. One can make
sure of this in the following wny. Differentiating Eq. (5.7a, b)
with respect to-r, we get
Ut ~~~ +~ ~~2 +u3 ~~1 +u4 ~~~ =0, (a)

(5.28)
uO ~u: - ul ~~~ - u2 ~~2 - u1 ~~a = 0. (b)
But according to Eq. (5.23) du,/d-r; = ij,Jm, with the first three
components of ~ being determined by Eq. (5.26) and the compo-
nents u, by Eq. (5.5). Hence, Eq. (5.28a), for example, can he
rewritten in the following form:
vv.:c= v~~ +vv11 v~11 +vvz~+icy~=O,

whence~~ (and similarly ij 0 ) can be derived at once:
~. ~ 4 (Pv); (a) I ~' ~ f (Pv). (b) (5.29)
We give the reader a chance to derive Eq. (5.29b) by himself..

144 Speckll Theory of Relatlt~llg
Thus, we have found the components of the 4-force F referred

to as a Minkowski force:
+(~' \l, \l, \l, )
F vF,vF,vF,if(Fv) ~(vF,if(Fo)), (a)
(5.30)
F(\l"f<Pv) \l' \i' \1' )~c~(Fv), vF). (b)
yFx yF 11 yF 11 c
To clear up the meaning of the fourth relation of Eq. (5.23) in

terms of Eq. (5.5a), or the zeroth one in terms of Eq. (5.5b), the
corresponding components in Eqs. (5.24) and (5.30) should be
equated:
iy ft. (mvc) = i? (Fv),
or otherwise
-Jr (myc = Pv.
2) (5.31)
Here we can go over the same reasoning which was evolved in

connection with Eq. (5.20). The right-hand side of Eq. (5.31) rep-
resents the work performed by the force; the left-hand side must
contain the energy change. Let us define the total energy of a
free relativistic particle as
8 = mc 2 y = mc2 (I - ~ 2 )-''•; (5.32)
~~~ar ~~~~fty wo7e~~e v p~~t}~~~- ai~s~~u~~~Ja~~e p~!n!~~ ~~~e~hd~:nf~;

energy of the particle is determined from Eq. (5.31) with an ac-
curacy of a constant value. Eq. (5.32) means that a motionless
particle (v = 0, ~ = 0) possesses the energy ~ 0 = mcz. Such ..
\'alue for the constant is not chosen at will, but comes from the
limit transition to the classical velocity summation formula.
We shall postpone the discussion of the relativistic motion equa-
tion (5.26) and the relativistic energy relation (Eq. (5.32)) till
§§ 5.3 and 5.4. In the meantime we shall dwell on the transfor-
mation of the 4-force and the consequences following from it.
Let us write out the force transformation law (in terms of Eq.
(5.30a))o
'151 = r m~ + iB64), ~~ = ~2. ~3 = g3, ~.; = r (~4 - tB61). (5.33)
We shall begin with a simple case. Let the three-dimensional
force F act on a particle which is at rest in the frame K0 . Then in
accordance with Eq. (5.30a) Fo
(f'O, 0). From Eq. (5.33) we ob·
Relalivisfic Mechanics of a Parflcle 145
tain
y'F~=fF~. y'F~=~. y'F~=F~. y'F'v'=-I'VF~.
In the considered case a particle moves relative to K' at the
velocity of the reference frame 1(0, i.e. at the velocity -V. Cons-:--
quently, y' = r, and we obtain the component transformation for-
mula for the force and the work accomplished by this force:
F;~F:, F; ~ F:o/1- H', F;~F',o,/1- B', F'v ~- VF',. (5.34)
lt is seen from the first three relations (5.34) that the force
components parallel to the relative motion velocity remain in-
variable. The force components normal to the relative motion ve-
locity change. It is easy to find out the meaning of the last relation
of (5.34). If the particle was at rest in the frame K0 , it moves at
the velocity -V in the frame K'. The work is performed only by
the force component Fx (all other components being normal ~o
the motion direction). The power developed by the force F~ in
the frame K0 is equal to -F~V which corresponds to the result
that we obtained.
From Eqs. (5.34) one sees that in non-relativistic case, whe11
n « I, the three-dimensional force does not change on transition
from one IFR to another. This fact wholly agrees with our intu-
itive ideas of the force invariance in any reference frame. However,
in the presentation of the STR and, in particular, in the derivation
of certain relations of the STR, when making use of the transfor-
mation of forces, one has to emphasize first of all the variation
of force components on transition from one IFR to another.
In a general case, using Eq. (5.30a), we obtain from Eq. (5.33)
y'F:=r[yF 11 -~y{Fv)]. y'f;=yf11, y'F:=yf;;,
y'(F'o')~ l']y(Fo)- VyF,].
Rewriting the last equations in the form
f~=fr[F:c--!}-(Fv)], F~='f-F 11 , F~=f,F 2 ,
(F'o'l~f.r!Fo- VF,] (5.35)
and taking into account Eq. (5.10), we obtain finally
F~= FII.Y~ -H2.

1-crvx
P'v' = (Pv) - 1tF"' (5.36)

1-crv.c
146 _ _ _ _ ____:cSP::"::''::'.::Th:::':::''"-Y2
aT_:R::''::'':::'':::"Y'------~
It is seen from the transformation formulae (Eq. (5.36)) for a

4-force that if there is no three-dimensional force in some IFR.
such a force cannot appear in any other IFR. Thus, forces trans-
form on transition from one IFR to another, but they never appear
or disappear.
A validity of the Jaw of inertia in all IFRs follows from this
immediately. If in one IFR no forces act on an object and it
moves due to inertia (v =canst), the same situation will he
observed in any other IFR (see Eqs. (5.27) and (5.32)).
§ 5.3. A three-dimensional relativistic equation of motion of a
particle (the second law of Newton in a relativistic form). Hav-
ing written down the motion equation in a 4-vector form (Eq.
(5.23)) and determined the components of the 4-force (the Min-
kowski force), we satisfied the principle of relativity for one thing,
and, for another, obtained the four components of the motion
tquation. The three components provided us with the "motion
equation" per se in a three-dimensional form (Eq. (5.27)), while
the fourth component permitted us to determine the relativistic
expression for energy (Eq. (5.32)). Eq. (5.27) was derived on the
assumption that equations of dynamics must retain their ap-
pearance in all IFRs, i.e. they must be covariant with respect to
the Lorentz transformation. However, even without passing from
one IFR to another, we are aware that the exact equation of
motion is represented by Eq. (5.26) and not by Eq. (5.19). Let
us write out these two equations side by side and clarify the
difference between them:
-#, (mo) ~ F; (a) I -#, (mvv) ~F. (b) (5.37)
First of all, it is clear that when ~ = v/c « I, i.e. y ~ I, Eq.

(5.37b) passes into Eq. (5.37a). This means that classical me-
chanics is the extreme case of relativistic mechanics when par-
ticles move at non-relativistic velocities. Moreover, to satisf\" the
Galilean principle of relativity in classical mechanics, the "Ga:i-
lean transformation has to be valid, requiring B « I (see § 2.7),
i.e. the relative velocity of the considered reference frames mu::.t
be non-relativistic as well.
Sometimes, from the comparison of Eqs. (5.37a) and (5.37b)
one may conclude that the difference between them consists in the-
fact that in Eq. (5.37b) mass depends on velocity, so that taking
my for a relativistic mass, we obtain a classical equation. Now
we shall see that the things are much more complicated, and in
Supplement IV we shall discuss why there is no sense in introduc-
infna ~:Jeern~~n~~~p~r~ssE~~-v(~~~~) and (5.37b), the left-hand
side of (5.37b) should be rewritten using Eq. (5.32) and the fol-
Relatlvisltc Mechanics of a Particfe 147
lowing identity (see also Eq. (5.31)):
f,<mvv)=f,(~v )=7~ +.fr!frr-=Tz<Fv)+mv 4lf-.

Rearranging the terms, one may rewrite Eqs. (5.37a) and (5.37b)
as follows:
m!fiT-=F, (a)
m*-+(F-{;(Fo)J-+[F-P(Fp)). (b) (5.38)
It is noteworthy that in an IFR co-moving with a particle (v = 0)

Eq. (5.38a) coincides with Eq. (5.38b). It is immediately seen
from these relations that the principal difference of the relativ-
istic law of dynamics, Eq. (5.37b), from the classical one, Eq.
(5.37a), lies in the fact that in the former the direction of a
3-acceleration does not coincide, generally speaking, with that of
a force. Consequently, the straightforward comparison of the force
and acceleration components, readily carried out for the case of
Eq. (5.37a), is quite impossible here. It is seen from Eq. (5.381J)
that there are still two cases for which an acceleration and a force
are oriented along the same direction and the definitions of mass
in Eqs. (5.37a) and (5.37b) can be directly intercompared.
(a) Let the force acting on a particle be always normal to its
velocity, i.e. F .l v. Then from Eq. (5.38b) the motion equation
is immediately obtained in the following form:
mv!fli--=F, (5.39)
where, according to Eqs. (5.31) and (5.32), y = const. This case

is realized when a charged particle moves in a constant magnetic
field. The Lorentz force F = e [vB] is so directed that Fv = 0
(always). It may be said that the motion of a relativistic particle
in a constant magnetic field proceeds in accordance with the clas-
sical motion equation (5.19), but with some effective (but con-
stant) mass my. This is valid, however, only for a special case,
when the condition Fv = 0 is satisfied. To make sure, let us con-
sider another case.
(b) Let a force acting on a particle be always directed along
its velocity. This implies, naturally, the rectilinear motion of a
particle (in case of a definite choice of the initial velocity).
A simple example of such a motion is provided by a charged par-
ticle moving in a plane capacitor when the initial velocity is
directed along the electric field. If F II v, then v (Fv) = F(vv) =
148 Special Theory of Relatir.lily
= Fv2, and from Eq. (5.38b) we obtain the following motion equd-
tion:
mv'4Jf-=F,
where y is a variable quantity.
Thus, in the two special cases permitting of an intercomparison
of Eqs. (5.38a) and (5.38b) we get a different dependence of mass
on velocity; this indicates that there is no universal dependen;:e
of mass on velocity. It is sound practice to use the invariant rest
mass (see Supplement IV).
As in classical mechanics, the equation of dynamics can al<>o
be written for the case when the rest mass of a particle varies
due to the exchange of energy and momentum with _,}he en-
vironment. If a particle loses the 4-vector momentum 11(111 ) =
= ( yll, f
yCD) per unit time due to convection, Eq. (5.22) should
be replaced by
1-T=(mu,) =!'it+ n,, (5.40)
wher~2nly ij,is a genuine mechanical force satisfying the condi-

tion FV = 0. Writing out the components of Eq. (5.40), we get
4/i-=F+l1. !!ff=Fv+CD.
Here II and CD are the momentum and the energy delivered to a
particle through convection per unit time. Having composed the
product
iili=v 2 (11v-¢)=-~D(I,
we see that the quantity CD0 is the energy delivery rate in the frame
in which the parhc\e is at rest; it is t>qual to the rate of change of
the particle's rest energy. Indeed, differentiating Eq. (5.40), we
obtain
(5.41)
Multiplying the left-hand and right-hand sides of Eq. (5.41) by u 1

and taking into account that ii(d\i/d-t) = 0 and FV
= 0, we get
(l}O;;;;,-ll,u,=c2 ~~ = dd~o.
Relatlulstic Mechanics af a Particle ...
Assuming that the genuine mechanical force must satisfy the
condition
the following quantity should be treated as a mechanical force

(see Eq. (5.41))'
m ~:1 ='J5,+llt-u 1 ~~ ='t5 1 +ll 1 -;u,=~ 1 +ll 1 +u 1 n~:k,

in the case of the convective transfer of momentum and energy.
When the genuine mechanical force is absent, the mechanic<~!
"reactive" force should be taken into consideration:
R,=n,+u, n:2"k ·
This force satisfies the condition RV = 0.

In a particular case the momentum n can be attained not by
mechanical means but, for example, via radiation or heat ex-
change between the particle and the environment. In the case of
a pure heat transfer the 4-momentum of heat transferred to th~·
particle during the time dr: is equal to 6Q,~n,dr:=(6p, i(6Q/c)).
Consequently, 6p =Ddt, 6Q = ctJdt are the momentum anrl
the energy of the heat transferred during the time dt. In this case
the 4-momentum of the transferred heat is a 4-vector.
§ 5.4. The relativistic expression for a particle's energy. Note
that Eq. (5.31) can be obtained not only as the fourth component
of Eq. (5.23), but directly from Eq. (5.37b) and exactly in the
same way as Eq. (5.20) follows from Eq. (5.19) in classical me-
chanics. To make sure, let us multiply scalarwise both the left-
hand and right·hand sides of Eq. (5.38b) by v. We obtain
mvV=~[Fv -*(Fv>]~7-Fv.
This is just Eq. (5.31) with Eq. (5.14) taken into consideration.
On the other hand, the fourth component of Eq. (5.23) turned ont
to yield the energy conservation law. The expression for energy
given by Eq. (5.32) differs essentially from the classical one. In
Newtonian mechanics the energy of a motionless free object (i.e.
an object possessing no potential energy) is assumed to be zero,
so that the kinetic energy is clearly defined as T = mv2/2. In re-
lativistic mechanics the total energy of a free particle is defin:!d
as 8 = mc 2y. We call this energy "total" because it includes the
energy of a resting object (the rest energy mc 2 ). But we mention-
ed earlier that Eq. (5.31) defines the energy with an accuracy of
a constant; consequently, having selected the appropriate value
150 Special Theory of Retativily
for it (by assuming 8o = - mc2 ), one could have regarded, as

in Newtonian mechanics, the energy of a resting object equal to
zero. But it is impossible to do so in the STR. In STR mechanics
one should not forget about the transformation rules for various
quanti~ies and the principle of correspondence with classical m·~
chanics: many classical and relativistic quantities must coincide
in the extreme case fl - 0. The Lorentz transformation is known
to turn into the Galilean transformation at small relative veloci-
tJes of reference frames (B- 0, r-+ I); at small velocities of par-
ticles (fl-+ 0) a three-dimensional relativistic momentum turns
into a three-dimensional classical one, myv- mv. Suppose we d~
t-ermined the total energy of a free relativistic particle in the form
+
lB = mc2y C; then in the extreme, case j:i-+ 0 we should obtain
+
lB = mc2 C. Let us examine now the transformation of the
4-momentum components on transition from one IFR to another. It
is performed in accordance with the following equations:
P!=r(P• +iBP4), P2=P2, PJ=P3, P~=I'(P4-iBP1). (5.42)
Substituting the values of the 4-momentum components P], P2, P3,
and P4 from Eq. (5.21), we obtain
p;=r(p,-{-<~'). p;=p,. p;=p,. <~''=l'(<~'+Vp,). (5.43)
where p,, p11 , P:~& and p~. p;. p~ are the components of the thre~
dimensional relativistic momentum p = myv. In the extreme case
corresponding to a transition to classical mechanics when B _,.. 0,
-13 _,.. 0 and p~ -+ mv;, p., _,.. mv.,, ~-+ mc 2, the first relation of
(5.43) would lead to mv~ = mv"'- mV- ~~ . But the latter equa·
tion must yield the classical law of velocity summation: v~ =
= u.,- V. It will indeed be the case if C=O. That is how we prove
the validity of Eq. (5.32). It should be noted that the principle of
.correspondence between the classical and the relativistic expres-
sions for energy is not valid only because in the framework of the
Newtonian mechanics one could not detect the existence of the rest
-energy, and the additive constant was chosen without allowing for
the rest energy (see below).
It is seen from Eq. (5.32) that the total energy does not turn
into zero even when the velocity of an object is equal to zero
(y = I at v = 0). The energy of a free particle in the frame in
"h•ch it is at rest is equal to mc2 and is called the rest energy
$ 0. Although we dealt with a particle until now, its elementary
character was not discussed. Therefore, all equations derived are
quite applicable to any complex object (system) composed of di·
verse components. Naturally, m will then represent the total mass
.of an object, and v the velocity of its motion as a whole. The
RelatlvisUc Mechanics of a Particle 151
equation 8 0 = mc2 is valid for any object that rests as a whole.

Consequently, the object's rest mass defines the total energy con-
fined in it regardless of the origin of this energy.
In classical mechanics the energy of a resting object may be-
either positive or negative: it is defined with an accuracy of a
constant value. In relativistic mechanics the energy of a free ob·
ject (or the energy of a closed system) is always positive and
related to the object's rest mass; the rest mass of an object defines
its rest energy. The inertia of an object proved to be a measure
of the object's energy. Whenever the object's energy changes by
M, the mass of this objects varies by tJ..m = M/c 2•
The question arises as to how such a high energy could remain
unnoticed, that is the rest energy confined within an object. In
~~~u~~eorr!:er~~ ac~~~~~~nc~if~i~ta!n~Y~~~~t i~02~o~r~~- ~shs~~!~~al~
however, as the part of energy that can be utilized. Although any
mass stores a huge amount of energy, its realization is far from
being easy. Only recently people have learned to employ the
atomic energy. Until a short while ago the rest energy had not
been realized (and, accordingly, mass had always been constant).
Since the realized energy always represents a difference of ener-
gies, the existence of the rest energy did not manifest itself in any
wasi-nce c2 is very large, a change of mass accompanying a change
of an object's energy is very small and cannot be experimentally
detected even though weighting was always one of the mo~t
precise kinds of measurement. For example, I kg of water heated
by 100 degrees gains only 5-IQ-9 g of mass. Such a minute mass
change cannot be detected even by means of a most sensitive
modern balance. The formation of nuclei, however, involves quile
perceptible mass changes; moreover, this mass defect determines
the binding energy (see§ 5.6).
In relativistic mechanics it is natural to define the part of th~
particle's energy that turns into zero at v = 0 as a kinetic energy.
It is obtained by deduction of the rest energy from the total energy
of a particle:
(5.44)
The same result can be obtained when a force's work is cal-
culated according to the equation of relativistic dynamics:
dT=Fvdt=vd(mvv) =mvvdv + mv2 dv=
= myu dv + mvv~ (6 d~) = mv3v dv ( 7- + f,) = mv3v dv.
Hence
152 Special Theory of RelatlrJltg
li T = 0 at v = 0 {i.e. at y = 1). canst=- mc 2 , whence it fol·

lows again that T = mc 2 (y-l).
Let us find the conditions under which Eq. (5.44) is converted
into an expression for the classical kinetic energy. Expanding v
m series
y= .y'l ~~2 =I ++J32.+fJ3i+
\\e see that
(5.45)
i.e. the classical kinetic energy has the meaning inasmuch as

j3 «: I and the term j3 4 can be ignored.
Having designated the classical kinetic energy by Tc1 and the
relativistic one by 1,81 , Eq. (5.45) can be rewritten as follows:
Tret=Tc~+{m?~+.
The ratio T, 8 J/T,, is given by the following expression:
T;,:' = 1 + { j3
2 +(terms of the order of j34 and higher).
In nuclear physics, where "the limits of applicability of New·
tonian mechanics" have to be defined more accurately, it is as-
sumed that 'fret= Tc~ provided the second term on the right-hand
side of the equation is less than one per cent (remember that
13 ~ I, and the series descends rapidly). Consequently, the bound·
ary limit (which is, of course, conditional) may be found from
the equation f ~ 2 = 0.01, ~ ~ 0.12.
Since y = y(v), and y = 8{mc 2 for a particle, one may speak
of relativistic velocities when the total energy 8 of a particle ex-
ceeds appreciably its rest mass, i.e. when the condition (8/mc 2 ) »
'» I is met. Surely both conditions are qualitatively equivalent,
but it should be borne in mind that when the velocity approaches
its limit (c), the energy tends to infinity. Therefore, very small
velocity variations in the vicinity of c may alter radically tht"
particle's energy.
In nuclear physics it is more convenient to deal with energies
of particles rather than with their velocities. Of course, particles
of different masses will have different energy limits defining tl,e
applicability of classical mechanics. For example, in the case o~
electrons this limit is equal to 3 kev, while for protons 7 Mev (it
would be useful for you to try to obtain these numbers yourself).
Eq. (5.32) may be rewritten as
ilf~vilfo (5.46)
Relafivlsflc Mechanics of a Parfltle 153
if the zero energy ;& 0 = mcJ 1s lal,cn 1nto account. The rest energ.J
(8 0 comprises all kinds of energy possessed by an object (or a
system). Eq. (5.46) shows that all kinds of energy increase y-fold
on transition from the proper (co-moving) frame to any oth'.!r
inertial frame. There was nothing o[ the kind in classical me-
chanics. On the other hand, the total energy o[ a particle (Eq.
(5.32)) and its kinetic energy (Eq. (5.45)) grow without bound
when v-+ c. This result has a plain physical meaning. A particle
whose rest mass differs from zero cannot attain the velocity equal
to c. This can be inferred from the fact that the particle would
require an infinite energy in order to achieve that velocity. Here
the limitedness of the velocity of light in vacuo shows up again.
When treating light quanta (photons) as relativistic particles (see
§ 7.6). one should bear in mind that they belong to another class
of particles and could not come into being as a result of acceler-
ation of conventional particles, that is through a dynamic transi-
tion. The limit transition v -+ c is carried out in nature, but t!le
1
te§m~~s~ xo~~~e~~~~i~r:;::~i;_r:n~r:n:t~mi.s ;h:efo~~~~h(~~
zeroth)
component of the 4-momentum of a free particle has a direct bear-
ing on the particle's energy. This is evident from the following
straightforward transformation:
P 4 =rnicy=7mc~=78; (a) I P(O)=mcy=f. (b)
+
Consequently, the 4-vector P is referred to as a 4-vector of energ,J-
momentum of a particle. From Eq. (5.7) and from the fact tllat
P V
= m it follows that
"jyl=-m2c2 ; (a) I 'P =m c
2 2 2• (b) (5.47}
+
The component transformation law for a 4·vector P is given by
Eqs. (5.42) and (5.43). It remains to rewrite Eq. (5.22) in th~
final form
P(p, if); (a) I :P(f. p). (b) (5.48}
where p = myv. Let us consider a particle in the reference frame

in which its relativistic 3-momentum p = myv is equal to zero.
The frame in which a particle is at rest (11 = 0) may be called
a proper reference frame. Let the particle's energy in tpis frame
be equal to ;&0 • Then in the frame K' in accordance with Eq.
(5.43) '
$'=r;so, p:=·-r~;so=-78'=-7V. (5.49}
154 Special Theory of Relatlvity
It is seen from Eq. (5.49) that the energy transferred by a par-

ticle is associated with the origination of a momentum. Indeed.
in the proper frame a particle possesses the energy 8° which does
not move in space. The momentum of a particle (an energy car-
rier) in this frame is equal to zero. In the frame K' the particle
moves; its velocity is equal to -V. This means that the energy
.. flows" at this velocity. Eq. (5.49) for p~ shows that the energy
flow involves the momentum p:=- r ~ V. This momentum
coincides with a relativistic three-dimensional momentum because
according to Eq. (5.32) 8°/c2 = m; the particle's velocity is equal
to -V, and r coincides withy in this case.
Thus, an energy carrier (a particle, in this case) needs a mo·
mentum to be attributed to. Although we have obtained this re-
sult for a particle, it has a general significance; we shall come
across it again when examining an electromagnetic field (Chap·
ter 6).
We would like to emphasize here that the very fact of integrat-
ing certain quantities into a 4-vector points to an intimate con·
nection between them. The quantities which are 4-vector campo·
nents (usually a 3-vector and a scalar) constitute in a sense a
closed combination: to calculate the energy and momentum of a
particle in the frame K', one should know those in the frame K
(see Eq. (5.43)). The fourth (or zeroth) component of a 4-vector
or energy-momentum cannot turn into zero. If in some frame the
energy and momentum of a particle turn into zero, they are equal
to zero in any other reference frame. This is where relativistic
and classical relations differ fundamentally. In classical me·
chanics the energy and momentum o[ a stationary particle are
equal to zero.
The square of a 4-vector is an important invariant. Let us write
it out:
P 2 <=p 2 -~=-m 2 c 2 ; (a) j j:,2::::=~-p 2 =m 2 c 2 , (b) (5.50)
(we have used Eqs. (5.47) and (5.48)). Needless to say that it is
equal in magnitude but opposite in sign in these two cases. What
is important, we have found an invariant relationship between a
relativistic momentum and a relativistic energy of a particle; it
is essential in this case that the invariant equation (5.50) de·
termines an invariant mass, a rest mass of a particle.
From Eq. (5.50) one can express the particle's energy in terms
of its momentum:
fG-c-Yp'+m'c'. (5.51)
The partkle's energy expressed in terms of its momentum is
referred to as the Hamiltonian function 3"1 of a particle. Thus,
Retallvisfit: Mechanics of a Particle 155
Eq. (5.51) gives the Hamiltonian function of a particle. It is com-

mon knowledge that the derivative of the Hamiltonian function
with respect to the momentum components yields the velocity com-
ponents of a particle:
:::r: == ;~ =%=v:r:, ... , or v=VpB. (5.52)
Eq. (5.52) may be derived by differentiating Eq. (5.50):

PxdP:r: + p11dp11+ pzdpz=-}rB dB, (5.53)
whence
:~ =p:r:f• :~ =p/1.=;' :~ =pzf• (5.54)
But inasmuch as B = mc 2y, and p = myv, then
p=~v (5.55)
for a particle, and we get back to Eq. (5.52).

In what follows we shall need a formula which is easy to
derive from Eq. (5.52): multiplying the left-hand and right-hand
sides of Eq. (5.52) by dp, we obtain vdp =dB. Those who are
not particularly inclined toward a treatment of gradients should
note that the same result follows at once from Eqs. (5.37a) and
(5.37b) both in the classical and relativistic cases although a
momentum is defined there in different ways, of course. Multi-
plying the left-hand and right-hand sides of Eqs. (5.37a) and
(5.37b) by v dt, we obtain F dr, i.e. dB. in the right-hand side,
and v dp in the left-hand one. Thus, in both classical and relativ-
istic mechanics we have the same formula
(5.56)
but the definition of energy and momentum will be different.
In the simplest case of a unidimensional motion B = B(p) is
determined according to Eq. (5.51). In the plane of the variables
B, p velocity is determined via a tangent of a slope angle of a
curve B = B(p) at a given point. When p »me, Eq. (5.51) b(!-
comes
B=cp. (5.51')
Particles with a finite rest mass are called ultra-relativistic if the
last relation holds for them. Eq. (5.51 ') is valid both for ultra-
relativistic particles and for photons. We shall see in § 7.6 that
light quanta (photons) may be treated as relativistic particles.
But now it is worthwhile to point out that they are quite special
particles. In any IFR these particles possess a finite momentum
15G Special Theory of Relativity
and a finite energy. Their velocity in vacuo is the same in any

IFR. They cannot originate from any of the particles possessing
a finite rest mass by means of their acceleration. Finally, we infer
from Eq. (5.50) that the photon's rest mass is equal to zero.
Let us consider some more relationships for a particle located
in an external potential field. Since the field is assumed to be
propagating at a finite velocity, the primary principle of the STR,
the finiteness of the signal transmission velocity, is observed. As
to the force acting on a particle, it is defined from the magnitude
of the potential function at the point in which the particle is
located (the field is stationary).
When a particle is located in a potential field, then Fv dt=-dU
and Eq. (5.31) turns into d(mc2y) = - dU, whence follows the
law of total energy conservation for a relativistic particle in a
potential field:
+
mc 2y U = const. (5.57)
(The energy is total in the sense that the sum of the relativistic
energy and potential energy of a particle remains constant.) In
relativistic mechanics the kinetic energy is equal to mc 2 (y-l);
having altered the magnitude of the constant in the right-hand
side by mc2 , the energy conservation law can be rewritten in the
form
mc 2 (y-I)+V=const. (5.58)
When a particle is located in a conservative field, its velocity
and potential energy may change in the course of motion but the
value of 8 = mc 2y + U remains constant (see Eq. (5.57)) in that
it is time independent in a given ·I FR. The quantity 8 may be
called a total energy of a particle in a conservative field. Surely,
this quantity remains constant in any IFR but changes its (in-
variable) value for another on transition from one IFR to an-
other. The definition of a 4-momentum for a particle in a conser·
vative field as P mV
= holds good, but Eq. (5.48) is to be re-
placed by P[mvv. f<8-U)]. since myc=~myc 2 = 8 -;u,
as it is seen from Eqs. (5.21) and (5.57); using the relation
P2 = - m 2c2. Eq. (5.51) yields the Hamiltonian function for a
particle in a conservative field:
fS ~ iM ~ c .YP' + m2c2 + U. (5.59)
Just as in the case of a free particle, a 3-relativistic momentum
can be expressed via the energy, velocity and potential energy:
p=myfl=+.-mc 2yfl= B;U v.
Retatlvlstlc Mechanics of a Partlcte 157
The transformation of the fourth component of the energy-mo-

mentum vector for a particle in a potential field shows that ac-
cording to Eq. (5.42) (and making use of Eq. (5.10) as well)
the total energy is equal to 8' = mc'ly' +
U' in the frame K', H
one would expect.
§ 5.6. The rest mass of a system. The binding energy. So far
we have been considering mechanics of a "particle", i.e. the be-
haviour of a certain single unit. However, the "elementariness"
(indivisibility) of a particle was, in fact, nowhere assumed, so
that all conclusions may be transferred to more complex system-;
consisting of .. subsystems".
The rest mass M of a complex system is defined in accordance
with the general formula of Eq. (5.50):
(5.60)
where E is now the total energy of a system and P its total m~
mentum.
Let us confine ourselves to the simplest systems consisting of
individual particles. First suppose that the particles do not in-
teract with one another. Then the energy of such a system will
be the sum of energies of particles comprising the system:
(5.61)
The additivity of energies signifies the absence of interaction. The

total momentum of a system always adds up vectorwise from the
momenta of individual particles, i.e. it is always additive:
(5.62)
In this case the rest mass of the system can be written as fol-
lows:
(5.63)
In order to find the relationship between the rest mass of a

system and the rest masses of particles comprising the system, one
should pass over to the reference frame in which the total mo-
mentum of the system is equal to zero: P = 0. Then from Eq.
(5.63) we obtain
(5.64)
,., Spec(af Theory of Relottvity
We sec that the rest mass of a system is expressed as the su:n

of energies of constituent particles (divided by c2 ). But the energy
of an individual particle can be represented according to Eq.
(5.44) as the sum of its rest energy and kinetic energy
<S, ~ m,c' T,.+ (5.65)
Then in accordance with Eq. (5.63) Wf! get
M= Lm~+-Jr Lr~. (5.66)
' '
Eq. (5.66) yields the important result: the rest mass of a sys-
~~~ t~f~~e~~elf~ e~~~gy0 ro{~~~ ;:r5t~~fes0f( d~~~~~iJu~~t ))a~~fi1~5at~~
in the reference frame in which the total momentum of the system
is equal to zero.
This way we reach a conclusion that in relativistic mechanicc;
the rest mass of the system composed of non-interacting particles
is not an additive quantity. Such a property of mass is uncommon
in classical mechanics. It is tempting to bring in some new
definition of mass for constituent particles, so that the rest mass
of a system will add up from these new masses called "relativis·
tic" sometimes. It is easy to see how this can be done. In the
reference frame where P = 0 we obtain, according to Eq. (5.64):
Me'~ :E <S, ~ :E m,y,c'. (5.67)
' '
Consequently, one can write down
M=Lmlyl=Lm~el, (5.68)
where miel = m1y1 is a relativistic mass. That is how we realize
additivity (which is by no means obligatory), but at the same
time we clear the way for various delusions. In fact, the introduc·
tion of a relativistic mass for a particle creates an illusion that
the increase of the particle's energy, or "relativistic mass", ac·
companying the increase of its velocity (momentum) is associated
with the changes of the internal structure of the particle. But
surely, there is no such thing at all (to make sure, one may ju~t
~~s~ ~~~~er'~r ar~~~~!~e f~~~~yw~:~~~t waiftt~~c~~~~ea~hee or~~tli~~~t1y
due to the special properties of the 4-space-time coming through
in the Lorentz transformation.
In terms of the four-dimensional approach the term "mass''
refers to the invariant norm pf the 4-vector of energy-momentum.
By bringing in a relativistic mass we actually apply the term
.. m.aSs'' (with an accuracy 'of a factor} to the time component of
Relativistic Ml!tllilnlcs of a Pflfticil! !59
the 4-vector of energy-momentum, which is, as we know, the

energy. However, the energy and the rest mass that we intend to
usc represent essentially different physical concepts.
Energy is a relative quantity; it depends on the IFR in which
the particle or the system of particles is considered. The rest mass
1emains the same in all IFRs; it is an absolute value of a 4-vector.
The time component of a 4-vector (energy) coincides with its
absolute value (rest mass) only when the spatial components of
this 4-vector arc equal to zero (which means that either the par-
ticle's momentum or the total momentum of a system of partici{:S
equals zero). And only in the case of the energy coinciding with
the rest energy it is proportional to the rest mass (with the con-
stant coefficient c2 ).
Thus, we can ascribe the precise four-dimensional meaning to
the momentum, energy and rest mass of a particle or a system,
provided the first two quantities are treated as components of a
4-vector of energy-momentum, and the last quantity as a norm of
the same vector. The methodical aspects of the problem are dis-
cussed in Supplement IV.
Let us examine now a system composed of interacting particles.
Eq. (5.63) remains valid, of course. Eq. (5.61) should be, how-
ever, replaced by
(5.69)
where U denotes the interaction energy of particles. This energy

is defined as the work required to disjoint the system into "initial"
non-interacting parts. In a stable system U < 0 since a "stable
equilibrium" state is characterized by a minimum of energy. In
such a system the quantity U is referred to as a binding energy.
Although the explicit analytical expression for the interaction
energy is often rather difficult to write out (see§ 5.8), its magni-
tude can be estimated. From Eq. (5.60) we obtain for the refer-
ence frame in which P = 0:
L;W.+U
M--'-,-,-, (5.70)
or expressed otherwise
M= Lm•+*Lr~
' '
+*· (5.71)
where we made use of the relation 8 1 =·m~c2 +T 1 which is cor-

rect for every individual particle.
If the condition ~ T1 « U is met, i.e. the total relativistic

kinetic energy of particles is small, then
M ~ f m, +(Vic'). (5.72)
It is evident from Eq. (5.72) that.in the system of interacting

particles the difference
(5.73)
customarily called a mass defect, is always different from zero.

In a stable system U < 0 and .1-M > 0. From the mass defect
one can calculate the binding energy:
U ~ !>M • c'. (5.74)
Such a calculation is meaningful only when binding energi~s
are substantial. Precisely such a case is realized in atomic nuclei.
It is well known that atomic nuclei are very stable, this being the
evidence of their considerable binding energy. Atomic nuclei a;e
composed of protons and neutrons, with each nucleus possessing
a quite definite number of protons and neutrons. The masses of
protons and neutrons in a free state (outside a nucleus) can be
experimentally determined. The mass of any atomic nucleus can
also be experimentally found. The difference between the sum-
marized mass of free protons and neutrons comprising the nucleus
and the measured mass of the nucleus yields the mass defect
and, according to Eq. (5.74), the binding energy. Exactly in this
manner the binding energies of nuclc1 are found in atomic phys-
ics.
Let us write out separately expressions for ultra-relativistic par-
ticles (v::::::::: c). In this case Eq. (5.51') holds:
I!~ cp (5.75)
2nd consequently from Eq. (5.51) we shall obtain m = 0. How-
ever, for two or more particles we shall obtain from Eq. (5.63)
M 2 c 2 =(~P~Y- (~ PtY* o. (5.76)
since ~ 8 1 = c ~ p 1• This points out that the rest mass of a

system composed of particles with the zero rest mass is not equal
to zero. There is nothing surprising in this since rest masses do
not add up.
Finally. ;r few' words about "composite" subsystems. When
determining a rest mass of a complex system according to Eqs.
Relativistic Mechanics of a Particle 161
(5.70) and (5.64), one should take a total energy of the system.
Assume that the system involves an electromagnetic field as well.
Designating the energy of the electromagnetic field by W, we
obtain from Eq. (5.70):
M= Lm,+-fr Lr,+*+}. (5.77)

I I
From this it is inferred that the energy of an electromagnetic
field, as any other energy. makes its contribution to the rest mass.
of a system.
§ 5.7. Some problems of relativistic mechanics of a particle..
Within the framework of a given inertial frame of reference there-
is no need to resort to four·dimensional relations; it is sufficient
to use the three·dimensional Eq. (5.26) and also Eq. (5.31). \VC"
shall recall that the general appearance of the second Jaw of
Newton remains invariable and only a momentum and energy of
a particle are defined otherwise. This new definition, however,
changes substantially the properties of the solution of the problem
as compared to that obtained for the same preblem from classical
mechanics equations. In particular, the solution of any problem
of relativistic mechanics does not permit of obtaining the velocity
of a particle exceeding that of light. Some other distinguishing
features come up as well; to eluci8ate them, we shall consider a
few problems, solving them with the aid of both the classical and
relativistic motion equations.
Since the equations obtained will not be needed afterwards, an·
individual numeration for each problem is adopted in this sec·
lion.
I. The elementary solution of the problem of the unidimensional
motion under the action of a constant force. The motion equation
takes the form (a classical equation on the left side and a rela-
tivistic one on the right side):
•
Tt(mv)~F; (a) I• Tt(myv)~F, (b) (I)
where F =canst. Integrating with the initial condition v = 0 at
t = 0. we obtain
mv=FI; (a) J mvv=FI (b) (2)
A velocity as a function of time is found from Eqs. (2a, b) al-

gebraically:
Ft/m
vel=;!-; (a} j Vrel (b)
(3)
6-9>
162 Spcclat Theory of Relativity
We shall discuss these results in Problem II as it will be-

come evident that Problem II is just a variant formulation of
Problem I.
11. The rectilinear and uniformly accelerated motion of a par-
ticle. If a particle moves along the x' axis (v~=v~=O) in the
frame K', then v 11 = Vz = 0 in any other IFR according to Eq.
(3.26). Let us consider a motion of ·a particle along the common
x, x' axis with a constant acceleration. If the acceleration of the
particle does not vary, the acceleration components in the ref-
erence frame in which this particle is at rest (the proper reference
frame) are (w0, 0, 0, 0). The quantity w0 is a customary thre·>
dimensional acceleration directed along the x' axis. The square
of a 4-vector acceleration is an invariant, so that the following
condition must hold in all reference frames for a uniformly ac-
celerated motion*:
du )'
( ___£ =w2 (I)
d-r: o•
where w 0 is the magnitude of the three-dimensional acceleration
in the proper reference frame. Surely, this condition differs from
the requirement V = 0 **. The 4-velocily components for a unidi-
mensional motion in an) reference frame acquire the form
V(u 1 = vvx, 0, 0, icv) \\"hence, according to Eq. (5.23), it follows
that ti 2 = ti 3 = 0. Consequently, the two equations that remain in
an arbt!rary IFR arc
m ~~~ =,•F, m ~~~ =fvFv.
We denoted here Fx = F and Vx = v. Eq. (I) will take the
form
C:":Y =~(v2F2 - ~v2F 2 v2)=;=Wo=const.
The motion equation will be written as
u;~~ = Yit'o, or (2)
In the three-dimensional notation
f, ( -J 1~ f, ) ~ w 0, or ,y-!:-t; ~ w<l+ C. (3)
• In this chapter we mt~ke use of the 4-1pace-time \\'hose fourth coordinate

is imagmary.
•• If w0 = const, ti is variable. In classical mechanics an object acquires a
constant acceleralion under the action of a constant force, whereas in relativis-
J~c a~Yf:s7~~~a~e~~nss~a0~~~~~~~ i~r:;et~c! froa~~~nt acceleration to an obj~t only
Relativistic Mechanrcs of a Particle 163
Solvmg Problem I, we got convinced, however, that these equa-

tions could be obtained at once if we substituted the constant
force F = mw 0 into the relativistic motion equation. But w 0 is not
the same as ti; this can be inferred from Eq. (2).
If the initial condition is such that the velocity u is equal to
zero at t = 0, then C = 0 and expressing the velocity through w~.
we obtain from Eq. (3)
(4)
where uc 1 denotes w 0 t. Integrating the latter relation at the initial

conditions, x = 0, t = 0, we obtain the following expression:
(5)
The soiution of the classical motion equation for a constant

force and the identical initial condition.s takes the form
VcJ=Wof, x=wrJ 2/2.
If the classical velocity grows indefinitely with time, it follows
from Eq. (4), due to the obvious inequality
• / x x2 = 'I/Aa2x+x2 a<a, if A>O,

rv A+QF
that the relativistic velocity always remains less than c as it, in
fact, must be in accordance with the principle of ultimate velocity
of signal transmission. The relativistic equations (4) and (5) for
the velocity u and coordinate x turn into classical ones at
Vcdc «I. If one rewrites Eq. (4) as v=c/.YJ c2/w'fl, it be- +
comes evident that v- c when t-oo.
Let us li.nd the relationship between the coordinate time t and
the proper time • of a particle. If one chooses the common origin
of the time count t 0 = o0 = 0, then
' m,l '

=-Arcsinh-=-ln
Wo c Wo
( m,t
c -+ ~ I+~
w)'' )
c
(6)
..
(see Eq. (3.16)) .
164 S(Jecial Theory of ReltJfit•if!l
Ignoring unity in the radicand as compared to Wolle, we ob·

tain, when t- oo,
'~"'""'.....E....in~.
wo <'
We see that the proper time of an object moving with a uni-

form acceleration ~ows substantially· slower than the time in a
.. motionless" reference frame relative to which the motion is co:I·
sidered.
There remains, of course, the physical question as to what clock
counts the particle's proper time given by Eq. (6) since the rela·
tion dr: = (1/v)dt pertains to a clock moving uniformly and
reclilmearly. We examined this problem in detail in § 3.3.
In conclusion we should note that a relativistic uniformly accel·
eratcd rectilinear motion is also called hyperbolic since the tim~
dependence of the path travelled (see Eq. (7) below) represents
hyperbola in terms of geometry. If a charged particle moves in
a uniform and constant electric field, or a heavy particle in a
unHm m and constant gravitational field, the motion is hyperbolic.
Let us write out finally the principal equations describing a hyper·
bolic motion:
x(l)=w; '' [-~I+
;-(w•t)'
---;- - I].
(7)
.../l+(w0 t/c) 2 '
From the expression for the time derivative of the velocity one
may -;ec the difference between a relativistic and non-relativistic
"constant" acceleration.
Ill. The motion of a charged particle in a constant uniform
electric field. Let us choose the following initial conditions: at the
moment t = 0 the coordinates of a charged particle x0 = y0 = 0,
.and its velocity v 0 is perpendicular to the field E. This corres-
ponds to the problem about a particle flying into a charged ca·
pacitor parallel to its plates (Fig. 5.1). Let us direct the x axis
along E and the y axis along v 0 . Then the motion of the particle
will take place in the plane (x, y). As far as it is possible, we
shall not discriminate between classical and relativistic equations
.of motion, writing them in the common form:
!f/;-=eE,
where F = eE is the force exerted by the electric field on th!.!

<harged particle. In components:
D~~;=eE, P11 =0.
Retatlvisflc Mt>chanlcs of a Parf(cle 165
Whence the momentum is found by integration:

p,.=eEt+Po"' p11 =Pov·
But according to the initial conditions both in classical and rei~-
Flg. 5.1. An eledron llymg into the uniform electric lield of a capacitor is at
the origin of coordinates at the moment t = 0 The force exerted by the lield is
t~~e~te:x~to~~el~ia~s~c~it"s!~ueti~~t~fl ~he~~~~b~~~a~of~~~~~~n w':\~st~~~e~\e~ a~~o"b~
lem of the motion of a heavy point thrown horizontally in the gravitational field
at the velocity v0
iivislic cases p"' = 0 and po11 =Po at t = 0. Consequently, one

may write
p"=eEt, Po 11 =po.
From here on we should take into account the difference in d~-
1lnilions of a momentum:
classical: relativistic:
p=mv. p=ni\V.
We have In this case
dx eEt
v.c=Tt=m·
(I)
p2 = P~ + p~ = (e£1)2 + p~.
dy
Vy=(jf=Vo. The particle's energy /! will
be defined as
Integrating, we get
et: ~~ ~=c-v'~=
x=m2+xo.
y=vo!+Yo·
= ..Jm2c4 + pgc2 + (ce£1) 2 -
But according to the initial
conditions x0 =y0 =0 at t=O,
~ -,j 8: + (ceEt) 2•
we finally obtain Using Eq. (5.55)

eE t 1
x=m2· y=r.Jo/. (•) (I')
166 Special Theory of Relatiuity
The velocity of a particle we obtain
v2= v!+ v;= ( ~ r + v~

12 v~rl=-%= p~ca
c~eEt
=
increases indefinitely with

time. The particle's path ob· ...j<8~+ (ceEt) 2
tained from Eq. (*) by elimi-
nation of the time t is repre· ,, (2')
sented by a parabola = ,Y(::.)'+(-';'-)''
(2) where vr;e' is defiined accord·
ing to Eq. ( 1'). lt follows from
Eq. (2') that, as in the pre-
vious section, vrel is alwa)'S
less than c because
vrel = ..!!JL = PoC 1 (3'}
II dt ...jf8~+ (ceEt)z
diminishes with time. Inte-

grating Eq. (2'), we obtain
with regard to the initial con-
ditions
x= ef-..J~~ + (ce£1) !;- 2-
Integrating Eq. (3') and tak-

ing into account the initial
conditions, we get
y = ~~ Arcsinh c;~t .
Eliminating t from the expres-
sions for x and y, we have
x=~(cosh :!~ -1).

Thus, when a classical path was a parabola, a relativistic one-
lurneC: out to be a catenary curve. In the case of v >> c a catenary
curve, however, turns into a parabola. Indeed, if v/c « I, we have
y ~ I, Pa = mva, and /So= mc2• Besides, at small x one may
assume cosh a ~ I + 92/2!, whence
x=-!:: e;:~~ = :;~ ~j~;: = ~~ tf,

which is the parabola of Eq. (2). This example, as well as all
subsequent ones, shows that problems of relativistic mech<~nir.s
Rela!IIJisfic Mechanics of a Particle 167
do not requ1re any velocity dependence of mass: the solution is

-obtained by integrating a motion equation.
IV. The motion of a charged particle In a constant uniform
magnetic field. The classical and relativistic motion equations for
a charged particle in a magnetic field
-¥,--e[vB[ (I)
.are the same not only in their appearance. The point is that a
magnetic field does not perform any work on a charged particle
and the energy of the particle remains constant (see Eqs. (5.20)
and (5.31)); of course, the expressions for energy in classical and
..
relativistic cases arc different. Using the relativistic relation
p=etV,
where' 8 = canst, Eq. (I) can be rewritten in the form

%= e;, [vB), (2)
whereas from the clac:;sical definition of a momentum p = mv it

follows that
%=f,-lvB). (3)
Therefore, relativistic and classical motion equations (2) <.~nd (3)

arc distinguished only by the constarits standing in front of tile
vector product. We shall recall how Eq. (2) or (3) is solved.
Orient the z axis along a magnetic field. Then B = Bk. DC'not~
the constant factors appearing in front of the product [vk} in
Eqs. (2) and (3) as follows:
roe~=!f, ro,e~= ec;B = ~~ =~roc~=wc~~· (4)
Now for definiteness let us solve Eq. (2). To define the vector
product [vk], rewrite Eq. (2) in components:
(5)
It is expedient to resort to a complex variable in the plane
(ur. u.,). Multiplying the second relation of (5) by the number i
and adding the result to the first one, we obtain -ft.
(vx +ivy) =
= -iw(vx+ ivy).
This equation can readily be integrated:
where a 1s 0:1 complex constant. If it is written in the form a =

= v 0 ,e-•o;r; with the real v0 , and a, the solution takes the following
form:
(6)
Obviously, v 0, is a modulus of the complex number in the left-
hand side of Eq. (6);
v~=v!+v~.
Consequently, the magnitude of the particle's velocity remains
constant in the plane (x, y). Eq. (6) can be rewritten in the form
ft (x + iy) = vo,e-' !~o~t+al'
permitting of direct integration:
(7)
Recalling the geometric representation of the complex quantity

w = x + iy = re"~. we see that a particle remains permanently
on a circumference of a constant radius r = v0 t/oo, while the angle
between its radius vector and the x axis increases evenly with
time: 'P = wt + canst. This means that the projection of the par·
tide's motion on the plane (x, y) is a uniform motion along a
circumference of the radius
(8)
where p, is the projection of a momentum on the plane (x, y)

and w is an angular velocity. As to the motion along the z axis,
it follows from the third relation (5) that
z=zo+voJ. (9)
From Eqs. (8) and (9) it follows that a charged particle in a uni·
form magnetic field moves along a helical line whose axis coin·
cides with the direction of a magnetic field, and whose radius
is determined according to Eq. (8). The velocity of a particle is
constant, as it should be in a magnetic field. If at the initial mo·
ment the velocity of a particle in the direction of a magnetic field
is equal to zero (voz = 0), the particle moves along the circum-
ference in the plane perpendicular to the field.
The quantity ro,~, defines the cyclic frequency of rotation of the
particle's projection on the plane (x, y) perpendicular to the di-
rection of a magnetic field. This frequency is referred to as cyclo-
tronic. As we have seen, w, 1 = yw,.e 1, i.e. the cyclotronic frequency
of relativistic particles is Jess than that of non-relativistic ones.
At small velocities y _,. I and we~_,. Wrel·
1n conclusion let us consider an acceleration gained by a par-

ticle in an electromagnetic field in terms of classical and rela-
tivistic mechanics. From the general motion equation
-'£1--e(E+[•B])
we obtain in the classical case (p = mv)
v"--;;. {E + [•B]).
In order to obtain an acceleration in the relativistic case we shall
make use of Eq. (5.55), whence
%=?-~~ +~%-
According to Eq. (5.31) d8/dt = Fv = eEv, and according to
Eq. (5.32) iB"!c' =my, so that (cf. Eq. (5.38))
-Dret= ~y { B+(vBJ-fa-(Bv) }=
=~Vc,- m;ca v(Bv)=~.Uct-iv(Ev).
Ttc second term in the last link of the equation can be regarded
as an emergence of a certain friction (proportional to the veloc-
it) ); owing to this, one may realize in qualitative terms that the
acceleration of a particle decreases sharply as the particle's veloc-
ity approaches that of light. It is obvious, of course, that vel =
= V,e 1 with an accuracy to within ~ 2 • The motion of a charged
particle in constant electric and magnetic fields is presented in
detail in [9]. § 22; we should only point out here that in the ca;;~
of crossed (mutually perpendicular) fields for which £2- c2B 2 9=-
+ 0 holds (cf. § 6.5), a transition to a certain inertial frame of
reference may eliminate one of the fields and leave either an
electric or a magnetic field. Then in this reference frame one may
utilize the results obtained here.
V. The reaction motion in relativistic mechanics. As in the
previous problems, we shall be examining classical and relativ-
istic cases simultaneously. As an example, let us consider thl!
motion of a rocket which (together with the ejected gas) can be
treated as a closed system. We shall recall that the rocket pro-
pulsion is brought about due to the fact that during each time in-
terval it ejects a certain amount of substance at a definite velocity
with respect to the rocket. In accordance with the momentum con-
servation law the rocket shell with the left-over fuel acquires a
momentum in the direction opposite to the direction in which the
gas jet is ejected. Both in classical and in relativistic cases the
problem is easier to solve in the inertial frame of reference

co-moving with the rocket. Since the velocity of the rocket chan-
ges, we deal with the instantaneous co-moving frame.
Let at the moment t a rocket In a relativistic case one-
mass (fuel and container) be should consider not a velocity
M(t) and a velocity V(t); dur- increment, but a velocity pa-
ing the time dt the rocket en- rartteter increment, since ve-
gine ejects a mass dM of gas locity parameters add up ( . He
at a constant velocity relative additive), while velocities do
to the rocket, v. Write out the not (see Eq. (3.37)).
momentum conservation law Write out the relation be-
in a co-moving frame (a unidi- tween the velocity and the ve-
mensional case): locity parameter e (see Eqs.
(M-dM)dV-vdM~o. (I) (2 -27 ), ( 2 -28 )):
tanh6=~. cosh6=v.
with dM > 0. Ignoring infini·
tesimal values of the second sino 8 ~ v~. (51
order (dM ·dV), we get when examining the velocity
of ejected gases, and
(2) tanh6=ll, cosh6=f,
Eq. (2) is easily integrated sinh8~fB, (6}
(v = const): when dealing with the velocity
V=vlnf, of the rocket. Obviously,
8=mc2cosh6, p=msmh6. (7)
where C is an integration con-
stant. Choosing the initial con- Suppose that during the time
ditions: at t=O V=O, M(O)= interval dt the mass dM is
=M 0, we finally obtain ejected at the velocity P=v/c.
The velocity of the rocket after
V=vln~. (3) the ejection of the mass dM
increases by dB= dV/c. We
Eq. (3) determines the velocity shall express the increment dB
V(t) of the rocket as a func- via the velocity parameter in-
tion of the velocity of ejected crement (see Eq. (6)):
gases and the change of the dB~ tano (d8), (8)
rocket mass (the mass of the
burnt fuel is equal to M 0 -M). Now write down the laws of
There is, however, one note- momentum and energy conser-
worthy feature in this deriva- vation. In a co-moving frame
tion. At any moment of time the momentum of a rocket "be-
the rocket has different co- fore the ejection" is equal to-
moving inertial frames. The zero (B = 0). After the ejec-
velocity increment in Eq. 2 tion of the mass dM the rocket
.during the time interval dt is acquires the momentum (M +
Relativistic Mechan{cs of a Particle 171
+ dM)
written for different inertial
frames of reference. In transi-
tion from one inertial frame of
I sinh (dO). while the
momentum of the mass dM
direded oppositely is equal to
reference to another all veloci- dM sinh(O). Hence.
ties (and velocity increments)
add up in classical mechanics. - dM sinh (6) +
Therefore, it is immaterial that + (M +dM)sinh(d6)-0, (9)
Eq. (2) refers to different
IFRs; the final velocity can be Relativistic mechanics makes
obtained [)y summing (inte- it possible to take into account
grating) velocity increments any energy transformations,
over the total time interval so that here we can write the
during which the velocity of energy conservation law as
the rocket varies from 0 to V: well:
+
r~=-~~dV.
dMc 2 cosh (0)
(4) +(M +dM)c'cosh(d6) -Me'·
M, 0
(10)
This is just the method that
was used in derivation of Eq. But we may regard dO as
(3). Hence it is clear how im- a small quantity, so th:.~t
portant in this derivation is sinh(d6)"<da cosh(d6)"< 1:
the additivity of velocities in consequently, ignoring infini-
transition from one IFR to tesimal values of the second
.another. order dM ·dO, we obtain from
In the foregoing calculations Eo. (9)
the mechanical energy conser- dM sinh (6)- M d6, (II)
vation law was not used be-
cause, on the one hand, it is and from Eq. (10)
insufficient (the thermal ener-
gy is also significant here), and cosh6--l. (12)
on the other, it is not neces-
sary when the velocity of a Dividing termwise Eq. (II) by
rocket is calculated. Eq. (12), we get
dO=- d: tanhO,
de--pd:. (t3)
where M = M (t) is the mass

of the rocket and fuel at the
moment t. Since in relativistic
mechanics a velocity parame-
ter is an additive value, the
172 Specla/ Theory of Relativity
final magnitude of a veloc1ty

parameter can be found by in·
tegration:
e~~ln{'f. 114)
wh.ere M 0 is the mass of the

rocket at the moment when its
velocity is equal to zero.
Eq. (14) also determines im-
plicitly the velocity of the
rocket at the moment when tie
mass of the burnt fuel is equal
to Mo-M.
It is easy to see that if a rocket moves at a non-relativistic

velocity, Eq. (14) turns into Eq. (3). In fact, in this case B « 1
and, consequently, tanh a = B « I, whence tanh e ~ e and Eq.
(14) coincides with Eq. (3). Just as in all solutions of rela-
tivistic mechanics, the velocity of a rocket cannot exceed that of
light c. Even if we manage to burn all the mass of the rocket
together with the fuel, i.e. M-+ Mo. ln(M 0/M)-+ lXI. From Eq. ( 14)
it only follows that e-+ lXI (with the maximum value of ~ being,
of course, equal to I). But B = tanh e and at e- lXI tanh e- 1.
i.e. V-+c.
Surely, the higher the ejection velocity, the more effective is
the rocket performance. Can the ejection velocity be made equal
to c, i.e.~= I? It can be, provided light serves as a reaction ga-:;:
only photons and neutrinos can move at the velocity c. These two
sorts of particles are peculiar in that their rest masses are equal
to zero (§ 7.6). The zero mass, however, is seen directly from
Eq. (I 0) in this case. As v-+ c a velocity parameter satisfies the
condition tanh e = ~ -+ I. But with tanh e -+ I cosh a-+ lXI, and
in order to satisfy Eq. (10) at a finite value of M, it is necessary
lhal dM = 0.
The more detailed analysis of photon rocket capabilities shows
its inapplicability for long-range space flights (see [11)).
VI. Colliding beams. The progress in nuclear physics depends
essentially on how high are the energies of interaction between
elementary particles that could be observed. There have been l\\o.
sources of high-energy particles so far: cosmic rays and accele-
rators. The accelerator designers are still very far from master-
ing the energies which particles in cosmic rays possess, whil:!
the systematic research in high-energy physics has been restricted
by the range of energies covered by accelerators. Accelerators are
complex and expensive machines whose construction continu~s.
Refatiuistic Mechanics of a Parllcfe
for years and whose cost amounts to an appreciable portion of a

national budget of any highly developed country.
Suppose that in a laboratory reference frame particles are accel-
erated to the energy 8. We are to succeed in colliding these par-
ticles with other particles of the same type (for example, we
examine proton-proton collisions). The proton beam, possessin~
the total energy 8 in a laboratory reference frame, can be directed
onto a target containing hydrogen in which protons are practically
motionless. It is sufficient to consider a collision of one bombard-
ing and one resting proton. Then the energy of the system con-
sisting of these two particles is equal to 8 +
m (we assume
c = I in this section). The following question arises: is it possible
to increase essentially the interaction energy by making two
beams, each consisting of particles possessing the energy 8 (in
a laboratory frame), mbve toward each other? How much higher
will be the "'useful" interaction energy? For the sake of diversity
\Ve shall be treating this problem in rather unusual time units.
which are light metres (see Chapter 2).
To simplify the problem, we shall not speak of beams any more.
but consider only two particles. The maximum useful energy
(spent on the generation of new particles, nuclear reactions, heat-
ing of a substance etc.) can be evaluated in the frame of the centre
of inertia, for it is in this frame that the internal energy of the
system is calculated (naturally, the motion of the system as a
\\'hole is "useless" from our point of view). Let us consider two
particles I and 2 possessing equal energies (velocities) in a lab-
oratory frame K, and moving toward each other. This frame will
be for them a frame of the centre of inertia, and the total energy
of particles in lhis frame provides the useful energy under dis-
cussion. This total energy is equal to 2T +2m, where T is the
kinetic energy of each particle, and 2m the rest energy of particles
(in the adopted time units c = 1). Let us find out how the same
collision would look in the frame K' in \vhich particle I is at rest.
This is just the picture of a particle impinging against a target.
We shall consider the energy of particle 2 calculated in the frame
where particle I is at rest. Let us make the appropriate recalcula-
tion according to Eq. (5.43). (Do not forget that the same colli-
sion is being considered in another reference frame.) Designate
the momentum and energy of particles I and 2 in the frame K by
(p, 8) and (-p, 8). In the frame K' the momentum and energy
of particle I are equal to p:=o, l!?=m. Fig. 5.2 illustrates the
frames K and K' and the velocities of particles in the frame K.
The frame K' ig the proper frame of particle I, and according to
Eq. (5.49)
(I)
174 Special Theory of RelaUvity
(in our case r = y since K' is associated with particle 1). It lol·
lows immediately from Eq. (I) that
v=~= T~m ~*·

We shall consider relativistic velocities of particles when~~ t;
consequently
In our case ~ = D ::::: I. whence the quantity fB entering in the

transformation of energy is approximately equal to r. Now it is
x' K
t:,.x'
Fig. 5.2. Two points move in a laboratory frame at equal and oppositely rli·
retied velocities. Particle 1 is at rest In the frame K'.
easy to obtain the formula for transformation of energy of par·

ticle 2 on transition from the reference frame K to K':
~;~ r (~,- Bp,)~ r(~ + Bp)"' r~ + rp"' 2r~.
28 is the energy realized in a head-on collision and accurate lo
a doubled rest energy of particles (which may be neglected at
relativistic velocities). To realize this energy with particle I rest·
ing, one needs the energy r times higher. This calculation shows
the advantage that we gain when we use collision beams.
The same result, howe\'er, can be obtained by simpler means.
We shall demonstrate that the recalculation of energy according
to Eq. (5.43) is equivalent to the calculation of energy according
to the formula 8 =my into which the relativistic expression of
the relative velocity of particles I and 2 is substituted.
Thus, let the momentum of particle 2 in the frame K be equal
to P2 = Px = - mv~ and energy 82 =my. We shall calculate
the energy of particle 2 in the frame K' in which particle 1 is at
rest.
According to Eq. (5 43)
~;~ r(~- Op,)~ r (my+ Bmy~)~ rmy(l +~B),
since in our case y = r (in the frame K both particles have the
same velocities). However, from Eq. (5.10) it follows (do not
Relaflvistic Mechanics af a Particle 175
forget that we have~< 0, see Fig. 5.2) that

fy(l +B~)=y';
therefore
8~=mv',
where v' is determined for the velocity of particle 2 in the frame

K' (i.e. the velocity of particle 2 relative to particle 1). Calculate
the relative velocity of particles 1 and 2. We have
t~;=P:= ~~~px~.
In the frame K particle 2 has the velocity -p and particle 1
the velocity p, and consequently
~:=- ~~~2.
This is just the relative velocity of particle 2. Find now v':
I +~2
v'=--'- I ,,=y'(I+~~ .
.'1/•-':'
Consequently,
lfff=mv' =my· y(l +~') "'2f/S,
for p ~ I and v = r. We have obtained the same result again, as
il should be.
§ 5.8. The conservation laws of relativistic mechanics. So far
we have discussed the energy and momentum conservation !awe;
for a material point, and now we have to dwell on the conserva-
tion laws for a system of n material points. The problem of con·
servation laws has two aspects. The first one involves the acquisi-
tion of relativistic laws of conservation in terms of a given I FR.
The second aspect pertains to the examination of the behaviour
of the quantities remaining constant on transition from one iner-
tial frame to another. Both of these problems are solved by ob-
vious methods for a system of non-interacting particles; in the case
of interacting particles these problems are very complicated.
Let us begin with a system of n non-interacting particles. The
motion equations and energy changes pertaining to the kth par~
ticle take the form (see Eqs. (5.27) and (5.31))
ft (m(kJv(kJvtkl) = ptkl, (5.78)
f, (mlklc2vtkl) = p,~<lvlkJ, (5.79)

176 Spector Theory of Relativity
where FU•I denotes a force acting on the kth particle (summatiort

over k is not performed here!).
If a particle non-interacting with any other particles is con-
sidered, FU•> = 0 and the momentum conservation law pU•> =
= m(l•>y(l•>v!~<l =canst and the energy conservation law BU•> =
= m!~<>y<,.>c 2 =canst follow directly from Eqs. (5.78) and (5.79).
In fact, this circumstance manif~ts itself in the following rela-
tion for an individual particl-e: P 2 = p2 - ~ 2 /c 2 =canst. In th;_
case of an individual particle which represents a closed system P'l
remains constant because p and ~/c do individually. Note here
once again that p and i8{c of an individual particle combine lo
form a 4-veotor.
When we deal with a system of n non-interacting material par-
ticles, the total momentum L p<~~"> and the total energy of a sys-
tem L 1! 1 ~~'> obviously remain constant because each individual
.addendum does.
The transformation laws for a total momentum and total energy
P= L tf". I!= L !!"' (5.80)
in the case of transition from one IFR to another are evident:
.a sum of vector components is transformed as a vector component.
The problem of conservation laws in a system of n interacting
particles is far more complicated. In conventional classical m..--
danics the interaction of particles in the case of conservative
forces could be described by means of the system's potential func-
tion U = (r< 1>, r< 2>, ... , r<n>), where r<~~">(t) determines the position
of the kth particle a,t the moment t, the positions of all n particles
being considered at the same moment of time. In the final analysis
a single moment of time can be chosen in classical mechanics du~
to the assumption of the infinite velocity of interaction propaga-
tion.
Since the interaction propagation velocity is finite in relativistic
mechanics, the calculation of the force at a given point requires
the positions occupied by all particles at some preceding moment
to be known. Hence, it is clear that the form which the function U
takes in a relativistic case is rather complicated.
If one writes down the expression for the energy of a system of
n objects in the form
/! = L m<~~'>c2y<~~'> (5.81)
and for the total momentum
P= L m<~~">y<~~'>v<~~">, (5.82)
the following assertion is possible. The quantities P and i~/c do
not form a 4-vector in contrast to what we had for an individual
pamcle. Besides, these quantities are not constant. The equatiO!l

8 = canst does not hold because in classical mechanics the sys-
tem's total energy, including its potential energy, remains con·
stant. Eq. (5.81) does not take into account the potential energy,
and there is no simple way of incorporating it rigorously. The
finite velocity of interaction propagation causes Eq. (5.82) to vary
with time. In the final analysis this circumstance clarifies the par-
~ro4x·~~~~o~a~~~hp~n~~~s q~~;t~~~s4~e~~o~ ~o~$~~!~~!i~~e~:el~~~~
Indeed, in any reference frame where the sums (5.81) and (5.82)
are composed, the summands are taken simultaneously in the
sense of simultaneity of a given reference frame. When passing
over to another inertial frame, one can find the values of moments
and energies of individual particles and add them according to
the rules of the 4·vector transformation. In a new frame, however,
recalculated events will not be simultaneous. In order to find 8
and P in a new frame, one has to reduce these sums to simlJI·
taneity in this new reference frame. It is this simultaneity reca!·
culation that deprives the quantities P and 8 of properties of
4·\'ector components.
Interacting relativistic systems feature ten integrals of motion:
!~:r~7~.e~~a~~~ee~te~yy~~~~~l~~n;t~~T~! ~~~~~i~~t~h:p~~~~:n~!
or these integrals is presented in the book [16], § 27, for example.
As to the behaviour of integrals of motion in transition from
one inertial frame to another, the energy and momentum form a
4·vector, and the integrals of motion of the centre of inertia and
of the moment of momentum form an antisymmetric 4-tensor, in
the approximation (pO•) = r.P,/c «I), where the terms (p1k))2 are
retained. Hence, it is clear that if all these integrals remain con-
stant in one reference frame, they will be constant in any other
reference frame. ,
There is a case in which the conservation laws for momentum
and energy can be put down in a simple form:
L mr•,yrk)vlk) = L mrk)y'll:)v'rk), (5.83)
L ml•)c2yfk) = L mr•)c2y'<kl. (5.84)
These equations are valid only in the case of fast particles
which interact weakly (or briefly). Eqs. (5.83) and (5.84) are not
valid during an interaction, but they are quite suitable before
and after it. In particular, they can be applied to ideal relativistic
gas and also to "collisions of microparticles".
Here is an example showing how the conservation laws in a
relativistic form are applied in the study of particle "collisions".
Let the particle of the mass m0 strike on a resting particle of the
mass m 1• The total mass of particles generated as a result of we

"collision" ("reaction") is equal to M. Reactions between par-
ticles are governed not only by the momentum and energy conser-
Yation laws, but also by other specific laws of conservation. We
shall not take them into consideration here. We shall be assum-
ing, e.g. from experimental data, that a reaction may take place.
The momentum and energy conservatiqn laws make it possible to
clear up the essential question: what is the minimum energy of
a striking particle sufficient to bring about the reaction we are
interested in?
"Before" and "after" the reaction the momentum and energy'
conservation laws are complied with. In four-dimensional terms
it means that the 4-vector energy-momentum of a system of par-
ticles remains constant. We consider the situation "before" the
collision in a laboratory reference frame. Before the collision the
particles do not interact, and therefore the energy of the system
of particles is equal to /8 0 +m1c2 and the momentum to p0 , where
/5 0 is the total energy of a striking particle and p 0 is its momen-
tum.
The situation after the collision is convenient to consider in th:!
frame of the centre of inertia. In accordance with the momentum
conservation law this reference frame moves uniformly and recti-
linearly relative to the laboratory one and therefore is also iner-
tial (provided the laboratory irame is inertial). The minimum
energy required for the accomplishment of the reaction is realized
in the case when all particles generated after the reaction are
resting in the frame of the centre of inertia (otherwise their total
energy would be higher). Consequently, if the minimum energy
is sought for, the energy of a system of particles generated after
the collision is equal to M&, and the momentum is equal to zero
(in the frame of the centre of inertia). In transitions from one
IFR to another the square of a 4-vector energy-momentum is an
invariant.
Write out 4-vectors of energy-momentum for a system of
particles before and after the collision: po (p 0, ltofc + m 1c),
P (0, Me). But the absolute values of these vectors are equal to
Po·=P2 or M 2 c 2 =(8 0jc+m 1 c) 2 -p~. Making use of the rela-
tions f!Uc2 -pg=m~c 2 , /50 =T0 +m 0c2 , we get
M2c2 = m~c 2 + m~c 2 + 2m 1 (T 0 +m c 0
2).
Hence, the minimum ("threshold") value of the kinetic energy o£

i:l striking particle follows directly:
To= 2 ~ 1 (M-ffl{l-mt)(M +mu+ m 1).

Retafir.oistit Mechanics of a Particle 179
Th1s ;ormula can be used in the treatment oi various reaction<>.

We shall quote three examples.
The production of a :n-meson in a collision of two nucleons:
N + N-+ N + + N n. The photoproduction of a :n-meson at a
+ +
nucleon: N v- N :n. The production oi the proton-anti proto:-~
pair (p+ p) on a bombardment of a proton (hydrogen)·contain·
ing target with protons: p + p-+ p + + +
p (p p).
The interpretation of these reactions has to be based on the fact
that the energy conservation law is always observed. In conse-
quence, in these reactions the kinetic energy of initial particle:>
turns (partially) into the rest energy of the particles produced.
Speaking of the "production" of mass from kinetic energy is, of
course, inaccurate.
CHAPTER 6
THE MAXWELL THEORY

IN A RELATIVISTIC FORM
The theory of relativity shows how to consider physical pheno-

mena in any inertial frame of reference. The STR is based on
complete equivalence of all inertial frames. Therefore, the basic
~~~it ;~n:lldf:~;;fairfrf~~!;c~: f~~;s~~fn"~ai~h "r~}~;!n~eur:a ~e if~~~
are written in requisite variables, i.e. using length and time scaleS
of a given reference frame.
The basic system of equations describing electromagnetic phe-
nomena was provided by Maxwell. It is remarkable that the sys-
tem of Maxwell's equations formulated firty years prior to the
advent of the special theory of relativity proved to be covariant
with 1espect to the Lorentz transformation, i.e. it retains its ap-
pearance, with the accuracy of variables' designations, under the
Lorentz transformation. This signifies that the system of J\'\ax-
well's equations retains its appearance in any inertial frame of
reference, and the principle of relativity holds automatically.
Thus, the equations of electrodynamics require no modifications
tn terms of the STR, and it might seem that the theory of rela-
tivity cannot introduce anything or importance. It is however far
from being the case.
First or all, previous to the advent of the theory of relativity it
\\aS not clear in which reference frames the system or Maxwell's
equations was valid. The theory or relativity indicated at once
that this system or equations fitted any inertial frame or reference.
Hence, it was natural to rewrite the system of Maxwell's equa·
tions in a four·dimensional form. Such a notation makes it pos-
sible to find the transformation equations for the basic quantities
or the theory when passing from one IFR to another. Using a
four-dimensional notation, we shall also find the inseparable unity
or charges and currents, electric and magnetic moments, electric
and magnetic fields. Relationship between some other physical
quantities will be discovered as well. Such a close relationship
between definite physical quantities remained unnoticed until the
emergence or a relativistic approach to electromagnetic ph·-·no-
mena.
Maxwell Theory !11 a Relativistic Form 181
As to the transformation of electric and magnetic field compo-

nents on transition from one inertial frame to another, it can h~
carried out systematically only in terms of the theory of relativity.
It is the theory of relativity that indicates a four-dimension·tl
antisymmetric tensor to be used to describe an electromagnetic field.
§ 6.1. The three-dimensional system of Maxwell's equations.
A 4-potential and 4-current *· The Maxwell theory represents a
~f~r~~~~f~c at~e~r:ct~~r:!~c~~~i~afe\~ti~n fi:~d. a~~it:~~~r~;~~u;;;i~~
described by the four vectors: the electric field strength £, mag-
netic field strength H. electric field induction D and magnetic
field induction B.
In a uniform isotropic medium the number of field vectors need-
ed to describe electromagnetic phenomena is reduced to two sinci!
field vectors turn out to be proportional to each other:
(6.1)
The constant coefficients e and J.l are called a dielectric permit-

tivity and magnetic permeability respectively. Vacuum is describ-
ed as a uniform isotropic medium possessing the definite valu~s
of e and J.l which are customarily denoted by e0 and J.lo and r~
ferred to as an electric and a magnetic constant respectively.
According to the Maxwell theory the field vectors obey the two
principal equations:
rolH~J+D. I rotH~J+eE,
rolli~-il; (a) rolli~-"it; (b)
(6.2)
the equations for an arbitrary medium are written on the \eft-hand

side; the same equations for a uniform and isotropic medium are
seen on the right-hand side.
In the Maxwell theory the average values of an electric and a
magnetic field (relative to .. actual" microscopic fields) are defined
by the vectors E and B. In the general case the vectors D and H
are related to the average fields by the following equations:
D~e 0 li+ P, B~l'o(H +M), (6.3)
where two more vectors are introduced: the polarization vector P
and magnetization vector M.
The charge conservation law is assumed to be valid in the Max-
well theory; in the case of a continuous charge distribution it is
• All equations of electrodynamics are written in this chapter by means of
tht Sl units. For the reader's convenience the principal formulae are also-
written out in the Gaussian system of units in Appendix II.
182 Special Theory of Retafivily
written down as the continuity equation:

~ + diV/=0. (6.4)
Here p is a charge density and j = pv a current density.

