Gener al Relativity and Gravitation, Vol. 31, No.

12, 1999

On the Curv ature of Space²

By A. Friedman in Petersburg *
With one ® gure. Received on 29. June 1922

§1. 1. In their well-known works on general cosmologic al questions,

Einstein 1 and de Sitter 2 arrive at two possible typ es of the universe; Ein-
stein obtains the so-called cylindrical world, in which space 3 has constant,
time-indep endent curvature, where the curvature radius is connected to
the total mass of matter present in space; de Sitter obtains a spherical
world in which not only space, but in a certain sense also the world can
be addressed as a world of constant curvature. 4 In doing so both Einstein
and de Sitter make certain presupp ositions about the matter tensor, which
correspond to the incoherence of matter and its relativ e rest, i.e. the ve-
locity of matter will be supp osed to be su ciently small in comparison to
the fundamental velocity 5 — the velocity of light.

² È
È Physik 10 , 377-386 (1922), with the title ª Uber
Originally published in Zeitschrift fur
die Kr umm
È ung des Raumes ” . Both pap ers are printed with the kind permission of
Springer-V erlag Gmb H & Co. KG, the cu rren t co pyrigh t owner, and tran slated by G.
F. R. Ellis and H . van Elst, Dep artmen t of M athemat ics and Applied M athematics,
Un iversit y of Cap e Town, Rondebo sch 7701, South Africa. Some obvious typ os have
b een corrected in this translation.
1
Einstein, Cosmological co nsiderations relating to the general theory of relativit y,
Sitzungsberichte Berl. Akad. 1917.
2
de Sitter, On Einstein’ s theory of gravitation and its astro nomical consequences.
M onthly Notices of the R. Astronom. Soc. 1916-1917 .
3
By ª space ” we understand here a space that is describ ed by a man ifold of three
dimensions; the ª world ” corresp onds to a man ifold of four dimensions.
4
Klein, On the integral form of the con servation theorems and the theory of the
spatially closed world. G Èotting. Nachr. 1918.
5
See this name by Edd ington given in his book: Espace, Temps et Gravitation , 2
P artie, S. 10. P aris 1921.
1991
0 0 01 - 770 1/9 9 /1 200 - 19 9 1$1 6 .0 0 /0 ° c 1 99 9 P len um Pu blishing C orp oratio n
1992 Friedman

The goal of this Notice is, ® rstly the derivation of the cylindrical and
spherical worlds (as special cases) from some general assumption s, and
secondly the proof of the possibility of a world whose space curvature is
constant with resp ect to three coordinates that serve as space coordinates,
and dependent on the time, i.e. on the fourth — the time coordinate; this
new typ e is, as concerning its other properties, an analogue of the Einstein
cylindrical world.
2. The assumption s on which we base our consideratio ns divide into
two classes. To the ® rst class belong assumptions which coincide with Ein-
stein’ s and de Sitter’ s assumption s; they relate to the equations which the
gravitational potentials obey, and to the state and the motion of matter.
To the second class belong assumption s on the general, so to speak geo-
metrical character of the world; from our hyp othesis follows as a special
case Einstein’ s cylindrical world and also de Sitter’ s spherical world.
The assumptions of the ® rst class are the following:
1. The gravitationa l potentials obey the Einstein equation system
with the cosmological term, which may also be set to zero:

R ik ¡ 1/ 2 gik R + l gik = ¡ k Tik ( i, k = 1, 2, 3, 4), (A )

here gik are the gravitational potentials, Tik the matter tensor, k — a
constant, R = gik R ik ; R ik is determined by the equations

{} { } { }{ }
p p
¶ 2 lg g ¶ lg g ik ¶ ik ia ks
R ik = ¡ ¡ + , (B )
¶ xi ¶ xk ¶ xs s
¶ xs s s a

where x i ( i = 1, 2, 3, 4) are the world coordinates, and

sym bols of the second kind.6
{} ik
l the ChristoŒel

2. Matter is incoherent and at relativ e rest; or, less strongly expressed,

the relativ e velocities of matter are vanishingly small in comparison to the
velocit y of light. As a consequenc e of these assumption s the matter tensor
is given by the equations:

Tik = 0 for i and k not = 4,

2
T44 = c r g44 , (C )

where r is the densit y of matter and c the fundamental velocity; further-

more the world coordinates are divided into three space coordinates x1 ,
x2 , x3 and the time coordinate x4 .

6
Th e sign of R ik and of R is diŒeren t in our case from the usual one.
On the Curv ature of Space 1993

3. The assumption s of the second class are the following:

I. After distributio n of the three space coordinates x 1 , x 2 , x 3 we have
a space of constant curvature, that however may dep end on x4 — the
time coordinate. The interval 7 ds, determined by ds2 = gik dxi dxk , can
be brought into the following form through intro duction of suitable space
coordinates:

ds2 = R 2 ( dx21 + sin 2 x1 dx22 + sin 2 x1 sin 2 x2 dx23 )

+ 2 g14 dx1 dx4 + 2 g24 dx2 dx4 + 2 g34 dx3 dx4 + g44 dx24 .

