PHYSICAL REVIEW VOLUME 68, N U M B E R S 11 A N D 12 DECEMBER 1 AND 15, 1945

A New Approach to Kinematic Cosmology

L. INFELD AND A. SCHILD
Department of Mathematics, University of Toronto, Toronto, Canada
(Received September 16, 1.945)

The kinematical aspect of relativistic cosmology is examined on the basis of three postulated
requirements: The constancy of the velocity of light, spatial isotropy, and homogeneity. Three
distinct types of cosmological models are obtained, characterized by different motions of
nebulae. The metric of any universe is conformal to Minkowski space and Maxwell's equations
are the same for all possible universes. In Part II, it is shown that the cosmological models
are metrically, though not topologically, equivalent to those of H. P. Robertson. Next, special
models are examined and their line elements brought into the conformal-Minkowskian form.
The problem of the displacement of the lines of nebular spectra is discussed; formulas are
obtained and applied to some special cosmological models. Finally, idealized experiments are
described which indicate the physical content of the cosmological coordinates.

I. General Theory

INTRODUCTION metric form and the motion of fundamental

B Y kinematical cosmology is usually under-
stood that part of relativistic cosmology
which deals with the metric form of our universe,
characterized by a four-dimensional space-time
manifold, and with the motion of free particles
and light rays in this universe. In this sense the
present paper deals with kinematical cosmology
and ignores its dynamical aspect, i.e., the connec-
tion between the Riemannian curvature tensor
on the one hand and the energy-momentum
tensor on the other.
The possible mathematical models of the
universe are usually deduced from a few simple
and convincing principles. In such models the
nebulae are represented by freely moving par-
ticles of a fluid and we imagine such a particle at
every point in space. Each of these particles
moves along a geodesic line. Such a particle,
representing a nebula, will be called a funda-
mental particle and an observer moving with it a
fundamental observer.1 The most important con-
cept by which cosmological models will be
described and characterized is that of funda-
mental particles and their motion. Next, light
rays or geodesic null lines must be considered;
they are the messengers of world events between
^fundamental particles. Finally, the universe is
characterized by a Riemannian metric. The
1
These terms seem to be originally due to E. A. Milne,
Relativity, Gravitation, and World-Structure (Clarendon
Press, Oxford, 1935).
250
particles fully determine the kinematical be-
havior of the universe. It might seem that the
motion of fundamental particles can be deduced
immediately from the Riemannian metric as the
particles move along geodesies. But this is not
so. A geodesic line is, at each world-point, deter-
mined by its direction there, whereas the world-
line of a fundamental particle is completely
determined by the choice of a world-point. The
world-line of a fundamental particle is a geo-
desic ; but the converse statement is not neces-
sarily true.
The questions which this paper tries to answer
are: What are the possible metric forms de-
scribing a universe? What are the possible
motions of fundamental particles? The guiding
principle, commonly accepted and leading to a
solution of these problems, is the principle of
, homogeneity, sometimes called the uniformity or
the cosmological principle. It states that every
fundamental observer sees and describes the
world in the same way.
Traditionally, the problem of relativistic
cosmology2 is attacked by choosing a coordinate
system in which all fundamental particles are at
rest. The fact that such a coordinate system
exists is by no means obvious and is closely
2
An excellent account of "Relativistic Cosmology,"
including a bibliography complete up to 1932, is to be
found in H. P . Robertson's article, Rev. Mod. Phys. 5,
62-90 (1933); also R. C. Tolman, Relativity Thermo-
dynamics and Cosmology (Clarendon Press, Oxford, 1934).
 KINEMATIC
related to the principle of homogeneity. T h u s the
line elements discussed in cosmology are usually
of the form
ds2 = dr2-R2(r)d(r2. (0.1)
Here R(T) is an a r b i t r a r y function of time and
da2 is the metric of a three-dimensional space of
constant curvature k=l, — 1 , or 0. T h u s the
universe is characterized by an arbitrary function
R(T) and by the choice of one a m o n g three
possible spaces. T h e problem of the motion of
fundamental particles disappears from such a
presentation because these particles are always
a t rest. I t is the space structure, i.e. the curvature
kR~2(r) of the 3-space r = constant, which
characterizes the cosmological model. I t should
be noted t h a t , in the coordinate system of (0.1),
t h e speed of light in a n y fixed direction is a
function of time and depends on both k and
Obviously, a discussion in which the four-
dimensional universe is characterized chiefly by
the curvature of a three-dimensional space is
contrary to the spirit of relativity theory in
which t h e world is represented b y a four-dimen-
sional space-time continuum. Historically, this
approach goes back to Einstein's first cosmo-
logical paper, 3 to the Einstein universe of which
all others seem to be natural generalizations.
This point of view, based on dynamic considera-
tion^ and on the generalization of the gravita-
tional equations, was long ago abandoned by its
originator.
W e believe t h a t a deeper insight into cos-
mological problems is gained b y a new approach.
Relativistic cosmology, a t least in its kinematical
aspect, should form a link between the restricted
and general theories of relativity. In restricted
relativity the world is represented by a Min-
kowski continuum. This is one of m a n y cos-
mological backgrounds which satisfy the prin-
ciple of homogeneity. We shall see t h a t an
approach to cosmology is possible in which the
structure of a three-dimensional space does not
enter the picture. W e believe t h a t this new
approach p u t s into the foreground the more
essential concepts of kinematical cosmology, i.e.,
3
A. Einstein, "Kosmologische Betrachtungen zur Allge-
meinen Relativitatstheorie," Sitz. Preuss. Akad. Wiss.
142-152 (1917).
COSMOLOGY 251
the type of motion of fundamental particles,
rather t h a n the space structure.
T h e ideas which have been sketched above
will become clearer if we now summarize some
of our results. 4
E v e r y cosmological background is a Rieman-
nian manifold with a metric of the form
ds2 = y(t, r)(dt2-dx2-dy2-dz2),
(0.2)
r2 = x2+y2+z2.
W e shall see t h a t y is not an arbitrary function
of t a n d r; b u t for the conclusions which we shall
now draw the particular form of y does n o t enter
the argument. A coordinate system in which the
Riemannian metric has the form (0.2) will be
called a cosmological coordinate system, or, briefly,
a c.c.s.
T h u s every cosmological background is con-
formal to a Minkowski background. Physically,
the light geometry is t h a t of a flat Minkowski
continuum. T h e line element (0.2) appears as a
natural generalization of the Minkowski space in
restricted relativity.
W e may, however, interpret (0.2) in a different
way and say t h a t every cosmological background
differs from a Minkowski background only b y a
gauging function determining the behavior of
clocks and measuring rods. T h i s s t a t e m e n t
requires some amplification. Starting from (0.2),
and without transforming the coordinate system,
we introduce new clocks and measuring rods b y
means of the gauge transformation, 5
ds' — \ds-, X = 7~*.
We then have 7 , = T _ 1 T = 1, and the vector field,
characterizing the gauging (i.e., the metrical
connection), changes from x% — 0 to
x / = d(log.A)/d**= - J d ( l o g y)/dx\
W e can, therefore, characterize a cosmological
space b y
ds'* = dt2-dx2-dy2-dz\
(0.3)
x/=-id(log7)/d*«,
i.e., b y a Minkowski metric a n d a n integrable
4
L . Infeld, "A New Approach to Relativistic Cos-
mology," Nature 156, 114 (1945).
5
H. Weyl, Space, Time, Matter (Methuen & Company,
Ltd., London, 1922), Chapter II, Section 16.
252 L. INFELD AND
gauging field %/• This is a new geometric picture
and, though (0.3) can always be replaced by
(0.2), this new interpretation is very suggestive
as it abandons all discussion of curved, expanding
universes and shifts the responsibility for cos-
mological phenomena to the gauging field, the
gradient of a scalar field, which determines the
behavior of clocks and measuring rods.
It should be added that this interpretation,
though based on Weyl's famous work, has
nothing in common with Weyl's unified field
theory. No connection is assumed between the
vector Xi a n d the vector potential of an electro-
magnetic field.
The next conclusion which may be drawn from
(0.2) and (0.3) is that Maxwell's equations are
the same for flat space as for any cosmological
space.6 This immediately follows from (0.3) and
from the fact that Maxwell's equations are gauge
invariant. But for the sake of clarity this simple
conclusion will be deduced from (0.2). We write
Maxwell's equations for empty space in the
usual form
dFii/Bxh+dF;k/dxi+dFki/dxi = Q,
(0.4)
d((-g)*/w)/&s/ = 0.
We have
(0.5)
= vikVilFkh
where
ri o o 0^
0 -1 0 0
0 0 - 1 0 servations. It is not claimed that the first pos-
[o o o - J
Therefore Maxwell's equations, which are dif-
ferential equations for Fijf have the following
form both for a Minkowski and a cosmological
background:
dFij/dtf+dFjk/dxi+dFki/dx'^O,
(0.6)
r}ikrj3ldFki/dxJ' = 0.
Another problem which suggests itself here is
that of Dirac's relativistic equations for an
6
This conclusion seems to be unknown. Compare E.
Schrodinger, "Maxwell's and Dirac's Equations in the
Expanding Universe," Proc. Roy. Irish Acad. 46A, 25-47
(1940).
A. SCHILD
electron. Indeed, much work was done in solving
Dirac's equations for cosmological spaces. From
our point of view, the problem, properly formu-
lated, is the following: Are Dirac's equations,
like Maxwell's equations, insensitive to the
choice of the y function or not? The answer to
this question is not as simple as in the case of
Maxwell's equations and requires a special
investigation which will be given elsewhere.
The most obvious conclusions resulting from
the fact that the cosmological background can
be represented by (0.2) or (0.3) have been formu-
lated above. It has been seen how the problem
of space structure disappears if the c.c.s. is used
and it will be shown how the problem of the
motion of fundamental particles appears instead.
In a c.c.s. the fundamental particles are, in
general, not at rest. Three types of motion are
possible, namely, the oscillating motion, the
converging-diverging motion, and the simple case
of rest. To each permissible form of the function
7 there belongs at least (and, in general, exactly)
one kind of motion. This motion, and not the
space structure, characterizes the universe. It is
the study of the permissible functions y and the
associated motions which forms the chief content
of this paper.
1. DERIVATION OF THE LINE-ELEMENTS
The derivation of the metrical forms which
describe the universe as a whole will be based on
three distinct postulates. All three assumptions
have immediate physical significance. The last
two are well justified by recent astronomical ob-
tulate is completely independent of the other
two; however, as it has a simple physical content,
we prefer to introduce it as a separate require-
ment.
Postulate I on light-geometry.—There exists a
coordinate system such that the geometry of
light rays is the same as in flat Minkowski
space; i.e., light rays travel along straight lines
with constant velocity c=l, say.
Mathematically, this implies that the four-
space must be conformal to Minkowski space,
i.e., the line element may be written in the form
ds2 = y{dt2-dx2-dy2-dz1)1 (1.1)
i 3
= yrjijdx dx \ (l.H)
 KINEMATIC
where 7 is a function of /, x, y, z. In (1.11) the
usual summation convention applies to double
indices,7 xl (i = 0, 1, 2, 3) = (t, x, y, s).
Postulate 11 on isotropy.—The universe, char-
acterized by the' line element (1.1), is spatially
isotropic.
Mathematically, the cosmological line element
must be invariant in form under a 3-parameter
continuous group of rotations which leaves every
point on the t axis fixed. An immediate conse-
quence is that 7 is a function of / and x2-\-y2-\-z2
only:
y = y(t,r), r2 = x2+y2+z2. (1.2)
The assumption of isotropy is well justified by
nebular counts, 8 which indicate that the dis-
tribution of nebulae, as observed from the earth,
exhibits spherical symmetry when averaged on a
sufficiently wide scale.
If suitable assumptions are made as to the
absolute average magnitude and the luminosity
of nebulae, and the density distribution of inter-
nebular matter, as yet unobservable, recent
astronomical research8 indicates that, on a scale
large compared to the mean distance between
nebular clusters, matter is uniformly distributed
throughout the universe. Thus, neglecting local
irregularities, the homogeneity of the material
universe leads, by what is usually known as
Mach's principle, to the following:
Postulate III on homogeneity.—The view of the
universe is the same for every fundamental
observer.
In our model of the universe we have funda-
mental particles, each moving along a geodesic
line. Our original coordinate system is related to
one fundamental particle; by this is meant that
the geodesic world-line of this particle is the /
axis x = y = z = 0. If we now change the coordinate
system, relating it to any other fundamental
particle, our view of the universe remains un-
changed ; in particular, ds2 is invariant in form.
Mathematically, the line element ds2 must be
invariant in form under a continuous group of
transformations, moving the / axis into world-
7
Latin indices will consistently range over 0, 1, 2, 3,
while Greek indices will range over 1, 2, 3.
8
A comprehensive account of astronomical observations
of cosmological interest is to be found in E. P. Hubble,
The Observational Approach to Cosmology (Clarendon
Press, Oxford, 1937).
COSMOLOGY 253
lines different from it. The existence of such a
group of coordinate transformations restricts the
possible functions y(t, r). In the following we
find line elements which satisfy our homogeneity
postulate. However, the treatment adopted here
gives no indication as to whether the admissible
forms have been exhausted or not. In the
appendix, the powerful methods of the theory
of continuous groups are applied to the problem
and it is there shown that the simpler arguments
of this section do, indeed, yield essentially all
possible line elements.
Consider the hyperquadric
z02-z12-z22-zz2-(l/K)zi2=-(l/K),
(1.3)
K = constant
in five-space and project stereographically from
the pole (0, 0, 0, 0, —1) onto the tangential
hyperplane 24 = 1. If the point (s0, Z\, s2, 23, 24)
on the hyperquadric is projected into (x°, x1, x2>
xz, 1), a simple calculation yields the equations
xi l+Ka/4:
Zi==: , 24 — ,
l-Ka/4 l-Ka/*
i = 0, 1,2,3, (1.31)
where i 2 2
a = rnjx x'' = i -r . (1.311)
It is well known that, for a suitable metric, the
stereographic projection is conformal. Thus we
obtain, by differentiating (1.31), the following
identity:
/(so, z4)(dz02-dzl2-dz22-dz32- {\/K)dz,2) =
y(t, r)(dt2-dx2-dy2-dz2)=ds2, (1.32)
where
2
y(t, r) = /(*„, 20/(1 -Ka/4) , (1.33)
s0, 24 being expressed in terms of t,r by (1.31).
The problem of finding transformations leaving
(1.1) invariant is obviously equivalent to that of
finding transformations of the five-dimensional
s space which leave (1.3) and the left-hand side
of (1.32) invariant in form.
One such transformation group, if / is a func-
tion of So only, is that of linear transformations
of the four variables zh s2, s3, s4 leaving Si 2 +s 2 2
+z32 + (l/K)zi2 invariant, with z0-^z0. The group
is, for positive K, that of real orthogonal trans-
formations, i.e., Euclidean rotations, of the
254 L. I N F E L D AND A. SCHILD

