Gen Relativ Gravit (2009) 41:1725–1736

DOI 10.1007/s10714-008-0743-0

RESEARCH ARTICLE

A planar perturbation of the Bianchi type I metric

Brian Wilson · Charles C. Dyer

Received: 4 March 2008 / Accepted: 8 December 2008 / Published online: 24 December 2008
© Springer Science+Business Media, LLC 2008

Abstract The standard procedure for finding analytic perturbations in General

Relativity suffers from the drawback that it is cumbersome to use beyond linear order
perturbations. Following up on our previous work, we continue to use an alternate
method of finding perturbations. We find a plane symmetric perturbation of the cos-
mological Bianchi type I metric. The perturbation corresponds to a fluid with heat flow
moving perpendicularly to a singular plane in a region which can be made to be either
overdense or underdense relative to the background spacetime. The fluid satisfies both
the Strong and Dominant energy conditions everywhere except the region close to the
singularity.

Keywords Perturbation · Bianchi type I metric · Plane symmetry ·

Fluid with heat flow

1 Introduction

The standard cosmological model is the Robertson–Walker model. This is a homo-

geneous and isotropic universe with only one dynamic feature, its rate of expan-
sion or contraction. Its high degree of symmetry makes it easy to analyze but also

B. Wilson (B)
Department of Physics, University of Toronto, 60 St George St.,
Toronto, ON M5S 1A7, Canada
e-mail: bwilson@[Link]

C. C. Dyer
Department of Astronomy and Astrophysics, University of Toronto,
50 St George St., Toronto, ON M5S 3H4, Canada
e-mail: dyer@[Link]

123
1726 B. Wilson, C. C. Dyer

makes it obviously unrealistic. It does model the average properties of the observed
universe, on the largest scales, quite well. One would like to find perturbations of the
Robertson–Walker model in order to increase its applicability to observations of the
universe.
One obvious approach is to look at a spherically symmetric perturbation as the
authors did in [1]. If one wishes to look at non-spherically symmetric perturbations,
one should first consider whether the Robertson–Walker model is the best model
to use. The Bianchi cosmological models, for example, may be a more realistic
background. These models are homogeneous but anisotropic cosmological models
that are essentially indistinguishable from the Robertson–Walker models when far
from the initial singularity or, in the case of the collapsing models, when far from
the future singularity. The present epoch is far from all such singularities. They
have three dynamic features, the expansion or contraction in the three spatial direc-
tions which are, in general, all different. The three rates of expansion or contrac-
tion asymptotically (in time) equalize, thus producing a model which looks like a
Robertson–Walker model. If the universe did not begin by expanding exactly iso-
tropically then the Bianchi models more accurately model its early behaviour than
the Robertson–Walker model. The Bianchi models also model a recollapsing uni-
verse better than the Robertson–Walker model unless the universe recollapses exactly
isotropically.
In this paper, we use the same method of generating exact perturbations ts was used
in our earlier paper [1]. We produce a plane symmetric perturbation of the axisymmet-
ric Bianchi type I cosmological model. This background model expands equally in two
directions while the third direction begins by contracting. The contraction eventually
halts and turns into an expansion, at which point the model becomes asymptotically
(in time) identical to a flat Robertson–Walker model with no cosmological constant. A
result of the underlying motion of the Bianchi type I model is that any initially spher-
ical perturbations will get flattened into pancakes. It is thus natural to begin looking
at plane-symmetric perturbations of the model.
Relativistic perturbation theory has many uses. It can help calibrate and extract
information from numerical computations [2]. It is necessary in order to properly ana-
lyze many important situations, such as structure formation [3]. For more details see
[1].
It is easy enough to describe how to perform standard perturbative calculations
to arbitrary order in General Relativity. It is very cumbersome to perform the actual
calculations. Second order perturbations are a topic of current investigation, see for
example [4–7]. The method of generating perturbations we use bypasses most of the
difficulties, but it has its own liabilities. The method does not allow for the study of
how a system evolves as N -body simulations do. This is a liability common to many
exact perturbation methods. It is, however, well-situated to modelling known systems
and generating plausible systems.
In Sect. 2, we review the background Bianchi type I metric, and setup the pertur-
bation. In Sect. 3, we analyze the fluid components to place constraints on the metric
perturbation components. In Sect. 4, we analyze a specific exact perturbation in terms
of the fluid components, thermodynamics and various energy conditions. In Sect. 5,
we make some concluding remarks.

