a

r
X
i
v
:
h
e
p
-
t
h
/
9
9
0
5
0
1
2
v
2


2
5

J
u
n

1
9
9
9
LPT-ORSAY 99/25
hep-th/9905012
Non-conventional cosmology from a
brane-universe
Pierre Binetruy
1
, Cedric Deayet
1
, David Langlois
2
1
LPT
1
, Universite Paris-XI, Batiment 210, F-91405 Orsay Cedex, France;
2
Departement dAstrophysique Relativiste et de Cosmologie, C.N.R.S.,
Observatoire de Paris, 92195, Meudon, France.
Abstract.
We consider brane-universes, where matter is conned to four-dimensional
hypersurfaces (three-branes) whereas one extra compact dimension is felt by grav-
ity only. We show that the cosmology of such branes is denitely dierent from
standard cosmology and identify the reasons behind this dierence. We give a
new class of exact solutions with a constant ve-dimensional radius and cosmo-
logically evolving brane. We discuss various consequences.
1 Introduction
There has been recently renewed activity in the domain of cosmology with ex-
tra dimensions. Historically, investigations on the possibility of extra dimensions
began with Kaluza-Klein type theories (see e.g. [1] for a review) and were then
revived with the advent of high energy physics theories, such as string and super-
string theories, where the existence of additional dimensions seemed necessary
to ensure a non-anomalous quantum behaviour. However, these extra dimen-
sions, since they remained undetected in experiments, were constrained to be
very small.
1
Unite mixte de recherche UMR n
o
8627.
1
Recently, it has been suggested [2] that additional dimensions could have a
very distinct nature from our familiar dimensions, in the sense that ordinary
matter would be conned to our four-dimensional universe while gravity would
live in the whole extended spacetime. In other words, our four-dimensional
universe would be a three-brane living in a (4+n)-dimensional spacetime (n being
the number of additional dimensions). One of the advantages of this picture would
be to relax the strong constraints on the size of extra dimensions and to allow for
large additional dimensions, with the associated property of a low fundamental
Planck scale, leading to many phenomenological consequences.
A strong motivation for considering such models comes from strongly coupled
string theories [3],[4]. For example, the strongly coupled E
8
E
8
heterotic string
theory is believed to be an 11-dimensional theory, the eld theory limit of which
was described by Horava and Witten [3]. The spacetime manifold includes a
compact dimension with an orbifold structure. Matter is conned on the two 10-
dimensional hypersurfaces (9-branes) which can be seen as forming the boundaries
of this spacetime.
For all such theories, an essential issue concerns the cosmological evolution of
our universe. Depending on the mass scales associated, this might indeed be the
only way to test such models. Several recent works have examined this question
[5] [6], with a special emphasis on ination [7]-[14]. The purpose of the present
work is to reconsider the cosmological evolution of a brane-like universe. Our
main result is that, contrarily to what has been implicitly assumed in previous
works, brane cosmology leads to Friedmann-like equations very dierent from the
standard ones.
Since this area of research lies at the interface between high energy physics
and general relativity, we have tried, as much as possible, to present our results
from both viewpoints. We have restricted our analysis to a single additional
dimension, leaving the more general case of several extra dimensions for future
work, but we expect that our main result, i.e. the non conventional nature of the
cosmological dynamics for a brane-universe, will still hold qualitatively. The plan
of this paper is the following. In section 2, we present the general framework for
our ve-dimensional eective theory. In section 3, we compute the corresponding
ve-dimensional Einstein equations and study the conditions imposed locally by
the presence of a brane. In section 4, we adopt a geometrical point of view and
rewrite the Einstein equations in the Gauss-Codacci form. In section 5, we adopt
a global point of view and show that this imposes some relations between the
energy-momentum of matter in our brane-universe and that of matter in the rest
of the ve-dimensional spacetime. We identify a class of exact solutions which
show explicitly the non-conventional dynamics discussed in this paper; this allows
us to discuss the behaviour of the zero modes in a Kaluza-Klein type approach.
Finally, section 6 discusses the consequences of our unconventional cosmological
equations.
2
2 Brane theory in ve-dimensional spacetime
The cosmological consequences of having n extra compact dimensions unseen by
ordinary matter depend on the number n of such dimensions [8, 10]. In this
article, we will restrict our attention to the case of a single such dimension. One
motivation is that this corresponds to the case studied by Horava and Witten
[3] and this explicit example may provide some insight and some explicit tests
of the properties discussed on more general terms. We will actually consider for
illustrative purposes somewhat simpler theories, obtained by compactication to
ve dimensions. Our four-dimensional Universe will be assimilated to a four-
dimensional hypersurface (three-brane).
The theory we are thus considering is ve-dimensional and has an action of
the form
S
(5)
=
1
2
2
(5)
_
d
5
x
_
g

R +
_
d
5
x
_
gL
m
, (1)
where the rst term corresponds to the Einstein-Hilbert action in ve dimensions
for a ve-dimensional metric g
AB
,

R being the ve-dimensional scalar curva-
ture for this metric, and the second term corresponds to the matter content.
Our notational conventions will be as follows: upper case Latin letters denote
ve-dimensional indices (0, 1, 2, 3, 5); lower case Greek indices run over the four
conventional dimensions (0, 1, 2, 3) whereas lower case Latin indices run over or-
dinary spatial dimensions (1, 2, 3). To avoid confusion, some ve-dimensional
elds (like g
AB
) will be tilted in contrast with four-dimensional elds. The sig-
nature of g
AB
is (++++).
A
will denote the covariant derivative associated
with the metric g
AB
. The constant
(5)
in the action (1) is directly related to the
ve-dimensional Newtons constant G
(5)
and the ve-dimensional reduced Planck
mass M
(5)
, by the relations