variables Z\, 22, zz, K~h±\ for negative K, that of hand side of (1.32) are invariant under Lorentz
Lorentz transformations of the " time-like" transformations of the variables z0, 21, z2y z% with
variable (—K)~h± and the "space-like" variables Zi—>Zt, i.e., with a—>a. By (1.31), we see that
zu £2, 23. Thus/=/(js 0 ) yields, by (1.33), the line such transformations of the five-space are
elements simply the Lorentz transformations in our
original space of variables x\ The new line
^ 2 = ( 1 -Ka/4)-*f(t/(l -Ka/tyrHjdxidx*, (1.4) elements are, by (1.33), of the form
which satisfy our three postulates. The com- ds2 = y(a)7]ijdxidxJ\ (1.5)
plicated transformations (see appendix) leaving where
this form invariant appear, by the above, simply 2
as ' 'rotations" of a four-dimensional subspace T (a) = ( l - J £ a / 4 ) -
about a line (the ZQ axis) in a five-dimensional
manifold. X/(-(-20-» ). (1.51)
The metric form (1.4) is admissible for all
non-zero values of K. It is natural to consider the From the way in which it was obtained, it is
limiting case when K = 0, and clear that (1.5) is equivalent to (1.4; K<0).
ds2 = f(t)rjijdxidx3'. (1.41) However, (1.5) is much simpler than the previous
forms (1.4), and this is also true of the group of
The limiting process K—>0 is difficult to carry coordinate transformations which leave the line
out on the five-dimensional linear transforma- element invariant and hence of the motion of
tions of the z considered above. However, it is fundamental particles. Thus we shall, whenever
immediately obvious that the line element (1.41) possible, prefer to characterize universes by the
conforms to the homogeneity requirement as it metric (1.5) rather than by (1.4) if K<0.
is invariant under spatial translations The last cosmological line element which we
(*, x, y, z)->(t, x + Z, y + rj, * + ? ) . (1.42) derive is equivalent to (1.41), where K = 0. It is
more cumbersome than (1.41) and need not be
We shall formally include (1.41) in the metric discussed in detail. However, we include it in
forms (1.4) by permitting K to become zero. this section for the sake, of completeness. The
Then, as will be shown in the following and in inversion
the appendix, the metric forms (1.4) describe all
possible universes which satisfy our three pos- {t.x^-^i-t/a.x^/a), < r = l , 2 , 3 , (1.61)
tulates. The line elements (1.4) can be written in is a conformal transformation of Minkowski
the form space and changes (1.41) into the form9
ds* = f(t/(l -Ka/4))ds0\ (1.43)
ds2=(\/a2)f(-t/a)rHjdxidxK (1.6)
where dso2 is the line element of an indefinite
4-space of constant curvature —K. This may be The minus-sign in the inversion (1.61) is neces-
contrasted with the forms (0.1) where a 3-space sary to preserve the sense of time. The group of
of constant curvature is multiplied by an arbi- transformations which leave (1.6) invariant are
trary function (of time). obtained from the spatial translations (1.42) by
We shall now obtain two further metric forms subjecting them to the inversion (1.61).
which are admissible. However, as they are ob- This completes our survey of the cosmological
tained from some of the line elements (1.4), for line elements satisfying our three requirements.
K<0, by coordinate transformations, they do The forms (1.4), (1.5), (1.6), are shown, in the
not yield new universes and are essentially appendix, to exhaust all possibilities, except for a
equivalent to those included above. trivial change of the temporal origin, i.e., except
If K<0, then z02 and ( — \/K)z£ have the same 9
It may be noted that if (1.4) be written in the form
signature in (1.3) and in (1.32). Thus we may ds* = <f>(Kt/±(l-i£a/4))(i£/4)2(l-Ka/^Sadx^x^ then,
interchange those two variables. Then / is a ifplicitly, the function <f> does not involve the parameter K im-
we obtain (1.6) from (1.4) by the limiting process
function of z± only, i.e., of a; (1.3) and the left-
 KINEMATIC COSMOLOGY 255

TABLE I. The cosmological line elements.

