PHYSICAL REVIEW D VOLUME 3 1, NUMBER 8 15 APRIL 1985

Physical approach to cosmological homogeneity

A. K. Raychaudhuri and Bijan Modak

Physics Department, Presidency College, Calcutta-700073, India
(Received 5 September 1984)
This paper shows that in general relativity space-time will admit three linearly independent space-
like Killing vectors if the cosmic matter is a perfect fluid with a functional relationship between
pressure and density and the spacelike eigenvectors of the shear tensor are parallel transported along
the world lines of the fluid. In the Newtonian case it is proved that for an irrotational motion of a
fluid with pressure and density functionally related spatial uniformity of the matter distribution
leads to the homogeneity of the velocity field.

I. INTRODUCTION G;pup=o,
where v, is the fluid velocity vector and subscript a num-
The study of cosmological models has often been based
bers the eigenvectors.
on the idea of spatial homogeneity which geometrically
We are thus excluding the case of shear-free motion,
means the existence of a three-parameter group of
but that case has been discussed exhaustively by Collins
isometrics with spacelike generators, commonly classified
and ~ a i n w r i ~ h As
t . ~will turn out, our assumptions also
into different types following Bianchi. When one supple-
exclude the cases where vorticity is present.
ments this with the Einstein field equations for a distribu-
In Sec. V, we show that in the case of a system obeying
tion of perfect fluid, one gets a spatial uniformity of phys-
Newtonian mechanics homogeneity of the velocity field
ical variables such as pressure, density, etc., and also the
follows from the assumptions of uniformity of pressure
geodesicity of world lines in the case in which there is nei-
and density if the motion is irrotational.
ther vorticity nor tilt. Here we want to invert the problem
and pose the following questions.
11. SOME CONSEQUENCES OF OUR ASSUMPTIONS
What specification of physical conditions would lead to
geometric homogeneity? In the case in which the eigenvalues of the shear tensor
In an earlier paper, Raychaudhuri and ~ a i t i deduced
' are distinct the eigenvectors will be mutually orthogonal.
geometric homogeneity in a number of cases. In the most In the case of degeneracy we may choose them to be
general case studied by them, the following assumptions orthogonal so that, in particular,
were made:
(a) The matter is a perfect fluid. CZu,=O, (la)
(b) There exists an equation of state p = p (p).
since v p is the timelike eigenvector of the shear tensor.
(c) The motion is geodetic, u', =O. Combining ( l a ) with assumption (c*), we get
(dl For the three-spaces orthogonal to the congruence of
world lines (such spaces were shown to exist because of Gv,=O. (lb)
the first three assumptions) one may have a diagonalized
metric tensor. This means that the shear tensor is also di- Also vad,=O. Thus the scalar product of d, with all the
agonal and the eigenvectors are along the coordinate axes four mutually orthogonal vectors <z,va vanish. Hence
for which the metric tensor is diagonal. v',=o. (2)
(e) The eigenvalues of the shear tensor are constant over
the above three-spaces. Again from the vanishing of the divergence of the
As we shall see, the last assumption is redun- energy-momentum tensor one has
dant as it can be deduced from the others. One may
wonder whether assumptions (c) and (dl may also be with- (P + p ) e = -p,,vP , (3)
drawn or weakened. Indeed, it is easy to see that they are ( p +p)vp=p,v(6;-vvv,) ) (4)
more than sufficient; for, with them, it follows that those
structure constants Cbn, for which a,b,c are all unequal where the expansion scalar O = U ~ ; , . From Eqs. (4) and (2)
vanish, but this is true only for a few Bianchi types. The we get
assumptions have thus left out many homogeneous types.
P,, = ( ~ , ~ v " ) v ,* (5)
In the following we shall prove the existence of spatial
homogeneity under assumptions (a) and (b) and a new as- Thus, if p,,#O, the motion is irrotational. If p,, = O and
sumption which we call (c*). pfO, then, in view of assumption (b) and Eq. (31, 6 = 0 .
(c*) The spacelike eigenvectors of the shear tensor We exclude such dust or nonexpanding cosmologies from
are parallel transported along the world lines, i.e., our present consideration. Thus our assumptions lead to