From Eqs. (6.2) and (6.4) the two equations follow which arc
easily incorporated in (6.2) and (6.4):·
divD=p,
divB~O; (a)
I divH~O.
divE=p/e,
(b)
(6.5)
The force density for an electromagnetic field acting on free char·

ges and currents is assumed to be equal to
f~p{E+[vB]}. (6.6)
It is referred to as the Lorentz force. From this expression it is
seen once again that E and B are average macroscopic fields.
The system of Maxwell's equations can be written down not
.only by means of field vectors but also via the scalar and vector
potentials <p and A. We shall consider the case of a uniform and
isotropic medium and combine the potentials cp and A with the
fields E and B by means of the relations
E~-V~-A. B~rotA. (6.7)
Having substituted these expressions into Eq. (6.2b) and hav-
ing additionally imposed the Lorentz condition
(6.8)
on potentials (one can show this condition to be easily met), we

obtain the equations which hold in the case of the potentials (p
and A:
(6.9)
where
0=6.-
a2
I I
- - , V=--=· (6.10)
al
02 .Yell
2
In Eqs. (6.9) it is assumed lhat p = p(r, t), j = j(r, t), i.e. the
current and charge densities are the assigned functions of coort.!i-
nates and time. Eqs. (6.7) and (6.9) are equivalent to (6.2).
Next, we shall impart a four-dimensional meaning to the quan-
tities involved in the Maxwell equations and rewrite the Maxwell
·quations themselves in a four-dimensional form. However, we
shall proceed gradually, and the Maxwell equations (6.2) and
Maxwell Theory in a Relalivlsfic Form 183
(6.4) will be written in a four-dimensional form only in § 6.7. Now

we begin with composing four-dimensional quantities from th~
potentials fJl, A and the densities p, pv.
Eqs. (6.9) are differential equations of the same kind, the
d'Aiembertian equations. Accordingly, they can be written at once
as a single four-dimensional equation, provid~d that the two 4-vec-
tors are intro~uced· the 4-potential .vector ¢:1 and the 4-current
density vector s.
For some time we shall be simultaneously writing out defini-
tions and relations in the real and imaginary forms, just the way
we did it in mechanics. So, let us define the 4-potential vedor dJ.
as follows:
q;e: ~: ~: (i:;J (a) I ${~c ~: ~ ~ }. (b)

(6 II)
and 4-current vector
+{'' s s, s,}
5
2
j" j 9 jz icp ' (a)
I +{'" s' s s'}
5 cp j"
2
ill iz •
(b) (6.12)
Recall the definition of the 4-radius vector:

->{x,x zx,x,}
R x y
2
ict ' (a)
1+{,(>
R = ct T x
x'i'i'}
z ·
y (b)
The 4-vector components <I>, s, R are written in two parallel line">,

in a conventional and a symmetric notation respectively; the ne-
cessary component values are immediately found from the com-
parison of these lines. Having determined the 4-potential ~nd
4-current density, one can rewrite Eq. (6.9) for vacuum (i.e. with
e = e0 and ll = 110. where c 2 = 1/(ewo)) as a unified formula:
0 <!>,=- ~oS.(k= I, 2, 3, 4), (a) I 0 <!>'=- ""'
(k=O, I, 2, 3). (b) (6.13)
Obviously, the three equations of (6.13a) at k = 1, 2, 3 coincide
with the three equations of (6.9) in the case of vacuum. At k = 4
Eq. (6.13a) gives ofqJ=-J.I.oicp, and since c;2 = 1/(toJ.I.o), we
obtain Eq. (6.9) again.
We suggest that the reader make sure that Eqs. (6.13b) also
coincide with Eqs. (6.9) in vacuum.
In the case of vacuum the Lorentz condition (Eq. (6.8)) and
the charge conservation law (Eq. (6.4)) can be written via the
184 SpecklE Theory of Relativity
4-divergence of the vectors Ill and s. Indeed, for example,
-divtil= ~~/ = ~~11 + ~~12 + ~~: + ~~44 =

= ~:ll + aa:ll + ~~z + aa(:~~) =divA;-Ql,
div;= ::: = ~; + ~: +*+ ~~i1:;: =divl+~-
Consequently, the Lorentz condition a~ the char~e conserva-
tion law in vacuum have the form div Ill = 0, div s = 0, when
presented in a four-dimensional notation. Surely, the same results
are obtained on the basis of the usual three-dimensional approach.
The conclusions just drawn are very significant. As it is shown
in Appendix I, § 4, the 4-divergence is the Lorentz transformation
invariant. As to Eqs. (6.13), these 4-vector relations are valid in
any inertial frame of reference in vacuum. Thus, the equations for
potentials, the Lorentz condition and the charge conservation law
can be so rewritten that it would become evident at once that they
retain their appearance in any inertial frame of reference. The
covariant notation of equations for potentials in the case of a
refractive medium (e +
l!o and f.1 +
f.lo) will be discussed below
(see§§ 6.14, 6.15).
§ 6.2. The transformation of a 4-potential and 4-current. The
very fact that we have managed to compose the 4-vectors $ and;
makes it possible to write down immediately the transformation
equations for components of these vectors. We shall write theo;e
equations here in both the real and the imaginary form (cf. Eq.
(4.10a, b)):
ltl,=l'(ltlf-iBltlf), ltll=ltll, ltl3=ltli. <ll•=f(lD4+iB$1);
(6.14a)
<D'~r(<D' +B<D''), <D'~r(<~>'' +B<D''), <D'~<D', <~>'~<~>'':
(6.14b)
s 1 =f(s;-iBs~). s2 =s;, s3 =s;. s 4 =f(s~+iBs;): (6.15a)
sO=f(sO'+Bs 1'), s 1 =f(s 1'+Bs0'), s 2 =s'~', s3=s3'. (6.15b)
Let us look more carefully into the current density transforma-
tion. A 4-current comprises a current density and a charge den-
sity. It is quite natural that a current and a charge density are
combined into a single 4-vector. Dealing with reference frames
moving relative to one another, one should bear in mind that a
-charge may be at rest only in one ("proper") reference frame. In
all other IFRs the charge moves, being in terms of these frames
Maxwell Theory in a Relativistic Form I~
nol only a charge but also a current. We see, therefore, how easy
is the transition from a motionless charge (electrostatics) to a
moving one (current): it is just a transition from the charge's
proper reference frame to any other IFR. When a current origi-
nates due to charges displaced together with a moving medium or
objects, it is called a convection current. It is just a convection
current that turns up on transition from a proper frame to an
arbitrary IFR.
The formula j = pv contains the density of charges moving at
the velocity v. Otherwise, misunderstandings may arise. For
example, a current flows in metals, even though p = 0. Indeed,
the total charge density in metals comprises those of ions and free
+
electrons and is equal to zero: p = P+ P- = 0. But certainly,
a current can flow provided there is a regular motion of electron:.:
j = P+V+ + p_v_ = p-v-. because the velocity of a regular motion
of ions is equal to zero.
From Eqs. (6.15) we obtain directly a convection current origi-
nating on transition from the charge's "proper" frame. Thus, let
the frame K' be characterized by a given charge density p' and
the absence of a current (i' = 0). Consequently, in the frame K'
the 4-current density has the components ;' (0, 0, 0, icp' = icp0 ),
i.e. s~ = s~=s~ = 0, s; =lcp0. Then in accordance with Eq. (6.15a)~
for example, in the frame K
s,=r(-iBicpo)=rVPo, s2=s3=0, s4=ricflo. (6.16)
In the developed form the last relation of Eq. (6.16) has the fol·
lowing appearance:
s4 ;:=::;;; icp = 1/1 ~P~Itc• ,
This way we obtain the charge density transformation law on

;ta~!!t0 fo f~~~ lrhaemceh~~y:;rv:Pr~p!r~:1~a~:· ~~a~;~~h ~~!~g:~ ~~!
velocity V:
p= -./l!:...ovitci, (6.17)
The first relation of (6.16) yields a current density
s,=i~= 1/1~~2tc~" =pV. (6.18)
As we have mentioned, a current associated with the motion of

a charged medium or charged object is referred to as a convec-
tion current.
The meaning of Eq. (6. 18) is very simple. The velocity of a
charge resting in the frame K' is equal to V with respect to the
frame K {that 1s the velocity of the frame K'; the same follows
from the relativistic velocity transformation formula (3.27)). There-
fore, Eq. (6.18) represents a convection current. As to the charg~
density transformation (see Eq. (6.17)), it is associated with the
volume change (since a density is a charge of a volume unit).
Since a volume transforms according to the law
d7"~d'll' 0 yl- V'lc',
and we consider the same physical volume containing the sam.:-
charge de, then
Po= d~o' and p= : ; = .,f!-IV2Jc2 d~o = v'l _P~2fc2 =por.

Of course, the total charge in a given Yolume remains constant
in any reference frame:
(6.19)
Eq. (6.19) expresses the invariance of a ch!rge confio~d in a
given volume. Using Eq. (6.17), the 4-vector s can be expressed
otherwise. Let us consider a small volume element of a moving
medium in the frame K. Then in the reference frame K0 co-moving
with this element the velocity of the clement v = 0 and p = P"·
In lht> frame K}he density p = p~y and, co~sequently, j = pv =
= piyv. Then s(p0yv, p0 icy) = paV, where V is the 4-ve\ocity of
an element of a medium. Thus,
-> ->
s = PoV, S1 = PoU1. (6.20)
For s 2 we have the following form:
(6.21)
The 4-vcctor s is a time-like vector because a charge velocity u

is always less than c.
If in the frame K' there is an uncharged conductor through
which a current flows, i.e.
(6.22)
in the frame K', a certain charge density p is observed in K. In~
deed, according to Eq. (6.15),
s,=fixo• S2=iJIO• Sa=iz0, St==icp=fiBixo• (6.23.)
The first three formulae of (6.23) define the current magnitude
in the frame K, the last one defines the charge density in the
Maxwell Tl!eary in a Relativistic Form 187
frame K:
(6.24)
Consequently, an observer in K will find the density p, even

though the charge density in the frame K' is equal to zero. This
result can easily be interpreted in geometric terms. Let the con-
ductor be at rest in the frame K; the ions of the conductor are
motionless and the electrons move at some average velocity u. In
,
('~ctro_ns_tun_e=f
Fig. 6.1. The Minkowski diagram illustrating the emergence of a charge density
in the frame K' in a current conductor which has a zero charge density in the
frame K. The total charge density caused by the ions and electrons in a conduc-
~~fthislh~q~:~s 1 be~~~0 in~~~~j~~si 5T~:n;~~~~d ~~~~~ ~( 1~: i:~ti~~e d~p1i~~e: 11by1 rd~r:
led lines, and the world lines of the electrons by slanted continuous lines. Be-
sides the reference frame K (with lhe x, t axes), also shown is the reference
[~!mxe :;:d <;}'~x~~e i;'·a~o a~~:~.na i~c~~ee hl&~~~-ot;n~~~~~~ -~~il=r~h:~~~e;:~s~l~
must be determined simultaneous[)' at all points, it has to be determined at all
~~~~!n~f i~h~h~b~~~~nKI~~n~a~~sx a~x~~u~js~~~b::Y~r'~o~s s~d ~~~~:~n~, u~i~~~:
tary segment in the frame K' will contain more ions than electrons. This fact
signifies the emergence of a positive charge density in the frame K'
the frame K the world lines of the ions are straight lines parallel
to the t axis while the world lines of the electrons are straig~tt
lines forming a certain angle a= arctan(v/c) with the r axis.
Fig. 6.1 shows the reference frames K(x, r) and K'(x', r'), the
world lines of ions (dotted lines), and the world lines of electrons
(thin continuous lines inclined to the "taxis at an angle 9). Inas-
much as metal is neutral on the average, each segment of the con-
ductor must emit an equal number of world lines of ions and
electrons. The charge density should be measured simultaneously
in each reference frame. In the frame K it is determined by the
number of world lines of ions and electrons crossing a unit of
length in this frame. For example, the charge density is defined
188 Spec-Ial Theory of RelallvUy
by the numller ul world lines of ions (taken wtth lilt: :.1gn ~+")
and the number of world lines of electrons (taken with the sign
"'-") going through the segment OA. A scale hyperbola cuts the
unit segments on the x and x' axes. It is these unit segments that
the charges must be related to. But in the frame K' the C(Jarge
den .. tly must be calculated for the whole conductor simultaneously.
In the frame K' simultaneous events ~re located on straight lin.•s
parallel to the x' axis, and on the x' axis itself. in particular. It
is seen, however, from Fig. 6.1 that the unit segment OA' accom-
modates more positive charges than negative ones. Accordingly,
the conductor will turn up to be charged positively in the frame
K' although it is neutral in the frame K. Surely, if one considers
a closed-type conductor in the frame K', its total charge will
remain c4ual to zero, but an electric dipole moment, not found
in the frame K, will be observed in the frame K' (see § 6.9 and
f,g, 6.4).
§ 6.3. An electromagnetic field tensor. In electrodynamics the
electric field strength E and magnetic induction B are convenient 4
ly expressed via the vector and scalar potentials A and 'P as fol-
lows*:
B- rot A, E- - grad~- OA/at. (6.25)
}he: ~~e~:~tt~~hesshaliqu~~:r;s o~fint~e 4 -ie~;~7~~~1 ~~m&~nec~t~P~~~

4-sp~ce:
Bx==BI = ~~z- aa:Y = ~~:- !~,2 • (6.26)
Ex==£1 =-~- a~x =-f ~~11 - ~~: ic=ic ( ~~;- ~~11 ).
(6.27)
The last terms of Eqs. (6.26) and (6.27) are written on the
basi~ of the definition of 4-potential components. Similarly, using
the <D components, one can write down the remaining components
of the vectors E and B as well. We shall obtain equations similar
to Eqs. (6.26) and (6.27) from which it follows that all compo-
nents of the vectors E and B can be expressed via certain combi-
nations of derivatives of the 4-vector iiJ components with respect
• In previous chapters we utilized the letter B to denote the ratio of the ve-
~~\~Y i~~~c~i~~r~~t~r sj,st:~ ~~sthpar~t~nf.~: ~~~rB;cBl~a'fo t~~;~~~o~t~c1t~n~=r~~
we shall not be using the designation B = 11/c in this chapter, apart from thP
cases when misunderstandings are ruled out
Maxu;el/ Theory In a Re/allvlsiic Form 189
to tour-dimensional coordinates. These combinations form the

antisymmetric 4-tensor of the second rank •:
Fa=c ( --ax;-- atll

atll•
ax• 1) •
(t, k= I, 2, 3, 4). (6.28)
Prior to discussing mathematical features and the meaning of

Eq. (6.28) we should examine the same transition in the real
4-space. As we have indicated, one has to differentiate between
covariant and contravariant components of vectors and tensors
in this case. The electromagnetic field tensor (Eq. (6.28)) is ex-
pediently written in covariant components. Then, if the vector cD
has the contravariant components (ll>o, A), its covariant compo-
nents will be (¢1°, -A). The differentiation with respect to con-
travariant coordinates leads to covariant components again (:>ee
Appendix I, § 8). Thus, Eq. (6.26) only changes the sign while
Eq. (6.27) is written as
E~ == £ 1 = - c ~~1° + c ~~~ = c( ~~~ - ~~~ ) , (6.27')
taking into account that q:. = cll> 0 , A:.:= -11> 11 t = xfJ/c, x = x 1•

Thus, in the case of the real 4-space we introduce the covariant
antisymmetric 4-tensor of the second rank
(6.28')
coinciding in its appearance with Eq. (6.28). The purpose of the

introduction of Eq. (6.28) or (6.28') is very remarkable. The two
Maxwellian field vectors E and B can be expressed in the 4-spacc
uniquely through a certai~ combination of space-time derivatives
of the 4-vector potential Ill. The behaviour of the quantities F,.
on transition from one IFR to another is most significant for us.
Their transformation, however, is very easy to find: the quanti-
ties F,k form a tensor since one can readily make sure (see Ap-
pendix I,§ 3) that derivatives of 4-vector components with respect
to coordinates transform in accordance with the tensor transfor-
mation rule. If the indices i and k in Eqs. (6.28) and (6.28') ac-
quire independently all values from I to 4 (and from 0 to 3 re-
spectively), we obtain 16 values of F,~ (four of which are equal
to zero) expressed through components of E and B. Let us write
• Just as in most books on relativistic electrodynamics. we use the subindl-
:u1rr~~: ~~~~i;i1de ~~~~;c~n~ 8~v~~e ai;a~:tnWl h~~~Yfha~~ 0th1i~si!~~ ~~~~ b{e!'d k~p!
misunderstanding.·
190 Special Theo•y of Relativity
down these components in the form of matrices:
F,, ~ ( - c~, cg,

cB, - cB,
-~=: =:~:)
- i£, ' (a)
iE~ i£11 iE, 0 (6.29)
F,k= ( - '~• 'cB~g• - ;~:

- c£11 0
;~:)
- cB" ' (b)
-cE" -c811 cB:c
0
We see that the components of an electric field strength and
magnetic induction are the components of one 4-tensor of an elec-
tromagnetic field. As usual, in the designation F,k the first sub-
index i denotes the line and the second k the column of the
matrix F,,..
Frequently the tensor (6.29a) is abbreviated as ij = (cB, -iE) ~
and the tensor (6.29b) as ij = (E, cB), assuming the components
of the vectors E and B are arranged as in (6.29a) and (6.29b)
respectively.
We have obtained the result which is dissimilar to the usual
three-dimensional case. The Maxwell theory deals customarily
with vector fields. In fact, the vectors E and B behave as 3-vecto:-s
as far as the transformation of a coordinate system, that is tOe
rotation of coordinate axes, is concerned. As soon as we pass
0\er to reference frames moving relative to one another, the sit-
uation changes drastically. In 4·space E and B are no long~r
vectors, not even four-dimensional ones. Although the vectors E
and B are expressed via components of a four-dimensional poten-
tial, there are no values to be added to the three-dimensional
vectors E and B in order to make them become 4-vectors. In
4-space an electromagnetic field is a single quantity of a mon~
complicated mathematical nature than a 4-vector. The fields E and
B have merged into a single 4-tensor which is referred to as an
electromagnetic field tensor.
The emergence of a single 4·tensor instead of two three-dimen-
sional vectors describing an electromagnetic field has a clear·cut
physical meaning. Electric and magnetic fields are intertwined <>O
inseparably that the "appearance" or "disappearance" of one of
the fields is determined by the choice of a reference frame. For
example, a "pure" ('lectric field generated by a charge occurs
under very speci!ic circumstances when the charge is considered
• The Gothic lett(" 5 dl'notes 4·force components in Chapter 5 In this chap-
ter the same letter 1s 1 ,._.d r'clu:.ivcl): [or the designation or a tensor.
Maxwell Theory In a Relativistic Form 191
in the frame in which it is at rest. In any other inertial frame,

however, this charge moves and consequently generates an elec-
tric current producing a magnetic field. On the other hand, we
have seen that although in a certain reference frame a current-
carrying conductor appears neutral, in other inertial frames of
reference it appears charged, and consequently, an electric field is
due to occur in these frames.
Thus, it is sufficient, for example, to have only an electric field
in the frame K for a magnetic field to appear in any other frame
K'. If in the frame K there is only a magnetic field, both magnetic
and electric fields will appear in any other frame K'. Had we at-
tempted to treat the fields E and B as vectors, that physical fact
could not have been expressed in mathematical terms. As we have
mentioned, there are no values to make the three-dimensional vee·
tors of an electromagnetic field become 4-vectors. Moreover, had.
each of the vectors E and B been contained in "its respective"
4-\•ector, the Lorentz transformation would have necessitated each
of these vectors in a "new" frame to be expressed through com-
ponents of "its respective" vector in an "old" frame. In such a
wav, the \'ectors E and B would turn out to be unrelated. Practice,
hO\\·c\'N, indicates an intimate connection between electric and
magnetic fields, that is bet\\"een the \ectors E and B.
(wo three-dimensional vectors possess six independent compo-
nent:.. An antisymmetric 4-tensor of the second rank possesses
exactly six independent components. We have found (see Eq.
(6.29a, b)) that the f1elds E and B form an antisymmetric 4-ten·
sor, an electromagnetic field tensor. Inasmuch as any component
of a tensor in a new reference frame is a linear combination of all
components of that tensor in an old reference frame, a transition
from one reference frame to another mar result in the appearance
of an electric field due to a magnetic field observed in another
frame, and vice versa. In a certain sense, an electromagnetic field
is a closed formation: if some inertial frame has no electric or
magnetic fields, an electromagnetic field will not appear in any
other inertial frame. We shall deal with the transformation of
electromagnetic field components in the next section while here
we shall examine brieRy an electromagnetic field in matter.
To describe a field in matter, one has to introduce, in addition
to the average fields E and B, two more vectors. These can be
either the electric induction vector D and the magnetic field
strength H, or the electric polarization vector P and the magne-
tization vector M. These four vectors are interrelated as follows:
D-•oE+P. H-B/~ 0 -M. (6.30)
The vectors H and D form a special tensor whose components
are customarily denoted by /111 while the tensor itself is abbrevjated
as f = (H, -icD). This tensor is derived from (6.29a) by substi-

tuting H components for cB components and -icD components
for -iE ones. Here is the matrix of this tensor components:
( -H
0 H,
0 H 11
-H, -icD,)
- icD 11
fa= Hy:r -H"' (6.31)
. 0 -icD:r. ·
icD" icD 11 icD:r. 0
In addition to the tensor f. it is also useful to introduce a tensor
of electric and magnetic moments of matter whose definition fol-
lows easily from Eq. (6.30):
m;~ = · /J;_ F,~~,- ft •. (6.32)
'V "'
It is abbreviated as follows: !In= (M, icP). Written in full it is
ma= (-~. M -M"

~· -:: icP,)
icP 11
icP11 •
(6.33)
-tC;z -uP 11 -icPz 0

It is the very fact of the origination of tensors (6.29). (6.31) and
(6.33) that points to a close pairwise connection between the
quantities E, B; H, D and M, P. We have written out here the
tensors f,,. and m,,. only for a 4-complex space; the corresponding
expressions for a real 4-space can be easily derived by the reader
himself.
§ 6.4. The transformation of electric and magnetic field compo-
nents. The four-dimensional approach is particularly convenient
because as soon as the mathematical nature of one or another
physical quantity is established (a scalar, 4-vector, 4-tensor), the
problem of its transformation on transition from one IFR to
another is solved automatically. In mechanics we dealt with
4-vectors. As we have established, components of the fields E
and B, H and D, and M and P are components of tensors
(6.29a, b), (6.31) and (6.33) respectively.
Consequently, components of three-dimensional vectors are
transformed according to the role of tensor component transfor-
mation. For example, the components F,,. in a 4-complex space
are transformed as follows:
(6.34)
where a,,. are components of the Lorentz transformation matrix

(2.41a) while components F,,. are defined by Eq. (6.29a). In order
Maxll'ell Theory in a Relativistic Form
"'
to transform components of matrix (6.29b), one has to make usc
of the Lorentz transformation matrix in the form (2.4lb).
Here we ought to make a purely methodological remark. Lec-
turers often avoid using tensors trying not to complicate a lecture
course. Indeed, explaining the meaning of tensors and their prop-
erties in the course of half an hour is a difficult task. One can-
not, however, disregard the fact that an electromagnetic field is a
tensor. The old question arises again: "shall we call a cat a ca~"
~0~ nt~~;e:; ~he!~~~~~f~r~~tT~~r=:~!\i~sn n~.3s4).mA;;a~~~t~a:~~~
equation can and should be obtained in the simplest way possibl~.
For example, it is easily derived as follows: we see from Eq. (6.28)
that the quantities F," are linear combinations of 4-vector compo-
nent derivatives with respect to 4-coordinates; the transformatio'l
of vector component derivatives due to the coordinate transforma-
tion follows from the simple analysis (see Appendix I,§ 3).
To memorize the transformation rules for tensor components,
one should keep in mind that they transform as a product of cor-
responding vector components. One way or another, we get Eq.
(6.34). And here it is the right time to call a tensor a tensor, hav-
ing disclosed, rather one-sidedly, of course, the meaning of tensor
components by means of vector component derivatives with re·
spect to coordinates.
We shall give an example of how the field transformation equa-
tion can be obtained for the case Bz. = F12 /c. According to Eq_
(6.34) the transformation equation for F 12 has the following form:
(6.35)
Recall that the summation is carried out here over the two in-
dependent pairs of the indices m and l, each of which runs from I
to 4. This way, Eq. (6.35) involves the sum of sixteen terms, each
being a product of two a, 11 and one of F, 11 components. We urg~
the readers who come across such equations for the first time to
write out (once in a lifetime) all sixteen terms. Here is the easiest
way to do this. First, we develop the sum with m taking the val-
ues I, 2, 3, 4. As before, the index l denotes summation. This way
we obtain a sum consisting of four terms in which the index m
is eliminated. Then we perform the summation over l in each of
these four terms. As a result all sixteen terms will be written out.
Then one should 5ubstitute a,~~ from the Lorentz matrix (see Eq.
(2.4la)) and components F, 11 from Eq. (6.29a) into these terms.
One can see at once that most terms of the sum (6.35) are equal
fo zero. Because of' this the summation in Eq. (6.35) can be much
s1mpler. Indeed, the q1,1antities a 1111, with m running from I to 4,
constitute the elements' of the first line of the Lorentz matrix (see
7-97
194 Special Theory of Relafivity
Eq. (2.4la)) while the second line is made up of the quantities cx 21

with I = I, 2, 3, 4. But the first line of the matrix has only two
elements, o: 11 and o: 14, which differ from zero. Consequently, one
must consider only the values of m equal to I and 4. The second
line has only one element differing from zero, o: 22 = I. Conse-
quently, one has to take only l = 2 and to rewrite Eq. (6.35) as
F 12 === cB 2 =a 22a 1 mF~ 2 = a 1 mF;,. 2 =a 1 jF~ 2 +a 14 F; 2 =
V cB;+~<.
~r{cB;-i-(iE:,)}~
c "'
~,
v i - V2jc2
lntercomparing the second and the last equation in this chain of
equations and dividing them by c, we get
8~+~~ v
Bz= .Yt-cV2Jc2 =r(a:+~t;}
In much the same way we obtain the transformation equations
for the other components:
E,~E;, E,~f(E;+VB;), E,~f(E;-vB;);
Bx= B~, 8 11 = r ( 8~ -{rt:), B:e =f( B:+*E~). (6-36)

Let us write out the transformation equations forD and H to be
used later on:
D.= o:. D!l= r ( D~ +f.- If:). D:e = f ( D:-7 H~); (6_37 )
H,~H;, H,~f(H;- VD;), H,~ f(H;+ vo;).
Exactly the same Eqs. (6.36) and (6.37) are of course obtained
in the real 4-space. We shall not mention this space any more be-
cause hereinafter we shall only use the final equations and they
are identical; moreover, substantial difference is noticed only in
transition from Eq. (6.27) to (6.27'). What follows is a simple
matter.
It is seen from Eq. (6.36) that all field vectors change their
magnitude and direction on transition from one inertial frame of
reference K' to another K. Only "longitudinal components" remain
invariable, i.e. components along the direction of the relative mo·
lion (along the x axis).
Let us expand the electric and magnetic fields E and 8 into
the components parallel and perpendicular to the motion direction
(the unit vectors i, j, k are directed along the x, y, z axes re 4
spectively), e.g.
Having noted that the velocity vector V of the coordinate system

K' has the components (V, 0, 0), we obtain
(VB'l ~I 8:~ 8~
:I~- JVB;+kvB;~ v (-JB; + kB;).
8~
(VE'(~ V(-JE;+kE;).
Then Eq. (6.36) can be rewritten in the vector form:
E 1 ~(E' -(VB'() 1,
B 1 ~ ( B' +-;'.- (VE'J) 1 •

E, ~r(E~ -(VB'l),;
BL ~ f ( B~ + * (VE'J) L. ( 6.38)
It is appropriate perhaps to recall here that a \I expressions of the

~~1~cfJ:AJjt~rteh:~~~~ot~ pzre:3u~til:e~fxr;re~s~~nl_ ~h~h~e~~~e [t~:Js~
formation equations are obtained by substitution of unprimed
quantities for primed ones and vice versa, and by changing the
sign of V to the opposite:
E;~(E+(VB() 1 , E,~l'(E+(VB(),;
(6.39)
B;~(B--;'.-JVEJ),. B~~r(B--;'.-(VEJ),.
In the case of non-relativistic velocities r ~ I, and we obtain

from Eq. (6.38)
E~E' + (B'Vj, B~B' --;'.-lE'Vj. (6.40)
The following designations are used: E = E 11 EJ. and B = +

= B 11 + B1.. The equations of reverse transformation from K to/('
are obtained as usual by substitution of unprimed quantities for
primed ones and vice versa with a simultaneous change of sign
of V;
E'~E+(VBJ, B'~B--;'.-(VE(. (6.41)
In conclusion let us write out the transformation equations for

D and H. One may not compute anything; it is sufficient to reca\1
that we have obtained the transformation equations for the com-
ponents of the tensor 5" = (cB, -iE), and now we are interested
in analogous equations for the tensor f = (H, -icD). Instead of
7'
196 Spectal Theory of RelatliJity
Eq. (6.39) we obtain the following expressions for the correspond·

ing components·
o; = (o+ ?- JVHJ) •. D'.L =r(D+ *!VHJL: (S.4 2)

n;~cH- [VDJ),, H', ~r(H-[VD[),.
In the case of non-relativistic veloctt.ies, when r ~ I, Eq. (6.41)

turns into
D'~D+,'.[VHJ, H'~H-[VDJ. (6.43)
Suppose that in the frame K' the magnetic field B' = 0. Then
in the frame K the relationship between E and 8 becomes very
simple. First of all, notice that [VE] = [VE.1] since [VEu] = 0.
From Eq. (6.38) we obtain
E=E;+rE'.L,
B~r+. [VE'J~-[,-[VE~]~,'.[V, fE~)~ (6.44)
~-f,-[V, Ei + f'E~]~,'.[VEJ,
Similarly, if in the frame K' the field E' is equal to zero or in the
frame K the field E is equal to zero, then
E~-[VBJ, E'~[VB'J, (6.45)
In both cases and in any inertial frame the fields turn out to be
mutually perpendicular. It follows from both the relativistic equa-
tions {6.38) and the approximate equations (6.41) for low veloci·
ties that if in one of the frames (say, K) an electric or a magnetic
field is equal to zero, electric and magnetic fields in all other
inertial frames of reference are perpendicular to each other. The
same result can be obtained by employing the Lorentz transforma-
tion invariants (see§ 6.5).
If the fields E' and B' are mutually perpendicular in a reference
frame K', there exists a reference frame K in which one of the
fields disappears. It will be shown in § 6.5 that the expression
c 8 2 - £2 remains invariant under the Lorentz transformation.
2
Consequently, if the condition c28' 2 - £' 2 < 0 is satisfied in the
frame K', one can obtain a purely electric field through the ap-
propriate choice of a reference frame, while in the case of
c2 8' 2 - £'2 > 0 one gels a purely magnetic field. We shall show
how to find the velocity V of the reference frame K. Suppose this
veloc1ty is perpendicular to 8' in the 'case of c28' 2 - £'2 < 0 and
to E' in the case of c28' 2 - £' 2 > 0. Then 8 11 = 0 in the form~r
case and Eu = 0 in the latter case. Now one has to ensure that
81. =0 in the former case; in order to do this, the following con-

dition should be met (see Eq. (6.38)):
B~ +
(1/c')(VE'J... = 0.
Multiplying both sides of this expression by E' vectorwise and
taking into account the relations 1VE'J.1. = [VEJ.J.l. = (V.B'.1.),
(E' (VEJ.]].l. = VE' 2 , BJ. = B', we obtain the reference frame velo-
city v
V = - (c'/E'') (E' B'(. (6.46)
In much the same manner, we obtain in the other case
V = (1/8''1 (E' B']. (6.47)
One can always find such an inertial frame of reference in which
an electric and a magnetic field are parallel to each other at a
given point (see, however, the com-
fj fi"
ment on light waves at the end of
§ 6.5). Obviously, provided there
exists one such frame, there should
be an infinite number of frames pos-
sessing the same property. In fact,
in any inertial frame of reference
K' moving rectilinearly and uniformly
relative to K in the direction coincid- ,. '
ing with the common direction of E
Fig. 6.2. A transition to the ref-
and B, the fields E' and B' will erence frame K in which an
remain parallel since the field com- electric and a magnehc field turn
ponents oriented along the motion out to be parallel.
direction do not vary.
In order to find at least one frame in which the fields are panli-
Jel, we shall proceed as follows. Suppose that the fields are paral-
lel in the frame K, i.e. rEB]= 0. Direct the velocity of the franll~
K' (in which the fields E' and B' are not parallel any more) along
the perpendicular to the fields E and B; assume that the x, x' axis
is directed along the velocity V (see Fig. 6.2). Then Ex= Bx = 0
and, since the vector cross product is equal to zero, E11 B:z- EzB 11 =
= 0. Substituting into this equation the components of E and 8,
expressed via the components of E' and B' according to Eq. (6.36),
we arrive at the following equation:
J'(s +VB:) I' ( B; +f.E;) =I' (S- VB;) I' ( B;- f. s)

The frame velocity V can be determined from this equation using
the given fields E' and B'. Taking into account that according to
Eq. (6.36) £~ = B~ = 0, v.e can immediately find the direction
198 Special Theory of Relafivily
of the velocity V relative toE' and B'. Indeed, [E'B']=i(E~B~

- E~B~) and V = V·i, so that solving the foregoing equation, one-
can write
V/c 1 [E'B']
1 + V2jc2 = - c2B'2 + £'2 · (6.48)
Thus, from the giv~n vectors E' anq B' in the frame K' one can
find the frame K in which E and B are parallel. The velocity di-
rection of this frame coincides with that of [E'B'], while the
'.'elocity magnitude is one of the roots of the quadratic equation
(6.48). Surely, from the two roots of Eq. (6.48) one should choose
the one for which V < c. The case E'B' = 0 was examined abov~.
one cannot obtain parallel fields here, but it is possible to g·~t
either a purely magnetic or a purely electric field.
§ 6.5. The electromagnetic field invariants. Although an elec·
tric field strength E and a magnetic field induction B vary under
the Lorentz transformation, there are some combinations of these
fields remaining invariable under it. These quantities are invari·
ants of antisymmetric 4-tensors of the second rank. We make use
of two such invariants (see Appendix I,§ 6):
li=F;k' 12=Fo,F:~o=~elktmF,_Flm'
Recalling the definitions of the tensorsF1~o and Fi~o
jj (cB, - iE), ~· (- iE, cB)
and taking into account that the first invariant is the sum of the
components F;~o squared and the second invariant represents the
pairwise products of the corresponding components of the tensors
F 1 ~o and Fi~o. we can write at once f 1 = 2(c 28 2 - £2), / 2 =
~ -2ic(BE).
Omitting immaterial constant factors, one can claim that an
electromagnetic field possesses two invariants (we shall not write
out the invariants of the tensor ~ and the combined invariants of
S: and f since they will not be needed):
11 =c 2 B2 -£2, 12 =BE.
From the existence of these! two invariants follow the results,
some of which have been mentioned before. If in some IFR the
fields E and Bare mutually orthogonal ([EB]= 0), they are also
orthogonal in any other inertial frame of reference. If in some
reference frame E = cB, this relationship holds in all inertial
It should be noted here that both invariants are equal to zero
for a light wave in vacuo. These properties, i.e. B .l E and cB=E.
are maintained in any I FR.
Maxwell Theory In a Relallvlslic Form 199
It is clear that if / 2 = 0 and / 1 =1= 0, one can always find a

reference frame in which either E = 0 or B = 0 (depending on
the sign of / 1), i.e. pass over to either a purely magnetic or a
purely electric field. Conversely, if either E or B is equal to zero
in some frame, these fields will be mutually orthogonal in all other
inertial frames. Note that the quantity BE is not a "real" scalar
since it changes sign on transition from the left coordinate system
to the right one and vice versa, while the quantity (BE)2 is a real
scalar.
§ 6.6. The Lorentz force. Now let us consider the forces acting
on electric charges in an electromagnetic field. To avoid confu-
sion, we shall confine our presentation to spatial distribution of
charges*. In a co·moving reference frame K' in which a con·
sidered space element rests together with a charge, this charge
experiences a force exerted by an electric field (a magnetic field
does not act on a charge at rest). The force acting on a charge
contained in a unit volume is referred to as a force density. If a
charge density in a co-moving reference frame K' is equal to p0,
the force density f' is defined by the equation
f'-p,E',
where E' is an electric field strength in K'.
The transition to any other IFR is associated with the variation
of the fields E and B; even if in a co·moving frame there was no
magnetic field and only an electric one was present, a magnetic
field will appear in any other IFR. Let us find the force density f'
expressed via the field components E and B in an arbitrary iner-
tial frame. First, let us examine the case of non-relativistic vela·
cities when r R: I; then according to Eq. (6.17) p = p0 f R: pJ,
and according to Eq. (6.41) E' = E +[VB], and because of this
f- f' ~ p0E'- p {E +[VB[). (6.49)
The last link of Eq. (6.49) defines the quantity which is usually
called the Lorentz force density in electrodynamics. The Lorentz
force defines the force acting on a unit volume containing a
charge; this force is generated by the electric and magnetic fields
of the frame K relative to which the charge moves at the velocity
V. It is not surprising that the force f' in the frame K' turned out
to be equal to the force f in the frame K since according to Eq.
(5.34) a force magnitude does not change on trt~nsition from one
IFR to another in a non-relativistic case.
Of course, Eq. (6.49) can also be used in the case when the
velocity of charge motion is different at various points in space.
In this case each element of space will have its own co-moving
• Point charges are discussed, ior example, in (8], § 29.
Speciol Theory of Relo.Uuily
"'"
reference frame and, consequently, the velocity V will be different
at various points.
Let us derive the expression for the Lorentz force by still
another method illustrating explicitly how Eq. (6.49) comes about.
Let the co-moving frame K' have an electric and a magnetic field
defined by the vectors E' and B'.
Making use of the 'Superposition prirtciple, we can describe each
of these fields as a sum of two fields:
I E]=O, B]=B';
II E2=E', 82=0.
Obviously, the initial field represents just the sum of the two
+
fields: E' = Ef E2, B' = B[ + Bi. However, the field transfor-
mation equations I and II are very simple and allow us to get th~
answer right away. In the frame K'
f = p,E' = p,E;. (6.50)
Using the first equation of (6.45), one can write down imme-
diately the electric field I in the frame K:
E,=-]VB,],
where 8 1 is the magnetic field in K. According to Eq. (6.36) the
electric field II in the frame K is equal to
E2 = Ef:d + r (E2vi + E~zk) ..... E~
in the case when r ~ I. The total electric field in K is equal to
the sum of E 1 and £2:
E=E,+E,=E;-]VB,]. (6.51)
inihc~:;!~~tifh~e~~c!/ncr~;s f~~~~~ [sv~Y~ [~B~]1 .t[:B2]~~

see from the second equation of (6.44) defining 8 2 that the prod-
uct I[VB2JI ~ (V 2/c 2 ) can be ignored in a non-relativistic case.
Therefore, [VBt] =[VB] and we obtain the Lorentz force (see
Eq. (6.49)) [rom Eq. (6.51).
If in the frame K an electric field is equal to zero (E = 0) and
a magnetic field differs from zero, it follows from Eq. (6.45) that
E' =(VB']; so, the Lorentz force, appearing to be produced b:>l
a pure magnetic field in the frame K. seems to be produced in a
co-movmg frame K' by a pure electric field. These examples show
once again a uniqueness of an el('ctromagnetic field and the rel-
ativity of its division into an electric and,a magnetic field.
A few words on lines of force of the field are relevant here. In
£>ach reference frame a vector field can be correlated with a family
of vector lines of force. These lines are formally defined as curves
Maxwell Theory itt a Refalfvislic Form 201
whose tangents at every point coincide with the direction of the

field vector at that point. A line of force is a useful assisting
notion allowing the properties of the field to be graphically exhib·
ited. In contrast to the ideas of the last century, however, no one
attaches any physical meaning to these lines now.
Suppose, a charge or a constant magnet moves in space. Should
one say in this case that a field and its lines of force move
along?
A field is a method to describe what happens at a given point
in space. The magnet motion merely causes the field to vary in
¥
"v
Fig. 6.3. The mteraclion of a charge q, moving at the velocity V parallel to 3

C!.!rrent-carrving conductor, and a current. (a) A conductor is at rest in the
frame K, thC charge and electrons move at the velocity V. (b) A conductor mo-
ves at the velocity V in the frame K', the charge and electrons are at rest.
time at a given point. And still, one may speak of the field motion
induced by a charge or magnet moving at a constant velocity,
since this field moves with them as a whole. The field transporta·
tion velocity is the velocity at which a charge or a magnet moves.
A motion of lines of force, however, is better not to be mentioned
since a motion velocity of lines of force has no physical meaning.
An auxiliary nature of lines of force is demonstrated particularly
well by the fact that they may just disappear in some reference
frame for a certain field.
Here is another example illustrating the relative character of
forces acting in an electromagnetic field. Consider a cylindrical
conductor carrying a current and a negative charge q moving
parallel to the conductor at the velocity V (Fig. 6.3). We shall
fix the frame K to the conductor, and the frame K' to the charge.
In the frame K the charge experiences the Lorentz force induced
by a magnetic field and directed at right angles to the conductor's
axis. Consequently, the charge approaches the conductor. In the
frame K', however, the charged particle is at rest and the magnetic
field has no effect on it. Then, what is the reason causing the
charge to deviate in terms of the frame K'?
Here one needs to re\'iew a microscopic description of what is
happening in a conductor. A current originates in a conductor
due to the motion of free electrons since positive ions and fixed
(valence) electrons cannot migrate along a conductor. Let the
density of conduction electrons be equal to p-. with their velocity

in K (relative to the conductor) equal to V-. The density of sta-
tionary conduction electrons is equal to P+• and due to the neutral-
ity of the conductor P+ + P- = 0. Inasmuch as the conductor is
neutral, there is no electric field outside of it, and the force acting
on the charge q arises only from a magnetic field:
F~q(VB(,
The magnitude of the magnetic field induced by a rectilinear cur·

rent at the distance r from its axis is known:
8= ~~~.
the vector B coincides with the tangent of a circle lying in a
plane perpendicular to the current's axis and having its centre on
it. The direction of the vector B is determined by the right-hand
screw rule. Hence, the force acting on the charge is directed
toward the conductor and is equal to
F= qVJlol =-~-~
2:rcr 4:rce 0 c~ r •
The current can be expressed via the conduction electron ve·

locily u-, their density and the cross-sectional areaS:
I~iS~p_v_S,
whence
F= 2:rcqeo p~S ~: = 2:eo p;S ; , (6.52)
provided that for the sake of simplicity we assume the velocity

of electrons in metal to be equal to that of the charge q, i.e.
V = V-.
Now let us consider the same situation in the frame K'. The
charge q and conduction electrons are at rest in K'. This time,
however, the charges connected with the conductor (and whose
density is equal to P+) move relative to the charge q. Although
they induce a certain magnetic field 8', it does not act on the
charge q any more, since the charge is motionless in K'. Whence
it is clear at once that an electric field must appear in the frame
K' since the charge must deviate toward the axis in K' as well.
Its origin is easy to understand from the results obtained by us
earlier. Conduction electrons are at rest in the frame K'. and
therefore p_ = fp~ (see Eq. (6.17) ). Positive charges connected
with the conductor move at the velocity - V in the frame !(', and
so p~=I'p+ (these charges were stationary in K). The resulting
Maxwell Theory in a Relativistic Form
''"
charge density p' is equal to p~ + p~ in the frame K', and, con-
sequently,
p'=p_JI'+ i'p+ =p+ (r- 1/f)= l'p+B 2,
where the relation P+ = - P- is allowed for; this equation coin-
cides with Eq. (6.24). Consequently, a moving conductor is
charged positively with the space density p'. But an electric field
K'
(a)
~ (6)
Fig. 6.4. (a) In the frame K the charge density p (p = P+ +

zero while the current density is equal to j '* p-) is equal to
0. Therefore, an electric field is
absent in the frame K and there is only a magnetic field B. (b) In the frame K'
~~~::n:~~~~o~o c~,~rie~~~~e~i~e~:~~ i~s eeq~~~l ~ B:: ~~:id';s~ ~~rr:l~t~i~n~i:rd be;
also appears.
of a uniformly charged cylinder is also known from electrody-

namics. It falls within the planes perpendicular to the cylinder's
axis and is oriented along the rays leaving the cylinder's axis. Its
magnitude
This implies that the force acting on a negative charge q is

directed toward the conductor, and its magnitude in the frame K'
is equal to
F'=qE'= 2:~o p;S 1'82.
Comparing this result with Eq. (6.52), we see that these forces
are equal in a non-relativistic approximation (f !=::::I). Recalling
that the forces transform according to Eq. (5.34), we find that
both ways of describing an observed phenomenon give identical
results at any velocity V. The results pertaining to fields in the
frames K and K' are explained in Fig. 6.4.
In conclusion it should be emphasized that all the results per-
taining to forces which an electromagnetic field exerts on spa~e
charges, are obtained quite easily provided that the Lorentz force
204 Special Theory of Relatlully
density (see Eq. (6.49))

1-p{B+[vBI}
is written in a four-dimensional form. To pass over to the four-
dimensional notation, rewrite the x-component of the Lorentz
force as follows:
f:c=ft =PE:s + pv 11 Bz- PVzB11 =
={ -7 )iF1 4 +s2~-s3 {- F; 3 )=
= 7(F12s2 + F135a + F14S4) = ~Fu.sk•
The following relations are taken into account in this chain of
equations:
i
p=-cs4, pv 11 =s2, pv.,=s3,
Bz.=!.p-. 811 =-F: 3 , F 11 =0.
Analogous expressions are obtained for {11 = f2 and fz = {3•

Hence it is clear that the 4-vector of a force density acting on a
charge in an electromagnetic field, to be denoted by f. has the
components*
(6.53)
We have already pointed out that the separation of forces exert-

ed on the charge by electric and magnetic fields into the parts pE
and p[vB] is relative. Both these forces constitute a united whole
combining naturally into a single four-dimensional expression
(Eq. (6.53) ).
As we have seen, the first three density components lead to th~
conven:ional three·dimensional equation (6.49). Let us find the
fourth component:
f4=+F41!s,.= +(F41S1 + F42s2 + F43s3)=-f(vE).

ThC" quantity p(vE) has a simple meaning which is immediately
seen as soon as the two sides of Eq. (6.49) are multiplied scalar-
WISt: by v. lahmg into consideration that [vB]v = 0, we get
(/v)- p (vB) .
.. We expect that the reader will not forget that the letter F supplemented
wtlh two submdkes represents a tensor f component. The components of the
4·force density have a single subindex,
Maxwell Theory ln a Relatlulsflc Form 205
fhe left-hand side of the last equation represents the power

of a Lorentz force per unit of volume (forces exerted by a mag·
netic field perform no work):
1·=+((1>).
Thus we have obtained a 4-force density vector whose compo·
nents are written down together as follows:
r{l'1. I,I, IaI, 0 1,(/•l }·

1 (6,54)
Let us consider the force exerted by an electromagnetic field

on a unit of volume containing a charge p0 in the reference frame
K0 co-moving with the charge. Then {(poE', 0). Passing over to
any other reference frame K, we get
11 = t;. f2 = rt;. /3 = rr~. / = - iDr{~ = - i ~ rp e:.

4 0
Here the equations for a force density in the frame K are ex-
pressed through the fields in the frame K'. Usually a force density
is expressed in terms of the quantities referred to the frame in
which the force density is determined. Making use of Eqs. (6.17)
and (6.36), we obtain f(p(E +[VB]). i/c p(Ev)) for non-relati-
vistic velocities (ignoring the terms V2/c 2 ).
In conclusion we shall write out the motion equation for a
charged particle in a four-dimensional form:
(6.55)
§ 6.7. Covariance of the system of the Maxwell equations. The

Maxwell equations define the behaviour of nn electromagnetic
field in the most adequate manner. They were proposed long be-
fore the advent of the theory of relativity and surely before the
Lorentz transformation was identified. According to the principle
of relativity the appearance of the Maxwell equations must remain
constant in all inertial frames of reference. Consequently, the
Maxwell equations must be covariant relative to the Lorentz trans·
formation. It so happened that the system of Maxwell's equations
satisfies these conditions when written in the form proposed by
its creator. To ascertain this, the system of Maxwell's equations,
written usually in terms cf three-dimensional equations, should
be rewritten in a four-dim<!nsional form. Now we shall be occu-
pied with just that. The system of the Maxwell equations is known
206 Special Theory of RelatlrJlty
to have the following form:

rotH~i+D. (a) I divD~p; (b) (6.56)
rotE~- B, (a) div B~o. (b) (6.57)
We have split the equations into two lines, having combined the
equations involving the average values of an electric and a mag-
netic field E and B, and the equations for the subsidiary vectors
Hand D.
In order to present Eqs. (6.56) and (6.57) in a four-dimensional
form, we shall need the tensors (6.29a) and (6.31); we shall abo
make use of the definition of a 4-current density vector (6.12a).
Note for the future use, by the way, that tensors 5' and fare linkd
in vacuo by the relation
(6.58)
Naturally, Eqs. (6.56) can be expressed via tensor (6.31) while

Eqs. (6.57) via tensor (6.29a).
Let us consider the x component of Eq. (6.56a):
b;t;- a~z +a~~~ =-i;r;- (6.59)
Recalling that according to Eq. (6.12a) ix

= s 1 and using th~
first line of Eq. (6.31) together with the definitions x 1 = x, x 2 = y,
x3 = z, X1 = ict, we shall rewrite Eq. (6.59) as - ~~; - ~~: -
- ~~; = - s 1• The two other components are given by similar ex-
pressions which can be written in a general form as (i = I. 2, 3):
a{tk
ax;= St, (6.60)
where the summation is performed over k running from I to 1.
It is readily seen that when i = 4. we get Eq. (6.56b). Thus Eq.
(6.56) is rewritten in the form of Eq. (6.60), but now in terms of
4-tensor components (6.31).
Now let us consider the components of Eq. (6.57). For example,
the x component of Eq. (6.57a) can be written as
(6.61)
Resorting to tensor (6.29a), one can rewrite Eq. (6.61) as

follows:
i (~:l + ~~:' + ~~;~)=o. (6.62)
Maxwell Theory In a Relativisfic Form 207
It 1s not d1rlicult to notice that consecutive terms in Eq. (6.62)

are obtained via a cyclic transposition of the three indices in each
of the preceding terms. The structure of Eq. (6.62), however, be-
comes quite evident if we introduce the tensor Fik which is dual
to the tensor F,k (see Appendix I,§ 6):
(6.63)
where e..ttrn IS a fully antisymmetric unit 4-tensor of the fourth

rank. One can easily see that the dual tensor Fik differs from the
tensor F.k only by transposed components of the imaginary and
real parts:
\l~(cB, -iE), \i·~(-iE, cB),
(6.64)
or, written in full,
o -i£, i£, cB,l
• ( iEz 0 -lEx c8u
(6.65)
Flk= -iEu iEx 0 cBz .
-cBx -cBu -CBz 0 J
Using this tensor, the pair of Maxwell's equations (6.57) can be
rewritten in a four-dimensional form as follows:
aF;k =0 (6.66)
ax, ·
Let us make sure that Eq. (6.66) corresponds to the four equations
ol (6.57).
Eq. (6.66) contains four equations (i = I, 2, 3, 4). Consider,
for example, the equation fori= I:
aF;~ = ~ + aF;z + aF;3 + aF;4 =
ax, dx 1 ax2 ax3 ax 4
=- i aa:z +i a:: + ~~::~) =0,
or ethel wise
aa:;: - a:: = - a!x , i. e. (rot E)x = - (B)x.
Eq. (6.66) yields the two other components of the equation
rotE=- B at i = 2, 3. For i = 4 we get the following equa-
tion:
dF;.t = aF;1+ aF;2 + aF;3 _ aF;4 =
ax, ax 1 ax2 dx3 ax4
= - a~:x) _ a~:u) _ a~:z) =O;
Special Theory of Relafivify
representing tq. (6.57b): div B = 0. So we see that Eq. (6.66)

comprises Maxwell's equations (6.57).
Quite often Eq. (6.57) is expressed directly through the tensor
f,k. We present this notation here since it will be nl;!cessary lat~r
on. The Maxwell equations (6.57) can be legitimately written both
in the form of Eq. (6.66) and as an equation of the type obtained
in (6.62)'
(6.67)
There is no summation involved in Eq. (6.67). Three different

values of the indices i, k, l are to be chosen from the four pos-
sible ones. The reader can make sure himself that if two of these
indices are taken to be the same, and antisymmetric propertie'i
of the tensor S'(F,k = - Fk,) are allowed for, Eq. (6.67) is seen
to turn into an identity. The structure of Eq. (6.67) shows that
the distribution of the chosen triad of numbers among the indic~s
i, k. l is insignificant. This implies that Eq. (6.67) contains sev·
era! independent equations, their number being equal to the
number of possible combinations, each containing three indices,
that can be formed from a collection of four indices. that i'i
~C3 = -C 1 = 4. We let the reader make sure for him,.elf that the
four equations (6.57) follow from Eq. (6.67).
Now the Maxwell equations can be readily proved to be co·
variant. We have seen that they can be written as Eq. (6.60) and
(6.66) or (6.60) and (6.67). But Eqs. (6.60) and (6.66) represent
the relations between 4-vectors since the expression a{,klaxk is a
vector (see Appendix I, § 5). Eq. (6.60) differs from Eq. (6.66)
by the zero vector featured on the right-hand side of Eq. (6.66).
As to Eq. (6.67), it is explicitly presented in a tensor form and
consequently is covariant. Thus, the four-dimensional notation of
Max\vell's equations itself indicates their covariance.
The system of Maxwell's equations is not confined to Eqs. (6.56)
and (6.57). We have already quoted the charge conservation law
in a covariant form (see p. 184). It remains only to rewrite the
umaterial equations" in a covariant form.
§ 6.8. The Minkowski equations for moving media (the tram-
formation of material equations). In the previous section we saw
that the system of Maxwell's equations (6.56) and (6.57) retains
its appearance in all inertial frames of reference. However, the
Maxwell equations yield an unambiguous picture of electromag-
~:~~u~~~o:~?cah ~~~~tr~~:ngn~fi~er~~~n~~=~i~n;~c~~.a~~~t~r~!~7fed~
As usual, the reference frame in which the medium (or its portion)
rests will be referred to as a co-moving one. In the case of a uni-
form isotropic medium the material equations have the following
Maxwell Theory in a Relativistic Form 209
form in a co-moving frame:

D'=eE', (6.68)
B'-"H', (6.69)
J'-aE', (6.70)
with a permittivity e, permeability ll and conductivity (J all being
constants. Consider the motion of a medium relative to a "labor-
atory" frame. In the frame co-moving with a medium the Maxwell
equations for a stationary medium are valid. Due to the principle
of relativity the material constants e, IJ., a must be the same both
in a stationary medium in a "laboratory" frame and in the ref-
erence frame co-moving with the medium. Inasmuch as the trans-
formation equations for the vectors E, B, H and D are known,
the relationship between them can be found in any other inertial
irame differing from K'. Let us write out the necessary transfor-
mation equations, having split them into the longitudinal and
transverse (relative to the reference frame velocity V) parts (see
§ 6.4):
E;-IE+(VB])1, E~-I'(E+[VB]),; (6.71)
s;- (s- {.-IVEI),. B~ -r (s-?-[VEJ), (6.71')
o;~·(D+{.-IVHI),. D~ -r ( D + {,-JVHI),; (6.72)
n;-(H-[VD[)1, H~-r(H-(VD]),. (6.73)

We should recall once more that all expressions of the type
[VA]u are equal to zero lor any A since we deal with projections
on the velocity direction and a \'ector cross product is perpen-
dicular to the velocity V. If the corresponding expressions are sub-
stituted into Eqs. (6.68) and (6.69), we obtain the identical re-
lationships for both the longitudinal and the transverse comp"J-
r
nents (in which the factor cancels out). These relationships can
be combined as follows:
D+?-JVHJ-e(E+(VB]), (6.74)
B-?-[VEJ-"(H-[VD]). (6.75)
Eqs. (6.74) and (6.75) are called the Minkowski equations: e

and ll appearing in these equations represent a permittivity and
permeability of a resting medium. These equations differ essen-
tially from Eqs. (6.68) and (6.69) in that they involve all of the
field vectors simultaneously. Using Eq. (6.75) one can easily elim-
inate B from Eq. (6.74) and obtain an equation involving only
210 Special Theory of Relatiuify
the three vectors E. D, H, or eliminateD from Eq. (6.75), using

Eq. {6.74).
The equations appear simpler when put down separately in
terms of longitudinal and transverse components:
D 1 =eE1, Ba=JJ.Ha. (6.76)
(1- e:~o *)DJ.=e(l -~)BJ.+(eJJ.-to~)[VHJ,
(1- e:~o 7v•) B.L=IJ. ( 1-cr
v•) HJ.+(e!l-eoJLo)[VEJ. (6.77)
The first equation of (6.77) is obtained from Eq. (6.74) into

which the expression forB is substituted from Eq. (6.75) and then
only a transverse component of the relation obtained is taken. In
much the sam~ manner one obtains the second equation of (6.77).
It is seen from these equations that if the vectors 8 and H, as
well as D and E, coincide in direction in an isotropic medium in
a co-moving frame K', this is not the case in other reference
frames.
Of course, when one examines the motion of a medium, the case
of non-relativistic velocities proves to be most interesting. Hence,
if one ignores the terms V2/c 2 and n2 V2/c 2 compared to unity
(n = 1/ tqJ./eoJJ.o is a medium refraction index, see Chapter 7) in
Eq. (6.77), then Eqs. (6.76) and (6.77) take a simpler form:
D~eE+T,(n'-l)[VH 1 , B~~H+T,(n'-l)[VE]. (6.78)
This form of material equations written for a moving medium is

employed \'ery often. Due to the equivalence of all IFRs in vacuo
the last equations turn into Eqs. (6.68) and (6.69) in this ca<>.::.
It is helpful to rewrite the material equations (6.68)-(6.70) in
a four-dimensional tensor form. We shall not derive these equa-
tions; we shall just write them out to check that we get Eqs.
t6.68) · (6.70) in the frame in which th~ medium is at rest. Let as
introduce a four-dimensional velocity V (l'V, icf) for a medium.
(r is written here since the velocity of an object or a medium V
is assumed to be equal to that of}he reference frame K'). In a
co-moving frame K' the 4-velocity V' has the components U] = 0,
V2= 0, VJ= 0, U~ =ic.
The reader can easily verify thai the tensor equations
+tt,,Pk=eFtkUk, (6.79)
7tFtkU, + Fk1U1 + F11Uk) =!J.(flkUt + fk,Vt + f11U:,), (6.80)
&1 = f FtkUk, (6.81)
Maxwell Theory In a Relatialstrc Form 211
lead to Eqs. (6.68). (6.69) and (6.70) respectively if compo- y,

nents are substituted into them.
In Eqs. (6.79) and (6.80) the summation is carried out over k.
The total number of equations amounts to 4, but since at i = 4 we
get an identity, there are only three equations, in fact. In Eq.
(6.80) one has to find the number of possible combinations of i, k
and l that can be formed from four values I, 2, 3, 4 taken three
at a time. This number is equal to 4 C3 = 4, but since the combi-
nation I, 2, 3 yields an identity, we come back again to three
equations as it should be. Having derived the correct expressions
for material equations in the frame K'. we prove that a tensor
notation is also correct.
We shall illustrate an application of a tensor notation by the
example of Eq. (6.81). Suppose that a current density is observed
in a co-moving frame K' (in which the medium is at rest) and the
charge density is equal to zero, i.e. s;
(i~._J;, j~. 0). The velocity
of a medium in the frame K' is equal to V' (0, 0, 0, ic). Here are
the s;
components:
s~ = f F~ku~ = 7 F; 4 V~ = 7(- iE;) (ic) = aE;.
i.e. j; = aE;; similarly ( = aE;. 1; =a£;. The fourth component
s~=7F~V~==0 (s:=icp'=O, ~ 4 = 0).