K
Case («>0) Transformations preserving ds
2
I 1/a (l-a/4a*)-y(t/(l-a/4**)) Rotations in (21, 0:24) plane
II - 1/a2 (l-fa/4a 2 )- 2 /(//(l-fo/4a s )) Lorentz transf. in (zi, «z4) plane
II' y(a) Lorentz transformations of x*
III 0 /(0 Spatial translations of x*
III' (l/a»)/(-//a) From III by inversion

for the forms obtained from them by the trans- The / axis x — 0 has, in the z space, the para-
formation /—>£+constant. metric equations
We summarize the results of this section in
Table I. Zo = V ( l - * 7 4 a 2 ) , *i = 0,
(2.3)
24=(l+/V4a2)/(l~/2/4a2),
2. THE MOTION OF FUNDAMENTAL PARTICLES
by (1.31). The equations are those of a rec-
The postulate of homogeneity ensures the
tangular hyperbola in the (z0, 01Z4) plane, as
existence of coordinate changes which leave the
shown in Fig. la. The path of a fundamental
metric form invariant and transform the / axis
particle in the z space is obtained by rotating
into other world-lines which are the t axes of the
the hyperbola about the z0 axis through some
new coordinate systems. It follows from the
angle p. In Fig. lb, this is the hyperbola in the
isotropy of the line element, which must hold in
plane p. We obtain the world-line of the funda-
all equivalent coordinate systems, that these
mental particle in the tx plane by projection
world-lines are geodesies. Thus it is natural to
from the pole 5 = ( 0 , 0 , - 1 ) onto the plane
identify such geodesies with the world-lines of
p . = = - £ 4 = l . In the figure, the points of the
the fundamental particles. If xl—>xfi(x) is a
hyperbolas in the planes p and P, which are
transformation leaving (1.1) invariant, then the
related by projection from S> are denoted by the
equations of motion of the fundamental particle,
same letter, lower case letters being used for
which is related to the new coordinate system,
points in p and capital letters for points in P .
are
It should be noticed that the branch abc of the
x'*(tyx, j , s ) = 0 , (7 = 1,2,3. (2.1) hyperbola in p is projected into the finite arc
The spatial isotropy of the universe implies ABC in P .
that the world-line of a fundamental particle Before obtaining the algebraic expressions for
must, in cosmological coordinates, lie in a plane the motion of fundamental particles, the im-
through the t axis; i.e., the fundamental particles 2.
move radially. Thus we may, without loss of
generality, consider fundamental world-lines in
the tx plane only, and, throughout this section,
assume the two equations of motion
y — z — 0. (2.2)
u—
The third equation, describing the motion along
the x axis, will be obtained and discussed in each
of the five cases enumerated in Table I.
Case I—K=l/a2>0.—We need consider only
Euclidean rotations in the (zu azi) plane, since,
by (1.31), 37 = 2 = 0 implies z2 = 23 = 0. Thus, sup-
pressing two dimensions, the rotations about the
z0 axis and the paths of fundamental particles can u-u
easily be visualized by means of a diagram. FIG. la. / axis in z space.
256 L. INFELD AND A. SCHILD

Plane P^

|\Plane p c i rc FIG. l b . The hyperbola in

plane p is the world-line of a
I I fundamental particle in z
space. The world-line in the
! ^ ^
<h—^
„ I tx plane is obtained by pro-
jection from S onto the tx
plane P. Points related by
this projection are denoted by
the same letter (lower case
letters for points in p, capitals
for points in P).
/
D \I
2,

•X—x-
eCZ*.

2.

portant problem of connectivity has to be dealt tified with the points, C, E, R (and c, e, r),
with. It follows immediately from the fact that respectively. The path of a fundamental particle
an operation of projection has been used that is the branch abc (c^a) of the hyperbola in p
the "ends at infinity" of any straight line must or the arc ABC (C=A) in P . The set of all
be identified. Thus both the z and x spaces have fundamental particles is obtained by letting p
the connectivity of projective spaces. We shall, assume all values from — w/2 to TT/2.
however, postulate a further connectivity. Re-
turning for the moment to the full five-dimen-
sional space, the points (s0, Si, z2, s3, z±) and
(so, — Si, —z2l —S3, — £4) on the hyperquadric
(1.3) will be identified. This is permissible as the
transformation (s0, %i, z2i s3, s4)—>(%, — zh —z%,
—S3, —ZA) leaves the form (1.32) invariant, /
being a function of ZQ only. Thus the three-
dimensional section
So = constant,
(2.31)
Si +S 2 + S32 + «2S42 = S02 + « 2
2 2

is a hypersphere on which antipodal points are

identified, i.e., it is an elliptic three-space. In
FIG. 1c. Section so=constant of hyperboloid
Fig. lb, the points A, B, Q (and a, bi q) are iden- — 2 0 2 - j - zi2 -f- OL2Z42 = a2,
 KINEMATIC COSMOLOGY 257

Proceeding now to t h e analytic t r e a t m e n t of

the motion of fundamental particles, we p u t
X=(l/2a)(t+x), Y={\/2a)(t-x) (2.32)

and have, b y (1.31),

z0/a=(X+Y)/(l-XY),
z1/a=(X-Y)/(l-XY)1 (2.33)
zt=(l+XY)/(l-XY).
Let
X = tan£, r=tan?7, (2.34)
and
%+y = T, £—n = p] (2.35)
then
2 0 /a: = t a n r , 3 i / a = sin p/cos r,
(2.36)
s 4 = cos p/cos r.
FIG. Id. The world-lines of a fundamental particle
(ABQA) and of a light ray (PQR, RS, SP) in elliptic
r and p are parameters on the hyperboloid ob- universes of type I.
tained b y rotation of the hyperbola in Fig. l a
a b o u t the z0 axis, p being, a s before, the angle of P u t t i n g p = ± ^ 7 r in (2.4) while r varies a n d
rotation; a / c o s r is t h e distance of a point on r = ± j 7 r while p varies, we obtain the two
the hyperboloid from the Zo axis. Fig. l c shows rectangular hyperbolas
a section z0 — constant, a n d the geometric content
of the parameters r a n d p is clearly indicated. -*2=±4c (2.42)
T h e elliptic connectivity identifies the pairs of
which bound our elliptic universe, a s shown in
p o i n t s (r, P)=S(T+X, p) a n d (r, P) = (T, P+TT),
Fig. Id. T h e points outside the concave quad-
as is immediately verified b y (2.36). T h u s we
rangle formed b y the four branches of the hyper-
m a y restrict both parameters to the interval
bolas (2.42) are each identified with a n interior
— 7r/2 to 7r/2. A straightforward calculation
point b y the elliptic connectivity (2.37). In
shows t h a t the pair of points (X, Y) and
Fig. Id t h e same letter is used to denote a pair
( — 1 / X , — 1 / F ) m u s t be identified, i.e.,
of points which are thus identified. T h e topology
(t, x) = (-4a2t/a, 4a2x/a), (2.37) of the tx plane is t h a t of a torus which can be
obtained b y first folding t h e plane along t h e
points connected b y an inversion which leaves
t axis, say, a n d connecting t h e branches of t h e
the line element (1.4, K=l/a2) invariant, a n d
hyperbolas (2.42) which meet the x axis; then
which changes (z0, Si, ZA) into (s 0 , ~zh —ZA).
folding along the x axis a n d connecting the two
T h e equation of motion of a fundamental branches of the other hyperbola with each other
particle is simply p = constant, or without crossing.
(l/2a)(*+*) = tan £ ( T + P ) , Returning to the motion of a fundamental
(2.4) particle, characterized b y the world-line ABQA,
(l/2a)(t-x) = tsini(T-p), say, in Fig. Id, t h e following simple conclusions
where r is a variable parameter. Eliminating r, are i m m e d i a t e : T h e fundamental particle, start-
these equations become ing from A, moves toward t h e spatial origin x = 0
with decreasing velocity until it is a t instan-
t2-x2-(4:a/v)x+4:a2 = 0, y = t a n p. (2.41) taneous rest a t B, i.e., when / = 0 and
T h u s the world-lines of fundamental particles x = xmin=(2oA)[+(l+i;2)*-l]. (2.43)
are rectangular hyperbolas which do n o t meet
the t axis; their common transverse axis is t h e I t then recedes with increasing velocity, first to
x axis a n d their a s y m p t o t e s are t h e null lines Q, then to A, a n d the cycle is repeated. T h u s , the
t=±(x+2a/v). observer a t the spatial origin sees the funda-
258 L. INFELD AND A. SCHILD

interval ( — 2a, 2a) of one period's d u r a t i o n m a y

correspond t o a n infinite proper time. T h e n ,
since all experiences of a n y observer are within
this finite / interval, the motion of fundamental
particles is no longer oscillating, and the light-
signal emitted by a fundamental observer does
not return throughout his entire proper life. By
T a b l e I (case I), the proper time corresponding
to the t interval ( — 2a, 2a) is
2a -2
(l-t2/4a2)~1
/ — 2/y •/ o
XfKt/(l-t2/4a2))dt. (2.44)
FIG. 2. Fundamental particles in universes of type II'.
W e have t h e following criterion: A universe of
mental particles move back and forth; and we type I is periodic or non-periodic in its physical
m a y well typify this motion b y the adjective behavior according as the integral 5 converges
oscillating. or diverges. Keeping this in mind, we shall
T h e first bounding hyperbola x — £ = 4a , nevertheless, for reasons of economy, retain the
2 2 2