123
The Bianchi type I metric 1727

2 Setting up the metric

A general axisymmetric, irrotational metric can be written in cylindrical coordinates

(t, z, ρ, φ) as

ds 2 = a dt 2 − b dz 2 − c (dρ 2 + ρ 2 dφ 2 ) (1)

where ρ is the radial coordinate, so that ρ ≥ 0, and a, b, c are functions of t, z and ρ.

We have chosen units such that the speed of light is unity.
The axisymmetric Bianchi type I metric (for more details see, for example [8]) in
these coordinates is given by a = 1, b = R1 (t), and c = R2 (t). In this Bianchi type
I metric the Einstein Field equations for a dust source require, using dots to represent
derivatives with respect to t,

Ṙ22
R̈2 = (2)
4R2

and

2
Ṙ 2 R1 Ṙ1 Ṙ2
R̈1 = 1 + − (3)
4R1 4 R1 R2

or equivalently,

2
S + Mt
R1 = R2 (4)
t

and

2/3
3
R2 = t 4/3 , (5)
4

with S and M positive constants. M can be set to 1 by rescaling the z coordinate.

Setting M thus we note that R1 > R2 ,

Ṙ1 Ṙ2 2S
= − 2 (6)
R1 R2 t + St

and hence

1 Ṙ2 Ṙ1 Ṙ2

− < < . (7)
2 R2 R1 R2

We also note that Ṙ1 = 0 at t = 2S , so S is a measure of the time it takes for the
initially collapsing R1 to turn around and start expanding. Present observations sug-
gest the universe is expanding isotropically so we require t  S ‘today’, which gives

123
1728 B. Wilson, C. C. Dyer

R1 ∼ R2 . We set t ‘today’ by requiring ṘR22 = 3t4 match the observed value of Hubble’s
parameter, approximately 70 km s−1 Mpc−1 . This gives an order of magnitude value
of t ∼ 1010 years. The particle horizon is the maximum distance a particle could have
traveled over the entire history of the universe. The particle horizon is of the order
l ∼ Ṙ /R
c
∼ 1010 ly. Finally, we note that the density, µ̄, of the co-moving dust is
2 2


Ṙ2 Ṙ1 Ṙ2
µ̄ = 2 + >0 (8)
4R2 R1 R2

where we will use the bar notation to indicate background quantities. The inequality
for µ̄ is a consequence of the inequality in Eq. (7).
The perturbed Bianchi type I cosmological metric is given by

ds 2 = (1 + A) dt 2 − R1 (1 + B) dz 2 − R2 (1 + C) (dρ 2 + ρ 2 dφ 2 ). (9)

We choose A, B, C to be functions of t and z alone. A prime will denote a derivative

with respect to z. We are looking at plane symmetric perturbations. Due to symmetry,
the fluid cannot flow in the ρ direction. If the fluid is infalling, it cannot escape and
we expect a singularity. If the fluid is an outflow, something needs to create matter
to keep the outflow running so we again expect a singularity. This singularity is of
necessity a plane singularity. We shall ignore it on the grounds that this solution is
only an approximate solution away from the singularity. Near the singularity we would
match this solution with one which allows for motion in the ρ direction if we could,
and be satisfied that the result was a realistic system.
We choose a fluid with pressure and heat flow. This gives the energy momentum
tensor

Tab = (µ + p) u a u b − p gab + u a qb + qa u b (10)

with µ the density, p the pressure, u a the fluid velocity and qa the heat flow. The
standard restrictions u a u a = 1 and u a qa = 0 apply. Due to symmetry we know
u 2 = u 3 = 0 and q2 = q3 = 0. This imposes restrictions on the metric through the
Einstein field equations, but the restrictions are identically satisfied for metrics of the
form we have chosen.