1807 @ 1985 The American Physical Society

1808 A. K. RAYCHAUDHURI AND BIJAN MODAK 31
-

the exclusion of rotating universes. where To's are functions of x 0 alone and 116's independent
With the velocity vector geodetic and irrotational, we of x O .
may introduce a comoving coordinate system in which the We then have
metric is of the form
d.~~=(dx~)~+g~~dx~dx~, (6)
uiK= 2sa<6<$=
2 saTa 2VaTa
a
i K , (10)
where x 0 lines are the world lines of matter and the in-
dices run from 1 to 3. From Eqs. (5), (3), and assumption
(b) we conclude that p,p,e are functions of x 0 alone. The giK=z
cf,,$f=XT
a a
a 211a11a
i K . (1 1)
Raychaudhuri equation for the present case ( Gl, =w p = 0 )
We shall now proceed to show that the following results
hold:
xaxb-xbxa=C:bxe ,
now shows that the shear scalar o is also a function of x 0
alone; 202=oa80a8, where gap is the shear tensor. c:bc$+ctfc,d,+ c ; ~ c , ~ ~ = o ,
We next show that the eigenvalues of the shear tensor where
capare also functions of x 0 alone. Calling the eigen-
values S 1 , S 2 , S 3(one of the eigenvalues is zero), we have,
because = 0 and 02is a function of x O alone,

Sl + S 2 + S 3 =O , and the C,eb's are functions of x 0 alone. The C,eb's are re-
lated to the Ricci rotation coefficients yabc when we
s12 + ~ 2 2 + ~ 3 2 = f ~ n of
~ t xi 0o alone
n . choose the unit eigenvectors of o i as ~ our orthonormal
In the case of degeneracy, it is obvious that the Si7sare tetrad:
functions of x O . For the nondegenerate case, if Si's are
functions of space coordinates, we have from the above
conditions or, equivalently,
sl,i s2,i s3,i
- - --4i (say).
s2-s3 s3-s1 s1-s2 A straightforward calculation now gives
The condition Sl,iK=S1,Ki gives d i , =$K,i
~ SO that &i
may be written as a gradient We may then integrate
the above to obtain
where the covariant vectors BF are defined by
4 = ~ - l l n ( c r ~ 1 + / 3 S 2 + y S 3 ),
where a,/3,y,K are given by
Thus the B y s also are functions of space coordinates
only, so that the C&'s can be written

where the Q's are independent of xO.

We next consider the field equations. The Raychau-
As complex values of Si are inadmissible, we conclude
dhuri equation which we have already used is simply the
that the space derivative of the Si's vanishes, i.e.,
R : equation. The Rb = O equation gives in the present
case (see, e.g., ~ a ~ c h a u d h u r i In
~ ) view
. of (lo),
(8),and (14a) this yields

111. PROOF OF THE THEOREM

The 6;'s are eigenvectors of gil as well [see Eq. (18) The Sa's and Ta's are, however, not independent. We
below] with eigenvalues dependent on x 0 alone. We now have, using (6),
have from assumption (c*) and Eq. (6)
8; +sg =o , so that using ( 11) and ( 10)
where 6 is a function of x O alone and overdots signify dif-
ferentiation with respect to xO. Hence the unit spacelike
eigenvectors are of the form
66=~a~6 (9) From Eqs. (15), (171, and (19) we get
-
31 PHYSICAL APPROACH TO COSMOLOGICAL HOMOGENEITY 1809

It is now easy to show the existence of the three linearly

independent Killing vectors. Consider the differential
equation
where a,b,c are unequal. Obviously K is a constant. It
thus follows that the shear eigenvalues bear a constant ra-
tio to one another, and on integration of (20) we get
where

where we have, reduced a constant to unity by a normali-

zation of the 7;'s. The other six field equations reduce to As is well known, Eq. (25) is completely integrable as the
7;'s form a group and any three linearly independent
solutions of (25) may be taken as the reciprocal group vec-
tors eo is en hart^). From (26) and the fact that the 77;'s are
where R': are the Ricci tensor components corresponding
orthogonal vectors of constant norm, we get
to the three-space metric. Introducing yabcddefined by
g l m , -gim
~ ril -glj rim =O .
Contracting with hf and using (251,
we have in view of (22)
ax;
h:gglm,~ +gim 7
ax + g ~ 7
j
ax = O ,
where the Oa's are functions of x O alone. Utilizing the which are just Killing equations.
symmetry properties of yabcd (which they inherit from the
Riemann-Christoffel tensor), we finally get the following IV. NECESSITY OF CONDITION ( c* )
conditions on the y's:
It is easy to see that the conditions <z;pvg=O,are neces-
a = 1,2,3, sary in the absence of vorticity or tilt. If the 7~;'s are the
yabba =function of x O alone afb , (24a) Killing vectors, then with the metric (61, 7L;,o=0 and then
b = 1,2,3, it can be shown (cf. ise en hart^) that the reciprocal group
vectors 17; satisfy the equation
yabcU=yacba (a,b,c all unequal) . (24b)
Writing out explicitly the condition (24a) and utilizing
(21), we get the equation of the form where uiKis the shear tensor. (In fact, the relation holds
for any tensor where components are merely functions of
2 aprP=function of xOalone , giK3sand their derivatives.)
P Also, as the differential equation satisfied by the 77;'s is
where r = T b / T a , and the up's are constructed out of the of the form
Q k ' s and hence are independent of x O . Consideration of
the three such equations that we obtain by using the three
pairs of values a,b leads us to the result that QL's are
constant and hence Ctc3sare functions of x O alone. The where the T's are independent of x O , the unit spacelike
relation (24b) yields for a = 1, using the constancy of eigenvectors of the sheaf tensor may be written in the
Q&'s just obtained, form Taq;, where the 7;'s [the solution of Eq. (2911 are
independent of x O and the To's are functions of x Oonly.
But this is of the form (9) to which we have been led from
assumption (c*) and the metric (6).