whose two branches are identified, is the world- term oscillating t o typify the general universe 1
line of the fundamental particle a t the greatest and t h e motion of fundamental particles.
distance from the spatial origin. For this particle T w o simple examples follow, t h e first of a
#min = 2a:. As is clear from Fig. I d , the coor- periodic a n d t h e second of a non-periodic
dinate distance 4 a m a y b e described a s t h e universe:
"perimeter of the universe" a t time £ = 0; i t is ds2= {(1 -a/^.a2)2-\-t2/a2}-l7lijdxidx^ (2.45)
the finite length of the x axis.
T h e second bounding hyperbola x2 — t2 = ~4=a2, is the line element of a cosmological model of
whose two branches are identified, is the locus type I, which will later (in p a r t II) be identified
of the points a t the greatest distance from the with the Einstein universe. W e immediately have
spatial origin on each world-line of a funda- 2a
mental particle. T o p u t i t differently, i t is the S~- dt/(l +t2/4:a2)=Tra. (2.46)
locus of the events where the fundamental par- / _9«
ticles reach the amplitudes of their oscillations,
T h u s S is finite and the universe is periodic.
stop receding, a n d s t a r t moving toward t h e
Next, we consider the De Sitter universe of
spatial origin. T h e coordinate time interval 4 a
type I, sometimes referred t o as the De Sitter-
is the period of the oscillating universe as deter-
Lanczos universe (see P a r t I I ) . I t s line element is
mined by the fundamental observer x = 0.
T h e periodicity of the oscillating universe is ds2= (1 -a/ia^Vijdx'dxK (2.47)
further exemplified b y following the p a t h of a
light ray emitted a t some event P in the direc- In this case /»2a
4U
tion of t h e positive x axis. W i t h o u t loss of dt/(l-t2/4a2) (2.48)
generality, we m a y take P on the t axis of our / -2a
coordinate system. Keeping t h e elliptic con-
TABLE II. The motion of fundamental particles.
nectivity in mind, we see t h a t the world-line is
PQR, RS, SP, a s shown in Fig. Id. T h u s the Case Equation of motion Type of motion
light emitted by a fundamental observer returns
I t2-X2-(4a/v)x+4a2 = 0 Oscillating
to him after a finite coordinate time 4a. II t2 - x2+ (4a/v)x - 4<x2 = 0 Converging-diverging
I t is i m p o r t a n t t o note t h a t t h e physical II' x = vt
III x— constant Rest
behavior of a cosmological model of t y p e I is n o t 2 2
necessarily of a periodic nature. T h e finite . t
nr t -x -x/v=o
 KINEMATIC
converges logarithmically at both limits. Thus
this universe is non-periodic in its physical be-
havior.
In conclusion, we may note that the spherical
universe in which antipodal points of the section
(2.31) are not identified is similar in its behavior
to the elliptic universe described above, and need
not be discussed in detail.
Case II—K = — 1/a2 < 0.—We shall discuss
this case briefly, mainly for the sake of com-
pleteness; in most applications, I I ' rather than
II will be appealed to. The equations of motion
of fundamental particles are immediately ob-
tained from those in Case I by applying the
transformation a—> — ia. However, the variables
X, Y, £, 7), r, p, introduced above, are now
imaginary. In order to avoid this, the definitions
will be slightly changed and these variables, in
Eqs. (2.32) to (2.36), replaced by iX> iY, i£, irj,
ir, ip. We then have
X=(l/2a)(t+x), Y=(l/2a)(t-x), (2.5)
z0/a=(X+Y)/(l+XY),
Zi/a=(X-Y)/(l+XY), (2.51)
z*=(l-XY)/(l+XY),
X = tanh f, Y= tanh 77, (2.52)
Z+v = r, f-i? = p, (2.53)
So/o: = tanh r, £i/a = sinh p/cosh r,
(2.54)
24 = cosh p/cosh r,
where all variables are now real.
The equation of motion of a fundamental
particle is p = constant, or,
t2-x2 + (4:a/v)x-4a2 = 0, y = tanh p. (2.55)
It should be noted that the parameter v is now
limited to the interval —1 to 1. The world-lines
(2.55) are rectangular hyperbolas meeting the
t axis in the two fixed points t= =k2a.
In contrast to the previous case, the only con-
nectivity of the tx plane in the present case is that
of a projective plane. All finite points represent
distinct events and no two are identified.
Rather than continue the examination of this
cosmological model in the c.c.s. II we shall find
it more convenient now to introduce the c.c.s.
ir.
COSMOLOGY 259
4 4
FIG. 3. Fundamental particles in universes of type III.
Case IF.—From the discussion immediately
preceding equation (1.5) we know that it is the
group of Lorentz transformations which leaves
the line element II' invariant. Under such trans-
formations the t axis is carried into the line
-Vty \V\ < 1 , (2.6)
which is the world-line of a fundamental particle.
The fundamental particles all move radially,
each with constant velocity v. They converge,
meet at the origin t = x = 0, and then diverge. This
simple behavior of the particles is shown in
Fig. 2. It should be noted that the whole ob-
servable universe, into which particles can
penetrate, is confined to the interior and the
surface of the double light-cone with vertex a t
the origin t = x = 0.
The explicit transformation which leads from
the coordinate system II to II' is easily obtained
in terms of the variables X, F. As we saw in
the first section, the transformation, in the
z space, is
(so, 2i, (xzA)—->( — az±, Zu Zo). (2.61)
The minus sign has to be introduced in order
that the sense of time be preserved. Using (2.51),
we obtain, by a short calculation,
(X, Y)-+«X-1)/(X+1),
(F-1)/(F+1)). (2.62)
If this transformation is applied to the equation
of motion (2.55), the equation of motion (2.6)
results. Thus the parameter v is actually the
same in both (2.55) and (2.6). It might be
260 L. INFELD AND A. SCHILD

noted that (2.62) transforms the time interval d of the complete group G, such that Gi moves
{ — 2a, 2a) in Case II into the interval (0, <*>) of
a plane through the / axis into itself. The trans-
the / axis in II'. formations of Gi may be combined with suitable
Case III—K — 0.—It follows from the trans- spatial rotations to yield the required three-
formation (1.42) that the equation of motion of parameter group G. The invariant plane of Gi,
fundamental particles is the tx plane, say, determines a preferred spatial
direction. Thus it is convenient .to introduce
x — constant. (2.7) cylindrical spatial coordinates (x, p, <p) in which
The fundamental particles are at rest, as shown the line element (1.1) becomes
in Fig. 3. ds2 = y(t, r)(dt2-dx2-dp2-p2d<p2)',
Case III'.—Subjecting equation (2.7) to the (A2)
inversion (1.61), we obtain the equation of r2 = x2+p2.
motion
As we are not interested in spatial rotations, we
t2-x2-x/v = 0, (2.8) may at the outset assume
where v is a constant. Thus the motion of funda- v = v, ^ = 0. (A21)
mental particles is along rectangular hyperbolas,
all touching the t axis at the origin. The equations of Killing are now, for i = j ,
The results of this section are summarized in y,t^+y,r^x/r+y,r^p/r+2y^t = 0y
Table II. ' t x p x
y,t% +y,r£ x/r+y,rZ p/r+2y£ ,z = 0,
APPENDIX (A22)
y.t¥+y,rS*x/r+ytrZ'p/r+2yZ'.p = 0t
In this section the theory of continuous groups
will be applied to the problem of finding all y,ttt+y,r£xx/r+yApp/r+2y^/p = 01
functions y(t, r) in the line element (1.1) which and for i^jy
satisfy the postulate of homogeneity. This prin-
ciple states that the space with the metric form
(1.1) admits a continuous group G of motions
which do not move the t axis into itself. The
contravariant components £* of the infinitesimal The last three equations of (A23) show that all
motion, which generates the group, satisfy the £* are independent of <p. From the Eqs. (A22),
equations of Killing10 we have
*S=**.*=*'.P={ p /p. (A24)
?gith+g*?. j+g*?. i = 0, (Al)
We shall now proceed as follows: The first
where the comma denotes partial differentiation;
three equations of (A23) and Eqs. (A24) will be
thus gij,k = dgij/dxk, etc.
solved and the functions £* obtained. Substi-
After solving Killing's equations for the £*,
tuting in any of the Eqs. (A22) we next obtain
the finite transformations of a one-parameter
a differential equation for y. This equation will
group are obtained by integrating the system of
be solved under different assumptions which
differential equations11
exhaust all possibilities. Finally, substituting for
l 3
dx /dp=^{x\ x\ x\ x ), (All) (•* in Eqs. (All), the finite transformations of the
group G\ may be obtained. However, as the
with the initial conditions xi — xi when p = 0\ main purpose of this appendix is to determine all
p is the parameter of the group. admissible functions 7, this last step will not be
It follows from the isotropy of the space that discussed in detail and only the results will be
it is sufficient to find a one-parameter subgroup given.
10 From the last of the relations (A24), we have
L. P. Eisenhart, Riemannian Geometry (Princeton
University Press, Princeton, 1926), Eq. (70.1).
11
Reference 10, Eq. (66.3).
 KINEMATIC COSMOLOGY 261

But, by (A24) and (A23), transformations of the group G\ are obtained by

solving the system of equations

Thus £ ^ = 0 and, similarly, £p,xs = 0. Therefore di/dp = x(2bt+e), (A31)