But in the reference frame relative to which the medium moves
St = fFtkUk=f<Ft2V2+ FuV3+ FtP4)=7 (- iE_.;)icf = ar£.1:,

s2 =fF:kVk=f(F21V 1+F24U4)=f(-c831'V +(-iEy)icl')=
= al' {E, + (V BJ,),
s, =or {E, +[VB],),
since
v~=rv. V:=V3=0, V4=icr.
The final result is obvious:
/=ar(E+[VBJ). (6.82)
Its meaning is quite clear: the current density in the medium
with a conductivity a is determined by the magnitude of an elec-
tric field in this medium; in accordance with Eq. (6.41) the magni-
tude of an electric field makes its appearance as a factor by a in
the case of r ;::::: I.
212 Special Theory 0/ Reloliuily
Th~ fourth equation defines the charge density associated \\'lth

the conductivity current:
St=iCPcot~d = %-FuV•=f(ifEV)=if ( -?). (6.83)
Pco11.t=r*
in complete agreement with Eq. (6.24).
It is worthwhile to consider Ohm's law in the case of moving
media, i.e. the material equation (6.70). We shall see that the
convection current pv and the conductivity current are closely
interlocked as it becomes obvious right after j and icp are com-
bined into a single 4-vector. The difference between a convection
current and a conduction current is caused by the choice of a ref-
erence frame. Therefore it is natural that both currents alike in-
duce a magnetic field.
We shall assume that a conductivity current represents a mo·
tion of charges with respect to a medium whereas a convection
current arises due to the presence of charges in a medium owing
to the motion of this medium.
re~tu~p:eok~a~~~. ab~:[J!!~ !r~~aerg~, ~~~~~t~s p~. t.f~:s~c~~~~ti~r;~
constitute jointly a 4-current which can be transformed to any
reference frame by means of Eq. (6.15a). Having expressed the j
components and the density p through j' and p' in the reference
frame K', we obtain
i,~r(i;+Vp'). i,~i;. i,~i;. p~r(p'+-;':-i;). (6.84)
It is seen from the first formula of (6.84) that a conductivity

current j.., incorporates a convection current rVp' = Vp so that 1t
is not proportional to cr any more. It is inconvenient because at
o = 0 a conductivity current must turn into zero. How to distin-
guish a conductivity current in the general case? To do this, one
must recall that if in the frame K' there is a charge density
p' = p0 we obtain a 4-current density in any other frame K (see
Eq. (6.20))
(6.85)
where u, is a 4-velocity of the charge. This current should be
called a convection current; accordingly, s, in Eq. (6.85) is sup-
plemented with a superscript "conv". Suppose, we have a 4-cur·
rent whose components are s., and we want to represent it as the
sum of a conductivity current and a convection current. First of
all, let us express p0 through ; and V.
Having multiplied both
Muxu.odl Tl!eon1 in a Relativislic Fnrm 213
sides of the equation s, = pou, by the corresponding components

u, and added up, we get sku~= p0u1; but according to Eq. (5.7}
u~ = - r?, so that
(6.86)
Consequently, the convection current can be put down as

(6.87)
In order to obtain components of a 4-conductivity current, oae

has to subtract components of Eq. (6.87) from s,:
s~ond=s,-s1o"v=s,+ s:~~ u,. (6.88)
On the other hand, in accordance with Eq. (6.81) the qu<Jntity

s~ 011 d can be put down as
(6.89)
Having equated these expressions, we obtain:

s, + s~~k = "7 F,ku,..
111 (6.90)
Utilizing the definitions ;(i ... j iz. icp), V(yv .. , yu

11 , 11 , yuz, icy), we
get
I+ v'v(%- p) ~cry (E + [vBJ) (6.91)
in a three-dimensional form.
Let us separate the terms proportional to a conductivity a in
Eq. (6.91). For this purpose the left-hand and right-hand sides of
Eq. (6.91) arc multiplied by v. Introducing the usual designa-
tinns y and ~. we get
or
~-p=-7-+ 0 ~~v) +· (6.92)
Substituting Eq. (6.92) into Eq. (6.91), we finally obtain
f~pv+crv{E+[VBJ-f,-(Ev)}. (6.93)
Thus, the term "conductivity current" can be attributed to th!!
quantity jcond = j - pv. A field existing in a substance moving
relative to a given reference frame is often denoted by E•
E'~E+[VB).
Then Eq. (6.93) can be rewritten in the following form:
;''"'~•v{E'--f;- (E'v)}. (6.95)
This equation resembles very much the force transformation

equation (5.35). To complete the transition to equations of mov·
ing media electrodynamics, one haS' to find out how to describe
boundary conditions when a media interface moves. The continuity
condition for normal components of induction follows from the
equations div D = 0 and div 8 = 0, which, according to Eqs.
(0.66) and (6.60), keep their appearance on transition from one
inertial frame of reference to another. Therefore, at the interfac~
(6.96)
Let us examine now the boundary conditions for tangent com·
poncnts of field strengths. Considering first the reference frame K'
co-moving with the interface, we obtain the continuity condition
for tangent components of E' and H' in this frame. But in terms
of the frame K relative to which the interface moves at the veloc-
ity u, the field:. E and H take the following form (see Eqs. (6.41)
and (6.43))o
+
E' ~ E [uB[, H' ~ H- [uD[. (6.97)
Let us draw a perpendicular n to the interface plane and denote
the projection of the velocity u on this perpendicular by urr. Let us
find the projections of Eq (6.97) on the plane perpendicular to n.
Remembering that [nE} = [n, En+ Et) =[nEt), we shall wrile
the equation E;1 = El2 as [nE;J] = (nE/ 2), i. e.
[nEd+ [n [uB,[[ ~ [nE,] + [n [uB,]],
or
[n, E,- E,[ ~ u (n (B,- B,)) + (B,- 8 1) (un).
Since according to Eq. (6.96) nB 1 = nB 2, we finally obtain
[n, E,- E,[ ~ "• (B,- 8 1), (6.98)
<1nd similarly,
(6.99)
This equation, together with Eq. (6.96), constitutes the boundary
condJtions for field vectors.
§ 6.9. The transformation of electric and magnetic moments. If
\\t;> combine an electric and a magnetic moment P and M into a
~ingle antisymmctric tensor (6.33) we can immediately write trans-
formatiOn equations for components of these quantities.
Maxwell Theory in a Relativistic Form ,.
Let us denote a polarization and magnetization, determined in
the reference frame co-moving with substance, by po and MO re-
spectively. Then an observer relative to whom the substance moves
at the velocity V will get
M,~M~. M,~r(M:+VP~). M,~l'(M~-VP:).
Px=P~, P11 =f(P~-~M~). Pz=f(P~+fM2).
(6.100)
These equations clear up at once the relationship between thP.
three-dimensional vectors P and M introduced earlier. Here one
can repeat everything that was said about the relationship between
electric and magnetic fields. As a rule, magnetization is always
accompanied with polarization, and vice versa. P or M can hE'
equal to zero only in a specially chosen coordinate system. A pol-
arized but not magnetized object is both polarized and magnetized
in terms of an observer relative to whom this object moves. In-
deed, suppose that in the frame K' in which the object is at rest
M0 = 0, P 0 (P~, P~. P~) =I= 0.
Then in the frame K relative to which the object moves at the
velocity V,
Px = P~. P11 = fP~, Pz = l'P~.
M..:=O, Mg=l'VP~. Mz=-I'VP~-
Consequently in the frame K magnetization of the object will be·

observed. If the object moves at a non-relativistic velocity, i.e.
V/c « I and f::::.: I,
This effect was found in experiments performed by Eichenwald

.(see (13], [29)). On the contrary, if in the frame K', relative to
which the object is at rest,
J>"~o. M'(M:. M:. M~).,co,
then in the frame K relative to which the object moves at the ve-
locity V
Mx=M~, M 11 =fM~. Mz=fM~.
(6.101)
P..:=O, P 11 =-1'fM~. Pz=f?-M~.
216 _ _ _ _ __:S:;.:P'c:"=:"l_:_T_:::><cc"'.=_Y_:_'lc_Rc:'.::''_:_livc:ile_Y_ _ _ _ _ __
Consequently, the object turns out to be polarized in the frame K.

If the object moves at a non-relativistic velocity, then
M~M'. P~-[M'-;';-]
This implies, for example, that a moving permanent magnet car-
ries an electric moment giving rise to the phenomenon of homo-
polar induction utilized in electrical engineering.
Here is an example illustrating these conclusions. Let the density
of a current flowing along the rectangular loop ABCD be equal
F
to j, and the loop itself move at
~~:.
the velocity V relative to the
frame K. Let us fix the frame K'
to the loop (Fig. 6.5). In accor-
K
;;---;-- dance with Eq. (6.24) a charge
p > 0 appears in the section BC
and a corresponding charge
I
p < 0 in the section AD. It is
obvious that the total charge
appearing in the loop ABCD i:c::
equal to zero. At the same time
z this loop possesses an electric
Fig. 6.5. Emergence of a dipole mo- moment directed along the y
ment of a current loop when cons•d- axis. We shall show that an ele-
ered m the reference frame I( relative mentarv calculation coincides
to wh1ch this loop moves. with cOnclusions of the STR.
Although there is no dipole mo-
ment m K' and only the zth component of M is present, Py =
=- r?- M~ emerges inK according to Eq. (6.101). In the frame
K' the rectangular current ABCD possesses the magnetic moment
IS where the vectorS is directed toward the negative z axis and is
Equal to ab (a and b being the sides of the rectangular loop). As-
suming for the sake of simplicity the cross-section of the conductor
to be equal to unity, we obtain M~=- i0 ab. The electric dipole mo-
ment emerging in the loop is not difficult to calculate. According
to Eq. (6.24) p = r.;.
i~. the distance between BC and AD is equal
to b. and the total charge in these sections is equal to pa. The di-
rection of this dipole moment coincides with that of the y axis.
Therefore, PH=pab=f~abi~=-?-rM~. just as it should be.
§ 6.10. Some problems involving the transformation of an elec-
tromagnetic field. The field of a uniformly moving charge. Tile
magnetic and electric fields of a uniformly moving charge are mo~t
-easily obtained by transformation of the fields existing in tht.!
fr2me K' in which the charge is at rf"!st. In the case of a point
Maxwell Theory in a Relati~;istic Form 217
electric charge e resting in the frame K' we face an electrostatic

problem since such a charge produces only an electric field.
However, when the same charge is considered in terms of the
frame K moving at the velocity - V relative to K', it is found to
generate a rectilinear current. A magnetic field induced by a recti-
linear current is very well known: lines of force of such field form
circles whose centres coincide with the current; the planes of these
circles are perpendicular to the current direction. Naturally, these
results follow from the field transformation equations.
Now, let a point charge be located at the origin of the frame K'.
Then in this frame
B' = 0, E' = 4: 8 -f-3~,
or, when expressed in projections on the coordinate axes,
where r'2
B~=o,
E~= 4:e *·
= x' + y' +
the frame K
2 2
s;=o,
E;=
z'2.
4:e *· B~=o.
E;= 4~e ~·
According to Eq. (6.36) we obtain in
£,~£',. E,~r£,. E,~r£'.. (6.102)

Bx=B~=O, 8 11 =-l'ci-E~, Bz=r{r-£;. (6.103)
As Bx = 0, the magnetic field in the frame K is located in the

planes perpendicular to the x axis, i.e. in the planes perpendicular
to the current direction. The equations describing the lines of force
of the magnetic field take the following form in the frame K:
dy dz dy B11
s;;=n;· or Tz=n;·
But
B 11 z' E; z
n:=-e:;= -y;= -y·
since z' = z and y' = y under the Lorentz transformation. Conse-
quently, the differential equation for the lines of force tal<es the
form dy/dz=-z/y, or ydy+zdz=O, i.e. d(y 2 +z2 )=0. Henc~.
it is obvious that we have the equation of a circle yz z2 = const +
in the capacity of a first integral. Consequently, the lines of force
represent circles with centres located on the current axis.
Surely, one can transform not only fields, but potentials as well.
In the frame K' the scalar potential is equal to
IP' = 4~e 7 =-Til>~.

whereas the vector potential is equal to zero: A'= 0. If the 4-po-

tential <iJ has the components (A', 7cp')
in the lrame K', its com·
ponents in the frame K take the following form in accordance
with Eq. (6.14a):
I!T1 = r (Ill;- i~ CD~). tl>2=~· <l>a= ttl~.
<l>4= r ( tt>~ + i~lll;).

Substituting the values of the 4-potential components in the
frame K', we get
A~=r(-if7cp')=r;cp', A2=0, A3=0, frp=r-fq>'.

Thus,
A=; rq{=;cp. cp=fq>'. (6.104)
Now \ve have to express r' entering into q/ via the charge coordi-
nates in the frame K. According to the Lorentz transformation
x'~r(x- Vt), y'~y. z'~z. (6.105)
and the expression for r' 2 \Viii be \Vritten as
r'2=x'2+ y'2+z''l= r2(x- Vt)2+y2+z2=
=1'2 [(x- Vt )+ Y~~z 2 ]=f2jR2,
2 r'=l'lli, (6.106)
\Vhere the following designation is introduced:
!J!'~(x-V/)'+(1-f.'-)<u'+i'). (6.107)
Making use of Eq. (6.107), one can express the scalar potential q>,
defined in Eq. (6.104), through !.R:
cp=fq>'=r 4~e 7= 4~e ft-.

Accordingly, the vector potential A can be written in the form
A=* q>= 4~e :2~ •
Let us rewrite the expressions for the E components of the

field, taking into account Eqs. (6.102), (6.105) and (6.107). We
Maxweff Theory in a Relativistic Form
'"
get:
Ex= ei~~:e~:)
Ev= P::e!JP' (6.108)
E:: = r 2 4~~9P ·
In the frame K' the charge is located at the origin 0' (i.e. at the
point x' = 0). Its coordinates at the moment tin the frame K will
be as follows: x0 = Vt, y0 = 0, z0 = 0. Let us introduce one more
vector, R. directed from the point 0', where the charge is located
.r-Vt•R·C0$8
Fig. 6.6. To the calculation of an electric and a magnetic 6eld of a uniformly
moving charge.
toward the observation point A whose coordinates are (x, y, z)

(Fig. 6.6). The vector R will take the form
R~(x-Vf)i+yf+zk, (6.109)
where ;, j, k are unit vectors along the x, y, z axes respectively.
Having multiplied the components of Eq. (6.108) by i, j, and k
respectively, we obtain
E=E"i+Eui+E::k= 4!e ~ifr.
If one introduces the angle 6 between the charge motion direction
(i.e. the x axis) and the radius vector R. then
x-Vt=Rcosa, R2 =R2 cos2 B+y2 +z2,
and consequently,
(6.110)
Taking into account Eqs. (6.109) and (6.110), one can rewrite
Eq. (6.107) as
\\hereupon the expression for E can he finally represented in the

form
V'
I eR. !-CT
E (6.111)
4Jte Rl ( 1- ;sin20 Y'2'
Eq. (6.111) presents the electric field of a moving charge in very
convenient variables, i.e. the distance R from the moving charge
1nd the angle a formed by the direction to the point at which the
field is sought and the charge motion dir('ction. Eq. (6.111) shows
that the magnitude of the field depends on the angle a. AI a
fixed R the minimum magnitude of the field corresponds to the
charge motion dtrection (a= 0, n):
£1= 4~e -ir(l- ~).

and the maximum magnitude of the field is observed in the direc-
!ton perpendicular to the motion (a= n/2):
E = _!__-=._ - - ' - -
J.. 4:1e R~ "'1/t- V1Jc2 •
A field strength magnitude depends on a charge motion velocity,

with Eu decreasing and £J. growing as the velocity increases. The
electric field of a charge moving at a relativistic velocity is local-
ized within two narrow solid angles whose boundary surface is
approximately determined from the relation ( V2 sin 2 9/c2 ) ~ I; the
axial line of these solid angles is perp('ndicular to the charge mo·
!ion direction.
The magnetic field 8 of a charge moving in the frame K can be
found by means of Eq. (6.44) (8' = 0 in the frame K'):
B~ f,- (VEJ. (6.112)
When the ,·elocity of the charge is low, the fields in vacuo ar~
described by the following approximate relations:
E= 4~eo 7{:-
and
8= 4Jt~oc2 ef:3R.l =f;i- fe:~l. (6.113)
Eq. (6.113) represents the Biot-Savart law.

The interaction of two moving charges. Let two charges e 1 and
e 2 mo\"e in parallel at the same velocity V. Let us determine the
jnteraction force between them in the reference frame I( relative
Maxwell Theory in a Retativtstlc Form
'"
to which they move. First, we shall find the force acting on the
charge e 1.
The charge e 1 experiences the action of an electric and a mag-
netic field induced by the charge e2. The force acting on thC'
ch11rge e 1 is the Lorentz force:
F, =e, {E,+[VB,)).
Taking into account Eq. (6.112), one can write down
F 1 =e 1E2 +?- [V[VE2ll =e E +% V (VE 1 2 2) - %- V E
2 2=
=e 1 (t-~)Ez+%-V(VE 2 ). (6.114)
E 2 can be found from Eq. (6.111) in which R is assumed to be

the radius vector drawn from the charge e2 to e 1 and e the angle
between R and the charge motion velocity direction V. Substituting
Eq. (6.111) into Eq. (6.114), we obtain
(I - ~:- yR e1e2VV cos 8 (I - ~)

pL = 4 ~~~3 ( l---;;r-sin
V' 28 )3.-2 + 4nsc2R2 (1-"""C2sin
V' 2 8 )'"
v•
''"
= 4neR1 (
lv:~.
-7 )arl
[( 1- C2
V') R VV
7[+7cos8'
l
1- 7 sm 1 8
whence the component along the motion direction is

e 1e 2 (1-{:--)cosB
. 2 8 )312'
Fx= 4nsR• ( I - V' stn
7
and the component perpendicular to the motion direction
( 1-~) 2 sin8
P - ..!..!.!!.._ -i------;rr'-'---____,,., (6.115)
~~- 4nsR~ (t-~sin:~ey'2.
Let the charges be located on a straight line parallel to the y

axis, with one of the charges being at the x axis, so that the dis-
tance between the charges is equal to y. Then e = n/2, F, = 0,
and
(6.116)
This equation can be obtained very simply. In the frame K'

where the two charges a.::e at rest, the interaction between them is
Special Theory of RelatiiJity
"'
of electrostahc nature and the interaction force IS equal to
e 1e 2/4ney 2• The transformation of this force on transition from the
frame K' to the frame K by means of Eq. (5.35) yields Eq. (6.116).
According to Eq. (6.116) the charges repel each other in the frame
K. But in the frame K the charges move producing two parallel
currC'nts flowing in the same direction. When such currents flow
along conductors, they attract each 9ther. There is no contradic-
tion here since the physical situations are different. Let us consider
the expression for a force (Eq. (6.116)) in vacuo in the case of
non-relativistic velocities V:
F 11 ~ 4 ~~:y2 (t- ~: + ... )= 4 :~:;~ - :~;1 11-of- (6.117)
On the other hand, following Ampere, the interaction force be-

tween two current elements e 1V and e 2V in vacuo can be written
as
p 12 =7} le1VI~2JV.RIJ = ~~~e2 IV1~RIJ =- llo:~e2 v2 -;r.
Taking into account that R = yj, we finally get
F 11 =-~":~~ 2 V 2 • (6.118)
The force defined by Eq. (6.117) and observed in the reference
frame K relative to which the charges move, consists of the Cou-
lomb repulsion and the Ampere attraction (with an accuracy of the
factor 1/2). The force expressed in the form of Eq. (6.117) can be
used for the explanation of current interactions in conductors only
with certain stipulations. Neutral current-carrying conductors must
attract each other in these conditions. However, a current-carrying
conductor is neutral only in one reference frame (§ 6.1). That is
why the Coulomb repulsion ought to be taken into consideration.
For all that, it usually seems to be weaker than the attraction.
§ 6.1 t. An energy-momentum-tension tensor of an electromag-
netic field in vacuo. A transition to four-dimensional quantitie<;
combines the quantities whose interrelationship was imperceptible
in a three-dimensional approach. In the case of a free particle one
4-vector combines energy and momentum. An electric and a mag-
netic field constitute an electromagnetic field tensor in 4-space. An
energy and a momentum of an electromagnetic field turn out to be
components of a tensor which, apart from an energy (a scalar m
a three-dimensional case) and a momentum (a three-dimensional
vector), comprises also a three-dimensional-tension tensor of Max-
well. Here we shall have to quote the results of the Maxwell theory
in a three-dimensional form.
I. The energy conservation law for charges and a field. This law
follows directly from Maxwell's equations: multiplying Eq. (6.56a)
Maxwell Theory in. a Relativistic Form 223
::.calarwise by E and Eq. (6.57a) by H and subtracting the expres-

sions thus obtained, we get
HrotE- EratH=- jE- DE- iJH.
Making use of the following identities H rotE- E rot H=div [EH]
and (d/dt) (ED+ BH) = 2 (DE +iJH) (the last one is valid for
an isotropic medium in which D = eE, B = tJ.H). we get
ft (ED~ BH) = - JE- div(EHJ,
whence, after integration 0\'er an arbitrary volume r and using
the Gauss-Ostrogradsky theorem, we come to
•:. ~- ).JEdY-~SdS. (6.119)
The left-hand side of Eq. (6.119) represents a time variation of

an electromagnetic field energy in a volume r. This energy is
defined in the Maxwell theory via an energy density (an energy
per unit of volume):
w= ED~BH (6.120)
by integrating over a vohme:

W==). wd'P. (6.121)
Let us consider the simplest case of charges in vacuo. In this

case j = pv, while the force density acting on these charges, i.e.
a Lorentz force density, is
f'~p(E+[vBJ)~pE+[iB[. (6.122)
This force is introduced into the theory in order to correlate the
field theor} with the field of force acting on charged objects lo-
cated in the field. Eq. (6.119) features an expression jE. Multiply-
ing Eq. (6.122) scalarwise by v, we obtain fLv = pEv = jE. There-
fore, one of the terms of the right-hand side represents a worl<
performed by a field on a charge in this case. In accordance with
the energy conservation law this work must turn into a kinetic
energy of particles T. Consequently,
i/Edr~\t'vdr~ 4ft. (6.123)
The second term of the right-hand side of Eq. (6.119) represents

the Poynting vector
s~[EH], (6.124)
'"
while the integral itseH represents a Hux of the vector S through
the surface enclosing the volume r. The integrand also include;;
dS = ndS, a product of a surface area dS and a unit vector n
of 1ts normal. Hence, the energy conservation law for charges and
a field can be written down as follows:
f,<r+W)~--r SdS. (6,125)
The Poynting vector (6.124) is usually interpreted as an energy

Oux per unit time through a unit area oriented normally to the
Poynting vector. Such an interpretation does not necessarily follow
from the Maxwell equations. The direct consequence of the Max-
well equations is the integral relation (6.119) which can be r~
garded as the energy conservation law. It is clear that any S' ad-
dition to the Poynting vectorS, satisfying the condition div S'=O,
does not vary the relation (6.119). The generally accepted inter-
pretation, however, is confirmed by experiment.
2. The momentum conservation law for charges and a field. The
momentum conservation law can be treated as follows. Multiplying
Eq. (6.56a) vectorwise by B and Eq. (6.57a) by D, and adding
the equations obtained termwise, we get
"IH cot H] + eiE cotE]~ -1/B] +.,,[HE]- e" [EiiJ
We have taken into account that D = eE and B = 11H; finally:
"IHcotH] + eiE rotE]~ -liB]- e"f,IEH]. (6.t26)
Let us make use of the vector identity
a diva -Ia rot a)= a~a ( aaa~- ~ 6a~a~) m 11 .

Subtracting the left-hand and right-hand sides of Eq. (6.126)
spectively from the identity
llH div H + eE divE ==pE
(see Eqs. (6.56b) and (6.57b)), we get the final equation
a:a ( t!.EaEII + llHaHII + 6all e£2~1J.H2) mil=
~ pE +liB]+ e" i,-JEH],

wh1ch can be rewritten in the form
~~ m 11 =f+~IDB). (6,t27)
Maxwell Theory ln a Relaflvlsttc Form 225
where the tension tensor of Maxwell is introduced

(6.128)
The tensor (6.128), being symmetric in vacuo and isotropic me-

dia, is asymmetric in anisotropic media where it is defined accord-
ing to the last equation of (6.128). Integrating Eq. (6.127) over
an arbitrary volume in the region where an electromagnetic field
exists, we get
It was again assumed that in Eq. (6.127) we deal with free

charges in vacuo, subjected to the Lorentz force (6.122).
According to the second law of Newton
IfdY=Tt·
7
dP
(6.130)
where P is the momentum of charges enclosed within the volume 7'.

The integral with respect to volume entering into the left-hand
side of Eq. (6.129) transforms into the integral with respect to the
surface enveloping the volume 7':
(6.131)
The expressions
represent a force acting on an infinitesimal surface area dS who~e

normal's components are na. The vectors mp are unit vectors
of the Cartesian coordinate system. We could already write down
the momentum conservation law if we knew what should be re-
garded as a momentum of an electromagnetic field in matter. So
let us confine ourselves to the case of vacuum where [DB]=
= {l/c2 ) [EH] = S/c2. Then taking into account Eqs. (6.130) and
(6.131), Eq. (6.129) can be rewritten in the form
ft. (P +G)=~s To.,no.m$ dS, (6.132)
._,
226 Special Theory of Relallvlty
having defined a field momentum density g in vacuo as

g=S/c2 (6.133)
and consequently a field momentum in the volume r as
o~jgdr.
Equation (6.132) and the definition (6.133) express the

momentum conservation law. For a complete field, when on the
~~/2t~8(fo' _;u;;~a~ o. 7TCe~e~~o:V}a~o~:a~~t ~~~ne~0 ~~e:;~V;~ou~~;
the Maxwell equations yield only the integral equation (6.132), and
if one adds a component of an arbitrary tensor nfj. satisfying the
condition (iJnp/dxa) = 0, to each component of the vector G,
,f.
':Y
T~fl namp dS = r~ adr~ll
Xa
mfl dY = 0,
and Eq. (6.132) retains its validity all the same. Here we proceed
in much the same way as we did when selecting an expression for
the Poynting vector from the energy conservation theorem or find·
ing an expression for a displacement current density. Our selection
depends on the correctness of all of its consequences. The tension
tensor (6.128) in uacuo where D = eoE, B =~H. together with
the definition of a momentum (Eq. (6.133) ), yields reasonable
physical results.
In conclusion note that Eq. (6.132) makes it clear that the defi·
nitions of a momentum density and a tension tensor are closely
interrelated. Having redefined a definition for a momentum density.
\\"e modify at once an expression of Ta~ (see§ 6.12).
Let us summarize the results that we obtained for the case of
vacuum: as a consequence of the Maxwell equations, the momen·
tum density defined by Eq. (6.133) ought to be assigned to an
-electromagnetic field in uacuo. Then Eq. (6.132) expresses the New-
ton law: an increment of the total momentum of charges and of a
field in a volume "'F is equal to the sum of forces acting upon this
"\"Olume. These forces can be written down in the form of surface
forces, i.e. the forces acting on a surface enveloping the volume "'F.
A transition to four-dimensional terms can be accomplished .-ts
f
follows. First, let us prove that a 4-force density (see Eq. (6.54))
can be rewritten as a four-dimensional divergence of a tensor T,~~;:
/1 == +Fu,sk = aJx~.
Maxwell Theory l.a a Relafiuistic Form 227
where T," is an energy-momentum-tension tensor •; the components

of this tensor have the following form:
(6.134)
In the first term of the right-hand side of Eq. (6.134) summa-

tion is performed over m, and in the second one overs and n; F1,..
and fs,. are the corresponding components of tensors (6.29a) and
(6.31).
In order to obtain Eq. (6.134) we shall need the Maxwell equa-
tions expressed as Eqs. (6.60) and (6.67); here we shall rewrite
them in a more convenient form:
(6.135)
(6.136)
Let us transform the four-dimensional force density:
ft=+Fuh=7Ftk Z;; =7{a~, (Ftkfkt)-fkt a;:;"}· (6.137)

Here we made use of Eq. (6.135) and applied the rule of a product
differentiation.
Now we shall deal with the second term in the last link of Eq.
(6.137):
fkt a;: =f (fkt a;;," + fu a;;;)=
1"
=f(tkta::: +fkt~:~)=~fkt(a;:1" + 0::~)=

1 aFkl 1 . /1,1; 1 afkt
= - 2 1 --ail=- 2'V !";'
kl
1 _ fp:; a ,
--ail=- 4 'V !";' tfil (1kl) =
kl
= --i a~ 1 <fs,.F 8,.). (6.138)
The following operations are carried out in the equation chain

(6.138). The transition to the ~cond link is based on antisym-
metry of the tensors f,.l and F,tt permitting of exchanging indices
in each term of the product without changing the product in the
process·
(6.139)
• We shall denote the four·dimensional tensor (6.134) by the same letter T

that we used in the case of the three·dLmE'nsional tensor (6 128); this should
not lead to any misunderstanding because, as it will turn out, nine components
of (6 134) lor i and k changing from I to 3 coincide with {6.128).
,.
Consequently, instead of one term we take a half-sum of two ~qual

-expressions given by the left-hand and right-hand sides of Eq.
{6.139). The third link involves a substitution of mute indices
which does not change a summation result: the index l is replaced
by k and vice versa, i.e. flll a:••
vx1
is replaced by f., 0: 11 • The com-
vxk
mon factor f~e 1 is taken out of the brackets in the fourth link, while
in the fifth link Eq. (6.136) is used. In the sixth link of the equa-
tion we replaced F,., in accordance with Eq. (6.58); this operation
is valid only for vacuum.
Eq. (6.138). however, remains valid also for a uniform isotrop-
Ic medium. It can be easily shown that the components of tensors
f and S: are proportional as before in such a medium although spa·
tia1 and temporal components have different proportionality fac·
tors. From the general appearance of the tensors 5= (cB, -iE)
and f= (H, -icD) it is clear that the spatial components tie to·
gether the vectors B and H, while the temporal ones the vectors D
and E. In order to get the necessary relations D = eE and B =
= 1-1H. one has to assume
(6.140)
(6.141)
But then, starting from the fifth link of Eq. (6.138), the subsequent
chain of equations will be rewritten as follows:
~f111 a::,' ==~a- 1 fa~ ~;: +~b-'f~r4 ~:: =

0~ 1 a:
1 (fa~) 2 +
+ b~' a~, (f114) 2 ={ a~, (fa~Fa~)+~ a!, (f~r4F~r4)=
={ a~ 1 (fHFk/)=7 a~ 1 (fsnFsn)· (6.142)
The last link of Eq. (6.138) and the third link of Eq. (6.142) are
written according to the rule of product differentiation:
fH :~~· =fa!, (fH)2•

Since the factors a and b in Eqs. (6.140) and (6.141) are con·
stant, they can be brought under the differential symbol. And
finally, for the sake of convenience some other mute indices are
introduced in summation; in so doing, we are not changing the
sum f,uFflt = fsnFsn. Therefore, the results obtained for vacuum
and a uniform isotropic medium prove to be the same.
Ma:cr~~ell Theory fn a Relativlstlc Form 229
Of course, this result is formally obvious because in the SI

aystem vacuum is just one of uniform and isotropic media as long
as only the relations D = eE and B = 1-1H are essential.
The first term of Eq. (6.137) can be combined with the second
term of Eq. (6.138) or Eq. (6.142), given in a final form, provided
both terms are differentiated with respect to the same variables.
However, a differentiation with respect to another variable can be
performed by means of the Kronecker delta
a~~ =Ou a~t.

Now we can write down the expression for f1 in a complete form
'(see Eq. (6.137)):
/J =~ { a~ 1 (Fa,fkl) + ~ 0~ 1 (fs,.Ff,.)} ==-

=+ 0~ 1 { FtkF,,+~Ou(f 8 ,.F 8 ,.) }· (6.143)
Let us substitute the mute summation indices in the first summand:
the index m will replace k, and the index k will replace l in the
first and second summands. Then we finally obtain
fJ= a~k { +Ftmfmk+-icoll1 (fs,.F8 ,.) }~ ~~:", (6.144)
where T1, is defined according to Eq. (6.134).
T
Thus, the components of a 4-force can be expressed through
the components of a tensor r,,
depending on the field vectors E
and B. Recall that the components of tensors W and f are propor-
tional in vacuo according to Eq. (6.58), the proportionality coeffi-
cient being the same.
Owing to this circumstance and the definition of the tensor Tu1
(Eq. (6.134)), one can see it is symmetric in vacuo, i.e. T1, = T, 1•
1t implies that this tensor has ten independent components. In the
case of matter a 4-tensor loses its symmetric properties.
maL:~e~i~ ~~~~ ~~;tot~~ l~n~i~~;~~:~\sh:x:;pe;::s~o~h{~~,,~ ~~~~r~~
precisely the sum of pairwise products of the respective compo-
nents of matrices (6.29a) and (6.31). The requisite components
are immediately seen from the definition of tensors f(H, -icD)
and W(cB, -iE). Designating the coefficient in Oik as A, we find
A=fc.f,,.F8 ,.=-Jc·2(cBH-cDE)=B2H- ~B. (6.145)

The digit 2 appeares in front of the parenthesis because due to
antisymmetric properties of f and F the product of pairwise com·
230 Special Theory of Relativlfy
ponents yield~ the same expression c (BH- DE) tw1ce. Now the
equation of T1t components can be rewritten as follows (-/smo are
substituted for fmt):
Tur= -+F1mf11.,..+ 6stA. 16.146)
Now let us consid~r the individual .components. For example.

we shall find T11 :
Tu = -{-F,mf.m +A=- +Fufu-+F12f12- 7F,3f13-
-7Ft4f14+A=-+(cBzH~+c8 11 H 11 -cE~~:D")+ ~H- ~E ::::::~-

= -BH+ H"B"+ ExD" + 828 - ~E =
=H"'B"+E"D.,- DE~BH =H 1B1 +EtD,-w. (6.147)
We have found T11 to be a component of a three-dimensional

tension tensor of Maxwell (6.128). Similarly, it can be shown that
all components of the tensor TaP• that is the components whose
indices i, k take on the values from I to 3, coincide with a tension
tensor of Maxwell (6.128). It remains to consider the T111 compo-
nents in which at least one of the indices is equal to 4. We shall
start from T44 :
T44=-+F4,f4,+A=
=E:r:Dx+E11 D11 +EzDz+ ~H- 028 = B~ + 02B =w. (6.148)
The component T44 turned out to be equal to the electromagnetic;:

field energy density. Now let us find T14:
T14=T41 = -7F~,f~ ... = -+(F12f42+ Fl3fot,)=
= - ic (D 11 Bz- DzB 11 ) = - iceollo [EHJ" =
=-1-S:~~=-ic~=-icg"'. (6.149)
In much the same way

T24=T 42=-icg11 ,
(6.150)
T34= T43 = - icgz.
The components T14, T24, T34 turned out to be proportional to the
components of the electromagnetic field momentum density g =
= S/c2• Later on it will be clear (see Eq. (6.153)) that we deal
here with a momentum density indeed, and not an energy flux to
Maxll!ell Theory rn a Re1atlv{stic Form 231
which a momentum is proportional. Write out the matrix of the

energy-momentum-tension tensor for a case of an electromagnetic
field in vacuo:
Tn T" r, -icgJ:
r, r, T, -icg11
-( r., -~cg).
r~
-fsJ:
r,. r,
-fs~~
r, -icg:r
= -+s
-fs" (6.151)
The upper left square comprising nine quantities defines a tension
tensor of Maxwell. It becomes a correct quantity in a relativistic
case when bordered with energy quantities 8 and w. Let us make
sure that having composed the tensor T,~e, we obtained the energy
and momentum conservation laws expressed in a three-dimensional
form by Eqs. (6.125) and (6.132). Consider now the spatial com-
ponents of a 4-force:
Iu =---ax; - C a
dTa!J 1 dSu
(icl) =
dTuiJ
ax;- - 7I dSu
---at
dTufl
=ax; dgu
-at· (6.152)
We took account of a three-dimensional momentum of an electro-
magnetic field in uacuo having the components ea = S«/c 2• Multi-
plying each component fa (a= I, 2, 3) by its respective unit
vector ma (a= I, 2, 3) and summing up the values thus ob-
tained, we get (f = f«m« is a three-dimensional Lorentz force}
(6.!53)
Integrating the identity (6.153) over an arbitrary volume, we

get
\tdV+\~dr~\•Jx·•
1" 1" 1" fl
m.dr. (6.!54)
The left-hand side of Eq. (6154) features a variation of a total mo-

mentum of particles as well as that of a total momentum of a
field:
~tdr=%. ~*dr=ft-~gdJ"= ~~. (6.155)
Applying the Gauss-Ostrogradsky theorem to the right-hand side
of Eq. (6.154), we get
~a~~~.~ n~gd:t'=~Tufln!JmadS=~T~~. 11 numpdS; (6.156)

.,. p s s
the last transition makes use of the symmetry of the tensor Te~.~·
Now then, we have arrived at the momentum conservation law
(Eq. (6.132)) and have made sure that the components T14, T24,
T34 are indeed proportional to electromagnetic field momentum
components. The expression TCl.~nClm~ can be considered not only
as a force acting on a surface element, but also as a momentum
flux through that surface element. The quantity Tce~m~ yields a
vector component of this flux. Surely, both these interpretations an~
equivalent.
Consider now f4• On the one hand, according to Eq. (6.54)
/,=~(vE)=+(fv), (6.157)
and on the other hand
f4 = aJx~ = ~::· + ~~; + ~~:' + ~~:4 =-fdivS+ a~~~f).

(6.158)
Consequently, Eq. (6.158) can be rewritten as
~~ +divS+(vf)=O. (6.159)
Integrating Eq. (6.159) with respect to an arbitrary volume of a

field and taking account of Eqs. (6.121) and (6.123), we get
f,(T+W)=-~SdS, (6.160)
s
with the Gauss theorem being applied to the term div S of Eq.
(6.159). This is precisely the energy conservation law (Eq. (6.125)).
Thus, in the relativistic theory the Maxwellian tensions, momen~
tum and energy of a field in uacuo amalgamate into a single tensor
quantity, the energy-momentum-tension tensor. The energy and
momentum conservation laws manifest themselves via a single re-
lation.
The symmetry of the energy-momentum-tension tensor consti-
tutes a fundamentally important property. Owing to this, the fun-
damental relationship between the energy and momentum flux
densities follows immediately for the case of an electromagnetic
field in uacuo:
(6.161)
One can readily make sure that the spur of the tensor T,,h i.e.
the sum of its diagonal components, is equal to zero.
Having established a tensor character of tensions, of a momen-
tum, and of an electromagnetic field energy flux and densitv. we
Maxwell Theory ln a Relatlulsf/c Form 233
automatically obtain the rules according to which these "quantities

are transformed on transition from one inertial frame of reference
to another. We shall write out only those transformation formulae
that will be needed later. Substituting the component values from
Eq. (6.151) into the general equations (A.I. 31), we get
Tu= Tn =f 2 (T~.~:- 2 ;.s~--;.-w'). (6.162)
T44=r 2 (w'+2~S~-{;-r:.~:). (6.163)
T.~: 11 =T•2=r(T~ 11 --;s~). (6.164)
gx=r2 {(1 +~)g~+~w'-~T~x}• (6.165)
g 11 =r(g~-;-T~ 11 ). (6.166)
g.~r (g;--f,r;,). (6.167)
§ 6.12. An energy·momentum·tension tensor of an electromag·

netic field in a medium. The Minkowskl tensor and Abraham
t"""~~- An energy-momentum-tension tensor (EMT) in a medium
attracts interest primarily because this tensor is associated with a
momentum of an electromagnetic field in a medium. The latter
quantity is directly related to the quantities observed in an ex-
periment, e. g. to the pressure of light. This tensor, however, is not
defined in a unique manner in a medium, and the discussion about
its "proper form" is still going on.
Let us find the general form of an energy-momentum tensor in
a uniform isotropic medium. It was shown in § 6.11 that the
general form of energy-momentum tensor components in a uniform
isotropic medium does not differ from that in the case of vacuum;
in both cases (see Eq. (6.146))
Tu:= -+Fimfllm +61,.A. (6.168)
However, the proportionality factors between spatial and tem-

poral compcnents f and ij are different in a medium (see Eqs.
(6140) and (6.141)), and the tensor To. defined according to Eq.
(6.168) turns out to be asymmetric in contrast to the tensor (6.149).
Asymmetry arises due to temporal components of the tensor; spa-
tial components are symmetric, at least in an isotropic medium.
It is indeed easy to see th:::tt spatial components of the tensor T 11,
ifl a medium differ from thos~ in vacuo only by the values of a
Special Theory of Relatiuity
and IJ.. So, for example

Tu=-+Ftmftm+A=
=-~(cB !..!..+cB !.JL -ceE2)+A =

c z "' g "' "'
= - flzBz-H 11 8 11 - H 11 Bx+ fl 11 B11'+ExDx+ BH;DE =
=flxBx+ExDx- ED~BH.
This expression coincides with the value of T11 given by Eq.

(6.147), the only difference being the quantities e and IJ. replacing
the quantities eo and llo of Eq. (6.147). Exactly in the same way
we shall find that
Ttt=-7F4mf4m+A= ED~BH =w.
However, if
T14 = -ice~J,[EH) 11 =- (i/c)(e1J./e0r.r.o)S 11,
then
T., ~-(ifc) fEHf, ~- (i/c) S,.
Consequently, the energy-momentum-tension tensor in a uniform
isotr'lpic medium, obtained by a direct transformation of a 4-force.
is not symmetric any more. It is called the Minkowski tensor, and
its components have the form
r.,
M ( r2,
Tl*= r,.
-(i/c)S,
(6.169)
Eq. (6.169) includes a refraction index n= .YeiJ.!SoiJ.o· The momen·

tum density corresponding to tensor (6.169) (i.e. the components.
T14, T24 , T34 ) turns out to be equal to
gM~ *S~fDBf. (6.170)
The superscript "M" appearing in a momentum density symbol

points to the fact that this density corresponds to the Minkowskl
tensor. The momentum density of a field in a medium (set: Eq.
(6.170)) exceeds n 2 times that in vacuo (Eq. (6.133)).
Maxwell Theory ln a Relatlr~lsflc Form 235
It ~~ often suggested to use Eq. (6.133) for the description of

the momentum density of a field in a medium as well. Surely, this
\':ay we subdivide a total momentum into a momentum of a field
.and a momentum of a medium itself. However, a momentum den-
sity of a field can be separated in the form of Eq. (6.133) only if
another energy-momentum tensor, different from the Minkowski
tensor, is utiltzed. It is also important to keep Eq. (6.133) because
it yields the most general formulation of the energy inertia law.
Since an energy flux is described by the components T4 a. of an
energy-momentum tensor and a momentum density by the compo-
nents Ta.t• Eq. (6.133) indicates the tensor's symmetry. Therefore,
we should construct a new symmetric tensor satisfying the follow-
ing conditions: Ta.t = -icga. and T4a. = -(i/c) Sa.; Eq. (6.133)
should be satisfied as well. The three-dimensional tension tensor
T a.P should coincide with the three-dimensional tension tensor of
Maxwell (6.128). Such a tensor was suggested by Abraham in the
following form:
A ( Tall -icgA) S
T11,= -(i/c)S w ' gA=g=cr· (6.171)
Since Td =Til« and owing to Eq. (6.133), this tensor is sym-

metric. The introduction of the Abraham tensor implies an ap-
f:r~~~n~~ ~~ ~h~olub~=h~%cfo;~!~nfo 0fin~ ~se%~~~iT~~~. f~~~:1J\~:t
the Lorentz force density components are related to the Minkowski
energy-momentum tensor by the ratio
Ia ~ ar:l,
axk •
(6.172)
This is precisely how the Minkowski tensor was obtained.

It follows directly from Eq. (6.172) that
ar""-
frzmaE:lfL= ax:~ nla· (6.173)
Writing out the sum on the right-hand side and interchanging
the terms of the equation, we get (see Eq. (6.169))
ar""-
a::
agM
mu-(jf""=fL, (6.174)
where fL denotes the Lorentz force density.
No\\ let us write the expression (aT~Jax~)mo. in full, taking into
account Eq. (6.171):
a~:~ mo.= a:xal! mo.- a:,A 51: aJxa(> nl.a- a:; + -:r (gM- gA).
• ' ' (6.175)

236 Special Theory of Relatl1.1ity
The second link of Eq. (6.175} takes into account that the tem.wn
tensor of Maxwell is the same both in the Minkowski and in the
Abraham tensor; the third link of Eq. (6.175) is the identical tran'i-
cription of the second link. But the first two terms of the last link
of Eq. (6.175) can now be substituted according to Eq. (6.174) to
yield
(6.176)
The right-hand side of Eq. (6.176) contains a term f4 representing

the derivative of a momentum density; according to the second law
of Newton the derivative of a momentum density with respect to
time makes up a force density
fA=f, (gM- gA) =f, {[DB]--;\- [EH] ). (6.177)
In an isotropic medium
/4 =-ft-(e!J.-eo!J.o)8=-ft-(n2;l 8). (6.17~}

A force density given by Eqs. (6.177} and (6.178) is referred to
as the Abraham force density.
The Min':owski tensor and the Abraham tensor furnish different
expressions for an electromagnetic field momentum density. Let us
write out the corresponding expressions for a plane electromag~
nel.ic wave momentum density. In the case of a plane electromag-
netic wave propagating in a uniform isotropic dielectric the rela-
tionship between the value of the Poynting vector S, the mono-
chromatic wave phase velocity v and the wave energy density w
is given by the simple equauon:
S=w•v, (6.179)
v = cfn = (e!J./e0tJo)'1•. (6.180)
Eqs. (6.170) and (6.171) furnish the momentum densities:
g"=(n'lc')S=(n'lc')w(c/n)=(w/c) n, (6.181)
gA=(S/c~=(lfc')w(c/n) =(w{cn). (6.182)
Let us assume that electromagnetic field energy can be quan~
tized, i.e. w = N1't.(ij, where N is the number of quanta in a unit of
volume. Then from Eqs. (6.181) and (6.182) we obtain the fol-
lowing quantum momenta in a medium:
p"' =(hoofc) n, (6.183)
pA = (h6>{cn). (6.184)
Wh1ch ot these two equations is "true"? The technique of secondary

quantization of an electromagnetic field in matter rE'sults in Eq.
(6.183). Let us assume that a quantum momentum in a medium p
is defined as follows:
(6.185)
where s is a unit vector in the direction of wave propagation,

and a 4-vector of a quantum energy-momentum takes the form
P(-¥-ns. -7-)- (6.186)
Using Eq. (6.186), one can obtain a correct expression stipulat-

ing the Vavilov-Cherenkov radiation (see Chapter 7). This circum-
stance seems to indicate that Eq. (6.186) and the Minkowski tensor
are to be preferred. However, a meticulous usage of any of these
tensors provides a correct result. The point is that both tensors
satisfy Eq. (6.127) which is the consequence of the field equations.
It is important to know how an electromagnetic field momentum
is defined in matter. In' a general case, it is impossible to split the
total momentum of a field into a fraction pertaining only to matter
and that pertaining only to a field. But this is precisely what it is
done when the Abraham momentum is introduced. When a light
wave passes from vacuum into a medium, its momentum is not
totally transferred into this medium; a fraction of the momentum
is transmitted to the medium itself. The Minkowski tensor is uti-
lized whenever a total transmitted momentum is considered; if we
deal with a momentum related to radiation in a medium, the Ab-
raham tensor should be used. Eq. (6.185) provides a correct value
for a photon momentum in the case of the Vavilov-Cherenkov ra-
diation (see Chapter 7) because an overall momentum which a
Cherenkov electron transmits to a medium is of prime importance
here. The overall momentum transmitted to a photon in a medium
is just equal to hwn/c. No wonder that quantization of an electro·
magnetic field in dielectrics results in Eq. (6.185) describing d
momentum. This expression represents an overall momentum of
an electromagnetic field, this momentum being associated both with
a field and with a matter (see§ 7.7).
As to a force acting on a matter, it is related to the Abrahar:1
force and, naturally, to the Abraham tensor. An attempt to measure
the Abraham force was undertaken in 1975 which apparently suc-
ceeded.
Fig. 6.7 illustrates the experimental arrangement. A disc made
of barium titanate (e ::::::: 4000, J.l.::::::: IJ.o) has a small hole in the
centre. Both rims of the disc are coated with aluminium; conse·
quently, it forms a cylindrical capacitor. The disc is so suspended
on a long tungsten filament that it can perform torsional oscilla-

tions between the poles of a de electromagnet generating a 10 kGs
field. An ac voltage with a 150 V peak value is applied to the m-
ternal rim electrode, while the outer electrode is grounded by
means of a thin gold wire not affecting the disc's oscillations. The
voltage is applied in phase with
Suspe'!jion o!ldo c volfop the characteristic oscillations of
the disc.
The Abraham force can also be
written in the form
I
fA=So!lo(XmXe-1)*=
unpten =e,!'o(Xmx, -1)[EHI, (6.187)
8 toment
where
~ 0.2mnftlia
Xe = efeo, Xm = !li!J.o (6.188)
and the fact that the magnetic field

s is constant (Ii = 0) is taken
into accouht. In the case of ba-
/Jrountflll§ rium titanate Xm ~ I and
quently
Fig. 6.7. An experimental observa·
lion of the Abraham force. fA=[e 0 (x,-1)E, ""HI=
=[P, ""HI; (6.189)
the last equation follows from the fact that in a uniform isotropic
medium
P=D-eoll=(e-e,)E=e,(x,-1)£. (6.190)
The physical meaning of the "Abraham force" is obvious in the
specific case just considered. P is the fraction of the displacement
current caused by the motion of bound charges. In essence, this
is the Ampere force. It is not difficult to find out that this force
induces a torque, the electric field being directed radially. Surely,
the essential point is that no other forces associated with the pres-
ence of an electromagnetic field contribute to the torque.
The authors who proposed this experiment claim that the ob-
served oscillations of the disc are consistent with the calculation
based on the presence of the Abraham force. We should point out
here again that this experimental result, however interesting per
~e. by no means "chooses" between the tensors (6.169) and (6.171).
Some additional remarks concerning a choice of an expression for
a photon momentum in a medium can be found in§ 7.7.
§ 6.13. An energy-momentum-tension tensor of a spherically
symmetric charge. If an electric charge in uacuo is at rest in the
frame K, there is only an electric field in the frame and the energy•
M11xwell Theory ln. a Relativislic Form 239
momentum·lension tensor can readily be wntten in the form
Tu.= ( r.,0 wo) . (6.191)
where w = e0 £2/2 and Ta.~ = e 0 E~~,Ee,- O(t 11 w. If the charge

moves relative to the frame K' at the velocity - V, its tensor T/11:
can be found through the use of the general equations for th~
tensor component transformation. Specifically, in order to trans-
form the momentum density along the x axis, that is (i!c) T14 , and
the energy density Tu, we have the following equations (s~e
A. I. 31):
+
Ti4 = - iBf2T11 f 2iBT44 = iill'2 (w- Tu),
T4-t=f2T44 -B2f 2T11 =f2 ( w- ~T 11 ).
Let us find the total energy and total momentum of a point
charge, taking into account that a transition from a volume ele·
ment dY' in the frame K' to a volume element dY in the frame K
is effected by means of the formula d"f'' =-Fdr:
U' -I r;, dY' -I r;,-j.-dr- r I (w -ll'Tu) dY, (6.192)
0~=+ ~ Tl4dY'={- ~ Tl 4 f-d"F=~~ (T 11 -w)dY. (6.193)

It is obvious that
~ wdY=~ e;£ 2
d'?=U.
If in the charge's proper frame of reference K it possesses a sphe 4
rica! symmetry, so that£!=£!=~= £ 2/3, then
I Tudr-s,l (e!-f)dr---T I E'dY- -{'

Consequently, Eqs. (6.192) and (6.193) take the form
u•-ru(1 +f). \0.194)
o~=-f~ru. (6.195)
The momentum components G~ and 0~ turn into zero, so that
G'=-~;ru. (6.196)
The appearance of the "minus" sign tn the last equatton is due to
the fact that the charge moves at the velocity -V relative to the
SpecliJl Theory of Relatioltg
frame K'. Comparing Eqs. (6.194) and (6.195) with the equations
transforming a momentum and energy of a particle on transition
from the proper frame of reference to an arbitrary one (see Eq.
(5.49)). we see that these equations are different. In former time<;
.an effort was made to treat an electron mass as an electromagnetic
one using the relation
m=U/c2• • (6.197)
Eq. (6.196) shows that such an interpretation gets into trouble
since to "confine" a charge some additional forces are required
neutralizing repulsion, i.e. the additional energy that was not ac-
counted for. Having taken into account mechanical stresses, we
can obtain the following relations:
G'~-rv~. u·~ru (6.198)
in a complete agreement with Eq. (5.49). The details can be seen

in [13].
§ 6.14. The field potentials in a moving non-conducting me-
dium.* In§ 6.1 we introduced a 4-potential of an electromagnetic
field in varuo. Surely, an electromagnetic field can be determined
immediately from t':e Maxwell equations, without resorting to po-
tentials. In many cases, however, the utilization of potentials as
intermediate qua· .tities determining the fields E and B, proves to
be very convenient if only because the number of functions to be
determined decreases. Knowing only four components of a vector
potential, one can find from them all components of an electric
and a magnetic field. The pot~ntials are still more convenient to
use in electrodynamics of moving media where the material equa-
tions (6.74) and (6.75) turn out to be much more complicated than
in the case of a stationary medium.
It will be shown below how one can obtain the expressions for
field potentials in a moving medium. To illustrate an application
of such potentials, we shall consider the propagation of a plane
electromagnetic wave in a medium moving relative to a stationary
observer. This example has a direct bearing on problems analysed
in Chapter 7. In this section we apply the methods of tensor al-
gebra described briefly in Appendix I, § 3.
Now let us derive equations for a 4-potential in moving media.
We shall be describing a field in a moving medium by two tensors:
F,~ (sec Eq. (6.29)) and f,~ (see Eq. (6.31)). The tensor F,~ is
sometimes referred tO as a ne[d fenSOf, and the tensor f111 as an in•
duction ten.'>Of.
• §§ 6.14 and 615 are wntten by 8 M Bolotovsky and S. N. Stolyarov.

Maxwell Theory ln a Relativistic Form 241
Let us introduce a four-dimensional field potential in a medium,

en. having defined it by the following relation
Fu=c(a~D*- aiD~). (6.199)
ax 1 ax,
coinciding with Eq. (6.28). Knowing the four components of the
potential Ill*• we can use this equation to determine all components
of the tensor F1~. that is the magnetic induction B and the electric
field E. In order to describe wholly an electromagnetic field in a
medium, one needs to know also the components of the tensor f.~.
I.e. the components of the vectors of the magnetic field and electric
induction. If the field tensor {j is known, the induction tensor r can
be determined by means of the material equations (6.79) and
(6.80) relating the components of these two tensors. Recall thRt
the correlation between the tensors {j and r in a vector form is
given by the Minkowski equations (6.74} and (6.75).
The material equations (6.79) and (6.80} defining the relation·
ship between the tensors {J and f can be written in the form of a
single tensor relation
(6.200)
where the tensor of the fourth rank e1k1rn is so chosen as to make
the Minkowski equations (6.74) and (6.75} valid. It is easy to show
that the ten~or of the form
ewm=-dc(llu-xc- 2UIUt}({)*m --x.c- 2UkUm) (6.201)
possesses the necessary properties. Here 6,1 is the Kronecker delb

defined by Eq. (A. I. 4), and u~ are the four-dimensional velocity
<'Omponents. This four-dimensional velocity was discussed in
V
Chapter 5 and shown to have the components (fV, icf) where V
is the three-dimensional velocity of a medium motion. The dimen-
sionless constant -x. is defined via the refraction index n:
-x.=....!l!....-t=n2 - l , n=.:_= · / e" . (6.202)
Eo"o v V eo"o
It is easy to see that -x. = 0 in uacuo, and the relation (6.200)
between the tensors f and {J takes the form
f 1 *="~ F'*' (6.203)
corresponding to the well-known relations between the fields E

and H, and the inductions D and B in uacuo:
(6.204)
242 Special Theory of Relatiuify
In a stationary medium the tensor e,kl"' of Eq. (6.201} yields the

following relations between the fields and inductions:
D~•E, B~~H. (6.205)
This is easy to ascertain, having assumed U 1 = U2 = U 3 = 0
an<l U 4 = ic in Eq. (6.201).
Since the components of the tensor. F,m are expressed via the
components of the four-dimensional potential <D,, and the compo-
nents of the induction tensor f,k are related to F1m by Eq. (6.200),
the components of the tensor f,k can also be expressed through the
components of the four-dimensional potential ¢1,. In other \\Ords.
to determine all components of fields and inductions in moving
media, it proves to be sufficient to know the four functions <D,.
Now let us derive the equations for field potentials in moving
media. For this purpose let us make use of Eq. (6.60):
ar,k
ax;;=s,.
Substituting f,t from Eq. (6.200), we get
iJFlm
e,klm ax;:= s,. (6.206)
Utilizing Eq. (6.201) which gives an explicit expression for the

tensor e,klm, as well as Eq. (6.28) expressing Ftm through the po·
tential components Q)t, we reduce Eq. (6.206) to the following
form by simple transformations:
.2...(6 11 -xc- 2U1U1){_j_[O(()- -xc- 2 U~zUm alD"']-
J.lC iJxk dx 1 dxk
- [-4-- xc-•(u, ....£.._)'] <1>

iJx~ iJxt
1} ~s,. (6.207)
Now multiplying both sides of Eq. (6.207) by the tensor
(6 1a + 1 :x c-2 U1Ua)
and making use of the relationship
(6u-xc- 2U,Ut)(6 1a+ l : x c- 2U1Ua)=ll 1a, {6.208)
which ~s easy to check, we finally obtain
[4-xc-
iJx
(U•...!!__)'
11 ax,.
2 ] Q)u.- ...!!..._(~- xc- 2 U~cUm
iJxa ax~
OlDm)
ax,.
~
=- j.Lc( ll a + 1~ x c- 2U 1Ua) s
1 1• (6.209)
Maxwell Theory ln a Relatluistic Form 243
The system of equations (6.209) defines all components of the

potential <Da from the given field sources s1 in a moving medium.
This system can be simplified provided a well chosen additional
condition is imposed on the potentials, e.g. it is required that the
following relation is to be valid:
~~: - xc- 2UkUm aa~; = 0. (6.210)
This condition is a generalization of the well-known Lorentz con-
dition imposed on potentials in vacuo (see Eq. (6.8)). The feasibi-
lity of Eq. (6.210) is demonstrated in the same way as in conven-
tional electrodynamics.
When condition (6.210) is satisfied, Eq. (6.209) becomes simpli-
fied and takes the form
{~
iJx 11
-xc- 2 (u~: _!_)'}<Da=-
ax 11
J.i.c(0 1a + -'-c- 2U1U,_)s,.
I+~
(6.211)
The system (6.211) is more convenient as compared to the sys·
tern (6.209) since it comprises four equations, each of which in·
eludes only one component of a vector potential (a= I, 2, 3, 4).
For given external sources, the solution of the system (6.211) fully
defines the field generated by these sources in a moving medium.
If a moving medium has an interface, the system (6.211) should
be supplemented by requisite boundary conditions (see§ 6.8).
As an example of solving the equations obtained, let us consider
an electromagnetic field in a moving medium in the absence of
external sources, both currents and charges. Since in this case all
s, = 0, the system (6.21 I) turns into a system of the four uniform
equations:
{ ~- xc-
iJxll.
2(Uk_j_)'}
cl>,_=O.
ax_
(6.212)
Due to the additional condition (6.210) only three out of the four
quantities CD,_ are independent. Accordingly, we can assume CD 4 =0,
and treat the remaining three quantities CD 1, CD 2, 11)3 as components
of some tensor which we shall denote by A. Thus we see that in
the case of such a calibration the vector potential <Da for a moving
medium is a three-dimensional vector potential A.
In this case the system of equations (6.212) can be rewritten
in terms of the potential A:
{ (fi--}r~)- c'~:=~2) [{VV) +iJ}A=O, (6.213)

where B = V/c, under the additional condition
divA- ,,~:~~'J [(VV)+f.](VA)~O, (6.214)
244 Spectal Theory of Relativity
which lollows from the additional condition (6.210) at 11>4 = 0. If

we know how to solve Eq. (6.213) with respect to the potential A,
the fields E and B can be expressed through A according to Eq.
(6.28) which in our case takes a simple form
B~rotA, E~-~~· (6.215)
Knowing E and B, we can find D ~nd H by means of the Min·

h.owski equations (6.74) and (6.75) for a moving medium.
Eq. (6.213) describes propagation of free electromagnetic waves
in a moving medium. Free electromagnetic waves usually imply
a field in the absence of charges and currents. Now let us pass
over to the solution of this equation. We shall be seeking the solu·
tion for the vector potential A in the form of a plane electromag-
netic wave:
A= A¢1 (wt-.tr), (6.216)
Substituting this expression into Eq. (6.213), we obtain
{(-k2 +*)+ c2 7:=~~) (kV-ID) 2 }Aoel(lllt-•r)=0. (6.217)
Eq. (6.217) shows that the plane wave amplitude Ao differs from
zero only for those waves that satisfy the condition
( - k2 + ~) + c27: =~2) (kV- ID)2 = 0. (6.218)
Eq. (6.218) can be readily derived from the dispersion equation

\\thich IS valtd for plane monochromatic waves in a stationary me 4
d1um:
k 2 -fJil/rP=O, k2 -=E/tJ,
\\'e shall rewrite it in the form
(k2-~)- n~:;l fJil=O.
Here the parentheses enclose the square of the four-dimensional

wave vector in vacuo k ( k, i -7"),
the quantity which is invariant
relative to the Lorentz transformation. The quantity enclosed in
the parentheses retains its appearance Jnd numerical value in all
inertial frames of reference. The second. !l.ddendum of the last equa~
tion transforms as the frequency ID. In the reference frame in which
a medium moves at the velocity V, ID should be replaced by
ID'=(ID- kV)/-y'l- V~jcl
tsee 4 7.2).
Maxwell Theory in a RelaUvistcc Form 243
Due to these considerations the dispersion equation takes the

form (6.218)
k 2 -~- ,~~:=~) ((J)-kV) 2 =0
in the reference frame relative to which the medium moves at the

velocity V. This condition defines a relationship between the wave
vector k and frequency (I) of a plane electromagnetic wave propa·
gating in a moving medium. The additional conrlit1on {6.214) for
such a wave takes the form
(Ao. k+V ,.~;:::~.> (ro-kVJ)~o. (6.219)
From Eq. (6.219) requiring that the scalar product turn into
zero, it follows that in a moving medium the vector Ao is perpen·
dicular not to the wave propagation direction defined by the wave
vector k, but to a linear combination of the wave vector k and the
velocity vector V of the medium. In the two specific cases, when
a wave propagates in vacuo (n =I) and when a medium is sta-
tionary (ll = 0), Eq. (6.219) turns into the well-known relation of
free transverse electromagnetic waves: Ank = 0 from which it fol·
lows that in a free electromagnetic wave lhe vectors E, H, B and D
are perpendicular to the wave vector, that is to the wave propaga·
tion direction. In a moving medium, however, the waves are not
transverse, generally speaking. Indeed, in the case of a plane wave
(Eq. (6.216)) the fields E and B are determined according to Eq.
(6.215):
B = - i [kAoJeHIIJI-krl, E = - i(J)Aoei{IIJI-krl.
Whence it is seen that the vector B is perpendicular to the wa\8
vector k while the vector E is not (since the vector A0 is not trans·
verse according to the condition (6.219)).
Eq. (6.217) relating the wave vector k and frequency (I) of a
wave in a moving medium includes the scalar product kV. This
means that the wave propagation conditions depend on the angle
between the propagation direction, or the wave vector k, and the
velocity of a medium V. This circumstance indicates the pheno·
menan of carrying away of light by a moving medium. Let us
consider this phenomenon in detail in the case of small velocities.
Since the quantity 8 = VIc is small, we shaJI ignore aJI values
of 8 having degrees higher than one in Eq. (6.128). We shaJI get
fz--k + 1[~-2w(kV)]=O,
2 2
n ,-;
or
~-2kV(t-*)w-*=0.
246 Special Theory of RelatiFJlty
Solving the obtained quadratic equation with respect to w m the

same approximation, we get
m~±__!!._+kV(I-_!_)·
..ye; n1.
(6.220)
From the two signs in front of the first addendum on the right·
hand side one must choose the pill§ sign since in the case of
V = 0 we have to get the well-known relationship between w and
k in a stationary medium.
m I ' (6.221)
k=-v'eji"=-;
here we introduced the refraction index of a stationary medium n.
The quantity c/n is the phase velocity of light in a stationary me·
dium.
The angle between the vectors k and V being denoted by 9, Eq.
(6.220) takes the following form
T=.;- + V cos a (I-~) (6.222)
for the indicated choice of a sign.
Just as in the case of a stationary medium (see Eq. (6.221)), the
quantity w/k in Eq. (6.222) defines the phase velocity of light but
this time in a moving isotropic medium. Comparing Eqs. (6.222)
and (6.221), we see that the phase velocity of light in a moving
medium is different in different directions. If light propagates
along the motion of a medium (cos a= 1), the phase velocity
is equal to
-T=.;;-+(1-~)v.
H light propagates against the motion of a medium (cos a =
~-I). then
T=f-(t-,!r)v.
The factor (I -l/n2 ) is the so-called light drag coefficient which
was experimentally measured by Fizeau with water serving as a
moving medium.
§ 6.15. The field potentials in a moving conducting medium.
Prior to dealing with field equations in a moving conducting me-
dium we shall recall the main facts concerning the propagation vf
v.aves through a stationary medium in the presence of conduct-
ance.
In this case the Maxwell equations take the form
rotE=-li, rotH=D+J+aE,
(6.223)
divD=p, divB=O.
Maxwell Theory rn a Relativistic Form 247
Here p and 1 are the charge density and current density induced by
.. extraneous" sources. Subsequently we shall confine ourselves lo
consideration of equations (6.223) in the absence of "extraneous"
sources, i.e. assume p = 0 and 1 = 0. The solution of equations
(6.223) will be assumed in the form
E = Eaef (.tr-wf), D = Daet l.tr-llt),
H = Haett.tr-wtl, B = Baef(.tr-wtl, (6.224)
\\'here E 0, D0, H 0, B 0 are constant amplitudes that do not depend

on either coordinates or time. Thus we seek the solution as a plane
\Vave with the wave vector k and frequency (!).
Substituting the expressions for fields (6.224) into equations
(6.223) and taking into account that p = 0 and j = 0, we obtain
~~p[~~~~:i~' fi::a:~raic, not differential, equations interrelating
(k, EoJ=CilB0 , [k, H 0 }=-CilD0 -iaEo.