2
(5)
= 8G
(5)
= M
3
(5)
. (2)
We shall consider ve-dimensional spacetime metrics of the form
ds
2
= g
AB
dx
A
dx
B
= g

dx

dx

+ b
2
dy
2
(3)
where y is the coordinate of the fth dimension which we assume to be compact;
its range interval I is chosen to be 1/2 y 1/2, with the two endpoints of
the interval identied.
Inspired by the compactication of Horava-Witten theory, we will sometimes
impose a symmetry y y. In the specic model of Horava-Witten [3], points
obtained from one another by this symmetry are identied. The relevant interval
is then 0 y 1/2 and there are two three-branes which form the boundary of
spacetime, respectively at y = 0 and y = 1/2.
Throughout this article, we will focus our attention on the hypersurface de-
ned by y = 0, which we identify with the world volume of the brane that forms
3
our universe. Since we are interested in cosmological solutions, we take a metric
of the form
ds
2
= n
2
(, y)d
2
+ a
2
(, y)
ij
dx
i
dx
j
+ b
2
(, y)dy
2
, (4)
where we have assumed for simplicity a at metric in the ordinary spatial dimen-
sions. Let us turn to the matter part of the theory. We shall distinguish two
kinds of sources: elds conned to the world-volume of our brane-universe and
elds living in the bulk spacetime. The energy-momentum tensor, derived from
the matter part of the action (1), will thus be decomposed into

T
A
B
=

T
A
B
|
bulk
+ T
A
B
|
brane
, (5)
where

T
A
B
|
bulk
is the energy momentum tensor of the bulk matter (and possi-
bly other branes), which we do not need to specify here, and the second term
T
A
B
|
brane
corresponds to the matter content in the brane (y = 0). The latter can
be expressed quite generally in the form
T
A
B
|
brane
=
(y)
b
diag (, p, p, p, 0) , (6)
where the energy density and pressure p are independent of the position in
the brane in order to recover a homogeneous cosmology in the brane. Here we
have assumed an idealized situation where the physical brane is assumed to be
innitely thin. Rigorously, the physical brane will have a width in the fth
dimension which is usually assumed to be given by the fundamental scale of the
underlying theory. In our case, the thin-brane approximation will thus be valid
when the energy scales at which we consider the theory are much smaller than the
fundamental scale (which is anyway required if one describes gravity classically).
Let us emphasize at this stage that the standard way to probe the metric
of the universe on cosmological scales is through the kinematics of the PPSM
(Particle Physics Standard Model) elds. If one assumes that the PPSM elds
live on the brane, they will be sensitive only to the induced metric (given here
by g

(, 0)). For example, in the case of redshift measurements, the redshift

of a given distant galaxy is
0
/
1
1 = (a(
0
, 0)/a(
1
, 0)) 1 (where
1
and
0
are respectively the time of emission and the time of reception). Similarly, the
equation of conservation for the energy-momentum in the brane, which will be
given below in (12), depends only on the metric in y = 0. The four-dimensional
eective cosmology for our world is thus given by the equations relating the
induced metric g

(, 0) to the sources. The determination of the dynamics of

this induced metric will be the object of the next two sections, whose essential
feature will be to solve locally the Einstein equations in the vicinity of the brane.
4
3 Eective brane cosmology from Einstein equa-
tions
The dynamics of the ve-dimensional geometry, thus including the brane geome-
try, is governed by the ve-dimensional Einstein equations, which can be derived
from the action (1) and take the usual form

G
AB
=
2
(5)

T
AB
. (7)
Inserting the ansatz (4) for the metric, the non-vanishing components of the
Einstein tensor

G
AB
are found to be

G
00
= 3
_
a
a
_
a
a
+

b
b
_

n
2
b
2
_
a

a
+
a

a
_
a

a

b

b
___
, (8)

G
ij
=
a
2
b
2

ij
_
a

a
_
a

a
+ 2
n

n
_

b
_
n

n
+ 2
a

a
_
+ 2
a

a
+
n

n
_
+
a
2
n
2

ij
_
a
a
_

a
a
+ 2
n
n
_
2
a
a
+

b
b
_
2
a
a
+
n
n
_

b
b
_
, (9)

G
05
= 3
_
n

n
a
a
+
a

b
b

a

a
_
, (10)

G
55
= 3
_
a

a
_
a

a
+
n

n
_

b
2
n
2
_
a
a
_
a
a

n
n
_
+
a
a
_
_
. (11)
In the above expressions, a prime stands for a derivative with respect to y, and
a dot for a derivative with respect to .
From the Bianchi identity
A

G
A
B
= 0 we can obtain, using the Einstein
equations (7), an equation of conservation of the form
+ 3(p + )
a
0
a
0
= 0, (12)
which is the usual four-dimensional equation of conservation for the energy den-
sity in cosmology (a
0
is the value of a on the brane). It will be shown in the next
section how one recovers more generally the conservation of the stress-energy-
momentum tensor of the brane.
We now seek a solution of Einsteins equations (7) in the vicinity of y = 0.
In order to have a well dened geometry, the metric is required to be continuous
across the brane localized in y = 0. However its derivatives with respect to y can
be discontinuous in y = 0. This will entail the existence of a Dirac delta function
in the second derivatives of the metric with respect to y. The resulting terms
with a delta function appearing in the Einstein tensor (in its components (8) and
(9)) must be matched with the distributional components from the stress-energy
tensor in order to satisfy Einsteins equations. For a, one has for example
a

=

a

+ [a

](y), (13)
5
where

a

is the non-distributional part of the double derivative of a (the standard

derivative) , and [a

] is the jump in the rst derivative across y = 0, dened by

[a

] = a

(0
+
) a

(0

). (14)
For future use we also dene the mean value of a given function f across y = 0
by
f =
f(0
+
) + f(0

)
2
. (15)
Matching the Dirac delta functions in the components (8) and (9) of the Einstein
tensor with those of the brane stress-energy tensor (6), one obtains the following
relations,
[a