/,«=/ i **= s 0 and 2 2 2
dx/dp = b(i +z -p )+ei+gy (A32)
£' = p(a**+26#+2rf+d) f dp/dp = 2bxpt (A33)
where a, 5, c, d, are constants. From this and with the initial conditions (i} x, p) = (t, x, p)
(A23), (A24), we have when p — 0.
£ttt=£pp We now have the following mutually exclusive
= axt+2bx+2d+d,
possibilities:
p
¥,P = £ ,t = axp+2cp. 1. b = e = &, g^O.—Then Y,r = 0; thus
Therefore 7 = 7(0- (A4)
2 2 2 2
^ = J^(/ +p )+c(^ +p )+2to+^+F(x), This is the universe III, Table I. Solving (A31)
and, similarly, to (A33), we immediately find that the trans-
formations of the group G\ are translations along
£x = %at(x2-p2)+b(x2-p2) + 2ctx+dx+<S>(t). the x axis.
But £%=£*,, by (A23), and thus 2. & = 0, e^O.—Making a suitable choice of
the temporal origin, we replace t+g/e by t. Then
Mt2+P2)+2bt+F\x) = Mx2-P2)+2cx+$'(t). (A3) simplifies to
Hence a = 0, F=cx2+ex+f, $ = bt2-\-et+g, e, /, g, r7,*+*7,r = 0.
being constants. Finally, we obtain Thus
yssy(P-r*)u (A5)
£t = c(t2+x2+p2)+2btx+dt+ex+ff (A25)
^ = 6(/2+x2_p2)+2^+^+^+gf This is the universe II', Table I. The finite
(A26)
transformations of G\ are easily obtained and
&> = p(2bx+2ct+d). (A27) seen to be Lorentz transformations in the
variables t and x.
Substituting in any of the Eqs. (A22), we have, 3. Z>^0, g/b-e2/4b2?*0— Changing the tem-
on multiplication by r, poral origin, we replace t+e/2b by t and also
ry,t(a2+cr2+dt+f)+r2y>r(2a+d) write K/4 for {g/b-e2/±b2)~\ Then (A3) becomes
+2ry(2ct+d)= -x{ry,t(2bt+e) 2trytt+y,r(t2+r2+4t/K)+4ry = 0. (A61)
2 2
+y,r(bt +br +et+g)+4bry}. (A28) Introducing
As 7 is a function of / and r only, we have the = */(l-(/*-r2)j?:/4) f
v
alternative c = d=f=0 or & = e = g = 0.'The latter
possibility does not lead to suitable transforma- and putting
tions: Using (All), a short calculation yields the y=f(t,v)v2/t2,
explicit form of the finite transformations and it
is seen that r — 0 implies f = 0, i.e., the t axis is a short computation shows that (A61) simplifies
moved into itself. We therefore put c = d=f=0 tof,t = 0. Therefore f=f(v), and we have
and have 7=(l~(/2_r2)jK:/4)-2
2 2
ry,t(2bt+e)+y,r(bt +br +et+g) Xf(t/(l-(t2-r2)K/A)). (A6)
+4bry = 0. (A3)
This is the universe I or II, Table I, according
This is the differential equation for y. The as K>0 or K<0. The transformations which
262 L. INFELD AND A. SCHILD

TABLE III. Line elements and motion.

Type Equation of motion Type of motion

I 4 (l-a)-«/(*/(l-a)) t -r2-(2/v)r+\=0
2
Oscillating
II -4 (l+a)-*/W(l+a)) t2-r2+(2/v)r-l=0
II' y(a) r — vt Converging-diverging
III 0 7(0 r — constant Rest
III' (\/a*)f{-t/a) t2-r2-(l/v)r =0

leave the form (A6) invariant are we obtain /, t = 0. Thus f=f(v), and
x±(Z 2 --p 2 )* = 2i£-*tan { t a n " 1 ^ y=(t*-r*)-*f(-t/(t*-r*)). (A7)
2 2
X(^±(^ -p )") + 2 X - ^ } , (A62)
This is the universe III', Table I. The finite
p/t = p/t. transformations which leave the line element
(A7) invariant can be shown to be
4. b T^O, g/b-e2/4b2 = 0 —This is the last pos-
sibility and concludes our examination of spaces x±(i2- p2)$=(xMt2~ P2)h-bp)~\
which admit a group G o f motions. From Eq. (A72)
(A61), we now have p/'t=p/t.

2rty)t+y,r(t2+r2)+4ry = 0. (A71) In each of the cases examined here the trans-

Putting formations of the group G can easily be shown
v=-t/(t2-r2), y=f(t,v)v2/t2, to agree with those discussed in Section 1.

II. Special Problems and Applications

3. INTRODUCTION We introduce two changes of notation. The

We give here a short summary of the main
results of Part I12 of this paper before proceeding
to applications.
In Part I, the metric forms and the associated
motion of fundamental particles (nebulae),
suitable for the kinematical description of our
universe at large, are examined on the basis of
three postulates, namely, the constancy of the
velocity of light, spatial isotropy, and homo-
geneity. The permissible line elements are of the
form
ds2 = y(t, r)(dt2-dx2-dy2-dz2)
(3.1)
= y(t, r)r\ijdxidx\ r2 = x2+y2+z2.
The function y is restricted by the homogeneity
requirement, and three distinct types of cos-
mological models are obtained. The permissible
functions y and the radial equations of motion
of fundamental particles are listed in Tables I
and II.
12
The numbering of sections, equations, and footnotes
is carried over from Part I. References to Eqs. (0.1)-(2.8)
and to footnotes 1-11 are to those of Part I.
equations of motion of fundamental particles,
which move radially, will be written in terms of
/ and r; it will be understood that they are to be
supplemented by the equations 0, <p = constant.
Next, in cases I and II, we change the coordinate
unit, measuring / and r in terms of the natural
cosmological unit 2a. This is achieved by the
coordinate transformation
x{/2a-*x\ (3.2)
The coordinates, being measured in natural units,
now have no physical dimensions. This intro-
duces a considerable simplification in most
formulas. Table III is a summary of the main
results of Sections 1 and 2.
In Table III, v is a parameter characterizing
the individual fundamental particle, and
a — t2 — r2 (3.3)
In universes of type I, the points
(t,r) = (-t/a,r/a) (3.4)
 KINEMATIC COSMOLOGY 263

are identified by virtue of the elliptic con- The cosmological line element (3.1), in spatial
nectivity. polar coordinates, is
The forms II' are obtained from II by the
ds2 = y(t, r){dt2-dr2-r2dB2-r2 sin2 6d<p2), (4.3)
coordinate transformation
where all permissible functions y are listed in
(X, F ) ^ ( ( X - 1 ) / ( X + 1 ) , ( F - 1 ) / ( F + 1 ) ) , (3.5)
Table III. It will be shown that the line elements
where (4.2) and (4.3) are equivalent and that they can
X = t+r, Y=t-r. (3.6) be obtained from each other by a coordinate
The forms III' are obtained from III by the transformation.
inversion We introduce the variables:
(/, r ) - » ( - * / a , r/a). (3.7) X=(t+r), Y=(t-r)y (4.31)
and
Using spherical spatial coordinates (r, 0, <p), the * = *(r+p), * = i ( r - p ) . (4.32)
transformation equations (3.5) and (3.7) must
be supplemented by It may be noted that X, F, and, as will be seen
shortly, £, rj, r, p, are essentially the same quan-
(0, <p)->(6, <p). (3.8) tities as were introduced in Section 2 (where
4. THE COSMOLOGICAL AND THE ROBERTSON y and z were both zero and therefore x appeared
COORDINATE SYSTEMS instead of r).
Case I—K = 4.—In terms of the variables X
The coordinate system by which cosmological and F, we have, by (4.3) and Table III,
models have usually been described in the past
is, in its general form, due to H. P. Robertson. 13 2_ / l X+Y\ lj /X+F\21
An a priori separation of space-time into space
\2 1-XY) 41 \ \ - X Y ) J
and time is ensured by writing the line element
in the form UXdY-(X- Y)2(d62+sm2 dd<p2)
ds2 = dT'2-Ri2da\ (4.1) (1+X2)(1+F2) '
2
where da is the positive definite metric of a The transformation to the Robertson coordinate
3-space. The assumptions of isotropy and system is
homogeneity of space impose two restrictions: X = tan J, Y= tan rj. (4.4)
Ri — RI(T') is a function of time only, and da2 is
the differential form of a 3-space of constant We immediately obtain
curvature k = l, —1, or 0. ds2 = f(± tan (£+*)) • \ sec2 (£+*?)
Let
• { 4 ^ 7 7 - s i n 2 (£-r7)(d0 2 +sin 2 dd?2)),
f
T=fdT'/R1(r'), R(r)^R1(r ). (4.11)
which, by (4.32), assumes the form (4.2-4.21),
where
Then the line element (4.1) assumes the form
R2(r) = I sec2 r/(J tan r), and k = 1. (4.41)
ds2 = R2(r)(dr2-d(T2)1 (4.2)
From this last relation we note that not all
where, with suitable spatial coordinates, we may
Robertson universes, with k = l, can actually be
write
brought into the cosmological form I. Only if the
d<r2 = dP2+sin2 P(dd2+sm2 dd<p2), function R(T) is periodic and of period ir is this
if * = 1, (4.21) possible. This can also be seen, and perhaps
2 2 2 2
da = dp +sinh p(d6 +sm 6d<p ),2 2 better, from the transformations (4.4). An in-
crease of T in r changes, by (4.32), both J and rj
ifjfe=-l, (4.22) by an amount §7r. The effect of this is to trans-
da2 = dp2+p2(dd2+sin2 6d<p2), if k = 0. (4.23) form X} Finto -1/X, - 1 / F ; and events (X, Y)
13 and (—1/X, —1/F) are identified in the elliptic
H. P. Robertson "On the Foundations of Relativistic
Cosmology," Proc, Nat. Acad. Sci. 15, 822-829 (1929). connectivity, given by Eq. (3.4).
264 L. INFELD
T h u s the forms (4.2-4.21) seem to include a
more extensive class of cosmological models t h a n
t h e corresponding forms I, T a b l e I I I . W e shall
show, however, t h a t this limitation is not very
serious.
First, the physical significance of the r-interval
7r will be examined. If (4.21) is interpreted as the
line element of an elliptic 3-space, the finite length
of a straight line is v, the linear unit being, as
usually in elliptic geometry, the radius of cur-
v a t u r e 1/k. E q u a t i o n s (4.2-4.21) show t h a t the
coordinate velocity of light rays traveling
radially is 1. T h u s the interval T on the r scale
is, physically, the time interval in which light
circumnavigates the universe and returns to the
fundamental particle which emitted it. T h i s is,
obviously, a considerable time interval and so,
even in a case where R(T) is not periodic, a large
portion of the Robertson universe can always be
represented b y our model I.
Next, it should be noted t h a t the periodicity
of R(T) does not necessarily imply the periodicity
of JRI(T'). In fact, in some i m p o r t a n t special
models the infinite proper interval (•—«>, °°) on
the r ' scale transforms into a finite interval T on
the r-scale, in which case the universe of t y p e I
is non-periodic in its physical behavior (see dis-
cussion in Section 2).
T o illustrate this, we examine the De Sitter-
Lanczos universe, 1 4 where k = l, and
Ri2(r') = l cosh 2 2 ( T ' - T 0 ' ) ,
ror being a constant. RI(T') is not periodic. W e
have
/ /
r = 2 f sech2(r'-To )dT
= 2 t a n " 1 e 2(T, ~ ro,) - \TC.
T h e interval — oo < / < oo corresponds to --ibr
<T<%W and, by (4.4), to -\<t<\. A short
calculation yields
£2(r)=isec2T.
t h e r e f o r e , b y (4.41), / = 1 , and, by I, T a b l e I I I ,
the De Sittef-Lanczos line element has the form
ds2 = (l-a)~2r}ijdxidxJ'
in the c.c.s.
14
H. P. Robertson, u Relativistic Cosmology," Rev.
Mod. Phys. 5, 62-90 (1933), Eq. (6.6), where we put c= 1,
AND A. SCHILD
W e have shown t h a t there is no essential
metrical difference between Robertson's uni-
verses and ours. However, there is an essential
topological difference. Suppressing 6 and <p, we
have in the Robertson case universes represented
by a (r, p) plane in which t h e events (r, p)
and (r, P+T) are identified; i.e., the universes
are topologically equivalent t o a cylinder, whose
r dimension goes to infinity. T h e topology of
t h e universe / is different; t h e events (r, p)
and (r+7r, p) m u s t also be identified, because
in the c.c.s. t h e y are connected b y t h e inver-
sion (3.4). T h u s we now have the topology,
not of a cylinder, b u t of a torus. Even in t h e case
of a "spherical u n i v e r s e / ' t h e topology remains
t h a t of a torus as the event (r, p) m u s t now be
identified with (r, p+27r) and with (r+27r, p).
T h o u g h the topological difference between Rob-
ertson's universes and ours does n o t reveal itself
in the problems considered in this paper, it will
become decisive in the t r e a t m e n t of Maxwell's
*and Dirac's equations on a cosmological back-
ground.
Case II—K = — 4.—The transformation to the
Robertson coordinate system is obtained from
(4.4) by replacing X, Y, £, v, by iX, iYy i£, irj:
X = tanh £, F = t a n h r\. (4.5)
Corresponding to (4.41), we now have
(4.42) ^ 2 ( r ) = i s e c h 2 r - / ( i tanh r),
and
•*=-l. (4.51)
T h e r e are in this case no limitations on the
function R{j).
Case IF.—The line element is, b y (4.3) and
(4.43.) Table I I I ,
ds2 = y(XY)-XY
dXdY-(±(X- F ) ) 2 ( d 0 2 + s i n 2 6d<p2)
(4.44) XY
T h e transformation to the Robertson coordinate
system 1 5 is
(4.45)
X = e2*y Y=e2\ (4.6)
15
This transformation has been obtained by A. G.
Walker, "On the Formal Comparison of Milne's Kine-
matical System with the Systems of General Relativity,"
M. N. R. A. S. 95, 263-269 (1935).
 KINEMATIC COSMOLOGY 265

