0475
0475f
The Meaning of Relativity
THE MEANING OF
RELATIVITY
FOUR LECTURES DELIVERED AT
PRINCETON UNIVERSITY, MAY, 1921
BY
ALBERT EINSTEIN
WITH FOUR DIAGRAMS
PRINCETON
PRINCETON UNIVERSITY PRESS
1923
Copyright 1922
Princeton University Press
Published 1922
PRINTED IN GREAT BRITAIN
AT THE ABERDEEN UNIVERSITY PRESS
ABERDEEN
Note.—The translation of these lectures into English
was made by Edwin Plimpton Adams, Professor of Physics in Princeton University
CONTENTS
Lecture I
PAGE
Space and Time in Pre-Relativity Physics . . . . . . . . . . . 1
Lecture II
The Theory of Special Relativity. . . . . . . . . . . . . . . . . . . 25
Lecture III
The General Theory of Relativity . . . . . . . . . . . . . . . . . 59
Lecture IV
The General Theory of Relativity (continued) . . . . . 84
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
iv
THE MEANING OF RELATIVITY
LECTURE I
SPACE AND TIME IN PRE-RELATIVITY
PHYSICS
The theory of relativity is intimately connected with the theory
of space and time. I shall therefore begin with a brief investigation of the origin of our ideas of space and time, although in
doing so I know that I introduce a controversial subject. The
object of all science, whether natural science or psychology, is
to co-ordinate our experiences and to bring them into a logical
system. How are our customary ideas of space and time related
to the character of our experiences?
The experiences of an individual appear to us arranged in a
series of events; in this series the single events which we remember appear to be ordered according to the criterion of “earlier”
and “later,” which cannot be analysed further. There exists,
therefore, for the individual, an I-time, or subjective time. This
in itself is not measurable. I can, indeed, associate numbers with
the events, in such a way that a greater number is associated
with the later event than with an earlier one; but the nature of
this association may be quite arbitrary. This association I can
define by means of a clock by comparing the order of events furnished by the clock with the order of the given series of events.
We understand by a clock something which provides a series of
events which can be counted, and which has other properties of
which we shall speak later.
1
THE MEANING OF RELATIVITY 2
By the aid of speech different individuals can, to a certain
extent, compare their experiences. In this way it is shown that
certain sense perceptions of different individuals correspond to
each other, while for other sense perceptions no such correspondence can be established. We are accustomed to regard as real
those sense perceptions which are common to different individuals, and which therefore are, in a measure, impersonal. The natural sciences, and in particular, the most fundamental of them,
physics, deal with such sense perceptions. The conception of
physical bodies, in particular of rigid bodies, is a relatively constant complex of such sense perceptions. A clock is also a body,
or a system, in the same sense, with the additional property that
the series of events which it counts is formed of elements all of
which can be regarded as equal.
The only justification for our concepts and system of concepts is that they serve to represent the complex of our experiences; beyond this they have no legitimacy. I am convinced that
the philosophers have had a harmful effect upon the progress
of scientific thinking in removing certain fundamental concepts
from the domain of empiricism, where they are under our control, to the intangible heights of the a priori. For even if it should
appear that the universe of ideas cannot be deduced from experience by logical means, but is, in a sense, a creation of the
human mind, without which no science is possible, nevertheless
this universe of ideas is just as little independent of the nature
of our experiences as clothes are of the form of the human body.
This is particularly true of our concepts of time and space, which
physicists have been obliged by the facts to bring down from the
Olympus of the a priori in order to adjust them and put them
in a serviceable condition.
PRE-RELATIVITY PHYSICS 3
We now come to our concepts and judgments of space. It
is essential here also to pay strict attention to the relation of
experience to our concepts. It seems to me that Poincar´e clearly
recognized the truth in the account he gave in his book, “La
Science et l’Hypothese.” Among all the changes which we can
perceive in a rigid body those are marked by their simplicity
which can be made reversibly by an arbitrary motion of the
body; Poincar´e calls these, changes in position. By means of
simple changes in position we can bring two bodies into contact.
The theorems of congruence, fundamental in geometry, have to
do with the laws that govern such changes in position. For the
concept of space the following seems essential. We can form new
bodies by bringing bodies B, C, . . . up to body A; we say that
we continue body A. We can continue body A in such a way that
it comes into contact with any other body, X. The ensemble of
all continuations of body A we can designate as the “space of
the body A.” Then it is true that all bodies are in the “space of
the (arbitrarily chosen) body A.” In this sense we cannot speak
of space in the abstract, but only of the “space belonging to a
body A.” The earth’s crust plays such a dominant rˆole in our
daily life in judging the relative positions of bodies that it has
led to an abstract conception of space which certainly cannot be
defended. In order to free ourselves from this fatal error we shall
speak only of “bodies of reference,” or “space of reference.” It
was only through the theory of general relativity that refinement
of these concepts became necessary, as we shall see later.
I shall not go into detail concerning those properties of the
space of reference which lead to our conceiving points as elements of space, and space as a continuum. Nor shall I attempt
to analyse further the properties of space which justify the con-
THE MEANING OF RELATIVITY 4
ception of continuous series of points, or lines. If these concepts
are assumed, together with their relation to the solid bodies of
experience, then it is easy to say what we mean by the threedimensionality of space; to each point three numbers, x1, x2, x3
(co-ordinates), may be associated, in such a way that this association is uniquely reciprocal, and that x1, x2 and x3 vary continuously when the point describes a continuous series of points
(a line).
It is assumed in pre-relativity physics that the laws of the
orientation of ideal rigid bodies are consistent with Euclidean
geometry. What this means may be expressed as follows: Two
points marked on a rigid body form an interval. Such an interval
can be oriented at rest, relatively to our space of reference, in
a multiplicity of ways. If, now, the points of this space can
be referred to co-ordinates x1, x2, x3, in such a way that the
differences of the co-ordinates, ∆x1, ∆x2, ∆x3, of the two ends
of the interval furnish the same sum of squares,
s
2 = ∆x1
2 + ∆x2
2 + ∆x3
2
, (1)
for every orientation of the interval, then the space of reference is called Euclidean, and the co-ordinates Cartesian.∗
It is
sufficient, indeed, to make this assumption in the limit for an
infinitely small interval. Involved in this assumption there are
some which are rather less special, to which we must call attention on account of their fundamental significance. In the first
place, it is assumed that one can move an ideal rigid body in an
arbitrary manner. In the second place, it is assumed that the behaviour of ideal rigid bodies towards orientation is independent
∗This relation must hold for an arbitrary choice of the origin and of the
direction (ratios ∆x1 : ∆x2 : ∆x3) of the interval.
PRE-RELATIVITY PHYSICS 5
of the material of the bodies and their changes of position, in the
sense that if two intervals can once be brought into coincidence,
they can always and everywhere be brought into coincidence.
Both of these assumptions, which are of fundamental importance for geometry and especially for physical measurements,
naturally arise from experience; in the theory of general relativity their validity needs to be assumed only for bodies and spaces
of reference which are infinitely small compared to astronomical
dimensions.
The quantity s we call the length of the interval. In order
that this may be uniquely determined it is necessary to fix arbitrarily the length of a definite interval; for example, we can put
it equal to 1 (unit of length). Then the lengths of all other intervals may be determined. If we make the xν linearly dependent
upon a parameter λ,
xν = aν + λbν,
we obtain a line which has all the properties of the straight
lines of the Euclidean geometry. In particular, it easily follows
that by laying off n times the interval s upon a straight line, an
interval of length n · s is obtained. A length, therefore, means
the result of a measurement carried out along a straight line by
means of a unit measuring rod. It has a significance which is as
independent of the system of co-ordinates as that of a straight
line, as will appear in the sequel.
We come now to a train of thought which plays an analogous
rˆole in the theories of special and general relativity. We ask
the question: besides the Cartesian co-ordinates which we have
used are there other equivalent co-ordinates? An interval has
THE MEANING OF RELATIVITY 6
a physical meaning which is independent of the choice of coordinates; and so has the spherical surface which we obtain as
the locus of the end points of all equal intervals that we lay off
from an arbitrary point of our space of reference. If xν as well
as x
0
ν (ν from 1 to 3) are Cartesian co-ordinates of our space
of reference, then the spherical surface will be expressed in our
two systems of co-ordinates by the equations
X∆xν
2 = const. (2)
X∆x
0
ν
2 = const. (2a)
How must the x
0
ν be expressed in terms of the xν in order that
equations (2) and (2a) may be equivalent to each other? Regarding the x
0
ν expressed as functions of the xν, we can write,
by Taylor’s theorem, for small values of the ∆xν,
∆x
0
ν =
X
α
∂x0
ν
∂xα
∆xα +
1
2
X
α,β
∂
2x
0
ν
∂xα ∂xβ
∆xα ∆xβ . . . .
If we substitute (2a) in this equation and compare with (1),
we see that the x
0
ν must be linear functions of the xν. If we
therefore put
x
0
ν = aν +
X
α
bναxα, (3)
or
∆x
0
ν =
X
α
bνα ∆xα, (3a)
PRE-RELATIVITY PHYSICS 7
then the equivalence of equations (2) and (2a) is expressed in
the form
X∆x
0
ν
2 = λ
X∆xν
2
(λ independent of ∆xν). (2b)
It therefore follows that λ must be a constant. If we put λ = 1,
(2b) and (3a) furnish the conditions
X
ν
bναbνβ = δαβ, (4)
in which δαβ = 1, or δαβ = 0, according as α = β or α 6= β. The
conditions (4) are called the conditions of orthogonality, and the
transformations (3), (4), linear orthogonal transformations. If
we stipulate that s
2 =
P∆xν
2
shall be equal to the square of
the length in every system of co-ordinates, and if we always measure with the same unit scale, then λ must be equal to 1. Therefore the linear orthogonal transformations are the only ones by
means of which we can pass from one Cartesian system of coordinates in our space of reference to another. We see that in
applying such transformations the equations of a straight line
become equations of a straight line. Reversing equations (3a)
by multiplying both sides by bνβ and summing for all the ν’s,
we obtain
X
ν
bνβ ∆x
0
ν =
X
ν,α
bναbνβ ∆xα =
X
α
δαβ ∆xα = ∆xβ. (5)
The same coefficients, b, also determine the inverse substitution
of ∆xν. Geometrically, bνα is the cosine of the angle between
the x
0
ν axis and the xα axis.
THE MEANING OF RELATIVITY 8
To sum up, we can say that in the Euclidean geometry
there are (in a given space of reference) preferred systems of
co-ordinates, the Cartesian systems, which transform into each
other by linear orthogonal transformations. The distance s between two points of our space of reference, measured by a measuring rod, is expressed in such co-ordinates in a particularly
simple manner. The whole of geometry may be founded upon
this conception of distance. In the present treatment, geometry
is related to actual things (rigid bodies), and its theorems are
statements concerning the behaviour of these things, which may
prove to be true or false.
One is ordinarily accustomed to study geometry divorced
from any relation between its concepts and experience. There
are advantages in isolating that which is purely logical and independent of what is, in principle, incomplete empiricism. This
is satisfactory to the pure mathematician. He is satisfied if he
can deduce his theorems from axioms correctly, that is, without
errors of logic. The question as to whether Euclidean geometry
is true or not does not concern him. But for our purpose it
is necessary to associate the fundamental concepts of geometry
with natural objects; without such an association geometry is
worthless for the physicist. The physicist is concerned with the
question as to whether the theorems of geometry are true or
not. That Euclidean geometry, from this point of view, affirms
something more than the mere deductions derived logically from
definitions may be seen from the following simple consideration.
Between n points of space there are n(n − 1)
2
distances, sµν;
between these and the 3n co-ordinates we have the relations
sµν
2 =
x1(µ) − x1(ν)
2
+
x2(µ) − x2(ν)
2
+ . . .
PRE-RELATIVITY PHYSICS 9
From these n(n − 1)
2
equations the 3n co-ordinates may be
eliminated, and from this elimination at least n(n − 1)
2
− 3n
equations in the sµν will result.∗ Since the sµν are measurable
quantities, and by definition are independent of each other, these
relations between the sµν are not necessary a priori.
From the foregoing it is evident that the equations of transformation (3), (4) have a fundamental significance in Euclidean
geometry, in that they govern the transformation from one
Cartesian system of co-ordinates to another. The Cartesian
systems of co-ordinates are characterized by the property that
in them the measurable distance between two points, s, is
expressed by the equation
s
2 =
X∆xν
2
.
If K(xν) and K0
(xν)
are two Cartesian systems of co-ordinates,
then
X∆xν
2 =
X∆x
0
ν
2
.
The right-hand side is identically equal to the left-hand side
on account of the equations of the linear orthogonal transformation, and the right-hand side differs from the left-hand side
only in that the xν are replaced by the x
0
ν. This is expressed
by the statement that P∆xν
2
is an invariant with respect to
linear orthogonal transformations. It is evident that in the Euclidean geometry only such, and all such, quantities have an
objective significance, independent of the particular choice of
∗
In reality there are n(n − 1)
2
− 3n + 6 equations.
THE MEANING OF RELATIVITY 10
the Cartesian co-ordinates, as can be expressed by an invariant with respect to linear orthogonal transformations. This is
the reason that the theory of invariants, which has to do with
the laws that govern the form of invariants, is so important for
analytical geometry.
As a second example of a geometrical invariant, consider a
volume. This is expressed by
V =
Z Z Z dx1 dx2 dx3.
By means of Jacobi’s theorem we may write
Z Z Z dx0
1 dx0
2 dx0
3 =
Z Z Z ∂(x
0
1, x0
2, x0
3)
∂(x1, x2, x3)
dx1 dx2 dx3
where the integrand in the last integral is the functional determinant of the x
0
ν with respect to the xν, and this by (3) is equal
to the determinant |bµν| of the coefficients of substitution, bνα. If
we form the determinant of the δµα from equation (4), we obtain,
by means of the theorem of multiplication of determinants,
1 = |δαβ| =
X
ν
bναbνβ
= |bµν|
2
; |bµν| = ±1. (6)
If we limit ourselves to those transformations which have the determinant +1,∗ and only these arise from continuous variations
of the systems of co-ordinates, then V is an invariant.
∗There are thus two kinds of Cartesian systems which are designated as
“right-handed” and “left-handed” systems. The difference between these is
familiar to every physicist and engineer. It is interesting to note that these
two kinds of systems cannot be defined geometrically, but only the contrast
between them.
PRE-RELATIVITY PHYSICS 11
Invariants, however, are not the only forms by means of
which we can give expression to the independence of the particular choice of the Cartesian co-ordinates. Vectors and tensors
are other forms of expression. Let us express the fact that the
point with the current co-ordinates xν lies upon a straight line.
We have
xν − Aν = λBν (ν from 1 to 3).
Without limiting the generality we can put
XBν
2 = 1.
If we multiply the equations by bβν (compare (3a) and (5))
and sum for all the ν’s, we get
x
0
β − A
0
β = λB0
β,
where we have written
B
0
β =
X
ν
bβνBν; A
0
β =
X
ν
bβνAν.
These are the equations of straight lines with respect to a
second Cartesian system of co-ordinates K0
. They have the
same form as the equations with respect to the original system of co-ordinates. It is therefore evident that straight lines
have a significance which is independent of the system of coordinates. Formally, this depends upon the fact that the quantities (xν − Aν) − λBν are transformed as the components of
an interval, ∆xν. The ensemble of three quantities, defined for
every system of Cartesian co-ordinates, and which transform as
the components of an interval, is called a vector. If the three
THE MEANING OF RELATIVITY 12
components of a vector vanish for one system of Cartesian coordinates, they vanish for all systems, because the equations of
transformation are homogeneous. We can thus get the meaning
of the concept of a vector without referring to a geometrical representation. This behaviour of the equations of a straight line
can be expressed by saying that the equation of a straight line
is co-variant with respect to linear orthogonal transformations.
We shall now show briefly that there are geometrical entities
which lead to the concept of tensors. Let P0 be the centre of a
surface of the second degree, P any point on the surface, and
ξν the projections of the interval P0P upon the co-ordinate axes.
Then the equation of the surface is
Xaµνξµξν = 1.
In this, and in analogous cases, we shall omit the sign of summation, and understand that the summation is to be carried out
for those indices that appear twice. We thus write the equation
of the surface
aµνξµξν = 1.
The quantities aµν determine the surface completely, for a given
position of the centre, with respect to the chosen system of
Cartesian co-ordinates. From the known law of transformation
for the ξν (3a) for linear orthogonal transformations, we easily
find the law of transformation for the aµν:
∗
a
0
στ = bσµbτ νaµν.
∗The equation a
0
στ ξ
0
σ
ξ
0
τ = 1 may, by (5), be replaced by
a
0
στ bµσbντ ξµξν = 1, from which the result stated immediately follows.
PRE-RELATIVITY PHYSICS 13
This transformation is homogeneous and of the first degree in
the aµν. On account of this transformation, the aµν are called
components of a tensor of the second rank (the latter on account
of the double index). If all the components, aµν, of a tensor with
respect to any system of Cartesian co-ordinates vanish, they
vanish with respect to every other Cartesian system. The form
and the position of the surface of the second degree is described
by this tensor (a).
Analytic tensors of higher rank (number of indices) may be
defined. It is possible and advantageous to regard vectors as
tensors of rank 1, and invariants (scalars) as tensors of rank 0.
In this respect, the problem of the theory of invariants may be so
formulated: according to what laws may new tensors be formed
from given tensors? We shall consider these laws now, in order
to be able to apply them later. We shall deal first only with the
properties of tensors with respect to the transformation from
one Cartesian system to another in the same space of reference,
by means of linear orthogonal transformations. As the laws are
wholly independent of the number of dimensions, we shall leave
this number, n, indefinite at first.
Definition. If a figure is defined with respect to every system of Cartesian co-ordinates in a space of reference of n dimensions by the n
α numbers Aµνρ · · · (α = number of indices), then
these numbers are the components of a tensor of rank α if the
transformation law is
A
0
µ0ν
0ρ
0
· · · = bµ0µbν
0νbρ
0ρ . . . Aµνρ · · ·
. (7)
THE MEANING OF RELATIVITY 14
Remark. From this definition it follows that
Aµνρ · · · = BµCνDρ . . . (8)
is an invariant, provided that (B), (C), (D) . . . are vectors.
Conversely, the tensor character of (A) may be inferred, if it
is known that the expression (8) leads to an invariant for an
arbitrary choice of the vectors (B), (C), etc.
Addition and Subtraction. By addition and subtraction of
the corresponding components of tensors of the same rank, a
tensor of equal rank results:
Aµνρ · · · ± Bµνρ · · · = Cµνρ · · ·
. (9)
The proof follows from the definition of a tensor given above.