]
a
0
b
0
=

(5)
2
3
, (16)
[n

]
n
0
b
0
=

(5)
2
3
(3p + 2) , (17)
where the subscript 0 for a, b, n means that these functions are taken in y = 0.
If we take the jump of the component (0, 5) of Einstein equations, and use the
above relations (16) and (17), we recover the energy conservation equation (12).
Let us now take the jump of the component (5, 5) of Einsteins equations. Using
(16-17), we obtain
a

a
0
p =
1
3

n
0
, (18)
whereas the mean value of the same equation gives
1
n
2
0
_
a
0
2
a
2
0

a
0
a
0
n
0
n
0
+
a
0
a
0
_
=

(5)
4
36
( + 3p) +
(a

)
2
a
2
0
b
2
0
_
1 +
3p

(5)
2

T
55
3b
2
0
, (19)
where we recall that

T
AB
is the bulk energy-momentum tensor.
If we restrict our analysis to solutions that are left invariant by the symmetry
y y (as in Horava-Witten supergravity), then things simplify considerably
since a is now a function of |y|, and a

= 0. The above equation then simplies

to the form
a
0
2
a
2
0
+
a
0
a
0
=

(5)
4
36
( + 3p)

(5)
2

T
55
3b
2
0
, (20)
where we have chosen the time , without any loss of generality, so that n
0
= 1
(this corresponds to the usual cosmic time in standard cosmology). This equation
is one of the main results of the present work and must be interpreted as a new
Friedmann-type equation. It must be emphasized that this equation (as well as
(19)) is fundamentally dierent from the standard cosmological Friedmann equa-
tion in the sense that the square of the Hubble parameter H depends quadratically
6
on the cosmological (i.e. the brane) energy density, in contrast with the usual
linear dependence.
2
Let us note from (20) that the universe dynamics will be dominated by the
brane energy density rather than the bulk pressure (assumed to be of the order
of the bulk energy density) when the condition

bulk


2
brane
M
3
(5)
(21)
is satised.
Assuming that (21) is veried, one important consequence is that, for the
same matter content in our four-dimensional universe, the scale factor evolution
will be dierent from the standard one. Indeed, let us consider an equation of
state of the form p = w, with w constant, which includes the main eective
equations of state used in cosmology, namely a radiation gas (with w = 1/3), a
non-relativistic gas (with w 0) and also a cosmological constant (with w = 1).
The energy conservation law (12), unchanged with respect to the standard case,
yields the usual relation between the energy density and the scale factor,
a
3(1+w)
0
. (22)
However, when one now looks for power-law solutions for the evolution of the
scale factor, i.e. such that
a
0
(t) t
q
, (23)
then the new Friedmann-type equation (20), neglecting the bulk term, leads to
q =
1
3(1 +w)
, (24)
which must be contrasted with the standard cosmological solutions for which
q
standard
= 2/(3(1 + w)) (the particular case w = 1 corresponding to a cosmo-
logical constant will be addressed below). The new cosmological equation thus
leads typically to slower evolution.
Another essential feature of the equation (19) is that the eective four-dimensional
Newtons constant does not appear in it. The dynamics of the universe, from this
equation, depends only on the fundamental ve-dimensional Newtons constant.
This result seems in contradiction with most of the recent litterature on brane
cosmology (see in particular [10] [12]), where the standard Friedmann equations
are recovered as an eective four-dimensional description. There exist, however,
some hints of unconventional cosmology (see [8] and the non-linear regime of
[10]) but they are presented as valid only for high energies, whereas our results
2
We will show later that the same quadratic dependence would hold, to leading order, if we
were to dene the Hubble parameter through the zero modes of the metric, in a Kaluza-Klein-
like way.
7
indicate that they are the rule as far as the ve-dimensional description makes
sense.
In what follows, we will try to get a better understanding of the above non
standard Friedmann equations, from dierent viewpoints including the derivation
of a new class of exact solutions.
4 Gauss-Codacci equations
The purpose of this section is to give a geometrical analysis of a three-brane living
in a ve-dimensional spacetime. It will be useful to resort to a well-known tech-
nique in general relativity, which consists in decomposing the Einstein equations
into a part tangential to the surface under consideration, a part normal to it and
nally a mixed part (since Einstein equations are tensorial of order two). This
leads in particular to the so-called Gauss-Codacci equations. This technique was
used to study thin shells in general relativity [15] and is particular t to study
the very analogous problem of branes.
Let us rst introduce the unit vector eld n
A
normal to the three-brane world-
sheet. Note that this vector eld is spacelike, i.e.
g
AB
n
A
n
B
= 1. (25)
The induced metric on the brane worldsheet will thus be dened by
h
AB
= g
AB
n
A
n
B
. (26)
A very useful quantity is the extrinsic curvature tensor K
AB
(also called second
fundamental form), which is dened by the expression
K
AB
= h
C
A

C
n
B
, (27)
and can be interpreted as representing the bending of the brane worlsheet in
the ve-dimensional spacetime. As will be shown below, the non-conventional
cosmological equations which we shall nd are a direct consequence of the fact
that K
AB
is non-vanishing in the presence of matter conned to the brane, which
means that the ve-dimensional metric depends necessarily on the fth dimen-
sion, in contrast with the usual assumption in the Kaluza-Klein approach.
The ve-dimensional Einstein tensor

G
AB
can then be decomposed into the
following projections,

G
AB
n
A
n
B
=
1
2
(4)
R +
1
2
_
K
2
K
AB
K
AB
_
, (28)
n
A

G
AB
h
B
C
= D
A
K
A
C
D
C
K, (29)