(6.225)
(kD0) - 0, (kB0) - 0.
We have made use of the following equations:
rot (Eae 1• ' ) = l(k, E0 Je 1•',
div (Eoet.tr) = l (kEo)ef*r.
Eqs. (6.225) should be supplemented with material equations
defining the relationship between field and induction values. In the
&implest case of an isotropic stationary medium we assume the
following relations:
(6.226)
We shall suppose that the quantities e and 1-1 do not depend on
field amplitudes. In so doing, we ensure a linear relationship be-
tween the fields E 0, H 0 and inductions D0, B0 respectively. However,
e and 1-1 may, generally speaking, depend not only on field values
but also on the frequency (t) and wavelength A = 2n/k. When e
and 1-1 depend only on frequency, we observe a frequency disper-
sion. But if e and 1-1 depend on a wavelength, the medium is said
to possess a spatial dispersion.
If the material equations (6.226) are substituted into Eqs.
(6.225), one gets the equations involving only the field amplitudes
Eo and Ho:
(k, E 0 }=J.1WH0, (kH0 J=-I!<ilE0 -iaEo,
(6.227)
e (kE0 ) - 0, "(kH0 ) - 0.
Heremafter we shall assume e and 1-1 not to turn into zero. Then
from the last two equations it follows that the fields Eo and H0
24~'------------s~~-"_•l_r_••-•~~-·~f_R,_w_u_oll~y-------------
are perpendicular to the wave vector k. Such waves are referred
to as transverse waves. We shall not discuss the longitudinal
waves arising when, for example, e = 0. (Then it is easy to see
that Eo II k).
Multiplying vectorwise the first equation of (6.227) by the wave
\'ector k, we obtain
(6.228)
Now the value of [kH0] can be taken from the second equation of
(6.227), and the double vector product is written out with the help
of the well-known formula
[k [kE0[] ~ k (kE0 ) - k 2E 0 ~- k 2E 0•
In the last equation we took account of the transverse character of
the field E 0, i.e. the relation (kEo) = 0. As a result of these trans·
formations we obtain
(k 2 - SIJ.m2 - iumjJ.) E0 = 0. (6.229)
Consequently, the amplitude Eo of a transverse electromagnetic
wave in a stationary conducting medium can differ from zero only
if the following equation is satisfied:
(6.230)
This condition is referred to as a di,;iJersion relation. One can read-
ily see that the same disper3ion relation (6.230) should be satis-
fied for the magnetic vector H 0 in a transverse wave to differ from
zero. In order to show this, one should multiply vectorwise the
second relation of (6.227) by k and then make use of the first re-
lation.
Let us assume that the wave frequency 0> is a given quantity and
the field depends only on the coordinate z and on time. Then we
can represent the fields E and H in the form
E=Eoetl<z-twl, H=Hoet""-twt, (6.231)
It follows from (6.227) that we can assume the vector Eo to be
directed in the positive direction of the x axis and the vector Ha
in the positivE' direction of the- y axis. Hence the three vectors k,
E<'· H 0 form the right-hand triad of vectors.
!n the case of a fixed frequency 0> the dispersion equaticn (6.230)
yzelds the :allowing values for the wave vedor k:
(6.232)
In a conducting medium the wave vector k turns out to be a

complex quantity. Hereinafter it will be important that the con-
ductivity u is always positive. This can be deduced even from the
Maxwell Theory in a Relatiuistlc Form
'"
fact thal Joule heat generated within a unit of volume of a con-
ducting medium per unit of time is equal to Q = iconaE = a£2 >
>0
Let us assume the frequency w to be a positive quantity. Then
in Eq. (6.232) the imaginary part of the radicand is also positive.
Assuming that permeabilities e and ll are also positive, we find
that the solution k 1 is located in the first quadrant of the plane of
imaginary variables, i.e. the imaginary and real parts of the solu-
tion k1 are positive:
+
k, = k\ lkf (kl, k'{ > 0). (6.233)
The second solution k2 differs from the first one only by sign:
k2 = - k, = - kl- ik'{, (6.234)
and we can present both solutions by means of the single equation
k - ± k' ± lk", (6.235)
where k' and k" are positive quantities. The quantity ±k' is the
real part of the wave vector k while the quantity ±k" is its ima-
ginary part.
Substituting Eq. (6.235) in Eq. (6.231), we obtain
E=Eo~ lwo:. el(:t: l'o:-(l)tl, H =Hoe"' ~two:. ell:t: l'<=-llltl. (6.236)
In these formulae one should take either all upper signs in the ex-
ponent, or all lower signs (wherever there is a choice between''±"
and "+"). Thus we obtain the two solutions for the fields E, H:
one proportional to
(6.237)
and the other proportional to
e-w-=. e-1(-'o:+(l)t). (6.238)
At first glance it seems that one of them, E~J- (6.237), attenuates
exponentially as z increases (the factor e-tt "'), while the second
one, Eq. (6.238), grows exponentially (the factor ek..-=). In actual
fact, both solutions represent damping waves. To make sure that
this is so, let us consider, for example, Eq. (6.238). It can be
treated as a wave of the type
(6.239)
whose amplitude is equal to e"·-=, i.e. grows exponentially with z.
It should be borne in mind that such a representation is valid only
if the imaginary part k" of the wave vector k is small compared
to the real part k', i.e. the wave amplitude changes little at
distances of the order of its wavelength.
250 Spedal Theory of Relai{V{Iy
The pha::.e of the wave (6.239) is defined by the expression en 4
tering into the exponent:

cp=k'z+mt. (6.240)
This phase is constant at
z=- .p-t +Canst, (6.241)
I.e. constant phase planes move along the axis at the velocity
Vp~t=-m/k'. (6.242)
This velocity up 11 is referred to as a phase velocity of a wave anJ
determines the wave propagation direction. The minus sign in Eq.
(6.242) means that the wave (6.238) propagates in the negative
direction of the z axis. However, if we move in the negative direc 4
tion of the z axis together with the wave, its amplitude, being
proportional to the factor ek""', will attenuate exponentially. It can
be seen that the wave (6.237) propagates in the positive direction
of the z axis at the same (in magnitude) phase velocity. The ampli·
tude of that wave is proportional to the factor e- 11 ~z and, conse·
quently, also attenuates in the propagation direction.
Thus, in a conducting medium there are two waves of a given
frequency propagating in opposite directions and possessing phase
velocities of equal magnitude. Amplitudes of each of these waves
attenuate exponentially in the propagation direction.
It follows from the Maxwell equations that the amplitudes EfJ
and H 0 are interrelated. This relationship can be expressed, for
example, by the first equation (6.227). Taking this equation into
account, we can write down the expressions for the fields (6.231)
in the form
E= Eoe"~' A•z. el (± k'z-t~m,
H =-de; [kEo] e"~' A"z. ef(± k'z-(1).1).

(6.243)
Thus, in order to determine a free plane electromagnetic wave

In a conducting medium, one has to specify not only the charac-
teristics of a medium e, J.l., o, but also the field polarization, i.e. the
direction and amplitude of the electric field £ 0•
From the solution of the dispersion equation (6.230) (see Eq.
(6.232)) it is seen that the wave vector k is a complex quantity.
Since we consider a unidimensional case when the field depends
only on the coordinate z and time, we can nssume the wave
vector k to be directed along the z axis:
k=k'+ik". (6.244)
Maxwell Theory {n a Relaflv{stlc Form 251
where both vectors k' and k" are directed along the z axis. Then
according to the third relation (6.227) the vector £ 0 is perpendi-
cular to the z axis. Let this vector be directed along the x axis:
Eo~ (£0, 0, 0). (6.245)
Then
H 0 ~(0. H 0 , 0), (6.246)
Ho=-dw kEo. (6.247)
i.e. the magnetic field is directed along the y axis as it is seen

from Eq. (6.243).
For the sake of simplicity let us assume the quantity Eo to be
real. Then H 0 is a complex quantity since the expression for H 0
(6.247) includes the complex quantity k (6.232).
Let us represent the wave vector k in the form
k~k' +ik"~l k le", (6.248)
+
where 1k I= ,Y(k')2 (k")2, tan cp = k"/k'. Then the field equations
(6.243) can be rewritten as
Ex= Eoe-lt*z.el (/t'z-(1)11,
Hy = -dwt k IEoe-lt*z.el rlt'z-Cilt+q~l. (6.249)
Here we have used only the upper signs in Eq. (6.243), to make
things simpler.
It is seen from these equations that in a conducting medium the
~v;v~e o!n~l~ e~e~r~rc~~~ (;~~~:~c ~~le~ ~~~d~i:t~~ffyedisin a~::~~
k" = 0 and the phase displacement disappears.
The real physical fields E and H cannot be complex quantities
so that a physical meaning may be ascribed either to real or to
imaginary parts of Eq. (6.249). Taking, for example. the real parts
of these expressions, we get
Ex= Eoe-lt"zcos (k'z- eDt),
(6.250)
Hv=-d; I k IEoe-lt"zcos (k'z -(i)f <p). +
The imaginary parts of Eq. (6.249) also yield the equivalent solu-
tions:
Ex= Eoe-lt"z. sin (k'z- eDt),
(6.251)
Hv=-fc;l k IEoe--": sin (k'z -mt <p). +
In conclusion, let us analyse Eq. (6.232) for the wave vector k
in cases of low and high conductivity of a medium. Let us write
252 Special Theory of Relatlaity
down Eq. (6.232), having chosen the plus sign in it for the sake
of simplicity:
(6.252)
If the absolute value of the second term of the radicand is much

less than of the first one (low conductivity), the following approxi-
mate expression is valid:
k ~ m , ; ; ; +I -J"f T~ k' + lk". (6.253)
In this case
k' ~ m ,;;;, k" ~ -J"f T. (6.254)
It is seen from these formulae that in the case of low conductivity

the wave (6.250) attenuates e times over the distance L which is
inversely proportional to conductivity:
(6.255)
If e, Jl and (] do not depend on frequency, the quantity L has the

same magnitude for waves of all frequencies.
In the opposite extreme case of high conductivity we may ignore
the first term of the radicand of Eq. (6.252) as compared to the
second one. This yields
k ~(I + {) ,.j •~m , (6.256)
In this case the imaginary and real parts of the wave vector k
are equal in magnitude. The distance L over which the wave at-
tenuates e times is equal to
L=..!...~
It' 'V
2 .
·I O'(l)jl (6.257)
J~~s t6u~::~itra:~ rh~~r~edp:~n~ 5 e!e~t~~m':~~~ti~ep~~~!hfaf~~~ ~~os:

highly conducting body (metal) attenuates drastically so that the
field differs from zero only in a thin surface layer, the depth of
this layer being of the same order of magnitude as L. If a and ~
do not depend on frequency, the skin layer depth is inversely pro-
portional to the square root of the incident wave frequency.
In the foregoing reasonings we assumed an electromagnetic
wave frequency w to be an assigned quantity and determined the
wave vector from the dispersion equation (6.230). We could pro·
ceed otherwise by assigning a wavelength or its corresponding real
wave vector k = 2n/A, and by calculating a wave frequency w ex·
Maxwell Theory Cn a RelatrvCstlc Form 253
pressed via the wave vector k and the parameters of a medium

'· JL, a:
a ~ka aa
6>t2=-i-±
• e
---::r·
e11 e
(6.258)
We shall not analyse this expression at length; note, however,

that both solutions rou always have the negative imaginary part
in the case of positive e, IJ., and o, implying the wave attenuation
with time. Indeed, the time dependence of the field has the form
e-"M. The quantity ro is defined by the complex expression (6.258)
+
which can be written as ro = ro' iw" where ro' is the real part
of frequency and ro" its imaginary part. When ro" < 0 ont" can
easily see that the factor e-ui)t attenuates with time as e+liJ·t. Do
not forget that ro" is a negative quantity while time varies only in
a positive direction!
Let us consider now the equations describing potentials of an
electromagnetic field in a moving conducting medium. It is known
that in a stationary uniform isotropic conducting medium the con·
ductivity current j, 0 ,a and the electric field E are interrelated:
/,ond=aB.
When passing to four-dimensional designations, we must assume
the current components to form a four-dimensional vector and the
electric field to be expressed via the elements of a tensor of the
second rank F,k (see Eq. (6.29)). Therefore, in the relativistic in-
variant notation the quantity cr must be expressed via the elements
of a certain tensor of the third rank:
(6.259)
One can easily check that the tensor of the third rank amkf can be
expressed as follows:
(6.260)
Substituting this expression in Eq. (6.259), we obtain /cond = crE

In the frame where a medium is stationary (U, = U2 = U3 = 0,
U4 = ic). If a medium moves at the velocity V, we get from Eqs.
(6.259) and (6.260), and taking into account Eq. (6.29a):
j 1, cond =aUkF tk· (6.261)
In accordance with Eq. (6.81) this relation is equivalent to the fol·

lowing formulae:
f,o..,=ar{E+(VB]), p"..,-crr(~E).
254 Spedol Thenrq of Rrlalioltg
The fir~t oi these relations has a straightforward ph)SICal me.w-

ing: this is Ohm's law for a moving conductor. The factor at a 10
the first relation defines the electric field in a stationary medium
frame The second relation indicates that if a stationary conductor
carrying a current is electrically neutral, an electric charge ap-
pears on this C'Onductor when it moves at the velocity V (see § 6 I
for the physical interpretation of this phenomenon).
In the case of a moving conducting medium Eq. (6.60) should
be written down as follows:
a!='• = Csm + s•• cond>•
where s,.., co1111 e. im, cond is defined by Eqs. (6.259), (6.260) and
(6.261). Taking into account Eq. (6.261), the last equation is re-
written as
a!,.,
vr,
-aF.,U =s,.
11
Let us substitute into this equation the quantity f,., from Eq.
(6.200) in which the tensor &m11111 is defined from Eq. (6.201). Then
the last equation becomes the equation describing the components
of the tensor F,.,. If we now express F,." in the obtained equation
through the field potentials according to Eq. (6.199), we shall
obtain the equation for field potentials in a moving conducting
medium:
{..;.-..;(u•....2....)'
iJxk c ax,
-a~(u,....2....)}<Dm=
iJx,
=-J£c(6mk+ ;cc.;: Umu,)s,. (6.262)
Here the potentials (J),. satisfy the following additional stipulation:

~~: -7u,u, :~; -af£[J,<n =0. 11
This stipulation is generalized from Eq. (6.210) to include the case

of a conducting medium.
When the sources s11 are absent in a medium, we get the fol-
lowing system of uniform equations from Eq. (6.262):
{ ::~ -7(u• a:,y -af£(u, a:,)}<n,=O.

This system defines propagation of free electromagnetic waves in
a moving medium which in a stationary frame has a dielectric
permittivity e, magnetic permeability f£ and conductivity a.
Maxwell Theory {n a R.etafivlsflc Form 255
If a plane electromagnetic wave of the type (6.216) propagates

In such a medium, the correlation between the frequency w and
wave vector k of that wave takes the form
k2- ~- xr 2 (w- kV) 2 + i<Jilr (w- kV) = 0. (6.263)
This dispersion equation follows from the foregoing differential

equation if one takes into account that the gradient operator
V = iJ/iJr is equivalent to multiplication by -ik, when applied to
a plane wave of the type (6.216), and the operator of differential·
~~e~i!~/~~~d~c~fvitf;n~ ~a~ i~:3i~i~a ~~~~! 0in~oul!~~~~~~~n (~-~~;)
passes into Eq. (6.218).
Having assigned the frequency w and propagation direction of
an electromagnetic wave, we can define the magnitude of the wa\'e
vector k (and thereby the wavelength A.= 2n/k) from Eq. (6.263).
On the other hand, having assigned the magnitude and direction
of the wave vector k, we can determine the wave frequency w. If
one of these quantities, k or w, is assumed to be given, Eq. (6.263)
becomes a quadratic equation with respect to the other quantity.
This quadratic equation has complex coefficients, and therefore
its solutions are also complex. It follows from the appearance of
the wave (6.216) that if its frequency is complex, the wave is no
more monochromatic and it either grows or attenuates exponen·
tially with time. In this case an attenuation (or growth) index is
equal to the imaginary part of the frequency w. When the imag-
inary part w" of the frequency w is positive, the wave attenuates
with ttme, and when it is negative, the wave grows with time.
When in Eq. (6.263) the frequency w and propagation direction
of a wave, that is the angle e between V and k, are assigned, we
obtain a quadratic equation for the absolute value of the wave
vector k. The solution of this equation yields, generally speaking,
complex values of k. In this case two conjugate complex roots of
Eq. (6.263) correspond to the exponential growing or attenuation
of the wave in space. From here on we shall confine ourselves to
the case of low attenuation when the imaginary part of the solu-
tion of Eq. (6.263) for k can be considered small as compared to
the real part. In this particular case the sign of the imaginary part
of k does not define whether the wave grows or attenuates in space.
Indeed, let one of the solutions of Eq. (6.263), with ffi and e as·
signed, be equal to k' + ik", where k' is a real and k" an imagin-
ary part of k. The wave vector direction is defined by the unit
vector n so that
k=kn=(k'+iP.")n.
Let us direct the z axis of the Cartesian system of coordinates

along the vector n. Then the wave (6.216) is written down as
A= Aoetcwt-•r) = Aatt'-z, etcwt-.t'zl, (6.264)
When the attenuation is low (k"« k'), Eq. (6.264) can be con·
sidered to define a wave possessing the wave vector k' and fre-
quency w, with its amplitude varying according to the exponential
law e~<-z. Suppose k" is a positive quantity. To make any conclu-
sions concerning the wave behaviour, one needs to know the wave
propagation direction, i.e. the sign of its phase velocity. The phase
velocity of the wave is equal to the ratio w/k'. Indeed, the plane
of a constant phase of the wave (6.264) is defined by the relation
CJ>t-k'z =canst, whence :z =f,-1- co;st. It is seen from the
last relation that the plane of a constant phase travels at the ve-
locity w!k'. When w/k' > 0, the wave (6.264) propagates in the
positive direction of the z axis. Then if k" > 0, the wave grows,
and if k" < 0, it attenuates. And when w/k' < 0, the wave prop-
agates in the negative direction of the z axis. Then if k" > 0,
the wave attenuates in its propagation direction (although its
amplitude does grow in the positive direction of the z axis). Thus,
to determine whether the wave grows or attenuates, it is not suffi-
cient to know the law according to which the wave amplitude
varies in space; one has to know the wave propagation direction
as well.
There is a simple method making it possible to find out whether
the wave grows or attenuates in the direction of its propagation.
Let us analyse the expression wk"!k'. When it is positive, the wave
grows in the direction of its propagation; in the opposite case the
wave attenuates. It can be easily seen that the expression wk"/k'
is the product of the phase velocity of the wave by the decrement
of its attenuation in space.
Let us consider now the solution of the dispersion equation
(6.263). Let the magnitude of the wave vector be equal to k and
Its direction form the angle 9 with the velocity vector V of a me-
diUm. In this case kV = k V cos e. The dispersion equation (6.263)
is a quadratic equation with respect to the frequency w. Solving
it for the case of low conductivity o and discarding all degrees of
o exceeding the first, we get
w1, 2 =(1+xf2)- 1{[xf2kV cos 9 ± ck -0f +if aJ.I.r [ 1 =F 8..0r9 ]}.
(6.265)
nc-;
where x= 1 1 , B=~. r=(l-82)-'h, and 6.= 1 +xf2 (1-
- 8 2 cos2 9). This equation shows that if in the frame of a sta-
tionary medium efJ. > BQJ10, i.e. x = (n 2 - 1)/c2 > 0, the imaginary
part of the frequency (I) is always positive for both solutions,
whatever the velocity V of motion of a medium. This means that
for a given wave vector k the wave (6.216) always attenuates with
time. The attenuation decrement is proportional to the conduc·
tivity o.
Now let us consider the case when the given characteristics of
the wave (6.216) are the frequency w and the wave propagation
direction defined by the angle 9. Then from the dispersion equa-
tion (6.263) one can determine the magnitude of the wave vector k
corresponding to the given values of w and e. The solutions of this
equation have the form k 1, 2 =kl. 2 +ik'{,2.ln the case of small o
we obtain after minor transformations
ck(-w(l +xf 2) {xBr2 cose+ yt;}- 1,
2ck'(-- cu~r (I -II' cos' e) (1 + Bcose v&r': (6.266)
ck~= -(I) (xBr2 cos a+ .../'6:) (I - xB2r 2 cos2 9)- 1,
2ckf{ = CO'flr (1 + Bcos a ..J'F:) (I - xB 2r 2cos29)-l 6. -'1•,
Here the quantity 6. is always positive due lr> the assumptions
made earlier. Using Eq. (6.266) one can obt<.• n expressions for
wk'{, dkb. They have the following form:
*It{=_ lJ~c 1 f2(~ 1 ~B:;:,s~ 1e~~;0:c:~6~ ,.!A)

~k'l=- o~c 1 f(i+Bcose,JA)
~ 2(xl:lrlcos8+1/6),.!6 ·
It is immediately seen from this that if x)>O, the products wk~~ 2/kl. 2
are always negative, whatever the velocity of motion of a medium.
This means that in a moving conducting medium the wave
(6.216) always attenuates in the direction of its propagation.
The only characteristic property of a moving medium is that when
the velocity of motion of a medium satisfies the condition
I - xB 2J'2 cos 2 a= 0 or B = 1/( I +
x cos 2 9), the real and imag-
inary parts of the second solution change sign simultaneously.
The potentials obtained by solving Eqs. (6.211) and (6.262) can
be successfully used to solve other problems (see the bibliography
at the end of the book), most of which, however, lie outside the
scope of this book.
9-~
CHAPTER 7
OPTICAL PHENOMENA
AND THE SPECIAL THEORY
OF RELATIVITY
Light, being a special case of electromagnetic waves, is des-

cribed by the Maxwell theory. As we have seen in the foregoing
chapter, the Maxwell theory meets all requirements of the theory of
relativity, and therefore must accurately describe the properties of
such a typical relativistic object as Hght. But even in the theory of
relativity the propagation of light in vacuo holds a special position.
We have already pointed out that the velocity of light in vacuo is
the ultimate feasible velocity of signal transmission and an unat-
tainable velocity limit for objects possessing a finite rest mass.
Besides, at the basis of the STR lies the statement about the ve-
locity of light in vacuo being the same in all IFRs.
The Maxwell theory is a macroscopic theory. In this chapter it
is convenient to examine a microscopic approach and to some
extent even quantum mechanical metho.ds. We mean the introduc·
tion or photons here. In some respect, the introduction of quantum
concepts leads to a very descriptive picture. The utilization of the
theory of relativity becomes indispensable when we consider op·
tical phenomena associated with a relative motion of bodies (the
Doppler effect, aberration).
§ 7.1. Properties of plane light waves. The Maxwell theory
shO\\·s that in a uniform isotropic medium (e =canst, f.1 =canst)
whose conductivity a is equal to zero, the time·dependent field
vectors E and H (as well as D and B which are proportional to
them) satisfy the wave equations
oE~tJ.E-~a;~=O,
(7.1)
oH~tJ.H-~ a;~ =0.
This signifies that in a uniform non-conducting medium the
wayes can propagate, whose phase velocity v= 1/~ is defined
exclusively by the properties of a medium. One of the possible so-
lutions of Eqs. (7.1) yields the plane waves:
E=Eoeltwt-.-,, H=Hoeltwt-.tr), (7.2)
where ro is the circular frequency. Here the field vectors are.
Optical Phenomena and Special Theory of Relativity 259
as:.um~d to depend harmonically on time, and the wave vector to

be directed along a normal to the surface of equal phases (the
wave front). It follows from Eqs. (7.1) that the absolute value of
the wave vector k is equal to w/v provided waves propagate in a
medium and to w/c when they propagate in vacuo. Since n =
= "1/e)l/eo)lo, one can write in a general case
k=7ns, (7.3)
J~het~ee sc~~ea ~ti~a~~~t~r ~r~n~~\h~onp~~~= ~for~ega~~~e d~~~gio?~

equal to wt- kr and therefore the surface of equal phases is
defined by the equation wt- kr =canst. At a given moment of
time it represents the plane kr =canst, with the vector of its
normal directed along k (r is a conventional three-dimensional ra·
dius vector). In the course of time this plane translates in space
parallel to itself in accordance with the equation kr =canst+ (l)t.
The plane waves (7.2) must satisfy not only Eqs. (7.1) but al.so
the Maxwell equations (6.56) and (6.57) in the absence of charges
(p = 0) and currents (j = 0); substituting Eq. (7.2) into the
Maxwell equations we obtain the following results. In a plane
wave propagating in a uniform medium the vectors E, H and k
form a clockwise triad, i.e. they are mutually perpendicular and
the vector cross product of any pair of them, taken in the order
indicated, defines the direction of the third vector.
As to the relationship between the amplitudes, the following
equation is valid: ~H = ,..;eE.Consequently, in the case of va-
cuum, when B = IJ.oH and e =eo. we get E = cB.
The direction of the Poynting vector S coincides with that of the
vector k while its absolute value is equal to the product of th~
energy density in the plane wave and the wave propagation ve-
locity v, i.e. S = wv, or S = wv(k!k) where w is the energy den-
sity in the electromagnetic wave. This result has a clear physical
meaning: the Poynting vector determines an energy Hux across a
unit of area oriented normally to the incident wave per unit of
time. But the energy flo,ving across a unit of area per unit of time
is contained within the cylinder whose directrix is formed by the
contour of that unit of area and generatrix by straight lines par-
allel to the wave propagation direction. The cylinder's height
should be taken equal to v. In this case the quantity v defines the
volume of the cylinder thus formed. and the product vw, the elec-
tromagnetic field energy contained in the cylinder. All this results
inS = vw. Note also that in a plane wave
W= e£2~)l.H: =e£2,
..
while in vacuo w = e0£2•
260 Special Theory of Relaflt•lty
The momentum of a unit of volume (the momentum density) ,)j

an electromagnetic field in vacuo g is equal to S/c2. In the case of
a plane wave in vacuo when S =
cw, we get g = (w/c) (ll/k),
whence
g=w/c. (7.4)
The field momentum density in a medium will be examined in § 7
of this chapter. Recalling the invariants / 1 and / 2 of an electro-
magnetic field (§ 6.5). we find that in the case of a plane wave in
vacuo both invariants turn to zero. This means that in any frame
the \"ectors E and H of a plane wave are orthogonal, and the ratio
of their amplitudes is always the same. In the frame K' a plane
wave must take the form
E' = E[;et (!.1'1'-~<'r'l. (7.5)
The phase of the wave at the world point R(r, icf) cannot
c!epend on the choice of a reference frame. Therefore, the phase
u)t- kr must be an invariant of the Lorentz transformation. Con-
sequenUy,
(J)f- k,x- k 11 y- kzz= u/t'- k~x'- k~y'- k~z'. (7.6)
Substituting the transformation formulae for x', y', z' and t' from
Eq. (2.37) into the right-hand side of Eq. (7.6), we obtain
~-~-~-~-~0-~~-~~-~-~-~
This is an identity with ~espect to 1:.0 x, y, z. T~ing in!~ accou~t
that k=(i)/C and kx=csx, k 11 =cs11 • kz=csz (s 1s a umt
\ector whose direction coincides with that of k), we get
w=(j)'f(l+Bs~), ws,=(j)T(B+s~). ws 11 =(j)'s~. wsz=(J)'S~.
(7.7)
Here k'=~s'.
From these equations one can easily obtain the formulae describ-
ing the Doppler effect, that is the light wa\'elength variation when
emitted by a source moving relative to an observer, and an aber-
ration of light, that is the change in the direction of a light !Jeam
on transition from one inertial frame of reference to another. To
eliminate reiteration, howe\·er, we shall derive Eqs. (7.7) in a
somewhat different fashion and then, in the next section, investigate
their consequences.
We shall take the four-dimensional approach from the very
beginning. It has been already pointed out that the phase (j)t- kr
mu~t be an invariant of the Lorentz transformation. But this ex-

pression becomes an invariant automatically if it is represented as
a scalar product of 4-vectors (the scalar product invariance is
shown in Appendix I,§ 4). For this purpose it is sufficient to in-
troduce the 4-wave vector k ( 7)
k, i along with the 4-radius vec-
tor R (~; ict). Then, oot- kr=- kR.The introduction of the 4-wave
vector k is com·enient mainly because we immediately obtain the
rule for its component transformation on transition from one IFR
to another. A plane light wave propagating in the frame K' changes
its direction and observed frequency on transition to the frame K.
We shall see that an amplitude of a plane wave also varies. The
transformation equations for quantities characterizing a light wave
in the reference frames K and K' can be readily obtained if one
takes into consideration that in a plane light wave the co'!.ventionnl
wave vector k, together with i(w/c), forms the 4-vector k.
§ 7.2. A 4-wave vector. The Doppler effect. Aberration of light.
Let us consider a plane light wa~~e obser\'ed in the reference frame
K' and described by the 4-vector k'. The frame K' is chosen so that
the light beam propagates in it in the plane (x', y') at the angle 9'
to the x' axis. Write out the 4-vector components:
kl = ll cos 9' =~ cos9', k~=k' sin 9' =fsin9',
(7.8)
kJ=O, k4=i~=i~.
Now let us find the components of the 4-vector k in the frame K.
In accordance with the general equations (4.10a)
kt=f(kl-iDk~). k2=kl. k3=kJ. k4=r(k4+iBkD. (7.9)
Since k3 = 0, in the frame K the beam propagates in the plane
(x, y) as well. Consequently, the 4-vector k
has the components
k (7-cos9, 7-sin9, 0, i-7-)
in the frame K. From the last for-
mula (7.9) we find that
i-7= r ( i f + iB fcose').
or
(I)= (I)' ~~~D~o~:' =oo'r(l +Bcos9'). (7.10)
Consequently, if in the frame K' the light frequency is equal to (1) 1
,
it will be different in the frame K in accordance with Eq. (7.10)

(cf. Eqs. (7.7)). It follows from the first formula (7.9) that
7-cos9= r (~cos9'- iBif).
262 Special Theory of Relafivily
or, if Eq. (7.10} is taken into account,

cos9=~f(cos9'+B)= ,c.:;:·c~s~'. (7.11)
The second formula (7.9), together with Eq. (7.10). yields
sin9=-fsin9'= 1 ~6 ,sin9' f(l~"86:os&')' (7.12)

Using Eqs. (7.11) and (7.12), one can easily find the expression
for sin 9' in terms of the angle 9:
sin 9' = r (I ~~Jecos 6) . (7.12')
Note that Eq. (7.12') is immediately obtained from Eq. (7.12) by
substituting unprimed quantities for primed ones and vice versa
and by taking the velocity V with the opposite sign. The equations
obtained make it possible to interpret quantitatively the two optical
effects: the Doppler effect and aberration of light. The Doppler
effect which is observed for waves of any kind consists in the fad
that in the case of a relative motion of a source and an observer
(receiver) the frequency (of sound or light) determined by the
observer differs from that measured in the reference frame in which
the source is at rest.
Let a source be at rest in the frame K'. Then the instruments.
resting in that frame will determine the natural frequency w0 of
the light source (wo = w').
When determining the frequency w in the frame K, we need t()
convert the angle 9' into the angle 9. It follows from Eq. (7.11)
that
cos9' = lc~~~sB6'
\\hence I+ B cos 9' =(I- B2 )/(l- B cos 9), and, consequently.
Eq. (7.10) can be rewritten in the final form:
(7.13)
{hheisf:=~~eKe~~i~:i~~s~~~~ri~~n~a~~=ti~~Pfr~e~u:~~~t.wA3iff~~~~~v~~o~
the natural frequency w0 of the source. The observed frequency v).
depends not only on the relative velocity of the source and the ob-
server (B = V/c), but also on the angle 9 at which light comes.
to the observer.
In particular, if the radiation comes along the relative velocity
direction, we observe the so-called radial Doppler effect. If the-
frame K' is to the right of K, the source moves away from the ob-
server and light propagates in the direction opposite to the x axis.
(see Fig. 7.1a). Consequently, cos a= cos n = -1. Then from

Eq. (7.13) we obtain the frequency w and period T = 2n/w:
w=mo~. T=To~~~~·
An observer receiving light from a source moving away !rom
him finds a frequency to decrease.
/(
Observer 8 =$ J'olfrce
~~
Light V x,.:z:'
(a)
/(
So/irce 9a0 Observer
~·
loi
v
light 8-:Jr/Z
x,x'
Observer
(C)
Fig. 7.1. The radial Doppler effect: (a) an observer and a source mo\·e away
irom each other; (b) an observer and a source draw together, (c) the transverse
Doppler effect.
On the other hand, when the frame K' is to the lert of K (see
Fig. 7.lb), cos a= I and a source approaches an observer:
./T+B ./~
w=mo/\Ji'=1.f• T=To'Vl+li'
The frequency of light received by an observer increases as com-
pared to the natural frequency w0 . To an accuracy of 8 2 terms the
last two formulae can be rewritten as follows (the easiest way is
to multiply both the radicand numerator and denominator by the
numerator):
c.>~c.> 0 (!-B), "~"<>(!+B).
Both formulae can be combined:

,.. Special Theory of Relativity
Thus, the radial Doppler effect proves to be an effect of the first

order with respect to B. To an accuracy of the second order with
respect to B the formulae obtained coincide with the classical ant'S
following from the fundamental considerations (§ 3.3).
When light is observed in the direction which is perpendicular
to the source velocity direction, i.e. 6 = n/2 (see Fig. 7.lc), we
witness the so-called transverse Doppler effect. It is described by
the formula
ond depends on 6 2 this time. H a source motion velocity is non-

relativistic, the binomial expansion yields
"'~"'(1-ll'/2).
Since this is a second-order effect, its qbservation is much more

difficult to perform as compared to the radial effect. No wonder.
the transverse Doppler effect was observed as late as 1938 (lves)
when the relativistic formula was wholly confirmed •. We would
like to point out here once more that according to the classical
theory no radial Doppler effect should exist (cf. § 3.3). The radial
Doppler effect arises only due to the relativity of time intervals
between events.
Let us rewrite Eq. (7.13) in the form that was used in § 6.15.
We shall group the quantities pertaining to the frame K on the
right-hand side:
eoo= I'w(t - .f cos a) =I' ((I) -fVcosa) =I' ((1)- kV). (J.I4)
The natural frequency appears on the left-hand side while the

1 ight-hand side contains the frequency observed in the reference
frame moving at the velocity V, the light propagation direction be-
ing defined by the vector k.
Eqs. (7.11) and (7.12) coincide with the formulae derived di-
rectly from the velocity transformation formulae; therefore they
fully describe the phenomenon of aberration that was mentioned
in§ 3.6.
In particular, the expression for an aberration angle follows from
Eqs. (7.11) and (7.12):
tan9= 51 ~8~~ (7.15)
To conclude this section, let us derive the transformation for-

mula for a solid angle element written in spherical coordinates.
• The details of Ives's experiments can be found in the monograph by Lands-
berg G. S, Optics, 1976, "Nauka" Publishing House (in Russian).
Optkal Phenomena and Special Theory of Relat(vlty 265
We shall orient the polar axis in the direction of relative motion of

two frames (the x, x' axis). In the frame K' the solid angle elt~·
ment dQ' is written down as dQ' =sin 0' dO' dfJJ' = - d(cos O')dfJJ'·
Since the y and z coordinates do not change, the fJl coordinate, that
is the projection on a plane perpendicular to the motion direction,
does not change either: fJl = fJJ' and dcp = dcp'. From the formula
preceding Eq. (7.13) it follows that
d(cosO')=- (l~;c::a) 2 sin9d9,
whence the sought for transformation formula is obtained:
dQ' = 1'2 (! ~cosB)2 sin Od9dcp= fZ(t d~cosB)I, (7.16)
since in the frame K the solid angle element dQ is equal to

dQ=sinOdOdcp.
§ 7.3. A plane wave limited In space. The transformation of the
plane wave energy and amplitude. Let us calculate the components
-of an energy-momentum-tension tensor for the case of a plane
wave. We shall orient the x' axis along the wave propagation di·
rection, the y' axis along the vector E' aud the z' axis along the
vector B'. With such a choice of axes,E~=D~=B~=H~=E~ =
=D~=H~=B~=O.The tensor Tl 11 takes the simple form (see
Eqs. (6.128), (6.148) and (6.151))'
-w' 0 0 -lw'
r,.~
' (
ooo o)·
0 0 0 0
(7.17)
-iw' 0 0 w'
We shall also need the components of the tensor r;,. in the case
when a plane wave propagates in the (x', y') plane at the angle 9'
to the x' axis. Such a transition is accomplished through a simple
rotation of a coordinate system; the matrix of this coordinate trans-
formation takes the form
cos
_ _ ( sin 9'
6' - 6' 00)
sin
cos 9' 0 0
a 1, . -0 0 I 0 · (7.18)
0 0 0 I
Transforming the components of the tensor (7.17) through the use
of the ma-trix (7.18) according to the general rules of tensor trans·
..
, Special Theory of Relativity
formation, we reach the tensor fl~:
_,
- w' cos29' - w' sin 9' cos 9'
( - w' sin 9' case'
-iw'cose')
-iw' sinS'
Ta~= 0 0 .
- iw' cos9' - iw' sin 9' w'
(7.19)
Therefore, specifically,
w' =w', S~=cw' case', t~.=-w' cos 2e'. (7.20)
Let us prove the following theorem: a plane wave limited in
space along its propagation direction (such a wave is som~
times called "a train of waves") possesses a momentum and
an energy making up a 4-vedor similar to a 4-vector of energy-
momentum of a material particle. (This theorem is a particular
case of the more general theorem*.) To prove the theorem we have
to know the formula defining a change in volume occupied by a
train of waves on transition from one inertial frame to another.
The difficulty arising here is caused by the fact that the train of
waves moves at the velocity of light c so that the volume of the
train cannot be measured in the proper frame of reference. It is.
impossible to introduce the reference frame moving at the velocity
of light! However, one can bypass the introduction of the proper
volume, having finally accomplished the limit transition to the ve-
locity of light.
Let a certain volume move as a whole in the frame K' at the
velocity v', \Vith its value being equal to 1"'0 in the proper frame of
reference. Then according to Eq. (3.28)
(7.21)
If one considers this volume in the frame K, its velocity v will be

determined by Eq. (3.41), and, consequently, the magnitude r ,)f
this volume in the frame K will be equal to
r' 'Vv~w •
l+cBcose'
• The general lheorem is presenled in 113] § 57 and W Heiller's Tile Quan-

tum Theory of Radiation, Oxford, 1954, § 2 This general theorem can be formu-
lated as follows in the space region 'Where the tensor Ttk satisfies the condi-
tion a~k TJk = 0 and Ttk = 0 at the boundanes, the components Tu consli-
tule the components of a 4-vector.
Optical Phenomena and Special Theory of RelatirJUy 261
the second equality follows from Eq. {3.41). Here v' is the motion
Yelocity in the frame K'. Now, passing to the limit r/ _. c, we get
the required formula:
'1"=1"',~~,. (7.22)
Thus, if a certain volume in the frame K' is equal to 1"' we shall

-observe in the frame K. moving at the velocity V relative to K',
.a \'olume whose magnitude is determined from Eq. (7.22). It is
understood that there is a similar relation for volume differen-
tials:
d'f'=d'F' ~~~~ = f(l+~costl') d'f''. (7.23)

Let us go back to proving the theorem. Applying the general
tensor transformation formulae (A.L31) to the tensor (7.17), we
shall obtain the fourth line components of the matrix T,,. in the
following form:
T~ 1 =-if 2 w'(l +B) 2, T~ 2 =0,
(7.24)
T43 =0, T~ 4 =f 2 w'(l +B) 2,
.and for the tensor (7.19) the components
T41 = - i f 2w' {B + cos9') (I+ Bcos9'),
T42 = - ifw' sin 9' (I + B cos 9'), (7.25)
r4l=0, TH=f2w'(I+Bcos8')2.
Naturally, when 8' = 0, Eqs. (7.25) pass into Eqs. (7.24). Let us
-show now that the components (7.25) and, consequently, in the
specific case, (7.24) as well, transform vectorwise in the necessary
frame of reference when multiplied by the volume or the volume
element. Indeed, for example,
T4 1d'F= r(-iw' cos9'- ill:o')dr' = r (T; 1 dr' -iBT~4dr'),
~a-~d~ ~a-~r-~ u~
T~ 1 d'F= f(w' + lhl cos9) d'f'' = f(T:4d'f'' +iBTl4d'F').
Comparing the obtained formulae (7.26) with the vector tran5·
formation formulae, we conclude that the quantities T~ 1 • T~ 2 • T43 ,
T44 , i.e. the fourth line components of the tensor (7.17) or (7.19)
multiplied by the corresponding volume element make up a 4-vec-
tor.
Of course, this result holds after integrating over volume or
multiplying by a total volume, provided the tensor components T,~~:
<lo not depend on coordinates as it is the case in a plane wave.
Let us find the total energy of a train of waves in the frame K'
(see Eq. (7. 19)):
(7.27)
The total momentum components of a train of waves are definrc:l

by the following formulae:
G~=f ~ Tt~ dY' =f ~ (-iw' c~s6')d'J"' =~cos 6',
G11
,
=c1 ~-, , U' ,
Tz4d1" =csm6,
, ,
Gz=O.
(7.28)
Similar calculations can also be accomplished in the frame K.

Eqs. (7.25) allow the total energy transformation to be performed
directly: ·
U= ~ T44 d.Y = f 2 (1 + U cosO'f ~ I' ~:~:sB'J =

(I
- r (I + B cos 8') U'. (7.29)

Calculating also the total momentum components
0 11 =-f-) T1 4 d1"=f<- ir)(B+ cos6')) w' dr' =

= 7r (B + cos6')U' =-~case, (7.30)
we used Eq. (7.11) in the last equality; in much the same \vay,
using Eq. (7.12), we get
0 11 = f) T2~dr=+U' sin 6' =7 r (I+ U sin 6') U' sin 0=-fsinO.

Thus, in any inertial frame of reference we can introduce tllC':
4-vector
p; (~cos 6', ~sin 0', 0, i ~), (7.31}
+
with P' 2 = 0 in all reference fr~mes.
It follows from the condition P' 2 = 0 that the light wa\e m vacuo
cannot be at rest in any inertial frame of reference. Comparing
the components p,
(7.31) to the components -k,
(7.8), we see tha.t
the transformation formulae for U' and w' must be the same. This
implies that the ratio U'/w' must be invariant. Consequently, the
energy of the same train of waves turns out to be different when
measured by different observers. The ratio of energies is equal to
the ratio of frequencies of the monochromatic radiation which
forms the train of waves. The frequencies are determined by the
same observers who measure the energy. The train is supposed to
Optical Phenamena and Special Theary of Relativity 269
be long enough since otherwise it will not be even approximately

monochromatic.
From the formulae obtained the amplitude transformation law
is easily established in the case of a plane wave. Indeed, from Eqs.
(7.25) we get the following formula for energy density transfor-
mation:
w = w'l' 2 (I+ B cos 8')2•
Comparing this expression to the frequency transformation for·
mula (7.10)
U>=U>'f(l +Bcos8'),
we see that the energy density transforms as the square of fre-
quency. Since the energy density is a quadratic function of field
amplitudes of a plane wave, they transform according to the same
rule as frequency does.
To illustrate the treatment of an electromagnetic wave as a syc;.
tern whose momentum and energy form a 4-vector, let us consid-:!r
how the angular distribution of radiation from a dipole oscillator
is transformed on transition from the frame K' in which the osci!-
lator's centre of inertia is at rest to any other I FR. In the referenc~
frame in which the oscillator's centre of inertia is at rest and the
polar axis is directed along the oscillator's axis, the radiation in-
~~~(~~ :~~' :::!nco~s~ · ~i~~~~i~Q'. (e~'ut'P'lhe isra~7~~r~n tfnt~~si~iu~} ~
fun~.e/1~· ~·e~e\~~i:~e~~~n~~~i.atft~ i~raansl~~~at~~r~c~~o\~ ~=r e~~~t ~~
establish; in the case of radiation de = c dP, where dP is a mo-
mentum fraction escaping with the radiation in a given direction,
it being known that de'= c dP'. According to the Lorentz trans·
formation
de'~ f(de- V dP) ~r (de- V dPcosa)~ I' de(l- Bcos9),
dl'=+dt,
where dt' is the proper time. Having divided termwise the upper
equality by the lower one, we get
dl= :~: r~{l- 1 Hcos9) = 1'1(1 ~~~cosO)·

-M-=
Then we immediately obtain the sought for result:
dl = canst • ri ~~na 9 ~~!B) cons! • r• ~:na 9~ ~~s W
=canst·
(r-;r
-(--v---)
sln 2 9
dO. 5
1-- cos9
7
where we have used Eqs. (7.16) and (7.12'). It ir. seen irom the
obtained formula that the angular dependence of radiation in the
frame K, relative to which the oscillator moves, differs essentially
from the angular dependence in the frame K', especially in the case
whPn V ~ c. In this case the maximum radiation is observed in a
direction forming an acute
angle with the oscillator's axis
(Fig. 7.2).
It is worthwhile to consid~r
how the radiation which is iso·
tropic in the reference frame K'
behaves in the ~rame K. All the
---'*'~~-,-----,--x,z' ~~~:~st:Y~~~r~8u~:ed~~(a~~~i})a~
= const-dQ' and
dl= f 2 (1 d/~cos9)
f2 (I con~\os 6) dfY =
const · dQ
Fig. 7.2. The variation in the angular =r"(l Bcos9)2
distribution of radiation emitted by a
~~r:/:n~:c;~:~~: ~~ :~a~s~:~hn t~e0 ~s~~~~ (1-+.'-)'
Iaior is at rest (B = 0) to the re£er· =canst· ( v )~dQ.
ence £rame K rel-'llive to which it t- 7 cos9
moves (B = 1/ 2 ). The maximum radia-
lLon dLrect1on IS seen to be inclined to-
ward the oscillator's motion direction From the last formula the
"searchlight effect" in K can be
The axis o£ the oscillat01 IS oriented
in the oscillator's motion direction.
seen. The radiation concentrate<;
around the direction 0 = 0,
since the value of the denominator is the least at cos 0 ~ I, with
the ratio V/c fixed.
§ 7.4. The pressure exerted by an electromagnetic wave (light)
on a surface. The pressure exerted on a surface of a body, i.e. the
force acting on a unit of area, is defined by the momentum flowing
across a unit of area and optical properties of the surface. The
momentum flow is expressed via the spatial components of the
energy-momentum-tension tensor T~IJ which for a plane wave takes
~i~fi~n t~~ r!~ri~n ~fJh~n (~h1e7 )w~~e(Jp~~~~::reesnd~ ~~;ntht:\pr~~i~:
then, as Eq. (7.17) shows, the tension tensor has only one com-
ponent differing from zero, that is T~;f = - w'. To find the momen-
tum flow across the given surface element, one has to define the
direction of the normal to this surface n(n~)- Then (see Chap-
ter 6) the momentum flow across the element dS with the normal n
Optical PhenomefUJ and Special Theory of Relativity 271
is equal to (Fig. 7.3)
because only one term of the double summation differs from zero.
The magnitude of pressure acting on the unit of area normal to the
x' axis is equal to p' = n~ = T~., =I w' f. If a light pulse propa-
gates at the velocity c, the unit of area receives the energy per
unit of time equal to w'c = ~·
But we have seen that w' = p',
whence ........., \ n'(-1,0,0)
p'=~'/c. (7.32)
'\
lig~fni~e~~~naVyto t~~ep~e;:::; ~~ ~ -JS+-~m;;!L-f'--x'
an electromagnetic wave inci- Ohje::/
~f~~:; a~dud:~id~dbC: /;~o~~J~ _..~
ed that the wave is absorbed. Fig. 7.3. The calculation of pressure
Now let us determine the exerted by an electromagnetic wave on
force exerted on a wall by a a surface.
light wave incident on this wall
at a certain angle and reflected from it. Let the incident angle be
equal to 8. We shall denote the normal to the wall by n, and the
unit vectors directed along the propagation of the incident and
reflected waves by s and s' respectively. The momentum flow across
a unit of area will yield the pressure p whose components are as
follows:
where Tall and n~ are the tension tensor components of incident

and reflected waves.
The components of T ..,_~ for the wave propagating at the angle 0
to the x axis are given by Eq. (7.19). The three-dimensional \vave
vector of the reflected beam differs from that of the incident beam
by the substitution of -8 for 8. Let us introduce the reflection
coefficient R so that w• = Rw. Keeping in mind that 1 11 =
= T4 4 cos 2 8 = - w cos 2 8, and 1 12 = n4 sin 8 cos 8=- w sin 8 cos e.
we obtain the following expression for the normal force (pressure
ol light):
Px=(w +Rw)cos2 8=w(l R) cos2 8+
and for the tangential force:
p.,=w ( I - R) sin 8cose.

.,, Special Theory of Relativity
Let us write out the magnitudes of normal pressure p/l for the
two most interesting cases. In the case of the normal incidence
(0 = 0) the pressure is equal to 2w if the wave is reHected com-
pletely and to w if it is fully absorbed. In the case of isotropic
radiation Px should be averaged over all directions, i.e. the average
value of cos2 a should be taken. But the average value of the
square of the direction cosine of the spatially-isotropic unit vector
is equal to 1!J. Thus in the case of a total absorption of isotropic
radiation the pressure is defined by the formula p = w/3.
Surely, all these formulae can be obtained by means of ele-
mentary reasoning. Proceeding from the magnitude of electromag-
netic field momentum density g = S/c 2 and that of the Poynting
vector in a plane wave S = we, we obtain g = w/c (w is the
energy density). When a plane wave falls on the wall at the
angle a. a unit of area takes up all the energy and momentum
per unit of time which are C'onfined within an oblique cylinder
whose base is formed by the unit of area and whose generatrix is
numerically equal to the light propagation velocity. The volume of
such a cylinder is equal to c cos a. Therefore I m2 of the wall takes
up the energy lC =we cos a during I s while the momentum Gx
transmitted in the direction perpendicular to the wall during I s
is equal to Gx = g cos a.c cos a= w cos 2 a. But the momentum
transmitted to a unit of area per unit of time is just equal to the
pressure p. Introducing as before the reHection coefficient R. we
get p = w(l + R)cos 2 a.
§ 7.5. The light frequency variation on reHection from a moving
surface (mirror). Let in the frame K a light beam propagate at
the angle ao to the x axis in the (x, y) plane. A mirror located
parallel to the y axis moves relative to the reference frame K at
the velocity V. The light beam reaching the mirror is reHected from
it. We shall find the frequency and propagation direction of the
reflected beam in terms of the frame K.
It is convenient to introduce the reference frame K' fixed to the
mirror. Then the problem is solved as follows. In the frame K the
4·vector of the light beam is specified, i.e. the frequency and prop-
agation direction of light are known. The frequency of light and
the beam direction in the frame K' are easy to find using the Lo-
rentz transformation formulae. In the frame K' in which the mirror
is at rest the routine law of reflection is valid: an angle of incid-
ence is equal to an angle of reHection. This implies that the 4-vec-
tor of the reflected beam differs from that of the incident beam
only by the sign of the wave vector component along the x axis.
To obtain the 4·vector of the reHected beam in the frame K. one
has lo apply the Lorentz transformation once more.
Now, let in the frame K the light beam of the frequency w0
propagate at the angle ao to the x axis in the plane (x, y). The
Optlcal Phenomena and Special Theory of Retali1.1ily 27J
components of the 4-vector k0 in the frame K will be
k~=k 0 cos60 ~7cos60 , k~=O,

(7.33)
k~=k 0 sin60 ==7sin6a, k~=i'7=ik 0 •
Let us find the 4-vector k' of the same beam in the frame K'.
According to the Lorentz transformation (4.10a) we get
k; = r (k?+ iBk~). k~= k~. k;= kg, k~=l' (k~- iBkn. (7.34)
The component kl alters the sign on reflection from the m1rror
which is stationary in the frame K'. Therefore, the 4-vector k" of
the reflected beam will take the form
k;'=- r (k~ +iBk~). k; = k~. k; =k~, k~' = l'(k~- iDkn.
(7.35)
ln the frame K the reflected beam will ~e described by the 4-vector
k which is derived from the 4-vector k" via the inverse Lorentz
transformation from the frame K' to the frame K:
k,- r (k~ -iBk:)-- 1' 2 ((1 + B1)k1 + 2iBk:)-
--f'7{(I+B')cos60-2B), (7.36)
k 2 = k~ =~sin 90, k3 = k~ = 0, (7.37)

k,- r (k;' +iBk;')-f'{(I + B2)k:- 2iBk1)-
=ir27{(1 +B2)-2Bcos60}. (7.38)
Since k3 =kg= 0, the reflected beam keeps propagating within

the plane (x, y). Assuming that
k 1 =-fcos9, k2 =...2J-sin9,
(7.39)
k3=0, k4=l-T.
we get from Eq. (7.38)
~ (I +B1 )-2Bcos00
6lo I B3
(7.40)
Consequently, the frequency w of reflected light observed in the

frame K is not equal to the frequency w0 of incident light.
274 Special TMory of RelatifJity
In the frame K the tangent of reflection angle is obtained in the

following form:
tane =~=-(I ~"s~)!!saoB2) 28. (7.41)
It is seen from Eq. {7.41) that e '=P 80.Therefore, the angle of

incidence and the angle of re·
flection prove to_ b~ different in
·j -~'
the frame K (F1g. 7.4a).
It is worthwhile to write out
the formulae pertaining to the
normal incidence of light on the
mirror. Let in the frame K the
-+~r'
angle of incidence 90 = 0. Then
we get
(0.) / oo=roo!~=· cose=-1.

x' sine~ 0. (7.42)
v It follows from Eq. (7.42) that
reflected light also propagates
along the normal to the mirror.
r' although in the direction op-
posite to the initial one. The
frame K' in which the mirror
is at rest moved in the same
(b) direction in which light propa-
gated. The frequency of light
fig. 7.4. ReHection of light from a mov- decreased on reflection.
ing mirror (11) A mirror moves along If the mirror moves toward
the normal to its plane As a result of the beam, the quantity B alters
reHeclion, the frequency changes and its sign, and therefore we get
the angle of incidence is not equal to
the angle of reHection. (b) A mirror
Cil=~ !~~, cos9=-l,
~u0e~e;y par~~~:ln~0 ~~~ch~~~~d !~ere~~~: sin9=0.
tion; the angle of incidence is equal to
the angle of reneclion The frequency of light increases
on refleclion.
Making use of this effect, one can determine the velocity of a
moving object, e.g. an automobile. When an automobile moves
toward an observer, the frequency change on reflection is found
from Eq. (7.40) with an accuracy to within the terms of the order
o( B2 (j.c.1 = w-Cilo):
!: = I:SB :::::28. (7.43)

Optical Phenomena and Special Theory of Relatiuily 275
II, for example, the velocity of an automobile is 72 km/h = 20 m/s,

then B = 20/3· 1()8~0.6· t0-7• This relative change of frequency
can readily be detected by means of standard instruments.
We have considered the case when the mirror moves along its
normal. However, the mirror may move parallel to its own plane
(Fig. 7.4b). In this case we have to modify somewhat the formulae
used. Eqs. (7.33) and (7.34) remain unchanged; however, the sign
of k 2 is altered on re~ection, and therefore,
k;'=f(k?+iBkn. k;=-k~. k;=kg, k~'=r(k~-iBk?).
(7.44)
Since k3 =kg=O, the re~ected beam remains in the plane (x, y)
as before. Returning to the frame K, we get
k 1 ~f(k;'- iBk;) ~k1. (7.45)
k2=-k~. (7.46)
k3=0, (7.47)
k4 =I' (k: + iBk~') = ~- (7.48)
From Eq. (7.48) we immediately obtain that oo = oo 0, and while
writing tan e = k2fkt. we find that tan e = -tan eo, i.e. a = -Bo.
~~~=~t~?~~t t~e e~1~~t~o ~~~~sof~~fl~~~:e~0 lii~shetl,f at~~ {~~q~~~~~ ~i
incidence is equal to the angle of reflection (in the frame K).
In conclusion Jet us write out the formulae describing the rc-
~€ction from the mirror modng along the normal to its plane in
the non-relativistic approximation, that is in the case when the
velocity of the mirror is small: B = V;c « I. Ignoring all terms
of the order B2, \\'e get
(i)=c:J>o(l-2Bcose0),
+
cos e = -cos e0 28 sin 2 e0,
+
sin 0 =sin 90 28 sin 00 cos 90•
In the case of the normal incidence on the mirror (OJ= 0) the
beam is reflected in the direction opposite to the initial one whereas
the f~;equency varies according to the following law:
ID=mo(t-2f). (7.49)
if the mirror moves in the same direction as the light beam. If

the mirror moves toward light, then
(7.50)
Eqs. (7.49) and (7.50) permit of a straightrorward Interpreta-

tion. Reflected light can be imagmed as going from an imaginary
source positioned behind the mirror, with the velocity of this imag-
inary source being equal to 2V. Consequently, if the imaginary
source is replaced by the real one with the same natural frequency
w 0 , the frequency change defined by Eqs. (7.49) and (7.50) will
correspond to the Doppler effect for this source.
The considered instances of light reflected from a moving mirror
represent specific cases of the general problem involving electro-
magnetic phenomena arising at a moving interface di\'iding two
media.
§ 7.6. Light quanta (photons) as relativistic particles. Relat!-
vistic mechanics presented in Chapter 5 dealt with particles pos·
sessing a finite (differing from zero) rest mas;. This~is manifested,
in particular, by a 4·momentum of a particle P = m V being mean·
ingful only on the condition that m =I= 0. The particles whose re.;t
mass differs from zero are referred to as tardyons. All such pi!r·
ltcles cannot reach the velocity c through acceleration. This can ile
seen from the fact that the infinite energy and momentum are rc·
quired for these particles to gain the ultimate velocity (8 = mc2 y,
p = rnyv, but if v- c, the factor v--+ oo). The solutions of all con·
crete problems cited in Chapter 5 testify that v remains less than c
in all cases.
While investigating an electromagnetic field interacting with
microparticles, physicists inferred that in such an interaction a
microparticle, e.g. an electron, always gains a definite energy and
definite momentum from the electromagnetic field. (For the sake
of simplicity \\'e discuss here a monochromatic radiation, i.e. a
radiation of a given frequency {1).) For the first time the assump·
tion about an electromagnetic field imparting energy to an electron
by definite portions (quanta) was made by A. Einstein in the
framework of the photoeffect theory (1905). In order to explain the
scattering of high·energy v-quanta by electrons, an electromagnetic
field had to be assumed to transfer not only a definite energy to
electron but a definite momentum as well (the Compton effect, 1923).
These properties of an electromagnetic field interacting with ~n
electron can be graphically described in terms of an interaction of
"particles of light", possessing a definite energy and momentum,
with an electron. Of course, it would be extremely naive to imagine
an electromagnetic field consisting of some kind of particles re·
sembling billiard balls. This "particle-of-light" concept is perfectly
suitable for describing energy and momentum exchange between
a field and microparticles. This being borne in mind, the concept of
particles of light, called light quanta, or photons, cannot lead to a
misunderstanding.
\\'hal properties should we attribute to a photon in order to treat

it as a relativistic particle? One of the photon's properties, to \Vit,
the relationship between its energy and momentum, can be ob·
tatned from macroscopic electrodynamics. Indeed, the pressure P
of light falling on a wall from vacuum was given by the relation
(7.32)
P~iff/c, (7.51)
\Vhere IS denotes the energy taken up by a unit of area of the W311
per umt of time. Now let us imagine that photons fall on the \Vall.
(Below it will be seen that the most essential is the fact that the
energy and momentum are transmitted by quantified portions.) Let
u~ depict light as a plane wave so that all photons move in the
same direction. Suppose each photon carries the energy e and
momentum p. If a unit of area of the wall absorbs N photons
falling on it every unit of time, the wall gains the energy Ne and
momentum N p. But the momentum gained by a umt of area of the
wall per unit of time is just the pressure of light P*sothat P=Np
and IS= Ne. That is why from Eq. (7.51) follows the relatio.t
between the energy and momentum of a photon:
p~e/c. (7.52)
But in the case of relativistic particles the relationship between
the energy, momentum and veloc1ty of motion is established by the
expression p = (e/c 2 )v, whence it is clear that the relation (7.52)
is \'alid only if v =c. Thus, a photon can be interpreted as a
relativistic particle pro\•idcd it moves at the velocity c.
Just as}or any relativistic particle, the 4-vector of energy-mo·
mentum P(p, ie/c) can be constructed for a photon in vacuo.
Using the general formulae for the calculation of the square of t~e
4-vector and taking into account Eq. (7 .52t, we get 2 = 0. On P
the other hand, for conventional particles P 2 = - m 2c2 (see Eq.
(5.47)). This means that the rest mass of a photon is equal to
zero. In order to permit the (imaginary) particle to reach the ui-
timate relativistic velocity, we had to discard the finite rest mass.
The rest mass of a photon proved to be equal to zero, and ,n
first sight this fact seems to be rather regrettable. We have got
used to all bodies and particles in nature to possess a rest mass.
Until quite recently mass was regarded as an indispensable attri-
bute of matter taken for actually existing reality. Physicists were
also inclined to believe that a rest mass defines individual features
of every body or object. In classical mechanics any mass-posses:'-
by ·t~c~o;e~in!n t~.~~htfh;s f~r~e Facls.d~~~-o~ti~i~i~~e b~~s:~~:~ f~. t~i~r~~~~