3 Constraining the perturbation

We now look at the fluid components by inverting equations (10) to get the components
in terms of the energy momentum tensor Tab . We then use the Einstein field equations,
choosing units to simplify the equations to Tab = G ab , to get the fluid components
in terms of the metric and its derivatives. We finally insert the metric of equation (9)
into the Einstein tensor G ab to get the fluid components in terms of the perturbations
A, B, C.

123
The Bianchi type I metric 1729

The pressure is

T33 1 1 1 1 1
p= ∼ Ȧ Ṙ1 R1−1 + Ȧ Ṙ2 R2−1 + A R1−1 − B̈ − Ḃ Ṙ1 R1−1
g33 4 4 2 2 2
1 1 1 1 1
− Ḃ Ṙ2 R2−1 − C̈ − Ċ Ṙ1 R1−1 − Ċ Ṙ2 R2−1 + C  R1−1 . (11)
4 2 4 2 2
The fluid velocity is given by

u −2
0 =g +α g
00 2 11
(12)

where

u1 T01 ± 2 − (T + g p)(T + g p)
T01 11 11 00 00
α≡ = . (13)
u0 T00 + g00 p

If we stay in the perturbative regime, T00 ∼ µ̄ will be much greater than the other
terms, and we get the approximation

T11 + g11 p
α∼± − . (14)
µ̄

The fluid velocity and heat flow components are, to first order in the perturbation,
1
u 1 ≡ α, u 0 ∼ 1 + (A + α 2 R1−1 ), (15)
2
q0 ∼ −µ̄α 2 R1−1 , and q1 ∼ −µ̄α. (16)
We define the density contrast to be δµ = µ − µ̄. T00 is approximately
T00 = (µ + p) u 20 − p g00 + 2 u 0 q0 ∼ µ̄ + δµ + Aµ̄ − µ̄α 2 R1−1 . (17)
We thus get
δµ ∼ T00 − µ̄ − Aµ̄ + p − T11 R1−1 , (18)
which expands to

1 1 1 1 1
δµ ∼ −Aµ̄ + Ȧ Ṙ1 R1−1 − Ȧ Ṙ2 R2−1 + A R1−1 − B̈ − Ḃ Ṙ1 R1−1
4 4 2 2 2
1 −1 1 1 −1 3 −1 1  −1
+ Ḃ Ṙ2 R2 + C̈ + Ċ Ṙ1 R1 + Ċ Ṙ2 R2 − C R1 . (19)
4 2 4 2 2

In addition, we calculate the fluid’s acceleration, u b ∇b u a , to have the components

α α̇
u b ∇b u 0 ∼ and u b ∇b u 1 ∼ α̇. (20)
R1
We will need this when we check whether the fluid has a non-decreasing entropy.

123
1730 B. Wilson, C. C. Dyer

For the fluid velocity to be real we need, from Eq. (14),

T11 + g11 p < 0. (21)

We also have, from Eq. (12),

T11 + g11 p 11
u −2
0 ∼g −
00
g (22)
T00

which means u 0 is real if and only if, to first order in the perturbation,

g 00
T00 ≥ T11 + g11 p (23)
g 11

or

R1
R1 T00 ≥ T11 − T22 . (24)
R2

Condition (24) is satisfied in the perturbative regime by virtue of T00 being the only
non-zero component of Tab in the background. This leaves condition (21) which is, to
first order,

Ȧ Ṙ1 R1−1 − Ȧ Ṙ2 R2−1 + 2 A R1−1 − 2 B̈ − 2 Ḃ Ṙ1 R1−1 − Ḃ Ṙ2 R2−1
+ 2C̈ − Ċ Ṙ1 R1−1 + 4Ċ Ṙ2 R2−1 + 2C  R1−1 > 0. (25)