G
AB
h
A
C
h
B
D
=
(4)
G
CD
KK
CD
n
E

E
K
CD
+ D
C
a
D
2n
(C
K
D)E
a
E
a
C
a
D
8
+
_
1
2
K
2
+
1
2
K
AB
K
AB
+ n
A

A
K
B
a
B
_
h
CD
, (30)
where a
B
= n
C

C
n
B
is the acceleration vector eld, D
A
is the covariant
derivative associated with the induced metric h
AB
, K = g
AB
K
AB
is the trace
of the extrinsic curvature tensor, and the parenthesis around two indices denote
symmetrisation with weight 1/2. The above equations directly follow from the
Gauss-Codacci relations (see e.g. [16]). Of course, these equations are simply a
rewriting of the equivalent relations (8)-(11), although more general in the sense
that we have not specied any ansatz for the metric. The components (8)-(11)
can be recovered directly from the above relations, by noting that, in the case of
a metric of the form (4), one nds n
A
= {0, 0, 0, 0, 1/b}, a
B
= {

b/b, 0, 0, 0, 0}
and K
A
B
= diag{n

/(nb),
i
j
a

/(ab), 0}.
Let us now examine the above equations from a distributional point of view.
The stress-energy tensor (6) associated with the brane can be rewritten in the
form
T
A
B
|
brane
= S
A
B
(y)
b
. (31)
The bulk stress-energy tensor is regular in the vicinity of y = 0 and is y-
dependent. The metric is assumed, as usual, to be continuous. Thus, only second
derivatives of the metric with respect to the normal coordinate y can contain a
Dirac distribution, whereas the extrinsic curvature can at most contain a Heavy-
side distribution. It is clear from the Einstein tensor decomposition (28-30) given
above that only the term n
E

E
K

n
A

A
K =
y
(K

) /b contains
second derivatives with respect to y. This implies the following equation
[K

] =
2
(5)
S

, (32)
where the brackets here denote the discontinuity between the two sides of the
brane, as dened in (14). The above relation is known to general relativists as
Israels junction condition [15]. It can be rewritten in the form
[K

] =
2
(5)
_
S

1
3
Sg

_
, (33)
where S S

is the trace of S

. This equation corresponds to the equations

(16)-(17) of the previous section.
If one assumes in addition a symmetry y y for the metric then the right
and left limits of the extrinsic curvature tensor are necessarily opposite, i.e.
[K

] = 2K
+

2

K

. (34)
Then (33) means that the extrinsic curvature of the brane is completely deter-
mined by its matter content. Note also that, except for the term with second
derivatives, all the other terms in

G
5
5
and

G

involving the extrinsic curvature

9
are quadratic in K

: this is compatible with having the same bulk matter on the

two sides of the brane if one has the symmetry K
+

= K

, which is the case

here.
Inserting the explicit expression of

K

into (28) leads to the following ex-

pression for the scalar four-dimensional curvature:
(4)
R =

4
(5)
4
_
1
3
S
2
S

_
2
2
(5)

T
5
5
. (35)
We thus recover the analogous of equation (20).
Let us now consider the mixed part of Einsteins equations, i.e.

G
5

, corre-
sponding to Eq. (29). Around the brane, this equation implies
D

K = 0, (36)
which, rewriting

K

in terms of the stress-energy tensor S

, according to (33),
yields simply the usual four-dimensional stress-energy conservation law, i.e.
D

= 0. (37)
To conclude this section, let us stress that the derivation of equation (35) is
more general than in the previous section because we did not need to assume
homogeneity and isotropy in the brane. Note also that this approach is local
and nothing was said or needed about the global nature of the ve-dimensional
spacetime.
5 Global solutions
In the previous sections, we have focused our attention to the behaviour of the
geometry on our brane-universe and in its neighbourhood. It is extremely in-
teresting, although dicult, to obtain some exact solutions of the global ve-
dimensional spacetime. We have managed to nd a class of global solutions with
a constant ve-dimensional radius and valid for a general equation of state for
the matter in our brane-universe. Before presenting them we discuss the general
properties of such solutions. We then discuss the implications in a Kaluza-Klein
perspective.
5.1 General properties of the solutions
An important constraint comes from the fact that the solutions to Einsteins
equations must be seen as dened globally throughout spacetime, i.e. for 1/2
y 1/2. This implies in particular that, from the global point of view that
we now adopt, the second derivative of a necessarily satises an equation more
complicated than (13).
10
Indeed there is no Green function G(y) on a circle, solution of G

= (y)
3
. On
the other hand, one can solve G

= (y) (y y
0
) or G

= (y) 1. This means

that, in order to make a metric solution of the Einstein equations globally dened,
one must introduce another brane (or matter in the bulk
4
). For simplicity, we
will put this other brane at y = 1/2. We thus consider an additional term in the
stress-energy tensor of the form
T
A
B
|
brane
= b
1
(y 1/2) diag (

, p

, p

, p

, 0) , (38)
where a star will be used to denote quantities associated with the second brane.
Keeping only the distributional part for the metric, the second derivative of
a then satises the following dierential equation:
a

= [a

]
0
((y) (y 1/2)) +
_
[a

]
0
+ [a

]
1/2
_
((y 1/2) 1) , (39)
where [a

]
0
, resp. [a

]
1/2
, is the jump of a

on the rst, resp. second, brane given

as in (16) by
[a

]
0
a
0
b
0
=

(5)
2
3
, (40)
[a

]
1/2
a
1/2
b
1/2
=

(5)
2
3

. (41)
Integrating (39) over y yields the following solution for a:
a = a
0
+
_
1
2
|y|
1
2
y
2
_
[a

]
0

1
2
y
2
[a

]
1/2
, (42)
where we recognize the quadratic behaviour found by Lukas , Ovrut and Waldram
[10]. A similar expression is obtained for n. Allowing a linear dependence in y
for the function b, we write:
b = b
0
+ 2|y|(b
1/2
b
0
), (43)
where b
0
is assumed to be constant in time. Then, plugging these functions in
the (0,0) Einstein equation on the brane at y = 0, one obtains from (8):
a
2
0
a
2
0
=
n
2
0
b
2
0
_