Multiplication. From a tensor of rank α and a tensor of
rank β we may obtain a tensor of rank α + β by multiplying all
the components of the first tensor by all the components of the
second tensor:
Tµνρ · · · αβ · · · = Aµνρ · · · Bαβγ · · ·
. (10)
Contraction. A tensor of rank α − 2 may be obtained from
one of rank α by putting two definite indices equal to each other
and then summing for this single index:
Tρ · · · = Aµµρ · · ·(= X
µ
Aµµρ · · ·). (11)
PRE-RELATIVITY PHYSICS 15
The proof is
A
0
µµρ · · · = bµαbµβbργ . . . Aαβγ · · · = δαβbργ . . . Aαβγ · · ·
= bργ . . . Aααγ · · ·
.
In addition to these elementary rules of operation there is
also the formation of tensors by differentiation (“erweiterung”):
Tµνρ · · · α =
∂Aµνρ · · ·
∂xα
. (12)
New tensors, in respect to linear orthogonal transformations,
may be formed from tensors according to these rules of operation.
Symmetrical Properties of Tensors. Tensors are called symmetrical or skew-symmetrical in respect to two of their indices,
µ and ν, if both the components which result from interchanging the indices µ and ν are equal to each other or equal with
opposite signs.
Condition for symmetry: Aµνρ = Aνµρ.
Condition for skew-symmetry: Aµνρ = −Aνµρ.
Theorem. The character of symmetry or skew-symmetry
exists independently of the choice of co-ordinates, and in this
lies its importance. The proof follows from the equation defining tensors.
Special Tensors.
I. The quantities δρσ (4) are tensor components (fundamental tensor).
THE MEANING OF RELATIVITY 16
Proof. If in the right-hand side of the equation of transformation A0
µν = bµαbνβAαβ, we substitute for Aαβ the quantities δαβ (which are equal to 1 or 0 according as α = β or α 6= β),
we get
A
0
µν = bµαbνα = δµν.
The justification for the last sign of equality becomes evident if
one applies (4) to the inverse substitution (5).
II. There is a tensor (δµνρ · · ·) skew-symmetrical with respect
to all pairs of indices, whose rank is equal to the number of
dimensions, n, and whose components are equal to +1 or −1
according as µ ν ρ . . . is an even or odd permutation of 1 2 3 . . . .
The proof follows with the aid of the theorem proved above
|bρσ| = 1.
These few simple theorems form the apparatus from the
theory of invariants for building the equations of pre-relativity
physics and the theory of special relativity.
We have seen that in pre-relativity physics, in order to specify relations in space, a body of reference, or a space of reference,
is required, and, in addition, a Cartesian system of co-ordinates.
We can fuse both these concepts into a single one by thinking
of a Cartesian system of co-ordinates as a cubical frame-work
formed of rods each of unit length. The co-ordinates of the lattice points of this frame are integral numbers. It follows from
the fundamental relation
s
2 = ∆x1
2 + ∆x2
2 + ∆x3
2
(13)
that the members of such a space-lattice are all of unit length.
To specify relations in time, we require in addition a standard
clock placed at the origin of our Cartesian system of co-ordinates
PRE-RELATIVITY PHYSICS 17
or frame of reference. If an event takes place anywhere we can
assign to it three co-ordinates, xν, and a time t, as soon as
we have specified the time of the clock at the origin which is
simultaneous with the event. We therefore give an objective significance to the statement of the simultaneity of distant events,
while previously we have been concerned only with the simultaneity of two experiences of an individual. The time so specified
is at all events independent of the position of the system of coordinates in our space of reference, and is therefore an invariant
with respect to the transformation (3).
It is postulated that the system of equations expressing the
laws of pre-relativity physics is co-variant with respect to the
transformation (3), as are the relations of Euclidean geometry.
The isotropy and homogeneity of space is expressed in this way.∗
We shall now consider some of the more important equations of
physics from this point of view.
The equations of motion of a material particle are
m
d
2xν
dt2
= Xν; (14)
(dxν) is a vector; dt, and therefore also 1
dt, an invariant; thus
∗The laws of physics could be expressed, even in case there were a
unique direction in space, in such a way as to be co-variant with respect
to the transformation (3); but such an expression would in this case be
unsuitable. If there were a unique direction in space it would simplify the
description of natural phenomena to orient the system of co-ordinates in
a definite way in this direction. But if, on the other hand, there is no
unique direction in space it is not logical to formulate the laws of nature
in such a way as to conceal the equivalence of systems of co-ordinates that
are oriented differently. We shall meet with this point of view again in the
theories of special and general relativity.
THE MEANING OF RELATIVITY 18
dxν
dt
is a vector; in the same way it may be shown that
d
2xν
dt2
is a vector. In general, the operation of differentiation
with respect to time does not alter the tensor character. Since
m is an invariant (tensor of rank 0),
m
d
2xν
dt2
is a vector, or
tensor of rank 1 (by the theorem of the multiplication of tensors). If the force (Xν) has a vector character, the same holds for
the difference
m
d
2xν
dt2
− Xν
. These equations of motion are
therefore valid in every other system of Cartesian co-ordinates
in the space of reference. In the case where the forces are conservative we can easily recognize the vector character of (Xν).
For a potential energy, Φ, exists, which depends only upon the
mutual distances of the particles, and is therefore an invariant.
The vector character of the force, Xν = −
∂Φ
∂xν
, is then a consequence of our general theorem about the derivative of a tensor
of rank 0.
Multiplying by the velocity, a tensor of rank 1, we obtain the
tensor equation
m
d
2xν
dt2
− Xν
dxν
dt = 0.
By contraction and multiplication by the scalar dt we obtain the
equation of kinetic energy
d
mq2
2
= Xν dxν.
PRE-RELATIVITY PHYSICS 19
If ξν denotes the difference of the co-ordinates of the material particle and a point fixed in space, then the ξν have the
character of vectors. We evidently have d
2xν
dt2
=
d
2
ξν
dt2
, so that
the equations of motion of the particle may be written
m
d
2
ξν
dt2
− Xν = 0.
Multiplying this equation by ξµ we obtain a tensor equation
m
d
2
ξν
dt2
− Xν
ξµ = 0.
Contracting the tensor on the left and taking the time average we obtain the virial theorem, which we shall not consider
further. By interchanging the indices and subsequent subtraction, we obtain, after a simple transformation, the theorem of
moments,
d
dt
m
ξµ
dξν
dt − ξν
dξµ
dt = ξµXν − ξνXµ. (15)
It is evident in this way that the moment of a vector is not a
vector but a tensor. On account of their skew-symmetrical character there are not nine, but only three independent equations of
this system. The possibility of replacing skew-symmetrical tensors of the second rank in space of three dimensions by vectors
depends upon the formation of the vector
Aµ =
1
2
Aστ δστµ.
THE MEANING OF RELATIVITY 20
If we multiply the skew-symmetrical tensor of rank 2 by the
special skew-symmetrical tensor δ introduced above, and contract twice, a vector results whose components are numerically
equal to those of the tensor. These are the so-called axial vectors which transform differently, from a right-handed system to
a left-handed system, from the ∆xν. There is a gain in picturesqueness in regarding a skew-symmetrical tensor of rank 2
as a vector in space of three dimensions, but it does not represent the exact nature of the corresponding quantity so well as
considering it a tensor.
We consider next the equations of motion of a continuous
medium. Let ρ be the density, uν the velocity components considered as functions of the co-ordinates and the time, Xν the
volume forces per unit of mass, and pνσ the stresses upon a surface perpendicular to the σ-axis in the direction of increasing xν.
Then the equations of motion are, by Newton’s law,
ρ
duν
dt = −
∂pνσ
∂xσ
+ ρXν,
in which duν
dt is the acceleration of the particle which at time t
has the co-ordinates xν. If we express this acceleration by partial
differential coefficients, we obtain, after dividing by ρ,
∂uν
dt +
∂uν
dxσ
uσ = −
1
ρ
∂pνσ
∂xσ
+ Xν. (16)
We must show that this equation holds independently of the
special choice of the Cartesian system of co-ordinates. (uν) is a
vector, and therefore ∂uν
∂t is also a vector. ∂uν
∂xσ
is a tensor of
PRE-RELATIVITY PHYSICS 21
rank 2, ∂uν
∂xσ
uτ is a tensor of rank 3. The second term on the left
results from contraction in the indices σ, τ . The vector character
of the second term on the right is obvious. In order that the first
term on the right may also be a vector it is necessary for pνσ to be
a tensor. Then by differentiation and contraction ∂pνσ
∂xσ
results,
and is therefore a vector, as it also is after multiplication by
the reciprocal scalar 1
ρ
. That pνσ is a tensor, and therefore
transforms according to the equation
p
0
µν = bµαbνβpαβ,
is proved in mechanics by integrating this equation over an infinitely small tetrahedron. It is also proved there, by application of the theorem of moments to an infinitely small parallelopipedon, that pνσ = pσν, and hence that the tensor of the stress is
a symmetrical tensor. From what has been said it follows that,
with the aid of the rules given above, the equation is co-variant
with respect to orthogonal transformations in space (rotational
transformations); and the rules according to which the quantities in the equation must be transformed in order that the
equation may be co-variant also become evident.
The co-variance of the equation of continuity,
∂ρ
∂t +
∂(ρuν)
∂xν
= 0, (17)
requires, from the foregoing, no particular discussion.
We shall also test for co-variance the equations which express
the dependence of the stress components upon the properties of
THE MEANING OF RELATIVITY 22
the matter, and set up these equations for the case of a compressible viscous fluid with the aid of the conditions of co-variance.
If we neglect the viscosity, the pressure, p, will be a scalar, and
will depend only upon the density and the temperature of the
fluid. The contribution to the stress tensor is then evidently
pδµν
in which δµν is the special symmetrical tensor. This term will
also be present in the case of a viscous fluid. But in this case
there will also be pressure terms, which depend upon the space
derivatives of the uν. We shall assume that this dependence is a
linear one. Since these terms must be symmetrical tensors, the
only ones which enter will be
α
∂uµ
∂xν
+
∂uν
∂xµ
+ βδµν
∂uα
∂xα
(for ∂uα
∂xα
is a scalar). For physical reasons (no slipping) it
is assumed that for symmetrical dilatations in all directions,
i.e. when
∂u1
∂x1
=
∂u2
∂x2
=
∂u3
∂x3
;
∂u1
∂x2
, etc., = 0,
there are no frictional forces present, from which it follows that
β = −
2
3
α. If only ∂u1
∂x3
is different from zero, let p31 = −α
∂u1
∂x3
,
by which α is determined. We then obtain for the complete
stress tensor,
pµν = pδµν−α
∂uµ
∂xν
+
∂uν
∂xµ
−
2
3
∂u1
∂x1
+
∂u2
∂x2
+
∂u3
∂x3
δµν
. (18)
PRE-RELATIVITY PHYSICS 23
The heuristic value of the theory of invariants, which arises
from the isotropy of space (equivalence of all directions), becomes evident from this example.
We consider, finally, Maxwell’s equations in the form which
are the foundation of the electron theory of Lorentz.
∂h3
∂x2
−
∂h2
∂x3
=
1
c
∂e1
∂t +
1
c
i1,
∂h1
∂x3
−
∂h3
∂x1
=
1
c
∂e2
∂t +
1
c
i2,
∂h2
∂x1
−
∂h1
∂x2
=
1
c
∂e3
∂t +
1
c
i3,
∂e1
∂x1
+
∂e2
∂x2
+
∂e3
∂x3
= ρ;
(19)
∂e3
∂x2
−
∂e2
∂x3
= −
1
c
∂h1
∂t ,
∂e1
∂x3
−
∂e3
∂x1
= −
1
c
∂h2
∂t ,
∂e2
∂x1
−
∂e1
∂x2
= −
1
c
∂h3
∂t ,
∂h1
∂x1
+
∂h2
∂x2
+
∂h3
∂x3
= 0.
(20)
i is a vector, because the current density is defined as the
density of electricity multiplied by the vector velocity of the
electricity. According to the first three equations it is evident
that e is also to be regarded as a vector. Then h cannot be
regarded as a vector.∗ The equations may, however, easily be
∗These considerations will make the reader familiar with tensor opera-
THE MEANING OF RELATIVITY 24
interpreted if h is regarded as a skew-symmetrical tensor of the
second rank. In this sense, we write h23, h31, h12, in place of
h1, h2, h3 respectively. Paying attention to the skew-symmetry
of hµν, the first three equations of (19) and (20) may be written
in the form
∂hµν
∂xν
=
1
c
∂eµ
∂t +
1
c
iµ, (19a)
∂eµ
∂xν
−
∂eν
∂xµ
= +
1
c
∂hµν
∂t . (20a)
In contrast to e, h appears as a quantity which has the same type
of symmetry as an angular velocity. The divergence equations
then take the form
∂eν
∂xν
= ρ, (19b)
∂hµν
∂xρ
+
∂hνρ
∂xµ
+
∂hρµ
∂xν
= 0. (20b)
The last equation is a skew-symmetrical tensor equation of the
third rank (the skew-symmetry of the left-hand side with respect to every pair of indices may easily be proved, if attention
is paid to the skew-symmetry of hµν). This notation is more
natural than the usual one, because, in contrast to the latter,
it is applicable to Cartesian left-handed systems as well as to
right-handed systems without change of sign.
tions without the special difficulties of the four-dimensional treatment; corresponding considerations in the theory of special relativity (Minkowski’s
interpretation of the field) will then offer fewer difficulties.
LECTURE II
THE THEORY OF SPECIAL RELATIVITY
The previous considerations concerning the configuration of
rigid bodies have been founded, irrespective of the assumption
as to the validity of the Euclidean geometry, upon the hypothesis
that all directions in space, or all configurations of Cartesian systems of co-ordinates, are physically equivalent. We may express
this as the “principle of relativity with respect to direction,” and
it has been shown how equations (laws of nature) may be found,
in accord with this principle, by the aid of the calculus of tensors. We now inquire whether there is a relativity with respect
to the state of motion of the space of reference; in other words,
whether there are spaces of reference in motion relatively to each
other which are physically equivalent. From the standpoint of
mechanics it appears that equivalent spaces of reference do exist. For experiments upon the earth tell us nothing of the fact
that we are moving about the sun with a velocity of approximately 30 kilometres a second. On the other hand, this physical
equivalence does not seem to hold for spaces of reference in arbitrary motion; for mechanical effects do not seem to be subject
to the same laws in a jolting railway train as in one moving with
uniform velocity; the rotation of the earth must be considered
in writing down the equations of motion relatively to the earth.
It appears, therefore, as if there were Cartesian systems of coordinates, the so-called inertial systems, with reference to which
the laws of mechanics (more generally the laws of physics) are
expressed in the simplest form. We may infer the validity of
the following theorem: If K is an inertial system, then every
25
THE MEANING OF RELATIVITY 26
other system K0 which moves uniformly and without rotation
relatively to K, is also an inertial system; the laws of nature are
in concordance for all inertial systems. This statement we shall
call the “principle of special relativity.” We shall draw certain
conclusions from this principle of “relativity of translation” just
as we have already done for relativity of direction.
In order to be able to do this, we must first solve the following
problem. If we are given the Cartesian co-ordinates, xν, and
the time, t, of an event relatively to one inertial system, K,
how can we calculate the co-ordinates, x
0
ν, and the time, t
0
, of
the same event relatively to an inertial system K0 which moves
with uniform translation relatively to K? In the pre-relativity
physics this problem was solved by making unconsciously two
hypotheses:—
1. The time is absolute; the time of an event, t
0
, relatively
to K0
is the same as the time relatively to K. If instantaneous
signals could be sent to a distance, and if one knew that the
state of motion of a clock had no influence on its rate, then this
assumption would be physically established. For then clocks,
similar to one another, and regulated alike, could be distributed
over the systems K and K0
, at rest relatively to them, and their
indications would be independent of the state of motion of the
systems; the time of an event would then be given by the clock
in its immediate neighbourhood.
2. Length is absolute; if an interval, at rest relatively to K,
has a length s, then it has the same length s relatively to a
system K0 which is in motion relatively to K.
If the axes of K and K0 are parallel to each other, a simple
calculation based on these two assumptions, gives the equations
SPECIAL RELATIVITY 27
of transformation
x
0
ν = xν − aν − bνt,
t
0 = t − b. )
(21)
This transformation is known as the “Galilean Transformation.” Differentiating twice by the time, we get
d
2x
0
ν
dt2
=
d
2xν
dt2
.
Further, it follows that for two simultaneous events,
x
0
ν
(1) − x
0
ν
(2) = xν
(1) − xν
(2)
.
The invariance of the distance between the two points results
from squaring and adding. From this easily follows the covariance of Newton’s equations of motion with respect to the
Galilean transformation (21). Hence it follows that classical
mechanics is in accord with the principle of special relativity if
the two hypotheses respecting scales and clocks are made.
But this attempt to found relativity of translation upon the
Galilean transformation fails when applied to electromagnetic
phenomena. The Maxwell-Lorentz electromagnetic equations
are not co-variant with respect to the Galilean transformation.
In particular, we note, by (21), that a ray of light which referred
to K has a velocity c, has a different velocity referred to K0
,
depending upon its direction. The space of reference of K is
therefore distinguished, with respect to its physical properties,
from all spaces of reference which are in motion relatively to it
(quiescent æther). But all experiments have shown that electromagnetic and optical phenomena, relatively to the earth as the
THE MEANING OF RELATIVITY 28
body of reference, are not influenced by the translational velocity of the earth. The most important of these experiments are
those of Michelson and Morley, which I shall assume are known.
The validity of the principle of special relativity can therefore
hardly be doubted.
On the other hand, the Maxwell-Lorentz equations have
proved their validity in the treatment of optical problems in
moving bodies. No other theory has satisfactorily explained the
facts of aberration, the propagation of light in moving bodies
(Fizeau), and phenomena observed in double stars (De Sitter).
The consequence of the Maxwell-Lorentz equations that in a
vacuum light is propagated with the velocity c, at least with respect to a definite inertial system K, must therefore be regarded
as proved. According to the principle of special relativity, we
must also assume the truth of this principle for every other
inertial system.