= ( 1/s) (dp/dt); it is the momentum increment transmitted to a unit of area
per unit of time that appears on the r1ght-hand side here.
ing entity could be traced, at least theoretically, in the rourse of

time.
Until the beginning of this century light was thought of as a
mysterious phenomenon; even physicists doubted its material n.'l·
ture. But in 1901 P. N. Lebedev discovered the pressure of light
experimentally. The pressure is caused by the momentum flux.
Prior to this, there were no particular doubts that )ight carried
energy. But if light possesses both energy and momentum, its rna·
terial nature cannot be questioned. Although the rest mass of an
individual light quantum is equal to zero, there is nothing re·
prehensible about it. In nature there are objects whose rest mass is
finite, and also objects of zero rest mass. The latter move at the
velocity of light and cannot be stopped; throughout all referenc~
frames their velocity is the same. When brought to a standstill,
they terminate their existence, passing into other forms of matter.
It is the very fact that the forms of matter possessing a zero re;:;t
mass convert into those with a finite rest mass (and back), thdt
illustrates an equivalence of these forms.
The expressions for the photon's energy and momentum cannot
be obtained in terms of relativistic mechanics. Contemporary
physics, however, discovered that in the processes of emission and
absorption, as well as in interactions with matter, light behaves
as an assembly of quasiparticles, each of which possesses the
energy hw and momentum ti(l)/c. Here rt is Planck's constant, rt =
= 6.626-I0-34 J-s, and w is the circular frequency of light. Every
elementary act of interaction with matter involves one quasiparticle
of this kind, called a light quantum by Einstein in his time. Tl-te
energy and momentum conservation Jaws are valid in these in·
teractions. The expression e = hw can be taken for the light quan·
tum energy and p = l!f-s for its momentum; here s is a unit
vector directed along the light beam. Thus, if one treats a light
quantum (photon) as a relativistic particle, its 4·vector energy-
momentum takes the form P{ il7-),
fl.k, wh~re k = ks, k =
= 2n/A = w/c. Cancelling all components of P by the commoz
factor, '' hich is Planck's constant h, we obtain the same 4·vector k
again which earlier denoted the wa\'e \"ector. This time. however,
!
it is defined to fit a photon: (k, ik), k = w/c, P= ttk. Since the:
four-dimensional momentum of a photon coincides, with an ac·
curacv to within the factor h, with the four·dimcnsional wave vee·
tor inhoduced in § 7.2, all results obtained for a wave are fully
applicable to a photon. We mean here the formulae describing the
Doppler effect, aberration of light, change of light frequency on
reflection from a moving mirror. In terms of the photon theory of
light one can easily derive the formula describing the pressure of
light. Indeed, let a photon fall on a surface of a body at an angle a.
The normal component of its momentum is equal to hw cos ate
(Fig. 7.5). On absorption of the photon the wall acquires just this
momentum along the normal's direction. If the photon is reflected,
the magnitude of the transmitted momentum depends on the reftec·
tion coefficient, that is the photon reflection probability; let us de·
note it by R. Then the momentum component transmitted normally
to the wall on reflection of the photon is equal to ( I+R) (hw/c)cos a
where R : :; ; ; I. If n denotes the number of photons in I m2, all the
photons confined within an oblique cylinder whose generatrix is.
equal to c will fall on I m2 ar<:!a
of the wall per I s. The base of
this cylinder is equal to I m2 ami
its volume to c cos a. Consequent·
ly, nc cos a photons will fall on
I m2 area of the wall per I s.
Provided all these photons get ab-
sorbed, the wall acquires the ener·
gy hwnc cos a and the normal mo-
mentum component
Fig. 7.5. Calculation of pressure-
nc cos a (I + R>-!7- cos a= produced by light The area of the
oblique cylinder base is equal to
=hoon(l +R)cos 2 a. S=l.
The momentum transmitted to a unit of area of the wall per unit

of time is just the pressure on the wall; therefore, p =
= w (I + R) cos2 a where w = nhw is the energy density in the
beam. This result coincides with thf" one derived in § 7.4.
Unfortunately, today a "photon mass" is still defined by the for·
mula taken from relativistic mech~nics of particles with a finite
rest mass, namely, mp, = e/c2 (e = hw). First of all, the formula
e = mc 2 holds explicitly for the case m -=1= 0 (see Chapter 5) and
is quite irrelevant to the case m = 0. Besides, the mass mp, has
no physical sense at all. There is also no sense in speaking of the
inertial properties of a photon: in all reference frames it moves at
the velocity c, i.e. a photon in vacuo cannot be either accelerated
or decelerated (it can only be exterminated). In quantum statistics
photons are treated as identical particles, this assumption giving
correct re!>ults. However, if a "photon mass" had any meaning, the
"blue" photon would have been "heavier" than the "red" one, thus
violating the identity of particles. On the contrary, the only com-
mon property of photons is their zero rest mass. Sometimes the
photon mass mph is used to explain the deviation of a light beam
in a gravitational field. It is inconsistent. howe\"er, to treat sucb
'"a relativistic object as light in terms of classical mechanics..\Iced-
less to say, the relativistic theory predicts the deviation of a light
beam in a gravitational field without resorting to any photon mass.
Finally, there are some who wish to have the "mass,conservation
law" in relativistic physics. Since a rest mass is not additi\·e
(§ 5.6), the new "masses" are introduced on the basis of the rela-
tion m = !!/c 2• This is, however, quite meaningless to undertake
since the conservation of such a "mass" is just a consequence of
the energy conservation law which is always valid. To summarize,
we can state that the introduction of a photon mass does not bring
any advantages, complicating needlessly the subtle concept of mass
(see§§ 5.6 and 5.7).
Commg back to the zero rest mass of a photon, let us make some
more remarks. There is no real IFR in which a photon would be
at rest, so that the photon's rest mass is an unobserved quanlity.
Just as meaningless is to speak of time flowing in the reference
h arne fixed to a photon. The zero mass of a photon does not at all
signifies the absence of mass. For example, the temperature O'"C
does not mean that the body lacks internal energy. It should be
recalled here that in the STR there are the world lines of the zero
length which are not less meaningful than all other lines. Surely,
this is all due to the velocity of light being distinguished among
other velocities. Apart from photons in vacuo there are also the
"real" particles, neutrinos, moving at the velocity c as well. Thetr
rest mass is also equal to zero and cannot be observed in an ex-
periment. After all, the question whether the photon's mass is equal
to zero or not can be solved experimentally. There are methods
capable of detecting the photon's rest mass if it differs from zero.
As more and more experiments of this kind are conducted, th~
lower limit of the photon's "rest mass" slides gradually lower and
lower. By the end of 1975 this limit reached the value rn< J0-6 3 kg.
§ 7.7. Light quanta in a medium. The Vavilov-Cherenkov effect.
The anomalous Doppler effect. It is seen from the previous section
and§ 6.12 that the photon's momentum in a medium is determined
according to Eq. (6.183) when proceeding from the Minkowsld
tensor and according to Eq. (6.184) when proceeding from the
Abraham tensor. The photon's energy remains constant on transi-
tion from one medium to another, provided the oscillation fre-
quency does not vary. Then what expression for the momentum
should be used when the momentum conservation law is applied
to "light quanta in a medium"? There is no direct answer to this
question, and some considerations in this respect will be given at
the end of this section. And now we sha 11 show that if light quanta
(photons) in a medium are utilized in the form given by Eq.
(6.183), we can obtain useful results concerning radiation kine-
matics, i.e. conditions imposed on a frequency and direction of ra-
diat10n. These conditions are defined by the energy and momentum

conservation laws. We shall begin with the elementary derivation
of the conditions for the Vavilov-Cherenkov radiation.
In this case radiation is emitted by a particle which has no in-
ternal degrees of freedom. We shall write down the conservation
laws for an electron-radiation svstem. Of course, the conservation
laws per se do not answer the· question as to whether radiation
will occur. This question can be solved by calculation on the basis
of equations of electrodynamics. However, if the conservation laws
do not hold, the radiation is absent a fortiori.
Suppose a light quantum has been radiated. If the energy and
momentum of an electron before the radiation were ~ 0 , p 0 and b~
came 8 1, P1 after it, the energy and momentum conservation law'>
take the form
6.~=8o-8L=hoo, (7.53)
6.p =Po- P1 = l7-ns. (7.54)

Written in such a form, the conservation laws presuppose that th~
change in the energy and momentum of an electron is connected
only with radiation. We can easily find the required consequence of
Eqs. (7.53) and (7.54) recalling that according to Newton's law
lip= F-6.t; multiplying both sides of this relation by v and re-
calling that Fv f).t = !1($, we get
MJ' = v 6p. (7.55)
Naturally, this relation is valid only for small momentum changes.
Substituting Eqs. (7.53) and (7.54) into Eq. (7.55), we can
cancel out hw, then note that vs = v cos a where a is the angle
between the directions of the electron motion and the radiation
propagation. Then the final kinematic condition for the radiation
angle a will be written as follows:
cos a=-;;. (7.56)
This is the condition for the Vavilov-Cherenkov radiation. It is not

satisfied if an electron moves uniformly itt vacuo (tt = I), since
Ieos a1 ~ I, and the electron's velocity v is always less than c.
Consequently, an electron moving uniformly in vacuo does not
radiate.
We obtained this result directly from the principle of relati\·it)·:
a charge resting in a certain IFR does not radiate. This charge
moves uniformly and rectilinearly relative to any other IFR. Ra-
diation, however, either occurs in all IFRs or does not occur in
any of them. Consequently, a uniformly moving electron does not
radiate. This reasoning is not correct, however, for an electro'l
moving in a medium since a new characteristic velocity, that IS
2112 Special Tlreorg of Relatit~ify
the velocity oi an electron relative to a medium, appears here.

defining at the same time the "privileged" reference frame fixed
to that medium.
Note that in our approximation the final results do not contain
the quantity fl in spite of the quantum mechanical ideas used. The
result obtained is a classical one. The quantum mechanical ideas
were employed for purely methodical reasons. We make use of the
two conservation laws which do not require an obligatory utiliza·
tion of quantum mechanical concepts.
It is easy to obtain the radiation condition with the recoil mo-
mentum taken into account. Let us introduce the 4-vector of
energy-momentum (or briefly, 4-momentum) of a light quantum in
a med1um
(7.57)
The photon radiation by an electron must obey the conservation

law for the 4-vector of energy-momentum, or, in other words, the
energy and momentum conservation laws. L~ the 4-momentum of
an electron before the radiation be equal to p 0, after the radiation
;, while the 4-momentum of the light quantum to;, that is
Po(mYoVo, imy 0 c), P(myv, imyc), ~(-¥-ns, i~).
The conservation law for the 4-momentum has the form
Po=P+;.
or in components,
Po1-:rt1=P1·
Squaring the latter relation, we get
p~l - 2pOI:rtl + ni = p~,
v.here each term involves the summation over the index i. How-
ever, due to the invariance of the particle momentum square pg1 =
= Pi• and we get
(7.58)
Cancelling by n, is prohibited here, of course: the left-hand and
right-hand sides involve independent summation. We shall calcu-
late the left-hand and right-hand sides of Eq. (7.58) separately:
P0 t:rt 1 =7nmy0 (v 0s) -myoflw, :rt~= ( ~ )\n2 - 1).
Equating these expressions and taking into account that VoS =

= Vo COS a, Where a is the angle between emitted light and the
Optical Phenomena and Special Theory of Relativity 283"
ele~.:tron propagation direction, we obtain
-f (17-r (n2- I) = 17- nmyoVo C05 9 - myohw.

Hence
COS9= nm~oVo {~(n 2 -J)+ myoe }=
= n:o {I+ 2:' (n2-J) .yl -4}.

When considering absorption of a quantum instead of its radia-
tion, we should alter the sign in front of hw in the latter formula.
When ttw/mc 2 «I, which is true for visible light and an electron,
we get back to the classical condition for radiation (Eq. 7.56):
cos a=_!_= vP,..

nvo Vo
Radiation kinematics deals also with the problem of light chang-
ing its frequency and propagation direction on transition from one
IFR to cmother. We mean the Doppler effect and aberration here.
Surely, these problems are solved easier in terms of the STR. In
§ 7.2 we considered the propagation of light in vacuo. Here we
shall obtain the requisite formulae for a uniform and isotropic
medium whose refraction index is equal to n; the formuli:le to be
obtained will prove to be quite different from the case of vacuum.
In fact, all calculations differ only slightly from those performed
in § 7.2 and therefore we shall present them only in a brief form;
in return+ we shall discuss the obtained results in detail. From the
4-vector n we obtain the 4-vector of a photon in a medium, propor-
tional to the former vector:
k(7ns, if). ~=lik. (7.59)

Assume that in the reference frame K' light propagates in the
plane (x', g') at the angle 9' to the x' axis; a medium is at rest
in the frame K'. Then
k'(-fncos9', {-nsin9', 0, i-f). (7.60)
+
The components of the vector k in the frame K are to be found
from the same Eqs. (7.9) from which it is seen that the beam
r{'mains. as before, in the plane (x, y) of the frame K. Instead of
Eq. (7.10) we shall gel
w d= w'r (I + Bn cos 9'), (7.61)
... Special Theory of Relatiuily
and instead of Eqs. (7.11) and (7.12)
ncos9= ~;ss!'c~=· , (7.62)

sin&'
sin 9 r(l +HncosO') ·
(7.63)
Hence, the final formula for the Doppier effect takes the form (d.
Eq. (7.13))
(7.64)
as before, ro' = r((a)-kV) (d. Eq. (7.14)), and for the aberration
angle (cf. Eq. (7.15))
tanO =:c.. sin 0'. (7.65)
let a monochromatic source possessing the natural frequency w0

be at rest in the frame K', i.e. move uniformly at the velocity V
{a} {6)
Fig. 7.6. (a) The Vavilov-Cherenkov radiation and the kinemallc explanallon of
its generation The positions of a unHormly mo\·ing particle are shown for the
moments of time It and lt. In the time Interval t,-1 1 the wave front will take
the position shown by a dotted circumference line (b) The Cherenkov cone di·
vides the space around the source into the regions of the anomalous and nor·
mal Doppler eftect
relative to K. Then w' = wo. It is seen from Eq. (7.64) that for
n > I, i.e. for such common media as, for example, water and
glass, Bn cos 9 may exceed unity even at V < c, and therefore the
denominator may turn into zero or even become a negative quan-
tit). Since an altered frequency sign implies, at the most, only an
alteration of an oscillation phase:
cos(-col)=cosco/, sin (-col)=-sincol=sin(col+ot/2),
a frequency may be always regarded as a positive quantity. Conse-

quently, the formula for the Doppler effect in a medium can u~
finally written down as follows:
(I)= 1~~~~- (7.66)

First of all, note that from Eq. (7.66) it is seen that a medium
does not affect the transverse Doppler effect: in the case of 9=n/2
we gC't exactly the same Eq. (7.14) as for vacuum. Thus, we get
one more evidence that the transverse Doppler effect arises only
due to the relativity of time intervals between events.
When I - Bn cos a = 0, the denominator of Eq. (7.66) turns
into zero, this being the condition for the Cherenkov radiation.
A moving charged particle having no internal degrees of freedom
radiates within a cone around its propagation direction. In thr.
case of a neutral source the Cherenkov cone divides space into
two parts with respect to the observed Doppler effect. The condi-
tion I - Bn cos 9 > 0 is valid outside the Cherenkov cone
(Fig. 7.6) where we observe the normal Doppler effect for which
(doJ/dO) > 0, as it is always the case in vacuo. "Within" the Che-
renkov cone I - Bn cos a< 0 and (dw/d9) < 0; this corresponds
to the anomalous Doppler effect.
It is interesting to compare the formulae for aberration of light
in vacuo and in a medium. In the case of a normal incidence in
the frame K' the aberration angle " in the frame K is determined
from the following relation:
tana=~-;. (7.67)
The Ollly difference from the case of vacuum (cf. Eq. (7.15)) con-
sists in the refraction index n appearing in the denominator. Therl!
is no singularity at V = c/n.
And in conclusion we shall decide which expression for the
photon's momentum should be regarded as "correct". As it w.n.c;
pointed out in § 6.12, two different expressions for the photon's
momentum in a medium, pM.=fT-n and pA=-7;..
correspond to
different subdh·isions of the momentum density of an electromag-
netic field in a medium into "the momentum density of a field"
and "'the momentum density of a medium itself". Since considering
the Cherenkov effect we are interested in a complete momentum
being lost by an electron, and such a momentum is defined by the
expression gM, the employment of gM leads to the correct result"'.
• For more details see V. L. Ginzburg, UFN 110, 309 (1973); V. L. Ginz-
burg, V. A. Ugarov, UFN 118, 175 (1976).
CHAPTER 8
ON CERTAIN PARADOXES
OF THE SPECIAL THEORY
OF RELATIVITY
If one consults a dictionary, he learns that the wor:d "paradox"

has at least three meanings: an unusual judgement differing from
a generally accepted one by its originality; a surprising deduction
from certain assumptions; and a result which seems mcredible at
first glance but proves to be correct on more careful consideration.
Studymg the STR. we can easily find examples illustrating all
meanings of the word "paradox".
As to the "generally accepted opinion", it is the ideas of classical
~~~s~f~:s s~;~rfio~!tfy~dt~~~ ~'::s~~a~~~odn~~\~~~ t~ff.e~t~:s ~~~~~i~~1
ideas of space and time coincide to a great extent with those con-
cepts that we acquire in our schoolyears and in our everyday prac-
tical life. These customary ideas have long become generally ac-
cepted. and their employment rests, we believe, on the "common
sense". But eventually, the common sense is an accumulation of
our settled down convictions among which there may also he false
ones. These false convictions are being exposed as science devel-
ops. Very often it turns out that certain ideas are true only
approximately, and the field of their application proves to be
limited. That is exactly what happened to some concepts when
the STR appeared.
In everyday life and in classical mechanics, for example, we got
accustomed to time having absolute meaning. Of course, this is
substantiated fairly well. The theory of relativity showed that the
time moments of an event measured in different IFRs are different,
i.e. relative. Howe\·er, it is psychologically difficult to readjust
oneself to relativity of time, especially as this relativity manifests
itself only at relati\'istic velocities (which macroscopic bodies never
attain) and never in e\'eryday life. Relativity of time, relativity of
simultaneity and time intervals between events correlate with r~l
ativity of lengths of scales moving with respect to one another
From the viewpoint of the common sense which makes us believe
~n~~~o~~~:r~i~~·y:~~s1h~~n~~~s~~ts p:~~lo~~~d~~i~rr J~l;~i~~~ z~
time is just a modern interpretation of measurement results oo-
On Certain Paradoxes of Special Tlleory of Relaliuily 287
tamed for the time of an event. By the way, th1s representation

conforms very well to the treatment of time in terms of dialectical
materialism according to which time, as a form of existence of
el('rnally moving matter, may depend on the motion of matter.
The "paradoxical" results of kinematics in the STR are very
well known and quoted in popular books. Here we mean relativity
of scale lengths ("contraction" of length-.), relativity of simulta·
neity and distances between events, and relativity of time intervals
between events. All these conclusions deviating from the "gene·
rally accepted" classical results were discussed in detail in
§§ 3.1·3.3. In the next sections we shall consider the paradoxes
which are further removed from the STR fundamentals.
§ 8.1. Faster-than-light velocities. As we saw in § 3.4 the prin-
ciple of causality requires the signal transmission velocity, that
x,x'
Fig.. S.t. Two particles moving toward each other can "approach" at the velocity
exceeding t.
is the velocity oi transmitting an energy and a momentum, to be

finite. The motion of any particle whose rest mass differs from
zero is, in fact, a signal since such a particle (when moving)
carries along energy and momentum. Hence, it is clear that the
motion velocity of such particles cannot exceed the velocity of light
in vacuo. Kinematics of the theory of relativity shows that if in a
gh•en IFR the velocity of a particle v < c, then in any other IFR
K' its velocity v' < c (see § 3.5). Let us consider another useful
example in this connection.
Let two particles move toward each other in the frame K at
equal velocities (Fig. 8.1). The only condition that the STR im-
poses on the velocities of these particles is that they should be less
than c. What is the relative velocity of these particles in K? Let the
velocity of particle 1 (denoted by v1) be equal to v; then the velo.::-
ity v 2 of particle 2 is equal to -v. The relative velocity of the
particles v,,., = v 1- v 2 = v -(-v) = 2v. Whence it follows that
if v > c/2, then Vret > c. Can this velocity mean that a signal is
traHlling faster than light?
In the considered case Vres is the velocity at which the distance
between the particles decreases. This distance actually decreases
288 Special Theory of RelatirJity
at the velocity exceeding that of light. No "signal", however, can

be transmitted at such a velocity.
To determine the possible velocity of a signal transmission, we
shall do as follows. An observer located at particle 1 wants to
transmit a signal (information) by means of particle 2. At the
moment when the particles come alongside each other he delivers
an "information package" to particle 2. But the velocity at which
the information leaves particle I is not at all the velocity at which
the distance between the particles varies in K, but it is the veloc·
ity of particle 2 relative to particle J. In other words, the velocity
at which information (a signal) is transmitted is the path covered
by an information carrier per unit of time. The distance per se
cannot be an information carrier.
Therefore, in order to calculate the velocity of a signal trans-
mission, one has to calculate the velocity of particle I relative lo
particle 2 (or vice versa). For this purpose, let us fix a frame 1(0
to particle /. Assuming V = v, we obtain v] = 0 from the general
formula
, v-V
v = 1-vV/ci·
This is an obvious result, of course: K' coincides with particle /.

As to v~. it is calculated as follows:
, -v-v 2v 2vfc
Vl= I +v2jc2 = - I +v2jc2 = - I +v2fc2. c.
This is the relative velocity of the considered particles. The last
link of the equality is written down to demonstrate that v~ <c.
The demonstration is presented below for the general case.
Let in the frame K the velocities of particles flying toward each
other be equal to v 1 = a 1c, v2 = a 2c; the theory of relativity re-
quires only that conditions a 1 < I and a 2 < I be satisfied. The
case when Gtt < 0.5 and ct2 < 0.5 is not of much interest since
+
even in K there is no velocity exceeding c. Suppose a 1 a 2 > J.
Let us introduce the frame K' where vl=O; we have already
seen that in this case V = v1. Then
, v2-V -a2c-a1c a 1 +~
V 2 =1-v2Vfc 2 =~=-!+a1a2c.
Let us prove that ext +a2 < I +

a 1a 2. We shall rearrange this
inequality by transferring all of its terms to the right-hand side:
a 1a2- at - ct2- I > 0. Grouping the terms, we get
(a 1 - I)(a,- I)> 0,
which is correct since the condition imposed on a 1 and a 2 is satis-
fied. The case considered earlier corresponds to a 1 = a 2•
On Certain Paradoxes of Special Theory of RelatifJitg 289
Thus, in a given IFR a conventional particle (m + 0) cannol

be accelerated to the velocity c. Nor this velocity can be reached
as a result of a transition from one IFR to another. But is it still
possible to find in nature velocities exceeding the velocity of light?
The first example suggests itself immediately. Let us take a
solid, absolutely rigid rod (body) and push it. Its both ends w1ll
simultaneously start moving which means that a signal will be
transmitted instantly. However, here the initial assumption is
erroneous. There are no absolutely rigid bodies in nature. All bo-
dies are similar to coilsprings of different rigidity. The transmis-
sion of a momentum (impact or push) from one end of a body to
the other is accomplished in the form of motion of an elastic wave.
And the velocity of elastic waves in solids is far less than the
velocity of light. Thus, the STR stresses once more that absolutely
rigid bodies do not exist in nature. By the way, an instantaneous
change of a momentum requires an infinite force, even in the
framework of Newton's mechanics.
Whereas the STR explicitly limits the velocity of signal trans-
mission, no restrictions are imposed on the velocities which are
not associated with the signal transmission and therefore they can
exceed c. Usually a paradox crops up when a certain velocity ex-
ceeding c is found and claimed to be that of signal transmission.
In the final analysis, it can always be demonstrated that the con-
sidered velocity has no relation to the signal transmission. We
shall discuss a few examples now.
The straight line AB moves parallel to itself at the velocity V1
directed normally to AB, and the straight line CD also moves
parallel to itself at the velocity V2 directed normally to CD. The
angle between the straight lines is equal to e. What is the dis-
placement velocity of the intersection point M of these straight
lines (Fig. 8.2)?
Fig. 8.2. The intersection point of two moving stra1ght lines can move faster
than light.
The relative velocity of the point M along the straight line AB

due to the motion of the straight line CD is equal to u2 = V2/sin e.
The velocity of motion of the point M along the straight line CD
due to the motion of the straight line AB is equal to u 1 =
= V.!sin e. The geometrical summation of the velocities u1 and u2
10-97
290 Special Theory of Relatif.!ity
yields (fig. 8.2)

VM= si~B -vfv~+ V~-2VtV2 cos8.
This formula shows that if a-+ 0, the velocity V.-'t-+ lXI, i.e. it
can exceed c. This fact, however, by no means contradicts the
theory of relativity. In the first plac~. the intersection point of the
lines is not a material body. Secondly, this point cannot be utilized
to transmit a signal (information) because at any given moment
it is being formed by the new points of the two lines, i.e. the in·
tersection point cannot be "marked".
The case of an oblique incidence of a plane light wave on a
plane surface (Fig. 8.3) is of somewhat greater interest. Consider
the point of intersection of the
wave front with the plane x = 0
(the point A in Fig. 8.3). In the
course of time this point moves
to the right. It is easy to find
the velocity of its displacement:
having chosen the section BD
equal to c, we obtain AD=
= c/sin a. But AD is just the
path covered by the point A per
unit or time, i.e. the velocity of
the point A. Since sin a~ l,
Fig. 8.3. The point of contact of an In· this velocity can be easily made
ddent electromagnetic wave with a greater than c. To dramatize the)
plane surface can move at a faster- situation, let us imagine the
than-light velocity. plane x = 0 covered with a
luminescent paint. Then a lu-
minous point will run along the axis at a faster-than-light velocity.
Surely, a luminous point moving at the velocity u > c can be
realized simpler, so to speak, "manually". Arrange electric bulbs
along the x axis and switch them on one after another (indepen-
dently) from left to right with a given time lag. Naturally, you can
get a luminous point moving at any velocity. From this second
example it is seen that in this process no information can be trans-
mitted since each source radiates independently. But is it possible-
to attain faster-than-light velocities by means of a relatively slow
rotation of a solid body of a considerable radius? For example,
a disc of radius r = c rotating at the angular velocity co J::::::: t
would possess the linear velocity u J::::::: c and over at its rim. How-
ever, such a velocity cannot be reached due to relativistic properties
of the motion equation. As the linear velocity of some sections of
a body increases, the forces required to accelerate these sections
become greater and greater, and consequently the linear velocity
On Certain Paradoxes of Special Theory of Relatioify 291
ot the farthest sections of the body cannot exceed c in this case

either.
If it is impossible to rotate a solid body, then let us try to rota:.e
a light beam. Let us place a searchlight at the origin of coordi-
nates and start rotating it at the angular velocity n. Let us ctr·
cumscribe a stationary sphere of radius c around the origin. Then
the light spot will run along the surface of this sphere at the linear
yelocity
v=O:c.
This velocity can exceed the velocity of light. The example of such
a beam is provided by a rotating pulsar. The light spot of the Crab
Nebula pulsar runs along the Earth surface at the velocity equal
approximately to 1022 m/s. But as in the previous cases no signal
is transmitted at such a velocity. As a matter of fact, every point
Fig. 8.4. The light spot reflected from a rotating mirror can run along the re-
moved screen at a faster-than-light velocity.
of a screen (the Earth) receives a new portion of light energy

from a searchlight (pulsar), but not from a neighbouring point of
the screen. Therefore, it is impossible to transmit information from
one point of the screen to another.
In fact, the same idea can also be realized as follows. A light
beam from the source I falls on a mirror consisting of several
facets and rotating at the angular velocity w. Depending on the
velocity w and the distance to the screen, one can get the motion
of the light spot (the source's image) at a linear velocity exceed-
ing that of light. Let us fabricate a reflecting mirror in the shape
of an ellipsoid and place a rotating mirror in one of its focal
points (Fig. 8.4). Then the beam reflected from the mirror will
pass through the second focal point in accordance with the well-
known property of an elliptical surface. This second focal point
can accommodate an analysing receiver. The light spot running
along the mirror represents the image of the source regardless of
the velocity of the spot.
The phase velocity of electromagnetic waves in a medium can
also exceed the velocity c. It is defined by the formula v = c/n
where n is the refraction index. There are cases when the refrac-
JO•
292 Special Theory of Relatioilg
tion index n < I and, consequently, v >c. All such cases relate
to a medium and to certain frequencies of electromagnetic waves.
For example, many substances have the refraction index n < I in
the range of hard X-rays. The same is true for plasma. But there
is no contradiction with the STR here again. The fact is that a
!i ~~:~e~~~~=~=~\~~~~~~c!t~~~i~~ !~fio~~dr~fra~tio~~~d~~~O:~~~-n~~
on the frequency of light passing through it, a signal can be trans-
mitted by means of electromagnetic waves whose frequency spec-
trum is sufficiently narrow (a group of waves). The velocity of a
signal is the velocity of energy transmission by such a group; .'IS
a more d~tailed analysis (see [36]) shows, the velocity of energy
transmission is defined by the group velocity. But the group veloc-
ity always turns out to be Jess than c, with the exception of th~
=~~:~~0 ~~ I~i~hi~~~0;io~~~~~e:e~~rt~et~~n~;~~~f vael;~~~p ~0:1~~~~
and consequently the signal transmission velocity, loses its mean-
ing. Thus, with the aid of wave processes, a signal is actually al·
ways transmitted at the velocity less than c.
(a) (b) (C)
Fig. 8.5. A rectangular l0op with an elastic thread stretching a sphere along the
frame's dtagonal. (a) The pict.ure observed m the proper reference frame J<O;
(b) the same pictur~ ob~ervcd u: the lrame K; (c) when the sphere is replaced
by a dumb·bcll, it expericnc{'s the actton oi a couple in terms of the reference
frame K.
§ 8.2. The thread~and-lever par.1dox. Let a plane rectangular

loop ABCD be at rest in its prope: reference irame 1(0. An elastic
~~~~o~:fr1~he(F?g~P~-~:) _oll~hetheafrsa:;;e fr~~· ~~~o d~~~~~i:~o~f ~~~
thread is found from the triangle ABC. De!>ignating AB = ao and
BC = be, we have
On Certain Paradoxes of Speclat Theory of Retatlr.rlly 293
Smce in the frame 1(0 the elastic forces are directed along the
thread, we can also write that
(8.1)
where P1 denotes the force directed toward the apex C. Similar

relations are valid for P~ as well.
Now let us pass to the frame K relative to which the frame K0
moves at the velocity V. As usual, we assume that the x0 and x
axes coincide and that the y 0, y and Zo, z axes are respectively
parallel. According to the formulae (3.5) and (5.34) of length and
force transformation we get
a=ao, b=b0(1-B2)'\ (8.2)
Fl"'=F?"'' F~~~=fll~~(l- 82)''•. (8.3)
It is seen from this that Eq. (8.1) does not hold any more; in
the frame K the angle defining the thread direction and the anglc
defining the direction of the forces are by no means equal:
tana'-b/a-(b.'a,)(t-B')''·-r-'b.'a,-tana,,I', (8.4)
tan•"- Fo,/F,-(F1,/f'l,)/(l- D')'''-l'F1,f f'1,- I' tan a,. (8.5)
Although the sum of the forces remains equal to zero as before,
these forces in the frame K are directed at a certain angle to the
thread (Fig. 8.5b). At first glance this circumstance seems sur-
prising. Indeed, what happens, for example, if we cut the thread
at the section 2. In the frame K0 the acceleration must be parallel
to the- force direction at the initial moment, i.e. it is directed along
the thread (since this is an explicitly non-relativistic case, the con·
ventional law of Newton is quite applicable). It seems that in the
frame K the acceleration should be directed at a certain angle to
the thread since the direction of the thread and the direction of
the force F 1 do not coincide. These statements are clearly contra-
dictory, but the paradox is resolved simply: in relativistic dynamicc;
the acceleration does not, generally speaking, coincide with the
acting force direction and although the force is directed at an
angle to the direction of the thread, the acceleration is oriented
along the thread. The paradox itself provides a useful illustration
of the characteristics of the relativistic equation of dynamics.
Let us make sure that in both frames the acceleration of the
sphere is directed along the thread. It is convenient to write the
relativistic equation of motion in the form
m do/dt- v-• [P- (o/c') (Po)];
294 Spec/at Theory of Relativity
here m is the mass of the sphere, F the conventional three-di-

mensional force acting on the sphere, v the velocity of the sphere,
v~(I-~~-'1• where ~~vic.
In the frame 1(0 at the moment t = 0 when the thread 2 is cut
mdv0/dt=F!.
or, in components,
m dv~jdt = F~x• m dv~jdt = F~ 11 •
Having divided the first relation by the second one termwise, we

obtain the formula defining the direction of motion at the initial
moment:
du~jdv~ = Ff1 jF~ 11 = tana0 •
In accordance with Eq. (8.1) this acceleration direction coincid~s
with the direction of the thread, as it should be. Thus, in 1(0 the
forces and the acceleration are parallel, and the motion is directed
along the thread at the initial moment.
Now let us pass over to the frame K. In this frame the sphere
moves at the velocity coinciding with that of the frame K0 , i.e. V.
Therefore v = r and the acceleration components can be written
here as follows:
mdo,/dt ~ [F,- (V/c 2) F,.VVf ~ F,,/f', (8.6}
mdv,/dl ~ F,,ff; (8.7}
we have taken into account here that the velocity of the sphere-
coincides with the velocity of the frame K, i.e. is equal to V and
has the components ( V, 0, 0); F 1x and F 111 are the force compo-
nents in the frame K *. In order to find the acceleration directioil
inK, we shall divide Eq. (8.6) by (8.7) termwise:
dux/du 11 = (F~x/F 111 ){r 2 = r tanaJr 2 = tanaJr =tan a'. (8.8)
In the thtrd link of this chain of equations we utilized the formula
(8.5) and in the last link the formula (8.4). But we see from Eq.
(8.8) that the acceleration in K at the initial moment is also di-
rected along the threads, and therefore no paradox can arise.
Let us imagine, however, that instead of the sphere that was
implied to be a point, the threads stretch some solid object, e.g~
• II is eas~· to notice that Eqs (8.6) and (8.7) correspond to two exceptional
cases of the relativistic equation when the force and acceleration are parallel;
formerly the corresponding masses were referred to in this case as "transverse'~
;md "radwl" nJa5ses Then these Jathtr incongruous terms have been practically
discarded although they con,•ty the tensor character of the rtlationshlp belweeru
the force and acceleration in rtlativistic me1:hanics quite satisfactorily.
On Certain Paradoxes of Special Theory of Relatiuify 295
a dumb-bell. Then in the frame K the spheres of the dumb-bell

would be subjected to the couple (Fig. 8.5c) and the dumb-bell
would be shifted relative to the diagonal of the loop.
It is obvious, however, that in the proper frame the dumb-bell
.axis coincides with the diagonal. Certainly, we come across a
'r
paradox here. This paradox, however, represents a version of the
well-known paradox of the lever which we shall discuss now. Let
.a lever be at rest in the frame K0 (Fig. 8.6). It is in the state of
.equilibrium in spite of the fact that two forces, F~ and FZ. act on
it, each of which is directed along the respective coordinate axis.
fj'
-
'
r Fx-Fx0
X (0) {b)
:Fig. 8.6. The lever paradox. In the frame 1(0 the lever is in equilibrium and the
resultant Force moment is equal to zero. When the same lever is considered in
terms of the Irame K, the Force moment difrercnt From zero arises in accordance
with the Formulae lor length and force transformations. The STR provides a ve·
ry elegant explanation why the lever is at rest in terms of the frame K (see the
text).
The equilibrium is ensured by the equality of the force moments

.in K0 :
(8.9)
these moments are oriented in opposite directions.
The same lever can be considered in terms of the frame K rela·
tive to which the lever moves as a whole at the velocity V. Having
formed an expression for the moments of the forces F:r and Fu in
K. we note that they are no longer equal, and, consequently, a
resultant force moment acting on the lever must appear. In fact,
in accordance with Eqs. (5.34) and (3.5) we have
F"=F~. Fu=F~....ji=Tff,
~.~f'.v~. t,~t:.
The difference of the moments of the forces Fll and Fu produces
.a torque in K
L~F,l,-F,t.~~f,-(l-B')~f'.~B'~f'.~-B'FJ,, (8.10)
296 Special Theory of RelativUy
wherE> Eq. (8.9) is used. Thus, the paradox consists in the follow·
ing. although it is well known that the lever is motionless, in the
frame K the lever is subjected to a force moment and, conse-
quently, must be rotating. The witty solution of this paradox
belongs to Laue. We have got used to a force moment inducing a
rotation, or, in other words, causing the appearance of the moment
of momentum in the system. In the frame K the force moment in-
deed defines the velocity at which the moment of momentum grO\vs,
but this growth is not associated with the rotation of the lever.
Then where does the increment of the moment of momentum come
from? Let us consider the work performed by the forces Fx and Fy
in the frame K. In the frame K the lever moves, and the force Fx
performs the work -F ..Y per unit of time. The force Fy performs
no work since it is oriented normally to the lever velocity direc-
tion. Consequently, at the end of the lever, i.e. at the point of ap-
plication of the force F x, the work is performed, and the energy of
the lever at this point increases by the quantity -Fx V per unit of
time. This means that the lever mass at the point of application of
the force increases by -FxV!c2 per unit of time. Multiplying this
quantity by the lever velocity V, we obtain the increment of the
momentum -fxB2. Hence, the moment of momentum grows by
-Fxl 11 B2 per unit of time. This is precisely the resultant moment
cited in Eq. (8.10). So, this additional moment does not describe
the rotation but defines the velocity at which the system's moment
of momentum varies. This explanation has some weak points. In
the STR there is no absolutely rigid bodies, and we must make
allowance for the deformation of the lever. In the foregoing rea.:;-
onings the lever was tacitly assumt>d to keep its form. In the
fr0ame K0 0 we must consider the lever's arms bent by the forces
Fx and Fu·
Considering this lever, we come across still another paradoxical
result. Suppose, no forces act on the lever till the moment t = 0
when F~ and F~ are "switched on" simultaneously in the frame K0 •
At ewry moment of time the equilibrium in the frame K0 will be
m<'intained. In the frame K, however, the forces will not be switched
on simultaneously since there will be a time interval when the force
F 1 is already ac-ting and the force F 2 has not been "switched on''
yet. Consequently, a force moment arises. The following simple
example shows that the forces applied at different points of the
body ue indeed essential. (The paradoxes appear, of course, when
a solid body is considered) Let in the frame Ko a solid bodv of the
length JO be located along the x 0 axis. No fofces act on th-is bod}
until the moment t = 0 when the oppositely directed equal forcel>
are switched on at both sides. In the frame K0 the equilibrium is
permanently maintained while in the frame K there is a time in-
terval during which the forces are not balanced and consequently
On Certain Paradoxes of Special Theory of Relativity 291
the body should start moving. We shall leave this paradox for the
reader to analyse.
§ 8.3. The tachyons. This is the name for the particles whose
velocity exceeds that of light in vacuo. From the very beginning
we should notify the reader that we speak of hypothetical particles:
the experimental attempts to observe such particles have not suc-
ceeded so far. But the very idea of their existence seems para·
doxical: the finite velocity of signal transmission is fundamental
for the STR, the ultimate value being that of the velocity c. Surely,
velocity per se has no limitations whatever (see § 8.1), but the
signal transmission is the propagation of energy and momentum.
The motion of particles to which we have got used can positively
serve as a signal. Besides, the conventional particles possessing a
finite rest mass, with which we have made ourselves familiar, can-
not reach the velocity of light. From the relativistic equation of
motion for such particles it follows that the velocity of light can
be reached only after an infinitely long time (not to mention the
fact that an infinitely high energy would be needed in that case).
Thus, the question about the faster-than-light velocity of particles
in our conventional world no longer arises.
One may, however, assume the existence of a special group of
particles whose conversion into conventional particles and back is
impossible. These particles, possessing faster-than-light velocities
from the very beginning of their existence could have been gen-
erated in certain nuclear transformations. The assumption con-
cerning the generation of tachyons was evoked by the picture of
photon generation: at the very beginning photons possess the ve-
locity o[ light, and do not emerge "dynamically", as a result of an
acceleration of conventional particles.
As in § 3.5 it can be shown that if the velocity of a particle v
exceeds c in one IFR, this is true in any other IFR. Consequently,
the conventional particles (photons) and tachyons form separate
groups of particles; the transition from one group of particles to
another by means of acceleration is impossible; the transition from
one IFR to another leaves a particle in the same group to which
it belonged in the initial IFR.
Assuming the existence of such particles, let us consider the
kinematic consequences of such an assumption.
So, let us assume that the velocity of a tachyon v determined by
the conventional means exceeds c, i.e. ~ = (v/c) > I. Then for the
interva I between two events, the positions of a tachyon at two
points in space at two moments of time, we get, as usual
ds 2 = c2 dt 2 - dx 2 =c2 (1- ~2 )dt 2 •
Here we consider the motion along the x, x' axis. As distinct
from the conventional particles, for a tachyon ds 2 < 0, i.e. the in·
298 Spet/al Theory of Relativity
terval is space-like. We saw in § 3.4 that in that case the concepts

"later" and "earlier" for two events are no longer absolute. Con-
sequently, there are such reference frames in which a tachyon
moves in one direction, and there are others in which it moves in
the opposite direction. One can find the condition making the ve·
Jocity of a tachyon reverse its direction in a certain reference-
frame 1('. In the frarne K' we obtain for a tachyon
u~r(t.t -f,-r.x)~r(l-~)M.
In any JFR we assume B < I. The time intervals tJ.t' and I!J.t have-
opposite signs and this means that the sequence of events varies.
ct K ct' K'
{b)
fig. 8.7. (a) The motion of a tachyon considered in two IFRs. In the frame I(
a tachyon moves to the right, in K' to the left. The bold line represents the
world line of the tachyon. (b) The reversal of the time sequence of events for
a moving tachyon.
with time, provided I - PB < 0. From this the required condition

v > c2/V follows; it is clear that v > c. The differences in de·
scribing the motion of a tachyon in the frames K and Ware clearly
seen in Fig. 8.7a. In the frame K the simultaneity lines are parallel
to the x axis, and drawing them farther and farther from the posi-
tive ct axis, we mark the position of the tachyon more and more
to the right: the tachyon moves to the right. In the frame K' the
simultaneity lines are parallel to the x' axts. Drawing these lines
till they intersect the ct' axis farther and farther along the positive
direction of the ct' axis, we find the tachyon located more and
more to the left: the tachyon moves to the left.
The same result can be presented in a more dramatic manner
(Fig. 8.7b). Let in the frame K a tachycn that left the point 0
.arrive at the world point P. In the frame K, as it is seen in the

figure, a tachyon was "emitted" at the moment t = 0 ("earlier")
.and arrived at the point P at the moment t 11 i.e. "later". The same
figure features the spatial and temporal axes of the frame K' in
which the simultaneity lines are parallel to the x' axis. lt is seen
in the figure that the tachyon in the frame K' was earlier at the
point P (at the moment - t]), then moved to the point 0 to be
.absorbed at the moment t' = 0. In this way, we can obtain the
movement of the tachyon in the opposite <liredion in space only
d A
Fig. 8.8. The observatian of a luminous particle maving at a faster-than-light

velocity.
by a proper choice of the reference frame. As a result, in a certain

reference frame we can observe the absorption of a tachyon instead
of its emission.
At the same time, we shall mention a curious picture of a
~'luminous tachyon", that is a tachyon which radiates light.
Fig. 8.8 illustrates that an observer at rest in the frame K will
"sec' two such tachyons diverging in the opposite directions.
Now let us go back to the reversion of the sequence of events in
time and, in particular, to the "emission" and "absorption" ex-
change. At first glance, this situation contradicts the conventional
cause-and-effect relations. Indeed, suppose it is known that the
source of tachyons is located at the point 0. The source is the
~·cause" of a tachyon generation. The motion of the tachyon toward
P is the "effect" of the tachyon generation. The observation in the
frame K' shows, however, that the tachyon leaves the point P and
is absorbed at the point 0. No matter how strange it may seem, it
must be admitted that the observed sequence does not contradict
300 Speda( Theory of Relatlvlty
the cause-and-effect relations, provided we define clearly what we