[a

]
1/2
a
0

b
1/2
b
0
[a

]
0
a
0
+
1
4
[a

]
2
0
a
2
0
_
. (44)
3
As is well known from electromagnetism, one cannot put a non zero charge alone in a
compact manifold with no boundary.
4
This would correspond to the presence of sources distributed over the entire bulk with an
energy-momentum tensor of the same order as that of the brane. The fundamental three-form
of eleven-dimensional supergravity would provide interesting candidates for such sources: one
could mention the Ramond-Ramond scalar. Studying this possibility goes beyond the scope of
this paper.
11
We have seen in section 3 that the (5,5) component of Einsteins equations im-
poses a
0
/a
0
to be of the order of the energy density . Thus the leading terms in
(44) must cancel, which leads to the equation:
[a

]
0
b
0
=
[a

]
1/2
b
1/2
, (45)
and similarly for n, using (9). In other words, using (40) and (41), we must
impose:
a
0
=

a
1/2
, (46)
(2 + 3p)n
0
= (2

+ 3p

)n
1/2
. (47)
Hence the matter on one brane is constrained by the matter on the other. This
is related to the fact that most of the information (even some global one) on
the solutions of the Einstein equations may be obtained locally, as shown in sec-
tion 3. The constraints between the two branes obtained above can probably be
seen as a particular example of topological constraints, which impose specic
restrictions on the distribution of localized matter in a space that contains com-
pact dimensions and which can be found in many dierent contexts (D-branes,
orientifolds or topological defects).
It turns out that, if we assume b
1/2
to be constant in time, in addition to b
0
,
the metric given by (42), the similar expression for n and (43), yields an exact
solution of Einsteins equations (the time evolution for a
0
being given by (44),
and that of by the energy conservation equation on the brane). However, it can
be shown that this class of solutions is simply a rewriting (up to a change of the
coordinate y) of the class of exact solutions, satisfying the condition b
0
= b
1/2
,
exhibited explicitly below.
5.2 Exact cosmological brane solution
Let us look for the simplest solutions one can envisage, namely solutions linear
in |y| for a and n, of the form
a = a
0
(t) (1 +|y|), (48)
n = n
0
(t) (1 +|y|), (49)
b = b
0
, (50)
where b
0
is assumed to be constant in time. and are in general functions of
time and depend directly on the matter content of the brane according to the
equations (16) and (17),
=

2
(5)
6
b
0
, =

2
(5)
2
_
w +
2
3
_
b
0
. (51)
12
We have introduced w p/ which contrarily to section 3, is not assumed here
to be constant. Note that the metric is well dened for
2
(5)
b
0
< 1.
To specify completely our solution, we need to determine the functions a
0
(t)
and (t) (note that n
0
(t) is a totally arbitrary function whose choice represents
simply a choice of time coordinate). This can be done by substituting the metric
ansatz (48-51) in the Einstein equations. The component

G
00
=
2
(5)

T
00
taken on
the brane, then yields the equation
a
2
0
a
2
0
=

4
(5)
36

2
. (52)
This equation, in addition to the energy conservation equation on the brane,
+ 3
a
0
a
0
(1 +w) = 0, (53)
completely determines, once an equation of state has been specied, the unknown
functions a
0
(t) and (t).
In the particular case of a constant w, the energy density conservation on the
brane implies a
3(1+w)
0
and (52) can then be integrated explicitly so that
a
0
(t) t
q
,
2
(5)
=
6q
t
, q =
1
3(1 +w)
(w = 1), (54)
or for a cosmological constant on the brane,
a
0
(t) e
Ht
, H =

2
(5)
6
(w = 1). (55)
We have kept only the expanding solutions of (52).
In the general case, it remains to check that the metric so obtained (48-51)
satises all Einstein equations everywhere. This can indeed be done explicitly
using (52) and (53).
An interesting consequence of our solution is the behaviour of the second
brane. As in the previous section, the matter content of the second brane is
totally determined by the topological constraints due to the compactness of
the additional dimension. Since our solution is a particular case of the general
form (42), the same relations (46) and (47) apply, which implies

=
_
1

2
(5)
12
b
0

_
1
, (56)
and
w

+
2
3
=
_
w +
2
3
_
1

2
(5)
12
b
0

1 +

2
(5)
4
b
0
(w + 2/3)
, (57)
13
where w

. In the general case, w and w

will be time-dependent. Even if

w is chosen to be constant, w

will be time-dependent, except in two particular

cases: w = w

= 1 corresponding to a cosmological constant on both branes

(although of opposite signs), and w = w

= 2/3. Note that the role of the

two branes is perfectly symmetric, as it is clear from (46) and (47). It is not
manifest on the two equations above, (56) and (57), because we have privileged
the position y = 0 in the form (48-50) given to the metric.
It can be checked explicitly that the energy conservation on the second brane,
i.e.