Using Eq. (141, after simplification this can be reduced to V. THE CASE OF NEWTONIAN COSMOLOGY

We shall in this section consider the case of Newtonian

cosmology which, in effect, means that we shall use
Euler's hydrodynamical equations and Poisson's equations
which is one of the Jacobi identities (13). The other Jaco-
for gravitation. We shall assume that pressure and densi-
bi identities are obtained by taking a = 2 and 3 in (24b).
ty are functions of time alone so that they are necessarily
If now we define the structure constants with respect to
functionally related. In this case there is no question of
the vectors 77; instead of <;, then, in view of (161, the new
geometric homogeneity, but what we shall show is that
C&'s will be simply Q&'s and thus be constants and satis-
the velocity field ui may be written
fy the Jacobi relations. Thus 7;'s are the generators of a
three-parameter group.
1810 A. K. RAYCHAUDHURI AND BIJAN MODAK 31
-

where aiK's are functions of time only. Another assump- where k is a function of t alone. Our problem is to satisfy
tion that we utilize is that vorticity vanishes, i.e., (35) and (37) simultaneously. Using spherical polar coor-
Vxv=O. dinates we write the solution of (35) in terms of spherical
One obtains quite generally, under the above assump- harmonics Ylm( 8,p 1,
tions,

The summation is restricted to positive values of 2 as oth-

where an overdot indicates the operator erwise there would be a singularity in the velocity field at
the origin. the Aim's are functions of time alone.
A straightforward calculation using the properties of
Ylm now yields

and oiKis the shear tensor ( ~ a ~ c h a u d h u r iR~ )is. defined

by the relation
(39)
where x =cos0. Also from (37)
From the equation of continuity and the assumed homo- 1 VlL I ' = k 2 y 2 + Z Z~lm~'~lm(0,~)
geneity in the density distribution, 0 and R are functions I m
of t alone and hence, from (31), a2 is also a function of
time alone. These, expressed in terms of the velocity po- which gives
tential 4, read J J ( ~ q b ) ~ dd. pw= 4 . r ; k 2 y 2 + ~ o.
~ ~ # = V . v = 0 = f u n c t i o nof t alone . (33) Comparing (39) and (401, we find
Also,
U,,KU,,K =#,rKd,lK=function of t alone . (34)
From (34)
v2*=0, (35) (a constant term in is of no significance), where the biK
are functions of time alone which leads to the desired re-
where sult (30).
*=+-8y2/6 :
VI. CONCLUDING REMARKS
so that
The problem that naturally comes to mind is the cases
where pressure vanishes [when Eq. (4) is trivially satis-
fied], when we cannot conclude the spatial uniformity of
and p and 8, or when vorticity is present. We propose to con-
sider these cases in a later paper.
ui,Kui,K=$,iK$,iK +82/3 9

so that *,;K*,iK is a function of t alone. Using (35) we ACKNOWLEDGMENT

then have One of the authors (B.M.) is indebted to the Council of
Scientific and Industrial Research (India) for financial as-
= 6 k, ~
~ ~ ( $ ~ l L ~ )(say) (37) sistance.

'A. K. Raychaudhuri and S. R. Maiti, Phys. Rev. D 18, 3595 ty Press, London, 1979).
(1978). 4L. P. Eisenhart, Riema~znianGeometry (Princeton University
2C. B. Collins and J. Wainwright, Phys. Rev. D 27, 1209 (1983). Press, Princeton, New Jersey, 1949).
3 ~ K.
. Raychaudhuri, Theoretical Cosmology (Oxford Universi-