3.1 Constraints due to local considerations

The time derivatives are of order l −1 , l being the particle horizon. At distances much
smaller than l, the time derivatives are much smaller than the z derivatives. Define
z 0 as the location where the space derivatives and the time derivatives are of equal
magnitude, i.e. Ḃ ∼ B  . The space derivatives dominate considerations for z  z 0 .
In this region, condition (25) becomes

A + C  ≥ 0. (26)

The pressure in this region is approximately

1 −1   
p∼ R1 A + C  . (27)
2

If condition (26) is satisfied then the pressure must be positive in this region. The
density contrast in this region is given by Eq. (19), which is approximately

1 −1   
δµ ∼ R1 A − C  . (28)
2

123
The Bianchi type I metric 1731

Fig. 1 Density profile of locally

overdense and underdense
perturbations. α < 0 forces the
fluid to be infalling. α > 0 only
changes the directions of both
u 1 and q 1 , forcing the fluid to be
outflowing. There is a curvature
singularity at g11 = 0

Condition (26) requires that the larger in magnitude of A and C  be positive.
Equation (28) gives an overdense region where A > C  and an underdense region
where C  > A .

3.2 Constraints due to far field considerations

We turn our attention to the region z  z 0 , where space derivatives are negligible. If
we look at late time evolution we get R1 → R2 → R. Condition (25) becomes

− Ȧ(t)R − 2 B̈ − 3 Ḃ Ṙ R −1 + 2C̈ + 3Ċ Ṙ R −1 ≥ 0 (29)

where (t) = t 22S +St

. Recall that S > 0 and, as measured ‘today’, t  S. This
makes (t) a small, positive function of time which is ignorable unless Ḃ = Ċ or
| Ȧ|  | Ḃ|, |Ċ|. The pressure in the same region is approximately

4 p ∼ 2 Ȧ Ṙ R −1 − 2 B̈ − 3 Ḃ Ṙ R −1 − 2C̈ − 3Ċ Ṙ R −1 . (30)

In this situation, the pressure given in Eq. (30) is not necessarily positive when con-
dition (29) is satisfied. If | Ḃ|  | Ȧ|, |Ċ| then satisfying condition (29) will guarantee
a positive pressure. With this assumption the density contrast given by Eq. (19) is
approximately

4δµ ∼ −2 B̈ − Ḃ Ṙ R −1 (31)

If condition (29) is satisfied, δµ will usually be positive, independent of the local con-
siderations. That is, the assumptions | Ḃ|  | Ȧ|, |Ċ| and Ḃ < 0 will give us a global
overdense universe, relative to the background, with either a local overdense or under-
dense region as determined by A and C  . The generic properties of the perturbation
with the above-given restrictions are illustrated in Fig. 1.

4 Specific perturbation

The easiest way to ensure that the pressure for z  z 0 given by Eq. (30) is positive while
simultaneously satisfying condition (29) is to choose | Ḃ|  | Ȧ|, |Ċ| and Ḃ, B̈ < 0.

123
1732 B. Wilson, C. C. Dyer

Fig. 2 Log–log graph of µ, δµ and p

If, in addition, we desire a metric perturbation which increases in magnitude with time
then we require B < 0 so that ḂB > 0. Similarly, we will require AȦ , C
Ċ
> 0 so that A
and C are either positive and increasing, or negative and decreasing. If we desire that
  
the metric perturbations vanish as z → ∞ then we need AA , BB , CC < 0 for large z.
We further recall that condition (26) requires A + C  > 0 for z  z 0 . Combining
all this, we choose