+ 3
a
1/2
a
1/2
(1 +w

= 0, (58)
is automatically satised upon using the rst brane energy conservation (53) and
the two relations (56) and (57).
The present class of solutions could be considered as an illustration of a holo-
graphic principle, by which all the properties of the bulk (here the metric) can
be determined solely by the specication of the elds on the boundary (here the
metric and the matter in our brane-universe).
Finally, it is interesting to note that this new class of exact solutions includes
as a particular case the solution given in [8]. The latter solution corresponds
to matter in the brane with the equation of state p = and represents an
extension of an earlier domain wall inationary solution (in a four-dimensional
spacetime) [18].
5.3 Kaluza-Klein approach
The geometrical picture of section 4 showed us that the y-dependence of the ve-
dimensional metric, embodied in the extrinsic curvature tensor, is unavoidable
because of the very presence of our brane-universe. The purpose of this section
is to show how this picture can be reconciled with the seemingly contradictory
Kaluza-Klein approach, for which, at an energy scale lower than the inverse radius
of the fth dimension, the so-called massive Kaluza-Klein modes (or in a more
geometrical language, the y-dependence of the elds) can be ignored.
The traditional Kaluza-Klein approach, whose purpose is to give a four-
dimensional interpretation of the ve-dimensional world, is supposed to work
when the energy scale of the system, dened more specically as the typical in-
verse length scale or time scale of variation of the elds, is much smaller than the
inverse of the radius of the fth dimension, dened by
R
(5)
() =
_
1/2
1/2
b(, y)dy. (59)
Since we are interested by cosmology, only time variations of the elds will be
considered from a four-dimensional point of view. And the corresponding (in-
verse) time scale is given by the Hubble parameter H(, y) = a/(an), so that the
14
limit of validity of a Kaluza-Klein reduction is
HR
(5)
1. (60)
For energy scales of the order or above R
1
(5)
, the fth dimension is expected to
open up.
The condition HR
(5)
1 can also be reexpressed in terms of using (20), in
which H is typically of the order of
2
(5)
. The low energy limit should thus be
valid when
2
(5)
R
1
(5)
.
5
In the traditional way of thinking, it is expected that, when HR
(5)
0, a
given solution of the equations of motion will converge towards its average value
(or zero-mode) over the compact space, the average value for any quantity Q
being dened by
Q =
_ 1
2

1
2
dy Q(y). (61)
This would mean in particular for the metric that g

(y, ) g

() for any y
and , and especially for y = 0 (on the brane)
6
. If this limit is true, then it is
equivalent to obtain the cosmological evolution from the equations of motion ver-
ied by the zero modes or from the equations of motion satised by the metric of
the brane. In this spirit, it would seem natural to average the Einstein equations
over the fth dimension.
The main diculty, however, in this procedure is that (when one neglects
bulk matter) the topological constraints mentioned previously will impose the
presence of a second brane, so that when one takes the average of Einsteins
equations, for example the (0,0) component, one will obtain on the right hand
side not only a term proportional to (which, were it be alone, would give back
the standard Friedmann equations) but also a second term proportional to

.
Moreover, as it can be seen explicitly with our exact solutions, and

conspire
so that their sum will be proportional to
2
.
After the results of sections 3 and 4, this conclusion should not be surprising.
Indeed, the non-standard features of the cosmological equations were shown,
through the Gauss-Codacci formalism, to come from a simple geometrical in-
terpretation, i.e. the embedding of our brane-universe in the ve-dimensional
spacetime. This geometrical eect was obtained by a purely local analysis so
that one sees no reason why it would cease to be valid when we compactify the
theory, i.e. when we assume (60).
5
Note that this is also the condition found in [10] when one requires that the |y|-dependent
part of the metric (given by [a]

0
a
0
|y| to the rst order) should be small with respect to a
2
0
,
using equation (16).
6
From a purely mathematical point of view, one could imagine a situation where the conver-
gence would not be true at any point of spacetime, but almost everywhere; so that the dynamics
of the zero modes could be dierent from the one of the induced metric. This will not be the
case for the solutions explicitly studied in this work.
15
Let us now try to be more specic and to exhibit, in the simple case b
0
= b
1/2
(and thus, from (45), [a

]
0
= [a

]
1/2
), a decomposition of the elds into zero
modes and massive modes as in the Kaluza-Klein approach. We take a standard
decomposition for the function b in the form
b = B
0
+

k=0
B
k
e
2iky
, (62)
where B
0
b is the zero mode. The analogous decomposition in the case
of a and n is more subtle because they must necessarily have a dependence on
y due to the presence of the brane. This means that a metric solution of the
ve-dimensional Einstein equations, because of the presence of a y-dependent
source term, will be necessarily y-dependent in order to keep the compactication
procedure consistent (see e.g. [17]). To make this property manifest, we choose
to adopt a distinction between what we call branic modes and the classical
Kaluza-Klein modes, and write
n = N
0
+

k=0
_

n
4
2
k
2
(1 (1)
k
) + N
k
_
e
2iky
, (63)
a = A
0
+

k=0
_

a
4
2
k
2
(1 (1)
k
) + A
k
_
e
2iky
. (64)
In the expressions above, the terms proportional to correspond to the branic
modes. According to (13),
a
= [a

] and
n
= [n

]. The branic terms in the

decomposition come from the Fourier transform of |y|. The branic modes are
dierent from the classical Kaluza-Klein massive modes in the sense that their
eective mass is independent of k. The expansions above imply that n = N
0
and a = A
0
.
It is now instructive to compare the zero-mode metric and the true metric
on the brane in the case of our explicit solutions of the previous subsection. The
average over the fth dimension of a and n yields respectively
A
0
= a
0
_
1

2
(5)
24
b
0

_
, N
0
= n
0
_
1 +

2
(5)
24
(3w + 2)b
0

_
. (65)
The rst remark is that the zero-modes of the metric indeed converge towards the
true metric on the brane (which is the meaningful one as far as observations
are concerned) in the limit
2
(5)
R
1
(5)
(here R
(5)
= b
0
). The same will be true
for the time variation of the metric. The Kaluza-Klein Hubble parameter is
indeed related to the true Hubble parameter by the expression

A
0
N
0
A
0
=
a
0
n
0
a
0
_
1

2
(5)
24
b
0

_
1
. (66)
16
Similar conclusions can be obtained from a variation principle for the case of
a cosmological constant (for which the action is well dened). Starting from the
action
S =
_
d
4
xdy
_
g
_
_

1
2
2
(5)

R +
(y)

g
55
+
(y 1/2)

g
55

_
_
, (67)
where and

are constants, one can then obtain the equation of motion for the
zero modes by a naive compactication, taking into account the constraint (56).
We get, after compactication
S = R
(5)
_
d
4
x

g
_
_

R
2
2
(5)

2
(5)
12

2
_
_
. (68)
This gives the following Friedmann-type equations:
3

A
0
A
0
=

4
(5)
12

2
,

A
0
A
0
+ 2

A
0
2
A
2
0
=

4
(5)
12

2
, (69)
which show that A
0
obeys the same equation (20) obeyed by a
0
(with = p
and