We immediately obtain Case I I ' : (t, r)—>(W, kr), (4.82)

2
<fc2 = 7(e «+*>).e*<H-n> Case I I I : (tfr)->(t+t0tr). (4.83)
2 2 2 2
•{4<^-sinh (£-rj)(dS +sin dd(p )},
These transformations change the cosmological
which, by (4.32), assumes the form (4.2-4.22), line elements into others which are formally
where different but of the same type; the equations of
motion of fundamental particles are preserved,
i?2(T)=e2TY(e2r), a n d k=~l. (4.61) and, in particular, the t axis is moved into itself.
Comparing (4.5) and (4.6), we easily regain the The event (t0, 0) is transformed into (0, 0), (1, 0),
transformation formula (3.5) leading from Case (0, 0), respectively. Thus, by a suitable choice of
II to Case II'. the line element and the function yf any event
Case Illy III'—K — 0.—The cosmological line on the t axis may be made to coincide with the
element III is of the form (4.2-4.23) and no natural origin (0, 0), in Cases I and III, and
transformation is required. For the sake of com- with (1, 0) in Case II'. This different behavior of
pleteness, we may write models of type II' is due to the fact that the
motion of fundamental particles has a singularity
X = 2$, F = 2 „ , (4.7) at / = 0 (when the universe shrinks to a point).
and The magnification (4.82) may also be applied
j R2( r ) ==7 ( T ) f £ = 0. (4.71)
to a model of type III. The event (/0, 0) is
In Case III', the transformation is the inversion transformed into (1, 0), if £o^0.,
(3.7) and need not be discussed further.
It has been shown that our cosmological 5. SOME SPECIAL UNIVERSES
models are the same as those considered by In this section some special cosmological
Robertson and his predecessors. However, the models will be examined and the line element
mathematical picture as well as the physical obtained in a c.cs.
emphasis differ considerably in the two repre- Universes of Type I.—The line elements of
sentations. The universes which, in Robertson's these universes are of the form
coordinates, are of type k = l, —1, 0, are, in a
c.cs., of type I, I I ' (or II), III (or III'), re- ds2=(l -a^fit/it-a^rjijdxidx3', (5.1)
spectively. In the Robertson coordinate system
the fundamental particles are at rest; in a c.cs. and the motion of fundamental particles is, in
their motion is, in the three cases, the oscillating general, the oscillating motion. As was seen in
motion, the converging-diverging motion, and Section 2, some of the universes I are, physically,
rest, respectively. In Robertson's coordinates the not of a periodic nature. In such cases the funda-
velocity of light is a function of position and mental particles only approach and then recede
direction; in a c.cs. it is constant. from a fundamental observer within his infinite
The form of the cosmological line element (4.2) proper life.
in the Robertson coordinate system suggests the The simplest model, in our representation, is
following consideration. The temporal translation the non-periodic De Sitter-Lanczos universe, which
we denote by Si. Its line element is
r+To (4.8)
ds2= (I -a^rjijdx^xK (5.11)
changes the function R(T), and therefore also
the line element (4.2), but preserves the type of The line element of Einstein's cylindrical
the universe, which is determined by da2. The universe E is usually obtained in the Robertson
translation (4.8) induces the following trans- coordinate system:
formations in the cosmological coordinates:
ds2 = dr'2-d(T2f (5.12)
Case I: (X, Y)-*((X+to)/(l-toX),
where da2 is the metric (4.21) of a 3-space of
(Y+to)/(l-t0Y))9 (4.81) constant curvature k — 1. By (4.1) and (4.11),
266 INF ELD AND A. SCHILD

TABLE IV. Special cosmological models.

Case * Universe T Type Transformation of line elements

(a) 2
(1-tf)- (a), (*)-(&), («), (n), (e),
i£ = 4 Oscillating s, 2
(* = 1) (b) 1/4/
Ci)-(*),
E (c) 4{(l-a)2+4/2}-1
by transformation
(d) 1 II'

(e) . 1/a2 TT r
(X, K ) - K ( l + X ) / ( 1 - X ) ,
JC=-4 Converging- Mi d+F)/(i-y)).
_a
(*«-l) diverging (/) 4(l+a-20
(g) 4(l-f-a+2/)- 2 (<Z), (J)-(«), («),
(*) (1-a)"2 II' (2>*(r),
52
1/4*2 by inversion
TT/
0") 1/a (t,r)-+(-t/a,r/a).
Mi 2 2 1
(*) 4{(l+a) -4/ }-

(I) 1
M, (a), (h)->(P),
(m) 1/a2 TTT/

i£ = 0 Rest by temporal translation

(*-0) (*) 1/4*2
*-*H-l.
S» 1/4/2 nr
(l/a2)(l+2//a)"2 nr (g)-Kw),
(2) V by temporal translation
ES
(0 fl/a* iir *-^-l.

* Motion of fundamental particles.