Here R depends only on x4 ; R is proportiona l to the curvature radius

of space, which therefore may change with time.
2. In the expression for ds2 , g14 , g24 , g34 can be made to vanish
by corresponding choice of the time coordinate, or, shortly said, time is
orthogonal to space. It seems to me that no physical or philosophi cal
reasons can be given for this second assumption; it serves exclusively to
simplify the calculation s. One must also remark that Einstein’ s and de
Sitter’ s world are contained as special cases in our assumptions.
In consequen ce of assumptions 1 and 2, ds2 can be brought into the
form

ds2 = R 2 ( dx21 + sin 2 x1 dx22 + sin 2 x1 sin 2 x2 dx23 ) + M 2 dx24 , (D )

where R is a function of x4 and M in the general case depends on all

four world coordinates. The Einstein universe is obtained if one replaces
2
in (D) R 2 by ¡ Rc 2 and furthermore sets M equal to 1, whereby R means
the constant (indep endent of x 4 ) curvature radius of space. De Sitter’ s
2
universe is obtained if one replaces in (D) R 2 by ¡ Rc 2 and M by cos x1 :

R2
dt 2 = ¡ 2 2 2 2 2 2
( dx1 + sin x1 dx2 + sin x 1 sin x 2 dx3 ) + dx4 ,
2
(D1 )
c2
R2
dt 2 = ¡ ( dx21 + sin 2 x1 dx22 + sin 2 x 1 sin 2 x 2 dx23 ) + cos2 x 1 dx24 . 8
c2
(D2 )

7
See e.g. Eddington, Espace, Temps et Gravitation , 2 P artie. P aris 1921.
8
The ds , which is assumed to have the dimension of time, we den ote by dt ; then the
Length
co nstan t k has the dimension and in CGS-units is equal to 1, 87.10 ± 27 . See
M ass
Lau e, Die Relativitatstheorie
È , Bd . II, S. 185. Braun sc hweig 1921.
1994 Friedman

4. Now we must make an agreement over the limits within which

the world coordinates are con® ned, i.e. over which points in the four-
dimensiona l manifold we will address as being diŒerent; without engaging
in a further justi® cation, we will supp ose that the space coordinates are
con® ned within the following intervals: x 1 in the interv al (0, p); x 2 in the
interv al (0, p) and x 3 in the interval (0, 2p); with resp ect to the time co-
ordinate we make no preliminary restrictiv e assumption, but will consider
this question further below.
§2. 1. It follows from the assumptions (C) and (D), if one sets in
equations (A) i = 1, 2, 3 and k = 4, that:

¶ M ¶ M ¶ M
R 9 ( x4 ) = R 9 ( x4 ) = R 9 ( x4 ) = 0;
¶ x1 ¶ x2 ¶ x3
from this there arise the two cases: (1.) R 9 ( x 4 ) = 0, R is independent of
x4 , we wish to designate this world as a stationary world; (2.) R 9 ( x4 ) not
= 0, M dep ends only on x 4 ; this shall be called the non-statio nary world.
We consider ® rst the stationary world and write down the equations
(A) for i, k = 1, 2, 3 and further i not = k , thus we obtain the following
system of formulae:

¶ 2M ¡ ¶ M
cotg x 1 = 0,
¶ x1 ¶ x2 ¶ x2
¶ 2M ¡ ¶ M
cotg x 1 = 0,
¶ x1 ¶ x3 ¶ x3
¶ 2M ¡ ¶ M
cotg x 2 = 0.
¶ x2 ¶ x3 ¶ x3
Integration of these equations yields the following expression for M :

M = A ( x3 , x4 ) sin x1 sin x2 + B ( x2 , x4 ) sin x1 + C ( x1 , x4 ), (1)

where A , B , C are arbitrary functions of their arguments. If we solve

the equations (A) for R ik and eliminate the unknown density r9 from the
still unused equations, we obtain, if we substitute for M the expression
(1), after somewhat lengthy, but completely elementary calculation s the
following two possibilitie s for M :

M = M 0 = const, (2)

9
The densit y r is in our case an unknown function of the world co ordinates x1 , x 2 , x 3 ,
x4 .
 On the Curv ature of Space 1995

M = ( A 0 x4 + B 0 ) cos x1 , (3)

where M 0 , A 0 , B 0 denote constants.

If M is equal to a constant, the stationary world is the cylindrical
world. Here it is advantageous to operate with the gravitational potentials
of formula ( D 1 ); if we determine the density and the quantit y l, Einstein’ s
well-known result will be obtained:

c2 2 4 p2
l= , r= , M = R,
R2 kR 2
k

where M means the total mass of space.