T
55
= 0).
To summarize, the Kaluza-Klein approach remains valid in a branic context
but one must be careful and take into account the topological constraints in the
average procedure over the matter sources. One then recovers the non-standard
equations obtained in the previous sections from a local point of view.
6 Cosmological consequences
The main purpose of this work was to show that, quite generically, the cosmology
arising from the assumption that our universe is conned to a three-brane within a
larger spacetime does not lead to the Friedmann equations of standard cosmology.
The most striking features of the new equations are on one hand that H
2
is not
proportional to as in standard cosmology but rather is related to an expression
quadratic in , on the other hand that the eective four-dimensional Newtons
constant G
(4)
does not enter, in other words the equations do not depend on the
radius of the fth dimension but only on the ve-dimensional Newtons constant
G
(5)
, i.e. the Planck mass scale of the fundamental theory.
One of the important questions is whether such a cosmology is viable when
confronted to observations. Let us discuss the consequences when we assume
that equation (20) applies to our present universe. It is convenient to rewrite the
energy density as a fraction of the critical density
c
whose usual denition is

c
=
3H
2
8G
(4)
, (70)
17
such that
=
c
. (71)
One can now rewrite (20) in the form

H + 2H
2
=
1
4

2
H
2

2
(1 + 3w)
1
3

2
(5)

T
5
5
, (72)
where is a dimensionless parameter dened by
=
m
2
p
H
M
3
(5)
, (73)
having introduced the reduced four-dimensional Planck mass m
p
=
1
(4)
, which
can be given in terms of the fundamental scale as
m
2
p
=
_
8G
(4)
_
1
= M
3
(5)
R
(5)
. (74)
Thus, can also be interpreted as the ratio of the fth dimension radius over the
Hubble radius. The left hand side of the above equation (72) can be reexpressed
today in terms of the so-called acceleration parameter (or rather deceleration
parameter) dened by
q =
_
aa
a
2
_
now
, (75)
so that
(

H + 2H
2
)
now
= (1 q)H
2
now
. (76)
Cosmological observations constrain |q| to be of the order 1. By comparing the
two sides of equations (72) (
now
being of the order 1), one thus arrives to the
conclusion that the present matter content of our universe could inuence the
dynamics of its geometry if was at least of order 1, which would signify that
the fth dimension radius is of the order of the present Hubble radius. Such
a possibility, however, must be rejected because we could then observe directly
ve-dimensional gravity, which is not the case. As we discuss below, a possible
way out is to assume that the bulk term dominates in equation (20) and is driving
the dynamics of our universe.
However, another possibility is that equation (20) does not apply today. In-
deed, an essential assumption of our model is strict homogeneity and isotropy of
our Universe which is only true today on very large scales. Therefore our model
probably breaks down when deviations from homogeneity are important on scales
of the order of the fth dimension radius.
One could then consider the possibility that this non-conventional cosmol-
ogy existed only for the early universe. How would it aect nucleosynthesis ?
Although detailed constraints would make necessary the use of a numerical evo-
lution of the chemical reactions, one can already get a rough constraint from the
18
abundance of Helium by taking the approximation that all free neutrons just be-
fore nucleosynthesis will end up in Helium nuclei. To see what this would entail
in our unconventional scenario, let us recall that nucleosynthesis can be roughly
divided into two steps: the rst step consists in the freezing-out of the proton to
neutron ratio, which takes place at a temperature T T
D
, the second step is the
production of light nuclei, which occurs at a temperature T T
N
0.1MeV (we
neglect here the nite time span for this step).
In the standard scenario (see, e.g., [20]), above the temperature T
D
(0.7
0.8) MeV, the neutron to proton ratio is xed by the thermodynamic equilibrium
relation
_
n
p
_
eq
(T) = exp (Q/T) , (77)
where Q m
n
m
p
1.3MeV. Below T
D
, as long as the production of light nuclei
has not begun, the neutron to proton ratio will be given by (n/p)
eq
(T
D
) corrected
by the decay of neutrons into protons (occuring on the time scale
n
10
3
s), so
that
_
n
p
_
(T)
_
n
p
_
eq
(T
D
) exp(t/
n
), (78)
where t is the age of the universe at the temperature T (assuming t t
D
).
Assuming all neutrons existing at T
N
, just before production of light nuclei, will
be essentially transformed into Helium nuclei, one gets Y = 2x/(1 + x) where
Y is the mass fraction of Helium 4 and x is the initial neutron to proton ratio.
Observations give us Y 0.25, which implies x = Y/(2 Y ) 1/7.
In the unconventional scenario (assuming the bulk contribution is negligible
at the time of nucleosynthesis), a(t) t
1/4
so that the age of the universe is
t H
1
/4. Moreover H T
4
(instead of H T
2
in the standard model), which
implies that the cooling in the universe is much slower than in the standard case.
An immediate consequence is that T
D
cannot be the same as in the standard
scenario, because it would lead to a much smaller neutron to proton ratio just
before nucleosynthesis due to a more eective decay of neutrons into protons. So
let us now evaluate the required new value for T
D
. For T Q, the rate
np
of
the reaction n p is given by

np
= T
5
, (79)
with 0.76s
1
MeV
5
. Let us also write the Hubble parameter in the form
H = T
4
. (80)
The freezing-out temperature T
D
is reached when the chemical reactions slow
down to the cosmological expansion rate, i.e. when H
np
, and is therefore
given by T
D
/. It thus follows that can be expressed in the form

_
Q+
1
4
T
4
N

1
n
_
/ lnx. (81)
19
Inserting numerical values, one nds T
D
2 3 MeV. More importantly, the
scale of M
(5)
is also xed by this constraint, substituting the value of in (80)
and then H in (20) with = 3p = (
2
/30)g