mean by them. For example, one may argue as follows.
We shall take A for the cause and B for the effect, provided th11t
a repetition of the event A at the moments t 1, t2, ••. chosen at will
leads unfailingly to the occurrence of the event 8 at the moments
+ +
t 1 T, t2 T, .... The essential point here is the controlled rep-
etition of the event A and its correlation with the event B. In this
sense the cause-and-effect relations do
not depend on which event occurs
N
earlier and which event occurs later.
The sequence of events in time is not
', involved in the definition of a cause-
ff· ------~--. ...........
c!1 and-effect relationship and cannot be
used in order to differentiate between
"-.,A' the cause and effect.
In our example the event to be con-
trolled in the frame K' is the absorp·
tion of a tachyon. This controlled ab-
I II sorption will always be preceded by
Fig. 8.9. The closed cause-and- the motion of the tachyon from the
effect cycle involving faster-than- point P toward 0. We shall have to
~~~h~~~~~alli~e}'~ist~o3~~f::e~~= for
take the absorption of the tachyon
the cause and its motion for the
frames. The first faster-than-light
signal AB is sent from the point effect. The cited definition of the cause
A; AA' is the simultaneity line and effect conHicts with the conven·
:e~ta~~er;!~~;~lifr~~ s\~~a~ 0frft g
tional statement that "the absolute
meaning of the notions 'earlier' and
of frame II which arrives at the
point D of frame I before the 'later' ... is a requisite condition for
first signal was sent (the point the concepts 'cause' and 'effect' to
A). The simultaneity lines and make sense". Of course, if the
the world lines of faster-than- "cause" and "effect" happen at one
light signals are drawn in accor-
dance with Fig 2 6b point in a given I FR. the cause must
precede the effect. But then the in-
terval between events is time-like a fortiori, and in any IFRs the
effect will happen "later" than the cause. The tachyons behave quite
differently. All "events" involving tachyons happen at different
points, when considered from our point of view. The reversal of the
sequence of events is of no significance.
Hence, the reversal of the sequence of events in time does not
contradict the conventional notions of the cause-and-effect relation-
ship. However, there is one condition that had to be satisfied on
all counts. It consists in the fact that it is impossible to exert in-
fluence on the past from the present. A signal sent from a given
point in space cannot get at it before it was sent.
If tachyons served as signals, it could be possible, as it is seen
from Fig. 8.9, to send a signal so that another signal caused !Jy
On Certain Paradoxes of Special Theory of Relatiolty 301
the former one will get at the initial point (the cause-and-effect
qcle) before the first signal was sent. Fig. 8.9 shows the world
lines of two bodies, I and //, which were at rest initially, then
moved uniformly and rectilinearly at equal velocities and finally
came to rest again. The world points A and A' are located at the
simultaneity line coinciding for both moving bodies. The world
points C and C' are located at the simultaneity line coinciding for
both resting bodies. The figure also illustrates the world lines of
two faster-than-light signals AB and CD. Having sent the signal
AB and then, on receiving it in another frame, the other signal CD,
we will receive the signal CD at the point D before the signal
from A was sent.
Thus, we have analysed the example of a closed cause-and-
effect cycle indicating explicitly the possibility of exerting influence
on the past. Certainly, this result pertains to any faster-than-light
signals, but when applied to tachyons, it implies that tachyons
themselves, unlike the conventional particles, cannot serve as
signals.
If one assumes that tachyons exist and the requirements of the
cause-and-effect cycle are satisfied, the resulting possibility of re-
versing the sequence of events in time for tachyons allows the
objectiOns concerning the "dynamic" properties of these particles
to be discarded. If one assumes the basic relations of the STR to
hold for tachyons, the transformation formulae for the velocity and
energy of a particle (see Chapters 3 and 5)
~'~ r~p~, l!''~fl!'(l-~B)
show that the tachyon's energy becomes negative in those ref-
erence frames where the sequence of events reverses its order and
where the sign of the velocity changes to the opposite since ~t
and !1t' are of opposite sign. The negative energy of a tachyon
is inadmissible since its existence would imply the possibility of
obtaining unlimited energy. In fact, the joint generation of a pair
of tachyons, one possessing the negative and the other the positive
energy, would not require any energy expense and the positi\·e
energy tachyon could perform useful work.
We have seen, however (see Fig. 8.8), that if in the reference
frame K a tachyon is emitted and absorbed, in the frame K' where
the tachyon's velocity obeys the condition v > c2/V the same pro-
cess can be described as an absorption of a tachyon moving in
the opposite direction and possessing the positive energy. This
circumstance makes it possible to avoid the difficulty associated
with the emergence of negative energies.
And, in conclusion, a few remarks concerning the momentum
and ene1gy of tachyons. In a unidimensional case (p"' = p) the
STR yields (see Chapter 5) the following equation.

~2-p2c2=m2c4, (8.11)
Plotting~= ~(p), we get a hyperbola; besides, as we hav.:!
seen in § 5.5,
u~dfff/dp. (8.12)
If the particle accelerates, it moves a1ong the hyperbola (8.11) in
the plane (~. p). The tangent slope is always less than c, irres-
pective of the manner in which the particle's energy increases:
whether via acceleration or due to the transition to another ref-
erence frame. Since the particle's energy is positive, the lower
branch of the hyperbola is disregarded. Note also that the hyper-
bola's asymptotes [!2- p 2c2 = 0 correspond to photons. H one as-
sumes tachyons to obey the basic formulae of relativistic me-
chanics (see Chapter 5), the quantities p = myv and ~ = mc2 y
become imaginary since v = ( I - ~2 )-•.r, and ~ > I, so that y = iy ..
,,. here 1/v. = -v'~ 2 - I. The real values of momentum and energy
can be obtained, provided we take the quantity im. for mass. Why
is an imaginary mass any better than an imaginary energy and
momentum? The point is, m. is the imaginary proper mass of a
tachyon, and there is no reference frame in which a tachyon could
be at rest (a reference frame consists of conventional particles and
its \·elocity never exceeds c). Therefore the tachyon's proper mass
cannot be observed and can be assumed to possess any magni-
tude.
bo~~t ~~~;e~~0~hdeiJ~a~~ ~~~ ~~;~~n~~;t Pc~nps~~e~~~~ ~0~ P~~f~
= -m~c 4 . Consequently, we must analyse three hyperbolas in the
plane (~. p) (Fig. 8.10). The slope of the tangent of these hyper-
bolas is more than c at any point. Of course, the factor y appears
not only in the expressions for a momentum and an energy, but
also in the definition of length via the proper length and the defi-
nition of time inten·a\s via the proper time. However, we can
readily reject the 'proper" quantities, assuming them unobservablr.
Ha,·ing referred the reader elsewhere for details, let us sum-
marize. In the past few years some attempts were made to analyse
the properties of faster-than-light particles in terms of the STR.
The STR says that the \"elocity that cannot be associated with the
real physical propagation of anything can have any magnitude.
Con\"entional particles always move at the velocity less than c,
that is any "signal" propagates at a velocity Jess than c. Conse-
quently, a tachyon cannot serve as a signal, i.e. its interaction
with our world is quite restricted. It might be possible that the
interaction of tachyons with our world is accomplished through
the exchange of electromagnetic signals.
On Certain Paradoxes of Special Theory of Relafivily 303
Proceeding from the motto "everything that is not forbidden has

the right to exist", we must admit the existence of tachyons. So
far, in theory their existence is not directly prohibited. Still it
seems highly unlikely that such particles really exist. The last
word is with an experiment.
,,
\,, Energy / .J
ll=#j / .l
\ NtJn-retativislitc"'
\ rarti11on.J JNlr/icles/ffi'
, ' ,Y'//
/
Momentum
Fig. 8.10. Tachyons and conventional particles depicted in the plane (8, p).
§ 8.4. The clock paradox. This paradox, provided there is a

paradox here at all, arises due to the difference in measurements
of time intervals between events in different IFRs, which was re·
peatedly discussed. We shall recall briefly the results that will be
needed later.
Let a body be at rest in the frame K' and two events be regis·
tered at the point x' at the moments J( and t2 by the clock mov·
ing together with that body and the frame K'. The interval t~- t\
is the proper-time interval to be denoted by .1.l". The same two
events will be registered by the observers of K at two points of
the frame K by two clocks at the moments t 1 and t2• The time in·
ten·al between the same two events will turn out to be equal to
D.t = t2- t1. We know that
"'' ~ -y'I- ~' M, (8.13)
i.e. the proper-time interval between events is less than the in·
terval between the same events registered by the clock of the
frame relative to which the body moves (cf. § 3.3).
Eq. (8.13) clearly shows an asymmetry in time readings. It
seems that we could argue as follows. Since all clocks in K are
304 Speclal Theory of Relatluily
S}nchronized, the time interval !it measured by different clocks of

K can be equated to the time interval registered by one clock of K.
Then it will turn out that the rates of the identical clocks in two
IFRs, K and K', are different. But the STR is based on the com-
plete symmetry of inertial frames! And this symmetry does exist!
We have missed an important point in our reasoning. Since the
simultaneity is relative, the clocks synchronized in one frame are
not synchronized in terms of another frame. The clock synchroni-
zation is relative! The quantity M is by no means the proper-time
interval for a clock of the frame K. Let us make the requisite cal-
culations.
Let clock Ill be at rest at the origin of the frame K' moving
relative to Kat the velocity V. Clock I synchronized in the frame K
rests at the point Xt =a and clock II synchronized in the same
frame rests at the point x2 = b.
The variable coordinate of clock JJI in the frame K is equal to
x 3 = Vt. Consequently, the coordinates of clocks I, //, //1 in the
frame K will be as follows:
x 1 =a (for clock I), (8.14)
x2 = b (for clock /I), (8.15)
x3 = Vt (for clock III), (8.16)
From the Lorentz transformation formula (2.11) X= r (x' + vn
we can obtain the dependence of the coordinate x' in the frame K'
~m lime t' in this frame and the coordinate x in K in the following
form:
x'=-Vt'+f·
In this way we shall find xl and x2; as to x3, it is obvious that
x3 = 0. Consequently,
x;--vt'+a!r (for clock/), (8.17)
x;--vt'+b!f (forelock//), (8.18)
X~=O (for clock I I I). (8.19)
As usual, we assume the readings of clocks from different frames
to be comparable, when the clocks are located at one point. ·1 nen
the following intercomparisons can be actually carried out. First,
the reading of clock III can be compared to that of clock I when
the clocks are passing each other; we shall denote the respective
readings of the clocks by tl and t 1; secondly, the readings of
clocks Iff and JJ can be compared when those clocks pass each
other; we shall denote these readings by 12 and f2 (Fig. 8.11).
When the reading of clock III coincides with that of clock I, both
On Certain Paradoxes of Special Theory of Relafioify 305
clocks are located at the point x' = 0; therefore, in accordance

with Eq. (8.17) the time moments tl and t2 will be given by the
followmg relations: tl =a/Vf and t2=b!Vf. At the same time
from Eq. (8.16) we obtain t1 = a/V, 12 = b/V, i.e. t 1 = rt~ and
12 = ft2. The readings 11 and 12 are the readings of different
clocks synchronized in the frame K.
Due to the synchronization in this frame clock 1 showed the mo-
ment l2 when clock JJ showed the same moment 12. The difference
l 2 - 11 is the time in the
frame K during which the K IK'
reading of clock Ill changed
by t2- t\. In terms of the ~v
frame K the rate of clock JJJ
is defined by the following
relation: -o/----------t'---
ll/:t'
t,
lll't'
t~-tl=+(t2-ti)= 1 I t2
-(1,-t,)-y'J-B', (8.20)
as it should be, since t2- ti
~in~:e,tf~~f~~~~~ ~~~r~~~
Fig. 8.11. The demonstration of the total
!~~m:.m; ~~s~~t ~~~rt~efr·~~~!~~:ar~~~·:
moving clock observed from In any reference frame the proper·time in·
the frame K is slow. We are terval b<'tween two events will turn out to
well aware of all this. Now be less than 1he 11m<! interval between the
we pass over to taking a de- same two e~ents registered by two clocks
cisive step: we need to com· of any other !FR.
pare the rates of clocks 1 and
JJ as observed from the frame K'. To pass judgement on the clock
rate, one should analyse the rate of one of the clocks, say clock 11.
However, we have only one direct reading of this clock: when
clock Jl was against clock 111, the latter showed 12 and the former
showed 12• The other reading of clock 11 has to be calculated (cf.
§ 2.4). We shall find where clock JJ was located and what it
showed at the moment clocks Ill and I were against each other.
Now Jet us analyse the situation in terms of the frame K'. When
clock Ill was against I, it showed the time tl =a/Vf. Clock 11
was at the distance x2- xl = (b- a)/r from clock I (see Eqs.
(8.17) and (8.18)). But when clock I was against Ill, their com-
mon coordinate was x\ = x3 = 0. Therefore, x2 = (b- a)/f is the
coordinate of clock 11 at the moment clocks 1 and Ill coincide. Now
it is easy to find the reading of clock 11 at the same moment of
time. We shall introduce x2 = (b- a)/f and t\ = a/Vf into the
formula
306 Sp.!cial Theory of Relativity
(Since the clocks are synchronized in the frame K'. tl ('Omcides

with the reading of the clock of K' located at the point x! so that
tl = t2.) As <1 result, we get the reading of clock 1/:
t=ft]+f* b~a =12+*(b-a). (8.21)
Were clocks I and II synchronized; they would show the same
time. But they are synchronized only in K and not in K'. We sec
that in terms of K' the clock of K is dissynchronized, and the fol~
lowing difference of readings appears
6=-;(b-a),
which grows as the clocks move away from each other. We already
obtained this result in § 2.4. Since in the frame K the distance
b- a= V(t2 - /J), the reading of clock II will be t = / 1 +.
+(V/c) 2 (t2 -tJ).Composing the difference of the marked time /2
and the calculated timet, we shall obtain
12-1=(12-11)--fr-.
or, in accordance with Eq. (8.20),
t2-t=(t~-~ny.
This means that the observer in the frame K' will note that the
clock moving relative to him is slow. Thus, the full equivalence of
the frames is proved.
This result confirms the full equivalence of the two inertial
frames considered: if in two IFRs there are two identical clocks,
the proper-time intervals registered by these clocks are equal.
Surely, it cannot be othenvise, since one of the basic principles of
the STR is the principle of relativity: if the rates of identical clocks
were different in two IFRs, such a physical method of distinguish-
ing these frames would be possible.
This is only the introductory explanation that had to be made.
The clock paradox is, of course, something different. Suppose we
compare the readings of two clocks: one from the frame K and the
other from K'. Naturally, immediately after the comparison the
clocks \Viii start differing from each other. Now the question arises:
if we bring somehow one of these clocks back to the point where
the other clock is located and intercompare their readings again,
what should we expect? It is the answer to this question that is
the clock paradox. This answer is far from being simple, and we
wish the reader to arm himself with patience.
First of all, we should point out that all formulae of the STR
imohe the quantities treated in terms of inertial frames of rc:-
ercnce. All time measurements carried out in the STR are made
by means of clocks resting in one or another IFR. Having once
compared two clocks, we are no more capable of getting them to-
gether at one point in space without taking them out of the rei-
erence frame in which they were at rest during the initial com·
parison. Indeed, if the motion is rectilinear, one of the clocks
should be decelerated and then accelerated in the opposite direction
to attain the same velocity. In that case the clock whose direction
of motion was reversed will
find itself at the same point
as the clock against which ct '(.x)
the initial comparison was
made. All this can be seen
very well in the Minkowski
diagram where the world
lines of two clocks, I and //,
are depicted (Fig. 8.12).
The "clock paradox" is
very convenient to analyse by
means of K calculus (§ 3.7).
We shall make use of the Fig. 8.12. The world lines of two clocks 1
space-time chart of Fig. 8.12. and II. The world line OD corresponds to
It illustrates the world lines clock 1 resting in K. Clock II first mo~·es
umformly from clock I (the line OT), then
of three clocks: clock I lo- having altered the velocity to the equal
cated at the origin of K (the and oppositely dir~led velocity at the
line OD), clock II resting at pomt T. approaches clock 1 again. AI the
the origin of K' (the line OT) point D they gel together and their read·
ings can be compar~d again (the first in·
and, finally, clock Ill rest- tercompanson was made at the point 0).
ing in K" (the line TD). L('t This comparison of clocks amounts to
.JS find the measured time in- what is called the clock paradox The m·
tervals directly. At the mo· set illustrates the world lme of one clock
ment t = t' = 0, when th~ gelling back to the point D
origins 0 and 0' coincide, the
initial exchange of light signals occurs which takes no time since
clocks from K and K' are positioned at one point. Clocks II and Ill
get together at the world point T; at this moment a light signal is
sent from the point T to clock I. Let clock /1 register the proper-
time interval .it 2 between its encounters with clocks I and Ill.
Then, as we know, clock I must register the time interval k 1.-r2
between the encounter of clocks I and II and the arrival of the
light signal from T. But the signal from Twas sent at the moment
when clocks II and Ill were against each other, and therefore if
clock Ill registers the proper-time interval 1.-r3 between the en-
counter of clocks II and Ill till its arrival at the point D. it is
possible to find the time interval between the reception of the light
signal by clock I at the point E and the encounter of clocks I and
Ill at the point D. We saw in § 3.7 that if the sign of the relative
velocity of the two reference frames changes to the opposite, the
coeffictent k changes to 1/k. Consequently, the time interval shown
in Fig. 8.12 by the section ED is equal to !i-r3fk. From the symmetry
of the imaginary experiment discussed here it is clear that 6.-r 2 =
=).,3. Designating the magnitude of this time interval by 6.-r, we
conclude that the time interval registered by clock I between its
encounters with clocks II and Ill is equal to
k (,h,)+ •;• ~ (k' +I)-¥-· (8.22)
The total time registered by the two observers (clocks II and

Ill) is equal to 26."t. Thts quantity is always less than the one
given by Eq. (8.22) since from the inequality (k -1) 2 > 0 tt
immediately follows that
k'+ I> 2k.
The obvious advantage of this approach lies in the fact that all
ttme measurements are performed by means of clocks resting in
inertial frames of reference. Thus, the time interval between events
appears shorter when measured by two inertial observers, as com-
pared to measurements made by one observer. Note that here, un-
like the case when the time interval measured by one clock is com-
pared to the time interval between the same events measured by
two clocks of another IFR, one compares the time intervals mea-
sured by clocks of three IFRs.
So. the use of two clocks, II and Ill, has led us to the conclu-
sion about different measurements of time intervals. Sometimes
they suggest to use the same clock in the reference frames K'
and K": at the point T clock II is just delivered to the frame K"
to make it possible to measure the time interval in question by the
same clock. This suggestion is worth dwelling upon. Although we
measure the time inten·at between the events 0 and D by means
of two clocks, I and II, these clocks are far from being equivalent
in the considered case. When clock II is delivered from K' to K",
it undergoes an acceleration and gets at a non-inertial frame. Its
world line transforms into a curve (see the inset in Fig. 8.12).
But the inertial motion is by no means equivalent to the non-iner-
tial one. It is quite possible that the clock which moved due to
inertia registers a longer time interval as compared to the clock
which participated in the non-inertial motion. There is no incon-
sistency here; the same conclusion is obtained from the Einstein
theory of gravitation.
We have already mentioned (see § 3.3) that any acceleration,
in principle, affects the clock rate Basically, the clock's rate is
"correct" in inertial frames of reference. Let the world line of a
particle be curved; th1s means that the particle undergoes an ac-

celeration. At any moment of the accelerated motion we can find an
inertial observer moving along a tangent to the actual motion path
and having the instantaneous velocity of that motion. The clock
moving with an acceleration has a "correct" rate provided it coin-
cides with the rate of the clock constructed identically, but moving
together with the inertial observer in the manner indicated.
At v:hat point of the world line does the difference between the
readings of "inertial" and "non-inertial" clocks come about? From
the principle of relativity it follows that the rates of the clock::>
constructed identically are the same in all IFRs. Whence it is clear
that the difference in the readings of two clocks brought to the
same point in space is caused by the clock acceleration, i.e. by the
LA B
Fig. 8.13. Path 1 between towns A and 8 is shorter than path fl although path
II differs from a straight line only at a short sechon. The difference m Length
is caused not so much by a curvilinear section as by the fact that path II is
not a straight line as a whole
curved portion of the world line. One often hears the objection that
the curved portion of the world line can be made as small a~
needed, i.e. an acceleration can be imparted for a very short time,
while the accumulated difference in time readings can be very
large. We should bear in mind, however, that an acceleration im-
parted during a short time interval in\'olves the immense forces,
and the reversal of a relativistic velocity direction is associated
with a considerable acceleration. Moreover, the difference in length
between a curved world line and a straight line connectmg the
same points is determined not by the length of the curved portion,
but by the overall curvature of the world line. This fad is very well
illustrated in Fig. 8.13: although path II from town A to town B
goes "along a straight line for almost all of the time", it is, no
doubt, longer than path I going from A to 8 along a straight line
connecting them. If an acceleration does not affect the clock's rate,
the length of the world line of a particle determines the proper-
time interval.
Until now we discussed time intervals registered by one or two
clocks. Going back to the initial problem, one may ask what
310 Spec/41 Theory of Relatitoity
clocks I and Ill show at the moment of their encounter at th~

point D. We remember that the sets of clocks in K, K' and K" arc
so synchronized that at the moment when the origins 0, 0' and 0"
of the frames coincide the three clocks from the three frames show
the reading t = t' = t" = 0. Now pay attention to the diagram
in Fig. 8.14. Here the world lines of clocks /, II and/// are sup-
plemented by the simultaneity lines _of the frames K' and K". The
transition from the frame K' to the frame K", that is to another set
,, of synchronized clocks, re-
sults in a jump of the simul-
D taneity line from AT to TB
(see the diagram). This tran-
sition explains a substantial
difference in readings of
clocks I and Ill. The substi-
tution of two living organism~
for two identical clocks leads
us to the so-called twins pa-
radox. The transition to living
~~io8 ·::~ T~:mt;af(~itl~~~~~~"!t 1~han~:~~ organisms,
a
however, evolves
series of complications so
the simultaneity line. From the line AT we
pass over to the tine TB. that we refer the reader else-
where [31].
§ 8.5. The "equivalence" of mass and energy. The zero rest
mass. In this section we shall go back to the problems that ha\'e
already been discussed; the main r<~ason for this repetition is not
the fact that new paradoxes will be disclosed, but that the oppor-
tunity arises to analyse jointly some results that were earlier pre-
sented separately. A few useful examples will also be given.
We know from § 5.6 that any physical system possessing th~
energy /! 0 in a proper reference frame (P 0 = 0) has <1 rest mass
M 0 = l! 0 /c 2 . In this relation /! 0 denotes all the energy contained
in the system. We shall illustrate this by two examples.
I. Consider a closed system consisting of n non-interacting ma-
terial points involved in elastic collisions (in classical physics this
model corresponds to ideal gas). Denote the rest masses of the
points by m611, m~2 1, ••• , m&n) and the four-dimensional velocities
in the proper frame K0 by ;~l). -;&21 , ••• , -;&n~. Let us pass over now
to another inertial frame of reference K whose relative velocitv
direction coincides with the X axis. In K the three-dimensional ve-
locity components of the points will be found by means of the fol-
lowing formulae:
11<1- u~~~o+V
vx - I+ Jx~loV •
,.
Let us also make use of Eq. (3.17):
-~c
-tcJ
)-
v·-(·;-·y~
~ r
I+* vlk)y
which we shall rewrite using the designations adopted in this

• (8.24)
book:
ylkl = fy~kl( 1 + v~~oV).
The resulting momentum of the system is defined as the sum of
the momenta of individual particles:
p= L m~k)y(klv(k) (Po= L ~kly~k)v~·) = o).
Therefore,
P.¥= Lm~k)yllllv~J= l:m~•lfy~"'(v~Jo+ V)=
= r L m~k)y~k)v~~~o + r -¥r L ~klc2-y~·~ =
=fP.r,o+ rfs-8 =f-!r- V.
0
It can be easily found that

P11 =P~I])=Q and P 2 =P~'=O.
Accordingly, the energy of the system is
8= Lmlklc2y~k)= L m~klc2fy~k{t + v")V)=

= r L ~•Jc2y~·~ = r8o•
since
L m~•lc2J'y~k) v.f~zoV = rv Lm~•ly~ll)v.r. o=O.
Consequently, in the case of a closed system
P=f~V, M 0 =~. (8.25)
where V is the motion velocity of the centre of inertia. This means

that the rest mass of the system M 0 is equal to 8/c 2• In terms of
the kinetic theory of matter the rest energy 8 0 of the system must
include the thermal energy as well. However, we have already
found that the rest mass of the system comprises not only the rest
masses of individual particles, but also their total kinetic energy.
that is the thermal energy when treated in macroscopic terms.
_•'-'------S-'-peclal Theory of Re_la_li"-".:.•-------
2. Consider an inelastic collision of two bodies. A system of two

bodies can be regarded closed and therefore the conservation law
for a 4-momentum can be applied to this process. Denote the rest
mass of a body formed after a collision by M 0 and the rest masses
of colliding bodies by m~ll and m~2 '. The energy-momentum conser-
vation law will be written in the four-dimensional form as fol-
lows:
(8.26)
where u, is the velocity of the single body formed after the col-
lision. The first three equations of (8.26) for i = I, 2, 3 permit th~
three velocity components of the single body to be found. As to the
fourth equation (i = 4), it is written as
mb''vu' + m~2Jy121 = M 0y.
In the reference frame where the newly formed body is at rest
m(ll m121
M,= -V~-(·~")' + -Vt-C'~')'.

This equality can also be written as follows:
Mo=?-[m~,tJc2(y(tJ _I) +m~2Jc2 (y121 _I)]+ m~IJ +m1,21. (8.27)
It is seen from Eq. (8.27) that the rest mass M 0 of the newly
formed system contains the sum of the rest masses of the initial
particles m~ll+ m~2 ' and a certain additional mass associated with
the fact that the relativistic kinetic energy of the two particles (the
expression in brackets) has been transformed into some other
kinds of energy (e.g. heat). Thus, in relativistic mechanics th~
energy conservation law includes all kinds of energy (and not only
those usually taken into account in mechanics).
Finally, it should be stressed once more that the relations ob·
tained indicate the proportionality of rest mass and rest energy;
it is far more important to remember that this fact is valid only
in a proper reference frame. Generally speaking, the rest energy
and rest mass possess different properties under the Lorentz trans·
formation when treated in four-dimensional terms (§ 5.7). Thus,
speaking of a conversion of "mass" into energy is meaningless
although sometimes one hears such a statement.
Now let us go back to a zero rest mass. Of course, from the
classical standpoint a zero rest mass seems rather strange. We
have seen (§ 7.6) that a zero rest mass should be attributed to the
particles moving at the velocity c. In accordance with the con·
On Certaln Paradoxes of Special Theory of Relativity 31~
temporary ideas the particles of this kind are light quanta (pho-
tons) and neutrinos. As we know, the velocity c holds a privileged
position in the STR since in all experimentally feasible IFR this
\'elocity retains its value. We could finish here, but we would like
to make some more remarks.
Obviously, there is no contradiction in regarding the matter (in
its philosophical meaning) possf'ssing a finite rest mass to be
equivalent to the matter possessing a zero rest mass. We shall see
that the latter case is realized comparatively rarely in nature, but
basically it is feasible. These two forms of matter just mentioned
can pass into one another. Now we shall dwell on one example of
such a conversion. This is the formation of electron-positron pairs
by gamma quanta (high energy photons) and the reverse reaction
of collision between an electron and a positron (this reaction is
known under somewhat obsolete name of "annihilation" of par-
ticles). This reaction brings to an end the existence of particles
possessing a finite rest mass (an electron and a positron), leading
to the appearance of two photons. What is essential, this reaction
satisfies the momentum and energy conservation laws. Just as
photons possessing a zero rest mass, so an electron and a positron
possessing a finite rest mass are characterized by definite momenta
and energies. The corresponding quantities resulting from thi-:;
reaction remain the same; a photon as an objective reality is
defined by its momentum and energy. The photon's rest mass which
is equal to zero characterizes a photon none the less than a finite
rest mass of an electron and a positron.
If the collision of an electron and a positron is considered in
the frame fixed to the centre of inertia, that is the frame in which
the particles mo\'e toward each other at equal, but oppositely di-
rected, velocities v 1 and v2, the energy conservation law takes the
form
moc2 + moc2 2hv. (8.28)
,jt-(!'f-)' ,jt-(!'f-)'
This equation shows that the total energy of an electron and a po-
sitron is equal to the energy of two photons formed. If we take into
account that in the frame of the centre of inertia v1 = v2, the ob-
served frequency of these photons will be equal to
(8.29)
From the momentum conservation law it follows that the ener-

gies of the photons formed are equal: the momenta of the photons
must be equal in magnitude (and oppositely directed), and the
photon's energy is proportion a I to its momentum.
314 Special Tlleory of Relatiulfy
When an electron and a positron move at non-relativistic ve-

locities, the frequency of photons resulting from the annihilation
of such an electron and a positron is equal to v = mc2/h which is
in a good agreement with experimental data.
The example presented here is far from being unique. We may
also mention the decay of a neutral (n:0 ) meson (possessing the
rest mass equal to about 200 rest masses of an electron): n:0 -+ 2y.
Let us consider now n photons of the same frequency moving in
various directions. The energy of this system of photons is equal
to the sum of the energies of individual photons: fC = e1 = nhv; L
the momentum of the system of photons P is equal to the sum of
thf' momenta of the photons:
P=L;p,=l!f-(sJ+s2+ ... +s"),

where s, is <1 unit vector oriented along the propagation direc·
tion of the ith photon. In accordnnce with the definition, the re"St
mass M of this set of photons can be found from the expression
M 2c 2 =~- p2= ( n:vr- (-¥- r (s 1 + s 2 + ... +s11)2. (8.30)
The right-hand side of Eq. (8.30) turns into zero only when all
the photons propagate in one direction. This result was obtained
in § 7.3· a limited train of plane waves has a zero rest mass.
However, two photons whose propagation directions form a certain
angle a possess a finite rest mass. Indeed, from the general for-
mula (8.30) we get
M 2c2 = ( 2 ~vr- ( l!:f- r(2 + 2cos0)= ( 2:v r (I- cos 2 {). (8.31)
Thus, a cloud of electromagnetic radiation consisting of photons

the rest mass of each of which is equal to zero possesses a positiv~
rest ma~s. and, accordingly, induces a gravitational field and ex-
periences a force of a gravitation<ll field.
Proceeding from the fact that even two photons possess a finit~
rest mass, we could try to avoid discussing the zero mass of 1'1
photon. But an individual photon can be observed in principle*,
find, therefore, a zero rest mass needs to be interpreted.
In order to clear up the cause that has led to the appearance
of a "zero rest mass" it is expedient to use the four-dimension?)
concepts. Let us consider the 4·momentum of a particle of a finite
rest mass m:
P{myv, f B). B=mc y. 2
• 0. Frisch, UFN 90, 379 (1966).

On Certain Paradoxes of Special Theory oi Relativity 315
The rest mass is the norm of the 4-vector P:

P 2 =~-p2 =m2c 2 , (8.32)
which is an invariant. In the 4-vector of energy-momentum the
energy is represented by a time component whereas the spatial
components are the compo-
nents of the three-dimension- !?adtalwn
~
al momentum. It is relevant
to recall that the basi:_ prop-
~t
erties of the 4-vector P coin-
cide with those of the 4-vec- HcZ The
tor V since P= m V. On the carria e
other hand, exper/?nce.
llTeCOit
Bel'ore theradtlltton AftertheaiMJrpfton

v =(* r-
2 (8.33)
p
Therefore, in the case of Fig. 8.15. A photon transfers mass al-

though its mass is equal to zero. Before
the world lines of zero length the radiation of a photon the energy of a
(ds = 0) the rest mass of carriage is equal to 8 0• In a closed sys-
the corresponding p!rticles tem the 4-momentum is retained, so that
the total 3-momentum of the carriage and
turns into zero since P 2 = 0. a photon is equal to zero as before and
Thus, photons move along
the lines of zero length. ~~eot~~a:s e;qe~:C t~f 8 :.hh~ar~!~se ~f1h~
There is still another ques- system remained constant although the
mass of the carriage decreased while the
tion which seems paradoxi- photon's mass is equal to zero (mass is
cal at first glance. How does not add1tLve!). When the photon is ab-
a photon, whose rest mass is sorbEd at the other end of the carriage,
the energy of the carriage becomes equal
equal to zero, transfer a finite to 8 o again, but th~ energy hv has already
rest mass from one point to been transferred from one end of the car-
another? An absorption of a riage to the other and the mass distribu-
photon proves that this is tion over the carriage differs from the
really the case. For example, initial one.
donating its energy to a
solid body, a photon warms that body up and thus increases its
rest mass.
Let us analyse a simple example. At one end of a carriage cap-
able of a frictionless motion a photon is emitted; then the photon
is absorbed at the other end of the carriage. Prior to the emission
of the photon the energy of the stationary carriage is /! 0 = Mc 2
(Fig. 8.15). Since the system is closed, its 4-momentum remains
constant and the sum of the 3-momentum of the carriage and the
photon is equal to zero as before. The cumulative energy of the
31< Special Tlu>org of RelatlrJilg
carriage and the photon is 8 0 . lhe mass oi the system remains

constant although the mass of the carriage decreased and the pho-
ton's mass is equal to zero. There is no reason to be frustrated
about: mass is not an additive quantity! When the photon is ab-
sorbed at the other end of the carriage, the energy of the carriage
is again 8 0, but by that time the energy hv had been transferred
from one end of the carriage to the other and the mass distribution
over the carriage had become different from the initial one.
Finally, we want to point out that the conclusions of the STR
make us define more accurately the concept of a "closed" system.
In mechanics a system is called closed if the constituent bodies do
not interact with. "external" bodies. An interacton is described by
means of forces. In chemistry they prefer to call a system closed
if it does not exchange any matter with the environment (then, ac-
cording to non-relativistic ideas, the mass remains constant). Pass-
ing over to thermal processes, we expect a closed system to be
heat-insulated. However, the STR declares any energy transfer to
be associated with a momentum transfer (this covers a heat trans-
fer as well): an energy transfer leads to a change of a system's
mass. These definitions can be combined into one, regarding a sys-
tem in whrch an energy and momentum (a 4-vector of energy-mo-
mentum) are retained as a closed one. An energy and a momentum
remain constant in a closed mechanical system. Such a system is
heat-insulated. In accordance with the conventional definition
Mc 2 = (8/c) 2 - P 2 a mass of the system remains constant. Of
course, the mass conservation law, when applied to a closed sys-
tem, does not imply the additivity of masses in the system. This
fact should be allowed for especially in the case of generation of
new particles.
SUPPLEMENT
1. Who developed the special theory of relativity, and how? *

(V. L. Ginzburg). The theory of relativity is one of the greatest
scientific discoveries of all times; moreover, it was made in our
century. The latter fact is especially significant in that the theory
is not so much a part of the history of science (or, if you wish,
not only a part of the history of science), but a physical theory
with direct and very extensive applications. That is the main reason
for the heightened interest in the story of how the theory of rela-
tivity evolved. It required a reappraisal of the fundamental con·
cepts of space and time, and thereby of the very foundations of
classical (pre-relativistic) physics. Old concepts die hard and new
positions are not easily won: the controversies and debates went on
for decades. Moreover, they involved representatives of other
sciences in addition to physicists. The theory of relativity has been
and remains the focus of intense scrutiny. This, of course, is also
true of its history - as an account or the evolution of ideas as well
as an issue of priority.
Thus it is that even today, seventy years after the enunciation
of the special theory of relativity, people are still asking, Who in
fact developed it, and how?
Special relativity is most frequently associated with the name of
Albert Einstein, with H. A. Lorentz, Henri Poincare, and a few
others mentioned as his predecessors. But there are other opinion-;
which, for example, name Lorentz, Poincare and Einstein as the
joint authors of STR. What view is more justified, and what, in
fact, is the argument all about? The answer to this question, as
well as to the one in the heading of this supplement, should be of
interest to the reader of this book. Below follow some remarks on
this score.
rel~~r:i1y~ 0;~! ~~~h~e;:o;tni~~~ at190t)ci~at~ :~: 6~f~~~ b~~1~~s~~
Hendnk Lorentz (1853-1928), one of the leading lights in theoret·
• This is a revised version of V. L. Ginzburg's article that appeared in the

final form in the 1974 Elnsleln CollectiOn (Moscow, 1976 in Russian),
219 Supplement
ical physics, w1nner of the 1902 Nobel Prize in physics. The author
of the second work ( 1906, a brief preview of which had been pub-
lished in 1905) was the celebrated French mathematician Henri
Poincare (1854-1912), also famous for his research in physics and
the methodology of science. Finally, the third work (1905) \Vas
written by a virtually unknown clerk of the Swiss Federal Patent
Office, Albert Einstein (1879-1955).
It is common knowledge that new works of popular and favourite
writers and poets immediately attract universal attention, whereas
novices hav-e to battle against stiff odds. In science this naturai
tendency is, if anything, more pronounced. How come that, in the
case of STR. it was the other way round, and it was Einstein's
work that gained acclaim, nay, renown? A clear answer to this
question was given by Wolfgang Pauli in his well-known article
''Theory of Relativity", first published in 1921 in the then presti-
gious Mathematical Encyclopedia. Pauli's article was subsequently
reprinted and translated into other languages (the Russian trans-
lt~tion appeared in 1947). Pauli concludes his account of the history
of the special theory of relativity with the words: "It was Einstein,
finally, who in a way completed the basic formulation of this new
discipline. His paper of 1905 was submitted at almost the same
time as Poincare's article and had been written without previous
knowledge of Lorentz's paper of 1904. It includes not only all the
essential results contained in the other two papers, but shows an
entirely novel, and much more profound, understanding of th~
whole problem" [8). Another eminent physicist, Max Born, recalls
his impression after reading Einstein's paper: "Although I was
quite familiar with the relativistic idea and the Lorentz transfor-
mations, Einstein's reasoning was a revelation to me."
It is in this entirely new and profound elucidation of the problem,
making it a relevation, that the success of Einstein's work is root-
ed, which is what made it fundamental to the enunciation of the
special theory of relativity.
A perusal of the history of science primarily focuses on two
questions. The first is, How? How did ideas appear and evolve,
how was a discovery prepared and made? The second question is,
Who? Who made the discovery, voiced the idea, turned it into
"Hesh and blood", elaborated it and drove it home to the scientific
community? The question How? would appear to be the basic,
primary one: it is connected with the very content of science and
the methods of scientific research. The question Who? may seem
secondary; indeed, it has no bearing on the essence of the matter,
if we take, say, physics and not the psychology of scientific creath--
ity, the sociology of the academic milieu, or the life of this or
that person. Actually, it is difficult, if not impossible, to draw the
line between How? and Who? Science is advanced by people, and
Supplement 319
H the end product- the totality of certain assertions, equations,

relationships, etc.- is depersonalized or, more precisely, almost
depersonalized, the initial process of the discovery of development
of the equations and relationships reHects the characteristic, most
typical traits of the discoverer. Thus, as far as the history of
science is concerned, the questions How? and Who? must be an·
swered simultaneously.
We shall preface further remarks on this score with a few words
about the special theory of relativity (this book, of course, covers
the topic in much greater detail, but it is useful to sum up the
situation here).
One of the fundamental physical concepts is that of inertial
frames of reference. A frame of reference used to define the coor·
dinates and time of events is inertial if the law of inertia holds in
_i{ namely that an isolated body (not subject to arfy fOrces) moves
Li]1torrntY-~~_rgfit Ttne.-To be sure, this definitioni'SnOf
immune from objections and must be clarified, insofar as it remains
unclear what body can be regarded as isolated? Broadly speaking,
a body can be considered isolated if all other bodies are sufficiently
far away. An example of a "good" inertial system is a coordinate
system with the origin at the centre of the Sun and the axes di·
reeled toward the remote stars. The inertia law holds with some·
\\hat less, but still sufficiently great, precision on the Earth (ne-
glecting gravity). A reference system rotating relative to an iner·
tial system is not inertial, the difference between the former and
the latter being the greater the higher the angular velocity.
If a given system is inertial, any other system moving uniformly
in a straight line relative to it is also inertial. The generalization
of this conclusion over all mechanical phenomena- the assertion
that all mechanical phenomena occur absolutely identically in all
inertial systems- is just what the classical. or Galilean, principle
of relativity is all about. More precisely, the definition and appli·
cation of the principle incorporates the quite definite prerelalivistic
assumption concerning the connection between the coordinates and
time of events in different inertial systems. Thus, if one such sy..,.
tem K' (coordinates x', y', z', and t') is moving relative to a given
inertial system K (coordinates x, y, z, and time t) with a velocity
V along the positive axes K, x' (the direction of which we assume
to coincide), then, as assumed before special relativity,
x'=K-VI, y'=y, z'=z, t'=t
(the Galilean transformations).
The absolute nature of time- its independence of the motion of
the reference system (whence the equality t' = t)- was, of course,
assumed to hold in all reference systems in general.
320 Supplement
In uniform motion a body's acceleration is, obviously, zero.

Hence, in the Galilean transformations, i.e. in any inertial system,
the acceleration is the same. Therefore, in these transformations,
the law of dynamics, Newton's second law (mass times accelera-
tion equals force), remains unchanged as long as the mass, force,
and acceleration remain the same in the K and K' systems. The
latter is assumed (and proved experimentally), and we come t•)
the conclusion that the classical principle of relativity holds in
Newtonian mechanics. Generally speaking, the invariance of the
equations expressing the fundamental physical laws in the Gali-
lean transformations is proof of the validity of the classical prin-
ciple of relativity.
Up till the end of the 19th century it was held that physics could
be constructed completely on the basis of the Newtonian equations
of motion. Thereby the classical principle of relativity was held to
be invariably valid. However, the development of electrodynamics
cast doubt on the classical principle of relativity. The equations of
electrodynamics (Maxwell's equations) do not retain their form
in the Galilean transformations, whence their application leads to
the conclusion that in electrodynamics the relativity principle
breaks down and, in particular, light and all other electromagnetic
waves propagate differently in different inertial systems, even in
vacuum. If the "luminiferous medium"- the ether- introduced
then is motionless in one inertial system (K), the velocity of light
in it is c = 3 X 108 m/s irrespective of the direction. In other iner-
tial systems K' moving with velocity V relative to the ether (along
the x and x' axes), the velocity of light is, as is obvious from the
Galilean transformations, c' = c- V along the x and x' axes and
+
c' = c V in the opposite direction, etc.
But experiments refuted that apparently obvious conclusion; all
experiments, starting with Michelson's famous experiment per-
formed in 1881 and repeated many times since, confirm the validity
of the relativity principle in electrodynamics as well as in physic'i
as a whole. But how, in accordance with the relativity principle, cau
the velocity of light be the same in different reference systems,
when the Galilean transformations lead to the opposite conclusion?
It took almost a quarter of a century of agonizing quest to arrive
at the solution constituting the core and basis of the STR: the
Galilean transformations had to be wrong. More precisely, as is
usual in such cases, they were not actually wrong, but approximatr.
The precise equations linking coordinates and time in the frames
K' and K have the form
xI X-
I= .
Vt
v2 ,
' '
y =y, z =z, t =~
' t -*X
r:--vz
V 1 -cr ~~-cr
Supplement 321
(the Lorentz transformations). If the relative velocity V of inertial

systems is small compared to the speed of light c, the Lorentz
transformations become the Galilean transformations; hence the
degree of accuracy given by the parameter V2/c 2• For a satellite in
orbit not far from the Earth V ~ 8 X 103 m/s, and V2Jc2 ~ ]Q-9.
The velocity of the Earth around the Sun is V ~ 3 X 104 m/s, and
V2fc2 ~ IQ- 8• It is obvious from these examples that in the domam
of the phenomena we encounter in everyday life the Galilean trans-
formations, and the Newtonian mechanics associated with them,
are valid to a high degree of accuracy. But in electrodynamics, and
in studying relativistic particles, travelling at velocities, v, com-
~a;t~~~~ ;;~h r~~~i~~~~~~~ ~F~~·e~' ~~r~~~~f~s tfse t~~r=~~~t~~~nsfor-
r + y2 + z2- c2t2 = x'2 + y''l + z'2- c2t'2.
Remembering that the equation of the front of a spherical light
wave has the form x2 + y 2 + x 2 - c2t 2 = 0, the above equation
immediately testifies to the validity of the relativity principle in
the propagation of light: in all inertial systems the speed of light
is the same and equal to c.
The special theory of relativity precisely represents the theoret-
ical constructions based on the principle of relativity and the
Lorentz transformations. The principal feature of the STR is the
G~~il!~~tiot·rt=~~~;~laf~~~ep~~ atsh;eH~~~~~tizn \~~~=r~:~~~~~~.of j~~
meaning of the latter, physically speaking, is not restricted to the
simple equations linking the coordinates and time x', y', z', t' with
x, y, z, t. As always in physics, it is necessary to establish the
meaning of all quantities, state the basis of the methods employed
to measure the coordinates and time, and clarify the properties of
the rulers and clocks used for this. One of the problems is that of
~~d~~:~e~i 7~~~ N:~e c:~fsa~~e~arc~n°fh~hL~;:~:str~n!?:r~iiJ~: ~~~
so defined that events simultaneous in the frame K (time t) arc
not simultaneous in the frame K' (time t'). The rejection of abso-
lute time is an especially radical conclusion (for which we are
indebted to Einstein). In importance and difficulty it can be com-
pared with the rejection of the idea that the Earth was stationary
on which Copernicus based his heliocentric system.
Now we can directly attack the question: Who developed the
special theory of relativity, and how?
The road to the STR lay, as is apparent from what has been
said, through a fundamental difficulty that had to be overcome:
the principle of relativity holds experimentally in electrodynamics
as well as in mechanics, but it is incompatible with the Galilean
transformations. To be sure, Lorentz and others sought to remove
11-91
322 Supplement
the contradicllou w1lhout rejecting the Galtl~an transformatiOns

by assuming that all bodies moving with respect to the ether con-
tract. If a ruler whose length at rest relative to the ether is La is of
length 10 ,YI- (V/c)2 when moving at velocity V, then we can
explain why some experiments do not reveal the motion of bodies
relative to the ether, and their results do not depend on the velocity
of the Earth's motion with respect to the Sun. However, the con-
traction hypothesis is not adequate for all experiments; new facts
kept coming to light which agreed with the relativity principle and
required additional hypotheses to explain them. This was, of course,
an intolerable situation, and Lorentz stubbornly strove to show
that many electromagnetic phenomena strictly, i.e. without neglect-
ing higher order terms, do not depend upon the motion of the sys-
tem. For this Lorentz had to show that for a body in uniform
tectilinear motion (relative to the ether) the equations of electro-
dynamics allow for solutions which in a certain way correspond
to the solutions for an identical body at rest. Correspondence :s
achieved by going over to new variables, x', y', z', and !', with the
help of the Lorentz transformations, as well as the introduction of
new (primed) electromagnetic field vectors. The field equations do
not change as a result of these transformations, and they have the
same form for the old (unprimed) and new (primed) quantities.
This property is known as invariance, in the present case in-
variance of the electromagnetic field equations with respect to the
Lorentz transformations.
Today, with special relativity, we know that this is precisely the
confirmation of the validity of the relativity principle in electrody-
namics, though Lorentz did not consider the time I' to be the time
in the moving reference frame; he called it local time and assumed
that he was dealing simply with supplementary quantities intro-
duced by means of a mathematical contri\'ance. In particular, the
\'ariable t' could not be called "time" in the same sense as the vari-
able /. In 1915 Lorentz reiterated the idea. He said that the main
reason for his failure had been that he had always held that only
the Yariable t could be taken as the true time, while his local time
t' should have been regarded as no more than supplementary
mathematical quantity. In Einst{'in's theory, on the other hand, t'
plays the same part as t. In 1927, a year before his death, Lorentz
stated this even more definitely, saying that for him there was only
one true time and that he regarded his time transformation as
merely a heuristic working hypothesis. Thus, the theory or relativ-
ity was actually the work of Einstein alone I may add that,
having reread the works or Lorentz and Poincare (70 years after
their publication), it was only with difficulty, and already knowing
the result (which, of course, greatly iacilitates understanding)
that I eeuld understand why the invariance of the electrodynamic
Supplement 323
equations with respect to the Lorentz transformations, proved in

these works, could at the time be regarded as proof of the validity
of the relativity principle.
Besides, Lorentz and Poincare saw the principle simply as an
assertion of the impossibility of observing uniform motion of a
body with respect to the ether. It requires no special effort to pro-
cc.ed from here to treating all inertial reference frames as com-
pletely equivalent (that is the contemporary formulation of th~
relativity principle) only if the Lorentz transformations are under-
stood as corresponding to going over to a moving frame of ref·
erence.
As we have seen, it was this that Lorentz definitely did not con·
sider. Poincare's stand is less clear. In his paper of 1906 he simply
asserts that the equations of electrodynamics "can be subjected
to a remarkable transformation discovered by Lorentz, the signif-
icance of which is that it explains why no experimental demon-
stration of the absolute motion of the universe is possible".* In my
view, this "explanation" goes no farther than Lorentz's. In general,
Poincare writes: "The results which l have obtained agree with
those of Lorentz in all the principal points, and I have needed only
to modify and augment them in certain details. These differences,
which are of but minor importance, will be shown in later sec·
tions." ** On the other hand, some of Poincare's remarks in earlier
works, papers and reports sound almost prophetic, notably regard-
ing the need to define the concept of simultaneity, the possibility
of using light signals for this, and his comments on the relativity
principle. However, he did not elaborate on this, and in his works
of 1905 and 1906 he followed Lorentz. As emphasized before, they
strove mainly to show, and showed, under what assumptions the
unHorm motion of bodies relative to the ether remained undetect-
able. But Einstein in his 1905 work reversed, one could say, lh~
whole issue by showing that, having accepted the relativity prin-
ciple and synchronized the clocks with the help of light (and also
postulating that the velocity of light does not depend on the motion
of the source), no additional hypotheses were required: the Lorentz
transformations follow directly from these assumptions. The con·
traction of moving rods and retardation of the rhythm or moving
clocks can also be postulated from them.
Thus, judging by published materials, Poincare was apparently
\ery close to enunciating the STR, but he failed to make the final
step. We can only surmise why. Perhaps it was because he was
primarily a mathematician, and it was therefore especially hard
• C W. K1lnuster, Spec1o.l Theory of Relo.IHJity, Qx[ord, Pergamon Press,

1970
•• Ibid.
II'
324 Supplement
for him to rise {or descend?) to a clear understanding of such

physically important aspects of the problem as adequate clarifica·
tion of the meaning of all introduced quantities and concepts.
Another similar hypothesis is that Poincare was prevented by his
predilection for convention, i.e. that school that emphasized (and
overrated) the role of conventional elements and definitions in
physics *. That convention plays a part in the development of
physical theories is indubitable. Length can be measured in
metres, feet or some other unusual or way-out units. The same is
true of time and other quantities, as well as of the definition of
simultaneity: there is no uniquely preordained definition. But the
end result, the content of a physical theory (as distinct from forms
of notation, etc.) is not a matter of convention, it is determined by
nature, a subject of investigation. Overestimation of the conven·
tional element in knowledge may prevent the clarification of con-
cepts. It could, in particular, explain why Poincare failed to clarify
the meaning of "true" time, t, and "local" time, t', which are in
fact equally true but are, if you care, "local" times for the frames K
and K', respectively.
• As far as I can judge, these comments comc1de w1th the view of Louis de
Broglie, expressed in an address on the occasion of the birth centenary of Poin·
care· "It needed but a little, and Henri Poincare rather than Albert Einstein
v.ould have been the first to enunc1ate the theory of relativity m all its general·
Hies, thereby giving French science the honour of the discovery .. However,
Poincare failed to make the decisiv~ step, leaving to Einstein the honour of per-
~e~·:~f~u~~ ~:a~}~;i~11~fi~he 0~:~!u~:~!~ti;yori;nc!~~e a~dd'ti~/~~\1~bl~~ ~~~0 ~e~~
physical nature of the connection between space and lime established by the
principle of relativity. Why did Poincare fail to pursue his conclusions to the
end? He doubtlessly possessed an extremely critical mind, perhaps because as
a scientist he was first and foremost a pure mathematician As mentioned before,
Poincare adopted a somewhat sceptical stance with regard to physical theories,
-holding that in grneral there existed an infinite numb<!r of logically equivalent
points of view and pictures of reality from which the scienllst, guided solely by
considerations of convenience. chose one Such nominalism probably prevented
h1m from concedmg that amidst all logically possible theories some were closer
to physical reality, or at least agreed better with the physicis\"s intuition and
\\ere therefore more useful That is why young Albert Emstein, who was only 25
at the lime and \\hose knowledge of mathematics \.\as m no way comparable
with the great French scientist's profound knowledge, was ablt'. beforl:! Poincare,
to find the synthesis \\hich at once removed all difficulties, using and Justifying
all the al\empts of his predecrssors. The coup de grdce was dealt by a r.1i~hty
intellect guided by a profound intuition of the nature of physical renlilv
"Yet Einstein's brilliant success should not let us forget that the problem of
relativity had been earlier and profoundly analysed by the viv1d mmd of Pam-
care. and that Poincare made a substantial contribution to the eventual solu-
tion of the problem Einstein would never have succeeded without Lorentz and
Poincare" (L De Broglie, YHenry Poincare, les theories de Ia physique." Le hv-
re du Centenaire de Ia Naissance de Henry Poincare 1854-1954, Paris, 1955)
I feel we must respect the point of view of de Broglie, whose attitude toward
the memory of Poincare was that of profound respect and maximum good-will.
Supplement_ _ _ _ _ _ _ _ _
,,
I must stress, however, that such hypothetical reasoning, in th1s

-case as applied to Poincare, is on the whole unjustified. There can
be no doubt that Poincare took an active part in the development
-of special relativity, and his contribution is indubitable. It is no
more legitimate to ask why he failed to do Einstein's work than it
is to ask the same question concerning other physicists of the time:
great works are great by virtue of the very fact that they are very
difficult.
In addition to what has been said of the part played by Ein·
stein's work, here is what he himself had to say, in a letter written
two months before he died: "Recalling the history of the elabora-
tion of the special theory of relativity, it can be stated conclusively
that by 1905 its discovery had been prepared. Lorentz already
knew that the transformation later named after him was of key
importance in analysing Maxwell's equations, and Poincare elab-
orated that idea. As for me, I knew only of Lorentz's important
work of 1895, but not Lorentz's later publication or the consecutive
investigations by Poincare. In this sense my work of 1905 was
indepeudent. The new element in it was the idea that the meaning
-of the Lorentz transformations went beyond the framework of Max-
·well's equations and involved the essence of space and time. Also
new was the conclusion that the 'Lorentz invariance' was a general
condition for all physical theories. That was especially important
to me becausE' I had already realized that Maxwell's theory d1d
not describe the microstructure of radiation, and therefore did not
.always hold."*
So, the reader wishing to receive a simple answer may ask,
after all is said and done, who developed the special theory of rel-
ativity? As in most such cases, special relativity is not a dis-
-covery or result attributable solely to one person. However, mo~t
physicists (myself included) unequivocally credit Einstein with
the principal role in elaborating it, for it was his work that con-
tained an "entirely novel, and much more profound, understanding
-of the whole problem" (8]; it was "the last and decisive element
in the foundation laid by Lorentz, Poincare and others, on which
the edifice could be built" (M. Born, Naturwiss Rundschau, 1956).
First among those "others" is Larmer, who derived the Lorentz
transformations back in 1900 (Vogt employed transformations very
similar in form even earlier, as far back as 1887).
There are other assessments of the part played by Einstein, Lo-
rentz and Poincare in elaborating the STR. And whereas extremist
views tn effect rejecting Einstein's contribution cannot be treated
-seriously, more moderate statements such as "special relativity
826 Supplement
was developed by Lorentz, Poincare and Einstein" are, in the final

analysis, their authors' affair: such things cannot be decreed, and
no one has invented an instrument for gauging scientific merits
with pharmaceutical precision.
To avoid any misunderstanding, one more comment regarding
the commonly used formula "Einstein's relativity theory" is ap·
propriate. It is a natural and legitimate formula, all the more so.
as it ts by no means the same as saying "Einstein's special theory
of relativity". For when we speak of relativity theory in general
\VC mean both the special and the general theory of relativity. The
general theory of relativity elaborates upon and advances special
relativity and is generally regarded as an unsurpassed pinnacle
of theoretical physics*. Max Born, for example, stated in 1955:
1'1 have held and continue to hold that this is the greatest dis-
covery of the human mind involving nature, which most remark·

ably combines philosophical depth, intuition, physics and mathc·
matical art. I admire it as I would a work of art." Noteworthy is
Einstein's own remark in a letter to A. Sommerfeld written in 1912,
when he was working on general relativity: "In comparison with
this problem the initial relativity theory (i.e. special relativity.-
V. 0.) was child's play." From another letter of Einstein's we know
that "the period from the origination of the idea of the special
theory of relativity to the completion of the paper which set it forth
was five or six weeks". It took Einstein eight or nine years (from
1906 or 1907 to 1915-1916) to elaborate the general theory of rel-
ativity, after which he continued to work on it until his death on
April 18, 1955. To this should be added that the general theory of
relativity is, more than any other theory in the history of science.
the creation of one person, Albert Einstein. Finally, relativity theory
emerged from the confines of the scientific community and reached
the general public only in 1919, when the deflection of light pass-
ing close to the Sun predicted by general relativity was actually
observed. Hence, relativity theory as a whole can be legitimately
associated with Einstein, End him alone.
Finally, a few words about who was the first. In 1952, Max Born
wrote to Einstein from Edinburgh: "The elderly mathematician
Whittaker, with whom 1 am friendly, and who resides here as
honorary professor, has prepared a new edition of his old book
A History of the Theories of Ether and Electricity, the second vol-
ume of which has already appeared. It includes, among other
things, a history of the theory of relativity, with the peculiarit>
• As space does not permit me to go in more detail into the place of general
relativity in the development of physics, I could refer the mJuJsiti\e reader 10
cxp::~\~~s 7oheE~n~~~e)'!!,ri~~j~~te~;ena~e~hein°1~:ral-,!~t~~!r Cot~~~~~~itl9il~o~
~ussian, Moscow, 1974.
Supplement 321
that tb discovery is ascribed to Poincare and Lorentz, whereas

your work is mentioned as secondary. Although the book comes
from Edinburgh, I am not really afraid that you may imagine that
I am behind it. In fact, for three years I have been doing all I can
to dissuade Whittaker from his intention, which he cherished for
a long time and which he liked to advertise. I reread the old orig·
inal papers, including some of Poincare's incidental works, and
provided Whittaker with English translations of the German
v.:orks .... But all was in vain. He insisted that everything of sub-
stance could be found in Poincare's work and that the physical
interpretation was obvious to Lorentz. But I know how sceptical
Lorentz really was, and how long it took for him to become a
•relativist'. I explained all this to Whittaker, but unsuccessfully.
This angers me, because he enjoys great prestige in the English·
speaking countries, and many will believe him. It is especially un-
pleasant to me that he makes all kinds of references to partial
<:ommunications regarding quantum mechanics in such a way as
to especially praise my role in it. So that many (if not you your-
self) may imagine that I am in some unsavory way involved in
this thing." •
Einstein's answer was: "Dear Born, don't give any thought lo
your friend's book. Everyone behaves as seems to him right, or.
~xpressed in deterministic language, as he has to. If he convinces
others, that is their problem. At any rate, I found satisfaction in
my efforts, and don't think it is sf>•tsible business to defend my few
results as 'property', like an old miser who has laboriously gath-
~red a few coins for himself. I don't think ill of him, to say nothin~.
Qf course, of you. And I don't have to read the thing."**
The answer is very typical of Einstein, and it can rlarify much
for those who are unfamiliar with his life. Actually, it explains the
main thing· the "secret" of his exceptional popularity in the mod-
ern world. The fact that he was the greatest of the great physicists
·of our, and not only our, age, is fundamental, but it is not all.
Einstein was also a champion of justice, freedom and other human
rights, he despised the forces of evil and offered an example of
nobleness and lofty human dignity. It is simply impossible to
imagine Einstein engaging in arguments about, still less bickering
onr, priority. The same is true of Lorentz and Poincare. Lorentz,
who contributed so much to the development of special relativity,
gave the credit of its enunciation ''solely to Einstein", and noted
Poincare's contribution. The latter spoke highly of Lorentz's role.
Einstein "tressed the contributions of Lorentz and Poincare. It
• Albert Ems/ern. Hedwiga und Max Born, Brfe[wechsel, 1916-1955 (Jetter

·dated 26 September. 1952).
•• Ibid (letter o£ 12 October, 1952),
323 Supplement
could be suggested that Poincare did not consider Einstein's con-

tribution to be so great and perhaps even felt that he had "done
it all" himself. But actually we can only speculate about what Poin-
care felt from what he did not say rather than any complaints he
ever voiced.*
So far we have spoken only of the initial works of Lorentz, Poin-
care and Einstein. It is to be hoped that this is adequate for a com-
parison of their relative significance. I should like to conclude hy
emphasizing that, naturally enough, work carried out after 190.1.
also contributed to the elaboration of special relativity. We cou!d
note papers by Einstein himself, as well as by Max Planck and,
especially, Hermann Minkowski (his four-dimensional interpreta-
tion of the theory proved extremely fruitful).
II. The unsuccessful search for a medium lor the propagation
of light. Light phenomena in uacuo play a special part in the spc·
cia! theory of relativity. The speed of light in vacuo is the limiting
speed with which signals can be transmitted, and it is right to say
that the history of relativity theory begins with the discovery that
the speed of light is finite.
As indicated in § 1.8, the theory of relativity proceeds from the-
consideration that the propagation of light (electromagnetic waves)
requires no material medium; in other words, light can travel
through vacuum. The idea entered physics with great difficulty,
and it is associated with the special theory of relativity. Today 1t
is part of the ABC of physics. The abandonment of the notion of
a "lummiferous medium" under pressure of experimental facts !S
an extremely instructive page in the history of physics which is
worth dwelling upon. We must, however, begin from afar and
briefly recall the development of notions regarding the nature of
light. In the early 17th century two points of view on the nature
of light appeared, neither of which has lost its significance to thi<>
day. One, the "corpuscular", belongs to Newton; the other, the
"wave", belongs to Huygens. Newton's initial premise can be rea-
dily appreciated: the success of his mechanics required a mechan·
ical interpretation of light. Newton held that light represented the
motion of special material particles called corpuscles. The basic
properties of light- propagation in a straight line through a ho·
mogeneous medium, the laws of reflection and refraction- can be
easily explained in terms of the corpuscular picture. In a homoge-
neous medium no forces act on the corpuscle, and it moves by
inertia, i.e. in a straight line. Reflection occurs according to the
law of elastic impact (like a billiard ball striking the cushion of
the table). The angle of incidence equals the angle of reflection,
~s remarkable that there ~~ not a single mention of E1nstem's ~~oork
on special relativity in any paper of Poincarfs although he d1ed seven years.
after it was enunciated.
Supplement .,.
which is precisely the law of reflection of light from a stationary
surface.
If a corpuscle at the interface of media I and II is subjected to
forces normal to the interface in the direction of the denser me-
dium, it changes the direction of motion. Indeed, let the velocity
components of the corpuscle in medium I be V11 and Vr. The forces
acting at the interface augment V11 , the direction of the velocity
changes, and "refraction" occurs (Fig. S.l).
f"tg. S.t. (a) The change in the direction of mohon of a corpuscle in crossing
the interlace of two media 1 and II Alter crossing the interface V; > V•. but
Vf = Vt. and the direction of the velocity changes (b) Refraction of light cros-
:&ing the interlace of two media.
We know from geometrical optics that when light passes from

.a medium with a refractive index n1 into a medium \Vith a refrac-
tive index n 2 refraction occurs. The connection between the angle
oi incidence i and the angle of refraction r is given by the Descar-
tes-Snell law:
Qualitatively, Newton's reasoning explains the refraction of light.

Newton also knew that light possessed properties which fitteC:
with difficulty into his scheme (recall "Newton's rings" in optics,
a typical interference phenomenon), but he devised ingenious ex-
planations to maintain the corpuscular picture. At about the same
time, Huygens put forward the idea of the wave nature of light.
He proceeded from an analogy with sound waves, although he harl
no idea of the nature of light waves. He knew that light could
travel where sound l'ould not (if you place a bell under a trans-
parent dome and evacuate the air you can see the hammer strike
the bell but do not hear the sound). Huygens (and all phystcists
.after him until Einstein) could not imagine vibrations propagating
in the absence of any medium. So it was essential to introduce a
330 Supplement
special medium through which light waves could tra,·el. Huygens

called it the ether. Thus appeared a concept the fallacy of which
was revealed only by the theory of relativity.
Newton rejected the wave theory of propagation of light. He
based his reasoning on the phenomenon of double refraction in
crystals. Newton showed that if light propagates as waves then
double refraction indicates a preferred direction of vibrations in
the beam. But in Newton's time o11ly longitudinal waves were
knO\vn, which do not possess this property. So he rejected the
wa\'e theory, although he conceded that it was plausible. Newton
also categorically rejected the ether*.
The corpuscular view of the nature of light was dominant for
a hundred years after Newton's death. Its success was to a con-
siderable degree due to his prestige. But a theory is good only as
long as it does not contradict facls and explains them. The begin-
ning of the 19th century brought with it the discovery of pheno-
mena which offered convincing testimony of the wave nature of
light. Light interference and diffraction were thoroughly investig-
ated, and rectilinear propagation was satisfactorily explained on
the basis of the wave theory. The discovery of polarization indi-
cated that light waves were transverse.
Thus, the 19th century ushured in the triumph of the wave theory
of light There seemed no doubt at all that light was a wave proc-
ess. But 19th-century physicists could not imagine oscillations in
the absence of some bodies or some medium. There had to be one.
Its name- the ether- had already been coined by Huygens; it
remained to establish its physical properties. Nineteenth-century
physics was dominated by the ideas of mechanics, an"d it is hardly
surprising that ether was endowed with the mechanical properties-
of a solid (transverse vibrations can propagate only through elastic
solid!>). It was, of course, a queer solid indeed: it could not be
sensed in motion, was invisible, could not be touched; but neithu
could it be endowed with other properties without coming into
contradiction with observations.
But even setting aside the difficulties with the properties of the-
ether, another pertinent problem was that of a frame of reference
\\ ith which the ether could be associated, i.e. the system in which
it was at rest. Obviously, it would have to be a preferred S)stem
o\"er all others, at least as far as optical phenomena were con-
cerned. Here, nature itself seemed to offer a direct answer to the
question, if the aberration of light was taken into account. The
phenomenon consists in the following. If a ray of light is observed
Supplement 331
from two rderencc systems moving relative to one another it will

be seen at different angles to some direction common to the two
systems (for instance, the direction of the relative velocity). If we
Qbserve the beams through a telescope, the visible direction coin·
<:ides with the axis of the telescope. But why does the motion of
the observer (telescope) affect the apparent direction of the in·
-cident light? This can be explained with the help of a simple
-example. Let a bead be falling vertically with a uniform velocity c.
We want it to pass through a pipe of length l moving horizontally
with a velocity V without hitting the walls. The way to achieve this
is by keeping the bead on the pipe's axis BB". For this, when the
bead reaches point B', the lower end of the pipe B should arrive
there at the same time. Obviously, the pipe should be tilted forward
in the direction of the motion. The angle of inclination <p to the
vertical is easily determined. Let the bead travel the distance
B"B' = l cos <pin time •· In the same time, the end B of the pipe
must travel the distance BB' = l sin <p. But l cos <p = ct, and
l sin <p = Vt, whence tan <p = VIc (Fig. S.2a).
In the corpuscular theory, the corpuscles play the part of the
bead. Consequently, the telescope must be tilted forward in the
-direction of the motion. But the same reasoning holds in the wave
theory of light. To keep the moving pipe from "indenting" the light
wave front (the velocity of the pipe is V and that of light is c) tt
must be tilted at the same angle <p, such that tan <p = V/c
(Fig. S.2b).
The aberration angle is defined as the change of the apparent
.angle at which the incident ray is observed in passing from one
inertial frame of reference to another. Obviously, it is impossible
to detect the aberration angle within one inertial frame, because
the direction of the beam (toward a distant star) is always the
-same. Nevertheless the aberration of light was discovered by oh·
serving stars from the Earth, because the Earth moves tn an el-
lipse, hence it is the same inertial reference system only over a
limited time interval. Every six months the Earth reverses its di-
rection of motion, and therefore the apparent sighting of the star
should change.
The English astronomer James Bradley was looking for the pa·
rallactic shift of stars: the apparent path traced by a star over a
year due to the change in the position of the observer. Fig. S.2c
-explains the appearance or the apparent parallactic shift of the
North Star. In the course of a year it should describe a small
-ellipse occupying a very specific position relative to the orbit or
the Earth.
In 1728, while trying to detect the parallactic shift, Bradley
discovered the aberration of light; he found that stars lying near
the pole of the ecliptic indeed describe an ellipse the major semi..
332 Supplement
axis of which is t:qual to 41". However, tht. ellip!'e did not lie as it
should have for parallactic shift (Fig. S.2c).
Fig. S.2d shows how the aberration of light affects the apparent
displacement of a star. Star s, lying perpendicular to the plane of
the Earth's orbit, is observed from two diametrically opposite po-
_t s
,,
l'c ~
{C) (d)
~~~-ta~i~' "~ft1 ~et:i~~at~Y(~j 1 ~"~-a~!8 tro~rs~J:~ 5 ~a~~r~f~o~l ~~~~o~i~~~~~r:~~~;

fa~~i~e s~if~i~f ah~~~z~~\a~~e ~~~ 1~f ~~~o~~Jfptfc ~JeT~~ ~~!~~fi~~e g{t1~: ~~~~~-
Thanks to this effect, in the course of a year the star describes a small ellipsr
Allention must be given to the observer's locatwn on the Earth; an observer at
point A, for example, sees star A on the celestial s~here (the position of the st:~r
~rb~~s~ls~~~~~~e ~he ~~ar (~lt1~e p~~:r~i~;~~ e~1ti~~Tc \oa~~s~~ilieE:~~~·l~i;~~t:~nt~~
course of the year Howe\•er, the respecilve positions of the Earth and the st~r
differ from those shown m (c). Th1s is the difference between aberrational n•1d
paralla~!~~ 0 ~h~f~~~:~h~~afh:£;~ht~e.~~~i~~ ~~~~!e!h(s~!~~~~~~ela~~/te aber-
sitions A and C. Over six months the angle of the direction toward
the star from these two points varies through 2q>. In Fig. S.2d,
angle q> is the angle between the direction at which an observ~r
\\ould see the stars from a stationary Earth (which, of course, is
impossible), and the apparent direction to the star s. Six months
later the same angle will be in the opposite direction, and the
difference between the apparent directions to the star s is 2cp. We
shall use a simple calculation to evaluate angle cp. Light from lhe
Supplement 333
North Star lalb normal to the plane of the Earth's orbit. The
motion of the Earth is perpendicular to the direction of the ray. The
;e~~iJr ;~s~h~e~~;,t~.!: a:c~n10~/cm~, 2~~~";v2~o~y41~~ T1~fshtw~!
the value Bradley obtained. He realized that he had failed to detect
the parallax of a fixed star (it was discovered a hundred years
later by Bessel) and instead discovered the aberration of light.
Bradley explained it on the basis of the corpuscular theory, which
we presented here. But the same result obtains for the wave theory
as well. Thus, the explanation of aberration posed no difficulty for
the wave theory. It did, however, involve an inevitable corollary
regarding the "luminiferous ether". It had to be assumed that light
travels through a medium at rest with respect to the heliocentric
system, otherwise it would not impinge normal to the plane of the
Earth's orbit.
It thus followed from the aberration of light that the ether was
stationary in a heliocentric frame of reference (Newton's absolute
system). The heliocentric system thus turns out to be a preferred,
r~~;~~,~~t 0 1
t~\ ~e r~~~~~~n~~e~~~:~~~~~r~i~~ g~eliff~:~v \~taf~i~~~
unpe
waves were electromagnetic waves of a specific frequency.
If we assume the existence of a luminiferous medium, its role
differs in no way from that of any material medium transmitting
\·ibrations. If the propagation speed of vibrations in a reference
system in which the medium is at rest is u, then in any other sys-
tem moving relative to the medium with the speed ± V the propa-
gation velocity of the oscillations will be c ± V. Under any wave
theory the velocity of the waves is independent of the motion of
the source, but it depends upon the motion of an observer relath·e
to the medium the oscillation of which produces the waves: in our
case the ether. Thus phenomena depend not only on the relative
velocities of bodies, but also on their velocities relath·e to the me·
dium
These assertions are best explained with the example of the Dop·
pier effect for sound waves in air. Let the air be at rest in a sys-
tem K where the velocity of sound is u; the source is moving re·
Jative to K (i.e. the air) with a velocity V, and the observer is at
rest in the system (Fig. S.Ja). Let us attach to the source a sys-
tem K' (motion toward the observer) or K" (motion frtJm the ob-
server). We can now reproduce the reasoning in § 3.4. Pulses are
sent from the source at intervals T' or T" (2n/T' and 2n/T" are
the proper frequencies w0 ). The receiver picks up t\\'O consecutive
signals at intervals T = T'- V~' = T' {I -f) in the first case.
and T= T" + v; = T"( I+ f) in the second. Thus, if the source

831 Supplement
is moving relative to the observer and the medium, when H ap-

proaches the observer the latter will observe an increase in the fre-
quency<ll=ffio/( --f),
t and when it recedes, a decrease ro=
=.,1(1 +f)
If, now, the source is at rest relative to the medium, while the
observer is receding from (or approadiing) it, then the propaga-
tion velocity of the vibrations relative to the observer will be u- V
+
or u V, respectively. Now the source is at rest in K, and the
/(I
/(1 : /(; I(" ':I( I('
! !
: t l ' ' '
~ ~I
f 0: 2 Obserrerat Obse!'Yer .roJ.ce at 0/lseroer
S011rce : SIJI.Jrce rest m K
11edllim at rest tn K rest inK
14} (6)
Fig. 5.3. Illustration of the derivation of the Doppler effect formula (or soun1.
(a} The observer and the air are at rest in the K. system, and the source 1s mo\'·
iug through the medium with the velocity V. (b) The source of the sound and
the air are at rest m the K system, and the ob~erver is moving through the air
with the velocity V
observers are connected with the systems K' and K". Signals ar~
emitted from the source at intervals T; they will reach the observer
inK' at intervals T'=T+ v~TV =(-'-v-)T· and the observ~r
1-v
in K" at intervals T"=T- v~TV =( 1 +If) T. Only now w0=
= 2n/T, wh~nce in receding we obtain a reduction in frequency
<il=Wo(l -f). and in approach an increase, ffi=Wo(l +f).
We find that th~ equations are different for the same relative
velocity of the source and the observer. Similarly to what was
done in § 3.4 for the case of wave radiation at an angle to thl!
direction of propagation, we obtain ro = Wo (I - ~cos 6) and w =
=-----v-.
t-vcos8
This example readily demonstrates that the ether assumption
violates the relativity principle. Of course, the relativity princip!~
Supplement
is valid in the prc::,t'Ille of a medium, but to make the cond1tiom

of the experiment identical, the velocity of the reference system re-
lative to the medium must be the same. In other words, every re-
ference system must be associated with a definite medium. Thus,
for the relativity principle to hold the ether would have to be en-
trained, or partially entrained by the reference system (the ether
drag). This is a strange assumption, but nevertheless it was de-
veloped, though in a different connection.
If the hypothesis of the stationary ether were correct it could
explain other optical phenomena. From this point of view, Fizeau's
experiment (1851) led to some very mysterious results We already
described the experiment in § 3.5, but let us take a look at it from
the point of view of a physicist of the 19th century. Let the water
be at rest in the system K', which is moving together with the-
water with velocity V relative to the laboratory. Let us also as-
sume that the laboratory can be considered the preferred reference
frame in which the ether is at rest. Light propagates through the
ether The substance changes its phase velocity, but the velocity
of the substance does not matter. Consequently, the velocity of
light in the laboratory system, v, is simply equal to the velocity
of light, u', in the stationary water. Of course, v' = c/n. Assume
for a moment that the ether is "dragged" together with the water.
Then, naturally, the velocity of light. u', would be compounded
with the velocity of the ether, i.e. the water, and we would g~t
v = v' + V. An ether endowed with the strange property of "par-
tial" drag would give u = v' ± kV, the sign depending upon the
relative direction of motion of the light and the medium. Fizeau's
experiment, which was subsequently repeatedly confirmed, gave
the result v = v' ± (I - ~) V. In any event, the stationary ether
hypothesis contradicted the results of Fizeau's experiment.
The battle for recognition of the ether did not end with this, of
course, but before describing further attempts to detect the ether
it is useful to dwell on the relationship between physical experi-
ment and theory. Cognition of nature imolves the search for laws
that correctly reflect existing relationships or, philosophically
speaking, the search for objective laws of the objective world exist-
ing independently of us. Nature is studied by people, and they in-
variably introduce their subjective notions and sensations, to say
nothing of inevitable errors. Consequently, the laws of nature must
be verified. What is the criterion of correctness of a physical law?
The correctness of physical laws is revealed in practical activity.
The guarantee that our knowledge agrees with the laws of nature
lies in their verification by different people in different places, and
repeated verification by one person The most important means of
\erifying physical laws and discovering the laws of nature is by
,,. Supplement
artificially creating the necessary conditions, i.e. by staging a

physical experiment.
?ut physics cannot be restricted to experimental results. Physics
is impossible without theories that can be used to systematize and
explain various natural phenomena on the basis of a small number
of fundamental laws. In turn, physical theory- and today it is a
science in its own right, theoretical ~hysics- is closely linked
'lhith mathematics. When experimental material accumulates, a
theory appears to explain a specific set of phenomena. Can an ex·
periment or series of experiments vindicate or rerute a theory? We
are not, of course, speaking of erroneous experiments, which may
always occur; ultimately their erroneousness is sure to be shown.
Sometimes in developing a theory the limits of its application are
apparent, and one should keep within them. With these reserva-
tions, the following can be asserted: if a single correct experiment
carried out within the limits of applicability of a given theory con-
tradicts that theory, the theory must be assumed wrong. As for
.. proof" of a law or theory through comparison of obtained con-
dusions with experimental data, no amount of experiments agree-
ing with a theory can be regarded as the ultimate proof. A theory
exists and is considered correct as long as its conclusions do not
contradict some new experiment within the domain of its applicabi-
lity. The situation is analogous to one of the rules of mathematics:
the validity of special cases is not proof of the validity of a general
theorem, but one example to the contrary refutes it.
Physics is essentially an experimental science. The process of
cognition of nature, of which physics is a part, is continuous and
unlimited. The question of establishing the ultimate truth is, rather,
a philosophical issue than one of some specific science. Individual
ph}sical experiments reveal various specific laws and regularities
or, from the philosophical point of view, relative truths. Although
the regularities contain elements of absolute truth, they do not
proYide exhaustive knowledge. In the final analysis, every theory
i~ either restricted or erroneous. At each stage a theory's validity
is determined by the absence of experimental data contradicting
it, while its value lies in its ability to explain and predict ob-
servable phenomena. These remarks, interesting in themselves,
need not have been cited if they were not useful in setting forth
the history of the ether. We shall rely mainly on "negative" experi-
ments pointing to the fallacy of various premises postulated in an
attempt to salvage or discover the ether.
Let us return to the assumption that the ether is stationary in
.a heliocentric system (which follo\\'S from the observed aberration
of light). If the ether is stationary in a heliocentric system, the
Earth in its motion around the Sun should experience an "ether
wind". It is not hard to conceive an experiment to discover motion
__________________s_u~~-~_m_rn_t_______________ M7
relative to the ether. It is shown graphically in f1g. 5.4. Two

photoelectric cells are sitting on an optical bench parallel to the
Earth's velocity in orbit. Halfway between them is a light source,
I, which emits flashes of light. Light travels through the stationary
ether with the velocity c. But cell C 1 is moving toward the light
beam, and cell C2 is receding from it. Light travels toward C1 with
the velocity c +
V, and toward C2 with the velocity c- V. Thus,
I
c, c~v ~lu c-V Cz
~71~---=v
Fig. S.4. A light source, /, and two photoelectric ~;:ells, c. and C2, are mounted
fo"ci~~ ~r'l~! 1 ~=~thh.i~~~s b~~bi~. i~f ~?:~;o;~~P~=:~~~eli~ 0t~~es~~~~~~r~ ~·1et~ee/ii
travels from the source I toward cell c. with the velocity c + V, and toward
cell Cz with the velocily c- V.
C 1 will register the arrival of the light beam before C2, the time
interval being
AI= c~V -c:f:v-=2l{r~. (S.II.l)
1-cr
Before considering the possibility of carrying out such an exper·
iment, note the following. In speaking of the velocity of light, the
ether assumption reduces solely to the assertion that there exi::.ts
a single reference frame in which the velocity of light in vacuo
is c. It is the one and only preferred system. In all other frames
moving relative to it with velocity V, the velocity of light in vacuo
is given by the classical rule of addition of velocities, c' = c ± V.
Our reasoning applies to a reference frame K connected with the
ether (or simply with the frame in which the velocity of light in
vacuo is c). It is assumed here that the signals received by the
photoelectric cells are registered in the K frame. But if we apply
the same reasoning to a frame K' in which the optical bench is at
rest we obtain exactly the same result. Simply, in that frame the
velocity of light in vacuo differs from that in K: c- V to the
right, and c +V to the l-eft (fig...: S.4). In the K frame, we recall,
+
c- V and c V would be simply the velocities with which light
reaches cells C 1 and C2 "'. It is not surprising that the rE>suits
ate the same in K and K', since in classical mechanics the timing
of events is absolute.
Thus, if we could measure the time difference .1t (S.Il.l) we
would thereby only prove the difference in the speed of light in
Supplement
frames K and K'. That, of course, would be indirect proof of the

ether (see further on).
As to the order of magnitude of M in such an experiment, if we
put. for example, 1 ~ 100 m, and V ~ 3 X 104 m/s (the speed
of the Earth on its orbit). then !J.t ~ 10- 10 s. Even modern hard-
ware cannot measure such a minute time interval.
In 1878. Maxwell suggested an experiment employing the phe-
nomenon of interference of light to detect the motion of the Earth
relail\'e to the ether, at rest in a heliocentric system (if light a..:-
tually propagated through it). Maxwell thought the required accu·
racy of measurement to be unattainable, but three years later Mi·
chelson huilt an interferometer capable of detecting the motion of
the experimental set-up relative to the stationary ether.
Michelson's experiment, carried out in 1881 "'. was as follows
(Fig. S.5). A beam of light from a source,/, is directed on a half-
silvered glass plate, P. Half the incident light is reHected, tlw
other half passes through the glass. Michelson's instrument (today
it is called Michelson's interferometer) had two mirrors, S 1 and 52,
located as shown in the drawing, at distances L 1 and L 2 from the
half-~ilvered glass P. All the parts of the interferometer were firmty
secured to a heavy block of stone floating on a disc of wood in a
tank of mercury so that the whole system could be turned smooth!}.
At the plate P the beam is split in two: beam I travelling to
mirror 5~o and beam 2 to mirror 5 2• Each beam reaches its mirror
and returns to plate P. As the plate is semi-transparent, a portion
of the light from both beams travels in direction 3. Since each uf
the beams I and 2 is a portion of the initial beam, the two ray~ I
and 2 travelling in the direction 3 are coherent and can interferP..
Let us determine the time it takes the beam of light to travel
from P to 5 2 and back. The interferometer is at rest in a svstcm
which is travelling with a velocity V relative to the ether·. The
distance between P and 5 2 is L 2, the speed of light to the right is
+
c- V, and to the left, c V. Hence, the requirf"d time is
[2 = c~V + c~V = 2
; 2
~ (S.II.2)
1-cr
Let us now find the time 11 it takes beam I to travel from P :o
mirror S 1• In the time 11 mirror 5 1 travels the distance V1 1, and
the light travels a distance c1 1 along the hypotenuse of tri<~nglt!
PSJP"· From this right-angle triangle it follows that (cll=Li +
+ (Vl 1) 2 , whence
i 1 =...jc2L:_vi=J:;-~. B=7·
-~lee A. A. Michelson and I. W. Morley, Phys. Mag. (5) 24, 449 (1887)-
{b)
fig. S.S. (a) Michelson's interferometer (b) Diagrammatic representation of Mi-

chelson's experiment. A beam of light from the source 1 is split on the semitrans-
parent plate P into two beams, I and 2, travelling along and perpendicular to
the direction of the Earth's motion in its orbit. The velocity of the Earth's orbi-
tal motion is indicated by the arrow and the Jetter V. Beams I and 2 reflect
from mirrors S 1 and S2, respectively, and return to plate P. After reflecting and
relraclmg, [\\'0 beams travel in the direction 3. That IS the direction in which the
:~~~~fre~~~etg:t!:h~t!s aop~:;:~d~ r;r~~:~·s~~; e~;~~n~ o~o~e ~~:~=~~~~n~h~o~~~~s-
!ton of the Earth's motion, and beam 2 normal to that direction. If light propa-
,a~~ut~~~~~g~e a5~~tif~:rfn:e'~:~e~~: ~:~=b~~r~~ ~~th;ir~t~e:~s ~o~it
ehange (the interference fringes would shift). The experiment, however, d•s-
covered no shift in the 1nterference fnnges.
... Supplement
The total time it takes the light beam to travel from P to S, and
back is double that amount, so that
(S.JI.3)
The dtfference between times 11 and t2 is due not only to the

difference in the lengths of the arms of the interferometer but also
to the motion of the apparatus.
One of the arms of the interferometer is aligned parallel to the
motion of the Earth along its orbit (which is known from astro·
nomtcal observatiOns). But we live on Earth which, according to
the assumption, is moving relative to the ether, hence the experi-
ment is inevitably conducted in an ''ether wind". Hence, too, it is
impossible to compare the interference pattern "without the ether
wind" and "in the ether wind". By aligning, for example, arm PS 1
in the direction of the velocity V, we obtain a certain interference
pattern: an alternation of light and dark fringes •, depending on
the difference in the propagation of beams 1 and 2:
111=12-lt=.!.(_L,____L,_)~
c I - B2 -v'!-W
~-' -(-L,
c .Vr:=-H"7
__ L,) .
....'r=!Ji (S.II.4)
This time the difference depends on both L1 and L 2 and the ve-
locity V. If the condition L 1 = L 2 could be guaranteed the differ-
ence would depend on L 1 = L2 = L in this way:
(5.11.5)
When L is known, the interference fringes obser·, t:d in the appara-

tus are uniquely determined by its motion. But, first, it is simpler
not to worry about assuring the condition L1 = L 2, and secondly,
it is more convenient to observe the change in the interference pat-
tern, that is, the displacement of the fringes. For that the whole
set-up 1s rotated through 90°. The arms of the interferometer change
places. We then obtain
tf= 2 ~ 1 ~. t2= 2 ~' ~· (S.II.6)

i).t'-1'-t'~~c-L,_ _ _
-2 l c ~
L_, )--'-(L __
1-Bl -c.Y~
L_,)
I-I32. 2
• On Michelson's interferometer see, for example, the already quoted OptiCS

by G. Landsberg.
Supplement 341
This means that in the rotation of the apparatus the time difference
changes <!S follows·
0.1• =lit' -lit= 2 (Lt
c~
+ /.2) (1 - _L_)
~
~- L +La
c
1 B 2, (S.II.7)
a:"~ the fringes shift.