Before we draw any conclusions from these two principles
we must first review the physical significance of the concepts
“time” and “velocity.” It follows from what has gone before, that
co-ordinates with respect to an inertial system are physically
defined by means of measurements and constructions with the
aid of rigid bodies. In order to measure time, we have supposed
a clock, U, present somewhere, at rest relatively to K. But
we cannot fix the time, by means of this clock, of an event
whose distance from the clock is not negligible; for there are no
“instantaneous signals” that we can use in order to compare the
time of the event with that of the clock. In order to complete the
definition of time we may employ the principle of the constancy
of the velocity of light in a vacuum. Let us suppose that we
place similar clocks at points of the system K, at rest relatively
SPECIAL RELATIVITY 29
to it, and regulated according to the following scheme. A ray
of light is sent out from one of the clocks, Um, at the instant
when it indicates the time tm, and travels through a vacuum a
distance rmn, to the clock Un; at the instant when this ray meets
the clock Un the latter is set to indicate the time tn = tm +
rmn
c
.
∗
The principle of the constancy of the velocity of light then states
that this adjustment of the clocks will not lead to contradictions.
With clocks so adjusted, we can assign the time to events which
take place near any one of them. It is essential to note that this
definition of time relates only to the inertial system K, since
we have used a system of clocks at rest relatively to K. The
assumption which was made in the pre-relativity physics of the
absolute character of time (i.e. the independence of time of the
choice of the inertial system) does not follow at all from this
definition.
The theory of relativity is often criticized for giving, without justification, a central theoretical rˆole to the propagation
of light, in that it founds the concept of time upon the law of
propagation of light. The situation, however, is somewhat as
follows. In order to give physical significance to the concept of
time, processes of some kind are required which enable relations
to be established between different places. It is immaterial what
kind of processes one chooses for such a definition of time. It
is advantageous, however, for the theory, to choose only those
∗Strictly speaking, it would be more correct to define simultaneity first,
somewhat as follows: two events taking place at the points A and B of
the system K are simultaneous if they appear at the same instant when
observed from the middle point, M, of the interval AB. Time is then
defined as the ensemble of the indications of similar clocks, at rest relatively
to K, which register the same simultaneously.
THE MEANING OF RELATIVITY 30
processes concerning which we know something certain. This
holds for the propagation of light in vacuo in a higher degree
than for any other process which could be considered, thanks to
the investigations of Maxwell and H. A. Lorentz.
From all of these considerations, space and time data have
a physically real, and not a mere fictitious, significance; in particular this holds for all the relations in which co-ordinates and
time enter, e.g. the relations (21). There is, therefore, sense in
asking whether those equations are true or not, as well as in
asking what the true equations of transformation are by which
we pass from one inertial system K to another, K0
, moving relatively to it. It may be shown that this is uniquely settled by
means of the principle of the constancy of the velocity of light
and the principle of special relativity.
To this end we think of space and time physically defined
with respect to two inertial systems, K and K0
, in the way that
has been shown. Further, let a ray of light pass from one point P1
to another point P2 of K through a vacuum. If r is the measured
distance between the two points, then the propagation of light
must satisfy the equation
r = c · ∆t.
If we square this equation, and express r
2 by the differences
of the co-ordinates, ∆xν, in place of this equation we can write
X(∆xν)
2 − c
2 ∆t
2 = 0. (22)
This equation formulates the principle of the constancy of the
velocity of light relatively to K. It must hold whatever may be
the motion of the source which emits the ray of light.
SPECIAL RELATIVITY 31
The same propagation of light may also be considered relatively to K0
, in which case also the principle of the constancy of
the velocity of light must be satisfied. Therefore, with respect
to K0
, we have the equation
X(∆x
0
ν)
2 − c
2 ∆t
02 = 0. (22a)
Equations (22a) and (22) must be mutually consistent with
each other with respect to the transformation which transforms
from K to K0
. A transformation which effects this we shall call
a “Lorentz transformation.”
Before considering these transformations in detail we shall
make a few general remarks about space and time. In the prerelativity physics space and time were separate entities. Specifications of time were independent of the choice of the space of
reference. The Newtonian mechanics was relative with respect
to the space of reference, so that, e.g. the statement that two
non-simultaneous events happened at the same place had no objective meaning (that is, independent of the space of reference).
But this relativity had no rˆole in building up the theory. One
spoke of points of space, as of instants of time, as if they were
absolute realities. It was not observed that the true element
of the space-time specification was the event, specified by the
four numbers x1, x2, x3, t. The conception of something happening was always that of a four-dimensional continuum; but
the recognition of this was obscured by the absolute character
of the pre-relativity time. Upon giving up the hypothesis of the
absolute character of time, particularly that of simultaneity, the
four-dimensionality of the time-space concept was immediately
recognized. It is neither the point in space, nor the instant in
THE MEANING OF RELATIVITY 32
time, at which something happens that has physical reality, but
only the event itself. There is no absolute (independent of the
space of reference) relation in space, and no absolute relation
in time between two events, but there is an absolute (independent of the space of reference) relation in space and time, as
will appear in the sequel. The circumstance that there is no
objective rational division of the four-dimensional continuum
into a three-dimensional space and a one-dimensional time continuum indicates that the laws of nature will assume a form
which is logically most satisfactory when expressed as laws in
the four-dimensional space-time continuum. Upon this depends
the great advance in method which the theory of relativity owes
to Minkowski. Considered from this standpoint, we must regard
x1, x2, x3, t as the four co-ordinates of an event in the fourdimensional continuum. We have far less success in picturing
to ourselves relations in this four-dimensional continuum than
in the three-dimensional Euclidean continuum; but it must be
emphasized that even in the Euclidean three-dimensional geometry its concepts and relations are only of an abstract nature in
our minds, and are not at all identical with the images we form
visually and through our sense of touch. The non-divisibility of
the four-dimensional continuum of events does not at all, however, involve the equivalence of the space co-ordinates with the
time co-ordinate. On the contrary, we must remember that the
time co-ordinate is defined physically wholly differently from the
space co-ordinates. The relations (22) and (22a) which when
equated define the Lorentz transformation show, further, a difference in the rˆole of the time co-ordinate from that of the space
co-ordinates; for the term ∆t
2 has the opposite sign to the space
terms, ∆x1
2
, ∆x2
2
, ∆x3
2
.
SPECIAL RELATIVITY 33
Before we analyse further the conditions which define the
Lorentz transformation, we shall introduce the light-time, l = ct,
in place of the time, t, in order that the constant c shall not
enter explicitly into the formulas to be developed later. Then
the Lorentz transformation is defined in such a way that, first,
it makes the equation
∆x1
2 + ∆x2
2 + ∆x3
2 − ∆l
2 = 0 (22b)
a co-variant equation, that is, an equation which is satisfied with
respect to every inertial system if it is satisfied in the inertial
system to which we refer the two given events (emission and
reception of the ray of light). Finally, with Minkowski, we introduce in place of the real time co-ordinate l = ct, the imaginary
time co-ordinate
x4 = il = ict (
√
−1 = i).
Then the equation defining the propagation of light, which must
be co-variant with respect to the Lorentz transformation, becomes
X
(4)
∆xν
2 = ∆x1
2 + ∆x2
2 + ∆x3
2 + ∆x4
2 = 0. (22c)
This condition is always satisfied∗
if we satisfy the more general
condition that
s
2 = ∆x1
2 + ∆x2
2 + ∆x3
2 + ∆x4
2
(23)
∗That this specialization lies in the nature of the case will be evident
later.
THE MEANING OF RELATIVITY 34
shall be an invariant with respect to the transformation. This
condition is satisfied only by linear transformations, that is,
transformations of the type
x
0
µ = aµ + bµαxα (24)
in which the summation over the α is to be extended from α = 1
to α = 4. A glance at equations (23) and (24) shows that the
Lorentz transformation so defined is identical with the translational and rotational transformations of the Euclidean geometry,
if we disregard the number of dimensions and the relations of reality. We can also conclude that the coefficients bµα must satisfy
the conditions
bµαbνα = δµν = bαµbαν. (25)
Since the ratios of the xν are real, it follows that all the aµ and
the bµα are real, except a4, b41, b42, b43, b14, b24 and b34, which
are purely imaginary.
Special Lorentz Transformation. We obtain the simplest
transformations of the type of (24) and (25) if only two of the
co-ordinates are to be transformed, and if all the aµ, which determine the new origin, vanish. We obtain then for the indices
1 and 2, on account of the three independent conditions which
the relations (25) furnish,
x
0
1 = x1 cos φ − x2 sin φ,
x
0
2 = x1 sin φ + x2 cos φ,
x
0
3 = x3,
x
0
4 = x4.
(26)
SPECIAL RELATIVITY 35
This is a simple rotation in space of the (space) co-ordinate
system about x3-axis. We see that the rotational transformation
in space (without the time transformation) which we studied
before is contained in the Lorentz transformation as a special
case. For the indices 1 and 4 we obtain, in an analogous manner,
x
0
1 = x1 cos ψ − x4 sin ψ,
x
0
4 = x1 sin ψ + x4 cos ψ,
x
0
2 = x2,
x
0
3 = x3.
(26a)
On account of the relations of reality ψ must be taken as
imaginary. To interpret these equations physically, we introduce
the real light-time l and the velocity v of K0
relatively to K,
instead of the imaginary angle ψ. We have, first,
x
0
1 = x1 cos ψ − ilsin ψ,
l
0 = −ix1 sin ψ + l cos ψ.
Since for the origin of K0
i.e., for x1 = 0, we must have x1 = vl,
it follows from the first of these equations that
v = itan ψ, (27)
and also
sin ψ =
−iv
√
1 − v
2
,
cos ψ =
1
√
1 − v
2
,
(28)
THE MEANING OF RELATIVITY 36
so that we obtain
x
0
1 =
x1 − vl
√
1 − v
2
,
l
0 =
l − vx1
√
1 − v
2
,
x
0
2 = x2,
x
0
3 = x3.
(29)
These equations form the well-known special Lorentz transformation, which in the general theory represents a rotation,
through an imaginary angle, of the four-dimensional system of
co-ordinates. If we introduce the ordinary time t, in place of the
light-time l, then in (29) we must replace l by ct and v by v
c
.
We must now fill in a gap. From the principle of the constancy of the velocity of light it follows that the equation
X∆xν
2 = 0
has a significance which is independent of the choice of the inertial system; but the invariance of the quantity P∆xν
2
does
not at all follow from this. This quantity might be transformed
with a factor. This depends upon the fact that the right-hand
side of (29) might be multiplied by a factor λ, independent of v.
But the principle of relativity does not permit this factor to be
different from 1, as we shall now show. Let us assume that we
have a rigid circular cylinder moving in the direction of its axis.
If its radius, measured at rest with a unit measuring rod is equal
to R0, its radius R in motion, might be different from R0, since
the theory of relativity does not make the assumption that the
shape of bodies with respect to a space of reference is independent of their motion relatively to this space of reference. But
SPECIAL RELATIVITY 37
all directions in space must be equivalent to each other. R may
therefore depend upon the magnitude q of the velocity, but not
upon its direction; R must therefore be an even function of q. If
the cylinder is at rest relatively to K0
the equation of its lateral
surface is
x
02 + y
02 = R0
2
.
If we write the last two equations of (29) more generally
x
0
2 = λx2,
x
0
3 = λx3,
then the lateral surface of the cylinder referred to K satisfies the
equation
x
2 + y
2 =
R0
2
λ
2
.
The factor λ therefore measures the lateral contraction of the
cylinder, and can thus, from the above, be only an even function
of v.
If we introduce a third system of co-ordinates, K00, which
moves relatively to K0 with velocity v in the direction of the
negative x-axis of K, we obtain, by applying (29) twice,
x
00
1 = λ(v)λ(−v)x1,
x
00
2 = λ(v)λ(−v)x2,
x
00
3 = λ(v)λ(−v)x3,
l
00 = λ(v)λ(−v)l.
Now, since λ(v) must be equal to λ(−v), and since we assume
that we use the same measuring rods in all the systems, it follows that the transformation of K00 to K must be the identical
THE MEANING OF RELATIVITY 38
transformation (since the possibility λ = −1 does not need to
be considered). It is essential for these considerations to assume
that the behaviour of the measuring rods does not depend upon
the history of their previous motion.
Moving Measuring Rods and Clocks. At the definite Ktime, l = 0, the position of the points given by the integers
x
0
1 = n, is with respect to K, given by x1 = n
√
1 − v
2
; this follows from the first of equations (29) and expresses the Lorentz
contraction. A clock at rest at the origin x1 = 0 of K, whose
beats are characterized by l = n, will, when observed from K0
,
have beats characterized by
l
0 =
n
√
1 − v
2
;
this follows from the second of equations (29) and shows that
the clock goes slower than if it were at rest relatively to K0
.
These two consequences, which hold, mutatis mutandis, for every system of reference, form the physical content, free from
convention, of the Lorentz transformation.
Addition Theorem for Velocities. If we combine two special
Lorentz transformations with the relative velocities v1 and v2,
then the velocity of the single Lorentz transformation which
takes the place of the two separate ones is, according to (27),
given by
v12 = itan(ψ1 + ψ2) = i
tan ψ1 + tan ψ2
1 − tan ψ1 tan ψ2
=
v1 + v2
1 + v1v2
. (30)
SPECIAL RELATIVITY 39
General Statements about the Lorentz Transformation and
its Theory of Invariants. The whole theory of invariants of the
special theory of relativity depends upon the invariant s
2
(23).
Formally, it has the same rˆole in the four-dimensional space-time
continuum as the invariant ∆x1
2+∆x2
2+∆x3
2
in the Euclidean
geometry and in the pre-relativity physics. The latter quantity
is not an invariant with respect to all the Lorentz transformations; the quantity s
2 of equation (23) assumes the rˆole of this
invariant. With respect to an arbitrary inertial system, s
2 may
be determined by measurements; with a given unit of measure
it is a completely determinate quantity, associated with an arbitrary pair of events.
The invariant s
2 differs, disregarding the number of dimensions, from the corresponding invariant of the Euclidean geometry in the following points. In the Euclidean geometry s
2
is
necessarily positive; it vanishes only when the two points concerned come together. On the other hand, from the vanishing
of
s
2 =
X
(4)
∆xν
2 = ∆x1
2 + ∆x2
2 + ∆x3
2 − ∆t
2
it cannot be concluded that the two space-time points fall together; the vanishing of this quantity s
2
, is the invariant condition that the two space-time points can be connected by a light
signal in vacuo. If P is a point (event) represented in the fourdimensional space of the x1, x2, x3, l, then all the “points” which
can be connected to P by means of a light signal lie upon the
cone s
2 = 0 (compare Fig. 1, in which the dimension x3 is suppressed). The “upper” half of the cone may contain the “points”
to which light signals can be sent from P; then the “lower” half
THE MEANING OF RELATIVITY 40
x1
x2
l
Fig. 1.
of the cone will contain the “points” from which light signals
can be sent to P. The points P
0
enclosed by the conical surface
furnish, with P, a negative s
2
; P P0
, as well as P
0P is then, according to Minkowski, of the nature of a time. Such intervals
represent elements of possible paths of motion, the velocity being less than that of light.∗
In this case the l-axis may be drawn
∗That material velocities exceeding that of light are not possible, follows from the appearance of the radical √
1 − v
2 in the special Lorentz
SPECIAL RELATIVITY 41
in the direction of P P0 by suitably choosing the state of motion
of the inertial system. If P
0
lies outside of the “light-cone” then
P P0
is of the nature of a space; in this case, by properly choosing
the inertial system, ∆l can be made to vanish.
By the introduction of the imaginary time variable, x4 =
il, Minkowski has made the theory of invariants for the fourdimensional continuum of physical phenomena fully analogous
to the theory of invariants for the three-dimensional continuum
of Euclidean space. The theory of four-dimensional tensors of
special relativity differs from the theory of tensors in threedimensional space, therefore, only in the number of dimensions
and the relations of reality.
A physical entity which is specified by four quantities, Aν,
in an arbitrary inertial system of the x1, x2, x3, x4, is called
a 4-vector, with the components Aν, if the Aν correspond in
their relations of reality and the properties of transformation to
the ∆xν; it may be of the nature of a space or of a time. The
sixteen quantities Aµν then form the components of a tensor of
the second rank, if they transform according to the scheme
A
0
µν = bµαbνβAαβ.
It follows from this that the Aµν behave, with respect to
their properties of transformation and their properties of reality, as the products of components, UµVν, of two 4-vectors,
(U) and (V ). All the components are real except those which
contain the index 4 once, those being purely imaginary. Tensors
of the third and higher ranks may be defined in an analogous
way. The operations of addition, subtraction, multiplication,
transformation (29).
THE MEANING OF RELATIVITY 42
contraction and differentiation for these tensors are wholly
analogous to the corresponding operations for tensors in threedimensional space.
Before we apply the tensor theory to the four-dimensional
space-time continuum, we shall examine more particularly the
skew-symmetrical tensors. The tensor of the second rank has, in
general, 16 = 4·4 components. In the case of skew-symmetry the
components with two equal indices vanish, and the components
with unequal indices are equal and opposite in pairs. There
exist, therefore, only six independent components, as is the case
in the electromagnetic field. In fact, it will be shown when we
consider Maxwell’s equations that these may be looked upon as
tensor equations, provided we regard the electromagnetic field
as a skew-symmetrical tensor. Further, it is clear that the skewsymmetrical tensor of the third rank (skew-symmetrical in all
pairs of indices) has only four independent components, since
there are only four combinations of three different indices.
We now turn to Maxwell’s equations (19a), (19b), (20a),
(20b), and introduce the notation:∗
φ23 φ31 φ12 φ14 φ24 φ34
h23 h31 h12 − iex − iey − iez
)
(30a)
J1 J2 J3 J4
1
c
ix
1
c
iy
1
c
iz iρ
(31)
with the convention that φµν shall be equal to −φνµ. Then
∗
In order to avoid confusion from now on we shall use the threedimensional space indices, x, y, z instead of 1, 2, 3, and we shall reserve the
numeral indices 1, 2, 3, 4 for the four-dimensional space-time continuum.
SPECIAL RELATIVITY 43
Maxwell’s equations may be combined into the forms
∂φµν
∂xν
= Jµ, (32)
∂φµν
∂xσ
+
∂φνσ
∂xµ
+
∂φσµ
∂xν
= 0, (33)
as one can easily verify by substituting from (30a) and (31).
Equations (32) and (33) have a tensor character, and are
therefore co-variant with respect to Lorentz transformations,
if the φµν and the Jν have a tensor character, which we assume.