In the second possible case, when M is given by (3), we arrive by
means of a well-behaved transformation of x4 10 at de Sitter’ s spherical
world, in which M = cos x1 ; with the help of ( D 2 ) we obtain de Sitter’ s
relations:
3 c2
l= , r = 0, M = 0.
2
R
Thus we have the following result: the stationary world is either the
Einstein cylindrical world or the de Sitter spherical world.
2. We now want to consider the non-statio nary world. M is now a
function of x4 ; by suitable choice of x 4 (without harming the generalit y of
the consideration), one can obtain that M = 1; to relate this to our usual
picture, we give ds2 a form that is analogous to ( D 1 ) and ( D 2 ):

R 2 ( x4 )
dt 2 = ¡ ( dx21 + sin 2 x 1 dx22 + sin 2 x 1 sin 2 x 2 dx23 ) + dx24 . (D3 )
c2

Our task is now the determination of R and r from the equations (A).
It is clear that the equations (A) with diŒering indices give nothing; the
equations (A) for i = k = 1, 2, 3 give one relation:

R9 2 2 R R9 9 c2
+ + 2 ¡ l = 0, (4)
R 2
R 2
R

the equation (A) with i = k = 4 gives the relation:

3 R9 2 3 c2
+ ¡ l = k c2 r, (5)
R2 R2

10 p
Th is tran sformation is given by the formula dx 4 = A 0 x4 + B 0 dx4 .
1996 Friedman

with
dR d2 R
R9 = and R9 9 = .
dx4 dx24
Since R 9 is not = 0, the integration of equation (4), when we also
write t for x4 , gives the following equation:

( )
2 l
1 dR A¡ R+ 3 c2 R3
= , (6)
c2 dt R

where A is an arbitrary constant. From this equation we obtain R by

inversion of an elliptic integral, i.e. by solving for R the equation

R
s
1 x
t=
c s a A¡ x+ l
3 c2 x3
dx + B , (7)

in which B and a are constants, and where also care must be taken of
the usual condition s of the change of sign of the square root. From the
equation (5), r can be determined to be

3A
r= ; (8)
k R3

the constant A is expressed in terms of the total mass of space M according

to:
kM
A= . (9)
6 p2
If M is positive, then also A will be positive.
3. We must base the consideration of the non-statio nary world on
equations (6) and (7); here the quantit y l is not determined; we will
assume that it can have arbitrary values. We determine now those values
of the variable x at which the square root of formula (7) can change its
sign. Restricting our considerations to positive curvature radii, it su ces
¥
to consider for x the interv al (0, ) and within this interv al those values
of x that make the quantit y under the square root equal to 0 or . One ¥
value of x for which the square root in (7) is equal to zero, is x = 0; the
remaining values of x , at which the square root in (7) changes its sign,
are given by the positive roots of the equation A ¡ x + 3 lc 2 x 3 = 0. We
denote 3 lc 2 by y and consider in the ( x, y)-plane the family of curves of
third degree:
y x3 ¡ x + A = 0 . (10)
On the Curv ature of Space 1997

A is here the parameter of the family, that varies in the interv al (0, ¥ ).
The curves of the family (s. Fig.) cut the x -axis at the point x = A , y = 0
and have a maxim um at the point
3A 4
x= , y= .
2 27 A 2
From the ® gure it is apparent that for negative l the equation A ¡ x +
l 3
3 c 2 x = 0 has one positive root x0 in the interv al (0, A ). Considering x 0
as a function of l and A :
x0 = H( l, A ) ,
one ® nds that H is an increasing function of l and an increasing function
(
of A . If l is located in the interv al 0, 4 c2
9 A2 ) , the equation has two positive
(
roots x 0 = H( l, A ) and x 90 = q (l, A ), where x0 falls in the interval A, 3 2A )
(
and x 90 in the interval 3 2A , ¥ )
; H( l, A ) is an increasing function both of l
and of A , q (l, A ) a decreasing function of l and A . Finally if l is greater
2
than 49 Ac 2 , the equation has no positive roots.
We now go over to the consideration of formula (7) and precede this
consideration by the following remark: let the curvature radius be equal
 1998 Friedman

to R 0 for t = t 0 ; the sign of the square root in (7) for t = t 0 is positive or

negative according to whether for t = t 0 the curvature radius is increasing
or decreasing; by replacing t by ¡ t if necessary, we can always make the
square root positive, i.e. by choice of the time we can always achieve that
the curvature radius for t = t 0 increases with increasing time.
2
4. We consider ® rst the case l > 49 Ac 2 , i.e. the case when the equation
A ¡ x + 3 lc2 x3 = 0 has no positive roots. The equation (7) can then be
written in the form
R
s
1 x
t¡ t0 =
c s R0 A ¡ x+ l
3 c2 x3
dx , (11)

where according to our remark, the square root is always positive. From
this it follows that R is an increasing function of t ; the positive initial value
R 0 is free of any restriction.
Since the curvature radius cannot be less than zero, so it must, de-
creasing from R 0 with decreasing time t , reach the value zero at the instan t
t 9 . The time of growth of R from 0 to R 0 we want to call the time since
the creation of the world;11 this time t 9 is given by:
R0
s
1 x
t9 =
c s 0 A¡ x+ l
3 c2 x3
dx. (12)