ir we want to obtain a difference in the distance travelled by the
oeams of the order A., 111• must be of the order of the vibration pe-
riod, T. But T='}.jc. For lhe motion of the Earth in orbit B ~ to-~;
for visible Jig}Jt A.:::::: 5 X IQ-5 em, whence the total length of the
+
interferometer arms L 1 L2 :::::: 50 metres. Such a path for the
light beam ca:1 be obtained by repeated reflection.
Michelson could have detected an "ether wind" blowing at
104 m/s. Together with all other physicists, he had nut the slightest
doubt that such an "eiht:r wind" would certainly be felt. But there
was .. one. Th€ ~xperiment was repeated many times, with greater
and greater precision: today an "ether wind" blowing at 30 m/s
could be detected, but Miche!son's result, or his nuli result, as it is
also called, stands fast. There is no doubt that it is correct.
However, it does not necessarily follow from Michelson's ex-
periment that there is no ether. Its results can be explained by
endowing the ether with certain properties. The coup de grace
against the ether required other observations. But IP.t us first draw
some conclusions from Michelson's experiment, without linking it
with the search for the "ether wind". The experiment showed that
a rotation of the interferometer on Earth does not cause a shift in
the interference fringes. In principle, however, such a shift could
be associated with the difference in the speed of light along the
two directions in the frame of reference of the interferometer.
Thus, regardless of whether the ether exists or not, Michelson's
experiment shows that the speed of light over a closed path meas-
ured on Earth is the same in all directions, i.e. it is isotropic. As
the Earth movl's around the Sun along a closed curve, and can
therefore be regarded as an inertial frame of reference only over
a short time interval, our measurements are actually carried out
in many inertial referem:e frames. It thus follows from the expe·
riment that the velocity of light along a closed path is isotropic in
any inertial frame of reference
A somewhat modified form of Michelson's experiment, staged in
1932 by Kennedy and Thorndike, offers confirmation of Einstein's
main postulate. From Pqualion (5.11.7) it is apparent that .lt also
depends upon c. H the absolute velocity of light in vacuo were
different in different inertial frames of reference, then a shift in the
fringes would be observed in passing from one IFR to another. The
interference pattern was on several occasions observed continuously
for periods ranging from eight days to one month (with a three·
342 Supplement
month interval). During that time the apparatus changed many

inertial frames of reference. It is so sensitive that a variation of
2 m/s for c could be detected. But nothing happened. It thus follows
from Michelson's experiment that the speed of light has the same
value (travelling there and back) in all directions within a given
IFR; moreover, it has the same value in vacuo in all inertial
At the risk of interrupting our disCourse, it is worth presenting
the results of Michelson's experiment in terms of the special theory
of relativity. If the experiment is staged in an inertial frame nf
reference its result is obvious. In every such system the velocity
of light in vacuo is isotropic and the interference pattern observed
In the experiment is due solely to geometric differences in path. In
short, in every IFR in which the interferometer is at rest we obtain
exactly the same picture as we would have in a classical consider·
otion in the preferred reference system in which the ether is
stationary.
The need for a more complex interpretation of the result arises
when the experiment is considered in an inertial reference frame
relative to which the interferometer is in motion. Let the inter-
ferometer be at rest in K' and the experiment be staged in K; for
the sake of simplicity we assume that in K' L 10 = L 20 . We have
added the subscript 0 to stress that we are dealing with proper
lengths. Obviously, equations (5.11.4) and (5.11.6) remain valid,
but L2 = L20 ~and L 1 = L 10• It is apparent then that !'it,
determined according to (5.11.6), vanishes. Naturally, a rotation
of the apparatus yields no effect. Thus, in special relativity the null
result of Michelson's experiment is explained by the relativity of
the length of the measuring rods (which is most directly linked
with the invariance of the velocity of light,§ 2.3).
To explain the result of Michelson's experiment and salvage
the ether, Lorentz and Fitzgerald assumed that in motion relative
to the stationary ether all bodies contract by the factor ...jl- B2
in the direction of motion (the "Lorentz contraction"). It is ap·
parent from the foregoing reasoning that such an explanation is
possible. But it is essential to emphasize the difference in the reb-
tivity of the length of measuring rulers in special relativity and in
the Lorentz contraction. In the 5TR the contraction is a conse-
quence of measurements made in the relati\'e motion of reference
frames. With Lorentz-Fitzgerald it is the consequence of motion
relative to the ether, which retains the preferred reference system.
The fallacy of the Lorentz contraction is revealed in a modification
of Michelson's experiment. Michelson's interferometer had arms
of the same length; Kennedy and Thorndike's had arms of
different length. If the interferometer arms are different, a shift
Supplement 343
in the tnterferencc fnnges should be observed \vhen tl1e mterfero-

mcter's velocity relative to the ether changes. But on Earth an
interferometer participates in three motions relative to the ether:
the Earth's motion relative to the Sun, its rotation, and finally, the
motion of the Sun. The resultant velocity varies by a certain value
e\·ery 12 hours (and six months). These variations should result in
a shift in the interference fringes. Indeed, if Lorentz's hypothesis
is correct, then L2 = L20 ,Y(l - B2), and from (5.11.4) we get
tlt=cn(L20-Lw). (5.11.8)
Without rotating the apparatus, let us see how 6.1 varies with
a variation of the velocity B relative to the ether by 1.8. Differen-
tiating (S.II.8) with respect to B, we obtain
~
.6.8
= ~ (L,- L,,) 6
c ~u
(----"--=)
.yr=e2 = L -c L,o 68 •
20 2 (S II 9)
· ·
The latter equation is written to the accuracy of D2, but prolonged

observations of the interference pattern revealed no variations.
Another way to square the ether with the results of Michelson's
experiment is to assume that the ether is "dragged along" by
moving bodies. But as we have seen, the aberration of light
"agrees" only with a "stationary ether in a heliocentric system".
An experiment specially staged by Fizeau (see§ 3.5, where it is ex~
plained in terms of special relativity) to determine the ether "drag"
led to the conclusion that there was a "partial drag".
Of course, the mentioned experiments and observations far from
exhaust the attempts to establish the properties of the ether. But it
is already obvious that the ether would have to be endowed with
extremely contradictory properties. But the ether's most significant
''contribution" to physics would probably have been the rejection
of the relativity principle in electrodynamics.
In 1905, Einstein's paper, On the Electrodynamics of Moving
Bodies, appeared. It virtually set forth the whole of the special
theory of relativity, which not only offered a natural explanation
of the result of Michelson's experiment but also correctly interpret-
ed all known mechanical, electrodynamic and optical phenomena.
From the outset it extended the relativity principle to all physics
and unambiguously asserted the equality of all inertial frames of
reference in vacuo, thereby making the ether redundant. Not a
single experimental fact contradicts special relativity.
In his initial experiments Michelson could have discovered a
variation in the speed of light in a change of its direction relative
to the motion of the Earth down to 0.15 m/s; later experimentets
could have detected a variation of O.QJ5 m/s; with lasers the de·
344 Suppfemenf
tectable change in speed is a mere 3 X IQ-5 m/s. Yet no variation

in the speed of light relative to a moving observer has ever been
detected.
That the speed of light in vacuo does not depend on the motion
of the source has been repeatedly verified. One experiment was
carried out with an extraterrestrial source, the Sun (A.M. Bonch-
Bruevich, 1956). If the velocity of light depends on the motion of
the source, then by measuring the velocity of light emitted from
two opposite points of the equator it should be possible to detect
the difference between those veloci-
ties. No such difference was detected.
A laboratory experiment was conduct-
ed in which the flight of gamma-
quanta over a certain distance was.
compared; gamma-quanta emitted by
a stationary and a moving source
(radioactive nuclei) were studied, and
again the independence of the velocity
of light was confirmed.
It can he confidently declared that,
despite the enormous increase in thl!
accuracy of experiments, there is no
indication of the existence of a pre-
ferred reference system, or of any
Fig. S.6. Diagrammatic represen· difference in the velocity of light
tation of the Sar.nac-Garress ex- in vacuo in different inertial frames
perirnent of reference, or of any manifestation
of the ether
In conclusion we should note that accelerated motion of a ree
ference system relative to an inertial frame or reference can, of
course, be detected. A mechanical experiment of this kind- Fou-
cault's experiment- was described in § 1.5. There are also optical
variants of the experiment, which we shall mention to round out
the picture. We shall describe Garress's experiment (1912), subse-
quently repeated by Sagnac. Three mir..ors, A, B, C. and a semi-
transparent plate D are mounted on a turntable (Fig. 5.6) together
with a light source L and photographic plate P. A beam of light Q
is split at plate D into two beams, DABCDP and DCBADP. tra-
velling along the path ABCD in opposite directions. If the system
is at rest, interference of the two beams travelling from L and
split at D occurs on the plate. When the turntable rotates the in-
terferl"nce fringes should displace owing to the change in path
lengths.
Disregarding the deviation of a geocentric system from '!n
inertial system, let us examine events in the reference system of
the Earth. For the sake of simplicity we shall assume that there
Supplement 345
are many mmors and the path of the beams is virtually a circle.
Then the speed with which the light catches up with the turntable
\\hen it travels in the direction of rotation is c- V = c-OR
(where R is the radius of the turntable and Q is the angular veloc-
ity of rotation); for the beam travelling in the opposite direction
the velocity is c+ V= c + QR. The time it takes the beam to
travel around the circumference is 'f1 = 2n.R/(c- V) in the first
case, and 'f2 = 2nRf(c + V), in the second. The difference between
the two times is
th='f,-'f2=2nR(cb--dv-)= 4nc~V -;-V~~.
--;r
where S is the area of the turntable. Sagnac observed a shift in the
fringes that agreed nicely with this formula. The shift can be used
to determine the angular velocity Q.
If the Earth is used as the turntable its angular velocity can be
determined. This experiment was carried out in 1925 by Michelson
and Hey!. The angular velocity corresponded to the component of
the angular velocity of rotation of the Earth along a plumb line
at the point of observation. For the experiment two kilometres of
pipes were laid and a second circuit was built to determine the zero
point of displacement of the fringes. Michelson's result was
0.230 ± 0.005, the theoretica I figure being 0.236. Excellent agree-
ment!
Thus, unlike uniform translational motion of the Earth, its rota-
tion can be detected by various physical experiments.
Ill. Was Michelson's experiment "decisive" for the creation of
the special theory of relativity? Michelson's experiment is given
great prominence in virtually all books on the history of special
relativity. Most authors assume, one way or another, that speci<~.l
relati\'ity was an upshot of attempts to explain Michelson's ex-
periment, \\ hich is the theory's principal experimental basis
That is the place assigned to Michelson's experiment in the
only book in Russian devoted to the experimental foundations of
the STR, written by S. I. Vavilov in 1928: "The story of [Mich~l
son's experimentJ is set forth here in fairly great detail becau,e
the basic postulates of the relativity theory were formulated on 1ts
basis." And in his foreword Vavilov wrote, "After reading this bo"Jk
the reader will understand why it is adorned with a picture of Mi-
chelson."
This claim is repe~:~ted in virtually all our textbooks dealing
with the history of special relativity. In Y. B. Rumer and M.S. Ryv-
kin's The Theory of Relativity (Moscow, 1960) we find: "Unlike all
preceding investigators, Einstein saw the negative result of Mi-
chelson's ~xperiment as .. ,". Foreign authors toe the same line on
... Supplement
thi& score. To cite bul one example, Laue, in a book written in

1911, states that Michelson's experiment ubecame, as it were, the
fundamental experiment for the relativity theory."
Besides textbooks and popular expositions there are, of course,
books on special relativity written by Einstein himself. In his book
on the special and general theories of relativity, which he subtitled,
A Comprehensible Exposition (gemeinverstiindlich), Michelson's
experiment is mentioned in § 16: "The Special Theory of Relativity
and Experiment". However, it is not clear from the text whether
there was any direct connection between Michelson's experiment
and the enunciation of the theory. It is not mentioned at all.
Nothing in other writings and statements of Einstein indicated any
contradiction with the accepted view that Michelson's experiment
was actually the point of departure for special relativity. Students
were taught so, and schoolchildren are told as much today. It was
therefore most surprising to read in an article by R. Shankland,
published in 1963, the following excerpt from his interview with
Einstein dating back to 1950:
"When I asked him how he had learned of the Mkhelson-Mor-
Iey experiment, he told me that he had become aware of it through
writings of H. A. Lorentz, but only after 1905 had it come to his
attention! 'Otherwise', he said, 'I would have mentioned it in my
paper!"' Indeed, Einstein's 1905 paper contains no mention of Mi-
chelson's experiment or references to Lorentz's papers.
We know that Lorentz and Poincare came very close to postulat-
ing special relatidty, but in fact it was enunciated by Einstein, and
virtually in that single work in 1905 Thus, if we speak of the
decisive significance of Michelson's experiment for developing spe-
cial relativity, we should determine its influence on Einstein's work.
And from Shankland's article we learn that Einstein first heard of
Michelson's experiment only after the creation of the special theory
of relativity.
Why should thC' question of the role of Michelson's experiment
in Einstein's efforts to formulate special relativity be of such con-
cern to the teacher? It is intriguing, of course, as a curious fact,
but hardly worth devoting a whole paragraph to it. But it is not
just a matter of curiosity. It is inevitable that the history of the
de\·elopment of a theory should be reflected in teaching. It is hard,
and in some cases impossible, to sidestep the history of an issue.
With time, of course. the presentation of a discipline becomes logi-
cally streamlined (teaching is never in vain!), but the real history
of a theory's development does not always reflect the logic of its
elaboration. Nature does not necessarilv reveal its secrets in a
sequence most com·enient for their interpretation. In the initial
F~l~~\~S a:;~~e~hfn et~~n~;:tt~~~ca~f :a~~e~nt!t!l:~~~=:r~~ r;:~e~~~~i;d~
Supplement 347
ing to the logical scheme which can be constructed after the theory
has been completed.
Thus, looking from this aspect at the academic presentation of
special relativity, we can see, that, if Shankland understood Ein-
stein correctly, the teaching of the theory did not even follow the
steps in its creation. A curious situation!
The question is not of denying the role of Michelson's experiment.
Its history and implementation cannot but arouse admiration. Mi-
chelson's experiment occupies an outstanding place in the history
of natural science. And yet it played a rather unfortunate part in
tne evolution of the traditional scheme of describing spec;al relo·
tivity.
When placed at the basis of instruction in the STR, Michelson's
experiment inevitably introduces the ether. It is impossible to ex-
plain 1ts meaning without speaking of the difficult search for a
material medium through which light propagates (see Supple-
ment II). But today we all know only too well that no such medium
1s necessary, and from the methodological point of view that is
where we should begin. There is absolutely no need to go back
to the intellectual atmosphere of the later nineteenth century.
Indeed, the ether played a prominent part in the physical views
of the 19th century. It was, in fact, the ether concept which sug-
gested to Maxwell the idea of the experiment ultimately carried
out by Michelson and MC'rley. But erroneous notions, which may
have played their part at a certain stage in the development of
science, are ultimately discarded. When Galilee enunciated his
inertia principle he immediately discarded Aristotle's doctrine that
motion had to be continuously supported. When the transformation
of mechanical energy into heat was discovered the phlogiston con-
cept had to be discarded. The theory of relativity began with the
rejection of absolute motion and the ether. But nowadays no one
brings up Aristotle's doctrines in expounding mechanics, no one
recalls phlogiston in lecturing on heat; why then should the ether
be kept in describing special relativity at school and college? The
introduction of the ether in modern expositions of special relativity
is, to say the least, strange. First one has to explain at length the
reason for introducing the ether, and then conclude by declaring
that it does not, after all, exist. Can such an approach be called
methodical?
It is sometimes said that there is no getting away from the ether:
it had to be postulated by analogy with the propagation of sound
or waves on water. One should only be sure to explain at the right
time that there is a substantial difference between the way electro-
magnetic and gravitational waves travel, on the one hand, and
elastic waves, on the other, and that matter is not required for the
propagation of electromagnetic waves. Light can propagate in the
::"'=-'-----------'S::tJ:pplemenl - - - - - - - - -
absence of matter m the conventional sense oi the word (posses.:iing
rest mass). So there is no reason for the ether arter alii
To this one could reply, "True, it is better not to introduce the
•ether', which doesn't exist anyhow, but special relativity is a com·
plicated thing. The road to it was long and difficult, and it lay
through the ether hypothesis, which was discarded when the need
for it passed. But it was a natural theory which appeared in the
search for a correct solution, it reflects the logic of human thinking,
and there is no harm in setting it forth." Everything in this reason·
ing is correct except one thing. Einstein arrived at special relativity
not ,·ia the ether (nor Michelson's experiment), but along a
simpler and clearer road. And if one speaks or the logic of human
thinking it is worth taking a closer look at Einstein's reasoning.
What role did Michelson's experiment play in Einstein's work?
We have already cited Einstein's statement as quoted by Shank-
land. But Shankland's article appeared after Einstein's death and
had not been, so to say, "authorized". Not long ago a letter writ-
ten by Einstein and containing a direct answer to the question was
discovered in the Einstein archive at Princeton, and it removes all
doubts. The story of the letter is as follows. On February 2, 1954,
a year before Einstein's death, a certain Mr Davenport wrote him,
saying that he was looking for evidence that Michelson had "in·
fluenced your thinking and perhaps helped you to work out your
theory of relativity." Not being a scientist, he asked Einstein for
"a brief statement in non-technical terms, indicating how Michelson
helped to pave the way, if he did, for your theory."
Einstein replied almost immediately, on February 9, 1954. It is
his last statement on the question. It seems obvious that he had
reflected about it before (after all, he had spoken of it with Shank·
land). The letter is clear and unequivocal. Here it is
"Dear Mr Davenport:
"Before Michelson's work it was already known that within
the limits of the precision of the experiments there was no influence
of the stale of motion of the coordinate system on the phenomen<l,
resp. their laws H. A. Lorentz has shown that this can be under-
stood on the basis of his formulation of Maxwell's theory for all
cases where the second power of the velocity of the system could
be neglected (effects of the first order)
"AcLOrding to the status of the theor}, tl was. however, natural
to expect that this independence would not hold for effects of
second and higher orders. To have shown that such expected effect
of the second order was de facto absent in one decisive case was
~~~~~~or~! t~r~t~~td ~r;~t~ ~~~u\l~~~~n°~f~~~h~l~~b~~~q~~ 11l~r~~e;~
the ingenious way by which he reached the very great required pre-
cision of measurement, is his immortal contribution to scientific
Supplement 349
knowledge. This contribution was a new strong argument for the

non-existence of 'absolute motion', resp. the principle of special
relativity which, since Newton, was never doubted in mechanics
but seemed incompatible with electro-dynamics.
"In my own development Michelson's result had not had a con~
siderable inHuence. I even do not remember if I knew of it at all
when I wrote my first paper on the subject (1905). The explana-
tion is that I was, for general reasons, firmly convinced how this
could be reconciled with our knowledge of electro-dynamics. One
can therefore understand why in my personal struggle Michelson's
experiment played no role or at least no decisive role.
"You have my permission to quote this letter. I am also willing
to give you further explanations if required
"Sincerely yours,
Albert Einstein"
The Jetter is clear, and there is nothing to add. True, it is con-

tradicted by one pronouncement of Einstein's known from B. Jaffe's
book. Einstein met Michelson but once, in 1931, in Pasadena. In
a short speech, before those present, he said addressing Michelson,
''Through your marvelous experimental work [you] paved the way
for the development of the theory of relativity." Jaffe's book, from
which these words are quoted, does not give Einstein's speech in
full. From an account of it in German it follows that the book
omits a whole sentence and that Einstein spoke of the "road" to the
general theory of relativity. Thus, there is not the slightest hint on
Einstein's part that Michelson's experiment was in any way de~
cisive, although Einstein invariably stresses its beauty and funda-
mental contribution to science. In a letter to Jaffe [26] he writes
that the experiment "strengthened my conviction concerning the
validity of the principle of the special theory of relativity." That is
probably the most correct assessment of the significance of Mi-
chelson's experiment for Einstein's work. The experiment's signi-
ficance in the history of physics was quite different, perhaps even
"decisive". The experiment's "null" result quite obviously domina-
ted the work of Lorentz and many others. But that was not the
road along which special relativity evolved. Paradoxically, Lo·
rentz, who discovered the famous "transformations" that bear his
name and which embody the very essence of special relativity,
remained a long way off from enunciating the STR.
Finally, of interest is the question why academic instruction has
been so persistent in following the "Lorentz way"? One fortuitous
consideration which probably played no mean part ought to be
mentioned. Here is a small quotation from the book by H. Bondi
[21]:
350 Supplement
"What has bedevilled this issue m textbooks ts the undue pro 4
mtnence given to the Michaelson-Marley experiment . . Einstein

said that at the time he wrote his basic paper on relativity ( 1905)
he had never heard of the experiment. Later on when it was decided
to reprint various es~a}S on relativity it was decided by the pub-
lisher<> (with the advice of somebody) to start in the middle of one
of Lorentz's essays. The first part that was included happened to
be the Michaelson-Marley experiment. For this reason since then
everybody, or nearly everybody, has felt obliged to start in the
same way. And what a complicated start it is!"
Indeed, if we take Lorentz's book Versuch einer Theorie der
elektrischen und optischen Erscheinungen in bewegten KOrpern
(leiden, 1895). we see in §§ 89-92 a description of Michelson's
experiment.
As for the exclamation, "And what a complicated start it is!",
far from everyone agrees with it. But the question of how most
reasonably to explain special relativity to the student is not .:n
idle one: today special relativity is not only a part of the college
course in general physics, it is also present in the high school cur-
riculum. Instruction should, doubtlessly, be based on the ideas of
modern physics and not include outdated notions of the past.
Einstein's elaboration of special relativity began with his rejec-
tion of the "luminiferous ether", and in that sense Michelson's ex 4
periment was certainly not "decisive". Einstein's reasoning is suf-

ficiently simple and logical, and there is every reason to use it in
expounding the special theory of relativity.
IV. Why shouldn't the mass velocity dependence, or the rela
4 4
tivistic mass, be introduced? Textbooks on special relativity (es 4
pecially the older ones) often introduce the "relativistic mass",
mrel=mv=m/~. P=v/c,
which, by definition, depends upon the velocity, and try to give it
independent meaning.
Whether the relativistic mass should be introduced or not is a
purely methodological issue. Whether m,el should simply be con 4
sidered an abbreviated notation poses no problem at all. But the

physical interpretation of relativistic mechanics is an entirely
different question. Here misunderstandings and vague interprela·
tions often arise. On this is our discourse.
It is not hard to understand how the temptation to introduce the
relativistic mass appeared. One need but juxtapose the Newtonian
and relativistic equations of motion (5.37a) and (5.37b),
Supplement 351
for the conclusion to suggest itself that the ·'only difference" be-
tween them is that in the relativistic equation the mass depends
upon the velocity. The expression my is then taken from under the
derivative sign and declared the relativistic mass, with an inde-
pendent meaning attached to it. There are many objections against
such an interpretation; they will be set forth later on. On the other
hand, emphasis will be laid on the advantages of employing the
invariant rest mass.
A reasonable relativistic interpretation of mass should, like all
relativistic mechanics, in the final analysis, rely on four-dimen-
sional concepts. Although in many cases academic instruction is
so concise that the introduction of four-dimensional concepts is
impossible, we cannot forget that the construction of relativistic
mechanics (Chapter 5) inevitably requires the introduction of a
four-dimensional world. If we are forced to restrict ourselves to
three-dimensional formulations of relativistic mechanics, then in
interpreting its results we must go back to its very sources.
Let us recall briefly what we did at the beginning of Ch. 5. We
defined the 4-momentum ~ector a! the product of the 4-velocity and
a scalar, the rest mass: P = mV. As for the equation of motion,
the derivative iP/d-c has entered its left-hand part. It can be seen
from this (see also § 5.1) that the relativistic factor y under the
derivative sign in (5.37b) appeared because in 4-space-time we
employ invariant proper time inste!d of non-invariant coordinate
time. The first three component; of P include simply the first three
components of the 4-velocity V, which bear no relation to dyna-
mics. Thus, the factor y refers to the properties of 4-space-time, not
to the internal state of the particle.
When we introduced a scalar, the rest mass, then in 4-space the
quantity obtained exact transformational properties; in other words,
it is at once possible to indicate the law according to which it
changes in passing from one inertial frame of reference to another.
The rest mass is a scalar, i.e. invariant. This is a very important
P
point. Since 2 = - m2c2 (see (5.47)), it is apparent that the rest
mass is proportional to the square of the absolute value of the
energy-momentum 4-vector of a particle.
As in classical mechanics, we want to associate the mass with
the properties of the particle itself, in which case the only reason-
able method of introducing the mass is in terms of the rest mass.
One could, of course, say that a particle's acceleration to relativis-
tic velocities causes changes in its internal properties, notably
mass. But even if we do not touch the particle and introduce an-
other inertial reference frame, the particle's equation of motion will
remain (5.37b). Thus, taking "relativisti.:: mass" at face value, it
352 Su.pplement
increases for no physical reason. Such a result is hardly satisfac-

tory.
There is no need to determine the "relativistic mass" experi•
mentally. Only microparticles actually attain relativistic velocities,
and the rest mass is sufficient to identify them. It is easily found
if we determine the particle's energy and momentum from equation
(5.50):
m2c2=~__:p2.
This is just what is done in high-energy physics.
But perhaps the "dependence of mass on velocity" can be veri-
fied directly? We should first note that no unique dependence of
mass on velocity follows from the mechanics of special relativity.
As pointed out in § 5.3, the essence of the matter is that, unlike
Newtonian mechanics, in relativistic mechanics the directions of
acceleration and force do not, in the most general case, coincide.
In Newtonian mechanics a body's mass can be determined from
the ratio of the magnitude of the force to the magnitude of the ac-
celeration it imparts to the body: m = F/(dv/dt). If we similarly
determine the mass in relativistic mechanics we arrive at the mass
tensor (regarding tensors see Appendix I, § 3). Indeed, let us
rewrite (5.38) in the form (a= I, 2, 3)
m ~~o =~ [Fo-*{F11 v11)J=+(oa11 - v:~ 11 )F1 ,

whence we see that the acceleration is a linear vector function of
the force, the factors of which (i.e. the tensor components) depend
upon the velocity of the body. These factors define the inverse mass
tensor:
m;;ll' = ~" (oall- o:~~~ ).

The appearance of the mass tensor has a simple physical mean-
~~~~c~~o~sm~F~~~d~r~~ t~~d a~h! 1 e\;eal~~Y~Ae~~~ti~fe~~ ~~focTt~t~:~
presents a kind of preferred direction. For simplicity's sake we
direct axis I along the velocity. Then va = 0 1av, where u is the
absolute value of the velocity. We have
m;;J = ~" ( 0011 -..;. 0100111 ).

or
Supplement
We find the mass tensor according to the conventional ruiEl (the

co-factors divided by the magnitude of the determinant):
m, 0 =mv'l ~ v~' ~I·

0 0 y- 2
whence it is obvious that the principal values of the mass tensor
mall are one "longitudinal" mass and two "transverse" mflsses.
But the "mass tensor" appeared solely from the desire to 1ntro·
duce the "dependence of mass on velocity"; the appearance of sut;h
a tensor is obviously unjustified: the rest mass is quite adequate
for interpreting any results.
Thus, the answer to the question regarding the "experimental
detection" of the dependence my is that indiddual components of
the tensor mr.~ll can be determined. In particular, the "transverse
mass" is easily found in the motion of a charged particle in a
magnetic field (sec § 5.5); the "longitudinal mass" is obtained in
the motion of a charged particle without an initial \·elocity in a
uniform electric field. But if we speak of all experiments with rc·
lativJstic particles in general, the simplest thing is to say that they
confirm the relatiYislic equation of motion. In the two considered
special cases when the equation of motion resembles the Newto-
nian (the directions of acceleration and force coincide) the equa·
lion of motion indeed looks as though the change in mass is due
to the velocity. It is, however, different for the two cases. In all
other cases the equation of motion is substantially different from
the Newtonian. Without taking this into account one can come up
against some paradoxes (see § 8.2).
If we employ only the rest mass in relativistic mechanics, it is
significant from the methodological point of view that the con·
cept of mass introduced at school remains the same. The concept
of rest mass is "visualized" in the usual way from Newtonian me-
chanics; it is the same mass that enters the relativistic relation-
ships of dynamics, though it can no longer be defined as the ratio
of force to acceleration, but it can be defined according to (5.68).
The rest mass is simply the mass used in Newtonian mechanics.
Indeed, at ~ <t: 1 equation (5.37b) becomes (5.37a). insofar as
in this case y ~ I. And the determination of mass in Newtonian
mechanics presents no difficulties. It is more important to stre;,s
that in relativistic mechanics the properties of space-time come
into play and the laws of mechanics change, but we retain the in-
variance of the rest mass as a characteristic of a particle.
Sometimes attempts are made to link the increase in the energy
of a particle (or a system) with an increase in mass. This is also
a redundant interpretation. The dependence (5.46) shows that all
354 Supplement
forms of energy increase equally when a particle (system) is exa-

mined not in its proper frame of reference.
The transformational properties of "relativistic mass" are also
highly unsatisfactory. "Relativistic mass", which is proportion<il
to the energy of a particle, should transform as the fourth compo-
nent of the energy-momentum 4-vector. As distinct from it, the
rest mass is, as mentioned before, an·im·ariant which, like charge,
characterizes an elementary particle. II is sometimes pointed out
that the rest mass can change (see § 5.6), whereas the "relati-
vistic mass" is always conserved, as long as the energy conserva-
tion law holds. But at the same lime the conservation of relativistic
mass yields absolutely nothing in comparison with the law of con-
servation of energy: it is simply a corollary of that law. Thus, the
law of "conservation" of relativistic mass is a redundant equation.
The introduction of the relativistic mass of particles and the
law of its "conservation" leads to the introduction of a "photon
mass", hv/c 2• In § 7.6 we specifically dealt with the inexpediency
of employing this quantity.
As is known, one can take as the primary principles of the spe-
cial theory of relativity, not Einstein's two postulates but, for
example, his first postulate and the mass-velocity dependence [32J.
Formally, special relativity can be developed on such a basis. But
this, of course, does not add clarity to the physical meaning of re-
lativistic mass. It is worth emphasizing that Einstein's postulates
possess a clear advantage over other possible postulates since they
permit a direct physical interpretation and explicitly stress rela-
tivistic features in determining the coordinates of an event.
Sometimes it is pointed out that many eminent physicists in-
troduced the relati\'istic mass. This is hardly a potent argument,
the truth being that most leading physicists were against it [9, II,
34, 35J. It is of inte-rest that after lengthy discussions of relativistic
mass Robert Feynman wrote that, strange as it may seem, the
equation m=mo/.Yl- v2/c 2 is rarely employed in practice. In-
stead, !her; are2 ~wo ir;eplaceable relationships which are easily
proved: g-- P c· =Moe and Pc = f!v/c.
Summing up, we can say that the im·ariant rest mass possesses
indubitable ad,·antages, whereas relativistic mass is a source of
numerous misunderstandings, whilC" adding nothing of substance.
V. Non-inertial frames of reference. The special theory of re-
lativity and the advance to gravitational theory (the general
theory of relativity). The laws of dynamics help us single out
inertial systems among all other possible frames of reference. We
have defined inertial systems as those in which all three of New-
ton's classical laws hold. Newton's third law explicitly stresses
that force is a consequence of interaction of bodies. In non-inertial
Supplement 35&
systems it is no longer possible to preserve all three laws. If the

second law is retained we must introduce forces which do not
satisfy the third law: inertia forces. Let us examine two examples
to recall how this is done.
In the STR a reference frame is a rigid body; the most general
motion of a rigid body is a combination of translational motion
and rotation. Arbitrary motion of a non-inertial frame of referent~
relative to an inertial system is a composition of accelerated trans-
A
K~.
ma ,
-/
m
/(
- ---- --
Ia) ro>
Fig. S.7. (a) A buggy \\.ith a suspended bead rofls down an inclined plane with-
constant acceleration a. In steady motion the thread of the pendulum is deOected
somewlwt from the normal to the inclined plane The Ioree of gravity and ten-
sion of the thread combine to yield a resultant imparting to the bead the accele-
rution, a, needed lor its motion together with the buggy This is the reasoning
of an observer in the mertial system K connected with the "stationary" inclined
plane. In this system the law of Newtonian d)namics holds: acceleration is
caused only by forces (b) The same buggy and bead ar<! eumined from the
point of view of system K' connected with the buggy, 11 hieh is moving wi\11
acceleration This IS a non-inertial system and the inertia Ioree -ma must be
introduced. The bead is at rest relative to the buggy. hence the resultant of the
three forces actmg on it- gravity, the tens1on of the thread, and the inertia
force- must vanish. It will readily be ob~en·ed that the angle of deOection of
the thread from the normal to the plane of the buggy is the same as in ca~e
(a). as it should he.
lational motion and rotation (uniform or accelerated). Uniform

translational motion leaves us within the confines of inertial sys-
tems. Our examples refer to accelerated translational motion and
uniform rotation.
Example I. A buggy of mass M is rolling without friction down
an inclined plane. Suspended from a bracket on the buggy is a
heavy bead of mass rn. Determine the angle between the thread
supporting the bead and the normal to the inclined plana-
(Fig. 5.7) for the case of steady motion.
(a) Reasoning from the point of view of the inertial coordinate-
system K (connected with the inclined plane). The buggy is mov-
ing with uniform acceleration a= g sin a directed parallel to the
Supplement
inclined plane. If the bead is at rest relative to the buggy it must

be subject to the same acceleration. But acceleration is due to
force, and acting on the bead are only two forces: gravity and the
tension of the thread. For them to yield a resultant parallel to the
jnclined plane they must be at an angle. Knowing the direction of
acceleration of one of them (the force of gravity) it is simple
graphically to construct the direction and magnitude of the second
Jorce (the tension of the thread).
M +:•'r-mw'r
mg mg
w
""(a) 0 {h)
ll'ig. S.8. (a) A suspended bf'ad rotates together With the turntable of a cenlrl·
h.:gal machine. In steady motion the thread of the pendulum is dellected some-
what from the vertical away from the ax•s of rotation Compounded. the tension
of the thread and the force of gravity ~·ield a resultant force which g-ives the
bead the centripetal force equal to mro~r and directed towHd the rotation allis,
n('eded to mal.e the bead rotate together with the turntable. This IS the reason·
ing of an obsen•er in an inertial s~·sh!m K located outside the turntable. In
~~i~e~Y~~e'OJ~"J~~~o~~n5J{n°;doi 31~ oh~·1r~~~ I'~ nf~r~~= c(b)1rA~t~b:~~\e~~r ~~ 0 01~s t ~~~:
table, that is, in tf:C! non·inerlial s~·stem !<.', will describe the same phenomenon
t~~:;:~~~t 1 ~e t~:r:· B5Jt'~~:\ t~her~C!~1 ~e i~h~!f' r~~~~;~~n~e 't~~~~~~ o1 1~~~~~~~~i~
the force of gra\'ily, and the mC!rlia force. equal to -mro 2r CompoundC!d, they
yield the resultant force equal to zC!ro. From the reasoning (b) it follows that
angle a has the same! \·alue as in (a).
(b) Reasoning from the point of view of the non·inertial coor-

dinate system K' (connected with the buggy and the bead). In
this system the bend is simply at rest, hence the resultant of all
the forces acting on it is zero. But in addition to the force of
gravity and the tension of the thread we must take into account
the inertia force -ma. It will be readily observed that we arrive
at the same result.
Example 2. A bead of mass m suspended on a thread is plnced
()n a centrifugal machine rotating at constant angular velocity O)
(Fig. 5.8). The bead is at a distance r from the rotation axis. 0~
termine the angle of deOection of the thread from the vertical.
(a) Reasoning from the point of view of the inertial coordinate
system K connected with the "stationary" stand of the centrifugal
machine. For the bead to move together with the thread it must
.experience a centripetal acceleration mw 2r. For this the thread
Supplement 357
must deviate from the vertical; then the resultant of the force of
gravity and the tension of the thread for a given deftection from
the vertical can make for the required centripetal acceleration.
(b) Reasoning from the point of view of the non-inertial coor-
dinate system K' connected with the turntable. In this system the
bead is at rest, hence the resultant of all the forces acting on it is
zero. In addition to the force of gravity and the tension of thl!
thread we must introduce the centrifugal inertia force -muh. The
angle of deftection of the thread from the vertical, a, is, of course,
in both cases the same.
These two examples show how inertia forces can be used to pre-
serve Newton's second law in non-inertial coordinate systems.
These examples do not include certain other types of "inertia
forces", but their essential features are apparent from them. Inertia
forces are proportional to the "inert" mass of a body; they are
either constant over all space (Example I) or they increase in-
finitely with infinite recession from the axis of rotation (Example 2).
Galileo already knew that all bodies on Earth fall at the same
rate, i.e. that the force of gravity imparts them the same accele-
ration. But inertia forces possess the same property. Thus, mate-
rial bodies react identically to inertia forces and gravity forces.
And another peculiarity of gravity forces is known from experi-
ence: there is no shielding from them (it is, in principle, possible
to get rid of all other forces). That is why no direct experimental
venfication of Newton's first law is possible on or near the Earth.
Newton himself pointed out that to verify that law one would have
to get to a place where there are no gravitational fields; that is
why it was stressed in Chapter I that Newton's first law is a
postulate.
Einstein's gravitational theory possibly originated when he got
the idea of the equality of all frames of reference. It seems to con-
tradict everything discussed in this book, which has repeatedly
emphasized the special role of inertial frames of reference. But let
us not be in too great a hurry.
If it is impossible to get rid of either gravity or inertia, we can
try to regard inertia and gravity as different aspects of the same
phenomenon. Then Newton's first Jaw must be formulated diffe-
rently. The first part of Newton's statement of the Jaw remains the
same: free motion of a bOdy is motion with no forces acting on the
body, gravitation being excluded from the category of "force".
Formerly, according to Newton, free motion meant uniform motion
in a straight line. Now, according to Einstein, free motion is iner-
tial and under the action of gravity forces. Gravitation is no longer
a force. Now the action of forces is considered only when a body's
motion is deftected from free motion, which in the Newtonian
schem~ was called free fall. According to Einstein, inertia and
12-91
358 Supptemenf
gravity together condition "free" motion, they constitute its "back-

ground".
Of course, free motion in the Einsteinian sense is by no means
in a straight line. In Euclidean geometry (on which Newtonian
mechanics rests) a straight line is the shortest distance between
two given points (or, as mathematicians call it, a geodesicalline).
We shall have to recall this a little later.
Let us get back to the conclusions obtained at the beginning of
this Supplement. Passing to non-inertial reference frames simu-
lated the appearance of inertia forces proportional to the inert
mass of the body. If we recall that the gravitational mass and
inert mass are equal (or proportional), it becomes apparent from
the first example that passing to a reference frame in translational
motion, but with acceleration, simulates the appearance of a uni-
form gravitational field of magnitude -ma. From the second
example it can be seen that going over to a uniformly rotating
frame of reference also leads to the appearance of a field of force
proportional to the body's mass. In the general case, it can be
asserted that passing to non-inertial frames of reference simulates
a gravitational field. These fields have a peculiarity that distin-
guishes them from "genuine" gravitational fields: they do not
vanish at infinity, but they do vanish in passing over to inertial
We have seen that a transition from one inertial frame of re·
ference to another does not affect the square of the interval between
events
(S.V.I)
In passing from an inertial to a non-inertial reference frame,
ds 2 changes its general form. Indeed, let us consider two examples
of passing from an inertial to a non-inertial frame.
Example I. A coordinate system K' is moving in a straight
line relative to K with uniform acceleration a:
x=x'+at2/2, dx=dx'-at'dt';
y~y', dy~dy';
z=z', dz=dz';
t~t', dt~dt',
If K is an inertial system and

ds 2 = c2 dP.- dx 2 - d!f- dil,
then in system K'
ds 2 = c2 dt' 2 - (dx'- at' dt'f- dy' 2 - dz'2 ,
ds~=(c 2 -a2t'2)dt'2 - 2at' dx' dt'- dx' 2 - dy' 2 - dz' 2•
Supplement 359
Example 2. A uniformly rotating coordinate system (the an-

gular velocity of rotation is Q). From equations (A.I.IO) of the
Appendix I we get (9 = Qt)
x=x' cosru- !I sin QJ,
y =x' sin ru +
y' cos Qt.
It is not hard to find that ds 2 transforms to the form
ds 2 = (c 2 - Q' (x' 2 + y'')] + 2Qy' dx' dl'- 2Qx' dy' dl'-
d/ 2
- dx 12 - dy'2- dz' 2•
It can be shown that in either case no time transformation can
reduce ds2 to the algebraic sum of the squares of the differentials
of the four coordinates.
Thus, in the general case, going over to non-inertial systems
changes the expression for the (invariant) interval between events,
and in such a way that it no longer reduces to the "Galilean" form
(S.V.l). Let the metric of 4-space be written down in the general
case as
(S.V.2)
where g 1k depends upon all four coordinates and summation is
assumed over the indices i and k. The difference between the
4-space metric (S.V.2) and the Galilean metric (S.V.l) is, ac-
cording to Einstein's ideas, due to gravity. Thus, the difference
between g,k and the Galilean values (g00 =c2, g 11 =g 22 =g33 = I}
reHects the existence of gravitational fields. But gravitational fields
are associated with matter. Thus, the geometric properties (me-
tric) of space-time are by no means invariab1e and they depend
upon the physical objects within it. Knowledge of the space-time
metric makes it possible to answer the fundamental questions that
usually interest physicists. The question arises: how can these glk 's
be found? Einstein was able to write a set of (non-linear) differ-
ential equations in partial derivatives which must be satisfied by
ten values of g,k. These values depend upon the distribution of
matter and electromagnetic radiation. Einstein's equations have
so far been solved only for a few special cases.
Let us brieHy sum up. Passing to non-inertial reference frames
causes the appearance of metric coefficients g,k differing from the
Galilean and simulates the appearance of a certain field of force
proportional to the mass. It can therefore be assumed that the
values of g,k reHect the existence of a field of force similar to a
gravitational field. But a similar assumption is made with respect
to "true" gravitational fields, which remain in going over to iner-
tial frames of reference. It is obvious from this that if, in an iner-
tial frame of reference, the square of the interval is determined
12'
360 'Supplement
according to (S.V.l), it means that there are no gravitational

fields.
The difference between fields appearing in a transition to non-
inertial frames of reference and true fields is that the g,,.'s corres-
ponding to "true" fields cannot be reduced to the Galilean form
(S.V.l} by any time or coordinate transformations. From the geo-
metrical point of view, 4-space-time which includes gravitational
fields is no longer flat. It is warped. But here we should stop and
refer the reader to special literature (see, for example, [31]).
It is only necessary to note the following. In terrestrial condi-
tions we successfully apply special relativity, i.e. employ the in-
terval (S.V.l) along with the Newtonian theory of gravitation,
i.e. we consider gravity a force. Newton's gravitational theory is
explicitly non-relativistic, being a theory of action-at-a-distance.
Nevertheless, the results are excellent (for example, in calculating
the motions of celestial bodies). But Einstein's theory predicts that
that is precisely as it should be, in certain conditions, of course.
It is in "weak" gravitational fields (and within the solar system
all gravitational fields are weak; there is no need to cite exact
criteria) that Einstein's gravitational equations reduce to Newton's
equation of gravity (Poisson's equation). As to the velocities of
celestial bodies, they are always non-relativistic.
MAIN EVENTS RELATED
TO THE HISTORY OF THE STR
Publication of Galileo's Dialogue Concerning the Two Chief World

Systems- Ptolemaic and Copernican, 1632.
The first determination of the speed of light, ROmer, 1676 (5?).
Publication of Newton's book Philosophiae Naturalis Principia
Mathematica, 1687.
Discovery of the aberration of light, Bradley, 1728.
Doppler effect, 1842.
FoucauH pendulum experiment, 1851.
Laboratory determination of the speed of light, Fizeau, 1849, Fou-
cault, 1862.
Determination of the speed of light propagation in moving water,
Fizeau, 1851.
Development of the theory of electromagnetic field, Maxwell,
1856-1864.
Michelson's first experiment, 1881.
Publication of E. Mach's book Die Mechanik in ihrer Entwicklung,
1883.
ImprO\'ed Michelson's experiment, 1887.
Discovery of radioactivity, Becquerel, 1896.
Discovery of electron, J. J. Thomson, 1894-1896.
Kaufmann's study of the movement of relativistic particles in elec-
tromagnetic field, 1902.
Lorentz's papers devoted to the electrodynamics of moving bodies,
1892-1904.
Poincare's papers on relativism, 1895-1905.
Poincare's speech in St. Louis, 1904.
Einstein's paper "On the Electrodynamics of Moving Bodies", Ann.
d. Phys. 17, 891 (1905).
Minkowski's lecture on space and time, Phys. Zs. 10, 104 (1909).
Mam Contributors to the Development of Spare
and Time Science
Copernicus (1473-1543) Doppler (1803-1853)
Galileo (1564-1642) Maxwell (1831-1879)
·Repler (1571-1630) Lorentz (1853-1928)
Descartes ( 1596-1650) Poincare ( 1854-1912)
Huygens (1629-1695) Minkowski (1864-1909)
Newton (1643-1727) Einstein (1879-1955)
APPENDIX I
Here we present some mathemahcal data needed tor reading

this book.
§ I. The symmetric notation. The summation rules. When a
rectilinear orthogonal (Cartesian) system of coordinates is intro-
duced in the three-dimensional space, the unit vectors oriented
along the x, y, z axes are denoted by i, j, k respectively. The posi-
tion of any point in space is determined by the radius vector
+
r = xi+ yj zk whose components are the coordinates of the
point. All directions in space being equivalent, it is expedient to
introduce the symmetric notation and to write, for example, x., x2,
x 3 instead of x, y, z, and ml> m 2, m3 instead of i, j, k. Then the
radius vector of a point will be written in the following form:
'
r~x 1 m 1 +x2m 2 +x3 m3 ={;1 xam 11 , (A.I.I)
where L denotes the summation carried out in this case over a

running from I to 3. The summation sign can be omitted, having
once and for all stipulated that the two identical Greek letter in-
dices appearing in one side of an equation imply the summation
from I to 3.
The special theory of relativity employs a four-dimensional space
with four coordinates x 1 = x, x2 = y, x3 = z, x~ = ict. The ra-
dius \"ector and other vectors have in this case four components.
The summation rule is valid, but the summation is carried out for
Latin letter indices running OYer all values from l to 4. For
example,
So, the concise summation rule prescribes the summation over two
identical indices appearing in one side of an equation; in the case
of Latin letter indices the summation runs over the values from
I to 4.
Appendix I
Let us go back to the three-dimensional case. The radms vector

(A. I. I} can be written in an abbreviated form as r = X 0 m 0 and
arbitrary vectors a and b as
a= a 0 m 0 = aPmll = a'Ym'Y, b = bama = b 11m11 = b'Ym'Y.
The same equality is written out several times in order to show
that the summation indices are "mute", i.e. the summation can be
carried out over any letter index with the result remaining con-
stant.
To illustrate the application of the abbreviated notation let Ui
derive the formula for the scalar product of two vectors a and tJ.
On the one hand
ab = a 0 m0 b11 m11 = a 0 b11m 0 m 11 • (A.I.2)
Here we take into account that the summation rule pertains to two
identical indices; there are two summations to be carried out in
Eq. (A.1.2), each of which being performed over its own letter
index. On the other hand, unit vectors are mutually orthogonal;
therefore, each vector yields unity when multiplied by itself and
zero when multiplied by any other vector. Consequently,
I, a~~.
{ (A.!.3)
m 0 m~= O, a=F~-
It is convenient to introduce the Kronecker delta possessing exactly
these properties:
(A.l.4)
This symbol differs from zero only at a= ~. and any summation

involving this symbol yields the simple result: aailaP =a~. Indeed
+ +
a 0 01a = a10u a2012 a3013 = a 1.
Now we can easily complete the derivation of Eq. (A. I. 2)
ab = lla,bpmamp = a11bp0op = a11b11
to obtain the conventional formula for a scalar product of vee..
tors.
The identical indices over which the summation is carried out
may appear in a numerator and denominator of a fraction. The
summation rule is valid in this case as well. Let us write, for
example, the expression for a gradient of the function f and a
divergence of the vector a:
grad/= a~a ma= aa! 1 m1 + ::, m + ::

2 3 ma=
=~l +:UJ+ .fz.k, diva=:::= ~:x + ~:~~ + ~:z.

361 Appendix I
When a Greek (or Latin) letter index appears alone, it is meant

to be "free", taking on any value out of the three (or four) possible
ones. For example, bq. denotes one of the coordinates of the vec-
tor b, that is b1, b2 or bs.
§ 2. The transformation of coordinates In the case of a rotation
of the Cartesian system of coordinates. Let the radius vector of
the point M be expressed as r = Xama: .in the "old" coordinate sys-
tem. After the rotation of the coordinate system the radius vector
of the same point M will be written in the "new" coordinate sys-
tem as r =x~mp where xp are the coordinates of the point
in the system after the rotation and mG are the new unit vec-
tors. It is not difficult to define the relationship between the coor-
dinates of the old and new systems. Let us write down the equality
~xpressing the "conservation" of the vector r,
xnmn = x~m~.
and multiply its both sides by m;, i.e. an arbitrary unit vector
of the new coordinate system. The left-hand side yields
xumnm~ = xuau'i;
' we used the designation mutnV =cos (mn.
here /'--.
mV) =any: thus, a<lY
represents a cosine of the angle between the vector m<l of the old
system and the vector mV of the new one. On the right-hand side
~ve obtain the following chain of equalities:
Thus.
(A.I.5)
The new coordinates are expressed via the old ones linearly,
the coefficients being the cosines of the angles between the old and
the new coordinate axes. We have to find the coefficients of the
expansion of tbe old unit vectors in terms of the new ones. Let us
expand the old vector my via the new ones:
(A.l.6)
where a~~ are unknown coefficients. To find them, let us multiply
both sides of this equation by mV. Similarly to the foregoing for-
mula
(A.!.?)
We have obtained an obvious result: the coefficients of the expan·
sion of the unit vector m<l with respect to the new unit vectors are
the cosines a<lY" ·
Appendix I
The cosines of the angles between the old and new vectors can
be combined in the matrix:
aa11=(:: :: ::); (A.l.8)

a31 tls2 tla3
in the designation aa:ll the first index, a, indicates the row, and the
second, ~. the column of the matrix (A. I. 8). Thus, the transfor-
mation of coordinates is determined by nine coefficients aa:11. It is
known, however, that the position of any solid body (in our case
the coordinate system) whose one point is motionless can be defined
by three parameters (three Eulerian angles). Whence it is clear
that there are only three independent coefficients among the nine
coefficients of the matrix aa:ll· It is easy to find the requisite rela-
tions between the coefficients aafl· Indeed, the rotation of the coor-
dinate system does not affect the distance from the origin to any
point: r2 = ~ = x~1 • But x~ = 1111 a x;. To square this expression, we
have to multiply the sums whose summation indices should be
different:
x~2 = x~x~ = a 1311 x 11ay11Xy = a 1111ay11x 11 x".
But, on the other hand, this expression is equal to x~. This can
be only if
a~ 11Qy 11 = O~y (~, V = l, 2, 3). (A.I.9)
Although at first glance there are nine conditions here, these
equations do not change when their indices ~ and v are inter·
changed. Consequently, there are only six independent equations
here. Each of them represents a product of the ~th and vth rows
of the matrix (A.I.8). (The matrix row multiplication consists in
the summation of pairwise products of the respective elements.)
Eq. (A.I.9) means that the product of any row by itself is equal
to unity and by any other row to zero. Since the order of the row
multiplication makes no difference, e.g. the product of the first row
by the second row is equal to the product of the second row by the
first one, the number of independent equations is equal to six and
not to nine, as it was indicated.
The relations obtained are best illustrated by the example of the
rotation of axes in the coordinate plane (x 1, x2). In this case
x;=a 11 x 1 +a21 x 2, x;=a 12 x 1 +tlz2x2•
In accordance with Fig. (A. I),
a 11 =cos8, ~ 1 =sin8,
ali<;=- sin a. 'a22= cos e.

366 Appendix 1
and therefore,
x~=x 1 cos9+x2 sin9, x~=-x 1 sin9+x2 cos9. (A.I.IO)
Passing to the conventional notation x 1 = x, x: = y, we obtain
the well-known equations of analytical geometry:
+
x' =xcos9 y sin 9, y' =-.x sin 9 ycos9. +
We used these equations in the derivation of the Lorentz trans-
formation. We have obtained the equations of the direct transition
(from the unprimed to primed system).
~~~ :yl~!~~sil~[s~r~~fa~i~~ ~e t~!"C~~:e~?~~~~sele~r i~h~ ~~~~~~ng~~~r:n:~f::Joa~

is defined by one parameter a. The angles between the old and new coordinate
vectors are seen m the figure.
The reverse transition equations are obtained in much the same

fashion. We shall write them out together with the direct transition
equations:
x~ =a~ 0 x~. m~=a~ 0 m~,
(A.!. II)
x0 = a0 ~x~, m0 = a 0 ~m~,
at the same time
...--..
a 011 = m 0 m~ = cos ( m0 m~).
Surely, the reverse transition equations (see (A.I.ll)) can be
obtained automatically from the direct transition equations by ex-
changing primed and unprimed quantities and replacing the angle
9 by -9 (which corresponds to the rotation in the opposite direr-
lion).
How are the vector components transformed on the coordinate
transformation? This can be easily found by the same technique
that we used when deriving equations for the transformation of
coordinates. We may not do this, however, having noticed that the
coordinates are also the components of a vector, that is, of the
radius vector. Therefore, it is clear that the vector components are
transformed as the coordinates are, i.e.
(A~.l2)
Appendix 1 367
As we have already mentioned, the four-dimensional (pseudo-

Euclidean) space used in the special theory of relativity includes
formally one imaginary coordinate associated with time
x 1 =x, x2 =y, x3 =z, x=ict.
The Lorentz transformation corresponds to linear transforma-
tions in this space:
(A.I.l3)
and
o o -iBr)
I 0 0 I
0 I 0 ' I'= . I v• , (A.I.I4)
o o r -v•-cr
where V is the relative velocity of two reference frames.
The Lorentz transformation coefficients a 111 satisfy the following
conditions:
(A.l.15)
These equalities mean that the product of the Lorentz transforma-
tion matrix rows yields unity when a row is multiplied by itself,
and zero when a row is multiplied by any other row.
Let us calculate the determinant of the Lorentz matrix •
r o o -iBr I
I 0
0
I 0
o I
0
O =1''+(-iBI')(-iBf)=f'(I-B')=I
;Br o o r
(the simplest way is to expand the determinant into the first row
elements). The Lorentz matrix determinant proves to be equal to
unity. This means that we deal with the proper Lorentz transfor-
mation, i.e. we stay in the systems of clockwise triads of un1t
coordinate vectors and do not pass to the systems of anticlockwise
triads.
Let us write the direct and the inverse Lorentz transformation
for coordinates in full:
x 1 =r(x~-iBx~). x2 =x~. x3 =x;. x4 =f(x;+iBx;). (A.1.16)
x~=r(x 1 +iBx~)· x~=x2 , x;=x3, x~=r(x~-iBx 1 ). (A.I.l7)
• Some data on determinants can be found in § 6 oi this Appendix, This de-

terminant is calculated there as well.
Appendtx 1
As to 4-vector components, they are transformed as coordinates,

and, consequently, we get for the vector A
(A 11 A2, A3, A4) (tht!;
arrow over a letter denotes a 4-vector):
A~=a.~~ 1 A~t, A 1 =a 1 ~tA~t, (A.l.IS)
Al = r (A1 + iBA4), AJ = A2, A3::::::; A3, A~= r (A4- iBA1).
(A.l.l9)
A1 = r (At+ iBA~). A2 = A2. A3 = A.l. A4 = r (A~+ iBAI).
(A.l.20)
From Eq. (A.I.l5) it follows that the scalar product of two

4-vectors is invariant under the Lorentz transformation. Indeed, if,
A1 =a 111Ak, B 1 =a 1mB:n. then
The comparison of the second and the last link of the writ~ chain
of equalities proves the invariance of the scalar product AB.
§ 3. The tensors. Vector quantities are a particular case of ten-
sors, mathematical quantities of much more complicated nature.
Prior to passing over to them, let us point out the most essentiat
points in the definition of a vector. In a given coordinate system
a vector represents a directed line segment characterized by its
own coordmates. But since a coordinate system is chosen at will,
the vector coordinates are random. What is important, however, is
that using the vector coordinates determined in one Cartesian sys-
tem, we can find its Cartesian coordinates in any other system by
means of Eq. (A.I.5). These are the transformation equation-;
which define the vector. Thus, the vectorial nature of quantities is
revealed under the transformation of coordinates.
To clear up the concept of a tensor by means of a specific
example, we shall recall how the 'relationship between the electric
induction vector D and the external electric field strength E is in-
troduced in electrostatics. By and large, the relation D = D(E) =
= D(£ 1, £ 2, £ 3 ) is unknown. Let us expand D into the compo-
nents Da:
Assuming that in the absence of the external field (E = 0) the

vector D is also equal to zero (D(O) = 0) and assuming the ex-
ternal. field to be small as compared to electric forci!S acting be-
tween molecules of a substance, we can expand the unknown vector
Appendtx 1 ,.
function Din a Taylor series:
D 1= a~~~O) £ 1+ a~~~O) £2 + a~~~O) £3 + ... = a~~~O) Er. + ... ,

(A.I.21)
D2= a~~~O) Er. + ... , D3= a~~~O) Er. + ... ,
where the summation is carried out over the index ~- Due to the
smallness of the field E its components £~> £ 2 and £ 3 are also
small (in fact, this is a good approximation in the case of real
fields, except for the fields generated in laser beams), and we can
take only the linear terms, neglecting all the others. Let us denotl!
the constant quantities, which are the derivatives at the zero point,
as follows:
a:e~o = enr.·
Then the expressions obtained can be written in the form

D1 = eu£1 + e12£2 + e1aEa,
D2 = e21E1 + ~£2 + e.zaEa. (A.I.22)
D3 = ea.E. + t.a2E2 + eaaEa
or in the abbreviated form
(A.I.2?)
Using its components, we can easily compose the vector D:
D = enr.Er.mn. (A.I.24)
The relationship between two vectors expressed by Eqs. (A.L23)
and (A.L24) is referred to as a linear vector function; in other
words, the vector D is a linear vector function of E.
Using Eq. (A.I.24), we can construct the vector D from the
given vector E at each point of a dielectric in the coordinate sys-
tem where the coefficients eg:r. are known. The coordinate system,
however, is chosen randomly. The rotation of the Cartesian coor-
dinate system changes the vector components without varying the
vectors themselves. The question is how the coefficients ectr. must
change to maintain the relationship D' = e{.>.EAm~ in the ne\v sys·
tern, with D = D'. This means that there must be two different
expansions of the same vector:
D = ea~JEr.mn = e~>-E~m~. (A.L25}
The vector components and unit vectors are known to be trans-
formed according to Eq. (A.l.ll): EtJ=ar,AE~. mn=anj.!m~- Io
370 Appendix 1
accordance with these equations the left-hand side of Eq. (A.I.25)

can be rewritten in the following form (the right-hand side is lert
unchanged):
Comparing the coeffidents of EAm[,. in the left-hand and right-hand

sides, we obtain the transformation law for the coefficients ed:
(A.I.26)
The comparison of this equation with the coordinate transformation

law
(A.I.27)
shows that each index of eo:P is transformed according to the Jaw

corresponding to the coordinate transformation law. Eq. (A.I.26)
represents the tensor transformation law. The inverse transforma-
tion obviously takes the form t,aJ. = aA.I'~ae~p.
Here is the general definition of a tensor: if in a given Cartesian
coordinate system we have nine quantities eo:B which, under the
coordinate transforJilation .\ o. = a11aXa, are transformed according
to the formulae
(A.I.28)
these quantities form a tensor of the second rank. It is not difficult

to realize that vectors are transformed as tensors of the first rank.
The rank of a tensor (or, as it is sometimes called, the valency of
a tensor) is defined by the number of its indices. There are two
such indices in our case. Tensors of a higher rank are almost never
used in this book. A tensor is defined for the space possessing a
definite number of dimensions since its transformation law in-
volves the components of the transformation matrix. We have been
considering the three-dimensional space, and the Greek letter in-
dices ct, p,y, 11 were varying from one to three.
We would like to point out two distinctive characteristics of a
tensor transformation:
(I) The transformation law for the coefficients of a linear vector
function (tensor) is obtained as a condition for the invariant phys-
ical relationship between vectors.
(2) Any component of a tensor in the .. new" coordinate system
is a linear combination of all components of this tensor in the "old"
system.
Appendix 1 371
As a useful special case, let us note the transformation of a

three-dimensional tensor of the second rank Ta;p whose only non-
zero component T11 corresponds to a rotation in a plane. In the
primed system the following components will differ from zero {see
Eqs. (A.I.28) and (A.I.IO)):
r; 1 =a11 aull.!l. =ailn =Tn cos2 9,
(A.I.29)
r; 2 =a12au. 2T1.11. =a 11 a 12Tu = - T11 sin 9cos9.
We shall come across these equations many times in the future.

The special theory of relativity deals with the four-dimensional
(pseudo-Euclidean) space. We have already considered the trans-
formation laws for 4-vectors in this space. (According to our defi-
nition, a vector is a tensor of the first rank.) The transformation
rules for tensors in the 4-space remain actually the same, but the
number of tensor components increases to sixteen while the sum-
mation is carried out from I to 4:
(A.I.30)
A tensor is referred to as symmetric if its components satisfy

the relation A,k = Akl· Such a tensor has only ten independent
components. The energy-momentum-tension tensor of an electro-
magnetic field may serve as an example of a symmetric tensor.
A tensor is referred to as antisymmetric if its components satisf~,
the relation A,k = - Ak•· It is clear that the elements of this tensor
having identical indices (i = k) are equal to zero since the only
quantity which is equal to itself when taken with the opposite sign
is zero. Thus, an antisymmetric tensor has only six independent
components (in this connection it is sometimes called a six-veclor
tensor). The electromagnetic field tensor provides an example of
an antisymmetric tensor.
A tensor is referred to as unitary if Ark= O;fl. It is easy to se-e
that the tensor O;k retains its form under the Lorentz transforma-
tion in all reference frames. Indeed, if Ami= Oml, then according
to Eq. (A.I.30)
the last transition is performed in accordance with Eq. (A.I.S).

For reference we shall write the transformation formulae for
the tensor of the second rank T;k in the case of the Lorentz tran::; ·
formation, i.e. the formulae describing the transition from the frame
K' to the frame K (the inverse transition formulae are obtained
by changing the sign of V and by replacing the primed quantities
"'---------A'-'PP_"_odlx_T_ _ _ _ _ _ _ __
by unprimed ones and vice versa):
T 11 = a~l~ 1 + a 1~au (r;, + T~4) + ai~T~ =

= f 2{T; 1 - iB (T~4 + ~ 1 )- B 2 T~4 },
T,z = a11~z + at4~2 = r (T;z- iB~z)•
TL3=a~l~s+ al4r;a=f (T~3- iBr~;),
TL4 =aua4l~t +aL4auT~, +aua44T~4 +aua.wr;t =
= f2 {~~ + iBT~1 + B r;, -
2 iBT; 4},
T21 =a 11 T~ 1 +a 14 T~ 4 = f (T~ 1 -iBT~4 ),
T22 =T~. T23 =T~,
T24 =a4 l~ 1 +a 44 T~ 4 =f (T~4 + iBT~1 ), (A.I.31)
T31 =a 11 T~ 1 +a 14 T~ 4 = f(T~ 1 - iBT'34),
TJ2=T~, T33 =T~ 3 ,
T34=a4lr;t +a44r~4 = f (T'34 + iBT~t)•
T41 =alla4LT~L +al4a4LT~4 +aua4l~l +alfaUT~=
= f 2 + iBT~ 1 + B2T~4 - iBT~),
(~ 1
Ttz=a4l~z + at4T'42= I' (T~z + iBT;z)•
T43 = a4l~s + a«T~a = f (T~a + iB~s)•
T « =ail;t + a4,a44r:t + a~T~ =
= r {T~ 4 + iB (r. + r~ 1 )- B 2 T~ 1 }.
2
4
The tensor quantities appear more often than it may seem &t
first glance. We shall give some examples of tensors of the second
rank in th~ 4-space. The products of the components of two vectors
c(c,) and b (bk) form a tensor. Indeed, let us compose the expres-
sion A,k = c,bk. The transformation formulae for the vector com-
ponents are known:
(A.l.32)
Consequently,
coinciding with the tensor transformation law (A.I.30).

Appendix I 373
Let us demonstrate that the derivative of a vector component

with respect to a coordinate.,.is transformed as a tensor component.
Let us consider the vector b (b;) and its component derivatives. It
will be easier for us to write out two formulae: b1 = a1mb:n. x/ =
ax;
=a,.tXk, whence ax 11 =a111 to be used in the following chain of
equalities:
!!!..!__=~ax; =-_!_,(a 1 mb~) ax; =a 1ma111 Ctb7 · (A.I.33)

Ctx 11 ax, ax,. Ctx1 axk Ctx 1
The first and the last links of Eq. (A.L33) show that the deriva-
tive ()b,fiJx,. is transformed according to the tensor component
transformation law.
From Eq. (A.J.33) it is also 5een that the vector derivative can
be transformed in succession. First. we can pass from b:n to b1 ac-
cording to the formulae b 1 :d::a 1 mb~,. Then from differentiating with
respect to x/ \\'e can proceed to differentiating with respect to x..,.
or course, the tensor character of the transformation is retained in
this process although in a rather toncealed form That is exactly
how they sometimes do when trying to avoid tensors in transfer·
ma\ions of an electrom<Jgnetic field
§ 4. The invariance of a 4-divergence and d'Alembert's operator.
Let us demonstrate the invariancc of a four-dimensional diver-
gence and d'Alembert's operator under the Lorentz transformation.
We shall ''rite out the necessary formulae in a convenient form:
and for the components of the vector b:

b; = a,. 1b,., b11 = a,. 1 b~.
First, let us prove that a 4-gradient is transformed as a vector.
We shall consider the function <p=<p (x;, x2. xl. xD or, briefly,
<p = <p(x/). Let the coordinate tram.formation be defined by t!Je
formulae xl=xi (x~o x2, x 3 , x~)- Then according to the differentia-
tion rule for composite functions
Ct~:p a" ax; a"
8;;; = ax; a;;= akt ax~ •
In this way we obtain the vector component transformation law

(A.I.32) agam.
374 Appet~dlx I
Now let us demonstrate the invariance of the 4-divergence. The

following chain of equalities proves the fact:
divA= iJAk = aA~ ax; = ak, _!_, (A~a,ll) = ak,alll a A; =
axk ax, ax,. ax, ax,
iJA~ iJA~
=()II iJx~ = ax~.
Here we have taken into account tbt in accordance with Eq.
(A.J.l5) (LJI,(X.IIf = 6,,.
D'Alembert's operator applied to the function ¢1,
0 (J)=:tifll-f:- ~2~~ = ~~ + ~:~ + ~:~ -+ ~~~'

will be written in the form
~:: = ~:: + ~~x~ + ~:1 + a::c~ 2 •
The expression &<1>/cJx~. however, is the divergence of a gra-

dient. Indeed, let A,= cJ<Dtax,, then
• _,.
diVA=a;;
a (&<D) o'<D
a;; =a;r·
But the divergence of a 4-vector is an invariant of the Lorenl.t.
transformation, and consequently
(A.I. 34)
Any component of a 4-vector 4J(I!l.r.) can be taken in the capa..:-

ity oi the function Ill. Let us assume the respective components
of the two 4-vectors (j}(¢1,.) and ;(s,.) to be related in the frame K
by the following equation:
0¢1,.~ 02~"
.,, =-~k· (A.I.35)
Then it follows from Eq. (A.I.34) that the relation
(A.I.36)
is valid provided Eq. (A.I.35) is. Multiplying the left-hand and

right-hand sides of Eq. (A.J.36) by the constant factor et~m and
summing over k, we immediately get
0'<1>~-- ~,s~. (A. I. 37)
Appendix 1 375
since akm<Dk = <D:n. The intercomparison of Eqs. (A.l.35) and

(A.I.37) shows that in the frame K.' we have exactly the same
equation as in K. in which the primed and unprimed quantities are
interchanged.
§ 5. The convolution ("rejuvenation") of tensor indices. In the
~~n~~~o~!c~~~\~e:~1~e ti~e a~a~:e~r!0 ~e~!~~i~f t~hee ;:c~kngt r!nte~1l~~
operation consists in summing up the tensor components possessmg
lwo identical indices. It is noteworthy that such an operation giYes
rise to the invariant expression. In the case of tensors of a higher
rank the convolution results in the tensor's rank getting lower by
two.
We can demonstrate this property very easily. Let us write the
tensor component transformation formula
A 1"' =a 1 ma 111 A~ 1
and sum up the components A,"' possessing the identical indices
having put i = k. Then due to Eq. (A.I.l5)
AH= A11 + A22 + A33 + A 44 =a1 ma, 1 A~ 1 =llm1 A~ 1 = A~m·
Thus,
~+k+k+~-~+~+~+4 ~L~
Although tensors of higher ranks are hardly used in this boo!;:,
we shall sometimes deal with the results of their convolution. We
have seen that the differentiation of a scalar function, i.e. an in-
variant expression, leads to the formation of the \'ector gradient
iJCl>/iJx,. The differentiation of a vector leads to the formution of a
tensor of second rank iJ2fl>!dx,dxk. We haYe already seen that the
convolution of this expression results in the invariant (A.l.34):
02!ll 02!ll
oxt =ox?·
If a tensor of the second rank has the components depending on
coordinates, the differentiation of these components leads to the
formation of a tensor of the third rank. For example, from the
tensor f,"' we get the tensor of the third rank Of,~<tax,.
Let us perform the convolution of this tensor over the indices k
and l and see that it results in four quantities forming the compo-
nents of a vector. Indeed,
!.!J!.=.!_(a a f') ax~ =a a a a!;p =
axk Ox~ is kp sp Ox" IS kp km ox;,
ofsP of~m
= a,s«pm ox;, = als Ox~ • (A.I.39)
S76 Appendix f
Since the indices k, m, s are mute, we see that the quantitie:;

at,~~.tax~~: are transformed according to the vector transformation
law (A.I.32).
Also let us consider now how to obtain an im·ariant expression
from the components of tensors of the second rank. If the product
of components of two vectors forms, as \Ve have seen, a tensor of
the second rank, it can readily be sh9wn that the product of com-
ponents of two tensors of the second rank yields a tensor of the
fourth rank. Let the components of the tensor & be denoted by F, 11
and those of the tensor f by {1m· Their product T,ktm = F,~~:ftm is the
tensor of the fourth rank. Let us perform the convolution of this
tensor over the indices i and l as well as over k and m, i.e. let us
compose the expression
(A.I.40)
which represents the sum of the pairwise products of the respective
components. We shall make sure that this expression does n.)t
change on transition from one reference frame to another. This
result is easy to prove since the transformation rule for the com·
ponents F,~~. and ft., is known:
Putting i = l, k = m, we shall get
Flkflll =a,salrallpalltF;Pf;t = OJ)ptF;irt= F;pf:p' (A .I. 41)
The equality (A.1.41) provides the evidence of the invariance of

(A.I.40). Of course, the invariance of F?11 or f;11 represents a
special case of (A.I.40).
Since we have repeatedly made use of the Gauss-Ostrogradsky
theorem to treat the vectors resulting from the three-dimensiondl
convolution of a tensor, we shall write out the requisite formulae.
In the three-dimensional space this theorem has a bearing on the
transformation of a vector flow across the closed surface S to the
integral of the volume "P' enveloped by that surface, e.g.,
~DdS~\divDdl". (A.I.42)
s r
The same theorem takes the following form in the symmetric no-
tation:
~ Dat~ dS = ~ ~~= d"P', (A. I. 43)

s r
Appendix I
where ntt are the components of the normal to the surface ele-
ment dS. Applying Eq. (A.l.43) to the vector
A~= ~x~ •
we shall obtain
I ~;: d1' ~ s~ t.,n.dS.
7
(A.I.44)
§ 6. Some data on determinants. The dual tensors. I. Let us

arrange n2 elements, denoted by the symbol a,.t, where i, k take
on all values from I to n, in the form of a square table. Let the
first index i in the symbol a,,. denote the number of the row and the
second index k the number of the column of an element. Thus, we
shall obtain the square matrix formed by the elements a,,.:
From this matrix we can form the determinant
a, a., ... a,. I

D,=lal,.l= I il2~ .a~ .··.·. ~n ,
a, 1 a, 2 • • • a,,
which implies a definite operation performed over the elements a,h
that is the formation of the sum of n! elements a,,.. This sum can
be obtained as follows. Take the product of elements involving
the elements from different rows, e.g. the rows I 2 ... n:
(A.I.45}
or the product of elements involving the elements from different

columns, e.g., I 2 ... n:
(A.I.46)
where the values of the indices a, p, ... , t will be now defined. To

obtain the value of the determinant, let us compose the algebraic
~urn of the terms (A.I.45) or (A.I.46) differing from one another
by the indices a, ~ •... , t forming a certain permutation of the nat-
13-97
378 Append{JC I
ural sequence of numbers I 2 . . n m each term of the sum Th1s

means that the indices a, ~ •... , T have different values in each
term of the sum. The sum is taken over all the permutations of
numbers I 2 ... n, the total number of such permutations being
.equal to nl.
The sign "+" or "-" is ascribed to each term of the sum de-
pending on whether an even or an odd number of pairwise permu-
tations (transpositions) of elements is needed to obtain a given
.arrangement a Jl T from the natural sequence of numbers
I 2 ... n. The pairwise transposition consists, for example, in the
transition from the sequence I 2 3 4 to the sequence I 3 2 4 in-
volving the transposition of the digits 2 and 3. The number of
transpositions needed to accomplish the transition from the nat-
ural sequence to a given one is denoted by the letter r. So, ac·
-cording to the definition, a determinant of the nth order is written
in full as follows:
D,.=laal=l::: :: .::•l=a.,..f!I ..,. 1

(-I)'aala 82 ... a1 ,.=
ani an2 · · · a,.,.

a.,..B.,.. ·
L: (A.l.47)
where the sum is taken over all the permutations of indices

a ~ ... t' taking on various values from I to n. The two last row-.
of the equality indicate one of the basic properties of a determinant,
that is the equivalence of rows and columns. Of course, in expres-
sions of the forms (A.I.45) and (A.I.46) it is not obligatory to
1ake respectively the first or second indices arranged in the natural
sequence. But then the bringing of specific elements to the canoni-
cal form (A.I.45) or (A.I.46) would require the indices a, ~ •... , 'C
to be re-denoted. This new notation would be reduced to the trans-
position of the indices; in the determinant itself this would mean
the transposition of rows or columns. Hence it is clear that t11~
transposition of an odd number of rows (or columns) alters the
sign of the determinant while the transposition of an even number
of rows (or columns) leaves the value of the determinant un-
changed. Having fixed a definite value n of indices i, k, ... , s, we
can compose the sum of the terms of transposed indices a p . . -c,
each term being accompanied with the requisite sign, i.e.
a.,..B.,...
L (-l)'aafa~ ...
a13 ,
ApperJJo; I 379
~hose value will be equal to +D, or -D, depending on what

number of transpositions, odd or even, is needed to obtain the se·
quence i k ... s from the natural sequence I 2 ... n.
This is how the determinant of the third rank D3 is written in
full:
a" a., a 0 ,1 . .,..a.,..v
0 1 = Ia:!l on a-n= l: (-I )r a1(1a2Aa1v =
a31 a32 033
a11aWl33
•u-t.B-l v-11
+ a13a21aJ2
!312•
+ a1~2Ja11-
•JJI•
a1.1a.12a11- a11a21a32-at202t033·
13'!1• •t3l• •213•
2. The determinant calculation technique. Let us pick from th~

sum (A.I.47) all the terms containing a certain element a,., group
them together and take this element as a common factor for this
combination of terms. The coefficient of the element a,t thus ob-
tained will be denoted by A,. and referred to as an adjunct or co-
factor of the element a,11-. The co-factor of a given element is cal-
culated in accordance with i:l simple rule. In the determinant D,.
we cross out the row and the column which contain the element
a••. whose co-factor A,. is sought for. Having crossed out the ith
row and the kth column, we obtain the determinant D,- 1 of the
(n- J)th rank which is referred to as a minor ~~~~- of the ele-
ment a,~~::
llu=D,_r= a,_t,t a1-1.2 ••• al-1.--1 al-1./t+l ••• a,_._,

af+l.t a1-1-1.2 ••• a, •• .t-1 a,+l.ll+l ••• al+l.n
a, 1 a,2 ... a~ --r a,, 11+1 ••• a,,
The co-factor A1k may differ from the minor ~ •• only in sign:
Au=(-1)'+-t ~~--
Each element has its own corresponding co-factor, but a givell

element does not enter every term of the sum (A.I.47). We can,
however, select a definite number of elements of a determinant
which together with their co-factors permit the value of the deter-
minant to be found. In fact, there is a theorem staling that th~
determinant can be expanded into the elements of any row or any
column as follows:
13'
380 Apperullx 1
the summation over k is not carried out here while k itself may
take on any value from I to n. If we compose the sum of the pro-
ducts of the elements of any row (or any column) and the minors
of another row (or another column), it will be equal to zero:
·-·i:. a(l.tA(I 1 =0
The last two formulae are combined into one:

(a-=Ft).
a~ I aa.tAat =DO,,.
For example, let us calculate the determinant of the Lorentz

transformation matrix, having expanded it into the elements of the
first row:
DL=
l•u a., a.,
a21a22UzJ~ 4 ""II
a 3,t1J2a33 a 34
=
r0 1o0o mr
0010
0 I =
au a 42 au a 44 -iBf 0 0 r
Io 0o 01r
=I' 0I I 0 -iBf I0 0 01
0I I = f2 -B2 f 2 =1' 2 (1-B2)=1.
-iBr o o
The reader can easily ascertain that the multiplication of the
elements of the first row by the co-factors of the elements of other
rows yields zero.
3. Let us introduce a fully antisymmetric unit tensor of thot
nth rank. This is the tensor 6ctll ... , whose components alter sign
when any two indices are interchanged and all components differ·
ing from zero are equal either to +I or to -I. Any component of
an antisymmetric tensor 6a:ll .. , whose two indices are equal turns
into zero (the transposition of two such indices alters the campo·
nent's sign due to the antisymmetry of the tensor, but at the same
time we get the same component; thus, only zero can be equal to
itself when taken with an opposite sign). Thus, only those campo·
nents of the tensor 6a:ll .. , differ from zero whose all indices are
different. Suppose 6 1 ~ • , = I; then all the components differing
from zero are equal to +I if the arrangement aj3 ... t" is obtained
from the sequence I 2 ... n through an even number of transposi·
lions. If the number of such transpositions in the arrangement
aj3 .. "T is odd, the component 6all ... T is equal to -I. Making use
of a fully antisymmetric unit tensor, \Ve can rewrite the expre~·
"Appendix I ..,
sian for the determinant D,. as follows:
where the summation over the pairs of indices ap ... '~" is implied
this time.
In particular, we can write the determinant corresponding to the
Lorentz matrix:
4. Now we shall deal with the 4-space of the STR- First of all,
it should be pointed out that we have defined a fully antisymmetric
unit tensor OpypJ£ without proving it to be a tensor. We must
make sure that the components of this tensor have the same values
in all IFRs, i.e. under the Lorentz transformation. This can be
done quite easily. In accordance with the tensor component trans·
formation rule
But according to what was said in item I of this section the

.quantity on the right-hand side is equal to DL6,~r 1 m, i.e. is equal to
±I depending on the number of transpositions needed to obtain
~~~ atheans~e:een~o~~~n~~t~ ~~e ;::u~~R~n~h~h~~;Pe~~:n:~a~fo~m~
tensor also do not change on transition from the left system of
-coordinates to the right one (i.e. when one or three spatial coordi·
nates change sign). According to Eq. (A.l.30) the components of
this tensor had to alter sign in this case. Consequently. O~YPJ£ is not
a tensor but a pseudotensor; its components behave in a different
'\Vay, as compared to tensors, when a coordinate alters sign (is
reflected). In all other transformations the behaviour of these com-
ponents coincides with that of the components of a tensor.
5. The cross and the mixed product of vectors in a three-dimen-
sional space. These questions are treated in order to have a good
analogy when considering some of the quantities in the 4-space of
lhe STR.
Consider three unit vectors m~o m 2, m3 of the orthogonal
Cartesian system of coordinates. Compose the cross product of any
pair of these vectors [ma:m~); as a result, the third vector will be
.obtained with the "plus" or "minus" sign, depending on the order
of co-factors in the cross product. This cross product can be easily
written via fully antisymmetric unit tensor of the third rank:
382 Appendix I
Now we can write the cross product of the vectors Gt = a 1 o:m~