Consequently, the laws for transforming these quantities from
one to another allowable (inertial) system of co-ordinates are
uniquely determined. The progress in method which electrodynamics owes to the theory of special relativity lies principally
in this, that the number of independent hypotheses is diminished. If we consider, for example, equations (19a) only from the
standpoint of relativity of direction, as we have done above, we
see that they have three logically independent terms. The way
in which the electric intensity enters these equations appears to
be wholly independent of the way in which the magnetic intensity enters them; it would not be surprising if instead of ∂eµ
∂l ,
we had, say, ∂
2eµ
∂l2
, or if this term were absent. On the other
hand, only two independent terms appear in equation (32). The
electromagnetic field appears as a formal unit; the way in which
the electric field enters this equation is determined by the way in
which the magnetic field enters it. Besides the electromagnetic
field, only the electric current density appears as an independent
entity. This advance in method arises from the fact that the
THE MEANING OF RELATIVITY 44
electric and magnetic fields draw their separate existences from
the relativity of motion. A field which appears to be purely an
electric field, judged from one system, has also magnetic field
components when judged from another inertial system. When
applied to an electromagnetic field, the general law of transformation furnishes, for the special case of the special Lorentz
transformation, the equations
e
0
x = ex h
0
x = hx,
e
0
y =
ey − vhz √
1 − v
2
h
0
y =
hy + vez √
1 − v
2
,
e
0
z =
ez + vhy
√
1 − v
2
h
0
z =
hz − vey
√
1 − v
2
.
(34)
If there exists with respect to K only a magnetic field, h, but
no electric field, e, then with respect to K0
there exists an electric
field e
0 as well, which would act upon an electric particle at rest
relatively to K0
. An observer at rest relatively to K would designate this force as the Biot-Savart force, or the Lorentz electromotive force. It therefore appears as if this electromotive force
had become fused with the electric field intensity into a single
entity.
In order to view this relation formally, let us consider the
expression for the force acting upon unit volume of electricity,
k = ρe + [i, h], (35)
in which i is the vector velocity of electricity, with the velocity
of light as the unit. If we introduce Jµ and φµν according to
(30a) and (31), we obtain for the first component the expression
φ12J2 + φ13J3 + φ14J4.
SPECIAL RELATIVITY 45
Observing that φ11 vanishes on account of the skew-symmetry of
the tensor (φ), the components of k are given by the first three
components of the four-dimensional vector
Kµ = φµνJν, (36)
and the fourth component is given by
K4 = φ41J1 + φ42J2 + φ43J3 = i(exix + eyiy + eziz) = iλ. (37)
There is, therefore, a four-dimensional vector of force per unit
volume, whose first three components, K1, K2, K3, are the ponderomotive force components per unit volume, and whose fourth
component is the rate of working of the field per unit volume,
multiplied by √
−1.
A comparison of (36) and (35) shows that the theory of relativity formally unites the ponderomotive force of the electric
field, ρe, and the Biot-Savart or Lorentz force [i, h].
Mass and Energy. An important conclusion can be drawn
from the existence and significance of the 4-vector Kµ. Let us
imagine a body upon which the electromagnetic field acts for
a time. In the symbolic figure (Fig. 2) Ox1 designates the x1-
axis, and is at the same time a substitute for the three space axes
Ox1, Ox2, Ox3; Ol designates the real time axis. In this diagram
a body of finite extent is represented, at a definite time l, by
the interval AB; the whole space-time existence of the body is
represented by a strip whose boundary is everywhere inclined
less than 45◦
to the l-axis. Between the time sections, l = l1
and l = l2, but not extending to them, a portion of the strip is
shaded. This represents the portion of the space-time manifold
THE MEANING OF RELATIVITY 46
x1
l
l1
l
l2
O
A B
Fig. 2.
in which the electromagnetic field acts upon the body, or upon
the electric charges contained in it, the action upon them being
transmitted to the body. We shall now consider the changes
which take place in the momentum and energy of the body as a
result of this action.
We shall assume that the principles of momentum and
energy are valid for the body. The change in momentum,
∆Ix, ∆Iy, ∆Iz, and the change in energy, ∆E, are then given
SPECIAL RELATIVITY 47
by the expressions
∆Ix =
Z l2
l1
dl Z
kx dx dy dz =
1
i
Z
K1 dx1 dx2 dx3 dx4,
∆Iy =
Z l2
l1
dl Z
ky dx dy dz =
1
i
Z
K2 dx1 dx2 dx3 dx4,
∆Iz =
Z l2
l1
dl Z
kz dx dy dz =
1
i
Z
K3 dx1 dx2 dx3 dx4,
∆E =
Z l2
l1
dl Z
λ dx dy dz =
1
i
Z
1
i
K4 dx1 dx2 dx3 dx4.
Since the four-dimensional element of volume is an invariant,
and (K1, K2, K3, K4) forms a 4-vector, the four-dimensional integral extended over the shaded portion transforms as a 4-vector,
as does also the integral between the limits l1 and l2, because
the portion of the region which is not shaded contributes nothing
to the integral. It follows, therefore, that ∆Ix, ∆Iy, ∆Iz, i∆E
form a 4-vector. Since the quantities themselves transform in
the same way as their increments, it follows that the aggregate
of the four quantities
Ix, Iy, Iz, iE
has itself the properties of a vector; these quantities are referred
to an instantaneous condition of the body (e.g. at the time l =
l1).
This 4-vector may also be expressed in terms of the mass m,
and the velocity of the body, considered as a material particle.
To form this expression, we note first, that
−ds2 = dτ 2 = −(dx1
2 + dx2
2 + dx3
2
) − dx4
2 = dl2
(1 − q
2
) (38)
THE MEANING OF RELATIVITY 48
is an invariant which refers to an infinitely short portion of the
four-dimensional line which represents the motion of the material particle. The physical significance of the invariant dτ may
easily be given. If the time axis is chosen in such a way that it
has the direction of the line differential which we are considering, or, in other words, if we reduce the material particle to rest,
we shall then have dτ = dl; this will therefore be measured by
the light-seconds clock which is at the same place, and at rest
relatively to the material particle. We therefore call τ the proper
time of the material particle. As opposed to dl, dτ is therefore an
invariant, and is practically equivalent to dl for motions whose
velocity is small compared to that of light. Hence we see that
uσ =
dxσ
dτ (39)
has, just as the dxν, the character of a vector; we shall designate (uσ) as the four-dimensional vector (in brief, 4-vector) of
velocity. Its components satisfy, by (38), the condition
Xuσ
2 = −1. (40)
We see that this 4-vector, whose components in the ordinary
notation are
p
qx
1 − q
2
, p
qy
1 − q
2
, p
qz
1 − q
2
,
i
p
1 − q
2
(41)
is the only 4-vector which can be formed from the velocity components of the material particle which are defined in three dimensions by
qx =
dx
dl , qy =
dy
dl , qz =
dz
dl .
SPECIAL RELATIVITY 49
We therefore see that
m
dxµ
dτ
(42)
must be that 4-vector which is to be equated to the 4-vector of
momentum and energy whose existence we have proved above.
By equating the components, we obtain, in three-dimensional
notation,
Ix = p
mqx
1 − q
2
,
Iy = p
mqy
1 − q
2
,
Iz = p
mqz
1 − q
2
,
E =
m
p
1 − q
2
.
(43)
We recognize, in fact, that these components of momentum
agree with those of classical mechanics for velocities which are
small compared to that of light. For large velocities the momentum increases more rapidly than linearly with the velocity, so as
to become infinite on approaching the velocity of light.
If we apply the last of equations (43) to a material particle
at rest (q = 0), we see that the energy, E0, of a body at rest is
equal to its mass. Had we chosen the second as our unit of time,
we would have obtained
E0 = mc2
. (44)
Mass and energy are therefore essentially alike; they are only
different expressions for the same thing. The mass of a body
THE MEANING OF RELATIVITY 50
is not a constant; it varies with changes in its energy.∗ We see
from the last of equations (43) that E becomes infinite when
q approaches 1, the velocity of light. If we develop E in powers
of q
2
, we obtain,
E = m +
m
2
q
2 +
3
8
mq4 + . . . . (45)
The second term of this expansion corresponds to the kinetic
energy of the material particle in classical mechanics.
Equations of Motion of Material Particles. From (43) we
obtain, by differentiating by the time l, and using the principle
of momentum, in the notation of three-dimensional vectors,
K =
d
dl
mq
p
1 − q
2
!
. (46)
This equation, which was previously employed by H. A.
Lorentz for the motion of electrons, has been proved to be true,
with great accuracy, by experiments with β-rays.
Energy Tensor of the Electromagnetic Field. Before the development of the theory of relativity it was known that the principles of energy and momentum could be expressed in a differential form for the electromagnetic field. The four-dimensional
formulation of these principles leads to an important conception,
∗The emission of energy in radioactive processes is evidently connected
with the fact that the atomic weights are not integers. Attempts have been
made to draw conclusions from this concerning the structure and stability
of the atomic nuclei.
SPECIAL RELATIVITY 51
that of the energy tensor, which is important for the further development of the theory of relativity.
If in the expression for the 4-vector of force per unit volume,
Kµ = φµνJν,
using the field equations (32), we express Jν in terms of the
field intensities, φµν, we obtain, after some transformations and
repeated application of the field equations (32) and (33), the
expression
Kµ = −
∂Tµν
∂xν
, (47)
where we have written∗
Tµν = −
1
4
φαβ
2
δµν + φµαφνα. (48)
The physical meaning of equation (47) becomes evident if in
place of this equation we write, using a new notation,
kx = −
∂pxx
∂x −
∂pxy
∂y −
∂pxz
∂z −
∂(ibx)
∂(il)
,
ky = −
∂pyx
∂x −
∂pyy
∂y −
∂pyz
∂z −
∂(iby)
∂(il)
,
kz = −
∂pzx
∂x −
∂pzy
∂y −
∂pzz
∂z −
∂(ibz)
∂(il)
,
iλ = −
∂(isx)
∂x −
∂(isy)
∂y −
∂(isz)
∂z −
∂(−η)
∂(il)
;
(47a)
∗To be summed for the indices α and β.
THE MEANING OF RELATIVITY 52
or, on eliminating the imaginary,
kx = −
∂pxx
∂x −
∂pxy
∂y −
∂pxz
∂z −
∂bx
∂l ,
ky = −
∂pyx
∂x −
∂pyy
∂y −
∂pyz
∂z −
∂by
∂l ,
kz = −
∂pzx
∂x −
∂pzy
∂y −
∂pzz
∂z −
∂bz
∂l ,
λ = −
∂sx
∂x −
∂sy
∂y −
∂sz
∂z −
∂η
∂l .
(47b)
When expressed in the latter form, we see that the first three
equations state the principle of momentum; pxx,. . . , pzx are the
Maxwell stresses in the electromagnetic field, and (bx, by, bz) is
the vector momentum per unit volume of the field. The last of
equations (47b) expresses the energy principle; s is the vector
flow of energy, and η the energy per unit volume of the field. In
fact, we get from (48) by introducing the well-known expressions
for the components of the field intensity from electrodynamics,
pxx = − hxhx +
1
2
(hx
2 + hy
2 + hz
2
)
− exex +
1
2
(ex
2 + ey
2 + ez
2
),
pxy = − hxhy pxz = − hxhz
− exey, − exez,
.
.
.
bx = sx = eyhz − ezhy,
by = sy = ezhx − exhz,
bz = sz = exhy − eyhx,
η = +1
2
(ex
2 + ey
2 + ez
2 + hx
2 + hy
2 + hz
2
).
(48a)
SPECIAL RELATIVITY 53
We conclude from (48) that the energy tensor of the electromagnetic field is symmetrical; with this is connected the fact
that the momentum per unit volume and the flow of energy are
equal to each other (relation between energy and inertia).
We therefore conclude from these considerations that the
energy per unit volume has the character of a tensor. This has
been proved directly only for an electromagnetic field, although
we may claim universal validity for it. Maxwell’s equations determine the electromagnetic field when the distribution of electric charges and currents is known. But we do not know the
laws which govern the currents and charges. We do know, indeed, that electricity consists of elementary particles (electrons,
positive nuclei), but from a theoretical point of view we cannot comprehend this. We do not know the energy factors which
determine the distribution of electricity in particles of definite
size and charge, and all attempts to complete the theory in this
direction have failed. If then we can build upon Maxwell’s equations in general, the energy tensor of the electromagnetic field
is known only outside the charged particles.∗
In these regions,
outside of charged particles, the only regions in which we can believe that we have the complete expression for the energy tensor,
we have, by (47),
∂Tµν
∂xν
= 0. (47c)
∗
It has been attempted to remedy this lack of knowledge by considering
the charged particles as proper singularities. But in my opinion this means
giving up a real understanding of the structure of matter. It seems to me
much better to give in to our present inability rather than to be satisfied
by a solution that is only apparent.
THE MEANING OF RELATIVITY 54
General Expressions for the Conservation Principles. We
can hardly avoid making the assumption that in all other cases,
also, the space distribution of energy is given by a symmetrical
tensor, Tµν, and that this complete energy tensor everywhere
satisfies the relation (47c). At any rate we shall see that by
means of this assumption we obtain the correct expression for
the integral energy principle.
Let us consider a spatially bounded, closed system, which,
four-dimensionally, we may represent as a strip, outside of which
the Tµν vanish. Integrate equation (47c) over a space section.
Since the integrals of ∂Tµ1
∂x1
,
∂Tµ2
∂x2
and ∂Tµ3
∂x3
vanish because
the Tµν vanish at the limits of integration, we obtain
∂
∂l Z
Tµ4 dx1 dx2 dx3
= 0. (49)
Inside the parentheses are the expressions for the momentum of
the whole system, multiplied by i, together with the negative
energy of the system, so that (49) expresses the conservation
principles in their integral form. That this gives the right conception of energy and the conservation principles will be seen
from the following considerations.
Phenomenological Representation of the
Energy Tensor of Matter.
Hydrodynamical Equations. We know that matter is built
up of electrically charged particles, but we do not know the laws
which govern the constitution of these particles. In treating mechanical problems, we are therefore obliged to make use of an
SPECIAL RELATIVITY 55
x1
l
Fig. 3.
inexact description of matter, which corresponds to that of classical mechanics. The density σ, of a material substance and the
hydrodynamical pressures are the fundamental concepts upon
which such a description is based.
Let σ0 be the density of matter at a place, estimated with
reference to a system of co-ordinates moving with the matter.
Then σ0, the density at rest, is an invariant. If we think of the
matter in arbitrary motion and neglect the pressures (particles
of dust in vacuo, neglecting the size of the particles and the
temperature), then the energy tensor will depend only upon the
THE MEANING OF RELATIVITY 56
velocity components, uν and σ0. We secure the tensor character
of Tµν by putting
Tµν = σ0uµuν, (50)
in which the uµ, in the three-dimensional representation, are
given by (41). In fact, it follows from (50) that for q = 0, T44 =
−σ0 (equal to the negative energy per unit volume), as it should,
according to the theorem of the equivalence of mass and energy,
and according to the physical interpretation of the energy tensor
given above. If an external force (four-dimensional vector, Kµ)
acts upon the matter, by the principles of momentum and energy
the equation
Kµ =
∂Tµν
∂xν
must hold. We shall now show that this equation leads to the
same law of motion of a material particle as that already obtained. Let us imagine the matter to be of infinitely small extent
in space, that is, a four-dimensional thread; then by integration
over the whole thread with respect to the space co-ordinates
x1, x2, x3, we obtain
Z
K1 dx1 dx2 dx3 =
Z
∂T14
∂x4
dx1 dx2 dx3
= −i
d
dl Z
σ0
dx1
dτ
dx4
dτ dx1 dx2 dx3
.
Now R
dx1 dx2 dx3 dx4 is an invariant, as is, therefore, also
R
σ0 dx1 dx2 dx3 dx4. We shall calculate this integral, first with
respect to the inertial system which we have chosen, and second,
with respect to a system relatively to which the matter has the
velocity zero. The integration is to be extended over a filament
SPECIAL RELATIVITY 57
of the thread for which σ0 may be regarded as constant over the
whole section. If the space volumes of the filament referred to
the two systems are dV and dV0 respectively, then we have
Z
σ0 dV dl =
Z
σ0 dV0 dτ
and therefore also
Z
σ0 dV =
Z
σ0 dV0
dτ
dl =
Z
dm i dτ
dx4
.
If we substitute the right-hand side for the left-hand side in
the former integral, and put dx1
dτ outside the sign of integration,
we obtain,
Kx =
d
dl
m
dx1
dτ
=
d
dl
p
mqx
1 − q
2
!
.
We see, therefore, that the generalized conception of the energy
tensor is in agreement with our former result.
The Eulerian Equations for Perfect Fluids. In order to get
nearer to the behaviour of real matter we must add to the energy
tensor a term which corresponds to the pressures. The simplest
case is that of a perfect fluid in which the pressure is determined
by a scalar p. Since the tangential stresses pxy, etc., vanish in
this case, the contribution to the energy tensor must be of the
form pδνµ. We must therefore put
Tµν = σuµuν + pδµν. (51)
THE MEANING OF RELATIVITY 58
At rest, the density of the matter, or the energy per unit volume,
is in this case, not σ but σ − p. For
−T44 = −σ
dx4
dτ
dx4
dτ − pδ44 = σ − p.
In the absence of any force, we have
∂Tµν
∂xν
= σuν
∂uµ
∂xν
+ uµ
∂(σuν)
∂xν
+
∂p
∂xµ
= 0.
If we multiply this equation by uµ
=
dxµ
dτ
and sum for the
µ’s we obtain, using (40),
−
∂(σuν)
∂xν
+
dp
dτ = 0, (52)
where we have put ∂p
∂xµ
dxµ
dτ =
dp
dτ . This is the equation of
continuity, which differs from that of classical mechanics by
the term dp
dτ , which, practically, is vanishingly small. Observing (52), the conservation principles take the form
σ
duµ
dτ + uµ
dp
dτ +
∂p
∂xµ
= 0. (53)
The equations for the first three indices evidently correspond to
the Eulerian equations. That the equations (52) and (53) correspond, to a first approximation, to the hydrodynamical equations of classical mechanics, is a further confirmation of the generalized energy principle. The density of matter and of energy
has the character of a symmetrical tensor.
LECTURE III
THE GENERAL THEORY OF RELATIVITY
All of the previous considerations have been based upon the
assumption that all inertial systems are equivalent for the description of physical phenomena, but that they are preferred, for
the formulation of the laws of nature, to spaces of reference in a
different state of motion. We can think of no cause for this preference for definite states of motion to all others, according to our
previous considerations, either in the perceptible bodies or in the
concept of motion; on the contrary, it must be regarded as an independent property of the space-time continuum. The principle
of inertia, in particular, seems to compel us to ascribe physically objective properties to the space-time continuum. Just as
it was necessary from the Newtonian standpoint to make both
the statements, tempus est absolutum, spatium est absolutum, so
from the standpoint of the special theory of relativity we must
say, continuum spatii et temporis est absolutum. In this latter
statement absolutum means not only “physically real,” but also
“independent in its physical properties, having a physical effect,
but not itself influenced by physical conditions.”