we have RI2(T')=R2(T) = 1, and Eq. (4.41) then and, on integration,

yields i? = 2 / ( 3 c o t h 2 ^ - l ) . (5.14)
/(*)=4(l+4*2)"1.
Thus the Einstein universe has the cosmological Thus R is not a periodic function of r and we
line element can, at best, represent only a portion of the
Lemaitre universe in a c.c.s.
ds2 = 4{(l-a)2+4:t2}~1r)iJdxidxK (5.13)
Universes of Type II, IV.—The line elements
We see that in the c.c.s. the Einstein universe of these universes are preferably taken in the
does not have a simple line element, and that it form II 7 :
seems more artificial than the De Sitter model. ds2 = y(a)y]ijdxidxj'. (5.2)
Finally, we examine very briefly Lemaitre's
The fundamental particles move with constant
universe. The Lemaitre line element in the
coordinate velocities and they all meet in a
Robertson coordinate system16 is of the type
point at time t = 0. We have described this motion
k = l, and the function Ri(r') is defined by
as converging-diverging.
ar / = (32? 1 )W2? 1 /(U 1 +2)*(i?i-l). The transition from the form II' to II is
achieved by the inverse of the transformation
Since dr'^Ridr, by (4.11), and i?i(r') =i?(r),
(3.5):
we immediately have
dr = 3*dR/(R2+2R)KR-l), (X, F ) - ( ( l + * ) / ( l - •x),
16
Reference 14, Eq. (8.2), where we put Re—\, (i+iO/(i- Y)). (5.21)
 KINEMATIC
The simplest model of Type II' is the Minkow-
ski universe (y(a) = 1) which we denote by M2.
Its metric is given by the pseudo-Euclidean line
element
ds2=7)ijdxidxJ'. (5.22)
The De Sitter line element (5.11), besides
being of type I, is also quite obviously of type
II'. Thus among the models II' we have a
De Sitter universe, which we denote by S2y and
whose line element is
ds2=(l-a)-2r}ijdxidxK (5.23)
Subjecting this form to the transformation (5.21),
we obtain the line element of S2 in the form I I :
ds2=(l/4t2)rjijdxidxK (5.24)
This is also a special case of the forms (5.1) of
type I, where f(x) = l/4x 2 . Thus (5.24) is an
alternative line element of the universe Si.17
Finally, we mention a model which we call
Milne's universe and denote by Mi. Its line
element is
ds2={\/a)y]ijdxidxj. (5.25)
In this universe, the motion of free particles
along any geodesic is identical with the motion
of particles in Milne's "kinematical relativity"
under the influence of the ' 'substratum" 18 ; in
particular, the fundamental particles, which form
Milne's substratum, behave like those in all our
models II'. We may add that in Robertson's
coordinates Milne's universe has a form analogous
to Einstein's universe; its line element is (5.12)
but do-2 is now the metric (4,22) of a 3-space of
constant curvature k — — 1.
Universes of Type III, III'.—It is convenient
to use the c.c.s. I l l :
ds2 = y(t)r)ijdxidxJ\ (5.3)
In these models the fundamental particles are at
rest.
The pseudo-Euclidean line element (5.22),
17
The transformation (5.21) is a special case of (4.81),
$0=1, and therefore changes any line element of type I into
another of the same type, and also preserves the equation
of motion of fundamental particles. A similar statement is
easily verified to hold for the inversion (3.7) applied to a
line element of type I I ' .
18
C. Gilbert, "On the Occurrence of Milne's Systems of
Particles in General Relativity," Quart. J. Math. 9, 187,
Eq. (9) (1938). See also reference 15.
COSMOLOGY 267
besides being of type II 7 , is also of type III.
Thus, among the models III, we have a Minkow-
ski universe Ms with line element
ds2 — r\ijdxidxj. (5.31)
The De Sitter line element (5.24) is immedi-
ately seen to be a special case of the forms (5.3).
Thus we have a third De Sitter universe o 3 of
type III and with line element
ds2={\/tt2)rHjdxidxK (5.32)
S3 is sometimes referred to as the "stationary"
De Sitter universe.19
The Einstein-De Sitter universe ES has, in the
Robertson coordinate system,20 a line element of
type k = 0, where
R^QT')*. (5.33)
Using (4.11), a short calculation yields the line
element in the c.c.s. I l l :
ds2 = tSijdxldxj. (5.34)
This cosmological model has also been proposed
by Dirac.21
It must be emphasized that the two Minkowski
universes M2 and Ms are quite distinct cos-
mological models and differ in their physical
behavior; e.g., M2 exhibits a nebular red shift,
M3 does not. M2 and Ms are characterized by
the same line element (5.22) but by different
motions of fundamental particles, i.e., the con-
verging-diverging motion and rest, respectively.
A corresponding statement holds for the three
De Sitter universes Si, S2, Ss, which may be
described by the same metric form (5.24) and by
the three different types of motion of funda-
mental particles.
We conclude this section with a short examina-
tion of stationary universes22 We may define this
term as follows: A universe is stationary if its
line element and the equation of motion of its
fundamental particles are invariant in form
under a coordinate transformation which moves
the / axis into itself and which transforms any
non-singular event on it into any other.
19
Reference 14, Eq. (6.1).
20
Reference 14, Eq. (8.8), where we put /c£ = 12.
21
P. A. M. Dirac, "A New Basis for Cosmology/' Proc.
Roy. Soc. A165, 199-208 (1938), Eq. (6).
22
Reference 13, p. 824, assumption II'.
268 L. INFELD AND
It is easily verified that the line elements of the
universes E, Mi, Mz, are invariant in form under
the transformations (4.81), (4.82), (4.83), re-
spectively, and are therefore stationary. The line
element of the universe £3 is invariant under
(4.82), which, however, preserves the origin
(0, 0). But this line element is singular at £ = 0,
and thus the definition of stationary cosmologies
applies to S%.
The different cosmological forms of the special
models examined in this section are summarized
in Table IV.
6. THE NEBULAR RED SHIFT
The displacement towards the red of the
spectral lines of nebulae, which is roughly pro-
portional to their distance from us, is now a well
established astronomical phenomenon.8 One of
the chief advantages of many relativistic models
of our universe is that this nebular red shift
emerges as a natural consequence of their struc-
ture. In this section expressions for the red shift
will be obtained in terms of the coordinates of
the fundamental particles under observation. No
comparison with observational data will be
attempted as the coordinates xl are not immedi-
ately interpretable in terms of physical time and
distance as estimated by the astronomer. Such
a comparison involves an examination of the
apparent magnitudes and luminosities of nebulae
in the model; the problem is not difficult but is
outside the scope of this report. However, in
most cosmological models where the red shift is
not altogether absent (as it is in the Einstein
universe) the first-order red shift effect is linear,
and agreement with observation can be achieved
by a suitable choice of some of the constants of
the model, such as 2a (the cosmological unit) or
the present time t0 (which may, for practical
purposes, be treated as a constant). Thus the red
shift phenomenon, at least to the first order, js
not a very effective criterion for narrowing down
the large number of possible cosmological models.
Most restrictions on the models, which are
suitable for describing our universe at large, are
obtained from the red shift effects of higher
order and from dynamical considerations which
are outside the domain of pure kinematics.
We base the derivation of the red shift formula
on a practical, corpuscular theory of light. This
A. SCHILD
is completely justified by the fact that Maxwell's
equations have in all cosmological spaces the
same form as in Minkowski space and that
therefore light is propagated exactly as in flat
space. Appeal is also made to a fundamental
principle of relativity which states that the
proper period of vibration (i.e., the period
measured in proper time ds) of an atom emitting
a sharp spectral line is constant, whatever the
motion or position of the atom. Consider an
atom, moving with a fundamental particle
P=(t,r) of radial velocity V, emitting a light
signal at time t and again, after one complete
vibration, at time t+dt. The two light signals
reach the observer 0 at the spatial origin at
times to^t+r and to+dt0, respectively, where
dto = dt(l+ V). This is immediately seen from the
fact that the second light signal is emitted from
the event (/+dt, r+ Vdt) and reaches 0 at
time t-\-r+dt(l+V). Remembering the form of
the cosmological line element (3.1), we find that
the proper period of vibration of the atom (which
equals the proper wave-length X, since the
velocity of light is 1) is given by
ds = \ = yKt,r)(l-V2)*dt.
The proper period of vibration as observed by 0
or
is dso = yo^dt0, >
d$o = Xo = 7o*(l+V)d/, 7o = 7(*o,0).
Thus the ratio of observed to proper wave-length
XoA=(Yo/7)*((l+n/(l-W (6.1)
This expression for the red shift may be ana-
lysed into two independent components, the
Doppler effect which contributes the factor
((1+V)/(1-V))K and the gravitational effect
which contributes the factor (YO/Y)*-
An explicit formula for the red shift will now
be obtained in each of the three types of universes
and the result expressed in terms of the time £0
of observation and the distance r of the funda-
mental particle P, determined at the time t
when the observed radiation left P.
Case /.—The velocity V of a fundamental
particle is obtained by differentiating the equa-
tion of motion, given in Table III.
V=2tr/(t2+r2+l). (6.2)
 KINEMATIC COSMOLOGY 269

TABLE V. First-order red shift effects.

Universe Type AX/X to first order Remarks

Si (a) d-a)-» (4; 0 /(l+*o 2 ))f -l</o<0:v.s.*

0<to<l :r.s.
E (fi) I 4{(l-a)2+4*2)-1 0 rigorously
M2 (d) II' 1 dAo)f /o<0 : v.s. / o > 0 : r.s.
/ 0 < - l , 0 < * 0 < 1 :r.s.
(h) II' (l-a)"2
- K / o < 0 , *o>l : v . s .
Mi (j) IV \/a 0 rigorously
Mz (I) III 1 0 rigorously
Sz (») III 1/4/2 -2(\to\/t0)f / o < 0 : r . s . ; * 0 >0 : v.s.
ES (q) III (2//o3)f *o<0 :v.s.; / 0 > 0 : r.s.

* r.s. is red shift; v.s. is violet shift.

Replacing / by t0—r, we have It is interesting to note that Milne's universe Mi,

V=2r(to-r)/((t0-ry+r*+l).
Thus the Doppler effect is given by the expression
((l+TO/tt-JO)*
= ((l+/o 2 )/(l + (/o-2f) 2 ))*.
Before considering the complete red shift in
the general case, it is convenient to examine the
special case of the Einstein universe E, whose
line element is given by (5.13). Remembering
that 7 = y (t, r) — y (to — r,r), we have
for which 7 = 1/a, exhibits no red shift.
(6.21)
Case III.—The fundamental particles in these
models are at rest and hence the red shift is
completely due to the gravitational effect.
Remembering that y = y(t), we have
(6.22)
Xo/X = (7(*o)/7(*o-r))*. (6.5)
The only universe of this type without red shift
is the Minkowski universe M 3 .
We proceed to obtain formulas for the first-
order red shift effect in the special cosmological