T
4
(where g

is the eective number

of relativistic degrees of freedom). One then nds
g

1
3

M
(5)

_

2
180
_1
3
3 TeV. (82)
For ve dimensions this fundamental scale is too low to be in agreement with
gravity such as it appears to us on large scales, but it would be worthwhile to
study the cases with more dimensions.
Going (much) further into the past, one can examine the issue of ination in
brane cosmology, a question which has attracted recently much attention [7]-[14].
A rst question is whether ination is triggered by the energy density in the
bulk or in the brane. It is more in the spirit of this paper to discuss brane
ination and we will restrict our attention to this case. The next issue deals with
the initial conditions. It has been stressed that brane ination might require an
almost empty bulk, following the condition
bulk
R
(5)

brane
. The study above
clearly shows that the domination of brane over bulk energy density need only be
imposed at a local level, i.e. on a typical distance scale of order M
1
(5)
. In other
words, we need only to impose the weaker constraint

bulk
M
1
(5)
<
brane
. (83)
This in fact ensures that the bulk energy in the neighbourhood of the brane does
not dominate over the conned energy. It is also consistent with our general
constraint (21), under the condition
brane
< M
4
(5)
, which represents a natural
bound for the energy density localized in the brane.
Similarly, recent discussions [21, 7, 8] on the necessity of having an untol-
erably light inaton eld do not apply in the context of our non-conventional
cosmological scenario. Indeed, using our relation (20) and assuming the natural
bound M
4
(5)
, we obtain
H M
3
(5)
M
(5)
. (84)
This only constraints the inaton mass m, which must be less than H during
ination in order to full slow-roll conditions, to be less than the fundamental
scale M
(5)
.
Throughout this work, we have focused our attention to the cases where the
matter content of the brane dominates that of the bulk. Let us however point
out that, if the bulk was dominating, then one would be back in the traditional
Kaluza-Klein approach, with the dierence that invisible ve-dimensional mat-
ter would control the geometry of the Universe whereas ordinary matter would
play only a passive role. However, an important constraint on such scenarios is
20
that they must recover the remarkable agreement, in standard cosmology, of the
present values of the Hubble parameter and of the radiation temperature with
their respective values at the time of nucleosynthesis, i.e. on a span of almost
ten orders of magnitude.
Finally, a word should be said about the dicult problem of the stabilisation
of the compact dimension [2, 22, 23]. Our previous considerations rest in a
crucial way on the assumption of invariance under dieomorphisms of the ve-
dimensional spacetime. The stabilisation of the radius R
(5)
is usually obtained
with the introduction of a potential for the corresponding b scalar eld, which
obviously breaks this reparametrisation invariance. Our formalism would thus
not be applicable directly once the radius has been stabilized. On the other
hand, this potential usually arises through quantum or thermal uctuations of
the ve-dimensional theory. In this case, the ve-dimensional framework that we
use should be applicable, although our considerations have clearly been limited
to a classical point of view.
It would be interesting to develop the present work in several directions. First,
as we mentioned earlier, our model does not necessarily apply to recent cosmology
because of deviations from homogeneity. An important extension would thus be
to allow for perturbations in the brane-universe. Another direction is to allow
for more than one extra dimension. Finally, in the case of string models, the
presence of scalar elds such as the dilaton or Ramond-Ramond scalars may play
a role in the early Universe dynamics.
Acknowledgments
We wish thank Brandon Carter, Emilian Dudas and Jihad Mourad for fruitful
discussions.
References
[1] D. Bailin, A. Love, Rep. Prog. Phys. 50 (1987) 1087.
[2] N. Arkani-Hamed, S. Dimopoulos, G. Dvali, Phys.Lett. B429, (1998) 263,
and hep-ph/9807344.
[3] P. Horava and E. Witten, Nucl. Phys. B460 (1996) 506, Nucl. Phys. B475
(1996) 94.
[4] I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos, G. Dvali, Phys.Lett. B436,
(1998) 257.
[5] N. Kaloper, I.I. Kogan, K.A. Olive, hep-th/9711027.
21
[6] A. Lukas, B.A. Ovrut, D. Waldram, hep-th/9806022.
[7] D.H. Lyth, Phys.Lett. B448 (1999) 191.
[8] N. Kaloper, A. Linde, hep-th/9811141.
[9] G. Dvali, S.H. Henry Tye, Phys.Lett B450 (1999) 72.
[10] A. Lukas, B.A. Ovrut, D. Waldram, hep-th/9902071.
[11] A. Mazumdar, hep-ph/9902381.
[12] N. Arkani-Hamed, S. Dimopoulos, N. Kaloper, J. March-Russel, hep-
ph/9903224.
[13] C. Csaki, M. Graesser, J. Terning, hep-ph/9903319.
[14] P. Kanti, K.A. Olive, hep-ph/9903524
[15] W. Israel, Nuovo Cimento B 44 (1966).
[16] R.M. Wald, General Relativity, The University of Chicago Press (1984).
[17] M.J. Du, B.E.W. Nilsson, C.N. Pope, N.P. Warner, Phys. Lett. B 149, 90
(1984).
[18] A. Vilenkin, Phys. Lett. B 133, 177 (1983); J. Ipser, P. Sikivie, Phys. Rev.
D 30, 712 (1984).
[19] A.G. Riess et al., Astroph. J 116, 1009 (1998); S. Perlmutter et al., Mea-
surements of Omega and Lambda from 42 High-Redshift Supernovae, astro-
ph/9812133.
[20] S. Weinberg, Gravitation and cosmology, Wiley, New York (1972).
[21] K. Benakli, S. Davidson, hep-ph/9810280.
[22] N. Arkani-Hamed, S. Dimopoulos, J. March-Russel, hep-th/9809124.
[23] K.R.Dienes, E.Dudas, T.Gherghetta, A.Riotto, Nucl.Phys B 543 (1999), 387-
442.
22