As long as the principle of inertia is regarded as the keystone of physics, this standpoint is certainly the only one which
is justified. But there are two serious criticisms of the ordinary
conception. In the first place, it is contrary to the mode of thinking in science to conceive of a thing (the space-time continuum)
which acts itself, but which cannot be acted upon. This is the
reason why E. Mach was led to make the attempt to eliminate
space as an active cause in the system of mechanics. Accord59
THE MEANING OF RELATIVITY 60
ing to him, a material particle does not move in unaccelerated
motion relatively to space, but relatively to the centre of all the
other masses in the universe; in this way the series of causes of
mechanical phenomena was closed, in contrast to the mechanics
of Newton and Galileo. In order to develop this idea within the
limits of the modern theory of action through a medium, the
properties of the space-time continuum which determine inertia
must be regarded as field properties of space, analogous to the
electromagnetic field. The concepts of classical mechanics afford no way of expressing this. For this reason Mach’s attempt
at a solution failed for the time being. We shall come back to
this point of view later. In the second place, classical mechanics
indicates a limitation which directly demands an extension of
the principle of relativity to spaces of reference which are not
in uniform motion relatively to each other. The ratio of the
masses of two bodies is defined in mechanics in two ways which
differ from each other fundamentally; in the first place, as the
reciprocal ratio of the accelerations which the same motional
force imparts to them (inert mass), and in the second place, as
the ratio of the forces which act upon them in the same gravitational field (gravitational mass). The equality of these two
masses, so differently defined, is a fact which is confirmed by
experiments of very high accuracy (experiments of E¨otv¨os), and
classical mechanics offers no explanation for this equality. It is,
however, clear that science is fully justified in assigning such a
numerical equality only after this numerical equality is reduced
to an equality of the real nature of the two concepts.
That this object may actually be attained by an extension
of the principle of relativity, follows from the following consideration. A little reflection will show that the theorem of the
THE GENERAL THEORY 61
equality of the inert and the gravitational mass is equivalent
to the theorem that the acceleration imparted to a body by a
gravitational field is independent of the nature of the body. For
Newton’s equation of motion in a gravitational field, written out
in full, is
(Inert mass) · (Acceleration) = (Intensity of the
gravitational field) · (Gravitational mass).
It is only when there is numerical equality between the inert
and gravitational mass that the acceleration is independent of
the nature of the body. Let now K be an inertial system. Masses
which are sufficiently far from each other and from other bodies are then, with respect to K, free from acceleration. We shall
also refer these masses to a system of co-ordinates K0
, uniformly
accelerated with respect to K. Relatively to K0 all the masses
have equal and parallel accelerations; with respect to K0
they
behave just as if a gravitational field were present and K0 were
unaccelerated. Overlooking for the present the question as to the
“cause” of such a gravitational field, which will occupy us later,
there is nothing to prevent our conceiving this gravitational field
as real, that is, the conception that K0
is “at rest” and a gravitational field is present we may consider as equivalent to the conception that only K is an “allowable” system of co-ordinates and
no gravitational field is present. The assumption of the complete
physical equivalence of the systems of co-ordinates, K and K0
,
we call the “principle of equivalence;” this principle is evidently
intimately connected with the theorem of the equality between
the inert and the gravitational mass, and signifies an extension
of the principle of relativity to co-ordinate systems which are in
THE MEANING OF RELATIVITY 62
non-uniform motion relatively to each other. In fact, through
this conception we arrive at the unity of the nature of inertia
and gravitation. For according to our way of looking at it, the
same masses may appear to be either under the action of inertia alone (with respect to K) or under the combined action of
inertia and gravitation (with respect to K0
). The possibility of
explaining the numerical equality of inertia and gravitation by
the unity of their nature gives to the general theory of relativity,
according to my conviction, such a superiority over the conceptions of classical mechanics, that all the difficulties encountered
in development must be considered as small in comparison.
What justifies us in dispensing with the preference for inertial systems over all other co-ordinate systems, a preference
that seems so securely established by experiment based upon
the principle of inertia? The weakness of the principle of inertia
lies in this, that it involves an argument in a circle: a mass moves
without acceleration if it is sufficiently far from other bodies; we
know that it is sufficiently far from other bodies only by the fact
that it moves without acceleration. Are there, in general, any
inertial systems for very extended portions of the space-time
continuum, or, indeed, for the whole universe? We may look
upon the principle of inertia as established, to a high degree of
approximation, for the space of our planetary system, provided
that we neglect the perturbations due to the sun and planets.
Stated more exactly, there are finite regions, where, with respect
to a suitably chosen space of reference, material particles move
freely without acceleration, and in which the laws of the special theory of relativity, which have been developed above, hold
with remarkable accuracy. Such regions we shall call “Galilean
regions.” We shall proceed from the consideration of such re-
THE GENERAL THEORY 63
gions as a special case of known properties.
The principle of equivalence demands that in dealing with
Galilean regions we may equally well make use of non-inertial
systems, that is, such co-ordinate systems as, relatively to inertial systems, are not free from acceleration and rotation. If,
further, we are going to do away completely with the difficult
question as to the objective reason for the preference of certain
systems of co-ordinates, then we must allow the use of arbitrarily moving systems of co-ordinates. As soon as we make this
attempt seriously we come into conflict with that physical interpretation of space and time to which we were led by the special
theory of relativity. For let K0 be a system of co-ordinates whose
z
0
-axis coincides with the z-axis of K, and which rotates about
the latter axis with constant angular velocity. Are the configurations of rigid bodies, at rest relatively to K0
, in accordance with
the laws of Euclidean geometry? Since K0
is not an inertial system, we do not know directly the laws of configuration of rigid
bodies with respect to K0
, nor the laws of nature, in general. But
we do know these laws with respect to the inertial system K,
and we can therefore estimate them with respect to K0
. Imagine
a circle drawn about the origin in the x
0
-y
0 plane of K0
, and a
diameter of this circle. Imagine, further, that we have given a
large number of rigid rods, all equal to each other. We suppose
these laid in series along the periphery and the diameter of the
circle, at rest relatively to K0
. If U is the number of these rods
along the periphery, D the number along the diameter, then, if
K0 does not rotate relatively to K, we shall have
U
D
= π.
THE MEANING OF RELATIVITY 64
But if K0
rotates we get a different result. Suppose that at
a definite time t of K we determine the ends of all the rods.
With respect to K all the rods upon the periphery experience
the Lorentz contraction, but the rods upon the diameter do not
experience this contraction (along their lengths!).∗
It therefore
follows that
U
D
> π.
It therefore follows that the laws of configuration of rigid
bodies with respect to K0 do not agree with the laws of configuration of rigid bodies that are in accordance with Euclidean geometry. If, further, we place two similar clocks (rotating with K0
),
one upon the periphery, and the other at the centre of the circle, then, judged from K, the clock on the periphery will go
slower than the clock at the centre. The same thing must take
place, judged from K0
, if we define time with respect to K0
in a
not wholly unnatural way, that is, in such a way that the laws
with respect to K0 depend explicitly upon the time. Space and
time, therefore, cannot be defined with respect to K0 as they
were in the special theory of relativity with respect to inertial
systems. But, according to the principle of equivalence, K0
is
also to be considered as a system at rest, with respect to which
there is a gravitational field (field of centrifugal force, and force
of Coriolis). We therefore arrive at the result: the gravitational
field influences and even determines the metrical laws of the
space-time continuum. If the laws of configuration of ideal rigid
bodies are to be expressed geometrically, then in the presence
∗These considerations assume that the behaviour of rods and clocks
depends only upon velocities, and not upon accelerations, or, at least, that
the influence of acceleration does not counteract that of velocity.
THE GENERAL THEORY 65
of a gravitational field the geometry is not Euclidean.
The case that we have been considering is analogous to that
which is presented in the two-dimensional treatment of surfaces. It is impossible in the latter case also, to introduce coordinates on a surface (e.g. the surface of an ellipsoid) which
have a simple metrical significance, while on a plane the Cartesian co-ordinates, x1, x2, signify directly lengths measured by a
unit measuring rod. Gauss overcame this difficulty, in his theory of surfaces, by introducing curvilinear co-ordinates which,
apart from satisfying conditions of continuity, were wholly arbitrary, and afterwards these co-ordinates were related to the
metrical properties of the surface. In an analogous way we
shall introduce in the general theory of relativity arbitrary coordinates, x1, x2, x3, x4, which shall number uniquely the spacetime points, so that neighbouring events are associated with
neighbouring values of the co-ordinates; otherwise, the choice
of co-ordinates is arbitrary. We shall be true to the principle
of relativity in its broadest sense if we give such a form to the
laws that they are valid in every such four-dimensional system
of co-ordinates, that is, if the equations expressing the laws are
co-variant with respect to arbitrary transformations.
The most important point of contact between Gauss’s theory
of surfaces and the general theory of relativity lies in the metrical properties upon which the concepts of both theories, in the
main, are based. In the case of the theory of surfaces, Gauss’s
argument is as follows. Plane geometry may be based upon the
concept of the distance ds, between two indefinitely near points.
The concept of this distance is physically significant because
the distance can be measured directly by means of a rigid measuring rod. By a suitable choice of Cartesian co-ordinates this
THE MEANING OF RELATIVITY 66
distance may be expressed by the formula ds2 = dx1
2 + dx2
2
.
We may base upon this quantity the concepts of the straight
line as the geodesic (δ
R
ds = 0), the interval, the circle, and the
angle, upon which the Euclidean plane geometry is built. A
geometry may be developed upon another continuously curved
surface, if we observe that an infinitesimally small portion of the
surface may be regarded as plane, to within relatively infinitesimal quantities. There are Cartesian co-ordinates, X1, X2, upon
such a small portion of the surface, and the distance between
two points, measured by a measuring rod, is given by
ds2 = dX1
2 + dX2
2
.
If we introduce arbitrary curvilinear co-ordinates, x1, x2, on the
surface, then dX1, dX2, may be expressed linearly in terms of
dx1, dx2. Then everywhere upon the surface we have
ds2 = g11 dx1
2 + 2g12 dx1 dx2 + g22 dx2
2
,
where g11, g12, g22 are determined by the nature of the surface
and the choice of co-ordinates; if these quantities are known,
then it is also known how networks of rigid rods may be laid
upon the surface. In other words, the geometry of surfaces may
be based upon this expression for ds2
exactly as plane geometry
is based upon the corresponding expression.
There are analogous relations in the four-dimensional spacetime continuum of physics. In the immediate neighbourhood of
an observer, falling freely in a gravitational field, there exists no
gravitational field. We can therefore always regard an infinitesimally small region of the space-time continuum as Galilean.
For such an infinitely small region there will be an inertial system (with the space co-ordinates, X1, X2, X3, and the time
THE GENERAL THEORY 67
co-ordinate X4) relatively to which we are to regard the laws of
the special theory of relativity as valid. The quantity which is
directly measurable by our unit measuring rods and clocks,
dX1
2 + dX2
2 + dX3
2 − dX4
2
,
or its negative,
ds2 = −dX1
2 − dX2
2 − dX3
2 + dX4
2
, (54)
is therefore a uniquely determinate invariant for two neighbouring events (points in the four-dimensional continuum), provided
that we use measuring rods that are equal to each other when
brought together and superimposed, and clocks whose rates are
the same when they are brought together. In this the physical
assumption is essential that the relative lengths of two measuring rods and the relative rates of two clocks are independent, in
principle, of their previous history. But this assumption is certainly warranted by experience; if it did not hold there could be
no sharp spectral lines; for the single atoms of the same element
certainly do not have the same history, and it would be absurd
to suppose any relative difference in the structure of the single
atoms due to their previous history if the mass and frequencies
of the single atoms of the same element were always the same.
Space-time regions of finite extent are, in general, not
Galilean, so that a gravitational field cannot be done away
with by any choice of co-ordinates in a finite region. There
is, therefore, no choice of co-ordinates for which the metrical
relations of the special theory of relativity hold in a finite region. But the invariant ds always exists for two neighbouring
points (events) of the continuum. This invariant ds may be
THE MEANING OF RELATIVITY 68
expressed in arbitrary co-ordinates. If one observes that the
local dXν may be expressed linearly in terms of the co-ordinate
differentials dxν, ds2 may be expressed in the form
ds2 = gµν dxµ dxν. (55)
The functions gµν describe, with respect to the arbitrarily chosen system of co-ordinates, the metrical relations of the
space-time continuum and also the gravitational field. As in
the special theory of relativity, we have to discriminate between
time-like and space-like line elements in the four-dimensional
continuum; owing to the change of sign introduced, time-like line
elements have a real, space-like line elements an imaginary ds.
The time-like ds can be measured directly by a suitably chosen
clock.
According to what has been said, it is evident that the formulation of the general theory of relativity assumes a generalization
of the theory of invariants and the theory of tensors; the question is raised as to the form of the equations which are co-variant
with respect to arbitrary point transformations. The generalized
calculus of tensors was developed by mathematicians long before the theory of relativity. Riemann first extended Gauss’s
train of thought to continua of any number of dimensions; with
prophetic vision he saw the physical meaning of this generalization of Euclid’s geometry. Then followed the development of
the theory in the form of the calculus of tensors, particularly by
Ricci and Levi-Civita. This is the place for a brief presentation
of the most important mathematical concepts and operations of
this calculus of tensors.
We designate four quantities, which are defined as functions
of the xν with respect to every system of co-ordinates, as com-
THE GENERAL THEORY 69
ponents, Aν
, of a contra-variant vector, if they transform in a
change of co-ordinates as the co-ordinate differentials dxν. We
therefore have
A
0µ =
∂x0
µ
∂xν
A
ν
. (56)
Besides these contra-variant vectors, there are also co-variant
vectors. If Bν are the components of a co-variant vector, these
vectors are transformed according to the rule
B
0
µ =
∂xν
∂x0
µ
Bν. (57)
The definition of a co-variant vector is chosen in such a way that
a co-variant vector and a contra-variant vector together form a
scalar according to the scheme,
φ = BνA
ν
(summed over the ν).
Accordingly,
B
0
µA
0µ =
∂xα
∂x0
µ
∂x0
µ
∂xβ
BαA
β = BαA
α
.
In particular, the derivatives ∂φ
∂xα
of a scalar φ, are components
of a co-variant vector, which, with the co-ordinate differentials,
form the scalar ∂φ
∂xα
dxα; we see from this example how natural
is the definition of the co-variant vectors.
There are here, also, tensors of any rank, which may have
co-variant or contra-variant character with respect to each index; as with vectors, the character is designated by the position
THE MEANING OF RELATIVITY 70
of the index. For example, Aν
µ denotes a tensor of the second
rank, which is co-variant with respect to the index µ, and contravariant with respect to the index ν. The tensor character indicates that the equation of transformation is
A
0ν
µ =
∂xα
∂x0
µ
∂x0
ν
∂xβ
A
β
α
. (58)
Tensors may be formed by the addition and subtraction of
tensors of equal rank and like character, as in the theory of
invariants of orthogonal linear substitutions, for example,
A
ν
µ + B
ν
µ = C
ν
µ
. (59)
The proof of the tensor character of C
ν
µ depends upon (58).
Tensors may be formed by multiplication, keeping the character of the indices, just as in the theory of invariants of linear
orthogonal transformations, for example,
A
ν
µBστ = C
ν
µστ . (60)
The proof follows directly from the rule of transformation.
Tensors may be formed by contraction with respect to two
indices of different character, for example,
A
µ
µστ = Bστ . (61)
The tensor character of Aµ
µστ determines the tensor character
of Bστ . Proof—
A
0µ
µστ =
∂xα
∂x0
µ
∂x0
µ
∂xβ
∂xs
∂x0
σ
∂xt
∂x0
τ
A
β
αst =
∂xs
∂x0
σ
∂xt
∂x0
τ
A
α
αst.
THE GENERAL THEORY 71
The properties of symmetry and skew-symmetry of a tensor with respect to two indices of like character have the same
significance as in the theory of invariants.
With this, everything essential has been said with regard to
the algebraic properties of tensors.
The Fundamental Tensor. It follows from the invariance
of ds2
for an arbitrary choice of the dxν, in connexion with
the condition of symmetry consistent with (55), that the gµν
are components of a symmetrical co-variant tensor (Fundamental Tensor). Let us form the determinant, g, of the gµν, and
also the minors, divided by g, corresponding to the single gµν.
These minors, divided by g, will be denoted by g
µν, and their
co-variant character is not yet known. Then we have
gµαg
µβ = δ
β
α =
(
1 if α = β,
0 if α 6= β.
(62)
If we form the infinitely small quantities (co-variant vectors)
dξµ = gµα dxα, (63)
multiply by g
µβ and sum over the µ, we obtain, by the use
of (62),
dxβ = g
βµ dξµ. (64)
Since the ratios of the dξµ are arbitrary, and the dxβ as well as
the dxµ are components of vectors, it follows that the g
µν are the
components of a contra-variant tensor∗
(contra-variant fundamental tensor). The tensor character of δ
β
α
(mixed fundamental
∗
If we multiply (64) by ∂x0
α
∂xβ
, sum over the β, and replace the dξµ by
THE MEANING OF RELATIVITY 72
tensor) accordingly follows, by (62). By means of the fundamental tensor, instead of tensors with co-variant index character, we
can introduce tensors with contra-variant index character, and
conversely. For example,
A
µ = g
µαAα,
Aµ = gµαA
α
,
T
σ
µ = g
σνTµν.
Volume Invariants. The volume element
Z
dx1 dx2 dx3 dx4 = dx
is not an invariant. For by Jacobi’s theorem,
dx0 =
dx0
µ
dxν
dx. (65)
But we can complement dx so that it becomes an invariant. If
we form the determinant of the quantities
g
0
µν =
∂xα
∂x0
µ
∂xβ
∂x0
ν
gαβ,
a transformation to the accented system, we obtain
dx0
α =
∂x0
σ
∂xµ
∂x0
α
∂xβ
g
µβ dξ0
σ
.
The statement made above follows from this, since, by (64), we must also
have dx0
α = g
σα0
dξ0
α, and both equations must hold for every choice of
the dξ0
σ
.
THE GENERAL THEORY 73
we obtain, by a double application of the theorem of multiplication of determinants,
g
0 = |g
0
µν| =
∂xν
∂x0
µ
2
· |gµν| =
∂x0
µ
∂xν
−2
g. (66)
We therefore get the invariant,
p
g
0 dx0 =
p
g dx.