A = k 2A z −2 R2 , B = −k 2B z −2 R2 and C = kC2 z −2 R2 , (32)

with k B  k A , kC all constants (with dimensions of length) much smaller than the
particle horizon l. k A > kC gives a local overdense region while kC > k A gives a local
underdense region.
Since the sign of α as given in (14) is chosen arbitrarily, there are four physically
different solutions. Two solutions have local underdense regions, one with an infall-
ing fluid and the other with an outflowing fluid. The other two have local overdense
regions, one with an infalling fluid and the other with an outflowing fluid. One might
think one could produce eight solutions by choosing α to have different signs for
z > 0 versus z < 0. However, our metric has a singularity at |B| = 1, equivalently
g11 = 0, where the Kretschmann curvature scalar is infinite. We are stuck with a plane
singularity (as we expected) and thus the z > 0 and z < 0 regions are not connected.
We therefore have only four solutions.
We plot the exact values of the fluid components given this exact metric in Figs. 2
and 3. We choose k 2A = 10−34 ly2 , k 2B = 10−30 ly2 , kC2 = 10−35 ly2 and α < 0. These
values put the singularity at z ∼ 10−8 ly ‘today’, comfortably far from the particle
horizon l ∼ 1010 ly, and they set z 0 ∼ 102 ly. We could increase z 0 by increasing
the ratio kk BA up to a maximum of around z 0 ∼ 103 ly without violating the | Ḃ| > |Ċ|

123
The Bianchi type I metric 1733

Fig. 3 Log–log graph of u 1 , q1 and q0

restriction. We note that z 0 ∝ t 1/3 while for the singularity we have z ∝ t 2/3 . The
singularity will eventually (at t ∼ 1040 years) catch z 0 . The log–log graphs show that
the fluid behaves as predicted in the various regions. The density contrast and pressure
are positive and go like 1/z 4 for z < z 0 and like 1/z 2 for z > z 0 . The fluid velocity is
always toward z = 0 and goes like 1/z for z < z 0 and 1/z 2 for z > z 0 . The velocity
is non-relativistic. This is all calculated ‘today’.

4.1 Entropy conservation

Entropy is guaranteed (see for example [9]) to be non-decreasing if a temperature T

can be found satisfying the heat flow equation

qa = κh ab (∂b T − T u c ∇c u b ) (33)

where h ab = gab − u b u a projects objects onto the three-space orthogonal to u a , and

κ is the thermal conductivity coefficient. Inserting equations (15), (16) and (20) into
(33) gives, to leading order,

T  − α Ṫ − T α̇ = −α µ̄κ −1 . (34)

Our chosen metric has

√ two −2regions. For z  z 0 we have α ∼ Ṙ t z −1 , while for
z  z 0 we have α ∼ R t z . In both cases we are assuming α  1, therefore
Eq. (34) requires either T  ∼ 0 or T  ∼ −α µ̄κ −1 . For z  z 0 we get the solution to
T  ∼ −α µ̄κ −1 to be

α µ̄z
T ∼ + K 1 (t). (35)
κ

123
1734 B. Wilson, C. C. Dyer

K 1 is a constant of the integration with respect to z. This temperature is only positive

for z → 0 if α > 0. For z  z 0 we get

α µ̄z log z
T ∼− + K 2 (t). (36)
κ

K 2 is another constant of the integration with respect to z. This temperature is only

positive for z → ∞ if α < 0.
If we assume T  ∼ 0 then Eq. (34) becomes

κ(αT )˙ ∼ α µ̄ (37)

or upon integration
 
α
καT ∼ α µ̄dt ∝ dt. (38)
t2

This temperature is positive as long as α increases at least as quickly as t, which is true

in both regions. We also note that this solution also gives T  ∼ 0 as we had assumed.
Thus the laws of thermodynamics are satisfied to leading order in the perturbation.
For α < 0, the infalling model, we have a temperature profile which is constant inside
z < z 0 and which increases slowly, as log z, outside this region. For α > 0, the out-
flowing model, we have a constant temperature far from the origin (z > z 0 ) and a z −1
temperature profile in the local region. In both cases the fluid moves from high to low
temperature.