We designate the world considered as a monotonic world of the ® rst

kind.
The time since the creation of the (monoton ic) world (of the ® rst
kind), considered as a function of R 0 , A , l, has the following properties:
1. it grows with growing R 0 ; 2. it decreases, when A increases, i.e. the
mass in space increases; 3. it decreases, when l increases. If A > 23 R 0 ,
then for an arbitrary l the time that has ¯ owed since the creation of the
world is ® nite; if A
2
2
£
3 R 0 , then there can always be found a value of
l = l 1 = 94 Ac 2 such that as l approaches this value, the time since the
creation of the world increases without limit. 2
( )
5. Now l shall lie in the interv al 0, 94 Ac 2 ; then the initial value of
the curvature radius can lie in the intervals: (0, x 0 ), ( x 0 , x 90 ), ( x 09 , ). If ¥
R 0 falls into the interval ( x0 , x90 ), then the square root in formula (7) is
imaginary; a space with this initial curvature is impossible.

11
The time since the creatio n of the world is the time that has ¯ owed from that instan t
when the space was one p oint ( R = 0) until the presen t state ( R = R 0 ); this time
may also b e in® nite.
 On the Curv ature of Space 1999

We will dedicate the next section to the case when R 0 lies in the
interv al (0, x 0 ); here we still consider the third case: R 0 > x 90 or R 0 >
q (l, A ) . By consideratio ns which are analogous to the preceding ones, it
can be shown that R is an increasing function of time, where R can begin
with the value x90 = q (l, A ). The time that has elapsed from the instan t
when R = x90 until the instant which corresponds to R = R 0 , we again call
the time since the creation of the world. Let this be t 9 , then it is
R0
s
1 x
t9 =
c s x 90 A¡ x+ l
3 c2 x3
dx. (13)

This world we call a monotonic world of the second kind.

6. We consider now the case when l falls between the limits ( ¡ , 0). ¥
In this case if R 0 > x 0 = H( l, A ), so the square root in (7) becomes
imaginary, the space with this R 0 is impossible. If R 0 < x0 , then the
considered case is identical with the one that we left aside in the previous
section. We thus suppose that l lies in the interv al ( ¡ ¥ , 4 c2
9 A2 ) and that
R 0 < x0 . Through known considerations12 one can now show that R is a
periodic function of t , with period t p , we call this the world period; t p is
given through the formula:
x0
s
2 x
tp =
c s 0 A¡ x+ l
3 c2 x3
dx . (14)

The curvature radius varies thereby between 0 and x0 . We want to call

this world the periodic world. The period of the periodic world increases,
when we increase l, and tends towards in® nity, when l tends towards the
2
value l 1 = 94 Ac 2 .
For small l the period is represented by the approximation formula

pA
tp = . (15)
c
With resp ect to the periodic world, two viewp oints are possible: if we
regard two events to be coincident if their space coordinates coincide and

12
See e.g. W eierstrass, On a class of real periodic functions. Monatsber. d. K Èonigl.
Akad. d. W issensch. 1866, an d H orn, On the theory of sm all ® nite oscillations.
ZS. f. Math. und Physik 47 , 400, 1902. In our case, the con sideration s of these
au thors m ust be altered ap propriately; the periodicit y in our case can be determin ed
by elemen tary con siderations.
2000 Friedman

the diŒerence of the time coordinates is an integer multiple of the period,

then the curvature radius increases from 0 to x 0 and then decreases to the
value 0; the time of the world’ s existence is ® nite; on the other hand, if
the time varies between ¡ ¥ and + ¥ (i.e. we consider two events to be
coincident only when not only their space coordinates but also their world
coordinates coincide), then we arrive at a true periodicit y of the space
curvature.
7. Our knowledge is completely insu cient to carry out numerical cal-
culations and to decide, which world our universe is; it is possible that the
causality problem and the problem of the centrifugal force will illuminate
these questions. It is left to remark that the ª cosmologica l ” quantit y l
remains undetermin ed in our formulae, since it is an extra constant in the
problem; possibly electrodynamical considerations can lead to its evalua-
tion. If we set l = 0 and M = 5.10 21 solar masses, then the world period
becomes of the order of 10 billion years. But these ® gures can surely only
serve as an illustration for our calculation s.

Petrograd, 29. May 1922.