Formation of Tensors by Differentiation. Although the algebraic operations of tensor formation have proved to be as
simple as in the special case of invariance with respect to linear orthogonal transformations, nevertheless in the general case,
the invariant differential operations are, unfortunately, considerably more complicated. The reason for this is as follows. If
Aµ
is a contra-variant vector, the coefficients of its transformation,
∂x0
µ
∂xν
, are independent of position only if the transformation
is a linear one. For then the vector components, Aµ +
∂Aµ
∂xα
dxα,
at a neighbouring point transform in the same way as the Aµ
,
from which follows the vector character of the vector differentials, and the tensor character of ∂Aµ
∂xα
. But if the ∂x0
µ
∂xν
are
variable this is no longer true.
That there are, nevertheless, in the general case, invariant
differential operations for tensors, is recognized most satisfactorily in the following way, introduced by Levi-Civita and Weyl.
Let (Aµ
) be a contra-variant vector whose components are given
with respect to the co-ordinate system of the xν. Let P1 and P2
THE MEANING OF RELATIVITY 74
be two infinitesimally near points of the continuum. For the infinitesimal region surrounding the point P1, there is, according
to our way of considering the matter, a co-ordinate system of
the Xν (with imaginary X4-co-ordinate) for which the continuum is Euclidean. Let A
µ
(1) be the co-ordinates of the vector at
the point P1. Imagine a vector drawn at the point P2, using the
local system of the Xν, with the same co-ordinates (parallel vector through P2), then this parallel vector is uniquely determined
by the vector at P1 and the displacement. We designate this operation, whose uniqueness will appear in the sequel, the parallel
displacement of the vector Aµ from P1 to the infinitesimally near
point P2. If we form the vector difference of the vector (Aµ
) at
the point P2 and the vector obtained by parallel displacement
from P1 to P2, we get a vector which may be regarded as the
differential of the vector (Aµ
) for the given displacement (dxν).
This vector displacement can naturally also be considered
with respect to the co-ordinate system of the xν. If Aν are the
co-ordinates of the vector at P1, Aν + δAν
the co-ordinates of
the vector displaced to P2 along the interval (dxν), then the δAν
do not vanish in this case. We know of these quantities, which
do not have a vector character, that they must depend linearly
and homogeneously upon the dxν and the Aν
. We therefore put
δAν = −Γ
ν
αβA
α
dxβ. (67)
In addition, we can state that the Γν
αβ must be symmetrical
with respect to the indices α and β. For we can assume from
a representation by the aid of a Euclidean system of local coordinates that the same parallelogram will be described by the
displacement of an element d
(1)xν along a second element d
(2)xν
THE GENERAL THEORY 75
as by a displacement of d
(2)xν along d
(1)xν. We must therefore
have
d
(2)xν + (d
(1)xν − Γ
ν
αβ d
(1)xα d
(2)xβ)
= d
(1)xν + (d
(2)xν − Γ
ν
αβ d
(2)xα d
(1)xβ).
The statement made above follows from this, after interchanging
the indices of summation, α and β, on the right-hand side.
Since the quantities gµν determine all the metrical properties
of the continuum, they must also determine the Γν
αβ. If we
consider the invariant of the vector Aν
, that is, the square of its
magnitude,
gµνA
µA
ν
,
which is an invariant, this cannot change in a parallel displacement. We therefore have
0 = δ(gµνA
µA
ν
) = ∂gµν
∂xα
A
µA
ν
dxα + gµνA
µ
δAν + gµνA
ν
δAµ
or, by (67),
∂gµν
∂xα
− gµβΓ
β
να − gνβΓ
β
µα
A
µA
ν
dxα = 0.
Owing to the symmetry of the expression in the brackets
with respect to the indices µ and ν, this equation can be valid
for an arbitrary choice of the vectors (Aµ
) and dxν only when
the expression in the brackets vanishes for all combinations of
the indices. By a cyclic interchange of the indices µ, ν, α, we
obtain thus altogether three equations, from which we obtain,
on taking into account the symmetrical property of the Γα
µν,
µν
α
= gαβΓ
β
µν, (68)
THE MEANING OF RELATIVITY 76
in which, following Christoffel, the abbreviation has been used,
µν
α
=
1
2
∂gµα
∂xν
+
∂gνα
∂xµ
−
∂gµν
∂xα
. (69)
If we multiply (68) by g
ασ and sum over the α, we obtain
Γ
α
µν =
1
2
g
σα
∂gµα
∂xν
+
∂gνα
∂xµ
−
∂gµν
∂xα
=
µν
σ
, (70)
in which µν
σ
is the Christoffel symbol of the second kind.
Thus the quantities Γ are deduced from the gµν. Equations
(67) and (70) are the foundation for the following discussion.
Co-variant Differentiation of Tensors. If (Aµ + δAµ
) is
the vector resulting from an infinitesimal parallel displacement
from P1 to P2, and (Aµ + dAµ
) the vector Aµ at the point P2,
then the difference of these two,
dAµ − δAµ =
∂Aµ
∂xσ
+ Γµ
σαA
α
dxσ,
is also a vector. Since this is the case for an arbitrary choice of
the dxσ, it follows that
A
µ
; σ =
∂Aµ
∂xσ
+ Γµ
σαA
α
(71)
is a tensor, which we designate as the co-variant derivative of
the tensor of the first rank (vector). Contracting this tensor, we
obtain the divergence of the contra-variant tensor Aµ
. In this
we must observe that according to (70),
Γ
σ
µσ =
1
2
g
σα ∂gσα
∂xµ
=
1
√g
∂
√g
∂xµ
. (72)
THE GENERAL THEORY 77
If we put, further,
A
µ√
g = A
µ
, (73)
a quantity designated by Weyl as the contra-variant tensor density∗ of the first rank, it follows that,
A =
∂ A
µ
∂xµ
(74)
is a scalar density.
We get the law of parallel displacement for the co-variant
vector Bµ by stipulating that the parallel displacement shall be
effected in such a way that the scalar
φ = A
µBµ
remains unchanged, and that therefore
A
µ
δBµ + Bµ δAµ
vanishes for every value assigned to (Aµ
). We therefore get
δBµ = Γα
µσAα dxσ. (75)
From this we arrive at the co-variant derivative of the covariant vector by the same process as that which led to (71),
Bµ; σ =
∂Bµ
∂xσ
− Γ
α
µσBα. (76)
∗This expression is justified, in that Aµ√g dx = A
µ
dx has a tensor
character. Every tensor, when multiplied by √g, changes into a tensor
density. We employ capital Gothic letters for tensor densities.
THE MEANING OF RELATIVITY 78
By interchanging the indices µ and β, and subtracting, we get
the skew-symmetrical tensor,
φµσ =
∂Bµ
∂xσ
−
∂Bσ
∂xµ
. (77)
For the co-variant differentiation of tensors of the second
and higher ranks we may use the process by which (75) was
deduced. Let, for example, (Aστ ) be a co-variant tensor of the
second rank. Then AστE
σF
τ
is a scalar, if E and F are vectors.
This expression must not be changed by the δ-displacement;
expressing this by a formula, we get, using (67), δAστ , whence
we get the desired co-variant derivative,
Aστ; ρ =
∂Aστ
∂xρ
− Γ
α
σρAατ − Γ
α
τ ρAσα. (78)
In order that the general law of co-variant differentiation of
tensors may be clearly seen, we shall write down two co-variant
derivatives deduced in an analogous way:
A
τ
σ; ρ =
∂Aτ
σ
∂xρ
− Γ
α
σρA
τ
α + Γτ
αρA
α
σ
, (79)
A
στ
; ρ =
∂Aστ
∂xρ
+ Γσ
αρA
ατ + Γτ
αρA
σα
. (80)
The general law of formation now becomes evident. From these
formulæ we shall deduce some others which are of interest for
the physical applications of the theory.
In case Aστ is skew-symmetrical, we obtain the tensor
Aστ ρ =
∂Aστ
∂xρ
+
∂Aτ ρ
∂xσ
+
∂Aρσ
∂xτ
, (81)
THE GENERAL THEORY 79
which is skew-symmetrical in all pairs of indices, by cyclic interchange and addition.
If, in (78), we replace Aστ by the fundamental tensor, gστ ,
then the right-hand side vanishes identically; an analogous statement holds for (80) with respect to g
στ ; that is, the co-variant
derivatives of the fundamental tensor vanish. That this must be
so we see directly in the local system of co-ordinates.
In case Aστ is skew-symmetrical, we obtain from (80), by
contraction with respect to τ and ρ,
A
σ =
∂ A
στ
∂xτ
. (82)
In the general case, from (79) and (80), by contraction with
respect to τ and ρ, we obtain the equations,
Aσ =
∂ A
α
σ
∂xα
− Γ
α
σβ A
β
α
, (83)
A
σ =
∂ A
σα
∂xα
+ Γσ
αβ A
αβ
. (84)
The Riemann Tensor. If we have given a curve extending
from the point P to the point G of the continuum, then a vector Aµ
, given at P, may, by a parallel displacement, be moved
along the curve to G. If the continuum is Euclidean (more generally, if by a suitable choice of co-ordinates the gµν are constants)
then the vector obtained at G as a result of this displacement
does not depend upon the choice of the curve joining P and G.
But otherwise, the result depends upon the path of the displacement. In this case, therefore, a vector suffers a change, ∆Aµ
(in its direction, not its magnitude), when it is carried from a
THE MEANING OF RELATIVITY 80
P G
Fig. 4.
point P of a closed curve, along the curve, and back to P. We
shall now calculate this vector change:
∆A
µ =
I
δAµ
.
As in Stokes’ theorem for the line integral of a vector around
a closed curve, this problem may be reduced to the integration
around a closed curve with infinitely small linear dimensions; we
shall limit ourselves to this case.
We have, first, by (67),
∆A
µ = −
I
Γ
µ
αβA
α
dxβ.
In this, Γµ
αβ is the value of this quantity at the variable
point G of the path of integration. If we put
ξ
µ = (xµ)G − (xµ)P
THE GENERAL THEORY 81
and denote the value of Γµ
αβ at P by Γ
µ
αβ, then we have, with
sufficient accuracy,
Γ
µ
αβ = Γ
µ
αβ +
∂Γ
µ
αβ
∂xν
ξ
ν
.
Let, further, Aα be the value obtained from Aα by a parallel
displacement along the curve from P to G. It may now easily
be proved by means of (67) that Aµ − Aµ
is infinitely small of
the first order, while, for a curve of infinitely small dimensions
of the first order, ∆Aµ
is infinitely small of the second order.
Therefore there is an error of only the second order if we put
A
α = Aα − Γ
α
στ Aσ
ξ
τ
.
If we introduce these values of Γµ
αβ and Aα
into the integral,
we obtain, neglecting all quantities of a higher order of small
quantities than the second,
∆A
µ = −
∂Γ
µ
σβ
∂xα
− Γ
µ
ρβΓ
ρ
σα
A
σ
I
ξ
α
dξβ
. (85)
The quantity removed from under the sign of integration refers
to the point P. Subtracting 1
2
d(ξ
α
ξ
β
) from the integrand, we
obtain
1
2
I
(ξ
α
dξβ − ξ
β
dξα
).
This skew-symmetrical tensor of the second rank, f
αβ, characterizes the surface element bounded by the curve in magnitude
and position. If the expression in the brackets in (85) were
skew-symmetrical with respect to the indices α and β, we could
THE MEANING OF RELATIVITY 82
conclude its tensor character from (85). We can accomplish this
by interchanging the summation indices α and β in (85) and
adding the resulting equation to (85). We obtain
2∆A
µ = −R
µ
σαβA
σ
f
αβ
, (86)
in which
R
µ
σαβ = −
∂Γ
µ
σα
∂xβ
+
∂Γ
µ
σβ
∂xα
+ Γµ
ραΓ
ρ
σβ − Γ
µ
ρβΓ
ρ
σα. (87)
The tensor character of R
µ
σαβ follows from (86); this is the
Riemann curvature tensor of the fourth rank, whose properties of
symmetry we do not need to go into. Its vanishing is a sufficient
condition (disregarding the reality of the chosen co-ordinates)
that the continuum is Euclidean.
By contraction of the Riemann tensor with respect to the
indices µ, β, we obtain the symmetrical tensor of the second
rank,
Rµν = −
∂Γ
α
µν
∂xα
+ Γα
µβΓ
β
να +
∂Γ
α
µα
∂xν
− Γ
α
µνΓ
β
αβ. (88)
The last two terms vanish if the system of co-ordinates is so
chosen that g = constant. From Rµν we can form the scalar,
R = g
µνRµν. (89)
Straightest Geodetic Lines. A line may be constructed in
such a way that its successive elements arise from each other by
parallel displacements. This is the natural generalization of the
straight line of the Euclidean geometry. For such a line, we have
δ
dxµ
ds
= −Γ
µ
αβ
dxα
ds dxβ.
THE GENERAL THEORY 83
The left-hand side is to be replaced by d
2xµ
ds2
,
∗
so that we have
d
2xµ
ds2
+ Γµ
αβ
dxα
ds
dxβ
ds = 0. (90)
We get the same line if we find the line which gives a stationary
value to the integral
Z
ds or Z p
gµν dxµ dxν
between two points (geodetic line).
∗The direction vector at a neighbouring point of the curve results, by
a parallel displacement along the line element (dxβ), from the direction
vector of each point considered.
LECTURE IV
THE GENERAL THEORY OF RELATIVITY
(continued)
We are now in possession of the mathematical apparatus which
is necessary to formulate the laws of the general theory of relativity. No attempt will be made in this presentation at systematic completeness, but single results and possibilities will
be developed progressively from what is known and from the results obtained. Such a presentation is most suited to the present
provisional state of our knowledge.
A material particle upon which no force acts moves, according to the principle of inertia, uniformly in a straight line. In
the four-dimensional continuum of the special theory of relativity (with real time co-ordinate) this is a real straight line. The
natural, that is, the simplest, generalization of the straight line
which is plausible in the system of concepts of Riemann’s general theory of invariants is that of the straightest, or geodetic,
line. We shall accordingly have to assume, in the sense of the
principle of equivalence, that the motion of a material particle,
under the action only of inertia and gravitation, is described by
the equation,
d
2xµ
ds2
+ Γµ
αβ
dxα
ds
dxβ
ds = 0. (90)
In fact, this equation reduces to that of a straight line if all the
components, Γµ
αβ, of the gravitational field vanish.
How are these equations connected with Newton’s equations
of motion? According to the special theory of relativity, the gµν
as well as the g
µν, have the values, with respect to an inertial
84
THE GENERAL THEORY 85
system (with real time co-ordinate and suitable choice of the
sign of ds2
),
−1 0 0 0
0 −1 0 0
0 0 −1 0
0 0 0 1
. (91)
The equations of motion then become
d
2xµ
ds2
= 0.
We shall call this the “first approximation” to the gµν-field. In
considering approximations it is often useful, as in the special
theory of relativity, to use an imaginary x4-co-ordinate, as then
the gµν, to the first approximation, assume the values
−1 0 0 0
0 −1 0 0
0 0 −1 0
0 0 0 −1
. (91a)
These values may be collected in the relation
gµν = −δµν.
To the second approximation we must then put
gµν = −δµν + γµν, (92)
where the γµν are to be regarded as small of the first order.
THE MEANING OF RELATIVITY 86
Both terms of our equation of motion are then small of the
first order. If we neglect terms which, relatively to these, are
small of the first order, we have to put
ds2 = −dxν
2 = dl2
(1 − q
2
), (93)
Γ
µ
αβ = −δµσ
αβ
σ
= −
αβ
µ
=
1
2
∂γαβ
∂xµ
−
∂γαµ
∂xβ
−
∂γβµ
∂xα
. (94)
We shall now introduce an approximation of a second kind. Let
the velocity of the material particles be very small compared to
that of light. Then ds will be the same as the time differential, dl. Further, dx1
ds ,
dx2
ds ,
dx3
ds will vanish compared to dx4
ds .
We shall assume, in addition, that the gravitational field varies
so little with the time that the derivatives of the γµν by x4 may
be neglected. Then the equation of motion (for µ = 1, 2, 3)
reduces to
d
2xµ
dl2
=
∂
∂xµ
γ44
2
. (90a)
This equation is identical with Newton’s equation of motion for
a material particle in a gravitational field, if we identify γ44
2
with the potential of the gravitational field; whether or not this
is allowable, naturally depends upon the field equations of gravitation, that is, it depends upon whether or not this quantity
satisfies, to a first approximation, the same laws of the field
as the gravitational potential in Newton’s theory. A glance at
(90) and (90a) shows that the Γµ
αβ actually do play the rˆole of
the intensity of the gravitational field. These quantities do not
have a tensor character.
Equations (90) express the influence of inertia and gravitation upon the material particle. The unity of inertia and gravi-
THE GENERAL THEORY 87
tation is formally expressed by the fact that the whole left-hand
side of (90) has the character of a tensor (with respect to any
transformation of co-ordinates), but the two terms taken separately do not have tensor character, so that, in analogy with
Newton’s equations, the first term would be regarded as the expression for inertia, and the second as the expression for the
gravitational force.
We must next attempt to find the laws of the gravitational
field. For this purpose, Poisson’s equation,
∆φ = 4πKρ
of the Newtonian theory must serve as a model. This equation has its foundation in the idea that the gravitational field
arises from the density ρ of ponderable matter. It must also
be so in the general theory of relativity. But our investigations
of the special theory of relativity have shown that in place of
the scalar density of matter we have the tensor of energy per
unit volume. In the latter is included not only the tensor of
the energy of ponderable matter, but also that of the electromagnetic energy. We have seen, indeed, that in a more complete
analysis the energy tensor can be regarded only as a provisional
means of representing matter. In reality, matter consists of electrically charged particles, and is to be regarded itself as a part,
in fact, the principal part, of the electromagnetic field. It is
only the circumstance that we have not sufficient knowledge of
the electromagnetic field of concentrated charges that compels
us, provisionally, to leave undetermined in presenting the theory, the true form of this tensor. From this point of view our
problem now is to introduce a tensor, Tµν, of the second rank,
THE MEANING OF RELATIVITY 88
whose structure we do not know provisionally, and which includes in itself the energy density of the electromagnetic field
and of ponderable matter; we shall denote this in the following
as the “energy tensor of matter.”
According to our previous results, the principles of momentum and energy are expressed by the statement that the divergence of this tensor vanishes (47c). In the general theory of relativity, we shall have to assume as valid the corresponding general
co-variant equation. If (Tµν) denotes the co-variant energy tensor of matter, T
ν
σ
the corresponding mixed tensor density, then,
in accordance with (83), we must require that
0 =
∂ T
α
σ
∂xα
− Γ
α
σβ T
β
α
(95)
be satisfied. It must be remembered that besides the energy density of the matter there must also be given an energy density of
the gravitational field, so that there can be no talk of principles
of conservation of energy and momentum for matter alone. This
is expressed mathematically by the presence of the second term
in (95), which makes it impossible to conclude the existence of
an integral equation of the form of (49). The gravitational field
transfers energy and momentum to the “matter,” in that it exerts forces upon it and gives it energy; this is expressed by the
second term in (95).