m
models listed in Table IV. The red shift will be
* /(l-(t0-ry-r*y+4(to-ry given by the usual expression
V , (6.23)
(l-<o 2 ) 2 +4/ 0 2 (6.6)
AX/X=(X 0 -X)/X.
which, after simplification, reduces to the
reciprocal of (6.22). Thus X0/X = 1 in the Einstein It will be expressed in terms of the time to of
universe and there is no red shift. observation and a new distance coordinate f,
The Doppler effect is the same for all universes defined by
of Type I. Writing y = F-yE, where yEVijdx^x3 f = |7o*k, 7o = 7(^o, 0). (6.61)
is the line element of the Einstein universe, we
see, from (6.1), and from the fact that (7^0/7^)* From the line element (4.3) we note that f is the
just cancels the Doppler effect, that proper distance at time t0 and in the immediate
Xo /gy_/ wa-w y 2 2
neighborhood of the spatial origin. Neglecting
second-order corrections, f is actually the dis-
X \F/ \F((to-r)/(l-to
F«torr)/(\-to +2t0r)),+2Ur))/ tance estimated by the astronomer, who always
assumes the rectilinear propagation of light with
In terms of the function / , introduced in Section constant velocity.
1, F is defined by With the formulas (6.3)-(6.5), (6.61), and the
F(x) = ti+x*)f(x), (6.31) functions 7 of Table IV, the calculations are
as is easily verified by Eqs. (5.1) and (5.13). quite straightforward. Therefore no computa-
Case IF.—The fundamental particles now tional details need be given and the first-order
move with constant velocity V—r/t and 7 = 7(a). red shift effects are collected in Table V.
Thus V=r/(t0 — r), and Eq. (6.1) becomes Finally, it is of interest to note that the three
models E, Mi, Ms, are the only universes with
Xo/X = (/o7(^o 2 )/(/o-2r)7(/o 2 -2^))». (6.4) no red shift.
270 L. INFELD AND
7. THE PHYSICAL CONTENT OF THE
COSMOLOGICAL COORDINATES
Suppose a physical cosmological model be
given which is of the kind considered in this
report. In this concluding section we deal with
the problem of finding the line element of the
model and of determining the coordinates of any
event in it. This is achieved by means of idealized
experiments. They give a clear indication of the
physical content of our time and distance coor-
dinates which so far were only convenient mathe-
matical constructs.
We choose a fundamental particle (arbitrarily)
and associate with it a coincident fundamental
observer 0. Without loss of generality, we may
assume the particle to be at the spatial origin.
The observer 0 is to be equipped with a theo-
dolite and with apparatus for sending and
receiving light signals. Also he is to carry a
mechanical or atomic clock whose vibrations, ac-
cording to the fundamental principles of rela-
tivity, measure proper time ds.
For the purposes of this section, the most con-
venient coordinate system is that introduced in
Section 4, Eq. (4.2). The line element is
ds2 = R2(r){dr2-dp2
-S2(p)(dd2+sm2ddcp2)}, (7.1)
where S(p) is sin p, sinh p, or p, according as the
model is of type I, IT, or III. From (7.1) simple
transformations lead to the form of the line
element in Robertson's coordinate system [Eq.
(4.11)] or in a c.c.s. [Eqs. (4.4), (4.6), (4.7), in
cases I, II', III, respectively].
Introducing new variables
(f, p) = (aT, ap), (7.11)
where a is a constant, the line element (7.1)
becomes
ds2 = <t>2(f){df2-dp2
-a2S2(p/ot)(d62+sm2 dd<p2)}, (7.2)
where <t>(f) = (l/a)R(f/a). In (7.1), r, p are
dimensionless numbers; R has the dimensions of
ds. In (7.2), <j> is dimensionless; f, p have the 2. The observer 0 selects two fundamental
dimensions of ds. By a suitable choice of a, we particles Pi and Pi close together and at equal
can always ensure that distance from him. By the procedure outlined
above, 0 determines the coordinate distance
0(0) = 1. (7.21) OPI = OP2 = P and the small angle PxOP^dB.
A. SCHILD
We know that in the coordinate systems of
(7.1) and (7.2) the fundamental particles are at
rest. Moreover, the velocity of radial light signals
is 1; this is an important property which these
coordinate systems share with the c.c.s.
We shall now proceed in the following order:
1. A f-clock is constructed which measures
f-time at 0; it is used to obtain the coordinates
(f, p) of any event. 2. The type of the universe
and the constant a are determined. 3. The line
element is obtained. 4. The transition to a c.c.s.
is examined.
1. The observer 0 sends a light signal to a
fundamental particle P in his immediate neigh-
borhood. P reflects the light ray back to 0 who
returns it to P , and this procedure is repeated
indefinitely. The fundamental observer 0 counts
the number of times the light-pulse has reached
him and this number serves as a measure of time,
f time. This is immediately seen from the fact
that, in the coordinates of (7.2), all fundamental
particles are at rest and that the radial velocity
of light is constant.
Since, in the general line element (7.2), there
is no preferred point on the temporal axis, 0 may,
without loss of generality, denote the initial
moment at which he starts his light ray experi-
ment by f = 0. In accordance with (7.21),
0 normalizes the f-time measure jto agree initially
with the proper time of his atomic clock; i.e.,
ds = df, when f = 0. (7.22)
The mechanism just described will be called a
f-clock.
In order to obtain the coordinates (f, p, 0, <p)
of an event E, the observer 0 sends, at time fi,
a light signal to E which is reflected by E and
returns to 0 at time T^. The direction of the
light ray, measured by O's theodolite, immedi-
ately yields the polar coordinates 0 and <p.
Since the light signal travels radially with unit
velocity, the coordinates f, p of E are given by
r = i(r2+fi), P = i ( f , - f , ) . (7-3)
 KINEMATIC COSMOLOGY 271

agree. 0 can, by prolonged observation, deter-

mine ds/df as a function of f. But, by (7.2),

ds/df = 4>(?). (7.5)

Thus the function <£* and therefore also the line
element, are known.
4. Applying the transformation (7.11) to (7.2),
we regain.the line element (7.1). Regraduating
the coordinate time scale in accordance with
FIG. 4. Arrangement of light signals to determine the type (7.11), 0 obtains a r-clock.
of universe.
From the line element (7.1), the coordinate
transformations (4.4), (4.6), or (4.7), according
0 sends a light signal to P i who reflects it to P2, as the cosmological model is of type I, II', or
who, in turn, sends it back to 0. This arrange- III, lead to the form of the line element in cos-
ment is sketched in Fig. 4. 0 measures the time mological coordinates.
interval Af in which the light ray completes the In a c.c.s. the observer 0 can determine the
circuit OP1P2O. Then the light ray travels the coordinates t, r of an event E by the procedure
distance P1P2 in time outlined above for the coordinates f, p. We have
<if = Af — 2p, t = i(h+h), r = J('2-/i), (7-6)
and 0 can compute the transverse velocity of the notation corresponding to that in Eqs. (7.3).
light pdd/df — Ct. From the line element (7.2) it However, the structure of O's t-clock is different.
follows that From the transformation equations it follows, by
ct=(p/a)/S(p/a). (7.4) putting r = p = 0, that the 2-clock is characterized
by the following equations
S(x) is sin x, sinh *v, X) 111 Cases 1,11', III, re-
spectively, and x / s i n x > l , x / s i n h x < l , x/x—1. Case I : t = tan Jr, (7.61)
Thus we have the following criterion: Case I I ' : t = er, (7.62)
The given cosmological model is of type I,
II', or III, according as ct is greater than, less Case I I I : t = r. (7.63)
than, or equal to 1.
In Cases I and II', the constant a can now be We see that the structure of the /-clock does not
obtained from Eq. (7.4). In Case III, a does not depend on the particular function P , as does the
enter the line element (7.2). r'-clock measuring Robertson's r'-time.
The experiment described above is based on This completes the examination which we
the observation "of transverse effects. These trans- undertook in this section. However, we now have
verse effects also enter such phenomena as the an opportunity to indicate how the trans-
distribution of nebulae in depth and the second- formations leading to a c.c.s., which we merely
order red shift effect, when the estimate of stated in Section 4, can be arrived at by logical
nebular distances is based on the observation of deduction and without guesswork. In Case III,
apparent luminosities. These phenomena may the line element (7.1) is automatically of the
serve as a practical means of determining the cosmological form, and no transformation is
type of the universe and the order of magnitude required. We shall limit ourselves to an examina-
of the constant a. tion of Case II', the treatment of Case I being-
3. As described above, O normalizes his similar.
f-clock so that, initially, its time measure df In the c.c.s. II', the equation of motion of a
coincides with the proper-time measure ds of his fundamental particle P was (in Section 1) shown
atomic clock. After a period of time, however, to be
he finds that his two time measures no longer r = vt. (7.7)
272 L. INFELD AND A. SCHILD

A light signal is, at time th sent by 0 to P who
reflects it, the light signal returning to 0 at time
h. By (7.6) and (7.7), we have
{h-h)/{t2+h)=v.
We assume that P is close to 0, i.e., that v is
small. We denote the short cosmological time
element h—h by dt. The constant (t2—h)/(h-\-h)»
or dt/2t, since OP is small, is proportional to a
constant r-interval dr. Thus
dr = pdt/t,
where /? is a constant. Integrating, we have
r = piog(t/p), t = Pe*">.
The constant of integration has, without loss of
generality, been chosen such that dr~dt when
r = 0. Regraduating the r-clock in accordance
with Eq. (7.8), we obtain a t clock measuring
cosmological time.
By use of the £-cloek the coordinates of an
event E may be obtained from Eq. (7.6).
Similarly, the coordinates r, p of E are given by
Thus
t2=(t+r)1 r2=(r+p),
h=(t-r)t ri = ( r - p ) .
The times t2, h are connected with r2, rh re-
spectively, by (7.8). Thus we obtain the equa-
tions of transformation
t±r=fie^±^1^. (7.9)
(7.71)
If this transformation is applied to the line
element (7.1), it is seen that the resulting line
element is of the cosmological form only if /J= 1.
Thus the constant 13 is determined and (7.9)
now agrees with (4.6).
Similarly, the transformation leading from the
(7.72) coordinates (f, p) to cosmological coordinates
(i, f), with dimensions, is given by
i±f = ae^±f)f<x'1 (i±f)/a = t±r; (7.91)
(7.8)
if we demand that di — df when f = 0. The con-
stant a which appeared in the line element (7.2)
as the negative radius of curvature of the spatial
section f = 0, appears now, in the c.c.s., as the
age of the universe on the cosmological time
scale.
In Case I, parallel considerations show that
the transformation equations are
(«±f)/2a = tan((f±/5)/2a);
(7.92)
0±f)/2a = /±r.
(7.81) The radius of curvature a of the spatial section
f = 0 appears, in the c.c.s., as one-quarter of the
coordinate period of oscillation.