and a2 = a211m13:
[aJ~) =[a lama. a2 8m 6 ) =lltoll2fl [mom11 ] = Oa11vlltoll<J'Imv. (A.I.48)o
It is seen from Eq. (A.I.48) that m., has the coefficients formed
by the products of vector components and convoluted with the
tensor lletBv· Let us rewrite Eq. (A.I.48):
lata.2] = 6Clflylltoll~m., = -4- (lloeyllJollt:l + o,.o.,all3ll2ol m., =
=+ lloflv llltalllJ- ll20llJ;Jl m., = ~6a 8 .,Ca~m.,.
In the third link of this equality a second term is added whicb
is equal to the first term with the mute indices a and ~ inter~
changed. Here in the third link we take into account that Ollav =
= - Octllv· In the fourth link Octlh is taken out of the parentheses.
The antisymmetric tensor, thus formed in the parentheses, is de-
noted by Call = a 1aa2p- a 1pa2a Consequently, the cross product
[a 1a 2] is a vector whose components are obtained from the anti·
symmetric tensor Cap according to the formulae
Cv = 1- 6a yCa
6 11 •
The vector C(Cy) is said to be dual to the antisymmetric tensor

Cal'l· This means that the vector C is orthogonal to two vectors at
and a2 defining a two-dimensional plane. The orthogonality can be-
proved analytically at once:
Ca1 = Cyaly =+ 6a 6yCaf1alv =+ llaflv {atall:! 11atv- al~!fZ2Gatv} = 0.
This expression is equal to zero since both the minuend and the
subtrahend are the determinants with two equal rows and such
determinants are equal to zero. In much the same way it is proved
that Ca2 = 0. In geometrical terms the norm of the vector C is
equal to the area of a parallelogram constructed on the vectors a,
and a2.
:,l
The mixed product of three vectors a 11 a2, as is denoted by
(at. a 2, as) and defined as follows:
(a" a,, a,)~ a, [a,a 3) ~ a,.m,6.0,a,.a,0m, ~,6.,a,.a,.a~:~·::
<= a 1 A6Aya2G6t~pv <= 6~;~ 6 yatvO:aa:JIJ <= D 3""" ~~ a:!:! Q:s •

a 31 a$2 a33
Here o,'Y is the Kronecker delta (A.I.4).

Appendix I
In geometriLal terms the mixed product of three vectors defines

the volume of the parallelepiped constructed on these vectors. This
volume is obtained with the "+" or "-" sign depending on the
order of the vectors a 1, a 2, a3 in the mixed product.
6. The dual tensors. Let two 4-vectors a 1 and a 2 be given in the
4-space. Then the projections of the parallelogram area on the co-
ordinate planes (x, x~) are defined by the antisymmetric tensor
~.~ = a 1,a 211 - a2 ,a, .... In the 4-space each area element ~.~~ can be
brought into correspondence with another normal area element 11 S;
such that all the straight lines lying in it are normal to all the
straight lines of the initial area element. If the element S~t ortho·
gonal to£," has the same area ass..... the element s;,.
is called dual
with respect to 5, 11• It can be shown that
(A.I.49)
With the aid of this formula any antisymmetric tensor can be

brought into correspondence with its dual tensor. The dual tensor
f; 11 is, in a certain sense, equivalent to the initial tensor f;t. We
bave seen that the second group of Maxwell's equations takes the
simpler form when written via the, dual tensor F;,..
The sum of the
products of the antisymmetric tensor components by their dual
-co-factors yields a pseudoscalar
F p• =i-OlklmFI,.Fim = i-a,aatbalcamdOilbcdat,a.tsF;sallamnF;" =
= + O,.,Ot.sbctOdnOil~cdF~sF/.n = +OrstnF~sFln. = F:sF;,.

In the-se equalities we made use of the definition (A.I.49), the
tensor transformation formulae (A.I.30) and the properties of the
Lorentz matrix coefficients (A.I.IS). It can be easily shown that
F~" = F1"F 111 is an invariant:
F,"F'" = a,"a"b~batca.tdF;d = O"cObdF~,f;d = F~~F~b·

These two invariants were used in§ 6.5.
§ 7. The stress tensor. The stress tensor is introduced in con-
tinuum mechanics to characterize the force acting on a volume as
a whole \'ia the force acting on the surface confining this volume.
We obtained the expressions of this type studying the forces in an
electromagnetic field; it is useful to consider this problem in me-
chanics where the physica I essence of phenomena is most obvious.
If an elastic body i&. subjected to deformation forces arise in it
tending to return it to the state of equilibrium. These force.'l are
Appendix f
referred to as the internal stresses and are caused by the forces of

interaction of the molecules of the body. The distinctive feature of
these forces is their "small radius of action"; in other words, their
inftuence is felt only over microscopic (atomic) distances. There·
fore, it is clear that when a certain volume r
is considered inside a
body, the forces acting on that volume reduce to the forces acting
across the surface confining that volume.
Indeed, let the force F act on a unit of volume of a body. Let us.
single out the volume r
within a body and consider the total force
acting on this volume. If the volume dr is subjected to the fore~
F dT', the total force acting on the volume is equal to
j P d:Y. (A.!. 50)
The various parts of the considered volume interact, but in ac·

cordance with the law of equal action and reaction the forces of
interaction cancel out and yield a zero resultant. Thus, the totat
force acting on the volume r arises due to the forces exerted by
the surrounding parts of the body. But as we have already men-
tioned, these forces act only across the surface confining the con-
sidered volume. Therefore, the total force will reduce to a certain
surface integral. In particular, the ~th component of the force
j F 0 d:Y (A.I.51)
must also turn into a surface integral. However, this is possibl~

only if Fr. can be represented in the form
(A.I.52)
where T11 r. denotes the components of a tensor (only in this cas~

the convolution results in a vector). Then in accordance with Eq.
(A.I.44)
I
'I"
F 0 dr~ I dT"'
r
. ,(;
iiXdr~'Yr",n"dS.
II s
(A.I.53)
Having multiplied both sides of Eq. (A.L53) by mr, and carried

out the actual summation, we obtain
~ Pdr = ~ aa:~ m11 lr= ~ Ta 11nam11 dS. (A.I.54)

' ' s
Appendix f 885
obtained can be formulated as follows: if the force F acting on a

unit of volume can be represented as
P~ --.z;
araiJ
m 0, (A.I.55)
its action on the whole volume can be described as the action of 3

surface force distributed over the surfflce confining this volume,
~~f~~~2 in (~h~~e th~r~;e~!~~~ 0;e~~~ai!d~ :~s 8:h!h~o~~a~df~Ytheur:~~?ac0: ~~~~

ment dS; p~ is the force acting on the area whose normal is 11. (b) On deriva·
tion of the condition for the equilibrium of a dosed tetrahedron-shaped elemen-
tary volume. The outward normal is chosen as a normal to the closed telrahed·
ron surface. At the faces BOG, AOC and AOB the unit vectors of the normal
are equal to -i, - / and -lr respectively. The areas of the faces BOG, AOC and
AOB are equal to dScos(.i);), dS cos(,(y) and dS cos(n-:-%), respectively.
with the surface element dS, whose normal n has the unit vec-
tor components n<t, being subjected to the force
r. 0n.m 0• (A.I.56)
Let us discuss briefly the physical meaning of the stress tensor
-components. Let us return to the volume r
within the body expe·
riencing a deformation. The force acting on the surface element dS
confining the volume r depends on the value and direction of the
element dS, i.e. on the direction of the normal n relative to this
element. Let us denote this force by p,. dS, having pointed out that
its direction, generally speaking, does not coincide with the direc-
tion of the normal of the surface element dS (Fig. A.2a). The
vector p,. is the force per unit of area; it depends on the orienta-
tion of the surface element and is called the stress on the surface
element dS with the normal n. At each point of the deformed elas·
tic body any direction n has its corresponding stress vector p,.
ln each Cartesian reference frame it is possible to determine the
stresses p., pg, p~ acting on unit of area elements whose normals
Appendix I
eoincide with the coordinate axes. We shall demonstrate that the

stress relating to any area element dS with the given vector n
cpn be expressed through nine components of the vectors P~~:, p 11 , P.z.
these nine components taken together form the stress tensor To;tji·
Let an elastically deformed body be in equilibrium. We shaH
consider the infinitesimal tetrahedron OABC (Fig. A.2b) who.$e
inclined side has the area dS. Let the normal n of this side be
directed at ari acute angle to the x 'axis. Then the area elemen~s
cut out by the coordinate planes will equal dS cos (n:':c),
dS cos (;;,g) and dS cos (n~). The normals of these area elements
are oriented in the direction opposite to the unitary coordinate
axes i, j, k, so that the side BOC is subjected to the force
-P~~: dS cos (;.'x). The value of P~~: can be taken at any point of
BOC since this side is infinitesimal. In much the same manner, the
forces acting on the sides AOC and AOB tutn out to be equal to
-p 11 dS cos (;.'g) and -P.z dS cos (n-:-z) respectively. In equi·
Jibrium the resultant of the forces acting on the tetrahedron is.
equal to zero·
dS IPn -piC cos <n:'x) - p 11 cos (n-:'y) - P.z cos (,;:---z)J = 0,
wh.ence the sought for stress Pn is expressed via Px, p11 , pz.:
Pn = P:c cos (n-:'x) + p 11 cos (n-:'y) + P.z cos (n-:-2). (A.I.57~
Passing to the symmetric notation in Eq. (A.I.57), we can show
that we obtained a tensor. Indeed, n is the vector of the normal or
an arbitrary side with the components no;t and therefore, Pn =
= p1n1 + +
p2n2 p3n3 = po;tno;t. But po;t = po;tpmp where Po;ttJ are the
components of the vector po;t, and consequently,
Pn =· Pg.r,mpna, (A.I.58)
U is seen from Eq. (A.I.58) _that the nine components of the
vectors po;t are transformed as a tensor (cf., e.g., Eq. (A.I.24)).
§ B. The rectilinear oblique-angled systems of coordinates. Un-
til now we have used the orthogonal rectilinear system of coordi-
nates, but a transition to rectilinear oblique-angled systems of co-
ordinates makes it possible to illustrate the features characteristic
of arbitrary coordinate systems dealt with in this book.
Let us choose, as before, a set of straight lines, not orthogonal
this time, as coordinate axes and denote them by x 1 and x 2
(Fig. A.3a). Then let us mark ~ basis unit vector m 11 on each-
of these axes x11.
An arbitrary vector A can be expanded into the non-colinear
J¥ectors m 11 :
(A.I.59)
Appendl% I 387
The quanhl!es A~-' are the components ol the vedor A which are
()btained by means of the parallel projection of this vector on the
coordinate axes; according to the definition they are called the
contravariant components of the vector A.
The quantities
(A.I.60)
.are the orthogonal projections of the vector A on the coordinate
.axes and are referred to as the covariant components of the vee-
(b)
Fig. A.3. The definiton of the CO· and contravariant coordinates In a rectilinear
oblique-angled system of coordinates on the plane.
tor A. Obviously, these definitions can be retained for any number

.of measurements. Designating the scalar product of the basis vec-
tors
(A.I.61)
we get g~'v = ev1-1 and in the case of the rectilinear coordinate axes
,g~'v= const. The covariant and contravariant coordinates relate
to the same vector and are interrelated:
(A.I.62)
This equation defines the transition from the contravariant compo-
nents of a vector to the covariant components.
Then, let us define the quantities g~'v by the condition:
vp={)v {)v={ 0, V=#=-p.,
giJ.I>g u.• I' I, v=p..
Now let us construct the expression
~ A0 = ~gpaA 11 =~A a= A~'. iA.I.64)
The last equality defines the transition from the covariant to
contravariant components. Thus, we have obtained two fundamen ..
tal formulae of transition:
(A.l.66)
Appendix I
The determinant formed by the quantities g111 is denoted by g

(see § 6 of this Appendix):
Using the formula (p. 380)

L Km~~Acsl = 6111K
we immediately obtain
(A.I.66)
where A 11 v is the cofactor of the element g 11 v.

It i<; easily seen that for the orthogonal rectilinear coordinates,
when m 11mv = 61lv• g 11 v = .S"'v and A 11 =All, i.e. there is no differ-
ence between the covariant and contravariant coordinates. That is
why in the case of the orthogonal Cartesian system of coordinates
we simply speak of the coordinates of vectors.
By the definition the scalar product of two vectors A and B is
the quantity
AB~ (A"m 0 ) (B•m.)~ A"B•(m 0m.)~g••A"B·~ A.B•. (A.I.67)
The scalar product of a vector by itself defines the square of the
vector's absolute value or the norm of the vector:
A2 = g 11 vA"' A"= AvA". (A.L68)
Thus, the norm is the square of the vector's length. If the norm
of the vector is equal to unity, the vector is called normed or uni-
tary. If the norm of any non-zero vector is positive, the space is
referred to as the proper Euclidean space.
In particular, the square of the infinitesimal vector dr possessing
the components dxv and connecting two infinitely close points of
space is equal to
(A.I.69)
Let us change the system of base axes and pass to new basis
vectors ~directed along the rectilinear axes x'11 (Fig. A.3b). Any
new vector ~ can be expanded into the old basis vectors:
m~ = a~mv, (A.I.70)
wllere a~ are constant coefficient3' dependent on the concrete trans-
Qflder any tra~sformation provided Ja~ a

old vector m 11 can be expanded into the new ones;
*
f,ormation of oblique-angled axes. The quantity rrr.f, does not change
I= 0. Of course, any
(A.I.71)
Appendix I ,..
It follows from Eqs. (A.I.70) and (A.J.71) that
m~ = a!mv = a:a~A.m~; mA. = a~A.m~ = a~A.a~m11 • (A. I. 72)
from Eq. (A.I.72) the coefficients a~ and a:r are seen to be re-
lated by the following expressions:
(A.l.73}
where 0:: is defined in accordance with Eq. (A. I. 63).

The radius vector r drawn from the origin of coordinates to the
point M (Fig. A.3b) can be written in two forms:
(A.l.74}
Taking into account Eqs. (A.l.70) and (A.I.71), the last equation
can be rewritten in two forms as well:
(A.l.75}
whence follow the formulae for the direct and inverse transforma-
tions of the contravariant coordinates of the vector r:
(A.I.76)
Here is the definition of a vector: the vector A is the quantity
whose covariant components are transformed under a transition:
to a new reference frame in the same manner as the basis vee·
tors m 11 • The contravariant components of vectors are transformed
as the contravariant coordinates X". Let us find the formulae for
the transformation of the components of the vector A. For cova-
riant components
A11 = Am11 = Aa~A.m~ = a:A. A~. (A.l.77}
The inverse transformation formula takes the form
A~= Am:= Aa!mv = a!Av. (A.I.78}
On the other hand, precisely as for the vector r we can write
(A.l.79}
whence
(A.l.80}
and consequently
(A.l.81}
We see that the transformation formulae for the covariant anrl
contravariant component-s of a vector are different.
390 Appefldlx 1
Let u::. write the transformation law lm the quantity (A.I.bl ).

IA.I.82i
According to the definition, this is the transformation law for a

covariant tensor. The quantity retaining its value when the basis
vectors (A.I.70) and (A.I.71) change is referred to as an in-
variant. In the considered case of the rectilinear oblique-angled
axes a;and a~11 are constant values. Let us demonstrate the in·
\'anance of the distance between points:
ds' 1 = g~ 11 dx'"' dx'v= a_~ dx'uazdx'v gDO = gDO dxl d.x4 (A.I.83)
(see Eq. (A. I. 81)).

The invariance of the operator !i = - 0-'- can also be easily
axu ax~A
verified.
When, for diverse reasons, vectors are introduced, both covariant
and contravaria.nl components may be found among the compo-
nertts of the vectors. We shall quote two important examples. Fron1
Eq. (A.I.Sl) it follows that
(A.I.84)
whence it is clear that the differentials of the contravariant coor-

<linates of a vector are transformed as contravariant vectors.
However, having considered the scalar function of contravariant
-components <p(x«) and in particular the components of the vector
arptax«, we immediately make sure that we deal with the covariant
components.
Indeed, <p = rp(x'") = q'l[x'"(x'l) ]; the coordinate transformation
is implied to be known; as usual, we shall write out the formulae:
with "convenient" indices. Thus, in accordance with Eq. (A.l.76)
x'"=a~".t<l, whence ax'"!ax«=a~"· In accordance with the for-
mulae for differentiating a composite function
(A.I.85)
this is precisely the transformation formula for the components of

the covariant vector (A. I. 77). On~ again we point out that all
formulae obtained are valid for the space of any number of di-
mensions.
Passing to the 4-space-time of Minkowski, we shall recall that
1he consequence of two Einstein's postulates is the invariance of the
.quadratic form
(A.l.86)
Appendix I 391
on transition from one IFR to another, i.e. under the Lorentz

transformation. The expression (A. I. 86) defines the square of the
elementary "distance" in 4-space. But the square of the distance
(A.I.86) is not necessarily positive. In this connection the Eucli-
dean space determined by the form (A.I.80) is referred to as the
Improper Euclidean or pseudo-Euclidean space. To take advantage
of the formalism of the proper Euclidean space, we can resort to
the technique used in this book and consisting in the introduction
of the imaginary coordinate (d. Chapter 3). This technique simpli-
fies the presentation, but at the same time it unintentionally in-
sinuates the idea of an imaginary nature of relativistic laws them-
selves which, of course, have nothing to do with the number i
utilized only for the purpose of making calculations easier.
Resorting to the real coordinates x 0 = ct, x1 = x, x2 = y, x3 = 2,
we can write Eq. (A.I.86) as
ds 2 = dx9' - dx 1' - dx 2 - dx 1'. (A.I.87)
In any case
iii=modX0 + ml dxl + ~dX2 + tnadX1. (A.I.88)
The expression for ds 2 from Eq. (A.I.87) coincides with that from
Eq. (A.I.86), provided the following conditions are satisfied:
ml= I, m~=m~=mi= -I; (A.I.89)
m1m,=O for i, k=O, 1,2, 3. (A.I.90)
All these conditions can be expressed by one formul<~
(A.I.91)
where
I 0 0 0)
0 -1 0 0
( (A.I.92)
g,,~ 0 0 -1 0 .
0 0 0 -1
Now go, defines the metric tensor of the pseudo-Euclidean space.
~j~;:rfa~~egJ~~~o~~~!~~~or:ntw~e:~~~i~ht~l~0~r~%ar~i::nd con-
A0=A0, A 1 =-A 1, A2 =-A 2, A3 =-A3 • (A.I.93)
The scalar product of two vectors and the norm of a vector are
defined by the following expressions:
AB=g All.Bv=A08°-A 18 1 -A 28 2 -A38 3 (A.I.94)
IA' I ~e..~"A'~ (A')'- (A')'- (A')'- (A')'.' (A.I.95).
Appendix I
It is seen from Eq. (A.I.95) that the norm of an arbitrary non-

zero real vector is not necessarily positive: it can also be zero or
negative. This indicates once more that the four-dimensional space
of the special theory of relativity is pseudo-Euclidean.
§ 9. The definition of hyperbolic functions and some relation-
ships between them. For real x the basic definitions are as fol-
lows:
sinhx= e}J;-2e-}J;, coshx= e}J;+2e-}J;,
tanhx=~= e}J;-e-JJ; = 1-e-b.

coshx e}J;+e-JJ; 1 +eb
It is seen at once that
cosh2 x - sinh2 x = f {(e}J; + e-}1;) 2- (&'- e-}1;) 2} =+. 4 =I.
Dividing the left-hand and the extreme right-hand sides of the last
equality by cosh2 x, we get
1- tanh 2 x = cos~' x , or cosh X= .Yt _ ~anh:l x •
To obtain the formulae

cosh (x 1 + x 2) =cosh x1 cosh x2 +sinh x1 sinh x2,
sinh (x 1 + x 2) =sinh x1 cosh x2 +sinh x2 coshxh
one needs to substitute into the definitions
cosh (x 1 + x 2) = e"•e"'= + ;-"·e-"> , sinh (x 1 + xJ
the values following from the definition of hyperbolic functions:
e 91 •1 =COSh 81,2 +Sinh 81,2• e-BI, I= COSh el, 2 - Sinh 81.2.
Finally, here is the very important formula for the real x:
tanh (x 1 + xJ = ~~:~ ~~ 11 ! ~:~ =

sinh x 1 cosh x1 +cosh x 1 sinh X1 tanh x, tanh x1
wsh x, cosh x1 + smh x, sinh x 1 I + tanh x 1 tanh x 1 •
In this book (Chapter 5) we also make use of t~e expansion
x2 x2
coshx= e'l~e-}J; ~ l+x+T ... ; 1 - x + T .••
=I +.f+ ... , sinhx:::::::x.

, Appendix I 393
Jshde~=~~~~~~h~s ~~f~~~es~ ~~~h~bd~fin~~~n:r!f~~~~~~~et~~c:~~~~

tions (the Euler rormulae)
sinz= e~~-2e-lz • cosz= e'z+2e-lz
we shall put the imaginary value of z, i.e. assume z = i!p. Then it

is immediately seen that
sin (i!p) = i sinh If, cos (i!p) =cosh If, tan (i!p) = i tanh If,
recalling that t"2 =-I and 1/i = -i.
BIBLIOGRAPHY TO APPENDIX I
1. McConnell, A. J. Application of Tensor Analysis, N.Y. 1957.
This rationally compiled book contains all the necessary infor-
mation on tensor analysis utilized in physics. Having studied the
book, the reader becomes quite prepared mathematically ror rearj.
ing books devoted to the Riemannian geometry and the general
theory of relativity.
2. Kochin, N. E. Vector Calculus and Elements of Tensor Cal-
culus, Moscow, 1965 (in Russian).
The book presents the basic information on tensors in very easy
terms.
APPENDIX Tl
THE BASIC FORMULAE

. OF ELECTRODYNAMICS
IN THE GAUSSIAN SYSTEM
In electrodynamics the Gaussian system is used as often as the

Sl system to which we kept in Chapter 6. Surely, the choice of the
system of units does not affed the essence of the matter, but the
appear.ance of formulae varies. For the readers' convenience we
shall cite the basic formulae in the Gaussian system. The formulae
_eossess the same numbering under which they are given in
Chapter 6.
The equations for the potentials A and q> have the following
form in vacuo: '
0 A==AA.-+r a;~=-~}.
(6.9)
D~-~liq>-7z~=-4np.
The Lorentz condition:

divA++*=O. (6.8)
The charge conservation law remains invariable:
-%7-+div/=0. (6.4)
The definition of the 4-potential and 4-current:
iii (A. i~). ii>(~. A), (6.11)

; U. icp), ; (cp, /). (6.12)
The relation between the average fields and potentials:
_.B=rotA, E=-gradcp-7..4.. (6.25)

Appe'l.d!K fl 395
-iD,)
The tensors of an electromagnetic field in matter:
1J H, -H,
( Hz 0 HA -tD11
•=
f1 H 11 H~ 0 -iD 2 '
iDA iDy iDZ 0
-08 8; ~~· =:~;)

F,,.= ( By" - B~ - iE: . (6.31)
tE" l£ 11 iEJl 0
The tensor of the momenta:

0 M, -M,
~,,.= ( -MM, -011

M,
(6.33)
' ''
0
-iP.: -iPy -iP2
now 8 = H + 4.nM and E = D - 4nP.
The field transformation formulae:
a,~a;,
E,~J (E,+-f-a;), B~=r(B~-fE;), (6.36)
E,~r(E;--f-a;). B,~ r(B;+ -f.Ev):

D,~D;, H,~H;,
D11 = f (oy +fH;), H>~r(H>-fv:). (6.37)
Dz = f ( D~-fH~). H,~r(H;+fv;).
Appendix !1
The same jormulae expressed via the prOJections on the relative

velocity direction and the direction perpendicular to it:
E 1 =EI, EJ.=r(EJ. -~(VB'J). B•=BI.
BL~r{B\.++(VE'J): (6.38)
D•=DI, DJ.=f (DJ. -~ (VH'J). H•=HI.

HL~r(H\.++(VD'J). (6.42)
The Lorentz force density·
(6.49)
The field invariants (§ '6.5):

l.=fi~~=if-H 2 , l2=2iEH.
The four-dimensional expression for the Lorentz force density is
the same in SI and the Gaussian system:
(6.53)
The Maxwell equations in the three-dimensional form
rotH=-¥- j + 7D. div D = 4np, (6.56)
rotll=-+iJ. divB=O. (6.57)
The Maxwell equations in the four·dimensional form:

a;,. 4n
~=cs,, (6.60)
a;;; + a;~~~/ + (:;; o. = (6.67)
The material equations io the lbrec-dimcnsional c·;:~ LJ not vary:

D'=eE', (6.68)
B'=1LII', (6.69)
j'~aE'. (6.70)
Appendix II
The material relations in a moving medium:
D++[VHI-•(!!++[YBI). (6.74}
8-+[VEI-•(H-+[VD[). (6.75)
Solving these equations with respect to D and B, we shall obtain

(B- V/c) ·
D- ,_:•"' { eE(I-B2)+(••-I){+IVHJ-7.- V(VE) j.

B- ,_:•"' { "H(l-B')-l•• -l){+[VHJ-f,- V(VE) }.
whence
D 1 =eE1, B 1 =p.H 1, (6.76)
(l-••D')DL - • ( l - B~EL + (••- l)+[VHJ,
(6.77)·
(I ~ •• D•) BL- •0- B')HL- ,•• -I)+ [VIl~
Ignoring the values of B2 and e1J.B 2 as compared to unity in Eq.

(6.77), we shall get
D-•E++(••-I)[VH],
(6.78)
B-•H --!-(••- l)[VE].
The material equations in the rour-dimensional form:
/ 11,u,=eF1,u,, {0.79)·
F1,u 1 + F,tu 1 + Fllu, = p. (/ 1,u 1 + /,1u 1 + fuu,), (6.80)·
s1=fF1,u,. (6.81)·
The electromagnetic field energy density:

BD+BH
w---,"-.
The Poynting-Umov vector:
s-+.-[EH).
398 Appendix II
The energy-momentum-tension tensor in uacuo.
r,.-( _r;s -~cg). g-4,-- .~, IEHI.

Ta,.,=-ia{EaDR + flaB~J-Oas(w).
The energy-momentum-tension tensors in matter (Minkowski and
Abraham):
M ( T~~ - icgM) .~-
T .-
~- -..;s w •
gM=
·-
IDBI.
T~= -in{EaDr. + HaBr.} -Oar,w,
r•1 ( f!, -
"= -fs
icg•)
w •
g' = -,~ = 4 ~, [EH],
·-
T:r.= -in {EaDr, + EflDa + HaBr.+ H11Ba} -O~w.

BIBL/OGRAPfj"(__ __ _
This list includes only those books which cover the subject of the theory of
relativity very substantially. Brief comments given for the readers' convenience
re8ect the personal views of the author of this book.
~The Principle of Relati~Jily, a collection of articles by the classics of the
felativistic theory, Moscow-Lenin~rad, 1935 (in Russian).
tic~~e b~~-ifhi;a~r:r~~~s a~th~~~e~n!keP?f"~~~~b~int~e~~r~"~nM~~~:i:~kiab!~~ ~~;
immediate forerunners of Einstein. The article by Einstein "On the Electrodyna-
mics of Moving Bodies" covers the special theory of relativity almost comple-
tely, except for the problems of thermodynamics Minkowski's report presents
the fundamentals of the STR in terms of four-dimensional geometrical physics.
2. The Principle of Relativity, a collection of articles on the special theory
ofT~i~lib~b'k ~~~~~~~ ~~~ ~~~ ~~~!~~~~-on
3 the STR published in the prc>vious
collecli.on of 1935. Some more articles by Poincare, Lorentz, Planck and Pauli
are added. The third part of the book is devoted to the history of the STR.
3. A. Einstein, Collected Works, Vol. I, Moscow, 1965; Vol. II, 1966; Vol. IV,
1967 (in Russian).
Volumes I and II include Einstein's works on the theory of relativity. Vol·
ume I contains papers- on both the STR and the general theory of relalivitr-
~~~uJec~~t~?;:r~h~af~;~~lyat~~~ a~tric:~~o 0~otr~~ 't~nEr~s:~~;r~ub\i:~!dti~~tflie~~~
separate editions (see refs. 4 and 5 below).
4. A Einstem, The Meaning of Relatiuity, Princeton, 1953.
A very concise and rather complicated presentation of the fundament.al ideas.
5. A. Einstein. Ober die spezieUe und die allgemeine Relalivifiilstheone (ge-
meinverstiindlich), Braunschweig, 1920.
licgn~~o~ids~=i~a~~~e~,i~~~~~~~he~~~bjf.feehxc;0~~~~~n~xposition" (gemein..,erslind-
6. L I Mandelshtam, Lectures on Physical Foundations of ttre Theory of Re-
ll:tivily. Complete Works, Vol. V, Moscow, 1950 A separate publication· Lec-
tures on Optics, Theory of Relativity and Quantum Mechanics, Moscow, 1972
(m Russtan).
The lectures given bh L. I. Mandelshtam and prepared for publication by his
~~!c%~Jordepa;ob1~1~sth:f t~:o~he~:Y:hThi'i~~~:~:C~r~f i~j~~t~~~Yr!~da~ 0 ~~~a~~~
audience and cannot be recommended to the reader who is new to the· subject.
The sophisticated reader, however, should get acquainted with these lectures by
all means.
/.'"C. Moeller, The Theory of Re/.atlui~, Oxford, 1952.
An extended course of university lectures covers a wide range of problems of
the STR and the GTR. II requires deep knowledge of physics and mathematics.
<00 Bibtiography
__k"W. Pauli, Theory oi Relatiuity, London, 1958.

The book is the translation of an article from the Mathematical Encyclopedia
\\ ntten in 1921. The presentation of the subject is fairly complete, even though
<Oncise. The article has a great number of references.
~Ssi~~~~au, E. Lifshits, The Field Theory, the last 6th ed. Moscow, 1973
The first four chapters of the book are in fact devoted to the STR. By and
large, the book systematically presents the gene~al and special theory of relativ-
ity .n terms of the least action principle. Fields in media are not considered
in this book; they are treated by the same authors in the book Continuum Elec-
./rvdynamics, Moscow, 1967.
The book is intended for readers who have a good grounding, for the exposi-
tion of some of the themes is rather concise; many details needed for the first
acquaintance with the STR are omitted. One of the best books on the STR as
concerns both factual data and the manner of presentation.
10 R. C. Tolman, Relatiuity, Thermodynamics and Cosmology, Oxford, 1934.
A classical work devoted both to the STR and to the GTR. II is written
thoroughly and deliberately and is distinguished for the wide scope of the
problems treated.
I I. E. F. Taylor, J. A. Wheeler, Spacetime Physics, San Francisco-London,
1966.
The book is remarkable for an attempt made by the authors to fresent the
STR at the very beginning of a college course. II presents the relahvistic con-
cepts of space and time as well as the relativistic mechanics in much detail. The
book is partially based on the lectures given by Wheeler at the refresher courses
for US teachers. It contains many superb illustrations, drawings and diagrams:
some chapters are provided with supplements including about a hundred very
useful problems pertaining to the most delicate themes of the STR. The book is
<Oncluded with a brief exposition of thE' general theory of relativity.
1 0 1
4~?bo~k ~~~~b~~· fi~~:r~~~c~T~ ~n~het~'::o;«e bT~ ':::~~~Yi:i {'h~ t:~ger
y 9
part. The first nine chapters are devoted to the exposition of the STR. While
-~1iR.gt:: :~fh;rer.lve~~~;~i~~~:~ t~~e~~~~e~~=li~~t~~~::isa~R~ackground for the
13. R. Becker, Theorle der Electrizttdt, Stuttgart, 1957.
A very clear and popular presentation of the principal problems of the STR:
the book is quite understandable for a student versed in eledrodynamics
14. W. Panofsky. M. Phillips, Classical Electricity and Magnetism, New York,
1955
In the second half of the book the reader will find an up-to-dale and interest-
~~fer~~~~ntation of the STR; every chapter is supplemented with problems and
15. A Sommerfeld, Lectures on Theoretical Physics, Vol. 3, Electrodynamics,

New York, 1952.
The third chapter of the book is devoted to the detailed study of the theory
of relativity and the electron theory. A rather peculiar and sufficiently sophisti·
<aled exposition, of great interest to a prepared reader
16.-V. A. Fok, Theory of Space, Time and Gravitation, Moscow, 1964 (in Rus·
si11n):
areA~~:o~~~ tt~eth~\k~ ri~~~:r ~~n~~~l,~~m~~~~e 1 ~fat~l~ ~hne o6r~~\n~~a~t=~~
.ner.
Bibliography 401
;:,:~er~·c;:~s~s~f· e~~:~~;~a;~~~~rC:.~;~;:~s.isNdee~o~~{~~Lfh~db~ief1 ~·osi-

tion of the STR. The applications are considered in Ch. 12 and partially m
Ch. 14. These chapters are supplied with problems.
18. M. A Tonnelat, Les principes de Ja thlorle etectromagnltiques et de lfl
relativitl, Paris, 1959.
The book gives a detailed exposition of many questions of the STR and !he
GTR: however, the presentation is overly comphcaled in notation and mathe-
matics.
?s;"~~~ t:~b~d;e L{Mt::~s. f9ei.hgslcs, R. P. Feynman, R. B Leighton.

The le<:tures delivered by Feynman is an interesting endevour to combine the
general course of physics with the course of theoretical physics. The theory of
relativity is dwelt upon in some chapters of this collection.
20. D. Bohm, The Special Theory of Relatiuity, New York-Am~terdam, 1965
A very interesting book stressing the basic philosophical problems of !hi.l
STR. The principal subje<:t of the book is the development of physical cqncepts
of space and time The concrete problems of the STR touched upon in the book
are presented in a clearcut and precise way. The book requires critical altitude
when the philosophical problems are discussed. It is intended for a sophisticat-
ed reader.
~~isH~:al?d~~~:~":ff!~~~s~~h~1~n~~:e~~!\c~tr~1!~!·ofat~:r~W· a1~2 7the

gravitational theory. It is very useful for the first acquaintance with the STR·
~- H. Bondi, Relatiuilg and Common Sense, London, 1965.
The book is remarkable for its methodical approach to the presentation of !he
STR. II justly points out that the presentation of the STR from the standpoint
of the 19th century physics is not expedient now In this connection the Michel-
son experiment is evaluated from the position of contemporary physics. ,.\[ the
same time, a new method of presentation of the STR is suggested which inlro·
duces the observers possessing identical clocks and radars. Thus, the use of
ri~tid scales is obviated. The K calculus allows the basic kine1na!ic results of thr
STR to be obtained directly from the Einstein postulates, bypassin~ the Lorentz
~r:S,~~r!h~~~nh;~c~h: ~r~s:l~~j~~ncae~t~~sW~liedu~~thg~h~t~~=~tz i~~;~~~~~~~:
lion as well as with the conventional ways of presentat1on.
_........2:J.P. G. Bergmann. The Riddle of Gravitation, New York, 1968.
This small book containing no technical details clearly points out the limited-
ness of the special theory of relativity and the inevitable transition in the gene-
ral case to the GTR. It indicates those conclusions of the STR thai served as
a starting point for the construction of the gravitational theory. The Riddle of
Gravitation consists of three parts: one devoted to the STR, the second to the
GTR, and the third, containing the most recent data of astronomical obsen a-
lions The reader of the book is supposed to be acquainted with the technical
college courses in physics and mathematics. The first part of the book cannot.
however, be used for the first acquaintance with the STR, but is an excellent
summary to be utilized for recapitulation
24. C. Lanczos, Albert Elnsteln and the Cosmic World Order, New York,
1965.
This book can by no means sene as a textbook on the STR However. the
principal approach to the construction of the STR is described clearly and pre-
cisely. The first three chaplers of the book should be recommended to everyone
who begins to study the STR.
"'25." Biblion.raphy
Yu. B. Rumer, M. S. Ryvkin, The Theory of RetatifJity, Moscow, 1960 (in

RUssian).
ST~; cth~pba:~~i~:'fnt~~~~ fo~e~h"y1:i~:de0:a:~~:;,i:~f~1e~~~~gi~~f :;,j~~!s.0 f the
26. B. Jaffe, Michelson and the Speed of Light, London, 1961.
~te~n gg1~~~~~:~tJ§7,1,P~~~co!~d~914 ~j;~R~s:l~n~~lativistic Velocities,
~<'~~~f~it~~~~~~n:~j',' t~i ~jg~f)_rov, Two Paradoxes of the Special Theo-
/ /1~:, ~g 7~amm, Fundamentals of the Theory of Electricity, Mir Publishers,
30 J. A. Stratton, Electromagnetic Theory, New York-Lonclon, 1941.

3r.'L. Marder, Time and Space-Traueller, London, 1971.
- A1. Ya. P. Terletsky, The Paradoxes of the Theory of Relativity, Moscow, 1966
\in Russian).
33. V. L. Ginzburg, Theoretical Physics and Astrophysics, Moscow, 1975 (in
Russian),
34- H. Goldstein, Classical Mechanics, Cambridge (Mass), 1950.
____ -as- C. Lanczos, The Variational Principles of Mechanics, Toronto, 1966
36. A. Sommerfeld, Lectures on Theoretical Physics, Vol. 4, Optics, New
York, 1954.
INDEX
Aberrallon anfle 103 Conservative field, total energy of a

~~~~h~'~£o~~e i~~j 103, 204 particle in t56
Continuity equation 182
Abraham momentum 237 Contravariant components of a vector
Abraham tensor 235 387
Absolute future 127 Convection current 183, 212
Absolute past 127 Convolution of tensor indices 375
Absolute space 29 Coordinate system 12
Absolute t1me 29 Coordinate transformation 16
Anomalous Doppler eftect 'JIIU Covariance of the system of Maxwell's
Antisymmetric: tensor 3110 equations 205
Covariant component~ of a vector 387
Current density 182, 211
Current loop, dipole moment of 216
Bind1ng energy 159 Cyclotronic I requency t67
Biot-Savart law 220
d'Alemberl's operator 374

Cartesian coordinates 12 Descartes-Snell law 329
~=~~e;~~~ ~~f:~i~!t92system 12
Dipole moment of a current loo1> !Ill
Dispersion relahon 248
Cause-and-eftec:t c:yc:le 300 Dis-synchronization of clocks 54
Cause·and-eftect relationship 92 Doppler effect 84, 233, 261
Central force 24 anomalous 280
Charge conservation law 181 radial 84, 262
Charge density 182 transverse 87, 264
Charged particle, motion in a constant
uniform electric field 164 g~~f tceo;:~~ij~~ 246
motion in a constant uniform mar~·
netic: field 167
Cherenkov cone 284
Classical momentum of a partlde 140 Emstein's postulates 38
'Classification of intervals 90 Elastic force 22
Clock paradox 303 Electric field strength 188
Clock synchronization 42 Electric: polarization 191
co\i/~~~~tbe~~~ 2 173
Electromagnetic field invariants 198
Electromagnetic: field strength 190
Conductivity 209 Eledromegnetic waves, phase veloclly
::Con4uctivity current 21~ of 291
Conservation laws of relahvistac me· Energy·mQmPnlum-tenslon tensor 227
c:hanic:s t75 Energy of a particle 140
''" lnde-c
Equivalence of mass and energy 310

Ether 107, 330 ,.
Instantaneous transttion of interaction
HEther wind" 336 Integrals of motion 177
Eulerian angles 365 Interaction of two movinR" charges 220
Event II Interval between events 61
Events, synchronism of 42, 44 classification of 90
Invariance of equallons 22
lnvariance of proper time 135
of the interval between events 63
~f:t~er~}h:~·Aiif~:mr:l~~~n~Zharge 216
Invariants of electromagnetic fi<'id 198
Isotropy of space 14
Fteld tensor 240
Fizeau experiment 31. 97
Force 21 K calculus 105
Abraham 237 Kennedy·Thorndike experiment 111
~~~~~k~~e~~rta1 t~. 363
central 24
elastic 22
friction 22
gravitational 22
Foucault experiment 26 Law of conservation of the total
4-acceleration 138 energy 141
4-current 184 Law of inertia 25
Four-dimensional equation of motion Length, relativity of 14, 113
141 Lifetime of muons 84
Four-dimensional field potential 241 Light aberralion 103, 264
Four-dimensional pseudo-Euclidean
space 118 ~t~~~tc~~~\;;xperiment 48
4-force 141 Light frequency variation on reftection
4-momenlum 141 272
4-potential 184 Light-like interval 94
4-space-time 118 Light metre 63
4-tensor 122, 189 Light, pressure or 270, 278
4-vector 120 Light quanta 276
4-vedor of energy-momentum 153 Line of force 200
4-veloclty 134 ~Longitudinal" mass 353
4-wave vector 261 Lorentz condition 182
Frequency dtspersion 247 Lorentz contraction 78, 342
Lorentz force 147, 182, 199
Lorentz force density 199
Galilean principle of relativity 19 Lorentz matrix 193, 367
Galilean transformation 15, 18, 319 Lorentz transformation 52, 59, 63, 86,
70. 114, 320, 367
g::;~?r~~r1~:~~s~~ theorem 223, 376
<iravitational theory 30, 357
Magnetic field strength 191
Magnetic induction 188
Magnetization 215
Hamiltonian function 154 Mass 21
Heliocentric frame 27 Mass and energy, equivalence of 310
Mass defect 160
Mass tensor 353
Material constants 209
Induction tensor 240 Material equations 208
, Inertia, Jaw of 25 Maxwell equations 181, 206
Inertial reference frame 16, 23, 319 Maxwell equations, covariance of the
Instantaneous CO·moving inertial system of 205
frame 38 Maxwell theory 181
IIUle"' 405
Metric coefficients 11!:1, .!5\:1 Reachon mot10n in relativistic mecha-

Metric tensor 392 nics 170
Michelson's interferometer 338 YReactive" Ioree 149
Michelson-Morley experiment 338 Reference frame 14
Minkowski diagram 187 Reflection of light from a moving mir-
Minkowski equations 209 ror 274
Minkowski equations for moving me- Relative velocity 22. 104
dia 208 Relativistic equation of motion 146
Minkowski space 119 Relativistic mechanics. conservation
Mmkowski tensor 234 laws of 175
Minkowski '>l.'orld 118 reaction motion m 170
Motion of a charged particle in a con- Relalivistic reference frame 41
stant uniform electric field 164 Relativistic transformation of veloci-
m t6~onstant uni~orm m~~~etic field ties 96
Rel.itiviStic velocity 33
Motion of a rocket 170 Relativity, Galilean principle of 19
Moving media, Minkowski equations Relativity of clock synchromtation 52
lor 208 Relativity of length 74, 113
Muons, lifetime of 84 Relativity of simultaneity 74
Relativity of time intervals between
events 83
Rest energy ISO
Newtonian mechanics 13 Rest mass 141
Newton's first law 25 Rest mass of a system 157
Newton's se<:ond law 21
Newton's third law 25
Non-inertial reference frame 27, 354
Sagnac-Garress experiment 344
Observer 19
~S:~~~y~eh~?~all~2( 210
Origin of coordinates 13 Signature 119
Simultaneity, relativity of 74
Skin effect 252
Pe1meability 209 Space-like interval 93
Pe1mittivity 209 Spatial dispersion 247
Phase velocity 250, 291 Stress tensor 383
Phase velocity cl t'leclromagnetic Synchronism of events 42, 44
waves 291 Synchronization of clocks 42
Photon 276
Photon rocket Ill
Sy~~~~ivi~r 0~ 05:-interacting particles
1· ~~ck·s constant 278 175
Hane wave limit~d h :;pace ZU~
Polarization 215
1-'oyutmg ~ector 223, ::!59
Pressure of light 270, Zl8 Tachyons 297
prn.-."r refe cr.:c frame 153 Tardyons 276
Proper til •. e 88, 112 Tensor mdices, convolution of 375
Propcr-1 ir.! Interval 49 Thread-and-level paradox 292
Proper !'me invariance 135 Time interval bt>tween events, relati-
Pseudo-Euclidean plane 123 vity of 83
Pseudo-Euclidean space liG, 120, 391 Time-like mterval 91
lour-dimensional !18 Total energy, law of conservation of
Pse!!do-E!!dit:!c::n thcore:n 127
"'
Total energy of a particle in a cOn·
'>cn·ative field 156
Radar method 105 Total energy of a system 157
Radial Doppler effect 84, 262 Total momentum oi a system 157
406
lrain of waves 266 Vav1l0v·Cherenkov efl"ect 280

Transformation of elt!Cinc and mag- Vavilov·Cherenkov radiation 2::57
netic field components 192 Velocity of light 24, 36, 39
Transformation of velocities, relativls· Vector components, «Jntravariant 3l:l7
tic 95 covariant 387
Transverse Doppler etle~l 87, 264
"Transverse" mass 353
Wave profile 61
Ultra-relativistic particles 155 WO'rld line 109, 123
Unidimensional motion due to a con·
slant force 161
UnUor.mi!y of space 14
Uniformly accelerated motion 162 Zero rest mass 311
TO THE READER
Mir Publishers would be grateful for your comments

on the content, translation and design ol this book. We
would also be pleased to receive any other sugges·
lions you may wish to make.
Our address is:
US$R, 129820, 1·110, GSP
Pervy Rizhsky Pereulok 2
Mir Publishers
Prlnfed ln the Union of Sor~let Soclallst Republics

THE THEORY OF PROBABILITY
by 8 GNEDENKO
The book presents the fundamentals of probab1111y

theory, the mathematical science that deals with the
laws of random phenomena. These laws play an ex-
tremely importa.nt role in phystcal and other fields of
natural science, in engineering, ~onomics, linguistics
and so forth.
The material covered ranges over the followmg
problems: the concept of probability, sequences of
lhdependent trials, Markov chains, random vanables
and dislrtbution funchons, numerical charactenstics of
random variables, the law of large numbers, charac·
tenstic functions, the classical limit theorems, the
theory of infinitely divisible d1slribuhon laws. the
theory of stochastic processes, and elements of the
theory of queues
The theory of probability is presented as a mathe-
matical discipline, however the examples given not
only illustrate the general propositions of the theory
but provide links with problems that occur in the
natural sciences.
The theory of probabtlity is a text for students
of mathematical departments of colleges and univer-
sities. It will also be found of definite interest to spe-
cialists in a wide range of lidds (physicists, engi-
neers, economists, linguists and others) that the
science of probability touches on
MIR PUBLISHERS MOSCOW

Mlr Publishers of Moscow publish Soviet scientific
and technical literature rn sixteen languages- English,
German, French, Italian, Spanish, Portuguese, Czech,
Slovak, Serbo-Croat, Hungarian, Mongol ran, Arabic,
fee;tt!~~k;F~~~ 1 h'rg~~~d~ec~~~~:r1 · :~o~l~m~~dT~~~S.~i~c~~~e
schools, literature on natural sciences and medicine,
Including textbooks for medical schools, popular science
and science fiction.
The contributors to Mir Publishers' list are leading
Soviet scientists and engineers In all fields of science
and technology and Include more than 40 Members
and Corresponding Members of the USSR Academy of Sciences.
Skilled translators provide a high standard of translation
from the original Russian.
Many of the t itles already issued by Mir Publishers have
been adopted as textbooks and manuals at educational
establishments In France, Switzerla:1d , Cuba, Egypt, India,
and many other countries.
Mlr Publishers' books in foreign la:1guages are expor ted
by V/ 0 "Mezhdunarodnaya Kniga" and can be purchased
or ordered through booksellers in your country dealing
with V/ 0 " M£>Zhdunarodnaya Kniga" .

Special Theory of Relativity by Ugarov

Uploaded by

Copyright:

Available Formats

Special Theory of Relativity by Ugarov

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

Special Theory of Relativity by Ugarov

Uploaded by

Copyright:

Available Formats

SPECIAL

@ f.IIIBH&Il peAIHUH11 $H3HKO•M&TeMITH'IeCKOII: JIBTepaT)'pW

It gives me pleasure to thank B. M. Bolotovsky and S. N. Sto-

Chapter I. CLASSICAL MECHANICS AND THE PRINCIPLE OF RELA-

§ 1.1. A coordinate system and a reference frame in classical mechanics II

Chapter 2 THE EINSTEIN POSTULATES THE INTERVAL BETWEEN

§ 2.1. Einstein's postulates 38

Chapter 3. CONSEQUENCES OF THE LORENTZ TRANSFORMATION.

Chapter 4. THE FOUR-DIMENSIONAL SPACE-TIME • 117

Chapter 5. RELATIVISTIC MECHANICS OF A PARTICLE , 133

Chapter 6. THE MAXWELL THEORY IN A RELATIVISTIC FORM. , 180

§ 6.9 The transformation of electric and magnetic moments . . 214

Chapter 7. OPTICAL PHENOMENA AND THE SPECIAL THEORY OF

Chapler 8. ON CERTAIN PARADOXES OF THE SPECIAL THEORY OF

MAIN EVENTS RELATED TO THE HISTORY OF THE STR . • 361

§ t.l. A coordinate system and a reference frame in classical

in space including those removed lar enough from the origin.

measurement tahen by an instrument produces a result that is

m;~n~u~f~~~tr~~e~~=e~::u~g c~nc;e~~rt~~~s~~~~r!nhc~n1~~~.n a~a

to perform this in the following way. First, rotating the reference

the initial moment of time both origins coincided. It is seen from

the event registered by the clocks of both systems will be the

linearly relative to each other and Jet us conduct a "mechanical

deed, let us examine the motion of a particle* in the frame 1{.

the factor m is called the mass of a body.

locities of objects I and II in the frames K and W. using the

--;-1 x,, y,, ,,; v, Ix;~x,-Vt, u;~y,, ,;~,, I v; -••- V

i.e. it remains permanent. In accordance with the Galrlcan trans·

m ~;~ = F (r 12, f).

Accordingly, in the frame K'

An equation which does not change in case of a transforma-

Assuming that an interaction, that is a force, is transmitted at

We begin by showing how important the first law of Newton

in the experiment the Earth withdraws from under the pendulum

• Hereinafter we shall often abbreviate the term '"inertial Frame oF reference"

an inertial one. We shall indicate another experiment of lhts kinO.

cally carried out? We have much to discuss before we can get

ciple within the framework of mechanics? The answer is very

As it is seen from here, the classical Eq. ( 1.4) is not correct in

of all known, that is light signals, are transmitted at a finite

principle of relativity is exactly what is called the first postulate

But what could be done in such a situation? Logically three

The most widely known principle of physics is that of conser·

Obviously, the velocity of light in vacuo has unique properties.

§ 2.1. Einstein's postulates. The final part of the foregoing

~~~ffr~int~;;~f~~o~i~To::~h~~ t~!is~~~~~!c~e: f~;h~o~~1n:; !~

The Lorentz transformation, however, follows directly from the

The construction of a relativistic frame of reference is funda-

clod. located at that point and the arrival at the point x2 by a

• One should not thmk that ~imaginary experiments'" (Gedankenexperimente)

ends of the train reach observer I just at the moment when he

the two ends of the rulers moving past them simultaneously in

traces out the two sides of an equilateral triangle. As the path

11t=2~z~+ ( v:t Y/c.

Finding !1t from the last equation, we get

rule of clock synchronization proposed by Einstein and described

does not _c~~~ultanci~...cl~m~nJs.on!J i~ the

end at the point 0 0• Both events are registered at the point o~

The same two events look somewhat different to the observer

These results are inherent in the theory of relativity and require

m;~n~u~f~tre~~=e~::u~g c~nc;ert~s~~~~r!nhc~n1~~~.n a~a

~~~- t!' 4be1 ~ho~~~i~a=~t:~t :r:;;/?eg o1r~h; 0 ~~~~0~o;~[:c~~~~ f~eaii~:'~r:J.

llt=rllt 0 =~ (3.9)

alv.~ay~o~etilti~:t amty0p~~arl~bsr~f~~~\st1~s~ussssnp5tL ~~!~s~~al~ar;li:Jcsm~~~u~hai~