4.2 The energy conditions

As shown in [1], both the Strong and Dominant energy conditions are satisfied for a
fluid with heat flow if

µ2 > 4q 2 . (39)

Inserting equation (16) into (39), and assuming µ ∼ µ̄, gives, to leading order,

µ̄2 > 4µ̄2 α 2 R1−1 (40)

or

1
α2 < R1 . (41)
4

This is the requirement that the fluid velocity be non-relativistic. We have R1 ∼

1013 ‘today’. Equation (13) defined α ≡ u 1 . The energy conditions are thus satisfied
wherever u 1 < 106 . Looking at Fig. 3, we see that this model satisfies the energy
conditions everywhere except close to the singularity.

123
The Bianchi type I metric 1735

5 Conclusions

We have derived a planar-symmetric metric which is a natural initial attempt at a cos-

mological perturbation of the axisymmetric–symmetric Bianchi type I model. This
metric is produced by a fluid with both pressure and heat flow. There are four different
models depending on whether there is a local overdensity or underdensity, and whether
the fluid is infalling or outflowing. We analyzed the thermodynamics of one explicit
form of the metric and found that entropy is non-decreasing and both the Strong and
Dominant energy conditions are satisfied everywhere except very near the singularity.
All models have a curvature singularity. This was expected based on the fact that
planar-symmetric perturbations with a fluid moving orthogonal to the planes of sym-
metry must either generate matter or indefinitely compress it depending on whether it
is an outflow or an infall.
The four different cases of the model are depicted in Fig. 4. The outflowing, under-
dense model (Fig. 4b) is representative of a fluid slowing down as it enters a region
where the fluid is more dense and moving more slowly. The outflowing, overdense
model (Fig. 4a) is representative of a fluid entering a region where the fluid is less
dense. These two systems could be a simple model of the asymmetric double lobes of
the radio galaxy MSH 05-22, one of which appears to be colliding with a sheet while
the other is aimed toward a void, as described in [10]. The outflowing, overdense
model also corresponds to a jet from a disc, albeit one of infinite width. It would be
nice to take a cylindrical section of this model and match its boundary with a different
model to get a true jet. However, such matchings are difficult to perform in general
relativity. The infalling, overdense model (Fig. 4c) represents a fluid falling orthogo-
nally onto an infinite disc. This could represent a model in which matter is falling onto

(a) (b)

(c) (d)

Fig. 4 Pictorial representation of the four cases of this solution: a the outflowing, overdense model; b the
outflowing, underdense model; c the inflowing, overdense model; and d the inflowing, underdense model

123
1736 B. Wilson, C. C. Dyer

the central region of a finite disc. In the central region of a finite disc one can safely
ignore edge effects, in which case one would expect exactly orthogonal fluid motion.
The outflowing, underdense model (Fig. 4d) is a model of a fluid falling onto a two
dimensional singularity. This is obviously an unlikely astronomical scenario, hence
this case is unlikely to prove useful.

Acknowledgments The work of BW was supported in part by the Ontario government through the OGS
program. CCD acknowledges the support of the Natural Sciences and Engineering Research Council of
Canada via a Discovery Grant.

References

1. Wilson, B., Dyer, C.C.: Gen. Relativ. Gravit. 39, 2001 (2007)
2. Seidel, E.: Classical Quantum Gravit. 21, S339 (2004)
3. Mars, M.: Class. Quantum Gravit. 22, 3325 (2005)
4. Nakamura, K.: Phys. Rev. D 74, 101301 (2006)
5. Herrero, A., Portilla, M.: Gen. Relativ. Gravit. 37, 1205 (2005)
6. Wiltshire, R., Messenger, P.: Gen. Relativ. Gravit. 36, 1213 (2004)
7. Cabezas, J.A., Martin, J., Molina, A., Ruiz, E.: Gen. Relativ. Gravit. 39, 707 (2007)
8. Stephani, H.: General Relativity—An Introduction to the Theory of the Gravitational Field, 2nd edn.
Cambridge University Press, Cambridge (1990)
9. Silva, R., Lima, S., Calvão, M.O.: Gen. Relativ. Gravit. 34, 865 (2002)
10. Subrahmanyan, R., Saripalli, L., Safouris, V., Hunstead, R.W.: Astrophys. J. 677, 63–78 (2008)

123