If there is an analogue of Poisson’s equation in the general
theory of relativity, then this equation must be a tensor equation for the tensor gµν of the gravitational potential; the energy tensor of matter must appear on the right-hand side of this
equation. On the left-hand side of the equation there must be
a differential tensor in the gµν. We have to find this differen-
THE GENERAL THEORY 89
tial tensor. It is completely determined by the following three
conditions:—
1. It may contain no differential coefficients of the gµν higher
than the second.
2. It must be linear and homogeneous in these second differential coefficients.
3. Its divergence must vanish identically.
The first two of these conditions are naturally taken from
Poisson’s equation. Since it may be proved mathematically
that all such differential tensors can be formed algebraically
(i.e. without differentiation) from Riemann’s tensor, our tensor
must be of the form
Rµν + agµνR,
in which Rµν and R are defined by (88) and (89) respectively.
Further, it may be proved that the third condition requires a
to have the value −
1
2
. For the law of the gravitational field we
therefore get the equation
Rµν −
1
2
gµνR = −κTµν. (96)
Equation (95) is a consequence of this equation. κ denotes a
constant, which is connected with the Newtonian gravitation
constant.
In the following I shall indicate the features of the theory
which are interesting from the point of view of physics, using as
little as possible of the rather involved mathematical method.
It must first be shown that the divergence of the left-hand side
actually vanishes. The energy principle for matter may be expressed, by (83),
0 =
∂ T
α
σ
∂xα
− Γ
α
σβ T
β
α
, (97)
THE MEANING OF RELATIVITY 90
in which
T
α
σ = Tστ g
τα√
−g.
The analogous operation, applied to the left-hand side of (96),
will lead to an identity.
In the region surrounding each world-point there are systems
of co-ordinates for which, choosing the x4-co-ordinate imaginary,
at the given point,
gµν = g
µν = −δµν =
(
−1 if µ = ν,
0 if µ 6= ν,
and for which the first derivatives of the gµν and the g
µν vanish.
We shall verify the vanishing of the divergence of the left-hand
side at this point. At this point the components Γα
σβ vanish, so
that we have to prove the vanishing only of
∂
∂xσ
√
−g gνσ(Rµν −
1
2
gµνR)
.
Introducing (88) and (70) into this expression, we see that the
only terms that remain are those in which third derivatives of
the g
µν enter. Since the gµν are to be replaced by −δµν, we obtain, finally, only a few terms which may easily be seen to cancel
each other. Since the quantity that we have formed has a tensor
character, its vanishing is proved for every other system of coordinates also, and naturally for every other four-dimensional
point. The energy principle of matter (97) is thus a mathematical consequence of the field equations (96).
In order to learn whether the equations (96) are consistent
with experience, we must, above all else, find out whether they
THE GENERAL THEORY 91
lead to the Newtonian theory as a first approximation. For this
purpose we must introduce various approximations into these
equations. We already know that Euclidean geometry and the
law of the constancy of the velocity of light are valid, to a certain
approximation, in regions of a great extent, as in the planetary
system. If, as in the special theory of relativity, we take the
fourth co-ordinate imaginary, this means that we must put
gµν = −δµν + γµν, (98)
in which the γµν are so small compared to 1 that we can neglect
the higher powers of the γµν and their derivatives. If we do this,
we learn nothing about the structure of the gravitational field, or
of metrical space of cosmical dimensions, but we do learn about
the influence of neighbouring masses upon physical phenomena.
Before carrying through this approximation we shall transform (96). We multiply (96) by g
µν, summed over the µ and ν;
observing the relation which follows from the definition of
the g
µν
,
gµνg
µν = 4,
we obtain the equation
R = κgµνTµν = κT.
If we put this value of R in (96) we obtain
Rµν = −κ(Tµν −
1
2
gµνT) = −κT∗
µν. (96a)
When the approximation which has been mentioned is carried
out, we obtain for the left-hand side,
−
1
2
∂
2γµν
∂xα
2 +
∂
2γαα
∂xµ ∂xν
−
∂
2γµα
∂xν ∂xα
−
∂
2γνα
∂xµ ∂xα
THE MEANING OF RELATIVITY 92
or
−
1
2
∂
2γµν
∂xα
2 +
1
2
∂
∂xν
∂γ0
µα
∂xα
+
1
2
∂
∂xµ
∂γ0
να
∂xα
,
in which has been put
γ
0
µν = γµν −
1
2
γσσδµν. (99)
We must now note that equation (96) is valid for any system of co-ordinates. We have already specialized the system of
co-ordinates in that we have chosen it so that within the region
considered the gµν differ infinitely little from the constant values −δµν. But this condition remains satisfied in any infinitesimal change of co-ordinates, so that there are still four conditions
to which the γµν may be subjected, provided these conditions
do not conflict with the conditions for the order of magnitude of
the γµν. We shall now assume that the system of co-ordinates
is so chosen that the four relations—
0 =
∂γ0
µν
∂xν
=
∂γµν
∂xν
−
1
2
∂γσσ
∂xµ
(100)
are satisfied. Then (96a) takes the form
∂
2γµν
∂xα
2 = 2κT∗
µν. (96b)
These equations may be solved by the method, familiar in
electrodynamics, of retarded potentials; we get, in an easily
understood notation,
γµν = −
κ
2π
Z
T
∗
µν(x0, y0, z0, t − r)
r
dV0. (101)
THE GENERAL THEORY 93
In order to see in what sense this theory contains the Newtonian theory, we must consider in greater detail the energy
tensor of matter. Considered phenomenologically, this energy
tensor is composed of that of the electromagnetic field and of
matter in the narrower sense. If we consider the different parts
of this energy tensor with respect to their order of magnitude,
it follows from the results of the special theory of relativity that
the contribution of the electromagnetic field practically vanishes
in comparison to that of ponderable matter. In our system of
units, the energy of one gram of matter is equal to 1, compared
to which the energy of the electric fields may be ignored, and
also the energy of deformation of matter, and even the chemical
energy. We get an approximation that is fully sufficient for our
purpose if we put
T
µν = σ
dxµ
ds
dxν
ds ,
ds2 = gµν dxµ dxν.
(102)
In this, σ is the density at rest, that is, the density of the ponderable matter, in the ordinary sense, measured with the aid
of a unit measuring rod, and referred to a Galilean system of
co-ordinates moving with the matter.
We observe, further, that in the co-ordinates we have chosen,
we shall make only a relatively small error if we replace the gµν
by −δµν, so that we put
ds2 = −
Xdxµ
2
. (102a)
The previous developments are valid however rapidly the
masses which generate the field may move relatively to our chosen system of quasi-Galilean co-ordinates. But in astronomy
THE MEANING OF RELATIVITY 94
we have to do with masses whose velocities, relatively to the
co-ordinate system employed, are always small compared to the
velocity of light, that is, small compared to 1, with our choice
of the unit of time. We therefore get an approximation which is
sufficient for nearly all practical purposes if in (101) we replace
the retarded potential by the ordinary (non-retarded) potential,
and if, for the masses which generate the field, we put
dx1
ds =
dx2
ds =
dx3
ds = 0,
dx4
ds =
√
−1 dl
dl =
√
−1. (103)
Then we get for T
µν and Tµν the values
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 −σ
. (104)
For T we get the value σ, and, finally, for T
∗
µν the values,
σ
2
0 0 0
0
σ
2
0 0
0 0
σ
2
0
0 0 0 −
σ
2
. (104a)
We thus get, from (101),
γ11 = γ22 = γ33 = −
κ
4π
Z
σ dV0
r
,
γ44 = +
κ
4π
Z
σ dV0
r
,
(101a)
THE GENERAL THEORY 95
while all the other γµν vanish. The least of these equations,
in connexion with equation (90a), contains Newton’s theory of
gravitation. If we replace l by ct we get
d
2xµ
dt2
=
κc2
8π
∂
∂xµ
Z
σ dV0
r
. (90b)
We see that the Newtonian gravitation constant K, is connected
with the constant κ that enters into our field equations by the
relation
K =
κc2
8π
. (105)
From the known numerical value of K, it therefore follows that
κ =
8πK
c
2
=
8π · 6.67 · 10−8
9 · 1020 = 1.86 · 10−27
. (105a)
From (101) we see that even in the first approximation the structure of the gravitational field differs fundamentally from that
which is consistent with the Newtonian theory; this difference
lies in the fact that the gravitational potential has the character
of a tensor and not a scalar. This was not recognized in the past
because only the component g44, to a first approximation, enters
the equations of motion of material particles.
In order now to be able to judge the behaviour of measuring
rods and clocks from our results, we must observe the following.
According to the principle of equivalence, the metrical relations
of the Euclidean geometry are valid relatively to a Cartesian
system of reference of infinitely small dimensions, and in a suitable state of motion (freely falling, and without rotation). We
can make the same statement for local systems of co-ordinates
THE MEANING OF RELATIVITY 96
which, relatively to these, have small accelerations, and therefore for such systems of co-ordinates as are at rest relatively to
the one we have selected. For such a local system, we have, for
two neighbouring point events,
ds2 = −dX1
2 − dX2
2 − dX3
2 + dT2 = −dS2 + dT2
,
where dS is measured directly by a measuring rod and dT by
a clock at rest relatively to the system; these are the naturally
measured lengths and times. Since ds2
, on the other hand, is
known in terms of the co-ordinates xν employed in finite regions,
in the form
ds2 = gµν dxµ dxν,
we have the possibility of getting the relation between naturally
measured lengths and times, on the one hand, and the corresponding differences of co-ordinates, on the other hand. As the
division into space and time is in agreement with respect to the
two systems of co-ordinates, so when we equate the two expressions for ds2 we get two relations. If, by (101a), we put
ds2 = −
1 +
κ
4π
Z
σ dV0
r
(dx1
2 + dx2
2 + dx3
2
)
+
1 −
κ
4π
Z
σ dV0
r
dl2
,
we obtain, to a sufficiently close approximation,
p
dX1
2 + dX2
2 + dX3
2
=
1 +
κ
8π
Z
σ dV0
r
p
dx1
2 + dx2
2 + dx3
2
,
dT =
1 −
κ
8π
Z
σ dV0
r
dl.
(106)
THE GENERAL THEORY 97
The unit measuring rod has therefore the length,
1 −
κ
8π
Z
σ dV0
r
in respect to the system of co-ordinates we have selected. The
particular system of co-ordinates we have selected insures that
this length shall depend only upon the place, and not upon the
direction. If we had chosen a different system of co-ordinates
this would not be so. But however we may choose a system of
co-ordinates, the laws of configuration of rigid rods do not agree
with those of Euclidean geometry; in other words, we cannot
choose any system of co-ordinates so that the co-ordinate differences, ∆x1, ∆x2, ∆x3, corresponding to the ends of a unit measuring rod, oriented in any way, shall always satisfy the relation
∆x1
2 + ∆x2
2 + ∆x3
2 = 1. In this sense space is not Euclidean,
but “curved.” It follows from the second of the relations above
that the interval between two beats of the unit clock (dT = 1)
corresponds to the “time”
1 +
κ
8π
Z
σ dV0
r
in the unit used in our system of co-ordinates. The rate of a
clock is accordingly slower the greater is the mass of the ponderable matter in its neighbourhood. We therefore conclude that
spectral lines which are produced on the sun’s surface will be
displaced towards the red, compared to the corresponding lines
produced on the earth, by about 2 · 10−6 of their wave-lengths.
At first, this important consequence of the theory appeared to
conflict with experiment; but results obtained during the past
year seem to make the existence of this effect more probable, and
THE MEANING OF RELATIVITY 98
it can hardly be doubted that this consequence of the theory will
be confirmed within the next year.
Another important consequence of the theory, which can be
tested experimentally, has to do with the path of rays of light.
In the general theory of relativity also the velocity of light is
everywhere the same, relatively to a local inertial system. This
velocity is unity in our natural measure of time. The law of
the propagation of light in general co-ordinates is therefore, according to the general theory of relativity, characterized, by the
equation
ds2 = 0.
To within the approximation which we are using, and in the
system of co-ordinates which we have selected, the velocity of
light is characterized, according to (106), by the equation
1+
κ
4π
Z
σ dV0
r
(dx1
2+dx2
2+dx3
2
) =
1−
κ
4π
Z
σ dV0
r
dl2
.
The velocity of light L, is therefore expressed in our co-ordinates
by
p
dx1
2 + dx2
2 + dx3
2
dl = 1 −
κ
4π
Z
σ dV0
r
. (107)
We can therefore draw the conclusion from this, that a ray of
light passing near a large mass is deflected. If we imagine the
sun, of mass M concentrated at the origin of our system of coordinates, then a ray of light, travelling parallel to the x3-axis,
in the x1-x3 plane, at a distance ∆ from the origin, will be
deflected, in all, by an amount
α =
Z +∞
−∞
1
L
∂L
∂x1
dx3
THE GENERAL THEORY 99
towards the sun. On performing the integration we get
α =
κM
2π∆
. (108)
The existence of this deflection, which amounts to 1.7
00 for
∆ equal to the radius of the sun, was confirmed, with remarkable
accuracy, by the English Solar Eclipse Expedition in 1919, and
most careful preparations have been made to get more exact
observational data at the solar eclipse in 1922. It should be
noted that this result, also, of the theory is not influenced by
our arbitrary choice of a system of co-ordinates.
This is the place to speak of the third consequence of the
theory which can be tested by observation, namely, that which
concerns the motion of the perihelion of the planet Mercury. The
secular changes in the planetary orbits are known with such accuracy that the approximation we have been using is no longer
sufficient for a comparison of theory and observation. It is necessary to go back to the general field equations (96). To solve
this problem I made use of the method of successive approximations. Since then, however, the problem of the central symmetrical statical gravitational field has been completely solved by
Schwarzschild and others; the derivation given by H. Weyl in his
book, “Raum-Zeit-Materie,” is particularly elegant. The calculation can be simplified somewhat if we do not go back directly
to the equation (96), but base it upon a principle of variation
that is equivalent to this equation. I shall indicate the procedure
only in so far as is necessary for understanding the method.
THE MEANING OF RELATIVITY 100
In the case of a statical field, ds2 must have the form
ds2 = −dσ2 + f
2
dx4
2
,
dσ2 =
X
1–3
γαβ dxα dxβ,
(109)
where the summation on the right-hand side of the last equation
is to be extended over the space variables only. The central
symmetry of the field requires the γµν to be of the form,
γαβ = µδαβ + λxαxβ; (110)
f
2
, µ and λ are functions of r =
√
x1
2 + x2
2 + x3
2 only. One
of these three functions can be chosen arbitrarily, because our
system of co-ordinates is, a priori, completely arbitrary; for by
a substitution
x
0
4 = x4,
x
0
α = F(r)xα,
we can always insure that one of these three functions shall be
an assigned function of r
0
. In place of (110) we can therefore
put, without limiting the generality,
γαβ = δαβ + λxαxβ. (110a)
In this way the gµν are expressed in terms of the two quantities λ and f. These are to be determined as functions of r,
by introducing them into equation (96), after first calculating
THE GENERAL THEORY 101
the Γσ
µν from (109) and (110a). We have
Γ
σ
αβ =
1
2
xσ
r
·
λ
0xαxβ + 2λr δαβ
1 + λr2
(for α, β, σ = 1, 2, 3),
Γ
4
44 = Γα
4β = Γ4
αβ = 0 (for α, β = 1, 2, 3),
Γ
4
4α =
1
2
f
−2
∂f 2
∂xα
, Γ
α
44 = −
1
2
g
αβ ∂f 2
∂xβ
.
(110b)
With the help of these results, the field equations furnish
Schwarzschild’s solution:
ds2 =
1 −
A
r
dl2 −
dr2
1 −
A
r
+ r
2
(sin2
θ dφ2 + dθ2
)
, (109a)
in which we have put
x4 = l,
x1 = r sin θ sin φ,
x2 = r sin θ cos φ,
x3 = r cos θ,
A =
κM
4π
.
(109b)
M denotes the sun’s mass, centrally symmetrically placed
about the origin of co-ordinates; the solution (109) is valid only
outside of this mass, where all the Tµν vanish. If the motion
of the planet takes place in the x1-x2 plane then we must replace (109a) by
ds2 =
1 −
A
r
dl2 −
dr2
1 −
A
r
− r
2
dφ2
. (109c)
THE MEANING OF RELATIVITY 102
The calculation of the planetary motion depends upon equation (90). From the first of equations (110b) and (90) we get,
for the indices 1, 2, 3,
d
ds
xα
dxβ
ds − xβ
dxα
ds
= 0,
or, if we integrate, and express the result in polar co-ordinates,
r
2
dφ
ds = constant. (111)
From (90), for µ = 4, we get
0 =
d
2
l
ds2
+
1
f
2
df 2
dxα
dxα
ds
dl
ds =
d
2
l
ds2
+
1
f
2
df 2
ds
dl
ds.
From this, after multiplication by f
2 and integration, we have
f
2
dl
ds = constant. (112)
In (109c), (111) and (112) we have three equations between
the four variables s, r, l and φ, from which the motion of the
planet may be calculated in the same way as in classical mechanics. The most important result we get from this is a secular
rotation of the elliptic orbit of the planet in the same sense as
the revolution of the planet, amounting in radians per revolution
to
24π
3a
2
(1 − e
2
)c
2T
2
, (113)
THE GENERAL THEORY 103
where
a = the semi-major axis of the planetary orbit in centimetres.
e = the numerical eccentricity.
c = 3 · 10+10
, the velocity of light in vacuo.
T = the period of revolution in seconds.
This expression furnishes the explanation of the motion of the
perihelion of the planet Mercury, which has been known for a
hundred years (since Leverrier), and for which theoretical astronomy has hitherto been unable satisfactorily to account.
There is no difficulty in expressing Maxwell’s theory of the
electromagnetic field in terms of the general theory of relativity;
this is done by application of the tensor formation (81), (82)
and (77). Let φµ be a tensor of the first rank, to be denoted
as an electromagnetic 4-potential; then an electromagnetic field
tensor may be defined by the relations,
φµν =
∂φµ
∂xν
−
∂φν
∂xµ
. (114)
The second of Maxwell’s systems of equations is then defined by
the tensor equation, resulting from this,
∂φµν
∂xρ
+
∂φνρ
∂xµ
+
∂φρµ
∂xν
= 0, (114a)
and the first of Maxwell’s systems of equations is defined by the
tensor-density relation
∂ F
µν
∂xν
= J
µ
, (115)
THE MEANING OF RELATIVITY 104
in which
F
µν =
√
−g gµνg
ντφστ ,
J
µ =
√
−g ρ
dxν
ds .
If we introduce the energy tensor of the electromagnetic field
into the right-hand side of (96), we obtain (115), for the special
case J
µ = 0, as a consequence of (96) by taking the divergence.
This inclusion of the theory of electricity in the scheme of the
general theory of relativity has been considered arbitrary and
unsatisfactory by many theoreticians. Nor can we in this way
conceive of the equilibrium of the electricity which constitutes
the elementary electrically charged particles. A theory in which
the gravitational field and the electromagnetic field enter as an
essential entity would be much preferable. H. Weyl, and recently
Th. Kaluza, have discovered some ingenious theorems along this
direction; but concerning them, I am convinced that they do not
bring us nearer to the true solution of the fundamental problem.
I shall not go into this further, but shall give a brief discussion
of the so-called cosmological problem, for without this, the considerations regarding the general theory of relativity would, in
a certain sense, remain unsatisfactory.
Our previous considerations, based upon the field equations (96), had for a foundation the conception that space on
the whole is Galilean-Euclidean, and that this character is disturbed only by masses embedded in it. This conception was
certainly justified as long as we were dealing with spaces of the
order of magnitude of those that astronomy has to do with.
But whether portions of the universe, however large they may
be, are quasi-Euclidean, is a wholly different question. We can
THE GENERAL THEORY 105
make this clear by using an example from the theory of surfaces
which we have employed many times. If a portion of a surface
is observed by the eye to be practically plane, it does not at all
follow that the whole surface has the form of a plane; the surface
might just as well be a sphere, for example, of sufficiently large
radius. The question as to whether the universe as a whole is
non-Euclidean was much discussed from the geometrical point of
view before the development of the theory of relativity. But with
the theory of relativity, this problem has entered upon a new
stage, for according to this theory the geometrical properties of
bodies are not independent, but depend upon the distribution
of masses.
If the universe were quasi-Euclidean, then Mach was wholly
wrong in his thought that inertia, as well as gravitation, depends
upon a kind of mutual action between bodies. For in this case,
with a suitably selected system of co-ordinates, the gµν would
be constant at infinity, as they are in the special theory of relativity, while within finite regions the gµν would differ from these
constant values by small amounts only, with a suitable choice
of co-ordinates, as a result of the influence of the masses in finite regions. The physical properties of space would not then be
wholly independent, that is, uninfluenced by matter, but in the
main they would be, and only in small measure, conditioned by
matter. Such a dualistic conception is even in itself not satisfactory; there are, however, some important physical arguments
against it, which we shall consider.
The hypothesis that the universe is infinite and Euclidean
at infinity, is, from the relativistic point of view, a complicated
hypothesis. In the language of the general theory of relativity
it demands that the Riemann tensor of the fourth rank Riklm
THE MEANING OF RELATIVITY 106
shall vanish at infinity, which furnishes twenty independent conditions, while only ten curvature components Rµν, enter into
the laws of the gravitational field. It is certainly unsatisfactory
to postulate such a far-reaching limitation without any physical
basis for it.
But in the second place, the theory of relativity makes it
appear probable that Mach was on the right road in his thought
that inertia depends upon a mutual action of matter. For we
shall show in the following that, according to our equations, inert
masses do act upon each other in the sense of the relativity of
inertia, even if only very feebly. What is to be expected along
the line of Mach’s thought?
1. The inertia of a body must increase when ponderable
masses are piled up in its neighbourhood.
2. A body must experience an accelerating force when
neighbouring masses are accelerated, and, in fact, the
force must be in the same direction as the acceleration.
3. A rotating hollow body must generate inside of itself
a “Coriolis field,” which deflects moving bodies in the
sense of the rotation, and a radial centrifugal field as
well.
We shall now show that these three effects, which are to be
expected in accordance with Mach’s ideas, are actually present
according to our theory, although their magnitude is so small
that confirmation of them by laboratory experiments is not to be
thought of. For this purpose we shall go back to the equations of
motion of a material particle (90), and carry the approximations
somewhat further than was done in equation (90a).
THE GENERAL THEORY 107
First, we consider γ44 as small of the first order. The square
of the velocity of masses moving under the influence of the gravitational force is of the same order, according to the energy
equation. It is therefore logical to regard the velocities of the
material particles we are considering, as well as the velocities
of the masses which generate the field, as small, of the order 1
2
.
We shall now carry out the approximation in the equations that
arise from the field equations (101) and the equations of motion (90) so far as to consider terms, in the second member
of (90), that are linear in those velocities. Further, we shall not
put ds and dl equal to each other, but, corresponding to the
higher approximation, we shall put
ds =
√
g44 dl =
1 −
γ44
2
dl.
From (90) we obtain, at first,
d
dl
1 +
γ44
2
dxµ
dl
= −Γ
µ
αβ
dxα
dl
dxβ
dl
1 +
γ44
2
. (116)
From (101) we get, to the approximation sought for,
−γ11 = −γ22 = −γ33 = γ44 =
κ
4π
Z
σ dV0
r
,
γ4α = −
iκ
2
Z σ
dxα
ds dV0
r
,
γαβ = 0,
(117)
in which, in (117), α and β denote the space indices only.
THE MEANING OF RELATIVITY 108
On the right-hand side of (116) we can replace 1 + γ44
2
by 1
and −Γ
µ
αβ by
αβ
µ
. It is easy to see, in addition, that to this
degree of approximation we must put
44
µ
= −
1
2
∂γ44
∂xµ
+
∂γ4µ
∂x4
,
α4
µ
=
1
2
∂γ4µ
∂xα
−
∂γ4α
∂xµ
,
αβ
µ
= 0,
in which α, β and µ denote space indices. We therefore obtain
from (116), in the usual vector notation,
d
dl
(1 + σ)v
= grad σ +
∂ A
∂l + [rot A, v],
σ =
κ
8π
Z
σ dV0
r
,
A =
κ
2π
Z σ
dxα
dl dV0
r
.
(118)
The equations of motion, (118), show now, in fact, that
1. The inert mass is proportional to 1 + σ, and therefore
increases when ponderable masses approach the test
body.
2. There is an inductive action of accelerated masses,
of the same sign, upon the test body. This is the
term
∂ A
∂l .
THE GENERAL THEORY 109
3. A material particle, moving perpendicularly to the axis
of rotation inside a rotating hollow body, is deflected in
the sense of the rotation (Coriolis field). The centrifugal action, mentioned above, inside a rotating hollow
body, also follows from the theory, as has been shown
by Thirring.∗
Although all of these effects are inaccessible to experiment,
because κ is so small, nevertheless they certainly exist according to the general theory of relativity. We must see in them a
strong support for Mach’s ideas as to the relativity of all inertial
actions. If we think these ideas consistently through to the end
we must expect the whole inertia, that is, the whole gµν-field, to
be determined by the matter of the universe, and not mainly by
the boundary conditions at infinity.
For a satisfactory conception of the gµν-field of cosmical dimensions, the fact seems to be of significance that the relative
velocity of the stars is small compared to the velocity of light.
It follows from this that, with a suitable choice of co-ordinates,
g44 is nearly constant in the universe, at least, in that part of
the universe in which there is matter. The assumption appears
natural, moreover, that there are stars in all parts of the universe, so that we may well assume that the inconstancy of g44 depends only upon the circumstance that matter is not distributed
continuously, but is concentrated in single celestial bodies and
systems of bodies. If we are willing to ignore these more local
∗That the centrifugal action must be inseparably connected with the
existence of the Coriolis field may be recognized, even without calculation,
in the special case of a co-ordinate system rotating uniformly relatively to
an inertial system; our general co-variant equations naturally must apply
to such a case.
THE MEANING OF RELATIVITY 110
non-uniformities of the density of matter and of the gµν-field, in
order to learn something of the geometrical properties of the universe as a whole, it appears natural to substitute for the actual
distribution of masses a continuous distribution, and furthermore to assign to this distribution a uniform density σ. In this
imagined universe all points with space directions will be geometrically equivalent; with respect to its space extension it will
have a constant curvature, and will be cylindrical with respect
to its x4-co-ordinate. The possibility seems to be particularly
satisfying that the universe is spatially bounded and thus, in
accordance with our assumption of the constancy of σ, is of
constant curvature, being either spherical or elliptical; for then
the boundary conditions at infinity which are so inconvenient
from the standpoint of the general theory of relativity, may be
replaced by the much more natural conditions for a closed surface.
According to what has been said, we are to put
ds2 = dx4
2 − γµν dxµ dxν, (119)
in which the indices µ and ν run from 1 to 3 only. The γµν
will be such functions of x1, x2, x3 as correspond to a threedimensional continuum of constant positive curvature. We must
now investigate whether such an assumption can satisfy the field
equations of gravitation.
In order to be able to investigate this, we must first find
what differential conditions the three-dimensional manifold of
constant curvature satisfies. A spherical manifold of three dimensions, embedded in a Euclidean continuum of four dimen-
THE GENERAL THEORY 111
sions,∗
is given by the equations
x1
2 + x2
2 + x3
2 + x4
2 = a
2
,
dx1
2 + dx2
2 + dx3
2 + dx4
2 = ds2
.
By eliminating x4, we get
ds2 = dx1
2 + dx2
2 + dx3
2 +
(x1 dx1 + x2 dx2 + x3 dx3)
2
a
2 − x1
2 − x2
2 − x3
2
.
As far as terms of the third and higher degrees in the xν, we
can put, in the neighbourhood of the origin of co-ordinates,
ds2 =
δµν +
xµxν
a
2
dxµ dxν.
Inside the brackets are the gµν of the manifold in the neighbourhood of the origin. Since the first derivatives of the gµν,
and therefore also the Γσ
µν, vanish at the origin, the calculation
of the Rµν for this manifold, by (88), is very simple at the origin.
We have
Rµν = −
2
a
2
δµν =
2
a
2
gµν.
Since the relation Rµν =
2
a
2
gµν is universally co-variant,
and since all points of the manifold are geometrically equivalent, this relation holds for every system of co-ordinates, and
everywhere in the manifold. In order to avoid confusion with
∗The aid of a fourth space dimension has naturally no significance except that of a mathematical artifice.
THE MEANING OF RELATIVITY 112
the four-dimensional continuum, we shall, in the following, designate quantities that refer to the three-dimensional continuum
by Greek letters, and put
Pµν = −
2
a
2
γµν. (120)
We now proceed to apply the field equations (96) to our special case. From (119) we get for the four-dimensional manifold,
Rµν = Pµν for the indices 1 to 3,
R14 = R24 = R34 = R44 = 0.
)
(121)
For the right-hand side of (96) we have to consider the energy
tensor for matter distributed like a cloud of dust. According to
what has gone before we must therefore put
T
µν = σ
dxµ
ds
dxν
ds
specialized for the case of rest. But in addition, we shall add
a pressure term that may be physically established as follows.
Matter consists of electrically charged particles. On the basis
of Maxwell’s theory these cannot be conceived of as electromagnetic fields free from singularities. In order to be consistent
with the facts, it is necessary to introduce energy terms, not
contained in Maxwell’s theory, so that the single electric particles may hold together in spite of the mutual repulsions between
their elements, charged with electricity of one sign. For the sake
of consistency with this fact, Poincar´e has assumed a pressure
to exist inside these particles which balances the electrostatic
repulsion. It cannot, however, be asserted that this pressure
THE GENERAL THEORY 113
vanishes outside the particles. We shall be consistent with this
circumstance if, in our phenomenological presentation, we add
a pressure term. This must not, however, be confused with a
hydrodynamical pressure, as it serves only for the energetic presentation of the dynamical relations inside matter. In this sense
we put
Tµν = gµσgνβ
dxα
ds
dxβ
ds − gµνp. (122)
In our special case we have, therefore, to put
Tµν = γµνp (for µ and ν from 1 to 3),
T44 = σ − p,
T = −γ
µνγµνp + σ − p = σ − 4p.
Observing that the field equation (96) may be written in the
form
Rµν = −κ(Tµν −
1
2
gµνT),
we get from (96) the equations,
+
2
a
2
γµν = κ
σ
2
− p
γµν,
0 = −κ
σ
2
+ p
.
From this follows
p = −
σ
2
,
a =
r
2
κσ
.
(123)
If the universe is quasi-Euclidean, and its radius of curvature
therefore infinite, then σ would vanish. But it is improbable that
THE MEANING OF RELATIVITY 114
the mean density of matter in the universe is actually zero; this
is our third argument against the assumption that the universe
is quasi-Euclidean. Nor does it seem possible that our hypothetical pressure can vanish; the physical nature of this pressure can
be appreciated only after we have a better theoretical knowledge
of the electromagnetic field. According to the second of equations (123) the radius, a, of the universe is determined in terms
of the total mass, M, of matter, by the equation
a =
Mκ
4π
2
. (124)
The complete dependence of the geometrical upon the physical
properties becomes clearly apparent by means of this equation.
Thus we may present the following arguments against the
conception of a space-infinite, and for the conception of a spacebounded, universe:—
1. From the standpoint of the theory of relativity, the condition for a closed surface is very much simpler than the corresponding boundary condition at infinity of the quasi-Euclidean
structure of the universe.
2. The idea that Mach expressed, that inertia depends upon
the mutual action of bodies, is contained, to a first approximation, in the equations of the theory of relativity; it follows
from these equations that inertia depends, at least in part, upon
mutual actions between masses. As it is an unsatisfactory assumption to make that inertia depends in part upon mutual
actions, and in part upon an independent property of space,
Mach’s idea gains in probability. But this idea of Mach’s corresponds only to a finite universe, bounded in space, and not to a
quasi-Euclidean, infinite universe. From the standpoint of epis-
THE GENERAL THEORY 115
temology it is more satisfying to have the mechanical properties
of space completely determined by matter, and this is the case
only in a space-bounded universe.
3. An infinite universe is possible only if the mean density of
matter in the universe vanishes. Although such an assumption
is logically possible, it is less probable than the assumption that
there is a finite mean density of matter in the universe.
INDEX
A
Accelerated masses, inductive
action of, 108
Addition and subtraction of
tensors, 14
— theorem of velocities, 38
B
Biot-Savart force, 44
C
Centrifugal force, 64
Clocks, moving, 38
Compressible viscous fluid, 22
Concept of space, 3
— time, 28
Conditions of orthogonality, 7
Congruence, theorems of, 3
Conservation principles, 54
Continuum, four-dimensional,
31
Contraction of tensors, 14
Contra-variant vectors, 69
— tensors, 71
Co-ordinates, preferred systems
of, 8
Co-variance of equation of
continuity, 21
Co-variant, 12 et seq.
— vector, 68
Criticism of principle of inertia,
62
Criticisms of theory of
relativity, 29
Curvilinear co-ordinates, 65
D
Differentiation of tensors, 73, 76
Displacement of spectral lines,
97
E
Energy and mass, 45, 49
— tensor of electromagnetic
field, 50
— — of matter, 54
Equation of continuity,
co-variance of, 21
Equations of motion of material
particle, 50
Equivalence of mass and
energy, 49
Equivalent spaces of reference,
25
Euclidean geometry, 4
F
Finiteness of universe, 105
Fizeau, 28
Four-dimensional continuum, 31
Four-vector, 41
116
INDEX 117
Fundamental tensor, 71
G
Galilean regions, 62
— transformation, 27
Gauss, 65
Geodetic lines, 82
Geometry, Euclidean, 4
Gravitation constant, 95
Gravitational mass, 60
H
Homogeneity of space, 17
Hydrodynamical equations, 54
Hypotheses of pre-relativity
physics, 73
I
Inductive action of accelerated
masses, 108
Inert and gravitational mass,
equality of, 60
Invariant, 9 et seq.
Isotropy of space, 17
K
Kaluza, 104
L
Levi-Civita, 73
Light-cone, 41
Light ray, path of, 98
Light-time, 33
Linear orthogonal
transformation, 7
Lorentz electromotive force, 44
— transformation, 31
M
Mach, 59, 105, 106, 109, 114
Mass and Energy, 45, 49
— equality of gravitational and
inert, 60
— gravitational, 60
Maxwell’s equations, 23
Mercury, perihelion of, 99, 103
Michelson and Morley, 28
Minkowski, 32
Motion of particle, equations of,
50
Moving measuring rods and
clocks, 38
Multiplication of tensors, 14
N
Newtonian gravitation
constant, 95
O
Operations on tensors, 13 et
seq.
Orthogonal transformations,
linear, 7
Orthogonality, conditions of, 7
THE MEANING OF RELATIVITY 118
P
Path of light ray, 98
Perihelion of Mercury, 99, 103
Poisson’s equation, 87
Preferred systems of
co-ordinates, 8
Pre-relativity physics,
hypotheses of, 26
Principle of equivalence, 61
— inertia, criticism of, 62
Principles of conservation, 54
R
Radius of Universe, 113
Rank of tensor, 13
Ray of light, path of, 98
Reference, space of, 3
Riemann, 68
— tensor, 79, 82, 105
Rods (measuring) and clocks in
motion, 38
Rotation, 63
S
Simultaneity, 17, 29
Sitter, 28
Skew-symmetrical tensor, 15
Solar Eclipse expedition (1919),
99
Space, Concept of, 2
— Homogeneity of, 17
— Isotropy of, 17
Spaces of reference, 3
— equivalence of, 25
Special Lorentz transformation,
34
Spectral lines, displacement of,
97
Straightest lines, 82
Stress tensor, 22
Symmetrical tensor, 15
Systems of co-ordinates,
preferred, 8
T
Tensor, 12 et seq., 68 et seq.
— Addition and subtraction of,
14
— Contraction of, 14
— Fundamental, 71
— Multiplication of, 14
— operations, 13 et seq.
— Rank of, 13
— Symmetrical and
Skew-symmetrical, 15
Tensors, formation by
differentiation, 73
Theorem for addition of
velocities, 38
Theorems of congruence, 3
Theory of relativity, criticisms
of, 29
Thirring, 109
Time-concept, 28
INDEX 119
Time-space concept, 31
Transformation, Galilean, 27
— Linear orthogonal, 7
U
Universe, Finiteness of, 105
— Radius of, 113
V
Vector, co-variant, 69
— contra-variant, 69
Velocities, addition theorem of,
38
Viscous compressible fluid, 22
W
Weyl, 73, 99, 104
PRINTED IN GREAT BRITAIN AT THE UNIVERSITY PRESS, ABERDEEN
LICENSING
